diff --git a/demo/webcam.js b/demo/webcam.js
index af468e8..8117fac 100644
--- a/demo/webcam.js
+++ b/demo/webcam.js
@@ -28,8 +28,11 @@ function log(...txt) {
   if (div) div.innerHTML += `<br>${txt}`;
 }
 
+
+
 // helper function to draw detected faces
 function drawFaces(canvas, data, fps) {
+  console.log(data);
   const ctx = canvas.getContext('2d');
   if (!ctx) return;
   ctx.clearRect(0, 0, canvas.width, canvas.height);
@@ -59,9 +62,11 @@ function drawFaces(canvas, data, fps) {
     ctx.fillText(`expression: ${Math.round(100 * expression[0][1])}% ${expression[0][0]}`, person.detection.box.x, person.detection.box.y - 42);
     ctx.fillText(`age: ${Math.round(person.age)} years`, person.detection.box.x, person.detection.box.y - 24);
     ctx.fillText(`roll:${person.angle.roll}° pitch:${person.angle.pitch}° yaw:${person.angle.yaw}°`, person.detection.box.x, person.detection.box.y - 6);
+    // draw person's name - M.S. 30/10/2022 
+    ctx.fillText(`${person.label}`, person.landmarks.positions[person.landmarks.positions.length - 1].x, person.landmarks.positions[person.landmarks.positions.length - 1].y + 95);
     // draw face points for each face
     ctx.globalAlpha = 0.8;
-    ctx.fillStyle = 'lightblue';
+    ctx.fillStyle = 'red';
     const pointSize = 2;
     for (let i = 0; i < person.landmarks.positions.length; i++) {
       ctx.beginPath();
@@ -71,30 +76,83 @@ function drawFaces(canvas, data, fps) {
   }
 }
 
-async function detectVideo(video, canvas) {
+
+
+const labels = ['Matan', 'Yehuda', 'Yoni_Open'];
+
+/**
+ * 
+ * @param {*} labels names of people that their photos exists 
+ * @returns encoding version of photos of the given labels array
+ */
+async function getLabelFaceDescriptions(labels = ['Matan', 'Yehuda', 'Yoni_Open']) {
+  return await Promise.all(
+    labels.map(async label => {
+      // fetch image data from urls and convert blob to HTMLImage element
+      const imgUrl = `${label}.jpg`
+      const img = await faceapi.fetchImage(imgUrl)
+
+      // detect the face with the highest score in the image and compute it's landmarks and face descriptor
+      const fullFaceDescription = await faceapi.detectSingleFace(img).withFaceLandmarks().withFaceDescriptor()
+
+      if (!fullFaceDescription) {
+        throw new Error(`no faces detected for ${label}`)
+      }
+
+      const faceDescriptors = [fullFaceDescription.descriptor]
+      return new faceapi.LabeledFaceDescriptors(label, faceDescriptors)
+    })
+  )
+}
+
+
+async function detectVideo(video, canvas, facesLables) {
   if (!video || video.paused) return false;
   const t0 = performance.now();
-  faceapi
+
+
+  const fullFaceDescriptions = faceapi
     .detectAllFaces(video, optionsSSDMobileNet)
     .withFaceLandmarks()
     .withFaceExpressions()
-    // .withFaceDescriptors()
-    .withAgeAndGender()
-    .then((result) => {
-      const fps = 1000 / (performance.now() - t0);
-      drawFaces(canvas, result, fps.toLocaleString());
-      requestAnimationFrame(() => detectVideo(video, canvas));
-      return true;
-    })
-    .catch((err) => {
-      log(`Detect Error: ${str(err)}`);
+    .withAgeAndGender().withFaceDescriptors();
+
+  console.log(fullFaceDescriptions);
+
+  fullFaceDescriptions.then(async result => {
+    console.log(result);
+    // log(fullFaceDescriptions); // TODO function to write entering person table per one
+
+    try {
+      // const fps = 1000 / (performance.now() - t0);
+      // 0.6 is a good distance threshold value to judge
+      // whether the descriptors match or not
+      const maxDescriptorDistance = 0.6;
+      const faceMatcher = new faceapi.FaceMatcher(facesLables, maxDescriptorDistance);
+      const recognitionResults = (await result).map(fd => faceMatcher.findBestMatch(fd.descriptor));
+
+
+      recognitionResults.forEach((bestMatch, i) => {
+        const text = bestMatch.toString();
+        // log(fullFaceDescriptions); // TODO function to write entering person table per one
+        console.log(text);
+        const fps = 1000 / (performance.now() - t0);
+        result[i]['label'] = text; // attached label to recognized person
+        drawFaces(canvas, result, fps.toLocaleString());
+        requestAnimationFrame(() => detectVideo(video, canvas, facesLables));
+
+        return true;
+      });
+    } catch (error) {
+      log(`Detect Error: ${str(error)}`);
+      console.error(error);
       return false;
-    });
-  return false;
+    }
+  })
 }
 
 // just initialize everything and call main function
-async function setupCamera() {
+async function setupCamera(facesLables) {
   const video = document.getElementById('video');
   const canvas = document.getElementById('canvas');
   if (!video || !canvas) return null;
@@ -132,7 +190,7 @@ async function setupCamera() {
     if (video && video.readyState >= 2) {
       if (video.paused) {
         video.play();
-        detectVideo(video, canvas);
+        detectVideo(video, canvas, facesLables);
       } else {
         video.pause();
       }
@@ -144,7 +202,7 @@ async function setupCamera() {
       canvas.width = video.videoWidth;
       canvas.height = video.videoHeight;
       video.play();
-      detectVideo(video, canvas);
+      detectVideo(video, canvas, facesLables);
       resolve(true);
     };
   });
@@ -159,6 +217,7 @@ async function setupFaceAPI() {
   await faceapi.nets.faceLandmark68Net.load(modelPath);
   await faceapi.nets.faceRecognitionNet.load(modelPath);
   await faceapi.nets.faceExpressionNet.load(modelPath);
+
   optionsSSDMobileNet = new faceapi.SsdMobilenetv1Options({ minConfidence: minScore, maxResults });
   // check tf engine state
   log(`Models loaded: ${str(faceapi.tf.engine().state.numTensors)} tensors`);
@@ -173,14 +232,16 @@ async function main() {
   // await faceapi.tf.setBackend('wasm');
 
   // default is webgl backend
-  await faceapi.tf.setBackend('webgl');
+  await faceapi.tf.setBackend('webgl'); //TODO check webgl in tensorflow
   await faceapi.tf.ready();
+  await setupFaceAPI();
+  let facesLables = await getLabelFaceDescriptions(labels);
 
   // check version
   log(`Version: FaceAPI ${str(faceapi?.version || '(not loaded)')} TensorFlow/JS ${str(faceapi?.tf?.version_core || '(not loaded)')} Backend: ${str(faceapi?.tf?.getBackend() || '(not loaded)')}`);
 
-  await setupFaceAPI();
-  await setupCamera();
+  await setupCamera(facesLables);
+
 }
 
 // start processing as soon as page is loaded
diff --git a/dist/face-api.esm-nobundle.js b/dist/face-api.esm-nobundle.js
index 487ce5c..ceebce2 100644
--- a/dist/face-api.esm-nobundle.js
+++ b/dist/face-api.esm-nobundle.js
@@ -4,4 +4,4698 @@
   author: <https://github.com/vladmandic>'
 */
 
-var lr=Object.defineProperty;var bo=Object.getOwnPropertyDescriptor;var go=Object.getOwnPropertyNames;var xo=Object.prototype.hasOwnProperty;var vo=(o=>typeof require!="undefined"?require:typeof Proxy!="undefined"?new Proxy(o,{get:(t,e)=>(typeof require!="undefined"?require:t)[e]}):o)(function(o){if(typeof require!="undefined")return require.apply(this,arguments);throw new Error('Dynamic require of "'+o+'" is not supported')});var Ve=(o,t)=>{for(var e in t)lr(o,e,{get:t[e],enumerable:!0})},fr=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let a of go(t))!xo.call(o,a)&&a!==e&&lr(o,a,{get:()=>t[a],enumerable:!(r=bo(t,a))||r.enumerable});return o},P=(o,t,e)=>(fr(o,t,"default"),e&&fr(e,t,"default"));var n={};Ve(n,{version:()=>Po});P(n,mn);P(n,pn);P(n,un);import*as mn from"@tensorflow/tfjs/dist/index.js";import*as pn from"@tensorflow/tfjs-backend-webgl/dist/index.js";import*as un from"@tensorflow/tfjs-backend-wasm/dist/index.js";var dr="4.0.0",yo="4.0.0",_o="4.0.0",To="4.0.0",wo="4.0.0",Po={tfjs:dr,"tfjs-core":dr,"tfjs-converter":yo,"tfjs-backend-cpu":_o,"tfjs-backend-webgl":To,"tfjs-backend-wasm":wo};var Wr={};Ve(Wr,{AnchorPosition:()=>Ke,DrawBox:()=>ee,DrawBoxOptions:()=>de,DrawFaceLandmarks:()=>De,DrawFaceLandmarksOptions:()=>Fe,DrawTextField:()=>et,DrawTextFieldOptions:()=>kt,drawContour:()=>Z,drawDetections:()=>No,drawFaceExpressions:()=>So,drawFaceLandmarks:()=>Ao});function Z(o,t,e=!1){if(o.beginPath(),t.slice(1).forEach(({x:r,y:a},s)=>{let i=t[s];o.moveTo(i.x,i.y),o.lineTo(r,a)}),e){let r=t[t.length-1],a=t[0];if(!r||!a)return;o.moveTo(r.x,r.y),o.lineTo(a.x,a.y)}o.stroke()}var hr={};Ve(hr,{computeReshapedDimensions:()=>je,getCenterPoint:()=>yt,isDimensions:()=>ue,isEven:()=>pe,isFloat:()=>Ge,isTensor:()=>xt,isTensor1D:()=>Fo,isTensor2D:()=>Ye,isTensor3D:()=>K,isTensor4D:()=>R,isValidNumber:()=>G,isValidProbablitiy:()=>It,range:()=>U,round:()=>vt});var S=class{constructor(t,e){if(!G(t)||!G(e))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:e})}`);this._width=t,this._height=e}get width(){return this._width}get height(){return this._height}reverse(){return new S(1/this.width,1/this.height)}};function xt(o,t){return o instanceof n.Tensor&&o.shape.length===t}function Fo(o){return xt(o,1)}function Ye(o){return xt(o,2)}function K(o){return xt(o,3)}function R(o){return xt(o,4)}function Ge(o){return o%1!==0}function pe(o){return o%2===0}function vt(o,t=2){let e=10**t;return Math.floor(o*e)/e}function ue(o){return o&&o.width&&o.height}function je({width:o,height:t},e){let r=e/Math.max(t,o);return new S(Math.round(o*r),Math.round(t*r))}function yt(o){return o.reduce((t,e)=>t.add(e),new g(0,0)).div(new g(o.length,o.length))}function U(o,t,e){return Array(o).fill(0).map((r,a)=>t+a*e)}function G(o){return!!o&&o!==1/0&&o!==-1/0&&!Number.isNaN(o)||o===0}function It(o){return G(o)&&o>=0&&o<=1}var g=class{constructor(t,e){this._x=t,this._y=e}get x(){return this._x}get y(){return this._y}add(t){return new g(this.x+t.x,this.y+t.y)}sub(t){return new g(this.x-t.x,this.y-t.y)}mul(t){return new g(this.x*t.x,this.y*t.y)}div(t){return new g(this.x/t.x,this.y/t.y)}abs(){return new g(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new g(Math.floor(this.x),Math.floor(this.y))}};var F=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(G)}static assertIsValidBox(t,e,r=!1){if(!F.isRect(t))throw new Error(`${e} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${e} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,e=!0){let r=t||{},a=[r.left,r.top,r.right,r.bottom].every(G),s=[r.x,r.y,r.width,r.height].every(G);if(!s&&!a)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[i,c,m,p]=s?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];F.assertIsValidBox({x:i,y:c,width:m,height:p},"Box.constructor",e),this._x=i,this._y=c,this._width=m,this._height=p}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new g(this.left,this.top)}get topRight(){return new g(this.right,this.top)}get bottomLeft(){return new g(this.left,this.bottom)}get bottomRight(){return new g(this.right,this.bottom)}round(){let[t,e,r,a]=[this.x,this.y,this.width,this.height].map(s=>Math.round(s));return new F({x:t,y:e,width:r,height:a})}floor(){let[t,e,r,a]=[this.x,this.y,this.width,this.height].map(s=>Math.floor(s));return new F({x:t,y:e,width:r,height:a})}toSquare(){let{x:t,y:e,width:r,height:a}=this,s=Math.abs(r-a);return r<a&&(t-=s/2,r+=s),a<r&&(e-=s/2,a+=s),new F({x:t,y:e,width:r,height:a})}rescale(t){let e=ue(t)?t.width:t,r=ue(t)?t.height:t;return new F({x:this.x*e,y:this.y*r,width:this.width*e,height:this.height*r})}pad(t,e){let[r,a,s,i]=[this.x-t/2,this.y-e/2,this.width+t,this.height+e];return new F({x:r,y:a,width:s,height:i})}clipAtImageBorders(t,e){let{x:r,y:a,right:s,bottom:i}=this,c=Math.max(r,0),m=Math.max(a,0),p=s-c,u=i-m,f=Math.min(p,t-c),l=Math.min(u,e-m);return new F({x:c,y:m,width:f,height:l}).floor()}shift(t,e){let{width:r,height:a}=this,s=this.x+t,i=this.y+e;return new F({x:s,y:i,width:r,height:a})}padAtBorders(t,e){let r=this.width+1,a=this.height+1,s=1,i=1,c=r,m=a,p=this.left,u=this.top,f=this.right,l=this.bottom;return f>e&&(c=-f+e+r,f=e),l>t&&(m=-l+t+a,l=t),p<1&&(m=2-p,p=1),u<1&&(m=2-u,u=1),{dy:i,edy:m,dx:s,edx:c,y:u,ey:l,x:p,ex:f,w:r,h:a}}calibrate(t){return new F({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var Nt=class extends F{constructor(t,e,r,a,s=!1){super({left:t,top:e,right:r,bottom:a},s)}};var ct=class{constructor(t,e,r,a,s){this._imageDims=new S(s.width,s.height),this._score=t,this._classScore=e,this._className=r,this._box=new F(a).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new F(this._box).rescale(this.imageDims.reverse())}forSize(t,e){return new ct(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:e})}};var E=class extends ct{constructor(t,e,r){super(t,t,"",e,r)}forSize(t,e){let{score:r,relativeBox:a,imageDims:s}=super.forSize(t,e);return new E(r,a,s)}};function br(o,t,e=!0){let r=Math.max(0,Math.min(o.right,t.right)-Math.max(o.left,t.left)),a=Math.max(0,Math.min(o.bottom,t.bottom)-Math.max(o.top,t.top)),s=r*a;return e?s/(o.area+t.area-s):s/Math.min(o.area,t.area)}function gr(o){let t=o.map(c=>c.x),e=o.map(c=>c.y),r=t.reduce((c,m)=>m<c?m:c,1/0),a=e.reduce((c,m)=>m<c?m:c,1/0),s=t.reduce((c,m)=>c<m?m:c,0),i=e.reduce((c,m)=>c<m?m:c,0);return new Nt(r,a,s,i)}function xr(o,t,e,r=!0){let a=t.map((i,c)=>({score:i,boxIndex:c})).sort((i,c)=>i.score-c.score).map(i=>i.boxIndex),s=[];for(;a.length>0;){let i=a.pop();s.push(i);let c=a,m=[];for(let p=0;p<c.length;p++){let u=c[p],f=o[i],l=o[u];m.push(br(f,l,r))}a=a.filter((p,u)=>m[u]<=e)}return s}function X(o,t){return n.tidy(()=>{let[e,r,a]=t,s=n.fill([...o.shape.slice(0,3),1],e,"float32"),i=n.fill([...o.shape.slice(0,3),1],r,"float32"),c=n.fill([...o.shape.slice(0,3),1],a,"float32"),m=n.concat([s,i,c],3);return n.sub(o,m)})}function vr(o,t=!1){return n.tidy(()=>{let[e,r]=o.shape.slice(1);if(e===r)return o;let a=Math.abs(e-r),s=Math.round(a*(t?.5:1)),i=e>r?2:1,c=l=>{let d=o.shape.slice();return d[i]=l,n.fill(d,0,"float32")},m=c(s),p=a-m.shape[i],f=[t&&p?c(p):null,o,m].filter(l=>!!l).map(l=>n.cast(l,"float32"));return n.concat(f,i)})}function Rn(o){let t=o.slice();for(let e=t.length-1;e>0;e--){let r=Math.floor(Math.random()*(e+1)),a=t[e];t[e]=t[r],t[r]=a}return t}function fe(o){return 1/(1+Math.exp(-o))}function On(o){return Math.log(o/(1-o))}var St=class extends F{constructor(t,e,r,a,s=!1){super({x:t,y:e,width:r,height:a},s)}};var Do=.5,Eo=.43,Mo=.45,$=class{constructor(t,e,r=new g(0,0)){let{width:a,height:s}=e;this._imgDims=new S(a,s),this._shift=r,this._positions=t.map(i=>i.mul(new g(a,s)).add(r))}get shift(){return new g(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new g(this.imageWidth,this.imageHeight)))}forSize(t,e){return new this.constructor(this.relativePositions,{width:t,height:e})}shiftBy(t,e){return new this.constructor(this.relativePositions,this._imgDims,new g(t,e))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,e={}){if(t){let s=t instanceof E?t.box.floor():new F(t);return this.shiftBy(s.x,s.y).align(null,e)}let{useDlibAlignment:r,minBoxPadding:a}={useDlibAlignment:!1,minBoxPadding:.2,...e};return r?this.alignDlib():this.alignMinBbox(a)}alignDlib(){let t=this.getRefPointsForAlignment(),[e,r,a]=t,s=f=>a.sub(f).magnitude(),i=(s(e)+s(r))/2,c=Math.floor(i/Mo),m=yt(t),p=Math.floor(Math.max(0,m.x-Do*c)),u=Math.floor(Math.max(0,m.y-Eo*c));return new St(p,u,Math.min(c,this.imageWidth+p),Math.min(c,this.imageHeight+u))}alignMinBbox(t){let e=gr(this.positions);return e.pad(e.width*t,e.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var yr=class extends ${getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],yt([t[3],t[4]])]}};var Lt=class extends ${getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(yt)}};var Kt=class{constructor(t,e){this._label=t,this._distance=e}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${vt(this.distance)})`:""}`}};var Qt=class extends F{constructor(e,r){super(e);this._label=r}static assertIsValidLabeledBox(e,r){if(F.assertIsValidBox(e,r),!G(e.label))throw new Error(`${r} - expected property label (${e.label}) to be a number`)}get label(){return this._label}};var Q=class{constructor(t,e){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(e)||e.some(r=>!(r instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=e}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let e=t.descriptors.map(r=>new Float32Array(r));return new Q(t.label,e)}};var _r=class extends Qt{constructor(e,r,a,s){super(e,r);this._score=a,this._classScore=s}static assertIsValidPredictedBox(e,r){if(Qt.assertIsValidLabeledBox(e,r),!It(e.score)||!It(e.classScore))throw new Error(`${r} - expected properties score (${e.score}) and (${e.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function tt(o){return o.detection instanceof E}function At(o,t){return{...o,...{detection:t}}}function Ue(){let o=window.fetch;if(!o)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:o,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function te(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function le(o){let t="";if(!o&&te())try{o=vo("fs")}catch(r){t=r.toString()}return{readFile:o?r=>new Promise((a,s)=>{o.readFile(r,(i,c)=>i?s(i):a(c))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function Xe(){let o=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,e=global.Video||global.HTMLVideoElement,r=()=>{if(o)return new o;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},a=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},s=()=>{if(e)return new e;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},i=global.fetch,c=le();return{Canvas:o||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:r,createImageElement:a,createVideoElement:s,fetch:i,...c}}function Je(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var C;function Co(){if(!C)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return C}function qe(o){C=o}function Ze(){return Je()?qe(Ue()):te()?qe(Xe()):null}function Io(o){if(C||Ze(),!C)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=C.Canvas,Image:e=C.Image}=o;C.Canvas=t,C.Image=e,C.createCanvasElement=o.createCanvasElement||(()=>new t),C.createImageElement=o.createImageElement||(()=>new e),C.ImageData=o.ImageData||C.ImageData,C.Video=o.Video||C.Video,C.fetch=o.fetch||C.fetch,C.readFile=o.readFile||C.readFile}var _={getEnv:Co,setEnv:qe,initialize:Ze,createBrowserEnv:Ue,createFileSystem:le,createNodejsEnv:Xe,monkeyPatch:Io,isBrowser:Je,isNodejs:te};Ze();function Wt(o){return!_.isNodejs()&&typeof o=="string"?document.getElementById(o):o}function k(o){let{Canvas:t,CanvasRenderingContext2D:e}=_.getEnv();if(o instanceof e)return o;let r=Wt(o);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let a=r.getContext("2d");if(!a)throw new Error("resolveContext2d - canvas 2d context is null");return a}var Ke=(a=>(a.TOP_LEFT="TOP_LEFT",a.TOP_RIGHT="TOP_RIGHT",a.BOTTOM_LEFT="BOTTOM_LEFT",a.BOTTOM_RIGHT="BOTTOM_RIGHT",a))(Ke||{}),kt=class{constructor(t={}){let{anchorPosition:e,backgroundColor:r,fontColor:a,fontSize:s,fontStyle:i,padding:c}=t;this.anchorPosition=e||"TOP_LEFT",this.backgroundColor=r||"rgba(0, 0, 0, 0.5)",this.fontColor=a||"rgba(255, 255, 255, 1)",this.fontSize=s||14,this.fontStyle=i||"Georgia",this.padding=c||4}},et=class{constructor(t,e,r={}){this.text=typeof t=="string"?[t]:t instanceof et?t.text:t,this.anchor=e,this.options=new kt(r)}measureWidth(t){let{padding:e}=this.options;return this.text.map(r=>t.measureText(r).width).reduce((r,a)=>r<a?a:r,0)+2*e}measureHeight(){let{fontSize:t,padding:e}=this.options;return this.text.length*t+2*e}getUpperLeft(t,e){let{anchorPosition:r}=this.options,a=r==="BOTTOM_RIGHT"||r==="TOP_RIGHT",s=r==="BOTTOM_LEFT"||r==="BOTTOM_RIGHT",i=this.measureWidth(t),c=this.measureHeight(),m=a?this.anchor.x-i:this.anchor.x,p=s?this.anchor.y-c:this.anchor.y;if(e){let{width:u,height:f}=e,l=Math.max(Math.min(m,u-i),0),d=Math.max(Math.min(p,f-c),0);return{x:l,y:d}}return{x:m,y:p}}draw(t){let e=Wt(t),r=k(e),{backgroundColor:a,fontColor:s,fontSize:i,fontStyle:c,padding:m}=this.options;r.font=`${i}px ${c}`;let p=this.measureWidth(r),u=this.measureHeight();r.fillStyle=a;let f=this.getUpperLeft(r,e);r.fillRect(f.x,f.y,p,u),r.fillStyle=s,this.text.forEach((l,d)=>{let b=m+f.x,x=m+f.y+(d+1)*i;r.fillText(l,b,x)})}};var de=class{constructor(t={}){let{boxColor:e,lineWidth:r,label:a,drawLabelOptions:s}=t;this.boxColor=e||"rgba(0, 0, 255, 1)",this.lineWidth=r||2,this.label=a;let i={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new kt({...i,...s})}},ee=class{constructor(t,e={}){this.box=new F(t),this.options=new de(e)}draw(t){let e=k(t),{boxColor:r,lineWidth:a}=this.options,{x:s,y:i,width:c,height:m}=this.box;e.strokeStyle=r,e.lineWidth=a,e.strokeRect(s,i,c,m);let{label:p}=this.options;p&&new et([p],{x:s-a/2,y:i},this.options.drawLabelOptions).draw(t)}};function No(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let a=r instanceof E?r.score:tt(r)?r.detection.score:void 0,s=r instanceof E?r.box:tt(r)?r.detection.box:new F(r),i=a?`${vt(a)}`:void 0;new ee(s,{label:i}).draw(o)})}function he(o){let{Image:t,Video:e}=_.getEnv();return o instanceof t&&o.complete||o instanceof e&&o.readyState>=3}function Tr(o){return new Promise((t,e)=>{(o instanceof _.getEnv().Canvas||he(o))&&t(null);function r(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",a),s.currentTarget.removeEventListener("error",r),e(s))}function a(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",a),s.currentTarget.removeEventListener("error",r),t(s))}o.addEventListener("load",a),o.addEventListener("error",r)})}function wr(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToImage - expected buf to be of type: Blob"));let r=new FileReader;r.onload=()=>{typeof r.result!="string"&&e(new Error("bufferToImage - expected reader.result to be a string, in onload"));let a=_.getEnv().createImageElement();a.onload=()=>t(a),a.onerror=e,a.src=r.result},r.onerror=e,r.readAsDataURL(o)})}function Bt(o){let{Image:t,Video:e}=_.getEnv();return o instanceof t?new S(o.naturalWidth,o.naturalHeight):o instanceof e?new S(o.videoWidth,o.videoHeight):new S(o.width,o.height)}function Rt({width:o,height:t}){let{createCanvasElement:e}=_.getEnv(),r=e();return r.width=o,r.height=t,r}function be(o,t){let{ImageData:e}=_.getEnv();if(!(o instanceof e)&&!he(o))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:a}=t||Bt(o),s=Rt({width:r,height:a});return o instanceof e?k(s).putImageData(o,0,0):k(s).drawImage(o,0,0,r,a),s}async function Pr(o,t){let e=t||_.getEnv().createCanvasElement(),[r,a,s]=o.shape.slice(R(o)?1:0),i=n.tidy(()=>o.as3D(r,a,s).toInt());return await n.browser.toPixels(i,e),i.dispose(),e}function Qe(o){let{Image:t,Canvas:e,Video:r}=_.getEnv();return o instanceof t||o instanceof e||o instanceof r}function Fr(o,t,e=!1){let{Image:r,Canvas:a}=_.getEnv();if(!(o instanceof r||o instanceof a))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Rt({width:1,height:1});let s=Bt(o),i=t/Math.max(s.height,s.width),c=i*s.width,m=i*s.height,p=Rt({width:t,height:t}),u=o instanceof a?o:be(o),f=Math.abs(c-m)/2,l=e&&c<m?f:0,d=e&&m<c?f:0;return u.width>0&&u.height>0&&k(p).drawImage(u,l,d,c,m),p}var rt=class{constructor(t,e=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=e,this._batchSize=t.length,t.forEach((r,a)=>{if(K(r)){this._imageTensors[a]=r,this._inputDimensions[a]=r.shape;return}if(R(r)){let i=r.shape[0];if(i!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${i} passed, but not supported in input array`);this._imageTensors[a]=r,this._inputDimensions[a]=r.shape.slice(1);return}let s=r instanceof _.getEnv().Canvas?r:be(r);this._canvases[a]=s,this._inputDimensions[a]=[s.height,s.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return U(this.batchSize,0,1).map((t,e)=>this.getReshapedInputDimensions(e))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let e=this.getInputWidth(t),r=this.getInputHeight(t);return je({width:e,height:r},this.inputSize)}toBatchTensor(t,e=!0){return this._inputSize=t,n.tidy(()=>{let r=U(this.batchSize,0,1).map(s=>{let i=this.getInput(s);if(i instanceof n.Tensor){let c=R(i)?i:n.expandDims(i);return c=vr(c,e),(c.shape[1]!==t||c.shape[2]!==t)&&(c=n.image.resizeBilinear(c,[t,t],!1,!1)),c.as3D(t,t,3)}if(i instanceof _.getEnv().Canvas)return n.browser.fromPixels(Fr(i,t,e));throw new Error(`toBatchTensor - at batchIdx ${s}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${i}`)});return n.stack(r.map(s=>n.cast(s,"float32"))).as4D(this.batchSize,t,t,3)})}};async function M(o){if(o instanceof rt)return o;let t=Array.isArray(o)?o:[o];if(!t.length)throw new Error("toNetInput - empty array passed as input");let e=a=>Array.isArray(o)?` at input index ${a}:`:"",r=t.map(Wt);return r.forEach((a,s)=>{if(!Qe(a)&&!K(a)&&!R(a))throw typeof t[s]=="string"?new Error(`toNetInput -${e(s)} string passed, but could not resolve HTMLElement for element id ${t[s]}`):new Error(`toNetInput -${e(s)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(R(a)){let i=a.shape[0];if(i!==1)throw new Error(`toNetInput -${e(s)} tf.Tensor4D with batchSize ${i} passed, but not supported in input array`)}}),await Promise.all(r.map(a=>Qe(a)&&Tr(a))),new rt(r,Array.isArray(o))}async function re(o,t){let{Canvas:e}=_.getEnv(),r=o;if(!(o instanceof e)){let i=await M(o);if(i.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let c=i.getInput(0);r=c instanceof e?c:await Pr(c)}let a=k(r);return t.map(i=>i instanceof E?i.forSize(r.width,r.height).box.floor():i).map(i=>i.clipAtImageBorders(r.width,r.height)).map(({x:i,y:c,width:m,height:p})=>{let u=Rt({width:m,height:p});return m>0&&p>0&&k(u).putImageData(a.getImageData(i,c,m,p),0,0),u})}async function oe(o,t){if(!K(o)&&!R(o))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(R(o)&&o.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return n.tidy(()=>{let[e,r,a]=o.shape.slice(R(o)?1:0);return t.map(c=>c instanceof E?c.forSize(r,e).box:c).map(c=>c.clipAtImageBorders(r,e)).filter(c=>c.width>0&&c.height>0).map(({x:c,y:m,width:p,height:u})=>n.slice3d(o.as3D(e,r,a),[m,c,0],[u,p,a]))})}async function mt(o,t){let{fetch:e}=_.getEnv(),r=await e(o,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}async function Fi(o){let t=await mt(o),e=await t.blob();if(!e.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${e.type}, for url: ${t.url}`);return wr(e)}async function Dr(o){return(await mt(o)).json()}async function Ii(o){return new Float32Array(await(await mt(o)).arrayBuffer())}function Er(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToVideo - expected buf to be of type: Blob"));let r=_.getEnv().createVideoElement();r.oncanplay=()=>t(r),r.onerror=e,r.playsInline=!0,r.muted=!0,r.src=URL.createObjectURL(o),r.play()})}async function ki(o){let t=await mt(o),e=await t.blob();if(!e.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${e.type}, for url: ${t.url}`);return Er(e)}function ge(o,t){let e=`${t}-weights_manifest.json`;if(!o)return{modelBaseUri:"",manifestUri:e};if(o==="/")return{modelBaseUri:"/",manifestUri:`/${e}`};let r=o.startsWith("http://")?"http://":o.startsWith("https://")?"https://":"";o=o.replace(r,"");let a=o.split("/").filter(c=>c),s=o.endsWith(".json")?a[a.length-1]:e,i=r+(o.endsWith(".json")?a.slice(0,a.length-1):a).join("/");return i=o.startsWith("/")?`/${i}`:i,{modelBaseUri:i,manifestUri:i==="/"?`/${s}`:`${i}/${s}`}}async function Mr(o,t){let{manifestUri:e,modelBaseUri:r}=ge(o,t),a=await Dr(e);return n.io.loadWeights(a,r)}function Vi(o,t,e=!1){let{width:r,height:a}=e?Bt(t):t;return o.width=r,o.height=a,{width:r,height:a}}var I=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:e,objProp:r}=this.traversePropertyPath(t);return e[r]}reassignParamFromPath(t,e){let{obj:r,objProp:a}=this.traversePropertyPath(t);r[a].dispose(),r[a]=e}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof n.Variable)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof n.Variable))}variable(){this.getFrozenParams().forEach(({path:t,tensor:e})=>{this.reassignParamFromPath(t,e.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:e})=>{let r=n.tensor(e.dataSync());e.dispose(),this.reassignParamFromPath(t,r)})}dispose(t=!0){this.getParamList().forEach(e=>{if(t&&e.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${e.path}`);e.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,e)=>t.concat(e)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let e=await Mr(t,this.getDefaultModelName());this.loadFromWeightMap(e)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:e}=_.getEnv(),{manifestUri:r,modelBaseUri:a}=ge(t,this.getDefaultModelName()),s=p=>Promise.all(p.map(u=>e(u).then(f=>f.buffer))),i=n.io.weightsLoaderFactory(s),c=JSON.parse((await e(r)).toString()),m=await i(c,a);this.loadFromWeightMap(m)}loadFromWeightMap(t){let{paramMappings:e,params:r}=this.extractParamsFromWeightMap(t);this._paramMappings=e,this._params=r}extractWeights(t){let{paramMappings:e,params:r}=this.extractParams(t);this._paramMappings=e,this._params=r}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let e=t.split("/").reduce((s,i)=>{if(!s.nextObj.hasOwnProperty(i))throw new Error(`traversePropertyPath - object does not have property ${i}, for path ${t}`);return{obj:s.nextObj,objProp:i,nextObj:s.nextObj[i]}},{nextObj:this.params}),{obj:r,objProp:a}=e;if(!r||!a||!(r[a]instanceof n.Tensor))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:r,objProp:a}}};function W(o,t,e){return n.tidy(()=>{let r=n.separableConv2d(o,t.depthwise_filter,t.pointwise_filter,e,"same");return r=n.add(r,t.bias),r})}function xe(o,t,e=!1){return n.tidy(()=>{let r=n.relu(e?n.add(n.conv2d(o,t.conv0.filters,[2,2],"same"),t.conv0.bias):W(o,t.conv0,[2,2])),a=W(r,t.conv1,[1,1]),s=n.relu(n.add(r,a)),i=W(s,t.conv2,[1,1]);return n.relu(n.add(r,n.add(a,i)))})}function ne(o,t,e=!1,r=!0){return n.tidy(()=>{let a=n.relu(e?n.add(n.conv2d(o,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):W(o,t.conv0,r?[2,2]:[1,1])),s=W(a,t.conv1,[1,1]),i=n.relu(n.add(a,s)),c=W(i,t.conv2,[1,1]),m=n.relu(n.add(a,n.add(s,c))),p=W(m,t.conv3,[1,1]);return n.relu(n.add(a,n.add(s,n.add(c,p))))})}function _t(o,t,e="same",r=!1){return n.tidy(()=>{let a=n.add(n.conv2d(o,t.filters,[1,1],e),t.bias);return r?n.relu(a):a})}function L(o,t){Object.keys(o).forEach(e=>{t.some(r=>r.originalPath===e)||o[e].dispose()})}function $t(o,t){return(e,r,a,s)=>{let i=n.tensor4d(o(e*r*a*a),[a,a,e,r]),c=n.tensor1d(o(r));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/bias`}),{filters:i,bias:c}}}function ve(o,t){return(e,r,a)=>{let s=n.tensor2d(o(e*r),[e,r]),i=n.tensor1d(o(r));return t.push({paramPath:`${a}/weights`},{paramPath:`${a}/bias`}),{weights:s,bias:i}}}var ae=class{constructor(t,e,r){this.depthwise_filter=t;this.pointwise_filter=e;this.bias=r}};function Ot(o,t){return(e,r,a)=>{let s=n.tensor4d(o(9*e),[3,3,e,1]),i=n.tensor4d(o(e*r),[1,1,e,r]),c=n.tensor1d(o(r));return t.push({paramPath:`${a}/depthwise_filter`},{paramPath:`${a}/pointwise_filter`},{paramPath:`${a}/bias`}),new ae(s,i,c)}}function Ht(o){return t=>{let e=o(`${t}/depthwise_filter`,4),r=o(`${t}/pointwise_filter`,4),a=o(`${t}/bias`,1);return new ae(e,r,a)}}function B(o,t){return(e,r,a)=>{let s=o[e];if(!xt(s,r))throw new Error(`expected weightMap[${e}] to be a Tensor${r}D, instead have ${s}`);return t.push({originalPath:e,paramPath:a||e}),s}}function A(o){let t=o;function e(a){let s=t.slice(0,a);return t=t.slice(a),s}function r(){return t}return{extractWeights:e,getRemainingWeights:r}}function ye(o,t){let e=$t(o,t),r=Ot(o,t);function a(i,c,m,p=!1){let u=p?e(i,c,3,`${m}/conv0`):r(i,c,`${m}/conv0`),f=r(c,c,`${m}/conv1`),l=r(c,c,`${m}/conv2`);return{conv0:u,conv1:f,conv2:l}}function s(i,c,m,p=!1){let{conv0:u,conv1:f,conv2:l}=a(i,c,m,p),d=r(c,c,`${m}/conv3`);return{conv0:u,conv1:f,conv2:l,conv3:d}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:s}}function Cr(o){let t=[],{extractWeights:e,getRemainingWeights:r}=A(o),{extractDenseBlock4Params:a}=ye(e,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),c=a(64,128,"dense2"),m=a(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:c,dense3:m}}}function _e(o){return t=>{let e=o(`${t}/filters`,4),r=o(`${t}/bias`,1);return{filters:e,bias:r}}}function Te(o,t){let e=B(o,t),r=_e(e),a=Ht(e);function s(c,m=!1){let p=m?r(`${c}/conv0`):a(`${c}/conv0`),u=a(`${c}/conv1`),f=a(`${c}/conv2`);return{conv0:p,conv1:u,conv2:f}}function i(c,m=!1){let p=m?r(`${c}/conv0`):a(`${c}/conv0`),u=a(`${c}/conv1`),f=a(`${c}/conv2`),l=a(`${c}/conv3`);return{conv0:p,conv1:u,conv2:f,conv3:l}}return{extractDenseBlock3Params:s,extractDenseBlock4Params:i}}function Ir(o){let t=[],{extractDenseBlock4Params:e}=Te(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2"),dense3:e("dense3")};return L(o,t),{params:r,paramMappings:t}}var zt=class extends I{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceFeatureExtractor - load model before inference");return n.tidy(()=>{let r=n.cast(t.toBatchTensor(112,!0),"float32"),s=X(r,[122.782,117.001,104.298]).div(255),i=ne(s,e.dense0,!0);return i=ne(i,e.dense1),i=ne(i,e.dense2),i=ne(i,e.dense3),i=n.avgPool(i,[7,7],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await M(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return Ir(t)}extractParams(t){return Cr(t)}};function se(o,t){return n.tidy(()=>n.add(n.matMul(o,t.weights),t.bias))}function Nr(o,t,e){let r=[],{extractWeights:a,getRemainingWeights:s}=A(o),c=ve(a,r)(t,e,"fc");if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{paramMappings:r,params:{fc:c}}}function Sr(o){let t=[],e=B(o,t);function r(s){let i=e(`${s}/weights`,2),c=e(`${s}/bias`,1);return{weights:i,bias:c}}let a={fc:r("fc")};return L(o,t),{params:a,paramMappings:t}}function we(o){let t={},e={};return Object.keys(o).forEach(r=>{let a=r.startsWith("fc")?e:t;a[r]=o[r]}),{featureExtractorMap:t,classifierMap:e}}var Vt=class extends I{constructor(e,r){super(e);this._faceFeatureExtractor=r}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return n.tidy(()=>{let a=e instanceof rt?this.faceFeatureExtractor.forwardInput(e):e;return se(a.as2D(a.shape[0],-1),r.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:a}=this.extractClassifierParams(e);this._params=r,this._paramMappings=a}extractClassifierParams(e){return Nr(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:a}=we(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Sr(a)}extractParams(e){let r=this.getClassifierChannelsIn(),a=this.getClassifierChannelsOut(),s=a*r+a,i=e.slice(0,e.length-s),c=e.slice(e.length-s);return this.faceFeatureExtractor.extractWeights(i),this.extractClassifierParams(c)}};var Lr=["neutral","happy","sad","angry","fearful","disgusted","surprised"],pt=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);Lr.forEach((e,r)=>{this[e]=t[r]})}asSortedArray(){return Lr.map(t=>({expression:t,probability:this[t]})).sort((t,e)=>e.probability-t.probability)}};var Pe=class extends Vt{constructor(t=new zt){super("FaceExpressionNet",t)}forwardInput(t){return n.tidy(()=>n.softmax(this.runNet(t)))}async forward(t){return this.forwardInput(await M(t))}async predictExpressions(t){let e=await M(t),r=await this.forwardInput(e),a=await Promise.all(n.unstack(r).map(async i=>{let c=i.dataSync();return i.dispose(),c}));r.dispose();let s=a.map(i=>new pt(i));return e.isBatchInput?s:s[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function Ar(o){return o.expressions instanceof pt}function tr(o,t){return{...o,...{expressions:t}}}function So(o,t,e=.1,r){(Array.isArray(t)?t:[t]).forEach(s=>{let i=s instanceof pt?s:Ar(s)?s.expressions:void 0;if(!i)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let m=i.asSortedArray().filter(f=>f.probability>e),p=tt(s)?s.detection.box.bottomLeft:r||new g(0,0);new et(m.map(f=>`${f.expression} (${vt(f.probability)})`),p).draw(o)})}function Yt(o){return tt(o)&&o.landmarks instanceof $&&o.unshiftedLandmarks instanceof $&&o.alignedRect instanceof E}function Lo(o){let t=m=>m*180/Math.PI,e=(m,p)=>Math.sqrt((m._x-p._x)**2+(m._y-p._y)**2),r={roll:void 0,pitch:void 0,yaw:void 0},a=(m,p,u)=>{let f=Math.floor(m._x-p._x),l=Math.floor(p._x-u._x);return f-l},s=(m,p)=>{let u=Math.hypot(p._x-m._x,p._y-m._y),f=p._y-m._y,l=Math.asin(f/u),d=t(l),b=Math.floor(90-d),x=p._x-m._x<0?-1:1;return b*x},i=(m,p,u)=>{let f=e(m,u),l={_x:(m._x+u._x)/2,_y:(m._y+u._y)/2},d=e(p,l),b=Math.atan(d/f),x=Math.floor(t(b)),w=l._y-p._y<0?-1:1;return x*w};if(!o||!o._positions||o._positions.length!==68)return r;let c=o._positions;return r.roll=s(c[27],c[66]),r.pitch=i(c[14],c[30],c[2]),r.yaw=a(c[14],c[33],c[2]),r}function ie(o,t){let{box:e}=o.detection,r=t.shiftBy(e.x,e.y),a=r.align(),{imageDims:s}=o.detection,i=new E(o.detection.score,a.rescale(s.reverse()),s),c=Lo(t);return{...o,...{landmarks:r,unshiftedLandmarks:t,alignedRect:i,angle:c}}}var Fe=class{constructor(t={}){let{drawLines:e=!0,drawPoints:r=!0,lineWidth:a,lineColor:s,pointSize:i,pointColor:c}=t;this.drawLines=e,this.drawPoints=r,this.lineWidth=a||1,this.pointSize=i||2,this.lineColor=s||"rgba(0, 255, 255, 1)",this.pointColor=c||"rgba(255, 0, 255, 1)"}},De=class{constructor(t,e={}){this.faceLandmarks=t,this.options=new Fe(e)}draw(t){let e=k(t),{drawLines:r,drawPoints:a,lineWidth:s,lineColor:i,pointSize:c,pointColor:m}=this.options;if(r&&this.faceLandmarks instanceof Lt&&(e.strokeStyle=i,e.lineWidth=s,Z(e,this.faceLandmarks.getJawOutline()),Z(e,this.faceLandmarks.getLeftEyeBrow()),Z(e,this.faceLandmarks.getRightEyeBrow()),Z(e,this.faceLandmarks.getNose()),Z(e,this.faceLandmarks.getLeftEye(),!0),Z(e,this.faceLandmarks.getRightEye(),!0),Z(e,this.faceLandmarks.getMouth(),!0)),a){e.strokeStyle=m,e.fillStyle=m;let p=u=>{e.beginPath(),e.arc(u.x,u.y,c,0,2*Math.PI),e.fill()};this.faceLandmarks.positions.forEach(p)}}};function Ao(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let a=r instanceof $?r:Yt(r)?r.landmarks:void 0;if(!a)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new De(a).draw(o)})}var kr="1.7.6";function Bo(o,t){let e=$t(o,t),r=Ot(o,t);function a(i,c,m){let p=r(i,c,`${m}/separable_conv0`),u=r(c,c,`${m}/separable_conv1`),f=e(i,c,1,`${m}/expansion_conv`);return{separable_conv0:p,separable_conv1:u,expansion_conv:f}}function s(i,c){let m=r(i,i,`${c}/separable_conv0`),p=r(i,i,`${c}/separable_conv1`),u=r(i,i,`${c}/separable_conv2`);return{separable_conv0:m,separable_conv1:p,separable_conv2:u}}return{extractConvParams:e,extractSeparableConvParams:r,extractReductionBlockParams:a,extractMainBlockParams:s}}function Br(o,t){let e=[],{extractWeights:r,getRemainingWeights:a}=A(o),{extractConvParams:s,extractSeparableConvParams:i,extractReductionBlockParams:c,extractMainBlockParams:m}=Bo(r,e),p=s(3,32,3,"entry_flow/conv_in"),u=c(32,64,"entry_flow/reduction_block_0"),f=c(64,128,"entry_flow/reduction_block_1"),l={conv_in:p,reduction_block_0:u,reduction_block_1:f},d={};U(t,0,1).forEach(h=>{d[`main_block_${h}`]=m(128,`middle_flow/main_block_${h}`)});let b=c(128,256,"exit_flow/reduction_block"),x=i(256,512,"exit_flow/separable_conv"),w={reduction_block:b,separable_conv:x};if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:e,params:{entry_flow:l,middle_flow:d,exit_flow:w}}}function Ro(o,t){let e=B(o,t),r=_e(e),a=Ht(e);function s(c){let m=a(`${c}/separable_conv0`),p=a(`${c}/separable_conv1`),u=r(`${c}/expansion_conv`);return{separable_conv0:m,separable_conv1:p,expansion_conv:u}}function i(c){let m=a(`${c}/separable_conv0`),p=a(`${c}/separable_conv1`),u=a(`${c}/separable_conv2`);return{separable_conv0:m,separable_conv1:p,separable_conv2:u}}return{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:s,extractMainBlockParams:i}}function Rr(o,t){let e=[],{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:s,extractMainBlockParams:i}=Ro(o,e),c=r("entry_flow/conv_in"),m=s("entry_flow/reduction_block_0"),p=s("entry_flow/reduction_block_1"),u={conv_in:c,reduction_block_0:m,reduction_block_1:p},f={};U(t,0,1).forEach(x=>{f[`main_block_${x}`]=i(`middle_flow/main_block_${x}`)});let l=s("exit_flow/reduction_block"),d=a("exit_flow/separable_conv"),b={reduction_block:l,separable_conv:d};return L(o,e),{params:{entry_flow:u,middle_flow:f,exit_flow:b},paramMappings:e}}function $r(o,t,e){return n.add(n.conv2d(o,t.filters,e,"same"),t.bias)}function er(o,t,e=!0){let r=e?n.relu(o):o;return r=W(r,t.separable_conv0,[1,1]),r=W(n.relu(r),t.separable_conv1,[1,1]),r=n.maxPool(r,[3,3],[2,2],"same"),r=n.add(r,$r(o,t.expansion_conv,[2,2])),r}function $o(o,t){let e=W(n.relu(o),t.separable_conv0,[1,1]);return e=W(n.relu(e),t.separable_conv1,[1,1]),e=W(n.relu(e),t.separable_conv2,[1,1]),e=n.add(e,o),e}var Ee=class extends I{constructor(e){super("TinyXception");this._numMainBlocks=e}forwardInput(e){let{params:r}=this;if(!r)throw new Error("TinyXception - load model before inference");return n.tidy(()=>{let a=n.cast(e.toBatchTensor(112,!0),"float32"),i=X(a,[122.782,117.001,104.298]).div(255),c=n.relu($r(i,r.entry_flow.conv_in,[2,2]));return c=er(c,r.entry_flow.reduction_block_0,!1),c=er(c,r.entry_flow.reduction_block_1),U(this._numMainBlocks,0,1).forEach(m=>{c=$o(c,r.middle_flow[`main_block_${m}`])}),c=er(c,r.exit_flow.reduction_block),c=n.relu(W(c,r.exit_flow.separable_conv,[1,1])),c})}async forward(e){return this.forwardInput(await M(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return Rr(e,this._numMainBlocks)}extractParams(e){return Br(e,this._numMainBlocks)}};function Or(o){let t=[],{extractWeights:e,getRemainingWeights:r}=A(o),a=ve(e,t),s=a(512,1,"fc/age"),i=a(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:s,gender:i}}}}function Hr(o){let t=[],e=B(o,t);function r(s){let i=e(`${s}/weights`,2),c=e(`${s}/bias`,1);return{weights:i,bias:c}}let a={fc:{age:r("fc/age"),gender:r("fc/gender")}};return L(o,t),{params:a,paramMappings:t}}var rr=(e=>(e.FEMALE="female",e.MALE="male",e))(rr||{});var Me=class extends I{constructor(e=new Ee(2)){super("AgeGenderNet");this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return n.tidy(()=>{let a=e instanceof rt?this.faceFeatureExtractor.forwardInput(e):e,s=n.avgPool(a,[7,7],[2,2],"valid").as2D(a.shape[0],-1),i=se(s,r.fc.age).as1D(),c=se(s,r.fc.gender);return{age:i,gender:c}})}forwardInput(e){return n.tidy(()=>{let{age:r,gender:a}=this.runNet(e);return{age:r,gender:n.softmax(a)}})}async forward(e){return this.forwardInput(await M(e))}async predictAgeAndGender(e){let r=await M(e),a=await this.forwardInput(r),s=n.unstack(a.age),i=n.unstack(a.gender),c=s.map((p,u)=>({ageTensor:p,genderTensor:i[u]})),m=await Promise.all(c.map(async({ageTensor:p,genderTensor:u})=>{let f=p.dataSync()[0],l=u.dataSync()[0],d=l>.5,b=d?"male":"female",x=d?l:1-l;return p.dispose(),u.dispose(),{age:f,gender:b,genderProbability:x}}));return a.age.dispose(),a.gender.dispose(),r.isBatchInput?m:m[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:a}=this.extractClassifierParams(e);this._params=r,this._paramMappings=a}extractClassifierParams(e){return Or(e)}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:a}=we(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Hr(a)}extractParams(e){let a=e.slice(0,e.length-1539),s=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(a),this.extractClassifierParams(s)}};var Gt=class extends Vt{postProcess(t,e,r){let a=r.map(({width:i,height:c})=>{let m=e/Math.max(c,i);return{width:i*m,height:c*m}}),s=a.length;return n.tidy(()=>{let i=(f,l)=>n.stack([n.fill([68],f,"float32"),n.fill([68],l,"float32")],1).as2D(1,136).as1D(),c=(f,l)=>{let{width:d,height:b}=a[f];return l(d,b)?Math.abs(d-b)/2:0},m=f=>c(f,(l,d)=>l<d),p=f=>c(f,(l,d)=>d<l);return t.mul(n.fill([s,136],e,"float32")).sub(n.stack(Array.from(Array(s),(f,l)=>i(m(l),p(l))))).div(n.stack(Array.from(Array(s),(f,l)=>i(a[l].width,a[l].height))))})}forwardInput(t){return n.tidy(()=>{let e=this.runNet(t);return this.postProcess(e,t.inputSize,t.inputDimensions.map(([r,a])=>({height:r,width:a})))})}async forward(t){return this.forwardInput(await M(t))}async detectLandmarks(t){let e=await M(t),r=n.tidy(()=>n.unstack(this.forwardInput(e))),a=await Promise.all(r.map(async(s,i)=>{let c=Array.from(s.dataSync()),m=c.filter((u,f)=>pe(f)),p=c.filter((u,f)=>!pe(f));return new Lt(Array(68).fill(0).map((u,f)=>new g(m[f],p[f])),{height:e.getInputHeight(i),width:e.getInputWidth(i)})}));return r.forEach(s=>s.dispose()),e.isBatchInput?a:a[0]}getClassifierChannelsOut(){return 136}};var jt=class extends Gt{constructor(t=new zt){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};function zr(o){let t=[],{extractDenseBlock3Params:e}=Te(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2")};return L(o,t),{params:r,paramMappings:t}}function Vr(o){let t=[],{extractWeights:e,getRemainingWeights:r}=A(o),{extractDenseBlock3Params:a}=ye(e,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),c=a(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:c}}}var Ce=class extends I{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("TinyFaceFeatureExtractor - load model before inference");return n.tidy(()=>{let r=n.cast(t.toBatchTensor(112,!0),"float32"),s=X(r,[122.782,117.001,104.298]).div(255),i=xe(s,e.dense0,!0);return i=xe(i,e.dense1),i=xe(i,e.dense2),i=n.avgPool(i,[14,14],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await M(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return zr(t)}extractParams(t){return Vr(t)}};var Ie=class extends Gt{constructor(t=new Ce){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var Yr=class extends jt{};function Gr(o,t){return n.add(n.mul(o,t.weights),t.biases)}function or(o,t,e,r,a="same"){let{filters:s,bias:i}=t.conv,c=n.conv2d(o,s,e,a);return c=n.add(c,i),c=Gr(c,t.scale),r?n.relu(c):c}function jr(o,t){return or(o,t,[1,1],!0)}function nr(o,t){return or(o,t,[1,1],!1)}function Ne(o,t){return or(o,t,[2,2],!0,"valid")}function Oo(o,t){function e(c,m,p){let u=o(c),f=u.length/(m*p*p);if(Ge(f))throw new Error(`depth has to be an integer: ${f}, weights.length: ${u.length}, numFilters: ${m}, filterSize: ${p}`);return n.tidy(()=>n.transpose(n.tensor4d(u,[m,f,p,p]),[2,3,1,0]))}function r(c,m,p,u){let f=e(c,m,p),l=n.tensor1d(o(m));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/bias`}),{filters:f,bias:l}}function a(c,m){let p=n.tensor1d(o(c)),u=n.tensor1d(o(c));return t.push({paramPath:`${m}/weights`},{paramPath:`${m}/biases`}),{weights:p,biases:u}}function s(c,m,p,u){let f=r(c,m,p,`${u}/conv`),l=a(m,`${u}/scale`);return{conv:f,scale:l}}function i(c,m,p,u,f=!1){let l=s((f?.5:1)*c,m,p,`${u}/conv1`),d=s(c,m,p,`${u}/conv2`);return{conv1:l,conv2:d}}return{extractConvLayerParams:s,extractResidualLayerParams:i}}function Ur(o){let{extractWeights:t,getRemainingWeights:e}=A(o),r=[],{extractConvLayerParams:a,extractResidualLayerParams:s}=Oo(t,r),i=a(4704,32,7,"conv32_down"),c=s(9216,32,3,"conv32_1"),m=s(9216,32,3,"conv32_2"),p=s(9216,32,3,"conv32_3"),u=s(36864,64,3,"conv64_down",!0),f=s(36864,64,3,"conv64_1"),l=s(36864,64,3,"conv64_2"),d=s(36864,64,3,"conv64_3"),b=s(147456,128,3,"conv128_down",!0),x=s(147456,128,3,"conv128_1"),w=s(147456,128,3,"conv128_2"),h=s(589824,256,3,"conv256_down",!0),y=s(589824,256,3,"conv256_1"),v=s(589824,256,3,"conv256_2"),D=s(589824,256,3,"conv256_down_out"),N=n.tidy(()=>n.transpose(n.tensor2d(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),e().length!==0)throw new Error(`weights remaing after extract: ${e().length}`);return{params:{conv32_down:i,conv32_1:c,conv32_2:m,conv32_3:p,conv64_down:u,conv64_1:f,conv64_2:l,conv64_3:d,conv128_down:b,conv128_1:x,conv128_2:w,conv256_down:h,conv256_1:y,conv256_2:v,conv256_down_out:D,fc:N},paramMappings:r}}function Ho(o,t){let e=B(o,t);function r(i){let c=e(`${i}/scale/weights`,1),m=e(`${i}/scale/biases`,1);return{weights:c,biases:m}}function a(i){let c=e(`${i}/conv/filters`,4),m=e(`${i}/conv/bias`,1),p=r(i);return{conv:{filters:c,bias:m},scale:p}}function s(i){return{conv1:a(`${i}/conv1`),conv2:a(`${i}/conv2`)}}return{extractConvLayerParams:a,extractResidualLayerParams:s}}function Xr(o){let t=[],{extractConvLayerParams:e,extractResidualLayerParams:r}=Ho(o,t),a=e("conv32_down"),s=r("conv32_1"),i=r("conv32_2"),c=r("conv32_3"),m=r("conv64_down"),p=r("conv64_1"),u=r("conv64_2"),f=r("conv64_3"),l=r("conv128_down"),d=r("conv128_1"),b=r("conv128_2"),x=r("conv256_down"),w=r("conv256_1"),h=r("conv256_2"),y=r("conv256_down_out"),{fc:v}=o;if(t.push({originalPath:"fc",paramPath:"fc"}),!Ye(v))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${v}`);let D={conv32_down:a,conv32_1:s,conv32_2:i,conv32_3:c,conv64_down:m,conv64_1:p,conv64_2:u,conv64_3:f,conv128_down:l,conv128_1:d,conv128_2:b,conv256_down:x,conv256_1:w,conv256_2:h,conv256_down_out:y,fc:v};return L(o,t),{params:D,paramMappings:t}}function j(o,t){let e=jr(o,t.conv1);return e=nr(e,t.conv2),e=n.add(e,o),e=n.relu(e),e}function ce(o,t){let e=Ne(o,t.conv1);e=nr(e,t.conv2);let r=n.avgPool(o,2,2,"valid"),a=n.zeros(r.shape),s=r.shape[3]!==e.shape[3];if(r.shape[1]!==e.shape[1]||r.shape[2]!==e.shape[2]){let c=[...e.shape];c[1]=1;let m=n.zeros(c);e=n.concat([e,m],1);let p=[...e.shape];p[2]=1;let u=n.zeros(p);e=n.concat([e,u],2)}return r=s?n.concat([r,a],3):r,e=n.add(r,e),e=n.relu(e),e}var Ut=class extends I{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceRecognitionNet - load model before inference");return n.tidy(()=>{let r=n.cast(t.toBatchTensor(150,!0),"float32"),s=X(r,[122.782,117.001,104.298]).div(255),i=Ne(s,e.conv32_down);i=n.maxPool(i,3,2,"valid"),i=j(i,e.conv32_1),i=j(i,e.conv32_2),i=j(i,e.conv32_3),i=ce(i,e.conv64_down),i=j(i,e.conv64_1),i=j(i,e.conv64_2),i=j(i,e.conv64_3),i=ce(i,e.conv128_down),i=j(i,e.conv128_1),i=j(i,e.conv128_2),i=ce(i,e.conv256_down),i=j(i,e.conv256_1),i=j(i,e.conv256_2),i=ce(i,e.conv256_down_out);let c=i.mean([1,2]);return n.matMul(c,e.fc)})}async forward(t){return this.forwardInput(await M(t))}async computeFaceDescriptor(t){var s;if((s=t==null?void 0:t.shape)!=null&&s.some(i=>i<=0))return new Float32Array(128);let e=await M(t),r=n.tidy(()=>n.unstack(this.forwardInput(e))),a=await Promise.all(r.map(i=>i.data()));return r.forEach(i=>i.dispose()),e.isBatchInput?a:a[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return Xr(t)}extractParams(t){return Ur(t)}};function bf(o){let t=new Ut;return t.extractWeights(o),t}function ar(o,t){return{...o,...{descriptor:t}}}function yf(o){return typeof o.age=="number"}function sr(o,t){return{...o,...{age:t}}}function Pf(o){return(o.gender==="male"||o.gender==="female")&&It(o.genderProbability)}function ir(o,t,e){return{...o,...{gender:t,genderProbability:e}}}function zo(o,t){function e(m,p){let u=n.tensor4d(o(9*m),[3,3,m,1]),f=n.tensor1d(o(m)),l=n.tensor1d(o(m)),d=n.tensor1d(o(m)),b=n.tensor1d(o(m));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/batch_norm_scale`},{paramPath:`${p}/batch_norm_offset`},{paramPath:`${p}/batch_norm_mean`},{paramPath:`${p}/batch_norm_variance`}),{filters:u,batch_norm_scale:f,batch_norm_offset:l,batch_norm_mean:d,batch_norm_variance:b}}function r(m,p,u,f,l){let d=n.tensor4d(o(m*p*u*u),[u,u,m,p]),b=n.tensor1d(o(p));return t.push({paramPath:`${f}/filters`},{paramPath:`${f}/${l?"batch_norm_offset":"bias"}`}),{filters:d,bias:b}}function a(m,p,u,f){let{filters:l,bias:d}=r(m,p,u,f,!0);return{filters:l,batch_norm_offset:d}}function s(m,p,u){let f=e(m,`${u}/depthwise_conv`),l=a(m,p,1,`${u}/pointwise_conv`);return{depthwise_conv:f,pointwise_conv:l}}function i(){let m=a(3,32,3,"mobilenetv1/conv_0"),p=s(32,64,"mobilenetv1/conv_1"),u=s(64,128,"mobilenetv1/conv_2"),f=s(128,128,"mobilenetv1/conv_3"),l=s(128,256,"mobilenetv1/conv_4"),d=s(256,256,"mobilenetv1/conv_5"),b=s(256,512,"mobilenetv1/conv_6"),x=s(512,512,"mobilenetv1/conv_7"),w=s(512,512,"mobilenetv1/conv_8"),h=s(512,512,"mobilenetv1/conv_9"),y=s(512,512,"mobilenetv1/conv_10"),v=s(512,512,"mobilenetv1/conv_11"),D=s(512,1024,"mobilenetv1/conv_12"),N=s(1024,1024,"mobilenetv1/conv_13");return{conv_0:m,conv_1:p,conv_2:u,conv_3:f,conv_4:l,conv_5:d,conv_6:b,conv_7:x,conv_8:w,conv_9:h,conv_10:y,conv_11:v,conv_12:D,conv_13:N}}function c(){let m=a(1024,256,1,"prediction_layer/conv_0"),p=a(256,512,3,"prediction_layer/conv_1"),u=a(512,128,1,"prediction_layer/conv_2"),f=a(128,256,3,"prediction_layer/conv_3"),l=a(256,128,1,"prediction_layer/conv_4"),d=a(128,256,3,"prediction_layer/conv_5"),b=a(256,64,1,"prediction_layer/conv_6"),x=a(64,128,3,"prediction_layer/conv_7"),w=r(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),h=r(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),y=r(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),v=r(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),D=r(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),N=r(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),Y=r(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),q=r(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),O=r(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),at=r(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),st=r(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),it=r(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:m,conv_1:p,conv_2:u,conv_3:f,conv_4:l,conv_5:d,conv_6:b,conv_7:x,box_predictor_0:{box_encoding_predictor:w,class_predictor:h},box_predictor_1:{box_encoding_predictor:y,class_predictor:v},box_predictor_2:{box_encoding_predictor:D,class_predictor:N},box_predictor_3:{box_encoding_predictor:Y,class_predictor:q},box_predictor_4:{box_encoding_predictor:O,class_predictor:at},box_predictor_5:{box_encoding_predictor:st,class_predictor:it}}}return{extractMobilenetV1Params:i,extractPredictionLayerParams:c}}function Jr(o){let t=[],{extractWeights:e,getRemainingWeights:r}=A(o),{extractMobilenetV1Params:a,extractPredictionLayerParams:s}=zo(e,t),i=a(),c=s(),p={extra_dim:n.tensor3d(e(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:i,prediction_layer:c,output_layer:p},paramMappings:t}}function Vo(o,t){let e=B(o,t);function r(p,u,f){let l=e(`${p}/Conv2d_${u}_pointwise/weights`,4,`${f}/filters`),d=e(`${p}/Conv2d_${u}_pointwise/convolution_bn_offset`,1,`${f}/batch_norm_offset`);return{filters:l,batch_norm_offset:d}}function a(p){let u=`mobilenetv1/conv_${p}`,f=`MobilenetV1/Conv2d_${p}_depthwise`,l=`${u}/depthwise_conv`,d=`${u}/pointwise_conv`,b=e(`${f}/depthwise_weights`,4,`${l}/filters`),x=e(`${f}/BatchNorm/gamma`,1,`${l}/batch_norm_scale`),w=e(`${f}/BatchNorm/beta`,1,`${l}/batch_norm_offset`),h=e(`${f}/BatchNorm/moving_mean`,1,`${l}/batch_norm_mean`),y=e(`${f}/BatchNorm/moving_variance`,1,`${l}/batch_norm_variance`);return{depthwise_conv:{filters:b,batch_norm_scale:x,batch_norm_offset:w,batch_norm_mean:h,batch_norm_variance:y},pointwise_conv:r("MobilenetV1",p,d)}}function s(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:a(1),conv_2:a(2),conv_3:a(3),conv_4:a(4),conv_5:a(5),conv_6:a(6),conv_7:a(7),conv_8:a(8),conv_9:a(9),conv_10:a(10),conv_11:a(11),conv_12:a(12),conv_13:a(13)}}function i(p,u){let f=e(`${p}/weights`,4,`${u}/filters`),l=e(`${p}/biases`,1,`${u}/bias`);return{filters:f,bias:l}}function c(p){let u=i(`Prediction/BoxPredictor_${p}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${p}/box_encoding_predictor`),f=i(`Prediction/BoxPredictor_${p}/ClassPredictor`,`prediction_layer/box_predictor_${p}/class_predictor`);return{box_encoding_predictor:u,class_predictor:f}}function m(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:c(0),box_predictor_1:c(1),box_predictor_2:c(2),box_predictor_3:c(3),box_predictor_4:c(4),box_predictor_5:c(5)}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:m}}function qr(o){let t=[],{extractMobilenetV1Params:e,extractPredictionLayerParams:r}=Vo(o,t),a=o["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!K(a))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${a}`);let s={mobilenetv1:e(),prediction_layer:r(),output_layer:{extra_dim:a}};return L(o,t),{params:s,paramMappings:t}}function H(o,t,e){return n.tidy(()=>{let r=n.conv2d(o,t.filters,e,"same");return r=n.add(r,t.batch_norm_offset),n.clipByValue(r,0,6)})}var Yo=.0010000000474974513;function Go(o,t,e){return n.tidy(()=>{let r=n.depthwiseConv2d(o,t.filters,e,"same");return r=n.batchNorm(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,Yo),n.clipByValue(r,0,6)})}function jo(o){return[2,4,6,12].some(t=>t===o)?[2,2]:[1,1]}function Zr(o,t){return n.tidy(()=>{let e,r=H(o,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((s,i)=>{let c=i+1,m=jo(c);r=Go(r,s.depthwise_conv,m),r=H(r,s.pointwise_conv,[1,1]),c===11&&(e=r)}),e===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:e}})}function Uo(o,t,e){let r=o.arraySync(),a=Math.min(r[t][0],r[t][2]),s=Math.min(r[t][1],r[t][3]),i=Math.max(r[t][0],r[t][2]),c=Math.max(r[t][1],r[t][3]),m=Math.min(r[e][0],r[e][2]),p=Math.min(r[e][1],r[e][3]),u=Math.max(r[e][0],r[e][2]),f=Math.max(r[e][1],r[e][3]),l=(i-a)*(c-s),d=(u-m)*(f-p);if(l<=0||d<=0)return 0;let b=Math.max(a,m),x=Math.max(s,p),w=Math.min(i,u),h=Math.min(c,f),y=Math.max(w-b,0)*Math.max(h-x,0);return y/(l+d-y)}function Kr(o,t,e,r,a){let s=o.shape[0],i=Math.min(e,s),c=t.map((u,f)=>({score:u,boxIndex:f})).filter(u=>u.score>a).sort((u,f)=>f.score-u.score),m=u=>u<=r?1:0,p=[];return c.forEach(u=>{if(p.length>=i)return;let f=u.score;for(let l=p.length-1;l>=0;--l){let d=Uo(o,u.boxIndex,p[l]);if(d!==0&&(u.score*=m(d),u.score<=a))break}f===u.score&&p.push(u.boxIndex)}),p}function Xo(o){let t=n.unstack(n.transpose(o,[1,0])),e=[n.sub(t[2],t[0]),n.sub(t[3],t[1])],r=[n.add(t[0],n.div(e[0],2)),n.add(t[1],n.div(e[1],2))];return{sizes:e,centers:r}}function Jo(o,t){let{sizes:e,centers:r}=Xo(o),a=n.unstack(n.transpose(t,[1,0])),s=n.div(n.mul(n.exp(n.div(a[2],5)),e[0]),2),i=n.add(n.mul(n.div(a[0],10),e[0]),r[0]),c=n.div(n.mul(n.exp(n.div(a[3],5)),e[1]),2),m=n.add(n.mul(n.div(a[1],10),e[1]),r[1]);return n.transpose(n.stack([n.sub(i,s),n.sub(m,c),n.add(i,s),n.add(m,c)]),[1,0])}function Qr(o,t,e){return n.tidy(()=>{let r=o.shape[0],a=Jo(n.reshape(n.tile(e.extra_dim,[r,1,1]),[-1,4]),n.reshape(o,[-1,4]));a=n.reshape(a,[r,a.shape[0]/r,4]);let s=n.sigmoid(n.slice(t,[0,0,1],[-1,-1,-1])),i=n.slice(s,[0,0,0],[-1,-1,1]);i=n.reshape(i,[r,i.shape[1]]);let c=n.unstack(a),m=n.unstack(i);return{boxes:c,scores:m}})}function Tt(o,t){return n.tidy(()=>{let e=o.shape[0],r=n.reshape(_t(o,t.box_encoding_predictor),[e,-1,1,4]),a=n.reshape(_t(o,t.class_predictor),[e,-1,3]);return{boxPredictionEncoding:r,classPrediction:a}})}function to(o,t,e){return n.tidy(()=>{let r=H(o,e.conv_0,[1,1]),a=H(r,e.conv_1,[2,2]),s=H(a,e.conv_2,[1,1]),i=H(s,e.conv_3,[2,2]),c=H(i,e.conv_4,[1,1]),m=H(c,e.conv_5,[2,2]),p=H(m,e.conv_6,[1,1]),u=H(p,e.conv_7,[2,2]),f=Tt(t,e.box_predictor_0),l=Tt(o,e.box_predictor_1),d=Tt(a,e.box_predictor_2),b=Tt(i,e.box_predictor_3),x=Tt(m,e.box_predictor_4),w=Tt(u,e.box_predictor_5),h=n.concat([f.boxPredictionEncoding,l.boxPredictionEncoding,d.boxPredictionEncoding,b.boxPredictionEncoding,x.boxPredictionEncoding,w.boxPredictionEncoding],1),y=n.concat([f.classPrediction,l.classPrediction,d.classPrediction,b.classPrediction,x.classPrediction,w.classPrediction],1);return{boxPredictions:h,classPredictions:y}})}var z=class{constructor({minConfidence:t,maxResults:e}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=e||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var wt=class extends I{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("SsdMobilenetv1 - load model before inference");return n.tidy(()=>{let r=n.cast(t.toBatchTensor(512,!1),"float32"),a=n.sub(n.div(r,127.5),1),s=Zr(a,e.mobilenetv1),{boxPredictions:i,classPredictions:c}=to(s.out,s.conv11,e.prediction_layer);return Qr(i,c,e.output_layer)})}async forward(t){return this.forwardInput(await M(t))}async locateFaces(t,e={}){let{maxResults:r,minConfidence:a}=new z(e),s=await M(t),{boxes:i,scores:c}=this.forwardInput(s),m=i[0],p=c[0];for(let v=1;v<i.length;v++)i[v].dispose(),c[v].dispose();let u=Array.from(p.dataSync()),l=Kr(m,u,r,.5,a),d=s.getReshapedInputDimensions(0),b=s.inputSize,x=b/d.width,w=b/d.height,h=m.arraySync(),y=l.map(v=>{let[D,N]=[Math.max(0,h[v][0]),Math.min(1,h[v][2])].map(O=>O*w),[Y,q]=[Math.max(0,h[v][1]),Math.min(1,h[v][3])].map(O=>O*x);return new E(u[v],new St(Y,D,q-Y,N-D),{height:s.getInputHeight(0),width:s.getInputWidth(0)})});return m.dispose(),p.dispose(),y}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return qr(t)}extractParams(t){return Jr(t)}};function qo(o){let t=new wt;return t.extractWeights(o),t}function bl(o){return qo(o)}var eo=class extends wt{};var ro=.4,oo=[new g(.738768,.874946),new g(2.42204,2.65704),new g(4.30971,7.04493),new g(10.246,4.59428),new g(12.6868,11.8741)],no=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],ao=[117.001,114.697,97.404],so="tiny_yolov2_model",io="tiny_yolov2_separable_conv_model";var Se=o=>typeof o=="number";function co(o){if(!o)throw new Error(`invalid config: ${o}`);if(typeof o.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${o.withSeparableConvs}`);if(!Se(o.iouThreshold)||o.iouThreshold<0||o.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${o.iouThreshold}`);if(!Array.isArray(o.classes)||!o.classes.length||!o.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(o.classes)}`);if(!Array.isArray(o.anchors)||!o.anchors.length||!o.anchors.map(t=>t||{}).every(t=>Se(t.x)&&Se(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(o.anchors)}`);if(o.meanRgb&&(!Array.isArray(o.meanRgb)||o.meanRgb.length!==3||!o.meanRgb.every(Se)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(o.meanRgb)}`)}function Xt(o){return n.tidy(()=>{let t=n.mul(o,n.scalar(.10000000149011612));return n.add(n.relu(n.sub(o,t)),t)})}function ot(o,t){return n.tidy(()=>{let e=n.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=n.conv2d(e,t.conv.filters,[1,1],"valid"),e=n.sub(e,t.bn.sub),e=n.mul(e,t.bn.truediv),e=n.add(e,t.conv.bias),Xt(e)})}function nt(o,t){return n.tidy(()=>{let e=n.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=n.separableConv2d(e,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),e=n.add(e,t.bias),Xt(e)})}function Zo(o,t){let e=$t(o,t);function r(i,c){let m=n.tensor1d(o(i)),p=n.tensor1d(o(i));return t.push({paramPath:`${c}/sub`},{paramPath:`${c}/truediv`}),{sub:m,truediv:p}}function a(i,c,m){let p=e(i,c,3,`${m}/conv`),u=r(c,`${m}/bn`);return{conv:p,bn:u}}let s=Ot(o,t);return{extractConvParams:e,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}}function mo(o,t,e,r){let{extractWeights:a,getRemainingWeights:s}=A(o),i=[],{extractConvParams:c,extractConvWithBatchNormParams:m,extractSeparableConvParams:p}=Zo(a,i),u;if(t.withSeparableConvs){let[f,l,d,b,x,w,h,y,v]=r,D=t.isFirstLayerConv2d?c(f,l,3,"conv0"):p(f,l,"conv0"),N=p(l,d,"conv1"),Y=p(d,b,"conv2"),q=p(b,x,"conv3"),O=p(x,w,"conv4"),at=p(w,h,"conv5"),st=y?p(h,y,"conv6"):void 0,it=v?p(y,v,"conv7"):void 0,gt=c(v||y||h,5*e,1,"conv8");u={conv0:D,conv1:N,conv2:Y,conv3:q,conv4:O,conv5:at,conv6:st,conv7:it,conv8:gt}}else{let[f,l,d,b,x,w,h,y,v]=r,D=m(f,l,"conv0"),N=m(l,d,"conv1"),Y=m(d,b,"conv2"),q=m(b,x,"conv3"),O=m(x,w,"conv4"),at=m(w,h,"conv5"),st=m(h,y,"conv6"),it=m(y,v,"conv7"),gt=c(v,5*e,1,"conv8");u={conv0:D,conv1:N,conv2:Y,conv3:q,conv4:O,conv5:at,conv6:st,conv7:it,conv8:gt}}if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{params:u,paramMappings:i}}function Ko(o,t){let e=B(o,t);function r(c){let m=e(`${c}/sub`,1),p=e(`${c}/truediv`,1);return{sub:m,truediv:p}}function a(c){let m=e(`${c}/filters`,4),p=e(`${c}/bias`,1);return{filters:m,bias:p}}function s(c){let m=a(`${c}/conv`),p=r(`${c}/bn`);return{conv:m,bn:p}}let i=Ht(e);return{extractConvParams:a,extractConvWithBatchNormParams:s,extractSeparableConvParams:i}}function po(o,t){let e=[],{extractConvParams:r,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}=Ko(o,e),i;if(t.withSeparableConvs){let c=t.filterSizes&&t.filterSizes.length||9;i={conv0:t.isFirstLayerConv2d?r("conv0"):s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:c>7?s("conv6"):void 0,conv7:c>8?s("conv7"):void 0,conv8:r("conv8")}}else i={conv0:a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:a("conv6"),conv7:a("conv7"),conv8:r("conv8")};return L(o,e),{params:i,paramMappings:e}}var J=class{constructor({inputSize:t,scoreThreshold:e}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=e||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var cr=class extends I{constructor(e){super("TinyYolov2");co(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,r){let a=ot(e,r.conv0);return a=n.maxPool(a,[2,2],[2,2],"same"),a=ot(a,r.conv1),a=n.maxPool(a,[2,2],[2,2],"same"),a=ot(a,r.conv2),a=n.maxPool(a,[2,2],[2,2],"same"),a=ot(a,r.conv3),a=n.maxPool(a,[2,2],[2,2],"same"),a=ot(a,r.conv4),a=n.maxPool(a,[2,2],[2,2],"same"),a=ot(a,r.conv5),a=n.maxPool(a,[2,2],[1,1],"same"),a=ot(a,r.conv6),a=ot(a,r.conv7),_t(a,r.conv8,"valid",!1)}runMobilenet(e,r){let a=this.config.isFirstLayerConv2d?Xt(_t(e,r.conv0,"valid",!1)):nt(e,r.conv0);return a=n.maxPool(a,[2,2],[2,2],"same"),a=nt(a,r.conv1),a=n.maxPool(a,[2,2],[2,2],"same"),a=nt(a,r.conv2),a=n.maxPool(a,[2,2],[2,2],"same"),a=nt(a,r.conv3),a=n.maxPool(a,[2,2],[2,2],"same"),a=nt(a,r.conv4),a=n.maxPool(a,[2,2],[2,2],"same"),a=nt(a,r.conv5),a=n.maxPool(a,[2,2],[1,1],"same"),a=r.conv6?nt(a,r.conv6):a,a=r.conv7?nt(a,r.conv7):a,_t(a,r.conv8,"valid",!1)}forwardInput(e,r){let{params:a}=this;if(!a)throw new Error("TinyYolov2 - load model before inference");return n.tidy(()=>{let s=n.cast(e.toBatchTensor(r,!1),"float32");return s=this.config.meanRgb?X(s,this.config.meanRgb):s,s=s.div(255),this.config.withSeparableConvs?this.runMobilenet(s,a):this.runTinyYolov2(s,a)})}async forward(e,r){return this.forwardInput(await M(e),r)}async detect(e,r={}){let{inputSize:a,scoreThreshold:s}=new J(r),i=await M(e),c=await this.forwardInput(i,a),m=n.tidy(()=>n.unstack(c)[0].expandDims()),p={width:i.getInputWidth(0),height:i.getInputHeight(0)},u=await this.extractBoxes(m,i.getReshapedInputDimensions(0),s);c.dispose(),m.dispose();let f=u.map(h=>h.box),l=u.map(h=>h.score),d=u.map(h=>h.classScore),b=u.map(h=>this.config.classes[h.label]);return xr(f.map(h=>h.rescale(a)),l,this.config.iouThreshold,!0).map(h=>new ct(l[h],d[h],b[h],f[h],p))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return po(e,this.config)}extractParams(e){let r=this.config.filterSizes||cr.DEFAULT_FILTER_SIZES,a=r?r.length:void 0;if(a!==7&&a!==8&&a!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${a} filterSizes in config`);return mo(e,this.config,this.boxEncodingSize,r)}async extractBoxes(e,r,a){let{width:s,height:i}=r,c=Math.max(s,i),m=c/s,p=c/i,u=e.shape[1],f=this.config.anchors.length,[l,d,b]=n.tidy(()=>{let y=e.reshape([u,u,f,this.boxEncodingSize]),v=y.slice([0,0,0,0],[u,u,f,4]),D=y.slice([0,0,0,4],[u,u,f,1]),N=this.withClassScores?n.softmax(y.slice([0,0,0,5],[u,u,f,this.config.classes.length]),3):n.scalar(0);return[v,D,N]}),x=[],w=await d.array(),h=await l.array();for(let y=0;y<u;y++)for(let v=0;v<u;v++)for(let D=0;D<f;D++){let N=fe(w[y][v][D][0]);if(!a||N>a){let Y=(v+fe(h[y][v][D][0]))/u*m,q=(y+fe(h[y][v][D][1]))/u*p,O=Math.exp(h[y][v][D][2])*this.config.anchors[D].x/u*m,at=Math.exp(h[y][v][D][3])*this.config.anchors[D].y/u*p,st=Y-O/2,it=q-at/2,gt={row:y,col:v,anchor:D},{classScore:pr,label:ur}=this.withClassScores?await this.extractPredictedClass(b,gt):{classScore:1,label:0};x.push({box:new Nt(st,it,st+O,it+at),score:N,classScore:N*pr,label:ur,...gt})}}return l.dispose(),d.dispose(),b.dispose(),x}async extractPredictedClass(e,r){let{row:a,col:s,anchor:i}=r,c=await e.array();return Array(this.config.classes.length).fill(0).map((m,p)=>c[a][s][i][p]).map((m,p)=>({classScore:m,label:p})).reduce((m,p)=>m.classScore>p.classScore?m:p)}},Pt=cr;Pt.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var Jt=class extends Pt{constructor(t=!0){let e={withSeparableConvs:t,iouThreshold:ro,classes:["face"],...t?{anchors:no,meanRgb:ao}:{anchors:oo,withClassScores:!0}};super(e)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(a=>new E(a.score,a.relativeBox,{width:a.imageWidth,height:a.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?io:so}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function id(o,t=!0){let e=new Jt(t);return e.extractWeights(o),e}var Le=class extends J{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var V=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function Ft(o,t,e,r,a=({alignedRect:s})=>s){let s=o.map(m=>Yt(m)?a(m):m.detection),i=r||(t instanceof n.Tensor?await oe(t,s):await re(t,s)),c=await e(i);return i.forEach(m=>m instanceof n.Tensor&&m.dispose()),c}async function qt(o,t,e,r,a){return Ft([o],t,async s=>e(s[0]),r,a)}var uo=.4,fo=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],lo=[117.001,114.697,97.404];var Zt=class extends Pt{constructor(){let t={withSeparableConvs:!0,iouThreshold:uo,classes:["face"],anchors:fo,meanRgb:lo,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(a=>new E(a.score,a.relativeBox,{width:a.imageWidth,height:a.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var T={ssdMobilenetv1:new wt,tinyFaceDetector:new Zt,tinyYolov2:new Jt,faceLandmark68Net:new jt,faceLandmark68TinyNet:new Ie,faceRecognitionNet:new Ut,faceExpressionNet:new Pe,ageGenderNet:new Me},Qo=(o,t)=>T.ssdMobilenetv1.locateFaces(o,t),Bd=(o,t)=>T.tinyFaceDetector.locateFaces(o,t),Rd=(o,t)=>T.tinyYolov2.locateFaces(o,t),tn=o=>T.faceLandmark68Net.detectLandmarks(o),$d=o=>T.faceLandmark68TinyNet.detectLandmarks(o),Od=o=>T.faceRecognitionNet.computeFaceDescriptor(o),Hd=o=>T.faceExpressionNet.predictExpressions(o),zd=o=>T.ageGenderNet.predictAgeAndGender(o),en=o=>T.ssdMobilenetv1.load(o),Vd=o=>T.tinyFaceDetector.load(o),Yd=o=>T.tinyYolov2.load(o),Gd=o=>T.faceLandmark68Net.load(o),jd=o=>T.faceLandmark68TinyNet.load(o),Ud=o=>T.faceRecognitionNet.load(o),Xd=o=>T.faceExpressionNet.load(o),Jd=o=>T.ageGenderNet.load(o),qd=en,Zd=Qo,Kd=tn;var Ae=class extends V{constructor(e,r,a){super();this.parentTask=e;this.input=r;this.extractedFaces=a}},Dt=class extends Ae{async run(){let t=await this.parentTask,e=await Ft(t,this.input,async r=>Promise.all(r.map(a=>T.faceExpressionNet.predictExpressions(a))),this.extractedFaces);return t.map((r,a)=>tr(r,e[a]))}withAgeAndGender(){return new Mt(this,this.input)}},Et=class extends Ae{async run(){let t=await this.parentTask;if(!t)return;let e=await qt(t,this.input,r=>T.faceExpressionNet.predictExpressions(r),this.extractedFaces);return tr(t,e)}withAgeAndGender(){return new Ct(this,this.input)}},ut=class extends Dt{withAgeAndGender(){return new lt(this,this.input)}withFaceDescriptors(){return new ht(this,this.input)}},ft=class extends Et{withAgeAndGender(){return new dt(this,this.input)}withFaceDescriptor(){return new bt(this,this.input)}};var We=class extends V{constructor(e,r,a){super();this.parentTask=e;this.input=r;this.extractedFaces=a}},Mt=class extends We{async run(){let t=await this.parentTask,e=await Ft(t,this.input,async r=>Promise.all(r.map(a=>T.ageGenderNet.predictAgeAndGender(a))),this.extractedFaces);return t.map((r,a)=>{let{age:s,gender:i,genderProbability:c}=e[a];return sr(ir(r,i,c),s)})}withFaceExpressions(){return new Dt(this,this.input)}},Ct=class extends We{async run(){let t=await this.parentTask;if(!t)return;let{age:e,gender:r,genderProbability:a}=await qt(t,this.input,s=>T.ageGenderNet.predictAgeAndGender(s),this.extractedFaces);return sr(ir(t,r,a),e)}withFaceExpressions(){return new Et(this,this.input)}},lt=class extends Mt{withFaceExpressions(){return new ut(this,this.input)}withFaceDescriptors(){return new ht(this,this.input)}},dt=class extends Ct{withFaceExpressions(){return new ft(this,this.input)}withFaceDescriptor(){return new bt(this,this.input)}};var ke=class extends V{constructor(e,r){super();this.parentTask=e;this.input=r}},ht=class extends ke{async run(){let t=await this.parentTask;return(await Ft(t,this.input,r=>Promise.all(r.map(a=>T.faceRecognitionNet.computeFaceDescriptor(a))),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}))).map((r,a)=>ar(t[a],r))}withFaceExpressions(){return new ut(this,this.input)}withAgeAndGender(){return new lt(this,this.input)}},bt=class extends ke{async run(){let t=await this.parentTask;if(!t)return;let e=await qt(t,this.input,r=>T.faceRecognitionNet.computeFaceDescriptor(r),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}));return ar(t,e)}withFaceExpressions(){return new ft(this,this.input)}withAgeAndGender(){return new dt(this,this.input)}};var Be=class extends V{constructor(e,r,a){super();this.parentTask=e;this.input=r;this.useTinyLandmarkNet=a}get landmarkNet(){return this.useTinyLandmarkNet?T.faceLandmark68TinyNet:T.faceLandmark68Net}},Re=class extends Be{async run(){let t=await this.parentTask,e=t.map(i=>i.detection),r=this.input instanceof n.Tensor?await oe(this.input,e):await re(this.input,e),a=await Promise.all(r.map(i=>this.landmarkNet.detectLandmarks(i)));return r.forEach(i=>i instanceof n.Tensor&&i.dispose()),t.filter((i,c)=>a[c]).map((i,c)=>ie(i,a[c]))}withFaceExpressions(){return new ut(this,this.input)}withAgeAndGender(){return new lt(this,this.input)}withFaceDescriptors(){return new ht(this,this.input)}},$e=class extends Be{async run(){let t=await this.parentTask;if(!t)return;let{detection:e}=t,r=this.input instanceof n.Tensor?await oe(this.input,[e]):await re(this.input,[e]),a=await this.landmarkNet.detectLandmarks(r[0]);return r.forEach(s=>s instanceof n.Tensor&&s.dispose()),ie(t,a)}withFaceExpressions(){return new ft(this,this.input)}withAgeAndGender(){return new dt(this,this.input)}withFaceDescriptor(){return new bt(this,this.input)}};var Oe=class extends V{constructor(e,r=new z){super();this.input=e;this.options=r}},me=class extends Oe{async run(){let{input:t,options:e}=this,r;if(e instanceof Le)r=T.tinyFaceDetector.locateFaces(t,e);else if(e instanceof z)r=T.ssdMobilenetv1.locateFaces(t,e);else if(e instanceof J)r=T.tinyYolov2.locateFaces(t,e);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return r}runAndExtendWithFaceDetections(){return new Promise((t,e)=>{this.run().then(r=>t(r.map(a=>At({},a)))).catch(r=>e(r))})}withFaceLandmarks(t=!1){return new Re(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new Dt(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new Mt(this.runAndExtendWithFaceDetections(),this.input)}},He=class extends Oe{async run(){let t=await new me(this.input,this.options),e=t[0];return t.forEach(r=>{r.score>e.score&&(e=r)}),e}runAndExtendWithFaceDetection(){return new Promise(async t=>{let e=await this.run();t(e?At({},e):void 0)})}withFaceLandmarks(t=!1){return new $e(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new Et(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new Ct(this.runAndExtendWithFaceDetection(),this.input)}};function Jh(o,t=new z){return new He(o,t)}function mr(o,t=new z){return new me(o,t)}async function rn(o,t){return mr(o,new z(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function eb(o,t={}){return mr(o,new J(t)).withFaceLandmarks().withFaceDescriptors()}var rb=rn;function ho(o,t){if(o.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let e=Array.from(o),r=Array.from(t);return Math.sqrt(e.map((a,s)=>a-r[s]).reduce((a,s)=>a+s*s,0))}var ze=class{constructor(t,e=.6){this._distanceThreshold=e;let r=Array.isArray(t)?t:[t];if(!r.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let a=1,s=()=>`person ${a++}`;this._labeledDescriptors=r.map(i=>{if(i instanceof Q)return i;if(i instanceof Float32Array)return new Q(s(),[i]);if(i.descriptor&&i.descriptor instanceof Float32Array)return new Q(s(),[i.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,e){return e.map(r=>ho(r,t)).reduce((r,a)=>r+a,0)/(e.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:e,label:r})=>new Kt(r,this.computeMeanDistance(t,e))).reduce((e,r)=>e.distance<r.distance?e:r)}findBestMatch(t){let e=this.matchDescriptor(t);return e.distance<this._distanceThreshold?e:new Kt("unknown",e.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let e=t.labeledDescriptors.map(r=>Q.fromJSON(r));return new ze(e,t.distanceThreshold)}};function yb(o){let t=new Zt;return t.extractWeights(o),t}function on(o,t){let{width:e,height:r}=new S(t.width,t.height);if(e<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:e,height:r})}`);if(Array.isArray(o))return o.map(a=>on(a,{width:e,height:r}));if(Yt(o)){let a=o.detection.forSize(e,r),s=o.unshiftedLandmarks.forSize(a.box.width,a.box.height);return ie(At(o,a),s)}return tt(o)?At(o,o.detection.forSize(e,r)):o instanceof $||o instanceof E?o.forSize(e,r):o}var Nb=kr;export{Me as AgeGenderNet,Nt as BoundingBox,F as Box,V as ComposableTask,ht as ComputeAllFaceDescriptorsTask,ke as ComputeFaceDescriptorsTaskBase,bt as ComputeSingleFaceDescriptorTask,Re as DetectAllFaceLandmarksTask,me as DetectAllFacesTask,Be as DetectFaceLandmarksTaskBase,Oe as DetectFacesTaskBase,$e as DetectSingleFaceLandmarksTask,He as DetectSingleFaceTask,S as Dimensions,Lr as FACE_EXPRESSION_LABELS,E as FaceDetection,eo as FaceDetectionNet,Pe as FaceExpressionNet,pt as FaceExpressions,jt as FaceLandmark68Net,Ie as FaceLandmark68TinyNet,Yr as FaceLandmarkNet,$ as FaceLandmarks,yr as FaceLandmarks5,Lt as FaceLandmarks68,Kt as FaceMatch,ze as FaceMatcher,Ut as FaceRecognitionNet,rr as Gender,Qt as LabeledBox,Q as LabeledFaceDescriptors,rt as NetInput,I as NeuralNetwork,ct as ObjectDetection,g as Point,_r as PredictedBox,St as Rect,wt as SsdMobilenetv1,z as SsdMobilenetv1Options,Zt as TinyFaceDetector,Le as TinyFaceDetectorOptions,Jt as TinyYolov2,J as TinyYolov2Options,rb as allFaces,rn as allFacesSsdMobilenetv1,eb as allFacesTinyYolov2,Tr as awaitMediaLoaded,wr as bufferToImage,Od as computeFaceDescriptor,Rt as createCanvas,be as createCanvasFromMedia,bl as createFaceDetectionNet,bf as createFaceRecognitionNet,qo as createSsdMobilenetv1,yb as createTinyFaceDetector,id as createTinyYolov2,mr as detectAllFaces,tn as detectFaceLandmarks,$d as detectFaceLandmarksTiny,Kd as detectLandmarks,Jh as detectSingleFace,Wr as draw,_ as env,ho as euclideanDistance,sr as extendWithAge,ar as extendWithFaceDescriptor,At as extendWithFaceDetection,tr as extendWithFaceExpressions,ie as extendWithFaceLandmarks,ir as extendWithGender,oe as extractFaceTensors,re as extractFaces,Fi as fetchImage,Dr as fetchJson,Ii as fetchNetWeights,mt as fetchOrThrow,ki as fetchVideo,k as getContext2dOrThrow,Bt as getMediaDimensions,Pr as imageTensorToCanvas,Fr as imageToSquare,On as inverseSigmoid,br as iou,Qe as isMediaElement,he as isMediaLoaded,yf as isWithAge,tt as isWithFaceDetection,Ar as isWithFaceExpressions,Yt as isWithFaceLandmarks,Pf as isWithGender,Jd as loadAgeGenderModel,qd as loadFaceDetectionModel,Xd as loadFaceExpressionModel,Gd as loadFaceLandmarkModel,jd as loadFaceLandmarkTinyModel,Ud as loadFaceRecognitionModel,en as loadSsdMobilenetv1Model,Vd as loadTinyFaceDetectorModel,Yd as loadTinyYolov2Model,Mr as loadWeightMap,Zd as locateFaces,Vi as matchDimensions,gr as minBbox,T as nets,xr as nonMaxSuppression,X as normalize,vr as padToSquare,zd as predictAgeAndGender,Hd as recognizeFaceExpressions,on as resizeResults,Wt as resolveInput,Rn as shuffleArray,fe as sigmoid,Qo as ssdMobilenetv1,n as tf,Bd as tinyFaceDetector,Rd as tinyYolov2,M as toNetInput,hr as utils,co as validateConfig,Nb as version};
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
+}) : x)(function(x) {
+  if (typeof require !== "undefined")
+    return require.apply(this, arguments);
+  throw new Error('Dynamic require of "' + x + '" is not supported');
+});
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __reExport = (target, mod, secondTarget) => (__copyProps(target, mod, "default"), secondTarget && __copyProps(secondTarget, mod, "default"));
+
+// dist/tfjs.esm.js
+var tfjs_esm_exports = {};
+__export(tfjs_esm_exports, {
+  version: () => version6
+});
+__reExport(tfjs_esm_exports, dist_star);
+__reExport(tfjs_esm_exports, dist_star2);
+__reExport(tfjs_esm_exports, dist_star3);
+import * as dist_star from "@tensorflow/tfjs/dist/index.js";
+import * as dist_star2 from "@tensorflow/tfjs-backend-webgl/dist/index.js";
+import * as dist_star3 from "@tensorflow/tfjs-backend-wasm/dist/index.js";
+var version = "4.0.0";
+var version2 = "4.0.0";
+var version3 = "4.0.0";
+var version4 = "4.0.0";
+var version5 = "4.0.0";
+var version6 = {
+  tfjs: version,
+  "tfjs-core": version,
+  "tfjs-converter": version2,
+  "tfjs-backend-cpu": version3,
+  "tfjs-backend-webgl": version4,
+  "tfjs-backend-wasm": version5
+};
+
+// src/draw/index.ts
+var draw_exports = {};
+__export(draw_exports, {
+  AnchorPosition: () => AnchorPosition,
+  DrawBox: () => DrawBox,
+  DrawBoxOptions: () => DrawBoxOptions,
+  DrawFaceLandmarks: () => DrawFaceLandmarks,
+  DrawFaceLandmarksOptions: () => DrawFaceLandmarksOptions,
+  DrawTextField: () => DrawTextField,
+  DrawTextFieldOptions: () => DrawTextFieldOptions,
+  drawContour: () => drawContour,
+  drawDetections: () => drawDetections,
+  drawFaceExpressions: () => drawFaceExpressions,
+  drawFaceLandmarks: () => drawFaceLandmarks
+});
+
+// src/draw/drawContour.ts
+function drawContour(ctx, points, isClosed = false) {
+  ctx.beginPath();
+  points.slice(1).forEach(({ x, y }, prevIdx) => {
+    const from = points[prevIdx];
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(x, y);
+  });
+  if (isClosed) {
+    const from = points[points.length - 1];
+    const to = points[0];
+    if (!from || !to) {
+      return;
+    }
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(to.x, to.y);
+  }
+  ctx.stroke();
+}
+
+// src/utils/index.ts
+var utils_exports = {};
+__export(utils_exports, {
+  computeReshapedDimensions: () => computeReshapedDimensions,
+  getCenterPoint: () => getCenterPoint,
+  isDimensions: () => isDimensions,
+  isEven: () => isEven,
+  isFloat: () => isFloat,
+  isTensor: () => isTensor,
+  isTensor1D: () => isTensor1D,
+  isTensor2D: () => isTensor2D,
+  isTensor3D: () => isTensor3D,
+  isTensor4D: () => isTensor4D,
+  isValidNumber: () => isValidNumber,
+  isValidProbablitiy: () => isValidProbablitiy,
+  range: () => range,
+  round: () => round
+});
+
+// src/classes/Dimensions.ts
+var Dimensions = class {
+  constructor(width, height) {
+    if (!isValidNumber(width) || !isValidNumber(height)) {
+      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);
+    }
+    this._width = width;
+    this._height = height;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  reverse() {
+    return new Dimensions(1 / this.width, 1 / this.height);
+  }
+};
+
+// src/utils/index.ts
+function isTensor(tensor2, dim) {
+  return tensor2 instanceof tfjs_esm_exports.Tensor && tensor2.shape.length === dim;
+}
+function isTensor1D(tensor2) {
+  return isTensor(tensor2, 1);
+}
+function isTensor2D(tensor2) {
+  return isTensor(tensor2, 2);
+}
+function isTensor3D(tensor2) {
+  return isTensor(tensor2, 3);
+}
+function isTensor4D(tensor2) {
+  return isTensor(tensor2, 4);
+}
+function isFloat(num) {
+  return num % 1 !== 0;
+}
+function isEven(num) {
+  return num % 2 === 0;
+}
+function round(num, prec = 2) {
+  const f = 10 ** prec;
+  return Math.floor(num * f) / f;
+}
+function isDimensions(obj) {
+  return obj && obj.width && obj.height;
+}
+function computeReshapedDimensions({ width, height }, inputSize) {
+  const scale2 = inputSize / Math.max(height, width);
+  return new Dimensions(Math.round(width * scale2), Math.round(height * scale2));
+}
+function getCenterPoint(pts) {
+  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0)).div(new Point(pts.length, pts.length));
+}
+function range(num, start, step) {
+  return Array(num).fill(0).map((_, i) => start + i * step);
+}
+function isValidNumber(num) {
+  return !!num && num !== Infinity && num !== -Infinity && !Number.isNaN(num) || num === 0;
+}
+function isValidProbablitiy(num) {
+  return isValidNumber(num) && num >= 0 && num <= 1;
+}
+
+// src/classes/Point.ts
+var Point = class {
+  constructor(x, y) {
+    this._x = x;
+    this._y = y;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  add(pt) {
+    return new Point(this.x + pt.x, this.y + pt.y);
+  }
+  sub(pt) {
+    return new Point(this.x - pt.x, this.y - pt.y);
+  }
+  mul(pt) {
+    return new Point(this.x * pt.x, this.y * pt.y);
+  }
+  div(pt) {
+    return new Point(this.x / pt.x, this.y / pt.y);
+  }
+  abs() {
+    return new Point(Math.abs(this.x), Math.abs(this.y));
+  }
+  magnitude() {
+    return Math.sqrt(this.x ** 2 + this.y ** 2);
+  }
+  floor() {
+    return new Point(Math.floor(this.x), Math.floor(this.y));
+  }
+};
+
+// src/classes/Box.ts
+var Box = class {
+  static isRect(rect) {
+    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);
+  }
+  static assertIsValidBox(box, callee, allowNegativeDimensions = false) {
+    if (!Box.isRect(box)) {
+      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);
+    }
+    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {
+      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);
+    }
+  }
+  constructor(_box, allowNegativeDimensions = true) {
+    const box = _box || {};
+    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);
+    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);
+    if (!isRect && !isBbox) {
+      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);
+    }
+    const [x, y, width, height] = isRect ? [box.x, box.y, box.width, box.height] : [box.left, box.top, box.right - box.left, box.bottom - box.top];
+    Box.assertIsValidBox({
+      x,
+      y,
+      width,
+      height
+    }, "Box.constructor", allowNegativeDimensions);
+    this._x = x;
+    this._y = y;
+    this._width = width;
+    this._height = height;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  get left() {
+    return this.x;
+  }
+  get top() {
+    return this.y;
+  }
+  get right() {
+    return this.x + this.width;
+  }
+  get bottom() {
+    return this.y + this.height;
+  }
+  get area() {
+    return this.width * this.height;
+  }
+  get topLeft() {
+    return new Point(this.left, this.top);
+  }
+  get topRight() {
+    return new Point(this.right, this.top);
+  }
+  get bottomLeft() {
+    return new Point(this.left, this.bottom);
+  }
+  get bottomRight() {
+    return new Point(this.right, this.bottom);
+  }
+  round() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.round(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  floor() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.floor(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  toSquare() {
+    let {
+      x,
+      y,
+      width,
+      height
+    } = this;
+    const diff = Math.abs(width - height);
+    if (width < height) {
+      x -= diff / 2;
+      width += diff;
+    }
+    if (height < width) {
+      y -= diff / 2;
+      height += diff;
+    }
+    return new Box({ x, y, width, height });
+  }
+  rescale(s) {
+    const scaleX = isDimensions(s) ? s.width : s;
+    const scaleY = isDimensions(s) ? s.height : s;
+    return new Box({
+      x: this.x * scaleX,
+      y: this.y * scaleY,
+      width: this.width * scaleX,
+      height: this.height * scaleY
+    });
+  }
+  pad(padX, padY) {
+    const [x, y, width, height] = [
+      this.x - padX / 2,
+      this.y - padY / 2,
+      this.width + padX,
+      this.height + padY
+    ];
+    return new Box({ x, y, width, height });
+  }
+  clipAtImageBorders(imgWidth, imgHeight) {
+    const { x, y, right, bottom } = this;
+    const clippedX = Math.max(x, 0);
+    const clippedY = Math.max(y, 0);
+    const newWidth = right - clippedX;
+    const newHeight = bottom - clippedY;
+    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);
+    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);
+    return new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight }).floor();
+  }
+  shift(sx, sy) {
+    const { width, height } = this;
+    const x = this.x + sx;
+    const y = this.y + sy;
+    return new Box({ x, y, width, height });
+  }
+  padAtBorders(imageHeight, imageWidth) {
+    const w = this.width + 1;
+    const h = this.height + 1;
+    const dx = 1;
+    const dy = 1;
+    let edx = w;
+    let edy = h;
+    let x = this.left;
+    let y = this.top;
+    let ex = this.right;
+    let ey = this.bottom;
+    if (ex > imageWidth) {
+      edx = -ex + imageWidth + w;
+      ex = imageWidth;
+    }
+    if (ey > imageHeight) {
+      edy = -ey + imageHeight + h;
+      ey = imageHeight;
+    }
+    if (x < 1) {
+      edy = 2 - x;
+      x = 1;
+    }
+    if (y < 1) {
+      edy = 2 - y;
+      y = 1;
+    }
+    return { dy, edy, dx, edx, y, ey, x, ex, w, h };
+  }
+  calibrate(region) {
+    return new Box({
+      left: this.left + region.left * this.width,
+      top: this.top + region.top * this.height,
+      right: this.right + region.right * this.width,
+      bottom: this.bottom + region.bottom * this.height
+    }).toSquare().round();
+  }
+};
+
+// src/classes/BoundingBox.ts
+var BoundingBox = class extends Box {
+  constructor(left, top, right, bottom, allowNegativeDimensions = false) {
+    super({ left, top, right, bottom }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/ObjectDetection.ts
+var ObjectDetection = class {
+  constructor(score, classScore, className, relativeBox, imageDims) {
+    this._imageDims = new Dimensions(imageDims.width, imageDims.height);
+    this._score = score;
+    this._classScore = classScore;
+    this._className = className;
+    this._box = new Box(relativeBox).rescale(this._imageDims);
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+  get className() {
+    return this._className;
+  }
+  get box() {
+    return this._box;
+  }
+  get imageDims() {
+    return this._imageDims;
+  }
+  get imageWidth() {
+    return this.imageDims.width;
+  }
+  get imageHeight() {
+    return this.imageDims.height;
+  }
+  get relativeBox() {
+    return new Box(this._box).rescale(this.imageDims.reverse());
+  }
+  forSize(width, height) {
+    return new ObjectDetection(
+      this.score,
+      this.classScore,
+      this.className,
+      this.relativeBox,
+      { width, height }
+    );
+  }
+};
+
+// src/classes/FaceDetection.ts
+var FaceDetection = class extends ObjectDetection {
+  constructor(score, relativeBox, imageDims) {
+    super(score, score, "", relativeBox, imageDims);
+  }
+  forSize(width, height) {
+    const { score, relativeBox, imageDims } = super.forSize(width, height);
+    return new FaceDetection(score, relativeBox, imageDims);
+  }
+};
+
+// src/ops/iou.ts
+function iou(box1, box2, isIOU = true) {
+  const width = Math.max(0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));
+  const height = Math.max(0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));
+  const interSection = width * height;
+  return isIOU ? interSection / (box1.area + box2.area - interSection) : interSection / Math.min(box1.area, box2.area);
+}
+
+// src/ops/minBbox.ts
+function minBbox(pts) {
+  const xs = pts.map((pt) => pt.x);
+  const ys = pts.map((pt) => pt.y);
+  const minX = xs.reduce((min, x) => x < min ? x : min, Infinity);
+  const minY = ys.reduce((min, y) => y < min ? y : min, Infinity);
+  const maxX = xs.reduce((max, x) => max < x ? x : max, 0);
+  const maxY = ys.reduce((max, y) => max < y ? y : max, 0);
+  return new BoundingBox(minX, minY, maxX, maxY);
+}
+
+// src/ops/nonMaxSuppression.ts
+function nonMaxSuppression(boxes, scores, iouThreshold, isIOU = true) {
+  let indicesSortedByScore = scores.map((score, boxIndex) => ({ score, boxIndex })).sort((c1, c2) => c1.score - c2.score).map((c) => c.boxIndex);
+  const pick = [];
+  while (indicesSortedByScore.length > 0) {
+    const curr = indicesSortedByScore.pop();
+    pick.push(curr);
+    const indices = indicesSortedByScore;
+    const outputs = [];
+    for (let i = 0; i < indices.length; i++) {
+      const idx = indices[i];
+      const currBox = boxes[curr];
+      const idxBox = boxes[idx];
+      outputs.push(iou(currBox, idxBox, isIOU));
+    }
+    indicesSortedByScore = indicesSortedByScore.filter(
+      (_, j) => outputs[j] <= iouThreshold
+    );
+  }
+  return pick;
+}
+
+// src/ops/normalize.ts
+function normalize(x, meanRgb) {
+  return tfjs_esm_exports.tidy(() => {
+    const [r, g, b] = meanRgb;
+    const avg_r = tfjs_esm_exports.fill([...x.shape.slice(0, 3), 1], r, "float32");
+    const avg_g = tfjs_esm_exports.fill([...x.shape.slice(0, 3), 1], g, "float32");
+    const avg_b = tfjs_esm_exports.fill([...x.shape.slice(0, 3), 1], b, "float32");
+    const avg_rgb = tfjs_esm_exports.concat([avg_r, avg_g, avg_b], 3);
+    return tfjs_esm_exports.sub(x, avg_rgb);
+  });
+}
+
+// src/ops/padToSquare.ts
+function padToSquare(imgTensor, isCenterImage = false) {
+  return tfjs_esm_exports.tidy(() => {
+    const [height, width] = imgTensor.shape.slice(1);
+    if (height === width)
+      return imgTensor;
+    const dimDiff = Math.abs(height - width);
+    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));
+    const paddingAxis = height > width ? 2 : 1;
+    const createPaddingTensor = (paddingAmountLocal) => {
+      const paddingTensorShape = imgTensor.shape.slice();
+      paddingTensorShape[paddingAxis] = paddingAmountLocal;
+      return tfjs_esm_exports.fill(paddingTensorShape, 0, "float32");
+    };
+    const paddingTensorAppend = createPaddingTensor(paddingAmount);
+    const remainingPaddingAmount = dimDiff - paddingTensorAppend.shape[paddingAxis];
+    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;
+    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend].filter((t) => !!t).map((t) => tfjs_esm_exports.cast(t, "float32"));
+    return tfjs_esm_exports.concat(tensorsToStack, paddingAxis);
+  });
+}
+
+// src/ops/shuffleArray.ts
+function shuffleArray(inputArray) {
+  const array = inputArray.slice();
+  for (let i = array.length - 1; i > 0; i--) {
+    const j = Math.floor(Math.random() * (i + 1));
+    const x = array[i];
+    array[i] = array[j];
+    array[j] = x;
+  }
+  return array;
+}
+
+// src/ops/index.ts
+function sigmoid(x) {
+  return 1 / (1 + Math.exp(-x));
+}
+function inverseSigmoid(x) {
+  return Math.log(x / (1 - x));
+}
+
+// src/classes/Rect.ts
+var Rect = class extends Box {
+  constructor(x, y, width, height, allowNegativeDimensions = false) {
+    super({ x, y, width, height }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/FaceLandmarks.ts
+var relX = 0.5;
+var relY = 0.43;
+var relScale = 0.45;
+var FaceLandmarks = class {
+  constructor(relativeFaceLandmarkPositions, imgDims, shift = new Point(0, 0)) {
+    const { width, height } = imgDims;
+    this._imgDims = new Dimensions(width, height);
+    this._shift = shift;
+    this._positions = relativeFaceLandmarkPositions.map(
+      (pt) => pt.mul(new Point(width, height)).add(shift)
+    );
+  }
+  get shift() {
+    return new Point(this._shift.x, this._shift.y);
+  }
+  get imageWidth() {
+    return this._imgDims.width;
+  }
+  get imageHeight() {
+    return this._imgDims.height;
+  }
+  get positions() {
+    return this._positions;
+  }
+  get relativePositions() {
+    return this._positions.map(
+      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight))
+    );
+  }
+  forSize(width, height) {
+    return new this.constructor(
+      this.relativePositions,
+      { width, height }
+    );
+  }
+  shiftBy(x, y) {
+    return new this.constructor(
+      this.relativePositions,
+      this._imgDims,
+      new Point(x, y)
+    );
+  }
+  shiftByPoint(pt) {
+    return this.shiftBy(pt.x, pt.y);
+  }
+  align(detection, options = {}) {
+    if (detection) {
+      const box = detection instanceof FaceDetection ? detection.box.floor() : new Box(detection);
+      return this.shiftBy(box.x, box.y).align(null, options);
+    }
+    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };
+    if (useDlibAlignment) {
+      return this.alignDlib();
+    }
+    return this.alignMinBbox(minBoxPadding);
+  }
+  alignDlib() {
+    const centers = this.getRefPointsForAlignment();
+    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;
+    const distToMouth = (pt) => mouthCenter.sub(pt).magnitude();
+    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;
+    const size = Math.floor(eyeToMouthDist / relScale);
+    const refPoint = getCenterPoint(centers);
+    const x = Math.floor(Math.max(0, refPoint.x - relX * size));
+    const y = Math.floor(Math.max(0, refPoint.y - relY * size));
+    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));
+  }
+  alignMinBbox(padding) {
+    const box = minBbox(this.positions);
+    return box.pad(box.width * padding, box.height * padding);
+  }
+  getRefPointsForAlignment() {
+    throw new Error("getRefPointsForAlignment not implemented by base class");
+  }
+};
+
+// src/classes/FaceLandmarks5.ts
+var FaceLandmarks5 = class extends FaceLandmarks {
+  getRefPointsForAlignment() {
+    const pts = this.positions;
+    return [
+      pts[0],
+      pts[1],
+      getCenterPoint([pts[3], pts[4]])
+    ];
+  }
+};
+
+// src/classes/FaceLandmarks68.ts
+var FaceLandmarks68 = class extends FaceLandmarks {
+  getJawOutline() {
+    return this.positions.slice(0, 17);
+  }
+  getLeftEyeBrow() {
+    return this.positions.slice(17, 22);
+  }
+  getRightEyeBrow() {
+    return this.positions.slice(22, 27);
+  }
+  getNose() {
+    return this.positions.slice(27, 36);
+  }
+  getLeftEye() {
+    return this.positions.slice(36, 42);
+  }
+  getRightEye() {
+    return this.positions.slice(42, 48);
+  }
+  getMouth() {
+    return this.positions.slice(48, 68);
+  }
+  getRefPointsForAlignment() {
+    return [
+      this.getLeftEye(),
+      this.getRightEye(),
+      this.getMouth()
+    ].map(getCenterPoint);
+  }
+};
+
+// src/classes/FaceMatch.ts
+var FaceMatch = class {
+  constructor(label, distance) {
+    this._label = label;
+    this._distance = distance;
+  }
+  get label() {
+    return this._label;
+  }
+  get distance() {
+    return this._distance;
+  }
+  toString(withDistance = true) {
+    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ""}`;
+  }
+};
+
+// src/classes/LabeledBox.ts
+var LabeledBox = class extends Box {
+  constructor(box, label) {
+    super(box);
+    this._label = label;
+  }
+  static assertIsValidLabeledBox(box, callee) {
+    Box.assertIsValidBox(box, callee);
+    if (!isValidNumber(box.label)) {
+      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);
+    }
+  }
+  get label() {
+    return this._label;
+  }
+};
+
+// src/classes/LabeledFaceDescriptors.ts
+var LabeledFaceDescriptors = class {
+  constructor(label, descriptors) {
+    if (!(typeof label === "string")) {
+      throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");
+    }
+    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {
+      throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");
+    }
+    this._label = label;
+    this._descriptors = descriptors;
+  }
+  get label() {
+    return this._label;
+  }
+  get descriptors() {
+    return this._descriptors;
+  }
+  toJSON() {
+    return {
+      label: this.label,
+      descriptors: this.descriptors.map((d) => Array.from(d))
+    };
+  }
+  static fromJSON(json) {
+    const descriptors = json.descriptors.map((d) => new Float32Array(d));
+    return new LabeledFaceDescriptors(json.label, descriptors);
+  }
+};
+
+// src/classes/PredictedBox.ts
+var PredictedBox = class extends LabeledBox {
+  constructor(box, label, score, classScore) {
+    super(box, label);
+    this._score = score;
+    this._classScore = classScore;
+  }
+  static assertIsValidPredictedBox(box, callee) {
+    LabeledBox.assertIsValidLabeledBox(box, callee);
+    if (!isValidProbablitiy(box.score) || !isValidProbablitiy(box.classScore)) {
+      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);
+    }
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+};
+
+// src/factories/WithFaceDetection.ts
+function isWithFaceDetection(obj) {
+  return obj.detection instanceof FaceDetection;
+}
+function extendWithFaceDetection(sourceObj, detection) {
+  const extension = { detection };
+  return { ...sourceObj, ...extension };
+}
+
+// src/env/createBrowserEnv.ts
+function createBrowserEnv() {
+  const fetch = window.fetch;
+  if (!fetch)
+    throw new Error("fetch - missing fetch implementation for browser environment");
+  const readFile = () => {
+    throw new Error("readFile - filesystem not available for browser environment");
+  };
+  return {
+    Canvas: HTMLCanvasElement,
+    CanvasRenderingContext2D,
+    Image: HTMLImageElement,
+    ImageData,
+    Video: HTMLVideoElement,
+    createCanvasElement: () => document.createElement("canvas"),
+    createImageElement: () => document.createElement("img"),
+    createVideoElement: () => document.createElement("video"),
+    fetch,
+    readFile
+  };
+}
+
+// src/env/isNodejs.ts
+function isNodejs() {
+  return typeof global === "object" && typeof process !== "undefined" && process.versions != null && process.versions.node != null;
+}
+
+// src/env/createFileSystem.ts
+function createFileSystem(fs) {
+  let requireFsError = "";
+  if (!fs && isNodejs()) {
+    try {
+      fs = __require("fs");
+    } catch (err) {
+      requireFsError = err.toString();
+    }
+  }
+  const readFile = fs ? (filePath) => new Promise((resolve, reject) => {
+    fs.readFile(filePath, (err, buffer) => err ? reject(err) : resolve(buffer));
+  }) : () => {
+    throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`);
+  };
+  return { readFile };
+}
+
+// src/env/createNodejsEnv.ts
+function createNodejsEnv() {
+  const Canvas = global["Canvas"] || global.HTMLCanvasElement;
+  const Image = global.Image || global.HTMLImageElement;
+  const Video = global["Video"] || global.HTMLVideoElement;
+  const createCanvasElement = () => {
+    if (Canvas)
+      return new Canvas();
+    throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment");
+  };
+  const createImageElement = () => {
+    if (Image)
+      return new Image();
+    throw new Error("createImageElement - missing Image implementation for nodejs environment");
+  };
+  const createVideoElement = () => {
+    if (Video)
+      return new Video();
+    throw new Error("createVideoElement - missing Video implementation for nodejs environment");
+  };
+  const fetch = global.fetch;
+  const fileSystem = createFileSystem();
+  return {
+    Canvas: Canvas || class {
+    },
+    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {
+    },
+    Image: Image || class {
+    },
+    ImageData: global.ImageData || class {
+    },
+    Video: global.HTMLVideoElement || class {
+    },
+    createCanvasElement,
+    createImageElement,
+    createVideoElement,
+    fetch,
+    ...fileSystem
+  };
+}
+
+// src/env/isBrowser.ts
+function isBrowser() {
+  return typeof window === "object" && typeof document !== "undefined" && typeof HTMLImageElement !== "undefined" && typeof HTMLCanvasElement !== "undefined" && typeof HTMLVideoElement !== "undefined" && typeof ImageData !== "undefined" && typeof CanvasRenderingContext2D !== "undefined";
+}
+
+// src/env/index.ts
+var environment;
+function getEnv() {
+  if (!environment) {
+    throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  return environment;
+}
+function setEnv(env2) {
+  environment = env2;
+}
+function initialize() {
+  if (isBrowser())
+    return setEnv(createBrowserEnv());
+  if (isNodejs())
+    return setEnv(createNodejsEnv());
+  return null;
+}
+function monkeyPatch(env2) {
+  if (!environment) {
+    initialize();
+  }
+  if (!environment) {
+    throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  const { Canvas = environment.Canvas, Image = environment.Image } = env2;
+  environment.Canvas = Canvas;
+  environment.Image = Image;
+  environment.createCanvasElement = env2.createCanvasElement || (() => new Canvas());
+  environment.createImageElement = env2.createImageElement || (() => new Image());
+  environment.ImageData = env2.ImageData || environment.ImageData;
+  environment.Video = env2.Video || environment.Video;
+  environment.fetch = env2.fetch || environment.fetch;
+  environment.readFile = env2.readFile || environment.readFile;
+}
+var env = {
+  getEnv,
+  setEnv,
+  initialize,
+  createBrowserEnv,
+  createFileSystem,
+  createNodejsEnv,
+  monkeyPatch,
+  isBrowser,
+  isNodejs
+};
+initialize();
+
+// src/dom/resolveInput.ts
+function resolveInput(arg) {
+  if (!env.isNodejs() && typeof arg === "string") {
+    return document.getElementById(arg);
+  }
+  return arg;
+}
+
+// src/dom/getContext2dOrThrow.ts
+function getContext2dOrThrow(canvasArg) {
+  const { Canvas, CanvasRenderingContext2D: CanvasRenderingContext2D2 } = env.getEnv();
+  if (canvasArg instanceof CanvasRenderingContext2D2) {
+    return canvasArg;
+  }
+  const canvas = resolveInput(canvasArg);
+  if (!(canvas instanceof Canvas)) {
+    throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+  }
+  const ctx = canvas.getContext("2d");
+  if (!ctx) {
+    throw new Error("resolveContext2d - canvas 2d context is null");
+  }
+  return ctx;
+}
+
+// src/draw/DrawTextField.ts
+var AnchorPosition = /* @__PURE__ */ ((AnchorPosition2) => {
+  AnchorPosition2["TOP_LEFT"] = "TOP_LEFT";
+  AnchorPosition2["TOP_RIGHT"] = "TOP_RIGHT";
+  AnchorPosition2["BOTTOM_LEFT"] = "BOTTOM_LEFT";
+  AnchorPosition2["BOTTOM_RIGHT"] = "BOTTOM_RIGHT";
+  return AnchorPosition2;
+})(AnchorPosition || {});
+var DrawTextFieldOptions = class {
+  constructor(options = {}) {
+    const {
+      anchorPosition,
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = options;
+    this.anchorPosition = anchorPosition || "TOP_LEFT" /* TOP_LEFT */;
+    this.backgroundColor = backgroundColor || "rgba(0, 0, 0, 0.5)";
+    this.fontColor = fontColor || "rgba(255, 255, 255, 1)";
+    this.fontSize = fontSize || 14;
+    this.fontStyle = fontStyle || "Georgia";
+    this.padding = padding || 4;
+  }
+};
+var DrawTextField = class {
+  constructor(text, anchor, options = {}) {
+    this.text = typeof text === "string" ? [text] : text instanceof DrawTextField ? text.text : text;
+    this.anchor = anchor;
+    this.options = new DrawTextFieldOptions(options);
+  }
+  measureWidth(ctx) {
+    const { padding } = this.options;
+    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => w0 < w1 ? w1 : w0, 0) + 2 * padding;
+  }
+  measureHeight() {
+    const { fontSize, padding } = this.options;
+    return this.text.length * fontSize + 2 * padding;
+  }
+  getUpperLeft(ctx, canvasDims) {
+    const { anchorPosition } = this.options;
+    const isShiftLeft = anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */ || anchorPosition === "TOP_RIGHT" /* TOP_RIGHT */;
+    const isShiftTop = anchorPosition === "BOTTOM_LEFT" /* BOTTOM_LEFT */ || anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */;
+    const textFieldWidth = this.measureWidth(ctx);
+    const textFieldHeight = this.measureHeight();
+    const x = isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x;
+    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;
+    if (canvasDims) {
+      const { width, height } = canvasDims;
+      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);
+      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);
+      return { x: newX, y: newY };
+    }
+    return { x, y };
+  }
+  draw(canvasArg) {
+    const canvas = resolveInput(canvasArg);
+    const ctx = getContext2dOrThrow(canvas);
+    const {
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = this.options;
+    ctx.font = `${fontSize}px ${fontStyle}`;
+    const maxTextWidth = this.measureWidth(ctx);
+    const textHeight = this.measureHeight();
+    ctx.fillStyle = backgroundColor;
+    const upperLeft = this.getUpperLeft(ctx, canvas);
+    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);
+    ctx.fillStyle = fontColor;
+    this.text.forEach((textLine, i) => {
+      const x = padding + upperLeft.x;
+      const y = padding + upperLeft.y + (i + 1) * fontSize;
+      ctx.fillText(textLine, x, y);
+    });
+  }
+};
+
+// src/draw/DrawBox.ts
+var DrawBoxOptions = class {
+  constructor(options = {}) {
+    const {
+      boxColor,
+      lineWidth,
+      label,
+      drawLabelOptions
+    } = options;
+    this.boxColor = boxColor || "rgba(0, 0, 255, 1)";
+    this.lineWidth = lineWidth || 2;
+    this.label = label;
+    const defaultDrawLabelOptions = {
+      anchorPosition: "BOTTOM_LEFT" /* BOTTOM_LEFT */,
+      backgroundColor: this.boxColor
+    };
+    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });
+  }
+};
+var DrawBox = class {
+  constructor(box, options = {}) {
+    this.box = new Box(box);
+    this.options = new DrawBoxOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const { boxColor, lineWidth } = this.options;
+    const {
+      x,
+      y,
+      width,
+      height
+    } = this.box;
+    ctx.strokeStyle = boxColor;
+    ctx.lineWidth = lineWidth;
+    ctx.strokeRect(x, y, width, height);
+    const { label } = this.options;
+    if (label) {
+      new DrawTextField([label], { x: x - lineWidth / 2, y }, this.options.drawLabelOptions).draw(canvasArg);
+    }
+  }
+};
+
+// src/draw/drawDetections.ts
+function drawDetections(canvasArg, detections) {
+  const detectionsArray = Array.isArray(detections) ? detections : [detections];
+  detectionsArray.forEach((det) => {
+    const score = det instanceof FaceDetection ? det.score : isWithFaceDetection(det) ? det.detection.score : void 0;
+    const box = det instanceof FaceDetection ? det.box : isWithFaceDetection(det) ? det.detection.box : new Box(det);
+    const label = score ? `${round(score)}` : void 0;
+    new DrawBox(box, { label }).draw(canvasArg);
+  });
+}
+
+// src/dom/isMediaLoaded.ts
+function isMediaLoaded(media) {
+  const { Image, Video } = env.getEnv();
+  return media instanceof Image && media.complete || media instanceof Video && media.readyState >= 3;
+}
+
+// src/dom/awaitMediaLoaded.ts
+function awaitMediaLoaded(media) {
+  return new Promise((resolve, reject) => {
+    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media))
+      resolve(null);
+    function onError(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      reject(e);
+    }
+    function onLoad(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      resolve(e);
+    }
+    media.addEventListener("load", onLoad);
+    media.addEventListener("error", onError);
+  });
+}
+
+// src/dom/bufferToImage.ts
+function bufferToImage(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToImage - expected buf to be of type: Blob"));
+    const reader = new FileReader();
+    reader.onload = () => {
+      if (typeof reader.result !== "string")
+        reject(new Error("bufferToImage - expected reader.result to be a string, in onload"));
+      const img = env.getEnv().createImageElement();
+      img.onload = () => resolve(img);
+      img.onerror = reject;
+      img.src = reader.result;
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(buf);
+  });
+}
+
+// src/dom/getMediaDimensions.ts
+function getMediaDimensions(input) {
+  const { Image, Video } = env.getEnv();
+  if (input instanceof Image) {
+    return new Dimensions(input.naturalWidth, input.naturalHeight);
+  }
+  if (input instanceof Video) {
+    return new Dimensions(input.videoWidth, input.videoHeight);
+  }
+  return new Dimensions(input.width, input.height);
+}
+
+// src/dom/createCanvas.ts
+function createCanvas({ width, height }) {
+  const { createCanvasElement } = env.getEnv();
+  const canvas = createCanvasElement();
+  canvas.width = width;
+  canvas.height = height;
+  return canvas;
+}
+function createCanvasFromMedia(media, dims) {
+  const { ImageData: ImageData2 } = env.getEnv();
+  if (!(media instanceof ImageData2) && !isMediaLoaded(media)) {
+    throw new Error("createCanvasFromMedia - media has not finished loading yet");
+  }
+  const { width, height } = dims || getMediaDimensions(media);
+  const canvas = createCanvas({ width, height });
+  if (media instanceof ImageData2) {
+    getContext2dOrThrow(canvas).putImageData(media, 0, 0);
+  } else {
+    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);
+  }
+  return canvas;
+}
+
+// src/dom/imageTensorToCanvas.ts
+async function imageTensorToCanvas(imgTensor, canvas) {
+  const targetCanvas = canvas || env.getEnv().createCanvasElement();
+  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);
+  const imgTensor3D = tfjs_esm_exports.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());
+  await tfjs_esm_exports["browser"].toPixels(imgTensor3D, targetCanvas);
+  imgTensor3D.dispose();
+  return targetCanvas;
+}
+
+// src/dom/isMediaElement.ts
+function isMediaElement(input) {
+  const { Image, Canvas, Video } = env.getEnv();
+  return input instanceof Image || input instanceof Canvas || input instanceof Video;
+}
+
+// src/dom/imageToSquare.ts
+function imageToSquare(input, inputSize, centerImage = false) {
+  const { Image, Canvas } = env.getEnv();
+  if (!(input instanceof Image || input instanceof Canvas)) {
+    throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+  }
+  if (inputSize <= 0)
+    return createCanvas({ width: 1, height: 1 });
+  const dims = getMediaDimensions(input);
+  const scale2 = inputSize / Math.max(dims.height, dims.width);
+  const width = scale2 * dims.width;
+  const height = scale2 * dims.height;
+  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });
+  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);
+  const offset = Math.abs(width - height) / 2;
+  const dx = centerImage && width < height ? offset : 0;
+  const dy = centerImage && height < width ? offset : 0;
+  if (inputCanvas.width > 0 && inputCanvas.height > 0)
+    getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);
+  return targetCanvas;
+}
+
+// src/dom/NetInput.ts
+var NetInput = class {
+  constructor(inputs, treatAsBatchInput = false) {
+    this._imageTensors = [];
+    this._canvases = [];
+    this._treatAsBatchInput = false;
+    this._inputDimensions = [];
+    this._inputSize = 0;
+    if (!Array.isArray(inputs)) {
+      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);
+    }
+    this._treatAsBatchInput = treatAsBatchInput;
+    this._batchSize = inputs.length;
+    inputs.forEach((input, idx) => {
+      if (isTensor3D(input)) {
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape;
+        return;
+      }
+      if (isTensor4D(input)) {
+        const batchSize = input.shape[0];
+        if (batchSize !== 1) {
+          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+        }
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape.slice(1);
+        return;
+      }
+      const canvas = input instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);
+      this._canvases[idx] = canvas;
+      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];
+    });
+  }
+  get imageTensors() {
+    return this._imageTensors;
+  }
+  get canvases() {
+    return this._canvases;
+  }
+  get isBatchInput() {
+    return this.batchSize > 1 || this._treatAsBatchInput;
+  }
+  get batchSize() {
+    return this._batchSize;
+  }
+  get inputDimensions() {
+    return this._inputDimensions;
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get reshapedInputDimensions() {
+    return range(this.batchSize, 0, 1).map(
+      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx)
+    );
+  }
+  getInput(batchIdx) {
+    return this.canvases[batchIdx] || this.imageTensors[batchIdx];
+  }
+  getInputDimensions(batchIdx) {
+    return this._inputDimensions[batchIdx];
+  }
+  getInputHeight(batchIdx) {
+    return this._inputDimensions[batchIdx][0];
+  }
+  getInputWidth(batchIdx) {
+    return this._inputDimensions[batchIdx][1];
+  }
+  getReshapedInputDimensions(batchIdx) {
+    if (typeof this.inputSize !== "number") {
+      throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+    }
+    const width = this.getInputWidth(batchIdx);
+    const height = this.getInputHeight(batchIdx);
+    return computeReshapedDimensions({ width, height }, this.inputSize);
+  }
+  toBatchTensor(inputSize, isCenterInputs = true) {
+    this._inputSize = inputSize;
+    return tfjs_esm_exports.tidy(() => {
+      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {
+        const input = this.getInput(batchIdx);
+        if (input instanceof tfjs_esm_exports.Tensor) {
+          let imgTensor = isTensor4D(input) ? input : tfjs_esm_exports.expandDims(input);
+          imgTensor = padToSquare(imgTensor, isCenterInputs);
+          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {
+            imgTensor = tfjs_esm_exports["image"].resizeBilinear(imgTensor, [inputSize, inputSize], false, false);
+          }
+          return imgTensor.as3D(inputSize, inputSize, 3);
+        }
+        if (input instanceof env.getEnv().Canvas) {
+          return tfjs_esm_exports["browser"].fromPixels(imageToSquare(input, inputSize, isCenterInputs));
+        }
+        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);
+      });
+      const batchTensor = tfjs_esm_exports.stack(inputTensors.map((t) => tfjs_esm_exports.cast(t, "float32"))).as4D(this.batchSize, inputSize, inputSize, 3);
+      return batchTensor;
+    });
+  }
+};
+
+// src/dom/toNetInput.ts
+async function toNetInput(inputs) {
+  if (inputs instanceof NetInput)
+    return inputs;
+  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];
+  if (!inputArgArray.length)
+    throw new Error("toNetInput - empty array passed as input");
+  const getIdxHint = (idx) => Array.isArray(inputs) ? ` at input index ${idx}:` : "";
+  const inputArray = inputArgArray.map(resolveInput);
+  inputArray.forEach((input, i) => {
+    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {
+      if (typeof inputArgArray[i] === "string")
+        throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);
+      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);
+    }
+    if (isTensor4D(input)) {
+      const batchSize = input.shape[0];
+      if (batchSize !== 1)
+        throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+    }
+  });
+  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));
+  return new NetInput(inputArray, Array.isArray(inputs));
+}
+
+// src/dom/extractFaces.ts
+async function extractFaces(input, detections) {
+  const { Canvas } = env.getEnv();
+  let canvas = input;
+  if (!(input instanceof Canvas)) {
+    const netInput = await toNetInput(input);
+    if (netInput.batchSize > 1)
+      throw new Error("extractFaces - batchSize > 1 not supported");
+    const tensorOrCanvas = netInput.getInput(0);
+    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);
+  }
+  const ctx = getContext2dOrThrow(canvas);
+  const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det).map((box) => box.clipAtImageBorders(canvas.width, canvas.height));
+  return boxes.map(({ x, y, width, height }) => {
+    const faceImg = createCanvas({ width, height });
+    if (width > 0 && height > 0)
+      getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);
+    return faceImg;
+  });
+}
+
+// src/dom/extractFaceTensors.ts
+async function extractFaceTensors(imageTensor, detections) {
+  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {
+    throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+  }
+  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {
+    throw new Error("extractFaceTensors - batchSize > 1 not supported");
+  }
+  return tfjs_esm_exports.tidy(() => {
+    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);
+    const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det).map((box) => box.clipAtImageBorders(imgWidth, imgHeight));
+    const faceTensors = boxes.filter((box) => box.width > 0 && box.height > 0).map(({ x, y, width, height }) => tfjs_esm_exports.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));
+    return faceTensors;
+  });
+}
+
+// src/dom/fetchOrThrow.ts
+async function fetchOrThrow(url, init) {
+  const { fetch } = env.getEnv();
+  const res = await fetch(url, init);
+  if (!(res.status < 400)) {
+    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);
+  }
+  return res;
+}
+
+// src/dom/fetchImage.ts
+async function fetchImage(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("image/")) {
+    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToImage(blob);
+}
+
+// src/dom/fetchJson.ts
+async function fetchJson(uri) {
+  return (await fetchOrThrow(uri)).json();
+}
+
+// src/dom/fetchNetWeights.ts
+async function fetchNetWeights(uri) {
+  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());
+}
+
+// src/dom/bufferToVideo.ts
+function bufferToVideo(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToVideo - expected buf to be of type: Blob"));
+    const video = env.getEnv().createVideoElement();
+    video.oncanplay = () => resolve(video);
+    video.onerror = reject;
+    video.playsInline = true;
+    video.muted = true;
+    video.src = URL.createObjectURL(buf);
+    video.play();
+  });
+}
+
+// src/dom/fetchVideo.ts
+async function fetchVideo(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("video/")) {
+    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToVideo(blob);
+}
+
+// src/common/getModelUris.ts
+function getModelUris(uri, defaultModelName) {
+  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;
+  if (!uri) {
+    return {
+      modelBaseUri: "",
+      manifestUri: defaultManifestFilename
+    };
+  }
+  if (uri === "/") {
+    return {
+      modelBaseUri: "/",
+      manifestUri: `/${defaultManifestFilename}`
+    };
+  }
+  const protocol = uri.startsWith("http://") ? "http://" : uri.startsWith("https://") ? "https://" : "";
+  uri = uri.replace(protocol, "");
+  const parts = uri.split("/").filter((s) => s);
+  const manifestFile = uri.endsWith(".json") ? parts[parts.length - 1] : defaultManifestFilename;
+  let modelBaseUri = protocol + (uri.endsWith(".json") ? parts.slice(0, parts.length - 1) : parts).join("/");
+  modelBaseUri = uri.startsWith("/") ? `/${modelBaseUri}` : modelBaseUri;
+  return {
+    modelBaseUri,
+    manifestUri: modelBaseUri === "/" ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`
+  };
+}
+
+// src/dom/loadWeightMap.ts
+async function loadWeightMap(uri, defaultModelName) {
+  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);
+  const manifest = await fetchJson(manifestUri);
+  return tfjs_esm_exports["io"].loadWeights(manifest, modelBaseUri);
+}
+
+// src/dom/matchDimensions.ts
+function matchDimensions(input, reference, useMediaDimensions = false) {
+  const { width, height } = useMediaDimensions ? getMediaDimensions(reference) : reference;
+  input.width = width;
+  input.height = height;
+  return { width, height };
+}
+
+// src/NeuralNetwork.ts
+var NeuralNetwork = class {
+  constructor(name) {
+    this._params = void 0;
+    this._paramMappings = [];
+    this._name = name;
+  }
+  get params() {
+    return this._params;
+  }
+  get paramMappings() {
+    return this._paramMappings;
+  }
+  get isLoaded() {
+    return !!this.params;
+  }
+  getParamFromPath(paramPath) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    return obj[objProp];
+  }
+  reassignParamFromPath(paramPath, tensor2) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    obj[objProp].dispose();
+    obj[objProp] = tensor2;
+  }
+  getParamList() {
+    return this._paramMappings.map(({ paramPath }) => ({
+      path: paramPath,
+      tensor: this.getParamFromPath(paramPath)
+    }));
+  }
+  getTrainableParams() {
+    return this.getParamList().filter((param) => param.tensor instanceof tfjs_esm_exports.Variable);
+  }
+  getFrozenParams() {
+    return this.getParamList().filter((param) => !(param.tensor instanceof tfjs_esm_exports.Variable));
+  }
+  variable() {
+    this.getFrozenParams().forEach(({ path, tensor: tensor2 }) => {
+      this.reassignParamFromPath(path, tensor2.variable());
+    });
+  }
+  freeze() {
+    this.getTrainableParams().forEach(({ path, tensor: variable }) => {
+      const tensor2 = tfjs_esm_exports.tensor(variable.dataSync());
+      variable.dispose();
+      this.reassignParamFromPath(path, tensor2);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.getParamList().forEach((param) => {
+      if (throwOnRedispose && param.tensor.isDisposed) {
+        throw new Error(`param tensor has already been disposed for path ${param.path}`);
+      }
+      param.tensor.dispose();
+    });
+    this._params = void 0;
+  }
+  serializeParams() {
+    return new Float32Array(
+      this.getParamList().map(({ tensor: tensor2 }) => Array.from(tensor2.dataSync())).reduce((flat, arr) => flat.concat(arr))
+    );
+  }
+  async load(weightsOrUrl) {
+    if (weightsOrUrl instanceof Float32Array) {
+      this.extractWeights(weightsOrUrl);
+      return;
+    }
+    await this.loadFromUri(weightsOrUrl);
+  }
+  async loadFromUri(uri) {
+    if (uri && typeof uri !== "string") {
+      throw new Error(`${this._name}.loadFromUri - expected model uri`);
+    }
+    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());
+    this.loadFromWeightMap(weightMap);
+  }
+  async loadFromDisk(filePath) {
+    if (filePath && typeof filePath !== "string") {
+      throw new Error(`${this._name}.loadFromDisk - expected model file path`);
+    }
+    const { readFile } = env.getEnv();
+    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());
+    const fetchWeightsFromDisk = (filePaths) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));
+    const loadWeights = tfjs_esm_exports["io"].weightsLoaderFactory(fetchWeightsFromDisk);
+    const manifest = JSON.parse((await readFile(manifestUri)).toString());
+    const weightMap = await loadWeights(manifest, modelBaseUri);
+    this.loadFromWeightMap(weightMap);
+  }
+  loadFromWeightMap(weightMap) {
+    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  extractWeights(weights) {
+    const { paramMappings, params } = this.extractParams(weights);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  traversePropertyPath(paramPath) {
+    if (!this.params) {
+      throw new Error("traversePropertyPath - model has no loaded params");
+    }
+    const result = paramPath.split("/").reduce((res, objProp2) => {
+      if (!res.nextObj.hasOwnProperty(objProp2)) {
+        throw new Error(`traversePropertyPath - object does not have property ${objProp2}, for path ${paramPath}`);
+      }
+      return { obj: res.nextObj, objProp: objProp2, nextObj: res.nextObj[objProp2] };
+    }, { nextObj: this.params });
+    const { obj, objProp } = result;
+    if (!obj || !objProp || !(obj[objProp] instanceof tfjs_esm_exports.Tensor)) {
+      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);
+    }
+    return { obj, objProp };
+  }
+};
+
+// src/common/depthwiseSeparableConv.ts
+function depthwiseSeparableConv(x, params, stride) {
+  return tfjs_esm_exports.tidy(() => {
+    let out = tfjs_esm_exports.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, "same");
+    out = tfjs_esm_exports.add(out, params.bias);
+    return out;
+  });
+}
+
+// src/faceFeatureExtractor/denseBlock.ts
+function denseBlock3(x, denseBlockParams, isFirstLayer = false) {
+  return tfjs_esm_exports.tidy(() => {
+    const out1 = tfjs_esm_exports.relu(
+      isFirstLayer ? tfjs_esm_exports.add(
+        tfjs_esm_exports.conv2d(x, denseBlockParams.conv0.filters, [2, 2], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, [2, 2])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tfjs_esm_exports.relu(tfjs_esm_exports.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    return tfjs_esm_exports.relu(tfjs_esm_exports.add(out1, tfjs_esm_exports.add(out2, out3)));
+  });
+}
+function denseBlock4(x, denseBlockParams, isFirstLayer = false, isScaleDown = true) {
+  return tfjs_esm_exports.tidy(() => {
+    const out1 = tfjs_esm_exports.relu(
+      isFirstLayer ? tfjs_esm_exports.add(
+        tfjs_esm_exports.conv2d(x, denseBlockParams.conv0.filters, isScaleDown ? [2, 2] : [1, 1], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, isScaleDown ? [2, 2] : [1, 1])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tfjs_esm_exports.relu(tfjs_esm_exports.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    const in4 = tfjs_esm_exports.relu(tfjs_esm_exports.add(out1, tfjs_esm_exports.add(out2, out3)));
+    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);
+    return tfjs_esm_exports.relu(tfjs_esm_exports.add(out1, tfjs_esm_exports.add(out2, tfjs_esm_exports.add(out3, out4))));
+  });
+}
+
+// src/common/convLayer.ts
+function convLayer(x, params, padding = "same", withRelu = false) {
+  return tfjs_esm_exports.tidy(() => {
+    const out = tfjs_esm_exports.add(
+      tfjs_esm_exports.conv2d(x, params.filters, [1, 1], padding),
+      params.bias
+    );
+    return withRelu ? tfjs_esm_exports.relu(out) : out;
+  });
+}
+
+// src/common/disposeUnusedWeightTensors.ts
+function disposeUnusedWeightTensors(weightMap, paramMappings) {
+  Object.keys(weightMap).forEach((path) => {
+    if (!paramMappings.some((pm) => pm.originalPath === path)) {
+      weightMap[path].dispose();
+    }
+  });
+}
+
+// src/common/extractConvParamsFactory.ts
+function extractConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, filterSize, mappedPrefix) => {
+    const filters = tfjs_esm_exports.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tfjs_esm_exports.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  };
+}
+
+// src/common/extractFCParamsFactory.ts
+function extractFCParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const fc_weights = tfjs_esm_exports.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);
+    const fc_bias = tfjs_esm_exports.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return {
+      weights: fc_weights,
+      bias: fc_bias
+    };
+  };
+}
+
+// src/common/types.ts
+var SeparableConvParams = class {
+  constructor(depthwise_filter, pointwise_filter, bias) {
+    this.depthwise_filter = depthwise_filter;
+    this.pointwise_filter = pointwise_filter;
+    this.bias = bias;
+  }
+};
+
+// src/common/extractSeparableConvParamsFactory.ts
+function extractSeparableConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const depthwise_filter = tfjs_esm_exports.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);
+    const pointwise_filter = tfjs_esm_exports.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);
+    const bias = tfjs_esm_exports.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/depthwise_filter` },
+      { paramPath: `${mappedPrefix}/pointwise_filter` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+function loadSeparableConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const depthwise_filter = extractWeightEntry(`${prefix}/depthwise_filter`, 4);
+    const pointwise_filter = extractWeightEntry(`${prefix}/pointwise_filter`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+
+// src/common/extractWeightEntryFactory.ts
+function extractWeightEntryFactory(weightMap, paramMappings) {
+  return (originalPath, paramRank, mappedPath) => {
+    const tensor2 = weightMap[originalPath];
+    if (!isTensor(tensor2, paramRank)) {
+      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor2}`);
+    }
+    paramMappings.push(
+      { originalPath, paramPath: mappedPath || originalPath }
+    );
+    return tensor2;
+  };
+}
+
+// src/common/extractWeightsFactory.ts
+function extractWeightsFactory(weights) {
+  let remainingWeights = weights;
+  function extractWeights(numWeights) {
+    const ret = remainingWeights.slice(0, numWeights);
+    remainingWeights = remainingWeights.slice(numWeights);
+    return ret;
+  }
+  function getRemainingWeights() {
+    return remainingWeights;
+  }
+  return {
+    extractWeights,
+    getRemainingWeights
+  };
+}
+
+// src/faceFeatureExtractor/extractorsFactory.ts
+function extractorsFactory(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`) : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);
+    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);
+    const conv22 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const { conv0, conv1, conv2: conv22 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);
+    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParams.ts
+function extractParams(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock4Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock4Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock4Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock4Params(64, 128, "dense2");
+  const dense3 = extractDenseBlock4Params(128, 256, "dense3");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: {
+      dense0,
+      dense1,
+      dense2,
+      dense3
+    }
+  };
+}
+
+// src/common/loadConvParamsFactory.ts
+function loadConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  };
+}
+
+// src/faceFeatureExtractor/loadParamsFactory.ts
+function loadParamsFactory(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractDenseBlock3Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock4Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock4Params("dense0", true),
+    dense1: extractDenseBlock4Params("dense1"),
+    dense2: extractDenseBlock4Params("dense2"),
+    dense3: extractDenseBlock4Params("dense3")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var FaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("FaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceFeatureExtractor - load model before inference");
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock4(normalized, params.dense0, true);
+      out = denseBlock4(out, params.dense1);
+      out = denseBlock4(out, params.dense2);
+      out = denseBlock4(out, params.dense3);
+      out = tfjs_esm_exports.avgPool(out, [7, 7], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams(weights);
+  }
+};
+
+// src/common/fullyConnectedLayer.ts
+function fullyConnectedLayer(x, params) {
+  return tfjs_esm_exports.tidy(() => tfjs_esm_exports.add(
+    tfjs_esm_exports.matMul(x, params.weights),
+    params.bias
+  ));
+}
+
+// src/faceProcessor/extractParams.ts
+function extractParams2(weights, channelsIn, channelsOut) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const fc = extractFCParams(channelsIn, channelsOut, "fc");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc }
+  };
+}
+
+// src/faceProcessor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap2(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: extractFcParams("fc")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceProcessor/util.ts
+function seperateWeightMaps(weightMap) {
+  const featureExtractorMap = {};
+  const classifierMap = {};
+  Object.keys(weightMap).forEach((key) => {
+    const map = key.startsWith("fc") ? classifierMap : featureExtractorMap;
+    map[key] = weightMap[key];
+  });
+  return { featureExtractorMap, classifierMap };
+}
+
+// src/faceProcessor/FaceProcessor.ts
+var FaceProcessor = class extends NeuralNetwork {
+  constructor(_name, faceFeatureExtractor) {
+    super(_name);
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams2(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap2(classifierMap);
+  }
+  extractParams(weights) {
+    const cIn = this.getClassifierChannelsIn();
+    const cOut = this.getClassifierChannelsOut();
+    const classifierWeightSize = cOut * cIn + cOut;
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressions.ts
+var FACE_EXPRESSION_LABELS = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"];
+var FaceExpressions = class {
+  constructor(probabilities) {
+    this.neutral = 0;
+    this.happy = 0;
+    this.sad = 0;
+    this.angry = 0;
+    this.fearful = 0;
+    this.disgusted = 0;
+    this.surprised = 0;
+    if (probabilities.length !== 7) {
+      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);
+    }
+    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {
+      this[expression] = probabilities[idx];
+    });
+  }
+  asSortedArray() {
+    return FACE_EXPRESSION_LABELS.map((expression) => ({ expression, probability: this[expression] })).sort((e0, e1) => e1.probability - e0.probability);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var FaceExpressionNet = class extends FaceProcessor {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceExpressionNet", faceFeatureExtractor);
+  }
+  forwardInput(input) {
+    return tfjs_esm_exports.tidy(() => tfjs_esm_exports.softmax(this.runNet(input)));
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictExpressions(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const probabilitesByBatch = await Promise.all(tfjs_esm_exports.unstack(out).map(async (t) => {
+      const data = t.dataSync();
+      t.dispose();
+      return data;
+    }));
+    out.dispose();
+    const predictionsByBatch = probabilitesByBatch.map((probabilites) => new FaceExpressions(probabilites));
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_expression_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+  getClassifierChannelsOut() {
+    return 7;
+  }
+};
+
+// src/factories/WithFaceExpressions.ts
+function isWithFaceExpressions(obj) {
+  return obj.expressions instanceof FaceExpressions;
+}
+function extendWithFaceExpressions(sourceObj, expressions) {
+  const extension = { expressions };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/drawFaceExpressions.ts
+function drawFaceExpressions(canvasArg, faceExpressions, minConfidence = 0.1, textFieldAnchor) {
+  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];
+  faceExpressionsArray.forEach((e) => {
+    const expr = e instanceof FaceExpressions ? e : isWithFaceExpressions(e) ? e.expressions : void 0;
+    if (!expr) {
+      throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+    }
+    const sorted = expr.asSortedArray();
+    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);
+    const anchor = isWithFaceDetection(e) ? e.detection.box.bottomLeft : textFieldAnchor || new Point(0, 0);
+    const drawTextField = new DrawTextField(
+      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),
+      anchor
+    );
+    drawTextField.draw(canvasArg);
+  });
+}
+
+// src/factories/WithFaceLandmarks.ts
+function isWithFaceLandmarks(obj) {
+  return isWithFaceDetection(obj) && obj["landmarks"] instanceof FaceLandmarks && obj["unshiftedLandmarks"] instanceof FaceLandmarks && obj["alignedRect"] instanceof FaceDetection;
+}
+function calculateFaceAngle(mesh) {
+  const degrees = (radians) => radians * 180 / Math.PI;
+  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);
+  const angle = {
+    roll: void 0,
+    pitch: void 0,
+    yaw: void 0
+  };
+  const calcYaw = (leftPoint, midPoint, rightPoint) => {
+    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);
+    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);
+    return leftToMidpoint - rightToMidpoint;
+  };
+  const calcRoll = (lever, pivot) => {
+    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);
+    const opposite = pivot._y - lever._y;
+    const angleInRadians = Math.asin(opposite / hypotenuse);
+    const angleInDegrees = degrees(angleInRadians);
+    const normalizeAngle = Math.floor(90 - angleInDegrees);
+    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;
+    const result = normalizeAngle * tiltDirection;
+    return result;
+  };
+  const calcPitch = (leftPoint, midPoint, rightPoint) => {
+    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);
+    const baseCoords = {
+      _x: (leftPoint._x + rightPoint._x) / 2,
+      _y: (leftPoint._y + rightPoint._y) / 2
+    };
+    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);
+    const angleInRadians = Math.atan(midToBaseLength / base);
+    const angleInDegrees = Math.floor(degrees(angleInRadians));
+    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;
+    const result = angleInDegrees * direction;
+    return result;
+  };
+  if (!mesh || !mesh._positions || mesh._positions.length !== 68)
+    return angle;
+  const pt = mesh._positions;
+  angle.roll = calcRoll(pt[27], pt[66]);
+  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);
+  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);
+  return angle;
+}
+function extendWithFaceLandmarks(sourceObj, unshiftedLandmarks) {
+  const { box: shift } = sourceObj.detection;
+  const landmarks = unshiftedLandmarks.shiftBy(shift.x, shift.y);
+  const rect = landmarks.align();
+  const { imageDims } = sourceObj.detection;
+  const alignedRect = new FaceDetection(
+    sourceObj.detection.score,
+    rect.rescale(imageDims.reverse()),
+    imageDims
+  );
+  const angle = calculateFaceAngle(unshiftedLandmarks);
+  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/DrawFaceLandmarks.ts
+var DrawFaceLandmarksOptions = class {
+  constructor(options = {}) {
+    const {
+      drawLines = true,
+      drawPoints = true,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = options;
+    this.drawLines = drawLines;
+    this.drawPoints = drawPoints;
+    this.lineWidth = lineWidth || 1;
+    this.pointSize = pointSize || 2;
+    this.lineColor = lineColor || "rgba(0, 255, 255, 1)";
+    this.pointColor = pointColor || "rgba(255, 0, 255, 1)";
+  }
+};
+var DrawFaceLandmarks = class {
+  constructor(faceLandmarks, options = {}) {
+    this.faceLandmarks = faceLandmarks;
+    this.options = new DrawFaceLandmarksOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const {
+      drawLines,
+      drawPoints,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = this.options;
+    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {
+      ctx.strokeStyle = lineColor;
+      ctx.lineWidth = lineWidth;
+      drawContour(ctx, this.faceLandmarks.getJawOutline());
+      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getNose());
+      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);
+      drawContour(ctx, this.faceLandmarks.getRightEye(), true);
+      drawContour(ctx, this.faceLandmarks.getMouth(), true);
+    }
+    if (drawPoints) {
+      ctx.strokeStyle = pointColor;
+      ctx.fillStyle = pointColor;
+      const drawPoint = (pt) => {
+        ctx.beginPath();
+        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);
+        ctx.fill();
+      };
+      this.faceLandmarks.positions.forEach(drawPoint);
+    }
+  }
+};
+function drawFaceLandmarks(canvasArg, faceLandmarks) {
+  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];
+  faceLandmarksArray.forEach((f) => {
+    const landmarks = f instanceof FaceLandmarks ? f : isWithFaceLandmarks(f) ? f.landmarks : void 0;
+    if (!landmarks) {
+      throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+    }
+    new DrawFaceLandmarks(landmarks).draw(canvasArg);
+  });
+}
+
+// package.json
+var version7 = "1.7.6";
+
+// src/xception/extractParams.ts
+function extractorsFactory2(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractReductionBlockParams(channelsIn, channelsOut, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(channels, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParams3(weights, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = extractorsFactory2(extractWeights, paramMappings);
+  const entry_flow_conv_in = extractConvParams(3, 32, 3, "entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, "entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, "entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, "exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, "exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { entry_flow, middle_flow, exit_flow }
+  };
+}
+
+// src/xception/extractParamsFromWeightMap.ts
+function loadParamsFactory2(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractReductionBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParamsFromWeightMap3(weightMap, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = loadParamsFactory2(weightMap, paramMappings);
+  const entry_flow_conv_in = extractConvParams("entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams("entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams("entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams("exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams("exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };
+}
+
+// src/xception/TinyXception.ts
+function conv(x, params, stride) {
+  return tfjs_esm_exports.add(tfjs_esm_exports.conv2d(x, params.filters, stride, "same"), params.bias);
+}
+function reductionBlock(x, params, isActivateInput = true) {
+  let out = isActivateInput ? tfjs_esm_exports.relu(x) : x;
+  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tfjs_esm_exports.relu(out), params.separable_conv1, [1, 1]);
+  out = tfjs_esm_exports.maxPool(out, [3, 3], [2, 2], "same");
+  out = tfjs_esm_exports.add(out, conv(x, params.expansion_conv, [2, 2]));
+  return out;
+}
+function mainBlock(x, params) {
+  let out = depthwiseSeparableConv(tfjs_esm_exports.relu(x), params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tfjs_esm_exports.relu(out), params.separable_conv1, [1, 1]);
+  out = depthwiseSeparableConv(tfjs_esm_exports.relu(out), params.separable_conv2, [1, 1]);
+  out = tfjs_esm_exports.add(out, x);
+  return out;
+}
+var TinyXception = class extends NeuralNetwork {
+  constructor(numMainBlocks) {
+    super("TinyXception");
+    this._numMainBlocks = numMainBlocks;
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyXception - load model before inference");
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = tfjs_esm_exports.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));
+      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);
+      out = reductionBlock(out, params.entry_flow.reduction_block_1);
+      range(this._numMainBlocks, 0, 1).forEach((idx) => {
+        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);
+      });
+      out = reductionBlock(out, params.exit_flow.reduction_block);
+      out = tfjs_esm_exports.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "tiny_xception_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap3(weightMap, this._numMainBlocks);
+  }
+  extractParams(weights) {
+    return extractParams3(weights, this._numMainBlocks);
+  }
+};
+
+// src/ageGenderNet/extractParams.ts
+function extractParams4(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const age = extractFCParams(512, 1, "fc/age");
+  const gender = extractFCParams(512, 2, "fc/gender");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc: { age, gender } }
+  };
+}
+
+// src/ageGenderNet/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap4(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: {
+      age: extractFcParams("fc/age"),
+      gender: extractFcParams("fc/gender")
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ageGenderNet/types.ts
+var Gender = /* @__PURE__ */ ((Gender2) => {
+  Gender2["FEMALE"] = "female";
+  Gender2["MALE"] = "male";
+  return Gender2;
+})(Gender || {});
+
+// src/ageGenderNet/AgeGenderNet.ts
+var AgeGenderNet = class extends NeuralNetwork {
+  constructor(faceFeatureExtractor = new TinyXception(2)) {
+    super("AgeGenderNet");
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      const pooled = tfjs_esm_exports.avgPool(bottleneckFeatures, [7, 7], [2, 2], "valid").as2D(bottleneckFeatures.shape[0], -1);
+      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();
+      const gender = fullyConnectedLayer(pooled, params.fc.gender);
+      return { age, gender };
+    });
+  }
+  forwardInput(input) {
+    return tfjs_esm_exports.tidy(() => {
+      const { age, gender } = this.runNet(input);
+      return { age, gender: tfjs_esm_exports.softmax(gender) };
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictAgeAndGender(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const ages = tfjs_esm_exports.unstack(out.age);
+    const genders = tfjs_esm_exports.unstack(out.gender);
+    const ageAndGenderTensors = ages.map((ageTensor, i) => ({
+      ageTensor,
+      genderTensor: genders[i]
+    }));
+    const predictionsByBatch = await Promise.all(
+      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {
+        const age = ageTensor.dataSync()[0];
+        const probMale = genderTensor.dataSync()[0];
+        const isMale = probMale > 0.5;
+        const gender = isMale ? "male" /* MALE */ : "female" /* FEMALE */;
+        const genderProbability = isMale ? probMale : 1 - probMale;
+        ageTensor.dispose();
+        genderTensor.dispose();
+        return { age, gender, genderProbability };
+      })
+    );
+    out.age.dispose();
+    out.gender.dispose();
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "age_gender_model";
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams4(weights);
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap4(classifierMap);
+  }
+  extractParams(weights) {
+    const classifierWeightSize = 512 * 1 + 1 + (512 * 2 + 2);
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68NetBase.ts
+var FaceLandmark68NetBase = class extends FaceProcessor {
+  postProcess(output, inputSize, originalDimensions) {
+    const inputDimensions = originalDimensions.map(({ width, height }) => {
+      const scale2 = inputSize / Math.max(height, width);
+      return {
+        width: width * scale2,
+        height: height * scale2
+      };
+    });
+    const batchSize = inputDimensions.length;
+    return tfjs_esm_exports.tidy(() => {
+      const createInterleavedTensor = (fillX, fillY) => tfjs_esm_exports.stack([tfjs_esm_exports.fill([68], fillX, "float32"), tfjs_esm_exports.fill([68], fillY, "float32")], 1).as2D(1, 136).as1D();
+      const getPadding = (batchIdx, cond) => {
+        const { width, height } = inputDimensions[batchIdx];
+        return cond(width, height) ? Math.abs(width - height) / 2 : 0;
+      };
+      const getPaddingX = (batchIdx) => getPadding(batchIdx, (w, h) => w < h);
+      const getPaddingY = (batchIdx) => getPadding(batchIdx, (w, h) => h < w);
+      const landmarkTensors = output.mul(tfjs_esm_exports.fill([batchSize, 136], inputSize, "float32")).sub(tfjs_esm_exports.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        getPaddingX(batchIdx),
+        getPaddingY(batchIdx)
+      )))).div(tfjs_esm_exports.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        inputDimensions[batchIdx].width,
+        inputDimensions[batchIdx].height
+      ))));
+      return landmarkTensors;
+    });
+  }
+  forwardInput(input) {
+    return tfjs_esm_exports.tidy(() => {
+      const out = this.runNet(input);
+      return this.postProcess(
+        out,
+        input.inputSize,
+        input.inputDimensions.map(([height, width]) => ({ height, width }))
+      );
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async detectLandmarks(input) {
+    const netInput = await toNetInput(input);
+    const landmarkTensors = tfjs_esm_exports.tidy(
+      () => tfjs_esm_exports.unstack(this.forwardInput(netInput))
+    );
+    const landmarksForBatch = await Promise.all(landmarkTensors.map(
+      async (landmarkTensor, batchIdx) => {
+        const landmarksArray = Array.from(landmarkTensor.dataSync());
+        const xCoords = landmarksArray.filter((_, i) => isEven(i));
+        const yCoords = landmarksArray.filter((_, i) => !isEven(i));
+        return new FaceLandmarks68(
+          Array(68).fill(0).map((_, i) => new Point(xCoords[i], yCoords[i])),
+          {
+            height: netInput.getInputHeight(batchIdx),
+            width: netInput.getInputWidth(batchIdx)
+          }
+        );
+      }
+    ));
+    landmarkTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? landmarksForBatch : landmarksForBatch[0];
+  }
+  getClassifierChannelsOut() {
+    return 136;
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68Net.ts
+var FaceLandmark68Net = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceLandmark68Net", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+};
+
+// src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts
+function extractParamsFromWeightMapTiny(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock3Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock3Params("dense0", true),
+    dense1: extractDenseBlock3Params("dense1"),
+    dense2: extractDenseBlock3Params("dense2")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/extractParamsTiny.ts
+function extractParamsTiny(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock3Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock3Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock3Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock3Params(64, 128, "dense2");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { dense0, dense1, dense2 }
+  };
+}
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var TinyFaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("TinyFaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyFaceFeatureExtractor - load model before inference");
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock3(normalized, params.dense0, true);
+      out = denseBlock3(out, params.dense1);
+      out = denseBlock3(out, params.dense2);
+      out = tfjs_esm_exports.avgPool(out, [14, 14], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_tiny_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMapTiny(weightMap);
+  }
+  extractParams(weights) {
+    return extractParamsTiny(weights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68TinyNet.ts
+var FaceLandmark68TinyNet = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new TinyFaceFeatureExtractor()) {
+    super("FaceLandmark68TinyNet", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_tiny_model";
+  }
+  getClassifierChannelsIn() {
+    return 128;
+  }
+};
+
+// src/faceLandmarkNet/index.ts
+var FaceLandmarkNet = class extends FaceLandmark68Net {
+};
+
+// src/faceRecognitionNet/scaleLayer.ts
+function scale(x, params) {
+  return tfjs_esm_exports.add(tfjs_esm_exports.mul(x, params.weights), params.biases);
+}
+
+// src/faceRecognitionNet/convLayer.ts
+function convLayer2(x, params, strides, withRelu, padding = "same") {
+  const { filters, bias } = params.conv;
+  let out = tfjs_esm_exports.conv2d(x, filters, strides, padding);
+  out = tfjs_esm_exports.add(out, bias);
+  out = scale(out, params.scale);
+  return withRelu ? tfjs_esm_exports.relu(out) : out;
+}
+function conv2(x, params) {
+  return convLayer2(x, params, [1, 1], true);
+}
+function convNoRelu(x, params) {
+  return convLayer2(x, params, [1, 1], false);
+}
+function convDown(x, params) {
+  return convLayer2(x, params, [2, 2], true, "valid");
+}
+
+// src/faceRecognitionNet/extractParams.ts
+function extractorsFactory3(extractWeights, paramMappings) {
+  function extractFilterValues(numFilterValues, numFilters, filterSize) {
+    const weights = extractWeights(numFilterValues);
+    const depth = weights.length / (numFilters * filterSize * filterSize);
+    if (isFloat(depth)) {
+      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);
+    }
+    return tfjs_esm_exports.tidy(
+      () => tfjs_esm_exports.transpose(
+        tfjs_esm_exports.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),
+        [2, 3, 1, 0]
+      )
+    );
+  }
+  function extractConvParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);
+    const bias = tfjs_esm_exports.tensor1d(extractWeights(numFilters));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  }
+  function extractScaleLayerParams(numWeights, mappedPrefix) {
+    const weights = tfjs_esm_exports.tensor1d(extractWeights(numWeights));
+    const biases = tfjs_esm_exports.tensor1d(extractWeights(numWeights));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/biases` }
+    );
+    return {
+      weights,
+      biases
+    };
+  }
+  function extractConvLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const conv3 = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);
+    const scale2 = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);
+    return { conv: conv3, scale: scale2 };
+  }
+  function extractResidualLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix, isDown = false) {
+    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);
+    const conv22 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);
+    return { conv1, conv2: conv22 };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParams5(weights) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory3(extractWeights, paramMappings);
+  const conv32_down = extractConvLayerParams(4704, 32, 7, "conv32_down");
+  const conv32_1 = extractResidualLayerParams(9216, 32, 3, "conv32_1");
+  const conv32_2 = extractResidualLayerParams(9216, 32, 3, "conv32_2");
+  const conv32_3 = extractResidualLayerParams(9216, 32, 3, "conv32_3");
+  const conv64_down = extractResidualLayerParams(36864, 64, 3, "conv64_down", true);
+  const conv64_1 = extractResidualLayerParams(36864, 64, 3, "conv64_1");
+  const conv64_2 = extractResidualLayerParams(36864, 64, 3, "conv64_2");
+  const conv64_3 = extractResidualLayerParams(36864, 64, 3, "conv64_3");
+  const conv128_down = extractResidualLayerParams(147456, 128, 3, "conv128_down", true);
+  const conv128_1 = extractResidualLayerParams(147456, 128, 3, "conv128_1");
+  const conv128_2 = extractResidualLayerParams(147456, 128, 3, "conv128_2");
+  const conv256_down = extractResidualLayerParams(589824, 256, 3, "conv256_down", true);
+  const conv256_1 = extractResidualLayerParams(589824, 256, 3, "conv256_1");
+  const conv256_2 = extractResidualLayerParams(589824, 256, 3, "conv256_2");
+  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, "conv256_down_out");
+  const fc = tfjs_esm_exports.tidy(
+    () => tfjs_esm_exports.transpose(tfjs_esm_exports.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0])
+  );
+  paramMappings.push({ paramPath: "fc" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/extractParamsFromWeightMap.ts
+function extractorsFactory4(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractScaleLayerParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);
+    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);
+    return { weights, biases };
+  }
+  function extractConvLayerParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);
+    const scale2 = extractScaleLayerParams(prefix);
+    return { conv: { filters, bias }, scale: scale2 };
+  }
+  function extractResidualLayerParams(prefix) {
+    return {
+      conv1: extractConvLayerParams(`${prefix}/conv1`),
+      conv2: extractConvLayerParams(`${prefix}/conv2`)
+    };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParamsFromWeightMap5(weightMap) {
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory4(weightMap, paramMappings);
+  const conv32_down = extractConvLayerParams("conv32_down");
+  const conv32_1 = extractResidualLayerParams("conv32_1");
+  const conv32_2 = extractResidualLayerParams("conv32_2");
+  const conv32_3 = extractResidualLayerParams("conv32_3");
+  const conv64_down = extractResidualLayerParams("conv64_down");
+  const conv64_1 = extractResidualLayerParams("conv64_1");
+  const conv64_2 = extractResidualLayerParams("conv64_2");
+  const conv64_3 = extractResidualLayerParams("conv64_3");
+  const conv128_down = extractResidualLayerParams("conv128_down");
+  const conv128_1 = extractResidualLayerParams("conv128_1");
+  const conv128_2 = extractResidualLayerParams("conv128_2");
+  const conv256_down = extractResidualLayerParams("conv256_down");
+  const conv256_1 = extractResidualLayerParams("conv256_1");
+  const conv256_2 = extractResidualLayerParams("conv256_2");
+  const conv256_down_out = extractResidualLayerParams("conv256_down_out");
+  const { fc } = weightMap;
+  paramMappings.push({ originalPath: "fc", paramPath: "fc" });
+  if (!isTensor2D(fc)) {
+    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/residualLayer.ts
+function residual(x, params) {
+  let out = conv2(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  out = tfjs_esm_exports.add(out, x);
+  out = tfjs_esm_exports.relu(out);
+  return out;
+}
+function residualDown(x, params) {
+  let out = convDown(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  let pooled = tfjs_esm_exports.avgPool(x, 2, 2, "valid");
+  const zeros2 = tfjs_esm_exports.zeros(pooled.shape);
+  const isPad = pooled.shape[3] !== out.shape[3];
+  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];
+  if (isAdjustShape) {
+    const padShapeX = [...out.shape];
+    padShapeX[1] = 1;
+    const zerosW = tfjs_esm_exports.zeros(padShapeX);
+    out = tfjs_esm_exports.concat([out, zerosW], 1);
+    const padShapeY = [...out.shape];
+    padShapeY[2] = 1;
+    const zerosH = tfjs_esm_exports.zeros(padShapeY);
+    out = tfjs_esm_exports.concat([out, zerosH], 2);
+  }
+  pooled = isPad ? tfjs_esm_exports.concat([pooled, zeros2], 3) : pooled;
+  out = tfjs_esm_exports.add(pooled, out);
+  out = tfjs_esm_exports.relu(out);
+  return out;
+}
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var FaceRecognitionNet = class extends NeuralNetwork {
+  constructor() {
+    super("FaceRecognitionNet");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceRecognitionNet - load model before inference");
+    }
+    return tfjs_esm_exports.tidy(() => {
+      const batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(150, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = convDown(normalized, params.conv32_down);
+      out = tfjs_esm_exports.maxPool(out, 3, 2, "valid");
+      out = residual(out, params.conv32_1);
+      out = residual(out, params.conv32_2);
+      out = residual(out, params.conv32_3);
+      out = residualDown(out, params.conv64_down);
+      out = residual(out, params.conv64_1);
+      out = residual(out, params.conv64_2);
+      out = residual(out, params.conv64_3);
+      out = residualDown(out, params.conv128_down);
+      out = residual(out, params.conv128_1);
+      out = residual(out, params.conv128_2);
+      out = residualDown(out, params.conv256_down);
+      out = residual(out, params.conv256_1);
+      out = residual(out, params.conv256_2);
+      out = residualDown(out, params.conv256_down_out);
+      const globalAvg = out.mean([1, 2]);
+      const fullyConnected = tfjs_esm_exports.matMul(globalAvg, params.fc);
+      return fullyConnected;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async computeFaceDescriptor(input) {
+    var _a;
+    if ((_a = input == null ? void 0 : input.shape) == null ? void 0 : _a.some((dim) => dim <= 0))
+      return new Float32Array(128);
+    const netInput = await toNetInput(input);
+    const faceDescriptorTensors = tfjs_esm_exports.tidy(() => tfjs_esm_exports.unstack(this.forwardInput(netInput)));
+    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data()));
+    faceDescriptorTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_recognition_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap5(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams5(weights);
+  }
+};
+
+// src/faceRecognitionNet/index.ts
+function createFaceRecognitionNet(weights) {
+  const net = new FaceRecognitionNet();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/factories/WithFaceDescriptor.ts
+function extendWithFaceDescriptor(sourceObj, descriptor) {
+  const extension = { descriptor };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithAge.ts
+function isWithAge(obj) {
+  return typeof obj.age === "number";
+}
+function extendWithAge(sourceObj, age) {
+  const extension = { age };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithGender.ts
+function isWithGender(obj) {
+  return (obj.gender === "male" /* MALE */ || obj.gender === "female" /* FEMALE */) && isValidProbablitiy(obj.genderProbability);
+}
+function extendWithGender(sourceObj, gender, genderProbability) {
+  const extension = { gender, genderProbability };
+  return { ...sourceObj, ...extension };
+}
+
+// src/ssdMobilenetv1/extractParams.ts
+function extractorsFactory5(extractWeights, paramMappings) {
+  function extractDepthwiseConvParams(numChannels, mappedPrefix) {
+    const filters = tfjs_esm_exports.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);
+    const batch_norm_scale = tfjs_esm_exports.tensor1d(extractWeights(numChannels));
+    const batch_norm_offset = tfjs_esm_exports.tensor1d(extractWeights(numChannels));
+    const batch_norm_mean = tfjs_esm_exports.tensor1d(extractWeights(numChannels));
+    const batch_norm_variance = tfjs_esm_exports.tensor1d(extractWeights(numChannels));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/batch_norm_scale` },
+      { paramPath: `${mappedPrefix}/batch_norm_offset` },
+      { paramPath: `${mappedPrefix}/batch_norm_mean` },
+      { paramPath: `${mappedPrefix}/batch_norm_variance` }
+    );
+    return {
+      filters,
+      batch_norm_scale,
+      batch_norm_offset,
+      batch_norm_mean,
+      batch_norm_variance
+    };
+  }
+  function extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, isPointwiseConv) {
+    const filters = tfjs_esm_exports.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tfjs_esm_exports.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/${isPointwiseConv ? "batch_norm_offset" : "bias"}` }
+    );
+    return { filters, bias };
+  }
+  function extractPointwiseConvParams(channelsIn, channelsOut, filterSize, mappedPrefix) {
+    const {
+      filters,
+      bias
+    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);
+    return {
+      filters,
+      batch_norm_offset: bias
+    };
+  }
+  function extractConvPairParams(channelsIn, channelsOut, mappedPrefix) {
+    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);
+    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);
+    return { depthwise_conv, pointwise_conv };
+  }
+  function extractMobilenetV1Params() {
+    const conv_0 = extractPointwiseConvParams(3, 32, 3, "mobilenetv1/conv_0");
+    const conv_1 = extractConvPairParams(32, 64, "mobilenetv1/conv_1");
+    const conv_2 = extractConvPairParams(64, 128, "mobilenetv1/conv_2");
+    const conv_3 = extractConvPairParams(128, 128, "mobilenetv1/conv_3");
+    const conv_4 = extractConvPairParams(128, 256, "mobilenetv1/conv_4");
+    const conv_5 = extractConvPairParams(256, 256, "mobilenetv1/conv_5");
+    const conv_6 = extractConvPairParams(256, 512, "mobilenetv1/conv_6");
+    const conv_7 = extractConvPairParams(512, 512, "mobilenetv1/conv_7");
+    const conv_8 = extractConvPairParams(512, 512, "mobilenetv1/conv_8");
+    const conv_9 = extractConvPairParams(512, 512, "mobilenetv1/conv_9");
+    const conv_10 = extractConvPairParams(512, 512, "mobilenetv1/conv_10");
+    const conv_11 = extractConvPairParams(512, 512, "mobilenetv1/conv_11");
+    const conv_12 = extractConvPairParams(512, 1024, "mobilenetv1/conv_12");
+    const conv_13 = extractConvPairParams(1024, 1024, "mobilenetv1/conv_13");
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      conv_8,
+      conv_9,
+      conv_10,
+      conv_11,
+      conv_12,
+      conv_13
+    };
+  }
+  function extractPredictionLayerParams() {
+    const conv_0 = extractPointwiseConvParams(1024, 256, 1, "prediction_layer/conv_0");
+    const conv_1 = extractPointwiseConvParams(256, 512, 3, "prediction_layer/conv_1");
+    const conv_2 = extractPointwiseConvParams(512, 128, 1, "prediction_layer/conv_2");
+    const conv_3 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_3");
+    const conv_4 = extractPointwiseConvParams(256, 128, 1, "prediction_layer/conv_4");
+    const conv_5 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_5");
+    const conv_6 = extractPointwiseConvParams(256, 64, 1, "prediction_layer/conv_6");
+    const conv_7 = extractPointwiseConvParams(64, 128, 3, "prediction_layer/conv_7");
+    const box_encoding_0_predictor = extractConvParams(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor");
+    const class_predictor_0 = extractConvParams(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor");
+    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor");
+    const class_predictor_1 = extractConvParams(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor");
+    const box_encoding_2_predictor = extractConvParams(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor");
+    const class_predictor_2 = extractConvParams(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor");
+    const box_encoding_3_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor");
+    const class_predictor_3 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor");
+    const box_encoding_4_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor");
+    const class_predictor_4 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor");
+    const box_encoding_5_predictor = extractConvParams(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor");
+    const class_predictor_5 = extractConvParams(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor");
+    const box_predictor_0 = {
+      box_encoding_predictor: box_encoding_0_predictor,
+      class_predictor: class_predictor_0
+    };
+    const box_predictor_1 = {
+      box_encoding_predictor: box_encoding_1_predictor,
+      class_predictor: class_predictor_1
+    };
+    const box_predictor_2 = {
+      box_encoding_predictor: box_encoding_2_predictor,
+      class_predictor: class_predictor_2
+    };
+    const box_predictor_3 = {
+      box_encoding_predictor: box_encoding_3_predictor,
+      class_predictor: class_predictor_3
+    };
+    const box_predictor_4 = {
+      box_encoding_predictor: box_encoding_4_predictor,
+      class_predictor: class_predictor_4
+    };
+    const box_predictor_5 = {
+      box_encoding_predictor: box_encoding_5_predictor,
+      class_predictor: class_predictor_5
+    };
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      box_predictor_0,
+      box_predictor_1,
+      box_predictor_2,
+      box_predictor_3,
+      box_predictor_4,
+      box_predictor_5
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParams6(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory5(extractWeights, paramMappings);
+  const mobilenetv1 = extractMobilenetV1Params();
+  const prediction_layer = extractPredictionLayerParams();
+  const extra_dim = tfjs_esm_exports.tensor3d(
+    extractWeights(5118 * 4),
+    [1, 5118, 4]
+  );
+  const output_layer = {
+    extra_dim
+  };
+  paramMappings.push({ paramPath: "output_layer/extra_dim" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    params: {
+      mobilenetv1,
+      prediction_layer,
+      output_layer
+    },
+    paramMappings
+  };
+}
+
+// src/ssdMobilenetv1/extractParamsFromWeightMap.ts
+function extractorsFactory6(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractPointwiseConvParams(prefix, idx, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);
+    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);
+    return { filters, batch_norm_offset };
+  }
+  function extractConvPairParams(idx) {
+    const mappedPrefix = `mobilenetv1/conv_${idx}`;
+    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;
+    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;
+    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;
+    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);
+    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);
+    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);
+    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);
+    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);
+    return {
+      depthwise_conv: {
+        filters,
+        batch_norm_scale,
+        batch_norm_offset,
+        batch_norm_mean,
+        batch_norm_variance
+      },
+      pointwise_conv: extractPointwiseConvParams("MobilenetV1", idx, mappedPrefixPointwiseConv)
+    };
+  }
+  function extractMobilenetV1Params() {
+    return {
+      conv_0: extractPointwiseConvParams("MobilenetV1", 0, "mobilenetv1/conv_0"),
+      conv_1: extractConvPairParams(1),
+      conv_2: extractConvPairParams(2),
+      conv_3: extractConvPairParams(3),
+      conv_4: extractConvPairParams(4),
+      conv_5: extractConvPairParams(5),
+      conv_6: extractConvPairParams(6),
+      conv_7: extractConvPairParams(7),
+      conv_8: extractConvPairParams(8),
+      conv_9: extractConvPairParams(9),
+      conv_10: extractConvPairParams(10),
+      conv_11: extractConvPairParams(11),
+      conv_12: extractConvPairParams(12),
+      conv_13: extractConvPairParams(13)
+    };
+  }
+  function extractConvParams(prefix, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);
+    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);
+    return { filters, bias };
+  }
+  function extractBoxPredictorParams(idx) {
+    const box_encoding_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,
+      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`
+    );
+    const class_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/ClassPredictor`,
+      `prediction_layer/box_predictor_${idx}/class_predictor`
+    );
+    return { box_encoding_predictor, class_predictor };
+  }
+  function extractPredictionLayerParams() {
+    return {
+      conv_0: extractPointwiseConvParams("Prediction", 0, "prediction_layer/conv_0"),
+      conv_1: extractPointwiseConvParams("Prediction", 1, "prediction_layer/conv_1"),
+      conv_2: extractPointwiseConvParams("Prediction", 2, "prediction_layer/conv_2"),
+      conv_3: extractPointwiseConvParams("Prediction", 3, "prediction_layer/conv_3"),
+      conv_4: extractPointwiseConvParams("Prediction", 4, "prediction_layer/conv_4"),
+      conv_5: extractPointwiseConvParams("Prediction", 5, "prediction_layer/conv_5"),
+      conv_6: extractPointwiseConvParams("Prediction", 6, "prediction_layer/conv_6"),
+      conv_7: extractPointwiseConvParams("Prediction", 7, "prediction_layer/conv_7"),
+      box_predictor_0: extractBoxPredictorParams(0),
+      box_predictor_1: extractBoxPredictorParams(1),
+      box_predictor_2: extractBoxPredictorParams(2),
+      box_predictor_3: extractBoxPredictorParams(3),
+      box_predictor_4: extractBoxPredictorParams(4),
+      box_predictor_5: extractBoxPredictorParams(5)
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParamsFromWeightMap6(weightMap) {
+  const paramMappings = [];
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory6(weightMap, paramMappings);
+  const extra_dim = weightMap["Output/extra_dim"];
+  paramMappings.push({ originalPath: "Output/extra_dim", paramPath: "output_layer/extra_dim" });
+  if (!isTensor3D(extra_dim)) {
+    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);
+  }
+  const params = {
+    mobilenetv1: extractMobilenetV1Params(),
+    prediction_layer: extractPredictionLayerParams(),
+    output_layer: {
+      extra_dim
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ssdMobilenetv1/pointwiseConvLayer.ts
+function pointwiseConvLayer(x, params, strides) {
+  return tfjs_esm_exports.tidy(() => {
+    let out = tfjs_esm_exports.conv2d(x, params.filters, strides, "same");
+    out = tfjs_esm_exports.add(out, params.batch_norm_offset);
+    return tfjs_esm_exports.clipByValue(out, 0, 6);
+  });
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var epsilon = 0.0010000000474974513;
+function depthwiseConvLayer(x, params, strides) {
+  return tfjs_esm_exports.tidy(() => {
+    let out = tfjs_esm_exports.depthwiseConv2d(x, params.filters, strides, "same");
+    out = tfjs_esm_exports.batchNorm(
+      out,
+      params.batch_norm_mean,
+      params.batch_norm_variance,
+      params.batch_norm_offset,
+      params.batch_norm_scale,
+      epsilon
+    );
+    return tfjs_esm_exports.clipByValue(out, 0, 6);
+  });
+}
+function getStridesForLayerIdx(layerIdx) {
+  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];
+}
+function mobileNetV1(x, params) {
+  return tfjs_esm_exports.tidy(() => {
+    let conv11;
+    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
+    const convPairParams = [
+      params.conv_1,
+      params.conv_2,
+      params.conv_3,
+      params.conv_4,
+      params.conv_5,
+      params.conv_6,
+      params.conv_7,
+      params.conv_8,
+      params.conv_9,
+      params.conv_10,
+      params.conv_11,
+      params.conv_12,
+      params.conv_13
+    ];
+    convPairParams.forEach((param, i) => {
+      const layerIdx = i + 1;
+      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
+      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
+      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
+      if (layerIdx === 11)
+        conv11 = out;
+    });
+    if (conv11 === null) {
+      throw new Error("mobileNetV1 - output of conv layer 11 is null");
+    }
+    return {
+      out,
+      conv11
+    };
+  });
+}
+
+// src/ssdMobilenetv1/nonMaxSuppression.ts
+function IOU(boxes, i, j) {
+  const boxesData = boxes.arraySync();
+  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);
+  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);
+  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);
+  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);
+  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);
+  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);
+  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);
+  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0)
+    return 0;
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function nonMaxSuppression2(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  const numBoxes = boxes.shape[0];
+  const outputSize = Math.min(maxOutputSize, numBoxes);
+  const candidates = scores.map((score, boxIndex) => ({ score, boxIndex })).filter((c) => c.score > scoreThreshold).sort((c1, c2) => c2.score - c1.score);
+  const suppressFunc = (x) => x <= iouThreshold ? 1 : 0;
+  const selected = [];
+  candidates.forEach((c) => {
+    if (selected.length >= outputSize)
+      return;
+    const originalScore = c.score;
+    for (let j = selected.length - 1; j >= 0; --j) {
+      const iou2 = IOU(boxes, c.boxIndex, selected[j]);
+      if (iou2 === 0)
+        continue;
+      c.score *= suppressFunc(iou2);
+      if (c.score <= scoreThreshold)
+        break;
+    }
+    if (originalScore === c.score) {
+      selected.push(c.boxIndex);
+    }
+  });
+  return selected;
+}
+
+// src/ssdMobilenetv1/outputLayer.ts
+function getCenterCoordinatesAndSizesLayer(x) {
+  const vec = tfjs_esm_exports.unstack(tfjs_esm_exports.transpose(x, [1, 0]));
+  const sizes = [
+    tfjs_esm_exports.sub(vec[2], vec[0]),
+    tfjs_esm_exports.sub(vec[3], vec[1])
+  ];
+  const centers = [
+    tfjs_esm_exports.add(vec[0], tfjs_esm_exports.div(sizes[0], 2)),
+    tfjs_esm_exports.add(vec[1], tfjs_esm_exports.div(sizes[1], 2))
+  ];
+  return { sizes, centers };
+}
+function decodeBoxesLayer(x0, x1) {
+  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);
+  const vec = tfjs_esm_exports.unstack(tfjs_esm_exports.transpose(x1, [1, 0]));
+  const div0_out = tfjs_esm_exports.div(tfjs_esm_exports.mul(tfjs_esm_exports.exp(tfjs_esm_exports.div(vec[2], 5)), sizes[0]), 2);
+  const add0_out = tfjs_esm_exports.add(tfjs_esm_exports.mul(tfjs_esm_exports.div(vec[0], 10), sizes[0]), centers[0]);
+  const div1_out = tfjs_esm_exports.div(tfjs_esm_exports.mul(tfjs_esm_exports.exp(tfjs_esm_exports.div(vec[3], 5)), sizes[1]), 2);
+  const add1_out = tfjs_esm_exports.add(tfjs_esm_exports.mul(tfjs_esm_exports.div(vec[1], 10), sizes[1]), centers[1]);
+  return tfjs_esm_exports.transpose(
+    tfjs_esm_exports.stack([
+      tfjs_esm_exports.sub(add0_out, div0_out),
+      tfjs_esm_exports.sub(add1_out, div1_out),
+      tfjs_esm_exports.add(add0_out, div0_out),
+      tfjs_esm_exports.add(add1_out, div1_out)
+    ]),
+    [1, 0]
+  );
+}
+function outputLayer(boxPredictions, classPredictions, params) {
+  return tfjs_esm_exports.tidy(() => {
+    const batchSize = boxPredictions.shape[0];
+    let boxes = decodeBoxesLayer(
+      tfjs_esm_exports.reshape(tfjs_esm_exports.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]),
+      tfjs_esm_exports.reshape(boxPredictions, [-1, 4])
+    );
+    boxes = tfjs_esm_exports.reshape(boxes, [batchSize, boxes.shape[0] / batchSize, 4]);
+    const scoresAndClasses = tfjs_esm_exports.sigmoid(tfjs_esm_exports.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));
+    let scores = tfjs_esm_exports.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]);
+    scores = tfjs_esm_exports.reshape(scores, [batchSize, scores.shape[1]]);
+    const boxesByBatch = tfjs_esm_exports.unstack(boxes);
+    const scoresByBatch = tfjs_esm_exports.unstack(scores);
+    return { boxes: boxesByBatch, scores: scoresByBatch };
+  });
+}
+
+// src/ssdMobilenetv1/boxPredictionLayer.ts
+function boxPredictionLayer(x, params) {
+  return tfjs_esm_exports.tidy(() => {
+    const batchSize = x.shape[0];
+    const boxPredictionEncoding = tfjs_esm_exports.reshape(
+      convLayer(x, params.box_encoding_predictor),
+      [batchSize, -1, 1, 4]
+    );
+    const classPrediction = tfjs_esm_exports.reshape(
+      convLayer(x, params.class_predictor),
+      [batchSize, -1, 3]
+    );
+    return { boxPredictionEncoding, classPrediction };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+function predictionLayer(x, conv11, params) {
+  return tfjs_esm_exports.tidy(() => {
+    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);
+    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);
+    const conv22 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);
+    const conv3 = pointwiseConvLayer(conv22, params.conv_3, [2, 2]);
+    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);
+    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);
+    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);
+    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);
+    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);
+    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);
+    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);
+    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);
+    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);
+    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);
+    const boxPredictions = tfjs_esm_exports.concat([
+      boxPrediction0.boxPredictionEncoding,
+      boxPrediction1.boxPredictionEncoding,
+      boxPrediction2.boxPredictionEncoding,
+      boxPrediction3.boxPredictionEncoding,
+      boxPrediction4.boxPredictionEncoding,
+      boxPrediction5.boxPredictionEncoding
+    ], 1);
+    const classPredictions = tfjs_esm_exports.concat([
+      boxPrediction0.classPrediction,
+      boxPrediction1.classPrediction,
+      boxPrediction2.classPrediction,
+      boxPrediction3.classPrediction,
+      boxPrediction4.classPrediction,
+      boxPrediction5.classPrediction
+    ], 1);
+    return {
+      boxPredictions,
+      classPredictions
+    };
+  });
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1Options.ts
+var SsdMobilenetv1Options = class {
+  constructor({ minConfidence, maxResults } = {}) {
+    this._name = "SsdMobilenetv1Options";
+    this._minConfidence = minConfidence || 0.5;
+    this._maxResults = maxResults || 100;
+    if (typeof this._minConfidence !== "number" || this._minConfidence <= 0 || this._minConfidence >= 1) {
+      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);
+    }
+    if (typeof this._maxResults !== "number") {
+      throw new Error(`${this._name} - expected maxResults to be a number`);
+    }
+  }
+  get minConfidence() {
+    return this._minConfidence;
+  }
+  get maxResults() {
+    return this._maxResults;
+  }
+};
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var SsdMobilenetv1 = class extends NeuralNetwork {
+  constructor() {
+    super("SsdMobilenetv1");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params)
+      throw new Error("SsdMobilenetv1 - load model before inference");
+    return tfjs_esm_exports.tidy(() => {
+      const batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(512, false), "float32");
+      const x = tfjs_esm_exports.sub(tfjs_esm_exports.div(batchTensor, 127.5), 1);
+      const features = mobileNetV1(x, params.mobilenetv1);
+      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);
+      return outputLayer(boxPredictions, classPredictions, params.output_layer);
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async locateFaces(input, options = {}) {
+    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);
+    const netInput = await toNetInput(input);
+    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);
+    const boxes = _boxes[0];
+    const scores = _scores[0];
+    for (let i = 1; i < _boxes.length; i++) {
+      _boxes[i].dispose();
+      _scores[i].dispose();
+    }
+    const scoresData = Array.from(scores.dataSync());
+    const iouThreshold = 0.5;
+    const indices = nonMaxSuppression2(boxes, scoresData, maxResults, iouThreshold, minConfidence);
+    const reshapedDims = netInput.getReshapedInputDimensions(0);
+    const inputSize = netInput.inputSize;
+    const padX = inputSize / reshapedDims.width;
+    const padY = inputSize / reshapedDims.height;
+    const boxesData = boxes.arraySync();
+    const results = indices.map((idx) => {
+      const [top, bottom] = [
+        Math.max(0, boxesData[idx][0]),
+        Math.min(1, boxesData[idx][2])
+      ].map((val) => val * padY);
+      const [left, right] = [
+        Math.max(0, boxesData[idx][1]),
+        Math.min(1, boxesData[idx][3])
+      ].map((val) => val * padX);
+      return new FaceDetection(
+        scoresData[idx],
+        new Rect(left, top, right - left, bottom - top),
+        { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) }
+      );
+    });
+    boxes.dispose();
+    scores.dispose();
+    return results;
+  }
+  getDefaultModelName() {
+    return "ssd_mobilenetv1_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap6(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams6(weights);
+  }
+};
+
+// src/ssdMobilenetv1/index.ts
+function createSsdMobilenetv1(weights) {
+  const net = new SsdMobilenetv1();
+  net.extractWeights(weights);
+  return net;
+}
+function createFaceDetectionNet(weights) {
+  return createSsdMobilenetv1(weights);
+}
+var FaceDetectionNet = class extends SsdMobilenetv1 {
+};
+
+// src/tinyYolov2/const.ts
+var IOU_THRESHOLD = 0.4;
+var BOX_ANCHORS = [
+  new Point(0.738768, 0.874946),
+  new Point(2.42204, 2.65704),
+  new Point(4.30971, 7.04493),
+  new Point(10.246, 4.59428),
+  new Point(12.6868, 11.8741)
+];
+var BOX_ANCHORS_SEPARABLE = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB_SEPARABLE = [117.001, 114.697, 97.404];
+var DEFAULT_MODEL_NAME = "tiny_yolov2_model";
+var DEFAULT_MODEL_NAME_SEPARABLE_CONV = "tiny_yolov2_separable_conv_model";
+
+// src/tinyYolov2/config.ts
+var isNumber = (arg) => typeof arg === "number";
+function validateConfig(config) {
+  if (!config) {
+    throw new Error(`invalid config: ${config}`);
+  }
+  if (typeof config.withSeparableConvs !== "boolean") {
+    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);
+  }
+  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1) {
+    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);
+  }
+  if (!Array.isArray(config.classes) || !config.classes.length || !config.classes.every((c) => typeof c === "string")) {
+    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);
+  }
+  if (!Array.isArray(config.anchors) || !config.anchors.length || !config.anchors.map((a) => a || {}).every((a) => isNumber(a.x) && isNumber(a.y))) {
+    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);
+  }
+  if (config.meanRgb && (!Array.isArray(config.meanRgb) || config.meanRgb.length !== 3 || !config.meanRgb.every(isNumber))) {
+    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);
+  }
+}
+
+// src/tinyYolov2/leaky.ts
+function leaky(x) {
+  return tfjs_esm_exports.tidy(() => {
+    const min = tfjs_esm_exports.mul(x, tfjs_esm_exports.scalar(0.10000000149011612));
+    return tfjs_esm_exports.add(tfjs_esm_exports.relu(tfjs_esm_exports.sub(x, min)), min);
+  });
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+function convWithBatchNorm(x, params) {
+  return tfjs_esm_exports.tidy(() => {
+    let out = tfjs_esm_exports.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tfjs_esm_exports.conv2d(out, params.conv.filters, [1, 1], "valid");
+    out = tfjs_esm_exports.sub(out, params.bn.sub);
+    out = tfjs_esm_exports.mul(out, params.bn.truediv);
+    out = tfjs_esm_exports.add(out, params.conv.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/depthwiseSeparableConv.ts
+function depthwiseSeparableConv2(x, params) {
+  return tfjs_esm_exports.tidy(() => {
+    let out = tfjs_esm_exports.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tfjs_esm_exports.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], "valid");
+    out = tfjs_esm_exports.add(out, params.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/extractParams.ts
+function extractorsFactory7(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  function extractBatchNormParams(size, mappedPrefix) {
+    const sub6 = tfjs_esm_exports.tensor1d(extractWeights(size));
+    const truediv = tfjs_esm_exports.tensor1d(extractWeights(size));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/sub` },
+      { paramPath: `${mappedPrefix}/truediv` }
+    );
+    return { sub: sub6, truediv };
+  }
+  function extractConvWithBatchNormParams(channelsIn, channelsOut, mappedPrefix) {
+    const conv3 = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);
+    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParams7(weights, config, boxEncodingSize, filterSizes) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory7(extractWeights, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = config.isFirstLayerConv2d ? extractConvParams(s0, s1, 3, "conv0") : extractSeparableConvParams(s0, s1, "conv0");
+    const conv1 = extractSeparableConvParams(s1, s2, "conv1");
+    const conv22 = extractSeparableConvParams(s2, s3, "conv2");
+    const conv3 = extractSeparableConvParams(s3, s4, "conv3");
+    const conv4 = extractSeparableConvParams(s4, s5, "conv4");
+    const conv5 = extractSeparableConvParams(s5, s6, "conv5");
+    const conv6 = s7 ? extractSeparableConvParams(s6, s7, "conv6") : void 0;
+    const conv7 = s8 ? extractSeparableConvParams(s7, s8, "conv7") : void 0;
+    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  } else {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = extractConvWithBatchNormParams(s0, s1, "conv0");
+    const conv1 = extractConvWithBatchNormParams(s1, s2, "conv1");
+    const conv22 = extractConvWithBatchNormParams(s2, s3, "conv2");
+    const conv3 = extractConvWithBatchNormParams(s3, s4, "conv3");
+    const conv4 = extractConvWithBatchNormParams(s4, s5, "conv4");
+    const conv5 = extractConvWithBatchNormParams(s5, s6, "conv5");
+    const conv6 = extractConvWithBatchNormParams(s6, s7, "conv6");
+    const conv7 = extractConvWithBatchNormParams(s7, s8, "conv7");
+    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  }
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/extractParamsFromWeightMap.ts
+function extractorsFactory8(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractBatchNormParams(prefix) {
+    const sub6 = extractWeightEntry(`${prefix}/sub`, 1);
+    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);
+    return { sub: sub6, truediv };
+  }
+  function extractConvParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  }
+  function extractConvWithBatchNormParams(prefix) {
+    const conv3 = extractConvParams(`${prefix}/conv`);
+    const bn = extractBatchNormParams(`${prefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParamsFromWeightMap7(weightMap, config) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory8(weightMap, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const numFilters = config.filterSizes && config.filterSizes.length || 9;
+    params = {
+      conv0: config.isFirstLayerConv2d ? extractConvParams("conv0") : extractSeparableConvParams("conv0"),
+      conv1: extractSeparableConvParams("conv1"),
+      conv2: extractSeparableConvParams("conv2"),
+      conv3: extractSeparableConvParams("conv3"),
+      conv4: extractSeparableConvParams("conv4"),
+      conv5: extractSeparableConvParams("conv5"),
+      conv6: numFilters > 7 ? extractSeparableConvParams("conv6") : void 0,
+      conv7: numFilters > 8 ? extractSeparableConvParams("conv7") : void 0,
+      conv8: extractConvParams("conv8")
+    };
+  } else {
+    params = {
+      conv0: extractConvWithBatchNormParams("conv0"),
+      conv1: extractConvWithBatchNormParams("conv1"),
+      conv2: extractConvWithBatchNormParams("conv2"),
+      conv3: extractConvWithBatchNormParams("conv3"),
+      conv4: extractConvWithBatchNormParams("conv4"),
+      conv5: extractConvWithBatchNormParams("conv5"),
+      conv6: extractConvWithBatchNormParams("conv6"),
+      conv7: extractConvWithBatchNormParams("conv7"),
+      conv8: extractConvParams("conv8")
+    };
+  }
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/TinyYolov2Options.ts
+var TinyYolov2Options = class {
+  constructor({ inputSize, scoreThreshold } = {}) {
+    this._name = "TinyYolov2Options";
+    this._inputSize = inputSize || 416;
+    this._scoreThreshold = scoreThreshold || 0.5;
+    if (typeof this._inputSize !== "number" || this._inputSize % 32 !== 0) {
+      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);
+    }
+    if (typeof this._scoreThreshold !== "number" || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {
+      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);
+    }
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get scoreThreshold() {
+    return this._scoreThreshold;
+  }
+};
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var _TinyYolov2Base = class extends NeuralNetwork {
+  constructor(config) {
+    super("TinyYolov2");
+    validateConfig(config);
+    this._config = config;
+  }
+  get config() {
+    return this._config;
+  }
+  get withClassScores() {
+    return this.config.withClassScores || this.config.classes.length > 1;
+  }
+  get boxEncodingSize() {
+    return 5 + (this.withClassScores ? this.config.classes.length : 0);
+  }
+  runTinyYolov2(x, params) {
+    let out = convWithBatchNorm(x, params.conv0);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv1);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv2);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv3);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv4);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv5);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [1, 1], "same");
+    out = convWithBatchNorm(out, params.conv6);
+    out = convWithBatchNorm(out, params.conv7);
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  runMobilenet(x, params) {
+    let out = this.config.isFirstLayerConv2d ? leaky(convLayer(x, params.conv0, "valid", false)) : depthwiseSeparableConv2(x, params.conv0);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv1);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv2);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv3);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv4);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv5);
+    out = tfjs_esm_exports.maxPool(out, [2, 2], [1, 1], "same");
+    out = params.conv6 ? depthwiseSeparableConv2(out, params.conv6) : out;
+    out = params.conv7 ? depthwiseSeparableConv2(out, params.conv7) : out;
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  forwardInput(input, inputSize) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyYolov2 - load model before inference");
+    }
+    return tfjs_esm_exports.tidy(() => {
+      let batchTensor = tfjs_esm_exports.cast(input.toBatchTensor(inputSize, false), "float32");
+      batchTensor = this.config.meanRgb ? normalize(batchTensor, this.config.meanRgb) : batchTensor;
+      batchTensor = batchTensor.div(255);
+      return this.config.withSeparableConvs ? this.runMobilenet(batchTensor, params) : this.runTinyYolov2(batchTensor, params);
+    });
+  }
+  async forward(input, inputSize) {
+    return this.forwardInput(await toNetInput(input), inputSize);
+  }
+  async detect(input, forwardParams = {}) {
+    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput, inputSize);
+    const out0 = tfjs_esm_exports.tidy(() => tfjs_esm_exports.unstack(out)[0].expandDims());
+    const inputDimensions = {
+      width: netInput.getInputWidth(0),
+      height: netInput.getInputHeight(0)
+    };
+    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);
+    out.dispose();
+    out0.dispose();
+    const boxes = results.map((res) => res.box);
+    const scores = results.map((res) => res.score);
+    const classScores = results.map((res) => res.classScore);
+    const classNames = results.map((res) => this.config.classes[res.label]);
+    const indices = nonMaxSuppression(
+      boxes.map((box) => box.rescale(inputSize)),
+      scores,
+      this.config.iouThreshold,
+      true
+    );
+    const detections = indices.map((idx) => new ObjectDetection(
+      scores[idx],
+      classScores[idx],
+      classNames[idx],
+      boxes[idx],
+      inputDimensions
+    ));
+    return detections;
+  }
+  getDefaultModelName() {
+    return "";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap7(weightMap, this.config);
+  }
+  extractParams(weights) {
+    const filterSizes = this.config.filterSizes || _TinyYolov2Base.DEFAULT_FILTER_SIZES;
+    const numFilters = filterSizes ? filterSizes.length : void 0;
+    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {
+      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);
+    }
+    return extractParams7(weights, this.config, this.boxEncodingSize, filterSizes);
+  }
+  async extractBoxes(outputTensor, inputBlobDimensions, scoreThreshold) {
+    const { width, height } = inputBlobDimensions;
+    const inputSize = Math.max(width, height);
+    const correctionFactorX = inputSize / width;
+    const correctionFactorY = inputSize / height;
+    const numCells = outputTensor.shape[1];
+    const numBoxes = this.config.anchors.length;
+    const [boxesTensor, scoresTensor, classScoresTensor] = tfjs_esm_exports.tidy(() => {
+      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);
+      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);
+      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);
+      const classScores = this.withClassScores ? tfjs_esm_exports.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3) : tfjs_esm_exports.scalar(0);
+      return [boxes, scores, classScores];
+    });
+    const results = [];
+    const scoresData = await scoresTensor.array();
+    const boxesData = await boxesTensor.array();
+    for (let row = 0; row < numCells; row++) {
+      for (let col = 0; col < numCells; col++) {
+        for (let anchor = 0; anchor < numBoxes; anchor++) {
+          const score = sigmoid(scoresData[row][col][anchor][0]);
+          if (!scoreThreshold || score > scoreThreshold) {
+            const ctX = (col + sigmoid(boxesData[row][col][anchor][0])) / numCells * correctionFactorX;
+            const ctY = (row + sigmoid(boxesData[row][col][anchor][1])) / numCells * correctionFactorY;
+            const widthLocal = Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x / numCells * correctionFactorX;
+            const heightLocal = Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y / numCells * correctionFactorY;
+            const x = ctX - widthLocal / 2;
+            const y = ctY - heightLocal / 2;
+            const pos = { row, col, anchor };
+            const { classScore, label } = this.withClassScores ? await this.extractPredictedClass(classScoresTensor, pos) : { classScore: 1, label: 0 };
+            results.push({
+              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),
+              score,
+              classScore: score * classScore,
+              label,
+              ...pos
+            });
+          }
+        }
+      }
+    }
+    boxesTensor.dispose();
+    scoresTensor.dispose();
+    classScoresTensor.dispose();
+    return results;
+  }
+  async extractPredictedClass(classesTensor, pos) {
+    const { row, col, anchor } = pos;
+    const classesData = await classesTensor.array();
+    return Array(this.config.classes.length).fill(0).map((_, i) => classesData[row][col][anchor][i]).map((classScore, label) => ({
+      classScore,
+      label
+    })).reduce((max, curr) => max.classScore > curr.classScore ? max : curr);
+  }
+};
+var TinyYolov2Base = _TinyYolov2Base;
+TinyYolov2Base.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+
+// src/tinyYolov2/TinyYolov2.ts
+var TinyYolov2 = class extends TinyYolov2Base {
+  constructor(withSeparableConvs = true) {
+    const config = {
+      withSeparableConvs,
+      iouThreshold: IOU_THRESHOLD,
+      classes: ["face"],
+      ...withSeparableConvs ? {
+        anchors: BOX_ANCHORS_SEPARABLE,
+        meanRgb: MEAN_RGB_SEPARABLE
+      } : {
+        anchors: BOX_ANCHORS,
+        withClassScores: true
+      }
+    };
+    super(config);
+  }
+  get withSeparableConvs() {
+    return this.config.withSeparableConvs;
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/tinyYolov2/index.ts
+function createTinyYolov2(weights, withSeparableConvs = true) {
+  const net = new TinyYolov2(withSeparableConvs);
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/tinyFaceDetector/TinyFaceDetectorOptions.ts
+var TinyFaceDetectorOptions = class extends TinyYolov2Options {
+  constructor() {
+    super(...arguments);
+    this._name = "TinyFaceDetectorOptions";
+  }
+};
+
+// src/globalApi/ComposableTask.ts
+var ComposableTask = class {
+  async then(onfulfilled) {
+    return onfulfilled(await this.run());
+  }
+  async run() {
+    throw new Error("ComposableTask - run is not implemented");
+  }
+};
+
+// src/globalApi/extractFacesAndComputeResults.ts
+async function extractAllFacesAndComputeResults(parentResults, input, computeResults, extractedFaces, getRectForAlignment = ({ alignedRect }) => alignedRect) {
+  const faceBoxes = parentResults.map((parentResult) => isWithFaceLandmarks(parentResult) ? getRectForAlignment(parentResult) : parentResult.detection);
+  const faces = extractedFaces || (input instanceof tfjs_esm_exports.Tensor ? await extractFaceTensors(input, faceBoxes) : await extractFaces(input, faceBoxes));
+  const results = await computeResults(faces);
+  faces.forEach((f) => f instanceof tfjs_esm_exports.Tensor && f.dispose());
+  return results;
+}
+async function extractSingleFaceAndComputeResult(parentResult, input, computeResult, extractedFaces, getRectForAlignment) {
+  return extractAllFacesAndComputeResults(
+    [parentResult],
+    input,
+    async (faces) => computeResult(faces[0]),
+    extractedFaces,
+    getRectForAlignment
+  );
+}
+
+// src/tinyFaceDetector/const.ts
+var IOU_THRESHOLD2 = 0.4;
+var BOX_ANCHORS2 = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB = [117.001, 114.697, 97.404];
+
+// src/tinyFaceDetector/TinyFaceDetector.ts
+var TinyFaceDetector = class extends TinyYolov2Base {
+  constructor() {
+    const config = {
+      withSeparableConvs: true,
+      iouThreshold: IOU_THRESHOLD2,
+      classes: ["face"],
+      anchors: BOX_ANCHORS2,
+      meanRgb: MEAN_RGB,
+      isFirstLayerConv2d: true,
+      filterSizes: [3, 16, 32, 64, 128, 256, 512]
+    };
+    super(config);
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return "tiny_face_detector_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/globalApi/nets.ts
+var nets = {
+  ssdMobilenetv1: new SsdMobilenetv1(),
+  tinyFaceDetector: new TinyFaceDetector(),
+  tinyYolov2: new TinyYolov2(),
+  faceLandmark68Net: new FaceLandmark68Net(),
+  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),
+  faceRecognitionNet: new FaceRecognitionNet(),
+  faceExpressionNet: new FaceExpressionNet(),
+  ageGenderNet: new AgeGenderNet()
+};
+var ssdMobilenetv1 = (input, options) => nets.ssdMobilenetv1.locateFaces(input, options);
+var tinyFaceDetector = (input, options) => nets.tinyFaceDetector.locateFaces(input, options);
+var tinyYolov2 = (input, options) => nets.tinyYolov2.locateFaces(input, options);
+var detectFaceLandmarks = (input) => nets.faceLandmark68Net.detectLandmarks(input);
+var detectFaceLandmarksTiny = (input) => nets.faceLandmark68TinyNet.detectLandmarks(input);
+var computeFaceDescriptor = (input) => nets.faceRecognitionNet.computeFaceDescriptor(input);
+var recognizeFaceExpressions = (input) => nets.faceExpressionNet.predictExpressions(input);
+var predictAgeAndGender = (input) => nets.ageGenderNet.predictAgeAndGender(input);
+var loadSsdMobilenetv1Model = (url) => nets.ssdMobilenetv1.load(url);
+var loadTinyFaceDetectorModel = (url) => nets.tinyFaceDetector.load(url);
+var loadTinyYolov2Model = (url) => nets.tinyYolov2.load(url);
+var loadFaceLandmarkModel = (url) => nets.faceLandmark68Net.load(url);
+var loadFaceLandmarkTinyModel = (url) => nets.faceLandmark68TinyNet.load(url);
+var loadFaceRecognitionModel = (url) => nets.faceRecognitionNet.load(url);
+var loadFaceExpressionModel = (url) => nets.faceExpressionNet.load(url);
+var loadAgeGenderModel = (url) => nets.ageGenderNet.load(url);
+var loadFaceDetectionModel = loadSsdMobilenetv1Model;
+var locateFaces = ssdMobilenetv1;
+var detectLandmarks = detectFaceLandmarks;
+
+// src/globalApi/PredictFaceExpressionsTask.ts
+var PredictFaceExpressionsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const faceExpressionsByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(
+        faces.map((face) => nets.faceExpressionNet.predictExpressions(face))
+      ),
+      this.extractedFaces
+    );
+    return parentResults.map(
+      (parentResult, i) => extendWithFaceExpressions(parentResult, faceExpressionsByFace[i])
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const faceExpressions = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceExpressionNet.predictExpressions(face),
+      this.extractedFaces
+    );
+    return extendWithFaceExpressions(parentResult, faceExpressions);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictAllFaceExpressionsWithFaceAlignmentTask = class extends PredictAllFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsWithFaceAlignmentTask = class extends PredictSingleFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/PredictAgeAndGenderTask.ts
+var PredictAgeAndGenderTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const ageAndGenderByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face))),
+      this.extractedFaces
+    );
+    return parentResults.map((parentResult, i) => {
+      const { age, gender, genderProbability } = ageAndGenderByFace[i];
+      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+    });
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.ageGenderNet.predictAgeAndGender(face),
+      this.extractedFaces
+    );
+    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictAllAgeAndGenderWithFaceAlignmentTask = class extends PredictAllAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderWithFaceAlignmentTask = class extends PredictSingleAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/ComputeFaceDescriptorsTasks.ts
+var ComputeFaceDescriptorsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+  }
+};
+var ComputeAllFaceDescriptorsTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const descriptors = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face))),
+      null,
+      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return descriptors.map((descriptor, i) => extendWithFaceDescriptor(parentResults[i], descriptor));
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+var ComputeSingleFaceDescriptorTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const descriptor = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face),
+      null,
+      (parentResult2) => parentResult2.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return extendWithFaceDescriptor(parentResult, descriptor);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var DetectFaceLandmarksTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, useTinyLandmarkNet) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.useTinyLandmarkNet = useTinyLandmarkNet;
+  }
+  get landmarkNet() {
+    return this.useTinyLandmarkNet ? nets.faceLandmark68TinyNet : nets.faceLandmark68Net;
+  }
+};
+var DetectAllFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const detections = parentResults.map((res) => res.detection);
+    const faces = this.input instanceof tfjs_esm_exports.Tensor ? await extractFaceTensors(this.input, detections) : await extractFaces(this.input, detections);
+    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face)));
+    faces.forEach((f) => f instanceof tfjs_esm_exports.Tensor && f.dispose());
+    const result = parentResults.filter((_parentResult, i) => faceLandmarksByFace[i]).map((parentResult, i) => extendWithFaceLandmarks(parentResult, faceLandmarksByFace[i]));
+    return result;
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var DetectSingleFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const { detection } = parentResult;
+    const faces = this.input instanceof tfjs_esm_exports.Tensor ? await extractFaceTensors(this.input, [detection]) : await extractFaces(this.input, [detection]);
+    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]);
+    faces.forEach((f) => f instanceof tfjs_esm_exports.Tensor && f.dispose());
+    return extendWithFaceLandmarks(parentResult, landmarks);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFacesTasks.ts
+var DetectFacesTaskBase = class extends ComposableTask {
+  constructor(input, options = new SsdMobilenetv1Options()) {
+    super();
+    this.input = input;
+    this.options = options;
+  }
+};
+var DetectAllFacesTask = class extends DetectFacesTaskBase {
+  async run() {
+    const { input, options } = this;
+    let result;
+    if (options instanceof TinyFaceDetectorOptions)
+      result = nets.tinyFaceDetector.locateFaces(input, options);
+    else if (options instanceof SsdMobilenetv1Options)
+      result = nets.ssdMobilenetv1.locateFaces(input, options);
+    else if (options instanceof TinyYolov2Options)
+      result = nets.tinyYolov2.locateFaces(input, options);
+    else
+      throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+    return result;
+  }
+  runAndExtendWithFaceDetections() {
+    return new Promise((resolve, reject) => {
+      this.run().then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection)))).catch((err) => reject(err));
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectAllFaceLandmarksTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+};
+var DetectSingleFaceTask = class extends DetectFacesTaskBase {
+  async run() {
+    const faceDetections = await new DetectAllFacesTask(this.input, this.options);
+    let faceDetectionWithHighestScore = faceDetections[0];
+    faceDetections.forEach((faceDetection) => {
+      if (faceDetection.score > faceDetectionWithHighestScore.score)
+        faceDetectionWithHighestScore = faceDetection;
+    });
+    return faceDetectionWithHighestScore;
+  }
+  runAndExtendWithFaceDetection() {
+    return new Promise(async (resolve) => {
+      const detection = await this.run();
+      resolve(detection ? extendWithFaceDetection({}, detection) : void 0);
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectSingleFaceLandmarksTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+};
+
+// src/globalApi/detectFaces.ts
+function detectSingleFace(input, options = new SsdMobilenetv1Options()) {
+  return new DetectSingleFaceTask(input, options);
+}
+function detectAllFaces(input, options = new SsdMobilenetv1Options()) {
+  return new DetectAllFacesTask(input, options);
+}
+
+// src/globalApi/allFaces.ts
+async function allFacesSsdMobilenetv1(input, minConfidence) {
+  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {})).withFaceLandmarks().withFaceDescriptors();
+}
+async function allFacesTinyYolov2(input, forwardParams = {}) {
+  return detectAllFaces(input, new TinyYolov2Options(forwardParams)).withFaceLandmarks().withFaceDescriptors();
+}
+var allFaces = allFacesSsdMobilenetv1;
+
+// src/euclideanDistance.ts
+function euclideanDistance(arr1, arr2) {
+  if (arr1.length !== arr2.length)
+    throw new Error("euclideanDistance: arr1.length !== arr2.length");
+  const desc1 = Array.from(arr1);
+  const desc2 = Array.from(arr2);
+  return Math.sqrt(
+    desc1.map((val, i) => val - desc2[i]).reduce((res, diff) => res + diff * diff, 0)
+  );
+}
+
+// src/globalApi/FaceMatcher.ts
+var FaceMatcher = class {
+  constructor(inputs, distanceThreshold = 0.6) {
+    this._distanceThreshold = distanceThreshold;
+    const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+    if (!inputArray.length)
+      throw new Error("FaceRecognizer.constructor - expected atleast one input");
+    let count = 1;
+    const createUniqueLabel = () => `person ${count++}`;
+    this._labeledDescriptors = inputArray.map((desc) => {
+      if (desc instanceof LabeledFaceDescriptors)
+        return desc;
+      if (desc instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);
+      if (desc.descriptor && desc.descriptor instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);
+      throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>");
+    });
+  }
+  get labeledDescriptors() {
+    return this._labeledDescriptors;
+  }
+  get distanceThreshold() {
+    return this._distanceThreshold;
+  }
+  computeMeanDistance(queryDescriptor, descriptors) {
+    return descriptors.map((d) => euclideanDistance(d, queryDescriptor)).reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);
+  }
+  matchDescriptor(queryDescriptor) {
+    return this.labeledDescriptors.map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors))).reduce((best, curr) => best.distance < curr.distance ? best : curr);
+  }
+  findBestMatch(queryDescriptor) {
+    const bestMatch = this.matchDescriptor(queryDescriptor);
+    return bestMatch.distance < this._distanceThreshold ? bestMatch : new FaceMatch("unknown", bestMatch.distance);
+  }
+  toJSON() {
+    return {
+      distanceThreshold: this._distanceThreshold,
+      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON())
+    };
+  }
+  static fromJSON(json) {
+    const labeledDescriptors = json.labeledDescriptors.map((ld) => LabeledFaceDescriptors.fromJSON(ld));
+    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);
+  }
+};
+
+// src/tinyFaceDetector/index.ts
+function createTinyFaceDetector(weights) {
+  const net = new TinyFaceDetector();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/resizeResults.ts
+function resizeResults(results, dimensions) {
+  const { width, height } = new Dimensions(dimensions.width, dimensions.height);
+  if (width <= 0 || height <= 0) {
+    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);
+  }
+  if (Array.isArray(results)) {
+    return results.map((obj) => resizeResults(obj, { width, height }));
+  }
+  if (isWithFaceLandmarks(results)) {
+    const resizedDetection = results.detection.forSize(width, height);
+    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);
+    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);
+  }
+  if (isWithFaceDetection(results)) {
+    return extendWithFaceDetection(results, results.detection.forSize(width, height));
+  }
+  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {
+    return results.forSize(width, height);
+  }
+  return results;
+}
+
+// src/index.ts
+var version8 = version7;
+export {
+  AgeGenderNet,
+  BoundingBox,
+  Box,
+  ComposableTask,
+  ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask,
+  DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask,
+  Dimensions,
+  FACE_EXPRESSION_LABELS,
+  FaceDetection,
+  FaceDetectionNet,
+  FaceExpressionNet,
+  FaceExpressions,
+  FaceLandmark68Net,
+  FaceLandmark68TinyNet,
+  FaceLandmarkNet,
+  FaceLandmarks,
+  FaceLandmarks5,
+  FaceLandmarks68,
+  FaceMatch,
+  FaceMatcher,
+  FaceRecognitionNet,
+  Gender,
+  LabeledBox,
+  LabeledFaceDescriptors,
+  NetInput,
+  NeuralNetwork,
+  ObjectDetection,
+  Point,
+  PredictedBox,
+  Rect,
+  SsdMobilenetv1,
+  SsdMobilenetv1Options,
+  TinyFaceDetector,
+  TinyFaceDetectorOptions,
+  TinyYolov2,
+  TinyYolov2Options,
+  allFaces,
+  allFacesSsdMobilenetv1,
+  allFacesTinyYolov2,
+  awaitMediaLoaded,
+  bufferToImage,
+  computeFaceDescriptor,
+  createCanvas,
+  createCanvasFromMedia,
+  createFaceDetectionNet,
+  createFaceRecognitionNet,
+  createSsdMobilenetv1,
+  createTinyFaceDetector,
+  createTinyYolov2,
+  detectAllFaces,
+  detectFaceLandmarks,
+  detectFaceLandmarksTiny,
+  detectLandmarks,
+  detectSingleFace,
+  draw_exports as draw,
+  env,
+  euclideanDistance,
+  extendWithAge,
+  extendWithFaceDescriptor,
+  extendWithFaceDetection,
+  extendWithFaceExpressions,
+  extendWithFaceLandmarks,
+  extendWithGender,
+  extractFaceTensors,
+  extractFaces,
+  fetchImage,
+  fetchJson,
+  fetchNetWeights,
+  fetchOrThrow,
+  fetchVideo,
+  getContext2dOrThrow,
+  getMediaDimensions,
+  imageTensorToCanvas,
+  imageToSquare,
+  inverseSigmoid,
+  iou,
+  isMediaElement,
+  isMediaLoaded,
+  isWithAge,
+  isWithFaceDetection,
+  isWithFaceExpressions,
+  isWithFaceLandmarks,
+  isWithGender,
+  loadAgeGenderModel,
+  loadFaceDetectionModel,
+  loadFaceExpressionModel,
+  loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel,
+  loadTinyYolov2Model,
+  loadWeightMap,
+  locateFaces,
+  matchDimensions,
+  minBbox,
+  nets,
+  nonMaxSuppression,
+  normalize,
+  padToSquare,
+  predictAgeAndGender,
+  recognizeFaceExpressions,
+  resizeResults,
+  resolveInput,
+  shuffleArray,
+  sigmoid,
+  ssdMobilenetv1,
+  tfjs_esm_exports as tf,
+  tinyFaceDetector,
+  tinyYolov2,
+  toNetInput,
+  utils_exports as utils,
+  validateConfig,
+  version8 as version
+};
diff --git a/dist/face-api.esm.js b/dist/face-api.esm.js
index 0e0af73..f8048dd 100644
--- a/dist/face-api.esm.js
+++ b/dist/face-api.esm.js
@@ -4,65 +4,48120 @@
   author: <https://github.com/vladmandic>'
 */
 
-var aF=Object.defineProperty;var Gr=(e=>typeof require!="undefined"?require:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof require!="undefined"?require:t)[n]}):e)(function(e){if(typeof require!="undefined")return require.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')});var ab=(e,t)=>{for(var n in t)aF(e,n,{get:t[n],enumerable:!0})};var Oe={};ab(Oe,{Abs:()=>wl,Acos:()=>kl,Acosh:()=>Il,AdadeltaOptimizer:()=>of,AdagradOptimizer:()=>lf,AdamOptimizer:()=>uf,AdamaxOptimizer:()=>pf,Add:()=>cs,AddN:()=>mi,All:()=>Sl,Any:()=>Tl,ArgMax:()=>fi,ArgMin:()=>sc,Asin:()=>Nl,Asinh:()=>Cl,Atan:()=>_l,Atan2:()=>Al,Atanh:()=>El,AvgPool:()=>gi,AvgPool3D:()=>ic,AvgPool3DGrad:()=>Yh,AvgPoolGrad:()=>Xh,BackendWasm:()=>kA,BatchMatMul:()=>yi,BatchToSpaceND:()=>$l,Bincount:()=>Zh,BroadcastArgs:()=>Jh,BroadcastTo:()=>RI,Callback:()=>JN,CallbackList:()=>nN,Cast:()=>bi,Ceil:()=>xi,ClipByValue:()=>ds,Complex:()=>Qh,ComplexAbs:()=>oc,Concat:()=>Fl,Conv2D:()=>vi,Conv2DBackpropFilter:()=>em,Conv2DBackpropInput:()=>wi,Conv3D:()=>lc,Conv3DBackpropFilterV2:()=>tm,Conv3DBackpropInputV2:()=>nm,Cos:()=>ki,Cosh:()=>Ii,CropAndResize:()=>Rl,Cumprod:()=>Dl,Cumsum:()=>Si,CustomCallback:()=>rN,DataStorage:()=>jh,DenseBincount:()=>am,DepthToSpace:()=>Ml,DepthwiseConv2dNative:()=>Ti,DepthwiseConv2dNativeBackpropFilter:()=>rm,DepthwiseConv2dNativeBackpropInput:()=>sm,Diag:()=>im,Dilation2D:()=>uc,Dilation2DBackpropFilter:()=>gh,Dilation2DBackpropInput:()=>fh,ENV:()=>Ix,EarlyStopping:()=>QN,Einsum:()=>om,Elu:()=>Ci,EluGrad:()=>lm,Environment:()=>FI,Equal:()=>Ol,Erf:()=>Pl,Exp:()=>_i,ExpandDims:()=>Ll,Expm1:()=>zl,FFT:()=>um,Fill:()=>pc,FlipLeftRight:()=>Wl,Floor:()=>Ei,FloorDiv:()=>Ai,FromPixels:()=>yh,FusedBatchNorm:()=>$i,FusedConv2D:()=>Qs,FusedDepthwiseConv2D:()=>ei,GPGPUContext:()=>ph,GatherNd:()=>Vl,GatherV2:()=>Bl,GraphModel:()=>k0,Greater:()=>Ul,GreaterEqual:()=>Fi,History:()=>aN,IFFT:()=>pm,Identity:()=>Di,Imag:()=>cm,InputSpec:()=>zt,IsFinite:()=>Gl,IsInf:()=>Hl,IsNan:()=>jl,KernelBackend:()=>rc,LRN:()=>cc,LRNGrad:()=>hm,LayerVariable:()=>XT,LayersModel:()=>Nr,LeakyRelu:()=>Ri,Less:()=>ql,LessEqual:()=>Kl,LinSpace:()=>dm,Log:()=>Mi,Log1p:()=>Xl,LogSoftmax:()=>PI,LogicalAnd:()=>Yl,LogicalNot:()=>Zl,LogicalOr:()=>Jl,LogicalXor:()=>MI,LowerBound:()=>qF,MathBackendWebGL:()=>Rf,Max:()=>Pi,MaxPool:()=>Li,MaxPool3D:()=>dc,MaxPool3DGrad:()=>fm,MaxPoolGrad:()=>mm,MaxPoolWithArgmax:()=>gm,Maximum:()=>Oi,Mean:()=>zi,Min:()=>Wi,Minimum:()=>Bi,MirrorPad:()=>Vi,Mod:()=>Ql,MomentumOptimizer:()=>cf,Multinomial:()=>ym,Multiply:()=>Ui,Neg:()=>eu,NonMaxSuppressionV3:()=>nu,NonMaxSuppressionV4:()=>au,NonMaxSuppressionV5:()=>ru,NotEqual:()=>tu,OP_SCOPE_SUFFIX:()=>Cx,OneHot:()=>Gi,OnesLike:()=>su,Optimizer:()=>$r,OptimizerConstructors:()=>Hr,Pack:()=>iu,PadV2:()=>Hi,Pool:()=>KF,Pow:()=>ji,Prelu:()=>qi,Prod:()=>Ki,RMSPropOptimizer:()=>df,RNN:()=>dr,RaggedGather:()=>bm,RaggedRange:()=>xm,RaggedTensorToTensor:()=>vm,Range:()=>hc,Rank:()=>wb,Real:()=>wm,RealDiv:()=>Ni,Reciprocal:()=>ou,Reduction:()=>xn,Relu:()=>Xi,Relu6:()=>Ji,Reshape:()=>lu,ResizeBilinear:()=>Zi,ResizeBilinearGrad:()=>Im,ResizeNearestNeighbor:()=>Yi,ResizeNearestNeighborGrad:()=>km,Reverse:()=>Qi,RotateWithOffset:()=>Su,Round:()=>eo,Rsqrt:()=>to,SGDOptimizer:()=>Bc,ScatterNd:()=>uu,SearchSorted:()=>Sm,Select:()=>pu,Selu:()=>cu,Sequential:()=>hl,Sigmoid:()=>ao,Sign:()=>mu,Sin:()=>no,Sinh:()=>hu,Slice:()=>du,Softmax:()=>io,Softplus:()=>fu,SpaceToBatchND:()=>gu,SparseFillEmptyRows:()=>mc,SparseReshape:()=>bu,SparseSegmentMean:()=>fc,SparseSegmentSum:()=>gc,SparseToDense:()=>Tm,SplitV:()=>yu,Sqrt:()=>ro,Square:()=>yc,SquaredDifference:()=>oo,Step:()=>ms,StridedSlice:()=>xu,StringNGrams:()=>bc,StringSplit:()=>xc,StringToHashBucketFast:()=>vc,Sub:()=>lo,Sum:()=>so,SymbolicTensor:()=>Ba,Tan:()=>uo,Tanh:()=>po,Tensor:()=>Te,TensorBuffer:()=>Ht,Tile:()=>hs,TopK:()=>vu,Transform:()=>wu,Transpose:()=>Tr,Unique:()=>Nm,Unpack:()=>ku,UnsortedSegmentSum:()=>wc,UpperBound:()=>XF,Variable:()=>ts,ZerosLike:()=>Iu,_FusedMatMul:()=>Js,abs:()=>Lt,acos:()=>Bx,acosh:()=>Vx,add:()=>Y,addN:()=>CS,all:()=>$m,any:()=>qp,argMax:()=>ai,argMin:()=>Ux,asin:()=>Gx,asinh:()=>Hx,atan:()=>jx,atan2:()=>qx,atanh:()=>Kx,avgPool:()=>ga,avgPool3d:()=>Yx,backend:()=>uS,backend_util:()=>N,basicLSTMCell:()=>$S,batchNorm:()=>ys,batchNorm2d:()=>Zx,batchNorm3d:()=>Jx,batchNorm4d:()=>Qx,batchToSpaceND:()=>Ac,bincount:()=>ev,booleanMaskAsync:()=>hT,broadcastArgs:()=>FS,broadcastTo:()=>Ks,broadcast_util:()=>Tu,browser:()=>co,buffer:()=>Pe,callbacks:()=>bH,cast:()=>oe,ceil:()=>tv,clipByValue:()=>en,clone:()=>rr,complex:()=>Cr,concat:()=>Ze,concat1d:()=>nv,concat2d:()=>av,concat3d:()=>rv,concat4d:()=>sv,constraints:()=>JT,conv1d:()=>Fm,conv2d:()=>$t,conv2dTranspose:()=>Dm,conv3d:()=>ov,conv3dTranspose:()=>lv,copyRegisteredKernels:()=>QF,cos:()=>$c,cosh:()=>Rm,cosineWindow:()=>tf,cumprod:()=>Kp,cumsum:()=>Mm,customGrad:()=>or,data:()=>I2,denseBincount:()=>Sh,deprecationWarn:()=>Rx,depthToSpace:()=>uv,depthwiseConv2d:()=>bs,deregisterOp:()=>wH,device_util:()=>Tc,diag:()=>RS,dilation2d:()=>pv,disableDeprecationWarnings:()=>CR,dispose:()=>_e,disposeVariables:()=>_R,div:()=>he,divNoNan:()=>cv,dot:()=>dv,dropout:()=>zv,einsum:()=>MS,elu:()=>Nu,enableDebugMode:()=>NR,enableProdMode:()=>TR,enclosingPowerOfTwo:()=>Wv,engine:()=>Na,env:()=>H,equal:()=>Qn,erf:()=>hv,euclideanNorm:()=>gv,exp:()=>fn,expandDims:()=>Zt,expm1:()=>yv,eye:()=>Pm,fft:()=>zc,fill:()=>gn,findBackend:()=>MR,findBackendFactory:()=>PR,floor:()=>_u,floorDiv:()=>Am,forceHalfFloat:()=>V_,fused:()=>pl,gather:()=>Eu,gatherND:()=>yT,gather_util:()=>Mx,getBackend:()=>DR,getGradient:()=>xb,getKernel:()=>bh,getKernelsForBackend:()=>xh,getThreadsCount:()=>mpe,gpgpu_util:()=>v_,grad:()=>iO,grads:()=>oO,greater:()=>Nn,greaterEqual:()=>Er,ifft:()=>ul,imag:()=>Cc,image:()=>za,inTopKAsync:()=>bT,initializers:()=>QT,input:()=>bN,io:()=>Ut,irfft:()=>Zm,isFinite:()=>bv,isInf:()=>xv,isNaN:()=>vv,keep:()=>Jt,kernel_impls:()=>cr,layers:()=>eN,leakyRelu:()=>Fc,less:()=>Om,lessEqual:()=>xs,linalg:()=>Uv,linspace:()=>WS,loadGraphModel:()=>S6,loadGraphModelSync:()=>T6,loadLayersModel:()=>xU,localResponseNormalization:()=>wv,log:()=>ea,log1p:()=>Dc,logSigmoid:()=>kv,logSoftmax:()=>zm,logSumExp:()=>Wm,logicalAnd:()=>Ea,logicalNot:()=>Rc,logicalOr:()=>Bm,logicalXor:()=>Iv,losses:()=>AT,lowerBound:()=>VS,matMul:()=>Fe,math:()=>lS,max:()=>ha,maxPool:()=>Dt,maxPool3d:()=>Sv,maxPoolWithArgmax:()=>US,maximum:()=>pr,mean:()=>Nt,memory:()=>kh,meshgrid:()=>GS,metrics:()=>XN,min:()=>il,minimum:()=>Au,mirrorPad:()=>Tv,mod:()=>Nv,model:()=>kU,models:()=>YN,moments:()=>Mc,movingAverage:()=>mT,mul:()=>z,multiRNNCell:()=>HS,multinomial:()=>jS,neg:()=>bt,nextFrame:()=>Gv,norm:()=>Cu,notEqual:()=>ii,oneHot:()=>rl,ones:()=>Zn,onesLike:()=>ta,op:()=>L,outerProduct:()=>qS,pad:()=>ya,pad1d:()=>KS,pad2d:()=>XS,pad3d:()=>YS,pad4d:()=>ZS,pool:()=>Cv,pow:()=>_r,prelu:()=>Oc,print:()=>$x,prod:()=>_v,profile:()=>ER,raggedGather:()=>JS,raggedRange:()=>QS,raggedTensorToTensor:()=>eT,rand:()=>tT,randomGamma:()=>nT,randomNormal:()=>Um,randomStandardNormal:()=>aT,randomUniform:()=>$u,range:()=>ol,ready:()=>FR,real:()=>sl,reciprocal:()=>$v,registerBackend:()=>Em,registerCallbackConstructor:()=>SU,registerGradient:()=>OI,registerKernel:()=>kc,registerOp:()=>vH,regularizers:()=>ZN,relu:()=>Xe,relu6:()=>Gm,removeBackend:()=>RR,reshape:()=>W,reverse:()=>fa,reverse1d:()=>rT,reverse2d:()=>sT,reverse3d:()=>iT,reverse4d:()=>oT,rfft:()=>Wc,round:()=>Hm,rsqrt:()=>jm,scalar:()=>be,scatterND:()=>fT,scatter_util:()=>Px,searchSorted:()=>Vm,selu:()=>qm,separableConv2d:()=>vs,sequential:()=>IU,serialization:()=>ne,setBackend:()=>$R,setPlatform:()=>OR,setThreadsCount:()=>hpe,setWasmPath:()=>cpe,setWasmPaths:()=>dpe,setWebGLContext:()=>GC,setdiff1dAsync:()=>lT,sigmoid:()=>da,sign:()=>Fv,signal:()=>ET,sin:()=>Km,sinh:()=>Xm,slice:()=>Be,slice1d:()=>Lc,slice2d:()=>Ym,slice3d:()=>mo,slice4d:()=>ll,slice_util:()=>jt,softmax:()=>Ka,softplus:()=>ho,spaceToBatchND:()=>Pc,sparse:()=>$T,sparseToDense:()=>gT,spectral:()=>_T,split:()=>zn,sqrt:()=>ln,square:()=>ot,squaredDifference:()=>Jm,squeeze:()=>ws,stack:()=>Ft,step:()=>fo,stridedSlice:()=>Dv,string:()=>FT,sub:()=>pe,sum:()=>fe,sumOutType:()=>_m,tan:()=>Rv,tanh:()=>ri,tensor:()=>kn,tensor1d:()=>Ke,tensor2d:()=>_a,tensor3d:()=>_c,tensor4d:()=>$a,tensor5d:()=>uT,tensor6d:()=>pT,tensor_util:()=>Va,test_util:()=>SS,tidy:()=>P,tile:()=>Ln,time:()=>AR,topk:()=>Mv,train:()=>zs,transpose:()=>Ee,truncatedNormal:()=>Qm,unique:()=>Pv,unregisterGradient:()=>JF,unregisterKernel:()=>ZF,unsortedSegmentSum:()=>ef,unstack:()=>ct,upcastType:()=>ma,upperBound:()=>cT,util:()=>v,valueAndGrad:()=>lO,valueAndGrads:()=>uO,variable:()=>Ov,variableGrads:()=>BS,version:()=>wpe,version_converter:()=>C6,version_core:()=>fM,version_layers:()=>dw,version_wasm:()=>fpe,version_webgl:()=>rJ,webgl:()=>sJ,webgl_util:()=>UC,where:()=>mn,whereAsync:()=>Lv,zeros:()=>It,zerosLike:()=>qe});var rF=Object.create,bx=Object.defineProperty,sF=Object.getOwnPropertyDescriptor,iF=Object.getOwnPropertyNames,oF=Object.getPrototypeOf,lF=Object.prototype.hasOwnProperty,vI=(e=>typeof Gr!="undefined"?Gr:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof Gr!="undefined"?Gr:t)[n]}):e)(function(e){if(typeof Gr!="undefined")return Gr.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')}),Bt=(e,t)=>()=>(t||e((t={exports:{}}).exports,t),t.exports),Ae=(e,t)=>{for(var n in t)bx(e,n,{get:t[n],enumerable:!0})},uF=(e,t,n,a)=>{if(t&&typeof t=="object"||typeof t=="function")for(let r of iF(t))!lF.call(e,r)&&r!==n&&bx(e,r,{get:()=>t[r],enumerable:!(a=sF(t,r))||a.enumerable});return e},us=(e,t,n)=>(n=e!=null?rF(oF(e)):{},uF(t||!e||!e.__esModule?bx(n,"default",{value:e,enumerable:!0}):n,e)),pF=Bt((e,t)=>{t.exports=a;var n=null;try{n=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(S){}function a(S,M,B){this.low=S|0,this.high=M|0,this.unsigned=!!B}a.prototype.__isLong__,Object.defineProperty(a.prototype,"__isLong__",{value:!0});function r(S){return(S&&S.__isLong__)===!0}a.isLong=r;var s={},i={};function o(S,M){var B,U,G;return M?(S>>>=0,(G=0<=S&&S<256)&&(U=i[S],U)?U:(B=u(S,(S|0)<0?-1:0,!0),G&&(i[S]=B),B)):(S|=0,(G=-128<=S&&S<128)&&(U=s[S],U)?U:(B=u(S,S<0?-1:0,!1),G&&(s[S]=B),B))}a.fromInt=o;function l(S,M){if(isNaN(S))return M?w:x;if(M){if(S<0)return w;if(S>=g)return A}else{if(S<=-y)return R;if(S+1>=y)return E}return S<0?l(-S,M).neg():u(S%f|0,S/f|0,M)}a.fromNumber=l;function u(S,M,B){return new a(S,M,B)}a.fromBits=u;var p=Math.pow;function d(S,M,B){if(S.length===0)throw Error("empty string");if(S==="NaN"||S==="Infinity"||S==="+Infinity"||S==="-Infinity")return x;if(typeof M=="number"?(B=M,M=!1):M=!!M,B=B||10,B<2||36<B)throw RangeError("radix");var U;if((U=S.indexOf("-"))>0)throw Error("interior hyphen");if(U===0)return d(S.substring(1),M,B).neg();for(var G=l(p(B,8)),q=x,K=0;K<S.length;K+=8){var Z=Math.min(8,S.length-K),Q=parseInt(S.substring(K,K+Z),B);if(Z<8){var ee=l(p(B,Z));q=q.mul(ee).add(l(Q))}else q=q.mul(G),q=q.add(l(Q))}return q.unsigned=M,q}a.fromString=d;function c(S,M){return typeof S=="number"?l(S,M):typeof S=="string"?d(S,M):u(S.low,S.high,typeof M=="boolean"?M:S.unsigned)}a.fromValue=c;var h=1<<16,m=1<<24,f=h*h,g=f*f,y=g/2,b=o(m),x=o(0);a.ZERO=x;var w=o(0,!0);a.UZERO=w;var I=o(1);a.ONE=I;var T=o(1,!0);a.UONE=T;var C=o(-1);a.NEG_ONE=C;var E=u(-1,2147483647,!1);a.MAX_VALUE=E;var A=u(-1,-1,!0);a.MAX_UNSIGNED_VALUE=A;var R=u(0,-2147483648,!1);a.MIN_VALUE=R;var F=a.prototype;F.toInt=function(){return this.unsigned?this.low>>>0:this.low},F.toNumber=function(){return this.unsigned?(this.high>>>0)*f+(this.low>>>0):this.high*f+(this.low>>>0)},F.toString=function(S){if(S=S||10,S<2||36<S)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative())if(this.eq(R)){var M=l(S),B=this.div(M),U=B.mul(M).sub(this);return B.toString(S)+U.toInt().toString(S)}else return"-"+this.neg().toString(S);for(var G=l(p(S,6),this.unsigned),q=this,K="";;){var Z=q.div(G),Q=q.sub(Z.mul(G)).toInt()>>>0,ee=Q.toString(S);if(q=Z,q.isZero())return ee+K;for(;ee.length<6;)ee="0"+ee;K=""+ee+K}},F.getHighBits=function(){return this.high},F.getHighBitsUnsigned=function(){return this.high>>>0},F.getLowBits=function(){return this.low},F.getLowBitsUnsigned=function(){return this.low>>>0},F.getNumBitsAbs=function(){if(this.isNegative())return this.eq(R)?64:this.neg().getNumBitsAbs();for(var S=this.high!=0?this.high:this.low,M=31;M>0&&(S&1<<M)==0;M--);return this.high!=0?M+33:M+1},F.isZero=function(){return this.high===0&&this.low===0},F.eqz=F.isZero,F.isNegative=function(){return!this.unsigned&&this.high<0},F.isPositive=function(){return this.unsigned||this.high>=0},F.isOdd=function(){return(this.low&1)===1},F.isEven=function(){return(this.low&1)===0},F.equals=function(S){return r(S)||(S=c(S)),this.unsigned!==S.unsigned&&this.high>>>31===1&&S.high>>>31===1?!1:this.high===S.high&&this.low===S.low},F.eq=F.equals,F.notEquals=function(S){return!this.eq(S)},F.neq=F.notEquals,F.ne=F.notEquals,F.lessThan=function(S){return this.comp(S)<0},F.lt=F.lessThan,F.lessThanOrEqual=function(S){return this.comp(S)<=0},F.lte=F.lessThanOrEqual,F.le=F.lessThanOrEqual,F.greaterThan=function(S){return this.comp(S)>0},F.gt=F.greaterThan,F.greaterThanOrEqual=function(S){return this.comp(S)>=0},F.gte=F.greaterThanOrEqual,F.ge=F.greaterThanOrEqual,F.compare=function(S){if(r(S)||(S=c(S)),this.eq(S))return 0;var M=this.isNegative(),B=S.isNegative();return M&&!B?-1:!M&&B?1:this.unsigned?S.high>>>0>this.high>>>0||S.high===this.high&&S.low>>>0>this.low>>>0?-1:1:this.sub(S).isNegative()?-1:1},F.comp=F.compare,F.negate=function(){return!this.unsigned&&this.eq(R)?R:this.not().add(I)},F.neg=F.negate,F.add=function(S){r(S)||(S=c(S));var M=this.high>>>16,B=this.high&65535,U=this.low>>>16,G=this.low&65535,q=S.high>>>16,K=S.high&65535,Z=S.low>>>16,Q=S.low&65535,ee=0,ae=0,te=0,le=0;return le+=G+Q,te+=le>>>16,le&=65535,te+=U+Z,ae+=te>>>16,te&=65535,ae+=B+K,ee+=ae>>>16,ae&=65535,ee+=M+q,ee&=65535,u(te<<16|le,ee<<16|ae,this.unsigned)},F.subtract=function(S){return r(S)||(S=c(S)),this.add(S.neg())},F.sub=F.subtract,F.multiply=function(S){if(this.isZero())return x;if(r(S)||(S=c(S)),n){var M=n.mul(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}if(S.isZero())return x;if(this.eq(R))return S.isOdd()?R:x;if(S.eq(R))return this.isOdd()?R:x;if(this.isNegative())return S.isNegative()?this.neg().mul(S.neg()):this.neg().mul(S).neg();if(S.isNegative())return this.mul(S.neg()).neg();if(this.lt(b)&&S.lt(b))return l(this.toNumber()*S.toNumber(),this.unsigned);var B=this.high>>>16,U=this.high&65535,G=this.low>>>16,q=this.low&65535,K=S.high>>>16,Z=S.high&65535,Q=S.low>>>16,ee=S.low&65535,ae=0,te=0,le=0,ie=0;return ie+=q*ee,le+=ie>>>16,ie&=65535,le+=G*ee,te+=le>>>16,le&=65535,le+=q*Q,te+=le>>>16,le&=65535,te+=U*ee,ae+=te>>>16,te&=65535,te+=G*Q,ae+=te>>>16,te&=65535,te+=q*Z,ae+=te>>>16,te&=65535,ae+=B*ee+U*Q+G*Z+q*K,ae&=65535,u(le<<16|ie,ae<<16|te,this.unsigned)},F.mul=F.multiply,F.divide=function(S){if(r(S)||(S=c(S)),S.isZero())throw Error("division by zero");if(n){if(!this.unsigned&&this.high===-2147483648&&S.low===-1&&S.high===-1)return this;var M=(this.unsigned?n.div_u:n.div_s)(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?w:x;var B,U,G;if(this.unsigned){if(S.unsigned||(S=S.toUnsigned()),S.gt(this))return w;if(S.gt(this.shru(1)))return T;G=w}else{if(this.eq(R)){if(S.eq(I)||S.eq(C))return R;if(S.eq(R))return I;var q=this.shr(1);return B=q.div(S).shl(1),B.eq(x)?S.isNegative()?I:C:(U=this.sub(S.mul(B)),G=B.add(U.div(S)),G)}else if(S.eq(R))return this.unsigned?w:x;if(this.isNegative())return S.isNegative()?this.neg().div(S.neg()):this.neg().div(S).neg();if(S.isNegative())return this.div(S.neg()).neg();G=x}for(U=this;U.gte(S);){B=Math.max(1,Math.floor(U.toNumber()/S.toNumber()));for(var K=Math.ceil(Math.log(B)/Math.LN2),Z=K<=48?1:p(2,K-48),Q=l(B),ee=Q.mul(S);ee.isNegative()||ee.gt(U);)B-=Z,Q=l(B,this.unsigned),ee=Q.mul(S);Q.isZero()&&(Q=I),G=G.add(Q),U=U.sub(ee)}return G},F.div=F.divide,F.modulo=function(S){if(r(S)||(S=c(S)),n){var M=(this.unsigned?n.rem_u:n.rem_s)(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}return this.sub(this.div(S).mul(S))},F.mod=F.modulo,F.rem=F.modulo,F.not=function(){return u(~this.low,~this.high,this.unsigned)},F.and=function(S){return r(S)||(S=c(S)),u(this.low&S.low,this.high&S.high,this.unsigned)},F.or=function(S){return r(S)||(S=c(S)),u(this.low|S.low,this.high|S.high,this.unsigned)},F.xor=function(S){return r(S)||(S=c(S)),u(this.low^S.low,this.high^S.high,this.unsigned)},F.shiftLeft=function(S){return r(S)&&(S=S.toInt()),(S&=63)===0?this:S<32?u(this.low<<S,this.high<<S|this.low>>>32-S,this.unsigned):u(0,this.low<<S-32,this.unsigned)},F.shl=F.shiftLeft,F.shiftRight=function(S){return r(S)&&(S=S.toInt()),(S&=63)===0?this:S<32?u(this.low>>>S|this.high<<32-S,this.high>>S,this.unsigned):u(this.high>>S-32,this.high>=0?0:-1,this.unsigned)},F.shr=F.shiftRight,F.shiftRightUnsigned=function(S){if(r(S)&&(S=S.toInt()),S&=63,S===0)return this;var M=this.high;if(S<32){var B=this.low;return u(B>>>S|M<<32-S,M>>>S,this.unsigned)}else return S===32?u(M,0,this.unsigned):u(M>>>S-32,0,this.unsigned)},F.shru=F.shiftRightUnsigned,F.shr_u=F.shiftRightUnsigned,F.toSigned=function(){return this.unsigned?u(this.low,this.high,!1):this},F.toUnsigned=function(){return this.unsigned?this:u(this.low,this.high,!0)},F.toBytes=function(S){return S?this.toBytesLE():this.toBytesBE()},F.toBytesLE=function(){var S=this.high,M=this.low;return[M&255,M>>>8&255,M>>>16&255,M>>>24,S&255,S>>>8&255,S>>>16&255,S>>>24]},F.toBytesBE=function(){var S=this.high,M=this.low;return[S>>>24,S>>>16&255,S>>>8&255,S&255,M>>>24,M>>>16&255,M>>>8&255,M&255]},a.fromBytes=function(S,M,B){return B?a.fromBytesLE(S,M):a.fromBytesBE(S,M)},a.fromBytesLE=function(S,M){return new a(S[0]|S[1]<<8|S[2]<<16|S[3]<<24,S[4]|S[5]<<8|S[6]<<16|S[7]<<24,M)},a.fromBytesBE=function(S,M){return new a(S[4]<<24|S[5]<<16|S[6]<<8|S[7],S[0]<<24|S[1]<<16|S[2]<<8|S[3],M)}}),cF=Bt(()=>{}),dF=Bt(()=>{}),hF=Bt((e,t)=>{(function(n,a,r){function s(u){var p=this,d=l();p.next=function(){var c=2091639*p.s0+p.c*23283064365386963e-26;return p.s0=p.s1,p.s1=p.s2,p.s2=c-(p.c=c|0)},p.c=1,p.s0=d(" "),p.s1=d(" "),p.s2=d(" "),p.s0-=d(u),p.s0<0&&(p.s0+=1),p.s1-=d(u),p.s1<0&&(p.s1+=1),p.s2-=d(u),p.s2<0&&(p.s2+=1),d=null}function i(u,p){return p.c=u.c,p.s0=u.s0,p.s1=u.s1,p.s2=u.s2,p}function o(u,p){var d=new s(u),c=p&&p.state,h=d.next;return h.int32=function(){return d.next()*4294967296|0},h.double=function(){return h()+(h()*2097152|0)*11102230246251565e-32},h.quick=h,c&&(typeof c=="object"&&i(c,d),h.state=function(){return i(d,{})}),h}function l(){var u=4022871197,p=function(d){d=String(d);for(var c=0;c<d.length;c++){u+=d.charCodeAt(c);var h=.02519603282416938*u;u=h>>>0,h-=u,h*=u,u=h>>>0,h-=u,u+=h*4294967296}return(u>>>0)*23283064365386963e-26};return p}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.alea=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),mF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.x=0,u.y=0,u.z=0,u.w=0,u.next=function(){var c=u.x^u.x<<11;return u.x=u.y,u.y=u.z,u.z=u.w,u.w^=u.w>>>19^c^c>>>8},l===(l|0)?u.x=l:p+=l;for(var d=0;d<p.length+64;d++)u.x^=p.charCodeAt(d)|0,u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xor128=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),fF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.next=function(){var c=u.x^u.x>>>2;return u.x=u.y,u.y=u.z,u.z=u.w,u.w=u.v,(u.d=u.d+362437|0)+(u.v=u.v^u.v<<4^(c^c<<1))|0},u.x=0,u.y=0,u.z=0,u.w=0,u.v=0,l===(l|0)?u.x=l:p+=l;for(var d=0;d<p.length+64;d++)u.x^=p.charCodeAt(d)|0,d==p.length&&(u.d=u.x<<10^u.x>>>4),u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u.v=l.v,u.d=l.d,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xorwow=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),gF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this;u.next=function(){var d=u.x,c=u.i,h,m,f;return h=d[c],h^=h>>>7,m=h^h<<24,h=d[c+1&7],m^=h^h>>>10,h=d[c+3&7],m^=h^h>>>3,h=d[c+4&7],m^=h^h<<7,h=d[c+7&7],h=h^h<<13,m^=h^h<<9,d[c]=m,u.i=c+1&7,m};function p(d,c){var h,m,f=[];if(c===(c|0))m=f[0]=c;else for(c=""+c,h=0;h<c.length;++h)f[h&7]=f[h&7]<<15^c.charCodeAt(h)+f[h+1&7]<<13;for(;f.length<8;)f.push(0);for(h=0;h<8&&f[h]===0;++h);for(h==8?m=f[7]=-1:m=f[h],d.x=f,d.i=0,h=256;h>0;--h)d.next()}p(u,l)}function i(l,u){return u.x=l.x.slice(),u.i=l.i,u}function o(l,u){l==null&&(l=+new Date);var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(d.x&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xorshift7=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),yF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this;u.next=function(){var d=u.w,c=u.X,h=u.i,m,f;return u.w=d=d+1640531527|0,f=c[h+34&127],m=c[h=h+1&127],f^=f<<13,m^=m<<17,f^=f>>>15,m^=m>>>12,f=c[h]=f^m,u.i=h,f+(d^d>>>16)|0};function p(d,c){var h,m,f,g,y,b=[],x=128;for(c===(c|0)?(m=c,c=null):(c=c+"\0",m=0,x=Math.max(x,c.length)),f=0,g=-32;g<x;++g)c&&(m^=c.charCodeAt((g+32)%c.length)),g===0&&(y=m),m^=m<<10,m^=m>>>15,m^=m<<4,m^=m>>>13,g>=0&&(y=y+1640531527|0,h=b[g&127]^=m+y,f=h==0?f+1:0);for(f>=128&&(b[(c&&c.length||0)&127]=-1),f=127,g=4*128;g>0;--g)m=b[f+34&127],h=b[f=f+1&127],m^=m<<13,h^=h<<17,m^=m>>>15,h^=h>>>12,b[f]=m^h;d.w=y,d.X=b,d.i=f}p(u,l)}function i(l,u){return u.i=l.i,u.w=l.w,u.X=l.X.slice(),u}function o(l,u){l==null&&(l=+new Date);var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(d.X&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xor4096=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),bF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.next=function(){var c=u.b,h=u.c,m=u.d,f=u.a;return c=c<<25^c>>>7^h,h=h-m|0,m=m<<24^m>>>8^f,f=f-c|0,u.b=c=c<<20^c>>>12^h,u.c=h=h-m|0,u.d=m<<16^h>>>16^f,u.a=f-c|0},u.a=0,u.b=0,u.c=-1640531527,u.d=1367130551,l===Math.floor(l)?(u.a=l/4294967296|0,u.b=l|0):p+=l;for(var d=0;d<p.length+20;d++)u.b^=p.charCodeAt(d)|0,u.next()}function i(l,u){return u.a=l.a,u.b=l.b,u.c=l.c,u.d=l.d,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.tychei=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),xF=Bt(()=>{}),vF=Bt((e,t)=>{(function(n,a,r){var s=256,i=6,o=52,l="random",u=r.pow(s,i),p=r.pow(2,o),d=p*2,c=s-1,h;function m(I,T,C){var E=[];T=T==!0?{entropy:!0}:T||{};var A=b(y(T.entropy?[I,w(a)]:I==null?x():I,3),E),R=new f(E),F=function(){for(var S=R.g(i),M=u,B=0;S<p;)S=(S+B)*s,M*=s,B=R.g(1);for(;S>=d;)S/=2,M/=2,B>>>=1;return(S+B)/M};return F.int32=function(){return R.g(4)|0},F.quick=function(){return R.g(4)/4294967296},F.double=F,b(w(R.S),a),(T.pass||C||function(S,M,B,U){return U&&(U.S&&g(U,R),S.state=function(){return g(R,{})}),B?(r[l]=S,M):S})(F,A,"global"in T?T.global:this==r,T.state)}function f(I){var T,C=I.length,E=this,A=0,R=E.i=E.j=0,F=E.S=[];for(C||(I=[C++]);A<s;)F[A]=A++;for(A=0;A<s;A++)F[A]=F[R=c&R+I[A%C]+(T=F[A])],F[R]=T;(E.g=function(S){for(var M,B=0,U=E.i,G=E.j,q=E.S;S--;)M=q[U=c&U+1],B=B*s+q[c&(q[U]=q[G=c&G+M])+(q[G]=M)];return E.i=U,E.j=G,B})(s)}function g(I,T){return T.i=I.i,T.j=I.j,T.S=I.S.slice(),T}function y(I,T){var C=[],E=typeof I,A;if(T&&E=="object")for(A in I)try{C.push(y(I[A],T-1))}catch(R){}return C.length?C:E=="string"?I:I+"\0"}function b(I,T){for(var C=I+"",E,A=0;A<C.length;)T[c&A]=c&(E^=T[c&A]*19)+C.charCodeAt(A++);return w(T)}function x(){try{var I;return h&&(I=h.randomBytes)?I=I(s):(I=new Uint8Array(s),(n.crypto||n.msCrypto).getRandomValues(I)),w(I)}catch(E){var T=n.navigator,C=T&&T.plugins;return[+new Date,n,C,n.screen,w(a)]}}function w(I){return String.fromCharCode.apply(0,I)}if(b(r.random(),a),typeof t=="object"&&t.exports){t.exports=m;try{h=xF()}catch(I){}}else typeof define=="function"&&define.amd?define(function(){return m}):r["seed"+l]=m})(typeof self!="undefined"?self:e,[],Math)}),Hh=Bt((e,t)=>{var n=hF(),a=mF(),r=fF(),s=gF(),i=yF(),o=bF(),l=vF();l.alea=n,l.xor128=a,l.xorwow=r,l.xorshift7=s,l.xor4096=i,l.tychei=o,t.exports=l}),wI=Bt(()=>{}),xx=Bt(()=>{}),dh=Bt(()=>{}),wF=Bt(()=>{}),kF=Bt(()=>{}),IF=Bt(()=>{}),SF=Bt((e,t)=>{var n=(()=>{var a=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(a=a||__filename),function(r){r=r||{};function s(){return ue.buffer!=ze&&it(ue.buffer),dt}function i(){return ue.buffer!=ze&&it(ue.buffer),Hn}function o(){return ue.buffer!=ze&&it(ue.buffer),Mt}function l(){return ue.buffer!=ze&&it(ue.buffer),sn}function u(){return ue.buffer!=ze&&it(ue.buffer),Fn}function p(){return ue.buffer!=ze&&it(ue.buffer),ia}function d(){return ue.buffer!=ze&&it(ue.buffer),Dn}var c=typeof r!="undefined"?r:{},h,m;c.ready=new Promise(function(D,j){h=D,m=j});var f;typeof process!="undefined"&&process.listeners&&(f={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var g=Object.assign({},c),y=[],b="./this.program",x=(D,j)=>{throw j},w=typeof window=="object",I=typeof importScripts=="function",T=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",C=c.ENVIRONMENT_IS_PTHREAD||!1,E="";function A(D){return c.locateFile?c.locateFile(D,E):E+D}var R,F,S,M;function B(D){D instanceof Rs||Q("exiting due to exception: "+D)}if(T){I?E=dh().dirname(E)+"/":E=__dirname+"/";var U,G;typeof vI=="function"&&(U=xx(),G=dh()),R=(j,re)=>(j=G.normalize(j),U.readFileSync(j,re?void 0:"utf8")),S=j=>{var re=R(j,!0);return re.buffer||(re=new Uint8Array(re)),re},F=(j,re,ce)=>{j=G.normalize(j),U.readFile(j,function(ke,je){ke?ce(ke):re(je.buffer)})},process.argv.length>1&&(b=process.argv[1].replace(/\\/g,"/")),y=process.argv.slice(2),process.on("uncaughtException",function(j){if(!(j instanceof Rs))throw j}),process.on("unhandledRejection",function(j){throw j}),x=(j,re)=>{if(wa())throw process.exitCode=j,re;B(re),process.exit(j)},c.inspect=function(){return"[Emscripten Module object]"};let D;try{D=wF()}catch(j){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),j}global.Worker=D.Worker}else(w||I)&&(I?E=self.location.href:typeof document!="undefined"&&document.currentScript&&(E=document.currentScript.src),typeof a!="undefined"&&a&&(E=a),E.indexOf("blob:")!==0?E=E.substr(0,E.replace(/[?#].*/,"").lastIndexOf("/")+1):E="",T||(R=D=>{var j=new XMLHttpRequest;return j.open("GET",D,!1),j.send(null),j.responseText},I&&(S=D=>{var j=new XMLHttpRequest;return j.open("GET",D,!1),j.responseType="arraybuffer",j.send(null),new Uint8Array(j.response)}),F=(D,j,re)=>{var ce=new XMLHttpRequest;ce.open("GET",D,!0),ce.responseType="arraybuffer",ce.onload=()=>{if(ce.status==200||ce.status==0&&ce.response){j(ce.response);return}re()},ce.onerror=re,ce.send(null)}),M=D=>document.title=D);T&&typeof performance=="undefined"&&(global.performance=kF().performance);var q=console.log.bind(console),K=console.warn.bind(console);T&&(q=D=>U.writeSync(1,D+`
-`),K=D=>U.writeSync(2,D+`
-`));var Z=c.print||q,Q=c.printErr||K;Object.assign(c,g),g=null,c.arguments&&(y=c.arguments),c.thisProgram&&(b=c.thisProgram),c.quit&&(x=c.quit);var ee=4,ae=Atomics.load,te=Atomics.store,le=Atomics.compareExchange,ie;c.wasmBinary&&(ie=c.wasmBinary);var ye=c.noExitRuntime||!0;typeof WebAssembly!="object"&&Lo("no native wasm support detected");var ue,xe,Ie=!1,Se;function Le(D,j){D||Lo(j)}var Ve=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function nt(D,j,re){for(var ce=j+re,ke=j;D[ke]&&!(ke>=ce);)++ke;if(ke-j>16&&D.buffer&&Ve)return Ve.decode(D.buffer instanceof SharedArrayBuffer?D.slice(j,ke):D.subarray(j,ke));for(var je="";j<ke;){var Ce=D[j++];if(!(Ce&128)){je+=String.fromCharCode(Ce);continue}var De=D[j++]&63;if((Ce&224)==192){je+=String.fromCharCode((Ce&31)<<6|De);continue}var Ot=D[j++]&63;if((Ce&240)==224?Ce=(Ce&15)<<12|De<<6|Ot:Ce=(Ce&7)<<18|De<<12|Ot<<6|D[j++]&63,Ce<65536)je+=String.fromCharCode(Ce);else{var la=Ce-65536;je+=String.fromCharCode(55296|la>>10,56320|la&1023)}}return je}function st(D,j){return D?nt(i(),D,j):""}function Je(D,j,re,ce){if(!(ce>0))return 0;for(var ke=re,je=re+ce-1,Ce=0;Ce<D.length;++Ce){var De=D.charCodeAt(Ce);if(De>=55296&&De<=57343){var Ot=D.charCodeAt(++Ce);De=65536+((De&1023)<<10)|Ot&1023}if(De<=127){if(re>=je)break;j[re++]=De}else if(De<=2047){if(re+1>=je)break;j[re++]=192|De>>6,j[re++]=128|De&63}else if(De<=65535){if(re+2>=je)break;j[re++]=224|De>>12,j[re++]=128|De>>6&63,j[re++]=128|De&63}else{if(re+3>=je)break;j[re++]=240|De>>18,j[re++]=128|De>>12&63,j[re++]=128|De>>6&63,j[re++]=128|De&63}}return j[re]=0,re-ke}function at(D,j,re){return Je(D,i(),j,re)}var ze,dt,Hn,Mt,sa,sn,Fn,ia,Dn;C&&(ze=c.buffer);function it(D){ze=D,c.HEAP8=dt=new Int8Array(D),c.HEAP16=Mt=new Int16Array(D),c.HEAP32=sn=new Int32Array(D),c.HEAPU8=Hn=new Uint8Array(D),c.HEAPU16=sa=new Uint16Array(D),c.HEAPU32=Fn=new Uint32Array(D),c.HEAPF32=ia=new Float32Array(D),c.HEAPF64=Dn=new Float64Array(D)}var Rn=c.INITIAL_MEMORY||16777216;if(C)ue=c.wasmMemory,ze=c.buffer;else if(c.wasmMemory)ue=c.wasmMemory;else if(ue=new WebAssembly.Memory({initial:Rn/65536,maximum:32768,shared:!0}),!(ue.buffer instanceof SharedArrayBuffer))throw Q("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),T&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");ue&&(ze=ue.buffer),Rn=ze.byteLength,it(ze);var jn,gr=[],Mo=[],Ja=[],mp=!1;function wa(){return ye}function Po(){if(c.preRun)for(typeof c.preRun=="function"&&(c.preRun=[c.preRun]);c.preRun.length;)bg(c.preRun.shift());bp(gr)}function Xt(){mp=!0,!C&&bp(Mo)}function bd(){if(!C){if(c.postRun)for(typeof c.postRun=="function"&&(c.postRun=[c.postRun]);c.postRun.length;)I1(c.postRun.shift());bp(Ja)}}function bg(D){gr.unshift(D)}function xg(D){Mo.unshift(D)}function I1(D){Ja.unshift(D)}var Br=0,Oo=null,yr=null;function S1(D){Br++,c.monitorRunDependencies&&c.monitorRunDependencies(Br)}function T1(D){if(Br--,c.monitorRunDependencies&&c.monitorRunDependencies(Br),Br==0&&(Oo!==null&&(clearInterval(Oo),Oo=null),yr)){var j=yr;yr=null,j()}}function Lo(D){C?postMessage({cmd:"onAbort",arg:D}):c.onAbort&&c.onAbort(D),D="Aborted("+D+")",Q(D),Ie=!0,Se=1,D+=". Build with -sASSERTIONS for more info.";var j=new WebAssembly.RuntimeError(D);throw m(j),j}var vg="data:application/octet-stream;base64,";function xd(D){return D.startsWith(vg)}function fp(D){return D.startsWith("file://")}var yn;yn="tfjs-backend-wasm-threaded-simd.wasm",xd(yn)||(yn=A(yn));function vd(D){try{if(D==yn&&ie)return new Uint8Array(ie);if(S)return S(D);throw"both async and sync fetching of the wasm failed"}catch(j){Lo(j)}}function wg(){if(!ie&&(w||I)){if(typeof fetch=="function"&&!fp(yn))return fetch(yn,{credentials:"same-origin"}).then(function(D){if(!D.ok)throw"failed to load wasm binary file at '"+yn+"'";return D.arrayBuffer()}).catch(function(){return vd(yn)});if(F)return new Promise(function(D,j){F(yn,function(re){D(new Uint8Array(re))},j)})}return Promise.resolve().then(function(){return vd(yn)})}function kg(){var D={env:Fd,wasi_snapshot_preview1:Fd};function j(Ce,De){var Ot=Ce.exports;if(c.asm=Ot,Fg(c.asm._emscripten_tls_init),jn=c.asm.__indirect_function_table,xg(c.asm.__wasm_call_ctors),xe=De,!C){var la=$e.unusedWorkers.length;$e.unusedWorkers.forEach(function(xr){$e.loadWasmModuleToWorker(xr,function(){--la||T1("wasm-instantiate")})})}}C||S1("wasm-instantiate");function re(Ce){j(Ce.instance,Ce.module)}function ce(Ce){return wg().then(function(De){return WebAssembly.instantiate(De,D)}).then(function(De){return De}).then(Ce,function(De){Q("failed to asynchronously prepare wasm: "+De),Lo(De)})}function ke(){return!ie&&typeof WebAssembly.instantiateStreaming=="function"&&!xd(yn)&&!fp(yn)&&!T&&typeof fetch=="function"?fetch(yn,{credentials:"same-origin"}).then(function(Ce){var De=WebAssembly.instantiateStreaming(Ce,D);return De.then(re,function(Ot){return Q("wasm streaming compile failed: "+Ot),Q("falling back to ArrayBuffer instantiation"),ce(re)})}):ce(re)}if(c.instantiateWasm)try{var je=c.instantiateWasm(D,j);return je}catch(Ce){Q("Module.instantiateWasm callback failed with error: "+Ce),m(Ce)}return ke().catch(m),{}}var Ig,N1,Sg={};function Rs(D){this.name="ExitStatus",this.message="Program terminated with exit("+D+")",this.status=D}function Tg(D){var j=$e.pthreads[D];delete $e.pthreads[D],j.terminate(),eb(D),$e.runningWorkers.splice($e.runningWorkers.indexOf(j),1),j.pthread_ptr=0}function Ng(D){var j=$e.pthreads[D];j.postMessage({cmd:"cancel"})}function gp(D){var j=$e.pthreads[D];Le(j),$e.returnWorkerToPool(j)}function wd(D){var j=$e.getNewWorker();if(!j)return 6;$e.runningWorkers.push(j),$e.pthreads[D.pthread_ptr]=j,j.pthread_ptr=D.pthread_ptr;var re={cmd:"run",start_routine:D.startRoutine,arg:D.arg,pthread_ptr:D.pthread_ptr};return j.runPthread=()=>{re.time=performance.now(),j.postMessage(re,D.transferList)},j.loaded&&(j.runPthread(),delete j.runPthread),0}var kd={varargs:void 0,get:function(){kd.varargs+=4;var D=l()[kd.varargs-4>>2];return D},getStr:function(D){var j=st(D);return j}};function yp(D){if(C)return Vr(1,1,D);Se=D,wa()||($e.terminateAllThreads(),c.onExit&&c.onExit(D),Ie=!0),x(D,new Rs(D))}function C1(D,j){if(Se=D,!j&&C)throw Sd(D),"unwind";yp(D)}var Id=C1;function Cg(D){if(D instanceof Rs||D=="unwind")return Se;x(1,D)}var $e={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){C?$e.initWorker():$e.initMainThread()},initMainThread:function(){for(var D=8;D--;)$e.allocateUnusedWorker()},initWorker:function(){ye=!1},setExitStatus:function(D){Se=D},terminateAllThreads:function(){for(var D of Object.values($e.pthreads))$e.returnWorkerToPool(D);for(var D of $e.unusedWorkers)D.terminate();$e.unusedWorkers=[]},returnWorkerToPool:function(D){var j=D.pthread_ptr;delete $e.pthreads[j],$e.unusedWorkers.push(D),$e.runningWorkers.splice($e.runningWorkers.indexOf(D),1),D.pthread_ptr=0,eb(j)},receiveObjectTransfer:function(D){},threadInitTLS:function(){$e.tlsInitFunctions.forEach(D=>D())},loadWasmModuleToWorker:function(D,j){D.onmessage=re=>{var ce=re.data,ke=ce.cmd;if(D.pthread_ptr&&($e.currentProxiedOperationCallerThread=D.pthread_ptr),ce.targetThread&&ce.targetThread!=Ld()){var je=$e.pthreads[ce.targetThread];je?je.postMessage(ce,ce.transferList):Q('Internal error! Worker sent a message "'+ke+'" to target pthread '+ce.targetThread+", but that thread no longer exists!"),$e.currentProxiedOperationCallerThread=void 0;return}ke==="processProxyingQueue"?xp(ce.queue):ke==="spawnThread"?wd(ce):ke==="cleanupThread"?gp(ce.thread):ke==="killThread"?Tg(ce.thread):ke==="cancelThread"?Ng(ce.thread):ke==="loaded"?(D.loaded=!0,j&&j(D),D.runPthread&&(D.runPthread(),delete D.runPthread)):ke==="print"?Z("Thread "+ce.threadId+": "+ce.text):ke==="printErr"?Q("Thread "+ce.threadId+": "+ce.text):ke==="alert"?alert("Thread "+ce.threadId+": "+ce.text):ce.target==="setimmediate"?D.postMessage(ce):ke==="onAbort"?c.onAbort&&c.onAbort(ce.arg):ke&&Q("worker sent an unknown command "+ke),$e.currentProxiedOperationCallerThread=void 0},D.onerror=re=>{var ce="worker sent an error!";throw Q(ce+" "+re.filename+":"+re.lineno+": "+re.message),re},T&&(D.on("message",function(re){D.onmessage({data:re})}),D.on("error",function(re){D.onerror(re)}),D.on("detachedExit",function(){})),D.postMessage({cmd:"load",urlOrBlob:c.mainScriptUrlOrBlob||a,wasmMemory:ue,wasmModule:xe})},allocateUnusedWorker:function(){var D=A("tfjs-backend-wasm-threaded-simd.worker.js");$e.unusedWorkers.push(new Worker(D))},getNewWorker:function(){return $e.unusedWorkers.length==0&&($e.allocateUnusedWorker(),$e.loadWasmModuleToWorker($e.unusedWorkers[0])),$e.unusedWorkers.pop()}};c.PThread=$e;function bp(D){for(;D.length>0;)D.shift()(c)}function _g(D){var j=tb(),re=D();return zd(j),re}function _1(D){return D}function E1(D){var j=/\b_Z[\w\d_]+/g;return D.replace(j,function(re){var ce=re;return re===ce?re:ce+" ["+re+"]"})}function Eg(){var D=Ld(),j=l()[D+44>>2],re=l()[D+48>>2],ce=j-re;P1(j,ce),zd(j)}c.establishStackSpace=Eg;function Sd(D){if(C)return Vr(2,0,D);try{Id(D)}catch(j){Cg(j)}}var zo=[];function Ag(D){var j=zo[D];return j||(D>=zo.length&&(zo.length=D+1),zo[D]=j=jn.get(D)),j}function $g(D,j){var re=Ag(D)(j);wa()?$e.setExitStatus(re):M1(re)}c.invokeEntryPoint=$g;function A1(){var D=new Error;if(!D.stack){try{throw new Error}catch(j){D=j}if(!D.stack)return"(no stack trace available)"}return D.stack.toString()}function Fg(D){$e.tlsInitFunctions.push(D)}function Dg(D,j){s().set(D,j)}function Rg(D){F1(D,!I,1,!w),$e.threadInitTLS()}function Mg(D){C?postMessage({cmd:"cleanupThread",thread:D}):gp(D)}function Td(D,j,re,ce){return C?Vr(3,1,D,j,re,ce):Nd(D,j,re,ce)}function Nd(D,j,re,ce){if(typeof SharedArrayBuffer=="undefined")return Q("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var ke=[],je=0;if(C&&(ke.length===0||je))return Td(D,j,re,ce);if(je)return je;var Ce={startRoutine:re,pthread_ptr:D,arg:ce,transferList:ke};return C?(Ce.cmd="spawnThread",postMessage(Ce,ke),0):wd(Ce)}function Pg(){return 2097152}var Og=!0;function Lg(){return Og}function xp(D){Atomics.store(l(),D>>2,1),Ld()&&R1(D),Atomics.compareExchange(l(),D>>2,1,0)}c.executeNotifiedProxyingQueue=xp;function zg(D,j,re,ce){if(D==j)setTimeout(()=>xp(ce));else if(C)postMessage({targetThread:D,cmd:"processProxyingQueue",queue:ce});else{var ke=$e.pthreads[D];if(!ke)return;ke.postMessage({cmd:"processProxyingQueue",queue:ce})}return 1}function Wg(D,j,re){return-1}function Bg(){Lo("")}function Ms(D){Ms.shown||(Ms.shown={}),Ms.shown[D]||(Ms.shown[D]=1,T&&(D="warning: "+D),Q(D))}function Vg(){T||I||Ms("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function Ug(){return Date.now()}function Cd(){return 2147483648}function Gg(){return Cd()}var Wo;T?Wo=()=>{var D=process.hrtime();return D[0]*1e3+D[1]/1e6}:C?Wo=()=>performance.now()-c.__performance_now_clock_drift:Wo=()=>performance.now();function Hg(D,j,re){i().copyWithin(D,j,j+re)}function jg(){return T?IF().cpus().length:navigator.hardwareConcurrency}function Vr(D,j){var re=arguments.length-2,ce=arguments;return _g(()=>{for(var ke=re,je=Wd(ke*8),Ce=je>>3,De=0;De<re;De++){var Ot=ce[2+De];d()[Ce+De]=Ot}return D1(D,ke,je,j)})}var vp=[];function qg(D,j,re){vp.length=j;for(var ce=re>>3,ke=0;ke<j;ke++)vp[ke]=d()[ce+ke];var je=D<0,Ce=je?Sg[-D-1]:ty[D];return Ce.apply(null,vp)}function Kg(D){try{return ue.grow(D-ze.byteLength+65535>>>16),it(ue.buffer),1}catch(j){}}function Xg(D){var j=i().length;if(D=D>>>0,D<=j)return!1;var re=Cd();if(D>re)return!1;let ce=(Ot,la)=>Ot+(la-Ot%la)%la;for(var ke=1;ke<=4;ke*=2){var je=j*(1+.2/ke);je=Math.min(je,D+100663296);var Ce=Math.min(re,ce(Math.max(D,je),65536)),De=Kg(Ce);if(De)return!0}return!1}function Yg(){throw"unwind"}function _d(D){return C?Vr(4,1,D):52}function Ed(D,j,re,ce,ke){return C?Vr(5,1,D,j,re,ce,ke):70}var Zg=[null,[],[]];function Jg(D,j){var re=Zg[D];j===0||j===10?((D===1?Z:Q)(nt(re,0)),re.length=0):re.push(j)}function Ad(D,j,re,ce){if(C)return Vr(6,1,D,j,re,ce);for(var ke=0,je=0;je<re;je++){var Ce=u()[j>>2],De=u()[j+4>>2];j+=8;for(var Ot=0;Ot<De;Ot++)Jg(D,i()[Ce+Ot]);ke+=De}return u()[ce>>2]=ke,0}function $d(D){var j=c["_"+D];return j}function Qg(D,j,re,ce,ke){var je={string:ua=>{var Go=0;if(ua!=null&&ua!==0){var z1=(ua.length<<2)+1;Go=Wd(z1),at(ua,Go,z1)}return Go},array:ua=>{var Go=Wd(ua.length);return Dg(ua,Go),Go}};function Ce(ua){return j==="string"?st(ua):j==="boolean"?Boolean(ua):ua}var De=$d(D),Ot=[],la=0;if(ce)for(var xr=0;xr<ce.length;xr++){var L1=je[re[xr]];L1?(la===0&&(la=tb()),Ot[xr]=L1(ce[xr])):Ot[xr]=ce[xr]}var nb=De.apply(null,Ot);function nF(ua){return la!==0&&zd(la),Ce(ua)}return nb=nF(nb),nb}function ey(D,j,re,ce){re=re||[];var ke=re.every(Ce=>Ce==="number"||Ce==="boolean"),je=j!=="string";return je&&ke&&!ce?$d(D):function(){return Qg(D,j,re,arguments,ce)}}$e.init();var ty=[null,yp,Sd,Td,_d,Ed,Ad],Fd={__emscripten_init_main_thread_js:Rg,__emscripten_thread_cleanup:Mg,__pthread_create_js:Nd,_emscripten_default_pthread_stack_size:Pg,_emscripten_get_now_is_monotonic:Lg,_emscripten_notify_task_queue:zg,_emscripten_set_offscreencanvas_size:Wg,abort:Bg,emscripten_check_blocking_allowed:Vg,emscripten_date_now:Ug,emscripten_get_heap_max:Gg,emscripten_get_now:Wo,emscripten_memcpy_big:Hg,emscripten_num_logical_cores:jg,emscripten_receive_on_main_thread_js:qg,emscripten_resize_heap:Xg,emscripten_unwind_to_js_event_loop:Yg,exit:Id,fd_close:_d,fd_seek:Ed,fd_write:Ad,memory:ue||c.wasmMemory},$1=kg(),ny=c.___wasm_call_ctors=function(){return(ny=c.___wasm_call_ctors=c.asm.__wasm_call_ctors).apply(null,arguments)},ay=c._init=function(){return(ay=c._init=c.asm.init).apply(null,arguments)},ry=c._init_with_threads_count=function(){return(ry=c._init_with_threads_count=c.asm.init_with_threads_count).apply(null,arguments)},sy=c._get_threads_count=function(){return(sy=c._get_threads_count=c.asm.get_threads_count).apply(null,arguments)},iy=c._register_tensor=function(){return(iy=c._register_tensor=c.asm.register_tensor).apply(null,arguments)},oy=c._dispose_data=function(){return(oy=c._dispose_data=c.asm.dispose_data).apply(null,arguments)},ly=c._dispose=function(){return(ly=c._dispose=c.asm.dispose).apply(null,arguments)},uy=c._Abs=function(){return(uy=c._Abs=c.asm.Abs).apply(null,arguments)},py=c._Add=function(){return(py=c._Add=c.asm.Add).apply(null,arguments)},cy=c._AddN=function(){return(cy=c._AddN=c.asm.AddN).apply(null,arguments)},dy=c._All=function(){return(dy=c._All=c.asm.All).apply(null,arguments)},hy=c._Any=function(){return(hy=c._Any=c.asm.Any).apply(null,arguments)},my=c._ArgMax=function(){return(my=c._ArgMax=c.asm.ArgMax).apply(null,arguments)},fy=c._AvgPool=function(){return(fy=c._AvgPool=c.asm.AvgPool).apply(null,arguments)},gy=c._BatchMatMul=function(){return(gy=c._BatchMatMul=c.asm.BatchMatMul).apply(null,arguments)},yy=c._Ceil=function(){return(yy=c._Ceil=c.asm.Ceil).apply(null,arguments)},by=c._ClipByValue=function(){return(by=c._ClipByValue=c.asm.ClipByValue).apply(null,arguments)},xy=c._Conv2D=function(){return(xy=c._Conv2D=c.asm.Conv2D).apply(null,arguments)},vy=c._Conv2DBackpropInput=function(){return(vy=c._Conv2DBackpropInput=c.asm.Conv2DBackpropInput).apply(null,arguments)},wy=c._Cos=function(){return(wy=c._Cos=c.asm.Cos).apply(null,arguments)},ky=c._Cosh=function(){return(ky=c._Cosh=c.asm.Cosh).apply(null,arguments)},Iy=c._CropAndResize=function(){return(Iy=c._CropAndResize=c.asm.CropAndResize).apply(null,arguments)},Sy=c._Cumprod=function(){return(Sy=c._Cumprod=c.asm.Cumprod).apply(null,arguments)},Ty=c._Cumsum=function(){return(Ty=c._Cumsum=c.asm.Cumsum).apply(null,arguments)},Ny=c._DepthToSpace=function(){return(Ny=c._DepthToSpace=c.asm.DepthToSpace).apply(null,arguments)},Cy=c._DepthwiseConv2dNative=function(){return(Cy=c._DepthwiseConv2dNative=c.asm.DepthwiseConv2dNative).apply(null,arguments)},_y=c._Elu=function(){return(_y=c._Elu=c.asm.Elu).apply(null,arguments)},Ey=c._Equal=function(){return(Ey=c._Equal=c.asm.Equal).apply(null,arguments)},Ay=c._Exp=function(){return(Ay=c._Exp=c.asm.Exp).apply(null,arguments)},$y=c._FlipLeftRight=function(){return($y=c._FlipLeftRight=c.asm.FlipLeftRight).apply(null,arguments)},Fy=c._Floor=function(){return(Fy=c._Floor=c.asm.Floor).apply(null,arguments)},Dy=c._FloorDiv=function(){return(Dy=c._FloorDiv=c.asm.FloorDiv).apply(null,arguments)},Ry=c._FusedBatchNorm=function(){return(Ry=c._FusedBatchNorm=c.asm.FusedBatchNorm).apply(null,arguments)},My=c._FusedConv2D=function(){return(My=c._FusedConv2D=c.asm.FusedConv2D).apply(null,arguments)},Py=c._FusedDepthwiseConv2D=function(){return(Py=c._FusedDepthwiseConv2D=c.asm.FusedDepthwiseConv2D).apply(null,arguments)},Oy=c._Gather=function(){return(Oy=c._Gather=c.asm.Gather).apply(null,arguments)},Ly=c._GatherNd=function(){return(Ly=c._GatherNd=c.asm.GatherNd).apply(null,arguments)},zy=c._Greater=function(){return(zy=c._Greater=c.asm.Greater).apply(null,arguments)},Wy=c._GreaterEqual=function(){return(Wy=c._GreaterEqual=c.asm.GreaterEqual).apply(null,arguments)},By=c._LeakyRelu=function(){return(By=c._LeakyRelu=c.asm.LeakyRelu).apply(null,arguments)},Vy=c._Less=function(){return(Vy=c._Less=c.asm.Less).apply(null,arguments)},Uy=c._LessEqual=function(){return(Uy=c._LessEqual=c.asm.LessEqual).apply(null,arguments)},Gy=c._Log=function(){return(Gy=c._Log=c.asm.Log).apply(null,arguments)},Hy=c._LogicalAnd=function(){return(Hy=c._LogicalAnd=c.asm.LogicalAnd).apply(null,arguments)},jy=c._LogicalNot=function(){return(jy=c._LogicalNot=c.asm.LogicalNot).apply(null,arguments)},qy=c._LogicalOr=function(){return(qy=c._LogicalOr=c.asm.LogicalOr).apply(null,arguments)},Ky=c._LogicalXor=function(){return(Ky=c._LogicalXor=c.asm.LogicalXor).apply(null,arguments)},Xy=c._Max=function(){return(Xy=c._Max=c.asm.Max).apply(null,arguments)},Dd=c._MaxPool=function(){return(Dd=c._MaxPool=c.asm.MaxPool).apply(null,arguments)},Rd=c._Maximum=function(){return(Rd=c._Maximum=c.asm.Maximum).apply(null,arguments)},wp=c._Mean=function(){return(wp=c._Mean=c.asm.Mean).apply(null,arguments)},Yy=c._Min=function(){return(Yy=c._Min=c.asm.Min).apply(null,arguments)},Zy=c._Minimum=function(){return(Zy=c._Minimum=c.asm.Minimum).apply(null,arguments)},Bo=c._MirrorPad=function(){return(Bo=c._MirrorPad=c.asm.MirrorPad).apply(null,arguments)},Md=c._Multiply=function(){return(Md=c._Multiply=c.asm.Multiply).apply(null,arguments)},Vo=c._Neg=function(){return(Vo=c._Neg=c.asm.Neg).apply(null,arguments)},Uo=c._NonMaxSuppressionV3=function(){return(Uo=c._NonMaxSuppressionV3=c.asm.NonMaxSuppressionV3).apply(null,arguments)},Jy=c._NonMaxSuppressionV4=function(){return(Jy=c._NonMaxSuppressionV4=c.asm.NonMaxSuppressionV4).apply(null,arguments)},X=c._NonMaxSuppressionV5=function(){return(X=c._NonMaxSuppressionV5=c.asm.NonMaxSuppressionV5).apply(null,arguments)},se=c._NotEqual=function(){return(se=c._NotEqual=c.asm.NotEqual).apply(null,arguments)},we=c._OneHot=function(){return(we=c._OneHot=c.asm.OneHot).apply(null,arguments)},He=c._PadV2=function(){return(He=c._PadV2=c.asm.PadV2).apply(null,arguments)},wt=c._Pow=function(){return(wt=c._Pow=c.asm.Pow).apply(null,arguments)},kt=c._Prelu=function(){return(kt=c._Prelu=c.asm.Prelu).apply(null,arguments)},Ue=c._Prod=function(){return(Ue=c._Prod=c.asm.Prod).apply(null,arguments)},We=c._RealDiv=function(){return(We=c._RealDiv=c.asm.RealDiv).apply(null,arguments)},Pt=c._Relu=function(){return(Pt=c._Relu=c.asm.Relu).apply(null,arguments)},oa=c._Relu6=function(){return(oa=c._Relu6=c.asm.Relu6).apply(null,arguments)},br=c._ResizeBilinear=function(){return(br=c._ResizeBilinear=c.asm.ResizeBilinear).apply(null,arguments)},Pd=c._ResizeNearestNeighbor=function(){return(Pd=c._ResizeNearestNeighbor=c.asm.ResizeNearestNeighbor).apply(null,arguments)},kp=c._Reverse=function(){return(kp=c._Reverse=c.asm.Reverse).apply(null,arguments)},Qy=c._RotateWithOffset=function(){return(Qy=c._RotateWithOffset=c.asm.RotateWithOffset).apply(null,arguments)},Mn=c._Round=function(){return(Mn=c._Round=c.asm.Round).apply(null,arguments)},Ur=c._Rsqrt=function(){return(Ur=c._Rsqrt=c.asm.Rsqrt).apply(null,arguments)},Od=c._ScatterNd=function(){return(Od=c._ScatterNd=c.asm.ScatterNd).apply(null,arguments)},S$=c._SelectV2=function(){return(S$=c._SelectV2=c.asm.SelectV2).apply(null,arguments)},T$=c._Sigmoid=function(){return(T$=c._Sigmoid=c.asm.Sigmoid).apply(null,arguments)},N$=c._Sin=function(){return(N$=c._Sin=c.asm.Sin).apply(null,arguments)},C$=c._Softmax=function(){return(C$=c._Softmax=c.asm.Softmax).apply(null,arguments)},_$=c._SparseFillEmptyRows=function(){return(_$=c._SparseFillEmptyRows=c.asm.SparseFillEmptyRows).apply(null,arguments)},E$=c._SparseReshape=function(){return(E$=c._SparseReshape=c.asm.SparseReshape).apply(null,arguments)},A$=c._SparseSegmentReduction=function(){return(A$=c._SparseSegmentReduction=c.asm.SparseSegmentReduction).apply(null,arguments)},$$=c._Sqrt=function(){return($$=c._Sqrt=c.asm.Sqrt).apply(null,arguments)},F$=c._Square=function(){return(F$=c._Square=c.asm.Square).apply(null,arguments)},D$=c._SquaredDifference=function(){return(D$=c._SquaredDifference=c.asm.SquaredDifference).apply(null,arguments)},R$=c._Step=function(){return(R$=c._Step=c.asm.Step).apply(null,arguments)},M$=c._StridedSlice=function(){return(M$=c._StridedSlice=c.asm.StridedSlice).apply(null,arguments)},P$=c._Sub=function(){return(P$=c._Sub=c.asm.Sub).apply(null,arguments)},O$=c._Sum=function(){return(O$=c._Sum=c.asm.Sum).apply(null,arguments)},L$=c._Tan=function(){return(L$=c._Tan=c.asm.Tan).apply(null,arguments)},z$=c._Tanh=function(){return(z$=c._Tanh=c.asm.Tanh).apply(null,arguments)},W$=c._Tile=function(){return(W$=c._Tile=c.asm.Tile).apply(null,arguments)},B$=c._TopK=function(){return(B$=c._TopK=c.asm.TopK).apply(null,arguments)},V$=c._Transform=function(){return(V$=c._Transform=c.asm.Transform).apply(null,arguments)},U$=c._Transpose=function(){return(U$=c._Transpose=c.asm.Transpose).apply(null,arguments)},G$=c.__FusedMatMul=function(){return(G$=c.__FusedMatMul=c.asm._FusedMatMul).apply(null,arguments)},H$=c._malloc=function(){return(H$=c._malloc=c.asm.malloc).apply(null,arguments)},j$=c._free=function(){return(j$=c._free=c.asm.free).apply(null,arguments)},q$=c.__emscripten_tls_init=function(){return(q$=c.__emscripten_tls_init=c.asm._emscripten_tls_init).apply(null,arguments)},Ld=c._pthread_self=function(){return(Ld=c._pthread_self=c.asm.pthread_self).apply(null,arguments)},K$=c.___errno_location=function(){return(K$=c.___errno_location=c.asm.__errno_location).apply(null,arguments)},F1=c.__emscripten_thread_init=function(){return(F1=c.__emscripten_thread_init=c.asm._emscripten_thread_init).apply(null,arguments)},X$=c.__emscripten_thread_crashed=function(){return(X$=c.__emscripten_thread_crashed=c.asm._emscripten_thread_crashed).apply(null,arguments)},Y$=c._emscripten_main_thread_process_queued_calls=function(){return(Y$=c._emscripten_main_thread_process_queued_calls=c.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},Z$=c._emscripten_main_browser_thread_id=function(){return(Z$=c._emscripten_main_browser_thread_id=c.asm.emscripten_main_browser_thread_id).apply(null,arguments)},D1=c._emscripten_run_in_main_runtime_thread_js=function(){return(D1=c._emscripten_run_in_main_runtime_thread_js=c.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},J$=c._emscripten_dispatch_to_thread_=function(){return(J$=c._emscripten_dispatch_to_thread_=c.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},R1=c.__emscripten_proxy_execute_task_queue=function(){return(R1=c.__emscripten_proxy_execute_task_queue=c.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)},eb=c.__emscripten_thread_free_data=function(){return(eb=c.__emscripten_thread_free_data=c.asm._emscripten_thread_free_data).apply(null,arguments)},M1=c.__emscripten_thread_exit=function(){return(M1=c.__emscripten_thread_exit=c.asm._emscripten_thread_exit).apply(null,arguments)},P1=c._emscripten_stack_set_limits=function(){return(P1=c._emscripten_stack_set_limits=c.asm.emscripten_stack_set_limits).apply(null,arguments)},tb=c.stackSave=function(){return(tb=c.stackSave=c.asm.stackSave).apply(null,arguments)},zd=c.stackRestore=function(){return(zd=c.stackRestore=c.asm.stackRestore).apply(null,arguments)},Wd=c.stackAlloc=function(){return(Wd=c.stackAlloc=c.asm.stackAlloc).apply(null,arguments)},Q$=c.dynCall_iijjiiii=function(){return(Q$=c.dynCall_iijjiiii=c.asm.dynCall_iijjiiii).apply(null,arguments)},eF=c.dynCall_jiji=function(){return(eF=c.dynCall_jiji=c.asm.dynCall_jiji).apply(null,arguments)};c.keepRuntimeAlive=wa,c.wasmMemory=ue,c.cwrap=ey,c.ExitStatus=Rs,c.PThread=$e;var Bd;yr=function D(){Bd||O1(),Bd||(yr=D)};function O1(D){if(D=D||y,Br>0)return;if(C){h(c),Xt(),postMessage({cmd:"loaded"});return}if(Po(),Br>0)return;function j(){Bd||(Bd=!0,c.calledRun=!0,!Ie&&(Xt(),h(c),c.onRuntimeInitialized&&c.onRuntimeInitialized(),bd()))}c.setStatus?(c.setStatus("Running..."),setTimeout(function(){setTimeout(function(){c.setStatus("")},1),j()},1)):j()}if(c.preInit)for(typeof c.preInit=="function"&&(c.preInit=[c.preInit]);c.preInit.length>0;)c.preInit.pop()();O1();var Vd;f&&(Vd={uncaughtException:process.listeners("uncaughtException").filter(function(D){return!f.uncaughtException.indexOf(D)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(D){return!f.unhandledRejection.indexOf(D)>-1})});var Ud;if(typeof WasmBackendModule!="undefined")Ud=WasmBackendModule;else if(typeof r!="undefined")Ud=r;else throw new Error("Could not find wasm module in post.js");if(Vd){var tF=Ud._dispose;Ud._dispose=function(){tF(),Vd.uncaughtException.forEach(function(D){process.removeListener("uncaughtException",D)}),Vd.unhandledRejection.forEach(function(D){process.removeListener("unhandledRejection",D)})}}return r.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModuleThreadedSimd=n)}),TF=Bt((e,t)=>{t.exports.wasmWorkerContents=`"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
-");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`}),NF=Bt((e,t)=>{var n=(()=>{var a=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(a=a||__filename),function(r){r=r||{};var s=typeof r!="undefined"?r:{},i,o;s.ready=new Promise(function(X,se){i=X,o=se});var l;typeof process!="undefined"&&process.listeners&&(l={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var u=Object.assign({},s),p=[],d="./this.program",c=(X,se)=>{throw se},h=typeof window=="object",m=typeof importScripts=="function",f=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",g="";function y(X){return s.locateFile?s.locateFile(X,g):g+X}var b,x,w,I;function T(X){X instanceof Oo||R("exiting due to exception: "+X)}if(f){m?g=dh().dirname(g)+"/":g=__dirname+"/";var C,E;typeof vI=="function"&&(C=xx(),E=dh()),b=(X,se)=>(X=E.normalize(X),C.readFileSync(X,se?void 0:"utf8")),w=X=>{var se=b(X,!0);return se.buffer||(se=new Uint8Array(se)),se},x=(X,se,we)=>{X=E.normalize(X),C.readFile(X,function(He,wt){He?we(He):se(wt.buffer)})},process.argv.length>1&&(d=process.argv[1].replace(/\\/g,"/")),p=process.argv.slice(2),process.on("uncaughtException",function(X){if(!(X instanceof Oo))throw X}),process.on("unhandledRejection",function(X){throw X}),c=(X,se)=>{if(Hn())throw process.exitCode=X,se;T(se),process.exit(X)},s.inspect=function(){return"[Emscripten Module object]"}}else(h||m)&&(m?g=self.location.href:typeof document!="undefined"&&document.currentScript&&(g=document.currentScript.src),a&&(g=a),g.indexOf("blob:")!==0?g=g.substr(0,g.replace(/[?#].*/,"").lastIndexOf("/")+1):g="",b=X=>{var se=new XMLHttpRequest;return se.open("GET",X,!1),se.send(null),se.responseText},m&&(w=X=>{var se=new XMLHttpRequest;return se.open("GET",X,!1),se.responseType="arraybuffer",se.send(null),new Uint8Array(se.response)}),x=(X,se,we)=>{var He=new XMLHttpRequest;He.open("GET",X,!0),He.responseType="arraybuffer",He.onload=()=>{if(He.status==200||He.status==0&&He.response){se(He.response);return}we()},He.onerror=we,He.send(null)},I=X=>document.title=X);var A=s.print||console.log.bind(console),R=s.printErr||console.warn.bind(console);Object.assign(s,u),u=null,s.arguments&&(p=s.arguments),s.thisProgram&&(d=s.thisProgram),s.quit&&(c=s.quit);var F=4,S;s.wasmBinary&&(S=s.wasmBinary);var M=s.noExitRuntime||!0;typeof WebAssembly!="object"&&Ja("no native wasm support detected");var B,U=!1,G;function q(X,se){X||Ja(se)}var K=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function Z(X,se,we){for(var He=se+we,wt=se;X[wt]&&!(wt>=He);)++wt;if(wt-se>16&&X.buffer&&K)return K.decode(X.subarray(se,wt));for(var kt="";se<wt;){var Ue=X[se++];if(!(Ue&128)){kt+=String.fromCharCode(Ue);continue}var We=X[se++]&63;if((Ue&224)==192){kt+=String.fromCharCode((Ue&31)<<6|We);continue}var Pt=X[se++]&63;if((Ue&240)==224?Ue=(Ue&15)<<12|We<<6|Pt:Ue=(Ue&7)<<18|We<<12|Pt<<6|X[se++]&63,Ue<65536)kt+=String.fromCharCode(Ue);else{var oa=Ue-65536;kt+=String.fromCharCode(55296|oa>>10,56320|oa&1023)}}return kt}function Q(X,se){return X?Z(ie,X,se):""}function ee(X,se,we,He){if(!(He>0))return 0;for(var wt=we,kt=we+He-1,Ue=0;Ue<X.length;++Ue){var We=X.charCodeAt(Ue);if(We>=55296&&We<=57343){var Pt=X.charCodeAt(++Ue);We=65536+((We&1023)<<10)|Pt&1023}if(We<=127){if(we>=kt)break;se[we++]=We}else if(We<=2047){if(we+1>=kt)break;se[we++]=192|We>>6,se[we++]=128|We&63}else if(We<=65535){if(we+2>=kt)break;se[we++]=224|We>>12,se[we++]=128|We>>6&63,se[we++]=128|We&63}else{if(we+3>=kt)break;se[we++]=240|We>>18,se[we++]=128|We>>12&63,se[we++]=128|We>>6&63,se[we++]=128|We&63}}return se[we]=0,we-wt}function ae(X,se,we){return ee(X,ie,se,we)}var te,le,ie,ye,ue,xe,Ie,Se,Le;function Ve(X){te=X,s.HEAP8=le=new Int8Array(X),s.HEAP16=ye=new Int16Array(X),s.HEAP32=xe=new Int32Array(X),s.HEAPU8=ie=new Uint8Array(X),s.HEAPU16=ue=new Uint16Array(X),s.HEAPU32=Ie=new Uint32Array(X),s.HEAPF32=Se=new Float32Array(X),s.HEAPF64=Le=new Float64Array(X)}var nt=s.INITIAL_MEMORY||16777216,st,Je=[],at=[],ze=[],dt=!1;function Hn(){return M}function Mt(){if(s.preRun)for(typeof s.preRun=="function"&&(s.preRun=[s.preRun]);s.preRun.length;)Fn(s.preRun.shift());yr(Je)}function sa(){dt=!0,yr(at)}function sn(){if(s.postRun)for(typeof s.postRun=="function"&&(s.postRun=[s.postRun]);s.postRun.length;)Dn(s.postRun.shift());yr(ze)}function Fn(X){Je.unshift(X)}function ia(X){at.unshift(X)}function Dn(X){ze.unshift(X)}var it=0,Rn=null,jn=null;function gr(X){it++,s.monitorRunDependencies&&s.monitorRunDependencies(it)}function Mo(X){if(it--,s.monitorRunDependencies&&s.monitorRunDependencies(it),it==0&&(Rn!==null&&(clearInterval(Rn),Rn=null),jn)){var se=jn;jn=null,se()}}function Ja(X){s.onAbort&&s.onAbort(X),X="Aborted("+X+")",R(X),U=!0,G=1,X+=". Build with -sASSERTIONS for more info.";var se=new WebAssembly.RuntimeError(X);throw o(se),se}var mp="data:application/octet-stream;base64,";function wa(X){return X.startsWith(mp)}function Po(X){return X.startsWith("file://")}var Xt;Xt="tfjs-backend-wasm.wasm",wa(Xt)||(Xt=y(Xt));function bd(X){try{if(X==Xt&&S)return new Uint8Array(S);if(w)return w(X);throw"both async and sync fetching of the wasm failed"}catch(se){Ja(se)}}function bg(){if(!S&&(h||m)){if(typeof fetch=="function"&&!Po(Xt))return fetch(Xt,{credentials:"same-origin"}).then(function(X){if(!X.ok)throw"failed to load wasm binary file at '"+Xt+"'";return X.arrayBuffer()}).catch(function(){return bd(Xt)});if(x)return new Promise(function(X,se){x(Xt,function(we){X(new Uint8Array(we))},se)})}return Promise.resolve().then(function(){return bd(Xt)})}function xg(){var X={env:yp,wasi_snapshot_preview1:yp};function se(Ue,We){var Pt=Ue.exports;s.asm=Pt,B=s.asm.memory,Ve(B.buffer),st=s.asm.__indirect_function_table,ia(s.asm.__wasm_call_ctors),Mo("wasm-instantiate")}gr("wasm-instantiate");function we(Ue){se(Ue.instance)}function He(Ue){return bg().then(function(We){return WebAssembly.instantiate(We,X)}).then(function(We){return We}).then(Ue,function(We){R("failed to asynchronously prepare wasm: "+We),Ja(We)})}function wt(){return!S&&typeof WebAssembly.instantiateStreaming=="function"&&!wa(Xt)&&!Po(Xt)&&!f&&typeof fetch=="function"?fetch(Xt,{credentials:"same-origin"}).then(function(Ue){var We=WebAssembly.instantiateStreaming(Ue,X);return We.then(we,function(Pt){return R("wasm streaming compile failed: "+Pt),R("falling back to ArrayBuffer instantiation"),He(we)})}):He(we)}if(s.instantiateWasm)try{var kt=s.instantiateWasm(X,se);return kt}catch(Ue){R("Module.instantiateWasm callback failed with error: "+Ue),o(Ue)}return wt().catch(o),{}}var I1,Br;function Oo(X){this.name="ExitStatus",this.message="Program terminated with exit("+X+")",this.status=X}function yr(X){for(;X.length>0;)X.shift()(s)}function S1(X){return X}function T1(X){var se=/\b_Z[\w\d_]+/g;return X.replace(se,function(we){var He=we;return we===He?we:He+" ["+we+"]"})}function Lo(){var X=new Error;if(!X.stack){try{throw new Error}catch(se){X=se}if(!X.stack)return"(no stack trace available)"}return X.stack.toString()}function vg(X,se){le.set(X,se)}function xd(){Ja("")}function fp(){return 2147483648}function yn(){return fp()}function vd(X,se,we){ie.copyWithin(X,se,se+we)}function wg(X){try{return B.grow(X-te.byteLength+65535>>>16),Ve(B.buffer),1}catch(se){}}function kg(X){var se=ie.length;X=X>>>0;var we=fp();if(X>we)return!1;let He=(Pt,oa)=>Pt+(oa-Pt%oa)%oa;for(var wt=1;wt<=4;wt*=2){var kt=se*(1+.2/wt);kt=Math.min(kt,X+100663296);var Ue=Math.min(we,He(Math.max(X,kt),65536)),We=wg(Ue);if(We)return!0}return!1}var Ig={varargs:void 0,get:function(){Ig.varargs+=4;var X=xe[Ig.varargs-4>>2];return X},getStr:function(X){var se=Q(X);return se}};function N1(X){return 52}function Sg(X,se,we,He,wt){return 70}var Rs=[null,[],[]];function Tg(X,se){var we=Rs[X];se===0||se===10?((X===1?A:R)(Z(we,0)),we.length=0):we.push(se)}function Ng(X,se,we,He){for(var wt=0,kt=0;kt<we;kt++){var Ue=Ie[se>>2],We=Ie[se+4>>2];se+=8;for(var Pt=0;Pt<We;Pt++)Tg(X,ie[Ue+Pt]);wt+=We}return Ie[He>>2]=wt,0}function gp(X){var se=s["_"+X];return se}function wd(X,se,we,He,wt){var kt={string:Mn=>{var Ur=0;if(Mn!=null&&Mn!==0){var Od=(Mn.length<<2)+1;Ur=wp(Od),ae(Mn,Ur,Od)}return Ur},array:Mn=>{var Ur=wp(Mn.length);return vg(Mn,Ur),Ur}};function Ue(Mn){return se==="string"?Q(Mn):se==="boolean"?Boolean(Mn):Mn}var We=gp(X),Pt=[],oa=0;if(He)for(var br=0;br<He.length;br++){var Pd=kt[we[br]];Pd?(oa===0&&(oa=Dd()),Pt[br]=Pd(He[br])):Pt[br]=He[br]}var kp=We.apply(null,Pt);function Qy(Mn){return oa!==0&&Rd(oa),Ue(Mn)}return kp=Qy(kp),kp}function kd(X,se,we,He){we=we||[];var wt=we.every(Ue=>Ue==="number"||Ue==="boolean"),kt=se!=="string";return kt&&wt&&!He?gp(X):function(){return wd(X,se,we,arguments,He)}}var yp={abort:xd,emscripten_get_heap_max:yn,emscripten_memcpy_big:vd,emscripten_resize_heap:kg,fd_close:N1,fd_seek:Sg,fd_write:Ng},C1=xg(),Id=s.___wasm_call_ctors=function(){return(Id=s.___wasm_call_ctors=s.asm.__wasm_call_ctors).apply(null,arguments)},Cg=s._init=function(){return(Cg=s._init=s.asm.init).apply(null,arguments)},$e=s._init_with_threads_count=function(){return($e=s._init_with_threads_count=s.asm.init_with_threads_count).apply(null,arguments)},bp=s._get_threads_count=function(){return(bp=s._get_threads_count=s.asm.get_threads_count).apply(null,arguments)},_g=s._register_tensor=function(){return(_g=s._register_tensor=s.asm.register_tensor).apply(null,arguments)},_1=s._dispose_data=function(){return(_1=s._dispose_data=s.asm.dispose_data).apply(null,arguments)},E1=s._dispose=function(){return(E1=s._dispose=s.asm.dispose).apply(null,arguments)},Eg=s._Abs=function(){return(Eg=s._Abs=s.asm.Abs).apply(null,arguments)},Sd=s._Add=function(){return(Sd=s._Add=s.asm.Add).apply(null,arguments)},zo=s._AddN=function(){return(zo=s._AddN=s.asm.AddN).apply(null,arguments)},Ag=s._All=function(){return(Ag=s._All=s.asm.All).apply(null,arguments)},$g=s._Any=function(){return($g=s._Any=s.asm.Any).apply(null,arguments)},A1=s._ArgMax=function(){return(A1=s._ArgMax=s.asm.ArgMax).apply(null,arguments)},Fg=s._AvgPool=function(){return(Fg=s._AvgPool=s.asm.AvgPool).apply(null,arguments)},Dg=s._BatchMatMul=function(){return(Dg=s._BatchMatMul=s.asm.BatchMatMul).apply(null,arguments)},Rg=s._Ceil=function(){return(Rg=s._Ceil=s.asm.Ceil).apply(null,arguments)},Mg=s._ClipByValue=function(){return(Mg=s._ClipByValue=s.asm.ClipByValue).apply(null,arguments)},Td=s._Conv2D=function(){return(Td=s._Conv2D=s.asm.Conv2D).apply(null,arguments)},Nd=s._Conv2DBackpropInput=function(){return(Nd=s._Conv2DBackpropInput=s.asm.Conv2DBackpropInput).apply(null,arguments)},Pg=s._Cos=function(){return(Pg=s._Cos=s.asm.Cos).apply(null,arguments)},Og=s._Cosh=function(){return(Og=s._Cosh=s.asm.Cosh).apply(null,arguments)},Lg=s._CropAndResize=function(){return(Lg=s._CropAndResize=s.asm.CropAndResize).apply(null,arguments)},xp=s._Cumprod=function(){return(xp=s._Cumprod=s.asm.Cumprod).apply(null,arguments)},zg=s._Cumsum=function(){return(zg=s._Cumsum=s.asm.Cumsum).apply(null,arguments)},Wg=s._DepthToSpace=function(){return(Wg=s._DepthToSpace=s.asm.DepthToSpace).apply(null,arguments)},Bg=s._DepthwiseConv2dNative=function(){return(Bg=s._DepthwiseConv2dNative=s.asm.DepthwiseConv2dNative).apply(null,arguments)},Ms=s._Elu=function(){return(Ms=s._Elu=s.asm.Elu).apply(null,arguments)},Vg=s._Equal=function(){return(Vg=s._Equal=s.asm.Equal).apply(null,arguments)},Ug=s._Exp=function(){return(Ug=s._Exp=s.asm.Exp).apply(null,arguments)},Cd=s._FlipLeftRight=function(){return(Cd=s._FlipLeftRight=s.asm.FlipLeftRight).apply(null,arguments)},Gg=s._Floor=function(){return(Gg=s._Floor=s.asm.Floor).apply(null,arguments)},Wo=s._FloorDiv=function(){return(Wo=s._FloorDiv=s.asm.FloorDiv).apply(null,arguments)},Hg=s._FusedBatchNorm=function(){return(Hg=s._FusedBatchNorm=s.asm.FusedBatchNorm).apply(null,arguments)},jg=s._FusedConv2D=function(){return(jg=s._FusedConv2D=s.asm.FusedConv2D).apply(null,arguments)},Vr=s._FusedDepthwiseConv2D=function(){return(Vr=s._FusedDepthwiseConv2D=s.asm.FusedDepthwiseConv2D).apply(null,arguments)},vp=s._Gather=function(){return(vp=s._Gather=s.asm.Gather).apply(null,arguments)},qg=s._GatherNd=function(){return(qg=s._GatherNd=s.asm.GatherNd).apply(null,arguments)},Kg=s._Greater=function(){return(Kg=s._Greater=s.asm.Greater).apply(null,arguments)},Xg=s._GreaterEqual=function(){return(Xg=s._GreaterEqual=s.asm.GreaterEqual).apply(null,arguments)},Yg=s._LeakyRelu=function(){return(Yg=s._LeakyRelu=s.asm.LeakyRelu).apply(null,arguments)},_d=s._Less=function(){return(_d=s._Less=s.asm.Less).apply(null,arguments)},Ed=s._LessEqual=function(){return(Ed=s._LessEqual=s.asm.LessEqual).apply(null,arguments)},Zg=s._Log=function(){return(Zg=s._Log=s.asm.Log).apply(null,arguments)},Jg=s._LogicalAnd=function(){return(Jg=s._LogicalAnd=s.asm.LogicalAnd).apply(null,arguments)},Ad=s._LogicalNot=function(){return(Ad=s._LogicalNot=s.asm.LogicalNot).apply(null,arguments)},$d=s._LogicalOr=function(){return($d=s._LogicalOr=s.asm.LogicalOr).apply(null,arguments)},Qg=s._LogicalXor=function(){return(Qg=s._LogicalXor=s.asm.LogicalXor).apply(null,arguments)},ey=s._Max=function(){return(ey=s._Max=s.asm.Max).apply(null,arguments)},ty=s._MaxPool=function(){return(ty=s._MaxPool=s.asm.MaxPool).apply(null,arguments)},Fd=s._Maximum=function(){return(Fd=s._Maximum=s.asm.Maximum).apply(null,arguments)},$1=s._Mean=function(){return($1=s._Mean=s.asm.Mean).apply(null,arguments)},ny=s._Min=function(){return(ny=s._Min=s.asm.Min).apply(null,arguments)},ay=s._Minimum=function(){return(ay=s._Minimum=s.asm.Minimum).apply(null,arguments)},ry=s._MirrorPad=function(){return(ry=s._MirrorPad=s.asm.MirrorPad).apply(null,arguments)},sy=s._Multiply=function(){return(sy=s._Multiply=s.asm.Multiply).apply(null,arguments)},iy=s._Neg=function(){return(iy=s._Neg=s.asm.Neg).apply(null,arguments)},oy=s._NonMaxSuppressionV3=function(){return(oy=s._NonMaxSuppressionV3=s.asm.NonMaxSuppressionV3).apply(null,arguments)},ly=s._NonMaxSuppressionV4=function(){return(ly=s._NonMaxSuppressionV4=s.asm.NonMaxSuppressionV4).apply(null,arguments)},uy=s._NonMaxSuppressionV5=function(){return(uy=s._NonMaxSuppressionV5=s.asm.NonMaxSuppressionV5).apply(null,arguments)},py=s._NotEqual=function(){return(py=s._NotEqual=s.asm.NotEqual).apply(null,arguments)},cy=s._OneHot=function(){return(cy=s._OneHot=s.asm.OneHot).apply(null,arguments)},dy=s._PadV2=function(){return(dy=s._PadV2=s.asm.PadV2).apply(null,arguments)},hy=s._Pow=function(){return(hy=s._Pow=s.asm.Pow).apply(null,arguments)},my=s._Prelu=function(){return(my=s._Prelu=s.asm.Prelu).apply(null,arguments)},fy=s._Prod=function(){return(fy=s._Prod=s.asm.Prod).apply(null,arguments)},gy=s._RealDiv=function(){return(gy=s._RealDiv=s.asm.RealDiv).apply(null,arguments)},yy=s._Relu=function(){return(yy=s._Relu=s.asm.Relu).apply(null,arguments)},by=s._Relu6=function(){return(by=s._Relu6=s.asm.Relu6).apply(null,arguments)},xy=s._ResizeBilinear=function(){return(xy=s._ResizeBilinear=s.asm.ResizeBilinear).apply(null,arguments)},vy=s._ResizeNearestNeighbor=function(){return(vy=s._ResizeNearestNeighbor=s.asm.ResizeNearestNeighbor).apply(null,arguments)},wy=s._Reverse=function(){return(wy=s._Reverse=s.asm.Reverse).apply(null,arguments)},ky=s._RotateWithOffset=function(){return(ky=s._RotateWithOffset=s.asm.RotateWithOffset).apply(null,arguments)},Iy=s._Round=function(){return(Iy=s._Round=s.asm.Round).apply(null,arguments)},Sy=s._Rsqrt=function(){return(Sy=s._Rsqrt=s.asm.Rsqrt).apply(null,arguments)},Ty=s._ScatterNd=function(){return(Ty=s._ScatterNd=s.asm.ScatterNd).apply(null,arguments)},Ny=s._SelectV2=function(){return(Ny=s._SelectV2=s.asm.SelectV2).apply(null,arguments)},Cy=s._Sigmoid=function(){return(Cy=s._Sigmoid=s.asm.Sigmoid).apply(null,arguments)},_y=s._Sin=function(){return(_y=s._Sin=s.asm.Sin).apply(null,arguments)},Ey=s._Softmax=function(){return(Ey=s._Softmax=s.asm.Softmax).apply(null,arguments)},Ay=s._SparseFillEmptyRows=function(){return(Ay=s._SparseFillEmptyRows=s.asm.SparseFillEmptyRows).apply(null,arguments)},$y=s._SparseReshape=function(){return($y=s._SparseReshape=s.asm.SparseReshape).apply(null,arguments)},Fy=s._SparseSegmentReduction=function(){return(Fy=s._SparseSegmentReduction=s.asm.SparseSegmentReduction).apply(null,arguments)},Dy=s._Sqrt=function(){return(Dy=s._Sqrt=s.asm.Sqrt).apply(null,arguments)},Ry=s._Square=function(){return(Ry=s._Square=s.asm.Square).apply(null,arguments)},My=s._SquaredDifference=function(){return(My=s._SquaredDifference=s.asm.SquaredDifference).apply(null,arguments)},Py=s._Step=function(){return(Py=s._Step=s.asm.Step).apply(null,arguments)},Oy=s._StridedSlice=function(){return(Oy=s._StridedSlice=s.asm.StridedSlice).apply(null,arguments)},Ly=s._Sub=function(){return(Ly=s._Sub=s.asm.Sub).apply(null,arguments)},zy=s._Sum=function(){return(zy=s._Sum=s.asm.Sum).apply(null,arguments)},Wy=s._Tan=function(){return(Wy=s._Tan=s.asm.Tan).apply(null,arguments)},By=s._Tanh=function(){return(By=s._Tanh=s.asm.Tanh).apply(null,arguments)},Vy=s._Tile=function(){return(Vy=s._Tile=s.asm.Tile).apply(null,arguments)},Uy=s._TopK=function(){return(Uy=s._TopK=s.asm.TopK).apply(null,arguments)},Gy=s._Transform=function(){return(Gy=s._Transform=s.asm.Transform).apply(null,arguments)},Hy=s._Transpose=function(){return(Hy=s._Transpose=s.asm.Transpose).apply(null,arguments)},jy=s.__FusedMatMul=function(){return(jy=s.__FusedMatMul=s.asm._FusedMatMul).apply(null,arguments)},qy=s._malloc=function(){return(qy=s._malloc=s.asm.malloc).apply(null,arguments)},Ky=s._free=function(){return(Ky=s._free=s.asm.free).apply(null,arguments)},Xy=s.___errno_location=function(){return(Xy=s.___errno_location=s.asm.__errno_location).apply(null,arguments)},Dd=s.stackSave=function(){return(Dd=s.stackSave=s.asm.stackSave).apply(null,arguments)},Rd=s.stackRestore=function(){return(Rd=s.stackRestore=s.asm.stackRestore).apply(null,arguments)},wp=s.stackAlloc=function(){return(wp=s.stackAlloc=s.asm.stackAlloc).apply(null,arguments)},Yy=s.dynCall_iijjiiii=function(){return(Yy=s.dynCall_iijjiiii=s.asm.dynCall_iijjiiii).apply(null,arguments)},Zy=s.dynCall_jiji=function(){return(Zy=s.dynCall_jiji=s.asm.dynCall_jiji).apply(null,arguments)};s.cwrap=kd;var Bo;jn=function X(){Bo||Md(),Bo||(jn=X)};function Md(X){if(X=X||p,it>0||(Mt(),it>0))return;function se(){Bo||(Bo=!0,s.calledRun=!0,!U&&(sa(),i(s),s.onRuntimeInitialized&&s.onRuntimeInitialized(),sn()))}s.setStatus?(s.setStatus("Running..."),setTimeout(function(){setTimeout(function(){s.setStatus("")},1),se()},1)):se()}if(s.preInit)for(typeof s.preInit=="function"&&(s.preInit=[s.preInit]);s.preInit.length>0;)s.preInit.pop()();Md();var Vo;l&&(Vo={uncaughtException:process.listeners("uncaughtException").filter(function(X){return!l.uncaughtException.indexOf(X)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(X){return!l.unhandledRejection.indexOf(X)>-1})});var Uo;if(typeof r!="undefined")Uo=r;else if(typeof WasmBackendModuleThreadedSimd!="undefined")Uo=WasmBackendModuleThreadedSimd;else throw new Error("Could not find wasm module in post.js");if(Vo){var Jy=Uo._dispose;Uo._dispose=function(){Jy(),Vo.uncaughtException.forEach(function(X){process.removeListener("uncaughtException",X)}),Vo.unhandledRejection.forEach(function(X){process.removeListener("unhandledRejection",X)})}}return r.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModule=n)}),jh=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},rc=class{refCount(e){return qn("refCount")}incRef(e){return qn("incRef")}timerAvailable(){return!0}time(e){return qn("time")}read(e){return qn("read")}readSync(e){return qn("readSync")}readToGPU(e,t){return qn("readToGPU")}numDataIds(){return qn("numDataIds")}disposeData(e,t){return qn("disposeData")}write(e,t,n){return qn("write")}move(e,t,n,a,r){return qn("move")}createTensorFromTexture(e,t,n){return qn("createTensorFromTexture")}memory(){return qn("memory")}floatPrecision(){return qn("floatPrecision")}epsilon(){return this.floatPrecision()===32?1e-7:1e-4}dispose(){return qn("dispose")}};function qn(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function kI(e){let t=e.length,n=0;for(;t>0;)n=Math.random()*t|0,t--,hh(e,t,n)}function CF(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let n=e.length,a=0;for(;n>0;)a=Math.random()*n|0,n--,hh(e,n,a),hh(t,n,a)}function Vp(e,t,n){return Math.max(e,Math.min(t,n))}function _F(e){return e%2===0?e:e+1}function hh(e,t,n){let a=e[t];e[t]=e[n],e[n]=a}function EF(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function AF(e,t){let n=Math.random();return t*n+(1-n)*e}function $F(e,t){let n=0;for(let a=0;a<e.length;a++){let r=Number(e[a])-Number(t[a]);n+=r*r}return n}function $(e,t){if(!e)throw new Error(typeof t=="string"?t:t())}function Sn(e,t,n=""){$(ps(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function hi(e){$(e!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Zs(e,t=[],n=!1){if(t==null&&(t=[]),Array.isArray(e)||hn(e)&&!n)for(let a=0;a<e.length;++a)Zs(e[a],t,n);else t.push(e);return t}function mt(e){if(e.length===0)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function FF(e){return e.length===0}function ps(e,t){if(e===t)return!0;if(e==null||t==null||e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function tl(e){return e%1===0}function DF(e){if(Math.tanh!=null)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{let t=Math.exp(2*e);return(t-1)/(t+1)}}function RF(e){let t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function MF(e){let t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return kI(t),t}function Lp(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function PF(e,t=r=>0,n,a){return new Promise((r,s)=>{let i=0,o=()=>{if(e()){r();return}i++;let l=t(i);if(n!=null&&i>=n){s();return}a!=null?a(o,l):setTimeout(o,l)};o()})}function OF(e,t){let n=1,a=-1;for(let s=0;s<e.length;++s)if(e[s]>=0)n*=e[s];else if(e[s]===-1){if(a!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${a} and dim ${s}`);a=s}else if(e[s]<0)throw Error(`Shapes can not be < 0. Found ${e[s]} at dim ${s}`);if(a===-1){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(n===0)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);let r=e.slice();return r[a]=t/n,r}function Aa(e,t){let n=t.length;return e=e==null?t.map((a,r)=>r):[].concat(e),$(e.every(a=>a>=-n&&a<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),$(e.every(a=>tl(a)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(a=>a<0?n+a:a)}function II(e,t){let n=[],a=[],r=t!=null&&Array.isArray(t)&&t.length===0,s=t==null||r?null:Aa(t,e).sort(),i=0;for(let o=0;o<e.length;++o){if(s!=null){if(s[i]===o&&e[o]!==1)throw new Error(`Can't squeeze axis ${o} since its dim '${e[o]}' is not 1`);(s[i]==null||s[i]>o)&&e[o]===1&&(n.push(e[o]),a.push(o)),s[i]<=o&&i++}e[o]!==1&&(n.push(e[o]),a.push(o))}return{newShape:n,keptDims:a}}function SI(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else throw new Error(`Unknown data type ${e}`);return n}function TI(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else if(e==="string")n=new Array(t);else throw new Error(`Unknown data type ${e}`);return n}function NI(e,t){for(let n=0;n<e.length;n++){let a=e[n];if(isNaN(a)||!isFinite(a))throw Error(`A tensor of type ${t} being uploaded contains ${a}.`)}}function CI(e){return e==="bool"||e==="complex64"||e==="float32"||e==="int32"||e==="string"}function LF(e,t){return!(t==="complex64"||t==="float32"&&e!=="complex64"||t==="int32"&&e!=="float32"&&e!=="complex64"||t==="bool"&&e==="bool")}function hn(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}function bb(e){if(e==="float32"||e==="int32")return 4;if(e==="complex64")return 8;if(e==="bool")return 1;throw new Error(`Unknown dtype ${e}`)}function _I(e){if(e==null)return 0;let t=0;return e.forEach(n=>t+=n.length),t}function Kr(e){return typeof e=="string"||e instanceof String}function EI(e){return typeof e=="boolean"}function AI(e){return typeof e=="number"}function qh(e){return Array.isArray(e)?qh(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":AI(e)?"float32":Kr(e)?"string":EI(e)?"bool":"float32"}function es(e){return!!(e&&e.constructor&&e.call&&e.apply)}function mh(e,t){for(let n=t;n<e;++n)if(e%n===0)return n;return e}function vl(e){let t=e.length;if(t<2)return[];let n=new Array(t-1);n[t-2]=e[t-1];for(let a=t-3;a>=0;--a)n[a]=n[a+1]*e[a+1];return n}function $I(e,t,n,a=!1){let r=new Array;if(t.length===1){let s=t[0]*(a?2:1);for(let i=0;i<s;i++)r[i]=n[e+i]}else{let s=t[0],i=t.slice(1),o=i.reduce((l,u)=>l*u)*(a?2:1);for(let l=0;l<s;l++)r[l]=$I(e+l*o,i,n,a)}return r}function Zo(e,t,n=!1){if(e.length===0)return t[0];let a=e.reduce((r,s)=>r*s)*(n?2:1);if(a===0)return[];if(a!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);return $I(0,e,t,n)}function vx(e,t){let n=Kh(e,t);for(let a=0;a<n.length;a++)n[a]=1;return n}function Kh(e,t){if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool")return new Uint8Array(e);throw new Error(`Unknown data type ${t}`)}function zF(e,t){let n=e.reduce((a,r)=>a*r,1);if(t==null||t==="float32")return Zo(e,new Float32Array(n));if(t==="int32")return Zo(e,new Int32Array(n));if(t==="bool")return Zo(e,new Uint8Array(n));throw new Error(`Unknown data type ${t}`)}function wx(e){e.forEach(t=>{$(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function WF(e,t,n){if(t===0)return 0;if(t===1)return e[0];let a=e[e.length-1];for(let r=0;r<e.length-1;++r)a+=n[r]*e[r];return a}function BF(e,t,n){if(t===0)return[];if(t===1)return[e];let a=new Array(t);for(let r=0;r<a.length-1;++r)a[r]=Math.floor(e/n[r]),e-=a[r]*n[r];return a[a.length-1]=e,a}function kx(e){return e&&e.then&&typeof e.then=="function"}var W1="tfjsflags",FI=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=VF,this.populateURLFlags()}setPlatform(e,t){this.platform!=null&&(H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},this.urlFlags[e]!=null){let a=this.urlFlags[e];H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${a}.`),this.set(e,a)}}async getAsync(e){return e in this.flags?this.flags[e]:(this.flags[e]=await this.evaluateFlag(e),this.flags[e])}get(e){if(e in this.flags)return this.flags[e];let t=this.evaluateFlag(e);if(kx(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(this.flagRegistry[e]==null)throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,this.flagRegistry[e].setHook!=null&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(this.flagRegistry[e]==null)throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);return this.flagRegistry[e].evaluationFn()}setFlags(e){this.flags=Object.assign({},e)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global=="undefined"||typeof this.global.location=="undefined"||typeof this.global.location.search=="undefined")return;let e=this.getQueryParams(this.global.location.search);W1 in e&&e[W1].split(",").forEach(t=>{let[n,a]=t.split(":");this.urlFlags[n]=GF(n,a)})}};function VF(e){let t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(n,...a)=>(UF(t,a[0],a[1]),a.join("="))),t}function UF(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}function GF(e,t){if(t=t.toLowerCase(),t==="true"||t==="false")return t==="true";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)}function H(){return Ix}var Ix=null;function HF(e){Ix=e}var rb;function DI(){if(rb==null){let e;if(typeof window!="undefined")e=window;else if(typeof global!="undefined")e=global;else if(typeof process!="undefined")e=process;else if(typeof self!="undefined")e=self;else throw new Error("Could not find a global object");rb=e}return rb}function jF(){let e=DI();return e._tfGlobals==null&&(e._tfGlobals=new Map),e._tfGlobals}function Sx(e,t){let n=jF();if(n.has(e))return n.get(e);{let a=t();return n.set(e,a),n.get(e)}}var wl="Abs",kl="Acos",Il="Acosh",cs="Add",mi="AddN",Sl="All",Tl="Any",fi="ArgMax",sc="ArgMin",Nl="Asin",Cl="Asinh",_l="Atan",El="Atanh",Al="Atan2",gi="AvgPool",Xh="AvgPoolGrad",ic="AvgPool3D",Yh="AvgPool3DGrad",yi="BatchMatMul",$l="BatchToSpaceND",Zh="Bincount",RI="BroadcastTo",Jh="BroadcastArgs",bi="Cast",xi="Ceil",ds="ClipByValue",Qh="Complex",oc="ComplexAbs",Fl="Concat",vi="Conv2D",em="Conv2DBackpropFilter",wi="Conv2DBackpropInput",lc="Conv3D",tm="Conv3DBackpropFilterV2",nm="Conv3DBackpropInputV2",ki="Cos",Ii="Cosh",Dl="Cumprod",Si="Cumsum",Rl="CropAndResize",am="DenseBincount",Ml="DepthToSpace",Ti="DepthwiseConv2dNative",rm="DepthwiseConv2dNativeBackpropFilter",sm="DepthwiseConv2dNativeBackpropInput",im="Diag",uc="Dilation2D",fh="Dilation2DBackpropInput",gh="Dilation2DBackpropFilter",Ni="RealDiv",om="Einsum",Ci="Elu",lm="EluGrad",Pl="Erf",Ol="Equal",_i="Exp",Ll="ExpandDims",zl="Expm1",um="FFT",pc="Fill",Wl="FlipLeftRight",Ei="Floor",Ai="FloorDiv",$i="FusedBatchNorm",Bl="GatherV2",Vl="GatherNd",Ul="Greater",Fi="GreaterEqual",Di="Identity",pm="IFFT",cm="Imag",Gl="IsFinite",Hl="IsInf",jl="IsNan",Ri="LeakyRelu",ql="Less",Kl="LessEqual",dm="LinSpace",Mi="Log",Xl="Log1p",Yl="LogicalAnd",Zl="LogicalNot",Jl="LogicalOr",MI="LogicalXor",PI="LogSoftmax",qF="LowerBound",cc="LRN",hm="LRNGrad",Pi="Max",Oi="Maximum",Li="MaxPool",mm="MaxPoolGrad",dc="MaxPool3D",fm="MaxPool3DGrad",gm="MaxPoolWithArgmax",zi="Mean",Wi="Min",Bi="Minimum",Vi="MirrorPad",Ql="Mod",ym="Multinomial",Ui="Multiply",eu="Neg",tu="NotEqual",nu="NonMaxSuppressionV3",au="NonMaxSuppressionV4",ru="NonMaxSuppressionV5",su="OnesLike",Gi="OneHot",iu="Pack",Hi="PadV2",KF="Pool",ji="Pow",qi="Prelu",Ki="Prod",bm="RaggedGather",xm="RaggedRange",vm="RaggedTensorToTensor",hc="Range",wm="Real",ou="Reciprocal",Xi="Relu",lu="Reshape",Yi="ResizeNearestNeighbor",km="ResizeNearestNeighborGrad",Zi="ResizeBilinear",Im="ResizeBilinearGrad",Ji="Relu6",Qi="Reverse",eo="Round",to="Rsqrt",uu="ScatterNd",Sm="SearchSorted",pu="Select",cu="Selu",du="Slice",no="Sin",hu="Sinh",mu="Sign",ao="Sigmoid",fu="Softplus",ro="Sqrt",so="Sum",gu="SpaceToBatchND",yu="SplitV",io="Softmax",mc="SparseFillEmptyRows",bu="SparseReshape",fc="SparseSegmentMean",gc="SparseSegmentSum",Tm="SparseToDense",oo="SquaredDifference",yc="Square",xu="StridedSlice",bc="StringNGrams",xc="StringSplit",vc="StringToHashBucketFast",lo="Sub",uo="Tan",po="Tanh",hs="Tile",vu="TopK",wu="Transform",Tr="Transpose",Nm="Unique",ku="Unpack",wc="UnsortedSegmentSum",XF="UpperBound",Iu="ZerosLike",ms="Step",yh="FromPixels",Su="RotateWithOffset",Js="_FusedMatMul",Qs="FusedConv2D",ei="FusedDepthwiseConv2D";function qr(...e){H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(...e)}function YF(...e){H().getBool("IS_TEST")||H().getBool("PROD")||console.log(...e)}var nl=Sx("kernelRegistry",()=>new Map),Up=Sx("gradRegistry",()=>new Map);function bh(e,t){let n=Tx(e,t);return nl.get(n)}function xb(e){return Up.get(e)}function xh(e){let t=nl.entries(),n=[];for(;;){let{done:a,value:r}=t.next();if(a)break;let[s,i]=r,[o]=s.split("_");o===e&&n.push(i)}return n}function kc(e){let{kernelName:t,backendName:n}=e,a=Tx(t,n);nl.has(a)&&qr(`The kernel '${t}' for backend '${n}' is already registered`),nl.set(a,e)}function OI(e){let{kernelName:t}=e;Up.has(t)&&H().getBool("DEBUG")&&qr(`Overriding the gradient for '${t}'`),Up.set(t,e)}function ZF(e,t){let n=Tx(e,t);if(!nl.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);nl.delete(n)}function JF(e){if(!Up.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);Up.delete(e)}function QF(e,t){xh(e).forEach(n=>{let a=Object.assign({},n,{backendName:t});kc(a)})}function Tx(e,t){return`${t}_${e}`}var v={};Ae(v,{arraysEqual:()=>ps,assert:()=>$,assertNonNegativeIntegerDimensions:()=>wx,assertNonNull:()=>hi,assertShapesMatch:()=>Sn,bytesFromStringArray:()=>_I,bytesPerElement:()=>bb,checkConversionForErrors:()=>NI,clamp:()=>Vp,computeStrides:()=>vl,createScalarValue:()=>sD,createShuffledIndices:()=>MF,decodeString:()=>vh,distSquared:()=>$F,encodeString:()=>Sc,fetch:()=>oD,fingerPrint64:()=>rD,flatten:()=>Zs,getArrayFromDType:()=>TI,getTypedArrayFromDType:()=>SI,hasEncodingLoss:()=>LF,hexToLong:()=>Ic,indexToLoc:()=>BF,inferDtype:()=>qh,inferFromImplicitShape:()=>OF,isBoolean:()=>EI,isFunction:()=>es,isInt:()=>tl,isNumber:()=>AI,isPromise:()=>kx,isScalarShape:()=>FF,isString:()=>Kr,isTypedArray:()=>hn,isValidDtype:()=>CI,locToIndex:()=>WF,makeOnesTypedArray:()=>vx,makeZerosNestedTypedArray:()=>zF,makeZerosTypedArray:()=>Kh,nearestDivisor:()=>mh,nearestLargerEven:()=>_F,now:()=>Gp,parseAxisParam:()=>Aa,randUniform:()=>AF,repeatedTry:()=>PF,rightPad:()=>Lp,shuffle:()=>kI,shuffleCombo:()=>CF,sizeFromShape:()=>mt,sizeToSquarishShape:()=>RF,squeezeShape:()=>II,sum:()=>EF,swap:()=>hh,tanh:()=>DF,toNestedArray:()=>Zo,toTypedArray:()=>Cm});var B1=us(pF()),Ws=B1.default||B1;function Ic(e){return Ws.fromString(e,!0,16)}var LI=Ic("c3a5c85c97cb3127"),Ls=Ic("b492b66fbe98f273"),bn=Ic("9ae16a3b2f90404f");function vb(e){return e.xor(e.shru(47))}function zI(e,t,n){let a=e.slice(t,t+n);return Ws.fromBytes(Array.from(a),!0,!0)}function ht(e,t){return zI(e,t,8)}function V1(e,t){return zI(e,t,4)}function Yt(e,t){return t===0?e:e.shru(t).or(e.shl(64-t))}function Zr(e,t,n=Ic("9ddfea08eb382d69")){let a=e.xor(t).mul(n);a=a.xor(a.shru(47));let r=t.xor(a).mul(n);return r=r.xor(r.shru(47)),r=r.mul(n),r}function eD(e,t,n,a,r,s){r=r.add(e),s=Yt(s.add(r).add(a),21);let i=r;return r=r.add(t),r=r.add(n),s=s.add(Yt(r,44)),[r.add(a),s.add(i)]}function Gd(e,t,n,a){return eD(ht(e,t),ht(e,t+8),ht(e,t+16),ht(e,t+24),n,a)}function tD(e,t=e.length){if(t>=8){let n=bn.add(t*2),a=ht(e,0).add(bn),r=ht(e,t-8),s=Yt(r,37).mul(n).add(a),i=Yt(a,25).add(r).mul(n);return Zr(s,i,n)}if(t>=4){let n=bn.add(t*2),a=V1(e,0);return Zr(a.shl(3).add(t),V1(e,t-4),n)}if(t>0){let n=e[0],a=e[t>>1],r=e[t-1],s=n+(a<<8),i=t+(r<<2);return vb(bn.mul(s).xor(LI.mul(i))).mul(bn)}return bn}function nD(e,t=e.length){let n=bn.add(t*2),a=ht(e,0).mul(Ls),r=ht(e,8),s=ht(e,t-8).mul(n),i=ht(e,t-16).mul(bn);return Zr(Yt(a.add(r),43).add(Yt(s,30)).add(i),a.add(Yt(r.add(bn),18)).add(s),n)}function aD(e,t=e.length){let n=bn.add(t*2),a=ht(e,0).mul(bn),r=ht(e,8),s=ht(e,t-8).mul(n),i=ht(e,t-16).mul(bn),o=Yt(a.add(r),43).add(Yt(s,30)).add(i),l=Zr(o,a.add(Yt(r.add(bn),18)).add(s),n),u=ht(e,16).mul(n),p=ht(e,24),d=o.add(ht(e,t-32)).mul(n),c=l.add(ht(e,t-24)).mul(n);return Zr(Yt(u.add(p),43).add(Yt(d,30)).add(c),u.add(Yt(p.add(a),18)).add(d),n)}function rD(e,t=e.length){let n=Ws.fromNumber(81,!0);if(t<=32)return t<=16?tD(e,t):nD(e,t);if(t<=64)return aD(e,t);let a=n,r=n.mul(Ls).add(113),s=vb(r.mul(bn).add(113)).mul(bn),i=[Ws.UZERO,Ws.UZERO],o=[Ws.UZERO,Ws.UZERO];a=a.mul(bn).add(ht(e,0));let l=0,u=(t-1>>6)*64,p=u+(t-1&63)-63;do a=Yt(a.add(r).add(i[0]).add(ht(e,l+8)),37).mul(Ls),r=Yt(r.add(i[1]).add(ht(e,l+48)),42).mul(Ls),a=a.xor(o[1]),r=r.add(i[0]).add(ht(e,l+40)),s=Yt(s.add(o[0]),33).mul(Ls),i=Gd(e,l,i[1].mul(Ls),a.add(o[0])),o=Gd(e,l+32,s.add(o[1]),r.add(ht(e,l+16))),[s,a]=[a,s],l+=64;while(l!==u);let d=Ls.add(s.and(255).shl(1));return l=p,o[0]=o[0].add(t-1&63),i[0]=i[0].add(o[0]),o[0]=o[0].add(i[0]),a=Yt(a.add(r).add(i[0]).add(ht(e,l+8)),37).mul(d),r=Yt(r.add(i[1]).add(ht(e,l+48)),42).mul(d),a=a.xor(o[1].mul(9)),r=r.add(i[0].mul(9).add(ht(e,l+40))),s=Yt(s.add(o[0]),33).mul(d),i=Gd(e,l,i[1].mul(d),a.add(o[0])),o=Gd(e,l+32,s.add(o[1]),r.add(ht(e,l+16))),[s,a]=[a,s],Zr(Zr(i[0],o[0],d).add(vb(r).mul(LI)).add(s),Zr(i[1],o[1],d).add(a),d)}function sD(e,t){return t==="string"?Sc(e):Cm([e],t)}function iD(e,t){return e instanceof Float32Array&&t==="float32"||e instanceof Int32Array&&t==="int32"||e instanceof Uint8Array&&t==="bool"}function Cm(e,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Zs(e)),H().getBool("DEBUG")&&NI(e,t),iD(e,t))return e;if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool"){let n=new Uint8Array(e.length);for(let a=0;a<n.length;++a)Math.round(e[a])!==0&&(n[a]=1);return n}else throw new Error(`Unknown data type ${t}`)}function Gp(){return H().platform.now()}function oD(e,t){return H().platform.fetch(e,t)}function Sc(e,t="utf-8"){return t=t||"utf-8",H().platform.encode(e,t)}function vh(e,t="utf-8"){return t=t||"utf-8",H().platform.decode(e,t)}var lD=class{constructor(e,t){this.backendTimer=e,this.logger=t,t==null&&(this.logger=new pD)}profileKernel(e,t,n){let a,r=()=>{a=n()},s,i=Gp();if(this.backendTimer.timerAvailable())s=this.backendTimer.time(r);else{r();for(let o of a)o.dataSync();s=Promise.resolve({kernelMs:Gp()-i})}if(H().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let o=0;o<a.length;o++){let l=a[o];l.data().then(u=>{uD(u,l.dtype,e)})}return{kernelName:e,outputs:a,inputs:t,timeMs:s.then(o=>o.kernelMs),extraInfo:s.then(o=>o.getExtraProfileInfo!=null?o.getExtraProfileInfo():"")}}logKernelProfile(e){let{kernelName:t,outputs:n,timeMs:a,inputs:r,extraInfo:s}=e;n.forEach(i=>{Promise.all([i.data(),a,s]).then(o=>{this.logger.logKernelProfile(t,i,o[0],o[1],r,o[2])})})}};function uD(e,t,n){if(t!=="float32")return!1;for(let a=0;a<e.length;a++){let r=e[a];if(isNaN(r)||!isFinite(r))return console.warn(`Found ${r} in the result of '${n}'`),!0}return!1}var pD=class{logKernelProfile(e,t,n,a,r,s){let i=typeof a=="number"?Lp(`${a}ms`,9):a.error,o=Lp(e,25),l=t.rank,u=t.size,p=Lp(t.shape.toString(),14),d="";for(let c in r){let h=r[c];if(h!=null){let m=h.shape||t.shape,f=m.length;d+=`${c}: ${f}D ${f>0?m:""} `}}console.log(`%c${o}	%c${i}	%c${l}D ${p}	%c${u}	%c${d}	%c${s}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function cD(e,t,n){let a={},r={};for(let l=0;l<t.length;l++)a[t[l].id]=!0;for(let l=0;l<e.length;l++){let u=e[l],p=u.inputs;for(let d in p){let c=p[d],h=!1;for(let m=0;m<t.length;m++)if(a[c.id]){u.outputs.forEach(f=>a[f.id]=!0),h=!0,r[u.id]=!0;break}if(h)break}}let s={};s[n.id]=!0;let i={};for(let l=e.length-1;l>=0;l--){let u=e[l],p=u.inputs;for(let d=0;d<u.outputs.length;d++)if(s[u.outputs[d].id]){for(let c in p)s[p[c].id]=!0,i[u.id]=!0;break}}let o=[];for(let l=0;l<e.length;l++){let u=e[l];if(r[u.id]&&i[u.id]){let p={};for(let c in u.inputs){let h=u.inputs[c];a[h.id]&&(p[c]=h)}let d=Object.assign({},u);d.inputs=p,d.outputs=u.outputs,o.push(d)}}return o}function dD(e,t,n,a){for(let r=t.length-1;r>=0;r--){let s=t[r],i=[];if(s.outputs.forEach(l=>{let u=e[l.id];u!=null?i.push(u):i.push(null)}),s.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let o=s.gradient(i);for(let l in s.inputs){if(!(l in o))throw new Error(`Cannot backprop through input ${l}. Available gradients found: ${Object.keys(o)}.`);let u=n(()=>o[l]());if(u.dtype!=="float32")throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${l} must have 'float32' dtype, but has '${u.dtype}'`);let p=s.inputs[l];if(!ps(u.shape,p.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${l}' has shape '${u.shape}', which does not match the shape of the input '${p.shape}'`);if(e[p.id]==null)e[p.id]=u;else{let d=e[p.id];e[p.id]=a(d,u),d.dispose()}}}}var U1=20,Ip=3,sb=7;function hD(e,t,n,a){let r=vl(t),s=mD(e,t,n,r),i=t.length,o=ah(e,t,n,r,s),l=["Tensor"];return a&&(l.push(`  dtype: ${n}`),l.push(`  rank: ${i}`),l.push(`  shape: [${t}]`),l.push("  values:")),l.push(o.map(u=>"    "+u).join(`
-`)),l.join(`
-`)}function mD(e,t,n,a){let r=mt(t),s=a[a.length-1],i=new Array(s).fill(0),o=t.length,l=n==="complex64"?_p(e):e;if(o>1)for(let u=0;u<r/s;u++){let p=u*s;for(let d=0;d<s;d++)i[d]=Math.max(i[d],Cp(l[p+d],0,n).length)}return i}function Cp(e,t,n){let a;return Array.isArray(e)?a=`${parseFloat(e[0].toFixed(sb))} + ${parseFloat(e[1].toFixed(sb))}j`:Kr(e)?a=`'${e}'`:n==="bool"?a=WI(e):a=parseFloat(e.toFixed(sb)).toString(),Lp(a,t)}function WI(e){return e===0?"false":"true"}function ah(e,t,n,a,r,s=!0){let i=n==="complex64"?2:1,o=t[0],l=t.length;if(l===0){if(n==="complex64"){let f=_p(e);return[Cp(f[0],0,n)]}return n==="bool"?[WI(e[0])]:[e[0].toString()]}if(l===1){if(o>U1){let g=Ip*i,y=Array.from(e.slice(0,g)),b=Array.from(e.slice((o-Ip)*i,o*i));return n==="complex64"&&(y=_p(y),b=_p(b)),["["+y.map((x,w)=>Cp(x,r[w],n)).join(", ")+", ..., "+b.map((x,w)=>Cp(x,r[o-Ip+w],n)).join(", ")+"]"]}let f=n==="complex64"?_p(e):Array.from(e);return["["+f.map((g,y)=>Cp(g,r[y],n)).join(", ")+"]"]}let u=t.slice(1),p=a.slice(1),d=a[0]*i,c=[];if(o>U1){for(let f=0;f<Ip;f++){let g=f*d,y=g+d;c.push(...ah(e.slice(g,y),u,n,p,r,!1))}c.push("...");for(let f=o-Ip;f<o;f++){let g=f*d,y=g+d;c.push(...ah(e.slice(g,y),u,n,p,r,f===o-1))}}else for(let f=0;f<o;f++){let g=f*d,y=g+d;c.push(...ah(e.slice(g,y),u,n,p,r,f===o-1))}let h=l===2?",":"";c[0]="["+c[0]+h;for(let f=1;f<c.length-1;f++)c[f]=" "+c[f]+h;let m=`,
-`;for(let f=2;f<l;f++)m+=`
-`;return c[c.length-1]=" "+c[c.length-1]+"]"+(s?"":m),c}function _p(e){let t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}var Ht=class{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=mt(e),n!=null){let a=n.length;$(a===this.size,()=>`Length of values '${a}' does not match the size inferred by the shape '${this.size}'.`)}if(t==="complex64")throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=n||TI(t,this.size),this.strides=vl(e)}set(e,...t){t.length===0&&(t=[0]),$(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);let n=this.locToIndex(t);this.values[n]=e}get(...e){e.length===0&&(e=[0]);let t=0;for(let a of e){if(a<0||a>=this.shape[t]){let r=`Requested out of range element at ${e}.   Buffer shape=${this.shape}`;throw new Error(r)}t++}let n=e[e.length-1];for(let a=0;a<e.length-1;++a)n+=this.strides[a]*e[a];return this.values[n]}locToIndex(e){if(this.rank===0)return 0;if(this.rank===1)return e[0];let t=e[e.length-1];for(let n=0;n<e.length-1;++n)t+=this.strides[n]*e[n];return t}indexToLoc(e){if(this.rank===0)return[];if(this.rank===1)return[e];let t=new Array(this.shape.length);for(let n=0;n<t.length-1;++n)t[n]=Math.floor(e/this.strides[n]),e-=t[n]*this.strides[n];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return Oa().makeTensor(this.values,this.shape,this.dtype)}},Oa=null,Ko=null,fD=null;function gD(e){Oa=e}function yD(e){Ko=e}function bD(e){fD=e}var Te=class{constructor(e,t,n,a){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=mt(e),this.strides=vl(e),this.dataId=n,this.id=a,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let e=await this.data();return Ko.buffer(this.shape,this.dtype,e)}bufferSync(){return Ko.buffer(this.shape,this.dtype,this.dataSync())}async array(){let e=await this.data();return Zo(this.shape,e,this.dtype==="complex64")}arraySync(){return Zo(this.shape,this.dataSync(),this.dtype==="complex64")}async data(){this.throwIfDisposed();let e=Oa().read(this.dataId);if(this.dtype==="string"){let t=await e;try{return t.map(n=>vh(n))}catch(n){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),Oa().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();let e=Oa().readSync(this.dataId);if(this.dtype==="string")try{return e.map(t=>vh(t))}catch(t){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();let e=await Oa().read(this.dataId);return this.dtype==="string"?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(Oa().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return Ko.print(this,e)}clone(){return this.throwIfDisposed(),Ko.clone(this)}toString(e=!1){let t=this.dataSync();return hD(t,this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),Ko.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),Oa().makeVariable(this,e,t,n)}};Object.defineProperty(Te,Symbol.hasInstance,{value:e=>!!e&&e.data!=null&&e.dataSync!=null&&e.throwIfDisposed!=null});function J(){return Sx("Tensor",()=>Te)}J();var ts=class extends Te{constructor(e,t,n,a){super(e.shape,e.dtype,e.dataId,a),this.trainable=t,this.name=n}assign(e){if(e.dtype!==this.dtype)throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);if(!ps(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);Oa().disposeTensor(this),this.dataId=e.dataId,Oa().incRef(this,null)}dispose(){Oa().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(ts,Symbol.hasInstance,{value:e=>e instanceof Te&&e.assign!=null&&e.assign instanceof Function});var Va={};Ae(Va,{assertTypesMatch:()=>BI,getTensorsInContainer:()=>Nx,isTensorInList:()=>vD,makeTypesMatch:()=>_t});var wb;(function(e){e.R0="R0",e.R1="R1",e.R2="R2",e.R3="R3",e.R4="R4",e.R5="R5",e.R6="R6"})(wb||(wb={}));var kb;(function(e){e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64"})(kb||(kb={}));var Ib;(function(e){e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64"})(Ib||(Ib={}));var Sb;(function(e){e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64"})(Sb||(Sb={}));var Tb;(function(e){e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64"})(Tb||(Tb={}));var xD={float32:Sb,int32:kb,bool:Ib,complex64:Tb};function ma(e,t){if(e==="string"||t==="string"){if(e==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return xD[e][t]}function _m(e){return ma(e,"int32")}function _t(e,t){if(e.dtype===t.dtype)return[e,t];let n=ma(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function BI(e,t){$(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function vD(e,t){return t.some(n=>n.id===e.id)}function Nx(e){let t=[];return VI(e,t,new Set),t}function VI(e,t,n){if(e==null)return;if(e instanceof Te){t.push(e);return}if(!wD(e))return;let a=e;for(let r in a){let s=a[r];n.has(s)||(n.add(s),VI(s,t,n))}}function wD(e){return Array.isArray(e)||typeof e=="object"}function ib(e){return e.kernelName!=null}var G1=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(let e in this.registeredVariables)this.registeredVariables[e].dispose()}},Hp=class{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new G1}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t];if(await this.initializeBackend(n).success){await this.setBackend(n);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry))if(e in this.registryFactory){let{asyncInit:t}=this.initializeBackend(e);if(t)return null}else return null;return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(qr(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:n},!0)}async setBackend(e){if(this.registryFactory[e]==null)throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,this.registry[e]==null){this.backendInstance=null;let{success:t,asyncInit:n}=this.initializeBackend(e);if(!(n?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new lD(this.backendInstance),!0}setupRegisteredKernels(){xh(this.backendName).forEach(e=>{e.setupFunc!=null&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){xh(e).forEach(t=>{t.disposeFunc!=null&&t.disposeFunc(this.registry[e])})}initializeBackend(e){let t=this.registryFactory[e];if(t==null)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{let n=t.factory();if(n&&!(n instanceof rc)&&typeof n.then=="function"){let a=++this.pendingBackendInitId,r=n.then(s=>a<this.pendingBackendInitId?!1:(this.registry[e]=s,this.pendingBackendInit=null,!0)).catch(s=>(a<this.pendingBackendInitId||(this.pendingBackendInit=null,qr(`Initialization of backend ${e} failed`),qr(s.stack||s.message)),!1));return this.pendingBackendInit=r,{success:r,asyncInit:!0}}else return this.registry[e]=n,{success:!0,asyncInit:!1}}catch(n){return qr(`Initialization of backend ${e} failed`),qr(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${e} backend not found in registry`);this.backendName===e&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,e in this.registry&&(this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e]),delete this.registryFactory[e],this.backendName===e&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t],{success:a,asyncInit:r}=this.initializeBackend(n);if(r||a)return{name:n,asyncInit:r}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){let n=this.state.tensorInfo.get(t),a=n.backend,r=this.readSync(t),s=a.refCount(t);a.disposeData(t,!0),n.backend=e,e.move(t,r,n.shape,n.dtype,s),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n=null;if(t==null){if(typeof e!="function")throw new Error("Please provide a function to tidy()");t=e}else{if(typeof e!="string"&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if(typeof t!="function")throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");n=e}let a;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(a),()=>(a=t(),a instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),a))}scopedRun(e,t,n){e();try{let a=n();return t(),a}catch(a){throw t(),a}}nextTensorId(){return Hp.nextTensorId++}nextVariableId(){return Hp.nextVariableId++}clone(e){let t=O.runKernel(Di,{x:e}),n={x:e},a=s=>({x:()=>{let i="float32",o={x:s},l={dtype:i};return O.runKernel(bi,o,l)}}),r=[];return this.addTapeNode(this.state.activeScope.name,n,[t],a,r,{}),t}runKernel(e,t,n){if(this.backendName==null&&this.backend,bh(e,this.backendName)==null)throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){let a=this.backend.numDataIds(),r=0;n.forEach(o=>{r+=o.dtype==="complex64"?3:1});let s=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],i=a-t-r-s;if(i>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)}runKernelFunc(e){let t,n=[],a=this.isTapeOn(),r=this.state.numBytes,s=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let i;this.backendName==null&&this.backend;let o,l=ib(e)?e.kernelName:this.state.activeScope!=null?this.state.activeScope.name:"";if(ib(e)){let{kernelName:h,inputs:m,attrs:f}=e;this.backendName==null&&this.backend;let g=bh(h,this.backendName);$(g!=null,()=>`Cannot find registered kernel '${h}' for backend '${this.backendName}'`),i=()=>{let y=this.backend.numDataIds();o=g.kernelFunc({inputs:m,attrs:f,backend:this.backend});let b=Array.isArray(o)?o:[o];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(h,y,b);let x=b.map(w=>w.rank!=null?w:this.makeTensorFromTensorInfo(w));if(a){let w=this.getTensorsForGradient(h,m,x);n=this.saveTensorsForBackwardMode(w)}return x}}else{let{forwardFunc:h}=e,m=f=>{!a||(n=f.map(g=>this.keep(this.clone(g))))};i=()=>{let f=this.backend.numDataIds();o=this.tidy(()=>h(this.backend,m));let g=Array.isArray(o)?o:[o];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,f,g),g}}let{inputs:u,attrs:p}=e,d=ib(e)?null:e.backwardsFunc,c;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?t=i():(c=this.profiler.profileKernel(l,u,()=>i()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(c),t=c.outputs)}),a&&this.addTapeNode(l,u,t,d,n,p),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-r,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-s,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(u).map(h=>u[h]!=null?u[h].shape:null),outputShapes:t.map(h=>h.shape),kernelTimeMs:c.timeMs,extraInfo:c.extraInfo}),Array.isArray(o)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(t=>this.keep(this.clone(t)))}getTensorsForGradient(e,t,n){let a=xb(e);if(a!=null){let r=a.inputsToSave||[],s=a.outputsToSave||[],i;a.saveAllInputs?($(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),i=Object.keys(t).map(l=>t[l])):i=r.map(l=>t[l]);let o=n.filter((l,u)=>s[u]);return i.concat(o)}return[]}makeTensor(e,t,n,a){if(e==null)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",a=a||this.backend;let r=e;n==="string"&&Kr(e[0])&&(r=e.map(o=>Sc(o)));let s=a.write(r,t,n),i=new Te(t,n,s,this.nextTensorId());if(this.trackTensor(i,a),n==="string"){let o=this.state.tensorInfo.get(s),l=_I(r);this.state.numBytes+=l-o.bytes,o.bytes=l}return i}makeTensorFromDataId(e,t,n,a){n=n||"float32";let r={dataId:e,shape:t,dtype:n};return this.makeTensorFromTensorInfo(r,a)}makeTensorFromTensorInfo(e,t){let{dataId:n,shape:a,dtype:r}=e,s=new Te(a,r,n,this.nextTensorId());return this.trackTensor(s,t),s}makeVariable(e,t=!0,n,a){n=n||this.nextVariableId().toString(),a!=null&&a!==e.dtype&&(e=e.cast(a));let r=new ts(e,t,n,this.nextTensorId());if(this.state.registeredVariables[r.name]!=null)throw new Error(`Variable with name ${r.name} was already registered`);return this.state.registeredVariables[r.name]=r,this.incRef(r,this.backend),r}trackTensor(e,t){this.state.numTensors++,e.dtype==="string"&&this.state.numStringTensors++;let n=0;e.dtype!=="complex64"&&e.dtype!=="string"&&(n=e.size*bb(e.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:n})),e instanceof ts||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;let t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,e.dtype==="string"&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),e.dtype!=="complex64"&&e.dtype!=="string"){let n=e.size*bb(e.dtype);this.state.numBytes-=n}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(let e in this.state.registeredVariables){let t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),this.state.registeredVariables[e.name]!=null&&delete this.state.registeredVariables[e.name]}memory(){let e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,e.reasons==null&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;let t=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await e(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(a=>a.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let a of this.state.activeProfile.kernels)a.kernelTimeMs=await a.kernelTimeMs,a.extraInfo=await a.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(e,t,n,a,r,s){let i={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:r},o=xb(e);o!=null&&(a=o.gradFunc),a!=null&&(i.gradient=l=>(l=l.map((u,p)=>{if(u==null){let d=n[p],c=Kh(d.size,d.dtype);return this.makeTensor(c,d.shape,d.dtype)}return u}),a(l.length>1?l:l[0],r,s))),this.state.activeTape.push(i)}keep(e){return e.kept=!0,e}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){let t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){let t=Nx(e),n=new Set(t.map(r=>r.id));for(let r=0;r<this.state.activeScope.track.length;r++){let s=this.state.activeScope.track[r];!s.kept&&!n.has(s.id)&&s.dispose()}let a=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(r=>{!r.kept&&r.scopeId===a.id&&this.track(r)})}gradients(e,t,n,a=!1){if($(t.length>0,()=>"gradients() received an empty list of xs."),n!=null&&n.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let r=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));$(r instanceof Te,()=>"The result y returned by f() must be a tensor.");let s=cD(this.state.activeTape,t,r);if(!a&&s.length===0&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let i={};i[r.id]=n==null?kD(r.shape):n,dD(i,s,l=>this.tidy(l),ID);let o=t.map(l=>i[l.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(l=>{for(let u of l.saved)u.dispose()}),this.state.activeTape=null),{value:r,grads:o}})}customGrad(e){return $(es(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{$(t.every(i=>i instanceof Te),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n,a={};t.forEach((i,o)=>{a[o]=i});let r=(i,o)=>(n=e(...t,o),$(n.value instanceof Te,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),$(es(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),s=(i,o)=>{let l=n.gradFunc(i,o),u=Array.isArray(l)?l:[l];$(u.length===t.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),$(u.every(d=>d instanceof Te),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let p={};return u.forEach((d,c)=>{p[c]=()=>d}),p};return this.runKernelFunc({forwardFunc:r,backwardsFunc:s,inputs:a})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){let t=Gp(),n=await this.backend.time(e);return n.wallMs=Gp()-t,n}track(e){return this.state.activeScope!=null&&(e.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(e)),e}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new G1;for(let e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};Hp.nextTensorId=0;Hp.nextVariableId=0;function kD(e){let t=vx(mt(e),"float32");return O.makeTensor(t,e,"float32")}function UI(){let e=DI();if(e._tfengine==null){let t=new FI(e);e._tfengine=new Hp(t)}return HF(e._tfengine.ENV),gD(()=>e._tfengine),e._tfengine}var O=UI();function ID(e,t){let n={a:e,b:t};return O.runKernel(cs,n)}var Tc={};Ae(Tc,{isBrowser:()=>GI,isMobile:()=>ND,mockIsMobile:()=>TD});function SD(){return typeof navigator!="undefined"&&navigator!=null}var Nb;function TD(e){Nb=e}function ND(e){if(Nb!==void 0)return Nb;if(e||SD()){if(e||(e=navigator),e.product==="ReactNative")return!0;let t=e.userAgent||e.vendor||(typeof window!="undefined"?window.opera:"");if(!t){let n=e;return n.userAgentData&&n.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function GI(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}var Jn=H();Jn.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")});Jn.registerFlag("IS_BROWSER",()=>GI());Jn.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");Jn.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));Jn.registerFlag("PROD",()=>!1);Jn.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>Jn.getBool("DEBUG"));Jn.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);Jn.registerFlag("IS_TEST",()=>!1);Jn.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0);Jn.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);Jn.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);Jn.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU",()=>!1);Jn.registerFlag("USE_SETTIMEOUTCUSTOM",()=>!1);function ir(e,t){let n=e;if(hn(e))return t==="string"?[]:[e.length];if(typeof e=="object"&&"texture"in e){let r=e.channels||"RGBA";return[e.height,e.width*r.length]}if(!Array.isArray(e))return[];let a=[];for(;Array.isArray(n)||hn(n)&&t!=="string";)a.push(n.length),n=n[0];return Array.isArray(e)&&H().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&HI(e,a,[]),a}function HI(e,t,n){if(n=n||[],!Array.isArray(e)&&!hn(e)){$(t.length===0,()=>`Element arr[${n.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}$(t.length>0,()=>`Element arr[${n.join("][")}] should be a primitive, but is an array of ${e.length} elements`),$(e.length===t[0],()=>`Element arr[${n.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);let a=t.slice(1);for(let r=0;r<e.length;++r)HI(e[r],a,n.concat(r))}function H1(e,t,n,a){if(e!=="string_or_numeric"){if(e==null)throw new Error("Expected dtype cannot be null.");if(e!=="numeric"&&e!==t||e==="numeric"&&t==="string")throw new Error(`Argument '${n}' passed to '${a}' must be ${e} tensor, but got ${t} tensor`)}}function _(e,t,n,a="numeric"){if(e instanceof Te)return H1(a,e.dtype,t,n),e;let r=qh(e);if(r!=="string"&&["bool","int32","float32"].indexOf(a)>=0&&(r=a),H1(a,r,t,n),e==null||!hn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string"){let o=e==null?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${o}'`)}let s=ir(e,r);!hn(e)&&!Array.isArray(e)&&(e=[e]);let i=r!=="string"?Cm(e,r):Zs(e,[],!0);return O.makeTensor(i,s,r)}function jp(e,t,n,a="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((r,s)=>_(r,`${t}[${s}]`,n,a))}var Cx="__op";function L(e){let t=Object.keys(e);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let n=t[0],a=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1)),n=n+Cx;let r=(...s)=>{O.startScope(n);try{let i=a(...s);return kx(i)&&console.error("Cannot return a Promise inside of tidy."),O.endScope(i),i}catch(i){throw O.endScope(null),i}};return Object.defineProperty(r,"name",{value:n,configurable:!0}),r}function CD(e,t){let n=_(e,"real","complex"),a=_(t,"imag","complex");Sn(n.shape,a.shape,`real and imag shapes, ${n.shape} and ${a.shape}, must match in call to tf.complex().`);let r={real:n,imag:a};return O.runKernel(Qh,r)}var Cr=L({complex_:CD});function fs(e,t,n,a){if(a==null&&(a=qh(e)),a==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(typeof e=="object"&&"texture"in e){if(a!=="float32"&&a!=="int32")throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${a}.`);return e.channels=e.channels||"RGBA",O.backend.createTensorFromTexture(e,t||n,a)}if(!hn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){wx(t);let r=mt(t),s=mt(n);$(r===s,()=>`Based on the provided shape, [${t}], the tensor should have ${r} values but has ${s}`);for(let i=0;i<n.length;++i){let o=n[i],l=i===n.length-1?o!==mt(t.slice(i)):!0;$(n[i]===t[i]||!l,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return!hn(e)&&!Array.isArray(e)&&(e=[e]),t=t||n,e=a!=="string"?Cm(e,a):Zs(e,[],!0),O.makeTensor(e,t,a)}function kn(e,t,n){let a=ir(e,n);return fs(e,t,a,n)}var Cb={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8},wh=4;async function _D(e,t){let n=[],a=[],r=Array.isArray(e)?e.map(i=>i.name):Object.keys(e);for(let i=0;i<r.length;++i){let o=r[i],l=Array.isArray(e)?e[i].tensor:e[o];if(l.dtype!=="float32"&&l.dtype!=="int32"&&l.dtype!=="bool"&&l.dtype!=="string"&&l.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${o}': ${l.dtype}`);let u={name:o,shape:l.shape,dtype:l.dtype};if(l.dtype==="string"){let p=new Promise(async d=>{let c=await l.bytes(),h=c.reduce((g,y)=>g+y.length,0)+wh*c.length,m=new Uint8Array(h),f=0;for(let g=0;g<c.length;g++){let y=c[g],b=new Uint8Array(new Uint32Array([y.length]).buffer);m.set(b,f),f+=wh,m.set(y,f),f+=y.length}d(m)});a.push(p)}else a.push(l.data());t!=null&&(u.group=t),n.push(u)}let s=await Promise.all(a);return{data:ED(s),specs:n}}function jI(e,t){let n={},a,r=0;for(let s of t){let i=s.name,o=s.dtype,l=s.shape,u=mt(l),p;if("quantization"in s){let d=s.quantization;if(d.dtype==="uint8"||d.dtype==="uint16"){if(!("min"in d&&"scale"in d))throw new Error(`Weight ${s.name} with quantization ${d.dtype} doesn't have corresponding metadata min and scale.`)}else if(d.dtype==="float16"){if(o!=="float32")throw new Error(`Weight ${s.name} is quantized with ${d.dtype} which only supports weights of type float32 not ${o}.`)}else throw new Error(`Weight ${s.name} has unknown quantization dtype ${d.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let c=Cb[d.dtype],h=e.slice(r,r+u*c),m=d.dtype==="uint8"?new Uint8Array(h):new Uint16Array(h);if(o==="float32")if(d.dtype==="uint8"||d.dtype==="uint16"){p=new Float32Array(m.length);for(let f=0;f<m.length;f++){let g=m[f];p[f]=g*d.scale+d.min}}else if(d.dtype==="float16")a===void 0&&(a=MD()),p=a(m);else throw new Error(`Unsupported quantization type ${d.dtype} for weight type float32.`);else if(o==="int32"){if(d.dtype!=="uint8"&&d.dtype!=="uint16")throw new Error(`Unsupported quantization type ${d.dtype} for weight type int32.`);p=new Int32Array(m.length);for(let f=0;f<m.length;f++){let g=m[f];p[f]=Math.round(g*d.scale+d.min)}}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);r+=u*c}else if(o==="string"){let d=mt(s.shape);p=[];for(let c=0;c<d;c++){let h=new Uint32Array(e.slice(r,r+wh))[0];r+=wh;let m=new Uint8Array(e.slice(r,r+h));p.push(m),r+=h}}else{let d=Cb[o],c=e.slice(r,r+u*d);if(o==="float32")p=new Float32Array(c);else if(o==="int32")p=new Int32Array(c);else if(o==="bool")p=new Uint8Array(c);else if(o==="complex64"){p=new Float32Array(c);let h=new Float32Array(p.length/2),m=new Float32Array(p.length/2);for(let y=0;y<h.length;y++)h[y]=p[y*2],m[y]=p[y*2+1];let f=kn(h,l,"float32"),g=kn(m,l,"float32");n[i]=Cr(f,g),f.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);r+=u*d}o!=="complex64"&&(n[i]=kn(p,l,o))}return n}function ED(e){if(e===null)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0,n=[];e.forEach(s=>{if(t+=s.byteLength,n.push(s.byteLength===s.buffer.byteLength?s:new s.constructor(s)),!(s instanceof Float32Array||s instanceof Int32Array||s instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${s.constructor.name}`)});let a=new Uint8Array(t),r=0;return n.forEach(s=>{a.set(new Uint8Array(s.buffer),r),r+=s.byteLength}),a.buffer}var _x=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function j1(e){return _x?Buffer.byteLength(e):new Blob([e]).size}function AD(e){if(_x)return Buffer.from(e).toString("base64");let t=new Uint8Array(e),n="";for(let a=0,r=t.length;a<r;a++)n+=String.fromCharCode(t[a]);return btoa(n)}function $D(e){if(_x){let a=Buffer.from(e,"base64");return a.buffer.slice(a.byteOffset,a.byteOffset+a.byteLength)}let t=atob(e),n=new Uint8Array(t.length);for(let a=0;a<t.length;++a)n.set([t.charCodeAt(a)],a);return n.buffer}function Ex(e){if(e.length===1)return e[0];let t=0;e.forEach(r=>{t+=r.byteLength});let n=new Uint8Array(t),a=0;return e.forEach(r=>{n.set(new Uint8Array(r),a),a+=r.byteLength}),n.buffer}function q1(e){let t="/";for(e=e.trim();e.endsWith(t);)e=e.slice(0,e.length-1);let n=e.split(t);return n[n.length-1]}function qI(e,t){let n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return e.signature!=null&&(n.signature=e.signature),e.userDefinedMetadata!=null&&(n.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(n.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(n.initializerSignature=e.initializerSignature),e.trainingConfig!=null&&(n.trainingConfig=e.trainingConfig),n}function KI(e,t,n){let a={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(e.trainingConfig!=null&&(a.trainingConfig=e.trainingConfig),e.weightsManifest!=null){if(!t)throw new Error("modelJSON has weightsManifest but weightSpecs is null");if(!n)throw new Error("modelJSON has weightsManifest but weightData is null");a.weightSpecs=t,a.weightData=n}return e.signature!=null&&(a.signature=e.signature),e.userDefinedMetadata!=null&&(a.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(a.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(a.initializerSignature=e.initializerSignature),a}async function Ax(e,t){let n,a;return e.weightsManifest!=null&&([n,a]=await t(e.weightsManifest)),KI(e,n,a)}function Nc(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:e.modelTopology==null?0:j1(JSON.stringify(e.modelTopology)),weightSpecsBytes:e.weightSpecs==null?0:j1(JSON.stringify(e.weightSpecs)),weightDataBytes:e.weightData==null?0:e.weightData.byteLength}}function XI(e){let t=[];for(let n of e)t.push(...n.weights);return t}function FD(){let e=n=>{let a=n<<13,r=0;for(;(a&8388608)===0;)r-=8388608,a<<=1;return a&=-8388609,r+=947912704,a|r},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let n=1024;n<2048;n++)t[n]=939524096+(n-1024<<13);return t}function DD(){let e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}function RD(){let e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}function MD(){let e=FD(),t=DD(),n=RD();return a=>{let r=new ArrayBuffer(4*a.length),s=new Uint32Array(r);for(let i=0;i<a.length;i++){let o=a[i],l=e[n[o>>10]+(o&1023)]+t[o>>10];s[i]=l}return new Float32Array(r)}}var At=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return At.instance==null&&(At.instance=new At),At.instance}static registerSaveRouter(e){At.getInstance().saveRouters.push(e)}static registerLoadRouter(e){At.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return At.getHandlers(e,"save")}static getLoadHandlers(e,t){return At.getHandlers(e,"load",t)}static getHandlers(e,t,n){let a=[];return(t==="load"?At.getInstance().loadRouters:At.getInstance().saveRouters).forEach(r=>{let s=r(e,n);s!==null&&a.push(s)}),a}},PD=e=>At.registerSaveRouter(e),OD=e=>At.registerLoadRouter(e),LD=e=>At.getSaveHandlers(e),zD=(e,t)=>At.getLoadHandlers(e,t),_b="tensorflowjs",Eb=1,Gs="models_store",Xr="model_info_store";function YI(){if(!H().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let e=typeof window=="undefined"?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(t==null)throw new Error("The current browser does not appear to support IndexedDB.");return t}function Ab(e){let t=e.result;t.createObjectStore(Gs,{keyPath:"modelPath"}),t.createObjectStore(Xr,{keyPath:"modelPath"})}var ti=class{constructor(e){if(this.indexedDB=YI(),e==null||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((n,a)=>{let r=this.indexedDB.open(_b,Eb);r.onupgradeneeded=()=>Ab(r),r.onsuccess=()=>{let s=r.result;if(t==null){let i=s.transaction(Gs,"readonly"),o=i.objectStore(Gs).get(this.modelPath);o.onsuccess=()=>{if(o.result==null)return s.close(),a(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));n(o.result.modelArtifacts)},o.onerror=l=>(s.close(),a(o.error)),i.oncomplete=()=>s.close()}else{let i=Nc(t),o=s.transaction(Xr,"readwrite"),l=o.objectStore(Xr),u=l.put({modelPath:this.modelPath,modelArtifactsInfo:i}),p;u.onsuccess=()=>{p=s.transaction(Gs,"readwrite");let d=p.objectStore(Gs).put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:i});d.onsuccess=()=>n({modelArtifactsInfo:i}),d.onerror=c=>{l=o.objectStore(Xr);let h=l.delete(this.modelPath);h.onsuccess=()=>(s.close(),a(d.error)),h.onerror=m=>(s.close(),a(d.error))}},u.onerror=d=>(s.close(),a(u.error)),o.oncomplete=()=>{p==null?s.close():p.oncomplete=()=>s.close()}}},r.onerror=s=>a(r.error)})}};ti.URL_SCHEME="indexeddb://";var ZI=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ti.URL_SCHEME)?WD(e.slice(ti.URL_SCHEME.length)):null;At.registerSaveRouter(ZI);At.registerLoadRouter(ZI);function WD(e){return new ti(e)}function BD(e){return e.startsWith(ti.URL_SCHEME)?e.slice(ti.URL_SCHEME.length):e}var VD=class{constructor(){this.indexedDB=YI()}async listModels(){return new Promise((e,t)=>{let n=this.indexedDB.open(_b,Eb);n.onupgradeneeded=()=>Ab(n),n.onsuccess=()=>{let a=n.result,r=a.transaction(Xr,"readonly"),s=r.objectStore(Xr).getAll();s.onsuccess=()=>{let i={};for(let o of s.result)i[o.modelPath]=o.modelArtifactsInfo;e(i)},s.onerror=i=>(a.close(),t(s.error)),r.oncomplete=()=>a.close()},n.onerror=a=>t(n.error)})}async removeModel(e){return e=BD(e),new Promise((t,n)=>{let a=this.indexedDB.open(_b,Eb);a.onupgradeneeded=()=>Ab(a),a.onsuccess=()=>{let r=a.result,s=r.transaction(Xr,"readwrite"),i=s.objectStore(Xr),o=i.get(e),l;o.onsuccess=()=>{if(o.result==null)return r.close(),n(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{let u=i.delete(e),p=()=>{l=r.transaction(Gs,"readwrite");let d=l.objectStore(Gs).delete(e);d.onsuccess=()=>t(o.result.modelArtifactsInfo),d.onerror=c=>n(o.error)};u.onsuccess=p,u.onerror=d=>(p(),r.close(),n(o.error))}},o.onerror=u=>(r.close(),n(o.error)),s.oncomplete=()=>{l==null?r.close():l.oncomplete=()=>r.close()}},a.onerror=r=>n(a.error)})}},Ir="/",Xo="tensorflowjs_models",JI="info",UD="model_topology",GD="weight_specs",HD="weight_data",jD="model_metadata";function QI(e){return{info:[Xo,e,JI].join(Ir),topology:[Xo,e,UD].join(Ir),weightSpecs:[Xo,e,GD].join(Ir),weightData:[Xo,e,HD].join(Ir),modelMetadata:[Xo,e,jD].join(Ir)}}function eS(e){for(let t of Object.values(e))window.localStorage.removeItem(t)}function qD(e){let t=e.split(Ir);if(t.length<3)throw new Error(`Invalid key format: ${e}`);return t.slice(1,t.length-1).join(Ir)}function KD(e){return e.startsWith(ni.URL_SCHEME)?e.slice(ni.URL_SCHEME.length):e}var ni=class{constructor(e){if(!H().getBool("IS_BROWSER")||typeof window=="undefined"||typeof window.localStorage=="undefined")throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,e==null||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=QI(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let t=JSON.stringify(e.modelTopology),n=JSON.stringify(e.weightSpecs),a=Nc(e);try{this.LS.setItem(this.keys.info,JSON.stringify(a)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,AD(e.weightData));let r={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:e.signature!=null?e.signature:void 0,userDefinedMetadata:e.userDefinedMetadata!=null?e.userDefinedMetadata:void 0,modelInitializer:e.modelInitializer!=null?e.modelInitializer:void 0,initializerSignature:e.initializerSignature!=null?e.initializerSignature:void 0,trainingConfig:e.trainingConfig!=null?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(r)),{modelArtifactsInfo:a}}catch(r){throw eS(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${a.modelTopologyBytes}, weightSpecsBytes=${a.weightSpecsBytes}, weightDataBytes=${a.weightDataBytes}.`)}}}async load(){let e=JSON.parse(this.LS.getItem(this.keys.info));if(e==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(e.modelTopologyType!=="JSON")throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let t={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(n==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=n;let a=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(a==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=a;let r=this.LS.getItem(this.keys.modelMetadata);if(r!=null){let i=JSON.parse(r);t.format=i.format,t.generatedBy=i.generatedBy,t.convertedBy=i.convertedBy,i.signature!=null&&(t.signature=i.signature),i.userDefinedMetadata!=null&&(t.userDefinedMetadata=i.userDefinedMetadata),i.modelInitializer!=null&&(t.modelInitializer=i.modelInitializer),i.initializerSignature!=null&&(t.initializerSignature=i.initializerSignature),i.trainingConfig!=null&&(t.trainingConfig=i.trainingConfig)}let s=this.LS.getItem(this.keys.weightData);if(s==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=$D(s),t}};ni.URL_SCHEME="localstorage://";var tS=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ni.URL_SCHEME)?XD(e.slice(ni.URL_SCHEME.length)):null;At.registerSaveRouter(tS);At.registerLoadRouter(tS);function XD(e){return new ni(e)}var YD=class{constructor(){$(H().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),$(typeof window=="undefined"||typeof window.localStorage!="undefined",()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let e={},t=Xo+Ir,n=Ir+JI;for(let a=0;a<this.LS.length;++a){let r=this.LS.key(a);if(r.startsWith(t)&&r.endsWith(n)){let s=qD(r);e[s]=JSON.parse(this.LS.getItem(r))}}return e}async removeModel(e){e=KD(e);let t=QI(e);if(this.LS.getItem(t.info)==null)throw new Error(`Cannot find model at path '${e}'`);let n=JSON.parse(this.LS.getItem(t.info));return eS(t),n}},Jo="://",Pn=class{constructor(){this.managers={}}static getInstance(){return Pn.instance==null&&(Pn.instance=new Pn),Pn.instance}static registerManager(e,t){$(e!=null,()=>"scheme must not be undefined or null."),e.endsWith(Jo)&&(e=e.slice(0,e.indexOf(Jo))),$(e.length>0,()=>"scheme must not be an empty string.");let n=Pn.getInstance();$(n.managers[e]==null,()=>`A model store manager is already registered for scheme '${e}'.`),n.managers[e]=t}static getManager(e){let t=Pn.getInstance().managers[e];if(t==null)throw new Error(`Cannot find model manager for scheme '${e}'`);return t}static getSchemes(){return Object.keys(Pn.getInstance().managers)}};function rh(e){if(e.indexOf(Jo)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Pn.getSchemes().join(",")}`);return{scheme:e.split(Jo)[0],path:e.split(Jo)[1]}}async function nS(e,t,n=!1){$(e!==t,()=>`Old path and new path are the same: '${e}'`);let a=At.getLoadHandlers(e);$(a.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),$(a.length<2,()=>`Copying failed because more than one (${a.length}) load handlers for source URL ${e}.`);let r=a[0],s=At.getSaveHandlers(t);$(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),$(s.length<2,()=>`Copying failed because more than one (${a.length}) save handlers for destination URL ${t}.`);let i=s[0],o=rh(e).scheme,l=rh(e).path,u=o===rh(e).scheme,p=await r.load();n&&u&&await Pn.getManager(o).removeModel(l);let d=await i.save(p);return n&&!u&&await Pn.getManager(o).removeModel(l),d.modelArtifactsInfo}async function ZD(){let e=Pn.getSchemes(),t={};for(let n of e){let a=await Pn.getManager(n).listModels();for(let r in a){let s=n+Jo+r;t[s]=a[r]}}return t}async function JD(e){let t=rh(e);return Pn.getManager(t.scheme).removeModel(t.path)}async function QD(e,t){return nS(e,t,!1)}async function eR(e,t){return nS(e,t,!0)}var tR=class{constructor(){this.messageName="setTimeoutCustom",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Browser's encoder only supports utf-8, but got ${t}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}setTimeoutCustom(e,t){if(typeof window=="undefined"||!H().getBool("USE_SETTIMEOUTCUSTOM")){setTimeout(e,t);return}this.functionRefs.push(e),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},"*")},t),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener("message",n=>{if(n.source===window&&n.data.name===this.messageName){n.stopPropagation();let a=this.functionRefs[n.data.index];a(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0)}},!0))}};if(H().get("IS_BROWSER")){H().setPlatform("browser",new tR);try{Pn.registerManager(ni.URL_SCHEME,new YD)}catch(e){}try{Pn.registerManager(ti.URL_SCHEME,new VD)}catch(e){}}var nR={importFetch:()=>cF()},ob,aR=class{constructor(){this.util=dF(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return H().global.fetch!=null?H().global.fetch(e,t):(ob==null&&(ob=nR.importFetch()),ob(e,t))}now(){let e=process.hrtime();return e[0]*1e3+e[1]/1e6}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Node built-in encoder only supports utf-8, but got ${t}`);return this.textEncoder.encode(e)}decode(e,t){return e.length===0?"":new this.util.TextDecoder(t).decode(e)}};H().get("IS_NODE")&&!H().get("IS_BROWSER")&&H().setPlatform("node",new aR);function Pe(e,t="float32",n){return t=t||"float32",wx(e),new Ht(e,t,n)}function rR(e,t){let n=_(e,"x","cast");if(!CI(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&n.dtype!=="string"||t!=="string"&&n.dtype==="string")throw new Error("Only strings can be casted to strings");let a={x:n},r={dtype:t};return O.runKernel(bi,a,r)}var oe=L({cast_:rR});function sR(e){let t={x:_(e,"x","clone","string_or_numeric")};return O.runKernel(Di,t)}var rr=L({clone_:sR});function $x(e,t=!1){console.log(e.toString(t))}UI();var iR={buffer:Pe,cast:oe,clone:rr,print:$x};yD(iR);var Ut={};Ae(Ut,{browserFiles:()=>hR,browserHTTPRequest:()=>bR,concatenateArrayBuffers:()=>Ex,copyModel:()=>QD,decodeWeights:()=>jI,encodeWeights:()=>_D,fromMemory:()=>vR,fromMemorySync:()=>oS,getLoadHandlers:()=>zD,getModelArtifactsForJSON:()=>Ax,getModelArtifactsForJSONSync:()=>KI,getModelArtifactsInfoForJSON:()=>Nc,getSaveHandlers:()=>LD,getWeightSpecs:()=>XI,http:()=>Dx,isHTTPScheme:()=>$b,listModels:()=>ZD,loadWeights:()=>mR,moveModel:()=>eR,registerLoadRouter:()=>OD,registerSaveRouter:()=>PD,removeModel:()=>JD,weightsLoaderFactory:()=>rS,withSaveHandler:()=>wR,withSaveHandlerSync:()=>kR});var oR="model",lR=".json",uR=".weights.bin";function K1(e){return new Promise(t=>setTimeout(t)).then(e)}var al=class{constructor(e){if(!H().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(al.URL_SCHEME)&&(e=e.slice(al.URL_SCHEME.length)),(e==null||e.length===0)&&(e=oR),this.modelJsonFileName=e+lR,this.weightDataFileName=e+uR}async save(e){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");let t=window.URL.createObjectURL(new Blob([e.weightData],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let n=[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}],a=qI(e,n),r=window.URL.createObjectURL(new Blob([JSON.stringify(a)],{type:"application/json"})),s=this.modelJsonAnchor==null?document.createElement("a"):this.modelJsonAnchor;if(s.download=this.modelJsonFileName,s.href=r,await K1(()=>s.dispatchEvent(new MouseEvent("click"))),e.weightData!=null){let i=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;i.download=this.weightDataFileName,i.href=t,await K1(()=>i.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Nc(e)}}}};al.URL_SCHEME="downloads://";var pR=class{constructor(e){if(e==null||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{let n=new FileReader;n.onload=a=>{let r=JSON.parse(a.target.result),s=r.modelTopology;if(s==null){t(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(r.weightsManifest==null){t(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){e({modelTopology:s});return}let i=Ax(r,o=>this.loadWeights(o));e(i)},n.onerror=a=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(e){let t=[],n=[];for(let s of e)t.push(...s.weights),n.push(...s.paths);let a=this.checkManifestAndWeightFiles(e),r=n.map(s=>this.loadWeightsFile(s,a[s]));return Promise.all(r).then(s=>[t,Ex(s)])}loadWeightsFile(e,t){return new Promise((n,a)=>{let r=new FileReader;r.onload=s=>{let i=s.target.result;n(i)},r.onerror=s=>a(`Failed to weights data from file of path '${e}'.`),r.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){let t=[],n=this.weightsFiles.map(r=>q1(r.name)),a={};for(let r of e)r.paths.forEach(s=>{let i=q1(s);if(t.indexOf(i)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${i}'`);if(t.push(i),n.indexOf(i)===-1)throw new Error(`Weight file with basename '${i}' is not provided.`);a[s]=this.weightsFiles[n.indexOf(i)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return a}},cR=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(al.URL_SCHEME)?dR(e.slice(al.URL_SCHEME.length)):null;At.registerSaveRouter(cR);function dR(e="model"){return new al(e)}function hR(e){return new pR(e)}function X1(e,t,n,a){i(e),n=n==null?0:n,a=a==null?1:a,o(n,a);let r=0,s=l=>(l.then(u=>{let p=n+ ++r/e.length*(a-n);return t(p),u}),l);function i(l){$(l!=null&&Array.isArray(l)&&l.length>0,()=>"promises must be a none empty array")}function o(l,u){$(l>=0&&l<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${l}`),$(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${u}`),$(u>=l,()=>`startFraction must be no more than endFraction, but got startFraction ${l} and endFraction ${u}`)}return Promise.all(e.map(s))}async function aS(e,t){t==null&&(t={});let n=t.fetchFunc==null?H().platform.fetch:t.fetchFunc,a=e.map(u=>n(u,t.requestInit,{isBinary:!0})),r=0,s=.5,i=(t.onProgress==null?await Promise.all(a):await X1(a,t.onProgress,r,s)).map(u=>u.arrayBuffer()),o=.5,l=1;return t.onProgress==null?await Promise.all(i):await X1(i,t.onProgress,o,l)}async function mR(e,t="",n,a){return rS(r=>aS(r,{requestInit:a}))(e,t,n)}function rS(e){return async(t,n="",a)=>{let r=t.map(()=>!1),s={},i=a!=null?a.map(()=>!1):[],o=[];if(t.forEach((h,m)=>{let f=0;h.weights.forEach(g=>{let y="quantization"in g?g.quantization.dtype:g.dtype,b=Cb[y]*mt(g.shape),x=()=>{r[m]=!0,s[m]==null&&(s[m]=[]),s[m].push({manifestEntry:g,groupOffset:f,sizeBytes:b})};a!=null?a.forEach((w,I)=>{w===g.name&&(x(),i[I]=!0)}):x(),o.push(g.name),f+=b})}),!i.every(h=>h)){let h=a.filter((m,f)=>!i[f]);throw new Error(`Could not find weights in manifest with names: ${h.join(", ")}. 
-Manifest JSON has weights with names: ${o.join(", ")}.`)}let l=r.reduce((h,m,f)=>(m&&h.push(f),h),[]),u=[];l.forEach(h=>{t[h].paths.forEach(m=>{let f=n+(n.endsWith("/")?"":"/")+m;u.push(f)})});let p=await e(u),d={},c=0;return l.forEach(h=>{let m=t[h].paths.length,f=0;for(let x=0;x<m;x++)f+=p[c+x].byteLength;let g=new ArrayBuffer(f),y=new Uint8Array(g),b=0;for(let x=0;x<m;x++){let w=new Uint8Array(p[c+x]);y.set(w,b),b+=w.byteLength}s[h].forEach(x=>{let w=g.slice(x.groupOffset,x.groupOffset+x.sizeBytes),I=jI(w,[x.manifestEntry]);for(let T in I)d[T]=I[T]}),c+=m}),d}}var fR="application/octet-stream",gR="application/json",Fx=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",t==null&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,this.weightUrlConverter=t.weightUrlConverter,t.fetchFunc!=null?($(typeof t.fetchFunc=="function",()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=t.fetchFunc):this.fetch=H().platform.fetch,$(e!=null&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&$(e.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,t.requestInit!=null&&t.requestInit.body!=null)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{}}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;let n=[{paths:["./model.weights.bin"],weights:e.weightSpecs}],a=qI(e,n);t.body.append("model.json",new Blob([JSON.stringify(a)],{type:gR}),"model.json"),e.weightData!=null&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:fR}),"model.weights.bin");let r=await this.fetch(this.path,t);if(r.ok)return{modelArtifactsInfo:Nc(e),responses:[r]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${r.status}.`)}async load(){let e=await this.fetch(this.path,this.requestInit);if(!e.ok)throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);let t;try{t=await e.json()}catch(r){let s=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?s+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":s+=" Please make sure the server is serving valid JSON for this request.",new Error(s)}let n=t.modelTopology,a=t.weightsManifest;if(n==null&&a==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return Ax(t,r=>this.loadWeights(r))}async loadWeights(e){let t=Array.isArray(this.path)?this.path[1]:this.path,[n,a]=yR(t),r=this.weightPathPrefix||n,s=XI(e),i=[],o=[];for(let u of e)for(let p of u.paths)this.weightUrlConverter!=null?o.push(this.weightUrlConverter(p)):i.push(r+p+a);this.weightUrlConverter&&i.push(...await Promise.all(o));let l=await aS(i,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[s,Ex(l)]}};Fx.URL_SCHEME_REGEX=/^https?:\/\//;function yR(e){let t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),a=e.substring(0,t),r=n>t?e.substring(n):"";return[a+"/",r]}function $b(e){return e.match(Fx.URL_SCHEME_REGEX)!=null}var sS=(e,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let n=!0;if(Array.isArray(e)?n=e.every(a=>$b(a)):n=$b(e),n)return Dx(e,t)}return null};At.registerSaveRouter(sS);At.registerLoadRouter(sS);function Dx(e,t){return new Fx(e,t)}function bR(e,t){return Dx(e,t)}var lb=class{constructor(e){this.modelArtifacts=e}load(){return this.modelArtifacts}},iS=class{constructor(e){this.saveHandler=e}save(e){return this.saveHandler(e)}},xR=class{constructor(e){e.load&&(this.load=()=>Promise.resolve(e.load())),e.save&&(this.save=t=>Promise.resolve(e.save(t)))}};function vR(e,t,n,a){let r=arguments;return new xR(oS(...r))}function oS(e,t,n,a){return arguments.length===1?e.modelTopology!=null||e.weightSpecs!=null?new lb(e):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new lb({modelTopology:e})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new lb({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:a}))}function wR(e){return new iS(e)}function kR(e){return new iS(e)}var lS={};Ae(lS,{confusionMatrix:()=>UR});function IR(e,t,n=!1,a=!1){let r=_(e,"a","matMul"),s=_(t,"b","matMul");[r,s]=_t(r,s);let i={a:r,b:s},o={transposeA:n,transposeB:a};return O.runKernel(yi,i,o)}var Fe=L({matMul_:IR});function SR(e,t,n=1,a=0,r="int32"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let s={indices:_(e,"indices","oneHot","int32")},i={dtype:r,depth:t,onValue:n,offValue:a};return O.runKernel(Gi,s,i)}var rl=L({oneHot_:SR});function TR(){H().set("PROD",!0)}function NR(){H().set("DEBUG",!0)}function CR(){H().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function Rx(e){H().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}bD(Rx);function _R(){O.disposeVariables()}function Na(){return O}function kh(){return O.memory()}function ER(e){return O.profile(e)}function P(e,t){return O.tidy(e,t)}function _e(e){Nx(e).forEach(t=>t.dispose())}function Jt(e){return O.keep(e)}function AR(e){return O.time(e)}function $R(e){return O.setBackend(e)}function FR(){return O.ready()}function DR(){return O.backendName}function RR(e){O.removeBackend(e)}function MR(e){return O.findBackend(e)}function PR(e){return O.findBackendFactory(e)}function Em(e,t,n=1){return O.registerBackend(e,t,n)}function uS(){return O.backend}function OR(e,t){H().setPlatform(e,t)}function LR(e){let t={input:_(e,"input","imag")};return O.runKernel(cm,t)}var Cc=L({imag_:LR});function zR(e){let t={x:_(e,"x","neg")};return O.runKernel(eu,t)}var bt=L({neg_:zR});function WR(e){let t={input:_(e,"input","real")};return O.runKernel(wm,t)}var sl=L({real_:WR});function BR(e,t,n){let a=_(e,"x","transpose");if(t==null&&(t=a.shape.map((i,o)=>o).reverse()),$(a.rank===t.length,()=>`Error in transpose: rank of input ${a.rank} must match length of perm ${t}.`),t.forEach(i=>{$(i>=0&&i<a.rank,()=>`All entries in 'perm' must be between 0 and ${a.rank-1} but got ${t}`)}),a.rank<=1)return a.clone();let r={x:a},s={perm:t};return a.dtype==="complex64"?P(()=>{let i=sl(a),o=Cc(a);return i=O.runKernel(Tr,{x:i},s),o=O.runKernel(Tr,{x:o},s),n&&(o=bt(o)),Cr(i,o)}):O.runKernel(Tr,r,s)}var Ee=L({transpose_:BR});function VR(e,t,n){let a=_(e,"labels","confusionMatrix"),r=_(t,"predictions","confusionMatrix");$(n==null||n>0&&Number.isInteger(n),()=>`If provided, numClasses must be a positive integer, but got ${n}`),$(a.rank===1,()=>`Expected the rank of labels to be 1, but got ${a.rank}`),$(r.rank===1,()=>`Expected the rank of predictions to be 1, but got ${r.rank}`),$(a.shape[0]===r.shape[0],()=>`Mismatch in the number of examples: ${a.shape[0]} vs. ${r.shape[0]}. Labels and predictions should have the same number of elements.`),$(n>0&&Number.isInteger(n),()=>`numClasses is required to be a positive integer, but got ${n}`);let s=rl(oe(a,"int32"),n),i=rl(oe(r,"int32"),n),o=Ee(s),l=Fe(o,i);return oe(l,"int32")}var UR=L({confusionMatrix_:VR}),Tu={};Ae(Tu,{assertAndGetBroadcastShape:()=>lt,getBroadcastDims:()=>pS,getReductionAxes:()=>Wt});function pS(e,t){let n=e.length,a=[];for(let r=0;r<n;r++){let s=n-1-r,i=e[s]||1;(t[t.length-1-r]||1)>1&&i===1&&a.unshift(s)}return a}function Wt(e,t){let n=[];for(let a=0;a<t.length;a++){let r=e[e.length-a-1],s=t.length-a-1,i=t[s];(r==null||r===1&&i>1)&&n.unshift(s)}return n}function lt(e,t){let n=[],a=Math.max(e.length,t.length);for(let r=0;r<a;r++){let s=e[e.length-r-1];s==null&&(s=1);let i=t[t.length-r-1];if(i==null&&(i=1),s===1)n.unshift(i);else if(i===1)n.unshift(s);else if(s!==i){let o=`Operands could not be broadcast together with shapes ${e} and ${t}.`;throw Error(o)}else n.unshift(s)}return n}var co={};Ae(co,{fromPixels:()=>YR,fromPixelsAsync:()=>KR,toPixels:()=>XR});function _c(e,t,n){if(hi(e),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");let a=ir(e,n);if(a.length!==3&&a.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return fs(e,t,a,n)}var Ps;function cS(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(e==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let n=!1,a=!1,r=!1,s=!1,i=!1,o=!1;if(e.data instanceof Uint8Array)n=!0;else if(typeof ImageData!="undefined"&&e instanceof ImageData)a=!0;else if(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)r=!0;else if(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)s=!0;else if(e.getContext!=null)i=!0;else if(typeof ImageBitmap!="undefined"&&e instanceof ImageBitmap)o=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${e.constructor.name}`);if(bh(yh,O.backendName)!=null){let c={pixels:e},h={numChannels:t};return O.runKernel(yh,c,h)}let[l,u]=r?[e.videoWidth,e.videoHeight]:[e.width,e.height],p;if(i)p=e.getContext("2d").getImageData(0,0,l,u).data;else if(a||n)p=e.data;else if(s||r||o){if(Ps==null)if(typeof document=="undefined")if(typeof OffscreenCanvas!="undefined"&&typeof OffscreenCanvasRenderingContext2D!="undefined")Ps=new OffscreenCanvas(1,1).getContext("2d");else throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");else Ps=document.createElement("canvas").getContext("2d",{willReadFrequently:!0});Ps.canvas.width=l,Ps.canvas.height=u,Ps.drawImage(e,0,0,l,u),p=Ps.getImageData(0,0,l,u).data}let d;if(t===4)d=new Int32Array(p);else{let c=l*u;d=new Int32Array(c*t);for(let h=0;h<c;h++)for(let m=0;m<t;++m)d[h*t+m]=p[h*4+m]}return _c(d,[u,l,t],"int32")}function GR(e){return e!=null&&e.data instanceof Uint8Array}function HR(){return typeof window!="undefined"&&typeof ImageBitmap!="undefined"&&window.hasOwnProperty("createImageBitmap")}function jR(e){return e!=null&&e.width!==0&&e.height!==0}function qR(e){return HR()&&!(e instanceof ImageBitmap)&&jR(e)&&!GR(e)}async function KR(e,t=3){let n=null;if(H().getBool("WRAP_TO_IMAGEBITMAP")&&qR(e)){let a;try{a=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(r){a=null}a!=null&&a.width===e.width&&a.height===e.height?n=a:n=e}else n=e;return cS(n,t)}async function XR(e,t){let n=_(e,"img","toPixels");if(!(e instanceof Te)){let u=n;n=oe(u,"int32"),u.dispose()}if(n.rank!==2&&n.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);let[a,r]=n.shape.slice(0,2),s=n.rank===2?1:n.shape[2];if(s>4||s===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${s}`);if(n.dtype!=="float32"&&n.dtype!=="int32")throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);let i=await n.data(),o=n.dtype==="float32"?255:1,l=new Uint8ClampedArray(r*a*4);for(let u=0;u<a*r;++u){let p=[0,0,0,255];for(let c=0;c<s;c++){let h=i[u*s+c];if(n.dtype==="float32"){if(h<0||h>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${h}.`)}else if(n.dtype==="int32"&&(h<0||h>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${h}.`);s===1?(p[0]=h*o,p[1]=h*o,p[2]=h*o):p[c]=h*o}let d=u*4;l[d+0]=Math.round(p[0]),l[d+1]=Math.round(p[1]),l[d+2]=Math.round(p[2]),l[d+3]=Math.round(p[3])}if(t!=null){t.width=r,t.height=a;let u=t.getContext("2d"),p=new ImageData(l,r,a);u.putImageData(p,0,0)}return n!==e&&n.dispose(),l}var YR=L({fromPixels_:cS}),Mx={};Ae(Mx,{prepareAndValidate:()=>dS});function dS(e,t){let n=e.shape.length,a=t.shape.length;if(n<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);if(a<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${a}.`);if(t.dtype!=="int32")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[a-1]>n)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[a-1]} vs. ${n}`);if(mt(e.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);let r=t.shape,s=r[r.length-1],i=1;for(let d=0;d<r.length-1;++d)i*=r[d];let o=e.shape,l=r.slice();l.pop();let u=1;for(let d=s;d<n;++d)u*=o[d],l.push(o[d]);let p=[...vl(e.shape).map(d=>d/u),1].slice(0,s);return[l,i,u,p]}var Px={};Ae(Px,{calculateShapes:()=>hS,validateInput:()=>Lx,validateUpdateShape:()=>Ox});function Ox(e,t,n){let a=t.rank>1?t.shape[t.rank-1]:1,r=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${a}, and batchDim: ${r}.`;if(n.rank<r)throw new Error(s+` update.rank < ${r}. `);if(e.length<a+(n.rank-r))throw new Error(s+` Output shape length < ${a+(n.rank-r)}`);if(n.rank!==r+e.length-a)throw new Error(s+` update.rank != ${r+e.length-a}`);for(let i=0;i<r;++i)if(n.shape[i]!==t.shape[i])throw new Error(s+` updates.shape[${i}] (${n.shape[i]}) != indices.shape[${i}] (${t.shape[i]}).`);for(let i=0;i<n.rank-r;++i)if(n.shape[i+r]!==e[i+a])throw new Error(s+` updates.shape[${i+r}] (${n.shape[i+r]}) != shape[${i+r}] (${e[i+r]})`)}function Lx(e,t,n){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(e.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);if(t.dtype!=="int32")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(n.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${n}`);if(n.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(e.size===0)throw new Error(`Updates specified for empty output. updates shape: ${e.shape}`)}Ox(n,t,e)}function hS(e,t,n){let a=t.shape.length,r=a>1?t.shape[a-1]:1,s=n.length,i=1;for(let d=r;d<s;++d)i*=n[d];let o=r<1?1:r,l=mt(t.shape)/o,u=[...vl(n.slice(0,r)),1],p=mt(n);return{sliceRank:r,numUpdates:l,sliceSize:i,strides:u,outputSize:p}}var jt={};Ae(jt,{assertParamsValid:()=>JR,computeFlatOffset:()=>aM,computeOutShape:()=>eM,getNormalizedAxes:()=>tM,isSliceContinous:()=>nM,maskToAxes:()=>QR,parseSliceParams:()=>kS,sliceInfo:()=>rM,startForAxis:()=>vS,startIndicesWithElidedDims:()=>yS,stopForAxis:()=>wS,stopIndicesWithElidedDims:()=>bS,stridesForAxis:()=>xS,stridesWithElidedDims:()=>mS});var Fb=-2,ZR=-1;function JR(e,t,n){let a=e.shape.length;$(a===t.length,()=>`Error in slice${a}D: Length of begin ${t} must match the rank of the array (${a}).`),$(a===n.length,()=>`Error in slice${a}D: Length of size ${n} must match the rank of the array (${a}).`);for(let r=0;r<a;++r)$(t[r]+n[r]<=e.shape[r],()=>`Error in slice${a}D: begin[${r}] + size[${r}] (${t[r]+n[r]}) would overflow input.shape[${r}] (${e.shape[r]})`)}function QR(e){let t=[],n=0;for(;e>0;)e&1&&t.push(n),e/=2,n++;return t}function eM(e,t,n){let a=[];for(let r=0;r<e.length;r++)a[r]=Math.ceil((t[r]-e[r])/n[r]);return a}function mS(e,t,n,a){let r=[...e];for(let s=r.length;s<a.length;s++)r.push(1);for(let s=0;s<n;s++)s===0?r[t]=1:(r.splice(t,0,1),r.pop());return r}function fS(e,t,n){return n<=e?n:n-(t-1)}function gS(e,t){let n=[];for(let a=0;a<e;a++)n.push(t+a);return n}function tM(e,t,n,a,r,s,i,o,l){let u=e.length,p=new Array(u),d=new Array(u),c=new Array(u);if(t.length&&n>0){let h=t[0],m=n+1;p=yS(i,h,m,a,e),d=bS(o,h,m,r,e),c=mS(s,h,m,e)}else for(let h=0;h<u;h++)p[h]=vS(i,a,s,e,h,l),d[h]=wS(o,r,s,e,h,l),c[h]=xS(s,h,l);return{begin:p,end:d,strides:c}}function yS(e,t,n,a,r){let s=[...r],i=gS(n,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=0;else{let l=fS(t,n,o),u=a[l];e&1<<l&&(u=0),s[o]=u}return s}function bS(e,t,n,a,r){let s=[...r],i=gS(n,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=Number.MAX_SAFE_INTEGER;else{let l=fS(t,n,o),u=a[l];e&1<<l&&(u=Number.MAX_SAFE_INTEGER),s[o]=u}for(let o=0;o<s.length;o++){let l=r[o];s[o]<0&&(s[o]+=l),s[o]=Vp(0,s[o],r[o])}return s}function xS(e,t,n){let a=e[t];return(n&1<<t||a==null)&&(a=1),a}function vS(e,t,n,a,r,s){let i=t[r],o=n[r]||1;(e&1<<r||s&1<<r||i==null)&&(o>0?i=Number.MIN_SAFE_INTEGER:i=Number.MAX_SAFE_INTEGER);let l=a[r];return i<0&&(i+=l),i=Vp(0,i,l-1),i}function wS(e,t,n,a,r,s){let i=t[r],o=n[r]||1;(e&1<<r||s&1<<r||i==null)&&(o>0?i=Number.MAX_SAFE_INTEGER:i=Number.MIN_SAFE_INTEGER);let l=a[r];return i<0&&(i+=l),o>0?i=Vp(0,i,l):i=Vp(-1,i,l-1),i}function nM(e,t,n){let a=n.length;for(let r=0;r<n.length;r++)if(n[r]>1){a=r;break}for(let r=a+1;r<n.length;r++)if(t[r]>0||n[r]!==e[r])return!1;return!0}function aM(e,t){let n=e.length>0?e[e.length-1]:1;for(let a=0;a<e.length-1;a++)n+=e[a]*t[a];return n}function kS(e,t,n){let a,r=e.shape.length;typeof t=="number"?a=[t,...new Array(r-1).fill(0)]:t.length<r?a=t.concat(new Array(r-t.length).fill(0)):a=t.slice(),a.forEach(i=>{$(i!==-1,()=>"slice() does not support negative begin indexing.")});let s;return n==null?s=new Array(r).fill(-1):typeof n=="number"?s=[n,...new Array(r-1).fill(-1)]:n.length<r?s=n.concat(new Array(r-n.length).fill(-1)):s=n,s=s.map((i,o)=>i>=0?i:($(i===-1,()=>`Negative size values should be exactly -1 but got ${i} for the slice() size at index ${o}.`),e.shape[o]-a[o])),[a,s]}function rM(e,t,n,a,r,s,i,o,l){let u;if(a==null?(u=new Array(t.length),u.fill(1)):u=a,i!=null&&(i&i-1)!==0)throw new Error("Multiple ellipses in slice is not allowed.");let p=!1,d={dims:u.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:n.slice(),strides:u.slice(),beginMask:r,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};for(let b=0;b<d.dims;b++)p&&(1<<b&o)!==0&&d.numAddAxisAfterEllipsis++,1<<b&i&&(p=!0);p||(d.ellipsisMask|=1<<d.dims,d.dims++);let c={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};sM(d,c);let h=!0,m=!0,f=!0,g=[],y=[];for(let b=0;b<e.length;++b){if(c.strides[b]===0)throw Error(`strides[${b}] must be non-zero`);let x=!!(c.shrinkAxisMask&1<<b),w=e[b];if(w===-1){g.push(x?1:-1);continue}let I=[c.beginMask&1<<b,c.endMask&1<<b],T=[c.strides[b]>0?0:-1,c.strides[b]>0?w:w-1];if(x&&c.strides[b]<=0)throw Error("only stride 1 allowed on non-range indexing.");f=f&&c.strides[b]===1;let C=!!(c.beginMask&1<<b&&c.endMask&1<<b);if(c.beginValid&&c.endValid){if(x){let F=c.begin[b]<0?w+c.begin[b]:c.begin[b];if(c.begin[b]=F,c.end[b]=c.begin[b]+1,F<0||F>=w)throw Error(`slice index ${c.begin[b]} of dimension ${b} out of bounds.`)}else c.begin[b]=Y1(c.begin[b],0,c.strides[b],w,I,T),c.end[b]=Y1(c.end[b],1,c.strides[b],w,I,T);let R=c.strides[b]===1&&c.begin[b]===0&&c.end[b]===w;h=h&&R,m=m&&(b===0&&c.strides[b]===1||R)}else h=h&&c.strides[b]===1&&C,m=m&&(b===0&&c.strides[b]===1||C);let E,A=!1;if(c.beginValid&&c.endValid?(E=c.end[b]-c.begin[b],A=!0):x?(E=1,A=!0):C&&w>=0&&(c.strides[b]<0?E=-w:E=w,A=!0),A){let R;E===0||E<0!=c.strides[b]<0?R=0:R=Math.trunc(E/c.strides[b])+(E%c.strides[b]!==0?1:0),g.push(R)}else g.push(-1)}for(let b=0;b<c.finalShapeGatherIndices.length;++b){let x=c.finalShapeGatherIndices[b];x>=0?y.push(g[x]):x===Fb&&y.push(1)}return{finalShapeSparse:y.filter((b,x)=>c.finalShapeGatherIndices[x]!==Fb),finalShape:y,isIdentity:h,sliceDim0:m,isSimpleSlice:f,begin:c.begin,end:c.end,strides:c.strides}}function sM(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let n=0;t.beginValid=e.begin!=null,t.endValid=e.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let a=0;a<e.dims;a++)if(1<<a&e.ellipsisMask){let r=Math.min(t.dims-(e.dims-a)+1+e.numAddAxisAfterEllipsis,t.dims);for(;n<r;n++)t.begin[n]=0,t.end[n]=0,t.strides[n]=1,t.beginMask|=1<<n,t.endMask|=1<<n,t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[n]=a}else if(1<<a&e.newAxisMask)t.finalShapeGatherIndices.push(Fb),t.finalShapeGatherIndicesSparse.push(-1);else{if(n===t.begin.length)throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);e.begin!=null&&(t.begin[n]=e.begin[a]),e.end!=null&&(t.end[n]=e.end[a]),t.strides[n]=e.strides[a],e.beginMask&1<<a&&(t.beginMask|=1<<n),e.endMask&1<<a&&(t.endMask|=1<<n),e.shrinkAxisMask&1<<a?(t.finalShapeGatherIndices.push(ZR),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<n):(t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(a)),t.inputShapeGatherIndicesSparse[n]=a,n++}}function Y1(e,t,n,a,r,s){if(r[t])return n>0?s[t]:s[t+1&1];{let i=e<0?a+e:e;return i<s[0]?s[0]:i>s[1]?s[1]:i}}var ne={};Ae(ne,{Serializable:()=>IS,SerializationMap:()=>Bs,registerClass:()=>gs});var IS=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},Bs=class{constructor(){this.classNameMap={}}static getMap(){return Bs.instance==null&&(Bs.instance=new Bs),Bs.instance}static register(e){Bs.getMap().classNameMap[e.className]=[e,e.fromConfig]}};function gs(e){$(e.className!=null,()=>"Class being registered does not have the static className property defined."),$(typeof e.className=="string",()=>"className is required to be a string, but got type "+typeof e.className),$(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),Bs.register(e)}var SS={};Ae(SS,{TEST_EPSILON_FLOAT16:()=>TS,createVideoElement:()=>hM,encodeStrings:()=>NS,expectArrayBuffersEqual:()=>dM,expectArraysClose:()=>oM,expectArraysEqual:()=>uM,expectNumbersClose:()=>pM,expectPromiseToFail:()=>lM,expectValuesInRange:()=>cM,play:()=>mM,testEpsilon:()=>zx});var iM=.001,TS=.1;function oM(e,t,n){return n==null&&(n=zx()),Db(e,t,(a,r)=>Wx(a,r,n))}function zx(){return O.backend.floatPrecision()===32?iM:TS}function Db(e,t,n){let a=!0;if((hn(e)||hn(t))&&(a=!1),hn(e)&&hn(t)&&(a=!0),a){let i=e.constructor.name,o=t.constructor.name;if(i!==o)throw new Error(`Arrays are of different type. Actual: ${i}. Expected: ${o}`)}if(Array.isArray(e)&&Array.isArray(t)){let i=ir(e),o=ir(t);if(!ps(i,o))throw new Error(`Arrays have different shapes. Actual: [${i}]. Expected: [${o}]`)}let r=hn(e)?e:Zs(e),s=hn(t)?t:Zs(t);if(r.length!==s.length)throw new Error(`Arrays have different lengths actual: ${r.length} vs expected: ${s.length}.
-Actual:   ${r}.
-Expected: ${s}.`);for(let i=0;i<s.length;++i){let o=r[i],l=s[i];if(!n(o,l))throw new Error(`Arrays differ: actual[${i}] = ${o}, expected[${i}] = ${l}.
-Actual:   ${r}.
-Expected: ${s}.`)}typeof expect!="undefined"&&expect().nothing()}function lM(e,t){e().then(()=>t.fail(),()=>t()),typeof expect!="undefined"&&expect().nothing()}function uM(e,t){let n=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return Kr(e)||Kr(e[0])||Kr(t)||Kr(t[0])?Db(e,n,(a,r)=>a==r):Db(e,t,(a,r)=>Wx(a,r,0))}function pM(e,t,n){if(n==null&&(n=zx()),!Wx(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`);typeof expect!="undefined"&&expect().nothing()}function Wx(e,t,n){return!isFinite(e)&&!isFinite(t)?!0:!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function cM(e,t,n){for(let a=0;a<e.length;a++)if(e[a]<t||e[a]>n)throw new Error(`Value out of range:${e[a]} low: ${t}, high: ${n}`)}function dM(e,t){let n=new Float32Array(e),a=new Float32Array(t);if(n.length!==a.length)throw new Error(`Expected ArrayBuffer to be of length ${a.length}, but it was ${n.length}`);for(let r=0;r<a.length;r++)if(n[r]!==a[r])throw new Error(`Expected ArrayBuffer value at ${r} to be ${a[r]} but got ${n[r]} instead`)}function NS(e){for(let t=0;t<e.length;t++){let n=e[t];Array.isArray(n)?NS(n):e[t]=Sc(n)}return e}function hM(e){let t=document.createElement("video");return"playsInline"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position="fixed",t.style.left="0px",t.style.top="0px",t.preload="auto",t.appendChild(e),new Promise(n=>{t.addEventListener("loadeddata",a=>n(t)),t.load()})}async function mM(e){await e.play(),"requestVideoFrameCallback"in e&&await new Promise(t=>{e.requestVideoFrameCallback(t)})}var fM="4.0.0";function gM(e,t){let n=_(e,"a","add"),a=_(t,"b","add");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(cs,r)}var Y=L({add_:gM});function yM(e,t){let n=_(e,"a","floorDiv"),a=_(t,"b","floorDiv");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Ai,r)}var Am=L({floorDiv_:yM});function bM(e,t){let n=_(e,"a","div"),a=_(t,"b","div");if([n,a]=_t(n,a),n.dtype==="int32"&&a.dtype==="int32")return Am(n,a);let r={a:n,b:a},s={};return O.runKernel(Ni,r,s)}var he=L({div_:bM});function xM(e,t){let n=_(e,"a","mul"),a=_(t,"b","mul");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Ui,r)}var z=L({mul_:xM});function vM(e){let t=_(e,"x","abs");if(t.dtype==="complex64"){let n={x:t};return O.runKernel(oc,n)}else{let n={x:t};return O.runKernel(wl,n)}}var Lt=L({abs_:vM});function wM(e){let t={x:_(e,"x","acos")};return O.runKernel(kl,t)}var Bx=L({acos_:wM});function kM(e){let t={x:_(e,"x","acosh")};return O.runKernel(Il,t)}var Vx=L({acosh_:kM});function IM(e){$(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),$(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);let t=e.map((r,s)=>_(r,`tensors${s}`,"addN")),n=t[0];t.forEach(r=>{if(r.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(r=>{if(!ps(r.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let a=t;return O.runKernel(mi,a)}var CS=L({addN_:IM});function SM(e,t=null,n=!1){let a={x:_(e,"x","all","bool")},r={axis:t,keepDims:n};return O.runKernel(Sl,a,r)}var $m=L({all_:SM});function TM(e,t=null,n=!1){let a={x:_(e,"x","any","bool")},r={axis:t,keepDims:n};return O.runKernel(Tl,a,r)}var qp=L({any_:TM});function NM(e,t=0){let n={x:_(e,"x","argMax")},a={axis:t};return O.runKernel(fi,n,a)}var ai=L({argMax_:NM});function CM(e,t=0){let n={x:_(e,"x","argMin")},a={axis:t};return O.runKernel(sc,n,a)}var Ux=L({argMin_:CM});function _M(e){let t={x:_(e,"x","asin")};return O.runKernel(Nl,t)}var Gx=L({asin_:_M});function EM(e){let t={x:_(e,"x","asinh")};return O.runKernel(Cl,t)}var Hx=L({asinh_:EM});function AM(e){let t={x:_(e,"x","atan")};return O.runKernel(_l,t)}var jx=L({atan_:AM});function $M(e,t){let n=_(e,"a","atan2"),a=_(t,"b","atan2");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Al,r)}var qx=L({atan2_:$M});function FM(e){let t={x:_(e,"x","atanh")};return O.runKernel(El,t)}var Kx=L({atanh_:FM});function DM(e,t,n,a,r="NHWC",s){let i=e[3],o=[...t,i],l=AS(r);return Ec(e,o,n,s,a,null,null,l)}function _S(e,t,n,a,r,s,i="channelsLast"){let[o,l]=Ih(t),u;if(i==="channelsLast")u=[o,l,e[3],e[3]];else if(i==="channelsFirst")u=[o,l,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return Ec(e,u,n,a,r,s,!1,i)}function RM(e,t,n,a,r,s,i="NDHWC"){let[o,l,u]=Rb(t),p,d;if(i==="NDHWC")d="channelsLast",p=[o,l,u,e[4],e[4]];else if(i==="NCDHW")d="channelsFirst",p=[o,l,u,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return ES(e,p,n,a,r,!1,d,s)}function Ec(e,t,n,a,r,s,i=!1,o="channelsLast"){let[l,u,p,d]=[-1,-1,-1,-1];if(o==="channelsLast")[l,u,p,d]=e;else if(o==="channelsFirst")[l,d,u,p]=e;else throw new Error(`Unknown dataFormat ${o}`);let[c,h,,m]=t,[f,g]=Ih(n),[y,b]=Ih(a),x=Qo(c,y),w=Qo(h,b),{padInfo:I,outHeight:T,outWidth:C}=OM(r,u,p,f,g,x,w,s,o),E=i?m*d:m,A;return o==="channelsFirst"?A=[l,E,T,C]:o==="channelsLast"&&(A=[l,T,C,E]),{batchSize:l,dataFormat:o,inHeight:u,inWidth:p,inChannels:d,outHeight:T,outWidth:C,outChannels:E,padInfo:I,strideHeight:f,strideWidth:g,filterHeight:c,filterWidth:h,effectiveFilterHeight:x,effectiveFilterWidth:w,dilationHeight:y,dilationWidth:b,inShape:e,outShape:A,filterShape:t}}function ES(e,t,n,a,r,s=!1,i="channelsLast",o){let[l,u,p,d,c]=[-1,-1,-1,-1,-1];if(i==="channelsLast")[l,u,p,d,c]=e;else if(i==="channelsFirst")[l,c,u,p,d]=e;else throw new Error(`Unknown dataFormat ${i}`);let[h,m,f,,g]=t,[y,b,x]=Rb(n),[w,I,T]=Rb(a),C=Qo(h,w),E=Qo(m,I),A=Qo(f,T),{padInfo:R,outDepth:F,outHeight:S,outWidth:M}=LM(r,u,p,d,y,b,x,C,E,A,o),B=s?g*c:g,U;return i==="channelsFirst"?U=[l,B,F,S,M]:i==="channelsLast"&&(U=[l,F,S,M,B]),{batchSize:l,dataFormat:i,inDepth:u,inHeight:p,inWidth:d,inChannels:c,outDepth:F,outHeight:S,outWidth:M,outChannels:B,padInfo:R,strideDepth:y,strideHeight:b,strideWidth:x,filterDepth:h,filterHeight:m,filterWidth:f,effectiveFilterDepth:C,effectiveFilterHeight:E,effectiveFilterWidth:A,dilationDepth:w,dilationHeight:I,dilationWidth:T,inShape:e,outShape:U,filterShape:t}}function MM(e,t,n,a,r){a==null&&(a=Xx(e,t,n));let s=e[0],i=e[1],o=qs((s-t+2*a)/n+1,r),l=qs((i-t+2*a)/n+1,r);return[o,l]}function PM(e,t,n,a,r,s){r==null&&(r=Xx(e,t,a));let i=e[0],o=e[1],l=e[2],u=qs((i-t+2*r)/a+1,s),p=qs((o-t+2*r)/a+1,s),d=qs((l-t+2*r)/a+1,s);return[u,p,d,n]}function Xx(e,t,n,a=1){let r=Qo(t,a);return Math.floor((e[0]*(n-1)-n+r)/2)}function Ih(e){return typeof e=="number"?[e,e,e]:e.length===2?[e[0],e[1],1]:e}function Rb(e){return typeof e=="number"?[e,e,e]:e}function Qo(e,t){return t<=1?e:e+(e-1)*(t-1)}function OM(e,t,n,a,r,s,i,o,l){let u,p,d;if(typeof e=="number"){u={top:e,bottom:e,left:e,right:e,type:e===0?"VALID":"NUMBER"};let c=MM([t,n],s,a,e,o);p=c[0],d=c[1]}else if(e==="same"){p=Math.ceil(t/a),d=Math.ceil(n/r);let c=Math.max(0,(p-1)*a+s-t),h=Math.max(0,(d-1)*r+i-n),m=Math.floor(c/2),f=c-m,g=Math.floor(h/2),y=h-g;u={top:m,bottom:f,left:g,right:y,type:"SAME"}}else if(e==="valid")u={top:0,bottom:0,left:0,right:0,type:"VALID"},p=Math.ceil((t-s+1)/a),d=Math.ceil((n-i+1)/r);else if(typeof e=="object"){let c=l==="channelsLast"?e[1][0]:e[2][0],h=l==="channelsLast"?e[1][1]:e[2][1],m=l==="channelsLast"?e[2][0]:e[3][0],f=l==="channelsLast"?e[2][1]:e[3][1];u={top:c,bottom:h,left:m,right:f,type:c===0&&h===0&&m===0&&f===0?"VALID":"EXPLICIT"},p=qs((t-s+c+h)/a+1,o),d=qs((n-i+m+f)/r+1,o)}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:u,outHeight:p,outWidth:d}}function LM(e,t,n,a,r,s,i,o,l,u,p){let d,c,h,m;if(typeof e=="number"){d={top:e,bottom:e,left:e,right:e,front:e,back:e,type:e===0?"VALID":"NUMBER"};let f=PM([t,n,a,1],o,1,r,e,p);c=f[0],h=f[1],m=f[2]}else if(e==="same"){c=Math.ceil(t/r),h=Math.ceil(n/s),m=Math.ceil(a/i);let f=(c-1)*r+o-t,g=(h-1)*s+l-n,y=(m-1)*i+u-a,b=Math.floor(f/2),x=f-b,w=Math.floor(g/2),I=g-w,T=Math.floor(y/2),C=y-T;d={top:w,bottom:I,left:T,right:C,front:b,back:x,type:"SAME"}}else if(e==="valid")d={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},c=Math.ceil((t-o+1)/r),h=Math.ceil((n-l+1)/s),m=Math.ceil((a-u+1)/i);else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:d,outDepth:c,outHeight:h,outWidth:m}}function qs(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error(`Unknown roundingMode ${t}`)}}function ns(e){let[t,n,a]=Ih(e);return t===1&&n===1&&a===1}function ur(e,t){return ns(e)||ns(t)}function AS(e){if(e==="NHWC")return"channelsLast";if(e==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function Tn(e,t,n){if(n!=null){if(typeof t=="string")throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);if(typeof t=="number")$(tl(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);else if(typeof t=="object")t.forEach(a=>{a.forEach(r=>{$(tl(r),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${r}.`)})});else throw Error(`Error in ${e}: Unknown padding parameter: ${t}`)}}function zM(e,t){let n={x:_(e,"x","reshape","string_or_numeric")},a={shape:t};return O.runKernel(lu,n,a)}var W=L({reshape_:zM});function WM(e,t,n,a,r){let s=_(e,"x","avgPool","float32"),i=1;$(ur(n,i),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${i}'`);let o=s,l=!1;s.rank===3&&(l=!0,o=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(o.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${o.rank}.`),Tn("avgPool",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r},d=O.runKernel(gi,u,p);return d=oe(d,s.dtype),l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var ga=L({avgPool_:WM});function BM(e,t,n,a,r,s="NDHWC"){let i=_(e,"x","avgPool3d","float32"),o=i,l=!1;i.rank===4&&(l=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(o.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`),$(s==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Tn("avgPool3d",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r,dataFormat:s},d=O.runKernel(ic,u,p);return d=oe(d,o.dtype),l?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var Yx=L({avgPool3d_:BM});function VM(e,t=0){$(e.length>=1,()=>"Pass at least one tensor to concat");let n=jp(e,"tensors","concat","string_or_numeric");if(n[0].dtype==="complex64"&&n.forEach(s=>{if(s.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
-          with dtype ${s.dtype}. `)}),n.length===1)return rr(n[0]);let a=n,r={axis:t};return O.runKernel(Fl,a,r)}var Ze=L({concat_:VM});function UM(e){let t={x:_(e,"x","sigmoid","float32")};return O.runKernel(ao,t)}var da=L({sigmoid_:UM});function GM(e,t,n){let a=_(e,"x","slice","string_or_numeric");if(a.rank===0)throw new Error("Slicing scalar is not possible");let r={x:a},s={begin:t,size:n};return O.runKernel(du,r,s)}var Be=L({slice_:GM});function HM(e){let t={x:_(e,"x","tanh","float32")};return O.runKernel(po,t)}var ri=L({tanh_:HM});function jM(e,t,n,a,r,s){let i=_(e,"forgetBias","basicLSTMCell"),o=_(t,"lstmKernel","basicLSTMCell"),l=_(n,"lstmBias","basicLSTMCell"),u=_(a,"data","basicLSTMCell"),p=_(r,"c","basicLSTMCell"),d=_(s,"h","basicLSTMCell"),c=Ze([u,d],1),h=Fe(c,o),m=Y(h,l),f=m.shape[0],g=m.shape[1]/4,y=[f,g],b=Be(m,[0,0],y),x=Be(m,[0,g],y),w=Be(m,[0,g*2],y),I=Be(m,[0,g*3],y),T=Y(z(da(b),ri(x)),z(p,da(Y(i,w)))),C=z(ri(T),da(I));return[T,C]}var $S=L({basicLSTMCell_:jM});function qM(e,t,n){let a=_(e,"x","batchToSpaceND"),r=t.reduce((o,l)=>o*l);$(a.rank>=1+t.length,()=>`input rank is ${a.rank} but should be > than blockShape.length ${t.length}`),$(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`),$(a.shape[0]%r===0,()=>`input tensor batch is ${a.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${r}`);let s={x:a},i={blockShape:t,crops:n};return O.runKernel($l,s,i)}var Ac=L({batchToSpaceND_:qM});function KM(e){let t;return e.rank===0||e.rank===1?t=W(e,[1,1,1,e.size]):e.rank===2?t=W(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?t=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]):t=e,t}function XM(e,t,n,a,r,s){s==null&&(s=.001);let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;a!=null&&(p=_(a,"offset","batchNorm")),$(o.rank===l.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),$(p==null||o.rank===p.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),$(u==null||o.rank===u.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let d={x:KM(i),scale:u,offset:p,mean:o,variance:l},c={varianceEpsilon:s},h=O.runKernel($i,d,c);return W(h,i.shape)}var ys=L({batchNorm_:XM});function YM(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),$(o.rank===2||o.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${o.rank}.`),$(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===2||p.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${p.rank}.`),ys(i,o,l,p,u,s)}var Zx=L({batchNorm2d_:YM});function ZM(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),$(o.rank===3||o.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${o.rank}.`),$(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===3||p.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${p.rank}.`),ys(i,o,l,p,u,s)}var Jx=L({batchNorm3d_:ZM});function JM(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),$(o.rank===4||o.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${o.rank}.`),$(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===4||p.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${p.rank}.`),ys(i,o,l,p,u,s)}var Qx=L({batchNorm4d_:JM});function QM(e,t,n){let a=_(e,"x","bincount"),r=_(t,"weights","bincount");$(a.dtype==="int32",()=>`Error in bincount: input dtype must be int32, but got ${a.dtype}`),$(n>=0,()=>`size must be non-negative, but got ${n}.`),$(r.size===a.size||r.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${a.shape}, weights shape: ${r.shape}.`);let s={x:a,weights:r},i={size:n};return O.runKernel(Zh,s,i)}var ev=L({bincount_:QM});function eP(e,t){let n=_(e,"s0","broadcastArgs","int32"),a=_(t,"s1","broadcastArgs","int32");if(n.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${n.rank}`);if(a.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${a.rank}`);let r={s0:n,s1:a};return O.runKernel(Jh,r)}var FS=L({broadcastArgs_:eP});function tP(e,t){let n=_(e,"broadcastTo","x"),a=n.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){let l=n.shape.slice();for(;l.length<t.length;)l.unshift(1);n=W(n,l)}let r=n.shape,s=Array.from(t);for(let l=t.length-1;l>=0;l--)if(r[l]===t[l])s[l]=1;else if(n.shape[l]!==1)throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${t}].`);if(s.map((l,u)=>l>1?u:-1).filter(l=>l>=0).length===0)return rr(n);let i={x:n},o={reps:s};return O.runKernel(hs,i,o)}var Ks=L({broadcastTo_:tP});function nP(e){let t={x:_(e,"x","ceil","float32")};return O.runKernel(xi,t)}var tv=L({ceil_:nP});function gn(e,t,n){let a={shape:e,value:t,dtype:n};return O.runKernel(pc,{},a)}function aP(e,t,n){let a=_(e,"x","clipByValue");if($(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`),t===n)return gn(a.shape,t,a.dtype);let r={x:a},s={clipValueMin:t,clipValueMax:n};return O.runKernel(ds,r,s)}var en=L({clipByValue_:aP});function rP(e){return Ze(e,0)}var nv=L({concat1d_:rP});function sP(e,t){return Ze(e,t)}var av=L({concat2d_:sP});function iP(e,t){return Ze(e,t)}var rv=L({concat3d_:iP});function oP(e,t){return Ze(e,t)}var sv=L({concat4d_:oP});function lP(e,t,n,a,r="NHWC",s=[1,1],i){let o=_(e,"x","conv2d","float32"),l=_(t,"filter","conv2d","float32"),u=o,p=!1;o.rank===3&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(u.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${u.rank}.`),$(l.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${l.rank}.`),Tn("conv2d",a,i);let d=r==="NHWC"?u.shape[3]:u.shape[1];$(d===l.shape[2],()=>`Error in conv2d: depth of input (${d}) must match input depth for filter ${l.shape[2]}.`),$(ur(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);let c={x:u,filter:l},h={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i},m=O.runKernel(vi,c,h);return p?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var $t=L({conv2d_:lP});function uP(e,t,n,a,r="NWC",s=1,i){let o=_(e,"x","conv1d"),l=_(t,"filter","conv1d"),u=o,p=!1;o.rank===2&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1]])),$(u.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${u.rank}.`),$(l.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${l.rank}.`),Tn("conv1d",a,i),$(u.shape[2]===l.shape[1],()=>`Error in conv1d: depth of input (${u.shape[2]}) must match input depth for filter ${l.shape[1]}.`),$(ur(n,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${s}'`),$(r==="NWC",()=>`Error in conv1d: got dataFormat of ${r} but only NWC is currently supported.`);let d=W(l,[1,l.shape[0],l.shape[1],l.shape[2]]),c=W(u,[u.shape[0],1,u.shape[1],u.shape[2]]),h=$t(c,d,[1,n],a,"NHWC",[1,s],i);return p?W(h,[h.shape[2],h.shape[3]]):W(h,[h.shape[0],h.shape[2],h.shape[3]])}var Fm=L({conv1d_:uP});function pP(e,t,n,a,r,s="NHWC",i){$(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,l=t,u=!1;t.rank===3&&(u=!0,l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]),o=[1,e[0],e[1],e[2]]),$(o.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`),$(l.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${l.rank}`),$(n.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${n.rank}`);let p=s==="NHWC"?o[3]:o[1],d=s==="NHWC"?l.shape[3]:l.shape[1];$(p===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${p}) must match input depth for filter ${n.shape[2]}.`),$(d===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${d}) must match output depth for filter ${n.shape[3]}.`),Tn("conv2dDerInput",r,i);let c={dy:l,filter:n},h={strides:a,pad:r,dataFormat:s,dimRoundingMode:i,inputShape:o},m=O.runKernel(wi,c,h);return u?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var iv=L({conv2DBackpropInput_:pP});function cP(e,t,n,a,r,s){let i=_(e,"x","conv2dTranspose"),o=_(t,"filter","conv2dTranspose");return iv(n,i,o,a,r,"NHWC",s)}var Dm=L({conv2dTranspose_:cP});function dP(e,t,n,a,r="NDHWC",s=[1,1,1]){let i=_(e,"x","conv3d"),o=_(t,"filter","conv3d"),l=i,u=!1;i.rank===4&&(u=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(l.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${l.rank}.`),$(o.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`),$(l.shape[4]===o.shape[3],()=>`Error in conv3d: depth of input (${l.shape[4]}) must match input depth for filter ${o.shape[3]}.`),$(ur(n,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),$(r==="NDHWC",()=>`Error in conv3d: got dataFormat of ${r} but only NDHWC is currently supported.`);let p={x:l,filter:o},d={strides:n,pad:a,dataFormat:r,dilations:s},c=O.runKernel(lc,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}var ov=L({conv3d_:dP});function hP(e,t,n,a,r){$(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,i=t,o=!1;t.rank===4&&(o=!0,i=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,e[0],e[1],e[2],e[3]]);let l=s[4],u=i.shape[4];$(s.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),$(i.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),$(n.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${n.rank}`),$(l===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${l}) must match input depth for filter ${n.shape[3]}.`),$(u===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${u}) must match output depth for filter ${n.shape[4]}.`);let p={dy:i,filter:n},d={pad:r,strides:a,inputShape:s},c=O.runKernel(nm,p,d);return o?W(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}var DS=L({conv3DBackpropInput_:hP});function mP(e,t,n,a,r){let s=_(e,"x","conv3dTranspose"),i=_(t,"filter","conv3dTranspose");return DS(n,s,i,a,r)}var lv=L({conv3dTranspose_:mP});function fP(e){let t={x:_(e,"x","cos","float32")};return O.runKernel(ki,t)}var $c=L({cos_:fP});function gP(e){let t={x:_(e,"x","cosh","float32")};return O.runKernel(Ii,t)}var Rm=L({cosh_:gP});function yP(e,t=0,n=!1,a=!1){let r={x:_(e,"x","cumprod")},s={axis:t,exclusive:n,reverse:a};return O.runKernel(Dl,r,s)}var Kp=L({cumprod_:yP});function bP(e,t=0,n=!1,a=!1){let r={x:_(e,"x","cumsum")},s={axis:t,exclusive:n,reverse:a};return O.runKernel(Si,r,s)}var Mm=L({cumsum_:bP});function xP(e,t,n,a=!1){let r=_(e,"x","denseBincount"),s=_(t,"weights","denseBincount");$(r.dtype==="int32",()=>`Error in denseBincount: input dtype must be int32, but got ${r.dtype}`),$(r.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${r.rank}.`),$(n>=0,()=>`size must be non-negative, but got ${n}.`),$(s.size===r.size||s.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${r.shape}, weights shape: ${s.shape}.`);let i={x:r,weights:s},o={size:n,binaryOutput:a};return O.runKernel(am,i,o)}var Sh=L({denseBincount_:xP});function vP(e,t,n="NHWC"){let a=_(e,"x","depthToSpace","float32"),r=n==="NHWC"?a.shape[1]:a.shape[2],s=n==="NHWC"?a.shape[2]:a.shape[3],i=n==="NHWC"?a.shape[3]:a.shape[1];$(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),$(r*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${r} and ${t}  for depthToSpace with input shape
-    ${a.shape}`),$(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${s} and ${t} for depthToSpace with input shape
-        ${a.shape}`),$(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${a.shape}`);let o={x:a},l={blockSize:t,dataFormat:n};return O.runKernel(Ml,o,l)}var uv=L({depthToSpace_:vP});function wP(e,t,n,a,r="NHWC",s=[1,1],i){let o=_(e,"x","depthwiseConv2d","float32"),l=_(t,"filter","depthwiseConv2d","float32"),u=o,p=!1;o.rank===3&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(u.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${u.rank}.`),$(l.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${l.rank}.`);let d=r==="NHWC"?u.shape[3]:u.shape[1];$(d===l.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${d}) must match the inChannels dimension in filter ${l.shape[2]}.`),Tn("depthwiseConv2d",a,i);let c={x:u,filter:l},h={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i},m=O.runKernel(Ti,c,h);return p?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var bs=L({depthwiseConv2d_:wP});function kP(e){let t={x:_(e,"x","diag")};return O.runKernel(im,t)}var RS=L({diag_:kP});function IP(e,t,n,a,r=[1,1],s="NHWC"){let i=_(e,"x","dilation2d"),o=_(t,"filter","dilation2d");$(i.rank===3||i.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),$(o.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`),$(s==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let l=i,u=!1;i.rank===3&&(l=W(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=!0);let p={x:l,filter:o},d={strides:n,pad:a,dilations:r},c=O.runKernel(uc,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var pv=L({dilation2d_:IP});function SP(e,t){let n=_(e,"a","equal","string_or_numeric"),a=_(t,"b","equal","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Ol,r)}var Qn=L({equal_:SP});function TP(e,t,n){let a=_(t,"a","where"),r=_(n,"b","where"),s=_(e,"condition","where","bool"),i=lt(lt(s.shape,a.shape),r.shape),o=Ks(s,i),l=Ks(a,i),u=Ks(r,i),p={condition:o,t:l,e:u};return O.runKernel(pu,p)}var mn=L({where_:TP});function NP(e){let t={x:_(e,"x","zerosLike")};return O.runKernel(Iu,t)}var qe=L({zerosLike_:NP});function CP(e,t){let n=_(e,"a","div"),a=_(t,"b","div");[n,a]=_t(n,a);let r=he(n,a),s=qe(r),i=Qn(a,s);return mn(i,s,r)}var cv=L({divNoNan_:CP});function _P(e,t){let n=_(e,"t1","dot"),a=_(t,"t2","dot");$((n.rank===1||n.rank===2)&&(a.rank===1||a.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${a.rank}.`);let r=n.rank===1?n.size:n.shape[1],s=a.rank===1?a.size:a.shape[0];if($(r===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${r} and ${s}.`),n.rank===1&&a.rank===1){let i=W(n,[1,-1]),o=W(a,[-1,1]),l=Fe(i,o);return W(l,[])}else if(n.rank===1&&a.rank===2){let i=W(n,[1,-1]),o=W(a,[a.shape[0],a.shape[1]]),l=Fe(i,o);return W(l,[l.size])}else if(n.rank===2&&a.rank===1){let i=W(a,[-1,1]),o=Fe(n,i);return W(o,[o.size])}else{let i=W(a,[a.shape[0],a.shape[1]]);return Fe(n,i)}}var dv=L({dot_:_P});function EP(e,...t){let n=t.map((r,s)=>_(r,`tensors${s}`,"einsum")),a={equation:e};return O.runKernel(om,n,a)}var MS=L({einsum_:EP});function AP(e){let t={x:_(e,"x","elu","float32")};return O.runKernel(Ci,t)}var Nu=L({elu_:AP});function $P(e){let t=_(e,"x","erf");$(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=oe(t,"float32"));let n={x:t};return O.runKernel(Pl,n)}var hv=L({erf_:$P});function mv(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function PS(e,t,n){let a=e.length+t.length,r=[],s=0,i=0;for(let o=0;o<a;o++)n.indexOf(o)===-1?r.push(e[s++]):r.push(t[i++]);return r}function OS(e,t){let n=[],a=e.length;for(let s=0;s<a;s++)t.indexOf(s)===-1&&n.push(e[s]);let r=t.map(s=>e[s]);return[n,r]}function si(e,t){let n=t.map(a=>1);return PS(e,n,t)}function FP(e,t,n){$(mv(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function LS(e,t){if(mv(e,t))return null;let n=[];for(let a=0;a<t;++a)e.indexOf(a)===-1&&n.push(a);return e.forEach(a=>n.push(a)),n}function fv(e){return e.map((t,n)=>[n,t]).sort((t,n)=>t[1]-n[1]).map(t=>t[0])}function DP(e,t){let n=[];for(let a=t-e;a<t;++a)n.push(a);return n}function RP(e,t=null,n=!1){let a={x:_(e,"x","max")},r={reductionIndices:t,keepDims:n};return O.runKernel(Pi,a,r)}var ha=L({max_:RP});function MP(e,t=null,n=!1){let a={x:_(e,"x","min")},r={axis:t,keepDims:n};return O.runKernel(Wi,a,r)}var il=L({min_:MP});function PP(e,t){let n=_(e,"base","pow"),a=_(t,"exp","pow");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(ji,r)}var _r=L({pow_:PP});function be(e,t){if((hn(e)&&t!=="string"||Array.isArray(e))&&t!=="complex64")throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if(t==="string"&&hn(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return fs(e,[],[],t)}function OP(e){let t={x:_(e,"x","sqrt","float32")};return O.runKernel(ro,t)}var ln=L({sqrt_:OP});function LP(e){let t=_(e,"x","square"),n={};return O.runKernel("Square",{x:t},n)}var ot=L({square_:LP});function zP(e,t=null,n=!1){let a=_(e,"x","sum");a.dtype==="bool"&&(a=oe(a,"int32"));let r={x:a},s={axis:t,keepDims:n};return O.runKernel(so,r,s)}var fe=L({sum_:zP});function WP(e,t="euclidean",n=null,a=!1){e=_(e,"x","norm");let r=zS(e,t,n),s=r.shape;if(a){let i=Aa(n,e.shape);s=si(r.shape,i)}return W(r,s)}function zS(e,t,n=null){if(e.rank===0)return Lt(e);if(e.rank!==1&&n===null)return zS(W(e,[-1]),t,n);if(e.rank===1||typeof n=="number"||Array.isArray(n)&&n.length===1){if(t===1)return fe(Lt(e),n);if(t===1/0)return ha(Lt(e),n);if(t===-1/0)return il(Lt(e),n);if(t==="euclidean"||t===2)return ln(fe(_r(Lt(e),be(2,"int32")),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(n)&&n.length===2){if(t===1)return ha(fe(Lt(e),n[0]),n[1]-1);if(t===1/0)return ha(fe(Lt(e),n[1]),n[0]);if(t===-1/0)return il(fe(Lt(e),n[1]),n[0]);if(t==="fro"||t==="euclidean")return ln(fe(ot(e),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${n}`)}var Cu=L({norm_:WP});function BP(e,t=null,n=!1){return Cu(e,"euclidean",t,n)}var gv=L({euclideanNorm_:BP});function VP(e){let t={x:_(e,"x","exp")};return O.runKernel(_i,t)}var fn=L({exp_:VP});function UP(e,t=0){let n=_(e,"x","expandDims","string_or_numeric");$(t<=n.rank,()=>"Axis must be <= rank of the tensor");let a={input:n},r={dim:t};return O.runKernel(Ll,a,r)}var Zt=L({expandDims_:UP});function GP(e){let t={x:_(e,"x","expm1")};return O.runKernel(zl,t)}var yv=L({expm1_:GP});function HP(e,t){let n=_(e,"x","tile","string_or_numeric");$(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);let a={x:n},r={reps:t};return O.runKernel(hs,a,r)}var Ln=L({tile_:HP});function jP(e,t,n,a="float32"){t==null&&(t=e);let r=Pe([e,t],a),s=e<=t?e:t;for(let o=0;o<s;++o)r.set(1,o,o);let i=W(r.toTensor(),[e,t]);if(n==null)return i;if(n.length===1)return Ln(Zt(i,0),[n[0],1,1]);if(n.length===2)return Ln(Zt(Zt(i,0),0),[n[0],n[1],1,1]);if(n.length===3)return Ln(Zt(Zt(Zt(i,0),0),0),[n[0],n[1],n[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)}var Pm=L({eye_:jP});function qP(e){let t={x:_(e,"x","floor","float32")};return O.runKernel(Ei,t)}var _u=L({floor_:qP});function KP(e,t,n=0,a=0){let r=_(e,"x","gather"),s=_(t,"indices","gather","int32"),i={x:r,indices:s},o={axis:n,batchDims:a};return O.runKernel(Bl,i,o)}var Eu=L({gather_:KP});function XP(e,t){let n=_(e,"a","greater","string_or_numeric"),a=_(t,"b","greater","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Ul,r)}var Nn=L({greater_:XP});function YP(e,t){let n=_(e,"a","greaterEqual","string_or_numeric"),a=_(t,"b","greaterEqual","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Fi,r)}var Er=L({greaterEqual_:YP});function ZP(e){let t={x:_(e,"x","isFinite")};return O.runKernel(Gl,t)}var bv=L({isFinite_:ZP});function JP(e){let t={x:_(e,"x","isInf")};return O.runKernel(Hl,t)}var xv=L({isInf_:JP});function QP(e){let t={x:_(e,"x","isNaN")};return O.runKernel(jl,t)}var vv=L({isNaN_:QP});function eO(e,t=.2){let n={x:_(e,"x","leakyRelu")},a={alpha:t};return O.runKernel(Ri,n,a)}var Fc=L({leakyRelu_:eO});function tO(e,t){let n=_(e,"a","less","string_or_numeric"),a=_(t,"b","less","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(ql,r)}var Om=L({less_:tO});function nO(e,t){let n=_(e,"a","lessEqual","string_or_numeric"),a=_(t,"b","lessEqual","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Kl,r)}var xs=L({lessEqual_:nO});function WS(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");let a={start:e,stop:t,num:n};return O.runKernel(dm,{},a)}function aO(e,t=5,n=1,a=1,r=.5){let s=_(e,"x","localResponseNormalization");$(s.rank===4||s.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
-               rank ${s.rank}.`),$(tl(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,o=!1;s.rank===3&&(o=!0,i=W(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let l={x:i},u={depthRadius:t,bias:n,alpha:a,beta:r},p=O.runKernel(cc,l,u);return o?W(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var wv=L({localResponseNormalization_:aO});function rO(e){let t={x:_(e,"x","log","float32")};return O.runKernel(Mi,t)}var ea=L({log_:rO});function sO(e){let t={x:_(e,"x","log1p")};return O.runKernel(Xl,t)}var Dc=L({log1p_:sO});function iO(e){return $(es(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{let a=_(t,"x","tf.grad","string_or_numeric"),r=n!=null?_(n,"dy","tf.grad"):null;return O.tidy(()=>{let{value:s,grads:i}=O.gradients(()=>e(a),[a],r);return r!=null&&Sn(s.shape,r.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Lm(i),i[0]})}}function oO(e){return $(es(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{$(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let a=jp(t,"args","tf.grads","string_or_numeric"),r=n!=null?_(n,"dy","tf.grads"):null;return O.tidy(()=>{let{value:s,grads:i}=O.gradients(()=>e(...a),a,r);return r!=null&&Sn(s.shape,r.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Lm(i),i})}}function lO(e){return $(es(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{$(t instanceof Te,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),$(n==null||n instanceof Te,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:a,value:r}=O.gradients(()=>e(t),[t],n);return Lm(a),{grad:a[0],value:r}}}function uO(e){return $(es(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{$(Array.isArray(t)&&t.every(r=>r instanceof Te),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),$(n==null||n instanceof Te,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let a=O.gradients(()=>e(...t),t,n);return n!=null&&Sn(a.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Lm(a.grads),a}}function BS(e,t){$(es(e),()=>"The f passed in variableGrads(f) must be a function"),$(t==null||Array.isArray(t)&&t.every(u=>u instanceof ts),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let n=t!=null;if(!n){t=[];for(let u in O.registeredVariables)t.push(O.registeredVariables[u])}let a=n?t.filter(u=>!u.trainable):null,r=t.length;t=t.filter(u=>u.trainable),$(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${r} variables is trainable.`);let s=!0,{value:i,grads:o}=O.gradients(e,t,null,s);$(o.some(u=>u!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),$(i.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${i.rank} tensor`);let l={};return t.forEach((u,p)=>{o[p]!=null&&(l[u.name]=o[p])}),a!=null&&a.forEach(u=>l[u.name]=null),{value:i,grads:l}}function or(e){return O.customGrad(e)}function Lm(e){if(e.filter(t=>t==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
-    the f you passed encloses all operations that lead from x to y.`)}function pO(e){let t={x:_(e,"x","softplus")};return O.runKernel(fu,t)}var ho=L({softplus_:pO});function cO(e){let t=_(e,"x","logSigmoid");return or(n=>({value:bt(ho(bt(n))),gradFunc:a=>z(a,da(bt(n)))}))(t)}var kv=L({logSigmoid_:cO});function dO(e,t){let n=_(e,"a","sub"),a=_(t,"b","sub");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(lo,r)}var pe=L({sub_:dO});function hO(e,t=-1){let n=_(e,"logits","logSoftmax");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);return or((a,r)=>{let s=ha(a,t,!0),i=pe(a,s),o=pe(oe(i,"float32"),ea(fe(fn(i),t,!0)));return r([o]),{value:o,gradFunc:(l,u)=>{let[p]=u,d=!0,c=fn(p);return pe(l,z(fe(l,t,d),c))}}})(n)}var zm=L({logSoftmax_:hO});function mO(e,t=null,n=!1){let a=_(e,"x","logSumExp"),r=Aa(t,a.shape),s=ha(a,r,!0),i=pe(a,s),o=fn(i),l=fe(o,r),u=ea(l),p=Y(W(s,u.shape),u);if(n){let d=si(p.shape,r);return W(p,d)}return p}var Wm=L({logSumExp_:mO});function fO(e,t){let n=_(e,"a","logicalAnd","bool"),a=_(t,"b","logicalAnd","bool");lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Yl,r)}var Ea=L({logicalAnd_:fO});function gO(e){let t={x:_(e,"x","logicalNot","bool")};return O.runKernel(Zl,t)}var Rc=L({logicalNot_:gO});function yO(e,t){let n=_(e,"a","logicalOr","bool"),a=_(t,"b","logicalOr","bool");lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Jl,r)}var Bm=L({logicalOr_:yO});function bO(e,t){let n=_(e,"a","logicalXor","bool"),a=_(t,"b","logicalXor","bool");return lt(n.shape,a.shape),Ea(Bm(e,t),Rc(Ea(e,t)))}var Iv=L({logicalXor_:bO}),Hd=2147483648;function xO(e,t,n="left"){let a=_(e,"sortedSequence","searchSorted"),r=_(t,"values","searchSorted"),s=a.shape[a.shape.length-1],i=r.shape[r.shape.length-1],o=W(a,[-1,s]),l=W(r,[-1,i]);if(o.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(o.shape[0]!==l.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(mt(l.shape)>=Hd)throw new Error(`values tensor size must less than ${Hd}`);if(o.shape[1]>=Hd)throw new Error(`trailing dim_size must less than ${Hd} for int32 output type, was ${o.shape[1]}`);let u={sortedSequence:o,values:l},p={side:n};return O.runKernel(Sm,u,p)}var Vm=L({searchSorted_:xO});function VS(e,t){return Vm(e,t,"left")}function vO(e,t,n,a,r){let s=_(e,"x","maxPool"),i=1,o=s,l=!1;s.rank===3&&(l=!0,o=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(o.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${o.rank}.`),$(ur(n,i),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${i}'`),Tn("maxPool",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r},d=O.runKernel(Li,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var Dt=L({maxPool_:vO});function wO(e,t=[1,1,1],n,a,r,s="NDHWC"){let i=_(e,"x","maxPool3d"),o=i,l=!1;i.rank===4&&(l=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(o.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`),$(s==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Tn("maxPool3d",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r,dataFormat:s},d=O.runKernel(dc,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var Sv=L({maxPool3d_:wO});function kO(e,t,n,a,r=!1){let s={x:_(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:n,pad:a,includeBatchInIndex:r},o=O.runKernel(gm,s,i);return{result:o[0],indexes:o[1]}}var US=L({maxPoolWithArgmax_:kO});function IO(e,t){let n=_(e,"a","maximum"),a=_(t,"b","maximum");[n,a]=_t(n,a),n.dtype==="bool"&&(n=oe(n,"int32"),a=oe(a,"int32")),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Oi,r)}var pr=L({maximum_:IO});function SO(e,t=null,n=!1){let a={x:_(e,"x","mean")},r={axis:t,keepDims:n};return O.runKernel(zi,a,r)}var Nt=L({mean_:SO});function It(e,t="float32"){if(t==="complex64"){let a=It(e,"float32"),r=It(e,"float32");return Cr(a,r)}let n=Kh(mt(e),t);return O.makeTensor(n,e,t)}function Zn(e,t="float32"){if(t==="complex64"){let a=Zn(e,"float32"),r=It(e,"float32");return Cr(a,r)}let n=vx(mt(e),t);return O.makeTensor(n,e,t)}function GS(e,t,{indexing:n="xy"}={}){if(n!=="xy"&&n!=="ij")throw new TypeError(`${n} is not a valid third argument to meshgrid`);if(e===void 0)return[];let a=_(e,"x","meshgrid",e instanceof Te?e.dtype:"float32");if(t===void 0)return[a];let r=_(t,"y","meshgrid",t instanceof Te?t.dtype:"float32"),s=mt(a.shape),i=mt(r.shape);return n==="xy"?(a=W(a,[1,-1]),r=W(r,[-1,1]),[Fe(Zn([i,1],a.dtype),a),Fe(r,Zn([1,s],r.dtype))]):(a=W(a,[-1,1]),r=W(r,[1,-1]),[Fe(a,Zn([1,i],a.dtype)),Fe(Zn([s,1],r.dtype),r)])}function TO(e,t){let n=_(e,"a","minimum"),a=_(t,"b","minimum");[n,a]=_t(n,a),n.dtype==="bool"&&(n=oe(n,"int32"),a=oe(a,"int32")),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Bi,r)}var Au=L({minimum_:TO});function NO(e,t,n){$(n==="reflect"||n==="symmetric",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);let a=_(e,"x","mirrorPad");if(a.rank===0)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");$(t.length===a.rank,()=>`Padding doesn't match input. Must be ${a.rank}. Got ${t.length}.`);let r=n==="reflect"?1:0;for(let o=0;o<a.rank;o++)$(t[o].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),$(t[o][0]>=0&&t[o][0]<=a.shape[o]-r&&t[o][1]>=0&&t[o][1]<=a.shape[o]-r,()=>`Padding in dimension ${o} cannot be greater than or equal to ${a.shape[o]-r} or less than 0 for input of shape ${a.shape}`);let s={paddings:t,mode:n},i={x:a};return O.runKernel(Vi,i,s)}var Tv=L({mirrorPad_:NO});function CO(e,t){let n=_(e,"a","mod"),a=_(t,"b","mod");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Ql,r)}var Nv=L({mod_:CO});function _O(e,t=null,n=!1){e=_(e,"x","moments");let a=Aa(t,e.shape),r=Nt(e,a,n),s=r.shape;n||(s=si(r.shape,a));let i=ot(pe(oe(e,"float32"),W(r,s))),o=Nt(i,a,n);return{mean:r,variance:o}}var Mc=L({moments_:_O});function EO(e,t,n,a){let r=_(t,"data","multiRNNCell"),s=jp(n,"c","multiRNNCell"),i=jp(a,"h","multiRNNCell"),o=r,l=[];for(let d=0;d<e.length;d++){let c=e[d](o,s[d],i[d]);l.push(c[0]),l.push(c[1]),o=c[1]}let u=[],p=[];for(let d=0;d<l.length;d+=2)u.push(l[d]),p.push(l[d+1]);return[u,p]}var HS=L({multiRNNCell_:EO});function AO(e,t,n,a=!1){let r=_(e,"logits","multinomial"),s=r.size,i=r.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);n=n||Math.random();let o={logits:i===1?W(r,[1,-1]):r},l={numSamples:t,seed:n,normalized:a},u=O.runKernel(ym,o,l);return i===1?W(u,[u.size]):u}var jS=L({multinomial_:AO});function $O(e,t){let n=_(e,"a","notEqual","string_or_numeric"),a=_(t,"b","notEqual","string_or_numeric");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(tu,r)}var ii=L({notEqual_:$O});function FO(e){let t={x:_(e,"x","onesLike")};return O.runKernel(su,t)}var ta=L({onesLike_:FO});function DO(e,t){let n=_(e,"v1","outerProduct"),a=_(t,"v2","outerProduct");$(n.rank===1&&a.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${a.rank}.`);let r=W(n,[-1,1]),s=W(a,[1,-1]);return Fe(r,s)}var qS=L({outerProduct_:DO});function RO(e,t,n=0){let a=_(e,"x","pad");if(a.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let r={paddings:t,constantValue:n},s={x:a};return O.runKernel(Hi,s,r)}var ya=L({pad_:RO});function MO(e,t,n=0){return $(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),ya(e,[t],n)}var KS=L({pad1d_:MO});function PO(e,t,n=0){return $(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ya(e,t,n)}var XS=L({pad2d_:PO});function OO(e,t,n=0){return $(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ya(e,t,n)}var YS=L({pad3d_:OO});function LO(e,t,n=0){return $(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ya(e,t,n)}var ZS=L({pad4d_:LO});function zO(e,t,n){let a=_(e,"x","spaceToBatchND");$(a.rank>=1+t.length,()=>`input rank ${a.rank} should be > than [blockShape] ${t.length}`),$(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),$(a.shape.reduce((i,o,l)=>l>0&&l<=t.length?i&&(o+n[l-1][0]+n[l-1][1])%t[l-1]===0:i,!0),()=>`input spatial dimensions ${a.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);let r={x:a},s={blockShape:t,paddings:n};return O.runKernel(gu,r,s)}var Pc=L({spaceToBatchND_:zO});function WO(e,t,n,a,r,s,i){r==null&&(r=[1,1]),s==null&&(s=1),a===0&&(a="valid");let o=_(e,"x","maxPool"),l=o,u=!1;o.rank===3&&(u=!0,l=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(ur(s,r),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${r}'`);let p=_S(l.shape,t,s,r,a),d=[p.dilationHeight,p.dilationWidth],c;a==="same"?c=VO([p.filterHeight,p.filterWidth],d):c=[[0,0],[0,0]];let h=d[0]===1&&d[1]===1,[m,f]=BO([p.inHeight,p.inWidth],d,c),g=h?a:"valid",y=h?l:Pc(l,d,m),b=(n==="avg"?()=>ga(y,t,s,g,i):()=>Dt(y,t,s,g,i))(),x=h?b:Ac(b,d,f);return u?W(x,[x.shape[1],x.shape[2],x.shape[3]]):x}function BO(e,t,n){let a=n.map(p=>p[0]),r=n.map(p=>p[1]),s=e.concat(a,r),i=t.map((p,d)=>(p-s[d]%p)%p),o=r.map((p,d)=>p+i[d]),l=t.map((p,d)=>[a[d],o[d]]),u=t.map((p,d)=>[0,i[d]]);return[l,u]}function VO(e,t){let n=e.map((s,i)=>s+(s-1)*(t[i]-1)).map(s=>s-1),a=n.map(s=>Math.floor(s/2)),r=n.map((s,i)=>s-a[i]);return n.map((s,i)=>[a[i],r[i]])}var Cv=L({pool_:WO});function UO(e,t){let n=_(e,"x","prelu"),a=_(t,"alpha","prelu"),r={x:n,alpha:a};return O.runKernel(qi,r)}var Oc=L({prelu_:UO});function GO(e,t=null,n=!1){let a=_(e,"x","prod");a.dtype==="bool"&&(a=oe(a,"int32"));let r={x:a},s={axis:t,keepDims:n};return O.runKernel(Ki,r,s)}var _v=L({prod_:GO});function HO(e,t,n,a){let r=e.map((p,d)=>_(p,`tensors${d}`,"raggedGather","int32")),s=_(t,"paramsDenseValues","raggedGather"),i=_(n,"indices","raggedGather","int32"),o={paramsNestedSplits:r,paramsDenseValues:s,indices:i},l={outputRaggedRank:a},u=O.runKernel(bm,o,l);return{outputNestedSplits:u.slice(0,u.length-1),outputDenseValues:u[u.length-1]}}var JS=L({raggedGather_:HO});function jO(e,t,n){let a=_(e,"starts","raggedRange"),r=_(t,"limits","raggedRange",a.dtype),s=_(n,"deltas","raggedRange",a.dtype),i={starts:a,limits:r,deltas:s},o=O.runKernel(xm,i);return{rtNestedSplits:o[0],rtDenseValues:o[1]}}var QS=L({raggedRange_:jO});function qO(e,t,n,a,r){let s=_(e,"shape","raggedTensorToTensor","int32"),i=_(t,"values","raggedTensorToTensor"),o=_(n,"defaultValue","raggedTensorToTensor",i.dtype),l=a.map((d,c)=>_(d,`tensors${c}`,"raggedTensorToTensor","int32")),u={shape:s,values:i,defaultValue:o,rowPartitionTensors:l},p={rowPartitionTypes:r};return O.runKernel(vm,u,p)}var eT=L({raggedTensorToTensor_:qO});function KO(e,t,n){let a=mt(e),r=null;if(n==null||n==="float32")r=new Float32Array(a);else if(n==="int32")r=new Int32Array(a);else if(n==="bool")r=new Uint8Array(a);else throw new Error(`Unknown data type ${n}`);for(let s=0;s<a;s++)r[s]=t();return O.makeTensor(r,e,n)}var tT=L({rand_:KO}),Ev=us(Hh()),Av=class{constructor(e,t,n,a,r){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=a,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let s=r||Math.random();this.random=Ev.alea(s.toString())}nextValue(){if(!isNaN(this.nextVal)){let a=this.nextVal;return this.nextVal=NaN,a}let e,t,n=!1;for(;!n;){let a,r,s;do a=2*this.random()-1,r=2*this.random()-1,s=a*a+r*r;while(s>=1||s===0);let i=Math.sqrt(-2*Math.log(s)/s);e=this.mean+this.stdDev*a*i,t=this.mean+this.stdDev*r*i,(!this.truncated||this.isValidTruncated(e))&&(n=!0)}return(!this.truncated||this.isValidTruncated(t))&&(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return this.dtype==null||this.dtype==="float32"?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}},XO=class{constructor(e,t,n,a){this.alpha=e,this.beta=1/t,this.dtype=n;let r=a||Math.random();this.randu=Ev.alea(r.toString()),this.randn=new Av(0,1,n,!1,this.randu()),e<1?this.d=e+2/3:this.d=e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,n,a,r,s;for(;;){do a=this.randn.nextValue(),s=1+this.c*a;while(s<=0);if(s*=s*s,e=a*a,t=1-.331*e*e,n=.5*e+this.d*(1-s+Math.log(s)),r=this.randu(),r<t||Math.log(r)<n)break}return s=1/this.beta*this.d*s,this.alpha<1&&(s*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(s)}convertValue(e){return this.dtype==="float32"?e:Math.round(e)}},YO=class{constructor(e=0,t=1,n,a){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=e,this.range=t-e,this.dtype=n,a==null&&(a=Math.random()),typeof a=="number"&&(a=a.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=Ev.alea(a)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function ZO(e,t,n=1,a="float32",r){if(n==null&&(n=1),a==null&&(a="float32"),a!=="float32"&&a!=="int32")throw new Error(`Unsupported data type ${a}`);let s=new XO(t,n,a,r),i=Pe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var nT=L({randomGamma_:ZO});function JO(e,t=0,n=1,a,r){if(a!=null&&a==="bool")throw new Error(`Unsupported data type ${a}`);let s=new Av(t,n,a,!1,r),i=Pe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var Um=L({randomNormal_:JO});function QO(e,t,n){if(t!=null&&t==="bool")throw new Error(`Unsupported data type ${t}`);return Um(e,0,1,t,n)}var aT=L({randomStandardNormal_:QO});function e3(e,t=0,n=1,a="float32",r){let s=Pe(e,a),i=new YO(t,n,null,r);for(let o=0;o<s.values.length;o++)s.values[o]=i.nextValue();return s.toTensor()}var $u=L({randomUniform_:e3});function ol(e,t,n=1,a="float32"){if(n===0)throw new Error("Cannot have a step of zero");let r={start:e,stop:t,step:n,dtype:a};return O.runKernel(hc,{},r)}function t3(e){let t={x:_(e,"x","reciprocal")};return O.runKernel(ou,t)}var $v=L({reciprocal_:t3});function n3(e){let t={x:_(e,"x","relu")};return O.runKernel(Xi,t)}var Xe=L({relu_:n3});function a3(e){let t={x:_(e,"x","relu6")};return O.runKernel(Ji,t)}var Gm=L({relu6_:a3});function r3(e,t){let n={x:_(e,"x","reverse")},a={dims:t};return O.runKernel(Qi,n,a)}var fa=L({reverse_:r3});function s3(e){let t=_(e,"x","reverse");return $(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),fa(t,0)}var rT=L({reverse1d_:s3});function i3(e,t){let n=_(e,"x","reverse");return $(n.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),fa(n,t)}var sT=L({reverse2d_:i3});function o3(e,t){let n=_(e,"x","reverse");return $(n.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),fa(n,t)}var iT=L({reverse3d_:o3});function l3(e,t){let n=_(e,"x","reverse");return $(n.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),fa(n,t)}var oT=L({reverse4d_:l3});function u3(e){let t={x:_(e,"x","round")};return O.runKernel(eo,t)}var Hm=L({round_:u3});function p3(e){let t={x:_(e,"x","rsqrt","float32")};return O.runKernel(to,t)}var jm=L({rsqrt_:p3});function c3(e){let t={x:_(e,"x","selu")};return O.runKernel(cu,t)}var qm=L({selu_:c3});function d3(e,t,n,a,r,s=[1,1],i="NHWC"){let o=_(e,"x","separableConv2d"),l=_(t,"depthwiseFilter","separableConv2d"),u=_(n,"pointwiseFilter","separableConv2d"),p=o,d=!1;if(o.rank===3&&(d=!0,p=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),i==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");$(p.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${p.rank}.`),$(l.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${l.rank}.`),$(u.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${l.rank}.`),$(u.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${u.shape[0]}.`),$(u.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${u.shape[1]}.`);let c=l.shape[2],h=l.shape[3];$(u.shape[2]===c*h,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${c*h}, but got ${u.shape[2]}.`);let m=bs(p,l,a,r,i,s),f=$t(m,u,1,"valid",i);return d?W(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var vs=L({separableConv2d_:d3});async function h3(e,t){let n=_(e,"x","setdiff1d"),a=_(t,"y","setdiff1d");$(n.dtype===a.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${a.dtype}).`),$(n.rank===1,()=>`x should be 1D tensor, but got x (${n.shape}).`),$(a.rank===1,()=>`y should be 1D tensor, but got y (${a.shape}).`);let r=await n.data(),s=await a.data(),i=new Set(s),o=0;for(let p=0;p<r.length;p++)i.has(r[p])||o++;let l=new Ht([o],n.dtype),u=new Ht([o],"int32");for(let p=0,d=0;p<r.length;p++)i.has(r[p])||(l.values[d]=r[p],u.values[d]=p,d++);return[l.toTensor(),u.toTensor()]}var lT=h3;function m3(e){let t={x:_(e,"x","sign")};return O.runKernel(mu,t)}var Fv=L({sign_:m3});function f3(e){let t={x:_(e,"x","sin","float32")};return O.runKernel(no,t)}var Km=L({sin_:f3});function g3(e){let t={x:_(e,"x","sinh")};return O.runKernel(hu,t)}var Xm=L({sinh_:g3});function y3(e,t,n){let a=_(e,"x","slice1d");return $(a.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${a.rank} tensor`),Be(a,[t],[n])}var Lc=L({slice1d_:y3});function b3(e,t,n){let a=_(e,"x","slice2d");return $(a.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var Ym=L({slice2d_:b3});function x3(e,t,n){let a=_(e,"x","slice3d");return $(a.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var mo=L({slice3d_:x3});function v3(e,t,n){let a=_(e,"x","slice4d");return $(a.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var ll=L({slice4d_:v3});function w3(e,t=-1){let n=_(e,"logits","softmax","float32");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);let a={logits:n},r={dim:t};return O.runKernel(io,a,r)}var Ka=L({softmax_:w3});function k3(e){$(e.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);let t={input:e};return O.runKernel(um,t)}var zc=L({fft_:k3});function I3(e){$(e.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);let t={input:e};return O.runKernel(pm,t)}var ul=L({ifft_:I3});function S3(e){let t=e.shape[e.shape.length-1],n=e.size/t,a;if(t<=2){let r=W(e,[n,t]);a=ul(r)}else{let r=[n,2*(t-1)],s=W(sl(e),[n,t]),i=W(Cc(e),[n,t]),o=fa(Be(s,[0,1],[n,t-2]),1),l=z(fa(Be(i,[0,1],[n,t-2]),1),be(-1)),u=Ze([s,o],1),p=Ze([i,l],1),d=W(Cr(u,p),[r[0],r[1]]);a=ul(d)}if(a=sl(a),e.rank===3&&e.shape[0]!==0){let r=a,s=e.shape[0];a=W(a,[s,a.shape[0]/s,a.shape[1]]),r.dispose()}return a}var Zm=L({irfft_:S3});function T3(e,t,n=0){let a={x:_(e,"x","split")},r={numOrSizeSplits:t,axis:n};return O.runKernel(yu,a,r)}var zn=L({split_:T3});function N3(e,t){$(e.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let n=e.shape[e.shape.length-1],a=e.size/n,r;if(t!=null&&t<n){let m=e.shape.map(g=>0),f=e.shape.map(g=>g);f[e.shape.length-1]=t,r=Be(e,m,f),n=t}else if(t!=null&&t>n){let m=e.shape.map(f=>f);m[e.shape.length-1]=t-n,r=Ze([e,It(m)],e.shape.length-1),n=t}else r=e;let s=qe(r),i=W(Cr(r,s),[a,n]),o=zc(i),l=Math.floor(n/2)+1,u=sl(o),p=Cc(o),d=zn(u,[l,n-l],u.shape.length-1),c=zn(p,[l,n-l],p.shape.length-1),h=r.shape.slice();return h[r.shape.length-1]=l,W(Cr(d[0],c[0]),h)}var Wc=L({rfft_:N3});function C3(e,t){let n=_(e,"a","squaredDifference"),a=_(t,"b","squaredDifference");[n,a]=_t(n,a),lt(n.shape,a.shape);let r={a:n,b:a},s={};return O.runKernel(oo,r,s)}var Jm=L({squaredDifference_:C3});function _3(e,t){let n=_(e,"x","squeeze","string_or_numeric");return W(n,II(n.shape,t).newShape)}var ws=L({squeeze_:_3});function E3(e,t=0){let n=jp(e,"tensors","stack","string_or_numeric");$(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length>0&&$(t<=n[0].rank,()=>"Axis must be <= rank of the tensor");let a=n,r={axis:t};return O.runKernel(iu,a,r)}var Ft=L({stack_:E3});function A3(e,t=0){let n={x:_(e,"x","step")},a={alpha:t};return O.runKernel(ms,n,a)}var fo=L({step_:A3});function $3(e,t,n,a,r=0,s=0,i=0,o=0,l=0){let u={x:_(e,"x","stridedSlice","string_or_numeric")},p={begin:t,end:n,strides:a,beginMask:r,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};return O.runKernel(xu,u,p)}var Dv=L({stridedSlice_:$3});function F3(e){let t={x:_(e,"x","tan","float32")};return O.runKernel(uo,t)}var Rv=L({tan_:F3});function Ke(e,t){hi(e);let n=ir(e,t);if(n.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");return fs(e,null,n,t)}function _a(e,t,n){if(hi(e),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");let a=ir(e,n);if(a.length!==2&&a.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return fs(e,t,a,n)}function $a(e,t,n){if(hi(e),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");let a=ir(e,n);if(a.length!==4&&a.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return fs(e,t,a,n)}function uT(e,t,n){if(hi(e),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");let a=ir(e,n);if(a.length!==5&&a.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return fs(e,t,a,n)}function pT(e,t,n){if(hi(e),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");let a=ir(e,n);if(a.length!==6&&a.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||a,fs(e,t,a,n)}function D3(e,t=1,n=!0){let a=_(e,"x","topk");if(a.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");let r=a.shape[a.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>r)throw new Error(`'k' passed to topk() must be <= the last dimension (${r}) but got ${t}`);let s={x:a},i={k:t,sorted:n},[o,l]=O.runKernel(vu,s,i);return{values:o,indices:l}}var Mv=L({topk_:D3});function R3(e,t=0,n=1,a,r){if(a!=null&&a==="bool")throw new Error("Unsupported data type $ { dtype }");let s=new Av(t,n,a,!0,r),i=Pe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var Qm=L({truncatedNormal_:R3});function M3(e,t=0){let n=_(e,"x","unique","string_or_numeric");$(n.rank>0,()=>"The input tensor must be at least 1D");let a={x:n},r={axis:t},[s,i]=O.runKernel(Nm,a,r);return{values:s,indices:i}}var Pv=L({unique_:M3});function P3(e,t,n){let a=_(e,"x","unsortedSegmentSum"),r=_(t,"segmentIds","unsortedSegmentSum","int32");$(tl(n),()=>"numSegments must be of dtype int");let s={x:a,segmentIds:r},i={numSegments:n};return O.runKernel(wc,s,i)}var ef=L({unsortedSegmentSum_:P3});function O3(e,t=0){let n=_(e,"x","unstack","string_or_numeric");$(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);let a={value:n},r={axis:t};return O.runKernel(ku,a,r)}var ct=L({unstack_:O3});function cT(e,t){return Vm(e,t,"right")}function Ov(e,t=!0,n,a){return O.makeVariable(e,t,n,a)}function dT(e,t){let n=[];for(let s=0;s<t.length;s++)t[s]&&n.push(s);let a=Pe(e,"int32"),r=Pe([n.length,e.length],"int32");for(let s=0;s<n.length;s++){let i=a.indexToLoc(n[s]),o=s*e.length;r.values.set(i,o)}return r.toTensor()}async function L3(e){let t=_(e,"condition","whereAsync","bool"),n=await t.data(),a=dT(t.shape,n);return e!==t&&t.dispose(),a}var Lv=L3;async function z3(e,t,n){let a=_(e,"tensor","boolMask"),r=_(t,"mask","boolMask","bool"),s=n==null?0:n,i=r.rank,o=a.shape;$(i>0,()=>"mask cannot be scalar"),Sn(o.slice(s,s+i),r.shape,"mask's shape must match the first K dimensions of tensor's shape,");let l=1;for(let f=s;f<s+i;f++)l*=o[f];let u=o.slice(0,s).concat([l],o.slice(s+i)),p=W(a,u),d=W(r,[-1]),c=await Lv(d),h=ws(c,[1]),m=Eu(p,h,s);return e!==a&&a.dispose(),t!==r&&r.dispose(),h.dispose(),p.dispose(),d.dispose(),c.dispose(),m}var hT=z3;function W3(e,t,n,a,r=!0){let s=_(e,"v","movingAverage"),i=_(t,"x","movingAverage"),o=_(n,"decay","movingAverage");BI(s,i),$(ps(s.shape,i.shape),()=>"Shape mismatch in v and x");let l=be(1),u=pe(l,o),p=z(pe(i,s),u);if(r){$(a!=null,()=>"When using zeroDebias: true, step is required.");let d=_(a,"step","movingAverage");p=he(p,pe(l,_r(o,d)))}return Y(s,p)}var mT=L({movingAverage_:W3});function B3(e,t,n){let a=_(e,"indices","scatterND","int32"),r=_(t,"updates","scatterND");Lx(r,a,n);let s={indices:a,updates:r},i={shape:n};return O.runKernel(uu,s,i)}var fT=L({scatterND_:B3});function V3(e,t,n,a){if(e.dtype!=="int32")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);let r=e.rank>0?e.shape[0]:1,s=e.rank>1?e.shape[1]:1;if(n.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${s}.`);let i=t.size;if(!(t.rank===0||t.rank===1&&i===r))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${r}]`);if(t.dtype!==a.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function U3(e,t,n,a=0){let r=_(e,"sparseIndices","sparseToDense","int32"),s=_(t,"sparseValues","sparseToDense","string_or_numeric"),i=_(a,"defaultValue","sparseToDense",s.dtype);V3(r,s,n,i);let o={sparseIndices:r,sparseValues:s,defaultValue:i},l={outputShape:n};return O.runKernel(Tm,o,l)}var gT=L({sparseToDense_:U3});function G3(e,t){let n=_(t,"indices","gatherND","int32"),a={params:_(e,"x","gatherND","string_or_numeric"),indices:n};return O.runKernel(Vl,a)}var yT=L({gatherND_:G3});function H3(e,t){if(t==null)return e.shape.slice();if(ps(e.shape,t))return t;if(e.shape.length===t.length){let n=[];for(let a=0;a<e.shape.length;a++)t[a]==null&&e.shape[a]!=null?n.push(e.shape[a]):n.push(t[a]);return n}return t}function j3(e,t,n,a){let r=_(e,"x","dropout");if($(r.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${r.dtype} tensor instead.`),$(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return e instanceof Te?r.clone():r;let s=H3(r,n),i=1-t,o=he(_u(Y($u(s,0,1,"float32",a),i)),i);return z(r,o)}var zv=L({dropout_:j3});function Wv(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function tf(e,t,n){let a=1-e%2,r=new Float32Array(e);for(let s=0;s<e;++s){let i=2*Math.PI*s/(e+a-1);r[s]=t-n*Math.cos(i)}return Ke(r,"float32")}async function q3(e,t,n=1){let a=_(e,"predictions","inTopK"),r=_(t,"targets","inTopK");$(a.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${a.rank}`),$(a.rank-1===r.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${a.rank} and targets rank ${r.rank}`),Sn(a.shape.slice(0,a.shape.length-1),r.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let s=a.shape[a.shape.length-1];$(n>0&&n<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${n}`);let i=await a.data(),o=await r.data(),[l,u]=[i.length/s,s],p=SI("bool",l);for(let d=0;d<l;d++){let c=d*u,h=i.subarray(c,c+u),m=[];for(let f=0;f<h.length;f++)m.push({value:h[f],index:f});m.sort((f,g)=>g.value-f.value),p[d]=0;for(let f=0;f<n;f++)if(m[f].index===o[d]){p[d]=1;break}}return e!==a&&a.dispose(),t!==r&&r.dispose(),kn(p,r.shape,"bool")}var bT=q3,pl={};Ae(pl,{conv2d:()=>Y3,depthwiseConv2d:()=>eL,matMul:()=>nL});function K3(e,t,n,a,r,s="NHWC",i){let o=e;e.rank===3&&(o=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]])),$(o.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`),$(l.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${l.shape}.`),$(n.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);let u=s==="NHWC"?o.shape[3]:o.shape[1],p=s==="NHWC"?l.shape[3]:l.shape[1];$(u===n[2],()=>`Error in conv2dDerFilter: depth of input ${u}) must match input depth in filter (${n[2]}.`),$(p===n[3],()=>`Error in conv2dDerFilter: depth of dy (${p}) must match output depth for filter (${n[3]}).`),Tn("conv2dDerFilter",r,i);let d={x:o,dy:l},c={strides:a,pad:r,dataFormat:s,dimRoundingMode:i,filterShape:n};return O.runKernel(em,d,c)}var Bv=L({conv2DBackpropFilter_:K3});function nf(e,t,n){if(n==null||n==="linear")return e;if(n==="relu")return z(e,fo(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function af(e,t){let n=t,a=Wt(e.shape,t.shape);return a.length>0&&(n=fe(n,a)),W(n,e.shape)}function rf(e,t,n,a){if(t==="linear")return e;if(t==="relu")return Xe(e);if(t==="elu")return Nu(e);if(t==="relu6")return Gm(e);if(t==="prelu")return Oc(e,n);if(t==="leakyrelu")return Fc(e,a);if(t==="sigmoid")return da(e);throw new Error(`Unknown fused activation ${t}.`)}var sf=(e,t)=>!(e>0)||t==="linear";function X3({x:e,filter:t,strides:n,pad:a,dataFormat:r="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:p}){if(l=l||"linear",sf(O.state.gradientDepth,l)===!1){$(r==="NHWC",()=>`Error in fused conv2d: got dataFormat of ${r} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let T=$t(e,t,n,a,r,s,i);return o!=null&&(T=Y(T,o)),rf(T,l,u,p)}let d=_(e,"x","conv2d","float32"),c=_(t,"filter","conv2d","float32"),h=d,m=!1;d.rank===3&&(m=!0,h=W(d,[1,d.shape[0],d.shape[1],d.shape[2]])),$(h.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${h.rank}.`),$(c.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${c.rank}.`),Tn("fused conv2d",a,i);let f=r==="NHWC"?h.shape[3]:h.shape[1];$(c.shape[2]===f,()=>`Error in conv2d: depth of input (${f}) must match input depth for filter ${c.shape[2]}.`),$(ur(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);let g=Ec(h.shape,c.shape,n,s,a,i),y;o!=null&&(y=_(o,"bias","fused conv2d"),[y]=_t(y,d),r==="NHWC"?lt(g.outShape,y.shape):($(y.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${y.shape.length}.`),$(y.shape.length===0||y.shape[0]===g.outChannels||y.shape[0]===1,()=>`Error in fused conv2d: bias shape (${y.shape}) is not compatible with the number of output channels (${g.outChannels})`)));let b;if(u!=null){let T=u.shape;if($(T.length<=1||T.length===3,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${T.length}.`),T.length===1)$(T[0]===1||T[0]===g.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${T}) is not compatible with the number of output channels (${g.outChannels}).`);else if(T.length===3)try{lt(T,g.outShape)}catch(C){let E=`Error in fused conv2d: PReLU activation weights (${T}) is not compatible with the output shape of the conv2d (${g.outShape}).`;throw Error(E)}b=_(u,"prelu weights","fused conv2d")}let x=(T,C)=>{$(r==="NHWC",()=>`Error in gradient of fused conv2D: got dataFormat of ${r} but only NHWC is currently supported.`);let[E,A,R,F]=C,S=nf(T,R,l);$(ns(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let M=iv(A.shape,S,E,n,a),B=Bv(A,S,E.shape,n,a),U=[M,B];if(F!=null){let G=af(F,S);U.push(G)}return U},w={x:h,filter:c,bias:y,preluActivationWeights:b},I={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:p};return o==null?or((T,C,E)=>{let A=O.runKernel(Qs,w,I);return E([C,T,A]),m&&(A=W(A,[A.shape[1],A.shape[2],A.shape[3]])),{value:A,gradFunc:x}})(h,c):or((T,C,E,A)=>{let R=O.runKernel(Qs,w,I);return A([C,T,R,E]),m&&(R=W(R,[R.shape[1],R.shape[2],R.shape[3]])),{value:R,gradFunc:x}})(h,c,y)}var Y3=L({fusedConv2d_:X3});function Z3(e,t,n,a,r,s=[1,1],i){let o=e;e.rank===3&&(o=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={x:o,dy:l},p={strides:a,pad:r,dimRoundingMode:i,dilations:s,filterShape:n};return O.runKernel(rm,u,p)}var xT=L({depthwiseConv2dNativeBackpropFilter_:Z3});function J3(e,t,n,a,r,s=[1,1],i){let o=t,l=!1;t.rank===3&&(l=!0,o=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={dy:o,filter:n},p={strides:a,pad:r,dimRoundingMode:i,dilations:s,inputShape:e},d=O.runKernel(sm,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var vT=L({depthwiseConv2dNativeBackpropInput_:J3});function Q3({x:e,filter:t,strides:n,pad:a,dataFormat:r="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:p}){if(sf(O.state.gradientDepth,l)===!1){let I=bs(e,t,n,a,r,s,i);return o!=null&&(I=Y(I,o)),rf(I,l,u,p)}let d=_(e,"x","depthwiseConv2d","float32"),c=_(t,"filter","depthwiseConv2d","float32"),h=d,m=!1;d.rank===3&&(m=!0,h=W(d,[1,d.shape[0],d.shape[1],d.shape[2]])),$(h.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${h.rank}.`),$(c.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${c.rank}.`),$(h.shape[3]===c.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${h.shape[3]}) must match the inChannels dimension in filter ${c.shape[2]}.`),s==null&&(s=[1,1]),$(ur(n,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),Tn("fused depthwiseConv2d",a,i);let f=Ec(h.shape,c.shape,n,s,a,i,!0),g;o!=null&&(g=_(o,"bias","fused conv2d"),[g]=_t(g,d),lt(f.outShape,g.shape));let y;u!=null&&(y=_(u,"prelu weights","fused depthwiseConv2d"));let b=(I,T)=>{$(ns(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[C,E,A,R]=T,F=nf(I,A,l),S=vT(E.shape,F,C,n,a,s,i),M=xT(E,F,C.shape,n,a,s,i);if(R!=null){let B=af(g,F);return[S,M,B]}return[S,M]},x={x:h,filter:c,bias:g,preluActivationWeights:y},w={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:p};return o==null?or((I,T,C)=>{let E=O.runKernel(ei,x,w);return C([T,I,E]),m&&(E=W(E,[E.shape[1],E.shape[2],E.shape[3]])),{value:E,gradFunc:b}})(h,c):or((I,T,C,E)=>{let A=O.runKernel(ei,x,w);return E([T,I,A,C]),m&&(A=W(A,[A.shape[1],A.shape[2],A.shape[3]])),{value:A,gradFunc:b}})(h,c,g)}var eL=L({fusedDepthwiseConv2d_:Q3});function tL({a:e,b:t,transposeA:n=!1,transposeB:a=!1,bias:r,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:o=.2}){if(sf(O.state.gradientDepth,s)===!1){let R=Fe(e,t,n,a);return r!=null&&(R=Y(R,r)),rf(R,s,i,o)}let l=_(e,"a","fused matMul"),u=_(t,"b","fused matMul");[l,u]=_t(l,u);let p=n?l.shape[l.rank-2]:l.shape[l.rank-1],d=a?u.shape[u.rank-1]:u.shape[u.rank-2],c=n?l.shape[l.rank-1]:l.shape[l.rank-2],h=a?u.shape[u.rank-2]:u.shape[u.rank-1],m=l.shape.slice(0,-2),f=u.shape.slice(0,-2),g=mt(m),y=mt(f);$(p===d,()=>`Error in fused matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${l.shape} and ${u.shape} and transposeA=${n} and transposeB=${a} must match.`);let b=lt(l.shape.slice(0,-2),u.shape.slice(0,-2)).concat([c,h]),x=n?W(l,[g,p,c]):W(l,[g,c,p]),w=a?W(u,[y,h,d]):W(u,[y,d,h]),I;r!=null&&(I=_(r,"bias","fused matMul"),[I]=_t(I,l),lt(b,I.shape));let T;i!=null&&(T=_(i,"prelu weights","fused matMul"));let C=(R,F)=>{let[S,M,B,U]=F,G=nf(W(R,B.shape),B,s),q,K;if(!n&&!a?(q=Fe(G,M,!1,!0),K=Fe(S,G,!0,!1)):!n&&a?(q=Fe(G,M,!1,!1),K=Fe(G,S,!0,!1)):n&&!a?(q=Fe(M,G,!1,!0),K=Fe(S,G,!1,!1)):(q=Fe(M,G,!0,!0),K=Fe(G,S,!0,!0)),r!=null){let Z=af(U,G);return[q,K,Z]}else return[q,K]},E={a:x,b:w,bias:I,preluActivationWeights:T},A={transposeA:n,transposeB:a,activation:s,leakyreluAlpha:o};return r==null?or((R,F,S)=>{let M=O.runKernel(Js,E,A);return S([R,F,M]),{value:W(M,b),gradFunc:C}})(x,w):or((R,F,S,M)=>{let B=O.runKernel(Js,E,A);return M([R,F,B,S]),{value:W(B,b),gradFunc:C}})(x,w,I)}var nL=L({fusedMatMul_:tL});function aL(e){return tf(e,.54,.46)}var rL=L({hammingWindow_:aL});function sL(e){return tf(e,.5,.5)}var wT=L({hannWindow_:sL});function iL(e,t,n,a=!1,r=0){let s=0,i=[];for(;s+t<=e.size;)i.push(Be(e,s,t)),s+=n;if(a)for(;s<e.size;){let o=s+t-e.size,l=Ze([Be(e,s,t-o),gn([o],r)]);i.push(l),s+=n}return i.length===0?_a([],[0,t]):W(Ze(i),[i.length,t])}var kT=L({frame_:iL});function oL(e,t,n,a,r=wT){a==null&&(a=Wv(t));let s=kT(e,t,n),i=z(s,r(t));return Wc(i,a)}var lL=L({stft_:oL});function uL(e,t,n,a,r="bilinear",s=0){let i=_(e,"image","cropAndResize"),o=_(t,"boxes","cropAndResize","float32"),l=_(n,"boxInd","cropAndResize","int32"),u=o.shape[0];$(i.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),$(o.rank===2&&o.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${u},4] but had shape ${o.shape}.`),$(l.rank===1&&l.shape[0]===u,()=>`Error in cropAndResize: boxInd must be have size [${u}] but had shape ${o.shape}.`),$(a.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${a.length}.`),$(a[0]>=1&&a[1]>=1,()=>`cropSize must be atleast [1,1], but was ${a}`),$(r==="bilinear"||r==="nearest",()=>`method must be bilinear or nearest, but was ${r}`);let p={image:i,boxes:o,boxInd:l},d={method:r,extrapolationValue:s,cropSize:a};return O.runKernel(Rl,p,d)}var pL=L({cropAndResize_:uL});function cL(e){let t=_(e,"image","flipLeftRight","float32");$(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let n={image:t};return O.runKernel(Wl,n,{})}var dL=L({flipLeftRight_:cL});function hL(e){let t=_(e,"image","grayscaleToRGB"),n=t.rank-1,a=t.shape[n];$(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),$(a===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${a}.`);let r=new Array(t.rank);return r.fill(1,0,n),r[n]=3,Ln(t,r)}var mL=L({grayscaleToRGB_:hL});function fL(e,t,n=0,a=.5){let r=_(e,"image","rotateWithOffset","float32");$(r.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${r.rank}.`);let s={image:r},i={radians:t,fillValue:n,center:a};return O.runKernel(Su,s,i)}var gL=L({rotateWithOffset_:fL});function Fu(e,t,n,a,r,s){a==null&&(a=.5),r==null&&(r=Number.NEGATIVE_INFINITY),s==null&&(s=0);let i=e.shape[0];return n=Math.min(n,i),$(0<=a&&a<=1,()=>`iouThreshold must be in [0, 1], but was '${a}'`),$(e.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),$(e.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),$(t.rank===1,()=>"scores must be a 1D tensor"),$(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),$(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:n,iouThreshold:a,scoreThreshold:r,softNmsSigma:s}}function yL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY){let s=_(e,"boxes","nonMaxSuppression","float32"),i=_(t,"scores","nonMaxSuppression","float32"),o=Fu(s,i,n,a,r);n=o.maxOutputSize,a=o.iouThreshold,r=o.scoreThreshold;let l={maxOutputSize:n,iouThreshold:a,scoreThreshold:r};return O.runKernel(nu,{boxes:s,scores:i},l)}var bL=L({nonMaxSuppression_:yL});function xL(e,t,n){let a=vL(e,t,n),r=a<0?-(a+1):a;e.splice(r,0,t)}function vL(e,t,n){return kL(e,t,n||wL)}function wL(e,t){return e>t?1:e<t?-1:0}function kL(e,t,n){let a=0,r=e.length,s=0,i=!1;for(;a<r;){s=a+(r-a>>>1);let o=n(t,e[s]);o>0?a=s+1:(r=s,i=!o)}return i?a:-a-1}function IT(e,t,n,a,r){return Vv(e,t,n,a,r,0)}function ST(e,t,n,a,r,s){return Vv(e,t,n,a,r,0,!1,s,!0)}function TT(e,t,n,a,r,s){return Vv(e,t,n,a,r,s,!0)}function Vv(e,t,n,a,r,s,i=!1,o=!1,l=!1){let u=[];for(let g=0;g<t.length;g++)t[g]>r&&u.push({score:t[g],boxIndex:g,suppressBeginIndex:0});u.sort(Z1);let p=s>0?-.5/s:0,d=[],c=[];for(;d.length<n&&u.length>0;){let g=u.pop(),{score:y,boxIndex:b,suppressBeginIndex:x}=g;if(y<r)break;let w=!1;for(let I=d.length-1;I>=x;--I){let T=IL(e,b,d[I]);if(T>=a){w=!0;break}if(g.score=g.score*SL(a,p,T),g.score<=r)break}g.suppressBeginIndex=d.length,w||(g.score===y?(d.push(b),c.push(g.score)):g.score>r&&xL(u,g,Z1))}let h=d.length,m=n-h;o&&m>0&&(d.push(...new Array(m).fill(0)),c.push(...new Array(m).fill(0)));let f={selectedIndices:d};return i&&(f.selectedScores=c),l&&(f.validOutputs=h),f}function IL(e,t,n){let a=e.subarray(t*4,t*4+4),r=e.subarray(n*4,n*4+4),s=Math.min(a[0],a[2]),i=Math.min(a[1],a[3]),o=Math.max(a[0],a[2]),l=Math.max(a[1],a[3]),u=Math.min(r[0],r[2]),p=Math.min(r[1],r[3]),d=Math.max(r[0],r[2]),c=Math.max(r[1],r[3]),h=(o-s)*(l-i),m=(d-u)*(c-p);if(h<=0||m<=0)return 0;let f=Math.max(s,u),g=Math.max(i,p),y=Math.min(o,d),b=Math.min(l,c),x=Math.max(y-f,0)*Math.max(b-g,0);return x/(h+m-x)}function SL(e,t,n){let a=Math.exp(t*n*n);return n<=e?a:0}function Z1(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}async function TL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY){let s=_(e,"boxes","nonMaxSuppressionAsync"),i=_(t,"scores","nonMaxSuppressionAsync"),o=Fu(s,i,n,a,r);n=o.maxOutputSize,a=o.iouThreshold,r=o.scoreThreshold;let l=await Promise.all([s.data(),i.data()]),u=l[0],p=l[1],{selectedIndices:d}=IT(u,p,n,a,r);return s!==e&&s.dispose(),i!==t&&i.dispose(),Ke(d,"int32")}var NL=TL;function CL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=0){let i=_(e,"boxes","nonMaxSuppression"),o=_(t,"scores","nonMaxSuppression"),l=Fu(i,o,n,a,r,s);n=l.maxOutputSize,a=l.iouThreshold,r=l.scoreThreshold,s=l.softNmsSigma;let u={boxes:i,scores:o},p={maxOutputSize:n,iouThreshold:a,scoreThreshold:r,softNmsSigma:s},d=O.runKernel(ru,u,p);return{selectedIndices:d[0],selectedScores:d[1]}}var _L=L({nonMaxSuppressionWithScore_:CL});async function EL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=0){let i=_(e,"boxes","nonMaxSuppressionAsync"),o=_(t,"scores","nonMaxSuppressionAsync"),l=Fu(i,o,n,a,r,s);n=l.maxOutputSize,a=l.iouThreshold,r=l.scoreThreshold,s=l.softNmsSigma;let u=await Promise.all([i.data(),o.data()]),p=u[0],d=u[1],{selectedIndices:c,selectedScores:h}=TT(p,d,n,a,r,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Ke(c,"int32"),selectedScores:Ke(h)}}var AL=EL;function $L(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=!1){let i=_(e,"boxes","nonMaxSuppression"),o=_(t,"scores","nonMaxSuppression"),l=Fu(i,o,n,a,r,null),u=l.maxOutputSize,p=l.iouThreshold,d=l.scoreThreshold,c={boxes:i,scores:o},h={maxOutputSize:u,iouThreshold:p,scoreThreshold:d,padToMaxOutputSize:s},m=O.runKernel(au,c,h);return{selectedIndices:m[0],validOutputs:m[1]}}var FL=L({nonMaxSuppressionPadded_:$L});async function DL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=!1){let i=_(e,"boxes","nonMaxSuppressionAsync"),o=_(t,"scores","nonMaxSuppressionAsync"),l=Fu(i,o,n,a,r,null),u=l.maxOutputSize,p=l.iouThreshold,d=l.scoreThreshold,[c,h]=await Promise.all([i.data(),o.data()]),{selectedIndices:m,validOutputs:f}=ST(c,h,u,p,d,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Ke(m,"int32"),validOutputs:be(f,"int32")}}var RL=DL;function ML(e,t,n=!1,a=!1){let r=_(e,"images","resizeBilinear");$(r.rank===3||r.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${r.rank}.`),$(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),$(a===!1||n===!1,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=r,i=!1;r.rank===3&&(i=!0,s=W(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:a,size:t},u=O.runKernel(Zi,o,l);return i?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var NT=L({resizeBilinear_:ML});function PL(e,t,n=!1,a=!1){let r=_(e,"images","resizeNearestNeighbor");$(r.rank===3||r.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${r.rank}.`),$(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),$(r.dtype==="float32"||r.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype"),$(a===!1||n===!1,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=r,i=!1;r.rank===3&&(i=!0,s=W(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:a,size:t},u=O.runKernel(Yi,o,l);return i?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var CT=L({resizeNearestNeighbor_:PL});function OL(e,t="binary",n=!1,a=.5){let r=_(e,"image","threshold"),s=.2989,i=.587,o=.114,l=r.shape[0]*r.shape[1],u=z(Ke([a]),255),p,d,c,h;if($(r.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${r.rank}.`),$(r.shape[2]===3||r.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${r.shape[2]}.`),$(r.dtype==="int32"||r.dtype==="float32",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${r.dtype}.`),$(t==="otsu"||t==="binary",()=>`Method must be binary or otsu, but was ${t}`),r.shape[2]===3){[p,d,c]=zn(r,[1,1,1],-1);let f=z(p,s),g=z(d,i),y=z(c,o);h=Y(Y(f,g),y)}else h=e;if(t==="otsu"){let f=ev(oe(Hm(h),"int32"),kn([]),256);u=LL(f,l)}let m=n?xs(h,u):Nn(h,u);return oe(z(m,255),"int32")}function LL(e,t){let n=Ke([-1]),a=Ke([0]),r=Ke([0]),s,i,o,l,u,p;for(let d=0;d<e.size-1;d++){s=Be(e,0,d+1),i=Be(e,d+1),u=he(fe(s),t),p=he(fe(i),t);let c=fe(z(s,ol(0,s.size)));o=he(c,fe(s));let h=gn(i.shape,s.size),m=Y(ol(0,i.size),h),f=z(i,m);l=he(fe(f),fe(i));let g=pe(o,l),y=pe(o,l),b=z(u,p);r=z(z(b,g),y);let x=Nn(r,a);a=mn(x,r,a),n=mn(x,Ke([d]),n)}return n}var zL=L({threshold_:OL});function WL(e,t,n="nearest",a="constant",r=0,s){let i=_(e,"image","transform","float32"),o=_(t,"transforms","transform","float32");$(i.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),$(o.rank===2&&(o.shape[0]===i.shape[0]||o.shape[0]===1)&&o.shape[1]===8,()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),$(s==null||s.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let l={image:i,transforms:o},u={interpolation:n,fillMode:a,fillValue:r,outputShape:s};return O.runKernel(wu,l,u)}var BL=L({transform_:WL});function VL(e,t,n){$(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),$(n%1===0,()=>`bandPart(): numUpper must be an integer, got ${n}.`);let a=_(e,"a","bandPart");$(a.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${a.rank}.`);let r=a.shape,[s,i]=a.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(n<=i))throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),n<0&&(n=i);let o=W(ol(0,s,1,"int32"),[-1,1]),l=ol(0,i,1,"int32"),u=pe(o,l),p=Ea(xs(u,be(+t,"int32")),Er(u,be(-n,"int32"))),d=It([s,i],a.dtype);return W(Ft(ct(W(a,[-1,s,i])).map(c=>mn(p,c,d))),r)}var UL=L({bandPart_:VL});function GL(e){let t;if(Array.isArray(e)){t=!1,$(e!=null&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let r=e[0].shape[0];for(let s=1;s<e.length;++s)$(e[s].shape[0]===r,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[s].shape[0]} vs. ${r})`)}else t=!0,e=zn(e,e.shape[0],0).map(r=>ws(r,[0]));$(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);let n=[],a=e;for(let r=0;r<e.length;++r)n.push(O.tidy(()=>{let s=a[r];if(r>0)for(let i=0;i<r;++i){let o=z(fe(z(n[i],s)),n[i]);s=pe(s,o)}return he(s,Cu(s,"euclidean"))}));return t?Ft(n,0):n}var HL=L({gramSchmidt_:GL});function jL(e,t=!1){if($(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),e.rank===2)return J1(e,t);{let n=e.shape.slice(0,e.shape.length-2).reduce((l,u)=>l*u),a=ct(W(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),r=[],s=[];a.forEach(l=>{let[u,p]=J1(l,t);r.push(u),s.push(p)});let i=W(Ft(r,0),e.shape),o=W(Ft(s,0),e.shape);return[i,o]}}function J1(e,t=!1){return O.tidy(()=>{$(e.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);let n=e.shape[0],a=e.shape[1],r=Pm(n),s=rr(e),i=_a([[1]],[1,1]),o=rr(i),l=n>=a?a:n;for(let u=0;u<l;++u){let p=s,d=o,c=r;[o,s,r]=O.tidy(()=>{let h=Be(s,[u,u],[n-u,1]),m=Cu(h),f=Be(s,[u,u],[1,1]),g=mn(Nn(f,0),_a([[-1]]),_a([[1]])),y=pe(f,z(g,m)),b=he(h,y);b.shape[0]===1?o=rr(i):o=Ze([i,Be(b,[1,0],[b.shape[0]-1,b.shape[1]])],0);let x=bt(he(Fe(g,y),m)),w=Be(s,[u,0],[n-u,a]),I=z(x,o),T=Ee(o);if(u===0)s=pe(w,Fe(I,Fe(T,w)));else{let A=pe(w,Fe(I,Fe(T,w)));s=Ze([Be(s,[0,0],[u,a]),A],0)}let C=Ee(I),E=Be(r,[0,u],[n,r.shape[1]-u]);if(u===0)r=pe(E,Fe(Fe(E,o),C));else{let A=pe(E,Fe(Fe(E,o),C));r=Ze([Be(r,[0,0],[n,u]),A],1)}return[o,s,r]}),_e([p,d,c])}return!t&&n>a&&(r=Be(r,[0,0],[n,a]),s=Be(s,[0,0],[a,a])),[r,s]})}var qL=L({qr_:jL}),xn;(function(e){e[e.NONE=0]="NONE",e[e.MEAN=1]="MEAN",e[e.SUM=2]="SUM",e[e.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS"})(xn||(xn={}));function KL(e,t,n=xn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"losses","computeWeightedLoss"),r=null;t!=null&&(r=_(t,"weights","computeWeightedLoss"));let s=r==null?a:z(a,r);if(n===xn.NONE)return s;if(n===xn.SUM)return fe(s);if(n===xn.MEAN){if(r==null)return Nt(s);{let i=a.size/r.size,o=he(fe(s),fe(r));return i>1?he(o,be(i)):o}}if(n===xn.SUM_BY_NONZERO_WEIGHTS){if(r==null)return he(fe(s),be(a.size));{let i=z(r,Zn(a.shape)),o=oe(fe(ii(i,be(0))),"float32");return he(fe(s),o)}}throw Error(`Unknown reduction: ${n}`)}var Ar=L({computeWeightedLoss_:KL});function XL(e,t,n,a=xn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","absoluteDifference"),s=_(t,"predictions","absoluteDifference"),i=null;n!=null&&(i=_(n,"weights","absoluteDifference")),Sn(r.shape,s.shape,"Error in absoluteDifference: ");let o=Lt(pe(r,s));return Ar(o,i,a)}var YL=L({absoluteDifference_:XL});function ZL(e,t,n,a,r=xn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","cosineDistance"),i=_(t,"predictions","cosineDistance"),o=null;a!=null&&(o=_(a,"weights","cosineDistance")),Sn(s.shape,i.shape,"Error in cosineDistance: ");let l=be(1),u=pe(l,fe(z(s,i),n,!0));return Ar(u,o,r)}var JL=L({cosineDistance_:ZL});function QL(e,t,n,a=xn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","hingeLoss"),s=_(t,"predictions","hingeLoss"),i=null;n!=null&&(i=_(n,"weights","hingeLoss")),Sn(r.shape,s.shape,"Error in hingeLoss: ");let o=be(1);r=pe(z(be(2),r),o);let l=Xe(pe(o,z(r,s)));return Ar(l,i,a)}var ez=L({hingeLoss_:QL});function tz(e,t,n,a=1,r=xn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","huberLoss"),i=_(t,"predictions","huberLoss"),o=null;n!=null&&(o=_(n,"weights","huberLoss")),Sn(s.shape,i.shape,"Error in huberLoss: ");let l=be(a),u=Lt(pe(i,s)),p=Au(u,l),d=pe(u,p),c=Y(z(be(.5),ot(p)),z(l,d));return Ar(c,o,r)}var nz=L({huberLoss_:tz});function az(e,t,n,a=1e-7,r=xn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","logLoss"),i=_(t,"predictions","logLoss"),o=null;n!=null&&(o=_(n,"weights","logLoss")),Sn(s.shape,i.shape,"Error in logLoss: ");let l=be(1),u=be(a),p=bt(z(s,ea(Y(i,u)))),d=z(pe(l,s),ea(Y(pe(l,i),u))),c=pe(p,d);return Ar(c,o,r)}var rz=L({logLoss_:az});function sz(e,t,n,a=xn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","meanSquaredError"),s=_(t,"predictions","meanSquaredError"),i=null;n!=null&&(i=_(n,"weights","meanSquaredError")),Sn(r.shape,s.shape,"Error in meanSquaredError: ");let o=Jm(r,s);return Ar(o,i,a)}var iz=L({meanSquaredError_:sz});function oz(e,t){let n=_(e,"labels","sigmoidCrossEntropyWithLogits"),a=_(t,"logits","sigmoidCrossEntropyWithLogits");Sn(n.shape,a.shape,"Error in sigmoidCrossEntropyWithLogits: ");let r=Xe(a),s=z(a,n),i=Dc(fn(bt(Lt(a))));return Y(pe(r,s),i)}function lz(e,t,n,a=0,r=xn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"multiClassLabels","sigmoidCrossEntropy"),i=_(t,"logits","sigmoidCrossEntropy"),o=null;if(n!=null&&(o=_(n,"weights","sigmoidCrossEntropy")),Sn(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),a>0){let u=be(a),p=be(1),d=be(.5);s=Y(z(s,pe(p,u)),z(d,u))}let l=oz(s,i);return Ar(l,o,r)}var uz=L({sigmoidCrossEntropy_:lz});function pz(e,t,n=-1){if(n===-1&&(n=t.rank-1),n!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${n}`);return or((a,r,s)=>{let i=Wm(r,[n],!0),o=pe(oe(r,"float32"),i);s([a,o]);let l=bt(z(o,a));return{value:fe(l,[n]),gradFunc:(u,p)=>{let[d,c]=p,h=si(u.shape,[n]);return[z(W(u,h),pe(oe(d,"float32"),fn(c))),z(W(u,h),pe(fn(c),oe(d,"float32")))]}}})(e,t)}function cz(e,t,n,a=0,r=xn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"onehotLabels","softmaxCrossEntropy"),i=_(t,"logits","softmaxCrossEntropy"),o=null;if(n!=null&&(o=_(n,"weights","softmaxCrossEntropy")),Sn(s.shape,i.shape,"Error in softmaxCrossEntropy: "),a>0){let u=be(a),p=be(1),d=be(s.shape[1]);s=Y(z(s,pe(p,u)),he(u,d))}let l=pz(s,i);return Ar(l,o,r)}var dz=L({softmaxCrossEntropy_:cz});function hz(e,t,n,a){let r=_(e,"indices","sparseFillEmptyRows","int32"),s=_(t,"values","sparseFillEmptyRows"),i=_(n,"denseShape","sparseFillEmptyRows","int32"),o=_(a,"defaultValue","sparseFillEmptyRows",s.dtype);if(r.rank!==2)throw new Error(`Indices should be Tensor2D but received shape
-        ${r.shape}`);if(s.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(i.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(o.rank!==0)throw new Error(`Default value should be a scalar but received shape ${o.shape}`);let l={indices:r,values:s,denseShape:i,defaultValue:o},u=O.runKernel(mc,l);return{outputIndices:u[0],outputValues:u[1],emptyRowIndicator:u[2],reverseIndexMap:u[3]}}var mz=L({sparseFillEmptyRows_:hz});function fz(e,t,n){let a=_(e,"inputIndices","sparseReshape","int32"),r=_(t,"inputShape","sparseReshape","int32"),s=_(n,"newShape","sparseReshape","int32");if(a.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape
-        ${a.shape}`);if(r.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${r.shape}`);if(s.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:a,inputShape:r,newShape:s},o=O.runKernel(bu,i);return{outputIndices:o[0],outputShape:o[1]}}var gz=L({sparseReshape_:fz});function yz(e,t,n){let a=_(e,"data","sparseSegmentMean"),r=_(t,"indices","sparseSegmentMean","int32"),s=_(n,"segmentIds","sparseSegmentMean","int32");if(a.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-          ${r.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-          ${s.shape}`);let i={data:a,indices:r,segmentIds:s};return O.runKernel(fc,i)}var bz=L({sparseSegmentMean_:yz});function xz(e,t,n){let a=_(e,"data","sparseSegmentSum"),r=_(t,"indices","sparseSegmentSum","int32"),s=_(n,"segmentIds","sparseSegmentSum","int32");if(a.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-         ${r.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-         ${s.shape}`);let i={data:a,indices:r,segmentIds:s};return O.runKernel(gc,i)}var vz=L({sparseSegmentSum_:xz});function wz(e,t,n,a,r,s,i,o){let l=_(e,"data","stringNGrams","string");if(l.dtype!=="string")throw new Error("Data must be of datatype string");if(l.shape.length!==1)throw new Error(`Data must be a vector, saw: ${l.shape}`);let u=_(t,"dataSplits","stringNGrams");if(u.dtype!=="int32")throw new Error("Data splits must be of datatype int32");let p={separator:n,nGramWidths:a,leftPad:r,rightPad:s,padWidth:i,preserveShortSequences:o},d={data:l,dataSplits:u},c=O.runKernel(bc,d,p);return{nGrams:c[0],nGramsSplits:c[1]}}var kz=L({stringNGrams_:wz});function Iz(e,t,n=!0){let a=_(e,"input","stringSplit","string"),r=_(t,"delimiter","stringSplit","string");if(a.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${a.shape}`);if(r.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${r.shape}`);let s={skipEmpty:n},i={input:a,delimiter:r},o=O.runKernel(xc,i,s);return{indices:o[0],values:o[1],shape:o[2]}}var Sz=L({stringSplit_:Iz});function Tz(e,t){let n=_(e,"input","stringToHashBucketFast","string"),a={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let r={input:n};return O.runKernel(vc,r,a)}var Nz=L({stringToHashBucketFast_:Tz}),_T={fft:zc,ifft:ul,rfft:Wc,irfft:Zm},ET={hammingWindow:rL,hannWindow:wT,frame:kT,stft:lL},za={flipLeftRight:dL,grayscaleToRGB:mL,resizeNearestNeighbor:CT,resizeBilinear:NT,rotateWithOffset:gL,cropAndResize:pL,nonMaxSuppression:bL,nonMaxSuppressionAsync:NL,nonMaxSuppressionWithScore:_L,nonMaxSuppressionWithScoreAsync:AL,nonMaxSuppressionPadded:FL,nonMaxSuppressionPaddedAsync:RL,threshold:zL,transform:BL},Uv={bandPart:UL,gramSchmidt:HL,qr:qL},AT={absoluteDifference:YL,computeWeightedLoss:Ar,cosineDistance:JL,hingeLoss:ez,huberLoss:nz,logLoss:rz,meanSquaredError:iz,sigmoidCrossEntropy:uz,softmaxCrossEntropy:dz},$T={sparseFillEmptyRows:mz,sparseReshape:gz,sparseSegmentMean:bz,sparseSegmentSum:vz},FT={stringNGrams:kz,stringSplit:Sz,stringToHashBucketFast:Nz},$r=class extends IS{minimize(e,t=!1,n){let{value:a,grads:r}=this.computeGradients(e,n);if(n!=null){let s=n.map(i=>({name:i.name,tensor:r[i.name]}));this.applyGradients(s)}else this.applyGradients(r);return _e(r),t?a:(a.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return BS(e,t)}dispose(){this.iterations_!=null&&_e(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:be(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(e){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty($r,Symbol.hasInstance,{value:e=>e.minimize!=null&&e.computeGradients!=null&&e.applyGradients!=null});var of=class extends $r{constructor(e,t,n=null){super(),this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t],r=!1;this.accumulatedGrads[n]==null&&(this.accumulatedGrads[n]={originalName:`${t}/accum_grad`,variable:P(()=>qe(a).variable(r))}),this.accumulatedUpdates[n]==null&&(this.accumulatedUpdates[n]={originalName:`${t}/accum_var`,variable:P(()=>qe(a).variable(r))});let s=Array.isArray(e)?e[n].tensor:e[t];if(s==null)return;let i=this.accumulatedGrads[n].variable,o=this.accumulatedUpdates[n].variable;P(()=>{let l=Y(z(i,this.rho),z(ot(s),1-this.rho)),u=z(he(ln(Y(o,this.epsilon)),ln(Y(i,this.epsilon))),s),p=Y(z(o,this.rho),z(ot(u),1-this.rho));i.assign(l),o.assign(p);let d=Y(z(u,-this.learningRate),a);a.assign(d)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(_e(this.accumulatedGrads.map(e=>e.variable)),_e(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){let e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=e.length/2,n=!1;this.accumulatedGrads=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedUpdates=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}};of.className="Adadelta";gs(of);var lf=class extends $r{constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t];this.accumulatedGrads[n]==null&&(this.accumulatedGrads[n]={originalName:`${t}/accumulator`,variable:P(()=>gn(a.shape,this.initialAccumulatorValue).variable(!1))});let r=Array.isArray(e)?e[n].tensor:e[t];if(r==null)return;let s=this.accumulatedGrads[n].variable;P(()=>{let i=Y(s,ot(r));s.assign(i);let o=Y(z(he(r,ln(Y(i,O.backend.epsilon()))),-this.learningRate),a);a.assign(o)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&_e(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulatedGrads=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}};lf.className="Adagrad";gs(lf);var uf=class extends $r{constructor(e,t,n,a=null){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=a,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],P(()=>{this.accBeta1=be(t).variable(),this.accBeta2=be(n).variable()}),a==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);P(()=>{let n=pe(1,this.accBeta1),a=pe(1,this.accBeta2);t.forEach((r,s)=>{let i=O.registeredVariables[r],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${r}/m`,variable:P(()=>qe(i).variable(o))}),this.accumulatedSecondMoment[s]==null&&(this.accumulatedSecondMoment[s]={originalName:`${r}/v`,variable:P(()=>qe(i).variable(o))});let l=Array.isArray(e)?e[s].tensor:e[r];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,p=this.accumulatedSecondMoment[s].variable,d=Y(z(u,this.beta1),z(l,1-this.beta1)),c=Y(z(p,this.beta2),z(ot(l),1-this.beta2)),h=he(d,n),m=he(c,a);u.assign(d),p.assign(c);let f=Y(z(he(h,Y(ln(m),this.epsilon)),-this.learningRate),i);i.assign(f)}),this.accBeta1.assign(z(this.accBeta1,this.beta1)),this.accBeta2.assign(z(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedSecondMoment!=null&&_e(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e),P(()=>{this.accBeta1.assign(_r(this.beta1,this.iterations_+1)),this.accBeta2.assign(_r(this.beta2,this.iterations_+1))});let t=e.length/2,n=!1;this.accumulatedFirstMoment=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedSecondMoment=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon)}};uf.className="Adam";gs(uf);var pf=class extends $r{constructor(e,t,n,a=null,r=0){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=a,this.decay=r,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],P(()=>{this.iteration=be(0).variable(),this.accBeta1=be(t).variable()}),a==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);P(()=>{let n=pe(1,this.accBeta1),a=he(-this.learningRate,Y(z(this.iteration,this.decay),1));t.forEach((r,s)=>{let i=O.registeredVariables[r],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${r}/m`,variable:qe(i).variable(o)}),this.accumulatedWeightedInfNorm[s]==null&&(this.accumulatedWeightedInfNorm[s]={originalName:`${r}/v`,variable:qe(i).variable(o)});let l=Array.isArray(e)?e[s].tensor:e[r];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,p=this.accumulatedWeightedInfNorm[s].variable,d=Y(z(u,this.beta1),z(l,1-this.beta1)),c=z(p,this.beta2),h=Lt(l),m=pr(c,h);u.assign(d),p.assign(m);let f=Y(z(he(a,n),he(d,Y(m,this.epsilon))),i);i.assign(f)}),this.iteration.assign(Y(this.iteration,1)),this.accBeta1.assign(z(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedWeightedInfNorm!=null&&_e(this.accumulatedWeightedInfNorm.map(e=>e.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon,t.decay)}};pf.className="Adamax";gs(pf);var Bc=class extends $r{constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=Array.isArray(e)?e[n].tensor:e[t];if(a==null)return;let r=O.registeredVariables[t];P(()=>{let s=Y(z(this.c,a),r);r.assign(s)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,this.c!=null&&this.c.dispose(),this.c=Jt(be(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(e=await this.extractIterations(e),e.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}};Bc.className="SGD";gs(Bc);var cf=class extends Bc{constructor(e,t,n=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=be(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t];this.accumulations[n]==null&&(this.accumulations[n]={originalName:`${t}/momentum`,variable:P(()=>qe(a).variable(!1))});let r=this.accumulations[n].variable,s=Array.isArray(e)?e[n].tensor:e[t];s!=null&&P(()=>{let i,o=Y(z(this.m,r),s);this.useNesterov?i=Y(z(this.c,Y(s,z(o,this.m))),a):i=Y(z(this.c,o),a),r.assign(o),a.assign(i)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&_e(this.accumulations.map(e=>e.variable))}setMomentum(e){this.momentum=e}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulations=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}};cf.className="Momentum";gs(cf);var df=class extends $r{constructor(e,t=.9,n=0,a=null,r=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=a,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=r,a==null&&(this.epsilon=O.backend.epsilon()),e==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t],r=!1;this.accumulatedMeanSquares[n]==null&&(this.accumulatedMeanSquares[n]={originalName:`${t}/rms`,variable:P(()=>qe(a).variable(r))}),this.accumulatedMoments[n]==null&&(this.accumulatedMoments[n]={originalName:`${t}/momentum`,variable:P(()=>qe(a).variable(r))}),this.accumulatedMeanGrads[n]==null&&this.centered&&(this.accumulatedMeanGrads[n]={originalName:`${t}/mg`,variable:P(()=>qe(a).variable(r))});let s=Array.isArray(e)?e[n].tensor:e[t];if(s==null)return;let i=this.accumulatedMeanSquares[n].variable,o=this.accumulatedMoments[n].variable;P(()=>{let l=Y(z(i,this.decay),z(ot(s),1-this.decay));if(this.centered){let u=this.accumulatedMeanGrads[n].variable,p=Y(z(u,this.decay),z(s,1-this.decay)),d=he(z(s,this.learningRate),ln(pe(l,Y(ot(p),this.epsilon)))),c=Y(z(o,this.momentum),d);i.assign(l),u.assign(p),o.assign(c);let h=pe(a,c);a.assign(h)}else{let u=Y(z(i,this.decay),z(ot(s),1-this.decay)),p=Y(z(o,this.momentum),he(z(s,this.learningRate),ln(Y(u,this.epsilon))));i.assign(u),o.assign(p);let d=pe(a,p);a.assign(d)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&_e(this.accumulatedMeanSquares.map(e=>e.variable)),this.accumulatedMeanGrads!=null&&this.centered&&_e(this.accumulatedMeanGrads.map(e=>e.variable)),this.accumulatedMoments!=null&&_e(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=this.centered?e.length/3:e.length/2,n=!1;this.accumulatedMeanSquares=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedMoments=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=e.slice(t*2,t*3).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}};df.className="RMSProp";gs(df);var Hr=class{static sgd(e){return new Bc(e)}static momentum(e,t,n=!1){return new cf(e,t,n)}static rmsprop(e,t=.9,n=0,a=null,r=!1){return new df(e,t,n,a,r)}static adam(e=.001,t=.9,n=.999,a=null){return new uf(e,t,n,a)}static adadelta(e=.001,t=.95,n=null){return new of(e,t,n)}static adamax(e=.002,t=.9,n=.999,a=null,r=0){return new pf(e,t,n,a,r)}static adagrad(e,t=.1){return new lf(e,t)}},zs={sgd:Hr.sgd,momentum:Hr.momentum,adadelta:Hr.adadelta,adagrad:Hr.adagrad,rmsprop:Hr.rmsprop,adamax:Hr.adamax,adam:Hr.adam},Cz=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:e=>e())();function Gv(){return new Promise(e=>Cz(()=>e()))}var N={};Ae(N,{ERF_A1:()=>Vz,ERF_A2:()=>Uz,ERF_A3:()=>Gz,ERF_A4:()=>Hz,ERF_A5:()=>jz,ERF_P:()=>Bz,PARALLELIZE_THRESHOLD:()=>Hv,RowPartitionType:()=>er,SELU_SCALE:()=>RT,SELU_SCALEALPHA:()=>DT,applyActivation:()=>rf,assertAndGetBroadcastShape:()=>lt,assertAxesAreInnerMostDims:()=>FP,assertParamsConsistent:()=>_z,assignToTypedArray:()=>Jz,axesAreInnerMostDims:()=>mv,calculateShapes:()=>hS,checkEinsumDimSizes:()=>rW,checkPadOnDimRoundingMode:()=>Tn,combineLocations:()=>PS,combineRaggedTensorToTensorShapes:()=>Az,complexWithEvenIndex:()=>Xz,complexWithOddIndex:()=>Yz,computeConv2DInfo:()=>Ec,computeConv3DInfo:()=>ES,computeDefaultPad:()=>Xx,computeDilation2DInfo:()=>DM,computeOptimalWindowSize:()=>Rz,computeOutAndReduceShapes:()=>OS,computeOutShape:()=>Ez,computePool2DInfo:()=>_S,computePool3DInfo:()=>RM,convertConv2DDataFormat:()=>AS,decodeEinsumEquation:()=>nW,eitherStridesOrDilationsAreOne:()=>ur,expandShapeToKeepDim:()=>si,exponent:()=>eW,exponents:()=>Qz,fromStringArrayToUint8:()=>TW,fromUint8ToStringArray:()=>SW,getAxesPermutation:()=>LS,getBroadcastDims:()=>pS,getComplexWithIndex:()=>Zz,getEinsumComputePath:()=>sW,getEinsumPermutation:()=>aW,getFusedBiasGradient:()=>af,getFusedDyActivation:()=>nf,getImageCenter:()=>Mz,getInnerMostAxes:()=>DP,getPermuted:()=>Oz,getRaggedRank:()=>Fz,getReductionAxes:()=>Wt,getReshaped:()=>Pz,getReshapedPermuted:()=>Lz,getRowPartitionTypesHelper:()=>$z,getSliceBeginCoords:()=>zz,getSliceSize:()=>Wz,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>uW,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>pW,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>cW,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>mW,getSparseReshapeInputOutputMismatchErrorMessage:()=>gW,getSparseReshapeInputOutputMultipleErrorMessage:()=>fW,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>dW,getSparseReshapeNegativeOutputDimErrorMessage:()=>hW,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>vW,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>yW,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>bW,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>xW,getUndoAxesPermutation:()=>fv,isIdentityPermutation:()=>iW,log:()=>YF,mergeRealAndImagArrays:()=>qz,prepareAndValidate:()=>dS,prepareSplitSize:()=>lW,segment_util:()=>MT,shouldFuse:()=>sf,slice_util:()=>jt,splitRealAndImagArrays:()=>Kz,tupleValuesAreOne:()=>ns,upcastType:()=>ma,validateDefaultValueShape:()=>Dz,validateInput:()=>Lx,validateUpdateShape:()=>Ox,warn:()=>qr});function _z(e,t){let n=e[0].length;e.forEach((r,s)=>{$(r.length===n,()=>`Error in concat${n}D: rank of tensors[${s}] must be the same as the rank of the rest (${n})`)}),$(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);let a=e[0];e.forEach((r,s)=>{for(let i=0;i<n;i++)$(i===t||r[i]===a[i],()=>`Error in concat${n}D: Shape of tensors[${s}] (${r}) does not match the shape of the rest (${a}) along the non-concatenated axis ${s}.`)})}function Ez(e,t){let n=e[0].slice();for(let a=1;a<e.length;a++)n[t]+=e[a][t];return n}var er;(function(e){e[e.FIRST_DIM_SIZE=0]="FIRST_DIM_SIZE",e[e.VALUE_ROWIDS=1]="VALUE_ROWIDS",e[e.ROW_LENGTHS=2]="ROW_LENGTHS",e[e.ROW_SPLITS=3]="ROW_SPLITS",e[e.ROW_LIMITS=4]="ROW_LIMITS",e[e.ROW_STARTS=5]="ROW_STARTS"})(er||(er={}));function Az(e,t,n){let a=new Array;if(n==null&&t==null)return a;if(t==null)for(;a.length<e+n.length;)a.push(-1);else a=t.slice();if(n==null)return a;if(e+n.length!==a.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${e+n.length}, but shape.rank = ${a.length}`);for(let r=1;r<n.length;++r){let s=n[r],i=a[a.length-n.length+r],o=a[i];if(s>=0)if(o>=0){if(o!==s)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${r+e}] = ${s} but shape[${r+e}] = ${o}`)}else a[i]=s}return a}function $z(e){let t={FIRST_DIM_SIZE:er.FIRST_DIM_SIZE,VALUE_ROWIDS:er.VALUE_ROWIDS,ROW_LENGTHS:er.ROW_LENGTHS,ROW_SPLITS:er.ROW_SPLITS,ROW_LIMITS:er.ROW_LIMITS,ROW_STARTS:er.ROW_STARTS},n=[];for(let a of e)if(a in t)n.push(t[a]);else break;return n}function Fz(e){return e.length===0?0:e[0]===er.FIRST_DIM_SIZE?e.length-1:e.length}function Dz(e,t){if(e==null||t==null)return;let n=e.length,a=t.length;if(n>=a)throw new Error(`defaultValue.shape=${e} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${n} must be less than ragged tensor input flatValues.rank = ${a})`);for(let r=0;r<Math.min(n,a-1);++r){let s=e[r],i=t[r+1];if(s>=0&&i>=0&&s!==1&&s!==i)throw new Error(`defaultValue.shape=${e}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${r-e.length}] = ${s} but ragged tensor input.flatValues.shape[${r-e.length}] = ${i}`)}}var Hv=30;function Rz(e){return e<=Hv?e:mh(e,Math.floor(Math.sqrt(e)))}function Mz(e,t,n){let a=n*(typeof e=="number"?e:e[0]),r=t*(typeof e=="number"?e:e[1]);return[a,r]}function Pz(e,t,n,a=!0){let r=[];if(a)r=r.concat(t.slice(0)),r.push(e[0]/n),r=r.concat(e.slice(1));else{r=r.concat(e[0]);let s=t.length;for(let i=0;i<s;++i)r=r.concat([e[i+1]/t[i],t[i]]);r=r.concat(e.slice(s+1))}return r}function Oz(e,t,n=!0){let a=[];if(n){a.push(t);for(let r=t+1;r<e;++r)r<=2*t?(a.push(r),a.push(r-(t+1))):a.push(r)}else{let r=[],s=[];for(let i=1;i<e;++i)i>=t*2+1||i%2===1?s.push(i):r.push(i);a.push(...r),a.push(0),a.push(...s)}return a}function Lz(e,t,n,a=!0){let r=[];a?r.push(e[0]/n):r.push(e[0]*n);for(let s=1;s<e.length;++s)s<=t.length?a?r.push(t[s-1]*e[s]):r.push(e[s]/t[s-1]):r.push(e[s]);return r}function zz(e,t){let n=[0];for(let a=0;a<t;++a)n.push(e[a][0]);return n}function Wz(e,t,n){let a=e.slice(0,1);for(let r=0;r<n;++r)a.push(e[r+1]-t[r][0]-t[r][1]);return a}var DT=1.7580993408473768,RT=1.0507009873554805,Bz=.3275911,Vz=.254829592,Uz=-.284496736,Gz=1.421413741,Hz=-1.453152027,jz=1.061405429;function qz(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);let n=new Float32Array(e.length*2);for(let a=0;a<n.length;a+=2)n[a]=e[a/2],n[a+1]=t[a/2];return n}function Kz(e){let t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let a=0;a<e.length;a+=2)t[a/2]=e[a],n[a/2]=e[a+1];return{real:t,imag:n}}function Xz(e){let t=Math.ceil(e.length/4),n=new Float32Array(t),a=new Float32Array(t);for(let r=0;r<e.length;r+=4)n[Math.floor(r/4)]=e[r],a[Math.floor(r/4)]=e[r+1];return{real:n,imag:a}}function Yz(e){let t=Math.floor(e.length/4),n=new Float32Array(t),a=new Float32Array(t);for(let r=2;r<e.length;r+=4)n[Math.floor(r/4)]=e[r],a[Math.floor(r/4)]=e[r+1];return{real:n,imag:a}}function Zz(e,t){let n=e[t*2],a=e[t*2+1];return{real:n,imag:a}}function Jz(e,t,n,a){e[a*2]=t,e[a*2+1]=n}function Qz(e,t){let n=new Float32Array(e/2),a=new Float32Array(e/2);for(let r=0;r<Math.ceil(e/2);r++){let s=(t?2:-2)*Math.PI*(r/e);n[r]=Math.cos(s),a[r]=Math.sin(s)}return{real:n,imag:a}}function eW(e,t,n){let a=(n?2:-2)*Math.PI*(e/t),r=Math.cos(a),s=Math.sin(a);return{real:r,imag:s}}var ub="->",tW=/->/g,Q1=",",ek="...";function nW(e,t){e=e.replace(/\s/g,"");let n=(e.length-e.replace(tW,"").length)/ub.length;if(n<1)throw new Error("Equations without an arrow are not supported.");if(n>1)throw new Error(`Equation must contain exactly one arrow ("${ub}").`);let[a,r]=e.split(ub);$(a.indexOf(ek)===-1,()=>`The ellipsis notation ("${ek}") is not supported yet.`);let s=a.split(Q1),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let o=[];for(let c=0;c<r.length;++c){let h=r[c];if(!s.some(m=>m.indexOf(h)!==-1))throw new Error(`Output subscripts contain the label ${h} not present in the input subscripts.`);o.indexOf(h)===-1&&o.push(h)}for(let c=0;c<a.length;++c){let h=a[c];o.indexOf(h)===-1&&h!==Q1&&o.push(h)}let l=new Array(s.length);for(let c=0;c<i;++c){if(new Set(s[c].split("")).size!==s[c].length)throw new Error(`Found duplicate axes in input component ${s[c]}. Support for duplicate axes in input is not implemented yet.`);l[c]=[];for(let h=0;h<s[c].length;++h)l[c].push(o.indexOf(s[c][h]))}let u=o.length,p=r.length,d=[];for(let c=p;c<u;++c)d.push(c);return{allDims:o,summedDims:d,idDims:l}}function aW(e,t){let n=new Array(e);n.fill(-1);for(let r=0;r<t.length;++r)n[t[r]]=r;let a=[];for(let r=0;r<e;++r)n[r]===-1&&a.push(r);return n=n.filter(r=>r!==-1),{permutationIndices:n,expandDims:a}}function rW(e,t,n){let a=new Array(e);for(let r=0;r<n.length;++r){let s=n[r].shape;for(let i=0;i<t[r].length;++i)a[t[r][i]]===void 0?a[t[r][i]]=s[i]:$(a[t[r][i]]===s[i],()=>`Expected dimension ${a[t[r][i]]} at axis ${i} of input shaped ${JSON.stringify(s)}, but got dimension ${s[i]}`)}}function sW(e,t){let n=e,a=[],r=0;e.length===0&&n.push(-1),r=e.length+1;for(let i=0;i<r;++i)a.push([]);let s=[];for(let i=0;i<n.length;++i){let o=n[i],l=oW(t,o);for(let u of l)s.indexOf(u)===-1&&(a[i].push(u),s.push(u))}return{path:n,steps:a}}function iW(e){return e.every((t,n)=>t===n)}function oW(e,t){let n=[];for(let a=0;a<e.length;++a)(e[a].length===0||e[a].indexOf(t)!==-1||t===-1)&&n.push(a);return n}function lW(e,t,n=0){let a=[];if(typeof t=="number")$(e.shape[n]%t===0,()=>"Number of splits must evenly divide the axis."),a=new Array(t).fill(e.shape[n]/t);else{let r=t.reduce((i,o)=>(o===-1&&(i+=1),i),0);$(r<=1,()=>"There should be only one negative value in split array.");let s=t.indexOf(-1);if(s!==-1){let i=t.reduce((o,l)=>l>0?o+l:o);t[s]=e.shape[n]-i}$(e.shape[n]===t.reduce((i,o)=>i+o),()=>"The sum of sizes must match the size of the axis dimension."),a=t}return a}function uW(e){return`Received SparseTensor with denseShape[0] = 0 but
-  indices.shape[0] = ${e}`}function pW(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function cW(e,t,n){return`indices(${e}, 0) is invalid: ${t} >= ${n}`}function dW(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function hW(e,t){return`size ${e} must be non-negative, not ${t}`}function mW(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function fW(e,t){let n=mt(e),a=mt(t);return`Input to reshape is a SparseTensor with ${n}
-  dense values, but the requested shape requires a multiple of ${a}. inputShape=${e} outputShape= ${t}`}function gW(e,t){let n=mt(e),a=mt(t);return`Input to reshape is a tensor with ${n} dense values, but the requested shape has ${a}. inputShape=${e} outputShape=${t}`}function yW(){return"segment ids must be >= 0"}function bW(){return"segment ids are not increasing"}function xW(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function vW(e,t,n){return`Bad: indices[${e}] == ${t} out of range [0, ${n})`}var MT={};Ae(MT,{collectGatherOpShapeInfo:()=>IW,computeOutShape:()=>kW,segOpComputeOptimalWindowSize:()=>wW});function wW(e,t){let n=!1,a;for(e<=Hv?(a=e,n=!0):a=mh(e,Math.floor(Math.sqrt(e)));!n;)a>t||a===e?n=!0:a=mh(e,a+1);return a}function kW(e,t,n){let a=[],r=e.length;for(let s=0;s<r;s++)s!==t?a.push(e[s]):a.push(n);return a}function IW(e,t,n,a){let r=t.shape.length,s=e.shape.length;if(a!==0&&(a<-r||a>r))throw new Error(`Expect batchDims in the range of [-${r}, ${r}], but got ${a}`);if(a<0&&(a+=r),a>s)throw new Error(`batchDims (${a}) must be less than rank(x) (
-    ${s}).`);if(n<a)throw new Error(`batchDims (${a}) must be less than or equal to axis (${n}).`);for(let d=0;d<a;++d)if(e.shape[d]!==t.shape[d])throw new Error(`x.shape[${d}]: ${e.shape[d]} should be equal to indices.shape[${d}]: ${t.shape[d]}.`);let i=e.shape[n],o=[],l=1,u=1,p=1;for(let d=0;d<a;++d)o.push(e.shape[d]),l*=e.shape[d];for(let d=a;d<n;d++)o.push(e.shape[d]),u*=e.shape[d];for(let d=a;d<r;d++)o.push(t.shape[d]);for(let d=n+1;d<s;d++)o.push(e.shape[d]),p*=e.shape[d];return{batchSize:l,sliceSize:p,outerSize:u,dimSize:i,outputShape:o}}function SW(e){try{return e.map(t=>vh(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function TW(e){return e.map(t=>Sc(t))}var cr={};Ae(cr,{nonMaxSuppressionV3Impl:()=>IT,nonMaxSuppressionV4Impl:()=>ST,nonMaxSuppressionV5Impl:()=>TT,whereImpl:()=>dT});var PT={kernelName:wl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,fo(oe(n,"float32"),-1))}}},NW={kernelName:kl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=ot(oe(n,"float32")),r=ln(pe(be(1),a));return bt(he(e,r))}}}},CW={kernelName:Il,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=ln(pe(ot(oe(n,"float32")),1));return he(e,a)}}}},_W={kernelName:cs,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=e,i=Wt(n.shape,r);return i.length>0&&(s=fe(s,i)),W(s,n.shape)},b:()=>{let s=e,i=Wt(a.shape,r);return i.length>0&&(s=fe(s,i)),W(s,a.shape)}}}},EW={kernelName:mi,saveAllInputs:!0,gradFunc:(e,t)=>{let n={};return t.forEach((a,r)=>{n[r]=()=>e.clone()}),n}},AW={kernelName:fi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},$W={kernelName:sc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},FW={kernelName:Nl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,ln(pe(be(1),ot(oe(n,"float32")))))}}},DW={kernelName:Cl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=ln(Y(be(1),ot(oe(n,"float32"))));return he(e,a)}}}},RW={kernelName:Al,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=Y(ot(n),ot(a)),i=z(e,he(a,s)),o=Wt(n.shape,r);return o.length>0&&(i=fe(i,o)),W(i,n.shape)},b:()=>{let s=Y(ot(n),ot(a)),i=bt(z(e,he(n,s))),o=Wt(a.shape,r);return o.length>0&&(i=fe(i,o)),W(i,a.shape)}}}},MW={kernelName:_l,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,Y(ot(oe(n,"float32")),1))}}},PW={kernelName:El,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,pe(be(1),ot(oe(n,"float32"))))}}};function OW(e,t,n,a,r,s){let i=_(e,"dy","avgPool3dGrad"),o=_(t,"input","avgPool3dGrad"),l=i,u=o,p=!1;o.rank===4&&(p=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),u=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),$(l.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${l.rank}.`),$(u.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${u.rank}.`),Tn("avgPool3dGrad",r,s);let d={dy:l,input:u},c={filterSize:n,strides:a,pad:r,dimRoundingMode:s},h=O.runKernel(Yh,d,c);return p?W(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}var LW=L({avgPool3dGrad_:OW}),zW={kernelName:ic,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{filterSize:r,strides:s,pad:i,dimRoundingMode:o}=n;return{x:()=>LW(e,a,r,s,i,o)}}};function WW(e,t,n,a,r){let s=_(e,"dy","avgPoolGrad"),i=_(t,"input","avgPoolGrad");$(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let o=i,l=s,u=!1;i.rank===3&&(u=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(l.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${l.rank}.`),$(o.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);let p={dy:l,input:o},d={filterSize:n,strides:a,pad:r},c=O.runKernel(Xh,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var BW=L({avgPoolGrad_:WW}),VW={kernelName:gi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{filterSize:r,strides:s,pad:i}=n;return{x:()=>BW(e,a,r,s,i)}}},UW={kernelName:yi,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{let[a,r]=t,{transposeA:s,transposeB:i}=n;return!s&&!i?{a:()=>Fe(e,r,!1,!0),b:()=>Fe(a,e,!0,!1)}:!s&&i?{a:()=>Fe(e,r,!1,!1),b:()=>Fe(e,a,!0,!1)}:s&&!i?{a:()=>Fe(r,e,!1,!0),b:()=>Fe(a,e,!1,!1)}:{a:()=>Fe(r,e,!0,!0),b:()=>Fe(e,a,!0,!0)}}},GW={kernelName:$l,gradFunc:(e,t,n)=>{let{blockShape:a,crops:r}=n;return{x:()=>Pc(e,a,r)}}},HW={kernelName:RI,gradFunc:(e,t,n)=>{let a=n,r=a.inputShape,s=a.shape,i=Array.from(s);for(let l=r.length-1;l>=0;l--)if(r[l]===s[l])i[l]=1;else if(r[l]!==1)throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${s}].`);let o=[];for(let l=0;l<i.length;l++)i[l]>1&&o.push(l);return{x:()=>fe(e,o,!0)}}},jW={kernelName:bi,gradFunc:e=>({x:()=>e.clone()})},qW={kernelName:xi,gradFunc:e=>({x:()=>qe(e)})},KW={kernelName:ds,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{clipValueMin:r,clipValueMax:s}=n;return{x:()=>mn(Ea(Er(a,r),xs(a,s)),e,qe(e))}}},XW={kernelName:oc,inputsToSave:["x"],gradFunc:PT.gradFunc},YW={kernelName:Fl,saveAllInputs:!0,gradFunc:(e,t,n)=>{let a=t.map(o=>o.shape),{axis:r}=n,s=Aa(r,t[0].shape)[0],i=a.map(o=>o[s]);return zn(e,i,s).map(o=>()=>o)}},ZW={kernelName:vi,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,{dilations:s,strides:i,pad:o,dataFormat:l}=n;return $(ns(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>iv(a.shape,e,r,i,o,l),filter:()=>Bv(a,e,r.shape,i,o,l)}}},JW={kernelName:wi,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,{strides:s,pad:i,dataFormat:o,dimRoundingMode:l}=n;return{dy:()=>$t(e,r,s,i,o,1,l),filter:()=>Bv(e,a,r.shape,s,i,o,l)}}};function QW(e,t,n,a,r){let s=e;e.rank===4&&(s=W(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let i=t;i.rank===4&&(i=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),$(s.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),$(i.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),$(n.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),$(s.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${n[3]}.`),$(i.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${n[4]}).`);let o={x:s,dy:i},l={strides:a,pad:r,filterShape:n};return O.runKernel(tm,o,l)}var eB=L({conv3DBackpropFilter_:QW}),tB={kernelName:lc,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:a,strides:r,pad:s}=n;$(ns(a),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${a}'`);let[i,o]=t;return{x:()=>DS(i.shape,e,o,r,s),filter:()=>eB(i,e,o.shape,r,s)}}},nB={kernelName:ki,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(bt(Km(oe(n,"float32"))),e)}}},aB={kernelName:Ii,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(Xm(oe(n,"float32")),e)}}},rB={kernelName:Si,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r,exclusive:s,reverse:i}=n;return{x:()=>{let o=LS([r],a.rank),l=Mm(e,r,s,!i);return o!=null&&(l=Ee(l,o)),l}}}},sB={kernelName:Ti,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:a,strides:r,pad:s,dimRoundingMode:i}=n,o=a==null?[1,1]:a;$(ns(o),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);let[l,u]=t;return $(l.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${l.rank}.`),$(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${u.rank}.`),$(l.shape[3]===u.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),$(ur(r,o),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${r} and dilations '${o}'.`),Tn("depthwiseConv2d",s,i),{x:()=>vT(l.shape,e,u,r,s,o,i),filter:()=>xT(l,e,u.shape,r,s,o,i)}}},iB={kernelName:uc,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,s={x:a,filter:r,dy:e},i={x:a,filter:r,dy:e};return{x:()=>O.runKernel(fh,s,n),filter:()=>O.runKernel(gh,i,n)}}},oB={kernelName:Ci,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t,a={dy:e,y:n};return{x:()=>O.runKernel(lm,a)}}},lB={kernelName:Pl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,a=z(fn(bt(ot(n))),2/Math.sqrt(Math.PI));return{x:()=>z(e,a)}}},uB={kernelName:_i,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,n)}}},pB={kernelName:Ll,inputsToSave:["input"],gradFunc:(e,t)=>{let[n]=t;return{input:()=>W(e,n.shape)}}},cB={kernelName:zl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,fn(n))}}},dB={kernelName:Ei,gradFunc:e=>({x:()=>qe(e)})},hB={kernelName:Ai,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=he(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);i.length>0&&(s=W(fe(s,i),a.shape));let o=ot(a);return bt(he(s,oe(o,"float32")))}}}},mB={kernelName:$i,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{let{varianceEpsilon:a}=n,[r,s,i,o]=t,l=o==null?be(1):o,u=Wt(s.shape,r.shape),p=[];if(s.rank===1){for(let f=0;f<r.shape.length-1;++f)p.push(r.shape[f]);p.push(1)}let d=pe(r,s),c=z(e,l),h=jm(Y(i,be(a))),m=z(z(z(h,h),h),be(-.5));return{x:()=>s.rank===1?W(z(z(e,Ln(W(h,[1,1,1,s.shape[0]]),p)),l),r.shape):W(z(z(e,h),l),r.shape),mean:()=>{let f=z(z(h,be(-1)),c);return s.rank===1&&(f=fe(f,u)),W(f,s.shape)},variance:()=>{let f=z(z(m,d),c);return s.rank===1&&(f=fe(f,u)),W(f,s.shape)},scale:()=>{let f=z(d,h),g=z(e,f);return s.rank===1&&(g=fe(g,u)),W(g,s.shape)},offset:()=>{let f=e;return s.rank===1&&(f=fe(f,u)),W(f,s.shape)}}}},fB={kernelName:Bl,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{let[a,r]=t,{axis:s}=n,i=Aa(s,a.shape)[0];return{x:()=>{let o=a.shape,l=r.size,u=o.slice(0,i),p=u.length,d=o.slice(s,o.length).slice(1),c=d.length,h=tk(0,p),m=tk(p+1,p+1+c),f=nk([u,[l],d]),g=W(e,f),y=W(r,[l]),b=nk([[p],h,m]),x=Ee(g,b),w=ef(x,y,a.shape[i]),I=fv(b);return w=Ee(w,I),w},indices:()=>r}}};function tk(e,t){let n=[];for(let a=e;a<t;++a)n.push(a);return n}function nk(e){let t=[];for(let n=0;n<e.length;++n)for(let a=0;a<e[n].length;++a)t.push(e[n][a]);return t}var gB={kernelName:Fi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>qe(n),b:()=>qe(a)}}},yB={kernelName:Di,gradFunc:e=>({x:()=>oe(e,"float32")})},bB={kernelName:Gl,gradFunc:e=>({x:()=>qe(e)})},xB={kernelName:Hl,gradFunc:e=>({x:()=>qe(e)})},vB={kernelName:jl,gradFunc:e=>({x:()=>qe(e)})},wB={kernelName:Ri,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{alpha:r}=n,s=Nn(a,0);return{x:()=>mn(s,e,z(e,r))}}},kB={kernelName:Xl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,Y(n,1))}}},IB={kernelName:Mi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,oe(n,"float32"))}}},SB={kernelName:PI,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n;return{logits:()=>{let s=fn(a);return pe(e,z(fe(e,r,!0),s))}}}};function TB(e,t,n,a=5,r=1,s=1,i=.5){let o={x:e,y:t,dy:n},l={depthRadius:a,bias:r,alpha:s,beta:i};return O.runKernel(hm,o,l)}var NB=L({localResponseNormalizationBackprop_:TB}),CB={kernelName:cc,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{depthRadius:s,bias:i,alpha:o,beta:l}=n;return{x:()=>NB(a,r,e,s,i,o,l)}}};function OT(e,t,n,a){return t.rank<n.rank&&(t=W(t,si(t.shape,a))),e.rank<n.rank&&(e=W(e,si(e.shape,a))),{x:()=>z(e,oe(Qn(n,t),e.dtype))}}var ak={kernelName:Pi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let a=n,{reductionIndices:r}=a,s=t[0],i=t[1],o=Aa(r,s.shape),l=OT(e,i,s,o);return{x:()=>l.x()}}},_B={kernelName:Oi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>z(e,oe(Er(n,a),"float32")),b:()=>z(e,oe(Om(n,a),"float32"))}}};function EB(e,t,n,a,r,s,i){let o=_(e,"dy","maxPool3dGrad"),l=_(t,"input","maxPool3dGrad"),u=_(n,"output","maxPool3dGrad"),p=o,d=l,c=u,h=!1;l.rank===4&&(h=!0,p=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),d=W(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]]),c=W(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]])),$(p.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${p.rank}.`),$(d.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${d.rank}.`),$(c.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${c.rank}.`),Tn("maxPool3dGrad",s,i);let m={dy:p,input:d,output:c},f={filterSize:a,strides:r,pad:s,dimRoundingMode:i},g=O.runKernel(fm,m,f);return h?W(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var AB=L({maxPool3dGrad_:EB}),$B={kernelName:dc,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n;return{x:()=>AB(e,a,r,s,i,o,l)}}};function FB(e,t,n,a,r,s,i){let o=_(e,"dy","maxPoolGrad"),l=_(t,"input","maxPoolGrad"),u=_(n,"output","maxPoolGrad");$(l.rank===o.rank,()=>`Rank of input (${l.rank}) does not match rank of dy (${o.rank})`),$(o.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`),$(l.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${l.rank}.`),Tn("maxPoolGrad",s,i);let p={dy:o,input:l,output:u},d={filterSize:a,strides:r,pad:s,dimRoundingMode:i};return O.runKernel(mm,p,d)}var DB=L({maxPoolGrad_:FB}),RB={kernelName:Li,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{filterSize:s,strides:i,pad:o}=n;return{x:()=>DB(e,a,r,s,i,o)}}},MB={kernelName:zi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n,s=Aa(r,a.shape),i=OS(a.shape,s)[1],o=mt(i);return{x:()=>{let l=a.shape.slice();s.forEach(p=>{l[p]=1});let u=W(e,l);return he(z(u,Zn(a.shape,"float32")),o)}}}},PB={kernelName:Wi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let a=n,{axis:r}=a,[s,i]=t,o=Aa(r,s.shape),l=OT(e,i,s,o);return{x:()=>l.x()}}},OB={kernelName:Bi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>z(e,oe(xs(n,a),"float32")),b:()=>z(e,oe(Nn(n,a),"float32"))}}},LB={kernelName:Vi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let a=t[0],{paddings:r}=n,s=r.map(i=>i[0]);return{x:()=>Be(e,s,a.shape)}}},zB={kernelName:Ql,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=Wt(n.shape,r);return s.length>0?W(fe(e,s),n.shape):e},b:()=>{let s=z(e,bt(_u(he(n,a)))),i=Wt(a.shape,r);return i.length>0?W(fe(s,i),a.shape):s}}}},WB={kernelName:Ui,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=z(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);return i.length>0?W(fe(s,i),a.shape):s}}}},BB={kernelName:eu,gradFunc:e=>({x:()=>bt(e)})},VB={kernelName:Gi,inputsToSave:["indices"],gradFunc:(e,t)=>{let n=t[0];return{indices:()=>It(n.shape,"float32")}}},UB={kernelName:su,gradFunc:e=>({x:()=>qe(e)})},GB={kernelName:iu,saveAllInputs:!0,gradFunc:(e,t,n)=>{let{axis:a}=n;return ct(e,a).map(r=>()=>r)}},rk={kernelName:Hi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let a=t[0],{paddings:r}=n,s=r.map(i=>i[0]);return{x:()=>Be(e,s,a.shape)}}},HB={kernelName:ji,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{let[n,a,r]=t,s=n,i=a,o=lt(s.shape,i.shape);return{a:()=>{let l=oe(i,"float32"),u=z(e,z(l,_r(s,pe(l,be(1))))),p=Wt(s.shape,o);return p.length>0&&(u=fe(u,p)),W(u,s.shape)},b:()=>{let l=Nn(s,0),u=mn(l,ea(s),qe(s)),p=z(e,z(r,u)),d=Wt(i.shape,o);return d.length>0&&(p=fe(p,d)),W(p,i.shape)}}}},jB={kernelName:qi,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{let[n,a]=t,r=Nn(n,0);return{x:()=>mn(r,e,z(e,a)),alpha:()=>{let s=mn(r,qe(e),z(e,n)),i=Wt(a.shape,e.shape);return i.length>0&&(s=fe(s,i)),W(s,a.shape)}}}};function qB(e,t,n){let a=e.shape.slice();a[n]=1;let r=W(t,a),s=Kp(e,n,!0,!1),i=Kp(e,n,!0,!0),o=z(s,i);return z(r,o)}function KB(e,t,n){let a=e.shape.length,r=a-n.length,s=N.getAxesPermutation(n,a),i=e;s!=null&&(i=Ee(e,s));let o=i.shape.slice(),l=o.splice(a-n.length,n.length).reduce((d,c)=>d*c,1);o.push(l);let u=i.reshape(o),p=qB(u,t,r);if(p=p.reshape(i.shape),s!=null){let d=N.getUndoAxesPermutation(s);p=Ee(p,d)}return p}var XB={kernelName:Ki,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n,s=[];return r==null?s=a.shape.map((i,o)=>o):typeof r=="number"?s=[r]:s=r,{x:()=>KB(a,e,s)}}},YB={kernelName:Ni,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=he(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);i.length>0&&(s=W(fe(s,i),a.shape));let o=ot(a);return bt(he(s,oe(o,"float32")))}}}},ZB={kernelName:ou,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,bt(ot(n)))}}},JB={kernelName:Ji,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,a=z(xs(n,6),fo(n));return{x:()=>z(e,oe(a,"float32"))}}},QB={kernelName:Xi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,oe(fo(n),"float32"))}}},e4={kernelName:lu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>W(e,n.shape)}}},t4={kernelName:Zi,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[a]=t,r={dy:e,images:a};return{images:()=>O.runKernel(Im,r,n)}}},n4={kernelName:Yi,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[a]=t,r={dy:e,images:a};return{images:()=>O.runKernel(km,r,n)}}},a4={kernelName:Qi,gradFunc:(e,t,n)=>{let{dims:a}=n,r=Aa(a,e.shape);return{x:()=>fa(e,r)}}},r4={kernelName:eo,gradFunc:e=>({x:()=>qe(e)})},s4={kernelName:to,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>bt(he(e,z(_r(n,1.5),2)))}}},i4={kernelName:pu,inputsToSave:["condition"],gradFunc:(e,t)=>{let[n]=t;return{condition:()=>oe(qe(n),"float32"),t:()=>z(e,oe(n,e.dtype)),e:()=>z(e,oe(Rc(n),e.dtype))}}},o4={kernelName:cu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=Nn(n,be(0)),r=be(DT),s=be(RT),i=z(e,s),o=z(z(e,r),fn(oe(n,"float32")));return mn(a,i,o)}}}},l4={kernelName:ao,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,z(n,pe(be(1),n)))}}},u4={kernelName:mu,gradFunc:e=>({x:()=>qe(e)})},p4={kernelName:no,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z($c(oe(n,"float32")),e)}}},c4={kernelName:hu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(Rm(oe(n,"float32")),e)}}},d4={kernelName:du,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{begin:r,size:s}=n,i=a.shape,[o,l]=kS(a,r,s),u=[];for(let p=0;p<e.rank;p++)u.push([o[p],i[p]-o[p]-l[p]]);return{x:()=>ya(e,u)}}},h4={kernelName:io,outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a]=t,{dim:r}=n,s=!0,i=z(e,a);return{logits:()=>pe(i,z(fe(i,[r],s),a))}}},m4={kernelName:fu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,da(n))}}},sk={kernelName:gu,gradFunc:(e,t,n)=>{let{blockShape:a,paddings:r}=n;return{x:()=>Ac(e,a,r)}}},ik={kernelName:yu,gradFunc:(e,t,n)=>{let{axis:a}=n;return{x:()=>Ze(e,a)}}},f4={kernelName:ro,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,z(ln(oe(n,"float32")),2))}}},g4={kernelName:yc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,z(oe(n,"float32"),2))}}},y4={kernelName:oo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=be(2);return{a:()=>z(e,z(r,pe(n,a))),b:()=>z(e,z(r,pe(a,n)))}}},b4={kernelName:ms,gradFunc:e=>({x:()=>qe(e)})},x4={kernelName:lo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=lt(n.shape,a.shape);return{a:()=>{let s=e,i=Wt(n.shape,r);return i.length>0&&(s=fe(s,i)),W(s,n.shape)},b:()=>{let s=e,i=Wt(a.shape,r);return i.length>0&&(s=fe(s,i)),W(bt(s),a.shape)}}}},v4={kernelName:so,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,r=a.shape.slice(),{axis:s}=n;Aa(s,a.shape).forEach(l=>{r[l]=1});let i=W(e,r),o=z(i,Zn(a.shape,"float32"));return{x:()=>o}}},w4={kernelName:uo,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,ot($c(n)))}}},k4={kernelName:po,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(pe(be(1),ot(n)),e)}}},I4={kernelName:hs,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{reps:r}=n;return{x:()=>{let s=qe(a);if(a.rank===1)for(let i=0;i<r[0];++i)s=Y(s,Be(e,[i*a.shape[0]],[a.shape[0]]));else if(a.rank===2)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1]],[a.shape[0],a.shape[1]]));else if(a.rank===3)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)for(let l=0;l<r[2];++l)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1],l*a.shape[2]],[a.shape[0],a.shape[1],a.shape[2]]));else if(a.rank===4)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)for(let l=0;l<r[2];++l)for(let u=0;u<r[3];++u)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1],l*a.shape[2],u*a.shape[3]],[a.shape[0],a.shape[1],a.shape[2],a.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${a.rank} tensors yet.`);return s}}}},S4={kernelName:Tr,gradFunc:(e,t,n)=>{let a=n,{perm:r}=a,s=fv(r);return{x:()=>Ee(e,s)}}},T4={kernelName:ku,gradFunc:(e,t,n)=>{let a=n,{axis:r}=a;return{value:()=>Ft(e,r)}}},N4={kernelName:wc,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>C4(e,n)}}};function C4(e,t){let n=pr(t,qe(t)),a=Eu(e,n),r=Er(t,be(0,"int32")),s=a.rank-r.rank;for(let o=0;o<s;++o)r=Zt(r,o+1);r=Ea(r,Zn(a.shape,"bool"));let i=qe(a);return mn(r,a,i)}var _4={kernelName:Iu,gradFunc:e=>({x:()=>qe(e)})},E4=[PT,NW,CW,_W,EW,AW,$W,FW,DW,RW,MW,PW,zW,VW,UW,GW,HW,jW,qW,KW,XW,YW,JW,ZW,tB,nB,aB,rB,sB,iB,YB,oB,lB,uB,pB,cB,hB,dB,mB,fB,gB,yB,bB,xB,vB,wB,kB,IB,SB,CB,ak,ak,_B,$B,RB,MB,PB,OB,LB,zB,WB,BB,VB,UB,GB,rk,rk,HB,jB,XB,ZB,JB,QB,e4,t4,n4,a4,r4,s4,i4,o4,l4,u4,p4,c4,d4,h4,m4,sk,sk,ik,ik,f4,y4,g4,b4,x4,v4,w4,k4,I4,S4,T4,N4,_4];for(let e of E4)OI(e);J().prototype.abs=function(){return this.throwIfDisposed(),Lt(this)};J().prototype.acos=function(){return this.throwIfDisposed(),Bx(this)};J().prototype.acosh=function(){return this.throwIfDisposed(),Vx(this)};J().prototype.add=function(e){return this.throwIfDisposed(),Y(this,e)};J().prototype.all=function(e,t){return this.throwIfDisposed(),$m(this,e,t)};J().prototype.any=function(e,t){return this.throwIfDisposed(),qp(this,e,t)};J().prototype.argMax=function(e){return this.throwIfDisposed(),ai(this,e)};J().prototype.argMin=function(e){return this.throwIfDisposed(),Ux(this,e)};J().prototype.asScalar=function(){return this.throwIfDisposed(),$(this.size===1,()=>"The array must have only 1 element."),W(this,[])};J().prototype.asType=function(e){return this.throwIfDisposed(),oe(this,e)};J().prototype.as1D=function(){return this.throwIfDisposed(),W(this,[this.size])};J().prototype.as2D=function(e,t){return this.throwIfDisposed(),W(this,[e,t])};J().prototype.as3D=function(e,t,n){return this.throwIfDisposed(),W(this,[e,t,n])};J().prototype.as4D=function(e,t,n,a){return this.throwIfDisposed(),W(this,[e,t,n,a])};J().prototype.as5D=function(e,t,n,a,r){return this.throwIfDisposed(),W(this,[e,t,n,a,r])};J().prototype.asin=function(){return this.throwIfDisposed(),Gx(this)};J().prototype.asinh=function(){return this.throwIfDisposed(),Hx(this)};J().prototype.atan=function(){return this.throwIfDisposed(),jx(this)};J().prototype.atan2=function(e){return this.throwIfDisposed(),qx(this,e)};J().prototype.atanh=function(){return this.throwIfDisposed(),Kx(this)};J().prototype.avgPool=function(e,t,n,a){return this.throwIfDisposed(),ga(this,e,t,n,a)};J().prototype.batchToSpaceND=function(e,t){return this.throwIfDisposed(),Ac(this,e,t)};J().prototype.batchNorm=function(e,t,n,a,r){return this.throwIfDisposed(),ys(this,e,t,n,a,r)};J().prototype.broadcastTo=function(e){return this.throwIfDisposed(),Ks(this,e)};J().prototype.cast=function(e){return this.throwIfDisposed(),oe(this,e)};J().prototype.ceil=function(){return this.throwIfDisposed(),tv(this)};J().prototype.clipByValue=function(e,t){return this.throwIfDisposed(),en(this,e,t)};J().prototype.concat=function(e,t){return this.throwIfDisposed(),e instanceof Te&&(e=[e]),Ze([this,...e],t)};J().prototype.conv1d=function(e,t,n,a,r,s){return this.throwIfDisposed(),Fm(this,e,t,n,a,r,s)};J().prototype.conv2dTranspose=function(e,t,n,a,r){return this.throwIfDisposed(),Dm(this,e,t,n,a,r)};J().prototype.conv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),$t(this,e,t,n,a,r,s)};J().prototype.cos=function(){return this.throwIfDisposed(),$c(this)};J().prototype.cosh=function(){return this.throwIfDisposed(),Rm(this)};J().prototype.cumprod=function(e,t,n){return this.throwIfDisposed(),Kp(this,e,t,n)};J().prototype.cumsum=function(e,t,n){return this.throwIfDisposed(),Mm(this,e,t,n)};J().prototype.depthToSpace=function(e,t){return this.throwIfDisposed(),uv(this,e,t)};J().prototype.depthwiseConv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),bs(this,e,t,n,a,r,s)};J().prototype.dilation2d=function(e,t,n,a,r){return this.throwIfDisposed(),pv(this,e,t,n,a,r)};J().prototype.divNoNan=function(e){return this.throwIfDisposed(),cv(this,e)};J().prototype.div=function(e){return this.throwIfDisposed(),he(this,e)};J().prototype.dot=function(e){return this.throwIfDisposed(),dv(this,e)};J().prototype.elu=function(){return this.throwIfDisposed(),Nu(this)};J().prototype.equal=function(e){return this.throwIfDisposed(),Qn(this,e)};J().prototype.erf=function(){return this.throwIfDisposed(),hv(this)};J().prototype.euclideanNorm=function(e,t){return this.throwIfDisposed(),gv(this,e,t)};J().prototype.exp=function(){return this.throwIfDisposed(),fn(this)};J().prototype.expandDims=function(e){return this.throwIfDisposed(),Zt(this,e)};J().prototype.expm1=function(){return this.throwIfDisposed(),yv(this)};J().prototype.fft=function(){return this.throwIfDisposed(),zc(this)};J().prototype.flatten=function(){return this.throwIfDisposed(),W(this,[this.size])};J().prototype.floor=function(){return this.throwIfDisposed(),_u(this)};J().prototype.floorDiv=function(e){return this.throwIfDisposed(),Am(this,e)};J().prototype.gather=function(e,t){return this.throwIfDisposed(),Eu(this,e,t)};J().prototype.greaterEqual=function(e){return this.throwIfDisposed(),Er(this,e)};J().prototype.greater=function(e){return this.throwIfDisposed(),Nn(this,e)};J().prototype.ifft=function(){return this.throwIfDisposed(),ul(this)};J().prototype.irfft=function(){return this.throwIfDisposed(),Zm(this)};J().prototype.isFinite=function(){return this.throwIfDisposed(),bv(this)};J().prototype.isInf=function(){return this.throwIfDisposed(),xv(this)};J().prototype.isNaN=function(){return this.throwIfDisposed(),vv(this)};J().prototype.leakyRelu=function(e){return this.throwIfDisposed(),Fc(this,e)};J().prototype.lessEqual=function(e){return this.throwIfDisposed(),xs(this,e)};J().prototype.less=function(e){return this.throwIfDisposed(),Om(this,e)};J().prototype.localResponseNormalization=function(e,t,n,a){return this.throwIfDisposed(),wv(this,e,t,n,a)};J().prototype.logSigmoid=function(){return this.throwIfDisposed(),kv(this)};J().prototype.logSoftmax=function(e){return this.throwIfDisposed(),zm(this,e)};J().prototype.logSumExp=function(e,t){return this.throwIfDisposed(),Wm(this,e,t)};J().prototype.log=function(){return this.throwIfDisposed(),ea(this)};J().prototype.log1p=function(){return this.throwIfDisposed(),Dc(this)};J().prototype.logicalAnd=function(e){return this.throwIfDisposed(),Ea(this,e)};J().prototype.logicalNot=function(){return this.throwIfDisposed(),Rc(this)};J().prototype.logicalOr=function(e){return this.throwIfDisposed(),Bm(this,e)};J().prototype.logicalXor=function(e){return this.throwIfDisposed(),Iv(this,e)};J().prototype.matMul=function(e,t,n){return this.throwIfDisposed(),Fe(this,e,t,n)};J().prototype.maxPool=function(e,t,n,a){return this.throwIfDisposed(),Dt(this,e,t,n,a)};J().prototype.max=function(e,t){return this.throwIfDisposed(),ha(this,e,t)};J().prototype.maximum=function(e){return this.throwIfDisposed(),pr(this,e)};J().prototype.mean=function(e,t){return this.throwIfDisposed(),Nt(this,e,t)};J().prototype.min=function(e,t){return this.throwIfDisposed(),il(this,e,t)};J().prototype.minimum=function(e){return this.throwIfDisposed(),Au(this,e)};J().prototype.mirrorPad=function(e,t){return this.throwIfDisposed(),Tv(this,e,t)};J().prototype.mod=function(e){return this.throwIfDisposed(),Nv(this,e)};J().prototype.mul=function(e){return this.throwIfDisposed(),z(this,e)};J().prototype.neg=function(){return this.throwIfDisposed(),bt(this)};J().prototype.norm=function(e,t,n){return this.throwIfDisposed(),Cu(this,e,t,n)};J().prototype.notEqual=function(e){return this.throwIfDisposed(),ii(this,e)};J().prototype.oneHot=function(e,t=1,n=0){return this.throwIfDisposed(),rl(this,e,t,n)};J().prototype.onesLike=function(){return this.throwIfDisposed(),ta(this)};J().prototype.pad=function(e,t){return this.throwIfDisposed(),ya(this,e,t)};J().prototype.pool=function(e,t,n,a,r,s){return this.throwIfDisposed(),Cv(this,e,t,n,a,r,s)};J().prototype.pow=function(e){return this.throwIfDisposed(),_r(this,e)};J().prototype.prelu=function(e){return this.throwIfDisposed(),Oc(this,e)};J().prototype.prod=function(e,t){return this.throwIfDisposed(),_v(this,e,t)};J().prototype.reciprocal=function(){return this.throwIfDisposed(),$v(this)};J().prototype.relu=function(){return this.throwIfDisposed(),Xe(this)};J().prototype.relu6=function(){return this.throwIfDisposed(),Gm(this)};J().prototype.reshapeAs=function(e){return this.throwIfDisposed(),W(this,e.shape)};J().prototype.reshape=function(e){return this.throwIfDisposed(),W(this,e)};J().prototype.resizeBilinear=function(e,t,n){return this.throwIfDisposed(),NT(this,e,t,n)};J().prototype.resizeNearestNeighbor=function(e,t,n){return this.throwIfDisposed(),CT(this,e,t,n)};J().prototype.reverse=function(e){return this.throwIfDisposed(),fa(this,e)};J().prototype.rfft=function(){return this.throwIfDisposed(),Wc(this)};J().prototype.round=function(){return this.throwIfDisposed(),Hm(this)};J().prototype.rsqrt=function(){return this.throwIfDisposed(),jm(this)};J().prototype.selu=function(){return this.throwIfDisposed(),qm(this)};J().prototype.separableConv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),vs(this,e,t,n,a,r,s)};J().prototype.sigmoid=function(){return this.throwIfDisposed(),da(this)};J().prototype.sign=function(){return this.throwIfDisposed(),Fv(this)};J().prototype.sin=function(){return this.throwIfDisposed(),Km(this)};J().prototype.sinh=function(){return this.throwIfDisposed(),Xm(this)};J().prototype.slice=function(e,t){return this.throwIfDisposed(),Be(this,e,t)};J().prototype.softmax=function(e){return this.throwIfDisposed(),Ka(this,e)};J().prototype.softplus=function(){return this.throwIfDisposed(),ho(this)};J().prototype.spaceToBatchND=function(e,t){return this.throwIfDisposed(),Pc(this,e,t)};J().prototype.split=function(e,t){return this.throwIfDisposed(),zn(this,e,t)};J().prototype.sqrt=function(){return this.throwIfDisposed(),ln(this)};J().prototype.square=function(){return this.throwIfDisposed(),ot(this)};J().prototype.squaredDifference=function(e){return this.throwIfDisposed(),Jm(this,e)};J().prototype.squeeze=function(e){return this.throwIfDisposed(),ws(this,e)};J().prototype.stack=function(e,t){this.throwIfDisposed();let n=e instanceof Te?[this,e]:[this,...e];return Ft(n,t)};J().prototype.step=function(e){return this.throwIfDisposed(),fo(this,e)};J().prototype.stridedSlice=function(e,t,n,a,r,s,i,o){return this.throwIfDisposed(),Dv(this,e,t,n,a,r,s,i,o)};J().prototype.sub=function(e){return this.throwIfDisposed(),pe(this,e)};J().prototype.sum=function(e,t){return this.throwIfDisposed(),fe(this,e,t)};J().prototype.tan=function(){return this.throwIfDisposed(),Rv(this)};J().prototype.tanh=function(){return this.throwIfDisposed(),ri(this)};J().prototype.tile=function(e){return this.throwIfDisposed(),Ln(this,e)};J().prototype.toBool=function(){return this.throwIfDisposed(),oe(this,"bool")};J().prototype.toFloat=function(){return this.throwIfDisposed(),oe(this,"float32")};J().prototype.toInt=function(){return this.throwIfDisposed(),oe(this,"int32")};J().prototype.topk=function(e,t){return this.throwIfDisposed(),Mv(this,e,t)};J().prototype.transpose=function(e){return this.throwIfDisposed(),Ee(this,e)};J().prototype.unique=function(e){return this.throwIfDisposed(),Pv(this,e)};J().prototype.unsortedSegmentSum=function(e,t){return this.throwIfDisposed(),ef(this,e,t)};J().prototype.unstack=function(e){return this.throwIfDisposed(),ct(this,e)};J().prototype.where=function(e,t){return this.throwIfDisposed(),mn(e,this,t)};J().prototype.zerosLike=function(){return this.throwIfDisposed(),qe(this)};var vr=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,vr.prototype)}},Wa=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Wa.prototype)}},V=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,V.prototype)}},Re=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Re.prototype)}},LT=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,LT.prototype)}},zT=class{constructor(e){this.maxEntries=e||100,this.cache=new Map}get(e){let t;return this.cache.has(e)&&(t=this.cache.get(e),this.cache.delete(e),this.cache.set(e,t)),t}put(e,t){if(this.cache.has(e))this.cache.delete(e);else if(this.cache.size>=this.maxEntries){let n=this.cache.keys().next().value;this.cache.delete(n)}this.cache.set(e,t)}getMaxEntries(){return this.maxEntries}setMaxEntries(e){if(e<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${e}.`);if(this.maxEntries>e)for(let t=0;t<this.maxEntries-e;t++){let n=this.cache.keys().next().value;this.cache.delete(n)}this.maxEntries=e}};function oi(e,t){if(Array.isArray(e)){let n=[];for(let a=0;a<t;a++)n=n.concat(e);return n}else{let n=new Array(t);return n.fill(e),n}}function tr(e,t){if(!e)throw new LT(t)}function ok(e,t){let n=0;for(let a of e)a===t&&n++;return n}function On(e){return e.length===1?e[0]:e}function yt(e){return Array.isArray(e)?e:[e]}function wr(e){let t=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return t[0]!=="_"?t:"private"+t}function Vs(e){return e.length<=1||e.indexOf("_")===-1?e:e.replace(/[_]+(\w|$)/g,(t,n)=>n.toUpperCase())}var ka={};function jv(e){if(e==null)return null;let t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function Mb(e){if(!(e==null||typeof e!="object"))if(Array.isArray(e))e.forEach(t=>Mb(t));else{let t=Object.keys(e);for(let n of t){let a=e[n];a!=null&&typeof a=="object"&&(!Array.isArray(a)&&a.type==="ndarray"&&typeof a.value=="number"?e[n]=a.value:Mb(a))}}}function Vc(e,t={},n={},a="object",r=!1){if(typeof e=="string"){let s=e,i;if(s in n)i=n[s];else if(s in ka)i=ka[s];else if(i=t[s],i==null)throw new V(`Unknown ${a}: ${e}. This may be due to one of the following reasons:
-1. The ${a} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${a} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return i}else{let s=e;if(s.className==null||s.config==null)throw new V(`${a}: Improper config format: ${JSON.stringify(s)}.
-'className' and 'config' must set.`);let i=s.className,o,l;if(i in n?[o,l]=n[i]:i in ka?[o,l]=ka.className:i in t&&([o,l]=t[i]),o==null)throw new V(`Unknown ${a}: ${i}. This may be due to one of the following reasons:
-1. The ${a} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${a} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(l!=null){let u={};for(let h of Object.keys(ka))u[h]=ka[h];for(let h of Object.keys(n))u[h]=n[h];let p=s.config;p.customObjects=u;let d=Object.assign({},ka);for(let h of Object.keys(n))ka[h]=n[h];Mb(s.config);let c=l(o,s.config,n,r);return ka=Object.assign({},d),c}else{let u=Object.assign({},ka);for(let d of Object.keys(n))ka[d]=n[d];let p=new o(s.config);return ka=Object.assign({},u),p}}}function A4(e,t){return e<t?-1:e>t?1:0}function jd(e,t){return-1*A4(e,t)}function Jr(e){if(e==null)return e;let t=[];for(let n of e)t.indexOf(n)===-1&&t.push(n);return t}function $4(e){if(e==null)throw new V(`Invalid value in obj: ${JSON.stringify(e)}`);for(let t in e)if(e.hasOwnProperty(t))return!1;return!0}function go(e,t,n){if(n!=null&&e.indexOf(n)<0)throw new V(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)}function qv(e,t,n=0,a=1/0){return tr(n>=0),tr(a>=n),Array.isArray(e)&&e.length>=n&&e.length<=a&&e.every(r=>typeof r===t)}function Qt(e,t){Array.isArray(e)?(v.assert(e.length>0,()=>`${t} is unexpectedly an empty array.`),e.forEach((n,a)=>Qt(n,`element ${a+1} of ${t}`))):v.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${WT(e)}.`)}function WT(e){return e===null?"null":Array.isArray(e)?"["+e.map(t=>WT(t)).join(",")+"]":typeof e=="string"?`"${e}"`:`${e}`}function F4(e,t,n){let a=n!=null?n():v.now(),r;return(...s)=>{let i=n!=null?n():v.now();return i-a<t||(a=i,r=e(...s)),r}}function BT(e){return e==="relu"?"relu":e==="linear"?"linear":e==="elu"?"elu":null}var D4=0;function VT(){return D4++}var qd={};function hf(e=""){return e in qd||(qd[e]=0),qd[e]+=1,e+qd[e].toString()}var R4=["channelsFirst","channelsLast"],M4=["nearest","bilinear"],P4=["valid","same","causal"],O4=["max","avg"],L4=["sum","mul","concat","ave"],Ho=new Map;function Rt(e){go(R4,"DataFormat",e)}function z4(e){go(M4,"InterpolationFormat",e)}function ba(e){go(P4,"PaddingMode",e)}function UT(e){go(O4,"PoolMode",e)}var zp=[],lk="/";function Xs(e,t){zp.push(e);try{let n=t();return zp.pop(),n}catch(n){throw zp.pop(),n}}function W4(){return zp.length===0?"":zp.join(lk)+lk}function GT(e){if(!jT(e))throw new Error("Not a valid tensor name: '"+e+"'");return W4()+e}function HT(e){if(!jT(e))throw new Error("Not a valid tensor name: '"+e+"'");Ho.has(e)||Ho.set(e,0);let t=Ho.get(e);if(Ho.set(e,Ho.get(e)+1),t>0){let n=`${e}_${t}`;return Ho.set(n,1),n}else return e}var B4=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function jT(e){return!!e.match(B4)}function V4(e){return e===parseInt(e.toString(),10)}function Qr(e,t,n){t==null&&(t=0),n==null&&(n=e.length);let a=1;for(let r=t;r<n;++r)a*=e[r];return a}function cl(e){if(e.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let n=0;n<e.length;n++){let a=e[n];a<t&&(t=a)}return t}function as(e){if(e.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let n=0;n<e.length;n++){let a=e[n];a>t&&(t=a)}return t}function Ha(e,t){if(t<e)throw new V(`end (${t}) < begin (${e}) is forbidden.`);let n=[];for(let a=e;a<t;++a)n.push(a);return n}var pb;function Gt(){return pb==null&&(pb=uS().epsilon()),pb}function ja(){return"channelsLast"}function yo(e,t){return oe(e,t)}function Uc(e,t=-1){let n=e.shape.slice();return t<0&&(t=n.length+t+1),n.splice(t,0,1),W(e,n)}function U4(e,t){return P(()=>{if(e.shape.length!==2)throw new V(`repeat() expects a rank-2 tensor, but received a rank-${e.shape.length} tensor.`);let n=Uc(e,1);return Pb(n,[1,t,1])})}function G4(e){let t=[Qr(e.shape)];return W(e,t)}function H4(e){if(e.rank<=1)throw new V(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);let t=[e.shape[0],Qr(e.shape,1)];return W(e,t)}function Ys(e,t,n){return P(()=>{switch(e.rank){case 1:return Lc(e,t,n);case 2:return Ym(e,[t,0],[n,e.shape[1]]);case 3:return mo(e,[t,0,0],[n,e.shape[1],e.shape[2]]);case 4:return ll(e,[t,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3]]);case 5:return Be(e,[t,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return Be(e,[t,0,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new V(`sliceAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}})}function cb(e,t,n){return P(()=>{switch(e.rank){case 1:return Lc(e,t,n);case 2:return Ym(e,[0,t],[e.shape[0],n]);case 3:return mo(e,[0,0,t],[e.shape[0],e.shape[1],n]);case 4:return ll(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],n]);default:throw new V(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function Kd(e,t,n,a){return P(()=>{switch(e.rank){case 1:return Lc(e,t,n);case 2:switch(a){case 1:return Ys(e,t,n);case 2:return cb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}case 3:switch(a){case 1:return Ys(e,t,n);case 2:return mo(e,[0,t,0],[e.shape[0],n,e.shape[2]]);case 3:return cb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}case 4:switch(a){case 1:return Ys(e,t,n);case 2:return ll(e,[0,t,0,0],[e.shape[0],n,e.shape[2],e.shape[3]]);case 3:return ll(e,[0,0,t,0],[e.shape[0],e.shape[1],n,e.shape[3]]);case 4:return cb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}default:throw new V(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function Kv(e,t=-1){let n;return t<0&&(n=e[0].rank,n!==0?t=n:t=0),t===e[0].rank&&(t=-1),Ze(e,t)}function uk(e,t){switch(e.rank){case 1:return nv([e,t]);case 2:return av([e,t],0);case 3:return rv([e,t],0);case 4:return sv([e,t],0);default:throw new V(`concatAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}}function Pb(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new V(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return Ln(e,t)}function mf(e,t=0,n=1,a,r){return Um(e,t,n,a,r)}function sr(e,t,n,a){if(e.rank<2||t.rank<2)throw new Re(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){let r=e.shape.slice(-1)[0],s=t.shape.slice(-2)[0];if(r!==s)throw new Re(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)}if(e.rank===2&&t.rank===2)return pl.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:a?Ob(e.rank,a,ja()):null,activation:n});{let r=e.shape.slice(),s=r.pop();e=W(e,[-1,s]);let i=t.shape.slice(),o=i.pop(),l=i.pop(),u=[...i,o],p=Array.from({length:t.rank},(m,f)=>f===0?t.rank-2:f<=t.rank-2?f-1:f);t=W(Ee(t,p),[l,-1]);let d=[...r,...u],c=!1,h=!1;return W(pl.matMul({a:e,b:t,transposeA:c,transposeB:h,bias:a?Ob(e.rank,a,ja()):null,activation:n}),d)}}function qT(e,t,n){return P(()=>(Array.isArray(t)?t=Ke(t,"int32"):t=oe(t,"int32"),Eu(e,t,n)))}function Gc(e){return z(e,e)}function Ob(e,t,n){let a=t.shape;if(t.rank!==1&&t.rank!==e)throw new V(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(e===5){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1,1,1]):W(t,[1,a[3],a[0],a[1],a[2]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,1,1,a[0]]):W(t,[1].concat(a))}else if(e===4){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1,1]):W(t,[1,a[2],a[0],a[1]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,1,a[0]]):W(t,[1].concat(a))}else if(e===3){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1]):W(t,[1,a[1],a[0]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,a[0]]):W(t,[1].concat(a))}else if(e<3)return t;throw new V(`Unsupported input rank by biasAdd: ${t.rank}`)}function Xa(e,t,n){return P(()=>(n==null&&(n=ja()),Rt(n),Y(e,Ob(e.rank,t,n))))}function j4(e,t=1){if(t!==1)throw new Re(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Nu(e)}function q4(e){return P(()=>he(e,Y(Lt(e),1)))}function KT(e,t,n,a){return P(()=>zv(e,t,n,a))}function K4(e){return P(()=>{let t=Y(.5,z(.2,e));return en(t,0,1)})}function Hc(e,t,n=!1){return n?e():t()}var X4=["fanIn","fanOut","fanAvg"],Y4=["normal","uniform","truncatedNormal"];function Z4(e){go(X4,"FanMode",e)}function J4(e){go(Y4,"Distribution",e)}var Fa=class extends ne.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},Xv=class extends Fa{apply(e,t){return It(e,t)}};Xv.className="Zeros";ne.registerClass(Xv);var ff=class extends Fa{apply(e,t){return Zn(e,t)}};ff.className="Ones";ne.registerClass(ff);var Yv=class extends Fa{constructor(e){if(super(),typeof e!="object")throw new V(`Expected argument of type ConstantConfig but got ${e}`);if(e.value===void 0)throw new V(`config must have value set but got ${e}`);this.value=e.value}apply(e,t){return P(()=>z(be(this.value),Zn(e,t)))}getConfig(){return{value:this.value}}};Yv.className="Constant";ne.registerClass(Yv);var Zv=class extends Fa{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return $u(e,this.minval,this.maxval,t)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};Zv.className="RandomUniform";ne.registerClass(Zv);var Jv=class extends Fa{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Re(`randomNormal does not support dType ${t}.`);return mf(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};Jv.className="RandomNormal";ne.registerClass(Jv);var Qv=class extends Fa{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Re(`truncatedNormal does not support dType ${t}.`);return Qm(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};Qv.className="TruncatedNormal";ne.registerClass(Qv);var ew=class extends Fa{constructor(e){super(),this.gain=e.gain!=null?e.gain:1}apply(e,t){return P(()=>{if(e.length!==2||e[0]!==e[1])throw new V("Identity matrix initializer can only be used for 2D square matrices.");return z(this.gain,Pm(e[0]))})}getConfig(){return{gain:this.gain}}};ew.className="Identity";ne.registerClass(ew);function Q4(e,t="channelsLast"){let n,a;if(Rt(t),e.length===2)n=e[0],a=e[1];else if([3,4,5].indexOf(e.length)!==-1){if(t==="channelsFirst"){let r=Qr(e,2);n=e[1]*r,a=e[0]*r}else if(t==="channelsLast"){let r=Qr(e,0,e.length-2);n=e[e.length-2]*r,a=e[e.length-1]*r}}else{let r=Qr(e);n=Math.sqrt(r),a=Math.sqrt(r)}return[n,a]}var Bn=class extends Fa{constructor(e){if(super(),e.scale<0)throw new V(`scale must be a positive float. Got: ${e.scale}`);this.scale=e.scale==null?1:e.scale,this.mode=e.mode==null?"fanIn":e.mode,Z4(this.mode),this.distribution=e.distribution==null?"normal":e.distribution,J4(this.distribution),this.seed=e.seed}apply(e,t){let n=Q4(e),a=n[0],r=n[1],s=this.scale;if(this.mode==="fanIn"?s/=Math.max(1,a):this.mode==="fanOut"?s/=Math.max(1,r):s/=Math.max(1,(a+r)/2),this.distribution==="normal"){let i=Math.sqrt(s);if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Re(`${this.getClassName()} does not support dType ${t}.`);return Qm(e,0,i,t,this.seed)}else{let i=Math.sqrt(3*s);return $u(e,-i,i,t)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};Bn.className="VarianceScaling";ne.registerClass(Bn);var gf=class extends Bn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};gf.className="GlorotUniform";ne.registerClass(gf);var yf=class extends Bn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};yf.className="GlorotNormal";ne.registerClass(yf);var bf=class extends Bn{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};bf.className="HeNormal";ne.registerClass(bf);var xf=class extends Bn{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};xf.className="HeUniform";ne.registerClass(xf);var vf=class extends Bn{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};vf.className="LeCunNormal";ne.registerClass(vf);var wf=class extends Bn{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};wf.className="LeCunNormal";ne.registerClass(wf);var tw=class extends Fa{constructor(e){if(super(),this.DEFAULT_GAIN=1,this.gain=e.gain==null?this.DEFAULT_GAIN:e.gain,this.seed=e.seed,this.seed!=null)throw new Re("Random seed is not implemented for Orthogonal Initializer yet.")}apply(e,t){return P(()=>{if(e.length<2)throw new Re("Shape must be at least 2D.");e[0]*e[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);let n=e[0]>e[1]?[e[1],e[0]]:e,a=mf(n,0,1,"float32"),r=Uv.gramSchmidt(a);return e[0]>e[1]&&(r=Ee(r)),z(this.gain,r)})}getConfig(){return{gain:this.gain,seed:this.seed}}};tw.className="Orthogonal";ne.registerClass(tw);var pk={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function ck(e,t={}){return Vc(e,ne.SerializationMap.getMap().classNameMap,t,"initializer")}function Ct(e){return jv(e)}function St(e){if(typeof e=="string"){let t=e in pk?pk[e]:e;if(t==="GlorotNormal")return new yf;if(t==="GlorotUniform")return new gf;if(t==="HeNormal")return new bf;if(t==="HeUniform")return new xf;if(t==="LeCunNormal")return new vf;if(t==="LeCunUniform")return new wf;{let n={};return n.className=t,n.config={},ck(n)}}else return e instanceof Fa?e:ck(e)}function Lb(e){return Array.isArray(e)&&Array.isArray(e[0])}function Th(e){return e.length===0?[]:Array.isArray(e[0])?e:[e]}function Ne(e){let t;if(Array.isArray(e)){if(e.length!==1)throw new V(`Expected Tensor length to be 1; got ${e.length}`);t=e[0]}else t=e;return t}function Qe(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(e.length===1)return e=e,e[0];throw new V(`Expected exactly 1 Shape; got ${e.length}`)}else return e}function Nh(e){let t=0;for(let n of e)n.shape.length===0?t+=1:t+=n.shape.reduce((a,r)=>a*r);return t}var dk="Variable",XT=class{constructor(e,t="float32",n=dk,a=!0,r=null){this.dtype=t==null?"float32":t,this.shape=e.shape,this.id=VT(),n=n==null?dk:n,this.originalName=GT(n),this.name=HT(this.originalName),this.trainable_=a,this.constraint=r,this.val=Ov(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),eV(this.val,e),this.val.id!==e.id&&(this.val.assign(e),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function eV(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}function zb(e){return e.map(t=>t.read())}function nw(e){e.forEach(t=>{t[0].write(t[1])})}var zt=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,e.shape!=null?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},Ba=class{constructor(e,t,n,a,r,s,i){this.dtype=e,this.shape=t,this.sourceLayer=n,this.inputs=a,this.callArgs=r,this.outputTensorIndex=i,this.id=VT(),s!=null&&(this.originalName=GT(s),this.name=HT(this.originalName)),this.rank=t.length}},tV=0,kf=class{constructor(e,t){this.callArgs=t,this.id=tV++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(let n of e.inboundLayers)n!=null&&n.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){let e=[];for(let t of this.inboundLayers)t!=null?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},nV=0,Ge=class extends ne.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=nV++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){let n=this.getClassName();t=wr(n)+"_"+hf(n)}if(this.name=t,this.trainable_=e.trainable==null?!0:e.trainable,e.inputShape!=null||e.batchInputShape!=null){let n;if(e.batchInputShape!=null)n=e.batchInputShape;else if(e.inputShape!=null){let r=null;e.batchSize!=null&&(r=e.batchSize),n=[r].concat(e.inputShape)}this.batchInputShape=n;let a=e.dtype;a==null&&(a=e.inputDType),a==null&&(a="float32"),this.dtype=a}e.weights!=null?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(this.inboundNodes.length===0)throw new Wa(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new V(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return On(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return On(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new vr(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(this.inboundNodes.length===0)throw new vr(`Layer ${this.name} is not connected, no input to return.`);return On(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new vr(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new vr(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return On(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){if(e=yt(e),this.inputSpec==null||this.inputSpec.length===0)return;let t=yt(this.inputSpec);if(e.length!==t.length)throw new V(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);for(let n=0;n<e.length;n++){let a=e[n],r=t[n];if(r==null)continue;let s=a.rank;if(r.ndim!=null&&s!==r.ndim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${r.ndim}, found ndim=${s}`);if(r.maxNDim!=null&&s>r.maxNDim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${r.maxNDim}, found ndim=${s}`);if(r.minNDim!=null&&s<r.minNDim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${r.minNDim}, found ndim=${s}.`);if(r.dtype!=null&&a.dtype!==r.dtype)throw new V(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${r.dtype}, found dtype=${a.dtype}.`);if(r.axes){let i=a.shape;for(let o in r.axes){let l=Number(o),u=r.axes[o],p=l>=0?i[l]:i[i.length+l];if(u!=null&&[u,null].indexOf(p)===-1)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected axis ${l} of input shape to have value ${u} but got shape ${i}.`)}}if(r.shape!=null)for(let i=0;i<r.shape.length;++i){let o=r.shape[i],l=a.shape[i];if(o!=null&&l!=null&&o!==l)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected shape=${r.shape}, found shape=${a.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){this._callHook!=null&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();let n=yt(e),a=!0;for(let s of n)if(!(s instanceof Ba)){a=!1;break}let r=!0;for(let s of n)if(s instanceof Ba){r=!1;break}if(a===r)throw new V("Arguments to apply() must be all SymbolicTensors or all Tensors");return Xs(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);let s=[];for(let i of yt(e))s.push(i.shape);this.build(On(s)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&r&&(this._refCount=1)}if(this.assertInputCompatibility(e),r){let s=this.call(e,t),i=yt(s),o=[];for(let l of i)n.indexOf(l)!==-1&&(l=l.clone()),o.push(l);if(s=On(o),this.activityRegularizer!=null)throw new Re("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return s}else{let s=aV(e),i=this.computeOutputShape(s),o,l=rV(e);if(this.warnOnIncompatibleInputShape(Array.isArray(e)?s[0]:s),i!=null&&i.length>0&&Array.isArray(i[0])?o=i.map((u,p)=>new Ba(l,u,this,yt(e),t,this.name,p)):o=new Ba(l,i,this,yt(e),t,this.name),this.addInboundNode(e,o,null,null,s,i,t),this._refCount++,this.activityRegularizer!=null)throw new Re("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return o}})}warnOnIncompatibleInputShape(e){if(this.batchInputShape!=null)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((n,a)=>{n!=null&&e[a]!=null&&e[a]!==n&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new vr(`The layer ${this.name} has never been called and thus has no defined output shape.`);let e=[];for(let t of this.inboundNodes){let n=JSON.stringify(t.outputShapes);e.indexOf(n)===-1&&e.push(n)}if(e.length===1){let t=this.inboundNodes[0].outputShapes;return Array.isArray(t)&&Array.isArray(t[0])&&t.length===1?t[0]:t}else throw new vr(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new Wa(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Nh(this.weights)}build(e){this.built=!0}getWeights(e=!1){return zb(e?this.trainableWeights:this.weights)}setWeights(e){P(()=>{let t=this.weights;if(t.length!==e.length)throw new V(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(t.length===0)return;let n=[],a=zb(t);for(let r=0;r<a.length;++r){let s=a[r],i=t[r],o=e[r];if(!v.arraysEqual(s.shape,o.shape))throw new V(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);n.push([i,o])}nw(n)})}addWeight(e,t,n,a,r,s,i,o){if(this._addedWeightNames.indexOf(e)!==-1)throw new V(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),n==null&&(n="float32"),this.fastWeightInitDuringBuild&&(a=o!=null?o():St("zeros"));let l=a.apply(t,n),u=new XT(l,n,e,s,i);return l.dispose(),r!=null&&this.addLoss(()=>r.apply(u.read())),s==null&&(s=!0),s?this._trainableWeights.push(u):this._nonTrainableWeights.push(u),u}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){e==null||Array.isArray(e)&&e.length===0||(e=yt(e),this._losses!==void 0&&this._losses!==null&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(t!=null)if(Array.isArray(t))t.forEach(n=>{if(n!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return t}addInboundNode(e,t,n,a,r,s,i=null){let o=yt(e);t=yt(t),n=yt(n),a=yt(a),r=Th(r),s=Th(s);let l=[],u=[],p=[];for(let d of o)l.push(d.sourceLayer),u.push(d.nodeIndex),p.push(d.tensorIndex);new kf({outboundLayer:this,inboundLayers:l,nodeIndices:u,tensorIndices:p,inputTensors:o,outputTensors:t,inputMasks:n,outputMasks:a,inputShapes:r,outputShapes:s},i);for(let d=0;d<t.length;d++)t[d].sourceLayer=this,t[d].nodeIndex=this.inboundNodes.length-1,t[d].tensorIndex=d}getConfig(){let e={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(e.batchInputShape=this.batchInputShape),this.dtype!=null&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return--this._refCount===0&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};function aV(e){e=yt(e);let t=[];for(let n of e)t.push(n.shape);return On(t)}function rV(e){return"float32"}function YT(e,t,n){if((t==null||n!=null&&n>0)&&(t=e.sourceLayer,n=e.nodeIndex),t.inboundNodes.length===0)return[e];{let a=t.inboundNodes[n];if(a.inboundLayers.length===0)return a.inputTensors;{let r=[];for(let s=0;s<a.inboundLayers.length;s++){let i=a.inputTensors[s],o=a.inboundLayers[s],l=a.nodeIndices[s],u=YT(i,o,l);for(let p of u)r.indexOf(p)===-1&&r.push(p)}return r}}}var Du=class extends Ge{constructor(e){if(super({dtype:e.dtype,name:e.name!=null?e.name:hf("input").toString()}),e.batchSize==null&&(e.batchSize=null),e.sparse==null&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,e.inputShape!=null&&e.batchInputShape!=null)throw new V("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(t==null){if(e.inputShape==null)throw new V("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(e.batchSize!=null)throw new V("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let n=e.dtype||"float32";this.batchInputShape=t,this.dtype=n,this.inputSpec=[{shape:t}];let a=new Ba(this.dtype,this.batchInputShape,this,[],{},this.name);a.nodeIndex=0,a.tensorIndex=0,new kf({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[a],outputTensors:[a],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new V(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};Du.className="InputLayer";ne.registerClass(Du);function ZT(e){if(e.batchShape==null&&e.shape==null)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(e.batchShape!=null&&e.shape!=null)throw new V("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;e.shape!=null&&t==null&&(t=[null].concat(e.shape));let n=e.dtype;return n==null&&(n="float32"),new Du({batchInputShape:t,name:e.name,dtype:n,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}function sV(e,t){if(e.dtype==null||e.dtype===t.dtype)return t;try{return oe(t,e.dtype)}catch(n){throw new V(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}var Hs=class{constructor(e){if(this.id2Value={},this.id2Mask={},this.name2Id={},e instanceof Hs)for(let t in e.id2Value)this.id2Value[t]=e.id2Value[t],t in e.id2Mask&&(this.id2Mask[t]=e.id2Mask[t]);else{if(e==null)return;for(let t of e)this.add(t.key,t.value)}}add(e,t,n){if(this.id2Value[e.id]==null)this.id2Value[e.id]=sV(e,t),this.name2Id[e.name]=e.id,n!=null&&(this.id2Mask[e.id]=n);else throw new V(`Duplicate key: name=${e.name}, id=${e.id}`);return this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return this.id2Value[e.id]!=null}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof Ba){if(this.id2Value[e.id]==null)throw new V(`Nonexistent key: ${e.name}`);return this.id2Value[e.id]}else{let t=this.name2Id[e];if(t==null)throw new V(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Value[t]}}getMask(e){if(e instanceof Ba){if(this.id2Value[e.id]==null)throw new V(`Nonexistent key: ${e.name}`);return this.id2Mask[e.id]}else{let t=this.name2Id[e];if(t==null)throw new V(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Mask[t]}}disposeMasks(){this.id2Mask!=null&&_e(this.id2Mask)}},Ch=new zT,_h=new zT;function iV(e){Ch!=null&&Ch.setMaxEntries(e),_h!=null&&_h.setMaxEntries(e)}function Ep(e,t,n,a){let r=n==null?!1:n.training,s=Array.isArray(e),i=s?e:[e],o=i.map(m=>m.name),l=[],u=t.names();for(let m of o)u.indexOf(m)!==-1?l.push(t.getValue(m)):l.push(null);a!=null&&(a.maxNumTensors=-1/0,a.minNumTensors=1/0);let p=o.join(",")+"|"+t.names().sort().join(","),d=Ch.get(p),c;if(d==null){let m=oV(i,t);d=m.sorted,c=m.recipientCounts,Ch.put(p,d),_h.put(p,c)}c={},r||Object.assign(c,_h.get(p));let h=new Hs(t);for(let m=0;m<d.length;++m){if(a!=null){let A=kh().numTensors;A>a.maxNumTensors&&(a.maxNumTensors=A),A<a.minNumTensors&&(a.minNumTensors=A)}let f=d[m],g=f.sourceLayer;if(g instanceof Du)continue;let y=[],b=[],x=[],w=!1;for(let A of f.inputs){let R=h.getValue(A),F=h.getMask(A);y.push(R),b.push(F),F!=null&&(w=!0),r||(c[A.name]--,c[A.name]===0&&!t.hasKey(A)&&o.indexOf(A.name)===-1&&!R.isDisposed&&A.sourceLayer.stateful!==!0&&x.push(R))}w&&(n=n||{},n.mask=b[0]);let I=yt(g.apply(y,n)),T=null;g.supportsMasking&&(T=g.computeMask(y,b));let C=uV(f),E=Array.isArray(C)?C:[C];for(let A=0;A<E.length;++A){h.hasKey(E[A])||h.add(E[A],I[A],Array.isArray(T)?T[0]:T);let R=o.indexOf(E[A].name);R!==-1&&(l[R]=I[A])}r||_e(x)}return h.disposeMasks(),s?l:l[0]}function oV(e,t){v.assert(e!=null&&e.length>0,()=>"Expected at least one fetch, got none");let n=[],a={};if(e.length===1){let r=hk(e[0],t);n=r.sorted,a=r.recipientMap}else{let r=new Set;for(let s of e){let{sorted:i,recipientMap:o}=hk(s,t);for(let l of i)r.has(l.name)||(n.push(l),r.add(l.name));for(let l in o)a[l]==null&&(a[l]=new Set),o[l].forEach(u=>a[l].add(u))}}return{sorted:n,recipientCounts:lV(a)}}function lV(e){let t={};for(let n in e)t[n]=e[n].size;return t}function hk(e,t){let n=new Set,a=[],r={};for(let o of t.names())n.add(o);let s=[],i=[];for(s.push(e);s.length>0;){let o=s[s.length-1];if(n.has(o.name)){s.pop();continue}let l=i[i.length-1]===s.length-1;if(o.inputs.length===0||l)s.pop(),a.push(o),n.add(o.name),l&&i.pop();else{i.push(s.length-1);for(let u of o.inputs)r[u.name]==null&&(r[u.name]=new Set),r[u.name].add(o.name),!n.has(u.name)&&s.push(u)}}return{sorted:a,recipientMap:r}}function uV(e){let t;if(e.sourceLayer.inboundNodes.length===1)t=e.sourceLayer.output;else{let n=null;for(let a=0;a<e.sourceLayer.inboundNodes.length;++a)for(let r of e.sourceLayer.inboundNodes[a].outputTensors)if(r.id===e.id){n=a;break}t=e.sourceLayer.getOutputAt(n)}return t}var pV=H();pV.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,iV);var JT={};Ae(JT,{maxNorm:()=>cV,minMaxNorm:()=>mV,nonNeg:()=>hV,unitNorm:()=>dV});function aw(e,t){return P(()=>ln(fe(z(e,e),t,!0)))}var jc=class extends ne.Serializable{getConfig(){return{}}},rw=class extends jc{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>{let t=aw(e,this.axis),n=en(t,0,this.maxValue);return z(e,he(n,Y(Gt(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};rw.className="MaxNorm";ne.registerClass(rw);var sw=class extends jc{constructor(e){super(),this.defaultAxis=0,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>he(e,Y(Gt(),aw(e,this.axis))))}getConfig(){return{axis:this.axis}}};sw.className="UnitNorm";ne.registerClass(sw);var iw=class extends jc{apply(e){return Xe(e)}};iw.className="NonNeg";ne.registerClass(iw);var ow=class extends jc{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=e.minValue!=null?e.minValue:this.defaultMinValue,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.rate=e.rate!=null?e.rate:this.defaultRate,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>{let t=aw(e,this.axis),n=Y(z(this.rate,en(t,this.minValue,this.maxValue)),z(1-this.rate,t));return z(e,he(n,Y(Gt(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};ow.className="MinMaxNorm";ne.registerClass(ow);var mk={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function qt(e){return jv(e)}function fk(e,t={}){return Vc(e,ne.SerializationMap.getMap().classNameMap,t,"constraint")}function Kt(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in mk?mk[e]:e,config:{}};return fk(t)}else return e instanceof jc?e:fk(e)}function cV(e){return new rw(e)}function dV(e){return new sw(e)}function hV(){return new iw}function mV(e){return new ow(e)}var QT={};Ae(QT,{constant:()=>yV,glorotNormal:()=>SV,glorotUniform:()=>IV,heNormal:()=>TV,heUniform:()=>NV,identity:()=>wV,leCunNormal:()=>CV,leCunUniform:()=>_V,ones:()=>gV,orthogonal:()=>EV,randomNormal:()=>xV,randomUniform:()=>bV,truncatedNormal:()=>vV,varianceScaling:()=>kV,zeros:()=>fV});function fV(){return new Xv}function gV(){return new ff}function yV(e){return new Yv(e)}function bV(e){return new Zv(e)}function xV(e){return new Jv(e)}function vV(e){return new Qv(e)}function wV(e){return new ew(e)}function kV(e){return new Bn(e)}function IV(e){return new gf(e)}function SV(e){return new yf(e)}function TV(e){return new bf(e)}function NV(e){return new xf(e)}function CV(e){return new vf(e)}function _V(e){return new wf(e)}function EV(e){return new tw(e)}var eN={};Ae(eN,{Layer:()=>Ge,RNN:()=>dr,RNNCell:()=>Zc,activation:()=>nG,add:()=>cG,alphaDropout:()=>KG,average:()=>dG,averagePooling1d:()=>y0,averagePooling2d:()=>b0,averagePooling3d:()=>x0,avgPool1d:()=>wG,avgPool2d:()=>IG,avgPool3d:()=>TG,avgPooling1d:()=>kG,avgPooling2d:()=>SG,avgPooling3d:()=>NG,batchNormalization:()=>bG,bidirectional:()=>WG,categoryEncoding:()=>JG,concatenate:()=>hG,conv1d:()=>qU,conv2d:()=>KU,conv2dTranspose:()=>XU,conv3d:()=>YU,conv3dTranspose:()=>ZU,convLstm2d:()=>PG,convLstm2dCell:()=>OG,cropping2D:()=>QU,dense:()=>aG,depthwiseConv2d:()=>tG,dot:()=>yG,dropout:()=>rG,elu:()=>BU,embedding:()=>pG,flatten:()=>iG,gaussianDropout:()=>qG,gaussianNoise:()=>jG,globalAveragePooling1d:()=>CG,globalAveragePooling2d:()=>_G,globalMaxPool1d:()=>VG,globalMaxPool2d:()=>UG,globalMaxPooling1d:()=>HN,globalMaxPooling2d:()=>jN,gru:()=>AG,gruCell:()=>$G,input:()=>bN,inputLayer:()=>WU,layerNormalization:()=>xG,leakyReLU:()=>UU,lstm:()=>FG,lstmCell:()=>DG,masking:()=>XG,maxPool1d:()=>GG,maxPool2d:()=>HG,maxPooling1d:()=>qN,maxPooling2d:()=>KN,maxPooling3d:()=>EG,maximum:()=>mG,minimum:()=>fG,multiply:()=>gG,permute:()=>uG,prelu:()=>GU,reLU:()=>VU,repeatVector:()=>oG,rescaling:()=>YG,reshape:()=>lG,resizing:()=>ZG,rnn:()=>LG,separableConv2d:()=>JU,simpleRNN:()=>RG,simpleRNNCell:()=>MG,softmax:()=>HU,spatialDropout1d:()=>sG,stackedRNNCells:()=>zG,thresholdedReLU:()=>jU,timeDistributed:()=>BG,upSampling2d:()=>eG,zeroPadding2d:()=>vG});async function jr(e){if(e==null)return;let t=[],n=[],a=[];for(let r in e){let s=e[r];if(typeof s!="number"){let i=s;t.push(i.data()),n.push(r),a.push(i)}}if(t.length>0){let r=await Promise.all(t);for(let s=0;s<r.length;++s)e[n[s]]=r[s][0];_e(a)}}function tN(e){if(e!=null)for(let t in e){let n=e[t];typeof n!="number"&&n.dispose()}}var gk;(function(e){e[e.SILENT=0]="SILENT",e[e.VERBOSE=1]="VERBOSE"})(gk||(gk={}));var AV=125,dl=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},nN=class{constructor(e,t=10){e==null&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(let t of this.callbacks)t.setParams(e)}setModel(e){for(let t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochBegin(e,t)}async onEpochEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochEnd(e,t)}async onBatchBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchBegin(e,t)}async onBatchEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchEnd(e,t)}async onTrainBegin(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainEnd(e)}},$V=class extends dl{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){t==null&&(t={});let n=t.size==null?0:t.size;this.seen+=n;for(let a in t){let r=t[a];if(typeof r=="number")this.totals.hasOwnProperty(a)||(this.totals[a]=0),this.totals[a]=this.totals[a]+r*n;else{let s;a in this.totals?s=this.totals[a]:this.totals[a]=0;let i=P(()=>Y(this.totals[a],z(r,n)));this.totals[a]=i,s!=null&&s.dispose()}}}async onEpochEnd(e,t){if(t!=null)for(let n of this.params.metrics)this.totals[n]!=null&&(typeof this.totals[n]=="number"?t[n]=this.totals[n]/this.seen:P(()=>{let a=z(he(1,this.seen),this.totals[n]);t[n]=a,this.totals[n].dispose(),Jt(t[n])}))}},aN=class extends dl{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){t==null&&(t={}),this.epoch.push(e);for(let n in t)this.history[n]==null&&(this.history[n]=[]),this.history[n].push(t[n])}async syncData(){let e=[],t=[],n=[];for(let r in this.history){let s=this.history[r];for(let i=0;i<s.length;++i)if(typeof s[i]!="number"){let o=s[i];e.push(o.data()),t.push(r),n.push(i)}}let a=await Promise.all(e);for(let r=0;r<a.length;++r)this.history[t[r]][n[r]].dispose(),this.history[t[r]][n[r]]=a[r][0]}},rN=class extends dl{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||Gv,this.yieldEvery=t||"auto",this.yieldEvery==="auto"&&(this.yieldEvery=AV),this.yieldEvery==="never"&&e.onYield!=null)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");v.isNumber(this.yieldEvery)&&(this.maybeWait=F4(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,n){let a=[];this.yield!=null&&(await jr(n),a.push(this.yield(e,t,n))),a.push(this.nextFrameFunc()),await Promise.all(a)}async onEpochBegin(e,t){this.currentEpoch=e,this.epochBegin!=null&&(await jr(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){let n=[];this.epochEnd!=null&&(await jr(t),n.push(this.epochEnd(e,t))),this.yieldEvery==="epoch"&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(e,t){this.batchBegin!=null&&(await jr(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){let n=[];this.batchEnd!=null&&(await jr(t),n.push(this.batchEnd(e,t))),this.yieldEvery==="batch"?n.push(this.nextFrameFunc()):v.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(n)}async onTrainBegin(e){this.trainBegin!=null&&(await jr(e),await this.trainBegin(e))}async onTrainEnd(e){this.trainEnd!=null&&(await jr(e),await this.trainEnd(e))}};function sN(e,t){return e==null&&(e={}),e instanceof dl?[e]:Array.isArray(e)&&e[0]instanceof dl?e:yt(e).map(n=>new rN(n,t))}var Ta=class{constructor(){}static registerCallbackConstructor(e,t){v.assert(e>=0&&Number.isInteger(e),()=>`Verbosity level is expected to be an integer >= 0, but got ${e}`),Ta.checkForDuplicate(t),Ta.constructors[e]==null&&(Ta.constructors[e]=[]),Ta.constructors[e].push(t)}static checkForDuplicate(e){for(let t in Ta.constructors)Ta.constructors[+t].forEach(n=>{if(n===e)throw new V("Duplicate callback constructor.")})}static clear(){Ta.constructors={}}static createCallbacks(e){let t=[];for(let n in Ta.constructors){let a=+n;e>=a&&t.push(...Ta.constructors[a])}return t.map(n=>new n)}};Ta.constructors={};function iN(e,t,n,a,r,s,i,o,l){let u=new aN,p=[new $V,...Ta.createCallbacks(t)];e!=null&&p.push(...e),p.push(u);let d=new nN(p);return d.setParams({epochs:n,initialEpoch:a,samples:r,steps:s,batchSize:i,verbose:t,doValidation:o,metrics:l}),{callbackList:d,history:u}}function Ua(e,t={},n=!1){return Vc(e,ne.SerializationMap.getMap().classNameMap,t,"layer",n)}function Eh(e,t){return P(()=>{e.dtype!=="float32"&&(e=oe(e,"float32"));let n=fe(Gc(e),t,!0),a=gn(n.shape,Gt()),r=ln(pr(n,a));return he(e,r)})}function bo(e,t){return P(()=>Nt(Gc(pe(t,e)),-1))}function If(e,t){return P(()=>Nt(Lt(pe(t,e)),-1))}function Ru(e,t){return P(()=>{let n=pe(e,t),a=en(Lt(e),Gt(),Number.MAX_VALUE),r=Lt(he(n,a));return z(100,Nt(r,-1))})}function FV(e,t){return P(()=>{let n=en(t,Gt(),Number.MAX_VALUE),a=ea(Y(1,n)),r=en(e,Gt(),Number.MAX_VALUE),s=ea(Y(1,r));return Nt(Gc(pe(a,s)),-1)})}function DV(e,t){return P(()=>{let n=pr(0,pe(1,z(e,t)));return Nt(Gc(n),-1)})}function RV(e,t){return P(()=>{let n=pr(0,pe(1,z(e,t)));return Nt(n,-1)})}function MV(e,t){return P(()=>{let n=fe(z(e,t),-1),a=ha(z(pe(1,e),t),-1);return pr(0,Y(1,pe(a,n)))})}function PV(e,t){return P(()=>{let n=Math.log(2),a=pe(t,e),r=pe(Y(a,ho(z(-2,a))),n);return Nt(r,-1)})}function Xp(e,t,n=!1){return P(()=>{if(n)t=Ka(t);else{let a=fe(t,t.shape.length-1,!0);t=he(t,a)}return t=en(t,Gt(),1-Gt()),bt(fe(z(oe(e,"float32"),ea(t)),t.shape.length-1))})}function Ah(e,t,n=!1){return P(()=>{let a=oe(_u(G4(e)),"int32");t=en(t,Gt(),1-Gt());let r=t.shape,s=W(rl(a,r[r.length-1]),r);return Xp(s,t,n)})}function OV(e,t){if(!v.arraysEqual(e.shape,t.shape))throw new V(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return P(()=>{let n=Xe(t),a=bt(Lt(t));return Y(pe(n,z(t,e)),Dc(fn(a)))})}function Sf(e,t){return P(()=>{let n;return n=en(t,Gt(),1-Gt()),n=ea(he(n,pe(1,n))),Nt(OV(e,n),-1)})}function LV(e,t){return P(()=>{let n=en(e,Gt(),1),a=en(t,Gt(),1);return fe(z(e,ea(he(n,a))),-1)})}function zV(e,t){return P(()=>{let n=ea(Y(Gt(),t));return Nt(pe(t,z(e,n)),-1)})}function lw(e,t){return P(()=>{let n=Eh(e,-1),a=Eh(t,-1),r=z(n,a);return bt(fe(r,-1))})}var $h={meanSquaredError:bo,meanAbsoluteError:If,meanAbsolutePercentageError:Ru,meanSquaredLogarithmicError:FV,squaredHinge:DV,hinge:RV,categoricalHinge:MV,logcosh:PV,categoricalCrossentropy:Xp,sparseCategoricalCrossentropy:Ah,binaryCrossentropy:Sf,kullbackLeiblerDivergence:LV,poisson:zV,cosineProximity:lw};function db(e){if(typeof e=="string"){if(e in $h)return $h[e];let t=`Unknown loss ${e}`;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new V(t)}else return e}function uw(e,t){return P(()=>{let n=z(.5,ta(t)),a=yo(Nn(t,n),e.dtype);return Nt(Qn(e,a),-1)})}function pw(e,t){return P(()=>yo(Qn(ai(e,-1),ai(t,-1)),"float32"))}function oN(e,t){return P(()=>oe(fe(Ea(Qn(e,1),Qn(t,1))),"float32"))}function WV(e,t){return P(()=>oe(fe(Ea(Qn(e,1),Qn(t,0))),"float32"))}function BV(e,t){return P(()=>oe(fe(Ea(Qn(e,0),Qn(t,1))),"float32"))}function lN(e,t){return P(()=>{let n=oN(e,t),a=BV(e,t),r=Y(n,a);return oe(mn(Nn(r,0),he(n,r),0),"float32")})}function VV(e,t){return P(()=>{let n=oN(e,t),a=WV(e,t),r=Y(n,a);return oe(mn(Nn(r,0),he(n,r),0),"float32")})}function uN(e,t){return Sf(e,t)}function pN(e,t){return e.rank===t.rank&&(e=ws(e,[e.rank-1])),t=ai(t,-1),t.dtype!==e.dtype&&(t=oe(t,e.dtype)),oe(Qn(e,t),"float32")}var UV=bo,GV=bo,HV=If,jV=If,qV=Ru,KV=Ru,cw=Xp,XV=lw,cN=Ah,Fh={binaryAccuracy:uw,categoricalAccuracy:pw,precision:lN,categoricalCrossentropy:cw,sparseCategoricalCrossentropy:cN,mse:UV,MSE:GV,mae:HV,MAE:jV,mape:qV,MAPE:KV,cosine:XV};function YV(e){if(typeof e=="string"&&e in Fh)return Fh[e];if(typeof e!="string"&&e!=null)return e;throw new V(`Unknown metric ${e}`)}function Xd(e){if(tr(e!==null,`Unknown LossOrMetricFn ${e}`),typeof e=="string")return e;{let t;for(let n of Object.keys($h))if($h[n]===e){t=n;break}if(t!==void 0)return t;for(let n of Object.keys(Fh))if(Fh[n]===e){t=n;break}return t!==void 0?t:e.name}}function ZV(e){let t={Adagrad:()=>zs.adagrad(.01),Adadelta:()=>zs.adadelta(1,.95,Gt()),Adam:()=>zs.adam(.001,.9,.999,Gt()),Adamax:()=>zs.adamax(.002,.9,.999,Gt(),0),RMSProp:()=>zs.rmsprop(.001,.9,0,Gt()),SGD:()=>zs.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new V(`Unknown Optimizer ${e}`)}function yk(e,t,n=!1){if(e==null||typeof e!="object"||Object.getPrototypeOf(e)!==Object.prototype||!Wb(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(n){let a=JSON.stringify(e);a.length>1048576&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${a.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${1048576}.`)}}function Wb(e){if(e===null)return!0;if(typeof e=="object")if(Object.getPrototypeOf(e)===Object.prototype){let t=Object.keys(e);for(let n of t)if(typeof n!="string"||!Wb(e[n]))return!1;return!0}else if(Array.isArray(e)){for(let t of e)if(!Wb(t))return!1;return!0}else return!1;else{let t=typeof e;return t==="string"||t==="number"||t==="boolean"}}function JV(e,t,n,a=console.log){let r=eU(e),s=["Layer (type)","Input Shape","Output shape","Param #"];r?(t=t||90,n=n||[.32,.61,.89,1]):(t=t||115,n=n||[.24,.48,.7,.8,1]),n[n.length-1]<=1&&(n=n.map(p=>Math.floor(t*p)));let i;if(!r){s.push("Receives inputs"),i=[];for(let p in e.nodesByDepth)i.push(...e.nodesByDepth[p])}a("_".repeat(t)),Dh(s,n,a),a("=".repeat(t));let o=e.layers;for(let p=0;p<o.length;++p)r?tU(o[p],n,a):nU(o[p],n,i,a),a((p===o.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();let l=QV(e),u=Nh(e.nonTrainableWeights);a(`Total params: ${l+u}`),a(`Trainable params: ${l}`),a(`Non-trainable params: ${u}`),a("_".repeat(t))}function QV(e){let t;return e.collectedTrainableWeights!=null?t=Nh(e.collectedTrainableWeights):t=Nh(e.trainableWeights),t}function eU(e){let t=!0,n=[],a=[];for(let r in e.nodesByDepth)n.push(e.nodesByDepth[r]);for(let r of n){if(r.length>1||r.length===1&&r[0].inboundLayers.length>1){t=!1;break}a.push(...r)}if(t)for(let r of e.layers){let s=!1;for(let i of r.inboundNodes)if(a.indexOf(i)!==-1)if(s){t=!1;break}else s=!0;if(!t)break}return t}function Dh(e,t,n=console.log){let a="";for(let r=0;r<e.length;++r)r>0&&(a=a.slice(0,a.length-1)+" "),a+=e[r],a=a.slice(0,t[r]),a+=" ".repeat(t[r]-a.length);n(a)}function tU(e,t,n){let a,r;try{r=e.inboundNodes.map(l=>JSON.stringify(l.inputShapes)).join(",")}catch(l){r="multiple"}try{a=JSON.stringify(e.outputShape)}catch(l){a="multiple"}let s=e.name,i=e.getClassName(),o=[`${s} (${i})`,r,a,e.countParams().toString()];Dh(o,t,n)}function nU(e,t,n,a){let r,s;try{s=e.inboundNodes.map(d=>JSON.stringify(d.inputShapes)).join(",")}catch(d){s="multiple"}try{r=JSON.stringify(e.outputShape)}catch(d){r="multiple"}let i=[];for(let d of e.inboundNodes)if(!(n!=null&&n.length>0&&n.indexOf(d)===-1))for(let c=0;c<d.inboundLayers.length;++c){let h=d.inboundLayers[c].name,m=d.nodeIndices[c],f=d.tensorIndices[c];i.push(`${h}[${m}][${f}]`)}let o=e.name,l=e.getClassName(),u=i.length===0?"":i[0],p=[`${o} (${l})`,s,r,e.countParams().toString(),u];Dh(p,t,a);for(let d=1;d<i.length;++d)Dh(["","","","",i[d]],t,a)}function dN(e,t,n){return(e==="inboundNodes"||e==="outputLayers"||e==="inputLayers")&&t===0&&typeof n=="string"}function Yp(e,t){if(e===null)return null;if(typeof e=="string")return Vs(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],a=e.length;for(let r=0;r<a;++r){let s=e[r];dN(t,r,s)?n.push(s):n.push(Yp(s,t))}return n}else{let n={};for(let a of Object.keys(e)){let r=e[a];if(a==="name"&&typeof r=="string")n[a]=r;else{let s=Vs(a);n[s]=Yp(r,s)}}return n}}function Bb(e,t){if(e==null)return null;if(typeof e=="string")return wr(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],a=e.length;for(let r=0;r<a;++r){let s=e[r];dN(t,r,s)?n.push(s):n.push(Bb(s,t))}return n}else{let n={};for(let a of Object.keys(e)){let r=e[a],s=wr(a);(a==="name"||a==="className")&&typeof r=="string"?n[s]=r:n[s]=Bb(r,a)}return n}}var dw="4.0.0",Qa=class extends Ge{constructor(e){if(super({}),this.containerNodes=new Set,this.name=e.name,this.name==null){let y=this.getClassName().toLowerCase();this.name=hf(y)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(e.inputs)?this.inputs=e.inputs.slice():this.inputs=[e.inputs],Array.isArray(e.outputs)?this.outputs=e.outputs.slice():this.outputs=[e.outputs],Jr(this.inputs).length!==this.inputs.length)throw new V(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(y=>y.name)}`);Jr(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(y=>y.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let y of this.outputs){let b=y.sourceLayer,x=y.nodeIndex,w=y.tensorIndex;this.outputLayers.push(b),this.outputLayersNodeIndices.push(x),this.outputLayersTensorIndices.push(w)}for(let y of this.inputs){let b=y.sourceLayer,x=y.nodeIndex,w=y.tensorIndex;tr(x===0,"input layer has >1 nodes"),tr(w===0,"input layer has >1 tensors"),this.inputLayers.push(b),this.inputLayersNodeIndices.push(x),this.inputLayersTensorIndices.push(w)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let y=0;y<this.inputLayers.length;y++){let b=this.inputLayers[y];if(!(b instanceof Du))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${y} (0-based) originates from layer type ${b.getClassName()}.`);this.inputNames.push(b.name),this.feedInputShapes.push(b.batchInputShape),this.feedInputNames.push(b.name)}for(let y of this.outputLayers)this.outputNames.push(y.name);this.internalInputShapes=this.inputs.map(y=>y.shape),this.internalOutputShapes=this.outputs.map(y=>y.shape);let t={},n={},a={},r={},s={},i=[],o=(y,b,x,w,I,T)=>{(w==null||I==null||T==null)&&(w=y.sourceLayer,I=y.nodeIndex,T=y.tensorIndex);let C=w.inboundNodes[I];if(x.indexOf(C)!==-1)throw new Wa(`The tensor ${y.name} at layer "${w.name}" is part of a cycle.`);if(b.indexOf(C)!==-1)return;this.containerNodes.add(Qa.nodeKey(w,I)),w.id in s||(s[w.id]=Object.keys(s).length),x.indexOf(C)===-1&&x.push(C);let E=C.inboundLayers.length;for(let A=0;A<E;A++){let R=C.inputTensors[A],F=C.inboundLayers[A],S=C.nodeIndices[A],M=C.tensorIndices[A];o(R,b,x,F,S,M)}for(b.push(C);x.indexOf(C)>=0;)x.splice(x.indexOf(C),1);i.push(C)},l=[],u=[];for(let y of this.outputs)o(y,l,u);let p=i.slice().reverse();for(let y of p){n[y.id]=y,y.id in t||(t[y.id]=0);let b=t[y.id],x=a[y.outboundLayer.id]==null?0:a[y.outboundLayer.id];b=Math.max(b,x),a[y.outboundLayer.id]=b,r[y.outboundLayer.id]=y.outboundLayer,t[y.id]=b;for(let w=0;w<y.inboundLayers.length;w++){let I=y.inboundLayers[w],T=y.nodeIndices[w],C=I.inboundNodes[T],E=t[C.id]==null?0:t[C.id];t[C.id]=Math.max(b+1,E),n[C.id]=C}}let d={};for(let y in t){let b=t[y];b in d||(d[b]=[]),d[b].push(n[y])}let c={};for(let y in a){let b=a[y];b in c||(c[b]=[]),c[b].push(r[y])}let h=Object.keys(c).map(y=>parseInt(y,10)).sort(jd);this.layers=[];for(let y of h){let b=c[y];b.sort((x,w)=>{let I=s[x.id],T=s[w.id];return I<T?-1:I>T?1:0});for(let x of b)x instanceof Qa&&this.internalContainerRefs.push(x),this.layers.push(x)}this.layersByDepth=c,h=Object.keys(d).map(y=>parseInt(y,10)).sort(jd);let m=this.inputs.slice(),f=[];for(let y of h)for(let b of d[y]){let x=b.outboundLayer;if(x!=null){for(let w of b.inputTensors)if(m.indexOf(w)===-1)throw new Wa(`Graph disconnected: cannot obtain value for tensor ${w} at layer "${x.name}". The following previous layers were accessed without issue: ${f}`);for(let w of b.outputTensors)m.push(w);f.push(x.name)}}this.nodesByDepth=d;let g=this.layers.map(y=>y.name);for(let y of g){let b=g.filter(x=>x===y).length;if(b!==1)throw new Wa(`The name "${y}" is used ${b} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(g))}this.outboundNodes=[],this.inboundNodes=[],new kf({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(y=>null),outputMasks:this.outputs.map(y=>null),inputShapes:this.inputs.map(y=>y.shape),outputShapes:this.outputs.map(y=>y.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let e={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(let t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(n=>n.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new V("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(let t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.layers)t.push(...n.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){let n={},a=0;for(let s of this.layers)for(let i of s.weights){if(n[i.originalName]!=null)throw new V(`Duplicate weight name: ${i.originalName}`);n[i.originalName]=i,a++}let r=[];for(let s in e){let i=s;if(n[s]==null){let o=s.split("/");i=o.slice(0,-2).concat([o[o.length-1]]).join("/")}if(n[i]!=null)r.push([n[i],e[s]]);else if(t)throw new V(`Provided weight data has no target variable: ${s}`);delete n[i]}if(t){let s=[];for(let i in n)s.push(i);if(s.length>0)throw new V(`${s.length} of ${a} weights are not set: ${s}`)}nw(r)}updatedConfig(){let e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion=`tfjs-layers ${dw}`,t.backend="TensorFlow.js",t}toJSON(e,t=!0){let n=Bb(this.updatedConfig());return t?JSON.stringify(n):n}call(e,t){return P(()=>{e=yt(e);let n=new Hs;for(let a=0;a<this.inputs.length;++a)n.add(this.inputs[a],e[a]);return Ep(this.outputs,n,t)})}computeMask(e,t){return P(()=>{e=yt(e);let n;return t==null?n=oi(null,e.length):n=yt(t),this.runInternalGraph(e,n)[1]})}computeOutputShape(e){let t=Th(e);if(t.length!==this.inputLayers.length)throw new V(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let i=0;i<t.length;i++){let o=this.inputLayers[i],l=t[i],u=o.name+"_0_0";n[u]=l}let a=Object.keys(this.nodesByDepth).map(i=>parseInt(i,10)).sort(jd);if(a.length>1)for(let i of a){let o=this.nodesByDepth[i];for(let l of o){let u=l.outboundLayer;if(this.inputLayers.map(m=>m.id).indexOf(u.id)!==-1)continue;let p=[];for(let m=0;m<l.inboundLayers.length;m++){let f=l.inboundLayers[m],g=l.nodeIndices[m],y=l.tensorIndices[m],b=`${f.name}_${g}_${y}`,x=n[b];p.push(x)}let d=u.computeOutputShape(On(p)),c=Th(d),h=u.inboundNodes.indexOf(l);for(let m=0;m<c.length;m++){let f=`${u.name}_${h}_${m}`;n[f]=c[m]}}}let r=[],s=[];for(let i=0;i<this.outputLayers.length;i++){let o=this.outputLayers[i],l=this.outputLayersNodeIndices[i],u=this.outputLayersTensorIndices[i],p=`${o.name}_${l}_${u}`;s.push(p)}for(let i=0;i<s.length;i++){let o=s[i];tr(o in n),r.push(n[o])}return On(r)}runInternalGraph(e,t){t==null&&(t=oi(null,e.length));let n={};for(let o=0;o<this.inputs.length;++o){let l=this.inputs[o],u=e[o],p=t[o];n[l.id]=[u,p]}let a=Object.keys(this.nodesByDepth).map(o=>parseInt(o,10)).sort(jd);for(let o of a){let l=this.nodesByDepth[o];for(let u of l){let p=u.outboundLayer,d=u.inputTensors,c=u.outputTensors,h=new Array;for(let m of d)m.id in n&&h.push(n[m.id]);if(h.length===d.length){let m={},f,g,y,b;if(u.callArgs!=null&&(m=u.callArgs),h.length===1){let[x,w]=h[0];m.mask==null&&(m.mask=w),y=yt(p.call(x,m)),b=yt(p.computeMask(x,w)),f=[x],g=[w]}else f=h.map(x=>x[0]),g=h.map(x=>x[1]),m.mask==null&&(m.mask=g),y=yt(p.call(f,m)),b=yt(p.computeMask(f,g));if(p.activityRegularizer)throw new Re("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let x=0;x<c.length;++x){let w=c[x],I=y[x],T=b[x];n[w.id]=[I,T]}}}}let r=[],s=[],i=[];for(let o of this.outputs){tr(o.id in n,`Could not compute output ${o.name} : ${o.id}`);let[l,u]=n[o.id];i.push(l.shape),r.push(l),s.push(u)}return[r,s,i]}buildNodeConversionMap(e){let t={},n;for(let a of this.layers){n=a instanceof Qa?1:0;for(let r=0;r<a.inboundNodes.length;r++){let s=Qa.nodeKey(a,r);this.containerNodes.has(s)&&(t[s]=n,n+=1)}}return t}getLayer(e,t){if(t!=null){if(this.layers.length<=t)throw new V(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);return this.layers[t]}else if(e==null)throw new V("Provide either a layer name or layer index");for(let n of this.layers)if(n.name===e)return n;throw new V(`No such layer: ${e}`)}calculateLosses(){return P(()=>{let e=[];for(let t of this.layers)for(let n=0;n<t.inboundNodes.length;++n){let a=Qa.nodeKey(t,n);this.containerNodes.has(a)&&e.push(...t.calculateLosses())}return e})}getConfig(){let e={name:this.name},t=this.buildNodeConversionMap(this.layers),n=[];for(let s of this.layers){let i=s.getClassName(),o=s.getConfig(),l=[];for(let p=0;p<s.inboundNodes.length;p++){let d=s.inboundNodes[p],c=Qa.nodeKey(s,p),h={};if(this.containerNodes.has(c)){if(d.callArgs)try{JSON.stringify(d.callArgs),h=d.callArgs}catch(m){console.warn(`Layer ${s.name} was passed non-serializable keyword arguments: ${d.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),h={}}if(d.inboundLayers.length>0){let m=[];for(let f=0;f<d.inboundLayers.length;f++){let g=d.inboundLayers[f],y=d.nodeIndices[f],b=d.tensorIndices[f],x=Qa.nodeKey(g,y),w=t[x];w==null&&(w=0),m.push([g.name,w,b,h])}l.push(m)}}}let u={};u.name=s.name,u.className=i,u.config=o,u.inboundNodes=l,n.push(u)}e.layers=n;let a=[];for(let s=0;s<this.inputLayers.length;s++){let i=this.inputLayers[s],o=this.inputLayersNodeIndices[s],l=Qa.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let p=this.inputLayersTensorIndices[s];a.push([i.name,u,p])}e.inputLayers=a;let r=[];for(let s=0;s<this.outputLayers.length;s++){let i=this.outputLayers[s],o=this.outputLayersNodeIndices[s],l=Qa.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let p=this.outputLayersTensorIndices[s];r.push([i.name,u,p])}return e.outputLayers=r,e}static fromConfig(e,t,n={},a=!1){let r={},s={};function i(f,g){f.name in s?s[f.name].push(g):s[f.name]=[g]}function o(f,g){let y=[],b;for(let x of g){let w=x[0],I=x[1],T=x[2];if(b=x[3]==null?{}:x[3],!(w in r)){i(f,g);return}let C=r[w];if(C.inboundNodes.length<=I){i(f,g);return}let E=C.inboundNodes[I];y.push(E.outputTensors[T])}y.length>0&&f.apply(On(y),b)}function l(f){let g=f.name,y=Ua(f,t.customObjects!=null?t.customObjects:{});y.setFastWeightInitDuringBuild(a),r[g]=y,f.inboundNodes.forEach(b=>{if(!(b instanceof Array))throw new V(`Corrupted configuration, expected array for nodeData: ${b}`);i(y,b)})}let u=t.name,p=t.layers;for(let f of p)l(f);for(;!$4(s);)for(let f of p){let g=r[f.name];if(g.name in s){let y=s[g.name];delete s[g.name];for(let b of y)o(g,b)}}let d=[],c=[],h=t.inputLayers;for(let f of h){let g=f[0],y=f[1],b=f[2];tr(g in r);let x=r[g].inboundNodes[y].outputTensors;d.push(x[b])}let m=t.outputLayers;for(let f of m){let g=f[0],y=f[1],b=f[2];tr(g in r);let x=r[g].inboundNodes[y].outputTensors;c.push(x[b])}return new e({inputs:d,outputs:c,name:u})}get stateful(){if(this._stateful)throw new V("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let e of this.layers)if(e.stateful)return!0;return!1}resetStates(){P(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function aU(e,t,n){let a=t.length;if(e==null||Array.isArray(e)&&e.length===0)return t.map(r=>null);if(a===1)return Array.isArray(e)&&e.length===1?e:typeof e=="object"&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==a)throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${a} outputs. Make sure a set of weights is provided for each model output.`);return e}else if(typeof e=="object"&&Object.keys(e).length>0&&typeof e[Object.keys(e)[0]]=="object"){let r=[];return t.forEach(s=>{s in e?r.push(e[s]):r.push(null)}),r}else throw new Error(`The model has multiple (${a}) outputs, so ${n} must be either an array with ${a} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)}function hN(e,t){return aU(e,t,"classWeight")}async function mN(e,t,n,a){if(t!=null||a!=null)throw new Error("Support sampleWeight is not implemented yet");if(n!=null){let r=P(()=>{if(e.shape.length===1)return rr(e);if(e.shape.length===2){if(e.shape[1]>1)return ai(e,1);if(e.shape[1]===1)return W(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),s=Array.from(await r.data());_e(r);let i=[];return s.forEach(o=>{if(n[o]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${o} exists in the data but not in classWeight`);i.push(n[o])}),Ke(i,"float32")}else return null}function rU(e,t){return z(e,t)}var sU=32;function fN(e,t){let n,a,r=t;n=r.xs,a=r.ys,v.assert(n!=null&&a!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let s=bk("input",e.inputNames,n),i=bk("output",e.outputNames,a),o=s[0].shape[0];v.assert(s.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`),v.assert(i.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let l=0;l<s.length;l++)v.assert(s[l].shape[0]===o,()=>`Batch size mismatch: input ${e.inputNames[l]} has ${s[l].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);for(let l=0;l<i.length;l++)v.assert(i[l].shape[0]===o,()=>`Batch size mismatch: output ${e.outputNames[l]} has ${i[l].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);return{xs:s,ys:i}}function bk(e,t,n){if(n instanceof Te)return[n];if(Array.isArray(n))return v.assert(n.length===t.length,()=>`Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),n;{let a=[];for(let r of t){if(n[r]==null)throw new V(`The feature data generated by the dataset lacks the required ${e} key '${r}'.`);a.push(n[r])}return a}}function iU(e){if(e.length===3)throw new Re("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}async function oU(e,t,n){let a=n.batchesPerEpoch!=null;if(v.assert(e.optimizer!=null,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),v.assert(n!=null,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),v.assert(n.epochs!=null&&n.epochs>0&&Number.isInteger(n.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${n.epochs}`),v.assert(!a||n.batchesPerEpoch>0&&Number.isInteger(n.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${n.batchesPerEpoch}`),v.assert(n.validationSplit==null,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{let r=n.validationData!=null,s,i;if(r)if(xk(n.validationData))v.assert(n.validationBatches==null||n.validationBatches>0&&Number.isInteger(n.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${n.validationBatches}`);else{let g=iU(n.validationData);s=g.xs,i=g.ys}let o=e.makeTrainFunction(),l=e.getDedupedMetricsNames(),u;r?u=l.slice().concat(l.map(g=>"val_"+g)):u=l.slice();let p=sN(n.callbacks,n.yieldEvery),d=n.verbose==null?1:n.verbose,{callbackList:c,history:h}=iN(p,d,n.epochs,null,null,lU(t,n),null,r,u);c.setModel(e),e.history=h,await c.onTrainBegin(),e.stopTraining_=!1;let m=n.initialEpoch==null?0:n.initialEpoch,f=await t.iterator();for(;m<n.epochs;){let g={};await c.onEpochBegin(m);let y=0,b=0;for(a||(f=await t.iterator());!a||y<n.batchesPerEpoch;){let x=await f.next();if(a&&x.done){console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${y} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(x.value!=null){let{xs:w,ys:I}=fN(e,x.value),T={};T.batch=b,T.size=w[0].shape[0],await c.onBatchBegin(b,T);let C=[];if(n.classWeight!=null){let R=hN(n.classWeight,e.outputNames);for(let F=0;F<R.length;++F)C.push(await mN(I[F],null,R[F]))}let E=w.concat(I).concat(C),A=o(E);_e(E);for(let R=0;R<l.length;++R){let F=l[R],S=A[R];T[F]=S,Jt(S)}await c.onBatchEnd(b,T),tN(T),b++,y++}if(a?y>=n.batchesPerEpoch:x.done){if(r){let w;xk(n.validationData)?w=yt(await e.evaluateDataset(n.validationData,{batches:n.validationBatches})):w=yt(e.evaluate(s,i,{batchSize:n.validationBatchSize==null?sU:n.validationBatchSize,verbose:0}));for(let I=0;I<e.metricsNames.length;++I)g[`val_${e.metricsNames[I]}`]=w[I]}break}if(e.stopTraining_)break}if(await c.onEpochEnd(m,g),m++,e.stopTraining_)break}return await c.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function lU(e,t){let n=null;return t.batchesPerEpoch!=null?n=t.batchesPerEpoch:Number.isFinite(e.size)&&(n=e.size),n}function xk(e){return typeof e.iterator=="function"}function uU(e){return typeof e.next=="function"}async function pU(e,t,n){n=n||{};let a=n.batches!=null,r=e.testFunction,s=[];if(n.verbose>0)throw new Re("Verbose mode is not implemented yet.");v.assert(!a||n.batches>0&&Number.isInteger(n.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(n.batches)}`);let i=uU(t)?t:await t.iterator(),o=0,l=0;for(;!a||l<n.batches;){let u=await i.next();if(s=P(()=>{if(u.value){let{xs:p,ys:d}=fN(e,u.value),c=p.concat(d),h=P(()=>r(c));if(_e(c),l===0)for(let f=0;f<h.length;++f)s.push(be(0));let m=c[0].shape[0];for(let f=0;f<h.length;++f){let g=h[f],y=s[f];s[f]=P(()=>Y(s[f],z(m,g))),l>0&&_e(y)}_e(h),o+=m,++l}return s}),u.done){a&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let u=0;u<s.length;++u){let p=s[u];s[u]=he(s[u],o),_e(p)}return On(s)}function Vb(e){v.assert(e>0&&Number.isInteger(e),()=>`batchSize is required to be a positive integer, but got ${e}`)}function Ap(e,t,n){return e==null?[null]:Array.isArray(e)?e.map(a=>Ys(a,t,n-t)):Ys(e,t,n-t)}function hw(e,t){return P(()=>e==null?null:Array.isArray(e)?e.map(n=>hw(n,t)):qT(e,t.dtype==="int32"?t:oe(t,"int32")))}function Ub(e,t){let n=[],a=0,r=null;for(;a<e;)r=a+t,r>=e&&(r=e),n.push([a,r]),a=r;return n}async function cU(e,t,n,a,r,s,i,o,l,u,p,d,c,h,m){r==null&&(r=32),s==null&&(s=1),p==null&&(p=!0),c==null&&(c=0);let f=!1;if(l!=null&&u!=null&&(f=!0),m!=null&&(f=!0,h==null))throw new V("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let g=e.checkNumSamples(n,r,h,"steps_per_epoch"),y;g!=null&&(y=Ha(0,g)),i==null&&(i=1);let{callbackList:b,history:x}=iN(o,i,s,c,g,h,r,f,d);b.setModel(e),e.history=x,await b.onTrainBegin(),e.stopTraining_=!1;for(let w=c;w<s;++w){await b.onEpochBegin(w);let I={};if(h!=null)throw new Re("stepsPerEpoch mode is not implemented yet.");{if(p==="batch")throw new Re("batch shuffling is not implemneted yet");p&&v.shuffle(y);let T=Ke(y),C=Ub(g,r);for(let E=0;E<C.length;++E){let A={};if(await b.onBatchBegin(E,A),P(()=>{let R=C[E][0],F=C[E][1],S=Ys(T,R,F-R);A.batch=E,A.size=F-R;let M=hw(n,S),B=t(M);for(let U=0;U<a.length;++U){let G=a[U],q=B[U];A[G]=q,Jt(q)}if(E===C.length-1&&f){let U=e.testLoop(l,u,r);for(let G=0;G<a.length;++G){let q=a[G],K=U[G];Jt(K),I["val_"+q]=K}}}),await b.onBatchEnd(E,A),tN(A),e.stopTraining_)break}T.dispose()}if(await b.onEpochEnd(w,I),e.stopTraining_)break}return await b.onTrainEnd(),await e.history.syncData(),e.history}async function dU(e,t,n,a={}){if(e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;let r,s,i,o,l,u,p,d,c;try{let h=a.batchSize==null?32:a.batchSize;Vb(h);let m=!1,f=await e.standardizeUserData(t,n,a.sampleWeight,a.classWeight,m,h);r=f[0],s=f[1],c=f[2];let g=!1,y;if(a.validationData!=null&&a.validationData.length>0){if(g=!0,a.validationData.length===2)l=a.validationData[0],u=a.validationData[1];else throw a.validationData.length===3?new Re("validationData including sample weights is not supported yet."):new V(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${a.validationData} is invalid.`);let E=!0,A=await e.standardizeUserData(l,u,null,null,E,h);p=A[0],d=A[1],y=p.concat(d)}else if(a.validationSplit!=null&&a.validationSplit>0&&a.validationSplit<1){g=!0;let E=Math.floor(r[0].shape[0]*(1-a.validationSplit)),A=r[0].shape[0];p=Ap(r,E,A),i=r,r=Ap(r,0,E),d=Ap(s,E,A),o=s,s=Ap(s,0,E),y=p.concat(d)}else a.validationSteps!=null&&(g=!0);let b=r.concat(s).concat(c);e.checkTrainableWeightsConsistency();let x=e.makeTrainFunction(),w=e.getDedupedMetricsNames(),I,T;g?(e.makeTestFunction(),I=e.testFunction,T=w.slice().concat(w.map(E=>"val_"+E))):(I=null,y=[],T=w.slice());let C=sN(a.callbacks,a.yieldEvery);return await cU(e,x,b,w,h,a.epochs,a.verbose,C,I,y,a.shuffle,T,a.initialEpoch,null,null)}finally{e.isTraining=!1,La(r,t),La(s,n),La(i,t),La(o,n),La(p,l),La(d,u),c!=null&&_e(c)}}function gN(e){let t=[];e instanceof Te&&(e=[e]);for(let n=0;n<e.length;++n){let a=e[n];if(a.rank===1)t.push(Uc(a,1));else{if(a.rank===0)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(a)}}return t}function La(e,t){if(e==null)return;let n=[];if(t instanceof Te)n.push(t.id);else if(Array.isArray(t))t.forEach(r=>n.push(r.id));else if(t!=null)for(let r in t){let s=t[r];n.push(s.id)}let a=[];if(e instanceof Te)n.indexOf(e.id)===-1&&a.push(e);else if(Array.isArray(e))e.forEach(r=>{n.indexOf(r.id)===-1&&a.push(r)});else if(e!=null)for(let r in e){let s=e[r];n.indexOf(s.id)===-1&&a.push(s)}a.forEach(r=>{r.isDisposed||r.dispose()})}function hU(e){return e instanceof Te}function Gb(e){return Array.isArray(e)}function vk(e){return!hU(e)&&!Gb(e)}function wk(e,t,n,a=!0,r=""){if(t==null||t.length===0){if(e!=null){let i=!1;if(Gb(e)&&e.length>0)i=!0;else if(vk(e)){for(let o in e)if(e.hasOwnProperty(o)){i=!0;break}}else i=!0;if(i)throw new V(`Error when checking model ${r} expected no data, but got ${e}`)}return[]}if(e==null)return t.map(i=>null);let s;if(vk(e)){e=e,s=[];for(let i of t){if(e[i]==null)throw new V(`No data provided for "${i}". Need data for each key in: ${t}`);s.push(e[i])}}else if(Gb(e)){if(e=e,e.length!==t.length)throw new V(`Error when checking model ${r}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);s=e}else{if(e=e,t.length>1)throw new V(`The model ${r} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);s=[e]}if(s=gN(s),n!=null)for(let i=0;i<t.length;++i){if(n[i]==null)continue;let o=s[i];if(o.shape.length!==n[i].length)throw new V(`Error when checking ${r}: expected ${t[i]} to have ${n[i].length} dimension(s). but got array with shape ${o.shape}`);for(let l=0;l<n[i].length;++l){if(l===0&&!a)continue;let u=o.shape[l],p=n[i][l];if(p!=null&&p>=0&&u!==p)throw new V(`${r} expected a batch of elements where each example has shape [${n[i].slice(1,n[i].length)}] (i.e.,tensor shape [*,${n[i].slice(1,n[i].length)}]) but the ${r} received an input with ${o.shape[0]} examples, each with shape [${o.shape.slice(1,o.shape.length)}] (tensor shape [${o.shape}])`)}}return s}function mU(e,t,n){let a=Jr(e.map(s=>s.shape[0]));a.sort();let r=Jr(t.map(s=>s.shape[0]));if(r.sort(),a.length>1)throw new V(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(e.map(s=>s.shape))}`);if(r.length>1)throw new V(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(s=>s.shape))}`);if(a.length>0&&r.length>0&&!v.arraysEqual(a,r))throw new V(`Input Tensors should have the same number of samples as target Tensors. Found ${a[0]} input sample(s) and ${r[0]} target sample(s).`)}function fU(e,t,n){let a=[bo,Sf,Xp];for(let r=0;r<e.length;++r){let s=e[r],i=t[r],o=n[r];if(i!=null){if(i===Xp&&s.shape[s.shape.length-1]===1)throw new V(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(a.indexOf(i)!==-1){let l=s.shape.slice(1),u=o.slice(1);for(let p=0;p<l.length;++p){let d=l[p],c=u[p];if(c!=null&&d!==c)throw new V(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function kk(e,t,n,a=!0,r=""){let s;if(Array.isArray(e)){if(e.length!==t.length)throw new V(`Error when checking model ${r}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);s=e}else{if(t.length>1)throw new V(`The model expects ${t.length} ${r} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);s=[e]}if(n!=null)for(let i=0;i<t.length;++i){if(n[i]==null)continue;let o=s[i];if(o.shape.length!==n[i].length)throw new V(`Error when checking ${r}: expected ${t[i]} to have ${n[i].length} dimension(s), but got array with shape ${JSON.stringify(o.shape)}`);for(let l=0;l<n[i].length;++l){if(l===0&&!a)continue;let u=o.shape[l],p=n[i][l];if(p!=null&&p!==u)throw new V(`Error when checking ${r}: expected ${t[i]} to have shape ${JSON.stringify(n[i])} but got array with shape ${JSON.stringify(o.shape)}.`)}}}function gU(e,t){if(e==null||Array.isArray(e)&&e.length===0)return t.map(a=>[]);let n;if(typeof e=="string"||typeof e=="function")n=[e];else if(Array.isArray(e)||typeof e=="object")n=e;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${e}`);if(Array.isArray(n))return t.map(a=>n);{let a=[];for(let r of t){let s=n.hasOwnProperty(r)?n[r]:[];Array.isArray(s)||(s=[s]),a.push(s)}return a}}var yU="layers-model",Nr=class extends Qa{constructor(e){super(e),this.isTraining=!1}summary(e,t,n=console.log){if(!this.built)throw new V("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");JV(this,e,t,n)}compile(e){if(e.loss==null&&(e.loss=[]),this.loss=e.loss,typeof e.optimizer=="string")this.optimizer_=ZV(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof $r))throw new V("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(!Array.isArray(e.loss)&&typeof e.loss!="string"&&typeof e.loss!="function"){e.loss=e.loss;for(let s in e.loss)if(this.outputNames.indexOf(s)===-1)throw new V(`Unknown entry in loss dictionary: "${s}". Only expected the following keys: ${this.outputNames}`);for(let s of this.outputNames)e.loss[s]==null&&console.warn(`Output "${s}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${s} during training`),t.push(db(e.loss[s]))}else if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new V(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);t=e.loss.map(s=>db(s))}else{let s=db(e.loss);this.outputs.forEach(i=>{t.push(s)})}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let s=0;s<this.outputs.length;++s){let i=this.internalOutputShapes[s],o=this.outputNames[s];this.feedOutputNames.push(o),this.feedOutputShapes.push(i),this.feedLossFns.push(this.lossFunctions[s])}let n=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Xs("loss",()=>{for(let s=0;s<this.outputs.length;++s){if(n.indexOf(s)!==-1)continue;let i=this.lossFunctions[s];this.outputs.length>1&&(this.metricsTensors.push([i,s]),this.metricsNames.push(this.outputNames[s]+"_loss"))}});let a=gU(e.metrics,this.outputNames),r=(s,i,o)=>{this.outputNames.length>1&&(i=this.outputNames[s]+"_"+i),this.metricsNames.push(i),this.metricsTensors.push([o,s])};Xs("metric",()=>{for(let s=0;s<this.outputs.length;++s){if(n.indexOf(s)!==-1)continue;let i=a[s];(o=>{let l="",u,p,d;for(let c of o){if(typeof c=="string"&&["accuracy","acc","crossentropy","ce"].indexOf(c)!==-1){let m=this.internalOutputShapes[s];m[m.length-1]===1||this.lossFunctions[s]===Sf?["accuracy","acc"].indexOf(c)!==-1?p=uw:["crossentropy","ce"].indexOf(c)!==-1&&(p=uN):this.lossFunctions[s]===Ah?["accuracy","acc"].indexOf(c)!==-1?p=pN:["crossentropy","ce"].indexOf(c)!==-1&&(p=cN):["accuracy","acc"].indexOf(c)!==-1?p=pw:["crossentropy","ce"].indexOf(c)!==-1&&(p=cw);let f;["accuracy","acc"].indexOf(c)!==-1?f="acc":["crossentropy","ce"].indexOf(c)!==-1&&(f="ce"),d=p,u=l+f}else d=YV(c),u=l+Xd(c);let h;Xs(u,()=>{h=d}),r(s,u,h)}})(i)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,n={}){let a=n.batchSize==null?32:n.batchSize;Vb(a);let r=!0,s=this.standardizeUserDataXY(e,t,r,a);try{let i=s[0].concat(s[1]);this.makeTestFunction();let o=this.testFunction,l=this.testLoop(o,i,a,n.verbose,n.steps);return On(l)}finally{La(s[0],e),La(s[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),pU(this,e,t)}checkNumSamples(e,t,n,a="steps"){let r;if(n!=null){if(r=null,t!=null)throw new V(`If ${a} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else if(e!=null)Array.isArray(e)?r=e[0].shape[0]:r=e.shape[0];else throw new V(`Either the input data should have a defined shape, or ${a} shoud be specified.`);return r}execute(e,t){if(Array.isArray(t)&&t.length===0)throw new V("`outputs` is an empty Array, which is not allowed.");let n=Array.isArray(t),a=n?t:[t],r=this.retrieveSymbolicTensors(a),s=new Hs;if(e instanceof Te&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new V(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let o=0;o<this.inputs.length;++o)s.add(this.inputs[o],e[o])}else for(let o of this.inputs){let l=e[o.name];if(l==null)throw new V(`No value is provided for the model's input ${o.name}`);s.add(o,l)}let i=Ep(r,s);return n?i:i[0]}retrieveSymbolicTensors(e){let t=oi(null,e.length),n=e.length;for(let a of this.layers){let r=Array.isArray(a.output)?a.output:[a.output],s=r.map(i=>i.name);for(let i=0;i<e.length;++i){let o=s.indexOf(e[i]);if(o!==-1&&(t[i]=r[o],n--),n===0)break}if(n===0)break}if(n>0){let a=[];throw t.forEach((r,s)=>{r==null&&a.push(e[s])}),new V(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(a)}`)}return t}predictLoop(e,t=32,n=!1){return P(()=>{let a=this.checkNumSamples(e);if(n)throw new Re("Verbose predictLoop() is not implemented yet.");let r=Ub(a,t),s=this.outputs.map(i=>[]);for(let i=0;i<r.length;++i)P(()=>{let o=r[i][0],l=r[i][1],u=Ap(e,o,l),p=[];if(Array.isArray(u))for(let c=0;c<u.length;++c)p.push({key:this.inputs[c],value:u[c]});else p.push({key:this.inputs[0],value:u});let d=new Hs(p);return Ep(this.outputs,d)}).forEach((o,l)=>s[l].push(o));return On(s.map(i=>Ze(i,0)))})}predict(e,t={}){let n=gN(e);kk(n,this.inputNames,this.feedInputShapes,!1);try{let a=t.batchSize==null?32:t.batchSize;return Vb(a),this.predictLoop(n,a)}finally{La(n,e)}}predictOnBatch(e){kk(e,this.inputNames,this.feedInputShapes,!0);let t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,n=!0,a){if(this.optimizer_==null)throw new Wa("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let r=[];for(let s=0;s<this.feedOutputShapes.length;++s){let i=this.feedOutputShapes[s];this.feedLossFns[s]===Ah?r.push(i.slice(0,i.length-1).concat([1])):r.push(i)}if(e=wk(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=wk(t,this.feedOutputNames,r,!1,"target"),mU(e,t,null),fU(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&a!=null&&a>0&&e[0].shape[0]%a!==0)throw new V(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${a}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,n,a,r=!0,s){let[i,o]=this.standardizeUserDataXY(e,t,r,s);if(n!=null)throw new Error("sample weight is not supported yet.");let l=null;if(a!=null){let u=hN(a,this.outputNames);l=[];for(let p=0;p<u.length;++p)l.push(await mN(o[p],null,u[p]))}return[i,o,l]}testLoop(e,t,n,a=0,r){return P(()=>{let s=this.checkNumSamples(t,n,r,"steps"),i=[];if(a>0)throw new Re("Verbose mode is not implemented yet.");if(r!=null)throw new Re("steps mode in testLoop() is not implemented yet");{let o=Ub(s,n),l=Ke(Ha(0,s));for(let u=0;u<o.length;++u){let p=o[u][0],d=o[u][1],c=Ys(l,p,d-p),h=hw(t,c),m=e(h);if(u===0)for(let f=0;f<m.length;++f)i.push(be(0));for(let f=0;f<m.length;++f){let g=m[f];i[f]=Y(i[f],z(d-p,g))}}for(let u=0;u<i.length;++u)i[u]=he(i[u],s)}return i})}getDedupedMetricsNames(){let e=this.metricsNames,t=[];for(let n=0;n<e.length;++n){let a=e[n],r=a;ok(e,a)>1&&(r+=`_${ok(e.slice(0,n),a)}`),t.push(r)}return t}makeTrainFunction(){return e=>{let t=[],n=e.slice(0,this.inputs.length),a=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),r=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),s=[],i=()=>{let u=[];for(let h=0;h<this.inputs.length;++h)u.push({key:this.inputs[h],value:n[h]});let p=new Hs(u),d=Ep(this.outputs,p,{training:!0}),c;for(let h=0;h<this.lossFunctions.length;++h){let m=this.lossFunctions[h],f=m(a[h],d[h]);r[h]!=null&&(f=rU(f,r[h]));let g=Nt(f);t.push(g),h===0?c=f:c=Y(c,f)}for(let h=0;h<this.metricsTensors.length;++h){let m;if(this.outputs.length>1&&h<this.outputs.length)m=t[h];else{let f=this.metricsTensors[h][0],g=this.metricsTensors[h][1];m=Nt(f(a[g],d[g]))}Jt(m),s.push(m)}return c=Nt(c),this.calculateLosses().forEach(h=>{c=Y(c,h)}),c},o=this.collectedTrainableWeights.map(u=>u.read()),l=!0;return[this.optimizer_.minimize(i,l,o)].concat(s)}}makeTestFunction(){this.testFunction=e=>P(()=>{let t=[],n,a=e.slice(0,this.inputs.length),r=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=[];for(let l=0;l<this.inputs.length;++l)s.push({key:this.inputs[l],value:a[l]});let i=new Hs(s),o=Ep(this.outputs,i);for(let l=0;l<this.lossFunctions.length;++l){let u=this.lossFunctions[l],p=Nt(u(r[l],o[l]));l===0?n=p:n=Y(n,p),t.push(n)}for(let l=0;l<this.metricsTensors.length;++l){let u=this.metricsTensors[l][0],p=this.metricsTensors[l][1],d=Nt(u(r[p],o[p]));t.push(d)}return t})}async fit(e,t,n={}){return dU(this,e,t,n)}async fitDataset(e,t){return oU(this,e,t)}async trainOnBatch(e,t){let n=await this.standardizeUserData(e,t),a=n[0],r=n[1],s=this.makeTrainFunction()(a.concat(r)),i=[];for(let o of s){let l=await o.data();i.push(l[0])}return _e(s),La(n[0],e),La(n[1],t),On(i)}getNamedWeights(e){let t=[],n=e!=null&&e.trainableOnly,a=n?this.trainableWeights:this.weights,r=this.getWeights(n);for(let s=0;s<a.length;++s)n&&!a[s].trainable||t.push({name:a[s].originalName,tensor:r[s]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){let e=super.dispose();if(e.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let t=kh().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-kh().numTensors}return e}getLossIdentifiers(){let e;if(typeof this.loss=="string")e=wr(this.loss);else if(Array.isArray(this.loss)){for(let t of this.loss)if(typeof t!="string")throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(t=>wr(t))}else{let t=Object.keys(this.loss);e={};let n=this.loss;for(let a of t)if(typeof n[a]=="string")e[a]=wr(n[a]);else throw new Error("Serialization of non-string loss is not supported.")}return e}getMetricIdentifiers(){if(typeof this.metrics=="string"||typeof this.metrics=="function")return[wr(Xd(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>wr(Xd(e)));{let e={};for(let t in this.metrics)e[t]=wr(Xd(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(e.weighted_metrics!=null)throw new Error("Loading weight_metrics is not supported yet.");if(e.loss_weights!=null)throw new Error("Loading loss_weights is not supported yet.");if(e.sample_weight_mode!=null)throw new Error("Loading sample_weight_mode is not supported yet.");let t=Yp(e.optimizer_config),n=Ua(t),a;if(typeof e.loss=="string")a=Vs(e.loss);else if(Array.isArray(e.loss))a=e.loss.map(s=>Vs(s));else if(e.loss!=null){a={};for(let s in e.loss)a[s]=Vs(e.loss[s])}let r;if(Array.isArray(e.metrics))r=e.metrics.map(s=>Vs(s));else if(e.metrics!=null){r={};for(let s in e.metrics)r[s]=Vs(e.metrics[s])}this.compile({loss:a,metrics:r,optimizer:n})}async save(e,t){if(typeof e=="string"){let i=Ut.getSaveHandlers(e);if(i.length===0)throw new V(`Cannot find any save handlers for URL '${e}'`);if(i.length>1)throw new V(`Found more than one (${i.length}) save handlers for URL '${e}'`);e=i[0]}if(e.save==null)throw new V("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let n=await Ut.encodeWeights(this.getNamedWeights(t)),a=!1,r=null,s={modelTopology:this.toJSON(r,a),format:yU,generatedBy:`TensorFlow.js tfjs-layers v${dw}`,convertedBy:null};if((t==null?!1:t.includeOptimizer)&&this.optimizer!=null){s.trainingConfig=this.getTrainingConfig();let i="optimizer",{data:o,specs:l}=await Ut.encodeWeights(await this.optimizer.getWeights(),i);n.specs.push(...l),n.data=Ut.concatenateArrayBuffers([n.data,o])}return this.userDefinedMetadata!=null&&(yk(this.userDefinedMetadata,this.name,!0),s.userDefinedMetadata=this.userDefinedMetadata),s.weightData=n.data,s.weightSpecs=n.specs,e.save(s)}setUserDefinedMetadata(e){yk(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Nr.className="Model";ne.registerClass(Nr);var yN=class extends Nr{};yN.className="Functional";ne.registerClass(yN);async function bU(e,t){"modelTopology"in e||(e={modelTopology:e}),e=e;let n=e.modelTopology;n.model_config!=null&&(n=n.model_config);let a=Yp(n),r=Ua(a,t);if(e.weightsManifest!=null){let s=await Ut.loadWeights(e.weightsManifest,e.pathPrefix,r.weights.map(o=>o.originalName)),i={};for(let o of r.weights)i[o.originalName]=s[o.originalName];r.loadWeights(i),_e(s)}return r}async function xU(e,t){if(t==null&&(t={}),typeof e=="string"){let n=Ut.getLoadHandlers(e,t);if(n.length===0)n.push(Ut.browserHTTPRequest(e,t));else if(n.length>1)throw new V(`Found more than one (${n.length}) load handlers for URL '${e}'`);e=n[0]}return vU(e,void 0,t)}async function vU(e,t,n){if(n==null&&(n={}),e.load==null)throw new V("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let a=await e.load(),r=a.modelTopology;r.model_config!=null&&(r=r.model_config);let s=n.strict==null?!0:n.strict,i=a.weightData!=null&&a.weightSpecs!=null&&s,o=Ua(Yp(r),t,i),l=a.trainingConfig;if(l!=null&&o.loadTrainingConfig(l),a.userDefinedMetadata!=null&&o.setUserDefinedMetadata(a.userDefinedMetadata),a.weightData!=null){if(a.weightSpecs==null)throw new V("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:u,optimizerWeights:p}=wU(a.weightData,a.weightSpecs);o.loadWeights(u,s),o.optimizer!=null&&p.length>0&&await o.optimizer.setWeights(p),_e(u),_e(p.map(d=>d.tensor))}return o}function wU(e,t){let n=Ut.decodeWeights(e,t),a={},r=[];return t.forEach(s=>{s.group==="optimizer"?r.push({name:s.name,tensor:n[s.name]}):a[s.name]=n[s.name]}),{modelWeights:a,optimizerWeights:r}}var hl=class extends Nr{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=e.name!=null?e.name:hf("sequential_"),e.layers!=null)for(let t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(t=>t<0))throw new V(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(e){let t=e instanceof hl||e instanceof Nr,n;if(t){if(n=e,n.outputs.length!==1)throw new V("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(n.inputs.length!==1)throw new V("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(this.outputs.length===0){if(e.inboundNodes.length===0){if(e.batchInputShape==null)throw new V("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let a=ZT({batchShape:e.batchInputShape,dtype:e.dtype,name:e.name+"_input"});e.apply(a)}if(t)this.outputs=n.outputs,this.inputs=n.inputs;else{if(e.inboundNodes.length!==1)throw new V(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);if(e.inboundNodes[0].outputTensors.length!==1)throw new V("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[e.inboundNodes[0].outputTensors[0]],this.inputs=YT(this.outputs[0])}this.inboundNodes=[],new kf({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:oi(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(a=>a.shape),outputShapes:this.outputs[0].shape})}else{let a=e.apply(this.outputs[0]);if(Array.isArray(a))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[a],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(e),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return this.model==null&&this.build(),this.model.call(e,t)}build(e){if(Qe(e),this.inputs.length===0||this.outputs.length===0)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Nr({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,n=console.log){this.built||this.build(),super.summary(e,t,n)}setWeights(e){this.model==null&&this.build(),this.model.setWeights(e)}evaluate(e,t,n={}){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.evaluate(e,t,n)}async evaluateDataset(e,t){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return this.model==null&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return this.model==null&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,n={}){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.fit(e,t,n)}async fitDataset(e,t){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(e,t,n={},a=!1){let r,s={};if(t instanceof Array){if(t[0].className==null||t[0].className==="Merge")throw new V("Legacy serialization format not supported yet.");r=t}else v.assert(t.layers!=null,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),r=t.layers,delete t.layers,s=t;let i=new e(s);if(!(i instanceof hl))throw new Re(`Sequential.fromConfig called on non-Sequential input: ${i}`);for(let o of r){let l=Ua(o,void 0,a);a&&l.setFastWeightInitDuringBuild(!0),i.add(l)}return i}set stopTraining(e){if(this.model==null)throw new V("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(this.model==null)throw new V("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let e=[];for(let t of this.layers){let n={};n.className=t.getClassName(),n.config=t.getConfig(),e.push(n)}return{name:this.name,layers:e}}};hl.className="Sequential";ne.registerClass(hl);function kU(e){return new Nr(e)}function IU(e){return new hl(e)}function bN(e){return ZT(e)}function SU(e,t){Ta.registerCallbackConstructor(e,t)}var Un=class extends ne.Serializable{getConfig(){return{}}},xN=class extends Un{apply(e,t=1){return j4(e,t)}};xN.className="elu";ne.registerClass(xN);var vN=class extends Un{apply(e){return qm(e)}};vN.className="selu";ne.registerClass(vN);var wN=class extends Un{apply(e){return Xe(e)}};wN.className="relu";ne.registerClass(wN);var kN=class extends Un{apply(e){return P(()=>Au(6,Xe(e)))}};kN.className="relu6";ne.registerClass(kN);var IN=class extends Un{apply(e){return e}};IN.className="linear";ne.registerClass(IN);var SN=class extends Un{apply(e){return da(e)}};SN.className="sigmoid";ne.registerClass(SN);var TN=class extends Un{apply(e){return K4(e)}};TN.className="hardSigmoid";ne.registerClass(TN);var NN=class extends Un{apply(e){return ho(e)}};NN.className="softplus";ne.registerClass(NN);var CN=class extends Un{apply(e){return q4(e)}};CN.className="softsign";ne.registerClass(CN);var _N=class extends Un{apply(e){return ri(e)}};_N.className="tanh";ne.registerClass(_N);var mw=class extends Un{apply(e,t=-1){return Ka(e,t)}};mw.className="softmax";ne.registerClass(mw);var EN=class extends Un{apply(e,t=-1){return zm(e,t)}};EN.className="logSoftmax";ne.registerClass(EN);var AN=class extends Un{apply(e,t=1){return P(()=>z(da(z(e,t)),e))}};AN.className="swish";ne.registerClass(AN);var $N=class extends Un{apply(e){return P(()=>z(e,ri(ho(e))))}};$N.className="mish";ne.registerClass($N);function rs(e){return e.getClassName()}function hb(e,t={}){return Vc(e,ne.SerializationMap.getMap().classNameMap,t,"activation")}function ss(e){if(e==null){let t={};return t.className="linear",t.config={},hb(t)}if(typeof e=="string"){let t={};return t.className=e,t.config={},hb(t)}else return e instanceof Un?e:hb(e)}function fw(e){if(e!=null&&typeof e!="object")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${e}`)}var FN=class extends ne.Serializable{},qc=class extends FN{constructor(e){super(),fw(e),this.l1=e==null||e.l1==null?.01:e.l1,this.l2=e==null||e.l2==null?.01:e.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(e){return P(()=>{let t=It([1]);return this.hasL1&&(t=Y(t,fe(z(this.l1,Lt(e))))),this.hasL2&&(t=Y(t,fe(z(this.l2,Gc(e))))),W(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};qc.className="L1L2";ne.registerClass(qc);function TU(e){return fw(e),new qc({l1:e!=null?e.l1:null,l2:0})}function NU(e){return fw(e),new qc({l2:e!=null?e.l2:null,l1:0})}var Ik={l1l2:"L1L2"};function pt(e){return jv(e)}function Sk(e,t={}){return Vc(e,ne.SerializationMap.getMap().classNameMap,t,"regularizer")}function Tt(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in Ik?Ik[e]:e,config:{}};return Sk(t)}else return e instanceof FN?e:Sk(e)}var gw=class extends Ge{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null&&(this.maxValue=e.maxValue)}call(e,t){e=Ne(e);let n=Xe(e);return this.maxValue!=null&&(n=en(n,0,this.maxValue)),n}computeOutputShape(e){return e}getConfig(){let e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};gw.className="ReLU";ne.registerClass(gw);var yw=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_ALPHA=.3,e==null&&(e={}),this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ne(e);return Fc(n,this.alpha)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};yw.className="LeakyReLU";ne.registerClass(yw);var bw=class extends Ge{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA_INITIALIZER="zeros",e==null&&(e={}),this.supportsMasking=!0,this.alphaInitializer=St(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=Tt(e.alphaRegularizer),this.alphaConstraint=Kt(e.alphaConstraint),e.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else if(typeof e.sharedAxes=="number")this.sharedAxes=[e.sharedAxes];else throw new V(`Expected sharedAxes to be a number or an array of numbers, but got ${e.sharedAxes}`)}build(e){e=Qe(e);let t=e.slice(1);if(this.sharedAxes!=null)for(let a of this.sharedAxes)t[a-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(this.sharedAxes!=null)for(let a=1;a<e.length;++a)n[a]=e[a];this.inputSpec=[new zt({ndim:e.length,axes:n})],this.built=!0}call(e,t){return e=Ne(e),Oc(e,this.alpha.read())}getConfig(){let e={alphaInitializer:Ct(this.alphaInitializer),alphaRegularizer:pt(this.alphaRegularizer),alphaConstraint:qt(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};bw.className="PReLU";ne.registerClass(bw);var xw=class extends Ge{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA=1,e==null&&(e={}),e.alpha!=null&&e.alpha!==this.DEFAULT_ALPHA)throw new Re(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ne(e);return Nu(n)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};xw.className="ELU";ne.registerClass(xw);var vw=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_THETA=1,e==null&&(e={}),this.theta=e.theta==null?this.DEFAULT_THETA:e.theta}call(e,t){let n=Ne(e);return z(n,oe(Nn(n,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){let e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};vw.className="ThresholdedReLU";ne.registerClass(vw);var ww=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_AXIS=1,e==null&&(e={}),this.softmax=new mw().apply,this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis}call(e,t){let n=Ne(e);return this.softmax(n,this.axis)}computeOutputShape(e){return e}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};ww.className="Softmax";ne.registerClass(ww);function el(e,t,n){if(typeof e=="number")return oi(e,t);if(e.length!==t)throw new V(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let a=0;a<t;++a){let r=e[a];if(!V4(r))throw new V(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${r}`)}return e}function Ga(e,t,n,a,r=1){if(e==null)return e;let s=t+(t-1)*(r-1),i;return n==="same"?i=e:i=e-s+1,Math.floor((i+a-1)/a)}function nr(e,t,n,a){if(e==null)return null;if(a==="valid")e=e*t+as([n-t,0]);else if(a==="same")e=e*t;else throw new V(`Unsupport padding mode: ${a}.`);return e}function kw(e,t){return P(()=>(Rt(t),t==="channelsFirst"?Ee(e,[0,2,3,1]):e))}function DN(e,t){return P(()=>(Rt(t),t==="channelsFirst"?Ee(e,[0,2,3,4,1]):e))}function CU(e,t,n,a=1,r="valid",s,i=1){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.shape.length!==3)throw new V(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(t.shape.length!==3)throw new V(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(n!=null&&n.shape.length!==1)throw new V(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(s==="channelsFirst"&&(e=Ee(e,[0,2,1])),r==="causal")throw new Re("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let o=Fm(e,t,a,r==="same"?"same":"valid","NWC",i);return n!=null&&(o=Xa(o,n)),o})}function Tk(e,t,n,a=[1,1],r="valid",s,i,o=null){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.rank!==3&&e.rank!==4)throw new V(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(t.rank!==3&&t.rank!==4)throw new V(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let l=kw(e,s);if(r==="causal")throw new Re("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return l=pl.conv2d({x:l,filter:t,strides:a,pad:r==="same"?"same":"valid",dilations:i,dataFormat:"NHWC",bias:n,activation:o}),s==="channelsFirst"&&(l=Ee(l,[0,3,1,2])),l})}function _U(e,t,n,a=[1,1,1],r="valid",s,i){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.rank!==4&&e.rank!==5)throw new V(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(t.rank!==4&&t.rank!==5)throw new V(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let o=DN(e,s);if(r==="causal")throw new Re("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return o=ov(o,t,a,r==="same"?"same":"valid","NDHWC",i),n!=null&&(o=Xa(o,n)),s==="channelsFirst"&&(o=Ee(o,[0,4,1,2,3])),o})}var Iw=class extends Ge{constructor(e,t){if(super(t),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",Iw.verifyArgs(t),this.rank=e,Qt(this.rank,"rank"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new Re(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=el(t.kernelSize,e,"kernelSize"),this.strides=el(t.strides==null?1:t.strides,e,"strides"),this.padding=t.padding==null?"valid":t.padding,ba(this.padding),this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Rt(this.dataFormat),this.activation=ss(t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.biasInitializer=St(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Kt(t.biasConstraint),this.biasRegularizer=Tt(t.biasRegularizer),this.activityRegularizer=Tt(t.activityRegularizer),this.dilationRate=el(t.dilationRate==null?1:t.dilationRate,e,"dilationRate"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new V(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new V(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new V(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(e){if(tr("kernelSize"in e,"required key 'kernelSize' not in config"),typeof e.kernelSize!="number"&&!qv(e.kernelSize,"number",1,3))throw new V(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){let e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:rs(this.activation),useBias:this.useBias,biasInitializer:Ct(this.biasInitializer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),biasConstraint:qt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},Kc=class extends Iw{constructor(e,t){super(e,t),this.kernel=null,Kc.verifyArgs(t),this.filters=t.filters,Qt(this.filters,"filters"),this.kernelInitializer=St(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Kt(t.kernelConstraint),this.kernelRegularizer=Tt(t.kernelRegularizer)}build(e){e=Qe(e);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V(`The channel dimension of the input should be defined. Found ${e[t]}`);let n=e[t],a=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",a,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:n}}],this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n,a=this.bias==null?null:this.bias.read(),r=BT(this.activation.getClassName());if(r!=null&&this.rank===2)n=Tk(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate,r);else{if(this.rank===1)n=CU(e,this.kernel.read(),a,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)n=Tk(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)n=_U(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new Re("convolutions greater than 3D are not implemented yet.");this.activation!=null&&(n=this.activation.apply(n))}return n})}computeOutputShape(e){e=Qe(e);let t=[],n=this.dataFormat==="channelsLast"?e.slice(1,e.length-1):e.slice(2);for(let r=0;r<n.length;++r){let s=Ga(n[r],this.kernelSize[r],this.padding,this.strides[r],typeof this.dilationRate=="number"?this.dilationRate:this.dilationRate[r]);t.push(s)}let a=[e[0]];return this.dataFormat==="channelsLast"?(a=a.concat(t),a.push(this.filters)):(a.push(this.filters),a=a.concat(t)),a}getConfig(){let e={filters:this.filters,kernelInitializer:Ct(this.kernelInitializer),kernelRegularizer:pt(this.kernelRegularizer),kernelConstraint:qt(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||typeof e.filters!="number"||e.filters<1)throw new V(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(e.filters)}`)}},Xc=class extends Kc{constructor(e){super(2,e),Xc.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!qv(e.kernelSize,"number",1,2))throw new V(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}};Xc.className="Conv2D";ne.registerClass(Xc);var Yc=class extends Kc{constructor(e){super(3,e),Yc.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!(Array.isArray(e.kernelSize)&&(e.kernelSize.length===1||e.kernelSize.length===3)))throw new V(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}};Yc.className="Conv3D";ne.registerClass(Yc);var Sw=class extends Xc{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:4})],this.padding!=="same"&&this.padding!=="valid")throw new V(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=Qe(e),e.length!==4)throw new V("Input should have rank 4; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],a=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",a,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:4,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{let n=Ne(e);if(n.shape.length!==4)throw new V(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let a=n.shape,r=a[0],s,i;this.dataFormat==="channelsFirst"?(s=2,i=3):(s=1,i=2);let o=a[s],l=a[i],u=this.kernelSize[0],p=this.kernelSize[1],d=this.strides[0],c=this.strides[1],h=nr(o,d,u,this.padding),m=nr(l,c,p,this.padding),f=[r,h,m,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,1]));let g=Dm(n,this.kernel.read(),f,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(g=Ee(g,[0,3,1,2])),this.bias!=null&&(g=Xa(g,this.bias.read(),this.dataFormat)),this.activation!=null&&(g=this.activation.apply(g)),g})}computeOutputShape(e){e=Qe(e);let t=e.slice(),n,a,r;this.dataFormat==="channelsFirst"?(n=1,a=2,r=3):(n=3,a=1,r=2);let s=this.kernelSize[0],i=this.kernelSize[1],o=this.strides[0],l=this.strides[1];return t[n]=this.filters,t[a]=nr(t[a],o,s,this.padding),t[r]=nr(t[r],l,i,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Sw.className="Conv2DTranspose";ne.registerClass(Sw);var Tw=class extends Yc{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:5})],this.padding!=="same"&&this.padding!=="valid")throw new V(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=Qe(e),e.length!==5)throw new V("Input should have rank 5; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],a=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",a,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:5,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{let n=Ne(e);if(n.shape.length!==5)throw new V(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let a=n.shape,r=a[0],s,i,o;this.dataFormat==="channelsFirst"?(o=2,s=3,i=4):(o=1,s=2,i=3);let l=a[o],u=a[s],p=a[i],d=this.kernelSize[0],c=this.kernelSize[1],h=this.kernelSize[2],m=this.strides[0],f=this.strides[1],g=this.strides[2],y=nr(l,m,d,this.padding),b=nr(u,f,c,this.padding),x=nr(p,g,h,this.padding),w=[r,y,b,x,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,4,1]));let I=lv(n,this.kernel.read(),w,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(I=Ee(I,[0,4,1,2,3])),this.bias!==null&&(I=Xa(I,this.bias.read(),this.dataFormat)),this.activation!==null&&(I=this.activation.apply(I)),I})}computeOutputShape(e){e=Qe(e);let t=e.slice(),n,a,r,s;this.dataFormat==="channelsFirst"?(n=1,a=2,r=3,s=4):(n=4,a=1,r=2,s=3);let i=this.kernelSize[0],o=this.kernelSize[1],l=this.kernelSize[2],u=this.strides[0],p=this.strides[1],d=this.strides[2];return t[n]=this.filters,t[a]=nr(t[a],u,i,this.padding),t[r]=nr(t[r],p,o,this.padding),t[s]=nr(t[s],d,l,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Tw.className="Conv3DTranspose";ne.registerClass(Tw);var RN=class extends Kc{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,t.filters==null)throw new V("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(t.kernelInitializer!=null||t.kernelRegularizer!=null||t.kernelConstraint!=null)throw new V("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(t.padding!=null&&t.padding!=="same"&&t.padding!=="valid")throw new V(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=Tt(t.depthwiseRegularizer),this.depthwiseConstraint=Kt(t.depthwiseConstraint),this.pointwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=Tt(t.pointwiseRegularizer),this.pointwiseConstraint=Kt(t.pointwiseConstraint)}build(e){if(e=Qe(e),e.length<this.rank+2)throw new V(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null||e[t]<0)throw new V(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(e[t])}`);let n=e[t],a=this.kernelSize.concat([n,this.depthMultiplier]),r=[];for(let i=0;i<this.rank;++i)r.push(1);r.push(n*this.depthMultiplier,this.filters);let s=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",a,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,s,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",r,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,s,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,s,this.biasConstraint):this.bias=null,this.inputSpec=[new zt({ndim:this.rank+2,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n;if(this.rank===1)throw new Re("1D separable convolution is not implemented yet.");return this.rank===2&&(this.dataFormat==="channelsFirst"&&(e=Ee(e,[0,2,3,1])),n=vs(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(n=Xa(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),this.dataFormat==="channelsFirst"&&(n=Ee(n,[0,3,1,2])),n})}getConfig(){let e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=Ct(this.depthwiseInitializer),e.pointwiseInitializer=Ct(this.pointwiseInitializer),e.depthwiseRegularizer=pt(this.depthwiseRegularizer),e.pointwiseRegularizer=pt(this.pointwiseRegularizer),e.depthwiseConstraint=qt(this.depthwiseConstraint),e.pointwiseConstraint=qt(this.pointwiseConstraint),e}};RN.className="SeparableConv";var Nw=class extends RN{constructor(e){super(2,e)}};Nw.className="SeparableConv2D";ne.registerClass(Nw);var Tf=class extends Kc{constructor(e){super(1,e),Tf.verifyArgs(e),this.inputSpec=[{ndim:3}]}getConfig(){let e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!qv(e.kernelSize,"number",1,1))throw new V(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}};Tf.className="Conv1D";ne.registerClass(Tf);var Cw=class extends Ge{constructor(e){super(e),typeof e.cropping=="number"?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:typeof e.cropping[0]=="number"?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=e.dataFormat===void 0?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return this.dataFormat==="channelsFirst"?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return P(()=>{if(e=Ne(e),this.dataFormat==="channelsLast"){let n=Kd(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return Kd(n,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let n=Kd(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return Kd(n,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Cw.className="Cropping2D";ne.registerClass(Cw);var _w=class extends Ge{constructor(e){super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=e.size==null?this.DEFAULT_SIZE:e.size,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),this.interpolation=e.interpolation==null?"nearest":e.interpolation,z4(this.interpolation)}computeOutputShape(e){if(this.dataFormat==="channelsFirst"){let t=e[2]==null?null:this.size[0]*e[2],n=e[3]==null?null:this.size[1]*e[3];return[e[0],e[1],t,n]}else{let t=e[1]==null?null:this.size[0]*e[1],n=e[2]==null?null:this.size[1]*e[2];return[e[0],t,n,e[3]]}}call(e,t){return P(()=>{let n=Ne(e),a=n.shape;if(this.dataFormat==="channelsFirst"){n=Ee(n,[0,2,3,1]);let r=this.size[0]*a[2],s=this.size[1]*a[3],i=this.interpolation==="nearest"?za.resizeNearestNeighbor(n,[r,s]):za.resizeBilinear(n,[r,s]);return Ee(i,[0,3,1,2])}else{let r=this.size[0]*a[1],s=this.size[1]*a[2];return this.interpolation==="nearest"?za.resizeNearestNeighbor(n,[r,s]):za.resizeBilinear(n,[r,s])}})}getConfig(){let e={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}};_w.className="UpSampling2D";ne.registerClass(_w);function EU(e,t,n=[1,1],a="valid",r,s){return P(()=>{r==null&&(r=ja()),Rt(r);let i=kw(e,r);if(e.rank!==4)throw new V(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(t.rank!==4)throw new V(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=bs(i,t,n,a==="same"?"same":"valid","NHWC",s),r==="channelsFirst"&&(i=Ee(i,[0,3,1,2])),i})}var Ew=class extends Iw{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=St(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Kt(e.depthwiseConstraint),this.depthwiseRegularizer=Tt(e.depthwiseRegularizer)}build(e){if(e=Qe(e),e.length<4)throw new V(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);let t=this.dataFormat==="channelsFirst"?1:3;if(e[t]==null||e[t]<0)throw new V(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);let n=e[t],a=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",a,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n=EU(e,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(n=Xa(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),n})}computeOutputShape(e){e=Qe(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],a=this.dataFormat==="channelsFirst"?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,r=Ga(t,this.kernelSize[0],this.padding,this.strides[0]),s=Ga(n,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat==="channelsFirst"?[e[0],a,r,s]:[e[0],r,s,a]}getConfig(){let e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=Ct(this.depthwiseInitializer),e.depthwiseRegularizer=pt(this.depthwiseRegularizer),e.depthwiseConstraint=qt(this.depthwiseRegularizer),e}};Ew.className="DepthwiseConv2D";ne.registerClass(Ew);function MN(e,t,n,a){if(Array.isArray(e)){if(t!=null||n!=null)throw new V("When inputs is an array, neither initialState or constants should be provided");a!=null&&(n=e.slice(e.length-a,e.length),e=e.slice(0,e.length-a)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function r(s){return s==null||Array.isArray(s)?s:[s]}return t=r(t),n=r(n),{inputs:e,initialState:t,constants:n}}function PN(e,t,n,a=!1,r,s,i=!1,o=!1){return P(()=>{let l=t.shape.length;if(l<3)throw new V(`Input should be at least 3D, but is ${l}D.`);let u=[1,0].concat(Ha(2,l));if(t=Ee(t,u),s!=null)throw new Re("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),r!=null&&(r=oe(oe(r,"bool"),"float32"),r.rank===l-1&&(r=Zt(r,-1)),r=Ee(r,u)),a&&(t=fa(t,0),r!=null&&(r=fa(r,0)));let p=[],d,c=n,h=t.shape[0],m=ct(t),f;r!=null&&(f=ct(r));for(let y=0;y<h;++y){let b=m[y],x=P(()=>e(b,c));if(r==null)d=x[0],c=x[1];else{let w=P(()=>{let I=f[y],T=pe(ta(I),I),C=Y(z(x[0],I),z(c[0],T)),E=c.map((A,R)=>Y(z(x[1][R],I),z(A,T)));return{output:C,newStates:E}});d=w.output,c=w.newStates}o&&p.push(d)}let g;return o&&(g=Ft(p,1)),[d,g,c]})}var dr=class extends Ge{constructor(e){super(e);let t;if(e.cell==null)throw new V("cell property is missing for the constructor of RNN.");if(Array.isArray(e.cell)?t=new _f({cells:e.cell}):t=e.cell,t.stateSize==null)throw new V("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=e.returnSequences==null?!1:e.returnSequences,this.returnState=e.returnState==null?!1:e.returnState,this.goBackwards=e.goBackwards==null?!1:e.goBackwards,this._stateful=e.stateful==null?!1:e.stateful,this.unroll=e.unroll==null?!1:e.unroll,this.supportsMasking=!0,this.inputSpec=[new zt({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return Ha(0,e).map(t=>null)}else return this.states_}setStates(e){this.states_=e}computeOutputShape(e){Lb(e)&&(e=e[0]),e=e;let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);let n=t[0],a;if(this.returnSequences?a=[e[0],e[1],n]:a=[e[0],n],this.returnState){let r=[];for(let s of t)r.push([e[0],s]);return[a].concat(r)}else return a}computeMask(e,t){return P(()=>{Array.isArray(t)&&(t=t[0]);let n=this.returnSequences?t:null;if(this.returnState){let a=this.states.map(r=>null);return[n].concat(a)}else return n})}get states(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let n=0;n<e;++n)t.push(null);return t}else return this.states_}set states(e){this.states_=e}build(e){if(this.numConstants!=null)throw new Re("Constants support is not implemented in RNN yet.");Lb(e)&&(e=e[0]),e=e;let t=this.stateful?e[0]:null,n=e.slice(2);this.inputSpec[0]=new zt({shape:[t,null,...n]});let a=[e[0]].concat(e.slice(2));this.cell.build(a);let r;if(Array.isArray(this.cell.stateSize)?r=this.cell.stateSize:r=[this.cell.stateSize],this.stateSpec!=null){if(!v.arraysEqual(this.stateSpec.map(s=>s.shape[s.shape.length-1]),r))throw new V(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=r.map(s=>new zt({shape:[null,s]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){P(()=>{if(!this.stateful)throw new vr("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape[0];if(n==null)throw new V("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(a=>It([n,a])):this.states_=[It([n,this.cell.stateSize])];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(a=>It([n,a])):this.states_[0]=It([n,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new V(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t===!0?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let a=0;a<this.states_.length;++a){let r=e[a],s=Array.isArray(this.cell.stateSize)?this.cell.stateSize[a]:this.cell.stateSize,i=[n,s];if(!v.arraysEqual(r.shape,i))throw new V(`State ${a} is incompatible with layer ${this.name}: expected shape=${i}, received shape=${r.shape}`);this.states_[a]=r}}this.states_=this.states_.map(a=>Jt(a.clone()))})}apply(e,t){let n=t==null?null:t.initialState,a=t==null?null:t.constants;t==null&&(t={});let r=MN(e,n,a,this.numConstants);e=r.inputs,n=r.initialState,a=r.constants;let s=[],i=[];if(n!=null){t.initialState=n,s=s.concat(n),this.stateSpec=[];for(let o of n)this.stateSpec.push(new zt({shape:o.shape}));i=i.concat(this.stateSpec)}if(a!=null&&(t.constants=a,s=s.concat(a),this.numConstants=a.length),s[0]instanceof Ba){let o=[e].concat(s),l=this.inputSpec.concat(i),u=this.inputSpec;this.inputSpec=l;let p=super.apply(o,t);return this.inputSpec=u,p}else return super.apply(e,t)}call(e,t){return P(()=>{let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;e=Ne(e),r==null&&(this.stateful?r=this.states_:r=this.getInitialState(e));let s=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(r.length!==s)throw new V(`RNN Layer has ${s} state(s) but was passed ${r.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let i={training:a},o=PN((c,h)=>{let m=this.cell.call([c].concat(h),i);return[m[0],m.slice(1)]},e,r,this.goBackwards,n,null,this.unroll,this.returnSequences),l=o[0],u=o[1],p=o[2];this.stateful&&this.resetStates(p,a);let d=this.returnSequences?u:l;return this.returnState?[d].concat(p):d})}getInitialState(e){return P(()=>{let t=It(e.shape);return t=fe(t,[1,2]),t=Uc(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(n=>n>1?Pb(t,[1,n]):t):this.cell.stateSize>1?[Pb(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){let e=super.getConfig(),t={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(t.numConstants=this.numConstants);let n=this.cell.getConfig();return this.getClassName()===dr.className&&(t.cell={className:this.cell.getClassName(),config:n}),Object.assign(Object.assign(Object.assign({},n),e),t)}static fromConfig(e,t,n={}){let a=t.cell,r=Ua(a,n);return new e(Object.assign(t,{cell:r}))}};dr.className="RNN";ne.registerClass(dr);var Zc=class extends Ge{},Nf=class extends Zc{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=ss(e.activation==null?this.DEFAULT_ACTIVATION:e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=cl([1,as([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=cl([1,as([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=Qe(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{if(e=e,e.length!==2)throw new V(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let n=e[1];e=e[0];let a=t.training==null?!1:t.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=is({ones:()=>ta(e),rate:this.dropout,training:a,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=is({ones:()=>ta(n),rate:this.recurrentDropout,training:a,dropoutFunc:this.dropoutFunc}));let r,s=this.dropoutMask,i=this.recurrentDropoutMask;s!=null?r=sr(z(e,s),this.kernel.read()):r=sr(e,this.kernel.read()),this.bias!=null&&(r=Xa(r,this.bias.read())),i!=null&&(n=z(n,i));let o=Y(r,sr(n,this.recurrentKernel.read()));return this.activation!=null&&(o=this.activation.apply(o)),[o,o]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:rs(this.activation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},e),t)}};Nf.className="SimpleRNNCell";ne.registerClass(Nf);var Aw=class extends dr{constructor(e){e.cell=new Nf(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return new e(t)}};Aw.className="SimpleRNN";ne.registerClass(Aw);var Cf=class extends Zc{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new V("GRUCell does not support reset_after parameter set to true.");this.units=e.units,Qt(this.units,"units"),this.activation=ss(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=ss(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=cl([1,as([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=cl([1,as([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=Qe(e);let t=e[e.length-1];this.kernel=this.addWeight("kernel",[t,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{if(e=e,e.length!==2)throw new V(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training==null?!1:t.training,a=e[1];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=is({ones:()=>ta(e),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=is({ones:()=>ta(a),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let r=this.dropoutMask,s=this.recurrentDropoutMask,i,o,l;0<this.dropout&&this.dropout<1&&(e=z(e,r[0]));let u=sr(e,this.kernel.read());this.useBias&&(u=Xa(u,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(a=z(a,s[0]));let p=this.recurrentKernel.read(),[d,c]=zn(p,[2*this.units,this.units],p.rank-1),h=sr(a,d),[m,f,g]=zn(u,3,u.rank-1),[y,b]=zn(h,2,h.rank-1);i=this.recurrentActivation.apply(Y(m,y)),o=this.recurrentActivation.apply(Y(f,b));let x=sr(z(o,a),c);l=this.activation.apply(Y(g,x));let w=Y(z(i,a),z(Y(1,bt(i)),l));return[w,w]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:rs(this.activation),recurrentActivation:rs(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},e),t)}};Cf.className="GRUCell";ne.registerClass(Cf);var $w=class extends dr{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new Cf(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};$w.className="GRU";ne.registerClass($w);var Jc=class extends Zc{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=ss(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=ss(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=cl([1,as([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=cl([1,as([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;e=Qe(e);let n=e[e.length-1];this.kernel=this.addWeight("kernel",[n,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let a;if(this.useBias){if(this.unitForgetBias){let r=this.biasInitializer,s=this.units;a=new(t=class extends Fa{apply(i,o){let l=r.apply([s]),u=new ff().apply([s]),p=r.apply([s*2]);return uk(uk(l,u),p)}},t.className="CustomInit",t)}else a=this.biasInitializer;this.bias=this.addWeight("bias",[this.units*4],null,a,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return P(()=>{let n=t.training==null?!1:t.training;if(e=e,e.length!==3)throw new V(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let a=e[1],r=e[2];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=is({ones:()=>ta(e),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=is({ones:()=>ta(a),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,i=this.recurrentDropoutMask,o,l,u,p;0<this.dropout&&this.dropout<1&&(e=z(e,s[0]));let d=sr(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(a=z(a,i[0])),d=Y(d,sr(a,this.recurrentKernel.read())),this.useBias&&(d=Xa(d,this.bias.read()));let[c,h,m,f]=zn(d,4,d.rank-1);o=this.recurrentActivation.apply(c),l=this.recurrentActivation.apply(h),u=Y(z(l,r),z(o,this.activation.apply(m))),p=this.recurrentActivation.apply(f);let g=z(p,this.activation.apply(u));return[g,g,u]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:rs(this.activation),recurrentActivation:rs(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},e),t)}};Jc.className="LSTMCell";ne.registerClass(Jc);var Fw=class extends dr{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new Jc(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};Fw.className="LSTM";ne.registerClass(Fw);var _f=class extends Zc{constructor(e){super(e),this.cells=e.cells}get stateSize(){let e=[];for(let t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return P(()=>{e=e;let n=e.slice(1),a=[];for(let i of this.cells.slice().reverse())Array.isArray(i.stateSize)?a.push(n.splice(0,i.stateSize.length)):a.push(n.splice(0,1));a.reverse();let r=[],s;for(let i=0;i<this.cells.length;++i){let o=this.cells[i];n=a[i],i===0?s=[e[0]].concat(n):s=[s[0]].concat(n),s=o.call(s,t),r.push(s.slice(1))}n=[];for(let i of r.slice().reverse())n.push(...i);return[s[0]].concat(n)})}build(e){Lb(e)&&(e=e[0]),e=e;let t;this.cells.forEach((n,a)=>{Xs(`RNNCell_${a}`,()=>{n.build(e),Array.isArray(n.stateSize)?t=n.stateSize[0]:t=n.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){let e=super.getConfig(),t=a=>({className:a.getClassName(),config:a.getConfig()}),n={cells:this.cells.map(t)};return Object.assign(Object.assign({},e),n)}static fromConfig(e,t,n={}){let a=[];for(let r of t.cells)a.push(Ua(r,n));return new e({cells:a})}get trainableWeights(){if(!this.trainable)return[];let e=[];for(let t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.cells)t.push(...n.trainableWeights);return t.concat(e)}return e}getWeights(){let e=[];for(let t of this.cells)e.push(...t.weights);return zb(e)}setWeights(e){let t=[];for(let n of this.cells){let a=n.weights.length,r=e.splice(a);for(let s=0;s<n.weights.length;++s)t.push([n.weights[s],r[s]])}nw(t)}};_f.className="StackedRNNCells";ne.registerClass(_f);function is(e){let{ones:t,rate:n,training:a=!1,count:r=1,dropoutFunc:s}=e,i=()=>s!=null?s(t(),n):KT(t(),n),o=()=>Hc(i,t,a);return!r||r<=1?Jt(o().clone()):Array(r).fill(void 0).map(o).map(l=>Jt(l.clone()))}var AU=function(e,t){var n={};for(var a in e)Object.prototype.hasOwnProperty.call(e,a)&&t.indexOf(a)<0&&(n[a]=e[a]);if(e!=null&&typeof Object.getOwnPropertySymbols=="function")for(var r=0,a=Object.getOwnPropertySymbols(e);r<a.length;r++)t.indexOf(a[r])<0&&Object.prototype.propertyIsEnumerable.call(e,a[r])&&(n[a[r]]=e[a[r]]);return n},ON=class extends dr{constructor(e){if(e.unroll)throw new Re("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new Re("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new zt({ndim:5})]}call(e,t){return P(()=>{if(this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new V("ConvRNN2D cell does not support constants");let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return P(()=>{let{stateSize:t}=this.cell,n=e.shape,a=this.computeSingleOutputShape(n),r=[a[0],...a.slice(2)],s=It(r);return Array.isArray(t)?Array(t.length).fill(s):[s]})}resetStates(e,t=!1){P(()=>{if(!this.stateful)throw new vr("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape,a=this.computeSingleOutputShape(n),r=[a[0],...a.slice(2)];if(n[0]==null)throw new V("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>It(r)):this.states_=[It(r)];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>It(r)):this.states_[0]=It(r);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new V(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let s=0;s<this.states_.length;++s){let i=e[s],o=r;if(!v.arraysEqual(i.shape,o))throw new V(`State ${s} is incompatible with layer ${this.name}: expected shape=${o}, received shape=${i.shape}`);this.states_[s]=i}}this.states_=this.states_.map(s=>Jt(s.clone()))})}computeSingleOutputShape(e){let{dataFormat:t,filters:n,kernelSize:a,padding:r,strides:s,dilationRate:i}=this.cell,o=t==="channelsFirst",l=e[o?3:2],u=e[o?4:3],p=Ga(l,a[0],r,s[0],i[0]),d=Ga(u,a[1],r,s[1],i[1]);return[...e.slice(0,2),...o?[n,p,d]:[p,d,n]]}};ON.className="ConvRNN2D";var Ef=class extends Jc{constructor(e){let{filters:t,kernelSize:n,strides:a,padding:r,dataFormat:s,dilationRate:i}=e;super(Object.assign(Object.assign({},e),{units:t})),this.filters=t,Qt(this.filters,"filters"),this.kernelSize=el(n,2,"kernelSize"),this.kernelSize.forEach(o=>Qt(o,"kernelSize")),this.strides=el(a||1,2,"strides"),this.strides.forEach(o=>Qt(o,"strides")),this.padding=r||"valid",ba(this.padding),this.dataFormat=s||"channelsLast",Rt(this.dataFormat),this.dilationRate=el(i||1,2,"dilationRate"),this.dilationRate.forEach(o=>Qt(o,"dilationRate"))}build(e){var t;e=Qe(e);let n=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[n]==null)throw new V(`The channel dimension of the input should be defined. Found ${e[n]}`);let a=e[n],r=4,s=this.kernelSize.concat([a,this.filters*r]);this.kernel=this.addWeight("kernel",s,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let i=this.kernelSize.concat([this.filters,this.filters*r]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",i,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let o;if(this.unitForgetBias){let l=this.biasInitializer,u=this.filters;o=new(t=class extends Fa{apply(p,d){let c=l.apply([u]),h=Zn([u]),m=l.apply([u*2]);return Kv([c,h,m])}},t.className="CustomInit",t)}else o=this.biasInitializer;this.bias=this.addWeight("bias",[this.filters*r],null,o,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return P(()=>{if(e.length!==3)throw new V(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training||!1,a=e[0],r=e[1],s=e[2],i=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=is({ones:()=>ta(a),rate:this.dropout,training:n,count:i,dropoutFunc:this.dropoutFunc}));let o=this.dropoutMask,l=(Z,Q,ee)=>!Q||!Q[ee]?Z:z(Q[ee],Z),u=l(a,o,0),p=l(a,o,1),d=l(a,o,2),c=l(a,o,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=is({ones:()=>ta(r),rate:this.recurrentDropout,training:n,count:i,dropoutFunc:this.dropoutFunc}));let h=this.recurrentDropoutMask,m=l(r,h,0),f=l(r,h,1),g=l(r,h,2),y=l(r,h,3),b=3,[x,w,I,T]=zn(this.kernel.read(),i,b),[C,E,A,R]=this.useBias?zn(this.bias.read(),i):[null,null,null,null];u=this.inputConv(u,x,C,this.padding),p=this.inputConv(p,w,E,this.padding),d=this.inputConv(d,I,A,this.padding),c=this.inputConv(c,T,R,this.padding);let[F,S,M,B]=zn(this.recurrentKernel.read(),i,b);m=this.recurrentConv(m,F),f=this.recurrentConv(f,S),g=this.recurrentConv(g,M),y=this.recurrentConv(y,B);let U=this.recurrentActivation.apply(Y(u,m)),G=this.recurrentActivation.apply(Y(p,f)),q=Y(z(G,s),z(U,this.activation.apply(Y(d,g)))),K=z(this.recurrentActivation.apply(Y(c,y)),this.activation.apply(q));return[K,K,q]})}getConfig(){let e=super.getConfig(),{units:t}=e,n=AU(e,["units"]),a={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),a)}inputConv(e,t,n,a){let r=$t(e,t,this.strides,a||"valid",this.dataFormat==="channelsFirst"?"NCHW":"NHWC",this.dilationRate);return n?Xa(r,n,this.dataFormat):r}recurrentConv(e,t){return $t(e,t,1,"same",this.dataFormat==="channelsFirst"?"NCHW":"NHWC")}};Ef.className="ConvLSTM2DCell";ne.registerClass(Ef);var Dw=class extends ON{constructor(e){let t=new Ef(e);super(Object.assign(Object.assign({},e),{cell:t}))}static fromConfig(e,t){return new e(t)}};Dw.className="ConvLSTM2D";ne.registerClass(Dw);var Af=class extends Ge{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(this.noiseShape==null)return this.noiseShape;let t=e.shape,n=[];for(let a=0;a<this.noiseShape.length;++a)n.push(this.noiseShape[a]==null?t[a]:this.noiseShape[a]);return n}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);if(0<this.rate&&this.rate<1){let a=t.training==null?!1:t.training,r=this.getNoiseShape(n);return Hc(()=>KT(n,this.rate,r,this.seed),()=>n,a)}return e})}getConfig(){let e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};Af.className="Dropout";ne.registerClass(Af);var Rw=class extends Af{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){let t=e.shape;return[t[0],1,t[2]]}};Rw.className="SpatialDropout1D";ne.registerClass(Rw);var Mw=class extends Ge{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.batchInputShape==null&&e.inputShape==null&&e.inputDim!=null){let t=null;e.batchSize!=null&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,Qt(this.units,"units"),this.activation=ss(e.activation),e.useBias!=null&&(this.useBias=e.useBias),this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Kt(e.kernelConstraint),this.biasConstraint=Kt(e.biasConstraint),this.kernelRegularizer=Tt(e.kernelRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){e=Qe(e);let t=e[e.length-1];this.kernel==null&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){e=Qe(e);let t=e.slice();return t[t.length-1]=this.units,t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=BT(this.activation.getClassName()),r;return a!=null?r=sr(n,this.kernel.read(),a,this.bias?this.bias.read():null):(r=sr(n,this.kernel.read()),this.bias!=null&&(r=Xa(r,this.bias.read())),this.activation!=null&&(r=this.activation.apply(r))),r})}getConfig(){let e={units:this.units,activation:rs(this.activation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),biasConstraint:qt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};Mw.className="Dense";ne.registerClass(Mw);var Pw=class extends Ge{constructor(e){e=e||{},super(e),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=Qe(e);for(let t of e.slice(1))if(t==null)throw new V(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],Qr(e,1)]}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);if(this.dataFormat==="channelsFirst"&&n.rank>1){let a=[0];for(let r=2;r<n.rank;++r)a.push(r);a.push(1),n=Ee(n,a)}return H4(n)})}getConfig(){let e={};this.dataFormat!=null&&(e.dataFormat=this.dataFormat);let t=super.getConfig();return Object.assign(e,t),e}};Pw.className="Flatten";ne.registerClass(Pw);var Ow=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.activation=ss(e.activation)}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return this.activation.apply(n)})}getConfig(){let e={activation:rs(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};Ow.className="Activation";ne.registerClass(Ow);var Lw=class extends Ge{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return P(()=>(e=Ne(e),U4(e,this.n)))}getConfig(){let e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};Lw.className="RepeatVector";ne.registerClass(Lw);var zw=class extends Ge{constructor(e){super(e),this.targetShape=e.targetShape;for(let t=0;t<this.targetShape.length;++t)this.isUnknown(this.targetShape[t])&&(this.targetShape[t]=null)}isUnknown(e){return e<0||e==null}fixUnknownDimension(e,t){let n="Total size of new array must be unchanged.",a=t.slice(),r=1,s=null;for(let o=0;o<a.length;++o){let l=a[o];if(this.isUnknown(l))if(s===null)s=o;else throw new V("Can only specifiy one unknown dimension.");else r*=l}let i=Qr(e);if(s!==null){if(r===0||i%r!==0)throw new V(n);a[s]=i/r}else if(i!==r)throw new V(n);return a}computeOutputShape(e){let t=!1;for(let n=0;n<e.length;++n)if(this.isUnknown(e[n])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=n.shape,r=a.slice(0,1).concat(this.fixUnknownDimension(a.slice(1),this.targetShape));return W(n,r)})}getConfig(){let e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};zw.className="Reshape";ne.registerClass(zw);var Ww=class extends Ge{constructor(e){if(super(e),e.dims==null)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);let t=Ha(1,e.dims.length+1);if(!v.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new zt({ndim:this.dims.length+1})]}computeOutputShape(e){e=Qe(e);let t=e.slice();return this.dims.forEach((n,a)=>{t[a+1]=e[n]}),t}call(e,t){return Ee(Ne(e),this.dimsIncludingBatch)}getConfig(){let e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};Ww.className="Permute";ne.registerClass(Ww);var Bw=class extends Ge{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null?this.maskValue=e.maskValue==null?0:e.maskValue:this.maskValue=0}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){let n=Ne(e),a=-1;return qp(ii(n,this.maskValue),a)}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=-1,r=!0,s=qp(ii(n,this.maskValue),a,r);return z(n,oe(s,n.dtype))})}};Bw.className="Masking";ne.registerClass(Bw);var Vw=class extends Ge{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",e.batchInputShape==null&&e.inputShape==null){let t=null;e.batchSize!=null&&(t=e.batchSize),e.inputLength==null?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(yt(e.inputLength))}this.inputDim=e.inputDim,Qt(this.inputDim,"inputDim"),this.outputDim=e.outputDim,Qt(this.outputDim,"outputDim"),this.embeddingsInitializer=St(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=Tt(e.embeddingsRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.embeddingsConstraint=Kt(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return P(()=>this.maskZero?(e=Ne(e),ii(e,qe(e))):null)}computeOutputShape(e){if(e=Qe(e),this.inputLength==null)return[...e,this.outputDim];let t=yt(this.inputLength);if(t.length!==e.length-1)throw new V(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let n=0;for(let a=0;a<t.length;++a){let r=t[a],s=e[a+1];if(r!=null&&s!=null&&r!==s)throw new V(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);r==null&&(t[n]=s),n++}}return[e[0],...t,this.outputDim]}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);n.dtype!=="int32"&&(n=yo(n,"int32"));let a=qT(this.embeddings.read(),W(n,[n.size]));return W(a,Qe(this.computeOutputShape(n.shape)))})}getConfig(){let e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Ct(this.embeddingsInitializer),embeddingsRegularizer:pt(this.embeddingsRegularizer),activityRegularizer:pt(this.activityRegularizer),embeddingsConstraint:qt(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};Vw.className="Embedding";ne.registerClass(Vw);var xo=class extends Ge{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new Re}computeElementwiseOpOutputShape(e,t){if(e==null||t==null)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(t.length===0)return e;let n=e.slice(0,e.length-t.length);for(let a=0;a<t.length;++a){let r=e[e.length-t.length+a],s=t[a];if(r==null||s==null||r<0||s<0)n.push(null);else if(r===1)n.push(s);else if(s===1)n.push(r);else{if(r!==s)throw new V("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));n.push(r)}}return n}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[Qe(e)]),e=e,e.length<2)throw new V(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(let r of e)r!=null&&r[0]!==null&&t.push(r[0]);if(t=Jr(t),t.length>1)throw new V(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let n=e[0]==null?null:e[0].slice(1);for(let r=1;r<e.length;++r){let s=e[r]==null?null:e[r].slice(1);n=this.computeElementwiseOpOutputShape(n,s)}let a=e.map(r=>r.length);e.indexOf(null)===-1&&Jr(a).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return P(()=>{if(e=e,this.reshapeRequired){let n=[],a=e.map(r=>r.rank);if(a.indexOf(null)===-1){let r=as(a);for(let s of e){let i=s.rank;for(let o=0;o<r-i;++o)s=Uc(s,1);n.push(s)}return this.mergeFunction(n)}else{let r=!1;for(let o of e){let l=o.rank;if(l==null){let u=o.shape,p=u[0],d=u.slice(1).concat([p]),c=W(o,[p].concat(Qr(u.slice(1))));c=Ee(c,[1,0]),c=W(c,d),n.push(c),r=!0}else if(l>1){let u=Ha(1,l).concat([0]);n.push(Ee(o,u)),r=!0}else n.push(o)}let s=this.mergeFunction(n),i=s.rank;if(r){if(i==null){let o=s.shape,l=o.length,u=o[l-1],p=[u].concat(o.slice(0,o.length-1));s=W(Ee(W(s,[-1,u]),[1,0]),p)}else if(i>1){let o=[i-1].concat(Ha(0,i-1));s=Ee(s,o)}}return s}}else return this.mergeFunction(e)})}computeOutputShape(e){e=e;let t;e[0]==null?t=null:t=e[0].slice(1);for(let a=1;a<e.length;++a){let r=e[a]==null?null:e[a].slice(1);t=this.computeElementwiseOpOutputShape(t,r)}let n=[];for(let a of e)a!=null&&a[0]!==null&&n.push(a[0]);return n=Jr(n),n.length===1?t=n.concat(t):t=[null].concat(t),t}computeMask(e,t){return P(()=>{if(t==null)return null;if(!Array.isArray(t))throw new V("`mask` should be an Array");if(!Array.isArray(e))throw new V("`inputs` should be an Array");if(t.length!==e.length)throw new V(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(a=>a==null))return null;t=t.map(a=>a==null?a:Zt(a,0));let n=t[0];for(let a=1;a<t.length-1;++a)n=Ea(n,t[a]);return n})}},Uw=class extends xo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return t})}};Uw.className="Add";ne.registerClass(Uw);var Gw=class extends xo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=z(t,e[n]);return t})}};Gw.className="Multiply";ne.registerClass(Gw);var Hw=class extends xo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return z(1/e.length,t)})}};Hw.className="Average";ne.registerClass(Hw);var jw=class extends xo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=pr(t,e[n]);return t})}};jw.className="Maximum";ne.registerClass(jw);var qw=class extends xo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=Au(t,e[n]);return t})}};qw.className="Minimum";ne.registerClass(qw);var Kw=class extends xo{constructor(e){super(e),this.DEFAULT_AXIS=-1,e==null&&(e={}),this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!(Array.isArray(e)&&Array.isArray(e[0]))||e.length===1)throw new V("A `Concatenate` layer should be called on a list of at least 2 inputs");e=e;let t=!0;for(let a of e)if(a!=null){t=!1;break}if(t)return;let n=[];for(let a=0;a<e.length;++a){let r=e[a].slice();r.splice(this.axis,1);let s=!1;for(let i of n)if(v.arraysEqual(i,r)){s=!0;break}s||n.push(r)}if(n.length>1)throw new V("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return P(()=>Kv(e,this.axis))}computeOutputShape(e){if(!(Array.isArray(e)&&Array.isArray(e[0])))throw new V("A `Concatenate` layer should be called on a list of inputs.");let t=e,n=t[0].slice(),a=this.axis<0?n.length+this.axis:this.axis;for(let r of t.slice(1)){if(n[a]==null||r[a]==null){n[a]=null;break}n[a]+=r[a]}return n}computeMask(e,t){if(t==null)return null;if(!Array.isArray(t))throw new V("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new V("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new V(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return P(()=>{let n=!0;if(t.forEach(s=>{if(s!=null){n=!1;return}}),n)return null;let a=[];for(let s=0;s<e.length;++s)t[s]==null?a.push(oe(ta(e[s]),"bool")):t[s].rank<e[s].rank?a.push(Zt(t[s],-1)):a.push(t[s]);let r=Ze(a,this.axis);return $m(r,-1,!1)})}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};Kw.className="Concatenate";ne.registerClass(Kw);function Sp(e,t){for(;e<0;)e+=t;return e}function $U(e,t,n){if(e.shape.length>3||t.shape.length>3)throw new Re("batchDot is not implemented for tensors of 4D or higher rank yet");if(v.assert(e.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${e.shape.length}`),v.assert(e.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof n=="number"&&(n=[n,n]),e.dtype==="complex64"||t.dtype==="complex64")throw new Re("batchDot is not implemented for complex64-type Tensors yet.");let a=e.shape.length,r=t.shape.length;n==null&&(n=[a-1,r-2]);let s=n;return P(()=>{let i;if(a>r){i=a-r;let l=[];for(let u=0;u<i;++u)l.push(1);t=W(t,t.shape.concat(l))}else if(r>a){i=r-a;let l=[];for(let u=0;u<i;++u)l.push(1);e=W(e,e.shape.concat(l))}else i=0;let o;if(e.shape.length===2&&t.shape.length===2)s[0]===s[1]?o=fe(z(e,t),s[0]):o=fe(z(Ee(e,[1,0]),t),s[1]);else{let l=s[0]!==e.shape.length-1,u=s[1]===t.shape.length-1;o=Fe(e,t,l,u)}if(i>0){let l;a>r?l=a+r-3:l=a-1;let u=[];for(let p=l;p<l+i;++p)u.push(p);o=ws(o,u)}return o.shape.length===1&&(o=Zt(o,1)),o})}var Xw=class extends xo{constructor(e){super(e),this.axes=e.axes,this.normalize=e.normalize==null?!1:e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0],n=e[1];if(t.length>3||n.length>3)throw new Re("Dot layer does not support tensors of 4D or higher rank yet.");let a=this.interpretAxes(t,n);if(t[a[0]]!==n[a[1]])throw new V(`Dimension incompatibility: ${t[a[0]]} !== ${n[a[1]]}`)}mergeFunction(e){if(e.length!==2)throw new V(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t=e[0],n=e[1],a;return Array.isArray(this.axes)?a=this.axes.map((r,s)=>Sp(r,e[s].shape.length)):a=[Sp(this.axes,t.shape.length),Sp(this.axes,n.shape.length)],this.normalize&&(t=Eh(t,a[0]),n=Eh(n,a[1])),$U(t,n,a)}interpretAxes(e,t){let n;return Array.isArray(this.axes)?n=this.axes:n=[Sp(this.axes,e.length),Sp(this.axes,t.length)],n}computeOutputShape(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0].slice(),n=e[1].slice();if(t.length>3||n.length>3)throw new Re("Dot layer does not support tensors of 4D or higher rank yet.");let a=this.interpretAxes(t,n);t.splice(a[0],1),n.splice(a[1],1),n.splice(0,1);let r=t.concat(n);return r.length===1&&r.push(1),r}computeMask(e,t){return null}getConfig(){let e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};Xw.className="Dot";ne.registerClass(Xw);var Yw=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return Hc(()=>Y(mf(n.shape,0,this.stddev),n),()=>n,t.training||!1)})}};Yw.className="GaussianNoise";ne.registerClass(Yw);var Zw=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return this.rate>0&&this.rate<1?Hc(()=>{let a=Math.sqrt(this.rate/(1-this.rate));return z(n,mf(n.shape,1,a))},()=>n,t.training||!1):n})}};Zw.className="GaussianDropout";ne.registerClass(Zw);var Jw=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||Ne(e).shape}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return P(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(e);return Hc(()=>{let a=Ne(e),r=1.6732632423543772,s=1.0507009873554805,i=-r*s,o=Er($u(n),this.rate);o=yo(o,"float32");let l=((1-this.rate)*(1+this.rate*i**2))**-.5,u=-l*i*this.rate,p=Y(z(a,o),z(Y(o,-1),i));return Y(z(p,l),u)},()=>Ne(e),t.training||!1)}return e})}};Jw.className="AlphaDropout";ne.registerClass(Jw);function Zp(e,t,n,a,r,s=.001){let i;if(e.rank===2)i=Zx(e,t,n,a,r,s);else if(e.rank===3)i=Jx(e,t,n,a,r,s);else if(e.rank===4)i=Qx(e,t,n,a,r,s);else throw new Re(`batchNormalization is not implemented for array of rank ${e.rank} yet`);return i}function FU(e,t,n,a,r=.001){return P(()=>{let s=Mc(e,a),i=s.mean,o=s.variance;return[Zp(e,i,o,n,t,r),i,o]})}function DU(e,t,n,a,r=.001){return P(()=>{let s=Mc(e,a),i=s.mean,o=s.variance,l=[];for(let h of Ha(0,e.rank))a.indexOf(h)!==-1?l.push(1):l.push(e.shape[h]);let u=W(i,l),p=W(o,l),d=t==null?null:W(t,l),c=n==null?null:W(n,l);return[Zp(e,u,p,c,d,r),i,o]})}function RU(e,t,n,a,r=.001){return v.arraysEqual(a.slice().sort(),Ha(0,e.rank-1))?FU(e,t,n,a,r):DU(e,t,n,a,r)}var Qw=class extends Ge{constructor(e){e==null&&(e={}),super(e),this.supportsMasking=!0,this.axis=e.axis==null?-1:e.axis,this.momentum=e.momentum==null?.99:e.momentum,this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.movingMeanInitializer=St(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=St(e.movingVarianceInitializer||"ones"),this.betaConstraint=Kt(e.betaConstraint),this.gammaConstraint=Kt(e.gammaConstraint),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer)}build(e){e=Qe(e);let t=this.axis>=0?this.axis:this.axis+e.length,n=e[t];if(n==null)throw new V(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new zt({ndim:e.length,axes:{[t]:n}})];let a=[n];this.scale&&(this.gamma=this.addWeight("gamma",a,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",a,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",a,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",a,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return P(()=>{let n=t.training==null?!1:t.training,a=Ne(e),r=a.shape,s=r.length,i=Ha(0,s),o=this.axis>=0?this.axis:this.axis+s;i.splice(o,1);let l=oi(1,s);l[o]=r[o];let u=i.slice();u.sort();let p=!v.arraysEqual(u,Ha(0,s).slice(0,s-1)),d=()=>{if(p){let g=W(this.movingMean.read(),l),y=W(this.movingVariance.read(),l),b=this.center?W(this.beta.read(),l):null,x=this.scale?W(this.gamma.read(),l):null;return Zp(a,g,y,b,x,this.epsilon)}else return Zp(a,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!n)return d();let[c,h,m]=RU(a,this.gamma.read(),this.beta.read(),i,this.epsilon),f=(g,y,b)=>{P(()=>{let x=1-b,w=g.read(),I=z(pe(w,y),x);g.write(pe(w,I))})};return f(this.movingMean,h,this.momentum),f(this.movingVariance,m,this.momentum),c})}getConfig(){let e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Ct(this.betaInitializer),gammaInitializer:Ct(this.gammaInitializer),movingMeanInitializer:Ct(this.movingMeanInitializer),movingVarianceInitializer:Ct(this.movingVarianceInitializer),betaRegularizer:pt(this.betaRegularizer),gammaRegularizer:pt(this.gammaRegularizer),betaConstraint:qt(this.betaConstraint),gammaConstraint:qt(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};Qw.className="BatchNormalization";ne.registerClass(Qw);var e0=class extends Ge{constructor(e){if(e==null&&(e={}),super(e),this.axis=e.axis==null?-1:e.axis,typeof this.axis=="number"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let t of this.axis)if(!Number.isInteger(t))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer),this.supportsMasking=!0}build(e){e=Qe(e);let t=e.length;typeof this.axis=="number"&&(this.axis=[this.axis]);for(let r=0;r<this.axis.length;++r)this.axis[r]<0&&(this.axis[r]+=t);for(let r of this.axis)if(r<0||r>=t)throw new Error(`Invalid axis: ${r}`);if(this.axis.length!==Jr(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(r=>e[r]),a=!0;this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,a):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,a):this.beta=null,this.built=!0}call(e,t){let n=Ne(e),a=n.shape,r=a.length;return P(()=>{let{mean:s,variance:i}=Mc(n,this.axis,!0),o=oi(1,r);for(let h of this.axis)o[h]=a[h];let l=h=>h!=null&&h.shape.length!==r?W(h,o):h,u=this.scale?l(this.gamma.read()):null,p=this.center?l(this.beta.read()):null,d=[],c=[];for(let h=0;h<r;++h)this.axis.indexOf(h)!==-1?(d.push(a[h]),c.push(1)):(d.push(1),c.push(a[h]));return s=Ln(s,d),i=Ln(i,d),u!=null&&(u=Ln(u,c)),p!=null&&(p=Ln(p,c)),Zp(n,s,i,p,u,this.epsilon)})}getConfig(){let e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Ct(this.betaInitializer),gammaInitializer:Ct(this.gammaInitializer),betaRegularizer:pt(this.betaRegularizer),gammaRegularizer:pt(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};e0.className="LayerNormalization";ne.registerClass(e0);function MU(e,t,n){return P(()=>{if(e.rank!==4)throw new V(`temporalPadding expects input tensor to be 4-D, but received a ${e.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new V("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(n==null&&(n=ja()),n!=="channelsLast"&&n!=="channelsFirst")throw new V(`Unknown data format: ${n}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let a;return n==="channelsFirst"?a=[[0,0],[0,0],t[0],t[1]]:a=[[0,0],t[0],t[1],[0,0]],ya(e,a)})}var t0=class extends Ge{constructor(e){if(e==null&&(e={}),super(e),this.dataFormat=e.dataFormat==null?ja():e.dataFormat,e.padding==null)this.padding=[[1,1],[1,1]];else if(typeof e.padding=="number")this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,e.padding.length!==2)throw new V(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,n;if(typeof e.padding[0]=="number")t=[e.padding[0],e.padding[0]],n=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,e.padding[0].length!==2)throw new V(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],e.padding[1].length!==2)throw new V(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);n=e.padding[1]}this.padding=[t,n]}this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=Qe(e);let t,n;return this.dataFormat==="channelsFirst"?(e[2]!=null&&e[2]>=0?t=e[2]+this.padding[0][0]+this.padding[0][1]:t=null,e[3]!=null&&e[3]>=0?n=e[3]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],e[1],t,n]):(e[1]!=null&&e[1]>=0?t=e[1]+this.padding[0][0]+this.padding[0][1]:t=null,e[2]!=null&&e[2]>=0?n=e[2]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],t,n,e[3]])}call(e,t){return P(()=>MU(Ne(e),this.padding,this.dataFormat))}getConfig(){let e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};t0.className="ZeroPadding2D";ne.registerClass(t0);function $f(e,t,n,a,r,s){return P(()=>{Rt(r),UT(s),ba(a),n==null&&(n=[1,1]),a==null&&(a="valid"),r==null&&(r=ja()),s==null&&(s="max"),e=kw(e,r);let i,o=a==="same"?"same":"valid";return s==="max"?i=Dt(e,t,n,o):i=ga(e,t,n,o),r==="channelsFirst"&&(i=Ee(i,[0,3,1,2])),i})}function LN(e,t,n,a,r,s){return P(()=>{Rt(r),UT(s),ba(a),n==null&&(n=[1,1,1]),a==null&&(a="valid"),r==null&&(r=ja()),s==null&&(s="max"),e=DN(e,r);let i,o=a==="same"?"same":"valid";return s==="max"?i=Sv(e,t,n,o):i=Yx(e,t,n,o),r==="channelsFirst"&&(i=Ee(i,[0,4,1,2,3])),i})}var zN=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=2),super(e),typeof e.poolSize=="number")this.poolSize=[e.poolSize];else if(Array.isArray(e.poolSize)&&e.poolSize.length===1&&typeof e.poolSize[0]=="number")this.poolSize=e.poolSize;else throw new V(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.poolSize)}`);if(Qt(this.poolSize,"poolSize"),e.strides==null)this.strides=this.poolSize;else if(typeof e.strides=="number")this.strides=[e.strides];else if(Array.isArray(e.strides)&&e.strides.length===1&&typeof e.strides[0]=="number")this.strides=e.strides;else throw new V(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.strides)}`);Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,ba(this.padding),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){e=Qe(e);let t=Ga(e[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return P(()=>{this.invokeCallHook(e,t),e=Uc(Ne(e),2);let n=this.poolingFunction(Ne(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return ws(n,[2])})}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},n0=class extends zN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),$f(e,t,n,a,r,"max")}};n0.className="MaxPooling1D";ne.registerClass(n0);var a0=class extends zN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),$f(e,t,n,a,r,"avg")}};a0.className="AveragePooling1D";ne.registerClass(a0);var WN=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==2)throw new V(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),ba(this.padding),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=Qe(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2];return t=Ga(t,this.poolSize[0],this.padding,this.strides[0]),n=Ga(n,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n]:[e[0],t,n,e[3]]}call(e,t){return P(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ne(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},r0=class extends WN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),$f(e,t,n,a,r,"max")}};r0.className="MaxPooling2D";ne.registerClass(r0);var s0=class extends WN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),$f(e,t,n,a,r,"avg")}};s0.className="AveragePooling2D";ne.registerClass(s0);var BN=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==3)throw new V(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),ba(this.padding),this.inputSpec=[new zt({ndim:5})]}computeOutputShape(e){e=Qe(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],a=this.dataFormat==="channelsFirst"?e[4]:e[3];return t=Ga(t,this.poolSize[0],this.padding,this.strides[0]),n=Ga(n,this.poolSize[1],this.padding,this.strides[1]),a=Ga(a,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n,a]:[e[0],t,n,a,e[4]]}call(e,t){return P(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ne(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},i0=class extends BN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),LN(e,t,n,a,r,"max")}};i0.className="MaxPooling3D";ne.registerClass(i0);var o0=class extends BN{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),ba(a),LN(e,t,n,a,r,"avg")}};o0.className="AveragePooling3D";ne.registerClass(o0);var VN=class extends Ge{constructor(e){super(e),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new Re}},l0=class extends VN{constructor(e){super(e||{})}call(e,t){return P(()=>{let n=Ne(e);return Nt(n,1)})}};l0.className="GlobalAveragePooling1D";ne.registerClass(l0);var u0=class extends VN{constructor(e){super(e||{})}call(e,t){return P(()=>{let n=Ne(e);return ha(n,1)})}};u0.className="GlobalMaxPooling1D";ne.registerClass(u0);var UN=class extends Ge{constructor(e){super(e),this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){return e=e,this.dataFormat==="channelsLast"?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new Re}getConfig(){let e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},p0=class extends UN{call(e,t){return P(()=>{let n=Ne(e);return this.dataFormat==="channelsLast"?Nt(n,[1,2]):Nt(n,[2,3])})}};p0.className="GlobalAveragePooling2D";ne.registerClass(p0);var c0=class extends UN{call(e,t){return P(()=>{let n=Ne(e);return this.dataFormat==="channelsLast"?ha(n,[1,2]):ha(n,[2,3])})}};c0.className="GlobalMaxPooling2D";ne.registerClass(c0);var GN=class extends Ge{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(e){this.layer!=null&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){let e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,n={}){let a=t.layer,r=Ua(a,n);delete t.layer;let s={layer:r};return Object.assign(s,t),new e(s)}},d0=class extends GN{constructor(e){super(e),this.supportsMasking=!0}build(e){if(e=Qe(e),e.length<3)throw new V(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(e)}`);this.inputSpec=[{shape:e}];let t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){e=Qe(e);let t=[e[0]].concat(e.slice(2)),n=this.layer.computeOutputShape(t),a=e[1];return[n[0],a].concat(n.slice(1))}call(e,t){return P(()=>(e=Ne(e),PN((n,a)=>[Ne(this.layer.call(n,t)),[]],e,[],!1,null,null,!1,!0)[1]))}};d0.className="TimeDistributed";ne.registerClass(d0);function PU(e){go(L4,"BidirectionalMergeMode",e)}var OU="concat",h0=class extends GN{constructor(e){super(e);let t=e.layer.getConfig(),n={};n.className=e.layer.getClassName(),n.config=t,this.forwardLayer=Ua(n),t.goBackwards=t.goBackwards!==!0;let a={};if(a.className=e.layer.getClassName(),a.config=t,this.backwardLayer=Ua(a),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=e.mergeMode===void 0?OU:e.mergeMode,PU(this.mergeMode),e.weights)throw new Re("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,this.forwardLayer!=null&&(this.forwardLayer.trainable=e),this.backwardLayer!=null&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){let t=e.length,n=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,n)),this.backwardLayer.setWeights(e.slice(n))}computeOutputShape(e){let t=this.forwardLayer.computeOutputShape(e);Array.isArray(t)&&Array.isArray(t[0])||(t=[t]),t=t;let n,a,r;return this.returnState&&(r=t.slice(1)),n=t[0],n=n,this.mergeMode==="concat"?(n[n.length-1]*=2,a=[n]):this.mergeMode==null?a=[n,n.slice()]:a=[n],this.returnState?this.mergeMode==null?a.concat(r).concat(r.slice()):[n].concat(r).concat(r.slice()):On(a)}apply(e,t){let n=t==null?null:t.initialState,a=t==null?null:t.constants;t==null&&(t={});let r=MN(e,n,a,this.numConstants);if(e=r.inputs,n=r.initialState,a=r.constants,Array.isArray(e)&&(n=e.slice(1),e=e[0]),(n==null||n.length===0)&&a==null)return super.apply(e,t);let s=[],i=[];if(n!=null){let l=n.length;if(l%2>0)throw new V("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=n,s.push(...n);let u=n.map(p=>new zt({shape:p.shape}));this.forwardLayer.stateSpec=u.slice(0,l/2),this.backwardLayer.stateSpec=u.slice(l/2),i.push(...u)}if(a!=null)throw new Re("Support for constants in Bidirectional layers is not implemented yet.");let o=s[0]instanceof Ba;for(let l of s)if(l instanceof Ba!==o)throw new V("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(o){let l=[e].concat(s),u=this.inputSpec.concat(i),p=this.inputSpec;this.inputSpec=u;let d=super.apply(l,t);return this.inputSpec=p,d}else return super.apply(e,t)}call(e,t){return P(()=>{let n=t.initialState,a,r;if(n==null)a=this.forwardLayer.call(e,t),r=this.backwardLayer.call(e,t);else{let o=n.slice(0,n.length/2),l=n.slice(n.length/2);a=this.forwardLayer.call(e,Object.assign(t,{initialState:o})),r=this.backwardLayer.call(e,Object.assign(t,{initialState:l}))}let s;this.returnState&&(Array.isArray(a)&&(s=a.slice(1).concat(r.slice(1))),a=a[0],r=r[0]),this.returnSequences&&(r=fa(r,1));let i;return this.mergeMode==="concat"?i=Kv([a,r]):this.mergeMode==="sum"?i=Y(a,r):this.mergeMode==="ave"?i=z(.5,Y(a,r)):this.mergeMode==="mul"?i=z(a,r):this.mergeMode==null&&(i=[a,r]),this.returnState?this.mergeMode==null?i.concat(s):[i].concat(s):i})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Xs(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Xs(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){Array.isArray(t)&&(t=t[0]);let n;if(this.returnSequences?this.mergeMode==null?n=[t,t]:n=t:this.mergeMode==null?n=[null,null]:n=null,this.returnState){let a=this.forwardLayer.states.map(r=>null);return Array.isArray(n)?n.concat(a).concat(a):[n].concat(a).concat(a)}else return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(e),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){let e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){let n=Ua(t.layer);if(delete t.layer,t.numConstants!=null)throw new Re("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let a=t;return a.layer=n,new e(a)}};h0.className="Bidirectional";ne.registerClass(h0);var m0=class extends Ge{constructor(e){super(e),this.scale=e.scale,e.offset?this.offset=e.offset:this.offset=0}getConfig(){let e={scale:this.scale,offset:this.offset},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return P(()=>(e=Ne(e),e.dtype!=="float32"&&(e=yo(e,"float32")),Y(z(e,this.scale),this.offset)))}};m0.className="Rescaling";ne.registerClass(m0);var LU=["bilinear","nearest"],Nk=new Set(LU),f0=class extends Ge{constructor(e){if(super(e),this.height=e.height,this.width=e.width,e.interpolation)if(Nk.has(e.interpolation))this.interpolation=e.interpolation;else throw new V(`Invalid interpolation parameter: ${e.interpolation} is not implemented`);else this.interpolation="bilinear";this.cropToAspectRatio=Boolean(e.cropToAspectRatio)}computeOutputShape(e){e=Qe(e);let t=e[2];return[this.height,this.width,t]}getConfig(){let e={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return P(()=>{let n=[this.height,this.width];if(this.interpolation==="bilinear")return za.resizeBilinear(e,n,!this.cropToAspectRatio);if(this.interpolation==="nearest")return za.resizeNearestNeighbor(e,n,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...Nk]} are supported`)})}};f0.className="Resizing";ne.registerClass(f0);function zU(e,t,n,a){let r=Ne(e);if(r.dtype!=="int32"&&(r=yo(r,"int32")),t==="int")return r;let s=r.shape;if(r.rank===0&&(r=Zt(r,-1)),t==="oneHot"&&r.shape[r.shape.length-1]!==1&&(r=Zt(r,-1)),r.rank>2)throw new V(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${s} which would result in output rank ${r.rank}.`);let i=["multiHot","oneHot"].includes(t),o=r,l;if(typeof a!="undefined"&&t==="count"?l=Sh(o,a,n,i):l=Sh(o,[],n,i),t!=="tfIdf")return l;if(a)return z(l,a);throw new V("When outputMode is 'tfIdf', weights must be provided.")}var g0=class extends Ge{constructor(e){super(e),this.numTokens=e.numTokens,e.outputMode?this.outputMode=e.outputMode:this.outputMode="multiHot"}getConfig(){let e={numTokens:this.numTokens,outputMode:this.outputMode},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){return e=Qe(e),e==null?[this.numTokens]:this.outputMode==="oneHot"&&e[e.length-1]!==1?(e.push(this.numTokens),e):(e[e.length-1]=this.numTokens,e)}call(e,t){return P(()=>{e=Ne(e),e.dtype!=="int32"&&(e=yo(e,"int32"));let n;if(typeof t.countWeights!="undefined"){if(this.outputMode!=="count")throw new V(`countWeights is not used when outputMode !== count.
-              Received countWeights=${t.countWeights}`);n=Ne(t.countWeights)}let a=ha(e),r=il(e),s=Nn(this.numTokens,a).bufferSync().get(0),i=Er(r,0).bufferSync().get(0);if(!(s&&i))throw new V(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return zU(e,this.outputMode,this.numTokens,n)})}};g0.className="CategoryEncoding";ne.registerClass(g0);function WU(e){return new Du(e)}function BU(e){return new xw(e)}function VU(e){return new gw(e)}function UU(e){return new yw(e)}function GU(e){return new bw(e)}function HU(e){return new ww(e)}function jU(e){return new vw(e)}function qU(e){return new Tf(e)}function KU(e){return new Xc(e)}function XU(e){return new Sw(e)}function YU(e){return new Yc(e)}function ZU(e){return new Tw(e)}function JU(e){return new Nw(e)}function QU(e){return new Cw(e)}function eG(e){return new _w(e)}function tG(e){return new Ew(e)}function nG(e){return new Ow(e)}function aG(e){return new Mw(e)}function rG(e){return new Af(e)}function sG(e){return new Rw(e)}function iG(e){return new Pw(e)}function oG(e){return new Lw(e)}function lG(e){return new zw(e)}function uG(e){return new Ww(e)}function pG(e){return new Vw(e)}function cG(e){return new Uw(e)}function dG(e){return new Hw(e)}function hG(e){return new Kw(e)}function mG(e){return new jw(e)}function fG(e){return new qw(e)}function gG(e){return new Gw(e)}function yG(e){return new Xw(e)}function bG(e){return new Qw(e)}function xG(e){return new e0(e)}function vG(e){return new t0(e)}function y0(e){return new a0(e)}function wG(e){return y0(e)}function kG(e){return y0(e)}function b0(e){return new s0(e)}function IG(e){return b0(e)}function SG(e){return b0(e)}function x0(e){return new o0(e)}function TG(e){return x0(e)}function NG(e){return x0(e)}function CG(e){return new l0(e)}function _G(e){return new p0(e)}function HN(e){return new u0(e)}function jN(e){return new c0(e)}function qN(e){return new n0(e)}function KN(e){return new r0(e)}function EG(e){return new i0(e)}function AG(e){return new $w(e)}function $G(e){return new Cf(e)}function FG(e){return new Fw(e)}function DG(e){return new Jc(e)}function RG(e){return new Aw(e)}function MG(e){return new Nf(e)}function PG(e){return new Dw(e)}function OG(e){return new Ef(e)}function LG(e){return new dr(e)}function zG(e){return new _f(e)}function WG(e){return new h0(e)}function BG(e){return new d0(e)}var VG=HN,UG=jN,GG=qN,HG=KN;function jG(e){return new Yw(e)}function qG(e){return new Zw(e)}function KG(e){return new Jw(e)}function XG(e){return new Bw(e)}function YG(e){return new m0(e)}function ZG(e){return new f0(e)}function JG(e){return new g0(e)}var XN={};Ae(XN,{MAPE:()=>uH,MSE:()=>dH,binaryAccuracy:()=>QG,binaryCrossentropy:()=>eH,categoricalAccuracy:()=>nH,categoricalCrossentropy:()=>aH,cosineProximity:()=>iH,mape:()=>pH,meanAbsoluteError:()=>oH,meanAbsolutePercentageError:()=>lH,meanSquaredError:()=>cH,mse:()=>hH,precision:()=>rH,recall:()=>sH,sparseCategoricalAccuracy:()=>tH});function QG(e,t){return uw(e,t)}function eH(e,t){return uN(e,t)}function tH(e,t){return pN(e,t)}function nH(e,t){return pw(e,t)}function aH(e,t){return cw(e,t)}function rH(e,t){return lN(e,t)}function sH(e,t){return VV(e,t)}function iH(e,t){return lw(e,t)}function oH(e,t){return If(e,t)}function lH(e,t){return Ru(e,t)}function uH(e,t){return Ru(e,t)}function pH(e,t){return Ru(e,t)}function cH(e,t){return bo(e,t)}function dH(e,t){return bo(e,t)}function hH(e,t){return bo(e,t)}var YN={};Ae(YN,{modelFromJSON:()=>bU});var ZN={};Ae(ZN,{l1:()=>fH,l1l2:()=>mH,l2:()=>gH});function mH(e){return new qc(e)}function fH(e){return TU(e)}function gH(e){return NU(e)}var JN=class extends dl{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Nr))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function Yd(e,t){return e<t}function Ck(e,t){return e>t}var QN=class extends JN{constructor(e){if(super(),e==null&&(e={}),e.restoreBestWeights)throw new Re("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,["auto","min","max"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),this.mode==="min"?this.monitorFunc=Yd:this.mode==="max"?this.monitorFunc=Ck:this.monitor.indexOf("acc")!==-1?this.monitorFunc=Ck:this.monitorFunc=Yd,this.monitorFunc===Yd&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===Yd?1/0:-1/0}async onEpochEnd(e,t){await jr(t);let n=this.getMonitorValue(t);n!=null&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){e==null&&(e={});let t=e[this.monitor];return t==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};function yH(e){return new QN(e)}var bH={earlyStopping:yH},xH=H();xH.registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")});var Sa;(function(e){e[e.DT_INVALID=0]="DT_INVALID",e[e.DT_FLOAT=1]="DT_FLOAT",e[e.DT_DOUBLE=2]="DT_DOUBLE",e[e.DT_INT32=3]="DT_INT32",e[e.DT_UINT8=4]="DT_UINT8",e[e.DT_INT16=5]="DT_INT16",e[e.DT_INT8=6]="DT_INT8",e[e.DT_STRING=7]="DT_STRING",e[e.DT_COMPLEX64=8]="DT_COMPLEX64",e[e.DT_INT64=9]="DT_INT64",e[e.DT_BOOL=10]="DT_BOOL",e[e.DT_QINT8=11]="DT_QINT8",e[e.DT_QUINT8=12]="DT_QUINT8",e[e.DT_QINT32=13]="DT_QINT32",e[e.DT_BFLOAT16=14]="DT_BFLOAT16",e[e.DT_QINT16=15]="DT_QINT16",e[e.DT_QUINT16=16]="DT_QUINT16",e[e.DT_UINT16=17]="DT_UINT16",e[e.DT_COMPLEX128=18]="DT_COMPLEX128",e[e.DT_HALF=19]="DT_HALF",e[e.DT_RESOURCE=20]="DT_RESOURCE",e[e.DT_VARIANT=21]="DT_VARIANT",e[e.DT_UINT32=22]="DT_UINT32",e[e.DT_UINT64=23]="DT_UINT64",e[e.DT_FLOAT_REF=101]="DT_FLOAT_REF",e[e.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",e[e.DT_INT32_REF=103]="DT_INT32_REF",e[e.DT_UINT8_REF=104]="DT_UINT8_REF",e[e.DT_INT16_REF=105]="DT_INT16_REF",e[e.DT_INT8_REF=106]="DT_INT8_REF",e[e.DT_STRING_REF=107]="DT_STRING_REF",e[e.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",e[e.DT_INT64_REF=109]="DT_INT64_REF",e[e.DT_BOOL_REF=110]="DT_BOOL_REF",e[e.DT_QINT8_REF=111]="DT_QINT8_REF",e[e.DT_QUINT8_REF=112]="DT_QUINT8_REF",e[e.DT_QINT32_REF=113]="DT_QINT32_REF",e[e.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",e[e.DT_QINT16_REF=115]="DT_QINT16_REF",e[e.DT_QUINT16_REF=116]="DT_QUINT16_REF",e[e.DT_UINT16_REF=117]="DT_UINT16_REF",e[e.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",e[e.DT_HALF_REF=119]="DT_HALF_REF",e[e.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",e[e.DT_VARIANT_REF=121]="DT_VARIANT_REF",e[e.DT_UINT32_REF=122]="DT_UINT32_REF",e[e.DT_UINT64_REF=123]="DT_UINT64_REF"})(Sa||(Sa={}));var _k;(function(e){let t;(function(n){n[n.LEGACY=0]="LEGACY",n[n.V1=1]="V1",n[n.V2=2]="V2"})(t=e.CheckpointFormatVersion||(e.CheckpointFormatVersion={}))})(_k||(_k={}));var v0={};function vH(e,t){let n={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};v0[e]=n}function e2(e){return v0[e]}function wH(e){delete v0[e]}function k(e,t,n,a,r){let s=t.inputParams[e];if(s&&s.inputIndexStart!==void 0){let o=s.inputIndexStart,l=s.inputIndexEnd===0?void 0:s.inputIndexEnd===void 0?o+1:s.inputIndexEnd;if(s.type==="tensor")return vn(t.inputNames[s.inputIndexStart],n,a,r);if(s.type==="tensors")return t.inputNames.slice(o,l).map(d=>vn(d,n,a,r));let u=vn(t.inputNames.slice(o)[0],n,a,r),p=u.dataSync();return s.type==="number"?p[0]:v.toNestedArray(u.shape,p)}let i=t.attrParams[e];return i&&i.value}function vn(e,t,n,a){let[r,s]=Xn(e);if(a!=null){let o=a.getHashTableHandleByName(r);if(o!=null)return o}let i=n.currentContextIds.find(o=>!!t[Rh(r,o)]);return i!==void 0?t[Rh(r,i)][s]:void 0}function kH(e,t,n){return t[Rh(e,n.currentContextId)]}function ar(e,t){let[n,a,r]=Xn(e);return[Rh(n,t&&t.currentContextId),a,r]}function Rh(e,t){return t?`${e}-${t}`:e}function Xn(e){let t=e.split(":");if(t.length===1)return[e,0,void 0];let n=t[0],a=t.length===3?t[1]:void 0,r=Number(t[t.length-1]);return[n,r,a]}function sh(e,t,n){let a=k("pad",e,t,n);if(a==="explicit"){a=k("explicitPaddings",e,t,n);let r=[[0,0],[0,0],[0,0],[0,0]];for(let s=0;s<4;s++)r[s][0]=a[s*2],r[s][1]=a[s*2+1];return r}return a}function kr(e){return e.kept?e:rr(e)}var t2={};Ae(t2,{json:()=>IH});var IH=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],n2={};Ae(n2,{json:()=>SH});var SH=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],a2={};Ae(a2,{json:()=>TH});var TH=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}],r2={};Ae(r2,{json:()=>NH});var NH=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],s2={};Ae(s2,{json:()=>CH});var CH=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomStandardNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],i2={};Ae(i2,{json:()=>_H});var _H=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],o2={};Ae(o2,{json:()=>EH});var EH=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],l2={};Ae(l2,{json:()=>AH});var AH=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],u2={};Ae(u2,{json:()=>$H});var $H=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}],p2={};Ae(p2,{json:()=>FH});var FH=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}],c2={};Ae(c2,{json:()=>DH});var DH=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],d2={};Ae(d2,{json:()=>RH});var RH=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}],h2={};Ae(h2,{json:()=>MH});var MH=[{tfOpName:"EuclideanNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",defaultValue:!1}]},{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}],m2={};Ae(m2,{json:()=>PH});var PH=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],f2={};Ae(f2,{json:()=>OH});var OH=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}],g2={};Ae(g2,{json:()=>LH});var LH=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],y2={};Ae(y2,{json:()=>zH});var zH=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],b2={};Ae(b2,{json:()=>WH});var WH=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],x2={};Ae(x2,{json:()=>BH});var BH=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],Ek=class{constructor(){let e=[t2,n2,a2,r2,s2,i2,o2,l2,u2,p2,c2,d2,h2,m2,f2,g2,y2,b2,x2],t=[].concat(...e.map(n=>n.json));this.opMappers=t.reduce((n,a)=>(n[a.tfOpName]=a,n),{})}static get Instance(){return this._instance||(this._instance=new this)}transformGraph(e,t={}){let n=e.node,a=[],r=[],s=[],i=n.reduce((m,f)=>(m[f.name]=this.mapNode(f),f.op.startsWith("Placeholder")?a.push(m[f.name]):f.op==="Const"?r.push(m[f.name]):(f.input==null||f.input.length===0)&&s.push(m[f.name]),m),{}),o=[],l=[],u={},p={};t!=null&&(u=this.mapSignatureEntries(t.inputs),p=this.mapSignatureEntries(t.outputs));let d=Object.keys(i);d.forEach(m=>{let f=i[m];f.inputNames.forEach((g,y)=>{let[b,,x]=ar(g),w=i[b];if(w.outputs!=null){let I=w.outputs.indexOf(x);if(I!==-1){let T=`${b}:${I}`;f.inputNames[y]=T}}f.inputs.push(w),w.children.push(f)})}),Object.keys(p).length===0?d.forEach(m=>{let f=i[m];f.children.length===0&&l.push(f)}):Object.keys(p).forEach(m=>{let[f]=ar(m),g=i[f];g!=null&&(g.signatureKey=p[m],l.push(g))}),Object.keys(u).length>0?Object.keys(u).forEach(m=>{let[f]=ar(m),g=i[f];g&&(g.signatureKey=u[m],o.push(g))}):o=a;let c={};e.library!=null&&e.library.function!=null&&(c=e.library.function.reduce((m,f)=>(m[f.signature.name]=this.mapFunction(f),m),{}));let h={nodes:i,inputs:o,outputs:l,weights:r,placeholders:a,signature:t,functions:c};return s.length>0&&(h.initNodes=s),h}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,n)=>(t[e[n].name]=n,t),{})}mapNode(e){let t=e2(e.op)||this.opMappers[e.op]||{};e.attr==null&&(e.attr={});let n={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(a=>a.startsWith("^")?a.slice(1):a),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return t.inputs!=null&&(n.inputParams=t.inputs.reduce((a,r)=>(a[r.name]={type:r.type,inputIndexStart:r.start,inputIndexEnd:r.end},a),{})),t.attrs!=null&&(n.attrParams=t.attrs.reduce((a,r)=>{let s=r.type,i;switch(r.type){case"string":i=Hb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Hb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"string[]":i=Jb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Jb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"number":i=qb(e.attr,r.tfName,r.defaultValue||0),i===void 0&&!!r.tfDeprecatedName&&(i=qb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"number[]":i=Zb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Zb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"bool":i=jb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=jb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"bool[]":i=ex(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=ex(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"shape":i=Yb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Yb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"shape[]":i=Qb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Qb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"dtype":i=Kb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Kb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"dtype[]":i=Xb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Xb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"func":i=Ak(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Ak(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${r.type} for op: ${e.op}`)}return a[r.name]={value:i,type:s},a},{})),n}mapFunction(e){let t=e.nodeDef,n=[],a=[],r={};t!=null&&(r=t.reduce((u,p)=>(u[p.name]=this.mapNode(p),p.op==="Const"&&a.push(u[p.name]),u),{}));let s=[],i=[];e.signature.inputArg.forEach(u=>{let[p]=ar(u.name),d={name:p,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:w0(u.type),type:"dtype"}},children:[]};d.signatureKey=u.name,s.push(d),r[p]=d}),Object.keys(r).forEach(u=>{let p=r[u];p.inputNames.forEach((d,c)=>{let[h,,m]=ar(d),f=r[h];if(f.outputs!=null){let g=f.outputs.indexOf(m);if(g!==-1){let y=`${h}:${g}`;p.inputNames[c]=y}}p.inputs.push(f),f.children.push(p)})});let o=e.ret;e.signature.outputArg.forEach(u=>{let[p,d]=ar(o[u.name]),c=r[p];c!=null&&(c.defaultOutput=d,i.push(c))});let l=this.mapArgsToSignature(e);return{nodes:r,inputs:s,outputs:i,weights:a,placeholders:n,signature:l}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n),t),{}),outputs:e.signature.outputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n,e.ret),t),{})}}mapArgToTensorInfo(e,t){let n=e.name;return t!=null&&(n=t[n]),{name:n,dtype:e.type}}};function VH(e){let t=H().global;if(typeof t.atob!="undefined")return t.atob(e);if(typeof Buffer!="undefined")return new Buffer(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}function v2(e,t){let n=Array.isArray(e)?String.fromCharCode.apply(null,e):VH(e);return t?n:n.toLowerCase()}function Hb(e,t,n,a=!1){let r=e[t];return r!=null?v2(r.s,a):n}function jb(e,t,n){let a=e[t];return a?a.b:n}function qb(e,t,n){let a=e[t]||{},r=a.i!=null?a.i:a.f!=null?a.f:n;return typeof r=="number"?r:parseInt(r,10)}function w0(e){switch(typeof e=="string"&&(e=Sa[e]),e){case Sa.DT_FLOAT:case Sa.DT_HALF:return"float32";case Sa.DT_INT32:case Sa.DT_INT64:case Sa.DT_INT8:case Sa.DT_UINT8:return"int32";case Sa.DT_BOOL:return"bool";case Sa.DT_DOUBLE:return"float32";case Sa.DT_STRING:return"string";default:return null}}function Ak(e,t,n){let a=e[t];return a&&a.func?a.func.name:n}function Kb(e,t,n){let a=e[t];return a&&a.type?w0(a.type):n}function Xb(e,t,n){let a=e[t];return a&&a.list&&a.list.type?a.list.type.map(r=>w0(r)):n}function w2(e){if(!e.unknownRank)return e.dim!=null?e.dim.map(t=>typeof t.size=="number"?t.size:parseInt(t.size,10)):[]}function Yb(e,t,n){let a=e[t];return a&&a.shape?w2(a.shape):n}function Zb(e,t,n){let a=e[t];return a?((a.list.f&&a.list.f.length?a.list.f:a.list.i)||[]).map(r=>typeof r=="number"?r:parseInt(r,10)):n}function Jb(e,t,n,a=!1){let r=e[t];return r&&r.list&&r.list.s?r.list.s.map(s=>v2(s,a)):n}function Qb(e,t,n){let a=e[t];return a&&a.list&&a.list.shape?a.list.shape.map(r=>w2(r)):n}function ex(e,t,n){let a=e[t];return a&&a.list&&a.list.b?a.list.b:n}var UH=class{constructor(e,t,n){this.node=e,this.tensorMap=t,this.context=n,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(a=>this.getInput(a)),e.rawAttrs!=null&&(this.attrs=Object.keys(e.rawAttrs).reduce((a,r)=>(a[r]=this.getAttr(r),a),{}))}getInput(e){return vn(e,this.tensorMap,this.context)}getAttr(e,t){let n=this.node.rawAttrs[e];if(n.tensor!=null)return vn(e,this.tensorMap,this.context);if(n.i!=null||n.f!=null)return qb(this.node.rawAttrs,e,t);if(n.s!=null)return Hb(this.node.rawAttrs,e,t);if(n.b!=null)return jb(this.node.rawAttrs,e,t);if(n.shape!=null)return Yb(this.node.rawAttrs,e,t);if(n.type!=null)return Kb(this.node.rawAttrs,e,t);if(n.list!=null){if(n.list.i!=null||n.list.f!=null)return Zb(this.node.rawAttrs,e,t);if(n.list.s!=null)return Jb(this.node.rawAttrs,e,t);if(n.list.shape!=null)return Qb(this.node.rawAttrs,e,t);if(n.list.b!=null)return ex(this.node.rawAttrs,e,t);if(n.list.type!=null)return Xb(this.node.rawAttrs,e,t)}return t}},un={};Ae(un,{OP_SCOPE_SUFFIX:()=>Cx,abs:()=>Lt,acos:()=>Bx,acosh:()=>Vx,add:()=>Y,addN:()=>CS,all:()=>$m,any:()=>qp,argMax:()=>ai,argMin:()=>Ux,asin:()=>Gx,asinh:()=>Hx,atan:()=>jx,atan2:()=>qx,atanh:()=>Kx,avgPool:()=>ga,avgPool3d:()=>Yx,basicLSTMCell:()=>$S,batchNorm:()=>ys,batchNorm2d:()=>Zx,batchNorm3d:()=>Jx,batchNorm4d:()=>Qx,batchToSpaceND:()=>Ac,bincount:()=>ev,booleanMaskAsync:()=>hT,broadcastArgs:()=>FS,broadcastTo:()=>Ks,buffer:()=>Pe,cast:()=>oe,ceil:()=>tv,clipByValue:()=>en,clone:()=>rr,complex:()=>Cr,concat:()=>Ze,concat1d:()=>nv,concat2d:()=>av,concat3d:()=>rv,concat4d:()=>sv,conv1d:()=>Fm,conv2d:()=>$t,conv2dTranspose:()=>Dm,conv3d:()=>ov,conv3dTranspose:()=>lv,cos:()=>$c,cosh:()=>Rm,cosineWindow:()=>tf,cumprod:()=>Kp,cumsum:()=>Mm,denseBincount:()=>Sh,depthToSpace:()=>uv,depthwiseConv2d:()=>bs,diag:()=>RS,dilation2d:()=>pv,div:()=>he,divNoNan:()=>cv,dot:()=>dv,dropout:()=>zv,einsum:()=>MS,elu:()=>Nu,enclosingPowerOfTwo:()=>Wv,equal:()=>Qn,erf:()=>hv,euclideanNorm:()=>gv,exp:()=>fn,expandDims:()=>Zt,expm1:()=>yv,eye:()=>Pm,fft:()=>zc,fill:()=>gn,floor:()=>_u,floorDiv:()=>Am,fused:()=>pl,gather:()=>Eu,gatherND:()=>yT,greater:()=>Nn,greaterEqual:()=>Er,ifft:()=>ul,imag:()=>Cc,image:()=>za,inTopKAsync:()=>bT,irfft:()=>Zm,isFinite:()=>bv,isInf:()=>xv,isNaN:()=>vv,leakyRelu:()=>Fc,less:()=>Om,lessEqual:()=>xs,linalg:()=>Uv,linspace:()=>WS,localResponseNormalization:()=>wv,log:()=>ea,log1p:()=>Dc,logSigmoid:()=>kv,logSoftmax:()=>zm,logSumExp:()=>Wm,logicalAnd:()=>Ea,logicalNot:()=>Rc,logicalOr:()=>Bm,logicalXor:()=>Iv,losses:()=>AT,lowerBound:()=>VS,matMul:()=>Fe,max:()=>ha,maxPool:()=>Dt,maxPool3d:()=>Sv,maxPoolWithArgmax:()=>US,maximum:()=>pr,mean:()=>Nt,meshgrid:()=>GS,min:()=>il,minimum:()=>Au,mirrorPad:()=>Tv,mod:()=>Nv,moments:()=>Mc,movingAverage:()=>mT,mul:()=>z,multiRNNCell:()=>HS,multinomial:()=>jS,neg:()=>bt,norm:()=>Cu,notEqual:()=>ii,oneHot:()=>rl,ones:()=>Zn,onesLike:()=>ta,op:()=>L,outerProduct:()=>qS,pad:()=>ya,pad1d:()=>KS,pad2d:()=>XS,pad3d:()=>YS,pad4d:()=>ZS,pool:()=>Cv,pow:()=>_r,prelu:()=>Oc,print:()=>$x,prod:()=>_v,raggedGather:()=>JS,raggedRange:()=>QS,raggedTensorToTensor:()=>eT,rand:()=>tT,randomGamma:()=>nT,randomNormal:()=>Um,randomStandardNormal:()=>aT,randomUniform:()=>$u,range:()=>ol,real:()=>sl,reciprocal:()=>$v,relu:()=>Xe,relu6:()=>Gm,reshape:()=>W,reverse:()=>fa,reverse1d:()=>rT,reverse2d:()=>sT,reverse3d:()=>iT,reverse4d:()=>oT,rfft:()=>Wc,round:()=>Hm,rsqrt:()=>jm,scalar:()=>be,scatterND:()=>fT,searchSorted:()=>Vm,selu:()=>qm,separableConv2d:()=>vs,setdiff1dAsync:()=>lT,sigmoid:()=>da,sign:()=>Fv,signal:()=>ET,sin:()=>Km,sinh:()=>Xm,slice:()=>Be,slice1d:()=>Lc,slice2d:()=>Ym,slice3d:()=>mo,slice4d:()=>ll,softmax:()=>Ka,softplus:()=>ho,spaceToBatchND:()=>Pc,sparse:()=>$T,sparseToDense:()=>gT,spectral:()=>_T,split:()=>zn,sqrt:()=>ln,square:()=>ot,squaredDifference:()=>Jm,squeeze:()=>ws,stack:()=>Ft,step:()=>fo,stridedSlice:()=>Dv,string:()=>FT,sub:()=>pe,sum:()=>fe,tan:()=>Rv,tanh:()=>ri,tensor:()=>kn,tensor1d:()=>Ke,tensor2d:()=>_a,tensor3d:()=>_c,tensor4d:()=>$a,tensor5d:()=>uT,tensor6d:()=>pT,tile:()=>Ln,topk:()=>Mv,transpose:()=>Ee,truncatedNormal:()=>Qm,unique:()=>Pv,unsortedSegmentSum:()=>ef,unstack:()=>ct,upperBound:()=>cT,variable:()=>Ov,where:()=>mn,whereAsync:()=>Lv,zeros:()=>It,zerosLike:()=>qe});var GH=(e,t,n,a=un)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[a.add(k("a",e,t,n),k("b",e,t,n))];case"AddN":return[a.addN(k("tensors",e,t,n))];case"FloorMod":case"Mod":return[a.mod(k("a",e,t,n),k("b",e,t,n))];case"Mul":return[a.mul(k("a",e,t,n),k("b",e,t,n))];case"RealDiv":case"Div":return[a.div(k("a",e,t,n),k("b",e,t,n))];case"DivNoNan":return[a.divNoNan(k("a",e,t,n),k("b",e,t,n))];case"FloorDiv":return[a.floorDiv(k("a",e,t,n),k("b",e,t,n))];case"Sub":return[a.sub(k("a",e,t,n),k("b",e,t,n))];case"Minimum":return[a.minimum(k("a",e,t,n),k("b",e,t,n))];case"Maximum":return[a.maximum(k("a",e,t,n),k("b",e,t,n))];case"Pow":return[a.pow(k("a",e,t,n),k("b",e,t,n))];case"SquaredDifference":return[a.squaredDifference(k("a",e,t,n),k("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},HH=(e,t,n,a=un)=>{switch(e.op){case"Abs":case"ComplexAbs":return[a.abs(k("x",e,t,n))];case"Acos":return[a.acos(k("x",e,t,n))];case"Acosh":return[a.acosh(k("x",e,t,n))];case"Asin":return[a.asin(k("x",e,t,n))];case"Asinh":return[a.asinh(k("x",e,t,n))];case"Atan":return[a.atan(k("x",e,t,n))];case"Atan2":return[a.atan2(k("x",e,t,n),k("y",e,t,n))];case"Atanh":return[a.atanh(k("x",e,t,n))];case"Ceil":return[a.ceil(k("x",e,t,n))];case"Complex":return[a.complex(k("real",e,t,n),k("imag",e,t,n))];case"Cos":return[a.cos(k("x",e,t,n))];case"Cosh":return[a.cosh(k("x",e,t,n))];case"Elu":return[a.elu(k("x",e,t,n))];case"Erf":return[a.erf(k("x",e,t,n))];case"Exp":return[a.exp(k("x",e,t,n))];case"Expm1":return[a.expm1(k("x",e,t,n))];case"Floor":return[a.floor(k("x",e,t,n))];case"Log":return[a.log(k("x",e,t,n))];case"Log1p":return[a.log1p(k("x",e,t,n))];case"Imag":return[a.imag(k("x",e,t,n))];case"Neg":return[a.neg(k("x",e,t,n))];case"Reciprocal":return[a.reciprocal(k("x",e,t,n))];case"Real":return[a.real(k("x",e,t,n))];case"Relu":return[a.relu(k("x",e,t,n))];case"Round":return[a.round(k("x",e,t,n))];case"Selu":return[a.selu(k("x",e,t,n))];case"Sigmoid":return[a.sigmoid(k("x",e,t,n))];case"Sin":return[a.sin(k("x",e,t,n))];case"Sign":return[a.sign(k("x",e,t,n))];case"Sinh":return[a.sinh(k("x",e,t,n))];case"Softplus":return[a.softplus(k("x",e,t,n))];case"Sqrt":return[a.sqrt(k("x",e,t,n))];case"Square":return[a.square(k("x",e,t,n))];case"Tanh":return[a.tanh(k("x",e,t,n))];case"Tan":return[a.tan(k("x",e,t,n))];case"ClipByValue":return[a.clipByValue(k("x",e,t,n),k("clipValueMin",e,t,n),k("clipValueMax",e,t,n))];case"Relu6":return[a.relu6(k("x",e,t,n))];case"Rsqrt":return[a.rsqrt(vn(e.inputNames[0],t,n))];case"Prod":return[a.prod(k("x",e,t,n),k("axes",e,t,n))];case"LeakyRelu":return[a.leakyRelu(k("x",e,t,n),k("alpha",e,t,n))];case"Prelu":return[a.prelu(k("x",e,t,n),k("alpha",e,t,n))];case"IsNan":return[a.isNaN(vn(e.inputNames[0],t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Ca(e,t,n=""){if(!(typeof e=="number"||typeof t=="number")){v.assert(e.length===t.length,()=>n+` Shapes ${e} and ${t} must match`);for(let a=0;a<e.length;a++){let r=e[a],s=t[a];v.assert(r<0||s<0||r===s,()=>n+` Shapes ${e} and ${t} must match`)}}}function $k(e){return!(typeof e=="number"||e.some(t=>t<0))}function Tp(e,t,n){let a=tx(e,n),r=!$k(a);if(r&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${a}`);if(r&&t.forEach(s=>{a=tx(s.shape,a)}),!$k(a))throw new Error(`Non-fully-defined elementShape: ${a}`);return a}function tx(e,t){if(typeof e=="number")return t;if(typeof t=="number")return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);let n=[];for(let a=0;a<e.length;++a){let r=e[a],s=t[a];if(r>=0&&s>=0&&r!==s)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);n[a]=r>=0?r:s}return n}var jH=class{constructor(e,t,n,a,r,s,i){this.name=e,this.dtype=t,this.maxSize=n,this.elementShape=a,this.identicalElementShapes=r,this.dynamicSize=s,this.clearAfterRead=i,this.tensors=[],this.closed_=!1,this.idTensor=be(0),Jt(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.tensor.id))&&t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);let t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(t=>this.read(t))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},
-          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=t.shape),Ca(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);n.tensor=t,Jt(t),n.written=!0,this.tensors[e]=n}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((n,a)=>this.write(n,t[a]))}gather(e,t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let a=0;a<this.size();a++)e.push(a)}if(e.length===0)return kn([],[0].concat(this.elementShape));let n=this.readMany(e);return Ca(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),Ft(n,0)}concat(e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(this.size()===0)return kn([],[0].concat(this.elementShape));let t=[];for(let a=0;a<this.size();a++)t.push(a);let n=this.readMany(t);return Ca(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),Ze(n,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);let n=Math.max(...e);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(e,ct(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let n=0,a=e.map(o=>(n+=o,n));if(n!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to
+var __defProp = Object.defineProperty;
+var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
+}) : x)(function(x) {
+  if (typeof require !== "undefined")
+    return require.apply(this, arguments);
+  throw new Error('Dynamic require of "' + x + '" is not supported');
+});
+var __export = (target, all5) => {
+  for (var name in all5)
+    __defProp(target, name, { get: all5[name], enumerable: true });
+};
+
+// dist/tfjs.esm.js
+var tfjs_esm_exports = {};
+__export(tfjs_esm_exports, {
+  Abs: () => Abs,
+  Acos: () => Acos,
+  Acosh: () => Acosh,
+  AdadeltaOptimizer: () => AdadeltaOptimizer,
+  AdagradOptimizer: () => AdagradOptimizer,
+  AdamOptimizer: () => AdamOptimizer,
+  AdamaxOptimizer: () => AdamaxOptimizer,
+  Add: () => Add,
+  AddN: () => AddN,
+  All: () => All,
+  Any: () => Any,
+  ArgMax: () => ArgMax,
+  ArgMin: () => ArgMin,
+  Asin: () => Asin,
+  Asinh: () => Asinh,
+  Atan: () => Atan,
+  Atan2: () => Atan2,
+  Atanh: () => Atanh,
+  AvgPool: () => AvgPool,
+  AvgPool3D: () => AvgPool3D,
+  AvgPool3DGrad: () => AvgPool3DGrad,
+  AvgPoolGrad: () => AvgPoolGrad,
+  BackendWasm: () => BackendWasm,
+  BatchMatMul: () => BatchMatMul,
+  BatchToSpaceND: () => BatchToSpaceND,
+  Bincount: () => Bincount,
+  BroadcastArgs: () => BroadcastArgs,
+  BroadcastTo: () => BroadcastTo,
+  Callback: () => Callback,
+  CallbackList: () => CallbackList,
+  Cast: () => Cast,
+  Ceil: () => Ceil,
+  ClipByValue: () => ClipByValue,
+  Complex: () => Complex,
+  ComplexAbs: () => ComplexAbs,
+  Concat: () => Concat,
+  Conv2D: () => Conv2D,
+  Conv2DBackpropFilter: () => Conv2DBackpropFilter,
+  Conv2DBackpropInput: () => Conv2DBackpropInput,
+  Conv3D: () => Conv3D,
+  Conv3DBackpropFilterV2: () => Conv3DBackpropFilterV2,
+  Conv3DBackpropInputV2: () => Conv3DBackpropInputV2,
+  Cos: () => Cos,
+  Cosh: () => Cosh,
+  CropAndResize: () => CropAndResize,
+  Cumprod: () => Cumprod,
+  Cumsum: () => Cumsum,
+  CustomCallback: () => CustomCallback,
+  DataStorage: () => DataStorage,
+  DenseBincount: () => DenseBincount,
+  DepthToSpace: () => DepthToSpace,
+  DepthwiseConv2dNative: () => DepthwiseConv2dNative,
+  DepthwiseConv2dNativeBackpropFilter: () => DepthwiseConv2dNativeBackpropFilter,
+  DepthwiseConv2dNativeBackpropInput: () => DepthwiseConv2dNativeBackpropInput,
+  Diag: () => Diag,
+  Dilation2D: () => Dilation2D,
+  Dilation2DBackpropFilter: () => Dilation2DBackpropFilter,
+  Dilation2DBackpropInput: () => Dilation2DBackpropInput,
+  ENV: () => ENV,
+  EarlyStopping: () => EarlyStopping,
+  Einsum: () => Einsum,
+  Elu: () => Elu,
+  EluGrad: () => EluGrad,
+  Environment: () => Environment,
+  Equal: () => Equal,
+  Erf: () => Erf,
+  Exp: () => Exp,
+  ExpandDims: () => ExpandDims,
+  Expm1: () => Expm1,
+  FFT: () => FFT,
+  Fill: () => Fill,
+  FlipLeftRight: () => FlipLeftRight,
+  Floor: () => Floor,
+  FloorDiv: () => FloorDiv,
+  FromPixels: () => FromPixels,
+  FusedBatchNorm: () => FusedBatchNorm,
+  FusedConv2D: () => FusedConv2D,
+  FusedDepthwiseConv2D: () => FusedDepthwiseConv2D,
+  GPGPUContext: () => GPGPUContext,
+  GatherNd: () => GatherNd,
+  GatherV2: () => GatherV2,
+  GraphModel: () => GraphModel,
+  Greater: () => Greater,
+  GreaterEqual: () => GreaterEqual,
+  History: () => History,
+  IFFT: () => IFFT,
+  Identity: () => Identity,
+  Imag: () => Imag,
+  InputSpec: () => InputSpec,
+  IsFinite: () => IsFinite,
+  IsInf: () => IsInf,
+  IsNan: () => IsNan,
+  KernelBackend: () => KernelBackend,
+  LRN: () => LRN,
+  LRNGrad: () => LRNGrad,
+  LayerVariable: () => LayerVariable,
+  LayersModel: () => LayersModel,
+  LeakyRelu: () => LeakyRelu,
+  Less: () => Less,
+  LessEqual: () => LessEqual,
+  LinSpace: () => LinSpace,
+  Log: () => Log,
+  Log1p: () => Log1p,
+  LogSoftmax: () => LogSoftmax,
+  LogicalAnd: () => LogicalAnd,
+  LogicalNot: () => LogicalNot,
+  LogicalOr: () => LogicalOr,
+  LogicalXor: () => LogicalXor,
+  LowerBound: () => LowerBound,
+  MathBackendWebGL: () => MathBackendWebGL,
+  Max: () => Max,
+  MaxPool: () => MaxPool,
+  MaxPool3D: () => MaxPool3D,
+  MaxPool3DGrad: () => MaxPool3DGrad,
+  MaxPoolGrad: () => MaxPoolGrad,
+  MaxPoolWithArgmax: () => MaxPoolWithArgmax,
+  Maximum: () => Maximum,
+  Mean: () => Mean,
+  Min: () => Min,
+  Minimum: () => Minimum,
+  MirrorPad: () => MirrorPad,
+  Mod: () => Mod,
+  MomentumOptimizer: () => MomentumOptimizer,
+  Multinomial: () => Multinomial,
+  Multiply: () => Multiply,
+  Neg: () => Neg,
+  NonMaxSuppressionV3: () => NonMaxSuppressionV3,
+  NonMaxSuppressionV4: () => NonMaxSuppressionV4,
+  NonMaxSuppressionV5: () => NonMaxSuppressionV5,
+  NotEqual: () => NotEqual,
+  OP_SCOPE_SUFFIX: () => OP_SCOPE_SUFFIX,
+  OneHot: () => OneHot,
+  OnesLike: () => OnesLike,
+  Optimizer: () => Optimizer,
+  OptimizerConstructors: () => OptimizerConstructors,
+  Pack: () => Pack,
+  PadV2: () => PadV2,
+  Pool: () => Pool,
+  Pow: () => Pow,
+  Prelu: () => Prelu,
+  Prod: () => Prod,
+  RMSPropOptimizer: () => RMSPropOptimizer,
+  RNN: () => RNN,
+  RaggedGather: () => RaggedGather,
+  RaggedRange: () => RaggedRange,
+  RaggedTensorToTensor: () => RaggedTensorToTensor,
+  Range: () => Range,
+  Rank: () => Rank,
+  Real: () => Real,
+  RealDiv: () => RealDiv,
+  Reciprocal: () => Reciprocal,
+  Reduction: () => Reduction,
+  Relu: () => Relu,
+  Relu6: () => Relu6,
+  Reshape: () => Reshape,
+  ResizeBilinear: () => ResizeBilinear,
+  ResizeBilinearGrad: () => ResizeBilinearGrad,
+  ResizeNearestNeighbor: () => ResizeNearestNeighbor,
+  ResizeNearestNeighborGrad: () => ResizeNearestNeighborGrad,
+  Reverse: () => Reverse,
+  RotateWithOffset: () => RotateWithOffset,
+  Round: () => Round,
+  Rsqrt: () => Rsqrt,
+  SGDOptimizer: () => SGDOptimizer,
+  ScatterNd: () => ScatterNd,
+  SearchSorted: () => SearchSorted,
+  Select: () => Select,
+  Selu: () => Selu,
+  Sequential: () => Sequential,
+  Sigmoid: () => Sigmoid,
+  Sign: () => Sign,
+  Sin: () => Sin,
+  Sinh: () => Sinh,
+  Slice: () => Slice,
+  Softmax: () => Softmax,
+  Softplus: () => Softplus,
+  SpaceToBatchND: () => SpaceToBatchND,
+  SparseFillEmptyRows: () => SparseFillEmptyRows,
+  SparseReshape: () => SparseReshape,
+  SparseSegmentMean: () => SparseSegmentMean,
+  SparseSegmentSum: () => SparseSegmentSum,
+  SparseToDense: () => SparseToDense,
+  SplitV: () => SplitV,
+  Sqrt: () => Sqrt,
+  Square: () => Square,
+  SquaredDifference: () => SquaredDifference,
+  Step: () => Step,
+  StridedSlice: () => StridedSlice,
+  StringNGrams: () => StringNGrams,
+  StringSplit: () => StringSplit,
+  StringToHashBucketFast: () => StringToHashBucketFast,
+  Sub: () => Sub,
+  Sum: () => Sum,
+  SymbolicTensor: () => SymbolicTensor,
+  Tan: () => Tan,
+  Tanh: () => Tanh,
+  Tensor: () => Tensor,
+  TensorBuffer: () => TensorBuffer,
+  Tile: () => Tile,
+  TopK: () => TopK,
+  Transform: () => Transform,
+  Transpose: () => Transpose,
+  Unique: () => Unique,
+  Unpack: () => Unpack,
+  UnsortedSegmentSum: () => UnsortedSegmentSum,
+  UpperBound: () => UpperBound,
+  Variable: () => Variable,
+  ZerosLike: () => ZerosLike,
+  _FusedMatMul: () => _FusedMatMul,
+  abs: () => abs,
+  acos: () => acos,
+  acosh: () => acosh,
+  add: () => add2,
+  addN: () => addN,
+  all: () => all,
+  any: () => any,
+  argMax: () => argMax,
+  argMin: () => argMin,
+  asin: () => asin,
+  asinh: () => asinh,
+  atan: () => atan,
+  atan2: () => atan2,
+  atanh: () => atanh,
+  avgPool: () => avgPool,
+  avgPool3d: () => avgPool3d,
+  backend: () => backend,
+  backend_util: () => backend_util_exports,
+  basicLSTMCell: () => basicLSTMCell,
+  batchNorm: () => batchNorm,
+  batchNorm2d: () => batchNorm2d,
+  batchNorm3d: () => batchNorm3d,
+  batchNorm4d: () => batchNorm4d,
+  batchToSpaceND: () => batchToSpaceND,
+  bincount: () => bincount,
+  booleanMaskAsync: () => booleanMaskAsync,
+  broadcastArgs: () => broadcastArgs,
+  broadcastTo: () => broadcastTo,
+  broadcast_util: () => broadcast_util_exports,
+  browser: () => browser_exports,
+  buffer: () => buffer,
+  callbacks: () => callbacks,
+  cast: () => cast,
+  ceil: () => ceil,
+  clipByValue: () => clipByValue,
+  clone: () => clone,
+  complex: () => complex,
+  concat: () => concat,
+  concat1d: () => concat1d,
+  concat2d: () => concat2d,
+  concat3d: () => concat3d,
+  concat4d: () => concat4d,
+  constraints: () => exports_constraints_exports,
+  conv1d: () => conv1d,
+  conv2d: () => conv2d,
+  conv2dTranspose: () => conv2dTranspose,
+  conv3d: () => conv3d,
+  conv3dTranspose: () => conv3dTranspose,
+  copyRegisteredKernels: () => copyRegisteredKernels,
+  cos: () => cos,
+  cosh: () => cosh,
+  cosineWindow: () => cosineWindow,
+  cumprod: () => cumprod,
+  cumsum: () => cumsum,
+  customGrad: () => customGrad,
+  data: () => dist_exports2,
+  denseBincount: () => denseBincount,
+  deprecationWarn: () => deprecationWarn,
+  depthToSpace: () => depthToSpace,
+  depthwiseConv2d: () => depthwiseConv2d,
+  deregisterOp: () => deregisterOp,
+  device_util: () => device_util_exports,
+  diag: () => diag,
+  dilation2d: () => dilation2d,
+  disableDeprecationWarnings: () => disableDeprecationWarnings,
+  dispose: () => dispose,
+  disposeVariables: () => disposeVariables,
+  div: () => div,
+  divNoNan: () => divNoNan,
+  dot: () => dot,
+  dropout: () => dropout,
+  einsum: () => einsum,
+  elu: () => elu,
+  enableDebugMode: () => enableDebugMode,
+  enableProdMode: () => enableProdMode,
+  enclosingPowerOfTwo: () => enclosingPowerOfTwo,
+  engine: () => engine,
+  env: () => env,
+  equal: () => equal,
+  erf: () => erf,
+  euclideanNorm: () => euclideanNorm,
+  exp: () => exp,
+  expandDims: () => expandDims,
+  expm1: () => expm1,
+  eye: () => eye,
+  fft: () => fft,
+  fill: () => fill,
+  findBackend: () => findBackend,
+  findBackendFactory: () => findBackendFactory,
+  floor: () => floor,
+  floorDiv: () => floorDiv,
+  forceHalfFloat: () => forceHalfFloat,
+  fused: () => fused_ops_exports,
+  gather: () => gather,
+  gatherND: () => gatherND,
+  gather_util: () => gather_nd_util_exports,
+  getBackend: () => getBackend,
+  getGradient: () => getGradient,
+  getKernel: () => getKernel,
+  getKernelsForBackend: () => getKernelsForBackend,
+  getThreadsCount: () => getThreadsCount,
+  gpgpu_util: () => gpgpu_util_exports,
+  grad: () => grad,
+  grads: () => grads,
+  greater: () => greater,
+  greaterEqual: () => greaterEqual,
+  ifft: () => ifft,
+  imag: () => imag,
+  image: () => image,
+  inTopKAsync: () => inTopKAsync,
+  initializers: () => exports_initializers_exports,
+  input: () => input,
+  io: () => io_exports,
+  irfft: () => irfft,
+  isFinite: () => isFinite2,
+  isInf: () => isInf,
+  isNaN: () => isNaN2,
+  keep: () => keep,
+  kernel_impls: () => kernel_impls_exports,
+  layers: () => exports_layers_exports,
+  leakyRelu: () => leakyRelu,
+  less: () => less,
+  lessEqual: () => lessEqual,
+  linalg: () => linalg,
+  linspace: () => linspace,
+  loadGraphModel: () => loadGraphModel,
+  loadGraphModelSync: () => loadGraphModelSync,
+  loadLayersModel: () => loadLayersModel,
+  localResponseNormalization: () => localResponseNormalization,
+  log: () => log2,
+  log1p: () => log1p,
+  logSigmoid: () => logSigmoid,
+  logSoftmax: () => logSoftmax,
+  logSumExp: () => logSumExp,
+  logicalAnd: () => logicalAnd,
+  logicalNot: () => logicalNot,
+  logicalOr: () => logicalOr,
+  logicalXor: () => logicalXor,
+  losses: () => losses,
+  lowerBound: () => lowerBound,
+  matMul: () => matMul,
+  math: () => math_exports,
+  max: () => max,
+  maxPool: () => maxPool,
+  maxPool3d: () => maxPool3d,
+  maxPoolWithArgmax: () => maxPoolWithArgmax,
+  maximum: () => maximum,
+  mean: () => mean,
+  memory: () => memory,
+  meshgrid: () => meshgrid,
+  metrics: () => exports_metrics_exports,
+  min: () => min,
+  minimum: () => minimum,
+  mirrorPad: () => mirrorPad,
+  mod: () => mod,
+  model: () => model,
+  models: () => exports_models_exports,
+  moments: () => moments,
+  movingAverage: () => movingAverage,
+  mul: () => mul,
+  multiRNNCell: () => multiRNNCell,
+  multinomial: () => multinomial,
+  neg: () => neg,
+  nextFrame: () => nextFrame,
+  norm: () => norm,
+  notEqual: () => notEqual,
+  oneHot: () => oneHot,
+  ones: () => ones2,
+  onesLike: () => onesLike,
+  op: () => op,
+  outerProduct: () => outerProduct,
+  pad: () => pad,
+  pad1d: () => pad1d,
+  pad2d: () => pad2d,
+  pad3d: () => pad3d,
+  pad4d: () => pad4d,
+  pool: () => pool,
+  pow: () => pow,
+  prelu: () => prelu,
+  print: () => print,
+  prod: () => prod,
+  profile: () => profile,
+  raggedGather: () => raggedGather,
+  raggedRange: () => raggedRange,
+  raggedTensorToTensor: () => raggedTensorToTensor,
+  rand: () => rand,
+  randomGamma: () => randomGamma,
+  randomNormal: () => randomNormal,
+  randomStandardNormal: () => randomStandardNormal,
+  randomUniform: () => randomUniform,
+  range: () => range,
+  ready: () => ready,
+  real: () => real,
+  reciprocal: () => reciprocal,
+  registerBackend: () => registerBackend,
+  registerCallbackConstructor: () => registerCallbackConstructor,
+  registerGradient: () => registerGradient,
+  registerKernel: () => registerKernel,
+  registerOp: () => registerOp,
+  regularizers: () => exports_regularizers_exports,
+  relu: () => relu,
+  relu6: () => relu6,
+  removeBackend: () => removeBackend,
+  reshape: () => reshape,
+  reverse: () => reverse,
+  reverse1d: () => reverse1d,
+  reverse2d: () => reverse2d,
+  reverse3d: () => reverse3d,
+  reverse4d: () => reverse4d,
+  rfft: () => rfft,
+  round: () => round2,
+  rsqrt: () => rsqrt,
+  scalar: () => scalar,
+  scatterND: () => scatterND,
+  scatter_util: () => scatter_nd_util_exports,
+  searchSorted: () => searchSorted,
+  selu: () => selu,
+  separableConv2d: () => separableConv2d,
+  sequential: () => sequential,
+  serialization: () => serialization_exports,
+  setBackend: () => setBackend,
+  setPlatform: () => setPlatform,
+  setThreadsCount: () => setThreadsCount,
+  setWasmPath: () => setWasmPath,
+  setWasmPaths: () => setWasmPaths,
+  setWebGLContext: () => setWebGLContext,
+  setdiff1dAsync: () => setdiff1dAsync,
+  sigmoid: () => sigmoid,
+  sign: () => sign,
+  signal: () => signal,
+  sin: () => sin,
+  sinh: () => sinh,
+  slice: () => slice,
+  slice1d: () => slice1d,
+  slice2d: () => slice2d,
+  slice3d: () => slice3d,
+  slice4d: () => slice4d,
+  slice_util: () => slice_util_exports,
+  softmax: () => softmax,
+  softplus: () => softplus,
+  spaceToBatchND: () => spaceToBatchND,
+  sparse: () => sparse,
+  sparseToDense: () => sparseToDense,
+  spectral: () => spectral,
+  split: () => split,
+  sqrt: () => sqrt,
+  square: () => square,
+  squaredDifference: () => squaredDifference,
+  squeeze: () => squeeze,
+  stack: () => stack,
+  step: () => step,
+  stridedSlice: () => stridedSlice,
+  string: () => string,
+  sub: () => sub,
+  sum: () => sum2,
+  sumOutType: () => sumOutType,
+  tan: () => tan,
+  tanh: () => tanh2,
+  tensor: () => tensor,
+  tensor1d: () => tensor1d,
+  tensor2d: () => tensor2d,
+  tensor3d: () => tensor3d,
+  tensor4d: () => tensor4d,
+  tensor5d: () => tensor5d,
+  tensor6d: () => tensor6d,
+  tensor_util: () => tensor_util_exports,
+  test_util: () => test_util_exports,
+  tidy: () => tidy,
+  tile: () => tile,
+  time: () => time,
+  topk: () => topk,
+  train: () => train,
+  transpose: () => transpose,
+  truncatedNormal: () => truncatedNormal,
+  unique: () => unique,
+  unregisterGradient: () => unregisterGradient,
+  unregisterKernel: () => unregisterKernel,
+  unsortedSegmentSum: () => unsortedSegmentSum,
+  unstack: () => unstack,
+  upcastType: () => upcastType,
+  upperBound: () => upperBound,
+  util: () => util_exports,
+  valueAndGrad: () => valueAndGrad,
+  valueAndGrads: () => valueAndGrads,
+  variable: () => variable,
+  variableGrads: () => variableGrads,
+  version: () => version62,
+  version_converter: () => version3,
+  version_core: () => version,
+  version_layers: () => version2,
+  version_wasm: () => version8,
+  version_webgl: () => version6,
+  webgl: () => webgl,
+  webgl_util: () => webgl_util_exports,
+  where: () => where,
+  whereAsync: () => whereAsync,
+  zeros: () => zeros,
+  zerosLike: () => zerosLike
+});
+var __create = Object.create;
+var __defProp2 = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __require2 = ((x) => typeof __require !== "undefined" ? __require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof __require !== "undefined" ? __require : a)[b]
+}) : x)(function(x) {
+  if (typeof __require !== "undefined")
+    return __require.apply(this, arguments);
+  throw new Error('Dynamic require of "' + x + '" is not supported');
+});
+var __commonJS = (cb, mod4) => function __require22() {
+  return mod4 || (0, cb[__getOwnPropNames(cb)[0]])((mod4 = { exports: {} }).exports, mod4), mod4.exports;
+};
+var __export2 = (target, all5) => {
+  for (var name in all5)
+    __defProp2(target, name, { get: all5[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp2(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod4, isNodeMode, target) => (target = mod4 != null ? __create(__getProtoOf(mod4)) : {}, __copyProps(
+  isNodeMode || !mod4 || !mod4.__esModule ? __defProp2(target, "default", { value: mod4, enumerable: true }) : target,
+  mod4
+));
+var require_long = __commonJS({
+  "node_modules/long/src/long.js"(exports, module) {
+    module.exports = Long2;
+    var wasm = null;
+    try {
+      wasm = new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([
+        0,
+        97,
+        115,
+        109,
+        1,
+        0,
+        0,
+        0,
+        1,
+        13,
+        2,
+        96,
+        0,
+        1,
+        127,
+        96,
+        4,
+        127,
+        127,
+        127,
+        127,
+        1,
+        127,
+        3,
+        7,
+        6,
+        0,
+        1,
+        1,
+        1,
+        1,
+        1,
+        6,
+        6,
+        1,
+        127,
+        1,
+        65,
+        0,
+        11,
+        7,
+        50,
+        6,
+        3,
+        109,
+        117,
+        108,
+        0,
+        1,
+        5,
+        100,
+        105,
+        118,
+        95,
+        115,
+        0,
+        2,
+        5,
+        100,
+        105,
+        118,
+        95,
+        117,
+        0,
+        3,
+        5,
+        114,
+        101,
+        109,
+        95,
+        115,
+        0,
+        4,
+        5,
+        114,
+        101,
+        109,
+        95,
+        117,
+        0,
+        5,
+        8,
+        103,
+        101,
+        116,
+        95,
+        104,
+        105,
+        103,
+        104,
+        0,
+        0,
+        10,
+        191,
+        1,
+        6,
+        4,
+        0,
+        35,
+        0,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        126,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        127,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        128,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        129,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        130,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11
+      ])), {}).exports;
+    } catch (e) {
+    }
+    function Long2(low, high, unsigned) {
+      this.low = low | 0;
+      this.high = high | 0;
+      this.unsigned = !!unsigned;
+    }
+    Long2.prototype.__isLong__;
+    Object.defineProperty(Long2.prototype, "__isLong__", { value: true });
+    function isLong(obj) {
+      return (obj && obj["__isLong__"]) === true;
+    }
+    Long2.isLong = isLong;
+    var INT_CACHE = {};
+    var UINT_CACHE = {};
+    function fromInt(value, unsigned) {
+      var obj, cachedObj, cache;
+      if (unsigned) {
+        value >>>= 0;
+        if (cache = 0 <= value && value < 256) {
+          cachedObj = UINT_CACHE[value];
+          if (cachedObj)
+            return cachedObj;
+        }
+        obj = fromBits(value, (value | 0) < 0 ? -1 : 0, true);
+        if (cache)
+          UINT_CACHE[value] = obj;
+        return obj;
+      } else {
+        value |= 0;
+        if (cache = -128 <= value && value < 128) {
+          cachedObj = INT_CACHE[value];
+          if (cachedObj)
+            return cachedObj;
+        }
+        obj = fromBits(value, value < 0 ? -1 : 0, false);
+        if (cache)
+          INT_CACHE[value] = obj;
+        return obj;
+      }
+    }
+    Long2.fromInt = fromInt;
+    function fromNumber(value, unsigned) {
+      if (isNaN(value))
+        return unsigned ? UZERO : ZERO;
+      if (unsigned) {
+        if (value < 0)
+          return UZERO;
+        if (value >= TWO_PWR_64_DBL)
+          return MAX_UNSIGNED_VALUE;
+      } else {
+        if (value <= -TWO_PWR_63_DBL)
+          return MIN_VALUE;
+        if (value + 1 >= TWO_PWR_63_DBL)
+          return MAX_VALUE;
+      }
+      if (value < 0)
+        return fromNumber(-value, unsigned).neg();
+      return fromBits(value % TWO_PWR_32_DBL | 0, value / TWO_PWR_32_DBL | 0, unsigned);
+    }
+    Long2.fromNumber = fromNumber;
+    function fromBits(lowBits, highBits, unsigned) {
+      return new Long2(lowBits, highBits, unsigned);
+    }
+    Long2.fromBits = fromBits;
+    var pow_dbl = Math.pow;
+    function fromString(str, unsigned, radix) {
+      if (str.length === 0)
+        throw Error("empty string");
+      if (str === "NaN" || str === "Infinity" || str === "+Infinity" || str === "-Infinity")
+        return ZERO;
+      if (typeof unsigned === "number") {
+        radix = unsigned, unsigned = false;
+      } else {
+        unsigned = !!unsigned;
+      }
+      radix = radix || 10;
+      if (radix < 2 || 36 < radix)
+        throw RangeError("radix");
+      var p2;
+      if ((p2 = str.indexOf("-")) > 0)
+        throw Error("interior hyphen");
+      else if (p2 === 0) {
+        return fromString(str.substring(1), unsigned, radix).neg();
+      }
+      var radixToPower = fromNumber(pow_dbl(radix, 8));
+      var result = ZERO;
+      for (var i = 0; i < str.length; i += 8) {
+        var size = Math.min(8, str.length - i), value = parseInt(str.substring(i, i + size), radix);
+        if (size < 8) {
+          var power = fromNumber(pow_dbl(radix, size));
+          result = result.mul(power).add(fromNumber(value));
+        } else {
+          result = result.mul(radixToPower);
+          result = result.add(fromNumber(value));
+        }
+      }
+      result.unsigned = unsigned;
+      return result;
+    }
+    Long2.fromString = fromString;
+    function fromValue(val, unsigned) {
+      if (typeof val === "number")
+        return fromNumber(val, unsigned);
+      if (typeof val === "string")
+        return fromString(val, unsigned);
+      return fromBits(val.low, val.high, typeof unsigned === "boolean" ? unsigned : val.unsigned);
+    }
+    Long2.fromValue = fromValue;
+    var TWO_PWR_16_DBL = 1 << 16;
+    var TWO_PWR_24_DBL = 1 << 24;
+    var TWO_PWR_32_DBL = TWO_PWR_16_DBL * TWO_PWR_16_DBL;
+    var TWO_PWR_64_DBL = TWO_PWR_32_DBL * TWO_PWR_32_DBL;
+    var TWO_PWR_63_DBL = TWO_PWR_64_DBL / 2;
+    var TWO_PWR_24 = fromInt(TWO_PWR_24_DBL);
+    var ZERO = fromInt(0);
+    Long2.ZERO = ZERO;
+    var UZERO = fromInt(0, true);
+    Long2.UZERO = UZERO;
+    var ONE = fromInt(1);
+    Long2.ONE = ONE;
+    var UONE = fromInt(1, true);
+    Long2.UONE = UONE;
+    var NEG_ONE = fromInt(-1);
+    Long2.NEG_ONE = NEG_ONE;
+    var MAX_VALUE = fromBits(4294967295 | 0, 2147483647 | 0, false);
+    Long2.MAX_VALUE = MAX_VALUE;
+    var MAX_UNSIGNED_VALUE = fromBits(4294967295 | 0, 4294967295 | 0, true);
+    Long2.MAX_UNSIGNED_VALUE = MAX_UNSIGNED_VALUE;
+    var MIN_VALUE = fromBits(0, 2147483648 | 0, false);
+    Long2.MIN_VALUE = MIN_VALUE;
+    var LongPrototype = Long2.prototype;
+    LongPrototype.toInt = function toInt() {
+      return this.unsigned ? this.low >>> 0 : this.low;
+    };
+    LongPrototype.toNumber = function toNumber() {
+      if (this.unsigned)
+        return (this.high >>> 0) * TWO_PWR_32_DBL + (this.low >>> 0);
+      return this.high * TWO_PWR_32_DBL + (this.low >>> 0);
+    };
+    LongPrototype.toString = function toString(radix) {
+      radix = radix || 10;
+      if (radix < 2 || 36 < radix)
+        throw RangeError("radix");
+      if (this.isZero())
+        return "0";
+      if (this.isNegative()) {
+        if (this.eq(MIN_VALUE)) {
+          var radixLong = fromNumber(radix), div3 = this.div(radixLong), rem1 = div3.mul(radixLong).sub(this);
+          return div3.toString(radix) + rem1.toInt().toString(radix);
+        } else
+          return "-" + this.neg().toString(radix);
+      }
+      var radixToPower = fromNumber(pow_dbl(radix, 6), this.unsigned), rem = this;
+      var result = "";
+      while (true) {
+        var remDiv = rem.div(radixToPower), intval = rem.sub(remDiv.mul(radixToPower)).toInt() >>> 0, digits = intval.toString(radix);
+        rem = remDiv;
+        if (rem.isZero())
+          return digits + result;
+        else {
+          while (digits.length < 6)
+            digits = "0" + digits;
+          result = "" + digits + result;
+        }
+      }
+    };
+    LongPrototype.getHighBits = function getHighBits() {
+      return this.high;
+    };
+    LongPrototype.getHighBitsUnsigned = function getHighBitsUnsigned() {
+      return this.high >>> 0;
+    };
+    LongPrototype.getLowBits = function getLowBits() {
+      return this.low;
+    };
+    LongPrototype.getLowBitsUnsigned = function getLowBitsUnsigned() {
+      return this.low >>> 0;
+    };
+    LongPrototype.getNumBitsAbs = function getNumBitsAbs() {
+      if (this.isNegative())
+        return this.eq(MIN_VALUE) ? 64 : this.neg().getNumBitsAbs();
+      var val = this.high != 0 ? this.high : this.low;
+      for (var bit = 31; bit > 0; bit--)
+        if ((val & 1 << bit) != 0)
+          break;
+      return this.high != 0 ? bit + 33 : bit + 1;
+    };
+    LongPrototype.isZero = function isZero() {
+      return this.high === 0 && this.low === 0;
+    };
+    LongPrototype.eqz = LongPrototype.isZero;
+    LongPrototype.isNegative = function isNegative() {
+      return !this.unsigned && this.high < 0;
+    };
+    LongPrototype.isPositive = function isPositive() {
+      return this.unsigned || this.high >= 0;
+    };
+    LongPrototype.isOdd = function isOdd() {
+      return (this.low & 1) === 1;
+    };
+    LongPrototype.isEven = function isEven22() {
+      return (this.low & 1) === 0;
+    };
+    LongPrototype.equals = function equals(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      if (this.unsigned !== other.unsigned && this.high >>> 31 === 1 && other.high >>> 31 === 1)
+        return false;
+      return this.high === other.high && this.low === other.low;
+    };
+    LongPrototype.eq = LongPrototype.equals;
+    LongPrototype.notEquals = function notEquals(other) {
+      return !this.eq(other);
+    };
+    LongPrototype.neq = LongPrototype.notEquals;
+    LongPrototype.ne = LongPrototype.notEquals;
+    LongPrototype.lessThan = function lessThan(other) {
+      return this.comp(other) < 0;
+    };
+    LongPrototype.lt = LongPrototype.lessThan;
+    LongPrototype.lessThanOrEqual = function lessThanOrEqual(other) {
+      return this.comp(other) <= 0;
+    };
+    LongPrototype.lte = LongPrototype.lessThanOrEqual;
+    LongPrototype.le = LongPrototype.lessThanOrEqual;
+    LongPrototype.greaterThan = function greaterThan(other) {
+      return this.comp(other) > 0;
+    };
+    LongPrototype.gt = LongPrototype.greaterThan;
+    LongPrototype.greaterThanOrEqual = function greaterThanOrEqual(other) {
+      return this.comp(other) >= 0;
+    };
+    LongPrototype.gte = LongPrototype.greaterThanOrEqual;
+    LongPrototype.ge = LongPrototype.greaterThanOrEqual;
+    LongPrototype.compare = function compare(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      if (this.eq(other))
+        return 0;
+      var thisNeg = this.isNegative(), otherNeg = other.isNegative();
+      if (thisNeg && !otherNeg)
+        return -1;
+      if (!thisNeg && otherNeg)
+        return 1;
+      if (!this.unsigned)
+        return this.sub(other).isNegative() ? -1 : 1;
+      return other.high >>> 0 > this.high >>> 0 || other.high === this.high && other.low >>> 0 > this.low >>> 0 ? -1 : 1;
+    };
+    LongPrototype.comp = LongPrototype.compare;
+    LongPrototype.negate = function negate() {
+      if (!this.unsigned && this.eq(MIN_VALUE))
+        return MIN_VALUE;
+      return this.not().add(ONE);
+    };
+    LongPrototype.neg = LongPrototype.negate;
+    LongPrototype.add = function add5(addend) {
+      if (!isLong(addend))
+        addend = fromValue(addend);
+      var a48 = this.high >>> 16;
+      var a32 = this.high & 65535;
+      var a16 = this.low >>> 16;
+      var a00 = this.low & 65535;
+      var b48 = addend.high >>> 16;
+      var b32 = addend.high & 65535;
+      var b16 = addend.low >>> 16;
+      var b00 = addend.low & 65535;
+      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;
+      c00 += a00 + b00;
+      c16 += c00 >>> 16;
+      c00 &= 65535;
+      c16 += a16 + b16;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c32 += a32 + b32;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c48 += a48 + b48;
+      c48 &= 65535;
+      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);
+    };
+    LongPrototype.subtract = function subtract(subtrahend) {
+      if (!isLong(subtrahend))
+        subtrahend = fromValue(subtrahend);
+      return this.add(subtrahend.neg());
+    };
+    LongPrototype.sub = LongPrototype.subtract;
+    LongPrototype.multiply = function multiply4(multiplier) {
+      if (this.isZero())
+        return ZERO;
+      if (!isLong(multiplier))
+        multiplier = fromValue(multiplier);
+      if (wasm) {
+        var low = wasm.mul(
+          this.low,
+          this.high,
+          multiplier.low,
+          multiplier.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      if (multiplier.isZero())
+        return ZERO;
+      if (this.eq(MIN_VALUE))
+        return multiplier.isOdd() ? MIN_VALUE : ZERO;
+      if (multiplier.eq(MIN_VALUE))
+        return this.isOdd() ? MIN_VALUE : ZERO;
+      if (this.isNegative()) {
+        if (multiplier.isNegative())
+          return this.neg().mul(multiplier.neg());
+        else
+          return this.neg().mul(multiplier).neg();
+      } else if (multiplier.isNegative())
+        return this.mul(multiplier.neg()).neg();
+      if (this.lt(TWO_PWR_24) && multiplier.lt(TWO_PWR_24))
+        return fromNumber(this.toNumber() * multiplier.toNumber(), this.unsigned);
+      var a48 = this.high >>> 16;
+      var a32 = this.high & 65535;
+      var a16 = this.low >>> 16;
+      var a00 = this.low & 65535;
+      var b48 = multiplier.high >>> 16;
+      var b32 = multiplier.high & 65535;
+      var b16 = multiplier.low >>> 16;
+      var b00 = multiplier.low & 65535;
+      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;
+      c00 += a00 * b00;
+      c16 += c00 >>> 16;
+      c00 &= 65535;
+      c16 += a16 * b00;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c16 += a00 * b16;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c32 += a32 * b00;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c32 += a16 * b16;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c32 += a00 * b32;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c48 += a48 * b00 + a32 * b16 + a16 * b32 + a00 * b48;
+      c48 &= 65535;
+      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);
+    };
+    LongPrototype.mul = LongPrototype.multiply;
+    LongPrototype.divide = function divide(divisor) {
+      if (!isLong(divisor))
+        divisor = fromValue(divisor);
+      if (divisor.isZero())
+        throw Error("division by zero");
+      if (wasm) {
+        if (!this.unsigned && this.high === -2147483648 && divisor.low === -1 && divisor.high === -1) {
+          return this;
+        }
+        var low = (this.unsigned ? wasm.div_u : wasm.div_s)(
+          this.low,
+          this.high,
+          divisor.low,
+          divisor.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      if (this.isZero())
+        return this.unsigned ? UZERO : ZERO;
+      var approx, rem, res;
+      if (!this.unsigned) {
+        if (this.eq(MIN_VALUE)) {
+          if (divisor.eq(ONE) || divisor.eq(NEG_ONE))
+            return MIN_VALUE;
+          else if (divisor.eq(MIN_VALUE))
+            return ONE;
+          else {
+            var halfThis = this.shr(1);
+            approx = halfThis.div(divisor).shl(1);
+            if (approx.eq(ZERO)) {
+              return divisor.isNegative() ? ONE : NEG_ONE;
+            } else {
+              rem = this.sub(divisor.mul(approx));
+              res = approx.add(rem.div(divisor));
+              return res;
+            }
+          }
+        } else if (divisor.eq(MIN_VALUE))
+          return this.unsigned ? UZERO : ZERO;
+        if (this.isNegative()) {
+          if (divisor.isNegative())
+            return this.neg().div(divisor.neg());
+          return this.neg().div(divisor).neg();
+        } else if (divisor.isNegative())
+          return this.div(divisor.neg()).neg();
+        res = ZERO;
+      } else {
+        if (!divisor.unsigned)
+          divisor = divisor.toUnsigned();
+        if (divisor.gt(this))
+          return UZERO;
+        if (divisor.gt(this.shru(1)))
+          return UONE;
+        res = UZERO;
+      }
+      rem = this;
+      while (rem.gte(divisor)) {
+        approx = Math.max(1, Math.floor(rem.toNumber() / divisor.toNumber()));
+        var log22 = Math.ceil(Math.log(approx) / Math.LN2), delta = log22 <= 48 ? 1 : pow_dbl(2, log22 - 48), approxRes = fromNumber(approx), approxRem = approxRes.mul(divisor);
+        while (approxRem.isNegative() || approxRem.gt(rem)) {
+          approx -= delta;
+          approxRes = fromNumber(approx, this.unsigned);
+          approxRem = approxRes.mul(divisor);
+        }
+        if (approxRes.isZero())
+          approxRes = ONE;
+        res = res.add(approxRes);
+        rem = rem.sub(approxRem);
+      }
+      return res;
+    };
+    LongPrototype.div = LongPrototype.divide;
+    LongPrototype.modulo = function modulo(divisor) {
+      if (!isLong(divisor))
+        divisor = fromValue(divisor);
+      if (wasm) {
+        var low = (this.unsigned ? wasm.rem_u : wasm.rem_s)(
+          this.low,
+          this.high,
+          divisor.low,
+          divisor.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      return this.sub(this.div(divisor).mul(divisor));
+    };
+    LongPrototype.mod = LongPrototype.modulo;
+    LongPrototype.rem = LongPrototype.modulo;
+    LongPrototype.not = function not() {
+      return fromBits(~this.low, ~this.high, this.unsigned);
+    };
+    LongPrototype.and = function and(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low & other.low, this.high & other.high, this.unsigned);
+    };
+    LongPrototype.or = function or(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low | other.low, this.high | other.high, this.unsigned);
+    };
+    LongPrototype.xor = function xor(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low ^ other.low, this.high ^ other.high, this.unsigned);
+    };
+    LongPrototype.shiftLeft = function shiftLeft(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      if ((numBits &= 63) === 0)
+        return this;
+      else if (numBits < 32)
+        return fromBits(this.low << numBits, this.high << numBits | this.low >>> 32 - numBits, this.unsigned);
+      else
+        return fromBits(0, this.low << numBits - 32, this.unsigned);
+    };
+    LongPrototype.shl = LongPrototype.shiftLeft;
+    LongPrototype.shiftRight = function shiftRight(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      if ((numBits &= 63) === 0)
+        return this;
+      else if (numBits < 32)
+        return fromBits(this.low >>> numBits | this.high << 32 - numBits, this.high >> numBits, this.unsigned);
+      else
+        return fromBits(this.high >> numBits - 32, this.high >= 0 ? 0 : -1, this.unsigned);
+    };
+    LongPrototype.shr = LongPrototype.shiftRight;
+    LongPrototype.shiftRightUnsigned = function shiftRightUnsigned(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      numBits &= 63;
+      if (numBits === 0)
+        return this;
+      else {
+        var high = this.high;
+        if (numBits < 32) {
+          var low = this.low;
+          return fromBits(low >>> numBits | high << 32 - numBits, high >>> numBits, this.unsigned);
+        } else if (numBits === 32)
+          return fromBits(high, 0, this.unsigned);
+        else
+          return fromBits(high >>> numBits - 32, 0, this.unsigned);
+      }
+    };
+    LongPrototype.shru = LongPrototype.shiftRightUnsigned;
+    LongPrototype.shr_u = LongPrototype.shiftRightUnsigned;
+    LongPrototype.toSigned = function toSigned() {
+      if (!this.unsigned)
+        return this;
+      return fromBits(this.low, this.high, false);
+    };
+    LongPrototype.toUnsigned = function toUnsigned() {
+      if (this.unsigned)
+        return this;
+      return fromBits(this.low, this.high, true);
+    };
+    LongPrototype.toBytes = function toBytes(le) {
+      return le ? this.toBytesLE() : this.toBytesBE();
+    };
+    LongPrototype.toBytesLE = function toBytesLE() {
+      var hi = this.high, lo = this.low;
+      return [
+        lo & 255,
+        lo >>> 8 & 255,
+        lo >>> 16 & 255,
+        lo >>> 24,
+        hi & 255,
+        hi >>> 8 & 255,
+        hi >>> 16 & 255,
+        hi >>> 24
+      ];
+    };
+    LongPrototype.toBytesBE = function toBytesBE() {
+      var hi = this.high, lo = this.low;
+      return [
+        hi >>> 24,
+        hi >>> 16 & 255,
+        hi >>> 8 & 255,
+        hi & 255,
+        lo >>> 24,
+        lo >>> 16 & 255,
+        lo >>> 8 & 255,
+        lo & 255
+      ];
+    };
+    Long2.fromBytes = function fromBytes(bytes, unsigned, le) {
+      return le ? Long2.fromBytesLE(bytes, unsigned) : Long2.fromBytesBE(bytes, unsigned);
+    };
+    Long2.fromBytesLE = function fromBytesLE(bytes, unsigned) {
+      return new Long2(
+        bytes[0] | bytes[1] << 8 | bytes[2] << 16 | bytes[3] << 24,
+        bytes[4] | bytes[5] << 8 | bytes[6] << 16 | bytes[7] << 24,
+        unsigned
+      );
+    };
+    Long2.fromBytesBE = function fromBytesBE(bytes, unsigned) {
+      return new Long2(
+        bytes[4] << 24 | bytes[5] << 16 | bytes[6] << 8 | bytes[7],
+        bytes[0] << 24 | bytes[1] << 16 | bytes[2] << 8 | bytes[3],
+        unsigned
+      );
+    };
+  }
+});
+var require_browser = __commonJS({
+  "(disabled):node_modules/node-fetch/browser.js"() {
+  }
+});
+var require_util = __commonJS({
+  "(disabled):util"() {
+  }
+});
+var require_alea = __commonJS({
+  "node_modules/seedrandom/lib/alea.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function Alea(seed) {
+        var me = this, mash = Mash();
+        me.next = function() {
+          var t = 2091639 * me.s0 + me.c * 23283064365386963e-26;
+          me.s0 = me.s1;
+          me.s1 = me.s2;
+          return me.s2 = t - (me.c = t | 0);
+        };
+        me.c = 1;
+        me.s0 = mash(" ");
+        me.s1 = mash(" ");
+        me.s2 = mash(" ");
+        me.s0 -= mash(seed);
+        if (me.s0 < 0) {
+          me.s0 += 1;
+        }
+        me.s1 -= mash(seed);
+        if (me.s1 < 0) {
+          me.s1 += 1;
+        }
+        me.s2 -= mash(seed);
+        if (me.s2 < 0) {
+          me.s2 += 1;
+        }
+        mash = null;
+      }
+      function copy(f, t) {
+        t.c = f.c;
+        t.s0 = f.s0;
+        t.s1 = f.s1;
+        t.s2 = f.s2;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new Alea(seed), state = opts && opts.state, prng = xg.next;
+        prng.int32 = function() {
+          return xg.next() * 4294967296 | 0;
+        };
+        prng.double = function() {
+          return prng() + (prng() * 2097152 | 0) * 11102230246251565e-32;
+        };
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      function Mash() {
+        var n = 4022871197;
+        var mash = function(data) {
+          data = String(data);
+          for (var i = 0; i < data.length; i++) {
+            n += data.charCodeAt(i);
+            var h = 0.02519603282416938 * n;
+            n = h >>> 0;
+            h -= n;
+            h *= n;
+            n = h >>> 0;
+            h -= n;
+            n += h * 4294967296;
+          }
+          return (n >>> 0) * 23283064365386963e-26;
+        };
+        return mash;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.alea = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_xor128 = __commonJS({
+  "node_modules/seedrandom/lib/xor128.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.x = 0;
+        me.y = 0;
+        me.z = 0;
+        me.w = 0;
+        me.next = function() {
+          var t = me.x ^ me.x << 11;
+          me.x = me.y;
+          me.y = me.z;
+          me.z = me.w;
+          return me.w ^= me.w >>> 19 ^ t ^ t >>> 8;
+        };
+        if (seed === (seed | 0)) {
+          me.x = seed;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 64; k++) {
+          me.x ^= strseed.charCodeAt(k) | 0;
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.x = f.x;
+        t.y = f.y;
+        t.z = f.z;
+        t.w = f.w;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xor128 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_xorwow = __commonJS({
+  "node_modules/seedrandom/lib/xorwow.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.next = function() {
+          var t = me.x ^ me.x >>> 2;
+          me.x = me.y;
+          me.y = me.z;
+          me.z = me.w;
+          me.w = me.v;
+          return (me.d = me.d + 362437 | 0) + (me.v = me.v ^ me.v << 4 ^ (t ^ t << 1)) | 0;
+        };
+        me.x = 0;
+        me.y = 0;
+        me.z = 0;
+        me.w = 0;
+        me.v = 0;
+        if (seed === (seed | 0)) {
+          me.x = seed;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 64; k++) {
+          me.x ^= strseed.charCodeAt(k) | 0;
+          if (k == strseed.length) {
+            me.d = me.x << 10 ^ me.x >>> 4;
+          }
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.x = f.x;
+        t.y = f.y;
+        t.z = f.z;
+        t.w = f.w;
+        t.v = f.v;
+        t.d = f.d;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xorwow = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_xorshift7 = __commonJS({
+  "node_modules/seedrandom/lib/xorshift7.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this;
+        me.next = function() {
+          var X = me.x, i = me.i, t, v, w;
+          t = X[i];
+          t ^= t >>> 7;
+          v = t ^ t << 24;
+          t = X[i + 1 & 7];
+          v ^= t ^ t >>> 10;
+          t = X[i + 3 & 7];
+          v ^= t ^ t >>> 3;
+          t = X[i + 4 & 7];
+          v ^= t ^ t << 7;
+          t = X[i + 7 & 7];
+          t = t ^ t << 13;
+          v ^= t ^ t << 9;
+          X[i] = v;
+          me.i = i + 1 & 7;
+          return v;
+        };
+        function init2(me2, seed2) {
+          var j, w, X = [];
+          if (seed2 === (seed2 | 0)) {
+            w = X[0] = seed2;
+          } else {
+            seed2 = "" + seed2;
+            for (j = 0; j < seed2.length; ++j) {
+              X[j & 7] = X[j & 7] << 15 ^ seed2.charCodeAt(j) + X[j + 1 & 7] << 13;
+            }
+          }
+          while (X.length < 8)
+            X.push(0);
+          for (j = 0; j < 8 && X[j] === 0; ++j)
+            ;
+          if (j == 8)
+            w = X[7] = -1;
+          else
+            w = X[j];
+          me2.x = X;
+          me2.i = 0;
+          for (j = 256; j > 0; --j) {
+            me2.next();
+          }
+        }
+        init2(me, seed);
+      }
+      function copy(f, t) {
+        t.x = f.x.slice();
+        t.i = f.i;
+        return t;
+      }
+      function impl(seed, opts) {
+        if (seed == null)
+          seed = +new Date();
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (state.x)
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xorshift7 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_xor4096 = __commonJS({
+  "node_modules/seedrandom/lib/xor4096.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this;
+        me.next = function() {
+          var w = me.w, X = me.X, i = me.i, t, v;
+          me.w = w = w + 1640531527 | 0;
+          v = X[i + 34 & 127];
+          t = X[i = i + 1 & 127];
+          v ^= v << 13;
+          t ^= t << 17;
+          v ^= v >>> 15;
+          t ^= t >>> 12;
+          v = X[i] = v ^ t;
+          me.i = i;
+          return v + (w ^ w >>> 16) | 0;
+        };
+        function init2(me2, seed2) {
+          var t, v, i, j, w, X = [], limit = 128;
+          if (seed2 === (seed2 | 0)) {
+            v = seed2;
+            seed2 = null;
+          } else {
+            seed2 = seed2 + "\0";
+            v = 0;
+            limit = Math.max(limit, seed2.length);
+          }
+          for (i = 0, j = -32; j < limit; ++j) {
+            if (seed2)
+              v ^= seed2.charCodeAt((j + 32) % seed2.length);
+            if (j === 0)
+              w = v;
+            v ^= v << 10;
+            v ^= v >>> 15;
+            v ^= v << 4;
+            v ^= v >>> 13;
+            if (j >= 0) {
+              w = w + 1640531527 | 0;
+              t = X[j & 127] ^= v + w;
+              i = 0 == t ? i + 1 : 0;
+            }
+          }
+          if (i >= 128) {
+            X[(seed2 && seed2.length || 0) & 127] = -1;
+          }
+          i = 127;
+          for (j = 4 * 128; j > 0; --j) {
+            v = X[i + 34 & 127];
+            t = X[i = i + 1 & 127];
+            v ^= v << 13;
+            t ^= t << 17;
+            v ^= v >>> 15;
+            t ^= t >>> 12;
+            X[i] = v ^ t;
+          }
+          me2.w = w;
+          me2.X = X;
+          me2.i = i;
+        }
+        init2(me, seed);
+      }
+      function copy(f, t) {
+        t.i = f.i;
+        t.w = f.w;
+        t.X = f.X.slice();
+        return t;
+      }
+      ;
+      function impl(seed, opts) {
+        if (seed == null)
+          seed = +new Date();
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (state.X)
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xor4096 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_tychei = __commonJS({
+  "node_modules/seedrandom/lib/tychei.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.next = function() {
+          var b = me.b, c = me.c, d = me.d, a = me.a;
+          b = b << 25 ^ b >>> 7 ^ c;
+          c = c - d | 0;
+          d = d << 24 ^ d >>> 8 ^ a;
+          a = a - b | 0;
+          me.b = b = b << 20 ^ b >>> 12 ^ c;
+          me.c = c = c - d | 0;
+          me.d = d << 16 ^ c >>> 16 ^ a;
+          return me.a = a - b | 0;
+        };
+        me.a = 0;
+        me.b = 0;
+        me.c = 2654435769 | 0;
+        me.d = 1367130551;
+        if (seed === Math.floor(seed)) {
+          me.a = seed / 4294967296 | 0;
+          me.b = seed | 0;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 20; k++) {
+          me.b ^= strseed.charCodeAt(k) | 0;
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.a = f.a;
+        t.b = f.b;
+        t.c = f.c;
+        t.d = f.d;
+        return t;
+      }
+      ;
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.tychei = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+var require_crypto = __commonJS({
+  "(disabled):crypto"() {
+  }
+});
+var require_seedrandom = __commonJS({
+  "node_modules/seedrandom/seedrandom.js"(exports, module) {
+    (function(global2, pool3, math) {
+      var width = 256, chunks = 6, digits = 52, rngname = "random", startdenom = math.pow(width, chunks), significance = math.pow(2, digits), overflow = significance * 2, mask = width - 1, nodecrypto;
+      function seedrandom5(seed, options, callback) {
+        var key = [];
+        options = options == true ? { entropy: true } : options || {};
+        var shortseed = mixkey(flatten4(
+          options.entropy ? [seed, tostring(pool3)] : seed == null ? autoseed() : seed,
+          3
+        ), key);
+        var arc4 = new ARC4(key);
+        var prng = function() {
+          var n = arc4.g(chunks), d = startdenom, x = 0;
+          while (n < significance) {
+            n = (n + x) * width;
+            d *= width;
+            x = arc4.g(1);
+          }
+          while (n >= overflow) {
+            n /= 2;
+            d /= 2;
+            x >>>= 1;
+          }
+          return (n + x) / d;
+        };
+        prng.int32 = function() {
+          return arc4.g(4) | 0;
+        };
+        prng.quick = function() {
+          return arc4.g(4) / 4294967296;
+        };
+        prng.double = prng;
+        mixkey(tostring(arc4.S), pool3);
+        return (options.pass || callback || function(prng2, seed2, is_math_call, state) {
+          if (state) {
+            if (state.S) {
+              copy(state, arc4);
+            }
+            prng2.state = function() {
+              return copy(arc4, {});
+            };
+          }
+          if (is_math_call) {
+            math[rngname] = prng2;
+            return seed2;
+          } else
+            return prng2;
+        })(
+          prng,
+          shortseed,
+          "global" in options ? options.global : this == math,
+          options.state
+        );
+      }
+      function ARC4(key) {
+        var t, keylen = key.length, me = this, i = 0, j = me.i = me.j = 0, s = me.S = [];
+        if (!keylen) {
+          key = [keylen++];
+        }
+        while (i < width) {
+          s[i] = i++;
+        }
+        for (i = 0; i < width; i++) {
+          s[i] = s[j = mask & j + key[i % keylen] + (t = s[i])];
+          s[j] = t;
+        }
+        (me.g = function(count2) {
+          var t2, r = 0, i2 = me.i, j2 = me.j, s2 = me.S;
+          while (count2--) {
+            t2 = s2[i2 = mask & i2 + 1];
+            r = r * width + s2[mask & (s2[i2] = s2[j2 = mask & j2 + t2]) + (s2[j2] = t2)];
+          }
+          me.i = i2;
+          me.j = j2;
+          return r;
+        })(width);
+      }
+      function copy(f, t) {
+        t.i = f.i;
+        t.j = f.j;
+        t.S = f.S.slice();
+        return t;
+      }
+      ;
+      function flatten4(obj, depth) {
+        var result = [], typ = typeof obj, prop;
+        if (depth && typ == "object") {
+          for (prop in obj) {
+            try {
+              result.push(flatten4(obj[prop], depth - 1));
+            } catch (e) {
+            }
+          }
+        }
+        return result.length ? result : typ == "string" ? obj : obj + "\0";
+      }
+      function mixkey(seed, key) {
+        var stringseed = seed + "", smear, j = 0;
+        while (j < stringseed.length) {
+          key[mask & j] = mask & (smear ^= key[mask & j] * 19) + stringseed.charCodeAt(j++);
+        }
+        return tostring(key);
+      }
+      function autoseed() {
+        try {
+          var out;
+          if (nodecrypto && (out = nodecrypto.randomBytes)) {
+            out = out(width);
+          } else {
+            out = new Uint8Array(width);
+            (global2.crypto || global2.msCrypto).getRandomValues(out);
+          }
+          return tostring(out);
+        } catch (e) {
+          var browser = global2.navigator, plugins = browser && browser.plugins;
+          return [+new Date(), global2, plugins, global2.screen, tostring(pool3)];
+        }
+      }
+      function tostring(a) {
+        return String.fromCharCode.apply(0, a);
+      }
+      mixkey(math.random(), pool3);
+      if (typeof module == "object" && module.exports) {
+        module.exports = seedrandom5;
+        try {
+          nodecrypto = require_crypto();
+        } catch (ex) {
+        }
+      } else if (typeof define == "function" && define.amd) {
+        define(function() {
+          return seedrandom5;
+        });
+      } else {
+        math["seed" + rngname] = seedrandom5;
+      }
+    })(
+      typeof self !== "undefined" ? self : exports,
+      [],
+      Math
+    );
+  }
+});
+var require_seedrandom2 = __commonJS({
+  "node_modules/seedrandom/index.js"(exports, module) {
+    var alea5 = require_alea();
+    var xor128 = require_xor128();
+    var xorwow = require_xorwow();
+    var xorshift7 = require_xorshift7();
+    var xor4096 = require_xor4096();
+    var tychei = require_tychei();
+    var sr = require_seedrandom();
+    sr.alea = alea5;
+    sr.xor128 = xor128;
+    sr.xorwow = xorwow;
+    sr.xorshift7 = xorshift7;
+    sr.xor4096 = xor4096;
+    sr.tychei = tychei;
+    module.exports = sr;
+  }
+});
+var require_string_decoder = __commonJS({
+  "(disabled):node_modules/string_decoder/lib/string_decoder.js"() {
+  }
+});
+var require_fs = __commonJS({
+  "(disabled):fs"() {
+  }
+});
+var require_path = __commonJS({
+  "(disabled):path"() {
+  }
+});
+var require_worker_threads = __commonJS({
+  "(disabled):worker_threads"() {
+  }
+});
+var require_perf_hooks = __commonJS({
+  "(disabled):perf_hooks"() {
+  }
+});
+var require_os = __commonJS({
+  "(disabled):os"() {
+  }
+});
+var require_tfjs_backend_wasm_threaded_simd = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(exports, module) {
+    var WasmBackendModuleThreadedSimd2 = (() => {
+      var _scriptDir = typeof document !== "undefined" && document.currentScript ? document.currentScript.src : void 0;
+      if (typeof __filename !== "undefined")
+        _scriptDir = _scriptDir || __filename;
+      return function(WasmBackendModuleThreadedSimd3) {
+        WasmBackendModuleThreadedSimd3 = WasmBackendModuleThreadedSimd3 || {};
+        function GROWABLE_HEAP_I8() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP8;
+        }
+        function GROWABLE_HEAP_U8() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPU8;
+        }
+        function GROWABLE_HEAP_I16() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP16;
+        }
+        function GROWABLE_HEAP_I32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP32;
+        }
+        function GROWABLE_HEAP_U32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPU32;
+        }
+        function GROWABLE_HEAP_F32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPF32;
+        }
+        function GROWABLE_HEAP_F64() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPF64;
+        }
+        var Module = typeof WasmBackendModuleThreadedSimd3 != "undefined" ? WasmBackendModuleThreadedSimd3 : {};
+        var readyPromiseResolve, readyPromiseReject;
+        Module["ready"] = new Promise(function(resolve, reject) {
+          readyPromiseResolve = resolve;
+          readyPromiseReject = reject;
+        });
+        var beforeListeners;
+        if (typeof process !== "undefined" && process.listeners) {
+          beforeListeners = { uncaughtException: process.listeners("uncaughtException"), unhandledRejection: process.listeners("unhandledRejection") };
+        }
+        var moduleOverrides = Object.assign({}, Module);
+        var arguments_ = [];
+        var thisProgram = "./this.program";
+        var quit_ = (status, toThrow) => {
+          throw toThrow;
+        };
+        var ENVIRONMENT_IS_WEB = typeof window == "object";
+        var ENVIRONMENT_IS_WORKER = typeof importScripts == "function";
+        var ENVIRONMENT_IS_NODE = typeof process == "object" && typeof process.versions == "object" && typeof process.versions.node == "string";
+        var ENVIRONMENT_IS_PTHREAD = Module["ENVIRONMENT_IS_PTHREAD"] || false;
+        var scriptDirectory = "";
+        function locateFile(path) {
+          if (Module["locateFile"]) {
+            return Module["locateFile"](path, scriptDirectory);
+          }
+          return scriptDirectory + path;
+        }
+        var read_, readAsync, readBinary, setWindowTitle;
+        function logExceptionOnExit(e) {
+          if (e instanceof ExitStatus)
+            return;
+          let toLog = e;
+          err("exiting due to exception: " + toLog);
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = require_path().dirname(scriptDirectory) + "/";
+          } else {
+            scriptDirectory = __dirname + "/";
+          }
+          var fs, nodePath;
+          if (typeof __require2 === "function") {
+            fs = require_fs();
+            nodePath = require_path();
+          }
+          read_ = (filename, binary) => {
+            filename = nodePath["normalize"](filename);
+            return fs.readFileSync(filename, binary ? void 0 : "utf8");
+          };
+          readBinary = (filename) => {
+            var ret = read_(filename, true);
+            if (!ret.buffer) {
+              ret = new Uint8Array(ret);
+            }
+            return ret;
+          };
+          readAsync = (filename, onload, onerror) => {
+            filename = nodePath["normalize"](filename);
+            fs.readFile(filename, function(err2, data) {
+              if (err2)
+                onerror(err2);
+              else
+                onload(data.buffer);
+            });
+          };
+          if (process["argv"].length > 1) {
+            thisProgram = process["argv"][1].replace(/\\/g, "/");
+          }
+          arguments_ = process["argv"].slice(2);
+          process["on"]("uncaughtException", function(ex) {
+            if (!(ex instanceof ExitStatus)) {
+              throw ex;
+            }
+          });
+          process["on"]("unhandledRejection", function(reason) {
+            throw reason;
+          });
+          quit_ = (status, toThrow) => {
+            if (keepRuntimeAlive()) {
+              process["exitCode"] = status;
+              throw toThrow;
+            }
+            logExceptionOnExit(toThrow);
+            process["exit"](status);
+          };
+          Module["inspect"] = function() {
+            return "[Emscripten Module object]";
+          };
+          let nodeWorkerThreads;
+          try {
+            nodeWorkerThreads = require_worker_threads();
+          } catch (e) {
+            console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?');
+            throw e;
+          }
+          global.Worker = nodeWorkerThreads.Worker;
+        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = self.location.href;
+          } else if (typeof document != "undefined" && document.currentScript) {
+            scriptDirectory = document.currentScript.src;
+          }
+          if (typeof _scriptDir !== "undefined" && _scriptDir) {
+            scriptDirectory = _scriptDir;
+          }
+          if (scriptDirectory.indexOf("blob:") !== 0) {
+            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, "").lastIndexOf("/") + 1);
+          } else {
+            scriptDirectory = "";
+          }
+          if (!ENVIRONMENT_IS_NODE) {
+            read_ = (url) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, false);
+              xhr.send(null);
+              return xhr.responseText;
+            };
+            if (ENVIRONMENT_IS_WORKER) {
+              readBinary = (url) => {
+                var xhr = new XMLHttpRequest();
+                xhr.open("GET", url, false);
+                xhr.responseType = "arraybuffer";
+                xhr.send(null);
+                return new Uint8Array(xhr.response);
+              };
+            }
+            readAsync = (url, onload, onerror) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, true);
+              xhr.responseType = "arraybuffer";
+              xhr.onload = () => {
+                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {
+                  onload(xhr.response);
+                  return;
+                }
+                onerror();
+              };
+              xhr.onerror = onerror;
+              xhr.send(null);
+            };
+          }
+          setWindowTitle = (title) => document.title = title;
+        } else {
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (typeof performance == "undefined") {
+            global.performance = require_perf_hooks().performance;
+          }
+        }
+        var defaultPrint = console.log.bind(console);
+        var defaultPrintErr = console.warn.bind(console);
+        if (ENVIRONMENT_IS_NODE) {
+          defaultPrint = (str) => fs.writeSync(1, str + "\n");
+          defaultPrintErr = (str) => fs.writeSync(2, str + "\n");
+        }
+        var out = Module["print"] || defaultPrint;
+        var err = Module["printErr"] || defaultPrintErr;
+        Object.assign(Module, moduleOverrides);
+        moduleOverrides = null;
+        if (Module["arguments"])
+          arguments_ = Module["arguments"];
+        if (Module["thisProgram"])
+          thisProgram = Module["thisProgram"];
+        if (Module["quit"])
+          quit_ = Module["quit"];
+        var POINTER_SIZE = 4;
+        var Atomics_load = Atomics.load;
+        var Atomics_store = Atomics.store;
+        var Atomics_compareExchange = Atomics.compareExchange;
+        var wasmBinary;
+        if (Module["wasmBinary"])
+          wasmBinary = Module["wasmBinary"];
+        var noExitRuntime = Module["noExitRuntime"] || true;
+        if (typeof WebAssembly != "object") {
+          abort("no native wasm support detected");
+        }
+        var wasmMemory;
+        var wasmModule;
+        var ABORT = false;
+        var EXITSTATUS;
+        function assert3(condition, text) {
+          if (!condition) {
+            abort(text);
+          }
+        }
+        var UTF8Decoder = typeof TextDecoder != "undefined" ? new TextDecoder("utf8") : void 0;
+        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {
+          var endIdx = idx + maxBytesToRead;
+          var endPtr = idx;
+          while (heapOrArray[endPtr] && !(endPtr >= endIdx))
+            ++endPtr;
+          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {
+            return UTF8Decoder.decode(heapOrArray.buffer instanceof SharedArrayBuffer ? heapOrArray.slice(idx, endPtr) : heapOrArray.subarray(idx, endPtr));
+          }
+          var str = "";
+          while (idx < endPtr) {
+            var u0 = heapOrArray[idx++];
+            if (!(u0 & 128)) {
+              str += String.fromCharCode(u0);
+              continue;
+            }
+            var u1 = heapOrArray[idx++] & 63;
+            if ((u0 & 224) == 192) {
+              str += String.fromCharCode((u0 & 31) << 6 | u1);
+              continue;
+            }
+            var u2 = heapOrArray[idx++] & 63;
+            if ((u0 & 240) == 224) {
+              u0 = (u0 & 15) << 12 | u1 << 6 | u2;
+            } else {
+              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;
+            }
+            if (u0 < 65536) {
+              str += String.fromCharCode(u0);
+            } else {
+              var ch = u0 - 65536;
+              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);
+            }
+          }
+          return str;
+        }
+        function UTF8ToString(ptr, maxBytesToRead) {
+          return ptr ? UTF8ArrayToString(GROWABLE_HEAP_U8(), ptr, maxBytesToRead) : "";
+        }
+        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {
+          if (!(maxBytesToWrite > 0))
+            return 0;
+          var startIdx = outIdx;
+          var endIdx = outIdx + maxBytesToWrite - 1;
+          for (var i = 0; i < str.length; ++i) {
+            var u = str.charCodeAt(i);
+            if (u >= 55296 && u <= 57343) {
+              var u1 = str.charCodeAt(++i);
+              u = 65536 + ((u & 1023) << 10) | u1 & 1023;
+            }
+            if (u <= 127) {
+              if (outIdx >= endIdx)
+                break;
+              heap[outIdx++] = u;
+            } else if (u <= 2047) {
+              if (outIdx + 1 >= endIdx)
+                break;
+              heap[outIdx++] = 192 | u >> 6;
+              heap[outIdx++] = 128 | u & 63;
+            } else if (u <= 65535) {
+              if (outIdx + 2 >= endIdx)
+                break;
+              heap[outIdx++] = 224 | u >> 12;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            } else {
+              if (outIdx + 3 >= endIdx)
+                break;
+              heap[outIdx++] = 240 | u >> 18;
+              heap[outIdx++] = 128 | u >> 12 & 63;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            }
+          }
+          heap[outIdx] = 0;
+          return outIdx - startIdx;
+        }
+        function stringToUTF8(str, outPtr, maxBytesToWrite) {
+          return stringToUTF8Array(str, GROWABLE_HEAP_U8(), outPtr, maxBytesToWrite);
+        }
+        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;
+        if (ENVIRONMENT_IS_PTHREAD) {
+          buffer2 = Module["buffer"];
+        }
+        function updateGlobalBufferAndViews(buf) {
+          buffer2 = buf;
+          Module["HEAP8"] = HEAP8 = new Int8Array(buf);
+          Module["HEAP16"] = HEAP16 = new Int16Array(buf);
+          Module["HEAP32"] = HEAP32 = new Int32Array(buf);
+          Module["HEAPU8"] = HEAPU8 = new Uint8Array(buf);
+          Module["HEAPU16"] = HEAPU16 = new Uint16Array(buf);
+          Module["HEAPU32"] = HEAPU32 = new Uint32Array(buf);
+          Module["HEAPF32"] = HEAPF32 = new Float32Array(buf);
+          Module["HEAPF64"] = HEAPF64 = new Float64Array(buf);
+        }
+        var INITIAL_MEMORY = Module["INITIAL_MEMORY"] || 16777216;
+        if (ENVIRONMENT_IS_PTHREAD) {
+          wasmMemory = Module["wasmMemory"];
+          buffer2 = Module["buffer"];
+        } else {
+          if (Module["wasmMemory"]) {
+            wasmMemory = Module["wasmMemory"];
+          } else {
+            wasmMemory = new WebAssembly.Memory({ "initial": INITIAL_MEMORY / 65536, "maximum": 2147483648 / 65536, "shared": true });
+            if (!(wasmMemory.buffer instanceof SharedArrayBuffer)) {
+              err("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag");
+              if (ENVIRONMENT_IS_NODE) {
+                console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)");
+              }
+              throw Error("bad memory");
+            }
+          }
+        }
+        if (wasmMemory) {
+          buffer2 = wasmMemory.buffer;
+        }
+        INITIAL_MEMORY = buffer2.byteLength;
+        updateGlobalBufferAndViews(buffer2);
+        var wasmTable;
+        var __ATPRERUN__ = [];
+        var __ATINIT__ = [];
+        var __ATPOSTRUN__ = [];
+        var runtimeInitialized = false;
+        function keepRuntimeAlive() {
+          return noExitRuntime;
+        }
+        function preRun() {
+          if (Module["preRun"]) {
+            if (typeof Module["preRun"] == "function")
+              Module["preRun"] = [Module["preRun"]];
+            while (Module["preRun"].length) {
+              addOnPreRun(Module["preRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPRERUN__);
+        }
+        function initRuntime() {
+          runtimeInitialized = true;
+          if (ENVIRONMENT_IS_PTHREAD)
+            return;
+          callRuntimeCallbacks(__ATINIT__);
+        }
+        function postRun() {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return;
+          if (Module["postRun"]) {
+            if (typeof Module["postRun"] == "function")
+              Module["postRun"] = [Module["postRun"]];
+            while (Module["postRun"].length) {
+              addOnPostRun(Module["postRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPOSTRUN__);
+        }
+        function addOnPreRun(cb) {
+          __ATPRERUN__.unshift(cb);
+        }
+        function addOnInit(cb) {
+          __ATINIT__.unshift(cb);
+        }
+        function addOnPostRun(cb) {
+          __ATPOSTRUN__.unshift(cb);
+        }
+        var runDependencies = 0;
+        var runDependencyWatcher = null;
+        var dependenciesFulfilled = null;
+        function addRunDependency(id) {
+          runDependencies++;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+        }
+        function removeRunDependency(id) {
+          runDependencies--;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+          if (runDependencies == 0) {
+            if (runDependencyWatcher !== null) {
+              clearInterval(runDependencyWatcher);
+              runDependencyWatcher = null;
+            }
+            if (dependenciesFulfilled) {
+              var callback = dependenciesFulfilled;
+              dependenciesFulfilled = null;
+              callback();
+            }
+          }
+        }
+        function abort(what) {
+          if (ENVIRONMENT_IS_PTHREAD) {
+            postMessage({ "cmd": "onAbort", "arg": what });
+          } else {
+            if (Module["onAbort"]) {
+              Module["onAbort"](what);
+            }
+          }
+          what = "Aborted(" + what + ")";
+          err(what);
+          ABORT = true;
+          EXITSTATUS = 1;
+          what += ". Build with -sASSERTIONS for more info.";
+          var e = new WebAssembly.RuntimeError(what);
+          readyPromiseReject(e);
+          throw e;
+        }
+        var dataURIPrefix = "data:application/octet-stream;base64,";
+        function isDataURI(filename) {
+          return filename.startsWith(dataURIPrefix);
+        }
+        function isFileURI(filename) {
+          return filename.startsWith("file://");
+        }
+        var wasmBinaryFile;
+        wasmBinaryFile = "tfjs-backend-wasm-threaded-simd.wasm";
+        if (!isDataURI(wasmBinaryFile)) {
+          wasmBinaryFile = locateFile(wasmBinaryFile);
+        }
+        function getBinary(file) {
+          try {
+            if (file == wasmBinaryFile && wasmBinary) {
+              return new Uint8Array(wasmBinary);
+            }
+            if (readBinary) {
+              return readBinary(file);
+            }
+            throw "both async and sync fetching of the wasm failed";
+          } catch (err2) {
+            abort(err2);
+          }
+        }
+        function getBinaryPromise() {
+          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {
+            if (typeof fetch == "function" && !isFileURI(wasmBinaryFile)) {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                if (!response["ok"]) {
+                  throw "failed to load wasm binary file at '" + wasmBinaryFile + "'";
+                }
+                return response["arrayBuffer"]();
+              }).catch(function() {
+                return getBinary(wasmBinaryFile);
+              });
+            } else {
+              if (readAsync) {
+                return new Promise(function(resolve, reject) {
+                  readAsync(wasmBinaryFile, function(response) {
+                    resolve(new Uint8Array(response));
+                  }, reject);
+                });
+              }
+            }
+          }
+          return Promise.resolve().then(function() {
+            return getBinary(wasmBinaryFile);
+          });
+        }
+        function createWasm() {
+          var info = { "env": asmLibraryArg, "wasi_snapshot_preview1": asmLibraryArg };
+          function receiveInstance(instance, module2) {
+            var exports3 = instance.exports;
+            Module["asm"] = exports3;
+            registerTLSInit(Module["asm"]["_emscripten_tls_init"]);
+            wasmTable = Module["asm"]["__indirect_function_table"];
+            addOnInit(Module["asm"]["__wasm_call_ctors"]);
+            wasmModule = module2;
+            if (!ENVIRONMENT_IS_PTHREAD) {
+              var numWorkersToLoad = PThread.unusedWorkers.length;
+              PThread.unusedWorkers.forEach(function(w) {
+                PThread.loadWasmModuleToWorker(w, function() {
+                  if (!--numWorkersToLoad)
+                    removeRunDependency("wasm-instantiate");
+                });
+              });
+            }
+          }
+          if (!ENVIRONMENT_IS_PTHREAD) {
+            addRunDependency("wasm-instantiate");
+          }
+          function receiveInstantiationResult(result) {
+            receiveInstance(result["instance"], result["module"]);
+          }
+          function instantiateArrayBuffer(receiver) {
+            return getBinaryPromise().then(function(binary) {
+              return WebAssembly.instantiate(binary, info);
+            }).then(function(instance) {
+              return instance;
+            }).then(receiver, function(reason) {
+              err("failed to asynchronously prepare wasm: " + reason);
+              abort(reason);
+            });
+          }
+          function instantiateAsync() {
+            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == "function" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == "function") {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                var result = WebAssembly.instantiateStreaming(response, info);
+                return result.then(receiveInstantiationResult, function(reason) {
+                  err("wasm streaming compile failed: " + reason);
+                  err("falling back to ArrayBuffer instantiation");
+                  return instantiateArrayBuffer(receiveInstantiationResult);
+                });
+              });
+            } else {
+              return instantiateArrayBuffer(receiveInstantiationResult);
+            }
+          }
+          if (Module["instantiateWasm"]) {
+            try {
+              var exports2 = Module["instantiateWasm"](info, receiveInstance);
+              return exports2;
+            } catch (e) {
+              err("Module.instantiateWasm callback failed with error: " + e);
+              readyPromiseReject(e);
+            }
+          }
+          instantiateAsync().catch(readyPromiseReject);
+          return {};
+        }
+        var tempDouble;
+        var tempI64;
+        var ASM_CONSTS = {};
+        function ExitStatus(status) {
+          this.name = "ExitStatus";
+          this.message = "Program terminated with exit(" + status + ")";
+          this.status = status;
+        }
+        function killThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          delete PThread.pthreads[pthread_ptr];
+          worker.terminate();
+          __emscripten_thread_free_data(pthread_ptr);
+          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);
+          worker.pthread_ptr = 0;
+        }
+        function cancelThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          worker.postMessage({ "cmd": "cancel" });
+        }
+        function cleanupThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          assert3(worker);
+          PThread.returnWorkerToPool(worker);
+        }
+        function spawnThread(threadParams) {
+          var worker = PThread.getNewWorker();
+          if (!worker) {
+            return 6;
+          }
+          PThread.runningWorkers.push(worker);
+          PThread.pthreads[threadParams.pthread_ptr] = worker;
+          worker.pthread_ptr = threadParams.pthread_ptr;
+          var msg = { "cmd": "run", "start_routine": threadParams.startRoutine, "arg": threadParams.arg, "pthread_ptr": threadParams.pthread_ptr };
+          worker.runPthread = () => {
+            msg.time = performance.now();
+            worker.postMessage(msg, threadParams.transferList);
+          };
+          if (worker.loaded) {
+            worker.runPthread();
+            delete worker.runPthread;
+          }
+          return 0;
+        }
+        var SYSCALLS = { varargs: void 0, get: function() {
+          SYSCALLS.varargs += 4;
+          var ret = GROWABLE_HEAP_I32()[SYSCALLS.varargs - 4 >> 2];
+          return ret;
+        }, getStr: function(ptr) {
+          var ret = UTF8ToString(ptr);
+          return ret;
+        } };
+        function _proc_exit(code) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(1, 1, code);
+          EXITSTATUS = code;
+          if (!keepRuntimeAlive()) {
+            PThread.terminateAllThreads();
+            if (Module["onExit"])
+              Module["onExit"](code);
+            ABORT = true;
+          }
+          quit_(code, new ExitStatus(code));
+        }
+        function exitJS(status, implicit) {
+          EXITSTATUS = status;
+          if (!implicit) {
+            if (ENVIRONMENT_IS_PTHREAD) {
+              exitOnMainThread(status);
+              throw "unwind";
+            } else {
+            }
+          }
+          _proc_exit(status);
+        }
+        var _exit = exitJS;
+        function handleException(e) {
+          if (e instanceof ExitStatus || e == "unwind") {
+            return EXITSTATUS;
+          }
+          quit_(1, e);
+        }
+        var PThread = { unusedWorkers: [], runningWorkers: [], tlsInitFunctions: [], pthreads: {}, init: function() {
+          if (ENVIRONMENT_IS_PTHREAD) {
+            PThread.initWorker();
+          } else {
+            PThread.initMainThread();
+          }
+        }, initMainThread: function() {
+          var pthreadPoolSize = 8;
+          while (pthreadPoolSize--) {
+            PThread.allocateUnusedWorker();
+          }
+        }, initWorker: function() {
+          noExitRuntime = false;
+        }, setExitStatus: function(status) {
+          EXITSTATUS = status;
+        }, terminateAllThreads: function() {
+          for (var worker of Object.values(PThread.pthreads)) {
+            PThread.returnWorkerToPool(worker);
+          }
+          for (var worker of PThread.unusedWorkers) {
+            worker.terminate();
+          }
+          PThread.unusedWorkers = [];
+        }, returnWorkerToPool: function(worker) {
+          var pthread_ptr = worker.pthread_ptr;
+          delete PThread.pthreads[pthread_ptr];
+          PThread.unusedWorkers.push(worker);
+          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);
+          worker.pthread_ptr = 0;
+          __emscripten_thread_free_data(pthread_ptr);
+        }, receiveObjectTransfer: function(data) {
+        }, threadInitTLS: function() {
+          PThread.tlsInitFunctions.forEach((f) => f());
+        }, loadWasmModuleToWorker: function(worker, onFinishedLoading) {
+          worker.onmessage = (e) => {
+            var d = e["data"];
+            var cmd = d["cmd"];
+            if (worker.pthread_ptr)
+              PThread.currentProxiedOperationCallerThread = worker.pthread_ptr;
+            if (d["targetThread"] && d["targetThread"] != _pthread_self()) {
+              var targetWorker = PThread.pthreads[d.targetThread];
+              if (targetWorker) {
+                targetWorker.postMessage(d, d["transferList"]);
+              } else {
+                err('Internal error! Worker sent a message "' + cmd + '" to target pthread ' + d["targetThread"] + ", but that thread no longer exists!");
+              }
+              PThread.currentProxiedOperationCallerThread = void 0;
+              return;
+            }
+            if (cmd === "processProxyingQueue") {
+              executeNotifiedProxyingQueue(d["queue"]);
+            } else if (cmd === "spawnThread") {
+              spawnThread(d);
+            } else if (cmd === "cleanupThread") {
+              cleanupThread(d["thread"]);
+            } else if (cmd === "killThread") {
+              killThread(d["thread"]);
+            } else if (cmd === "cancelThread") {
+              cancelThread(d["thread"]);
+            } else if (cmd === "loaded") {
+              worker.loaded = true;
+              if (onFinishedLoading)
+                onFinishedLoading(worker);
+              if (worker.runPthread) {
+                worker.runPthread();
+                delete worker.runPthread;
+              }
+            } else if (cmd === "print") {
+              out("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (cmd === "printErr") {
+              err("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (cmd === "alert") {
+              alert("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (d.target === "setimmediate") {
+              worker.postMessage(d);
+            } else if (cmd === "onAbort") {
+              if (Module["onAbort"]) {
+                Module["onAbort"](d["arg"]);
+              }
+            } else if (cmd) {
+              err("worker sent an unknown command " + cmd);
+            }
+            PThread.currentProxiedOperationCallerThread = void 0;
+          };
+          worker.onerror = (e) => {
+            var message = "worker sent an error!";
+            err(message + " " + e.filename + ":" + e.lineno + ": " + e.message);
+            throw e;
+          };
+          if (ENVIRONMENT_IS_NODE) {
+            worker.on("message", function(data) {
+              worker.onmessage({ data });
+            });
+            worker.on("error", function(e) {
+              worker.onerror(e);
+            });
+            worker.on("detachedExit", function() {
+            });
+          }
+          worker.postMessage({ "cmd": "load", "urlOrBlob": Module["mainScriptUrlOrBlob"] || _scriptDir, "wasmMemory": wasmMemory, "wasmModule": wasmModule });
+        }, allocateUnusedWorker: function() {
+          var pthreadMainJs = locateFile("tfjs-backend-wasm-threaded-simd.worker.js");
+          PThread.unusedWorkers.push(new Worker(pthreadMainJs));
+        }, getNewWorker: function() {
+          if (PThread.unusedWorkers.length == 0) {
+            PThread.allocateUnusedWorker();
+            PThread.loadWasmModuleToWorker(PThread.unusedWorkers[0]);
+          }
+          return PThread.unusedWorkers.pop();
+        } };
+        Module["PThread"] = PThread;
+        function callRuntimeCallbacks(callbacks2) {
+          while (callbacks2.length > 0) {
+            callbacks2.shift()(Module);
+          }
+        }
+        function withStackSave(f) {
+          var stack2 = stackSave();
+          var ret = f();
+          stackRestore(stack2);
+          return ret;
+        }
+        function demangle(func2) {
+          return func2;
+        }
+        function demangleAll(text) {
+          var regex = /\b_Z[\w\d_]+/g;
+          return text.replace(regex, function(x) {
+            var y = demangle(x);
+            return x === y ? x : y + " [" + x + "]";
+          });
+        }
+        function establishStackSpace() {
+          var pthread_ptr = _pthread_self();
+          var stackTop = GROWABLE_HEAP_I32()[pthread_ptr + 44 >> 2];
+          var stackSize = GROWABLE_HEAP_I32()[pthread_ptr + 48 >> 2];
+          var stackMax = stackTop - stackSize;
+          _emscripten_stack_set_limits(stackTop, stackMax);
+          stackRestore(stackTop);
+        }
+        Module["establishStackSpace"] = establishStackSpace;
+        function exitOnMainThread(returnCode) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(2, 0, returnCode);
+          try {
+            _exit(returnCode);
+          } catch (e) {
+            handleException(e);
+          }
+        }
+        var wasmTableMirror = [];
+        function getWasmTableEntry(funcPtr) {
+          var func2 = wasmTableMirror[funcPtr];
+          if (!func2) {
+            if (funcPtr >= wasmTableMirror.length)
+              wasmTableMirror.length = funcPtr + 1;
+            wasmTableMirror[funcPtr] = func2 = wasmTable.get(funcPtr);
+          }
+          return func2;
+        }
+        function invokeEntryPoint(ptr, arg) {
+          var result = getWasmTableEntry(ptr)(arg);
+          if (keepRuntimeAlive()) {
+            PThread.setExitStatus(result);
+          } else {
+            __emscripten_thread_exit(result);
+          }
+        }
+        Module["invokeEntryPoint"] = invokeEntryPoint;
+        function jsStackTrace() {
+          var error = new Error();
+          if (!error.stack) {
+            try {
+              throw new Error();
+            } catch (e) {
+              error = e;
+            }
+            if (!error.stack) {
+              return "(no stack trace available)";
+            }
+          }
+          return error.stack.toString();
+        }
+        function registerTLSInit(tlsInitFunc) {
+          PThread.tlsInitFunctions.push(tlsInitFunc);
+        }
+        function writeArrayToMemory(array2, buffer3) {
+          GROWABLE_HEAP_I8().set(array2, buffer3);
+        }
+        function ___emscripten_init_main_thread_js(tb) {
+          __emscripten_thread_init(tb, !ENVIRONMENT_IS_WORKER, 1, !ENVIRONMENT_IS_WEB);
+          PThread.threadInitTLS();
+        }
+        function ___emscripten_thread_cleanup(thread) {
+          if (!ENVIRONMENT_IS_PTHREAD)
+            cleanupThread(thread);
+          else
+            postMessage({ "cmd": "cleanupThread", "thread": thread });
+        }
+        function pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(3, 1, pthread_ptr, attr, startRoutine, arg);
+          return ___pthread_create_js(pthread_ptr, attr, startRoutine, arg);
+        }
+        function ___pthread_create_js(pthread_ptr, attr, startRoutine, arg) {
+          if (typeof SharedArrayBuffer == "undefined") {
+            err("Current environment does not support SharedArrayBuffer, pthreads are not available!");
+            return 6;
+          }
+          var transferList = [];
+          var error = 0;
+          if (ENVIRONMENT_IS_PTHREAD && (transferList.length === 0 || error)) {
+            return pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg);
+          }
+          if (error)
+            return error;
+          var threadParams = { startRoutine, pthread_ptr, arg, transferList };
+          if (ENVIRONMENT_IS_PTHREAD) {
+            threadParams.cmd = "spawnThread";
+            postMessage(threadParams, transferList);
+            return 0;
+          }
+          return spawnThread(threadParams);
+        }
+        function __emscripten_default_pthread_stack_size() {
+          return 2097152;
+        }
+        var nowIsMonotonic = true;
+        function __emscripten_get_now_is_monotonic() {
+          return nowIsMonotonic;
+        }
+        function executeNotifiedProxyingQueue(queue) {
+          Atomics.store(GROWABLE_HEAP_I32(), queue >> 2, 1);
+          if (_pthread_self()) {
+            __emscripten_proxy_execute_task_queue(queue);
+          }
+          Atomics.compareExchange(GROWABLE_HEAP_I32(), queue >> 2, 1, 0);
+        }
+        Module["executeNotifiedProxyingQueue"] = executeNotifiedProxyingQueue;
+        function __emscripten_notify_task_queue(targetThreadId, currThreadId, mainThreadId, queue) {
+          if (targetThreadId == currThreadId) {
+            setTimeout(() => executeNotifiedProxyingQueue(queue));
+          } else if (ENVIRONMENT_IS_PTHREAD) {
+            postMessage({ "targetThread": targetThreadId, "cmd": "processProxyingQueue", "queue": queue });
+          } else {
+            var worker = PThread.pthreads[targetThreadId];
+            if (!worker) {
+              return;
+            }
+            worker.postMessage({ "cmd": "processProxyingQueue", "queue": queue });
+          }
+          return 1;
+        }
+        function __emscripten_set_offscreencanvas_size(target, width, height) {
+          return -1;
+        }
+        function _abort() {
+          abort("");
+        }
+        function warnOnce(text) {
+          if (!warnOnce.shown)
+            warnOnce.shown = {};
+          if (!warnOnce.shown[text]) {
+            warnOnce.shown[text] = 1;
+            if (ENVIRONMENT_IS_NODE)
+              text = "warning: " + text;
+            err(text);
+          }
+        }
+        function _emscripten_check_blocking_allowed() {
+          if (ENVIRONMENT_IS_NODE)
+            return;
+          if (ENVIRONMENT_IS_WORKER)
+            return;
+          warnOnce("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread");
+        }
+        function _emscripten_date_now() {
+          return Date.now();
+        }
+        function getHeapMax() {
+          return 2147483648;
+        }
+        function _emscripten_get_heap_max() {
+          return getHeapMax();
+        }
+        var _emscripten_get_now;
+        if (ENVIRONMENT_IS_NODE) {
+          _emscripten_get_now = () => {
+            var t = process["hrtime"]();
+            return t[0] * 1e3 + t[1] / 1e6;
+          };
+        } else if (ENVIRONMENT_IS_PTHREAD) {
+          _emscripten_get_now = () => performance.now() - Module["__performance_now_clock_drift"];
+        } else
+          _emscripten_get_now = () => performance.now();
+        function _emscripten_memcpy_big(dest, src, num) {
+          GROWABLE_HEAP_U8().copyWithin(dest, src, src + num);
+        }
+        function _emscripten_num_logical_cores() {
+          if (ENVIRONMENT_IS_NODE)
+            return require_os().cpus().length;
+          return navigator["hardwareConcurrency"];
+        }
+        function _emscripten_proxy_to_main_thread_js(index, sync) {
+          var numCallArgs = arguments.length - 2;
+          var outerArgs = arguments;
+          return withStackSave(() => {
+            var serializedNumCallArgs = numCallArgs;
+            var args = stackAlloc(serializedNumCallArgs * 8);
+            var b = args >> 3;
+            for (var i = 0; i < numCallArgs; i++) {
+              var arg = outerArgs[2 + i];
+              GROWABLE_HEAP_F64()[b + i] = arg;
+            }
+            return _emscripten_run_in_main_runtime_thread_js(index, serializedNumCallArgs, args, sync);
+          });
+        }
+        var _emscripten_receive_on_main_thread_js_callArgs = [];
+        function _emscripten_receive_on_main_thread_js(index, numCallArgs, args) {
+          _emscripten_receive_on_main_thread_js_callArgs.length = numCallArgs;
+          var b = args >> 3;
+          for (var i = 0; i < numCallArgs; i++) {
+            _emscripten_receive_on_main_thread_js_callArgs[i] = GROWABLE_HEAP_F64()[b + i];
+          }
+          var isEmAsmConst = index < 0;
+          var func2 = !isEmAsmConst ? proxiedFunctionTable[index] : ASM_CONSTS[-index - 1];
+          return func2.apply(null, _emscripten_receive_on_main_thread_js_callArgs);
+        }
+        function emscripten_realloc_buffer(size) {
+          try {
+            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            return 1;
+          } catch (e) {
+          }
+        }
+        function _emscripten_resize_heap(requestedSize) {
+          var oldSize = GROWABLE_HEAP_U8().length;
+          requestedSize = requestedSize >>> 0;
+          if (requestedSize <= oldSize) {
+            return false;
+          }
+          var maxHeapSize = getHeapMax();
+          if (requestedSize > maxHeapSize) {
+            return false;
+          }
+          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;
+          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {
+            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);
+            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);
+            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));
+            var replacement = emscripten_realloc_buffer(newSize);
+            if (replacement) {
+              return true;
+            }
+          }
+          return false;
+        }
+        function _emscripten_unwind_to_js_event_loop() {
+          throw "unwind";
+        }
+        function _fd_close(fd) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(4, 1, fd);
+          return 52;
+        }
+        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(5, 1, fd, offset_low, offset_high, whence, newOffset);
+          return 70;
+        }
+        var printCharBuffers = [null, [], []];
+        function printChar(stream, curr) {
+          var buffer3 = printCharBuffers[stream];
+          if (curr === 0 || curr === 10) {
+            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));
+            buffer3.length = 0;
+          } else {
+            buffer3.push(curr);
+          }
+        }
+        function _fd_write(fd, iov, iovcnt, pnum) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(6, 1, fd, iov, iovcnt, pnum);
+          var num = 0;
+          for (var i = 0; i < iovcnt; i++) {
+            var ptr = GROWABLE_HEAP_U32()[iov >> 2];
+            var len = GROWABLE_HEAP_U32()[iov + 4 >> 2];
+            iov += 8;
+            for (var j = 0; j < len; j++) {
+              printChar(fd, GROWABLE_HEAP_U8()[ptr + j]);
+            }
+            num += len;
+          }
+          GROWABLE_HEAP_U32()[pnum >> 2] = num;
+          return 0;
+        }
+        function getCFunc(ident) {
+          var func2 = Module["_" + ident];
+          return func2;
+        }
+        function ccall(ident, returnType, argTypes, args, opts) {
+          var toC = { "string": (str) => {
+            var ret2 = 0;
+            if (str !== null && str !== void 0 && str !== 0) {
+              var len = (str.length << 2) + 1;
+              ret2 = stackAlloc(len);
+              stringToUTF8(str, ret2, len);
+            }
+            return ret2;
+          }, "array": (arr) => {
+            var ret2 = stackAlloc(arr.length);
+            writeArrayToMemory(arr, ret2);
+            return ret2;
+          } };
+          function convertReturnValue(ret2) {
+            if (returnType === "string") {
+              return UTF8ToString(ret2);
+            }
+            if (returnType === "boolean")
+              return Boolean(ret2);
+            return ret2;
+          }
+          var func2 = getCFunc(ident);
+          var cArgs = [];
+          var stack2 = 0;
+          if (args) {
+            for (var i = 0; i < args.length; i++) {
+              var converter = toC[argTypes[i]];
+              if (converter) {
+                if (stack2 === 0)
+                  stack2 = stackSave();
+                cArgs[i] = converter(args[i]);
+              } else {
+                cArgs[i] = args[i];
+              }
+            }
+          }
+          var ret = func2.apply(null, cArgs);
+          function onDone(ret2) {
+            if (stack2 !== 0)
+              stackRestore(stack2);
+            return convertReturnValue(ret2);
+          }
+          ret = onDone(ret);
+          return ret;
+        }
+        function cwrap(ident, returnType, argTypes, opts) {
+          argTypes = argTypes || [];
+          var numericArgs = argTypes.every((type) => type === "number" || type === "boolean");
+          var numericRet = returnType !== "string";
+          if (numericRet && numericArgs && !opts) {
+            return getCFunc(ident);
+          }
+          return function() {
+            return ccall(ident, returnType, argTypes, arguments, opts);
+          };
+        }
+        PThread.init();
+        var proxiedFunctionTable = [null, _proc_exit, exitOnMainThread, pthreadCreateProxied, _fd_close, _fd_seek, _fd_write];
+        var asmLibraryArg = { "__emscripten_init_main_thread_js": ___emscripten_init_main_thread_js, "__emscripten_thread_cleanup": ___emscripten_thread_cleanup, "__pthread_create_js": ___pthread_create_js, "_emscripten_default_pthread_stack_size": __emscripten_default_pthread_stack_size, "_emscripten_get_now_is_monotonic": __emscripten_get_now_is_monotonic, "_emscripten_notify_task_queue": __emscripten_notify_task_queue, "_emscripten_set_offscreencanvas_size": __emscripten_set_offscreencanvas_size, "abort": _abort, "emscripten_check_blocking_allowed": _emscripten_check_blocking_allowed, "emscripten_date_now": _emscripten_date_now, "emscripten_get_heap_max": _emscripten_get_heap_max, "emscripten_get_now": _emscripten_get_now, "emscripten_memcpy_big": _emscripten_memcpy_big, "emscripten_num_logical_cores": _emscripten_num_logical_cores, "emscripten_receive_on_main_thread_js": _emscripten_receive_on_main_thread_js, "emscripten_resize_heap": _emscripten_resize_heap, "emscripten_unwind_to_js_event_loop": _emscripten_unwind_to_js_event_loop, "exit": _exit, "fd_close": _fd_close, "fd_seek": _fd_seek, "fd_write": _fd_write, "memory": wasmMemory || Module["wasmMemory"] };
+        var asm = createWasm();
+        var ___wasm_call_ctors = Module["___wasm_call_ctors"] = function() {
+          return (___wasm_call_ctors = Module["___wasm_call_ctors"] = Module["asm"]["__wasm_call_ctors"]).apply(null, arguments);
+        };
+        var _init = Module["_init"] = function() {
+          return (_init = Module["_init"] = Module["asm"]["init"]).apply(null, arguments);
+        };
+        var _init_with_threads_count = Module["_init_with_threads_count"] = function() {
+          return (_init_with_threads_count = Module["_init_with_threads_count"] = Module["asm"]["init_with_threads_count"]).apply(null, arguments);
+        };
+        var _get_threads_count = Module["_get_threads_count"] = function() {
+          return (_get_threads_count = Module["_get_threads_count"] = Module["asm"]["get_threads_count"]).apply(null, arguments);
+        };
+        var _register_tensor = Module["_register_tensor"] = function() {
+          return (_register_tensor = Module["_register_tensor"] = Module["asm"]["register_tensor"]).apply(null, arguments);
+        };
+        var _dispose_data = Module["_dispose_data"] = function() {
+          return (_dispose_data = Module["_dispose_data"] = Module["asm"]["dispose_data"]).apply(null, arguments);
+        };
+        var _dispose = Module["_dispose"] = function() {
+          return (_dispose = Module["_dispose"] = Module["asm"]["dispose"]).apply(null, arguments);
+        };
+        var _Abs = Module["_Abs"] = function() {
+          return (_Abs = Module["_Abs"] = Module["asm"]["Abs"]).apply(null, arguments);
+        };
+        var _Add = Module["_Add"] = function() {
+          return (_Add = Module["_Add"] = Module["asm"]["Add"]).apply(null, arguments);
+        };
+        var _AddN = Module["_AddN"] = function() {
+          return (_AddN = Module["_AddN"] = Module["asm"]["AddN"]).apply(null, arguments);
+        };
+        var _All = Module["_All"] = function() {
+          return (_All = Module["_All"] = Module["asm"]["All"]).apply(null, arguments);
+        };
+        var _Any = Module["_Any"] = function() {
+          return (_Any = Module["_Any"] = Module["asm"]["Any"]).apply(null, arguments);
+        };
+        var _ArgMax = Module["_ArgMax"] = function() {
+          return (_ArgMax = Module["_ArgMax"] = Module["asm"]["ArgMax"]).apply(null, arguments);
+        };
+        var _AvgPool = Module["_AvgPool"] = function() {
+          return (_AvgPool = Module["_AvgPool"] = Module["asm"]["AvgPool"]).apply(null, arguments);
+        };
+        var _BatchMatMul = Module["_BatchMatMul"] = function() {
+          return (_BatchMatMul = Module["_BatchMatMul"] = Module["asm"]["BatchMatMul"]).apply(null, arguments);
+        };
+        var _Ceil = Module["_Ceil"] = function() {
+          return (_Ceil = Module["_Ceil"] = Module["asm"]["Ceil"]).apply(null, arguments);
+        };
+        var _ClipByValue = Module["_ClipByValue"] = function() {
+          return (_ClipByValue = Module["_ClipByValue"] = Module["asm"]["ClipByValue"]).apply(null, arguments);
+        };
+        var _Conv2D = Module["_Conv2D"] = function() {
+          return (_Conv2D = Module["_Conv2D"] = Module["asm"]["Conv2D"]).apply(null, arguments);
+        };
+        var _Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = function() {
+          return (_Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = Module["asm"]["Conv2DBackpropInput"]).apply(null, arguments);
+        };
+        var _Cos = Module["_Cos"] = function() {
+          return (_Cos = Module["_Cos"] = Module["asm"]["Cos"]).apply(null, arguments);
+        };
+        var _Cosh = Module["_Cosh"] = function() {
+          return (_Cosh = Module["_Cosh"] = Module["asm"]["Cosh"]).apply(null, arguments);
+        };
+        var _CropAndResize = Module["_CropAndResize"] = function() {
+          return (_CropAndResize = Module["_CropAndResize"] = Module["asm"]["CropAndResize"]).apply(null, arguments);
+        };
+        var _Cumprod = Module["_Cumprod"] = function() {
+          return (_Cumprod = Module["_Cumprod"] = Module["asm"]["Cumprod"]).apply(null, arguments);
+        };
+        var _Cumsum = Module["_Cumsum"] = function() {
+          return (_Cumsum = Module["_Cumsum"] = Module["asm"]["Cumsum"]).apply(null, arguments);
+        };
+        var _DepthToSpace = Module["_DepthToSpace"] = function() {
+          return (_DepthToSpace = Module["_DepthToSpace"] = Module["asm"]["DepthToSpace"]).apply(null, arguments);
+        };
+        var _DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = function() {
+          return (_DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = Module["asm"]["DepthwiseConv2dNative"]).apply(null, arguments);
+        };
+        var _Elu = Module["_Elu"] = function() {
+          return (_Elu = Module["_Elu"] = Module["asm"]["Elu"]).apply(null, arguments);
+        };
+        var _Equal = Module["_Equal"] = function() {
+          return (_Equal = Module["_Equal"] = Module["asm"]["Equal"]).apply(null, arguments);
+        };
+        var _Exp = Module["_Exp"] = function() {
+          return (_Exp = Module["_Exp"] = Module["asm"]["Exp"]).apply(null, arguments);
+        };
+        var _FlipLeftRight = Module["_FlipLeftRight"] = function() {
+          return (_FlipLeftRight = Module["_FlipLeftRight"] = Module["asm"]["FlipLeftRight"]).apply(null, arguments);
+        };
+        var _Floor = Module["_Floor"] = function() {
+          return (_Floor = Module["_Floor"] = Module["asm"]["Floor"]).apply(null, arguments);
+        };
+        var _FloorDiv = Module["_FloorDiv"] = function() {
+          return (_FloorDiv = Module["_FloorDiv"] = Module["asm"]["FloorDiv"]).apply(null, arguments);
+        };
+        var _FusedBatchNorm = Module["_FusedBatchNorm"] = function() {
+          return (_FusedBatchNorm = Module["_FusedBatchNorm"] = Module["asm"]["FusedBatchNorm"]).apply(null, arguments);
+        };
+        var _FusedConv2D = Module["_FusedConv2D"] = function() {
+          return (_FusedConv2D = Module["_FusedConv2D"] = Module["asm"]["FusedConv2D"]).apply(null, arguments);
+        };
+        var _FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = function() {
+          return (_FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = Module["asm"]["FusedDepthwiseConv2D"]).apply(null, arguments);
+        };
+        var _Gather = Module["_Gather"] = function() {
+          return (_Gather = Module["_Gather"] = Module["asm"]["Gather"]).apply(null, arguments);
+        };
+        var _GatherNd = Module["_GatherNd"] = function() {
+          return (_GatherNd = Module["_GatherNd"] = Module["asm"]["GatherNd"]).apply(null, arguments);
+        };
+        var _Greater = Module["_Greater"] = function() {
+          return (_Greater = Module["_Greater"] = Module["asm"]["Greater"]).apply(null, arguments);
+        };
+        var _GreaterEqual = Module["_GreaterEqual"] = function() {
+          return (_GreaterEqual = Module["_GreaterEqual"] = Module["asm"]["GreaterEqual"]).apply(null, arguments);
+        };
+        var _LeakyRelu = Module["_LeakyRelu"] = function() {
+          return (_LeakyRelu = Module["_LeakyRelu"] = Module["asm"]["LeakyRelu"]).apply(null, arguments);
+        };
+        var _Less = Module["_Less"] = function() {
+          return (_Less = Module["_Less"] = Module["asm"]["Less"]).apply(null, arguments);
+        };
+        var _LessEqual = Module["_LessEqual"] = function() {
+          return (_LessEqual = Module["_LessEqual"] = Module["asm"]["LessEqual"]).apply(null, arguments);
+        };
+        var _Log = Module["_Log"] = function() {
+          return (_Log = Module["_Log"] = Module["asm"]["Log"]).apply(null, arguments);
+        };
+        var _LogicalAnd = Module["_LogicalAnd"] = function() {
+          return (_LogicalAnd = Module["_LogicalAnd"] = Module["asm"]["LogicalAnd"]).apply(null, arguments);
+        };
+        var _LogicalNot = Module["_LogicalNot"] = function() {
+          return (_LogicalNot = Module["_LogicalNot"] = Module["asm"]["LogicalNot"]).apply(null, arguments);
+        };
+        var _LogicalOr = Module["_LogicalOr"] = function() {
+          return (_LogicalOr = Module["_LogicalOr"] = Module["asm"]["LogicalOr"]).apply(null, arguments);
+        };
+        var _LogicalXor = Module["_LogicalXor"] = function() {
+          return (_LogicalXor = Module["_LogicalXor"] = Module["asm"]["LogicalXor"]).apply(null, arguments);
+        };
+        var _Max = Module["_Max"] = function() {
+          return (_Max = Module["_Max"] = Module["asm"]["Max"]).apply(null, arguments);
+        };
+        var _MaxPool = Module["_MaxPool"] = function() {
+          return (_MaxPool = Module["_MaxPool"] = Module["asm"]["MaxPool"]).apply(null, arguments);
+        };
+        var _Maximum = Module["_Maximum"] = function() {
+          return (_Maximum = Module["_Maximum"] = Module["asm"]["Maximum"]).apply(null, arguments);
+        };
+        var _Mean = Module["_Mean"] = function() {
+          return (_Mean = Module["_Mean"] = Module["asm"]["Mean"]).apply(null, arguments);
+        };
+        var _Min = Module["_Min"] = function() {
+          return (_Min = Module["_Min"] = Module["asm"]["Min"]).apply(null, arguments);
+        };
+        var _Minimum = Module["_Minimum"] = function() {
+          return (_Minimum = Module["_Minimum"] = Module["asm"]["Minimum"]).apply(null, arguments);
+        };
+        var _MirrorPad = Module["_MirrorPad"] = function() {
+          return (_MirrorPad = Module["_MirrorPad"] = Module["asm"]["MirrorPad"]).apply(null, arguments);
+        };
+        var _Multiply = Module["_Multiply"] = function() {
+          return (_Multiply = Module["_Multiply"] = Module["asm"]["Multiply"]).apply(null, arguments);
+        };
+        var _Neg = Module["_Neg"] = function() {
+          return (_Neg = Module["_Neg"] = Module["asm"]["Neg"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = function() {
+          return (_NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = Module["asm"]["NonMaxSuppressionV3"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = function() {
+          return (_NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = Module["asm"]["NonMaxSuppressionV4"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = function() {
+          return (_NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = Module["asm"]["NonMaxSuppressionV5"]).apply(null, arguments);
+        };
+        var _NotEqual = Module["_NotEqual"] = function() {
+          return (_NotEqual = Module["_NotEqual"] = Module["asm"]["NotEqual"]).apply(null, arguments);
+        };
+        var _OneHot = Module["_OneHot"] = function() {
+          return (_OneHot = Module["_OneHot"] = Module["asm"]["OneHot"]).apply(null, arguments);
+        };
+        var _PadV2 = Module["_PadV2"] = function() {
+          return (_PadV2 = Module["_PadV2"] = Module["asm"]["PadV2"]).apply(null, arguments);
+        };
+        var _Pow = Module["_Pow"] = function() {
+          return (_Pow = Module["_Pow"] = Module["asm"]["Pow"]).apply(null, arguments);
+        };
+        var _Prelu = Module["_Prelu"] = function() {
+          return (_Prelu = Module["_Prelu"] = Module["asm"]["Prelu"]).apply(null, arguments);
+        };
+        var _Prod = Module["_Prod"] = function() {
+          return (_Prod = Module["_Prod"] = Module["asm"]["Prod"]).apply(null, arguments);
+        };
+        var _RealDiv = Module["_RealDiv"] = function() {
+          return (_RealDiv = Module["_RealDiv"] = Module["asm"]["RealDiv"]).apply(null, arguments);
+        };
+        var _Relu = Module["_Relu"] = function() {
+          return (_Relu = Module["_Relu"] = Module["asm"]["Relu"]).apply(null, arguments);
+        };
+        var _Relu6 = Module["_Relu6"] = function() {
+          return (_Relu6 = Module["_Relu6"] = Module["asm"]["Relu6"]).apply(null, arguments);
+        };
+        var _ResizeBilinear = Module["_ResizeBilinear"] = function() {
+          return (_ResizeBilinear = Module["_ResizeBilinear"] = Module["asm"]["ResizeBilinear"]).apply(null, arguments);
+        };
+        var _ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = function() {
+          return (_ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = Module["asm"]["ResizeNearestNeighbor"]).apply(null, arguments);
+        };
+        var _Reverse = Module["_Reverse"] = function() {
+          return (_Reverse = Module["_Reverse"] = Module["asm"]["Reverse"]).apply(null, arguments);
+        };
+        var _RotateWithOffset = Module["_RotateWithOffset"] = function() {
+          return (_RotateWithOffset = Module["_RotateWithOffset"] = Module["asm"]["RotateWithOffset"]).apply(null, arguments);
+        };
+        var _Round = Module["_Round"] = function() {
+          return (_Round = Module["_Round"] = Module["asm"]["Round"]).apply(null, arguments);
+        };
+        var _Rsqrt = Module["_Rsqrt"] = function() {
+          return (_Rsqrt = Module["_Rsqrt"] = Module["asm"]["Rsqrt"]).apply(null, arguments);
+        };
+        var _ScatterNd = Module["_ScatterNd"] = function() {
+          return (_ScatterNd = Module["_ScatterNd"] = Module["asm"]["ScatterNd"]).apply(null, arguments);
+        };
+        var _SelectV2 = Module["_SelectV2"] = function() {
+          return (_SelectV2 = Module["_SelectV2"] = Module["asm"]["SelectV2"]).apply(null, arguments);
+        };
+        var _Sigmoid = Module["_Sigmoid"] = function() {
+          return (_Sigmoid = Module["_Sigmoid"] = Module["asm"]["Sigmoid"]).apply(null, arguments);
+        };
+        var _Sin = Module["_Sin"] = function() {
+          return (_Sin = Module["_Sin"] = Module["asm"]["Sin"]).apply(null, arguments);
+        };
+        var _Softmax = Module["_Softmax"] = function() {
+          return (_Softmax = Module["_Softmax"] = Module["asm"]["Softmax"]).apply(null, arguments);
+        };
+        var _SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = function() {
+          return (_SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = Module["asm"]["SparseFillEmptyRows"]).apply(null, arguments);
+        };
+        var _SparseReshape = Module["_SparseReshape"] = function() {
+          return (_SparseReshape = Module["_SparseReshape"] = Module["asm"]["SparseReshape"]).apply(null, arguments);
+        };
+        var _SparseSegmentReduction = Module["_SparseSegmentReduction"] = function() {
+          return (_SparseSegmentReduction = Module["_SparseSegmentReduction"] = Module["asm"]["SparseSegmentReduction"]).apply(null, arguments);
+        };
+        var _Sqrt = Module["_Sqrt"] = function() {
+          return (_Sqrt = Module["_Sqrt"] = Module["asm"]["Sqrt"]).apply(null, arguments);
+        };
+        var _Square = Module["_Square"] = function() {
+          return (_Square = Module["_Square"] = Module["asm"]["Square"]).apply(null, arguments);
+        };
+        var _SquaredDifference = Module["_SquaredDifference"] = function() {
+          return (_SquaredDifference = Module["_SquaredDifference"] = Module["asm"]["SquaredDifference"]).apply(null, arguments);
+        };
+        var _Step = Module["_Step"] = function() {
+          return (_Step = Module["_Step"] = Module["asm"]["Step"]).apply(null, arguments);
+        };
+        var _StridedSlice = Module["_StridedSlice"] = function() {
+          return (_StridedSlice = Module["_StridedSlice"] = Module["asm"]["StridedSlice"]).apply(null, arguments);
+        };
+        var _Sub = Module["_Sub"] = function() {
+          return (_Sub = Module["_Sub"] = Module["asm"]["Sub"]).apply(null, arguments);
+        };
+        var _Sum = Module["_Sum"] = function() {
+          return (_Sum = Module["_Sum"] = Module["asm"]["Sum"]).apply(null, arguments);
+        };
+        var _Tan = Module["_Tan"] = function() {
+          return (_Tan = Module["_Tan"] = Module["asm"]["Tan"]).apply(null, arguments);
+        };
+        var _Tanh = Module["_Tanh"] = function() {
+          return (_Tanh = Module["_Tanh"] = Module["asm"]["Tanh"]).apply(null, arguments);
+        };
+        var _Tile = Module["_Tile"] = function() {
+          return (_Tile = Module["_Tile"] = Module["asm"]["Tile"]).apply(null, arguments);
+        };
+        var _TopK = Module["_TopK"] = function() {
+          return (_TopK = Module["_TopK"] = Module["asm"]["TopK"]).apply(null, arguments);
+        };
+        var _Transform = Module["_Transform"] = function() {
+          return (_Transform = Module["_Transform"] = Module["asm"]["Transform"]).apply(null, arguments);
+        };
+        var _Transpose = Module["_Transpose"] = function() {
+          return (_Transpose = Module["_Transpose"] = Module["asm"]["Transpose"]).apply(null, arguments);
+        };
+        var __FusedMatMul = Module["__FusedMatMul"] = function() {
+          return (__FusedMatMul = Module["__FusedMatMul"] = Module["asm"]["_FusedMatMul"]).apply(null, arguments);
+        };
+        var _malloc = Module["_malloc"] = function() {
+          return (_malloc = Module["_malloc"] = Module["asm"]["malloc"]).apply(null, arguments);
+        };
+        var _free = Module["_free"] = function() {
+          return (_free = Module["_free"] = Module["asm"]["free"]).apply(null, arguments);
+        };
+        var __emscripten_tls_init = Module["__emscripten_tls_init"] = function() {
+          return (__emscripten_tls_init = Module["__emscripten_tls_init"] = Module["asm"]["_emscripten_tls_init"]).apply(null, arguments);
+        };
+        var _pthread_self = Module["_pthread_self"] = function() {
+          return (_pthread_self = Module["_pthread_self"] = Module["asm"]["pthread_self"]).apply(null, arguments);
+        };
+        var ___errno_location = Module["___errno_location"] = function() {
+          return (___errno_location = Module["___errno_location"] = Module["asm"]["__errno_location"]).apply(null, arguments);
+        };
+        var __emscripten_thread_init = Module["__emscripten_thread_init"] = function() {
+          return (__emscripten_thread_init = Module["__emscripten_thread_init"] = Module["asm"]["_emscripten_thread_init"]).apply(null, arguments);
+        };
+        var __emscripten_thread_crashed = Module["__emscripten_thread_crashed"] = function() {
+          return (__emscripten_thread_crashed = Module["__emscripten_thread_crashed"] = Module["asm"]["_emscripten_thread_crashed"]).apply(null, arguments);
+        };
+        var _emscripten_main_thread_process_queued_calls = Module["_emscripten_main_thread_process_queued_calls"] = function() {
+          return (_emscripten_main_thread_process_queued_calls = Module["_emscripten_main_thread_process_queued_calls"] = Module["asm"]["emscripten_main_thread_process_queued_calls"]).apply(null, arguments);
+        };
+        var _emscripten_main_browser_thread_id = Module["_emscripten_main_browser_thread_id"] = function() {
+          return (_emscripten_main_browser_thread_id = Module["_emscripten_main_browser_thread_id"] = Module["asm"]["emscripten_main_browser_thread_id"]).apply(null, arguments);
+        };
+        var _emscripten_run_in_main_runtime_thread_js = Module["_emscripten_run_in_main_runtime_thread_js"] = function() {
+          return (_emscripten_run_in_main_runtime_thread_js = Module["_emscripten_run_in_main_runtime_thread_js"] = Module["asm"]["emscripten_run_in_main_runtime_thread_js"]).apply(null, arguments);
+        };
+        var _emscripten_dispatch_to_thread_ = Module["_emscripten_dispatch_to_thread_"] = function() {
+          return (_emscripten_dispatch_to_thread_ = Module["_emscripten_dispatch_to_thread_"] = Module["asm"]["emscripten_dispatch_to_thread_"]).apply(null, arguments);
+        };
+        var __emscripten_proxy_execute_task_queue = Module["__emscripten_proxy_execute_task_queue"] = function() {
+          return (__emscripten_proxy_execute_task_queue = Module["__emscripten_proxy_execute_task_queue"] = Module["asm"]["_emscripten_proxy_execute_task_queue"]).apply(null, arguments);
+        };
+        var __emscripten_thread_free_data = Module["__emscripten_thread_free_data"] = function() {
+          return (__emscripten_thread_free_data = Module["__emscripten_thread_free_data"] = Module["asm"]["_emscripten_thread_free_data"]).apply(null, arguments);
+        };
+        var __emscripten_thread_exit = Module["__emscripten_thread_exit"] = function() {
+          return (__emscripten_thread_exit = Module["__emscripten_thread_exit"] = Module["asm"]["_emscripten_thread_exit"]).apply(null, arguments);
+        };
+        var _emscripten_stack_set_limits = Module["_emscripten_stack_set_limits"] = function() {
+          return (_emscripten_stack_set_limits = Module["_emscripten_stack_set_limits"] = Module["asm"]["emscripten_stack_set_limits"]).apply(null, arguments);
+        };
+        var stackSave = Module["stackSave"] = function() {
+          return (stackSave = Module["stackSave"] = Module["asm"]["stackSave"]).apply(null, arguments);
+        };
+        var stackRestore = Module["stackRestore"] = function() {
+          return (stackRestore = Module["stackRestore"] = Module["asm"]["stackRestore"]).apply(null, arguments);
+        };
+        var stackAlloc = Module["stackAlloc"] = function() {
+          return (stackAlloc = Module["stackAlloc"] = Module["asm"]["stackAlloc"]).apply(null, arguments);
+        };
+        var dynCall_iijjiiii = Module["dynCall_iijjiiii"] = function() {
+          return (dynCall_iijjiiii = Module["dynCall_iijjiiii"] = Module["asm"]["dynCall_iijjiiii"]).apply(null, arguments);
+        };
+        var dynCall_jiji = Module["dynCall_jiji"] = function() {
+          return (dynCall_jiji = Module["dynCall_jiji"] = Module["asm"]["dynCall_jiji"]).apply(null, arguments);
+        };
+        Module["keepRuntimeAlive"] = keepRuntimeAlive;
+        Module["wasmMemory"] = wasmMemory;
+        Module["cwrap"] = cwrap;
+        Module["ExitStatus"] = ExitStatus;
+        Module["PThread"] = PThread;
+        var calledRun;
+        dependenciesFulfilled = function runCaller() {
+          if (!calledRun)
+            run();
+          if (!calledRun)
+            dependenciesFulfilled = runCaller;
+        };
+        function run(args) {
+          args = args || arguments_;
+          if (runDependencies > 0) {
+            return;
+          }
+          if (ENVIRONMENT_IS_PTHREAD) {
+            readyPromiseResolve(Module);
+            initRuntime();
+            postMessage({ "cmd": "loaded" });
+            return;
+          }
+          preRun();
+          if (runDependencies > 0) {
+            return;
+          }
+          function doRun() {
+            if (calledRun)
+              return;
+            calledRun = true;
+            Module["calledRun"] = true;
+            if (ABORT)
+              return;
+            initRuntime();
+            readyPromiseResolve(Module);
+            if (Module["onRuntimeInitialized"])
+              Module["onRuntimeInitialized"]();
+            postRun();
+          }
+          if (Module["setStatus"]) {
+            Module["setStatus"]("Running...");
+            setTimeout(function() {
+              setTimeout(function() {
+                Module["setStatus"]("");
+              }, 1);
+              doRun();
+            }, 1);
+          } else {
+            doRun();
+          }
+        }
+        if (Module["preInit"]) {
+          if (typeof Module["preInit"] == "function")
+            Module["preInit"] = [Module["preInit"]];
+          while (Module["preInit"].length > 0) {
+            Module["preInit"].pop()();
+          }
+        }
+        run();
+        var listenersAdded;
+        if (beforeListeners) {
+          listenersAdded = { uncaughtException: process.listeners("uncaughtException").filter(function(listener) {
+            return !beforeListeners.uncaughtException.indexOf(listener) > -1;
+          }), unhandledRejection: process.listeners("unhandledRejection").filter(function(listener) {
+            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;
+          }) };
+        }
+        var actualModule;
+        if (typeof WasmBackendModule !== "undefined") {
+          actualModule = WasmBackendModule;
+        } else if (typeof WasmBackendModuleThreadedSimd3 !== "undefined") {
+          actualModule = WasmBackendModuleThreadedSimd3;
+        } else {
+          throw new Error("Could not find wasm module in post.js");
+        }
+        if (listenersAdded) {
+          var tmpDispose = actualModule["_dispose"];
+          actualModule["_dispose"] = function() {
+            tmpDispose();
+            listenersAdded.uncaughtException.forEach(function(listener) {
+              process.removeListener("uncaughtException", listener);
+            });
+            listenersAdded.unhandledRejection.forEach(function(listener) {
+              process.removeListener("unhandledRejection", listener);
+            });
+          };
+        }
+        return WasmBackendModuleThreadedSimd3.ready;
+      };
+    })();
+    if (typeof exports === "object" && typeof module === "object")
+      module.exports = WasmBackendModuleThreadedSimd2;
+    else if (typeof define === "function" && define["amd"])
+      define([], function() {
+        return WasmBackendModuleThreadedSimd2;
+      });
+    else if (typeof exports === "object")
+      exports["WasmBackendModuleThreadedSimd"] = WasmBackendModuleThreadedSimd2;
+  }
+});
+var require_tfjs_backend_wasm_threaded_simd_worker = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js"(exports, module) {
+    module.exports.wasmWorkerContents = `"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
+");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`;
+  }
+});
+var require_tfjs_backend_wasm = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(exports, module) {
+    var WasmBackendModule2 = (() => {
+      var _scriptDir = typeof document !== "undefined" && document.currentScript ? document.currentScript.src : void 0;
+      if (typeof __filename !== "undefined")
+        _scriptDir = _scriptDir || __filename;
+      return function(WasmBackendModule3) {
+        WasmBackendModule3 = WasmBackendModule3 || {};
+        var Module = typeof WasmBackendModule3 != "undefined" ? WasmBackendModule3 : {};
+        var readyPromiseResolve, readyPromiseReject;
+        Module["ready"] = new Promise(function(resolve, reject) {
+          readyPromiseResolve = resolve;
+          readyPromiseReject = reject;
+        });
+        var beforeListeners;
+        if (typeof process !== "undefined" && process.listeners) {
+          beforeListeners = { uncaughtException: process.listeners("uncaughtException"), unhandledRejection: process.listeners("unhandledRejection") };
+        }
+        var moduleOverrides = Object.assign({}, Module);
+        var arguments_ = [];
+        var thisProgram = "./this.program";
+        var quit_ = (status, toThrow) => {
+          throw toThrow;
+        };
+        var ENVIRONMENT_IS_WEB = typeof window == "object";
+        var ENVIRONMENT_IS_WORKER = typeof importScripts == "function";
+        var ENVIRONMENT_IS_NODE = typeof process == "object" && typeof process.versions == "object" && typeof process.versions.node == "string";
+        var scriptDirectory = "";
+        function locateFile(path) {
+          if (Module["locateFile"]) {
+            return Module["locateFile"](path, scriptDirectory);
+          }
+          return scriptDirectory + path;
+        }
+        var read_, readAsync, readBinary, setWindowTitle;
+        function logExceptionOnExit(e) {
+          if (e instanceof ExitStatus)
+            return;
+          let toLog = e;
+          err("exiting due to exception: " + toLog);
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = require_path().dirname(scriptDirectory) + "/";
+          } else {
+            scriptDirectory = __dirname + "/";
+          }
+          var fs, nodePath;
+          if (typeof __require2 === "function") {
+            fs = require_fs();
+            nodePath = require_path();
+          }
+          read_ = (filename, binary) => {
+            filename = nodePath["normalize"](filename);
+            return fs.readFileSync(filename, binary ? void 0 : "utf8");
+          };
+          readBinary = (filename) => {
+            var ret = read_(filename, true);
+            if (!ret.buffer) {
+              ret = new Uint8Array(ret);
+            }
+            return ret;
+          };
+          readAsync = (filename, onload, onerror) => {
+            filename = nodePath["normalize"](filename);
+            fs.readFile(filename, function(err2, data) {
+              if (err2)
+                onerror(err2);
+              else
+                onload(data.buffer);
+            });
+          };
+          if (process["argv"].length > 1) {
+            thisProgram = process["argv"][1].replace(/\\/g, "/");
+          }
+          arguments_ = process["argv"].slice(2);
+          process["on"]("uncaughtException", function(ex) {
+            if (!(ex instanceof ExitStatus)) {
+              throw ex;
+            }
+          });
+          process["on"]("unhandledRejection", function(reason) {
+            throw reason;
+          });
+          quit_ = (status, toThrow) => {
+            if (keepRuntimeAlive()) {
+              process["exitCode"] = status;
+              throw toThrow;
+            }
+            logExceptionOnExit(toThrow);
+            process["exit"](status);
+          };
+          Module["inspect"] = function() {
+            return "[Emscripten Module object]";
+          };
+        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = self.location.href;
+          } else if (typeof document != "undefined" && document.currentScript) {
+            scriptDirectory = document.currentScript.src;
+          }
+          if (_scriptDir) {
+            scriptDirectory = _scriptDir;
+          }
+          if (scriptDirectory.indexOf("blob:") !== 0) {
+            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, "").lastIndexOf("/") + 1);
+          } else {
+            scriptDirectory = "";
+          }
+          {
+            read_ = (url) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, false);
+              xhr.send(null);
+              return xhr.responseText;
+            };
+            if (ENVIRONMENT_IS_WORKER) {
+              readBinary = (url) => {
+                var xhr = new XMLHttpRequest();
+                xhr.open("GET", url, false);
+                xhr.responseType = "arraybuffer";
+                xhr.send(null);
+                return new Uint8Array(xhr.response);
+              };
+            }
+            readAsync = (url, onload, onerror) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, true);
+              xhr.responseType = "arraybuffer";
+              xhr.onload = () => {
+                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {
+                  onload(xhr.response);
+                  return;
+                }
+                onerror();
+              };
+              xhr.onerror = onerror;
+              xhr.send(null);
+            };
+          }
+          setWindowTitle = (title) => document.title = title;
+        } else {
+        }
+        var out = Module["print"] || console.log.bind(console);
+        var err = Module["printErr"] || console.warn.bind(console);
+        Object.assign(Module, moduleOverrides);
+        moduleOverrides = null;
+        if (Module["arguments"])
+          arguments_ = Module["arguments"];
+        if (Module["thisProgram"])
+          thisProgram = Module["thisProgram"];
+        if (Module["quit"])
+          quit_ = Module["quit"];
+        var POINTER_SIZE = 4;
+        var wasmBinary;
+        if (Module["wasmBinary"])
+          wasmBinary = Module["wasmBinary"];
+        var noExitRuntime = Module["noExitRuntime"] || true;
+        if (typeof WebAssembly != "object") {
+          abort("no native wasm support detected");
+        }
+        var wasmMemory;
+        var ABORT = false;
+        var EXITSTATUS;
+        function assert3(condition, text) {
+          if (!condition) {
+            abort(text);
+          }
+        }
+        var UTF8Decoder = typeof TextDecoder != "undefined" ? new TextDecoder("utf8") : void 0;
+        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {
+          var endIdx = idx + maxBytesToRead;
+          var endPtr = idx;
+          while (heapOrArray[endPtr] && !(endPtr >= endIdx))
+            ++endPtr;
+          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {
+            return UTF8Decoder.decode(heapOrArray.subarray(idx, endPtr));
+          }
+          var str = "";
+          while (idx < endPtr) {
+            var u0 = heapOrArray[idx++];
+            if (!(u0 & 128)) {
+              str += String.fromCharCode(u0);
+              continue;
+            }
+            var u1 = heapOrArray[idx++] & 63;
+            if ((u0 & 224) == 192) {
+              str += String.fromCharCode((u0 & 31) << 6 | u1);
+              continue;
+            }
+            var u2 = heapOrArray[idx++] & 63;
+            if ((u0 & 240) == 224) {
+              u0 = (u0 & 15) << 12 | u1 << 6 | u2;
+            } else {
+              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;
+            }
+            if (u0 < 65536) {
+              str += String.fromCharCode(u0);
+            } else {
+              var ch = u0 - 65536;
+              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);
+            }
+          }
+          return str;
+        }
+        function UTF8ToString(ptr, maxBytesToRead) {
+          return ptr ? UTF8ArrayToString(HEAPU8, ptr, maxBytesToRead) : "";
+        }
+        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {
+          if (!(maxBytesToWrite > 0))
+            return 0;
+          var startIdx = outIdx;
+          var endIdx = outIdx + maxBytesToWrite - 1;
+          for (var i = 0; i < str.length; ++i) {
+            var u = str.charCodeAt(i);
+            if (u >= 55296 && u <= 57343) {
+              var u1 = str.charCodeAt(++i);
+              u = 65536 + ((u & 1023) << 10) | u1 & 1023;
+            }
+            if (u <= 127) {
+              if (outIdx >= endIdx)
+                break;
+              heap[outIdx++] = u;
+            } else if (u <= 2047) {
+              if (outIdx + 1 >= endIdx)
+                break;
+              heap[outIdx++] = 192 | u >> 6;
+              heap[outIdx++] = 128 | u & 63;
+            } else if (u <= 65535) {
+              if (outIdx + 2 >= endIdx)
+                break;
+              heap[outIdx++] = 224 | u >> 12;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            } else {
+              if (outIdx + 3 >= endIdx)
+                break;
+              heap[outIdx++] = 240 | u >> 18;
+              heap[outIdx++] = 128 | u >> 12 & 63;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            }
+          }
+          heap[outIdx] = 0;
+          return outIdx - startIdx;
+        }
+        function stringToUTF8(str, outPtr, maxBytesToWrite) {
+          return stringToUTF8Array(str, HEAPU8, outPtr, maxBytesToWrite);
+        }
+        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;
+        function updateGlobalBufferAndViews(buf) {
+          buffer2 = buf;
+          Module["HEAP8"] = HEAP8 = new Int8Array(buf);
+          Module["HEAP16"] = HEAP16 = new Int16Array(buf);
+          Module["HEAP32"] = HEAP32 = new Int32Array(buf);
+          Module["HEAPU8"] = HEAPU8 = new Uint8Array(buf);
+          Module["HEAPU16"] = HEAPU16 = new Uint16Array(buf);
+          Module["HEAPU32"] = HEAPU32 = new Uint32Array(buf);
+          Module["HEAPF32"] = HEAPF32 = new Float32Array(buf);
+          Module["HEAPF64"] = HEAPF64 = new Float64Array(buf);
+        }
+        var INITIAL_MEMORY = Module["INITIAL_MEMORY"] || 16777216;
+        var wasmTable;
+        var __ATPRERUN__ = [];
+        var __ATINIT__ = [];
+        var __ATPOSTRUN__ = [];
+        var runtimeInitialized = false;
+        function keepRuntimeAlive() {
+          return noExitRuntime;
+        }
+        function preRun() {
+          if (Module["preRun"]) {
+            if (typeof Module["preRun"] == "function")
+              Module["preRun"] = [Module["preRun"]];
+            while (Module["preRun"].length) {
+              addOnPreRun(Module["preRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPRERUN__);
+        }
+        function initRuntime() {
+          runtimeInitialized = true;
+          callRuntimeCallbacks(__ATINIT__);
+        }
+        function postRun() {
+          if (Module["postRun"]) {
+            if (typeof Module["postRun"] == "function")
+              Module["postRun"] = [Module["postRun"]];
+            while (Module["postRun"].length) {
+              addOnPostRun(Module["postRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPOSTRUN__);
+        }
+        function addOnPreRun(cb) {
+          __ATPRERUN__.unshift(cb);
+        }
+        function addOnInit(cb) {
+          __ATINIT__.unshift(cb);
+        }
+        function addOnPostRun(cb) {
+          __ATPOSTRUN__.unshift(cb);
+        }
+        var runDependencies = 0;
+        var runDependencyWatcher = null;
+        var dependenciesFulfilled = null;
+        function addRunDependency(id) {
+          runDependencies++;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+        }
+        function removeRunDependency(id) {
+          runDependencies--;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+          if (runDependencies == 0) {
+            if (runDependencyWatcher !== null) {
+              clearInterval(runDependencyWatcher);
+              runDependencyWatcher = null;
+            }
+            if (dependenciesFulfilled) {
+              var callback = dependenciesFulfilled;
+              dependenciesFulfilled = null;
+              callback();
+            }
+          }
+        }
+        function abort(what) {
+          {
+            if (Module["onAbort"]) {
+              Module["onAbort"](what);
+            }
+          }
+          what = "Aborted(" + what + ")";
+          err(what);
+          ABORT = true;
+          EXITSTATUS = 1;
+          what += ". Build with -sASSERTIONS for more info.";
+          var e = new WebAssembly.RuntimeError(what);
+          readyPromiseReject(e);
+          throw e;
+        }
+        var dataURIPrefix = "data:application/octet-stream;base64,";
+        function isDataURI(filename) {
+          return filename.startsWith(dataURIPrefix);
+        }
+        function isFileURI(filename) {
+          return filename.startsWith("file://");
+        }
+        var wasmBinaryFile;
+        wasmBinaryFile = "tfjs-backend-wasm.wasm";
+        if (!isDataURI(wasmBinaryFile)) {
+          wasmBinaryFile = locateFile(wasmBinaryFile);
+        }
+        function getBinary(file) {
+          try {
+            if (file == wasmBinaryFile && wasmBinary) {
+              return new Uint8Array(wasmBinary);
+            }
+            if (readBinary) {
+              return readBinary(file);
+            }
+            throw "both async and sync fetching of the wasm failed";
+          } catch (err2) {
+            abort(err2);
+          }
+        }
+        function getBinaryPromise() {
+          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {
+            if (typeof fetch == "function" && !isFileURI(wasmBinaryFile)) {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                if (!response["ok"]) {
+                  throw "failed to load wasm binary file at '" + wasmBinaryFile + "'";
+                }
+                return response["arrayBuffer"]();
+              }).catch(function() {
+                return getBinary(wasmBinaryFile);
+              });
+            } else {
+              if (readAsync) {
+                return new Promise(function(resolve, reject) {
+                  readAsync(wasmBinaryFile, function(response) {
+                    resolve(new Uint8Array(response));
+                  }, reject);
+                });
+              }
+            }
+          }
+          return Promise.resolve().then(function() {
+            return getBinary(wasmBinaryFile);
+          });
+        }
+        function createWasm() {
+          var info = { "env": asmLibraryArg, "wasi_snapshot_preview1": asmLibraryArg };
+          function receiveInstance(instance, module2) {
+            var exports3 = instance.exports;
+            Module["asm"] = exports3;
+            wasmMemory = Module["asm"]["memory"];
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            wasmTable = Module["asm"]["__indirect_function_table"];
+            addOnInit(Module["asm"]["__wasm_call_ctors"]);
+            removeRunDependency("wasm-instantiate");
+          }
+          addRunDependency("wasm-instantiate");
+          function receiveInstantiationResult(result) {
+            receiveInstance(result["instance"]);
+          }
+          function instantiateArrayBuffer(receiver) {
+            return getBinaryPromise().then(function(binary) {
+              return WebAssembly.instantiate(binary, info);
+            }).then(function(instance) {
+              return instance;
+            }).then(receiver, function(reason) {
+              err("failed to asynchronously prepare wasm: " + reason);
+              abort(reason);
+            });
+          }
+          function instantiateAsync() {
+            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == "function" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == "function") {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                var result = WebAssembly.instantiateStreaming(response, info);
+                return result.then(receiveInstantiationResult, function(reason) {
+                  err("wasm streaming compile failed: " + reason);
+                  err("falling back to ArrayBuffer instantiation");
+                  return instantiateArrayBuffer(receiveInstantiationResult);
+                });
+              });
+            } else {
+              return instantiateArrayBuffer(receiveInstantiationResult);
+            }
+          }
+          if (Module["instantiateWasm"]) {
+            try {
+              var exports2 = Module["instantiateWasm"](info, receiveInstance);
+              return exports2;
+            } catch (e) {
+              err("Module.instantiateWasm callback failed with error: " + e);
+              readyPromiseReject(e);
+            }
+          }
+          instantiateAsync().catch(readyPromiseReject);
+          return {};
+        }
+        var tempDouble;
+        var tempI64;
+        function ExitStatus(status) {
+          this.name = "ExitStatus";
+          this.message = "Program terminated with exit(" + status + ")";
+          this.status = status;
+        }
+        function callRuntimeCallbacks(callbacks2) {
+          while (callbacks2.length > 0) {
+            callbacks2.shift()(Module);
+          }
+        }
+        function demangle(func2) {
+          return func2;
+        }
+        function demangleAll(text) {
+          var regex = /\b_Z[\w\d_]+/g;
+          return text.replace(regex, function(x) {
+            var y = demangle(x);
+            return x === y ? x : y + " [" + x + "]";
+          });
+        }
+        function jsStackTrace() {
+          var error = new Error();
+          if (!error.stack) {
+            try {
+              throw new Error();
+            } catch (e) {
+              error = e;
+            }
+            if (!error.stack) {
+              return "(no stack trace available)";
+            }
+          }
+          return error.stack.toString();
+        }
+        function writeArrayToMemory(array2, buffer3) {
+          HEAP8.set(array2, buffer3);
+        }
+        function _abort() {
+          abort("");
+        }
+        function getHeapMax() {
+          return 2147483648;
+        }
+        function _emscripten_get_heap_max() {
+          return getHeapMax();
+        }
+        function _emscripten_memcpy_big(dest, src, num) {
+          HEAPU8.copyWithin(dest, src, src + num);
+        }
+        function emscripten_realloc_buffer(size) {
+          try {
+            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            return 1;
+          } catch (e) {
+          }
+        }
+        function _emscripten_resize_heap(requestedSize) {
+          var oldSize = HEAPU8.length;
+          requestedSize = requestedSize >>> 0;
+          var maxHeapSize = getHeapMax();
+          if (requestedSize > maxHeapSize) {
+            return false;
+          }
+          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;
+          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {
+            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);
+            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);
+            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));
+            var replacement = emscripten_realloc_buffer(newSize);
+            if (replacement) {
+              return true;
+            }
+          }
+          return false;
+        }
+        var SYSCALLS = { varargs: void 0, get: function() {
+          SYSCALLS.varargs += 4;
+          var ret = HEAP32[SYSCALLS.varargs - 4 >> 2];
+          return ret;
+        }, getStr: function(ptr) {
+          var ret = UTF8ToString(ptr);
+          return ret;
+        } };
+        function _fd_close(fd) {
+          return 52;
+        }
+        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {
+          return 70;
+        }
+        var printCharBuffers = [null, [], []];
+        function printChar(stream, curr) {
+          var buffer3 = printCharBuffers[stream];
+          if (curr === 0 || curr === 10) {
+            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));
+            buffer3.length = 0;
+          } else {
+            buffer3.push(curr);
+          }
+        }
+        function _fd_write(fd, iov, iovcnt, pnum) {
+          var num = 0;
+          for (var i = 0; i < iovcnt; i++) {
+            var ptr = HEAPU32[iov >> 2];
+            var len = HEAPU32[iov + 4 >> 2];
+            iov += 8;
+            for (var j = 0; j < len; j++) {
+              printChar(fd, HEAPU8[ptr + j]);
+            }
+            num += len;
+          }
+          HEAPU32[pnum >> 2] = num;
+          return 0;
+        }
+        function getCFunc(ident) {
+          var func2 = Module["_" + ident];
+          return func2;
+        }
+        function ccall(ident, returnType, argTypes, args, opts) {
+          var toC = { "string": (str) => {
+            var ret2 = 0;
+            if (str !== null && str !== void 0 && str !== 0) {
+              var len = (str.length << 2) + 1;
+              ret2 = stackAlloc(len);
+              stringToUTF8(str, ret2, len);
+            }
+            return ret2;
+          }, "array": (arr) => {
+            var ret2 = stackAlloc(arr.length);
+            writeArrayToMemory(arr, ret2);
+            return ret2;
+          } };
+          function convertReturnValue(ret2) {
+            if (returnType === "string") {
+              return UTF8ToString(ret2);
+            }
+            if (returnType === "boolean")
+              return Boolean(ret2);
+            return ret2;
+          }
+          var func2 = getCFunc(ident);
+          var cArgs = [];
+          var stack2 = 0;
+          if (args) {
+            for (var i = 0; i < args.length; i++) {
+              var converter = toC[argTypes[i]];
+              if (converter) {
+                if (stack2 === 0)
+                  stack2 = stackSave();
+                cArgs[i] = converter(args[i]);
+              } else {
+                cArgs[i] = args[i];
+              }
+            }
+          }
+          var ret = func2.apply(null, cArgs);
+          function onDone(ret2) {
+            if (stack2 !== 0)
+              stackRestore(stack2);
+            return convertReturnValue(ret2);
+          }
+          ret = onDone(ret);
+          return ret;
+        }
+        function cwrap(ident, returnType, argTypes, opts) {
+          argTypes = argTypes || [];
+          var numericArgs = argTypes.every((type) => type === "number" || type === "boolean");
+          var numericRet = returnType !== "string";
+          if (numericRet && numericArgs && !opts) {
+            return getCFunc(ident);
+          }
+          return function() {
+            return ccall(ident, returnType, argTypes, arguments, opts);
+          };
+        }
+        var asmLibraryArg = { "abort": _abort, "emscripten_get_heap_max": _emscripten_get_heap_max, "emscripten_memcpy_big": _emscripten_memcpy_big, "emscripten_resize_heap": _emscripten_resize_heap, "fd_close": _fd_close, "fd_seek": _fd_seek, "fd_write": _fd_write };
+        var asm = createWasm();
+        var ___wasm_call_ctors = Module["___wasm_call_ctors"] = function() {
+          return (___wasm_call_ctors = Module["___wasm_call_ctors"] = Module["asm"]["__wasm_call_ctors"]).apply(null, arguments);
+        };
+        var _init = Module["_init"] = function() {
+          return (_init = Module["_init"] = Module["asm"]["init"]).apply(null, arguments);
+        };
+        var _init_with_threads_count = Module["_init_with_threads_count"] = function() {
+          return (_init_with_threads_count = Module["_init_with_threads_count"] = Module["asm"]["init_with_threads_count"]).apply(null, arguments);
+        };
+        var _get_threads_count = Module["_get_threads_count"] = function() {
+          return (_get_threads_count = Module["_get_threads_count"] = Module["asm"]["get_threads_count"]).apply(null, arguments);
+        };
+        var _register_tensor = Module["_register_tensor"] = function() {
+          return (_register_tensor = Module["_register_tensor"] = Module["asm"]["register_tensor"]).apply(null, arguments);
+        };
+        var _dispose_data = Module["_dispose_data"] = function() {
+          return (_dispose_data = Module["_dispose_data"] = Module["asm"]["dispose_data"]).apply(null, arguments);
+        };
+        var _dispose = Module["_dispose"] = function() {
+          return (_dispose = Module["_dispose"] = Module["asm"]["dispose"]).apply(null, arguments);
+        };
+        var _Abs = Module["_Abs"] = function() {
+          return (_Abs = Module["_Abs"] = Module["asm"]["Abs"]).apply(null, arguments);
+        };
+        var _Add = Module["_Add"] = function() {
+          return (_Add = Module["_Add"] = Module["asm"]["Add"]).apply(null, arguments);
+        };
+        var _AddN = Module["_AddN"] = function() {
+          return (_AddN = Module["_AddN"] = Module["asm"]["AddN"]).apply(null, arguments);
+        };
+        var _All = Module["_All"] = function() {
+          return (_All = Module["_All"] = Module["asm"]["All"]).apply(null, arguments);
+        };
+        var _Any = Module["_Any"] = function() {
+          return (_Any = Module["_Any"] = Module["asm"]["Any"]).apply(null, arguments);
+        };
+        var _ArgMax = Module["_ArgMax"] = function() {
+          return (_ArgMax = Module["_ArgMax"] = Module["asm"]["ArgMax"]).apply(null, arguments);
+        };
+        var _AvgPool = Module["_AvgPool"] = function() {
+          return (_AvgPool = Module["_AvgPool"] = Module["asm"]["AvgPool"]).apply(null, arguments);
+        };
+        var _BatchMatMul = Module["_BatchMatMul"] = function() {
+          return (_BatchMatMul = Module["_BatchMatMul"] = Module["asm"]["BatchMatMul"]).apply(null, arguments);
+        };
+        var _Ceil = Module["_Ceil"] = function() {
+          return (_Ceil = Module["_Ceil"] = Module["asm"]["Ceil"]).apply(null, arguments);
+        };
+        var _ClipByValue = Module["_ClipByValue"] = function() {
+          return (_ClipByValue = Module["_ClipByValue"] = Module["asm"]["ClipByValue"]).apply(null, arguments);
+        };
+        var _Conv2D = Module["_Conv2D"] = function() {
+          return (_Conv2D = Module["_Conv2D"] = Module["asm"]["Conv2D"]).apply(null, arguments);
+        };
+        var _Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = function() {
+          return (_Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = Module["asm"]["Conv2DBackpropInput"]).apply(null, arguments);
+        };
+        var _Cos = Module["_Cos"] = function() {
+          return (_Cos = Module["_Cos"] = Module["asm"]["Cos"]).apply(null, arguments);
+        };
+        var _Cosh = Module["_Cosh"] = function() {
+          return (_Cosh = Module["_Cosh"] = Module["asm"]["Cosh"]).apply(null, arguments);
+        };
+        var _CropAndResize = Module["_CropAndResize"] = function() {
+          return (_CropAndResize = Module["_CropAndResize"] = Module["asm"]["CropAndResize"]).apply(null, arguments);
+        };
+        var _Cumprod = Module["_Cumprod"] = function() {
+          return (_Cumprod = Module["_Cumprod"] = Module["asm"]["Cumprod"]).apply(null, arguments);
+        };
+        var _Cumsum = Module["_Cumsum"] = function() {
+          return (_Cumsum = Module["_Cumsum"] = Module["asm"]["Cumsum"]).apply(null, arguments);
+        };
+        var _DepthToSpace = Module["_DepthToSpace"] = function() {
+          return (_DepthToSpace = Module["_DepthToSpace"] = Module["asm"]["DepthToSpace"]).apply(null, arguments);
+        };
+        var _DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = function() {
+          return (_DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = Module["asm"]["DepthwiseConv2dNative"]).apply(null, arguments);
+        };
+        var _Elu = Module["_Elu"] = function() {
+          return (_Elu = Module["_Elu"] = Module["asm"]["Elu"]).apply(null, arguments);
+        };
+        var _Equal = Module["_Equal"] = function() {
+          return (_Equal = Module["_Equal"] = Module["asm"]["Equal"]).apply(null, arguments);
+        };
+        var _Exp = Module["_Exp"] = function() {
+          return (_Exp = Module["_Exp"] = Module["asm"]["Exp"]).apply(null, arguments);
+        };
+        var _FlipLeftRight = Module["_FlipLeftRight"] = function() {
+          return (_FlipLeftRight = Module["_FlipLeftRight"] = Module["asm"]["FlipLeftRight"]).apply(null, arguments);
+        };
+        var _Floor = Module["_Floor"] = function() {
+          return (_Floor = Module["_Floor"] = Module["asm"]["Floor"]).apply(null, arguments);
+        };
+        var _FloorDiv = Module["_FloorDiv"] = function() {
+          return (_FloorDiv = Module["_FloorDiv"] = Module["asm"]["FloorDiv"]).apply(null, arguments);
+        };
+        var _FusedBatchNorm = Module["_FusedBatchNorm"] = function() {
+          return (_FusedBatchNorm = Module["_FusedBatchNorm"] = Module["asm"]["FusedBatchNorm"]).apply(null, arguments);
+        };
+        var _FusedConv2D = Module["_FusedConv2D"] = function() {
+          return (_FusedConv2D = Module["_FusedConv2D"] = Module["asm"]["FusedConv2D"]).apply(null, arguments);
+        };
+        var _FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = function() {
+          return (_FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = Module["asm"]["FusedDepthwiseConv2D"]).apply(null, arguments);
+        };
+        var _Gather = Module["_Gather"] = function() {
+          return (_Gather = Module["_Gather"] = Module["asm"]["Gather"]).apply(null, arguments);
+        };
+        var _GatherNd = Module["_GatherNd"] = function() {
+          return (_GatherNd = Module["_GatherNd"] = Module["asm"]["GatherNd"]).apply(null, arguments);
+        };
+        var _Greater = Module["_Greater"] = function() {
+          return (_Greater = Module["_Greater"] = Module["asm"]["Greater"]).apply(null, arguments);
+        };
+        var _GreaterEqual = Module["_GreaterEqual"] = function() {
+          return (_GreaterEqual = Module["_GreaterEqual"] = Module["asm"]["GreaterEqual"]).apply(null, arguments);
+        };
+        var _LeakyRelu = Module["_LeakyRelu"] = function() {
+          return (_LeakyRelu = Module["_LeakyRelu"] = Module["asm"]["LeakyRelu"]).apply(null, arguments);
+        };
+        var _Less = Module["_Less"] = function() {
+          return (_Less = Module["_Less"] = Module["asm"]["Less"]).apply(null, arguments);
+        };
+        var _LessEqual = Module["_LessEqual"] = function() {
+          return (_LessEqual = Module["_LessEqual"] = Module["asm"]["LessEqual"]).apply(null, arguments);
+        };
+        var _Log = Module["_Log"] = function() {
+          return (_Log = Module["_Log"] = Module["asm"]["Log"]).apply(null, arguments);
+        };
+        var _LogicalAnd = Module["_LogicalAnd"] = function() {
+          return (_LogicalAnd = Module["_LogicalAnd"] = Module["asm"]["LogicalAnd"]).apply(null, arguments);
+        };
+        var _LogicalNot = Module["_LogicalNot"] = function() {
+          return (_LogicalNot = Module["_LogicalNot"] = Module["asm"]["LogicalNot"]).apply(null, arguments);
+        };
+        var _LogicalOr = Module["_LogicalOr"] = function() {
+          return (_LogicalOr = Module["_LogicalOr"] = Module["asm"]["LogicalOr"]).apply(null, arguments);
+        };
+        var _LogicalXor = Module["_LogicalXor"] = function() {
+          return (_LogicalXor = Module["_LogicalXor"] = Module["asm"]["LogicalXor"]).apply(null, arguments);
+        };
+        var _Max = Module["_Max"] = function() {
+          return (_Max = Module["_Max"] = Module["asm"]["Max"]).apply(null, arguments);
+        };
+        var _MaxPool = Module["_MaxPool"] = function() {
+          return (_MaxPool = Module["_MaxPool"] = Module["asm"]["MaxPool"]).apply(null, arguments);
+        };
+        var _Maximum = Module["_Maximum"] = function() {
+          return (_Maximum = Module["_Maximum"] = Module["asm"]["Maximum"]).apply(null, arguments);
+        };
+        var _Mean = Module["_Mean"] = function() {
+          return (_Mean = Module["_Mean"] = Module["asm"]["Mean"]).apply(null, arguments);
+        };
+        var _Min = Module["_Min"] = function() {
+          return (_Min = Module["_Min"] = Module["asm"]["Min"]).apply(null, arguments);
+        };
+        var _Minimum = Module["_Minimum"] = function() {
+          return (_Minimum = Module["_Minimum"] = Module["asm"]["Minimum"]).apply(null, arguments);
+        };
+        var _MirrorPad = Module["_MirrorPad"] = function() {
+          return (_MirrorPad = Module["_MirrorPad"] = Module["asm"]["MirrorPad"]).apply(null, arguments);
+        };
+        var _Multiply = Module["_Multiply"] = function() {
+          return (_Multiply = Module["_Multiply"] = Module["asm"]["Multiply"]).apply(null, arguments);
+        };
+        var _Neg = Module["_Neg"] = function() {
+          return (_Neg = Module["_Neg"] = Module["asm"]["Neg"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = function() {
+          return (_NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = Module["asm"]["NonMaxSuppressionV3"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = function() {
+          return (_NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = Module["asm"]["NonMaxSuppressionV4"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = function() {
+          return (_NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = Module["asm"]["NonMaxSuppressionV5"]).apply(null, arguments);
+        };
+        var _NotEqual = Module["_NotEqual"] = function() {
+          return (_NotEqual = Module["_NotEqual"] = Module["asm"]["NotEqual"]).apply(null, arguments);
+        };
+        var _OneHot = Module["_OneHot"] = function() {
+          return (_OneHot = Module["_OneHot"] = Module["asm"]["OneHot"]).apply(null, arguments);
+        };
+        var _PadV2 = Module["_PadV2"] = function() {
+          return (_PadV2 = Module["_PadV2"] = Module["asm"]["PadV2"]).apply(null, arguments);
+        };
+        var _Pow = Module["_Pow"] = function() {
+          return (_Pow = Module["_Pow"] = Module["asm"]["Pow"]).apply(null, arguments);
+        };
+        var _Prelu = Module["_Prelu"] = function() {
+          return (_Prelu = Module["_Prelu"] = Module["asm"]["Prelu"]).apply(null, arguments);
+        };
+        var _Prod = Module["_Prod"] = function() {
+          return (_Prod = Module["_Prod"] = Module["asm"]["Prod"]).apply(null, arguments);
+        };
+        var _RealDiv = Module["_RealDiv"] = function() {
+          return (_RealDiv = Module["_RealDiv"] = Module["asm"]["RealDiv"]).apply(null, arguments);
+        };
+        var _Relu = Module["_Relu"] = function() {
+          return (_Relu = Module["_Relu"] = Module["asm"]["Relu"]).apply(null, arguments);
+        };
+        var _Relu6 = Module["_Relu6"] = function() {
+          return (_Relu6 = Module["_Relu6"] = Module["asm"]["Relu6"]).apply(null, arguments);
+        };
+        var _ResizeBilinear = Module["_ResizeBilinear"] = function() {
+          return (_ResizeBilinear = Module["_ResizeBilinear"] = Module["asm"]["ResizeBilinear"]).apply(null, arguments);
+        };
+        var _ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = function() {
+          return (_ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = Module["asm"]["ResizeNearestNeighbor"]).apply(null, arguments);
+        };
+        var _Reverse = Module["_Reverse"] = function() {
+          return (_Reverse = Module["_Reverse"] = Module["asm"]["Reverse"]).apply(null, arguments);
+        };
+        var _RotateWithOffset = Module["_RotateWithOffset"] = function() {
+          return (_RotateWithOffset = Module["_RotateWithOffset"] = Module["asm"]["RotateWithOffset"]).apply(null, arguments);
+        };
+        var _Round = Module["_Round"] = function() {
+          return (_Round = Module["_Round"] = Module["asm"]["Round"]).apply(null, arguments);
+        };
+        var _Rsqrt = Module["_Rsqrt"] = function() {
+          return (_Rsqrt = Module["_Rsqrt"] = Module["asm"]["Rsqrt"]).apply(null, arguments);
+        };
+        var _ScatterNd = Module["_ScatterNd"] = function() {
+          return (_ScatterNd = Module["_ScatterNd"] = Module["asm"]["ScatterNd"]).apply(null, arguments);
+        };
+        var _SelectV2 = Module["_SelectV2"] = function() {
+          return (_SelectV2 = Module["_SelectV2"] = Module["asm"]["SelectV2"]).apply(null, arguments);
+        };
+        var _Sigmoid = Module["_Sigmoid"] = function() {
+          return (_Sigmoid = Module["_Sigmoid"] = Module["asm"]["Sigmoid"]).apply(null, arguments);
+        };
+        var _Sin = Module["_Sin"] = function() {
+          return (_Sin = Module["_Sin"] = Module["asm"]["Sin"]).apply(null, arguments);
+        };
+        var _Softmax = Module["_Softmax"] = function() {
+          return (_Softmax = Module["_Softmax"] = Module["asm"]["Softmax"]).apply(null, arguments);
+        };
+        var _SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = function() {
+          return (_SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = Module["asm"]["SparseFillEmptyRows"]).apply(null, arguments);
+        };
+        var _SparseReshape = Module["_SparseReshape"] = function() {
+          return (_SparseReshape = Module["_SparseReshape"] = Module["asm"]["SparseReshape"]).apply(null, arguments);
+        };
+        var _SparseSegmentReduction = Module["_SparseSegmentReduction"] = function() {
+          return (_SparseSegmentReduction = Module["_SparseSegmentReduction"] = Module["asm"]["SparseSegmentReduction"]).apply(null, arguments);
+        };
+        var _Sqrt = Module["_Sqrt"] = function() {
+          return (_Sqrt = Module["_Sqrt"] = Module["asm"]["Sqrt"]).apply(null, arguments);
+        };
+        var _Square = Module["_Square"] = function() {
+          return (_Square = Module["_Square"] = Module["asm"]["Square"]).apply(null, arguments);
+        };
+        var _SquaredDifference = Module["_SquaredDifference"] = function() {
+          return (_SquaredDifference = Module["_SquaredDifference"] = Module["asm"]["SquaredDifference"]).apply(null, arguments);
+        };
+        var _Step = Module["_Step"] = function() {
+          return (_Step = Module["_Step"] = Module["asm"]["Step"]).apply(null, arguments);
+        };
+        var _StridedSlice = Module["_StridedSlice"] = function() {
+          return (_StridedSlice = Module["_StridedSlice"] = Module["asm"]["StridedSlice"]).apply(null, arguments);
+        };
+        var _Sub = Module["_Sub"] = function() {
+          return (_Sub = Module["_Sub"] = Module["asm"]["Sub"]).apply(null, arguments);
+        };
+        var _Sum = Module["_Sum"] = function() {
+          return (_Sum = Module["_Sum"] = Module["asm"]["Sum"]).apply(null, arguments);
+        };
+        var _Tan = Module["_Tan"] = function() {
+          return (_Tan = Module["_Tan"] = Module["asm"]["Tan"]).apply(null, arguments);
+        };
+        var _Tanh = Module["_Tanh"] = function() {
+          return (_Tanh = Module["_Tanh"] = Module["asm"]["Tanh"]).apply(null, arguments);
+        };
+        var _Tile = Module["_Tile"] = function() {
+          return (_Tile = Module["_Tile"] = Module["asm"]["Tile"]).apply(null, arguments);
+        };
+        var _TopK = Module["_TopK"] = function() {
+          return (_TopK = Module["_TopK"] = Module["asm"]["TopK"]).apply(null, arguments);
+        };
+        var _Transform = Module["_Transform"] = function() {
+          return (_Transform = Module["_Transform"] = Module["asm"]["Transform"]).apply(null, arguments);
+        };
+        var _Transpose = Module["_Transpose"] = function() {
+          return (_Transpose = Module["_Transpose"] = Module["asm"]["Transpose"]).apply(null, arguments);
+        };
+        var __FusedMatMul = Module["__FusedMatMul"] = function() {
+          return (__FusedMatMul = Module["__FusedMatMul"] = Module["asm"]["_FusedMatMul"]).apply(null, arguments);
+        };
+        var _malloc = Module["_malloc"] = function() {
+          return (_malloc = Module["_malloc"] = Module["asm"]["malloc"]).apply(null, arguments);
+        };
+        var _free = Module["_free"] = function() {
+          return (_free = Module["_free"] = Module["asm"]["free"]).apply(null, arguments);
+        };
+        var ___errno_location = Module["___errno_location"] = function() {
+          return (___errno_location = Module["___errno_location"] = Module["asm"]["__errno_location"]).apply(null, arguments);
+        };
+        var stackSave = Module["stackSave"] = function() {
+          return (stackSave = Module["stackSave"] = Module["asm"]["stackSave"]).apply(null, arguments);
+        };
+        var stackRestore = Module["stackRestore"] = function() {
+          return (stackRestore = Module["stackRestore"] = Module["asm"]["stackRestore"]).apply(null, arguments);
+        };
+        var stackAlloc = Module["stackAlloc"] = function() {
+          return (stackAlloc = Module["stackAlloc"] = Module["asm"]["stackAlloc"]).apply(null, arguments);
+        };
+        var dynCall_iijjiiii = Module["dynCall_iijjiiii"] = function() {
+          return (dynCall_iijjiiii = Module["dynCall_iijjiiii"] = Module["asm"]["dynCall_iijjiiii"]).apply(null, arguments);
+        };
+        var dynCall_jiji = Module["dynCall_jiji"] = function() {
+          return (dynCall_jiji = Module["dynCall_jiji"] = Module["asm"]["dynCall_jiji"]).apply(null, arguments);
+        };
+        Module["cwrap"] = cwrap;
+        var calledRun;
+        dependenciesFulfilled = function runCaller() {
+          if (!calledRun)
+            run();
+          if (!calledRun)
+            dependenciesFulfilled = runCaller;
+        };
+        function run(args) {
+          args = args || arguments_;
+          if (runDependencies > 0) {
+            return;
+          }
+          preRun();
+          if (runDependencies > 0) {
+            return;
+          }
+          function doRun() {
+            if (calledRun)
+              return;
+            calledRun = true;
+            Module["calledRun"] = true;
+            if (ABORT)
+              return;
+            initRuntime();
+            readyPromiseResolve(Module);
+            if (Module["onRuntimeInitialized"])
+              Module["onRuntimeInitialized"]();
+            postRun();
+          }
+          if (Module["setStatus"]) {
+            Module["setStatus"]("Running...");
+            setTimeout(function() {
+              setTimeout(function() {
+                Module["setStatus"]("");
+              }, 1);
+              doRun();
+            }, 1);
+          } else {
+            doRun();
+          }
+        }
+        if (Module["preInit"]) {
+          if (typeof Module["preInit"] == "function")
+            Module["preInit"] = [Module["preInit"]];
+          while (Module["preInit"].length > 0) {
+            Module["preInit"].pop()();
+          }
+        }
+        run();
+        var listenersAdded;
+        if (beforeListeners) {
+          listenersAdded = { uncaughtException: process.listeners("uncaughtException").filter(function(listener) {
+            return !beforeListeners.uncaughtException.indexOf(listener) > -1;
+          }), unhandledRejection: process.listeners("unhandledRejection").filter(function(listener) {
+            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;
+          }) };
+        }
+        var actualModule;
+        if (typeof WasmBackendModule3 !== "undefined") {
+          actualModule = WasmBackendModule3;
+        } else if (typeof WasmBackendModuleThreadedSimd !== "undefined") {
+          actualModule = WasmBackendModuleThreadedSimd;
+        } else {
+          throw new Error("Could not find wasm module in post.js");
+        }
+        if (listenersAdded) {
+          var tmpDispose = actualModule["_dispose"];
+          actualModule["_dispose"] = function() {
+            tmpDispose();
+            listenersAdded.uncaughtException.forEach(function(listener) {
+              process.removeListener("uncaughtException", listener);
+            });
+            listenersAdded.unhandledRejection.forEach(function(listener) {
+              process.removeListener("unhandledRejection", listener);
+            });
+          };
+        }
+        return WasmBackendModule3.ready;
+      };
+    })();
+    if (typeof exports === "object" && typeof module === "object")
+      module.exports = WasmBackendModule2;
+    else if (typeof define === "function" && define["amd"])
+      define([], function() {
+        return WasmBackendModule2;
+      });
+    else if (typeof exports === "object")
+      exports["WasmBackendModule"] = WasmBackendModule2;
+  }
+});
+var EPSILON_FLOAT32 = 1e-7;
+var EPSILON_FLOAT16 = 1e-4;
+var DataStorage = class {
+  constructor(backend2, dataMover) {
+    this.backend = backend2;
+    this.dataMover = dataMover;
+    this.data = /* @__PURE__ */ new WeakMap();
+    this.dataIdsCount = 0;
+  }
+  get(dataId) {
+    if (!this.data.has(dataId)) {
+      this.dataMover.moveData(this.backend, dataId);
+    }
+    return this.data.get(dataId);
+  }
+  set(dataId, value) {
+    this.dataIdsCount++;
+    this.data.set(dataId, value);
+  }
+  has(dataId) {
+    return this.data.has(dataId);
+  }
+  delete(dataId) {
+    this.dataIdsCount--;
+    return this.data.delete(dataId);
+  }
+  numDataIds() {
+    return this.dataIdsCount;
+  }
+};
+var KernelBackend = class {
+  refCount(dataId) {
+    return notYetImplemented("refCount");
+  }
+  incRef(dataId) {
+    return notYetImplemented("incRef");
+  }
+  timerAvailable() {
+    return true;
+  }
+  time(f) {
+    return notYetImplemented("time");
+  }
+  read(dataId) {
+    return notYetImplemented("read");
+  }
+  readSync(dataId) {
+    return notYetImplemented("readSync");
+  }
+  readToGPU(dataId, options) {
+    return notYetImplemented("readToGPU");
+  }
+  numDataIds() {
+    return notYetImplemented("numDataIds");
+  }
+  disposeData(dataId, force) {
+    return notYetImplemented("disposeData");
+  }
+  write(values, shape, dtype) {
+    return notYetImplemented("write");
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    return notYetImplemented("move");
+  }
+  createTensorFromTexture(values, shape, dtype) {
+    return notYetImplemented("createTensorFromTexture");
+  }
+  memory() {
+    return notYetImplemented("memory");
+  }
+  floatPrecision() {
+    return notYetImplemented("floatPrecision");
+  }
+  epsilon() {
+    return this.floatPrecision() === 32 ? EPSILON_FLOAT32 : EPSILON_FLOAT16;
+  }
+  dispose() {
+    return notYetImplemented("dispose");
+  }
+};
+function notYetImplemented(kernelName) {
+  throw new Error(`'${kernelName}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`);
+}
+function shuffle(array2) {
+  let counter = array2.length;
+  let index = 0;
+  while (counter > 0) {
+    index = Math.random() * counter | 0;
+    counter--;
+    swap(array2, counter, index);
+  }
+}
+function shuffleCombo(array2, array22) {
+  if (array2.length !== array22.length) {
+    throw new Error(`Array sizes must match to be shuffled together First array length was ${array2.length}Second array length was ${array22.length}`);
+  }
+  let counter = array2.length;
+  let index = 0;
+  while (counter > 0) {
+    index = Math.random() * counter | 0;
+    counter--;
+    swap(array2, counter, index);
+    swap(array22, counter, index);
+  }
+}
+function clamp(min6, x, max6) {
+  return Math.max(min6, Math.min(x, max6));
+}
+function nearestLargerEven(val) {
+  return val % 2 === 0 ? val : val + 1;
+}
+function swap(object, left, right) {
+  const temp = object[left];
+  object[left] = object[right];
+  object[right] = temp;
+}
+function sum(arr) {
+  let sum6 = 0;
+  for (let i = 0; i < arr.length; i++) {
+    sum6 += arr[i];
+  }
+  return sum6;
+}
+function randUniform(a, b) {
+  const r = Math.random();
+  return b * r + (1 - r) * a;
+}
+function distSquared(a, b) {
+  let result = 0;
+  for (let i = 0; i < a.length; i++) {
+    const diff = Number(a[i]) - Number(b[i]);
+    result += diff * diff;
+  }
+  return result;
+}
+function assert(expr, msg) {
+  if (!expr) {
+    throw new Error(typeof msg === "string" ? msg : msg());
+  }
+}
+function assertShapesMatch(shapeA, shapeB, errorMessagePrefix = "") {
+  assert(arraysEqual(shapeA, shapeB), () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+}
+function assertNonNull(a) {
+  assert(a != null, () => `The input to the tensor constructor must be a non-null value.`);
+}
+function flatten(arr, result = [], skipTypedArray = false) {
+  if (result == null) {
+    result = [];
+  }
+  if (Array.isArray(arr) || isTypedArray(arr) && !skipTypedArray) {
+    for (let i = 0; i < arr.length; ++i) {
+      flatten(arr[i], result, skipTypedArray);
+    }
+  } else {
+    result.push(arr);
+  }
+  return result;
+}
+function sizeFromShape(shape) {
+  if (shape.length === 0) {
+    return 1;
+  }
+  let size = shape[0];
+  for (let i = 1; i < shape.length; i++) {
+    size *= shape[i];
+  }
+  return size;
+}
+function isScalarShape(shape) {
+  return shape.length === 0;
+}
+function arraysEqual(n1, n2) {
+  if (n1 === n2) {
+    return true;
+  }
+  if (n1 == null || n2 == null) {
+    return false;
+  }
+  if (n1.length !== n2.length) {
+    return false;
+  }
+  for (let i = 0; i < n1.length; i++) {
+    if (n1[i] !== n2[i]) {
+      return false;
+    }
+  }
+  return true;
+}
+function isInt(a) {
+  return a % 1 === 0;
+}
+function tanh(x) {
+  if (Math.tanh != null) {
+    return Math.tanh(x);
+  }
+  if (x === Infinity) {
+    return 1;
+  } else if (x === -Infinity) {
+    return -1;
+  } else {
+    const e2x = Math.exp(2 * x);
+    return (e2x - 1) / (e2x + 1);
+  }
+}
+function sizeToSquarishShape(size) {
+  const width = Math.ceil(Math.sqrt(size));
+  return [width, Math.ceil(size / width)];
+}
+function createShuffledIndices(n) {
+  const shuffledIndices = new Uint32Array(n);
+  for (let i = 0; i < n; ++i) {
+    shuffledIndices[i] = i;
+  }
+  shuffle(shuffledIndices);
+  return shuffledIndices;
+}
+function rightPad(a, size) {
+  if (size <= a.length) {
+    return a;
+  }
+  return a + " ".repeat(size - a.length);
+}
+function repeatedTry(checkFn, delayFn = (counter) => 0, maxCounter, scheduleFn) {
+  return new Promise((resolve, reject) => {
+    let tryCount = 0;
+    const tryFn = () => {
+      if (checkFn()) {
+        resolve();
+        return;
+      }
+      tryCount++;
+      const nextBackoff = delayFn(tryCount);
+      if (maxCounter != null && tryCount >= maxCounter) {
+        reject();
+        return;
+      }
+      if (scheduleFn != null) {
+        scheduleFn(tryFn, nextBackoff);
+      } else {
+        setTimeout(tryFn, nextBackoff);
+      }
+    };
+    tryFn();
+  });
+}
+function inferFromImplicitShape(shape, size) {
+  let shapeProd = 1;
+  let implicitIdx = -1;
+  for (let i = 0; i < shape.length; ++i) {
+    if (shape[i] >= 0) {
+      shapeProd *= shape[i];
+    } else if (shape[i] === -1) {
+      if (implicitIdx !== -1) {
+        throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${implicitIdx} and dim ${i}`);
+      }
+      implicitIdx = i;
+    } else if (shape[i] < 0) {
+      throw Error(`Shapes can not be < 0. Found ${shape[i]} at dim ${i}`);
+    }
+  }
+  if (implicitIdx === -1) {
+    if (size > 0 && size !== shapeProd) {
+      throw Error(`Size(${size}) must match the product of shape ${shape}`);
+    }
+    return shape;
+  }
+  if (shapeProd === 0) {
+    throw Error(`Cannot infer the missing size in [${shape}] when there are 0 elements`);
+  }
+  if (size % shapeProd !== 0) {
+    throw Error(`The implicit shape can't be a fractional number. Got ${size} / ${shapeProd}`);
+  }
+  const newShape = shape.slice();
+  newShape[implicitIdx] = size / shapeProd;
+  return newShape;
+}
+function parseAxisParam(axis, shape) {
+  const rank = shape.length;
+  axis = axis == null ? shape.map((s, i) => i) : [].concat(axis);
+  assert(axis.every((ax) => ax >= -rank && ax < rank), () => `All values in axis param must be in range [-${rank}, ${rank}) but got axis ${axis}`);
+  assert(axis.every((ax) => isInt(ax)), () => `All values in axis param must be integers but got axis ${axis}`);
+  return axis.map((a) => a < 0 ? rank + a : a);
+}
+function squeezeShape(shape, axis) {
+  const newShape = [];
+  const keptDims = [];
+  const isEmptyArray = axis != null && Array.isArray(axis) && axis.length === 0;
+  const axes = axis == null || isEmptyArray ? null : parseAxisParam(axis, shape).sort();
+  let j = 0;
+  for (let i = 0; i < shape.length; ++i) {
+    if (axes != null) {
+      if (axes[j] === i && shape[i] !== 1) {
+        throw new Error(`Can't squeeze axis ${i} since its dim '${shape[i]}' is not 1`);
+      }
+      if ((axes[j] == null || axes[j] > i) && shape[i] === 1) {
+        newShape.push(shape[i]);
+        keptDims.push(i);
+      }
+      if (axes[j] <= i) {
+        j++;
+      }
+    }
+    if (shape[i] !== 1) {
+      newShape.push(shape[i]);
+      keptDims.push(i);
+    }
+  }
+  return { newShape, keptDims };
+}
+function getTypedArrayFromDType(dtype, size) {
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  return values;
+}
+function getArrayFromDType(dtype, size) {
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else if (dtype === "string") {
+    values = new Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  return values;
+}
+function checkConversionForErrors(vals, dtype) {
+  for (let i = 0; i < vals.length; i++) {
+    const num = vals[i];
+    if (isNaN(num) || !isFinite(num)) {
+      throw Error(`A tensor of type ${dtype} being uploaded contains ${num}.`);
+    }
+  }
+}
+function isValidDtype(dtype) {
+  return dtype === "bool" || dtype === "complex64" || dtype === "float32" || dtype === "int32" || dtype === "string";
+}
+function hasEncodingLoss(oldType, newType) {
+  if (newType === "complex64") {
+    return false;
+  }
+  if (newType === "float32" && oldType !== "complex64") {
+    return false;
+  }
+  if (newType === "int32" && oldType !== "float32" && oldType !== "complex64") {
+    return false;
+  }
+  if (newType === "bool" && oldType === "bool") {
+    return false;
+  }
+  return true;
+}
+function isTypedArray(a) {
+  return a instanceof Float32Array || a instanceof Int32Array || a instanceof Uint8Array || a instanceof Uint8ClampedArray;
+}
+function bytesPerElement(dtype) {
+  if (dtype === "float32" || dtype === "int32") {
+    return 4;
+  } else if (dtype === "complex64") {
+    return 8;
+  } else if (dtype === "bool") {
+    return 1;
+  } else {
+    throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+function bytesFromStringArray(arr) {
+  if (arr == null) {
+    return 0;
+  }
+  let bytes = 0;
+  arr.forEach((x) => bytes += x.length);
+  return bytes;
+}
+function isString(value) {
+  return typeof value === "string" || value instanceof String;
+}
+function isBoolean(value) {
+  return typeof value === "boolean";
+}
+function isNumber(value) {
+  return typeof value === "number";
+}
+function inferDtype(values) {
+  if (Array.isArray(values)) {
+    return inferDtype(values[0]);
+  }
+  if (values instanceof Float32Array) {
+    return "float32";
+  } else if (values instanceof Int32Array || values instanceof Uint8Array || values instanceof Uint8ClampedArray) {
+    return "int32";
+  } else if (isNumber(values)) {
+    return "float32";
+  } else if (isString(values)) {
+    return "string";
+  } else if (isBoolean(values)) {
+    return "bool";
+  }
+  return "float32";
+}
+function isFunction(f) {
+  return !!(f && f.constructor && f.call && f.apply);
+}
+function nearestDivisor(size, start) {
+  for (let i = start; i < size; ++i) {
+    if (size % i === 0) {
+      return i;
+    }
+  }
+  return size;
+}
+function computeStrides(shape) {
+  const rank = shape.length;
+  if (rank < 2) {
+    return [];
+  }
+  const strides = new Array(rank - 1);
+  strides[rank - 2] = shape[rank - 1];
+  for (let i = rank - 3; i >= 0; --i) {
+    strides[i] = strides[i + 1] * shape[i + 1];
+  }
+  return strides;
+}
+function createNestedArray(offset, shape, a, isComplex = false) {
+  const ret = new Array();
+  if (shape.length === 1) {
+    const d = shape[0] * (isComplex ? 2 : 1);
+    for (let i = 0; i < d; i++) {
+      ret[i] = a[offset + i];
+    }
+  } else {
+    const d = shape[0];
+    const rest = shape.slice(1);
+    const len = rest.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);
+    for (let i = 0; i < d; i++) {
+      ret[i] = createNestedArray(offset + i * len, rest, a, isComplex);
+    }
+  }
+  return ret;
+}
+function toNestedArray(shape, a, isComplex = false) {
+  if (shape.length === 0) {
+    return a[0];
+  }
+  const size = shape.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);
+  if (size === 0) {
+    return [];
+  }
+  if (size !== a.length) {
+    throw new Error(`[${shape}] does not match the input size ${a.length}${isComplex ? " for a complex tensor" : ""}.`);
+  }
+  return createNestedArray(0, shape, a, isComplex);
+}
+function makeOnesTypedArray(size, dtype) {
+  const array2 = makeZerosTypedArray(size, dtype);
+  for (let i = 0; i < array2.length; i++) {
+    array2[i] = 1;
+  }
+  return array2;
+}
+function makeZerosTypedArray(size, dtype) {
+  if (dtype == null || dtype === "float32" || dtype === "complex64") {
+    return new Float32Array(size);
+  } else if (dtype === "int32") {
+    return new Int32Array(size);
+  } else if (dtype === "bool") {
+    return new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function makeZerosNestedTypedArray(shape, dtype) {
+  const size = shape.reduce((prev, curr) => prev * curr, 1);
+  if (dtype == null || dtype === "float32") {
+    return toNestedArray(shape, new Float32Array(size));
+  } else if (dtype === "int32") {
+    return toNestedArray(shape, new Int32Array(size));
+  } else if (dtype === "bool") {
+    return toNestedArray(shape, new Uint8Array(size));
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function assertNonNegativeIntegerDimensions(shape) {
+  shape.forEach((dimSize) => {
+    assert(Number.isInteger(dimSize) && dimSize >= 0, () => `Tensor must have a shape comprised of positive integers but got shape [${shape}].`);
+  });
+}
+function locToIndex(locs, rank, strides) {
+  if (rank === 0) {
+    return 0;
+  } else if (rank === 1) {
+    return locs[0];
+  }
+  let index = locs[locs.length - 1];
+  for (let i = 0; i < locs.length - 1; ++i) {
+    index += strides[i] * locs[i];
+  }
+  return index;
+}
+function indexToLoc(index, rank, strides) {
+  if (rank === 0) {
+    return [];
+  } else if (rank === 1) {
+    return [index];
+  }
+  const locs = new Array(rank);
+  for (let i = 0; i < locs.length - 1; ++i) {
+    locs[i] = Math.floor(index / strides[i]);
+    index -= locs[i] * strides[i];
+  }
+  locs[locs.length - 1] = index;
+  return locs;
+}
+function isPromise(object) {
+  return object && object.then && typeof object.then === "function";
+}
+var TENSORFLOWJS_FLAGS_PREFIX = "tfjsflags";
+var Environment = class {
+  constructor(global2) {
+    this.global = global2;
+    this.flags = {};
+    this.flagRegistry = {};
+    this.urlFlags = {};
+    this.getQueryParams = getQueryParams;
+    this.populateURLFlags();
+  }
+  setPlatform(platformName, platform) {
+    if (this.platform != null) {
+      if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+        console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${platformName}.`);
+      }
+    }
+    this.platformName = platformName;
+    this.platform = platform;
+  }
+  registerFlag(flagName, evaluationFn, setHook) {
+    this.flagRegistry[flagName] = { evaluationFn, setHook };
+    if (this.urlFlags[flagName] != null) {
+      const flagValue = this.urlFlags[flagName];
+      if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+        console.warn(`Setting feature override from URL ${flagName}: ${flagValue}.`);
+      }
+      this.set(flagName, flagValue);
+    }
+  }
+  async getAsync(flagName) {
+    if (flagName in this.flags) {
+      return this.flags[flagName];
+    }
+    this.flags[flagName] = await this.evaluateFlag(flagName);
+    return this.flags[flagName];
+  }
+  get(flagName) {
+    if (flagName in this.flags) {
+      return this.flags[flagName];
+    }
+    const flagValue = this.evaluateFlag(flagName);
+    if (isPromise(flagValue)) {
+      throw new Error(`Flag ${flagName} cannot be synchronously evaluated. Please use getAsync() instead.`);
+    }
+    this.flags[flagName] = flagValue;
+    return this.flags[flagName];
+  }
+  getNumber(flagName) {
+    return this.get(flagName);
+  }
+  getBool(flagName) {
+    return this.get(flagName);
+  }
+  getFlags() {
+    return this.flags;
+  }
+  get features() {
+    return this.flags;
+  }
+  set(flagName, value) {
+    if (this.flagRegistry[flagName] == null) {
+      throw new Error(`Cannot set flag ${flagName} as it has not been registered.`);
+    }
+    this.flags[flagName] = value;
+    if (this.flagRegistry[flagName].setHook != null) {
+      this.flagRegistry[flagName].setHook(value);
+    }
+  }
+  evaluateFlag(flagName) {
+    if (this.flagRegistry[flagName] == null) {
+      throw new Error(`Cannot evaluate flag '${flagName}': no evaluation function found.`);
+    }
+    return this.flagRegistry[flagName].evaluationFn();
+  }
+  setFlags(flags) {
+    this.flags = Object.assign({}, flags);
+  }
+  reset() {
+    this.flags = {};
+    this.urlFlags = {};
+    this.populateURLFlags();
+  }
+  populateURLFlags() {
+    if (typeof this.global === "undefined" || typeof this.global.location === "undefined" || typeof this.global.location.search === "undefined") {
+      return;
+    }
+    const urlParams = this.getQueryParams(this.global.location.search);
+    if (TENSORFLOWJS_FLAGS_PREFIX in urlParams) {
+      const keyValues = urlParams[TENSORFLOWJS_FLAGS_PREFIX].split(",");
+      keyValues.forEach((keyValue) => {
+        const [key, value] = keyValue.split(":");
+        this.urlFlags[key] = parseValue(key, value);
+      });
+    }
+  }
+};
+function getQueryParams(queryString) {
+  const params = {};
+  queryString.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g, (s, ...t) => {
+    decodeParam(params, t[0], t[1]);
+    return t.join("=");
+  });
+  return params;
+}
+function decodeParam(params, name, value) {
+  params[decodeURIComponent(name)] = decodeURIComponent(value || "");
+}
+function parseValue(flagName, value) {
+  value = value.toLowerCase();
+  if (value === "true" || value === "false") {
+    return value === "true";
+  } else if (`${+value}` === value) {
+    return +value;
+  }
+  throw new Error(`Could not parse value flag value ${value} for flag ${flagName}.`);
+}
+function env() {
+  return ENV;
+}
+var ENV = null;
+function setEnvironmentGlobal(environment2) {
+  ENV = environment2;
+}
+var globalNameSpace;
+function getGlobalNamespace() {
+  if (globalNameSpace == null) {
+    let ns;
+    if (typeof window !== "undefined") {
+      ns = window;
+    } else if (typeof global !== "undefined") {
+      ns = global;
+    } else if (typeof process !== "undefined") {
+      ns = process;
+    } else if (typeof self !== "undefined") {
+      ns = self;
+    } else {
+      throw new Error("Could not find a global object");
+    }
+    globalNameSpace = ns;
+  }
+  return globalNameSpace;
+}
+function getGlobalMap() {
+  const ns = getGlobalNamespace();
+  if (ns._tfGlobals == null) {
+    ns._tfGlobals = /* @__PURE__ */ new Map();
+  }
+  return ns._tfGlobals;
+}
+function getGlobal(key, init2) {
+  const globalMap = getGlobalMap();
+  if (globalMap.has(key)) {
+    return globalMap.get(key);
+  } else {
+    const singleton = init2();
+    globalMap.set(key, singleton);
+    return globalMap.get(key);
+  }
+}
+var Abs = "Abs";
+var Acos = "Acos";
+var Acosh = "Acosh";
+var Add = "Add";
+var AddN = "AddN";
+var All = "All";
+var Any = "Any";
+var ArgMax = "ArgMax";
+var ArgMin = "ArgMin";
+var Asin = "Asin";
+var Asinh = "Asinh";
+var Atan = "Atan";
+var Atanh = "Atanh";
+var Atan2 = "Atan2";
+var AvgPool = "AvgPool";
+var AvgPoolGrad = "AvgPoolGrad";
+var AvgPool3D = "AvgPool3D";
+var AvgPool3DGrad = "AvgPool3DGrad";
+var BatchMatMul = "BatchMatMul";
+var BatchToSpaceND = "BatchToSpaceND";
+var Bincount = "Bincount";
+var BroadcastTo = "BroadcastTo";
+var BroadcastArgs = "BroadcastArgs";
+var Cast = "Cast";
+var Ceil = "Ceil";
+var ClipByValue = "ClipByValue";
+var Complex = "Complex";
+var ComplexAbs = "ComplexAbs";
+var Concat = "Concat";
+var Conv2D = "Conv2D";
+var Conv2DBackpropFilter = "Conv2DBackpropFilter";
+var Conv2DBackpropInput = "Conv2DBackpropInput";
+var Conv3D = "Conv3D";
+var Conv3DBackpropFilterV2 = "Conv3DBackpropFilterV2";
+var Conv3DBackpropInputV2 = "Conv3DBackpropInputV2";
+var Cos = "Cos";
+var Cosh = "Cosh";
+var Cumprod = "Cumprod";
+var Cumsum = "Cumsum";
+var CropAndResize = "CropAndResize";
+var DenseBincount = "DenseBincount";
+var DepthToSpace = "DepthToSpace";
+var DepthwiseConv2dNative = "DepthwiseConv2dNative";
+var DepthwiseConv2dNativeBackpropFilter = "DepthwiseConv2dNativeBackpropFilter";
+var DepthwiseConv2dNativeBackpropInput = "DepthwiseConv2dNativeBackpropInput";
+var Diag = "Diag";
+var Dilation2D = "Dilation2D";
+var Dilation2DBackpropInput = "Dilation2DBackpropInput";
+var Dilation2DBackpropFilter = "Dilation2DBackpropFilter";
+var RealDiv = "RealDiv";
+var Einsum = "Einsum";
+var Elu = "Elu";
+var EluGrad = "EluGrad";
+var Erf = "Erf";
+var Equal = "Equal";
+var Exp = "Exp";
+var ExpandDims = "ExpandDims";
+var Expm1 = "Expm1";
+var FFT = "FFT";
+var Fill = "Fill";
+var FlipLeftRight = "FlipLeftRight";
+var Floor = "Floor";
+var FloorDiv = "FloorDiv";
+var FusedBatchNorm = "FusedBatchNorm";
+var GatherV2 = "GatherV2";
+var GatherNd = "GatherNd";
+var Greater = "Greater";
+var GreaterEqual = "GreaterEqual";
+var Identity = "Identity";
+var IFFT = "IFFT";
+var Imag = "Imag";
+var IsFinite = "IsFinite";
+var IsInf = "IsInf";
+var IsNan = "IsNan";
+var LeakyRelu = "LeakyRelu";
+var Less = "Less";
+var LessEqual = "LessEqual";
+var LinSpace = "LinSpace";
+var Log = "Log";
+var Log1p = "Log1p";
+var LogicalAnd = "LogicalAnd";
+var LogicalNot = "LogicalNot";
+var LogicalOr = "LogicalOr";
+var LogicalXor = "LogicalXor";
+var LogSoftmax = "LogSoftmax";
+var LowerBound = "LowerBound";
+var LRN = "LRN";
+var LRNGrad = "LRNGrad";
+var Max = "Max";
+var Maximum = "Maximum";
+var MaxPool = "MaxPool";
+var MaxPoolGrad = "MaxPoolGrad";
+var MaxPool3D = "MaxPool3D";
+var MaxPool3DGrad = "MaxPool3DGrad";
+var MaxPoolWithArgmax = "MaxPoolWithArgmax";
+var Mean = "Mean";
+var Min = "Min";
+var Minimum = "Minimum";
+var MirrorPad = "MirrorPad";
+var Mod = "Mod";
+var Multinomial = "Multinomial";
+var Multiply = "Multiply";
+var Neg = "Neg";
+var NotEqual = "NotEqual";
+var NonMaxSuppressionV3 = "NonMaxSuppressionV3";
+var NonMaxSuppressionV4 = "NonMaxSuppressionV4";
+var NonMaxSuppressionV5 = "NonMaxSuppressionV5";
+var OnesLike = "OnesLike";
+var OneHot = "OneHot";
+var Pack = "Pack";
+var PadV2 = "PadV2";
+var Pool = "Pool";
+var Pow = "Pow";
+var Prelu = "Prelu";
+var Prod = "Prod";
+var RaggedGather = "RaggedGather";
+var RaggedRange = "RaggedRange";
+var RaggedTensorToTensor = "RaggedTensorToTensor";
+var Range = "Range";
+var Real = "Real";
+var Reciprocal = "Reciprocal";
+var Relu = "Relu";
+var Reshape = "Reshape";
+var ResizeNearestNeighbor = "ResizeNearestNeighbor";
+var ResizeNearestNeighborGrad = "ResizeNearestNeighborGrad";
+var ResizeBilinear = "ResizeBilinear";
+var ResizeBilinearGrad = "ResizeBilinearGrad";
+var Relu6 = "Relu6";
+var Reverse = "Reverse";
+var Round = "Round";
+var Rsqrt = "Rsqrt";
+var ScatterNd = "ScatterNd";
+var SearchSorted = "SearchSorted";
+var Select = "Select";
+var Selu = "Selu";
+var Slice = "Slice";
+var Sin = "Sin";
+var Sinh = "Sinh";
+var Sign = "Sign";
+var Sigmoid = "Sigmoid";
+var Softplus = "Softplus";
+var Sqrt = "Sqrt";
+var Sum = "Sum";
+var SpaceToBatchND = "SpaceToBatchND";
+var SplitV = "SplitV";
+var Softmax = "Softmax";
+var SparseFillEmptyRows = "SparseFillEmptyRows";
+var SparseReshape = "SparseReshape";
+var SparseSegmentMean = "SparseSegmentMean";
+var SparseSegmentSum = "SparseSegmentSum";
+var SparseToDense = "SparseToDense";
+var SquaredDifference = "SquaredDifference";
+var Square = "Square";
+var StridedSlice = "StridedSlice";
+var StringNGrams = "StringNGrams";
+var StringSplit = "StringSplit";
+var StringToHashBucketFast = "StringToHashBucketFast";
+var Sub = "Sub";
+var Tan = "Tan";
+var Tanh = "Tanh";
+var Tile = "Tile";
+var TopK = "TopK";
+var Transform = "Transform";
+var Transpose = "Transpose";
+var Unique = "Unique";
+var Unpack = "Unpack";
+var UnsortedSegmentSum = "UnsortedSegmentSum";
+var UpperBound = "UpperBound";
+var ZerosLike = "ZerosLike";
+var Step = "Step";
+var FromPixels = "FromPixels";
+var RotateWithOffset = "RotateWithOffset";
+var _FusedMatMul = "_FusedMatMul";
+var FusedConv2D = "FusedConv2D";
+var FusedDepthwiseConv2D = "FusedDepthwiseConv2D";
+function warn(...msg) {
+  if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+    console.warn(...msg);
+  }
+}
+function log(...msg) {
+  if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+    console.log(...msg);
+  }
+}
+var kernelRegistry = getGlobal("kernelRegistry", () => /* @__PURE__ */ new Map());
+var gradRegistry = getGlobal("gradRegistry", () => /* @__PURE__ */ new Map());
+function getKernel(kernelName, backendName) {
+  const key = makeKey(kernelName, backendName);
+  return kernelRegistry.get(key);
+}
+function getGradient(kernelName) {
+  return gradRegistry.get(kernelName);
+}
+function getKernelsForBackend(backendName) {
+  const it = kernelRegistry.entries();
+  const result = [];
+  while (true) {
+    const { done, value } = it.next();
+    if (done) {
+      break;
+    }
+    const [key, config] = value;
+    const [backend2] = key.split("_");
+    if (backend2 === backendName) {
+      result.push(config);
+    }
+  }
+  return result;
+}
+function registerKernel(config) {
+  const { kernelName, backendName } = config;
+  const key = makeKey(kernelName, backendName);
+  if (kernelRegistry.has(key)) {
+    warn(`The kernel '${kernelName}' for backend '${backendName}' is already registered`);
+  }
+  kernelRegistry.set(key, config);
+}
+function registerGradient(config) {
+  const { kernelName } = config;
+  if (gradRegistry.has(kernelName)) {
+    if (env().getBool("DEBUG")) {
+      warn(`Overriding the gradient for '${kernelName}'`);
+    }
+  }
+  gradRegistry.set(kernelName, config);
+}
+function unregisterKernel(kernelName, backendName) {
+  const key = makeKey(kernelName, backendName);
+  if (!kernelRegistry.has(key)) {
+    throw new Error(`The kernel '${kernelName}' for backend '${backendName}' is not registered`);
+  }
+  kernelRegistry.delete(key);
+}
+function unregisterGradient(kernelName) {
+  if (!gradRegistry.has(kernelName)) {
+    throw new Error(`The gradient '${kernelName}' for backend is not registered`);
+  }
+  gradRegistry.delete(kernelName);
+}
+function copyRegisteredKernels(registeredBackendName, newBackendName) {
+  const kernels = getKernelsForBackend(registeredBackendName);
+  kernels.forEach((kernelConfig) => {
+    const newKernelConfig = Object.assign({}, kernelConfig, { backendName: newBackendName });
+    registerKernel(newKernelConfig);
+  });
+}
+function makeKey(kernelName, backendName) {
+  return `${backendName}_${kernelName}`;
+}
+var util_exports = {};
+__export2(util_exports, {
+  arraysEqual: () => arraysEqual,
+  assert: () => assert,
+  assertNonNegativeIntegerDimensions: () => assertNonNegativeIntegerDimensions,
+  assertNonNull: () => assertNonNull,
+  assertShapesMatch: () => assertShapesMatch,
+  bytesFromStringArray: () => bytesFromStringArray,
+  bytesPerElement: () => bytesPerElement,
+  checkConversionForErrors: () => checkConversionForErrors,
+  clamp: () => clamp,
+  computeStrides: () => computeStrides,
+  createScalarValue: () => createScalarValue,
+  createShuffledIndices: () => createShuffledIndices,
+  decodeString: () => decodeString,
+  distSquared: () => distSquared,
+  encodeString: () => encodeString,
+  fetch: () => fetch3,
+  fingerPrint64: () => fingerPrint64,
+  flatten: () => flatten,
+  getArrayFromDType: () => getArrayFromDType,
+  getTypedArrayFromDType: () => getTypedArrayFromDType,
+  hasEncodingLoss: () => hasEncodingLoss,
+  hexToLong: () => hexToLong,
+  indexToLoc: () => indexToLoc,
+  inferDtype: () => inferDtype,
+  inferFromImplicitShape: () => inferFromImplicitShape,
+  isBoolean: () => isBoolean,
+  isFunction: () => isFunction,
+  isInt: () => isInt,
+  isNumber: () => isNumber,
+  isPromise: () => isPromise,
+  isScalarShape: () => isScalarShape,
+  isString: () => isString,
+  isTypedArray: () => isTypedArray,
+  isValidDtype: () => isValidDtype,
+  locToIndex: () => locToIndex,
+  makeOnesTypedArray: () => makeOnesTypedArray,
+  makeZerosNestedTypedArray: () => makeZerosNestedTypedArray,
+  makeZerosTypedArray: () => makeZerosTypedArray,
+  nearestDivisor: () => nearestDivisor,
+  nearestLargerEven: () => nearestLargerEven,
+  now: () => now,
+  parseAxisParam: () => parseAxisParam,
+  randUniform: () => randUniform,
+  repeatedTry: () => repeatedTry,
+  rightPad: () => rightPad,
+  shuffle: () => shuffle,
+  shuffleCombo: () => shuffleCombo,
+  sizeFromShape: () => sizeFromShape,
+  sizeToSquarishShape: () => sizeToSquarishShape,
+  squeezeShape: () => squeezeShape,
+  sum: () => sum,
+  swap: () => swap,
+  tanh: () => tanh,
+  toNestedArray: () => toNestedArray,
+  toTypedArray: () => toTypedArray
+});
+var LongExports = __toESM(require_long());
+var Long = LongExports.default || LongExports;
+function hexToLong(hex) {
+  return Long.fromString(hex, true, 16);
+}
+var k0 = hexToLong("c3a5c85c97cb3127");
+var k1 = hexToLong("b492b66fbe98f273");
+var k2 = hexToLong("9ae16a3b2f90404f");
+function shiftMix(val) {
+  return val.xor(val.shru(47));
+}
+function fetch2(s, offset, numBytes) {
+  const bytes = s.slice(offset, offset + numBytes);
+  return Long.fromBytes(Array.from(bytes), true, true);
+}
+function fetch64(s, offset) {
+  return fetch2(s, offset, 8);
+}
+function fetch32(s, offset) {
+  return fetch2(s, offset, 4);
+}
+function rotate64(val, shift) {
+  return shift === 0 ? val : val.shru(shift).or(val.shl(64 - shift));
+}
+function hashLen16(u, v, mul2 = hexToLong("9ddfea08eb382d69")) {
+  let a = u.xor(v).mul(mul2);
+  a = a.xor(a.shru(47));
+  let b = v.xor(a).mul(mul2);
+  b = b.xor(b.shru(47));
+  b = b.mul(mul2);
+  return b;
+}
+function weakHashLen32WithSeeds(w, x, y, z, a, b) {
+  a = a.add(w);
+  b = rotate64(b.add(a).add(z), 21);
+  const c = a;
+  a = a.add(x);
+  a = a.add(y);
+  b = b.add(rotate64(a, 44));
+  return [a.add(z), b.add(c)];
+}
+function weakHashLen32WithSeedsStr(s, offset, a, b) {
+  return weakHashLen32WithSeeds(fetch64(s, offset), fetch64(s, offset + 8), fetch64(s, offset + 16), fetch64(s, offset + 24), a, b);
+}
+function hashLen0to16(s, len = s.length) {
+  if (len >= 8) {
+    const mul2 = k2.add(len * 2);
+    const a = fetch64(s, 0).add(k2);
+    const b = fetch64(s, len - 8);
+    const c = rotate64(b, 37).mul(mul2).add(a);
+    const d = rotate64(a, 25).add(b).mul(mul2);
+    return hashLen16(c, d, mul2);
+  }
+  if (len >= 4) {
+    const mul2 = k2.add(len * 2);
+    const a = fetch32(s, 0);
+    return hashLen16(a.shl(3).add(len), fetch32(s, len - 4), mul2);
+  }
+  if (len > 0) {
+    const a = s[0];
+    const b = s[len >> 1];
+    const c = s[len - 1];
+    const y = a + (b << 8);
+    const z = len + (c << 2);
+    return shiftMix(k2.mul(y).xor(k0.mul(z))).mul(k2);
+  }
+  return k2;
+}
+function hashLen17to32(s, len = s.length) {
+  const mul2 = k2.add(len * 2);
+  const a = fetch64(s, 0).mul(k1);
+  const b = fetch64(s, 8);
+  const c = fetch64(s, len - 8).mul(mul2);
+  const d = fetch64(s, len - 16).mul(k2);
+  return hashLen16(rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d), a.add(rotate64(b.add(k2), 18)).add(c), mul2);
+}
+function hashLen33to64(s, len = s.length) {
+  const mul2 = k2.add(len * 2);
+  const a = fetch64(s, 0).mul(k2);
+  const b = fetch64(s, 8);
+  const c = fetch64(s, len - 8).mul(mul2);
+  const d = fetch64(s, len - 16).mul(k2);
+  const y = rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d);
+  const z = hashLen16(y, a.add(rotate64(b.add(k2), 18)).add(c), mul2);
+  const e = fetch64(s, 16).mul(mul2);
+  const f = fetch64(s, 24);
+  const g = y.add(fetch64(s, len - 32)).mul(mul2);
+  const h = z.add(fetch64(s, len - 24)).mul(mul2);
+  return hashLen16(rotate64(e.add(f), 43).add(rotate64(g, 30)).add(h), e.add(rotate64(f.add(a), 18)).add(g), mul2);
+}
+function fingerPrint64(s, len = s.length) {
+  const seed = Long.fromNumber(81, true);
+  if (len <= 32) {
+    if (len <= 16) {
+      return hashLen0to16(s, len);
+    } else {
+      return hashLen17to32(s, len);
+    }
+  } else if (len <= 64) {
+    return hashLen33to64(s, len);
+  }
+  let x = seed;
+  let y = seed.mul(k1).add(113);
+  let z = shiftMix(y.mul(k2).add(113)).mul(k2);
+  let v = [Long.UZERO, Long.UZERO];
+  let w = [Long.UZERO, Long.UZERO];
+  x = x.mul(k2).add(fetch64(s, 0));
+  let offset = 0;
+  const end = (len - 1 >> 6) * 64;
+  const last64 = end + (len - 1 & 63) - 63;
+  do {
+    x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(k1);
+    y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(k1);
+    x = x.xor(w[1]);
+    y = y.add(v[0]).add(fetch64(s, offset + 40));
+    z = rotate64(z.add(w[0]), 33).mul(k1);
+    v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(k1), x.add(w[0]));
+    w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));
+    [z, x] = [x, z];
+    offset += 64;
+  } while (offset !== end);
+  const mul2 = k1.add(z.and(255).shl(1));
+  offset = last64;
+  w[0] = w[0].add(len - 1 & 63);
+  v[0] = v[0].add(w[0]);
+  w[0] = w[0].add(v[0]);
+  x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(mul2);
+  y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(mul2);
+  x = x.xor(w[1].mul(9));
+  y = y.add(v[0].mul(9).add(fetch64(s, offset + 40)));
+  z = rotate64(z.add(w[0]), 33).mul(mul2);
+  v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(mul2), x.add(w[0]));
+  w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));
+  [z, x] = [x, z];
+  return hashLen16(hashLen16(v[0], w[0], mul2).add(shiftMix(y).mul(k0)).add(z), hashLen16(v[1], w[1], mul2).add(x), mul2);
+}
+function createScalarValue(value, dtype) {
+  if (dtype === "string") {
+    return encodeString(value);
+  }
+  return toTypedArray([value], dtype);
+}
+function noConversionNeeded(a, dtype) {
+  return a instanceof Float32Array && dtype === "float32" || a instanceof Int32Array && dtype === "int32" || a instanceof Uint8Array && dtype === "bool";
+}
+function toTypedArray(a, dtype) {
+  if (dtype === "string") {
+    throw new Error("Cannot convert a string[] to a TypedArray");
+  }
+  if (Array.isArray(a)) {
+    a = flatten(a);
+  }
+  if (env().getBool("DEBUG")) {
+    checkConversionForErrors(a, dtype);
+  }
+  if (noConversionNeeded(a, dtype)) {
+    return a;
+  }
+  if (dtype == null || dtype === "float32" || dtype === "complex64") {
+    return new Float32Array(a);
+  } else if (dtype === "int32") {
+    return new Int32Array(a);
+  } else if (dtype === "bool") {
+    const bool = new Uint8Array(a.length);
+    for (let i = 0; i < bool.length; ++i) {
+      if (Math.round(a[i]) !== 0) {
+        bool[i] = 1;
+      }
+    }
+    return bool;
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function now() {
+  return env().platform.now();
+}
+function fetch3(path, requestInits) {
+  return env().platform.fetch(path, requestInits);
+}
+function encodeString(s, encoding = "utf-8") {
+  encoding = encoding || "utf-8";
+  return env().platform.encode(s, encoding);
+}
+function decodeString(bytes, encoding = "utf-8") {
+  encoding = encoding || "utf-8";
+  return env().platform.decode(bytes, encoding);
+}
+var Profiler = class {
+  constructor(backendTimer, logger) {
+    this.backendTimer = backendTimer;
+    this.logger = logger;
+    if (logger == null) {
+      this.logger = new Logger();
+    }
+  }
+  profileKernel(kernelName, inputs, f) {
+    let outputs;
+    const holdResultWrapperFn = () => {
+      outputs = f();
+    };
+    let timer;
+    const start = now();
+    if (this.backendTimer.timerAvailable()) {
+      timer = this.backendTimer.time(holdResultWrapperFn);
+    } else {
+      holdResultWrapperFn();
+      for (const output of outputs) {
+        output.dataSync();
+      }
+      timer = Promise.resolve({ kernelMs: now() - start });
+    }
+    if (env().getBool("CHECK_COMPUTATION_FOR_ERRORS")) {
+      for (let i = 0; i < outputs.length; i++) {
+        const output = outputs[i];
+        output.data().then((tensorVals) => {
+          checkComputationForErrors(tensorVals, output.dtype, kernelName);
+        });
+      }
+    }
+    const kernelProfile = {
+      kernelName,
+      outputs,
+      inputs,
+      timeMs: timer.then((timing) => timing.kernelMs),
+      extraInfo: timer.then((timing) => timing.getExtraProfileInfo != null ? timing.getExtraProfileInfo() : "")
+    };
+    return kernelProfile;
+  }
+  logKernelProfile(kernelProfile) {
+    const { kernelName, outputs, timeMs, inputs, extraInfo } = kernelProfile;
+    outputs.forEach((result) => {
+      Promise.all([result.data(), timeMs, extraInfo]).then((valueContainer) => {
+        this.logger.logKernelProfile(kernelName, result, valueContainer[0], valueContainer[1], inputs, valueContainer[2]);
+      });
+    });
+  }
+};
+function checkComputationForErrors(vals, dtype, kernelName) {
+  if (dtype !== "float32") {
+    return false;
+  }
+  for (let i = 0; i < vals.length; i++) {
+    const num = vals[i];
+    if (isNaN(num) || !isFinite(num)) {
+      console.warn(`Found ${num} in the result of '${kernelName}'`);
+      return true;
+    }
+  }
+  return false;
+}
+var Logger = class {
+  logKernelProfile(name, result, vals, timeMs, inputs, extraInfo) {
+    const time2 = typeof timeMs === "number" ? rightPad(`${timeMs}ms`, 9) : timeMs["error"];
+    const paddedName = rightPad(name, 25);
+    const rank = result.rank;
+    const size = result.size;
+    const shape = rightPad(result.shape.toString(), 14);
+    let inputShapesDescription = "";
+    for (const name2 in inputs) {
+      const input2 = inputs[name2];
+      if (input2 != null) {
+        const inputShape = input2.shape || result.shape;
+        const inputRank = inputShape.length;
+        inputShapesDescription += `${name2}: ${inputRank}D ${inputRank > 0 ? inputShape : ""} `;
+      }
+    }
+    console.log(`%c${paddedName}	%c${time2}	%c${rank}D ${shape}	%c${size}	%c${inputShapesDescription}	%c${extraInfo}`, "font-weight:bold", "color:red", "color:blue", "color: orange", "color: green", "color: steelblue");
+  }
+};
+function getFilteredNodesXToY(tape, xs, y) {
+  const tensorsFromX = {};
+  const nodesFromX = {};
+  for (let i = 0; i < xs.length; i++) {
+    tensorsFromX[xs[i].id] = true;
+  }
+  for (let i = 0; i < tape.length; i++) {
+    const node = tape[i];
+    const nodeInputs = node.inputs;
+    for (const inputName in nodeInputs) {
+      const input2 = nodeInputs[inputName];
+      let anyInputFromX = false;
+      for (let j = 0; j < xs.length; j++) {
+        if (tensorsFromX[input2.id]) {
+          node.outputs.forEach((output) => tensorsFromX[output.id] = true);
+          anyInputFromX = true;
+          nodesFromX[node.id] = true;
+          break;
+        }
+      }
+      if (anyInputFromX) {
+        break;
+      }
+    }
+  }
+  const tensorsLeadToY = {};
+  tensorsLeadToY[y.id] = true;
+  const nodesToY = {};
+  for (let i = tape.length - 1; i >= 0; i--) {
+    const node = tape[i];
+    const nodeInputs = node.inputs;
+    for (let j = 0; j < node.outputs.length; j++) {
+      if (tensorsLeadToY[node.outputs[j].id]) {
+        for (const inputName in nodeInputs) {
+          tensorsLeadToY[nodeInputs[inputName].id] = true;
+          nodesToY[node.id] = true;
+        }
+        break;
+      }
+    }
+  }
+  const filteredTape = [];
+  for (let i = 0; i < tape.length; i++) {
+    const node = tape[i];
+    if (nodesFromX[node.id] && nodesToY[node.id]) {
+      const prunedInputs = {};
+      for (const inputName in node.inputs) {
+        const nodeInput = node.inputs[inputName];
+        if (tensorsFromX[nodeInput.id]) {
+          prunedInputs[inputName] = nodeInput;
+        }
+      }
+      const prunedNode = Object.assign({}, node);
+      prunedNode.inputs = prunedInputs;
+      prunedNode.outputs = node.outputs;
+      filteredTape.push(prunedNode);
+    }
+  }
+  return filteredTape;
+}
+function backpropagateGradients(tensorAccumulatedGradientMap, filteredTape, tidy2, add5) {
+  for (let i = filteredTape.length - 1; i >= 0; i--) {
+    const node = filteredTape[i];
+    const dys = [];
+    node.outputs.forEach((o) => {
+      const gradTensor = tensorAccumulatedGradientMap[o.id];
+      if (gradTensor != null) {
+        dys.push(gradTensor);
+      } else {
+        dys.push(null);
+      }
+    });
+    if (node.gradient == null) {
+      throw new Error(`Cannot compute gradient: gradient function not found for ${node.kernelName}.`);
+    }
+    const inputGradients = node.gradient(dys);
+    for (const inputName in node.inputs) {
+      if (!(inputName in inputGradients)) {
+        throw new Error(`Cannot backprop through input ${inputName}. Available gradients found: ${Object.keys(inputGradients)}.`);
+      }
+      const dx = tidy2(() => inputGradients[inputName]());
+      if (dx.dtype !== "float32") {
+        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input ${inputName} must have 'float32' dtype, but has '${dx.dtype}'`);
+      }
+      const x = node.inputs[inputName];
+      if (!arraysEqual(dx.shape, x.shape)) {
+        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input '${inputName}' has shape '${dx.shape}', which does not match the shape of the input '${x.shape}'`);
+      }
+      if (tensorAccumulatedGradientMap[x.id] == null) {
+        tensorAccumulatedGradientMap[x.id] = dx;
+      } else {
+        const curGradient = tensorAccumulatedGradientMap[x.id];
+        tensorAccumulatedGradientMap[x.id] = add5(curGradient, dx);
+        curGradient.dispose();
+      }
+    }
+  }
+}
+var FORMAT_LIMIT_NUM_VALS = 20;
+var FORMAT_NUM_FIRST_LAST_VALS = 3;
+var FORMAT_NUM_SIG_DIGITS = 7;
+function tensorToString(vals, shape, dtype, verbose) {
+  const strides = computeStrides(shape);
+  const padPerCol = computeMaxSizePerColumn(vals, shape, dtype, strides);
+  const rank = shape.length;
+  const valsLines = subTensorToString(vals, shape, dtype, strides, padPerCol);
+  const lines = ["Tensor"];
+  if (verbose) {
+    lines.push(`  dtype: ${dtype}`);
+    lines.push(`  rank: ${rank}`);
+    lines.push(`  shape: [${shape}]`);
+    lines.push(`  values:`);
+  }
+  lines.push(valsLines.map((l) => "    " + l).join("\n"));
+  return lines.join("\n");
+}
+function computeMaxSizePerColumn(vals, shape, dtype, strides) {
+  const n = sizeFromShape(shape);
+  const numCols = strides[strides.length - 1];
+  const padPerCol = new Array(numCols).fill(0);
+  const rank = shape.length;
+  const valuesOrTuples = dtype === "complex64" ? createComplexTuples(vals) : vals;
+  if (rank > 1) {
+    for (let row = 0; row < n / numCols; row++) {
+      const offset = row * numCols;
+      for (let j = 0; j < numCols; j++) {
+        padPerCol[j] = Math.max(padPerCol[j], valToString(valuesOrTuples[offset + j], 0, dtype).length);
+      }
+    }
+  }
+  return padPerCol;
+}
+function valToString(val, pad3, dtype) {
+  let valStr;
+  if (Array.isArray(val)) {
+    valStr = `${parseFloat(val[0].toFixed(FORMAT_NUM_SIG_DIGITS))} + ${parseFloat(val[1].toFixed(FORMAT_NUM_SIG_DIGITS))}j`;
+  } else if (isString(val)) {
+    valStr = `'${val}'`;
+  } else if (dtype === "bool") {
+    valStr = boolNumToString(val);
+  } else {
+    valStr = parseFloat(val.toFixed(FORMAT_NUM_SIG_DIGITS)).toString();
+  }
+  return rightPad(valStr, pad3);
+}
+function boolNumToString(v) {
+  return v === 0 ? "false" : "true";
+}
+function subTensorToString(vals, shape, dtype, strides, padPerCol, isLast = true) {
+  const storagePerElement = dtype === "complex64" ? 2 : 1;
+  const size = shape[0];
+  const rank = shape.length;
+  if (rank === 0) {
+    if (dtype === "complex64") {
+      const complexTuple = createComplexTuples(vals);
+      return [valToString(complexTuple[0], 0, dtype)];
+    }
+    if (dtype === "bool") {
+      return [boolNumToString(vals[0])];
+    }
+    return [vals[0].toString()];
+  }
+  if (rank === 1) {
+    if (size > FORMAT_LIMIT_NUM_VALS) {
+      const firstValsSize = FORMAT_NUM_FIRST_LAST_VALS * storagePerElement;
+      let firstVals = Array.from(vals.slice(0, firstValsSize));
+      let lastVals = Array.from(vals.slice((size - FORMAT_NUM_FIRST_LAST_VALS) * storagePerElement, size * storagePerElement));
+      if (dtype === "complex64") {
+        firstVals = createComplexTuples(firstVals);
+        lastVals = createComplexTuples(lastVals);
+      }
+      return [
+        "[" + firstVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(", ") + ", ..., " + lastVals.map((x, i) => valToString(x, padPerCol[size - FORMAT_NUM_FIRST_LAST_VALS + i], dtype)).join(", ") + "]"
+      ];
+    }
+    const displayVals = dtype === "complex64" ? createComplexTuples(vals) : Array.from(vals);
+    return [
+      "[" + displayVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(", ") + "]"
+    ];
+  }
+  const subshape = shape.slice(1);
+  const substrides = strides.slice(1);
+  const stride = strides[0] * storagePerElement;
+  const lines = [];
+  if (size > FORMAT_LIMIT_NUM_VALS) {
+    for (let i = 0; i < FORMAT_NUM_FIRST_LAST_VALS; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, false));
+    }
+    lines.push("...");
+    for (let i = size - FORMAT_NUM_FIRST_LAST_VALS; i < size; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));
+    }
+  } else {
+    for (let i = 0; i < size; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));
+    }
+  }
+  const sep = rank === 2 ? "," : "";
+  lines[0] = "[" + lines[0] + sep;
+  for (let i = 1; i < lines.length - 1; i++) {
+    lines[i] = " " + lines[i] + sep;
+  }
+  let newLineSep = ",\n";
+  for (let i = 2; i < rank; i++) {
+    newLineSep += "\n";
+  }
+  lines[lines.length - 1] = " " + lines[lines.length - 1] + "]" + (isLast ? "" : newLineSep);
+  return lines;
+}
+function createComplexTuples(vals) {
+  const complexTuples = [];
+  for (let i = 0; i < vals.length; i += 2) {
+    complexTuples.push([vals[i], vals[i + 1]]);
+  }
+  return complexTuples;
+}
+var TensorBuffer = class {
+  constructor(shape, dtype, values) {
+    this.dtype = dtype;
+    this.shape = shape.slice();
+    this.size = sizeFromShape(shape);
+    if (values != null) {
+      const n = values.length;
+      assert(n === this.size, () => `Length of values '${n}' does not match the size inferred by the shape '${this.size}'.`);
+    }
+    if (dtype === "complex64") {
+      throw new Error(`complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).`);
+    }
+    this.values = values || getArrayFromDType(dtype, this.size);
+    this.strides = computeStrides(shape);
+  }
+  set(value, ...locs) {
+    if (locs.length === 0) {
+      locs = [0];
+    }
+    assert(locs.length === this.rank, () => `The number of provided coordinates (${locs.length}) must match the rank (${this.rank})`);
+    const index = this.locToIndex(locs);
+    this.values[index] = value;
+  }
+  get(...locs) {
+    if (locs.length === 0) {
+      locs = [0];
+    }
+    let i = 0;
+    for (const loc of locs) {
+      if (loc < 0 || loc >= this.shape[i]) {
+        const msg = `Requested out of range element at ${locs}.   Buffer shape=${this.shape}`;
+        throw new Error(msg);
+      }
+      i++;
+    }
+    let index = locs[locs.length - 1];
+    for (let i2 = 0; i2 < locs.length - 1; ++i2) {
+      index += this.strides[i2] * locs[i2];
+    }
+    return this.values[index];
+  }
+  locToIndex(locs) {
+    if (this.rank === 0) {
+      return 0;
+    } else if (this.rank === 1) {
+      return locs[0];
+    }
+    let index = locs[locs.length - 1];
+    for (let i = 0; i < locs.length - 1; ++i) {
+      index += this.strides[i] * locs[i];
+    }
+    return index;
+  }
+  indexToLoc(index) {
+    if (this.rank === 0) {
+      return [];
+    } else if (this.rank === 1) {
+      return [index];
+    }
+    const locs = new Array(this.shape.length);
+    for (let i = 0; i < locs.length - 1; ++i) {
+      locs[i] = Math.floor(index / this.strides[i]);
+      index -= locs[i] * this.strides[i];
+    }
+    locs[locs.length - 1] = index;
+    return locs;
+  }
+  get rank() {
+    return this.shape.length;
+  }
+  toTensor() {
+    return trackerFn().makeTensor(this.values, this.shape, this.dtype);
+  }
+};
+var trackerFn = null;
+var opHandler = null;
+var deprecationWarningFn = null;
+function setTensorTracker(fn) {
+  trackerFn = fn;
+}
+function setOpHandler(handler) {
+  opHandler = handler;
+}
+function setDeprecationWarningFn(fn) {
+  deprecationWarningFn = fn;
+}
+var Tensor = class {
+  constructor(shape, dtype, dataId, id) {
+    this.kept = false;
+    this.isDisposedInternal = false;
+    this.shape = shape.slice();
+    this.dtype = dtype || "float32";
+    this.size = sizeFromShape(shape);
+    this.strides = computeStrides(shape);
+    this.dataId = dataId;
+    this.id = id;
+    this.rankType = this.rank < 5 ? this.rank.toString() : "higher";
+  }
+  get rank() {
+    return this.shape.length;
+  }
+  async buffer() {
+    const vals = await this.data();
+    return opHandler.buffer(this.shape, this.dtype, vals);
+  }
+  bufferSync() {
+    return opHandler.buffer(this.shape, this.dtype, this.dataSync());
+  }
+  async array() {
+    const vals = await this.data();
+    return toNestedArray(this.shape, vals, this.dtype === "complex64");
+  }
+  arraySync() {
+    return toNestedArray(this.shape, this.dataSync(), this.dtype === "complex64");
+  }
+  async data() {
+    this.throwIfDisposed();
+    const data = trackerFn().read(this.dataId);
+    if (this.dtype === "string") {
+      const bytes = await data;
+      try {
+        return bytes.map((b) => decodeString(b));
+      } catch (_a) {
+        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
+      }
+    }
+    return data;
+  }
+  dataToGPU(options) {
+    this.throwIfDisposed();
+    return trackerFn().readToGPU(this.dataId, options);
+  }
+  dataSync() {
+    this.throwIfDisposed();
+    const data = trackerFn().readSync(this.dataId);
+    if (this.dtype === "string") {
+      try {
+        return data.map((b) => decodeString(b));
+      } catch (_a) {
+        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
+      }
+    }
+    return data;
+  }
+  async bytes() {
+    this.throwIfDisposed();
+    const data = await trackerFn().read(this.dataId);
+    if (this.dtype === "string") {
+      return data;
+    } else {
+      return new Uint8Array(data.buffer);
+    }
+  }
+  dispose() {
+    if (this.isDisposed) {
+      return;
+    }
+    trackerFn().disposeTensor(this);
+    this.isDisposedInternal = true;
+  }
+  get isDisposed() {
+    return this.isDisposedInternal;
+  }
+  throwIfDisposed() {
+    if (this.isDisposed) {
+      throw new Error(`Tensor is disposed.`);
+    }
+  }
+  print(verbose = false) {
+    return opHandler.print(this, verbose);
+  }
+  clone() {
+    this.throwIfDisposed();
+    return opHandler.clone(this);
+  }
+  toString(verbose = false) {
+    const vals = this.dataSync();
+    return tensorToString(vals, this.shape, this.dtype, verbose);
+  }
+  cast(dtype) {
+    this.throwIfDisposed();
+    return opHandler.cast(this, dtype);
+  }
+  variable(trainable = true, name, dtype) {
+    this.throwIfDisposed();
+    return trackerFn().makeVariable(this, trainable, name, dtype);
+  }
+};
+Object.defineProperty(Tensor, Symbol.hasInstance, {
+  value: (instance) => {
+    return !!instance && instance.data != null && instance.dataSync != null && instance.throwIfDisposed != null;
+  }
+});
+function getGlobalTensorClass() {
+  return getGlobal("Tensor", () => {
+    return Tensor;
+  });
+}
+getGlobalTensorClass();
+var Variable = class extends Tensor {
+  constructor(initialValue, trainable, name, tensorId) {
+    super(initialValue.shape, initialValue.dtype, initialValue.dataId, tensorId);
+    this.trainable = trainable;
+    this.name = name;
+  }
+  assign(newValue) {
+    if (newValue.dtype !== this.dtype) {
+      throw new Error(`dtype of the new value (${newValue.dtype}) and previous value (${this.dtype}) must match`);
+    }
+    if (!arraysEqual(newValue.shape, this.shape)) {
+      throw new Error(`shape of the new value (${newValue.shape}) and previous value (${this.shape}) must match`);
+    }
+    trackerFn().disposeTensor(this);
+    this.dataId = newValue.dataId;
+    trackerFn().incRef(this, null);
+  }
+  dispose() {
+    trackerFn().disposeVariable(this);
+    this.isDisposedInternal = true;
+  }
+};
+Object.defineProperty(Variable, Symbol.hasInstance, {
+  value: (instance) => {
+    return instance instanceof Tensor && instance.assign != null && instance.assign instanceof Function;
+  }
+});
+var tensor_util_exports = {};
+__export2(tensor_util_exports, {
+  assertTypesMatch: () => assertTypesMatch,
+  getTensorsInContainer: () => getTensorsInContainer,
+  isTensorInList: () => isTensorInList,
+  makeTypesMatch: () => makeTypesMatch
+});
+var Rank;
+(function(Rank2) {
+  Rank2["R0"] = "R0";
+  Rank2["R1"] = "R1";
+  Rank2["R2"] = "R2";
+  Rank2["R3"] = "R3";
+  Rank2["R4"] = "R4";
+  Rank2["R5"] = "R5";
+  Rank2["R6"] = "R6";
+})(Rank || (Rank = {}));
+var UpcastInt32AndMap;
+(function(UpcastInt32AndMap2) {
+  UpcastInt32AndMap2["float32"] = "float32";
+  UpcastInt32AndMap2["int32"] = "int32";
+  UpcastInt32AndMap2["bool"] = "int32";
+  UpcastInt32AndMap2["complex64"] = "complex64";
+})(UpcastInt32AndMap || (UpcastInt32AndMap = {}));
+var UpcastBoolAndMap;
+(function(UpcastBoolAndMap2) {
+  UpcastBoolAndMap2["float32"] = "float32";
+  UpcastBoolAndMap2["int32"] = "int32";
+  UpcastBoolAndMap2["bool"] = "bool";
+  UpcastBoolAndMap2["complex64"] = "complex64";
+})(UpcastBoolAndMap || (UpcastBoolAndMap = {}));
+var UpcastFloat32AndMap;
+(function(UpcastFloat32AndMap2) {
+  UpcastFloat32AndMap2["float32"] = "float32";
+  UpcastFloat32AndMap2["int32"] = "float32";
+  UpcastFloat32AndMap2["bool"] = "float32";
+  UpcastFloat32AndMap2["complex64"] = "complex64";
+})(UpcastFloat32AndMap || (UpcastFloat32AndMap = {}));
+var UpcastComplex64AndMap;
+(function(UpcastComplex64AndMap2) {
+  UpcastComplex64AndMap2["float32"] = "complex64";
+  UpcastComplex64AndMap2["int32"] = "complex64";
+  UpcastComplex64AndMap2["bool"] = "complex64";
+  UpcastComplex64AndMap2["complex64"] = "complex64";
+})(UpcastComplex64AndMap || (UpcastComplex64AndMap = {}));
+var upcastTypeMap = {
+  "float32": UpcastFloat32AndMap,
+  "int32": UpcastInt32AndMap,
+  "bool": UpcastBoolAndMap,
+  "complex64": UpcastComplex64AndMap
+};
+function upcastType(typeA, typeB) {
+  if (typeA === "string" || typeB === "string") {
+    if (typeA === "string" && typeB === "string") {
+      return "string";
+    }
+    throw new Error(`Can not upcast ${typeA} with ${typeB}`);
+  }
+  return upcastTypeMap[typeA][typeB];
+}
+function sumOutType(type) {
+  return upcastType(type, "int32");
+}
+function makeTypesMatch(a, b) {
+  if (a.dtype === b.dtype) {
+    return [a, b];
+  }
+  const dtype = upcastType(a.dtype, b.dtype);
+  return [a.cast(dtype), b.cast(dtype)];
+}
+function assertTypesMatch(a, b) {
+  assert(a.dtype === b.dtype, () => `The dtypes of the first(${a.dtype}) and second(${b.dtype}) input must match`);
+}
+function isTensorInList(tensor2, tensorList) {
+  return tensorList.some((x) => x.id === tensor2.id);
+}
+function getTensorsInContainer(result) {
+  const list = [];
+  const seen = /* @__PURE__ */ new Set();
+  walkTensorContainer(result, list, seen);
+  return list;
+}
+function walkTensorContainer(container, list, seen) {
+  if (container == null) {
+    return;
+  }
+  if (container instanceof Tensor) {
+    list.push(container);
+    return;
+  }
+  if (!isIterable(container)) {
+    return;
+  }
+  const iterable = container;
+  for (const k in iterable) {
+    const val = iterable[k];
+    if (!seen.has(val)) {
+      seen.add(val);
+      walkTensorContainer(val, list, seen);
+    }
+  }
+}
+function isIterable(obj) {
+  return Array.isArray(obj) || typeof obj === "object";
+}
+function isRegisteredKernelInvocation(kernelInvocation) {
+  return kernelInvocation.kernelName != null;
+}
+var EngineState = class {
+  constructor() {
+    this.registeredVariables = {};
+    this.nextTapeNodeId = 0;
+    this.numBytes = 0;
+    this.numTensors = 0;
+    this.numStringTensors = 0;
+    this.numDataBuffers = 0;
+    this.gradientDepth = 0;
+    this.kernelDepth = 0;
+    this.scopeStack = [];
+    this.numDataMovesStack = [];
+    this.nextScopeId = 0;
+    this.tensorInfo = /* @__PURE__ */ new WeakMap();
+    this.profiling = false;
+    this.activeProfile = {
+      newBytes: 0,
+      newTensors: 0,
+      peakBytes: 0,
+      kernels: [],
+      result: null,
+      get kernelNames() {
+        return Array.from(new Set(this.kernels.map((k) => k.name)));
+      }
+    };
+  }
+  dispose() {
+    for (const variableName in this.registeredVariables) {
+      this.registeredVariables[variableName].dispose();
+    }
+  }
+};
+var Engine = class {
+  constructor(ENV7) {
+    this.ENV = ENV7;
+    this.registry = {};
+    this.registryFactory = {};
+    this.pendingBackendInitId = 0;
+    this.state = new EngineState();
+  }
+  async ready() {
+    if (this.pendingBackendInit != null) {
+      return this.pendingBackendInit.then(() => {
+      });
+    }
+    if (this.backendInstance != null) {
+      return;
+    }
+    const sortedBackends = this.getSortedBackends();
+    for (let i = 0; i < sortedBackends.length; i++) {
+      const backendName = sortedBackends[i];
+      const success = await this.initializeBackend(backendName).success;
+      if (success) {
+        await this.setBackend(backendName);
+        return;
+      }
+    }
+    throw new Error(`Could not initialize any backends, all backend initializations failed.`);
+  }
+  get backend() {
+    if (this.pendingBackendInit != null) {
+      throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+    }
+    if (this.backendInstance == null) {
+      const { name, asyncInit } = this.initializeBackendsAndReturnBest();
+      if (asyncInit) {
+        throw new Error(`The highest priority backend '${name}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+      }
+      this.setBackend(name);
+    }
+    return this.backendInstance;
+  }
+  backendNames() {
+    return Object.keys(this.registryFactory);
+  }
+  findBackend(backendName) {
+    if (!(backendName in this.registry)) {
+      if (backendName in this.registryFactory) {
+        const { asyncInit } = this.initializeBackend(backendName);
+        if (asyncInit) {
+          return null;
+        }
+      } else {
+        return null;
+      }
+    }
+    return this.registry[backendName];
+  }
+  findBackendFactory(backendName) {
+    if (!(backendName in this.registryFactory)) {
+      return null;
+    }
+    return this.registryFactory[backendName].factory;
+  }
+  registerBackend(backendName, factory, priority = 1) {
+    if (backendName in this.registryFactory) {
+      warn(`${backendName} backend was already registered. Reusing existing backend factory.`);
+      return false;
+    }
+    this.registryFactory[backendName] = { factory, priority };
+    return true;
+  }
+  async setBackend(backendName) {
+    if (this.registryFactory[backendName] == null) {
+      throw new Error(`Backend name '${backendName}' not found in registry`);
+    }
+    this.backendName = backendName;
+    if (this.registry[backendName] == null) {
+      this.backendInstance = null;
+      const { success, asyncInit } = this.initializeBackend(backendName);
+      const result = asyncInit ? await success : success;
+      if (!result) {
+        return false;
+      }
+    }
+    this.backendInstance = this.registry[backendName];
+    this.setupRegisteredKernels();
+    this.profiler = new Profiler(this.backendInstance);
+    return true;
+  }
+  setupRegisteredKernels() {
+    const kernels = getKernelsForBackend(this.backendName);
+    kernels.forEach((kernel) => {
+      if (kernel.setupFunc != null) {
+        kernel.setupFunc(this.backendInstance);
+      }
+    });
+  }
+  disposeRegisteredKernels(backendName) {
+    const kernels = getKernelsForBackend(backendName);
+    kernels.forEach((kernel) => {
+      if (kernel.disposeFunc != null) {
+        kernel.disposeFunc(this.registry[backendName]);
+      }
+    });
+  }
+  initializeBackend(backendName) {
+    const registryFactoryEntry = this.registryFactory[backendName];
+    if (registryFactoryEntry == null) {
+      throw new Error(`Cannot initialize backend ${backendName}, no registration found.`);
+    }
+    try {
+      const backend2 = registryFactoryEntry.factory();
+      if (backend2 && !(backend2 instanceof KernelBackend) && typeof backend2.then === "function") {
+        const promiseId = ++this.pendingBackendInitId;
+        const success = backend2.then((backendInstance) => {
+          if (promiseId < this.pendingBackendInitId) {
+            return false;
+          }
+          this.registry[backendName] = backendInstance;
+          this.pendingBackendInit = null;
+          return true;
+        }).catch((err) => {
+          if (promiseId < this.pendingBackendInitId) {
+            return false;
+          }
+          this.pendingBackendInit = null;
+          warn(`Initialization of backend ${backendName} failed`);
+          warn(err.stack || err.message);
+          return false;
+        });
+        this.pendingBackendInit = success;
+        return { success, asyncInit: true };
+      } else {
+        this.registry[backendName] = backend2;
+        return { success: true, asyncInit: false };
+      }
+    } catch (err) {
+      warn(`Initialization of backend ${backendName} failed`);
+      warn(err.stack || err.message);
+      return { success: false, asyncInit: false };
+    }
+  }
+  removeBackend(backendName) {
+    if (!(backendName in this.registryFactory)) {
+      throw new Error(`${backendName} backend not found in registry`);
+    }
+    if (this.backendName === backendName && this.pendingBackendInit != null) {
+      this.pendingBackendInitId++;
+    }
+    if (backendName in this.registry) {
+      this.disposeRegisteredKernels(backendName);
+      this.registry[backendName].dispose();
+      delete this.registry[backendName];
+    }
+    delete this.registryFactory[backendName];
+    if (this.backendName === backendName) {
+      this.pendingBackendInit = null;
+      this.backendName = null;
+      this.backendInstance = null;
+    }
+  }
+  getSortedBackends() {
+    if (Object.keys(this.registryFactory).length === 0) {
+      throw new Error("No backend found in registry.");
+    }
+    return Object.keys(this.registryFactory).sort((a, b) => {
+      return this.registryFactory[b].priority - this.registryFactory[a].priority;
+    });
+  }
+  initializeBackendsAndReturnBest() {
+    const sortedBackends = this.getSortedBackends();
+    for (let i = 0; i < sortedBackends.length; i++) {
+      const backendName = sortedBackends[i];
+      const { success, asyncInit } = this.initializeBackend(backendName);
+      if (asyncInit || success) {
+        return { name: backendName, asyncInit };
+      }
+    }
+    throw new Error(`Could not initialize any backends, all backend initializations failed.`);
+  }
+  moveData(backend2, dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    const srcBackend = info.backend;
+    const values = this.readSync(dataId);
+    const refCount = srcBackend.refCount(dataId);
+    srcBackend.disposeData(dataId, true);
+    info.backend = backend2;
+    backend2.move(dataId, values, info.shape, info.dtype, refCount);
+    if (this.shouldCheckForMemLeaks()) {
+      this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1]++;
+    }
+  }
+  tidy(nameOrFn, fn) {
+    let name = null;
+    if (fn == null) {
+      if (typeof nameOrFn !== "function") {
+        throw new Error("Please provide a function to tidy()");
+      }
+      fn = nameOrFn;
+    } else {
+      if (typeof nameOrFn !== "string" && !(nameOrFn instanceof String)) {
+        throw new Error("When calling with two arguments, the first argument to tidy() must be a string");
+      }
+      if (typeof fn !== "function") {
+        throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");
+      }
+      name = nameOrFn;
+    }
+    let result;
+    return this.scopedRun(() => this.startScope(name), () => this.endScope(result), () => {
+      result = fn();
+      if (result instanceof Promise) {
+        console.error("Cannot return a Promise inside of tidy.");
+      }
+      return result;
+    });
+  }
+  scopedRun(start, end, f) {
+    start();
+    try {
+      const res = f();
+      end();
+      return res;
+    } catch (ex) {
+      end();
+      throw ex;
+    }
+  }
+  nextTensorId() {
+    return Engine.nextTensorId++;
+  }
+  nextVariableId() {
+    return Engine.nextVariableId++;
+  }
+  clone(x) {
+    const y = ENGINE.runKernel(Identity, { x });
+    const inputs = { x };
+    const grad2 = (dy) => ({
+      x: () => {
+        const dtype = "float32";
+        const gradInputs = { x: dy };
+        const attrs = { dtype };
+        return ENGINE.runKernel(
+          Cast,
+          gradInputs,
+          attrs
+        );
+      }
+    });
+    const saved = [];
+    this.addTapeNode(this.state.activeScope.name, inputs, [y], grad2, saved, {});
+    return y;
+  }
+  runKernel(kernelName, inputs, attrs) {
+    if (this.backendName == null) {
+      this.backend;
+    }
+    const hasKernel = getKernel(kernelName, this.backendName) != null;
+    if (!hasKernel) {
+      throw new Error(`Kernel '${kernelName}' not registered for backend '${this.backendName}'`);
+    }
+    return this.runKernelFunc({ kernelName, inputs, attrs });
+  }
+  shouldCheckForMemLeaks() {
+    return this.ENV.getBool("IS_TEST");
+  }
+  checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos) {
+    const numDataIdsAfter = this.backend.numDataIds();
+    let numOutputDataIds = 0;
+    outInfos.forEach((info) => {
+      numOutputDataIds += info.dtype === "complex64" ? 3 : 1;
+    });
+    const numMoves = this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1];
+    const dataIdsLeaked = numDataIdsAfter - numDataIdsBefore - numOutputDataIds - numMoves;
+    if (dataIdsLeaked > 0) {
+      throw new Error(`Backend '${this.backendName}' has an internal memory leak (${dataIdsLeaked} data ids) after running '${kernelName}'`);
+    }
+  }
+  runKernelFunc(kernelParams) {
+    let outputs;
+    let saved = [];
+    const isTapeOn = this.isTapeOn();
+    const startingBytecount = this.state.numBytes;
+    const startingNumTensors = this.state.numTensors;
+    if (this.shouldCheckForMemLeaks()) {
+      this.state.numDataMovesStack.push(0);
+    }
+    let kernelFunc3;
+    if (this.backendName == null) {
+      this.backend;
+    }
+    let out;
+    const kernelOrScopeName = isRegisteredKernelInvocation(kernelParams) ? kernelParams.kernelName : this.state.activeScope != null ? this.state.activeScope.name : "";
+    if (isRegisteredKernelInvocation(kernelParams)) {
+      const { kernelName, inputs: inputs2, attrs: attrs2 } = kernelParams;
+      if (this.backendName == null) {
+        this.backend;
+      }
+      const kernel = getKernel(kernelName, this.backendName);
+      assert(kernel != null, () => `Cannot find registered kernel '${kernelName}' for backend '${this.backendName}'`);
+      kernelFunc3 = () => {
+        const numDataIdsBefore = this.backend.numDataIds();
+        out = kernel.kernelFunc({ inputs: inputs2, attrs: attrs2, backend: this.backend });
+        const outInfos = Array.isArray(out) ? out : [out];
+        if (this.shouldCheckForMemLeaks()) {
+          this.checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos);
+        }
+        const outTensors = outInfos.map((outInfo) => {
+          if (outInfo.rank != null) {
+            return outInfo;
+          }
+          return this.makeTensorFromTensorInfo(outInfo);
+        });
+        if (isTapeOn) {
+          const tensorsToSave = this.getTensorsForGradient(kernelName, inputs2, outTensors);
+          saved = this.saveTensorsForBackwardMode(tensorsToSave);
+        }
+        return outTensors;
+      };
+    } else {
+      const { forwardFunc } = kernelParams;
+      const saveFunc = (tensors) => {
+        if (!isTapeOn) {
+          return;
+        }
+        saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));
+      };
+      kernelFunc3 = () => {
+        const numDataIdsBefore = this.backend.numDataIds();
+        out = this.tidy(() => forwardFunc(this.backend, saveFunc));
+        const outs = Array.isArray(out) ? out : [out];
+        if (this.shouldCheckForMemLeaks()) {
+          this.checkKernelForMemLeak(kernelOrScopeName, numDataIdsBefore, outs);
+        }
+        return outs;
+      };
+    }
+    const { inputs, attrs } = kernelParams;
+    const backwardsFunc = isRegisteredKernelInvocation(kernelParams) ? null : kernelParams.backwardsFunc;
+    let kernelProfile;
+    this.scopedRun(
+      () => this.state.kernelDepth++,
+      () => this.state.kernelDepth--,
+      () => {
+        if (!this.ENV.getBool("DEBUG") && !this.state.profiling) {
+          outputs = kernelFunc3();
+        } else {
+          kernelProfile = this.profiler.profileKernel(kernelOrScopeName, inputs, () => kernelFunc3());
+          if (this.ENV.getBool("DEBUG")) {
+            this.profiler.logKernelProfile(kernelProfile);
+          }
+          outputs = kernelProfile.outputs;
+        }
+      }
+    );
+    if (isTapeOn) {
+      this.addTapeNode(kernelOrScopeName, inputs, outputs, backwardsFunc, saved, attrs);
+    }
+    if (this.state.profiling) {
+      this.state.activeProfile.kernels.push({
+        name: kernelOrScopeName,
+        bytesAdded: this.state.numBytes - startingBytecount,
+        totalBytesSnapshot: this.state.numBytes,
+        tensorsAdded: this.state.numTensors - startingNumTensors,
+        totalTensorsSnapshot: this.state.numTensors,
+        inputShapes: Object.keys(inputs).map((key) => inputs[key] != null ? inputs[key].shape : null),
+        outputShapes: outputs.map((item) => item.shape),
+        kernelTimeMs: kernelProfile.timeMs,
+        extraInfo: kernelProfile.extraInfo
+      });
+    }
+    return Array.isArray(out) ? outputs : outputs[0];
+  }
+  saveTensorsForBackwardMode(tensors) {
+    const saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));
+    return saved;
+  }
+  getTensorsForGradient(kernelName, inputs, outputs) {
+    const gradConfig = getGradient(kernelName);
+    if (gradConfig != null) {
+      const inputsToSave = gradConfig.inputsToSave || [];
+      const outputsToSave = gradConfig.outputsToSave || [];
+      let inputTensorsToSave;
+      if (gradConfig.saveAllInputs) {
+        assert(Array.isArray(inputs), () => "saveAllInputs is true, expected inputs to be an array.");
+        inputTensorsToSave = Object.keys(inputs).map((key) => inputs[key]);
+      } else {
+        inputTensorsToSave = inputsToSave.map((inputName) => inputs[inputName]);
+      }
+      const outputTensorsToSave = outputs.filter((_, i) => outputsToSave[i]);
+      return inputTensorsToSave.concat(outputTensorsToSave);
+    }
+    return [];
+  }
+  makeTensor(values, shape, dtype, backend2) {
+    if (values == null) {
+      throw new Error("Values passed to engine.makeTensor() are null");
+    }
+    dtype = dtype || "float32";
+    backend2 = backend2 || this.backend;
+    let backendVals = values;
+    if (dtype === "string" && isString(values[0])) {
+      backendVals = values.map((d) => encodeString(d));
+    }
+    const dataId = backend2.write(backendVals, shape, dtype);
+    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());
+    this.trackTensor(t, backend2);
+    if (dtype === "string") {
+      const info = this.state.tensorInfo.get(dataId);
+      const newBytes = bytesFromStringArray(backendVals);
+      this.state.numBytes += newBytes - info.bytes;
+      info.bytes = newBytes;
+    }
+    return t;
+  }
+  makeTensorFromDataId(dataId, shape, dtype, backend2) {
+    dtype = dtype || "float32";
+    const tensorInfo = { dataId, shape, dtype };
+    return this.makeTensorFromTensorInfo(tensorInfo, backend2);
+  }
+  makeTensorFromTensorInfo(tensorInfo, backend2) {
+    const { dataId, shape, dtype } = tensorInfo;
+    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());
+    this.trackTensor(t, backend2);
+    return t;
+  }
+  makeVariable(initialValue, trainable = true, name, dtype) {
+    name = name || this.nextVariableId().toString();
+    if (dtype != null && dtype !== initialValue.dtype) {
+      initialValue = initialValue.cast(dtype);
+    }
+    const v = new Variable(initialValue, trainable, name, this.nextTensorId());
+    if (this.state.registeredVariables[v.name] != null) {
+      throw new Error(`Variable with name ${v.name} was already registered`);
+    }
+    this.state.registeredVariables[v.name] = v;
+    this.incRef(v, this.backend);
+    return v;
+  }
+  trackTensor(a, backend2) {
+    this.state.numTensors++;
+    if (a.dtype === "string") {
+      this.state.numStringTensors++;
+    }
+    let bytes = 0;
+    if (a.dtype !== "complex64" && a.dtype !== "string") {
+      bytes = a.size * bytesPerElement(a.dtype);
+    }
+    this.state.numBytes += bytes;
+    if (!this.state.tensorInfo.has(a.dataId)) {
+      this.state.numDataBuffers++;
+      this.state.tensorInfo.set(a.dataId, {
+        backend: backend2 || this.backend,
+        dtype: a.dtype,
+        shape: a.shape,
+        bytes
+      });
+    }
+    if (!(a instanceof Variable)) {
+      this.track(a);
+    }
+  }
+  incRef(a, backend2) {
+    this.trackTensor(a, backend2);
+    this.backend.incRef(a.dataId);
+  }
+  removeDataId(dataId, backend2) {
+    if (this.state.tensorInfo.has(dataId) && this.state.tensorInfo.get(dataId).backend === backend2) {
+      this.state.tensorInfo.delete(dataId);
+      this.state.numDataBuffers--;
+    }
+  }
+  disposeTensor(a) {
+    if (!this.state.tensorInfo.has(a.dataId)) {
+      return;
+    }
+    const info = this.state.tensorInfo.get(a.dataId);
+    this.state.numTensors--;
+    if (a.dtype === "string") {
+      this.state.numStringTensors--;
+      this.state.numBytes -= info.bytes;
+    }
+    if (a.dtype !== "complex64" && a.dtype !== "string") {
+      const bytes = a.size * bytesPerElement(a.dtype);
+      this.state.numBytes -= bytes;
+    }
+    if (info.backend.disposeData(a.dataId)) {
+      this.removeDataId(a.dataId, info.backend);
+    }
+  }
+  disposeVariables() {
+    for (const varName in this.state.registeredVariables) {
+      const v = this.state.registeredVariables[varName];
+      this.disposeVariable(v);
+    }
+  }
+  disposeVariable(v) {
+    this.disposeTensor(v);
+    if (this.state.registeredVariables[v.name] != null) {
+      delete this.state.registeredVariables[v.name];
+    }
+  }
+  memory() {
+    const info = this.backend.memory();
+    info.numTensors = this.state.numTensors;
+    info.numDataBuffers = this.state.numDataBuffers;
+    info.numBytes = this.state.numBytes;
+    if (this.state.numStringTensors > 0) {
+      info.unreliable = true;
+      if (info.reasons == null) {
+        info.reasons = [];
+      }
+      info.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)");
+    }
+    return info;
+  }
+  async profile(query) {
+    this.state.profiling = true;
+    const startBytes = this.state.numBytes;
+    const startNumTensors = this.state.numTensors;
+    this.state.activeProfile.kernels = [];
+    this.state.activeProfile.result = await query();
+    this.state.profiling = false;
+    this.state.activeProfile.peakBytes = Math.max(...this.state.activeProfile.kernels.map((d) => d.totalBytesSnapshot));
+    this.state.activeProfile.newBytes = this.state.numBytes - startBytes;
+    this.state.activeProfile.newTensors = this.state.numTensors - startNumTensors;
+    for (const kernel of this.state.activeProfile.kernels) {
+      kernel.kernelTimeMs = await kernel.kernelTimeMs;
+      kernel.extraInfo = await kernel.extraInfo;
+    }
+    return this.state.activeProfile;
+  }
+  isTapeOn() {
+    return this.state.gradientDepth > 0 && this.state.kernelDepth === 0;
+  }
+  addTapeNode(kernelName, inputs, outputs, gradientsFunc, saved, attrs) {
+    const tapeNode = { id: this.state.nextTapeNodeId++, kernelName, inputs, outputs, saved };
+    const gradConfig = getGradient(kernelName);
+    if (gradConfig != null) {
+      gradientsFunc = gradConfig.gradFunc;
+    }
+    if (gradientsFunc != null) {
+      tapeNode.gradient = (dys) => {
+        dys = dys.map((dy, i) => {
+          if (dy == null) {
+            const output = outputs[i];
+            const vals = makeZerosTypedArray(output.size, output.dtype);
+            return this.makeTensor(vals, output.shape, output.dtype);
+          }
+          return dy;
+        });
+        return gradientsFunc(dys.length > 1 ? dys : dys[0], saved, attrs);
+      };
+    }
+    this.state.activeTape.push(tapeNode);
+  }
+  keep(result) {
+    result.kept = true;
+    return result;
+  }
+  startTape() {
+    if (this.state.gradientDepth === 0) {
+      this.state.activeTape = [];
+    }
+    this.state.gradientDepth++;
+  }
+  endTape() {
+    this.state.gradientDepth--;
+  }
+  startScope(name) {
+    const scopeInfo = {
+      track: [],
+      name: "unnamed scope",
+      id: this.state.nextScopeId++
+    };
+    if (name) {
+      scopeInfo.name = name;
+    }
+    this.state.scopeStack.push(scopeInfo);
+    this.state.activeScope = scopeInfo;
+  }
+  endScope(result) {
+    const tensorsToTrackInParent = getTensorsInContainer(result);
+    const tensorsToTrackInParentSet = new Set(tensorsToTrackInParent.map((t) => t.id));
+    for (let i = 0; i < this.state.activeScope.track.length; i++) {
+      const tensor2 = this.state.activeScope.track[i];
+      if (!tensor2.kept && !tensorsToTrackInParentSet.has(tensor2.id)) {
+        tensor2.dispose();
+      }
+    }
+    const oldScope = this.state.scopeStack.pop();
+    this.state.activeScope = this.state.scopeStack.length === 0 ? null : this.state.scopeStack[this.state.scopeStack.length - 1];
+    tensorsToTrackInParent.forEach((tensor2) => {
+      if (!tensor2.kept && tensor2.scopeId === oldScope.id) {
+        this.track(tensor2);
+      }
+    });
+  }
+  gradients(f, xs, dy, allowNoGradients = false) {
+    assert(xs.length > 0, () => "gradients() received an empty list of xs.");
+    if (dy != null && dy.dtype !== "float32") {
+      throw new Error(`dy must have 'float32' dtype, but has '${dy.dtype}'`);
+    }
+    const y = this.scopedRun(() => this.startTape(), () => this.endTape(), () => this.tidy("forward", f));
+    assert(y instanceof Tensor, () => "The result y returned by f() must be a tensor.");
+    const filteredTape = getFilteredNodesXToY(this.state.activeTape, xs, y);
+    if (!allowNoGradients && filteredTape.length === 0 && xs.length > 0) {
+      throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");
+    }
+    return this.tidy("backward", () => {
+      const accumulatedGradientMap = {};
+      accumulatedGradientMap[y.id] = dy == null ? ones(y.shape) : dy;
+      backpropagateGradients(
+        accumulatedGradientMap,
+        filteredTape,
+        (f2) => this.tidy(f2),
+        add
+      );
+      const grads2 = xs.map((x) => accumulatedGradientMap[x.id]);
+      if (this.state.gradientDepth === 0) {
+        this.state.activeTape.forEach((node) => {
+          for (const tensor2 of node.saved) {
+            tensor2.dispose();
+          }
+        });
+        this.state.activeTape = null;
+      }
+      return { value: y, grads: grads2 };
+    });
+  }
+  customGrad(f) {
+    assert(isFunction(f), () => "The f passed in customGrad(f) must be a function.");
+    return (...inputs) => {
+      assert(inputs.every((t) => t instanceof Tensor), () => "The args passed in customGrad(f)(x1, x2,...) must all be tensors");
+      let res;
+      const inputMap = {};
+      inputs.forEach((input2, i) => {
+        inputMap[i] = input2;
+      });
+      const forwardFunc = (_, save) => {
+        res = f(...[...inputs, save]);
+        assert(res.value instanceof Tensor, () => "The function f passed in customGrad(f) must return an object where `obj.value` is a tensor");
+        assert(isFunction(res.gradFunc), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function.");
+        return res.value;
+      };
+      const backwardsFunc = (dy, saved) => {
+        const gradRes = res.gradFunc(dy, saved);
+        const grads2 = Array.isArray(gradRes) ? gradRes : [gradRes];
+        assert(grads2.length === inputs.length, () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...).");
+        assert(grads2.every((t) => t instanceof Tensor), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");
+        const gradMap = {};
+        grads2.forEach((grad2, i) => {
+          gradMap[i] = () => grad2;
+        });
+        return gradMap;
+      };
+      return this.runKernelFunc({
+        forwardFunc,
+        backwardsFunc,
+        inputs: inputMap
+      });
+    };
+  }
+  readSync(dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.readSync(dataId);
+  }
+  read(dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.read(dataId);
+  }
+  readToGPU(dataId, options) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.readToGPU(dataId, options);
+  }
+  async time(query) {
+    const start = now();
+    const timingInfo = await this.backend.time(query);
+    timingInfo.wallMs = now() - start;
+    return timingInfo;
+  }
+  track(result) {
+    if (this.state.activeScope != null) {
+      result.scopeId = this.state.activeScope.id;
+      this.state.activeScope.track.push(result);
+    }
+    return result;
+  }
+  get registeredVariables() {
+    return this.state.registeredVariables;
+  }
+  reset() {
+    this.pendingBackendInitId++;
+    this.state.dispose();
+    this.ENV.reset();
+    this.state = new EngineState();
+    for (const backendName in this.registry) {
+      this.disposeRegisteredKernels(backendName);
+      this.registry[backendName].dispose();
+      delete this.registry[backendName];
+    }
+    this.backendName = null;
+    this.backendInstance = null;
+    this.pendingBackendInit = null;
+  }
+};
+Engine.nextTensorId = 0;
+Engine.nextVariableId = 0;
+function ones(shape) {
+  const values = makeOnesTypedArray(sizeFromShape(shape), "float32");
+  return ENGINE.makeTensor(values, shape, "float32");
+}
+function getOrMakeEngine() {
+  const ns = getGlobalNamespace();
+  if (ns._tfengine == null) {
+    const environment2 = new Environment(ns);
+    ns._tfengine = new Engine(environment2);
+  }
+  setEnvironmentGlobal(ns._tfengine.ENV);
+  setTensorTracker(() => ns._tfengine);
+  return ns._tfengine;
+}
+var ENGINE = getOrMakeEngine();
+function add(a, b) {
+  const inputs = { a, b };
+  return ENGINE.runKernel(Add, inputs);
+}
+var device_util_exports = {};
+__export2(device_util_exports, {
+  isBrowser: () => isBrowser,
+  isMobile: () => isMobile,
+  mockIsMobile: () => mockIsMobile
+});
+function _isNavigatorDefined() {
+  return typeof navigator !== "undefined" && navigator != null;
+}
+var isMobileMockValue;
+function mockIsMobile(value) {
+  isMobileMockValue = value;
+}
+function isMobile(nav) {
+  if (isMobileMockValue !== void 0) {
+    return isMobileMockValue;
+  }
+  if (nav || _isNavigatorDefined()) {
+    if (!nav) {
+      nav = navigator;
+    }
+    if (nav.product === "ReactNative") {
+      return true;
+    }
+    const a = nav.userAgent || nav.vendor || (typeof window !== "undefined" ? window.opera : "");
+    if (!a) {
+      const navAny = nav;
+      return navAny.userAgentData && navAny.userAgentData.mobile;
+    }
+    return /(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(a) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(a.substr(0, 4));
+  }
+  return false;
+}
+function isBrowser() {
+  return typeof window !== "undefined" && window.document != null || typeof WorkerGlobalScope !== "undefined";
+}
+var ENV2 = env();
+ENV2.registerFlag("DEBUG", () => false, (debugValue) => {
+  if (debugValue) {
+    console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.");
+  }
+});
+ENV2.registerFlag("IS_BROWSER", () => isBrowser());
+ENV2.registerFlag("IS_NODE", () => typeof process !== "undefined" && typeof process.versions !== "undefined" && typeof process.versions.node !== "undefined");
+ENV2.registerFlag("IS_CHROME", () => typeof navigator !== "undefined" && navigator != null && navigator.userAgent != null && /Chrome/.test(navigator.userAgent) && /Google Inc/.test(navigator.vendor));
+ENV2.registerFlag("PROD", () => false);
+ENV2.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY", () => ENV2.getBool("DEBUG"));
+ENV2.registerFlag("DEPRECATION_WARNINGS_ENABLED", () => true);
+ENV2.registerFlag("IS_TEST", () => false);
+ENV2.registerFlag("CHECK_COMPUTATION_FOR_ERRORS", () => true);
+ENV2.registerFlag("WRAP_TO_IMAGEBITMAP", () => false);
+ENV2.registerFlag("ENGINE_COMPILE_ONLY", () => false);
+ENV2.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU", () => false);
+ENV2.registerFlag("USE_SETTIMEOUTCUSTOM", () => false);
+function inferShape(val, dtype) {
+  let firstElem = val;
+  if (isTypedArray(val)) {
+    return dtype === "string" ? [] : [val.length];
+  }
+  if (typeof val === "object" && "texture" in val) {
+    const usedChannels = val.channels || "RGBA";
+    return [val.height, val.width * usedChannels.length];
+  }
+  if (!Array.isArray(val)) {
+    return [];
+  }
+  const shape = [];
+  while (Array.isArray(firstElem) || isTypedArray(firstElem) && dtype !== "string") {
+    shape.push(firstElem.length);
+    firstElem = firstElem[0];
+  }
+  if (Array.isArray(val) && env().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")) {
+    deepAssertShapeConsistency(val, shape, []);
+  }
+  return shape;
+}
+function deepAssertShapeConsistency(val, shape, indices) {
+  indices = indices || [];
+  if (!Array.isArray(val) && !isTypedArray(val)) {
+    assert(shape.length === 0, () => `Element arr[${indices.join("][")}] is a primitive, but should be an array/TypedArray of ${shape[0]} elements`);
+    return;
+  }
+  assert(shape.length > 0, () => `Element arr[${indices.join("][")}] should be a primitive, but is an array of ${val.length} elements`);
+  assert(val.length === shape[0], () => `Element arr[${indices.join("][")}] should have ${shape[0]} elements, but has ${val.length} elements`);
+  const subShape = shape.slice(1);
+  for (let i = 0; i < val.length; ++i) {
+    deepAssertShapeConsistency(val[i], subShape, indices.concat(i));
+  }
+}
+function assertDtype(expectedDtype, actualDType, argName, functionName) {
+  if (expectedDtype === "string_or_numeric") {
+    return;
+  }
+  if (expectedDtype == null) {
+    throw new Error(`Expected dtype cannot be null.`);
+  }
+  if (expectedDtype !== "numeric" && expectedDtype !== actualDType || expectedDtype === "numeric" && actualDType === "string") {
+    throw new Error(`Argument '${argName}' passed to '${functionName}' must be ${expectedDtype} tensor, but got ${actualDType} tensor`);
+  }
+}
+function convertToTensor(x, argName, functionName, parseAsDtype = "numeric") {
+  if (x instanceof Tensor) {
+    assertDtype(parseAsDtype, x.dtype, argName, functionName);
+    return x;
+  }
+  let inferredDtype = inferDtype(x);
+  if (inferredDtype !== "string" && ["bool", "int32", "float32"].indexOf(parseAsDtype) >= 0) {
+    inferredDtype = parseAsDtype;
+  }
+  assertDtype(parseAsDtype, inferredDtype, argName, functionName);
+  if (x == null || !isTypedArray(x) && !Array.isArray(x) && typeof x !== "number" && typeof x !== "boolean" && typeof x !== "string") {
+    const type = x == null ? "null" : x.constructor.name;
+    throw new Error(`Argument '${argName}' passed to '${functionName}' must be a Tensor or TensorLike, but got '${type}'`);
+  }
+  const inferredShape = inferShape(x, inferredDtype);
+  if (!isTypedArray(x) && !Array.isArray(x)) {
+    x = [x];
+  }
+  const skipTypedArray = true;
+  const values = inferredDtype !== "string" ? toTypedArray(x, inferredDtype) : flatten(x, [], skipTypedArray);
+  return ENGINE.makeTensor(values, inferredShape, inferredDtype);
+}
+function convertToTensorArray(arg, argName, functionName, parseAsDtype = "numeric") {
+  if (!Array.isArray(arg)) {
+    throw new Error(`Argument ${argName} passed to ${functionName} must be a \`Tensor[]\` or \`TensorLike[]\``);
+  }
+  const tensors = arg;
+  return tensors.map((t, i) => convertToTensor(t, `${argName}[${i}]`, functionName, parseAsDtype));
+}
+var OP_SCOPE_SUFFIX = "__op";
+function op(f) {
+  const keys = Object.keys(f);
+  if (keys.length !== 1) {
+    throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${keys.length} keys.`);
+  }
+  let opName = keys[0];
+  const fn = f[opName];
+  if (opName.endsWith("_")) {
+    opName = opName.substring(0, opName.length - 1);
+  }
+  opName = opName + OP_SCOPE_SUFFIX;
+  const f2 = (...args) => {
+    ENGINE.startScope(opName);
+    try {
+      const result = fn(...args);
+      if (isPromise(result)) {
+        console.error("Cannot return a Promise inside of tidy.");
+      }
+      ENGINE.endScope(result);
+      return result;
+    } catch (ex) {
+      ENGINE.endScope(null);
+      throw ex;
+    }
+  };
+  Object.defineProperty(f2, "name", { value: opName, configurable: true });
+  return f2;
+}
+function complex_(real4, imag4) {
+  const $real = convertToTensor(real4, "real", "complex");
+  const $imag = convertToTensor(imag4, "imag", "complex");
+  assertShapesMatch($real.shape, $imag.shape, `real and imag shapes, ${$real.shape} and ${$imag.shape}, must match in call to tf.complex().`);
+  const inputs = { real: $real, imag: $imag };
+  return ENGINE.runKernel(Complex, inputs);
+}
+var complex = op({ complex_ });
+function makeTensor(values, shape, inferredShape, dtype) {
+  if (dtype == null) {
+    dtype = inferDtype(values);
+  }
+  if (dtype === "complex64") {
+    throw new Error(`Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).`);
+  }
+  if (typeof values === "object" && "texture" in values) {
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${dtype}.`);
+    }
+    values.channels = values.channels || "RGBA";
+    return ENGINE.backend.createTensorFromTexture(values, shape || inferredShape, dtype);
+  }
+  if (!isTypedArray(values) && !Array.isArray(values) && typeof values !== "number" && typeof values !== "boolean" && typeof values !== "string") {
+    throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");
+  }
+  if (shape != null) {
+    assertNonNegativeIntegerDimensions(shape);
+    const providedSize = sizeFromShape(shape);
+    const inferredSize = sizeFromShape(inferredShape);
+    assert(providedSize === inferredSize, () => `Based on the provided shape, [${shape}], the tensor should have ${providedSize} values but has ${inferredSize}`);
+    for (let i = 0; i < inferredShape.length; ++i) {
+      const inferred = inferredShape[i];
+      const flatDimsDontMatch = i === inferredShape.length - 1 ? inferred !== sizeFromShape(shape.slice(i)) : true;
+      assert(inferredShape[i] === shape[i] || !flatDimsDontMatch, () => `Error creating a new Tensor. Inferred shape (${inferredShape}) does not match the provided shape (${shape}). `);
+    }
+  }
+  if (!isTypedArray(values) && !Array.isArray(values)) {
+    values = [values];
+  }
+  shape = shape || inferredShape;
+  values = dtype !== "string" ? toTypedArray(values, dtype) : flatten(values, [], true);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+function tensor(values, shape, dtype) {
+  const inferredShape = inferShape(values, dtype);
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+var DTYPE_VALUE_SIZE_MAP = {
+  "float32": 4,
+  "float16": 2,
+  "int32": 4,
+  "uint16": 2,
+  "uint8": 1,
+  "bool": 1,
+  "complex64": 8
+};
+var NUM_BYTES_STRING_LENGTH = 4;
+async function encodeWeights(tensors, group) {
+  const specs = [];
+  const dataPromises = [];
+  const names = Array.isArray(tensors) ? tensors.map((tensor2) => tensor2.name) : Object.keys(tensors);
+  for (let i = 0; i < names.length; ++i) {
+    const name = names[i];
+    const t = Array.isArray(tensors) ? tensors[i].tensor : tensors[name];
+    if (t.dtype !== "float32" && t.dtype !== "int32" && t.dtype !== "bool" && t.dtype !== "string" && t.dtype !== "complex64") {
+      throw new Error(`Unsupported dtype in weight '${name}': ${t.dtype}`);
+    }
+    const spec = { name, shape: t.shape, dtype: t.dtype };
+    if (t.dtype === "string") {
+      const utf8bytes = new Promise(async (resolve) => {
+        const vals = await t.bytes();
+        const totalNumBytes = vals.reduce((p2, c) => p2 + c.length, 0) + NUM_BYTES_STRING_LENGTH * vals.length;
+        const bytes = new Uint8Array(totalNumBytes);
+        let offset = 0;
+        for (let i2 = 0; i2 < vals.length; i2++) {
+          const val = vals[i2];
+          const bytesOfLength = new Uint8Array(new Uint32Array([val.length]).buffer);
+          bytes.set(bytesOfLength, offset);
+          offset += NUM_BYTES_STRING_LENGTH;
+          bytes.set(val, offset);
+          offset += val.length;
+        }
+        resolve(bytes);
+      });
+      dataPromises.push(utf8bytes);
+    } else {
+      dataPromises.push(t.data());
+    }
+    if (group != null) {
+      spec.group = group;
+    }
+    specs.push(spec);
+  }
+  const tensorValues = await Promise.all(dataPromises);
+  return { data: concatenateTypedArrays(tensorValues), specs };
+}
+function decodeWeights(buffer2, specs) {
+  const out = {};
+  let float16Decode;
+  let offset = 0;
+  for (const spec of specs) {
+    const name = spec.name;
+    const dtype = spec.dtype;
+    const shape = spec.shape;
+    const size = sizeFromShape(shape);
+    let values;
+    if ("quantization" in spec) {
+      const quantization = spec.quantization;
+      if (quantization.dtype === "uint8" || quantization.dtype === "uint16") {
+        if (!("min" in quantization && "scale" in quantization)) {
+          throw new Error(`Weight ${spec.name} with quantization ${quantization.dtype} doesn't have corresponding metadata min and scale.`);
+        }
+      } else if (quantization.dtype === "float16") {
+        if (dtype !== "float32") {
+          throw new Error(`Weight ${spec.name} is quantized with ${quantization.dtype} which only supports weights of type float32 not ${dtype}.`);
+        }
+      } else {
+        throw new Error(`Weight ${spec.name} has unknown quantization dtype ${quantization.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);
+      }
+      const quantizationSizeFactor = DTYPE_VALUE_SIZE_MAP[quantization.dtype];
+      const byteBuffer = buffer2.slice(offset, offset + size * quantizationSizeFactor);
+      const quantizedArray = quantization.dtype === "uint8" ? new Uint8Array(byteBuffer) : new Uint16Array(byteBuffer);
+      if (dtype === "float32") {
+        if (quantization.dtype === "uint8" || quantization.dtype === "uint16") {
+          values = new Float32Array(quantizedArray.length);
+          for (let i = 0; i < quantizedArray.length; i++) {
+            const v = quantizedArray[i];
+            values[i] = v * quantization.scale + quantization.min;
+          }
+        } else if (quantization.dtype === "float16") {
+          if (float16Decode === void 0) {
+            float16Decode = getFloat16Decoder();
+          }
+          values = float16Decode(quantizedArray);
+        } else {
+          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type float32.`);
+        }
+      } else if (dtype === "int32") {
+        if (quantization.dtype !== "uint8" && quantization.dtype !== "uint16") {
+          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type int32.`);
+        }
+        values = new Int32Array(quantizedArray.length);
+        for (let i = 0; i < quantizedArray.length; i++) {
+          const v = quantizedArray[i];
+          values[i] = Math.round(v * quantization.scale + quantization.min);
+        }
+      } else {
+        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);
+      }
+      offset += size * quantizationSizeFactor;
+    } else if (dtype === "string") {
+      const size2 = sizeFromShape(spec.shape);
+      values = [];
+      for (let i = 0; i < size2; i++) {
+        const byteLength = new Uint32Array(buffer2.slice(offset, offset + NUM_BYTES_STRING_LENGTH))[0];
+        offset += NUM_BYTES_STRING_LENGTH;
+        const bytes = new Uint8Array(buffer2.slice(offset, offset + byteLength));
+        values.push(bytes);
+        offset += byteLength;
+      }
+    } else {
+      const dtypeFactor = DTYPE_VALUE_SIZE_MAP[dtype];
+      const byteBuffer = buffer2.slice(offset, offset + size * dtypeFactor);
+      if (dtype === "float32") {
+        values = new Float32Array(byteBuffer);
+      } else if (dtype === "int32") {
+        values = new Int32Array(byteBuffer);
+      } else if (dtype === "bool") {
+        values = new Uint8Array(byteBuffer);
+      } else if (dtype === "complex64") {
+        values = new Float32Array(byteBuffer);
+        const real4 = new Float32Array(values.length / 2);
+        const image2 = new Float32Array(values.length / 2);
+        for (let i = 0; i < real4.length; i++) {
+          real4[i] = values[i * 2];
+          image2[i] = values[i * 2 + 1];
+        }
+        const realTensor = tensor(real4, shape, "float32");
+        const imageTensor = tensor(image2, shape, "float32");
+        out[name] = complex(realTensor, imageTensor);
+        realTensor.dispose();
+        imageTensor.dispose();
+      } else {
+        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);
+      }
+      offset += size * dtypeFactor;
+    }
+    if (dtype !== "complex64") {
+      out[name] = tensor(values, shape, dtype);
+    }
+  }
+  return out;
+}
+function concatenateTypedArrays(xs) {
+  if (xs === null) {
+    throw new Error(`Invalid input value: ${JSON.stringify(xs)}`);
+  }
+  let totalByteLength = 0;
+  const normalizedXs = [];
+  xs.forEach((x) => {
+    totalByteLength += x.byteLength;
+    normalizedXs.push(x.byteLength === x.buffer.byteLength ? x : new x.constructor(x));
+    if (!(x instanceof Float32Array || x instanceof Int32Array || x instanceof Uint8Array)) {
+      throw new Error(`Unsupported TypedArray subtype: ${x.constructor.name}`);
+    }
+  });
+  const y = new Uint8Array(totalByteLength);
+  let offset = 0;
+  normalizedXs.forEach((x) => {
+    y.set(new Uint8Array(x.buffer), offset);
+    offset += x.byteLength;
+  });
+  return y.buffer;
+}
+var useNodeBuffer = typeof Buffer !== "undefined" && (typeof Blob === "undefined" || typeof atob === "undefined" || typeof btoa === "undefined");
+function stringByteLength(str) {
+  if (useNodeBuffer) {
+    return Buffer.byteLength(str);
+  }
+  return new Blob([str]).size;
+}
+function arrayBufferToBase64String(buffer2) {
+  if (useNodeBuffer) {
+    return Buffer.from(buffer2).toString("base64");
+  }
+  const buf = new Uint8Array(buffer2);
+  let s = "";
+  for (let i = 0, l = buf.length; i < l; i++) {
+    s += String.fromCharCode(buf[i]);
+  }
+  return btoa(s);
+}
+function base64StringToArrayBuffer(str) {
+  if (useNodeBuffer) {
+    const buf = Buffer.from(str, "base64");
+    return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);
+  }
+  const s = atob(str);
+  const buffer2 = new Uint8Array(s.length);
+  for (let i = 0; i < s.length; ++i) {
+    buffer2.set([s.charCodeAt(i)], i);
+  }
+  return buffer2.buffer;
+}
+function concatenateArrayBuffers(buffers) {
+  if (buffers.length === 1) {
+    return buffers[0];
+  }
+  let totalByteLength = 0;
+  buffers.forEach((buffer2) => {
+    totalByteLength += buffer2.byteLength;
+  });
+  const temp = new Uint8Array(totalByteLength);
+  let offset = 0;
+  buffers.forEach((buffer2) => {
+    temp.set(new Uint8Array(buffer2), offset);
+    offset += buffer2.byteLength;
+  });
+  return temp.buffer;
+}
+function basename(path) {
+  const SEPARATOR = "/";
+  path = path.trim();
+  while (path.endsWith(SEPARATOR)) {
+    path = path.slice(0, path.length - 1);
+  }
+  const items = path.split(SEPARATOR);
+  return items[items.length - 1];
+}
+function getModelJSONForModelArtifacts(artifacts, manifest) {
+  const result = {
+    modelTopology: artifacts.modelTopology,
+    format: artifacts.format,
+    generatedBy: artifacts.generatedBy,
+    convertedBy: artifacts.convertedBy,
+    weightsManifest: manifest
+  };
+  if (artifacts.signature != null) {
+    result.signature = artifacts.signature;
+  }
+  if (artifacts.userDefinedMetadata != null) {
+    result.userDefinedMetadata = artifacts.userDefinedMetadata;
+  }
+  if (artifacts.modelInitializer != null) {
+    result.modelInitializer = artifacts.modelInitializer;
+  }
+  if (artifacts.initializerSignature != null) {
+    result.initializerSignature = artifacts.initializerSignature;
+  }
+  if (artifacts.trainingConfig != null) {
+    result.trainingConfig = artifacts.trainingConfig;
+  }
+  return result;
+}
+function getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData) {
+  const modelArtifacts = {
+    modelTopology: modelJSON.modelTopology,
+    format: modelJSON.format,
+    generatedBy: modelJSON.generatedBy,
+    convertedBy: modelJSON.convertedBy
+  };
+  if (modelJSON.trainingConfig != null) {
+    modelArtifacts.trainingConfig = modelJSON.trainingConfig;
+  }
+  if (modelJSON.weightsManifest != null) {
+    if (!weightSpecs) {
+      throw new Error("modelJSON has weightsManifest but weightSpecs is null");
+    }
+    if (!weightData) {
+      throw new Error("modelJSON has weightsManifest but weightData is null");
+    }
+    modelArtifacts.weightSpecs = weightSpecs;
+    modelArtifacts.weightData = weightData;
+  }
+  if (modelJSON.signature != null) {
+    modelArtifacts.signature = modelJSON.signature;
+  }
+  if (modelJSON.userDefinedMetadata != null) {
+    modelArtifacts.userDefinedMetadata = modelJSON.userDefinedMetadata;
+  }
+  if (modelJSON.modelInitializer != null) {
+    modelArtifacts.modelInitializer = modelJSON.modelInitializer;
+  }
+  if (modelJSON.initializerSignature != null) {
+    modelArtifacts.initializerSignature = modelJSON.initializerSignature;
+  }
+  return modelArtifacts;
+}
+async function getModelArtifactsForJSON(modelJSON, loadWeights2) {
+  let weightSpecs;
+  let weightData;
+  if (modelJSON.weightsManifest != null) {
+    [weightSpecs, weightData] = await loadWeights2(modelJSON.weightsManifest);
+  }
+  return getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData);
+}
+function getModelArtifactsInfoForJSON(modelArtifacts) {
+  if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+    throw new Error("Expected JSON model topology, received ArrayBuffer.");
+  }
+  return {
+    dateSaved: new Date(),
+    modelTopologyType: "JSON",
+    modelTopologyBytes: modelArtifacts.modelTopology == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.modelTopology)),
+    weightSpecsBytes: modelArtifacts.weightSpecs == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.weightSpecs)),
+    weightDataBytes: modelArtifacts.weightData == null ? 0 : modelArtifacts.weightData.byteLength
+  };
+}
+function getWeightSpecs(weightsManifest) {
+  const weightSpecs = [];
+  for (const entry of weightsManifest) {
+    weightSpecs.push(...entry.weights);
+  }
+  return weightSpecs;
+}
+function computeFloat16MantisaTable() {
+  const convertMantissa = (i) => {
+    let m = i << 13;
+    let e = 0;
+    while ((m & 8388608) === 0) {
+      e -= 8388608;
+      m <<= 1;
+    }
+    m &= ~8388608;
+    e += 947912704;
+    return m | e;
+  };
+  const mantisaTable = new Uint32Array(2048);
+  mantisaTable[0] = 0;
+  for (let i = 1; i < 1024; i++) {
+    mantisaTable[i] = convertMantissa(i);
+  }
+  for (let i = 1024; i < 2048; i++) {
+    mantisaTable[i] = 939524096 + (i - 1024 << 13);
+  }
+  return mantisaTable;
+}
+function computeFloat16ExponentTable() {
+  const exponentTable = new Uint32Array(64);
+  exponentTable[0] = 0;
+  exponentTable[31] = 1199570944;
+  exponentTable[32] = 2147483648;
+  exponentTable[63] = 3347054592;
+  for (let i = 1; i < 31; i++) {
+    exponentTable[i] = i << 23;
+  }
+  for (let i = 33; i < 63; i++) {
+    exponentTable[i] = 2147483648 + (i - 32 << 23);
+  }
+  return exponentTable;
+}
+function computeFloat16OffsetTable() {
+  const offsetTable = new Uint32Array(64);
+  for (let i = 0; i < 64; i++) {
+    offsetTable[i] = 1024;
+  }
+  offsetTable[0] = offsetTable[32] = 0;
+  return offsetTable;
+}
+function getFloat16Decoder() {
+  const mantisaTable = computeFloat16MantisaTable();
+  const exponentTable = computeFloat16ExponentTable();
+  const offsetTable = computeFloat16OffsetTable();
+  return (quantizedArray) => {
+    const buffer2 = new ArrayBuffer(4 * quantizedArray.length);
+    const bufferUint32View = new Uint32Array(buffer2);
+    for (let index = 0; index < quantizedArray.length; index++) {
+      const float16Bits = quantizedArray[index];
+      const float32Bits = mantisaTable[offsetTable[float16Bits >> 10] + (float16Bits & 1023)] + exponentTable[float16Bits >> 10];
+      bufferUint32View[index] = float32Bits;
+    }
+    return new Float32Array(buffer2);
+  };
+}
+var IORouterRegistry = class {
+  constructor() {
+    this.saveRouters = [];
+    this.loadRouters = [];
+  }
+  static getInstance() {
+    if (IORouterRegistry.instance == null) {
+      IORouterRegistry.instance = new IORouterRegistry();
+    }
+    return IORouterRegistry.instance;
+  }
+  static registerSaveRouter(saveRouter) {
+    IORouterRegistry.getInstance().saveRouters.push(saveRouter);
+  }
+  static registerLoadRouter(loadRouter) {
+    IORouterRegistry.getInstance().loadRouters.push(loadRouter);
+  }
+  static getSaveHandlers(url) {
+    return IORouterRegistry.getHandlers(url, "save");
+  }
+  static getLoadHandlers(url, loadOptions) {
+    return IORouterRegistry.getHandlers(url, "load", loadOptions);
+  }
+  static getHandlers(url, handlerType, loadOptions) {
+    const validHandlers = [];
+    const routers = handlerType === "load" ? IORouterRegistry.getInstance().loadRouters : IORouterRegistry.getInstance().saveRouters;
+    routers.forEach((router) => {
+      const handler = router(url, loadOptions);
+      if (handler !== null) {
+        validHandlers.push(handler);
+      }
+    });
+    return validHandlers;
+  }
+};
+var registerSaveRouter = (loudRouter) => IORouterRegistry.registerSaveRouter(loudRouter);
+var registerLoadRouter = (loudRouter) => IORouterRegistry.registerLoadRouter(loudRouter);
+var getSaveHandlers = (url) => IORouterRegistry.getSaveHandlers(url);
+var getLoadHandlers = (url, loadOptions) => IORouterRegistry.getLoadHandlers(url, loadOptions);
+var DATABASE_NAME = "tensorflowjs";
+var DATABASE_VERSION = 1;
+var MODEL_STORE_NAME = "models_store";
+var INFO_STORE_NAME = "model_info_store";
+function getIndexedDBFactory() {
+  if (!env().getBool("IS_BROWSER")) {
+    throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");
+  }
+  const theWindow = typeof window === "undefined" ? self : window;
+  const factory = theWindow.indexedDB || theWindow.mozIndexedDB || theWindow.webkitIndexedDB || theWindow.msIndexedDB || theWindow.shimIndexedDB;
+  if (factory == null) {
+    throw new Error("The current browser does not appear to support IndexedDB.");
+  }
+  return factory;
+}
+function setUpDatabase(openRequest) {
+  const db = openRequest.result;
+  db.createObjectStore(MODEL_STORE_NAME, { keyPath: "modelPath" });
+  db.createObjectStore(INFO_STORE_NAME, { keyPath: "modelPath" });
+}
+var BrowserIndexedDB = class {
+  constructor(modelPath) {
+    this.indexedDB = getIndexedDBFactory();
+    if (modelPath == null || !modelPath) {
+      throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");
+    }
+    this.modelPath = modelPath;
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");
+    }
+    return this.databaseAction(this.modelPath, modelArtifacts);
+  }
+  async load() {
+    return this.databaseAction(this.modelPath);
+  }
+  databaseAction(modelPath, modelArtifacts) {
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        if (modelArtifacts == null) {
+          const modelTx = db.transaction(MODEL_STORE_NAME, "readonly");
+          const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+          const getRequest = modelStore.get(this.modelPath);
+          getRequest.onsuccess = () => {
+            if (getRequest.result == null) {
+              db.close();
+              return reject(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));
+            } else {
+              resolve(getRequest.result.modelArtifacts);
+            }
+          };
+          getRequest.onerror = (error) => {
+            db.close();
+            return reject(getRequest.error);
+          };
+          modelTx.oncomplete = () => db.close();
+        } else {
+          const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);
+          const infoTx = db.transaction(INFO_STORE_NAME, "readwrite");
+          let infoStore = infoTx.objectStore(INFO_STORE_NAME);
+          const putInfoRequest = infoStore.put({ modelPath: this.modelPath, modelArtifactsInfo });
+          let modelTx;
+          putInfoRequest.onsuccess = () => {
+            modelTx = db.transaction(MODEL_STORE_NAME, "readwrite");
+            const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+            const putModelRequest = modelStore.put({
+              modelPath: this.modelPath,
+              modelArtifacts,
+              modelArtifactsInfo
+            });
+            putModelRequest.onsuccess = () => resolve({ modelArtifactsInfo });
+            putModelRequest.onerror = (error) => {
+              infoStore = infoTx.objectStore(INFO_STORE_NAME);
+              const deleteInfoRequest = infoStore.delete(this.modelPath);
+              deleteInfoRequest.onsuccess = () => {
+                db.close();
+                return reject(putModelRequest.error);
+              };
+              deleteInfoRequest.onerror = (error2) => {
+                db.close();
+                return reject(putModelRequest.error);
+              };
+            };
+          };
+          putInfoRequest.onerror = (error) => {
+            db.close();
+            return reject(putInfoRequest.error);
+          };
+          infoTx.oncomplete = () => {
+            if (modelTx == null) {
+              db.close();
+            } else {
+              modelTx.oncomplete = () => db.close();
+            }
+          };
+        }
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+};
+BrowserIndexedDB.URL_SCHEME = "indexeddb://";
+var indexedDBRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserIndexedDB.URL_SCHEME)) {
+      return browserIndexedDB(url.slice(BrowserIndexedDB.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(indexedDBRouter);
+IORouterRegistry.registerLoadRouter(indexedDBRouter);
+function browserIndexedDB(modelPath) {
+  return new BrowserIndexedDB(modelPath);
+}
+function maybeStripScheme(key) {
+  return key.startsWith(BrowserIndexedDB.URL_SCHEME) ? key.slice(BrowserIndexedDB.URL_SCHEME.length) : key;
+}
+var BrowserIndexedDBManager = class {
+  constructor() {
+    this.indexedDB = getIndexedDBFactory();
+  }
+  async listModels() {
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        const tx = db.transaction(INFO_STORE_NAME, "readonly");
+        const store = tx.objectStore(INFO_STORE_NAME);
+        const getAllInfoRequest = store.getAll();
+        getAllInfoRequest.onsuccess = () => {
+          const out = {};
+          for (const item of getAllInfoRequest.result) {
+            out[item.modelPath] = item.modelArtifactsInfo;
+          }
+          resolve(out);
+        };
+        getAllInfoRequest.onerror = (error) => {
+          db.close();
+          return reject(getAllInfoRequest.error);
+        };
+        tx.oncomplete = () => db.close();
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+  async removeModel(path) {
+    path = maybeStripScheme(path);
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        const infoTx = db.transaction(INFO_STORE_NAME, "readwrite");
+        const infoStore = infoTx.objectStore(INFO_STORE_NAME);
+        const getInfoRequest = infoStore.get(path);
+        let modelTx;
+        getInfoRequest.onsuccess = () => {
+          if (getInfoRequest.result == null) {
+            db.close();
+            return reject(new Error(`Cannot find model with path '${path}' in IndexedDB.`));
+          } else {
+            const deleteInfoRequest = infoStore.delete(path);
+            const deleteModelData = () => {
+              modelTx = db.transaction(MODEL_STORE_NAME, "readwrite");
+              const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+              const deleteModelRequest = modelStore.delete(path);
+              deleteModelRequest.onsuccess = () => resolve(getInfoRequest.result.modelArtifactsInfo);
+              deleteModelRequest.onerror = (error) => reject(getInfoRequest.error);
+            };
+            deleteInfoRequest.onsuccess = deleteModelData;
+            deleteInfoRequest.onerror = (error) => {
+              deleteModelData();
+              db.close();
+              return reject(getInfoRequest.error);
+            };
+          }
+        };
+        getInfoRequest.onerror = (error) => {
+          db.close();
+          return reject(getInfoRequest.error);
+        };
+        infoTx.oncomplete = () => {
+          if (modelTx == null) {
+            db.close();
+          } else {
+            modelTx.oncomplete = () => db.close();
+          }
+        };
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+};
+var PATH_SEPARATOR = "/";
+var PATH_PREFIX = "tensorflowjs_models";
+var INFO_SUFFIX = "info";
+var MODEL_TOPOLOGY_SUFFIX = "model_topology";
+var WEIGHT_SPECS_SUFFIX = "weight_specs";
+var WEIGHT_DATA_SUFFIX = "weight_data";
+var MODEL_METADATA_SUFFIX = "model_metadata";
+function getModelKeys(path) {
+  return {
+    info: [PATH_PREFIX, path, INFO_SUFFIX].join(PATH_SEPARATOR),
+    topology: [PATH_PREFIX, path, MODEL_TOPOLOGY_SUFFIX].join(PATH_SEPARATOR),
+    weightSpecs: [PATH_PREFIX, path, WEIGHT_SPECS_SUFFIX].join(PATH_SEPARATOR),
+    weightData: [PATH_PREFIX, path, WEIGHT_DATA_SUFFIX].join(PATH_SEPARATOR),
+    modelMetadata: [PATH_PREFIX, path, MODEL_METADATA_SUFFIX].join(PATH_SEPARATOR)
+  };
+}
+function removeItems(keys) {
+  for (const key of Object.values(keys)) {
+    window.localStorage.removeItem(key);
+  }
+}
+function getModelPathFromKey(key) {
+  const items = key.split(PATH_SEPARATOR);
+  if (items.length < 3) {
+    throw new Error(`Invalid key format: ${key}`);
+  }
+  return items.slice(1, items.length - 1).join(PATH_SEPARATOR);
+}
+function maybeStripScheme2(key) {
+  return key.startsWith(BrowserLocalStorage.URL_SCHEME) ? key.slice(BrowserLocalStorage.URL_SCHEME.length) : key;
+}
+var BrowserLocalStorage = class {
+  constructor(modelPath) {
+    if (!env().getBool("IS_BROWSER") || typeof window === "undefined" || typeof window.localStorage === "undefined") {
+      throw new Error("The current environment does not support local storage.");
+    }
+    this.LS = window.localStorage;
+    if (modelPath == null || !modelPath) {
+      throw new Error("For local storage, modelPath must not be null, undefined or empty.");
+    }
+    this.modelPath = modelPath;
+    this.keys = getModelKeys(this.modelPath);
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");
+    } else {
+      const topology = JSON.stringify(modelArtifacts.modelTopology);
+      const weightSpecs = JSON.stringify(modelArtifacts.weightSpecs);
+      const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);
+      try {
+        this.LS.setItem(this.keys.info, JSON.stringify(modelArtifactsInfo));
+        this.LS.setItem(this.keys.topology, topology);
+        this.LS.setItem(this.keys.weightSpecs, weightSpecs);
+        this.LS.setItem(this.keys.weightData, arrayBufferToBase64String(modelArtifacts.weightData));
+        const metadata = {
+          format: modelArtifacts.format,
+          generatedBy: modelArtifacts.generatedBy,
+          convertedBy: modelArtifacts.convertedBy,
+          signature: modelArtifacts.signature != null ? modelArtifacts.signature : void 0,
+          userDefinedMetadata: modelArtifacts.userDefinedMetadata != null ? modelArtifacts.userDefinedMetadata : void 0,
+          modelInitializer: modelArtifacts.modelInitializer != null ? modelArtifacts.modelInitializer : void 0,
+          initializerSignature: modelArtifacts.initializerSignature != null ? modelArtifacts.initializerSignature : void 0,
+          trainingConfig: modelArtifacts.trainingConfig != null ? modelArtifacts.trainingConfig : void 0
+        };
+        this.LS.setItem(this.keys.modelMetadata, JSON.stringify(metadata));
+        return { modelArtifactsInfo };
+      } catch (err) {
+        removeItems(this.keys);
+        throw new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${modelArtifactsInfo.modelTopologyBytes}, weightSpecsBytes=${modelArtifactsInfo.weightSpecsBytes}, weightDataBytes=${modelArtifactsInfo.weightDataBytes}.`);
+      }
+    }
+  }
+  async load() {
+    const info = JSON.parse(this.LS.getItem(this.keys.info));
+    if (info == null) {
+      throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);
+    }
+    if (info.modelTopologyType !== "JSON") {
+      throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");
+    }
+    const out = {};
+    const topology = JSON.parse(this.LS.getItem(this.keys.topology));
+    if (topology == null) {
+      throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);
+    }
+    out.modelTopology = topology;
+    const weightSpecs = JSON.parse(this.LS.getItem(this.keys.weightSpecs));
+    if (weightSpecs == null) {
+      throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);
+    }
+    out.weightSpecs = weightSpecs;
+    const metadataString = this.LS.getItem(this.keys.modelMetadata);
+    if (metadataString != null) {
+      const metadata = JSON.parse(metadataString);
+      out.format = metadata.format;
+      out.generatedBy = metadata.generatedBy;
+      out.convertedBy = metadata.convertedBy;
+      if (metadata.signature != null) {
+        out.signature = metadata.signature;
+      }
+      if (metadata.userDefinedMetadata != null) {
+        out.userDefinedMetadata = metadata.userDefinedMetadata;
+      }
+      if (metadata.modelInitializer != null) {
+        out.modelInitializer = metadata.modelInitializer;
+      }
+      if (metadata.initializerSignature != null) {
+        out.initializerSignature = metadata.initializerSignature;
+      }
+      if (metadata.trainingConfig != null) {
+        out.trainingConfig = metadata.trainingConfig;
+      }
+    }
+    const weightDataBase64 = this.LS.getItem(this.keys.weightData);
+    if (weightDataBase64 == null) {
+      throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);
+    }
+    out.weightData = base64StringToArrayBuffer(weightDataBase64);
+    return out;
+  }
+};
+BrowserLocalStorage.URL_SCHEME = "localstorage://";
+var localStorageRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserLocalStorage.URL_SCHEME)) {
+      return browserLocalStorage(url.slice(BrowserLocalStorage.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(localStorageRouter);
+IORouterRegistry.registerLoadRouter(localStorageRouter);
+function browserLocalStorage(modelPath) {
+  return new BrowserLocalStorage(modelPath);
+}
+var BrowserLocalStorageManager = class {
+  constructor() {
+    assert(env().getBool("IS_BROWSER"), () => "Current environment is not a web browser");
+    assert(typeof window === "undefined" || typeof window.localStorage !== "undefined", () => "Current browser does not appear to support localStorage");
+    this.LS = window.localStorage;
+  }
+  async listModels() {
+    const out = {};
+    const prefix = PATH_PREFIX + PATH_SEPARATOR;
+    const suffix = PATH_SEPARATOR + INFO_SUFFIX;
+    for (let i = 0; i < this.LS.length; ++i) {
+      const key = this.LS.key(i);
+      if (key.startsWith(prefix) && key.endsWith(suffix)) {
+        const modelPath = getModelPathFromKey(key);
+        out[modelPath] = JSON.parse(this.LS.getItem(key));
+      }
+    }
+    return out;
+  }
+  async removeModel(path) {
+    path = maybeStripScheme2(path);
+    const keys = getModelKeys(path);
+    if (this.LS.getItem(keys.info) == null) {
+      throw new Error(`Cannot find model at path '${path}'`);
+    }
+    const info = JSON.parse(this.LS.getItem(keys.info));
+    removeItems(keys);
+    return info;
+  }
+};
+var URL_SCHEME_SUFFIX = "://";
+var ModelStoreManagerRegistry = class {
+  constructor() {
+    this.managers = {};
+  }
+  static getInstance() {
+    if (ModelStoreManagerRegistry.instance == null) {
+      ModelStoreManagerRegistry.instance = new ModelStoreManagerRegistry();
+    }
+    return ModelStoreManagerRegistry.instance;
+  }
+  static registerManager(scheme, manager) {
+    assert(scheme != null, () => "scheme must not be undefined or null.");
+    if (scheme.endsWith(URL_SCHEME_SUFFIX)) {
+      scheme = scheme.slice(0, scheme.indexOf(URL_SCHEME_SUFFIX));
+    }
+    assert(scheme.length > 0, () => "scheme must not be an empty string.");
+    const registry = ModelStoreManagerRegistry.getInstance();
+    assert(registry.managers[scheme] == null, () => `A model store manager is already registered for scheme '${scheme}'.`);
+    registry.managers[scheme] = manager;
+  }
+  static getManager(scheme) {
+    const manager = ModelStoreManagerRegistry.getInstance().managers[scheme];
+    if (manager == null) {
+      throw new Error(`Cannot find model manager for scheme '${scheme}'`);
+    }
+    return manager;
+  }
+  static getSchemes() {
+    return Object.keys(ModelStoreManagerRegistry.getInstance().managers);
+  }
+};
+function parseURL(url) {
+  if (url.indexOf(URL_SCHEME_SUFFIX) === -1) {
+    throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${ModelStoreManagerRegistry.getSchemes().join(",")}`);
+  }
+  return {
+    scheme: url.split(URL_SCHEME_SUFFIX)[0],
+    path: url.split(URL_SCHEME_SUFFIX)[1]
+  };
+}
+async function cloneModelInternal(sourceURL, destURL, deleteSource = false) {
+  assert(sourceURL !== destURL, () => `Old path and new path are the same: '${sourceURL}'`);
+  const loadHandlers = IORouterRegistry.getLoadHandlers(sourceURL);
+  assert(loadHandlers.length > 0, () => `Copying failed because no load handler is found for source URL ${sourceURL}.`);
+  assert(loadHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) load handlers for source URL ${sourceURL}.`);
+  const loadHandler = loadHandlers[0];
+  const saveHandlers = IORouterRegistry.getSaveHandlers(destURL);
+  assert(saveHandlers.length > 0, () => `Copying failed because no save handler is found for destination URL ${destURL}.`);
+  assert(saveHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) save handlers for destination URL ${destURL}.`);
+  const saveHandler = saveHandlers[0];
+  const sourceScheme = parseURL(sourceURL).scheme;
+  const sourcePath = parseURL(sourceURL).path;
+  const sameMedium = sourceScheme === parseURL(sourceURL).scheme;
+  const modelArtifacts = await loadHandler.load();
+  if (deleteSource && sameMedium) {
+    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);
+  }
+  const saveResult = await saveHandler.save(modelArtifacts);
+  if (deleteSource && !sameMedium) {
+    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);
+  }
+  return saveResult.modelArtifactsInfo;
+}
+async function listModels() {
+  const schemes = ModelStoreManagerRegistry.getSchemes();
+  const out = {};
+  for (const scheme of schemes) {
+    const schemeOut = await ModelStoreManagerRegistry.getManager(scheme).listModels();
+    for (const path in schemeOut) {
+      const url = scheme + URL_SCHEME_SUFFIX + path;
+      out[url] = schemeOut[path];
+    }
+  }
+  return out;
+}
+async function removeModel(url) {
+  const schemeAndPath = parseURL(url);
+  const manager = ModelStoreManagerRegistry.getManager(schemeAndPath.scheme);
+  return manager.removeModel(schemeAndPath.path);
+}
+async function copyModel(sourceURL, destURL) {
+  const deleteSource = false;
+  return cloneModelInternal(sourceURL, destURL, deleteSource);
+}
+async function moveModel(sourceURL, destURL) {
+  const deleteSource = true;
+  return cloneModelInternal(sourceURL, destURL, deleteSource);
+}
+var PlatformBrowser = class {
+  constructor() {
+    this.messageName = "setTimeoutCustom";
+    this.functionRefs = [];
+    this.handledMessageCount = 0;
+    this.hasEventListener = false;
+  }
+  fetch(path, init2) {
+    return fetch(path, init2);
+  }
+  now() {
+    return performance.now();
+  }
+  encode(text, encoding) {
+    if (encoding !== "utf-8" && encoding !== "utf8") {
+      throw new Error(`Browser's encoder only supports utf-8, but got ${encoding}`);
+    }
+    if (this.textEncoder == null) {
+      this.textEncoder = new TextEncoder();
+    }
+    return this.textEncoder.encode(text);
+  }
+  decode(bytes, encoding) {
+    return new TextDecoder(encoding).decode(bytes);
+  }
+  setTimeoutCustom(functionRef, delay) {
+    if (typeof window === "undefined" || !env().getBool("USE_SETTIMEOUTCUSTOM")) {
+      setTimeout(functionRef, delay);
+      return;
+    }
+    this.functionRefs.push(functionRef);
+    setTimeout(() => {
+      window.postMessage({ name: this.messageName, index: this.functionRefs.length - 1 }, "*");
+    }, delay);
+    if (!this.hasEventListener) {
+      this.hasEventListener = true;
+      window.addEventListener("message", (event) => {
+        if (event.source === window && event.data.name === this.messageName) {
+          event.stopPropagation();
+          const functionRef2 = this.functionRefs[event.data.index];
+          functionRef2();
+          this.handledMessageCount++;
+          if (this.handledMessageCount === this.functionRefs.length) {
+            this.functionRefs = [];
+            this.handledMessageCount = 0;
+          }
+        }
+      }, true);
+    }
+  }
+};
+if (env().get("IS_BROWSER")) {
+  env().setPlatform("browser", new PlatformBrowser());
+  try {
+    ModelStoreManagerRegistry.registerManager(BrowserLocalStorage.URL_SCHEME, new BrowserLocalStorageManager());
+  } catch (err) {
+  }
+  try {
+    ModelStoreManagerRegistry.registerManager(BrowserIndexedDB.URL_SCHEME, new BrowserIndexedDBManager());
+  } catch (err) {
+  }
+}
+var getNodeFetch = {
+  importFetch: () => require_browser()
+};
+var systemFetch;
+var PlatformNode = class {
+  constructor() {
+    this.util = require_util();
+    this.textEncoder = new this.util.TextEncoder();
+  }
+  fetch(path, requestInits) {
+    if (env().global.fetch != null) {
+      return env().global.fetch(path, requestInits);
+    }
+    if (systemFetch == null) {
+      systemFetch = getNodeFetch.importFetch();
+    }
+    return systemFetch(path, requestInits);
+  }
+  now() {
+    const time2 = process.hrtime();
+    return time2[0] * 1e3 + time2[1] / 1e6;
+  }
+  encode(text, encoding) {
+    if (encoding !== "utf-8" && encoding !== "utf8") {
+      throw new Error(`Node built-in encoder only supports utf-8, but got ${encoding}`);
+    }
+    return this.textEncoder.encode(text);
+  }
+  decode(bytes, encoding) {
+    if (bytes.length === 0) {
+      return "";
+    }
+    return new this.util.TextDecoder(encoding).decode(bytes);
+  }
+};
+if (env().get("IS_NODE") && !env().get("IS_BROWSER")) {
+  env().setPlatform("node", new PlatformNode());
+}
+function buffer(shape, dtype = "float32", values) {
+  dtype = dtype || "float32";
+  assertNonNegativeIntegerDimensions(shape);
+  return new TensorBuffer(shape, dtype, values);
+}
+function cast_(x, dtype) {
+  const $x = convertToTensor(x, "x", "cast");
+  if (!isValidDtype(dtype)) {
+    throw new Error(`Failed to cast to unknown dtype ${dtype}`);
+  }
+  if (dtype === "string" && $x.dtype !== "string" || dtype !== "string" && $x.dtype === "string") {
+    throw new Error("Only strings can be casted to strings");
+  }
+  const inputs = { x: $x };
+  const attrs = { dtype };
+  return ENGINE.runKernel(Cast, inputs, attrs);
+}
+var cast = op({ cast_ });
+function clone_(x) {
+  const $x = convertToTensor(x, "x", "clone", "string_or_numeric");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Identity, inputs);
+}
+var clone = op({ clone_ });
+function print(x, verbose = false) {
+  console.log(x.toString(verbose));
+}
+getOrMakeEngine();
+var opHandler2 = {
+  buffer,
+  cast,
+  clone,
+  print
+};
+setOpHandler(opHandler2);
+var io_exports = {};
+__export2(io_exports, {
+  browserFiles: () => browserFiles,
+  browserHTTPRequest: () => browserHTTPRequest,
+  concatenateArrayBuffers: () => concatenateArrayBuffers,
+  copyModel: () => copyModel,
+  decodeWeights: () => decodeWeights,
+  encodeWeights: () => encodeWeights,
+  fromMemory: () => fromMemory,
+  fromMemorySync: () => fromMemorySync,
+  getLoadHandlers: () => getLoadHandlers,
+  getModelArtifactsForJSON: () => getModelArtifactsForJSON,
+  getModelArtifactsForJSONSync: () => getModelArtifactsForJSONSync,
+  getModelArtifactsInfoForJSON: () => getModelArtifactsInfoForJSON,
+  getSaveHandlers: () => getSaveHandlers,
+  getWeightSpecs: () => getWeightSpecs,
+  http: () => http,
+  isHTTPScheme: () => isHTTPScheme,
+  listModels: () => listModels,
+  loadWeights: () => loadWeights,
+  moveModel: () => moveModel,
+  registerLoadRouter: () => registerLoadRouter,
+  registerSaveRouter: () => registerSaveRouter,
+  removeModel: () => removeModel,
+  weightsLoaderFactory: () => weightsLoaderFactory,
+  withSaveHandler: () => withSaveHandler,
+  withSaveHandlerSync: () => withSaveHandlerSync
+});
+var DEFAULT_FILE_NAME_PREFIX = "model";
+var DEFAULT_JSON_EXTENSION_NAME = ".json";
+var DEFAULT_WEIGHT_DATA_EXTENSION_NAME = ".weights.bin";
+function defer(f) {
+  return new Promise((resolve) => setTimeout(resolve)).then(f);
+}
+var BrowserDownloads = class {
+  constructor(fileNamePrefix) {
+    if (!env().getBool("IS_BROWSER")) {
+      throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");
+    }
+    if (fileNamePrefix.startsWith(BrowserDownloads.URL_SCHEME)) {
+      fileNamePrefix = fileNamePrefix.slice(BrowserDownloads.URL_SCHEME.length);
+    }
+    if (fileNamePrefix == null || fileNamePrefix.length === 0) {
+      fileNamePrefix = DEFAULT_FILE_NAME_PREFIX;
+    }
+    this.modelJsonFileName = fileNamePrefix + DEFAULT_JSON_EXTENSION_NAME;
+    this.weightDataFileName = fileNamePrefix + DEFAULT_WEIGHT_DATA_EXTENSION_NAME;
+  }
+  async save(modelArtifacts) {
+    if (typeof document === "undefined") {
+      throw new Error("Browser downloads are not supported in this environment since `document` is not present");
+    }
+    const weightsURL = window.URL.createObjectURL(new Blob([modelArtifacts.weightData], { type: "application/octet-stream" }));
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");
+    } else {
+      const weightsManifest = [{
+        paths: ["./" + this.weightDataFileName],
+        weights: modelArtifacts.weightSpecs
+      }];
+      const modelJSON = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);
+      const modelJsonURL = window.URL.createObjectURL(new Blob([JSON.stringify(modelJSON)], { type: "application/json" }));
+      const jsonAnchor = this.modelJsonAnchor == null ? document.createElement("a") : this.modelJsonAnchor;
+      jsonAnchor.download = this.modelJsonFileName;
+      jsonAnchor.href = modelJsonURL;
+      await defer(() => jsonAnchor.dispatchEvent(new MouseEvent("click")));
+      if (modelArtifacts.weightData != null) {
+        const weightDataAnchor = this.weightDataAnchor == null ? document.createElement("a") : this.weightDataAnchor;
+        weightDataAnchor.download = this.weightDataFileName;
+        weightDataAnchor.href = weightsURL;
+        await defer(() => weightDataAnchor.dispatchEvent(new MouseEvent("click")));
+      }
+      return { modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts) };
+    }
+  }
+};
+BrowserDownloads.URL_SCHEME = "downloads://";
+var BrowserFiles = class {
+  constructor(files) {
+    if (files == null || files.length < 1) {
+      throw new Error(`When calling browserFiles, at least 1 file is required, but received ${files}`);
+    }
+    this.jsonFile = files[0];
+    this.weightsFiles = files.slice(1);
+  }
+  async load() {
+    return new Promise((resolve, reject) => {
+      const jsonReader = new FileReader();
+      jsonReader.onload = (event) => {
+        const modelJSON = JSON.parse(event.target.result);
+        const modelTopology = modelJSON.modelTopology;
+        if (modelTopology == null) {
+          reject(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));
+          return;
+        }
+        const weightsManifest = modelJSON.weightsManifest;
+        if (weightsManifest == null) {
+          reject(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));
+          return;
+        }
+        if (this.weightsFiles.length === 0) {
+          resolve({ modelTopology });
+          return;
+        }
+        const modelArtifactsPromise = getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));
+        resolve(modelArtifactsPromise);
+      };
+      jsonReader.onerror = (error) => reject(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`);
+      jsonReader.readAsText(this.jsonFile);
+    });
+  }
+  loadWeights(weightsManifest) {
+    const weightSpecs = [];
+    const paths = [];
+    for (const entry of weightsManifest) {
+      weightSpecs.push(...entry.weights);
+      paths.push(...entry.paths);
+    }
+    const pathToFile = this.checkManifestAndWeightFiles(weightsManifest);
+    const promises = paths.map((path) => this.loadWeightsFile(path, pathToFile[path]));
+    return Promise.all(promises).then((buffers) => [weightSpecs, concatenateArrayBuffers(buffers)]);
+  }
+  loadWeightsFile(path, file) {
+    return new Promise((resolve, reject) => {
+      const weightFileReader = new FileReader();
+      weightFileReader.onload = (event) => {
+        const weightData = event.target.result;
+        resolve(weightData);
+      };
+      weightFileReader.onerror = (error) => reject(`Failed to weights data from file of path '${path}'.`);
+      weightFileReader.readAsArrayBuffer(file);
+    });
+  }
+  checkManifestAndWeightFiles(manifest) {
+    const basenames = [];
+    const fileNames = this.weightsFiles.map((file) => basename(file.name));
+    const pathToFile = {};
+    for (const group of manifest) {
+      group.paths.forEach((path) => {
+        const pathBasename = basename(path);
+        if (basenames.indexOf(pathBasename) !== -1) {
+          throw new Error(`Duplicate file basename found in weights manifest: '${pathBasename}'`);
+        }
+        basenames.push(pathBasename);
+        if (fileNames.indexOf(pathBasename) === -1) {
+          throw new Error(`Weight file with basename '${pathBasename}' is not provided.`);
+        } else {
+          pathToFile[path] = this.weightsFiles[fileNames.indexOf(pathBasename)];
+        }
+      });
+    }
+    if (basenames.length !== this.weightsFiles.length) {
+      throw new Error(`Mismatch in the number of files in weights manifest (${basenames.length}) and the number of weight files provided (${this.weightsFiles.length}).`);
+    }
+    return pathToFile;
+  }
+};
+var browserDownloadsRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserDownloads.URL_SCHEME)) {
+      return browserDownloads(url.slice(BrowserDownloads.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(browserDownloadsRouter);
+function browserDownloads(fileNamePrefix = "model") {
+  return new BrowserDownloads(fileNamePrefix);
+}
+function browserFiles(files) {
+  return new BrowserFiles(files);
+}
+function monitorPromisesProgress(promises, onProgress, startFraction, endFraction) {
+  checkPromises(promises);
+  startFraction = startFraction == null ? 0 : startFraction;
+  endFraction = endFraction == null ? 1 : endFraction;
+  checkFraction(startFraction, endFraction);
+  let resolvedPromise = 0;
+  const registerMonitor = (promise) => {
+    promise.then((value) => {
+      const fraction = startFraction + ++resolvedPromise / promises.length * (endFraction - startFraction);
+      onProgress(fraction);
+      return value;
+    });
+    return promise;
+  };
+  function checkPromises(promises2) {
+    assert(promises2 != null && Array.isArray(promises2) && promises2.length > 0, () => "promises must be a none empty array");
+  }
+  function checkFraction(startFraction2, endFraction2) {
+    assert(startFraction2 >= 0 && startFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got startFraction ${startFraction2}`);
+    assert(endFraction2 >= 0 && endFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got endFraction ${endFraction2}`);
+    assert(endFraction2 >= startFraction2, () => `startFraction must be no more than endFraction, but got startFraction ${startFraction2} and endFraction ${endFraction2}`);
+  }
+  return Promise.all(promises.map(registerMonitor));
+}
+async function loadWeightsAsArrayBuffer(fetchURLs, loadOptions) {
+  if (loadOptions == null) {
+    loadOptions = {};
+  }
+  const fetchFunc = loadOptions.fetchFunc == null ? env().platform.fetch : loadOptions.fetchFunc;
+  const requests = fetchURLs.map((fetchURL) => fetchFunc(fetchURL, loadOptions.requestInit, { isBinary: true }));
+  const fetchStartFraction = 0;
+  const fetchEndFraction = 0.5;
+  const responses = loadOptions.onProgress == null ? await Promise.all(requests) : await monitorPromisesProgress(requests, loadOptions.onProgress, fetchStartFraction, fetchEndFraction);
+  const bufferPromises = responses.map((response) => response.arrayBuffer());
+  const bufferStartFraction = 0.5;
+  const bufferEndFraction = 1;
+  const buffers = loadOptions.onProgress == null ? await Promise.all(bufferPromises) : await monitorPromisesProgress(bufferPromises, loadOptions.onProgress, bufferStartFraction, bufferEndFraction);
+  return buffers;
+}
+async function loadWeights(manifest, filePathPrefix = "", weightNames, requestInit) {
+  const fetchWeights = (fetchUrls) => loadWeightsAsArrayBuffer(fetchUrls, { requestInit });
+  const loadWeights2 = weightsLoaderFactory(fetchWeights);
+  return loadWeights2(manifest, filePathPrefix, weightNames);
+}
+function weightsLoaderFactory(fetchWeightsFunction) {
+  return async (manifest, filePathPrefix = "", weightNames) => {
+    const groupIndicesToFetchMap = manifest.map(() => false);
+    const groupWeightsToFetch = {};
+    const weightsFound = weightNames != null ? weightNames.map(() => false) : [];
+    const allManifestWeightNames = [];
+    manifest.forEach((manifestGroupConfig, groupIndex) => {
+      let groupOffset = 0;
+      manifestGroupConfig.weights.forEach((weightsEntry) => {
+        const rawDtype = "quantization" in weightsEntry ? weightsEntry.quantization.dtype : weightsEntry.dtype;
+        const weightsBytes = DTYPE_VALUE_SIZE_MAP[rawDtype] * sizeFromShape(weightsEntry.shape);
+        const enqueueWeightsForFetchingFn = () => {
+          groupIndicesToFetchMap[groupIndex] = true;
+          if (groupWeightsToFetch[groupIndex] == null) {
+            groupWeightsToFetch[groupIndex] = [];
+          }
+          groupWeightsToFetch[groupIndex].push({
+            manifestEntry: weightsEntry,
+            groupOffset,
+            sizeBytes: weightsBytes
+          });
+        };
+        if (weightNames != null) {
+          weightNames.forEach((weightName, weightIndex) => {
+            if (weightName === weightsEntry.name) {
+              enqueueWeightsForFetchingFn();
+              weightsFound[weightIndex] = true;
+            }
+          });
+        } else {
+          enqueueWeightsForFetchingFn();
+        }
+        allManifestWeightNames.push(weightsEntry.name);
+        groupOffset += weightsBytes;
+      });
+    });
+    if (!weightsFound.every((found) => found)) {
+      const weightsNotFound = weightNames.filter((_, i) => !weightsFound[i]);
+      throw new Error(`Could not find weights in manifest with names: ${weightsNotFound.join(", ")}. 
+Manifest JSON has weights with names: ${allManifestWeightNames.join(", ")}.`);
+    }
+    const groupIndicesToFetch = groupIndicesToFetchMap.reduce((accumulator, shouldFetch, i) => {
+      if (shouldFetch) {
+        accumulator.push(i);
+      }
+      return accumulator;
+    }, []);
+    const fetchUrls = [];
+    groupIndicesToFetch.forEach((i) => {
+      manifest[i].paths.forEach((filepath) => {
+        const fetchUrl = filePathPrefix + (!filePathPrefix.endsWith("/") ? "/" : "") + filepath;
+        fetchUrls.push(fetchUrl);
+      });
+    });
+    const buffers = await fetchWeightsFunction(fetchUrls);
+    const weightsTensorMap = {};
+    let bufferIndexOffset = 0;
+    groupIndicesToFetch.forEach((i) => {
+      const numBuffers = manifest[i].paths.length;
+      let groupBytes = 0;
+      for (let i2 = 0; i2 < numBuffers; i2++) {
+        groupBytes += buffers[bufferIndexOffset + i2].byteLength;
+      }
+      const groupBuffer = new ArrayBuffer(groupBytes);
+      const groupByteBuffer = new Uint8Array(groupBuffer);
+      let groupBufferOffset = 0;
+      for (let i2 = 0; i2 < numBuffers; i2++) {
+        const buffer2 = new Uint8Array(buffers[bufferIndexOffset + i2]);
+        groupByteBuffer.set(buffer2, groupBufferOffset);
+        groupBufferOffset += buffer2.byteLength;
+      }
+      const weightsEntries = groupWeightsToFetch[i];
+      weightsEntries.forEach((weightsEntry) => {
+        const byteBuffer = groupBuffer.slice(weightsEntry.groupOffset, weightsEntry.groupOffset + weightsEntry.sizeBytes);
+        const nameToTensorMap = decodeWeights(byteBuffer, [weightsEntry.manifestEntry]);
+        for (const name in nameToTensorMap) {
+          weightsTensorMap[name] = nameToTensorMap[name];
+        }
+      });
+      bufferIndexOffset += numBuffers;
+    });
+    return weightsTensorMap;
+  };
+}
+var OCTET_STREAM_MIME_TYPE = "application/octet-stream";
+var JSON_TYPE = "application/json";
+var HTTPRequest = class {
+  constructor(path, loadOptions) {
+    this.DEFAULT_METHOD = "POST";
+    if (loadOptions == null) {
+      loadOptions = {};
+    }
+    this.weightPathPrefix = loadOptions.weightPathPrefix;
+    this.onProgress = loadOptions.onProgress;
+    this.weightUrlConverter = loadOptions.weightUrlConverter;
+    if (loadOptions.fetchFunc != null) {
+      assert(typeof loadOptions.fetchFunc === "function", () => "Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)");
+      this.fetch = loadOptions.fetchFunc;
+    } else {
+      this.fetch = env().platform.fetch;
+    }
+    assert(path != null && path.length > 0, () => "URL path for http must not be null, undefined or empty.");
+    if (Array.isArray(path)) {
+      assert(path.length === 2, () => `URL paths for http must have a length of 2, (actual length is ${path.length}).`);
+    }
+    this.path = path;
+    if (loadOptions.requestInit != null && loadOptions.requestInit.body != null) {
+      throw new Error("requestInit is expected to have no pre-existing body, but has one.");
+    }
+    this.requestInit = loadOptions.requestInit || {};
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");
+    }
+    const init2 = Object.assign({ method: this.DEFAULT_METHOD }, this.requestInit);
+    init2.body = new FormData();
+    const weightsManifest = [{
+      paths: ["./model.weights.bin"],
+      weights: modelArtifacts.weightSpecs
+    }];
+    const modelTopologyAndWeightManifest = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);
+    init2.body.append("model.json", new Blob([JSON.stringify(modelTopologyAndWeightManifest)], { type: JSON_TYPE }), "model.json");
+    if (modelArtifacts.weightData != null) {
+      init2.body.append("model.weights.bin", new Blob([modelArtifacts.weightData], { type: OCTET_STREAM_MIME_TYPE }), "model.weights.bin");
+    }
+    const response = await this.fetch(this.path, init2);
+    if (response.ok) {
+      return {
+        modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts),
+        responses: [response]
+      };
+    } else {
+      throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${response.status}.`);
+    }
+  }
+  async load() {
+    const modelConfigRequest = await this.fetch(this.path, this.requestInit);
+    if (!modelConfigRequest.ok) {
+      throw new Error(`Request to ${this.path} failed with status code ${modelConfigRequest.status}. Please verify this URL points to the model JSON of the model to load.`);
+    }
+    let modelJSON;
+    try {
+      modelJSON = await modelConfigRequest.json();
+    } catch (e) {
+      let message = `Failed to parse model JSON of response from ${this.path}.`;
+      if (this.path.endsWith(".pb")) {
+        message += " Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.";
+      } else {
+        message += " Please make sure the server is serving valid JSON for this request.";
+      }
+      throw new Error(message);
+    }
+    const modelTopology = modelJSON.modelTopology;
+    const weightsManifest = modelJSON.weightsManifest;
+    if (modelTopology == null && weightsManifest == null) {
+      throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);
+    }
+    return getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));
+  }
+  async loadWeights(weightsManifest) {
+    const weightPath = Array.isArray(this.path) ? this.path[1] : this.path;
+    const [prefix, suffix] = parseUrl(weightPath);
+    const pathPrefix = this.weightPathPrefix || prefix;
+    const weightSpecs = getWeightSpecs(weightsManifest);
+    const fetchURLs = [];
+    const urlPromises = [];
+    for (const weightsGroup of weightsManifest) {
+      for (const path of weightsGroup.paths) {
+        if (this.weightUrlConverter != null) {
+          urlPromises.push(this.weightUrlConverter(path));
+        } else {
+          fetchURLs.push(pathPrefix + path + suffix);
+        }
+      }
+    }
+    if (this.weightUrlConverter) {
+      fetchURLs.push(...await Promise.all(urlPromises));
+    }
+    const buffers = await loadWeightsAsArrayBuffer(fetchURLs, {
+      requestInit: this.requestInit,
+      fetchFunc: this.fetch,
+      onProgress: this.onProgress
+    });
+    return [weightSpecs, concatenateArrayBuffers(buffers)];
+  }
+};
+HTTPRequest.URL_SCHEME_REGEX = /^https?:\/\//;
+function parseUrl(url) {
+  const lastSlash = url.lastIndexOf("/");
+  const lastSearchParam = url.lastIndexOf("?");
+  const prefix = url.substring(0, lastSlash);
+  const suffix = lastSearchParam > lastSlash ? url.substring(lastSearchParam) : "";
+  return [prefix + "/", suffix];
+}
+function isHTTPScheme(url) {
+  return url.match(HTTPRequest.URL_SCHEME_REGEX) != null;
+}
+var httpRouter = (url, loadOptions) => {
+  if (typeof fetch === "undefined" && (loadOptions == null || loadOptions.fetchFunc == null)) {
+    return null;
+  } else {
+    let isHTTP = true;
+    if (Array.isArray(url)) {
+      isHTTP = url.every((urlItem) => isHTTPScheme(urlItem));
+    } else {
+      isHTTP = isHTTPScheme(url);
+    }
+    if (isHTTP) {
+      return http(url, loadOptions);
+    }
+  }
+  return null;
+};
+IORouterRegistry.registerSaveRouter(httpRouter);
+IORouterRegistry.registerLoadRouter(httpRouter);
+function http(path, loadOptions) {
+  return new HTTPRequest(path, loadOptions);
+}
+function browserHTTPRequest(path, loadOptions) {
+  return http(path, loadOptions);
+}
+var PassthroughLoader = class {
+  constructor(modelArtifacts) {
+    this.modelArtifacts = modelArtifacts;
+  }
+  load() {
+    return this.modelArtifacts;
+  }
+};
+var PassthroughSaver = class {
+  constructor(saveHandler) {
+    this.saveHandler = saveHandler;
+  }
+  save(modelArtifacts) {
+    return this.saveHandler(modelArtifacts);
+  }
+};
+var PassthroughAsync = class {
+  constructor(handler) {
+    if (handler.load) {
+      this.load = () => Promise.resolve(handler.load());
+    }
+    if (handler.save) {
+      this.save = (modelArtifacts) => Promise.resolve(handler.save(modelArtifacts));
+    }
+  }
+};
+function fromMemory(modelArtifacts, weightSpecs, weightData, trainingConfig) {
+  const args = arguments;
+  return new PassthroughAsync(fromMemorySync(...args));
+}
+function fromMemorySync(modelArtifacts, weightSpecs, weightData, trainingConfig) {
+  if (arguments.length === 1) {
+    const isModelArtifacts = modelArtifacts.modelTopology != null || modelArtifacts.weightSpecs != null;
+    if (isModelArtifacts) {
+      return new PassthroughLoader(modelArtifacts);
+    } else {
+      console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.");
+      return new PassthroughLoader({ modelTopology: modelArtifacts });
+    }
+  } else {
+    console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.");
+    return new PassthroughLoader({
+      modelTopology: modelArtifacts,
+      weightSpecs,
+      weightData,
+      trainingConfig
+    });
+  }
+}
+function withSaveHandler(saveHandler) {
+  return new PassthroughSaver(saveHandler);
+}
+function withSaveHandlerSync(saveHandler) {
+  return new PassthroughSaver(saveHandler);
+}
+var math_exports = {};
+__export2(math_exports, {
+  confusionMatrix: () => confusionMatrix
+});
+function matMul_(a, b, transposeA = false, transposeB = false) {
+  let $a = convertToTensor(a, "a", "matMul");
+  let $b = convertToTensor(b, "b", "matMul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  const attrs = { transposeA, transposeB };
+  return ENGINE.runKernel(BatchMatMul, inputs, attrs);
+}
+var matMul = op({ matMul_ });
+function oneHot_(indices, depth, onValue = 1, offValue = 0, dtype = "int32") {
+  if (depth < 2) {
+    throw new Error(`Error in oneHot: depth must be >=2, but it is ${depth}`);
+  }
+  const $indices = convertToTensor(indices, "indices", "oneHot", "int32");
+  const inputs = { indices: $indices };
+  const attrs = { dtype, depth, onValue, offValue };
+  return ENGINE.runKernel(OneHot, inputs, attrs);
+}
+var oneHot = op({ oneHot_ });
+function enableProdMode() {
+  env().set("PROD", true);
+}
+function enableDebugMode() {
+  env().set("DEBUG", true);
+}
+function disableDeprecationWarnings() {
+  env().set("DEPRECATION_WARNINGS_ENABLED", false);
+  console.warn(`TensorFlow.js deprecation warnings have been disabled.`);
+}
+function deprecationWarn(msg) {
+  if (env().getBool("DEPRECATION_WARNINGS_ENABLED")) {
+    console.warn(msg + " You can disable deprecation warnings with tf.disableDeprecationWarnings().");
+  }
+}
+setDeprecationWarningFn(deprecationWarn);
+function disposeVariables() {
+  ENGINE.disposeVariables();
+}
+function engine() {
+  return ENGINE;
+}
+function memory() {
+  return ENGINE.memory();
+}
+function profile(f) {
+  return ENGINE.profile(f);
+}
+function tidy(nameOrFn, fn) {
+  return ENGINE.tidy(nameOrFn, fn);
+}
+function dispose(container) {
+  const tensors = getTensorsInContainer(container);
+  tensors.forEach((tensor2) => tensor2.dispose());
+}
+function keep(result) {
+  return ENGINE.keep(result);
+}
+function time(f) {
+  return ENGINE.time(f);
+}
+function setBackend(backendName) {
+  return ENGINE.setBackend(backendName);
+}
+function ready() {
+  return ENGINE.ready();
+}
+function getBackend() {
+  return ENGINE.backendName;
+}
+function removeBackend(name) {
+  ENGINE.removeBackend(name);
+}
+function findBackend(name) {
+  return ENGINE.findBackend(name);
+}
+function findBackendFactory(name) {
+  return ENGINE.findBackendFactory(name);
+}
+function registerBackend(name, factory, priority = 1) {
+  return ENGINE.registerBackend(name, factory, priority);
+}
+function backend() {
+  return ENGINE.backend;
+}
+function setPlatform(platformName, platform) {
+  env().setPlatform(platformName, platform);
+}
+function imag_(input2) {
+  const $input = convertToTensor(input2, "input", "imag");
+  const inputs = { input: $input };
+  return ENGINE.runKernel(Imag, inputs);
+}
+var imag = op({ imag_ });
+function neg_(x) {
+  const $x = convertToTensor(x, "x", "neg");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Neg, inputs);
+}
+var neg = op({ neg_ });
+function real_(input2) {
+  const $input = convertToTensor(input2, "input", "real");
+  const inputs = { input: $input };
+  return ENGINE.runKernel(Real, inputs);
+}
+var real = op({ real_ });
+function transpose_(x, perm, conjugate) {
+  const $x = convertToTensor(x, "x", "transpose");
+  if (perm == null) {
+    perm = $x.shape.map((s, i) => i).reverse();
+  }
+  assert($x.rank === perm.length, () => `Error in transpose: rank of input ${$x.rank} must match length of perm ${perm}.`);
+  perm.forEach((axis) => {
+    assert(axis >= 0 && axis < $x.rank, () => `All entries in 'perm' must be between 0 and ${$x.rank - 1} but got ${perm}`);
+  });
+  if ($x.rank <= 1) {
+    return $x.clone();
+  }
+  const inputs = { x: $x };
+  const attrs = { perm };
+  if ($x.dtype === "complex64") {
+    return tidy(() => {
+      let $real = real($x);
+      let $imag = imag($x);
+      $real = ENGINE.runKernel(Transpose, { x: $real }, attrs);
+      $imag = ENGINE.runKernel(Transpose, { x: $imag }, attrs);
+      if (conjugate) {
+        $imag = neg($imag);
+      }
+      return complex($real, $imag);
+    });
+  }
+  return ENGINE.runKernel(Transpose, inputs, attrs);
+}
+var transpose = op({ transpose_ });
+function confusionMatrix_(labels, predictions, numClasses) {
+  const $labels = convertToTensor(labels, "labels", "confusionMatrix");
+  const $predictions = convertToTensor(predictions, "predictions", "confusionMatrix");
+  assert(numClasses == null || numClasses > 0 && Number.isInteger(numClasses), () => `If provided, numClasses must be a positive integer, but got ${numClasses}`);
+  assert($labels.rank === 1, () => `Expected the rank of labels to be 1, but got ${$labels.rank}`);
+  assert($predictions.rank === 1, () => `Expected the rank of predictions to be 1, but got ${$predictions.rank}`);
+  assert($labels.shape[0] === $predictions.shape[0], () => `Mismatch in the number of examples: ${$labels.shape[0]} vs. ${$predictions.shape[0]}. Labels and predictions should have the same number of elements.`);
+  assert(numClasses > 0 && Number.isInteger(numClasses), () => `numClasses is required to be a positive integer, but got ${numClasses}`);
+  const oneHotLabels = oneHot(cast($labels, "int32"), numClasses);
+  const oneHotPredictions = oneHot(cast($predictions, "int32"), numClasses);
+  const oneHotLabelsT = transpose(oneHotLabels);
+  const product = matMul(oneHotLabelsT, oneHotPredictions);
+  return cast(product, "int32");
+}
+var confusionMatrix = op({ confusionMatrix_ });
+var broadcast_util_exports = {};
+__export2(broadcast_util_exports, {
+  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,
+  getBroadcastDims: () => getBroadcastDims,
+  getReductionAxes: () => getReductionAxes
+});
+function getBroadcastDims(inShape, outShape) {
+  const inRank = inShape.length;
+  const dims = [];
+  for (let i = 0; i < inRank; i++) {
+    const dim = inRank - 1 - i;
+    const a = inShape[dim] || 1;
+    const b = outShape[outShape.length - 1 - i] || 1;
+    if (b > 1 && a === 1) {
+      dims.unshift(dim);
+    }
+  }
+  return dims;
+}
+function getReductionAxes(inShape, outShape) {
+  const result = [];
+  for (let i = 0; i < outShape.length; i++) {
+    const inDim = inShape[inShape.length - i - 1];
+    const outAxis = outShape.length - i - 1;
+    const outDim = outShape[outAxis];
+    if (inDim == null || inDim === 1 && outDim > 1) {
+      result.unshift(outAxis);
+    }
+  }
+  return result;
+}
+function assertAndGetBroadcastShape(shapeA, shapeB) {
+  const result = [];
+  const l = Math.max(shapeA.length, shapeB.length);
+  for (let i = 0; i < l; i++) {
+    let a = shapeA[shapeA.length - i - 1];
+    if (a == null) {
+      a = 1;
+    }
+    let b = shapeB[shapeB.length - i - 1];
+    if (b == null) {
+      b = 1;
+    }
+    if (a === 1) {
+      result.unshift(b);
+    } else if (b === 1) {
+      result.unshift(a);
+    } else if (a !== b) {
+      const errMsg = `Operands could not be broadcast together with shapes ${shapeA} and ${shapeB}.`;
+      throw Error(errMsg);
+    } else {
+      result.unshift(a);
+    }
+  }
+  return result;
+}
+var browser_exports = {};
+__export2(browser_exports, {
+  fromPixels: () => fromPixels,
+  fromPixelsAsync: () => fromPixelsAsync,
+  toPixels: () => toPixels
+});
+function tensor3d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 3) {
+    throw new Error("tensor3d() requires shape to have three numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 3 && inferredShape.length !== 1) {
+    throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+var fromPixels2DContext;
+function fromPixels_(pixels, numChannels = 3) {
+  if (numChannels > 4) {
+    throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");
+  }
+  if (pixels == null) {
+    throw new Error("pixels passed to tf.browser.fromPixels() can not be null");
+  }
+  let isPixelData2 = false;
+  let isImageData = false;
+  let isVideo = false;
+  let isImage = false;
+  let isCanvasLike = false;
+  let isImageBitmap = false;
+  if (pixels.data instanceof Uint8Array) {
+    isPixelData2 = true;
+  } else if (typeof ImageData !== "undefined" && pixels instanceof ImageData) {
+    isImageData = true;
+  } else if (typeof HTMLVideoElement !== "undefined" && pixels instanceof HTMLVideoElement) {
+    isVideo = true;
+  } else if (typeof HTMLImageElement !== "undefined" && pixels instanceof HTMLImageElement) {
+    isImage = true;
+  } else if (pixels.getContext != null) {
+    isCanvasLike = true;
+  } else if (typeof ImageBitmap !== "undefined" && pixels instanceof ImageBitmap) {
+    isImageBitmap = true;
+  } else {
+    throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${pixels.constructor.name}`);
+  }
+  const kernel = getKernel(FromPixels, ENGINE.backendName);
+  if (kernel != null) {
+    const inputs = { pixels };
+    const attrs = { numChannels };
+    return ENGINE.runKernel(FromPixels, inputs, attrs);
+  }
+  const [width, height] = isVideo ? [
+    pixels.videoWidth,
+    pixels.videoHeight
+  ] : [pixels.width, pixels.height];
+  let vals;
+  if (isCanvasLike) {
+    vals = pixels.getContext("2d").getImageData(0, 0, width, height).data;
+  } else if (isImageData || isPixelData2) {
+    vals = pixels.data;
+  } else if (isImage || isVideo || isImageBitmap) {
+    if (fromPixels2DContext == null) {
+      if (typeof document === "undefined") {
+        if (typeof OffscreenCanvas !== "undefined" && typeof OffscreenCanvasRenderingContext2D !== "undefined") {
+          fromPixels2DContext = new OffscreenCanvas(1, 1).getContext("2d");
+        } else {
+          throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");
+        }
+      } else {
+        fromPixels2DContext = document.createElement("canvas").getContext("2d", { willReadFrequently: true });
+      }
+    }
+    fromPixels2DContext.canvas.width = width;
+    fromPixels2DContext.canvas.height = height;
+    fromPixels2DContext.drawImage(pixels, 0, 0, width, height);
+    vals = fromPixels2DContext.getImageData(0, 0, width, height).data;
+  }
+  let values;
+  if (numChannels === 4) {
+    values = new Int32Array(vals);
+  } else {
+    const numPixels = width * height;
+    values = new Int32Array(numPixels * numChannels);
+    for (let i = 0; i < numPixels; i++) {
+      for (let channel = 0; channel < numChannels; ++channel) {
+        values[i * numChannels + channel] = vals[i * 4 + channel];
+      }
+    }
+  }
+  const outShape = [height, width, numChannels];
+  return tensor3d(values, outShape, "int32");
+}
+function isPixelData(pixels) {
+  return pixels != null && pixels.data instanceof Uint8Array;
+}
+function isImageBitmapFullySupported() {
+  return typeof window !== "undefined" && typeof ImageBitmap !== "undefined" && window.hasOwnProperty("createImageBitmap");
+}
+function isNonEmptyPixels(pixels) {
+  return pixels != null && pixels.width !== 0 && pixels.height !== 0;
+}
+function canWrapPixelsToImageBitmap(pixels) {
+  return isImageBitmapFullySupported() && !(pixels instanceof ImageBitmap) && isNonEmptyPixels(pixels) && !isPixelData(pixels);
+}
+async function fromPixelsAsync(pixels, numChannels = 3) {
+  let inputs = null;
+  if (env().getBool("WRAP_TO_IMAGEBITMAP") && canWrapPixelsToImageBitmap(pixels)) {
+    let imageBitmap;
+    try {
+      imageBitmap = await createImageBitmap(pixels, { premultiplyAlpha: "none" });
+    } catch (e) {
+      imageBitmap = null;
+    }
+    if (imageBitmap != null && imageBitmap.width === pixels.width && imageBitmap.height === pixels.height) {
+      inputs = imageBitmap;
+    } else {
+      inputs = pixels;
+    }
+  } else {
+    inputs = pixels;
+  }
+  return fromPixels_(inputs, numChannels);
+}
+async function toPixels(img, canvas) {
+  let $img = convertToTensor(img, "img", "toPixels");
+  if (!(img instanceof Tensor)) {
+    const originalImgTensor = $img;
+    $img = cast(originalImgTensor, "int32");
+    originalImgTensor.dispose();
+  }
+  if ($img.rank !== 2 && $img.rank !== 3) {
+    throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${$img.rank}.`);
+  }
+  const [height, width] = $img.shape.slice(0, 2);
+  const depth = $img.rank === 2 ? 1 : $img.shape[2];
+  if (depth > 4 || depth === 2) {
+    throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${depth}`);
+  }
+  if ($img.dtype !== "float32" && $img.dtype !== "int32") {
+    throw new Error(`Unsupported type for toPixels: ${$img.dtype}. Please use float32 or int32 tensors.`);
+  }
+  const data = await $img.data();
+  const multiplier = $img.dtype === "float32" ? 255 : 1;
+  const bytes = new Uint8ClampedArray(width * height * 4);
+  for (let i = 0; i < height * width; ++i) {
+    const rgba = [0, 0, 0, 255];
+    for (let d = 0; d < depth; d++) {
+      const value = data[i * depth + d];
+      if ($img.dtype === "float32") {
+        if (value < 0 || value > 1) {
+          throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${value}.`);
+        }
+      } else if ($img.dtype === "int32") {
+        if (value < 0 || value > 255) {
+          throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${value}.`);
+        }
+      }
+      if (depth === 1) {
+        rgba[0] = value * multiplier;
+        rgba[1] = value * multiplier;
+        rgba[2] = value * multiplier;
+      } else {
+        rgba[d] = value * multiplier;
+      }
+    }
+    const j = i * 4;
+    bytes[j + 0] = Math.round(rgba[0]);
+    bytes[j + 1] = Math.round(rgba[1]);
+    bytes[j + 2] = Math.round(rgba[2]);
+    bytes[j + 3] = Math.round(rgba[3]);
+  }
+  if (canvas != null) {
+    canvas.width = width;
+    canvas.height = height;
+    const ctx = canvas.getContext("2d");
+    const imageData = new ImageData(bytes, width, height);
+    ctx.putImageData(imageData, 0, 0);
+  }
+  if ($img !== img) {
+    $img.dispose();
+  }
+  return bytes;
+}
+var fromPixels = op({ fromPixels_ });
+var gather_nd_util_exports = {};
+__export2(gather_nd_util_exports, {
+  prepareAndValidate: () => prepareAndValidate
+});
+function prepareAndValidate(tensor2, indices) {
+  const tensorRank = tensor2.shape.length;
+  const indicesRank = indices.shape.length;
+  if (tensorRank < 1) {
+    throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${tensorRank}.`);
+  }
+  if (indicesRank < 1) {
+    throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${indicesRank}.`);
+  }
+  if (indices.dtype !== "int32") {
+    throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${indices.dtype}.`);
+  }
+  if (indices.shape[indicesRank - 1] > tensorRank) {
+    throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${indices.shape[indicesRank - 1]} vs. ${tensorRank}`);
+  }
+  if (sizeFromShape(tensor2.shape) === 0) {
+    throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${tensor2.shape}.`);
+  }
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  let nResult = 1;
+  for (let i = 0; i < indicesShape.length - 1; ++i) {
+    nResult *= indicesShape[i];
+  }
+  const inputShape = tensor2.shape;
+  const resultShape = indicesShape.slice();
+  resultShape.pop();
+  let sliceSize = 1;
+  for (let i = sliceRank; i < tensorRank; ++i) {
+    sliceSize *= inputShape[i];
+    resultShape.push(inputShape[i]);
+  }
+  const strides = [
+    ...computeStrides(tensor2.shape).map((stride) => stride / sliceSize),
+    1
+  ].slice(0, sliceRank);
+  return [resultShape, nResult, sliceSize, strides];
+}
+var scatter_nd_util_exports = {};
+__export2(scatter_nd_util_exports, {
+  calculateShapes: () => calculateShapes,
+  validateInput: () => validateInput,
+  validateUpdateShape: () => validateUpdateShape
+});
+function validateUpdateShape(shape, indices, updates) {
+  const sliceDim = indices.rank > 1 ? indices.shape[indices.rank - 1] : 1;
+  const batchDim = indices.rank > 1 ? indices.rank - 1 : 1;
+  const shapeError = `Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${updates.shape}, indices.shape: ${indices.shape}, shape: ${shape}, sliceDim: ${sliceDim}, and batchDim: ${batchDim}.`;
+  if (updates.rank < batchDim) {
+    throw new Error(shapeError + ` update.rank < ${batchDim}. `);
+  }
+  if (shape.length < sliceDim + (updates.rank - batchDim)) {
+    throw new Error(shapeError + ` Output shape length < ${sliceDim + (updates.rank - batchDim)}`);
+  }
+  if (updates.rank !== batchDim + shape.length - sliceDim) {
+    throw new Error(shapeError + ` update.rank != ${batchDim + shape.length - sliceDim}`);
+  }
+  for (let d = 0; d < batchDim; ++d) {
+    if (updates.shape[d] !== indices.shape[d]) {
+      throw new Error(shapeError + ` updates.shape[${d}] (${updates.shape[d]}) != indices.shape[${d}] (${indices.shape[d]}).`);
+    }
+  }
+  for (let d = 0; d < updates.rank - batchDim; ++d) {
+    if (updates.shape[d + batchDim] !== shape[d + sliceDim]) {
+      throw new Error(shapeError + ` updates.shape[${d + batchDim}] (${updates.shape[d + batchDim]}) != shape[${d + batchDim}] (${shape[d + batchDim]})`);
+    }
+  }
+}
+function validateInput(updates, indices, shape) {
+  if (indices.rank < 1) {
+    throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${indices.rank}.`);
+  }
+  if (updates.rank < 1) {
+    throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${updates.rank}.`);
+  }
+  if (indices.dtype !== "int32") {
+    throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${indices.dtype}`);
+  }
+  if (shape.length < 1) {
+    throw new Error(`Output rank must be greater or equal to 1, but got shape: ${shape}`);
+  }
+  if (shape.length === 0) {
+    if (indices.size === 0) {
+      throw new Error(`Indices specified for empty output. indices shape: ${indices.shape}`);
+    }
+    if (updates.size === 0) {
+      throw new Error(`Updates specified for empty output. updates shape: ${updates.shape}`);
+    }
+  }
+  validateUpdateShape(shape, indices, updates);
+}
+function calculateShapes(updates, indices, shape) {
+  const indicesRank = indices.shape.length;
+  const sliceRank = indicesRank > 1 ? indices.shape[indicesRank - 1] : 1;
+  const totalNd = shape.length;
+  let sliceSize = 1;
+  for (let i = sliceRank; i < totalNd; ++i) {
+    sliceSize *= shape[i];
+  }
+  const safeSliceDim = sliceRank < 1 ? 1 : sliceRank;
+  const numUpdates = sizeFromShape(indices.shape) / safeSliceDim;
+  const strides = [...computeStrides(shape.slice(0, sliceRank)), 1];
+  const outputSize = sizeFromShape(shape);
+  return { sliceRank, numUpdates, sliceSize, strides, outputSize };
+}
+var slice_util_exports = {};
+__export2(slice_util_exports, {
+  assertParamsValid: () => assertParamsValid,
+  computeFlatOffset: () => computeFlatOffset,
+  computeOutShape: () => computeOutShape,
+  getNormalizedAxes: () => getNormalizedAxes,
+  isSliceContinous: () => isSliceContinous,
+  maskToAxes: () => maskToAxes,
+  parseSliceParams: () => parseSliceParams,
+  sliceInfo: () => sliceInfo,
+  startForAxis: () => startForAxis,
+  startIndicesWithElidedDims: () => startIndicesWithElidedDims,
+  stopForAxis: () => stopForAxis,
+  stopIndicesWithElidedDims: () => stopIndicesWithElidedDims,
+  stridesForAxis: () => stridesForAxis,
+  stridesWithElidedDims: () => stridesWithElidedDims
+});
+var NEW_AXIS = -2;
+var SHRINK_AXIS = -1;
+function assertParamsValid(input2, begin, size) {
+  const inputRank = input2.shape.length;
+  assert(inputRank === begin.length, () => `Error in slice${inputRank}D: Length of begin ${begin} must match the rank of the array (${inputRank}).`);
+  assert(inputRank === size.length, () => `Error in slice${inputRank}D: Length of size ${size} must match the rank of the array (${inputRank}).`);
+  for (let i = 0; i < inputRank; ++i) {
+    assert(begin[i] + size[i] <= input2.shape[i], () => `Error in slice${inputRank}D: begin[${i}] + size[${i}] (${begin[i] + size[i]}) would overflow input.shape[${i}] (${input2.shape[i]})`);
+  }
+}
+function maskToAxes(mask) {
+  const axes = [];
+  let axis = 0;
+  while (mask > 0) {
+    if (mask & 1) {
+      axes.push(axis);
+    }
+    mask /= 2;
+    axis++;
+  }
+  return axes;
+}
+function computeOutShape(begin, end, strides) {
+  const size = [];
+  for (let axis = 0; axis < begin.length; axis++) {
+    size[axis] = Math.ceil((end[axis] - begin[axis]) / strides[axis]);
+  }
+  return size;
+}
+function stridesWithElidedDims(strides, ellipsisInsertionIndex, numElidedAxes, inputShape) {
+  const newStrides = [...strides];
+  for (let i = newStrides.length; i < inputShape.length; i++) {
+    newStrides.push(1);
+  }
+  for (let i = 0; i < numElidedAxes; i++) {
+    if (i === 0) {
+      newStrides[ellipsisInsertionIndex] = 1;
+    } else {
+      newStrides.splice(ellipsisInsertionIndex, 0, 1);
+      newStrides.pop();
+    }
+  }
+  return newStrides;
+}
+function unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, normalizedAxis) {
+  if (normalizedAxis <= ellipsisInsertionIndex) {
+    return normalizedAxis;
+  }
+  return normalizedAxis - (numElidedAxes - 1);
+}
+function getElidedAxes(numElidedAxes, ellipsisInsertionIndex) {
+  const elidedAxes = [];
+  for (let i = 0; i < numElidedAxes; i++) {
+    elidedAxes.push(ellipsisInsertionIndex + i);
+  }
+  return elidedAxes;
+}
+function getNormalizedAxes(inputShape, ellipsisAxes, numInterpolatedAxes, begin, end, strides, beginMask, endMask, ellipsisMask) {
+  const inputRank = inputShape.length;
+  let normalizedBegin = new Array(inputRank), normalizedEnd = new Array(inputRank), normalizedStrides = new Array(inputRank);
+  if (ellipsisAxes.length && numInterpolatedAxes > 0) {
+    const fullIndex = ellipsisAxes[0];
+    const numElidedAxes = numInterpolatedAxes + 1;
+    normalizedBegin = startIndicesWithElidedDims(beginMask, fullIndex, numElidedAxes, begin, inputShape);
+    normalizedEnd = stopIndicesWithElidedDims(endMask, fullIndex, numElidedAxes, end, inputShape);
+    normalizedStrides = stridesWithElidedDims(strides, fullIndex, numElidedAxes, inputShape);
+  } else {
+    for (let axis = 0; axis < inputRank; axis++) {
+      normalizedBegin[axis] = startForAxis(beginMask, begin, strides, inputShape, axis, ellipsisMask);
+      normalizedEnd[axis] = stopForAxis(endMask, end, strides, inputShape, axis, ellipsisMask);
+      normalizedStrides[axis] = stridesForAxis(strides, axis, ellipsisMask);
+    }
+  }
+  return {
+    begin: normalizedBegin,
+    end: normalizedEnd,
+    strides: normalizedStrides
+  };
+}
+function startIndicesWithElidedDims(beginMask, ellipsisInsertionIndex, numElidedAxes, originalBegin, inputShape) {
+  const newIndices = [...inputShape];
+  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);
+  for (let axis = 0; axis < newIndices.length; axis++) {
+    if (elidedAxes.indexOf(axis) > -1) {
+      newIndices[axis] = 0;
+    } else {
+      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);
+      let originalValue = originalBegin[originalAxis];
+      if (beginMask & 1 << originalAxis) {
+        originalValue = 0;
+      }
+      newIndices[axis] = originalValue;
+    }
+  }
+  return newIndices;
+}
+function stopIndicesWithElidedDims(endMask, ellipsisInsertionIndex, numElidedAxes, originalEnd, inputShape) {
+  const newIndices = [...inputShape];
+  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);
+  for (let axis = 0; axis < newIndices.length; axis++) {
+    if (elidedAxes.indexOf(axis) > -1) {
+      newIndices[axis] = Number.MAX_SAFE_INTEGER;
+    } else {
+      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);
+      let originalValue = originalEnd[originalAxis];
+      if (endMask & 1 << originalAxis) {
+        originalValue = Number.MAX_SAFE_INTEGER;
+      }
+      newIndices[axis] = originalValue;
+    }
+  }
+  for (let i = 0; i < newIndices.length; i++) {
+    const axisSize = inputShape[i];
+    if (newIndices[i] < 0) {
+      newIndices[i] += axisSize;
+    }
+    newIndices[i] = clamp(0, newIndices[i], inputShape[i]);
+  }
+  return newIndices;
+}
+function stridesForAxis(strides, axis, ellipsisMask) {
+  let stride = strides[axis];
+  if (ellipsisMask & 1 << axis || stride == null) {
+    stride = 1;
+  }
+  return stride;
+}
+function startForAxis(beginMask, startIndices, strides, inputShape, axis, ellipsisMask) {
+  let start = startIndices[axis];
+  const stride = strides[axis] || 1;
+  if (beginMask & 1 << axis || ellipsisMask & 1 << axis || start == null) {
+    if (stride > 0) {
+      start = Number.MIN_SAFE_INTEGER;
+    } else {
+      start = Number.MAX_SAFE_INTEGER;
+    }
+  }
+  const axisSize = inputShape[axis];
+  if (start < 0) {
+    start += axisSize;
+  }
+  start = clamp(0, start, axisSize - 1);
+  return start;
+}
+function stopForAxis(endMask, stopIndices, strides, inputShape, axis, ellipsisMask) {
+  let stop = stopIndices[axis];
+  const stride = strides[axis] || 1;
+  if (endMask & 1 << axis || ellipsisMask & 1 << axis || stop == null) {
+    if (stride > 0) {
+      stop = Number.MAX_SAFE_INTEGER;
+    } else {
+      stop = Number.MIN_SAFE_INTEGER;
+    }
+  }
+  const axisSize = inputShape[axis];
+  if (stop < 0) {
+    stop += axisSize;
+  }
+  if (stride > 0) {
+    stop = clamp(0, stop, axisSize);
+  } else {
+    stop = clamp(-1, stop, axisSize - 1);
+  }
+  return stop;
+}
+function isSliceContinous(shape, begin, size) {
+  let firstNonOneAxis = size.length;
+  for (let i = 0; i < size.length; i++) {
+    if (size[i] > 1) {
+      firstNonOneAxis = i;
+      break;
+    }
+  }
+  for (let i = firstNonOneAxis + 1; i < size.length; i++) {
+    if (begin[i] > 0 || size[i] !== shape[i]) {
+      return false;
+    }
+  }
+  return true;
+}
+function computeFlatOffset(begin, strides) {
+  let flatOffset = begin.length > 0 ? begin[begin.length - 1] : 1;
+  for (let i = 0; i < begin.length - 1; i++) {
+    flatOffset += begin[i] * strides[i];
+  }
+  return flatOffset;
+}
+function parseSliceParams(x, begin, size) {
+  let begin_;
+  const xRank = x.shape.length;
+  if (typeof begin === "number") {
+    begin_ = [begin, ...new Array(xRank - 1).fill(0)];
+  } else if (begin.length < xRank) {
+    begin_ = begin.concat(new Array(xRank - begin.length).fill(0));
+  } else {
+    begin_ = begin.slice();
+  }
+  begin_.forEach((d) => {
+    assert(d !== -1, () => "slice() does not support negative begin indexing.");
+  });
+  let size_;
+  if (size == null) {
+    size_ = new Array(xRank).fill(-1);
+  } else if (typeof size === "number") {
+    size_ = [size, ...new Array(xRank - 1).fill(-1)];
+  } else if (size.length < xRank) {
+    size_ = size.concat(new Array(xRank - size.length).fill(-1));
+  } else {
+    size_ = size;
+  }
+  size_ = size_.map((d, i) => {
+    if (d >= 0) {
+      return d;
+    } else {
+      assert(d === -1, () => `Negative size values should be exactly -1 but got ${d} for the slice() size at index ${i}.`);
+      return x.shape[i] - begin_[i];
+    }
+  });
+  return [begin_, size_];
+}
+function sliceInfo(xShape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {
+  let stridesNonNull;
+  if (strides == null) {
+    stridesNonNull = new Array(begin.length);
+    stridesNonNull.fill(1);
+  } else {
+    stridesNonNull = strides;
+  }
+  if (ellipsisMask != null && (ellipsisMask & ellipsisMask - 1) !== 0) {
+    throw new Error("Multiple ellipses in slice is not allowed.");
+  }
+  let ellipsisSeen = false;
+  const sparseSpec = {
+    dims: stridesNonNull.length,
+    numAddAxisAfterEllipsis: 0,
+    begin: begin.slice(),
+    end: end.slice(),
+    strides: stridesNonNull.slice(),
+    beginMask,
+    endMask,
+    ellipsisMask,
+    newAxisMask,
+    shrinkAxisMask
+  };
+  for (let i = 0; i < sparseSpec.dims; i++) {
+    if (ellipsisSeen && (1 << i & newAxisMask) !== 0) {
+      sparseSpec.numAddAxisAfterEllipsis++;
+    }
+    if (1 << i & ellipsisMask) {
+      ellipsisSeen = true;
+    }
+  }
+  if (!ellipsisSeen) {
+    sparseSpec.ellipsisMask |= 1 << sparseSpec.dims;
+    sparseSpec.dims++;
+  }
+  const denseSpec = {
+    dims: xShape.length,
+    beginMask: 0,
+    endMask: 0,
+    beginValid: false,
+    endValid: false
+  };
+  buildDenseSpec(sparseSpec, denseSpec);
+  let isIdentity = true;
+  let sliceDim0 = true;
+  let isSimpleSlice = true;
+  const processingShape = [];
+  const finalShape = [];
+  for (let i = 0; i < xShape.length; ++i) {
+    if (denseSpec.strides[i] === 0) {
+      throw Error(`strides[${i}] must be non-zero`);
+    }
+    const shrinkI = !!(denseSpec.shrinkAxisMask & 1 << i);
+    const dimI = xShape[i];
+    if (dimI === -1) {
+      processingShape.push(shrinkI ? 1 : -1);
+      continue;
+    }
+    const masks = [denseSpec.beginMask & 1 << i, denseSpec.endMask & 1 << i];
+    const validRange = [
+      denseSpec.strides[i] > 0 ? 0 : -1,
+      denseSpec.strides[i] > 0 ? dimI : dimI - 1
+    ];
+    if (shrinkI && denseSpec.strides[i] <= 0) {
+      throw Error("only stride 1 allowed on non-range indexing.");
+    }
+    isSimpleSlice = isSimpleSlice && denseSpec.strides[i] === 1;
+    const beginAndEndMasked = !!(denseSpec.beginMask & 1 << i && denseSpec.endMask & 1 << i);
+    if (denseSpec.beginValid && denseSpec.endValid) {
+      if (shrinkI) {
+        const xFwd = denseSpec.begin[i] < 0 ? dimI + denseSpec.begin[i] : denseSpec.begin[i];
+        denseSpec.begin[i] = xFwd;
+        denseSpec.end[i] = denseSpec.begin[i] + 1;
+        if (xFwd < 0 || xFwd >= dimI) {
+          throw Error(`slice index ${denseSpec.begin[i]} of dimension ${i} out of bounds.`);
+        }
+      } else {
+        denseSpec.begin[i] = canonical(denseSpec.begin[i], 0, denseSpec.strides[i], dimI, masks, validRange);
+        denseSpec.end[i] = canonical(denseSpec.end[i], 1, denseSpec.strides[i], dimI, masks, validRange);
+      }
+      const takeAllInDimension = denseSpec.strides[i] === 1 && denseSpec.begin[i] === 0 && denseSpec.end[i] === dimI;
+      isIdentity = isIdentity && takeAllInDimension;
+      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || takeAllInDimension);
+    } else {
+      isIdentity = isIdentity && (denseSpec.strides[i] === 1 && beginAndEndMasked);
+      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || beginAndEndMasked);
+    }
+    let intervalLength;
+    let knownInterval = false;
+    if (denseSpec.beginValid && denseSpec.endValid) {
+      intervalLength = denseSpec.end[i] - denseSpec.begin[i];
+      knownInterval = true;
+    } else if (shrinkI) {
+      intervalLength = 1;
+      knownInterval = true;
+    } else if (beginAndEndMasked) {
+      if (dimI >= 0) {
+        if (denseSpec.strides[i] < 0) {
+          intervalLength = -dimI;
+        } else {
+          intervalLength = dimI;
+        }
+        knownInterval = true;
+      }
+    }
+    if (knownInterval) {
+      let sizeI;
+      if (intervalLength === 0 || intervalLength < 0 !== denseSpec.strides[i] < 0) {
+        sizeI = 0;
+      } else {
+        sizeI = Math.trunc(intervalLength / denseSpec.strides[i]) + (intervalLength % denseSpec.strides[i] !== 0 ? 1 : 0);
+      }
+      processingShape.push(sizeI);
+    } else {
+      processingShape.push(-1);
+    }
+  }
+  for (let denseDim = 0; denseDim < denseSpec.finalShapeGatherIndices.length; ++denseDim) {
+    const gatherIndex = denseSpec.finalShapeGatherIndices[denseDim];
+    if (gatherIndex >= 0) {
+      finalShape.push(processingShape[gatherIndex]);
+    } else if (gatherIndex === NEW_AXIS) {
+      finalShape.push(1);
+    }
+  }
+  const finalShapeSparse = finalShape.filter((dim, i) => denseSpec.finalShapeGatherIndices[i] !== NEW_AXIS);
+  return {
+    finalShapeSparse,
+    finalShape,
+    isIdentity,
+    sliceDim0,
+    isSimpleSlice,
+    begin: denseSpec.begin,
+    end: denseSpec.end,
+    strides: denseSpec.strides
+  };
+}
+function buildDenseSpec(sparse2, dense2) {
+  dense2.beginMask = 0;
+  dense2.endMask = 0;
+  dense2.shrinkAxisMask = 0;
+  let fullIndex = 0;
+  dense2.beginValid = sparse2.begin != null;
+  dense2.endValid = sparse2.end != null;
+  dense2.begin = new Array(dense2.dims);
+  dense2.end = new Array(dense2.dims);
+  dense2.strides = new Array(dense2.dims);
+  dense2.finalShapeGatherIndices = [];
+  dense2.finalShapeGatherIndicesSparse = [];
+  dense2.inputShapeGatherIndicesSparse = new Array(dense2.dims);
+  for (let i = 0; i < sparse2.dims; i++) {
+    if (1 << i & sparse2.ellipsisMask) {
+      const nextIndex = Math.min(dense2.dims - (sparse2.dims - i) + 1 + sparse2.numAddAxisAfterEllipsis, dense2.dims);
+      for (; fullIndex < nextIndex; fullIndex++) {
+        dense2.begin[fullIndex] = 0;
+        dense2.end[fullIndex] = 0;
+        dense2.strides[fullIndex] = 1;
+        dense2.beginMask |= 1 << fullIndex;
+        dense2.endMask |= 1 << fullIndex;
+        dense2.finalShapeGatherIndices.push(fullIndex);
+        dense2.finalShapeGatherIndicesSparse.push(-1);
+        dense2.inputShapeGatherIndicesSparse[fullIndex] = i;
+      }
+    } else if (1 << i & sparse2.newAxisMask) {
+      dense2.finalShapeGatherIndices.push(NEW_AXIS);
+      dense2.finalShapeGatherIndicesSparse.push(-1);
+    } else {
+      if (fullIndex === dense2.begin.length) {
+        throw Error(`Index out of range using input dim ${fullIndex}; input has only ${dense2.dims} dims, ${dense2.begin.length}.`);
+      }
+      if (sparse2.begin != null) {
+        dense2.begin[fullIndex] = sparse2.begin[i];
+      }
+      if (sparse2.end != null) {
+        dense2.end[fullIndex] = sparse2.end[i];
+      }
+      dense2.strides[fullIndex] = sparse2.strides[i];
+      if (sparse2.beginMask & 1 << i) {
+        dense2.beginMask |= 1 << fullIndex;
+      }
+      if (sparse2.endMask & 1 << i) {
+        dense2.endMask |= 1 << fullIndex;
+      }
+      if (sparse2.shrinkAxisMask & 1 << i) {
+        dense2.finalShapeGatherIndices.push(SHRINK_AXIS);
+        dense2.finalShapeGatherIndicesSparse.push(-1);
+        dense2.shrinkAxisMask |= 1 << fullIndex;
+      } else {
+        dense2.finalShapeGatherIndices.push(fullIndex);
+        dense2.finalShapeGatherIndicesSparse.push(i);
+      }
+      dense2.inputShapeGatherIndicesSparse[fullIndex] = i;
+      fullIndex++;
+    }
+  }
+}
+function canonical(x, c, strideI, dimI, masks, validRange) {
+  if (masks[c]) {
+    return strideI > 0 ? validRange[c] : validRange[c + 1 & 1];
+  } else {
+    const xFwd = x < 0 ? dimI + x : x;
+    return xFwd < validRange[0] ? validRange[0] : xFwd > validRange[1] ? validRange[1] : xFwd;
+  }
+}
+var serialization_exports = {};
+__export2(serialization_exports, {
+  Serializable: () => Serializable,
+  SerializationMap: () => SerializationMap,
+  registerClass: () => registerClass
+});
+var Serializable = class {
+  getClassName() {
+    return this.constructor.className;
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+var SerializationMap = class {
+  constructor() {
+    this.classNameMap = {};
+  }
+  static getMap() {
+    if (SerializationMap.instance == null) {
+      SerializationMap.instance = new SerializationMap();
+    }
+    return SerializationMap.instance;
+  }
+  static register(cls) {
+    SerializationMap.getMap().classNameMap[cls.className] = [cls, cls.fromConfig];
+  }
+};
+function registerClass(cls) {
+  assert(cls.className != null, () => `Class being registered does not have the static className property defined.`);
+  assert(typeof cls.className === "string", () => `className is required to be a string, but got type ` + typeof cls.className);
+  assert(cls.className.length > 0, () => `Class being registered has an empty-string as its className, which is disallowed.`);
+  SerializationMap.register(cls);
+}
+var test_util_exports = {};
+__export2(test_util_exports, {
+  TEST_EPSILON_FLOAT16: () => TEST_EPSILON_FLOAT16,
+  createVideoElement: () => createVideoElement,
+  encodeStrings: () => encodeStrings,
+  expectArrayBuffersEqual: () => expectArrayBuffersEqual,
+  expectArraysClose: () => expectArraysClose,
+  expectArraysEqual: () => expectArraysEqual,
+  expectNumbersClose: () => expectNumbersClose,
+  expectPromiseToFail: () => expectPromiseToFail,
+  expectValuesInRange: () => expectValuesInRange,
+  play: () => play,
+  testEpsilon: () => testEpsilon
+});
+var TEST_EPSILON_FLOAT32 = 1e-3;
+var TEST_EPSILON_FLOAT16 = 0.1;
+function expectArraysClose(actual, expected, epsilon32) {
+  if (epsilon32 == null) {
+    epsilon32 = testEpsilon();
+  }
+  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, epsilon32));
+}
+function testEpsilon() {
+  return ENGINE.backend.floatPrecision() === 32 ? TEST_EPSILON_FLOAT32 : TEST_EPSILON_FLOAT16;
+}
+function expectArraysPredicate(actual, expected, predicate) {
+  let checkClassType = true;
+  if (isTypedArray(actual) || isTypedArray(expected)) {
+    checkClassType = false;
+  }
+  if (isTypedArray(actual) && isTypedArray(expected)) {
+    checkClassType = true;
+  }
+  if (checkClassType) {
+    const aType = actual.constructor.name;
+    const bType = expected.constructor.name;
+    if (aType !== bType) {
+      throw new Error(`Arrays are of different type. Actual: ${aType}. Expected: ${bType}`);
+    }
+  }
+  if (Array.isArray(actual) && Array.isArray(expected)) {
+    const actualShape = inferShape(actual);
+    const expectedShape = inferShape(expected);
+    if (!arraysEqual(actualShape, expectedShape)) {
+      throw new Error(`Arrays have different shapes. Actual: [${actualShape}]. Expected: [${expectedShape}]`);
+    }
+  }
+  const actualFlat = isTypedArray(actual) ? actual : flatten(actual);
+  const expectedFlat = isTypedArray(expected) ? expected : flatten(expected);
+  if (actualFlat.length !== expectedFlat.length) {
+    throw new Error(`Arrays have different lengths actual: ${actualFlat.length} vs expected: ${expectedFlat.length}.
+Actual:   ${actualFlat}.
+Expected: ${expectedFlat}.`);
+  }
+  for (let i = 0; i < expectedFlat.length; ++i) {
+    const a = actualFlat[i];
+    const e = expectedFlat[i];
+    if (!predicate(a, e)) {
+      throw new Error(`Arrays differ: actual[${i}] = ${a}, expected[${i}] = ${e}.
+Actual:   ${actualFlat}.
+Expected: ${expectedFlat}.`);
+    }
+  }
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function expectPromiseToFail(fn, done) {
+  fn().then(() => done.fail(), () => done());
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function expectArraysEqual(actual, expected) {
+  const exp4 = typeof expected === "string" || typeof expected === "number" || typeof expected === "boolean" ? [expected] : expected;
+  if (isString(actual) || isString(actual[0]) || isString(expected) || isString(expected[0])) {
+    return expectArraysPredicate(actual, exp4, (a, b) => a == b);
+  }
+  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, 0));
+}
+function expectNumbersClose(a, e, epsilon32) {
+  if (epsilon32 == null) {
+    epsilon32 = testEpsilon();
+  }
+  if (!areClose(a, e, epsilon32)) {
+    throw new Error(`Numbers differ: actual === ${a}, expected === ${e}`);
+  }
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function areClose(a, e, epsilon32) {
+  if (!isFinite(a) && !isFinite(e)) {
+    return true;
+  }
+  if (isNaN(a) || isNaN(e) || Math.abs(a - e) > epsilon32) {
+    return false;
+  }
+  return true;
+}
+function expectValuesInRange(actual, low, high) {
+  for (let i = 0; i < actual.length; i++) {
+    if (actual[i] < low || actual[i] > high) {
+      throw new Error(`Value out of range:${actual[i]} low: ${low}, high: ${high}`);
+    }
+  }
+}
+function expectArrayBuffersEqual(actual, expected) {
+  const actualArray = new Float32Array(actual);
+  const expectedArray = new Float32Array(expected);
+  if (actualArray.length !== expectedArray.length) {
+    throw new Error(`Expected ArrayBuffer to be of length ${expectedArray.length}, but it was ${actualArray.length}`);
+  }
+  for (let i = 0; i < expectedArray.length; i++) {
+    if (actualArray[i] !== expectedArray[i]) {
+      throw new Error(`Expected ArrayBuffer value at ${i} to be ${expectedArray[i]} but got ${actualArray[i]} instead`);
+    }
+  }
+}
+function encodeStrings(a) {
+  for (let i = 0; i < a.length; i++) {
+    const val = a[i];
+    if (Array.isArray(val)) {
+      encodeStrings(val);
+    } else {
+      a[i] = encodeString(val);
+    }
+  }
+  return a;
+}
+function createVideoElement(source) {
+  const video = document.createElement("video");
+  if ("playsInline" in video) {
+    video.playsInline = true;
+  }
+  video.muted = true;
+  video.loop = true;
+  video.style.position = "fixed";
+  video.style.left = "0px";
+  video.style.top = "0px";
+  video.preload = "auto";
+  video.appendChild(source);
+  return new Promise((resolve) => {
+    video.addEventListener("loadeddata", (_) => resolve(video));
+    video.load();
+  });
+}
+async function play(video) {
+  await video.play();
+  if ("requestVideoFrameCallback" in video) {
+    await new Promise((resolve) => {
+      video.requestVideoFrameCallback(resolve);
+    });
+  }
+}
+var version = "4.0.0";
+function add_(a, b) {
+  let $a = convertToTensor(a, "a", "add");
+  let $b = convertToTensor(b, "b", "add");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Add, inputs);
+}
+var add2 = op({ add_ });
+function floorDiv_(a, b) {
+  let $a = convertToTensor(a, "a", "floorDiv");
+  let $b = convertToTensor(b, "b", "floorDiv");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(FloorDiv, inputs);
+}
+var floorDiv = op({ floorDiv_ });
+function div_(a, b) {
+  let $a = convertToTensor(a, "a", "div");
+  let $b = convertToTensor(b, "b", "div");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "int32" && $b.dtype === "int32") {
+    return floorDiv($a, $b);
+  }
+  const inputs = { a: $a, b: $b };
+  const attrs = {};
+  return ENGINE.runKernel(RealDiv, inputs, attrs);
+}
+var div = op({ div_ });
+function mul_(a, b) {
+  let $a = convertToTensor(a, "a", "mul");
+  let $b = convertToTensor(b, "b", "mul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Multiply, inputs);
+}
+var mul = op({ mul_ });
+function abs_(x) {
+  const $x = convertToTensor(x, "x", "abs");
+  if ($x.dtype === "complex64") {
+    const inputs = { x: $x };
+    return ENGINE.runKernel(ComplexAbs, inputs);
+  } else {
+    const inputs = { x: $x };
+    return ENGINE.runKernel(Abs, inputs);
+  }
+}
+var abs = op({ abs_ });
+function acos_(x) {
+  const $x = convertToTensor(x, "x", "acos");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Acos, inputs);
+}
+var acos = op({ acos_ });
+function acosh_(x) {
+  const $x = convertToTensor(x, "x", "acosh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Acosh, inputs);
+}
+var acosh = op({ acosh_ });
+function addN_(tensors) {
+  assert(Array.isArray(tensors), () => "The argument passed to tf.addN() must be a list of tensors");
+  assert(tensors.length >= 1, () => `Must pass at least one tensor to tf.addN(), but got ${tensors.length}`);
+  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, "addN"));
+  const firstTensor = $tensors[0];
+  $tensors.forEach((t) => {
+    if (t.dtype !== firstTensor.dtype) {
+      throw new Error("All tensors passed to tf.addN() must have the same dtype");
+    }
+  });
+  $tensors.forEach((t) => {
+    if (!arraysEqual(t.shape, firstTensor.shape)) {
+      throw new Error("All tensors passed to tf.addN() must have the same shape");
+    }
+  });
+  const inputs = $tensors;
+  return ENGINE.runKernel(AddN, inputs);
+}
+var addN = op({ addN_ });
+function all_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "all", "bool");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(All, inputs, attrs);
+}
+var all = op({ all_ });
+function any_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "any", "bool");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Any, inputs, attrs);
+}
+var any = op({ any_ });
+function argMax_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "argMax");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(ArgMax, inputs, attrs);
+}
+var argMax = op({ argMax_ });
+function argMin_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "argMin");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(ArgMin, inputs, attrs);
+}
+var argMin = op({ argMin_ });
+function asin_(x) {
+  const $x = convertToTensor(x, "x", "asin");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Asin, inputs);
+}
+var asin = op({ asin_ });
+function asinh_(x) {
+  const $x = convertToTensor(x, "x", "asinh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Asinh, inputs);
+}
+var asinh = op({ asinh_ });
+function atan_(x) {
+  const $x = convertToTensor(x, "x", "atan");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Atan, inputs);
+}
+var atan = op({ atan_ });
+function atan2_(a, b) {
+  let $a = convertToTensor(a, "a", "atan2");
+  let $b = convertToTensor(b, "b", "atan2");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Atan2, inputs);
+}
+var atan2 = op({ atan2_ });
+function atanh_(x) {
+  const $x = convertToTensor(x, "x", "atanh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Atanh, inputs);
+}
+var atanh = op({ atanh_ });
+function computeDilation2DInfo(inputShape, filterShape, strides, pad3, dataFormat = "NHWC", dilations) {
+  const inputChannels = inputShape[3];
+  const $filterShape = [...filterShape, inputChannels];
+  const $dataFormat = convertConv2DDataFormat(dataFormat);
+  return computeConv2DInfo(inputShape, $filterShape, strides, dilations, pad3, null, null, $dataFormat);
+}
+function computePool2DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = "channelsLast") {
+  const [filterHeight, filterWidth] = parseTupleParam(filterSize);
+  let filterShape;
+  if (dataFormat === "channelsLast") {
+    filterShape = [filterHeight, filterWidth, inShape[3], inShape[3]];
+  } else if (dataFormat === "channelsFirst") {
+    filterShape = [filterHeight, filterWidth, inShape[1], inShape[1]];
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  return computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, false, dataFormat);
+}
+function computePool3DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = "NDHWC") {
+  const [filterDepth, filterHeight, filterWidth] = parse3TupleParam(filterSize);
+  let filterShape;
+  let $dataFormat;
+  if (dataFormat === "NDHWC") {
+    $dataFormat = "channelsLast";
+    filterShape = [filterDepth, filterHeight, filterWidth, inShape[4], inShape[4]];
+  } else if (dataFormat === "NCDHW") {
+    $dataFormat = "channelsFirst";
+    filterShape = [filterDepth, filterHeight, filterWidth, inShape[1], inShape[1]];
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  return computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, false, $dataFormat, roundingMode);
+}
+function computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, depthwise = false, dataFormat = "channelsLast") {
+  let [batchSize, inHeight, inWidth, inChannels] = [-1, -1, -1, -1];
+  if (dataFormat === "channelsLast") {
+    [batchSize, inHeight, inWidth, inChannels] = inShape;
+  } else if (dataFormat === "channelsFirst") {
+    [batchSize, inChannels, inHeight, inWidth] = inShape;
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  const [filterHeight, filterWidth, , filterChannels] = filterShape;
+  const [strideHeight, strideWidth] = parseTupleParam(strides);
+  const [dilationHeight, dilationWidth] = parseTupleParam(dilations);
+  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);
+  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);
+  const { padInfo, outHeight, outWidth } = getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, effectiveFilterHeight, effectiveFilterWidth, roundingMode, dataFormat);
+  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;
+  let outShape;
+  if (dataFormat === "channelsFirst") {
+    outShape = [batchSize, outChannels, outHeight, outWidth];
+  } else if (dataFormat === "channelsLast") {
+    outShape = [batchSize, outHeight, outWidth, outChannels];
+  }
+  return {
+    batchSize,
+    dataFormat,
+    inHeight,
+    inWidth,
+    inChannels,
+    outHeight,
+    outWidth,
+    outChannels,
+    padInfo,
+    strideHeight,
+    strideWidth,
+    filterHeight,
+    filterWidth,
+    effectiveFilterHeight,
+    effectiveFilterWidth,
+    dilationHeight,
+    dilationWidth,
+    inShape,
+    outShape,
+    filterShape
+  };
+}
+function computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, depthwise = false, dataFormat = "channelsLast", roundingMode) {
+  let [batchSize, inDepth, inHeight, inWidth, inChannels] = [-1, -1, -1, -1, -1];
+  if (dataFormat === "channelsLast") {
+    [batchSize, inDepth, inHeight, inWidth, inChannels] = inShape;
+  } else if (dataFormat === "channelsFirst") {
+    [batchSize, inChannels, inDepth, inHeight, inWidth] = inShape;
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  const [filterDepth, filterHeight, filterWidth, , filterChannels] = filterShape;
+  const [strideDepth, strideHeight, strideWidth] = parse3TupleParam(strides);
+  const [dilationDepth, dilationHeight, dilationWidth] = parse3TupleParam(dilations);
+  const effectiveFilterDepth = getEffectiveFilterSize(filterDepth, dilationDepth);
+  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);
+  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);
+  const { padInfo, outDepth, outHeight, outWidth } = get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, effectiveFilterDepth, effectiveFilterHeight, effectiveFilterWidth, roundingMode);
+  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;
+  let outShape;
+  if (dataFormat === "channelsFirst") {
+    outShape = [batchSize, outChannels, outDepth, outHeight, outWidth];
+  } else if (dataFormat === "channelsLast") {
+    outShape = [batchSize, outDepth, outHeight, outWidth, outChannels];
+  }
+  return {
+    batchSize,
+    dataFormat,
+    inDepth,
+    inHeight,
+    inWidth,
+    inChannels,
+    outDepth,
+    outHeight,
+    outWidth,
+    outChannels,
+    padInfo,
+    strideDepth,
+    strideHeight,
+    strideWidth,
+    filterDepth,
+    filterHeight,
+    filterWidth,
+    effectiveFilterDepth,
+    effectiveFilterHeight,
+    effectiveFilterWidth,
+    dilationDepth,
+    dilationHeight,
+    dilationWidth,
+    inShape,
+    outShape,
+    filterShape
+  };
+}
+function computeOutputShape2D(inShape, fieldSize, stride, zeroPad, roundingMode) {
+  if (zeroPad == null) {
+    zeroPad = computeDefaultPad(inShape, fieldSize, stride);
+  }
+  const inputRows = inShape[0];
+  const inputCols = inShape[1];
+  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  return [outputRows, outputCols];
+}
+function computeOutputShape4D(inShape, fieldSize, outChannels, stride, zeroPad, roundingMode) {
+  if (zeroPad == null) {
+    zeroPad = computeDefaultPad(inShape, fieldSize, stride);
+  }
+  const inputDepth = inShape[0];
+  const inputRows = inShape[1];
+  const inputCols = inShape[2];
+  const outputDepths = round((inputDepth - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  return [outputDepths, outputRows, outputCols, outChannels];
+}
+function computeDefaultPad(inputShape, fieldSize, stride, dilation = 1) {
+  const effectiveFieldSize = getEffectiveFilterSize(fieldSize, dilation);
+  return Math.floor((inputShape[0] * (stride - 1) - stride + effectiveFieldSize) / 2);
+}
+function parseTupleParam(param) {
+  if (typeof param === "number") {
+    return [param, param, param];
+  }
+  if (param.length === 2) {
+    return [param[0], param[1], 1];
+  }
+  return param;
+}
+function parse3TupleParam(param) {
+  return typeof param === "number" ? [param, param, param] : param;
+}
+function getEffectiveFilterSize(filterSize, dilation) {
+  if (dilation <= 1) {
+    return filterSize;
+  }
+  return filterSize + (filterSize - 1) * (dilation - 1);
+}
+function getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, filterHeight, filterWidth, roundingMode, dataFormat) {
+  let padInfo;
+  let outHeight;
+  let outWidth;
+  if (typeof pad3 === "number") {
+    const padType = pad3 === 0 ? "VALID" : "NUMBER";
+    padInfo = { top: pad3, bottom: pad3, left: pad3, right: pad3, type: padType };
+    const outShape = computeOutputShape2D([inHeight, inWidth], filterHeight, strideHeight, pad3, roundingMode);
+    outHeight = outShape[0];
+    outWidth = outShape[1];
+  } else if (pad3 === "same") {
+    outHeight = Math.ceil(inHeight / strideHeight);
+    outWidth = Math.ceil(inWidth / strideWidth);
+    const padAlongHeight = Math.max(0, (outHeight - 1) * strideHeight + filterHeight - inHeight);
+    const padAlongWidth = Math.max(0, (outWidth - 1) * strideWidth + filterWidth - inWidth);
+    const top = Math.floor(padAlongHeight / 2);
+    const bottom = padAlongHeight - top;
+    const left = Math.floor(padAlongWidth / 2);
+    const right = padAlongWidth - left;
+    padInfo = { top, bottom, left, right, type: "SAME" };
+  } else if (pad3 === "valid") {
+    padInfo = { top: 0, bottom: 0, left: 0, right: 0, type: "VALID" };
+    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);
+    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);
+  } else if (typeof pad3 === "object") {
+    const top = dataFormat === "channelsLast" ? pad3[1][0] : pad3[2][0];
+    const bottom = dataFormat === "channelsLast" ? pad3[1][1] : pad3[2][1];
+    const left = dataFormat === "channelsLast" ? pad3[2][0] : pad3[3][0];
+    const right = dataFormat === "channelsLast" ? pad3[2][1] : pad3[3][1];
+    const padType = top === 0 && bottom === 0 && left === 0 && right === 0 ? "VALID" : "EXPLICIT";
+    padInfo = { top, bottom, left, right, type: padType };
+    outHeight = round((inHeight - filterHeight + top + bottom) / strideHeight + 1, roundingMode);
+    outWidth = round((inWidth - filterWidth + left + right) / strideWidth + 1, roundingMode);
+  } else {
+    throw Error(`Unknown padding parameter: ${pad3}`);
+  }
+  return { padInfo, outHeight, outWidth };
+}
+function get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, filterDepth, filterHeight, filterWidth, roundingMode) {
+  let padInfo;
+  let outDepth;
+  let outHeight;
+  let outWidth;
+  if (typeof pad3 === "number") {
+    const padType = pad3 === 0 ? "VALID" : "NUMBER";
+    padInfo = {
+      top: pad3,
+      bottom: pad3,
+      left: pad3,
+      right: pad3,
+      front: pad3,
+      back: pad3,
+      type: padType
+    };
+    const outShape = computeOutputShape4D([inDepth, inHeight, inWidth, 1], filterDepth, 1, strideDepth, pad3, roundingMode);
+    outDepth = outShape[0];
+    outHeight = outShape[1];
+    outWidth = outShape[2];
+  } else if (pad3 === "same") {
+    outDepth = Math.ceil(inDepth / strideDepth);
+    outHeight = Math.ceil(inHeight / strideHeight);
+    outWidth = Math.ceil(inWidth / strideWidth);
+    const padAlongDepth = (outDepth - 1) * strideDepth + filterDepth - inDepth;
+    const padAlongHeight = (outHeight - 1) * strideHeight + filterHeight - inHeight;
+    const padAlongWidth = (outWidth - 1) * strideWidth + filterWidth - inWidth;
+    const front = Math.floor(padAlongDepth / 2);
+    const back = padAlongDepth - front;
+    const top = Math.floor(padAlongHeight / 2);
+    const bottom = padAlongHeight - top;
+    const left = Math.floor(padAlongWidth / 2);
+    const right = padAlongWidth - left;
+    padInfo = { top, bottom, left, right, front, back, type: "SAME" };
+  } else if (pad3 === "valid") {
+    padInfo = {
+      top: 0,
+      bottom: 0,
+      left: 0,
+      right: 0,
+      front: 0,
+      back: 0,
+      type: "VALID"
+    };
+    outDepth = Math.ceil((inDepth - filterDepth + 1) / strideDepth);
+    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);
+    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);
+  } else {
+    throw Error(`Unknown padding parameter: ${pad3}`);
+  }
+  return { padInfo, outDepth, outHeight, outWidth };
+}
+function round(value, roundingMode) {
+  if (!roundingMode) {
+    return Math.trunc(value);
+  }
+  switch (roundingMode) {
+    case "round":
+      return Math.round(value);
+    case "ceil":
+      return Math.ceil(value);
+    case "floor":
+      return Math.floor(value);
+    default:
+      throw new Error(`Unknown roundingMode ${roundingMode}`);
+  }
+}
+function tupleValuesAreOne(param) {
+  const [dimA, dimB, dimC] = parseTupleParam(param);
+  return dimA === 1 && dimB === 1 && dimC === 1;
+}
+function eitherStridesOrDilationsAreOne(strides, dilations) {
+  return tupleValuesAreOne(strides) || tupleValuesAreOne(dilations);
+}
+function convertConv2DDataFormat(dataFormat) {
+  if (dataFormat === "NHWC") {
+    return "channelsLast";
+  } else if (dataFormat === "NCHW") {
+    return "channelsFirst";
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+}
+function checkPadOnDimRoundingMode(opDesc, pad3, dimRoundingMode) {
+  if (dimRoundingMode != null) {
+    if (typeof pad3 === "string") {
+      throw Error(`Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);
+    } else if (typeof pad3 === "number") {
+      assert(isInt(pad3), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);
+    } else if (typeof pad3 === "object") {
+      pad3.forEach((p2) => {
+        p2.forEach((v) => {
+          assert(isInt(v), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${v}.`);
+        });
+      });
+    } else {
+      throw Error(`Error in ${opDesc}: Unknown padding parameter: ${pad3}`);
+    }
+  }
+}
+function reshape_(x, shape) {
+  const $x = convertToTensor(x, "x", "reshape", "string_or_numeric");
+  const inputs = { x: $x };
+  const attrs = { shape };
+  return ENGINE.runKernel(Reshape, inputs, attrs);
+}
+var reshape = op({ reshape_ });
+function avgPool_(x, filterSize, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "avgPool", "float32");
+  const dilations = 1;
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in avgPool: x must be rank 4 but got rank ${x4D.rank}.`);
+  checkPadOnDimRoundingMode("avgPool", pad3, dimRoundingMode);
+  const inputs = { x: x4D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  let res = ENGINE.runKernel(AvgPool, inputs, attrs);
+  res = cast(res, $x.dtype);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var avgPool = op({ avgPool_ });
+function avgPool3d_(x, filterSize, strides, pad3, dimRoundingMode, dataFormat = "NDHWC") {
+  const $x = convertToTensor(x, "x", "avgPool3d", "float32");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in avgPool3d: x must be rank 5 but got rank ${x5D.rank}.`);
+  assert(dataFormat === "NDHWC", () => `Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  checkPadOnDimRoundingMode("avgPool3d", pad3, dimRoundingMode);
+  const inputs = { x: x5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };
+  let res = ENGINE.runKernel(AvgPool3D, inputs, attrs);
+  res = cast(res, x5D.dtype);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var avgPool3d = op({ avgPool3d_ });
+function concat_(tensors, axis = 0) {
+  assert(tensors.length >= 1, () => "Pass at least one tensor to concat");
+  const $tensors = convertToTensorArray(tensors, "tensors", "concat", "string_or_numeric");
+  if ($tensors[0].dtype === "complex64") {
+    $tensors.forEach((tensor2) => {
+      if (tensor2.dtype !== "complex64") {
+        throw new Error(`Cannot concatenate complex64 tensors with a tensor
+          with dtype ${tensor2.dtype}. `);
+      }
+    });
+  }
+  if ($tensors.length === 1) {
+    return clone($tensors[0]);
+  }
+  const inputs = $tensors;
+  const attr = { axis };
+  return ENGINE.runKernel(Concat, inputs, attr);
+}
+var concat = op({ concat_ });
+function sigmoid_(x) {
+  const $x = convertToTensor(x, "x", "sigmoid", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sigmoid, inputs);
+}
+var sigmoid = op({ sigmoid_ });
+function slice_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice", "string_or_numeric");
+  if ($x.rank === 0) {
+    throw new Error("Slicing scalar is not possible");
+  }
+  const inputs = { x: $x };
+  const attrs = { begin, size };
+  return ENGINE.runKernel(Slice, inputs, attrs);
+}
+var slice = op({ slice_ });
+function tanh_(x) {
+  const $x = convertToTensor(x, "x", "tanh", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Tanh, inputs);
+}
+var tanh2 = op({ tanh_ });
+function basicLSTMCell_(forgetBias, lstmKernel, lstmBias, data, c, h) {
+  const $forgetBias = convertToTensor(forgetBias, "forgetBias", "basicLSTMCell");
+  const $lstmKernel = convertToTensor(lstmKernel, "lstmKernel", "basicLSTMCell");
+  const $lstmBias = convertToTensor(lstmBias, "lstmBias", "basicLSTMCell");
+  const $data = convertToTensor(data, "data", "basicLSTMCell");
+  const $c = convertToTensor(c, "c", "basicLSTMCell");
+  const $h = convertToTensor(h, "h", "basicLSTMCell");
+  const combined = concat([$data, $h], 1);
+  const weighted = matMul(combined, $lstmKernel);
+  const res = add2(weighted, $lstmBias);
+  const batchSize = res.shape[0];
+  const sliceCols = res.shape[1] / 4;
+  const sliceSize = [batchSize, sliceCols];
+  const i = slice(res, [0, 0], sliceSize);
+  const j = slice(res, [0, sliceCols], sliceSize);
+  const f = slice(res, [0, sliceCols * 2], sliceSize);
+  const o = slice(res, [0, sliceCols * 3], sliceSize);
+  const newC = add2(mul(sigmoid(i), tanh2(j)), mul($c, sigmoid(add2($forgetBias, f))));
+  const newH = mul(tanh2(newC), sigmoid(o));
+  return [newC, newH];
+}
+var basicLSTMCell = op({ basicLSTMCell_ });
+function batchToSpaceND_(x, blockShape, crops) {
+  const $x = convertToTensor(x, "x", "batchToSpaceND");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  assert($x.rank >= 1 + blockShape.length, () => `input rank is ${$x.rank} but should be > than blockShape.length ${blockShape.length}`);
+  assert(crops.length === blockShape.length, () => `crops.length is ${crops.length} but should be equal to blockShape.length  ${blockShape.length}`);
+  assert($x.shape[0] % prod5 === 0, () => `input tensor batch is ${$x.shape[0]} but is not divisible by the product of the elements of blockShape ${blockShape.join(" * ")} === ${prod5}`);
+  const inputs = { x: $x };
+  const attrs = { blockShape, crops };
+  return ENGINE.runKernel(BatchToSpaceND, inputs, attrs);
+}
+var batchToSpaceND = op({ batchToSpaceND_ });
+function xAs4D(x) {
+  let x4D;
+  if (x.rank === 0 || x.rank === 1) {
+    x4D = reshape(x, [1, 1, 1, x.size]);
+  } else if (x.rank === 2) {
+    x4D = reshape(x, [1, 1, x.shape[0], x.shape[1]]);
+  } else if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  } else {
+    x4D = x;
+  }
+  return x4D;
+}
+function batchNorm_(x, mean4, variance, offset, scale22, varianceEpsilon) {
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale22 != null) {
+    $scale = convertToTensor(scale22, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($mean.rank === $variance.rank, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  assert($offset == null || $mean.rank === $offset.rank, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  assert($scale == null || $mean.rank === $scale.rank, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  const x4D = xAs4D($x);
+  const inputs = {
+    x: x4D,
+    scale: $scale,
+    offset: $offset,
+    mean: $mean,
+    variance: $variance
+  };
+  const attrs = { varianceEpsilon };
+  const res = ENGINE.runKernel(FusedBatchNorm, inputs, attrs);
+  return reshape(res, $x.shape);
+}
+var batchNorm = op({ batchNorm_ });
+function batchNorm2d_(x, mean4, variance, offset, scale22, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale22 != null) {
+    $scale = convertToTensor(scale22, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 2, () => `Error in batchNorm2D: x must be rank 2 but got rank ${$x.rank}.`);
+  assert($mean.rank === 2 || $mean.rank === 1, () => `Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 2 || $variance.rank === 1, () => `Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 2 || $scale.rank === 1, () => `Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 2 || $offset.rank === 1, () => `Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm2d = op({ batchNorm2d_ });
+function batchNorm3d_(x, mean4, variance, offset, scale22, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale22 != null) {
+    $scale = convertToTensor(scale22, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 3, () => `Error in batchNorm3D: x must be rank 3 but got rank ${$x.rank}.`);
+  assert($mean.rank === 3 || $mean.rank === 1, () => `Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 3 || $variance.rank === 1, () => `Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 3 || $scale.rank === 1, () => `Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 3 || $offset.rank === 1, () => `Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm3d = op({ batchNorm3d_ });
+function batchNorm4d_(x, mean4, variance, offset, scale22, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale22 != null) {
+    $scale = convertToTensor(scale22, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 4, () => `Error in batchNorm4D: x must be rank 4 but got rank ${$x.rank}.`);
+  assert($mean.rank === 4 || $mean.rank === 1, () => `Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 4 || $variance.rank === 1, () => `Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 4 || $scale.rank === 1, () => `Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 4 || $offset.rank === 1, () => `Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm4d = op({ batchNorm4d_ });
+function bincount_(x, weights, size) {
+  const $x = convertToTensor(x, "x", "bincount");
+  const $weights = convertToTensor(weights, "weights", "bincount");
+  assert($x.dtype === "int32", () => `Error in bincount: input dtype must be int32, but got ${$x.dtype}`);
+  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);
+  assert($weights.size === $x.size || $weights.size === 0, () => `Error in bincount: weights must have the same size as input or0-length, but got input shape: ${$x.shape}, weights shape: ${$weights.shape}.`);
+  const inputs = { x: $x, weights: $weights };
+  const attrs = { size };
+  return ENGINE.runKernel(Bincount, inputs, attrs);
+}
+var bincount = op({ bincount_ });
+function broadcastArgs_(s0, s1) {
+  const shape1Input = convertToTensor(s0, "s0", "broadcastArgs", "int32");
+  const shape2Input = convertToTensor(s1, "s1", "broadcastArgs", "int32");
+  if (shape1Input.rank !== 1) {
+    throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${shape1Input.rank}`);
+  }
+  if (shape2Input.rank !== 1) {
+    throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${shape2Input.rank}`);
+  }
+  const inputs = { s0: shape1Input, s1: shape2Input };
+  return ENGINE.runKernel(BroadcastArgs, inputs);
+}
+var broadcastArgs = op({ broadcastArgs_ });
+function broadcastTo_(x, shape) {
+  let input2 = convertToTensor(x, "broadcastTo", "x");
+  const xShape = input2.shape;
+  if (shape.some((d) => !(d > 0) || d % 1 !== 0)) {
+    throw new Error(`broadcastTo(): Invalid broadcast shape [${shape}].`);
+  }
+  if (shape.length < input2.rank) {
+    throw new Error(`broadcastTo(): shape.length=${shape.length} < input.rank=${input2.rank}.`);
+  }
+  if (shape.length > input2.rank) {
+    const newShape = input2.shape.slice();
+    while (newShape.length < shape.length) {
+      newShape.unshift(1);
+    }
+    input2 = reshape(input2, newShape);
+  }
+  const inputShape = input2.shape;
+  const reps = Array.from(shape);
+  for (let i = shape.length - 1; i >= 0; i--) {
+    if (inputShape[i] === shape[i]) {
+      reps[i] = 1;
+    } else if (input2.shape[i] !== 1) {
+      throw new Error(`broadcastTo(): [${xShape}] cannot be broadcast to [${shape}].`);
+    }
+  }
+  const axes = reps.map((n, i) => n > 1 ? i : -1).filter((i) => i >= 0);
+  if (axes.length === 0) {
+    return clone(input2);
+  }
+  const inputs = { x: input2 };
+  const attrs = { reps };
+  return ENGINE.runKernel(Tile, inputs, attrs);
+}
+var broadcastTo = op({ broadcastTo_ });
+function ceil_(x) {
+  const $x = convertToTensor(x, "x", "ceil", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Ceil, inputs);
+}
+var ceil = op({ ceil_ });
+function fill(shape, value, dtype) {
+  const attrs = { shape, value, dtype };
+  return ENGINE.runKernel(Fill, {}, attrs);
+}
+function clipByValue_(x, clipValueMin, clipValueMax) {
+  const $x = convertToTensor(x, "x", "clipByValue");
+  assert(clipValueMin <= clipValueMax, () => `Error in clip: min (${clipValueMin}) must be less than or equal to max (${clipValueMax}).`);
+  if (clipValueMin === clipValueMax) {
+    return fill($x.shape, clipValueMin, $x.dtype);
+  }
+  const inputs = { x: $x };
+  const attrs = { clipValueMin, clipValueMax };
+  return ENGINE.runKernel(ClipByValue, inputs, attrs);
+}
+var clipByValue = op({ clipByValue_ });
+function concat1d_(tensors) {
+  return concat(tensors, 0);
+}
+var concat1d = op({ concat1d_ });
+function concat2d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat2d = op({ concat2d_ });
+function concat3d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat3d = op({ concat3d_ });
+function concat4d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat4d = op({ concat4d_ });
+function conv2d_(x, filter, strides, pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "conv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in conv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("conv2d", pad3, dimRoundingMode);
+  const inDepth = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert(inDepth === $filter.shape[2], () => `Error in conv2d: depth of input (${inDepth}) must match input depth for filter ${$filter.shape[2]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };
+  const res = ENGINE.runKernel(Conv2D, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var conv2d = op({ conv2d_ });
+function conv1d_(x, filter, stride, pad3, dataFormat = "NWC", dilation = 1, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv1d");
+  const $filter = convertToTensor(filter, "filter", "conv1d");
+  let x3D = $x;
+  let reshapedTo3D = false;
+  if ($x.rank === 2) {
+    reshapedTo3D = true;
+    x3D = reshape($x, [1, $x.shape[0], $x.shape[1]]);
+  }
+  assert(x3D.rank === 3, () => `Error in conv1d: input must be rank 3, but got rank ${x3D.rank}.`);
+  assert($filter.rank === 3, () => `Error in conv1d: filter must be rank 3, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("conv1d", pad3, dimRoundingMode);
+  assert(x3D.shape[2] === $filter.shape[1], () => `Error in conv1d: depth of input (${x3D.shape[2]}) must match input depth for filter ${$filter.shape[1]}.`);
+  assert(eitherStridesOrDilationsAreOne(stride, dilation), () => `Error in conv1D: Either stride or dilation must be 1. Got stride ${stride} and dilation '${dilation}'`);
+  assert(dataFormat === "NWC", () => `Error in conv1d: got dataFormat of ${dataFormat} but only NWC is currently supported.`);
+  const filter4D = reshape($filter, [1, $filter.shape[0], $filter.shape[1], $filter.shape[2]]);
+  const input4D = reshape(x3D, [x3D.shape[0], 1, x3D.shape[1], x3D.shape[2]]);
+  const strides = [1, stride];
+  const dilations = [1, dilation];
+  const conv2dDataFormat = "NHWC";
+  const res = conv2d(input4D, filter4D, strides, pad3, conv2dDataFormat, dilations, dimRoundingMode);
+  if (reshapedTo3D) {
+    return reshape(res, [res.shape[2], res.shape[3]]);
+  }
+  return reshape(res, [res.shape[0], res.shape[2], res.shape[3]]);
+}
+var conv1d = op({ conv1d_ });
+function conv2DBackpropInput_(xShape, dy, filter, strides, pad3, dataFormat = "NHWC", dimRoundingMode) {
+  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);
+  let xShape4D = xShape;
+  let dy4D = dy;
+  let reshapedTo4D = false;
+  if (dy.rank === 3) {
+    reshapedTo4D = true;
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+    xShape4D = [1, xShape[0], xShape[1], xShape[2]];
+  }
+  assert(xShape4D.length === 4, () => `Error in conv2dDerInput: inShape must be length 4, but got length ${xShape4D.length}.`);
+  assert(dy4D.rank === 4, () => `Error in conv2dDerInput: dy must be rank 4, but got rank ${dy4D.rank}`);
+  assert(filter.rank === 4, () => `Error in conv2dDerInput: filter must be rank 4, but got rank ${filter.rank}`);
+  const inDepth = dataFormat === "NHWC" ? xShape4D[3] : xShape4D[1];
+  const outDepth = dataFormat === "NHWC" ? dy4D.shape[3] : dy4D.shape[1];
+  assert(inDepth === filter.shape[2], () => `Error in conv2dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[2]}.`);
+  assert(outDepth === filter.shape[3], () => `Error in conv2dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[3]}.`);
+  checkPadOnDimRoundingMode("conv2dDerInput", pad3, dimRoundingMode);
+  const inputs = { dy: dy4D, filter };
+  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape: xShape4D };
+  const res = ENGINE.runKernel(Conv2DBackpropInput, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var conv2DBackpropInput = op({ conv2DBackpropInput_ });
+function conv2dTranspose_(x, filter, outputShape, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv2dTranspose");
+  const $filter = convertToTensor(filter, "filter", "conv2dTranspose");
+  return conv2DBackpropInput(outputShape, $x, $filter, strides, pad3, "NHWC", dimRoundingMode);
+}
+var conv2dTranspose = op({ conv2dTranspose_ });
+function conv3d_(x, filter, strides, pad3, dataFormat = "NDHWC", dilations = [1, 1, 1]) {
+  const $x = convertToTensor(x, "x", "conv3d");
+  const $filter = convertToTensor(filter, "filter", "conv3d");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in conv3d: input must be rank 5, but got rank ${x5D.rank}.`);
+  assert($filter.rank === 5, () => `Error in conv3d: filter must be rank 5, but got rank ${$filter.rank}.`);
+  assert(x5D.shape[4] === $filter.shape[3], () => `Error in conv3d: depth of input (${x5D.shape[4]}) must match input depth for filter ${$filter.shape[3]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv3D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  assert(dataFormat === "NDHWC", () => `Error in conv3d: got dataFormat of ${dataFormat} but only NDHWC is currently supported.`);
+  const inputs = { x: x5D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations };
+  const res = ENGINE.runKernel(Conv3D, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var conv3d = op({ conv3d_ });
+function conv3DBackpropInput_(xShape, dy, filter, strides, pad3) {
+  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);
+  let xShape5D = xShape;
+  let dy5D = dy;
+  let reshapedTo5D = false;
+  if (dy.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);
+    xShape5D = [1, xShape[0], xShape[1], xShape[2], xShape[3]];
+  }
+  const inDepth = xShape5D[4];
+  const outDepth = dy5D.shape[4];
+  assert(xShape5D.length === 5, () => `Error in conv3dDerInput: inShape must be length 5, but got length ${xShape5D.length}.`);
+  assert(dy5D.rank === 5, () => `Error in conv3dDerInput: dy must be rank 5, but got rank ${dy5D.rank}`);
+  assert(filter.rank === 5, () => `Error in conv3dDerInput: filter must be rank 5, but got rank ${filter.rank}`);
+  assert(inDepth === filter.shape[3], () => `Error in conv3dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[3]}.`);
+  assert(outDepth === filter.shape[4], () => `Error in conv3dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[4]}.`);
+  const inputs = { dy: dy5D, filter };
+  const attrs = { pad: pad3, strides, inputShape: xShape5D };
+  const res = ENGINE.runKernel(Conv3DBackpropInputV2, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var conv3DBackpropInput = op({ conv3DBackpropInput_ });
+function conv3dTranspose_(x, filter, outputShape, strides, pad3) {
+  const $x = convertToTensor(x, "x", "conv3dTranspose");
+  const $filter = convertToTensor(filter, "filter", "conv3dTranspose");
+  return conv3DBackpropInput(outputShape, $x, $filter, strides, pad3);
+}
+var conv3dTranspose = op({ conv3dTranspose_ });
+function cos_(x) {
+  const $x = convertToTensor(x, "x", "cos", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Cos, inputs);
+}
+var cos = op({ cos_ });
+function cosh_(x) {
+  const $x = convertToTensor(x, "x", "cosh", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Cosh, inputs);
+}
+var cosh = op({ cosh_ });
+function cumprod_(x, axis = 0, exclusive = false, reverse5 = false) {
+  const $x = convertToTensor(x, "x", "cumprod");
+  const inputs = { x: $x };
+  const attrs = { axis, exclusive, reverse: reverse5 };
+  return ENGINE.runKernel(Cumprod, inputs, attrs);
+}
+var cumprod = op({ cumprod_ });
+function cumsum_(x, axis = 0, exclusive = false, reverse5 = false) {
+  const $x = convertToTensor(x, "x", "cumsum");
+  const inputs = { x: $x };
+  const attrs = { axis, exclusive, reverse: reverse5 };
+  return ENGINE.runKernel(Cumsum, inputs, attrs);
+}
+var cumsum = op({ cumsum_ });
+function denseBincount_(x, weights, size, binaryOutput = false) {
+  const $x = convertToTensor(x, "x", "denseBincount");
+  const $weights = convertToTensor(weights, "weights", "denseBincount");
+  assert($x.dtype === "int32", () => `Error in denseBincount: input dtype must be int32, but got ${$x.dtype}`);
+  assert($x.rank <= 2, () => `Error in denseBincount: input must be at most rank 2, but got rank ${$x.rank}.`);
+  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);
+  assert($weights.size === $x.size || $weights.size === 0, () => `Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${$x.shape}, weights shape: ${$weights.shape}.`);
+  const inputs = { x: $x, weights: $weights };
+  const attrs = { size, binaryOutput };
+  return ENGINE.runKernel(DenseBincount, inputs, attrs);
+}
+var denseBincount = op({ denseBincount_ });
+function depthToSpace_(x, blockSize, dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "depthToSpace", "float32");
+  const inputHeight = dataFormat === "NHWC" ? $x.shape[1] : $x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? $x.shape[2] : $x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? $x.shape[3] : $x.shape[1];
+  assert(blockSize > 1, () => `blockSize should be > 1 for depthToSpace, but was: ${blockSize}`);
+  assert(inputHeight * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying
+    ${inputHeight} and ${blockSize}  for depthToSpace with input shape
+    ${$x.shape}`);
+  assert(inputWidth * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying
+    ${inputWidth} and ${blockSize} for depthToSpace with input shape
+        ${$x.shape}`);
+  assert(inputDepth % (blockSize * blockSize) === 0, () => `Dimension size must be evenly divisible by ${blockSize * blockSize} but is ${inputDepth} for depthToSpace with input shape ${$x.shape}`);
+  const inputs = { x: $x };
+  const attrs = { blockSize, dataFormat };
+  return ENGINE.runKernel(DepthToSpace, inputs, attrs);
+}
+var depthToSpace = op({ depthToSpace_ });
+function depthwiseConv2d_(x, filter, strides, pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "depthwiseConv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "depthwiseConv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  const inChannels = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert(inChannels === $filter.shape[2], () => `Error in depthwiseConv2d: number of input channels (${inChannels}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);
+  checkPadOnDimRoundingMode("depthwiseConv2d", pad3, dimRoundingMode);
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };
+  const res = ENGINE.runKernel(DepthwiseConv2dNative, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var depthwiseConv2d = op({ depthwiseConv2d_ });
+function diag_(x) {
+  const $x = convertToTensor(x, "x", "diag");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Diag, inputs);
+}
+var diag = op({ diag_ });
+function dilation2d_(x, filter, strides, pad3, dilations = [1, 1], dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "dilation2d");
+  const $filter = convertToTensor(filter, "filter", "dilation2d");
+  assert($x.rank === 3 || $x.rank === 4, () => `Error in dilation2d: input must be rank 3 or 4, but got rank ${$x.rank}.`);
+  assert($filter.rank === 3, () => `Error in dilation2d: filter must be rank 3, but got rank ${$filter.rank}.`);
+  assert(dataFormat === "NHWC", () => `Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+    reshapedTo4D = true;
+  }
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dilations };
+  const res = ENGINE.runKernel(Dilation2D, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var dilation2d = op({ dilation2d_ });
+function equal_(a, b) {
+  let $a = convertToTensor(a, "a", "equal", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "equal", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Equal, inputs);
+}
+var equal = op({ equal_ });
+function where_(condition, a, b) {
+  const $a = convertToTensor(a, "a", "where");
+  const $b = convertToTensor(b, "b", "where");
+  const $condition = convertToTensor(condition, "condition", "where", "bool");
+  const broadcastShape = assertAndGetBroadcastShape(assertAndGetBroadcastShape($condition.shape, $a.shape), $b.shape);
+  const $broadcastedCondition = broadcastTo($condition, broadcastShape);
+  const $broadcastedA = broadcastTo($a, broadcastShape);
+  const $broadcastedB = broadcastTo($b, broadcastShape);
+  const inputs = {
+    condition: $broadcastedCondition,
+    t: $broadcastedA,
+    e: $broadcastedB
+  };
+  return ENGINE.runKernel(Select, inputs);
+}
+var where = op({ where_ });
+function zerosLike_(x) {
+  const $x = convertToTensor(x, "x", "zerosLike");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(ZerosLike, inputs);
+}
+var zerosLike = op({ zerosLike_ });
+function divNoNan_(a, b) {
+  let $a = convertToTensor(a, "a", "div");
+  let $b = convertToTensor(b, "b", "div");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const divResult = div($a, $b);
+  const zeros4 = zerosLike(divResult);
+  const bEqualsZero = equal($b, zeros4);
+  return where(bEqualsZero, zeros4, divResult);
+}
+var divNoNan = op({ divNoNan_ });
+function dot_(t1, t2) {
+  const $t1 = convertToTensor(t1, "t1", "dot");
+  const $t2 = convertToTensor(t2, "t2", "dot");
+  assert(($t1.rank === 1 || $t1.rank === 2) && ($t2.rank === 1 || $t2.rank === 2), () => `Error in dot: inputs must all be rank 1 or 2, but got ranks ${$t1.rank} and ${$t2.rank}.`);
+  const t1Inner = $t1.rank === 1 ? $t1.size : $t1.shape[1];
+  const t2Inner = $t2.rank === 1 ? $t2.size : $t2.shape[0];
+  assert(t1Inner === t2Inner, () => `Error in dot: inner dimensions of inputs must match, but got ${t1Inner} and ${t2Inner}.`);
+  if ($t1.rank === 1 && $t2.rank === 1) {
+    const t12D = reshape($t1, [1, -1]);
+    const t22D = reshape($t2, [-1, 1]);
+    const t1t2 = matMul(t12D, t22D);
+    return reshape(t1t2, []);
+  } else if ($t1.rank === 1 && $t2.rank === 2) {
+    const t12D = reshape($t1, [1, -1]);
+    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);
+    const t1t2 = matMul(t12D, t22D);
+    return reshape(t1t2, [t1t2.size]);
+  } else if ($t1.rank === 2 && $t2.rank === 1) {
+    const t22D = reshape($t2, [-1, 1]);
+    const t1t2 = matMul($t1, t22D);
+    return reshape(t1t2, [t1t2.size]);
+  } else {
+    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);
+    const t1t2 = matMul($t1, t22D);
+    return t1t2;
+  }
+}
+var dot = op({ dot_ });
+function einsum_(equation, ...tensors) {
+  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, "einsum"));
+  const attrs = { equation };
+  return ENGINE.runKernel(Einsum, $tensors, attrs);
+}
+var einsum = op({ einsum_ });
+function elu_(x) {
+  const $x = convertToTensor(x, "x", "elu", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Elu, inputs);
+}
+var elu = op({ elu_ });
+function erf_(x) {
+  let $x = convertToTensor(x, "x", "erf");
+  assert($x.dtype === "int32" || $x.dtype === "float32", () => "Input dtype must be `int32` or `float32`.");
+  if ($x.dtype === "int32") {
+    $x = cast($x, "float32");
+  }
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Erf, inputs);
+}
+var erf = op({ erf_ });
+function axesAreInnerMostDims(axes, rank) {
+  for (let i = 0; i < axes.length; ++i) {
+    if (axes[axes.length - i - 1] !== rank - 1 - i) {
+      return false;
+    }
+  }
+  return true;
+}
+function combineLocations(outputLoc, reduceLoc, axes) {
+  const rank = outputLoc.length + reduceLoc.length;
+  const loc = [];
+  let outIdx = 0;
+  let reduceIdx = 0;
+  for (let dim = 0; dim < rank; dim++) {
+    if (axes.indexOf(dim) === -1) {
+      loc.push(outputLoc[outIdx++]);
+    } else {
+      loc.push(reduceLoc[reduceIdx++]);
+    }
+  }
+  return loc;
+}
+function computeOutAndReduceShapes(aShape, axes) {
+  const outShape = [];
+  const rank = aShape.length;
+  for (let dim = 0; dim < rank; dim++) {
+    if (axes.indexOf(dim) === -1) {
+      outShape.push(aShape[dim]);
+    }
+  }
+  const reduceShape = axes.map((dim) => aShape[dim]);
+  return [outShape, reduceShape];
+}
+function expandShapeToKeepDim(shape, axes) {
+  const reduceSubShape = axes.map((x) => 1);
+  return combineLocations(shape, reduceSubShape, axes);
+}
+function assertAxesAreInnerMostDims(msg, axes, rank) {
+  assert(axesAreInnerMostDims(axes, rank), () => `${msg} supports only inner-most axes for now. Got axes ${axes} and rank-${rank} input.`);
+}
+function getAxesPermutation(axes, rank) {
+  if (axesAreInnerMostDims(axes, rank)) {
+    return null;
+  }
+  const result = [];
+  for (let i = 0; i < rank; ++i) {
+    if (axes.indexOf(i) === -1) {
+      result.push(i);
+    }
+  }
+  axes.forEach((axis) => result.push(axis));
+  return result;
+}
+function getUndoAxesPermutation(axes) {
+  return axes.map((axis, i) => [i, axis]).sort((a, b) => a[1] - b[1]).map((x) => x[0]);
+}
+function getInnerMostAxes(numAxes, rank) {
+  const res = [];
+  for (let i = rank - numAxes; i < rank; ++i) {
+    res.push(i);
+  }
+  return res;
+}
+function max_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "max");
+  const inputs = { x: $x };
+  const attrs = { reductionIndices: axis, keepDims };
+  return ENGINE.runKernel(Max, inputs, attrs);
+}
+var max = op({ max_ });
+function min_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "min");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Min, inputs, attrs);
+}
+var min = op({ min_ });
+function pow_(base, exp4) {
+  let $base = convertToTensor(base, "base", "pow");
+  let $exp = convertToTensor(exp4, "exp", "pow");
+  [$base, $exp] = makeTypesMatch($base, $exp);
+  const inputs = { a: $base, b: $exp };
+  return ENGINE.runKernel(Pow, inputs);
+}
+var pow = op({ pow_ });
+function scalar(value, dtype) {
+  if ((isTypedArray(value) && dtype !== "string" || Array.isArray(value)) && dtype !== "complex64") {
+    throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");
+  }
+  if (dtype === "string" && isTypedArray(value) && !(value instanceof Uint8Array)) {
+    throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");
+  }
+  const shape = [];
+  const inferredShape = [];
+  return makeTensor(value, shape, inferredShape, dtype);
+}
+function sqrt_(x) {
+  const $x = convertToTensor(x, "x", "sqrt", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sqrt, inputs);
+}
+var sqrt = op({ sqrt_ });
+function square_(x) {
+  const $x = convertToTensor(x, "x", "square");
+  const attrs = {};
+  return ENGINE.runKernel("Square", { x: $x }, attrs);
+}
+var square = op({ square_ });
+function sum_(x, axis = null, keepDims = false) {
+  let $x = convertToTensor(x, "x", "sum");
+  if ($x.dtype === "bool") {
+    $x = cast($x, "int32");
+  }
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Sum, inputs, attrs);
+}
+var sum2 = op({ sum_ });
+function norm_(x, ord = "euclidean", axis = null, keepDims = false) {
+  x = convertToTensor(x, "x", "norm");
+  const norm2 = normImpl(x, ord, axis);
+  let keepDimsShape = norm2.shape;
+  if (keepDims) {
+    const axes = parseAxisParam(axis, x.shape);
+    keepDimsShape = expandShapeToKeepDim(norm2.shape, axes);
+  }
+  return reshape(norm2, keepDimsShape);
+}
+function normImpl(x, p2, axis = null) {
+  if (x.rank === 0) {
+    return abs(x);
+  }
+  if (x.rank !== 1 && axis === null) {
+    return normImpl(reshape(x, [-1]), p2, axis);
+  }
+  if (x.rank === 1 || typeof axis === "number" || Array.isArray(axis) && axis.length === 1) {
+    if (p2 === 1) {
+      return sum2(abs(x), axis);
+    }
+    if (p2 === Infinity) {
+      return max(abs(x), axis);
+    }
+    if (p2 === -Infinity) {
+      return min(abs(x), axis);
+    }
+    if (p2 === "euclidean" || p2 === 2) {
+      return sqrt(sum2(pow(abs(x), scalar(2, "int32")), axis));
+    }
+    throw new Error(`Error in norm: invalid ord value: ${p2}`);
+  }
+  if (Array.isArray(axis) && axis.length === 2) {
+    if (p2 === 1) {
+      return max(sum2(abs(x), axis[0]), axis[1] - 1);
+    }
+    if (p2 === Infinity) {
+      return max(sum2(abs(x), axis[1]), axis[0]);
+    }
+    if (p2 === -Infinity) {
+      return min(sum2(abs(x), axis[1]), axis[0]);
+    }
+    if (p2 === "fro" || p2 === "euclidean") {
+      return sqrt(sum2(square(x), axis));
+    }
+    throw new Error(`Error in norm: invalid ord value: ${p2}`);
+  }
+  throw new Error(`Error in norm: invalid axis: ${axis}`);
+}
+var norm = op({ norm_ });
+function euclideanNorm_(x, axis = null, keepDims = false) {
+  return norm(x, "euclidean", axis, keepDims);
+}
+var euclideanNorm = op({ euclideanNorm_ });
+function exp_(x) {
+  const $x = convertToTensor(x, "x", "exp");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Exp, inputs);
+}
+var exp = op({ exp_ });
+function expandDims_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "expandDims", "string_or_numeric");
+  assert(axis <= $x.rank, () => "Axis must be <= rank of the tensor");
+  const inputs = { input: $x };
+  const attrs = { dim: axis };
+  return ENGINE.runKernel(ExpandDims, inputs, attrs);
+}
+var expandDims = op({ expandDims_ });
+function expm1_(x) {
+  const $x = convertToTensor(x, "x", "expm1");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Expm1, inputs);
+}
+var expm1 = op({ expm1_ });
+function tile_(x, reps) {
+  const $x = convertToTensor(x, "x", "tile", "string_or_numeric");
+  assert($x.rank === reps.length, () => `Error in transpose: rank of input ${$x.rank} must match length of reps ${reps}.`);
+  const inputs = { x: $x };
+  const attrs = { reps };
+  return ENGINE.runKernel(Tile, inputs, attrs);
+}
+var tile = op({ tile_ });
+function eye_(numRows, numColumns, batchShape, dtype = "float32") {
+  if (numColumns == null) {
+    numColumns = numRows;
+  }
+  const buff = buffer([numRows, numColumns], dtype);
+  const n = numRows <= numColumns ? numRows : numColumns;
+  for (let i = 0; i < n; ++i) {
+    buff.set(1, i, i);
+  }
+  const out = reshape(buff.toTensor(), [numRows, numColumns]);
+  if (batchShape == null) {
+    return out;
+  } else {
+    if (batchShape.length === 1) {
+      return tile(expandDims(out, 0), [batchShape[0], 1, 1]);
+    } else if (batchShape.length === 2) {
+      return tile(expandDims(expandDims(out, 0), 0), [batchShape[0], batchShape[1], 1, 1]);
+    } else if (batchShape.length === 3) {
+      return tile(expandDims(expandDims(expandDims(out, 0), 0), 0), [
+        batchShape[0],
+        batchShape[1],
+        batchShape[2],
+        1,
+        1
+      ]);
+    } else {
+      throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${batchShape.length}D.`);
+    }
+  }
+}
+var eye = op({ eye_ });
+function floor_(x) {
+  const $x = convertToTensor(x, "x", "floor", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Floor, inputs);
+}
+var floor = op({ floor_ });
+function gather_(x, indices, axis = 0, batchDims = 0) {
+  const $x = convertToTensor(x, "x", "gather");
+  const $indices = convertToTensor(indices, "indices", "gather", "int32");
+  const inputs = { x: $x, indices: $indices };
+  const attrs = { axis, batchDims };
+  return ENGINE.runKernel(GatherV2, inputs, attrs);
+}
+var gather = op({ gather_ });
+function greater_(a, b) {
+  let $a = convertToTensor(a, "a", "greater", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "greater", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Greater, inputs);
+}
+var greater = op({ greater_ });
+function greaterEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "greaterEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "greaterEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(GreaterEqual, inputs);
+}
+var greaterEqual = op({ greaterEqual_ });
+function isFinite_(x) {
+  const $x = convertToTensor(x, "x", "isFinite");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsFinite, inputs);
+}
+var isFinite2 = op({ isFinite_ });
+function isInf_(x) {
+  const $x = convertToTensor(x, "x", "isInf");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsInf, inputs);
+}
+var isInf = op({ isInf_ });
+function isNaN_(x) {
+  const $x = convertToTensor(x, "x", "isNaN");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsNan, inputs);
+}
+var isNaN2 = op({ isNaN_ });
+function leakyRelu_(x, alpha = 0.2) {
+  const $x = convertToTensor(x, "x", "leakyRelu");
+  const inputs = { x: $x };
+  const attrs = { alpha };
+  return ENGINE.runKernel(LeakyRelu, inputs, attrs);
+}
+var leakyRelu = op({ leakyRelu_ });
+function less_(a, b) {
+  let $a = convertToTensor(a, "a", "less", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "less", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Less, inputs);
+}
+var less = op({ less_ });
+function lessEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "lessEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "lessEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LessEqual, inputs);
+}
+var lessEqual = op({ lessEqual_ });
+function linspace(start, stop, num) {
+  if (num <= 0) {
+    throw new Error("The number of values should be positive.");
+  }
+  const attrs = { start, stop, num };
+  return ENGINE.runKernel(LinSpace, {}, attrs);
+}
+function localResponseNormalization_(x, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {
+  const $x = convertToTensor(x, "x", "localResponseNormalization");
+  assert($x.rank === 4 || $x.rank === 3, () => `Error in localResponseNormalization: x must be rank 3 or 4 but got
+               rank ${$x.rank}.`);
+  assert(isInt(depthRadius), () => `Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${depthRadius}.`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  const inputs = { x: x4D };
+  const attrs = { depthRadius, bias, alpha, beta };
+  const res = ENGINE.runKernel(LRN, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  } else {
+    return res;
+  }
+}
+var localResponseNormalization = op({ localResponseNormalization_ });
+function log_(x) {
+  const $x = convertToTensor(x, "x", "log", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Log, inputs);
+}
+var log2 = op({ log_ });
+function log1p_(x) {
+  const $x = convertToTensor(x, "x", "log1p");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Log1p, inputs);
+}
+var log1p = op({ log1p_ });
+function grad(f) {
+  assert(isFunction(f), () => "The f passed in grad(f) must be a function");
+  return (x, dy) => {
+    const $x = convertToTensor(x, "x", "tf.grad", "string_or_numeric");
+    const $dy = dy != null ? convertToTensor(dy, "dy", "tf.grad") : null;
+    return ENGINE.tidy(() => {
+      const { value, grads: grads2 } = ENGINE.gradients(() => f($x), [$x], $dy);
+      if ($dy != null) {
+        assertShapesMatch(value.shape, $dy.shape, "The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)");
+      }
+      checkGrads(grads2);
+      return grads2[0];
+    });
+  };
+}
+function grads(f) {
+  assert(isFunction(f), () => "The f passed in grads(f) must be a function");
+  return (args, dy) => {
+    assert(Array.isArray(args), () => "The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");
+    const $args = convertToTensorArray(args, "args", "tf.grads", "string_or_numeric");
+    const $dy = dy != null ? convertToTensor(dy, "dy", "tf.grads") : null;
+    return ENGINE.tidy(() => {
+      const { value, grads: grads2 } = ENGINE.gradients(() => f(...$args), $args, $dy);
+      if ($dy != null) {
+        assertShapesMatch(value.shape, $dy.shape, "The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])");
+      }
+      checkGrads(grads2);
+      return grads2;
+    });
+  };
+}
+function valueAndGrad(f) {
+  assert(isFunction(f), () => "The f passed in valueAndGrad(f) must be a function");
+  return (x, dy) => {
+    assert(x instanceof Tensor, () => "The x passed in valueAndGrad(f)(x) must be a tensor");
+    assert(dy == null || dy instanceof Tensor, () => "The dy passed in valueAndGrad(f)(x, dy) must be a tensor");
+    const { grads: grads2, value } = ENGINE.gradients(() => f(x), [x], dy);
+    checkGrads(grads2);
+    return { grad: grads2[0], value };
+  };
+}
+function valueAndGrads(f) {
+  assert(isFunction(f), () => "The f passed in valueAndGrads(f) must be a function");
+  return (args, dy) => {
+    assert(Array.isArray(args) && args.every((arg) => arg instanceof Tensor), () => "The args passed in valueAndGrads(f)(args) must be array of tensors");
+    assert(dy == null || dy instanceof Tensor, () => "The dy passed in valueAndGrads(f)(args, dy) must be a tensor");
+    const res = ENGINE.gradients(() => f(...args), args, dy);
+    if (dy != null) {
+      assertShapesMatch(res.value.shape, dy.shape, "The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])");
+    }
+    checkGrads(res.grads);
+    return res;
+  };
+}
+function variableGrads(f, varList) {
+  assert(isFunction(f), () => "The f passed in variableGrads(f) must be a function");
+  assert(varList == null || Array.isArray(varList) && varList.every((v) => v instanceof Variable), () => "The varList passed in variableGrads(f, varList) must be an array of variables");
+  const specifiedVarList = varList != null;
+  if (!specifiedVarList) {
+    varList = [];
+    for (const varName in ENGINE.registeredVariables) {
+      varList.push(ENGINE.registeredVariables[varName]);
+    }
+  }
+  const specifiedNonTrainable = specifiedVarList ? varList.filter((variable2) => !variable2.trainable) : null;
+  const originalVarCount = varList.length;
+  varList = varList.filter((variable2) => variable2.trainable);
+  assert(varList.length > 0, () => `variableGrads() expects at least one of the input variables to be trainable, but none of the ${originalVarCount} variables is trainable.`);
+  const allowNoGradients = true;
+  const { value, grads: grads2 } = ENGINE.gradients(f, varList, null, allowNoGradients);
+  assert(grads2.some((g) => g != null), () => "Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize().");
+  assert(value.rank === 0, () => `The f passed in variableGrads(f) must return a scalar, but it returned a rank-${value.rank} tensor`);
+  const namedGrads = {};
+  varList.forEach((v, i) => {
+    if (grads2[i] != null) {
+      namedGrads[v.name] = grads2[i];
+    }
+  });
+  if (specifiedNonTrainable != null) {
+    specifiedNonTrainable.forEach((v) => namedGrads[v.name] = null);
+  }
+  return { value, grads: namedGrads };
+}
+function customGrad(f) {
+  return ENGINE.customGrad(f);
+}
+function checkGrads(grads2) {
+  const numNullGradients = grads2.filter((g) => g == null).length;
+  if (numNullGradients > 0) {
+    throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
+    the f you passed encloses all operations that lead from x to y.`);
+  }
+}
+function softplus_(x) {
+  const $x = convertToTensor(x, "x", "softplus");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Softplus, inputs);
+}
+var softplus = op({ softplus_ });
+function logSigmoid_(x) {
+  const $x = convertToTensor(x, "x", "logSigmoid");
+  const customOp = customGrad((x2) => {
+    const value = neg(softplus(neg(x2)));
+    const gradFunc = (dy) => {
+      const derX = mul(dy, sigmoid(neg(x2)));
+      return derX;
+    };
+    return { value, gradFunc };
+  });
+  return customOp($x);
+}
+var logSigmoid = op({ logSigmoid_ });
+function sub_(a, b) {
+  let $a = convertToTensor(a, "a", "sub");
+  let $b = convertToTensor(b, "b", "sub");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Sub, inputs);
+}
+var sub = op({ sub_ });
+function logSoftmax_(logits, axis = -1) {
+  const $logits = convertToTensor(logits, "logits", "logSoftmax");
+  if (axis === -1) {
+    axis = $logits.rank - 1;
+  }
+  if (axis !== $logits.rank - 1) {
+    throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and axis was ${axis}`);
+  }
+  const customOp = customGrad((logits2, save) => {
+    const keepDims = true;
+    const xMax = max(logits2, axis, true);
+    const shifted = sub(logits2, xMax);
+    const value = sub(cast(shifted, "float32"), log2(sum2(exp(shifted), axis, keepDims)));
+    save([value]);
+    const gradFunc = (dy, saved) => {
+      const [value2] = saved;
+      const keepDims2 = true;
+      const softmax6 = exp(value2);
+      return sub(dy, mul(sum2(dy, axis, keepDims2), softmax6));
+    };
+    return { value, gradFunc };
+  });
+  return customOp($logits);
+}
+var logSoftmax = op({ logSoftmax_ });
+function logSumExp_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "logSumExp");
+  const axes = parseAxisParam(axis, $x.shape);
+  const xMax = max($x, axes, true);
+  const a = sub($x, xMax);
+  const b = exp(a);
+  const c = sum2(b, axes);
+  const d = log2(c);
+  const res = add2(reshape(xMax, d.shape), d);
+  if (keepDims) {
+    const newShape = expandShapeToKeepDim(res.shape, axes);
+    return reshape(res, newShape);
+  }
+  return res;
+}
+var logSumExp = op({ logSumExp_ });
+function logicalAnd_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalAnd", "bool");
+  const $b = convertToTensor(b, "b", "logicalAnd", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LogicalAnd, inputs);
+}
+var logicalAnd = op({ logicalAnd_ });
+function logicalNot_(x) {
+  const $x = convertToTensor(x, "x", "logicalNot", "bool");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(LogicalNot, inputs);
+}
+var logicalNot = op({ logicalNot_ });
+function logicalOr_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalOr", "bool");
+  const $b = convertToTensor(b, "b", "logicalOr", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LogicalOr, inputs);
+}
+var logicalOr = op({ logicalOr_ });
+function logicalXor_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalXor", "bool");
+  const $b = convertToTensor(b, "b", "logicalXor", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  return logicalAnd(logicalOr(a, b), logicalNot(logicalAnd(a, b)));
+}
+var logicalXor = op({ logicalXor_ });
+var INT32_MAX = 2147483648;
+function searchSorted_(sortedSequence, values, side = "left") {
+  const $sortedSequence = convertToTensor(sortedSequence, "sortedSequence", "searchSorted");
+  const $values = convertToTensor(values, "values", "searchSorted");
+  const sequenceSize = $sortedSequence.shape[$sortedSequence.shape.length - 1];
+  const valuesSize = $values.shape[$values.shape.length - 1];
+  const $sortedSequence2D = reshape($sortedSequence, [-1, sequenceSize]);
+  const $values2D = reshape($values, [-1, valuesSize]);
+  if ($sortedSequence2D.rank < 2) {
+    throw new Error(`Sorted input argument must be at least 2-dimensional`);
+  }
+  if ($sortedSequence2D.shape[0] !== $values2D.shape[0]) {
+    throw new Error(`Leading dimension of 'sortedSequence' and 'values' must match.`);
+  }
+  if (sizeFromShape($values2D.shape) >= INT32_MAX) {
+    throw new Error(`values tensor size must less than ${INT32_MAX}`);
+  }
+  if ($sortedSequence2D.shape[1] >= INT32_MAX) {
+    throw new Error(`trailing dim_size must less than ${INT32_MAX} for int32 output type, was ${$sortedSequence2D.shape[1]}`);
+  }
+  const inputs = {
+    sortedSequence: $sortedSequence2D,
+    values: $values2D
+  };
+  const attrs = { side };
+  return ENGINE.runKernel(SearchSorted, inputs, attrs);
+}
+var searchSorted = op({ searchSorted_ });
+function lowerBound(sortedSequence, values) {
+  return searchSorted(sortedSequence, values, "left");
+}
+function maxPool_(x, filterSize, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "maxPool");
+  const dilations = 1;
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x4D.rank}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  checkPadOnDimRoundingMode("maxPool", pad3, dimRoundingMode);
+  const inputs = { x: x4D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(MaxPool, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var maxPool = op({ maxPool_ });
+function maxPool3d_(x, filterSize = [1, 1, 1], strides, pad3, dimRoundingMode, dataFormat = "NDHWC") {
+  const $x = convertToTensor(x, "x", "maxPool3d");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in maxPool3d: x must be rank 5 but got rank ${x5D.rank}.`);
+  assert(dataFormat === "NDHWC", () => `Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  checkPadOnDimRoundingMode("maxPool3d", pad3, dimRoundingMode);
+  const inputs = { x: x5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };
+  const res = ENGINE.runKernel(MaxPool3D, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var maxPool3d = op({ maxPool3d_ });
+function maxPoolWithArgmax_(x, filterSize, strides, pad3, includeBatchInIndex = false) {
+  const $x = convertToTensor(x, "x", "maxPoolWithArgmax");
+  const inputs = { x: $x };
+  const attrs = { filterSize, strides, pad: pad3, includeBatchInIndex };
+  const result = ENGINE.runKernel(MaxPoolWithArgmax, inputs, attrs);
+  return { result: result[0], indexes: result[1] };
+}
+var maxPoolWithArgmax = op({ maxPoolWithArgmax_ });
+function maximum_(a, b) {
+  let $a = convertToTensor(a, "a", "maximum");
+  let $b = convertToTensor(b, "b", "maximum");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "bool") {
+    $a = cast($a, "int32");
+    $b = cast($b, "int32");
+  }
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Maximum, inputs);
+}
+var maximum = op({ maximum_ });
+function mean_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "mean");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Mean, inputs, attrs);
+}
+var mean = op({ mean_ });
+function zeros(shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = zeros(shape, "float32");
+    const imag4 = zeros(shape, "float32");
+    return complex(real4, imag4);
+  }
+  const values = makeZerosTypedArray(sizeFromShape(shape), dtype);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+function ones2(shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = ones2(shape, "float32");
+    const imag4 = zeros(shape, "float32");
+    return complex(real4, imag4);
+  }
+  const values = makeOnesTypedArray(sizeFromShape(shape), dtype);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+function meshgrid(x, y, { indexing = "xy" } = {}) {
+  if (indexing !== "xy" && indexing !== "ij") {
+    throw new TypeError(`${indexing} is not a valid third argument to meshgrid`);
+  }
+  if (x === void 0) {
+    return [];
+  }
+  let $x = convertToTensor(x, "x", "meshgrid", x instanceof Tensor ? x.dtype : "float32");
+  if (y === void 0) {
+    return [$x];
+  }
+  let $y = convertToTensor(y, "y", "meshgrid", y instanceof Tensor ? y.dtype : "float32");
+  const w = sizeFromShape($x.shape);
+  const h = sizeFromShape($y.shape);
+  if (indexing === "xy") {
+    $x = reshape($x, [1, -1]);
+    $y = reshape($y, [-1, 1]);
+    return [
+      matMul(ones2([h, 1], $x.dtype), $x),
+      matMul($y, ones2([1, w], $y.dtype))
+    ];
+  }
+  $x = reshape($x, [-1, 1]);
+  $y = reshape($y, [1, -1]);
+  return [
+    matMul($x, ones2([1, h], $x.dtype)),
+    matMul(ones2([w, 1], $y.dtype), $y)
+  ];
+}
+function minimum_(a, b) {
+  let $a = convertToTensor(a, "a", "minimum");
+  let $b = convertToTensor(b, "b", "minimum");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "bool") {
+    $a = cast($a, "int32");
+    $b = cast($b, "int32");
+  }
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Minimum, inputs);
+}
+var minimum = op({ minimum_ });
+function mirrorPad_(x, paddings, mode) {
+  assert(mode === "reflect" || mode === "symmetric", () => `Invalid mode. Mode must be either reflect or symmetric. Got ${mode}.`);
+  const $x = convertToTensor(x, "x", "mirrorPad");
+  if ($x.rank === 0) {
+    throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");
+  }
+  assert(paddings.length === $x.rank, () => `Padding doesn't match input. Must be ${$x.rank}. Got ${paddings.length}.`);
+  const shapeOffset = mode === "reflect" ? 1 : 0;
+  for (let i = 0; i < $x.rank; i++) {
+    assert(paddings[i].length === 2, () => `Invalid number of paddings. Must be length of 2 each.`);
+    assert(paddings[i][0] >= 0 && paddings[i][0] <= $x.shape[i] - shapeOffset && paddings[i][1] >= 0 && paddings[i][1] <= $x.shape[i] - shapeOffset, () => `Padding in dimension ${i} cannot be greater than or equal to ${$x.shape[i] - shapeOffset} or less than 0 for input of shape ${$x.shape}`);
+  }
+  const attrs = { paddings, mode };
+  const inputs = { x: $x };
+  return ENGINE.runKernel(MirrorPad, inputs, attrs);
+}
+var mirrorPad = op({ mirrorPad_ });
+function mod_(a, b) {
+  let $a = convertToTensor(a, "a", "mod");
+  let $b = convertToTensor(b, "b", "mod");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Mod, inputs);
+}
+var mod = op({ mod_ });
+function moments_(x, axis = null, keepDims = false) {
+  x = convertToTensor(x, "x", "moments");
+  const axes = parseAxisParam(axis, x.shape);
+  const xMean = mean(x, axes, keepDims);
+  let keepDimsShape = xMean.shape;
+  if (!keepDims) {
+    keepDimsShape = expandShapeToKeepDim(xMean.shape, axes);
+  }
+  const devSquared = square(sub(cast(x, "float32"), reshape(xMean, keepDimsShape)));
+  const variance = mean(devSquared, axes, keepDims);
+  return { mean: xMean, variance };
+}
+var moments = op({ moments_ });
+function multiRNNCell_(lstmCells, data, c, h) {
+  const $data = convertToTensor(data, "data", "multiRNNCell");
+  const $c = convertToTensorArray(c, "c", "multiRNNCell");
+  const $h = convertToTensorArray(h, "h", "multiRNNCell");
+  let input2 = $data;
+  const newStates = [];
+  for (let i = 0; i < lstmCells.length; i++) {
+    const output = lstmCells[i](input2, $c[i], $h[i]);
+    newStates.push(output[0]);
+    newStates.push(output[1]);
+    input2 = output[1];
+  }
+  const newC = [];
+  const newH = [];
+  for (let i = 0; i < newStates.length; i += 2) {
+    newC.push(newStates[i]);
+    newH.push(newStates[i + 1]);
+  }
+  return [newC, newH];
+}
+var multiRNNCell = op({ multiRNNCell_ });
+function multinomial_(logits, numSamples, seed, normalized = false) {
+  const $logits = convertToTensor(logits, "logits", "multinomial");
+  const numOutcomes = $logits.size;
+  const origRank = $logits.rank;
+  if (numOutcomes < 2) {
+    throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${numOutcomes}.`);
+  }
+  if (origRank > 2) {
+    throw new Error(`Rank of probabilities must be 1 or 2, but is ${origRank}`);
+  }
+  seed = seed || Math.random();
+  const logits2D = origRank === 1 ? reshape($logits, [1, -1]) : $logits;
+  const inputs = { logits: logits2D };
+  const attrs = { numSamples, seed, normalized };
+  const res = ENGINE.runKernel(Multinomial, inputs, attrs);
+  return origRank === 1 ? reshape(res, [res.size]) : res;
+}
+var multinomial = op({ multinomial_ });
+function notEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "notEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "notEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(NotEqual, inputs);
+}
+var notEqual = op({ notEqual_ });
+function onesLike_(x) {
+  const $x = convertToTensor(x, "x", "onesLike");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(OnesLike, inputs);
+}
+var onesLike = op({ onesLike_ });
+function outerProduct_(v1, v2) {
+  const $v1 = convertToTensor(v1, "v1", "outerProduct");
+  const $v2 = convertToTensor(v2, "v2", "outerProduct");
+  assert($v1.rank === 1 && $v2.rank === 1, () => `Error in outerProduct: inputs must be rank 1, but got ranks ${$v1.rank} and ${$v2.rank}.`);
+  const v12D = reshape($v1, [-1, 1]);
+  const v22D = reshape($v2, [1, -1]);
+  return matMul(v12D, v22D);
+}
+var outerProduct = op({ outerProduct_ });
+function pad_(x, paddings, constantValue = 0) {
+  const $x = convertToTensor(x, "x", "pad");
+  if ($x.rank === 0) {
+    throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");
+  }
+  const attrs = { paddings, constantValue };
+  const inputs = { x: $x };
+  return ENGINE.runKernel(PadV2, inputs, attrs);
+}
+var pad = op({ pad_ });
+function pad1d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 2, () => "Invalid number of paddings. Must be length of 2.");
+  return pad(x, [paddings], constantValue);
+}
+var pad1d = op({ pad1d_ });
+function pad2d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 2 && paddings[0].length === 2 && paddings[1].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad2d = op({ pad2d_ });
+function pad3d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 3 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad3d = op({ pad3d_ });
+function pad4d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 4 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2 && paddings[3].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad4d = op({ pad4d_ });
+function spaceToBatchND_(x, blockShape, paddings) {
+  const $x = convertToTensor(x, "x", "spaceToBatchND");
+  assert($x.rank >= 1 + blockShape.length, () => `input rank ${$x.rank} should be > than [blockShape] ${blockShape.length}`);
+  assert(paddings.length === blockShape.length, () => `paddings.shape[0] ${paddings.length} must be equal to [blockShape] ${blockShape.length}`);
+  assert($x.shape.reduce((a, b, i) => {
+    if (i > 0 && i <= blockShape.length) {
+      return a && (b + paddings[i - 1][0] + paddings[i - 1][1]) % blockShape[i - 1] === 0;
+    }
+    return a;
+  }, true), () => `input spatial dimensions ${$x.shape.slice(1)} with paddings ${paddings.toString()} must be divisible by blockShapes ${blockShape.toString()}`);
+  const inputs = { x: $x };
+  const attrs = { blockShape, paddings };
+  return ENGINE.runKernel(SpaceToBatchND, inputs, attrs);
+}
+var spaceToBatchND = op({ spaceToBatchND_ });
+function pool_(input2, windowShape, poolingType, pad3, dilations, strides, dimRoundingMode) {
+  if (dilations == null) {
+    dilations = [1, 1];
+  }
+  if (strides == null) {
+    strides = 1;
+  }
+  if (pad3 === 0) {
+    pad3 = "valid";
+  }
+  const $x = convertToTensor(input2, "x", "maxPool");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in pool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = computePool2DInfo(x4D.shape, windowShape, strides, dilations, pad3);
+  const dilation = [convInfo.dilationHeight, convInfo.dilationWidth];
+  let basePadding;
+  if (pad3 === "same") {
+    basePadding = withSpaceToBatchBasePaddings([convInfo.filterHeight, convInfo.filterWidth], dilation);
+  } else {
+    basePadding = [[0, 0], [0, 0]];
+  }
+  const isDilationOne = dilation[0] === 1 && dilation[1] === 1;
+  const [adjustedPadding, adjustedCrops] = requiredSpaceToBatchPaddings([convInfo.inHeight, convInfo.inWidth], dilation, basePadding);
+  const convertedPad = isDilationOne ? pad3 : "valid";
+  const convertedX = isDilationOne ? x4D : spaceToBatchND(x4D, dilation, adjustedPadding);
+  const forwardOp = poolingType === "avg" ? () => avgPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode) : () => maxPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode);
+  const y = forwardOp();
+  const res = isDilationOne ? y : batchToSpaceND(y, dilation, adjustedCrops);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+function requiredSpaceToBatchPaddings(inputShape, blockShape, basePadding) {
+  const padStart = basePadding.map((b) => b[0]);
+  const origPadEnd = basePadding.map((b) => b[1]);
+  const fullInputShape = inputShape.concat(padStart, origPadEnd);
+  const padEndExtra = blockShape.map((b, i) => (b - fullInputShape[i] % b) % b);
+  const padEnd = origPadEnd.map((s, i) => s + padEndExtra[i]);
+  const paddings = blockShape.map((_, i) => [padStart[i], padEnd[i]]);
+  const crops = blockShape.map((_, i) => [0, padEndExtra[i]]);
+  return [paddings, crops];
+}
+function withSpaceToBatchBasePaddings(filterShape, dilation) {
+  const dilatedFilterShape = filterShape.map((s, i) => {
+    return s + (s - 1) * (dilation[i] - 1);
+  });
+  const padExtraShape = dilatedFilterShape.map((s) => s - 1);
+  const padExtraStart = padExtraShape.map((s) => Math.floor(s / 2));
+  const padExtraEnd = padExtraShape.map((s, i) => s - padExtraStart[i]);
+  return padExtraShape.map((_, i) => {
+    return [padExtraStart[i], padExtraEnd[i]];
+  });
+}
+var pool = op({ pool_ });
+function prelu_(x, alpha) {
+  const $x = convertToTensor(x, "x", "prelu");
+  const $alpha = convertToTensor(alpha, "alpha", "prelu");
+  const inputs = { x: $x, alpha: $alpha };
+  return ENGINE.runKernel(Prelu, inputs);
+}
+var prelu = op({ prelu_ });
+function prod_(x, axis = null, keepDims = false) {
+  let $x = convertToTensor(x, "x", "prod");
+  if ($x.dtype === "bool") {
+    $x = cast($x, "int32");
+  }
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Prod, inputs, attrs);
+}
+var prod = op({ prod_ });
+function raggedGather_(paramsNestedSplits, paramsDenseValues, indices, outputRaggedRank) {
+  const $paramsNestedSplits = paramsNestedSplits.map((t, i) => convertToTensor(t, `tensors${i}`, "raggedGather", "int32"));
+  const $paramsDenseValues = convertToTensor(paramsDenseValues, "paramsDenseValues", "raggedGather");
+  const $indices = convertToTensor(indices, "indices", "raggedGather", "int32");
+  const inputs = {
+    paramsNestedSplits: $paramsNestedSplits,
+    paramsDenseValues: $paramsDenseValues,
+    indices: $indices
+  };
+  const attrs = { outputRaggedRank };
+  const result = ENGINE.runKernel(RaggedGather, inputs, attrs);
+  return {
+    outputNestedSplits: result.slice(0, result.length - 1),
+    outputDenseValues: result[result.length - 1]
+  };
+}
+var raggedGather = op({ raggedGather_ });
+function raggedRange_(starts, limits, deltas) {
+  const $starts = convertToTensor(starts, "starts", "raggedRange");
+  const $limits = convertToTensor(limits, "limits", "raggedRange", $starts.dtype);
+  const $deltas = convertToTensor(deltas, "deltas", "raggedRange", $starts.dtype);
+  const inputs = {
+    starts: $starts,
+    limits: $limits,
+    deltas: $deltas
+  };
+  const result = ENGINE.runKernel(RaggedRange, inputs);
+  return {
+    rtNestedSplits: result[0],
+    rtDenseValues: result[1]
+  };
+}
+var raggedRange = op({ raggedRange_ });
+function raggedTensorToTensor_(shape, values, defaultValue, rowPartitionTensors, rowPartitionTypes) {
+  const $shape = convertToTensor(shape, "shape", "raggedTensorToTensor", "int32");
+  const $values = convertToTensor(values, "values", "raggedTensorToTensor");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "raggedTensorToTensor", $values.dtype);
+  const $rowPartitionTensors = rowPartitionTensors.map((t, i) => convertToTensor(t, `tensors${i}`, "raggedTensorToTensor", "int32"));
+  const inputs = {
+    shape: $shape,
+    values: $values,
+    defaultValue: $defaultValue,
+    rowPartitionTensors: $rowPartitionTensors
+  };
+  const attrs = { rowPartitionTypes };
+  return ENGINE.runKernel(RaggedTensorToTensor, inputs, attrs);
+}
+var raggedTensorToTensor = op({ raggedTensorToTensor_ });
+function rand_(shape, randFunction, dtype) {
+  const size = sizeFromShape(shape);
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  for (let i = 0; i < size; i++) {
+    values[i] = randFunction();
+  }
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+var rand = op({ rand_ });
+var seedrandom = __toESM(require_seedrandom2());
+var MPRandGauss = class {
+  constructor(mean4, stdDeviation, dtype, truncated, seed) {
+    this.mean = mean4;
+    this.stdDev = stdDeviation;
+    this.dtype = dtype;
+    this.nextVal = NaN;
+    this.truncated = truncated;
+    if (this.truncated) {
+      this.upper = this.mean + this.stdDev * 2;
+      this.lower = this.mean - this.stdDev * 2;
+    }
+    const seedValue = seed ? seed : Math.random();
+    this.random = seedrandom.alea(seedValue.toString());
+  }
+  nextValue() {
+    if (!isNaN(this.nextVal)) {
+      const value = this.nextVal;
+      this.nextVal = NaN;
+      return value;
+    }
+    let resultX, resultY;
+    let isValid = false;
+    while (!isValid) {
+      let v1, v2, s;
+      do {
+        v1 = 2 * this.random() - 1;
+        v2 = 2 * this.random() - 1;
+        s = v1 * v1 + v2 * v2;
+      } while (s >= 1 || s === 0);
+      const mul2 = Math.sqrt(-2 * Math.log(s) / s);
+      resultX = this.mean + this.stdDev * v1 * mul2;
+      resultY = this.mean + this.stdDev * v2 * mul2;
+      if (!this.truncated || this.isValidTruncated(resultX)) {
+        isValid = true;
+      }
+    }
+    if (!this.truncated || this.isValidTruncated(resultY)) {
+      this.nextVal = this.convertValue(resultY);
+    }
+    return this.convertValue(resultX);
+  }
+  convertValue(value) {
+    if (this.dtype == null || this.dtype === "float32") {
+      return value;
+    }
+    return Math.round(value);
+  }
+  isValidTruncated(value) {
+    return value <= this.upper && value >= this.lower;
+  }
+};
+var RandGamma = class {
+  constructor(alpha, beta, dtype, seed) {
+    this.alpha = alpha;
+    this.beta = 1 / beta;
+    this.dtype = dtype;
+    const seedValue = seed ? seed : Math.random();
+    this.randu = seedrandom.alea(seedValue.toString());
+    this.randn = new MPRandGauss(0, 1, dtype, false, this.randu());
+    if (alpha < 1) {
+      this.d = alpha + 2 / 3;
+    } else {
+      this.d = alpha - 1 / 3;
+    }
+    this.c = 1 / Math.sqrt(9 * this.d);
+  }
+  nextValue() {
+    let x2, v0, v1, x, u, v;
+    while (true) {
+      do {
+        x = this.randn.nextValue();
+        v = 1 + this.c * x;
+      } while (v <= 0);
+      v *= v * v;
+      x2 = x * x;
+      v0 = 1 - 0.331 * x2 * x2;
+      v1 = 0.5 * x2 + this.d * (1 - v + Math.log(v));
+      u = this.randu();
+      if (u < v0 || Math.log(u) < v1) {
+        break;
+      }
+    }
+    v = 1 / this.beta * this.d * v;
+    if (this.alpha < 1) {
+      v *= Math.pow(this.randu(), 1 / this.alpha);
+    }
+    return this.convertValue(v);
+  }
+  convertValue(value) {
+    if (this.dtype === "float32") {
+      return value;
+    }
+    return Math.round(value);
+  }
+};
+var UniformRandom = class {
+  constructor(min6 = 0, max6 = 1, dtype, seed) {
+    this.canReturnFloat = () => this.dtype == null || this.dtype === "float32";
+    this.min = min6;
+    this.range = max6 - min6;
+    this.dtype = dtype;
+    if (seed == null) {
+      seed = Math.random();
+    }
+    if (typeof seed === "number") {
+      seed = seed.toString();
+    }
+    if (!this.canReturnFloat() && this.range <= 1) {
+      throw new Error(`The difference between ${min6} - ${max6} <= 1 and dtype is not float`);
+    }
+    this.random = seedrandom.alea(seed);
+  }
+  convertValue(value) {
+    if (this.canReturnFloat()) {
+      return value;
+    }
+    return Math.round(value);
+  }
+  nextValue() {
+    return this.convertValue(this.min + this.range * this.random());
+  }
+};
+function randomGamma_(shape, alpha, beta = 1, dtype = "float32", seed) {
+  if (beta == null) {
+    beta = 1;
+  }
+  if (dtype == null) {
+    dtype = "float32";
+  }
+  if (dtype !== "float32" && dtype !== "int32") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  const rgamma = new RandGamma(alpha, beta, dtype, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = rgamma.nextValue();
+  }
+  return res.toTensor();
+}
+var randomGamma = op({ randomGamma_ });
+function randomNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  const randGauss = new MPRandGauss(mean4, stdDev, dtype, false, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = randGauss.nextValue();
+  }
+  return res.toTensor();
+}
+var randomNormal = op({ randomNormal_ });
+function randomStandardNormal_(shape, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  return randomNormal(shape, 0, 1, dtype, seed);
+}
+var randomStandardNormal = op({ randomStandardNormal_ });
+function randomUniform_(shape, minval = 0, maxval = 1, dtype = "float32", seed) {
+  const res = buffer(shape, dtype);
+  const random = new UniformRandom(minval, maxval, null, seed);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = random.nextValue();
+  }
+  return res.toTensor();
+}
+var randomUniform = op({ randomUniform_ });
+function range(start, stop, step5 = 1, dtype = "float32") {
+  if (step5 === 0) {
+    throw new Error("Cannot have a step of zero");
+  }
+  const attrs = { start, stop, step: step5, dtype };
+  return ENGINE.runKernel(Range, {}, attrs);
+}
+function reciprocal_(x) {
+  const $x = convertToTensor(x, "x", "reciprocal");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Reciprocal, inputs);
+}
+var reciprocal = op({ reciprocal_ });
+function relu_(x) {
+  const $x = convertToTensor(x, "x", "relu");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Relu, inputs);
+}
+var relu = op({ relu_ });
+function relu6_(x) {
+  const $x = convertToTensor(x, "x", "relu6");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Relu6, inputs);
+}
+var relu6 = op({ relu6_ });
+function reverse_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  const inputs = { x: $x };
+  const attrs = { dims: axis };
+  return ENGINE.runKernel(Reverse, inputs, attrs);
+}
+var reverse = op({ reverse_ });
+function reverse1d_(x) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 1, () => `Error in reverse1D: x must be rank 1 but got rank ${$x.rank}.`);
+  return reverse($x, 0);
+}
+var reverse1d = op({ reverse1d_ });
+function reverse2d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 2, () => `Error in reverse2D: x must be rank 2 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse2d = op({ reverse2d_ });
+function reverse3d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 3, () => `Error in reverse3D: x must be rank 3 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse3d = op({ reverse3d_ });
+function reverse4d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 4, () => `Error in reverse4D: x must be rank 4 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse4d = op({ reverse4d_ });
+function round_(x) {
+  const $x = convertToTensor(x, "x", "round");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Round, inputs);
+}
+var round2 = op({ round_ });
+function rsqrt_(x) {
+  const $x = convertToTensor(x, "x", "rsqrt", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Rsqrt, inputs);
+}
+var rsqrt = op({ rsqrt_ });
+function selu_(x) {
+  const $x = convertToTensor(x, "x", "selu");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Selu, inputs);
+}
+var selu = op({ selu_ });
+function separableConv2d_(x, depthwiseFilter, pointwiseFilter, strides, pad3, dilation = [1, 1], dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "separableConv2d");
+  const $depthwiseFilter = convertToTensor(depthwiseFilter, "depthwiseFilter", "separableConv2d");
+  const $pointwiseFilter = convertToTensor(pointwiseFilter, "pointwiseFilter", "separableConv2d");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  if (dataFormat === "NCHW") {
+    throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");
+  }
+  assert(x4D.rank === 4, () => `Error in separableConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($depthwiseFilter.rank === 4, () => `Error in separableConv2d: depthwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);
+  assert($pointwiseFilter.rank === 4, () => `Error in separableConv2d: pointwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);
+  assert($pointwiseFilter.shape[0] === 1, () => `Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${$pointwiseFilter.shape[0]}.`);
+  assert($pointwiseFilter.shape[1] === 1, () => `Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${$pointwiseFilter.shape[1]}.`);
+  const inChannels = $depthwiseFilter.shape[2];
+  const channelMultiplier = $depthwiseFilter.shape[3];
+  assert($pointwiseFilter.shape[2] === inChannels * channelMultiplier, () => `Error in separableConv2d: the third dimension of pointwise filter must be ${inChannels * channelMultiplier}, but got ${$pointwiseFilter.shape[2]}.`);
+  const depthwise = depthwiseConv2d(x4D, $depthwiseFilter, strides, pad3, dataFormat, dilation);
+  const pointwiseStride = 1;
+  const res = conv2d(depthwise, $pointwiseFilter, pointwiseStride, "valid", dataFormat);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var separableConv2d = op({ separableConv2d_ });
+async function setdiff1dAsync_(x, y) {
+  const $x = convertToTensor(x, "x", "setdiff1d");
+  const $y = convertToTensor(y, "y", "setdiff1d");
+  assert($x.dtype === $y.dtype, () => `x and y should have the same dtype, but got x (${$x.dtype}) and y (${$y.dtype}).`);
+  assert($x.rank === 1, () => `x should be 1D tensor, but got x (${$x.shape}).`);
+  assert($y.rank === 1, () => `y should be 1D tensor, but got y (${$y.shape}).`);
+  const xVals = await $x.data();
+  const yVals = await $y.data();
+  const ySet = new Set(yVals);
+  let outputSize = 0;
+  for (let i = 0; i < xVals.length; i++) {
+    if (!ySet.has(xVals[i])) {
+      outputSize++;
+    }
+  }
+  const buffer2 = new TensorBuffer([outputSize], $x.dtype);
+  const indices = new TensorBuffer([outputSize], "int32");
+  for (let i = 0, p2 = 0; i < xVals.length; i++) {
+    if (!ySet.has(xVals[i])) {
+      buffer2.values[p2] = xVals[i];
+      indices.values[p2] = i;
+      p2++;
+    }
+  }
+  return [buffer2.toTensor(), indices.toTensor()];
+}
+var setdiff1dAsync = setdiff1dAsync_;
+function sign_(x) {
+  const $x = convertToTensor(x, "x", "sign");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sign, inputs);
+}
+var sign = op({ sign_ });
+function sin_(x) {
+  const $x = convertToTensor(x, "x", "sin", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sin, inputs);
+}
+var sin = op({ sin_ });
+function sinh_(x) {
+  const $x = convertToTensor(x, "x", "sinh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sinh, inputs);
+}
+var sinh = op({ sinh_ });
+function slice1d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice1d");
+  assert($x.rank === 1, () => `slice1d expects a rank-1 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, [begin], [size]);
+}
+var slice1d = op({ slice1d_ });
+function slice2d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice2d");
+  assert($x.rank === 2, () => `slice2d expects a rank-2 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice2d = op({ slice2d_ });
+function slice3d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice3d");
+  assert($x.rank === 3, () => `slice3d expects a rank-3 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice3d = op({ slice3d_ });
+function slice4d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice4d");
+  assert($x.rank === 4, () => `slice4d expects a rank-4 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice4d = op({ slice4d_ });
+function softmax_(logits, dim = -1) {
+  const $logits = convertToTensor(logits, "logits", "softmax", "float32");
+  if (dim === -1) {
+    dim = $logits.rank - 1;
+  }
+  if (dim !== $logits.rank - 1) {
+    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and dim was ${dim}`);
+  }
+  const inputs = { logits: $logits };
+  const attrs = { dim };
+  return ENGINE.runKernel(Softmax, inputs, attrs);
+}
+var softmax = op({ softmax_ });
+function fft_(input2) {
+  assert(input2.dtype === "complex64", () => `The dtype for tf.spectral.fft() must be complex64 but got ${input2.dtype}.`);
+  const inputs = { input: input2 };
+  return ENGINE.runKernel(FFT, inputs);
+}
+var fft = op({ fft_ });
+function ifft_(input2) {
+  assert(input2.dtype === "complex64", () => `The dtype for tf.spectral.ifft() must be complex64 but got ${input2.dtype}.`);
+  const inputs = { input: input2 };
+  return ENGINE.runKernel(IFFT, inputs);
+}
+var ifft = op({ ifft_ });
+function irfft_(input2) {
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = input2.size / innerDimensionSize;
+  let ret;
+  if (innerDimensionSize <= 2) {
+    const complexInput = reshape(input2, [batch, innerDimensionSize]);
+    ret = ifft(complexInput);
+  } else {
+    const outputShape = [batch, 2 * (innerDimensionSize - 1)];
+    const realInput = reshape(real(input2), [batch, innerDimensionSize]);
+    const imagInput = reshape(imag(input2), [batch, innerDimensionSize]);
+    const realConjugate = reverse(slice(realInput, [0, 1], [batch, innerDimensionSize - 2]), 1);
+    const imagConjugate = mul(reverse(slice(imagInput, [0, 1], [batch, innerDimensionSize - 2]), 1), scalar(-1));
+    const r = concat([realInput, realConjugate], 1);
+    const i = concat([imagInput, imagConjugate], 1);
+    const complexInput = reshape(complex(r, i), [outputShape[0], outputShape[1]]);
+    ret = ifft(complexInput);
+  }
+  ret = real(ret);
+  if (input2.rank === 3 && input2.shape[0] !== 0) {
+    const temp = ret;
+    const batch2 = input2.shape[0];
+    ret = reshape(ret, [batch2, ret.shape[0] / batch2, ret.shape[1]]);
+    temp.dispose();
+  }
+  return ret;
+}
+var irfft = op({ irfft_ });
+function split_(x, numOrSizeSplits, axis = 0) {
+  const $x = convertToTensor(x, "x", "split");
+  const inputs = { x: $x };
+  const attr = { numOrSizeSplits, axis };
+  return ENGINE.runKernel(SplitV, inputs, attr);
+}
+var split = op({ split_ });
+function rfft_(input2, fftLength) {
+  assert(input2.dtype === "float32", () => `The dtype for rfft() must be real value but got ${input2.dtype}`);
+  let innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = input2.size / innerDimensionSize;
+  let adjustedInput;
+  if (fftLength != null && fftLength < innerDimensionSize) {
+    const begin = input2.shape.map((v) => 0);
+    const size = input2.shape.map((v) => v);
+    size[input2.shape.length - 1] = fftLength;
+    adjustedInput = slice(input2, begin, size);
+    innerDimensionSize = fftLength;
+  } else if (fftLength != null && fftLength > innerDimensionSize) {
+    const zerosShape = input2.shape.map((v) => v);
+    zerosShape[input2.shape.length - 1] = fftLength - innerDimensionSize;
+    adjustedInput = concat([input2, zeros(zerosShape)], input2.shape.length - 1);
+    innerDimensionSize = fftLength;
+  } else {
+    adjustedInput = input2;
+  }
+  const zerosInput = zerosLike(adjustedInput);
+  const complexInput = reshape(complex(adjustedInput, zerosInput), [batch, innerDimensionSize]);
+  const ret = fft(complexInput);
+  const half = Math.floor(innerDimensionSize / 2) + 1;
+  const realValues = real(ret);
+  const imagValues = imag(ret);
+  const realComplexConjugate = split(realValues, [half, innerDimensionSize - half], realValues.shape.length - 1);
+  const imagComplexConjugate = split(imagValues, [half, innerDimensionSize - half], imagValues.shape.length - 1);
+  const outputShape = adjustedInput.shape.slice();
+  outputShape[adjustedInput.shape.length - 1] = half;
+  return reshape(complex(realComplexConjugate[0], imagComplexConjugate[0]), outputShape);
+}
+var rfft = op({ rfft_ });
+function squaredDifference_(a, b) {
+  let $a = convertToTensor(a, "a", "squaredDifference");
+  let $b = convertToTensor(b, "b", "squaredDifference");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  const attrs = {};
+  return ENGINE.runKernel(SquaredDifference, inputs, attrs);
+}
+var squaredDifference = op({ squaredDifference_ });
+function squeeze_(x, axis) {
+  const $x = convertToTensor(x, "x", "squeeze", "string_or_numeric");
+  return reshape($x, squeezeShape($x.shape, axis).newShape);
+}
+var squeeze = op({ squeeze_ });
+function stack_(tensors, axis = 0) {
+  const $tensors = convertToTensorArray(tensors, "tensors", "stack", "string_or_numeric");
+  assert($tensors.length >= 1, () => "Pass at least one tensor to tf.stack");
+  if ($tensors.length > 0) {
+    assert(axis <= $tensors[0].rank, () => "Axis must be <= rank of the tensor");
+  }
+  const inputs = $tensors;
+  const attrs = { axis };
+  return ENGINE.runKernel(Pack, inputs, attrs);
+}
+var stack = op({ stack_ });
+function step_(x, alpha = 0) {
+  const $x = convertToTensor(x, "x", "step");
+  const inputs = { x: $x };
+  const attrs = { alpha };
+  return ENGINE.runKernel(Step, inputs, attrs);
+}
+var step = op({ step_ });
+function stridedSlice_(x, begin, end, strides, beginMask = 0, endMask = 0, ellipsisMask = 0, newAxisMask = 0, shrinkAxisMask = 0) {
+  const $x = convertToTensor(x, "x", "stridedSlice", "string_or_numeric");
+  const inputs = { x: $x };
+  const attrs = {
+    begin,
+    end,
+    strides,
+    beginMask,
+    endMask,
+    ellipsisMask,
+    newAxisMask,
+    shrinkAxisMask
+  };
+  return ENGINE.runKernel(StridedSlice, inputs, attrs);
+}
+var stridedSlice = op({ stridedSlice_ });
+function tan_(x) {
+  const $x = convertToTensor(x, "x", "tan", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Tan, inputs);
+}
+var tan = op({ tan_ });
+function tensor1d(values, dtype) {
+  assertNonNull(values);
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 1) {
+    throw new Error("tensor1d() requires values to be a flat/TypedArray");
+  }
+  const shape = null;
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+function tensor2d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 2) {
+    throw new Error("tensor2d() requires shape to have two numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 2 && inferredShape.length !== 1) {
+    throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+function tensor4d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 4) {
+    throw new Error("tensor4d() requires shape to have four numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 4 && inferredShape.length !== 1) {
+    throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+function tensor5d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 5) {
+    throw new Error("tensor5d() requires shape to have five numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 5 && inferredShape.length !== 1) {
+    throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+function tensor6d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 6) {
+    throw new Error("tensor6d() requires shape to have six numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 6 && inferredShape.length !== 1) {
+    throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");
+  }
+  shape = shape || inferredShape;
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+function topk_(x, k = 1, sorted = true) {
+  const $x = convertToTensor(x, "x", "topk");
+  if ($x.rank === 0) {
+    throw new Error("topk() expects the input to be of rank 1 or higher");
+  }
+  const lastDim = $x.shape[$x.shape.length - 1];
+  if (k < 0) {
+    throw new Error(`'k' passed to topk() must be >= 0 but got ${k}`);
+  }
+  if (k > lastDim) {
+    throw new Error(`'k' passed to topk() must be <= the last dimension (${lastDim}) but got ${k}`);
+  }
+  const inputs = { x: $x };
+  const attrs = { k, sorted };
+  const [values, indices] = ENGINE.runKernel(TopK, inputs, attrs);
+  return { values, indices };
+}
+var topk = op({ topk_ });
+function truncatedNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type $ { dtype }`);
+  }
+  const randGauss = new MPRandGauss(mean4, stdDev, dtype, true, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = randGauss.nextValue();
+  }
+  return res.toTensor();
+}
+var truncatedNormal = op({ truncatedNormal_ });
+function unique_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "unique", "string_or_numeric");
+  assert($x.rank > 0, () => "The input tensor must be at least 1D");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  const [values, indices] = ENGINE.runKernel(Unique, inputs, attrs);
+  return { values, indices };
+}
+var unique = op({ unique_ });
+function unsortedSegmentSum_(x, segmentIds, numSegments) {
+  const $x = convertToTensor(x, "x", "unsortedSegmentSum");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "unsortedSegmentSum", "int32");
+  assert(isInt(numSegments), () => "numSegments must be of dtype int");
+  const inputs = { x: $x, segmentIds: $segmentIds };
+  const attrs = { numSegments };
+  return ENGINE.runKernel(UnsortedSegmentSum, inputs, attrs);
+}
+var unsortedSegmentSum = op({ unsortedSegmentSum_ });
+function unstack_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "unstack", "string_or_numeric");
+  assert(axis >= -$x.shape.length && axis < $x.shape.length, () => `Axis = ${axis} is not in [-${$x.shape.length}, ${$x.shape.length})`);
+  const inputs = { value: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(Unpack, inputs, attrs);
+}
+var unstack = op({ unstack_ });
+function upperBound(sortedSequence, values) {
+  return searchSorted(sortedSequence, values, "right");
+}
+function variable(initialValue, trainable = true, name, dtype) {
+  return ENGINE.makeVariable(initialValue, trainable, name, dtype);
+}
+function whereImpl(condShape, condVals) {
+  const indices = [];
+  for (let i = 0; i < condVals.length; i++) {
+    if (condVals[i]) {
+      indices.push(i);
+    }
+  }
+  const inBuffer = buffer(condShape, "int32");
+  const out = buffer([indices.length, condShape.length], "int32");
+  for (let i = 0; i < indices.length; i++) {
+    const loc = inBuffer.indexToLoc(indices[i]);
+    const offset = i * condShape.length;
+    out.values.set(loc, offset);
+  }
+  return out.toTensor();
+}
+async function whereAsync_(condition) {
+  const $condition = convertToTensor(condition, "condition", "whereAsync", "bool");
+  const vals = await $condition.data();
+  const res = whereImpl($condition.shape, vals);
+  if (condition !== $condition) {
+    $condition.dispose();
+  }
+  return res;
+}
+var whereAsync = whereAsync_;
+async function booleanMaskAsync_(tensor2, mask, axis) {
+  const $tensor = convertToTensor(tensor2, "tensor", "boolMask");
+  const $mask = convertToTensor(mask, "mask", "boolMask", "bool");
+  const axisFrom = axis == null ? 0 : axis;
+  const maskDim = $mask.rank;
+  const tensorShape = $tensor.shape;
+  assert(maskDim > 0, () => "mask cannot be scalar");
+  assertShapesMatch(tensorShape.slice(axisFrom, axisFrom + maskDim), $mask.shape, `mask's shape must match the first K dimensions of tensor's shape,`);
+  let leadingSize = 1;
+  for (let i = axisFrom; i < axisFrom + maskDim; i++) {
+    leadingSize *= tensorShape[i];
+  }
+  const targetTensorShape = tensorShape.slice(0, axisFrom).concat([leadingSize], tensorShape.slice(axisFrom + maskDim));
+  const reshapedTensor = reshape($tensor, targetTensorShape);
+  const reshapedMask = reshape($mask, [-1]);
+  const positivePositions = await whereAsync(reshapedMask);
+  const indices = squeeze(positivePositions, [1]);
+  const res = gather(reshapedTensor, indices, axisFrom);
+  if (tensor2 !== $tensor) {
+    $tensor.dispose();
+  }
+  if (mask !== $mask) {
+    $mask.dispose();
+  }
+  indices.dispose();
+  reshapedTensor.dispose();
+  reshapedMask.dispose();
+  positivePositions.dispose();
+  return res;
+}
+var booleanMaskAsync = booleanMaskAsync_;
+function movingAverage_(v, x, decay, step5, zeroDebias = true) {
+  const $v = convertToTensor(v, "v", "movingAverage");
+  const $x = convertToTensor(x, "x", "movingAverage");
+  const $decay = convertToTensor(decay, "decay", "movingAverage");
+  assertTypesMatch($v, $x);
+  assert(arraysEqual($v.shape, $x.shape), () => "Shape mismatch in v and x");
+  const one = scalar(1);
+  const oneMinusDecay = sub(one, $decay);
+  let update = mul(sub($x, $v), oneMinusDecay);
+  if (zeroDebias) {
+    assert(step5 != null, () => "When using zeroDebias: true, step is required.");
+    const $step = convertToTensor(step5, "step", "movingAverage");
+    update = div(update, sub(one, pow($decay, $step)));
+  }
+  return add2($v, update);
+}
+var movingAverage = op({ movingAverage_ });
+function scatterND_(indices, updates, shape) {
+  const $indices = convertToTensor(indices, "indices", "scatterND", "int32");
+  const $updates = convertToTensor(updates, "updates", "scatterND");
+  validateInput($updates, $indices, shape);
+  const inputs = { indices: $indices, updates: $updates };
+  const attrs = { shape };
+  return ENGINE.runKernel(ScatterNd, inputs, attrs);
+}
+var scatterND = op({ scatterND_ });
+function validateInput2(sparseIndices, sparseValues, outputShape, defaultValues) {
+  if (sparseIndices.dtype !== "int32") {
+    throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${sparseIndices.dtype}.`);
+  }
+  if (sparseIndices.rank > 2) {
+    throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${sparseIndices.shape}.`);
+  }
+  const numElems = sparseIndices.rank > 0 ? sparseIndices.shape[0] : 1;
+  const numDims = sparseIndices.rank > 1 ? sparseIndices.shape[1] : 1;
+  if (outputShape.length !== numDims) {
+    throw new Error(`outputShape has incorrect number of elements:, ${outputShape.length}, should be: ${numDims}.`);
+  }
+  const numValues = sparseValues.size;
+  if (!(sparseValues.rank === 0 || sparseValues.rank === 1 && numValues === numElems)) {
+    throw new Error(`sparseValues has incorrect shape ${sparseValues.shape}, should be [] or [${numElems}]`);
+  }
+  if (sparseValues.dtype !== defaultValues.dtype) {
+    throw new Error("sparseValues.dtype must match defaultValues.dtype");
+  }
+}
+function sparseToDense_(sparseIndices, sparseValues, outputShape, defaultValue = 0) {
+  const $sparseIndices = convertToTensor(sparseIndices, "sparseIndices", "sparseToDense", "int32");
+  const $sparseValues = convertToTensor(sparseValues, "sparseValues", "sparseToDense", "string_or_numeric");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "sparseToDense", $sparseValues.dtype);
+  validateInput2($sparseIndices, $sparseValues, outputShape, $defaultValue);
+  const inputs = {
+    sparseIndices: $sparseIndices,
+    sparseValues: $sparseValues,
+    defaultValue: $defaultValue
+  };
+  const attrs = { outputShape };
+  return ENGINE.runKernel(SparseToDense, inputs, attrs);
+}
+var sparseToDense = op({ sparseToDense_ });
+function gatherND_(x, indices) {
+  const $indices = convertToTensor(indices, "indices", "gatherND", "int32");
+  const $x = convertToTensor(x, "x", "gatherND", "string_or_numeric");
+  const inputs = { params: $x, indices: $indices };
+  return ENGINE.runKernel(GatherNd, inputs);
+}
+var gatherND = op({ gatherND_ });
+function getNoiseShape(x, noiseShape) {
+  if (noiseShape == null) {
+    return x.shape.slice();
+  }
+  if (arraysEqual(x.shape, noiseShape)) {
+    return noiseShape;
+  }
+  if (x.shape.length === noiseShape.length) {
+    const newDimension = [];
+    for (let i = 0; i < x.shape.length; i++) {
+      if (noiseShape[i] == null && x.shape[i] != null) {
+        newDimension.push(x.shape[i]);
+      } else {
+        newDimension.push(noiseShape[i]);
+      }
+    }
+    return newDimension;
+  }
+  return noiseShape;
+}
+function dropout_(x, rate, noiseShape, seed) {
+  const $x = convertToTensor(x, "x", "dropout");
+  assert($x.dtype === "float32", () => `x has to be a floating point tensor since it's going to be scaled, but got a ${$x.dtype} tensor instead.`);
+  assert(rate >= 0 && rate < 1, () => `rate must be a float in the range [0, 1), but got ${rate}.`);
+  if (rate === 0) {
+    return x instanceof Tensor ? $x.clone() : $x;
+  }
+  const $noiseShape = getNoiseShape($x, noiseShape);
+  const keepProb = 1 - rate;
+  const multiplier = div(floor(add2(randomUniform($noiseShape, 0, 1, "float32", seed), keepProb)), keepProb);
+  return mul($x, multiplier);
+}
+var dropout = op({ dropout_ });
+function enclosingPowerOfTwo(value) {
+  return Math.floor(Math.pow(2, Math.ceil(Math.log(value) / Math.log(2))));
+}
+function cosineWindow(windowLength, a, b) {
+  const even = 1 - windowLength % 2;
+  const newValues = new Float32Array(windowLength);
+  for (let i = 0; i < windowLength; ++i) {
+    const cosArg = 2 * Math.PI * i / (windowLength + even - 1);
+    newValues[i] = a - b * Math.cos(cosArg);
+  }
+  return tensor1d(newValues, "float32");
+}
+async function inTopKAsync_(predictions, targets, k = 1) {
+  const $predictions = convertToTensor(predictions, "predictions", "inTopK");
+  const $targets = convertToTensor(targets, "targets", "inTopK");
+  assert($predictions.rank > 1, () => `inTopK() expects the predictions to be of rank 2 or higher, but got ${$predictions.rank}`);
+  assert($predictions.rank - 1 === $targets.rank, () => `predictions rank should be 1 larger than targets rank, but got predictions rank ${$predictions.rank} and targets rank ${$targets.rank}`);
+  assertShapesMatch($predictions.shape.slice(0, $predictions.shape.length - 1), $targets.shape, `predictions's shape should be align with the targets' shape, except the last dimension.`);
+  const lastDim = $predictions.shape[$predictions.shape.length - 1];
+  assert(k > 0 && k <= lastDim, () => `'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${lastDim}), but got ${k}`);
+  const predictionsVals = await $predictions.data();
+  const targetsVals = await $targets.data();
+  const [batch, size] = [predictionsVals.length / lastDim, lastDim];
+  const precision3 = getTypedArrayFromDType("bool", batch);
+  for (let b = 0; b < batch; b++) {
+    const offset = b * size;
+    const vals = predictionsVals.subarray(offset, offset + size);
+    const valAndInd = [];
+    for (let i = 0; i < vals.length; i++) {
+      valAndInd.push({ value: vals[i], index: i });
+    }
+    valAndInd.sort((a, b2) => b2.value - a.value);
+    precision3[b] = 0;
+    for (let i = 0; i < k; i++) {
+      if (valAndInd[i].index === targetsVals[b]) {
+        precision3[b] = 1;
+        break;
+      }
+    }
+  }
+  if (predictions !== $predictions) {
+    $predictions.dispose();
+  }
+  if (targets !== $targets) {
+    $targets.dispose();
+  }
+  return tensor(precision3, $targets.shape, "bool");
+}
+var inTopKAsync = inTopKAsync_;
+var fused_ops_exports = {};
+__export2(fused_ops_exports, {
+  conv2d: () => conv2d2,
+  depthwiseConv2d: () => depthwiseConv2d2,
+  matMul: () => matMul2
+});
+function conv2DBackpropFilter_(x, dy, filterShape, strides, pad3, dataFormat = "NHWC", dimRoundingMode) {
+  let x4D = x;
+  if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  }
+  let dy4D = dy;
+  if (dy4D.rank === 3) {
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in conv2dDerFilter: input must be rank 4, but got shape ${x4D.shape}.`);
+  assert(dy4D.rank === 4, () => `Error in conv2dDerFilter: dy must be rank 4, but got shape ${dy4D.shape}.`);
+  assert(filterShape.length === 4, () => `Error in conv2dDerFilter: filterShape must be length 4, but got ${filterShape}.`);
+  const inDepth = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  const outDepth = dataFormat === "NHWC" ? dy4D.shape[3] : dy4D.shape[1];
+  assert(inDepth === filterShape[2], () => `Error in conv2dDerFilter: depth of input ${inDepth}) must match input depth in filter (${filterShape[2]}.`);
+  assert(outDepth === filterShape[3], () => `Error in conv2dDerFilter: depth of dy (${outDepth}) must match output depth for filter (${filterShape[3]}).`);
+  checkPadOnDimRoundingMode("conv2dDerFilter", pad3, dimRoundingMode);
+  const inputs = { x: x4D, dy: dy4D };
+  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape };
+  return ENGINE.runKernel(Conv2DBackpropFilter, inputs, attrs);
+}
+var conv2DBackpropFilter = op({ conv2DBackpropFilter_ });
+function getFusedDyActivation(dy, y, activation2) {
+  if (activation2 == null || activation2 === "linear") {
+    return dy;
+  }
+  if (activation2 === "relu") {
+    return mul(dy, step(y));
+  }
+  throw new Error(`Cannot compute gradient for fused activation ${activation2}.`);
+}
+function getFusedBiasGradient(bias, dyActivation) {
+  let res = dyActivation;
+  const reduceAxes = getReductionAxes(bias.shape, dyActivation.shape);
+  if (reduceAxes.length > 0) {
+    res = sum2(res, reduceAxes);
+  }
+  return reshape(res, bias.shape);
+}
+function applyActivation(x, activation2, preluActivationWeights, leakyreluAlpha) {
+  if (activation2 === "linear") {
+    return x;
+  } else if (activation2 === "relu") {
+    return relu(x);
+  } else if (activation2 === "elu") {
+    return elu(x);
+  } else if (activation2 === "relu6") {
+    return relu6(x);
+  } else if (activation2 === "prelu") {
+    return prelu(x, preluActivationWeights);
+  } else if (activation2 === "leakyrelu") {
+    return leakyRelu(x, leakyreluAlpha);
+  } else if (activation2 === "sigmoid") {
+    return sigmoid(x);
+  }
+  throw new Error(`Unknown fused activation ${activation2}.`);
+}
+var shouldFuse = (gradientDepth, activation2) => {
+  const gradientMode = gradientDepth > 0;
+  return !gradientMode || activation2 === "linear";
+};
+function fusedConv2d_({ x, filter, strides, pad: pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha }) {
+  activation2 = activation2 || "linear";
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    assert(dataFormat === "NHWC", () => `Error in fused conv2d: got dataFormat of ${dataFormat} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);
+    let result = conv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  const $x = convertToTensor(x, "x", "conv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "conv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in fused conv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in fused conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("fused conv2d", pad3, dimRoundingMode);
+  const inputChannels = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert($filter.shape[2] === inputChannels, () => `Error in conv2d: depth of input (${inputChannels}) must match input depth for filter ${$filter.shape[2]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused conv2d");
+    [$bias] = makeTypesMatch($bias, $x);
+    if (dataFormat === "NHWC") {
+      assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);
+    } else {
+      assert($bias.shape.length <= 1, () => `Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${$bias.shape.length}.`);
+      assert($bias.shape.length === 0 || $bias.shape[0] === convInfo.outChannels || $bias.shape[0] === 1, () => `Error in fused conv2d: bias shape (${$bias.shape}) is not compatible with the number of output channels (${convInfo.outChannels})`);
+    }
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    const alphaShape = preluActivationWeights.shape;
+    assert(alphaShape.length <= 1 || alphaShape.length === 3, () => `Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${alphaShape.length}.`);
+    if (alphaShape.length === 1) {
+      assert(alphaShape[0] === 1 || alphaShape[0] === convInfo.outChannels, () => `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the number of output channels (${convInfo.outChannels}).`);
+    } else if (alphaShape.length === 3) {
+      try {
+        assertAndGetBroadcastShape(alphaShape, convInfo.outShape);
+      } catch (e) {
+        const errMsg = `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the output shape of the conv2d (${convInfo.outShape}).`;
+        throw Error(errMsg);
+      }
+    }
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused conv2d");
+  }
+  const grad2 = (dy, saved) => {
+    assert(dataFormat === "NHWC", () => `Error in gradient of fused conv2D: got dataFormat of ${dataFormat} but only NHWC is currently supported.`);
+    const [$filter2, x4D2, y, $bias2] = saved;
+    const dyActivation = getFusedDyActivation(dy, y, activation2);
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    const xDer = conv2DBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3);
+    const filterDer = conv2DBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3);
+    const der = [xDer, filterDer];
+    if ($bias2 != null) {
+      const biasDer = getFusedBiasGradient($bias2, dyActivation);
+      der.push(biasDer);
+    }
+    return der;
+  };
+  const inputs = {
+    x: x4D,
+    filter: $filter,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = {
+    strides,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    dimRoundingMode,
+    activation: activation2,
+    leakyreluAlpha
+  };
+  if (bias == null) {
+    const customOp = customGrad((x4D2, filter2, save) => {
+      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);
+      save([filter2, x4D2, res]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOp(x4D, $filter);
+  } else {
+    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {
+      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);
+      save([filter2, x4D2, res, bias2]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOpWithBias(x4D, $filter, $bias);
+  }
+}
+var conv2d2 = op({ fusedConv2d_ });
+function depthwiseConv2dNativeBackpropFilter_(x, dy, filterShape, strides, pad3, dilations = [1, 1], dimRoundingMode) {
+  let x4D = x;
+  if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  }
+  let dy4D = dy;
+  if (dy4D.rank === 3) {
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  const inputs = { x: x4D, dy: dy4D };
+  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, filterShape };
+  return ENGINE.runKernel(DepthwiseConv2dNativeBackpropFilter, inputs, attrs);
+}
+var depthwiseConv2dNativeBackpropFilter = op({ depthwiseConv2dNativeBackpropFilter_ });
+function depthwiseConv2dNativeBackpropInput_(xShape, dy, filter, strides, pad3, dilations = [1, 1], dimRoundingMode) {
+  let dy4D = dy;
+  let reshapedTo4D = false;
+  if (dy.rank === 3) {
+    reshapedTo4D = true;
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  const inputs = { dy: dy4D, filter };
+  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, inputShape: xShape };
+  const res = ENGINE.runKernel(DepthwiseConv2dNativeBackpropInput, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var depthwiseConv2dNativeBackpropInput = op({ depthwiseConv2dNativeBackpropInput_ });
+function fusedDepthwiseConv2d_({ x, filter, strides, pad: pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha }) {
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    let result = depthwiseConv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  const $x = convertToTensor(x, "x", "depthwiseConv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "depthwiseConv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in fused depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  assert(x4D.shape[3] === $filter.shape[2], () => `Error in fused depthwiseConv2d: number of input channels (${x4D.shape[3]}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);
+  if (dilations == null) {
+    dilations = [1, 1];
+  }
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  checkPadOnDimRoundingMode("fused depthwiseConv2d", pad3, dimRoundingMode);
+  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused conv2d");
+    [$bias] = makeTypesMatch($bias, $x);
+    assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused depthwiseConv2d");
+  }
+  const grad2 = (dy, saved) => {
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${dilations}'`);
+    const [$filter2, x4D2, y, bias2] = saved;
+    const dyActivation = getFusedDyActivation(dy, y, activation2);
+    const xDer = depthwiseConv2dNativeBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3, dilations, dimRoundingMode);
+    const filterDer = depthwiseConv2dNativeBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3, dilations, dimRoundingMode);
+    if (bias2 != null) {
+      const biasDer = getFusedBiasGradient($bias, dyActivation);
+      return [xDer, filterDer, biasDer];
+    }
+    return [xDer, filterDer];
+  };
+  const inputs = {
+    x: x4D,
+    filter: $filter,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = {
+    strides,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    dimRoundingMode,
+    activation: activation2,
+    leakyreluAlpha
+  };
+  if (bias == null) {
+    const customOp = customGrad((x4D2, filter2, save) => {
+      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);
+      save([filter2, x4D2, res]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOp(x4D, $filter);
+  } else {
+    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {
+      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);
+      save([filter2, x4D2, res, bias2]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOpWithBias(x4D, $filter, $bias);
+  }
+}
+var depthwiseConv2d2 = op({ fusedDepthwiseConv2d_ });
+function fusedMatMul_({ a, b, transposeA = false, transposeB = false, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha = 0.2 }) {
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    let result = matMul(a, b, transposeA, transposeB);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  let $a = convertToTensor(a, "a", "fused matMul");
+  let $b = convertToTensor(b, "b", "fused matMul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const innerShapeA = transposeA ? $a.shape[$a.rank - 2] : $a.shape[$a.rank - 1];
+  const innerShapeB = transposeB ? $b.shape[$b.rank - 1] : $b.shape[$b.rank - 2];
+  const outerShapeA = transposeA ? $a.shape[$a.rank - 1] : $a.shape[$a.rank - 2];
+  const outerShapeB = transposeB ? $b.shape[$b.rank - 2] : $b.shape[$b.rank - 1];
+  const outerDimsA = $a.shape.slice(0, -2);
+  const outerDimsB = $b.shape.slice(0, -2);
+  const batchDimA = sizeFromShape(outerDimsA);
+  const batchDimB = sizeFromShape(outerDimsB);
+  assert(innerShapeA === innerShapeB, () => `Error in fused matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${$a.shape} and ${$b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const outShapeOuterDims = assertAndGetBroadcastShape($a.shape.slice(0, -2), $b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  const a3D = transposeA ? reshape($a, [batchDimA, innerShapeA, outerShapeA]) : reshape($a, [batchDimA, outerShapeA, innerShapeA]);
+  const b3D = transposeB ? reshape($b, [batchDimB, outerShapeB, innerShapeB]) : reshape($b, [batchDimB, innerShapeB, outerShapeB]);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused matMul");
+    [$bias] = makeTypesMatch($bias, $a);
+    assertAndGetBroadcastShape(outShape, $bias.shape);
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused matMul");
+  }
+  const grad2 = (dy, saved) => {
+    const [a3D2, b3D2, y, $bias2] = saved;
+    const dyActivation = getFusedDyActivation(reshape(dy, y.shape), y, activation2);
+    let aDer;
+    let bDer;
+    if (!transposeA && !transposeB) {
+      aDer = matMul(dyActivation, b3D2, false, true);
+      bDer = matMul(a3D2, dyActivation, true, false);
+    } else if (!transposeA && transposeB) {
+      aDer = matMul(dyActivation, b3D2, false, false);
+      bDer = matMul(dyActivation, a3D2, true, false);
+    } else if (transposeA && !transposeB) {
+      aDer = matMul(b3D2, dyActivation, false, true);
+      bDer = matMul(a3D2, dyActivation, false, false);
+    } else {
+      aDer = matMul(b3D2, dyActivation, true, true);
+      bDer = matMul(dyActivation, a3D2, true, true);
+    }
+    if (bias != null) {
+      const biasDer = getFusedBiasGradient($bias2, dyActivation);
+      return [aDer, bDer, biasDer];
+    } else {
+      return [aDer, bDer];
+    }
+  };
+  const inputs = {
+    a: a3D,
+    b: b3D,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = { transposeA, transposeB, activation: activation2, leakyreluAlpha };
+  if (bias == null) {
+    const customOp = customGrad((a3D2, b3D2, save) => {
+      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);
+      save([a3D2, b3D2, res]);
+      return { value: reshape(res, outShape), gradFunc: grad2 };
+    });
+    return customOp(a3D, b3D);
+  } else {
+    const customOpWithBias = customGrad((a3D2, b3D2, $bias2, save) => {
+      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);
+      save([a3D2, b3D2, res, $bias2]);
+      return { value: reshape(res, outShape), gradFunc: grad2 };
+    });
+    return customOpWithBias(a3D, b3D, $bias);
+  }
+}
+var matMul2 = op({ fusedMatMul_ });
+function hammingWindow_(windowLength) {
+  return cosineWindow(windowLength, 0.54, 0.46);
+}
+var hammingWindow = op({ hammingWindow_ });
+function hannWindow_(windowLength) {
+  return cosineWindow(windowLength, 0.5, 0.5);
+}
+var hannWindow = op({ hannWindow_ });
+function frame_(signal2, frameLength, frameStep, padEnd = false, padValue = 0) {
+  let start = 0;
+  const output = [];
+  while (start + frameLength <= signal2.size) {
+    output.push(slice(signal2, start, frameLength));
+    start += frameStep;
+  }
+  if (padEnd) {
+    while (start < signal2.size) {
+      const padLen = start + frameLength - signal2.size;
+      const pad3 = concat([
+        slice(signal2, start, frameLength - padLen),
+        fill([padLen], padValue)
+      ]);
+      output.push(pad3);
+      start += frameStep;
+    }
+  }
+  if (output.length === 0) {
+    return tensor2d([], [0, frameLength]);
+  }
+  return reshape(concat(output), [output.length, frameLength]);
+}
+var frame = op({ frame_ });
+function stft_(signal2, frameLength, frameStep, fftLength, windowFn = hannWindow) {
+  if (fftLength == null) {
+    fftLength = enclosingPowerOfTwo(frameLength);
+  }
+  const framedSignal = frame(signal2, frameLength, frameStep);
+  const windowedSignal = mul(framedSignal, windowFn(frameLength));
+  return rfft(windowedSignal, fftLength);
+}
+var stft = op({ stft_ });
+function cropAndResize_(image2, boxes, boxInd, cropSize, method = "bilinear", extrapolationValue = 0) {
+  const $image = convertToTensor(image2, "image", "cropAndResize");
+  const $boxes = convertToTensor(boxes, "boxes", "cropAndResize", "float32");
+  const $boxInd = convertToTensor(boxInd, "boxInd", "cropAndResize", "int32");
+  const numBoxes = $boxes.shape[0];
+  assert($image.rank === 4, () => `Error in cropAndResize: image must be rank 4,but got rank ${$image.rank}.`);
+  assert($boxes.rank === 2 && $boxes.shape[1] === 4, () => `Error in cropAndResize: boxes must be have size [${numBoxes},4] but had shape ${$boxes.shape}.`);
+  assert($boxInd.rank === 1 && $boxInd.shape[0] === numBoxes, () => `Error in cropAndResize: boxInd must be have size [${numBoxes}] but had shape ${$boxes.shape}.`);
+  assert(cropSize.length === 2, () => `Error in cropAndResize: cropSize must be of length 2, but got length ${cropSize.length}.`);
+  assert(cropSize[0] >= 1 && cropSize[1] >= 1, () => `cropSize must be atleast [1,1], but was ${cropSize}`);
+  assert(method === "bilinear" || method === "nearest", () => `method must be bilinear or nearest, but was ${method}`);
+  const inputs = { image: $image, boxes: $boxes, boxInd: $boxInd };
+  const attrs = { method, extrapolationValue, cropSize };
+  const res = ENGINE.runKernel(CropAndResize, inputs, attrs);
+  return res;
+}
+var cropAndResize = op({ cropAndResize_ });
+function flipLeftRight_(image2) {
+  const $image = convertToTensor(image2, "image", "flipLeftRight", "float32");
+  assert($image.rank === 4, () => `Error in flipLeftRight: image must be rank 4,but got rank ${$image.rank}.`);
+  const inputs = { image: $image };
+  const res = ENGINE.runKernel(FlipLeftRight, inputs, {});
+  return res;
+}
+var flipLeftRight = op({ flipLeftRight_ });
+function grayscaleToRGB_(image2) {
+  const $image = convertToTensor(image2, "image", "grayscaleToRGB");
+  const lastDimsIdx = $image.rank - 1;
+  const lastDims = $image.shape[lastDimsIdx];
+  assert($image.rank >= 2, () => `Error in grayscaleToRGB: images must be at least rank 2, but got rank ${$image.rank}.`);
+  assert(lastDims === 1, () => `Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${lastDims}.`);
+  const reps = new Array($image.rank);
+  reps.fill(1, 0, lastDimsIdx);
+  reps[lastDimsIdx] = 3;
+  return tile($image, reps);
+}
+var grayscaleToRGB = op({ grayscaleToRGB_ });
+function rotateWithOffset_(image2, radians, fillValue = 0, center = 0.5) {
+  const $image = convertToTensor(image2, "image", "rotateWithOffset", "float32");
+  assert($image.rank === 4, () => `Error in rotateWithOffset: image must be rank 4,but got rank ${$image.rank}.`);
+  const inputs = { image: $image };
+  const attrs = { radians, fillValue, center };
+  const res = ENGINE.runKernel(RotateWithOffset, inputs, attrs);
+  return res;
+}
+var rotateWithOffset = op({ rotateWithOffset_ });
+function nonMaxSuppSanityCheck(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {
+  if (iouThreshold == null) {
+    iouThreshold = 0.5;
+  }
+  if (scoreThreshold == null) {
+    scoreThreshold = Number.NEGATIVE_INFINITY;
+  }
+  if (softNmsSigma == null) {
+    softNmsSigma = 0;
+  }
+  const numBoxes = boxes.shape[0];
+  maxOutputSize = Math.min(maxOutputSize, numBoxes);
+  assert(0 <= iouThreshold && iouThreshold <= 1, () => `iouThreshold must be in [0, 1], but was '${iouThreshold}'`);
+  assert(boxes.rank === 2, () => `boxes must be a 2D tensor, but was of rank '${boxes.rank}'`);
+  assert(boxes.shape[1] === 4, () => `boxes must have 4 columns, but 2nd dimension was ${boxes.shape[1]}`);
+  assert(scores.rank === 1, () => "scores must be a 1D tensor");
+  assert(scores.shape[0] === numBoxes, () => `scores has incompatible shape with boxes. Expected ${numBoxes}, but was ${scores.shape[0]}`);
+  assert(0 <= softNmsSigma && softNmsSigma <= 1, () => `softNmsSigma must be in [0, 1], but was '${softNmsSigma}'`);
+  return { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };
+}
+function nonMaxSuppression_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression", "float32");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression", "float32");
+  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);
+  maxOutputSize = inputs.maxOutputSize;
+  iouThreshold = inputs.iouThreshold;
+  scoreThreshold = inputs.scoreThreshold;
+  const attrs = { maxOutputSize, iouThreshold, scoreThreshold };
+  return ENGINE.runKernel(NonMaxSuppressionV3, { boxes: $boxes, scores: $scores }, attrs);
+}
+var nonMaxSuppression = op({ nonMaxSuppression_ });
+function binaryInsert(arr, element, comparator) {
+  const index = binarySearch(arr, element, comparator);
+  const insertionPoint = index < 0 ? -(index + 1) : index;
+  arr.splice(insertionPoint, 0, element);
+}
+function binarySearch(arr, target, comparator) {
+  return binarySearch_(arr, target, comparator || defaultComparator);
+}
+function defaultComparator(a, b) {
+  return a > b ? 1 : a < b ? -1 : 0;
+}
+function binarySearch_(arr, target, comparator) {
+  let left = 0;
+  let right = arr.length;
+  let middle = 0;
+  let found = false;
+  while (left < right) {
+    middle = left + (right - left >>> 1);
+    const compareResult = comparator(target, arr[middle]);
+    if (compareResult > 0) {
+      left = middle + 1;
+    } else {
+      right = middle;
+      found = !compareResult;
+    }
+  }
+  return found ? left : -left - 1;
+}
+function nonMaxSuppressionV3Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, 0);
+}
+function nonMaxSuppressionV4Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize) {
+  return nonMaxSuppressionImpl_(
+    boxes,
+    scores,
+    maxOutputSize,
+    iouThreshold,
+    scoreThreshold,
+    0,
+    false,
+    padToMaxOutputSize,
+    true
+  );
+}
+function nonMaxSuppressionV5Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {
+  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, true);
+}
+function nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, returnScoresTensor = false, padToMaxOutputSize = false, returnValidOutputs = false) {
+  const candidates = [];
+  for (let i = 0; i < scores.length; i++) {
+    if (scores[i] > scoreThreshold) {
+      candidates.push({ score: scores[i], boxIndex: i, suppressBeginIndex: 0 });
+    }
+  }
+  candidates.sort(ascendingComparator);
+  const scale22 = softNmsSigma > 0 ? -0.5 / softNmsSigma : 0;
+  const selectedIndices = [];
+  const selectedScores = [];
+  while (selectedIndices.length < maxOutputSize && candidates.length > 0) {
+    const candidate = candidates.pop();
+    const { score: originalScore, boxIndex, suppressBeginIndex } = candidate;
+    if (originalScore < scoreThreshold) {
+      break;
+    }
+    let ignoreCandidate = false;
+    for (let j = selectedIndices.length - 1; j >= suppressBeginIndex; --j) {
+      const iou2 = intersectionOverUnion(boxes, boxIndex, selectedIndices[j]);
+      if (iou2 >= iouThreshold) {
+        ignoreCandidate = true;
+        break;
+      }
+      candidate.score = candidate.score * suppressWeight(iouThreshold, scale22, iou2);
+      if (candidate.score <= scoreThreshold) {
+        break;
+      }
+    }
+    candidate.suppressBeginIndex = selectedIndices.length;
+    if (!ignoreCandidate) {
+      if (candidate.score === originalScore) {
+        selectedIndices.push(boxIndex);
+        selectedScores.push(candidate.score);
+      } else if (candidate.score > scoreThreshold) {
+        binaryInsert(candidates, candidate, ascendingComparator);
+      }
+    }
+  }
+  const validOutputs = selectedIndices.length;
+  const elemsToPad = maxOutputSize - validOutputs;
+  if (padToMaxOutputSize && elemsToPad > 0) {
+    selectedIndices.push(...new Array(elemsToPad).fill(0));
+    selectedScores.push(...new Array(elemsToPad).fill(0));
+  }
+  const result = { selectedIndices };
+  if (returnScoresTensor) {
+    result["selectedScores"] = selectedScores;
+  }
+  if (returnValidOutputs) {
+    result["validOutputs"] = validOutputs;
+  }
+  return result;
+}
+function intersectionOverUnion(boxes, i, j) {
+  const iCoord = boxes.subarray(i * 4, i * 4 + 4);
+  const jCoord = boxes.subarray(j * 4, j * 4 + 4);
+  const yminI = Math.min(iCoord[0], iCoord[2]);
+  const xminI = Math.min(iCoord[1], iCoord[3]);
+  const ymaxI = Math.max(iCoord[0], iCoord[2]);
+  const xmaxI = Math.max(iCoord[1], iCoord[3]);
+  const yminJ = Math.min(jCoord[0], jCoord[2]);
+  const xminJ = Math.min(jCoord[1], jCoord[3]);
+  const ymaxJ = Math.max(jCoord[0], jCoord[2]);
+  const xmaxJ = Math.max(jCoord[1], jCoord[3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0) {
+    return 0;
+  }
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function suppressWeight(iouThreshold, scale22, iou2) {
+  const weight = Math.exp(scale22 * iou2 * iou2);
+  return iou2 <= iouThreshold ? weight : 0;
+}
+function ascendingComparator(c1, c2) {
+  return c1.score - c2.score || c1.score === c2.score && c2.boxIndex - c1.boxIndex;
+}
+async function nonMaxSuppressionAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);
+  maxOutputSize = inputs.maxOutputSize;
+  iouThreshold = inputs.iouThreshold;
+  scoreThreshold = inputs.scoreThreshold;
+  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);
+  const boxesVals = boxesAndScores[0];
+  const scoresVals = boxesAndScores[1];
+  const { selectedIndices } = nonMaxSuppressionV3Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return tensor1d(selectedIndices, "int32");
+}
+var nonMaxSuppressionAsync = nonMaxSuppressionAsync_;
+function nonMaxSuppressionWithScore_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  maxOutputSize = params.maxOutputSize;
+  iouThreshold = params.iouThreshold;
+  scoreThreshold = params.scoreThreshold;
+  softNmsSigma = params.softNmsSigma;
+  const inputs = { boxes: $boxes, scores: $scores };
+  const attrs = { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };
+  const result = ENGINE.runKernel(NonMaxSuppressionV5, inputs, attrs);
+  return { selectedIndices: result[0], selectedScores: result[1] };
+}
+var nonMaxSuppressionWithScore = op({ nonMaxSuppressionWithScore_ });
+async function nonMaxSuppressionWithScoreAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  maxOutputSize = params.maxOutputSize;
+  iouThreshold = params.iouThreshold;
+  scoreThreshold = params.scoreThreshold;
+  softNmsSigma = params.softNmsSigma;
+  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);
+  const boxesVals = boxesAndScores[0];
+  const scoresVals = boxesAndScores[1];
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return {
+    selectedIndices: tensor1d(selectedIndices, "int32"),
+    selectedScores: tensor1d(selectedScores)
+  };
+}
+var nonMaxSuppressionWithScoreAsync = nonMaxSuppressionWithScoreAsync_;
+function nonMaxSuppressionPadded_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);
+  const $maxOutputSize = params.maxOutputSize;
+  const $iouThreshold = params.iouThreshold;
+  const $scoreThreshold = params.scoreThreshold;
+  const inputs = { boxes: $boxes, scores: $scores };
+  const attrs = {
+    maxOutputSize: $maxOutputSize,
+    iouThreshold: $iouThreshold,
+    scoreThreshold: $scoreThreshold,
+    padToMaxOutputSize
+  };
+  const result = ENGINE.runKernel(NonMaxSuppressionV4, inputs, attrs);
+  return { selectedIndices: result[0], validOutputs: result[1] };
+}
+var nonMaxSuppressionPadded = op({ nonMaxSuppressionPadded_ });
+async function nonMaxSuppressionPaddedAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);
+  const $maxOutputSize = params.maxOutputSize;
+  const $iouThreshold = params.iouThreshold;
+  const $scoreThreshold = params.scoreThreshold;
+  const [boxesVals, scoresVals] = await Promise.all([$boxes.data(), $scores.data()]);
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl(boxesVals, scoresVals, $maxOutputSize, $iouThreshold, $scoreThreshold, padToMaxOutputSize);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return {
+    selectedIndices: tensor1d(selectedIndices, "int32"),
+    validOutputs: scalar(validOutputs, "int32")
+  };
+}
+var nonMaxSuppressionPaddedAsync = nonMaxSuppressionPaddedAsync_;
+function resizeBilinear_(images, size, alignCorners = false, halfPixelCenters = false) {
+  const $images = convertToTensor(images, "images", "resizeBilinear");
+  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeBilinear: x must be rank 3 or 4, but got rank ${$images.rank}.`);
+  assert(size.length === 2, () => `Error in resizeBilinear: new shape must 2D, but got shape ${size}.`);
+  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.`);
+  let batchImages = $images;
+  let reshapedTo4D = false;
+  if ($images.rank === 3) {
+    reshapedTo4D = true;
+    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);
+  }
+  const [] = size;
+  const inputs = { images: batchImages };
+  const attrs = { alignCorners, halfPixelCenters, size };
+  const res = ENGINE.runKernel(ResizeBilinear, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var resizeBilinear = op({ resizeBilinear_ });
+function resizeNearestNeighbor_(images, size, alignCorners = false, halfPixelCenters = false) {
+  const $images = convertToTensor(images, "images", "resizeNearestNeighbor");
+  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${$images.rank}.`);
+  assert(size.length === 2, () => `Error in resizeNearestNeighbor: new shape must 2D, but got shape ${size}.`);
+  assert($images.dtype === "float32" || $images.dtype === "int32", () => "`images` must have `int32` or `float32` as dtype");
+  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.`);
+  let batchImages = $images;
+  let reshapedTo4D = false;
+  if ($images.rank === 3) {
+    reshapedTo4D = true;
+    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);
+  }
+  const [] = size;
+  const inputs = { images: batchImages };
+  const attrs = { alignCorners, halfPixelCenters, size };
+  const res = ENGINE.runKernel(ResizeNearestNeighbor, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var resizeNearestNeighbor = op({ resizeNearestNeighbor_ });
+function threshold_(image2, method = "binary", inverted = false, threshValue = 0.5) {
+  const $image = convertToTensor(image2, "image", "threshold");
+  const RED_INTENCITY_COEF = 0.2989;
+  const GREEN_INTENCITY_COEF = 0.587;
+  const BLUE_INTENCITY_COEF = 0.114;
+  const totalPixelsInImage = $image.shape[0] * $image.shape[1];
+  let $threshold = mul(tensor1d([threshValue]), 255);
+  let r, g, b, grayscale;
+  assert($image.rank === 3, () => `Error in threshold: image must be rank 3,but got rank ${$image.rank}.`);
+  assert($image.shape[2] === 3 || $image.shape[2] === 1, () => `Error in threshold: image color channel must be equal to 3 or 1but got ${$image.shape[2]}.`);
+  assert($image.dtype === "int32" || $image.dtype === "float32", () => `Error in dtype: image dtype must be int32 or float32,but got dtype ${$image.dtype}.`);
+  assert(method === "otsu" || method === "binary", () => `Method must be binary or otsu, but was ${method}`);
+  if ($image.shape[2] === 3) {
+    [r, g, b] = split($image, [1, 1, 1], -1);
+    const $r = mul(r, RED_INTENCITY_COEF);
+    const $g = mul(g, GREEN_INTENCITY_COEF);
+    const $b = mul(b, BLUE_INTENCITY_COEF);
+    grayscale = add2(add2($r, $g), $b);
+  } else {
+    grayscale = image2;
+  }
+  if (method === "otsu") {
+    const $histogram = bincount(cast(round2(grayscale), "int32"), tensor([]), 256);
+    $threshold = otsu($histogram, totalPixelsInImage);
+  }
+  const invCondition = inverted ? lessEqual(grayscale, $threshold) : greater(grayscale, $threshold);
+  const result = cast(mul(invCondition, 255), "int32");
+  return result;
+}
+function otsu(histogram, total) {
+  let bestThresh = tensor1d([-1]);
+  let bestInBetVar = tensor1d([0]);
+  let cInBetVar = tensor1d([0]);
+  let classFirst, classSecond, meanFirst, meanSec, weightForeground, weightBack;
+  for (let index = 0; index < histogram.size - 1; index++) {
+    classFirst = slice(histogram, 0, index + 1);
+    classSecond = slice(histogram, index + 1);
+    weightForeground = div(sum2(classFirst), total);
+    weightBack = div(sum2(classSecond), total);
+    const meanFirstDivA = sum2(mul(classFirst, range(0, classFirst.size)));
+    meanFirst = div(meanFirstDivA, sum2(classFirst));
+    const meanSecFill = fill(classSecond.shape, classFirst.size);
+    const meanSecAdd = add2(range(0, classSecond.size), meanSecFill);
+    const meanSecMul = mul(classSecond, meanSecAdd);
+    meanSec = div(sum2(meanSecMul), sum2(classSecond));
+    const cInBetVarSubA = sub(meanFirst, meanSec);
+    const cInBetVarSubB = sub(meanFirst, meanSec);
+    const cInBetVarMul = mul(weightForeground, weightBack);
+    cInBetVar = mul(mul(cInBetVarMul, cInBetVarSubA), cInBetVarSubB);
+    const condition = greater(cInBetVar, bestInBetVar);
+    bestInBetVar = where(condition, cInBetVar, bestInBetVar);
+    bestThresh = where(condition, tensor1d([index]), bestThresh);
+  }
+  return bestThresh;
+}
+var threshold = op({ threshold_ });
+function transform_(image2, transforms, interpolation = "nearest", fillMode = "constant", fillValue = 0, outputShape) {
+  const $image = convertToTensor(image2, "image", "transform", "float32");
+  const $transforms = convertToTensor(transforms, "transforms", "transform", "float32");
+  assert($image.rank === 4, () => `Error in transform: image must be rank 4,but got rank ${$image.rank}.`);
+  assert($transforms.rank === 2 && ($transforms.shape[0] === $image.shape[0] || $transforms.shape[0] === 1) && $transforms.shape[1] === 8, () => `Error in transform: Input transform should be batch x 8 or 1 x 8`);
+  assert(outputShape == null || outputShape.length === 2, () => `Error in transform: outputShape must be [height, width] or null, but got ${outputShape}.`);
+  const inputs = { image: $image, transforms: $transforms };
+  const attrs = { interpolation, fillMode, fillValue, outputShape };
+  return ENGINE.runKernel(Transform, inputs, attrs);
+}
+var transform = op({ transform_ });
+function bandPart_(a, numLower, numUpper) {
+  assert(numLower % 1 === 0, () => `bandPart(): numLower must be an integer, got ${numLower}.`);
+  assert(numUpper % 1 === 0, () => `bandPart(): numUpper must be an integer, got ${numUpper}.`);
+  const $a = convertToTensor(a, "a", "bandPart");
+  assert($a.rank >= 2, () => `bandPart(): Rank must be at least 2, got ${$a.rank}.`);
+  const shape = $a.shape;
+  const [M, N] = $a.shape.slice(-2);
+  if (!(numLower <= M)) {
+    throw new Error(`bandPart(): numLower (${numLower}) must not be greater than the number of rows (${M}).`);
+  }
+  if (!(numUpper <= N)) {
+    throw new Error(`bandPart(): numUpper (${numUpper}) must not be greater than the number of columns (${N}).`);
+  }
+  if (numLower < 0) {
+    numLower = M;
+  }
+  if (numUpper < 0) {
+    numUpper = N;
+  }
+  const i = reshape(range(0, M, 1, "int32"), [-1, 1]);
+  const j = range(0, N, 1, "int32");
+  const ij = sub(i, j);
+  const inBand = logicalAnd(lessEqual(ij, scalar(+numLower, "int32")), greaterEqual(ij, scalar(-numUpper, "int32")));
+  const zero = zeros([M, N], $a.dtype);
+  return reshape(stack(unstack(reshape($a, [-1, M, N])).map((mat) => where(inBand, mat, zero))), shape);
+}
+var bandPart = op({ bandPart_ });
+function gramSchmidt_(xs) {
+  let inputIsTensor2D;
+  if (Array.isArray(xs)) {
+    inputIsTensor2D = false;
+    assert(xs != null && xs.length > 0, () => "Gram-Schmidt process: input must not be null, undefined, or empty");
+    const dim = xs[0].shape[0];
+    for (let i = 1; i < xs.length; ++i) {
+      assert(xs[i].shape[0] === dim, () => `Gram-Schmidt: Non-unique lengths found in the input vectors: (${xs[i].shape[0]} vs. ${dim})`);
+    }
+  } else {
+    inputIsTensor2D = true;
+    xs = split(xs, xs.shape[0], 0).map((x) => squeeze(x, [0]));
+  }
+  assert(xs.length <= xs[0].shape[0], () => `Gram-Schmidt: Number of vectors (${xs.length}) exceeds number of dimensions (${xs[0].shape[0]}).`);
+  const ys = [];
+  const xs1d = xs;
+  for (let i = 0; i < xs.length; ++i) {
+    ys.push(ENGINE.tidy(() => {
+      let x = xs1d[i];
+      if (i > 0) {
+        for (let j = 0; j < i; ++j) {
+          const proj = mul(sum2(mul(ys[j], x)), ys[j]);
+          x = sub(x, proj);
+        }
+      }
+      return div(x, norm(x, "euclidean"));
+    }));
+  }
+  if (inputIsTensor2D) {
+    return stack(ys, 0);
+  } else {
+    return ys;
+  }
+}
+var gramSchmidt = op({ gramSchmidt_ });
+function qr_(x, fullMatrices = false) {
+  assert(x.rank >= 2, () => `qr() requires input tensor to have a rank >= 2, but got rank ${x.rank}`);
+  if (x.rank === 2) {
+    return qr2d(x, fullMatrices);
+  } else {
+    const outerDimsProd = x.shape.slice(0, x.shape.length - 2).reduce((value, prev) => value * prev);
+    const x2ds = unstack(reshape(x, [
+      outerDimsProd,
+      x.shape[x.shape.length - 2],
+      x.shape[x.shape.length - 1]
+    ]), 0);
+    const q2ds = [];
+    const r2ds = [];
+    x2ds.forEach((x2d) => {
+      const [q2d, r2d] = qr2d(x2d, fullMatrices);
+      q2ds.push(q2d);
+      r2ds.push(r2d);
+    });
+    const q = reshape(stack(q2ds, 0), x.shape);
+    const r = reshape(stack(r2ds, 0), x.shape);
+    return [q, r];
+  }
+}
+function qr2d(x, fullMatrices = false) {
+  return ENGINE.tidy(() => {
+    assert(x.shape.length === 2, () => `qr2d() requires a 2D Tensor, but got a ${x.shape.length}D Tensor.`);
+    const m = x.shape[0];
+    const n = x.shape[1];
+    let q = eye(m);
+    let r = clone(x);
+    const one2D = tensor2d([[1]], [1, 1]);
+    let w = clone(one2D);
+    const iters = m >= n ? n : m;
+    for (let j = 0; j < iters; ++j) {
+      const rTemp = r;
+      const wTemp = w;
+      const qTemp = q;
+      [w, r, q] = ENGINE.tidy(() => {
+        const rjEnd1 = slice(r, [j, j], [m - j, 1]);
+        const normX = norm(rjEnd1);
+        const rjj = slice(r, [j, j], [1, 1]);
+        const s = where(greater(rjj, 0), tensor2d([[-1]]), tensor2d([[1]]));
+        const u1 = sub(rjj, mul(s, normX));
+        const wPre = div(rjEnd1, u1);
+        if (wPre.shape[0] === 1) {
+          w = clone(one2D);
+        } else {
+          w = concat([
+            one2D,
+            slice(wPre, [1, 0], [wPre.shape[0] - 1, wPre.shape[1]])
+          ], 0);
+        }
+        const tau = neg(div(matMul(s, u1), normX));
+        const rjEndAll = slice(r, [j, 0], [m - j, n]);
+        const tauTimesW = mul(tau, w);
+        const wT = transpose(w);
+        if (j === 0) {
+          r = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));
+        } else {
+          const rTimesTau = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));
+          r = concat([slice(r, [0, 0], [j, n]), rTimesTau], 0);
+        }
+        const tawTimesWT = transpose(tauTimesW);
+        const qAllJEnd = slice(q, [0, j], [m, q.shape[1] - j]);
+        if (j === 0) {
+          q = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));
+        } else {
+          const qTimesTau = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));
+          q = concat([slice(q, [0, 0], [m, j]), qTimesTau], 1);
+        }
+        return [w, r, q];
+      });
+      dispose([rTemp, wTemp, qTemp]);
+    }
+    if (!fullMatrices && m > n) {
+      q = slice(q, [0, 0], [m, n]);
+      r = slice(r, [0, 0], [n, n]);
+    }
+    return [q, r];
+  });
+}
+var qr = op({ qr_ });
+var Reduction;
+(function(Reduction2) {
+  Reduction2[Reduction2["NONE"] = 0] = "NONE";
+  Reduction2[Reduction2["MEAN"] = 1] = "MEAN";
+  Reduction2[Reduction2["SUM"] = 2] = "SUM";
+  Reduction2[Reduction2["SUM_BY_NONZERO_WEIGHTS"] = 3] = "SUM_BY_NONZERO_WEIGHTS";
+})(Reduction || (Reduction = {}));
+function computeWeightedLoss_(losses2, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $losses = convertToTensor(losses2, "losses", "computeWeightedLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "computeWeightedLoss");
+  }
+  const weightedLoss = $weights == null ? $losses : mul($losses, $weights);
+  if (reduction === Reduction.NONE) {
+    return weightedLoss;
+  }
+  if (reduction === Reduction.SUM) {
+    return sum2(weightedLoss);
+  }
+  if (reduction === Reduction.MEAN) {
+    if ($weights == null) {
+      return mean(weightedLoss);
+    } else {
+      const broadcastFactor = $losses.size / $weights.size;
+      const result = div(sum2(weightedLoss), sum2($weights));
+      return broadcastFactor > 1 ? div(result, scalar(broadcastFactor)) : result;
+    }
+  }
+  if (reduction === Reduction.SUM_BY_NONZERO_WEIGHTS) {
+    if ($weights == null) {
+      return div(sum2(weightedLoss), scalar($losses.size));
+    } else {
+      const broadcastedWeights = mul($weights, ones2($losses.shape));
+      const numNonZeros = cast(sum2(notEqual(broadcastedWeights, scalar(0))), "float32");
+      return div(sum2(weightedLoss), numNonZeros);
+    }
+  }
+  throw Error(`Unknown reduction: ${reduction}`);
+}
+var computeWeightedLoss = op({ computeWeightedLoss_ });
+function absoluteDifference_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "absoluteDifference");
+  const $predictions = convertToTensor(predictions, "predictions", "absoluteDifference");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "absoluteDifference");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in absoluteDifference: ");
+  const losses2 = abs(sub($labels, $predictions));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var absoluteDifference = op({ absoluteDifference_ });
+function cosineDistance_(labels, predictions, axis, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "cosineDistance");
+  const $predictions = convertToTensor(predictions, "predictions", "cosineDistance");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "cosineDistance");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in cosineDistance: ");
+  const one = scalar(1);
+  const losses2 = sub(one, sum2(mul($labels, $predictions), axis, true));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var cosineDistance = op({ cosineDistance_ });
+function hingeLoss_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $labels = convertToTensor(labels, "labels", "hingeLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "hingeLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "hingeLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in hingeLoss: ");
+  const one = scalar(1);
+  $labels = sub(mul(scalar(2), $labels), one);
+  const losses2 = relu(sub(one, mul($labels, $predictions)));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var hingeLoss = op({ hingeLoss_ });
+function huberLoss_(labels, predictions, weights, delta = 1, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "huberLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "huberLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "huberLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in huberLoss: ");
+  const deltaScalar = scalar(delta);
+  const error = abs(sub($predictions, $labels));
+  const quadratic = minimum(error, deltaScalar);
+  const linear = sub(error, quadratic);
+  const losses2 = add2(mul(scalar(0.5), square(quadratic)), mul(deltaScalar, linear));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var huberLoss = op({ huberLoss_ });
+function logLoss_(labels, predictions, weights, epsilon32 = 1e-7, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "logLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "logLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "logLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in logLoss: ");
+  const one = scalar(1);
+  const epsilonScalar = scalar(epsilon32);
+  const l13 = neg(mul($labels, log2(add2($predictions, epsilonScalar))));
+  const l23 = mul(sub(one, $labels), log2(add2(sub(one, $predictions), epsilonScalar)));
+  const losses2 = sub(l13, l23);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var logLoss = op({ logLoss_ });
+function meanSquaredError_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "meanSquaredError");
+  const $predictions = convertToTensor(predictions, "predictions", "meanSquaredError");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "meanSquaredError");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in meanSquaredError: ");
+  const losses2 = squaredDifference($labels, $predictions);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var meanSquaredError = op({ meanSquaredError_ });
+function sigmoidCrossEntropyWithLogits_(labels, logits) {
+  const $labels = convertToTensor(labels, "labels", "sigmoidCrossEntropyWithLogits");
+  const $logits = convertToTensor(logits, "logits", "sigmoidCrossEntropyWithLogits");
+  assertShapesMatch($labels.shape, $logits.shape, "Error in sigmoidCrossEntropyWithLogits: ");
+  const maxOutput = relu($logits);
+  const outputXTarget = mul($logits, $labels);
+  const sigmoidOutput = log1p(exp(neg(abs($logits))));
+  return add2(sub(maxOutput, outputXTarget), sigmoidOutput);
+}
+function sigmoidCrossEntropy_(multiClassLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $multiClassLabels = convertToTensor(multiClassLabels, "multiClassLabels", "sigmoidCrossEntropy");
+  const $logits = convertToTensor(logits, "logits", "sigmoidCrossEntropy");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "sigmoidCrossEntropy");
+  }
+  assertShapesMatch($multiClassLabels.shape, $logits.shape, "Error in sigmoidCrossEntropy: ");
+  if (labelSmoothing > 0) {
+    const labelSmoothingScalar = scalar(labelSmoothing);
+    const one = scalar(1);
+    const half = scalar(0.5);
+    $multiClassLabels = add2(mul($multiClassLabels, sub(one, labelSmoothingScalar)), mul(half, labelSmoothingScalar));
+  }
+  const losses2 = sigmoidCrossEntropyWithLogits_($multiClassLabels, $logits);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var sigmoidCrossEntropy = op({ sigmoidCrossEntropy_ });
+function softmaxCrossEntropyWithLogits_(labels, logits, dim = -1) {
+  if (dim === -1) {
+    dim = logits.rank - 1;
+  }
+  if (dim !== logits.rank - 1) {
+    throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${logits.rank} and dim was ${dim}`);
+  }
+  const customOp = customGrad((labels2, logits2, save) => {
+    const keepDims = true;
+    const lse = logSumExp(logits2, [dim], keepDims);
+    const logResult = sub(cast(logits2, "float32"), lse);
+    save([labels2, logResult]);
+    const costVector = neg(mul(logResult, labels2));
+    const value = sum2(costVector, [dim]);
+    const gradFunc = (dy, saved) => {
+      const [labels3, logResult2] = saved;
+      const dyShape = expandShapeToKeepDim(dy.shape, [dim]);
+      return [
+        mul(reshape(dy, dyShape), sub(cast(labels3, "float32"), exp(logResult2))),
+        mul(reshape(dy, dyShape), sub(exp(logResult2), cast(labels3, "float32")))
+      ];
+    };
+    return { value, gradFunc };
+  });
+  return customOp(labels, logits);
+}
+function softmaxCrossEntropy_(onehotLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $onehotLabels = convertToTensor(onehotLabels, "onehotLabels", "softmaxCrossEntropy");
+  const $logits = convertToTensor(logits, "logits", "softmaxCrossEntropy");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "softmaxCrossEntropy");
+  }
+  assertShapesMatch($onehotLabels.shape, $logits.shape, "Error in softmaxCrossEntropy: ");
+  if (labelSmoothing > 0) {
+    const labelSmoothingScalar = scalar(labelSmoothing);
+    const one = scalar(1);
+    const numClasses = scalar($onehotLabels.shape[1]);
+    $onehotLabels = add2(mul($onehotLabels, sub(one, labelSmoothingScalar)), div(labelSmoothingScalar, numClasses));
+  }
+  const losses2 = softmaxCrossEntropyWithLogits_($onehotLabels, $logits);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var softmaxCrossEntropy = op({ softmaxCrossEntropy_ });
+function sparseFillEmptyRows_(indices, values, denseShape, defaultValue) {
+  const $indices = convertToTensor(indices, "indices", "sparseFillEmptyRows", "int32");
+  const $values = convertToTensor(values, "values", "sparseFillEmptyRows");
+  const $denseShape = convertToTensor(denseShape, "denseShape", "sparseFillEmptyRows", "int32");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "sparseFillEmptyRows", $values.dtype);
+  if ($indices.rank !== 2) {
+    throw new Error(`Indices should be Tensor2D but received shape
+        ${$indices.shape}`);
+  }
+  if ($values.rank !== 1) {
+    throw new Error(`Values should be Tensor1D but received shape ${$values.shape}`);
+  }
+  if ($denseShape.rank !== 1) {
+    throw new Error(`Dense shape should be Tensor1D but received shape ${$denseShape.shape}`);
+  }
+  if ($defaultValue.rank !== 0) {
+    throw new Error(`Default value should be a scalar but received shape ${$defaultValue.shape}`);
+  }
+  const inputs = {
+    indices: $indices,
+    values: $values,
+    denseShape: $denseShape,
+    defaultValue: $defaultValue
+  };
+  const result = ENGINE.runKernel(SparseFillEmptyRows, inputs);
+  return {
+    outputIndices: result[0],
+    outputValues: result[1],
+    emptyRowIndicator: result[2],
+    reverseIndexMap: result[3]
+  };
+}
+var sparseFillEmptyRows = op({ sparseFillEmptyRows_ });
+function sparseReshape_(inputIndices, inputShape, newShape) {
+  const $inputIndices = convertToTensor(inputIndices, "inputIndices", "sparseReshape", "int32");
+  const $inputShape = convertToTensor(inputShape, "inputShape", "sparseReshape", "int32");
+  const $newShape = convertToTensor(newShape, "newShape", "sparseReshape", "int32");
+  if ($inputIndices.rank !== 2) {
+    throw new Error(`Input indices should be Tensor2D but received shape
+        ${$inputIndices.shape}`);
+  }
+  if ($inputShape.rank !== 1) {
+    throw new Error(`Input shape should be Tensor1D but received shape ${$inputShape.shape}`);
+  }
+  if ($newShape.rank !== 1) {
+    throw new Error(`New shape should be Tensor1D but received shape ${$newShape.shape}`);
+  }
+  const inputs = {
+    inputIndices: $inputIndices,
+    inputShape: $inputShape,
+    newShape: $newShape
+  };
+  const result = ENGINE.runKernel(SparseReshape, inputs);
+  return { outputIndices: result[0], outputShape: result[1] };
+}
+var sparseReshape = op({ sparseReshape_ });
+function sparseSegmentMean_(data, indices, segmentIds) {
+  const $data = convertToTensor(data, "data", "sparseSegmentMean");
+  const $indices = convertToTensor(indices, "indices", "sparseSegmentMean", "int32");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "sparseSegmentMean", "int32");
+  if ($data.rank < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if ($indices.rank !== 1) {
+    throw new Error(`Indices should be Tensor1D but received shape
+          ${$indices.shape}`);
+  }
+  if ($segmentIds.rank !== 1) {
+    throw new Error(`Segment ids should be Tensor1D but received shape
+          ${$segmentIds.shape}`);
+  }
+  const inputs = {
+    data: $data,
+    indices: $indices,
+    segmentIds: $segmentIds
+  };
+  return ENGINE.runKernel(SparseSegmentMean, inputs);
+}
+var sparseSegmentMean = op({ sparseSegmentMean_ });
+function sparseSegmentSum_(data, indices, segmentIds) {
+  const $data = convertToTensor(data, "data", "sparseSegmentSum");
+  const $indices = convertToTensor(indices, "indices", "sparseSegmentSum", "int32");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "sparseSegmentSum", "int32");
+  if ($data.rank < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if ($indices.rank !== 1) {
+    throw new Error(`Indices should be Tensor1D but received shape
+         ${$indices.shape}`);
+  }
+  if ($segmentIds.rank !== 1) {
+    throw new Error(`Segment ids should be Tensor1D but received shape
+         ${$segmentIds.shape}`);
+  }
+  const inputs = {
+    data: $data,
+    indices: $indices,
+    segmentIds: $segmentIds
+  };
+  return ENGINE.runKernel(SparseSegmentSum, inputs);
+}
+var sparseSegmentSum = op({ sparseSegmentSum_ });
+function stringNGrams_(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+  const $data = convertToTensor(data, "data", "stringNGrams", "string");
+  if ($data.dtype !== "string") {
+    throw new Error("Data must be of datatype string");
+  }
+  if ($data.shape.length !== 1) {
+    throw new Error(`Data must be a vector, saw: ${$data.shape}`);
+  }
+  const $dataSplits = convertToTensor(dataSplits, "dataSplits", "stringNGrams");
+  if ($dataSplits.dtype !== "int32") {
+    throw new Error("Data splits must be of datatype int32");
+  }
+  const attrs = {
+    separator,
+    nGramWidths,
+    leftPad,
+    rightPad: rightPad2,
+    padWidth,
+    preserveShortSequences
+  };
+  const inputs = { data: $data, dataSplits: $dataSplits };
+  const result = ENGINE.runKernel(StringNGrams, inputs, attrs);
+  return { nGrams: result[0], nGramsSplits: result[1] };
+}
+var stringNGrams = op({ stringNGrams_ });
+function stringSplit_(input2, delimiter, skipEmpty = true) {
+  const $input = convertToTensor(input2, "input", "stringSplit", "string");
+  const $delimiter = convertToTensor(delimiter, "delimiter", "stringSplit", "string");
+  if ($input.rank !== 1) {
+    throw new Error(`Input should be Tensor1D but received shape ${$input.shape}`);
+  }
+  if ($delimiter.rank !== 0) {
+    throw new Error(`Delimiter should be a scalar but received shape ${$delimiter.shape}`);
+  }
+  const attrs = { skipEmpty };
+  const inputs = { input: $input, delimiter: $delimiter };
+  const result = ENGINE.runKernel(StringSplit, inputs, attrs);
+  return { indices: result[0], values: result[1], shape: result[2] };
+}
+var stringSplit = op({ stringSplit_ });
+function stringToHashBucketFast_(input2, numBuckets) {
+  const $input = convertToTensor(input2, "input", "stringToHashBucketFast", "string");
+  const attrs = { numBuckets };
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const inputs = { input: $input };
+  return ENGINE.runKernel(StringToHashBucketFast, inputs, attrs);
+}
+var stringToHashBucketFast = op({ stringToHashBucketFast_ });
+var spectral = {
+  fft,
+  ifft,
+  rfft,
+  irfft
+};
+var signal = {
+  hammingWindow,
+  hannWindow,
+  frame,
+  stft
+};
+var image = {
+  flipLeftRight,
+  grayscaleToRGB,
+  resizeNearestNeighbor,
+  resizeBilinear,
+  rotateWithOffset,
+  cropAndResize,
+  nonMaxSuppression,
+  nonMaxSuppressionAsync,
+  nonMaxSuppressionWithScore,
+  nonMaxSuppressionWithScoreAsync,
+  nonMaxSuppressionPadded,
+  nonMaxSuppressionPaddedAsync,
+  threshold,
+  transform
+};
+var linalg = {
+  bandPart,
+  gramSchmidt,
+  qr
+};
+var losses = {
+  absoluteDifference,
+  computeWeightedLoss,
+  cosineDistance,
+  hingeLoss,
+  huberLoss,
+  logLoss,
+  meanSquaredError,
+  sigmoidCrossEntropy,
+  softmaxCrossEntropy
+};
+var sparse = {
+  sparseFillEmptyRows,
+  sparseReshape,
+  sparseSegmentMean,
+  sparseSegmentSum
+};
+var string = {
+  stringNGrams,
+  stringSplit,
+  stringToHashBucketFast
+};
+var Optimizer = class extends Serializable {
+  minimize(f, returnCost = false, varList) {
+    const { value, grads: grads2 } = this.computeGradients(f, varList);
+    if (varList != null) {
+      const gradArray = varList.map((v) => ({ name: v.name, tensor: grads2[v.name] }));
+      this.applyGradients(gradArray);
+    } else {
+      this.applyGradients(grads2);
+    }
+    dispose(grads2);
+    if (returnCost) {
+      return value;
+    } else {
+      value.dispose();
+      return null;
+    }
+  }
+  get iterations() {
+    if (this.iterations_ == null) {
+      this.iterations_ = 0;
+    }
+    return this.iterations_;
+  }
+  incrementIterations() {
+    this.iterations_ = this.iterations + 1;
+  }
+  computeGradients(f, varList) {
+    return variableGrads(f, varList);
+  }
+  dispose() {
+    if (this.iterations_ != null) {
+      dispose(this.iterations_);
+    }
+  }
+  async saveIterations() {
+    if (this.iterations_ == null) {
+      this.iterations_ = 0;
+    }
+    return {
+      name: "iter",
+      tensor: scalar(this.iterations_, "int32")
+    };
+  }
+  async getWeights() {
+    throw new Error("getWeights() is not implemented for this optimizer yet.");
+  }
+  async setWeights(weightValues) {
+    throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`);
+  }
+  async extractIterations(weightValues) {
+    this.iterations_ = (await weightValues[0].tensor.data())[0];
+    return weightValues.slice(1);
+  }
+};
+Object.defineProperty(Optimizer, Symbol.hasInstance, {
+  value: (instance) => {
+    return instance.minimize != null && instance.computeGradients != null && instance.applyGradients != null;
+  }
+});
+var AdadeltaOptimizer = class extends Optimizer {
+  constructor(learningRate, rho, epsilon32 = null) {
+    super();
+    this.learningRate = learningRate;
+    this.rho = rho;
+    this.epsilon = epsilon32;
+    this.accumulatedGrads = [];
+    this.accumulatedUpdates = [];
+    if (epsilon32 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      const trainable = false;
+      if (this.accumulatedGrads[i] == null) {
+        this.accumulatedGrads[i] = {
+          originalName: `${name}/accum_grad`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedUpdates[i] == null) {
+        this.accumulatedUpdates[i] = {
+          originalName: `${name}/accum_var`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedGrad = this.accumulatedGrads[i].variable;
+      const accumulatedUpdate = this.accumulatedUpdates[i].variable;
+      tidy(() => {
+        const newAccumulatedGrad = add2(mul(accumulatedGrad, this.rho), mul(square(gradient), 1 - this.rho));
+        const updates = mul(div(sqrt(add2(accumulatedUpdate, this.epsilon)), sqrt(add2(accumulatedGrad, this.epsilon))), gradient);
+        const newAccumulatedUpdate = add2(mul(accumulatedUpdate, this.rho), mul(square(updates), 1 - this.rho));
+        accumulatedGrad.assign(newAccumulatedGrad);
+        accumulatedUpdate.assign(newAccumulatedUpdate);
+        const newValue = add2(mul(updates, -this.learningRate), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedUpdates != null) {
+      dispose(this.accumulatedGrads.map((v) => v.variable));
+      dispose(this.accumulatedUpdates.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedGrads, ...this.accumulatedUpdates];
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const variableCount = weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedGrads = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedUpdates = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "rho": this.rho,
+      "epsilon": this.epsilon
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["rho"], config["epsilon"]);
+  }
+};
+AdadeltaOptimizer.className = "Adadelta";
+registerClass(AdadeltaOptimizer);
+var AdagradOptimizer = class extends Optimizer {
+  constructor(learningRate, initialAccumulatorValue = 0.1) {
+    super();
+    this.learningRate = learningRate;
+    this.initialAccumulatorValue = initialAccumulatorValue;
+    this.accumulatedGrads = [];
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      if (this.accumulatedGrads[i] == null) {
+        const trainable = false;
+        this.accumulatedGrads[i] = {
+          originalName: `${name}/accumulator`,
+          variable: tidy(() => fill(value.shape, this.initialAccumulatorValue).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedGrad = this.accumulatedGrads[i].variable;
+      tidy(() => {
+        const newAccumulatedGrad = add2(accumulatedGrad, square(gradient));
+        accumulatedGrad.assign(newAccumulatedGrad);
+        const newValue = add2(mul(div(gradient, sqrt(add2(newAccumulatedGrad, ENGINE.backend.epsilon()))), -this.learningRate), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedGrads != null) {
+      dispose(this.accumulatedGrads.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    return [await this.saveIterations()].concat(this.accumulatedGrads.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const trainable = false;
+    this.accumulatedGrads = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "initialAccumulatorValue": this.initialAccumulatorValue
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["initialAccumulatorValue"]);
+  }
+};
+AdagradOptimizer.className = "Adagrad";
+registerClass(AdagradOptimizer);
+var AdamOptimizer = class extends Optimizer {
+  constructor(learningRate, beta1, beta2, epsilon32 = null) {
+    super();
+    this.learningRate = learningRate;
+    this.beta1 = beta1;
+    this.beta2 = beta2;
+    this.epsilon = epsilon32;
+    this.accumulatedFirstMoment = [];
+    this.accumulatedSecondMoment = [];
+    tidy(() => {
+      this.accBeta1 = scalar(beta1).variable();
+      this.accBeta2 = scalar(beta2).variable();
+    });
+    if (epsilon32 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);
+    tidy(() => {
+      const oneMinusAccBeta1 = sub(1, this.accBeta1);
+      const oneMinusAccBeta2 = sub(1, this.accBeta2);
+      varNames.forEach((name, i) => {
+        const value = ENGINE.registeredVariables[name];
+        const trainable = false;
+        if (this.accumulatedFirstMoment[i] == null) {
+          this.accumulatedFirstMoment[i] = {
+            originalName: `${name}/m`,
+            variable: tidy(() => zerosLike(value).variable(trainable))
+          };
+        }
+        if (this.accumulatedSecondMoment[i] == null) {
+          this.accumulatedSecondMoment[i] = {
+            originalName: `${name}/v`,
+            variable: tidy(() => zerosLike(value).variable(trainable))
+          };
+        }
+        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+        if (gradient == null) {
+          return;
+        }
+        const firstMoment = this.accumulatedFirstMoment[i].variable;
+        const secondMoment = this.accumulatedSecondMoment[i].variable;
+        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));
+        const newSecondMoment = add2(mul(secondMoment, this.beta2), mul(square(gradient), 1 - this.beta2));
+        const biasCorrectedFirstMoment = div(newFirstMoment, oneMinusAccBeta1);
+        const biasCorrectedSecondMoment = div(newSecondMoment, oneMinusAccBeta2);
+        firstMoment.assign(newFirstMoment);
+        secondMoment.assign(newSecondMoment);
+        const newValue = add2(mul(div(biasCorrectedFirstMoment, add2(sqrt(biasCorrectedSecondMoment), this.epsilon)), -this.learningRate), value);
+        value.assign(newValue);
+      });
+      this.accBeta1.assign(mul(this.accBeta1, this.beta1));
+      this.accBeta2.assign(mul(this.accBeta2, this.beta2));
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.accBeta1.dispose();
+    this.accBeta2.dispose();
+    if (this.accumulatedFirstMoment != null) {
+      dispose(this.accumulatedFirstMoment.map((v) => v.variable));
+    }
+    if (this.accumulatedSecondMoment != null) {
+      dispose(this.accumulatedSecondMoment.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedFirstMoment, ...this.accumulatedSecondMoment];
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    tidy(() => {
+      this.accBeta1.assign(pow(this.beta1, this.iterations_ + 1));
+      this.accBeta2.assign(pow(this.beta2, this.iterations_ + 1));
+    });
+    const variableCount = weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedFirstMoment = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedSecondMoment = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "beta1": this.beta1,
+      "beta2": this.beta2,
+      "epsilon": this.epsilon
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["beta1"], config["beta2"], config["epsilon"]);
+  }
+};
+AdamOptimizer.className = "Adam";
+registerClass(AdamOptimizer);
+var AdamaxOptimizer = class extends Optimizer {
+  constructor(learningRate, beta1, beta2, epsilon32 = null, decay = 0) {
+    super();
+    this.learningRate = learningRate;
+    this.beta1 = beta1;
+    this.beta2 = beta2;
+    this.epsilon = epsilon32;
+    this.decay = decay;
+    this.accumulatedFirstMoment = [];
+    this.accumulatedWeightedInfNorm = [];
+    tidy(() => {
+      this.iteration = scalar(0).variable();
+      this.accBeta1 = scalar(beta1).variable();
+    });
+    if (epsilon32 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    tidy(() => {
+      const oneMinusAccBeta1 = sub(1, this.accBeta1);
+      const lr = div(-this.learningRate, add2(mul(this.iteration, this.decay), 1));
+      variableNames.forEach((name, i) => {
+        const value = ENGINE.registeredVariables[name];
+        const trainable = false;
+        if (this.accumulatedFirstMoment[i] == null) {
+          this.accumulatedFirstMoment[i] = {
+            originalName: `${name}/m`,
+            variable: zerosLike(value).variable(trainable)
+          };
+        }
+        if (this.accumulatedWeightedInfNorm[i] == null) {
+          this.accumulatedWeightedInfNorm[i] = {
+            originalName: `${name}/v`,
+            variable: zerosLike(value).variable(trainable)
+          };
+        }
+        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+        if (gradient == null) {
+          return;
+        }
+        const firstMoment = this.accumulatedFirstMoment[i].variable;
+        const weightedInfNorm = this.accumulatedWeightedInfNorm[i].variable;
+        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));
+        const ut0 = mul(weightedInfNorm, this.beta2);
+        const ut1 = abs(gradient);
+        const newWeightedInfNorm = maximum(ut0, ut1);
+        firstMoment.assign(newFirstMoment);
+        weightedInfNorm.assign(newWeightedInfNorm);
+        const newValue = add2(mul(div(lr, oneMinusAccBeta1), div(newFirstMoment, add2(newWeightedInfNorm, this.epsilon))), value);
+        value.assign(newValue);
+      });
+      this.iteration.assign(add2(this.iteration, 1));
+      this.accBeta1.assign(mul(this.accBeta1, this.beta1));
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.accBeta1.dispose();
+    this.iteration.dispose();
+    if (this.accumulatedFirstMoment != null) {
+      dispose(this.accumulatedFirstMoment.map((v) => v.variable));
+    }
+    if (this.accumulatedWeightedInfNorm != null) {
+      dispose(this.accumulatedWeightedInfNorm.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    throw new Error("getWeights() is not implemented for Adamax yet.");
+  }
+  async setWeights(weightValues) {
+    throw new Error("setWeights() is not implemented for Adamax yet.");
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "beta1": this.beta1,
+      "beta2": this.beta2,
+      "epsilon": this.epsilon,
+      "decay": this.decay
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["beta1"], config["beta2"], config["epsilon"], config["decay"]);
+  }
+};
+AdamaxOptimizer.className = "Adamax";
+registerClass(AdamaxOptimizer);
+var SGDOptimizer = class extends Optimizer {
+  constructor(learningRate) {
+    super();
+    this.learningRate = learningRate;
+    this.setLearningRate(learningRate);
+  }
+  applyGradients(variableGradients) {
+    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);
+    varNames.forEach((name, i) => {
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const value = ENGINE.registeredVariables[name];
+      tidy(() => {
+        const newValue = add2(mul(this.c, gradient), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  setLearningRate(learningRate) {
+    this.learningRate = learningRate;
+    if (this.c != null) {
+      this.c.dispose();
+    }
+    this.c = keep(scalar(-learningRate));
+  }
+  dispose() {
+    this.c.dispose();
+  }
+  async getWeights() {
+    return [await this.saveIterations()];
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    if (weightValues.length !== 0) {
+      throw new Error("SGD optimizer does not have settable weights.");
+    }
+  }
+  getConfig() {
+    return { "learningRate": this.learningRate };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"]);
+  }
+};
+SGDOptimizer.className = "SGD";
+registerClass(SGDOptimizer);
+var MomentumOptimizer = class extends SGDOptimizer {
+  constructor(learningRate, momentum, useNesterov = false) {
+    super(learningRate);
+    this.learningRate = learningRate;
+    this.momentum = momentum;
+    this.useNesterov = useNesterov;
+    this.accumulations = [];
+    this.m = scalar(this.momentum);
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      if (this.accumulations[i] == null) {
+        const trainable = false;
+        this.accumulations[i] = {
+          originalName: `${name}/momentum`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const accumulation = this.accumulations[i].variable;
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      tidy(() => {
+        let newValue;
+        const newAccumulation = add2(mul(this.m, accumulation), gradient);
+        if (this.useNesterov) {
+          newValue = add2(mul(this.c, add2(gradient, mul(newAccumulation, this.m))), value);
+        } else {
+          newValue = add2(mul(this.c, newAccumulation), value);
+        }
+        accumulation.assign(newAccumulation);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.m.dispose();
+    if (this.accumulations != null) {
+      dispose(this.accumulations.map((v) => v.variable));
+    }
+  }
+  setMomentum(momentum) {
+    this.momentum = momentum;
+  }
+  async getWeights() {
+    return [await this.saveIterations()].concat(this.accumulations.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const trainable = false;
+    this.accumulations = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "momentum": this.momentum,
+      "useNesterov": this.useNesterov
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["momentum"], config["useNesterov"]);
+  }
+};
+MomentumOptimizer.className = "Momentum";
+registerClass(MomentumOptimizer);
+var RMSPropOptimizer = class extends Optimizer {
+  constructor(learningRate, decay = 0.9, momentum = 0, epsilon32 = null, centered = false) {
+    super();
+    this.learningRate = learningRate;
+    this.decay = decay;
+    this.momentum = momentum;
+    this.epsilon = epsilon32;
+    this.accumulatedMeanSquares = [];
+    this.accumulatedMoments = [];
+    this.accumulatedMeanGrads = [];
+    this.centered = centered;
+    if (epsilon32 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+    if (learningRate == null) {
+      throw new Error(`learningRate for RMSPropOptimizer must be defined.`);
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      const trainable = false;
+      if (this.accumulatedMeanSquares[i] == null) {
+        this.accumulatedMeanSquares[i] = {
+          originalName: `${name}/rms`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedMoments[i] == null) {
+        this.accumulatedMoments[i] = {
+          originalName: `${name}/momentum`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedMeanGrads[i] == null && this.centered) {
+        this.accumulatedMeanGrads[i] = {
+          originalName: `${name}/mg`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedMeanSquare = this.accumulatedMeanSquares[i].variable;
+      const accumulatedMoments = this.accumulatedMoments[i].variable;
+      tidy(() => {
+        const newAccumulatedMeanSquare = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));
+        if (this.centered) {
+          const accumulatedMeanGrad = this.accumulatedMeanGrads[i].variable;
+          const newAccumulatedMeanGrad = add2(mul(accumulatedMeanGrad, this.decay), mul(gradient, 1 - this.decay));
+          const gradContribution = div(mul(gradient, this.learningRate), sqrt(sub(newAccumulatedMeanSquare, add2(square(newAccumulatedMeanGrad), this.epsilon))));
+          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), gradContribution);
+          accumulatedMeanSquare.assign(newAccumulatedMeanSquare);
+          accumulatedMeanGrad.assign(newAccumulatedMeanGrad);
+          accumulatedMoments.assign(newAccumulatedMoments);
+          const newValue = sub(value, newAccumulatedMoments);
+          value.assign(newValue);
+        } else {
+          const newAccumulatedMeanSquare2 = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));
+          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), div(mul(gradient, this.learningRate), sqrt(add2(newAccumulatedMeanSquare2, this.epsilon))));
+          accumulatedMeanSquare.assign(newAccumulatedMeanSquare2);
+          accumulatedMoments.assign(newAccumulatedMoments);
+          const newValue = sub(value, newAccumulatedMoments);
+          value.assign(newValue);
+        }
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedMeanSquares != null) {
+      dispose(this.accumulatedMeanSquares.map((v) => v.variable));
+    }
+    if (this.accumulatedMeanGrads != null && this.centered) {
+      dispose(this.accumulatedMeanGrads.map((v) => v.variable));
+    }
+    if (this.accumulatedMoments != null) {
+      dispose(this.accumulatedMoments.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedMeanSquares, ...this.accumulatedMoments];
+    if (this.centered) {
+      variables.push(...this.accumulatedMeanGrads);
+    }
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const variableCount = this.centered ? weightValues.length / 3 : weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedMeanSquares = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedMoments = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    if (this.centered) {
+      this.accumulatedMeanGrads = weightValues.slice(variableCount * 2, variableCount * 3).map((v) => ({
+        originalName: v.name,
+        variable: v.tensor.variable(trainable)
+      }));
+    }
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "decay": this.decay,
+      "momentum": this.momentum,
+      "epsilon": this.epsilon,
+      "centered": this.centered
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["decay"], config["momentum"], config["epsilon"], config["centered"]);
+  }
+};
+RMSPropOptimizer.className = "RMSProp";
+registerClass(RMSPropOptimizer);
+var OptimizerConstructors = class {
+  static sgd(learningRate) {
+    return new SGDOptimizer(learningRate);
+  }
+  static momentum(learningRate, momentum, useNesterov = false) {
+    return new MomentumOptimizer(learningRate, momentum, useNesterov);
+  }
+  static rmsprop(learningRate, decay = 0.9, momentum = 0, epsilon32 = null, centered = false) {
+    return new RMSPropOptimizer(learningRate, decay, momentum, epsilon32, centered);
+  }
+  static adam(learningRate = 1e-3, beta1 = 0.9, beta2 = 0.999, epsilon32 = null) {
+    return new AdamOptimizer(learningRate, beta1, beta2, epsilon32);
+  }
+  static adadelta(learningRate = 1e-3, rho = 0.95, epsilon32 = null) {
+    return new AdadeltaOptimizer(learningRate, rho, epsilon32);
+  }
+  static adamax(learningRate = 2e-3, beta1 = 0.9, beta2 = 0.999, epsilon32 = null, decay = 0) {
+    return new AdamaxOptimizer(learningRate, beta1, beta2, epsilon32, decay);
+  }
+  static adagrad(learningRate, initialAccumulatorValue = 0.1) {
+    return new AdagradOptimizer(learningRate, initialAccumulatorValue);
+  }
+};
+var train = {
+  sgd: OptimizerConstructors.sgd,
+  momentum: OptimizerConstructors.momentum,
+  adadelta: OptimizerConstructors.adadelta,
+  adagrad: OptimizerConstructors.adagrad,
+  rmsprop: OptimizerConstructors.rmsprop,
+  adamax: OptimizerConstructors.adamax,
+  adam: OptimizerConstructors.adam
+};
+var delayCallback = (() => {
+  if (typeof requestAnimationFrame !== "undefined") {
+    return requestAnimationFrame;
+  } else if (typeof setImmediate !== "undefined") {
+    return setImmediate;
+  }
+  return (f) => f();
+})();
+function nextFrame() {
+  return new Promise((resolve) => delayCallback(() => resolve()));
+}
+var backend_util_exports = {};
+__export2(backend_util_exports, {
+  ERF_A1: () => ERF_A1,
+  ERF_A2: () => ERF_A2,
+  ERF_A3: () => ERF_A3,
+  ERF_A4: () => ERF_A4,
+  ERF_A5: () => ERF_A5,
+  ERF_P: () => ERF_P,
+  PARALLELIZE_THRESHOLD: () => PARALLELIZE_THRESHOLD,
+  RowPartitionType: () => RowPartitionType,
+  SELU_SCALE: () => SELU_SCALE,
+  SELU_SCALEALPHA: () => SELU_SCALEALPHA,
+  applyActivation: () => applyActivation,
+  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,
+  assertAxesAreInnerMostDims: () => assertAxesAreInnerMostDims,
+  assertParamsConsistent: () => assertParamsConsistent,
+  assignToTypedArray: () => assignToTypedArray,
+  axesAreInnerMostDims: () => axesAreInnerMostDims,
+  calculateShapes: () => calculateShapes,
+  checkEinsumDimSizes: () => checkEinsumDimSizes,
+  checkPadOnDimRoundingMode: () => checkPadOnDimRoundingMode,
+  combineLocations: () => combineLocations,
+  combineRaggedTensorToTensorShapes: () => combineRaggedTensorToTensorShapes,
+  complexWithEvenIndex: () => complexWithEvenIndex,
+  complexWithOddIndex: () => complexWithOddIndex,
+  computeConv2DInfo: () => computeConv2DInfo,
+  computeConv3DInfo: () => computeConv3DInfo,
+  computeDefaultPad: () => computeDefaultPad,
+  computeDilation2DInfo: () => computeDilation2DInfo,
+  computeOptimalWindowSize: () => computeOptimalWindowSize,
+  computeOutAndReduceShapes: () => computeOutAndReduceShapes,
+  computeOutShape: () => computeOutShape2,
+  computePool2DInfo: () => computePool2DInfo,
+  computePool3DInfo: () => computePool3DInfo,
+  convertConv2DDataFormat: () => convertConv2DDataFormat,
+  decodeEinsumEquation: () => decodeEinsumEquation,
+  eitherStridesOrDilationsAreOne: () => eitherStridesOrDilationsAreOne,
+  expandShapeToKeepDim: () => expandShapeToKeepDim,
+  exponent: () => exponent,
+  exponents: () => exponents,
+  fromStringArrayToUint8: () => fromStringArrayToUint8,
+  fromUint8ToStringArray: () => fromUint8ToStringArray,
+  getAxesPermutation: () => getAxesPermutation,
+  getBroadcastDims: () => getBroadcastDims,
+  getComplexWithIndex: () => getComplexWithIndex,
+  getEinsumComputePath: () => getEinsumComputePath,
+  getEinsumPermutation: () => getEinsumPermutation,
+  getFusedBiasGradient: () => getFusedBiasGradient,
+  getFusedDyActivation: () => getFusedDyActivation,
+  getImageCenter: () => getImageCenter,
+  getInnerMostAxes: () => getInnerMostAxes,
+  getPermuted: () => getPermuted,
+  getRaggedRank: () => getRaggedRank,
+  getReductionAxes: () => getReductionAxes,
+  getReshaped: () => getReshaped,
+  getReshapedPermuted: () => getReshapedPermuted,
+  getRowPartitionTypesHelper: () => getRowPartitionTypesHelper,
+  getSliceBeginCoords: () => getSliceBeginCoords,
+  getSliceSize: () => getSliceSize,
+  getSparseFillEmptyRowsIndicesDenseShapeMismatch: () => getSparseFillEmptyRowsIndicesDenseShapeMismatch,
+  getSparseFillEmptyRowsNegativeIndexErrorMessage: () => getSparseFillEmptyRowsNegativeIndexErrorMessage,
+  getSparseFillEmptyRowsOutOfRangeIndexErrorMessage: () => getSparseFillEmptyRowsOutOfRangeIndexErrorMessage,
+  getSparseReshapeEmptyTensorZeroOutputDimErrorMessage: () => getSparseReshapeEmptyTensorZeroOutputDimErrorMessage,
+  getSparseReshapeInputOutputMismatchErrorMessage: () => getSparseReshapeInputOutputMismatchErrorMessage,
+  getSparseReshapeInputOutputMultipleErrorMessage: () => getSparseReshapeInputOutputMultipleErrorMessage,
+  getSparseReshapeMultipleNegativeOneOutputDimErrorMessage: () => getSparseReshapeMultipleNegativeOneOutputDimErrorMessage,
+  getSparseReshapeNegativeOutputDimErrorMessage: () => getSparseReshapeNegativeOutputDimErrorMessage,
+  getSparseSegmentReductionIndicesOutOfRangeErrorMessage: () => getSparseSegmentReductionIndicesOutOfRangeErrorMessage,
+  getSparseSegmentReductionNegativeSegmentIdsErrorMessage: () => getSparseSegmentReductionNegativeSegmentIdsErrorMessage,
+  getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage: () => getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage,
+  getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage: () => getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage,
+  getUndoAxesPermutation: () => getUndoAxesPermutation,
+  isIdentityPermutation: () => isIdentityPermutation,
+  log: () => log,
+  mergeRealAndImagArrays: () => mergeRealAndImagArrays,
+  prepareAndValidate: () => prepareAndValidate,
+  prepareSplitSize: () => prepareSplitSize,
+  segment_util: () => segment_util_exports,
+  shouldFuse: () => shouldFuse,
+  slice_util: () => slice_util_exports,
+  splitRealAndImagArrays: () => splitRealAndImagArrays,
+  tupleValuesAreOne: () => tupleValuesAreOne,
+  upcastType: () => upcastType,
+  validateDefaultValueShape: () => validateDefaultValueShape,
+  validateInput: () => validateInput,
+  validateUpdateShape: () => validateUpdateShape,
+  warn: () => warn
+});
+function assertParamsConsistent(shapes, axis) {
+  const rank = shapes[0].length;
+  shapes.forEach((shape, i) => {
+    assert(shape.length === rank, () => `Error in concat${rank}D: rank of tensors[${i}] must be the same as the rank of the rest (${rank})`);
+  });
+  assert(axis >= 0 && axis < rank, () => `Error in concat${rank}D: axis must be between 0 and ${rank - 1}.`);
+  const firstShape = shapes[0];
+  shapes.forEach((shape, i) => {
+    for (let r = 0; r < rank; r++) {
+      assert(r === axis || shape[r] === firstShape[r], () => `Error in concat${rank}D: Shape of tensors[${i}] (${shape}) does not match the shape of the rest (${firstShape}) along the non-concatenated axis ${i}.`);
+    }
+  });
+}
+function computeOutShape2(shapes, axis) {
+  const outputShape = shapes[0].slice();
+  for (let i = 1; i < shapes.length; i++) {
+    outputShape[axis] += shapes[i][axis];
+  }
+  return outputShape;
+}
+var RowPartitionType;
+(function(RowPartitionType3) {
+  RowPartitionType3[RowPartitionType3["FIRST_DIM_SIZE"] = 0] = "FIRST_DIM_SIZE";
+  RowPartitionType3[RowPartitionType3["VALUE_ROWIDS"] = 1] = "VALUE_ROWIDS";
+  RowPartitionType3[RowPartitionType3["ROW_LENGTHS"] = 2] = "ROW_LENGTHS";
+  RowPartitionType3[RowPartitionType3["ROW_SPLITS"] = 3] = "ROW_SPLITS";
+  RowPartitionType3[RowPartitionType3["ROW_LIMITS"] = 4] = "ROW_LIMITS";
+  RowPartitionType3[RowPartitionType3["ROW_STARTS"] = 5] = "ROW_STARTS";
+})(RowPartitionType || (RowPartitionType = {}));
+function combineRaggedTensorToTensorShapes(raggedRank, shape, valueShape) {
+  let outputShape = new Array();
+  if (valueShape == null && shape == null) {
+    return outputShape;
+  }
+  if (shape == null) {
+    while (outputShape.length < raggedRank + valueShape.length) {
+      outputShape.push(-1);
+    }
+  } else {
+    outputShape = shape.slice();
+  }
+  if (valueShape == null) {
+    return outputShape;
+  }
+  if (raggedRank + valueShape.length !== outputShape.length) {
+    throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.rank = ${raggedRank + valueShape.length}, but shape.rank = ${outputShape.length}`);
+  }
+  for (let i = 1; i < valueShape.length; ++i) {
+    const valueDim = valueShape[i];
+    const outputShapeDimIndex = outputShape[outputShape.length - valueShape.length + i];
+    const outputShapeDim = outputShape[outputShapeDimIndex];
+    if (valueDim >= 0) {
+      if (outputShapeDim >= 0) {
+        if (outputShapeDim !== valueDim) {
+          throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.shape[${i + raggedRank}] = ${valueDim} but shape[${i + raggedRank}] = ${outputShapeDim}`);
+        }
+      } else {
+        outputShape[outputShapeDimIndex] = valueDim;
+      }
+    }
+  }
+  return outputShape;
+}
+function getRowPartitionTypesHelper(rowPartitionTypeStrings) {
+  const stringToType = {
+    "FIRST_DIM_SIZE": RowPartitionType.FIRST_DIM_SIZE,
+    "VALUE_ROWIDS": RowPartitionType.VALUE_ROWIDS,
+    "ROW_LENGTHS": RowPartitionType.ROW_LENGTHS,
+    "ROW_SPLITS": RowPartitionType.ROW_SPLITS,
+    "ROW_LIMITS": RowPartitionType.ROW_LIMITS,
+    "ROW_STARTS": RowPartitionType.ROW_STARTS
+  };
+  const result = [];
+  for (const typeStr of rowPartitionTypeStrings) {
+    if (typeStr in stringToType) {
+      result.push(stringToType[typeStr]);
+    } else {
+      break;
+    }
+  }
+  return result;
+}
+function getRaggedRank(rowPartitionTypes) {
+  if (rowPartitionTypes.length === 0) {
+    return 0;
+  }
+  if (rowPartitionTypes[0] === RowPartitionType.FIRST_DIM_SIZE) {
+    return rowPartitionTypes.length - 1;
+  }
+  return rowPartitionTypes.length;
+}
+function validateDefaultValueShape(defaultValueShape, valueShape) {
+  if (defaultValueShape == null || valueShape == null) {
+    return;
+  }
+  const defaultNDims = defaultValueShape.length;
+  const valuesNDims = valueShape.length;
+  if (defaultNDims >= valuesNDims) {
+    throw new Error(`defaultValue.shape=${defaultValueShape} and ragged tensor flatValues.shape=${valueShape}, are incompatible: defaultValue.rank = ${defaultNDims} must be less than ragged tensor input flatValues.rank = ${valuesNDims})`);
+  }
+  for (let i = 0; i < Math.min(defaultNDims, valuesNDims - 1); ++i) {
+    const defaultDim = defaultValueShape[i];
+    const valueDim = valueShape[i + 1];
+    if (defaultDim >= 0 && valueDim >= 0 && defaultDim !== 1 && defaultDim !== valueDim) {
+      throw new Error(`defaultValue.shape=${defaultValueShape}, and ragged tensor input flatValues.shape=${valueShape} are incompatible: defaultValue.shape[${i - defaultValueShape.length}] = ${defaultDim} but ragged tensor input.flatValues.shape[${i - defaultValueShape.length}] = ${valueDim}`);
+    }
+  }
+}
+var PARALLELIZE_THRESHOLD = 30;
+function computeOptimalWindowSize(inSize) {
+  if (inSize <= PARALLELIZE_THRESHOLD) {
+    return inSize;
+  }
+  return nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));
+}
+function getImageCenter(center, imageHeight, imageWidth) {
+  const centerX = imageWidth * (typeof center === "number" ? center : center[0]);
+  const centerY = imageHeight * (typeof center === "number" ? center : center[1]);
+  return [centerX, centerY];
+}
+function getReshaped(inputShape, blockShape, prod5, batchToSpace = true) {
+  let reshaped = [];
+  if (batchToSpace) {
+    reshaped = reshaped.concat(blockShape.slice(0));
+    reshaped.push(inputShape[0] / prod5);
+    reshaped = reshaped.concat(inputShape.slice(1));
+  } else {
+    reshaped = reshaped.concat(inputShape[0]);
+    const spatialLength = blockShape.length;
+    for (let i = 0; i < spatialLength; ++i) {
+      reshaped = reshaped.concat([inputShape[i + 1] / blockShape[i], blockShape[i]]);
+    }
+    reshaped = reshaped.concat(inputShape.slice(spatialLength + 1));
+  }
+  return reshaped;
+}
+function getPermuted(reshapedRank, blockShapeRank, batchToSpace = true) {
+  const permuted = [];
+  if (batchToSpace) {
+    permuted.push(blockShapeRank);
+    for (let i = blockShapeRank + 1; i < reshapedRank; ++i) {
+      if (i <= 2 * blockShapeRank) {
+        permuted.push(i);
+        permuted.push(i - (blockShapeRank + 1));
+      } else {
+        permuted.push(i);
+      }
+    }
+  } else {
+    const permutedBeforeBatch = [];
+    const permutedAfterBatch = [];
+    for (let i = 1; i < reshapedRank; ++i) {
+      if (i >= blockShapeRank * 2 + 1 || i % 2 === 1) {
+        permutedAfterBatch.push(i);
+      } else {
+        permutedBeforeBatch.push(i);
+      }
+    }
+    permuted.push(...permutedBeforeBatch);
+    permuted.push(0);
+    permuted.push(...permutedAfterBatch);
+  }
+  return permuted;
+}
+function getReshapedPermuted(inputShape, blockShape, prod5, batchToSpace = true) {
+  const reshapedPermuted = [];
+  if (batchToSpace) {
+    reshapedPermuted.push(inputShape[0] / prod5);
+  } else {
+    reshapedPermuted.push(inputShape[0] * prod5);
+  }
+  for (let i = 1; i < inputShape.length; ++i) {
+    if (i <= blockShape.length) {
+      if (batchToSpace) {
+        reshapedPermuted.push(blockShape[i - 1] * inputShape[i]);
+      } else {
+        reshapedPermuted.push(inputShape[i] / blockShape[i - 1]);
+      }
+    } else {
+      reshapedPermuted.push(inputShape[i]);
+    }
+  }
+  return reshapedPermuted;
+}
+function getSliceBeginCoords(crops, blockShape) {
+  const sliceBeginCoords = [0];
+  for (let i = 0; i < blockShape; ++i) {
+    sliceBeginCoords.push(crops[i][0]);
+  }
+  return sliceBeginCoords;
+}
+function getSliceSize(uncroppedShape, crops, blockShape) {
+  const sliceSize = uncroppedShape.slice(0, 1);
+  for (let i = 0; i < blockShape; ++i) {
+    sliceSize.push(uncroppedShape[i + 1] - crops[i][0] - crops[i][1]);
+  }
+  return sliceSize;
+}
+var SELU_SCALEALPHA = 1.7580993408473768;
+var SELU_SCALE = 1.0507009873554805;
+var ERF_P = 0.3275911;
+var ERF_A1 = 0.254829592;
+var ERF_A2 = -0.284496736;
+var ERF_A3 = 1.421413741;
+var ERF_A4 = -1.453152027;
+var ERF_A5 = 1.061405429;
+function mergeRealAndImagArrays(real4, imag4) {
+  if (real4.length !== imag4.length) {
+    throw new Error(`Cannot merge real and imag arrays of different lengths. real:${real4.length}, imag: ${imag4.length}.`);
+  }
+  const result = new Float32Array(real4.length * 2);
+  for (let i = 0; i < result.length; i += 2) {
+    result[i] = real4[i / 2];
+    result[i + 1] = imag4[i / 2];
+  }
+  return result;
+}
+function splitRealAndImagArrays(complex4) {
+  const real4 = new Float32Array(complex4.length / 2);
+  const imag4 = new Float32Array(complex4.length / 2);
+  for (let i = 0; i < complex4.length; i += 2) {
+    real4[i / 2] = complex4[i];
+    imag4[i / 2] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function complexWithEvenIndex(complex4) {
+  const len = Math.ceil(complex4.length / 4);
+  const real4 = new Float32Array(len);
+  const imag4 = new Float32Array(len);
+  for (let i = 0; i < complex4.length; i += 4) {
+    real4[Math.floor(i / 4)] = complex4[i];
+    imag4[Math.floor(i / 4)] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function complexWithOddIndex(complex4) {
+  const len = Math.floor(complex4.length / 4);
+  const real4 = new Float32Array(len);
+  const imag4 = new Float32Array(len);
+  for (let i = 2; i < complex4.length; i += 4) {
+    real4[Math.floor(i / 4)] = complex4[i];
+    imag4[Math.floor(i / 4)] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function getComplexWithIndex(complex4, index) {
+  const real4 = complex4[index * 2];
+  const imag4 = complex4[index * 2 + 1];
+  return { real: real4, imag: imag4 };
+}
+function assignToTypedArray(data, real4, imag4, index) {
+  data[index * 2] = real4;
+  data[index * 2 + 1] = imag4;
+}
+function exponents(n, inverse) {
+  const real4 = new Float32Array(n / 2);
+  const imag4 = new Float32Array(n / 2);
+  for (let i = 0; i < Math.ceil(n / 2); i++) {
+    const x = (inverse ? 2 : -2) * Math.PI * (i / n);
+    real4[i] = Math.cos(x);
+    imag4[i] = Math.sin(x);
+  }
+  return { real: real4, imag: imag4 };
+}
+function exponent(k, n, inverse) {
+  const x = (inverse ? 2 : -2) * Math.PI * (k / n);
+  const real4 = Math.cos(x);
+  const imag4 = Math.sin(x);
+  return { real: real4, imag: imag4 };
+}
+var ARROW = "->";
+var ARROW_REGEX = /->/g;
+var COMMA = ",";
+var ELLIPSIS = "...";
+function decodeEinsumEquation(equation, numTensors) {
+  equation = equation.replace(/\s/g, "");
+  const numArrows = (equation.length - equation.replace(ARROW_REGEX, "").length) / ARROW.length;
+  if (numArrows < 1) {
+    throw new Error("Equations without an arrow are not supported.");
+  } else if (numArrows > 1) {
+    throw new Error(`Equation must contain exactly one arrow ("${ARROW}").`);
+  }
+  const [inputString, outputString] = equation.split(ARROW);
+  assert(inputString.indexOf(ELLIPSIS) === -1, () => `The ellipsis notation ("${ELLIPSIS}") is not supported yet.`);
+  const inputTerms = inputString.split(COMMA);
+  const numInputs = inputTerms.length;
+  if (numTensors !== numInputs) {
+    throw new Error(`Expected ${numInputs} input tensors, received ${numTensors}`);
+  }
+  if (numInputs > 2) {
+    throw new Error("Support for more than 2 input tensors is not implemented yet.");
+  }
+  const allDims = [];
+  for (let i = 0; i < outputString.length; ++i) {
+    const dimName = outputString[i];
+    if (!inputTerms.some((inputTerm) => inputTerm.indexOf(dimName) !== -1)) {
+      throw new Error(`Output subscripts contain the label ${dimName} not present in the input subscripts.`);
+    }
+    if (allDims.indexOf(dimName) === -1) {
+      allDims.push(dimName);
+    }
+  }
+  for (let i = 0; i < inputString.length; ++i) {
+    const dimName = inputString[i];
+    if (allDims.indexOf(dimName) === -1 && dimName !== COMMA) {
+      allDims.push(dimName);
+    }
+  }
+  const idDims = new Array(inputTerms.length);
+  for (let i = 0; i < numInputs; ++i) {
+    if (new Set(inputTerms[i].split("")).size !== inputTerms[i].length) {
+      throw new Error(`Found duplicate axes in input component ${inputTerms[i]}. Support for duplicate axes in input is not implemented yet.`);
+    }
+    idDims[i] = [];
+    for (let j = 0; j < inputTerms[i].length; ++j) {
+      idDims[i].push(allDims.indexOf(inputTerms[i][j]));
+    }
+  }
+  const numDims = allDims.length;
+  const numOutDims = outputString.length;
+  const summedDims = [];
+  for (let i = numOutDims; i < numDims; ++i) {
+    summedDims.push(i);
+  }
+  return { allDims, summedDims, idDims };
+}
+function getEinsumPermutation(nDims, idDims) {
+  let permutationIndices = new Array(nDims);
+  permutationIndices.fill(-1);
+  for (let i = 0; i < idDims.length; ++i) {
+    permutationIndices[idDims[i]] = i;
+  }
+  const expandDims6 = [];
+  for (let i = 0; i < nDims; ++i) {
+    if (permutationIndices[i] === -1) {
+      expandDims6.push(i);
+    }
+  }
+  permutationIndices = permutationIndices.filter((d) => d !== -1);
+  return { permutationIndices, expandDims: expandDims6 };
+}
+function checkEinsumDimSizes(nDims, idDims, tensors) {
+  const dimSizes = new Array(nDims);
+  for (let i = 0; i < tensors.length; ++i) {
+    const shape = tensors[i].shape;
+    for (let j = 0; j < idDims[i].length; ++j) {
+      if (dimSizes[idDims[i][j]] === void 0) {
+        dimSizes[idDims[i][j]] = shape[j];
+      } else {
+        assert(dimSizes[idDims[i][j]] === shape[j], () => `Expected dimension ${dimSizes[idDims[i][j]]} at axis ${j} of input shaped ${JSON.stringify(shape)}, but got dimension ${shape[j]}`);
+      }
+    }
+  }
+}
+function getEinsumComputePath(summedDims, idDims) {
+  const path = summedDims;
+  const steps = [];
+  let nSteps = 0;
+  if (summedDims.length === 0) {
+    path.push(-1);
+  }
+  nSteps = summedDims.length + 1;
+  for (let i = 0; i < nSteps; ++i) {
+    steps.push([]);
+  }
+  const computedTermIndices = [];
+  for (let i = 0; i < path.length; ++i) {
+    const summedDim = path[i];
+    const termIndices = findTermsWithDim(idDims, summedDim);
+    for (const termIndex of termIndices) {
+      if (computedTermIndices.indexOf(termIndex) === -1) {
+        steps[i].push(termIndex);
+        computedTermIndices.push(termIndex);
+      }
+    }
+  }
+  return { path, steps };
+}
+function isIdentityPermutation(perm) {
+  return perm.every((dim, index) => dim === index);
+}
+function findTermsWithDim(idDims, dim) {
+  const termIndices = [];
+  for (let i = 0; i < idDims.length; ++i) {
+    if (idDims[i].length === 0 || idDims[i].indexOf(dim) !== -1 || dim === -1) {
+      termIndices.push(i);
+    }
+  }
+  return termIndices;
+}
+function prepareSplitSize(x, numOrSizeSplits, axis = 0) {
+  let splitSizes = [];
+  if (typeof numOrSizeSplits === "number") {
+    assert(x.shape[axis] % numOrSizeSplits === 0, () => "Number of splits must evenly divide the axis.");
+    splitSizes = new Array(numOrSizeSplits).fill(x.shape[axis] / numOrSizeSplits);
+  } else {
+    const numOfNegs = numOrSizeSplits.reduce((count2, value) => {
+      if (value === -1) {
+        count2 += 1;
+      }
+      return count2;
+    }, 0);
+    assert(numOfNegs <= 1, () => "There should be only one negative value in split array.");
+    const negIndex = numOrSizeSplits.indexOf(-1);
+    if (negIndex !== -1) {
+      const total = numOrSizeSplits.reduce((a, b) => b > 0 ? a + b : a);
+      numOrSizeSplits[negIndex] = x.shape[axis] - total;
+    }
+    assert(x.shape[axis] === numOrSizeSplits.reduce((a, b) => a + b), () => "The sum of sizes must match the size of the axis dimension.");
+    splitSizes = numOrSizeSplits;
+  }
+  return splitSizes;
+}
+function getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesLength) {
+  return `Received SparseTensor with denseShape[0] = 0 but
+  indices.shape[0] = ${indicesLength}`;
+}
+function getSparseFillEmptyRowsNegativeIndexErrorMessage(index, value) {
+  return `indices(${index}, 0) is invalid: ${value} < 0`;
+}
+function getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(index, value, limit) {
+  return `indices(${index}, 0) is invalid: ${value} >= ${limit}`;
+}
+function getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(dim1, dim2) {
+  return `only one output dimension may be -1, not both ${dim1} and ${dim2}`;
+}
+function getSparseReshapeNegativeOutputDimErrorMessage(dim, value) {
+  return `size ${dim} must be non-negative, not ${value}`;
+}
+function getSparseReshapeEmptyTensorZeroOutputDimErrorMessage() {
+  return "reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero";
+}
+function getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape) {
+  const inputSize = sizeFromShape(inputShape);
+  const outputSize = sizeFromShape(outputShape);
+  return `Input to reshape is a SparseTensor with ${inputSize}
+  dense values, but the requested shape requires a multiple of ${outputSize}. inputShape=${inputShape} outputShape= ${outputShape}`;
+}
+function getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape) {
+  const inputSize = sizeFromShape(inputShape);
+  const outputSize = sizeFromShape(outputShape);
+  return `Input to reshape is a tensor with ${inputSize} dense values, but the requested shape has ${outputSize}. inputShape=${inputShape} outputShape=${outputShape}`;
+}
+function getSparseSegmentReductionNegativeSegmentIdsErrorMessage() {
+  return `segment ids must be >= 0`;
+}
+function getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage() {
+  return `segment ids are not increasing`;
+}
+function getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(segmentId, outputRows) {
+  return `Segment id ${segmentId} out of range [0, ${outputRows}), possibly because segmentIds input is not sorted.`;
+}
+function getSparseSegmentReductionIndicesOutOfRangeErrorMessage(index, indexValue, inputRows) {
+  return `Bad: indices[${index}] == ${indexValue} out of range [0, ${inputRows})`;
+}
+var segment_util_exports = {};
+__export2(segment_util_exports, {
+  collectGatherOpShapeInfo: () => collectGatherOpShapeInfo,
+  computeOutShape: () => computeOutShape3,
+  segOpComputeOptimalWindowSize: () => segOpComputeOptimalWindowSize
+});
+function segOpComputeOptimalWindowSize(inSize, numSegments) {
+  let done = false;
+  let res;
+  if (inSize <= PARALLELIZE_THRESHOLD) {
+    res = inSize;
+    done = true;
+  } else {
+    res = nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));
+  }
+  while (!done) {
+    if (res > numSegments || res === inSize) {
+      done = true;
+    } else {
+      res = nearestDivisor(inSize, res + 1);
+    }
+  }
+  return res;
+}
+function computeOutShape3(aShape, axis, numSegments) {
+  const outShape = [];
+  const rank = aShape.length;
+  for (let dim = 0; dim < rank; dim++) {
+    if (dim !== axis) {
+      outShape.push(aShape[dim]);
+    } else {
+      outShape.push(numSegments);
+    }
+  }
+  return outShape;
+}
+function collectGatherOpShapeInfo(x, indices, axis, batchDims) {
+  const indicesRank = indices.shape.length;
+  const xRank = x.shape.length;
+  if (batchDims !== 0) {
+    if (batchDims < -indicesRank || batchDims > indicesRank) {
+      throw new Error(`Expect batchDims in the range of [-${indicesRank}, ${indicesRank}], but got ${batchDims}`);
+    }
+  }
+  if (batchDims < 0) {
+    batchDims += indicesRank;
+  }
+  if (batchDims > xRank) {
+    throw new Error(`batchDims (${batchDims}) must be less than rank(x) (
+    ${xRank}).`);
+  }
+  if (axis < batchDims) {
+    throw new Error(`batchDims (${batchDims}) must be less than or equal to axis (${axis}).`);
+  }
+  for (let i = 0; i < batchDims; ++i) {
+    if (x.shape[i] !== indices.shape[i]) {
+      throw new Error(`x.shape[${i}]: ${x.shape[i]} should be equal to indices.shape[${i}]: ${indices.shape[i]}.`);
+    }
+  }
+  const dimSize = x.shape[axis];
+  const outputShape = [];
+  let batchSize = 1;
+  let outerSize = 1;
+  let sliceSize = 1;
+  for (let i = 0; i < batchDims; ++i) {
+    outputShape.push(x.shape[i]);
+    batchSize *= x.shape[i];
+  }
+  for (let i = batchDims; i < axis; i++) {
+    outputShape.push(x.shape[i]);
+    outerSize *= x.shape[i];
+  }
+  for (let i = batchDims; i < indicesRank; i++) {
+    outputShape.push(indices.shape[i]);
+  }
+  for (let i = axis + 1; i < xRank; i++) {
+    outputShape.push(x.shape[i]);
+    sliceSize *= x.shape[i];
+  }
+  return { batchSize, sliceSize, outerSize, dimSize, outputShape };
+}
+function fromUint8ToStringArray(vals) {
+  try {
+    return vals.map((val) => decodeString(val));
+  } catch (err) {
+    throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${err}`);
+  }
+}
+function fromStringArrayToUint8(strings) {
+  return strings.map((s) => encodeString(s));
+}
+var kernel_impls_exports = {};
+__export2(kernel_impls_exports, {
+  nonMaxSuppressionV3Impl: () => nonMaxSuppressionV3Impl,
+  nonMaxSuppressionV4Impl: () => nonMaxSuppressionV4Impl,
+  nonMaxSuppressionV5Impl: () => nonMaxSuppressionV5Impl,
+  whereImpl: () => whereImpl
+});
+var absGradConfig = {
+  kernelName: Abs,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, step(cast(x, "float32"), -1)) };
+  }
+};
+var acosGradConfig = {
+  kernelName: Acos,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = square(cast(x, "float32"));
+        const b = sqrt(sub(scalar(1), a));
+        return neg(div(dy, b));
+      }
+    };
+  }
+};
+var acoshGradConfig = {
+  kernelName: Acosh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = sqrt(sub(square(cast(x, "float32")), 1));
+        return div(dy, a);
+      }
+    };
+  }
+};
+var addGradConfig = {
+  kernelName: Add,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+var addNGradConfig = {
+  kernelName: AddN,
+  saveAllInputs: true,
+  gradFunc: (dy, saved) => {
+    const ders = {};
+    saved.forEach((_, i) => {
+      ders[i] = () => dy.clone();
+    });
+    return ders;
+  }
+};
+var argMaxGradConfig = {
+  kernelName: ArgMax,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => zerosLike(x) };
+  }
+};
+var argMinGradConfig = {
+  kernelName: ArgMin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => zerosLike(x) };
+  }
+};
+var asinGradConfig = {
+  kernelName: Asin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, sqrt(sub(scalar(1), square(cast(x, "float32"))))) };
+  }
+};
+var asinhGradConfig = {
+  kernelName: Asinh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = sqrt(add2(scalar(1), square(cast(x, "float32"))));
+        return div(dy, a);
+      }
+    };
+  }
+};
+var atan2GradConfig = {
+  kernelName: Atan2,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const d = add2(square(a), square(b));
+      let res = mul(dy, div(b, d));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      const d = add2(square(a), square(b));
+      let res = neg(mul(dy, div(a, d)));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+var atanGradConfig = {
+  kernelName: Atan,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, add2(square(cast(x, "float32")), 1)) };
+  }
+};
+var atanhGradConfig = {
+  kernelName: Atanh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, sub(scalar(1), square(cast(x, "float32")))) };
+  }
+};
+function avgPool3dGrad_(dy, input2, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "avgPool3dGrad");
+  const $input = convertToTensor(input2, "input", "avgPool3dGrad");
+  let dy5D = $dy;
+  let input5D = $input;
+  let reshapedTo5D = false;
+  if ($input.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);
+    input5D = reshape($input, [
+      1,
+      $input.shape[0],
+      $input.shape[1],
+      $input.shape[2],
+      $input.shape[3]
+    ]);
+  }
+  assert(dy5D.rank === 5, () => `Error in avgPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);
+  assert(input5D.rank === 5, () => `Error in avgPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);
+  checkPadOnDimRoundingMode("avgPool3dGrad", pad3, dimRoundingMode);
+  const inputs = { dy: dy5D, input: input5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(AvgPool3DGrad, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var avgPool3dGrad = op({ avgPool3dGrad_ });
+var avgPool3DGradConfig = {
+  kernelName: AvgPool3D,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+    return {
+      x: () => avgPool3dGrad(dy, x, filterSize, strides, pad3, dimRoundingMode)
+    };
+  }
+};
+function avgPoolGrad_(dy, input2, filterSize, strides, pad3) {
+  const $dy = convertToTensor(dy, "dy", "avgPoolGrad");
+  const $input = convertToTensor(input2, "input", "avgPoolGrad");
+  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);
+  let input4D = $input;
+  let dy4D = $dy;
+  let reshapedTo4D = false;
+  if ($input.rank === 3) {
+    reshapedTo4D = true;
+    input4D = reshape($input, [1, $input.shape[0], $input.shape[1], $input.shape[2]]);
+    dy4D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2]]);
+  }
+  assert(dy4D.rank === 4, () => `Error in avgPoolGrad: dy must be rank 4 but got rank ${dy4D.rank}.`);
+  assert(input4D.rank === 4, () => `Error in avgPoolGrad: input must be rank 4 but got rank ${input4D.rank}.`);
+  const inputs = { dy: dy4D, input: input4D };
+  const attrs = { filterSize, strides, pad: pad3 };
+  const res = ENGINE.runKernel(AvgPoolGrad, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var avgPoolGrad = op({ avgPoolGrad_ });
+var avgPoolGradConfig = {
+  kernelName: AvgPool,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { filterSize, strides, pad: pad3 } = attrs;
+    return { x: () => avgPoolGrad(dy, x, filterSize, strides, pad3) };
+  }
+};
+var batchMatMulGradConfig = {
+  kernelName: BatchMatMul,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved, attrs) => {
+    const [a, b] = saved;
+    const { transposeA, transposeB } = attrs;
+    if (!transposeA && !transposeB) {
+      return {
+        a: () => matMul(dy, b, false, true),
+        b: () => matMul(a, dy, true, false)
+      };
+    } else if (!transposeA && transposeB) {
+      return {
+        a: () => matMul(dy, b, false, false),
+        b: () => matMul(dy, a, true, false)
+      };
+    } else if (transposeA && !transposeB) {
+      return {
+        a: () => matMul(b, dy, false, true),
+        b: () => matMul(a, dy, false, false)
+      };
+    } else {
+      return {
+        a: () => matMul(b, dy, true, true),
+        b: () => matMul(dy, a, true, true)
+      };
+    }
+  }
+};
+var batchToSpaceNDGradConfig = {
+  kernelName: BatchToSpaceND,
+  gradFunc: (dy, saved, attrs) => {
+    const { blockShape, crops } = attrs;
+    return { x: () => spaceToBatchND(dy, blockShape, crops) };
+  }
+};
+var broadcastToGradConfig = {
+  kernelName: BroadcastTo,
+  gradFunc: (dy, saved, attrs) => {
+    const broadCastToAttrs = attrs;
+    const inputShape = broadCastToAttrs.inputShape;
+    const outputShape = broadCastToAttrs.shape;
+    const reps = Array.from(outputShape);
+    for (let i = inputShape.length - 1; i >= 0; i--) {
+      if (inputShape[i] === outputShape[i]) {
+        reps[i] = 1;
+      } else if (inputShape[i] !== 1) {
+        throw new Error(`broadcastTo(): [${inputShape}] cannot be broadcast to [${outputShape}].`);
+      }
+    }
+    const axes = [];
+    for (let i = 0; i < reps.length; i++) {
+      if (reps[i] > 1) {
+        axes.push(i);
+      }
+    }
+    return { x: () => sum2(dy, axes, true) };
+  }
+};
+var castGradConfig = {
+  kernelName: Cast,
+  gradFunc: (dy) => {
+    return { x: () => dy.clone() };
+  }
+};
+var ceilGradConfig = {
+  kernelName: Ceil,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var clipByValueGradConfig = {
+  kernelName: ClipByValue,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { clipValueMin, clipValueMax } = attrs;
+    return {
+      x: () => where(logicalAnd(greaterEqual(x, clipValueMin), lessEqual(x, clipValueMax)), dy, zerosLike(dy))
+    };
+  }
+};
+var complexAbsGradConfig = {
+  kernelName: ComplexAbs,
+  inputsToSave: ["x"],
+  gradFunc: absGradConfig.gradFunc
+};
+var concatGradConfig = {
+  kernelName: Concat,
+  saveAllInputs: true,
+  gradFunc: (dy, saved, attrs) => {
+    const shapes = saved.map((t) => t.shape);
+    const { axis } = attrs;
+    const $axis = parseAxisParam(axis, saved[0].shape)[0];
+    const sizeSplits = shapes.map((s) => s[$axis]);
+    const derTensors = split(dy, sizeSplits, $axis);
+    return derTensors.map((t) => () => t);
+  }
+};
+var conv2DGradConfig = {
+  kernelName: Conv2D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x4D, $filter] = saved;
+    const { dilations, strides, pad: pad3, dataFormat } = attrs;
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    return {
+      x: () => conv2DBackpropInput(x4D.shape, dy, $filter, strides, pad3, dataFormat),
+      filter: () => conv2DBackpropFilter(x4D, dy, $filter.shape, strides, pad3, dataFormat)
+    };
+  }
+};
+var conv2DBackpropInputGradConfig = {
+  kernelName: Conv2DBackpropInput,
+  inputsToSave: ["dy", "filter"],
+  gradFunc: (ddx, saved, attrs) => {
+    const [dy, filter] = saved;
+    const { strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+    return {
+      dy: () => conv2d(ddx, filter, strides, pad3, dataFormat, 1, dimRoundingMode),
+      filter: () => conv2DBackpropFilter(ddx, dy, filter.shape, strides, pad3, dataFormat, dimRoundingMode)
+    };
+  }
+};
+function conv3DBackpropFilter_(x, dy, filterShape, strides, pad3) {
+  let x5D = x;
+  if (x.rank === 4) {
+    x5D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2], x.shape[3]]);
+  }
+  let dy5D = dy;
+  if (dy5D.rank === 4) {
+    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in conv3dDerFilter: input must be rank 5, but got shape ${x5D.shape}.`);
+  assert(dy5D.rank === 5, () => `Error in conv3dDerFilter: dy must be rank 5, but got shape ${dy5D.shape}.`);
+  assert(filterShape.length === 5, () => `Error in conv3dDerFilter: filterShape must be length 5, but got ${filterShape}.`);
+  assert(x5D.shape[4] === filterShape[3], () => `Error in conv3dDerFilter: depth of input ${x5D.shape[4]}) must match input depth in filter (${filterShape[3]}.`);
+  assert(dy5D.shape[4] === filterShape[4], () => `Error in conv3dDerFilter: depth of dy (${dy5D.shape[4]}) must match output depth for filter (${filterShape[4]}).`);
+  const inputs = { x: x5D, dy: dy5D };
+  const attrs = { strides, pad: pad3, filterShape };
+  return ENGINE.runKernel(Conv3DBackpropFilterV2, inputs, attrs);
+}
+var conv3DBackpropFilter = op({ conv3DBackpropFilter_ });
+var conv3DGradConfig = {
+  kernelName: Conv3D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const { dilations, strides, pad: pad3 } = attrs;
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    const [x5D, $filter] = saved;
+    return {
+      x: () => conv3DBackpropInput(x5D.shape, dy, $filter, strides, pad3),
+      filter: () => conv3DBackpropFilter(x5D, dy, $filter.shape, strides, pad3)
+    };
+  }
+};
+var cosGradConfig = {
+  kernelName: Cos,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(neg(sin(cast(x, "float32"))), dy) };
+  }
+};
+var coshGradConfig = {
+  kernelName: Cosh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(sinh(cast(x, "float32")), dy) };
+  }
+};
+var cumsumGradConfig = {
+  kernelName: Cumsum,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis, exclusive, reverse: reverse5 } = attrs;
+    return {
+      x: () => {
+        const permutation = getAxesPermutation([axis], x.rank);
+        let out = cumsum(dy, axis, exclusive, !reverse5);
+        if (permutation != null) {
+          out = transpose(out, permutation);
+        }
+        return out;
+      }
+    };
+  }
+};
+var depthwiseConv2dNativeGradConfig = {
+  kernelName: DepthwiseConv2dNative,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const { dilations, strides, pad: pad3, dimRoundingMode } = attrs;
+    const $dilations = dilations == null ? [1, 1] : dilations;
+    assert(tupleValuesAreOne($dilations), () => `Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${$dilations}'`);
+    const [x, filter] = saved;
+    assert(x.rank === 4, () => `Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${x.rank}.`);
+    assert(filter.rank === 4, () => `Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${filter.rank}.`);
+    assert(x.shape[3] === filter.shape[2], () => `Error in gradient of depthwiseConv2d: number of input channels (${x.shape[3]}) must match the inChannels dimension in filter ${filter.shape[2]}.`);
+    assert(eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${strides} and dilations '${$dilations}'.`);
+    checkPadOnDimRoundingMode("depthwiseConv2d", pad3, dimRoundingMode);
+    return {
+      x: () => depthwiseConv2dNativeBackpropInput(x.shape, dy, filter, strides, pad3, $dilations, dimRoundingMode),
+      filter: () => depthwiseConv2dNativeBackpropFilter(x, dy, filter.shape, strides, pad3, $dilations, dimRoundingMode)
+    };
+  }
+};
+var dilation2dGradConfig = {
+  kernelName: Dilation2D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, filter] = saved;
+    const inputInputs = { x, filter, dy };
+    const filterInputs = { x, filter, dy };
+    return {
+      x: () => ENGINE.runKernel(Dilation2DBackpropInput, inputInputs, attrs),
+      filter: () => ENGINE.runKernel(Dilation2DBackpropFilter, filterInputs, attrs)
+    };
+  }
+};
+var eluGradConfig = {
+  kernelName: Elu,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    const inputs = { dy, y };
+    return { x: () => ENGINE.runKernel(EluGrad, inputs) };
+  }
+};
+var erfGradConfig = {
+  kernelName: Erf,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    const a = mul(exp(neg(square(x))), 2 / Math.sqrt(Math.PI));
+    return { x: () => mul(dy, a) };
+  }
+};
+var expGradConfig = {
+  kernelName: Exp,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(dy, y) };
+  }
+};
+var expandDimsGradConfig = {
+  kernelName: ExpandDims,
+  inputsToSave: ["input"],
+  gradFunc: (dy, saved) => {
+    const [input2] = saved;
+    return { input: () => reshape(dy, input2.shape) };
+  }
+};
+var expm1GradConfig = {
+  kernelName: Expm1,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, exp(x)) };
+  }
+};
+var floorGradConfig = {
+  kernelName: Floor,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var floorDivGradConfig = {
+  kernelName: FloorDiv,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = div(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      let res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = reshape(sum2(res, reduceAxes), b.shape);
+      }
+      const tmp = square(b);
+      return neg(div(res, cast(tmp, "float32")));
+    };
+    return { a: derA, b: derB };
+  }
+};
+var fusedBatchNormGradConfig = {
+  kernelName: FusedBatchNorm,
+  inputsToSave: ["x", "mean", "variance", "scale"],
+  gradFunc: (dy, saved, attrs) => {
+    const { varianceEpsilon } = attrs;
+    const [x, mean4, variance, scale22] = saved;
+    const scaleValue = scale22 == null ? scalar(1) : scale22;
+    const reductionAxes = getReductionAxes(mean4.shape, x.shape);
+    const tileShape = [];
+    if (mean4.rank === 1) {
+      for (let i = 0; i < x.shape.length - 1; ++i) {
+        tileShape.push(x.shape[i]);
+      }
+      tileShape.push(1);
+    }
+    const xMinusMean = sub(x, mean4);
+    const dyTimesScaleValue = mul(dy, scaleValue);
+    const oneOverSqrtVariance = rsqrt(add2(variance, scalar(varianceEpsilon)));
+    const minusHalfRCube = mul(mul(mul(oneOverSqrtVariance, oneOverSqrtVariance), oneOverSqrtVariance), scalar(-0.5));
+    const derX = () => {
+      if (mean4.rank === 1) {
+        return reshape(mul(mul(dy, tile(reshape(oneOverSqrtVariance, [1, 1, 1, mean4.shape[0]]), tileShape)), scaleValue), x.shape);
+      } else {
+        return reshape(mul(mul(dy, oneOverSqrtVariance), scaleValue), x.shape);
+      }
+    };
+    const derMean = () => {
+      let meanDer = mul(mul(oneOverSqrtVariance, scalar(-1)), dyTimesScaleValue);
+      if (mean4.rank === 1) {
+        meanDer = sum2(meanDer, reductionAxes);
+      }
+      return reshape(meanDer, mean4.shape);
+    };
+    const derVariance = () => {
+      let varianceDer = mul(mul(minusHalfRCube, xMinusMean), dyTimesScaleValue);
+      if (mean4.rank === 1) {
+        varianceDer = sum2(varianceDer, reductionAxes);
+      }
+      return reshape(varianceDer, mean4.shape);
+    };
+    const derScale = () => {
+      const xMinusMean2TimesRsqrt = mul(xMinusMean, oneOverSqrtVariance);
+      let scaleDer = mul(dy, xMinusMean2TimesRsqrt);
+      if (mean4.rank === 1) {
+        scaleDer = sum2(scaleDer, reductionAxes);
+      }
+      return reshape(scaleDer, mean4.shape);
+    };
+    const derOffset = () => {
+      let offsetDer = dy;
+      if (mean4.rank === 1) {
+        offsetDer = sum2(offsetDer, reductionAxes);
+      }
+      return reshape(offsetDer, mean4.shape);
+    };
+    return {
+      x: derX,
+      mean: derMean,
+      variance: derVariance,
+      scale: derScale,
+      offset: derOffset
+    };
+  }
+};
+var gatherGradConfig = {
+  kernelName: GatherV2,
+  inputsToSave: ["x", "indices"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, indices] = saved;
+    const { axis } = attrs;
+    const parsedAxis = parseAxisParam(axis, x.shape)[0];
+    const derX = () => {
+      const paramsShape = x.shape;
+      const indicesSize = indices.size;
+      const outerShape = paramsShape.slice(0, parsedAxis);
+      const outerDims = outerShape.length;
+      const innerShape = paramsShape.slice(axis, paramsShape.length).slice(1);
+      const innerDims = innerShape.length;
+      const outerAxesIndices = arrayRange(0, outerDims);
+      const innerAxesIndices = arrayRange(outerDims + 1, outerDims + 1 + innerDims);
+      const valuesShape = arrayConcat([outerShape, [indicesSize], innerShape]);
+      const values = reshape(dy, valuesShape);
+      const reshapedIndices = reshape(indices, [indicesSize]);
+      const transposeDims = arrayConcat([[outerDims], outerAxesIndices, innerAxesIndices]);
+      const valuesTranspose = transpose(values, transposeDims);
+      let paramsGrad = unsortedSegmentSum(valuesTranspose, reshapedIndices, x.shape[parsedAxis]);
+      const invertTransposeDims = getUndoAxesPermutation(transposeDims);
+      paramsGrad = transpose(paramsGrad, invertTransposeDims);
+      return paramsGrad;
+    };
+    return { x: derX, indices: () => indices };
+  }
+};
+function arrayRange(start, stop) {
+  const result = [];
+  for (let i = start; i < stop; ++i) {
+    result.push(i);
+  }
+  return result;
+}
+function arrayConcat(arrays) {
+  const result = [];
+  for (let i = 0; i < arrays.length; ++i) {
+    for (let j = 0; j < arrays[i].length; ++j) {
+      result.push(arrays[i][j]);
+    }
+  }
+  return result;
+}
+var greaterEqualGradConfig = {
+  kernelName: GreaterEqual,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    return { a: () => zerosLike(a), b: () => zerosLike(b) };
+  }
+};
+var identityGradConfig = {
+  kernelName: Identity,
+  gradFunc: (dy) => {
+    return { x: () => cast(dy, "float32") };
+  }
+};
+var isFiniteGradConfig = {
+  kernelName: IsFinite,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var isInfGradConfig = {
+  kernelName: IsInf,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var isNanGradConfig = {
+  kernelName: IsNan,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var leakyReluGradConfig = {
+  kernelName: LeakyRelu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { alpha } = attrs;
+    const mask = greater(x, 0);
+    return { x: () => where(mask, dy, mul(dy, alpha)) };
+  }
+};
+var log1pGradConfig = {
+  kernelName: Log1p,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, add2(x, 1)) };
+  }
+};
+var logGradConfig = {
+  kernelName: Log,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, cast(x, "float32")) };
+  }
+};
+var logSoftmaxGradConfig = {
+  kernelName: LogSoftmax,
+  inputsToSave: [],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [value] = saved;
+    const { axis } = attrs;
+    return {
+      logits: () => {
+        const keepDims = true;
+        const softmax6 = exp(value);
+        return sub(dy, mul(sum2(dy, axis, keepDims), softmax6));
+      }
+    };
+  }
+};
+function localResponseNormalizationBackprop_(x, y, dy, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {
+  const inputs = { x, y, dy };
+  const attrs = { depthRadius, bias, alpha, beta };
+  return ENGINE.runKernel(LRNGrad, inputs, attrs);
+}
+var localResponseNormalizationBackprop = op({ localResponseNormalizationBackprop_ });
+var lrnGradConfig = {
+  kernelName: LRN,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { depthRadius, bias, alpha, beta } = attrs;
+    return {
+      x: () => localResponseNormalizationBackprop(x, y, dy, depthRadius, bias, alpha, beta)
+    };
+  }
+};
+function gradForMinAndMax(dy, y, xOrig, origAxes) {
+  if (y.rank < xOrig.rank) {
+    y = reshape(y, expandShapeToKeepDim(y.shape, origAxes));
+  }
+  if (dy.rank < xOrig.rank) {
+    dy = reshape(dy, expandShapeToKeepDim(dy.shape, origAxes));
+  }
+  return {
+    x: () => {
+      const dx = mul(dy, cast(equal(xOrig, y), dy.dtype));
+      return dx;
+    }
+  };
+}
+var maxGradConfig = {
+  kernelName: Max,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const maxAttrs = attrs;
+    const { reductionIndices } = maxAttrs;
+    const x = saved[0];
+    const y = saved[1];
+    const origAxes = parseAxisParam(reductionIndices, x.shape);
+    const maxGrad = gradForMinAndMax(dy, y, x, origAxes);
+    return {
+      x: () => {
+        return maxGrad["x"]();
+      }
+    };
+  }
+};
+var maximumGradConfig = {
+  kernelName: Maximum,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const derA = () => mul(dy, cast(greaterEqual(a, b), "float32"));
+    const derB = () => mul(dy, cast(less(a, b), "float32"));
+    return { a: derA, b: derB };
+  }
+};
+function maxPool3dGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "maxPool3dGrad");
+  const $input = convertToTensor(input2, "input", "maxPool3dGrad");
+  const $output = convertToTensor(output, "output", "maxPool3dGrad");
+  let dy5D = $dy;
+  let input5D = $input;
+  let output5D = $output;
+  let reshapedTo5D = false;
+  if ($input.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);
+    input5D = reshape($input, [
+      1,
+      $input.shape[0],
+      $input.shape[1],
+      $input.shape[2],
+      $input.shape[3]
+    ]);
+    output5D = reshape($output, [
+      1,
+      $output.shape[0],
+      $output.shape[1],
+      $output.shape[2],
+      $output.shape[3]
+    ]);
+  }
+  assert(dy5D.rank === 5, () => `Error in maxPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);
+  assert(input5D.rank === 5, () => `Error in maxPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);
+  assert(output5D.rank === 5, () => `Error in maxPool3dGrad: output must be rank 5 but got rank ${output5D.rank}.`);
+  checkPadOnDimRoundingMode("maxPool3dGrad", pad3, dimRoundingMode);
+  const inputs = { dy: dy5D, input: input5D, output: output5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(MaxPool3DGrad, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var maxPool3dGrad = op({ maxPool3dGrad_ });
+var maxPool3DGradConfig = {
+  kernelName: MaxPool3D,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+    return {
+      x: () => maxPool3dGrad(dy, x, y, filterSize, strides, pad3, dimRoundingMode)
+    };
+  }
+};
+function maxPoolGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "maxPoolGrad");
+  const $input = convertToTensor(input2, "input", "maxPoolGrad");
+  const $output = convertToTensor(output, "output", "maxPoolGrad");
+  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);
+  assert($dy.rank === 4, () => `Error in maxPoolGrad: dy must be rank 4 but got rank ${$dy.rank}.`);
+  assert($input.rank === 4, () => `Error in maxPoolGrad: input must be rank 4 but got rank ${$input.rank}.`);
+  checkPadOnDimRoundingMode("maxPoolGrad", pad3, dimRoundingMode);
+  const inputs = { dy: $dy, input: $input, output: $output };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  return ENGINE.runKernel(MaxPoolGrad, inputs, attrs);
+}
+var maxPoolGrad = op({ maxPoolGrad_ });
+var maxPoolGradConfig = {
+  kernelName: MaxPool,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { filterSize, strides, pad: pad3 } = attrs;
+    return {
+      x: () => maxPoolGrad(dy, x, y, filterSize, strides, pad3)
+    };
+  }
+};
+var meanGradConfig = {
+  kernelName: Mean,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis } = attrs;
+    const axes = parseAxisParam(axis, x.shape);
+    const shapes = computeOutAndReduceShapes(x.shape, axes);
+    const reduceShape = shapes[1];
+    const reduceSize = sizeFromShape(reduceShape);
+    const derX = () => {
+      const expandedDyShape = x.shape.slice();
+      axes.forEach((axis2) => {
+        expandedDyShape[axis2] = 1;
+      });
+      const expandedDy = reshape(dy, expandedDyShape);
+      const res = div(mul(expandedDy, ones2(x.shape, "float32")), reduceSize);
+      return res;
+    };
+    return { x: derX };
+  }
+};
+var minGradConfig = {
+  kernelName: Min,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const minAttrs = attrs;
+    const { axis } = minAttrs;
+    const [x, y] = saved;
+    const origAxes = parseAxisParam(axis, x.shape);
+    const minGrad = gradForMinAndMax(dy, y, x, origAxes);
+    return {
+      x: () => {
+        return minGrad["x"]();
+      }
+    };
+  }
+};
+var minimumGradConfig = {
+  kernelName: Minimum,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const derA = () => mul(dy, cast(lessEqual(a, b), "float32"));
+    const derB = () => mul(dy, cast(greater(a, b), "float32"));
+    return { a: derA, b: derB };
+  }
+};
+var mirrorPadGradConfig = {
+  kernelName: MirrorPad,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const x = saved[0];
+    const { paddings } = attrs;
+    const begin = paddings.map((p2) => p2[0]);
+    return { x: () => slice(dy, begin, x.shape) };
+  }
+};
+var modGradConfig = {
+  kernelName: Mod,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(dy, reduceAxes), a.shape);
+      }
+      return dy;
+    };
+    const derB = () => {
+      const res = mul(dy, neg(floor(div(a, b))));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), b.shape);
+      }
+      return res;
+    };
+    return { a: derA, b: derB };
+  }
+};
+var multiplyGradConfig = {
+  kernelName: Multiply,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = mul(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      const res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), b.shape);
+      }
+      return res;
+    };
+    return { a: derA, b: derB };
+  }
+};
+var negGradConfig = {
+  kernelName: Neg,
+  gradFunc: (dy) => {
+    return { x: () => neg(dy) };
+  }
+};
+var oneHotGradConfig = {
+  kernelName: OneHot,
+  inputsToSave: ["indices"],
+  gradFunc: (dy, saved) => {
+    const indices = saved[0];
+    return { indices: () => zeros(indices.shape, "float32") };
+  }
+};
+var onesLikeGradConfig = {
+  kernelName: OnesLike,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var packGradConfig = {
+  kernelName: Pack,
+  saveAllInputs: true,
+  gradFunc: (dy, saved, attrs) => {
+    const { axis } = attrs;
+    const derTensors = unstack(dy, axis);
+    return derTensors.map((t) => () => t);
+  }
+};
+var padV2GradConfig = {
+  kernelName: PadV2,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const x = saved[0];
+    const { paddings } = attrs;
+    const begin = paddings.map((p2) => p2[0]);
+    return { x: () => slice(dy, begin, x.shape) };
+  }
+};
+var powGradConfig = {
+  kernelName: Pow,
+  inputsToSave: ["a", "b"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [a, b, y] = saved;
+    const base = a;
+    const exp4 = b;
+    const outShape = assertAndGetBroadcastShape(base.shape, exp4.shape);
+    const derBase = () => {
+      const expFloat = cast(exp4, "float32");
+      let res = mul(dy, mul(expFloat, pow(base, sub(expFloat, scalar(1)))));
+      const reduceAxes = getReductionAxes(base.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, base.shape);
+    };
+    const derExp = () => {
+      const condition = greater(base, 0);
+      const logBase = where(condition, log2(base), zerosLike(base));
+      let res = mul(dy, mul(y, logBase));
+      const reduceAxes = getReductionAxes(exp4.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, exp4.shape);
+    };
+    return { a: derBase, b: derExp };
+  }
+};
+var preluGradConfig = {
+  kernelName: Prelu,
+  inputsToSave: ["x", "alpha"],
+  gradFunc: (dy, saved) => {
+    const [x, alpha] = saved;
+    const mask = greater(x, 0);
+    return {
+      x: () => where(mask, dy, mul(dy, alpha)),
+      alpha: () => {
+        let res = where(mask, zerosLike(dy), mul(dy, x));
+        const reduceAxes = getReductionAxes(alpha.shape, dy.shape);
+        if (reduceAxes.length > 0) {
+          res = sum2(res, reduceAxes);
+        }
+        return reshape(res, alpha.shape);
+      }
+    };
+  }
+};
+function prodGradFn_(x, dy, axis) {
+  const expandedYShape = x.shape.slice();
+  expandedYShape[axis] = 1;
+  const expandedDy = reshape(dy, expandedYShape);
+  const xCumProd = cumprod(x, axis, true, false);
+  const xCumRevProd = cumprod(x, axis, true, true);
+  const dx = mul(xCumProd, xCumRevProd);
+  return mul(expandedDy, dx);
+}
+function prodsGradFn_(x, dy, axis) {
+  const xRank = x.shape.length;
+  const finalProdAxis = xRank - axis.length;
+  const xPermutation = backend_util_exports.getAxesPermutation(axis, xRank);
+  let permutedX = x;
+  if (xPermutation != null) {
+    permutedX = transpose(x, xPermutation);
+  }
+  const newShape = permutedX.shape.slice();
+  const removedShape = newShape.splice(xRank - axis.length, axis.length);
+  const endPartShape = removedShape.reduce((p2, c) => p2 * c, 1);
+  newShape.push(endPartShape);
+  const reshapedPermutedX = permutedX.reshape(newShape);
+  let prodGrad = prodGradFn_(reshapedPermutedX, dy, finalProdAxis);
+  prodGrad = prodGrad.reshape(permutedX.shape);
+  if (xPermutation != null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(xPermutation);
+    prodGrad = transpose(prodGrad, undoPermutation);
+  }
+  return prodGrad;
+}
+var prodGradConfig = {
+  kernelName: Prod,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis } = attrs;
+    let axisArr = [];
+    if (axis === void 0 || axis === null) {
+      axisArr = x.shape.map((_, i) => i);
+    } else if (typeof axis === "number") {
+      axisArr = [axis];
+    } else {
+      axisArr = axis;
+    }
+    return { x: () => prodsGradFn_(x, dy, axisArr) };
+  }
+};
+var divGradConfig = {
+  kernelName: RealDiv,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = div(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      let res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = reshape(sum2(res, reduceAxes), b.shape);
+      }
+      const tmp = square(b);
+      return neg(div(res, cast(tmp, "float32")));
+    };
+    return { a: derA, b: derB };
+  }
+};
+var reciprocalGradConfig = {
+  kernelName: Reciprocal,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, neg(square(x))) };
+  }
+};
+var relu6GradConfig = {
+  kernelName: Relu6,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    const mask = mul(lessEqual(x, 6), step(x));
+    return { x: () => mul(dy, cast(mask, "float32")) };
+  }
+};
+var reluGradConfig = {
+  kernelName: Relu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, cast(step(x), "float32")) };
+  }
+};
+var reshapeGradConfig = {
+  kernelName: Reshape,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => reshape(dy, x.shape) };
+  }
+};
+var resizeBilinearGradConfig = {
+  kernelName: ResizeBilinear,
+  inputsToSave: ["images"],
+  gradFunc: (dy, saved, attrs) => {
+    const [images] = saved;
+    const inputs = { dy, images };
+    const imagesDer = () => ENGINE.runKernel(ResizeBilinearGrad, inputs, attrs);
+    return { images: imagesDer };
+  }
+};
+var resizeNearestNeighborGradConfig = {
+  kernelName: ResizeNearestNeighbor,
+  inputsToSave: ["images"],
+  gradFunc: (dy, saved, attrs) => {
+    const [images] = saved;
+    const inputs = { dy, images };
+    const imagesDer = () => ENGINE.runKernel(ResizeNearestNeighborGrad, inputs, attrs);
+    return { images: imagesDer };
+  }
+};
+var reverseGradConfig = {
+  kernelName: Reverse,
+  gradFunc: (dy, saved, attrs) => {
+    const { dims } = attrs;
+    const axes = parseAxisParam(dims, dy.shape);
+    return { x: () => reverse(dy, axes) };
+  }
+};
+var roundGradConfig = {
+  kernelName: Round,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var rsqrtGradConfig = {
+  kernelName: Rsqrt,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => neg(div(dy, mul(pow(x, 1.5), 2))) };
+  }
+};
+var selectGradConfig = {
+  kernelName: Select,
+  inputsToSave: ["condition"],
+  gradFunc: (dy, saved) => {
+    const [condition] = saved;
+    return {
+      condition: () => cast(zerosLike(condition), "float32"),
+      t: () => mul(dy, cast(condition, dy.dtype)),
+      e: () => mul(dy, cast(logicalNot(condition), dy.dtype))
+    };
+  }
+};
+var seluGradConfig = {
+  kernelName: Selu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const mask = greater(x, scalar(0));
+        const scaleAlpha2 = scalar(SELU_SCALEALPHA);
+        const scale22 = scalar(SELU_SCALE);
+        const greaterThanZeroDer = mul(dy, scale22);
+        const lessEqualZeroDer = mul(mul(dy, scaleAlpha2), exp(cast(x, "float32")));
+        return where(mask, greaterThanZeroDer, lessEqualZeroDer);
+      }
+    };
+  }
+};
+var sigmoidGradConfig = {
+  kernelName: Sigmoid,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(dy, mul(y, sub(scalar(1), y))) };
+  }
+};
+var signGradConfig = {
+  kernelName: Sign,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var sinGradConfig = {
+  kernelName: Sin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(cos(cast(x, "float32")), dy) };
+  }
+};
+var sinhGradConfig = {
+  kernelName: Sinh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(cosh(cast(x, "float32")), dy) };
+  }
+};
+var sliceGradConfig = {
+  kernelName: Slice,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { begin, size } = attrs;
+    const inputShape = x.shape;
+    const [begin_, size_] = parseSliceParams(x, begin, size);
+    const paddings = [];
+    for (let i = 0; i < dy.rank; i++) {
+      paddings.push([begin_[i], inputShape[i] - begin_[i] - size_[i]]);
+    }
+    return { x: () => pad(dy, paddings) };
+  }
+};
+var softmaxGradConfig = {
+  kernelName: Softmax,
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [y] = saved;
+    const { dim } = attrs;
+    const keepDims = true;
+    const dyTimesY = mul(dy, y);
+    return {
+      logits: () => sub(dyTimesY, mul(sum2(dyTimesY, [dim], keepDims), y))
+    };
+  }
+};
+var softplusGradConfig = {
+  kernelName: Softplus,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, sigmoid(x)) };
+  }
+};
+var spaceToBatchNDGradConfig = {
+  kernelName: SpaceToBatchND,
+  gradFunc: (dy, saved, attrs) => {
+    const { blockShape, paddings } = attrs;
+    return { x: () => batchToSpaceND(dy, blockShape, paddings) };
+  }
+};
+var splitVGradConfig = {
+  kernelName: SplitV,
+  gradFunc: (dy, saved, attrs) => {
+    const { axis } = attrs;
+    return { x: () => concat(dy, axis) };
+  }
+};
+var sqrtGradConfig = {
+  kernelName: Sqrt,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, mul(sqrt(cast(x, "float32")), 2)) };
+  }
+};
+var squareGradConfig = {
+  kernelName: Square,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, mul(cast(x, "float32"), 2)) };
+  }
+};
+var squaredDifferenceGradConfig = {
+  kernelName: SquaredDifference,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const two = scalar(2);
+    const derA = () => mul(dy, mul(two, sub(a, b)));
+    const derB = () => mul(dy, mul(two, sub(b, a)));
+    return { a: derA, b: derB };
+  }
+};
+var stepGradConfig = {
+  kernelName: Step,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var subGradConfig = {
+  kernelName: Sub,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(neg(res), b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+var sumGradConfig = {
+  kernelName: Sum,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const expandedDyShape = x.shape.slice();
+    const { axis } = attrs;
+    const axes = parseAxisParam(axis, x.shape);
+    axes.forEach((axis2) => {
+      expandedDyShape[axis2] = 1;
+    });
+    const expandedDy = reshape(dy, expandedDyShape);
+    const derX = mul(expandedDy, ones2(x.shape, "float32"));
+    return { x: () => derX };
+  }
+};
+var tanGradConfig = {
+  kernelName: Tan,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, square(cos(x))) };
+  }
+};
+var tanhGradConfig = {
+  kernelName: Tanh,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(sub(scalar(1), square(y)), dy) };
+  }
+};
+var tileGradConfig = {
+  kernelName: Tile,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { reps } = attrs;
+    const derX = () => {
+      let xGrad = zerosLike(x);
+      if (x.rank === 1) {
+        for (let i = 0; i < reps[0]; ++i) {
+          xGrad = add2(xGrad, slice(dy, [i * x.shape[0]], [x.shape[0]]));
+        }
+      } else if (x.rank === 2) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1]], [
+              x.shape[0],
+              x.shape[1]
+            ]));
+          }
+        }
+      } else if (x.rank === 3) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            for (let k = 0; k < reps[2]; ++k) {
+              xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1], k * x.shape[2]], [x.shape[0], x.shape[1], x.shape[2]]));
+            }
+          }
+        }
+      } else if (x.rank === 4) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            for (let k = 0; k < reps[2]; ++k) {
+              for (let l = 0; l < reps[3]; ++l) {
+                xGrad = add2(xGrad, slice(dy, [
+                  i * x.shape[0],
+                  j * x.shape[1],
+                  k * x.shape[2],
+                  l * x.shape[3]
+                ], [x.shape[0], x.shape[1], x.shape[2], x.shape[3]]));
+              }
+            }
+          }
+        }
+      } else {
+        throw new Error(`Gradient for tile operation is not implemented for rank-${x.rank} tensors yet.`);
+      }
+      return xGrad;
+    };
+    return { x: derX };
+  }
+};
+var transposeGradConfig = {
+  kernelName: Transpose,
+  gradFunc: (dy, saved, attrs) => {
+    const transposeAttrs = attrs;
+    const { perm } = transposeAttrs;
+    const undoPerm = getUndoAxesPermutation(perm);
+    return { x: () => transpose(dy, undoPerm) };
+  }
+};
+var unpackGradConfig = {
+  kernelName: Unpack,
+  gradFunc: (dy, saved, attrs) => {
+    const unpackAttrs = attrs;
+    const { axis } = unpackAttrs;
+    return { value: () => stack(dy, axis) };
+  }
+};
+var unsortedSegmentSumGradConfig = {
+  kernelName: UnsortedSegmentSum,
+  inputsToSave: ["segmentIds"],
+  gradFunc: (dy, saved) => {
+    const [segmentIds] = saved;
+    const derX = () => {
+      return gatherDropNegatives(dy, segmentIds);
+    };
+    return { x: derX };
+  }
+};
+function gatherDropNegatives(x, indices) {
+  const zeroClippedIndices = maximum(indices, zerosLike(indices));
+  const gathered = gather(x, zeroClippedIndices);
+  let isPositive = greaterEqual(indices, scalar(0, "int32"));
+  const numIters = gathered.rank - isPositive.rank;
+  for (let i = 0; i < numIters; ++i) {
+    isPositive = expandDims(isPositive, i + 1);
+  }
+  isPositive = logicalAnd(isPositive, ones2(gathered.shape, "bool"));
+  const zeroSlice = zerosLike(gathered);
+  return where(isPositive, gathered, zeroSlice);
+}
+var zerosLikeGradConfig = {
+  kernelName: ZerosLike,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+var gradConfigs = [
+  absGradConfig,
+  acosGradConfig,
+  acoshGradConfig,
+  addGradConfig,
+  addNGradConfig,
+  argMaxGradConfig,
+  argMinGradConfig,
+  asinGradConfig,
+  asinhGradConfig,
+  atan2GradConfig,
+  atanGradConfig,
+  atanhGradConfig,
+  avgPool3DGradConfig,
+  avgPoolGradConfig,
+  batchMatMulGradConfig,
+  batchToSpaceNDGradConfig,
+  broadcastToGradConfig,
+  castGradConfig,
+  ceilGradConfig,
+  clipByValueGradConfig,
+  complexAbsGradConfig,
+  concatGradConfig,
+  conv2DBackpropInputGradConfig,
+  conv2DGradConfig,
+  conv3DGradConfig,
+  cosGradConfig,
+  coshGradConfig,
+  cumsumGradConfig,
+  depthwiseConv2dNativeGradConfig,
+  dilation2dGradConfig,
+  divGradConfig,
+  eluGradConfig,
+  erfGradConfig,
+  expGradConfig,
+  expandDimsGradConfig,
+  expm1GradConfig,
+  floorDivGradConfig,
+  floorGradConfig,
+  fusedBatchNormGradConfig,
+  gatherGradConfig,
+  greaterEqualGradConfig,
+  identityGradConfig,
+  isFiniteGradConfig,
+  isInfGradConfig,
+  isNanGradConfig,
+  leakyReluGradConfig,
+  log1pGradConfig,
+  logGradConfig,
+  logSoftmaxGradConfig,
+  lrnGradConfig,
+  maxGradConfig,
+  maxGradConfig,
+  maximumGradConfig,
+  maxPool3DGradConfig,
+  maxPoolGradConfig,
+  meanGradConfig,
+  minGradConfig,
+  minimumGradConfig,
+  mirrorPadGradConfig,
+  modGradConfig,
+  multiplyGradConfig,
+  negGradConfig,
+  oneHotGradConfig,
+  onesLikeGradConfig,
+  packGradConfig,
+  padV2GradConfig,
+  padV2GradConfig,
+  powGradConfig,
+  preluGradConfig,
+  prodGradConfig,
+  reciprocalGradConfig,
+  relu6GradConfig,
+  reluGradConfig,
+  reshapeGradConfig,
+  resizeBilinearGradConfig,
+  resizeNearestNeighborGradConfig,
+  reverseGradConfig,
+  roundGradConfig,
+  rsqrtGradConfig,
+  selectGradConfig,
+  seluGradConfig,
+  sigmoidGradConfig,
+  signGradConfig,
+  sinGradConfig,
+  sinhGradConfig,
+  sliceGradConfig,
+  softmaxGradConfig,
+  softplusGradConfig,
+  spaceToBatchNDGradConfig,
+  spaceToBatchNDGradConfig,
+  splitVGradConfig,
+  splitVGradConfig,
+  sqrtGradConfig,
+  squaredDifferenceGradConfig,
+  squareGradConfig,
+  stepGradConfig,
+  subGradConfig,
+  sumGradConfig,
+  tanGradConfig,
+  tanhGradConfig,
+  tileGradConfig,
+  transposeGradConfig,
+  unpackGradConfig,
+  unsortedSegmentSumGradConfig,
+  zerosLikeGradConfig
+];
+for (const gradientConfig of gradConfigs) {
+  registerGradient(gradientConfig);
+}
+getGlobalTensorClass().prototype.abs = function() {
+  this.throwIfDisposed();
+  return abs(this);
+};
+getGlobalTensorClass().prototype.acos = function() {
+  this.throwIfDisposed();
+  return acos(this);
+};
+getGlobalTensorClass().prototype.acosh = function() {
+  this.throwIfDisposed();
+  return acosh(this);
+};
+getGlobalTensorClass().prototype.add = function(b) {
+  this.throwIfDisposed();
+  return add2(this, b);
+};
+getGlobalTensorClass().prototype.all = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return all(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.any = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return any(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.argMax = function(axis) {
+  this.throwIfDisposed();
+  return argMax(this, axis);
+};
+getGlobalTensorClass().prototype.argMin = function(axis) {
+  this.throwIfDisposed();
+  return argMin(this, axis);
+};
+getGlobalTensorClass().prototype.asScalar = function() {
+  this.throwIfDisposed();
+  assert(this.size === 1, () => "The array must have only 1 element.");
+  return reshape(this, []);
+};
+getGlobalTensorClass().prototype.asType = function(dtype) {
+  this.throwIfDisposed();
+  return cast(this, dtype);
+};
+getGlobalTensorClass().prototype.as1D = function() {
+  this.throwIfDisposed();
+  return reshape(this, [this.size]);
+};
+getGlobalTensorClass().prototype.as2D = function(rows, columns) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns]);
+};
+getGlobalTensorClass().prototype.as3D = function(rows, columns, depth) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth]);
+};
+getGlobalTensorClass().prototype.as4D = function(rows, columns, depth, depth2) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth, depth2]);
+};
+getGlobalTensorClass().prototype.as5D = function(rows, columns, depth, depth2, depth3) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth, depth2, depth3]);
+};
+getGlobalTensorClass().prototype.asin = function() {
+  this.throwIfDisposed();
+  return asin(this);
+};
+getGlobalTensorClass().prototype.asinh = function() {
+  this.throwIfDisposed();
+  return asinh(this);
+};
+getGlobalTensorClass().prototype.atan = function() {
+  this.throwIfDisposed();
+  return atan(this);
+};
+getGlobalTensorClass().prototype.atan2 = function(b) {
+  this.throwIfDisposed();
+  return atan2(this, b);
+};
+getGlobalTensorClass().prototype.atanh = function() {
+  this.throwIfDisposed();
+  return atanh(this);
+};
+getGlobalTensorClass().prototype.avgPool = function(filterSize, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return avgPool(this, filterSize, strides, pad3, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.batchToSpaceND = function(blockShape, crops) {
+  this.throwIfDisposed();
+  return batchToSpaceND(this, blockShape, crops);
+};
+getGlobalTensorClass().prototype.batchNorm = function(mean4, variance, offset, scale22, varianceEpsilon) {
+  this.throwIfDisposed();
+  return batchNorm(this, mean4, variance, offset, scale22, varianceEpsilon);
+};
+getGlobalTensorClass().prototype.broadcastTo = function(shape) {
+  this.throwIfDisposed();
+  return broadcastTo(this, shape);
+};
+getGlobalTensorClass().prototype.cast = function(dtype) {
+  this.throwIfDisposed();
+  return cast(this, dtype);
+};
+getGlobalTensorClass().prototype.ceil = function() {
+  this.throwIfDisposed();
+  return ceil(this);
+};
+getGlobalTensorClass().prototype.clipByValue = function(min6, max6) {
+  this.throwIfDisposed();
+  return clipByValue(this, min6, max6);
+};
+getGlobalTensorClass().prototype.concat = function(x, axis) {
+  this.throwIfDisposed();
+  if (x instanceof Tensor) {
+    x = [x];
+  }
+  return concat([this, ...x], axis);
+};
+getGlobalTensorClass().prototype.conv1d = function(filter, stride, pad3, dataFormat, dilation, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv1d(this, filter, stride, pad3, dataFormat, dilation, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.conv2dTranspose = function(filter, outputShape, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv2dTranspose(this, filter, outputShape, strides, pad3, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.conv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.cos = function() {
+  this.throwIfDisposed();
+  return cos(this);
+};
+getGlobalTensorClass().prototype.cosh = function() {
+  this.throwIfDisposed();
+  return cosh(this);
+};
+getGlobalTensorClass().prototype.cumprod = function(axis, exclusive, reverse5) {
+  this.throwIfDisposed();
+  return cumprod(this, axis, exclusive, reverse5);
+};
+getGlobalTensorClass().prototype.cumsum = function(axis, exclusive, reverse5) {
+  this.throwIfDisposed();
+  return cumsum(this, axis, exclusive, reverse5);
+};
+getGlobalTensorClass().prototype.depthToSpace = function(blockSize, dataFormat) {
+  this.throwIfDisposed();
+  return depthToSpace(this, blockSize, dataFormat);
+};
+getGlobalTensorClass().prototype.depthwiseConv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {
+  this.throwIfDisposed();
+  return depthwiseConv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.dilation2d = function(filter, strides, pad3, dilations, dataFormat) {
+  this.throwIfDisposed();
+  return dilation2d(this, filter, strides, pad3, dilations, dataFormat);
+};
+getGlobalTensorClass().prototype.divNoNan = function(b) {
+  this.throwIfDisposed();
+  return divNoNan(this, b);
+};
+getGlobalTensorClass().prototype.div = function(b) {
+  this.throwIfDisposed();
+  return div(this, b);
+};
+getGlobalTensorClass().prototype.dot = function(b) {
+  this.throwIfDisposed();
+  return dot(this, b);
+};
+getGlobalTensorClass().prototype.elu = function() {
+  this.throwIfDisposed();
+  return elu(this);
+};
+getGlobalTensorClass().prototype.equal = function(b) {
+  this.throwIfDisposed();
+  return equal(this, b);
+};
+getGlobalTensorClass().prototype.erf = function() {
+  this.throwIfDisposed();
+  return erf(this);
+};
+getGlobalTensorClass().prototype.euclideanNorm = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return euclideanNorm(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.exp = function() {
+  this.throwIfDisposed();
+  return exp(this);
+};
+getGlobalTensorClass().prototype.expandDims = function(axis) {
+  this.throwIfDisposed();
+  return expandDims(this, axis);
+};
+getGlobalTensorClass().prototype.expm1 = function() {
+  this.throwIfDisposed();
+  return expm1(this);
+};
+getGlobalTensorClass().prototype.fft = function() {
+  this.throwIfDisposed();
+  return fft(this);
+};
+getGlobalTensorClass().prototype.flatten = function() {
+  this.throwIfDisposed();
+  return reshape(this, [this.size]);
+};
+getGlobalTensorClass().prototype.floor = function() {
+  this.throwIfDisposed();
+  return floor(this);
+};
+getGlobalTensorClass().prototype.floorDiv = function(b) {
+  this.throwIfDisposed();
+  return floorDiv(this, b);
+};
+getGlobalTensorClass().prototype.gather = function(indices, axis) {
+  this.throwIfDisposed();
+  return gather(this, indices, axis);
+};
+getGlobalTensorClass().prototype.greaterEqual = function(b) {
+  this.throwIfDisposed();
+  return greaterEqual(this, b);
+};
+getGlobalTensorClass().prototype.greater = function(b) {
+  this.throwIfDisposed();
+  return greater(this, b);
+};
+getGlobalTensorClass().prototype.ifft = function() {
+  this.throwIfDisposed();
+  return ifft(this);
+};
+getGlobalTensorClass().prototype.irfft = function() {
+  this.throwIfDisposed();
+  return irfft(this);
+};
+getGlobalTensorClass().prototype.isFinite = function() {
+  this.throwIfDisposed();
+  return isFinite2(this);
+};
+getGlobalTensorClass().prototype.isInf = function() {
+  this.throwIfDisposed();
+  return isInf(this);
+};
+getGlobalTensorClass().prototype.isNaN = function() {
+  this.throwIfDisposed();
+  return isNaN2(this);
+};
+getGlobalTensorClass().prototype.leakyRelu = function(alpha) {
+  this.throwIfDisposed();
+  return leakyRelu(this, alpha);
+};
+getGlobalTensorClass().prototype.lessEqual = function(b) {
+  this.throwIfDisposed();
+  return lessEqual(this, b);
+};
+getGlobalTensorClass().prototype.less = function(b) {
+  this.throwIfDisposed();
+  return less(this, b);
+};
+getGlobalTensorClass().prototype.localResponseNormalization = function(depthRadius, bias, alpha, beta) {
+  this.throwIfDisposed();
+  return localResponseNormalization(this, depthRadius, bias, alpha, beta);
+};
+getGlobalTensorClass().prototype.logSigmoid = function() {
+  this.throwIfDisposed();
+  return logSigmoid(this);
+};
+getGlobalTensorClass().prototype.logSoftmax = function(axis) {
+  this.throwIfDisposed();
+  return logSoftmax(this, axis);
+};
+getGlobalTensorClass().prototype.logSumExp = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return logSumExp(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.log = function() {
+  this.throwIfDisposed();
+  return log2(this);
+};
+getGlobalTensorClass().prototype.log1p = function() {
+  this.throwIfDisposed();
+  return log1p(this);
+};
+getGlobalTensorClass().prototype.logicalAnd = function(b) {
+  this.throwIfDisposed();
+  return logicalAnd(this, b);
+};
+getGlobalTensorClass().prototype.logicalNot = function() {
+  this.throwIfDisposed();
+  return logicalNot(this);
+};
+getGlobalTensorClass().prototype.logicalOr = function(b) {
+  this.throwIfDisposed();
+  return logicalOr(this, b);
+};
+getGlobalTensorClass().prototype.logicalXor = function(b) {
+  this.throwIfDisposed();
+  return logicalXor(this, b);
+};
+getGlobalTensorClass().prototype.matMul = function(b, transposeA, transposeB) {
+  this.throwIfDisposed();
+  return matMul(this, b, transposeA, transposeB);
+};
+getGlobalTensorClass().prototype.maxPool = function(filterSize, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return maxPool(this, filterSize, strides, pad3, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.max = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return max(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.maximum = function(b) {
+  this.throwIfDisposed();
+  return maximum(this, b);
+};
+getGlobalTensorClass().prototype.mean = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return mean(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.min = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return min(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.minimum = function(b) {
+  this.throwIfDisposed();
+  return minimum(this, b);
+};
+getGlobalTensorClass().prototype.mirrorPad = function(paddings, mode) {
+  this.throwIfDisposed();
+  return mirrorPad(this, paddings, mode);
+};
+getGlobalTensorClass().prototype.mod = function(b) {
+  this.throwIfDisposed();
+  return mod(this, b);
+};
+getGlobalTensorClass().prototype.mul = function(b) {
+  this.throwIfDisposed();
+  return mul(this, b);
+};
+getGlobalTensorClass().prototype.neg = function() {
+  this.throwIfDisposed();
+  return neg(this);
+};
+getGlobalTensorClass().prototype.norm = function(ord, axis, keepDims) {
+  this.throwIfDisposed();
+  return norm(this, ord, axis, keepDims);
+};
+getGlobalTensorClass().prototype.notEqual = function(b) {
+  this.throwIfDisposed();
+  return notEqual(this, b);
+};
+getGlobalTensorClass().prototype.oneHot = function(depth, onValue = 1, offValue = 0) {
+  this.throwIfDisposed();
+  return oneHot(this, depth, onValue, offValue);
+};
+getGlobalTensorClass().prototype.onesLike = function() {
+  this.throwIfDisposed();
+  return onesLike(this);
+};
+getGlobalTensorClass().prototype.pad = function(paddings, constantValue) {
+  this.throwIfDisposed();
+  return pad(this, paddings, constantValue);
+};
+getGlobalTensorClass().prototype.pool = function(windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode) {
+  this.throwIfDisposed();
+  return pool(this, windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode);
+};
+getGlobalTensorClass().prototype.pow = function(exp4) {
+  this.throwIfDisposed();
+  return pow(this, exp4);
+};
+getGlobalTensorClass().prototype.prelu = function(alpha) {
+  this.throwIfDisposed();
+  return prelu(this, alpha);
+};
+getGlobalTensorClass().prototype.prod = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return prod(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.reciprocal = function() {
+  this.throwIfDisposed();
+  return reciprocal(this);
+};
+getGlobalTensorClass().prototype.relu = function() {
+  this.throwIfDisposed();
+  return relu(this);
+};
+getGlobalTensorClass().prototype.relu6 = function() {
+  this.throwIfDisposed();
+  return relu6(this);
+};
+getGlobalTensorClass().prototype.reshapeAs = function(x) {
+  this.throwIfDisposed();
+  return reshape(this, x.shape);
+};
+getGlobalTensorClass().prototype.reshape = function(shape) {
+  this.throwIfDisposed();
+  return reshape(this, shape);
+};
+getGlobalTensorClass().prototype.resizeBilinear = function(newShape2D, alignCorners, halfPixelCenters) {
+  this.throwIfDisposed();
+  return resizeBilinear(this, newShape2D, alignCorners, halfPixelCenters);
+};
+getGlobalTensorClass().prototype.resizeNearestNeighbor = function(newShape2D, alignCorners, halfFloatCenters) {
+  this.throwIfDisposed();
+  return resizeNearestNeighbor(this, newShape2D, alignCorners, halfFloatCenters);
+};
+getGlobalTensorClass().prototype.reverse = function(axis) {
+  this.throwIfDisposed();
+  return reverse(this, axis);
+};
+getGlobalTensorClass().prototype.rfft = function() {
+  this.throwIfDisposed();
+  return rfft(this);
+};
+getGlobalTensorClass().prototype.round = function() {
+  this.throwIfDisposed();
+  return round2(this);
+};
+getGlobalTensorClass().prototype.rsqrt = function() {
+  this.throwIfDisposed();
+  return rsqrt(this);
+};
+getGlobalTensorClass().prototype.selu = function() {
+  this.throwIfDisposed();
+  return selu(this);
+};
+getGlobalTensorClass().prototype.separableConv2d = function(depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat) {
+  this.throwIfDisposed();
+  return separableConv2d(this, depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat);
+};
+getGlobalTensorClass().prototype.sigmoid = function() {
+  this.throwIfDisposed();
+  return sigmoid(this);
+};
+getGlobalTensorClass().prototype.sign = function() {
+  this.throwIfDisposed();
+  return sign(this);
+};
+getGlobalTensorClass().prototype.sin = function() {
+  this.throwIfDisposed();
+  return sin(this);
+};
+getGlobalTensorClass().prototype.sinh = function() {
+  this.throwIfDisposed();
+  return sinh(this);
+};
+getGlobalTensorClass().prototype.slice = function(begin, size) {
+  this.throwIfDisposed();
+  return slice(this, begin, size);
+};
+getGlobalTensorClass().prototype.softmax = function(dim) {
+  this.throwIfDisposed();
+  return softmax(this, dim);
+};
+getGlobalTensorClass().prototype.softplus = function() {
+  this.throwIfDisposed();
+  return softplus(this);
+};
+getGlobalTensorClass().prototype.spaceToBatchND = function(blockShape, paddings) {
+  this.throwIfDisposed();
+  return spaceToBatchND(this, blockShape, paddings);
+};
+getGlobalTensorClass().prototype.split = function(numOrSizeSplits, axis) {
+  this.throwIfDisposed();
+  return split(this, numOrSizeSplits, axis);
+};
+getGlobalTensorClass().prototype.sqrt = function() {
+  this.throwIfDisposed();
+  return sqrt(this);
+};
+getGlobalTensorClass().prototype.square = function() {
+  this.throwIfDisposed();
+  return square(this);
+};
+getGlobalTensorClass().prototype.squaredDifference = function(b) {
+  this.throwIfDisposed();
+  return squaredDifference(this, b);
+};
+getGlobalTensorClass().prototype.squeeze = function(axis) {
+  this.throwIfDisposed();
+  return squeeze(this, axis);
+};
+getGlobalTensorClass().prototype.stack = function(x, axis) {
+  this.throwIfDisposed();
+  const tensorsToBeStacked = x instanceof Tensor ? [this, x] : [this, ...x];
+  return stack(tensorsToBeStacked, axis);
+};
+getGlobalTensorClass().prototype.step = function(alpha) {
+  this.throwIfDisposed();
+  return step(this, alpha);
+};
+getGlobalTensorClass().prototype.stridedSlice = function(begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {
+  this.throwIfDisposed();
+  return stridedSlice(this, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+};
+getGlobalTensorClass().prototype.sub = function(b) {
+  this.throwIfDisposed();
+  return sub(this, b);
+};
+getGlobalTensorClass().prototype.sum = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return sum2(this, axis, keepDims);
+};
+getGlobalTensorClass().prototype.tan = function() {
+  this.throwIfDisposed();
+  return tan(this);
+};
+getGlobalTensorClass().prototype.tanh = function() {
+  this.throwIfDisposed();
+  return tanh2(this);
+};
+getGlobalTensorClass().prototype.tile = function(reps) {
+  this.throwIfDisposed();
+  return tile(this, reps);
+};
+getGlobalTensorClass().prototype.toBool = function() {
+  this.throwIfDisposed();
+  return cast(this, "bool");
+};
+getGlobalTensorClass().prototype.toFloat = function() {
+  this.throwIfDisposed();
+  return cast(this, "float32");
+};
+getGlobalTensorClass().prototype.toInt = function() {
+  this.throwIfDisposed();
+  return cast(this, "int32");
+};
+getGlobalTensorClass().prototype.topk = function(k, sorted) {
+  this.throwIfDisposed();
+  return topk(this, k, sorted);
+};
+getGlobalTensorClass().prototype.transpose = function(perm) {
+  this.throwIfDisposed();
+  return transpose(this, perm);
+};
+getGlobalTensorClass().prototype.unique = function(axis) {
+  this.throwIfDisposed();
+  return unique(this, axis);
+};
+getGlobalTensorClass().prototype.unsortedSegmentSum = function(segmentIds, numSegments) {
+  this.throwIfDisposed();
+  return unsortedSegmentSum(this, segmentIds, numSegments);
+};
+getGlobalTensorClass().prototype.unstack = function(axis) {
+  this.throwIfDisposed();
+  return unstack(this, axis);
+};
+getGlobalTensorClass().prototype.where = function(condition, x) {
+  this.throwIfDisposed();
+  return where(condition, this, x);
+};
+getGlobalTensorClass().prototype.zerosLike = function() {
+  this.throwIfDisposed();
+  return zerosLike(this);
+};
+var AttributeError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, AttributeError.prototype);
+  }
+};
+var RuntimeError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, RuntimeError.prototype);
+  }
+};
+var ValueError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, ValueError.prototype);
+  }
+};
+var NotImplementedError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, NotImplementedError.prototype);
+  }
+};
+var AssertionError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, AssertionError.prototype);
+  }
+};
+var LruCache = class {
+  constructor(maxEntries) {
+    this.maxEntries = maxEntries || 100;
+    this.cache = /* @__PURE__ */ new Map();
+  }
+  get(key) {
+    let entry;
+    if (this.cache.has(key)) {
+      entry = this.cache.get(key);
+      this.cache.delete(key);
+      this.cache.set(key, entry);
+    }
+    return entry;
+  }
+  put(key, value) {
+    if (this.cache.has(key)) {
+      this.cache.delete(key);
+    } else if (this.cache.size >= this.maxEntries) {
+      const keyToDelete = this.cache.keys().next().value;
+      this.cache.delete(keyToDelete);
+    }
+    this.cache.set(key, value);
+  }
+  getMaxEntries() {
+    return this.maxEntries;
+  }
+  setMaxEntries(maxEntries) {
+    if (maxEntries < 0) {
+      throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${maxEntries}.`);
+    }
+    if (this.maxEntries > maxEntries) {
+      for (let i = 0; i < this.maxEntries - maxEntries; i++) {
+        const keyToDelete = this.cache.keys().next().value;
+        this.cache.delete(keyToDelete);
+      }
+    }
+    this.maxEntries = maxEntries;
+  }
+};
+function pyListRepeat(value, numValues) {
+  if (Array.isArray(value)) {
+    let newArray = [];
+    for (let i = 0; i < numValues; i++) {
+      newArray = newArray.concat(value);
+    }
+    return newArray;
+  } else {
+    const newArray = new Array(numValues);
+    newArray.fill(value);
+    return newArray;
+  }
+}
+function assert2(val, message) {
+  if (!val) {
+    throw new AssertionError(message);
+  }
+}
+function count(array2, refernce) {
+  let counter = 0;
+  for (const item of array2) {
+    if (item === refernce) {
+      counter++;
+    }
+  }
+  return counter;
+}
+function singletonOrArray(xs) {
+  if (xs.length === 1) {
+    return xs[0];
+  }
+  return xs;
+}
+function toList(x) {
+  if (Array.isArray(x)) {
+    return x;
+  }
+  return [x];
+}
+function toSnakeCase(name) {
+  const intermediate = name.replace(/(.)([A-Z][a-z0-9]+)/g, "$1_$2");
+  const insecure = intermediate.replace(/([a-z])([A-Z])/g, "$1_$2").toLowerCase();
+  if (insecure[0] !== "_") {
+    return insecure;
+  }
+  return "private" + insecure;
+}
+function toCamelCase(identifier) {
+  if (identifier.length <= 1) {
+    return identifier;
+  }
+  if (identifier.indexOf("_") === -1) {
+    return identifier;
+  }
+  return identifier.replace(/[_]+(\w|$)/g, (m, p1) => p1.toUpperCase());
+}
+var _GLOBAL_CUSTOM_OBJECTS = {};
+function serializeKerasObject(instance) {
+  if (instance === null || instance === void 0) {
+    return null;
+  }
+  const dict = {};
+  dict["className"] = instance.getClassName();
+  dict["config"] = instance.getConfig();
+  return dict;
+}
+function convertNDArrayScalarsInConfig(config) {
+  if (config == null || typeof config !== "object") {
+    return;
+  } else if (Array.isArray(config)) {
+    config.forEach((configItem) => convertNDArrayScalarsInConfig(configItem));
+  } else {
+    const fields = Object.keys(config);
+    for (const field of fields) {
+      const value = config[field];
+      if (value != null && typeof value === "object") {
+        if (!Array.isArray(value) && value["type"] === "ndarray" && typeof value["value"] === "number") {
+          config[field] = value["value"];
+        } else {
+          convertNDArrayScalarsInConfig(value);
+        }
+      }
+    }
+  }
+}
+function deserializeKerasObject(identifier, moduleObjects = {}, customObjects = {}, printableModuleName = "object", fastWeightInit = false) {
+  if (typeof identifier === "string") {
+    const functionName = identifier;
+    let fn;
+    if (functionName in customObjects) {
+      fn = customObjects[functionName];
+    } else if (functionName in _GLOBAL_CUSTOM_OBJECTS) {
+      fn = _GLOBAL_CUSTOM_OBJECTS[functionName];
+    } else {
+      fn = moduleObjects[functionName];
+      if (fn == null) {
+        throw new ValueError(`Unknown ${printableModuleName}: ${identifier}. This may be due to one of the following reasons:
+1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+      }
+    }
+    return fn;
+  } else {
+    const config = identifier;
+    if (config["className"] == null || config["config"] == null) {
+      throw new ValueError(`${printableModuleName}: Improper config format: ${JSON.stringify(config)}.
+'className' and 'config' must set.`);
+    }
+    const className = config["className"];
+    let cls, fromConfig;
+    if (className in customObjects) {
+      [cls, fromConfig] = customObjects[className];
+    } else if (className in _GLOBAL_CUSTOM_OBJECTS) {
+      [cls, fromConfig] = _GLOBAL_CUSTOM_OBJECTS["className"];
+    } else if (className in moduleObjects) {
+      [cls, fromConfig] = moduleObjects[className];
+    }
+    if (cls == null) {
+      throw new ValueError(`Unknown ${printableModuleName}: ${className}. This may be due to one of the following reasons:
+1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+    }
+    if (fromConfig != null) {
+      const customObjectsCombined = {};
+      for (const key of Object.keys(_GLOBAL_CUSTOM_OBJECTS)) {
+        customObjectsCombined[key] = _GLOBAL_CUSTOM_OBJECTS[key];
+      }
+      for (const key of Object.keys(customObjects)) {
+        customObjectsCombined[key] = customObjects[key];
+      }
+      const nestedConfig = config["config"];
+      nestedConfig["customObjects"] = customObjectsCombined;
+      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);
+      for (const key of Object.keys(customObjects)) {
+        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];
+      }
+      convertNDArrayScalarsInConfig(config["config"]);
+      const returnObj = fromConfig(cls, config["config"], customObjects, fastWeightInit);
+      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);
+      return returnObj;
+    } else {
+      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);
+      for (const key of Object.keys(customObjects)) {
+        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];
+      }
+      const returnObj = new cls(config["config"]);
+      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);
+      return returnObj;
+    }
+  }
+}
+function numberCompare(a, b) {
+  return a < b ? -1 : a > b ? 1 : 0;
+}
+function reverseNumberCompare(a, b) {
+  return -1 * numberCompare(a, b);
+}
+function unique2(xs) {
+  if (xs == null) {
+    return xs;
+  }
+  const out = [];
+  for (const x of xs) {
+    if (out.indexOf(x) === -1) {
+      out.push(x);
+    }
+  }
+  return out;
+}
+function isObjectEmpty(obj) {
+  if (obj == null) {
+    throw new ValueError(`Invalid value in obj: ${JSON.stringify(obj)}`);
+  }
+  for (const key in obj) {
+    if (obj.hasOwnProperty(key)) {
+      return false;
+    }
+  }
+  return true;
+}
+function checkStringTypeUnionValue(values, label, value) {
+  if (value == null) {
+    return;
+  }
+  if (values.indexOf(value) < 0) {
+    throw new ValueError(`${value} is not a valid ${label}.  Valid values are ${values} or null/undefined.`);
+  }
+}
+function checkArrayTypeAndLength(x, expectedType, minLength = 0, maxLength = Infinity) {
+  assert2(minLength >= 0);
+  assert2(maxLength >= minLength);
+  return Array.isArray(x) && x.length >= minLength && x.length <= maxLength && x.every((e) => typeof e === expectedType);
+}
+function assertPositiveInteger(value, name) {
+  if (Array.isArray(value)) {
+    util_exports.assert(value.length > 0, () => `${name} is unexpectedly an empty array.`);
+    value.forEach((v, i) => assertPositiveInteger(v, `element ${i + 1} of ${name}`));
+  } else {
+    util_exports.assert(Number.isInteger(value) && value > 0, () => `Expected ${name} to be a positive integer, but got ${formatAsFriendlyString(value)}.`);
+  }
+}
+function formatAsFriendlyString(value) {
+  if (value === null) {
+    return "null";
+  } else if (Array.isArray(value)) {
+    return "[" + value.map((v) => formatAsFriendlyString(v)).join(",") + "]";
+  } else if (typeof value === "string") {
+    return `"${value}"`;
+  } else {
+    return `${value}`;
+  }
+}
+function debounce(f, waitMs, nowFunc) {
+  let lastTime = nowFunc != null ? nowFunc() : util_exports.now();
+  let lastResult;
+  const f2 = (...args) => {
+    const now2 = nowFunc != null ? nowFunc() : util_exports.now();
+    if (now2 - lastTime < waitMs) {
+      return lastResult;
+    }
+    lastTime = now2;
+    lastResult = f(...args);
+    return lastResult;
+  };
+  return f2;
+}
+function mapActivationToFusedKernel(activationName) {
+  if (activationName === "relu") {
+    return "relu";
+  }
+  if (activationName === "linear") {
+    return "linear";
+  }
+  if (activationName === "elu") {
+    return "elu";
+  }
+  return null;
+}
+var _nextUniqueTensorId = 0;
+function getNextUniqueTensorId() {
+  return _nextUniqueTensorId++;
+}
+var _uidPrefixes = {};
+function getUid(prefix = "") {
+  if (!(prefix in _uidPrefixes)) {
+    _uidPrefixes[prefix] = 0;
+  }
+  _uidPrefixes[prefix] += 1;
+  return prefix + _uidPrefixes[prefix].toString();
+}
+var VALID_DATA_FORMAT_VALUES = ["channelsFirst", "channelsLast"];
+var VALID_INTERPOLATION_FORMAT_VALUES = ["nearest", "bilinear"];
+var VALID_PADDING_MODE_VALUES = ["valid", "same", "causal"];
+var VALID_POOL_MODE_VALUES = ["max", "avg"];
+var VALID_BIDIRECTIONAL_MERGE_MODES = ["sum", "mul", "concat", "ave"];
+var nameMap = /* @__PURE__ */ new Map();
+function checkDataFormat(value) {
+  checkStringTypeUnionValue(VALID_DATA_FORMAT_VALUES, "DataFormat", value);
+}
+function checkInterpolationFormat(value) {
+  checkStringTypeUnionValue(VALID_INTERPOLATION_FORMAT_VALUES, "InterpolationFormat", value);
+}
+function checkPaddingMode(value) {
+  checkStringTypeUnionValue(VALID_PADDING_MODE_VALUES, "PaddingMode", value);
+}
+function checkPoolMode(value) {
+  checkStringTypeUnionValue(VALID_POOL_MODE_VALUES, "PoolMode", value);
+}
+var _nameScopeStack = [];
+var _nameScopeDivider = "/";
+function nameScope(name, fn) {
+  _nameScopeStack.push(name);
+  try {
+    const val = fn();
+    _nameScopeStack.pop();
+    return val;
+  } catch (e) {
+    _nameScopeStack.pop();
+    throw e;
+  }
+}
+function currentNameScopePrefix() {
+  if (_nameScopeStack.length === 0) {
+    return "";
+  } else {
+    return _nameScopeStack.join(_nameScopeDivider) + _nameScopeDivider;
+  }
+}
+function getScopedTensorName(tensorName) {
+  if (!isValidTensorName(tensorName)) {
+    throw new Error("Not a valid tensor name: '" + tensorName + "'");
+  }
+  return currentNameScopePrefix() + tensorName;
+}
+function getUniqueTensorName(scopedName) {
+  if (!isValidTensorName(scopedName)) {
+    throw new Error("Not a valid tensor name: '" + scopedName + "'");
+  }
+  if (!nameMap.has(scopedName)) {
+    nameMap.set(scopedName, 0);
+  }
+  const index = nameMap.get(scopedName);
+  nameMap.set(scopedName, nameMap.get(scopedName) + 1);
+  if (index > 0) {
+    const result = `${scopedName}_${index}`;
+    nameMap.set(result, 1);
+    return result;
+  } else {
+    return scopedName;
+  }
+}
+var tensorNameRegex = new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);
+function isValidTensorName(name) {
+  return !!name.match(tensorNameRegex);
+}
+function isInteger(x) {
+  return x === parseInt(x.toString(), 10);
+}
+function arrayProd(array2, begin, end) {
+  if (begin == null) {
+    begin = 0;
+  }
+  if (end == null) {
+    end = array2.length;
+  }
+  let prod5 = 1;
+  for (let i = begin; i < end; ++i) {
+    prod5 *= array2[i];
+  }
+  return prod5;
+}
+function min2(array2) {
+  if (array2.length === 0) {
+    return Number.NaN;
+  }
+  let min6 = Number.POSITIVE_INFINITY;
+  for (let i = 0; i < array2.length; i++) {
+    const value = array2[i];
+    if (value < min6) {
+      min6 = value;
+    }
+  }
+  return min6;
+}
+function max2(array2) {
+  if (array2.length === 0) {
+    return Number.NaN;
+  }
+  let max6 = Number.NEGATIVE_INFINITY;
+  for (let i = 0; i < array2.length; i++) {
+    const value = array2[i];
+    if (value > max6) {
+      max6 = value;
+    }
+  }
+  return max6;
+}
+function range2(begin, end) {
+  if (end < begin) {
+    throw new ValueError(`end (${end}) < begin (${begin}) is forbidden.`);
+  }
+  const out = [];
+  for (let i = begin; i < end; ++i) {
+    out.push(i);
+  }
+  return out;
+}
+var _epsilon;
+function epsilon() {
+  if (_epsilon == null) {
+    _epsilon = backend().epsilon();
+  }
+  return _epsilon;
+}
+function imageDataFormat() {
+  return "channelsLast";
+}
+function cast2(x, dtype) {
+  return cast(x, dtype);
+}
+function expandDims2(x, axis = -1) {
+  const outShape = x.shape.slice();
+  if (axis < 0) {
+    axis = outShape.length + axis + 1;
+  }
+  outShape.splice(axis, 0, 1);
+  return reshape(x, outShape);
+}
+function repeat(x, n) {
+  return tidy(() => {
+    if (x.shape.length !== 2) {
+      throw new ValueError(`repeat() expects a rank-2 tensor, but received a rank-${x.shape.length} tensor.`);
+    }
+    const y = expandDims2(x, 1);
+    return tile2(y, [1, n, 1]);
+  });
+}
+function flatten2(x) {
+  const newShape = [arrayProd(x.shape)];
+  return reshape(x, newShape);
+}
+function batchFlatten(x) {
+  if (x.rank <= 1) {
+    throw new ValueError(`batchFlatten requires a minimum rank of 2. Got rank: ${x.rank}.`);
+  }
+  const newShape = [x.shape[0], arrayProd(x.shape, 1)];
+  return reshape(x, newShape);
+}
+function sliceAlongFirstAxis(array2, start, size) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        return slice2d(array2, [start, 0], [size, array2.shape[1]]);
+      case 3:
+        return slice3d(array2, [start, 0, 0], [size, array2.shape[1], array2.shape[2]]);
+      case 4:
+        return slice4d(array2, [start, 0, 0, 0], [size, array2.shape[1], array2.shape[2], array2.shape[3]]);
+      case 5:
+        return slice(array2, [start, 0, 0, 0, 0], [
+          size,
+          array2.shape[1],
+          array2.shape[2],
+          array2.shape[3],
+          array2.shape[4]
+        ]);
+      case 6:
+        return slice(array2, [start, 0, 0, 0, 0, 0], [
+          size,
+          array2.shape[1],
+          array2.shape[2],
+          array2.shape[3],
+          array2.shape[4],
+          array2.shape[5]
+        ]);
+      default:
+        throw new ValueError(`sliceAlongFirstAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function sliceAlongLastAxis(array2, start, size) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        return slice2d(array2, [0, start], [array2.shape[0], size]);
+      case 3:
+        return slice3d(array2, [0, 0, start], [array2.shape[0], array2.shape[1], size]);
+      case 4:
+        return slice4d(array2, [0, 0, 0, start], [array2.shape[0], array2.shape[1], array2.shape[2], size]);
+      default:
+        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function sliceAlongAxis(array2, start, size, axis) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      case 3:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return slice3d(array2, [0, start, 0], [array2.shape[0], size, array2.shape[2]]);
+          case 3:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      case 4:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return slice4d(array2, [0, start, 0, 0], [array2.shape[0], size, array2.shape[2], array2.shape[3]]);
+          case 3:
+            return slice4d(array2, [0, 0, start, 0], [array2.shape[0], array2.shape[1], size, array2.shape[3]]);
+          case 4:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      default:
+        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function concatenate(tensors, axis = -1) {
+  let rank;
+  if (axis < 0) {
+    rank = tensors[0].rank;
+    if (rank !== 0) {
+      axis = rank;
+    } else {
+      axis = 0;
+    }
+  }
+  if (axis === tensors[0].rank) {
+    axis = -1;
+  }
+  return concat(tensors, axis);
+}
+function concatAlongFirstAxis(a, b) {
+  switch (a.rank) {
+    case 1:
+      return concat1d([a, b]);
+    case 2:
+      return concat2d([a, b], 0);
+    case 3:
+      return concat3d([a, b], 0);
+    case 4:
+      return concat4d([a, b], 0);
+    default:
+      throw new ValueError(`concatAlongFirstAxis() received an unsupported tensor rank: ${a.rank}`);
+  }
+}
+function tile2(x, n) {
+  if (!Array.isArray(n)) {
+    n = [n];
+  }
+  if (x.rank !== n.length) {
+    throw new ValueError(`The length of input n (${n.length}) does not match the number of dimensions in input x (${x.rank})`);
+  }
+  return tile(x, n);
+}
+function randomNormal2(shape, mean4 = 0, stddev = 1, dtype, seed) {
+  return randomNormal(shape, mean4, stddev, dtype, seed);
+}
+function dot2(a, b, activation2, bias) {
+  if (a.rank < 2 || b.rank < 2) {
+    throw new NotImplementedError(`dot requires both inputs to be rank >= 2 but got x shape = ${a.shape} and y shape = ${b.shape}`);
+  }
+  if (b.rank >= 3) {
+    const xLastDim = a.shape.slice(-1)[0];
+    const ySecondLastDim = b.shape.slice(-2)[0];
+    if (xLastDim !== ySecondLastDim) {
+      throw new NotImplementedError(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${a.shape} and  y shape = ${b.shape}`);
+    }
+  }
+  if (a.rank === 2 && b.rank === 2) {
+    const transposeA = false;
+    const transposeB = false;
+    return fused_ops_exports.matMul({
+      a,
+      b,
+      transposeA,
+      transposeB,
+      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,
+      activation: activation2
+    });
+  } else {
+    const aFirstDims = a.shape.slice();
+    const aLastDim = aFirstDims.pop();
+    a = reshape(a, [-1, aLastDim]);
+    const bShape = b.shape.slice();
+    const bLastDim = bShape.pop();
+    const ySecondLastDim = bShape.pop();
+    const yOtherDims = [...bShape, bLastDim];
+    const perm = Array.from({ length: b.rank }, (_, i) => {
+      if (i === 0) {
+        return b.rank - 2;
+      } else if (i <= b.rank - 2) {
+        return i - 1;
+      }
+      return i;
+    });
+    b = reshape(transpose(b, perm), [ySecondLastDim, -1]);
+    const outputShape = [...aFirstDims, ...yOtherDims];
+    const transposeA = false;
+    const transposeB = false;
+    return reshape(fused_ops_exports.matMul({
+      a,
+      b,
+      transposeA,
+      transposeB,
+      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,
+      activation: activation2
+    }), outputShape);
+  }
+}
+function gather2(reference, indices, axis) {
+  return tidy(() => {
+    if (Array.isArray(indices)) {
+      indices = tensor1d(indices, "int32");
+    } else {
+      indices = cast(indices, "int32");
+    }
+    return gather(reference, indices, axis);
+  });
+}
+function square2(x) {
+  return mul(x, x);
+}
+function reshapeBias(xRank, bias, dataFormat) {
+  const biasShape = bias.shape;
+  if (bias.rank !== 1 && bias.rank !== xRank) {
+    throw new ValueError(`Unexpected bias dimensions: ${bias.rank}; expected it to be 1 or ${xRank}`);
+  }
+  if (xRank === 5) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1, 1, 1]);
+      } else {
+        return reshape(bias, [1, biasShape[3], biasShape[0], biasShape[1], biasShape[2]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, 1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank === 4) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1, 1]);
+      } else {
+        return reshape(bias, [1, biasShape[2], biasShape[0], biasShape[1]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank === 3) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1]);
+      } else {
+        return reshape(bias, [1, biasShape[1], biasShape[0]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank < 3) {
+    return bias;
+  }
+  throw new ValueError(`Unsupported input rank by biasAdd: ${bias.rank}`);
+}
+function biasAdd(x, bias, dataFormat) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    return add2(x, reshapeBias(x.rank, bias, dataFormat));
+  });
+}
+function elu2(x, alpha = 1) {
+  if (alpha !== 1) {
+    throw new NotImplementedError(`Support for alpha values other than 1 (${alpha}) is not implemented yet.`);
+  }
+  return elu(x);
+}
+function softsign(x) {
+  return tidy(() => div(x, add2(abs(x), 1)));
+}
+function dropout2(x, level, noiseShape, seed) {
+  return tidy(() => dropout(x, level, noiseShape, seed));
+}
+function hardSigmoid(x) {
+  return tidy(() => {
+    const y = add2(0.5, mul(0.2, x));
+    return clipByValue(y, 0, 1);
+  });
+}
+function inTrainPhase(x, alt, training = false) {
+  return training ? x() : alt();
+}
+var VALID_FAN_MODE_VALUES = ["fanIn", "fanOut", "fanAvg"];
+var VALID_DISTRIBUTION_VALUES = ["normal", "uniform", "truncatedNormal"];
+function checkFanMode(value) {
+  checkStringTypeUnionValue(VALID_FAN_MODE_VALUES, "FanMode", value);
+}
+function checkDistribution(value) {
+  checkStringTypeUnionValue(VALID_DISTRIBUTION_VALUES, "Distribution", value);
+}
+var Initializer = class extends serialization_exports.Serializable {
+  fromConfigUsesCustomObjects() {
+    return false;
+  }
+  getConfig() {
+    return {};
+  }
+};
+var Zeros = class extends Initializer {
+  apply(shape, dtype) {
+    return zeros(shape, dtype);
+  }
+};
+Zeros.className = "Zeros";
+serialization_exports.registerClass(Zeros);
+var Ones = class extends Initializer {
+  apply(shape, dtype) {
+    return ones2(shape, dtype);
+  }
+};
+Ones.className = "Ones";
+serialization_exports.registerClass(Ones);
+var Constant = class extends Initializer {
+  constructor(args) {
+    super();
+    if (typeof args !== "object") {
+      throw new ValueError(`Expected argument of type ConstantConfig but got ${args}`);
+    }
+    if (args.value === void 0) {
+      throw new ValueError(`config must have value set but got ${args}`);
+    }
+    this.value = args.value;
+  }
+  apply(shape, dtype) {
+    return tidy(() => mul(scalar(this.value), ones2(shape, dtype)));
+  }
+  getConfig() {
+    return {
+      value: this.value
+    };
+  }
+};
+Constant.className = "Constant";
+serialization_exports.registerClass(Constant);
+var RandomUniform = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MINVAL = -0.05;
+    this.DEFAULT_MAXVAL = 0.05;
+    this.minval = args.minval || this.DEFAULT_MINVAL;
+    this.maxval = args.maxval || this.DEFAULT_MAXVAL;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    return randomUniform(shape, this.minval, this.maxval, dtype);
+  }
+  getConfig() {
+    return { minval: this.minval, maxval: this.maxval, seed: this.seed };
+  }
+};
+RandomUniform.className = "RandomUniform";
+serialization_exports.registerClass(RandomUniform);
+var RandomNormal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MEAN = 0;
+    this.DEFAULT_STDDEV = 0.05;
+    this.mean = args.mean || this.DEFAULT_MEAN;
+    this.stddev = args.stddev || this.DEFAULT_STDDEV;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    dtype = dtype || "float32";
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new NotImplementedError(`randomNormal does not support dType ${dtype}.`);
+    }
+    return randomNormal2(shape, this.mean, this.stddev, dtype, this.seed);
+  }
+  getConfig() {
+    return { mean: this.mean, stddev: this.stddev, seed: this.seed };
+  }
+};
+RandomNormal.className = "RandomNormal";
+serialization_exports.registerClass(RandomNormal);
+var TruncatedNormal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MEAN = 0;
+    this.DEFAULT_STDDEV = 0.05;
+    this.mean = args.mean || this.DEFAULT_MEAN;
+    this.stddev = args.stddev || this.DEFAULT_STDDEV;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    dtype = dtype || "float32";
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new NotImplementedError(`truncatedNormal does not support dType ${dtype}.`);
+    }
+    return truncatedNormal(shape, this.mean, this.stddev, dtype, this.seed);
+  }
+  getConfig() {
+    return { mean: this.mean, stddev: this.stddev, seed: this.seed };
+  }
+};
+TruncatedNormal.className = "TruncatedNormal";
+serialization_exports.registerClass(TruncatedNormal);
+var Identity2 = class extends Initializer {
+  constructor(args) {
+    super();
+    this.gain = args.gain != null ? args.gain : 1;
+  }
+  apply(shape, dtype) {
+    return tidy(() => {
+      if (shape.length !== 2 || shape[0] !== shape[1]) {
+        throw new ValueError("Identity matrix initializer can only be used for 2D square matrices.");
+      } else {
+        return mul(this.gain, eye(shape[0]));
+      }
+    });
+  }
+  getConfig() {
+    return { gain: this.gain };
+  }
+};
+Identity2.className = "Identity";
+serialization_exports.registerClass(Identity2);
+function computeFans(shape, dataFormat = "channelsLast") {
+  let fanIn;
+  let fanOut;
+  checkDataFormat(dataFormat);
+  if (shape.length === 2) {
+    fanIn = shape[0];
+    fanOut = shape[1];
+  } else if ([3, 4, 5].indexOf(shape.length) !== -1) {
+    if (dataFormat === "channelsFirst") {
+      const receptiveFieldSize = arrayProd(shape, 2);
+      fanIn = shape[1] * receptiveFieldSize;
+      fanOut = shape[0] * receptiveFieldSize;
+    } else if (dataFormat === "channelsLast") {
+      const receptiveFieldSize = arrayProd(shape, 0, shape.length - 2);
+      fanIn = shape[shape.length - 2] * receptiveFieldSize;
+      fanOut = shape[shape.length - 1] * receptiveFieldSize;
+    }
+  } else {
+    const shapeProd = arrayProd(shape);
+    fanIn = Math.sqrt(shapeProd);
+    fanOut = Math.sqrt(shapeProd);
+  }
+  return [fanIn, fanOut];
+}
+var VarianceScaling = class extends Initializer {
+  constructor(args) {
+    super();
+    if (args.scale < 0) {
+      throw new ValueError(`scale must be a positive float. Got: ${args.scale}`);
+    }
+    this.scale = args.scale == null ? 1 : args.scale;
+    this.mode = args.mode == null ? "fanIn" : args.mode;
+    checkFanMode(this.mode);
+    this.distribution = args.distribution == null ? "normal" : args.distribution;
+    checkDistribution(this.distribution);
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    const fans = computeFans(shape);
+    const fanIn = fans[0];
+    const fanOut = fans[1];
+    let scale22 = this.scale;
+    if (this.mode === "fanIn") {
+      scale22 /= Math.max(1, fanIn);
+    } else if (this.mode === "fanOut") {
+      scale22 /= Math.max(1, fanOut);
+    } else {
+      scale22 /= Math.max(1, (fanIn + fanOut) / 2);
+    }
+    if (this.distribution === "normal") {
+      const stddev = Math.sqrt(scale22);
+      dtype = dtype || "float32";
+      if (dtype !== "float32" && dtype !== "int32") {
+        throw new NotImplementedError(`${this.getClassName()} does not support dType ${dtype}.`);
+      }
+      return truncatedNormal(shape, 0, stddev, dtype, this.seed);
+    } else {
+      const limit = Math.sqrt(3 * scale22);
+      return randomUniform(shape, -limit, limit, dtype);
+    }
+  }
+  getConfig() {
+    return {
+      scale: this.scale,
+      mode: this.mode,
+      distribution: this.distribution,
+      seed: this.seed
+    };
+  }
+};
+VarianceScaling.className = "VarianceScaling";
+serialization_exports.registerClass(VarianceScaling);
+var GlorotUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanAvg",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+GlorotUniform.className = "GlorotUniform";
+serialization_exports.registerClass(GlorotUniform);
+var GlorotNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanAvg",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+GlorotNormal.className = "GlorotNormal";
+serialization_exports.registerClass(GlorotNormal);
+var HeNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 2,
+      mode: "fanIn",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+HeNormal.className = "HeNormal";
+serialization_exports.registerClass(HeNormal);
+var HeUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 2,
+      mode: "fanIn",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+HeUniform.className = "HeUniform";
+serialization_exports.registerClass(HeUniform);
+var LeCunNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanIn",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+LeCunNormal.className = "LeCunNormal";
+serialization_exports.registerClass(LeCunNormal);
+var LeCunUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanIn",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+LeCunUniform.className = "LeCunNormal";
+serialization_exports.registerClass(LeCunUniform);
+var Orthogonal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_GAIN = 1;
+    this.gain = args.gain == null ? this.DEFAULT_GAIN : args.gain;
+    this.seed = args.seed;
+    if (this.seed != null) {
+      throw new NotImplementedError("Random seed is not implemented for Orthogonal Initializer yet.");
+    }
+  }
+  apply(shape, dtype) {
+    return tidy(() => {
+      if (shape.length < 2) {
+        throw new NotImplementedError("Shape must be at least 2D.");
+      }
+      if (shape[0] * shape[1] > 2e3) {
+        console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${shape[0] * shape[1]}) elements: Slowness may result.`);
+      }
+      const normalizedShape = shape[0] > shape[1] ? [shape[1], shape[0]] : shape;
+      const a = randomNormal2(normalizedShape, 0, 1, "float32");
+      let q = linalg.gramSchmidt(a);
+      if (shape[0] > shape[1]) {
+        q = transpose(q);
+      }
+      return mul(this.gain, q);
+    });
+  }
+  getConfig() {
+    return {
+      gain: this.gain,
+      seed: this.seed
+    };
+  }
+};
+Orthogonal.className = "Orthogonal";
+serialization_exports.registerClass(Orthogonal);
+var INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "constant": "Constant",
+  "glorotNormal": "GlorotNormal",
+  "glorotUniform": "GlorotUniform",
+  "heNormal": "HeNormal",
+  "heUniform": "HeUniform",
+  "identity": "Identity",
+  "leCunNormal": "LeCunNormal",
+  "leCunUniform": "LeCunUniform",
+  "ones": "Ones",
+  "orthogonal": "Orthogonal",
+  "randomNormal": "RandomNormal",
+  "randomUniform": "RandomUniform",
+  "truncatedNormal": "TruncatedNormal",
+  "varianceScaling": "VarianceScaling",
+  "zeros": "Zeros"
+};
+function deserializeInitializer(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "initializer");
+}
+function serializeInitializer(initializer) {
+  return serializeKerasObject(initializer);
+}
+function getInitializer(identifier) {
+  if (typeof identifier === "string") {
+    const className = identifier in INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    if (className === "GlorotNormal") {
+      return new GlorotNormal();
+    } else if (className === "GlorotUniform") {
+      return new GlorotUniform();
+    } else if (className === "HeNormal") {
+      return new HeNormal();
+    } else if (className === "HeUniform") {
+      return new HeUniform();
+    } else if (className === "LeCunNormal") {
+      return new LeCunNormal();
+    } else if (className === "LeCunUniform") {
+      return new LeCunUniform();
+    } else {
+      const config = {};
+      config["className"] = className;
+      config["config"] = {};
+      return deserializeInitializer(config);
+    }
+  } else if (identifier instanceof Initializer) {
+    return identifier;
+  } else {
+    return deserializeInitializer(identifier);
+  }
+}
+function isArrayOfShapes(x) {
+  return Array.isArray(x) && Array.isArray(x[0]);
+}
+function normalizeShapeList(x) {
+  if (x.length === 0) {
+    return [];
+  }
+  if (!Array.isArray(x[0])) {
+    return [x];
+  }
+  return x;
+}
+function getExactlyOneTensor(xs) {
+  let x;
+  if (Array.isArray(xs)) {
+    if (xs.length !== 1) {
+      throw new ValueError(`Expected Tensor length to be 1; got ${xs.length}`);
+    }
+    x = xs[0];
+  } else {
+    x = xs;
+  }
+  return x;
+}
+function getExactlyOneShape(shapes) {
+  if (Array.isArray(shapes) && Array.isArray(shapes[0])) {
+    if (shapes.length === 1) {
+      shapes = shapes;
+      return shapes[0];
+    } else {
+      throw new ValueError(`Expected exactly 1 Shape; got ${shapes.length}`);
+    }
+  } else {
+    return shapes;
+  }
+}
+function countParamsInWeights(weights) {
+  let count2 = 0;
+  for (const weight of weights) {
+    if (weight.shape.length === 0) {
+      count2 += 1;
+    } else {
+      count2 += weight.shape.reduce((a, b) => a * b);
+    }
+  }
+  return count2;
+}
+var DEFAULT_VARIABLE_NAME_PREFIX = "Variable";
+var LayerVariable = class {
+  constructor(val, dtype = "float32", name = DEFAULT_VARIABLE_NAME_PREFIX, trainable = true, constraint = null) {
+    this.dtype = dtype == null ? "float32" : dtype;
+    this.shape = val.shape;
+    this.id = getNextUniqueTensorId();
+    name = name == null ? DEFAULT_VARIABLE_NAME_PREFIX : name;
+    this.originalName = getScopedTensorName(name);
+    this.name = getUniqueTensorName(this.originalName);
+    this.trainable_ = trainable;
+    this.constraint = constraint;
+    this.val = variable(val, this.trainable_, this.name, this.dtype);
+  }
+  read() {
+    this.assertNotDisposed();
+    return this.val;
+  }
+  write(newVal) {
+    this.assertNotDisposed();
+    checkShapesMatch(this.val, newVal);
+    if (this.val.id !== newVal.id) {
+      this.val.assign(newVal);
+      if (this.constraint != null) {
+        this.val.assign(this.constraint.apply(this.val));
+      }
+    }
+    return this;
+  }
+  dispose() {
+    this.assertNotDisposed();
+    this.val.dispose();
+  }
+  assertNotDisposed() {
+    if (this.val.isDisposed) {
+      throw new Error(`LayersVariable ${this.name} is already disposed.`);
+    }
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this.trainable_ = trainable;
+    this.val.trainable = trainable;
+  }
+};
+function checkShapesMatch(x, y) {
+  if (x.shape.toString() !== y.shape.toString()) {
+    throw new Error("Shape mismatch: " + JSON.stringify(x.shape) + " vs. " + JSON.stringify(y.shape));
+  }
+}
+function batchGetValue(xs) {
+  return xs.map((x) => x.read());
+}
+function batchSetValue(variablesAndValues) {
+  variablesAndValues.forEach((variableAndValue) => {
+    const variable2 = variableAndValue[0];
+    variable2.write(variableAndValue[1]);
+  });
+}
+var InputSpec = class {
+  constructor(args) {
+    this.dtype = args.dtype;
+    this.shape = args.shape;
+    if (args.shape != null) {
+      this.ndim = args.shape.length;
+    } else {
+      this.ndim = args.ndim;
+    }
+    this.maxNDim = args.maxNDim;
+    this.minNDim = args.minNDim;
+    this.axes = args.axes || {};
+  }
+};
+var SymbolicTensor = class {
+  constructor(dtype, shape, sourceLayer, inputs, callArgs, name, outputTensorIndex) {
+    this.dtype = dtype;
+    this.shape = shape;
+    this.sourceLayer = sourceLayer;
+    this.inputs = inputs;
+    this.callArgs = callArgs;
+    this.outputTensorIndex = outputTensorIndex;
+    this.id = getNextUniqueTensorId();
+    if (name != null) {
+      this.originalName = getScopedTensorName(name);
+      this.name = getUniqueTensorName(this.originalName);
+    }
+    this.rank = shape.length;
+  }
+};
+var _nextNodeID = 0;
+var Node = class {
+  constructor(args, callArgs) {
+    this.callArgs = callArgs;
+    this.id = _nextNodeID++;
+    this.outboundLayer = args.outboundLayer;
+    this.inboundLayers = args.inboundLayers;
+    this.nodeIndices = args.nodeIndices;
+    this.tensorIndices = args.tensorIndices;
+    this.inputTensors = args.inputTensors;
+    this.outputTensors = args.outputTensors;
+    this.inputMasks = args.inputMasks;
+    this.outputMasks = args.outputMasks;
+    this.inputShapes = args.inputShapes;
+    this.outputShapes = args.outputShapes;
+    for (const layer of args.inboundLayers) {
+      if (layer != null) {
+        layer.outboundNodes.push(this);
+      }
+    }
+    args.outboundLayer.inboundNodes.push(this);
+  }
+  getConfig() {
+    const inboundNames = [];
+    for (const layer of this.inboundLayers) {
+      if (layer != null) {
+        inboundNames.push(layer.name);
+      } else {
+        inboundNames.push(null);
+      }
+    }
+    return {
+      outboundLayer: this.outboundLayer ? this.outboundLayer.name : null,
+      inboundLayers: inboundNames,
+      nodeIndices: this.nodeIndices,
+      tensorIndices: this.tensorIndices
+    };
+  }
+};
+var _nextLayerID = 0;
+var Layer = class extends serialization_exports.Serializable {
+  constructor(args = {}) {
+    super();
+    this._callHook = null;
+    this._addedWeightNames = [];
+    this._stateful = false;
+    this.id = _nextLayerID++;
+    this.activityRegularizer = null;
+    this.inputSpec = null;
+    this.supportsMasking = false;
+    this._trainableWeights = [];
+    this._nonTrainableWeights = [];
+    this._losses = [];
+    this._updates = [];
+    this._built = false;
+    this.inboundNodes = [];
+    this.outboundNodes = [];
+    let name = args.name;
+    if (!name) {
+      const prefix = this.getClassName();
+      name = toSnakeCase(prefix) + "_" + getUid(prefix);
+    }
+    this.name = name;
+    this.trainable_ = args.trainable == null ? true : args.trainable;
+    if (args.inputShape != null || args.batchInputShape != null) {
+      let batchInputShape;
+      if (args.batchInputShape != null) {
+        batchInputShape = args.batchInputShape;
+      } else if (args.inputShape != null) {
+        let batchSize = null;
+        if (args.batchSize != null) {
+          batchSize = args.batchSize;
+        }
+        batchInputShape = [batchSize].concat(args.inputShape);
+      }
+      this.batchInputShape = batchInputShape;
+      let dtype = args.dtype;
+      if (dtype == null) {
+        dtype = args.inputDType;
+      }
+      if (dtype == null) {
+        dtype = "float32";
+      }
+      this.dtype = dtype;
+    }
+    if (args.weights != null) {
+      this.initialWeights = args.weights;
+    } else {
+      this.initialWeights = null;
+    }
+    this._refCount = null;
+    this.fastWeightInitDuringBuild = false;
+  }
+  static nodeKey(layer, nodeIndex) {
+    return layer.name + "_ib-" + nodeIndex.toString();
+  }
+  getNodeAtIndex(nodeIndex, attrName) {
+    if (this.inboundNodes.length === 0) {
+      throw new RuntimeError(`The layer has never been called and thus has no defined ${attrName}.`);
+    }
+    if (this.inboundNodes.length <= nodeIndex) {
+      throw new ValueError(`Asked to get ${attrName} at node ${nodeIndex}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);
+    }
+    return this.inboundNodes[nodeIndex];
+  }
+  getInputAt(nodeIndex) {
+    return singletonOrArray(this.getNodeAtIndex(nodeIndex, "input").inputTensors);
+  }
+  getOutputAt(nodeIndex) {
+    return singletonOrArray(this.getNodeAtIndex(nodeIndex, "output").outputTensors);
+  }
+  get input() {
+    if (this.inboundNodes.length > 1) {
+      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);
+    } else if (this.inboundNodes.length === 0) {
+      throw new AttributeError(`Layer ${this.name} is not connected, no input to return.`);
+    }
+    return singletonOrArray(this.getNodeAtIndex(0, "input").inputTensors);
+  }
+  get output() {
+    if (this.inboundNodes.length === 0) {
+      throw new AttributeError(`Layer ${this.name} has no inbound nodes.`);
+    }
+    if (this.inboundNodes.length > 1) {
+      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);
+    }
+    return singletonOrArray(this.getNodeAtIndex(0, "output").outputTensors);
+  }
+  get losses() {
+    return this._losses;
+  }
+  calculateLosses() {
+    return this.losses.map((lossFn) => lossFn());
+  }
+  get updates() {
+    return this._updates;
+  }
+  get built() {
+    return this._built;
+  }
+  set built(built) {
+    this._built = built;
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this._trainableWeights.forEach((w) => w.trainable = trainable);
+    this.trainable_ = trainable;
+  }
+  get trainableWeights() {
+    if (this.trainable_) {
+      return this._trainableWeights.filter((w) => w.trainable);
+    } else {
+      return [];
+    }
+  }
+  set trainableWeights(weights) {
+    this._trainableWeights = weights;
+  }
+  get nonTrainableWeights() {
+    if (this.trainable) {
+      return this._trainableWeights.filter((w) => !w.trainable).concat(this._nonTrainableWeights);
+    } else {
+      return this._trainableWeights.concat(this._nonTrainableWeights);
+    }
+  }
+  set nonTrainableWeights(weights) {
+    this._nonTrainableWeights = weights;
+  }
+  get weights() {
+    return this.trainableWeights.concat(this.nonTrainableWeights);
+  }
+  get stateful() {
+    return this._stateful;
+  }
+  resetStates() {
+    if (!this.stateful) {
+      throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.");
+    }
+  }
+  assertInputCompatibility(inputs) {
+    inputs = toList(inputs);
+    if (this.inputSpec == null || this.inputSpec.length === 0) {
+      return;
+    }
+    const inputSpec = toList(this.inputSpec);
+    if (inputs.length !== inputSpec.length) {
+      throw new ValueError(`Layer ${this.name} expects ${inputSpec.length} inputs, but it received ${inputs.length} input tensors. Input received: ${inputs}`);
+    }
+    for (let inputIndex = 0; inputIndex < inputs.length; inputIndex++) {
+      const x = inputs[inputIndex];
+      const spec = inputSpec[inputIndex];
+      if (spec == null) {
+        continue;
+      }
+      const ndim = x.rank;
+      if (spec.ndim != null) {
+        if (ndim !== spec.ndim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected ndim=${spec.ndim}, found ndim=${ndim}`);
+        }
+      }
+      if (spec.maxNDim != null) {
+        if (ndim > spec.maxNDim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected max_ndim=${spec.maxNDim}, found ndim=${ndim}`);
+        }
+      }
+      if (spec.minNDim != null) {
+        if (ndim < spec.minNDim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected min_ndim=${spec.minNDim}, found ndim=${ndim}.`);
+        }
+      }
+      if (spec.dtype != null) {
+        if (x.dtype !== spec.dtype) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name} : expected dtype=${spec.dtype}, found dtype=${x.dtype}.`);
+        }
+      }
+      if (spec.axes) {
+        const xShape = x.shape;
+        for (const key in spec.axes) {
+          const axis = Number(key);
+          const value = spec.axes[key];
+          const xShapeAtAxis = axis >= 0 ? xShape[axis] : xShape[xShape.length + axis];
+          if (value != null && [value, null].indexOf(xShapeAtAxis) === -1) {
+            throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected axis ${axis} of input shape to have value ${value} but got shape ${xShape}.`);
+          }
+        }
+      }
+      if (spec.shape != null) {
+        for (let i = 0; i < spec.shape.length; ++i) {
+          const specDim = spec.shape[i];
+          const dim = x.shape[i];
+          if (specDim != null && dim != null) {
+            if (specDim !== dim) {
+              throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected shape=${spec.shape}, found shape=${x.shape}.`);
+            }
+          }
+        }
+      }
+    }
+  }
+  call(inputs, kwargs) {
+    return inputs;
+  }
+  invokeCallHook(inputs, kwargs) {
+    if (this._callHook != null) {
+      this._callHook(inputs, kwargs);
+    }
+  }
+  setCallHook(callHook) {
+    this._callHook = callHook;
+  }
+  clearCallHook() {
+    this._callHook = null;
+  }
+  apply(inputs, kwargs) {
+    kwargs = kwargs || {};
+    this.assertNotDisposed();
+    const inputsList = toList(inputs);
+    let allAreSymbolic = true;
+    for (const input2 of inputsList) {
+      if (!(input2 instanceof SymbolicTensor)) {
+        allAreSymbolic = false;
+        break;
+      }
+    }
+    let noneAreSymbolic = true;
+    for (const input2 of inputsList) {
+      if (input2 instanceof SymbolicTensor) {
+        noneAreSymbolic = false;
+        break;
+      }
+    }
+    if (allAreSymbolic === noneAreSymbolic) {
+      throw new ValueError("Arguments to apply() must be all SymbolicTensors or all Tensors");
+    }
+    return nameScope(this.name, () => {
+      if (!this.built) {
+        this.assertInputCompatibility(inputs);
+        const inputShapes = [];
+        for (const xElem of toList(inputs)) {
+          inputShapes.push(xElem.shape);
+        }
+        this.build(singletonOrArray(inputShapes));
+        this.built = true;
+        if (this.initialWeights) {
+          this.setWeights(this.initialWeights);
+        }
+        if (this._refCount === null && noneAreSymbolic) {
+          this._refCount = 1;
+        }
+      }
+      this.assertInputCompatibility(inputs);
+      if (noneAreSymbolic) {
+        let output = this.call(inputs, kwargs);
+        const outputList = toList(output);
+        const outputListCopy = [];
+        for (let x of outputList) {
+          if (inputsList.indexOf(x) !== -1) {
+            x = x.clone();
+          }
+          outputListCopy.push(x);
+        }
+        output = singletonOrArray(outputListCopy);
+        if (this.activityRegularizer != null) {
+          throw new NotImplementedError("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+        }
+        return output;
+      } else {
+        const inputShape = collectInputShape(inputs);
+        const outputShape = this.computeOutputShape(inputShape);
+        let output;
+        const outputDType = guessOutputDType(inputs);
+        this.warnOnIncompatibleInputShape(Array.isArray(inputs) ? inputShape[0] : inputShape);
+        if (outputShape != null && outputShape.length > 0 && Array.isArray(outputShape[0])) {
+          output = outputShape.map((shape, index) => new SymbolicTensor(outputDType, shape, this, toList(inputs), kwargs, this.name, index));
+        } else {
+          output = new SymbolicTensor(outputDType, outputShape, this, toList(inputs), kwargs, this.name);
+        }
+        this.addInboundNode(inputs, output, null, null, inputShape, outputShape, kwargs);
+        this._refCount++;
+        if (this.activityRegularizer != null) {
+          throw new NotImplementedError("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+        }
+        return output;
+      }
+    });
+  }
+  warnOnIncompatibleInputShape(inputShape) {
+    if (this.batchInputShape == null) {
+      return;
+    } else if (inputShape.length !== this.batchInputShape.length) {
+      console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(inputShape)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);
+    } else {
+      let dimMismatch = false;
+      this.batchInputShape.forEach((dimension, i) => {
+        if (dimension != null && inputShape[i] != null && inputShape[i] !== dimension) {
+          dimMismatch = true;
+        }
+      });
+      if (dimMismatch) {
+        console.warn(`The shape of the input tensor (${JSON.stringify(inputShape)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`);
+      }
+    }
+  }
+  get outputShape() {
+    if (this.inboundNodes == null || this.inboundNodes.length === 0) {
+      throw new AttributeError(`The layer ${this.name} has never been called and thus has no defined output shape.`);
+    }
+    const allOutputShapes = [];
+    for (const node of this.inboundNodes) {
+      const shapeString = JSON.stringify(node.outputShapes);
+      if (allOutputShapes.indexOf(shapeString) === -1) {
+        allOutputShapes.push(shapeString);
+      }
+    }
+    if (allOutputShapes.length === 1) {
+      const outputShapes = this.inboundNodes[0].outputShapes;
+      if (Array.isArray(outputShapes) && Array.isArray(outputShapes[0]) && outputShapes.length === 1) {
+        return outputShapes[0];
+      } else {
+        return outputShapes;
+      }
+    } else {
+      throw new AttributeError(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`);
+    }
+  }
+  countParams() {
+    if (!this.built) {
+      throw new RuntimeError(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);
+    }
+    return countParamsInWeights(this.weights);
+  }
+  build(inputShape) {
+    this.built = true;
+  }
+  getWeights(trainableOnly = false) {
+    return batchGetValue(trainableOnly ? this.trainableWeights : this.weights);
+  }
+  setWeights(weights) {
+    tidy(() => {
+      const params = this.weights;
+      if (params.length !== weights.length) {
+        throw new ValueError(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${weights.length}, but the layer was expecting ${params.length} weights. Provided weights: ${weights}...`);
+      }
+      if (params.length === 0) {
+        return;
+      }
+      const weightValueTuples = [];
+      const paramValues = batchGetValue(params);
+      for (let i = 0; i < paramValues.length; ++i) {
+        const pv = paramValues[i];
+        const p2 = params[i];
+        const w = weights[i];
+        if (!util_exports.arraysEqual(pv.shape, w.shape)) {
+          throw new ValueError(`Layer weight shape ${pv.shape} not compatible with provided weight shape ${w.shape}`);
+        }
+        weightValueTuples.push([p2, w]);
+      }
+      batchSetValue(weightValueTuples);
+    });
+  }
+  addWeight(name, shape, dtype, initializer, regularizer, trainable, constraint, getInitializerFunc) {
+    if (this._addedWeightNames.indexOf(name) !== -1) {
+      throw new ValueError(`Duplicate weight name ${name} for layer ${this.name}`);
+    }
+    this._addedWeightNames.push(name);
+    if (dtype == null) {
+      dtype = "float32";
+    }
+    if (this.fastWeightInitDuringBuild) {
+      initializer = getInitializerFunc != null ? getInitializerFunc() : getInitializer("zeros");
+    }
+    const initValue = initializer.apply(shape, dtype);
+    const weight = new LayerVariable(initValue, dtype, name, trainable, constraint);
+    initValue.dispose();
+    if (regularizer != null) {
+      this.addLoss(() => regularizer.apply(weight.read()));
+    }
+    if (trainable == null) {
+      trainable = true;
+    }
+    if (trainable) {
+      this._trainableWeights.push(weight);
+    } else {
+      this._nonTrainableWeights.push(weight);
+    }
+    return weight;
+  }
+  setFastWeightInitDuringBuild(value) {
+    this.fastWeightInitDuringBuild = value;
+  }
+  addLoss(losses2) {
+    if (losses2 == null || Array.isArray(losses2) && losses2.length === 0) {
+      return;
+    }
+    losses2 = toList(losses2);
+    if (this._losses !== void 0 && this._losses !== null) {
+      this.losses.push(...losses2);
+    }
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  computeMask(inputs, mask) {
+    if (!this.supportsMasking) {
+      if (mask != null) {
+        if (Array.isArray(mask)) {
+          mask.forEach((maskElement) => {
+            if (maskElement != null) {
+              throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);
+            }
+          });
+        } else {
+          throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);
+        }
+      }
+      return null;
+    }
+    return mask;
+  }
+  addInboundNode(inputTensors, outputTensors, inputMasks, outputMasks, inputShapes, outputShapes, kwargs = null) {
+    const inputTensorList = toList(inputTensors);
+    outputTensors = toList(outputTensors);
+    inputMasks = toList(inputMasks);
+    outputMasks = toList(outputMasks);
+    inputShapes = normalizeShapeList(inputShapes);
+    outputShapes = normalizeShapeList(outputShapes);
+    const inboundLayers = [];
+    const nodeIndices = [];
+    const tensorIndices = [];
+    for (const x of inputTensorList) {
+      inboundLayers.push(x.sourceLayer);
+      nodeIndices.push(x.nodeIndex);
+      tensorIndices.push(x.tensorIndex);
+    }
+    new Node({
+      outboundLayer: this,
+      inboundLayers,
+      nodeIndices,
+      tensorIndices,
+      inputTensors: inputTensorList,
+      outputTensors,
+      inputMasks,
+      outputMasks,
+      inputShapes,
+      outputShapes
+    }, kwargs);
+    for (let i = 0; i < outputTensors.length; i++) {
+      outputTensors[i].sourceLayer = this;
+      outputTensors[i].nodeIndex = this.inboundNodes.length - 1;
+      outputTensors[i].tensorIndex = i;
+    }
+  }
+  getConfig() {
+    const config = { name: this.name, trainable: this.trainable };
+    if (this.batchInputShape != null) {
+      config["batchInputShape"] = this.batchInputShape;
+    }
+    if (this.dtype != null) {
+      config["dtype"] = this.dtype;
+    }
+    return config;
+  }
+  disposeWeights() {
+    this.weights.forEach((weight) => weight.dispose());
+    return this.weights.length;
+  }
+  assertNotDisposed() {
+    if (this._refCount === 0) {
+      throw new Error(`Layer '${this.name}' is already disposed.`);
+    }
+  }
+  dispose() {
+    if (!this.built) {
+      throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);
+    }
+    if (this._refCount === null) {
+      throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);
+    }
+    this.assertNotDisposed();
+    let numDisposedVariables = 0;
+    if (--this._refCount === 0) {
+      numDisposedVariables = this.disposeWeights();
+    }
+    return { refCountAfterDispose: this._refCount, numDisposedVariables };
+  }
+};
+function collectInputShape(inputTensors) {
+  inputTensors = toList(inputTensors);
+  const shapes = [];
+  for (const x of inputTensors) {
+    shapes.push(x.shape);
+  }
+  return singletonOrArray(shapes);
+}
+function guessOutputDType(inputTensors) {
+  return "float32";
+}
+function getSourceInputs(tensor2, layer, nodeIndex) {
+  if (layer == null || nodeIndex != null && nodeIndex > 0) {
+    layer = tensor2.sourceLayer;
+    nodeIndex = tensor2.nodeIndex;
+  }
+  if (layer.inboundNodes.length === 0) {
+    return [tensor2];
+  } else {
+    const node = layer.inboundNodes[nodeIndex];
+    if (node.inboundLayers.length === 0) {
+      return node.inputTensors;
+    } else {
+      const sourceTensors = [];
+      for (let i = 0; i < node.inboundLayers.length; i++) {
+        const x = node.inputTensors[i];
+        const layer2 = node.inboundLayers[i];
+        const nodeIndex2 = node.nodeIndices[i];
+        const previousSources = getSourceInputs(x, layer2, nodeIndex2);
+        for (const x2 of previousSources) {
+          if (sourceTensors.indexOf(x2) === -1) {
+            sourceTensors.push(x2);
+          }
+        }
+      }
+      return sourceTensors;
+    }
+  }
+}
+var InputLayer = class extends Layer {
+  constructor(args) {
+    super({
+      dtype: args.dtype,
+      name: args.name != null ? args.name : getUid("input").toString()
+    });
+    if (args.batchSize == null) {
+      args.batchSize = null;
+    }
+    if (args.sparse == null) {
+      args.sparse = false;
+    }
+    this.trainable = false;
+    this.built = true;
+    this.sparse = args.sparse;
+    if (args.inputShape != null && args.batchInputShape != null) {
+      throw new ValueError("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");
+    }
+    let batchInputShape = args.batchInputShape;
+    if (batchInputShape == null) {
+      if (args.inputShape == null) {
+        throw new ValueError("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");
+      } else {
+        batchInputShape = [args.batchSize].concat(args.inputShape);
+      }
+    } else {
+      if (args.batchSize != null) {
+        throw new ValueError("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");
+      }
+    }
+    const dtype = args.dtype || "float32";
+    this.batchInputShape = batchInputShape;
+    this.dtype = dtype;
+    this.inputSpec = [{ shape: batchInputShape }];
+    const inputTensor = new SymbolicTensor(this.dtype, this.batchInputShape, this, [], {}, this.name);
+    inputTensor.nodeIndex = 0;
+    inputTensor.tensorIndex = 0;
+    new Node({
+      outboundLayer: this,
+      inboundLayers: [],
+      nodeIndices: [],
+      tensorIndices: [],
+      inputTensors: [inputTensor],
+      outputTensors: [inputTensor],
+      inputMasks: [null],
+      outputMasks: [null],
+      inputShapes: [batchInputShape],
+      outputShapes: [batchInputShape]
+    });
+  }
+  apply(inputs, kwargs) {
+    throw new ValueError(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`);
+  }
+  dispose() {
+    return { refCountAfterDispose: this._refCount, numDisposedVariables: 0 };
+  }
+  getConfig() {
+    return {
+      batchInputShape: this.batchInputShape,
+      dtype: this.dtype,
+      sparse: this.sparse,
+      name: this.name
+    };
+  }
+};
+InputLayer.className = "InputLayer";
+serialization_exports.registerClass(InputLayer);
+function Input(config) {
+  if (config.batchShape == null && config.shape == null) {
+    throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");
+  }
+  if (config.batchShape != null && config.shape != null) {
+    throw new ValueError("Please provide either a `shape` or `batchShape` argument to Input, but not both.");
+  }
+  let batchShape = config.batchShape;
+  if (config.shape != null && batchShape == null) {
+    batchShape = [null].concat(config.shape);
+  }
+  let dtype = config.dtype;
+  if (dtype == null) {
+    dtype = "float32";
+  }
+  const inputLayer2 = new InputLayer({
+    batchInputShape: batchShape,
+    name: config.name,
+    dtype,
+    sparse: config.sparse
+  });
+  const outputs = inputLayer2.inboundNodes[0].outputTensors;
+  return outputs[0];
+}
+function assertFeedCompatibility(key, val) {
+  if (key.dtype == null || key.dtype === val.dtype) {
+    return val;
+  }
+  try {
+    return cast(val, key.dtype);
+  } catch (err) {
+    throw new ValueError(`The dtype of the feed (${val.dtype}) can not be cast to the dtype of the key '${key.name}' (${key.dtype}).`);
+  }
+}
+var FeedDict = class {
+  constructor(feeds) {
+    this.id2Value = {};
+    this.id2Mask = {};
+    this.name2Id = {};
+    if (feeds instanceof FeedDict) {
+      for (const id in feeds.id2Value) {
+        this.id2Value[id] = feeds.id2Value[id];
+        if (id in feeds.id2Mask) {
+          this.id2Mask[id] = feeds.id2Mask[id];
+        }
+      }
+    } else {
+      if (feeds == null) {
+        return;
+      }
+      for (const feed of feeds) {
+        this.add(feed.key, feed.value);
+      }
+    }
+  }
+  add(key, value, mask) {
+    if (this.id2Value[key.id] == null) {
+      this.id2Value[key.id] = assertFeedCompatibility(key, value);
+      this.name2Id[key.name] = key.id;
+      if (mask != null) {
+        this.id2Mask[key.id] = mask;
+      }
+    } else {
+      throw new ValueError(`Duplicate key: name=${key.name}, id=${key.id}`);
+    }
+    return this;
+  }
+  addFeed(feed) {
+    this.add(feed.key, feed.value);
+  }
+  hasKey(key) {
+    return this.id2Value[key.id] != null;
+  }
+  names() {
+    return Object.keys(this.name2Id);
+  }
+  getValue(key) {
+    if (key instanceof SymbolicTensor) {
+      if (this.id2Value[key.id] == null) {
+        throw new ValueError(`Nonexistent key: ${key.name}`);
+      } else {
+        return this.id2Value[key.id];
+      }
+    } else {
+      const id = this.name2Id[key];
+      if (id == null) {
+        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);
+      }
+      return this.id2Value[id];
+    }
+  }
+  getMask(key) {
+    if (key instanceof SymbolicTensor) {
+      if (this.id2Value[key.id] == null) {
+        throw new ValueError(`Nonexistent key: ${key.name}`);
+      } else {
+        return this.id2Mask[key.id];
+      }
+    } else {
+      const id = this.name2Id[key];
+      if (id == null) {
+        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);
+      }
+      return this.id2Mask[id];
+    }
+  }
+  disposeMasks() {
+    if (this.id2Mask != null) {
+      dispose(this.id2Mask);
+    }
+  }
+};
+var cachedSorted = new LruCache();
+var cachedRecipientCounts = new LruCache();
+function updateCacheMaxEntries(maxEntries) {
+  if (cachedSorted != null) {
+    cachedSorted.setMaxEntries(maxEntries);
+  }
+  if (cachedRecipientCounts != null) {
+    cachedRecipientCounts.setMaxEntries(maxEntries);
+  }
+}
+function execute(fetches, feedDict, kwargs, probe) {
+  const training = kwargs == null ? false : kwargs["training"];
+  const arrayFetches = Array.isArray(fetches);
+  const fetchArray = arrayFetches ? fetches : [fetches];
+  const outputNames = fetchArray.map((t) => t.name);
+  const finalOutputs = [];
+  const feedNames = feedDict.names();
+  for (const outputName of outputNames) {
+    if (feedNames.indexOf(outputName) !== -1) {
+      finalOutputs.push(feedDict.getValue(outputName));
+    } else {
+      finalOutputs.push(null);
+    }
+  }
+  if (probe != null) {
+    probe.maxNumTensors = -Infinity;
+    probe.minNumTensors = Infinity;
+  }
+  const fetchAndFeedKey = outputNames.join(",") + "|" + feedDict.names().sort().join(",");
+  let sorted = cachedSorted.get(fetchAndFeedKey);
+  let recipientCounts;
+  if (sorted == null) {
+    const out = getTopologicalSortAndRecipientCounts(fetchArray, feedDict);
+    sorted = out.sorted;
+    recipientCounts = out.recipientCounts;
+    cachedSorted.put(fetchAndFeedKey, sorted);
+    cachedRecipientCounts.put(fetchAndFeedKey, recipientCounts);
+  }
+  recipientCounts = {};
+  if (!training) {
+    Object.assign(recipientCounts, cachedRecipientCounts.get(fetchAndFeedKey));
+  }
+  const internalFeedDict = new FeedDict(feedDict);
+  for (let i = 0; i < sorted.length; ++i) {
+    if (probe != null) {
+      const numTensors = memory().numTensors;
+      if (numTensors > probe.maxNumTensors) {
+        probe.maxNumTensors = numTensors;
+      }
+      if (numTensors < probe.minNumTensors) {
+        probe.minNumTensors = numTensors;
+      }
+    }
+    const symbolic = sorted[i];
+    const srcLayer = symbolic.sourceLayer;
+    if (srcLayer instanceof InputLayer) {
+      continue;
+    }
+    const inputValues = [];
+    const inputMasks = [];
+    const tensorsToDispose = [];
+    let maskExists = false;
+    for (const input2 of symbolic.inputs) {
+      const value = internalFeedDict.getValue(input2);
+      const mask = internalFeedDict.getMask(input2);
+      inputValues.push(value);
+      inputMasks.push(mask);
+      if (mask != null) {
+        maskExists = true;
+      }
+      if (!training) {
+        recipientCounts[input2.name]--;
+        if (recipientCounts[input2.name] === 0 && !feedDict.hasKey(input2) && outputNames.indexOf(input2.name) === -1 && !value.isDisposed && input2.sourceLayer.stateful !== true) {
+          tensorsToDispose.push(value);
+        }
+      }
+    }
+    if (maskExists) {
+      kwargs = kwargs || {};
+      kwargs["mask"] = inputMasks[0];
+    }
+    const outputTensors = toList(srcLayer.apply(inputValues, kwargs));
+    let outputMask = null;
+    if (srcLayer.supportsMasking) {
+      outputMask = srcLayer.computeMask(inputValues, inputMasks);
+    }
+    const layerOutputs = getNodeOutputs(symbolic);
+    const outputSymbolicTensors = Array.isArray(layerOutputs) ? layerOutputs : [layerOutputs];
+    for (let i2 = 0; i2 < outputSymbolicTensors.length; ++i2) {
+      if (!internalFeedDict.hasKey(outputSymbolicTensors[i2])) {
+        internalFeedDict.add(outputSymbolicTensors[i2], outputTensors[i2], Array.isArray(outputMask) ? outputMask[0] : outputMask);
+      }
+      const index = outputNames.indexOf(outputSymbolicTensors[i2].name);
+      if (index !== -1) {
+        finalOutputs[index] = outputTensors[i2];
+      }
+    }
+    if (!training) {
+      dispose(tensorsToDispose);
+    }
+  }
+  internalFeedDict.disposeMasks();
+  return arrayFetches ? finalOutputs : finalOutputs[0];
+}
+function getTopologicalSortAndRecipientCounts(fetches, feedDict) {
+  util_exports.assert(fetches != null && fetches.length > 0, () => `Expected at least one fetch, got none`);
+  let finalSorted = [];
+  let finalRecipientMap = {};
+  if (fetches.length === 1) {
+    const out = getTopologicalSortAndRecipientCountsForOneFetch(fetches[0], feedDict);
+    finalSorted = out.sorted;
+    finalRecipientMap = out.recipientMap;
+  } else {
+    const visited = /* @__PURE__ */ new Set();
+    for (const fetch4 of fetches) {
+      const { sorted, recipientMap } = getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict);
+      for (const symbolicTensor of sorted) {
+        if (!visited.has(symbolicTensor.name)) {
+          finalSorted.push(symbolicTensor);
+          visited.add(symbolicTensor.name);
+        }
+      }
+      for (const name in recipientMap) {
+        if (finalRecipientMap[name] == null) {
+          finalRecipientMap[name] = /* @__PURE__ */ new Set();
+        }
+        recipientMap[name].forEach((recipient) => finalRecipientMap[name].add(recipient));
+      }
+    }
+  }
+  return {
+    sorted: finalSorted,
+    recipientCounts: recipientMap2Counts(finalRecipientMap)
+  };
+}
+function recipientMap2Counts(recipientMap) {
+  const recipientCounts = {};
+  for (const name in recipientMap) {
+    recipientCounts[name] = recipientMap[name].size;
+  }
+  return recipientCounts;
+}
+function getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict) {
+  const visited = /* @__PURE__ */ new Set();
+  const sorted = [];
+  const recipientMap = {};
+  for (const key of feedDict.names()) {
+    visited.add(key);
+  }
+  const stack2 = [];
+  const marks = [];
+  stack2.push(fetch4);
+  while (stack2.length > 0) {
+    const top = stack2[stack2.length - 1];
+    if (visited.has(top.name)) {
+      stack2.pop();
+      continue;
+    }
+    const topIsMarked = marks[marks.length - 1] === stack2.length - 1;
+    if (top.inputs.length === 0 || topIsMarked) {
+      stack2.pop();
+      sorted.push(top);
+      visited.add(top.name);
+      if (topIsMarked) {
+        marks.pop();
+      }
+    } else {
+      marks.push(stack2.length - 1);
+      for (const input2 of top.inputs) {
+        if (recipientMap[input2.name] == null) {
+          recipientMap[input2.name] = /* @__PURE__ */ new Set();
+        }
+        recipientMap[input2.name].add(top.name);
+        if (visited.has(input2.name)) {
+          continue;
+        }
+        stack2.push(input2);
+      }
+    }
+  }
+  return { sorted, recipientMap };
+}
+function getNodeOutputs(fetch4) {
+  let layerOutputs;
+  if (fetch4.sourceLayer.inboundNodes.length === 1) {
+    layerOutputs = fetch4.sourceLayer.output;
+  } else {
+    let nodeIndex = null;
+    for (let i = 0; i < fetch4.sourceLayer.inboundNodes.length; ++i) {
+      for (const outputTensor of fetch4.sourceLayer.inboundNodes[i].outputTensors) {
+        if (outputTensor.id === fetch4.id) {
+          nodeIndex = i;
+          break;
+        }
+      }
+    }
+    layerOutputs = fetch4.sourceLayer.getOutputAt(nodeIndex);
+  }
+  return layerOutputs;
+}
+var ENV3 = env();
+ENV3.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES", () => 100, updateCacheMaxEntries);
+var exports_constraints_exports = {};
+__export2(exports_constraints_exports, {
+  maxNorm: () => maxNorm,
+  minMaxNorm: () => minMaxNorm,
+  nonNeg: () => nonNeg,
+  unitNorm: () => unitNorm
+});
+function calcL2Norms(w, axis) {
+  return tidy(() => sqrt(sum2(mul(w, w), axis, true)));
+}
+var Constraint = class extends serialization_exports.Serializable {
+  getConfig() {
+    return {};
+  }
+};
+var MaxNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultMaxValue = 2;
+    this.defaultAxis = 0;
+    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => {
+      const norms = calcL2Norms(w, this.axis);
+      const desired = clipByValue(norms, 0, this.maxValue);
+      return mul(w, div(desired, add2(epsilon(), norms)));
+    });
+  }
+  getConfig() {
+    return { maxValue: this.maxValue, axis: this.axis };
+  }
+};
+MaxNorm.className = "MaxNorm";
+serialization_exports.registerClass(MaxNorm);
+var UnitNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultAxis = 0;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => div(w, add2(epsilon(), calcL2Norms(w, this.axis))));
+  }
+  getConfig() {
+    return { axis: this.axis };
+  }
+};
+UnitNorm.className = "UnitNorm";
+serialization_exports.registerClass(UnitNorm);
+var NonNeg = class extends Constraint {
+  apply(w) {
+    return relu(w);
+  }
+};
+NonNeg.className = "NonNeg";
+serialization_exports.registerClass(NonNeg);
+var MinMaxNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultMinValue = 0;
+    this.defaultMaxValue = 1;
+    this.defaultRate = 1;
+    this.defaultAxis = 0;
+    this.minValue = args.minValue != null ? args.minValue : this.defaultMinValue;
+    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;
+    this.rate = args.rate != null ? args.rate : this.defaultRate;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => {
+      const norms = calcL2Norms(w, this.axis);
+      const desired = add2(mul(this.rate, clipByValue(norms, this.minValue, this.maxValue)), mul(1 - this.rate, norms));
+      return mul(w, div(desired, add2(epsilon(), norms)));
+    });
+  }
+  getConfig() {
+    return {
+      minValue: this.minValue,
+      maxValue: this.maxValue,
+      rate: this.rate,
+      axis: this.axis
+    };
+  }
+};
+MinMaxNorm.className = "MinMaxNorm";
+serialization_exports.registerClass(MinMaxNorm);
+var CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "maxNorm": "MaxNorm",
+  "minMaxNorm": "MinMaxNorm",
+  "nonNeg": "NonNeg",
+  "unitNorm": "UnitNorm"
+};
+function serializeConstraint(constraint) {
+  return serializeKerasObject(constraint);
+}
+function deserializeConstraint(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "constraint");
+}
+function getConstraint(identifier) {
+  if (identifier == null) {
+    return null;
+  }
+  if (typeof identifier === "string") {
+    const className = identifier in CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP ? CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    const config = { className, config: {} };
+    return deserializeConstraint(config);
+  } else if (identifier instanceof Constraint) {
+    return identifier;
+  } else {
+    return deserializeConstraint(identifier);
+  }
+}
+function maxNorm(args) {
+  return new MaxNorm(args);
+}
+function unitNorm(args) {
+  return new UnitNorm(args);
+}
+function nonNeg() {
+  return new NonNeg();
+}
+function minMaxNorm(config) {
+  return new MinMaxNorm(config);
+}
+var exports_initializers_exports = {};
+__export2(exports_initializers_exports, {
+  constant: () => constant,
+  glorotNormal: () => glorotNormal,
+  glorotUniform: () => glorotUniform,
+  heNormal: () => heNormal,
+  heUniform: () => heUniform,
+  identity: () => identity,
+  leCunNormal: () => leCunNormal,
+  leCunUniform: () => leCunUniform,
+  ones: () => ones3,
+  orthogonal: () => orthogonal,
+  randomNormal: () => randomNormal3,
+  randomUniform: () => randomUniform2,
+  truncatedNormal: () => truncatedNormal2,
+  varianceScaling: () => varianceScaling,
+  zeros: () => zeros2
+});
+function zeros2() {
+  return new Zeros();
+}
+function ones3() {
+  return new Ones();
+}
+function constant(args) {
+  return new Constant(args);
+}
+function randomUniform2(args) {
+  return new RandomUniform(args);
+}
+function randomNormal3(args) {
+  return new RandomNormal(args);
+}
+function truncatedNormal2(args) {
+  return new TruncatedNormal(args);
+}
+function identity(args) {
+  return new Identity2(args);
+}
+function varianceScaling(config) {
+  return new VarianceScaling(config);
+}
+function glorotUniform(args) {
+  return new GlorotUniform(args);
+}
+function glorotNormal(args) {
+  return new GlorotNormal(args);
+}
+function heNormal(args) {
+  return new HeNormal(args);
+}
+function heUniform(args) {
+  return new HeUniform(args);
+}
+function leCunNormal(args) {
+  return new LeCunNormal(args);
+}
+function leCunUniform(args) {
+  return new LeCunUniform(args);
+}
+function orthogonal(args) {
+  return new Orthogonal(args);
+}
+var exports_layers_exports = {};
+__export2(exports_layers_exports, {
+  Layer: () => Layer,
+  RNN: () => RNN,
+  RNNCell: () => RNNCell,
+  activation: () => activation,
+  add: () => add3,
+  alphaDropout: () => alphaDropout,
+  average: () => average,
+  averagePooling1d: () => averagePooling1d,
+  averagePooling2d: () => averagePooling2d,
+  averagePooling3d: () => averagePooling3d,
+  avgPool1d: () => avgPool1d,
+  avgPool2d: () => avgPool2d,
+  avgPool3d: () => avgPool3d2,
+  avgPooling1d: () => avgPooling1d,
+  avgPooling2d: () => avgPooling2d,
+  avgPooling3d: () => avgPooling3d,
+  batchNormalization: () => batchNormalization2,
+  bidirectional: () => bidirectional,
+  categoryEncoding: () => categoryEncoding,
+  concatenate: () => concatenate2,
+  conv1d: () => conv1d2,
+  conv2d: () => conv2d3,
+  conv2dTranspose: () => conv2dTranspose2,
+  conv3d: () => conv3d2,
+  conv3dTranspose: () => conv3dTranspose2,
+  convLstm2d: () => convLstm2d,
+  convLstm2dCell: () => convLstm2dCell,
+  cropping2D: () => cropping2D,
+  dense: () => dense,
+  depthwiseConv2d: () => depthwiseConv2d4,
+  dot: () => dot3,
+  dropout: () => dropout3,
+  elu: () => elu3,
+  embedding: () => embedding,
+  flatten: () => flatten3,
+  gaussianDropout: () => gaussianDropout,
+  gaussianNoise: () => gaussianNoise,
+  globalAveragePooling1d: () => globalAveragePooling1d,
+  globalAveragePooling2d: () => globalAveragePooling2d,
+  globalMaxPool1d: () => globalMaxPool1d,
+  globalMaxPool2d: () => globalMaxPool2d,
+  globalMaxPooling1d: () => globalMaxPooling1d,
+  globalMaxPooling2d: () => globalMaxPooling2d,
+  gru: () => gru,
+  gruCell: () => gruCell,
+  input: () => input,
+  inputLayer: () => inputLayer,
+  layerNormalization: () => layerNormalization,
+  leakyReLU: () => leakyReLU,
+  lstm: () => lstm,
+  lstmCell: () => lstmCell,
+  masking: () => masking,
+  maxPool1d: () => maxPool1d,
+  maxPool2d: () => maxPool2d,
+  maxPooling1d: () => maxPooling1d,
+  maxPooling2d: () => maxPooling2d,
+  maxPooling3d: () => maxPooling3d,
+  maximum: () => maximum2,
+  minimum: () => minimum2,
+  multiply: () => multiply,
+  permute: () => permute,
+  prelu: () => prelu2,
+  reLU: () => reLU,
+  repeatVector: () => repeatVector,
+  rescaling: () => rescaling,
+  reshape: () => reshape2,
+  resizing: () => resizing,
+  rnn: () => rnn2,
+  separableConv2d: () => separableConv2d2,
+  simpleRNN: () => simpleRNN,
+  simpleRNNCell: () => simpleRNNCell,
+  softmax: () => softmax2,
+  spatialDropout1d: () => spatialDropout1d,
+  stackedRNNCells: () => stackedRNNCells,
+  thresholdedReLU: () => thresholdedReLU,
+  timeDistributed: () => timeDistributed,
+  upSampling2d: () => upSampling2d,
+  zeroPadding2d: () => zeroPadding2d
+});
+async function resolveScalarsInLogs(logs) {
+  if (logs == null) {
+    return;
+  }
+  const promises = [];
+  const keys = [];
+  const scalarsToDispose = [];
+  for (const key in logs) {
+    const value = logs[key];
+    if (typeof value !== "number") {
+      const valueScalar = value;
+      promises.push(valueScalar.data());
+      keys.push(key);
+      scalarsToDispose.push(valueScalar);
+    }
+  }
+  if (promises.length > 0) {
+    const values = await Promise.all(promises);
+    for (let i = 0; i < values.length; ++i) {
+      logs[keys[i]] = values[i][0];
+    }
+    dispose(scalarsToDispose);
+  }
+}
+function disposeTensorsInLogs(logs) {
+  if (logs == null) {
+    return;
+  }
+  for (const key in logs) {
+    const value = logs[key];
+    if (typeof value !== "number") {
+      value.dispose();
+    }
+  }
+}
+var ModelLoggingVerbosity;
+(function(ModelLoggingVerbosity2) {
+  ModelLoggingVerbosity2[ModelLoggingVerbosity2["SILENT"] = 0] = "SILENT";
+  ModelLoggingVerbosity2[ModelLoggingVerbosity2["VERBOSE"] = 1] = "VERBOSE";
+})(ModelLoggingVerbosity || (ModelLoggingVerbosity = {}));
+var DEFAULT_YIELD_EVERY_MS = 125;
+var BaseCallback = class {
+  constructor() {
+    this.validationData = null;
+  }
+  setParams(params) {
+    this.params = params;
+  }
+  async onEpochBegin(epoch, logs) {
+  }
+  async onEpochEnd(epoch, logs) {
+  }
+  async onBatchBegin(batch, logs) {
+  }
+  async onBatchEnd(batch, logs) {
+  }
+  async onTrainBegin(logs) {
+  }
+  async onTrainEnd(logs) {
+  }
+  setModel(model2) {
+  }
+};
+var CallbackList = class {
+  constructor(callbacks2, queueLength = 10) {
+    if (callbacks2 == null) {
+      callbacks2 = [];
+    }
+    this.callbacks = callbacks2;
+    this.queueLength = queueLength;
+  }
+  append(callback) {
+    this.callbacks.push(callback);
+  }
+  setParams(params) {
+    for (const callback of this.callbacks) {
+      callback.setParams(params);
+    }
+  }
+  setModel(model2) {
+    for (const callback of this.callbacks) {
+      callback.setModel(model2);
+    }
+  }
+  async onEpochBegin(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onEpochBegin(epoch, logs);
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onEpochEnd(epoch, logs);
+    }
+  }
+  async onBatchBegin(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onBatchBegin(batch, logs);
+    }
+  }
+  async onBatchEnd(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onBatchEnd(batch, logs);
+    }
+  }
+  async onTrainBegin(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onTrainBegin(logs);
+    }
+  }
+  async onTrainEnd(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onTrainEnd(logs);
+    }
+  }
+};
+var BaseLogger = class extends BaseCallback {
+  constructor() {
+    super();
+  }
+  async onEpochBegin(epoch) {
+    this.seen = 0;
+    this.totals = {};
+  }
+  async onBatchEnd(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    const batchSize = logs["size"] == null ? 0 : logs["size"];
+    this.seen += batchSize;
+    for (const key in logs) {
+      const value = logs[key];
+      if (typeof value === "number") {
+        if (!this.totals.hasOwnProperty(key)) {
+          this.totals[key] = 0;
+        }
+        this.totals[key] = this.totals[key] + value * batchSize;
+      } else {
+        let oldTotalsToDispose;
+        if (key in this.totals) {
+          oldTotalsToDispose = this.totals[key];
+        } else {
+          this.totals[key] = 0;
+        }
+        const total = tidy(() => add2(this.totals[key], mul(value, batchSize)));
+        this.totals[key] = total;
+        if (oldTotalsToDispose != null) {
+          oldTotalsToDispose.dispose();
+        }
+      }
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs != null) {
+      for (const key of this.params["metrics"]) {
+        if (this.totals[key] == null) {
+          continue;
+        }
+        if (typeof this.totals[key] === "number") {
+          logs[key] = this.totals[key] / this.seen;
+        } else {
+          tidy(() => {
+            const log5 = mul(div(1, this.seen), this.totals[key]);
+            logs[key] = log5;
+            this.totals[key].dispose();
+            keep(logs[key]);
+          });
+        }
+      }
+    }
+  }
+};
+var History = class extends BaseCallback {
+  async onTrainBegin(logs) {
+    this.epoch = [];
+    this.history = {};
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    this.epoch.push(epoch);
+    for (const key in logs) {
+      if (this.history[key] == null) {
+        this.history[key] = [];
+      }
+      this.history[key].push(logs[key]);
+    }
+  }
+  async syncData() {
+    const promises = [];
+    const keys = [];
+    const indices = [];
+    for (const key in this.history) {
+      const valueArray = this.history[key];
+      for (let i = 0; i < valueArray.length; ++i) {
+        if (typeof valueArray[i] !== "number") {
+          const valueScalar = valueArray[i];
+          promises.push(valueScalar.data());
+          keys.push(key);
+          indices.push(i);
+        }
+      }
+    }
+    const values = await Promise.all(promises);
+    for (let n = 0; n < values.length; ++n) {
+      const tensorToDispose = this.history[keys[n]][indices[n]];
+      tensorToDispose.dispose();
+      this.history[keys[n]][indices[n]] = values[n][0];
+    }
+  }
+};
+var CustomCallback = class extends BaseCallback {
+  constructor(args, yieldEvery) {
+    super();
+    this.currentEpoch = 0;
+    this.nowFunc = args.nowFunc;
+    this.nextFrameFunc = args.nextFrameFunc || nextFrame;
+    this.yieldEvery = yieldEvery || "auto";
+    if (this.yieldEvery === "auto") {
+      this.yieldEvery = DEFAULT_YIELD_EVERY_MS;
+    }
+    if (this.yieldEvery === "never" && args.onYield != null) {
+      throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");
+    }
+    if (util_exports.isNumber(this.yieldEvery)) {
+      this.maybeWait = debounce(this.maybeWait.bind(this), this.yieldEvery, this.nowFunc);
+    }
+    this.trainBegin = args.onTrainBegin;
+    this.trainEnd = args.onTrainEnd;
+    this.epochBegin = args.onEpochBegin;
+    this.epochEnd = args.onEpochEnd;
+    this.batchBegin = args.onBatchBegin;
+    this.batchEnd = args.onBatchEnd;
+    this.yield = args.onYield;
+  }
+  async maybeWait(epoch, batch, logs) {
+    const ps = [];
+    if (this.yield != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.yield(epoch, batch, logs));
+    }
+    ps.push(this.nextFrameFunc());
+    await Promise.all(ps);
+  }
+  async onEpochBegin(epoch, logs) {
+    this.currentEpoch = epoch;
+    if (this.epochBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.epochBegin(epoch, logs);
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    const ps = [];
+    if (this.epochEnd != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.epochEnd(epoch, logs));
+    }
+    if (this.yieldEvery === "epoch") {
+      ps.push(this.nextFrameFunc());
+    }
+    await Promise.all(ps);
+  }
+  async onBatchBegin(batch, logs) {
+    if (this.batchBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.batchBegin(batch, logs);
+    }
+  }
+  async onBatchEnd(batch, logs) {
+    const ps = [];
+    if (this.batchEnd != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.batchEnd(batch, logs));
+    }
+    if (this.yieldEvery === "batch") {
+      ps.push(this.nextFrameFunc());
+    } else if (util_exports.isNumber(this.yieldEvery)) {
+      ps.push(this.maybeWait(this.currentEpoch, batch, logs));
+    }
+    await Promise.all(ps);
+  }
+  async onTrainBegin(logs) {
+    if (this.trainBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.trainBegin(logs);
+    }
+  }
+  async onTrainEnd(logs) {
+    if (this.trainEnd != null) {
+      await resolveScalarsInLogs(logs);
+      await this.trainEnd(logs);
+    }
+  }
+};
+function standardizeCallbacks(callbacks2, yieldEvery) {
+  if (callbacks2 == null) {
+    callbacks2 = {};
+  }
+  if (callbacks2 instanceof BaseCallback) {
+    return [callbacks2];
+  }
+  if (Array.isArray(callbacks2) && callbacks2[0] instanceof BaseCallback) {
+    return callbacks2;
+  }
+  const callbackConfigs = toList(callbacks2);
+  return callbackConfigs.map((callbackConfig) => new CustomCallback(callbackConfig, yieldEvery));
+}
+var CallbackConstructorRegistry = class {
+  constructor() {
+  }
+  static registerCallbackConstructor(verbosityLevel, callbackConstructor) {
+    util_exports.assert(verbosityLevel >= 0 && Number.isInteger(verbosityLevel), () => `Verbosity level is expected to be an integer >= 0, but got ${verbosityLevel}`);
+    CallbackConstructorRegistry.checkForDuplicate(callbackConstructor);
+    if (CallbackConstructorRegistry.constructors[verbosityLevel] == null) {
+      CallbackConstructorRegistry.constructors[verbosityLevel] = [];
+    }
+    CallbackConstructorRegistry.constructors[verbosityLevel].push(callbackConstructor);
+  }
+  static checkForDuplicate(callbackConstructor) {
+    for (const levelName in CallbackConstructorRegistry.constructors) {
+      const constructors = CallbackConstructorRegistry.constructors[+levelName];
+      constructors.forEach((ctor) => {
+        if (ctor === callbackConstructor) {
+          throw new ValueError("Duplicate callback constructor.");
+        }
+      });
+    }
+  }
+  static clear() {
+    CallbackConstructorRegistry.constructors = {};
+  }
+  static createCallbacks(verbosityLevel) {
+    const constructors = [];
+    for (const levelName in CallbackConstructorRegistry.constructors) {
+      const level = +levelName;
+      if (verbosityLevel >= level) {
+        constructors.push(...CallbackConstructorRegistry.constructors[level]);
+      }
+    }
+    return constructors.map((ctor) => new ctor());
+  }
+};
+CallbackConstructorRegistry.constructors = {};
+function configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics) {
+  const history = new History();
+  const actualCallbacks = [
+    new BaseLogger(),
+    ...CallbackConstructorRegistry.createCallbacks(verbose)
+  ];
+  if (callbacks2 != null) {
+    actualCallbacks.push(...callbacks2);
+  }
+  actualCallbacks.push(history);
+  const callbackList = new CallbackList(actualCallbacks);
+  callbackList.setParams({
+    epochs,
+    initialEpoch,
+    samples: numTrainSamples,
+    steps: stepsPerEpoch,
+    batchSize,
+    verbose,
+    doValidation,
+    metrics: callbackMetrics
+  });
+  return { callbackList, history };
+}
+function deserialize(config, customObjects = {}, fastWeightInit = false) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "layer", fastWeightInit);
+}
+function l2Normalize(x, axis) {
+  return tidy(() => {
+    if (x.dtype !== "float32") {
+      x = cast(x, "float32");
+    }
+    const squareSum = sum2(square2(x), axis, true);
+    const epsilonTensor = fill(squareSum.shape, epsilon());
+    const norm2 = sqrt(maximum(squareSum, epsilonTensor));
+    return div(x, norm2);
+  });
+}
+function meanSquaredError2(yTrue, yPred) {
+  return tidy(() => mean(square2(sub(yPred, yTrue)), -1));
+}
+function meanAbsoluteError(yTrue, yPred) {
+  return tidy(() => mean(abs(sub(yPred, yTrue)), -1));
+}
+function meanAbsolutePercentageError(yTrue, yPred) {
+  return tidy(() => {
+    const diff = sub(yTrue, yPred);
+    const clippedTrue = clipByValue(abs(yTrue), epsilon(), Number.MAX_VALUE);
+    const absResult = abs(div(diff, clippedTrue));
+    return mul(100, mean(absResult, -1));
+  });
+}
+function meanSquaredLogarithmicError(yTrue, yPred) {
+  return tidy(() => {
+    const clippedPred = clipByValue(yPred, epsilon(), Number.MAX_VALUE);
+    const firstLog = log2(add2(1, clippedPred));
+    const clippedTrue = clipByValue(yTrue, epsilon(), Number.MAX_VALUE);
+    const secondLog = log2(add2(1, clippedTrue));
+    return mean(square2(sub(firstLog, secondLog)), -1);
+  });
+}
+function squaredHinge(yTrue, yPred) {
+  return tidy(() => {
+    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));
+    return mean(square2(maxResult), -1);
+  });
+}
+function hinge(yTrue, yPred) {
+  return tidy(() => {
+    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));
+    return mean(maxResult, -1);
+  });
+}
+function categoricalHinge(yTrue, yPred) {
+  return tidy(() => {
+    const pos = sum2(mul(yTrue, yPred), -1);
+    const neg4 = max(mul(sub(1, yTrue), yPred), -1);
+    return maximum(0, add2(1, sub(neg4, pos)));
+  });
+}
+function logcosh(yTrue, yPred) {
+  return tidy(() => {
+    const log22 = Math.log(2);
+    const predictionDiff = sub(yPred, yTrue);
+    const logcoshResult = sub(add2(predictionDiff, softplus(mul(-2, predictionDiff))), log22);
+    return mean(logcoshResult, -1);
+  });
+}
+function categoricalCrossentropy(target, output, fromLogits = false) {
+  return tidy(() => {
+    if (fromLogits) {
+      output = softmax(output);
+    } else {
+      const outputSum = sum2(output, output.shape.length - 1, true);
+      output = div(output, outputSum);
+    }
+    output = clipByValue(output, epsilon(), 1 - epsilon());
+    return neg(sum2(mul(cast(target, "float32"), log2(output)), output.shape.length - 1));
+  });
+}
+function sparseCategoricalCrossentropy(target, output, fromLogits = false) {
+  return tidy(() => {
+    const flatTarget = cast(floor(flatten2(target)), "int32");
+    output = clipByValue(output, epsilon(), 1 - epsilon());
+    const outputShape = output.shape;
+    const oneHotTarget = reshape(oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);
+    return categoricalCrossentropy(oneHotTarget, output, fromLogits);
+  });
+}
+function sigmoidCrossEntropyWithLogits(labels, logits) {
+  if (!util_exports.arraysEqual(labels.shape, logits.shape)) {
+    throw new ValueError(`logits and labels must have the same shape, but got shapes ${JSON.stringify(labels.shape)} and ${JSON.stringify(logits.shape)}`);
+  }
+  return tidy(() => {
+    const reluLogits = relu(logits);
+    const negAbsLogits = neg(abs(logits));
+    return add2(sub(reluLogits, mul(logits, labels)), log1p(exp(negAbsLogits)));
+  });
+}
+function binaryCrossentropy(yTrue, yPred) {
+  return tidy(() => {
+    let y;
+    y = clipByValue(yPred, epsilon(), 1 - epsilon());
+    y = log2(div(y, sub(1, y)));
+    return mean(sigmoidCrossEntropyWithLogits(yTrue, y), -1);
+  });
+}
+function kullbackLeiblerDivergence(yTrue, yPred) {
+  return tidy(() => {
+    const clippedTrue = clipByValue(yTrue, epsilon(), 1);
+    const clippedPred = clipByValue(yPred, epsilon(), 1);
+    return sum2(mul(yTrue, log2(div(clippedTrue, clippedPred))), -1);
+  });
+}
+function poisson(yTrue, yPred) {
+  return tidy(() => {
+    const logPred = log2(add2(epsilon(), yPred));
+    return mean(sub(yPred, mul(yTrue, logPred)), -1);
+  });
+}
+function cosineProximity(yTrue, yPred) {
+  return tidy(() => {
+    const trueNormalized = l2Normalize(yTrue, -1);
+    const predNormalized = l2Normalize(yPred, -1);
+    const trueXPred = mul(trueNormalized, predNormalized);
+    return neg(sum2(trueXPred, -1));
+  });
+}
+var lossesMap = {
+  meanSquaredError: meanSquaredError2,
+  meanAbsoluteError,
+  meanAbsolutePercentageError,
+  meanSquaredLogarithmicError,
+  squaredHinge,
+  hinge,
+  categoricalHinge,
+  logcosh,
+  categoricalCrossentropy,
+  sparseCategoricalCrossentropy,
+  binaryCrossentropy,
+  kullbackLeiblerDivergence,
+  poisson,
+  cosineProximity
+};
+function get(identifierOrFn) {
+  if (typeof identifierOrFn === "string") {
+    if (identifierOrFn in lossesMap) {
+      return lossesMap[identifierOrFn];
+    }
+    let errMsg = `Unknown loss ${identifierOrFn}`;
+    if (identifierOrFn.toLowerCase().includes("softmaxcrossentropy")) {
+      errMsg = `Unknown loss ${identifierOrFn}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`;
+    }
+    throw new ValueError(errMsg);
+  } else {
+    return identifierOrFn;
+  }
+}
+function binaryAccuracy(yTrue, yPred) {
+  return tidy(() => {
+    const threshold3 = mul(0.5, onesLike(yPred));
+    const yPredThresholded = cast2(greater(yPred, threshold3), yTrue.dtype);
+    return mean(equal(yTrue, yPredThresholded), -1);
+  });
+}
+function categoricalAccuracy(yTrue, yPred) {
+  return tidy(() => cast2(equal(argMax(yTrue, -1), argMax(yPred, -1)), "float32"));
+}
+function truePositives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 1))), "float32");
+  });
+}
+function falseNegatives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 0))), "float32");
+  });
+}
+function falsePositives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 0), equal(yPred, 1))), "float32");
+  });
+}
+function precision(yTrue, yPred) {
+  return tidy(() => {
+    const tp = truePositives(yTrue, yPred);
+    const fp = falsePositives(yTrue, yPred);
+    const denominator = add2(tp, fp);
+    return cast(where(greater(denominator, 0), div(tp, denominator), 0), "float32");
+  });
+}
+function recall(yTrue, yPred) {
+  return tidy(() => {
+    const tp = truePositives(yTrue, yPred);
+    const fn = falseNegatives(yTrue, yPred);
+    const denominator = add2(tp, fn);
+    return cast(where(greater(denominator, 0), div(tp, denominator), 0), "float32");
+  });
+}
+function binaryCrossentropy2(yTrue, yPred) {
+  return binaryCrossentropy(yTrue, yPred);
+}
+function sparseCategoricalAccuracy(yTrue, yPred) {
+  if (yTrue.rank === yPred.rank) {
+    yTrue = squeeze(yTrue, [yTrue.rank - 1]);
+  }
+  yPred = argMax(yPred, -1);
+  if (yPred.dtype !== yTrue.dtype) {
+    yPred = cast(yPred, yTrue.dtype);
+  }
+  return cast(equal(yTrue, yPred), "float32");
+}
+var mse = meanSquaredError2;
+var MSE = meanSquaredError2;
+var mae = meanAbsoluteError;
+var MAE = meanAbsoluteError;
+var mape = meanAbsolutePercentageError;
+var MAPE = meanAbsolutePercentageError;
+var categoricalCrossentropy2 = categoricalCrossentropy;
+var cosine = cosineProximity;
+var sparseCategoricalCrossentropy2 = sparseCategoricalCrossentropy;
+var metricsMap = {
+  binaryAccuracy,
+  categoricalAccuracy,
+  precision,
+  categoricalCrossentropy: categoricalCrossentropy2,
+  sparseCategoricalCrossentropy: sparseCategoricalCrossentropy2,
+  mse,
+  MSE,
+  mae,
+  MAE,
+  mape,
+  MAPE,
+  cosine
+};
+function get2(identifier) {
+  if (typeof identifier === "string" && identifier in metricsMap) {
+    return metricsMap[identifier];
+  } else if (typeof identifier !== "string" && identifier != null) {
+    return identifier;
+  } else {
+    throw new ValueError(`Unknown metric ${identifier}`);
+  }
+}
+function getLossOrMetricName(fn) {
+  assert2(fn !== null, `Unknown LossOrMetricFn ${fn}`);
+  if (typeof fn === "string") {
+    return fn;
+  } else {
+    let fnName;
+    for (const key of Object.keys(lossesMap)) {
+      if (lossesMap[key] === fn) {
+        fnName = key;
+        break;
+      }
+    }
+    if (fnName !== void 0) {
+      return fnName;
+    }
+    for (const key of Object.keys(metricsMap)) {
+      if (metricsMap[key] === fn) {
+        fnName = key;
+        break;
+      }
+    }
+    if (fnName !== void 0) {
+      return fnName;
+    }
+    return fn.name;
+  }
+}
+function getOptimizer(identifier) {
+  const optimizerMap = {
+    "Adagrad": () => train.adagrad(0.01),
+    "Adadelta": () => train.adadelta(1, 0.95, epsilon()),
+    "Adam": () => train.adam(1e-3, 0.9, 0.999, epsilon()),
+    "Adamax": () => train.adamax(2e-3, 0.9, 0.999, epsilon(), 0),
+    "RMSProp": () => train.rmsprop(1e-3, 0.9, 0, epsilon()),
+    "SGD": () => train.sgd(0.01)
+  };
+  optimizerMap["adagrad"] = optimizerMap["Adagrad"];
+  optimizerMap["adadelta"] = optimizerMap["Adadelta"];
+  optimizerMap["adam"] = optimizerMap["Adam"];
+  optimizerMap["adamax"] = optimizerMap["Adamax"];
+  optimizerMap["rmsprop"] = optimizerMap["RMSProp"];
+  optimizerMap["sgd"] = optimizerMap["SGD"];
+  if (identifier in optimizerMap) {
+    return optimizerMap[identifier]();
+  }
+  throw new ValueError(`Unknown Optimizer ${identifier}`);
+}
+var MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH = 1 * 1024 * 1024;
+function checkUserDefinedMetadata(userDefinedMetadata, modelName, checkSize = false) {
+  if (userDefinedMetadata == null || typeof userDefinedMetadata !== "object" || Object.getPrototypeOf(userDefinedMetadata) !== Object.prototype || !plainObjectCheck(userDefinedMetadata)) {
+    throw new Error("User-defined metadata is expected to be a JSON object, but is not.");
+  }
+  if (checkSize) {
+    const out = JSON.stringify(userDefinedMetadata);
+    if (out.length > MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH) {
+      console.warn(`User-defined metadata of model "${modelName}" is too large in size (length=${out.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH}.`);
+    }
+  }
+}
+function plainObjectCheck(x) {
+  if (x === null) {
+    return true;
+  } else if (typeof x === "object") {
+    if (Object.getPrototypeOf(x) === Object.prototype) {
+      const keys = Object.keys(x);
+      for (const key of keys) {
+        if (typeof key !== "string") {
+          return false;
+        }
+        if (!plainObjectCheck(x[key])) {
+          return false;
+        }
+      }
+      return true;
+    } else {
+      if (Array.isArray(x)) {
+        for (const item of x) {
+          if (!plainObjectCheck(item)) {
+            return false;
+          }
+        }
+        return true;
+      } else {
+        return false;
+      }
+    }
+  } else {
+    const xType = typeof x;
+    return xType === "string" || xType === "number" || xType === "boolean";
+  }
+}
+function printSummary(model2, lineLength, positions, printFn = console.log) {
+  const sequentialLike = isModelSequentialLike(model2);
+  const toDisplay = ["Layer (type)", "Input Shape", "Output shape", "Param #"];
+  if (sequentialLike) {
+    lineLength = lineLength || 90;
+    positions = positions || [0.32, 0.61, 0.89, 1];
+  } else {
+    lineLength = lineLength || 115;
+    positions = positions || [0.24, 0.48, 0.7, 0.8, 1];
+  }
+  if (positions[positions.length - 1] <= 1) {
+    positions = positions.map((p2) => Math.floor(lineLength * p2));
+  }
+  let relevantNodes;
+  if (!sequentialLike) {
+    toDisplay.push("Receives inputs");
+    relevantNodes = [];
+    for (const depth in model2.nodesByDepth) {
+      relevantNodes.push(...model2.nodesByDepth[depth]);
+    }
+  }
+  printFn("_".repeat(lineLength));
+  printRow(toDisplay, positions, printFn);
+  printFn("=".repeat(lineLength));
+  const layers = model2.layers;
+  for (let i = 0; i < layers.length; ++i) {
+    if (sequentialLike) {
+      printLayerSummary(layers[i], positions, printFn);
+    } else {
+      printLayerSummaryWithConnections(layers[i], positions, relevantNodes, printFn);
+    }
+    printFn((i === layers.length - 1 ? "=" : "_").repeat(lineLength));
+  }
+  model2.checkTrainableWeightsConsistency();
+  const trainableCount = countTrainableParams(model2);
+  const nonTrainableCount = countParamsInWeights(model2.nonTrainableWeights);
+  printFn(`Total params: ${trainableCount + nonTrainableCount}`);
+  printFn(`Trainable params: ${trainableCount}`);
+  printFn(`Non-trainable params: ${nonTrainableCount}`);
+  printFn("_".repeat(lineLength));
+}
+function countTrainableParams(model2) {
+  let trainableCount;
+  if (model2.collectedTrainableWeights != null) {
+    trainableCount = countParamsInWeights(model2.collectedTrainableWeights);
+  } else {
+    trainableCount = countParamsInWeights(model2.trainableWeights);
+  }
+  return trainableCount;
+}
+function isModelSequentialLike(model2) {
+  let sequentialLike = true;
+  const nodesByDepth = [];
+  const nodes = [];
+  for (const depth in model2.nodesByDepth) {
+    nodesByDepth.push(model2.nodesByDepth[depth]);
+  }
+  for (const depthNodes of nodesByDepth) {
+    if (depthNodes.length > 1 || depthNodes.length === 1 && depthNodes[0].inboundLayers.length > 1) {
+      sequentialLike = false;
+      break;
+    }
+    nodes.push(...depthNodes);
+  }
+  if (sequentialLike) {
+    for (const layer of model2.layers) {
+      let flag = false;
+      for (const node of layer.inboundNodes) {
+        if (nodes.indexOf(node) !== -1) {
+          if (flag) {
+            sequentialLike = false;
+            break;
+          } else {
+            flag = true;
+          }
+        }
+      }
+      if (!sequentialLike) {
+        break;
+      }
+    }
+  }
+  return sequentialLike;
+}
+function printRow(fields, positions, printFn = console.log) {
+  let line = "";
+  for (let i = 0; i < fields.length; ++i) {
+    if (i > 0) {
+      line = line.slice(0, line.length - 1) + " ";
+    }
+    line += fields[i];
+    line = line.slice(0, positions[i]);
+    line += " ".repeat(positions[i] - line.length);
+  }
+  printFn(line);
+}
+function printLayerSummary(layer, positions, printFn) {
+  let outputShape;
+  let inputShape;
+  try {
+    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(",");
+  } catch (err) {
+    inputShape = "multiple";
+  }
+  try {
+    outputShape = JSON.stringify(layer.outputShape);
+  } catch (err) {
+    outputShape = "multiple";
+  }
+  const name = layer.name;
+  const className = layer.getClassName();
+  const fields = [
+    `${name} (${className})`,
+    inputShape,
+    outputShape,
+    layer.countParams().toString()
+  ];
+  printRow(fields, positions, printFn);
+}
+function printLayerSummaryWithConnections(layer, positions, relevantNodes, printFn) {
+  let outputShape;
+  let inputShape;
+  try {
+    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(",");
+  } catch (err) {
+    inputShape = "multiple";
+  }
+  try {
+    outputShape = JSON.stringify(layer.outputShape);
+  } catch (err) {
+    outputShape = "multiple";
+  }
+  const connections = [];
+  for (const node of layer.inboundNodes) {
+    if (relevantNodes != null && relevantNodes.length > 0 && relevantNodes.indexOf(node) === -1) {
+      continue;
+    }
+    for (let i = 0; i < node.inboundLayers.length; ++i) {
+      const inboundLayer = node.inboundLayers[i].name;
+      const inboundLayerIndex = node.nodeIndices[i];
+      const inboundTensorIndex = node.tensorIndices[i];
+      connections.push(`${inboundLayer}[${inboundLayerIndex}][${inboundTensorIndex}]`);
+    }
+  }
+  const name = layer.name;
+  const className = layer.getClassName();
+  const firstConnection = connections.length === 0 ? "" : connections[0];
+  const fields = [
+    `${name} (${className})`,
+    inputShape,
+    outputShape,
+    layer.countParams().toString(),
+    firstConnection
+  ];
+  printRow(fields, positions, printFn);
+  for (let i = 1; i < connections.length; ++i) {
+    printRow(["", "", "", "", connections[i]], positions, printFn);
+  }
+}
+function isArrayItemInputOrOutputName(key, index, value) {
+  return (key === "inboundNodes" || key === "outputLayers" || key === "inputLayers") && index === 0 && typeof value === "string";
+}
+function convertPythonicToTs(pythonicConfig, key) {
+  if (pythonicConfig === null) {
+    return null;
+  } else if (typeof pythonicConfig === "string") {
+    return toCamelCase(pythonicConfig);
+  } else if (typeof pythonicConfig === "number" || typeof pythonicConfig === "boolean") {
+    return pythonicConfig;
+  } else if (pythonicConfig instanceof Array) {
+    const tsArray = [];
+    const arrayLength = pythonicConfig.length;
+    for (let i = 0; i < arrayLength; ++i) {
+      const item = pythonicConfig[i];
+      if (isArrayItemInputOrOutputName(key, i, item)) {
+        tsArray.push(item);
+      } else {
+        tsArray.push(convertPythonicToTs(item, key));
+      }
+    }
+    return tsArray;
+  } else {
+    const tsDict = {};
+    for (const pythonicKey of Object.keys(pythonicConfig)) {
+      const pythonicValue = pythonicConfig[pythonicKey];
+      if (pythonicKey === "name" && typeof pythonicValue === "string") {
+        tsDict[pythonicKey] = pythonicValue;
+      } else {
+        const tsKey = toCamelCase(pythonicKey);
+        tsDict[tsKey] = convertPythonicToTs(pythonicValue, tsKey);
+      }
+    }
+    return tsDict;
+  }
+}
+function convertTsToPythonic(tsConfig, key) {
+  if (tsConfig === null || tsConfig === void 0) {
+    return null;
+  } else if (typeof tsConfig === "string") {
+    return toSnakeCase(tsConfig);
+  } else if (typeof tsConfig === "number" || typeof tsConfig === "boolean") {
+    return tsConfig;
+  } else if (tsConfig instanceof Array) {
+    const pyArray = [];
+    const arrayLength = tsConfig.length;
+    for (let i = 0; i < arrayLength; ++i) {
+      const item = tsConfig[i];
+      if (isArrayItemInputOrOutputName(key, i, item)) {
+        pyArray.push(item);
+      } else {
+        pyArray.push(convertTsToPythonic(item, key));
+      }
+    }
+    return pyArray;
+  } else {
+    const pyDict = {};
+    for (const tsKey of Object.keys(tsConfig)) {
+      const tsValue = tsConfig[tsKey];
+      const pyKey = toSnakeCase(tsKey);
+      if ((tsKey === "name" || tsKey === "className") && typeof tsValue === "string") {
+        pyDict[pyKey] = tsValue;
+      } else {
+        pyDict[pyKey] = convertTsToPythonic(tsValue, tsKey);
+      }
+    }
+    return pyDict;
+  }
+}
+var version2 = "4.0.0";
+var Container = class extends Layer {
+  constructor(args) {
+    super({});
+    this.containerNodes = /* @__PURE__ */ new Set();
+    this.name = args.name;
+    if (this.name == null) {
+      const prefix = this.getClassName().toLowerCase();
+      this.name = getUid(prefix);
+    }
+    this.supportsMasking = false;
+    this.trainable_ = true;
+    if (Array.isArray(args.inputs)) {
+      this.inputs = args.inputs.slice();
+    } else {
+      this.inputs = [args.inputs];
+    }
+    if (Array.isArray(args.outputs)) {
+      this.outputs = args.outputs.slice();
+    } else {
+      this.outputs = [args.outputs];
+    }
+    if (unique2(this.inputs).length !== this.inputs.length) {
+      throw new ValueError(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map((x) => x.name)}`);
+    }
+    if (unique2(this.outputs).length !== this.outputs.length) {
+      console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map((x) => x.name)}`);
+    }
+    this.inputLayers = [];
+    this.inputLayersNodeIndices = [];
+    this.inputLayersTensorIndices = [];
+    this.outputLayers = [];
+    this.outputLayersNodeIndices = [];
+    this.outputLayersTensorIndices = [];
+    this.layers = [];
+    this.internalContainerRefs = [];
+    for (const x of this.outputs) {
+      const layer = x.sourceLayer;
+      const nodeIndex = x.nodeIndex;
+      const tensorIndex = x.tensorIndex;
+      this.outputLayers.push(layer);
+      this.outputLayersNodeIndices.push(nodeIndex);
+      this.outputLayersTensorIndices.push(tensorIndex);
+    }
+    for (const x of this.inputs) {
+      const layer = x.sourceLayer;
+      const nodeIndex = x.nodeIndex;
+      const tensorIndex = x.tensorIndex;
+      assert2(nodeIndex === 0, "input layer has >1 nodes");
+      assert2(tensorIndex === 0, "input layer has >1 tensors");
+      this.inputLayers.push(layer);
+      this.inputLayersNodeIndices.push(nodeIndex);
+      this.inputLayersTensorIndices.push(tensorIndex);
+    }
+    this.inputNames = [];
+    this.outputNames = [];
+    this.feedInputShapes = [];
+    this.feedInputNames = [];
+    this.feedOutputNames = [];
+    for (let i = 0; i < this.inputLayers.length; i++) {
+      const layer = this.inputLayers[i];
+      if (!(layer instanceof InputLayer)) {
+        throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${args.inputs}. Input ${i} (0-based) originates from layer type ${layer.getClassName()}.`);
+      }
+      this.inputNames.push(layer.name);
+      this.feedInputShapes.push(layer.batchInputShape);
+      this.feedInputNames.push(layer.name);
+    }
+    for (const layer of this.outputLayers) {
+      this.outputNames.push(layer.name);
+    }
+    this.internalInputShapes = this.inputs.map((x) => x.shape);
+    this.internalOutputShapes = this.outputs.map((x) => x.shape);
+    const nodesDepths = {};
+    const nodeIDToNode = {};
+    const layersDepths = {};
+    const layerIDToLayer = {};
+    const layerIndices = {};
+    const nodesInDecreasingDepth = [];
+    const buildMapOfGraph = (tensor2, finishedNodes2, nodesInProgress2, layer, nodeIndex, tensorIndex) => {
+      if (layer == null || nodeIndex == null || tensorIndex == null) {
+        layer = tensor2.sourceLayer;
+        nodeIndex = tensor2.nodeIndex;
+        tensorIndex = tensor2.tensorIndex;
+      }
+      const node = layer.inboundNodes[nodeIndex];
+      if (nodesInProgress2.indexOf(node) !== -1) {
+        throw new RuntimeError(`The tensor ${tensor2.name} at layer "${layer.name}" is part of a cycle.`);
+      }
+      if (finishedNodes2.indexOf(node) !== -1) {
+        return;
+      }
+      this.containerNodes.add(Container.nodeKey(layer, nodeIndex));
+      if (!(layer.id in layerIndices)) {
+        layerIndices[layer.id] = Object.keys(layerIndices).length;
+      }
+      if (nodesInProgress2.indexOf(node) === -1) {
+        nodesInProgress2.push(node);
+      }
+      const numInboundLayers = node.inboundLayers.length;
+      for (let i = 0; i < numInboundLayers; i++) {
+        const x = node.inputTensors[i];
+        const layer2 = node.inboundLayers[i];
+        const nodeIndex2 = node.nodeIndices[i];
+        const tensorIndex2 = node.tensorIndices[i];
+        buildMapOfGraph(x, finishedNodes2, nodesInProgress2, layer2, nodeIndex2, tensorIndex2);
+      }
+      finishedNodes2.push(node);
+      while (nodesInProgress2.indexOf(node) >= 0) {
+        nodesInProgress2.splice(nodesInProgress2.indexOf(node), 1);
+      }
+      nodesInDecreasingDepth.push(node);
+    };
+    const finishedNodes = [];
+    const nodesInProgress = [];
+    for (const x of this.outputs) {
+      buildMapOfGraph(x, finishedNodes, nodesInProgress);
+    }
+    const reversedNodesInDecreasingDepth = nodesInDecreasingDepth.slice().reverse();
+    for (const node of reversedNodesInDecreasingDepth) {
+      nodeIDToNode[node.id] = node;
+      if (!(node.id in nodesDepths)) {
+        nodesDepths[node.id] = 0;
+      }
+      let depth = nodesDepths[node.id];
+      const previousDepth = layersDepths[node.outboundLayer.id] == null ? 0 : layersDepths[node.outboundLayer.id];
+      depth = Math.max(depth, previousDepth);
+      layersDepths[node.outboundLayer.id] = depth;
+      layerIDToLayer[node.outboundLayer.id] = node.outboundLayer;
+      nodesDepths[node.id] = depth;
+      for (let i = 0; i < node.inboundLayers.length; i++) {
+        const inboundLayer = node.inboundLayers[i];
+        const nodeIndex = node.nodeIndices[i];
+        const inboundNode = inboundLayer.inboundNodes[nodeIndex];
+        const previousDepth2 = nodesDepths[inboundNode.id] == null ? 0 : nodesDepths[inboundNode.id];
+        nodesDepths[inboundNode.id] = Math.max(depth + 1, previousDepth2);
+        nodeIDToNode[inboundNode.id] = inboundNode;
+      }
+    }
+    const nodesByDepth = {};
+    for (const nodeID in nodesDepths) {
+      const depth = nodesDepths[nodeID];
+      if (!(depth in nodesByDepth)) {
+        nodesByDepth[depth] = [];
+      }
+      nodesByDepth[depth].push(nodeIDToNode[nodeID]);
+    }
+    const layersByDepth = {};
+    for (const layerID in layersDepths) {
+      const depth = layersDepths[layerID];
+      if (!(depth in layersByDepth)) {
+        layersByDepth[depth] = [];
+      }
+      layersByDepth[depth].push(layerIDToLayer[layerID]);
+    }
+    let depthKeys = Object.keys(layersByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    this.layers = [];
+    for (const depth of depthKeys) {
+      const layersForDepth = layersByDepth[depth];
+      layersForDepth.sort((a, b) => {
+        const aIndex = layerIndices[a.id];
+        const bIndex = layerIndices[b.id];
+        if (aIndex < bIndex) {
+          return -1;
+        }
+        if (aIndex > bIndex) {
+          return 1;
+        }
+        return 0;
+      });
+      for (const layer of layersForDepth) {
+        if (layer instanceof Container) {
+          this.internalContainerRefs.push(layer);
+        }
+        this.layers.push(layer);
+      }
+    }
+    this.layersByDepth = layersByDepth;
+    depthKeys = Object.keys(nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    const computableTensors = this.inputs.slice();
+    const layersWithCompleteInput = [];
+    for (const depth of depthKeys) {
+      for (const node of nodesByDepth[depth]) {
+        const layer = node.outboundLayer;
+        if (layer != null) {
+          for (const x of node.inputTensors) {
+            if (computableTensors.indexOf(x) === -1) {
+              throw new RuntimeError(`Graph disconnected: cannot obtain value for tensor ${x} at layer "${layer.name}". The following previous layers were accessed without issue: ${layersWithCompleteInput}`);
+            }
+          }
+          for (const x of node.outputTensors) {
+            computableTensors.push(x);
+          }
+          layersWithCompleteInput.push(layer.name);
+        }
+      }
+    }
+    this.nodesByDepth = nodesByDepth;
+    const allNames = this.layers.map((x) => x.name);
+    for (const name of allNames) {
+      const numOccurrences = allNames.filter((x) => x === name).length;
+      if (numOccurrences !== 1) {
+        throw new RuntimeError(`The name "${name}" is used ${numOccurrences} times in the model. All layer names should be unique. Layer names: ` + JSON.stringify(allNames));
+      }
+    }
+    this.outboundNodes = [];
+    this.inboundNodes = [];
+    new Node({
+      outboundLayer: this,
+      inboundLayers: [],
+      nodeIndices: [],
+      tensorIndices: [],
+      inputTensors: this.inputs,
+      outputTensors: this.outputs,
+      inputMasks: this.inputs.map((x) => null),
+      outputMasks: this.outputs.map((x) => null),
+      inputShapes: this.inputs.map((x) => x.shape),
+      outputShapes: this.outputs.map((x) => x.shape)
+    });
+    this.built = true;
+    this._refCount = 1;
+  }
+  assertNotDisposed() {
+    if (this._refCount === 0) {
+      throw new Error(`Container '${this.name}' is already disposed.`);
+    }
+  }
+  dispose() {
+    this.assertNotDisposed();
+    const result = { refCountAfterDispose: null, numDisposedVariables: 0 };
+    if (--this._refCount === 0) {
+      for (const layer of this.layers) {
+        result.numDisposedVariables += layer.dispose().numDisposedVariables;
+      }
+      for (const container of this.internalContainerRefs) {
+        result.numDisposedVariables += container.dispose().numDisposedVariables;
+      }
+    }
+    result.refCountAfterDispose = this._refCount;
+    return result;
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this.layers.forEach((layer) => {
+      layer._trainableWeights.forEach((w) => w.trainable = trainable);
+    });
+    this.trainable_ = trainable;
+  }
+  get trainableWeights() {
+    if (this._trainableWeights.length > 0) {
+      throw new ValueError("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");
+    }
+    if (!this.trainable) {
+      return [];
+    }
+    let weights = [];
+    for (const layer of this.layers) {
+      weights = weights.concat(layer.trainableWeights);
+    }
+    return weights;
+  }
+  get nonTrainableWeights() {
+    const weights = [];
+    for (const layer of this.layers) {
+      weights.push(...layer.nonTrainableWeights);
+    }
+    if (!this.trainable) {
+      const trainableWeights = [];
+      for (const layer of this.layers) {
+        trainableWeights.push(...layer.trainableWeights);
+      }
+      return trainableWeights.concat(weights);
+    }
+    return weights;
+  }
+  get weights() {
+    return this.trainableWeights.concat(this.nonTrainableWeights);
+  }
+  loadWeights(weights, strict = true) {
+    const nameToWeight = {};
+    let totalWeightsCount = 0;
+    for (const layer of this.layers) {
+      for (const weight of layer.weights) {
+        if (nameToWeight[weight.originalName] != null) {
+          throw new ValueError(`Duplicate weight name: ${weight.originalName}`);
+        }
+        nameToWeight[weight.originalName] = weight;
+        totalWeightsCount++;
+      }
+    }
+    const weightValueTuples = [];
+    for (const name in weights) {
+      let validatedName = name;
+      if (nameToWeight[name] == null) {
+        const tokens = name.split("/");
+        const shortenNameArray = tokens.slice(0, -2).concat([tokens[tokens.length - 1]]);
+        validatedName = shortenNameArray.join("/");
+      }
+      if (nameToWeight[validatedName] != null) {
+        weightValueTuples.push([nameToWeight[validatedName], weights[name]]);
+      } else if (strict) {
+        throw new ValueError(`Provided weight data has no target variable: ${name}`);
+      }
+      delete nameToWeight[validatedName];
+    }
+    if (strict) {
+      const unsetNames = [];
+      for (const name in nameToWeight) {
+        unsetNames.push(name);
+      }
+      if (unsetNames.length > 0) {
+        throw new ValueError(`${unsetNames.length} of ${totalWeightsCount} weights are not set: ${unsetNames}`);
+      }
+    }
+    batchSetValue(weightValueTuples);
+  }
+  updatedConfig() {
+    const theConfig = this.getConfig();
+    const modelConfig = {};
+    modelConfig["className"] = this.getClassName();
+    modelConfig["config"] = theConfig;
+    modelConfig["kerasVersion"] = `tfjs-layers ${version2}`;
+    modelConfig["backend"] = "TensorFlow.js";
+    return modelConfig;
+  }
+  toJSON(unused, returnString = true) {
+    const modelConfig = convertTsToPythonic(this.updatedConfig());
+    return returnString ? JSON.stringify(modelConfig) : modelConfig;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = toList(inputs);
+      const feedDict = new FeedDict();
+      for (let i = 0; i < this.inputs.length; ++i) {
+        feedDict.add(this.inputs[i], inputs[i]);
+      }
+      return execute(this.outputs, feedDict, kwargs);
+    });
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      inputs = toList(inputs);
+      let masks;
+      if (mask == null) {
+        masks = pyListRepeat(null, inputs.length);
+      } else {
+        masks = toList(mask);
+      }
+      return this.runInternalGraph(inputs, masks)[1];
+    });
+  }
+  computeOutputShape(inputShape) {
+    const inputShapes = normalizeShapeList(inputShape);
+    if (inputShapes.length !== this.inputLayers.length) {
+      throw new ValueError(`Invalid inputShape argument ${inputShape}: model has ${this.inputLayers.length} tensor inputs.`);
+    }
+    const layersToOutputShapes = {};
+    for (let i = 0; i < inputShapes.length; i++) {
+      const layer = this.inputLayers[i];
+      const inputShape2 = inputShapes[i];
+      const shapeKey = layer.name + "_0_0";
+      layersToOutputShapes[shapeKey] = inputShape2;
+    }
+    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    if (depthKeys.length > 1) {
+      for (const depth of depthKeys) {
+        const nodes = this.nodesByDepth[depth];
+        for (const node of nodes) {
+          const layer = node.outboundLayer;
+          if (this.inputLayers.map((x) => x.id).indexOf(layer.id) !== -1) {
+            continue;
+          }
+          const inputShapes2 = [];
+          for (let j = 0; j < node.inboundLayers.length; j++) {
+            const inboundLayer = node.inboundLayers[j];
+            const nodeIndex2 = node.nodeIndices[j];
+            const tensorIndex = node.tensorIndices[j];
+            const shapeKey = `${inboundLayer.name}_${nodeIndex2}_${tensorIndex}`;
+            const inputShape2 = layersToOutputShapes[shapeKey];
+            inputShapes2.push(inputShape2);
+          }
+          const outputShape = layer.computeOutputShape(singletonOrArray(inputShapes2));
+          const outputShapes2 = normalizeShapeList(outputShape);
+          const nodeIndex = layer.inboundNodes.indexOf(node);
+          for (let j = 0; j < outputShapes2.length; j++) {
+            const shapeKey = `${layer.name}_${nodeIndex}_${j}`;
+            layersToOutputShapes[shapeKey] = outputShapes2[j];
+          }
+        }
+      }
+    }
+    const outputShapes = [];
+    const outputShapeKeys = [];
+    for (let i = 0; i < this.outputLayers.length; i++) {
+      const layer = this.outputLayers[i];
+      const nodeIndex = this.outputLayersNodeIndices[i];
+      const tensorIndex = this.outputLayersTensorIndices[i];
+      const shapeKey = `${layer.name}_${nodeIndex}_${tensorIndex}`;
+      outputShapeKeys.push(shapeKey);
+    }
+    for (let i = 0; i < outputShapeKeys.length; i++) {
+      const key = outputShapeKeys[i];
+      assert2(key in layersToOutputShapes);
+      outputShapes.push(layersToOutputShapes[key]);
+    }
+    return singletonOrArray(outputShapes);
+  }
+  runInternalGraph(inputs, masks) {
+    if (masks == null) {
+      masks = pyListRepeat(null, inputs.length);
+    }
+    const tensorMap = {};
+    for (let i = 0; i < this.inputs.length; ++i) {
+      const x = this.inputs[i];
+      const y = inputs[i];
+      const mask = masks[i];
+      tensorMap[x.id] = [y, mask];
+    }
+    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    for (const depth of depthKeys) {
+      const nodes = this.nodesByDepth[depth];
+      for (const node of nodes) {
+        const layer = node.outboundLayer;
+        const referenceInputTensors = node.inputTensors;
+        const referenceOutputTensors = node.outputTensors;
+        const computedData = new Array();
+        for (const x of referenceInputTensors) {
+          if (x.id in tensorMap) {
+            computedData.push(tensorMap[x.id]);
+          }
+        }
+        if (computedData.length === referenceInputTensors.length) {
+          let kwargs = {};
+          let computedTensors;
+          let computedMasks;
+          let outputTensors2;
+          let outputMasks2;
+          if (node.callArgs != null) {
+            kwargs = node.callArgs;
+          }
+          if (computedData.length === 1) {
+            const [computedTensor, computedMask] = computedData[0];
+            if (kwargs["mask"] == null) {
+              kwargs["mask"] = computedMask;
+            }
+            outputTensors2 = toList(layer.call(computedTensor, kwargs));
+            outputMasks2 = toList(layer.computeMask(computedTensor, computedMask));
+            computedTensors = [computedTensor];
+            computedMasks = [computedMask];
+          } else {
+            computedTensors = computedData.map((x) => x[0]);
+            computedMasks = computedData.map((x) => x[1]);
+            if (kwargs["mask"] == null) {
+              kwargs["mask"] = computedMasks;
+            }
+            outputTensors2 = toList(layer.call(computedTensors, kwargs));
+            outputMasks2 = toList(layer.computeMask(computedTensors, computedMasks));
+          }
+          if (layer.activityRegularizer) {
+            throw new NotImplementedError("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");
+          }
+          for (let i = 0; i < referenceOutputTensors.length; ++i) {
+            const x = referenceOutputTensors[i];
+            const y = outputTensors2[i];
+            const mask = outputMasks2[i];
+            tensorMap[x.id] = [y, mask];
+          }
+        }
+      }
+    }
+    const outputTensors = [];
+    const outputMasks = [];
+    const outputShapes = [];
+    for (const x of this.outputs) {
+      assert2(x.id in tensorMap, `Could not compute output ${x.name} : ${x.id}`);
+      const [tensor2, mask] = tensorMap[x.id];
+      outputShapes.push(tensor2.shape);
+      outputTensors.push(tensor2);
+      outputMasks.push(mask);
+    }
+    return [outputTensors, outputMasks, outputShapes];
+  }
+  buildNodeConversionMap(layers) {
+    const nodeConversionMap = {};
+    let keptNodes;
+    for (const layer of this.layers) {
+      keptNodes = layer instanceof Container ? 1 : 0;
+      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {
+        const nodeKey = Container.nodeKey(layer, originalNodeIndex);
+        if (this.containerNodes.has(nodeKey)) {
+          nodeConversionMap[nodeKey] = keptNodes;
+          keptNodes += 1;
+        }
+      }
+    }
+    return nodeConversionMap;
+  }
+  getLayer(name, index) {
+    if (index != null) {
+      if (this.layers.length <= index) {
+        throw new ValueError(`Was asked to retrieve layer at index ${index}, but model only has ${this.layers.length} layer(s).`);
+      } else {
+        return this.layers[index];
+      }
+    } else {
+      if (name == null) {
+        throw new ValueError("Provide either a layer name or layer index");
+      }
+    }
+    for (const layer of this.layers) {
+      if (layer.name === name) {
+        return layer;
+      }
+    }
+    throw new ValueError(`No such layer: ${name}`);
+  }
+  calculateLosses() {
+    return tidy(() => {
+      const losses2 = [];
+      for (const layer of this.layers) {
+        for (let nodeIndex = 0; nodeIndex < layer.inboundNodes.length; ++nodeIndex) {
+          const nodeKey = Container.nodeKey(layer, nodeIndex);
+          if (this.containerNodes.has(nodeKey)) {
+            losses2.push(...layer.calculateLosses());
+          }
+        }
+      }
+      return losses2;
+    });
+  }
+  getConfig() {
+    const config = { name: this.name };
+    const nodeConversionMap = this.buildNodeConversionMap(this.layers);
+    const layerConfigs = [];
+    for (const layer of this.layers) {
+      const layerClassName = layer.getClassName();
+      const layerConfig = layer.getConfig();
+      const filteredInboundNodes = [];
+      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {
+        const node = layer.inboundNodes[originalNodeIndex];
+        const nodeKey = Container.nodeKey(layer, originalNodeIndex);
+        let kwargs = {};
+        if (this.containerNodes.has(nodeKey)) {
+          if (node.callArgs) {
+            try {
+              JSON.stringify(node.callArgs);
+              kwargs = node.callArgs;
+            } catch (err) {
+              console.warn(`Layer ${layer.name} was passed non-serializable keyword arguments: ${node.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`);
+              kwargs = {};
+            }
+          }
+          if (node.inboundLayers.length > 0) {
+            const nodeData = [];
+            for (let i = 0; i < node.inboundLayers.length; i++) {
+              const inboundLayer = node.inboundLayers[i];
+              const nodeIndex = node.nodeIndices[i];
+              const tensorIndex = node.tensorIndices[i];
+              const nodeKey2 = Container.nodeKey(inboundLayer, nodeIndex);
+              let newNodeIndex = nodeConversionMap[nodeKey2];
+              if (newNodeIndex == null) {
+                newNodeIndex = 0;
+              }
+              nodeData.push([inboundLayer.name, newNodeIndex, tensorIndex, kwargs]);
+            }
+            filteredInboundNodes.push(nodeData);
+          }
+        }
+      }
+      const dict = {};
+      dict["name"] = layer.name;
+      dict["className"] = layerClassName;
+      dict["config"] = layerConfig;
+      dict["inboundNodes"] = filteredInboundNodes;
+      layerConfigs.push(dict);
+    }
+    config["layers"] = layerConfigs;
+    const modelInputs = [];
+    for (let i = 0; i < this.inputLayers.length; i++) {
+      const layer = this.inputLayers[i];
+      const nodeIndex = this.inputLayersNodeIndices[i];
+      const nodeKey = Container.nodeKey(layer, nodeIndex);
+      if (!this.containerNodes.has(nodeKey)) {
+        continue;
+      }
+      let newNodeIndex = nodeConversionMap[nodeKey];
+      if (newNodeIndex === null || newNodeIndex === void 0) {
+        newNodeIndex = 0;
+      }
+      const tensorIndex = this.inputLayersTensorIndices[i];
+      modelInputs.push([layer.name, newNodeIndex, tensorIndex]);
+    }
+    config["inputLayers"] = modelInputs;
+    const modelOutputs = [];
+    for (let i = 0; i < this.outputLayers.length; i++) {
+      const layer = this.outputLayers[i];
+      const nodeIndex = this.outputLayersNodeIndices[i];
+      const nodeKey = Container.nodeKey(layer, nodeIndex);
+      if (!this.containerNodes.has(nodeKey)) {
+        continue;
+      }
+      let newNodeIndex = nodeConversionMap[nodeKey];
+      if (newNodeIndex === null || newNodeIndex === void 0) {
+        newNodeIndex = 0;
+      }
+      const tensorIndex = this.outputLayersTensorIndices[i];
+      modelOutputs.push([layer.name, newNodeIndex, tensorIndex]);
+    }
+    config["outputLayers"] = modelOutputs;
+    return config;
+  }
+  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {
+    const createdLayers = {};
+    const unprocessedNodes = {};
+    function addUnprocessedNode(layer, nodeData) {
+      if (!(layer.name in unprocessedNodes)) {
+        unprocessedNodes[layer.name] = [nodeData];
+      } else {
+        unprocessedNodes[layer.name].push(nodeData);
+      }
+    }
+    function processNode(layer, nodeData) {
+      const inputTensors2 = [];
+      let kwargs;
+      for (const inputData of nodeData) {
+        const inboundLayerName = inputData[0];
+        const inboundNodeIndex = inputData[1];
+        const inboundTensorIndex = inputData[2];
+        kwargs = inputData[3] == null ? {} : inputData[3];
+        if (!(inboundLayerName in createdLayers)) {
+          addUnprocessedNode(layer, nodeData);
+          return;
+        }
+        const inboundLayer = createdLayers[inboundLayerName];
+        if (inboundLayer.inboundNodes.length <= inboundNodeIndex) {
+          addUnprocessedNode(layer, nodeData);
+          return;
+        }
+        const inboundNode = inboundLayer.inboundNodes[inboundNodeIndex];
+        inputTensors2.push(inboundNode.outputTensors[inboundTensorIndex]);
+      }
+      if (inputTensors2.length > 0) {
+        layer.apply(singletonOrArray(inputTensors2), kwargs);
+      }
+    }
+    function processLayer(layerData) {
+      const layerName = layerData["name"];
+      const layer = deserialize(layerData, config["customObjects"] != null ? config["customObjects"] : {});
+      layer.setFastWeightInitDuringBuild(fastWeightInit);
+      createdLayers[layerName] = layer;
+      const inboundNodesData = layerData["inboundNodes"];
+      inboundNodesData.forEach((nodeData) => {
+        if (!(nodeData instanceof Array)) {
+          throw new ValueError(`Corrupted configuration, expected array for nodeData: ${nodeData}`);
+        }
+        addUnprocessedNode(layer, nodeData);
+      });
+    }
+    const name = config["name"];
+    const layersFromConfig = config["layers"];
+    for (const layerData of layersFromConfig) {
+      processLayer(layerData);
+    }
+    while (!isObjectEmpty(unprocessedNodes)) {
+      for (const layerData of layersFromConfig) {
+        const layer = createdLayers[layerData["name"]];
+        if (layer.name in unprocessedNodes) {
+          const currentUnprocessedNodesForLayer = unprocessedNodes[layer.name];
+          delete unprocessedNodes[layer.name];
+          for (const nodeData of currentUnprocessedNodesForLayer) {
+            processNode(layer, nodeData);
+          }
+        }
+      }
+    }
+    const inputTensors = [];
+    const outputTensors = [];
+    const inputLayersFromConfig = config["inputLayers"];
+    for (const layerData of inputLayersFromConfig) {
+      const layerName = layerData[0];
+      const nodeIndex = layerData[1];
+      const tensorIndex = layerData[2];
+      assert2(layerName in createdLayers);
+      const layer = createdLayers[layerName];
+      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;
+      inputTensors.push(layerOutputTensors[tensorIndex]);
+    }
+    const outputLayersFromConfig = config["outputLayers"];
+    for (const layerData of outputLayersFromConfig) {
+      const layerName = layerData[0];
+      const nodeIndex = layerData[1];
+      const tensorIndex = layerData[2];
+      assert2(layerName in createdLayers);
+      const layer = createdLayers[layerName];
+      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;
+      outputTensors.push(layerOutputTensors[tensorIndex]);
+    }
+    return new cls({ inputs: inputTensors, outputs: outputTensors, name });
+  }
+  get stateful() {
+    if (this._stateful) {
+      throw new ValueError("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");
+    }
+    for (const layer of this.layers) {
+      if (layer.stateful) {
+        return true;
+      }
+    }
+    return false;
+  }
+  resetStates() {
+    tidy(() => {
+      this.layers.forEach((layer) => {
+        if (layer.stateful) {
+          layer.resetStates();
+        }
+      });
+    });
+  }
+};
+function standardizeSampleOrClassWeights(xWeight, outputNames, weightType) {
+  const numOutputs = outputNames.length;
+  if (xWeight == null || Array.isArray(xWeight) && xWeight.length === 0) {
+    return outputNames.map((name) => null);
+  }
+  if (numOutputs === 1) {
+    if (Array.isArray(xWeight) && xWeight.length === 1) {
+      return xWeight;
+    } else if (typeof xWeight === "object" && outputNames[0] in xWeight) {
+      return [xWeight[outputNames[0]]];
+    } else {
+      return [xWeight];
+    }
+  }
+  if (Array.isArray(xWeight)) {
+    if (xWeight.length !== numOutputs) {
+      throw new Error(`Provided ${weightType} is an array of ${xWeight.length} element(s), but the model has ${numOutputs} outputs. Make sure a set of weights is provided for each model output.`);
+    }
+    return xWeight;
+  } else if (typeof xWeight === "object" && Object.keys(xWeight).length > 0 && typeof xWeight[Object.keys(xWeight)[0]] === "object") {
+    const output = [];
+    outputNames.forEach((outputName) => {
+      if (outputName in xWeight) {
+        output.push(xWeight[outputName]);
+      } else {
+        output.push(null);
+      }
+    });
+    return output;
+  } else {
+    throw new Error(`The model has multiple (${numOutputs}) outputs, so ${weightType} must be either an array with ${numOutputs} elements or an object with ${outputNames} keys. Provided ${weightType} not understood: ${JSON.stringify(xWeight)}`);
+  }
+}
+function standardizeClassWeights(classWeight, outputNames) {
+  return standardizeSampleOrClassWeights(classWeight, outputNames, "classWeight");
+}
+async function standardizeWeights(y, sampleWeight, classWeight, sampleWeightMode) {
+  if (sampleWeight != null || sampleWeightMode != null) {
+    throw new Error("Support sampleWeight is not implemented yet");
+  }
+  if (classWeight != null) {
+    const yClasses = tidy(() => {
+      if (y.shape.length === 1) {
+        return clone(y);
+      } else if (y.shape.length === 2) {
+        if (y.shape[1] > 1) {
+          const axis = 1;
+          return argMax(y, axis);
+        } else if (y.shape[1] === 1) {
+          return reshape(y, [y.shape[0]]);
+        } else {
+          throw new Error(`Encountered unexpected last-dimension size (${y.shape[1]}) during handling of class weights. The size is expected to be >= 1.`);
+        }
+      } else {
+        throw new Error(`Unexpected rank of target (y) tensor (${y.rank}) during handling of class weights. The rank is expected to be 1 or 2.`);
+      }
+    });
+    const yClassIndices = Array.from(await yClasses.data());
+    dispose(yClasses);
+    const classSampleWeight = [];
+    yClassIndices.forEach((classIndex) => {
+      if (classWeight[classIndex] == null) {
+        throw new Error(`classWeight must contain all classes in the training data. The class ${classIndex} exists in the data but not in classWeight`);
+      } else {
+        classSampleWeight.push(classWeight[classIndex]);
+      }
+    });
+    return tensor1d(classSampleWeight, "float32");
+  } else {
+    return null;
+  }
+}
+function computeWeightedLoss2(losses2, sampleWeights) {
+  return mul(losses2, sampleWeights);
+}
+var DEFAULT_VALIDATION_BATCH_SIZE = 32;
+function standardizeDataIteratorOutput(model2, iteratorOut) {
+  let xs;
+  let ys;
+  const iteratorOutObj = iteratorOut;
+  xs = iteratorOutObj["xs"];
+  ys = iteratorOutObj["ys"];
+  util_exports.assert(xs != null && ys != null, () => `A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${iteratorOut}`);
+  const flattenedXs = flattenTensorOrArrayOrMap("input", model2.inputNames, xs);
+  const flattenedYs = flattenTensorOrArrayOrMap("output", model2.outputNames, ys);
+  const batchSize = flattenedXs[0].shape[0];
+  util_exports.assert(flattenedXs.length === model2.inputs.length, () => `LayersModel has ${model2.inputs.length} inputs, but the dataset provides ${flattenedXs.length} inputs.  (Expected input keys: ${JSON.stringify(model2.inputNames)})`);
+  util_exports.assert(flattenedYs.length === model2.outputs.length, () => `LayersModel has ${model2.outputs.length} outputs, but the dataset provides ${flattenedYs.length} outputs.  (Expected output keys: ${JSON.stringify(model2.outputNames)})`);
+  for (let xIndex = 0; xIndex < flattenedXs.length; xIndex++) {
+    util_exports.assert(flattenedXs[xIndex].shape[0] === batchSize, () => `Batch size mismatch: input ${model2.inputNames[xIndex]} has ${flattenedXs[xIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);
+  }
+  for (let yIndex = 0; yIndex < flattenedYs.length; yIndex++) {
+    util_exports.assert(flattenedYs[yIndex].shape[0] === batchSize, () => `Batch size mismatch: output ${model2.outputNames[yIndex]} has ${flattenedYs[yIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);
+  }
+  return { xs: flattenedXs, ys: flattenedYs };
+}
+function flattenTensorOrArrayOrMap(inputOrOutput, names, values) {
+  if (values instanceof Tensor) {
+    return [values];
+  } else if (Array.isArray(values)) {
+    util_exports.assert(values.length === names.length, () => `Received an array of ${values.length} Tensors, but expected ${names.length} to match the ${inputOrOutput} keys ${names}.`);
+    return values;
+  } else {
+    const result = [];
+    for (const name of names) {
+      if (values[name] == null) {
+        throw new ValueError(`The feature data generated by the dataset lacks the required ${inputOrOutput} key '${name}'.`);
+      }
+      result.push(values[name]);
+    }
+    return result;
+  }
+}
+function standardizeTensorValidationData(data) {
+  if (data.length === 3) {
+    throw new NotImplementedError("Validation with sample weights is not implemented yet.");
+  }
+  return { xs: data[0], ys: data[1] };
+}
+async function fitDataset(model2, dataset, args) {
+  const hasBatchesPerEpoch = args.batchesPerEpoch != null;
+  util_exports.assert(model2.optimizer != null, () => "You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig).");
+  util_exports.assert(args != null, () => `For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call.`);
+  util_exports.assert(args.epochs != null && args.epochs > 0 && Number.isInteger(args.epochs), () => `For fitDataset(), config.epochs is expected to be a positive integer, but got ${args.epochs}`);
+  util_exports.assert(!hasBatchesPerEpoch || args.batchesPerEpoch > 0 && Number.isInteger(args.batchesPerEpoch), () => `For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${args.batchesPerEpoch}`);
+  util_exports.assert(
+    args["validationSplit"] == null,
+    () => "`validationSplit` is not supported by `fitDataset()`. Use validationData instead."
+  );
+  if (model2.isTraining) {
+    throw new Error("Cannot start training because another fit() call is ongoing.");
+  }
+  model2.isTraining = true;
+  try {
+    const doValidation = args.validationData != null;
+    let valXs;
+    let valYs;
+    if (doValidation) {
+      if (isDatasetObject(args.validationData)) {
+        util_exports.assert(args.validationBatches == null || args.validationBatches > 0 && Number.isInteger(args.validationBatches), () => `For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${args.validationBatches}`);
+      } else {
+        const validationData = standardizeTensorValidationData(args.validationData);
+        valXs = validationData.xs;
+        valYs = validationData.ys;
+      }
+    }
+    const trainFunction = model2.makeTrainFunction();
+    const outLabels = model2.getDedupedMetricsNames();
+    let callbackMetrics;
+    if (doValidation) {
+      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => "val_" + n));
+    } else {
+      callbackMetrics = outLabels.slice();
+    }
+    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);
+    const verbose = args.verbose == null ? 1 : args.verbose;
+    const { callbackList, history } = configureCallbacks(
+      callbacks2,
+      verbose,
+      args.epochs,
+      null,
+      null,
+      getStepsPerEpoch(dataset, args),
+      null,
+      doValidation,
+      callbackMetrics
+    );
+    callbackList.setModel(model2);
+    model2.history = history;
+    await callbackList.onTrainBegin();
+    model2.stopTraining_ = false;
+    let epoch = args.initialEpoch == null ? 0 : args.initialEpoch;
+    let dataIterator = await dataset.iterator();
+    while (epoch < args.epochs) {
+      const epochLogs = {};
+      await callbackList.onEpochBegin(epoch);
+      let stepsDone = 0;
+      let batchIndex = 0;
+      if (!hasBatchesPerEpoch) {
+        dataIterator = await dataset.iterator();
+      }
+      while (hasBatchesPerEpoch ? stepsDone < args.batchesPerEpoch : true) {
+        const iteratorOut = await dataIterator.next();
+        if (hasBatchesPerEpoch && iteratorOut.done) {
+          console.warn(`You provided \`batchesPerEpoch\` as ${args.batchesPerEpoch}, but your dataset iterator ran out of data after ${stepsDone} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${args.batchesPerEpoch * args.epochs} batches). You may need to use the repeat() function when building your dataset.`);
+          break;
+        }
+        if (iteratorOut.value != null) {
+          const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);
+          const batchLogs = {};
+          batchLogs["batch"] = batchIndex;
+          batchLogs["size"] = xs[0].shape[0];
+          await callbackList.onBatchBegin(batchIndex, batchLogs);
+          const sampleWeights = [];
+          if (args.classWeight != null) {
+            const standardClassWeights = standardizeClassWeights(args.classWeight, model2.outputNames);
+            for (let i = 0; i < standardClassWeights.length; ++i) {
+              sampleWeights.push(await standardizeWeights(ys[i], null, standardClassWeights[i]));
+            }
+          }
+          const ins = xs.concat(ys).concat(sampleWeights);
+          const outs = trainFunction(ins);
+          dispose(ins);
+          for (let i = 0; i < outLabels.length; ++i) {
+            const label = outLabels[i];
+            const out = outs[i];
+            batchLogs[label] = out;
+            keep(out);
+          }
+          await callbackList.onBatchEnd(batchIndex, batchLogs);
+          disposeTensorsInLogs(batchLogs);
+          batchIndex++;
+          stepsDone++;
+        }
+        if (hasBatchesPerEpoch ? stepsDone >= args.batchesPerEpoch : iteratorOut.done) {
+          if (doValidation) {
+            let valOuts;
+            if (isDatasetObject(args.validationData)) {
+              valOuts = toList(await model2.evaluateDataset(args.validationData, { batches: args.validationBatches }));
+            } else {
+              valOuts = toList(model2.evaluate(valXs, valYs, {
+                batchSize: args.validationBatchSize == null ? DEFAULT_VALIDATION_BATCH_SIZE : args.validationBatchSize,
+                verbose: 0
+              }));
+            }
+            for (let i = 0; i < model2.metricsNames.length; ++i) {
+              epochLogs[`val_${model2.metricsNames[i]}`] = valOuts[i];
+            }
+          }
+          break;
+        }
+        if (model2.stopTraining_) {
+          break;
+        }
+      }
+      await callbackList.onEpochEnd(epoch, epochLogs);
+      epoch++;
+      if (model2.stopTraining_) {
+        break;
+      }
+    }
+    await callbackList.onTrainEnd();
+    await model2.history.syncData();
+    return model2.history;
+  } finally {
+    model2.isTraining = false;
+  }
+}
+function getStepsPerEpoch(dataset, args) {
+  let stepsPerEpoch = null;
+  if (args.batchesPerEpoch != null) {
+    stepsPerEpoch = args.batchesPerEpoch;
+  } else if (Number.isFinite(dataset.size)) {
+    stepsPerEpoch = dataset.size;
+  }
+  return stepsPerEpoch;
+}
+function isDatasetObject(dataset) {
+  return typeof dataset.iterator === "function";
+}
+function isLazyIteratorObject(iterator) {
+  return typeof iterator.next === "function";
+}
+async function evaluateDataset(model2, dataset, args) {
+  args = args || {};
+  const hasBatches = args.batches != null;
+  const f = model2.testFunction;
+  let outs = [];
+  if (args.verbose > 0) {
+    throw new NotImplementedError("Verbose mode is not implemented yet.");
+  }
+  util_exports.assert(!hasBatches || args.batches > 0 && Number.isInteger(args.batches), () => `Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(args.batches)}`);
+  const dataIterator = isLazyIteratorObject(dataset) ? dataset : await dataset.iterator();
+  let numExamples = 0;
+  let batch = 0;
+  while (hasBatches ? batch < args.batches : true) {
+    const iteratorOut = await dataIterator.next();
+    outs = tidy(() => {
+      if (iteratorOut.value) {
+        const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);
+        const xsAndYs = xs.concat(ys);
+        const batchOuts = tidy(() => f(xsAndYs));
+        dispose(xsAndYs);
+        if (batch === 0) {
+          for (let i = 0; i < batchOuts.length; ++i) {
+            outs.push(scalar(0));
+          }
+        }
+        const batchSize = xsAndYs[0].shape[0];
+        for (let i = 0; i < batchOuts.length; ++i) {
+          const batchOut = batchOuts[i];
+          const oldScalar = outs[i];
+          outs[i] = tidy(() => add2(outs[i], mul(batchSize, batchOut)));
+          if (batch > 0) {
+            dispose(oldScalar);
+          }
+        }
+        dispose(batchOuts);
+        numExamples += batchSize;
+        ++batch;
+      }
+      return outs;
+    });
+    if (iteratorOut.done) {
+      if (hasBatches) {
+        console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${args.batches} batches). You may need to use the repeat() function when building your dataset.`);
+      }
+      break;
+    }
+  }
+  for (let i = 0; i < outs.length; ++i) {
+    const oldScalar = outs[i];
+    outs[i] = div(outs[i], numExamples);
+    dispose(oldScalar);
+  }
+  return singletonOrArray(outs);
+}
+function checkBatchSize(batchSize) {
+  util_exports.assert(batchSize > 0 && Number.isInteger(batchSize), () => `batchSize is required to be a positive integer, but got ${batchSize}`);
+}
+function sliceArrays(arrays, start, stop) {
+  if (arrays == null) {
+    return [null];
+  } else if (Array.isArray(arrays)) {
+    return arrays.map((array2) => sliceAlongFirstAxis(array2, start, stop - start));
+  } else {
+    return sliceAlongFirstAxis(arrays, start, stop - start);
+  }
+}
+function sliceArraysByIndices(arrays, indices) {
+  return tidy(() => {
+    if (arrays == null) {
+      return null;
+    } else if (Array.isArray(arrays)) {
+      return arrays.map((array2) => sliceArraysByIndices(array2, indices));
+    } else {
+      return gather2(arrays, indices.dtype === "int32" ? indices : cast(indices, "int32"));
+    }
+  });
+}
+function makeBatches(size, batchSize) {
+  const output = [];
+  let batchStart = 0;
+  let batchEnd = null;
+  while (batchStart < size) {
+    batchEnd = batchStart + batchSize;
+    if (batchEnd >= size) {
+      batchEnd = size;
+    }
+    output.push([batchStart, batchEnd]);
+    batchStart = batchEnd;
+  }
+  return output;
+}
+async function fitLoop(model2, f, ins, outLabels, batchSize, epochs, verbose, callbacks2, valF, valIns, shuffle2, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps) {
+  if (batchSize == null) {
+    batchSize = 32;
+  }
+  if (epochs == null) {
+    epochs = 1;
+  }
+  if (shuffle2 == null) {
+    shuffle2 = true;
+  }
+  if (initialEpoch == null) {
+    initialEpoch = 0;
+  }
+  let doValidation = false;
+  if (valF != null && valIns != null) {
+    doValidation = true;
+  }
+  if (validationSteps != null) {
+    doValidation = true;
+    if (stepsPerEpoch == null) {
+      throw new ValueError("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");
+    }
+  }
+  const numTrainSamples = model2.checkNumSamples(ins, batchSize, stepsPerEpoch, "steps_per_epoch");
+  let indexArray;
+  if (numTrainSamples != null) {
+    indexArray = range2(0, numTrainSamples);
+  }
+  if (verbose == null) {
+    verbose = 1;
+  }
+  const { callbackList, history } = configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics);
+  callbackList.setModel(model2);
+  model2.history = history;
+  await callbackList.onTrainBegin();
+  model2.stopTraining_ = false;
+  for (let epoch = initialEpoch; epoch < epochs; ++epoch) {
+    await callbackList.onEpochBegin(epoch);
+    const epochLogs = {};
+    if (stepsPerEpoch != null) {
+      throw new NotImplementedError("stepsPerEpoch mode is not implemented yet.");
+    } else {
+      if (shuffle2 === "batch") {
+        throw new NotImplementedError("batch shuffling is not implemneted yet");
+      } else if (shuffle2) {
+        util_exports.shuffle(indexArray);
+      }
+      const epochIndexArray1D = tensor1d(indexArray);
+      const batches = makeBatches(numTrainSamples, batchSize);
+      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+        const batchLogs = {};
+        await callbackList.onBatchBegin(batchIndex, batchLogs);
+        tidy(() => {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const batchIds = sliceAlongFirstAxis(epochIndexArray1D, batchStart, batchEnd - batchStart);
+          batchLogs["batch"] = batchIndex;
+          batchLogs["size"] = batchEnd - batchStart;
+          const insBatch = sliceArraysByIndices(ins, batchIds);
+          const outs = f(insBatch);
+          for (let i = 0; i < outLabels.length; ++i) {
+            const label = outLabels[i];
+            const out = outs[i];
+            batchLogs[label] = out;
+            keep(out);
+          }
+          if (batchIndex === batches.length - 1) {
+            if (doValidation) {
+              const valOuts = model2.testLoop(valF, valIns, batchSize);
+              for (let i = 0; i < outLabels.length; ++i) {
+                const label = outLabels[i];
+                const out = valOuts[i];
+                keep(out);
+                epochLogs["val_" + label] = out;
+              }
+            }
+          }
+        });
+        await callbackList.onBatchEnd(batchIndex, batchLogs);
+        disposeTensorsInLogs(batchLogs);
+        if (model2.stopTraining_) {
+          break;
+        }
+      }
+      epochIndexArray1D.dispose();
+    }
+    await callbackList.onEpochEnd(epoch, epochLogs);
+    if (model2.stopTraining_) {
+      break;
+    }
+  }
+  await callbackList.onTrainEnd();
+  await model2.history.syncData();
+  return model2.history;
+}
+async function fitTensors(model2, x, y, args = {}) {
+  if (model2.isTraining) {
+    throw new Error("Cannot start training because another fit() call is ongoing.");
+  }
+  model2.isTraining = true;
+  let inputs;
+  let targets;
+  let originalInputs;
+  let originalTargets;
+  let inputValX;
+  let inputValY;
+  let valX;
+  let valY;
+  let sampleWeights;
+  try {
+    const batchSize = args.batchSize == null ? 32 : args.batchSize;
+    checkBatchSize(batchSize);
+    const checkBatchAxis = false;
+    const standardizedOuts = await model2.standardizeUserData(x, y, args.sampleWeight, args.classWeight, checkBatchAxis, batchSize);
+    inputs = standardizedOuts[0];
+    targets = standardizedOuts[1];
+    sampleWeights = standardizedOuts[2];
+    let doValidation = false;
+    let valIns;
+    if (args.validationData != null && args.validationData.length > 0) {
+      doValidation = true;
+      if (args.validationData.length === 2) {
+        inputValX = args.validationData[0];
+        inputValY = args.validationData[1];
+      } else if (args.validationData.length === 3) {
+        throw new NotImplementedError("validationData including sample weights is not supported yet.");
+      } else {
+        throw new ValueError(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${args.validationData} is invalid.`);
+      }
+      const checkBatchAxis2 = true;
+      const valStandardized = await model2.standardizeUserData(inputValX, inputValY, null, null, checkBatchAxis2, batchSize);
+      valX = valStandardized[0];
+      valY = valStandardized[1];
+      valIns = valX.concat(valY);
+    } else if (args.validationSplit != null && args.validationSplit > 0 && args.validationSplit < 1) {
+      doValidation = true;
+      const splitAt = Math.floor(inputs[0].shape[0] * (1 - args.validationSplit));
+      const originalBatchSize = inputs[0].shape[0];
+      valX = sliceArrays(inputs, splitAt, originalBatchSize);
+      originalInputs = inputs;
+      inputs = sliceArrays(inputs, 0, splitAt);
+      valY = sliceArrays(targets, splitAt, originalBatchSize);
+      originalTargets = targets;
+      targets = sliceArrays(targets, 0, splitAt);
+      valIns = valX.concat(valY);
+    } else if (args.validationSteps != null) {
+      doValidation = true;
+    }
+    const ins = inputs.concat(targets).concat(sampleWeights);
+    model2.checkTrainableWeightsConsistency();
+    const trainFunction = model2.makeTrainFunction();
+    const outLabels = model2.getDedupedMetricsNames();
+    let valFunction;
+    let callbackMetrics;
+    if (doValidation) {
+      model2.makeTestFunction();
+      valFunction = model2.testFunction;
+      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => "val_" + n));
+    } else {
+      valFunction = null;
+      valIns = [];
+      callbackMetrics = outLabels.slice();
+    }
+    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);
+    const out = await fitLoop(model2, trainFunction, ins, outLabels, batchSize, args.epochs, args.verbose, callbacks2, valFunction, valIns, args.shuffle, callbackMetrics, args.initialEpoch, null, null);
+    return out;
+  } finally {
+    model2.isTraining = false;
+    disposeNewTensors(inputs, x);
+    disposeNewTensors(targets, y);
+    disposeNewTensors(originalInputs, x);
+    disposeNewTensors(originalTargets, y);
+    disposeNewTensors(valX, inputValX);
+    disposeNewTensors(valY, inputValY);
+    if (sampleWeights != null) {
+      dispose(sampleWeights);
+    }
+  }
+}
+function ensureTensorsRank2OrHigher(tensors) {
+  const outs = [];
+  if (tensors instanceof Tensor) {
+    tensors = [tensors];
+  }
+  for (let i = 0; i < tensors.length; ++i) {
+    const tensor2 = tensors[i];
+    if (tensor2.rank === 1) {
+      outs.push(expandDims2(tensor2, 1));
+    } else if (tensor2.rank === 0) {
+      throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");
+    } else {
+      outs.push(tensor2);
+    }
+  }
+  return outs;
+}
+function disposeNewTensors(tensors, refTensors) {
+  if (tensors == null) {
+    return;
+  }
+  const oldTensorIds = [];
+  if (refTensors instanceof Tensor) {
+    oldTensorIds.push(refTensors.id);
+  } else if (Array.isArray(refTensors)) {
+    refTensors.forEach((t) => oldTensorIds.push(t.id));
+  } else if (refTensors != null) {
+    for (const name in refTensors) {
+      const oldTensor = refTensors[name];
+      oldTensorIds.push(oldTensor.id);
+    }
+  }
+  const tensorsToDispose = [];
+  if (tensors instanceof Tensor) {
+    if (oldTensorIds.indexOf(tensors.id) === -1) {
+      tensorsToDispose.push(tensors);
+    }
+  } else if (Array.isArray(tensors)) {
+    tensors.forEach((t) => {
+      if (oldTensorIds.indexOf(t.id) === -1) {
+        tensorsToDispose.push(t);
+      }
+    });
+  } else if (tensors != null) {
+    for (const name in tensors) {
+      const tensor2 = tensors[name];
+      if (oldTensorIds.indexOf(tensor2.id) === -1) {
+        tensorsToDispose.push(tensor2);
+      }
+    }
+  }
+  tensorsToDispose.forEach((t) => {
+    if (!t.isDisposed) {
+      t.dispose();
+    }
+  });
+}
+function isDataTensor(x) {
+  return x instanceof Tensor;
+}
+function isDataArray(x) {
+  return Array.isArray(x);
+}
+function isDataDict(x) {
+  return !isDataTensor(x) && !isDataArray(x);
+}
+function standardizeInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = "") {
+  if (names == null || names.length === 0) {
+    if (data != null) {
+      let gotUnexpectedData = false;
+      if (isDataArray(data) && data.length > 0) {
+        gotUnexpectedData = true;
+      } else if (isDataDict(data)) {
+        for (const key in data) {
+          if (data.hasOwnProperty(key)) {
+            gotUnexpectedData = true;
+            break;
+          }
+        }
+      } else {
+        gotUnexpectedData = true;
+      }
+      if (gotUnexpectedData) {
+        throw new ValueError(`Error when checking model ${exceptionPrefix} expected no data, but got ${data}`);
+      }
+    }
+    return [];
+  }
+  if (data == null) {
+    return names.map((name) => null);
+  }
+  let arrays;
+  if (isDataDict(data)) {
+    data = data;
+    arrays = [];
+    for (const name of names) {
+      if (data[name] == null) {
+        throw new ValueError(`No data provided for "${name}". Need data for each key in: ${names}`);
+      }
+      arrays.push(data[name]);
+    }
+  } else if (isDataArray(data)) {
+    data = data;
+    if (data.length !== names.length) {
+      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${names.length} Tensor(s), but instead got the following list of Tensor(s): ${data}`);
+    }
+    arrays = data;
+  } else {
+    data = data;
+    if (names.length > 1) {
+      throw new ValueError(`The model ${exceptionPrefix} expects ${names.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${data.shape}`);
+    }
+    arrays = [data];
+  }
+  arrays = ensureTensorsRank2OrHigher(arrays);
+  if (shapes != null) {
+    for (let i = 0; i < names.length; ++i) {
+      if (shapes[i] == null) {
+        continue;
+      }
+      const array2 = arrays[i];
+      if (array2.shape.length !== shapes[i].length) {
+        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s). but got array with shape ${array2.shape}`);
+      }
+      for (let j = 0; j < shapes[i].length; ++j) {
+        if (j === 0 && !checkBatchAxis) {
+          continue;
+        }
+        const dim = array2.shape[j];
+        const refDim = shapes[i][j];
+        if (refDim != null && refDim >= 0 && dim !== refDim) {
+          throw new ValueError(`${exceptionPrefix} expected a batch of elements where each example has shape [${shapes[i].slice(1, shapes[i].length)}] (i.e.,tensor shape [*,${shapes[i].slice(1, shapes[i].length)}]) but the ${exceptionPrefix} received an input with ${array2.shape[0]} examples, each with shape [${array2.shape.slice(1, array2.shape.length)}] (tensor shape [${array2.shape}])`);
+        }
+      }
+    }
+  }
+  return arrays;
+}
+function checkArrayLengths(inputs, targets, weights) {
+  const setX = unique2(inputs.map((input2) => input2.shape[0]));
+  setX.sort();
+  const setY = unique2(targets.map((target) => target.shape[0]));
+  setY.sort();
+  if (setX.length > 1) {
+    throw new ValueError(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(inputs.map((input2) => input2.shape))}`);
+  }
+  if (setY.length > 1) {
+    throw new ValueError(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(targets.map((target) => target.shape))}`);
+  }
+  if (setX.length > 0 && setY.length > 0 && !util_exports.arraysEqual(setX, setY)) {
+    throw new ValueError(`Input Tensors should have the same number of samples as target Tensors. Found ${setX[0]} input sample(s) and ${setY[0]} target sample(s).`);
+  }
+}
+function checkLossAndTargetCompatibility(targets, lossFns, outputShapes) {
+  const keyLosses = [
+    meanSquaredError2,
+    binaryCrossentropy,
+    categoricalCrossentropy
+  ];
+  for (let i = 0; i < targets.length; ++i) {
+    const y = targets[i];
+    const loss = lossFns[i];
+    const shape = outputShapes[i];
+    if (loss == null) {
+      continue;
+    }
+    if (loss === categoricalCrossentropy) {
+      if (y.shape[y.shape.length - 1] === 1) {
+        throw new ValueError(`You are passing a target array of shape ${y.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);
+      }
+    }
+    if (keyLosses.indexOf(loss) !== -1) {
+      const slicedYShape = y.shape.slice(1);
+      const slicedShape = shape.slice(1);
+      for (let j = 0; j < slicedYShape.length; ++j) {
+        const targetDim = slicedYShape[j];
+        const outDim = slicedShape[j];
+        if (outDim != null && targetDim !== outDim) {
+          throw new ValueError(`A target Tensor with shape ${y.shape} was passed for an output of shape ${shape}, while using a loss function that expects targets to have the same shape as the output.`);
+        }
+      }
+    }
+  }
+}
+function checkInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = "") {
+  let arrays;
+  if (Array.isArray(data)) {
+    if (data.length !== names.length) {
+      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${names.length} Tensor(s), but instead got ${data.length} Tensors(s).`);
+    }
+    arrays = data;
+  } else {
+    if (names.length > 1) {
+      throw new ValueError(`The model expects ${names.length} ${exceptionPrefix} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(data.shape)}.`);
+    }
+    arrays = [data];
+  }
+  if (shapes != null) {
+    for (let i = 0; i < names.length; ++i) {
+      if (shapes[i] == null) {
+        continue;
+      }
+      const array2 = arrays[i];
+      if (array2.shape.length !== shapes[i].length) {
+        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s), but got array with shape ${JSON.stringify(array2.shape)}`);
+      }
+      for (let j = 0; j < shapes[i].length; ++j) {
+        if (j === 0 && !checkBatchAxis) {
+          continue;
+        }
+        const dim = array2.shape[j];
+        const refDim = shapes[i][j];
+        if (refDim != null) {
+          if (refDim !== dim) {
+            throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have shape ${JSON.stringify(shapes[i])} but got array with shape ${JSON.stringify(array2.shape)}.`);
+          }
+        }
+      }
+    }
+  }
+}
+function collectMetrics(metrics, outputNames) {
+  if (metrics == null || Array.isArray(metrics) && metrics.length === 0) {
+    return outputNames.map((name) => []);
+  }
+  let wrappedMetrics;
+  if (typeof metrics === "string" || typeof metrics === "function") {
+    wrappedMetrics = [metrics];
+  } else if (Array.isArray(metrics) || typeof metrics === "object") {
+    wrappedMetrics = metrics;
+  } else {
+    throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${metrics}`);
+  }
+  if (Array.isArray(wrappedMetrics)) {
+    return outputNames.map((name) => wrappedMetrics);
+  } else {
+    const nestedMetrics = [];
+    for (const name of outputNames) {
+      let outputMetrics = wrappedMetrics.hasOwnProperty(name) ? wrappedMetrics[name] : [];
+      if (!Array.isArray(outputMetrics)) {
+        outputMetrics = [outputMetrics];
+      }
+      nestedMetrics.push(outputMetrics);
+    }
+    return nestedMetrics;
+  }
+}
+var LAYERS_MODEL_FORMAT_NAME = "layers-model";
+var LayersModel = class extends Container {
+  constructor(args) {
+    super(args);
+    this.isTraining = false;
+  }
+  summary(lineLength, positions, printFn = console.log) {
+    if (!this.built) {
+      throw new ValueError(`This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).`);
+    }
+    printSummary(this, lineLength, positions, printFn);
+  }
+  compile(args) {
+    if (args.loss == null) {
+      args.loss = [];
+    }
+    this.loss = args.loss;
+    if (typeof args.optimizer === "string") {
+      this.optimizer_ = getOptimizer(args.optimizer);
+      this.isOptimizerOwned = true;
+    } else {
+      if (!(args.optimizer instanceof Optimizer)) {
+        throw new ValueError(`User-defined optimizer must be an instance of tf.Optimizer.`);
+      }
+      this.optimizer_ = args.optimizer;
+      this.isOptimizerOwned = false;
+    }
+    let lossFunctions = [];
+    if (!Array.isArray(args.loss) && typeof args.loss !== "string" && typeof args.loss !== "function") {
+      args.loss = args.loss;
+      for (const name in args.loss) {
+        if (this.outputNames.indexOf(name) === -1) {
+          throw new ValueError(`Unknown entry in loss dictionary: "${name}". Only expected the following keys: ${this.outputNames}`);
+        }
+      }
+      for (const name of this.outputNames) {
+        if (args.loss[name] == null) {
+          console.warn(`Output "${name}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${name} during training`);
+        }
+        lossFunctions.push(get(args.loss[name]));
+      }
+    } else if (Array.isArray(args.loss)) {
+      if (args.loss.length !== this.outputs.length) {
+        throw new ValueError(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${args.loss}.`);
+      }
+      const theLosses = args.loss;
+      lossFunctions = theLosses.map((l) => get(l));
+    } else {
+      const lossFunction = get(args.loss);
+      this.outputs.forEach((_) => {
+        lossFunctions.push(lossFunction);
+      });
+    }
+    this.lossFunctions = lossFunctions;
+    this.feedOutputNames = [];
+    this.feedOutputShapes = [];
+    this.feedLossFns = [];
+    for (let i = 0; i < this.outputs.length; ++i) {
+      const shape = this.internalOutputShapes[i];
+      const name = this.outputNames[i];
+      this.feedOutputNames.push(name);
+      this.feedOutputShapes.push(shape);
+      this.feedLossFns.push(this.lossFunctions[i]);
+    }
+    const skipTargetIndices = [];
+    this.metrics = args.metrics;
+    this.metricsNames = ["loss"];
+    this.metricsTensors = [];
+    nameScope("loss", () => {
+      for (let i = 0; i < this.outputs.length; ++i) {
+        if (skipTargetIndices.indexOf(i) !== -1) {
+          continue;
+        }
+        const weightedLoss = this.lossFunctions[i];
+        if (this.outputs.length > 1) {
+          this.metricsTensors.push([weightedLoss, i]);
+          this.metricsNames.push(this.outputNames[i] + "_loss");
+        }
+      }
+    });
+    const nestedMetrics = collectMetrics(args.metrics, this.outputNames);
+    const appendMetric = (outputIndex, metricName, metricTensor) => {
+      if (this.outputNames.length > 1) {
+        metricName = this.outputNames[outputIndex] + "_" + metricName;
+      }
+      this.metricsNames.push(metricName);
+      this.metricsTensors.push([metricTensor, outputIndex]);
+    };
+    nameScope("metric", () => {
+      for (let i = 0; i < this.outputs.length; ++i) {
+        if (skipTargetIndices.indexOf(i) !== -1) {
+          continue;
+        }
+        const outputMetrics = nestedMetrics[i];
+        const handleMetrics = (metrics) => {
+          const metricNamePrefix = "";
+          let metricName;
+          let accFn;
+          let weightedMetricFn;
+          for (const metric of metrics) {
+            if (typeof metric === "string" && ["accuracy", "acc", "crossentropy", "ce"].indexOf(metric) !== -1) {
+              const outputShape = this.internalOutputShapes[i];
+              if (outputShape[outputShape.length - 1] === 1 || this.lossFunctions[i] === binaryCrossentropy) {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = binaryAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = binaryCrossentropy2;
+                }
+              } else if (this.lossFunctions[i] === sparseCategoricalCrossentropy) {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = sparseCategoricalAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = sparseCategoricalCrossentropy2;
+                }
+              } else {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = categoricalAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = categoricalCrossentropy2;
+                }
+              }
+              let suffix;
+              if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                suffix = "acc";
+              } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                suffix = "ce";
+              }
+              weightedMetricFn = accFn;
+              metricName = metricNamePrefix + suffix;
+            } else {
+              const metricFn = get2(metric);
+              weightedMetricFn = metricFn;
+              metricName = metricNamePrefix + getLossOrMetricName(metric);
+            }
+            let metricResult;
+            nameScope(metricName, () => {
+              metricResult = weightedMetricFn;
+            });
+            appendMetric(i, metricName, metricResult);
+          }
+        };
+        handleMetrics(outputMetrics);
+      }
+    });
+    this.collectedTrainableWeights = this.trainableWeights;
+  }
+  checkTrainableWeightsConsistency() {
+    if (this.collectedTrainableWeights == null) {
+      return;
+    }
+    if (this.trainableWeights.length !== this.collectedTrainableWeights.length) {
+      console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?");
+    }
+  }
+  evaluate(x, y, args = {}) {
+    const batchSize = args.batchSize == null ? 32 : args.batchSize;
+    checkBatchSize(batchSize);
+    const checkBatchAxis = true;
+    const standardizedOuts = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);
+    try {
+      const ins = standardizedOuts[0].concat(standardizedOuts[1]);
+      this.makeTestFunction();
+      const f = this.testFunction;
+      const testOuts = this.testLoop(f, ins, batchSize, args.verbose, args.steps);
+      return singletonOrArray(testOuts);
+    } finally {
+      disposeNewTensors(standardizedOuts[0], x);
+      disposeNewTensors(standardizedOuts[1], y);
+    }
+  }
+  async evaluateDataset(dataset, args) {
+    this.makeTestFunction();
+    return evaluateDataset(this, dataset, args);
+  }
+  checkNumSamples(ins, batchSize, steps, stepsName = "steps") {
+    let numSamples;
+    if (steps != null) {
+      numSamples = null;
+      if (batchSize != null) {
+        throw new ValueError(`If ${stepsName} is set, batchSize must be null or undefined.Got batchSize = ${batchSize}`);
+      }
+    } else if (ins != null) {
+      if (Array.isArray(ins)) {
+        numSamples = ins[0].shape[0];
+      } else {
+        numSamples = ins.shape[0];
+      }
+    } else {
+      throw new ValueError(`Either the input data should have a defined shape, or ${stepsName} shoud be specified.`);
+    }
+    return numSamples;
+  }
+  execute(inputs, outputs) {
+    if (Array.isArray(outputs) && outputs.length === 0) {
+      throw new ValueError("`outputs` is an empty Array, which is not allowed.");
+    }
+    const outputsIsArray = Array.isArray(outputs);
+    const outputNames = outputsIsArray ? outputs : [outputs];
+    const outputSymbolicTensors = this.retrieveSymbolicTensors(outputNames);
+    const feedDict = new FeedDict();
+    if (inputs instanceof Tensor) {
+      inputs = [inputs];
+    }
+    if (Array.isArray(inputs)) {
+      if (inputs.length !== this.inputs.length) {
+        throw new ValueError(`The number of inputs provided (${inputs.length}) does not match the number of inputs of this model (${this.inputs.length}).`);
+      }
+      for (let i = 0; i < this.inputs.length; ++i) {
+        feedDict.add(this.inputs[i], inputs[i]);
+      }
+    } else {
+      for (const input2 of this.inputs) {
+        const tensorValue = inputs[input2.name];
+        if (tensorValue == null) {
+          throw new ValueError(`No value is provided for the model's input ${input2.name}`);
+        }
+        feedDict.add(input2, tensorValue);
+      }
+    }
+    const executeOutputs = execute(outputSymbolicTensors, feedDict);
+    return outputsIsArray ? executeOutputs : executeOutputs[0];
+  }
+  retrieveSymbolicTensors(symbolicTensorNames) {
+    const outputSymbolicTensors = pyListRepeat(null, symbolicTensorNames.length);
+    let outputsRemaining = symbolicTensorNames.length;
+    for (const layer of this.layers) {
+      const layerOutputs = Array.isArray(layer.output) ? layer.output : [layer.output];
+      const layerOutputNames = layerOutputs.map((output) => output.name);
+      for (let i = 0; i < symbolicTensorNames.length; ++i) {
+        const index = layerOutputNames.indexOf(symbolicTensorNames[i]);
+        if (index !== -1) {
+          outputSymbolicTensors[i] = layerOutputs[index];
+          outputsRemaining--;
+        }
+        if (outputsRemaining === 0) {
+          break;
+        }
+      }
+      if (outputsRemaining === 0) {
+        break;
+      }
+    }
+    if (outputsRemaining > 0) {
+      const remainingNames = [];
+      outputSymbolicTensors.forEach((tensor2, i) => {
+        if (tensor2 == null) {
+          remainingNames.push(symbolicTensorNames[i]);
+        }
+      });
+      throw new ValueError(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(remainingNames)}`);
+    }
+    return outputSymbolicTensors;
+  }
+  predictLoop(ins, batchSize = 32, verbose = false) {
+    return tidy(() => {
+      const numSamples = this.checkNumSamples(ins);
+      if (verbose) {
+        throw new NotImplementedError("Verbose predictLoop() is not implemented yet.");
+      }
+      const batches = makeBatches(numSamples, batchSize);
+      const outsBatches = this.outputs.map((output) => []);
+      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+        const batchOuts = tidy(() => {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const insBatch = sliceArrays(ins, batchStart, batchEnd);
+          const feeds = [];
+          if (Array.isArray(insBatch)) {
+            for (let i = 0; i < insBatch.length; ++i) {
+              feeds.push({ key: this.inputs[i], value: insBatch[i] });
+            }
+          } else {
+            feeds.push({ key: this.inputs[0], value: insBatch });
+          }
+          const feedDict = new FeedDict(feeds);
+          return execute(this.outputs, feedDict);
+        });
+        batchOuts.forEach((batchOut, i) => outsBatches[i].push(batchOut));
+      }
+      return singletonOrArray(outsBatches.map((batches2) => concat(batches2, 0)));
+    });
+  }
+  predict(x, args = {}) {
+    const xsRank2OrHigher = ensureTensorsRank2OrHigher(x);
+    checkInputData(xsRank2OrHigher, this.inputNames, this.feedInputShapes, false);
+    try {
+      const batchSize = args.batchSize == null ? 32 : args.batchSize;
+      checkBatchSize(batchSize);
+      return this.predictLoop(xsRank2OrHigher, batchSize);
+    } finally {
+      disposeNewTensors(xsRank2OrHigher, x);
+    }
+  }
+  predictOnBatch(x) {
+    checkInputData(x, this.inputNames, this.feedInputShapes, true);
+    const batchSize = (Array.isArray(x) ? x[0] : x).shape[0];
+    return this.predictLoop(x, batchSize);
+  }
+  standardizeUserDataXY(x, y, checkBatchAxis = true, batchSize) {
+    if (this.optimizer_ == null) {
+      throw new RuntimeError("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");
+    }
+    const outputShapes = [];
+    for (let i = 0; i < this.feedOutputShapes.length; ++i) {
+      const outputShape = this.feedOutputShapes[i];
+      const lossFn = this.feedLossFns[i];
+      if (lossFn === sparseCategoricalCrossentropy) {
+        outputShapes.push(outputShape.slice(0, outputShape.length - 1).concat([1]));
+      } else {
+        outputShapes.push(outputShape);
+      }
+    }
+    x = standardizeInputData(x, this.feedInputNames, this.feedInputShapes, false, "input");
+    y = standardizeInputData(y, this.feedOutputNames, outputShapes, false, "target");
+    checkArrayLengths(x, y, null);
+    checkLossAndTargetCompatibility(y, this.feedLossFns, this.feedOutputShapes);
+    if (this.stateful && batchSize != null && batchSize > 0) {
+      if (x[0].shape[0] % batchSize !== 0) {
+        throw new ValueError(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${batchSize}. Found: ${x[0].shape[0]} sample(s).`);
+      }
+    }
+    return [x, y];
+  }
+  async standardizeUserData(x, y, sampleWeight, classWeight, checkBatchAxis = true, batchSize) {
+    const [standardXs, standardYs] = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);
+    if (sampleWeight != null) {
+      throw new Error("sample weight is not supported yet.");
+    }
+    let standardSampleWeights = null;
+    if (classWeight != null) {
+      const classWeights = standardizeClassWeights(classWeight, this.outputNames);
+      standardSampleWeights = [];
+      for (let i = 0; i < classWeights.length; ++i) {
+        standardSampleWeights.push(await standardizeWeights(standardYs[i], null, classWeights[i]));
+      }
+    }
+    return [standardXs, standardYs, standardSampleWeights];
+  }
+  testLoop(f, ins, batchSize, verbose = 0, steps) {
+    return tidy(() => {
+      const numSamples = this.checkNumSamples(ins, batchSize, steps, "steps");
+      const outs = [];
+      if (verbose > 0) {
+        throw new NotImplementedError("Verbose mode is not implemented yet.");
+      }
+      if (steps != null) {
+        throw new NotImplementedError("steps mode in testLoop() is not implemented yet");
+      } else {
+        const batches = makeBatches(numSamples, batchSize);
+        const indexArray = tensor1d(range2(0, numSamples));
+        for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const batchIds = sliceAlongFirstAxis(indexArray, batchStart, batchEnd - batchStart);
+          const insBatch = sliceArraysByIndices(ins, batchIds);
+          const batchOuts = f(insBatch);
+          if (batchIndex === 0) {
+            for (let i = 0; i < batchOuts.length; ++i) {
+              outs.push(scalar(0));
+            }
+          }
+          for (let i = 0; i < batchOuts.length; ++i) {
+            const batchOut = batchOuts[i];
+            outs[i] = add2(outs[i], mul(batchEnd - batchStart, batchOut));
+          }
+        }
+        for (let i = 0; i < outs.length; ++i) {
+          outs[i] = div(outs[i], numSamples);
+        }
+      }
+      return outs;
+    });
+  }
+  getDedupedMetricsNames() {
+    const outLabels = this.metricsNames;
+    const dedupedOutLabels = [];
+    for (let i = 0; i < outLabels.length; ++i) {
+      const label = outLabels[i];
+      let newLabel = label;
+      if (count(outLabels, label) > 1) {
+        const dupIndex = count(outLabels.slice(0, i), label);
+        newLabel += `_${dupIndex}`;
+      }
+      dedupedOutLabels.push(newLabel);
+    }
+    return dedupedOutLabels;
+  }
+  makeTrainFunction() {
+    return (data) => {
+      const lossValues = [];
+      const inputs = data.slice(0, this.inputs.length);
+      const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);
+      const sampleWeights = data.slice(this.inputs.length + this.outputs.length, this.inputs.length + this.outputs.length * 2);
+      const metricsValues = [];
+      const totalLossFunction = () => {
+        const feeds = [];
+        for (let i = 0; i < this.inputs.length; ++i) {
+          feeds.push({ key: this.inputs[i], value: inputs[i] });
+        }
+        const feedDict = new FeedDict(feeds);
+        const outputs = execute(this.outputs, feedDict, { "training": true });
+        let totalLoss;
+        for (let i = 0; i < this.lossFunctions.length; ++i) {
+          const lossFunction = this.lossFunctions[i];
+          let loss = lossFunction(targets[i], outputs[i]);
+          if (sampleWeights[i] != null) {
+            loss = computeWeightedLoss2(loss, sampleWeights[i]);
+          }
+          const meanLoss = mean(loss);
+          lossValues.push(meanLoss);
+          if (i === 0) {
+            totalLoss = loss;
+          } else {
+            totalLoss = add2(totalLoss, loss);
+          }
+        }
+        for (let i = 0; i < this.metricsTensors.length; ++i) {
+          let weightedMetric;
+          if (this.outputs.length > 1 && i < this.outputs.length) {
+            weightedMetric = lossValues[i];
+          } else {
+            const metric = this.metricsTensors[i][0];
+            const outputIndex = this.metricsTensors[i][1];
+            weightedMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));
+          }
+          keep(weightedMetric);
+          metricsValues.push(weightedMetric);
+        }
+        totalLoss = mean(totalLoss);
+        this.calculateLosses().forEach((regularizerLoss) => {
+          totalLoss = add2(totalLoss, regularizerLoss);
+        });
+        return totalLoss;
+      };
+      const variables = this.collectedTrainableWeights.map((param) => param.read());
+      const returnCost = true;
+      const totalLossValue = this.optimizer_.minimize(totalLossFunction, returnCost, variables);
+      return [totalLossValue].concat(metricsValues);
+    };
+  }
+  makeTestFunction() {
+    this.testFunction = (data) => {
+      return tidy(() => {
+        const valOutputs = [];
+        let totalLoss;
+        const inputs = data.slice(0, this.inputs.length);
+        const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);
+        const feeds = [];
+        for (let i = 0; i < this.inputs.length; ++i) {
+          feeds.push({ key: this.inputs[i], value: inputs[i] });
+        }
+        const feedDict = new FeedDict(feeds);
+        const outputs = execute(this.outputs, feedDict);
+        for (let i = 0; i < this.lossFunctions.length; ++i) {
+          const lossFunction = this.lossFunctions[i];
+          const loss = mean(lossFunction(targets[i], outputs[i]));
+          if (i === 0) {
+            totalLoss = loss;
+          } else {
+            totalLoss = add2(totalLoss, loss);
+          }
+          valOutputs.push(totalLoss);
+        }
+        for (let i = 0; i < this.metricsTensors.length; ++i) {
+          const metric = this.metricsTensors[i][0];
+          const outputIndex = this.metricsTensors[i][1];
+          const meanMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));
+          valOutputs.push(meanMetric);
+        }
+        return valOutputs;
+      });
+    };
+  }
+  async fit(x, y, args = {}) {
+    return fitTensors(this, x, y, args);
+  }
+  async fitDataset(dataset, args) {
+    return fitDataset(this, dataset, args);
+  }
+  async trainOnBatch(x, y) {
+    const standardizeOut = await this.standardizeUserData(x, y);
+    const inputs = standardizeOut[0];
+    const targets = standardizeOut[1];
+    const trainFunction = this.makeTrainFunction();
+    const losses2 = trainFunction(inputs.concat(targets));
+    const lossValues = [];
+    for (const loss of losses2) {
+      const v = await loss.data();
+      lossValues.push(v[0]);
+    }
+    dispose(losses2);
+    disposeNewTensors(standardizeOut[0], x);
+    disposeNewTensors(standardizeOut[1], y);
+    return singletonOrArray(lossValues);
+  }
+  getNamedWeights(config) {
+    const namedWeights = [];
+    const trainableOnly = config != null && config.trainableOnly;
+    const weights = trainableOnly ? this.trainableWeights : this.weights;
+    const weightValues = this.getWeights(trainableOnly);
+    for (let i = 0; i < weights.length; ++i) {
+      if (trainableOnly && !weights[i].trainable) {
+        continue;
+      }
+      namedWeights.push({ name: weights[i].originalName, tensor: weightValues[i] });
+    }
+    return namedWeights;
+  }
+  set stopTraining(stop) {
+    this.stopTraining_ = stop;
+  }
+  get stopTraining() {
+    return this.stopTraining_;
+  }
+  get optimizer() {
+    return this.optimizer_;
+  }
+  set optimizer(optimizer) {
+    if (this.optimizer_ !== optimizer) {
+      this.optimizer_ = optimizer;
+      this.isOptimizerOwned = false;
+    }
+  }
+  dispose() {
+    const result = super.dispose();
+    if (result.refCountAfterDispose === 0 && this.optimizer != null && this.isOptimizerOwned) {
+      const numTensorsBeforeOptmizerDisposal = memory().numTensors;
+      this.optimizer_.dispose();
+      result.numDisposedVariables += numTensorsBeforeOptmizerDisposal - memory().numTensors;
+    }
+    return result;
+  }
+  getLossIdentifiers() {
+    let lossNames;
+    if (typeof this.loss === "string") {
+      lossNames = toSnakeCase(this.loss);
+    } else if (Array.isArray(this.loss)) {
+      for (const loss of this.loss) {
+        if (typeof loss !== "string") {
+          throw new Error("Serialization of non-string loss is not supported.");
+        }
+      }
+      lossNames = this.loss.map((name) => toSnakeCase(name));
+    } else {
+      const outputNames = Object.keys(this.loss);
+      lossNames = {};
+      const losses2 = this.loss;
+      for (const outputName of outputNames) {
+        if (typeof losses2[outputName] === "string") {
+          lossNames[outputName] = toSnakeCase(losses2[outputName]);
+        } else {
+          throw new Error("Serialization of non-string loss is not supported.");
+        }
+      }
+    }
+    return lossNames;
+  }
+  getMetricIdentifiers() {
+    if (typeof this.metrics === "string" || typeof this.metrics === "function") {
+      return [toSnakeCase(getLossOrMetricName(this.metrics))];
+    } else if (Array.isArray(this.metrics)) {
+      return this.metrics.map((metric) => toSnakeCase(getLossOrMetricName(metric)));
+    } else {
+      const metricsIdentifiers = {};
+      for (const key in this.metrics) {
+        metricsIdentifiers[key] = toSnakeCase(getLossOrMetricName(this.metrics[key]));
+      }
+      return metricsIdentifiers;
+    }
+  }
+  getTrainingConfig() {
+    return {
+      loss: this.getLossIdentifiers(),
+      metrics: this.getMetricIdentifiers(),
+      optimizer_config: {
+        class_name: this.optimizer.getClassName(),
+        config: this.optimizer.getConfig()
+      }
+    };
+  }
+  loadTrainingConfig(trainingConfig) {
+    if (trainingConfig.weighted_metrics != null) {
+      throw new Error("Loading weight_metrics is not supported yet.");
+    }
+    if (trainingConfig.loss_weights != null) {
+      throw new Error("Loading loss_weights is not supported yet.");
+    }
+    if (trainingConfig.sample_weight_mode != null) {
+      throw new Error("Loading sample_weight_mode is not supported yet.");
+    }
+    const tsConfig = convertPythonicToTs(trainingConfig.optimizer_config);
+    const optimizer = deserialize(tsConfig);
+    let loss;
+    if (typeof trainingConfig.loss === "string") {
+      loss = toCamelCase(trainingConfig.loss);
+    } else if (Array.isArray(trainingConfig.loss)) {
+      loss = trainingConfig.loss.map((lossEntry) => toCamelCase(lossEntry));
+    } else if (trainingConfig.loss != null) {
+      loss = {};
+      for (const key in trainingConfig.loss) {
+        loss[key] = toCamelCase(trainingConfig.loss[key]);
+      }
+    }
+    let metrics;
+    if (Array.isArray(trainingConfig.metrics)) {
+      metrics = trainingConfig.metrics.map((metric) => toCamelCase(metric));
+    } else if (trainingConfig.metrics != null) {
+      metrics = {};
+      for (const key in trainingConfig.metrics) {
+        metrics[key] = toCamelCase(trainingConfig.metrics[key]);
+      }
+    }
+    this.compile({ loss, metrics, optimizer });
+  }
+  async save(handlerOrURL, config) {
+    if (typeof handlerOrURL === "string") {
+      const handlers = io_exports.getSaveHandlers(handlerOrURL);
+      if (handlers.length === 0) {
+        throw new ValueError(`Cannot find any save handlers for URL '${handlerOrURL}'`);
+      } else if (handlers.length > 1) {
+        throw new ValueError(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);
+      }
+      handlerOrURL = handlers[0];
+    }
+    if (handlerOrURL.save == null) {
+      throw new ValueError("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+    }
+    const weightDataAndSpecs = await io_exports.encodeWeights(this.getNamedWeights(config));
+    const returnString = false;
+    const unusedArg = null;
+    const modelConfig = this.toJSON(unusedArg, returnString);
+    const modelArtifacts = {
+      modelTopology: modelConfig,
+      format: LAYERS_MODEL_FORMAT_NAME,
+      generatedBy: `TensorFlow.js tfjs-layers v${version2}`,
+      convertedBy: null
+    };
+    const includeOptimizer = config == null ? false : config.includeOptimizer;
+    if (includeOptimizer && this.optimizer != null) {
+      modelArtifacts.trainingConfig = this.getTrainingConfig();
+      const weightType = "optimizer";
+      const { data: optimizerWeightData, specs: optimizerWeightSpecs } = await io_exports.encodeWeights(await this.optimizer.getWeights(), weightType);
+      weightDataAndSpecs.specs.push(...optimizerWeightSpecs);
+      weightDataAndSpecs.data = io_exports.concatenateArrayBuffers([weightDataAndSpecs.data, optimizerWeightData]);
+    }
+    if (this.userDefinedMetadata != null) {
+      const checkSize = true;
+      checkUserDefinedMetadata(this.userDefinedMetadata, this.name, checkSize);
+      modelArtifacts.userDefinedMetadata = this.userDefinedMetadata;
+    }
+    modelArtifacts.weightData = weightDataAndSpecs.data;
+    modelArtifacts.weightSpecs = weightDataAndSpecs.specs;
+    return handlerOrURL.save(modelArtifacts);
+  }
+  setUserDefinedMetadata(userDefinedMetadata) {
+    checkUserDefinedMetadata(userDefinedMetadata, this.name);
+    this.userDefinedMetadata = userDefinedMetadata;
+  }
+  getUserDefinedMetadata() {
+    return this.userDefinedMetadata;
+  }
+};
+LayersModel.className = "Model";
+serialization_exports.registerClass(LayersModel);
+var Functional = class extends LayersModel {
+};
+Functional.className = "Functional";
+serialization_exports.registerClass(Functional);
+async function modelFromJSON(modelAndWeightsConfig, customObjects) {
+  if (!("modelTopology" in modelAndWeightsConfig)) {
+    modelAndWeightsConfig = { modelTopology: modelAndWeightsConfig };
+  }
+  modelAndWeightsConfig = modelAndWeightsConfig;
+  let modelTopology = modelAndWeightsConfig.modelTopology;
+  if (modelTopology["model_config"] != null) {
+    modelTopology = modelTopology["model_config"];
+  }
+  const tsConfig = convertPythonicToTs(modelTopology);
+  const model2 = deserialize(tsConfig, customObjects);
+  if (modelAndWeightsConfig.weightsManifest != null) {
+    const weightValues = await io_exports.loadWeights(modelAndWeightsConfig.weightsManifest, modelAndWeightsConfig.pathPrefix, model2.weights.map((weight) => weight.originalName));
+    const uniqueWeightValues = {};
+    for (const weight of model2.weights) {
+      uniqueWeightValues[weight.originalName] = weightValues[weight.originalName];
+    }
+    model2.loadWeights(uniqueWeightValues);
+    dispose(weightValues);
+  }
+  return model2;
+}
+async function loadLayersModel(pathOrIOHandler, options) {
+  if (options == null) {
+    options = {};
+  }
+  if (typeof pathOrIOHandler === "string") {
+    const handlers = io_exports.getLoadHandlers(pathOrIOHandler, options);
+    if (handlers.length === 0) {
+      handlers.push(io_exports.browserHTTPRequest(pathOrIOHandler, options));
+    } else if (handlers.length > 1) {
+      throw new ValueError(`Found more than one (${handlers.length}) load handlers for URL '${pathOrIOHandler}'`);
+    }
+    pathOrIOHandler = handlers[0];
+  }
+  return loadLayersModelFromIOHandler(pathOrIOHandler, void 0, options);
+}
+async function loadLayersModelFromIOHandler(handler, customObjects, options) {
+  if (options == null) {
+    options = {};
+  }
+  if (handler.load == null) {
+    throw new ValueError("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+  }
+  const artifacts = await handler.load();
+  let modelTopology = artifacts.modelTopology;
+  if (modelTopology["model_config"] != null) {
+    modelTopology = modelTopology["model_config"];
+  }
+  const strict = options.strict == null ? true : options.strict;
+  const fastWeightInit = artifacts.weightData != null && artifacts.weightSpecs != null && strict;
+  const model2 = deserialize(convertPythonicToTs(modelTopology), customObjects, fastWeightInit);
+  const trainingConfig = artifacts.trainingConfig;
+  if (trainingConfig != null) {
+    model2.loadTrainingConfig(trainingConfig);
+  }
+  if (artifacts.userDefinedMetadata != null) {
+    model2.setUserDefinedMetadata(artifacts.userDefinedMetadata);
+  }
+  if (artifacts.weightData != null) {
+    if (artifacts.weightSpecs == null) {
+      throw new ValueError("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");
+    }
+    const { modelWeights, optimizerWeights } = decodeModelAndOptimizerWeights(artifacts.weightData, artifacts.weightSpecs);
+    model2.loadWeights(modelWeights, strict);
+    if (model2.optimizer != null && optimizerWeights.length > 0) {
+      await model2.optimizer.setWeights(optimizerWeights);
+    }
+    dispose(modelWeights);
+    dispose(optimizerWeights.map((w) => w.tensor));
+  }
+  return model2;
+}
+function decodeModelAndOptimizerWeights(buffer2, specs) {
+  const name2Tensor = io_exports.decodeWeights(buffer2, specs);
+  const modelWeights = {};
+  const optimizerWeights = [];
+  specs.forEach((spec) => {
+    if (spec.group === "optimizer") {
+      optimizerWeights.push({ name: spec.name, tensor: name2Tensor[spec.name] });
+    } else {
+      modelWeights[spec.name] = name2Tensor[spec.name];
+    }
+  });
+  return { modelWeights, optimizerWeights };
+}
+var Sequential = class extends LayersModel {
+  constructor(args) {
+    super({ inputs: [], outputs: [] });
+    args = args || {};
+    this.trainable = true;
+    this.built = false;
+    this.name = args.name != null ? args.name : getUid("sequential_");
+    if (args.layers != null) {
+      for (const layer of args.layers) {
+        this.add(layer);
+      }
+    }
+  }
+  checkShape(layer) {
+    const shape = layer.inboundNodes[0].outputTensors[0].shape;
+    if (shape.some((x) => x < 0)) {
+      throw new ValueError(`Negative dimension size caused by adding layer ${layer.name} with input shape [${layer.inboundNodes[0].inputTensors[0].shape}]`);
+    }
+  }
+  add(layer) {
+    const isLayerModelInstance = layer instanceof Sequential || layer instanceof LayersModel;
+    let modelLayer;
+    if (isLayerModelInstance) {
+      modelLayer = layer;
+      if (modelLayer.outputs.length !== 1) {
+        throw new ValueError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+      }
+      if (modelLayer.inputs.length !== 1) {
+        throw new ValueError("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.");
+      }
+    }
+    if (this.outputs.length === 0) {
+      if (layer.inboundNodes.length === 0) {
+        if (layer.batchInputShape == null) {
+          throw new ValueError("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");
+        }
+        const x = Input({
+          batchShape: layer.batchInputShape,
+          dtype: layer.dtype,
+          name: layer.name + "_input"
+        });
+        layer.apply(x);
+      }
+      if (isLayerModelInstance) {
+        this.outputs = modelLayer.outputs;
+        this.inputs = modelLayer.inputs;
+      } else {
+        if (layer.inboundNodes.length !== 1) {
+          throw new ValueError(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${layer.name} which has ${layer.inboundNodes.length} pre-existing inbound connections.`);
+        }
+        if (layer.inboundNodes[0].outputTensors.length !== 1) {
+          throw new ValueError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+        }
+        this.checkShape(layer);
+        this.outputs = [layer.inboundNodes[0].outputTensors[0]];
+        this.inputs = getSourceInputs(this.outputs[0]);
+      }
+      this.inboundNodes = [];
+      new Node({
+        outboundLayer: this,
+        inboundLayers: [],
+        nodeIndices: [],
+        tensorIndices: [],
+        inputTensors: this.inputs,
+        outputTensors: this.outputs,
+        inputMasks: pyListRepeat(null, this.inputs.length),
+        outputMasks: [null],
+        inputShapes: this.inputs.map((x) => x.shape),
+        outputShapes: this.outputs[0].shape
+      });
+    } else {
+      const outputTensor = layer.apply(this.outputs[0]);
+      if (Array.isArray(outputTensor)) {
+        throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+      }
+      this.checkShape(layer);
+      this.outputs = [outputTensor];
+      this.inboundNodes[0].outputTensors = this.outputs;
+      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];
+    }
+    this.layers.push(layer);
+    this.built = false;
+  }
+  pop() {
+    if (this.layers.length === 0) {
+      throw new TypeError("There are no layers in the model.");
+    }
+    this.layers.pop();
+    if (this.layers.length === 0) {
+      this.outputs = [];
+      this.inboundNodes = [];
+      this.outboundNodes = [];
+    } else {
+      const lastLayerIndex = this.layers.length - 1;
+      this.layers[lastLayerIndex].outboundNodes = [];
+      this.outputs = [this.layers[lastLayerIndex].output];
+      this.inboundNodes[0].outputTensors = this.outputs;
+      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];
+    }
+  }
+  call(inputs, kwargs) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.call(inputs, kwargs);
+  }
+  build(inputShape) {
+    getExactlyOneShape(inputShape);
+    if (this.inputs.length === 0 || this.outputs.length === 0) {
+      throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");
+    }
+    this.model = new LayersModel({
+      inputs: this.inputs,
+      outputs: this.outputs[0],
+      name: this.name + "_model"
+    });
+    this.model.trainable = this.trainable;
+    this.supportsMasking = this.model.supportsMasking;
+    this.inputLayers = this.model.inputLayers;
+    this.inputLayersNodeIndices = this.model.inputLayersNodeIndices;
+    this.inputLayersTensorIndices = this.model.inputLayersTensorIndices;
+    this.outputLayers = this.model.outputLayers;
+    this.outputLayersNodeIndices = this.model.outputLayersNodeIndices;
+    this.outputLayersTensorIndices = this.model.outputLayersTensorIndices;
+    this.nodesByDepth = this.model.nodesByDepth;
+    this.containerNodes = this.model.containerNodes;
+    this.outputNames = this.model.outputNames;
+    this.inputNames = this.model.inputNames;
+    this.built = true;
+  }
+  countParams() {
+    if (!this.built) {
+      this.build();
+    }
+    return super.countParams();
+  }
+  summary(lineLength, positions, printFn = console.log) {
+    if (!this.built) {
+      this.build();
+    }
+    super.summary(lineLength, positions, printFn);
+  }
+  setWeights(weights) {
+    if (this.model == null) {
+      this.build();
+    }
+    this.model.setWeights(weights);
+  }
+  evaluate(x, y, args = {}) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.evaluate(x, y, args);
+  }
+  async evaluateDataset(dataset, args) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.evaluateDataset(dataset, args);
+  }
+  predict(x, args = {}) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.predict(x, args);
+  }
+  predictOnBatch(x) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.predictOnBatch(x);
+  }
+  compile(args) {
+    this.build();
+    this.model.compile(args);
+    this.optimizer_ = this.model.optimizer;
+    this.isOptimizerOwned = this.model.isOptimizerOwned;
+    this.loss = this.model.loss;
+    this.metrics = this.model.metrics;
+    this.metricsTensors = this.model.metricsTensors;
+    this.metricsNames = this.model.metricsNames;
+  }
+  get optimizer() {
+    return this.model == null ? void 0 : this.model.optimizer;
+  }
+  set optimizer(optimizer) {
+    this.model.optimizer = optimizer;
+  }
+  async fit(x, y, args = {}) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.fit(x, y, args);
+  }
+  async fitDataset(dataset, args) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.fitDataset(dataset, args);
+  }
+  async trainOnBatch(x, y) {
+    return this.model.trainOnBatch(x, y);
+  }
+  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {
+    let configArray;
+    let extraModelConfig = {};
+    if (config instanceof Array) {
+      if (!(config[0].className != null) || config[0]["className"] === "Merge") {
+        throw new ValueError("Legacy serialization format not supported yet.");
+      }
+      configArray = config;
+    } else {
+      util_exports.assert(config["layers"] != null, () => `When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field.`);
+      configArray = config["layers"];
+      delete config["layers"];
+      extraModelConfig = config;
+    }
+    const model2 = new cls(extraModelConfig);
+    if (!(model2 instanceof Sequential)) {
+      throw new NotImplementedError(`Sequential.fromConfig called on non-Sequential input: ${model2}`);
+    }
+    for (const conf of configArray) {
+      const customObjects2 = void 0;
+      const layer = deserialize(conf, customObjects2, fastWeightInit);
+      if (fastWeightInit) {
+        layer.setFastWeightInitDuringBuild(true);
+      }
+      model2.add(layer);
+    }
+    return model2;
+  }
+  set stopTraining(stop) {
+    if (this.model == null) {
+      throw new ValueError("Cannot set the stopTraining property of a sequential model before it is compiled.");
+    }
+    this.model.stopTraining = stop;
+  }
+  get stopTraining() {
+    if (this.model == null) {
+      throw new ValueError("Cannot get the stopTraining property of a sequential model before it is compiled.");
+    }
+    return this.model.stopTraining;
+  }
+  getConfig() {
+    const layers = [];
+    for (const layer of this.layers) {
+      const dict = {};
+      dict["className"] = layer.getClassName();
+      dict["config"] = layer.getConfig();
+      layers.push(dict);
+    }
+    return { name: this.name, layers };
+  }
+};
+Sequential.className = "Sequential";
+serialization_exports.registerClass(Sequential);
+function model(args) {
+  return new LayersModel(args);
+}
+function sequential(config) {
+  return new Sequential(config);
+}
+function input(config) {
+  return Input(config);
+}
+function registerCallbackConstructor(verbosityLevel, callbackConstructor) {
+  CallbackConstructorRegistry.registerCallbackConstructor(verbosityLevel, callbackConstructor);
+}
+var Activation = class extends serialization_exports.Serializable {
+  getConfig() {
+    return {};
+  }
+};
+var Elu2 = class extends Activation {
+  apply(x, alpha = 1) {
+    return elu2(x, alpha);
+  }
+};
+Elu2.className = "elu";
+serialization_exports.registerClass(Elu2);
+var Selu2 = class extends Activation {
+  apply(x) {
+    return selu(x);
+  }
+};
+Selu2.className = "selu";
+serialization_exports.registerClass(Selu2);
+var Relu2 = class extends Activation {
+  apply(x) {
+    return relu(x);
+  }
+};
+Relu2.className = "relu";
+serialization_exports.registerClass(Relu2);
+var Relu62 = class extends Activation {
+  apply(x) {
+    return tidy(() => minimum(6, relu(x)));
+  }
+};
+Relu62.className = "relu6";
+serialization_exports.registerClass(Relu62);
+var Linear = class extends Activation {
+  apply(x) {
+    return x;
+  }
+};
+Linear.className = "linear";
+serialization_exports.registerClass(Linear);
+var Sigmoid2 = class extends Activation {
+  apply(x) {
+    return sigmoid(x);
+  }
+};
+Sigmoid2.className = "sigmoid";
+serialization_exports.registerClass(Sigmoid2);
+var HardSigmoid = class extends Activation {
+  apply(x) {
+    return hardSigmoid(x);
+  }
+};
+HardSigmoid.className = "hardSigmoid";
+serialization_exports.registerClass(HardSigmoid);
+var Softplus2 = class extends Activation {
+  apply(x) {
+    return softplus(x);
+  }
+};
+Softplus2.className = "softplus";
+serialization_exports.registerClass(Softplus2);
+var Softsign = class extends Activation {
+  apply(x) {
+    return softsign(x);
+  }
+};
+Softsign.className = "softsign";
+serialization_exports.registerClass(Softsign);
+var Tanh2 = class extends Activation {
+  apply(x) {
+    return tanh2(x);
+  }
+};
+Tanh2.className = "tanh";
+serialization_exports.registerClass(Tanh2);
+var Softmax2 = class extends Activation {
+  apply(x, axis = -1) {
+    return softmax(x, axis);
+  }
+};
+Softmax2.className = "softmax";
+serialization_exports.registerClass(Softmax2);
+var LogSoftmax2 = class extends Activation {
+  apply(x, axis = -1) {
+    return logSoftmax(x, axis);
+  }
+};
+LogSoftmax2.className = "logSoftmax";
+serialization_exports.registerClass(LogSoftmax2);
+var Swish = class extends Activation {
+  apply(x, alpha = 1) {
+    return tidy(() => mul(sigmoid(mul(x, alpha)), x));
+  }
+};
+Swish.className = "swish";
+serialization_exports.registerClass(Swish);
+var Mish = class extends Activation {
+  apply(x) {
+    return tidy(() => mul(x, tanh2(softplus(x))));
+  }
+};
+Mish.className = "mish";
+serialization_exports.registerClass(Mish);
+function serializeActivation(activation2) {
+  return activation2.getClassName();
+}
+function deserializeActivation(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "activation");
+}
+function getActivation(identifier) {
+  if (identifier == null) {
+    const config = {};
+    config["className"] = "linear";
+    config["config"] = {};
+    return deserializeActivation(config);
+  }
+  if (typeof identifier === "string") {
+    const config = {};
+    config["className"] = identifier;
+    config["config"] = {};
+    return deserializeActivation(config);
+  } else if (identifier instanceof Activation) {
+    return identifier;
+  } else {
+    return deserializeActivation(identifier);
+  }
+}
+function assertObjectArgs(args) {
+  if (args != null && typeof args !== "object") {
+    throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${args}`);
+  }
+}
+var Regularizer = class extends serialization_exports.Serializable {
+};
+var L1L2 = class extends Regularizer {
+  constructor(args) {
+    super();
+    assertObjectArgs(args);
+    this.l1 = args == null || args.l1 == null ? 0.01 : args.l1;
+    this.l2 = args == null || args.l2 == null ? 0.01 : args.l2;
+    this.hasL1 = this.l1 !== 0;
+    this.hasL2 = this.l2 !== 0;
+  }
+  apply(x) {
+    return tidy(() => {
+      let regularization = zeros([1]);
+      if (this.hasL1) {
+        regularization = add2(regularization, sum2(mul(this.l1, abs(x))));
+      }
+      if (this.hasL2) {
+        regularization = add2(regularization, sum2(mul(this.l2, square2(x))));
+      }
+      return reshape(regularization, []);
+    });
+  }
+  getConfig() {
+    return { "l1": this.l1, "l2": this.l2 };
+  }
+  static fromConfig(cls, config) {
+    return new cls({ l1: config["l1"], l2: config["l2"] });
+  }
+};
+L1L2.className = "L1L2";
+serialization_exports.registerClass(L1L2);
+function l1(args) {
+  assertObjectArgs(args);
+  return new L1L2({ l1: args != null ? args.l1 : null, l2: 0 });
+}
+function l2(args) {
+  assertObjectArgs(args);
+  return new L1L2({ l2: args != null ? args.l2 : null, l1: 0 });
+}
+var REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "l1l2": "L1L2"
+};
+function serializeRegularizer(constraint) {
+  return serializeKerasObject(constraint);
+}
+function deserializeRegularizer(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "regularizer");
+}
+function getRegularizer(identifier) {
+  if (identifier == null) {
+    return null;
+  }
+  if (typeof identifier === "string") {
+    const className = identifier in REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    const config = { className, config: {} };
+    return deserializeRegularizer(config);
+  } else if (identifier instanceof Regularizer) {
+    return identifier;
+  } else {
+    return deserializeRegularizer(identifier);
+  }
+}
+var ReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.supportsMasking = true;
+    if (args != null) {
+      this.maxValue = args.maxValue;
+    }
+  }
+  call(inputs, kwargs) {
+    inputs = getExactlyOneTensor(inputs);
+    let output = relu(inputs);
+    if (this.maxValue != null) {
+      output = clipByValue(output, 0, this.maxValue);
+    }
+    return output;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { maxValue: this.maxValue };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ReLU.className = "ReLU";
+serialization_exports.registerClass(ReLU);
+var LeakyReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA = 0.3;
+    if (args == null) {
+      args = {};
+    }
+    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return leakyRelu(x, this.alpha);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { alpha: this.alpha };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+LeakyReLU.className = "LeakyReLU";
+serialization_exports.registerClass(LeakyReLU);
+var PReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA_INITIALIZER = "zeros";
+    if (args == null) {
+      args = {};
+    }
+    this.supportsMasking = true;
+    this.alphaInitializer = getInitializer(args.alphaInitializer || this.DEFAULT_ALPHA_INITIALIZER);
+    this.alphaRegularizer = getRegularizer(args.alphaRegularizer);
+    this.alphaConstraint = getConstraint(args.alphaConstraint);
+    if (args.sharedAxes == null) {
+      this.sharedAxes = null;
+    } else if (Array.isArray(args.sharedAxes)) {
+      this.sharedAxes = args.sharedAxes;
+    } else if (typeof args.sharedAxes === "number") {
+      this.sharedAxes = [args.sharedAxes];
+    } else {
+      throw new ValueError(`Expected sharedAxes to be a number or an array of numbers, but got ${args.sharedAxes}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const paramShape = inputShape.slice(1);
+    if (this.sharedAxes != null) {
+      for (const i of this.sharedAxes) {
+        paramShape[i - 1] = 1;
+      }
+    }
+    this.alpha = this.addWeight("alpha", paramShape, "float32", this.alphaInitializer, this.alphaRegularizer, true, this.alphaConstraint);
+    const axes = {};
+    if (this.sharedAxes != null) {
+      for (let i = 1; i < inputShape.length; ++i) {
+        axes[i] = inputShape[i];
+      }
+    }
+    this.inputSpec = [new InputSpec({
+      ndim: inputShape.length,
+      axes
+    })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    inputs = getExactlyOneTensor(inputs);
+    return prelu(inputs, this.alpha.read());
+  }
+  getConfig() {
+    const config = {
+      alphaInitializer: serializeInitializer(this.alphaInitializer),
+      alphaRegularizer: serializeRegularizer(this.alphaRegularizer),
+      alphaConstraint: serializeConstraint(this.alphaConstraint),
+      sharedAxes: this.sharedAxes
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+PReLU.className = "PReLU";
+serialization_exports.registerClass(PReLU);
+var ELU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA = 1;
+    if (args == null) {
+      args = {};
+    }
+    if (args.alpha != null && args.alpha !== this.DEFAULT_ALPHA) {
+      throw new NotImplementedError(`Non-default alpha value (${args.alpha}) is not supported by the ELU layer yet.`);
+    }
+    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return elu(x);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { alpha: this.alpha };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ELU.className = "ELU";
+serialization_exports.registerClass(ELU);
+var ThresholdedReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_THETA = 1;
+    if (args == null) {
+      args = {};
+    }
+    this.theta = args.theta == null ? this.DEFAULT_THETA : args.theta;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return mul(x, cast(greater(x, this.theta), "float32"));
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { theta: this.theta };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ThresholdedReLU.className = "ThresholdedReLU";
+serialization_exports.registerClass(ThresholdedReLU);
+var Softmax3 = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_AXIS = 1;
+    if (args == null) {
+      args = {};
+    }
+    this.softmax = new Softmax2().apply;
+    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return this.softmax(x, this.axis);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { axis: this.axis };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Softmax3.className = "Softmax";
+serialization_exports.registerClass(Softmax3);
+function normalizeArray(value, n, name) {
+  if (typeof value === "number") {
+    return pyListRepeat(value, n);
+  } else {
+    if (value.length !== n) {
+      throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${value.length} elements.`);
+    }
+    for (let i = 0; i < n; ++i) {
+      const singleValue = value[i];
+      if (!isInteger(singleValue)) {
+        throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${JSON.stringify(value)} including a non-integer number ${singleValue}`);
+      }
+    }
+    return value;
+  }
+}
+function convOutputLength(inputLength, filterSize, padding, stride, dilation = 1) {
+  if (inputLength == null) {
+    return inputLength;
+  }
+  const dilatedFilterSize = filterSize + (filterSize - 1) * (dilation - 1);
+  let outputLength;
+  if (padding === "same") {
+    outputLength = inputLength;
+  } else {
+    outputLength = inputLength - dilatedFilterSize + 1;
+  }
+  return Math.floor((outputLength + stride - 1) / stride);
+}
+function deconvLength(dimSize, strideSize, kernelSize, padding) {
+  if (dimSize == null) {
+    return null;
+  }
+  if (padding === "valid") {
+    dimSize = dimSize * strideSize + max2([kernelSize - strideSize, 0]);
+  } else if (padding === "same") {
+    dimSize = dimSize * strideSize;
+  } else {
+    throw new ValueError(`Unsupport padding mode: ${padding}.`);
+  }
+  return dimSize;
+}
+function preprocessConv2DInput(x, dataFormat) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    if (dataFormat === "channelsFirst") {
+      return transpose(x, [0, 2, 3, 1]);
+    } else {
+      return x;
+    }
+  });
+}
+function preprocessConv3DInput(x, dataFormat) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    if (dataFormat === "channelsFirst") {
+      return transpose(x, [0, 2, 3, 4, 1]);
+    } else {
+      return x;
+    }
+  });
+}
+function conv1dWithBias(x, kernel, bias, strides = 1, padding = "valid", dataFormat, dilationRate = 1) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.shape.length !== 3) {
+      throw new ValueError(`The input of a conv1dWithBias operation should be 3, but is ${x.shape.length} instead.`);
+    }
+    if (kernel.shape.length !== 3) {
+      throw new ValueError(`The kernel for a conv1dWithBias operation should be 3, but is ${kernel.shape.length} instead`);
+    }
+    if (bias != null && bias.shape.length !== 1) {
+      throw new ValueError(`The bias for a conv1dWithBias operation should be 1, but is ${kernel.shape.length} instead`);
+    }
+    if (dataFormat === "channelsFirst") {
+      x = transpose(x, [0, 2, 1]);
+    }
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+    }
+    let y = conv1d(x, kernel, strides, padding === "same" ? "same" : "valid", "NWC", dilationRate);
+    if (bias != null) {
+      y = biasAdd(y, bias);
+    }
+    return y;
+  });
+}
+function conv2dWithBiasActivation(x, kernel, bias, strides = [1, 1], padding = "valid", dataFormat, dilationRate, activation2 = null) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.rank !== 3 && x.rank !== 4) {
+      throw new ValueError(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${x.rank}.`);
+    }
+    if (kernel.rank !== 3 && kernel.rank !== 4) {
+      throw new ValueError(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${x.rank}.`);
+    }
+    let y = preprocessConv2DInput(x, dataFormat);
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+    }
+    y = fused_ops_exports.conv2d({
+      x: y,
+      filter: kernel,
+      strides,
+      pad: padding === "same" ? "same" : "valid",
+      dilations: dilationRate,
+      dataFormat: "NHWC",
+      bias,
+      activation: activation2
+    });
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+function conv3dWithBias(x, kernel, bias, strides = [1, 1, 1], padding = "valid", dataFormat, dilationRate) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.rank !== 4 && x.rank !== 5) {
+      throw new ValueError(`conv3dWithBias expects input to be of rank 4 or 5, but received ${x.rank}.`);
+    }
+    if (kernel.rank !== 4 && kernel.rank !== 5) {
+      throw new ValueError(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${x.rank}.`);
+    }
+    let y = preprocessConv3DInput(x, dataFormat);
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");
+    }
+    y = conv3d(y, kernel, strides, padding === "same" ? "same" : "valid", "NDHWC", dilationRate);
+    if (bias != null) {
+      y = biasAdd(y, bias);
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 4, 1, 2, 3]);
+    }
+    return y;
+  });
+}
+var BaseConv = class extends Layer {
+  constructor(rank, args) {
+    super(args);
+    this.bias = null;
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    BaseConv.verifyArgs(args);
+    this.rank = rank;
+    assertPositiveInteger(this.rank, "rank");
+    if (this.rank !== 1 && this.rank !== 2 && this.rank !== 3) {
+      throw new NotImplementedError(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);
+    }
+    this.kernelSize = normalizeArray(args.kernelSize, rank, "kernelSize");
+    this.strides = normalizeArray(args.strides == null ? 1 : args.strides, rank, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    checkPaddingMode(this.padding);
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.activation = getActivation(args.activation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.dilationRate = normalizeArray(args.dilationRate == null ? 1 : args.dilationRate, rank, "dilationRate");
+    if (this.rank === 1 && (Array.isArray(this.dilationRate) && this.dilationRate.length !== 1)) {
+      throw new ValueError(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+    } else if (this.rank === 2) {
+      if (typeof this.dilationRate === "number") {
+        this.dilationRate = [this.dilationRate, this.dilationRate];
+      } else if (this.dilationRate.length !== 2) {
+        throw new ValueError(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+      }
+    } else if (this.rank === 3) {
+      if (typeof this.dilationRate === "number") {
+        this.dilationRate = [this.dilationRate, this.dilationRate, this.dilationRate];
+      } else if (this.dilationRate.length !== 3) {
+        throw new ValueError(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+      }
+    }
+  }
+  static verifyArgs(args) {
+    assert2("kernelSize" in args, `required key 'kernelSize' not in config`);
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 3)) {
+      throw new ValueError(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+  getConfig() {
+    const config = {
+      kernelSize: this.kernelSize,
+      strides: this.strides,
+      padding: this.padding,
+      dataFormat: this.dataFormat,
+      dilationRate: this.dilationRate,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      biasConstraint: serializeConstraint(this.biasConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var Conv = class extends BaseConv {
+  constructor(rank, args) {
+    super(rank, args);
+    this.kernel = null;
+    Conv.verifyArgs(args);
+    this.filters = args.filters;
+    assertPositiveInteger(this.filters, "filters");
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([inputDim, this.filters]);
+    this.kernel = this.addWeight("kernel", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [{ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } }];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let outputs;
+      const biasValue = this.bias == null ? null : this.bias.read();
+      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());
+      if (fusedActivationName != null && this.rank === 2) {
+        outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate, fusedActivationName);
+      } else {
+        if (this.rank === 1) {
+          outputs = conv1dWithBias(inputs, this.kernel.read(), biasValue, this.strides[0], this.padding, this.dataFormat, this.dilationRate[0]);
+        } else if (this.rank === 2) {
+          outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);
+        } else if (this.rank === 3) {
+          outputs = conv3dWithBias(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);
+        } else {
+          throw new NotImplementedError("convolutions greater than 3D are not implemented yet.");
+        }
+        if (this.activation != null) {
+          outputs = this.activation.apply(outputs);
+        }
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const newSpace = [];
+    const space = this.dataFormat === "channelsLast" ? inputShape.slice(1, inputShape.length - 1) : inputShape.slice(2);
+    for (let i = 0; i < space.length; ++i) {
+      const newDim = convOutputLength(space[i], this.kernelSize[i], this.padding, this.strides[i], typeof this.dilationRate === "number" ? this.dilationRate : this.dilationRate[i]);
+      newSpace.push(newDim);
+    }
+    let outputShape = [inputShape[0]];
+    if (this.dataFormat === "channelsLast") {
+      outputShape = outputShape.concat(newSpace);
+      outputShape.push(this.filters);
+    } else {
+      outputShape.push(this.filters);
+      outputShape = outputShape.concat(newSpace);
+    }
+    return outputShape;
+  }
+  getConfig() {
+    const config = {
+      filters: this.filters,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  static verifyArgs(args) {
+    if (!("filters" in args) || typeof args.filters !== "number" || args.filters < 1) {
+      throw new ValueError(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(args.filters)}`);
+    }
+  }
+};
+var Conv2D2 = class extends Conv {
+  constructor(args) {
+    super(2, args);
+    Conv2D2.verifyArgs(args);
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 2)) {
+      throw new ValueError(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+};
+Conv2D2.className = "Conv2D";
+serialization_exports.registerClass(Conv2D2);
+var Conv3D2 = class extends Conv {
+  constructor(args) {
+    super(3, args);
+    Conv3D2.verifyArgs(args);
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number") {
+      if (!(Array.isArray(args.kernelSize) && (args.kernelSize.length === 1 || args.kernelSize.length === 3))) {
+        throw new ValueError(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(args.kernelSize)}.`);
+      }
+    }
+  }
+};
+Conv3D2.className = "Conv3D";
+serialization_exports.registerClass(Conv3D2);
+var Conv2DTranspose = class extends Conv2D2 {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+    if (this.padding !== "same" && this.padding !== "valid") {
+      throw new ValueError(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length !== 4) {
+      throw new ValueError("Input should have rank 4; Received input shape: " + JSON.stringify(inputShape));
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError("The channel dimension of the inputs should be defined. Found `None`.");
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);
+    this.kernel = this.addWeight("kernel", kernelShape, "float32", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [new InputSpec({ ndim: 4, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.shape.length !== 4) {
+        throw new ValueError(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);
+      }
+      const inputShape = input2.shape;
+      const batchSize = inputShape[0];
+      let hAxis;
+      let wAxis;
+      if (this.dataFormat === "channelsFirst") {
+        hAxis = 2;
+        wAxis = 3;
+      } else {
+        hAxis = 1;
+        wAxis = 2;
+      }
+      const height = inputShape[hAxis];
+      const width = inputShape[wAxis];
+      const kernelH = this.kernelSize[0];
+      const kernelW = this.kernelSize[1];
+      const strideH = this.strides[0];
+      const strideW = this.strides[1];
+      const outHeight = deconvLength(height, strideH, kernelH, this.padding);
+      const outWidth = deconvLength(width, strideW, kernelW, this.padding);
+      const outputShape = [batchSize, outHeight, outWidth, this.filters];
+      if (this.dataFormat !== "channelsLast") {
+        input2 = transpose(input2, [0, 2, 3, 1]);
+      }
+      let outputs = conv2dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);
+      if (this.dataFormat !== "channelsLast") {
+        outputs = transpose(outputs, [0, 3, 1, 2]);
+      }
+      if (this.bias != null) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    let channelAxis;
+    let heightAxis;
+    let widthAxis;
+    if (this.dataFormat === "channelsFirst") {
+      channelAxis = 1;
+      heightAxis = 2;
+      widthAxis = 3;
+    } else {
+      channelAxis = 3;
+      heightAxis = 1;
+      widthAxis = 2;
+    }
+    const kernelH = this.kernelSize[0];
+    const kernelW = this.kernelSize[1];
+    const strideH = this.strides[0];
+    const strideW = this.strides[1];
+    outputShape[channelAxis] = this.filters;
+    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);
+    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);
+    return outputShape;
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["dilationRate"];
+    return config;
+  }
+};
+Conv2DTranspose.className = "Conv2DTranspose";
+serialization_exports.registerClass(Conv2DTranspose);
+var Conv3DTranspose = class extends Conv3D2 {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+    if (this.padding !== "same" && this.padding !== "valid") {
+      throw new ValueError(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length !== 5) {
+      throw new ValueError("Input should have rank 5; Received input shape: " + JSON.stringify(inputShape));
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError("The channel dimension of the inputs should be defined. Found `None`.");
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);
+    this.kernel = this.addWeight("kernel", kernelShape, "float32", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [new InputSpec({ ndim: 5, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.shape.length !== 5) {
+        throw new ValueError(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);
+      }
+      const inputShape = input2.shape;
+      const batchSize = inputShape[0];
+      let hAxis;
+      let wAxis;
+      let dAxis;
+      if (this.dataFormat === "channelsFirst") {
+        dAxis = 2;
+        hAxis = 3;
+        wAxis = 4;
+      } else {
+        dAxis = 1;
+        hAxis = 2;
+        wAxis = 3;
+      }
+      const depth = inputShape[dAxis];
+      const height = inputShape[hAxis];
+      const width = inputShape[wAxis];
+      const kernelD = this.kernelSize[0];
+      const kernelH = this.kernelSize[1];
+      const kernelW = this.kernelSize[2];
+      const strideD = this.strides[0];
+      const strideH = this.strides[1];
+      const strideW = this.strides[2];
+      const outDepth = deconvLength(depth, strideD, kernelD, this.padding);
+      const outHeight = deconvLength(height, strideH, kernelH, this.padding);
+      const outWidth = deconvLength(width, strideW, kernelW, this.padding);
+      const outputShape = [batchSize, outDepth, outHeight, outWidth, this.filters];
+      if (this.dataFormat !== "channelsLast") {
+        input2 = transpose(input2, [0, 2, 3, 4, 1]);
+      }
+      let outputs = conv3dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);
+      if (this.dataFormat !== "channelsLast") {
+        outputs = transpose(outputs, [0, 4, 1, 2, 3]);
+      }
+      if (this.bias !== null) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation !== null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    let channelAxis;
+    let depthAxis;
+    let heightAxis;
+    let widthAxis;
+    if (this.dataFormat === "channelsFirst") {
+      channelAxis = 1;
+      depthAxis = 2;
+      heightAxis = 3;
+      widthAxis = 4;
+    } else {
+      channelAxis = 4;
+      depthAxis = 1;
+      heightAxis = 2;
+      widthAxis = 3;
+    }
+    const kernelD = this.kernelSize[0];
+    const kernelH = this.kernelSize[1];
+    const kernelW = this.kernelSize[2];
+    const strideD = this.strides[0];
+    const strideH = this.strides[1];
+    const strideW = this.strides[2];
+    outputShape[channelAxis] = this.filters;
+    outputShape[depthAxis] = deconvLength(outputShape[depthAxis], strideD, kernelD, this.padding);
+    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);
+    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);
+    return outputShape;
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["dilationRate"];
+    return config;
+  }
+};
+Conv3DTranspose.className = "Conv3DTranspose";
+serialization_exports.registerClass(Conv3DTranspose);
+var SeparableConv = class extends Conv {
+  constructor(rank, config) {
+    super(rank, config);
+    this.DEFAULT_DEPTHWISE_INITIALIZER = "glorotUniform";
+    this.DEFAULT_POINTWISE_INITIALIZER = "glorotUniform";
+    this.depthwiseKernel = null;
+    this.pointwiseKernel = null;
+    if (config.filters == null) {
+      throw new ValueError("The `filters` configuration field is required by SeparableConv, but is unspecified.");
+    }
+    if (config.kernelInitializer != null || config.kernelRegularizer != null || config.kernelConstraint != null) {
+      throw new ValueError("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");
+    }
+    if (config.padding != null && config.padding !== "same" && config.padding !== "valid") {
+      throw new ValueError(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(config.padding)}`);
+    }
+    this.depthMultiplier = config.depthMultiplier == null ? 1 : config.depthMultiplier;
+    this.depthwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_DEPTHWISE_INITIALIZER);
+    this.depthwiseRegularizer = getRegularizer(config.depthwiseRegularizer);
+    this.depthwiseConstraint = getConstraint(config.depthwiseConstraint);
+    this.pointwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_POINTWISE_INITIALIZER);
+    this.pointwiseRegularizer = getRegularizer(config.pointwiseRegularizer);
+    this.pointwiseConstraint = getConstraint(config.pointwiseConstraint);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < this.rank + 2) {
+      throw new ValueError(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank + 2}, but received input shape: ${JSON.stringify(inputShape)}`);
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {
+      throw new ValueError(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(inputShape[channelAxis])}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const depthwiseKernelShape = this.kernelSize.concat([inputDim, this.depthMultiplier]);
+    const pointwiseKernelShape = [];
+    for (let i = 0; i < this.rank; ++i) {
+      pointwiseKernelShape.push(1);
+    }
+    pointwiseKernelShape.push(inputDim * this.depthMultiplier, this.filters);
+    const trainable = true;
+    this.depthwiseKernel = this.addWeight("depthwise_kernel", depthwiseKernelShape, "float32", this.depthwiseInitializer, this.depthwiseRegularizer, trainable, this.depthwiseConstraint);
+    this.pointwiseKernel = this.addWeight("pointwise_kernel", pointwiseKernelShape, "float32", this.pointwiseInitializer, this.pointwiseRegularizer, trainable, this.pointwiseConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, trainable, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.inputSpec = [new InputSpec({ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let output;
+      if (this.rank === 1) {
+        throw new NotImplementedError("1D separable convolution is not implemented yet.");
+      } else if (this.rank === 2) {
+        if (this.dataFormat === "channelsFirst") {
+          inputs = transpose(inputs, [0, 2, 3, 1]);
+        }
+        output = separableConv2d(inputs, this.depthwiseKernel.read(), this.pointwiseKernel.read(), this.strides, this.padding, this.dilationRate, "NHWC");
+      }
+      if (this.useBias) {
+        output = biasAdd(output, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        output = this.activation.apply(output);
+      }
+      if (this.dataFormat === "channelsFirst") {
+        output = transpose(output, [0, 3, 1, 2]);
+      }
+      return output;
+    });
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    delete config["kernelInitializer"];
+    delete config["kernelRegularizer"];
+    delete config["kernelConstraint"];
+    config["depthwiseInitializer"] = serializeInitializer(this.depthwiseInitializer);
+    config["pointwiseInitializer"] = serializeInitializer(this.pointwiseInitializer);
+    config["depthwiseRegularizer"] = serializeRegularizer(this.depthwiseRegularizer);
+    config["pointwiseRegularizer"] = serializeRegularizer(this.pointwiseRegularizer);
+    config["depthwiseConstraint"] = serializeConstraint(this.depthwiseConstraint);
+    config["pointwiseConstraint"] = serializeConstraint(this.pointwiseConstraint);
+    return config;
+  }
+};
+SeparableConv.className = "SeparableConv";
+var SeparableConv2D = class extends SeparableConv {
+  constructor(args) {
+    super(2, args);
+  }
+};
+SeparableConv2D.className = "SeparableConv2D";
+serialization_exports.registerClass(SeparableConv2D);
+var Conv1D = class extends Conv {
+  constructor(args) {
+    super(1, args);
+    Conv1D.verifyArgs(args);
+    this.inputSpec = [{ ndim: 3 }];
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    delete config["dataFormat"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 1)) {
+      throw new ValueError(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+};
+Conv1D.className = "Conv1D";
+serialization_exports.registerClass(Conv1D);
+var Cropping2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    if (typeof args.cropping === "number") {
+      this.cropping = [[args.cropping, args.cropping], [args.cropping, args.cropping]];
+    } else if (typeof args.cropping[0] === "number") {
+      this.cropping = [
+        [args.cropping[0], args.cropping[0]],
+        [args.cropping[1], args.cropping[1]]
+      ];
+    } else {
+      this.cropping = args.cropping;
+    }
+    this.dataFormat = args.dataFormat === void 0 ? "channelsLast" : args.dataFormat;
+    this.inputSpec = [{ ndim: 4 }];
+  }
+  computeOutputShape(inputShape) {
+    if (this.dataFormat === "channelsFirst") {
+      return [
+        inputShape[0],
+        inputShape[1],
+        inputShape[2] - this.cropping[0][0] - this.cropping[0][1],
+        inputShape[3] - this.cropping[1][0] - this.cropping[1][1]
+      ];
+    } else {
+      return [
+        inputShape[0],
+        inputShape[1] - this.cropping[0][0] - this.cropping[0][1],
+        inputShape[2] - this.cropping[1][0] - this.cropping[1][1],
+        inputShape[3]
+      ];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[1] - this.cropping[0][0] - this.cropping[0][1], 2);
+        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[2] - this.cropping[1][1] - this.cropping[1][0], 3);
+      } else {
+        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[2] - this.cropping[0][0] - this.cropping[0][1], 3);
+        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[3] - this.cropping[1][1] - this.cropping[1][0], 4);
+      }
+    });
+  }
+  getConfig() {
+    const config = { cropping: this.cropping, dataFormat: this.dataFormat };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Cropping2D.className = "Cropping2D";
+serialization_exports.registerClass(Cropping2D);
+var UpSampling2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_SIZE = [2, 2];
+    this.inputSpec = [{ ndim: 4 }];
+    this.size = args.size == null ? this.DEFAULT_SIZE : args.size;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.interpolation = args.interpolation == null ? "nearest" : args.interpolation;
+    checkInterpolationFormat(this.interpolation);
+  }
+  computeOutputShape(inputShape) {
+    if (this.dataFormat === "channelsFirst") {
+      const height = inputShape[2] == null ? null : this.size[0] * inputShape[2];
+      const width = inputShape[3] == null ? null : this.size[1] * inputShape[3];
+      return [inputShape[0], inputShape[1], height, width];
+    } else {
+      const height = inputShape[1] == null ? null : this.size[0] * inputShape[1];
+      const width = inputShape[2] == null ? null : this.size[1] * inputShape[2];
+      return [inputShape[0], height, width, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      if (this.dataFormat === "channelsFirst") {
+        input2 = transpose(input2, [0, 2, 3, 1]);
+        const height = this.size[0] * inputShape[2];
+        const width = this.size[1] * inputShape[3];
+        const resized = this.interpolation === "nearest" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);
+        return transpose(resized, [0, 3, 1, 2]);
+      } else {
+        const height = this.size[0] * inputShape[1];
+        const width = this.size[1] * inputShape[2];
+        return this.interpolation === "nearest" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);
+      }
+    });
+  }
+  getConfig() {
+    const config = {
+      size: this.size,
+      dataFormat: this.dataFormat,
+      interpolation: this.interpolation
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+UpSampling2D.className = "UpSampling2D";
+serialization_exports.registerClass(UpSampling2D);
+function depthwiseConv2d3(x, depthwiseKernel, strides = [1, 1], padding = "valid", dataFormat, dilationRate) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    let y = preprocessConv2DInput(x, dataFormat);
+    if (x.rank !== 4) {
+      throw new ValueError(`Input for depthwiseConv2d is required to be 4-D, but is instead ${x.rank}-D`);
+    }
+    if (depthwiseKernel.rank !== 4) {
+      throw new ValueError(`depthwiseKernel is required to be 4-D, but is instead ${depthwiseKernel.rank}-D`);
+    }
+    y = depthwiseConv2d(y, depthwiseKernel, strides, padding === "same" ? "same" : "valid", "NHWC", dilationRate);
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+var DepthwiseConv2D = class extends BaseConv {
+  constructor(args) {
+    super(2, args);
+    this.depthwiseKernel = null;
+    this.depthMultiplier = args.depthMultiplier == null ? 1 : args.depthMultiplier;
+    this.depthwiseInitializer = getInitializer(args.depthwiseInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.depthwiseConstraint = getConstraint(args.depthwiseConstraint);
+    this.depthwiseRegularizer = getRegularizer(args.depthwiseRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < 4) {
+      throw new ValueError(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(inputShape)}.`);
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : 3;
+    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {
+      throw new ValueError(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${inputShape[channelAxis]}).`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const depthwiseKernelShape = [
+      this.kernelSize[0],
+      this.kernelSize[1],
+      inputDim,
+      this.depthMultiplier
+    ];
+    this.depthwiseKernel = this.addWeight("depthwise_kernel", depthwiseKernelShape, null, this.depthwiseInitializer, this.depthwiseRegularizer, true, this.depthwiseConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [inputDim * this.depthMultiplier], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let outputs = depthwiseConv2d3(inputs, this.depthwiseKernel.read(), this.strides, this.padding, this.dataFormat, null);
+      if (this.useBias) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const rows = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    const cols = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    const outFilters = this.dataFormat === "channelsFirst" ? inputShape[1] * this.depthMultiplier : inputShape[3] * this.depthMultiplier;
+    const outRows = convOutputLength(rows, this.kernelSize[0], this.padding, this.strides[0]);
+    const outCols = convOutputLength(cols, this.kernelSize[1], this.padding, this.strides[1]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], outFilters, outRows, outCols];
+    } else {
+      return [inputShape[0], outRows, outCols, outFilters];
+    }
+  }
+  getConfig() {
+    const config = super.getConfig();
+    config["depthMultiplier"] = this.depthMultiplier;
+    config["depthwiseInitializer"] = serializeInitializer(this.depthwiseInitializer);
+    config["depthwiseRegularizer"] = serializeRegularizer(this.depthwiseRegularizer);
+    config["depthwiseConstraint"] = serializeConstraint(this.depthwiseRegularizer);
+    return config;
+  }
+};
+DepthwiseConv2D.className = "DepthwiseConv2D";
+serialization_exports.registerClass(DepthwiseConv2D);
+function standardizeArgs(inputs, initialState, constants, numConstants) {
+  if (Array.isArray(inputs)) {
+    if (initialState != null || constants != null) {
+      throw new ValueError("When inputs is an array, neither initialState or constants should be provided");
+    }
+    if (numConstants != null) {
+      constants = inputs.slice(inputs.length - numConstants, inputs.length);
+      inputs = inputs.slice(0, inputs.length - numConstants);
+    }
+    if (inputs.length > 1) {
+      initialState = inputs.slice(1, inputs.length);
+    }
+    inputs = inputs[0];
+  }
+  function toListOrNull(x) {
+    if (x == null || Array.isArray(x)) {
+      return x;
+    } else {
+      return [x];
+    }
+  }
+  initialState = toListOrNull(initialState);
+  constants = toListOrNull(constants);
+  return { inputs, initialState, constants };
+}
+function rnn(stepFunction, inputs, initialStates, goBackwards = false, mask, constants, unroll = false, needPerStepOutputs = false) {
+  return tidy(() => {
+    const ndim = inputs.shape.length;
+    if (ndim < 3) {
+      throw new ValueError(`Input should be at least 3D, but is ${ndim}D.`);
+    }
+    const axes = [1, 0].concat(range2(2, ndim));
+    inputs = transpose(inputs, axes);
+    if (constants != null) {
+      throw new NotImplementedError("The rnn() functoin of the deeplearn.js backend does not support constants yet.");
+    }
+    if (unroll) {
+      console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend.");
+    }
+    if (mask != null) {
+      mask = cast(cast(mask, "bool"), "float32");
+      if (mask.rank === ndim - 1) {
+        mask = expandDims(mask, -1);
+      }
+      mask = transpose(mask, axes);
+    }
+    if (goBackwards) {
+      inputs = reverse(inputs, 0);
+      if (mask != null) {
+        mask = reverse(mask, 0);
+      }
+    }
+    const perStepOutputs = [];
+    let lastOutput;
+    let states = initialStates;
+    const timeSteps = inputs.shape[0];
+    const perStepInputs = unstack(inputs);
+    let perStepMasks;
+    if (mask != null) {
+      perStepMasks = unstack(mask);
+    }
+    for (let t = 0; t < timeSteps; ++t) {
+      const currentInput = perStepInputs[t];
+      const stepOutputs = tidy(() => stepFunction(currentInput, states));
+      if (mask == null) {
+        lastOutput = stepOutputs[0];
+        states = stepOutputs[1];
+      } else {
+        const maskedOutputs = tidy(() => {
+          const stepMask = perStepMasks[t];
+          const negStepMask = sub(onesLike(stepMask), stepMask);
+          const output = add2(mul(stepOutputs[0], stepMask), mul(states[0], negStepMask));
+          const newStates = states.map((state, i) => {
+            return add2(mul(stepOutputs[1][i], stepMask), mul(state, negStepMask));
+          });
+          return { output, newStates };
+        });
+        lastOutput = maskedOutputs.output;
+        states = maskedOutputs.newStates;
+      }
+      if (needPerStepOutputs) {
+        perStepOutputs.push(lastOutput);
+      }
+    }
+    let outputs;
+    if (needPerStepOutputs) {
+      const axis = 1;
+      outputs = stack(perStepOutputs, axis);
+    }
+    return [lastOutput, outputs, states];
+  });
+}
+var RNN = class extends Layer {
+  constructor(args) {
+    super(args);
+    let cell;
+    if (args.cell == null) {
+      throw new ValueError("cell property is missing for the constructor of RNN.");
+    } else if (Array.isArray(args.cell)) {
+      cell = new StackedRNNCells({ cells: args.cell });
+    } else {
+      cell = args.cell;
+    }
+    if (cell.stateSize == null) {
+      throw new ValueError("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");
+    }
+    this.cell = cell;
+    this.returnSequences = args.returnSequences == null ? false : args.returnSequences;
+    this.returnState = args.returnState == null ? false : args.returnState;
+    this.goBackwards = args.goBackwards == null ? false : args.goBackwards;
+    this._stateful = args.stateful == null ? false : args.stateful;
+    this.unroll = args.unroll == null ? false : args.unroll;
+    this.supportsMasking = true;
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+    this.stateSpec = null;
+    this.states_ = null;
+    this.numConstants = null;
+    this.keptStates = [];
+  }
+  getStates() {
+    if (this.states_ == null) {
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      return range2(0, numStates).map((x) => null);
+    } else {
+      return this.states_;
+    }
+  }
+  setStates(states) {
+    this.states_ = states;
+  }
+  computeOutputShape(inputShape) {
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    let stateSize = this.cell.stateSize;
+    if (!Array.isArray(stateSize)) {
+      stateSize = [stateSize];
+    }
+    const outputDim = stateSize[0];
+    let outputShape;
+    if (this.returnSequences) {
+      outputShape = [inputShape[0], inputShape[1], outputDim];
+    } else {
+      outputShape = [inputShape[0], outputDim];
+    }
+    if (this.returnState) {
+      const stateShape = [];
+      for (const dim of stateSize) {
+        stateShape.push([inputShape[0], dim]);
+      }
+      return [outputShape].concat(stateShape);
+    } else {
+      return outputShape;
+    }
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (Array.isArray(mask)) {
+        mask = mask[0];
+      }
+      const outputMask = this.returnSequences ? mask : null;
+      if (this.returnState) {
+        const stateMask = this.states.map((s) => null);
+        return [outputMask].concat(stateMask);
+      } else {
+        return outputMask;
+      }
+    });
+  }
+  get states() {
+    if (this.states_ == null) {
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      const output = [];
+      for (let i = 0; i < numStates; ++i) {
+        output.push(null);
+      }
+      return output;
+    } else {
+      return this.states_;
+    }
+  }
+  set states(s) {
+    this.states_ = s;
+  }
+  build(inputShape) {
+    const constantShape = null;
+    if (this.numConstants != null) {
+      throw new NotImplementedError("Constants support is not implemented in RNN yet.");
+    }
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    const batchSize = this.stateful ? inputShape[0] : null;
+    const inputDim = inputShape.slice(2);
+    this.inputSpec[0] = new InputSpec({ shape: [batchSize, null, ...inputDim] });
+    const stepInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    if (constantShape != null) {
+      throw new NotImplementedError("Constants support is not implemented in RNN yet.");
+    } else {
+      this.cell.build(stepInputShape);
+    }
+    let stateSize;
+    if (Array.isArray(this.cell.stateSize)) {
+      stateSize = this.cell.stateSize;
+    } else {
+      stateSize = [this.cell.stateSize];
+    }
+    if (this.stateSpec != null) {
+      if (!util_exports.arraysEqual(this.stateSpec.map((spec) => spec.shape[spec.shape.length - 1]), stateSize)) {
+        throw new ValueError(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`);
+      }
+    } else {
+      this.stateSpec = stateSize.map((dim) => new InputSpec({ shape: [null, dim] }));
+    }
+    if (this.stateful) {
+      this.resetStates();
+    }
+  }
+  resetStates(states, training = false) {
+    tidy(() => {
+      if (!this.stateful) {
+        throw new AttributeError("Cannot call resetStates() on an RNN Layer that is not stateful.");
+      }
+      const batchSize = this.inputSpec[0].shape[0];
+      if (batchSize == null) {
+        throw new ValueError("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+      }
+      if (this.states_ == null) {
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));
+        } else {
+          this.states_ = [zeros([batchSize, this.cell.stateSize])];
+        }
+      } else if (states == null) {
+        dispose(this.states_);
+        if (this.keptStates != null) {
+          dispose(this.keptStates);
+          this.keptStates = [];
+        }
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));
+        } else {
+          this.states_[0] = zeros([batchSize, this.cell.stateSize]);
+        }
+      } else {
+        if (!Array.isArray(states)) {
+          states = [states];
+        }
+        if (states.length !== this.states_.length) {
+          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);
+        }
+        if (training === true) {
+          this.keptStates.push(this.states_.slice());
+        } else {
+          dispose(this.states_);
+        }
+        for (let index = 0; index < this.states_.length; ++index) {
+          const value = states[index];
+          const dim = Array.isArray(this.cell.stateSize) ? this.cell.stateSize[index] : this.cell.stateSize;
+          const expectedShape = [batchSize, dim];
+          if (!util_exports.arraysEqual(value.shape, expectedShape)) {
+            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);
+          }
+          this.states_[index] = value;
+        }
+      }
+      this.states_ = this.states_.map((state) => keep(state.clone()));
+    });
+  }
+  apply(inputs, kwargs) {
+    let initialState = kwargs == null ? null : kwargs["initialState"];
+    let constants = kwargs == null ? null : kwargs["constants"];
+    if (kwargs == null) {
+      kwargs = {};
+    }
+    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);
+    inputs = standardized.inputs;
+    initialState = standardized.initialState;
+    constants = standardized.constants;
+    let additionalInputs = [];
+    let additionalSpecs = [];
+    if (initialState != null) {
+      kwargs["initialState"] = initialState;
+      additionalInputs = additionalInputs.concat(initialState);
+      this.stateSpec = [];
+      for (const state of initialState) {
+        this.stateSpec.push(new InputSpec({ shape: state.shape }));
+      }
+      additionalSpecs = additionalSpecs.concat(this.stateSpec);
+    }
+    if (constants != null) {
+      kwargs["constants"] = constants;
+      additionalInputs = additionalInputs.concat(constants);
+      this.numConstants = constants.length;
+    }
+    const isTensor2 = additionalInputs[0] instanceof SymbolicTensor;
+    if (isTensor2) {
+      const fullInput = [inputs].concat(additionalInputs);
+      const fullInputSpec = this.inputSpec.concat(additionalSpecs);
+      const originalInputSpec = this.inputSpec;
+      this.inputSpec = fullInputSpec;
+      const output = super.apply(fullInput, kwargs);
+      this.inputSpec = originalInputSpec;
+      return output;
+    } else {
+      return super.apply(inputs, kwargs);
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      let initialState = kwargs == null ? null : kwargs["initialState"];
+      inputs = getExactlyOneTensor(inputs);
+      if (initialState == null) {
+        if (this.stateful) {
+          initialState = this.states_;
+        } else {
+          initialState = this.getInitialState(inputs);
+        }
+      }
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      if (initialState.length !== numStates) {
+        throw new ValueError(`RNN Layer has ${numStates} state(s) but was passed ${initialState.length} initial state(s).`);
+      }
+      if (this.unroll) {
+        console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");
+      }
+      const cellCallKwargs = { training };
+      const step5 = (inputs2, states2) => {
+        const outputs2 = this.cell.call([inputs2].concat(states2), cellCallKwargs);
+        return [outputs2[0], outputs2.slice(1)];
+      };
+      const rnnOutputs = rnn(step5, inputs, initialState, this.goBackwards, mask, null, this.unroll, this.returnSequences);
+      const lastOutput = rnnOutputs[0];
+      const outputs = rnnOutputs[1];
+      const states = rnnOutputs[2];
+      if (this.stateful) {
+        this.resetStates(states, training);
+      }
+      const output = this.returnSequences ? outputs : lastOutput;
+      if (this.returnState) {
+        return [output].concat(states);
+      } else {
+        return output;
+      }
+    });
+  }
+  getInitialState(inputs) {
+    return tidy(() => {
+      let initialState = zeros(inputs.shape);
+      initialState = sum2(initialState, [1, 2]);
+      initialState = expandDims2(initialState);
+      if (Array.isArray(this.cell.stateSize)) {
+        return this.cell.stateSize.map((dim) => dim > 1 ? tile2(initialState, [1, dim]) : initialState);
+      } else {
+        return this.cell.stateSize > 1 ? [tile2(initialState, [1, this.cell.stateSize])] : [initialState];
+      }
+    });
+  }
+  get trainableWeights() {
+    if (!this.trainable) {
+      return [];
+    }
+    return this.cell.trainableWeights;
+  }
+  get nonTrainableWeights() {
+    if (!this.trainable) {
+      return this.cell.weights;
+    }
+    return this.cell.nonTrainableWeights;
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.cell != null) {
+      this.cell.setFastWeightInitDuringBuild(value);
+    }
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      returnSequences: this.returnSequences,
+      returnState: this.returnState,
+      goBackwards: this.goBackwards,
+      stateful: this.stateful,
+      unroll: this.unroll
+    };
+    if (this.numConstants != null) {
+      config["numConstants"] = this.numConstants;
+    }
+    const cellConfig = this.cell.getConfig();
+    if (this.getClassName() === RNN.className) {
+      config["cell"] = {
+        "className": this.cell.getClassName(),
+        "config": cellConfig
+      };
+    }
+    return Object.assign(Object.assign(Object.assign({}, cellConfig), baseConfig), config);
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const cellConfig = config["cell"];
+    const cell = deserialize(cellConfig, customObjects);
+    return new cls(Object.assign(config, { cell }));
+  }
+};
+RNN.className = "RNN";
+serialization_exports.registerClass(RNN);
+var RNNCell = class extends Layer {
+};
+var SimpleRNNCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    this.units = args.units;
+    assertPositiveInteger(this.units, `units`);
+    this.activation = getActivation(args.activation == null ? this.DEFAULT_ACTIVATION : args.activation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.stateSize = this.units;
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    this.kernel = this.addWeight("kernel", [inputShape[inputShape.length - 1], this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (inputs.length !== 2) {
+        throw new ValueError(`SimpleRNNCell expects 2 input Tensors, got ${inputs.length}.`);
+      }
+      let prevOutput = inputs[1];
+      inputs = inputs[0];
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(prevOutput),
+          rate: this.recurrentDropout,
+          training,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      let h;
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      if (dpMask != null) {
+        h = dot2(mul(inputs, dpMask), this.kernel.read());
+      } else {
+        h = dot2(inputs, this.kernel.read());
+      }
+      if (this.bias != null) {
+        h = biasAdd(h, this.bias.read());
+      }
+      if (recDpMask != null) {
+        prevOutput = mul(prevOutput, recDpMask);
+      }
+      let output = add2(h, dot2(prevOutput, this.recurrentKernel.read()));
+      if (this.activation != null) {
+        output = this.activation.apply(output);
+      }
+      return [output, output];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+SimpleRNNCell.className = "SimpleRNNCell";
+serialization_exports.registerClass(SimpleRNNCell);
+var SimpleRNN = class extends RNN {
+  constructor(args) {
+    args.cell = new SimpleRNNCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+SimpleRNN.className = "SimpleRNN";
+serialization_exports.registerClass(SimpleRNN);
+var GRUCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    if (args.resetAfter) {
+      throw new ValueError(`GRUCell does not support reset_after parameter set to true.`);
+    }
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);
+    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.implementation = args.implementation;
+    this.stateSize = this.units;
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const inputDim = inputShape[inputShape.length - 1];
+    this.kernel = this.addWeight("kernel", [inputDim, this.units * 3], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units * 3], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.units * 3], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (inputs.length !== 2) {
+        throw new ValueError(`GRUCell expects 2 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      let hTMinus1 = inputs[1];
+      inputs = inputs[0];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          count: 3,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: 3,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      let z;
+      let r;
+      let hh;
+      if (0 < this.dropout && this.dropout < 1) {
+        inputs = mul(inputs, dpMask[0]);
+      }
+      let matrixX = dot2(inputs, this.kernel.read());
+      if (this.useBias) {
+        matrixX = biasAdd(matrixX, this.bias.read());
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {
+        hTMinus1 = mul(hTMinus1, recDpMask[0]);
+      }
+      const recurrentKernelValue = this.recurrentKernel.read();
+      const [rk1, rk2] = split(recurrentKernelValue, [2 * this.units, this.units], recurrentKernelValue.rank - 1);
+      const matrixInner = dot2(hTMinus1, rk1);
+      const [xZ, xR, xH] = split(matrixX, 3, matrixX.rank - 1);
+      const [recurrentZ, recurrentR] = split(matrixInner, 2, matrixInner.rank - 1);
+      z = this.recurrentActivation.apply(add2(xZ, recurrentZ));
+      r = this.recurrentActivation.apply(add2(xR, recurrentR));
+      const recurrentH = dot2(mul(r, hTMinus1), rk2);
+      hh = this.activation.apply(add2(xH, recurrentH));
+      const h = add2(mul(z, hTMinus1), mul(add2(1, neg(z)), hh));
+      return [h, h];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      recurrentActivation: serializeActivation(this.recurrentActivation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout,
+      implementation: this.implementation,
+      resetAfter: false
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+GRUCell.className = "GRUCell";
+serialization_exports.registerClass(GRUCell);
+var GRU = class extends RNN {
+  constructor(args) {
+    if (args.implementation === 0) {
+      console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.");
+    }
+    args.cell = new GRUCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    if (config["implmentation"] === 0) {
+      config["implementation"] = 1;
+    }
+    return new cls(config);
+  }
+};
+GRU.className = "GRU";
+serialization_exports.registerClass(GRU);
+var LSTMCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);
+    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.unitForgetBias = args.unitForgetBias;
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.implementation = args.implementation;
+    this.stateSize = [this.units, this.units];
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    var _a;
+    inputShape = getExactlyOneShape(inputShape);
+    const inputDim = inputShape[inputShape.length - 1];
+    this.kernel = this.addWeight("kernel", [inputDim, this.units * 4], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units * 4], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    let biasInitializer;
+    if (this.useBias) {
+      if (this.unitForgetBias) {
+        const capturedBiasInit = this.biasInitializer;
+        const capturedUnits = this.units;
+        biasInitializer = new (_a = class CustomInit extends Initializer {
+          apply(shape, dtype) {
+            const bI = capturedBiasInit.apply([capturedUnits]);
+            const bF = new Ones().apply([capturedUnits]);
+            const bCAndH = capturedBiasInit.apply([capturedUnits * 2]);
+            return concatAlongFirstAxis(concatAlongFirstAxis(bI, bF), bCAndH);
+          }
+        }, _a.className = "CustomInit", _a)();
+      } else {
+        biasInitializer = this.biasInitializer;
+      }
+      this.bias = this.addWeight("bias", [this.units * 4], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      inputs = inputs;
+      if (inputs.length !== 3) {
+        throw new ValueError(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      let hTMinus1 = inputs[1];
+      const cTMinus1 = inputs[2];
+      inputs = inputs[0];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          count: 4,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: 4,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      let i;
+      let f;
+      let c;
+      let o;
+      if (0 < this.dropout && this.dropout < 1) {
+        inputs = mul(inputs, dpMask[0]);
+      }
+      let z = dot2(inputs, this.kernel.read());
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {
+        hTMinus1 = mul(hTMinus1, recDpMask[0]);
+      }
+      z = add2(z, dot2(hTMinus1, this.recurrentKernel.read()));
+      if (this.useBias) {
+        z = biasAdd(z, this.bias.read());
+      }
+      const [z0, z1, z2, z3] = split(z, 4, z.rank - 1);
+      i = this.recurrentActivation.apply(z0);
+      f = this.recurrentActivation.apply(z1);
+      c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(z2)));
+      o = this.recurrentActivation.apply(z3);
+      const h = mul(o, this.activation.apply(c));
+      return [h, h, c];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      recurrentActivation: serializeActivation(this.recurrentActivation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      unitForgetBias: this.unitForgetBias,
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout,
+      implementation: this.implementation
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+LSTMCell.className = "LSTMCell";
+serialization_exports.registerClass(LSTMCell);
+var LSTM = class extends RNN {
+  constructor(args) {
+    if (args.implementation === 0) {
+      console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.");
+    }
+    args.cell = new LSTMCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    if (config["implmentation"] === 0) {
+      config["implementation"] = 1;
+    }
+    return new cls(config);
+  }
+};
+LSTM.className = "LSTM";
+serialization_exports.registerClass(LSTM);
+var StackedRNNCells = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.cells = args.cells;
+  }
+  get stateSize() {
+    const stateSize = [];
+    for (const cell of this.cells.slice().reverse()) {
+      if (Array.isArray(cell.stateSize)) {
+        stateSize.push(...cell.stateSize);
+      } else {
+        stateSize.push(cell.stateSize);
+      }
+    }
+    return stateSize;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      let states = inputs.slice(1);
+      const nestedStates = [];
+      for (const cell of this.cells.slice().reverse()) {
+        if (Array.isArray(cell.stateSize)) {
+          nestedStates.push(states.splice(0, cell.stateSize.length));
+        } else {
+          nestedStates.push(states.splice(0, 1));
+        }
+      }
+      nestedStates.reverse();
+      const newNestedStates = [];
+      let callInputs;
+      for (let i = 0; i < this.cells.length; ++i) {
+        const cell = this.cells[i];
+        states = nestedStates[i];
+        if (i === 0) {
+          callInputs = [inputs[0]].concat(states);
+        } else {
+          callInputs = [callInputs[0]].concat(states);
+        }
+        callInputs = cell.call(callInputs, kwargs);
+        newNestedStates.push(callInputs.slice(1));
+      }
+      states = [];
+      for (const cellStates of newNestedStates.slice().reverse()) {
+        states.push(...cellStates);
+      }
+      return [callInputs[0]].concat(states);
+    });
+  }
+  build(inputShape) {
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    let outputDim;
+    this.cells.forEach((cell, i) => {
+      nameScope(`RNNCell_${i}`, () => {
+        cell.build(inputShape);
+        if (Array.isArray(cell.stateSize)) {
+          outputDim = cell.stateSize[0];
+        } else {
+          outputDim = cell.stateSize;
+        }
+        inputShape = [inputShape[0], outputDim];
+      });
+    });
+    this.built = true;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const getCellConfig = (cell) => {
+      return {
+        "className": cell.getClassName(),
+        "config": cell.getConfig()
+      };
+    };
+    const cellConfigs = this.cells.map(getCellConfig);
+    const config = { "cells": cellConfigs };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const cells = [];
+    for (const cellConfig of config["cells"]) {
+      cells.push(deserialize(cellConfig, customObjects));
+    }
+    return new cls({ cells });
+  }
+  get trainableWeights() {
+    if (!this.trainable) {
+      return [];
+    }
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.trainableWeights);
+    }
+    return weights;
+  }
+  get nonTrainableWeights() {
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.nonTrainableWeights);
+    }
+    if (!this.trainable) {
+      const trainableWeights = [];
+      for (const cell of this.cells) {
+        trainableWeights.push(...cell.trainableWeights);
+      }
+      return trainableWeights.concat(weights);
+    }
+    return weights;
+  }
+  getWeights() {
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.weights);
+    }
+    return batchGetValue(weights);
+  }
+  setWeights(weights) {
+    const tuples = [];
+    for (const cell of this.cells) {
+      const numParams = cell.weights.length;
+      const inputWeights = weights.splice(numParams);
+      for (let i = 0; i < cell.weights.length; ++i) {
+        tuples.push([cell.weights[i], inputWeights[i]]);
+      }
+    }
+    batchSetValue(tuples);
+  }
+};
+StackedRNNCells.className = "StackedRNNCells";
+serialization_exports.registerClass(StackedRNNCells);
+function generateDropoutMask(args) {
+  const { ones: ones4, rate, training = false, count: count2 = 1, dropoutFunc } = args;
+  const droppedInputs = () => dropoutFunc != null ? dropoutFunc(ones4(), rate) : dropout2(ones4(), rate);
+  const createMask = () => inTrainPhase(droppedInputs, ones4, training);
+  if (!count2 || count2 <= 1) {
+    return keep(createMask().clone());
+  }
+  const masks = Array(count2).fill(void 0).map(createMask);
+  return masks.map((m) => keep(m.clone()));
+}
+var __rest = function(s, e) {
+  var t = {};
+  for (var p2 in s)
+    if (Object.prototype.hasOwnProperty.call(s, p2) && e.indexOf(p2) < 0)
+      t[p2] = s[p2];
+  if (s != null && typeof Object.getOwnPropertySymbols === "function")
+    for (var i = 0, p2 = Object.getOwnPropertySymbols(s); i < p2.length; i++) {
+      if (e.indexOf(p2[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p2[i]))
+        t[p2[i]] = s[p2[i]];
+    }
+  return t;
+};
+var ConvRNN2D = class extends RNN {
+  constructor(args) {
+    if (args.unroll) {
+      throw new NotImplementedError("Unrolling is not possible with convolutional RNNs.");
+    }
+    if (Array.isArray(args.cell)) {
+      throw new NotImplementedError("It is not possible at the moment to stack convolutional cells.");
+    }
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      if (kwargs && kwargs["constants"]) {
+        throw new ValueError("ConvRNN2D cell does not support constants");
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  computeOutputShape(inputShape) {
+    let outShape = this.computeSingleOutputShape(inputShape);
+    if (!this.returnSequences) {
+      outShape = [outShape[0], ...outShape.slice(2)];
+    }
+    if (this.returnState) {
+      outShape = [outShape, ...Array(2).fill([inputShape[0], ...outShape.slice(-3)])];
+    }
+    return outShape;
+  }
+  getInitialState(inputs) {
+    return tidy(() => {
+      const { stateSize } = this.cell;
+      const inputShape = inputs.shape;
+      const outputShape = this.computeSingleOutputShape(inputShape);
+      const stateShape = [outputShape[0], ...outputShape.slice(2)];
+      const initialState = zeros(stateShape);
+      if (Array.isArray(stateSize)) {
+        return Array(stateSize.length).fill(initialState);
+      }
+      return [initialState];
+    });
+  }
+  resetStates(states, training = false) {
+    tidy(() => {
+      if (!this.stateful) {
+        throw new AttributeError("Cannot call resetStates() on an RNN Layer that is not stateful.");
+      }
+      const inputShape = this.inputSpec[0].shape;
+      const outputShape = this.computeSingleOutputShape(inputShape);
+      const stateShape = [outputShape[0], ...outputShape.slice(2)];
+      const batchSize = inputShape[0];
+      if (batchSize == null) {
+        throw new ValueError("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+      }
+      if (this.getStates() == null) {
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));
+        } else {
+          this.states_ = [zeros(stateShape)];
+        }
+      } else if (states == null) {
+        dispose(this.states_);
+        if (this.keptStates != null) {
+          dispose(this.keptStates);
+          this.keptStates = [];
+        }
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));
+        } else {
+          this.states_[0] = zeros(stateShape);
+        }
+      } else {
+        if (!Array.isArray(states)) {
+          states = [states];
+        }
+        if (states.length !== this.states_.length) {
+          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);
+        }
+        if (training) {
+          this.keptStates.push(this.states_.slice());
+        } else {
+          dispose(this.states_);
+        }
+        for (let index = 0; index < this.states_.length; ++index) {
+          const value = states[index];
+          const expectedShape = stateShape;
+          if (!util_exports.arraysEqual(value.shape, expectedShape)) {
+            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);
+          }
+          this.states_[index] = value;
+        }
+      }
+      this.states_ = this.states_.map((state) => keep(state.clone()));
+    });
+  }
+  computeSingleOutputShape(inputShape) {
+    const { dataFormat, filters, kernelSize, padding, strides, dilationRate } = this.cell;
+    const isChannelsFirst = dataFormat === "channelsFirst";
+    const h = inputShape[isChannelsFirst ? 3 : 2];
+    const w = inputShape[isChannelsFirst ? 4 : 3];
+    const hOut = convOutputLength(h, kernelSize[0], padding, strides[0], dilationRate[0]);
+    const wOut = convOutputLength(w, kernelSize[1], padding, strides[1], dilationRate[1]);
+    const outShape = [
+      ...inputShape.slice(0, 2),
+      ...isChannelsFirst ? [filters, hOut, wOut] : [hOut, wOut, filters]
+    ];
+    return outShape;
+  }
+};
+ConvRNN2D.className = "ConvRNN2D";
+var ConvLSTM2DCell = class extends LSTMCell {
+  constructor(args) {
+    const { filters, kernelSize, strides, padding, dataFormat, dilationRate } = args;
+    super(Object.assign(Object.assign({}, args), { units: filters }));
+    this.filters = filters;
+    assertPositiveInteger(this.filters, "filters");
+    this.kernelSize = normalizeArray(kernelSize, 2, "kernelSize");
+    this.kernelSize.forEach((size) => assertPositiveInteger(size, "kernelSize"));
+    this.strides = normalizeArray(strides || 1, 2, "strides");
+    this.strides.forEach((stride) => assertPositiveInteger(stride, "strides"));
+    this.padding = padding || "valid";
+    checkPaddingMode(this.padding);
+    this.dataFormat = dataFormat || "channelsLast";
+    checkDataFormat(this.dataFormat);
+    this.dilationRate = normalizeArray(dilationRate || 1, 2, "dilationRate");
+    this.dilationRate.forEach((rate) => assertPositiveInteger(rate, "dilationRate"));
+  }
+  build(inputShape) {
+    var _a;
+    inputShape = getExactlyOneShape(inputShape);
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const numOfKernels = 4;
+    const kernelShape = this.kernelSize.concat([inputDim, this.filters * numOfKernels]);
+    this.kernel = this.addWeight("kernel", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    const recurrentKernelShape = this.kernelSize.concat([this.filters, this.filters * numOfKernels]);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", recurrentKernelShape, null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      let biasInitializer;
+      if (this.unitForgetBias) {
+        const init2 = this.biasInitializer;
+        const filters = this.filters;
+        biasInitializer = new (_a = class CustomInit extends Initializer {
+          apply(shape, dtype) {
+            const biasI = init2.apply([filters]);
+            const biasF = ones2([filters]);
+            const biasCAndO = init2.apply([filters * 2]);
+            return concatenate([biasI, biasF, biasCAndO]);
+          }
+        }, _a.className = "CustomInit", _a)();
+      } else {
+        biasInitializer = this.biasInitializer;
+      }
+      this.bias = this.addWeight("bias", [this.filters * numOfKernels], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (inputs.length !== 3) {
+        throw new ValueError(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      const training = kwargs["training"] || false;
+      const x = inputs[0];
+      const hTMinus1 = inputs[1];
+      const cTMinus1 = inputs[2];
+      const numOfKernels = 4;
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(x),
+          rate: this.dropout,
+          training,
+          count: numOfKernels,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dropoutMask = this.dropoutMask;
+      const applyDropout = (x2, mask, index) => {
+        if (!mask || !mask[index]) {
+          return x2;
+        }
+        return mul(mask[index], x2);
+      };
+      let xI = applyDropout(x, dropoutMask, 0);
+      let xF = applyDropout(x, dropoutMask, 1);
+      let xC = applyDropout(x, dropoutMask, 2);
+      let xO = applyDropout(x, dropoutMask, 3);
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: numOfKernels,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const recDropoutMask = this.recurrentDropoutMask;
+      let hI = applyDropout(hTMinus1, recDropoutMask, 0);
+      let hF = applyDropout(hTMinus1, recDropoutMask, 1);
+      let hC = applyDropout(hTMinus1, recDropoutMask, 2);
+      let hO = applyDropout(hTMinus1, recDropoutMask, 3);
+      const kernelChannelAxis = 3;
+      const [kernelI, kernelF, kernelC, kernelO] = split(this.kernel.read(), numOfKernels, kernelChannelAxis);
+      const [biasI, biasF, biasC, biasO] = this.useBias ? split(this.bias.read(), numOfKernels) : [null, null, null, null];
+      xI = this.inputConv(xI, kernelI, biasI, this.padding);
+      xF = this.inputConv(xF, kernelF, biasF, this.padding);
+      xC = this.inputConv(xC, kernelC, biasC, this.padding);
+      xO = this.inputConv(xO, kernelO, biasO, this.padding);
+      const [recKernelI, recKernelF, recKernelC, recKernelO] = split(this.recurrentKernel.read(), numOfKernels, kernelChannelAxis);
+      hI = this.recurrentConv(hI, recKernelI);
+      hF = this.recurrentConv(hF, recKernelF);
+      hC = this.recurrentConv(hC, recKernelC);
+      hO = this.recurrentConv(hO, recKernelO);
+      const i = this.recurrentActivation.apply(add2(xI, hI));
+      const f = this.recurrentActivation.apply(add2(xF, hF));
+      const c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(add2(xC, hC))));
+      const h = mul(this.recurrentActivation.apply(add2(xO, hO)), this.activation.apply(c));
+      return [h, h, c];
+    });
+  }
+  getConfig() {
+    const _a = super.getConfig(), { "units": _ } = _a, baseConfig = __rest(_a, ["units"]);
+    const config = {
+      filters: this.filters,
+      kernelSize: this.kernelSize,
+      padding: this.padding,
+      dataFormat: this.dataFormat,
+      dilationRate: this.dilationRate,
+      strides: this.strides
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+  inputConv(x, w, b, padding) {
+    const out = conv2d(x, w, this.strides, padding || "valid", this.dataFormat === "channelsFirst" ? "NCHW" : "NHWC", this.dilationRate);
+    if (b) {
+      return biasAdd(out, b, this.dataFormat);
+    }
+    return out;
+  }
+  recurrentConv(x, w) {
+    const strides = 1;
+    return conv2d(x, w, strides, "same", this.dataFormat === "channelsFirst" ? "NCHW" : "NHWC");
+  }
+};
+ConvLSTM2DCell.className = "ConvLSTM2DCell";
+serialization_exports.registerClass(ConvLSTM2DCell);
+var ConvLSTM2D = class extends ConvRNN2D {
+  constructor(args) {
+    const cell = new ConvLSTM2DCell(args);
+    super(Object.assign(Object.assign({}, args), { cell }));
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+ConvLSTM2D.className = "ConvLSTM2D";
+serialization_exports.registerClass(ConvLSTM2D);
+var Dropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.rate = Math.max(Math.min(args.rate, 1), 0);
+    this.noiseShape = args.noiseShape;
+    this.seed = args.seed;
+    this.supportsMasking = true;
+  }
+  getNoiseShape(input2) {
+    if (this.noiseShape == null) {
+      return this.noiseShape;
+    }
+    const inputShape = input2.shape;
+    const noiseShape = [];
+    for (let i = 0; i < this.noiseShape.length; ++i) {
+      noiseShape.push(this.noiseShape[i] == null ? inputShape[i] : this.noiseShape[i]);
+    }
+    return noiseShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      if (0 < this.rate && this.rate < 1) {
+        const training = kwargs["training"] == null ? false : kwargs["training"];
+        const noiseShape = this.getNoiseShape(input2);
+        const output = inTrainPhase(() => dropout2(input2, this.rate, noiseShape, this.seed), () => input2, training);
+        return output;
+      }
+      return inputs;
+    });
+  }
+  getConfig() {
+    const config = {
+      rate: this.rate,
+      noiseShape: this.noiseShape,
+      seed: this.seed
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  dispose() {
+    return super.dispose();
+  }
+};
+Dropout.className = "Dropout";
+serialization_exports.registerClass(Dropout);
+var SpatialDropout1D = class extends Dropout {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [{ ndim: 3 }];
+  }
+  getNoiseShape(input2) {
+    const inputShape = input2.shape;
+    return [inputShape[0], 1, inputShape[2]];
+  }
+};
+SpatialDropout1D.className = "SpatialDropout1D";
+serialization_exports.registerClass(SpatialDropout1D);
+var Dense = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.activation = null;
+    this.useBias = true;
+    this.kernel = null;
+    this.bias = null;
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    if (args.batchInputShape == null && args.inputShape == null && args.inputDim != null) {
+      let batchSize = null;
+      if (args.batchSize != null) {
+        batchSize = args.batchSize;
+      }
+      this.batchInputShape = [batchSize, args.inputDim];
+    }
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation);
+    if (args.useBias != null) {
+      this.useBias = args.useBias;
+    }
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.supportsMasking = true;
+    this.inputSpec = [{ minNDim: 2 }];
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const inputLastDim = inputShape[inputShape.length - 1];
+    if (this.kernel == null) {
+      this.kernel = this.addWeight("kernel", [inputLastDim, this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+      if (this.useBias) {
+        this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+      }
+    }
+    this.inputSpec = [{ minNDim: 2, axes: { [-1]: inputLastDim } }];
+    this.built = true;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    outputShape[outputShape.length - 1] = this.units;
+    return outputShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());
+      let output;
+      if (fusedActivationName != null) {
+        output = dot2(input2, this.kernel.read(), fusedActivationName, this.bias ? this.bias.read() : null);
+      } else {
+        output = dot2(input2, this.kernel.read());
+        if (this.bias != null) {
+          output = biasAdd(output, this.bias.read());
+        }
+        if (this.activation != null) {
+          output = this.activation.apply(output);
+        }
+      }
+      return output;
+    });
+  }
+  getConfig() {
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Dense.className = "Dense";
+serialization_exports.registerClass(Dense);
+var Flatten = class extends Layer {
+  constructor(args) {
+    args = args || {};
+    super(args);
+    this.inputSpec = [{ minNDim: 3 }];
+    this.dataFormat = args.dataFormat;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    for (const dim of inputShape.slice(1)) {
+      if (dim == null) {
+        throw new ValueError(`The shape of the input to "Flatten" is not fully defined (got ${inputShape.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);
+      }
+    }
+    return [inputShape[0], arrayProd(inputShape, 1)];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      let input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsFirst" && input2.rank > 1) {
+        const permutation = [0];
+        for (let i = 2; i < input2.rank; ++i) {
+          permutation.push(i);
+        }
+        permutation.push(1);
+        input2 = transpose(input2, permutation);
+      }
+      return batchFlatten(input2);
+    });
+  }
+  getConfig() {
+    const config = {};
+    if (this.dataFormat != null) {
+      config["dataFormat"] = this.dataFormat;
+    }
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Flatten.className = "Flatten";
+serialization_exports.registerClass(Flatten);
+var Activation2 = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.activation = getActivation(args.activation);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      return this.activation.apply(input2);
+    });
+  }
+  getConfig() {
+    const config = { activation: serializeActivation(this.activation) };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Activation2.className = "Activation";
+serialization_exports.registerClass(Activation2);
+var RepeatVector = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.n = args.n;
+    this.inputSpec = [{ ndim: 2 }];
+  }
+  computeOutputShape(inputShape) {
+    return [inputShape[0], this.n, inputShape[1]];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      return repeat(inputs, this.n);
+    });
+  }
+  getConfig() {
+    const config = {
+      n: this.n
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+RepeatVector.className = "RepeatVector";
+serialization_exports.registerClass(RepeatVector);
+var Reshape2 = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.targetShape = args.targetShape;
+    for (let i = 0; i < this.targetShape.length; ++i) {
+      if (this.isUnknown(this.targetShape[i])) {
+        this.targetShape[i] = null;
+      }
+    }
+  }
+  isUnknown(dim) {
+    return dim < 0 || dim == null;
+  }
+  fixUnknownDimension(inputShape, outputShape) {
+    const errorMsg = "Total size of new array must be unchanged.";
+    const finalShape = outputShape.slice();
+    let known = 1;
+    let unknown = null;
+    for (let i = 0; i < finalShape.length; ++i) {
+      const dim = finalShape[i];
+      if (this.isUnknown(dim)) {
+        if (unknown === null) {
+          unknown = i;
+        } else {
+          throw new ValueError("Can only specifiy one unknown dimension.");
+        }
+      } else {
+        known *= dim;
+      }
+    }
+    const originalSize = arrayProd(inputShape);
+    if (unknown !== null) {
+      if (known === 0 || originalSize % known !== 0) {
+        throw new ValueError(errorMsg);
+      }
+      finalShape[unknown] = originalSize / known;
+    } else if (originalSize !== known) {
+      throw new ValueError(errorMsg);
+    }
+    return finalShape;
+  }
+  computeOutputShape(inputShape) {
+    let anyUnknownDims = false;
+    for (let i = 0; i < inputShape.length; ++i) {
+      if (this.isUnknown(inputShape[i])) {
+        anyUnknownDims = true;
+        break;
+      }
+    }
+    if (anyUnknownDims) {
+      return inputShape.slice(0, 1).concat(this.targetShape);
+    } else {
+      return inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      const outputShape = inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));
+      return reshape(input2, outputShape);
+    });
+  }
+  getConfig() {
+    const config = {
+      targetShape: this.targetShape
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Reshape2.className = "Reshape";
+serialization_exports.registerClass(Reshape2);
+var Permute = class extends Layer {
+  constructor(args) {
+    super(args);
+    if (args.dims == null) {
+      throw new Error("Required configuration field `dims` is missing during Permute constructor call.");
+    }
+    if (!Array.isArray(args.dims)) {
+      throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${args.dims} instead.`);
+    }
+    const expectedSortedIndices = range2(1, args.dims.length + 1);
+    if (!util_exports.arraysEqual(args.dims.slice().sort(), expectedSortedIndices)) {
+      throw new Error("Invalid permutation `dims`: " + JSON.stringify(args.dims) + " `dims` must contain consecutive integers starting from 1.");
+    }
+    this.dims = args.dims;
+    this.dimsIncludingBatch = [0].concat(this.dims);
+    this.inputSpec = [new InputSpec({ ndim: this.dims.length + 1 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    this.dims.forEach((dim, i) => {
+      outputShape[i + 1] = inputShape[dim];
+    });
+    return outputShape;
+  }
+  call(inputs, kwargs) {
+    return transpose(getExactlyOneTensor(inputs), this.dimsIncludingBatch);
+  }
+  getConfig() {
+    const config = {
+      dims: this.dims
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Permute.className = "Permute";
+serialization_exports.registerClass(Permute);
+var Masking = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.supportsMasking = true;
+    if (args != null) {
+      this.maskValue = args.maskValue == null ? 0 : args.maskValue;
+    } else {
+      this.maskValue = 0;
+    }
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { maskValue: this.maskValue };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  computeMask(inputs, mask) {
+    const input2 = getExactlyOneTensor(inputs);
+    const axis = -1;
+    return any(notEqual(input2, this.maskValue), axis);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const axis = -1;
+      const keepDims = true;
+      const booleanMask = any(notEqual(input2, this.maskValue), axis, keepDims);
+      const output = mul(input2, cast(booleanMask, input2.dtype));
+      return output;
+    });
+  }
+};
+Masking.className = "Masking";
+serialization_exports.registerClass(Masking);
+var Embedding = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.embeddings = null;
+    this.DEFAULT_EMBEDDINGS_INITIALIZER = "randomUniform";
+    if (args.batchInputShape == null && args.inputShape == null) {
+      let batchSize = null;
+      if (args.batchSize != null) {
+        batchSize = args.batchSize;
+      }
+      if (args.inputLength == null) {
+        this.batchInputShape = [batchSize, null];
+      } else {
+        this.batchInputShape = [batchSize].concat(toList(args.inputLength));
+      }
+    }
+    this.inputDim = args.inputDim;
+    assertPositiveInteger(this.inputDim, "inputDim");
+    this.outputDim = args.outputDim;
+    assertPositiveInteger(this.outputDim, "outputDim");
+    this.embeddingsInitializer = getInitializer(args.embeddingsInitializer || this.DEFAULT_EMBEDDINGS_INITIALIZER);
+    this.embeddingsRegularizer = getRegularizer(args.embeddingsRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.embeddingsConstraint = getConstraint(args.embeddingsConstraint);
+    this.maskZero = args.maskZero;
+    this.supportsMasking = args.maskZero;
+    this.inputLength = args.inputLength;
+  }
+  build(inputShape) {
+    this.embeddings = this.addWeight("embeddings", [this.inputDim, this.outputDim], this.dtype, this.embeddingsInitializer, this.embeddingsRegularizer, true, this.embeddingsConstraint);
+    this.built = true;
+  }
+  warnOnIncompatibleInputShape(inputShape) {
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (!this.maskZero) {
+        return null;
+      } else {
+        inputs = getExactlyOneTensor(inputs);
+        return notEqual(inputs, zerosLike(inputs));
+      }
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (this.inputLength == null) {
+      return [...inputShape, this.outputDim];
+    }
+    const inLens = toList(this.inputLength);
+    if (inLens.length !== inputShape.length - 1) {
+      throw new ValueError(`"inputLength" is ${this.inputLength}, but received input shape has shape ${inputShape}`);
+    } else {
+      let i = 0;
+      for (let k = 0; k < inLens.length; ++k) {
+        const s1 = inLens[k];
+        const s2 = inputShape[k + 1];
+        if (s1 != null && s2 != null && s1 !== s2) {
+          throw new ValueError(`"inputLength" is ${this.inputLength}, but received input shape has shape ${inputShape}`);
+        } else if (s1 == null) {
+          inLens[i] = s2;
+        }
+        i++;
+      }
+    }
+    return [inputShape[0], ...inLens, this.outputDim];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.dtype !== "int32") {
+        input2 = cast2(input2, "int32");
+      }
+      const output = gather2(this.embeddings.read(), reshape(input2, [input2.size]));
+      return reshape(output, getExactlyOneShape(this.computeOutputShape(input2.shape)));
+    });
+  }
+  getConfig() {
+    const config = {
+      inputDim: this.inputDim,
+      outputDim: this.outputDim,
+      embeddingsInitializer: serializeInitializer(this.embeddingsInitializer),
+      embeddingsRegularizer: serializeRegularizer(this.embeddingsRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      embeddingsConstraint: serializeConstraint(this.embeddingsConstraint),
+      maskZero: this.maskZero,
+      inputLength: this.inputLength
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Embedding.className = "Embedding";
+serialization_exports.registerClass(Embedding);
+var Merge = class extends Layer {
+  constructor(args) {
+    super(args || {});
+    this.supportsMasking = true;
+  }
+  mergeFunction(inputs) {
+    throw new NotImplementedError();
+  }
+  computeElementwiseOpOutputShape(shape1, shape2) {
+    if (shape1 == null || shape2 == null) {
+      return null;
+    } else if (shape1.length < shape2.length) {
+      return this.computeElementwiseOpOutputShape(shape2, shape1);
+    } else if (shape2.length === 0) {
+      return shape1;
+    }
+    const outputShape = shape1.slice(0, shape1.length - shape2.length);
+    for (let k = 0; k < shape2.length; ++k) {
+      const i = shape1[shape1.length - shape2.length + k];
+      const j = shape2[k];
+      if (i == null || j == null || i < 0 || j < 0) {
+        outputShape.push(null);
+      } else if (i === 1) {
+        outputShape.push(j);
+      } else if (j === 1) {
+        outputShape.push(i);
+      } else {
+        if (i !== j) {
+          throw new ValueError("Operands could not be broadcast together with shapes " + JSON.stringify(shape1) + " " + JSON.stringify(shape2));
+        }
+        outputShape.push(i);
+      }
+    }
+    return outputShape;
+  }
+  build(inputShape) {
+    if (Array.isArray(inputShape) && !Array.isArray(inputShape[0])) {
+      inputShape = [getExactlyOneShape(inputShape)];
+    }
+    inputShape = inputShape;
+    if (inputShape.length < 2) {
+      throw new ValueError(`A merge layer should be called on an Array of at least 2 inputs. Got ${inputShape.length} input(s).`);
+    }
+    let batchSizes = [];
+    for (const shape of inputShape) {
+      if (shape != null && shape[0] !== null) {
+        batchSizes.push(shape[0]);
+      }
+    }
+    batchSizes = unique2(batchSizes);
+    if (batchSizes.length > 1) {
+      throw new ValueError(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(inputShape)}.`);
+    }
+    let outputShape = inputShape[0] == null ? null : inputShape[0].slice(1);
+    for (let i = 1; i < inputShape.length; ++i) {
+      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);
+      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);
+    }
+    const allRanks = inputShape.map((shape) => shape.length);
+    if (inputShape.indexOf(null) === -1 && unique2(allRanks).length === 1) {
+      this.reshapeRequired = false;
+    } else {
+      this.reshapeRequired = true;
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (this.reshapeRequired) {
+        const reshapedInputs = [];
+        const inputDims = inputs.map((input2) => input2.rank);
+        if (inputDims.indexOf(null) === -1) {
+          const maxNDim = max2(inputDims);
+          for (let x of inputs) {
+            const xNDim = x.rank;
+            for (let k = 0; k < maxNDim - xNDim; ++k) {
+              x = expandDims2(x, 1);
+            }
+            reshapedInputs.push(x);
+          }
+          return this.mergeFunction(reshapedInputs);
+        } else {
+          let transposed = false;
+          for (const x of inputs) {
+            const xNDim = x.rank;
+            if (xNDim == null) {
+              const xShape = x.shape;
+              const batchSize = xShape[0];
+              const newShape = xShape.slice(1).concat([batchSize]);
+              let xTransposed = reshape(x, [batchSize].concat(arrayProd(xShape.slice(1))));
+              xTransposed = transpose(xTransposed, [1, 0]);
+              xTransposed = reshape(xTransposed, newShape);
+              reshapedInputs.push(xTransposed);
+              transposed = true;
+            } else if (xNDim > 1) {
+              const dims = range2(1, xNDim).concat([0]);
+              reshapedInputs.push(transpose(x, dims));
+              transposed = true;
+            } else {
+              reshapedInputs.push(x);
+            }
+          }
+          let y = this.mergeFunction(reshapedInputs);
+          const yNDim = y.rank;
+          if (transposed) {
+            if (yNDim == null) {
+              const yShape = y.shape;
+              const yNDim2 = yShape.length;
+              const batchSize = yShape[yNDim2 - 1];
+              const newShape = [batchSize].concat(yShape.slice(0, yShape.length - 1));
+              y = reshape(transpose(reshape(y, [-1, batchSize]), [1, 0]), newShape);
+            } else if (yNDim > 1) {
+              const dims = [yNDim - 1].concat(range2(0, yNDim - 1));
+              y = transpose(y, dims);
+            }
+          }
+          return y;
+        }
+      } else {
+        return this.mergeFunction(inputs);
+      }
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = inputShape;
+    let outputShape;
+    if (inputShape[0] == null) {
+      outputShape = null;
+    } else {
+      outputShape = inputShape[0].slice(1);
+    }
+    for (let i = 1; i < inputShape.length; ++i) {
+      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);
+      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);
+    }
+    let batchSizes = [];
+    for (const shape of inputShape) {
+      if (shape != null && shape[0] !== null) {
+        batchSizes.push(shape[0]);
+      }
+    }
+    batchSizes = unique2(batchSizes);
+    if (batchSizes.length === 1) {
+      outputShape = batchSizes.concat(outputShape);
+    } else {
+      outputShape = [null].concat(outputShape);
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (mask == null) {
+        return null;
+      }
+      if (!Array.isArray(mask)) {
+        throw new ValueError("`mask` should be an Array");
+      }
+      if (!Array.isArray(inputs)) {
+        throw new ValueError("`inputs` should be an Array");
+      }
+      if (mask.length !== inputs.length) {
+        throw new ValueError(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${inputs.length} vs ${mask.length})`);
+      }
+      if (mask.every((m) => m == null)) {
+        return null;
+      }
+      mask = mask.map((m) => m == null ? m : expandDims(m, 0));
+      let output = mask[0];
+      for (let i = 1; i < mask.length - 1; ++i) {
+        output = logicalAnd(output, mask[i]);
+      }
+      return output;
+    });
+  }
+};
+var Add2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = add2(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Add2.className = "Add";
+serialization_exports.registerClass(Add2);
+var Multiply2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = mul(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Multiply2.className = "Multiply";
+serialization_exports.registerClass(Multiply2);
+var Average = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = add2(output, inputs[i]);
+      }
+      return mul(1 / inputs.length, output);
+    });
+  }
+};
+Average.className = "Average";
+serialization_exports.registerClass(Average);
+var Maximum2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0];
+      for (let i = 1; i < inputs.length; ++i) {
+        output = maximum(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Maximum2.className = "Maximum";
+serialization_exports.registerClass(Maximum2);
+var Minimum2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0];
+      for (let i = 1; i < inputs.length; ++i) {
+        output = minimum(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Minimum2.className = "Minimum";
+serialization_exports.registerClass(Minimum2);
+var Concatenate = class extends Merge {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_AXIS = -1;
+    if (args == null) {
+      args = {};
+    }
+    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;
+    this.supportsMasking = true;
+    this.reshapeRequired = false;
+  }
+  build(inputShape) {
+    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0])) || inputShape.length === 1) {
+      throw new ValueError("A `Concatenate` layer should be called on a list of at least 2 inputs");
+    }
+    inputShape = inputShape;
+    let allNoneShape = true;
+    for (const shape of inputShape) {
+      if (shape != null) {
+        allNoneShape = false;
+        break;
+      }
+    }
+    if (allNoneShape) {
+      return;
+    }
+    const shapeSet = [];
+    for (let i = 0; i < inputShape.length; ++i) {
+      const shapeWithoutConcatAxis = inputShape[i].slice();
+      shapeWithoutConcatAxis.splice(this.axis, 1);
+      let exists = false;
+      for (const shape of shapeSet) {
+        if (util_exports.arraysEqual(shape, shapeWithoutConcatAxis)) {
+          exists = true;
+          break;
+        }
+      }
+      if (!exists) {
+        shapeSet.push(shapeWithoutConcatAxis);
+      }
+    }
+    if (shapeSet.length > 1) {
+      throw new ValueError("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: " + JSON.stringify(inputShape));
+    }
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      return concatenate(inputs, this.axis);
+    });
+  }
+  computeOutputShape(inputShape) {
+    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0]))) {
+      throw new ValueError("A `Concatenate` layer should be called on a list of inputs.");
+    }
+    const inputShapes = inputShape;
+    const outputShape = inputShapes[0].slice();
+    const axis = this.axis < 0 ? outputShape.length + this.axis : this.axis;
+    for (const shape of inputShapes.slice(1)) {
+      if (outputShape[axis] == null || shape[axis] == null) {
+        outputShape[axis] = null;
+        break;
+      }
+      outputShape[axis] += shape[axis];
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    if (mask == null) {
+      return null;
+    }
+    if (!Array.isArray(mask)) {
+      throw new ValueError("`mask` should be an array for Concatenate");
+    }
+    if (!Array.isArray(inputs)) {
+      throw new ValueError("`inputs` should be an array for Concatenate");
+    }
+    if (mask.length !== inputs.length) {
+      throw new ValueError(`Mismatch in the length of mask (${mask.length}) and the legnth of inputs (${inputs.length})`);
+    }
+    return tidy(() => {
+      let allNullMasks = true;
+      mask.forEach((m) => {
+        if (m != null) {
+          allNullMasks = false;
+          return;
+        }
+      });
+      if (allNullMasks) {
+        return null;
+      }
+      const outputMasks = [];
+      for (let i = 0; i < inputs.length; ++i) {
+        if (mask[i] == null) {
+          outputMasks.push(cast(onesLike(inputs[i]), "bool"));
+        } else if (mask[i].rank < inputs[i].rank) {
+          outputMasks.push(expandDims(mask[i], -1));
+        } else {
+          outputMasks.push(mask[i]);
+        }
+      }
+      const concatenatedMasks = concat(outputMasks, this.axis);
+      return all(concatenatedMasks, -1, false);
+    });
+  }
+  getConfig() {
+    const config = {
+      "axis": this.axis
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Concatenate.className = "Concatenate";
+serialization_exports.registerClass(Concatenate);
+function interpretAxis(axis, dim) {
+  while (axis < 0) {
+    axis += dim;
+  }
+  return axis;
+}
+function batchDot(x, y, axes) {
+  if (x.shape.length > 3 || y.shape.length > 3) {
+    throw new NotImplementedError("batchDot is not implemented for tensors of 4D or higher rank yet");
+  }
+  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of x to be >= 2, but got ${x.shape.length}`);
+  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of y to be >= 2, but got ${y.shape.length}`);
+  if (typeof axes === "number") {
+    axes = [axes, axes];
+  }
+  if (x.dtype === "complex64" || y.dtype === "complex64") {
+    throw new NotImplementedError("batchDot is not implemented for complex64-type Tensors yet.");
+  }
+  const xNDim = x.shape.length;
+  const yNDim = y.shape.length;
+  if (axes == null) {
+    axes = [xNDim - 1, yNDim - 2];
+  }
+  const axesArray = axes;
+  return tidy(() => {
+    let diff;
+    if (xNDim > yNDim) {
+      diff = xNDim - yNDim;
+      const diffShape = [];
+      for (let i = 0; i < diff; ++i) {
+        diffShape.push(1);
+      }
+      y = reshape(y, y.shape.concat(diffShape));
+    } else if (yNDim > xNDim) {
+      diff = yNDim - xNDim;
+      const diffShape = [];
+      for (let i = 0; i < diff; ++i) {
+        diffShape.push(1);
+      }
+      x = reshape(x, x.shape.concat(diffShape));
+    } else {
+      diff = 0;
+    }
+    let out;
+    if (x.shape.length === 2 && y.shape.length === 2) {
+      if (axesArray[0] === axesArray[1]) {
+        out = sum2(mul(x, y), axesArray[0]);
+      } else {
+        out = sum2(mul(transpose(x, [1, 0]), y), axesArray[1]);
+      }
+    } else {
+      const adjX = axesArray[0] !== x.shape.length - 1;
+      const adjY = axesArray[1] === y.shape.length - 1;
+      out = matMul(x, y, adjX, adjY);
+    }
+    if (diff > 0) {
+      let idx;
+      if (xNDim > yNDim) {
+        idx = xNDim + yNDim - 3;
+      } else {
+        idx = xNDim - 1;
+      }
+      const squeezeAxes = [];
+      for (let i = idx; i < idx + diff; ++i) {
+        squeezeAxes.push(i);
+      }
+      out = squeeze(out, squeezeAxes);
+    }
+    if (out.shape.length === 1) {
+      out = expandDims(out, 1);
+    }
+    return out;
+  });
+}
+var Dot = class extends Merge {
+  constructor(args) {
+    super(args);
+    this.axes = args.axes;
+    this.normalize = args.normalize == null ? false : args.normalize;
+    this.supportsMasking = true;
+    this.reshapeRequired = false;
+  }
+  build(inputShape) {
+    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+    const shape1 = inputShape[0];
+    const shape2 = inputShape[1];
+    if (shape1.length > 3 || shape2.length > 3) {
+      throw new NotImplementedError("Dot layer does not support tensors of 4D or higher rank yet.");
+    }
+    const axes = this.interpretAxes(shape1, shape2);
+    if (shape1[axes[0]] !== shape2[axes[1]]) {
+      throw new ValueError(`Dimension incompatibility: ${shape1[axes[0]]} !== ${shape2[axes[1]]}`);
+    }
+  }
+  mergeFunction(inputs) {
+    if (inputs.length !== 2) {
+      throw new ValueError(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${inputs.length} input(s).`);
+    }
+    let x1 = inputs[0];
+    let x2 = inputs[1];
+    let axes;
+    if (!Array.isArray(this.axes)) {
+      axes = [
+        interpretAxis(this.axes, x1.shape.length),
+        interpretAxis(this.axes, x2.shape.length)
+      ];
+    } else {
+      axes = this.axes.map((axis, i) => interpretAxis(axis, inputs[i].shape.length));
+    }
+    if (this.normalize) {
+      x1 = l2Normalize(x1, axes[0]);
+      x2 = l2Normalize(x2, axes[1]);
+    }
+    return batchDot(x1, x2, axes);
+  }
+  interpretAxes(shape1, shape2) {
+    let axes;
+    if (!Array.isArray(this.axes)) {
+      axes = [
+        interpretAxis(this.axes, shape1.length),
+        interpretAxis(this.axes, shape2.length)
+      ];
+    } else {
+      axes = this.axes;
+    }
+    return axes;
+  }
+  computeOutputShape(inputShape) {
+    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+    const shape1 = inputShape[0].slice();
+    const shape2 = inputShape[1].slice();
+    if (shape1.length > 3 || shape2.length > 3) {
+      throw new NotImplementedError("Dot layer does not support tensors of 4D or higher rank yet.");
+    }
+    const axes = this.interpretAxes(shape1, shape2);
+    shape1.splice(axes[0], 1);
+    shape2.splice(axes[1], 1);
+    shape2.splice(0, 1);
+    const outputShape = shape1.concat(shape2);
+    if (outputShape.length === 1) {
+      outputShape.push(1);
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    return null;
+  }
+  getConfig() {
+    const config = {
+      "axes": this.axes,
+      "normalize": this.normalize
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Dot.className = "Dot";
+serialization_exports.registerClass(Dot);
+var GaussianNoise = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.stddev = args.stddev;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { stddev: this.stddev };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const noised = () => add2(randomNormal2(input2.shape, 0, this.stddev), input2);
+      const output = inTrainPhase(noised, () => input2, kwargs["training"] || false);
+      return output;
+    });
+  }
+};
+GaussianNoise.className = "GaussianNoise";
+serialization_exports.registerClass(GaussianNoise);
+var GaussianDropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.rate = args.rate;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { rate: this.rate };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.rate > 0 && this.rate < 1) {
+        const noised = () => {
+          const stddev = Math.sqrt(this.rate / (1 - this.rate));
+          return mul(input2, randomNormal2(input2.shape, 1, stddev));
+        };
+        return inTrainPhase(noised, () => input2, kwargs["training"] || false);
+      }
+      return input2;
+    });
+  }
+};
+GaussianDropout.className = "GaussianDropout";
+serialization_exports.registerClass(GaussianDropout);
+var AlphaDropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.rate = args.rate;
+    this.noiseShape = args.noiseShape;
+  }
+  _getNoiseShape(inputs) {
+    return this.noiseShape || getExactlyOneTensor(inputs).shape;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { rate: this.rate };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.rate < 1 && this.rate > 0) {
+        const noiseShape = this._getNoiseShape(inputs);
+        const droppedInputs = () => {
+          const input2 = getExactlyOneTensor(inputs);
+          const alpha = 1.6732632423543772;
+          const scale22 = 1.0507009873554805;
+          const alphaP = -alpha * scale22;
+          let keptIdx = greaterEqual(randomUniform(noiseShape), this.rate);
+          keptIdx = cast2(keptIdx, "float32");
+          const a = ((1 - this.rate) * (1 + this.rate * alphaP ** 2)) ** -0.5;
+          const b = -a * alphaP * this.rate;
+          const x = add2(mul(input2, keptIdx), mul(add2(keptIdx, -1), alphaP));
+          return add2(mul(x, a), b);
+        };
+        return inTrainPhase(droppedInputs, () => getExactlyOneTensor(inputs), kwargs["training"] || false);
+      }
+      return inputs;
+    });
+  }
+};
+AlphaDropout.className = "AlphaDropout";
+serialization_exports.registerClass(AlphaDropout);
+function batchNormalization(x, mean4, variance, beta, gamma, epsilon32 = 1e-3) {
+  let out;
+  if (x.rank === 2) {
+    out = batchNorm2d(x, mean4, variance, beta, gamma, epsilon32);
+  } else if (x.rank === 3) {
+    out = batchNorm3d(x, mean4, variance, beta, gamma, epsilon32);
+  } else if (x.rank === 4) {
+    out = batchNorm4d(x, mean4, variance, beta, gamma, epsilon32);
+  } else {
+    throw new NotImplementedError(`batchNormalization is not implemented for array of rank ${x.rank} yet`);
+  }
+  return out;
+}
+function regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon32 = 1e-3) {
+  return tidy(() => {
+    const meanAndVariance = moments(x, reductionAxes);
+    const mean4 = meanAndVariance.mean;
+    const variance = meanAndVariance.variance;
+    const normed = batchNormalization(x, mean4, variance, beta, gamma, epsilon32);
+    return [normed, mean4, variance];
+  });
+}
+function broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon32 = 1e-3) {
+  return tidy(() => {
+    const meanAndVariance = moments(x, reductionAxes);
+    const mean4 = meanAndVariance.mean;
+    const variance = meanAndVariance.variance;
+    const targetShape = [];
+    for (const axis of range2(0, x.rank)) {
+      if (reductionAxes.indexOf(axis) !== -1) {
+        targetShape.push(1);
+      } else {
+        targetShape.push(x.shape[axis]);
+      }
+    }
+    const broadcastMean = reshape(mean4, targetShape);
+    const broadcastVariance = reshape(variance, targetShape);
+    const broadcastGamma = gamma == null ? null : reshape(gamma, targetShape);
+    const broadcastBeta = beta == null ? null : reshape(beta, targetShape);
+    const normed = batchNormalization(x, broadcastMean, broadcastVariance, broadcastBeta, broadcastGamma, epsilon32);
+    return [normed, mean4, variance];
+  });
+}
+function normalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon32 = 1e-3) {
+  if (util_exports.arraysEqual(reductionAxes.slice().sort(), range2(0, x.rank - 1))) {
+    return regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon32);
+  } else {
+    return broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon32);
+  }
+}
+var BatchNormalization = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.supportsMasking = true;
+    this.axis = args.axis == null ? -1 : args.axis;
+    this.momentum = args.momentum == null ? 0.99 : args.momentum;
+    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;
+    this.center = args.center == null ? true : args.center;
+    this.scale = args.scale == null ? true : args.scale;
+    this.betaInitializer = getInitializer(args.betaInitializer || "zeros");
+    this.gammaInitializer = getInitializer(args.gammaInitializer || "ones");
+    this.movingMeanInitializer = getInitializer(args.movingMeanInitializer || "zeros");
+    this.movingVarianceInitializer = getInitializer(args.movingVarianceInitializer || "ones");
+    this.betaConstraint = getConstraint(args.betaConstraint);
+    this.gammaConstraint = getConstraint(args.gammaConstraint);
+    this.betaRegularizer = getRegularizer(args.betaRegularizer);
+    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const axis = this.axis >= 0 ? this.axis : this.axis + inputShape.length;
+    const dim = inputShape[axis];
+    if (dim == null) {
+      throw new ValueError(`Axis ${axis} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(inputShape)}.`);
+    }
+    this.inputSpec = [new InputSpec({ ndim: inputShape.length, axes: { [axis]: dim } })];
+    const shape = [dim];
+    if (this.scale) {
+      this.gamma = this.addWeight("gamma", shape, null, this.gammaInitializer, this.gammaRegularizer, true, this.gammaConstraint);
+    }
+    if (this.center) {
+      this.beta = this.addWeight("beta", shape, null, this.betaInitializer, this.betaRegularizer, true, this.betaConstraint);
+    }
+    this.movingMean = this.addWeight("moving_mean", shape, null, this.movingMeanInitializer, null, false);
+    this.movingVariance = this.addWeight("moving_variance", shape, null, this.movingVarianceInitializer, null, false);
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      const input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      const ndim = inputShape.length;
+      const reductionAxes = range2(0, ndim);
+      const axis = this.axis >= 0 ? this.axis : this.axis + ndim;
+      reductionAxes.splice(axis, 1);
+      const broadcastShape = pyListRepeat(1, ndim);
+      broadcastShape[axis] = inputShape[axis];
+      const sortedReductionAxes = reductionAxes.slice();
+      sortedReductionAxes.sort();
+      const needsBroadcasting = !util_exports.arraysEqual(sortedReductionAxes, range2(0, ndim).slice(0, ndim - 1));
+      const normalizeInference = () => {
+        if (needsBroadcasting) {
+          const broadcastMovingMean = reshape(this.movingMean.read(), broadcastShape);
+          const broadcastMovingVariance = reshape(this.movingVariance.read(), broadcastShape);
+          const broadcastBeta = this.center ? reshape(this.beta.read(), broadcastShape) : null;
+          const broadcastGamma = this.scale ? reshape(this.gamma.read(), broadcastShape) : null;
+          return batchNormalization(input2, broadcastMovingMean, broadcastMovingVariance, broadcastBeta, broadcastGamma, this.epsilon);
+        } else {
+          return batchNormalization(input2, this.movingMean.read(), this.movingVariance.read(), this.beta == null ? null : this.beta.read(), this.gamma == null ? null : this.gamma.read(), this.epsilon);
+        }
+      };
+      if (!training) {
+        return normalizeInference();
+      }
+      const [normedTraining, mean4, variance] = normalizeBatchInTraining(input2, this.gamma.read(), this.beta.read(), reductionAxes, this.epsilon);
+      const doMovingAverage = (variable2, value, momentum) => {
+        tidy(() => {
+          const decay = 1 - momentum;
+          const origValue = variable2.read();
+          const updateDelta = mul(sub(origValue, value), decay);
+          variable2.write(sub(origValue, updateDelta));
+        });
+      };
+      const updateMovingMeanAndVariance = () => {
+        doMovingAverage(this.movingMean, mean4, this.momentum);
+        doMovingAverage(this.movingVariance, variance, this.momentum);
+      };
+      updateMovingMeanAndVariance();
+      return normedTraining;
+    });
+  }
+  getConfig() {
+    const config = {
+      axis: this.axis,
+      momentum: this.momentum,
+      epsilon: this.epsilon,
+      center: this.center,
+      scale: this.scale,
+      betaInitializer: serializeInitializer(this.betaInitializer),
+      gammaInitializer: serializeInitializer(this.gammaInitializer),
+      movingMeanInitializer: serializeInitializer(this.movingMeanInitializer),
+      movingVarianceInitializer: serializeInitializer(this.movingVarianceInitializer),
+      betaRegularizer: serializeRegularizer(this.betaRegularizer),
+      gammaRegularizer: serializeRegularizer(this.gammaRegularizer),
+      betaConstraint: serializeConstraint(this.betaConstraint),
+      gammaConstraint: serializeConstraint(this.gammaConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+BatchNormalization.className = "BatchNormalization";
+serialization_exports.registerClass(BatchNormalization);
+var LayerNormalization = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.axis = args.axis == null ? -1 : args.axis;
+    if (typeof this.axis === "number") {
+      if (!Number.isInteger(this.axis)) {
+        throw new Error(`Expected axis to be an integer, but received ${this.axis}`);
+      }
+    } else if (Array.isArray(this.axis)) {
+      for (const axis of this.axis) {
+        if (!Number.isInteger(axis)) {
+          throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`);
+        }
+      }
+    } else {
+      throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);
+    }
+    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;
+    this.center = args.center == null ? true : args.center;
+    this.scale = args.scale == null ? true : args.scale;
+    this.betaInitializer = getInitializer(args.betaInitializer || "zeros");
+    this.gammaInitializer = getInitializer(args.gammaInitializer || "ones");
+    this.betaRegularizer = getRegularizer(args.betaRegularizer);
+    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);
+    this.supportsMasking = true;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const nDims = inputShape.length;
+    if (typeof this.axis === "number") {
+      this.axis = [this.axis];
+    }
+    for (let i = 0; i < this.axis.length; ++i) {
+      if (this.axis[i] < 0) {
+        this.axis[i] += nDims;
+      }
+    }
+    for (const axis of this.axis) {
+      if (axis < 0 || axis >= nDims) {
+        throw new Error(`Invalid axis: ${axis}`);
+      }
+    }
+    if (this.axis.length !== unique2(this.axis).length) {
+      throw new Error(`Found duplicate axes in: ${this.axis}`);
+    }
+    const paramShape = this.axis.map((axis) => inputShape[axis]);
+    const trainable = true;
+    if (this.scale) {
+      this.gamma = this.addWeight("gamma", paramShape, "float32", this.gammaInitializer, this.gammaRegularizer, trainable);
+    } else {
+      this.gamma = null;
+    }
+    if (this.center) {
+      this.beta = this.addWeight("beta", paramShape, "float32", this.betaInitializer, this.betaRegularizer, trainable);
+    } else {
+      this.beta = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    const input2 = getExactlyOneTensor(inputs);
+    const inputShape = input2.shape;
+    const nDims = inputShape.length;
+    return tidy(() => {
+      const keepDims = true;
+      let { mean: mean4, variance } = moments(input2, this.axis, keepDims);
+      const broadcastShape = pyListRepeat(1, nDims);
+      for (const dim of this.axis) {
+        broadcastShape[dim] = inputShape[dim];
+      }
+      const broadcast = (v) => {
+        if (v != null && v.shape.length !== nDims) {
+          return reshape(v, broadcastShape);
+        } else {
+          return v;
+        }
+      };
+      let scale22 = this.scale ? broadcast(this.gamma.read()) : null;
+      let offset = this.center ? broadcast(this.beta.read()) : null;
+      const momentsTiling = [];
+      const scaleOffsetTiling = [];
+      for (let i = 0; i < nDims; ++i) {
+        if (this.axis.indexOf(i) !== -1) {
+          momentsTiling.push(inputShape[i]);
+          scaleOffsetTiling.push(1);
+        } else {
+          momentsTiling.push(1);
+          scaleOffsetTiling.push(inputShape[i]);
+        }
+      }
+      mean4 = tile(mean4, momentsTiling);
+      variance = tile(variance, momentsTiling);
+      if (scale22 != null) {
+        scale22 = tile(scale22, scaleOffsetTiling);
+      }
+      if (offset != null) {
+        offset = tile(offset, scaleOffsetTiling);
+      }
+      return batchNormalization(input2, mean4, variance, offset, scale22, this.epsilon);
+    });
+  }
+  getConfig() {
+    const config = {
+      axis: this.axis,
+      epsilon: this.epsilon,
+      center: this.center,
+      scale: this.scale,
+      betaInitializer: serializeInitializer(this.betaInitializer),
+      gammaInitializer: serializeInitializer(this.gammaInitializer),
+      betaRegularizer: serializeRegularizer(this.betaRegularizer),
+      gammaRegularizer: serializeRegularizer(this.gammaRegularizer)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+LayerNormalization.className = "LayerNormalization";
+serialization_exports.registerClass(LayerNormalization);
+function spatial2dPadding(x, padding, dataFormat) {
+  return tidy(() => {
+    if (x.rank !== 4) {
+      throw new ValueError(`temporalPadding expects input tensor to be 4-D, but received a ${x.rank}-D tensor.`);
+    }
+    if (padding == null) {
+      padding = [[1, 1], [1, 1]];
+    }
+    if (padding.length !== 2 || padding[0].length !== 2 || padding[1].length !== 2) {
+      throw new ValueError("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (dataFormat !== "channelsLast" && dataFormat !== "channelsFirst") {
+      throw new ValueError(`Unknown data format: ${dataFormat}. Supported data formats are 'channelsLast' and 'channelsFirst.`);
+    }
+    let pattern;
+    if (dataFormat === "channelsFirst") {
+      pattern = [[0, 0], [0, 0], padding[0], padding[1]];
+    } else {
+      pattern = [[0, 0], padding[0], padding[1], [0, 0]];
+    }
+    return pad(x, pattern);
+  });
+}
+var ZeroPadding2D = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.dataFormat = args.dataFormat == null ? imageDataFormat() : args.dataFormat;
+    if (args.padding == null) {
+      this.padding = [[1, 1], [1, 1]];
+    } else if (typeof args.padding === "number") {
+      this.padding = [[args.padding, args.padding], [args.padding, args.padding]];
+    } else {
+      args.padding = args.padding;
+      if (args.padding.length !== 2) {
+        throw new ValueError(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${args.padding.length} array.`);
+      }
+      let heightPadding;
+      let widthPadding;
+      if (typeof args.padding[0] === "number") {
+        heightPadding = [args.padding[0], args.padding[0]];
+        widthPadding = [args.padding[1], args.padding[1]];
+      } else {
+        args.padding = args.padding;
+        if (args.padding[0].length !== 2) {
+          throw new ValueError(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${args.padding[0].length} array.`);
+        }
+        heightPadding = args.padding[0];
+        if (args.padding[1].length !== 2) {
+          throw new ValueError(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${args.padding[1].length} array.`);
+        }
+        widthPadding = args.padding[1];
+      }
+      this.padding = [heightPadding, widthPadding];
+    }
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let rows;
+    let cols;
+    if (this.dataFormat === "channelsFirst") {
+      if (inputShape[2] != null && inputShape[2] >= 0) {
+        rows = inputShape[2] + this.padding[0][0] + this.padding[0][1];
+      } else {
+        rows = null;
+      }
+      if (inputShape[3] != null && inputShape[3] >= 0) {
+        cols = inputShape[3] + this.padding[1][0] + this.padding[1][1];
+      } else {
+        cols = null;
+      }
+      return [inputShape[0], inputShape[1], rows, cols];
+    } else {
+      if (inputShape[1] != null && inputShape[1] >= 0) {
+        rows = inputShape[1] + this.padding[0][0] + this.padding[0][1];
+      } else {
+        rows = null;
+      }
+      if (inputShape[2] != null && inputShape[2] >= 0) {
+        cols = inputShape[2] + this.padding[1][0] + this.padding[1][1];
+      } else {
+        cols = null;
+      }
+      return [inputShape[0], rows, cols, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => spatial2dPadding(getExactlyOneTensor(inputs), this.padding, this.dataFormat));
+  }
+  getConfig() {
+    const config = {
+      padding: this.padding,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ZeroPadding2D.className = "ZeroPadding2D";
+serialization_exports.registerClass(ZeroPadding2D);
+function pool2d(x, poolSize, strides, padding, dataFormat, poolMode) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    checkPoolMode(poolMode);
+    checkPaddingMode(padding);
+    if (strides == null) {
+      strides = [1, 1];
+    }
+    if (padding == null) {
+      padding = "valid";
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (poolMode == null) {
+      poolMode = "max";
+    }
+    x = preprocessConv2DInput(x, dataFormat);
+    let y;
+    const paddingString = padding === "same" ? "same" : "valid";
+    if (poolMode === "max") {
+      y = maxPool(x, poolSize, strides, paddingString);
+    } else {
+      y = avgPool(
+        x,
+        poolSize,
+        strides,
+        paddingString
+      );
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+function pool3d(x, poolSize, strides, padding, dataFormat, poolMode) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    checkPoolMode(poolMode);
+    checkPaddingMode(padding);
+    if (strides == null) {
+      strides = [1, 1, 1];
+    }
+    if (padding == null) {
+      padding = "valid";
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (poolMode == null) {
+      poolMode = "max";
+    }
+    x = preprocessConv3DInput(x, dataFormat);
+    let y;
+    const paddingString = padding === "same" ? "same" : "valid";
+    if (poolMode === "max") {
+      y = maxPool3d(x, poolSize, strides, paddingString);
+    } else {
+      y = avgPool3d(x, poolSize, strides, paddingString);
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 4, 1, 2, 3]);
+    }
+    return y;
+  });
+}
+var Pooling1D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = 2;
+    }
+    super(args);
+    if (typeof args.poolSize === "number") {
+      this.poolSize = [args.poolSize];
+    } else if (Array.isArray(args.poolSize) && args.poolSize.length === 1 && typeof args.poolSize[0] === "number") {
+      this.poolSize = args.poolSize;
+    } else {
+      throw new ValueError(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.poolSize)}`);
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else {
+      if (typeof args.strides === "number") {
+        this.strides = [args.strides];
+      } else if (Array.isArray(args.strides) && args.strides.length === 1 && typeof args.strides[0] === "number") {
+        this.strides = args.strides;
+      } else {
+        throw new ValueError(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.strides)}`);
+      }
+    }
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const length = convOutputLength(inputShape[1], this.poolSize[0], this.padding, this.strides[0]);
+    return [inputShape[0], length, inputShape[2]];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      inputs = expandDims2(getExactlyOneTensor(inputs), 2);
+      const output = this.poolingFunction(getExactlyOneTensor(inputs), [this.poolSize[0], 1], [this.strides[0], 1], this.padding, "channelsLast");
+      return squeeze(output, [2]);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling1D = class extends Pooling1D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling1D.className = "MaxPooling1D";
+serialization_exports.registerClass(MaxPooling1D);
+var AveragePooling1D = class extends Pooling1D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling1D.className = "AveragePooling1D";
+serialization_exports.registerClass(AveragePooling1D);
+var Pooling2D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = [2, 2];
+    }
+    super(args);
+    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize];
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else if (Array.isArray(args.strides)) {
+      if (args.strides.length !== 2) {
+        throw new ValueError(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${args.strides.length}.`);
+      }
+      this.strides = args.strides;
+    } else {
+      this.strides = [args.strides, args.strides];
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let rows = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    let cols = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    rows = convOutputLength(rows, this.poolSize[0], this.padding, this.strides[0]);
+    cols = convOutputLength(cols, this.poolSize[1], this.padding, this.strides[1]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], inputShape[1], rows, cols];
+    } else {
+      return [inputShape[0], rows, cols, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling2D = class extends Pooling2D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling2D.className = "MaxPooling2D";
+serialization_exports.registerClass(MaxPooling2D);
+var AveragePooling2D = class extends Pooling2D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling2D.className = "AveragePooling2D";
+serialization_exports.registerClass(AveragePooling2D);
+var Pooling3D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = [2, 2, 2];
+    }
+    super(args);
+    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize, args.poolSize];
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else if (Array.isArray(args.strides)) {
+      if (args.strides.length !== 3) {
+        throw new ValueError(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${args.strides.length}.`);
+      }
+      this.strides = args.strides;
+    } else {
+      this.strides = [args.strides, args.strides, args.strides];
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let depths = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    let rows = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    let cols = this.dataFormat === "channelsFirst" ? inputShape[4] : inputShape[3];
+    depths = convOutputLength(depths, this.poolSize[0], this.padding, this.strides[0]);
+    rows = convOutputLength(rows, this.poolSize[1], this.padding, this.strides[1]);
+    cols = convOutputLength(cols, this.poolSize[2], this.padding, this.strides[2]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], inputShape[1], depths, rows, cols];
+    } else {
+      return [inputShape[0], depths, rows, cols, inputShape[4]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling3D = class extends Pooling3D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool3d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling3D.className = "MaxPooling3D";
+serialization_exports.registerClass(MaxPooling3D);
+var AveragePooling3D = class extends Pooling3D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool3d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling3D.className = "AveragePooling3D";
+serialization_exports.registerClass(AveragePooling3D);
+var GlobalPooling1D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+  }
+  computeOutputShape(inputShape) {
+    return [inputShape[0], inputShape[2]];
+  }
+  call(inputs, kwargs) {
+    throw new NotImplementedError();
+  }
+};
+var GlobalAveragePooling1D = class extends GlobalPooling1D {
+  constructor(args) {
+    super(args || {});
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      return mean(input2, 1);
+    });
+  }
+};
+GlobalAveragePooling1D.className = "GlobalAveragePooling1D";
+serialization_exports.registerClass(GlobalAveragePooling1D);
+var GlobalMaxPooling1D = class extends GlobalPooling1D {
+  constructor(args) {
+    super(args || {});
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      return max(input2, 1);
+    });
+  }
+};
+GlobalMaxPooling1D.className = "GlobalMaxPooling1D";
+serialization_exports.registerClass(GlobalMaxPooling1D);
+var GlobalPooling2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = inputShape;
+    if (this.dataFormat === "channelsLast") {
+      return [inputShape[0], inputShape[3]];
+    } else {
+      return [inputShape[0], inputShape[1]];
+    }
+  }
+  call(inputs, kwargs) {
+    throw new NotImplementedError();
+  }
+  getConfig() {
+    const config = { dataFormat: this.dataFormat };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var GlobalAveragePooling2D = class extends GlobalPooling2D {
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        return mean(input2, [1, 2]);
+      } else {
+        return mean(input2, [2, 3]);
+      }
+    });
+  }
+};
+GlobalAveragePooling2D.className = "GlobalAveragePooling2D";
+serialization_exports.registerClass(GlobalAveragePooling2D);
+var GlobalMaxPooling2D = class extends GlobalPooling2D {
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        return max(input2, [1, 2]);
+      } else {
+        return max(input2, [2, 3]);
+      }
+    });
+  }
+};
+GlobalMaxPooling2D.className = "GlobalMaxPooling2D";
+serialization_exports.registerClass(GlobalMaxPooling2D);
+var Wrapper = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.layer = args.layer;
+  }
+  build(inputShape) {
+    this.built = true;
+  }
+  get trainable() {
+    if (this.layer != null) {
+      return this.layer.trainable;
+    } else {
+      return false;
+    }
+  }
+  set trainable(value) {
+    if (this.layer != null) {
+      this.layer.trainable = value;
+    }
+  }
+  get trainableWeights() {
+    return this.layer.trainableWeights;
+  }
+  get nonTrainableWeights() {
+    return this.layer.nonTrainableWeights;
+  }
+  get updates() {
+    return this.layer._updates;
+  }
+  get losses() {
+    return this.layer.losses;
+  }
+  getWeights() {
+    return this.layer.getWeights();
+  }
+  setWeights(weights) {
+    this.layer.setWeights(weights);
+  }
+  getConfig() {
+    const config = {
+      "layer": {
+        "className": this.layer.getClassName(),
+        "config": this.layer.getConfig()
+      }
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.layer != null) {
+      this.layer.setFastWeightInitDuringBuild(value);
+    }
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const layerConfig = config["layer"];
+    const layer = deserialize(layerConfig, customObjects);
+    delete config["layer"];
+    const newConfig = { layer };
+    Object.assign(newConfig, config);
+    return new cls(newConfig);
+  }
+};
+var TimeDistributed = class extends Wrapper {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < 3) {
+      throw new ValueError(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(inputShape)}`);
+    }
+    this.inputSpec = [{ shape: inputShape }];
+    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    if (!this.layer.built) {
+      this.layer.build(childInputShape);
+      this.layer.built = true;
+    }
+    super.build(inputShape);
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    const childOutputShape = this.layer.computeOutputShape(childInputShape);
+    const timesteps = inputShape[1];
+    return [childOutputShape[0], timesteps].concat(childOutputShape.slice(1));
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      const step5 = (inputs2, states) => {
+        const output = getExactlyOneTensor(this.layer.call(inputs2, kwargs));
+        return [output, []];
+      };
+      const rnnOutputs = rnn(step5, inputs, [], false, null, null, false, true);
+      const y = rnnOutputs[1];
+      return y;
+    });
+  }
+};
+TimeDistributed.className = "TimeDistributed";
+serialization_exports.registerClass(TimeDistributed);
+function checkBidirectionalMergeMode(value) {
+  checkStringTypeUnionValue(VALID_BIDIRECTIONAL_MERGE_MODES, "BidirectionalMergeMode", value);
+}
+var DEFAULT_BIDIRECTIONAL_MERGE_MODE = "concat";
+var Bidirectional = class extends Wrapper {
+  constructor(args) {
+    super(args);
+    const layerConfig = args.layer.getConfig();
+    const forwDict = {};
+    forwDict["className"] = args.layer.getClassName();
+    forwDict["config"] = layerConfig;
+    this.forwardLayer = deserialize(forwDict);
+    layerConfig["goBackwards"] = layerConfig["goBackwards"] === true ? false : true;
+    const backDict = {};
+    backDict["className"] = args.layer.getClassName();
+    backDict["config"] = layerConfig;
+    this.backwardLayer = deserialize(backDict);
+    this.forwardLayer.name = "forward_" + this.forwardLayer.name;
+    this.backwardLayer.name = "backward_" + this.backwardLayer.name;
+    this.mergeMode = args.mergeMode === void 0 ? DEFAULT_BIDIRECTIONAL_MERGE_MODE : args.mergeMode;
+    checkBidirectionalMergeMode(this.mergeMode);
+    if (args.weights) {
+      throw new NotImplementedError("weights support is not implemented for Bidirectional layer yet.");
+    }
+    this._stateful = args.layer.stateful;
+    this.returnSequences = args.layer.returnSequences;
+    this.returnState = args.layer.returnState;
+    this.supportsMasking = true;
+    this._trainable = true;
+    this.inputSpec = args.layer.inputSpec;
+    this.numConstants = null;
+  }
+  get trainable() {
+    return this._trainable;
+  }
+  set trainable(value) {
+    this._trainable = value;
+    if (this.forwardLayer != null) {
+      this.forwardLayer.trainable = value;
+    }
+    if (this.backwardLayer != null) {
+      this.backwardLayer.trainable = value;
+    }
+  }
+  getWeights() {
+    return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights());
+  }
+  setWeights(weights) {
+    const numWeights = weights.length;
+    const numeightsOver2 = Math.floor(numWeights / 2);
+    this.forwardLayer.setWeights(weights.slice(0, numeightsOver2));
+    this.backwardLayer.setWeights(weights.slice(numeightsOver2));
+  }
+  computeOutputShape(inputShape) {
+    let layerShapes = this.forwardLayer.computeOutputShape(inputShape);
+    if (!(Array.isArray(layerShapes) && Array.isArray(layerShapes[0]))) {
+      layerShapes = [layerShapes];
+    }
+    layerShapes = layerShapes;
+    let outputShape;
+    let outputShapes;
+    let stateShape;
+    if (this.returnState) {
+      stateShape = layerShapes.slice(1);
+      outputShape = layerShapes[0];
+    } else {
+      outputShape = layerShapes[0];
+    }
+    outputShape = outputShape;
+    if (this.mergeMode === "concat") {
+      outputShape[outputShape.length - 1] *= 2;
+      outputShapes = [outputShape];
+    } else if (this.mergeMode == null) {
+      outputShapes = [outputShape, outputShape.slice()];
+    } else {
+      outputShapes = [outputShape];
+    }
+    if (this.returnState) {
+      if (this.mergeMode == null) {
+        return outputShapes.concat(stateShape).concat(stateShape.slice());
+      }
+      return [outputShape].concat(stateShape).concat(stateShape.slice());
+    }
+    return singletonOrArray(outputShapes);
+  }
+  apply(inputs, kwargs) {
+    let initialState = kwargs == null ? null : kwargs["initialState"];
+    let constants = kwargs == null ? null : kwargs["constants"];
+    if (kwargs == null) {
+      kwargs = {};
+    }
+    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);
+    inputs = standardized.inputs;
+    initialState = standardized.initialState;
+    constants = standardized.constants;
+    if (Array.isArray(inputs)) {
+      initialState = inputs.slice(1);
+      inputs = inputs[0];
+    }
+    if ((initialState == null || initialState.length === 0) && constants == null) {
+      return super.apply(inputs, kwargs);
+    }
+    const additionalInputs = [];
+    const additionalSpecs = [];
+    if (initialState != null) {
+      const numStates = initialState.length;
+      if (numStates % 2 > 0) {
+        throw new ValueError("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");
+      }
+      kwargs["initialState"] = initialState;
+      additionalInputs.push(...initialState);
+      const stateSpecs = initialState.map((state) => new InputSpec({ shape: state.shape }));
+      this.forwardLayer.stateSpec = stateSpecs.slice(0, numStates / 2);
+      this.backwardLayer.stateSpec = stateSpecs.slice(numStates / 2);
+      additionalSpecs.push(...stateSpecs);
+    }
+    if (constants != null) {
+      throw new NotImplementedError("Support for constants in Bidirectional layers is not implemented yet.");
+    }
+    const isSymbolicTensor = additionalInputs[0] instanceof SymbolicTensor;
+    for (const tensor2 of additionalInputs) {
+      if (tensor2 instanceof SymbolicTensor !== isSymbolicTensor) {
+        throw new ValueError("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");
+      }
+    }
+    if (isSymbolicTensor) {
+      const fullInput = [inputs].concat(additionalInputs);
+      const fullInputSpec = this.inputSpec.concat(additionalSpecs);
+      const originalInputSpec = this.inputSpec;
+      this.inputSpec = fullInputSpec;
+      const output = super.apply(fullInput, kwargs);
+      this.inputSpec = originalInputSpec;
+      return output;
+    } else {
+      return super.apply(inputs, kwargs);
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const initialState = kwargs["initialState"];
+      let y;
+      let yRev;
+      if (initialState == null) {
+        y = this.forwardLayer.call(inputs, kwargs);
+        yRev = this.backwardLayer.call(inputs, kwargs);
+      } else {
+        const forwardState = initialState.slice(0, initialState.length / 2);
+        const backwardState = initialState.slice(initialState.length / 2);
+        y = this.forwardLayer.call(inputs, Object.assign(kwargs, { initialState: forwardState }));
+        yRev = this.backwardLayer.call(inputs, Object.assign(kwargs, { initialState: backwardState }));
+      }
+      let states;
+      if (this.returnState) {
+        if (Array.isArray(y)) {
+          states = y.slice(1).concat(yRev.slice(1));
+        } else {
+        }
+        y = y[0];
+        yRev = yRev[0];
+      }
+      if (this.returnSequences) {
+        yRev = reverse(yRev, 1);
+      }
+      let output;
+      if (this.mergeMode === "concat") {
+        output = concatenate([y, yRev]);
+      } else if (this.mergeMode === "sum") {
+        output = add2(y, yRev);
+      } else if (this.mergeMode === "ave") {
+        output = mul(0.5, add2(y, yRev));
+      } else if (this.mergeMode === "mul") {
+        output = mul(y, yRev);
+      } else if (this.mergeMode == null) {
+        output = [y, yRev];
+      }
+      if (this.returnState) {
+        if (this.mergeMode == null) {
+          return output.concat(states);
+        }
+        return [output].concat(states);
+      }
+      return output;
+    });
+  }
+  resetStates(states) {
+    this.forwardLayer.resetStates();
+    this.backwardLayer.resetStates();
+  }
+  build(inputShape) {
+    nameScope(this.forwardLayer.name, () => {
+      this.forwardLayer.build(inputShape);
+    });
+    nameScope(this.backwardLayer.name, () => {
+      this.backwardLayer.build(inputShape);
+    });
+    this.built = true;
+  }
+  computeMask(inputs, mask) {
+    if (Array.isArray(mask)) {
+      mask = mask[0];
+    }
+    let outputMask;
+    if (this.returnSequences) {
+      if (this.mergeMode == null) {
+        outputMask = [mask, mask];
+      } else {
+        outputMask = mask;
+      }
+    } else {
+      if (this.mergeMode == null) {
+        outputMask = [null, null];
+      } else {
+        outputMask = null;
+      }
+    }
+    if (this.returnState) {
+      const states = this.forwardLayer.states;
+      const stateMask = states.map((state) => null);
+      if (Array.isArray(outputMask)) {
+        return outputMask.concat(stateMask).concat(stateMask);
+      } else {
+        return [outputMask].concat(stateMask).concat(stateMask);
+      }
+    } else {
+      return outputMask;
+    }
+  }
+  get trainableWeights() {
+    return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights);
+  }
+  get nonTrainableWeights() {
+    return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights);
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.forwardLayer != null) {
+      this.forwardLayer.setFastWeightInitDuringBuild(value);
+    }
+    if (this.backwardLayer != null) {
+      this.backwardLayer.setFastWeightInitDuringBuild(value);
+    }
+  }
+  getConfig() {
+    const config = {
+      "mergeMode": this.mergeMode
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  static fromConfig(cls, config) {
+    const rnnLayer = deserialize(config["layer"]);
+    delete config["layer"];
+    if (config["numConstants"] != null) {
+      throw new NotImplementedError(`Deserialization of a Bidirectional layer with numConstants present is not supported yet.`);
+    }
+    const newConfig = config;
+    newConfig["layer"] = rnnLayer;
+    return new cls(newConfig);
+  }
+};
+Bidirectional.className = "Bidirectional";
+serialization_exports.registerClass(Bidirectional);
+var Rescaling = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.scale = args.scale;
+    if (args.offset) {
+      this.offset = args.offset;
+    } else {
+      this.offset = 0;
+    }
+  }
+  getConfig() {
+    const config = {
+      "scale": this.scale,
+      "offset": this.offset
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (inputs.dtype !== "float32") {
+        inputs = cast2(inputs, "float32");
+      }
+      return add2(mul(inputs, this.scale), this.offset);
+    });
+  }
+};
+Rescaling.className = "Rescaling";
+serialization_exports.registerClass(Rescaling);
+var INTERPOLATION_KEYS = ["bilinear", "nearest"];
+var INTERPOLATION_METHODS = new Set(INTERPOLATION_KEYS);
+var Resizing = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.height = args.height;
+    this.width = args.width;
+    if (args.interpolation) {
+      if (INTERPOLATION_METHODS.has(args.interpolation)) {
+        this.interpolation = args.interpolation;
+      } else {
+        throw new ValueError(`Invalid interpolation parameter: ${args.interpolation} is not implemented`);
+      }
+    } else {
+      this.interpolation = "bilinear";
+    }
+    this.cropToAspectRatio = Boolean(args.cropToAspectRatio);
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const numChannels = inputShape[2];
+    return [this.height, this.width, numChannels];
+  }
+  getConfig() {
+    const config = {
+      "height": this.height,
+      "width": this.width,
+      "interpolation": this.interpolation,
+      "cropToAspectRatio": this.cropToAspectRatio
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const size = [this.height, this.width];
+      if (this.interpolation === "bilinear") {
+        return image.resizeBilinear(inputs, size, !this.cropToAspectRatio);
+      } else if (this.interpolation === "nearest") {
+        return image.resizeNearestNeighbor(inputs, size, !this.cropToAspectRatio);
+      } else {
+        throw new Error(`Interpolation is ${this.interpolation} but only ${[...INTERPOLATION_METHODS]} are supported`);
+      }
+    });
+  }
+};
+Resizing.className = "Resizing";
+serialization_exports.registerClass(Resizing);
+function encodeCategoricalInputs(inputs, outputMode, depth, weights) {
+  let input2 = getExactlyOneTensor(inputs);
+  if (input2.dtype !== "int32") {
+    input2 = cast2(input2, "int32");
+  }
+  if (outputMode === "int") {
+    return input2;
+  }
+  const originalShape = input2.shape;
+  if (input2.rank === 0) {
+    input2 = expandDims(input2, -1);
+  }
+  if (outputMode === "oneHot") {
+    if (input2.shape[input2.shape.length - 1] !== 1) {
+      input2 = expandDims(input2, -1);
+    }
+  }
+  if (input2.rank > 2) {
+    throw new ValueError(`When outputMode is not int, maximum output rank is 2 Received outputMode ${outputMode} and input shape ${originalShape} which would result in output rank ${input2.rank}.`);
+  }
+  const binaryOutput = ["multiHot", "oneHot"].includes(outputMode);
+  const denseBincountInput = input2;
+  let binCounts;
+  if (typeof weights !== "undefined" && outputMode === "count") {
+    binCounts = denseBincount(denseBincountInput, weights, depth, binaryOutput);
+  } else {
+    binCounts = denseBincount(denseBincountInput, [], depth, binaryOutput);
+  }
+  if (outputMode !== "tfIdf") {
+    return binCounts;
+  }
+  if (weights) {
+    return mul(binCounts, weights);
+  } else {
+    throw new ValueError(`When outputMode is 'tfIdf', weights must be provided.`);
+  }
+}
+var CategoryEncoding = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.numTokens = args.numTokens;
+    if (args.outputMode) {
+      this.outputMode = args.outputMode;
+    } else {
+      this.outputMode = "multiHot";
+    }
+  }
+  getConfig() {
+    const config = {
+      "numTokens": this.numTokens,
+      "outputMode": this.outputMode
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape == null) {
+      return [this.numTokens];
+    }
+    if (this.outputMode === "oneHot" && inputShape[inputShape.length - 1] !== 1) {
+      inputShape.push(this.numTokens);
+      return inputShape;
+    }
+    inputShape[inputShape.length - 1] = this.numTokens;
+    return inputShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (inputs.dtype !== "int32") {
+        inputs = cast2(inputs, "int32");
+      }
+      let countWeights;
+      if (typeof kwargs["countWeights"] !== "undefined") {
+        if (this.outputMode !== "count") {
+          throw new ValueError(`countWeights is not used when outputMode !== count.
+              Received countWeights=${kwargs["countWeights"]}`);
+        }
+        countWeights = getExactlyOneTensor(kwargs["countWeights"]);
+      }
+      const maxValue = max(inputs);
+      const minValue = min(inputs);
+      const greaterEqualMax = greater(this.numTokens, maxValue).bufferSync().get(0);
+      const greaterMin = greaterEqual(minValue, 0).bufferSync().get(0);
+      if (!(greaterEqualMax && greaterMin)) {
+        throw new ValueError(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);
+      }
+      return encodeCategoricalInputs(inputs, this.outputMode, this.numTokens, countWeights);
+    });
+  }
+};
+CategoryEncoding.className = "CategoryEncoding";
+serialization_exports.registerClass(CategoryEncoding);
+function inputLayer(args) {
+  return new InputLayer(args);
+}
+function elu3(args) {
+  return new ELU(args);
+}
+function reLU(args) {
+  return new ReLU(args);
+}
+function leakyReLU(args) {
+  return new LeakyReLU(args);
+}
+function prelu2(args) {
+  return new PReLU(args);
+}
+function softmax2(args) {
+  return new Softmax3(args);
+}
+function thresholdedReLU(args) {
+  return new ThresholdedReLU(args);
+}
+function conv1d2(args) {
+  return new Conv1D(args);
+}
+function conv2d3(args) {
+  return new Conv2D2(args);
+}
+function conv2dTranspose2(args) {
+  return new Conv2DTranspose(args);
+}
+function conv3d2(args) {
+  return new Conv3D2(args);
+}
+function conv3dTranspose2(args) {
+  return new Conv3DTranspose(args);
+}
+function separableConv2d2(args) {
+  return new SeparableConv2D(args);
+}
+function cropping2D(args) {
+  return new Cropping2D(args);
+}
+function upSampling2d(args) {
+  return new UpSampling2D(args);
+}
+function depthwiseConv2d4(args) {
+  return new DepthwiseConv2D(args);
+}
+function activation(args) {
+  return new Activation2(args);
+}
+function dense(args) {
+  return new Dense(args);
+}
+function dropout3(args) {
+  return new Dropout(args);
+}
+function spatialDropout1d(args) {
+  return new SpatialDropout1D(args);
+}
+function flatten3(args) {
+  return new Flatten(args);
+}
+function repeatVector(args) {
+  return new RepeatVector(args);
+}
+function reshape2(args) {
+  return new Reshape2(args);
+}
+function permute(args) {
+  return new Permute(args);
+}
+function embedding(args) {
+  return new Embedding(args);
+}
+function add3(args) {
+  return new Add2(args);
+}
+function average(args) {
+  return new Average(args);
+}
+function concatenate2(args) {
+  return new Concatenate(args);
+}
+function maximum2(args) {
+  return new Maximum2(args);
+}
+function minimum2(args) {
+  return new Minimum2(args);
+}
+function multiply(args) {
+  return new Multiply2(args);
+}
+function dot3(args) {
+  return new Dot(args);
+}
+function batchNormalization2(args) {
+  return new BatchNormalization(args);
+}
+function layerNormalization(args) {
+  return new LayerNormalization(args);
+}
+function zeroPadding2d(args) {
+  return new ZeroPadding2D(args);
+}
+function averagePooling1d(args) {
+  return new AveragePooling1D(args);
+}
+function avgPool1d(args) {
+  return averagePooling1d(args);
+}
+function avgPooling1d(args) {
+  return averagePooling1d(args);
+}
+function averagePooling2d(args) {
+  return new AveragePooling2D(args);
+}
+function avgPool2d(args) {
+  return averagePooling2d(args);
+}
+function avgPooling2d(args) {
+  return averagePooling2d(args);
+}
+function averagePooling3d(args) {
+  return new AveragePooling3D(args);
+}
+function avgPool3d2(args) {
+  return averagePooling3d(args);
+}
+function avgPooling3d(args) {
+  return averagePooling3d(args);
+}
+function globalAveragePooling1d(args) {
+  return new GlobalAveragePooling1D(args);
+}
+function globalAveragePooling2d(args) {
+  return new GlobalAveragePooling2D(args);
+}
+function globalMaxPooling1d(args) {
+  return new GlobalMaxPooling1D(args);
+}
+function globalMaxPooling2d(args) {
+  return new GlobalMaxPooling2D(args);
+}
+function maxPooling1d(args) {
+  return new MaxPooling1D(args);
+}
+function maxPooling2d(args) {
+  return new MaxPooling2D(args);
+}
+function maxPooling3d(args) {
+  return new MaxPooling3D(args);
+}
+function gru(args) {
+  return new GRU(args);
+}
+function gruCell(args) {
+  return new GRUCell(args);
+}
+function lstm(args) {
+  return new LSTM(args);
+}
+function lstmCell(args) {
+  return new LSTMCell(args);
+}
+function simpleRNN(args) {
+  return new SimpleRNN(args);
+}
+function simpleRNNCell(args) {
+  return new SimpleRNNCell(args);
+}
+function convLstm2d(args) {
+  return new ConvLSTM2D(args);
+}
+function convLstm2dCell(args) {
+  return new ConvLSTM2DCell(args);
+}
+function rnn2(args) {
+  return new RNN(args);
+}
+function stackedRNNCells(args) {
+  return new StackedRNNCells(args);
+}
+function bidirectional(args) {
+  return new Bidirectional(args);
+}
+function timeDistributed(args) {
+  return new TimeDistributed(args);
+}
+var globalMaxPool1d = globalMaxPooling1d;
+var globalMaxPool2d = globalMaxPooling2d;
+var maxPool1d = maxPooling1d;
+var maxPool2d = maxPooling2d;
+function gaussianNoise(args) {
+  return new GaussianNoise(args);
+}
+function gaussianDropout(args) {
+  return new GaussianDropout(args);
+}
+function alphaDropout(args) {
+  return new AlphaDropout(args);
+}
+function masking(args) {
+  return new Masking(args);
+}
+function rescaling(args) {
+  return new Rescaling(args);
+}
+function resizing(args) {
+  return new Resizing(args);
+}
+function categoryEncoding(args) {
+  return new CategoryEncoding(args);
+}
+var exports_metrics_exports = {};
+__export2(exports_metrics_exports, {
+  MAPE: () => MAPE2,
+  MSE: () => MSE2,
+  binaryAccuracy: () => binaryAccuracy2,
+  binaryCrossentropy: () => binaryCrossentropy3,
+  categoricalAccuracy: () => categoricalAccuracy2,
+  categoricalCrossentropy: () => categoricalCrossentropy3,
+  cosineProximity: () => cosineProximity2,
+  mape: () => mape2,
+  meanAbsoluteError: () => meanAbsoluteError2,
+  meanAbsolutePercentageError: () => meanAbsolutePercentageError2,
+  meanSquaredError: () => meanSquaredError3,
+  mse: () => mse2,
+  precision: () => precision2,
+  recall: () => recall2,
+  sparseCategoricalAccuracy: () => sparseCategoricalAccuracy2
+});
+function binaryAccuracy2(yTrue, yPred) {
+  return binaryAccuracy(yTrue, yPred);
+}
+function binaryCrossentropy3(yTrue, yPred) {
+  return binaryCrossentropy2(yTrue, yPred);
+}
+function sparseCategoricalAccuracy2(yTrue, yPred) {
+  return sparseCategoricalAccuracy(yTrue, yPred);
+}
+function categoricalAccuracy2(yTrue, yPred) {
+  return categoricalAccuracy(yTrue, yPred);
+}
+function categoricalCrossentropy3(yTrue, yPred) {
+  return categoricalCrossentropy2(yTrue, yPred);
+}
+function precision2(yTrue, yPred) {
+  return precision(yTrue, yPred);
+}
+function recall2(yTrue, yPred) {
+  return recall(yTrue, yPred);
+}
+function cosineProximity2(yTrue, yPred) {
+  return cosineProximity(yTrue, yPred);
+}
+function meanAbsoluteError2(yTrue, yPred) {
+  return meanAbsoluteError(yTrue, yPred);
+}
+function meanAbsolutePercentageError2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function MAPE2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function mape2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function meanSquaredError3(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+function MSE2(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+function mse2(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+var exports_models_exports = {};
+__export2(exports_models_exports, {
+  modelFromJSON: () => modelFromJSON
+});
+var exports_regularizers_exports = {};
+__export2(exports_regularizers_exports, {
+  l1: () => l12,
+  l1l2: () => l1l2,
+  l2: () => l22
+});
+function l1l2(config) {
+  return new L1L2(config);
+}
+function l12(config) {
+  return l1(config);
+}
+function l22(config) {
+  return l2(config);
+}
+var Callback = class extends BaseCallback {
+  constructor() {
+    super(...arguments);
+    this.model = null;
+  }
+  setModel(model2) {
+    if (!(model2 instanceof LayersModel)) {
+      throw new Error("model must be a LayersModel, not some other Container");
+    }
+    this.model = model2;
+  }
+};
+function less2(currVal, prevVal) {
+  return currVal < prevVal;
+}
+function greater2(currVal, prevVal) {
+  return currVal > prevVal;
+}
+var EarlyStopping = class extends Callback {
+  constructor(args) {
+    super();
+    if (args == null) {
+      args = {};
+    }
+    if (args.restoreBestWeights) {
+      throw new NotImplementedError("restoreBestWeights = True is not implemented in EarlyStopping yet.");
+    }
+    this.monitor = args.monitor || "val_loss";
+    this.minDelta = Math.abs(args.minDelta || 0);
+    this.patience = args.patience || 0;
+    this.verbose = args.verbose || 0;
+    this.mode = args.mode || "auto";
+    this.baseline = args.baseline;
+    if (["auto", "min", "max"].indexOf(this.mode) === -1) {
+      console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`);
+      this.mode = "auto";
+    }
+    if (this.mode === "min") {
+      this.monitorFunc = less2;
+    } else if (this.mode === "max") {
+      this.monitorFunc = greater2;
+    } else {
+      if (this.monitor.indexOf("acc") !== -1) {
+        this.monitorFunc = greater2;
+      } else {
+        this.monitorFunc = less2;
+      }
+    }
+    if (this.monitorFunc === less2) {
+      this.minDelta *= -1;
+    }
+  }
+  async onTrainBegin(logs) {
+    this.wait = 0;
+    this.stoppedEpoch = 0;
+    if (this.baseline != null) {
+      this.best = this.baseline;
+    } else {
+      this.best = this.monitorFunc === less2 ? Infinity : -Infinity;
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    await resolveScalarsInLogs(logs);
+    const current = this.getMonitorValue(logs);
+    if (current == null) {
+      return;
+    }
+    if (this.monitorFunc(current - this.minDelta, this.best)) {
+      this.best = current;
+      this.wait = 0;
+    } else {
+      this.wait++;
+      if (this.wait >= this.patience) {
+        this.stoppedEpoch = epoch;
+        this.model.stopTraining = true;
+      }
+    }
+  }
+  async onTrainEnd(logs) {
+    if (this.stoppedEpoch > 0 && this.verbose) {
+      console.log(`Epoch ${this.stoppedEpoch}: early stopping.`);
+    }
+  }
+  getMonitorValue(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    const monitorValue = logs[this.monitor];
+    if (monitorValue == null) {
+      console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(logs)}`);
+    }
+    return monitorValue;
+  }
+};
+function earlyStopping(args) {
+  return new EarlyStopping(args);
+}
+var callbacks = { earlyStopping };
+var ENV4 = env();
+ENV4.registerFlag("KEEP_INTERMEDIATE_TENSORS", () => false, (debugValue) => {
+  if (debugValue) {
+    console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.");
+  }
+});
+var DataType;
+(function(DataType2) {
+  DataType2[DataType2["DT_INVALID"] = 0] = "DT_INVALID";
+  DataType2[DataType2["DT_FLOAT"] = 1] = "DT_FLOAT";
+  DataType2[DataType2["DT_DOUBLE"] = 2] = "DT_DOUBLE";
+  DataType2[DataType2["DT_INT32"] = 3] = "DT_INT32";
+  DataType2[DataType2["DT_UINT8"] = 4] = "DT_UINT8";
+  DataType2[DataType2["DT_INT16"] = 5] = "DT_INT16";
+  DataType2[DataType2["DT_INT8"] = 6] = "DT_INT8";
+  DataType2[DataType2["DT_STRING"] = 7] = "DT_STRING";
+  DataType2[DataType2["DT_COMPLEX64"] = 8] = "DT_COMPLEX64";
+  DataType2[DataType2["DT_INT64"] = 9] = "DT_INT64";
+  DataType2[DataType2["DT_BOOL"] = 10] = "DT_BOOL";
+  DataType2[DataType2["DT_QINT8"] = 11] = "DT_QINT8";
+  DataType2[DataType2["DT_QUINT8"] = 12] = "DT_QUINT8";
+  DataType2[DataType2["DT_QINT32"] = 13] = "DT_QINT32";
+  DataType2[DataType2["DT_BFLOAT16"] = 14] = "DT_BFLOAT16";
+  DataType2[DataType2["DT_QINT16"] = 15] = "DT_QINT16";
+  DataType2[DataType2["DT_QUINT16"] = 16] = "DT_QUINT16";
+  DataType2[DataType2["DT_UINT16"] = 17] = "DT_UINT16";
+  DataType2[DataType2["DT_COMPLEX128"] = 18] = "DT_COMPLEX128";
+  DataType2[DataType2["DT_HALF"] = 19] = "DT_HALF";
+  DataType2[DataType2["DT_RESOURCE"] = 20] = "DT_RESOURCE";
+  DataType2[DataType2["DT_VARIANT"] = 21] = "DT_VARIANT";
+  DataType2[DataType2["DT_UINT32"] = 22] = "DT_UINT32";
+  DataType2[DataType2["DT_UINT64"] = 23] = "DT_UINT64";
+  DataType2[DataType2["DT_FLOAT_REF"] = 101] = "DT_FLOAT_REF";
+  DataType2[DataType2["DT_DOUBLE_REF"] = 102] = "DT_DOUBLE_REF";
+  DataType2[DataType2["DT_INT32_REF"] = 103] = "DT_INT32_REF";
+  DataType2[DataType2["DT_UINT8_REF"] = 104] = "DT_UINT8_REF";
+  DataType2[DataType2["DT_INT16_REF"] = 105] = "DT_INT16_REF";
+  DataType2[DataType2["DT_INT8_REF"] = 106] = "DT_INT8_REF";
+  DataType2[DataType2["DT_STRING_REF"] = 107] = "DT_STRING_REF";
+  DataType2[DataType2["DT_COMPLEX64_REF"] = 108] = "DT_COMPLEX64_REF";
+  DataType2[DataType2["DT_INT64_REF"] = 109] = "DT_INT64_REF";
+  DataType2[DataType2["DT_BOOL_REF"] = 110] = "DT_BOOL_REF";
+  DataType2[DataType2["DT_QINT8_REF"] = 111] = "DT_QINT8_REF";
+  DataType2[DataType2["DT_QUINT8_REF"] = 112] = "DT_QUINT8_REF";
+  DataType2[DataType2["DT_QINT32_REF"] = 113] = "DT_QINT32_REF";
+  DataType2[DataType2["DT_BFLOAT16_REF"] = 114] = "DT_BFLOAT16_REF";
+  DataType2[DataType2["DT_QINT16_REF"] = 115] = "DT_QINT16_REF";
+  DataType2[DataType2["DT_QUINT16_REF"] = 116] = "DT_QUINT16_REF";
+  DataType2[DataType2["DT_UINT16_REF"] = 117] = "DT_UINT16_REF";
+  DataType2[DataType2["DT_COMPLEX128_REF"] = 118] = "DT_COMPLEX128_REF";
+  DataType2[DataType2["DT_HALF_REF"] = 119] = "DT_HALF_REF";
+  DataType2[DataType2["DT_RESOURCE_REF"] = 120] = "DT_RESOURCE_REF";
+  DataType2[DataType2["DT_VARIANT_REF"] = 121] = "DT_VARIANT_REF";
+  DataType2[DataType2["DT_UINT32_REF"] = 122] = "DT_UINT32_REF";
+  DataType2[DataType2["DT_UINT64_REF"] = 123] = "DT_UINT64_REF";
+})(DataType || (DataType = {}));
+var SaverDef;
+(function(SaverDef2) {
+  let CheckpointFormatVersion;
+  (function(CheckpointFormatVersion2) {
+    CheckpointFormatVersion2[CheckpointFormatVersion2["LEGACY"] = 0] = "LEGACY";
+    CheckpointFormatVersion2[CheckpointFormatVersion2["V1"] = 1] = "V1";
+    CheckpointFormatVersion2[CheckpointFormatVersion2["V2"] = 2] = "V2";
+  })(CheckpointFormatVersion = SaverDef2.CheckpointFormatVersion || (SaverDef2.CheckpointFormatVersion = {}));
+})(SaverDef || (SaverDef = {}));
+var CUSTOM_OPS = {};
+function registerOp(name, opFunc) {
+  const opMapper = {
+    tfOpName: name,
+    category: "custom",
+    inputs: [],
+    attrs: [],
+    customExecutor: opFunc
+  };
+  CUSTOM_OPS[name] = opMapper;
+}
+function getRegisteredOp(name) {
+  return CUSTOM_OPS[name];
+}
+function deregisterOp(name) {
+  delete CUSTOM_OPS[name];
+}
+function getParamValue(paramName, node, tensorMap, context, resourceManager) {
+  const inputParam = node.inputParams[paramName];
+  if (inputParam && inputParam.inputIndexStart !== void 0) {
+    const start = inputParam.inputIndexStart;
+    const end = inputParam.inputIndexEnd === 0 ? void 0 : inputParam.inputIndexEnd === void 0 ? start + 1 : inputParam.inputIndexEnd;
+    if (inputParam.type === "tensor") {
+      return getTensor(node.inputNames[inputParam.inputIndexStart], tensorMap, context, resourceManager);
+    }
+    if (inputParam.type === "tensors") {
+      const inputs = node.inputNames.slice(start, end);
+      return inputs.map((name) => getTensor(name, tensorMap, context, resourceManager));
+    }
+    const tensor2 = getTensor(node.inputNames.slice(start)[0], tensorMap, context, resourceManager);
+    const data = tensor2.dataSync();
+    return inputParam.type === "number" ? data[0] : util_exports.toNestedArray(tensor2.shape, data);
+  }
+  const attrParam = node.attrParams[paramName];
+  return attrParam && attrParam.value;
+}
+function getTensor(name, tensorsMap, context, resourceManager) {
+  const [nodeName, index] = parseNodeName(name);
+  if (resourceManager != null) {
+    const tensor2 = resourceManager.getHashTableHandleByName(nodeName);
+    if (tensor2 != null) {
+      return tensor2;
+    }
+  }
+  const contextId = context.currentContextIds.find((contextId2) => {
+    return !!tensorsMap[getNodeNameWithContextId(nodeName, contextId2)];
+  });
+  return contextId !== void 0 ? tensorsMap[getNodeNameWithContextId(nodeName, contextId)][index] : void 0;
+}
+function getTensorsForCurrentContenxt(name, tensorsMap, context) {
+  return tensorsMap[getNodeNameWithContextId(name, context.currentContextId)];
+}
+function getNodeNameAndIndex(inputName, context) {
+  const [nodeName, index, outputName] = parseNodeName(inputName);
+  return [
+    getNodeNameWithContextId(nodeName, context && context.currentContextId),
+    index,
+    outputName
+  ];
+}
+function getNodeNameWithContextId(name, contextId) {
+  return !!contextId ? `${name}-${contextId}` : name;
+}
+function parseNodeName(name) {
+  const parts = name.split(":");
+  if (parts.length === 1) {
+    return [name, 0, void 0];
+  }
+  const nodeName = parts[0];
+  const outputName = parts.length === 3 ? parts[1] : void 0;
+  const index = Number(parts[parts.length - 1]);
+  return [nodeName, index, outputName];
+}
+function getPadding(node, tensorMap, context) {
+  let pad3 = getParamValue("pad", node, tensorMap, context);
+  if (pad3 === "explicit") {
+    pad3 = getParamValue("explicitPaddings", node, tensorMap, context);
+    const explicitPadding = [[0, 0], [0, 0], [0, 0], [0, 0]];
+    for (let i = 0; i < 4; i++) {
+      explicitPadding[i][0] = pad3[i * 2];
+      explicitPadding[i][1] = pad3[i * 2 + 1];
+    }
+    return explicitPadding;
+  }
+  return pad3;
+}
+function cloneTensor(tensor2) {
+  return tensor2.kept ? tensor2 : clone(tensor2);
+}
+var arithmetic_exports = {};
+__export2(arithmetic_exports, {
+  json: () => json
+});
+var json = [
+  {
+    "tfOpName": "Add",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AddV2",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AddN",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BiasAdd",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sub",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "RealDiv",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Div",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "DivNoNan",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FloorDiv",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mul",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Maximum",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Minimum",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pow",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SquaredDifference",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mod",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FloorMod",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+var basic_math_exports = {};
+__export2(basic_math_exports, {
+  json: () => json2
+});
+var json2 = [
+  {
+    "tfOpName": "Abs",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Acos",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Asin",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atan2",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "y",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Ceil",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ClipByValue",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "clipValueMin",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "clipValueMax",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Complex",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "real",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "imag",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ComplexAbs",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cos",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cosh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Elu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Exp",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Floor",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Log",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Imag",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "outputType",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Neg",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Real",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "outputType",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prelu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "alpha",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Relu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Relu6",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Selu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sigmoid",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sin",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sinh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sqrt",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Rsqrt",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Square",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tanh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sign",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Round",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Expm1",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Log1p",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reciprocal",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Softplus",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Asinh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Acosh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atanh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Erf",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prod",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axes",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool",
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LeakyRelu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "alpha",
+        "name": "alpha",
+        "type": "number",
+        "defaultValue": 0.2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "IsNan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+var control_exports = {};
+__export2(control_exports, {
+  json: () => json3
+});
+var json3 = [
+  {
+    "tfOpName": "EmptyTensorList",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 1,
+        "name": "maxNumElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LoopCond",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "pred",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Switch",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "pred",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Merge",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Enter",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "frame_name",
+        "name": "frameName",
+        "type": "string"
+      },
+      {
+        "tfName": "is_constant",
+        "name": "isConstant",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Exit",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "NextIteration",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "size",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dynamic_size",
+        "name": "dynamicSize",
+        "type": "bool"
+      },
+      {
+        "tfName": "clear_after_read",
+        "name": "clearAfterRead",
+        "type": "bool"
+      },
+      {
+        "tfName": "identical_element_shapes",
+        "name": "identicalElementShapes",
+        "type": "bool"
+      },
+      {
+        "tfName": "tensor_array_name",
+        "name": "name",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayWriteV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayReadV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayGatherV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayScatterV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayConcatV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape_except0",
+        "name": "elementShapeExcept0",
+        "type": "shape",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArraySplitV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "lengths",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArraySizeV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayCloseV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StatelessIf",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "cond",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "then_branch",
+        "name": "thenBranch",
+        "type": "func"
+      },
+      {
+        "tfName": "else_branch",
+        "name": "elseBranch",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "If",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "cond",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "then_branch",
+        "name": "thenBranch",
+        "type": "func"
+      },
+      {
+        "tfName": "else_branch",
+        "name": "elseBranch",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StatelessWhile",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "cond",
+        "name": "cond",
+        "type": "func"
+      },
+      {
+        "tfName": "body",
+        "name": "body",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "While",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "cond",
+        "name": "cond",
+        "type": "func"
+      },
+      {
+        "tfName": "body",
+        "name": "body",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListScatter",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListScatterV2",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 3,
+        "name": "numElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListGather",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListGetItem",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListSetItem",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListReserve",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 1,
+        "name": "numElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListFromTensor",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListStack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "num_elements",
+        "name": "numElements",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListSplit",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 2,
+        "name": "lengths",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListConcat",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListConcatV2",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListPopBack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListPushBack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListLength",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListResize",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      }
+    ]
+  }
+];
+var convolution_exports = {};
+__export2(convolution_exports, {
+  json: () => json4
+});
+var json4 = [
+  {
+    "tfOpName": "AvgPool",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPool",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": [],
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPoolWithArgmax",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "include_batch_in_index",
+        "name": "includeBatchInIndex",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AvgPool3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPool3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv1D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "stride",
+        "name": "stride",
+        "type": "number"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NWC"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "dilation",
+        "name": "dilation",
+        "type": "number",
+        "defaultValue": 1
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "useCudnnOnGpu",
+        "name": "useCudnnOnGpu",
+        "type": "bool"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "_FusedConv2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "use_cudnn_on_gpu",
+        "name": "useCudnnOnGpu",
+        "type": "bool",
+        "defaultValue": true
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "defaultValue": [
+          1,
+          1,
+          1,
+          1
+        ]
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-4
+      },
+      {
+        "tfName": "leakyrelu_alpha",
+        "name": "leakyreluAlpha",
+        "type": "number",
+        "defaultValue": 0.2
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv2DBackpropInput",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 2,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 0,
+        "name": "outputShape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthwiseConv2d",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthwiseConv2dNative",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedDepthwiseConv2dNative",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "defaultValue": [
+          1,
+          1,
+          1,
+          1
+        ]
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Dilation2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "rates",
+        "name": "dilations",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      }
+    ]
+  }
+];
+var creation_exports = {};
+__export2(creation_exports, {
+  json: () => json5
+});
+var json5 = [
+  {
+    "tfOpName": "Fill",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      },
+      {
+        "start": 1,
+        "name": "value",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LinSpace",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "start",
+        "type": "number"
+      },
+      {
+        "start": 1,
+        "name": "stop",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "num",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "OneHot",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "depth",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "onValue",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "start": 3,
+        "name": "offValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Ones",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "OnesLike",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "RandomStandardNormal",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "RandomUniform",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "minval",
+        "name": "minval",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "maxval",
+        "name": "maxval",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Range",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "start",
+        "type": "number"
+      },
+      {
+        "start": 1,
+        "name": "stop",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "step",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tidx",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TruncatedNormal",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "means",
+        "name": "mean",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "stddev",
+        "name": "stdDev",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number"
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Zeros",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ZerosLike",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Multinomial",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "logits",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "numSamples",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number"
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "output_dtype",
+        "name": "output_dtype",
+        "type": "dtype"
+      }
+    ]
+  }
+];
+var dynamic_exports = {};
+__export2(dynamic_exports, {
+  json: () => json6
+});
+var json6 = [
+  {
+    "tfOpName": "NonMaxSuppressionV2",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV3",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV4",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "T_threshold",
+        "name": "threshold",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "pad_to_max_output_size",
+        "name": "padToMaxOutputSize",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV5",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      },
+      {
+        "start": 5,
+        "name": "softNmsSigma",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Where",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ListDiff",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "y",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+var evaluation_exports = {};
+__export2(evaluation_exports, {
+  json: () => json7
+});
+var json7 = [
+  {
+    "tfOpName": "LowerBound",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sortedSequence",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TopKV2",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "k",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "sorted",
+        "name": "sorted",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "UpperBound",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sortedSequence",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Unique",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "UniqueV2",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  }
+];
+var graph_exports = {};
+__export2(graph_exports, {
+  json: () => json8
+});
+var json8 = [
+  {
+    "tfOpName": "PlaceholderWithDefault",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "default",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "shape",
+        "name": "shape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Placeholder",
+    "category": "graph",
+    "attrs": [
+      {
+        "tfName": "shape",
+        "name": "shape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Const",
+    "category": "graph"
+  },
+  {
+    "tfOpName": "Identity",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "IdentityN",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "x",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Snapshot",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Rank",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Size",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Shape",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ShapeN",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "x",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Print",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "data",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "message",
+        "name": "message",
+        "type": "string"
+      },
+      {
+        "tfName": "first_n",
+        "name": "firstN",
+        "type": "number",
+        "notSupported": true
+      },
+      {
+        "tfName": "summarize",
+        "name": "summarize",
+        "type": "number",
+        "defaultValue": 3
+      }
+    ]
+  },
+  {
+    "tfOpName": "NoOp",
+    "category": "graph",
+    "inputs": []
+  },
+  {
+    "tfOpName": "StopGradient",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "FakeQuantWithMinMaxVars",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "min",
+        "name": "min",
+        "type": "number"
+      },
+      {
+        "tfName": "max",
+        "name": "max",
+        "type": "number"
+      }
+    ]
+  }
+];
+var hash_table_exports = {};
+__export2(hash_table_exports, {
+  json: () => json9
+});
+var json9 = [
+  {
+    "tfOpName": "HashTable",
+    "category": "hash_table",
+    "inputs": [],
+    "attrs": [
+      {
+        "tfName": "shared_name",
+        "name": "sharedName",
+        "type": "string"
+      },
+      {
+        "tfName": "use_node_name_sharing",
+        "name": "useNodeNameSharing",
+        "type": "bool"
+      },
+      {
+        "tfName": "key_dtype",
+        "name": "keyDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "value_dtype",
+        "name": "valueDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "HashTableV2",
+    "category": "hash_table",
+    "inputs": [],
+    "attrs": [
+      {
+        "tfName": "shared_name",
+        "name": "sharedName",
+        "type": "string"
+      },
+      {
+        "tfName": "use_node_name_sharing",
+        "name": "useNodeNameSharing",
+        "type": "bool"
+      },
+      {
+        "tfName": "key_dtype",
+        "name": "keyDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "value_dtype",
+        "name": "valueDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableImport",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "values",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableImportV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "values",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableFind",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableFindV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableSize",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableSizeV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      }
+    ]
+  }
+];
+var image_exports = {};
+__export2(image_exports, {
+  json: () => json10
+});
+var json10 = [
+  {
+    "tfOpName": "ResizeBilinear",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "align_corners",
+        "name": "alignCorners",
+        "type": "bool"
+      },
+      {
+        "tfName": "half_pixel_centers",
+        "name": "halfPixelCenters",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ResizeNearestNeighbor",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "align_corners",
+        "name": "alignCorners",
+        "type": "bool"
+      },
+      {
+        "tfName": "half_pixel_centers",
+        "name": "halfPixelCenters",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "CropAndResize",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "image",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "boxInd",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "cropSize",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "method",
+        "name": "method",
+        "type": "string"
+      },
+      {
+        "tfName": "extrapolation_value",
+        "name": "extrapolationValue",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ImageProjectiveTransformV3",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "transforms",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "fillValue",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "interpolation",
+        "name": "interpolation",
+        "type": "string"
+      },
+      {
+        "tfName": "fill_mode",
+        "name": "fillMode",
+        "type": "string"
+      }
+    ]
+  }
+];
+var logical_exports = {};
+__export2(logical_exports, {
+  json: () => json11
+});
+var json11 = [
+  {
+    "tfOpName": "Equal",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "NotEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Greater",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "GreaterEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Less",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LessEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalAnd",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalNot",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalOr",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Select",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SelectV2",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+var matrices_exports = {};
+__export2(matrices_exports, {
+  json: () => json12
+});
+var json12 = [
+  {
+    "tfOpName": "_FusedMatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-4
+      },
+      {
+        "tfName": "transpose_a",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "transpose_b",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "leakyrelu_alpha",
+        "name": "leakyreluAlpha",
+        "type": "number",
+        "defaultValue": 0.2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "transpose_a",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "transpose_b",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchMatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "adj_x",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "adj_y",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchMatMulV2",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "adj_x",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "adj_y",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Transpose",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "perm",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Einsum",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "equation",
+        "name": "equation",
+        "type": "string"
+      },
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  }
+];
+var normalization_exports = {};
+__export2(normalization_exports, {
+  json: () => json13
+});
+var json13 = [
+  {
+    "tfOpName": "EuclideanNorm",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool",
+        "defaultValue": false
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNorm",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNormV2",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNormV3",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LRN",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "depth_radius",
+        "name": "radius",
+        "type": "number",
+        "defaultValue": 5
+      },
+      {
+        "tfName": "bias",
+        "name": "bias",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "alpha",
+        "name": "alpha",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "beta",
+        "name": "beta",
+        "type": "number",
+        "defaultValue": 0.5
+      }
+    ]
+  },
+  {
+    "tfOpName": "Softmax",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogSoftmax",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseToDense",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sparseIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "sparseValues",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "defaultValue": true,
+        "notSupported": true
+      }
+    ]
+  }
+];
+var reduction_exports = {};
+__export2(reduction_exports, {
+  json: () => json14
+});
+var json14 = [
+  {
+    "tfOpName": "Bincount",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "weights",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DenseBincount",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "weights",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "binary_output",
+        "name": "binaryOutput",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Max",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mean",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Min",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sum",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "All",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Any",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ArgMax",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ArgMin",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prod",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cumprod",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "exclusive",
+        "name": "exclusive",
+        "type": "bool"
+      },
+      {
+        "tfName": "reverse",
+        "name": "reverse",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cumsum",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "exclusive",
+        "name": "exclusive",
+        "type": "bool"
+      },
+      {
+        "tfName": "reverse",
+        "name": "reverse",
+        "type": "bool"
+      }
+    ]
+  }
+];
+var slice_join_exports = {};
+__export2(slice_join_exports, {
+  json: () => json15
+});
+var json15 = [
+  {
+    "tfOpName": "ConcatV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "end": -1,
+        "name": "tensors",
+        "type": "tensors"
+      },
+      {
+        "start": -1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      }
+    ]
+  },
+  {
+    "tfOpName": "Concat",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 1,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      },
+      {
+        "start": 0,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      }
+    ]
+  },
+  {
+    "tfOpName": "GatherV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "batch_dims",
+        "name": "batchDims",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Gather",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reverse",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "dims",
+        "type": "bool[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ReverseV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Slice",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "begin",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "size",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StridedSlice",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "begin",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "end",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "strides",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "begin_mask",
+        "name": "beginMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "end_mask",
+        "name": "endMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "new_axis_mask",
+        "name": "newAxisMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "ellipsis_mask",
+        "name": "ellipsisMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "shrink_axis_mask",
+        "name": "shrinkAxisMask",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pack",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Unpack",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "num",
+        "name": "num",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tile",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "reps",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Split",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "start": 1,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_split",
+        "name": "numOrSizeSplits",
+        "type": "number",
+        "defaultValue": 1
+      }
+    ]
+  },
+  {
+    "tfOpName": "SplitV",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "numOrSizeSplits",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "ScatterNd",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "GatherNd",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseToDense",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sparseIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "sparseValues",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "defaultValue": false,
+        "notSupported": true
+      }
+    ]
+  }
+];
+var sparse_exports = {};
+__export2(sparse_exports, {
+  json: () => json16
+});
+var json16 = [
+  {
+    "tfOpName": "SparseFillEmptyRows",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "denseShape",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseReshape",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "inputIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "inputShape",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "newShape",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseSegmentMean",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "segmentIds",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseSegmentSum",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "segmentIds",
+        "type": "tensor"
+      }
+    ]
+  }
+];
+var spectral_exports = {};
+__export2(spectral_exports, {
+  json: () => json17
+});
+var json17 = [
+  {
+    "tfOpName": "FFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "IFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "RFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "fft_length",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "IRFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "fft_length",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  }
+];
+var string_exports = {};
+__export2(string_exports, {
+  json: () => json18
+});
+var json18 = [
+  {
+    "tfOpName": "StringNGrams",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "dataSplits",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "separator",
+        "name": "separator",
+        "type": "string"
+      },
+      {
+        "tfName": "ngram_widths",
+        "name": "nGramWidths",
+        "type": "number[]"
+      },
+      {
+        "tfName": "left_pad",
+        "name": "leftPad",
+        "type": "string"
+      },
+      {
+        "tfName": "right_pad",
+        "name": "rightPad",
+        "type": "string"
+      },
+      {
+        "tfName": "pad_width",
+        "name": "padWidth",
+        "type": "number"
+      },
+      {
+        "tfName": "preserve_short_sequences",
+        "name": "preserveShortSequences",
+        "type": "bool"
+      }
+    ],
+    "outputs": [
+      "ngrams",
+      "ngrams_splits"
+    ]
+  },
+  {
+    "tfOpName": "StringSplit",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "delimiter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "skip_empty",
+        "name": "skipEmpty",
+        "type": "bool"
+      }
+    ],
+    "outputs": [
+      "indices",
+      "values",
+      "shape"
+    ]
+  },
+  {
+    "tfOpName": "StringToHashBucketFast",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_buckets",
+        "name": "numBuckets",
+        "type": "number"
+      }
+    ]
+  }
+];
+var transformation_exports = {};
+__export2(transformation_exports, {
+  json: () => json19
+});
+var json19 = [
+  {
+    "tfOpName": "Cast",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "SrcT",
+        "name": "sdtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "DstT",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ExpandDims",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "MirrorPad",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "mode",
+        "name": "mode",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pad",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "constant_value",
+        "name": "constantValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "PadV2",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "constantValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reshape",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Squeeze",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "tfDeprecatedName": "squeeze_dims",
+        "name": "axis",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SpaceToBatchND",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "blockShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "paddings",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchToSpaceND",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "blockShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "crops",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthToSpace",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "block_size",
+        "name": "blockSize",
+        "type": "number"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BroadcastTo",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": []
+  },
+  {
+    "tfOpName": "BroadcastArgs",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "s0",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "s1",
+        "type": "tensor"
+      }
+    ],
+    "attrs": []
+  }
+];
+var OperationMapper = class {
+  constructor() {
+    const ops = [
+      arithmetic_exports,
+      basic_math_exports,
+      control_exports,
+      convolution_exports,
+      creation_exports,
+      dynamic_exports,
+      evaluation_exports,
+      graph_exports,
+      hash_table_exports,
+      image_exports,
+      logical_exports,
+      matrices_exports,
+      normalization_exports,
+      reduction_exports,
+      slice_join_exports,
+      sparse_exports,
+      spectral_exports,
+      string_exports,
+      transformation_exports
+    ];
+    const mappersJson = [].concat(...ops.map((op2) => op2.json));
+    this.opMappers = mappersJson.reduce((map, mapper) => {
+      map[mapper.tfOpName] = mapper;
+      return map;
+    }, {});
+  }
+  static get Instance() {
+    return this._instance || (this._instance = new this());
+  }
+  transformGraph(graph, signature = {}) {
+    const tfNodes = graph.node;
+    const placeholders = [];
+    const weights = [];
+    const initNodes = [];
+    const nodes = tfNodes.reduce((map, node) => {
+      map[node.name] = this.mapNode(node);
+      if (node.op.startsWith("Placeholder")) {
+        placeholders.push(map[node.name]);
+      } else if (node.op === "Const") {
+        weights.push(map[node.name]);
+      } else if (node.input == null || node.input.length === 0) {
+        initNodes.push(map[node.name]);
+      }
+      return map;
+    }, {});
+    let inputs = [];
+    const outputs = [];
+    let inputNodeNameToKey = {};
+    let outputNodeNameToKey = {};
+    if (signature != null) {
+      inputNodeNameToKey = this.mapSignatureEntries(signature.inputs);
+      outputNodeNameToKey = this.mapSignatureEntries(signature.outputs);
+    }
+    const allNodes = Object.keys(nodes);
+    allNodes.forEach((key) => {
+      const node = nodes[key];
+      node.inputNames.forEach((name, index) => {
+        const [nodeName, , outputName] = getNodeNameAndIndex(name);
+        const inputNode = nodes[nodeName];
+        if (inputNode.outputs != null) {
+          const outputIndex = inputNode.outputs.indexOf(outputName);
+          if (outputIndex !== -1) {
+            const inputName = `${nodeName}:${outputIndex}`;
+            node.inputNames[index] = inputName;
+          }
+        }
+        node.inputs.push(inputNode);
+        inputNode.children.push(node);
+      });
+    });
+    if (Object.keys(outputNodeNameToKey).length === 0) {
+      allNodes.forEach((key) => {
+        const node = nodes[key];
+        if (node.children.length === 0) {
+          outputs.push(node);
+        }
+      });
+    } else {
+      Object.keys(outputNodeNameToKey).forEach((name) => {
+        const [nodeName] = getNodeNameAndIndex(name);
+        const node = nodes[nodeName];
+        if (node != null) {
+          node.signatureKey = outputNodeNameToKey[name];
+          outputs.push(node);
+        }
+      });
+    }
+    if (Object.keys(inputNodeNameToKey).length > 0) {
+      Object.keys(inputNodeNameToKey).forEach((name) => {
+        const [nodeName] = getNodeNameAndIndex(name);
+        const node = nodes[nodeName];
+        if (node) {
+          node.signatureKey = inputNodeNameToKey[name];
+          inputs.push(node);
+        }
+      });
+    } else {
+      inputs = placeholders;
+    }
+    let functions = {};
+    if (graph.library != null && graph.library.function != null) {
+      functions = graph.library.function.reduce((functions2, func2) => {
+        functions2[func2.signature.name] = this.mapFunction(func2);
+        return functions2;
+      }, {});
+    }
+    const result = { nodes, inputs, outputs, weights, placeholders, signature, functions };
+    if (initNodes.length > 0) {
+      result.initNodes = initNodes;
+    }
+    return result;
+  }
+  mapSignatureEntries(entries) {
+    return Object.keys(entries || {}).reduce((prev, curr) => {
+      prev[entries[curr].name] = curr;
+      return prev;
+    }, {});
+  }
+  mapNode(node) {
+    const mapper = getRegisteredOp(node.op) || this.opMappers[node.op] || {};
+    if (node.attr == null) {
+      node.attr = {};
+    }
+    const newNode = {
+      name: node.name,
+      op: node.op,
+      category: mapper.category,
+      inputNames: (node.input || []).map((input2) => input2.startsWith("^") ? input2.slice(1) : input2),
+      inputs: [],
+      children: [],
+      inputParams: {},
+      attrParams: {},
+      rawAttrs: node.attr,
+      outputs: mapper.outputs
+    };
+    if (mapper.inputs != null) {
+      newNode.inputParams = mapper.inputs.reduce((map, param) => {
+        map[param.name] = {
+          type: param.type,
+          inputIndexStart: param.start,
+          inputIndexEnd: param.end
+        };
+        return map;
+      }, {});
+    }
+    if (mapper.attrs != null) {
+      newNode.attrParams = mapper.attrs.reduce((map, param) => {
+        const type = param.type;
+        let value = void 0;
+        switch (param.type) {
+          case "string":
+            value = getStringParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getStringParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "string[]":
+            value = getStringArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getStringArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "number":
+            value = getNumberParam(node.attr, param.tfName, param.defaultValue || 0);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getNumberParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "number[]":
+            value = getNumericArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getNumericArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "bool":
+            value = getBoolParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getBoolParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "bool[]":
+            value = getBoolArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getBoolArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "shape":
+            value = getTensorShapeParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getTensorShapeParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "shape[]":
+            value = getTensorShapeArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getTensorShapeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "dtype":
+            value = getDtypeParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getDtypeParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "dtype[]":
+            value = getDtypeArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getDtypeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "func":
+            value = getFuncParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getFuncParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "tensor":
+          case "tensors":
+            break;
+          default:
+            throw new Error(`Unsupported param type: ${param.type} for op: ${node.op}`);
+        }
+        map[param.name] = { value, type };
+        return map;
+      }, {});
+    }
+    return newNode;
+  }
+  mapFunction(functionDef) {
+    const tfNodes = functionDef.nodeDef;
+    const placeholders = [];
+    const weights = [];
+    let nodes = {};
+    if (tfNodes != null) {
+      nodes = tfNodes.reduce((map, node) => {
+        map[node.name] = this.mapNode(node);
+        if (node.op === "Const") {
+          weights.push(map[node.name]);
+        }
+        return map;
+      }, {});
+    }
+    const inputs = [];
+    const outputs = [];
+    functionDef.signature.inputArg.forEach((arg) => {
+      const [nodeName] = getNodeNameAndIndex(arg.name);
+      const node = {
+        name: nodeName,
+        op: "Placeholder",
+        inputs: [],
+        inputNames: [],
+        category: "graph",
+        inputParams: {},
+        attrParams: { dtype: { value: parseDtypeParam(arg.type), type: "dtype" } },
+        children: []
+      };
+      node.signatureKey = arg.name;
+      inputs.push(node);
+      nodes[nodeName] = node;
+    });
+    const allNodes = Object.keys(nodes);
+    allNodes.forEach((key) => {
+      const node = nodes[key];
+      node.inputNames.forEach((name, index) => {
+        const [nodeName, , outputName] = getNodeNameAndIndex(name);
+        const inputNode = nodes[nodeName];
+        if (inputNode.outputs != null) {
+          const outputIndex = inputNode.outputs.indexOf(outputName);
+          if (outputIndex !== -1) {
+            const inputName = `${nodeName}:${outputIndex}`;
+            node.inputNames[index] = inputName;
+          }
+        }
+        node.inputs.push(inputNode);
+        inputNode.children.push(node);
+      });
+    });
+    const returnNodeMap = functionDef.ret;
+    functionDef.signature.outputArg.forEach((output) => {
+      const [nodeName, index] = getNodeNameAndIndex(returnNodeMap[output.name]);
+      const node = nodes[nodeName];
+      if (node != null) {
+        node.defaultOutput = index;
+        outputs.push(node);
+      }
+    });
+    const signature = this.mapArgsToSignature(functionDef);
+    return { nodes, inputs, outputs, weights, placeholders, signature };
+  }
+  mapArgsToSignature(functionDef) {
+    return {
+      methodName: functionDef.signature.name,
+      inputs: functionDef.signature.inputArg.reduce((map, arg) => {
+        map[arg.name] = this.mapArgToTensorInfo(arg);
+        return map;
+      }, {}),
+      outputs: functionDef.signature.outputArg.reduce((map, arg) => {
+        map[arg.name] = this.mapArgToTensorInfo(arg, functionDef.ret);
+        return map;
+      }, {})
+    };
+  }
+  mapArgToTensorInfo(arg, nameMap2) {
+    let name = arg.name;
+    if (nameMap2 != null) {
+      name = nameMap2[name];
+    }
+    return { name, dtype: arg.type };
+  }
+};
+function decodeBase64(text) {
+  const global2 = env().global;
+  if (typeof global2.atob !== "undefined") {
+    return global2.atob(text);
+  } else if (typeof Buffer !== "undefined") {
+    return new Buffer(text, "base64").toString();
+  } else {
+    throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()");
+  }
+}
+function parseStringParam(s, keepCase) {
+  const value = Array.isArray(s) ? String.fromCharCode.apply(null, s) : decodeBase64(s);
+  return keepCase ? value : value.toLowerCase();
+}
+function getStringParam(attrs, name, def, keepCase = false) {
+  const param = attrs[name];
+  if (param != null) {
+    return parseStringParam(param.s, keepCase);
+  }
+  return def;
+}
+function getBoolParam(attrs, name, def) {
+  const param = attrs[name];
+  return param ? param.b : def;
+}
+function getNumberParam(attrs, name, def) {
+  const param = attrs[name] || {};
+  const value = param["i"] != null ? param["i"] : param["f"] != null ? param["f"] : def;
+  return typeof value === "number" ? value : parseInt(value, 10);
+}
+function parseDtypeParam(value) {
+  if (typeof value === "string") {
+    value = DataType[value];
+  }
+  switch (value) {
+    case DataType.DT_FLOAT:
+    case DataType.DT_HALF:
+      return "float32";
+    case DataType.DT_INT32:
+    case DataType.DT_INT64:
+    case DataType.DT_INT8:
+    case DataType.DT_UINT8:
+      return "int32";
+    case DataType.DT_BOOL:
+      return "bool";
+    case DataType.DT_DOUBLE:
+      return "float32";
+    case DataType.DT_STRING:
+      return "string";
+    default:
+      return null;
+  }
+}
+function getFuncParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.func) {
+    return param.func.name;
+  }
+  return def;
+}
+function getDtypeParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.type) {
+    return parseDtypeParam(param.type);
+  }
+  return def;
+}
+function getDtypeArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.type) {
+    return param.list.type.map((v) => parseDtypeParam(v));
+  }
+  return def;
+}
+function parseTensorShapeParam(shape) {
+  if (shape.unknownRank) {
+    return void 0;
+  }
+  if (shape.dim != null) {
+    return shape.dim.map((dim) => typeof dim.size === "number" ? dim.size : parseInt(dim.size, 10));
+  }
+  return [];
+}
+function getTensorShapeParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.shape) {
+    return parseTensorShapeParam(param.shape);
+  }
+  return def;
+}
+function getNumericArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param) {
+    return ((param.list.f && param.list.f.length ? param.list.f : param.list.i) || []).map((v) => typeof v === "number" ? v : parseInt(v, 10));
+  }
+  return def;
+}
+function getStringArrayParam(attrs, name, def, keepCase = false) {
+  const param = attrs[name];
+  if (param && param.list && param.list.s) {
+    return param.list.s.map((v) => {
+      return parseStringParam(v, keepCase);
+    });
+  }
+  return def;
+}
+function getTensorShapeArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.shape) {
+    return param.list.shape.map((v) => {
+      return parseTensorShapeParam(v);
+    });
+  }
+  return def;
+}
+function getBoolArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.b) {
+    return param.list.b;
+  }
+  return def;
+}
+var NodeValueImpl = class {
+  constructor(node, tensorMap, context) {
+    this.node = node;
+    this.tensorMap = tensorMap;
+    this.context = context;
+    this.inputs = [];
+    this.attrs = {};
+    this.inputs = node.inputNames.map((name) => this.getInput(name));
+    if (node.rawAttrs != null) {
+      this.attrs = Object.keys(node.rawAttrs).reduce((attrs, key) => {
+        attrs[key] = this.getAttr(key);
+        return attrs;
+      }, {});
+    }
+  }
+  getInput(name) {
+    return getTensor(name, this.tensorMap, this.context);
+  }
+  getAttr(name, defaultValue) {
+    const value = this.node.rawAttrs[name];
+    if (value.tensor != null) {
+      return getTensor(name, this.tensorMap, this.context);
+    }
+    if (value.i != null || value.f != null) {
+      return getNumberParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.s != null) {
+      return getStringParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.b != null) {
+      return getBoolParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.shape != null) {
+      return getTensorShapeParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.type != null) {
+      return getDtypeParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.list != null) {
+      if (value.list.i != null || value.list.f != null) {
+        return getNumericArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.s != null) {
+        return getStringArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.shape != null) {
+        return getTensorShapeArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.b != null) {
+        return getBoolArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.type != null) {
+        return getDtypeArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+    }
+    return defaultValue;
+  }
+};
+var ops_for_converter_exports = {};
+__export2(ops_for_converter_exports, {
+  OP_SCOPE_SUFFIX: () => OP_SCOPE_SUFFIX,
+  abs: () => abs,
+  acos: () => acos,
+  acosh: () => acosh,
+  add: () => add2,
+  addN: () => addN,
+  all: () => all,
+  any: () => any,
+  argMax: () => argMax,
+  argMin: () => argMin,
+  asin: () => asin,
+  asinh: () => asinh,
+  atan: () => atan,
+  atan2: () => atan2,
+  atanh: () => atanh,
+  avgPool: () => avgPool,
+  avgPool3d: () => avgPool3d,
+  basicLSTMCell: () => basicLSTMCell,
+  batchNorm: () => batchNorm,
+  batchNorm2d: () => batchNorm2d,
+  batchNorm3d: () => batchNorm3d,
+  batchNorm4d: () => batchNorm4d,
+  batchToSpaceND: () => batchToSpaceND,
+  bincount: () => bincount,
+  booleanMaskAsync: () => booleanMaskAsync,
+  broadcastArgs: () => broadcastArgs,
+  broadcastTo: () => broadcastTo,
+  buffer: () => buffer,
+  cast: () => cast,
+  ceil: () => ceil,
+  clipByValue: () => clipByValue,
+  clone: () => clone,
+  complex: () => complex,
+  concat: () => concat,
+  concat1d: () => concat1d,
+  concat2d: () => concat2d,
+  concat3d: () => concat3d,
+  concat4d: () => concat4d,
+  conv1d: () => conv1d,
+  conv2d: () => conv2d,
+  conv2dTranspose: () => conv2dTranspose,
+  conv3d: () => conv3d,
+  conv3dTranspose: () => conv3dTranspose,
+  cos: () => cos,
+  cosh: () => cosh,
+  cosineWindow: () => cosineWindow,
+  cumprod: () => cumprod,
+  cumsum: () => cumsum,
+  denseBincount: () => denseBincount,
+  depthToSpace: () => depthToSpace,
+  depthwiseConv2d: () => depthwiseConv2d,
+  diag: () => diag,
+  dilation2d: () => dilation2d,
+  div: () => div,
+  divNoNan: () => divNoNan,
+  dot: () => dot,
+  dropout: () => dropout,
+  einsum: () => einsum,
+  elu: () => elu,
+  enclosingPowerOfTwo: () => enclosingPowerOfTwo,
+  equal: () => equal,
+  erf: () => erf,
+  euclideanNorm: () => euclideanNorm,
+  exp: () => exp,
+  expandDims: () => expandDims,
+  expm1: () => expm1,
+  eye: () => eye,
+  fft: () => fft,
+  fill: () => fill,
+  floor: () => floor,
+  floorDiv: () => floorDiv,
+  fused: () => fused_ops_exports,
+  gather: () => gather,
+  gatherND: () => gatherND,
+  greater: () => greater,
+  greaterEqual: () => greaterEqual,
+  ifft: () => ifft,
+  imag: () => imag,
+  image: () => image,
+  inTopKAsync: () => inTopKAsync,
+  irfft: () => irfft,
+  isFinite: () => isFinite2,
+  isInf: () => isInf,
+  isNaN: () => isNaN2,
+  leakyRelu: () => leakyRelu,
+  less: () => less,
+  lessEqual: () => lessEqual,
+  linalg: () => linalg,
+  linspace: () => linspace,
+  localResponseNormalization: () => localResponseNormalization,
+  log: () => log2,
+  log1p: () => log1p,
+  logSigmoid: () => logSigmoid,
+  logSoftmax: () => logSoftmax,
+  logSumExp: () => logSumExp,
+  logicalAnd: () => logicalAnd,
+  logicalNot: () => logicalNot,
+  logicalOr: () => logicalOr,
+  logicalXor: () => logicalXor,
+  losses: () => losses,
+  lowerBound: () => lowerBound,
+  matMul: () => matMul,
+  max: () => max,
+  maxPool: () => maxPool,
+  maxPool3d: () => maxPool3d,
+  maxPoolWithArgmax: () => maxPoolWithArgmax,
+  maximum: () => maximum,
+  mean: () => mean,
+  meshgrid: () => meshgrid,
+  min: () => min,
+  minimum: () => minimum,
+  mirrorPad: () => mirrorPad,
+  mod: () => mod,
+  moments: () => moments,
+  movingAverage: () => movingAverage,
+  mul: () => mul,
+  multiRNNCell: () => multiRNNCell,
+  multinomial: () => multinomial,
+  neg: () => neg,
+  norm: () => norm,
+  notEqual: () => notEqual,
+  oneHot: () => oneHot,
+  ones: () => ones2,
+  onesLike: () => onesLike,
+  op: () => op,
+  outerProduct: () => outerProduct,
+  pad: () => pad,
+  pad1d: () => pad1d,
+  pad2d: () => pad2d,
+  pad3d: () => pad3d,
+  pad4d: () => pad4d,
+  pool: () => pool,
+  pow: () => pow,
+  prelu: () => prelu,
+  print: () => print,
+  prod: () => prod,
+  raggedGather: () => raggedGather,
+  raggedRange: () => raggedRange,
+  raggedTensorToTensor: () => raggedTensorToTensor,
+  rand: () => rand,
+  randomGamma: () => randomGamma,
+  randomNormal: () => randomNormal,
+  randomStandardNormal: () => randomStandardNormal,
+  randomUniform: () => randomUniform,
+  range: () => range,
+  real: () => real,
+  reciprocal: () => reciprocal,
+  relu: () => relu,
+  relu6: () => relu6,
+  reshape: () => reshape,
+  reverse: () => reverse,
+  reverse1d: () => reverse1d,
+  reverse2d: () => reverse2d,
+  reverse3d: () => reverse3d,
+  reverse4d: () => reverse4d,
+  rfft: () => rfft,
+  round: () => round2,
+  rsqrt: () => rsqrt,
+  scalar: () => scalar,
+  scatterND: () => scatterND,
+  searchSorted: () => searchSorted,
+  selu: () => selu,
+  separableConv2d: () => separableConv2d,
+  setdiff1dAsync: () => setdiff1dAsync,
+  sigmoid: () => sigmoid,
+  sign: () => sign,
+  signal: () => signal,
+  sin: () => sin,
+  sinh: () => sinh,
+  slice: () => slice,
+  slice1d: () => slice1d,
+  slice2d: () => slice2d,
+  slice3d: () => slice3d,
+  slice4d: () => slice4d,
+  softmax: () => softmax,
+  softplus: () => softplus,
+  spaceToBatchND: () => spaceToBatchND,
+  sparse: () => sparse,
+  sparseToDense: () => sparseToDense,
+  spectral: () => spectral,
+  split: () => split,
+  sqrt: () => sqrt,
+  square: () => square,
+  squaredDifference: () => squaredDifference,
+  squeeze: () => squeeze,
+  stack: () => stack,
+  step: () => step,
+  stridedSlice: () => stridedSlice,
+  string: () => string,
+  sub: () => sub,
+  sum: () => sum2,
+  tan: () => tan,
+  tanh: () => tanh2,
+  tensor: () => tensor,
+  tensor1d: () => tensor1d,
+  tensor2d: () => tensor2d,
+  tensor3d: () => tensor3d,
+  tensor4d: () => tensor4d,
+  tensor5d: () => tensor5d,
+  tensor6d: () => tensor6d,
+  tile: () => tile,
+  topk: () => topk,
+  transpose: () => transpose,
+  truncatedNormal: () => truncatedNormal,
+  unique: () => unique,
+  unsortedSegmentSum: () => unsortedSegmentSum,
+  unstack: () => unstack,
+  upperBound: () => upperBound,
+  variable: () => variable,
+  where: () => where,
+  whereAsync: () => whereAsync,
+  zeros: () => zeros,
+  zerosLike: () => zerosLike
+});
+var executeOp = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "BiasAdd":
+    case "AddV2":
+    case "Add": {
+      return [ops.add(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "AddN": {
+      return [ops.addN(getParamValue("tensors", node, tensorMap, context))];
+    }
+    case "FloorMod":
+    case "Mod":
+      return [ops.mod(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    case "Mul":
+      return [ops.mul(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    case "RealDiv":
+    case "Div": {
+      return [ops.div(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "DivNoNan": {
+      return [ops.divNoNan(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "FloorDiv": {
+      return [ops.floorDiv(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Sub": {
+      return [ops.sub(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Minimum": {
+      return [ops.minimum(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Maximum": {
+      return [ops.maximum(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Pow": {
+      return [ops.pow(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "SquaredDifference": {
+      return [ops.squaredDifference(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp2 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Abs":
+    case "ComplexAbs":
+      return [ops.abs(getParamValue("x", node, tensorMap, context))];
+    case "Acos":
+      return [ops.acos(getParamValue("x", node, tensorMap, context))];
+    case "Acosh":
+      return [ops.acosh(getParamValue("x", node, tensorMap, context))];
+    case "Asin":
+      return [ops.asin(getParamValue("x", node, tensorMap, context))];
+    case "Asinh":
+      return [ops.asinh(getParamValue("x", node, tensorMap, context))];
+    case "Atan":
+      return [ops.atan(getParamValue("x", node, tensorMap, context))];
+    case "Atan2":
+      return [ops.atan2(getParamValue("x", node, tensorMap, context), getParamValue("y", node, tensorMap, context))];
+    case "Atanh":
+      return [ops.atanh(getParamValue("x", node, tensorMap, context))];
+    case "Ceil":
+      return [ops.ceil(getParamValue("x", node, tensorMap, context))];
+    case "Complex":
+      return [ops.complex(getParamValue("real", node, tensorMap, context), getParamValue("imag", node, tensorMap, context))];
+    case "Cos":
+      return [ops.cos(getParamValue("x", node, tensorMap, context))];
+    case "Cosh":
+      return [ops.cosh(getParamValue("x", node, tensorMap, context))];
+    case "Elu":
+      return [ops.elu(getParamValue("x", node, tensorMap, context))];
+    case "Erf":
+      return [ops.erf(getParamValue("x", node, tensorMap, context))];
+    case "Exp":
+      return [ops.exp(getParamValue("x", node, tensorMap, context))];
+    case "Expm1": {
+      return [ops.expm1(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Floor":
+      return [ops.floor(getParamValue("x", node, tensorMap, context))];
+    case "Log":
+      return [ops.log(getParamValue("x", node, tensorMap, context))];
+    case "Log1p": {
+      return [ops.log1p(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Imag":
+      return [ops.imag(getParamValue("x", node, tensorMap, context))];
+    case "Neg":
+      return [ops.neg(getParamValue("x", node, tensorMap, context))];
+    case "Reciprocal": {
+      return [ops.reciprocal(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Real":
+      return [ops.real(getParamValue("x", node, tensorMap, context))];
+    case "Relu":
+      return [ops.relu(getParamValue("x", node, tensorMap, context))];
+    case "Round": {
+      return [ops.round(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Selu":
+      return [ops.selu(getParamValue("x", node, tensorMap, context))];
+    case "Sigmoid":
+      return [ops.sigmoid(getParamValue("x", node, tensorMap, context))];
+    case "Sin":
+      return [ops.sin(getParamValue("x", node, tensorMap, context))];
+    case "Sign": {
+      return [ops.sign(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Sinh": {
+      return [ops.sinh(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Softplus": {
+      return [ops.softplus(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Sqrt": {
+      return [ops.sqrt(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Square": {
+      return [ops.square(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Tanh": {
+      return [ops.tanh(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Tan":
+      return [ops.tan(getParamValue("x", node, tensorMap, context))];
+    case "ClipByValue":
+      return [ops.clipByValue(getParamValue("x", node, tensorMap, context), getParamValue("clipValueMin", node, tensorMap, context), getParamValue("clipValueMax", node, tensorMap, context))];
+    case "Relu6":
+      return [ops.relu6(getParamValue("x", node, tensorMap, context))];
+    case "Rsqrt":
+      return [ops.rsqrt(getTensor(node.inputNames[0], tensorMap, context))];
+    case "Prod":
+      return [ops.prod(getParamValue("x", node, tensorMap, context), getParamValue("axes", node, tensorMap, context))];
+    case "LeakyRelu":
+      return [ops.leakyRelu(getParamValue("x", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context))];
+    case "Prelu":
+      return [ops.prelu(getParamValue("x", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context))];
+    case "IsNan":
+      return [ops.isNaN(getTensor(node.inputNames[0], tensorMap, context))];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+function assertShapesMatchAllowUndefinedSize(shapeA, shapeB, errorMessagePrefix = "") {
+  if (typeof shapeA === "number" || typeof shapeB === "number") {
+    return;
+  }
+  util_exports.assert(shapeA.length === shapeB.length, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+  for (let i = 0; i < shapeA.length; i++) {
+    const dim0 = shapeA[i];
+    const dim1 = shapeB[i];
+    util_exports.assert(dim0 < 0 || dim1 < 0 || dim0 === dim1, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+  }
+}
+function fullDefinedShape(elementShape) {
+  if (typeof elementShape === "number" || elementShape.some((dim) => dim < 0)) {
+    return false;
+  }
+  return true;
+}
+function inferElementShape(listElementShape, tensors, elementShape) {
+  let partialShape = mergeElementShape(listElementShape, elementShape);
+  const notfullDefinedShape = !fullDefinedShape(partialShape);
+  if (notfullDefinedShape && tensors.length === 0) {
+    throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${partialShape}`);
+  }
+  if (notfullDefinedShape) {
+    tensors.forEach((tensor2) => {
+      partialShape = mergeElementShape(tensor2.shape, partialShape);
+    });
+  }
+  if (!fullDefinedShape(partialShape)) {
+    throw new Error(`Non-fully-defined elementShape: ${partialShape}`);
+  }
+  return partialShape;
+}
+function mergeElementShape(elementShapeA, elementShapeB) {
+  if (typeof elementShapeA === "number") {
+    return elementShapeB;
+  }
+  if (typeof elementShapeB === "number") {
+    return elementShapeA;
+  }
+  if (elementShapeA.length !== elementShapeB.length) {
+    throw new Error(`Incompatible ranks during merge: ${elementShapeA} vs. ${elementShapeB}`);
+  }
+  const result = [];
+  for (let i = 0; i < elementShapeA.length; ++i) {
+    const dim0 = elementShapeA[i];
+    const dim1 = elementShapeB[i];
+    if (dim0 >= 0 && dim1 >= 0 && dim0 !== dim1) {
+      throw new Error(`Incompatible shape during merge: ${elementShapeA} vs. ${elementShapeB}`);
+    }
+    result[i] = dim0 >= 0 ? dim0 : dim1;
+  }
+  return result;
+}
+var TensorArray = class {
+  constructor(name, dtype, maxSize, elementShape, identicalElementShapes, dynamicSize, clearAfterRead) {
+    this.name = name;
+    this.dtype = dtype;
+    this.maxSize = maxSize;
+    this.elementShape = elementShape;
+    this.identicalElementShapes = identicalElementShapes;
+    this.dynamicSize = dynamicSize;
+    this.clearAfterRead = clearAfterRead;
+    this.tensors = [];
+    this.closed_ = false;
+    this.idTensor = scalar(0);
+    keep(this.idTensor);
+  }
+  get id() {
+    return this.idTensor.id;
+  }
+  get closed() {
+    return this.closed_;
+  }
+  clearAndClose(keepIds) {
+    this.tensors.forEach((tensor2) => {
+      if (keepIds == null || !keepIds.has(tensor2.tensor.id)) {
+        tensor2.tensor.dispose();
+      }
+    });
+    this.tensors = [];
+    this.closed_ = true;
+    this.idTensor.dispose();
+  }
+  size() {
+    return this.tensors.length;
+  }
+  read(index) {
+    if (this.closed_) {
+      throw new Error(`TensorArray ${this.name} has already been closed.`);
+    }
+    if (index < 0 || index >= this.size()) {
+      throw new Error(`Tried to read from index ${index}, but array size is: ${this.size()}`);
+    }
+    const tensorWithState = this.tensors[index];
+    if (tensorWithState.cleared) {
+      throw new Error(`TensorArray ${this.name}: Could not read index ${index} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);
+    }
+    if (this.clearAfterRead) {
+      tensorWithState.cleared = true;
+    }
+    tensorWithState.read = true;
+    return tensorWithState.tensor;
+  }
+  readMany(indices) {
+    return indices.map((index) => this.read(index));
+  }
+  write(index, tensor2) {
+    if (this.closed_) {
+      throw new Error(`TensorArray ${this.name} has already been closed.`);
+    }
+    if (index < 0 || !this.dynamicSize && index >= this.maxSize) {
+      throw new Error(`Tried to write to index ${index}, but array is not resizeable and size is: ${this.maxSize}`);
+    }
+    const t = this.tensors[index] || {};
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index},
+          because the value dtype is ${tensor2.dtype}, but TensorArray dtype is ${this.dtype}.`);
+    }
+    if (this.size() === 0 && (this.elementShape == null || this.elementShape.length === 0)) {
+      this.elementShape = tensor2.shape;
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, `TensorArray ${this.name}: Could not write to TensorArray index ${index}.`);
+    if (t.read) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been read.`);
+    }
+    if (t.written) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been written.`);
+    }
+    t.tensor = tensor2;
+    keep(tensor2);
+    t.written = true;
+    this.tensors[index] = t;
+  }
+  writeMany(indices, tensors) {
+    if (indices.length !== tensors.length) {
+      throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${indices.length} is not the same as tensors size: ${tensors.length}.`);
+    }
+    indices.forEach((i, index) => this.write(i, tensors[index]));
+  }
+  gather(indices, dtype) {
+    if (!!dtype && dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${dtype}`);
+    }
+    if (!indices) {
+      indices = [];
+      for (let i = 0; i < this.size(); i++) {
+        indices.push(i);
+      }
+    } else {
+      indices = indices.slice(0, this.size());
+    }
+    if (indices.length === 0) {
+      return tensor([], [0].concat(this.elementShape));
+    }
+    const tensors = this.readMany(indices);
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, "TensorArray shape mismatch: ");
+    return stack(tensors, 0);
+  }
+  concat(dtype) {
+    if (!!dtype && dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${dtype}`);
+    }
+    if (this.size() === 0) {
+      return tensor([], [0].concat(this.elementShape));
+    }
+    const indices = [];
+    for (let i = 0; i < this.size(); i++) {
+      indices.push(i);
+    }
+    const tensors = this.readMany(indices);
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, `TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${tensors[0].shape})`);
+    return concat(tensors, 0);
+  }
+  scatter(indices, tensor2) {
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);
+    }
+    if (indices.length !== tensor2.shape[0]) {
+      throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);
+    }
+    const maxIndex = Math.max(...indices);
+    if (!this.dynamicSize && maxIndex >= this.maxSize) {
+      throw new Error(`Max index must be < array size (${maxIndex}  vs. ${this.maxSize})`);
+    }
+    this.writeMany(indices, unstack(tensor2, 0));
+  }
+  split(length, tensor2) {
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);
+    }
+    let totalLength = 0;
+    const cumulativeLengths = length.map((len) => {
+      totalLength += len;
+      return totalLength;
+    });
+    if (totalLength !== tensor2.shape[0]) {
+      throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${n}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);let r=n===0?0:t.size/n,s=[];P(()=>{t=W(t,[1,n,r]);for(let o=0;o<e.length;++o){let l=o===0?0:a[o-1],u=[0,l,0],p=[1,e[o],r];s[o]=W(Be(t,u,p),this.elementShape)}return s});let i=[];for(let o=0;o<e.length;o++)i[o]=o;this.writeMany(i,s)}},ml=class{constructor(e,t,n,a=-1){this.tensors=e,this.elementShape=t,this.elementDtype=n,e!=null&&e.forEach(r=>{if(n!==r.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${r.dtype}`);Ca(t,r.shape,"TensorList shape mismatch: "),Jt(r)}),this.idTensor=be(0),this.maxNumElements=a,Jt(this.idTensor)}get id(){return this.idTensor.id}copy(){return new ml([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.id))&&t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,n=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(n!==-1&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);Ca(e,this.elementShape,"TensorList shape mismatch: ");let a=Tp(this.elementShape,this.tensors,e);return P(()=>{let r=this.tensors.map(s=>W(s,a));return Ft(r,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error("Trying to pop from an empty list.");let n=Tp(this.elementShape,this.tensors,e),a=this.tensors.pop();return a.kept=!1,Ca(a.shape,e,"TensorList shape mismatch: "),W(a,n)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(Ca(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");Jt(e),this.tensors.push(e)}resize(e){if(e<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${e}`);if(this.maxNumElements!==-1&&e>this.maxNumElements)throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);let t=new ml([],this.elementShape,this.elementDtype,this.maxNumElements);t.tensors.length=e;for(let n=0;n<Math.min(this.tensors.length,e);++n)t.tensors[n]=this.tensors[n];return t}getItem(e,t,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(this.tensors[e]==null)throw new Error(`element at index ${e} is null.`);Ca(this.tensors[e].shape,t,"TensorList shape mismatch: ");let a=Tp(this.elementShape,this.tensors,t);return W(this.tensors[e],a)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||this.maxNumElements!==-1&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);Ca(this.elementShape,t.shape,"TensorList shape mismatch: "),Jt(t),this.tensors[e]!=null&&(this.tensors[e].kept=!1),this.tensors[e]=t}gather(e,t,n){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);Ca(this.elementShape,n,"TensorList shape mismatch: "),e=e.slice(0,this.size());let a=Tp(this.elementShape,this.tensors,n);return e.length===0?kn([],[0].concat(a)):P(()=>{let r=e.map(s=>W(this.tensors[s],a));return Ft(r,0)})}concat(e,t){if(!!e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);Ca(this.elementShape,t,"TensorList shape mismatch: ");let n=Tp(this.elementShape,this.tensors,t);return this.size()===0?kn([],[0].concat(n)):P(()=>{let a=this.tensors.map(r=>W(r,n));return Ze(a,0)})}};function qH(e,t,n){let a=e.dtype;if(e.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${e.shape}`);if(e.dtype!==n)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);let r=e.shape.slice(1);Ca(r,t,"TensorList shape mismatch: ");let s=ct(e);return new ml(s,t,a)}function KH(e,t,n,a){return new ml([],e,t,a)}function XH(e,t,n,a){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let r=Math.max(...t);if(a!=null&&a!==-1&&r>=a)throw new Error(`Max index must be < array size (${r}  vs. ${a})`);let s=new ml([],n,e.dtype,a),i=ct(e,0);return t.forEach((o,l)=>{s.setItem(o,i[l])}),s}function YH(e,t,n){let a=0,r=t.map(p=>(a+=p,a));if(a!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to
+        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);
+    }
+    if (!this.dynamicSize && length.length !== this.maxSize) {
+      throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${length.length}), and the TensorArray is not marked as dynamically resizeable`);
+    }
+    const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;
+    const tensors = [];
+    tidy(() => {
+      tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);
+      for (let i = 0; i < length.length; ++i) {
+        const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];
+        const indices2 = [0, previousLength, 0];
+        const sizes = [1, length[i], elementPerRow];
+        tensors[i] = reshape(slice(tensor2, indices2, sizes), this.elementShape);
+      }
+      return tensors;
+    });
+    const indices = [];
+    for (let i = 0; i < length.length; i++) {
+      indices[i] = i;
+    }
+    this.writeMany(indices, tensors);
+  }
+};
+var TensorList = class {
+  constructor(tensors, elementShape, elementDtype, maxNumElements = -1) {
+    this.tensors = tensors;
+    this.elementShape = elementShape;
+    this.elementDtype = elementDtype;
+    if (tensors != null) {
+      tensors.forEach((tensor2) => {
+        if (elementDtype !== tensor2.dtype) {
+          throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${tensor2.dtype}`);
+        }
+        assertShapesMatchAllowUndefinedSize(elementShape, tensor2.shape, "TensorList shape mismatch: ");
+        keep(tensor2);
+      });
+    }
+    this.idTensor = scalar(0);
+    this.maxNumElements = maxNumElements;
+    keep(this.idTensor);
+  }
+  get id() {
+    return this.idTensor.id;
+  }
+  copy() {
+    return new TensorList([...this.tensors], this.elementShape, this.elementDtype);
+  }
+  clearAndClose(keepIds) {
+    this.tensors.forEach((tensor2) => {
+      if (keepIds == null || !keepIds.has(tensor2.id)) {
+        tensor2.dispose();
+      }
+    });
+    this.tensors.length = 0;
+    this.idTensor.dispose();
+  }
+  size() {
+    return this.tensors.length;
+  }
+  stack(elementShape, elementDtype, numElements = -1) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (numElements !== -1 && this.tensors.length !== numElements) {
+      throw new Error(`Operation expected a list with ${numElements} elements but got a list with ${this.tensors.length} elements.`);
+    }
+    assertShapesMatchAllowUndefinedSize(elementShape, this.elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    return tidy(() => {
+      const reshapedTensors = this.tensors.map((tensor2) => reshape(tensor2, outputElementShape));
+      return stack(reshapedTensors, 0);
+    });
+  }
+  popBack(elementShape, elementDtype) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (this.size() === 0) {
+      throw new Error("Trying to pop from an empty list.");
+    }
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    const tensor2 = this.tensors.pop();
+    tensor2.kept = false;
+    assertShapesMatchAllowUndefinedSize(tensor2.shape, elementShape, "TensorList shape mismatch: ");
+    return reshape(tensor2, outputElementShape);
+  }
+  pushBack(tensor2) {
+    if (tensor2.dtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(tensor2.shape, this.elementShape, "TensorList shape mismatch: ");
+    if (this.maxNumElements === this.size()) {
+      throw new Error(`Trying to push element into a full list.`);
+    }
+    keep(tensor2);
+    this.tensors.push(tensor2);
+  }
+  resize(size) {
+    if (size < 0) {
+      throw new Error(`TensorListResize expects size to be non-negative. Got: ${size}`);
+    }
+    if (this.maxNumElements !== -1 && size > this.maxNumElements) {
+      throw new Error(`TensorListResize input size ${size} is greater maxNumElement ${this.maxNumElements}.`);
+    }
+    const destTensorList = new TensorList([], this.elementShape, this.elementDtype, this.maxNumElements);
+    destTensorList.tensors.length = size;
+    for (let i = 0; i < Math.min(this.tensors.length, size); ++i) {
+      destTensorList.tensors[i] = this.tensors[i];
+    }
+    return destTensorList;
+  }
+  getItem(elementIndex, elementShape, elementDtype) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (elementIndex < 0 || elementIndex > this.tensors.length) {
+      throw new Error(`Trying to access element ${elementIndex} in a list with ${this.tensors.length} elements.`);
+    }
+    if (this.tensors[elementIndex] == null) {
+      throw new Error(`element at index ${elementIndex} is null.`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.tensors[elementIndex].shape, elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    return reshape(this.tensors[elementIndex], outputElementShape);
+  }
+  setItem(elementIndex, tensor2) {
+    if (tensor2.dtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);
+    }
+    if (elementIndex < 0 || this.maxNumElements !== -1 && elementIndex >= this.maxNumElements) {
+      throw new Error(`Trying to set element ${elementIndex} in a list with max ${this.maxNumElements} elements.`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, "TensorList shape mismatch: ");
+    keep(tensor2);
+    if (this.tensors[elementIndex] != null) {
+      this.tensors[elementIndex].kept = false;
+    }
+    this.tensors[elementIndex] = tensor2;
+  }
+  gather(indices, elementDtype, elementShape) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, "TensorList shape mismatch: ");
+    indices = indices.slice(0, this.size());
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    if (indices.length === 0) {
+      return tensor([], [0].concat(outputElementShape));
+    }
+    return tidy(() => {
+      const tensors = indices.map((i) => reshape(this.tensors[i], outputElementShape));
+      return stack(tensors, 0);
+    });
+  }
+  concat(elementDtype, elementShape) {
+    if (!!elementDtype && elementDtype !== this.elementDtype) {
+      throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    if (this.size() === 0) {
+      return tensor([], [0].concat(outputElementShape));
+    }
+    return tidy(() => {
+      const tensors = this.tensors.map((t) => reshape(t, outputElementShape));
+      return concat(tensors, 0);
+    });
+  }
+};
+function fromTensor(tensor2, elementShape, elementDtype) {
+  const dtype = tensor2.dtype;
+  if (tensor2.shape.length < 1) {
+    throw new Error(`Tensor must be at least a vector, but saw shape: ${tensor2.shape}`);
+  }
+  if (tensor2.dtype !== elementDtype) {
+    throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${elementDtype}`);
+  }
+  const tensorElementShape = tensor2.shape.slice(1);
+  assertShapesMatchAllowUndefinedSize(tensorElementShape, elementShape, "TensorList shape mismatch: ");
+  const tensorList = unstack(tensor2);
+  return new TensorList(tensorList, elementShape, dtype);
+}
+function reserve(elementShape, elementDtype, numElements, maxNumElements) {
+  return new TensorList([], elementShape, elementDtype, maxNumElements);
+}
+function scatter(tensor2, indices, elementShape, numElements) {
+  if (indices.length !== tensor2.shape[0]) {
+    throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);
+  }
+  const maxIndex = Math.max(...indices);
+  if (numElements != null && numElements !== -1 && maxIndex >= numElements) {
+    throw new Error(`Max index must be < array size (${maxIndex}  vs. ${numElements})`);
+  }
+  const list = new TensorList([], elementShape, tensor2.dtype, numElements);
+  const tensors = unstack(tensor2, 0);
+  indices.forEach((value, index) => {
+    list.setItem(value, tensors[index]);
+  });
+  return list;
+}
+function split2(tensor2, length, elementShape) {
+  let totalLength = 0;
+  const cumulativeLengths = length.map((len) => {
+    totalLength += len;
+    return totalLength;
+  });
+  if (totalLength !== tensor2.shape[0]) {
+    throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${a}, and tensor's shape is: ${e.shape}`);let s=e.shape.slice(1),i=tx(s,n),o=a===0?0:e.size/a,l=P(()=>{let p=[];e=W(e,[1,a,o]);for(let d=0;d<t.length;++d){let c=d===0?0:r[d-1],h=[0,c,0],m=[1,t[d],o];p[d]=W(Be(e,h,m),i)}return e.dispose(),p}),u=new ml([],n,e.dtype,t.length);for(let p=0;p<l.length;p++)u.setItem(p,l[p]);return u}var ZH=async(e,t,n)=>{switch(e.op){case"If":case"StatelessIf":{let a=k("thenBranch",e,t,n),r=k("elseBranch",e,t,n),s=k("cond",e,t,n),i=k("args",e,t,n);return(await s.data())[0]?n.functionMap[a].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap):n.functionMap[r].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap)}case"While":case"StatelessWhile":{let a=k("body",e,t,n),r=k("cond",e,t,n),s=k("args",e,t,n),i=await n.functionMap[r].executeFunctionAsync(s,n.tensorArrayMap,n.tensorListMap),o=s.map(p=>p.id),l=await i[0].data();i.forEach(p=>{!p.kept&&o.indexOf(p.id)===-1&&p.dispose()});let u=s;for(;l[0];){let p=u;u=await n.functionMap[a].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);let d=u.map(h=>h.id);p.forEach(h=>{!h.kept&&o.indexOf(h.id)===-1&&d.indexOf(h.id)===-1&&h.dispose()});let c=await n.functionMap[r].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);l=await c[0].data(),c.forEach(h=>{!h.kept&&o.indexOf(h.id)===-1&&d.indexOf(h.id)===-1&&h.dispose()})}return u}case"LoopCond":{let a=k("pred",e,t,n);return[kr(a)]}case"Switch":{let a=k("pred",e,t,n),r=k("data",e,t,n);return r.kept||(r=kr(r)),(await a.data())[0]?[void 0,r]:[r,void 0]}case"Merge":{let a=e.inputNames.find(r=>vn(r,t,n)!==void 0);if(a){let r=vn(a,t,n);return[kr(r)]}return}case"Enter":{let a=k("frameName",e,t,n),r=k("tensor",e,t,n);return n.enterFrame(a),[kr(r)]}case"Exit":{let a=k("tensor",e,t,n);return n.exitFrame(),[kr(a)]}case"NextIteration":{let a=k("tensor",e,t,n);return n.nextIteration(),[kr(a)]}case"TensorArrayV3":{let a=k("size",e,t,n),r=k("dtype",e,t,n),s=k("elementShape",e,t,n),i=k("dynamicSize",e,t,n),o=k("clearAfterRead",e,t,n),l=k("identicalElementShapes",e,t,n),u=k("name",e,t,n),p=new jH(u,r,a,s,l,i,o);return n.addTensorArray(p),[p.idTensor,be(1)]}case"TensorArrayWriteV3":{let a=k("tensorArrayId",e,t,n),r=k("index",e,t,n),s=k("tensor",e,t,n),i=n.getTensorArray(a.id);return i.write(r,s),[i.idTensor]}case"TensorArrayReadV3":{let a=k("tensorArrayId",e,t,n),r=k("index",e,t,n);return[n.getTensorArray(a.id).read(r)]}case"TensorArrayGatherV3":{let a=k("tensorArrayId",e,t,n),r=k("indices",e,t,n),s=k("dtype",e,t,n);return[n.getTensorArray(a.id).gather(r,s)]}case"TensorArrayScatterV3":{let a=k("tensorArrayId",e,t,n),r=k("indices",e,t,n),s=k("tensor",e,t,n),i=n.getTensorArray(a.id);return i.scatter(r,s),[i.idTensor]}case"TensorArrayConcatV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id),s=k("dtype",e,t,n);return[r.concat(s)]}case"TensorArraySplitV3":{let a=k("tensorArrayId",e,t,n),r=k("tensor",e,t,n),s=k("lengths",e,t,n),i=n.getTensorArray(a.id);return i.split(s,r),[i.idTensor]}case"TensorArraySizeV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id);return[be(r.size(),"int32")]}case"TensorArrayCloseV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id);return r.clearAndClose(),[r.idTensor]}case"TensorListSetItem":{let a=k("tensorListId",e,t,n),r=k("index",e,t,n),s=k("tensor",e,t,n),i=n.getTensorList(a.id);return i.setItem(r,s),[i.idTensor]}case"TensorListGetItem":{let a=k("tensorListId",e,t,n),r=k("index",e,t,n),s=k("elementShape",e,t,n),i=k("elementDType",e,t,n);return[n.getTensorList(a.id).getItem(r,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{let a=k("indices",e,t,n),r=k("tensor",e,t,n),s=k("elementShape",e,t,n),i=k("numElements",e,t,n),o=XH(r,a,s,i);return n.addTensorList(o),[o.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let a=k("elementShape",e,t,n),r=k("elementDType",e,t,n),s;e.op==="TensorListReserve"?s="numElements":s="maxNumElements";let i=k(s,e,t,n),o=e.op==="TensorListReserve"?-1:i,l=KH(a,r,i,o);return n.addTensorList(l),[l.idTensor]}case"TensorListGather":{let a=k("tensorListId",e,t,n),r=k("indices",e,t,n),s=k("elementShape",e,t,n),i=k("elementDType",e,t,n);return[n.getTensorList(a.id).gather(r,i,s)]}case"TensorListStack":{let a=k("tensorListId",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n),i=k("numElements",e,t,n);return[n.getTensorList(a.id).stack(r,s,i)]}case"TensorListFromTensor":{let a=k("tensor",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n),i=qH(a,r,s);return n.addTensorList(i),[i.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let a=k("tensorListId",e,t,n),r=n.getTensorList(a.id),s=k("dtype",e,t,n),i=k("elementShape",e,t,n);return[r.concat(s,i)]}case"TensorListPushBack":{let a=k("tensorListId",e,t,n),r=k("tensor",e,t,n),s=n.getTensorList(a.id);return s.pushBack(r),[s.idTensor]}case"TensorListPopBack":{let a=k("tensorListId",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n);return[n.getTensorList(a.id).popBack(r,s)]}case"TensorListSplit":{let a=k("tensor",e,t,n),r=k("elementShape",e,t,n),s=k("lengths",e,t,n),i=YH(a,s,r);return n.addTensorList(i),[i.idTensor]}case"TensorListLength":{let a=k("tensorListId",e,t,n),r=n.getTensorList(a.id);return[be(r.size(),"int32")]}case"TensorListResize":{let a=k("tensorListId",e,t,n),r=k("size",e,t,n),s=n.getTensorList(a.id).resize(r);return n.addTensorList(s),[s.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Fk(e,t,n){let[a,r]=k("fusedOps",e,t,n),s=a==="biasadd",i=!s,o=r==="prelu",l=a==="fusedbatchnorm",u=k("numArgs",e,t,n);if(s){if(o&&u!==2)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&s&&u!==1)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(l)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let p=k("strides",e,t,n),d=sh(e,t,n),c=k("dataFormat",e,t,n).toUpperCase(),h=k("dilations",e,t,n),[m,f]=k("args",e,t,n);i&&(f=m,m=void 0);let g=k("leakyreluAlpha",e,t,n);return{stride:p,pad:d,dataFormat:c,dilations:h,biasArg:m,preluArg:f,activationFunc:r,leakyreluAlpha:g}}var JH=(e,t,n,a=un)=>{switch(e.op){case"Conv1D":{let r=k("stride",e,t,n),s=k("pad",e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilation",e,t,n);return[a.conv1d(k("x",e,t,n),k("filter",e,t,n),r,s,i,o)]}case"Conv2D":{let r=k("strides",e,t,n),s=sh(e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilations",e,t,n);return[a.conv2d(k("x",e,t,n),k("filter",e,t,n),[r[1],r[2]],s,i,[o[1],o[2]])]}case"_FusedConv2D":{let{stride:r,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:p,leakyreluAlpha:d}=Fk(e,t,n);return[a.fused.conv2d({x:k("x",e,t,n),filter:k("filter",e,t,n),strides:[r[1],r[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:p,preluActivationWeights:u,leakyreluAlpha:d})]}case"FusedDepthwiseConv2dNative":{let{stride:r,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:p,leakyreluAlpha:d}=Fk(e,t,n);return[a.fused.depthwiseConv2d({x:k("x",e,t,n),filter:k("filter",e,t,n),strides:[r[1],r[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:p,preluActivationWeights:u,leakyreluAlpha:d})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let r=k("outputShape",e,t,n),s=k("strides",e,t,n),i=sh(e,t,n);return[a.conv2dTranspose(k("x",e,t,n),k("filter",e,t,n),r,[s[1],s[2]],i)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let r=k("strides",e,t,n),s=sh(e,t,n),i=k("dilations",e,t,n),o=k("dataFormat",e,t,n).toUpperCase();return[a.depthwiseConv2d(k("input",e,t,n),k("filter",e,t,n),[r[1],r[2]],s,o,[i[1],i[2]])]}case"Conv3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilations",e,t,n);return[a.conv3d(k("x",e,t,n),k("filter",e,t,n),[r[1],r[2],r[3]],s,i,[o[1],o[2],o[3]])]}case"AvgPool":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.avgPool(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s)]}case"MaxPool":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.maxPool(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s)]}case"MaxPoolWithArgmax":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n),o=k("includeBatchInIndex",e,t,n),{result:l,indexes:u}=a.maxPoolWithArgmax(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s,o);return[l,u]}case"AvgPool3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.avgPool3d(k("x",e,t,n),[i[1],i[2],i[3]],[r[1],r[2],r[3]],s)]}case"MaxPool3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.maxPool3d(k("x",e,t,n),[i[1],i[2],i[3]],[r[1],r[2],r[3]],s)]}case"Dilation2D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("dilations",e,t,n),o=r[1],l=r[2],u=i[1],p=i[2];return[a.dilation2d(k("x",e,t,n),k("filter",e,t,n),[o,l],s,[u,p],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},QH=(e,t,n,a=un)=>{switch(e.op){case"Fill":{let r=k("shape",e,t,n),s=k("dtype",e,t,n),i=k("value",e,t,n);return[a.fill(r,i,s)]}case"LinSpace":{let r=k("start",e,t,n),s=k("stop",e,t,n),i=k("num",e,t,n);return[a.linspace(r,s,i)]}case"Multinomial":{let r=k("logits",e,t,n),s=k("numSamples",e,t,n),i=k("seed",e,t,n);return[a.multinomial(r,s,i)]}case"OneHot":{let r=k("indices",e,t,n),s=k("depth",e,t,n),i=k("onValue",e,t,n),o=k("offValue",e,t,n),l=k("dtype",e,t,n);return[a.oneHot(r,s,i,o,l)]}case"Ones":return[a.ones(k("shape",e,t,n),k("dtype",e,t,n))];case"OnesLike":return[a.onesLike(k("x",e,t,n))];case"RandomStandardNormal":return[a.randomStandardNormal(k("shape",e,t,n),k("dtype",e,t,n),k("seed",e,t,n))];case"RandomUniform":return[a.randomUniform(k("shape",e,t,n),k("minval",e,t,n),k("maxval",e,t,n),k("dtype",e,t,n))];case"Range":{let r=k("start",e,t,n),s=k("stop",e,t,n),i=k("step",e,t,n);return[a.range(r,s,i,k("dtype",e,t,n))]}case"TruncatedNormal":{let r=k("shape",e,t,n),s=k("mean",e,t,n),i=k("stdDev",e,t,n),o=k("seed",e,t,n);return[a.truncatedNormal(r,s,i,k("dtype",e,t,n),o)]}case"Zeros":return[a.zeros(k("shape",e,t,n),k("dtype",e,t,n))];case"ZerosLike":return[a.zerosLike(k("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function mb(e,t,n){let a=k("boxes",e,t,n),r=k("scores",e,t,n),s=k("maxOutputSize",e,t,n),i=k("iouThreshold",e,t,n),o=k("scoreThreshold",e,t,n),l=k("softNmsSigma",e,t,n);return{boxes:a,scores:r,maxOutputSize:s,iouThreshold:i,scoreThreshold:o,softNmsSigma:l}}var e6=async(e,t,n,a,r=un)=>{switch(e.op){case"NonMaxSuppressionV5":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u,softNmsSigma:p}=mb(e,t,n),d=await r.image.nonMaxSuppressionWithScoreAsync(s,i,o,l,u,p);return[d.selectedIndices,d.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u}=mb(e,t,n),p=k("padToMaxOutputSize",e,t,n),d=await r.image.nonMaxSuppressionPaddedAsync(s,i,o,l,u,p);return[d.selectedIndices,d.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u}=mb(e,t,n);return[await r.image.nonMaxSuppressionAsync(s,i,o,l,u)]}case"Where":{let s=r.cast(k("condition",e,t,n),"bool"),i=[await r.whereAsync(s)];return s.dispose(),i}case"ListDiff":return r.setdiff1dAsync(k("x",e,t,n),k("y",e,t,n));default:throw TypeError(`Node type ${e.op} is not implemented`)}},t6=(e,t,n,a=un)=>{switch(e.op){case"LowerBound":{let r=k("sortedSequence",e,t,n),s=k("values",e,t,n);return[a.lowerBound(r,s)]}case"TopKV2":{let r=k("x",e,t,n),s=k("k",e,t,n),i=k("sorted",e,t,n),o=a.topk(r,s,i);return[o.values,o.indices]}case"UpperBound":{let r=k("sortedSequence",e,t,n),s=k("values",e,t,n);return[a.upperBound(r,s)]}case"Unique":{let r=k("x",e,t,n),s=a.unique(r);return[s.values,s.indices]}case"UniqueV2":{let r=k("x",e,t,n),s=k("axis",e,t,n),i=a.unique(r,s);return[i.values,i.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},n6=(e,t,n,a=un)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":let r=k("default",e,t,n);return[vn(e.name,t,n)||r];case"Placeholder":return[vn(e.name,t,n)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":{let p=k("x",e,t,n);return[kr(p)]}case"IdentityN":return k("x",e,t,n).map(p=>kr(p));case"Snapshot":let s=k("x",e,t,n);return[kr(s)];case"Shape":return[a.tensor1d(k("x",e,t,n).shape,"int32")];case"ShapeN":return k("x",e,t,n).map(p=>a.tensor1d(p.shape));case"Size":return[a.scalar(k("x",e,t,n).size,"int32")];case"Rank":return[a.scalar(k("x",e,t,n).rank,"int32")];case"NoOp":return[a.scalar(1)];case"Print":let i=k("x",e,t,n),o=k("data",e,t,n),l=k("message",e,t,n),u=k("summarize",e,t,n);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(l);for(let p=0;p<o.length;p++)console.log(Array.prototype.slice.call(o[p].dataSync()).slice(0,u));return[i];default:throw TypeError(`Node type ${e.op} is not implemented`)}},a6=class{constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=be(0),this.tensorMap=new Map,Jt(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return be(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return this.tensorMap.forEach(a=>a.dispose()),this.tensorMap.clear(),P(()=>{let a=ct(t),r=n.length,s=a.length;v.assert(r===s,()=>`The number of elements doesn't match, keys has ${r} elements, the values has ${s} elements.`);for(let i=0;i<r;i++){let o=n[i],l=a[i];Jt(l),this.tensorMap.set(o,l)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return P(()=>{let a=[];for(let r=0;r<n.length;r++){let s=n[r],i=this.findWithDefault(s,t);a.push(i)}return Ft(a)})}findWithDefault(e,t){let n=this.tensorMap.get(e);return n!=null?n:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}},r6=async(e,t,n,a)=>{switch(e.op){case"HashTable":case"HashTableV2":{let r=a.getHashTableHandleByName(e.name);if(r!=null)return[r];{let s=k("keyDType",e,t,n),i=k("valueDType",e,t,n),o=new a6(s,i);return a.addHashTable(e.name,o),[o.handle]}}case"LookupTableImport":case"LookupTableImportV2":{let r=k("tableHandle",e,t,n,a),s=k("keys",e,t,n),i=k("values",e,t,n);return[await a.getHashTableById(r.id).import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{let r=k("tableHandle",e,t,n,a),s=k("keys",e,t,n),i=k("defaultValue",e,t,n);return[await a.getHashTableById(r.id).find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{let r=k("tableHandle",e,t,n,a);return[a.getHashTableById(r.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},s6=(e,t,n,a=un)=>{switch(e.op){case"ResizeBilinear":{let r=k("images",e,t,n),s=k("size",e,t,n),i=k("alignCorners",e,t,n),o=k("halfPixelCenters",e,t,n);return[a.image.resizeBilinear(r,[s[0],s[1]],i,o)]}case"ResizeNearestNeighbor":{let r=k("images",e,t,n),s=k("size",e,t,n),i=k("alignCorners",e,t,n),o=k("halfPixelCenters",e,t,n);return[a.image.resizeNearestNeighbor(r,[s[0],s[1]],i,o)]}case"CropAndResize":{let r=k("image",e,t,n),s=k("boxes",e,t,n),i=k("boxInd",e,t,n),o=k("cropSize",e,t,n),l=k("method",e,t,n),u=k("extrapolationValue",e,t,n);return[a.image.cropAndResize(r,s,i,o,l,u)]}case"ImageProjectiveTransformV3":{let r=k("images",e,t,n),s=k("transforms",e,t,n),i=k("outputShape",e,t,n),o=k("fillValue",e,t,n),l=k("interpolation",e,t,n),u=k("fillMode",e,t,n);return[a.image.transform(r,s,l.toLowerCase(),u.toLowerCase(),o,i)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},i6=(e,t,n,a=un)=>{switch(e.op){case"Equal":return[a.equal(k("a",e,t,n),k("b",e,t,n))];case"NotEqual":return[a.notEqual(k("a",e,t,n),k("b",e,t,n))];case"Greater":return[a.greater(k("a",e,t,n),k("b",e,t,n))];case"GreaterEqual":return[a.greaterEqual(k("a",e,t,n),k("b",e,t,n))];case"Less":return[a.less(k("a",e,t,n),k("b",e,t,n))];case"LessEqual":return[a.lessEqual(k("a",e,t,n),k("b",e,t,n))];case"LogicalAnd":return[a.logicalAnd(k("a",e,t,n),k("b",e,t,n))];case"LogicalNot":return[a.logicalNot(k("a",e,t,n))];case"LogicalOr":return[a.logicalOr(k("a",e,t,n),k("b",e,t,n))];case"Select":case"SelectV2":return[a.where(k("condition",e,t,n),k("a",e,t,n),k("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},o6=(e,t,n,a=un)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[a.matMul(k("a",e,t,n),k("b",e,t,n),k("transposeA",e,t,n),k("transposeB",e,t,n))];case"Einsum":return[a.einsum(k("equation",e,t,n),...k("tensors",e,t,n))];case"Transpose":return[a.transpose(k("x",e,t,n),k("perm",e,t,n))];case"_FusedMatMul":let[r,s]=k("fusedOps",e,t,n),i=r==="biasadd",o=s==="prelu",l=k("numArgs",e,t,n),u=k("leakyreluAlpha",e,t,n);if(i){if(o&&l!==2)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&l!==1)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[p,d]=k("args",e,t,n);return[a.fused.matMul({a:k("a",e,t,n),b:k("b",e,t,n),transposeA:k("transposeA",e,t,n),transposeB:k("transposeB",e,t,n),bias:p,activation:s,preluActivationWeights:d,leakyreluAlpha:u})];default:throw TypeError(`Node type ${e.op} is not implemented`)}},l6=(e,t,n,a=un)=>{switch(e.op){case"EuclideanNorm":return[a.euclideanNorm(k("x",e,t,n),k("axis",e,t,n),k("keepDims",e,t,n))];case"FusedBatchNorm":case"FusedBatchNormV2":return[a.batchNorm(k("x",e,t,n),k("mean",e,t,n),k("variance",e,t,n),k("offset",e,t,n),k("scale",e,t,n),k("epsilon",e,t,n))];case"FusedBatchNormV3":return[a.batchNorm(k("x",e,t,n),k("mean",e,t,n),k("variance",e,t,n),k("offset",e,t,n),k("scale",e,t,n),k("epsilon",e,t,n))];case"LRN":return[a.localResponseNormalization(k("x",e,t,n),k("radius",e,t,n),k("bias",e,t,n),k("alpha",e,t,n),k("beta",e,t,n))];case"Softmax":return[a.softmax(k("x",e,t,n))];case"LogSoftmax":return[a.logSoftmax(k("x",e,t,n))];case"SparseToDense":return[a.sparseToDense(k("sparseIndices",e,t,n),k("outputShape",e,t,n),k("sparseValues",e,t,n),k("defaultValue",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},u6=(e,t,n,a=un)=>{switch(e.op){case"Max":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.max(k("x",e,t,n),o,l)]}case"Mean":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.mean(k("x",e,t,n),o,l)]}case"Min":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.min(k("x",e,t,n),o,l)]}case"Sum":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.sum(k("x",e,t,n),o,l)]}case"All":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.all(k("x",e,t,n),o,l)]}case"Any":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.any(k("x",e,t,n),o,l)]}case"ArgMax":{let o=k("axis",e,t,n);return[a.argMax(k("x",e,t,n),o)]}case"ArgMin":{let o=k("axis",e,t,n);return[a.argMin(k("x",e,t,n),o)]}case"Prod":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.prod(k("x",e,t,n),o,l)]}case"Cumprod":{let o=k("axis",e,t,n),l=k("exclusive",e,t,n),u=k("reverse",e,t,n);return[a.cumprod(k("x",e,t,n),o,l,u)]}case"Cumsum":{let o=k("axis",e,t,n),l=k("exclusive",e,t,n),u=k("reverse",e,t,n);return[a.cumsum(k("x",e,t,n),o,l,u)]}case"Bincount":let r=k("x",e,t,n),s=k("weights",e,t,n),i=k("size",e,t,n);return[a.bincount(r,s,i)];case"DenseBincount":{let o=k("x",e,t,n),l=k("weights",e,t,n),u=k("size",e,t,n),p=k("binaryOutput",e,t,n);return[a.denseBincount(o,l,u,p)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},p6=(e,t,n,a=un)=>{switch(e.op){case"ConcatV2":case"Concat":{let r=k("n",e,t,n),s=k("axis",e,t,n),i=k("tensors",e,t,n);return i=i.slice(0,r),[a.concat(i,s)]}case"Gather":{let r=k("x",e,t,n),s=k("indices",e,t,n);return[a.gather(r,a.cast(s,"int32"),0)]}case"GatherV2":{let r=k("axis",e,t,n),s=k("batchDims",e,t,n),i=k("x",e,t,n),o=k("indices",e,t,n);return[a.gather(i,a.cast(o,"int32"),r,s)]}case"Reverse":{let r=k("dims",e,t,n),s=[];for(let o=0;o<r.length;o++)r[o]&&s.push(o);let i=k("x",e,t,n);return[a.reverse(i,s)]}case"ReverseV2":{let r=k("axis",e,t,n),s=k("x",e,t,n);return[a.reverse(s,r)]}case"Slice":{let r=k("begin",e,t,n),s=k("size",e,t,n);return[a.slice(k("x",e,t,n),r,s)]}case"StridedSlice":{let r=k("begin",e,t,n),s=k("end",e,t,n),i=k("strides",e,t,n),o=k("beginMask",e,t,n),l=k("endMask",e,t,n),u=k("ellipsisMask",e,t,n),p=k("newAxisMask",e,t,n),d=k("shrinkAxisMask",e,t,n),c=k("x",e,t,n);return[a.stridedSlice(c,r,s,i,o,l,u,p,d)]}case"Pack":return P(()=>{let r=k("axis",e,t,n),s=k("tensors",e,t,n),i=s[0].shape,o=a.squeeze(s[0]).shape,l=s.map(u=>{let p=v.arraysEqual(u.shape,i);if(!p&&!v.arraysEqual(a.squeeze(u).shape,o))throw new Error("the input tensors shape does not match");return p?u:a.reshape(u,i)});return[a.stack(l,r)]});case"Unpack":{let r=k("axis",e,t,n),s=k("tensor",e,t,n);return a.unstack(s,r)}case"Tile":{let r=k("reps",e,t,n);return[a.tile(k("x",e,t,n),r)]}case"Split":case"SplitV":{let r=k("axis",e,t,n),s=k("numOrSizeSplits",e,t,n),i=k("x",e,t,n);return a.split(i,s,r)}case"ScatterNd":{let r=k("indices",e,t,n),s=k("values",e,t,n),i=k("shape",e,t,n);return[a.scatterND(r,s,i)]}case"GatherNd":{let r=k("x",e,t,n),s=k("indices",e,t,n);return[a.gatherND(r,s)]}case"SparseToDense":{let r=k("sparseIndices",e,t,n),s=k("outputShape",e,t,n),i=k("sparseValues",e,t,n),o=k("defaultValue",e,t,n);return[a.sparseToDense(r,i,s,i.dtype===o.dtype?o:a.cast(o,i.dtype))]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},c6=(e,t,n,a=un)=>{switch(e.op){case"SparseFillEmptyRows":{let{outputIndices:r,outputValues:s,emptyRowIndicator:i,reverseIndexMap:o}=a.sparse.sparseFillEmptyRows(k("indices",e,t,n),k("values",e,t,n),k("denseShape",e,t,n),k("defaultValue",e,t,n));return[r,s,i,o]}case"SparseReshape":{let{outputIndices:r,outputShape:s}=a.sparse.sparseReshape(k("inputIndices",e,t,n),k("inputShape",e,t,n),k("newShape",e,t,n));return[r,s]}case"SparseSegmentMean":return[a.sparse.sparseSegmentMean(k("data",e,t,n),k("indices",e,t,n),k("segmentIds",e,t,n))];case"SparseSegmentSum":return[a.sparse.sparseSegmentSum(k("data",e,t,n),k("indices",e,t,n),k("segmentIds",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},d6=(e,t,n,a=un)=>{switch(e.op){case"FFT":return[a.fft(k("x",e,t,n))];case"IFFT":return[a.ifft(k("x",e,t,n))];case"RFFT":return[a.rfft(k("x",e,t,n))];case"IRFFT":return[a.irfft(k("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},h6=(e,t,n,a=un)=>{switch(e.op){case"StringNGrams":{let{nGrams:r,nGramsSplits:s}=a.string.stringNGrams(k("data",e,t,n),k("dataSplits",e,t,n),k("separator",e,t,n),k("nGramWidths",e,t,n),k("leftPad",e,t,n),k("rightPad",e,t,n),k("padWidth",e,t,n),k("preserveShortSequences",e,t,n));return[r,s]}case"StringSplit":{let{indices:r,values:s,shape:i}=a.string.stringSplit(k("input",e,t,n),k("delimiter",e,t,n),k("skipEmpty",e,t,n));return[r,s,i]}case"StringToHashBucketFast":return[a.string.stringToHashBucketFast(k("input",e,t,n),k("numBuckets",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},m6=(e,t,n,a=un)=>{switch(e.op){case"Cast":return[a.cast(k("x",e,t,n),k("dtype",e,t,n))];case"ExpandDims":{let r=k("axis",e,t,n);return[a.expandDims(k("x",e,t,n),r)]}case"Squeeze":{let r=k("axis",e,t,n);return[a.squeeze(k("x",e,t,n),r)]}case"Reshape":return[a.reshape(k("x",e,t,n),k("shape",e,t,n))];case"MirrorPad":return[a.mirrorPad(k("x",e,t,n),k("padding",e,t,n),k("mode",e,t,n))];case"PadV2":case"Pad":return[a.pad(k("x",e,t,n),k("padding",e,t,n),k("constantValue",e,t,n))];case"SpaceToBatchND":{let r=k("blockShape",e,t,n),s=k("paddings",e,t,n);return[a.spaceToBatchND(k("x",e,t,n),r,s)]}case"BatchToSpaceND":{let r=k("blockShape",e,t,n),s=k("crops",e,t,n);return[a.batchToSpaceND(k("x",e,t,n),r,s)]}case"DepthToSpace":{let r=k("blockSize",e,t,n),s=k("dataFormat",e,t,n).toUpperCase();return[a.depthToSpace(k("x",e,t,n),r,s)]}case"BroadcastTo":return[a.broadcastTo(k("x",e,t,n),k("shape",e,t,n))];case"BroadcastArgs":return[a.broadcastArgs(k("s0",e,t,n),k("s1",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Dk(e,t,n,a,r=P){let s=((i,o,l)=>{switch(i.category){case"arithmetic":return r(()=>GH(i,o,l));case"basic_math":return r(()=>HH(i,o,l));case"control":return ZH(i,o,l);case"convolution":return r(()=>JH(i,o,l));case"creation":return r(()=>QH(i,o,l));case"dynamic":return e6(i,o,l);case"evaluation":return r(()=>t6(i,o,l));case"image":return r(()=>s6(i,o,l));case"graph":return r(()=>n6(i,o,l));case"logical":return r(()=>i6(i,o,l));case"matrices":return r(()=>o6(i,o,l));case"normalization":return r(()=>l6(i,o,l));case"reduction":return r(()=>u6(i,o,l));case"slice_join":return r(()=>p6(i,o,l));case"sparse":return r(()=>c6(i,o,l));case"spectral":return r(()=>d6(i,o,l));case"string":return r(()=>h6(i,o,l));case"transformation":return r(()=>m6(i,o,l));case"hash_table":return r6(i,o,l,a);case"custom":let u=e2(i.op);if(u&&u.customExecutor)return u.customExecutor(new UH(i,o,l));throw TypeError(`Custom op ${i.op} is not registered.`);default:throw TypeError(`Unknown op '${i.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,n);return v.isPromise(s)?s.then(i=>[].concat(i)):[].concat(s)}var Rk=class{constructor(e={},t={},n={},a={}){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=n,this.functionMap=a,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let e=[];for(let t=0;t<this.contexts.length-1;t++){let n=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(n))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(t=>t.id===0&&t.iterationId===0?"":`${t.frameName}-${t.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error("Cannot exit frame, the context is empty")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error("Cannot increase frame iteration, the context is empty")}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(let t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(let t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function Mk(e,t,n,a){let r=new Set,s=[],i=null,o=null,l=new Set,u=Object.keys(e).map(c=>Xn(c)[0]),p=[];a!=null&&(p=a.map(c=>Xn(c.name)[0]));let d=[...t];for(;d.length>0;){let c=d.pop();if((k2(c)||x6(c)||v6(c))&&i==null&&(i=c,o=i.children.map(h=>h.name).filter(h=>r.has(h))),r.add(c.name),n[c.name]==null&&u.indexOf(c.name)===-1&&p.indexOf(c.name)===-1){if(c.inputs.length===0){s.push(c.name);continue}c.inputs.forEach(h=>{l.has(h.name)||(l.add(h.name),d.push(h))})}}return{inputs:e,outputs:t,usedNodes:r,missingInputs:s,dynamicNode:i,syncInputs:o}}function f6(e,t,n){let{usedNodes:a,inputs:r}=n,s=[],i=Object.keys(r).map(p=>Xn(p)[0]).map(p=>e.nodes[p]),o=e.initNodes;i.forEach(p=>{a.has(p.name)&&s.push(p)}),e.weights.forEach(p=>{a.has(p.name)&&s.push(p)}),o!=null&&o.forEach(p=>{a.has(p.name)&&s.push(p)});let l=new Set,u=[];for(;s.length>0;){let p=s.pop();l.add(p.name),t[p.name]||u.push(p),p.children.forEach(d=>{!l.has(d.name)&&a.has(d.name)&&d.inputs.every(c=>l.has(c.name))&&s.push(d)})}return u}var g6=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],y6=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],b6=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function k2(e){return g6.indexOf(e.op)>=0}function x6(e){return y6.indexOf(e.op)>=0}function v6(e){return b6.indexOf(e.op)>=0}var nx=class{constructor(e,t){this.graph=e,this.parent=t,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=e.outputs,this._inputs=e.inputs,this._initNodes=e.initNodes,this._signature=e.signature,this._functions=e.functions,e.functions!=null&&Object.keys(e.functions).forEach(n=>{this._functionExecutorMap[n]=new nx(e.functions[n],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){let t=Object.keys(e).map(n=>e[n].map(a=>a.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{let t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}getCompilationKey(e,t){let n=e.map(r=>r.name).sort(),a=t.map(r=>r.name).sort();return n.join(this.SEPERATOR)+"--"+a.join(this.SEPERATOR)}compile(e,t){let n=Mk(e,t,this.weightMap,this._initNodes),{missingInputs:a,dynamicNode:r,syncInputs:s}=n;if(r!=null)throw new Error(`This execution contains the node '${r.name}', which has the dynamic op '${r.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);if(a.length>0){let i=t.map(l=>l.name),o=Object.keys(e);throw new Error(`Cannot compute the outputs [${i}] from the provided inputs [${o}]. Missing the following inputs: [${a}]`)}return f6(this.graph,this.weightMap,n)}execute(e,t){e=this.mapInputs(e);let n=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);let a=n.map(p=>this.graph.nodes[Xn(p)[0]]),r=t.map(p=>Xn(p)[0]),s=r.map(p=>this.graph.nodes[p]);this.resetIntermediateTensors(),s.length===0&&(s=this._outputs);let i=this.getCompilationKey(a,s),o=this.compiledMap.get(i);o==null&&(o=this.compile(e,s),this.compiledMap.set(i,o));let l={},u={};return P(()=>{let p=new Rk(this.weightMap,l,u,this.functionExecutorMap),d=Object.assign({},this.weightMap);Object.keys(e).forEach(m=>{let[f,g]=Xn(m),y=[];y[g]=e[m],d[f]=y});let c=this.getFrozenTensorIds(d),h={};for(let m=0;m<o.length;m++){let f=o[m];if(!d[f.name]){let g=Dk(f,d,p,this._resourceManager);if(v.isPromise(g))throw new Error(`The execution of the op '${f.op}' returned a promise. Please use model.executeAsync() instead.`);d[f.name]=g,this.checkTensorForDisposal(f.name,f,d,p,c,r,h)}}return this.parent==null&&p.dispose(c),t.map(m=>vn(m,d,p))})}getFrozenTensorIds(e){let t=[].concat.apply([],Object.keys(e).map(n=>e[n]).map(n=>n.map(a=>a.id)));return new Set(t)}checkTensorForDisposal(e,t,n,a,r,s,i){t.category==="control"||s.indexOf(e)!==-1||(n[e].forEach(o=>{o!=null&&(i[o.id]=(i[o.id]||0)+t.children.length)}),t.inputs.forEach(o=>{if(o.category!=="control"){let l=kH(o.name,n,a);l!=null&&l.forEach(u=>{if(u&&!u.kept&&!r.has(u.id)){let p=i[u.id];if(p===1){if(!this.keepTensorForDebug)u.dispose();else{let[d,c]=ar(t.name,a);this.intermediateTensors[d]?this.intermediateTensors[d][c]=u:(this.intermediateTensors[d]=[],this.intermediateTensors[d][c]=u)}delete i[u.id]}else p!=null&&i[u.id]--}})}}))}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(e=>this.intermediateTensors[e].forEach(t=>t.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(e=>{this.tensorsMap[e].forEach(t=>{t&&!t.kept&&!t.isDisposed&&!this.keepIds.has(t.id)&&t.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let e in this.intermediateTensors)this.intermediateTensors[e].forEach(t=>t.dispose()),delete this.intermediateTensors[e]}async _executeAsync(e,t,n=!1,a={},r={}){n||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepTensorForDebug=H().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(u){console.warn(u.message)}this.resetIntermediateTensors();let s=new Rk(this.weightMap,a,r,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(e,s,t,n);let i=t.map(u=>vn(u,this.tensorsMap,s)),o=i.map(u=>u.id),l=Object.keys(e).map(u=>e[u].id);return this.keepIds=new Set([...o,...l,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&s.dispose(this.keepIds),i}async executeFunctionAsync(e,t,n){let a=e.reduce((r,s,i)=>(r[this.inputs[i].name]=s,r),{});return this._executeAsync(a,this.outputNodes,!0,t,n)}async executeWithControlFlow(e,t,n,a){let r=Object.keys(e),s=r.map(b=>this.graph.nodes[Xn(b)[0]]),i=n.map(b=>Xn(b)[0]),o=i.map(b=>this.graph.nodes[b]);o.length===0&&(o=this._outputs);let{usedNodes:l,missingInputs:u,dynamicNode:p,syncInputs:d}=Mk(e,o,this.weightMap,this._initNodes),c=[...s,...this.graph.weights,...this._initNodes||[]].map(b=>({node:b,contexts:t.currentContext})),h=Object.assign({},this.weightMap);Object.keys(e).forEach(b=>{let[x,w]=Xn(b),I=[];I[w]=e[b],h[x]=I});let m={},f=this.getFrozenTensorIds(h),g={};for(;c.length>0;){let b=this.processStack(s,c,t,h,g,f,i,m,l);await Promise.all(b)}p==null&&!a&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let y=o.filter(b=>!k2(b)&&!vn(b.name,h,t)).map(b=>b.name);if(y.length>0){let b="";throw p!=null&&(b=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${d}]`),new Error(`Cannot compute the outputs [${y}] from the provided inputs [${r}]. Consider providing the following inputs: [${u}]. ${b}`)}return h}processStack(e,t,n,a,r,s,i,o,l){let u=[];for(;t.length>0;){let p=t.pop();n.currentContext=p.contexts;let d="";if(p.node.op==="Enter"&&k("isConstant",p.node,a,n)&&([d]=ar(p.node.name,n)),a[p.node.name]==null){let c=Dk(p.node,a,n,this._resourceManager);d||([d]=ar(p.node.name,n));let h=n.currentContext;v.isPromise(c)?u.push(c.then(m=>(a[d]=m,n.currentContext=h,this.checkTensorForDisposal(d,p.node,a,n,s,i,o),this.processChildNodes(p.node,t,n,a,r,l),m))):(a[d]=c,this.checkTensorForDisposal(d,p.node,a,n,s,i,o),this.processChildNodes(p.node,t,n,a,r,l))}else this.processChildNodes(p.node,t,n,a,r,l)}return u}processChildNodes(e,t,n,a,r,s){e.children.forEach(i=>{let[o]=ar(i.name,n);r[o]||!s.has(i.name)||(i.op==="Merge"?i.inputNames.some(l=>!!vn(l,a,n))&&(r[o]=!0,t.push({contexts:n.currentContext,node:i})):i.inputNames.every(l=>!!vn(l,a,n))&&(r[o]=!0,t.push({contexts:n.currentContext,node:i})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(t=>t.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{let n=e[t],[a]=Xn(t),r=this.graph.nodes[a];if(r.attrParams.shape&&r.attrParams.shape.value){let s=r.attrParams.shape.value,i=s.length===n.shape.length&&n.shape.every((o,l)=>s[l]===-1||s[l]===o);v.assert(i,()=>`The shape of dict['${r.name}'] provided in model.execute(dict) must be [${s}], but was [${n.shape}]`)}r.attrParams.dtype&&r.attrParams.dtype.value&&v.assert(n.dtype===r.attrParams.dtype.value,()=>`The dtype of dict['${r.name}'] provided in model.execute(dict) must be ${r.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(e){let t={};for(let n in e)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[n]!=null){let a=this._signature.inputs[n];t[a.name]=e[n]}else t[n]=e[n];return t}checkInputs(e){let t=Object.keys(e).filter(n=>{let[a]=Xn(n);return this.graph.nodes[a]==null});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(t=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[t]!=null?this._signature.outputs[t].name:t,{})}checkOutputs(e){e.forEach(t=>{let[n]=Xn(t);if(!this.graph.nodes[n])throw new Error(`The output '${t}' is not found in the graph`)})}},w6=class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(let e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(let e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}},k6="?tfjs-format=file",I6="model.json",k0=class{constructor(e,t={},n=Ut){this.modelUrl=e,this.loadOptions=t,this.version="n/a",this.io=n,t==null&&(this.loadOptions={}),this.resourceManager=new w6}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}findIOHandler(){let e=this.modelUrl;if(e.load!=null)this.handler=e;else if(this.loadOptions.requestInit!=null)this.handler=this.io.browserHTTPRequest(e,this.loadOptions);else{let t=this.io.getLoadHandlers(e,this.loadOptions);if(t.length===0)t.push(this.io.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}load(){if(this.findIOHandler(),this.handler.load==null)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let e=this.handler.load();return v.isPromise(e)?e.then(t=>this.loadSync(t)):this.loadSync(e)}loadSync(e){this.artifacts=e;let t=this.artifacts.modelTopology,n=this.artifacts.signature;if(this.artifacts.userDefinedMetadata!=null){let r=this.artifacts.userDefinedMetadata;r.signature!=null&&(n=r.signature),r.structuredOutputKeys!=null&&(this.structuredOutputKeys=r.structuredOutputKeys)}this.signature=n,this.version=`${t.versions.producer}.${t.versions.minConsumer}`;let a=this.io.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new nx(Ek.Instance.transformGraph(t,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(a),this.executor.resourceManager=this.resourceManager,e.modelInitializer!=null&&e.modelInitializer.node!=null){let r=Ek.Instance.transformGraph(e.modelInitializer);this.initializer=new nx(r),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=e.initializerSignature}return!0}async save(e,t){if(typeof e=="string"){let n=this.io.getSaveHandlers(e);if(n.length===0)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(n.length>1)throw new Error(`Found more than one (${n.length}) save handlers for URL '${e}'`);e=n[0]}if(e.save==null)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}predict(e,t){let n=this.execute(e,this.outputNodes);if(this.structuredOutputKeys){let a=n instanceof Te?[n]:n,r={};return a.forEach((s,i)=>r[this.structuredOutputKeys[i]]=s),r}return n}normalizeInputs(e){if(!(e instanceof Te)&&!Array.isArray(e)){if(this.signature!=null&&this.signature.inputs!=null)for(let a in this.signature.inputs){let r=this.signature.inputs[a];r.resourceId!=null&&(e[a]=this.resourceIdToCapturedInput[r.resourceId])}return e}e=Array.isArray(e)?e:[e];let t=Object.keys(this.resourceIdToCapturedInput).length;if(e.length+t!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-t} non-resource placeholders, while there are ${e.length} input tensors provided.`);let n=0;return this.inputNodes.reduce((a,r)=>{let s=this.signature?this.signature.inputs[r]:null;return s!=null&&s.resourceId!=null?a[r]=this.resourceIdToCapturedInput[s.resourceId]:a[r]=e[n++],a},{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}executeInitializerGraph(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(e){if(this.resourceIdToCapturedInput={},this.initializerSignature){let t=Object.keys(this.initializerSignature.outputs);for(let n=0;n<t.length;n++){let a=t[n],r=this.initializerSignature.outputs[a];this.resourceIdToCapturedInput[r.resourceId]=e[n]}}}execute(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=this.executor.execute(e,t);return n.length>1?n:n[0]}async executeAsync(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=await this.executor.executeAsync(e,t);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,n)=>(t[n]=[e[n]],t),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&_e(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function S6(e,t={},n=Ut){if(e==null)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");t==null&&(t={}),t.fromTFHub&&typeof e=="string"&&(e=N6(e));let a=new k0(e,t,n);return await a.load(),a}function T6(e){if(e==null)throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");let t;if(e instanceof Array){let[a,r]=e;if(!a)throw new Error("modelJSON must be the first element of the array");if(!r||!(r instanceof ArrayBuffer))throw new Error("An ArrayBuffer of weights must be the second element of the array");if(!("modelTopology"in a))throw new Error("Model JSON is missing 'modelTopology'");if(!("weightsManifest"in a))throw new Error("Model JSON is missing 'weightsManifest'");let s=Ut.getWeightSpecs(a.weightsManifest),i=Ut.getModelArtifactsForJSONSync(a,s,r);t=Ut.fromMemorySync(i)}else if("load"in e)t=e;else if("modelTopology"in e&&"weightSpecs"in e&&"weightData"in e)t=Ut.fromMemorySync(e);else throw new Error("Unknown model format");let n=new k0(t);return n.load(),n}function N6(e){return e.endsWith("/")||(e=e+"/"),`${e}${I6}${k6}`}var C6="4.0.0",I2={};Ae(I2,{CSVDataset:()=>F2,Dataset:()=>Mu,FileDataSource:()=>z2,TextLineDataset:()=>$2,URLDataSource:()=>W2,array:()=>Y6,csv:()=>oj,func:()=>lj,generator:()=>uj,microphone:()=>cj,version_data:()=>dj,webcam:()=>pj,zip:()=>Z6});var _6=us(Hh()),E6=us(Hh());function A6(e,t){return Mh(e,t)}function Mh(e,t,n=new Map,a=new Set){if(e==null)return null;if(typeof Blob=="function"&&e instanceof Blob)return e.slice();if(a.has(e))throw new Error("Circular references are not supported.");if(n.has(e))return n.get(e);let r=t(e);if(r.recurse&&r.value!==null)throw new Error("A deep map function may not return both a value and recurse=true.");if(r.recurse)if(fl(e)){let s=Array.isArray(e)?[]:{};a.add(e);for(let i in e){let o=e[i],l=Mh(o,t,n,a);s[i]=l}return a.delete(e),e.__proto__&&(s.__proto__=e.__proto__),s}else throw new Error(`Can't recurse into non-iterable type: ${e}`);else return n.set(e,r.value),r.value}function $6(e,t=T2){return S2(e,t)}function S2(e,t,n=new Set){let a=e[0];if(n.has(a))throw new Error("Circular references are not supported.");let r=t(e);if(r.recurse&&r.value!==null)throw new Error("A deep zip function may not return both a value and recurse=true.");if(r.recurse)if(fl(a)){let s=Array.isArray(a)?[]:{};n.add(a);for(let i in a){let o=e.map(u=>u[i]),l=S2(o,t,n);s[i]=l}return n.delete(a),s}else throw new Error(`Can't recurse into non-iterable type: ${a}`);else return r.value}function T2(e){return e===null?null:fl(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function N2(e,t){let n=new Map;Mh(e,t,n);for(let a of Array.from(n.keys())){let r=n.get(a);if(v.isPromise(r)){let s=await r;n.set(a,s)}}return Mh(e,t,n)}function fl(e){let t=!1;if(H().get("IS_BROWSER"))t=e instanceof TextDecoder;else{let{StringDecoder:n}=wI();t=e instanceof n}return e!=null&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||typeof e=="object"&&!(e instanceof Te)&&!(e instanceof Promise)&&!t)}function F6(e){return e==null||D6(e)||Array.isArray(e)||typeof e=="object"&&e instanceof Te||v.isTypedArray(e)}function D6(e){return e===null||typeof e!="object"&&typeof e!="function"}function R6(e){return A6(e,M6)}function M6(e){return e instanceof Te?{value:e.clone(),recurse:!1}:fl(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var C2=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,e==null)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(let t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.wrap(this.begin+e),n=this.get(t);return this.set(t,this.pop()),n}},I0=class extends C2{constructor(){super(I0.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){let e=this.capacity*2,t=new Array(e),n=this.length();for(let a=0;a<n;a++)t[a]=this.get(this.wrap(this.begin+a));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};I0.INITIAL_CAPACITY=32;function _2(e){return new L6(e)}function S0(e){return new z6(e)}function P6(e,t){return new E2(e,t)}function O6(e,t=Yr.FAIL){return new K6(e,t)}var tn=class{async toArray(){let e=[],t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){let e=this.prefetch(100),t=[],n=await e.next();for(;!n.done;)t.push(n.value),n=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),n=e(t.value);for(;!t.done&&n;)t=await this.next(),n=e(t.value)}handleErrors(e){return new j6(this,e)}filter(e){return new G6(this,e)}map(e){return new H6(this,e)}mapAsync(e){return new Pk(this,e)}serialMapAsync(e){return new Pk(this,e).serial()}flatmap(e){return new q6(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(t=>t===!0)}rowMajorBatch(e,t=!0){return new U6(this,e,t)}columnMajorBatch(e,t=!0,n=T2){return this.rowMajorBatch(e,t).map(a=>$6(a,n))}concatenate(e,t){return new E2(_2([this,e]),t)}take(e){return e<0||e==null?this:new V6(this,e)}skip(e){return e<0||e==null?this:new B6(this,e)}prefetch(e){return new A2(this,e)}shuffle(e,t){return new X6(this,e,t)}serial(){return new W6(this)}},L6=class extends tn{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let e=this.items[this.trav];return this.trav++,{value:R6(e),done:!1}}},z6=class extends tn{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message=`Error thrown while iterating through a dataset: ${e.message}`,e}}},W6=class extends tn{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},B6=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let e=await this.upstream.next();if(e.done)return e;_e(e.value)}return this.upstream.next()}},V6=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},U6=class extends tn{constructor(e,t,n=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let e=[];for(;e.length<this.batchSize;){let t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},G6=class extends tn{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;_e(e.value)}}},H6=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Map`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Va.getTensorsInContainer(e.value),n=this.transform(e.value),a=Va.getTensorsInContainer(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return{value:n,done:!1}}},j6=class extends tn{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},Pk=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Va.getTensorsInContainer(e.value),n=await this.transform(e.value),a=Va.getTensorsInContainer(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return{value:n,done:!1}}},T0=class extends tn{constructor(){super(),this.outputQueue=new I0,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},q6=class extends T0{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let e=await this.upstream.next();if(e.done)return!1;let t=Va.getTensorsInContainer(e.value),n=this.transform(e.value),a=Va.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return!0}},E2=class extends tn{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,this.iterator==null){let n=await this.moreIterators.next();if(n.done)return{value:null,done:!0};this.iterator=n.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}},Yr;(function(e){e[e.FAIL=0]="FAIL",e[e.SHORTEST=1]="SHORTEST",e[e.LONGEST=2]="LONGEST"})(Yr||(Yr={}));var K6=class extends tn{constructor(e,t=Yr.FAIL){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,n=0;function a(s){return s instanceof tn?{value:s.next().then(i=>(t++,i.done&&n++,i.value)),recurse:!1}:{value:null,recurse:!0}}let r=await N2(this.iterators,a);if(t===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case Yr.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case Yr.SHORTEST:return{value:null,done:!0};case Yr.LONGEST:default:}return this.count++,{value:r,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},A2=class extends tn{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new C2(t)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},X6=class extends A2{constructor(e,t,n){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=E6.alea(n||v.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(t.done)this.upstreamExhausted=!0;else return this.refill(),t}return{value:null,done:!0}}},Mu=class{constructor(){this.size=null}batch(e,t=!0){let n=this;v.assert(e>0,()=>`batchSize needs to be positive, but it is
-      ${e}`);let a;return this.size===1/0||this.size==null?a=this.size:t?a=Math.ceil(this.size/e):a=Math.floor(this.size/e),Kn(async()=>(await n.iterator()).columnMajorBatch(e,t,J6),a)}concatenate(e){let t=this,n;return this.size===1/0||e.size===1/0?n=1/0:this.size!=null&&e.size!=null?n=this.size+e.size:n=null,Kn(async()=>(await t.iterator()).concatenate(await e.iterator()),n)}filter(e){let t=this,n;return this.size===1/0?n=1/0:n=null,Kn(async()=>(await t.iterator()).filter(a=>P(()=>e(a))),n)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){let t=this;return Kn(async()=>(await t.iterator()).map(n=>P(()=>e(n))),this.size)}mapAsync(e){let t=this;return Kn(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(e==null)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let t=this;return Kn(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){let t=this,n;return this.size!=null&&e>0?n=this.size*e:e===0?n=0:this.size!=null&&(e===void 0||e<0)?n=1/0:n=null,Kn(async()=>{let a=S0(async()=>({value:await t.iterator(),done:!1}));return P6(a.take(e))},n)}skip(e){let t=this,n;return this.size!=null&&e>=0&&this.size>=e?n=this.size-e:this.size!=null&&(this.size<e||e===void 0||e<0)?n=0:n=null,Kn(async()=>(await t.iterator()).skip(e),n)}shuffle(e,t,n=!0){if(e==null||e<0)throw this.size==null?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let a=this,r=_6.alea(t||v.now().toString());return Kn(async()=>{let s=r.int32();return n&&(s+=r.int32()),(await a.iterator()).shuffle(e,s.toString())},this.size)}take(e){let t=this,n;return this.size!=null&&this.size>e?n=e:this.size!=null&&this.size<=e?n=this.size:n=null,Kn(async()=>(await t.iterator()).take(e),n)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};Mu.MAX_BUFFER_SIZE=1e4;function Kn(e,t=null){return new class extends Mu{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function Y6(e){return Kn(async()=>_2(e),e.length)}function Z6(e){if(!fl(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let n=0;n<e.length;n++)t=t==null?e[n].size:Math.min(t,e[n].size);else if(e instanceof Object)for(let n in e)t=t==null?e[n].size:Math.min(t,e[n].size);return Kn(async()=>{let n=await N2(e,a=>{if(a instanceof Mu)return{value:a.iterator(),recurse:!1};if(fl(a))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")});return O6(n,Yr.SHORTEST)},t)}function J6(e){if(e===null)return null;let t=e[0];return F6(t)?{value:Q6(e),recurse:!1}:{value:null,recurse:!0}}function Q6(e){if(e.length===0)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof Te?Ft(e):kn(e)}var $2=class extends Mu{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`
-`).map(e=>(e.endsWith("\r")&&(e=e.slice(0,-1)),e))}},Zd='"',Np=Symbol("out"),Ok=Symbol("field"),Jd=Symbol("quote"),fb=Symbol("quoteafterquote"),Lk=Symbol("quoteinquote"),F2=class extends Mu{constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new $2(e),t||(t={}),this.hasHeader=t.hasHeader!==!1,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(v.assert(t.delimiter==null,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&v.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);let t=this.fullColumnNames.reduce((a,r)=>(a[r]=a[r]+1||1,a),{}),n=Object.keys(t).filter(a=>t[a]>1);if(v.assert(n.length===0,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs){for(let a of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(a)===-1)throw new Error('The key "'+a+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error("No data was found for CSV parsing.");let t=e.value;return this.parseRow(t,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(t=>this.makeDataElement(t))}makeDataElement(e){let t=this.parseRow(e),n={},a={};for(let r=0;r<this.fullColumnNames.length;r++){let s=this.fullColumnNames[r],i=this.columnConfigs?this.columnConfigs[s]:null;if(!(this.configuredColumnsOnly&&!i)){let o=t[r],l=null;if(o==="")if(i&&i.default!==void 0)l=i.default;else{if(i&&(i.required||i.isLabel))throw new Error(`Required column ${s} is empty in this line: ${e}`);l=void 0}else{let u=Number(o);if(isNaN(u))i&&i.dtype==="bool"?l=this.getBoolean(o):l=o;else if(!i||!i.dtype)l=u;else switch(i.dtype){case"float32":l=u;break;case"int32":l=Math.floor(u);break;case"bool":l=this.getBoolean(o);break;default:l=u}}i&&i.isLabel?a[s]=l:n[s]=l}}return Object.keys(a).length===0?n:{xs:n,ys:a}}getBoolean(e){return e==="1"||e.toLowerCase()==="true"?1:0}parseRow(e,t=!0){let n=[],a=0,r=e.length,s=Np;for(let i=0;i<r;i++)switch(s){case Np:switch(e.charAt(i)){case Zd:a=i+1,s=Jd;break;case this.delimiter:if(a=i+1,this.delimiter===" "&&this.delimWhitespace)break;n.push(""),s=Np;break;default:s=Ok,a=i;break}break;case Ok:switch(e.charAt(i)){case this.delimiter:n.push(e.substring(a,i)),s=Np,a=i+1;break;default:}break;case Jd:switch(e.charAt(i)){case Zd:s=fb;break;default:}break;case fb:switch(e.charAt(i)){case this.delimiter:n.push(e.substring(a,i-1)),s=Np,a=i+1;break;case Zd:s=Jd;break;default:s=Lk;break}break;case Lk:switch(e.charAt(i)){case Zd:s=Jd;break;default:}break;default:}if(s===fb?n.push(e.substring(a,r-1)):n.push(e.substring(a)),t&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}},D2=class extends tn{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;let t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=e.includeSpectrogram!==!1,this.includeWaveform=e.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(e={}){if(!H().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let t=new D2(e);return await t.start(),t}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(n){throw new Error(`Error thrown while initializing video stream: ${n.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t,n=await this.getAudioData();if(this.includeSpectrogram){let a=this.flattenQueue(n.freqDataQueue);e=this.getTensorFromAudioDataArray(a,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let a=this.flattenQueue(n.timeDataQueue);t=this.getTensorFromAudioDataArray(a,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let e=[],t=[],n=0;return new Promise(a=>{let r=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&a({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(r),a({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){let t=e[0].length,n=new Float32Array(e.length*t);return e.forEach((a,r)=>n.set(a,r*t)),n}getTensorFromAudioDataArray(e,t){let n=new Float32Array(v.sizeFromShape(t));return n.set(e,n.length-e.length),kn(n,t)}},R2=class extends tn{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Ke([0],"int32"),this.webcamConfig.centerCrop){let n=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,a=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,r=(1-n)/2,s=(1-a)/2,i=r+n,o=a+s;this.cropBox=_a([s,r,o,i],[1,4])}else this.cropBox=_a([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(e,t={}){if(!H().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!e){if(e=document.createElement("video"),!t.resizeWidth||!t.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");e.width=t.resizeWidth,e.height=t.resizeHeight}let n=new R2(e,t);return await n.start(),n}async start(){this.webcamConfig.facingMode&&v.assert(this.webcamConfig.facingMode==="user"||this.webcamConfig.facingMode==="environment",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message=`Error thrown while initializing video stream: ${e.message}`,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=co.fromPixels(this.webcamVideoElement)}catch(t){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(t)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(t){throw new Error(`Error thrown cropping the video: ${t.message}`)}finally{e.dispose()}else return{value:e,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(e){return P(()=>{let t=Zt(oe(e,"float32"),0),n;n=za.cropAndResize(t,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let a=n.shape;return W(n,a.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},M2=class{},P2=class extends tn{split(e){return new ej(this,e)}},ej=class extends P2{constructor(e,t){super(),this.upstream=e,this.impl=new tj(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},tj=class extends T0{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let e=await this.upstream.next();if(e.done)return this.carryover===""?!1:(this.outputQueue.push(this.carryover),this.carryover="",!0);let t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(let n of t.slice(0,-1))this.outputQueue.push(n);return this.carryover=t[t.length-1],!0}},nj=class extends tn{decodeUTF8(){return new aj(this)}},aj=class extends P2{constructor(e){super(),this.upstream=e,this.impl=new rj(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},rj=class extends T0{constructor(e){if(super(),this.upstream=e,H().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:t}=wI();this.decoder=new t("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let e=await this.upstream.next(),t;if(e.done)return!1;t=e.value;let n;return H().get("IS_BROWSER")?n=this.decoder.decode(t,{stream:!0}):n=this.decoder.write(Buffer.from(t.buffer)),this.outputQueue.push(n),!0}},O2=class extends nj{constructor(e,t={}){super(),this.file=e,this.options=t,v.assert(e instanceof Uint8Array||(H().get("IS_BROWSER")?e instanceof File||e instanceof Blob:!1),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,t)=>{let n=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,n)));else{let a=new FileReader;a.onload=s=>{let i=a.result;if(i instanceof ArrayBuffer&&(i=new Uint8Array(i)),!(i instanceof Uint8Array))return t(new TypeError("FileReader returned unknown type."));e(i)},a.onabort=s=>t(new Error("Aborted")),a.onerror=s=>t(new Error(s.type));let r=this.file.slice(this.offset,n);a.readAsArrayBuffer(r)}this.offset=n}),done:!1}}};async function sj(e,t={},n){let a,r;typeof e=="string"?a=e:(a=e.url,r=ij(e));let s=await(n||v.fetch)(a,r);if(s.ok){let i=new Uint8Array(await s.arrayBuffer());return new O2(i,t)}else throw new Error(s.statusText)}var ij=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function L2(e){return typeof e=="string"&&e.slice(0,7)==="file://"}var z2=class extends M2{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(L2(this.input)&&H().get("IS_NODE")){let e=xx();this.input=e.readFileSync(this.input.slice(7))}return new O2(this.input,this.options)}},W2=class extends M2{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return L2(this.url)?new z2(this.url,this.fileOptions).iterator():sj(this.url,this.fileOptions)}};function oj(e,t={}){return new F2(new W2(e),t)}function lj(e){let t=S0(e);return Kn(async()=>t)}function uj(e){return Kn(async()=>{let t=await e();return S0(()=>t.next())})}async function pj(e,t){return R2.create(e,t)}async function cj(e){return D2.create(e)}var dj="4.0.0";function ge(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&v.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var hj=cr.whereImpl,N0=class extends rc{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new jh(this,Na())}nextDataId(){return N0.nextDataId++}write(e,t,n){this.firstUse&&(this.firstUse=!1,H().get("IS_NODE")&&N.warn(`
-============================
-Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. 
-============================`));let a={id:this.nextDataId()};return this.data.set(a,{values:e,dtype:n,refCount:1}),a}makeTensorInfo(e,t,n){let a;if(t==="string"&&n!=null&&n.length>0&&v.isString(n[0])){let r=n.map(s=>v.encodeString(s));a=this.write(r,e,t)}else a=this.write(n,e,t);return{dataId:a,shape:e,dtype:t}}refCount(e){return this.data.has(e)?this.data.get(e).refCount:0}incRef(e){let t=this.data.get(e);t.refCount++}decRef(e){if(this.data.has(e)){let t=this.data.get(e);t.refCount--}}move(e,t,n,a,r){this.data.set(e,{values:t,dtype:a,refCount:r})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){let{dtype:t,complexTensorInfos:n}=this.data.get(e);if(t==="complex64"){let a=this.readSync(n.real.dataId),r=this.readSync(n.imag.dataId);return N.mergeRealAndImagArrays(a,r)}return this.data.get(e).values}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(a=>v.decodeString(a));return Pe(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Pe(e.shape,e.dtype,t)}makeOutput(e,t,n){return Na().makeTensorFromTensorInfo(this.makeTensorInfo(t,n,e),this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(e);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){ge([e],"where");let t=this.readSync(e.dataId);return hj(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};N0.nextDataId=0;var B2={};Ae(B2,{addImpl:()=>G2,bincountImpl:()=>_0,bincountReduceImpl:()=>H2,castImpl:()=>U2,ceilImpl:()=>j2,concatImpl:()=>E0,equalImpl:()=>q2,expImpl:()=>X2,expm1Impl:()=>Z2,floorImpl:()=>J2,gatherNdImpl:()=>Q2,gatherV2Impl:()=>eC,greaterEqualImpl:()=>nC,greaterImpl:()=>tC,lessEqualImpl:()=>rC,lessImpl:()=>aC,linSpaceImpl:()=>sC,logImpl:()=>iC,maxImpl:()=>oC,maximumImpl:()=>lC,minimumImpl:()=>uC,multiplyImpl:()=>A0,negImpl:()=>pC,notEqualImpl:()=>cC,prodImpl:()=>dC,raggedGatherImpl:()=>hC,raggedRangeImpl:()=>mC,raggedTensorToTensorImpl:()=>fC,rangeImpl:()=>F0,rsqrtImpl:()=>gC,scatterImpl:()=>Yo,sigmoidImpl:()=>oq,simpleAbsImpl:()=>V2,sliceImpl:()=>Oh,sparseFillEmptyRowsImpl:()=>bC,sparseReshapeImpl:()=>xC,sparseSegmentReductionImpl:()=>D0,sqrtImpl:()=>pq,squaredDifferenceImpl:()=>vC,stridedSliceImpl:()=>wC,stringNGramsImpl:()=>R0,stringSplitImpl:()=>M0,stringToHashBucketFastImpl:()=>P0,subImpl:()=>kC,tileImpl:()=>IC,topKImpl:()=>TC,transposeImpl:()=>$0,uniqueImpl:()=>NC});function V2(e){let t=new Float32Array(e.length);for(let n=0;n<e.length;++n)t[n]=Math.abs(e[n]);return t}var mj=e=>{let{x:t}=e.inputs,n=e.backend;ge(t,"abs");let a=new Float32Array(v.sizeFromShape(t.shape)),r=n.data.get(t.dataId).values;return a=V2(r),n.makeOutput(a,t.shape,t.dtype)},fj={kernelName:wl,backendName:"cpu",kernelFunc:mj};function Vt(e){return(t,n,a,r,s)=>{let i=N.assertAndGetBroadcastShape(t,n),o=i.length,l=v.computeStrides(i),u=v.sizeFromShape(i),p=v.getTypedArrayFromDType(s,u),d=t.length,c=n.length,h=v.computeStrides(t),m=v.computeStrides(n),f=N.getBroadcastDims(t,i),g=N.getBroadcastDims(n,i);if(f.length+g.length===0)for(let y=0;y<p.length;++y)p[y]=e(a[y%a.length],r[y%r.length]);else for(let y=0;y<p.length;++y){let b=v.indexToLoc(y,o,l),x=b.slice(-d);f.forEach(C=>x[C]=0);let w=v.locToIndex(x,d,h),I=b.slice(-c);g.forEach(C=>I[C]=0);let T=v.locToIndex(I,c,m);p[y]=e(a[w],r[T])}return[p,i]}}function Yn(e){let{inputs:t,backend:n}=e,{real:a,imag:r}=t,s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,o=n.makeTensorInfo(a.shape,"complex64"),l=n.data.get(o.dataId);return l.complexTensorInfos={real:n.makeTensorInfo(a.shape,"float32",s),imag:n.makeTensorInfo(r.shape,"float32",i)},o}var gj={kernelName:Qh,backendName:"cpu",kernelFunc:Yn};function Ph(e,t,n="float32"){if(n==="complex64"){let r=Ph(e,t,"float32"),s=Ph(e,t,"float32");return Yn({inputs:{real:r,imag:s},backend:e})}let a=v.makeZerosTypedArray(v.sizeFromShape(t),n);return e.makeTensorInfo(t,n,a)}function lr(e){let{inputs:t,backend:n}=e,{x:a}=t;return n.incRef(a.dataId),{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}var yj={kernelName:Di,backendName:"cpu",kernelFunc:lr};function li(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.data.get(a.dataId).complexTensorInfos.real,s=n.data.get(r.dataId).values;return n.makeTensorInfo(r.shape,r.dtype,s)}var bj={kernelName:wm,backendName:"cpu",kernelFunc:li};function U2(e,t,n,a){if(a==="int32"){let r=Int32Array.from(e);return[t,"int32",r]}if(a==="bool"){let r=v.toTypedArray([0],n),[s,i]=Vt((o,l)=>o!==l?1:0)(t,[],e,r,"bool");return[i,"bool",s]}throw new Error(`Error in Cast: failed to cast ${n} to ${a}`)}function os(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dtype:s}=a;if(s==="complex64"){if(r.dtype==="complex64")return lr({inputs:{x:r},backend:n});let p=Ph(n,r.shape,r.dtype),d=os({inputs:{x:r},backend:n,attrs:{dtype:"float32"}}),c=Yn({inputs:{real:d,imag:p},backend:n});return n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),c}if(r.dtype==="complex64"){let p=li({inputs:{input:r},backend:n}),d=os({inputs:{x:p},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(p),d}if(!v.hasEncodingLoss(r.dtype,s)){let p=lr({inputs:{x:r},backend:n});return{dataId:p.dataId,shape:p.shape,dtype:s}}let i=n.data.get(r.dataId).values,[o,l,u]=U2(i,r.shape,r.dtype,s);return n.makeTensorInfo(o,l,u)}var xj={kernelName:bi,backendName:"cpu",kernelFunc:os};function nn(e,t,n,a){return n==null?({inputs:r,backend:s})=>{let{a:i,b:o}=r,l=s;ge([i,o],e);let u=l.data.get(i.dataId).values,p=l.data.get(o.dataId).values,d=i.dtype==="string"?N.fromUint8ToStringArray(u):u,c=i.dtype==="string"?N.fromUint8ToStringArray(p):p,h=a||i.dtype,[m,f]=t(i.shape,o.shape,d,c,h);return l.makeTensorInfo(f,h,m)}:({inputs:r,backend:s})=>{let{a:i,b:o}=r,l=s;if(i.dtype==="complex64"||o.dtype==="complex64"){let u=os({inputs:{x:i},backend:l,attrs:{dtype:"complex64"}}),p=l.data.get(u.dataId),d=p.complexTensorInfos.real,c=p.complexTensorInfos.imag,h=l.data.get(d.dataId).values,m=l.data.get(c.dataId).values,f=os({inputs:{x:o},backend:l,attrs:{dtype:"complex64"}}),g=l.data.get(f.dataId),y=g.complexTensorInfos.real,b=g.complexTensorInfos.imag,x=l.data.get(y.dataId).values,w=l.data.get(b.dataId).values,[I,T,C]=n(i.shape,o.shape,h,m,x,w),E=l.makeTensorInfo(C,"float32",I),A=l.makeTensorInfo(C,"float32",T),R=Yn({inputs:{real:E,imag:A},backend:l});return l.disposeIntermediateTensorInfo(u),l.disposeIntermediateTensorInfo(f),l.disposeIntermediateTensorInfo(E),l.disposeIntermediateTensorInfo(A),R}else{let u=l.data.get(i.dataId).values,p=l.data.get(o.dataId).values,d=a||i.dtype,[c,h]=t(i.shape,o.shape,u,p,d);return l.makeTensorInfo(h,d,c)}}}function C0(e){return(t,n,a,r,s,i)=>{let o=N.assertAndGetBroadcastShape(t,n),l=v.sizeFromShape(o),u=o.length,p=v.computeStrides(o),d=v.getTypedArrayFromDType("float32",l),c=v.getTypedArrayFromDType("float32",l),h=N.getBroadcastDims(t,o),m=N.getBroadcastDims(n,o),f=N.mergeRealAndImagArrays(a,r),g=N.mergeRealAndImagArrays(s,i),y=t.length,b=v.computeStrides(t),x=n.length,w=v.computeStrides(n);if(h.length+m.length===0)for(let I=0;I<d.length;I++){let T=I%f.length,C=I%g.length,E=e(f[T*2],f[T*2+1],g[C*2],g[C*2+1]);d[I]=E.real,c[I]=E.imag}else for(let I=0;I<d.length;I++){let T=v.indexToLoc(I,u,p),C=T.slice(-y);h.forEach(S=>C[S]=0);let E=v.locToIndex(C,y,b),A=T.slice(-x);m.forEach(S=>A[S]=0);let R=v.locToIndex(A,x,w),F=e(f[E*2],f[E*2+1],g[R*2],g[R*2+1]);d[I]=F.real,c[I]=F.imag}return[d,c,o]}}var G2=Vt((e,t)=>e+t),vj=C0((e,t,n,a)=>({real:e+n,imag:t+a})),gl=nn(cs,G2,vj),wj={kernelName:cs,backendName:"cpu",kernelFunc:gl};function _0(e,t,n,a,r){let s=v.sizeFromShape(a),i=v.makeZerosTypedArray(r,n);for(let o=0;o<e.length;o++){let l=e[o];if(l<0)throw new Error("Input x must be non-negative!");l>=r||(s>0?i[l]+=t[o]:i[l]+=1)}return i}function H2(e,t,n,a=!1){let r=e.shape[0],s=e.shape[1],i=Pe([r,n],t.dtype);for(let o=0;o<r;o++)for(let l=0;l<s;l++){let u=e.get(o,l);if(u<0)throw new Error("Input x must be non-negative!");u>=n||(a?i.set(1,o,u):t.size>0?i.set(i.get(o,u)+t.get(o,l),o,u):i.set(i.get(o,u)+1,o,u))}return i}function ks(e){return(t,n,a)=>{let r=v.getTypedArrayFromDType(n,t.length);for(let s=0;s<t.length;++s)r[s]=e(t[s],a);return r}}function rt(e,t,n){return({inputs:a,attrs:r,backend:s})=>{let{x:i}=a;if(ge(i,e),i.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=v.sizeFromShape(i.shape),p=n||i.dtype,d=v.getArrayFromDType(p,u);for(let c=0;c<u;++c)d[c]=t(l[c],r);return o.makeTensorInfo(i.shape,p,d)}}function Pu(e,t,n){return({inputs:a,attrs:r,backend:s})=>{let{x:i}=a;if(ge(i,e),i.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=n||i.dtype,p=t(l,u,r);return o.makeTensorInfo(i.shape,u,p)}}var j2=ks(e=>Math.ceil(e)),kj=Pu(xi,j2),Ij={kernelName:xi,backendName:"cpu",kernelFunc:kj};function E0(e,t,n,a){let r=v.getArrayFromDType(n,v.sizeFromShape(t));if(a&&n!=="string"){let s=0;e.forEach(i=>{let o=v.sizeFromShape(i.shape);r.set(i.vals,s),s+=o})}else{let s=0;e.forEach(i=>{let o=n==="string"?N.fromUint8ToStringArray(i.vals):i.vals,l=0;for(let u=0;u<i.shape[0];++u){let p=u*t[1]+s;for(let d=0;d<i.shape[1];++d)r[p+d]=o[l++]}s+=i.shape[1]})}return r}var q2=Vt((e,t)=>e===t?1:0),K2=nn(Ol,q2,null,"bool"),Sj={kernelName:Ol,backendName:"cpu",kernelFunc:K2},X2=ks(e=>Math.exp(e)),Y2=Pu(_i,X2,"float32"),Tj={kernelName:_i,backendName:"cpu",kernelFunc:Y2},Z2=ks(e=>Math.expm1(e)),Nj=Pu(zl,Z2),Cj={kernelName:zl,backendName:"cpu",kernelFunc:Nj},J2=ks(e=>Math.floor(e)),_j=Pu(Ei,J2),Ej={kernelName:Ei,backendName:"cpu",kernelFunc:_j};function Q2(e,t,n,a,r,s,i,o,l){let u=Pe([a,s],n);for(let p=0;p<a;p++){let d=[],c=0;for(let h=0;h<r;h++){let m=e[p*r+h];c+=m*i[h],d.push(m)}if(c<0||c>=l/s)throw new Error(`Invalid indices: ${d} does not index into ${o}`);for(let h=0;h<s;h++)u.values[p*s+h]=t.get(...t.indexToLoc(c*s+h))}return u}function eC(e,t,n){let a=Pe(n,e.dtype);for(let r=0;r<a.size;++r){let s=a.indexToLoc(r).slice(),i=s[0],o=s[2],l=t.locToIndex([i,o]);s[2]=t.values[l];let u=e.locToIndex(s);0<=u&&u<e.values.length&&(a.values[r]=e.values[u])}return a}var tC=Vt((e,t)=>e>t?1:0),Aj=nn(Ul,tC,null,"bool"),$j={kernelName:Ul,backendName:"cpu",kernelFunc:Aj},nC=Vt((e,t)=>e>=t?1:0),Fj=nn(Fi,nC,null,"bool"),Dj={kernelName:Fi,backendName:"cpu",kernelFunc:Fj},aC=Vt((e,t)=>e<t?1:0),Rj=nn(ql,aC,null,"bool"),Mj={kernelName:ql,backendName:"cpu",kernelFunc:Rj},rC=Vt((e,t)=>e<=t?1:0),Pj=nn(Kl,rC,null,"bool"),Oj={kernelName:Kl,backendName:"cpu",kernelFunc:Pj};function sC(e,t,n){let a=(t-e)/(n-1),r=v.makeZerosTypedArray(n,"float32");r[0]=e;for(let s=1;s<r.length;s++)r[s]=r[s-1]+a;return r}var iC=ks(e=>Math.log(e)),Lj=Pu(Mi,iC),zj={kernelName:Mi,backendName:"cpu",kernelFunc:Lj};function oC(e,t,n,a){let r=v.getTypedArrayFromDType(a,v.sizeFromShape(n));for(let s=0;s<r.length;++s){let i=s*t,o=e[i];for(let l=0;l<t;++l){let u=e[i+l];(Number.isNaN(u)||u>o)&&(o=u)}r[s]=o}return r}var lC=Vt((e,t)=>Math.max(e,t)),Wj=nn(Oi,lC),Bj={kernelName:Oi,backendName:"cpu",kernelFunc:Wj},uC=Vt((e,t)=>Math.min(e,t)),Vj=nn(Bi,uC),Uj={kernelName:Bi,backendName:"cpu",kernelFunc:Vj},A0=Vt((e,t)=>e*t),Gj=C0((e,t,n,a)=>({real:e*n-t*a,imag:e*a+t*n})),Ff=nn(Ui,A0,Gj),Hj={kernelName:Ui,backendName:"cpu",kernelFunc:Ff};function pC(e,t,n){let a=v.createScalarValue(-1,n);return A0([],t,a,e,n)}function jj(e){let{inputs:t,backend:n}=e,{x:a}=t;ge(a,"neg");let r=n.data.get(a.dataId).values,[s,i]=pC(r,a.shape,a.dtype);return n.makeTensorInfo(i,a.dtype,s)}var qj={kernelName:eu,backendName:"cpu",kernelFunc:jj},cC=Vt((e,t)=>e!==t?1:0),Kj=nn(tu,cC,null,"bool"),Xj={kernelName:tu,backendName:"cpu",kernelFunc:Kj};function $0(e,t,n,a,r){let s=t.length,i=v.sizeFromShape(t),o=v.computeStrides(t),l=v.computeStrides(r),u=v.getTypedArrayFromDType(n,v.sizeFromShape(r));for(let p=0;p<i;++p){let d=v.indexToLoc(p,s,o),c=new Array(d.length);for(let m=0;m<c.length;m++)c[m]=d[a[m]];let h=v.locToIndex(c,s,l);u[h]=e[p]}return u}function Vn(e){let{inputs:t,attrs:n,backend:a}=e,{x:r}=t,{perm:s}=n;ge(r,"transpose");let i=r.shape.length,o=new Array(i);for(let p=0;p<o.length;p++)o[p]=r.shape[s[p]];let l=a.data.get(r.dataId).values,u=$0(l,r.shape,r.dtype,s,o);return{dataId:a.write(u,o,r.dtype),shape:o,dtype:r.dtype}}var Yj={kernelName:Tr,backendName:"cpu",kernelFunc:Vn};function dC(e,t,n,a){let[r,s]=N.computeOutAndReduceShapes(e,a),i=ma(t,"int32"),o=v.makeZerosTypedArray(v.sizeFromShape(r),i),l=v.sizeFromShape(s);for(let u=0;u<o.length;++u){let p=u*l,d=1;for(let c=0;c<l;++c)d*=n[p+c];o[u]=d}return{outVals:o,outShape:r,outDtype:i}}function Zj(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"prod");let o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=N.getAxesPermutation(l,o),p=l,d=r,c=[];u!=null&&(d=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),c.push(d),p=N.getInnerMostAxes(p.length,o));let h=n.data.get(d.dataId).values,{outVals:m,outShape:f,outDtype:g}=dC(d.shape,d.dtype,h,p),y=f;return i&&(y=N.expandShapeToKeepDim(f,l)),c.forEach(b=>n.disposeIntermediateTensorInfo(b)),n.makeTensorInfo(y,g,m)}var Jj={kernelName:Ki,backendName:"cpu",kernelFunc:Zj};function Qj(e,t,n){e.forEach((a,r)=>{if(a<0||a>=n){let s=v.indexToLoc(r,t.length,v.computeStrides(t)).join(",");throw new Error(`indices[${s}] = ${a} is not in [0, ${n})`)}})}function eq(e,t){for(let n=0;n<e.length;++n){let a=e[n],r=n===e.length-1?t:e[n+1].length;if(a.length===0)throw new Error("Ragged splits may not be empty");if(a[0]<0)throw new Error("Ragged splits must be non-negative");if(a[a.length-1]>r)throw new Error("Ragged splits must not point past values");for(let s=1;s<a.length;++s)if(a[s-1]>a[s])throw new Error("Ragged splits must be sorted in ascending order")}}function tq(e,t,n,a){let r=[],s=0,i=t.length-1+n.length,o=new Array(i).fill(null).map(()=>[0]);eq(n,a);let l=1;for(let u=0;u<t.length-1;++u){l*=t[u];let p=t[u+1];for(let d=1;d<l+1;++d)o[u].push(d*p)}for(let u=0;u<e.length;++u){let p=e[u],d=e[u]+1;for(let c=0;c<n.length;++c){let h=n[c],m=c+t.length-1;if(m>=0){let f=o[m],g=f[f.length-1]-h[p];for(let y=p;y<d;++y)o[m].push(h[y+1]+g)}p=h[p],d=h[d]}d!==p&&(r.push([p,d]),s+=d-p)}return{outSplits:o,valueSlices:r,numValues:s}}function nq(e){let t=[];for(let n=0;n<e.length;++n){let a=e[n].length,r=v.getArrayFromDType("int32",a);t.push(r),e[n].forEach((s,i)=>r[i]=s)}return t}function zk(e,t){let n=e.slice(0,t);for(;n.length<t;)n.push(1);for(let a=t;a<e.length;a++)n[t-1]*=e[a];return n}function aq(e,t,n,a,r,s){let i=zk(t,2)[1],o=zk(s,2)[1],l=0;for(let u of n)for(let p=u[0];p<u[1];++p){for(let d=0;d<a;++d)r[l*o+d]=e[p*i+d];++l}}function rq(e,t,n,a,r){let s=t.slice();s[0]=r;let i=v.getArrayFromDType(n,v.sizeFromShape(s)),o=e.length,l=o===0?0:o/t[0];return aq(e,t,a,l,i,s),[i,s]}function hC(e,t,n,a,r,s,i,o){if(e.length===0)throw new Error("paramsNestedSplits must be non empty");if(t[0].length===0)throw new Error("Split tensors must not be scalars");let l=t[0][0]-1;if(Qj(s,i,l),a.length===0)throw new Error("params.rank must be nonzero");let u=a[0],{outSplits:p,valueSlices:d,numValues:c}=tq(s,i,e,u),h=nq(p),m=rq(n,a,r,d,c);return[h,m[0],m[1]]}var Wk=2147483647;function mC(e,t,n,a,r,s,i){if(t.length>1)throw new Error("starts must be a scalar or vector");if(r.length>1)throw new Error("limits must be a scalar or vector");if(i.length>1)throw new Error("deltas must be a scalar or vector");let o=t.length===0,l=r.length===0,u=i.length===0,p=[];o||p.push(t[0]),l||p.push(r[0]),u||p.push(i[0]);for(let g=1;g<p.length;++g)if(p[g]!==p[g-1])throw new Error("starts, limits, and deltas must have the same shape");let d=p.length===0?1:p[0],c=v.getArrayFromDType("int32",d+1);c[0]=0;for(let g=0;g<d;++g){let y=o?e[0]:e[g],b=l?a[0]:a[g],x=u?s[0]:s[g];if(x===0)throw new Error("Requires delta != 0");let w;if(x>0&&b<y||x<0&&b>y)w=0;else if(w=Math.ceil(Math.abs((b-y)/x)),w>Wk)throw new Error(`Requires ((limit - start) / delta) <= ${Wk}`);c[g+1]=c[g]+w}let h=c[d],m=v.getArrayFromDType(n,h),f=0;for(let g=0;g<d;++g){let y=c[g+1]-c[g],b=o?e[0]:e[g],x=u?s[0]:s[g];for(let w=0;w<y;++w)m[f++]=b,b+=x}return[c,m]}var Ia=N.RowPartitionType,ax=class{constructor(e,t,n,a,r,s,i,o,l,u){this.shape=e,this.shapeShape=t,this.values=n,this.valuesShape=a,this.valuesDType=r,this.defaultValue=s,this.defaultValueShape=i,this.rowPartitionValues=o,this.rowPartitionValuesShapes=l,this.rowPartitionTypes=N.getRowPartitionTypesHelper(u),this.raggedRank=N.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(e){return this.rowPartitionTypes[0]===Ia.FIRST_DIM_SIZE?this.rowPartitionTypes[e+1]:this.rowPartitionTypes[e]}getRowPartitionTensor(e){return this.rowPartitionTypes[0]===Ia.FIRST_DIM_SIZE?this.rowPartitionValues[e+1]:this.rowPartitionValues[e]}getMaxWidth(e){let t=this.getRowPartitionTensor(e-1);switch(this.getRowPartitionTypeByDimension(e-1)){case Ia.VALUE_ROWIDS:return ax.getMaxWidthValueRowID(t);case Ia.ROW_SPLITS:return ax.getMaxWidthRowSplit(t);default:throw new Error(`Cannot handle partition type ${Ia[this.getRowPartitionTypeByDimension(e-1)]}`)}}static getMaxWidthRowSplit(e){let t=e.length;if(t===0||t===1)return 0;let n=0;for(let a=0;a<t-1;++a){let r=e[a+1]-e[a];r>n&&(n=r)}return n}static getMaxWidthValueRowID(e){let t=e.length;if(t===0)return 0;let n=0,a=e[0],r=0;for(let s=1;s<t;++s){let i=e[s];i!==a&&(a=i,r=Math.max(s-n,r),n=s)}return Math.max(t-n,r)}tensorShapeFromTensor(e,t,n=!0){if(t.length===0){if(e[0]===-1)return[];throw new Error("The only valid scalar shape tensor is the fully unknown shape specified as -1.")}return Vk(e,n)}calculateOutputSize(e){let t=this.valuesShape,n=this.defaultValueShape;N.validateDefaultValueShape(n,t);let a=this.tensorShapeFromTensor(this.shape,this.shapeShape),r=N.combineRaggedTensorToTensorShapes(this.raggedRank,a,t);r[0]<0&&(r[0]=e);for(let s=1;s<=this.raggedRank;++s)r[s]<0&&(r[s]=this.getMaxWidth(s));return r}calculateFirstParentOutputIndex(e,t,n){let a=Math.min(e,n),r=[],s=0;for(let i=0;i<a;++i,s+=t)r.push(s);for(let i=a;i<e;++i)r.push(-1);return v.assert(r.length===e,()=>"Final length of result must be equal to firstDimension."),r}calculateOutputIndexRowSplit(e,t,n,a){let r=e.length,s=[];for(let i=0;i<r-1;++i){let o=e[i+1]-e[i],l=Math.min(a,o),u=t[i];u===-1&&(l=0);for(let p=0;p<l;++p)s.push(u),u+=n;for(let p=0;p<o-l;++p)s.push(-1)}if(r>0&&s.length!==e[r-1])throw new Error("Invalid row split size.");return s}calculateOutputIndexValueRowID(e,t,n,a){let r=e.length,s=[];if(r===0)return[];let i=0,o=e[0];if(o>=t.length)throw new Error(`Got currentValueRowId=${o}, which is not less than ${t.length}`);let l=t[o];s.push(l);for(let u=1;u<r;++u){let p=e[u];if(p===o)l>=0&&(++i,i<a?l+=n:l=-1);else{if(i=0,o=p,p>=t.length)throw new Error(`Got nextValueRowId=${p} which is not less than ${t.length}`);l=t[p]}s.push(l)}if(s.length!==e.length)throw new Error("Invalid row ids.");return s}calculateOutputIndex(e,t,n,a){let r=this.getRowPartitionTensor(e),s=this.getRowPartitionTypeByDimension(e);switch(s){case Ia.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(r,t,n,a);case Ia.ROW_SPLITS:if(r.length-1>t.length)throw new Error(`Row partition size is greater than output size: ${r.length-1} > ${t.length}`);return this.calculateOutputIndexRowSplit(r,t,n,a);default:throw new Error(`Unsupported partition type: ${Ia[s]}`)}}getFirstDimensionSize(){let e=this.rowPartitionValues[0];if(this.rowPartitionTypes.length===0)throw new Error("No row_partition_types given.");let t=this.rowPartitionTypes[0];switch(t){case Ia.FIRST_DIM_SIZE:return e[0];case Ia.VALUE_ROWIDS:throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");case Ia.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${Ia[t]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error("Invalid first partition input. Tensor requires at least one element.");let e=this.getFirstDimensionSize(),t=this.calculateOutputSize(e),n=new Array(this.raggedRank+1);n[n.length-1]=1;for(let s=n.length-2;s>=0;--s)n[s]=n[s+1]*t[s+1];let a=Vk(t,!1),r=v.getArrayFromDType(this.valuesDType,v.sizeFromShape(a));if(n[0]*t[0]>0){let s=this.calculateFirstParentOutputIndex(e,n[0],t[0]);for(let i=1;i<=this.raggedRank;++i)s=this.calculateOutputIndex(i-1,s,n[i],t[i]);this.setOutput(this.raggedRank,s,r,a)}return[a,r]}setOutput(e,t,n,a){if(n.length===0)return;let r=this.values,s=n,i=a.slice();i=i.slice(e+1);let o=v.sizeFromShape(i),l=t.length,u=this.defaultValue;if(u.length!==o&&u.length!==1){let h=this.defaultValueShape;P(()=>{let m=W(u,h);u=Ks(m,i).dataSync()})}let p=0,d=0,c=0;for(let h=0;h<=l;++h){let m=h<l?t[h]:-1;if(m===c){++c;continue}if(d<c){let f=r.subarray(p*o),g=s.subarray(d*o),y=(c-d)*o;Bk(g,f,y)}if(h>=l){let f=n.length;m=Math.floor(f/o)}if(m>c)if(this.defaultValue.length===1)s.subarray(c*o,m*o).fill(this.defaultValue[0]),c=m;else for(;m>c;){let f=s.slice(c*o);Bk(f,u,o),++c}m<0?(p=h+1,d=c):(p=h,d=c,c=d+1)}}};function Bk(e,t,n){for(let a=0;a<n;a++)e[a]=t[a]}function Vk(e,t){let n=[];for(let a of e){if(a<0){if(!t)throw new Error(`Dimension ${a} must be >= 0`);if(a<-1)throw new Error(`Dimension ${a} must be >= -1`);a=-1}n.push(a)}return n}function fC(e,t,n,a,r,s,i,o,l,u){return new ax(e,t,n,a,r,s,i,o,l,u).compute()}function F0(e,t,n,a){let r=e===t,s=e<t&&n<0,i=t<e&&n>1;if(r||s||i)return v.makeZerosTypedArray(0,a);let o=Math.abs(Math.ceil((t-e)/n)),l=v.makeZerosTypedArray(o,a);t<e&&n===1&&(n=-1),l[0]=e;for(let u=1;u<l.length;u++)l[u]=l[u-1]+n;return l}var gC=ks(e=>1/Math.sqrt(e)),sq=Pu(to,gC),iq={kernelName:to,backendName:"cpu",kernelFunc:sq};function Yo(e,t,n,a,r,s,i,o,l,u){let p=[a/r,r],d=e.values,c=t.values;if(a===0)return Pe(n,t.dtype);let h=Pe(p,t.dtype);typeof l=="string"||typeof l=="number"?h.values.fill(l):typeof l=="boolean"&&h.values.fill(+l);for(let m=0;m<s;m++){let f=[],g=0;for(let y=0;y<i;y++){let b=d[m*i+y];f.push(b),g+=b*o[y]}if(g<0||g>=a/r)throw new Error(`Invalid indices: ${f} does not index into ${n}`);for(let y=0;y<r;y++)u?h.values[g*r+y]+=c[m*r+y]:h.values[g*r+y]=t.rank===0?c[0]:c[m*r+y]}return h}var oq=ks(e=>1/(1+Math.exp(-e))),yC=rt(ao,e=>1/(1+Math.exp(-e))),lq={kernelName:ao,backendName:"cpu",kernelFunc:yC};function Oh(e,t,n,a,r){let s=jt.isSliceContinous(a,t,n),i=v.sizeFromShape(n),o=v.computeStrides(a);if(s){let d=jt.computeFlatOffset(t,o);return r==="string"?e.slice(d,d+i):e.subarray(d,d+i)}let l=r==="string"?N.fromUint8ToStringArray(e):e,u=Pe(a,r,l),p=Pe(n,r);for(let d=0;d<p.size;++d){let c=p.indexToLoc(d),h=c.map((m,f)=>m+t[f]);p.set(u.get(...h),...c)}return r==="string"?N.fromStringArrayToUint8(p.values):p.values}function ui(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,size:i}=a;ge(r,"slice");let[o,l]=jt.parseSliceParams(r,s,i);jt.assertParamsValid(r,o,l);let u=n.data.get(r.dataId).values,p=Oh(u,o,l,r.shape,r.dtype);return n.makeTensorInfo(l,r.dtype,p)}var uq={kernelName:du,backendName:"cpu",kernelFunc:ui};function bC(e,t,n,a,r,s,i){let o=t[0],l=s[0],u=new Array(l),p=new Array(o),d=t[1];if(l===0){if(o!==0)throw new Error(N.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));let g=v.getArrayFromDType(n,0),y=v.getArrayFromDType(r,0);return[g,[0,d],y,u,p]}let c=!0,h=0,m=new Array(l).fill(0);for(let g=0;g<o;++g){let y=e[g*d];if(y<0)throw new Error(N.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,y));if(y>=l)throw new Error(N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,y,l));++m[y],c=c&&y>=h,h=y}let f=!0;for(let g=0;g<l;++g){let y=m[g]===0;u[g]=y,f=f&&!y,m[g]=Math.max(m[g],1),g>0&&(m[g]+=m[g-1])}if(f&&c){let g=e,y=a;for(let b=0;b<o;++b)p[b]=b;return[g,[o,d],y,u,p]}else{let g=m[l-1],y=v.getArrayFromDType(n,g*d),b=v.getArrayFromDType(r,g),x=new Array(l).fill(0);for(let w=0;w<o;++w){let I=e[w*d],T=x[I],C=(I===0?0:m[I-1])+T;x[I]++;for(let E=0;E<d;++E)y[C*d+E]=e[w*d+E];b[C]=a[w],p[w]=C}for(let w=0;w<l;++w)if(x[w]===0){let I=w===0?0:m[w-1];y[I*d+0]=w;for(let T=1;T<d;++T)y[I*d+T]=0;b[I]=i}return[y,[g,d],b,u,p]}}function xC(e,t,n,a,r){let s=v.sizeFromShape(a),i=t[0],o=r.length,l=[],u=1,p=-1;for(let f=0;f<o;++f){let g=r[f];if(g===-1){if(p!==-1)throw new Error(N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(p,f));p=f,l.push(1)}else{if(g<0)throw new Error(N.getSparseReshapeNegativeOutputDimErrorMessage(f,g));u*=g,l.push(g)}}if(p!==-1){if(u<=0)throw new Error(N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let f=Math.trunc(s/u);if(u*f!==s)throw new Error(N.getSparseReshapeInputOutputMultipleErrorMessage(a,l));l[p]=f}if(v.sizeFromShape(l)!==s)throw new Error(N.getSparseReshapeInputOutputMismatchErrorMessage(a,l));let d=a.length,c=[];if(d>0){c[d-1]=1;for(let f=d-2;f>=0;--f)c[f]=c[f+1]*a[f+1]}let h=[];if(o>0){h[o-1]=1;for(let f=o-2;f>=0;--f)h[f]=h[f+1]*l[f+1]}let m=v.getArrayFromDType(n,i*o);for(let f=0;f<i;++f){let g=0;for(let y=0;y<d;++y)g+=e[f*d+y]*c[y];for(let y=0;y<o;++y)m[f*o+y]=Math.trunc(g/h[y]),g%=h[y]}return[m,[i,o],l]}function D0(e,t,n,a,r,s=!1,i=0){let o=a.length,l=[t[0],e.length/t[0]],u=l[1],p=o>0?r[o-1]+1:0;if(p<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let d=t.slice();d[0]=p;let c=d.reduce((b,x)=>b*x,1),h=v.getArrayFromDType(n,c);if(o===0)return p>0&&h.fill(i),[h,d];if(p<=0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let m=0,f=1,g=0,y=r[m];for(;;){let b=0;if(f<o){if(b=r[f],y===b){++f;continue}if(y>=b)throw new Error(N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=p)throw new Error(N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,p));y>g&&h.fill(i,g*u,y*u);for(let x=m;x<f;++x){let w=a[x];if(w<0||w>=l[0])throw new Error(N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(x,a[x],l[0]));for(let I=0;I<u;I++)h[y*u+I]+=e[w*u+I]}if(s)for(let x=0;x<u;x++)h[y*u+x]/=f-m;if(m=f,++f,g=y+1,y=b,f>o)break}return g<p&&h.fill(i,g*u,p*u),[h,d]}var pq=ks(e=>Math.sqrt(e)),cq=rt(ro,e=>Math.sqrt(e)),dq={kernelName:ro,backendName:"cpu",kernelFunc:cq},vC=Vt((e,t)=>{let n=e-t;return n*n}),hq=nn(oo,vC),mq={kernelName:oo,backendName:"cpu",kernelFunc:hq};function wC(e,t,n,a){let r=Pe(e,t.dtype);for(let s=0;s<r.size;s++){let i=r.indexToLoc(s),o=new Array(i.length);for(let l=0;l<o.length;l++)o[l]=i[l]*n[l]+a[l];r.set(t.get(...o),...i)}return r}var fq=class{constructor(e,t,n,a,r,s){this.separator=v.encodeString(e),this.nGramWidths=t,this.leftPad=v.encodeString(n),this.rightPad=v.encodeString(a),this.padWidth=r,this.preserveShort=s}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){let n=this.getPadWidth(t);return Math.max(0,e+2*n-t+1)}createNGrams(e,t,n,a,r,s){for(let i=0;i<r;++i){let o=this.getPadWidth(s),l=Math.max(0,o-i),u=Math.max(0,o-(r-(i+1))),p=s-(l+u),d=t+(l>0?0:i-o),c=0;c+=l*this.leftPad.length;for(let g=0;g<p;++g)c+=e[d+g].length;c+=u*this.rightPad.length,c+=(l+u+p-1)*this.separator.length,n[a+i]=new Uint8Array(c);let h=n[a+i],m=0,f=g=>g.forEach(y=>h[m++]=y);for(let g=0;g<l;++g)f(this.leftPad),f(this.separator);for(let g=0;g<p-1;++g)f(e[d+g]),f(this.separator);if(p>0){f(e[d+p-1]);for(let g=0;g<u;++g)f(this.separator),f(this.rightPad)}else{for(let g=0;g<u-1;++g)f(this.rightPad),f(this.separator);f(this.rightPad)}}}compute(e,t){let n=e.length,a=t.length;if(a>0){let o=t[0];if(o!==0)throw new Error(`First split value must be 0, got ${o}`);for(let l=1;l<a;++l){let u=t[l]>=o;if(u=u&&t[l]<=n,!u)throw new Error(`Invalid split value ${t[l]}, must be in [${o}, ${n}]`);o=t[l]}if(o!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${o}`)}let r=a-1,s=v.getArrayFromDType("int32",a);if(n===0||a===0){let o=new Array(n);for(let l=0;l<=r;++l)s[l]=0;return[o,s]}s[0]=0;for(let o=1;o<=r;++o){let l=t[o]-t[o-1],u=0;this.nGramWidths.forEach(p=>{u+=this.getNumNGrams(l,p)}),this.preserveShort&&l>0&&u===0&&(u=1),s[o]=s[o-1]+u}let i=new Array(s[r]);for(let o=0;o<r;++o){let l=t[o],u=s[o];if(this.nGramWidths.forEach(p=>{let d=t[o+1]-t[o],c=this.getNumNGrams(d,p);this.createNGrams(e,l,i,u,c,p),u+=c}),this.preserveShort&&u===s[o]){let p=t[o+1]-t[o];if(p===0)continue;let d=p+2*this.padWidth,c=1;this.createNGrams(e,l,i,u,c,d)}}return[i,s]}};function R0(e,t,n,a,r,s,i,o){return new fq(n,a,r,s,i,o).compute(e,t)}function gq(e,t,n,a){if(!e.length)return;if(t.length===0){for(let s=0;s<e.length;++s)a.push(e.subarray(s,s+1));return}if(t.length===1){let s=t[0],i=e.indexOf(s);for(;i!==-1;){let o=e.subarray(0,i);(!n||o.length!==0)&&a.push(o),e=e.subarray(i+1),i=e.indexOf(s)}(!n||e.length!==0)&&a.push(e);return}let r=0;for(let s=0;s<e.length+1;s++)if(s===e.length||t.indexOf(e[s])!==-1){let i=e.subarray(r,s);(!n||i.length!==0)&&a.push(i),r=s+1}}function M0(e,t,n){let a=e.length,r=[],s=0,i=0,o=new Array(a);for(let c=0;c<a;++c){let h=r.length;gq(e[c],t,n,r);let m=r.length-h;o[c]=m,s+=m,i=Math.max(i,m)}let l=v.getArrayFromDType("int32",s*2),u=new Array(s),p=[a,i],d=0;for(let c=0;c<a;++c)for(let h=0;h<o[c];++h)l[d*2]=c,l[d*2+1]=h,u[d]=r[d],++d;return[l,u,p]}function P0(e,t){let n=v.getArrayFromDType("int32",e.length);for(let a=0;a<e.length;++a)n[a]=v.fingerPrint64(e[a]).modulo(t).getLowBitsUnsigned();return n}var kC=Vt((e,t)=>e-t),yq=C0((e,t,n,a)=>({real:e-n,imag:t-a})),O0=nn(lo,kC,yq),bq={kernelName:lo,backendName:"cpu",kernelFunc:O0};function IC(e,t){let n=new Array(e.rank);for(let r=0;r<n.length;r++)n[r]=e.shape[r]*t[r];let a=Pe(n,e.dtype);for(let r=0;r<a.values.length;++r){let s=a.indexToLoc(r),i=new Array(e.rank);for(let l=0;l<i.length;l++)i[l]=s[l]%e.shape[l];let o=e.locToIndex(i);a.values[r]=e.values[o]}return a}var $p=(e,t)=>{let n=t.value-e.value;return n===0?e.index-t.index:n};function SC(e,t,n=0,a=e.length-1){for(;a>n;){if(a-n>600){let o=a-n+1,l=t-n+1,u=Math.log(o),p=.5*Math.exp(2*u/3),d=.5*Math.sqrt(u*p*(o-p)/o)*Math.sign(l-o/2),c=Math.max(n,Math.floor(t-l*p/o+d)),h=Math.min(a,Math.floor(t+(o-l)*p/o+d));SC(e,t,c,h)}let r=e[t],s=n,i=a;for(v.swap(e,n,t),$p(e[a],r)>0&&v.swap(e,n,a);s<i;){for(v.swap(e,s,i),s++,i--;$p(e[s],r)<0;)s=s+1;for(;$p(e[i],r)>0;)i=i-1}$p(e[n],r)===0?v.swap(e,n,i):(i=i+1,v.swap(e,i,a)),i<=t&&(n=i+1),t<=i&&(a=i-1)}}function TC(e,t,n,a,r){let s=t[t.length-1],[i,o]=[e.length/s,s],l=v.getTypedArrayFromDType(n,i*a),u=v.getTypedArrayFromDType("int32",i*a);for(let d=0;d<i;d++){let c=d*o,h=e.subarray(c,c+o),m=new Array(h.length);h.forEach((b,x)=>m[x]={value:b,index:x}),a<m.length&&(SC(m,a),m=m.slice(0,a)),r&&m.sort($p);let f=d*a,g=l.subarray(f,f+a),y=u.subarray(f,f+a);for(let b=0;b<a;b++)g[b]=m[b].value,y[b]=m[b].index}let p=t.slice();return p[p.length-1]=a,[Pe(p,n,l),Pe(p,"int32",u)]}function NC(e,t,n,a){let r=v.parseAxisParam(t,n)[0],s=[1,n[0],1];for(let m=0;m<r;m++)s[0]*=n[m];s[1]=n[r];for(let m=r+1;m<n.length;m++)s[2]*=n[m];let i={},o=new Int32Array(n[r]),l=new Ht(s,a,e),u=[],p=s[0]===1&&s[2]===1;for(let m=0;m<n[r];m++){let f;if(p)f=e[m].toString();else{let g=[];for(let y=0;y<s[0];y++)for(let b=0;b<s[2];b++)g.push(l.get(y,m,b));f=g.join(",")}if(i[f]!==void 0)o[m]=i[f];else{let g=Object.keys(i).length;i[f]=g,o[m]=g,u.push(m)}}let d=s.slice();d[1]=Object.keys(i).length;let c=new Ht(d,a);u.forEach((m,f)=>{for(let g=0;g<s[0];g++)for(let y=0;y<s[2];y++)c.set(l.get(g,m,y),g,f,y)});let h=n.slice();return h[r]=d[1],{outputValues:c.values,outputShape:h,indices:o}}Em("cpu",()=>new N0,1);var CC=rt(Ci,e=>e>=0?e:Math.exp(e)-1),xq={kernelName:Ci,backendName:"cpu",kernelFunc:CC};function _C(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{alpha:s}=a;ge([r],"leakyRelu");let i=v.sizeFromShape(r.shape),o=n.data.get(r.dataId).values,l=v.getTypedArrayFromDType("float32",i);for(let u=0;u<o.length;u++)l[u]=o[u]<0?s*o[u]:o[u];return n.makeTensorInfo(r.shape,"float32",l)}var vq={kernelName:Ri,backendName:"cpu",kernelFunc:_C},wq=Vt((e,t)=>e<0?t*e:e);function EC(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t;ge([a,r],"prelu");let s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,[o,l]=wq(a.shape,r.shape,s,i,"float32");return n.makeTensorInfo(l,"float32",o)}var kq={kernelName:qi,backendName:"cpu",kernelFunc:EC},AC=rt(Xi,e=>Math.max(0,e)),Iq={kernelName:Xi,backendName:"cpu",kernelFunc:AC},$C=rt(Ji,e=>Math.min(Math.max(0,e),6)),Sq={kernelName:Ji,backendName:"cpu",kernelFunc:$C};function Lh(e,t,n,a,r){if(n==="linear")return lr({inputs:{x:t},backend:e});if(n==="relu")return AC({inputs:{x:t},backend:e});if(n==="elu")return CC({inputs:{x:t},backend:e});if(n==="relu6")return $C({inputs:{x:t},backend:e});if(n==="prelu")return EC({inputs:{x:t,alpha:a},backend:e});if(n==="leakyrelu")return _C({inputs:{x:t},backend:e,attrs:{alpha:r}});if(n==="sigmoid")return yC({inputs:{x:t},backend:e});throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)}function ft(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{shape:s}=a,i=v.sizeFromShape(r.shape),o=v.inferFromImplicitShape(s,i),l=v.sizeFromShape(o);v.assert(i===l,()=>`The new shape (${o}) has ${l} elements and the old shape (${r.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),n.incRef(r.dataId);let u=n.data.get(r.dataId);if(u.complexTensorInfos!=null){let p=u.complexTensorInfos.real,d=u.complexTensorInfos.imag;p.shape=o,d.shape=o}return{dataId:r.dataId,shape:o,dtype:r.dtype}}var Tq={kernelName:lu,backendName:"cpu",kernelFunc:ft};function FC(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;ge([r,s],"matMul");let l=r.shape.length,u=s.shape.length,p=i?r.shape[l-2]:r.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],c=i?r.shape[l-1]:r.shape[l-2],h=o?s.shape[u-2]:s.shape[u-1],m=r.shape.slice(0,-2),f=s.shape.slice(0,-2),g=v.sizeFromShape(m),y=v.sizeFromShape(f),b=Tu.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)).concat([c,h]);v.assert(p===d,()=>`Error in matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${r.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,p,c]:[g,c,p],w=o?[y,h,d]:[y,d,h],I=ft({inputs:{x:r},backend:n,attrs:{shape:x}}),T=ft({inputs:{x:s},backend:n,attrs:{shape:w}}),C=i?I.shape[1]:I.shape[2],E=i?I.shape[2]:I.shape[1],A=o?T.shape[1]:T.shape[2],R=Math.max(g,y),F=n.data.get(I.dataId).values,S=n.data.get(T.dataId).values,M=v.computeStrides(I.shape),B=v.computeStrides(T.shape),[U,G,q]=i?[M[0],1,M[1]]:[M[0],M[1],1],[K,Z,Q]=o?[1,B[1],B[0]]:[B[1],1,B[0]],ee=E*A,ae=Pe([R,E,A],I.dtype),te=ae.values,le=n.blockSize;for(let ie=0;ie<R;ie++)for(let ye=0;ye<E;ye+=le)for(let ue=0;ue<A;ue+=le)for(let xe=0;xe<C;xe+=le){let Ie=Math.min(ye+le,E),Se=Math.min(ue+le,A),Le=Math.min(xe+le,C);for(let Ve=ye;Ve<Ie;Ve++)for(let nt=ue;nt<Se;nt++){let st=0;for(let Je=xe;Je<Le;Je++){let at=Math.min(ie,g-1)*U,ze=Math.min(ie,y-1)*Q,dt=F[at+Ve*G+Je*q],Hn=S[Je*K+nt*Z+ze];st+=dt*Hn}te[ie*ee+(Ve*A+nt)]+=st}}return n.disposeIntermediateTensorInfo(I),n.disposeIntermediateTensorInfo(T),n.makeTensorInfo(b,ae.dtype,ae.values)}var Nq={kernelName:yi,backendName:"cpu",kernelFunc:FC};function Cq(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a,c,h,m,f=[];c=FC({inputs:{a:r,b:s},attrs:{transposeA:l,transposeB:u},backend:n}),i&&(h=gl({inputs:{a:c,b:i},backend:n}),f.push(c),c=h),p&&(m=Lh(n,c,p,o,d),f.push(c),c=m);for(let g of f)n.disposeIntermediateTensorInfo(g);return c}var _q={kernelName:Js,backendName:"cpu",kernelFunc:Cq},Eq=rt(kl,e=>Math.acos(e)),Aq={kernelName:kl,backendName:"cpu",kernelFunc:Eq},$q=rt(Il,e=>Math.acosh(e)),Fq={kernelName:Il,backendName:"cpu",kernelFunc:$q};function Dq(e){let{inputs:t,backend:n}=e,a=t;ge(t,"addN");let r=a.map(o=>n.data.get(o.dataId).values),s=Pe(a[0].shape,a[0].dtype),i=s.values;for(let o=0;o<a.length;o++){let l=r[o];for(let u=0;u<i.length;u++)i[u]+=l[u]}return n.makeTensorInfo(s.shape,s.dtype,s.values)}var Rq={kernelName:mi,backendName:"cpu",kernelFunc:Dq};function Mq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"all");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("all",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];x=x&&I}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var Pq={kernelName:Sl,backendName:"cpu",kernelFunc:Mq};function Oq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"any");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("any",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];x=x||I}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var Lq={kernelName:Tl,backendName:"cpu",kernelFunc:Oq};function zq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a;ge(r,"argMax");let i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Vn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],N.assertAxesAreInnerMostDims("argMax",i,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,i),c=v.sizeFromShape(p),h=v.makeZerosTypedArray(c,"int32"),m=v.sizeFromShape(d),f=n.data.get(l.dataId).values;for(let g=0;g<h.length;++g){let y=g*m,b=f[y],x=0;for(let w=0;w<m;++w){let I=f[y+w];I>b&&(b=I,x=w)}h[g]=x}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(p,"int32",h)}var Wq={kernelName:fi,backendName:"cpu",kernelFunc:zq};function Bq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a;ge(r,"argMin");let i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Vn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],N.assertAxesAreInnerMostDims("argMin",i,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,i),c=v.sizeFromShape(p),h=v.makeZerosTypedArray(c,"int32"),m=v.sizeFromShape(d),f=n.data.get(l.dataId).values;for(let g=0;g<h.length;++g){let y=g*m,b=f[y],x=0;for(let w=0;w<m;++w){let I=f[y+w];I<b&&(b=I,x=w)}h[g]=x}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(p,"int32",h)}var Vq={kernelName:sc,backendName:"cpu",kernelFunc:Bq},Uq=rt(Nl,e=>Math.asin(e)),Gq={kernelName:Nl,backendName:"cpu",kernelFunc:Uq},Hq=rt(Cl,e=>Math.asinh(e)),jq={kernelName:Cl,backendName:"cpu",kernelFunc:Hq},qq=rt(_l,e=>Math.atan(e)),Kq={kernelName:_l,backendName:"cpu",kernelFunc:qq},Xq=Vt((e,t)=>Math.atan2(e,t)),Yq=nn(Al,Xq),Zq={kernelName:Al,backendName:"cpu",kernelFunc:Yq},Jq=rt(El,e=>Math.atanh(e)),Qq={kernelName:El,backendName:"cpu",kernelFunc:Jq};function L0(e,t,n,a,r,s){let i=r.strideHeight,o=r.strideWidth,l=r.dilationHeight,u=r.dilationWidth,p=r.effectiveFilterHeight,d=r.effectiveFilterWidth,c=r.padInfo.top,h=r.padInfo.left,m=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,f=Pe(r.outShape,n),g=f.values,y=r.outShape[1]*r.outShape[2]*r.outShape[3],b=r.outShape[2]*r.outShape[3],x=r.outShape[3];for(let w=0;w<r.batchSize;++w){let I=w*y,T=w*a[0];for(let C=0;C<r.inChannels;++C)for(let E=0;E<r.outHeight;++E){let A=E*i-c,R=Math.max(0,A),F=Math.min(r.inHeight,p+A),S=I+E*b;for(let M=0;M<r.outWidth;++M){let B=M*o-h,U=Math.max(0,B),G=Math.min(r.inWidth,d+B),q=m,K=0,Z=0;for(let ee=R;ee<F;ee+=l){let ae=T+ee*a[1];for(let te=U;te<G;te+=u){let le=ae+te*a[2],ie=e[le+C];s==="max"&&ie>q?q=ie:s==="avg"&&(K+=ie,Z++)}if(isNaN(q))break}let Q=S+M*x+C;g[Q]=s==="avg"?K/Z:q}}}return f}function DC(e,t,n,a,r=!1,s=!1){let i=Pe(a.outShape,"int32"),o=a.strideHeight,l=a.strideWidth,u=a.dilationHeight,p=a.dilationWidth,d=a.effectiveFilterHeight,c=a.effectiveFilterWidth,h=a.padInfo.top,m=a.padInfo.left,f=Pe(t,n,e);for(let g=0;g<a.batchSize;++g)for(let y=0;y<a.inChannels;++y)for(let b=0;b<a.outHeight;++b){let x=b*o-h,w=x;for(;w<0;)w+=u;let I=Math.min(a.inHeight,d+x);for(let T=0;T<a.outWidth;++T){let C=T*l-m,E=C;for(;E<0;)E+=p;let A=Math.min(a.inWidth,c+C),R=Number.NEGATIVE_INFINITY,F=-1;for(let S=w;S<I;S+=u){let M=S-x;for(let B=E;B<A;B+=p){let U=B-C,G=f.get(g,S,B,y);G>R&&(R=G,r?F=s?((g*a.inHeight+S)*a.inWidth+B)*a.inChannels+y:(S*a.inWidth+B)*a.inChannels+y:F=M*c+U)}}i.set(F,g,b,T,y)}}return i}function RC(e,t,n,a,r,s){let i=r.strideDepth,o=r.strideHeight,l=r.strideWidth,u=r.dilationDepth,p=r.dilationHeight,d=r.dilationWidth,c=r.effectiveFilterDepth,h=r.effectiveFilterHeight,m=r.effectiveFilterWidth,f=r.padInfo.front,g=r.padInfo.top,y=r.padInfo.left,b=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,x=Pe(r.outShape,n),w=x.values,I=r.outShape[1]*r.outShape[2]*r.outShape[3]*r.outShape[4],T=r.outShape[2]*r.outShape[3]*r.outShape[4],C=r.outShape[3]*r.outShape[4],E=r.outShape[4];for(let A=0;A<r.batchSize;++A){let R=A*I,F=A*a[0];for(let S=0;S<r.inChannels;++S)for(let M=0;M<r.outDepth;++M){let B=M*i-f,U=B;for(;U<0;)U+=u;let G=Math.min(r.inDepth,c+B),q=R+M*T;for(let K=0;K<r.outHeight;++K){let Z=K*o-g,Q=Z;for(;Q<0;)Q+=p;let ee=Math.min(r.inHeight,h+Z),ae=q+K*C;for(let te=0;te<r.outWidth;++te){let le=te*l-y,ie=le;for(;ie<0;)ie+=d;let ye=Math.min(r.inWidth,m+le),ue=ae+te*E,xe=b,Ie=0,Se=0;for(let Ve=U;Ve<G;Ve+=u){let nt=F+Ve*a[1];for(let st=Q;st<ee;st+=p){let Je=nt+st*a[2];for(let at=ie;at<ye;at+=d){let ze=Je+at*a[3],dt=e[ze+S];if(s==="max"&&dt>xe?xe=dt:s==="avg"&&(Ie+=dt,Se++),isNaN(xe))break}if(isNaN(xe))break}if(isNaN(xe))break}let Le=ue+S;w[Le]=s==="avg"?Ie/Se:xe}}}}return x}function e5(e,t){let n=Pe(t.outShape,"int32"),a=t.strideDepth,r=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,o=t.dilationHeight,l=t.dilationWidth,u=t.effectiveFilterDepth,p=t.effectiveFilterHeight,d=t.effectiveFilterWidth,c=t.padInfo.front,h=t.padInfo.top,m=t.padInfo.left;for(let f=0;f<t.batchSize;++f)for(let g=0;g<t.inChannels;++g)for(let y=0;y<t.outDepth;++y){let b=y*a-c,x=b;for(;x<0;)x+=i;let w=Math.min(t.inDepth,u+b);for(let I=0;I<t.outHeight;++I){let T=I*r-h,C=T;for(;C<0;)C+=o;let E=Math.min(t.inHeight,p+T);for(let A=0;A<t.outWidth;++A){let R=A*s-m,F=R;for(;F<0;)F+=l;let S=Math.min(t.inWidth,d+R),M=Number.NEGATIVE_INFINITY,B=-1;for(let U=x;U<w;U+=i){let G=U-b;for(let q=C;q<E;q+=o){let K=q-T;for(let Z=F;Z<S;Z+=l){let Q=Z-R,ee=e.get(f,U,q,Z,g);ee>=M&&(M=ee,B=G*p*d+K*p+Q)}}}n.set(B,f,y,I,A,g)}}}return n}function t5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;ge(r,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l),d;if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))d=lr({inputs:{x:r},backend:n});else{let c=n.data.get(r.dataId).values,h=v.computeStrides(r.shape),m=L0(c,r.shape,r.dtype,h,p,"avg");d=n.makeTensorInfo(p.outShape,r.dtype,m.values)}return d}var n5={kernelName:gi,backendName:"cpu",kernelFunc:t5};function a5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a;ge(r,"avgPool3d");let p=N.computePool3DInfo(r.shape,s,i,1,o,l,u),d=n.data.get(r.dataId).values,c=RC(d,r.shape,r.dtype,v.computeStrides(r.shape),p,"avg");return n.makeTensorInfo(c.shape,"float32",c.values)}var r5={kernelName:ic,backendName:"cpu",kernelFunc:a5};function s5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=a;ge([r,s],"avgPool3DGrad");let p=N.computePool3DInfo(s.shape,i,o,1,l,u),d=p.strideDepth,c=p.strideHeight,h=p.strideWidth,m=p.filterDepth,f=p.filterHeight,g=p.filterWidth,y=p.dilationDepth,b=p.dilationHeight,x=p.dilationWidth,w=p.effectiveFilterDepth,I=p.effectiveFilterHeight,T=p.effectiveFilterWidth,C=w-1-p.padInfo.front,E=T-1-p.padInfo.left,A=I-1-p.padInfo.top,R=Pe(s.shape,"float32"),F=1/(m*f*g),S=n.bufferSync(r);for(let M=0;M<p.batchSize;++M)for(let B=0;B<p.inChannels;++B)for(let U=0;U<p.inDepth;++U)for(let G=0;G<p.inHeight;++G)for(let q=0;q<p.inWidth;++q){let K=U-C,Z=G-A,Q=q-E,ee=0;for(let ae=0;ae<w;ae+=y){let te=(K+ae)/d;if(!(te<0||te>=p.outDepth||Math.floor(te)!==te))for(let le=0;le<I;le+=b){let ie=(Z+le)/c;if(!(ie<0||ie>=p.outHeight||Math.floor(ie)!==ie))for(let ye=0;ye<T;ye+=x){let ue=(Q+ye)/h;ue<0||ue>=p.outWidth||Math.floor(ue)!==ue||(ee+=S.get(M,te,ie,ue,B))}}}R.set(ee*F,M,U,G,q,B)}return n.makeTensorInfo(R.shape,R.dtype,R.values)}var i5={kernelName:Yh,backendName:"cpu",kernelFunc:s5};function o5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s;ge([r,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=a,p=N.computePool2DInfo(i.shape,o,l,1,u),d=p.strideHeight,c=p.strideWidth,h=p.filterHeight,m=p.filterWidth,f=p.dilationHeight,g=p.dilationWidth,y=p.effectiveFilterHeight,b=p.effectiveFilterWidth,x=b-1-p.padInfo.left,w=y-1-p.padInfo.top,I=Pe(i.shape,"float32"),T=1/(h*m),C=n.data.get(r.dataId).values,E=Pe(r.shape,"float32",C);for(let A=0;A<p.batchSize;++A)for(let R=0;R<p.inChannels;++R)for(let F=0;F<p.inHeight;++F)for(let S=0;S<p.inWidth;++S){let M=F-w,B=S-x,U=0;for(let G=0;G<y;G+=f){let q=(M+G)/d;if(!(q<0||q>=p.outHeight||Math.floor(q)!==q))for(let K=0;K<b;K+=g){let Z=(B+K)/c;Z<0||Z>=p.outWidth||Math.floor(Z)!==Z||(U+=E.get(A,q,Z,R))}}I.set(U*T,A,F,S,R)}return n.makeTensorInfo(I.shape,I.dtype,I.values)}var l5={kernelName:Xh,backendName:"cpu",kernelFunc:o5};function u5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,scale:s,offset:i,mean:o,variance:l}=t;v.assert(o.shape.length===l.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(s==null||o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),ge([r,o,l,s,i],"batchNorm");let{varianceEpsilon:u}=a;u==null&&(u=.001);let p=n.data.get(r.dataId).values,d=n.data.get(o.dataId).values,c=n.data.get(l.dataId).values,h=s?n.data.get(s.dataId).values:new Float32Array([1]),m=i?n.data.get(i.dataId).values:new Float32Array([0]),f=new Float32Array(p.length),g=m.length,y=h.length,b=c.length,x=d.length,w=0,I=0,T=0,C=0;for(let E=0;E<p.length;++E)f[E]=m[w++]+(p[E]-d[I++])*h[T++]/Math.sqrt(c[C++]+u),w>=g&&(w=0),I>=x&&(I=0),T>=y&&(T=0),C>=b&&(C=0);return n.makeTensorInfo(r.shape,r.dtype,f)}var p5={kernelName:$i,backendName:"cpu",kernelFunc:u5};function c5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a;ge([r],"batchToSpaceND");let o=s.reduce((y,b)=>y*b),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=ft({inputs:{x:r},backend:n,attrs:{shape:l}}),m=Vn({inputs:{x:h},backend:n,attrs:{perm:u}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:p}}),g=ui({inputs:{x:f},backend:n,attrs:{begin:d,size:c}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),g}var d5={kernelName:$l,backendName:"cpu",kernelFunc:c5};function h5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i}=a,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,u=_0(o,l,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,u)}var m5={kernelName:Zh,backendName:"cpu",kernelFunc:h5};function f5(e){let{inputs:t,backend:n}=e,{s0:a,s1:r}=t,s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,o=N.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var g5={kernelName:Jh,backendName:"cpu",kernelFunc:f5},y5=rt(ds,(e,t)=>{let n=t;return e>n.clipValueMax?n.clipValueMax:e<n.clipValueMin?n.clipValueMin:e}),b5={kernelName:ds,backendName:"cpu",kernelFunc:y5},x5=e=>{let{x:t}=e.inputs,n=e.backend,a=new Float32Array(v.sizeFromShape(t.shape)),r=n.data.get(t.dataId),s=r.complexTensorInfos.real,i=r.complexTensorInfos.imag,o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values;for(let u=0;u<o.length;u++){let p=o[u],d=l[u];a[u]=Math.hypot(p,d)}return n.makeOutput(a,t.shape,"float32")},v5={kernelName:oc,backendName:"cpu",kernelFunc:x5};function yl(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.data.get(a.dataId).complexTensorInfos.imag,s=n.data.get(r.dataId).values;return n.makeTensorInfo(r.shape,r.dtype,s)}var w5={kernelName:cm,backendName:"cpu",kernelFunc:yl};function bl(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a,s=v.parseAxisParam(r,t[0].shape)[0],i=t.map(f=>f.shape);N.assertParamsConsistent(i,s);let o=N.computeOutShape(t.map(f=>f.shape),s);if(v.sizeFromShape(o)===0)return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(f=>v.sizeFromShape(f.shape)>0);if(l.length===1)return lr({inputs:{x:l[0]},backend:n});if(l[0].dtype==="complex64"){let f=l.map(w=>li({inputs:{input:w},backend:n})),g=l.map(w=>yl({inputs:{input:w},backend:n})),y=bl({inputs:f,backend:n,attrs:{axis:s}}),b=bl({inputs:g,backend:n,attrs:{axis:s}}),x=Yn({inputs:{real:y,imag:b},backend:n});return f.forEach(w=>n.disposeIntermediateTensorInfo(w)),g.forEach(w=>n.disposeIntermediateTensorInfo(w)),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(b),x}let u=l.map(f=>{let g=v.sizeFromShape(f.shape.slice(s));return ft({inputs:{x:f},backend:n,attrs:{shape:[-1,g]}})}),p=u.map(f=>({vals:n.data.get(f.dataId).values,shape:f.shape}));o=N.computeOutShape(u.map(f=>f.shape),1);let d=u[0].shape[0]===1,c=E0(p,o,t[0].dtype,d),h=N.computeOutShape(l.map(f=>f.shape),s),m=n.makeTensorInfo(h,t[0].dtype,c);return u.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var k5={kernelName:Fl,backendName:"cpu",kernelFunc:bl};function MC(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:p}=a;ge([r,s],"conv2d");let d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,s.shape,i,u,o,p,!1,d),h=c.filterHeight,m=c.filterWidth,f=c.dilationHeight,g=c.dilationWidth,y=c.padInfo.left,b=c.padInfo.top,x=c.dataFormat==="channelsLast",w=new Ht(c.outShape,r.dtype),I=v.computeStrides(r.shape),T=v.computeStrides(s.shape),C=I[0],E=x?I[1]:I[2],A=x?I[2]:1,R=x?1:I[1],F=w.strides[0],S=x?w.strides[1]:w.strides[2],M=x?w.strides[2]:1,B=x?1:w.strides[1],U=n.data.get(r.dataId).values,G=n.data.get(s.dataId).values,q=w.values;for(let K=0;K<c.batchSize;++K){let Z=K*C,Q=K*F;for(let ee=0;ee<c.outHeight;++ee){let ae=Q+ee*S,te=ee*c.strideHeight-b;for(let le=0;le<h;++le){let ie=te+le*f;if(ie<0||ie>=c.inHeight)continue;let ye=le*T[0],ue=Z+ie*E;for(let xe=0;xe<c.outWidth;++xe){let Ie=ae+xe*M,Se=xe*c.strideWidth-y;for(let Le=0;Le<m;++Le){let Ve=Se+Le*g;if(Ve<0||Ve>=c.inWidth)continue;let nt=ye+Le*T[1],st=ue+Ve*A,Je=nt;for(let at=0;at<c.inChannels;++at){let ze=U[st+at*R];for(let dt=0;dt<c.outChannels;++dt)q[Ie+dt*B]+=ze*G[Je+dt];Je+=c.outChannels}}}}}}return n.makeTensorInfo(w.shape,w.dtype,q)}var I5={kernelName:vi,backendName:"cpu",kernelFunc:MC};function S5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:p}=a;ge([r,s],"conv2dBackpropFilter");let d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,p,i,1,o,u,!1,d),{strideHeight:h,strideWidth:m,filterHeight:f,filterWidth:g}=c,y=c.dataFormat==="channelsLast",b=new Ht(c.filterShape,"float32"),x=c.padInfo.left,w=c.padInfo.top,I=n.data.get(r.dataId).values,T=n.data.get(s.dataId).values,C=new Ht(r.shape,r.dtype,I),E=new Ht(s.shape,s.dtype,T);for(let A=0;A<f;++A){let R=Math.max(0,Math.ceil((w-A)/h)),F=Math.min(c.outHeight,(c.inHeight+w-A)/h);for(let S=0;S<g;++S){let M=Math.max(0,Math.ceil((x-S)/m)),B=Math.min(c.outWidth,(c.inWidth+x-S)/m);for(let U=0;U<c.inChannels;++U)for(let G=0;G<c.outChannels;++G){let q=0;for(let K=0;K<c.batchSize;++K)for(let Z=R;Z<F;++Z){let Q=A+Z*h-w;for(let ee=M;ee<B;++ee){let ae=S+ee*m-x;y?q+=C.get(K,Q,ae,U)*E.get(K,Z,ee,G):q+=C.get(K,U,Q,ae)*E.get(K,G,Z,ee)}}b.set(q,A,S,U,G)}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}var T5={kernelName:em,backendName:"cpu",kernelFunc:S5};function N5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:p}=a;ge([r,s],"conv2dBackpropInput");let d=v.computeStrides(s.shape),c=v.computeStrides(r.shape),h=N.convertConv2DDataFormat(u),m=N.computeConv2DInfo(i,s.shape,o,1,l,p,!1,h),f=new Ht(m.inShape,"float32"),g=f.values,y=n.data.get(r.dataId).values,b=n.data.get(s.dataId).values,[x,w,I]=d,{batchSize:T,filterHeight:C,filterWidth:E,inChannels:A,inHeight:R,inWidth:F,outChannels:S,outHeight:M,outWidth:B,strideHeight:U,strideWidth:G}=m;h=m.dataFormat;let q=C-1-m.padInfo.top,K=E-1-m.padInfo.left,Z=h==="channelsLast",Q=f.strides[0],ee=Z?f.strides[1]:f.strides[2],ae=Z?f.strides[2]:1,te=Z?1:f.strides[1],le=c[0],ie=Z?c[1]:c[2],ye=Z?c[2]:1,ue=Z?1:c[1];for(let xe=0;xe<T;++xe)for(let Ie=0;Ie<A;++Ie)for(let Se=0;Se<R;++Se){let Le=Se-q,Ve=Math.max(0,Math.ceil(Le/U)),nt=Math.min(M,(C+Le)/U);for(let st=0;st<F;++st){let Je=st-K,at=Math.max(0,Math.ceil(Je/G)),ze=Math.min(B,(E+Je)/G),dt=0;for(let Mt=Ve;Mt<nt;++Mt){let sa=Mt*U-Le;for(let sn=at;sn<ze;++sn){let Fn=sn*G-Je,ia=le*xe+ie*Mt+ye*sn,Dn=x*(C-1-sa)+w*(E-1-Fn)+I*Ie;for(let it=0;it<S;++it){let Rn=y[ia+ue*it],jn=b[Dn+it];dt+=Rn*jn}}}let Hn=Q*xe+ee*Se+ae*st+te*Ie;g[Hn]=dt}}return n.makeTensorInfo(f.shape,f.dtype,f.values)}var C5={kernelName:wi,backendName:"cpu",kernelFunc:N5};function _5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a;ge([r,s],"conv3d");let u=N.computeConv3DInfo(r.shape,s.shape,i,l,o),{filterDepth:p,filterHeight:d,filterWidth:c,dilationDepth:h,dilationHeight:m,dilationWidth:f,padInfo:g}=u,y=g.front,b=g.left,x=g.top,w=new Ht(u.outShape,r.dtype),I=n.data.get(r.dataId).values,T=n.data.get(s.dataId).values,C=w.values,E=v.computeStrides(r.shape),A=v.computeStrides(s.shape);for(let R=0;R<u.batchSize;++R){let F=R*E[0],S=R*w.strides[0];for(let M=0;M<u.outDepth;++M){let B=S+M*w.strides[1],U=M*u.strideDepth-y;for(let G=0;G<p;++G){let q=U+G*h;if(q<0||q>=u.inDepth)continue;let K=G*A[0],Z=F+q*E[1];for(let Q=0;Q<u.outHeight;++Q){let ee=B+Q*w.strides[2],ae=Q*u.strideHeight-x;for(let te=0;te<d;++te){let le=ae+te*m;if(le<0||le>=u.inHeight)continue;let ie=K+te*A[1],ye=Z+le*E[2];for(let ue=0;ue<u.outWidth;++ue){let xe=ee+ue*u.outChannels,Ie=ue*u.strideWidth-b;for(let Se=0;Se<c;++Se){let Le=Ie+Se*f;if(Le<0||Le>=u.inWidth)continue;let Ve=ie+Se*A[2],nt=ye+Le*u.inChannels,st=Ve;for(let Je=0;Je<u.inChannels;++Je){let at=I[nt+Je];for(let ze=0;ze<u.outChannels;++ze)C[xe+ze]+=at*T[st+ze];st+=u.outChannels}}}}}}}}return n.makeTensorInfo(w.shape,w.dtype,w.values)}var E5={kernelName:lc,backendName:"cpu",kernelFunc:_5};function A5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,filterShape:l}=a;ge([r,s],"conv3dBackpropFilterV2");let u=v.computeStrides(r.shape),p=v.computeStrides(s.shape),d=N.computeConv3DInfo(r.shape,l,i,1,o),c=d.strideDepth,h=d.strideHeight,m=d.strideWidth,f=d.filterDepth,g=d.filterHeight,y=d.filterWidth,b=new Ht(d.filterShape,"float32"),x=b.values,[w,I,T,C]=b.strides,E=n.data.get(s.dataId).values,[A,R,F,S]=p,M=n.data.get(r.dataId).values,[B,U,G,q]=u,K=d.padInfo.front,Z=d.padInfo.left,Q=d.padInfo.top;for(let ee=0;ee<f;++ee){let ae=Math.max(0,Math.ceil((K-ee)/c)),te=Math.min(d.outDepth,(d.inDepth+K-ee)/c),le=ee*w;for(let ie=0;ie<g;++ie){let ye=Math.max(0,Math.ceil((Q-ie)/h)),ue=Math.min(d.outHeight,(d.inHeight+Q-ie)/h),xe=ie*I+le;for(let Ie=0;Ie<y;++Ie){let Se=Math.max(0,Math.ceil((Z-Ie)/m)),Le=Math.min(d.outWidth,(d.inWidth+Z-Ie)/m),Ve=Ie*T+xe;for(let nt=0;nt<d.inChannels;++nt){let st=nt*C+Ve;for(let Je=0;Je<d.outChannels;++Je){let at=0;for(let ze=0;ze<d.batchSize;++ze){let dt=ze*B,Hn=ze*A;for(let Mt=ae;Mt<te;++Mt){let sa=(ee+Mt*c-K)*U+dt,sn=Mt*R+Hn;for(let Fn=ye;Fn<ue;++Fn){let ia=(ie+Fn*h-Q)*G+sa,Dn=Fn*F+sn;for(let it=Se;it<Le;++it){let Rn=(Ie+it*m-Z)*q+ia,jn=it*S+Dn;at+=M[Rn+nt]*E[jn+Je]}}}}x[st+Je]=at}}}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}var $5={kernelName:tm,backendName:"cpu",kernelFunc:A5};function F5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{pad:i,strides:o,inputShape:l}=a;ge([r],"conv3dBackpropInputV2");let u=v.computeStrides(r.shape),p=v.computeStrides(s.shape),d=N.computeConv3DInfo(l,s.shape,o,1,i),c=new Ht(d.inShape,"float32"),h=c.values,[m,f,g,y]=c.strides,b=n.data.get(r.dataId).values,[x,w,I,T]=u,C=n.data.get(s.dataId).values,[E,A,R,F]=p,{batchSize:S,filterDepth:M,filterHeight:B,filterWidth:U,inChannels:G,inDepth:q,inHeight:K,inWidth:Z,outChannels:Q,outDepth:ee,outHeight:ae,outWidth:te,strideDepth:le,strideHeight:ie,strideWidth:ye}=d,ue=M-1-d.padInfo.front,xe=B-1-d.padInfo.top,Ie=U-1-d.padInfo.left;for(let Se=0;Se<S;++Se)for(let Le=0;Le<G;++Le)for(let Ve=0;Ve<q;++Ve){let nt=Ve-ue,st=Math.max(0,Math.ceil(nt/le)),Je=Math.min(ee,(M+nt)/le);for(let at=0;at<K;++at){let ze=at-xe,dt=Math.max(0,Math.ceil(ze/ie)),Hn=Math.min(ae,(B+ze)/ie);for(let Mt=0;Mt<Z;++Mt){let sa=Mt-Ie,sn=Math.max(0,Math.ceil(sa/ye)),Fn=Math.min(te,(U+sa)/ye),ia=0;for(let Dn=st;Dn<Je;++Dn){let it=Dn*le-nt;for(let Rn=dt;Rn<Hn;++Rn){let jn=Rn*ie-ze;for(let gr=sn;gr<Fn;++gr){let Mo=gr*ye-sa,Ja=x*Se+w*Dn+I*Rn+T*gr,mp=E*(M-1-it)+A*(B-1-jn)+R*(U-1-Mo)+F*Le;for(let wa=0;wa<Q;++wa){let Po=b[Ja+wa],Xt=C[mp+wa];ia+=Po*Xt}}}}h[m*Se+f*Ve+g*at+y*Mt+Le]=ia}}}return n.makeTensorInfo(c.shape,c.dtype,c.values)}var D5={kernelName:nm,backendName:"cpu",kernelFunc:F5},R5=rt(ki,e=>Math.cos(e)),M5={kernelName:ki,backendName:"cpu",kernelFunc:R5},P5=rt(Ii,e=>Math.cosh(e)),O5={kernelName:Ii,backendName:"cpu",kernelFunc:P5};function L5(e){let{inputs:t,backend:n,attrs:a}=e,{image:r,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=a,[p,d,c,h]=r.shape,m=s.shape[0],[f,g]=o,y=Pe([m,f,g,h],"float32"),b=n.data.get(s.dataId).values,x=n.data.get(i.dataId).values,w=n.data.get(r.dataId).values,I=v.computeStrides(r.shape),T=v.computeStrides(y.shape);for(let C=0;C<m;C++){let E=C*4,A=b[E],R=b[E+1],F=b[E+2],S=b[E+3],M=x[C];if(M>=p)continue;let B=f>1?(F-A)*(d-1)/(f-1):0,U=g>1?(S-R)*(c-1)/(g-1):0;for(let G=0;G<f;G++){let q=f>1?A*(d-1)+G*B:.5*(A+F)*(d-1);if(q<0||q>d-1){for(let K=0;K<g;K++)for(let Z=0;Z<h;Z++){let Q=Z+K*T[2]+G*T[1]+C*T[0];y.values[Q]=u}continue}if(l==="bilinear"){let K=Math.floor(q),Z=Math.ceil(q),Q=q-K;for(let ee=0;ee<g;ee++){let ae=g>1?R*(c-1)+ee*U:.5*(R+S)*(c-1);if(ae<0||ae>c-1){for(let ye=0;ye<h;ye++){let ue=ye+ee*T[2]+G*T[1]+C*T[0];y.values[ue]=u}continue}let te=Math.floor(ae),le=Math.ceil(ae),ie=ae-te;for(let ye=0;ye<h;ye++){let ue=ye+te*I[2]+K*I[1]+M*I[0],xe=w[ue];ue=ye+le*I[2]+K*I[1]+M*I[0];let Ie=w[ue];ue=ye+te*I[2]+Z*I[1]+M*I[0];let Se=w[ue];ue=ye+le*I[2]+Z*I[1]+M*I[0];let Le=w[ue],Ve=xe+(Ie-xe)*ie,nt=Se+(Le-Se)*ie;ue=ye+ee*T[2]+G*T[1]+C*T[0],y.values[ue]=Ve+(nt-Ve)*Q}}}else for(let K=0;K<g;++K){let Z=g>1?R*(c-1)+K*U:.5*(R+S)*(c-1);if(Z<0||Z>c-1){for(let ae=0;ae<h;ae++){let te=ae+K*T[2]+G*T[1]+C*T[0];y.values[te]=u}continue}let Q=Math.round(Z),ee=Math.round(q);for(let ae=0;ae<h;ae++){let te=ae+Q*I[2]+ee*I[1]+M*I[0],le=ae+K*T[2]+G*T[1]+C*T[0];y.values[le]=w[te]}}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}var z5={kernelName:Rl,backendName:"cpu",kernelFunc:L5};function W5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;ge(r,"cumprod");let l=N.getAxesPermutation([s],r.shape.length),u=r;l!=null&&(u=Vn({inputs:{x:r},backend:n,attrs:{perm:l}}));let p=N.getInnerMostAxes(1,r.shape.length)[0];if(p!==u.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${p}`);let d=ma(u.dtype,"int32"),c=v.makeOnesTypedArray(v.sizeFromShape(u.shape),d),h=n.data.get(u.dataId).values,m=u.shape[u.shape.length-1],f=o?(y,b)=>y+m-b-1:(y,b)=>y+b;for(let y=0;y<h.length;y+=m)for(let b=0;b<m;b++){let x=f(y,b);if(b===0)c[x]=i?1:h[x];else{let w=f(y,b-1);c[x]=i?h[w]*c[w]:h[x]*c[w]}}let g=n.makeTensorInfo(u.shape,d,c);if(l!=null){let y=N.getUndoAxesPermutation(l),b=Vn({inputs:{x:g},backend:n,attrs:{perm:y}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),b}return g}var B5={kernelName:Dl,backendName:"cpu",kernelFunc:W5};function V5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;ge(r,"cumsum");let l=N.getAxesPermutation([s],r.shape.length),u=r;l!=null&&(u=Vn({inputs:{x:r},backend:n,attrs:{perm:l}}));let p=N.getInnerMostAxes(1,r.shape.length)[0];if(p!==u.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${p}`);let d=ma(u.dtype,"int32"),c=v.makeZerosTypedArray(v.sizeFromShape(u.shape),d),h=n.data.get(u.dataId).values,m=u.shape[u.shape.length-1],f=o?(y,b)=>y+m-b-1:(y,b)=>y+b;for(let y=0;y<h.length;y+=m)for(let b=0;b<m;b++){let x=f(y,b);if(b===0)c[x]=i?0:h[x];else{let w=f(y,b-1);c[x]=i?h[w]+c[w]:h[x]+c[w]}}let g=n.makeTensorInfo(u.shape,d,c);if(l!=null){let y=N.getUndoAxesPermutation(l),b=Vn({inputs:{x:g},backend:n,attrs:{perm:y}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),b}return g}var U5={kernelName:Si,backendName:"cpu",kernelFunc:V5};function G5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i,binaryOutput:o}=a;if(r.shape.length===1){let l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=_0(l,u,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,p)}else if(r.shape.length===2){let l=n.bufferSync(r),u=n.bufferSync(s),p=H2(l,u,i,o);return n.makeTensorInfo(p.shape,s.dtype,p.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${r.shape.length}.`)}var H5={kernelName:am,backendName:"cpu",kernelFunc:G5};function j5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockSize:s,dataFormat:i}=a;v.assert(i==="NHWC",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let o=r.shape[0],l=r.shape[1],u=r.shape[2],p=r.shape[3],d=l*s,c=u*s,h=p/(s*s),m=n.data.get(r.dataId).values,f=new Float32Array(o*d*c*h),g=0;for(let y=0;y<o;++y)for(let b=0;b<d;++b){let x=Math.floor(b/s),w=b%s;for(let I=0;I<c;++I){let T=Math.floor(I/s),C=I%s,E=(w*s+C)*h;for(let A=0;A<h;++A){let R=A+E+p*(T+u*(x+l*y));f[g++]=m[R]}}}return n.makeTensorInfo([o,d,c,h],r.dtype,f)}var q5={kernelName:Ml,backendName:"cpu",kernelFunc:j5};function PC(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=a;ge([r,s],"depthwiseConv2DNative");let p=v.computeStrides(r.shape),d=v.computeStrides(s.shape),c=l;c==null&&(c=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(i,c),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);let h=N.computeConv2DInfo(r.shape,s.shape,i,c,o,u,!0),{filterHeight:m,filterWidth:f,dilationHeight:g,dilationWidth:y,padInfo:b}=h,x=b.left,w=b.top,I=h.outChannels/h.inChannels,T=new Ht(h.outShape,r.dtype),C=n.data.get(r.dataId).values,E=n.data.get(s.dataId).values,A=T.values;for(let R=0;R<h.batchSize;++R){let F=R*p[0],S=R*T.strides[0];for(let M=0;M<h.outHeight;++M){let B=S+M*T.strides[1],U=M*h.strideHeight-w;for(let G=0;G<m;++G){let q=U+G*g;if(q<0||q>=h.inHeight)continue;let K=G*d[0],Z=F+q*p[1];for(let Q=0;Q<h.outWidth;++Q){let ee=B+Q*T.strides[2],ae=Q*h.strideWidth-x;for(let te=0;te<f;++te){let le=ae+te*y;if(le<0||le>=h.inWidth)continue;let ie=K+te*d[1],ye=Z+le*h.inChannels,ue=ee,xe=ie;for(let Ie=0;Ie<h.inChannels;++Ie){let Se=C[ye+Ie];for(let Le=0;Le<I;++Le)A[ue+Le]+=Se*E[xe+Le];ue+=I,xe+=I}}}}}}return n.makeTensorInfo(T.shape,T.dtype,T.values)}var K5={kernelName:Ti,backendName:"cpu",kernelFunc:PC};function X5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:p}=a;ge([r,s],"depthwiseConv2dNativeBackpropFilter");let d=N.computeConv2DInfo(r.shape,p,i,o,l,u,!0),{strideHeight:c,strideWidth:h,filterHeight:m,filterWidth:f}=d,g=new Ht(d.filterShape,"float32"),y=d.padInfo.left,b=d.padInfo.top,x=d.outChannels/d.inChannels,w=n.data.get(r.dataId).values,I=new Ht(r.shape,r.dtype,w),T=n.data.get(s.dataId).values,C=new Ht(s.shape,s.dtype,T);for(let E=0;E<m;++E){let A=Math.max(0,Math.ceil((b-E)/c)),R=Math.min(d.outHeight,(d.inHeight+b-E)/c);for(let F=0;F<f;++F){let S=Math.max(0,Math.ceil((y-F)/h)),M=Math.min(d.outWidth,(d.inWidth+y-F)/h);for(let B=0;B<d.outChannels;++B){let U=Math.trunc(B/x),G=B%x,q=0;for(let K=0;K<d.batchSize;++K)for(let Z=A;Z<R;++Z){let Q=E+Z*c-b;for(let ee=S;ee<M;++ee){let ae=F+ee*h-y;q+=I.get(K,Q,ae,U)*C.get(K,Z,ee,B)}}g.set(q,E,F,U,G)}}}return n.makeTensorInfo(g.shape,g.dtype,g.values)}var Y5={kernelName:rm,backendName:"cpu",kernelFunc:X5};function Z5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:p}=a;ge([r,s],"depthwiseConv2DNativeBackpropInput");let d=v.computeStrides(r.shape),c=v.computeStrides(s.shape),h=N.computeConv2DInfo(p,s.shape,i,o,l,u,!0),m=new Ht(h.inShape,"float32"),f=m.values,[g,y,b]=m.strides,x=n.data.get(r.dataId).values,[w,I,T]=d,C=n.data.get(s.dataId).values,[E,A,R]=c,{batchSize:F,filterHeight:S,filterWidth:M,inChannels:B,inHeight:U,inWidth:G,outChannels:q,outHeight:K,outWidth:Z,strideHeight:Q,strideWidth:ee}=h,ae=S-1-h.padInfo.top,te=M-1-h.padInfo.left,le=q/B;for(let ie=0;ie<F;++ie)for(let ye=0;ye<B;++ye)for(let ue=0;ue<U;++ue){let xe=ue-ae,Ie=Math.max(0,Math.ceil(xe/Q)),Se=Math.min(K,(S+xe)/Q);for(let Le=0;Le<G;++Le){let Ve=Le-te,nt=Math.max(0,Math.ceil(Ve/ee)),st=Math.min(Z,(M+Ve)/ee),Je=0;for(let at=Ie;at<Se;++at){let ze=at*Q-xe;for(let dt=nt;dt<st;++dt){let Hn=dt*ee-Ve,Mt=w*ie+I*at+T*dt,sa=E*(S-1-ze)+A*(M-1-Hn)+R*ye;for(let sn=0;sn<le;++sn){let Fn=ye*le+sn,ia=x[Mt+Fn],Dn=C[sa+sn];Je+=ia*Dn}}}f[g*ie+y*ue+b*Le+ye]=Je}}return n.makeTensorInfo(m.shape,m.dtype,m.values)}var J5={kernelName:sm,backendName:"cpu",kernelFunc:Z5};function Q5(e){let{inputs:t,backend:n}=e,{x:a}=t,r=v.sizeFromShape(a.shape),s=n.data.get(a.dataId).values,i=Pe([r,r],a.dtype),o=i.values;for(let u=0;u<s.length;u++)o[u*r+u]=s[u];let l=[...a.shape,...a.shape];return n.makeTensorInfo(l,i.dtype,i.values)}var e8={kernelName:im,backendName:"cpu",kernelFunc:Q5},t8={kernelName:uc,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r}=e,{strides:s,pad:i,dilations:o}=n,l=t,u=l.data.get(a.dataId).values,p=a.shape.length,d=l.data.get(r.dataId).values,c=r.shape.length,{batchSize:h,inHeight:m,inWidth:f,inChannels:g,outHeight:y,outWidth:b,padInfo:x,strideHeight:w,strideWidth:I,filterHeight:T,filterWidth:C,dilationHeight:E,dilationWidth:A,outShape:R}=N.computeDilation2DInfo(a.shape,r.shape,s,i,"NHWC",o),F=v.sizeFromShape(R),S=R.length,M=v.getArrayFromDType(a.dtype,F);for(let B=0;B<h;++B)for(let U=0;U<y;++U){let G=U*w-x.top;for(let q=0;q<b;++q){let K=q*I-x.left;for(let Z=0;Z<g;++Z){let Q=Number.MIN_SAFE_INTEGER;for(let ae=0;ae<T;++ae){let te=G+ae*E;if(te>=0&&te<m)for(let le=0;le<C;++le){let ie=K+le*A;if(ie>=0&&ie<f){let ye=v.locToIndex([B,te,ie,Z],p,v.computeStrides(a.shape)),ue=v.locToIndex([ae,le,Z],c,v.computeStrides(r.shape)),xe=u[ye]+d[ue];xe>Q&&(Q=xe)}}}let ee=v.locToIndex([B,U,q,Z],S,v.computeStrides(R));M[ee]=Q}}}return{dataId:l.write(v.toTypedArray(M,a.dtype),R,a.dtype),shape:R,dtype:a.dtype}}},n8={kernelName:gh,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,p=v.toNestedArray(a.shape,u.data.get(a.dataId).values),d=v.toNestedArray(r.shape,u.data.get(r.dataId).values),{batchSize:c,inHeight:h,inWidth:m,inChannels:f,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:w,filterHeight:I,filterWidth:T,dilationHeight:C,dilationWidth:E,outShape:A}=N.computeDilation2DInfo(a.shape,r.shape,i,o,"NHWC",l);v.assert(s.rank===A.length,()=>`Error in ${gh}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);let R=v.toNestedArray(A,u.data.get(s.dataId).values),F=v.makeZerosNestedTypedArray(r.shape,r.dtype);for(let S=0;S<c;++S)for(let M=0;M<g;++M){let B=M*x-b.top;for(let U=0;U<y;++U){let G=U*w-b.left;for(let q=0;q<f;++q){let K=Number.MIN_SAFE_INTEGER,Z=0,Q=0;for(let ee=0;ee<I;++ee){let ae=B+ee*C;if(ae>=0&&ae<h)for(let te=0;te<T;++te){let le=G+te*E;if(le>=0&&le<m){let ie=p[S][ae][le][q]+d[ee][te][q];ie>K&&(K=ie,Z=ee,Q=te)}}}F[Z][Q][q]+=R[S][M][U][q]}}}return{dataId:u.write(v.toTypedArray(F,a.dtype),r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},a8={kernelName:fh,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,p=v.toNestedArray(a.shape,u.data.get(a.dataId).values),d=v.toNestedArray(r.shape,u.data.get(r.dataId).values),{batchSize:c,inHeight:h,inWidth:m,inChannels:f,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:w,filterHeight:I,filterWidth:T,dilationHeight:C,dilationWidth:E,outShape:A}=N.computeDilation2DInfo(a.shape,r.shape,i,o,"NHWC",l);v.assert(s.rank===A.length,()=>`Error in ${fh}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);let R=v.toNestedArray(A,u.data.get(s.dataId).values),F=v.makeZerosNestedTypedArray(a.shape,a.dtype);for(let S=0;S<c;++S)for(let M=0;M<g;++M){let B=M*x-b.top;for(let U=0;U<y;++U){let G=U*w-b.left;for(let q=0;q<f;++q){let K=Number.MIN_SAFE_INTEGER,Z=B<0?0:B,Q=G<0?0:G;for(let ee=0;ee<I;++ee){let ae=B+ee*C;if(ae>=0&&ae<h)for(let te=0;te<T;++te){let le=G+te*E;if(le>=0&&le<m){let ie=p[S][ae][le][q]+d[ee][te][q];ie>K&&(K=ie,Z=ae,Q=le)}}}F[S][Z][Q][q]+=R[S][M][U][q]}}}return{dataId:u.write(v.toTypedArray(F,a.dtype),a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}};function Qc(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"sum");let o;r.dtype==="bool"?o=os({inputs:{x:r},backend:n,attrs:{dtype:"int32"}}):o=lr({inputs:{x:r},backend:n});let l=o.shape.length,u=v.parseAxisParam(s,o.shape),p=N.getAxesPermutation(u,l),d=u,c=o;p!=null&&(c=Vn({inputs:{x:o},backend:n,attrs:{perm:p}}),d=N.getInnerMostAxes(d.length,l)),N.assertAxesAreInnerMostDims("sum",d,c.shape.length);let[h,m]=N.computeOutAndReduceShapes(c.shape,d),f=N.upcastType(c.dtype,"int32"),g=Ph(n,h,f),y=v.sizeFromShape(m),b=n.data.get(g.dataId).values,x=n.data.get(c.dataId).values;for(let w=0;w<b.length;++w){let I=w*y,T=0;for(let C=0;C<y;++C)T+=x[I+C];b[w]=T}if(i){let w=N.expandShapeToKeepDim(g.shape,u),I=g;g=ft({inputs:{x:g},backend:n,attrs:{shape:w}}),n.disposeIntermediateTensorInfo(I)}return n.disposeIntermediateTensorInfo(o),p!=null&&n.disposeIntermediateTensorInfo(c),g}var r8={kernelName:so,backendName:"cpu",kernelFunc:Qc};function s8(e){let{inputs:t,backend:n,attrs:a}=e,{equation:r}=a,s=t,{allDims:i,summedDims:o,idDims:l}=N.decodeEinsumEquation(r,s.length);N.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:p}=N.getEinsumComputePath(o,l),d=p.length,c=null,h=i.length,m=[];for(let f=0;f<d;++f){for(let g of p[f]){let{permutationIndices:y,expandDims:b}=N.getEinsumPermutation(h,l[g]),x;N.isIdentityPermutation(y)?x=s[g]:(x=Vn({inputs:{x:s[g]},backend:n,attrs:{perm:y}}),m.push(x));let w=x.shape.slice();for(let I=0;I<b.length;++I)w.splice(b[I],0,1);v.arraysEqual(x.shape,w)||(x=ft({inputs:{x},backend:n,attrs:{shape:w}}),m.push(x)),c===null?c=x:(c=Ff({inputs:{a:x,b:c},backend:n}),m.push(c))}f<d-1&&(u[f]>=0&&(c=Qc({inputs:{x:c},backend:n,attrs:{axis:u[f]-(i.length-h),keepDims:!1}}),m.push(c)),h--)}for(let f of m)f!==c&&n.disposeIntermediateTensorInfo(f);return c}var i8={kernelName:om,backendName:"cpu",kernelFunc:s8};function o8(e){let{inputs:t,backend:n}=e,{dy:a,y:r}=t;ge([a,r],"eluGrad");let s=new Float32Array(v.sizeFromShape(r.shape)),i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values;for(let l=0;l<i.length;++l){let u=i[l];u>=1?s[l]=o[l]:s[l]=o[l]*(u+1)}return n.makeTensorInfo(r.shape,"float32",s)}var l8={kernelName:lm,backendName:"cpu",kernelFunc:o8},u8=N.ERF_P,p8=N.ERF_A1,c8=N.ERF_A2,d8=N.ERF_A3,h8=N.ERF_A4,m8=N.ERF_A5,f8=rt(Pl,e=>{let t=Math.sign(e),n=Math.abs(e),a=1/(1+u8*n);return t*(1-((((m8*a+h8)*a+d8)*a+c8)*a+p8)*a*Math.exp(-n*n))}),g8={kernelName:Pl,backendName:"cpu",kernelFunc:f8};function zh(e){let{inputs:t,backend:n,attrs:a}=e,{input:r}=t,{dim:s}=a,i=r.shape.length,o=r.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),ft({inputs:{x:r},backend:n,attrs:{shape:o}})}var y8={kernelName:Ll,backendName:"cpu",kernelFunc:zh},b8=Vt((e,t)=>e/t),z0=nn(Ni,b8),rx={kernelName:Ni,backendName:"cpu",kernelFunc:z0};function OC(e,t,n){let a=e.shape,r=a[0],s=a[1],i=n.data.get(e.dataId),o=i.complexTensorInfos.real,l=i.complexTensorInfos.imag,u=[r,s],p=v.sizeFromShape(u),d=v.getTypedArrayFromDType("float32",p),c=v.getTypedArrayFromDType("float32",p);for(let g=0;g<r;g++){let y=ui({inputs:{x:o},backend:n,attrs:{begin:[g,0],size:[1,s]}}),b=ui({inputs:{x:l},backend:n,attrs:{begin:[g,0],size:[1,s]}}),x=Yn({inputs:{real:y,imag:b},backend:n}),{real:w,imag:I}=x8(x,t,n),T=N.mergeRealAndImagArrays(w,I);for(let C=0;C<s;C++){let E=N.getComplexWithIndex(T,C);d[g*s+C]=E.real,c[g*s+C]=E.imag}n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(b),n.disposeIntermediateTensorInfo(x)}let h=n.makeTensorInfo(u,"float32",d),m=n.makeTensorInfo(u,"float32",c),f=Yn({inputs:{real:h,imag:m},backend:n});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}function x8(e,t,n){let a=v.sizeFromShape(e.shape),r=n.data.get(e.dataId),s=n.data.get(r.complexTensorInfos.real.dataId).values,i=n.data.get(r.complexTensorInfos.imag.dataId).values;if(v8(a)){let o=sx(s,i,a,t,n),l=[e.shape[0],e.shape[1]];if(t){let u=n.makeTensorInfo(l,"float32",o.real),p=n.makeTensorInfo(l,"float32",o.imag),d=n.makeTensorInfo([],"float32",v.createScalarValue(a,"float32")),c=lr({inputs:{x:d},backend:n}),h=rx.kernelFunc({inputs:{a:u,b:d},backend:n}),m=rx.kernelFunc({inputs:{a:p,b:c},backend:n}),f=n.data.get(h.dataId).values,g=n.data.get(m.dataId).values;return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),{real:f,imag:g}}return o}else{let o=N.mergeRealAndImagArrays(s,i),l=w8(o,a,t);return N.splitRealAndImagArrays(l)}}function v8(e){return(e&e-1)===0}function sx(e,t,n,a,r){if(n===1)return{real:e,imag:t};let s=N.mergeRealAndImagArrays(e,t),i=n/2,o=N.complexWithEvenIndex(s),l=o.real,u=o.imag,p=[l.length],d=r.makeTensorInfo(p,"float32",l),c=r.makeTensorInfo(p,"float32",u),h=Yn({inputs:{real:d,imag:c},backend:r}),m=N.complexWithOddIndex(s),f=m.real,g=m.imag,y=[f.length],b=r.makeTensorInfo(y,"float32",f),x=r.makeTensorInfo(y,"float32",g),w=Yn({inputs:{real:b,imag:x},backend:r}),I=sx(l,u,i,a,r),T=I.real,C=I.imag,E=[T.length],A=r.makeTensorInfo(E,"float32",T),R=r.makeTensorInfo(E,"float32",C),F=Yn({inputs:{real:A,imag:R},backend:r}),S=sx(f,g,i,a,r),M=S.real,B=S.imag,U=[M.length],G=r.makeTensorInfo(U,"float32",M),q=r.makeTensorInfo(U,"float32",B),K=Yn({inputs:{real:G,imag:q},backend:r}),Z=N.exponents(n,a),Q=[Z.real.length],ee=r.makeTensorInfo(Q,"float32",Z.real),ae=r.makeTensorInfo(Q,"float32",Z.imag),te=Yn({inputs:{real:ee,imag:ae},backend:r}),le=Ff({inputs:{a:te,b:K},backend:r}),ie=gl({inputs:{a:F,b:le},backend:r}),ye=O0({inputs:{a:F,b:le},backend:r}),ue=li({inputs:{input:ie},backend:r}),xe=li({inputs:{input:ye},backend:r}),Ie=yl({inputs:{input:ie},backend:r}),Se=yl({inputs:{input:ye},backend:r}),Le=bl({inputs:[ue,xe],backend:r,attrs:{axis:0}}),Ve=bl({inputs:[Ie,Se],backend:r,attrs:{axis:0}}),nt=r.data.get(Le.dataId).values,st=r.data.get(Ve.dataId).values;return r.disposeIntermediateTensorInfo(d),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(b),r.disposeIntermediateTensorInfo(x),r.disposeIntermediateTensorInfo(w),r.disposeIntermediateTensorInfo(A),r.disposeIntermediateTensorInfo(R),r.disposeIntermediateTensorInfo(F),r.disposeIntermediateTensorInfo(G),r.disposeIntermediateTensorInfo(q),r.disposeIntermediateTensorInfo(K),r.disposeIntermediateTensorInfo(ee),r.disposeIntermediateTensorInfo(ae),r.disposeIntermediateTensorInfo(te),r.disposeIntermediateTensorInfo(le),r.disposeIntermediateTensorInfo(ie),r.disposeIntermediateTensorInfo(ye),r.disposeIntermediateTensorInfo(ue),r.disposeIntermediateTensorInfo(Ie),r.disposeIntermediateTensorInfo(xe),r.disposeIntermediateTensorInfo(Se),r.disposeIntermediateTensorInfo(Le),r.disposeIntermediateTensorInfo(Ve),{real:nt,imag:st}}function w8(e,t,n){let a=new Float32Array(t*2);for(let r=0;r<t;r++){let s=0,i=0;for(let o=0;o<t;o++){let l=N.exponent(r*o,t,n),u=N.getComplexWithIndex(e,o);s+=u.real*l.real-u.imag*l.imag,i+=u.real*l.imag+u.imag*l.real}n&&(s/=t,i/=t),N.assignToTypedArray(a,s,i,r)}return a}function k8(e){let{inputs:t,backend:n}=e,{input:a}=t,r=v.sizeFromShape(a.shape),s=a.shape[a.shape.length-1],i=r/s,o=ft({inputs:{x:a},backend:n,attrs:{shape:[i,s]}}),l=OC(o,!1,n),u=ft({inputs:{x:l},backend:n,attrs:{shape:a.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(l),u}var I8={kernelName:um,backendName:"cpu",kernelFunc:k8};function W0(e){let{backend:t,attrs:n}=e,{shape:a,value:r,dtype:s}=n,i=s||v.inferDtype(r),o=v.getArrayFromDType(i,v.sizeFromShape(a));return T8(o,r,i),t.makeTensorInfo(a,i,o)}var S8={kernelName:pc,backendName:"cpu",kernelFunc:W0};function T8(e,t,n){e.fill(t)}var N8={kernelName:Wl,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,r=n,s=v.getTypedArrayFromDType(a.dtype,v.sizeFromShape(a.shape)),[i,o,l,u]=a.shape,p=r.data.get(a.dataId).values;for(let d=0;d<i;d++){let c=d*l*o*u;for(let h=0;h<o;h++){let m=h*(l*u);for(let f=0;f<l;f++){let g=f*u;for(let y=0;y<u;y++){let b=Math.round(l-f-1),x=c+m+g+y,w=p[x];if(b>=0&&b<l){let I=b*u,T=c+m+I+y;w=p[T]}s[x]=w}}}}return{dataId:r.write(s,a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},C8=Vt((e,t)=>Math.floor(e/t)),_8=nn(Ai,C8,null,"int32"),E8={kernelName:Ai,backendName:"cpu",kernelFunc:_8};function A8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=MC({inputs:{x:r,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c}});if(i){let g=f;if(p==="NCHW"&&i.shape.length===1&&i.shape[0]!==1){let y=ft({inputs:{x:i},backend:n,attrs:{shape:[i.shape[0],1,1]}});f=gl({inputs:{a:f,b:y},backend:n}),n.disposeIntermediateTensorInfo(y)}else f=gl({inputs:{a:f,b:i},backend:n});n.disposeIntermediateTensorInfo(g)}if(h){let g=f;if(p==="NCHW"&&h==="prelu"&&o.shape.length===1&&o.shape[0]!==1){let y=ft({inputs:{x:o},backend:n,attrs:{shape:[o.shape[0],1,1]}});f=Lh(n,f,h,y,m),n.disposeIntermediateTensorInfo(y)}else f=Lh(n,f,h,o,m);n.disposeIntermediateTensorInfo(g)}return f}var $8={kernelName:Qs,backendName:"cpu",kernelFunc:A8};function F8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=PC({inputs:{x:r,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c}});if(i){let g=f;f=gl({inputs:{a:f,b:i},backend:n}),n.disposeIntermediateTensorInfo(g)}if(h){let g=f;f=Lh(n,f,h,o,m),n.disposeIntermediateTensorInfo(g)}return f}var D8={kernelName:ei,backendName:"cpu",kernelFunc:F8};function R8(e){let{inputs:t,backend:n}=e,{params:a,indices:r}=t,s=v.sizeFromShape(a.shape),i=r.shape,o=i[i.length-1],[l,u,p,d]=N.prepareAndValidate(a,r);if(u===0)return n.makeTensorInfo(l,a.dtype,[]);let c=n.data.get(r.dataId).values,h=n.bufferSync(a),m=Q2(c,h,a.dtype,u,o,p,d,a.shape,s);return n.makeTensorInfo(l,a.dtype,m.values)}var M8={kernelName:Vl,backendName:"cpu",kernelFunc:R8};function P8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,indices:s}=t,{axis:i,batchDims:o}=a;ge([r,s],"gatherV2");let l=v.parseAxisParam(i,r.shape)[0],u=n.data.get(s.dataId).values,p=r.shape[l];for(let w=0;w<u.length;++w){let I=u[w];v.assert(I<=p-1&&I>=0,()=>`GatherV2: the index value ${I} is not in [0, ${p-1}]`)}let d=o;o==null&&(d=0);let c=v.sizeFromShape(s.shape),h=N.segment_util.collectGatherOpShapeInfo(r,s,l,d),m=ft({inputs:{x:r},backend:n,attrs:{shape:[h.batchSize,h.outerSize,h.dimSize,h.sliceSize]}}),f=ft({inputs:{x:s},backend:n,attrs:{shape:[h.batchSize,c/h.batchSize]}}),g=[h.batchSize,h.outerSize,c/h.batchSize,h.sliceSize],y=n.bufferSync(f),b=n.bufferSync(m),x=eC(b,y,g);return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),n.makeTensorInfo(h.outputShape,x.dtype,x.values)}var O8={kernelName:Bl,backendName:"cpu",kernelFunc:P8};function L8(e){let{inputs:t,backend:n}=e,{input:a}=t,r=v.sizeFromShape(a.shape),s=a.shape[a.shape.length-1],i=r/s,o=ft({inputs:{x:a},backend:n,attrs:{shape:[i,s]}}),l=OC(o,!0,n),u=ft({inputs:{x:l},backend:n,attrs:{shape:a.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(l),u}var z8={kernelName:pm,backendName:"cpu",kernelFunc:L8},W8=rt(Gl,e=>Number.isFinite(e)?1:0,"bool"),B8={kernelName:Gl,backendName:"cpu",kernelFunc:W8},V8=rt(Hl,e=>Math.abs(e)===1/0?1:0,"bool"),U8={kernelName:Hl,backendName:"cpu",kernelFunc:V8},G8=rt(jl,e=>Number.isNaN(e)?1:0,"bool"),H8={kernelName:jl,backendName:"cpu",kernelFunc:G8};function j8(e){let{backend:t,attrs:n}=e,{start:a,stop:r,num:s}=n,i=sC(a,r,s);return t.makeTensorInfo([i.length],"float32",i)}var q8={kernelName:dm,backendName:"cpu",kernelFunc:j8},K8=rt(Xl,e=>Math.log1p(e)),X8={kernelName:Xl,backendName:"cpu",kernelFunc:K8},Y8=Vt((e,t)=>e&&t),Z8=nn(Yl,Y8,null,"bool"),J8={kernelName:Yl,backendName:"cpu",kernelFunc:Z8},Q8=rt(Zl,e=>e?0:1,"bool"),eK={kernelName:Zl,backendName:"cpu",kernelFunc:Q8},tK=Vt((e,t)=>e||t),nK=nn(Jl,tK,null,"bool"),aK={kernelName:Jl,backendName:"cpu",kernelFunc:nK};function rK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=a;ge(r,"LRN");let u=r.shape[3],p=u-1,d=n.data.get(r.dataId).values,c=v.sizeFromShape(r.shape),h=new Float32Array(c);function m(f){let g=f%u,y=f-g+Math.max(0,g-s),b=f-g+Math.min(g+s,p),x=0;for(;y<=b;y++){let w=d[y];x+=w*w}return x}for(let f=0;f<c;f++){let g=m(f),y=d[f]*Math.pow(i+o*g,-l);h[f]=y}return n.makeTensorInfo(r.shape,r.dtype,h)}var sK={kernelName:cc,backendName:"cpu",kernelFunc:rK};function iK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:p}=a;ge(i,"LRNGrad");let d=v.sizeFromShape(i.shape),c=i.shape[3],h=n.data.get(i.dataId).values,m=n.data.get(r.dataId).values,f=n.data.get(s.dataId).values,g=new Float32Array(d),y=d;for(let b=0;b<y;b++){let x=b%c,w=b-x+Math.max(0,x-o),I=b-x+Math.min(c,x+o+1),T=0;for(let C=w;C<I;C++)T+=Math.pow(m[C],2);T=u*T+l;for(let C=w;C<I;C++){let E=-2*u*p*m[C]*f[b]/T;b===C&&(E+=Math.pow(T,-p)),E*=h[b],g[C]+=E}}return n.makeTensorInfo(i.shape,r.dtype,g)}var oK={kernelName:hm,backendName:"cpu",kernelFunc:iK};function LC(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reductionIndices:s,keepDims:i}=a,o=n,l=r.shape,u=l.length,p=v.parseAxisParam(s,l),d=p,c=N.getAxesPermutation(d,u),h=o.data.get(r.dataId).values;if(c!=null){let w=new Array(u);for(let I=0;I<w.length;I++)w[I]=l[c[I]];h=$0(h,l,r.dtype,c,w),d=N.getInnerMostAxes(d.length,u),l=w}ge(r,"max"),N.assertAxesAreInnerMostDims("max",d,u);let[m,f]=N.computeOutAndReduceShapes(l,d),g=v.sizeFromShape(f),y=oC(h,g,m,r.dtype),b=o.write(y,m,r.dtype),x=m;return i&&(x=N.expandShapeToKeepDim(m,p)),{dataId:b,shape:x,dtype:r.dtype}}var lK={kernelName:Pi,backendName:"cpu",kernelFunc:LC};function uK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;ge(r,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l),d;if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))d=lr({inputs:{x:r},backend:n});else{let c=n.data.get(r.dataId).values,h=v.computeStrides(r.shape),m=L0(c,r.shape,r.dtype,h,p,"max");d=n.makeTensorInfo(p.outShape,r.dtype,m.values)}return d}var pK={kernelName:Li,backendName:"cpu",kernelFunc:uK};function cK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a;ge(r,"maxPool3d");let p=N.computePool3DInfo(r.shape,s,i,1,o,l,u),d=n.data.get(r.dataId).values,c=RC(d,r.shape,r.dtype,v.computeStrides(r.shape),p,"max");return n.makeTensorInfo(c.shape,"float32",c.values)}var dK={kernelName:dc,backendName:"cpu",kernelFunc:cK};function hK(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=a;ge([r,s],"maxPool3DGrad");let p=N.computePool3DInfo(s.shape,i,o,1,l,u),d=n.bufferSync(s),c=e5(d,p),h=p.strideDepth,m=p.strideHeight,f=p.strideWidth,g=p.dilationDepth,y=p.dilationHeight,b=p.dilationWidth,x=p.effectiveFilterDepth,w=p.effectiveFilterHeight,I=p.effectiveFilterWidth,T=x-1-p.padInfo.front,C=I-1-p.padInfo.left,E=w-1-p.padInfo.top,A=Pe(s.shape,"float32"),R=n.bufferSync(r);for(let F=0;F<p.batchSize;++F)for(let S=0;S<p.inChannels;++S)for(let M=0;M<p.inDepth;++M)for(let B=0;B<p.inHeight;++B)for(let U=0;U<p.inWidth;++U){let G=M-T,q=B-E,K=U-C,Z=0;for(let Q=0;Q<x;Q+=g){let ee=(G+Q)/h;if(!(ee<0||ee>=p.outDepth||Math.floor(ee)!==ee))for(let ae=0;ae<w;ae+=y){let te=(q+ae)/m;if(!(te<0||te>=p.outHeight||Math.floor(te)!==te))for(let le=0;le<I;le+=b){let ie=(K+le)/f;if(ie<0||ie>=p.outWidth||Math.floor(ie)!==ie)continue;let ye=x*w*I-1-c.get(F,ee,te,ie,S),ue=Q*w*I+ae*I+le,xe=ye===ue?1:0;xe!==0&&(Z+=R.get(F,ee,te,ie,S)*xe)}}}A.set(Z,F,M,B,U,S)}return n.makeTensorInfo(A.shape,A.dtype,A.values)}var mK={kernelName:fm,backendName:"cpu",kernelFunc:hK};function fK(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s,output:i}=t,o=s;ge([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:p,dimRoundingMode:d}=a,c=N.computePool2DInfo(o.shape,l,u,1,p,d),h=n.data.get(o.dataId).values,m=Pe(c.outShape,o.dtype,DC(h,o.shape,o.dtype,c).values),f=c.strideHeight,g=c.strideWidth,y=c.dilationHeight,b=c.dilationWidth,x=c.effectiveFilterHeight,w=c.effectiveFilterWidth,I=w-1-c.padInfo.left,T=x-1-c.padInfo.top,C=Pe(o.shape,"float32"),E=n.data.get(r.dataId).values,A=Pe(r.shape,"float32",E);for(let R=0;R<c.batchSize;++R)for(let F=0;F<c.inChannels;++F)for(let S=0;S<c.inHeight;++S)for(let M=0;M<c.inWidth;++M){let B=S-T,U=M-I,G=0;for(let q=0;q<x;q+=y){let K=(B+q)/f;if(!(K<0||K>=c.outHeight||Math.floor(K)!==K))for(let Z=0;Z<w;Z+=b){let Q=(U+Z)/g;if(Q<0||Q>=c.outWidth||Math.floor(Q)!==Q)continue;let ee=x*w-1-m.get(R,K,Q,F),ae=q*w+Z,te=ee===ae?1:0;te!==0&&(G+=A.get(R,K,Q,F)*te)}}C.set(G,R,S,M,F)}return n.makeTensorInfo(C.shape,C.dtype,C.values)}var gK={kernelName:mm,backendName:"cpu",kernelFunc:fK};function yK(e,t,n,a,r){let s=v.computeStrides(t),i=L0(e,t,n,s,r,"max"),o=DC(e,t,n,r,!0,a);return[i.values,o.values]}var bK={kernelName:gm,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{filterSize:r,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;ge(a,"MaxPoolWithArgmax");let u=l.data.get(a.dataId).values,p=N.computePool2DInfo(a.shape,r,s,[1,1],i),[d,c]=yK(u,a.shape,a.dtype,o,p),h=l.write(d,p.outShape,a.dtype),m=l.write(c,p.outShape,a.dtype);return[{dataId:h,shape:p.outShape,dtype:a.dtype},{dataId:m,shape:p.outShape,dtype:"int32"}]}};function xK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=v.parseAxisParam(s,r.shape),l=N.computeOutAndReduceShapes(r.shape,o)[1],u=v.sizeFromShape(l),p=[],d=n.makeTensorInfo([],"float32",new Float32Array([u]));p.push(d);let c=os({inputs:{x:r},backend:n,attrs:{dtype:"float32"}});p.push(c);let h=z0({inputs:{a:c,b:d},backend:n});p.push(h);let m=Qc({inputs:{x:h},backend:n,attrs:{axis:s,keepDims:i}});return p.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var vK={kernelName:zi,backendName:"cpu",kernelFunc:xK};function wK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"min");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("min",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];(Number.isNaN(I)||I<x)&&(x=I)}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var kK={kernelName:Wi,backendName:"cpu",kernelFunc:wK};function IK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,mode:i}=a;ge(r,"mirrorPad");let o=s.map((b,x)=>b[0]+r.shape[x]+b[1]),l=s.map(b=>b[0]),u=s.map((b,x)=>b[0]+r.shape[x]),p=i==="reflect"?0:1,d=n.data.get(r.dataId).values,c=r.shape.length,h=v.computeStrides(r.shape),m=v.sizeFromShape(o),f=o.length,g=v.computeStrides(o),y=v.getTypedArrayFromDType(r.dtype,m);for(let b=0;b<m;b++){let x=v.indexToLoc(b,f,g);for(let I=0;I<f;I++)x[I]<l[I]?x[I]=l[I]*2-x[I]-p:x[I]>=u[I]&&(x[I]=(u[I]-1)*2-x[I]+p);x=x.map((I,T)=>I-l[T]);let w=v.locToIndex(x,c,h);y[b]=d[w]}return{dataId:n.write(y,o,r.dtype),shape:o,dtype:r.dtype}}var SK={kernelName:Vi,backendName:"cpu",kernelFunc:IK},TK=Vt((e,t)=>{let n=e%t;return e<0&&t<0||e>=0&&t>=0?n:(n+t)%t}),NK=nn(Ql,TK),CK={kernelName:Ql,backendName:"cpu",kernelFunc:NK},_K=us(Hh());function zC(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{dim:s}=a,i=r.shape.length,o=s;if(o===-1&&(o=i-1),o!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);let l=v.parseAxisParam([o],r.shape),u=LC({inputs:{x:r},backend:n,attrs:{reductionIndices:l,keepDims:!1}}),p=N.expandShapeToKeepDim(u.shape,l),d=ft({inputs:{x:u},backend:n,attrs:{shape:p}}),c=O0({inputs:{a:r,b:d},backend:n}),h=Y2({inputs:{x:c},backend:n}),m=Qc({inputs:{x:h},backend:n,attrs:{axis:l,keepDims:!1}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:p}}),g=z0({inputs:{a:h,b:f},backend:n});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),g}var EK={kernelName:io,backendName:"cpu",kernelFunc:zC};function AK(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{numSamples:s,seed:i,normalized:o}=a;ge(r,"multinomial");let l=o?r:zC({inputs:{logits:r},backend:n,attrs:{dim:-1}}),u=l.shape[0],p=l.shape[1],d=n.data.get(l.dataId).values,c=[u,s],h=v.makeZerosTypedArray(v.sizeFromShape(c),"int32");for(let m=0;m<u;++m){let f=m*p,g=new Float32Array(p-1);g[0]=d[f];for(let x=1;x<g.length;++x)g[x]=g[x-1]+d[f+x];let y=_K.alea(i.toString()),b=m*s;for(let x=0;x<s;++x){let w=y();h[b+x]=g.length;for(let I=0;I<g.length;I++)if(w<g[I]){h[b+x]=I;break}}}return o||n.disposeIntermediateTensorInfo(l),n.makeTensorInfo(c,"int32",h)}var $K={kernelName:ym,backendName:"cpu",kernelFunc:AK},FK=cr.nonMaxSuppressionV3Impl;function DK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=a;ge(r,"NonMaxSuppression");let u=n.data.get(r.dataId).values,p=n.data.get(s.dataId).values,{selectedIndices:d}=FK(u,p,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}var RK={kernelName:nu,backendName:"cpu",kernelFunc:DK},MK=cr.nonMaxSuppressionV4Impl;function PK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=a;ge(r,"NonMaxSuppressionPadded");let p=n.data.get(r.dataId).values,d=n.data.get(s.dataId).values,{selectedIndices:c,validOutputs:h}=MK(p,d,i,o,l,u);return[n.makeTensorInfo([c.length],"int32",new Int32Array(c)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var OK={kernelName:au,backendName:"cpu",kernelFunc:PK},LK=cr.nonMaxSuppressionV5Impl;function zK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=a;ge(r,"NonMaxSuppressionWithScore");let p=n.data.get(r.dataId).values,d=n.data.get(s.dataId).values,c=i,h=o,m=l,f=u,{selectedIndices:g,selectedScores:y}=LK(p,d,c,h,m,f);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var WK={kernelName:ru,backendName:"cpu",kernelFunc:zK};function BK(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a;ge(r,"oneHot");let u=v.sizeFromShape(r.shape),p=new Float32Array(u*i);p.fill(l);let d=n.data.get(r.dataId).values;for(let c=0;c<u;++c)d[c]>=0&&d[c]<i&&(p[c*i+d[c]]=o);return n.makeTensorInfo([...r.shape,i],s,p)}var VK={kernelName:Gi,backendName:"cpu",kernelFunc:BK};function Wh(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("zerosLike is not supported for string tensors");if(a.dtype==="complex64"){let r=li({inputs:{input:a},backend:n}),s=Wh({inputs:{x:r},backend:n}),i=yl({inputs:{input:a},backend:n}),o=Wh({inputs:{x:i},backend:n}),l=Yn({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return W0({backend:n,attrs:{shape:a.shape,value:0,dtype:a.dtype}})}var UK={kernelName:Iu,backendName:"cpu",kernelFunc:Wh};function WC(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("onesLike is not supported for string tensors");if(a.dtype==="complex64"){let r=li({inputs:{input:a},backend:n}),s=WC({inputs:{x:r},backend:n}),i=yl({inputs:{input:a},backend:n}),o=Wh({inputs:{x:i},backend:n}),l=Yn({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return W0({backend:n,attrs:{shape:a.shape,value:1,dtype:a.dtype}})}var GK={kernelName:su,backendName:"cpu",kernelFunc:WC};function BC(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return zh({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=zh({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=bl({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeIntermediateTensorInfo(p)),u}var HK={kernelName:iu,backendName:"cpu",kernelFunc:BC};function jK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,constantValue:i}=a;ge(r,"pad");let o=s.map((y,b)=>y[0]+r.shape[b]+y[1]),l=s.map(y=>y[0]),u=n.data.get(r.dataId).values,p=v.sizeFromShape(r.shape),d=r.shape.length,c=v.computeStrides(r.shape),h=v.sizeFromShape(o),m=o.length,f=v.computeStrides(o),g=v.getTypedArrayFromDType(r.dtype,h);i!==0&&g.fill(i);for(let y=0;y<p;y++){let b=v.indexToLoc(y,d,c).map((w,I)=>w+l[I]),x=v.locToIndex(b,m,f);g[x]=u[y]}return{dataId:n.write(g,o,r.dtype),shape:o,dtype:r.dtype}}var VC={kernelName:Hi,backendName:"cpu",kernelFunc:jK},qK=Vt((e,t)=>Math.pow(e,t)),KK=nn(ji,qK),XK={kernelName:ji,backendName:"cpu",kernelFunc:KK};function YK(e){let{inputs:t,backend:n,attrs:a}=e,{paramsNestedSplits:r,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=a,l=r.map(y=>n.data.get(y.dataId).values),u=r.map(y=>y.shape),p=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,[c,h,m]=hC(l,u,p,s.shape,s.dtype,d,i.shape,o),f=c.map(y=>n.makeTensorInfo([y.length],"int32",y)),g=n.makeTensorInfo(m,s.dtype,h);return f.concat([g])}var ZK={kernelName:bm,backendName:"cpu",kernelFunc:YK};function JK(e){let{inputs:t,backend:n}=e,{starts:a,limits:r,deltas:s}=t,i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=mC(i,a.shape,a.dtype,o,r.shape,l,s.shape),d=n.makeTensorInfo([u.length],"int32",u),c=n.makeTensorInfo([p.length],a.dtype,p);return[d,c]}var QK={kernelName:xm,backendName:"cpu",kernelFunc:JK};function eX(e){let{inputs:t,backend:n,attrs:a}=e,{shape:r,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=a,u=n.data.get(r.dataId).values,p=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,c=o.map(g=>n.data.get(g.dataId).values),h=o.map(g=>g.shape),[m,f]=fC(u,r.shape,p,s.shape,s.dtype,d,i.shape,c,h,l);return n.makeTensorInfo(m,s.dtype,f)}var tX={kernelName:vm,backendName:"cpu",kernelFunc:eX};function nX(e){let{backend:t,attrs:n}=e,{start:a,stop:r,dtype:s,step:i}=n,o=F0(a,r,i,s);return t.makeTensorInfo([o.length],s,o)}var aX={kernelName:hc,backendName:"cpu",kernelFunc:nX},rX=rt(ou,e=>1/e),sX={kernelName:ou,backendName:"cpu",kernelFunc:rX};function iX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a;ge(r,"resizeBilinear");let l=v.computeStrides(r.shape),[u,p]=o,[d,c,h,m]=r.shape,f=n.data.get(r.dataId).values,g=new Float32Array(v.sizeFromShape([d,u,p,m])),y=[s&&u>1?c-1:c,s&&p>1?h-1:h],b=[s&&u>1?u-1:u,s&&p>1?p-1:p],x=0,w=y[0]/b[0],I=y[1]/b[1];for(let T=0;T<d;T++)for(let C=0;C<u;C++){let E;i?E=w*(C+.5)-.5:E=w*C;let A=Math.max(0,Math.floor(E)),R=E-A,F=Math.min(c-1,Math.ceil(E)),S=T*l[0]+A*l[1],M=T*l[0]+F*l[1];for(let B=0;B<p;B++){let U;i?U=I*(B+.5)-.5:U=I*B;let G=Math.max(0,Math.floor(U)),q=U-G,K=Math.min(h-1,Math.ceil(U)),Z=S+G*l[2],Q=M+G*l[2],ee=S+K*l[2],ae=M+K*l[2];for(let te=0;te<m;te++){let le=f[Z+te],ie=f[Q+te],ye=f[ee+te],ue=f[ae+te],xe=le+(ye-le)*q,Ie=ie+(ue-ie)*q,Se=xe+(Ie-xe)*R;g[x++]=Se}}}return n.makeTensorInfo([d,u,p,m],"float32",g)}var oX={kernelName:Zi,backendName:"cpu",kernelFunc:iX};function lX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a;ge([s,r],"resizeBilinearGrad");let o=v.computeStrides(r.shape),[l,u,p,d]=r.shape,[,c,h]=s.shape,m=new Float32Array(l*u*p*d),f=[i&&c>1?u-1:u,i&&h>1?p-1:p],g=[i&&c>1?c-1:c,i&&h>1?h-1:h],y=f[0]/g[0],b=f[1]/g[1],x=n.data.get(s.dataId).values,w=0;for(let I=0;I<l;I++){let T=I*o[0];for(let C=0;C<c;C++){let E=C*y,A=Math.floor(E),R=Math.min(Math.ceil(E),u-1),F=T+A*o[1],S=T+R*o[1],M=E-A,B=1-M;for(let U=0;U<h;U++){let G=U*b,q=Math.floor(G),K=Math.min(Math.ceil(G),p-1),Z=G-q,Q=1-Z,ee=F+q*o[2],ae=F+K*o[2],te=S+q*o[2],le=S+K*o[2],ie=B*Q,ye=B*Z,ue=M*Q,xe=M*Z;for(let Ie=0;Ie<d;Ie++){let Se=x[w++];m[ee+Ie]+=Se*ie,m[ae+Ie]+=Se*ye,m[te+Ie]+=Se*ue,m[le+Ie]+=Se*xe}}}}return n.makeTensorInfo([l,p,u,d],"float32",m)}var uX={kernelName:Im,backendName:"cpu",kernelFunc:lX};function pX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a;ge(r,"resizeNearestNeighbor");let l=v.computeStrides(r.shape),[u,p]=o,[d,c,h,m]=r.shape,f=n.data.get(r.dataId).values,g=new Float32Array(d*u*p*m),y=[s&&u>1?c-1:c,s&&p>1?h-1:h],b=[s&&u>1?u-1:u,s&&p>1?p-1:p],x=y[0]/b[0],w=y[1]/b[1],I=0;for(let T=0;T<d;T++){let C=T*l[0];for(let E=0;E<u;E++){let A=i?x*(E+.5):x*E,R=Math.min(c-1,s?Math.round(A):Math.floor(A));i&&(R=Math.max(0,R));let F=C+R*l[1];for(let S=0;S<p;S++){let M=i?w*(S+.5):w*S,B=Math.min(h-1,s?Math.round(M):Math.floor(M));i&&(B=Math.max(0,B));let U=F+B*l[2];for(let G=0;G<m;G++){let q=f[U+G];g[I++]=q}}}}return n.makeTensorInfo([d,u,p,m],r.dtype,g)}var cX={kernelName:Yi,backendName:"cpu",kernelFunc:pX};function dX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a;ge([s,r],"resizeNearestNeighborGrad");let o=v.computeStrides(r.shape),l=v.computeStrides(s.shape),[u,p,d,c]=r.shape,[,h,m]=s.shape,f=new Float32Array(u*p*d*c),g=n.data.get(s.dataId).values,y=[i&&h>1?p-1:p,i&&m>1?d-1:d],b=[i&&h>1?h-1:h,i&&m>1?m-1:m],x=y[0]/b[0],w=y[1]/b[1],I=1/x,T=1/w,C=Math.ceil(I)*2+2,E=Math.ceil(T)*2+2;for(let A=0;A<u;A++){let R=A*o[0];for(let F=0;F<p;F++){let S=R+F*o[1],M=Math.floor(F*I),B=Math.floor(M-C/2);for(let U=0;U<d;U++){let G=S+U*o[2],q=Math.floor(U*T),K=Math.floor(q-E/2);for(let Z=0;Z<c;Z++){let Q=0;for(let ee=0;ee<C;ee++){let ae=ee+B;if(ae<0||ae>=h)continue;let te=R+ae*l[1],le=ae*x,ie=Math.min(p-1,i?Math.round(le):Math.floor(le));if(F===ie)for(let ye=0;ye<E;ye++){let ue=ye+K;if(ue<0||ue>=m)continue;let xe=te+ue*l[2],Ie=ue*w,Se=Math.min(d-1,i?Math.round(Ie):Math.floor(Ie));U===Se&&(Q+=g[xe+Z])}}f[G+Z]=Q}}}}return n.makeTensorInfo(r.shape,r.dtype,f)}var hX={kernelName:km,backendName:"cpu",kernelFunc:dX};function mX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a;ge(r,"reverse");let i=r.shape.length,o=v.parseAxisParam(s,r.shape);if(i===0)return lr({inputs:{x:r},backend:n});let l=new Ht(r.shape,r.dtype),u=n.bufferSync(r);for(let p=0;p<l.size;p++){let d=l.indexToLoc(p),c=d.slice();o.forEach(h=>c[h]=r.shape[h]-1-c[h]),l.set(u.get(...c),...d)}return n.makeTensorInfo(l.shape,l.dtype,l.values)}var fX={kernelName:Qi,backendName:"cpu",kernelFunc:mX},gX={kernelName:Su,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,{radians:r,fillValue:s,center:i}=t,o=n,l=v.getTypedArrayFromDType(a.dtype,v.sizeFromShape(a.shape)),[u,p,d,c]=a.shape,[h,m]=N.getImageCenter(i,p,d),f=255,g=Math.sin(r),y=Math.cos(r),b=o.data.get(a.dataId).values;for(let x=0;x<u;x++){let w=x*d*p*c;for(let I=0;I<p;I++){let T=I*(d*c);for(let C=0;C<d;C++){let E=C*c;for(let A=0;A<c;A++){let R=[u,I,C,A],F=R[2],S=R[1],M=(F-h)*y-(S-m)*g,B=(F-h)*g+(S-m)*y;M=Math.round(M+h),B=Math.round(B+m);let U=s;if(typeof s!="number"&&(A===3?U=f:U=s[A]),M>=0&&M<d&&B>=0&&B<p){let q=B*(d*c),K=M*c,Z=w+q+K+A;U=b[Z]}let G=w+T+E+A;l[G]=U}}}}return{dataId:o.write(l,a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},yX=rt(eo,e=>{let t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2===0?t:t+1}),bX={kernelName:eo,backendName:"cpu",kernelFunc:yX};function xX(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r,updates:s}=t,{shape:i}=a,{sliceRank:o,numUpdates:l,sliceSize:u,strides:p,outputSize:d}=N.calculateShapes(s,r,i),c=!0,h=n.bufferSync(r),m=n.bufferSync(s),f=Yo(h,m,i,d,u,l,o,p,0,c);return n.makeTensorInfo(i,f.dtype,f.values)}var vX={kernelName:uu,backendName:"cpu",kernelFunc:xX};function wX(e,t){let n=0,a=e.length,r=0;for(;n<a;)r=Math.floor((n+a)/2),e[r]<t?n=r+1:a=r;return a}function kX(e,t){let n=0,a=e.length,r=0;for(;n<a;)r=Math.floor((n+a)/2),e[r]<=t?n=r+1:a=r;return a}function IX(e,t,n,a,r,s){let i=v.getArrayFromDType("int32",n*r);for(let o=0;o<n;++o){let l=e.slice(o*a,(o+1)*a),u=o*r;for(let p=0;p<r;++p)i[u+p]=s==="left"?wX(l,t[p+u]):kX(l,t[p+u])}return i}function SX(e){let{inputs:t,backend:n,attrs:a}=e,{sortedSequence:r,values:s}=t,{side:i}=a,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,u=IX(o,l,r.shape[0],r.shape[1],s.shape[1],i);return n.makeTensorInfo(s.shape,"int32",u)}var TX={kernelName:Sm,backendName:"cpu",kernelFunc:SX};function NX(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t;ge([a,r,s],"select");let i=a.shape.length,o=n.data.get(a.dataId).values,l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=ma(r.dtype,s.dtype),d=v.makeZerosTypedArray(v.sizeFromShape(r.shape),p),c=0,h=i===0||i>1||r.shape.length===1?1:v.sizeFromShape(r.shape.slice(1));for(let m=0;m<o.length;m++)for(let f=0;f<h;f++)o[m]===1?d[c++]=l[m]:d[c++]=u[m];return n.makeTensorInfo(r.shape,p,d)}var CX={kernelName:pu,backendName:"cpu",kernelFunc:NX},_X=N.SELU_SCALEALPHA,EX=N.SELU_SCALE,AX=rt(cu,e=>e>=0?EX*e:_X*(Math.exp(e)-1)),$X={kernelName:cu,backendName:"cpu",kernelFunc:AX},FX=rt(mu,e=>e<0?-1:e>0?1:0),DX={kernelName:mu,backendName:"cpu",kernelFunc:FX},RX=rt(no,e=>Math.sin(e)),MX={kernelName:no,backendName:"cpu",kernelFunc:RX},PX=rt(hu,e=>Math.sinh(e)),OX={kernelName:hu,backendName:"cpu",kernelFunc:PX},LX=11920928955078125e-23,Uk=Math.log(LX)+2,zX=rt(fu,e=>{let t=e>-Uk,n=e<Uk,a=Math.exp(e),r;return n?r=a:t?r=e:r=Math.log(1+a),r}),WX={kernelName:fu,backendName:"cpu",kernelFunc:zX};function BX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a;ge([r],"spaceToBatchND");let o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<r.shape.length;++g)l.push([0,0]);let u=VC.kernelFunc({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),p=N.getReshaped(u.shape,s,o,!1),d=N.getPermuted(p.length,s.length,!1),c=N.getReshapedPermuted(u.shape,s,o,!1),h=ft({inputs:{x:u},backend:n,attrs:{shape:p}}),m=Vn({inputs:{x:h},backend:n,attrs:{perm:d}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}var VX={kernelName:gu,backendName:"cpu",kernelFunc:BX};function UX(e){let{inputs:t,backend:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-        ${s.shape}`);if(a.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Values must be a vector, saw:
-        ${r.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let o=n.data.get(a.dataId).values,l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=n.data.get(i.dataId).values[0],[d,c,h,m,f]=bC(o,a.shape,a.dtype,l,r.dtype,u,p);return[n.makeTensorInfo(c,a.dtype,d),n.makeTensorInfo([c[0]],r.dtype,h),n.makeTensorInfo([m.length],"bool",new Uint8Array(m.map(g=>Number(g)))),n.makeTensorInfo([f.length],a.dtype,new Int32Array(f))]}var GX={kernelName:mc,backendName:"cpu",kernelFunc:UX};function HX(e){let{inputs:t,backend:n}=e,{inputIndices:a,inputShape:r,newShape:s}=t;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.data.get(r.dataId).values),o=n.data.get(a.dataId).values,l=Array.from(n.data.get(s.dataId).values),[u,p,d]=xC(o,a.shape,a.dtype,i,l);return[n.makeTensorInfo(p,a.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}var jX={kernelName:bu,backendName:"cpu",kernelFunc:HX};function qX(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-          ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-          ${s.shape}`);if(r.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=D0(i,a.shape,a.dtype,o,l,!0);return n.makeTensorInfo(p,a.dtype,u)}var KX={kernelName:fc,backendName:"cpu",kernelFunc:qX};function XX(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-         ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-         ${s.shape}`);if(r.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=D0(i,a.shape,a.dtype,o,l);return n.makeTensorInfo(p,a.dtype,u)}var YX={kernelName:gc,backendName:"cpu",kernelFunc:XX};function ZX(e){let{inputs:t,backend:n,attrs:a}=e,{sparseIndices:r,sparseValues:s,defaultValue:i}=t,{outputShape:o}=a,{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=N.calculateShapes(s,r,o),h=!1,m=n.bufferSync(r),f;switch(s.dtype){case"bool":{let g=n.bufferSync(s),y=Boolean(n.data.get(i.dataId).values[0]);f=Yo(m,g,o,c,p,u,l,d,y,h);break}case"float32":{let g=n.bufferSync(s),y=n.data.get(i.dataId).values[0];f=Yo(m,g,o,c,p,u,l,d,y,h);break}case"int32":{let g=n.bufferSync(s),y=n.data.get(i.dataId).values[0];f=Yo(m,g,o,c,p,u,l,d,y,h);break}case"string":{let g=n.bufferSync(s),y=v.decodeString(n.data.get(i.dataId).values[0]);f=Yo(m,g,o,c,p,u,l,d,y,h);break}default:throw new Error(`Unsupported type ${s.dtype}`)}return n.makeTensorInfo(o,f.dtype,f.values)}var JX={kernelName:Tm,backendName:"cpu",kernelFunc:ZX};function QX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=a,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=new Array(r.shape.length).fill(0),p=r.shape.slice();return l.map(d=>{let c=[...p];c[o]=d;let h=ui({inputs:{x:r},backend:n,attrs:{begin:u,size:c}});return u[o]+=d,h})}var eY={kernelName:yu,backendName:"cpu",kernelFunc:QX},tY={kernelName:yc,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{let{x:n}=e,a=t;ge(n,"square");let r=a.data.get(n.dataId).values,s=new Float32Array(r.length);for(let i=0;i<r.length;++i){let o=r[i];s[i]=o*o}return{dataId:a.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},nY=rt(ms,(e,t)=>{let n=t;return isNaN(e)?NaN:e>0?1:n.alpha}),aY={kernelName:ms,backendName:"cpu",kernelFunc:nY};function rY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a;ge(r,"stridedSlice");let{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=ft({inputs:{x:r},backend:n,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let T=jt.computeOutShape(b,x,w),C=ui({inputs:{x:r},backend:n,attrs:{begin:b,size:T}});I=ft({inputs:{x:C},backend:n,attrs:{shape:m}}),n.disposeIntermediateTensorInfo(C)}else{let T=n.bufferSync(r),C=wC(h,T,w,b);I=n.makeTensorInfo(m,C.dtype,C.values)}return I}var sY={kernelName:xu,backendName:"cpu",kernelFunc:rY};function iY(e){let{inputs:t,backend:n,attrs:a}=e,{separator:r,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=a,{data:p,dataSplits:d}=t,c=n.data.get(p.dataId).values,h=n.data.get(d.dataId).values,[m,f]=R0(c,h,r,s,i,o,l,u);return[n.makeTensorInfo([m.length],"string",m),n.makeTensorInfo(d.shape,"int32",f)]}var oY={kernelName:bc,backendName:"cpu",kernelFunc:iY};function lY(e){let{inputs:t,backend:n,attrs:a}=e,{skipEmpty:r}=a,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values[0],[u,p,d]=M0(o,l,r),c=p.length;return[n.makeTensorInfo([c,2],"int32",u),n.makeTensorInfo([c],"string",p),n.makeTensorInfo([2],"int32",new Int32Array(d))]}var uY={kernelName:xc,backendName:"cpu",kernelFunc:lY};function pY(e){let{inputs:t,backend:n,attrs:a}=e,{numBuckets:r}=a,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(r<=0)throw new Error("Number of buckets must be at least 1");let i=n.data.get(s.dataId).values,o=P0(i,r);return n.makeTensorInfo(s.shape,"int32",o)}var cY={kernelName:vc,backendName:"cpu",kernelFunc:pY},dY=rt(uo,e=>Math.tan(e)),hY={kernelName:uo,backendName:"cpu",kernelFunc:dY},mY=rt(po,e=>Math.tanh(e)),fY={kernelName:po,backendName:"cpu",kernelFunc:mY};function gY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reps:s}=a;ge(r,"tile");let i=IC(n.bufferSync(r),s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}var yY={kernelName:hs,backendName:"cpu",kernelFunc:gY};function bY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{k:s,sorted:i}=a;ge(r,"topk");let o=n.data.get(r.dataId).values,[l,u]=TC(o,r.shape,r.dtype,s,i);return[n.makeTensorInfo(l.shape,l.dtype,l.values),n.makeTensorInfo(u.shape,u.dtype,u.values)]}var xY={kernelName:vu,backendName:"cpu",kernelFunc:bY};function vY(e){let{inputs:t,attrs:n,backend:a}=e,{image:r,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=v.computeStrides(r.shape),b=y[0],x=y[1],w=y[2],I=v.computeStrides(g),T=I[0],C=I[1],E=I[2],A=v.getTypedArrayFromDType(r.dtype,v.sizeFromShape(g));A.fill(l);let R=a.data.get(r.dataId).values,F=a.data.get(s.dataId).values;for(let S=0;S<p;++S){let M=s.shape[0]===1?F:F.subarray(S*8,S*8+8);for(let B=0;B<m;++B)for(let U=0;U<f;++U)for(let G=0;G<h;++G){let q,K=M[6]*U+M[7]*B+1;if(K===0)continue;let Z=(M[0]*U+M[1]*B+M[2])/K,Q=(M[3]*U+M[4]*B+M[5])/K,ee=Gk(Z,c,o),ae=Gk(Q,d,o);switch(i){case"nearest":q=NY(R,d,c,b,x,w,S,ae,ee,G,l);break;case"bilinear":q=CY(R,d,c,b,x,w,S,ae,ee,G,l);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}let te=S*T+B*C+U*E+G;A[te]=q}return a.makeTensorInfo(g,r.dtype,A)}return{dataId:a.write(A,g,r.dtype),shape:r.shape,dtype:r.dtype}}var wY={kernelName:wu,backendName:"cpu",kernelFunc:vY};function Gk(e,t,n){switch(n){case"reflect":return kY(e,t);case"wrap":return IY(e,t);case"nearest":return TY(e,t);case"constant":default:return SY(e,t)}}function kY(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let a=2*t;n<a&&(n=a*Math.trunc(-n/a)+n),n=n<-t?n+a:-n-1}else if(n>t-1)if(t<=1)n=0;else{let a=2*t;n-=a*Math.trunc(n/a),n>=t&&(n=a-n-1)}return v.clamp(0,n,t-1)}function IY(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let a=t-1;n+=t*(Math.trunc(-n/a)+1)}else if(n>t-1)if(t<=1)n=0;else{let a=t-1;n-=t*Math.trunc(n/a)}return v.clamp(0,n,t-1)}function SY(e,t){return e}function TY(e,t){return v.clamp(0,e,t-1)}function Fp(e,t,n,a,r,s,i,o,l,u,p){let d=i*a+o*r+l*s+u;return 0<=o&&o<t&&0<=l&&l<n?e[d]:p}function NY(e,t,n,a,r,s,i,o,l,u,p){let d=Math.round(o),c=Math.round(l);return Fp(e,t,n,a,r,s,i,d,c,u,p)}function CY(e,t,n,a,r,s,i,o,l,u,p){let d=Math.floor(o),c=Math.floor(l),h=d+1,m=c+1,f=(m-l)*Fp(e,t,n,a,r,s,i,d,c,u,p)+(l-c)*Fp(e,t,n,a,r,s,i,d,m,u,p),g=(m-l)*Fp(e,t,n,a,r,s,i,h,c,u,p)+(l-c)*Fp(e,t,n,a,r,s,i,h,m,u,p);return(h-o)*f+(o-d)*g}function _Y(e){let{inputs:t,attrs:n,backend:a}=e,{axis:r}=n,{x:s}=t;ge(s,"unique");let i=a.data.get(s.dataId).values,{outputValues:o,outputShape:l,indices:u}=NC(i,r,s.shape,s.dtype);return[a.makeTensorInfo(l,s.dtype,o),a.makeTensorInfo([u.length],"int32",u)]}var EY={kernelName:Nm,backendName:"cpu",kernelFunc:_Y};function AY(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r.shape.length,o=r.shape[s],l=new Array(i-1),u=0;for(let h=0;h<i;h++)h!==s&&(l[u++]=r.shape[h]);let p=new Array(i).fill(0),d=r.shape.slice();d[s]=1;let c=new Array(o);for(let h=0;h<c.length;h++){p[s]=h;let m=ui({inputs:{x:r},backend:n,attrs:{begin:p,size:d}});c[h]=ft({inputs:{x:m},backend:n,attrs:{shape:l}}),n.disposeIntermediateTensorInfo(m)}return c}var $Y={kernelName:ku,backendName:"cpu",kernelFunc:AY};function FY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,segmentIds:s}=t,{numSegments:i}=a;ge(r,"unsortedSegmentSum");let o=r.shape.length,l=s.shape.length,u=[],p=[],d=o-l,c=s;for(let m=0;m<d;++m){let f=zh({inputs:{input:c},backend:n,attrs:{dim:m+1}});c=f,p.push(f)}for(let m=0;m<i;++m){let f=v.createScalarValue(m,"int32"),g=n.makeTensorInfo([],"int32",f),y=K2({inputs:{a:g,b:c},backend:n}),b=os({inputs:{x:y},backend:n,attrs:{dtype:"float32"}}),x=Ff({inputs:{a:b,b:r},backend:n}),w=Qc({inputs:{x},backend:n,attrs:{axis:0,keepDims:!1}});u.push(w),p.push(g),p.push(y),p.push(b),p.push(x),p.push(w)}let h=BC({inputs:u,backend:n,attrs:{axis:0}});return p.forEach(m=>n.disposeIntermediateTensorInfo(m)),h}var DY={kernelName:wc,backendName:"cpu",kernelFunc:FY},RY=[_q,fj,Aq,Fq,wj,Rq,Pq,Lq,Wq,Vq,Gq,jq,Kq,Zq,Qq,n5,r5,i5,l5,Nq,p5,d5,m5,g5,xj,Ij,b5,gj,v5,k5,I5,T5,C5,E5,$5,D5,M5,O5,z5,B5,U5,H5,q5,K5,Y5,J5,e8,t8,n8,a8,i8,xq,l8,Sj,g8,Tj,y8,Cj,I8,S8,N8,Ej,E8,$8,D8,M8,O8,$j,Dj,yj,z8,w5,B8,U8,H8,vq,Mj,Oj,q8,zj,X8,J8,eK,aK,sK,oK,lK,Bj,pK,dK,mK,gK,bK,vK,kK,Uj,SK,CK,$K,Hj,qj,RK,OK,WK,Xj,VK,GK,HK,VC,XK,kq,Jj,ZK,QK,tX,aX,bj,rx,sX,Iq,Sq,Tq,oX,uX,cX,hX,fX,gX,bX,iq,vX,TX,CX,$X,lq,DX,MX,OX,uq,EK,WX,VX,GX,jX,KX,YX,JX,eY,dq,tY,mq,aY,sY,oY,uY,cY,bq,r8,hY,fY,yY,xY,wY,Yj,EY,$Y,DY,UK];for(let e of RY)kc(e);var UC={};Ae(UC,{assertNotComplex:()=>Lu,bindCanvasToFramebuffer:()=>HY,bindColorTextureToFramebuffer:()=>oh,bindTextureToProgramUniformSampler:()=>s_,bindTextureUnit:()=>n_,bindVertexBufferToProgramAttribute:()=>ix,callAndCheck:()=>me,canBeRepresented:()=>HC,createFragmentShader:()=>KC,createFramebuffer:()=>t_,createProgram:()=>XC,createStaticIndexBuffer:()=>JC,createStaticVertexBuffer:()=>ZC,createTexture:()=>QC,createVertexShader:()=>qC,getBatchDim:()=>pi,getExtensionOrThrow:()=>Dp,getFramebufferErrorMessage:()=>i_,getMaxTexturesInShader:()=>p_,getNumChannels:()=>UY,getProgramUniformLocation:()=>r_,getProgramUniformLocationOrThrow:()=>a_,getRowsCols:()=>ci,getShapeAs3D:()=>Mp,getTextureShapeFromLogicalShape:()=>l_,getWebGLDisjointQueryTimerVersion:()=>c_,getWebGLErrorMessage:()=>jC,getWebGLMaxTextureSize:()=>u_,hasExtension:()=>ca,isCapableOfRenderingToFloatTexture:()=>d_,isDownloadFloatTextureEnabled:()=>h_,isReshapeFree:()=>Qp,isWebGLFenceEnabled:()=>m_,isWebGLVersionEnabled:()=>lx,linkProgram:()=>YC,logShaderSourceAndInfoLog:()=>V0,resetMaxTextureSize:()=>jY,resetMaxTexturesInShader:()=>qY,unbindColorTextureFromFramebuffer:()=>ox,unbindTextureUnit:()=>GY,validateFramebuffer:()=>Rp,validateProgram:()=>ih,validateTextureSize:()=>e_});var Us={},Qd={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function GC(e,t){Us[e]=t}function qa(e,t){if(!(e in Us)||t!=null){let a=PY(e,t);if(a!==null)Us[e]=a;else return console.log("Could not get context for WebGL version",e),null}let n=Us[e];return n==null||n.isContextLost()?(delete Us[e],qa(e)):(n.disable(n.DEPTH_TEST),n.disable(n.STENCIL_TEST),n.disable(n.BLEND),n.disable(n.DITHER),n.disable(n.POLYGON_OFFSET_FILL),n.disable(n.SAMPLE_COVERAGE),n.enable(n.SCISSOR_TEST),n.enable(n.CULL_FACE),n.cullFace(n.BACK),Us[e])}function MY(e){if(typeof OffscreenCanvas!="undefined"&&e===2)return new OffscreenCanvas(300,150);if(typeof document!="undefined")return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}function PY(e,t){if(e!==1&&e!==2)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let n=t==null?MY(e):t;return n.addEventListener("webglcontextlost",a=>{a.preventDefault(),delete Us[e]},!1),H().getBool("SOFTWARE_WEBGL_ENABLED")&&(Qd.failIfMajorPerformanceCaveat=!1),e===1?n.getContext("webgl",Qd)||n.getContext("experimental-webgl",Qd):n.getContext("webgl2",Qd)}var Jp;(function(e){e[e.DENSE=0]="DENSE",e[e.SHARED_BATCH=1]="SHARED_BATCH"})(Jp||(Jp={}));var pa;(function(e){e[e.RENDER=0]="RENDER",e[e.UPLOAD=1]="UPLOAD",e[e.PIXELS=2]="PIXELS",e[e.DOWNLOAD=3]="DOWNLOAD"})(pa||(pa={}));var on;(function(e){e[e.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",e[e.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",e[e.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",e[e.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",e[e.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16"})(on||(on={}));function ed(e,t){return[t,e]}function OY(e,t){return e*t}function eh(e){let t=v.sizeFromShape(e),n=Math.ceil(t/4);return v.sizeToSquarishShape(n)}function Ou(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function LY(e,t){let[n,a]=Ou(e,t);return n*a*4}function B0(e,t){let n=e,a,r,s,i,o,l,u,p,d,c;return H().getNumber("WEBGL_VERSION")===2?(a=n.R32F,r=n.R16F,s=n.RGBA16F,i=n.RGBA32F,o=n.RED,u=4,p=1,d=n.HALF_FLOAT,c=n.FLOAT,l=n.RGBA8):(a=e.RGBA,r=e.RGBA,s=e.RGBA,i=n.RGBA,o=e.RGBA,u=4,p=4,d=t!=null?t.HALF_FLOAT_OES:null,c=e.FLOAT,l=e.RGBA),{internalFormatFloat:a,internalFormatHalfFloat:r,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:o,downloadTextureFormat:l,downloadUnpackNumChannels:u,defaultNumChannels:p,textureTypeHalfFloat:d,textureTypeFloat:c}}function me(e,t){let n=t();return H().getBool("DEBUG")&&zY(e),n}function zY(e){let t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+jC(e,t))}var WY=596e-10,BY=65504;function HC(e){return!!(H().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||e===0||WY<Math.abs(e)&&Math.abs(e)<BY)}function jC(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function Dp(e,t){return Fr(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function qC(e,t){let n=Fr(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw console.log(e.getShaderInfoLog(n)),new Error("Failed to compile vertex shader.");return n}function KC(e,t){let n=Fr(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),H().get("ENGINE_COMPILE_ONLY"))return n;if(e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw V0(t,e.getShaderInfoLog(n)),new Error("Failed to compile fragment shader.");return n}var VY=/ERROR: [0-9]+:([0-9]+):/g;function V0(e,t){let n=VY.exec(t);if(n==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(e);return}let a=+n[1],r=e.split(`
-`),s=r.length.toString().length+2,i=r.map((d,c)=>v.rightPad((c+1).toString(),s)+d),o=0;for(let d=0;d<i.length;d++)o=Math.max(i[d].length,o);let l=i.slice(0,a-1),u=i.slice(a-1,a),p=i.slice(a);console.log(l.join(`
-`)),console.log(t.split(`
-`)[0]),console.log(`%c ${v.rightPad(u[0],o)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(p.join(`
-`))}function XC(e){return Fr(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function YC(e,t){if(me(e,()=>e.linkProgram(t)),!H().get("ENGINE_COMPILE_ONLY")&&e.getProgramParameter(t,e.LINK_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function ih(e,t){if(me(e,()=>e.validateProgram(t)),e.getProgramParameter(t,e.VALIDATE_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function ZC(e,t){let n=Fr(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function JC(e,t){let n=Fr(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function UY(){return H().getNumber("WEBGL_VERSION")===2?1:4}function QC(e){return Fr(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function e_(e,t){let n=H().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){let a=`[${e}x${t}]`;throw new Error("Requested texture size "+a+" is invalid.")}if(e>n||t>n){let a=`[${e}x${t}]`,r=`[${n}x${n}]`;throw new Error("Requested texture size "+a+" greater than WebGL maximum on this browser / GPU "+r+".")}}function t_(e){return Fr(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function ix(e,t,n,a,r,s,i){let o=e.getAttribLocation(t,n);return o===-1?!1:(me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,a)),me(e,()=>e.vertexAttribPointer(o,r,e.FLOAT,!1,s,i)),me(e,()=>e.enableVertexAttribArray(o)),!0)}function n_(e,t,n){o_(e,n),me(e,()=>e.activeTexture(e.TEXTURE0+n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function GY(e,t){o_(e,t),me(e,()=>e.activeTexture(e.TEXTURE0+t)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function a_(e,t,n){return Fr(e,()=>e.getUniformLocation(t,n),'uniform "'+n+'" not present in program.')}function r_(e,t,n){return e.getUniformLocation(t,n)}function s_(e,t,n,a){me(e,()=>n_(e,t,a)),me(e,()=>e.uniform1i(n,a))}function HY(e){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),me(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function oh(e,t,n){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,n)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function ox(e,t){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function Rp(e){let t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+i_(e,t))}function i_(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function Fr(e,t,n){let a=me(e,()=>t());if(a==null)throw new Error(n);return a}function o_(e,t){let n=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,a=t+e.TEXTURE0;if(a<e.TEXTURE0||a>n){let r=`[gl.TEXTURE0, gl.TEXTURE${n}]`;throw new Error(`textureUnit must be in ${r}.`)}}function pi(e,t=2){return v.sizeFromShape(e.slice(0,e.length-t))}function ci(e){if(e.length===0)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function Mp(e){let t=[1,1,1];return e.length===0||e.length===1&&e[0]===1||(t=[pi(e),...ci(e)]),t}function l_(e,t=!1){let n=H().getNumber("WEBGL_MAX_TEXTURE_SIZE"),a=H().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");a===1/0&&H().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")&&(a=n/2),t&&(n=n*2,a=a*2,e=e.map((o,l)=>l>=e.length-2?v.nearestLargerEven(e[l]):e[l]),e.length===1&&(e=[2,e[0]])),e.length!==2&&(e=v.squeezeShape(e).newShape);let r=v.sizeFromShape(e),s=null;e.length<=1&&r<=n?s=[1,r]:e.length===2&&e[0]<=n&&e[1]<=n?s=e:e.length===3&&e[0]*e[1]<=n&&e[2]<=n?s=[e[0]*e[1],e[2]]:e.length===3&&e[0]<=n&&e[1]*e[2]<=n?s=[e[0],e[1]*e[2]]:e.length===4&&e[0]*e[1]*e[2]<=n&&e[3]<=n?s=[e[0]*e[1]*e[2],e[3]]:e.length===4&&e[0]<=n&&e[1]*e[2]*e[3]<=n&&(s=[e[0],e[1]*e[2]*e[3]]);let i=s!=null&&Math.max(...s)>a&&Math.min(...s)<=(t?2:1)&&Math.min(...s)>0;if(s==null||i)if(t){let o=pi(e),l=2,u=2;e.length&&([l,u]=ci(e)),r=o*(l/2)*(u/2),s=v.sizeToSquarishShape(r).map(p=>p*2)}else s=v.sizeToSquarishShape(r);return s}function th(e){return e%2===0}function Qp(e,t){if(e=e.slice(-2),t=t.slice(-2),v.arraysEqual(e,t)||!e.length||!t.length||e[0]===0||e[1]===0||t[0]===0||t[1]===0)return!0;if(e.length!==t.length){let n=e.slice(-1)[0],a=t.slice(-1)[0];if(n===a||th(n)&&th(a)&&(e[0]===1||t[0]===1))return!0}return e[1]===t[1]&&th(e[0])&&th(t[0])}var lh,uh;function u_(e){if(lh==null){let t=qa(e);lh=t.getParameter(t.MAX_TEXTURE_SIZE)}return lh}function jY(){lh=null}function qY(){uh=null}function p_(e){if(uh==null){let t=qa(e);uh=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,uh)}function c_(e){if(e===0)return 0;let t,n=qa(e);return ca(n,"EXT_disjoint_timer_query_webgl2")&&e===2?t=2:ca(n,"EXT_disjoint_timer_query")?t=1:t=0,t}function ca(e,t){return e.getExtension(t)!=null}function lx(e){try{if(qa(e)!=null)return!0}catch(t){return console.log("Error when getting WebGL context: ",t),!1}return!1}function d_(e){if(e===0)return!1;let t=qa(e);if(e===1){if(!ca(t,"OES_texture_float"))return!1}else if(!ca(t,"EXT_color_buffer_float"))return!1;return ux(t)}function h_(e){if(e===0)return!1;let t=qa(e);if(e===1){if(!ca(t,"OES_texture_float")||!ca(t,"WEBGL_color_buffer_float"))return!1}else{if(ca(t,"EXT_color_buffer_float"))return ux(t);let n="EXT_color_buffer_half_float";if(ca(t,n)){let a=t.getExtension(n);return KY(t,a)}return!1}return ux(t)}function ux(e){let t=B0(e),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n);let a=1,r=1;e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,a,r,0,t.textureFormatFloat,t.textureTypeFloat,null);let s=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,s),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);let i=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(s),i}function KY(e,t){let n=B0(e,t),a=e.createTexture();e.bindTexture(e.TEXTURE_2D,a);let r=1,s=1;e.texImage2D(e.TEXTURE_2D,0,n.internalFormatHalfFloat,r,s,0,n.textureFormatFloat,n.textureTypeHalfFloat,null);let i=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,i),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,a,0);let o=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(a),e.deleteFramebuffer(i),o}function m_(e){return e!==2?!1:qa(e).fenceSync!=null}function Lu(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&v.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var ve=H();ve.registerFlag("HAS_WEBGL",()=>ve.getNumber("WEBGL_VERSION")>0);ve.registerFlag("WEBGL_VERSION",()=>lx(2)?2:lx(1)?1:0);ve.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1);ve.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>ve.get("WEBGL_VERSION")===2);ve.registerFlag("WEBGL_CPU_FORWARD",()=>!0);ve.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1);ve.registerFlag("WEBGL_PACK",()=>ve.getBool("HAS_WEBGL"));ve.registerFlag("WEBGL_PACK_NORMALIZATION",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_CLIP",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_REDUCE",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_LAZILY_UNPACK",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_CONV_IM2COL",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>u_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>p_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let e=ve.getNumber("WEBGL_VERSION");return e===0?0:c_(e)});ve.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>ve.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!Tc.isMobile());ve.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>d_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>ve.getBool("WEBGL_FORCE_F16_TEXTURES")?!1:ve.getBool("WEBGL_RENDER_FLOAT32_CAPABLE"));ve.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>h_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_FENCE_API_ENABLED",()=>m_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>ve.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0);ve.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)});ve.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>Tc.isMobile()?1:-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)});ve.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128);ve.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1);ve.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5);ve.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);ve.registerFlag("WEBGL_EXP_CONV",()=>!1);ve.registerFlag("SOFTWARE_WEBGL_ENABLED",()=>ve.getBool("IS_TEST"));ve.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE",()=>1/0);ve.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE",()=>!1);ve.registerFlag("WEBGL2_ISNAN_CUSTOM",()=>!1);function Cn(){let e,t,n,a,r,s,i,o,l,u;return H().getNumber("WEBGL_VERSION")===2?(e="#version 300 es",t="in",n="out",a="in",r="texture",s="outputColor",i="out vec4 outputColor;",o=H().getBool("WEBGL2_ISNAN_CUSTOM")?`
+        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);
+  }
+  const shapeWithoutFirstDim = tensor2.shape.slice(1);
+  const outputElementShape = mergeElementShape(shapeWithoutFirstDim, elementShape);
+  const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;
+  const tensors = tidy(() => {
+    const tensors2 = [];
+    tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);
+    for (let i = 0; i < length.length; ++i) {
+      const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];
+      const indices = [0, previousLength, 0];
+      const sizes = [1, length[i], elementPerRow];
+      tensors2[i] = reshape(slice(tensor2, indices, sizes), outputElementShape);
+    }
+    tensor2.dispose();
+    return tensors2;
+  });
+  const list = new TensorList([], elementShape, tensor2.dtype, length.length);
+  for (let i = 0; i < tensors.length; i++) {
+    list.setItem(i, tensors[i]);
+  }
+  return list;
+}
+var executeOp3 = async (node, tensorMap, context) => {
+  switch (node.op) {
+    case "If":
+    case "StatelessIf": {
+      const thenFunc = getParamValue("thenBranch", node, tensorMap, context);
+      const elseFunc = getParamValue("elseBranch", node, tensorMap, context);
+      const cond = getParamValue("cond", node, tensorMap, context);
+      const args = getParamValue("args", node, tensorMap, context);
+      const condValue = await cond.data();
+      if (condValue[0]) {
+        return context.functionMap[thenFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      } else {
+        return context.functionMap[elseFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      }
+    }
+    case "While":
+    case "StatelessWhile": {
+      const bodyFunc = getParamValue("body", node, tensorMap, context);
+      const condFunc = getParamValue("cond", node, tensorMap, context);
+      const args = getParamValue("args", node, tensorMap, context);
+      const condResult = await context.functionMap[condFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      const argIds = args.map((tensor2) => tensor2.id);
+      let condValue = await condResult[0].data();
+      condResult.forEach((tensor2) => {
+        if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1) {
+          tensor2.dispose();
+        }
+      });
+      let result = args;
+      while (condValue[0]) {
+        const origResult = result;
+        result = await context.functionMap[bodyFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);
+        const resultIds = result.map((tensor2) => tensor2.id);
+        origResult.forEach((tensor2) => {
+          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {
+            tensor2.dispose();
+          }
+        });
+        const condResult2 = await context.functionMap[condFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);
+        condValue = await condResult2[0].data();
+        condResult2.forEach((tensor2) => {
+          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {
+            tensor2.dispose();
+          }
+        });
+      }
+      return result;
+    }
+    case "LoopCond": {
+      const pred = getParamValue("pred", node, tensorMap, context);
+      return [cloneTensor(pred)];
+    }
+    case "Switch": {
+      const pred = getParamValue("pred", node, tensorMap, context);
+      let data = getParamValue("data", node, tensorMap, context);
+      if (!data.kept) {
+        data = cloneTensor(data);
+      }
+      return (await pred.data())[0] ? [void 0, data] : [data, void 0];
+    }
+    case "Merge": {
+      const inputName = node.inputNames.find((name) => getTensor(name, tensorMap, context) !== void 0);
+      if (inputName) {
+        const data = getTensor(inputName, tensorMap, context);
+        return [cloneTensor(data)];
+      }
+      return void 0;
+    }
+    case "Enter": {
+      const frameId = getParamValue("frameName", node, tensorMap, context);
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.enterFrame(frameId);
+      return [cloneTensor(data)];
+    }
+    case "Exit": {
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.exitFrame();
+      return [cloneTensor(data)];
+    }
+    case "NextIteration": {
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.nextIteration();
+      return [cloneTensor(data)];
+    }
+    case "TensorArrayV3": {
+      const size = getParamValue("size", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const dynamicSize = getParamValue("dynamicSize", node, tensorMap, context);
+      const clearAfterRead = getParamValue("clearAfterRead", node, tensorMap, context);
+      const identicalElementShapes = getParamValue("identicalElementShapes", node, tensorMap, context);
+      const name = getParamValue("name", node, tensorMap, context);
+      const tensorArray = new TensorArray(name, dtype, size, elementShape, identicalElementShapes, dynamicSize, clearAfterRead);
+      context.addTensorArray(tensorArray);
+      return [tensorArray.idTensor, scalar(1)];
+    }
+    case "TensorArrayWriteV3": {
+      const id = getParamValue("tensorArrayId", node, tensorMap, context);
+      const index = getParamValue("index", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const writeTensorArray = context.getTensorArray(id.id);
+      writeTensorArray.write(index, writeTensor);
+      return [writeTensorArray.idTensor];
+    }
+    case "TensorArrayReadV3": {
+      const readId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const readIndex = getParamValue("index", node, tensorMap, context);
+      const readTensorArray = context.getTensorArray(readId.id);
+      return [readTensorArray.read(readIndex)];
+    }
+    case "TensorArrayGatherV3": {
+      const gatherId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const gatherIndices = getParamValue("indices", node, tensorMap, context);
+      const gatherDtype = getParamValue("dtype", node, tensorMap, context);
+      const gatherTensorArray = context.getTensorArray(gatherId.id);
+      return [gatherTensorArray.gather(gatherIndices, gatherDtype)];
+    }
+    case "TensorArrayScatterV3": {
+      const scatterId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const scatterIndices = getParamValue("indices", node, tensorMap, context);
+      const scatterTensor = getParamValue("tensor", node, tensorMap, context);
+      const scatterTensorArray = context.getTensorArray(scatterId.id);
+      scatterTensorArray.scatter(scatterIndices, scatterTensor);
+      return [scatterTensorArray.idTensor];
+    }
+    case "TensorArrayConcatV3": {
+      const concatId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const concatTensorArray = context.getTensorArray(concatId.id);
+      const concatDtype = getParamValue("dtype", node, tensorMap, context);
+      return [concatTensorArray.concat(concatDtype)];
+    }
+    case "TensorArraySplitV3": {
+      const splitId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const splitTensor = getParamValue("tensor", node, tensorMap, context);
+      const lengths = getParamValue("lengths", node, tensorMap, context);
+      const splitTensorArray = context.getTensorArray(splitId.id);
+      splitTensorArray.split(lengths, splitTensor);
+      return [splitTensorArray.idTensor];
+    }
+    case "TensorArraySizeV3": {
+      const sizeId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const sizeTensorArray = context.getTensorArray(sizeId.id);
+      return [scalar(sizeTensorArray.size(), "int32")];
+    }
+    case "TensorArrayCloseV3": {
+      const closeId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const closeTensorArray = context.getTensorArray(closeId.id);
+      closeTensorArray.clearAndClose();
+      return [closeTensorArray.idTensor];
+    }
+    case "TensorListSetItem": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const index = getParamValue("index", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      tensorList.setItem(index, writeTensor);
+      return [tensorList.idTensor];
+    }
+    case "TensorListGetItem": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const readIndex = getParamValue("index", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDType = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.getItem(readIndex, elementShape, elementDType)];
+    }
+    case "TensorListScatterV2":
+    case "TensorListScatter": {
+      const scatterIndices = getParamValue("indices", node, tensorMap, context);
+      const scatterTensor = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const numElements = getParamValue("numElements", node, tensorMap, context);
+      const tensorList = scatter(scatterTensor, scatterIndices, elementShape, numElements);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListReserve":
+    case "EmptyTensorList": {
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      let numElementsParam;
+      if (node.op === "TensorListReserve") {
+        numElementsParam = "numElements";
+      } else {
+        numElementsParam = "maxNumElements";
+      }
+      const numElements = getParamValue(numElementsParam, node, tensorMap, context);
+      const maxNumElements = node.op === "TensorListReserve" ? -1 : numElements;
+      const tensorList = reserve(elementShape, elementDtype, numElements, maxNumElements);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListGather": {
+      const gatherId = getParamValue("tensorListId", node, tensorMap, context);
+      const gatherIndices = getParamValue("indices", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(gatherId.id);
+      return [tensorList.gather(gatherIndices, elementDtype, elementShape)];
+    }
+    case "TensorListStack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const numElements = getParamValue("numElements", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.stack(elementShape, elementDtype, numElements)];
+    }
+    case "TensorListFromTensor": {
+      const tensor2 = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = fromTensor(tensor2, elementShape, elementDtype);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListConcat":
+    case "TensorListConcatV2": {
+      const concatId = getParamValue("tensorListId", node, tensorMap, context);
+      const tensorList = context.getTensorList(concatId.id);
+      const concatDtype = getParamValue("dtype", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      return [tensorList.concat(concatDtype, elementShape)];
+    }
+    case "TensorListPushBack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      tensorList.pushBack(writeTensor);
+      return [tensorList.idTensor];
+    }
+    case "TensorListPopBack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDType = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.popBack(elementShape, elementDType)];
+    }
+    case "TensorListSplit": {
+      const splitTensor = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const lengths = getParamValue("lengths", node, tensorMap, context);
+      const tensorList = split2(splitTensor, lengths, elementShape);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListLength": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [scalar(tensorList.size(), "int32")];
+    }
+    case "TensorListResize": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const srcTensorList = context.getTensorList(idTensor.id);
+      const destTensorList = srcTensorList.resize(size);
+      context.addTensorList(destTensorList);
+      return [destTensorList.idTensor];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+function fusedConvAndDepthWiseParams(node, tensorMap, context) {
+  const [extraOp, activationFunc] = getParamValue("fusedOps", node, tensorMap, context);
+  const isBiasAdd = extraOp === "biasadd";
+  const noBiasAdd = !isBiasAdd;
+  const isPrelu = activationFunc === "prelu";
+  const isBatchNorm = extraOp === "fusedbatchnorm";
+  const numArgs = getParamValue("numArgs", node, tensorMap, context);
+  if (isBiasAdd) {
+    if (isPrelu && numArgs !== 2) {
+      throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+    }
+    if (!isPrelu && isBiasAdd && numArgs !== 1) {
+      throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.");
+    }
+  }
+  if (isBatchNorm) {
+    throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");
+  }
+  const stride = getParamValue("strides", node, tensorMap, context);
+  const pad3 = getPadding(node, tensorMap, context);
+  const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+  const dilations = getParamValue("dilations", node, tensorMap, context);
+  let [biasArg, preluArg] = getParamValue("args", node, tensorMap, context);
+  if (noBiasAdd) {
+    preluArg = biasArg;
+    biasArg = void 0;
+  }
+  const leakyreluAlpha = getParamValue("leakyreluAlpha", node, tensorMap, context);
+  return {
+    stride,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    biasArg,
+    preluArg,
+    activationFunc,
+    leakyreluAlpha
+  };
+}
+var executeOp4 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Conv1D": {
+      const stride = getParamValue("stride", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilation = getParamValue("dilation", node, tensorMap, context);
+      return [ops.conv1d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), stride, pad3, dataFormat, dilation)];
+    }
+    case "Conv2D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      return [ops.conv2d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];
+    }
+    case "_FusedConv2D": {
+      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);
+      return [ops.fused.conv2d({
+        x: getParamValue("x", node, tensorMap, context),
+        filter: getParamValue("filter", node, tensorMap, context),
+        strides: [stride[1], stride[2]],
+        pad: pad3,
+        dataFormat,
+        dilations: [dilations[1], dilations[2]],
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    }
+    case "FusedDepthwiseConv2dNative": {
+      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);
+      return [ops.fused.depthwiseConv2d({
+        x: getParamValue("x", node, tensorMap, context),
+        filter: getParamValue("filter", node, tensorMap, context),
+        strides: [stride[1], stride[2]],
+        pad: pad3,
+        dataFormat,
+        dilations: [dilations[1], dilations[2]],
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    }
+    case "Conv2DBackpropInput":
+    case "Conv2dTranspose": {
+      const shape = getParamValue("outputShape", node, tensorMap, context);
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      return [ops.conv2dTranspose(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), shape, [stride[1], stride[2]], pad3)];
+    }
+    case "DepthwiseConv2dNative":
+    case "DepthwiseConv2d": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      return [ops.depthwiseConv2d(getParamValue("input", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];
+    }
+    case "Conv3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      return [ops.conv3d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2], stride[3]], pad3, dataFormat, [dilations[1], dilations[2], dilations[3]])];
+    }
+    case "AvgPool": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.avgPool(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];
+    }
+    case "MaxPool": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.maxPool(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];
+    }
+    case "MaxPoolWithArgmax": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      const includeBatchInIndex = getParamValue("includeBatchInIndex", node, tensorMap, context);
+      const { result, indexes } = ops.maxPoolWithArgmax(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3, includeBatchInIndex);
+      return [result, indexes];
+    }
+    case "AvgPool3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.avgPool3d(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];
+    }
+    case "MaxPool3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.maxPool3d(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];
+    }
+    case "Dilation2D": {
+      const strides = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      const strideHeight = strides[1];
+      const strideWidth = strides[2];
+      const dilationHeight = dilations[1];
+      const dilationWidth = dilations[2];
+      return [ops.dilation2d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [strideHeight, strideWidth], pad3, [dilationHeight, dilationWidth], "NHWC")];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp5 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Fill": {
+      const shape = getParamValue("shape", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      const value = getParamValue("value", node, tensorMap, context);
+      return [ops.fill(shape, value, dtype)];
+    }
+    case "LinSpace": {
+      const start = getParamValue("start", node, tensorMap, context);
+      const stop = getParamValue("stop", node, tensorMap, context);
+      const num = getParamValue("num", node, tensorMap, context);
+      return [ops.linspace(start, stop, num)];
+    }
+    case "Multinomial": {
+      const logits = getParamValue("logits", node, tensorMap, context);
+      const numSamples = getParamValue("numSamples", node, tensorMap, context);
+      const seed = getParamValue("seed", node, tensorMap, context);
+      return [ops.multinomial(logits, numSamples, seed)];
+    }
+    case "OneHot": {
+      const indices = getParamValue("indices", node, tensorMap, context);
+      const depth = getParamValue("depth", node, tensorMap, context);
+      const onValue = getParamValue("onValue", node, tensorMap, context);
+      const offValue = getParamValue("offValue", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      return [ops.oneHot(indices, depth, onValue, offValue, dtype)];
+    }
+    case "Ones": {
+      return [ops.ones(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "OnesLike": {
+      return [ops.onesLike(getParamValue("x", node, tensorMap, context))];
+    }
+    case "RandomStandardNormal": {
+      return [ops.randomStandardNormal(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context), getParamValue("seed", node, tensorMap, context))];
+    }
+    case "RandomUniform": {
+      return [ops.randomUniform(
+        getParamValue("shape", node, tensorMap, context),
+        getParamValue("minval", node, tensorMap, context),
+        getParamValue("maxval", node, tensorMap, context),
+        getParamValue("dtype", node, tensorMap, context)
+      )];
+    }
+    case "Range": {
+      const start = getParamValue("start", node, tensorMap, context);
+      const stop = getParamValue("stop", node, tensorMap, context);
+      const step5 = getParamValue("step", node, tensorMap, context);
+      return [ops.range(start, stop, step5, getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "TruncatedNormal": {
+      const shape = getParamValue("shape", node, tensorMap, context);
+      const mean4 = getParamValue("mean", node, tensorMap, context);
+      const stdDev = getParamValue("stdDev", node, tensorMap, context);
+      const seed = getParamValue("seed", node, tensorMap, context);
+      return [ops.truncatedNormal(shape, mean4, stdDev, getParamValue("dtype", node, tensorMap, context), seed)];
+    }
+    case "Zeros": {
+      return [ops.zeros(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "ZerosLike": {
+      return [ops.zerosLike(getParamValue("x", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+function nmsParams(node, tensorMap, context) {
+  const boxes = getParamValue("boxes", node, tensorMap, context);
+  const scores = getParamValue("scores", node, tensorMap, context);
+  const maxOutputSize = getParamValue("maxOutputSize", node, tensorMap, context);
+  const iouThreshold = getParamValue("iouThreshold", node, tensorMap, context);
+  const scoreThreshold = getParamValue("scoreThreshold", node, tensorMap, context);
+  const softNmsSigma = getParamValue("softNmsSigma", node, tensorMap, context);
+  return {
+    boxes,
+    scores,
+    maxOutputSize,
+    iouThreshold,
+    scoreThreshold,
+    softNmsSigma
+  };
+}
+var executeOp6 = async (node, tensorMap, context, resourceManager, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "NonMaxSuppressionV5": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = nmsParams(node, tensorMap, context);
+      const result = await ops.image.nonMaxSuppressionWithScoreAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+      return [result.selectedIndices, result.selectedScores];
+    }
+    case "NonMaxSuppressionV4": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);
+      const padToMaxOutputSize = getParamValue("padToMaxOutputSize", node, tensorMap, context);
+      const result = await ops.image.nonMaxSuppressionPaddedAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+      return [result.selectedIndices, result.validOutputs];
+    }
+    case "NonMaxSuppressionV3":
+    case "NonMaxSuppressionV2": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);
+      return [await ops.image.nonMaxSuppressionAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold)];
+    }
+    case "Where": {
+      const condition = ops.cast(getParamValue("condition", node, tensorMap, context), "bool");
+      const result = [await ops.whereAsync(condition)];
+      condition.dispose();
+      return result;
+    }
+    case "ListDiff": {
+      return ops.setdiff1dAsync(getParamValue("x", node, tensorMap, context), getParamValue("y", node, tensorMap, context));
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp7 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "LowerBound": {
+      const sortedSequence = getParamValue("sortedSequence", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      return [ops.lowerBound(sortedSequence, values)];
+    }
+    case "TopKV2": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const k = getParamValue("k", node, tensorMap, context);
+      const sorted = getParamValue("sorted", node, tensorMap, context);
+      const result = ops.topk(x, k, sorted);
+      return [result.values, result.indices];
+    }
+    case "UpperBound": {
+      const sortedSequence = getParamValue("sortedSequence", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      return [ops.upperBound(sortedSequence, values)];
+    }
+    case "Unique": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const result = ops.unique(x);
+      return [result.values, result.indices];
+    }
+    case "UniqueV2": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const result = ops.unique(x, axis);
+      return [result.values, result.indices];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp8 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Const": {
+      return tensorMap[node.name];
+    }
+    case "PlaceholderWithDefault":
+      const def = getParamValue("default", node, tensorMap, context);
+      return [getTensor(node.name, tensorMap, context) || def];
+    case "Placeholder":
+      return [getTensor(node.name, tensorMap, context)];
+    case "Identity":
+    case "StopGradient":
+    case "FakeQuantWithMinMaxVars": {
+      const data2 = getParamValue("x", node, tensorMap, context);
+      return [cloneTensor(data2)];
+    }
+    case "IdentityN":
+      return getParamValue("x", node, tensorMap, context).map((t) => cloneTensor(t));
+    case "Snapshot":
+      const snapshot = getParamValue("x", node, tensorMap, context);
+      return [cloneTensor(snapshot)];
+    case "Shape":
+      return [ops.tensor1d(getParamValue("x", node, tensorMap, context).shape, "int32")];
+    case "ShapeN":
+      return getParamValue("x", node, tensorMap, context).map((t) => ops.tensor1d(t.shape));
+    case "Size":
+      return [ops.scalar(getParamValue("x", node, tensorMap, context).size, "int32")];
+    case "Rank":
+      return [ops.scalar(getParamValue("x", node, tensorMap, context).rank, "int32")];
+    case "NoOp":
+      return [ops.scalar(1)];
+    case "Print":
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const data = getParamValue("data", node, tensorMap, context);
+      const message = getParamValue("message", node, tensorMap, context);
+      const summarize = getParamValue("summarize", node, tensorMap, context);
+      console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance.");
+      console.log(message);
+      for (let i = 0; i < data.length; i++) {
+        console.log(Array.prototype.slice.call(data[i].dataSync()).slice(0, summarize));
+      }
+      return [input2];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var HashTable = class {
+  constructor(keyDType, valueDType) {
+    this.keyDType = keyDType;
+    this.valueDType = valueDType;
+    this.handle = scalar(0);
+    this.tensorMap = /* @__PURE__ */ new Map();
+    keep(this.handle);
+  }
+  get id() {
+    return this.handle.id;
+  }
+  clearAndClose() {
+    this.tensorMap.forEach((value) => value.dispose());
+    this.tensorMap.clear();
+    this.handle.dispose();
+  }
+  size() {
+    return this.tensorMap.size;
+  }
+  tensorSize() {
+    return scalar(this.size(), "int32");
+  }
+  async import(keys, values) {
+    this.checkKeyAndValueTensor(keys, values);
+    const $keys = await keys.data();
+    this.tensorMap.forEach((value) => value.dispose());
+    this.tensorMap.clear();
+    return tidy(() => {
+      const $values = unstack(values);
+      const keysLength = $keys.length;
+      const valuesLength = $values.length;
+      util_exports.assert(keysLength === valuesLength, () => `The number of elements doesn't match, keys has ${keysLength} elements, the values has ${valuesLength} elements.`);
+      for (let i = 0; i < keysLength; i++) {
+        const key = $keys[i];
+        const value = $values[i];
+        keep(value);
+        this.tensorMap.set(key, value);
+      }
+      return this.handle;
+    });
+  }
+  async find(keys, defaultValue) {
+    this.checkKeyAndValueTensor(keys, defaultValue);
+    const $keys = await keys.data();
+    return tidy(() => {
+      const result = [];
+      for (let i = 0; i < $keys.length; i++) {
+        const key = $keys[i];
+        const value = this.findWithDefault(key, defaultValue);
+        result.push(value);
+      }
+      return stack(result);
+    });
+  }
+  findWithDefault(key, defaultValue) {
+    const result = this.tensorMap.get(key);
+    return result != null ? result : defaultValue;
+  }
+  checkKeyAndValueTensor(key, value) {
+    if (key.dtype !== this.keyDType) {
+      throw new Error(`Expect key dtype ${this.keyDType}, but got ${key.dtype}`);
+    }
+    if (value.dtype !== this.valueDType) {
+      throw new Error(`Expect value dtype ${this.valueDType}, but got ${value.dtype}`);
+    }
+  }
+};
+var executeOp9 = async (node, tensorMap, context, resourceManager) => {
+  switch (node.op) {
+    case "HashTable":
+    case "HashTableV2": {
+      const existingTableHandle = resourceManager.getHashTableHandleByName(node.name);
+      if (existingTableHandle != null) {
+        return [existingTableHandle];
+      } else {
+        const keyDType = getParamValue("keyDType", node, tensorMap, context);
+        const valueDType = getParamValue("valueDType", node, tensorMap, context);
+        const hashTable = new HashTable(keyDType, valueDType);
+        resourceManager.addHashTable(node.name, hashTable);
+        return [hashTable.handle];
+      }
+    }
+    case "LookupTableImport":
+    case "LookupTableImportV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const keys = getParamValue("keys", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [await hashTable.import(keys, values)];
+    }
+    case "LookupTableFind":
+    case "LookupTableFindV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const keys = getParamValue("keys", node, tensorMap, context);
+      const defaultValue = getParamValue("defaultValue", node, tensorMap, context);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [await hashTable.find(keys, defaultValue)];
+    }
+    case "LookupTableSize":
+    case "LookupTableSizeV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [hashTable.tensorSize()];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp10 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "ResizeBilinear": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const alignCorners = getParamValue("alignCorners", node, tensorMap, context);
+      const halfPixelCenters = getParamValue("halfPixelCenters", node, tensorMap, context);
+      return [ops.image.resizeBilinear(images, [size[0], size[1]], alignCorners, halfPixelCenters)];
+    }
+    case "ResizeNearestNeighbor": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const alignCorners = getParamValue("alignCorners", node, tensorMap, context);
+      const halfPixelCenters = getParamValue("halfPixelCenters", node, tensorMap, context);
+      return [ops.image.resizeNearestNeighbor(images, [size[0], size[1]], alignCorners, halfPixelCenters)];
+    }
+    case "CropAndResize": {
+      const image2 = getParamValue("image", node, tensorMap, context);
+      const boxes = getParamValue("boxes", node, tensorMap, context);
+      const boxInd = getParamValue("boxInd", node, tensorMap, context);
+      const cropSize = getParamValue("cropSize", node, tensorMap, context);
+      const method = getParamValue("method", node, tensorMap, context);
+      const extrapolationValue = getParamValue("extrapolationValue", node, tensorMap, context);
+      return [ops.image.cropAndResize(image2, boxes, boxInd, cropSize, method, extrapolationValue)];
+    }
+    case "ImageProjectiveTransformV3": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const transforms = getParamValue("transforms", node, tensorMap, context);
+      const outputShape = getParamValue("outputShape", node, tensorMap, context);
+      const fillValue = getParamValue("fillValue", node, tensorMap, context);
+      const interpolation = getParamValue("interpolation", node, tensorMap, context);
+      const fillMode = getParamValue("fillMode", node, tensorMap, context);
+      return [ops.image.transform(images, transforms, interpolation.toLowerCase(), fillMode.toLowerCase(), fillValue, outputShape)];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp11 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Equal": {
+      return [ops.equal(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "NotEqual": {
+      return [ops.notEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Greater": {
+      return [ops.greater(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "GreaterEqual": {
+      return [ops.greaterEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Less": {
+      return [ops.less(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LessEqual": {
+      return [ops.lessEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LogicalAnd": {
+      return [ops.logicalAnd(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LogicalNot": {
+      return [ops.logicalNot(getParamValue("a", node, tensorMap, context))];
+    }
+    case "LogicalOr": {
+      return [ops.logicalOr(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Select":
+    case "SelectV2": {
+      return [ops.where(getParamValue("condition", node, tensorMap, context), getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp12 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "BatchMatMul":
+    case "BatchMatMulV2":
+    case "MatMul":
+      return [ops.matMul(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context), getParamValue("transposeA", node, tensorMap, context), getParamValue("transposeB", node, tensorMap, context))];
+    case "Einsum":
+      return [ops.einsum(getParamValue("equation", node, tensorMap, context), ...getParamValue("tensors", node, tensorMap, context))];
+    case "Transpose":
+      return [ops.transpose(getParamValue("x", node, tensorMap, context), getParamValue("perm", node, tensorMap, context))];
+    case "_FusedMatMul":
+      const [extraOp, activationFunc] = getParamValue("fusedOps", node, tensorMap, context);
+      const isBiasAdd = extraOp === "biasadd";
+      const isPrelu = activationFunc === "prelu";
+      const numArgs = getParamValue("numArgs", node, tensorMap, context);
+      const leakyreluAlpha = getParamValue("leakyreluAlpha", node, tensorMap, context);
+      if (isBiasAdd) {
+        if (isPrelu && numArgs !== 2) {
+          throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+        }
+        if (!isPrelu && numArgs !== 1) {
+          throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.");
+        }
+      }
+      const [biasArg, preluArg] = getParamValue("args", node, tensorMap, context);
+      return [ops.fused.matMul({
+        a: getParamValue("a", node, tensorMap, context),
+        b: getParamValue("b", node, tensorMap, context),
+        transposeA: getParamValue("transposeA", node, tensorMap, context),
+        transposeB: getParamValue("transposeB", node, tensorMap, context),
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp13 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "EuclideanNorm":
+      return [ops.euclideanNorm(getParamValue("x", node, tensorMap, context), getParamValue("axis", node, tensorMap, context), getParamValue("keepDims", node, tensorMap, context))];
+    case "FusedBatchNorm":
+    case "FusedBatchNormV2": {
+      return [ops.batchNorm(getParamValue("x", node, tensorMap, context), getParamValue("mean", node, tensorMap, context), getParamValue("variance", node, tensorMap, context), getParamValue("offset", node, tensorMap, context), getParamValue("scale", node, tensorMap, context), getParamValue("epsilon", node, tensorMap, context))];
+    }
+    case "FusedBatchNormV3": {
+      return [ops.batchNorm(getParamValue("x", node, tensorMap, context), getParamValue("mean", node, tensorMap, context), getParamValue("variance", node, tensorMap, context), getParamValue("offset", node, tensorMap, context), getParamValue("scale", node, tensorMap, context), getParamValue("epsilon", node, tensorMap, context))];
+    }
+    case "LRN": {
+      return [ops.localResponseNormalization(getParamValue("x", node, tensorMap, context), getParamValue("radius", node, tensorMap, context), getParamValue("bias", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context), getParamValue("beta", node, tensorMap, context))];
+    }
+    case "Softmax": {
+      return [ops.softmax(getParamValue("x", node, tensorMap, context))];
+    }
+    case "LogSoftmax": {
+      return [ops.logSoftmax(getParamValue("x", node, tensorMap, context))];
+    }
+    case "SparseToDense": {
+      return [ops.sparseToDense(getParamValue("sparseIndices", node, tensorMap, context), getParamValue("outputShape", node, tensorMap, context), getParamValue("sparseValues", node, tensorMap, context), getParamValue("defaultValue", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp14 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Max": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.max(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Mean": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.mean(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Min": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.min(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Sum": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.sum(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "All": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.all(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Any": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.any(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "ArgMax": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.argMax(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "ArgMin": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.argMin(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Prod": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.prod(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Cumprod": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const exclusive = getParamValue("exclusive", node, tensorMap, context);
+      const reverse5 = getParamValue("reverse", node, tensorMap, context);
+      return [ops.cumprod(getParamValue("x", node, tensorMap, context), axis, exclusive, reverse5)];
+    }
+    case "Cumsum": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const exclusive = getParamValue("exclusive", node, tensorMap, context);
+      const reverse5 = getParamValue("reverse", node, tensorMap, context);
+      return [ops.cumsum(getParamValue("x", node, tensorMap, context), axis, exclusive, reverse5)];
+    }
+    case "Bincount":
+      const x = getParamValue("x", node, tensorMap, context);
+      const weights = getParamValue("weights", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      return [ops.bincount(x, weights, size)];
+    case "DenseBincount": {
+      const x2 = getParamValue("x", node, tensorMap, context);
+      const weights2 = getParamValue("weights", node, tensorMap, context);
+      const size2 = getParamValue("size", node, tensorMap, context);
+      const binaryOutput = getParamValue("binaryOutput", node, tensorMap, context);
+      return [ops.denseBincount(x2, weights2, size2, binaryOutput)];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp15 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "ConcatV2":
+    case "Concat": {
+      const n = getParamValue("n", node, tensorMap, context);
+      const axis = getParamValue("axis", node, tensorMap, context);
+      let inputs = getParamValue("tensors", node, tensorMap, context);
+      inputs = inputs.slice(0, n);
+      return [ops.concat(inputs, axis)];
+    }
+    case "Gather": {
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gather(input2, ops.cast(indices, "int32"), 0)];
+    }
+    case "GatherV2": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const batchDims = getParamValue("batchDims", node, tensorMap, context);
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gather(input2, ops.cast(indices, "int32"), axis, batchDims)];
+    }
+    case "Reverse": {
+      const dims = getParamValue("dims", node, tensorMap, context);
+      const axis = [];
+      for (let i = 0; i < dims.length; i++) {
+        if (dims[i]) {
+          axis.push(i);
+        }
+      }
+      const input2 = getParamValue("x", node, tensorMap, context);
+      return [ops.reverse(input2, axis)];
+    }
+    case "ReverseV2": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const input2 = getParamValue("x", node, tensorMap, context);
+      return [ops.reverse(input2, axis)];
+    }
+    case "Slice": {
+      const begin = getParamValue("begin", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      return [ops.slice(getParamValue("x", node, tensorMap, context), begin, size)];
+    }
+    case "StridedSlice": {
+      const begin = getParamValue("begin", node, tensorMap, context);
+      const end = getParamValue("end", node, tensorMap, context);
+      const strides = getParamValue("strides", node, tensorMap, context);
+      const beginMask = getParamValue("beginMask", node, tensorMap, context);
+      const endMask = getParamValue("endMask", node, tensorMap, context);
+      const ellipsisMask = getParamValue("ellipsisMask", node, tensorMap, context);
+      const newAxisMask = getParamValue("newAxisMask", node, tensorMap, context);
+      const shrinkAxisMask = getParamValue("shrinkAxisMask", node, tensorMap, context);
+      const tensor2 = getParamValue("x", node, tensorMap, context);
+      return [ops.stridedSlice(tensor2, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask)];
+    }
+    case "Pack": {
+      return tidy(() => {
+        const axis = getParamValue("axis", node, tensorMap, context);
+        const tensors = getParamValue("tensors", node, tensorMap, context);
+        const shape = tensors[0].shape;
+        const squeezedShape = ops.squeeze(tensors[0]).shape;
+        const mapped = tensors.map((tensor2) => {
+          const sameShape = util_exports.arraysEqual(tensor2.shape, shape);
+          if (!sameShape && !util_exports.arraysEqual(ops.squeeze(tensor2).shape, squeezedShape)) {
+            throw new Error("the input tensors shape does not match");
+          }
+          return sameShape ? tensor2 : ops.reshape(tensor2, shape);
+        });
+        return [ops.stack(mapped, axis)];
+      });
+    }
+    case "Unpack": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const tensor2 = getParamValue("tensor", node, tensorMap, context);
+      return ops.unstack(tensor2, axis);
+    }
+    case "Tile": {
+      const reps = getParamValue("reps", node, tensorMap, context);
+      return [ops.tile(getParamValue("x", node, tensorMap, context), reps)];
+    }
+    case "Split":
+    case "SplitV": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const numOrSizeSplits = getParamValue("numOrSizeSplits", node, tensorMap, context);
+      const tensor2 = getParamValue("x", node, tensorMap, context);
+      return ops.split(tensor2, numOrSizeSplits, axis);
+    }
+    case "ScatterNd": {
+      const indices = getParamValue("indices", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      const shape = getParamValue("shape", node, tensorMap, context);
+      return [ops.scatterND(indices, values, shape)];
+    }
+    case "GatherNd": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gatherND(x, indices)];
+    }
+    case "SparseToDense": {
+      const indices = getParamValue("sparseIndices", node, tensorMap, context);
+      const shape = getParamValue("outputShape", node, tensorMap, context);
+      const sparseValues = getParamValue("sparseValues", node, tensorMap, context);
+      const defaultValue = getParamValue("defaultValue", node, tensorMap, context);
+      return [ops.sparseToDense(indices, sparseValues, shape, sparseValues.dtype === defaultValue.dtype ? defaultValue : ops.cast(defaultValue, sparseValues.dtype))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp16 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "SparseFillEmptyRows": {
+      const { outputIndices, outputValues, emptyRowIndicator, reverseIndexMap } = ops.sparse.sparseFillEmptyRows(getParamValue("indices", node, tensorMap, context), getParamValue("values", node, tensorMap, context), getParamValue("denseShape", node, tensorMap, context), getParamValue("defaultValue", node, tensorMap, context));
+      return [
+        outputIndices,
+        outputValues,
+        emptyRowIndicator,
+        reverseIndexMap
+      ];
+    }
+    case "SparseReshape": {
+      const { outputIndices, outputShape } = ops.sparse.sparseReshape(getParamValue("inputIndices", node, tensorMap, context), getParamValue("inputShape", node, tensorMap, context), getParamValue("newShape", node, tensorMap, context));
+      return [outputIndices, outputShape];
+    }
+    case "SparseSegmentMean": {
+      const outputData = ops.sparse.sparseSegmentMean(getParamValue("data", node, tensorMap, context), getParamValue("indices", node, tensorMap, context), getParamValue("segmentIds", node, tensorMap, context));
+      return [outputData];
+    }
+    case "SparseSegmentSum": {
+      const outputData = ops.sparse.sparseSegmentSum(getParamValue("data", node, tensorMap, context), getParamValue("indices", node, tensorMap, context), getParamValue("segmentIds", node, tensorMap, context));
+      return [outputData];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp17 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "FFT": {
+      return [ops.fft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "IFFT": {
+      return [ops.ifft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "RFFT": {
+      return [ops.rfft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "IRFFT": {
+      return [ops.irfft(getParamValue("x", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp18 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "StringNGrams": {
+      const { nGrams, nGramsSplits } = ops.string.stringNGrams(getParamValue("data", node, tensorMap, context), getParamValue("dataSplits", node, tensorMap, context), getParamValue("separator", node, tensorMap, context), getParamValue("nGramWidths", node, tensorMap, context), getParamValue("leftPad", node, tensorMap, context), getParamValue("rightPad", node, tensorMap, context), getParamValue("padWidth", node, tensorMap, context), getParamValue("preserveShortSequences", node, tensorMap, context));
+      return [nGrams, nGramsSplits];
+    }
+    case "StringSplit": {
+      const { indices, values, shape } = ops.string.stringSplit(getParamValue("input", node, tensorMap, context), getParamValue("delimiter", node, tensorMap, context), getParamValue("skipEmpty", node, tensorMap, context));
+      return [indices, values, shape];
+    }
+    case "StringToHashBucketFast": {
+      const output = ops.string.stringToHashBucketFast(getParamValue("input", node, tensorMap, context), getParamValue("numBuckets", node, tensorMap, context));
+      return [output];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+var executeOp19 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Cast": {
+      return [ops.cast(getParamValue("x", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "ExpandDims": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.expandDims(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Squeeze": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.squeeze(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Reshape": {
+      return [ops.reshape(getParamValue("x", node, tensorMap, context), getParamValue("shape", node, tensorMap, context))];
+    }
+    case "MirrorPad": {
+      return [ops.mirrorPad(getParamValue("x", node, tensorMap, context), getParamValue("padding", node, tensorMap, context), getParamValue("mode", node, tensorMap, context))];
+    }
+    case "PadV2":
+    case "Pad": {
+      return [ops.pad(getParamValue("x", node, tensorMap, context), getParamValue("padding", node, tensorMap, context), getParamValue("constantValue", node, tensorMap, context))];
+    }
+    case "SpaceToBatchND": {
+      const blockShape = getParamValue("blockShape", node, tensorMap, context);
+      const paddings = getParamValue("paddings", node, tensorMap, context);
+      return [ops.spaceToBatchND(getParamValue("x", node, tensorMap, context), blockShape, paddings)];
+    }
+    case "BatchToSpaceND": {
+      const blockShape = getParamValue("blockShape", node, tensorMap, context);
+      const crops = getParamValue("crops", node, tensorMap, context);
+      return [ops.batchToSpaceND(getParamValue("x", node, tensorMap, context), blockShape, crops)];
+    }
+    case "DepthToSpace": {
+      const blockSize = getParamValue("blockSize", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      return [ops.depthToSpace(getParamValue("x", node, tensorMap, context), blockSize, dataFormat)];
+    }
+    case "BroadcastTo": {
+      return [ops.broadcastTo(getParamValue("x", node, tensorMap, context), getParamValue("shape", node, tensorMap, context))];
+    }
+    case "BroadcastArgs": {
+      return [ops.broadcastArgs(getParamValue("s0", node, tensorMap, context), getParamValue("s1", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+function executeOp20(node, tensorMap, context, resourceManager, tidy2 = tidy) {
+  const value = ((node2, tensorMap2, context2) => {
+    switch (node2.category) {
+      case "arithmetic":
+        return tidy2(() => executeOp(node2, tensorMap2, context2));
+      case "basic_math":
+        return tidy2(() => executeOp2(node2, tensorMap2, context2));
+      case "control":
+        return executeOp3(node2, tensorMap2, context2);
+      case "convolution":
+        return tidy2(() => executeOp4(node2, tensorMap2, context2));
+      case "creation":
+        return tidy2(() => executeOp5(node2, tensorMap2, context2));
+      case "dynamic":
+        return executeOp6(node2, tensorMap2, context2);
+      case "evaluation":
+        return tidy2(() => executeOp7(node2, tensorMap2, context2));
+      case "image":
+        return tidy2(() => executeOp10(node2, tensorMap2, context2));
+      case "graph":
+        return tidy2(() => executeOp8(node2, tensorMap2, context2));
+      case "logical":
+        return tidy2(() => executeOp11(node2, tensorMap2, context2));
+      case "matrices":
+        return tidy2(() => executeOp12(node2, tensorMap2, context2));
+      case "normalization":
+        return tidy2(() => executeOp13(node2, tensorMap2, context2));
+      case "reduction":
+        return tidy2(() => executeOp14(node2, tensorMap2, context2));
+      case "slice_join":
+        return tidy2(() => executeOp15(node2, tensorMap2, context2));
+      case "sparse":
+        return tidy2(() => executeOp16(node2, tensorMap2, context2));
+      case "spectral":
+        return tidy2(() => executeOp17(node2, tensorMap2, context2));
+      case "string":
+        return tidy2(() => executeOp18(node2, tensorMap2, context2));
+      case "transformation":
+        return tidy2(() => executeOp19(node2, tensorMap2, context2));
+      case "hash_table":
+        return executeOp9(node2, tensorMap2, context2, resourceManager);
+      case "custom":
+        const opMapper = getRegisteredOp(node2.op);
+        if (opMapper && opMapper.customExecutor) {
+          return opMapper.customExecutor(new NodeValueImpl(node2, tensorMap2, context2));
+        } else {
+          throw TypeError(`Custom op ${node2.op} is not registered.`);
+        }
+      default:
+        throw TypeError(`Unknown op '${node2.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`);
+    }
+  })(node, tensorMap, context);
+  if (util_exports.isPromise(value)) {
+    return value.then((data) => [].concat(data));
+  }
+  return [].concat(value);
+}
+var ExecutionContext = class {
+  constructor(weightMap = {}, tensorArrayMap = {}, tensorListMap = {}, functionMap = {}) {
+    this.weightMap = weightMap;
+    this.tensorArrayMap = tensorArrayMap;
+    this.tensorListMap = tensorListMap;
+    this.functionMap = functionMap;
+    this.rootContext = { id: 0, frameName: "", iterationId: 0 };
+    this.contexts = [this.rootContext];
+    this.lastId = 0;
+    this.generateCurrentContextIds();
+  }
+  newFrame(id, frameName) {
+    return { id, frameName, iterationId: 0 };
+  }
+  set currentContext(contexts2) {
+    if (this.contexts !== contexts2) {
+      this.contexts = contexts2;
+      this.generateCurrentContextIds();
+    }
+  }
+  get currentContext() {
+    return this.contexts;
+  }
+  get currentContextId() {
+    return this._currentContextIds[0];
+  }
+  get currentContextIds() {
+    return this._currentContextIds;
+  }
+  generateCurrentContextIds() {
+    const names = [];
+    for (let i = 0; i < this.contexts.length - 1; i++) {
+      const contexts2 = this.contexts.slice(0, this.contexts.length - i);
+      names.push(this.contextIdforContexts(contexts2));
+    }
+    names.push("");
+    this._currentContextIds = names;
+  }
+  contextIdforContexts(contexts2) {
+    return contexts2 ? contexts2.map((context) => context.id === 0 && context.iterationId === 0 ? "" : `${context.frameName}-${context.iterationId}`).join("/") : "";
+  }
+  enterFrame(frameId) {
+    if (this.contexts) {
+      this.lastId++;
+      this.contexts = this.contexts.slice();
+      this.contexts.push(this.newFrame(this.lastId, frameId));
+      this._currentContextIds.unshift(this.contextIdforContexts(this.contexts));
+    }
+  }
+  exitFrame() {
+    if (this.contexts && this.contexts.length > 1) {
+      this.contexts = this.contexts.slice();
+      this.contexts.splice(-1);
+      this.currentContextIds.shift();
+    } else {
+      throw new Error("Cannot exit frame, the context is empty");
+    }
+  }
+  nextIteration() {
+    if (this.contexts && this.contexts.length > 0) {
+      this.contexts = this.contexts.slice();
+      this.lastId++;
+      const context = Object.assign({}, this.contexts[this.contexts.length - 1]);
+      context.iterationId += 1;
+      context.id = this.lastId;
+      this.contexts.splice(-1, 1, context);
+      this._currentContextIds.splice(0, 1, this.contextIdforContexts(this.contexts));
+    } else {
+      throw new Error("Cannot increase frame iteration, the context is empty");
+    }
+  }
+  getWeight(name) {
+    return this.weightMap[name];
+  }
+  addTensorArray(tensorArray) {
+    this.tensorArrayMap[tensorArray.id] = tensorArray;
+  }
+  getTensorArray(id) {
+    return this.tensorArrayMap[id];
+  }
+  addTensorList(tensorList) {
+    this.tensorListMap[tensorList.id] = tensorList;
+  }
+  getTensorList(id) {
+    return this.tensorListMap[id];
+  }
+  dispose(keepIds) {
+    for (const key in this.tensorArrayMap) {
+      this.tensorArrayMap[key].clearAndClose(keepIds);
+    }
+    for (const key in this.tensorListMap) {
+      this.tensorListMap[key].clearAndClose(keepIds);
+    }
+  }
+};
+function getExecutionSubgraph(inputs, outputs, weightMap, initNodes) {
+  const usedNodes = /* @__PURE__ */ new Set();
+  const missingInputs = [];
+  let dynamicNode = null;
+  let syncInputs = null;
+  const seen = /* @__PURE__ */ new Set();
+  const inputNodeNames = Object.keys(inputs).map((name) => parseNodeName(name)[0]);
+  let initNodeNames = [];
+  if (initNodes != null) {
+    initNodeNames = initNodes.map((node) => parseNodeName(node.name)[0]);
+  }
+  const frontier = [...outputs];
+  while (frontier.length > 0) {
+    const node = frontier.pop();
+    if (isControlFlow(node) || isDynamicShape(node) || isHashTable(node)) {
+      if (dynamicNode == null) {
+        dynamicNode = node;
+        syncInputs = dynamicNode.children.map((child) => child.name).filter((name) => usedNodes.has(name));
+      }
+    }
+    usedNodes.add(node.name);
+    if (weightMap[node.name] != null) {
+      continue;
+    }
+    if (inputNodeNames.indexOf(node.name) !== -1) {
+      continue;
+    }
+    if (initNodeNames.indexOf(node.name) !== -1) {
+      continue;
+    }
+    if (node.inputs.length === 0) {
+      missingInputs.push(node.name);
+      continue;
+    }
+    node.inputs.forEach((input2) => {
+      if (seen.has(input2.name)) {
+        return;
+      }
+      seen.add(input2.name);
+      frontier.push(input2);
+    });
+  }
+  return { inputs, outputs, usedNodes, missingInputs, dynamicNode, syncInputs };
+}
+function getNodesInTopologicalOrder(graph, weightMap, executionInfo) {
+  const { usedNodes, inputs } = executionInfo;
+  const frontier = [];
+  const inputNodes = Object.keys(inputs).map((name) => parseNodeName(name)[0]).map((name) => graph.nodes[name]);
+  const initNodes = graph.initNodes;
+  inputNodes.forEach((input2) => {
+    if (usedNodes.has(input2.name)) {
+      frontier.push(input2);
+    }
+  });
+  graph.weights.forEach((weight) => {
+    if (usedNodes.has(weight.name)) {
+      frontier.push(weight);
+    }
+  });
+  if (initNodes != null) {
+    initNodes.forEach((node) => {
+      if (usedNodes.has(node.name)) {
+        frontier.push(node);
+      }
+    });
+  }
+  const seen = /* @__PURE__ */ new Set();
+  const orderedNodes = [];
+  while (frontier.length > 0) {
+    const node = frontier.pop();
+    seen.add(node.name);
+    if (!weightMap[node.name]) {
+      orderedNodes.push(node);
+    }
+    node.children.forEach((child) => {
+      if (!seen.has(child.name) && usedNodes.has(child.name) && child.inputs.every((input2) => seen.has(input2.name))) {
+        frontier.push(child);
+      }
+    });
+  }
+  return orderedNodes;
+}
+var CONTROL_FLOW_OPS = [
+  "Switch",
+  "Merge",
+  "Enter",
+  "Exit",
+  "NextIteration",
+  "StatelessIf",
+  "StatelessWhile",
+  "if",
+  "While"
+];
+var DYNAMIC_SHAPE_OPS = [
+  "NonMaxSuppressionV2",
+  "NonMaxSuppressionV3",
+  "NonMaxSuppressionV5",
+  "Where"
+];
+var HASH_TABLE_OPS = [
+  "HashTable",
+  "HashTableV2",
+  "LookupTableImport",
+  "LookupTableImportV2",
+  "LookupTableFind",
+  "LookupTableFindV2",
+  "LookupTableSize",
+  "LookupTableSizeV2"
+];
+function isControlFlow(node) {
+  return CONTROL_FLOW_OPS.indexOf(node.op) >= 0;
+}
+function isDynamicShape(node) {
+  return DYNAMIC_SHAPE_OPS.indexOf(node.op) >= 0;
+}
+function isHashTable(node) {
+  return HASH_TABLE_OPS.indexOf(node.op) >= 0;
+}
+var GraphExecutor = class {
+  constructor(graph, parent) {
+    this.graph = graph;
+    this.parent = parent;
+    this.compiledMap = /* @__PURE__ */ new Map();
+    this._weightMap = {};
+    this.SEPERATOR = ",";
+    this._functions = {};
+    this._functionExecutorMap = {};
+    this.intermediateTensors = {};
+    this.keepTensorForDebug = false;
+    this._outputs = graph.outputs;
+    this._inputs = graph.inputs;
+    this._initNodes = graph.initNodes;
+    this._signature = graph.signature;
+    this._functions = graph.functions;
+    if (graph.functions != null) {
+      Object.keys(graph.functions).forEach((name) => {
+        this._functionExecutorMap[name] = new GraphExecutor(graph.functions[name], this);
+      });
+    }
+  }
+  get weightIds() {
+    return this.parent ? this.parent.weightIds : this._weightIds;
+  }
+  get functionExecutorMap() {
+    return this.parent ? this.parent.functionExecutorMap : this._functionExecutorMap;
+  }
+  get weightMap() {
+    return this.parent ? this.parent.weightMap : this._weightMap;
+  }
+  set weightMap(weightMap) {
+    const weightIds = Object.keys(weightMap).map((key) => weightMap[key].map((tensor2) => tensor2.id));
+    this._weightIds = [].concat(...weightIds);
+    this._weightMap = weightMap;
+  }
+  set resourceManager(resourceManager) {
+    this._resourceManager = resourceManager;
+  }
+  get inputs() {
+    return this._inputs.map((node) => {
+      return {
+        name: node.name,
+        shape: node.attrParams["shape"] ? node.attrParams["shape"].value : void 0,
+        dtype: node.attrParams["dtype"] ? node.attrParams["dtype"].value : void 0
+      };
+    });
+  }
+  get outputs() {
+    return this._outputs.map((node) => {
+      return {
+        name: node.name,
+        shape: node.attrParams["shape"] ? node.attrParams["shape"].value : void 0,
+        dtype: node.attrParams["dtype"] ? node.attrParams["dtype"].value : void 0
+      };
+    });
+  }
+  get inputNodes() {
+    return this._inputs.map((node) => node.signatureKey || node.name);
+  }
+  get outputNodes() {
+    return this._outputs.map((node) => {
+      const name = node.signatureKey || node.name;
+      return node.defaultOutput ? `${name}:${node.defaultOutput}` : name;
+    });
+  }
+  get functions() {
+    return Object.keys(this._functions).reduce((map, key) => {
+      map[key] = this._functions[key].signature;
+      return map;
+    }, {});
+  }
+  getCompilationKey(inputs, outputs) {
+    const sortedInputs = inputs.map((node) => node.name).sort();
+    const sortedOutputs = outputs.map((node) => node.name).sort();
+    return sortedInputs.join(this.SEPERATOR) + "--" + sortedOutputs.join(this.SEPERATOR);
+  }
+  compile(inputs, outputs) {
+    const executionInfo = getExecutionSubgraph(inputs, outputs, this.weightMap, this._initNodes);
+    const { missingInputs, dynamicNode, syncInputs } = executionInfo;
+    if (dynamicNode != null) {
+      throw new Error(`This execution contains the node '${dynamicNode.name}', which has the dynamic op '${dynamicNode.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${syncInputs}]`);
+    }
+    if (missingInputs.length > 0) {
+      const outNames = outputs.map((n) => n.name);
+      const inNames = Object.keys(inputs);
+      throw new Error(`Cannot compute the outputs [${outNames}] from the provided inputs [${inNames}]. Missing the following inputs: [${missingInputs}]`);
+    }
+    return getNodesInTopologicalOrder(this.graph, this.weightMap, executionInfo);
+  }
+  execute(inputs, outputs) {
+    inputs = this.mapInputs(inputs);
+    const names = Object.keys(inputs).sort();
+    this.checkInputs(inputs);
+    this.checkInputShapeAndType(inputs);
+    outputs = this.mapOutputs(outputs);
+    this.checkOutputs(outputs);
+    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);
+    const outputNodeNames = outputs.map((name) => parseNodeName(name)[0]);
+    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);
+    this.resetIntermediateTensors();
+    if (outputNodes.length === 0) {
+      outputNodes = this._outputs;
+    }
+    const compilationKey = this.getCompilationKey(inputNodes, outputNodes);
+    let orderedNodes = this.compiledMap.get(compilationKey);
+    if (orderedNodes == null) {
+      orderedNodes = this.compile(inputs, outputNodes);
+      this.compiledMap.set(compilationKey, orderedNodes);
+    }
+    const tensorArrayMap = {};
+    const tensorListMap = {};
+    return tidy(() => {
+      const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);
+      const tensorsMap = Object.assign({}, this.weightMap);
+      Object.keys(inputs).forEach((name) => {
+        const [nodeName, index] = parseNodeName(name);
+        const tensors = [];
+        tensors[index] = inputs[name];
+        tensorsMap[nodeName] = tensors;
+      });
+      const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);
+      const intermediateTensorConsumerCount = {};
+      for (let i = 0; i < orderedNodes.length; i++) {
+        const node = orderedNodes[i];
+        if (!tensorsMap[node.name]) {
+          const tensors = executeOp20(node, tensorsMap, context, this._resourceManager);
+          if (util_exports.isPromise(tensors)) {
+            throw new Error(`The execution of the op '${node.op}' returned a promise. Please use model.executeAsync() instead.`);
+          }
+          tensorsMap[node.name] = tensors;
+          this.checkTensorForDisposal(node.name, node, tensorsMap, context, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount);
+        }
+      }
+      if (this.parent == null) {
+        context.dispose(tensorsToKeep);
+      }
+      return outputs.map((name) => getTensor(name, tensorsMap, context));
+    });
+  }
+  getFrozenTensorIds(tensorMap) {
+    const ids = [].concat.apply([], Object.keys(tensorMap).map((key) => tensorMap[key]).map((tensors) => tensors.map((tensor2) => tensor2.id)));
+    return new Set(ids);
+  }
+  checkTensorForDisposal(nodeName, node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount) {
+    if (node.category === "control" || outputNames.indexOf(nodeName) !== -1) {
+      return;
+    }
+    tensorMap[nodeName].forEach((tensor2) => {
+      if (tensor2 != null) {
+        intermediateTensorConsumerCount[tensor2.id] = (intermediateTensorConsumerCount[tensor2.id] || 0) + node.children.length;
+      }
+    });
+    node.inputs.forEach((input2) => {
+      if (input2.category !== "control") {
+        const tensors = getTensorsForCurrentContenxt(input2.name, tensorMap, context);
+        if (tensors != null) {
+          tensors.forEach((tensor2) => {
+            if (tensor2 && !tensor2.kept && !tensorsToKeep.has(tensor2.id)) {
+              const count2 = intermediateTensorConsumerCount[tensor2.id];
+              if (count2 === 1) {
+                if (!this.keepTensorForDebug) {
+                  tensor2.dispose();
+                } else {
+                  const [nodeName2, index] = getNodeNameAndIndex(node.name, context);
+                  if (this.intermediateTensors[nodeName2]) {
+                    this.intermediateTensors[nodeName2][index] = tensor2;
+                  } else {
+                    this.intermediateTensors[nodeName2] = [];
+                    this.intermediateTensors[nodeName2][index] = tensor2;
+                  }
+                }
+                delete intermediateTensorConsumerCount[tensor2.id];
+              } else if (count2 != null) {
+                intermediateTensorConsumerCount[tensor2.id]--;
+              }
+            }
+          });
+        }
+      }
+    });
+  }
+  async executeAsync(inputs, outputs) {
+    return this._executeAsync(inputs, outputs);
+  }
+  disposeIntermediateTensors() {
+    if (!this.intermediateTensors) {
+      return;
+    }
+    Object.keys(this.intermediateTensors).forEach((key) => this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose()));
+    this.disposeTensorsMap();
+  }
+  disposeTensorsMap() {
+    if (!this.tensorsMap) {
+      return;
+    }
+    Object.keys(this.tensorsMap).forEach((key) => {
+      const tensorArray = this.tensorsMap[key];
+      tensorArray.forEach((tensor2) => {
+        if (tensor2 && !tensor2.kept && !tensor2.isDisposed && !this.keepIds.has(tensor2.id)) {
+          tensor2.dispose();
+        }
+      });
+    });
+  }
+  getIntermediateTensors() {
+    return this.tensorsMap;
+  }
+  resetIntermediateTensors() {
+    for (const key in this.intermediateTensors) {
+      this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose());
+      delete this.intermediateTensors[key];
+    }
+  }
+  async _executeAsync(inputs, outputs, isFunctionExecution = false, tensorArrayMap = {}, tensorListMap = {}) {
+    if (!isFunctionExecution) {
+      inputs = this.mapInputs(inputs);
+      this.checkInputs(inputs);
+      this.checkInputShapeAndType(inputs);
+      outputs = this.mapOutputs(outputs);
+      this.checkOutputs(outputs);
+    }
+    try {
+      this.keepTensorForDebug = env().getBool("KEEP_INTERMEDIATE_TENSORS");
+    } catch (e) {
+      console.warn(e.message);
+    }
+    this.resetIntermediateTensors();
+    const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);
+    this.tensorsMap = await this.executeWithControlFlow(inputs, context, outputs, isFunctionExecution);
+    const results = outputs.map((name) => getTensor(name, this.tensorsMap, context));
+    const outputIds = results.map((t) => t.id);
+    const inputIds = Object.keys(inputs).map((name) => inputs[name].id);
+    this.keepIds = /* @__PURE__ */ new Set([...outputIds, ...inputIds, ...this.weightIds]);
+    if (!this.keepTensorForDebug) {
+      this.disposeTensorsMap();
+    }
+    if (this.parent == null) {
+      context.dispose(this.keepIds);
+    }
+    return results;
+  }
+  async executeFunctionAsync(inputs, tensorArrayMap, tensorListMap) {
+    const mappedInputs = inputs.reduce((map, tensor2, index) => {
+      map[this.inputs[index].name] = tensor2;
+      return map;
+    }, {});
+    return this._executeAsync(mappedInputs, this.outputNodes, true, tensorArrayMap, tensorListMap);
+  }
+  async executeWithControlFlow(inputs, context, outputNames, isFunctionExecution) {
+    const names = Object.keys(inputs);
+    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);
+    const outputNodeNames = outputNames.map((name) => parseNodeName(name)[0]);
+    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);
+    if (outputNodes.length === 0) {
+      outputNodes = this._outputs;
+    }
+    const { usedNodes, missingInputs, dynamicNode, syncInputs } = getExecutionSubgraph(inputs, outputNodes, this.weightMap, this._initNodes);
+    const stack2 = [
+      ...inputNodes,
+      ...this.graph.weights,
+      ...this._initNodes || []
+    ].map((node) => {
+      return { node, contexts: context.currentContext };
+    });
+    const tensorsMap = Object.assign({}, this.weightMap);
+    Object.keys(inputs).forEach((name) => {
+      const [nodeName, index] = parseNodeName(name);
+      const tensors = [];
+      tensors[index] = inputs[name];
+      tensorsMap[nodeName] = tensors;
+    });
+    const intermediateTensorConsumerCount = {};
+    const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);
+    const added = {};
+    while (stack2.length > 0) {
+      const promises = this.processStack(inputNodes, stack2, context, tensorsMap, added, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount, usedNodes);
+      await Promise.all(promises);
+    }
+    if (dynamicNode == null && !isFunctionExecution) {
+      console.warn(`This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.`);
+    }
+    const missingOutputs = outputNodes.filter((node) => !isControlFlow(node) && !getTensor(node.name, tensorsMap, context)).map((node) => node.name);
+    if (missingOutputs.length > 0) {
+      let alternativeMsg = "";
+      if (dynamicNode != null) {
+        alternativeMsg = `Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${syncInputs}]`;
+      }
+      throw new Error(`Cannot compute the outputs [${missingOutputs}] from the provided inputs [${names}]. Consider providing the following inputs: [${missingInputs}]. ${alternativeMsg}`);
+    }
+    return tensorsMap;
+  }
+  processStack(inputNodes, stack2, context, tensorMap, added, tensorsToKeep, outputNames, intermediateTensorConsumerCount, usedNodes) {
+    const promises = [];
+    while (stack2.length > 0) {
+      const item = stack2.pop();
+      context.currentContext = item.contexts;
+      let nodeName = "";
+      if (item.node.op === "Enter" && getParamValue("isConstant", item.node, tensorMap, context)) {
+        [nodeName] = getNodeNameAndIndex(item.node.name, context);
+      }
+      if (tensorMap[item.node.name] == null) {
+        const tensors = executeOp20(item.node, tensorMap, context, this._resourceManager);
+        if (!nodeName) {
+          [nodeName] = getNodeNameAndIndex(item.node.name, context);
+        }
+        const currentContext = context.currentContext;
+        if (util_exports.isPromise(tensors)) {
+          promises.push(tensors.then((t) => {
+            tensorMap[nodeName] = t;
+            context.currentContext = currentContext;
+            this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);
+            this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+            return t;
+          }));
+        } else {
+          tensorMap[nodeName] = tensors;
+          this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);
+          this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+        }
+      } else {
+        this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+      }
+    }
+    return promises;
+  }
+  processChildNodes(node, stack2, context, tensorMap, added, usedNodes) {
+    node.children.forEach((childNode) => {
+      const [nodeName] = getNodeNameAndIndex(childNode.name, context);
+      if (added[nodeName] || !usedNodes.has(childNode.name)) {
+        return;
+      }
+      if (childNode.op === "Merge") {
+        if (childNode.inputNames.some((name) => {
+          return !!getTensor(name, tensorMap, context);
+        })) {
+          added[nodeName] = true;
+          stack2.push({ contexts: context.currentContext, node: childNode });
+        }
+      } else if (childNode.inputNames.every((name) => {
+        return !!getTensor(name, tensorMap, context);
+      })) {
+        added[nodeName] = true;
+        stack2.push({ contexts: context.currentContext, node: childNode });
+      }
+    });
+  }
+  dispose() {
+    Object.keys(this.weightMap).forEach((key) => this.weightMap[key].forEach((tensor2) => tensor2.dispose()));
+  }
+  checkInputShapeAndType(inputs) {
+    Object.keys(inputs).forEach((name) => {
+      const input2 = inputs[name];
+      const [nodeName] = parseNodeName(name);
+      const node = this.graph.nodes[nodeName];
+      if (node.attrParams["shape"] && node.attrParams["shape"].value) {
+        const shape = node.attrParams["shape"].value;
+        const match = shape.length === input2.shape.length && input2.shape.every((dim, index) => shape[index] === -1 || shape[index] === dim);
+        util_exports.assert(match, () => `The shape of dict['${node.name}'] provided in model.execute(dict) must be [${shape}], but was [${input2.shape}]`);
+      }
+      if (node.attrParams["dtype"] && node.attrParams["dtype"].value) {
+        util_exports.assert(input2.dtype === node.attrParams["dtype"].value, () => `The dtype of dict['${node.name}'] provided in model.execute(dict) must be ${node.attrParams["dtype"].value}, but was ${input2.dtype}`);
+      }
+    });
+  }
+  mapInputs(inputs) {
+    const result = {};
+    for (const inputName in inputs) {
+      if (this._signature != null && this._signature.inputs != null && this._signature.inputs[inputName] != null) {
+        const tensor2 = this._signature.inputs[inputName];
+        result[tensor2.name] = inputs[inputName];
+      } else {
+        result[inputName] = inputs[inputName];
+      }
+    }
+    return result;
+  }
+  checkInputs(inputs) {
+    const notInGraph = Object.keys(inputs).filter((name) => {
+      const [nodeName] = parseNodeName(name);
+      return this.graph.nodes[nodeName] == null;
+    });
+    if (notInGraph.length > 0) {
+      throw new Error(`The dict provided in model.execute(dict) has keys: [${notInGraph}] that are not part of graph`);
+    }
+  }
+  mapOutputs(outputs) {
+    return outputs.map((name) => {
+      if (this._signature != null && this._signature.outputs != null && this._signature.outputs[name] != null) {
+        const tensor2 = this._signature.outputs[name];
+        return tensor2.name;
+      }
+      return name;
+    }, {});
+  }
+  checkOutputs(outputs) {
+    outputs.forEach((name) => {
+      const [normalizedName] = parseNodeName(name);
+      if (!this.graph.nodes[normalizedName]) {
+        throw new Error(`The output '${name}' is not found in the graph`);
+      }
+    });
+  }
+};
+var ResourceManager = class {
+  constructor(hashTableNameToHandle = {}, hashTableMap = {}) {
+    this.hashTableNameToHandle = hashTableNameToHandle;
+    this.hashTableMap = hashTableMap;
+  }
+  addHashTable(name, hashTable) {
+    this.hashTableNameToHandle[name] = hashTable.handle;
+    this.hashTableMap[hashTable.id] = hashTable;
+  }
+  getHashTableHandleByName(name) {
+    return this.hashTableNameToHandle[name];
+  }
+  getHashTableById(id) {
+    return this.hashTableMap[id];
+  }
+  dispose() {
+    for (const key in this.hashTableMap) {
+      this.hashTableMap[key].clearAndClose();
+      delete this.hashTableMap[key];
+    }
+    for (const name in this.hashTableNameToHandle) {
+      this.hashTableNameToHandle[name].dispose();
+      delete this.hashTableNameToHandle[name];
+    }
+  }
+};
+var TFHUB_SEARCH_PARAM = "?tfjs-format=file";
+var DEFAULT_MODEL_NAME = "model.json";
+var GraphModel = class {
+  constructor(modelUrl, loadOptions = {}, tfio = io_exports) {
+    this.modelUrl = modelUrl;
+    this.loadOptions = loadOptions;
+    this.version = "n/a";
+    this.io = tfio;
+    if (loadOptions == null) {
+      this.loadOptions = {};
+    }
+    this.resourceManager = new ResourceManager();
+  }
+  get modelVersion() {
+    return this.version;
+  }
+  get inputNodes() {
+    return this.executor.inputNodes;
+  }
+  get outputNodes() {
+    return this.executor.outputNodes;
+  }
+  get inputs() {
+    return this.executor.inputs;
+  }
+  get outputs() {
+    return this.executor.outputs;
+  }
+  get weights() {
+    return this.executor.weightMap;
+  }
+  get metadata() {
+    return this.artifacts.userDefinedMetadata;
+  }
+  get modelSignature() {
+    return this.signature;
+  }
+  get modelStructuredOutputKeys() {
+    return this.structuredOutputKeys;
+  }
+  findIOHandler() {
+    const path = this.modelUrl;
+    if (path.load != null) {
+      this.handler = path;
+    } else if (this.loadOptions.requestInit != null) {
+      this.handler = this.io.browserHTTPRequest(path, this.loadOptions);
+    } else {
+      const handlers = this.io.getLoadHandlers(path, this.loadOptions);
+      if (handlers.length === 0) {
+        handlers.push(this.io.browserHTTPRequest(path, this.loadOptions));
+      } else if (handlers.length > 1) {
+        throw new Error(`Found more than one (${handlers.length}) load handlers for URL '${[path]}'`);
+      }
+      this.handler = handlers[0];
+    }
+  }
+  load() {
+    this.findIOHandler();
+    if (this.handler.load == null) {
+      throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+    }
+    const loadResult = this.handler.load();
+    if (util_exports.isPromise(loadResult)) {
+      return loadResult.then((artifacts) => this.loadSync(artifacts));
+    }
+    return this.loadSync(loadResult);
+  }
+  loadSync(artifacts) {
+    this.artifacts = artifacts;
+    const graph = this.artifacts.modelTopology;
+    let signature = this.artifacts.signature;
+    if (this.artifacts.userDefinedMetadata != null) {
+      const metadata = this.artifacts.userDefinedMetadata;
+      if (metadata.signature != null) {
+        signature = metadata.signature;
+      }
+      if (metadata.structuredOutputKeys != null) {
+        this.structuredOutputKeys = metadata.structuredOutputKeys;
+      }
+    }
+    this.signature = signature;
+    this.version = `${graph.versions.producer}.${graph.versions.minConsumer}`;
+    const weightMap = this.io.decodeWeights(this.artifacts.weightData, this.artifacts.weightSpecs);
+    this.executor = new GraphExecutor(OperationMapper.Instance.transformGraph(graph, this.signature));
+    this.executor.weightMap = this.convertTensorMapToTensorsMap(weightMap);
+    this.executor.resourceManager = this.resourceManager;
+    if (artifacts.modelInitializer != null && artifacts.modelInitializer.node != null) {
+      const initializer = OperationMapper.Instance.transformGraph(artifacts.modelInitializer);
+      this.initializer = new GraphExecutor(initializer);
+      this.initializer.weightMap = this.executor.weightMap;
+      this.initializer.resourceManager = this.resourceManager;
+      this.initializerSignature = artifacts.initializerSignature;
+    }
+    return true;
+  }
+  async save(handlerOrURL, config) {
+    if (typeof handlerOrURL === "string") {
+      const handlers = this.io.getSaveHandlers(handlerOrURL);
+      if (handlers.length === 0) {
+        throw new Error(`Cannot find any save handlers for URL '${handlerOrURL}'`);
+      } else if (handlers.length > 1) {
+        throw new Error(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);
+      }
+      handlerOrURL = handlers[0];
+    }
+    if (handlerOrURL.save == null) {
+      throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+    }
+    return handlerOrURL.save(this.artifacts);
+  }
+  predict(inputs, config) {
+    const outputTensors = this.execute(inputs, this.outputNodes);
+    if (this.structuredOutputKeys) {
+      const outputTensorsArray = outputTensors instanceof Tensor ? [outputTensors] : outputTensors;
+      const outputTensorMap = {};
+      outputTensorsArray.forEach((outputTensor, i) => outputTensorMap[this.structuredOutputKeys[i]] = outputTensor);
+      return outputTensorMap;
+    }
+    return outputTensors;
+  }
+  normalizeInputs(inputs) {
+    if (!(inputs instanceof Tensor) && !Array.isArray(inputs)) {
+      if (this.signature != null && this.signature.inputs != null) {
+        for (const input2 in this.signature.inputs) {
+          const tensor2 = this.signature.inputs[input2];
+          if (tensor2.resourceId != null) {
+            inputs[input2] = this.resourceIdToCapturedInput[tensor2.resourceId];
+          }
+        }
+      }
+      return inputs;
+    }
+    inputs = Array.isArray(inputs) ? inputs : [inputs];
+    const numCapturedInputs = Object.keys(this.resourceIdToCapturedInput).length;
+    if (inputs.length + numCapturedInputs !== this.inputNodes.length) {
+      throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length - numCapturedInputs} non-resource placeholders, while there are ${inputs.length} input tensors provided.`);
+    }
+    let inputIndex = 0;
+    return this.inputNodes.reduce((map, inputName) => {
+      const signature = this.signature ? this.signature.inputs[inputName] : null;
+      if (signature != null && signature.resourceId != null) {
+        map[inputName] = this.resourceIdToCapturedInput[signature.resourceId];
+      } else {
+        map[inputName] = inputs[inputIndex++];
+      }
+      return map;
+    }, {});
+  }
+  normalizeOutputs(outputs) {
+    outputs = outputs || this.outputNodes;
+    return !Array.isArray(outputs) ? [outputs] : outputs;
+  }
+  executeInitializerGraph() {
+    if (this.initializer == null) {
+      return [];
+    }
+    if (this.initializerSignature == null) {
+      return this.initializer.execute({}, []);
+    } else {
+      return this.initializer.execute({}, Object.keys(this.initializerSignature.outputs));
+    }
+  }
+  async executeInitializerGraphAsync() {
+    if (this.initializer == null) {
+      return [];
+    }
+    if (this.initializerSignature == null) {
+      return this.initializer.executeAsync({}, []);
+    } else {
+      return this.initializer.executeAsync({}, Object.keys(this.initializerSignature.outputs));
+    }
+  }
+  setResourceIdToCapturedInput(outputs) {
+    this.resourceIdToCapturedInput = {};
+    if (this.initializerSignature) {
+      const outputNames = Object.keys(this.initializerSignature.outputs);
+      for (let i = 0; i < outputNames.length; i++) {
+        const outputName = outputNames[i];
+        const tensorInfo = this.initializerSignature.outputs[outputName];
+        this.resourceIdToCapturedInput[tensorInfo.resourceId] = outputs[i];
+      }
+    }
+  }
+  execute(inputs, outputs) {
+    if (this.resourceIdToCapturedInput == null) {
+      this.setResourceIdToCapturedInput(this.executeInitializerGraph());
+    }
+    inputs = this.normalizeInputs(inputs);
+    outputs = this.normalizeOutputs(outputs);
+    const result = this.executor.execute(inputs, outputs);
+    return result.length > 1 ? result : result[0];
+  }
+  async executeAsync(inputs, outputs) {
+    if (this.resourceIdToCapturedInput == null) {
+      this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync());
+    }
+    inputs = this.normalizeInputs(inputs);
+    outputs = this.normalizeOutputs(outputs);
+    const result = await this.executor.executeAsync(inputs, outputs);
+    return result.length > 1 ? result : result[0];
+  }
+  getIntermediateTensors() {
+    return this.executor.getIntermediateTensors();
+  }
+  disposeIntermediateTensors() {
+    this.executor.disposeIntermediateTensors();
+  }
+  convertTensorMapToTensorsMap(map) {
+    return Object.keys(map).reduce((newMap, key) => {
+      newMap[key] = [map[key]];
+      return newMap;
+    }, {});
+  }
+  dispose() {
+    this.executor.dispose();
+    if (this.initializer) {
+      this.initializer.dispose();
+      if (this.resourceIdToCapturedInput) {
+        dispose(this.resourceIdToCapturedInput);
+      }
+    }
+    this.resourceManager.dispose();
+  }
+};
+async function loadGraphModel(modelUrl, options = {}, tfio = io_exports) {
+  if (modelUrl == null) {
+    throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");
+  }
+  if (options == null) {
+    options = {};
+  }
+  if (options.fromTFHub && typeof modelUrl === "string") {
+    modelUrl = getTFHubUrl(modelUrl);
+  }
+  const model2 = new GraphModel(modelUrl, options, tfio);
+  await model2.load();
+  return model2;
+}
+function loadGraphModelSync(modelSource) {
+  if (modelSource == null) {
+    throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");
+  }
+  let ioHandler;
+  if (modelSource instanceof Array) {
+    const [modelJSON, weights] = modelSource;
+    if (!modelJSON) {
+      throw new Error("modelJSON must be the first element of the array");
+    }
+    if (!weights || !(weights instanceof ArrayBuffer)) {
+      throw new Error("An ArrayBuffer of weights must be the second element of the array");
+    }
+    if (!("modelTopology" in modelJSON)) {
+      throw new Error("Model JSON is missing 'modelTopology'");
+    }
+    if (!("weightsManifest" in modelJSON)) {
+      throw new Error("Model JSON is missing 'weightsManifest'");
+    }
+    const weightSpecs = io_exports.getWeightSpecs(modelJSON.weightsManifest);
+    const modelArtifacts = io_exports.getModelArtifactsForJSONSync(modelJSON, weightSpecs, weights);
+    ioHandler = io_exports.fromMemorySync(modelArtifacts);
+  } else if ("load" in modelSource) {
+    ioHandler = modelSource;
+  } else if ("modelTopology" in modelSource && "weightSpecs" in modelSource && "weightData" in modelSource) {
+    ioHandler = io_exports.fromMemorySync(modelSource);
+  } else {
+    throw new Error("Unknown model format");
+  }
+  const model2 = new GraphModel(ioHandler);
+  model2.load();
+  return model2;
+}
+function getTFHubUrl(modelUrl) {
+  if (!modelUrl.endsWith("/")) {
+    modelUrl = modelUrl + "/";
+  }
+  return `${modelUrl}${DEFAULT_MODEL_NAME}${TFHUB_SEARCH_PARAM}`;
+}
+var version3 = "4.0.0";
+var dist_exports2 = {};
+__export2(dist_exports2, {
+  CSVDataset: () => CSVDataset,
+  Dataset: () => Dataset,
+  FileDataSource: () => FileDataSource,
+  TextLineDataset: () => TextLineDataset,
+  URLDataSource: () => URLDataSource,
+  array: () => array,
+  csv: () => csv,
+  func: () => func,
+  generator: () => generator,
+  microphone: () => microphone,
+  version_data: () => version4,
+  webcam: () => webcam,
+  zip: () => zip
+});
+var seedrandom3 = __toESM(require_seedrandom2());
+var seedrandom2 = __toESM(require_seedrandom2());
+function deepMap(input2, mapFn) {
+  return deepMapInternal(input2, mapFn);
+}
+function deepMapInternal(input2, mapFn, seen = /* @__PURE__ */ new Map(), containedIn = /* @__PURE__ */ new Set()) {
+  if (input2 == null) {
+    return null;
+  }
+  if (typeof Blob === "function" && input2 instanceof Blob) {
+    return input2.slice();
+  }
+  if (containedIn.has(input2)) {
+    throw new Error("Circular references are not supported.");
+  }
+  if (seen.has(input2)) {
+    return seen.get(input2);
+  }
+  const result = mapFn(input2);
+  if (result.recurse && result.value !== null) {
+    throw new Error("A deep map function may not return both a value and recurse=true.");
+  }
+  if (!result.recurse) {
+    seen.set(input2, result.value);
+    return result.value;
+  } else if (isIterable2(input2)) {
+    const mappedIterable = Array.isArray(input2) ? [] : {};
+    containedIn.add(input2);
+    for (const k in input2) {
+      const child = input2[k];
+      const childResult = deepMapInternal(child, mapFn, seen, containedIn);
+      mappedIterable[k] = childResult;
+    }
+    containedIn.delete(input2);
+    if (input2.__proto__) {
+      mappedIterable.__proto__ = input2.__proto__;
+    }
+    return mappedIterable;
+  } else {
+    throw new Error(`Can't recurse into non-iterable type: ${input2}`);
+  }
+}
+function deepZip(inputs, zipFn = zipToList) {
+  return deepZipInternal(inputs, zipFn);
+}
+function deepZipInternal(inputs, zipFn, containedIn = /* @__PURE__ */ new Set()) {
+  const input2 = inputs[0];
+  if (containedIn.has(input2)) {
+    throw new Error("Circular references are not supported.");
+  }
+  const result = zipFn(inputs);
+  if (result.recurse && result.value !== null) {
+    throw new Error("A deep zip function may not return both a value and recurse=true.");
+  }
+  if (!result.recurse) {
+    return result.value;
+  } else if (isIterable2(input2)) {
+    const mappedIterable = Array.isArray(input2) ? [] : {};
+    containedIn.add(input2);
+    for (const k in input2) {
+      const children = inputs.map((x) => x[k]);
+      const childResult = deepZipInternal(children, zipFn, containedIn);
+      mappedIterable[k] = childResult;
+    }
+    containedIn.delete(input2);
+    return mappedIterable;
+  } else {
+    throw new Error(`Can't recurse into non-iterable type: ${input2}`);
+  }
+}
+function zipToList(x) {
+  if (x === null) {
+    return null;
+  }
+  if (isIterable2(x[0])) {
+    return { value: null, recurse: true };
+  } else {
+    return { value: x, recurse: false };
+  }
+}
+async function deepMapAndAwaitAll(input2, mapFn) {
+  const seen = /* @__PURE__ */ new Map();
+  deepMapInternal(input2, mapFn, seen);
+  for (const key of Array.from(seen.keys())) {
+    const value = seen.get(key);
+    if (util_exports.isPromise(value)) {
+      const mappedValue = await value;
+      seen.set(key, mappedValue);
+    }
+  }
+  const result = deepMapInternal(input2, mapFn, seen);
+  return result;
+}
+function isIterable2(obj) {
+  let isTextDecoder = false;
+  if (env().get("IS_BROWSER")) {
+    isTextDecoder = obj instanceof TextDecoder;
+  } else {
+    const { StringDecoder } = require_string_decoder();
+    isTextDecoder = obj instanceof StringDecoder;
+  }
+  return obj != null && !ArrayBuffer.isView(obj) && (Array.isArray(obj) || typeof obj === "object" && !(obj instanceof Tensor) && !(obj instanceof Promise) && !isTextDecoder);
+}
+function canTensorify(obj) {
+  return obj == null || isPrimitive(obj) || Array.isArray(obj) || typeof obj === "object" && obj instanceof Tensor || util_exports.isTypedArray(obj);
+}
+function isPrimitive(value) {
+  return value === null || typeof value !== "object" && typeof value !== "function";
+}
+function deepClone(container) {
+  return deepMap(container, cloneIfTensor);
+}
+function cloneIfTensor(item) {
+  if (item instanceof Tensor) {
+    return { value: item.clone(), recurse: false };
+  } else if (isIterable2(item)) {
+    return { value: null, recurse: true };
+  } else {
+    return { value: item, recurse: false };
+  }
+}
+var RingBuffer = class {
+  constructor(capacity) {
+    this.capacity = capacity;
+    this.begin = 0;
+    this.end = 0;
+    if (capacity == null) {
+      throw new RangeError("Can't create a ring buffer of unknown capacity.");
+    }
+    if (capacity < 1) {
+      throw new RangeError("Can't create ring buffer of capacity < 1.");
+    }
+    this.data = new Array(capacity);
+    this.doubledCapacity = 2 * capacity;
+  }
+  wrap(index) {
+    while (index < 0) {
+      index += this.doubledCapacity;
+    }
+    return index % this.doubledCapacity;
+  }
+  get(index) {
+    if (index < 0) {
+      throw new RangeError("Can't get item at a negative index.");
+    }
+    return this.data[index % this.capacity];
+  }
+  set(index, value) {
+    if (index < 0) {
+      throw new RangeError("Can't set item at a negative index.");
+    }
+    this.data[index % this.capacity] = value;
+  }
+  length() {
+    let length = this.end - this.begin;
+    if (length < 0) {
+      length = this.doubledCapacity + length;
+    }
+    return length;
+  }
+  isFull() {
+    return this.length() === this.capacity;
+  }
+  isEmpty() {
+    return this.length() === 0;
+  }
+  push(value) {
+    if (this.isFull()) {
+      throw new RangeError("Ring buffer is full.");
+    }
+    this.set(this.end, value);
+    this.end = this.wrap(this.end + 1);
+  }
+  pushAll(values) {
+    for (const value of values) {
+      this.push(value);
+    }
+  }
+  pop() {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    this.end = this.wrap(this.end - 1);
+    const result = this.get(this.end);
+    this.set(this.end, void 0);
+    return result;
+  }
+  unshift(value) {
+    if (this.isFull()) {
+      throw new RangeError("Ring buffer is full.");
+    }
+    this.begin = this.wrap(this.begin - 1);
+    this.set(this.begin, value);
+  }
+  shift() {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    const result = this.get(this.begin);
+    this.set(this.begin, void 0);
+    this.begin = this.wrap(this.begin + 1);
+    return result;
+  }
+  shuffleExcise(relativeIndex) {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    const index = this.wrap(this.begin + relativeIndex);
+    const result = this.get(index);
+    this.set(index, this.pop());
+    return result;
+  }
+};
+var GrowingRingBuffer = class extends RingBuffer {
+  constructor() {
+    super(GrowingRingBuffer.INITIAL_CAPACITY);
+  }
+  isFull() {
+    return false;
+  }
+  push(value) {
+    if (super.isFull()) {
+      this.expand();
+    }
+    super.push(value);
+  }
+  unshift(value) {
+    if (super.isFull()) {
+      this.expand();
+    }
+    super.unshift(value);
+  }
+  expand() {
+    const newCapacity = this.capacity * 2;
+    const newData = new Array(newCapacity);
+    const len = this.length();
+    for (let i = 0; i < len; i++) {
+      newData[i] = this.get(this.wrap(this.begin + i));
+    }
+    this.data = newData;
+    this.capacity = newCapacity;
+    this.doubledCapacity = 2 * this.capacity;
+    this.begin = 0;
+    this.end = len;
+  }
+};
+GrowingRingBuffer.INITIAL_CAPACITY = 32;
+function iteratorFromItems(items) {
+  return new ArrayIterator(items);
+}
+function iteratorFromFunction(func2) {
+  return new FunctionCallIterator(func2);
+}
+function iteratorFromConcatenated(baseIterators, baseErrorHandler) {
+  return new ChainedIterator(baseIterators, baseErrorHandler);
+}
+function iteratorFromZipped(iterators, mismatchMode = ZipMismatchMode.FAIL) {
+  return new ZipIterator(iterators, mismatchMode);
+}
+var LazyIterator = class {
+  async toArray() {
+    const result = [];
+    let x = await this.next();
+    while (!x.done) {
+      result.push(x.value);
+      x = await this.next();
+    }
+    return result;
+  }
+  async toArrayForTest() {
+    const stream = this.prefetch(100);
+    const result = [];
+    let x = await stream.next();
+    while (!x.done) {
+      result.push(x.value);
+      x = await stream.next();
+    }
+    return result;
+  }
+  async resolveFully() {
+    let x = await this.next();
+    while (!x.done) {
+      x = await this.next();
+    }
+  }
+  async resolveWhile(predicate) {
+    let x = await this.next();
+    let shouldContinue = predicate(x.value);
+    while (!x.done && shouldContinue) {
+      x = await this.next();
+      shouldContinue = predicate(x.value);
+    }
+  }
+  handleErrors(handler) {
+    return new ErrorHandlingLazyIterator(this, handler);
+  }
+  filter(predicate) {
+    return new FilterIterator(this, predicate);
+  }
+  map(transform5) {
+    return new MapIterator(this, transform5);
+  }
+  mapAsync(transform5) {
+    return new AsyncMapIterator(this, transform5);
+  }
+  serialMapAsync(transform5) {
+    return new AsyncMapIterator(this, transform5).serial();
+  }
+  flatmap(transform5) {
+    return new FlatmapIterator(this, transform5);
+  }
+  async forEachAsync(f) {
+    return this.map(f).resolveFully();
+  }
+  async serialForEach(f) {
+    return this.serialMapAsync(f).resolveWhile((x) => x === true);
+  }
+  rowMajorBatch(batchSize, smallLastBatch = true) {
+    return new RowMajorBatchIterator(this, batchSize, smallLastBatch);
+  }
+  columnMajorBatch(batchSize, smallLastBatch = true, zipFn = zipToList) {
+    const rowBatches = this.rowMajorBatch(batchSize, smallLastBatch);
+    return rowBatches.map((x) => deepZip(x, zipFn));
+  }
+  concatenate(iterator, baseErrorHandler) {
+    return new ChainedIterator(iteratorFromItems([this, iterator]), baseErrorHandler);
+  }
+  take(count2) {
+    if (count2 < 0 || count2 == null) {
+      return this;
+    }
+    return new TakeIterator(this, count2);
+  }
+  skip(count2) {
+    if (count2 < 0 || count2 == null) {
+      return this;
+    }
+    return new SkipIterator(this, count2);
+  }
+  prefetch(bufferSize) {
+    return new PrefetchIterator(this, bufferSize);
+  }
+  shuffle(windowSize, seed) {
+    return new ShuffleIterator(this, windowSize, seed);
+  }
+  serial() {
+    return new SerialIterator(this);
+  }
+};
+var ArrayIterator = class extends LazyIterator {
+  constructor(items) {
+    super();
+    this.items = items;
+    this.trav = 0;
+  }
+  summary() {
+    return `Array of ${this.items.length} items`;
+  }
+  async next() {
+    if (this.trav >= this.items.length) {
+      return { value: null, done: true };
+    }
+    const item = this.items[this.trav];
+    this.trav++;
+    return { value: deepClone(item), done: false };
+  }
+};
+var FunctionCallIterator = class extends LazyIterator {
+  constructor(nextFn) {
+    super();
+    this.nextFn = nextFn;
+  }
+  summary() {
+    return `Function call`;
+  }
+  async next() {
+    try {
+      return this.nextFn();
+    } catch (e) {
+      e.message = `Error thrown while iterating through a dataset: ${e.message}`;
+      throw e;
+    }
+  }
+};
+var SerialIterator = class extends LazyIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Serial`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    return this.upstream.next();
+  }
+};
+var SkipIterator = class extends LazyIterator {
+  constructor(upstream, maxCount) {
+    super();
+    this.upstream = upstream;
+    this.maxCount = maxCount;
+    this.count = 0;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Skip`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (this.count++ < this.maxCount) {
+      const skipped = await this.upstream.next();
+      if (skipped.done) {
+        return skipped;
+      }
+      dispose(skipped.value);
+    }
+    return this.upstream.next();
+  }
+};
+var TakeIterator = class extends LazyIterator {
+  constructor(upstream, maxCount) {
+    super();
+    this.upstream = upstream;
+    this.maxCount = maxCount;
+    this.count = 0;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Take`;
+  }
+  async next() {
+    if (this.count++ >= this.maxCount) {
+      return { value: null, done: true };
+    }
+    return this.upstream.next();
+  }
+};
+var RowMajorBatchIterator = class extends LazyIterator {
+  constructor(upstream, batchSize, enableSmallLastBatch = true) {
+    super();
+    this.upstream = upstream;
+    this.batchSize = batchSize;
+    this.enableSmallLastBatch = enableSmallLastBatch;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> RowMajorBatch`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    const batch = [];
+    while (batch.length < this.batchSize) {
+      const item = await this.upstream.next();
+      if (item.done) {
+        if (this.enableSmallLastBatch && batch.length > 0) {
+          return { value: batch, done: false };
+        }
+        return { value: null, done: true };
+      }
+      batch.push(item.value);
+    }
+    return { value: batch, done: false };
+  }
+};
+var FilterIterator = class extends LazyIterator {
+  constructor(upstream, predicate) {
+    super();
+    this.upstream = upstream;
+    this.predicate = predicate;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Filter`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (true) {
+      const item = await this.upstream.next();
+      if (item.done || this.predicate(item.value)) {
+        return item;
+      }
+      dispose(item.value);
+    }
+  }
+};
+var MapIterator = class extends LazyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Map`;
+  }
+  async next() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return { value: null, done: true };
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mapped = this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return { value: mapped, done: false };
+  }
+};
+var ErrorHandlingLazyIterator = class extends LazyIterator {
+  constructor(upstream, handler) {
+    super();
+    this.upstream = upstream;
+    this.handler = handler;
+    this.count = 0;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> handleErrors`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (true) {
+      try {
+        return await this.upstream.next();
+      } catch (e) {
+        if (!this.handler(e)) {
+          return { value: null, done: true };
+        }
+      }
+    }
+  }
+};
+var AsyncMapIterator = class extends LazyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> AsyncMap`;
+  }
+  async next() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return { value: null, done: true };
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mapped = await this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return { value: mapped, done: false };
+  }
+};
+var OneToManyIterator = class extends LazyIterator {
+  constructor() {
+    super();
+    this.outputQueue = new GrowingRingBuffer();
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (this.outputQueue.length() === 0) {
+      if (!await this.pump()) {
+        return { value: null, done: true };
+      }
+    }
+    return { value: this.outputQueue.shift(), done: false };
+  }
+};
+var FlatmapIterator = class extends OneToManyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Flatmap`;
+  }
+  async pump() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return false;
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mappedArray = this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mappedArray);
+    this.outputQueue.pushAll(mappedArray);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return true;
+  }
+};
+var ChainedIterator = class extends LazyIterator {
+  constructor(iterators, baseErrorHandler) {
+    super();
+    this.baseErrorHandler = baseErrorHandler;
+    this.lastRead = null;
+    this.iterator = null;
+    this.moreIterators = iterators;
+  }
+  summary() {
+    const upstreamSummaries = "TODO: fill in upstream of chained summaries";
+    return `${upstreamSummaries} -> Chained`;
+  }
+  async next() {
+    this.lastRead = this.readFromChain(this.lastRead);
+    return this.lastRead;
+  }
+  async readFromChain(lastRead) {
+    await lastRead;
+    if (this.iterator == null) {
+      const iteratorResult = await this.moreIterators.next();
+      if (iteratorResult.done) {
+        return { value: null, done: true };
+      }
+      this.iterator = iteratorResult.value;
+      if (this.baseErrorHandler != null) {
+        this.iterator = this.iterator.handleErrors(this.baseErrorHandler);
+      }
+    }
+    const itemResult = await this.iterator.next();
+    if (itemResult.done) {
+      this.iterator = null;
+      return this.readFromChain(lastRead);
+    }
+    return itemResult;
+  }
+};
+var ZipMismatchMode;
+(function(ZipMismatchMode2) {
+  ZipMismatchMode2[ZipMismatchMode2["FAIL"] = 0] = "FAIL";
+  ZipMismatchMode2[ZipMismatchMode2["SHORTEST"] = 1] = "SHORTEST";
+  ZipMismatchMode2[ZipMismatchMode2["LONGEST"] = 2] = "LONGEST";
+})(ZipMismatchMode || (ZipMismatchMode = {}));
+var ZipIterator = class extends LazyIterator {
+  constructor(iterators, mismatchMode = ZipMismatchMode.FAIL) {
+    super();
+    this.iterators = iterators;
+    this.mismatchMode = mismatchMode;
+    this.count = 0;
+    this.currentPromise = null;
+  }
+  summary() {
+    const upstreamSummaries = "TODO: fill in upstream of zip summaries";
+    return `{${upstreamSummaries}} -> Zip`;
+  }
+  async nextState(afterState) {
+    await afterState;
+    let numIterators = 0;
+    let iteratorsDone = 0;
+    function getNext(container) {
+      if (container instanceof LazyIterator) {
+        const result = container.next();
+        return {
+          value: result.then((x) => {
+            numIterators++;
+            if (x.done) {
+              iteratorsDone++;
+            }
+            return x.value;
+          }),
+          recurse: false
+        };
+      } else {
+        return { value: null, recurse: true };
+      }
+    }
+    const mapped = await deepMapAndAwaitAll(this.iterators, getNext);
+    if (numIterators === iteratorsDone) {
+      return { value: null, done: true };
+    }
+    if (iteratorsDone > 0) {
+      switch (this.mismatchMode) {
+        case ZipMismatchMode.FAIL:
+          throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);
+        case ZipMismatchMode.SHORTEST:
+          return { value: null, done: true };
+        case ZipMismatchMode.LONGEST:
+        default:
+      }
+    }
+    this.count++;
+    return { value: mapped, done: false };
+  }
+  async next() {
+    this.currentPromise = this.nextState(this.currentPromise);
+    return this.currentPromise;
+  }
+};
+var PrefetchIterator = class extends LazyIterator {
+  constructor(upstream, bufferSize) {
+    super();
+    this.upstream = upstream;
+    this.bufferSize = bufferSize;
+    this.buffer = new RingBuffer(bufferSize);
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Prefetch`;
+  }
+  refill() {
+    while (!this.buffer.isFull()) {
+      const v = this.upstream.next();
+      this.buffer.push(v);
+    }
+  }
+  next() {
+    this.refill();
+    return this.buffer.shift();
+  }
+};
+var ShuffleIterator = class extends PrefetchIterator {
+  constructor(upstream, windowSize, seed) {
+    super(upstream, windowSize);
+    this.upstream = upstream;
+    this.windowSize = windowSize;
+    this.upstreamExhausted = false;
+    this.random = seedrandom2.alea(seed || util_exports.now().toString());
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  randomInt(max6) {
+    return Math.floor(this.random() * max6);
+  }
+  chooseIndex() {
+    return this.randomInt(this.buffer.length());
+  }
+  async serialNext() {
+    if (!this.upstreamExhausted) {
+      this.refill();
+    }
+    while (!this.buffer.isEmpty()) {
+      const chosenIndex = this.chooseIndex();
+      const result = await this.buffer.shuffleExcise(chosenIndex);
+      if (result.done) {
+        this.upstreamExhausted = true;
+      } else {
+        this.refill();
+        return result;
+      }
+    }
+    return { value: null, done: true };
+  }
+};
+var Dataset = class {
+  constructor() {
+    this.size = null;
+  }
+  batch(batchSize, smallLastBatch = true) {
+    const base = this;
+    util_exports.assert(batchSize > 0, () => `batchSize needs to be positive, but it is
+      ${batchSize}`);
+    let size;
+    if (this.size === Infinity || this.size == null) {
+      size = this.size;
+    } else if (smallLastBatch) {
+      size = Math.ceil(this.size / batchSize);
+    } else {
+      size = Math.floor(this.size / batchSize);
+    }
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).columnMajorBatch(batchSize, smallLastBatch, deepBatchConcat);
+    }, size);
+  }
+  concatenate(dataset) {
+    const base = this;
+    let size;
+    if (this.size === Infinity || dataset.size === Infinity) {
+      size = Infinity;
+    } else if (this.size != null && dataset.size != null) {
+      size = this.size + dataset.size;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).concatenate(await dataset.iterator()), size);
+  }
+  filter(predicate) {
+    const base = this;
+    let size;
+    if (this.size === Infinity) {
+      size = Infinity;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).filter((x) => tidy(() => predicate(x)));
+    }, size);
+  }
+  async forEachAsync(f) {
+    return (await this.iterator()).forEachAsync(f);
+  }
+  map(transform5) {
+    const base = this;
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).map((x) => tidy(() => transform5(x)));
+    }, this.size);
+  }
+  mapAsync(transform5) {
+    const base = this;
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).mapAsync(transform5);
+    }, this.size);
+  }
+  prefetch(bufferSize) {
+    if (bufferSize == null) {
+      throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");
+    }
+    const base = this;
+    return datasetFromIteratorFn(async () => (await base.iterator()).prefetch(bufferSize), this.size);
+  }
+  repeat(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && count2 > 0) {
+      size = this.size * count2;
+    } else if (count2 === 0) {
+      size = 0;
+    } else if (this.size != null && (count2 === void 0 || count2 < 0)) {
+      size = Infinity;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => {
+      const iteratorIterator = iteratorFromFunction(async () => ({ value: await base.iterator(), done: false }));
+      return iteratorFromConcatenated(iteratorIterator.take(count2));
+    }, size);
+  }
+  skip(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && count2 >= 0 && this.size >= count2) {
+      size = this.size - count2;
+    } else if (this.size != null && (this.size < count2 || count2 === void 0 || count2 < 0)) {
+      size = 0;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).skip(count2), size);
+  }
+  shuffle(bufferSize, seed, reshuffleEachIteration = true) {
+    if (bufferSize == null || bufferSize < 0) {
+      if (this.size == null) {
+        throw new RangeError("`Dataset.shuffle()` requires bufferSize to be specified.");
+      } else {
+        throw new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);
+      }
+    }
+    const base = this;
+    const random = seedrandom3.alea(seed || util_exports.now().toString());
+    return datasetFromIteratorFn(async () => {
+      let seed2 = random.int32();
+      if (reshuffleEachIteration) {
+        seed2 += random.int32();
+      }
+      return (await base.iterator()).shuffle(bufferSize, seed2.toString());
+    }, this.size);
+  }
+  take(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && this.size > count2) {
+      size = count2;
+    } else if (this.size != null && this.size <= count2) {
+      size = this.size;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).take(count2), size);
+  }
+  async toArray() {
+    if (this.size === Infinity) {
+      throw new Error("Can not convert infinite data stream to array.");
+    }
+    return (await this.iterator()).toArray();
+  }
+  async toArrayForTest() {
+    if (this.size === Infinity) {
+      throw new Error("Can not convert infinite data stream to array.");
+    }
+    return (await this.iterator()).toArrayForTest();
+  }
+};
+Dataset.MAX_BUFFER_SIZE = 1e4;
+function datasetFromIteratorFn(iteratorFn, size = null) {
+  return new class extends Dataset {
+    constructor() {
+      super(...arguments);
+      this.size = size;
+    }
+    async iterator() {
+      return iteratorFn();
+    }
+  }();
+}
+function array(items) {
+  return datasetFromIteratorFn(async () => iteratorFromItems(items), items.length);
+}
+function zip(datasets) {
+  if (!isIterable2(datasets)) {
+    throw new Error("The argument to zip() must be an object or array.");
+  }
+  let size;
+  if (Array.isArray(datasets)) {
+    for (let i = 0; i < datasets.length; i++) {
+      size = size == null ? datasets[i].size : Math.min(size, datasets[i].size);
+    }
+  } else if (datasets instanceof Object) {
+    for (const ds in datasets) {
+      size = size == null ? datasets[ds].size : Math.min(size, datasets[ds].size);
+    }
+  }
+  return datasetFromIteratorFn(async () => {
+    const streams = await deepMapAndAwaitAll(datasets, (d) => {
+      if (d instanceof Dataset) {
+        return { value: d.iterator(), recurse: false };
+      } else if (isIterable2(d)) {
+        return { value: null, recurse: true };
+      } else {
+        throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.");
+      }
+    });
+    return iteratorFromZipped(streams, ZipMismatchMode.SHORTEST);
+  }, size);
+}
+function deepBatchConcat(rows) {
+  if (rows === null) {
+    return null;
+  }
+  const exampleRow = rows[0];
+  if (canTensorify(exampleRow)) {
+    const value = batchConcat(rows);
+    return { value, recurse: false };
+  }
+  return { value: null, recurse: true };
+}
+function batchConcat(arrays) {
+  if (arrays.length === 0) {
+    throw new Error("Can't make a batch of zero elements.");
+  }
+  if (arrays[0] instanceof Tensor) {
+    return stack(arrays);
+  } else {
+    return tensor(arrays);
+  }
+}
+var TextLineDataset = class extends Dataset {
+  constructor(input2) {
+    super();
+    this.input = input2;
+  }
+  async iterator() {
+    const inputIterator = await this.input.iterator();
+    const utf8Iterator = inputIterator.decodeUTF8();
+    const lineIterator = utf8Iterator.split("\n").map((line) => {
+      if (line.endsWith("\r")) {
+        line = line.slice(0, -1);
+      }
+      return line;
+    });
+    return lineIterator;
+  }
+};
+var CODE_QUOTE = '"';
+var STATE_OUT = Symbol("out");
+var STATE_FIELD = Symbol("field");
+var STATE_QUOTE = Symbol("quote");
+var STATE_QUOTE_AFTER_QUOTE = Symbol("quoteafterquote");
+var STATE_WITHIN_QUOTE_IN_QUOTE = Symbol("quoteinquote");
+var CSVDataset = class extends Dataset {
+  constructor(input2, csvConfig) {
+    super();
+    this.input = input2;
+    this.hasHeader = true;
+    this.fullColumnNames = null;
+    this.columnNamesValidated = false;
+    this.columnConfigs = null;
+    this.configuredColumnsOnly = false;
+    this.delimiter = ",";
+    this.delimWhitespace = false;
+    this.base = new TextLineDataset(input2);
+    if (!csvConfig) {
+      csvConfig = {};
+    }
+    this.hasHeader = csvConfig.hasHeader === false ? false : true;
+    this.fullColumnNames = csvConfig.columnNames;
+    this.columnConfigs = csvConfig.columnConfigs;
+    this.configuredColumnsOnly = csvConfig.configuredColumnsOnly;
+    if (csvConfig.delimWhitespace) {
+      util_exports.assert(csvConfig.delimiter == null, () => "Delimiter should not be provided when delimWhitespace is true.");
+      this.delimWhitespace = true;
+      this.delimiter = " ";
+    } else {
+      this.delimiter = csvConfig.delimiter ? csvConfig.delimiter : ",";
+    }
+  }
+  async columnNames() {
+    if (!this.columnNamesValidated) {
+      await this.setColumnNames();
+    }
+    return this.configuredColumnsOnly ? Object.keys(this.columnConfigs) : this.fullColumnNames;
+  }
+  async setColumnNames() {
+    const columnNamesFromFile = await this.maybeReadHeaderLine();
+    if (!this.fullColumnNames && !columnNamesFromFile) {
+      throw new Error("Column names must be provided if there is no header line.");
+    } else if (this.fullColumnNames && columnNamesFromFile) {
+      util_exports.assert(columnNamesFromFile.length === this.fullColumnNames.length, () => "The length of provided columnNames (" + this.fullColumnNames.length.toString() + ") does not match the length of the header line read from file (" + columnNamesFromFile.length.toString() + ").");
+    }
+    if (!this.fullColumnNames) {
+      this.fullColumnNames = columnNamesFromFile;
+    }
+    const counts = this.fullColumnNames.reduce((countAcc, name) => {
+      countAcc[name] = countAcc[name] + 1 || 1;
+      return countAcc;
+    }, {});
+    const duplicateNames = Object.keys(counts).filter((name) => counts[name] > 1);
+    util_exports.assert(duplicateNames.length === 0, () => "Duplicate column names found: " + duplicateNames.toString());
+    if (this.columnConfigs) {
+      for (const key of Object.keys(this.columnConfigs)) {
+        const index = this.fullColumnNames.indexOf(key);
+        if (index === -1) {
+          throw new Error('The key "' + key + '" provided in columnConfigs does not match any of the column names (' + this.fullColumnNames.toString() + ").");
+        }
+      }
+    }
+    this.columnNamesValidated = true;
+  }
+  async maybeReadHeaderLine() {
+    if (this.hasHeader) {
+      const iter = await this.base.iterator();
+      const firstElement = await iter.next();
+      if (firstElement.done) {
+        throw new Error("No data was found for CSV parsing.");
+      }
+      const firstLine = firstElement.value;
+      const headers = this.parseRow(firstLine, false);
+      return headers;
+    } else {
+      return null;
+    }
+  }
+  async iterator() {
+    if (!this.columnNamesValidated) {
+      await this.setColumnNames();
+    }
+    let lines = await this.base.iterator();
+    if (this.hasHeader) {
+      lines = lines.skip(1);
+    }
+    return lines.map((x) => this.makeDataElement(x));
+  }
+  makeDataElement(line) {
+    const values = this.parseRow(line);
+    const features = {};
+    const labels = {};
+    for (let i = 0; i < this.fullColumnNames.length; i++) {
+      const key = this.fullColumnNames[i];
+      const config = this.columnConfigs ? this.columnConfigs[key] : null;
+      if (this.configuredColumnsOnly && !config) {
+        continue;
+      } else {
+        const value = values[i];
+        let parsedValue = null;
+        if (value === "") {
+          if (config && config.default !== void 0) {
+            parsedValue = config.default;
+          } else if (config && (config.required || config.isLabel)) {
+            throw new Error(`Required column ${key} is empty in this line: ${line}`);
+          } else {
+            parsedValue = void 0;
+          }
+        } else {
+          const valueAsNum = Number(value);
+          if (isNaN(valueAsNum)) {
+            if (config && config.dtype === "bool") {
+              parsedValue = this.getBoolean(value);
+            } else {
+              parsedValue = value;
+            }
+          } else if (!config || !config.dtype) {
+            parsedValue = valueAsNum;
+          } else {
+            switch (config.dtype) {
+              case "float32":
+                parsedValue = valueAsNum;
+                break;
+              case "int32":
+                parsedValue = Math.floor(valueAsNum);
+                break;
+              case "bool":
+                parsedValue = this.getBoolean(value);
+                break;
+              default:
+                parsedValue = valueAsNum;
+            }
+          }
+        }
+        config && config.isLabel ? labels[key] = parsedValue : features[key] = parsedValue;
+      }
+    }
+    if (Object.keys(labels).length === 0) {
+      return features;
+    } else {
+      return { xs: features, ys: labels };
+    }
+  }
+  getBoolean(value) {
+    if (value === "1" || value.toLowerCase() === "true") {
+      return 1;
+    } else {
+      return 0;
+    }
+  }
+  parseRow(line, validateElementCount = true) {
+    const result = [];
+    let readOffset = 0;
+    const readLength = line.length;
+    let currentState = STATE_OUT;
+    for (let i = 0; i < readLength; i++) {
+      switch (currentState) {
+        case STATE_OUT:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              readOffset = i + 1;
+              currentState = STATE_QUOTE;
+              break;
+            case this.delimiter:
+              readOffset = i + 1;
+              if (this.delimiter === " " && this.delimWhitespace) {
+                break;
+              }
+              result.push("");
+              currentState = STATE_OUT;
+              break;
+            default:
+              currentState = STATE_FIELD;
+              readOffset = i;
+              break;
+          }
+          break;
+        case STATE_FIELD:
+          switch (line.charAt(i)) {
+            case this.delimiter:
+              result.push(line.substring(readOffset, i));
+              currentState = STATE_OUT;
+              readOffset = i + 1;
+              break;
+            default:
+          }
+          break;
+        case STATE_QUOTE:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE_AFTER_QUOTE;
+              break;
+            default:
+          }
+          break;
+        case STATE_QUOTE_AFTER_QUOTE:
+          switch (line.charAt(i)) {
+            case this.delimiter:
+              result.push(line.substring(readOffset, i - 1));
+              currentState = STATE_OUT;
+              readOffset = i + 1;
+              break;
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE;
+              break;
+            default:
+              currentState = STATE_WITHIN_QUOTE_IN_QUOTE;
+              break;
+          }
+          break;
+        case STATE_WITHIN_QUOTE_IN_QUOTE:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE;
+              break;
+            default:
+          }
+          break;
+        default:
+      }
+    }
+    if (currentState === STATE_QUOTE_AFTER_QUOTE) {
+      result.push(line.substring(readOffset, readLength - 1));
+    } else {
+      result.push(line.substring(readOffset));
+    }
+    if (validateElementCount && result.length !== this.fullColumnNames.length) {
+      throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${result}`);
+    }
+    return result;
+  }
+};
+var MicrophoneIterator = class extends LazyIterator {
+  constructor(microphoneConfig) {
+    super();
+    this.microphoneConfig = microphoneConfig;
+    this.isClosed = false;
+    this.fftSize = microphoneConfig.fftSize || 1024;
+    const fftSizeLog2 = Math.log2(this.fftSize);
+    if (this.fftSize < 0 || fftSizeLog2 < 4 || fftSizeLog2 > 14 || !Number.isInteger(fftSizeLog2)) {
+      throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);
+    }
+    this.numFrames = microphoneConfig.numFramesPerSpectrogram || 43;
+    this.sampleRateHz = microphoneConfig.sampleRateHz;
+    this.columnTruncateLength = microphoneConfig.columnTruncateLength || this.fftSize;
+    this.audioTrackConstraints = microphoneConfig.audioTrackConstraints;
+    this.smoothingTimeConstant = microphoneConfig.smoothingTimeConstant || 0;
+    this.includeSpectrogram = microphoneConfig.includeSpectrogram === false ? false : true;
+    this.includeWaveform = microphoneConfig.includeWaveform === true ? true : false;
+    if (!this.includeSpectrogram && !this.includeWaveform) {
+      throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.");
+    }
+  }
+  summary() {
+    return `microphone`;
+  }
+  static async create(microphoneConfig = {}) {
+    if (!env().get("IS_BROWSER")) {
+      throw new Error("microphone API is only supported in browser environment.");
+    }
+    const microphoneIterator = new MicrophoneIterator(microphoneConfig);
+    await microphoneIterator.start();
+    return microphoneIterator;
+  }
+  async start() {
+    try {
+      this.stream = await navigator.mediaDevices.getUserMedia({
+        audio: this.audioTrackConstraints == null ? true : this.audioTrackConstraints,
+        video: false
+      });
+    } catch (e) {
+      throw new Error(`Error thrown while initializing video stream: ${e.message}`);
+    }
+    if (!this.stream) {
+      throw new Error("Could not obtain audio from microphone.");
+    }
+    const ctxConstructor = window.AudioContext || window.webkitAudioContext;
+    this.audioContext = new ctxConstructor();
+    if (!this.sampleRateHz) {
+      this.sampleRateHz = this.audioContext.sampleRate;
+    } else if (this.audioContext.sampleRate !== this.sampleRateHz) {
+      throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);
+    }
+    const streamSource = this.audioContext.createMediaStreamSource(this.stream);
+    this.analyser = this.audioContext.createAnalyser();
+    this.analyser.fftSize = this.fftSize * 2;
+    this.analyser.smoothingTimeConstant = this.smoothingTimeConstant;
+    streamSource.connect(this.analyser);
+    this.freqData = new Float32Array(this.fftSize);
+    this.timeData = new Float32Array(this.fftSize);
+    return;
+  }
+  async next() {
+    if (this.isClosed) {
+      return { value: null, done: true };
+    }
+    let spectrogramTensor;
+    let waveformTensor;
+    const audioDataQueue = await this.getAudioData();
+    if (this.includeSpectrogram) {
+      const freqData = this.flattenQueue(audioDataQueue.freqDataQueue);
+      spectrogramTensor = this.getTensorFromAudioDataArray(freqData, [this.numFrames, this.columnTruncateLength, 1]);
+    }
+    if (this.includeWaveform) {
+      const timeData = this.flattenQueue(audioDataQueue.timeDataQueue);
+      waveformTensor = this.getTensorFromAudioDataArray(timeData, [this.numFrames * this.fftSize, 1]);
+    }
+    return {
+      value: { "spectrogram": spectrogramTensor, "waveform": waveformTensor },
+      done: false
+    };
+  }
+  async capture() {
+    return (await this.next()).value;
+  }
+  async getAudioData() {
+    const freqDataQueue = [];
+    const timeDataQueue = [];
+    let currentFrames = 0;
+    return new Promise((resolve) => {
+      const intervalID = setInterval(() => {
+        if (this.includeSpectrogram) {
+          this.analyser.getFloatFrequencyData(this.freqData);
+          if (this.freqData[0] === -Infinity) {
+            resolve({ freqDataQueue, timeDataQueue });
+          }
+          freqDataQueue.push(this.freqData.slice(0, this.columnTruncateLength));
+        }
+        if (this.includeWaveform) {
+          this.analyser.getFloatTimeDomainData(this.timeData);
+          timeDataQueue.push(this.timeData.slice());
+        }
+        if (++currentFrames === this.numFrames) {
+          clearInterval(intervalID);
+          resolve({ freqDataQueue, timeDataQueue });
+        }
+      }, this.fftSize / this.sampleRateHz * 1e3);
+    });
+  }
+  stop() {
+    if (!this.isClosed) {
+      this.isClosed = true;
+      this.analyser.disconnect();
+      this.audioContext.close();
+      if (this.stream != null && this.stream.getTracks().length > 0) {
+        this.stream.getTracks()[0].stop();
+      }
+    }
+  }
+  toArray() {
+    throw new Error("Can not convert infinite audio stream to array.");
+  }
+  getSampleRate() {
+    return this.sampleRateHz;
+  }
+  flattenQueue(queue) {
+    const frameSize = queue[0].length;
+    const freqData = new Float32Array(queue.length * frameSize);
+    queue.forEach((data, i) => freqData.set(data, i * frameSize));
+    return freqData;
+  }
+  getTensorFromAudioDataArray(freqData, shape) {
+    const vals = new Float32Array(util_exports.sizeFromShape(shape));
+    vals.set(freqData, vals.length - freqData.length);
+    return tensor(vals, shape);
+  }
+};
+var WebcamIterator = class extends LazyIterator {
+  constructor(webcamVideoElement, webcamConfig) {
+    super();
+    this.webcamVideoElement = webcamVideoElement;
+    this.webcamConfig = webcamConfig;
+    this.isClosed = true;
+    this.resize = false;
+    if (this.needToResize()) {
+      this.resize = true;
+      this.cropSize = [this.webcamConfig.resizeHeight, this.webcamConfig.resizeWidth];
+      this.cropBoxInd = tensor1d([0], "int32");
+      if (this.webcamConfig.centerCrop) {
+        const widthCroppingRatio = this.webcamConfig.resizeWidth * 1 / this.webcamVideoElement.width;
+        const heightCroppingRatio = this.webcamConfig.resizeHeight * 1 / this.webcamVideoElement.height;
+        const widthCropStart = (1 - widthCroppingRatio) / 2;
+        const heightCropStart = (1 - heightCroppingRatio) / 2;
+        const widthCropEnd = widthCropStart + widthCroppingRatio;
+        const heightCropEnd = heightCroppingRatio + heightCropStart;
+        this.cropBox = tensor2d([heightCropStart, widthCropStart, heightCropEnd, widthCropEnd], [1, 4]);
+      } else {
+        this.cropBox = tensor2d([0, 0, 1, 1], [1, 4]);
+      }
+    }
+  }
+  summary() {
+    return `webcam`;
+  }
+  static async create(webcamVideoElement, webcamConfig = {}) {
+    if (!env().get("IS_BROWSER")) {
+      throw new Error("tf.data.webcam is only supported in browser environment.");
+    }
+    if (!webcamVideoElement) {
+      webcamVideoElement = document.createElement("video");
+      if (!webcamConfig.resizeWidth || !webcamConfig.resizeHeight) {
+        throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");
+      }
+      webcamVideoElement.width = webcamConfig.resizeWidth;
+      webcamVideoElement.height = webcamConfig.resizeHeight;
+    }
+    const webcamIterator = new WebcamIterator(webcamVideoElement, webcamConfig);
+    await webcamIterator.start();
+    return webcamIterator;
+  }
+  async start() {
+    if (this.webcamConfig.facingMode) {
+      util_exports.assert(this.webcamConfig.facingMode === "user" || this.webcamConfig.facingMode === "environment", () => `Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);
+    }
+    try {
+      this.stream = await navigator.mediaDevices.getUserMedia({
+        video: {
+          deviceId: this.webcamConfig.deviceId,
+          facingMode: this.webcamConfig.facingMode ? this.webcamConfig.facingMode : "user",
+          width: this.webcamVideoElement.width,
+          height: this.webcamVideoElement.height
+        }
+      });
+    } catch (e) {
+      e.message = `Error thrown while initializing video stream: ${e.message}`;
+      throw e;
+    }
+    if (!this.stream) {
+      throw new Error("Could not obtain video from webcam.");
+    }
+    try {
+      this.webcamVideoElement.srcObject = this.stream;
+    } catch (error) {
+      console.log(error);
+      this.webcamVideoElement.src = window.URL.createObjectURL(this.stream);
+    }
+    this.webcamVideoElement.play();
+    this.isClosed = false;
+    return new Promise((resolve) => {
+      this.webcamVideoElement.onloadedmetadata = () => {
+        resolve();
+      };
+    });
+  }
+  async next() {
+    if (this.isClosed) {
+      return { value: null, done: true };
+    }
+    let img;
+    try {
+      img = browser_exports.fromPixels(this.webcamVideoElement);
+    } catch (e) {
+      throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`);
+    }
+    if (this.resize) {
+      try {
+        return { value: this.cropAndResizeFrame(img), done: false };
+      } catch (e) {
+        throw new Error(`Error thrown cropping the video: ${e.message}`);
+      } finally {
+        img.dispose();
+      }
+    } else {
+      return { value: img, done: false };
+    }
+  }
+  needToResize() {
+    if (this.webcamConfig.resizeWidth && this.webcamConfig.resizeHeight && (this.webcamVideoElement.width !== this.webcamConfig.resizeWidth || this.webcamVideoElement.height !== this.webcamConfig.resizeHeight)) {
+      return true;
+    }
+    return false;
+  }
+  cropAndResizeFrame(img) {
+    return tidy(() => {
+      const expandedImage = expandDims(cast(img, "float32"), 0);
+      let resizedImage;
+      resizedImage = image.cropAndResize(expandedImage, this.cropBox, this.cropBoxInd, this.cropSize, "bilinear");
+      const shape = resizedImage.shape;
+      return reshape(resizedImage, shape.slice(1));
+    });
+  }
+  async capture() {
+    return (await this.next()).value;
+  }
+  stop() {
+    const tracks = this.stream.getTracks();
+    tracks.forEach((track) => track.stop());
+    try {
+      this.webcamVideoElement.srcObject = null;
+    } catch (error) {
+      console.log(error);
+      this.webcamVideoElement.src = null;
+    }
+    this.isClosed = true;
+  }
+  toArray() {
+    throw new Error("Can not convert infinite video stream to array.");
+  }
+};
+var DataSource = class {
+};
+var StringIterator = class extends LazyIterator {
+  split(separator) {
+    return new SplitIterator(this, separator);
+  }
+};
+var SplitIterator = class extends StringIterator {
+  constructor(upstream, separator) {
+    super();
+    this.upstream = upstream;
+    this.impl = new SplitIteratorImpl(upstream, separator);
+  }
+  summary() {
+    return this.impl.summary();
+  }
+  async next() {
+    return this.impl.next();
+  }
+};
+var SplitIteratorImpl = class extends OneToManyIterator {
+  constructor(upstream, separator) {
+    super();
+    this.upstream = upstream;
+    this.separator = separator;
+    this.carryover = "";
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Split('${this.separator}')`;
+  }
+  async pump() {
+    const chunkResult = await this.upstream.next();
+    if (chunkResult.done) {
+      if (this.carryover === "") {
+        return false;
+      }
+      this.outputQueue.push(this.carryover);
+      this.carryover = "";
+      return true;
+    }
+    const lines = chunkResult.value.split(this.separator);
+    lines[0] = this.carryover + lines[0];
+    for (const line of lines.slice(0, -1)) {
+      this.outputQueue.push(line);
+    }
+    this.carryover = lines[lines.length - 1];
+    return true;
+  }
+};
+var ByteChunkIterator = class extends LazyIterator {
+  decodeUTF8() {
+    return new Utf8Iterator(this);
+  }
+};
+var Utf8Iterator = class extends StringIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    this.impl = new Utf8IteratorImpl(upstream);
+  }
+  summary() {
+    return this.impl.summary();
+  }
+  async next() {
+    return this.impl.next();
+  }
+};
+var Utf8IteratorImpl = class extends OneToManyIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    if (env().get("IS_BROWSER")) {
+      this.decoder = new TextDecoder("utf-8");
+    } else {
+      const { StringDecoder } = require_string_decoder();
+      this.decoder = new StringDecoder("utf8");
+    }
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Utf8`;
+  }
+  async pump() {
+    const chunkResult = await this.upstream.next();
+    let chunk;
+    if (chunkResult.done) {
+      return false;
+    } else {
+      chunk = chunkResult.value;
+    }
+    let text;
+    if (env().get("IS_BROWSER")) {
+      text = this.decoder.decode(chunk, { stream: true });
+    } else {
+      text = this.decoder.write(Buffer.from(chunk.buffer));
+    }
+    this.outputQueue.push(text);
+    return true;
+  }
+};
+var FileChunkIterator = class extends ByteChunkIterator {
+  constructor(file, options = {}) {
+    super();
+    this.file = file;
+    this.options = options;
+    util_exports.assert(file instanceof Uint8Array || (env().get("IS_BROWSER") ? file instanceof File || file instanceof Blob : false), () => "FileChunkIterator only supports File, Blob and Uint8Array right now.");
+    this.offset = options.offset || 0;
+    this.chunkSize = options.chunkSize || 1024 * 1024;
+  }
+  summary() {
+    return `FileChunks ${this.file}`;
+  }
+  async next() {
+    if (this.offset >= (this.file instanceof Uint8Array ? this.file.byteLength : this.file.size)) {
+      return { value: null, done: true };
+    }
+    const chunk = new Promise((resolve, reject) => {
+      const end = this.offset + this.chunkSize;
+      if (this.file instanceof Uint8Array) {
+        resolve(new Uint8Array(this.file.slice(this.offset, end)));
+      } else {
+        const fileReader = new FileReader();
+        fileReader.onload = (event) => {
+          let data = fileReader.result;
+          if (data instanceof ArrayBuffer) {
+            data = new Uint8Array(data);
+          }
+          if (!(data instanceof Uint8Array)) {
+            return reject(new TypeError("FileReader returned unknown type."));
+          }
+          resolve(data);
+        };
+        fileReader.onabort = (event) => {
+          return reject(new Error("Aborted"));
+        };
+        fileReader.onerror = (event) => {
+          return reject(new Error(event.type));
+        };
+        const slice5 = this.file.slice(this.offset, end);
+        fileReader.readAsArrayBuffer(slice5);
+      }
+      this.offset = end;
+    });
+    return { value: await chunk, done: false };
+  }
+};
+async function urlChunkIterator(url, options = {}, fetchFunc) {
+  let urlString;
+  let requestInit;
+  if (typeof url === "string") {
+    urlString = url;
+  } else {
+    urlString = url.url;
+    requestInit = getRequestInitFromRequest(url);
+  }
+  const response = await (fetchFunc || util_exports.fetch)(urlString, requestInit);
+  if (response.ok) {
+    const uint8Array = new Uint8Array(await response.arrayBuffer());
+    return new FileChunkIterator(uint8Array, options);
+  } else {
+    throw new Error(response.statusText);
+  }
+}
+var getRequestInitFromRequest = (request) => {
+  const init2 = {
+    method: request.method,
+    headers: request.headers,
+    body: request.body,
+    mode: request.mode,
+    credentials: request.credentials,
+    cache: request.cache,
+    redirect: request.redirect,
+    referrer: request.referrer,
+    integrity: request.integrity
+  };
+  return init2;
+};
+function isLocalPath(source) {
+  return typeof source === "string" && source.slice(0, 7) === "file://";
+}
+var FileDataSource = class extends DataSource {
+  constructor(input2, options = {}) {
+    super();
+    this.input = input2;
+    this.options = options;
+  }
+  async iterator() {
+    if (isLocalPath(this.input) && env().get("IS_NODE")) {
+      const fs = require_fs();
+      this.input = fs.readFileSync(this.input.slice(7));
+    }
+    return new FileChunkIterator(this.input, this.options);
+  }
+};
+var URLDataSource = class extends DataSource {
+  constructor(url, fileOptions = {}) {
+    super();
+    this.url = url;
+    this.fileOptions = fileOptions;
+  }
+  async iterator() {
+    if (isLocalPath(this.url)) {
+      return new FileDataSource(this.url, this.fileOptions).iterator();
+    } else {
+      return urlChunkIterator(this.url, this.fileOptions);
+    }
+  }
+};
+function csv(source, csvConfig = {}) {
+  return new CSVDataset(new URLDataSource(source), csvConfig);
+}
+function func(f) {
+  const iter = iteratorFromFunction(f);
+  return datasetFromIteratorFn(async () => iter);
+}
+function generator(generator2) {
+  return datasetFromIteratorFn(async () => {
+    const gen = await generator2();
+    return iteratorFromFunction(() => gen.next());
+  });
+}
+async function webcam(webcamVideoElement, webcamConfig) {
+  return WebcamIterator.create(webcamVideoElement, webcamConfig);
+}
+async function microphone(microphoneConfig) {
+  return MicrophoneIterator.create(microphoneConfig);
+}
+var version4 = "4.0.0";
+function assertNotComplex(tensor2, opName) {
+  if (!Array.isArray(tensor2)) {
+    tensor2 = [tensor2];
+  }
+  tensor2.forEach((t) => {
+    if (t != null) {
+      util_exports.assert(t.dtype !== "complex64", () => `${opName} does not support complex64 tensors in the CPU backend.`);
+    }
+  });
+}
+var whereImpl2 = kernel_impls_exports.whereImpl;
+var MathBackendCPU = class extends KernelBackend {
+  constructor() {
+    super();
+    this.blockSize = 48;
+    this.firstUse = true;
+    this.data = new DataStorage(this, engine());
+  }
+  nextDataId() {
+    return MathBackendCPU.nextDataId++;
+  }
+  write(values, shape, dtype) {
+    if (this.firstUse) {
+      this.firstUse = false;
+      if (env().get("IS_NODE")) {
+        backend_util_exports.warn("\n============================\nHi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. \n============================");
+      }
+    }
+    const dataId = { id: this.nextDataId() };
+    this.data.set(dataId, { values, dtype, refCount: 1 });
+    return dataId;
+  }
+  makeTensorInfo(shape, dtype, values) {
+    let outId;
+    if (dtype === "string" && values != null && values.length > 0 && util_exports.isString(values[0])) {
+      const encodedValues = values.map((d) => util_exports.encodeString(d));
+      outId = this.write(encodedValues, shape, dtype);
+    } else {
+      outId = this.write(values, shape, dtype);
+    }
+    return { dataId: outId, shape, dtype };
+  }
+  refCount(dataId) {
+    if (this.data.has(dataId)) {
+      const tensorData = this.data.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const tensorData = this.data.get(dataId);
+    tensorData.refCount++;
+  }
+  decRef(dataId) {
+    if (this.data.has(dataId)) {
+      const tensorData = this.data.get(dataId);
+      tensorData.refCount--;
+    }
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    this.data.set(dataId, { values, dtype, refCount });
+  }
+  numDataIds() {
+    return this.data.numDataIds();
+  }
+  async read(dataId) {
+    return this.readSync(dataId);
+  }
+  readSync(dataId) {
+    const { dtype, complexTensorInfos } = this.data.get(dataId);
+    if (dtype === "complex64") {
+      const realValues = this.readSync(complexTensorInfos.real.dataId);
+      const imagValues = this.readSync(complexTensorInfos.imag.dataId);
+      return backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    }
+    return this.data.get(dataId).values;
+  }
+  bufferSync(t) {
+    const data = this.readSync(t.dataId);
+    if (t.dtype === "string") {
+      try {
+        const strings = data.map((d) => util_exports.decodeString(d));
+        return buffer(t.shape, t.dtype, strings);
+      } catch (_a) {
+        throw new Error("Failed to decode encoded string bytes into utf-8");
+      }
+    }
+    return buffer(t.shape, t.dtype, data);
+  }
+  makeOutput(values, shape, dtype) {
+    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);
+  }
+  disposeData(dataId, force = false) {
+    if (this.data.has(dataId)) {
+      this.data.get(dataId).refCount--;
+      if (!force && this.data.get(dataId).refCount > 0) {
+        return false;
+      }
+      const { complexTensorInfos } = this.data.get(dataId);
+      if (complexTensorInfos != null) {
+        this.disposeData(complexTensorInfos.real.dataId, true);
+        this.disposeData(complexTensorInfos.imag.dataId, true);
+      }
+      this.data.delete(dataId);
+    }
+    return true;
+  }
+  disposeIntermediateTensorInfo(tensorInfo) {
+    this.disposeData(tensorInfo.dataId);
+  }
+  async time(f) {
+    const start = util_exports.now();
+    f();
+    const kernelMs = util_exports.now() - start;
+    return { kernelMs };
+  }
+  memory() {
+    return {
+      unreliable: true,
+      reasons: ["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]
+    };
+  }
+  where(condition) {
+    assertNotComplex([condition], "where");
+    const condVals = this.readSync(condition.dataId);
+    return whereImpl2(condition.shape, condVals);
+  }
+  dispose() {
+  }
+  floatPrecision() {
+    return 32;
+  }
+  epsilon() {
+    return super.epsilon();
+  }
+};
+MathBackendCPU.nextDataId = 0;
+var shared_exports = {};
+__export2(shared_exports, {
+  addImpl: () => addImpl,
+  bincountImpl: () => bincountImpl,
+  bincountReduceImpl: () => bincountReduceImpl,
+  castImpl: () => castImpl,
+  ceilImpl: () => ceilImpl,
+  concatImpl: () => concatImpl,
+  equalImpl: () => equalImpl,
+  expImpl: () => expImpl,
+  expm1Impl: () => expm1Impl,
+  floorImpl: () => floorImpl,
+  gatherNdImpl: () => gatherNdImpl,
+  gatherV2Impl: () => gatherV2Impl,
+  greaterEqualImpl: () => greaterEqualImpl,
+  greaterImpl: () => greaterImpl,
+  lessEqualImpl: () => lessEqualImpl,
+  lessImpl: () => lessImpl,
+  linSpaceImpl: () => linSpaceImpl,
+  logImpl: () => logImpl,
+  maxImpl: () => maxImpl,
+  maximumImpl: () => maximumImpl,
+  minimumImpl: () => minimumImpl,
+  multiplyImpl: () => multiplyImpl,
+  negImpl: () => negImpl,
+  notEqualImpl: () => notEqualImpl,
+  prodImpl: () => prodImpl,
+  raggedGatherImpl: () => raggedGatherImpl,
+  raggedRangeImpl: () => raggedRangeImpl,
+  raggedTensorToTensorImpl: () => raggedTensorToTensorImpl,
+  rangeImpl: () => rangeImpl,
+  rsqrtImpl: () => rsqrtImpl,
+  scatterImpl: () => scatterImpl,
+  sigmoidImpl: () => sigmoidImpl,
+  simpleAbsImpl: () => simpleAbsImpl,
+  sliceImpl: () => sliceImpl,
+  sparseFillEmptyRowsImpl: () => sparseFillEmptyRowsImpl,
+  sparseReshapeImpl: () => sparseReshapeImpl,
+  sparseSegmentReductionImpl: () => sparseSegmentReductionImpl,
+  sqrtImpl: () => sqrtImpl,
+  squaredDifferenceImpl: () => squaredDifferenceImpl,
+  stridedSliceImpl: () => stridedSliceImpl,
+  stringNGramsImpl: () => stringNGramsImpl,
+  stringSplitImpl: () => stringSplitImpl,
+  stringToHashBucketFastImpl: () => stringToHashBucketFastImpl,
+  subImpl: () => subImpl,
+  tileImpl: () => tileImpl,
+  topKImpl: () => topKImpl,
+  transposeImpl: () => transposeImpl,
+  uniqueImpl: () => uniqueImpl
+});
+function simpleAbsImpl(vals) {
+  const resultValues = new Float32Array(vals.length);
+  for (let i = 0; i < vals.length; ++i) {
+    resultValues[i] = Math.abs(vals[i]);
+  }
+  return resultValues;
+}
+var abs2 = (args) => {
+  const { x } = args.inputs;
+  const cpuBackend = args.backend;
+  assertNotComplex(x, "abs");
+  let resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));
+  const values = cpuBackend.data.get(x.dataId).values;
+  resultValues = simpleAbsImpl(values);
+  return cpuBackend.makeOutput(resultValues, x.shape, x.dtype);
+};
+var absConfig = {
+  kernelName: Abs,
+  backendName: "cpu",
+  kernelFunc: abs2
+};
+function createSimpleBinaryKernelImpl(op2) {
+  return (aShape, bShape, aVals, bVals, dtype) => {
+    const newShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const resultRank = newShape.length;
+    const resultStrides = util_exports.computeStrides(newShape);
+    const resultSize = util_exports.sizeFromShape(newShape);
+    const result = util_exports.getTypedArrayFromDType(dtype, resultSize);
+    const aRank = aShape.length;
+    const bRank = bShape.length;
+    const aStrides = util_exports.computeStrides(aShape);
+    const bStrides = util_exports.computeStrides(bShape);
+    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, newShape);
+    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, newShape);
+    if (aBroadcastDims.length + bBroadcastDims.length === 0) {
+      for (let i = 0; i < result.length; ++i) {
+        result[i] = op2(aVals[i % aVals.length], bVals[i % bVals.length]);
+      }
+    } else {
+      for (let i = 0; i < result.length; ++i) {
+        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);
+        const aLoc = loc.slice(-aRank);
+        aBroadcastDims.forEach((d) => aLoc[d] = 0);
+        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);
+        const bLoc = loc.slice(-bRank);
+        bBroadcastDims.forEach((d) => bLoc[d] = 0);
+        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);
+        result[i] = op2(aVals[aIndex], bVals[bIndex]);
+      }
+    }
+    return [result, newShape];
+  };
+}
+function complex2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { real: real4, imag: imag4 } = inputs;
+  const realVals = backend2.data.get(real4.dataId).values;
+  const imagVals = backend2.data.get(imag4.dataId).values;
+  const complexInfo = backend2.makeTensorInfo(real4.shape, "complex64");
+  const complex4 = backend2.data.get(complexInfo.dataId);
+  complex4.complexTensorInfos = {
+    real: backend2.makeTensorInfo(real4.shape, "float32", realVals),
+    imag: backend2.makeTensorInfo(imag4.shape, "float32", imagVals)
+  };
+  return complexInfo;
+}
+var complexConfig = {
+  kernelName: Complex,
+  backendName: "cpu",
+  kernelFunc: complex2
+};
+function zeros3(backend2, shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = zeros3(backend2, shape, "float32");
+    const imag4 = zeros3(backend2, shape, "float32");
+    return complex2({ inputs: { real: real4, imag: imag4 }, backend: backend2 });
+  }
+  const values = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(shape), dtype);
+  return backend2.makeTensorInfo(shape, dtype, values);
+}
+function identity2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  backend2.incRef(x.dataId);
+  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };
+}
+var identityConfig = {
+  kernelName: Identity,
+  backendName: "cpu",
+  kernelFunc: identity2
+};
+function real2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const real4 = backend2.data.get(input2.dataId).complexTensorInfos.real;
+  const realVal = backend2.data.get(real4.dataId).values;
+  return backend2.makeTensorInfo(real4.shape, real4.dtype, realVal);
+}
+var realConfig = {
+  kernelName: Real,
+  backendName: "cpu",
+  kernelFunc: real2
+};
+function castImpl(values, shape, inputType, dtype) {
+  if (dtype === "int32") {
+    const resultValues = Int32Array.from(values);
+    return [shape, "int32", resultValues];
+  }
+  if (dtype === "bool") {
+    const zero = util_exports.toTypedArray([0], inputType);
+    const [resultData, resultShape] = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0)(shape, [], values, zero, "bool");
+    return [resultShape, "bool", resultData];
+  }
+  throw new Error(`Error in Cast: failed to cast ${inputType} to ${dtype}`);
+}
+function cast3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dtype } = attrs;
+  if (dtype === "complex64") {
+    if (x.dtype === "complex64") {
+      return identity2({ inputs: { x }, backend: backend2 });
+    }
+    const zerosTensorInfo = zeros3(backend2, x.shape, x.dtype);
+    const floatX = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+    const result = complex2({ inputs: { real: floatX, imag: zerosTensorInfo }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);
+    backend2.disposeIntermediateTensorInfo(floatX);
+    return result;
+  }
+  if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const result = cast3({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    return result;
+  }
+  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {
+    const result = identity2({ inputs: { x }, backend: backend2 });
+    return { dataId: result.dataId, shape: result.shape, dtype };
+  }
+  const values = backend2.data.get(x.dataId).values;
+  const [resultShape, resultType, resultData] = castImpl(values, x.shape, x.dtype, dtype);
+  return backend2.makeTensorInfo(resultShape, resultType, resultData);
+}
+var castConfig = {
+  kernelName: Cast,
+  backendName: "cpu",
+  kernelFunc: cast3
+};
+function binaryKernelFunc(name, simpleImpl, complexImpl, dtype) {
+  if (complexImpl == null) {
+    return ({ inputs, backend: backend2 }) => {
+      const { a, b } = inputs;
+      const cpuBackend = backend2;
+      assertNotComplex([a, b], name);
+      const aVals = cpuBackend.data.get(a.dataId).values;
+      const bVals = cpuBackend.data.get(b.dataId).values;
+      const decodedAVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;
+      const decodedBVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;
+      const $dtype = dtype || a.dtype;
+      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);
+      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);
+    };
+  }
+  return ({ inputs, backend: backend2 }) => {
+    const { a, b } = inputs;
+    const cpuBackend = backend2;
+    if (a.dtype === "complex64" || b.dtype === "complex64") {
+      const $aComplex = cast3({ inputs: { x: a }, backend: cpuBackend, attrs: { dtype: "complex64" } });
+      const $aComplexVals = cpuBackend.data.get($aComplex.dataId);
+      const aReal = $aComplexVals.complexTensorInfos.real;
+      const aImag = $aComplexVals.complexTensorInfos.imag;
+      const aRealVals = cpuBackend.data.get(aReal.dataId).values;
+      const aImagVals = cpuBackend.data.get(aImag.dataId).values;
+      const $bComplex = cast3({ inputs: { x: b }, backend: cpuBackend, attrs: { dtype: "complex64" } });
+      const $bComplexVals = cpuBackend.data.get($bComplex.dataId);
+      const bReal = $bComplexVals.complexTensorInfos.real;
+      const bImag = $bComplexVals.complexTensorInfos.imag;
+      const bRealVals = cpuBackend.data.get(bReal.dataId).values;
+      const bImagVals = cpuBackend.data.get(bImag.dataId).values;
+      const [resultRealData, resultImagData, resultShape] = complexImpl(a.shape, b.shape, aRealVals, aImagVals, bRealVals, bImagVals);
+      const resultReal = cpuBackend.makeTensorInfo(resultShape, "float32", resultRealData);
+      const resultImag = cpuBackend.makeTensorInfo(resultShape, "float32", resultImagData);
+      const result = complex2({ inputs: { real: resultReal, imag: resultImag }, backend: cpuBackend });
+      cpuBackend.disposeIntermediateTensorInfo($aComplex);
+      cpuBackend.disposeIntermediateTensorInfo($bComplex);
+      cpuBackend.disposeIntermediateTensorInfo(resultReal);
+      cpuBackend.disposeIntermediateTensorInfo(resultImag);
+      return result;
+    } else {
+      const aVals = cpuBackend.data.get(a.dataId).values;
+      const bVals = cpuBackend.data.get(b.dataId).values;
+      const $dtype = dtype || a.dtype;
+      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, aVals, bVals, $dtype);
+      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);
+    }
+  };
+}
+function createComplexBinaryKernelImpl(op2) {
+  return (aShape, bShape, aRealVals, aImagVals, bRealVals, bImagVals) => {
+    const resultShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const resultSize = util_exports.sizeFromShape(resultShape);
+    const resultRank = resultShape.length;
+    const resultStrides = util_exports.computeStrides(resultShape);
+    const resultRealVals = util_exports.getTypedArrayFromDType("float32", resultSize);
+    const resultImagVals = util_exports.getTypedArrayFromDType("float32", resultSize);
+    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, resultShape);
+    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, resultShape);
+    const aVals = backend_util_exports.mergeRealAndImagArrays(aRealVals, aImagVals);
+    const bVals = backend_util_exports.mergeRealAndImagArrays(bRealVals, bImagVals);
+    const aRank = aShape.length;
+    const aStrides = util_exports.computeStrides(aShape);
+    const bRank = bShape.length;
+    const bStrides = util_exports.computeStrides(bShape);
+    if (aBroadcastDims.length + bBroadcastDims.length === 0) {
+      for (let i = 0; i < resultRealVals.length; i++) {
+        const aIdx = i % aVals.length;
+        const bIdx = i % bVals.length;
+        const result = op2(aVals[aIdx * 2], aVals[aIdx * 2 + 1], bVals[bIdx * 2], bVals[bIdx * 2 + 1]);
+        resultRealVals[i] = result.real;
+        resultImagVals[i] = result.imag;
+      }
+    } else {
+      for (let i = 0; i < resultRealVals.length; i++) {
+        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);
+        const aLoc = loc.slice(-aRank);
+        aBroadcastDims.forEach((d) => aLoc[d] = 0);
+        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);
+        const bLoc = loc.slice(-bRank);
+        bBroadcastDims.forEach((d) => bLoc[d] = 0);
+        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);
+        const opResult = op2(aVals[aIndex * 2], aVals[aIndex * 2 + 1], bVals[bIndex * 2], bVals[bIndex * 2 + 1]);
+        resultRealVals[i] = opResult.real;
+        resultImagVals[i] = opResult.imag;
+      }
+    }
+    return [resultRealVals, resultImagVals, resultShape];
+  };
+}
+var addImpl = createSimpleBinaryKernelImpl((a, b) => a + b);
+var addComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return { real: aReal + bReal, imag: aImag + bImag };
+});
+var add4 = binaryKernelFunc(Add, addImpl, addComplexImpl);
+var addConfig = {
+  kernelName: Add,
+  backendName: "cpu",
+  kernelFunc: add4
+};
+function bincountImpl(xVals, weightsVals, weightsDtype, weightsShape, size) {
+  const weightsSize = util_exports.sizeFromShape(weightsShape);
+  const outVals = util_exports.makeZerosTypedArray(size, weightsDtype);
+  for (let i = 0; i < xVals.length; i++) {
+    const value = xVals[i];
+    if (value < 0) {
+      throw new Error("Input x must be non-negative!");
+    }
+    if (value >= size) {
+      continue;
+    }
+    if (weightsSize > 0) {
+      outVals[value] += weightsVals[i];
+    } else {
+      outVals[value] += 1;
+    }
+  }
+  return outVals;
+}
+function bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput = false) {
+  const numRows = xBuf.shape[0];
+  const numCols = xBuf.shape[1];
+  const outBuf = buffer([numRows, size], weightsBuf.dtype);
+  for (let i = 0; i < numRows; i++) {
+    for (let j = 0; j < numCols; j++) {
+      const value = xBuf.get(i, j);
+      if (value < 0) {
+        throw new Error("Input x must be non-negative!");
+      }
+      if (value >= size) {
+        continue;
+      }
+      if (binaryOutput) {
+        outBuf.set(1, i, value);
+      } else {
+        if (weightsBuf.size > 0) {
+          outBuf.set(outBuf.get(i, value) + weightsBuf.get(i, j), i, value);
+        } else {
+          outBuf.set(outBuf.get(i, value) + 1, i, value);
+        }
+      }
+    }
+  }
+  return outBuf;
+}
+function createSimpleUnaryImpl(op2) {
+  return (values, dtype, attrs) => {
+    const newValues = util_exports.getTypedArrayFromDType(dtype, values.length);
+    for (let i = 0; i < values.length; ++i) {
+      newValues[i] = op2(values[i], attrs);
+    }
+    return newValues;
+  };
+}
+function unaryKernelFunc(name, op2, dtype) {
+  return ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    assertNotComplex(x, name);
+    if (x.dtype === "string" || dtype === "string") {
+      throw new Error("unaryKernelFunc does not support string input/output");
+    }
+    const cpuBackend = backend2;
+    const values = cpuBackend.data.get(x.dataId).values;
+    const xSize = util_exports.sizeFromShape(x.shape);
+    const $dtype = dtype || x.dtype;
+    const newValues = util_exports.getArrayFromDType($dtype, xSize);
+    for (let i = 0; i < xSize; ++i) {
+      newValues[i] = op2(values[i], attrs);
+    }
+    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);
+  };
+}
+function unaryKernelFuncFromImpl(name, unaryImpl, dtype) {
+  return ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    assertNotComplex(x, name);
+    if (x.dtype === "string" || dtype === "string") {
+      throw new Error("unaryKernelFunc does not support string input/output");
+    }
+    const cpuBackend = backend2;
+    const values = cpuBackend.data.get(x.dataId).values;
+    const $dtype = dtype || x.dtype;
+    const newValues = unaryImpl(values, $dtype, attrs);
+    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);
+  };
+}
+var ceilImpl = createSimpleUnaryImpl((xi) => Math.ceil(xi));
+var ceil2 = unaryKernelFuncFromImpl(Ceil, ceilImpl);
+var ceilConfig = {
+  kernelName: Ceil,
+  backendName: "cpu",
+  kernelFunc: ceil2
+};
+function concatImpl(inputs, outShape, dtype, simplyConcat) {
+  const outVals = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(outShape));
+  if (simplyConcat && dtype !== "string") {
+    let offset = 0;
+    inputs.forEach((input2) => {
+      const size = util_exports.sizeFromShape(input2.shape);
+      outVals.set(input2.vals, offset);
+      offset += size;
+    });
+  } else {
+    let colOffset = 0;
+    inputs.forEach((input2) => {
+      const decodedData = dtype === "string" ? backend_util_exports.fromUint8ToStringArray(input2.vals) : input2.vals;
+      let tIdx = 0;
+      for (let row = 0; row < input2.shape[0]; ++row) {
+        const resIdx = row * outShape[1] + colOffset;
+        for (let col = 0; col < input2.shape[1]; ++col) {
+          outVals[resIdx + col] = decodedData[tIdx++];
+        }
+      }
+      colOffset += input2.shape[1];
+    });
+  }
+  return outVals;
+}
+var equalImpl = createSimpleBinaryKernelImpl((a, b) => a === b ? 1 : 0);
+var equal2 = binaryKernelFunc(Equal, equalImpl, null, "bool");
+var equalConfig = {
+  kernelName: Equal,
+  backendName: "cpu",
+  kernelFunc: equal2
+};
+var expImpl = createSimpleUnaryImpl((xi) => Math.exp(xi));
+var exp2 = unaryKernelFuncFromImpl(Exp, expImpl, "float32");
+var expConfig = {
+  kernelName: Exp,
+  backendName: "cpu",
+  kernelFunc: exp2
+};
+var expm1Impl = createSimpleUnaryImpl((xi) => Math.expm1(xi));
+var expm12 = unaryKernelFuncFromImpl(Expm1, expm1Impl);
+var expm1Config = {
+  kernelName: Expm1,
+  backendName: "cpu",
+  kernelFunc: expm12
+};
+var floorImpl = createSimpleUnaryImpl((xi) => Math.floor(xi));
+var floor2 = unaryKernelFuncFromImpl(Floor, floorImpl);
+var floorConfig = {
+  kernelName: Floor,
+  backendName: "cpu",
+  kernelFunc: floor2
+};
+function gatherNdImpl(indicesData, paramsBuf, dtype, numSlices, sliceRank, sliceSize, strides, paramsShape, paramsSize) {
+  const outBuf = buffer([numSlices, sliceSize], dtype);
+  for (let i = 0; i < numSlices; i++) {
+    const index = [];
+    let flattenIndex = 0;
+    for (let j = 0; j < sliceRank; j++) {
+      const dim = indicesData[i * sliceRank + j];
+      flattenIndex += dim * strides[j];
+      index.push(dim);
+    }
+    if (flattenIndex < 0 || flattenIndex >= paramsSize / sliceSize) {
+      throw new Error(`Invalid indices: ${index} does not index into ${paramsShape}`);
+    }
+    for (let k = 0; k < sliceSize; k++) {
+      outBuf.values[i * sliceSize + k] = paramsBuf.get(...paramsBuf.indexToLoc(flattenIndex * sliceSize + k));
+    }
+  }
+  return outBuf;
+}
+function gatherV2Impl(xBuf, indicesBuf, flattenOutputShape) {
+  const outBuf = buffer(flattenOutputShape, xBuf.dtype);
+  for (let i = 0; i < outBuf.size; ++i) {
+    const newLoc = outBuf.indexToLoc(i);
+    const originalLoc = newLoc.slice();
+    const batchIdx = originalLoc[0];
+    const indicesIdx = originalLoc[2];
+    const indicesIndex = indicesBuf.locToIndex([batchIdx, indicesIdx]);
+    originalLoc[2] = indicesBuf.values[indicesIndex];
+    const originalIndex = xBuf.locToIndex(originalLoc);
+    if (0 <= originalIndex && originalIndex < xBuf.values.length) {
+      outBuf.values[i] = xBuf.values[originalIndex];
+    }
+  }
+  return outBuf;
+}
+var greaterImpl = createSimpleBinaryKernelImpl((a, b) => a > b ? 1 : 0);
+var greater3 = binaryKernelFunc(Greater, greaterImpl, null, "bool");
+var greaterConfig = {
+  kernelName: Greater,
+  backendName: "cpu",
+  kernelFunc: greater3
+};
+var greaterEqualImpl = createSimpleBinaryKernelImpl((a, b) => a >= b ? 1 : 0);
+var greaterEqual2 = binaryKernelFunc(GreaterEqual, greaterEqualImpl, null, "bool");
+var greaterEqualConfig = {
+  kernelName: GreaterEqual,
+  backendName: "cpu",
+  kernelFunc: greaterEqual2
+};
+var lessImpl = createSimpleBinaryKernelImpl((a, b) => a < b ? 1 : 0);
+var less3 = binaryKernelFunc(Less, lessImpl, null, "bool");
+var lessConfig = {
+  kernelName: Less,
+  backendName: "cpu",
+  kernelFunc: less3
+};
+var lessEqualImpl = createSimpleBinaryKernelImpl((a, b) => a <= b ? 1 : 0);
+var lessEqual2 = binaryKernelFunc(LessEqual, lessEqualImpl, null, "bool");
+var lessEqualConfig = {
+  kernelName: LessEqual,
+  backendName: "cpu",
+  kernelFunc: lessEqual2
+};
+function linSpaceImpl(start, stop, num) {
+  const step5 = (stop - start) / (num - 1);
+  const values = util_exports.makeZerosTypedArray(num, "float32");
+  values[0] = start;
+  for (let i = 1; i < values.length; i++) {
+    values[i] = values[i - 1] + step5;
+  }
+  return values;
+}
+var logImpl = createSimpleUnaryImpl((xi) => Math.log(xi));
+var log3 = unaryKernelFuncFromImpl(Log, logImpl);
+var logConfig = {
+  kernelName: Log,
+  backendName: "cpu",
+  kernelFunc: log3
+};
+function maxImpl(aVals, reduceSize, outShape, dtype) {
+  const vals = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(outShape));
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let max6 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (Number.isNaN(value) || value > max6) {
+        max6 = value;
+      }
+    }
+    vals[i] = max6;
+  }
+  return vals;
+}
+var maximumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.max(aValue, bValue));
+var maximum3 = binaryKernelFunc(Maximum, maximumImpl);
+var maximumConfig = {
+  kernelName: Maximum,
+  backendName: "cpu",
+  kernelFunc: maximum3
+};
+var minimumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.min(aValue, bValue));
+var minimum3 = binaryKernelFunc(Minimum, minimumImpl);
+var minimumConfig = {
+  kernelName: Minimum,
+  backendName: "cpu",
+  kernelFunc: minimum3
+};
+var multiplyImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue * bValue);
+var multiplyComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return {
+    real: aReal * bReal - aImag * bImag,
+    imag: aReal * bImag + aImag * bReal
+  };
+});
+var multiply2 = binaryKernelFunc(Multiply, multiplyImpl, multiplyComplexImpl);
+var multiplyConfig = {
+  kernelName: Multiply,
+  backendName: "cpu",
+  kernelFunc: multiply2
+};
+function negImpl(xVals, xShape, xDtype) {
+  const minusOne = util_exports.createScalarValue(-1, xDtype);
+  return multiplyImpl([], xShape, minusOne, xVals, xDtype);
+}
+function neg2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "neg");
+  const xVals = backend2.data.get(x.dataId).values;
+  const [res, newShape] = negImpl(xVals, x.shape, x.dtype);
+  return backend2.makeTensorInfo(newShape, x.dtype, res);
+}
+var negConfig = {
+  kernelName: Neg,
+  backendName: "cpu",
+  kernelFunc: neg2
+};
+var notEqualImpl = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0);
+var notEqual2 = binaryKernelFunc(NotEqual, notEqualImpl, null, "bool");
+var notEqualConfig = {
+  kernelName: NotEqual,
+  backendName: "cpu",
+  kernelFunc: notEqual2
+};
+function transposeImpl(xVals, xShape, dtype, perm, newShape) {
+  const xRank = xShape.length;
+  const xSize = util_exports.sizeFromShape(xShape);
+  const xStrides = util_exports.computeStrides(xShape);
+  const newStrides = util_exports.computeStrides(newShape);
+  const result = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(newShape));
+  for (let i = 0; i < xSize; ++i) {
+    const loc = util_exports.indexToLoc(i, xRank, xStrides);
+    const newLoc = new Array(loc.length);
+    for (let i2 = 0; i2 < newLoc.length; i2++) {
+      newLoc[i2] = loc[perm[i2]];
+    }
+    const newIndex = util_exports.locToIndex(newLoc, xRank, newStrides);
+    result[newIndex] = xVals[i];
+  }
+  return result;
+}
+function transpose2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x } = inputs;
+  const { perm } = attrs;
+  assertNotComplex(x, "transpose");
+  const xRank = x.shape.length;
+  const newShape = new Array(xRank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[perm[i]];
+  }
+  const values = backend2.data.get(x.dataId).values;
+  const result = transposeImpl(values, x.shape, x.dtype, perm, newShape);
+  const dataId = backend2.write(result, newShape, x.dtype);
+  return { dataId, shape: newShape, dtype: x.dtype };
+}
+var transposeConfig = {
+  kernelName: Transpose,
+  backendName: "cpu",
+  kernelFunc: transpose2
+};
+function prodImpl(xShape, xDtype, xVals, reductionAxes) {
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, reductionAxes);
+  const outDtype = upcastType(xDtype, "int32");
+  const outVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), outDtype);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  for (let i = 0; i < outVals.length; ++i) {
+    const offset = i * reduceSize;
+    let prod5 = 1;
+    for (let j = 0; j < reduceSize; ++j) {
+      prod5 *= xVals[offset + j];
+    }
+    outVals[i] = prod5;
+  }
+  return { outVals, outShape, outDtype };
+}
+function prod2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "prod");
+  const xRank = x.shape.length;
+  const axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);
+  let reductionAxes = axes;
+  let permutedX = x;
+  const intermediateTensorInfos = [];
+  if (permutation != null) {
+    permutedX = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+    intermediateTensorInfos.push(permutedX);
+    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);
+  }
+  const xVals = backend2.data.get(permutedX.dataId).values;
+  const { outVals, outShape, outDtype } = prodImpl(permutedX.shape, permutedX.dtype, xVals, reductionAxes);
+  let resultShape = outShape;
+  if (keepDims) {
+    resultShape = backend_util_exports.expandShapeToKeepDim(outShape, axes);
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(resultShape, outDtype, outVals);
+}
+var prodConfig = {
+  kernelName: Prod,
+  backendName: "cpu",
+  kernelFunc: prod2
+};
+function validateIndices(indices, indicesShape, numParams) {
+  indices.forEach((index, i) => {
+    if (index < 0 || index >= numParams) {
+      const locString = util_exports.indexToLoc(i, indicesShape.length, util_exports.computeStrides(indicesShape)).join(",");
+      throw new Error(`indices[${locString}] = ${index} is not in [0, ${numParams})`);
+    }
+  });
+}
+function validateSplits(paramsNestedSplits, numParamsDenseValues) {
+  for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {
+    const splits = paramsNestedSplits[dim];
+    const lastSplit = dim === paramsNestedSplits.length - 1 ? numParamsDenseValues : paramsNestedSplits[dim + 1].length;
+    if (splits.length === 0) {
+      throw new Error("Ragged splits may not be empty");
+    }
+    if (splits[0] < 0) {
+      throw new Error("Ragged splits must be non-negative");
+    }
+    if (splits[splits.length - 1] > lastSplit) {
+      throw new Error("Ragged splits must not point past values");
+    }
+    for (let i = 1; i < splits.length; ++i) {
+      if (splits[i - 1] > splits[i]) {
+        throw new Error("Ragged splits must be sorted in ascending order");
+      }
+    }
+  }
+}
+function makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues) {
+  const valueSlices = [];
+  let numValues = 0;
+  const numSplits = indicesShape.length - 1 + paramsNestedSplits.length;
+  const outSplits = new Array(numSplits).fill(null).map(() => [0]);
+  validateSplits(paramsNestedSplits, numParamsDenseValues);
+  let nrows = 1;
+  for (let dim = 0; dim < indicesShape.length - 1; ++dim) {
+    nrows *= indicesShape[dim];
+    const rowLength = indicesShape[dim + 1];
+    for (let i = 1; i < nrows + 1; ++i) {
+      outSplits[dim].push(i * rowLength);
+    }
+  }
+  for (let i = 0; i < indices.length; ++i) {
+    let start = indices[i];
+    let limit = indices[i] + 1;
+    for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {
+      const splits = paramsNestedSplits[dim];
+      const outDim = dim + indicesShape.length - 1;
+      if (outDim >= 0) {
+        const outSplitsOutDim = outSplits[outDim];
+        const delta = outSplitsOutDim[outSplitsOutDim.length - 1] - splits[start];
+        for (let j = start; j < limit; ++j) {
+          outSplits[outDim].push(splits[j + 1] + delta);
+        }
+      }
+      start = splits[start];
+      limit = splits[limit];
+    }
+    if (limit !== start) {
+      valueSlices.push([start, limit]);
+      numValues += limit - start;
+    }
+  }
+  return { outSplits, valueSlices, numValues };
+}
+function getSplits(outSplits) {
+  const splitsOut = [];
+  for (let i = 0; i < outSplits.length; ++i) {
+    const numSplits = outSplits[i].length;
+    const splits = util_exports.getArrayFromDType("int32", numSplits);
+    splitsOut.push(splits);
+    outSplits[i].forEach((value, j) => splits[j] = value);
+  }
+  return splitsOut;
+}
+function computeFlatOuterDims(orig, numOutDims) {
+  const outDims = orig.slice(0, numOutDims);
+  while (outDims.length < numOutDims) {
+    outDims.push(1);
+  }
+  for (let inDim = numOutDims; inDim < orig.length; inDim++) {
+    outDims[numOutDims - 1] *= orig[inDim];
+  }
+  return outDims;
+}
+function writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, values, valuesShape) {
+  const denseM = computeFlatOuterDims(paramsDenseValuesShape, 2)[1];
+  const valuesM = computeFlatOuterDims(valuesShape, 2)[1];
+  let outPos = 0;
+  for (const slice5 of valueSlices) {
+    for (let i = slice5[0]; i < slice5[1]; ++i) {
+      for (let j = 0; j < valueSize; ++j) {
+        values[outPos * valuesM + j] = paramsDenseValues[i * denseM + j];
+      }
+      ++outPos;
+    }
+  }
+}
+function getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues) {
+  const valuesShape = paramsDenseValuesShape.slice();
+  valuesShape[0] = numValues;
+  const valuesOut = util_exports.getArrayFromDType(paramsDenseValuesDType, util_exports.sizeFromShape(valuesShape));
+  const numElements = paramsDenseValues.length;
+  const valueSize = numElements === 0 ? 0 : numElements / paramsDenseValuesShape[0];
+  writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, valuesOut, valuesShape);
+  return [valuesOut, valuesShape];
+}
+function raggedGatherImpl(paramsNestedSplits, paramsNestedSplitsShapes, paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, indices, indicesShape, outputRaggedRank) {
+  if (paramsNestedSplits.length === 0) {
+    throw new Error("paramsNestedSplits must be non empty");
+  }
+  if (paramsNestedSplitsShapes[0].length === 0) {
+    throw new Error("Split tensors must not be scalars");
+  }
+  const numParams = paramsNestedSplitsShapes[0][0] - 1;
+  validateIndices(indices, indicesShape, numParams);
+  if (paramsDenseValuesShape.length === 0) {
+    throw new Error("params.rank must be nonzero");
+  }
+  const numParamsDenseValues = paramsDenseValuesShape[0];
+  const { outSplits, valueSlices, numValues } = makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues);
+  const outputNestedSplits = getSplits(outSplits);
+  const outputDenseValues = getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues);
+  return [outputNestedSplits, outputDenseValues[0], outputDenseValues[1]];
+}
+var INT32_MAX2 = 2147483647;
+function raggedRangeImpl(starts, startsShape, startsDType, limits, limitsShape, deltas, deltasShape) {
+  if (startsShape.length > 1) {
+    throw new Error("starts must be a scalar or vector");
+  }
+  if (limitsShape.length > 1) {
+    throw new Error("limits must be a scalar or vector");
+  }
+  if (deltasShape.length > 1) {
+    throw new Error("deltas must be a scalar or vector");
+  }
+  const broadcastStarts = startsShape.length === 0;
+  const broadcastLimits = limitsShape.length === 0;
+  const broadcastDeltas = deltasShape.length === 0;
+  const inSizes = [];
+  if (!broadcastStarts) {
+    inSizes.push(startsShape[0]);
+  }
+  if (!broadcastLimits) {
+    inSizes.push(limitsShape[0]);
+  }
+  if (!broadcastDeltas) {
+    inSizes.push(deltasShape[0]);
+  }
+  for (let i = 1; i < inSizes.length; ++i) {
+    if (inSizes[i] !== inSizes[i - 1]) {
+      throw new Error("starts, limits, and deltas must have the same shape");
+    }
+  }
+  const nRows = inSizes.length === 0 ? 1 : inSizes[0];
+  const rtNestedSplits = util_exports.getArrayFromDType("int32", nRows + 1);
+  rtNestedSplits[0] = 0;
+  for (let row = 0; row < nRows; ++row) {
+    const start = broadcastStarts ? starts[0] : starts[row];
+    const limit = broadcastLimits ? limits[0] : limits[row];
+    const delta = broadcastDeltas ? deltas[0] : deltas[row];
+    if (delta === 0) {
+      throw new Error("Requires delta != 0");
+    }
+    let size;
+    if (delta > 0 && limit < start || delta < 0 && limit > start) {
+      size = 0;
+    } else {
+      size = Math.ceil(Math.abs((limit - start) / delta));
+      if (size > INT32_MAX2) {
+        throw new Error(`Requires ((limit - start) / delta) <= ${INT32_MAX2}`);
+      }
+    }
+    rtNestedSplits[row + 1] = rtNestedSplits[row] + size;
+  }
+  const nVals = rtNestedSplits[nRows];
+  const rtDenseValues = util_exports.getArrayFromDType(startsDType, nVals);
+  let valueIndex = 0;
+  for (let row = 0; row < nRows; ++row) {
+    const rowSize = rtNestedSplits[row + 1] - rtNestedSplits[row];
+    let value = broadcastStarts ? starts[0] : starts[row];
+    const delta = broadcastDeltas ? deltas[0] : deltas[row];
+    for (let i = 0; i < rowSize; ++i) {
+      rtDenseValues[valueIndex++] = value;
+      value += delta;
+    }
+  }
+  return [rtNestedSplits, rtDenseValues];
+}
+var RowPartitionType2 = backend_util_exports.RowPartitionType;
+var RaggedTensorToTensorOp = class {
+  constructor(shape, shapeShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypeStrings) {
+    this.shape = shape;
+    this.shapeShape = shapeShape;
+    this.values = values;
+    this.valuesShape = valuesShape;
+    this.valuesDType = valuesDType;
+    this.defaultValue = defaultValue;
+    this.defaultValueShape = defaultValueShape;
+    this.rowPartitionValues = rowPartitionValues;
+    this.rowPartitionValuesShapes = rowPartitionValuesShapes;
+    this.rowPartitionTypes = backend_util_exports.getRowPartitionTypesHelper(rowPartitionTypeStrings);
+    this.raggedRank = backend_util_exports.getRaggedRank(this.rowPartitionTypes);
+  }
+  getRowPartitionTypeByDimension(dimension) {
+    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {
+      return this.rowPartitionTypes[dimension + 1];
+    } else {
+      return this.rowPartitionTypes[dimension];
+    }
+  }
+  getRowPartitionTensor(dimension) {
+    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {
+      return this.rowPartitionValues[dimension + 1];
+    } else {
+      return this.rowPartitionValues[dimension];
+    }
+  }
+  getMaxWidth(dimension) {
+    const rowPartitionTensor = this.getRowPartitionTensor(dimension - 1);
+    switch (this.getRowPartitionTypeByDimension(dimension - 1)) {
+      case RowPartitionType2.VALUE_ROWIDS:
+        return RaggedTensorToTensorOp.getMaxWidthValueRowID(rowPartitionTensor);
+      case RowPartitionType2.ROW_SPLITS:
+        return RaggedTensorToTensorOp.getMaxWidthRowSplit(rowPartitionTensor);
+      default:
+        throw new Error(`Cannot handle partition type ${RowPartitionType2[this.getRowPartitionTypeByDimension(dimension - 1)]}`);
+    }
+  }
+  static getMaxWidthRowSplit(rowSplit) {
+    const tensorLength = rowSplit.length;
+    if (tensorLength === 0 || tensorLength === 1) {
+      return 0;
+    }
+    let maxWidth = 0;
+    for (let i = 0; i < tensorLength - 1; ++i) {
+      const currentWidth = rowSplit[i + 1] - rowSplit[i];
+      if (currentWidth > maxWidth) {
+        maxWidth = currentWidth;
+      }
+    }
+    return maxWidth;
+  }
+  static getMaxWidthValueRowID(valueRowIds) {
+    const indexLength = valueRowIds.length;
+    if (indexLength === 0) {
+      return 0;
+    }
+    let firstEqualIndex = 0;
+    let firstEqualIndexValue = valueRowIds[0];
+    let maxWidth = 0;
+    for (let i = 1; i < indexLength; ++i) {
+      const value = valueRowIds[i];
+      if (value !== firstEqualIndexValue) {
+        firstEqualIndexValue = value;
+        maxWidth = Math.max(i - firstEqualIndex, maxWidth);
+        firstEqualIndex = i;
+      }
+    }
+    return Math.max(indexLength - firstEqualIndex, maxWidth);
+  }
+  tensorShapeFromTensor(t, tShape, isPartial = true) {
+    if (tShape.length === 0) {
+      if (t[0] === -1) {
+        return [];
+      }
+      throw new Error(`The only valid scalar shape tensor is the fully unknown shape specified as -1.`);
+    }
+    return makeShape(t, isPartial);
+  }
+  calculateOutputSize(firstDim) {
+    const valueShape = this.valuesShape;
+    const defaultValueShape = this.defaultValueShape;
+    backend_util_exports.validateDefaultValueShape(defaultValueShape, valueShape);
+    const shape = this.tensorShapeFromTensor(this.shape, this.shapeShape);
+    const outputShape = backend_util_exports.combineRaggedTensorToTensorShapes(this.raggedRank, shape, valueShape);
+    const result = outputShape;
+    if (result[0] < 0) {
+      result[0] = firstDim;
+    }
+    for (let i = 1; i <= this.raggedRank; ++i) {
+      if (result[i] < 0) {
+        result[i] = this.getMaxWidth(i);
+      }
+    }
+    return result;
+  }
+  calculateFirstParentOutputIndex(firstDimension, outputIndexMultiplier, firstDimensionOutput) {
+    const minDimension = Math.min(firstDimension, firstDimensionOutput);
+    const result = [];
+    let currentOutputIndex = 0;
+    for (let i = 0; i < minDimension; ++i, currentOutputIndex += outputIndexMultiplier) {
+      result.push(currentOutputIndex);
+    }
+    for (let i = minDimension; i < firstDimension; ++i) {
+      result.push(-1);
+    }
+    util_exports.assert(result.length === firstDimension, () => "Final length of result must be equal to firstDimension.");
+    return result;
+  }
+  calculateOutputIndexRowSplit(rowSplit, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const rowSplitSize = rowSplit.length;
+    const result = [];
+    for (let i = 0; i < rowSplitSize - 1; ++i) {
+      const rowLength = rowSplit[i + 1] - rowSplit[i];
+      let realLength = Math.min(outputSize, rowLength);
+      let parentOutputIndexCurrent = parentOutputIndex[i];
+      if (parentOutputIndexCurrent === -1) {
+        realLength = 0;
+      }
+      for (let j = 0; j < realLength; ++j) {
+        result.push(parentOutputIndexCurrent);
+        parentOutputIndexCurrent += outputIndexMultiplier;
+      }
+      for (let j = 0; j < rowLength - realLength; ++j) {
+        result.push(-1);
+      }
+    }
+    if (rowSplitSize > 0 && result.length !== rowSplit[rowSplitSize - 1]) {
+      throw new Error("Invalid row split size.");
+    }
+    return result;
+  }
+  calculateOutputIndexValueRowID(valueRowIds, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const indexSize = valueRowIds.length;
+    const result = [];
+    if (indexSize === 0) {
+      return [];
+    }
+    let currentOutputColumn = 0;
+    let currentValueRowId = valueRowIds[0];
+    if (currentValueRowId >= parentOutputIndex.length) {
+      throw new Error(`Got currentValueRowId=${currentValueRowId}, which is not less than ${parentOutputIndex.length}`);
+    }
+    let currentOutputIndex = parentOutputIndex[currentValueRowId];
+    result.push(currentOutputIndex);
+    for (let i = 1; i < indexSize; ++i) {
+      const nextValueRowId = valueRowIds[i];
+      if (nextValueRowId === currentValueRowId) {
+        if (currentOutputIndex >= 0) {
+          ++currentOutputColumn;
+          if (currentOutputColumn < outputSize) {
+            currentOutputIndex += outputIndexMultiplier;
+          } else {
+            currentOutputIndex = -1;
+          }
+        }
+      } else {
+        currentOutputColumn = 0;
+        currentValueRowId = nextValueRowId;
+        if (nextValueRowId >= parentOutputIndex.length) {
+          throw new Error(`Got nextValueRowId=${nextValueRowId} which is not less than ${parentOutputIndex.length}`);
+        }
+        currentOutputIndex = parentOutputIndex[nextValueRowId];
+      }
+      result.push(currentOutputIndex);
+    }
+    if (result.length !== valueRowIds.length) {
+      throw new Error("Invalid row ids.");
+    }
+    return result;
+  }
+  calculateOutputIndex(dimension, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const rowPartitionTensor = this.getRowPartitionTensor(dimension);
+    const partitionType = this.getRowPartitionTypeByDimension(dimension);
+    switch (partitionType) {
+      case RowPartitionType2.VALUE_ROWIDS:
+        return this.calculateOutputIndexValueRowID(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);
+      case RowPartitionType2.ROW_SPLITS:
+        if (rowPartitionTensor.length - 1 > parentOutputIndex.length) {
+          throw new Error(`Row partition size is greater than output size: ${rowPartitionTensor.length - 1} > ${parentOutputIndex.length}`);
+        }
+        return this.calculateOutputIndexRowSplit(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);
+      default:
+        throw new Error(`Unsupported partition type: ${RowPartitionType2[partitionType]}`);
+    }
+  }
+  getFirstDimensionSize() {
+    const firstPartitionTensor = this.rowPartitionValues[0];
+    if (this.rowPartitionTypes.length === 0) {
+      throw new Error("No row_partition_types given.");
+    }
+    const firstPartitionType = this.rowPartitionTypes[0];
+    switch (firstPartitionType) {
+      case RowPartitionType2.FIRST_DIM_SIZE:
+        return firstPartitionTensor[0];
+      case RowPartitionType2.VALUE_ROWIDS:
+        throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");
+      case RowPartitionType2.ROW_SPLITS:
+        return this.rowPartitionValuesShapes[0][0] - 1;
+      default:
+        throw new Error(`Cannot handle type ${RowPartitionType2[firstPartitionType]}`);
+    }
+  }
+  compute() {
+    const firstPartitionTensor = this.rowPartitionValues[0];
+    if (firstPartitionTensor.length <= 0) {
+      throw new Error("Invalid first partition input. Tensor requires at least one element.");
+    }
+    const firstDimension = this.getFirstDimensionSize();
+    const outputSize = this.calculateOutputSize(firstDimension);
+    const multiplier = new Array(this.raggedRank + 1);
+    multiplier[multiplier.length - 1] = 1;
+    for (let i = multiplier.length - 2; i >= 0; --i) {
+      multiplier[i] = multiplier[i + 1] * outputSize[i + 1];
+    }
+    const outputShape = makeShape(outputSize, false);
+    const outputTensor = util_exports.getArrayFromDType(this.valuesDType, util_exports.sizeFromShape(outputShape));
+    const fullSize = multiplier[0] * outputSize[0];
+    if (fullSize > 0) {
+      let outputIndex = this.calculateFirstParentOutputIndex(firstDimension, multiplier[0], outputSize[0]);
+      for (let i = 1; i <= this.raggedRank; ++i) {
+        const newOutputIndex = this.calculateOutputIndex(i - 1, outputIndex, multiplier[i], outputSize[i]);
+        outputIndex = newOutputIndex;
+      }
+      this.setOutput(this.raggedRank, outputIndex, outputTensor, outputShape);
+    }
+    return [outputShape, outputTensor];
+  }
+  setOutput(raggedRank, outputIndex, outputTensor, outputShape) {
+    if (outputTensor.length === 0) {
+      return;
+    }
+    const valuesBase = this.values;
+    const outputBase = outputTensor;
+    let elementShape = outputShape.slice();
+    elementShape = elementShape.slice(raggedRank + 1);
+    const valueElementSize = util_exports.sizeFromShape(elementShape);
+    const outputIndexSize = outputIndex.length;
+    let defaultValue = this.defaultValue;
+    if (defaultValue.length !== valueElementSize && defaultValue.length !== 1) {
+      const srcShape = this.defaultValueShape;
+      tidy(() => {
+        const defaultValueTensor = reshape(defaultValue, srcShape);
+        const bCastDefault = broadcastTo(defaultValueTensor, elementShape);
+        defaultValue = bCastDefault.dataSync();
+      });
+    }
+    let srcStart = 0;
+    let dstStart = 0;
+    let dstEnd = 0;
+    for (let srcI = 0; srcI <= outputIndexSize; ++srcI) {
+      let dstI = srcI < outputIndexSize ? outputIndex[srcI] : -1;
+      if (dstI === dstEnd) {
+        ++dstEnd;
+        continue;
+      }
+      if (dstStart < dstEnd) {
+        const src = valuesBase.subarray(srcStart * valueElementSize);
+        const dst = outputBase.subarray(dstStart * valueElementSize);
+        const nVals = (dstEnd - dstStart) * valueElementSize;
+        copyArray(dst, src, nVals);
+      }
+      if (srcI >= outputIndexSize) {
+        const outputSize = outputTensor.length;
+        dstI = Math.floor(outputSize / valueElementSize);
+      }
+      if (dstI > dstEnd) {
+        if (this.defaultValue.length === 1) {
+          outputBase.subarray(dstEnd * valueElementSize, dstI * valueElementSize).fill(this.defaultValue[0]);
+          dstEnd = dstI;
+        } else {
+          while (dstI > dstEnd) {
+            const dst = outputBase.slice(dstEnd * valueElementSize);
+            copyArray(dst, defaultValue, valueElementSize);
+            ++dstEnd;
+          }
+        }
+      }
+      if (dstI < 0) {
+        srcStart = srcI + 1;
+        dstStart = dstEnd;
+      } else {
+        srcStart = srcI;
+        dstStart = dstEnd;
+        dstEnd = dstStart + 1;
+      }
+    }
+  }
+};
+function copyArray(dst, src, size) {
+  for (let i = 0; i < size; i++) {
+    dst[i] = src[i];
+  }
+}
+function makeShape(shape, isPartial) {
+  const out = [];
+  for (let dim of shape) {
+    if (dim < 0) {
+      if (!isPartial) {
+        throw new Error(`Dimension ${dim} must be >= 0`);
+      }
+      if (dim < -1) {
+        throw new Error(`Dimension ${dim} must be >= -1`);
+      }
+      dim = -1;
+    }
+    out.push(dim);
+  }
+  return out;
+}
+function raggedTensorToTensorImpl(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes) {
+  return new RaggedTensorToTensorOp(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes).compute();
+}
+function rangeImpl(start, stop, step5, dtype) {
+  const sameStartStop = start === stop;
+  const increasingRangeNegativeStep = start < stop && step5 < 0;
+  const decreasingRangePositiveStep = stop < start && step5 > 1;
+  if (sameStartStop || increasingRangeNegativeStep || decreasingRangePositiveStep) {
+    return util_exports.makeZerosTypedArray(0, dtype);
+  }
+  const numElements = Math.abs(Math.ceil((stop - start) / step5));
+  const values = util_exports.makeZerosTypedArray(numElements, dtype);
+  if (stop < start && step5 === 1) {
+    step5 = -1;
+  }
+  values[0] = start;
+  for (let i = 1; i < values.length; i++) {
+    values[i] = values[i - 1] + step5;
+  }
+  return values;
+}
+var rsqrtImpl = createSimpleUnaryImpl((xi) => 1 / Math.sqrt(xi));
+var rsqrt2 = unaryKernelFuncFromImpl(Rsqrt, rsqrtImpl);
+var rsqrtConfig = {
+  kernelName: Rsqrt,
+  backendName: "cpu",
+  kernelFunc: rsqrt2
+};
+function scatterImpl(indices, updates, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, defaultValue, sumDupeIndices) {
+  const flattenShape = [outputSize / sliceSize, sliceSize];
+  const indicesData = indices.values;
+  const updatesData = updates.values;
+  if (outputSize === 0) {
+    return buffer(shape, updates.dtype);
+  }
+  const outBuf = buffer(flattenShape, updates.dtype);
+  if (typeof defaultValue === "string") {
+    outBuf.values.fill(defaultValue);
+  } else if (typeof defaultValue === "number") {
+    outBuf.values.fill(defaultValue);
+  } else if (typeof defaultValue === "boolean") {
+    outBuf.values.fill(+defaultValue);
+  }
+  for (let i = 0; i < numUpdates; i++) {
+    const index = [];
+    let flattenIndex = 0;
+    for (let j = 0; j < sliceRank; j++) {
+      const dim = indicesData[i * sliceRank + j];
+      index.push(dim);
+      flattenIndex += dim * strides[j];
+    }
+    if (flattenIndex < 0 || flattenIndex >= outputSize / sliceSize) {
+      throw new Error(`Invalid indices: ${index} does not index into ${shape}`);
+    }
+    for (let k = 0; k < sliceSize; k++) {
+      if (sumDupeIndices) {
+        outBuf.values[flattenIndex * sliceSize + k] += updatesData[i * sliceSize + k];
+      } else {
+        outBuf.values[flattenIndex * sliceSize + k] = updates.rank === 0 ? updatesData[0] : updatesData[i * sliceSize + k];
+      }
+    }
+  }
+  return outBuf;
+}
+var sigmoidImpl = createSimpleUnaryImpl((xi) => 1 / (1 + Math.exp(-xi)));
+var sigmoid2 = unaryKernelFunc(Sigmoid, (xi) => 1 / (1 + Math.exp(-xi)));
+var sigmoidConfig = {
+  kernelName: Sigmoid,
+  backendName: "cpu",
+  kernelFunc: sigmoid2
+};
+function sliceImpl(vals, begin, size, shape, dtype) {
+  const isContinous = slice_util_exports.isSliceContinous(shape, begin, size);
+  const length = util_exports.sizeFromShape(size);
+  const xStrides = util_exports.computeStrides(shape);
+  if (isContinous) {
+    const flatOffset = slice_util_exports.computeFlatOffset(begin, xStrides);
+    if (dtype === "string") {
+      return vals.slice(flatOffset, flatOffset + length);
+    }
+    return vals.subarray(flatOffset, flatOffset + length);
+  }
+  const decodedData = dtype === "string" ? backend_util_exports.fromUint8ToStringArray(vals) : vals;
+  const inBuf = buffer(shape, dtype, decodedData);
+  const outBuf = buffer(size, dtype);
+  for (let i = 0; i < outBuf.size; ++i) {
+    const outLoc = outBuf.indexToLoc(i);
+    const inLoc = outLoc.map((idx, j) => idx + begin[j]);
+    outBuf.set(inBuf.get(...inLoc), ...outLoc);
+  }
+  if (dtype === "string") {
+    return backend_util_exports.fromStringArrayToUint8(outBuf.values);
+  }
+  return outBuf.values;
+}
+function slice2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, size } = attrs;
+  assertNotComplex(x, "slice");
+  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);
+  slice_util_exports.assertParamsValid(x, $begin, $size);
+  const vals = backend2.data.get(x.dataId).values;
+  const outVals = sliceImpl(vals, $begin, $size, x.shape, x.dtype);
+  return backend2.makeTensorInfo($size, x.dtype, outVals);
+}
+var sliceConfig = {
+  kernelName: Slice,
+  backendName: "cpu",
+  kernelFunc: slice2
+};
+function sparseFillEmptyRowsImpl(indices, indicesShape, indicesDType, values, valuesDType, denseShape, defaultValue) {
+  const indicesCount = indicesShape[0];
+  const denseRows = denseShape[0];
+  const emptyRowIndicator = new Array(denseRows);
+  const reverseIndexMap = new Array(indicesCount);
+  const rank = indicesShape[1];
+  if (denseRows === 0) {
+    if (indicesCount !== 0) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesCount));
+    }
+    const outputIndices = util_exports.getArrayFromDType(indicesDType, 0);
+    const outputValues = util_exports.getArrayFromDType(valuesDType, 0);
+    return [
+      outputIndices,
+      [0, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  }
+  let rowsAreOrdered = true;
+  let lastIndicesRow = 0;
+  const csrOffset = new Array(denseRows).fill(0);
+  for (let i = 0; i < indicesCount; ++i) {
+    const row = indices[i * rank];
+    if (row < 0) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(i, row));
+    }
+    if (row >= denseRows) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(i, row, denseRows));
+    }
+    ++csrOffset[row];
+    rowsAreOrdered = rowsAreOrdered && row >= lastIndicesRow;
+    lastIndicesRow = row;
+  }
+  let allRowsFull = true;
+  for (let row = 0; row < denseRows; ++row) {
+    const rowEmpty = csrOffset[row] === 0;
+    emptyRowIndicator[row] = rowEmpty;
+    allRowsFull = allRowsFull && !rowEmpty;
+    csrOffset[row] = Math.max(csrOffset[row], 1);
+    if (row > 0) {
+      csrOffset[row] += csrOffset[row - 1];
+    }
+  }
+  if (allRowsFull && rowsAreOrdered) {
+    const outputIndices = indices;
+    const outputValues = values;
+    for (let i = 0; i < indicesCount; ++i) {
+      reverseIndexMap[i] = i;
+    }
+    return [
+      outputIndices,
+      [indicesCount, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  } else {
+    const fullIndicesCount = csrOffset[denseRows - 1];
+    const outputIndices = util_exports.getArrayFromDType(indicesDType, fullIndicesCount * rank);
+    const outputValues = util_exports.getArrayFromDType(valuesDType, fullIndicesCount);
+    const filledCount = new Array(denseRows).fill(0);
+    for (let i = 0; i < indicesCount; ++i) {
+      const row = indices[i * rank];
+      const offset = filledCount[row];
+      const outputI = (row === 0 ? 0 : csrOffset[row - 1]) + offset;
+      filledCount[row]++;
+      for (let j = 0; j < rank; ++j) {
+        outputIndices[outputI * rank + j] = indices[i * rank + j];
+      }
+      outputValues[outputI] = values[i];
+      reverseIndexMap[i] = outputI;
+    }
+    for (let row = 0; row < denseRows; ++row) {
+      const rowCount = filledCount[row];
+      if (rowCount === 0) {
+        const startingIndex = row === 0 ? 0 : csrOffset[row - 1];
+        outputIndices[startingIndex * rank + 0] = row;
+        for (let col = 1; col < rank; ++col) {
+          outputIndices[startingIndex * rank + col] = 0;
+        }
+        outputValues[startingIndex] = defaultValue;
+      }
+    }
+    return [
+      outputIndices,
+      [fullIndicesCount, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  }
+}
+function sparseReshapeImpl(inputIndices, inputIndicesShape, inputDType, inputShape, targetShape) {
+  const denseSize = util_exports.sizeFromShape(inputShape);
+  const nnz = inputIndicesShape[0];
+  const outputRank = targetShape.length;
+  const outputShape = [];
+  let product = 1;
+  let unknownIndex = -1;
+  for (let d = 0; d < outputRank; ++d) {
+    const size = targetShape[d];
+    if (size === -1) {
+      if (unknownIndex !== -1) {
+        throw new Error(backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(unknownIndex, d));
+      }
+      unknownIndex = d;
+      outputShape.push(1);
+    } else {
+      if (size < 0) {
+        throw new Error(backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(d, size));
+      }
+      product *= size;
+      outputShape.push(size);
+    }
+  }
+  if (unknownIndex !== -1) {
+    if (product <= 0) {
+      throw new Error(backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());
+    }
+    const missing = Math.trunc(denseSize / product);
+    if (product * missing !== denseSize) {
+      throw new Error(backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape));
+    }
+    outputShape[unknownIndex] = missing;
+  }
+  const outputSize = util_exports.sizeFromShape(outputShape);
+  if (outputSize !== denseSize) {
+    throw new Error(backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape));
+  }
+  const inputRank = inputShape.length;
+  const inputStrides = [];
+  if (inputRank > 0) {
+    inputStrides[inputRank - 1] = 1;
+    for (let d = inputRank - 2; d >= 0; --d) {
+      inputStrides[d] = inputStrides[d + 1] * inputShape[d + 1];
+    }
+  }
+  const outputStrides = [];
+  if (outputRank > 0) {
+    outputStrides[outputRank - 1] = 1;
+    for (let d = outputRank - 2; d >= 0; --d) {
+      outputStrides[d] = outputStrides[d + 1] * outputShape[d + 1];
+    }
+  }
+  const newIndices = util_exports.getArrayFromDType(inputDType, nnz * outputRank);
+  for (let i = 0; i < nnz; ++i) {
+    let id = 0;
+    for (let j = 0; j < inputRank; ++j) {
+      id += inputIndices[i * inputRank + j] * inputStrides[j];
+    }
+    for (let j = 0; j < outputRank; ++j) {
+      newIndices[i * outputRank + j] = Math.trunc(id / outputStrides[j]);
+      id %= outputStrides[j];
+    }
+  }
+  return [newIndices, [nnz, outputRank], outputShape];
+}
+function sparseSegmentReductionImpl(input2, inputShape, inputDType, indices, segmentIds, isMean = false, defaultValue = 0) {
+  const numIndices = indices.length;
+  const inputFlat = [inputShape[0], input2.length / inputShape[0]];
+  const numCol = inputFlat[1];
+  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIds[numIndices - 1] + 1 : 0;
+  const outputRows = lastSegmentIdPlusOne;
+  if (outputRows < 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  const outputShape = inputShape.slice();
+  outputShape[0] = outputRows;
+  const outputLength = outputShape.reduce((product, value) => product * value, 1);
+  const output = util_exports.getArrayFromDType(inputDType, outputLength);
+  if (numIndices === 0) {
+    if (outputRows > 0) {
+      output.fill(defaultValue);
+    }
+    return [output, outputShape];
+  }
+  if (outputRows <= 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  let start = 0, end = 1;
+  let uninitializedIndex = 0;
+  let outIndex = segmentIds[start];
+  while (true) {
+    let nextIndex = 0;
+    if (end < numIndices) {
+      nextIndex = segmentIds[end];
+      if (outIndex === nextIndex) {
+        ++end;
+        continue;
+      }
+      if (outIndex >= nextIndex) {
+        throw new Error(backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage());
+      }
+    }
+    if (outIndex < 0 || outIndex >= outputRows) {
+      throw new Error(backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(outIndex, outputRows));
+    }
+    if (outIndex > uninitializedIndex) {
+      output.fill(defaultValue, uninitializedIndex * numCol, outIndex * numCol);
+    }
+    for (let i = start; i < end; ++i) {
+      const index = indices[i];
+      if (index < 0 || index >= inputFlat[0]) {
+        throw new Error(backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(i, indices[i], inputFlat[0]));
+      }
+      for (let j = 0; j < numCol; j++) {
+        output[outIndex * numCol + j] += input2[index * numCol + j];
+      }
+    }
+    if (isMean) {
+      for (let j = 0; j < numCol; j++) {
+        output[outIndex * numCol + j] /= end - start;
+      }
+    }
+    start = end;
+    ++end;
+    uninitializedIndex = outIndex + 1;
+    outIndex = nextIndex;
+    if (end > numIndices) {
+      break;
+    }
+  }
+  if (uninitializedIndex < outputRows) {
+    output.fill(defaultValue, uninitializedIndex * numCol, outputRows * numCol);
+  }
+  return [output, outputShape];
+}
+var sqrtImpl = createSimpleUnaryImpl((xi) => Math.sqrt(xi));
+var sqrt2 = unaryKernelFunc(Sqrt, (xi) => Math.sqrt(xi));
+var sqrtConfig = {
+  kernelName: Sqrt,
+  backendName: "cpu",
+  kernelFunc: sqrt2
+};
+var squaredDifferenceImpl = createSimpleBinaryKernelImpl((a, b) => {
+  const diff = a - b;
+  return diff * diff;
+});
+var squaredDifference2 = binaryKernelFunc(SquaredDifference, squaredDifferenceImpl);
+var squaredDifferenceConfig = {
+  kernelName: SquaredDifference,
+  backendName: "cpu",
+  kernelFunc: squaredDifference2
+};
+function stridedSliceImpl(outShape, xBuf, strides, begin) {
+  const outBuf = buffer(outShape, xBuf.dtype);
+  for (let i = 0; i < outBuf.size; i++) {
+    const loc = outBuf.indexToLoc(i);
+    const newLoc = new Array(loc.length);
+    for (let j = 0; j < newLoc.length; j++) {
+      newLoc[j] = loc[j] * strides[j] + begin[j];
+    }
+    outBuf.set(xBuf.get(...newLoc), ...loc);
+  }
+  return outBuf;
+}
+var StringNGramsOp = class {
+  constructor(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+    this.separator = util_exports.encodeString(separator);
+    this.nGramWidths = nGramWidths;
+    this.leftPad = util_exports.encodeString(leftPad);
+    this.rightPad = util_exports.encodeString(rightPad2);
+    this.padWidth = padWidth;
+    this.preserveShort = preserveShortSequences;
+  }
+  getPadWidth(nGramWidth) {
+    return Math.min(this.padWidth < 0 ? nGramWidth - 1 : this.padWidth, nGramWidth - 1);
+  }
+  getNumNGrams(length, nGramWidth) {
+    const padWidth = this.getPadWidth(nGramWidth);
+    return Math.max(0, length + 2 * padWidth - nGramWidth + 1);
+  }
+  createNGrams(data, splitIndex, output, outputStartIndex, numNGrams, nGramWidth) {
+    for (let nGramIndex = 0; nGramIndex < numNGrams; ++nGramIndex) {
+      const padWidth = this.getPadWidth(nGramWidth);
+      const leftPadding = Math.max(0, padWidth - nGramIndex);
+      const rightPadding = Math.max(0, padWidth - (numNGrams - (nGramIndex + 1)));
+      const numTokens = nGramWidth - (leftPadding + rightPadding);
+      const dataStartIndex = splitIndex + (leftPadding > 0 ? 0 : nGramIndex - padWidth);
+      let nGramSize = 0;
+      nGramSize += leftPadding * this.leftPad.length;
+      for (let n = 0; n < numTokens; ++n) {
+        nGramSize += data[dataStartIndex + n].length;
+      }
+      nGramSize += rightPadding * this.rightPad.length;
+      const numSeparators = leftPadding + rightPadding + numTokens - 1;
+      nGramSize += numSeparators * this.separator.length;
+      output[outputStartIndex + nGramIndex] = new Uint8Array(nGramSize);
+      const nGram = output[outputStartIndex + nGramIndex];
+      let nextNGramIndex = 0;
+      const appendToNGram = (str) => str.forEach((value) => nGram[nextNGramIndex++] = value);
+      for (let n = 0; n < leftPadding; ++n) {
+        appendToNGram(this.leftPad);
+        appendToNGram(this.separator);
+      }
+      for (let n = 0; n < numTokens - 1; ++n) {
+        appendToNGram(data[dataStartIndex + n]);
+        appendToNGram(this.separator);
+      }
+      if (numTokens > 0) {
+        appendToNGram(data[dataStartIndex + numTokens - 1]);
+        for (let n = 0; n < rightPadding; ++n) {
+          appendToNGram(this.separator);
+          appendToNGram(this.rightPad);
+        }
+      } else {
+        for (let n = 0; n < rightPadding - 1; ++n) {
+          appendToNGram(this.rightPad);
+          appendToNGram(this.separator);
+        }
+        appendToNGram(this.rightPad);
+      }
+    }
+  }
+  compute(data, splits) {
+    const inputDataSize = data.length;
+    const splitsSize = splits.length;
+    if (splitsSize > 0) {
+      let prevSplit = splits[0];
+      if (prevSplit !== 0) {
+        throw new Error(`First split value must be 0, got ${prevSplit}`);
+      }
+      for (let i = 1; i < splitsSize; ++i) {
+        let validSplits = splits[i] >= prevSplit;
+        validSplits = validSplits && splits[i] <= inputDataSize;
+        if (!validSplits) {
+          throw new Error(`Invalid split value ${splits[i]}, must be in [${prevSplit}, ${inputDataSize}]`);
+        }
+        prevSplit = splits[i];
+      }
+      if (prevSplit !== inputDataSize) {
+        throw new Error(`Last split value must be data size. Expected ${inputDataSize}, got ${prevSplit}`);
+      }
+    }
+    const numBatchItems = splitsSize - 1;
+    const nGramsSplits = util_exports.getArrayFromDType("int32", splitsSize);
+    if (inputDataSize === 0 || splitsSize === 0) {
+      const empty = new Array(inputDataSize);
+      for (let i = 0; i <= numBatchItems; ++i) {
+        nGramsSplits[i] = 0;
+      }
+      return [empty, nGramsSplits];
+    }
+    nGramsSplits[0] = 0;
+    for (let i = 1; i <= numBatchItems; ++i) {
+      const length = splits[i] - splits[i - 1];
+      let numNGrams = 0;
+      this.nGramWidths.forEach((nGramWidth) => {
+        numNGrams += this.getNumNGrams(length, nGramWidth);
+      });
+      if (this.preserveShort && length > 0 && numNGrams === 0) {
+        numNGrams = 1;
+      }
+      nGramsSplits[i] = nGramsSplits[i - 1] + numNGrams;
+    }
+    const nGrams = new Array(nGramsSplits[numBatchItems]);
+    for (let i = 0; i < numBatchItems; ++i) {
+      const splitIndex = splits[i];
+      let outputStartIdx = nGramsSplits[i];
+      this.nGramWidths.forEach((nGramWidth) => {
+        const length = splits[i + 1] - splits[i];
+        const numNGrams = this.getNumNGrams(length, nGramWidth);
+        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);
+        outputStartIdx += numNGrams;
+      });
+      if (this.preserveShort && outputStartIdx === nGramsSplits[i]) {
+        const dataLength = splits[i + 1] - splits[i];
+        if (dataLength === 0) {
+          continue;
+        }
+        const nGramWidth = dataLength + 2 * this.padWidth;
+        const numNGrams = 1;
+        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);
+      }
+    }
+    return [nGrams, nGramsSplits];
+  }
+};
+function stringNGramsImpl(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+  return new StringNGramsOp(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences).compute(data, dataSplits);
+}
+function split3(str, delimiters, skipEmpty, result) {
+  if (!str.length) {
+    return;
+  }
+  if (delimiters.length === 0) {
+    for (let i = 0; i < str.length; ++i) {
+      result.push(str.subarray(i, i + 1));
+    }
+    return;
+  }
+  if (delimiters.length === 1) {
+    const delimiter = delimiters[0];
+    let f = str.indexOf(delimiter);
+    while (f !== -1) {
+      const token = str.subarray(0, f);
+      if (!skipEmpty || token.length !== 0) {
+        result.push(token);
+      }
+      str = str.subarray(f + 1);
+      f = str.indexOf(delimiter);
+    }
+    if (!skipEmpty || str.length !== 0) {
+      result.push(str);
+    }
+    return;
+  }
+  let tokenStart = 0;
+  for (let i = 0; i < str.length + 1; i++) {
+    if (i === str.length || delimiters.indexOf(str[i]) !== -1) {
+      const token = str.subarray(tokenStart, i);
+      if (!skipEmpty || token.length !== 0) {
+        result.push(token);
+      }
+      tokenStart = i + 1;
+    }
+  }
+}
+function stringSplitImpl(input2, delimiter, skipEmpty) {
+  const batchSize = input2.length;
+  const tokens = [];
+  let outputSize = 0;
+  let maxNumEntries = 0;
+  const numIndices = new Array(batchSize);
+  for (let i = 0; i < batchSize; ++i) {
+    const prevTokensLength = tokens.length;
+    split3(input2[i], delimiter, skipEmpty, tokens);
+    const nEntries = tokens.length - prevTokensLength;
+    numIndices[i] = nEntries;
+    outputSize += nEntries;
+    maxNumEntries = Math.max(maxNumEntries, nEntries);
+  }
+  const indices = util_exports.getArrayFromDType("int32", outputSize * 2);
+  const values = new Array(outputSize);
+  const shape = [batchSize, maxNumEntries];
+  let c = 0;
+  for (let i = 0; i < batchSize; ++i) {
+    for (let j = 0; j < numIndices[i]; ++j) {
+      indices[c * 2] = i;
+      indices[c * 2 + 1] = j;
+      values[c] = tokens[c];
+      ++c;
+    }
+  }
+  return [indices, values, shape];
+}
+function stringToHashBucketFastImpl(input2, numBuckets) {
+  const output = util_exports.getArrayFromDType("int32", input2.length);
+  for (let i = 0; i < input2.length; ++i) {
+    output[i] = util_exports.fingerPrint64(input2[i]).modulo(numBuckets).getLowBitsUnsigned();
+  }
+  return output;
+}
+var subImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue - bValue);
+var subComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return { real: aReal - bReal, imag: aImag - bImag };
+});
+var sub2 = binaryKernelFunc(Sub, subImpl, subComplexImpl);
+var subConfig = {
+  kernelName: Sub,
+  backendName: "cpu",
+  kernelFunc: sub2
+};
+function tileImpl(xBuf, reps) {
+  const newShape = new Array(xBuf.rank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = xBuf.shape[i] * reps[i];
+  }
+  const result = buffer(newShape, xBuf.dtype);
+  for (let i = 0; i < result.values.length; ++i) {
+    const newLoc = result.indexToLoc(i);
+    const originalLoc = new Array(xBuf.rank);
+    for (let j = 0; j < originalLoc.length; j++) {
+      originalLoc[j] = newLoc[j] % xBuf.shape[j];
+    }
+    const originalIndex = xBuf.locToIndex(originalLoc);
+    result.values[i] = xBuf.values[originalIndex];
+  }
+  return result;
+}
+var comparePair = (a, b) => {
+  const valueDiff = b.value - a.value;
+  return valueDiff === 0 ? a.index - b.index : valueDiff;
+};
+function select(array2, k, left = 0, right = array2.length - 1) {
+  while (right > left) {
+    if (right - left > 600) {
+      const n = right - left + 1;
+      const i2 = k - left + 1;
+      const z = Math.log(n);
+      const s = 0.5 * Math.exp(2 * z / 3);
+      const sd = 0.5 * Math.sqrt(z * s * (n - s) / n) * Math.sign(i2 - n / 2);
+      const newLeft = Math.max(left, Math.floor(k - i2 * s / n + sd));
+      const newRight = Math.min(right, Math.floor(k + (n - i2) * s / n + sd));
+      select(array2, k, newLeft, newRight);
+    }
+    const t = array2[k];
+    let i = left;
+    let j = right;
+    util_exports.swap(array2, left, k);
+    if (comparePair(array2[right], t) > 0) {
+      util_exports.swap(array2, left, right);
+    }
+    while (i < j) {
+      util_exports.swap(array2, i, j);
+      i++;
+      j--;
+      while (comparePair(array2[i], t) < 0) {
+        i = i + 1;
+      }
+      while (comparePair(array2[j], t) > 0) {
+        j = j - 1;
+      }
+    }
+    if (comparePair(array2[left], t) === 0) {
+      util_exports.swap(array2, left, j);
+    } else {
+      j = j + 1;
+      util_exports.swap(array2, j, right);
+    }
+    if (j <= k) {
+      left = j + 1;
+    }
+    if (k <= j) {
+      right = j - 1;
+    }
+  }
+}
+function topKImpl(x, xShape, xDtype, k, sorted) {
+  const lastDim = xShape[xShape.length - 1];
+  const [batch, size] = [x.length / lastDim, lastDim];
+  const allTopKVals = util_exports.getTypedArrayFromDType(xDtype, batch * k);
+  const allTopKIndices = util_exports.getTypedArrayFromDType("int32", batch * k);
+  for (let b = 0; b < batch; b++) {
+    const offset = b * size;
+    const vals = x.subarray(offset, offset + size);
+    let valAndInd = new Array(vals.length);
+    vals.forEach((value, index) => valAndInd[index] = { value, index });
+    if (k < valAndInd.length) {
+      select(valAndInd, k);
+      valAndInd = valAndInd.slice(0, k);
+    }
+    if (sorted) {
+      valAndInd.sort(comparePair);
+    }
+    const outOffset = b * k;
+    const topKVals = allTopKVals.subarray(outOffset, outOffset + k);
+    const topKIndices = allTopKIndices.subarray(outOffset, outOffset + k);
+    for (let i = 0; i < k; i++) {
+      topKVals[i] = valAndInd[i].value;
+      topKIndices[i] = valAndInd[i].index;
+    }
+  }
+  const outputShape = xShape.slice();
+  outputShape[outputShape.length - 1] = k;
+  return [
+    buffer(outputShape, xDtype, allTopKVals),
+    buffer(outputShape, "int32", allTopKIndices)
+  ];
+}
+function uniqueImpl(values, axis, shape, dtype) {
+  const $axis = util_exports.parseAxisParam(axis, shape)[0];
+  const newShape = [1, shape[0], 1];
+  for (let i = 0; i < $axis; i++) {
+    newShape[0] *= shape[i];
+  }
+  newShape[1] = shape[$axis];
+  for (let i = $axis + 1; i < shape.length; i++) {
+    newShape[2] *= shape[i];
+  }
+  const uniqueElements = {};
+  const indices = new Int32Array(shape[$axis]);
+  const inputBuffer = new TensorBuffer(newShape, dtype, values);
+  const uniqueIndices = [];
+  const is1DTensor = newShape[0] === 1 && newShape[2] === 1;
+  for (let i = 0; i < shape[$axis]; i++) {
+    let element;
+    if (is1DTensor) {
+      element = values[i].toString();
+    } else {
+      const axisValues = [];
+      for (let m = 0; m < newShape[0]; m++) {
+        for (let n = 0; n < newShape[2]; n++) {
+          axisValues.push(inputBuffer.get(m, i, n));
+        }
+      }
+      element = axisValues.join(",");
+    }
+    if (uniqueElements[element] !== void 0) {
+      indices[i] = uniqueElements[element];
+    } else {
+      const uniqueIndex = Object.keys(uniqueElements).length;
+      uniqueElements[element] = uniqueIndex;
+      indices[i] = uniqueIndex;
+      uniqueIndices.push(i);
+    }
+  }
+  const outputTmpShape = newShape.slice();
+  outputTmpShape[1] = Object.keys(uniqueElements).length;
+  const outputBuffer = new TensorBuffer(outputTmpShape, dtype);
+  uniqueIndices.forEach((uniqueElementIndex, i) => {
+    for (let m = 0; m < newShape[0]; m++) {
+      for (let n = 0; n < newShape[2]; n++) {
+        outputBuffer.set(inputBuffer.get(m, uniqueElementIndex, n), m, i, n);
+      }
+    }
+  });
+  const outputShape = shape.slice();
+  outputShape[$axis] = outputTmpShape[1];
+  return {
+    outputValues: outputBuffer.values,
+    outputShape,
+    indices
+  };
+}
+registerBackend("cpu", () => new MathBackendCPU(), 1);
+var elu4 = unaryKernelFunc(Elu, (xi) => xi >= 0 ? xi : Math.exp(xi) - 1);
+var eluConfig = {
+  kernelName: Elu,
+  backendName: "cpu",
+  kernelFunc: elu4
+};
+function leakyRelu2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { alpha } = attrs;
+  assertNotComplex([x], "leakyRelu");
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xVals = backend2.data.get(x.dataId).values;
+  const outVals = util_exports.getTypedArrayFromDType("float32", xSize);
+  for (let i = 0; i < xVals.length; i++) {
+    outVals[i] = xVals[i] < 0 ? alpha * xVals[i] : xVals[i];
+  }
+  return backend2.makeTensorInfo(x.shape, "float32", outVals);
+}
+var leakyReluConfig = {
+  kernelName: LeakyRelu,
+  backendName: "cpu",
+  kernelFunc: leakyRelu2
+};
+var preluImpl = createSimpleBinaryKernelImpl((xValue, aValue) => xValue < 0 ? aValue * xValue : xValue);
+function prelu3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  assertNotComplex([x, alpha], "prelu");
+  const aVals = backend2.data.get(x.dataId).values;
+  const bVals = backend2.data.get(alpha.dataId).values;
+  const [resultData, resultShape] = preluImpl(x.shape, alpha.shape, aVals, bVals, "float32");
+  return backend2.makeTensorInfo(resultShape, "float32", resultData);
+}
+var preluConfig = {
+  kernelName: Prelu,
+  backendName: "cpu",
+  kernelFunc: prelu3
+};
+var relu2 = unaryKernelFunc(Relu, (xi) => Math.max(0, xi));
+var reluConfig = {
+  kernelName: Relu,
+  backendName: "cpu",
+  kernelFunc: relu2
+};
+var relu62 = unaryKernelFunc(Relu6, (xi) => Math.min(Math.max(0, xi), 6));
+var relu6Config = {
+  kernelName: Relu6,
+  backendName: "cpu",
+  kernelFunc: relu62
+};
+function applyActivation2(backend2, x, activation2, preluActivationWeights, leakyreluAlpha) {
+  if (activation2 === "linear") {
+    return identity2({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "relu") {
+    return relu2({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "elu") {
+    return elu4({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "relu6") {
+    return relu62({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "prelu") {
+    return prelu3({ inputs: { x, alpha: preluActivationWeights }, backend: backend2 });
+  } else if (activation2 === "leakyrelu") {
+    return leakyRelu2({ inputs: { x }, backend: backend2, attrs: { alpha: leakyreluAlpha } });
+  } else if (activation2 === "sigmoid") {
+    return sigmoid2({ inputs: { x }, backend: backend2 });
+  }
+  throw new Error(`Activation ${activation2} has not been implemented for the CPU backend.`);
+}
+function reshape3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  const $xSize = util_exports.sizeFromShape($shape);
+  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);
+  backend2.incRef(x.dataId);
+  const xData = backend2.data.get(x.dataId);
+  if (xData.complexTensorInfos != null) {
+    const real4 = xData.complexTensorInfos.real;
+    const imag4 = xData.complexTensorInfos.imag;
+    real4.shape = $shape;
+    imag4.shape = $shape;
+  }
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig = {
+  kernelName: Reshape,
+  backendName: "cpu",
+  kernelFunc: reshape3
+};
+function batchMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  assertNotComplex([a, b], "matMul");
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape3({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape3({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];
+  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];
+  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const a3dValues = backend2.data.get(a3d.dataId).values;
+  const b3dValues = backend2.data.get(b3d.dataId).values;
+  const a3dStrides = util_exports.computeStrides(a3d.shape);
+  const b3dStrides = util_exports.computeStrides(b3d.shape);
+  const [aBatch, aOuterStep, aInnerStep] = transposeA ? [a3dStrides[0], 1, a3dStrides[1]] : [a3dStrides[0], a3dStrides[1], 1];
+  const [bInnerStep, bOuterStep, bBatch] = transposeB ? [1, b3dStrides[1], b3dStrides[0]] : [b3dStrides[1], 1, b3dStrides[0]];
+  const size = leftDim * rightDim;
+  const result = buffer([batchDim, leftDim, rightDim], a3d.dtype);
+  const resVals = result.values;
+  const blockSize = backend2.blockSize;
+  for (let bi = 0; bi < batchDim; bi++) {
+    for (let i0 = 0; i0 < leftDim; i0 += blockSize) {
+      for (let j0 = 0; j0 < rightDim; j0 += blockSize) {
+        for (let k02 = 0; k02 < sharedDim; k02 += blockSize) {
+          const iBlock = Math.min(i0 + blockSize, leftDim);
+          const jBlock = Math.min(j0 + blockSize, rightDim);
+          const kBlock = Math.min(k02 + blockSize, sharedDim);
+          for (let i = i0; i < iBlock; i++) {
+            for (let j = j0; j < jBlock; j++) {
+              let sum6 = 0;
+              for (let k = k02; k < kBlock; k++) {
+                const batchOffsetA = Math.min(bi, batchDimA - 1) * aBatch;
+                const batchOffsetB = Math.min(bi, batchDimB - 1) * bBatch;
+                const aVal = a3dValues[batchOffsetA + i * aOuterStep + k * aInnerStep];
+                const bVal = b3dValues[k * bInnerStep + j * bOuterStep + batchOffsetB];
+                sum6 += aVal * bVal;
+              }
+              resVals[bi * size + (i * rightDim + j)] += sum6;
+            }
+          }
+        }
+      }
+    }
+  }
+  backend2.disposeIntermediateTensorInfo(a3d);
+  backend2.disposeIntermediateTensorInfo(b3d);
+  return backend2.makeTensorInfo(outShape, result.dtype, result.values);
+}
+var batchMatMulConfig = {
+  kernelName: BatchMatMul,
+  backendName: "cpu",
+  kernelFunc: batchMatMul
+};
+function _fusedMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  let current;
+  let addRes;
+  let activationRes;
+  const intermediates = [];
+  const matMulRes = batchMatMul({ inputs: { a, b }, attrs: { transposeA, transposeB }, backend: backend2 });
+  current = matMulRes;
+  if (bias) {
+    addRes = add4({ inputs: { a: current, b: bias }, backend: backend2 });
+    intermediates.push(current);
+    current = addRes;
+  }
+  if (activation2) {
+    activationRes = applyActivation2(backend2, current, activation2, preluActivationWeights, leakyreluAlpha);
+    intermediates.push(current);
+    current = activationRes;
+  }
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return current;
+}
+var _fusedMatMulConfig = {
+  kernelName: _FusedMatMul,
+  backendName: "cpu",
+  kernelFunc: _fusedMatMul
+};
+var acos2 = unaryKernelFunc(Acos, (xi) => Math.acos(xi));
+var acosConfig = {
+  kernelName: Acos,
+  backendName: "cpu",
+  kernelFunc: acos2
+};
+var acosh2 = unaryKernelFunc(Acosh, (xi) => Math.acosh(xi));
+var acoshConfig = {
+  kernelName: Acosh,
+  backendName: "cpu",
+  kernelFunc: acosh2
+};
+function addN2(args) {
+  const { inputs, backend: backend2 } = args;
+  const tensors = inputs;
+  assertNotComplex(inputs, "addN");
+  const vals = tensors.map((t) => backend2.data.get(t.dataId).values);
+  const outBuf = buffer(tensors[0].shape, tensors[0].dtype);
+  const outVals = outBuf.values;
+  for (let i = 0; i < tensors.length; i++) {
+    const currVals = vals[i];
+    for (let j = 0; j < outVals.length; j++) {
+      outVals[j] += currVals[j];
+    }
+  }
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var addNConfig = {
+  kernelName: AddN,
+  backendName: "cpu",
+  kernelFunc: addN2
+};
+function all2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "all");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let all5 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      all5 = all5 && value;
+    }
+    vals[i] = all5;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var allConfig = {
+  kernelName: All,
+  backendName: "cpu",
+  kernelFunc: all2
+};
+function any2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "any");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let anyVal = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      anyVal = anyVal || value;
+    }
+    vals[i] = anyVal;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var anyConfig = {
+  kernelName: Any,
+  backendName: "cpu",
+  kernelFunc: any2
+};
+function argMax2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  assertNotComplex(x, "argMax");
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  axes = [axes[0]];
+  backend_util_exports.assertAxesAreInnerMostDims("argMax", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const outSize = util_exports.sizeFromShape(outShape);
+  const vals = util_exports.makeZerosTypedArray(outSize, "int32");
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let max6 = aVals[offset];
+    let maxIndex = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (value > max6) {
+        max6 = value;
+        maxIndex = j;
+      }
+    }
+    vals[i] = maxIndex;
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(outShape, "int32", vals);
+}
+var argMaxConfig = {
+  kernelName: ArgMax,
+  backendName: "cpu",
+  kernelFunc: argMax2
+};
+function argMin2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  assertNotComplex(x, "argMin");
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  axes = [axes[0]];
+  backend_util_exports.assertAxesAreInnerMostDims("argMin", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const outSize = util_exports.sizeFromShape(outShape);
+  const vals = util_exports.makeZerosTypedArray(outSize, "int32");
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let min6 = aVals[offset];
+    let minIndex = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (value < min6) {
+        min6 = value;
+        minIndex = j;
+      }
+    }
+    vals[i] = minIndex;
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(outShape, "int32", vals);
+}
+var argMinConfig = {
+  kernelName: ArgMin,
+  backendName: "cpu",
+  kernelFunc: argMin2
+};
+var asin2 = unaryKernelFunc(Asin, (xi) => Math.asin(xi));
+var asinConfig = {
+  kernelName: Asin,
+  backendName: "cpu",
+  kernelFunc: asin2
+};
+var asinh2 = unaryKernelFunc(Asinh, (xi) => Math.asinh(xi));
+var asinhConfig = {
+  kernelName: Asinh,
+  backendName: "cpu",
+  kernelFunc: asinh2
+};
+var atan3 = unaryKernelFunc(Atan, (xi) => Math.atan(xi));
+var atanConfig = {
+  kernelName: Atan,
+  backendName: "cpu",
+  kernelFunc: atan3
+};
+var atan2Impl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.atan2(aValue, bValue));
+var atan22 = binaryKernelFunc(Atan2, atan2Impl);
+var atan2Config = {
+  kernelName: Atan2,
+  backendName: "cpu",
+  kernelFunc: atan22
+};
+var atanh2 = unaryKernelFunc(Atanh, (xi) => Math.atanh(xi));
+var atanhConfig = {
+  kernelName: Atanh,
+  backendName: "cpu",
+  kernelFunc: atanh2
+};
+function pool2(xValues, xShape, dtype, strides, convInfo, poolType) {
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const initialValue = poolType === "max" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;
+  const output = buffer(convInfo.outShape, dtype);
+  const outputVals = output.values;
+  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3];
+  const outputRowStrides = convInfo.outShape[2] * convInfo.outShape[3];
+  const outputColStrides = convInfo.outShape[3];
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const outputBatchOffset = b * outputBatchStrides;
+    const inputBatchOffset = b * strides[0];
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+        const xRCorner = yR * strideHeight - padTop;
+        const xRMin = Math.max(0, xRCorner);
+        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);
+        const outputRowOffset = outputBatchOffset + yR * outputRowStrides;
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const xCCorner = yC * strideWidth - padLeft;
+          const xCMin = Math.max(0, xCCorner);
+          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);
+          let minMaxValue = initialValue;
+          let avgValue = 0;
+          let count2 = 0;
+          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {
+            const xROffset = inputBatchOffset + xR * strides[1];
+            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {
+              const xCOffset = xROffset + xC * strides[2];
+              const pixel = xValues[xCOffset + d];
+              if (poolType === "max" && pixel > minMaxValue) {
+                minMaxValue = pixel;
+              } else if (poolType === "avg") {
+                avgValue += pixel;
+                count2++;
+              }
+            }
+            if (isNaN(minMaxValue)) {
+              break;
+            }
+          }
+          const outputOffset = outputRowOffset + yC * outputColStrides + d;
+          outputVals[outputOffset] = poolType === "avg" ? avgValue / count2 : minMaxValue;
+        }
+      }
+    }
+  }
+  return output;
+}
+function maxPoolPositions(xValues, xShape, dtype, convInfo, flattenPositions = false, includeBatchInIndex = false) {
+  const maxPositions = buffer(convInfo.outShape, "int32");
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const xBuf = buffer(xShape, dtype, xValues);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+        const xRCorner = yR * strideHeight - padTop;
+        let xRMin = xRCorner;
+        while (xRMin < 0) {
+          xRMin += dilationHeight;
+        }
+        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const xCCorner = yC * strideWidth - padLeft;
+          let xCMin = xCCorner;
+          while (xCMin < 0) {
+            xCMin += dilationWidth;
+          }
+          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);
+          let maxValue = Number.NEGATIVE_INFINITY;
+          let maxPosition = -1;
+          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {
+            const wR = xR - xRCorner;
+            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {
+              const wC = xC - xCCorner;
+              const pixel = xBuf.get(b, xR, xC, d);
+              if (pixel > maxValue) {
+                maxValue = pixel;
+                if (flattenPositions) {
+                  maxPosition = includeBatchInIndex ? ((b * convInfo.inHeight + xR) * convInfo.inWidth + xC) * convInfo.inChannels + d : (xR * convInfo.inWidth + xC) * convInfo.inChannels + d;
+                } else {
+                  maxPosition = wR * effectiveFilterWidth + wC;
+                }
+              }
+            }
+          }
+          maxPositions.set(maxPosition, b, yR, yC, d);
+        }
+      }
+    }
+  }
+  return maxPositions;
+}
+function pool3d2(xValues, xShape, dtype, strides, convInfo, poolType) {
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = convInfo.padInfo.front;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const initialValue = poolType === "max" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;
+  const output = buffer(convInfo.outShape, dtype);
+  const outputVals = output.values;
+  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];
+  const outputDepthStrides = convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];
+  const outputRowStrides = convInfo.outShape[3] * convInfo.outShape[4];
+  const outputColStrides = convInfo.outShape[4];
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    const outputBatchOffset = batch * outputBatchStrides;
+    const inputBatchOffset = batch * strides[0];
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {
+        const xDepthCorner = yDepth * strideDepth - padFront;
+        let xDepthMin = xDepthCorner;
+        while (xDepthMin < 0) {
+          xDepthMin += dilationDepth;
+        }
+        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);
+        const outputDepthOffset = outputBatchOffset + yDepth * outputDepthStrides;
+        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {
+          const xRowCorner = yRow * strideHeight - padTop;
+          let xRowMin = xRowCorner;
+          while (xRowMin < 0) {
+            xRowMin += dilationHeight;
+          }
+          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);
+          const outputRowOffset = outputDepthOffset + yRow * outputRowStrides;
+          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {
+            const xColCorner = yCol * strideWidth - padLeft;
+            let xColMin = xColCorner;
+            while (xColMin < 0) {
+              xColMin += dilationWidth;
+            }
+            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);
+            const outputColOffset = outputRowOffset + yCol * outputColStrides;
+            let minMaxValue = initialValue;
+            let avgValue = 0;
+            let count2 = 0;
+            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {
+              const xDepthOffset = inputBatchOffset + xDepth * strides[1];
+              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {
+                const xRowOffset = xDepthOffset + xRow * strides[2];
+                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {
+                  const xColOffset = xRowOffset + xCol * strides[3];
+                  const pixel = xValues[xColOffset + channel];
+                  if (poolType === "max" && pixel > minMaxValue) {
+                    minMaxValue = pixel;
+                  } else if (poolType === "avg") {
+                    avgValue += pixel;
+                    count2++;
+                  }
+                  if (isNaN(minMaxValue)) {
+                    break;
+                  }
+                }
+                if (isNaN(minMaxValue)) {
+                  break;
+                }
+              }
+              if (isNaN(minMaxValue)) {
+                break;
+              }
+            }
+            const outputOffset = outputColOffset + channel;
+            outputVals[outputOffset] = poolType === "avg" ? avgValue / count2 : minMaxValue;
+          }
+        }
+      }
+    }
+  }
+  return output;
+}
+function maxPool3dPositions(xBuf, convInfo) {
+  const maxPositions = buffer(convInfo.outShape, "int32");
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = convInfo.padInfo.front;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {
+        const xDepthCorner = yDepth * strideDepth - padFront;
+        let xDepthMin = xDepthCorner;
+        while (xDepthMin < 0) {
+          xDepthMin += dilationDepth;
+        }
+        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);
+        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {
+          const xRowCorner = yRow * strideHeight - padTop;
+          let xRowMin = xRowCorner;
+          while (xRowMin < 0) {
+            xRowMin += dilationHeight;
+          }
+          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);
+          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {
+            const xColCorner = yCol * strideWidth - padLeft;
+            let xColMin = xColCorner;
+            while (xColMin < 0) {
+              xColMin += dilationWidth;
+            }
+            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);
+            let maxValue = Number.NEGATIVE_INFINITY;
+            let maxPosition = -1;
+            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {
+              const wDepth = xDepth - xDepthCorner;
+              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {
+                const wRow = xRow - xRowCorner;
+                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {
+                  const wCol = xCol - xColCorner;
+                  const pixel = xBuf.get(batch, xDepth, xRow, xCol, channel);
+                  if (pixel >= maxValue) {
+                    maxValue = pixel;
+                    maxPosition = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterHeight + wCol;
+                  }
+                }
+              }
+            }
+            maxPositions.set(maxPosition, batch, yDepth, yRow, yCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return maxPositions;
+}
+function avgPool2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "avgPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  let res;
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    res = identity2({ inputs: { x }, backend: backend2 });
+  } else {
+    const xValues = backend2.data.get(x.dataId).values;
+    const strides2 = util_exports.computeStrides(x.shape);
+    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, "avg");
+    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);
+  }
+  return res;
+}
+var avgPoolConfig = {
+  kernelName: AvgPool,
+  backendName: "cpu",
+  kernelFunc: avgPool2
+};
+function avgPool3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  assertNotComplex(x, "avgPool3d");
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);
+  const xValues = backend2.data.get(x.dataId).values;
+  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, "avg");
+  return backend2.makeTensorInfo(outBuf.shape, "float32", outBuf.values);
+}
+var avgPool3DConfig = {
+  kernelName: AvgPool3D,
+  backendName: "cpu",
+  kernelFunc: avgPool3D
+};
+function avgPool3DGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  assertNotComplex([dy, input2], "avgPool3DGrad");
+  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterDepth = convInfo.filterDepth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(input2.shape, "float32");
+  const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);
+  const dyBuf = backend2.bufferSync(dy);
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {
+        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {
+          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {
+            const dyDepthCorner = dxDepth - padFront;
+            const dyRowCorner = dxRow - padTop;
+            const dyColCorner = dxCol - padLeft;
+            let dotProd = 0;
+            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {
+              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;
+              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {
+                continue;
+              }
+              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {
+                const dyRow = (dyRowCorner + wRow) / strideHeight;
+                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {
+                  continue;
+                }
+                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {
+                  const dyCol = (dyColCorner + wCol) / strideWidth;
+                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {
+                    continue;
+                  }
+                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  dotProd += pixel;
+                }
+              }
+            }
+            dx.set(dotProd * avgMultiplier, batch, dxDepth, dxRow, dxCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var avgPool3DGradConfig2 = {
+  kernelName: AvgPool3DGrad,
+  backendName: "cpu",
+  kernelFunc: avgPool3DGrad
+};
+function avgPoolGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  assertNotComplex([dy, input2], "avgPoolGrad");
+  const { filterSize, strides, pad: pad3 } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(x.shape, "float32");
+  const avgMultiplier = 1 / (filterHeight * filterWidth);
+  const dyData = backend2.data.get(dy.dataId).values;
+  const dyBuf = buffer(dy.shape, "float32", dyData);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {
+        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {
+          const dyRCorner = dxR - padTop;
+          const dyCCorner = dxC - padLeft;
+          let dotProd = 0;
+          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {
+            const dyR = (dyRCorner + wR) / strideHeight;
+            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {
+              continue;
+            }
+            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {
+              const dyC = (dyCCorner + wC) / strideWidth;
+              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {
+                continue;
+              }
+              const pixel = dyBuf.get(b, dyR, dyC, d);
+              dotProd += pixel;
+            }
+          }
+          dx.set(dotProd * avgMultiplier, b, dxR, dxC, d);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var avgPoolGradConfig2 = {
+  kernelName: AvgPoolGrad,
+  backendName: "cpu",
+  kernelFunc: avgPoolGrad2
+};
+function batchNorm2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, scale: scale22, offset, mean: mean4, variance } = inputs;
+  util_exports.assert(mean4.shape.length === variance.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  util_exports.assert(scale22 == null || mean4.shape.length === scale22.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  assertNotComplex([x, mean4, variance, scale22, offset], "batchNorm");
+  let { varianceEpsilon } = attrs;
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const xVals = backend2.data.get(x.dataId).values;
+  const mVals = backend2.data.get(mean4.dataId).values;
+  const varVals = backend2.data.get(variance.dataId).values;
+  const sVals = scale22 ? backend2.data.get(scale22.dataId).values : new Float32Array([1]);
+  const offVals = offset ? backend2.data.get(offset.dataId).values : new Float32Array([0]);
+  const outVals = new Float32Array(xVals.length);
+  const offValsLength = offVals.length;
+  const sValsLength = sVals.length;
+  const varValsLength = varVals.length;
+  const mValsLength = mVals.length;
+  let offi = 0;
+  let mi = 0;
+  let si = 0;
+  let vi = 0;
+  for (let i = 0; i < xVals.length; ++i) {
+    outVals[i] = offVals[offi++] + (xVals[i] - mVals[mi++]) * sVals[si++] / Math.sqrt(varVals[vi++] + varianceEpsilon);
+    if (offi >= offValsLength) {
+      offi = 0;
+    }
+    if (mi >= mValsLength) {
+      mi = 0;
+    }
+    if (si >= sValsLength) {
+      si = 0;
+    }
+    if (vi >= varValsLength) {
+      vi = 0;
+    }
+  }
+  return backend2.makeTensorInfo(x.shape, x.dtype, outVals);
+}
+var batchNormConfig = {
+  kernelName: FusedBatchNorm,
+  backendName: "cpu",
+  kernelFunc: batchNorm2
+};
+function batchToSpaceND2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  assertNotComplex([x], "batchToSpaceND");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const xReshaped = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const xTransposed = transpose2({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });
+  const xTransposedReshaped = reshape3({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });
+  const result = slice2({
+    inputs: { x: xTransposedReshaped },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  backend2.disposeIntermediateTensorInfo(xReshaped);
+  backend2.disposeIntermediateTensorInfo(xTransposed);
+  backend2.disposeIntermediateTensorInfo(xTransposedReshaped);
+  return result;
+}
+var batchToSpaceNDConfig = {
+  kernelName: BatchToSpaceND,
+  backendName: "cpu",
+  kernelFunc: batchToSpaceND2
+};
+function bincount2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size } = attrs;
+  const xVals = backend2.data.get(x.dataId).values;
+  const weightsVals = backend2.data.get(weights.dataId).values;
+  const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);
+  return backend2.makeTensorInfo([size], weights.dtype, outVals);
+}
+var bincountConfig = {
+  kernelName: Bincount,
+  backendName: "cpu",
+  kernelFunc: bincount2
+};
+function broadcastArgs2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { s0, s1 } = inputs;
+  const s0Vals = backend2.data.get(s0.dataId).values;
+  const s1Vals = backend2.data.get(s1.dataId).values;
+  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));
+  return backend2.makeTensorInfo([broadcastShape.length], "int32", Int32Array.from(broadcastShape));
+}
+var broadcastArgsConfig = {
+  kernelName: BroadcastArgs,
+  backendName: "cpu",
+  kernelFunc: broadcastArgs2
+};
+var clipByValue2 = unaryKernelFunc(ClipByValue, (xi, attrs) => {
+  const clipAttrs = attrs;
+  if (xi > clipAttrs.clipValueMax) {
+    return clipAttrs.clipValueMax;
+  }
+  return xi < clipAttrs.clipValueMin ? clipAttrs.clipValueMin : xi;
+});
+var clipByValueConfig = {
+  kernelName: ClipByValue,
+  backendName: "cpu",
+  kernelFunc: clipByValue2
+};
+var complexAbs = (args) => {
+  const { x } = args.inputs;
+  const cpuBackend = args.backend;
+  const resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));
+  const complexVals = cpuBackend.data.get(x.dataId);
+  const real4 = complexVals.complexTensorInfos.real;
+  const imag4 = complexVals.complexTensorInfos.imag;
+  const realVals = cpuBackend.data.get(real4.dataId).values;
+  const imagVals = cpuBackend.data.get(imag4.dataId).values;
+  for (let i = 0; i < realVals.length; i++) {
+    const real5 = realVals[i];
+    const imag5 = imagVals[i];
+    resultValues[i] = Math.hypot(real5, imag5);
+  }
+  return cpuBackend.makeOutput(resultValues, x.shape, "float32");
+};
+var complexAbsConfig = {
+  kernelName: ComplexAbs,
+  backendName: "cpu",
+  kernelFunc: complexAbs
+};
+function imag2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const imag4 = backend2.data.get(input2.dataId).complexTensorInfos.imag;
+  const imagVal = backend2.data.get(imag4.dataId).values;
+  return backend2.makeTensorInfo(imag4.shape, imag4.dtype, imagVal);
+}
+var imagConfig = {
+  kernelName: Imag,
+  backendName: "cpu",
+  kernelFunc: imag2
+};
+function concat2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, $axis);
+  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);
+  }
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity2({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  if ($inputs[0].dtype === "complex64") {
+    const reals = $inputs.map((t) => real2({ inputs: { input: t }, backend: backend2 }));
+    const imags = $inputs.map((t) => imag2({ inputs: { input: t }, backend: backend2 }));
+    const realConcated = concat2({ inputs: reals, backend: backend2, attrs: { axis: $axis } });
+    const imagConcated = concat2({ inputs: imags, backend: backend2, attrs: { axis: $axis } });
+    const result = complex2({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });
+    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));
+    backend2.disposeIntermediateTensorInfo(realConcated);
+    backend2.disposeIntermediateTensorInfo(imagConcated);
+    return result;
+  }
+  const inputs2D = $inputs.map((t) => {
+    const innerSize = util_exports.sizeFromShape(t.shape.slice($axis));
+    const shape = [-1, innerSize];
+    return reshape3({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+  });
+  const inputsValShapes = inputs2D.map((t) => {
+    return { vals: backend2.data.get(t.dataId).values, shape: t.shape };
+  });
+  outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);
+  const simplyConcat = inputs2D[0].shape[0] === 1;
+  const outVals = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);
+  const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), $axis);
+  const outInfo = backend2.makeTensorInfo(finalOutShape, inputs[0].dtype, outVals);
+  inputs2D.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return outInfo;
+}
+var concatConfig = {
+  kernelName: Concat,
+  backendName: "cpu",
+  kernelFunc: concat2
+};
+function conv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;
+  assertNotComplex([x, filter], "conv2d");
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const padLeft = convInfo.padInfo.left;
+  const padTop = convInfo.padInfo.top;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const xBatchStride = xStrides[0];
+  const xRowStride = isChannelsLast ? xStrides[1] : xStrides[2];
+  const xColStride = isChannelsLast ? xStrides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : xStrides[1];
+  const yBatchStride = y.strides[0];
+  const yRowStride = isChannelsLast ? y.strides[1] : y.strides[2];
+  const yColStride = isChannelsLast ? y.strides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : y.strides[1];
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xBatchStride;
+    const yOffset1 = b * yBatchStride;
+    for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+      const yOffset2 = yOffset1 + yR * yRowStride;
+      const xRCorner = yR * convInfo.strideHeight - padTop;
+      for (let wR = 0; wR < filterHeight; ++wR) {
+        const xR = xRCorner + wR * dilationHeight;
+        if (xR < 0 || xR >= convInfo.inHeight) {
+          continue;
+        }
+        const wOffset1 = wR * filterStrides[0];
+        const xOffset2 = xOffset1 + xR * xRowStride;
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const yOffset3 = yOffset2 + yC * yColStride;
+          const xCCorner = yC * convInfo.strideWidth - padLeft;
+          for (let wC = 0; wC < filterWidth; ++wC) {
+            const xC = xCCorner + wC * dilationWidth;
+            if (xC < 0 || xC >= convInfo.inWidth) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wC * filterStrides[1];
+            const xOffset3 = xOffset2 + xC * xColStride;
+            let wOffset3 = wOffset2;
+            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+              const xVal = xVals[xOffset3 + d1 * xChannelStride];
+              for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+                yVals[yOffset3 + d2 * yChannelStride] += xVal * wVals[wOffset3 + d2];
+              }
+              wOffset3 += convInfo.outChannels;
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, yVals);
+}
+var conv2DConfig = {
+  kernelName: Conv2D,
+  backendName: "cpu",
+  kernelFunc: conv2D
+};
+function conv2DBackpropFilter2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;
+  assertNotComplex([x, dy], "conv2dBackpropFilter");
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const dW = new TensorBuffer(convInfo.filterShape, "float32");
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  const xVals = backend2.data.get(x.dataId).values;
+  const dyVals = backend2.data.get(dy.dataId).values;
+  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);
+  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);
+  for (let wR = 0; wR < filterHeight; ++wR) {
+    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+    for (let wC = 0; wC < filterWidth; ++wC) {
+      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+      for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+        for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+          let dotProd = 0;
+          for (let b = 0; b < convInfo.batchSize; ++b) {
+            for (let yR = yRMin; yR < yRMax; ++yR) {
+              const xR = wR + yR * strideHeight - topPad;
+              for (let yC = yCMin; yC < yCMax; ++yC) {
+                const xC = wC + yC * strideWidth - leftPad;
+                if (isChannelsLast) {
+                  dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);
+                } else {
+                  dotProd += xBuf.get(b, d1, xR, xC) * dyBuf.get(b, d2, yR, yC);
+                }
+              }
+            }
+          }
+          dW.set(dotProd, wR, wC, d1, d2);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);
+}
+var conv2DBackpropFilterConfig = {
+  kernelName: Conv2DBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: conv2DBackpropFilter2
+};
+function conv2DBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  assertNotComplex([dy, filter], "conv2dBackpropInput");
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  let $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2] = filterStrides;
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  $dataFormat = convInfo.dataFormat;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const isChannelsLast = $dataFormat === "channelsLast";
+  const xBatchStride = dx.strides[0];
+  const xRowStride = isChannelsLast ? dx.strides[1] : dx.strides[2];
+  const xColStride = isChannelsLast ? dx.strides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : dx.strides[1];
+  const yBatchStride = dyStrides[0];
+  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];
+  const yColStride = isChannelsLast ? dyStrides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xR = 0; xR < inHeight; ++xR) {
+        const xRCorner = xR - topPad;
+        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+        for (let xC = 0; xC < inWidth; ++xC) {
+          const xCCorner = xC - leftPad;
+          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+          let dotProd = 0;
+          for (let yR = xRMin; yR < yRMax; ++yR) {
+            const wR = yR * strideHeight - xRCorner;
+            for (let yC = xCMin; yC < yCMax; ++yC) {
+              const wC = yC * strideWidth - xCCorner;
+              const dyOffset = yBatchStride * b + yRowStride * yR + yColStride * yC;
+              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;
+              for (let d2 = 0; d2 < outChannels; ++d2) {
+                const pixel = dyValues[dyOffset + yChannelStride * d2];
+                const weight = fltValues[fltOffset + d2];
+                dotProd += pixel * weight;
+              }
+            }
+          }
+          const dxOffset = xBatchStride * b + xRowStride * xR + xColStride * xC + xChannelStride * d1;
+          dxValues[dxOffset] = dotProd;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var conv2DBackpropInputConfig = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "cpu",
+  kernelFunc: conv2DBackpropInput2
+};
+function conv3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  assertNotComplex([x, filter], "conv3d");
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);
+  const { filterDepth, filterHeight, filterWidth, dilationDepth, dilationHeight, dilationWidth, padInfo } = convInfo;
+  const padFront = padInfo.front;
+  const padLeft = padInfo.left;
+  const padTop = padInfo.top;
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xStrides[0];
+    const yOffset1 = b * y.strides[0];
+    for (let yF = 0; yF < convInfo.outDepth; ++yF) {
+      const yOffset2 = yOffset1 + yF * y.strides[1];
+      const xFCorner = yF * convInfo.strideDepth - padFront;
+      for (let wF = 0; wF < filterDepth; ++wF) {
+        const xF = xFCorner + wF * dilationDepth;
+        if (xF < 0 || xF >= convInfo.inDepth) {
+          continue;
+        }
+        const wOffset1 = wF * filterStrides[0];
+        const xOffset2 = xOffset1 + xF * xStrides[1];
+        for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+          const yOffset3 = yOffset2 + yR * y.strides[2];
+          const xRCorner = yR * convInfo.strideHeight - padTop;
+          for (let wR = 0; wR < filterHeight; ++wR) {
+            const xR = xRCorner + wR * dilationHeight;
+            if (xR < 0 || xR >= convInfo.inHeight) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wR * filterStrides[1];
+            const xOffset3 = xOffset2 + xR * xStrides[2];
+            for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+              const yOffset4 = yOffset3 + yC * convInfo.outChannels;
+              const xCCorner = yC * convInfo.strideWidth - padLeft;
+              for (let wC = 0; wC < filterWidth; ++wC) {
+                const xC = xCCorner + wC * dilationWidth;
+                if (xC < 0 || xC >= convInfo.inWidth) {
+                  continue;
+                }
+                const wOffset3 = wOffset2 + wC * filterStrides[2];
+                const xOffset4 = xOffset3 + xC * convInfo.inChannels;
+                let wOffset4 = wOffset3;
+                for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+                  const xVal = xVals[xOffset4 + d1];
+                  for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+                    yVals[yOffset4 + d2] += xVal * wVals[wOffset4 + d2];
+                  }
+                  wOffset4 += convInfo.outChannels;
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);
+}
+var conv3DConfig = {
+  kernelName: Conv3D,
+  backendName: "cpu",
+  kernelFunc: conv3D
+};
+function conv3DBackpropFilterV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, filterShape } = attrs;
+  assertNotComplex([x, dy], "conv3dBackpropFilterV2");
+  const xStrides = util_exports.computeStrides(x.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterDepth = convInfo.filterDepth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dw = new TensorBuffer(convInfo.filterShape, "float32");
+  const dwValues = dw.values;
+  const [dwS0, dwS1, dwS2, dwS3] = dw.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;
+  const xValues = backend2.data.get(x.dataId).values;
+  const [xS0, xS1, xS2, xS3] = xStrides;
+  const frontPad = convInfo.padInfo.front;
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  for (let wF = 0; wF < filterDepth; ++wF) {
+    const yFMin = Math.max(0, Math.ceil((frontPad - wF) / strideDepth));
+    const yFMax = Math.min(convInfo.outDepth, (convInfo.inDepth + frontPad - wF) / strideDepth);
+    const wOffset1 = wF * dwS0;
+    for (let wR = 0; wR < filterHeight; ++wR) {
+      const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+      const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+      const wOffset2 = wR * dwS1 + wOffset1;
+      for (let wC = 0; wC < filterWidth; ++wC) {
+        const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+        const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+        const wOffset3 = wC * dwS2 + wOffset2;
+        for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+          const wOffset4 = d1 * dwS3 + wOffset3;
+          for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+            let dotProd = 0;
+            for (let b = 0; b < convInfo.batchSize; ++b) {
+              const xOffset1 = b * xS0;
+              const yOffset1 = b * dyS0;
+              for (let yF = yFMin; yF < yFMax; ++yF) {
+                const xF = wF + yF * strideDepth - frontPad;
+                const xOffset2 = xF * xS1 + xOffset1;
+                const yOffset2 = yF * dyS1 + yOffset1;
+                for (let yR = yRMin; yR < yRMax; ++yR) {
+                  const xR = wR + yR * strideHeight - topPad;
+                  const xOffset3 = xR * xS2 + xOffset2;
+                  const yOffset3 = yR * dyS2 + yOffset2;
+                  for (let yC = yCMin; yC < yCMax; ++yC) {
+                    const xC = wC + yC * strideWidth - leftPad;
+                    const xOffset4 = xC * xS3 + xOffset3;
+                    const yOffset4 = yC * dyS3 + yOffset3;
+                    dotProd += xValues[xOffset4 + d1] * dyValues[yOffset4 + d2];
+                  }
+                }
+              }
+            }
+            dwValues[wOffset4 + d2] = dotProd;
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dw.shape, dw.dtype, dw.values);
+}
+var conv3DBackpropFilterV2Config = {
+  kernelName: Conv3DBackpropFilterV2,
+  backendName: "cpu",
+  kernelFunc: conv3DBackpropFilterV2
+};
+function conv3DBackpropInputV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { pad: pad3, strides, inputShape } = attrs;
+  assertNotComplex([dy], "conv3dBackpropInputV2");
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const [dxS0, dxS1, dxS2, dxS3] = dx.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2, fltS3] = filterStrides;
+  const { batchSize, filterDepth, filterHeight, filterWidth, inChannels, inDepth, inHeight, inWidth, outChannels, outDepth, outHeight, outWidth, strideDepth, strideHeight, strideWidth } = convInfo;
+  const frontPad = filterDepth - 1 - convInfo.padInfo.front;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xF = 0; xF < inDepth; ++xF) {
+        const xFCorner = xF - frontPad;
+        const xFMin = Math.max(0, Math.ceil(xFCorner / strideDepth));
+        const yFMax = Math.min(outDepth, (filterDepth + xFCorner) / strideDepth);
+        for (let xR = 0; xR < inHeight; ++xR) {
+          const xRCorner = xR - topPad;
+          const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+          const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+          for (let xC = 0; xC < inWidth; ++xC) {
+            const xCCorner = xC - leftPad;
+            const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+            const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+            let dotProd = 0;
+            for (let yF = xFMin; yF < yFMax; ++yF) {
+              const wF = yF * strideDepth - xFCorner;
+              for (let yR = xRMin; yR < yRMax; ++yR) {
+                const wR = yR * strideHeight - xRCorner;
+                for (let yC = xCMin; yC < yCMax; ++yC) {
+                  const wC = yC * strideWidth - xCCorner;
+                  const dyOffset = dyS0 * b + dyS1 * yF + dyS2 * yR + dyS3 * yC;
+                  const fltOffset = fltS0 * (filterDepth - 1 - wF) + fltS1 * (filterHeight - 1 - wR) + fltS2 * (filterWidth - 1 - wC) + fltS3 * d1;
+                  for (let d2 = 0; d2 < outChannels; ++d2) {
+                    const pixel = dyValues[dyOffset + d2];
+                    const weight = fltValues[fltOffset + d2];
+                    dotProd += pixel * weight;
+                  }
+                }
+              }
+            }
+            dxValues[dxS0 * b + dxS1 * xF + dxS2 * xR + dxS3 * xC + d1] = dotProd;
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var conv3DBackpropInputV2Config = {
+  kernelName: Conv3DBackpropInputV2,
+  backendName: "cpu",
+  kernelFunc: conv3DBackpropInputV2
+};
+var cos2 = unaryKernelFunc(Cos, (xi) => Math.cos(xi));
+var cosConfig = {
+  kernelName: Cos,
+  backendName: "cpu",
+  kernelFunc: cos2
+};
+var cosh2 = unaryKernelFunc(Cosh, (xi) => Math.cosh(xi));
+var coshConfig = {
+  kernelName: Cosh,
+  backendName: "cpu",
+  kernelFunc: cosh2
+};
+function cropAndResize2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, boxes, boxInd } = inputs;
+  const { cropSize, method, extrapolationValue } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const numBoxes = boxes.shape[0];
+  const [cropHeight, cropWidth] = cropSize;
+  const output = buffer([numBoxes, cropHeight, cropWidth, numChannels], "float32");
+  const boxVals = backend2.data.get(boxes.dataId).values;
+  const boxIndVals = backend2.data.get(boxInd.dataId).values;
+  const imageVals = backend2.data.get(image2.dataId).values;
+  const inStride = util_exports.computeStrides(image2.shape);
+  const outStride = util_exports.computeStrides(output.shape);
+  for (let b = 0; b < numBoxes; b++) {
+    const startInd = b * 4;
+    const y1 = boxVals[startInd];
+    const x1 = boxVals[startInd + 1];
+    const y2 = boxVals[startInd + 2];
+    const x2 = boxVals[startInd + 3];
+    const bInd = boxIndVals[b];
+    if (bInd >= batch) {
+      continue;
+    }
+    const heightScale = cropHeight > 1 ? (y2 - y1) * (imageHeight - 1) / (cropHeight - 1) : 0;
+    const widthScale = cropWidth > 1 ? (x2 - x1) * (imageWidth - 1) / (cropWidth - 1) : 0;
+    for (let y = 0; y < cropHeight; y++) {
+      const yInd = cropHeight > 1 ? y1 * (imageHeight - 1) + y * heightScale : 0.5 * (y1 + y2) * (imageHeight - 1);
+      if (yInd < 0 || yInd > imageHeight - 1) {
+        for (let x = 0; x < cropWidth; x++) {
+          for (let c = 0; c < numChannels; c++) {
+            const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[ind] = extrapolationValue;
+          }
+        }
+        continue;
+      }
+      if (method === "bilinear") {
+        const topInd = Math.floor(yInd);
+        const bottomInd = Math.ceil(yInd);
+        const yLerp = yInd - topInd;
+        for (let x = 0; x < cropWidth; x++) {
+          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);
+          if (xInd < 0 || xInd > imageWidth - 1) {
+            for (let c = 0; c < numChannels; c++) {
+              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+              output.values[ind] = extrapolationValue;
+            }
+            continue;
+          }
+          const leftInd = Math.floor(xInd);
+          const rightInd = Math.ceil(xInd);
+          const xLerp = xInd - leftInd;
+          for (let c = 0; c < numChannels; c++) {
+            let ind = c + leftInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];
+            const topLeft = imageVals[ind];
+            ind = c + rightInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];
+            const topRight = imageVals[ind];
+            ind = c + leftInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];
+            const bottomLeft = imageVals[ind];
+            ind = c + rightInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];
+            const bottomRight = imageVals[ind];
+            const top = topLeft + (topRight - topLeft) * xLerp;
+            const bottom = bottomLeft + (bottomRight - bottomLeft) * xLerp;
+            ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[ind] = top + (bottom - top) * yLerp;
+          }
+        }
+      } else {
+        for (let x = 0; x < cropWidth; ++x) {
+          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);
+          if (xInd < 0 || xInd > imageWidth - 1) {
+            for (let c = 0; c < numChannels; c++) {
+              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+              output.values[ind] = extrapolationValue;
+            }
+            continue;
+          }
+          const closestX = Math.round(xInd);
+          const closestY = Math.round(yInd);
+          for (let c = 0; c < numChannels; c++) {
+            const inInd = c + closestX * inStride[2] + closestY * inStride[1] + bInd * inStride[0];
+            const outInd = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[outInd] = imageVals[inInd];
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(output.shape, output.dtype, output.values);
+}
+var cropAndResizeConfig = {
+  kernelName: CropAndResize,
+  backendName: "cpu",
+  kernelFunc: cropAndResize2
+};
+function cumprod2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  assertNotComplex(x, "cumprod");
+  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);
+  let $x = x;
+  if (permutation != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];
+  if (permutedAxis !== $x.shape.length - 1) {
+    throw new Error(`backend.cumprod in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);
+  }
+  const resultDtype = upcastType($x.dtype, "int32");
+  const vals = util_exports.makeOnesTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  const finalDim = $x.shape[$x.shape.length - 1];
+  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;
+  for (let i = 0; i < aVals.length; i += finalDim) {
+    for (let j = 0; j < finalDim; j++) {
+      const idx = indexAdjuster(i, j);
+      if (j === 0) {
+        vals[idx] = exclusive ? 1 : aVals[idx];
+      } else {
+        const prevIdx = indexAdjuster(i, j - 1);
+        vals[idx] = exclusive ? aVals[prevIdx] * vals[prevIdx] : aVals[idx] * vals[prevIdx];
+      }
+    }
+  }
+  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo($x);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+var cumprodConfig = {
+  kernelName: Cumprod,
+  backendName: "cpu",
+  kernelFunc: cumprod2
+};
+function cumsum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  assertNotComplex(x, "cumsum");
+  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);
+  let $x = x;
+  if (permutation != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];
+  if (permutedAxis !== $x.shape.length - 1) {
+    throw new Error(`backend.cumsum in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);
+  }
+  const resultDtype = upcastType($x.dtype, "int32");
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  const finalDim = $x.shape[$x.shape.length - 1];
+  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;
+  for (let i = 0; i < aVals.length; i += finalDim) {
+    for (let j = 0; j < finalDim; j++) {
+      const idx = indexAdjuster(i, j);
+      if (j === 0) {
+        vals[idx] = exclusive ? 0 : aVals[idx];
+      } else {
+        const prevIdx = indexAdjuster(i, j - 1);
+        vals[idx] = exclusive ? aVals[prevIdx] + vals[prevIdx] : aVals[idx] + vals[prevIdx];
+      }
+    }
+  }
+  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo($x);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+var cumsumConfig = {
+  kernelName: Cumsum,
+  backendName: "cpu",
+  kernelFunc: cumsum2
+};
+function denseBincount2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size, binaryOutput } = attrs;
+  if (x.shape.length === 1) {
+    const xVals = backend2.data.get(x.dataId).values;
+    const weightsVals = backend2.data.get(weights.dataId).values;
+    const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);
+    return backend2.makeTensorInfo([size], weights.dtype, outVals);
+  } else if (x.shape.length === 2) {
+    const xBuf = backend2.bufferSync(x);
+    const weightsBuf = backend2.bufferSync(weights);
+    const outBuf = bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput);
+    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);
+  }
+  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);
+}
+var denseBincountConfig = {
+  kernelName: DenseBincount,
+  backendName: "cpu",
+  kernelFunc: denseBincount2
+};
+function depthToSpace2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  util_exports.assert(dataFormat === "NHWC", () => `Only NHWC dataFormat supported on CPU for depthToSpace. Got ${dataFormat}`);
+  const batchSize = x.shape[0];
+  const inputHeight = x.shape[1];
+  const inputWidth = x.shape[2];
+  const inputDepth = x.shape[3];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const xValues = backend2.data.get(x.dataId).values;
+  const result = new Float32Array(batchSize * outputHeight * outputWidth * outputDepth);
+  let outputIdx = 0;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let h = 0; h < outputHeight; ++h) {
+      const inH = Math.floor(h / blockSize);
+      const offsetH = h % blockSize;
+      for (let w = 0; w < outputWidth; ++w) {
+        const inW = Math.floor(w / blockSize);
+        const offsetW = w % blockSize;
+        const offsetD = (offsetH * blockSize + offsetW) * outputDepth;
+        for (let d = 0; d < outputDepth; ++d) {
+          const inD = d + offsetD;
+          const inputIdx = inD + inputDepth * (inW + inputWidth * (inH + inputHeight * b));
+          result[outputIdx++] = xValues[inputIdx];
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batchSize, outputHeight, outputWidth, outputDepth], x.dtype, result);
+}
+var depthToSpaceConfig = {
+  kernelName: DepthToSpace,
+  backendName: "cpu",
+  kernelFunc: depthToSpace2
+};
+function depthwiseConv2dNative(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;
+  assertNotComplex([x, filter], "depthwiseConv2DNative");
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const { filterHeight, filterWidth, dilationHeight, dilationWidth, padInfo } = convInfo;
+  const padLeft = padInfo.left;
+  const padTop = padInfo.top;
+  const chMul = convInfo.outChannels / convInfo.inChannels;
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xStrides[0];
+    const yOffset1 = b * y.strides[0];
+    for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+      const yOffset2 = yOffset1 + yR * y.strides[1];
+      const xRCorner = yR * convInfo.strideHeight - padTop;
+      for (let wR = 0; wR < filterHeight; ++wR) {
+        const xR = xRCorner + wR * dilationHeight;
+        if (xR < 0 || xR >= convInfo.inHeight) {
+          continue;
+        }
+        const wOffset1 = wR * filterStrides[0];
+        const xOffset2 = xOffset1 + xR * xStrides[1];
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const yOffset3 = yOffset2 + yC * y.strides[2];
+          const xCCorner = yC * convInfo.strideWidth - padLeft;
+          for (let wC = 0; wC < filterWidth; ++wC) {
+            const xC = xCCorner + wC * dilationWidth;
+            if (xC < 0 || xC >= convInfo.inWidth) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wC * filterStrides[1];
+            const xOffset3 = xOffset2 + xC * convInfo.inChannels;
+            let yOffset4 = yOffset3;
+            let wOffset3 = wOffset2;
+            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+              const xVal = xVals[xOffset3 + d1];
+              for (let q = 0; q < chMul; ++q) {
+                yVals[yOffset4 + q] += xVal * wVals[wOffset3 + q];
+              }
+              yOffset4 += chMul;
+              wOffset3 += chMul;
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);
+}
+var depthwiseConv2dNativeConfig = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNative
+};
+function depthwiseConv2dNativeBackpropFilter2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;
+  assertNotComplex([x, dy], "depthwiseConv2dNativeBackpropFilter");
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);
+  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;
+  const dW = new TensorBuffer(convInfo.filterShape, "float32");
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  const chMul = convInfo.outChannels / convInfo.inChannels;
+  const xVals = backend2.data.get(x.dataId).values;
+  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);
+  const dyVals = backend2.data.get(dy.dataId).values;
+  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);
+  for (let wR = 0; wR < filterHeight; ++wR) {
+    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+    for (let wC = 0; wC < filterWidth; ++wC) {
+      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+      for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+        const d1 = Math.trunc(d2 / chMul);
+        const dm = d2 % chMul;
+        let dotProd = 0;
+        for (let b = 0; b < convInfo.batchSize; ++b) {
+          for (let yR = yRMin; yR < yRMax; ++yR) {
+            const xR = wR + yR * strideHeight - topPad;
+            for (let yC = yCMin; yC < yCMax; ++yC) {
+              const xC = wC + yC * strideWidth - leftPad;
+              dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);
+            }
+          }
+        }
+        dW.set(dotProd, wR, wC, d1, dm);
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);
+}
+var depthwiseConv2dNativeBackpropFilterConfig = {
+  kernelName: DepthwiseConv2dNativeBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNativeBackpropFilter2
+};
+function depthwiseConv2dNativeBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;
+  assertNotComplex([dy, filter], "depthwiseConv2DNativeBackpropInput");
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const [dxS0, dxS1, dxS2] = dx.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2] = dyStrides;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2] = filterStrides;
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const chMul = outChannels / inChannels;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xR = 0; xR < inHeight; ++xR) {
+        const xRCorner = xR - topPad;
+        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+        for (let xC = 0; xC < inWidth; ++xC) {
+          const xCCorner = xC - leftPad;
+          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+          let dotProd = 0;
+          for (let yR = xRMin; yR < yRMax; ++yR) {
+            const wR = yR * strideHeight - xRCorner;
+            for (let yC = xCMin; yC < yCMax; ++yC) {
+              const wC = yC * strideWidth - xCCorner;
+              const dyOffset = dyS0 * b + dyS1 * yR + dyS2 * yC;
+              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;
+              for (let dm = 0; dm < chMul; ++dm) {
+                const d2 = d1 * chMul + dm;
+                const pixel = dyValues[dyOffset + d2];
+                const weight = fltValues[fltOffset + dm];
+                dotProd += pixel * weight;
+              }
+            }
+          }
+          dxValues[dxS0 * b + dxS1 * xR + dxS2 * xC + d1] = dotProd;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var depthwiseConv2dNativeBackpropInputConfig = {
+  kernelName: DepthwiseConv2dNativeBackpropInput,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNativeBackpropInput2
+};
+function diag2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xVals = backend2.data.get(x.dataId).values;
+  const outBuf = buffer([xSize, xSize], x.dtype);
+  const vals = outBuf.values;
+  for (let i = 0; i < xVals.length; i++) {
+    vals[i * xSize + i] = xVals[i];
+  }
+  const outShape = [...x.shape, ...x.shape];
+  return backend2.makeTensorInfo(outShape, outBuf.dtype, outBuf.values);
+}
+var diagConfig = {
+  kernelName: Diag,
+  backendName: "cpu",
+  kernelFunc: diag2
+};
+var dilation2DConfig = {
+  kernelName: Dilation2D,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const xVals = cpuBackend.data.get(x.dataId).values;
+    const xRank = x.shape.length;
+    const filterVals = cpuBackend.data.get(filter.dataId).values;
+    const filterRank = filter.shape.length;
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    const outSize = util_exports.sizeFromShape(outShape);
+    const outRank = outShape.length;
+    const outputVals = util_exports.getArrayFromDType(x.dtype, outSize);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const xIndex = util_exports.locToIndex([b, hIn, wIn, d], xRank, util_exports.computeStrides(x.shape));
+                    const filterIndex = util_exports.locToIndex([h, w, d], filterRank, util_exports.computeStrides(filter.shape));
+                    const val = xVals[xIndex] + filterVals[filterIndex];
+                    if (val > curVal) {
+                      curVal = val;
+                    }
+                  }
+                }
+              }
+            }
+            const outputIndex = util_exports.locToIndex([b, hOut, wOut, d], outRank, util_exports.computeStrides(outShape));
+            outputVals[outputIndex] = curVal;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(outputVals, x.dtype), outShape, x.dtype);
+    return { dataId, shape: outShape, dtype: x.dtype };
+  }
+};
+var dilation2DBackpropFilterConfig = {
+  kernelName: Dilation2DBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter, dy } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);
+    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropFilter}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);
+    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);
+    const gradients = util_exports.makeZerosNestedTypedArray(filter.shape, filter.dtype);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            let hMax = 0;
+            let wMax = 0;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];
+                    if (val > curVal) {
+                      curVal = val;
+                      hMax = h;
+                      wMax = w;
+                    }
+                  }
+                }
+              }
+            }
+            gradients[hMax][wMax][d] += $dy[b][hOut][wOut][d];
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), filter.shape, filter.dtype);
+    return { dataId, shape: filter.shape, dtype: filter.dtype };
+  }
+};
+var dilation2DBackpropInputConfig = {
+  kernelName: Dilation2DBackpropInput,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter, dy } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);
+    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropInput}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);
+    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);
+    const gradients = util_exports.makeZerosNestedTypedArray(x.shape, x.dtype);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            let hInMax = hBeg < 0 ? 0 : hBeg;
+            let wInMax = wBeg < 0 ? 0 : wBeg;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];
+                    if (val > curVal) {
+                      curVal = val;
+                      hInMax = hIn;
+                      wInMax = wIn;
+                    }
+                  }
+                }
+              }
+            }
+            gradients[b][hInMax][wInMax][d] += $dy[b][hOut][wOut][d];
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), x.shape, x.dtype);
+    return { dataId, shape: x.shape, dtype: x.dtype };
+  }
+};
+function sum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "sum");
+  let $x;
+  if (x.dtype === "bool") {
+    $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "int32" } });
+  } else {
+    $x = identity2({ inputs: { x }, backend: backend2 });
+  }
+  const xRank = $x.shape.length;
+  const axes = util_exports.parseAxisParam(axis, $x.shape);
+  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);
+  let reductionAxes = axes;
+  let permutedX = $x;
+  if (permutation != null) {
+    permutedX = transpose2({ inputs: { x: $x }, backend: backend2, attrs: { perm: permutation } });
+    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", reductionAxes, permutedX.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, reductionAxes);
+  const resultDtype = backend_util_exports.upcastType(permutedX.dtype, "int32");
+  let result = zeros3(backend2, outShape, resultDtype);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = backend2.data.get(result.dataId).values;
+  const aVals = backend2.data.get(permutedX.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let sum6 = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      sum6 += aVals[offset + j];
+    }
+    vals[i] = sum6;
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(result.shape, axes);
+    const oldResult = result;
+    result = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: newShape } });
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  backend2.disposeIntermediateTensorInfo($x);
+  if (permutation != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return result;
+}
+var sumConfig = {
+  kernelName: Sum,
+  backendName: "cpu",
+  kernelFunc: sum3
+};
+function einsum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { equation } = attrs;
+  const tensors = inputs;
+  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);
+  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);
+  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);
+  const nSteps = steps.length;
+  let out = null;
+  let numDimsRemaining = allDims.length;
+  const tensorsToDispose = [];
+  for (let i = 0; i < nSteps; ++i) {
+    for (const idTerm of steps[i]) {
+      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);
+      let x;
+      if (backend_util_exports.isIdentityPermutation(perm)) {
+        x = tensors[idTerm];
+      } else {
+        x = transpose2({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });
+        tensorsToDispose.push(x);
+      }
+      const targetShape = x.shape.slice();
+      for (let k = 0; k < dimsToExpand.length; ++k) {
+        targetShape.splice(dimsToExpand[k], 0, 1);
+      }
+      if (!util_exports.arraysEqual(x.shape, targetShape)) {
+        x = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });
+        tensorsToDispose.push(x);
+      }
+      if (out === null) {
+        out = x;
+      } else {
+        out = multiply2({ inputs: { a: x, b: out }, backend: backend2 });
+        tensorsToDispose.push(out);
+      }
+    }
+    if (i < nSteps - 1) {
+      if (path[i] >= 0) {
+        out = sum3({
+          inputs: { x: out },
+          backend: backend2,
+          attrs: {
+            axis: path[i] - (allDims.length - numDimsRemaining),
+            keepDims: false
+          }
+        });
+        tensorsToDispose.push(out);
+      }
+      numDimsRemaining--;
+    }
+  }
+  for (const tensorInfo of tensorsToDispose) {
+    if (tensorInfo === out) {
+      continue;
+    }
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+  return out;
+}
+var einsumConfig = {
+  kernelName: Einsum,
+  backendName: "cpu",
+  kernelFunc: einsum2
+};
+function eluGrad(args) {
+  const { inputs, backend: backend2 } = args;
+  const { dy, y } = inputs;
+  assertNotComplex([dy, y], "eluGrad");
+  const resultValues = new Float32Array(util_exports.sizeFromShape(y.shape));
+  const values = backend2.data.get(y.dataId).values;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  for (let i = 0; i < values.length; ++i) {
+    const v = values[i];
+    if (v >= 1) {
+      resultValues[i] = dyValues[i];
+    } else {
+      resultValues[i] = dyValues[i] * (v + 1);
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, "float32", resultValues);
+}
+var eluGradConfig2 = {
+  kernelName: EluGrad,
+  backendName: "cpu",
+  kernelFunc: eluGrad
+};
+var p = backend_util_exports.ERF_P;
+var a1 = backend_util_exports.ERF_A1;
+var a2 = backend_util_exports.ERF_A2;
+var a3 = backend_util_exports.ERF_A3;
+var a4 = backend_util_exports.ERF_A4;
+var a5 = backend_util_exports.ERF_A5;
+var erf2 = unaryKernelFunc(Erf, (xi) => {
+  const sign4 = Math.sign(xi);
+  const v = Math.abs(xi);
+  const t = 1 / (1 + p * v);
+  return sign4 * (1 - ((((a5 * t + a4) * t + a3) * t + a2) * t + a1) * t * Math.exp(-v * v));
+});
+var erfConfig = {
+  kernelName: Erf,
+  backendName: "cpu",
+  kernelFunc: erf2
+};
+function expandDims3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { input: input2 } = inputs;
+  const { dim } = attrs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape3({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig = {
+  kernelName: ExpandDims,
+  backendName: "cpu",
+  kernelFunc: expandDims3
+};
+var realDivImpl = createSimpleBinaryKernelImpl((a, b) => a / b);
+var div2 = binaryKernelFunc(RealDiv, realDivImpl);
+var realDivConfig = {
+  kernelName: RealDiv,
+  backendName: "cpu",
+  kernelFunc: div2
+};
+function fftBatch(input2, inverse, cpuBackend) {
+  const inputShape = input2.shape;
+  const batch = inputShape[0];
+  const innerDim = inputShape[1];
+  const inputVals = cpuBackend.data.get(input2.dataId);
+  const real2D = inputVals.complexTensorInfos.real;
+  const imag2D = inputVals.complexTensorInfos.imag;
+  const resultShape = [batch, innerDim];
+  const resultSize = util_exports.sizeFromShape(resultShape);
+  const resultReal = util_exports.getTypedArrayFromDType("float32", resultSize);
+  const resultImag = util_exports.getTypedArrayFromDType("float32", resultSize);
+  for (let b = 0; b < batch; b++) {
+    const r = slice2({
+      inputs: { x: real2D },
+      backend: cpuBackend,
+      attrs: { begin: [b, 0], size: [1, innerDim] }
+    });
+    const i = slice2({
+      inputs: { x: imag2D },
+      backend: cpuBackend,
+      attrs: { begin: [b, 0], size: [1, innerDim] }
+    });
+    const input3 = complex2({ inputs: { real: r, imag: i }, backend: cpuBackend });
+    const { real: real4, imag: imag4 } = fftImpl(input3, inverse, cpuBackend);
+    const res = backend_util_exports.mergeRealAndImagArrays(real4, imag4);
+    for (let d = 0; d < innerDim; d++) {
+      const c = backend_util_exports.getComplexWithIndex(res, d);
+      resultReal[b * innerDim + d] = c.real;
+      resultImag[b * innerDim + d] = c.imag;
+    }
+    cpuBackend.disposeIntermediateTensorInfo(r);
+    cpuBackend.disposeIntermediateTensorInfo(i);
+    cpuBackend.disposeIntermediateTensorInfo(input3);
+  }
+  const $realInfo = cpuBackend.makeTensorInfo(resultShape, "float32", resultReal);
+  const $imagInfo = cpuBackend.makeTensorInfo(resultShape, "float32", resultImag);
+  const result = complex2({ inputs: { real: $realInfo, imag: $imagInfo }, backend: cpuBackend });
+  cpuBackend.disposeIntermediateTensorInfo($realInfo);
+  cpuBackend.disposeIntermediateTensorInfo($imagInfo);
+  return result;
+}
+function fftImpl(input2, inverse, cpuBackend) {
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const inputVals = cpuBackend.data.get(input2.dataId);
+  const realVals = cpuBackend.data.get(inputVals.complexTensorInfos.real.dataId).values;
+  const imagVals = cpuBackend.data.get(inputVals.complexTensorInfos.imag.dataId).values;
+  if (isExponentOf2(inputSize)) {
+    const result = fftRadix2(realVals, imagVals, inputSize, inverse, cpuBackend);
+    const resultShape = [input2.shape[0], input2.shape[1]];
+    if (inverse) {
+      const realInfo = cpuBackend.makeTensorInfo(resultShape, "float32", result.real);
+      const imagInfo = cpuBackend.makeTensorInfo(resultShape, "float32", result.imag);
+      const sizeInfo = cpuBackend.makeTensorInfo([], "float32", util_exports.createScalarValue(inputSize, "float32"));
+      const sizeInfoCopy = identity2({ inputs: { x: sizeInfo }, backend: cpuBackend });
+      const divRealInfo = realDivConfig.kernelFunc({ inputs: { a: realInfo, b: sizeInfo }, backend: cpuBackend });
+      const divImagInfo = realDivConfig.kernelFunc({ inputs: { a: imagInfo, b: sizeInfoCopy }, backend: cpuBackend });
+      const divRealVals = cpuBackend.data.get(divRealInfo.dataId).values;
+      const divImagVals = cpuBackend.data.get(divImagInfo.dataId).values;
+      cpuBackend.disposeIntermediateTensorInfo(realInfo);
+      cpuBackend.disposeIntermediateTensorInfo(imagInfo);
+      cpuBackend.disposeIntermediateTensorInfo(sizeInfo);
+      cpuBackend.disposeIntermediateTensorInfo(sizeInfoCopy);
+      cpuBackend.disposeIntermediateTensorInfo(divRealInfo);
+      cpuBackend.disposeIntermediateTensorInfo(divImagInfo);
+      return { real: divRealVals, imag: divImagVals };
+    }
+    return result;
+  } else {
+    const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);
+    const rawOutput = fourierTransformByMatmul(data, inputSize, inverse);
+    return backend_util_exports.splitRealAndImagArrays(rawOutput);
+  }
+}
+function isExponentOf2(size) {
+  return (size & size - 1) === 0;
+}
+function fftRadix2(realVals, imagVals, size, inverse, cpuBackend) {
+  if (size === 1) {
+    return { real: realVals, imag: imagVals };
+  }
+  const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);
+  const half = size / 2;
+  const evenComplex = backend_util_exports.complexWithEvenIndex(data);
+  const evenRealVals = evenComplex.real;
+  const evenImagVals = evenComplex.imag;
+  const evenShape = [evenRealVals.length];
+  const evenRealInfo = cpuBackend.makeTensorInfo(evenShape, "float32", evenRealVals);
+  const evenImagInfo = cpuBackend.makeTensorInfo(evenShape, "float32", evenImagVals);
+  const evenTensorInfo = complex2({ inputs: { real: evenRealInfo, imag: evenImagInfo }, backend: cpuBackend });
+  const oddComplex = backend_util_exports.complexWithOddIndex(data);
+  const oddRealVals = oddComplex.real;
+  const oddImagVals = oddComplex.imag;
+  const oddShape = [oddRealVals.length];
+  const oddRealInfo = cpuBackend.makeTensorInfo(oddShape, "float32", oddRealVals);
+  const oddImagInfo = cpuBackend.makeTensorInfo(oddShape, "float32", oddImagVals);
+  const oddTensorInfo = complex2({ inputs: { real: oddRealInfo, imag: oddImagInfo }, backend: cpuBackend });
+  const $evenComplex = fftRadix2(evenRealVals, evenImagVals, half, inverse, cpuBackend);
+  const $evenRealVals = $evenComplex.real;
+  const $evenImagVals = $evenComplex.imag;
+  const $evenShape = [$evenRealVals.length];
+  const $evenRealInfo = cpuBackend.makeTensorInfo($evenShape, "float32", $evenRealVals);
+  const $evenImagInfo = cpuBackend.makeTensorInfo($evenShape, "float32", $evenImagVals);
+  const $evenTensorInfo = complex2({
+    inputs: { real: $evenRealInfo, imag: $evenImagInfo },
+    backend: cpuBackend
+  });
+  const $oddComplex = fftRadix2(oddRealVals, oddImagVals, half, inverse, cpuBackend);
+  const $oddRealVals = $oddComplex.real;
+  const $oddImagVals = $oddComplex.imag;
+  const $oddShape = [$oddRealVals.length];
+  const $oddRealInfo = cpuBackend.makeTensorInfo($oddShape, "float32", $oddRealVals);
+  const $oddImagInfo = cpuBackend.makeTensorInfo($oddShape, "float32", $oddImagVals);
+  const $oddTensorInfo = complex2({ inputs: { real: $oddRealInfo, imag: $oddImagInfo }, backend: cpuBackend });
+  const e = backend_util_exports.exponents(size, inverse);
+  const eShape = [e.real.length];
+  const eRealInfo = cpuBackend.makeTensorInfo(eShape, "float32", e.real);
+  const eImagInfo = cpuBackend.makeTensorInfo(eShape, "float32", e.imag);
+  const complexInfo = complex2({ inputs: { real: eRealInfo, imag: eImagInfo }, backend: cpuBackend });
+  const exponentInfo = multiply2({ inputs: { a: complexInfo, b: $oddTensorInfo }, backend: cpuBackend });
+  const addPart = add4({
+    inputs: { a: $evenTensorInfo, b: exponentInfo },
+    backend: cpuBackend
+  });
+  const subPart = sub2({
+    inputs: { a: $evenTensorInfo, b: exponentInfo },
+    backend: cpuBackend
+  });
+  const addPartReal = real2({ inputs: { input: addPart }, backend: cpuBackend });
+  const subPartReal = real2({ inputs: { input: subPart }, backend: cpuBackend });
+  const addPartImag = imag2({ inputs: { input: addPart }, backend: cpuBackend });
+  const subPartImag = imag2({ inputs: { input: subPart }, backend: cpuBackend });
+  const $real = concat2({
+    inputs: [addPartReal, subPartReal],
+    backend: cpuBackend,
+    attrs: { axis: 0 }
+  });
+  const $imag = concat2({
+    inputs: [addPartImag, subPartImag],
+    backend: cpuBackend,
+    attrs: { axis: 0 }
+  });
+  const $realVals = cpuBackend.data.get($real.dataId).values;
+  const $imagVals = cpuBackend.data.get($imag.dataId).values;
+  cpuBackend.disposeIntermediateTensorInfo(evenRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(evenImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(evenTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo(eRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(eImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(complexInfo);
+  cpuBackend.disposeIntermediateTensorInfo(exponentInfo);
+  cpuBackend.disposeIntermediateTensorInfo(addPart);
+  cpuBackend.disposeIntermediateTensorInfo(subPart);
+  cpuBackend.disposeIntermediateTensorInfo(addPartReal);
+  cpuBackend.disposeIntermediateTensorInfo(addPartImag);
+  cpuBackend.disposeIntermediateTensorInfo(subPartReal);
+  cpuBackend.disposeIntermediateTensorInfo(subPartImag);
+  cpuBackend.disposeIntermediateTensorInfo($real);
+  cpuBackend.disposeIntermediateTensorInfo($imag);
+  return { real: $realVals, imag: $imagVals };
+}
+function fourierTransformByMatmul(data, size, inverse) {
+  const ret = new Float32Array(size * 2);
+  for (let r = 0; r < size; r++) {
+    let real4 = 0;
+    let imag4 = 0;
+    for (let c = 0; c < size; c++) {
+      const e = backend_util_exports.exponent(r * c, size, inverse);
+      const term = backend_util_exports.getComplexWithIndex(data, c);
+      real4 += term.real * e.real - term.imag * e.imag;
+      imag4 += term.real * e.imag + term.imag * e.real;
+    }
+    if (inverse) {
+      real4 /= size;
+      imag4 /= size;
+    }
+    backend_util_exports.assignToTypedArray(ret, real4, imag4, r);
+  }
+  return ret;
+}
+function fft2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape3({
+    inputs: { x: input2 },
+    backend: backend2,
+    attrs: { shape: [batch, innerDimensionSize] }
+  });
+  const result = fftBatch(input2D, false, backend2);
+  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var fftConfig = {
+  kernelName: FFT,
+  backendName: "cpu",
+  kernelFunc: fft2
+};
+function fill2(args) {
+  const { backend: backend2, attrs } = args;
+  const { shape, value, dtype } = attrs;
+  const $dtype = dtype || util_exports.inferDtype(value);
+  const values = util_exports.getArrayFromDType($dtype, util_exports.sizeFromShape(shape));
+  fillValues(values, value, $dtype);
+  return backend2.makeTensorInfo(shape, $dtype, values);
+}
+var fillConfig = {
+  kernelName: Fill,
+  backendName: "cpu",
+  kernelFunc: fill2
+};
+function fillValues(values, value, dtype) {
+  if (dtype === "string") {
+    values.fill(value);
+  } else {
+    values.fill(value);
+  }
+}
+var flipLeftRightConfig = {
+  kernelName: FlipLeftRight,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const cpuBackend = backend2;
+    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));
+    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+    const imageVals = cpuBackend.data.get(image2.dataId).values;
+    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {
+      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;
+      for (let row = 0; row < imageHeight; row++) {
+        const rowOffset = row * (imageWidth * numChannels);
+        for (let col = 0; col < imageWidth; col++) {
+          const colOffset = col * numChannels;
+          for (let channel = 0; channel < numChannels; channel++) {
+            const coordX = Math.round(imageWidth - col - 1);
+            const outIdx = batchOffset + rowOffset + colOffset + channel;
+            let outputValue = imageVals[outIdx];
+            if (coordX >= 0 && coordX < imageWidth) {
+              const rotatedColOffset = coordX * numChannels;
+              const imageIdx = batchOffset + rowOffset + rotatedColOffset + channel;
+              outputValue = imageVals[imageIdx];
+            }
+            output[outIdx] = outputValue;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);
+    return { dataId, shape: image2.shape, dtype: image2.dtype };
+  }
+};
+var floorDivImpl = createSimpleBinaryKernelImpl((a, b) => Math.floor(a / b));
+var floorDiv2 = binaryKernelFunc(FloorDiv, floorDivImpl, null, "int32");
+var floorDivConfig = {
+  kernelName: FloorDiv,
+  backendName: "cpu",
+  kernelFunc: floorDiv2
+};
+function fusedConv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  let result = conv2D({
+    inputs: { x, filter },
+    backend: backend2,
+    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }
+  });
+  if (bias) {
+    const resultOld = result;
+    if (dataFormat === "NCHW" && bias.shape.length === 1 && bias.shape[0] !== 1) {
+      const reshapedBias = reshape3({ inputs: { x: bias }, backend: backend2, attrs: { shape: [bias.shape[0], 1, 1] } });
+      result = add4({ inputs: { a: result, b: reshapedBias }, backend: backend2 });
+      backend2.disposeIntermediateTensorInfo(reshapedBias);
+    } else {
+      result = add4({ inputs: { a: result, b: bias }, backend: backend2 });
+    }
+    backend2.disposeIntermediateTensorInfo(resultOld);
+  }
+  if (activation2) {
+    const resultOld = result;
+    if (dataFormat === "NCHW" && activation2 === "prelu" && preluActivationWeights.shape.length === 1 && preluActivationWeights.shape[0] !== 1) {
+      const reshapedAlpha = reshape3({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: [preluActivationWeights.shape[0], 1, 1] }
+      });
+      result = applyActivation2(backend2, result, activation2, reshapedAlpha, leakyreluAlpha);
+      backend2.disposeIntermediateTensorInfo(reshapedAlpha);
+    } else {
+      result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);
+    }
+    backend2.disposeIntermediateTensorInfo(resultOld);
+  }
+  return result;
+}
+var fusedConv2DConfig = {
+  kernelName: FusedConv2D,
+  backendName: "cpu",
+  kernelFunc: fusedConv2D
+};
+function fusedDepthwiseConv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  let result = depthwiseConv2dNative({
+    inputs: { x, filter },
+    backend: backend2,
+    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }
+  });
+  if (bias) {
+    const oldResult = result;
+    result = add4({ inputs: { a: result, b: bias }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  if (activation2) {
+    const oldResult = result;
+    result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  return result;
+}
+var fusedDepthwiseConv2DConfig = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "cpu",
+  kernelFunc: fusedDepthwiseConv2D
+};
+function gatherNd(args) {
+  const { inputs, backend: backend2 } = args;
+  const { params, indices } = inputs;
+  const paramsSize = util_exports.sizeFromShape(params.shape);
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);
+  if (numSlices === 0) {
+    return backend2.makeTensorInfo(resultShape, params.dtype, []);
+  }
+  const indicesData = backend2.data.get(indices.dataId).values;
+  const paramsBuf = backend2.bufferSync(params);
+  const outBuf = gatherNdImpl(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);
+  return backend2.makeTensorInfo(resultShape, params.dtype, outBuf.values);
+}
+var gatherNdConfig = {
+  kernelName: GatherNd,
+  backendName: "cpu",
+  kernelFunc: gatherNd
+};
+function gatherV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  assertNotComplex([x, indices], "gatherV2");
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const indicesVals = backend2.data.get(indices.dataId).values;
+  const axisDim = x.shape[parsedAxis];
+  for (let i = 0; i < indicesVals.length; ++i) {
+    const index = indicesVals[i];
+    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+  }
+  let $batchDims = batchDims;
+  if (batchDims == null) {
+    $batchDims = 0;
+  }
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, $batchDims);
+  const flattenX = reshape3({
+    inputs: { x },
+    backend: backend2,
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    }
+  });
+  const flattenIndex = reshape3({
+    inputs: { x: indices },
+    backend: backend2,
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }
+  });
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  const indicesBuf = backend2.bufferSync(flattenIndex);
+  const xBuf = backend2.bufferSync(flattenX);
+  const outBuf = gatherV2Impl(xBuf, indicesBuf, flattenOutputShape);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(flattenIndex);
+  return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);
+}
+var gatherV2Config = {
+  kernelName: GatherV2,
+  backendName: "cpu",
+  kernelFunc: gatherV2
+};
+function ifft2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape3({
+    inputs: { x: input2 },
+    backend: backend2,
+    attrs: { shape: [batch, innerDimensionSize] }
+  });
+  const result = fftBatch(input2D, true, backend2);
+  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var ifftConfig = {
+  kernelName: IFFT,
+  backendName: "cpu",
+  kernelFunc: ifft2
+};
+var isFinite3 = unaryKernelFunc(IsFinite, (xi) => Number.isFinite(xi) ? 1 : 0, "bool");
+var isFiniteConfig = {
+  kernelName: IsFinite,
+  backendName: "cpu",
+  kernelFunc: isFinite3
+};
+var isInf2 = unaryKernelFunc(IsInf, (xi) => Math.abs(xi) === Infinity ? 1 : 0, "bool");
+var isInfConfig = {
+  kernelName: IsInf,
+  backendName: "cpu",
+  kernelFunc: isInf2
+};
+var isNaN3 = unaryKernelFunc(IsNan, (xi) => Number.isNaN(xi) ? 1 : 0, "bool");
+var isNaNConfig = {
+  kernelName: IsNan,
+  backendName: "cpu",
+  kernelFunc: isNaN3
+};
+function linSpace(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, num } = attrs;
+  const outVals = linSpaceImpl(start, stop, num);
+  return backend2.makeTensorInfo([outVals.length], "float32", outVals);
+}
+var linSpaceConfig = {
+  kernelName: LinSpace,
+  backendName: "cpu",
+  kernelFunc: linSpace
+};
+var log1p2 = unaryKernelFunc(Log1p, (xi) => Math.log1p(xi));
+var log1pConfig = {
+  kernelName: Log1p,
+  backendName: "cpu",
+  kernelFunc: log1p2
+};
+var logicalAndImpl = createSimpleBinaryKernelImpl((a, b) => a && b);
+var logicalAnd2 = binaryKernelFunc(LogicalAnd, logicalAndImpl, null, "bool");
+var logicalAndConfig = {
+  kernelName: LogicalAnd,
+  backendName: "cpu",
+  kernelFunc: logicalAnd2
+};
+var logicalNot2 = unaryKernelFunc(LogicalNot, (xi) => xi ? 0 : 1, "bool");
+var logicalNotConfig = {
+  kernelName: LogicalNot,
+  backendName: "cpu",
+  kernelFunc: logicalNot2
+};
+var logicalOrImpl = createSimpleBinaryKernelImpl((a, b) => a || b);
+var logicalOr2 = binaryKernelFunc(LogicalOr, logicalOrImpl, null, "bool");
+var logicalOrConfig = {
+  kernelName: LogicalOr,
+  backendName: "cpu",
+  kernelFunc: logicalOr2
+};
+function lRN(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  assertNotComplex(x, "LRN");
+  const channels = x.shape[3];
+  const maxD = channels - 1;
+  const xValues = backend2.data.get(x.dataId).values;
+  const size = util_exports.sizeFromShape(x.shape);
+  const result = new Float32Array(size);
+  function sumAcrossChannels(offset) {
+    const currentChannel = offset % channels;
+    let beginSumOffset = offset - currentChannel + Math.max(0, currentChannel - depthRadius);
+    const endSumOffset = offset - currentChannel + Math.min(currentChannel + depthRadius, maxD);
+    let sum6 = 0;
+    for (; beginSumOffset <= endSumOffset; beginSumOffset++) {
+      const z = xValues[beginSumOffset];
+      sum6 += z * z;
+    }
+    return sum6;
+  }
+  for (let offset = 0; offset < size; offset++) {
+    const sum6 = sumAcrossChannels(offset);
+    const val = xValues[offset] * Math.pow(bias + alpha * sum6, -beta);
+    result[offset] = val;
+  }
+  return backend2.makeTensorInfo(x.shape, x.dtype, result);
+}
+var LRNConfig = {
+  kernelName: LRN,
+  backendName: "cpu",
+  kernelFunc: lRN
+};
+function lRNGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, y, dy } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  assertNotComplex(dy, "LRNGrad");
+  const dySize = util_exports.sizeFromShape(dy.shape);
+  const channels = dy.shape[3];
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const xValues = backend2.data.get(x.dataId).values;
+  const yValues = backend2.data.get(y.dataId).values;
+  const result = new Float32Array(dySize);
+  const size = dySize;
+  for (let offset = 0; offset < size; offset++) {
+    const currentChannel = offset % channels;
+    const depthBegin = offset - currentChannel + Math.max(0, currentChannel - depthRadius);
+    const depthEnd = offset - currentChannel + Math.min(channels, currentChannel + depthRadius + 1);
+    let norm2 = 0;
+    for (let k = depthBegin; k < depthEnd; k++) {
+      norm2 += Math.pow(xValues[k], 2);
+    }
+    norm2 = alpha * norm2 + bias;
+    for (let k = depthBegin; k < depthEnd; k++) {
+      let dyi = -2 * alpha * beta * xValues[k] * yValues[offset] / norm2;
+      if (offset === k) {
+        dyi += Math.pow(norm2, -beta);
+      }
+      dyi *= dyValues[offset];
+      result[k] += dyi;
+    }
+  }
+  return backend2.makeTensorInfo(dy.shape, x.dtype, result);
+}
+var LRNGradConfig = {
+  kernelName: LRNGrad,
+  backendName: "cpu",
+  kernelFunc: lRNGrad
+};
+function max3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reductionIndices, keepDims } = attrs;
+  const cpuBackend = backend2;
+  let xShape = x.shape;
+  const xRank = xShape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, xShape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let xVals = cpuBackend.data.get(x.dataId).values;
+  if (permutedAxes != null) {
+    const newShape = new Array(xRank);
+    for (let i = 0; i < newShape.length; i++) {
+      newShape[i] = xShape[permutedAxes[i]];
+    }
+    xVals = transposeImpl(xVals, xShape, x.dtype, permutedAxes, newShape);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    xShape = newShape;
+  }
+  assertNotComplex(x, "max");
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, xRank);
+  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const result = maxImpl(xVals, reduceSize, maxOutShape, x.dtype);
+  const dataId = cpuBackend.write(result, maxOutShape, x.dtype);
+  let outShape = maxOutShape;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);
+    outShape = newShape;
+  }
+  return { dataId, shape: outShape, dtype: x.dtype };
+}
+var maxConfig = {
+  kernelName: Max,
+  backendName: "cpu",
+  kernelFunc: max3
+};
+function maxPool2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "maxPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  let res;
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    res = identity2({ inputs: { x }, backend: backend2 });
+  } else {
+    const xValues = backend2.data.get(x.dataId).values;
+    const strides2 = util_exports.computeStrides(x.shape);
+    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, "max");
+    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);
+  }
+  return res;
+}
+var maxPoolConfig = {
+  kernelName: MaxPool,
+  backendName: "cpu",
+  kernelFunc: maxPool2
+};
+function maxPool3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  assertNotComplex(x, "maxPool3d");
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);
+  const xValues = backend2.data.get(x.dataId).values;
+  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, "max");
+  return backend2.makeTensorInfo(outBuf.shape, "float32", outBuf.values);
+}
+var maxPool3DConfig = {
+  kernelName: MaxPool3D,
+  backendName: "cpu",
+  kernelFunc: maxPool3D
+};
+function maxPool3DGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  assertNotComplex([dy, input2], "maxPool3DGrad");
+  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const inputBuf = backend2.bufferSync(input2);
+  const maxPosBuf = maxPool3dPositions(inputBuf, convInfo);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(input2.shape, "float32");
+  const dyBuf = backend2.bufferSync(dy);
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {
+        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {
+          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {
+            const dyDepthCorner = dxDepth - padFront;
+            const dyRowCorner = dxRow - padTop;
+            const dyColCorner = dxCol - padLeft;
+            let dotProd = 0;
+            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {
+              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;
+              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {
+                continue;
+              }
+              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {
+                const dyRow = (dyRowCorner + wRow) / strideHeight;
+                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {
+                  continue;
+                }
+                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {
+                  const dyCol = (dyColCorner + wCol) / strideWidth;
+                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {
+                    continue;
+                  }
+                  const maxPos = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  const curPos = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterWidth + wCol;
+                  const mask = maxPos === curPos ? 1 : 0;
+                  if (mask === 0) {
+                    continue;
+                  }
+                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  dotProd += pixel * mask;
+                }
+              }
+            }
+            dx.set(dotProd, batch, dxDepth, dxRow, dxCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var maxPool3DGradConfig2 = {
+  kernelName: MaxPool3DGrad,
+  backendName: "cpu",
+  kernelFunc: maxPool3DGrad
+};
+function maxPoolGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2, output } = inputs;
+  const x = input2;
+  assertNotComplex([input2, output], "maxPoolGrad");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const xValues = backend2.data.get(x.dataId).values;
+  const maxPosBuf = buffer(convInfo.outShape, x.dtype, maxPoolPositions(xValues, x.shape, x.dtype, convInfo).values);
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(x.shape, "float32");
+  const dyData = backend2.data.get(dy.dataId).values;
+  const dyBuf = buffer(dy.shape, "float32", dyData);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {
+        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {
+          const dyRCorner = dxR - padTop;
+          const dyCCorner = dxC - padLeft;
+          let dotProd = 0;
+          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {
+            const dyR = (dyRCorner + wR) / strideHeight;
+            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {
+              continue;
+            }
+            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {
+              const dyC = (dyCCorner + wC) / strideWidth;
+              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {
+                continue;
+              }
+              const maxPos = effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(b, dyR, dyC, d);
+              const curPos = wR * effectiveFilterWidth + wC;
+              const mask = maxPos === curPos ? 1 : 0;
+              if (mask === 0) {
+                continue;
+              }
+              const pixel = dyBuf.get(b, dyR, dyC, d);
+              dotProd += pixel * mask;
+            }
+          }
+          dx.set(dotProd, b, dxR, dxC, d);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var maxPoolGradConfig2 = {
+  kernelName: MaxPoolGrad,
+  backendName: "cpu",
+  kernelFunc: maxPoolGrad2
+};
+function maxPoolWithArgmaxImpl(xValues, xShape, dtype, includeBatchInIndex, convInfo) {
+  const strides = util_exports.computeStrides(xShape);
+  const maxPools = pool2(xValues, xShape, dtype, strides, convInfo, "max");
+  const maxPositions = maxPoolPositions(xValues, xShape, dtype, convInfo, true, includeBatchInIndex);
+  return [maxPools.values, maxPositions.values];
+}
+var maxPoolWithArgmaxConfig = {
+  kernelName: MaxPoolWithArgmax,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;
+    const cpuBackend = backend2;
+    assertNotComplex(x, "MaxPoolWithArgmax");
+    const values = cpuBackend.data.get(x.dataId).values;
+    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, [1, 1], pad3);
+    const [pooled, indexes] = maxPoolWithArgmaxImpl(values, x.shape, x.dtype, includeBatchInIndex, convInfo);
+    const pooledDataId = cpuBackend.write(pooled, convInfo.outShape, x.dtype);
+    const indexesDataId = cpuBackend.write(indexes, convInfo.outShape, x.dtype);
+    return [
+      { dataId: pooledDataId, shape: convInfo.outShape, dtype: x.dtype },
+      { dataId: indexesDataId, shape: convInfo.outShape, dtype: "int32" }
+    ];
+  }
+};
+function mean2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const axes = util_exports.parseAxisParam(axis, x.shape);
+  const shapes = backend_util_exports.computeOutAndReduceShapes(x.shape, axes);
+  const reduceShape = shapes[1];
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const toDispose = [];
+  const reduceSizeScalar = backend2.makeTensorInfo([], "float32", new Float32Array([reduceSize]));
+  toDispose.push(reduceSizeScalar);
+  const $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+  toDispose.push($x);
+  const res = div2({ inputs: { a: $x, b: reduceSizeScalar }, backend: backend2 });
+  toDispose.push(res);
+  const result = sum3({ inputs: { x: res }, backend: backend2, attrs: { axis, keepDims } });
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var meanConfig = {
+  kernelName: Mean,
+  backendName: "cpu",
+  kernelFunc: mean2
+};
+function min3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "min");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let min6 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (Number.isNaN(value) || value < min6) {
+        min6 = value;
+      }
+    }
+    vals[i] = min6;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var minConfig = {
+  kernelName: Min,
+  backendName: "cpu",
+  kernelFunc: min3
+};
+function mirrorPad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, mode } = attrs;
+  assertNotComplex(x, "mirrorPad");
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const start = paddings.map((p2) => p2[0]);
+  const end = paddings.map((p2, i) => p2[0] + x.shape[i]);
+  const offset = mode === "reflect" ? 0 : 1;
+  const xVals = backend2.data.get(x.dataId).values;
+  const xRank = x.shape.length;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const resultSize = util_exports.sizeFromShape(outShape);
+  const resultRank = outShape.length;
+  const resultStrides = util_exports.computeStrides(outShape);
+  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);
+  for (let i = 0; i < resultSize; i++) {
+    let coords2 = util_exports.indexToLoc(i, resultRank, resultStrides);
+    for (let i2 = 0; i2 < resultRank; i2++) {
+      if (coords2[i2] < start[i2]) {
+        coords2[i2] = start[i2] * 2 - coords2[i2] - offset;
+      } else if (coords2[i2] >= end[i2]) {
+        coords2[i2] = (end[i2] - 1) * 2 - coords2[i2] + offset;
+      }
+    }
+    coords2 = coords2.map((c, i2) => c - start[i2]);
+    const inIndex = util_exports.locToIndex(coords2, xRank, xStrides);
+    resVals[i] = xVals[inIndex];
+  }
+  const outId = backend2.write(resVals, outShape, x.dtype);
+  return { dataId: outId, shape: outShape, dtype: x.dtype };
+}
+var mirrorPadConfig = {
+  kernelName: MirrorPad,
+  backendName: "cpu",
+  kernelFunc: mirrorPad2
+};
+var modImpl = createSimpleBinaryKernelImpl((aValue, bValue) => {
+  const rem = aValue % bValue;
+  if (aValue < 0 && bValue < 0 || aValue >= 0 && bValue >= 0) {
+    return rem;
+  } else {
+    return (rem + bValue) % bValue;
+  }
+});
+var mod2 = binaryKernelFunc(Mod, modImpl);
+var modConfig = {
+  kernelName: Mod,
+  backendName: "cpu",
+  kernelFunc: mod2
+};
+var seedrandom4 = __toESM(require_seedrandom2());
+function softmax3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { dim } = attrs;
+  const logitsRank = logits.shape.length;
+  let $dim = dim;
+  if ($dim === -1) {
+    $dim = logitsRank - 1;
+  }
+  if ($dim !== logitsRank - 1) {
+    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${logitsRank} and dim was ${$dim}`);
+  }
+  const axes = util_exports.parseAxisParam([$dim], logits.shape);
+  const maxLogit = max3({
+    inputs: { x: logits },
+    backend: backend2,
+    attrs: { reductionIndices: axes, keepDims: false }
+  });
+  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);
+  const maxLogitReshaped = reshape3({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });
+  const a = sub2({ inputs: { a: logits, b: maxLogitReshaped }, backend: backend2 });
+  const b = exp2({ inputs: { x: a }, backend: backend2 });
+  const sumExp = sum3({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });
+  const sumReshaped = reshape3({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });
+  const result = div2({ inputs: { a: b, b: sumReshaped }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(maxLogit);
+  backend2.disposeIntermediateTensorInfo(maxLogitReshaped);
+  backend2.disposeIntermediateTensorInfo(a);
+  backend2.disposeIntermediateTensorInfo(b);
+  backend2.disposeIntermediateTensorInfo(sumExp);
+  backend2.disposeIntermediateTensorInfo(sumReshaped);
+  return result;
+}
+var softmaxConfig = {
+  kernelName: Softmax,
+  backendName: "cpu",
+  kernelFunc: softmax3
+};
+function multinomial2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { numSamples, seed, normalized } = attrs;
+  assertNotComplex(logits, "multinomial");
+  const probabilities = normalized ? logits : softmax3({ inputs: { logits }, backend: backend2, attrs: { dim: -1 } });
+  const batchSize = probabilities.shape[0];
+  const numEvents = probabilities.shape[1];
+  const probVals = backend2.data.get(probabilities.dataId).values;
+  const resShape = [batchSize, numSamples];
+  const resVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(resShape), "int32");
+  for (let b = 0; b < batchSize; ++b) {
+    const offset = b * numEvents;
+    const cdf = new Float32Array(numEvents - 1);
+    cdf[0] = probVals[offset];
+    for (let event = 1; event < cdf.length; ++event) {
+      cdf[event] = cdf[event - 1] + probVals[offset + event];
+    }
+    const random = seedrandom4.alea(seed.toString());
+    const outOffset = b * numSamples;
+    for (let sampleId = 0; sampleId < numSamples; ++sampleId) {
+      const r = random();
+      resVals[outOffset + sampleId] = cdf.length;
+      for (let event = 0; event < cdf.length; event++) {
+        if (r < cdf[event]) {
+          resVals[outOffset + sampleId] = event;
+          break;
+        }
+      }
+    }
+  }
+  if (!normalized) {
+    backend2.disposeIntermediateTensorInfo(probabilities);
+  }
+  return backend2.makeTensorInfo(resShape, "int32", resVals);
+}
+var multinomialConfig = {
+  kernelName: Multinomial,
+  backendName: "cpu",
+  kernelFunc: multinomial2
+};
+var nonMaxSuppressionV3Impl2 = kernel_impls_exports.nonMaxSuppressionV3Impl;
+function nonMaxSuppressionV3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppression");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const { selectedIndices } = nonMaxSuppressionV3Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  return backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices));
+}
+var nonMaxSuppressionV3Config = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV3
+};
+var nonMaxSuppressionV4Impl2 = kernel_impls_exports.nonMaxSuppressionV4Impl;
+function nonMaxSuppressionV4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppressionPadded");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([], "int32", new Int32Array([validOutputs]))
+  ];
+}
+var nonMaxSuppressionV4Config = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV4
+};
+var nonMaxSuppressionV5Impl2 = kernel_impls_exports.nonMaxSuppressionV5Impl;
+function nonMaxSuppressionV5(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppressionWithScore");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const maxOutputSizeVal = maxOutputSize;
+  const iouThresholdVal = iouThreshold;
+  const scoreThresholdVal = scoreThreshold;
+  const softNmsSigmaVal = softNmsSigma;
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl2(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([selectedScores.length], "float32", new Float32Array(selectedScores))
+  ];
+}
+var nonMaxSuppressionV5Config = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV5
+};
+function oneHot2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  assertNotComplex(indices, "oneHot");
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const res = new Float32Array(indicesSize * depth);
+  res.fill(offValue);
+  const indicesVal = backend2.data.get(indices.dataId).values;
+  for (let event = 0; event < indicesSize; ++event) {
+    if (indicesVal[event] >= 0 && indicesVal[event] < depth) {
+      res[event * depth + indicesVal[event]] = onValue;
+    }
+  }
+  return backend2.makeTensorInfo([...indices.shape, depth], dtype, res);
+}
+var oneHotConfig = {
+  kernelName: OneHot,
+  backendName: "cpu",
+  kernelFunc: oneHot2
+};
+function zerosLike2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("zerosLike is not supported for string tensors");
+  } else if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const r = zerosLike2({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 0, dtype: x.dtype } });
+  }
+}
+var zerosLikeConfig = {
+  kernelName: ZerosLike,
+  backendName: "cpu",
+  kernelFunc: zerosLike2
+};
+function onesLike2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("onesLike is not supported for string tensors");
+  } else if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const r = onesLike2({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 1, dtype: x.dtype } });
+  }
+}
+var onesLikeConfig = {
+  kernelName: OnesLike,
+  backendName: "cpu",
+  kernelFunc: onesLike2
+};
+function pack(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims3({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims3({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat2({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var packConfig = {
+  kernelName: Pack,
+  backendName: "cpu",
+  kernelFunc: pack
+};
+function padV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, constantValue } = attrs;
+  assertNotComplex(x, "pad");
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const start = paddings.map((p2) => p2[0]);
+  const xVals = backend2.data.get(x.dataId).values;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xRank = x.shape.length;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const resultSize = util_exports.sizeFromShape(outShape);
+  const resultRank = outShape.length;
+  const resultStrides = util_exports.computeStrides(outShape);
+  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);
+  if (constantValue !== 0) {
+    resVals.fill(constantValue);
+  }
+  for (let i = 0; i < xSize; i++) {
+    const coords2 = util_exports.indexToLoc(i, xRank, xStrides);
+    const outCoords = coords2.map((c, i2) => c + start[i2]);
+    const outIndex = util_exports.locToIndex(outCoords, resultRank, resultStrides);
+    resVals[outIndex] = xVals[i];
+  }
+  const outId = backend2.write(resVals, outShape, x.dtype);
+  return { dataId: outId, shape: outShape, dtype: x.dtype };
+}
+var padV2Config = {
+  kernelName: PadV2,
+  backendName: "cpu",
+  kernelFunc: padV2
+};
+var powImpl = createSimpleBinaryKernelImpl((a, b) => Math.pow(a, b));
+var pow2 = binaryKernelFunc(Pow, powImpl);
+var powConfig = {
+  kernelName: Pow,
+  backendName: "cpu",
+  kernelFunc: pow2
+};
+function raggedGather2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;
+  const { outputRaggedRank } = attrs;
+  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.data.get(t.dataId).values);
+  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);
+  const $paramsDenseValues = backend2.data.get(paramsDenseValues.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImpl($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);
+  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], "int32", splits));
+  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);
+  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);
+}
+var raggedGatherConfig = {
+  kernelName: RaggedGather,
+  backendName: "cpu",
+  kernelFunc: raggedGather2
+};
+function raggedRange2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { starts, limits, deltas } = inputs;
+  const $starts = backend2.data.get(starts.dataId).values;
+  const $limits = backend2.data.get(limits.dataId).values;
+  const $deltas = backend2.data.get(deltas.dataId).values;
+  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImpl($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);
+  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], "int32", rtNestedSplitsData);
+  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);
+  return [rtNestedSplits, rtDenseValues];
+}
+var raggedRangeConfig = {
+  kernelName: RaggedRange,
+  backendName: "cpu",
+  kernelFunc: raggedRange2
+};
+function raggedTensorToTensor2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { shape, values, defaultValue, rowPartitionTensors } = inputs;
+  const { rowPartitionTypes } = attrs;
+  const $shape = backend2.data.get(shape.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const $defaultValue = backend2.data.get(defaultValue.dataId).values;
+  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.data.get(t.dataId).values);
+  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);
+  const [outputShape, output] = raggedTensorToTensorImpl($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);
+  return backend2.makeTensorInfo(outputShape, values.dtype, output);
+}
+var raggedTensorToTensorConfig = {
+  kernelName: RaggedTensorToTensor,
+  backendName: "cpu",
+  kernelFunc: raggedTensorToTensor2
+};
+function range3(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, dtype, step: step5 } = attrs;
+  const values = rangeImpl(start, stop, step5, dtype);
+  return backend2.makeTensorInfo([values.length], dtype, values);
+}
+var rangeConfig = {
+  kernelName: Range,
+  backendName: "cpu",
+  kernelFunc: range3
+};
+var reciprocal2 = unaryKernelFunc(Reciprocal, (xi) => 1 / xi);
+var reciprocalConfig = {
+  kernelName: Reciprocal,
+  backendName: "cpu",
+  kernelFunc: reciprocal2
+};
+function resizeBilinear2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  assertNotComplex(images, "resizeBilinear");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const xValues = backend2.data.get(images.dataId).values;
+  const result = new Float32Array(util_exports.sizeFromShape([batch, newHeight, newWidth, numChannels]));
+  const effectiveInputSize = [
+    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+  ];
+  const effectiveOutputSize = [
+    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+  ];
+  let outputIdx = 0;
+  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];
+  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];
+  for (let b = 0; b < batch; b++) {
+    for (let r = 0; r < newHeight; r++) {
+      let sourceFracRow;
+      if (halfPixelCenters) {
+        sourceFracRow = effectiveRowSizeRatio * (r + 0.5) - 0.5;
+      } else {
+        sourceFracRow = effectiveRowSizeRatio * r;
+      }
+      const sourceRowFloor = Math.max(0, Math.floor(sourceFracRow));
+      const rowFrac = sourceFracRow - sourceRowFloor;
+      const sourceRowCeil = Math.min(oldHeight - 1, Math.ceil(sourceFracRow));
+      const topRowOffset = b * imagesStrides[0] + sourceRowFloor * imagesStrides[1];
+      const botRowOffset = b * imagesStrides[0] + sourceRowCeil * imagesStrides[1];
+      for (let c = 0; c < newWidth; c++) {
+        let sourceFracCol;
+        if (halfPixelCenters) {
+          sourceFracCol = effectiveColSizeRatio * (c + 0.5) - 0.5;
+        } else {
+          sourceFracCol = effectiveColSizeRatio * c;
+        }
+        const sourceColFloor = Math.max(0, Math.floor(sourceFracCol));
+        const colFrac = sourceFracCol - sourceColFloor;
+        const sourceColCeil = Math.min(oldWidth - 1, Math.ceil(sourceFracCol));
+        const topLeftOffest = topRowOffset + sourceColFloor * imagesStrides[2];
+        const botLeftOffset = botRowOffset + sourceColFloor * imagesStrides[2];
+        const topRightOffset = topRowOffset + sourceColCeil * imagesStrides[2];
+        const botRightOffest = botRowOffset + sourceColCeil * imagesStrides[2];
+        for (let d = 0; d < numChannels; d++) {
+          const topLeft = xValues[topLeftOffest + d];
+          const bottomLeft = xValues[botLeftOffset + d];
+          const topRight = xValues[topRightOffset + d];
+          const bottomRight = xValues[botRightOffest + d];
+          const top = topLeft + (topRight - topLeft) * colFrac;
+          const bottom = bottomLeft + (bottomRight - bottomLeft) * colFrac;
+          const newValue = top + (bottom - top) * rowFrac;
+          result[outputIdx++] = newValue;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], "float32", result);
+}
+var resizeBilinearConfig = {
+  kernelName: ResizeBilinear,
+  backendName: "cpu",
+  kernelFunc: resizeBilinear2
+};
+function resizeBilinearGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  assertNotComplex([dy, images], "resizeBilinearGrad");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [batch, xHeight, xWidth, depth] = images.shape;
+  const [, yHeight, yWidth] = dy.shape;
+  const output = new Float32Array(batch * xHeight * xWidth * depth);
+  const effectiveXSize = [
+    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+  ];
+  const effectiveYSize = [
+    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+  ];
+  const heightScale = effectiveXSize[0] / effectiveYSize[0];
+  const widthScale = effectiveXSize[1] / effectiveYSize[1];
+  const dyValues = backend2.data.get(dy.dataId).values;
+  let offset = 0;
+  for (let b = 0; b < batch; b++) {
+    const bOffset = b * imagesStrides[0];
+    for (let r = 0; r < yHeight; r++) {
+      const dxR = r * heightScale;
+      const topDxRIndex = Math.floor(dxR);
+      const bottomDxRIndex = Math.min(Math.ceil(dxR), xHeight - 1);
+      const topDxROffset = bOffset + topDxRIndex * imagesStrides[1];
+      const bottomDxROffset = bOffset + bottomDxRIndex * imagesStrides[1];
+      const dxRLerp = dxR - topDxRIndex;
+      const inverseDxRLerp = 1 - dxRLerp;
+      for (let c = 0; c < yWidth; c++) {
+        const dxC = c * widthScale;
+        const leftDxCIndex = Math.floor(dxC);
+        const rightDxCIndex = Math.min(Math.ceil(dxC), xWidth - 1);
+        const dxCLerp = dxC - leftDxCIndex;
+        const inverseDxCLerp = 1 - dxCLerp;
+        const topLeftRCOffset = topDxROffset + leftDxCIndex * imagesStrides[2];
+        const topRightRCOffset = topDxROffset + rightDxCIndex * imagesStrides[2];
+        const bottomLeftRCOffset = bottomDxROffset + leftDxCIndex * imagesStrides[2];
+        const bottomRightRCOffset = bottomDxROffset + rightDxCIndex * imagesStrides[2];
+        const inverseDxRLerpTimesInverseDxCLerp = inverseDxRLerp * inverseDxCLerp;
+        const inverseDxRLerpTimesDxCLerp = inverseDxRLerp * dxCLerp;
+        const dxRLerpTimesInverseDxCLerp = dxRLerp * inverseDxCLerp;
+        const dxRLerpTimesDxCLerp = dxRLerp * dxCLerp;
+        for (let d = 0; d < depth; d++) {
+          const dyVal = dyValues[offset++];
+          output[topLeftRCOffset + d] += dyVal * inverseDxRLerpTimesInverseDxCLerp;
+          output[topRightRCOffset + d] += dyVal * inverseDxRLerpTimesDxCLerp;
+          output[bottomLeftRCOffset + d] += dyVal * dxRLerpTimesInverseDxCLerp;
+          output[bottomRightRCOffset + d] += dyVal * dxRLerpTimesDxCLerp;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, xWidth, xHeight, depth], "float32", output);
+}
+var resizeBilinearGradConfig2 = {
+  kernelName: ResizeBilinearGrad,
+  backendName: "cpu",
+  kernelFunc: resizeBilinearGrad
+};
+function resizeNearestNeighbor2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  assertNotComplex(images, "resizeNearestNeighbor");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const xValues = backend2.data.get(images.dataId).values;
+  const output = new Float32Array(batch * newHeight * newWidth * numChannels);
+  const effectiveInputSize = [
+    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+  ];
+  const effectiveOutputSize = [
+    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+  ];
+  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];
+  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];
+  let outputOffset = 0;
+  for (let b = 0; b < batch; b++) {
+    const batchOffset = b * imagesStrides[0];
+    for (let r = 0; r < newHeight; r++) {
+      const sourceFracRow = halfPixelCenters ? effectiveRowSizeRatio * (r + 0.5) : effectiveRowSizeRatio * r;
+      let sourceNearestRow = Math.min(oldHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));
+      if (halfPixelCenters) {
+        sourceNearestRow = Math.max(0, sourceNearestRow);
+      }
+      const rowOffset = batchOffset + sourceNearestRow * imagesStrides[1];
+      for (let c = 0; c < newWidth; c++) {
+        const sourceFracCol = halfPixelCenters ? effectiveColSizeRatio * (c + 0.5) : effectiveColSizeRatio * c;
+        let sourceNearestCol = Math.min(oldWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));
+        if (halfPixelCenters) {
+          sourceNearestCol = Math.max(0, sourceNearestCol);
+        }
+        const colOffset = rowOffset + sourceNearestCol * imagesStrides[2];
+        for (let d = 0; d < numChannels; d++) {
+          const newVal = xValues[colOffset + d];
+          output[outputOffset++] = newVal;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], images.dtype, output);
+}
+var resizeNearestNeighborConfig = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "cpu",
+  kernelFunc: resizeNearestNeighbor2
+};
+function resizeNearestNeighborGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  assertNotComplex([dy, images], "resizeNearestNeighborGrad");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const [batch, xHeight, xWidth, depth] = images.shape;
+  const [, yHeight, yWidth] = dy.shape;
+  const output = new Float32Array(batch * xHeight * xWidth * depth);
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const effectiveXSize = [
+    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+  ];
+  const effectiveYSize = [
+    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+  ];
+  const heightScale = effectiveXSize[0] / effectiveYSize[0];
+  const widthScale = effectiveXSize[1] / effectiveYSize[1];
+  const invHeightScale = 1 / heightScale;
+  const invWidthScale = 1 / widthScale;
+  const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+  const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+  for (let b = 0; b < batch; b++) {
+    const batchOffset = b * imagesStrides[0];
+    for (let r = 0; r < xHeight; r++) {
+      const rowOffset = batchOffset + r * imagesStrides[1];
+      const startRLerp = Math.floor(r * invHeightScale);
+      const startDyR = Math.floor(startRLerp - winHeight / 2);
+      for (let c = 0; c < xWidth; c++) {
+        const colOffset = rowOffset + c * imagesStrides[2];
+        const startCLerp = Math.floor(c * invWidthScale);
+        const startDyC = Math.floor(startCLerp - winWidth / 2);
+        for (let d = 0; d < depth; d++) {
+          let accum = 0;
+          for (let dyRIndex = 0; dyRIndex < winHeight; dyRIndex++) {
+            const dyR = dyRIndex + startDyR;
+            if (dyR < 0 || dyR >= yHeight) {
+              continue;
+            }
+            const dyROffset = batchOffset + dyR * dyStrides[1];
+            const sourceFracRow = dyR * heightScale;
+            const sourceNearestRow = Math.min(xHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));
+            if (r !== sourceNearestRow) {
+              continue;
+            }
+            for (let dyCIndex = 0; dyCIndex < winWidth; dyCIndex++) {
+              const dyC = dyCIndex + startDyC;
+              if (dyC < 0 || dyC >= yWidth) {
+                continue;
+              }
+              const dyCOffset = dyROffset + dyC * dyStrides[2];
+              const sourceFracCol = dyC * widthScale;
+              const sourceNearestCol = Math.min(xWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));
+              if (c === sourceNearestCol) {
+                accum += dyValues[dyCOffset + d];
+              }
+            }
+          }
+          output[colOffset + d] = accum;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(images.shape, images.dtype, output);
+}
+var resizeNearestNeighborGradConfig2 = {
+  kernelName: ResizeNearestNeighborGrad,
+  backendName: "cpu",
+  kernelFunc: resizeNearestNeighborGrad
+};
+function reverse2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  assertNotComplex(x, "reverse");
+  const xRank = x.shape.length;
+  const $dims = util_exports.parseAxisParam(dims, x.shape);
+  if (xRank === 0) {
+    return identity2({ inputs: { x }, backend: backend2 });
+  }
+  const outBuf = new TensorBuffer(x.shape, x.dtype);
+  const xBuf = backend2.bufferSync(x);
+  for (let i = 0; i < outBuf.size; i++) {
+    const outLoc = outBuf.indexToLoc(i);
+    const inLoc = outLoc.slice();
+    $dims.forEach((d) => inLoc[d] = x.shape[d] - 1 - inLoc[d]);
+    outBuf.set(xBuf.get(...inLoc), ...outLoc);
+  }
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var reverseConfig = {
+  kernelName: Reverse,
+  backendName: "cpu",
+  kernelFunc: reverse2
+};
+var rotateWithOffsetConfig = {
+  kernelName: RotateWithOffset,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const { radians, fillValue, center } = attrs;
+    const cpuBackend = backend2;
+    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));
+    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+    const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);
+    const fullOpacityValue = 255;
+    const sinFactor = Math.sin(radians);
+    const cosFactor = Math.cos(radians);
+    const imageVals = cpuBackend.data.get(image2.dataId).values;
+    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {
+      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;
+      for (let row = 0; row < imageHeight; row++) {
+        const rowOffset = row * (imageWidth * numChannels);
+        for (let col = 0; col < imageWidth; col++) {
+          const colOffset = col * numChannels;
+          for (let channel = 0; channel < numChannels; channel++) {
+            const coords2 = [batch, row, col, channel];
+            const x = coords2[2];
+            const y = coords2[1];
+            let coordX = (x - centerX) * cosFactor - (y - centerY) * sinFactor;
+            let coordY = (x - centerX) * sinFactor + (y - centerY) * cosFactor;
+            coordX = Math.round(coordX + centerX);
+            coordY = Math.round(coordY + centerY);
+            let outputValue = fillValue;
+            if (typeof fillValue !== "number") {
+              if (channel === 3) {
+                outputValue = fullOpacityValue;
+              } else {
+                outputValue = fillValue[channel];
+              }
+            }
+            if (coordX >= 0 && coordX < imageWidth && coordY >= 0 && coordY < imageHeight) {
+              const rotatedRowOffset = coordY * (imageWidth * numChannels);
+              const rotatedColOffset = coordX * numChannels;
+              const imageIdx = batchOffset + rotatedRowOffset + rotatedColOffset + channel;
+              outputValue = imageVals[imageIdx];
+            }
+            const outIdx = batchOffset + rowOffset + colOffset + channel;
+            output[outIdx] = outputValue;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);
+    return { dataId, shape: image2.shape, dtype: image2.dtype };
+  }
+};
+var round3 = unaryKernelFunc(Round, (xi) => {
+  const base = Math.floor(xi);
+  if (xi - base < 0.5) {
+    return Math.floor(xi);
+  } else if (xi - base > 0.5) {
+    return Math.ceil(xi);
+  } else {
+    if (base % 2 === 0) {
+      return base;
+    } else {
+      return base + 1;
+    }
+  }
+});
+var roundConfig = {
+  kernelName: Round,
+  backendName: "cpu",
+  kernelFunc: round3
+};
+function scatterNd(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);
+  const sumDupeIndices = true;
+  const indicesBuf = backend2.bufferSync(indices);
+  const updatesBuf = backend2.bufferSync(updates);
+  const outBuf = scatterImpl(indicesBuf, updatesBuf, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, 0, sumDupeIndices);
+  return backend2.makeTensorInfo(shape, outBuf.dtype, outBuf.values);
+}
+var scatterNdConfig = {
+  kernelName: ScatterNd,
+  backendName: "cpu",
+  kernelFunc: scatterNd
+};
+function lowerBound2(array2, value) {
+  let left = 0;
+  let right = array2.length;
+  let mid = 0;
+  while (left < right) {
+    mid = Math.floor((left + right) / 2);
+    if (array2[mid] < value) {
+      left = mid + 1;
+    } else {
+      right = mid;
+    }
+  }
+  return right;
+}
+function upperBound2(array2, value) {
+  let left = 0;
+  let right = array2.length;
+  let mid = 0;
+  while (left < right) {
+    mid = Math.floor((left + right) / 2);
+    if (array2[mid] <= value) {
+      left = mid + 1;
+    } else {
+      right = mid;
+    }
+  }
+  return right;
+}
+function searchSortedImpl(sortedInputs, values, batchSize, numInputs, numValues, side) {
+  const output = util_exports.getArrayFromDType("int32", batchSize * numValues);
+  for (let b = 0; b < batchSize; ++b) {
+    const sortedInputsSlice = sortedInputs.slice(b * numInputs, (b + 1) * numInputs);
+    const outputOffset = b * numValues;
+    for (let i = 0; i < numValues; ++i) {
+      output[outputOffset + i] = side === "left" ? lowerBound2(sortedInputsSlice, values[i + outputOffset]) : upperBound2(sortedInputsSlice, values[i + outputOffset]);
+    }
+  }
+  return output;
+}
+function searchSorted2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sortedSequence, values } = inputs;
+  const { side } = attrs;
+  const $sortedSequence = backend2.data.get(sortedSequence.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const output = searchSortedImpl($sortedSequence, $values, sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);
+  return backend2.makeTensorInfo(values.shape, "int32", output);
+}
+var searchSortedConfig = {
+  kernelName: SearchSorted,
+  backendName: "cpu",
+  kernelFunc: searchSorted2
+};
+function select2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  assertNotComplex([condition, t, e], "select");
+  const conditionRank = condition.shape.length;
+  const values = backend2.data.get(condition.dataId).values;
+  const tValues = backend2.data.get(t.dataId).values;
+  const eValues = backend2.data.get(e.dataId).values;
+  const resultDtype = upcastType(t.dtype, e.dtype);
+  const newValues = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(t.shape), resultDtype);
+  let index = 0;
+  const offset = conditionRank === 0 || conditionRank > 1 || t.shape.length === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));
+  for (let i = 0; i < values.length; i++) {
+    for (let j = 0; j < offset; j++) {
+      if (values[i] === 1) {
+        newValues[index++] = tValues[i];
+      } else {
+        newValues[index++] = eValues[i];
+      }
+    }
+  }
+  return backend2.makeTensorInfo(t.shape, resultDtype, newValues);
+}
+var selectConfig = {
+  kernelName: Select,
+  backendName: "cpu",
+  kernelFunc: select2
+};
+var scaleAlpha = backend_util_exports.SELU_SCALEALPHA;
+var scale = backend_util_exports.SELU_SCALE;
+var selu2 = unaryKernelFunc(Selu, (xi) => {
+  if (xi >= 0) {
+    return scale * xi;
+  } else {
+    return scaleAlpha * (Math.exp(xi) - 1);
+  }
+});
+var seluConfig = {
+  kernelName: Selu,
+  backendName: "cpu",
+  kernelFunc: selu2
+};
+var sign2 = unaryKernelFunc(Sign, (xi) => {
+  if (xi < 0) {
+    return -1;
+  } else if (xi > 0) {
+    return 1;
+  } else {
+    return 0;
+  }
+});
+var signConfig = {
+  kernelName: Sign,
+  backendName: "cpu",
+  kernelFunc: sign2
+};
+var sin2 = unaryKernelFunc(Sin, (xi) => Math.sin(xi));
+var sinConfig = {
+  kernelName: Sin,
+  backendName: "cpu",
+  kernelFunc: sin2
+};
+var sinh2 = unaryKernelFunc(Sinh, (xi) => Math.sinh(xi));
+var sinhConfig = {
+  kernelName: Sinh,
+  backendName: "cpu",
+  kernelFunc: sinh2
+};
+var epsilon2 = 11920928955078125e-23;
+var threshold2 = Math.log(epsilon2) + 2;
+var softplus2 = unaryKernelFunc(Softplus, (xi) => {
+  const tooLarge = xi > -threshold2;
+  const tooSmall = xi < threshold2;
+  const expX = Math.exp(xi);
+  let result;
+  if (tooSmall) {
+    result = expX;
+  } else if (tooLarge) {
+    result = xi;
+  } else {
+    result = Math.log(1 + expX);
+  }
+  return result;
+});
+var softplusConfig = {
+  kernelName: Softplus,
+  backendName: "cpu",
+  kernelFunc: softplus2
+};
+function spaceToBatchND2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  assertNotComplex([x], "spaceToBatchND");
+  const prod5 = util_exports.sizeFromShape(blockShape);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const paddedX = padV2Config.kernelFunc({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapeInputs = { x: paddedX };
+  const reshapeAttrs = { shape: reshapedPaddedShape };
+  const paddedXReshaped = reshape3({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });
+  const transposeInputs = { x: paddedXReshaped };
+  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };
+  const paddedXT = transpose2({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });
+  const resultReshapeInputs = { x: paddedXT };
+  const resultReshapeAttrs = { shape: flattenShape };
+  const result = reshape3({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });
+  backend2.disposeIntermediateTensorInfo(paddedX);
+  backend2.disposeIntermediateTensorInfo(paddedXReshaped);
+  backend2.disposeIntermediateTensorInfo(paddedXT);
+  return result;
+}
+var spaceToBatchNDConfig = {
+  kernelName: SpaceToBatchND,
+  backendName: "cpu",
+  kernelFunc: spaceToBatchND2
+};
+function sparseFillEmptyRows2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  if (denseShape.shape.length !== 1) {
+    throw new Error(`Dense shape must be a vector, saw:
+        ${denseShape.shape}`);
+  }
+  if (indices.shape.length !== 2) {
+    throw new Error(`Indices must be a matrix, saw:
+        ${indices.shape}`);
+  }
+  if (values.shape.length !== 1) {
+    throw new Error(`Values must be a vector, saw:
+        ${values.shape}`);
+  }
+  if (defaultValue.shape.length !== 0) {
+    throw new Error(`Default value must be a scalar, saw:
+        ${defaultValue.shape}`);
+  }
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const $denseShape = backend2.data.get(denseShape.dataId).values;
+  const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImpl($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);
+  return [
+    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),
+    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),
+    backend2.makeTensorInfo([emptyRowIndicator.length], "bool", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),
+    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))
+  ];
+}
+var sparseFillEmptyRowsConfig = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "cpu",
+  kernelFunc: sparseFillEmptyRows2
+};
+function sparseReshape2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape
+        ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape
+        ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const $inputShape = Array.from(backend2.data.get(inputShape.dataId).values);
+  const $inputIndices = backend2.data.get(inputIndices.dataId).values;
+  const targetShape = Array.from(backend2.data.get(newShape.dataId).values);
+  const [newIndices, indicesShape, outputShape] = sparseReshapeImpl($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);
+  return [
+    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),
+    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))
+  ];
+}
+var sparseReshapeConfig = {
+  kernelName: SparseReshape,
+  backendName: "cpu",
+  kernelFunc: sparseReshape2
+};
+function sparseSegmentMean2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+          ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+          ${segmentIds.shape}`);
+  }
+  if (indices.shape[0] !== segmentIds.shape[0]) {
+    throw new Error(`segmentIds and indices should have same size.`);
+  }
+  const $data = backend2.data.get(data.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $segmentIds = backend2.data.get(segmentIds.dataId).values;
+  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds, true);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentMeanConfig = {
+  kernelName: SparseSegmentMean,
+  backendName: "cpu",
+  kernelFunc: sparseSegmentMean2
+};
+function sparseSegmentSum2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+         ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+         ${segmentIds.shape}`);
+  }
+  if (indices.shape[0] !== segmentIds.shape[0]) {
+    throw new Error(`segmentIds and indices should have same size.`);
+  }
+  const $data = backend2.data.get(data.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $segmentIds = backend2.data.get(segmentIds.dataId).values;
+  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentSumConfig = {
+  kernelName: SparseSegmentSum,
+  backendName: "cpu",
+  kernelFunc: sparseSegmentSum2
+};
+function sparseToDense2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sparseIndices, sparseValues, defaultValue } = inputs;
+  const { outputShape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);
+  const sumDupeIndices = false;
+  const indicesBuf = backend2.bufferSync(sparseIndices);
+  let outBuf;
+  switch (sparseValues.dtype) {
+    case "bool": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = Boolean(backend2.data.get(defaultValue.dataId).values[0]);
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "float32": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "int32": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "string": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = util_exports.decodeString(backend2.data.get(defaultValue.dataId).values[0]);
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    default:
+      throw new Error(`Unsupported type ${sparseValues.dtype}`);
+  }
+  return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);
+}
+var sparseToDenseConfig = {
+  kernelName: SparseToDense,
+  backendName: "cpu",
+  kernelFunc: sparseToDense2
+};
+function splitV(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const begin = new Array(x.shape.length).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const sliceSize = [...size];
+    sliceSize[$axis] = s;
+    const sliceT = slice2({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });
+    begin[$axis] += s;
+    return sliceT;
+  });
+}
+var splitVConfig = {
+  kernelName: SplitV,
+  backendName: "cpu",
+  kernelFunc: splitV
+};
+var squareConfig = {
+  kernelName: Square,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2 }) => {
+    const { x } = inputs;
+    const cpuBackend = backend2;
+    assertNotComplex(x, "square");
+    const values = cpuBackend.data.get(x.dataId).values;
+    const newValues = new Float32Array(values.length);
+    for (let i = 0; i < values.length; ++i) {
+      const value = values[i];
+      newValues[i] = value * value;
+    }
+    const dataId = cpuBackend.write(newValues, x.shape, x.dtype);
+    return { dataId, shape: x.shape, dtype: x.dtype };
+  }
+};
+var step2 = unaryKernelFunc(Step, (xi, attrs) => {
+  const stepAttrs = attrs;
+  if (isNaN(xi)) {
+    return NaN;
+  } else {
+    return xi > 0 ? 1 : stepAttrs.alpha;
+  }
+});
+var stepConfig = {
+  kernelName: Step,
+  backendName: "cpu",
+  kernelFunc: step2
+};
+function stridedSlice2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  assertNotComplex(x, "stridedSlice");
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice2({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape3({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeIntermediateTensorInfo(sliced);
+  } else {
+    const xBuf = backend2.bufferSync(x);
+    const outBuf = stridedSliceImpl(finalShapeSparse, xBuf, $strides, $begin);
+    result = backend2.makeTensorInfo(finalShape, outBuf.dtype, outBuf.values);
+  }
+  return result;
+}
+var stridedSliceConfig = {
+  kernelName: StridedSlice,
+  backendName: "cpu",
+  kernelFunc: stridedSlice2
+};
+function stringNGrams2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const { data, dataSplits } = inputs;
+  const $data = backend2.data.get(data.dataId).values;
+  const $dataSplits = backend2.data.get(dataSplits.dataId).values;
+  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  return [
+    backend2.makeTensorInfo([nGrams.length], "string", nGrams),
+    backend2.makeTensorInfo(dataSplits.shape, "int32", nGramsSplits)
+  ];
+}
+var stringNGramsConfig = {
+  kernelName: StringNGrams,
+  backendName: "cpu",
+  kernelFunc: stringNGrams2
+};
+function stringSplit2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { skipEmpty } = attrs;
+  const { input: input2, delimiter } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (input2.shape.length !== 1) {
+    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);
+  }
+  if (delimiter.shape.length !== 0) {
+    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);
+  }
+  const $input = backend2.data.get(input2.dataId).values;
+  const $delimiter = backend2.data.get(delimiter.dataId).values[0];
+  const [indices, values, shape] = stringSplitImpl($input, $delimiter, skipEmpty);
+  const outputSize = values.length;
+  return [
+    backend2.makeTensorInfo([outputSize, 2], "int32", indices),
+    backend2.makeTensorInfo([outputSize], "string", values),
+    backend2.makeTensorInfo([2], "int32", new Int32Array(shape))
+  ];
+}
+var stringSplitConfig = {
+  kernelName: StringSplit,
+  backendName: "cpu",
+  kernelFunc: stringSplit2
+};
+function stringToHashBucketFast2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { numBuckets } = attrs;
+  const { input: input2 } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const $input = backend2.data.get(input2.dataId).values;
+  const output = stringToHashBucketFastImpl($input, numBuckets);
+  return backend2.makeTensorInfo(input2.shape, "int32", output);
+}
+var stringToHashBucketFastConfig = {
+  kernelName: StringToHashBucketFast,
+  backendName: "cpu",
+  kernelFunc: stringToHashBucketFast2
+};
+var tan2 = unaryKernelFunc(Tan, (xi) => Math.tan(xi));
+var tanConfig = {
+  kernelName: Tan,
+  backendName: "cpu",
+  kernelFunc: tan2
+};
+var tanh3 = unaryKernelFunc(Tanh, (xi) => Math.tanh(xi));
+var tanhConfig = {
+  kernelName: Tanh,
+  backendName: "cpu",
+  kernelFunc: tanh3
+};
+function tile3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reps } = attrs;
+  assertNotComplex(x, "tile");
+  const outBuf = tileImpl(backend2.bufferSync(x), reps);
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var tileConfig = {
+  kernelName: Tile,
+  backendName: "cpu",
+  kernelFunc: tile3
+};
+function topK(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  assertNotComplex(x, "topk");
+  const xVals = backend2.data.get(x.dataId).values;
+  const [allTopKVals, allTopKIndices] = topKImpl(xVals, x.shape, x.dtype, k, sorted);
+  return [
+    backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),
+    backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)
+  ];
+}
+var topKConfig = {
+  kernelName: TopK,
+  backendName: "cpu",
+  kernelFunc: topK
+};
+function transform2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [batch, outHeight, outWidth, numChannels];
+  const inStrides = util_exports.computeStrides(image2.shape);
+  const batchInStride = inStrides[0];
+  const rowInStride = inStrides[1];
+  const colInStride = inStrides[2];
+  const outStrides = util_exports.computeStrides(outShape);
+  const batchOutStride = outStrides[0];
+  const rowOutStride = outStrides[1];
+  const colOutStride = outStrides[2];
+  const outVals = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(outShape));
+  outVals.fill(fillValue);
+  const imageVals = backend2.data.get(image2.dataId).values;
+  const transformVals = backend2.data.get(transforms.dataId).values;
+  for (let b = 0; b < batch; ++b) {
+    const transform5 = transforms.shape[0] === 1 ? transformVals : transformVals.subarray(b * 8, b * 8 + 8);
+    for (let outY = 0; outY < outHeight; ++outY) {
+      for (let outX = 0; outX < outWidth; ++outX) {
+        for (let channel = 0; channel < numChannels; ++channel) {
+          let val;
+          const projection = transform5[6] * outX + transform5[7] * outY + 1;
+          if (projection === 0) {
+            continue;
+          }
+          const inX = (transform5[0] * outX + transform5[1] * outY + transform5[2]) / projection;
+          const inY = (transform5[3] * outX + transform5[4] * outY + transform5[5]) / projection;
+          const x = mapCoord(inX, imageWidth, fillMode);
+          const y = mapCoord(inY, imageHeight, fillMode);
+          switch (interpolation) {
+            case "nearest":
+              val = nearestInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);
+              break;
+            case "bilinear":
+              val = bilinearInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);
+              break;
+            default:
+              throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${interpolation}`);
+          }
+          const ind = b * batchOutStride + outY * rowOutStride + outX * colOutStride + channel;
+          outVals[ind] = val;
+        }
+      }
+    }
+    return backend2.makeTensorInfo(outShape, image2.dtype, outVals);
+  }
+  const dataId = backend2.write(outVals, outShape, image2.dtype);
+  return { dataId, shape: image2.shape, dtype: image2.dtype };
+}
+var transformConfig = {
+  kernelName: Transform,
+  backendName: "cpu",
+  kernelFunc: transform2
+};
+function mapCoord(outCoord, len, mode) {
+  switch (mode) {
+    case "reflect":
+      return mapCoordReflect(outCoord, len);
+    case "wrap":
+      return mapCoordWrap(outCoord, len);
+    case "nearest":
+      return mapCoordNearest(outCoord, len);
+    case "constant":
+    default:
+      return mapCoordConstant(outCoord, len);
+  }
+}
+function mapCoordReflect(outCoord, len) {
+  let inCoord = outCoord;
+  if (inCoord < 0) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz2 = 2 * len;
+      if (inCoord < sz2) {
+        inCoord = sz2 * Math.trunc(-inCoord / sz2) + inCoord;
+      }
+      inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1;
+    }
+  } else if (inCoord > len - 1) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz2 = 2 * len;
+      inCoord -= sz2 * Math.trunc(inCoord / sz2);
+      if (inCoord >= len) {
+        inCoord = sz2 - inCoord - 1;
+      }
+    }
+  }
+  return util_exports.clamp(0, inCoord, len - 1);
+}
+function mapCoordWrap(outCoord, len) {
+  let inCoord = outCoord;
+  if (inCoord < 0) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz = len - 1;
+      inCoord += len * (Math.trunc(-inCoord / sz) + 1);
+    }
+  } else if (inCoord > len - 1) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz = len - 1;
+      inCoord -= len * Math.trunc(inCoord / sz);
+    }
+  }
+  return util_exports.clamp(0, inCoord, len - 1);
+}
+function mapCoordConstant(outCoord, len) {
+  return outCoord;
+}
+function mapCoordNearest(outCoord, len) {
+  return util_exports.clamp(0, outCoord, len - 1);
+}
+function readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const ind = batch * batchStride + y * rowStride + x * colStride + channel;
+  if (0 <= y && y < imageHeight && 0 <= x && x < imageWidth) {
+    return imageVals[ind];
+  } else {
+    return fillValue;
+  }
+}
+function nearestInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const $y = Math.round(y);
+  const $x = Math.round(x);
+  return readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, $y, $x, channel, fillValue);
+}
+function bilinearInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const yFloor = Math.floor(y);
+  const xFloor = Math.floor(x);
+  const yCeil = yFloor + 1;
+  const xCeil = xFloor + 1;
+  const valueYFloor = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xCeil, channel, fillValue);
+  const valueYCeil = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xCeil, channel, fillValue);
+  return (yCeil - y) * valueYFloor + (y - yFloor) * valueYCeil;
+}
+function unique3(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  assertNotComplex(x, "unique");
+  const values = backend2.data.get(x.dataId).values;
+  const { outputValues, outputShape, indices } = uniqueImpl(values, axis, x.shape, x.dtype);
+  return [
+    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),
+    backend2.makeTensorInfo([indices.length], "int32", indices)
+  ];
+}
+var uniqueConfig = {
+  kernelName: Unique,
+  backendName: "cpu",
+  kernelFunc: unique3
+};
+function unpack(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const valueRank = value.shape.length;
+  const num = value.shape[axis];
+  const outShape = new Array(valueRank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < valueRank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = value.shape[i];
+    }
+  }
+  const begin = new Array(valueRank).fill(0);
+  const size = value.shape.slice();
+  size[axis] = 1;
+  const res = new Array(num);
+  for (let i = 0; i < res.length; i++) {
+    begin[axis] = i;
+    const tempRes = slice2({ inputs: { x: value }, backend: backend2, attrs: { begin, size } });
+    res[i] = reshape3({ inputs: { x: tempRes }, backend: backend2, attrs: { shape: outShape } });
+    backend2.disposeIntermediateTensorInfo(tempRes);
+  }
+  return res;
+}
+var unpackConfig = {
+  kernelName: Unpack,
+  backendName: "cpu",
+  kernelFunc: unpack
+};
+function unsortedSegmentSum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, segmentIds } = inputs;
+  const { numSegments } = attrs;
+  assertNotComplex(x, "unsortedSegmentSum");
+  const xRank = x.shape.length;
+  const segmentIdsRank = segmentIds.shape.length;
+  const res = [];
+  const intermediates = [];
+  const numIters = xRank - segmentIdsRank;
+  let $segmentIds = segmentIds;
+  for (let i = 0; i < numIters; ++i) {
+    const expanded = expandDims3({ inputs: { input: $segmentIds }, backend: backend2, attrs: { dim: i + 1 } });
+    $segmentIds = expanded;
+    intermediates.push(expanded);
+  }
+  for (let i = 0; i < numSegments; ++i) {
+    const scalarValue = util_exports.createScalarValue(i, "int32");
+    const segmentId = backend2.makeTensorInfo([], "int32", scalarValue);
+    const mask = equal2({ inputs: { a: segmentId, b: $segmentIds }, backend: backend2 });
+    const maskCasted = cast3({ inputs: { x: mask }, backend: backend2, attrs: { dtype: "float32" } });
+    const mul2 = multiply2({ inputs: { a: maskCasted, b: x }, backend: backend2 });
+    const sumTensorInfo = sum3({ inputs: { x: mul2 }, backend: backend2, attrs: { axis: 0, keepDims: false } });
+    res.push(sumTensorInfo);
+    intermediates.push(segmentId);
+    intermediates.push(mask);
+    intermediates.push(maskCasted);
+    intermediates.push(mul2);
+    intermediates.push(sumTensorInfo);
+  }
+  const result = pack({ inputs: res, backend: backend2, attrs: { axis: 0 } });
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var unsortedSegmentSumConfig = {
+  kernelName: UnsortedSegmentSum,
+  backendName: "cpu",
+  kernelFunc: unsortedSegmentSum2
+};
+var kernelConfigs = [
+  _fusedMatMulConfig,
+  absConfig,
+  acosConfig,
+  acoshConfig,
+  addConfig,
+  addNConfig,
+  allConfig,
+  anyConfig,
+  argMaxConfig,
+  argMinConfig,
+  asinConfig,
+  asinhConfig,
+  atanConfig,
+  atan2Config,
+  atanhConfig,
+  avgPoolConfig,
+  avgPool3DConfig,
+  avgPool3DGradConfig2,
+  avgPoolGradConfig2,
+  batchMatMulConfig,
+  batchNormConfig,
+  batchToSpaceNDConfig,
+  bincountConfig,
+  broadcastArgsConfig,
+  castConfig,
+  ceilConfig,
+  clipByValueConfig,
+  complexConfig,
+  complexAbsConfig,
+  concatConfig,
+  conv2DConfig,
+  conv2DBackpropFilterConfig,
+  conv2DBackpropInputConfig,
+  conv3DConfig,
+  conv3DBackpropFilterV2Config,
+  conv3DBackpropInputV2Config,
+  cosConfig,
+  coshConfig,
+  cropAndResizeConfig,
+  cumprodConfig,
+  cumsumConfig,
+  denseBincountConfig,
+  depthToSpaceConfig,
+  depthwiseConv2dNativeConfig,
+  depthwiseConv2dNativeBackpropFilterConfig,
+  depthwiseConv2dNativeBackpropInputConfig,
+  diagConfig,
+  dilation2DConfig,
+  dilation2DBackpropFilterConfig,
+  dilation2DBackpropInputConfig,
+  einsumConfig,
+  eluConfig,
+  eluGradConfig2,
+  equalConfig,
+  erfConfig,
+  expConfig,
+  expandDimsConfig,
+  expm1Config,
+  fftConfig,
+  fillConfig,
+  flipLeftRightConfig,
+  floorConfig,
+  floorDivConfig,
+  fusedConv2DConfig,
+  fusedDepthwiseConv2DConfig,
+  gatherNdConfig,
+  gatherV2Config,
+  greaterConfig,
+  greaterEqualConfig,
+  identityConfig,
+  ifftConfig,
+  imagConfig,
+  isFiniteConfig,
+  isInfConfig,
+  isNaNConfig,
+  leakyReluConfig,
+  lessConfig,
+  lessEqualConfig,
+  linSpaceConfig,
+  logConfig,
+  log1pConfig,
+  logicalAndConfig,
+  logicalNotConfig,
+  logicalOrConfig,
+  LRNConfig,
+  LRNGradConfig,
+  maxConfig,
+  maximumConfig,
+  maxPoolConfig,
+  maxPool3DConfig,
+  maxPool3DGradConfig2,
+  maxPoolGradConfig2,
+  maxPoolWithArgmaxConfig,
+  meanConfig,
+  minConfig,
+  minimumConfig,
+  mirrorPadConfig,
+  modConfig,
+  multinomialConfig,
+  multiplyConfig,
+  negConfig,
+  nonMaxSuppressionV3Config,
+  nonMaxSuppressionV4Config,
+  nonMaxSuppressionV5Config,
+  notEqualConfig,
+  oneHotConfig,
+  onesLikeConfig,
+  packConfig,
+  padV2Config,
+  powConfig,
+  preluConfig,
+  prodConfig,
+  raggedGatherConfig,
+  raggedRangeConfig,
+  raggedTensorToTensorConfig,
+  rangeConfig,
+  realConfig,
+  realDivConfig,
+  reciprocalConfig,
+  reluConfig,
+  relu6Config,
+  reshapeConfig,
+  resizeBilinearConfig,
+  resizeBilinearGradConfig2,
+  resizeNearestNeighborConfig,
+  resizeNearestNeighborGradConfig2,
+  reverseConfig,
+  rotateWithOffsetConfig,
+  roundConfig,
+  rsqrtConfig,
+  scatterNdConfig,
+  searchSortedConfig,
+  selectConfig,
+  seluConfig,
+  sigmoidConfig,
+  signConfig,
+  sinConfig,
+  sinhConfig,
+  sliceConfig,
+  softmaxConfig,
+  softplusConfig,
+  spaceToBatchNDConfig,
+  sparseFillEmptyRowsConfig,
+  sparseReshapeConfig,
+  sparseSegmentMeanConfig,
+  sparseSegmentSumConfig,
+  sparseToDenseConfig,
+  splitVConfig,
+  sqrtConfig,
+  squareConfig,
+  squaredDifferenceConfig,
+  stepConfig,
+  stridedSliceConfig,
+  stringNGramsConfig,
+  stringSplitConfig,
+  stringToHashBucketFastConfig,
+  subConfig,
+  sumConfig,
+  tanConfig,
+  tanhConfig,
+  tileConfig,
+  topKConfig,
+  transformConfig,
+  transposeConfig,
+  uniqueConfig,
+  unpackConfig,
+  unsortedSegmentSumConfig,
+  zerosLikeConfig
+];
+for (const kernelConfig of kernelConfigs) {
+  registerKernel(kernelConfig);
+}
+var webgl_util_exports = {};
+__export2(webgl_util_exports, {
+  assertNotComplex: () => assertNotComplex2,
+  bindCanvasToFramebuffer: () => bindCanvasToFramebuffer,
+  bindColorTextureToFramebuffer: () => bindColorTextureToFramebuffer,
+  bindTextureToProgramUniformSampler: () => bindTextureToProgramUniformSampler,
+  bindTextureUnit: () => bindTextureUnit,
+  bindVertexBufferToProgramAttribute: () => bindVertexBufferToProgramAttribute,
+  callAndCheck: () => callAndCheck,
+  canBeRepresented: () => canBeRepresented,
+  createFragmentShader: () => createFragmentShader,
+  createFramebuffer: () => createFramebuffer,
+  createProgram: () => createProgram,
+  createStaticIndexBuffer: () => createStaticIndexBuffer,
+  createStaticVertexBuffer: () => createStaticVertexBuffer,
+  createTexture: () => createTexture,
+  createVertexShader: () => createVertexShader,
+  getBatchDim: () => getBatchDim,
+  getExtensionOrThrow: () => getExtensionOrThrow,
+  getFramebufferErrorMessage: () => getFramebufferErrorMessage,
+  getMaxTexturesInShader: () => getMaxTexturesInShader,
+  getNumChannels: () => getNumChannels,
+  getProgramUniformLocation: () => getProgramUniformLocation,
+  getProgramUniformLocationOrThrow: () => getProgramUniformLocationOrThrow,
+  getRowsCols: () => getRowsCols,
+  getShapeAs3D: () => getShapeAs3D,
+  getTextureShapeFromLogicalShape: () => getTextureShapeFromLogicalShape,
+  getWebGLDisjointQueryTimerVersion: () => getWebGLDisjointQueryTimerVersion,
+  getWebGLErrorMessage: () => getWebGLErrorMessage,
+  getWebGLMaxTextureSize: () => getWebGLMaxTextureSize,
+  hasExtension: () => hasExtension,
+  isCapableOfRenderingToFloatTexture: () => isCapableOfRenderingToFloatTexture,
+  isDownloadFloatTextureEnabled: () => isDownloadFloatTextureEnabled,
+  isReshapeFree: () => isReshapeFree,
+  isWebGLFenceEnabled: () => isWebGLFenceEnabled,
+  isWebGLVersionEnabled: () => isWebGLVersionEnabled,
+  linkProgram: () => linkProgram,
+  logShaderSourceAndInfoLog: () => logShaderSourceAndInfoLog,
+  resetMaxTextureSize: () => resetMaxTextureSize,
+  resetMaxTexturesInShader: () => resetMaxTexturesInShader,
+  unbindColorTextureFromFramebuffer: () => unbindColorTextureFromFramebuffer,
+  unbindTextureUnit: () => unbindTextureUnit,
+  validateFramebuffer: () => validateFramebuffer,
+  validateProgram: () => validateProgram,
+  validateTextureSize: () => validateTextureSize
+});
+var contexts = {};
+var WEBGL_ATTRIBUTES = {
+  alpha: false,
+  antialias: false,
+  premultipliedAlpha: false,
+  preserveDrawingBuffer: false,
+  depth: false,
+  stencil: false,
+  failIfMajorPerformanceCaveat: true
+};
+function setWebGLContext(webGLVersion, gl) {
+  contexts[webGLVersion] = gl;
+}
+function getWebGLContext(webGLVersion, customCanvas) {
+  if (!(webGLVersion in contexts) || customCanvas != null) {
+    const newCtx = getWebGLRenderingContext(webGLVersion, customCanvas);
+    if (newCtx !== null) {
+      contexts[webGLVersion] = newCtx;
+    } else {
+      console.log("Could not get context for WebGL version", webGLVersion);
+      return null;
+    }
+  }
+  const gl = contexts[webGLVersion];
+  if (gl == null || gl.isContextLost()) {
+    delete contexts[webGLVersion];
+    return getWebGLContext(webGLVersion);
+  }
+  gl.disable(gl.DEPTH_TEST);
+  gl.disable(gl.STENCIL_TEST);
+  gl.disable(gl.BLEND);
+  gl.disable(gl.DITHER);
+  gl.disable(gl.POLYGON_OFFSET_FILL);
+  gl.disable(gl.SAMPLE_COVERAGE);
+  gl.enable(gl.SCISSOR_TEST);
+  gl.enable(gl.CULL_FACE);
+  gl.cullFace(gl.BACK);
+  return contexts[webGLVersion];
+}
+function createCanvas(webGLVersion) {
+  if (typeof OffscreenCanvas !== "undefined" && webGLVersion === 2) {
+    return new OffscreenCanvas(300, 150);
+  } else if (typeof document !== "undefined") {
+    return document.createElement("canvas");
+  } else {
+    throw new Error("Cannot create a canvas in this context");
+  }
+}
+function getWebGLRenderingContext(webGLVersion, customCanvas) {
+  if (webGLVersion !== 1 && webGLVersion !== 2) {
+    throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");
+  }
+  const canvas = customCanvas == null ? createCanvas(webGLVersion) : customCanvas;
+  canvas.addEventListener("webglcontextlost", (ev) => {
+    ev.preventDefault();
+    delete contexts[webGLVersion];
+  }, false);
+  if (env().getBool("SOFTWARE_WEBGL_ENABLED")) {
+    WEBGL_ATTRIBUTES.failIfMajorPerformanceCaveat = false;
+  }
+  if (webGLVersion === 1) {
+    return canvas.getContext("webgl", WEBGL_ATTRIBUTES) || canvas.getContext("experimental-webgl", WEBGL_ATTRIBUTES);
+  }
+  return canvas.getContext("webgl2", WEBGL_ATTRIBUTES);
+}
+var PackingScheme;
+(function(PackingScheme2) {
+  PackingScheme2[PackingScheme2["DENSE"] = 0] = "DENSE";
+  PackingScheme2[PackingScheme2["SHARED_BATCH"] = 1] = "SHARED_BATCH";
+})(PackingScheme || (PackingScheme = {}));
+var TextureUsage;
+(function(TextureUsage2) {
+  TextureUsage2[TextureUsage2["RENDER"] = 0] = "RENDER";
+  TextureUsage2[TextureUsage2["UPLOAD"] = 1] = "UPLOAD";
+  TextureUsage2[TextureUsage2["PIXELS"] = 2] = "PIXELS";
+  TextureUsage2[TextureUsage2["DOWNLOAD"] = 3] = "DOWNLOAD";
+})(TextureUsage || (TextureUsage = {}));
+var PhysicalTextureType;
+(function(PhysicalTextureType2) {
+  PhysicalTextureType2[PhysicalTextureType2["UNPACKED_FLOAT16"] = 0] = "UNPACKED_FLOAT16";
+  PhysicalTextureType2[PhysicalTextureType2["UNPACKED_FLOAT32"] = 1] = "UNPACKED_FLOAT32";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_4X1_UNSIGNED_BYTE"] = 2] = "PACKED_4X1_UNSIGNED_BYTE";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_2X2_FLOAT32"] = 3] = "PACKED_2X2_FLOAT32";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_2X2_FLOAT16"] = 4] = "PACKED_2X2_FLOAT16";
+})(PhysicalTextureType || (PhysicalTextureType = {}));
+function getUnpackedMatrixTextureShapeWidthHeight(rows, columns) {
+  return [columns, rows];
+}
+function getUnpackedArraySizeFromMatrixSize(matrixSize, channelsPerTexture) {
+  return matrixSize * channelsPerTexture;
+}
+function getDenseTexShape(shape) {
+  const size = util_exports.sizeFromShape(shape);
+  const texelsNeeded = Math.ceil(size / 4);
+  return util_exports.sizeToSquarishShape(texelsNeeded);
+}
+function getPackedMatrixTextureShapeWidthHeight(rows, columns) {
+  return [
+    Math.max(1, Math.ceil(columns / 2)),
+    Math.max(1, Math.ceil(rows / 2))
+  ];
+}
+function getPackedRGBAArraySizeFromMatrixShape(rows, columns) {
+  const [w, h] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return w * h * 4;
+}
+function getTextureConfig(gl, textureHalfFloatExtension) {
+  const glany = gl;
+  let internalFormatFloat;
+  let internalFormatHalfFloat;
+  let internalFormatPackedHalfFloat;
+  let internalFormatPackedFloat;
+  let textureFormatFloat;
+  let downloadTextureFormat;
+  let downloadUnpackNumChannels;
+  let defaultNumChannels;
+  let textureTypeHalfFloat;
+  let textureTypeFloat;
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    internalFormatFloat = glany.R32F;
+    internalFormatHalfFloat = glany.R16F;
+    internalFormatPackedHalfFloat = glany.RGBA16F;
+    internalFormatPackedFloat = glany.RGBA32F;
+    textureFormatFloat = glany.RED;
+    downloadUnpackNumChannels = 4;
+    defaultNumChannels = 1;
+    textureTypeHalfFloat = glany.HALF_FLOAT;
+    textureTypeFloat = glany.FLOAT;
+    downloadTextureFormat = glany.RGBA8;
+  } else {
+    internalFormatFloat = gl.RGBA;
+    internalFormatHalfFloat = gl.RGBA;
+    internalFormatPackedHalfFloat = gl.RGBA;
+    internalFormatPackedFloat = glany.RGBA;
+    textureFormatFloat = gl.RGBA;
+    downloadUnpackNumChannels = 4;
+    defaultNumChannels = 4;
+    textureTypeHalfFloat = textureHalfFloatExtension != null ? textureHalfFloatExtension.HALF_FLOAT_OES : null;
+    textureTypeFloat = gl.FLOAT;
+    downloadTextureFormat = gl.RGBA;
+  }
+  return {
+    internalFormatFloat,
+    internalFormatHalfFloat,
+    internalFormatPackedHalfFloat,
+    internalFormatPackedFloat,
+    textureFormatFloat,
+    downloadTextureFormat,
+    downloadUnpackNumChannels,
+    defaultNumChannels,
+    textureTypeHalfFloat,
+    textureTypeFloat
+  };
+}
+function callAndCheck(gl, func2) {
+  const returnValue = func2();
+  if (env().getBool("DEBUG")) {
+    checkWebGLError(gl);
+  }
+  return returnValue;
+}
+function checkWebGLError(gl) {
+  const error = gl.getError();
+  if (error !== gl.NO_ERROR) {
+    throw new Error("WebGL Error: " + getWebGLErrorMessage(gl, error));
+  }
+}
+var MIN_FLOAT16 = 596e-10;
+var MAX_FLOAT16 = 65504;
+function canBeRepresented(num) {
+  if (env().getBool("WEBGL_RENDER_FLOAT32_ENABLED") || num === 0 || MIN_FLOAT16 < Math.abs(num) && Math.abs(num) < MAX_FLOAT16) {
+    return true;
+  }
+  return false;
+}
+function getWebGLErrorMessage(gl, status) {
+  switch (status) {
+    case gl.NO_ERROR:
+      return "NO_ERROR";
+    case gl.INVALID_ENUM:
+      return "INVALID_ENUM";
+    case gl.INVALID_VALUE:
+      return "INVALID_VALUE";
+    case gl.INVALID_OPERATION:
+      return "INVALID_OPERATION";
+    case gl.INVALID_FRAMEBUFFER_OPERATION:
+      return "INVALID_FRAMEBUFFER_OPERATION";
+    case gl.OUT_OF_MEMORY:
+      return "OUT_OF_MEMORY";
+    case gl.CONTEXT_LOST_WEBGL:
+      return "CONTEXT_LOST_WEBGL";
+    default:
+      return `Unknown error code ${status}`;
+  }
+}
+function getExtensionOrThrow(gl, extensionName) {
+  return throwIfNull(gl, () => gl.getExtension(extensionName), 'Extension "' + extensionName + '" not supported on this browser.');
+}
+function createVertexShader(gl, vertexShaderSource) {
+  const vertexShader = throwIfNull(gl, () => gl.createShader(gl.VERTEX_SHADER), "Unable to create vertex WebGLShader.");
+  callAndCheck(gl, () => gl.shaderSource(vertexShader, vertexShaderSource));
+  callAndCheck(gl, () => gl.compileShader(vertexShader));
+  if (gl.getShaderParameter(vertexShader, gl.COMPILE_STATUS) === false) {
+    console.log(gl.getShaderInfoLog(vertexShader));
+    throw new Error("Failed to compile vertex shader.");
+  }
+  return vertexShader;
+}
+function createFragmentShader(gl, fragmentShaderSource) {
+  const fragmentShader = throwIfNull(gl, () => gl.createShader(gl.FRAGMENT_SHADER), "Unable to create fragment WebGLShader.");
+  callAndCheck(gl, () => gl.shaderSource(fragmentShader, fragmentShaderSource));
+  callAndCheck(gl, () => gl.compileShader(fragmentShader));
+  if (env().get("ENGINE_COMPILE_ONLY")) {
+    return fragmentShader;
+  }
+  if (gl.getShaderParameter(fragmentShader, gl.COMPILE_STATUS) === false) {
+    logShaderSourceAndInfoLog(fragmentShaderSource, gl.getShaderInfoLog(fragmentShader));
+    throw new Error("Failed to compile fragment shader.");
+  }
+  return fragmentShader;
+}
+var lineNumberRegex = /ERROR: [0-9]+:([0-9]+):/g;
+function logShaderSourceAndInfoLog(shaderSource, shaderInfoLog) {
+  const lineNumberRegexResult = lineNumberRegex.exec(shaderInfoLog);
+  if (lineNumberRegexResult == null) {
+    console.log(`Couldn't parse line number in error: ${shaderInfoLog}`);
+    console.log(shaderSource);
+    return;
+  }
+  const lineNumber = +lineNumberRegexResult[1];
+  const shaderLines = shaderSource.split("\n");
+  const pad3 = shaderLines.length.toString().length + 2;
+  const linesWithLineNumbers = shaderLines.map((line, lineNumber2) => util_exports.rightPad((lineNumber2 + 1).toString(), pad3) + line);
+  let maxLineLength = 0;
+  for (let i = 0; i < linesWithLineNumbers.length; i++) {
+    maxLineLength = Math.max(linesWithLineNumbers[i].length, maxLineLength);
+  }
+  const beforeErrorLines = linesWithLineNumbers.slice(0, lineNumber - 1);
+  const errorLine = linesWithLineNumbers.slice(lineNumber - 1, lineNumber);
+  const afterErrorLines = linesWithLineNumbers.slice(lineNumber);
+  console.log(beforeErrorLines.join("\n"));
+  console.log(shaderInfoLog.split("\n")[0]);
+  console.log(`%c ${util_exports.rightPad(errorLine[0], maxLineLength)}`, "border:1px solid red; background-color:#e3d2d2; color:#a61717");
+  console.log(afterErrorLines.join("\n"));
+}
+function createProgram(gl) {
+  return throwIfNull(gl, () => gl.createProgram(), "Unable to create WebGLProgram.");
+}
+function linkProgram(gl, program) {
+  callAndCheck(gl, () => gl.linkProgram(program));
+  if (env().get("ENGINE_COMPILE_ONLY")) {
+    return;
+  }
+  if (gl.getProgramParameter(program, gl.LINK_STATUS) === false) {
+    console.log(gl.getProgramInfoLog(program));
+    throw new Error("Failed to link vertex and fragment shaders.");
+  }
+}
+function validateProgram(gl, program) {
+  callAndCheck(gl, () => gl.validateProgram(program));
+  if (gl.getProgramParameter(program, gl.VALIDATE_STATUS) === false) {
+    console.log(gl.getProgramInfoLog(program));
+    throw new Error("Shader program validation failed.");
+  }
+}
+function createStaticVertexBuffer(gl, data) {
+  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), "Unable to create WebGLBuffer");
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.bufferData(gl.ARRAY_BUFFER, data, gl.STATIC_DRAW));
+  return buffer2;
+}
+function createStaticIndexBuffer(gl, data) {
+  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), "Unable to create WebGLBuffer");
+  callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, data, gl.STATIC_DRAW));
+  return buffer2;
+}
+function getNumChannels() {
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    return 1;
+  }
+  return 4;
+}
+function createTexture(gl) {
+  return throwIfNull(gl, () => gl.createTexture(), "Unable to create WebGLTexture.");
+}
+function validateTextureSize(width, height) {
+  const maxTextureSize = env().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+  if (width <= 0 || height <= 0) {
+    const requested = `[${width}x${height}]`;
+    throw new Error("Requested texture size " + requested + " is invalid.");
+  }
+  if (width > maxTextureSize || height > maxTextureSize) {
+    const requested = `[${width}x${height}]`;
+    const max6 = `[${maxTextureSize}x${maxTextureSize}]`;
+    throw new Error("Requested texture size " + requested + " greater than WebGL maximum on this browser / GPU " + max6 + ".");
+  }
+}
+function createFramebuffer(gl) {
+  return throwIfNull(gl, () => gl.createFramebuffer(), "Unable to create WebGLFramebuffer.");
+}
+function bindVertexBufferToProgramAttribute(gl, program, attribute, buffer2, arrayEntriesPerItem, itemStrideInBytes, itemOffsetInBytes) {
+  const loc = gl.getAttribLocation(program, attribute);
+  if (loc === -1) {
+    return false;
+  }
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.vertexAttribPointer(loc, arrayEntriesPerItem, gl.FLOAT, false, itemStrideInBytes, itemOffsetInBytes));
+  callAndCheck(gl, () => gl.enableVertexAttribArray(loc));
+  return true;
+}
+function bindTextureUnit(gl, texture, textureUnit) {
+  validateTextureUnit(gl, textureUnit);
+  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+}
+function unbindTextureUnit(gl, textureUnit) {
+  validateTextureUnit(gl, textureUnit);
+  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function getProgramUniformLocationOrThrow(gl, program, uniformName) {
+  return throwIfNull(gl, () => gl.getUniformLocation(program, uniformName), 'uniform "' + uniformName + '" not present in program.');
+}
+function getProgramUniformLocation(gl, program, uniformName) {
+  return gl.getUniformLocation(program, uniformName);
+}
+function bindTextureToProgramUniformSampler(gl, texture, uniformSamplerLocation, textureUnit) {
+  callAndCheck(gl, () => bindTextureUnit(gl, texture, textureUnit));
+  callAndCheck(gl, () => gl.uniform1i(uniformSamplerLocation, textureUnit));
+}
+function bindCanvasToFramebuffer(gl) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));
+  callAndCheck(gl, () => gl.viewport(0, 0, gl.canvas.width, gl.canvas.height));
+  callAndCheck(gl, () => gl.scissor(0, 0, gl.canvas.width, gl.canvas.height));
+}
+function bindColorTextureToFramebuffer(gl, texture, framebuffer) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));
+  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0));
+}
+function unbindColorTextureFromFramebuffer(gl, framebuffer) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));
+  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, null, 0));
+}
+function validateFramebuffer(gl) {
+  const status = gl.checkFramebufferStatus(gl.FRAMEBUFFER);
+  if (status !== gl.FRAMEBUFFER_COMPLETE) {
+    throw new Error("Error binding framebuffer: " + getFramebufferErrorMessage(gl, status));
+  }
+}
+function getFramebufferErrorMessage(gl, status) {
+  switch (status) {
+    case gl.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:
+      return "FRAMEBUFFER_INCOMPLETE_ATTACHMENT";
+    case gl.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:
+      return "FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";
+    case gl.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:
+      return "FRAMEBUFFER_INCOMPLETE_DIMENSIONS";
+    case gl.FRAMEBUFFER_UNSUPPORTED:
+      return "FRAMEBUFFER_UNSUPPORTED";
+    default:
+      return `unknown error ${status}`;
+  }
+}
+function throwIfNull(gl, returnTOrNull, failureMessage) {
+  const tOrNull = callAndCheck(gl, () => returnTOrNull());
+  if (tOrNull == null) {
+    throw new Error(failureMessage);
+  }
+  return tOrNull;
+}
+function validateTextureUnit(gl, textureUnit) {
+  const maxTextureUnit = gl.MAX_COMBINED_TEXTURE_IMAGE_UNITS - 1;
+  const glTextureUnit = textureUnit + gl.TEXTURE0;
+  if (glTextureUnit < gl.TEXTURE0 || glTextureUnit > maxTextureUnit) {
+    const textureUnitRange = `[gl.TEXTURE0, gl.TEXTURE${maxTextureUnit}]`;
+    throw new Error(`textureUnit must be in ${textureUnitRange}.`);
+  }
+}
+function getBatchDim(shape, dimsToSkip = 2) {
+  return util_exports.sizeFromShape(shape.slice(0, shape.length - dimsToSkip));
+}
+function getRowsCols(shape) {
+  if (shape.length === 0) {
+    throw Error("Cannot get rows and columns of an empty shape array.");
+  }
+  return [
+    shape.length > 1 ? shape[shape.length - 2] : 1,
+    shape[shape.length - 1]
+  ];
+}
+function getShapeAs3D(shape) {
+  let shapeAs3D = [1, 1, 1];
+  const isScalar = shape.length === 0 || shape.length === 1 && shape[0] === 1;
+  if (!isScalar) {
+    shapeAs3D = [getBatchDim(shape), ...getRowsCols(shape)];
+  }
+  return shapeAs3D;
+}
+function getTextureShapeFromLogicalShape(logShape, isPacked = false) {
+  let maxTexSize = env().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+  let maxSizeForNarrowTex = env().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");
+  if (maxSizeForNarrowTex === Infinity && env().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")) {
+    maxSizeForNarrowTex = maxTexSize / 2;
+  }
+  if (isPacked) {
+    maxTexSize = maxTexSize * 2;
+    maxSizeForNarrowTex = maxSizeForNarrowTex * 2;
+    logShape = logShape.map((d, i) => i >= logShape.length - 2 ? util_exports.nearestLargerEven(logShape[i]) : logShape[i]);
+    if (logShape.length === 1) {
+      logShape = [2, logShape[0]];
+    }
+  }
+  if (logShape.length !== 2) {
+    const squeezeResult = util_exports.squeezeShape(logShape);
+    logShape = squeezeResult.newShape;
+  }
+  let size = util_exports.sizeFromShape(logShape);
+  let textureShape = null;
+  if (logShape.length <= 1 && size <= maxTexSize) {
+    textureShape = [1, size];
+  } else if (logShape.length === 2 && logShape[0] <= maxTexSize && logShape[1] <= maxTexSize) {
+    textureShape = logShape;
+  } else if (logShape.length === 3 && logShape[0] * logShape[1] <= maxTexSize && logShape[2] <= maxTexSize) {
+    textureShape = [logShape[0] * logShape[1], logShape[2]];
+  } else if (logShape.length === 3 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] <= maxTexSize) {
+    textureShape = [logShape[0], logShape[1] * logShape[2]];
+  } else if (logShape.length === 4 && logShape[0] * logShape[1] * logShape[2] <= maxTexSize && logShape[3] <= maxTexSize) {
+    textureShape = [logShape[0] * logShape[1] * logShape[2], logShape[3]];
+  } else if (logShape.length === 4 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] * logShape[3] <= maxTexSize) {
+    textureShape = [logShape[0], logShape[1] * logShape[2] * logShape[3]];
+  }
+  const isLongNarrowTex = textureShape != null && Math.max(...textureShape) > maxSizeForNarrowTex && Math.min(...textureShape) <= (isPacked ? 2 : 1) && Math.min(...textureShape) > 0;
+  if (textureShape == null || isLongNarrowTex) {
+    if (isPacked) {
+      const batchDim = getBatchDim(logShape);
+      let rows = 2, cols = 2;
+      if (logShape.length) {
+        [rows, cols] = getRowsCols(logShape);
+      }
+      size = batchDim * (rows / 2) * (cols / 2);
+      textureShape = util_exports.sizeToSquarishShape(size).map((d) => d * 2);
+    } else {
+      textureShape = util_exports.sizeToSquarishShape(size);
+    }
+  }
+  return textureShape;
+}
+function isEven(n) {
+  return n % 2 === 0;
+}
+function isReshapeFree(shape1, shape2) {
+  shape1 = shape1.slice(-2);
+  shape2 = shape2.slice(-2);
+  if (util_exports.arraysEqual(shape1, shape2)) {
+    return true;
+  }
+  if (!shape1.length || !shape2.length) {
+    return true;
+  }
+  if (shape1[0] === 0 || shape1[1] === 0 || shape2[0] === 0 || shape2[1] === 0) {
+    return true;
+  }
+  if (shape1.length !== shape2.length) {
+    const shape1Cols = shape1.slice(-1)[0];
+    const shape2Cols = shape2.slice(-1)[0];
+    if (shape1Cols === shape2Cols) {
+      return true;
+    }
+    if (isEven(shape1Cols) && isEven(shape2Cols) && (shape1[0] === 1 || shape2[0] === 1)) {
+      return true;
+    }
+  }
+  return shape1[1] === shape2[1] && isEven(shape1[0]) && isEven(shape2[0]);
+}
+var MAX_TEXTURE_SIZE;
+var MAX_TEXTURES_IN_SHADER;
+function getWebGLMaxTextureSize(webGLVersion) {
+  if (MAX_TEXTURE_SIZE == null) {
+    const gl = getWebGLContext(webGLVersion);
+    MAX_TEXTURE_SIZE = gl.getParameter(gl.MAX_TEXTURE_SIZE);
+  }
+  return MAX_TEXTURE_SIZE;
+}
+function resetMaxTextureSize() {
+  MAX_TEXTURE_SIZE = null;
+}
+function resetMaxTexturesInShader() {
+  MAX_TEXTURES_IN_SHADER = null;
+}
+function getMaxTexturesInShader(webGLVersion) {
+  if (MAX_TEXTURES_IN_SHADER == null) {
+    const gl = getWebGLContext(webGLVersion);
+    MAX_TEXTURES_IN_SHADER = gl.getParameter(gl.MAX_TEXTURE_IMAGE_UNITS);
+  }
+  return Math.min(16, MAX_TEXTURES_IN_SHADER);
+}
+function getWebGLDisjointQueryTimerVersion(webGLVersion) {
+  if (webGLVersion === 0) {
+    return 0;
+  }
+  let queryTimerVersion;
+  const gl = getWebGLContext(webGLVersion);
+  if (hasExtension(gl, "EXT_disjoint_timer_query_webgl2") && webGLVersion === 2) {
+    queryTimerVersion = 2;
+  } else if (hasExtension(gl, "EXT_disjoint_timer_query")) {
+    queryTimerVersion = 1;
+  } else {
+    queryTimerVersion = 0;
+  }
+  return queryTimerVersion;
+}
+function hasExtension(gl, extensionName) {
+  const ext = gl.getExtension(extensionName);
+  return ext != null;
+}
+function isWebGLVersionEnabled(webGLVersion) {
+  try {
+    const gl = getWebGLContext(webGLVersion);
+    if (gl != null) {
+      return true;
+    }
+  } catch (e) {
+    console.log("Error when getting WebGL context: ", e);
+    return false;
+  }
+  return false;
+}
+function isCapableOfRenderingToFloatTexture(webGLVersion) {
+  if (webGLVersion === 0) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  if (webGLVersion === 1) {
+    if (!hasExtension(gl, "OES_texture_float")) {
+      return false;
+    }
+  } else {
+    if (!hasExtension(gl, "EXT_color_buffer_float")) {
+      return false;
+    }
+  }
+  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);
+  return isFrameBufferComplete;
+}
+function isDownloadFloatTextureEnabled(webGLVersion) {
+  if (webGLVersion === 0) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  if (webGLVersion === 1) {
+    if (!hasExtension(gl, "OES_texture_float")) {
+      return false;
+    }
+    if (!hasExtension(gl, "WEBGL_color_buffer_float")) {
+      return false;
+    }
+  } else {
+    if (hasExtension(gl, "EXT_color_buffer_float")) {
+      return createFloatTextureAndBindToFramebuffer(gl);
+    }
+    const COLOR_BUFFER_HALF_FLOAT = "EXT_color_buffer_half_float";
+    if (hasExtension(gl, COLOR_BUFFER_HALF_FLOAT)) {
+      const textureHalfFloatExtension = gl.getExtension(COLOR_BUFFER_HALF_FLOAT);
+      return createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension);
+    }
+    return false;
+  }
+  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);
+  return isFrameBufferComplete;
+}
+function createFloatTextureAndBindToFramebuffer(gl) {
+  const texConfig = getTextureConfig(gl);
+  const texture = gl.createTexture();
+  gl.bindTexture(gl.TEXTURE_2D, texture);
+  const width = 1;
+  const height = 1;
+  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeFloat, null);
+  const frameBuffer = gl.createFramebuffer();
+  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);
+  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);
+  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;
+  gl.bindTexture(gl.TEXTURE_2D, null);
+  gl.bindFramebuffer(gl.FRAMEBUFFER, null);
+  gl.deleteTexture(texture);
+  gl.deleteFramebuffer(frameBuffer);
+  return isFrameBufferComplete;
+}
+function createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension) {
+  const texConfig = getTextureConfig(gl, textureHalfFloatExtension);
+  const texture = gl.createTexture();
+  gl.bindTexture(gl.TEXTURE_2D, texture);
+  const width = 1;
+  const height = 1;
+  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatHalfFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeHalfFloat, null);
+  const frameBuffer = gl.createFramebuffer();
+  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);
+  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);
+  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;
+  gl.bindTexture(gl.TEXTURE_2D, null);
+  gl.bindFramebuffer(gl.FRAMEBUFFER, null);
+  gl.deleteTexture(texture);
+  gl.deleteFramebuffer(frameBuffer);
+  return isFrameBufferComplete;
+}
+function isWebGLFenceEnabled(webGLVersion) {
+  if (webGLVersion !== 2) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  const isEnabled = gl.fenceSync != null;
+  return isEnabled;
+}
+function assertNotComplex2(tensor2, opName) {
+  if (!Array.isArray(tensor2)) {
+    tensor2 = [tensor2];
+  }
+  tensor2.forEach((t) => {
+    if (t != null) {
+      util_exports.assert(t.dtype !== "complex64", () => `${opName} does not support complex64 tensors in the WebGL backend.`);
+    }
+  });
+}
+var ENV5 = env();
+ENV5.registerFlag("HAS_WEBGL", () => ENV5.getNumber("WEBGL_VERSION") > 0);
+ENV5.registerFlag("WEBGL_VERSION", () => {
+  if (isWebGLVersionEnabled(2)) {
+    return 2;
+  } else if (isWebGLVersionEnabled(1)) {
+    return 1;
+  }
+  return 0;
+});
+ENV5.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS", () => false);
+ENV5.registerFlag("WEBGL_BUFFER_SUPPORTED", () => ENV5.get("WEBGL_VERSION") === 2);
+ENV5.registerFlag("WEBGL_CPU_FORWARD", () => true);
+ENV5.registerFlag("WEBGL_FORCE_F16_TEXTURES", () => false);
+ENV5.registerFlag("WEBGL_PACK", () => ENV5.getBool("HAS_WEBGL"));
+ENV5.registerFlag("WEBGL_PACK_NORMALIZATION", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_CLIP", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_DEPTHWISECONV", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_BINARY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_UNARY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_REDUCE", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_LAZILY_UNPACK", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_CONV_IM2COL", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_MAX_TEXTURE_SIZE", () => getWebGLMaxTextureSize(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER", () => getMaxTexturesInShader(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION", () => {
+  const webGLVersion = ENV5.getNumber("WEBGL_VERSION");
+  if (webGLVersion === 0) {
+    return 0;
+  }
+  return getWebGLDisjointQueryTimerVersion(webGLVersion);
+});
+ENV5.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE", () => ENV5.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0 && !device_util_exports.isMobile());
+ENV5.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE", () => isCapableOfRenderingToFloatTexture(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED", () => {
+  return ENV5.getBool("WEBGL_FORCE_F16_TEXTURES") ? false : ENV5.getBool("WEBGL_RENDER_FLOAT32_CAPABLE");
+});
+ENV5.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED", () => isDownloadFloatTextureEnabled(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_FENCE_API_ENABLED", () => isWebGLFenceEnabled(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM", () => {
+  const useUniforms = ENV5.getBool("WEBGL_RENDER_FLOAT32_ENABLED");
+  return useUniforms ? 4 : 0;
+});
+ENV5.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD", () => {
+  return -1;
+}, (threshold3) => {
+  if (threshold3 < 0 && threshold3 !== -1) {
+    throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${threshold3}.`);
+  }
+});
+ENV5.registerFlag("WEBGL_FLUSH_THRESHOLD", () => {
+  return device_util_exports.isMobile() ? 1 : -1;
+}, (threshold3) => {
+  if (threshold3 < 0 && threshold3 !== -1) {
+    throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${threshold3}.`);
+  }
+});
+ENV5.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD", () => 128);
+ENV5.registerFlag("WEBGL_USE_SHAPES_UNIFORMS", () => false);
+ENV5.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD", () => 1e5);
+ENV5.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD", () => 128);
+ENV5.registerFlag("WEBGL_EXP_CONV", () => false);
+ENV5.registerFlag("SOFTWARE_WEBGL_ENABLED", () => ENV5.getBool("IS_TEST"));
+ENV5.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE", () => Infinity);
+ENV5.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE", () => false);
+ENV5.registerFlag("WEBGL2_ISNAN_CUSTOM", () => false);
+function getGlslDifferences() {
+  let version10;
+  let attribute;
+  let varyingVs;
+  let varyingFs;
+  let texture2D;
+  let output;
+  let defineOutput;
+  let defineSpecialNaN;
+  let defineSpecialInf;
+  let defineRound;
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    version10 = "#version 300 es";
+    attribute = "in";
+    varyingVs = "out";
+    varyingFs = "in";
+    texture2D = "texture";
+    output = "outputColor";
+    defineOutput = "out vec4 outputColor;";
+    defineSpecialNaN = env().getBool("WEBGL2_ISNAN_CUSTOM") ? `
       bool isnan_custom(float val) {
         uint floatToUint = floatBitsToUint(val);
         return (floatToUint & 0x7fffffffu) > 0x7f800000u;
@@ -74,7 +48129,9 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       }
 
       #define isnan(value) isnan_custom(value)
-    `:"",l="",u=`
+    ` : "";
+    defineSpecialInf = ``;
+    defineRound = `
       #define round(value) newRound(value)
       int newRound(float value) {
         return int(floor(value + 0.5));
@@ -83,7 +48140,16 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 newRound(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `):(e="",t="attribute",n="varying",a="varying",r="texture2D",s="gl_FragColor",i="",o=`
+    `;
+  } else {
+    version10 = "";
+    attribute = "attribute";
+    varyingVs = "varying";
+    varyingFs = "varying";
+    texture2D = "texture2D";
+    output = "gl_FragColor";
+    defineOutput = "";
+    defineSpecialNaN = `
       #define isnan(value) isnan_custom(value)
       bool isnan_custom(float val) {
         return (val > 0. || val < 1. || val == 0.) ? false : true;
@@ -91,7 +48157,8 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       bvec4 isnan_custom(vec4 val) {
         return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));
       }
-    `,l=`
+    `;
+    defineSpecialInf = `
       uniform float INFINITY;
 
       bool isinf(float val) {
@@ -100,7 +48167,8 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       bvec4 isinf(vec4 val) {
         return equal(abs(val), vec4(INFINITY));
       }
-    `,u=`
+    `;
+    defineRound = `
       int round(float value) {
         return int(floor(value + 0.5));
       }
@@ -108,15 +48176,72 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 round(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `),{version:e,attribute:t,varyingVs:n,varyingFs:a,texture2D:r,output:s,defineOutput:i,defineSpecialNaN:o,defineSpecialInf:l,defineRound:u}}function vo(e,t,n="index"){let a=v.computeStrides(t);return a.map((r,s)=>{let i=`int ${e[s]} = ${n} / ${r}`,o=s===a.length-1?`int ${e[s+1]} = ${n} - ${e[s]} * ${r}`:`index -= ${e[s]} * ${r}`;return`${i}; ${o};`}).join("")}function Df(e,t,n="index"){let a=v.computeStrides(t);return a.map((r,s)=>{let i=`int ${e[s]} = ${n} / outShapeStrides[${s}]`,o=s===a.length-1?`int ${e[s+1]} = ${n} - ${e[s]} * outShapeStrides[${s}]`:`index -= ${e[s]} * outShapeStrides[${s}]`;return`${i}; ${o};`}).join("")}function XY(e,t){let n=e.length,a=e.map(s=>`${t}[${s}]`),r=new Array(n-1);r[n-2]=a[n-1];for(let s=n-3;s>=0;--s)r[s]=`(${r[s+1]} * ${a[s+1]})`;return r}function YY(e,t,n="index"){let a=e.map((s,i)=>i),r=XY(a,t);return r.map((s,i)=>{let o=`int ${e[i]} = ${n} / ${r[i]}`,l=i===r.length-1?`int ${e[i+1]} = ${n} - ${e[i]} * ${r[i]}`:`index -= ${e[i]} * ${r[i]}`;return`${o}; ${l};`}).join("")}function U0(e){let t=v.computeStrides(e).map(n=>n.toString());return`
-  int getFlatIndex(ivec3 coords) {
-    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;
+    `;
   }
-`}function G0(){return`
+  return {
+    version: version10,
+    attribute,
+    varyingVs,
+    varyingFs,
+    texture2D,
+    output,
+    defineOutput,
+    defineSpecialNaN,
+    defineSpecialInf,
+    defineRound
+  };
+}
+function getLogicalCoordinatesFromFlatIndex(coords2, shape, index = "index") {
+  const strides = util_exports.computeStrides(shape);
+  return strides.map((stride, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / ${stride}`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${stride}` : `index -= ${coords2[i]} * ${stride}`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function getOutputLogicalCoordinatesFromFlatIndexByUniform(coords2, shape, index = "index") {
+  const strides = util_exports.computeStrides(shape);
+  return strides.map((_, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / outShapeStrides[${i}]`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * outShapeStrides[${i}]` : `index -= ${coords2[i]} * outShapeStrides[${i}]`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function symbolicallyComputeStrides(indicesArr, variableName) {
+  const numCoords = indicesArr.length;
+  const shape = indicesArr.map((d) => `${variableName}[${d}]`);
+  const strides = new Array(numCoords - 1);
+  strides[numCoords - 2] = shape[numCoords - 1];
+  for (let i = numCoords - 3; i >= 0; --i) {
+    strides[i] = `(${strides[i + 1]} * ${shape[i + 1]})`;
+  }
+  return strides;
+}
+function getLogicalCoordinatesFromFlatIndexByUniform(coords2, variableName, index = "index") {
+  const indicesArray = coords2.map((_, i) => i);
+  const strides = symbolicallyComputeStrides(indicesArray, variableName);
+  return strides.map((_, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / ${strides[i]}`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${strides[i]}` : `index -= ${coords2[i]} * ${strides[i]}`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function getFlatIndexFrom3D(shape) {
+  const strides = util_exports.computeStrides(shape).map((d) => d.toString());
+  return `
+  int getFlatIndex(ivec3 coords) {
+    return coords.x * ${strides[0]} + coords.y * ${strides[1]} + coords.z;
+  }
+`;
+}
+function getFlatIndexFrom3DOutput() {
+  return `
   int getFlatIndex(ivec3 coords) {
     return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;
   }
-`}var f_=`
+`;
+}
+var ENCODE_FLOAT_SNIPPET = `
   const float FLOAT_MAX = 1.70141184e38;
   const float FLOAT_MIN = 1.17549435e-38;
 
@@ -155,27 +48280,211 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
 
     return c / 255.0;
   }
-`,{getBroadcastDims:g_}=N;function ZY(e,t,n){let a=[];if(e.forEach(c=>{let h=v.sizeFromShape(c.shapeInfo.logicalShape);if(c.shapeInfo.isUniform?a.push(`uniform float ${c.name}${h>1?`[${h}]`:""};`):(a.push(`uniform sampler2D ${c.name};`),a.push(`uniform int offset${c.name};`)),n.enableShapeUniforms){let{uniformShape:m}=H0(n.packedInputs,c.shapeInfo.logicalShape,c.shapeInfo.texShape);switch(m.length){case 1:a.push(`uniform int ${c.name}Shape;`);break;case 2:a.push(`uniform ivec2 ${c.name}Shape;`);break;case 3:a.push(`uniform ivec3 ${c.name}Shape;`);break;case 4:a.push(`uniform ivec4 ${c.name}Shape;`);break;default:break}a.push(`uniform ivec2 ${c.name}TexShape;`)}}),n.enableShapeUniforms){switch(t.logicalShape.length){case 1:a.push("uniform int outShape;");break;case 2:a.push("uniform ivec2 outShape;"),a.push("uniform int outShapeStrides;");break;case 3:a.push("uniform ivec3 outShape;"),a.push("uniform ivec2 outShapeStrides;");break;case 4:a.push("uniform ivec4 outShape;"),a.push("uniform ivec3 outShapeStrides;");break;default:break}a.push("uniform ivec2 outTexShape;")}n.customUniforms&&n.customUniforms.forEach(c=>{a.push(`uniform ${c.type} ${c.name}${c.arrayIndex?`[${c.arrayIndex}]`:""};`)});let r=a.join(`
-`),s=e.map(c=>JY(c,t,n.packedInputs,n.enableShapeUniforms)).join(`
-`),i=t.texShape,o=Cn(),l=tZ(o),u,p,d=rZ(o);return t.isPacked?(u=QY(t.logicalShape,i,n.enableShapeUniforms),p=aZ(o)):(u=eZ(t.logicalShape,i,n.enableShapeUniforms),p=nZ(o)),n.packedInputs&&(d+=lZ),[d,l,p,r,u,s,n.userCode].join(`
-`)}function zu(e,t=!1){let n=e.shapeInfo.logicalShape;switch(n.length){case 0:return vZ(e,t);case 1:return kZ(e,t);case 2:return SZ(e,t);case 3:return NZ(e,t);case 4:return _Z(e,t);case 5:return EZ(e);case 6:return AZ(e);default:throw new Error(`${n.length}-D input sampling is not yet supported`)}}function y_(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return xZ(e);case 1:return wZ(e,t);case 2:return IZ(e,t);case 3:return TZ(e,t);default:return CZ(e,t)}}function JY(e,t,n=!1,a){let r="";n?r+=y_(e,a):r+=zu(e,a);let s=e.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(n?r+=$Z(e,t):r+=FZ(e,t)),r}function QY(e,t,n){switch(e.length){case 0:return b_();case 1:return uZ(e,t,n);case 2:return yZ(e,t,n);case 3:return cZ(e,t,n);default:return hZ(e,t,n)}}function eZ(e,t,n){switch(e.length){case 0:return b_();case 1:return pZ(e,t,n);case 2:return bZ(e,t,n);case 3:return dZ(e,t,n);case 4:return mZ(e,t,n);case 5:return fZ(e,t);case 6:return gZ(e,t);default:throw new Error(`${e.length}-D output sampling is not yet supported`)}}function tZ(e){return`
+`;
+var { getBroadcastDims: getBroadcastDims2 } = backend_util_exports;
+function makeShader(inputsInfo, outputShape, program) {
+  const prefixSnippets = [];
+  inputsInfo.forEach((x) => {
+    const size = util_exports.sizeFromShape(x.shapeInfo.logicalShape);
+    if (x.shapeInfo.isUniform) {
+      prefixSnippets.push(`uniform float ${x.name}${size > 1 ? `[${size}]` : ""};`);
+    } else {
+      prefixSnippets.push(`uniform sampler2D ${x.name};`);
+      prefixSnippets.push(`uniform int offset${x.name};`);
+    }
+    if (program.enableShapeUniforms) {
+      const { uniformShape } = getUniformInfoFromShape(program.packedInputs, x.shapeInfo.logicalShape, x.shapeInfo.texShape);
+      switch (uniformShape.length) {
+        case 1:
+          prefixSnippets.push(`uniform int ${x.name}Shape;`);
+          break;
+        case 2:
+          prefixSnippets.push(`uniform ivec2 ${x.name}Shape;`);
+          break;
+        case 3:
+          prefixSnippets.push(`uniform ivec3 ${x.name}Shape;`);
+          break;
+        case 4:
+          prefixSnippets.push(`uniform ivec4 ${x.name}Shape;`);
+          break;
+        default:
+          break;
+      }
+      prefixSnippets.push(`uniform ivec2 ${x.name}TexShape;`);
+    }
+  });
+  if (program.enableShapeUniforms) {
+    switch (outputShape.logicalShape.length) {
+      case 1:
+        prefixSnippets.push(`uniform int outShape;`);
+        break;
+      case 2:
+        prefixSnippets.push(`uniform ivec2 outShape;`);
+        prefixSnippets.push(`uniform int outShapeStrides;`);
+        break;
+      case 3:
+        prefixSnippets.push(`uniform ivec3 outShape;`);
+        prefixSnippets.push(`uniform ivec2 outShapeStrides;`);
+        break;
+      case 4:
+        prefixSnippets.push(`uniform ivec4 outShape;`);
+        prefixSnippets.push(`uniform ivec3 outShapeStrides;`);
+        break;
+      default:
+        break;
+    }
+    prefixSnippets.push(`uniform ivec2 outTexShape;`);
+  }
+  if (program.customUniforms) {
+    program.customUniforms.forEach((d) => {
+      prefixSnippets.push(`uniform ${d.type} ${d.name}${d.arrayIndex ? `[${d.arrayIndex}]` : ""};`);
+    });
+  }
+  const inputPrefixSnippet = prefixSnippets.join("\n");
+  const inputSamplingSnippet = inputsInfo.map((x) => getInputSamplingSnippet(x, outputShape, program.packedInputs, program.enableShapeUniforms)).join("\n");
+  const outTexShape = outputShape.texShape;
+  const glsl = getGlslDifferences();
+  const floatTextureSampleSnippet = getFloatTextureSampleSnippet(glsl);
+  let outputSamplingSnippet;
+  let floatTextureSetOutputSnippet;
+  let shaderPrefix = getShaderPrefix(glsl);
+  if (outputShape.isPacked) {
+    outputSamplingSnippet = getPackedOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);
+    floatTextureSetOutputSnippet = getFloatTextureSetRGBASnippet(glsl);
+  } else {
+    outputSamplingSnippet = getOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);
+    floatTextureSetOutputSnippet = getFloatTextureSetRSnippet(glsl);
+  }
+  if (program.packedInputs) {
+    shaderPrefix += SHADER_PACKED_PREFIX;
+  }
+  const source = [
+    shaderPrefix,
+    floatTextureSampleSnippet,
+    floatTextureSetOutputSnippet,
+    inputPrefixSnippet,
+    outputSamplingSnippet,
+    inputSamplingSnippet,
+    program.userCode
+  ].join("\n");
+  return source;
+}
+function getSamplerFromInInfo(inInfo, enableShapeUniforms = false) {
+  const shape = inInfo.shapeInfo.logicalShape;
+  switch (shape.length) {
+    case 0:
+      return getSamplerScalar(inInfo, enableShapeUniforms);
+    case 1:
+      return getSampler1D(inInfo, enableShapeUniforms);
+    case 2:
+      return getSampler2D(inInfo, enableShapeUniforms);
+    case 3:
+      return getSampler3D(inInfo, enableShapeUniforms);
+    case 4:
+      return getSampler4D(inInfo, enableShapeUniforms);
+    case 5:
+      return getSampler5D(inInfo);
+    case 6:
+      return getSampler6D(inInfo);
+    default:
+      throw new Error(`${shape.length}-D input sampling is not yet supported`);
+  }
+}
+function getPackedSamplerFromInInfo(inInfo, enableShapeUniforms) {
+  const shape = inInfo.shapeInfo.logicalShape;
+  switch (shape.length) {
+    case 0:
+      return getPackedSamplerScalar(inInfo);
+    case 1:
+      return getPackedSampler1D(inInfo, enableShapeUniforms);
+    case 2:
+      return getPackedSampler2D(inInfo, enableShapeUniforms);
+    case 3:
+      return getPackedSampler3D(inInfo, enableShapeUniforms);
+    default:
+      return getPackedSamplerND(inInfo, enableShapeUniforms);
+  }
+}
+function getInputSamplingSnippet(inInfo, outShapeInfo, usesPackedTextures = false, enableShapeUniforms) {
+  let res = "";
+  if (usesPackedTextures) {
+    res += getPackedSamplerFromInInfo(inInfo, enableShapeUniforms);
+  } else {
+    res += getSamplerFromInInfo(inInfo, enableShapeUniforms);
+  }
+  const inShape = inInfo.shapeInfo.logicalShape;
+  const outShape = outShapeInfo.logicalShape;
+  if (inShape.length <= outShape.length) {
+    if (usesPackedTextures) {
+      res += getPackedSamplerAtOutputCoords(inInfo, outShapeInfo);
+    } else {
+      res += getSamplerAtOutputCoords(inInfo, outShapeInfo);
+    }
+  }
+  return res;
+}
+function getPackedOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {
+  switch (outShape.length) {
+    case 0:
+      return getOutputScalarCoords();
+    case 1:
+      return getOutputPacked1DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 2:
+      return getOutputPacked2DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 3:
+      return getOutputPacked3DCoords(outShape, outTexShape, enableShapeUniforms);
+    default:
+      return getOutputPackedNDCoords(outShape, outTexShape, enableShapeUniforms);
+  }
+}
+function getOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {
+  switch (outShape.length) {
+    case 0:
+      return getOutputScalarCoords();
+    case 1:
+      return getOutput1DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 2:
+      return getOutput2DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 3:
+      return getOutput3DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 4:
+      return getOutput4DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 5:
+      return getOutput5DCoords(outShape, outTexShape);
+    case 6:
+      return getOutput6DCoords(outShape, outTexShape);
+    default:
+      throw new Error(`${outShape.length}-D output sampling is not yet supported`);
+  }
+}
+function getFloatTextureSampleSnippet(glsl) {
+  return `
     float sampleTexture(sampler2D textureSampler, vec2 uv) {
-      return ${e.texture2D}(textureSampler, uv).r;
+      return ${glsl.texture2D}(textureSampler, uv).r;
     }
-  `}function nZ(e){return`
+  `;
+}
+function getFloatTextureSetRSnippet(glsl) {
+  return `
     void setOutput(float val) {
-      ${e.output} = vec4(val, 0, 0, 0);
+      ${glsl.output} = vec4(val, 0, 0, 0);
     }
-  `}function aZ(e){return`
+  `;
+}
+function getFloatTextureSetRGBASnippet(glsl) {
+  return `
     void setOutput(vec4 val) {
-      ${e.output} = val;
+      ${glsl.output} = val;
     }
-  `}function rZ(e){return`${e.version}
+  `;
+}
+function getShaderPrefix(glsl) {
+  const SHADER_PREFIX = `${glsl.version}
     precision highp float;
     precision highp int;
     precision highp sampler2D;
-    ${e.varyingFs} vec2 resultUV;
-    ${e.defineOutput}
+    ${glsl.varyingFs} vec2 resultUV;
+    ${glsl.defineOutput}
     const vec2 halfCR = vec2(0.5, 0.5);
 
     struct ivec5
@@ -198,9 +48507,9 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
     };
 
     uniform float NAN;
-    ${e.defineSpecialNaN}
-    ${e.defineSpecialInf}
-    ${e.defineRound}
+    ${glsl.defineSpecialNaN}
+    ${glsl.defineSpecialInf}
+    ${glsl.defineRound}
 
     int imod(int x, int y) {
       return x - y * (x / y);
@@ -225,10 +48534,13 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       return fract((p3.x + p3.y) * p3.z);
     }
 
-    ${sZ}
-    ${iZ}
-    ${oZ}
-  `}var sZ=`
+    ${SAMPLE_1D_SNIPPET}
+    ${SAMPLE_2D_SNIPPET}
+    ${SAMPLE_3D_SNIPPET}
+  `;
+  return SHADER_PREFIX;
+}
+var SAMPLE_1D_SNIPPET = `
 vec2 uvFromFlat(int texNumR, int texNumC, int index) {
   int texR = index / texNumC;
   int texC = index - texR * texNumC;
@@ -240,7 +48552,8 @@ vec2 packedUVfrom1D(int texNumR, int texNumC, int index) {
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,iZ=`
+`;
+var SAMPLE_2D_SNIPPET = `
 vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texNumC, int row, int col) {
   int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);
@@ -248,7 +48561,8 @@ vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,oZ=`
+`;
+var SAMPLE_3D_SNIPPET = `
 vec2 packedUVfrom3D(int texNumR, int texNumC,
     int texelsInBatch, int texelsInLogicalRow, int b,
     int row, int col) {
@@ -257,7 +48571,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   int texC = index - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,lZ=`
+`;
+var SHADER_PACKED_PREFIX = `
   float getChannel(vec4 frag, vec2 innerDims) {
     vec2 modCoord = mod(innerDims, 2.);
     return modCoord.x == 0. ?
@@ -268,68 +48583,111 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
     float modCoord = mod(float(dim), 2.);
     return modCoord == 0. ? frag.r : frag.g;
   }
-`;function b_(){return`
+`;
+function getOutputScalarCoords() {
+  return `
     int getOutputCoords() {
       return 0;
     }
-  `}function uZ(e,t,n){let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return a[0]===1?n?`
+  `;
+}
+function getOutputPacked1DCoords(shape, texShape, enableShapeUniforms) {
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (packedTexShape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return 2 * int(resultUV.x * ${a[1]}.0);
+        return 2 * int(resultUV.x * ${packedTexShape[1]}.0);
       }
-    `:a[1]===1?n?`
+    `;
+  }
+  if (packedTexShape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return 2 * int(resultUV.y * ${a[0]}.0);
+        return 2 * int(resultUV.y * ${packedTexShape[0]}.0);
       }
-    `:n?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     int getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(packedTexShape[0], packedTexShape[1]));
       return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);
     }
-  `:`
+  `;
+  }
+  return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      return 2 * (resTexRC.x * ${a[1]} + resTexRC.y);
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      return 2 * (resTexRC.x * ${packedTexShape[1]} + resTexRC.y);
     }
-  `}function pZ(e,t,n){return t[0]===1?n?`
+  `;
+}
+function getOutput1DCoords(shape, texShape, enableShapeUniforms) {
+  if (texShape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return int(resultUV.x * float(outTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return int(resultUV.x * ${t[1]}.0);
+        return int(resultUV.x * ${texShape[1]}.0);
       }
-    `:t[1]===1?n?`
+    `;
+  }
+  if (texShape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return int(resultUV.y * float(outTexShape[0]));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return int(resultUV.y * ${t[0]}.0);
+        return int(resultUV.y * ${texShape[0]}.0);
       }
-    `:n?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(outTexShape[0], outTexShape[1]));
       return resTexRC.x * outTexShape[1] + resTexRC.y;
     }
-  `:`
+  `;
+  }
+  return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      return resTexRC.x * ${t[1]} + resTexRC.y;
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      return resTexRC.x * ${texShape[1]} + resTexRC.y;
     }
-  `}function cZ(e,t,n){if(n)return`
+  `;
+}
+function getOutputPacked3DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    return `
     ivec3 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));
@@ -346,37 +48704,54 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec3(b, r, c);
     }
-  `;let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],r=Math.ceil(e[2]/2),s=r*Math.ceil(e[1]/2);return`
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texelsInLogicalRow = Math.ceil(shape[2] / 2);
+  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[1] / 2);
+  return `
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${texelsInBatch};
+      index -= b * ${texelsInBatch};
 
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
       return ivec3(b, r, c);
     }
-  `}function dZ(e,t,n){if(n)return`
+  `;
+}
+function getOutput3DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], shape);
+    return `
   ivec3 getOutputCoords() {
     ivec2 resTexRC = ivec2(resultUV.yx *
                            vec2(outTexShape[0], outTexShape[1]));
     int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-    ${Df(["r","c","d"],e)}
+    ${coordsFromIndexSnippet2}
     return ivec3(r, c, d);
   }
-`;let a=vo(["r","c","d"],e);return`
+`;
+  }
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], shape);
+  return `
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${a}
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      ${coordsFromIndexSnippet}
       return ivec3(r, c, d);
     }
-  `}function hZ(e,t,n){if(n)return`
+  `;
+}
+function getOutputPackedNDCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    return `
     ivec4 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       ivec2 resTexRC = ivec2(resultUV.yx *
@@ -398,74 +48773,115 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec4(b2, b, r, c);
     }
-  `;let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],r=Math.ceil(e[e.length-1]/2),s=r*Math.ceil(e[e.length-2]/2),i=s,o="",l="b, r, c";for(let u=2;u<e.length-1;u++)i*=e[e.length-u-1],o=`
-      int b${u} = index / ${i};
-      index -= b${u} * ${i};
-    `+o,l=`b${u}, `+l;return`
-    ivec${e.length} getOutputCoords() {
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texelsInLogicalRow = Math.ceil(shape[shape.length - 1] / 2);
+  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[shape.length - 2] / 2);
+  let texelsInBatchN = texelsInBatch;
+  let batches = ``;
+  let coords2 = "b, r, c";
+  for (let b = 2; b < shape.length - 1; b++) {
+    texelsInBatchN *= shape[shape.length - b - 1];
+    batches = `
+      int b${b} = index / ${texelsInBatchN};
+      index -= b${b} * ${texelsInBatchN};
+    ` + batches;
+    coords2 = `b${b}, ` + coords2;
+  }
+  return `
+    ivec${shape.length} getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
 
-      ${o}
+      ${batches}
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${texelsInBatch};
+      index -= b * ${texelsInBatch};
 
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
-      return ivec${e.length}(${l});
+      return ivec${shape.length}(${coords2});
     }
-  `}function mZ(e,t,n){if(n)return`
+  `;
+}
+function getOutput4DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d", "d2"], shape);
+    return `
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
         vec2(outTexShape[0], outTexShape[1]));
       int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-      ${Df(["r","c","d","d2"],e)}
+      ${coordsFromIndexSnippet2}
       return ivec4(r, c, d, d2);
     }
-  `;let a=vo(["r","c","d","d2"],e);return`
+  `;
+  }
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2"], shape);
+  return `
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-        vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${a}
+        vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      ${coordsFromIndexSnippet}
       return ivec4(r, c, d, d2);
     }
-  `}function fZ(e,t){let n=vo(["r","c","d","d2","d3"],e);return`
+  `;
+}
+function getOutput5DCoords(shape, texShape) {
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2", "d3"], shape);
+  return `
     ivec5 getOutputCoords() {
-      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},
-                             ${t[1]}));
+      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${texShape[0]},
+                             ${texShape[1]}));
 
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
 
-      ${n}
+      ${coordsFromIndexSnippet}
 
       ivec5 outShape = ivec5(r, c, d, d2, d3);
       return outShape;
     }
-  `}function gZ(e,t){let n=vo(["r","c","d","d2","d3","d4"],e);return`
+  `;
+}
+function getOutput6DCoords(shape, texShape) {
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2", "d3", "d4"], shape);
+  return `
     ivec6 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-        vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
+        vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
 
-      ${n}
+      ${coordsFromIndexSnippet}
 
       ivec6 result = ivec6(r, c, d, d2, d3, d4);
       return result;
     }
-  `}function yZ(e,t,n){let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(v.arraysEqual(e,t))return n?`
+  `;
+}
+function getOutputPacked2DCoords(shape, texShape, enableShapeUniforms) {
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
         return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
-        return 2 * ivec2(resultUV.yx * vec2(${a[0]}, ${a[1]}));
+        return 2 * ivec2(resultUV.yx * vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
       }
-    `;let r=Math.ceil(e[1]/2);return n?`
+    `;
+  }
+  const texelsInLogicalRow = Math.ceil(shape[1] / 2);
+  if (enableShapeUniforms) {
+    return `
     ivec2 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));
@@ -478,54 +48894,78 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec2(r, c);
     }
-  `:`
+  `;
+  }
+  return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
 
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
       return ivec2(r, c);
     }
-  `}function bZ(e,t,n){return v.arraysEqual(e,t)?n?`
+  `;
+}
+function getOutput2DCoords(shape, texShape, enableShapeUniforms) {
+  if (util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
-        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));
+        return ivec2(resultUV.yx * vec2(${texShape[0]}, ${texShape[1]}));
       }
-    `:e[1]===1?n?`
+    `;
+  }
+  if (shape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
                                vec2(outTexShape[0], outTexShape[1]));
         int index = resTexRC.x * outTexShape[1] + resTexRC.y;
         return ivec2(index, 0);
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
-                               vec2(${t[0]}, ${t[1]}));
-        int index = resTexRC.x * ${t[1]} + resTexRC.y;
+                               vec2(${texShape[0]}, ${texShape[1]}));
+        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
         return ivec2(index, 0);
       }
-    `:e[0]===1?n?`
+    `;
+  }
+  if (shape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
                                vec2(outTexShape[0], outTexShape[1]));
         int index = resTexRC.x * outTexShape[1] + resTexRC.y;
         return ivec2(0, index);
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
-                               vec2(${t[0]}, ${t[1]}));
-        int index = resTexRC.x * ${t[1]} + resTexRC.y;
+                               vec2(${texShape[0]}, ${texShape[1]}));
+        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
         return ivec2(0, index);
       }
-    `:n?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(outTexShape[0], outTexShape[1]));
@@ -534,459 +48974,1204 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       int c = index - r * outShape[1];
       return ivec2(r, c);
     }
-  `:`
+  `;
+  }
+  return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      int r = index / ${e[1]};
-      int c = index - r * ${e[1]};
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      int r = index / ${shape[1]};
+      int c = index - r * ${shape[1]};
       return ivec2(r, c);
     }
-  `}function wo(e){return`offset${e}`}function xZ(e){let t=e.name,n="get"+t.charAt(0).toUpperCase()+t.slice(1),a=Cn();return`
-    vec4 ${n}() {
-      return ${a.texture2D}(${t}, halfCR);
+  `;
+}
+function getFlatOffsetUniformName(texName) {
+  return `offset${texName}`;
+}
+function getPackedSamplerScalar(inputInfo) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const glsl = getGlslDifferences();
+  return `
+    vec4 ${funcName}() {
+      return ${glsl.texture2D}(${texName}, halfCR);
     }
-  `}function vZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`float ${a}() {return ${n};}`;let[r,s]=e.shapeInfo.texShape;if(r===1&&s===1)return`
-      float ${a}() {
-        return sampleTexture(${n}, halfCR);
+  `;
+}
+function getSamplerScalar(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  if (inputInfo.shapeInfo.isUniform) {
+    return `float ${funcName}() {return ${texName};}`;
+  }
+  const [texNumR, texNumC] = inputInfo.shapeInfo.texShape;
+  if (texNumR === 1 && texNumC === 1) {
+    return `
+      float ${funcName}() {
+        return sampleTexture(${texName}, halfCR);
       }
-    `;let i=wo(n);if(t)return`
-    float ${a}() {
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${i});
-      return sampleTexture(${n}, uv);
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}() {
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `;let[o,l]=e.shapeInfo.texShape;return`
-    float ${a}() {
-      vec2 uv = uvFromFlat(${o}, ${l}, ${i});
-      return sampleTexture(${n}, uv);
+  `;
+  }
+  const [tNumR, tNumC] = inputInfo.shapeInfo.texShape;
+  return `
+    float ${funcName}() {
+      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `}function wZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=e.shapeInfo.texShape,s=Cn();if(t)return`
-    vec4 ${a}(int index) {
-      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
+  `;
+}
+function getPackedSampler1D(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int index) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
       vec2 uv = packedUVfrom1D(
         packedTexShape[0], packedTexShape[1], index);
-      return ${s.texture2D}(${n}, uv);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let i=[Math.ceil(r[0]/2),Math.ceil(r[1]/2)];return`
-    vec4 ${a}(int index) {
-      vec2 uv = packedUVfrom1D(
-        ${i[0]}, ${i[1]}, index);
-      return ${s.texture2D}(${n}, uv);
-    }
-  `}function kZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`
-      float ${a}(int index) {
-        ${Wu(e)}
-      }
-    `;let r=e.shapeInfo.texShape,s=r[0],i=r[1];if(i===1&&s===1)return`
-      float ${a}(int index) {
-        return sampleTexture(${n}, halfCR);
-      }
-    `;let o=wo(n);return i===1?t?`
-      float ${a}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${n}TexShape[0]));
-        return sampleTexture(${n}, uv);
-      }
-    `:`
-      float ${a}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);
-        return sampleTexture(${n}, uv);
-      }
-    `:s===1?t?`
-      float ${a}(int index) {
-        vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${n}TexShape[1]), 0.5);
-        return sampleTexture(${n}, uv);
-      }
-    `:`
-      float ${a}(int index) {
-        vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);
-        return sampleTexture(${n}, uv);
-      }
-    `:t?`
-    float ${a}(int index) {
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${o});
-      return sampleTexture(${n}, uv);
-    }
-  `:`
-    float ${a}(int index) {
-      vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});
-      return sampleTexture(${n}, uv);
-    }
-  `}function IZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape,i=s[0],o=s[1],l=Cn();if(s!=null&&v.arraysEqual(n,s))return t?`
-      vec4 ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
-
-        return ${l.texture2D}(${a}, uv);
-      }
-    `:`
-      vec4 ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);
-
-        return ${l.texture2D}(${a}, uv);
-      }
-    `;if(t)return`
-    vec4 ${r}(int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${a}TexShape[0]) / 2.0), ceil(float(${a}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${a}Shape[1]) / 2.0));
-      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
-      return ${l.texture2D}(${a}, uv);
-    }
-  `;let u=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],p=Math.ceil(n[1]/2);return`
-    vec4 ${r}(int row, int col) {
-      vec2 uv = packedUVfrom2D(${p}, ${u[0]}, ${u[1]}, row, col);
-      return ${l.texture2D}(${a}, uv);
-    }
-  `}function SZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape;if(s!=null&&v.arraysEqual(n,s)){if(t)return`
-      float ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
-      }
-    `;let c=s[0],h=s[1];return`
-    float ${r}(int row, int col) {
-      vec2 uv = (vec2(col, row) + halfCR) / vec2(${h}.0, ${c}.0);
-      return sampleTexture(${a}, uv);
-    }
-  `}let{newShape:i,keptDims:o}=v.squeezeShape(n),l=i;if(l.length<n.length){let c=Bu(e,l),h=["row","col"];return`
-      ${zu(c,t)}
-      float ${r}(int row, int col) {
-        return ${r}(${Vu(h,o)});
-      }
-    `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col) {
-        int index = round(dot(vec2(row, col), vec2(${n[1]}, 1)));
-        ${Wu(e)}
-      }
-    `;let u=s[0],p=s[1],d=wo(a);return p===1?t?`
-      float ${r}(int row, int col) {
-        float index = dot(vec3(row, col, ${d}), vec3(${a}Shape[1], 1, 1));
-        vec2 uv = vec2(0.5, (index + 0.5) / float(${a}TexShape[0]));
-        return sampleTexture(${a}, uv);
-      }
-    `:`
-    float ${r}(int row, int col) {
-      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));
-      vec2 uv = vec2(0.5, (index + 0.5) / ${u}.0);
-      return sampleTexture(${a}, uv);
-    }
-  `:u===1?t?`
-      float ${r}(int row, int col) {
-        float index = dot(vec3(row, col, ${d}), vec3(${a}Shape[1], 1, 1));
-        vec2 uv = vec2((index + 0.5) / float(${a}TexShape[1]), 0.5);
-        return sampleTexture(${a}, uv);
-      }
-    `:`
-    float ${r}(int row, int col) {
-      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));
-      vec2 uv = vec2((index + 0.5) / ${p}.0, 0.5);
-      return sampleTexture(${a}, uv);
-    }
-  `:t?`
-      float ${r}(int row, int col) {
-        // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${a}Shape[1] + col + ${d};
-        vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index);
-        return sampleTexture(${a}, uv);
-      }
-    `:`
-  float ${r}(int row, int col) {
-    // Explicitly use integer operations as dot() only works on floats.
-    int index = row * ${n[1]} + col + ${d};
-    vec2 uv = uvFromFlat(${u}, ${p}, index);
-    return sampleTexture(${a}, uv);
+  `;
   }
-`}function TZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(n[0]===1){let c=n.slice(1),h=[1,2],m=Bu(e,c),f=["b","row","col"];return`
-        ${y_(m,t)}
-        vec4 ${r}(int b, int row, int col) {
-          return ${r}(${Vu(f,h)});
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  return `
+    vec4 ${funcName}(int index) {
+      vec2 uv = packedUVfrom1D(
+        ${packedTexShape[0]}, ${packedTexShape[1]}, index);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+}
+function getSampler1D(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int index) {
+        ${getUniformSampler(inputInfo)}
+      }
+    `;
+  }
+  const texShape = inputInfo.shapeInfo.texShape;
+  const tNumR = texShape[0];
+  const tNumC = texShape[1];
+  if (tNumC === 1 && tNumR === 1) {
+    return `
+      float ${funcName}(int index) {
+        return sampleTexture(${texName}, halfCR);
+      }
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (tNumC === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / float(${texName}TexShape[0]));
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / ${tNumR}.0);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  if (tNumR === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2((float(index + ${offset}) + 0.5) / float(${texName}TexShape[1]), 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2((float(index + ${offset}) + 0.5) / ${tNumC}.0, 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int index) {
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  return `
+    float ${funcName}(int index) {
+      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, index + ${offset});
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+}
+function getPackedSampler2D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const glsl = getGlslDifferences();
+  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
+      vec4 ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+
+        return ${glsl.texture2D}(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      vec4 ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);
+
+        return ${glsl.texture2D}(${texName}, uv);
+      }
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${texName}Shape[1]) / 2.0));
+      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const valuesPerRow = Math.ceil(shape[1] / 2);
+  return `
+    vec4 ${funcName}(int row, int col) {
+      vec2 uv = packedUVfrom2D(${valuesPerRow}, ${packedTexShape[0]}, ${packedTexShape[1]}, row, col);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+}
+function getSampler2D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    const texNumR2 = texShape[0];
+    const texNumC2 = texShape[1];
+    return `
+    float ${funcName}(int row, int col) {
+      vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC2}.0, ${texNumR2}.0);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const squeezedShape = newShape;
+  if (squeezedShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["row", "col"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+      float ${funcName}(int row, int col) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
+      }
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col) {
+        int index = round(dot(vec2(row, col), vec2(${shape[1]}, 1)));
+        ${getUniformSampler(inputInfo)}
+      }
+    `;
+  }
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const offset = getFlatOffsetUniformName(texName);
+  if (texNumC === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));
+        vec2 uv = vec2(0.5, (index + 0.5) / float(${texName}TexShape[0]));
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col) {
+      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));
+      vec2 uv = vec2(0.5, (index + 0.5) / ${texNumR}.0);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  if (texNumR === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));
+        vec2 uv = vec2((index + 0.5) / float(${texName}TexShape[1]), 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col) {
+      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));
+      vec2 uv = vec2((index + 0.5) / ${texNumC}.0, 0.5);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  if (enableShapeUniforms) {
+    return `
+      float ${funcName}(int row, int col) {
+        // Explicitly use integer operations as dot() only works on floats.
+        int index = row * ${texName}Shape[1] + col + ${offset};
+        vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  return `
+  float ${funcName}(int row, int col) {
+    // Explicitly use integer operations as dot() only works on floats.
+    int index = row * ${shape[1]} + col + ${offset};
+    vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+    return sampleTexture(${texName}, uv);
+  }
+`;
+}
+function getPackedSampler3D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (shape[0] === 1) {
+    const squeezedShape = shape.slice(1);
+    const keptDims = [1, 2];
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["b", "row", "col"];
+    return `
+        ${getPackedSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+        vec4 ${funcName}(int b, int row, int col) {
+          return ${funcName}(${getSqueezedParams(params, keptDims)});
         }
-      `}let o=Cn();if(t)return`
-    vec4 ${r}(int b, int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${a}TexShape[0]) / 2.0), ceil(float(${a}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${a}Shape[2]) / 2.0));
-      int texelsInBatch = valuesPerRow * int(ceil(float(${a}Shape[1]) / 2.0));
+      `;
+  }
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int b, int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${texName}Shape[2]) / 2.0));
+      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[1]) / 2.0));
       vec2 uv = packedUVfrom3D(
         packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);
-      return ${o.texture2D}(${a}, uv);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let l=i[0],u=i[1],p=Math.ceil(n[2]/2),d=p*Math.ceil(n[1]/2);return`
-    vec4 ${r}(int b, int row, int col) {
+  `;
+  }
+  const texNumR = packedTexShape[0];
+  const texNumC = packedTexShape[1];
+  const valuesPerRow = Math.ceil(shape[2] / 2);
+  const texelsInBatch = valuesPerRow * Math.ceil(shape[1] / 2);
+  return `
+    vec4 ${funcName}(int b, int row, int col) {
       vec2 uv = packedUVfrom3D(
-        ${l}, ${u}, ${d}, ${p}, b, row, col);
-      return ${o.texture2D}(${a}, uv);
+        ${texNumR}, ${texNumC}, ${texelsInBatch}, ${valuesPerRow}, b, row, col);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `}function NZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=n[1]*n[2],i=n[2],{newShape:o,keptDims:l}=v.squeezeShape(n),u=o;if(u.length<n.length){let f=Bu(e,u),g=["row","col","depth"];return`
-        ${zu(f,t)}
-        float ${r}(int row, int col, int depth) {
-          return ${r}(${Vu(g,l)});
+  `;
+}
+function getSampler3D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride0 = shape[1] * shape[2];
+  const stride1 = shape[2];
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const squeezedShape = newShape;
+  if (squeezedShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["row", "col", "depth"];
+    return `
+        ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+        float ${funcName}(int row, int col, int depth) {
+          return ${funcName}(${getSqueezedParams(params, keptDims)});
         }
-      `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col, int depth) {
+      `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth) {
         int index = round(dot(vec3(row, col, depth),
-                          vec3(${s}, ${i}, 1)));
-        ${Wu(e)}
+                          vec3(${stride0}, ${stride1}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let p=e.shapeInfo.texShape,d=p[0],c=p[1],h=e.shapeInfo.flatOffset;if(c===s&&h==null)return t?`
-      float ${r}(int row, int col, int depth) {
-        int stride1 = ${a}Shape[2];
+    `;
+  }
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  if (texNumC === stride0 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth) {
+        int stride1 = ${texName}Shape[2];
         float texR = float(row);
         float texC = dot(vec2(col, depth), vec2(stride1, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-        float ${r}(int row, int col, int depth) {
+    `;
+    }
+    return `
+        float ${funcName}(int row, int col, int depth) {
           float texR = float(row);
-          float texC = dot(vec2(col, depth), vec2(${i}, 1));
+          float texC = dot(vec2(col, depth), vec2(${stride1}, 1));
           vec2 uv = (vec2(texC, texR) + halfCR) /
-                     vec2(${c}.0, ${d}.0);
-          return sampleTexture(${a}, uv);
+                     vec2(${texNumC}.0, ${texNumR}.0);
+          return sampleTexture(${texName}, uv);
         }
-      `;if(c===i&&h==null)return t?`
-      float ${r}(int row, int col, int depth) {
-        float texR = dot(vec2(row, col), vec2(${a}Shape[1], 1));
+      `;
+  }
+  if (texNumC === stride1 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth) {
+        float texR = dot(vec2(row, col), vec2(${texName}Shape[1], 1));
         float texC = float(depth);
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-    float ${r}(int row, int col, int depth) {
-      float texR = dot(vec2(row, col), vec2(${n[1]}, 1));
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col, int depth) {
+      float texR = dot(vec2(row, col), vec2(${shape[1]}, 1));
       float texC = float(depth);
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}.0, ${d}.0);
-      return sampleTexture(${a}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);
+      return sampleTexture(${texName}, uv);
     }
-  `;let m=wo(a);return t?`
-    float ${r}(int row, int col, int depth) {
+  `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int row, int col, int depth) {
       // Explicitly use integer operations as dot() only works on floats.
-      int stride0 = ${a}Shape[1] * ${a}Shape[2];
-      int stride1 = ${a}Shape[2];
-      int index = row * stride0 + col * stride1 + depth + ${m};
-      vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index);
-      return sampleTexture(${a}, uv);
+      int stride0 = ${texName}Shape[1] * ${texName}Shape[2];
+      int stride1 = ${texName}Shape[2];
+      int index = row * stride0 + col * stride1 + depth + ${offset};
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);
+      return sampleTexture(${texName}, uv);
     }
-    `:`
-      float ${r}(int row, int col, int depth) {
+    `;
+  }
+  return `
+      float ${funcName}(int row, int col, int depth) {
         // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${s} + col * ${i} + depth + ${m};
-        vec2 uv = uvFromFlat(${d}, ${c}, index);
-        return sampleTexture(${a}, uv);
+        int index = row * ${stride0} + col * ${stride1} + depth + ${offset};
+        vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+        return sampleTexture(${texName}, uv);
       }
-  `}function CZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=Cn();if(t)return`
-    vec4 ${a}(int b2, int b, int row, int col) {
-      int valuesPerRow = int(ceil(float(${n}Shape[3]) / 2.0));
-      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[2]) / 2.0));
+  `;
+}
+function getPackedSamplerND(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int b2, int b, int row, int col) {
+      int valuesPerRow = int(ceil(float(${texName}Shape[3]) / 2.0));
+      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[2]) / 2.0));
       int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);
-      texelsInBatch *= ${n}Shape[1];
+      texelsInBatch *= ${texName}Shape[1];
       index = b2 * texelsInBatch + index;
-      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
       int texR = index / packedTexShape[1];
       int texC = index - texR * packedTexShape[1];
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${r.texture2D}(${n}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let s=e.shapeInfo.logicalShape,i=s.length,o=e.shapeInfo.texShape,l=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)],u=l[0],p=l[1],d=Math.ceil(s[i-1]/2),c=d*Math.ceil(s[i-2]/2),h="int b, int row, int col",m=`b * ${c} + (row / 2) * ${d} + (col / 2)`;for(let f=2;f<i-1;f++)h=`int b${f}, `+h,c*=s[i-f-1],m=`b${f} * ${c} + `+m;return`
-    vec4 ${a}(${h}) {
-      int index = ${m};
-      int texR = index / ${p};
-      int texC = index - texR * ${p};
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}, ${u});
-      return ${r.texture2D}(${n}, uv);
+  `;
+  }
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const rank = shape.length;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texNumR = packedTexShape[0];
+  const texNumC = packedTexShape[1];
+  const valuesPerRow = Math.ceil(shape[rank - 1] / 2);
+  let texelsInBatch = valuesPerRow * Math.ceil(shape[rank - 2] / 2);
+  let params = `int b, int row, int col`;
+  let index = `b * ${texelsInBatch} + (row / 2) * ${valuesPerRow} + (col / 2)`;
+  for (let b = 2; b < rank - 1; b++) {
+    params = `int b${b}, ` + params;
+    texelsInBatch *= shape[rank - b - 1];
+    index = `b${b} * ${texelsInBatch} + ` + index;
+  }
+  return `
+    vec4 ${funcName}(${params}) {
+      int index = ${index};
+      int texR = index / ${texNumC};
+      int texC = index - texR * ${texNumC};
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}, ${texNumR});
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `}function _Z(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=n[3],i=n[2]*s,o=n[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(n);if(l.length<n.length){let b=Bu(e,l),x=["row","col","depth","depth2"];return`
-      ${zu(b,t)}
-      float ${r}(int row, int col, int depth, int depth2) {
-        return ${r}(${Vu(x,u)});
+  `;
+}
+function getSampler4D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride2 = shape[3];
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+      float ${funcName}(int row, int col, int depth, int depth2) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         int index = round(dot(vec4(row, col, depth, depth2),
-                          vec4(${o}, ${i}, ${s}, 1)));
-        ${Wu(e)}
+                          vec4(${stride0}, ${stride1}, ${stride2}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let p=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,c=d[0],h=d[1],m=`int stride2 = ${a}Shape[3];`,f=`int stride1 = ${a}Shape[2] * stride2;`,g=`int stride0 = ${a}Shape[1] * stride1;`;if(h===o&&p==null)return t?`
-      float ${r}(int row, int col, int depth, int depth2) {
-        ${m}
-        ${f}
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const stride2Str = `int stride2 = ${texName}Shape[3];`;
+  const stride1Str = `int stride1 = ${texName}Shape[2] * stride2;`;
+  const stride0Str = `int stride0 = ${texName}Shape[1] * stride1;`;
+  if (texNumC === stride0 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
+        ${stride2Str}
+        ${stride1Str}
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
                 vec3(stride1, stride2, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-      float ${r}(int row, int col, int depth, int depth2) {
+    `;
+    }
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
-                vec3(${i}, ${s}, 1));
+                vec3(${stride1}, ${stride2}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${h}.0, ${c}.0);
-        return sampleTexture(${a}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(h===s&&p==null)return t?`
-      float ${r}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  if (texNumC === stride2 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
-                         vec3(${a}Shape[1] * ${a}Shape[2], ${a}Shape[2], 1));
+                         vec3(${texName}Shape[1] * ${texName}Shape[2], ${texName}Shape[2], 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                  vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-      float ${r}(int row, int col, int depth, int depth2) {
+    `;
+    }
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
-                         vec3(${n[1]*n[2]}, ${n[2]}, 1));
+                         vec3(${shape[1] * shape[2]}, ${shape[2]}, 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${h}.0, ${c}.0);
-        return sampleTexture(${a}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let y=wo(a);return t?`
-    float ${r}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      ${m}
-      ${f}
-      ${g}
+      ${stride2Str}
+      ${stride1Str}
+      ${stride0Str}
       int index = row * stride0 + col * stride1 +
           depth * stride2 + depth2;
-      vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index + ${y});
-      return sampleTexture(${a}, uv);
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `:`
-    float ${r}(int row, int col, int depth, int depth2) {
+  `;
+  }
+  return `
+    float ${funcName}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${o} + col * ${i} +
-          depth * ${s} + depth2;
-      vec2 uv = uvFromFlat(${c}, ${h}, index + ${y});
-      return sampleTexture(${a}, uv);
+      int index = row * ${stride0} + col * ${stride1} +
+          depth * ${stride2} + depth2;
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index + ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `}function EZ(e){let t=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=t[4],s=t[3]*r,i=t[2]*s,o=t[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(t);if(l.length<t.length){let f=Bu(e,l),g=["row","col","depth","depth2","depth3"];return`
-      ${zu(f)}
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
-        return ${a}(${Vu(g,u)});
+  `;
+}
+function getSampler5D(inputInfo) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride3 = shape[4];
+  const stride2 = shape[3] * stride3;
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2", "depth3"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo)}
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}if(e.shapeInfo.isUniform)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         float index = dot(
           vec4(row, col, depth, depth2),
-          vec4(${o}, ${i}, ${s}, ${r})) +
+          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +
           depth3;
-        ${Wu(e)}
+        ${getUniformSampler(inputInfo)}
       }
-    `;let p=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,c=d[0],h=d[1];if(h===o&&p==null)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  if (texNumC === stride0 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-                         vec4(${i}, ${s}, ${r}, 1));
+                         vec4(${stride1}, ${stride2}, ${stride3}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${h}.0, ${c}.0);
-        return sampleTexture(${n}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(h===r&&p==null)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  if (texNumC === stride3 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         float texR = dot(
           vec4(row, col, depth, depth2),
-          vec4(${t[1]*t[2]*t[3]},
-               ${t[2]*t[3]}, ${t[3]}, 1));
+          vec4(${shape[1] * shape[2] * shape[3]},
+               ${shape[2] * shape[3]}, ${shape[3]}, 1));
         int texC = depth3;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${h}.0, ${c}.0);
-        return sampleTexture(${n}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let m=wo(n);return`
-    float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  return `
+    float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${o} + col * ${i} + depth * ${s} +
-          depth2 * ${r} + depth3 + ${m};
-      vec2 uv = uvFromFlat(${c}, ${h}, index);
-      return sampleTexture(${n}, uv);
+      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +
+          depth2 * ${stride3} + depth3 + ${offset};
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+      return sampleTexture(${texName}, uv);
     }
-  `}function AZ(e){let t=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),{newShape:r,keptDims:s}=v.squeezeShape(t);if(r.length<t.length){let g=Bu(e,r),y=["row","col","depth","depth2","depth3","depth4"];return`
-      ${zu(g)}
-      float ${a}(int row, int col, int depth,
+  `;
+}
+function getSampler6D(inputInfo) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2", "depth3", "depth4"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo)}
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
-        return ${a}(${Vu(y,s)});
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}let i=t[5],o=t[4]*i,l=t[3]*o,u=t[2]*l,p=t[1]*u;if(e.shapeInfo.isUniform)return`
-      float ${a}(int row, int col, int depth,
+    `;
+  }
+  const stride4 = shape[5];
+  const stride3 = shape[4] * stride4;
+  const stride2 = shape[3] * stride3;
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
         int index = round(dot(
           vec4(row, col, depth, depth2),
-          vec4(${p}, ${u}, ${l}, ${o})) +
+          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +
           dot(
             vec2(depth3, depth4),
-            vec2(${i}, 1)));
-        ${Wu(e)}
+            vec2(${stride4}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let d=e.shapeInfo.flatOffset,c=e.shapeInfo.texShape,h=c[0],m=c[1];if(m===p&&d==null)return`
-      float ${a}(int row, int col, int depth,
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  if (texNumC === stride0 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-          vec4(${u}, ${l}, ${o}, ${i})) +
+          vec4(${stride1}, ${stride2}, ${stride3}, ${stride4})) +
                float(depth4);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${m}.0, ${h}.0);
-        return sampleTexture(${n}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(m===i&&d==null)return`
-      float ${a}(int row, int col, int depth,
+    `;
+  }
+  if (texNumC === stride4 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         float texR = dot(vec4(row, col, depth, depth2),
-          vec4(${t[1]*t[2]*t[3]*t[4]},
-               ${t[2]*t[3]*t[4]},
-               ${t[3]*t[4]},
-               ${t[4]})) + float(depth3);
+          vec4(${shape[1] * shape[2] * shape[3] * shape[4]},
+               ${shape[2] * shape[3] * shape[4]},
+               ${shape[3] * shape[4]},
+               ${shape[4]})) + float(depth3);
         int texC = depth4;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${m}.0, ${h}.0);
-        return sampleTexture(${n}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let f=wo(n);return`
-    float ${a}(int row, int col, int depth,
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  return `
+    float ${funcName}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${p} + col * ${u} + depth * ${l} +
-          depth2 * ${o} + depth3 * ${i} + depth4 + ${f};
-      vec2 uv = uvFromFlat(${h}, ${m}, index);
-      return sampleTexture(${n}, uv);
+      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +
+          depth2 * ${stride3} + depth3 * ${stride4} + depth4 + ${offset};
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+      return sampleTexture(${texName}, uv);
     }
-  `}function Wu(e){let t=e.name,n=v.sizeFromShape(e.shapeInfo.logicalShape);return n<2?`return ${t};`:`
-    for (int i = 0; i < ${n}; i++) {
+  `;
+}
+function getUniformSampler(inputInfo) {
+  const texName = inputInfo.name;
+  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);
+  if (inSize < 2) {
+    return `return ${texName};`;
+  }
+  return `
+    for (int i = 0; i < ${inSize}; i++) {
       if (i == index) {
-        return ${t}[i];
+        return ${texName}[i];
       }
     }
-  `}function $Z(e,t){let n=e.name,a=n.charAt(0).toUpperCase()+n.slice(1),r="get"+a+"AtOutCoords",s=e.shapeInfo.logicalShape.length,i=t.logicalShape.length,o=g_(e.shapeInfo.logicalShape,t.logicalShape),l=gt(i),u=i-s,p,d=["x","y","z","w","u","v"];s===0?p="":i<2&&o.length>=1?p="coords = 0;":p=o.map(g=>`coords.${d[g+u]} = 0;`).join(`
-`);let c="";i<2&&s>0?c="coords":c=e.shapeInfo.logicalShape.map((g,y)=>`coords.${d[y+u]}`).join(", ");let h="return outputValue;",m=v.sizeFromShape(e.shapeInfo.logicalShape)===1,f=v.sizeFromShape(t.logicalShape)===1;if(s===1&&!m&&!f)h=`
+  `;
+}
+function getPackedSamplerAtOutputCoords(inputInfo, outShapeInfo) {
+  const texName = inputInfo.name;
+  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);
+  const funcName = "get" + texFuncSnippet + "AtOutCoords";
+  const inRank = inputInfo.shapeInfo.logicalShape.length;
+  const outRank = outShapeInfo.logicalShape.length;
+  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);
+  const type = getCoordsDataType(outRank);
+  const rankDiff = outRank - inRank;
+  let coordsSnippet;
+  const fields = ["x", "y", "z", "w", "u", "v"];
+  if (inRank === 0) {
+    coordsSnippet = "";
+  } else if (outRank < 2 && broadcastDims.length >= 1) {
+    coordsSnippet = "coords = 0;";
+  } else {
+    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join("\n");
+  }
+  let unpackedCoordsSnippet = "";
+  if (outRank < 2 && inRank > 0) {
+    unpackedCoordsSnippet = "coords";
+  } else {
+    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(", ");
+  }
+  let output = `return outputValue;`;
+  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);
+  const isInputScalar = inSize === 1;
+  const outSize = util_exports.sizeFromShape(outShapeInfo.logicalShape);
+  const isOutputScalar = outSize === 1;
+  if (inRank === 1 && !isInputScalar && !isOutputScalar) {
+    output = `
       return vec4(outputValue.xy, outputValue.xy);
-    `;else if(m&&!f)i===1?h=`
+    `;
+  } else if (isInputScalar && !isOutputScalar) {
+    if (outRank === 1) {
+      output = `
         return vec4(outputValue.x, outputValue.x, 0., 0.);
-      `:h=`
+      `;
+    } else {
+      output = `
         return vec4(outputValue.x);
-      `;else if(o.length){let g=s-2,y=s-1;o.indexOf(g)>-1&&o.indexOf(y)>-1?h="return vec4(outputValue.x);":o.indexOf(g)>-1?h="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":o.indexOf(y)>-1&&(h="return vec4(outputValue.xx, outputValue.zz);")}return`
-    vec4 ${r}() {
-      ${l} coords = getOutputCoords();
-      ${p}
-      vec4 outputValue = get${a}(${c});
-      ${h}
+      `;
     }
-  `}function FZ(e,t){let n=e.name,a=n.charAt(0).toUpperCase()+n.slice(1),r="get"+a+"AtOutCoords",s=t.texShape,i=e.shapeInfo.texShape,o=e.shapeInfo.logicalShape.length,l=t.logicalShape.length;if(!e.shapeInfo.isUniform&&o===l&&e.shapeInfo.flatOffset==null&&v.arraysEqual(i,s))return`
-      float ${r}() {
-        return sampleTexture(${n}, resultUV);
+  } else if (broadcastDims.length) {
+    const rows = inRank - 2;
+    const cols = inRank - 1;
+    if (broadcastDims.indexOf(rows) > -1 && broadcastDims.indexOf(cols) > -1) {
+      output = `return vec4(outputValue.x);`;
+    } else if (broadcastDims.indexOf(rows) > -1) {
+      output = `return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);`;
+    } else if (broadcastDims.indexOf(cols) > -1) {
+      output = `return vec4(outputValue.xx, outputValue.zz);`;
+    }
+  }
+  return `
+    vec4 ${funcName}() {
+      ${type} coords = getOutputCoords();
+      ${coordsSnippet}
+      vec4 outputValue = get${texFuncSnippet}(${unpackedCoordsSnippet});
+      ${output}
+    }
+  `;
+}
+function getSamplerAtOutputCoords(inputInfo, outShapeInfo) {
+  const texName = inputInfo.name;
+  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);
+  const funcName = "get" + texFuncSnippet + "AtOutCoords";
+  const outTexShape = outShapeInfo.texShape;
+  const inTexShape = inputInfo.shapeInfo.texShape;
+  const inRank = inputInfo.shapeInfo.logicalShape.length;
+  const outRank = outShapeInfo.logicalShape.length;
+  if (!inputInfo.shapeInfo.isUniform && inRank === outRank && inputInfo.shapeInfo.flatOffset == null && util_exports.arraysEqual(inTexShape, outTexShape)) {
+    return `
+      float ${funcName}() {
+        return sampleTexture(${texName}, resultUV);
       }
-    `;let u=gt(l),p=g_(e.shapeInfo.logicalShape,t.logicalShape),d=l-o,c,h=["x","y","z","w","u","v"];o===0?c="":l<2&&p.length>=1?c="coords = 0;":c=p.map(f=>`coords.${h[f+d]} = 0;`).join(`
-`);let m="";return l<2&&o>0?m="coords":m=e.shapeInfo.logicalShape.map((f,g)=>`coords.${h[g+d]}`).join(", "),`
-    float ${r}() {
-      ${u} coords = getOutputCoords();
-      ${c}
-      return get${a}(${m});
+    `;
+  }
+  const type = getCoordsDataType(outRank);
+  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);
+  const rankDiff = outRank - inRank;
+  let coordsSnippet;
+  const fields = ["x", "y", "z", "w", "u", "v"];
+  if (inRank === 0) {
+    coordsSnippet = "";
+  } else if (outRank < 2 && broadcastDims.length >= 1) {
+    coordsSnippet = "coords = 0;";
+  } else {
+    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join("\n");
+  }
+  let unpackedCoordsSnippet = "";
+  if (outRank < 2 && inRank > 0) {
+    unpackedCoordsSnippet = "coords";
+  } else {
+    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(", ");
+  }
+  return `
+    float ${funcName}() {
+      ${type} coords = getOutputCoords();
+      ${coordsSnippet}
+      return get${texFuncSnippet}(${unpackedCoordsSnippet});
     }
-  `}function gt(e){if(e<=1)return"int";if(e===2)return"ivec2";if(e===3)return"ivec3";if(e===4)return"ivec4";if(e===5)return"ivec5";if(e===6)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function H0(e,t,n){let{newShape:a,keptDims:r}=v.squeezeShape(t),s=t.length,i=e&&s===3&&t[0]===1,o=i?t.slice(1):a,l=!e&&s>1&&!v.arraysEqual(t,n)&&a.length<s||i;return{useSqueezeShape:l,uniformShape:l?o:t,keptDims:r}}function Bu(e,t){let n=JSON.parse(JSON.stringify(e));return n.shapeInfo.logicalShape=t,n}function Vu(e,t){return t.map(n=>e[n]).join(", ")}function DZ(e,t,n,a){let r=n.map((p,d)=>{let c={logicalShape:p.shape,texShape:p.isUniform?null:p.texData.texShape,isUniform:p.isUniform,isPacked:p.isUniform?!1:p.texData.isPacked,flatOffset:null};return p.texData!=null&&p.texData.slice!=null&&p.texData.slice.flatOffset>0&&(c.flatOffset=p.texData.slice.flatOffset),{name:t.variableNames[d],shapeInfo:c}}),s=r.map(p=>p.shapeInfo),i={logicalShape:a.shape,texShape:a.texData.texShape,isUniform:!1,isPacked:a.texData.isPacked,flatOffset:null},o=ZY(r,i,t),l=KC(e.gl,o),u=e.createProgram(l);return H().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:Object.assign({program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i},x_(e,t,u))}function x_(e,t,n){let a={},r={},s={},i=[],o,l,u,p=null,d=null;d=e.getUniformLocation(n,"NAN",!1),H().getNumber("WEBGL_VERSION")===1&&(p=e.getUniformLocation(n,"INFINITY",!1));let c=!1;for(let h=0;h<t.variableNames.length;h++){let m=t.variableNames[h];a[m]=e.getUniformLocation(n,m,c),a[`offset${m}`]=e.getUniformLocation(n,`offset${m}`,c),t.enableShapeUniforms&&(r[`${m}Shape`]=e.getUniformLocation(n,`${m}Shape`,c),s[`${m}TexShape`]=e.getUniformLocation(n,`${m}TexShape`,c))}return t.enableShapeUniforms&&(o=e.getUniformLocation(n,"outShape",c),u=e.getUniformLocation(n,"outShapeStrides",c),l=e.getUniformLocation(n,"outTexShape",c)),t.customUniforms&&t.customUniforms.forEach((h,m)=>{i[m]=e.getUniformLocation(n,h.name,c)}),{uniformLocations:a,customUniformLocations:i,infLoc:p,nanLoc:d,inShapesLocations:r,inTexShapesLocations:s,outShapeLocation:o,outShapeStridesLocation:u,outTexShapeLocation:l}}function Hk(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((n,a)=>{let r=n.logicalShape,s=t[a],i=s.shape;if(!v.arraysEqual(r,i))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${r} and ${i} must match`);if(n.isUniform&&s.isUniform)return;let o=n.texShape,l=s.isUniform?null:s.texData.texShape;if(!v.arraysEqual(o,l))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${o} and ${l} must match`)})}function RZ(e,t,n,a,r){t.program.enableShapeUniforms||(Hk(t.inShapeInfos,n),Hk([t.outShapeInfo],[a]));let s=a.texData.texture,i=a.texData.texShape;a.texData.isPacked?e.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):e.setOutputMatrixTexture(s.texture,i[0],i[1]),e.setProgram(t.webGLProgram),H().getNumber("WEBGL_VERSION")===1&&t.infLoc!==null&&e.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&e.gl.uniform1f(t.nanLoc,NaN),n.forEach((l,u)=>{let p=t.program.variableNames[u],d=t.uniformLocations[p],c=t.uniformLocations[`offset${p}`],h=t.inShapesLocations[`${p}Shape`],m=t.inTexShapesLocations[`${p}TexShape`];if(h){let{uniformShape:f}=H0(t.program.packedInputs,l.shape,l.texData.texShape);switch(f.length){case 1:e.gl.uniform1iv(h,new Int32Array(f));break;case 2:e.gl.uniform2iv(h,new Int32Array(f));break;case 3:e.gl.uniform3iv(h,new Int32Array(f));break;case 4:e.gl.uniform4iv(h,new Int32Array(f));break;default:break}}if(m&&e.gl.uniform2i(m,l.texData.texShape[0],l.texData.texShape[1]),d!=null){if(l.isUniform){if(v.sizeFromShape(l.shape)<2)e.gl.uniform1f(d,l.uniformValues[0]);else{let f=l.uniformValues;f instanceof Float32Array||(f=new Float32Array(f)),e.gl.uniform1fv(d,f)}return}l.texData.slice!=null&&c!=null&&e.gl.uniform1i(c,l.texData.slice.flatOffset),e.setInputMatrixTexture(l.texData.texture.texture,d,u)}});let o=t.outShapeLocation;if(o)switch(a.shape.length){case 1:e.gl.uniform1iv(o,new Int32Array(a.shape));break;case 2:e.gl.uniform2iv(o,new Int32Array(a.shape));break;case 3:e.gl.uniform3iv(o,new Int32Array(a.shape));break;case 4:e.gl.uniform4iv(o,new Int32Array(a.shape));break;default:break}if(t.outShapeStridesLocation){let l=v.computeStrides(a.shape);switch(a.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(l));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(l));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(l));break;default:break}}t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,a.texData.texShape[0],a.texData.texShape[1]),t.program.customUniforms&&r&&t.program.customUniforms.forEach((l,u)=>{let p=t.customUniformLocations[u],d=r[u];if(l.type==="float")e.gl.uniform1fv(p,d);else if(l.type==="vec2")e.gl.uniform2fv(p,d);else if(l.type==="vec3")e.gl.uniform3fv(p,d);else if(l.type==="vec4")e.gl.uniform4fv(p,d);else if(l.type==="int")e.gl.uniform1iv(p,d);else if(l.type==="ivec2")e.gl.uniform2iv(p,d);else if(l.type==="ivec3")e.gl.uniform3iv(p,d);else if(l.type==="ivec4")e.gl.uniform4iv(p,d);else throw Error(`uniform type ${l.type} is not supported yet.`)}),e.executeProgram()}function MZ(e,t,n){let a="";t.concat(n).forEach(i=>{let o=i.texData!=null&&i.texData.slice!=null&&i.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!i.isUniform){let l=i.texData.texShape,{useSqueezeShape:u,uniformShape:p,keptDims:d}=H0(e.packedInputs,i.shape,l),c="",h="",m="";if(p.length===1&&e.packedInputs){let I=[Math.ceil(l[0]/2),Math.ceil(l[1]/2)];c=`${I[0]>1}_${I[1]>1}`}else if(p.length===2&&!e.packedInputs)h=`${p[0]>1}_${p[1]>1}`;else if(p.length>2&&!e.packedInputs){let I=v.computeStrides(p);m=`${I[0]===l[1]}_${I[I.length-1]===l[1]}`}let f=i.shape.length,g=p.length===2&&v.arraysEqual(i.shape,l),y=v.sizeFromShape(i.shape)===1,b=N.getBroadcastDims(i.shape,n.shape),x=!e.packedInputs&&f===n.shape.length&&v.arraysEqual(l,n.texData.texShape),w=e.packedInputs||p.length>2?"":`${l[0]>1}_${l[1]>1}`;a+=`${f}_${x}_${u?d:""}_${p.length}_${y}_${b}_${g}_${c}_${h}_${m}_${w}_${o}`}else{let l=i.isUniform?"uniform":i.texData.texShape;a+=`${i.shape}_${l}_${o}`}});let r=e.userCode,s=e.constructor.name;return s+="_"+a+"_"+r+`${H().getNumber("WEBGL_VERSION")}`,s}function _n(e){return H().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var PZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=Jp.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=Cn();this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length),this.userCode=`
+  `;
+}
+function getCoordsDataType(rank) {
+  if (rank <= 1) {
+    return "int";
+  } else if (rank === 2) {
+    return "ivec2";
+  } else if (rank === 3) {
+    return "ivec3";
+  } else if (rank === 4) {
+    return "ivec4";
+  } else if (rank === 5) {
+    return "ivec5";
+  } else if (rank === 6) {
+    return "ivec6";
+  } else {
+    throw Error(`GPU for rank ${rank} is not yet supported`);
+  }
+}
+function getUniformInfoFromShape(isPacked, shape, texShape) {
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const rank = shape.length;
+  const useSqueezePackedShape = isPacked && rank === 3 && shape[0] === 1;
+  const squeezeShape2 = useSqueezePackedShape ? shape.slice(1) : newShape;
+  const useSqueezeShape = !isPacked && rank > 1 && !util_exports.arraysEqual(shape, texShape) && newShape.length < rank || useSqueezePackedShape;
+  const uniformShape = useSqueezeShape ? squeezeShape2 : shape;
+  return { useSqueezeShape, uniformShape, keptDims };
+}
+function squeezeInputInfo(inInfo, squeezedShape) {
+  const newInputInfo = JSON.parse(JSON.stringify(inInfo));
+  newInputInfo.shapeInfo.logicalShape = squeezedShape;
+  return newInputInfo;
+}
+function getSqueezedParams(params, keptDims) {
+  return keptDims.map((d) => params[d]).join(", ");
+}
+function compileProgram(gpgpu, program, inputs, output) {
+  const inputInfos = inputs.map((input2, i) => {
+    const shapeInfo = {
+      logicalShape: input2.shape,
+      texShape: input2.isUniform ? null : input2.texData.texShape,
+      isUniform: input2.isUniform,
+      isPacked: input2.isUniform ? false : input2.texData.isPacked,
+      flatOffset: null
+    };
+    if (input2.texData != null && input2.texData.slice != null && input2.texData.slice.flatOffset > 0) {
+      shapeInfo.flatOffset = input2.texData.slice.flatOffset;
+    }
+    return { name: program.variableNames[i], shapeInfo };
+  });
+  const inShapeInfos = inputInfos.map((x) => x.shapeInfo);
+  const outShapeInfo = {
+    logicalShape: output.shape,
+    texShape: output.texData.texShape,
+    isUniform: false,
+    isPacked: output.texData.isPacked,
+    flatOffset: null
+  };
+  const source = makeShader(inputInfos, outShapeInfo, program);
+  const fragmentShader = createFragmentShader(gpgpu.gl, source);
+  const webGLProgram = gpgpu.createProgram(fragmentShader);
+  if (!env().get("ENGINE_COMPILE_ONLY")) {
+    return Object.assign({
+      program,
+      fragmentShader,
+      source,
+      webGLProgram,
+      inShapeInfos,
+      outShapeInfo
+    }, getUniformLocations(gpgpu, program, webGLProgram));
+  } else {
+    return {
+      program,
+      fragmentShader,
+      source,
+      webGLProgram,
+      inShapeInfos,
+      outShapeInfo,
+      uniformLocations: null,
+      customUniformLocations: null,
+      infLoc: null,
+      nanLoc: null,
+      inShapesLocations: null,
+      inTexShapesLocations: null,
+      outShapeLocation: null,
+      outShapeStridesLocation: null,
+      outTexShapeLocation: null
+    };
+  }
+}
+function getUniformLocations(gpgpu, program, webGLProgram) {
+  const uniformLocations = {};
+  const inShapesLocations = {};
+  const inTexShapesLocations = {};
+  const customUniformLocations = [];
+  let outShapeLocation;
+  let outTexShapeLocation;
+  let outShapeStridesLocation;
+  let infLoc = null;
+  let nanLoc = null;
+  nanLoc = gpgpu.getUniformLocation(webGLProgram, "NAN", false);
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    infLoc = gpgpu.getUniformLocation(webGLProgram, "INFINITY", false);
+  }
+  const shouldThrow = false;
+  for (let i = 0; i < program.variableNames.length; i++) {
+    const varName = program.variableNames[i];
+    uniformLocations[varName] = gpgpu.getUniformLocation(webGLProgram, varName, shouldThrow);
+    uniformLocations[`offset${varName}`] = gpgpu.getUniformLocation(webGLProgram, `offset${varName}`, shouldThrow);
+    if (program.enableShapeUniforms) {
+      inShapesLocations[`${varName}Shape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}Shape`, shouldThrow);
+      inTexShapesLocations[`${varName}TexShape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}TexShape`, shouldThrow);
+    }
+  }
+  if (program.enableShapeUniforms) {
+    outShapeLocation = gpgpu.getUniformLocation(webGLProgram, "outShape", shouldThrow);
+    outShapeStridesLocation = gpgpu.getUniformLocation(webGLProgram, "outShapeStrides", shouldThrow);
+    outTexShapeLocation = gpgpu.getUniformLocation(webGLProgram, "outTexShape", shouldThrow);
+  }
+  if (program.customUniforms) {
+    program.customUniforms.forEach((d, i) => {
+      customUniformLocations[i] = gpgpu.getUniformLocation(webGLProgram, d.name, shouldThrow);
+    });
+  }
+  return {
+    uniformLocations,
+    customUniformLocations,
+    infLoc,
+    nanLoc,
+    inShapesLocations,
+    inTexShapesLocations,
+    outShapeLocation,
+    outShapeStridesLocation,
+    outTexShapeLocation
+  };
+}
+function validateBinaryAndProgram(shapeInfos, inputs) {
+  if (shapeInfos.length !== inputs.length) {
+    throw Error(`Binary was compiled with ${shapeInfos.length} inputs, but was executed with ${inputs.length} inputs`);
+  }
+  shapeInfos.forEach((s, i) => {
+    const shapeA = s.logicalShape;
+    const input2 = inputs[i];
+    const shapeB = input2.shape;
+    if (!util_exports.arraysEqual(shapeA, shapeB)) {
+      throw Error(`Binary was compiled with different shapes than the current args. Shapes ${shapeA} and ${shapeB} must match`);
+    }
+    if (s.isUniform && input2.isUniform) {
+      return;
+    }
+    const texShapeA = s.texShape;
+    const texShapeB = input2.isUniform ? null : input2.texData.texShape;
+    if (!util_exports.arraysEqual(texShapeA, texShapeB)) {
+      throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${texShapeA} and ${texShapeB} must match`);
+    }
+  });
+}
+function runProgram(gpgpu, binary, inputs, output, customUniformValues) {
+  if (!binary.program.enableShapeUniforms) {
+    validateBinaryAndProgram(binary.inShapeInfos, inputs);
+    validateBinaryAndProgram([binary.outShapeInfo], [output]);
+  }
+  const outTex = output.texData.texture;
+  const outTexShape = output.texData.texShape;
+  if (output.texData.isPacked) {
+    gpgpu.setOutputPackedMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);
+  } else {
+    gpgpu.setOutputMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);
+  }
+  gpgpu.setProgram(binary.webGLProgram);
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    if (binary.infLoc !== null) {
+      gpgpu.gl.uniform1f(binary.infLoc, Infinity);
+    }
+  }
+  if (binary.nanLoc !== null) {
+    gpgpu.gl.uniform1f(binary.nanLoc, NaN);
+  }
+  inputs.forEach((input2, i) => {
+    const varName = binary.program.variableNames[i];
+    const varLoc = binary.uniformLocations[varName];
+    const varOffsetLoc = binary.uniformLocations[`offset${varName}`];
+    const varShapeLoc = binary.inShapesLocations[`${varName}Shape`];
+    const varTexShapeLoc = binary.inTexShapesLocations[`${varName}TexShape`];
+    if (varShapeLoc) {
+      const { uniformShape } = getUniformInfoFromShape(binary.program.packedInputs, input2.shape, input2.texData.texShape);
+      switch (uniformShape.length) {
+        case 1:
+          gpgpu.gl.uniform1iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 2:
+          gpgpu.gl.uniform2iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 3:
+          gpgpu.gl.uniform3iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 4:
+          gpgpu.gl.uniform4iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        default:
+          break;
+      }
+    }
+    if (varTexShapeLoc) {
+      gpgpu.gl.uniform2i(varTexShapeLoc, input2.texData.texShape[0], input2.texData.texShape[1]);
+    }
+    if (varLoc == null) {
+      return;
+    }
+    if (input2.isUniform) {
+      if (util_exports.sizeFromShape(input2.shape) < 2) {
+        gpgpu.gl.uniform1f(varLoc, input2.uniformValues[0]);
+      } else {
+        let vals = input2.uniformValues;
+        if (!(vals instanceof Float32Array)) {
+          vals = new Float32Array(vals);
+        }
+        gpgpu.gl.uniform1fv(varLoc, vals);
+      }
+      return;
+    }
+    if (input2.texData.slice != null && varOffsetLoc != null) {
+      gpgpu.gl.uniform1i(varOffsetLoc, input2.texData.slice.flatOffset);
+    }
+    gpgpu.setInputMatrixTexture(input2.texData.texture.texture, varLoc, i);
+  });
+  const outShapeLoc = binary.outShapeLocation;
+  if (outShapeLoc) {
+    switch (output.shape.length) {
+      case 1:
+        gpgpu.gl.uniform1iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 2:
+        gpgpu.gl.uniform2iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 3:
+        gpgpu.gl.uniform3iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 4:
+        gpgpu.gl.uniform4iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      default:
+        break;
+    }
+  }
+  if (binary.outShapeStridesLocation) {
+    const strides = util_exports.computeStrides(output.shape);
+    switch (output.shape.length) {
+      case 2:
+        gpgpu.gl.uniform1iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      case 3:
+        gpgpu.gl.uniform2iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      case 4:
+        gpgpu.gl.uniform3iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      default:
+        break;
+    }
+  }
+  if (binary.outTexShapeLocation) {
+    gpgpu.gl.uniform2i(binary.outTexShapeLocation, output.texData.texShape[0], output.texData.texShape[1]);
+  }
+  if (binary.program.customUniforms && customUniformValues) {
+    binary.program.customUniforms.forEach((d, i) => {
+      const customLoc = binary.customUniformLocations[i];
+      const customValue = customUniformValues[i];
+      if (d.type === "float") {
+        gpgpu.gl.uniform1fv(customLoc, customValue);
+      } else if (d.type === "vec2") {
+        gpgpu.gl.uniform2fv(customLoc, customValue);
+      } else if (d.type === "vec3") {
+        gpgpu.gl.uniform3fv(customLoc, customValue);
+      } else if (d.type === "vec4") {
+        gpgpu.gl.uniform4fv(customLoc, customValue);
+      } else if (d.type === "int") {
+        gpgpu.gl.uniform1iv(customLoc, customValue);
+      } else if (d.type === "ivec2") {
+        gpgpu.gl.uniform2iv(customLoc, customValue);
+      } else if (d.type === "ivec3") {
+        gpgpu.gl.uniform3iv(customLoc, customValue);
+      } else if (d.type === "ivec4") {
+        gpgpu.gl.uniform4iv(customLoc, customValue);
+      } else {
+        throw Error(`uniform type ${d.type} is not supported yet.`);
+      }
+    });
+  }
+  gpgpu.executeProgram();
+}
+function makeShaderKey(program, inputs, output) {
+  let keyInputs = "";
+  inputs.concat(output).forEach((x) => {
+    const hasOffset = x.texData != null && x.texData.slice != null && x.texData.slice.flatOffset > 0;
+    if (program.enableShapeUniforms && !x.isUniform) {
+      const xTexShape = x.texData.texShape;
+      const { useSqueezeShape, uniformShape, keptDims } = getUniformInfoFromShape(program.packedInputs, x.shape, xTexShape);
+      let rank1 = "", rank2 = "", rank34 = "";
+      if (uniformShape.length === 1 && program.packedInputs) {
+        const packedTexShape = [Math.ceil(xTexShape[0] / 2), Math.ceil(xTexShape[1] / 2)];
+        rank1 = `${packedTexShape[0] > 1}_${packedTexShape[1] > 1}`;
+      } else if (uniformShape.length === 2 && !program.packedInputs) {
+        rank2 = `${uniformShape[0] > 1}_${uniformShape[1] > 1}`;
+      } else if (uniformShape.length > 2 && !program.packedInputs) {
+        const strides = util_exports.computeStrides(uniformShape);
+        rank34 = `${strides[0] === xTexShape[1]}_${strides[strides.length - 1] === xTexShape[1]}`;
+      }
+      const xRank = x.shape.length;
+      const isLogicalShapTexShapeEqual = uniformShape.length === 2 && util_exports.arraysEqual(x.shape, xTexShape);
+      const isScalar = util_exports.sizeFromShape(x.shape) === 1;
+      const broadcastDims = backend_util_exports.getBroadcastDims(x.shape, output.shape);
+      const isInOutTexShapeEqual = !program.packedInputs && xRank === output.shape.length && util_exports.arraysEqual(xTexShape, output.texData.texShape);
+      const isTexShapeGreaterThanOne = program.packedInputs || uniformShape.length > 2 ? "" : `${xTexShape[0] > 1}_${xTexShape[1] > 1}`;
+      keyInputs += `${xRank}_${isInOutTexShapeEqual}_${useSqueezeShape ? keptDims : ""}_${uniformShape.length}_${isScalar}_${broadcastDims}_${isLogicalShapTexShapeEqual}_${rank1}_${rank2}_${rank34}_${isTexShapeGreaterThanOne}_${hasOffset}`;
+    } else {
+      const texShape = x.isUniform ? "uniform" : x.texData.texShape;
+      keyInputs += `${x.shape}_${texShape}_${hasOffset}`;
+    }
+  });
+  const keyUserCode = program.userCode;
+  let key = program.constructor.name;
+  key += "_" + keyInputs + "_" + keyUserCode + `${env().getNumber("WEBGL_VERSION")}`;
+  return key;
+}
+function useShapeUniforms(rank) {
+  return env().getBool("WEBGL_USE_SHAPES_UNIFORMS") && rank <= 4;
+}
+var DecodeMatrixProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.outPackingScheme = PackingScheme.DENSE;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Df(["r","c","d"],e):vo(["r","c","d"],e)}
+        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], outputShape) : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], outputShape)}
         return ivec3(r, c, d);
       }
 
@@ -1002,11 +50187,24 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           result[i] = getA(rc.x, rc.y, rc.z);
         }
 
-        ${t.output} = result;
+        ${glsl.output} = result;
       }
-    `}},OZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=Jp.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=Cn();this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length),this.userCode=`
+    `;
+  }
+};
+var DecodeMatrixPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outPackingScheme = PackingScheme.DENSE;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Df(["r","c","d"],e):vo(["r","c","d"],e)}
+        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], outputShape) : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], outputShape)}
         return ivec3(r, c, d);
       }
 
@@ -1022,52 +50220,117 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));
         }
 
-        ${t.output} = result;
+        ${glsl.output} = result;
       }
-    `}},LZ=class{constructor(e){this.variableNames=["A"],this.outTexUsage=pa.DOWNLOAD;let t=Cn();this.outputShape=e,this.userCode=`
-      ${f_}
+    `;
+  }
+};
+var EncodeFloatProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.outTexUsage = TextureUsage.DOWNLOAD;
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.userCode = `
+      ${ENCODE_FLOAT_SNIPPET}
 
       void main() {
         float x = getAAtOutCoords();
-        ${t.output} = encode_float(x);
+        ${glsl.output} = encode_float(x);
       }
-    `}},zZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=pa.DOWNLOAD;let t=Cn();this.outputShape=e,this.userCode=`
-      ${f_}
+    `;
+  }
+};
+var EncodeFloatPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = false;
+    this.outTexUsage = TextureUsage.DOWNLOAD;
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.userCode = `
+      ${ENCODE_FLOAT_SNIPPET}
 
       void main() {
         ivec3 coords = getOutputCoords();
         float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));
-        ${t.output} = encode_float(x);
+        ${glsl.output} = encode_float(x);
       }
-    `}},WZ={R:0,G:1,B:2,A:3},jk=class{constructor(e,t=!1,n="RGBA"){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let a=Cn();this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length);let r="result";t&&(r="floor(result * 255. + 0.5)");let s="";for(let i=0;i<n.length;i++){let o=n[i];s+=`
-          if(offset == ${i}) {
-            result = values[${WZ[o]}];
-          }`}this.userCode=`
-      ${this.enableShapeUniforms?G0():U0(e)}
+    `;
+  }
+};
+var CHANNEL_CHAR_TO_INDEX_MAP = {
+  "R": 0,
+  "G": 1,
+  "B": 2,
+  "A": 3
+};
+var EncodeMatrixProgram = class {
+  constructor(outputShape, inputIsUnsignedByte = false, usedChannels = "RGBA") {
+    this.variableNames = ["A"];
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let output = `result`;
+    if (inputIsUnsignedByte) {
+      output = `floor(result * 255. + 0.5)`;
+    }
+    let mainLoop = "";
+    for (let usedChannelIndex = 0; usedChannelIndex < usedChannels.length; usedChannelIndex++) {
+      const curChannel = usedChannels[usedChannelIndex];
+      mainLoop += `
+          if(offset == ${usedChannelIndex}) {
+            result = values[${CHANNEL_CHAR_TO_INDEX_MAP[curChannel]}];
+          }`;
+    }
+    this.userCode = `
+      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
       void main() {
         ivec3 coords = getOutputCoords();
         int flatIndex = getFlatIndex(coords);
         float result = 0.;
-        int offset = imod(flatIndex, ${n.length});
+        int offset = imod(flatIndex, ${usedChannels.length});
 
-        flatIndex = idiv(flatIndex, ${n.length}, 1.);
+        flatIndex = idiv(flatIndex, ${usedChannels.length}, 1.);
 
         int r = flatIndex / texShape[1];
         if (r < texShape[0]) {
           int c = imod(flatIndex, texShape[1]);
           vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
-          vec4 values = ${a.texture2D}(A, uv);
-          ${s}
+          vec4 values = ${glsl.texture2D}(A, uv);
+          ${mainLoop}
         }
-        ${a.output} = vec4(${r}, 0., 0., 0.);
+        ${glsl.output} = vec4(${output}, 0., 0., 0.);
       }
-    `}},BZ=class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let n=Cn();this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length);let a="",r="result";t&&(r="floor(result * 255. + 0.5)");for(let s=0;s<=1;s++)for(let i=0;i<=1;i++){let o=s*2+i;a+=`
+    `;
+  }
+};
+var EncodeMatrixPackedProgram = class {
+  constructor(outputShape, inputIsUnsignedByte = false) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let mainLoop = "";
+    let output = "result";
+    if (inputIsUnsignedByte) {
+      output = "floor(result * 255. + 0.5)";
+    }
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        const channel = row * 2 + col;
+        mainLoop += `
           localCoords = coords;
-          if(localCoords[2] + ${i} < ${this.enableShapeUniforms?"outShape[2]":`${e[2]}`}) {
-          localCoords[2] += ${i};
-          if (localCoords[1] + ${s} < ${this.enableShapeUniforms?"outShape[1]":`${e[1]}`}) {
-            localCoords[1] += ${s};
+          if(localCoords[2] + ${col} < ${this.enableShapeUniforms ? "outShape[2]" : `${outputShape[2]}`}) {
+          localCoords[2] += ${col};
+          if (localCoords[1] + ${row} < ${this.enableShapeUniforms ? "outShape[1]" : `${outputShape[1]}`}) {
+            localCoords[1] += ${row};
 
             flatIndex = getFlatIndex(localCoords);
             offset = imod(flatIndex, 4);
@@ -1077,21 +50340,24 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             int r = flatIndex / texShape[1];
             int c = imod(flatIndex, texShape[1]);
             vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
-            values = ${n.texture2D}(A, uv);
+            values = ${glsl.texture2D}(A, uv);
 
             if (offset == 0) {
-              result[${o}] = values[0];
+              result[${channel}] = values[0];
             } else if (offset == 1) {
-              result[${o}] = values[1];
+              result[${channel}] = values[1];
             } else if (offset == 2) {
-              result[${o}] = values[2];
+              result[${channel}] = values[2];
             } else {
-              result[${o}] = values[3];
+              result[${channel}] = values[3];
             }
           }
         }
-        `}this.userCode=`
-        ${this.enableShapeUniforms?G0():U0(e)}
+        `;
+      }
+    }
+    this.userCode = `
+        ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
         void main() {
           ivec3 coords = getOutputCoords();
@@ -1102,60 +50368,739 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           vec2 uv;
           vec4 values;
 
-          ${a}
+          ${mainLoop}
 
-          ${n.output} = ${r};
+          ${glsl.output} = ${output};
         }
-    `}},v_={};Ae(v_,{bindVertexProgramAttributeStreams:()=>E_,createBufferFromOutputTexture:()=>F_,createFloat16MatrixTexture:()=>T_,createFloat16PackedMatrixTexture:()=>__,createFloat32MatrixTexture:()=>S_,createIndexBuffer:()=>I_,createPackedMatrixTexture:()=>C_,createUnsignedBytesMatrixTexture:()=>N_,createVertexBuffer:()=>k_,createVertexShader:()=>w_,downloadByteEncodedFloatMatrixFromOutputTexture:()=>R_,downloadFloat32MatrixFromBuffer:()=>D_,downloadMatrixFromPackedOutputTexture:()=>P_,downloadPackedMatrixFromBuffer:()=>M_,getInternalFormatForFloat16MatrixTexture:()=>q0,getInternalFormatForFloat16PackedMatrixTexture:()=>Y0,getInternalFormatForFloat32MatrixTexture:()=>j0,getInternalFormatForPackedMatrixTexture:()=>X0,getInternalFormatForUnsignedBytesMatrixTexture:()=>K0,uploadDenseMatrixToTexture:()=>A_,uploadPixelDataToTexture:()=>$_});function w_(e){let t=Cn(),n=`${t.version}
+    `;
+  }
+};
+var gpgpu_util_exports = {};
+__export2(gpgpu_util_exports, {
+  bindVertexProgramAttributeStreams: () => bindVertexProgramAttributeStreams,
+  createBufferFromOutputTexture: () => createBufferFromOutputTexture,
+  createFloat16MatrixTexture: () => createFloat16MatrixTexture,
+  createFloat16PackedMatrixTexture: () => createFloat16PackedMatrixTexture,
+  createFloat32MatrixTexture: () => createFloat32MatrixTexture,
+  createIndexBuffer: () => createIndexBuffer,
+  createPackedMatrixTexture: () => createPackedMatrixTexture,
+  createUnsignedBytesMatrixTexture: () => createUnsignedBytesMatrixTexture,
+  createVertexBuffer: () => createVertexBuffer,
+  createVertexShader: () => createVertexShader2,
+  downloadByteEncodedFloatMatrixFromOutputTexture: () => downloadByteEncodedFloatMatrixFromOutputTexture,
+  downloadFloat32MatrixFromBuffer: () => downloadFloat32MatrixFromBuffer,
+  downloadMatrixFromPackedOutputTexture: () => downloadMatrixFromPackedOutputTexture,
+  downloadPackedMatrixFromBuffer: () => downloadPackedMatrixFromBuffer,
+  getInternalFormatForFloat16MatrixTexture: () => getInternalFormatForFloat16MatrixTexture,
+  getInternalFormatForFloat16PackedMatrixTexture: () => getInternalFormatForFloat16PackedMatrixTexture,
+  getInternalFormatForFloat32MatrixTexture: () => getInternalFormatForFloat32MatrixTexture,
+  getInternalFormatForPackedMatrixTexture: () => getInternalFormatForPackedMatrixTexture,
+  getInternalFormatForUnsignedBytesMatrixTexture: () => getInternalFormatForUnsignedBytesMatrixTexture,
+  uploadDenseMatrixToTexture: () => uploadDenseMatrixToTexture,
+  uploadPixelDataToTexture: () => uploadPixelDataToTexture
+});
+function createVertexShader2(gl) {
+  const glsl = getGlslDifferences();
+  const vertexShaderSource = `${glsl.version}
     precision highp float;
-    ${t.attribute} vec3 clipSpacePos;
-    ${t.attribute} vec2 uv;
-    ${t.varyingVs} vec2 resultUV;
+    ${glsl.attribute} vec3 clipSpacePos;
+    ${glsl.attribute} vec2 uv;
+    ${glsl.varyingVs} vec2 resultUV;
 
     void main() {
       gl_Position = vec4(clipSpacePos, 1);
       resultUV = uv;
-    }`;return qC(e,n)}function k_(e){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return ZC(e,t)}function I_(e){let t=new Uint16Array([0,1,2,2,1,3]);return JC(e,t)}function td(e,t,n,a,r,s){e_(t,n);let i=QC(e),o=e.TEXTURE_2D;return me(e,()=>e.bindTexture(o,i)),me(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(o,e.TEXTURE_MIN_FILTER,e.NEAREST)),me(e,()=>e.texParameteri(o,e.TEXTURE_MAG_FILTER,e.NEAREST)),H().getNumber("WEBGL_VERSION")===1?me(e,()=>e.texImage2D(o,0,a,t,n,0,r,s,null)):me(e,()=>e.texStorage2D(o,1,a,t,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:i,texShape:[n,t]}}function j0(e){return e.internalFormatFloat}function S_(e,t,n,a){let[r,s]=ed(t,n);return td(e,r,s,j0(a),a.textureFormatFloat,e.FLOAT)}function q0(e){return e.internalFormatHalfFloat}function T_(e,t,n,a){let[r,s]=ed(t,n);return td(e,r,s,q0(a),a.textureFormatFloat,a.textureTypeHalfFloat)}function K0(e){return e.downloadTextureFormat}function N_(e,t,n,a){let[r,s]=ed(t,n);return td(e,r,s,K0(a),e.RGBA,e.UNSIGNED_BYTE)}function X0(e){return e.internalFormatPackedFloat}function C_(e,t,n,a){let[r,s]=Ou(t,n);return td(e,r,s,X0(a),e.RGBA,e.FLOAT)}function Y0(e){return e.internalFormatPackedHalfFloat}function __(e,t,n,a){let[r,s]=Ou(t,n);return td(e,r,s,Y0(a),e.RGBA,a.textureTypeHalfFloat)}function E_(e,t,n){return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),ix(e,t,"clipSpacePos",n,3,20,0)&&ix(e,t,"uv",n,2,20,12)}function A_(e,t,n,a,r,s){me(e,()=>e.bindTexture(e.TEXTURE_2D,t));let i,o,l;r instanceof Uint8Array?(i=new Uint8Array(n*a*4),o=e.UNSIGNED_BYTE,l=e.RGBA):(i=new Float32Array(n*a*4),o=e.FLOAT,l=s.internalFormatPackedFloat),i.set(r),H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n,a,e.RGBA,o,i)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,l,n,a,0,e.RGBA,o,i)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function $_(e,t,n){me(e,()=>e.bindTexture(e.TEXTURE_2D,t)),n.data instanceof Uint8Array?H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n.width,n.height,e.RGBA,e.UNSIGNED_BYTE,n.data)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,n.width,n.height,0,e.RGBA,e.UNSIGNED_BYTE,n.data)):H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,n)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function F_(e,t,n,a){let r=e.createBuffer();me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,r));let s=4*4*t*n;return me(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,s,e.STREAM_READ)),me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,0)),me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),r}function D_(e,t,n){let a=e,r=new Float32Array(n);return a.bindBuffer(a.PIXEL_PACK_BUFFER,t),a.getBufferSubData(a.PIXEL_PACK_BUFFER,0,r),a.bindBuffer(a.PIXEL_PACK_BUFFER,null),r}function R_(e,t,n,a){let[r,s]=ed(t,n),i=4,o=new Uint8Array(OY(t*n,i));return me(e,()=>e.readPixels(0,0,r,s,a.downloadTextureFormat,e.UNSIGNED_BYTE,o)),new Float32Array(o.buffer)}function M_(e,t,n,a,r,s,i,o){let l=e,u=new Float32Array(LY(s,i));return l.bindBuffer(l.PIXEL_PACK_BUFFER,t),l.getBufferSubData(l.PIXEL_PACK_BUFFER,0,u),l.bindBuffer(l.PIXEL_PACK_BUFFER,null),u}function P_(e,t,n){let a=new Float32Array(t*n*4);return me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,a)),a}var ph=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let t=H().getNumber("WEBGL_VERSION");e!=null?(this.gl=e,GC(t,e)):this.gl=qa(t);let n="WEBGL_color_buffer_float",a="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),H().getNumber("WEBGL_VERSION")===1){let r="OES_texture_float",s="OES_texture_half_float";if(this.textureFloatExtension=Dp(this.gl,r),ca(this.gl,s))this.textureHalfFloatExtension=Dp(this.gl,s);else if(H().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),ca(this.gl,a))this.colorBufferHalfFloatExtension=Dp(this.gl,a);else if(H().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",ca(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else if(ca(this.gl,a))this.colorBufferHalfFloatExtension=this.gl.getExtension(a);else throw new Error("GL context does not support color renderable floats");this.vertexBuffer=k_(this.gl),this.indexBuffer=I_(this.gl),this.framebuffer=t_(this.gl),this.textureConfig=B0(this.gl,this.textureHalfFloatExtension)}get debug(){return H().getBool("DEBUG")}dispose(){if(this.disposed)return;this.program!=null&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),this.outputTexture!=null&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let e=this.gl;me(e,()=>e.finish()),me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.deleteFramebuffer(this.framebuffer)),me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),me(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),S_(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),T_(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),N_(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),$_(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,n,a){this.throwIfDisposed(),A_(this.gl,e,t,n,a,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),__(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),C_(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(ox(this.gl,this.framebuffer),this.outputTexture=null),me(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,n){return this.downloadMatrixDriver(e,()=>R_(this.gl,t,n,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,n,a,r,s){return M_(this.gl,e,t,n,a,r,s,this.textureConfig)}downloadFloat32MatrixFromBuffer(e,t){return D_(this.gl,e,t)}createBufferFromTexture(e,t,n){this.bindTextureToFrameBuffer(e);let a=F_(this.gl,t,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),a}createAndWaitForFence(){let e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,n;if(H().getBool("WEBGL_FENCE_API_ENABLED")){let a=e,r=a.fenceSync(a.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),n=()=>{let s=a.clientWaitSync(r,0,0);return s===a.ALREADY_SIGNALED||s===a.CONDITION_SATISFIED},t=r}else H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(t,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:t,isFencePassed:n}}downloadMatrixFromPackedTexture(e,t,n){return this.downloadMatrixDriver(e,()=>P_(this.gl,t,n))}createProgram(e){this.throwIfDisposed();let t=this.gl;this.vertexShader==null&&(this.vertexShader=w_(t));let n=XC(t);return me(t,()=>t.attachShader(n,this.vertexShader)),me(t,()=>t.attachShader(n,e)),YC(t,n),this.debug&&ih(t,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=E_(t,this.program,this.vertexBuffer)),n}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),e!=null&&me(this.gl,()=>this.gl.deleteProgram(e))}setProgram(e){this.throwIfDisposed(),this.program=e,this.program!=null&&this.debug&&ih(this.gl,this.program),me(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,n=!0){return this.throwIfDisposed(),n?a_(this.gl,e,t):r_(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),me(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,n){this.throwIfDisposed(),this.throwIfNoProgram(),s_(this.gl,e,t,n)}setOutputMatrixTexture(e,t,n){this.setOutputMatrixTextureDriver(e,n,t)}setOutputPackedMatrixTexture(e,t,n){this.throwIfDisposed();let[a,r]=Ou(t,n);this.setOutputMatrixTextureDriver(e,a,r)}setOutputMatrixWriteRegion(e,t,n,a){this.setOutputMatrixWriteRegionDriver(n,e,a,t)}setOutputPackedMatrixWriteRegion(e,t,n,a){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){this.program!=null&&ih(this.gl,this.program),Rp(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let e=this.gl;this.debug&&this.debugValidate(),me(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),me(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=Dp(this.gl,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let n=this.gl,a=this.getQueryTimerExtensionWebGL2(),r=n.createQuery();return n.beginQuery(a.TIME_ELAPSED_EXT,r),r}let e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let t=this.gl,n=this.getQueryTimerExtensionWebGL2();t.endQuery(n.TIME_ELAPSED_EXT);return}let e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await v.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(t===0)return null;if(t===2){let n=this.gl;return n.getQueryParameter(e,n.QUERY_RESULT)/1e6}else{let n=this.getQueryTimerExtensionWebGL1();return n.getQueryObjectEXT(e,n.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(t===0)return!0;if(t===2){let n=this.gl,a=this.getQueryTimerExtensionWebGL2(),r=n.getQueryParameter(e,n.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(a.GPU_DISJOINT_EXT)),r&&!this.disjoint}else{let n=this.getQueryTimerExtensionWebGL1(),a=n.getQueryObjectEXT(e,n.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),a&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){let e=VZ(this.itemsToPoll.map(t=>t.isDoneFn));for(let t=0;t<=e;++t){let{resolveFn:n}=this.itemsToPoll[t];n()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){if(this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),this.itemsToPoll.length>1)return;let n;"setTimeoutCustom"in H().platform&&(n=H().platform.setTimeoutCustom.bind(H().platform)),v.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0),()=>0,null,n)}bindTextureToFrameBuffer(e){this.throwIfDisposed(),oh(this.gl,e,this.framebuffer),this.debug&&Rp(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(oh(this.gl,this.outputTexture,this.framebuffer),this.debug&&Rp(this.gl)):ox(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);let n=t();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(e,t,n){this.throwIfDisposed();let a=this.gl;oh(a,e,this.framebuffer),this.debug&&Rp(a),this.outputTexture=e,me(a,()=>a.viewport(0,0,t,n)),me(a,()=>a.scissor(0,0,t,n))}setOutputMatrixWriteRegionDriver(e,t,n,a){this.throwIfDisposed(),me(this.gl,()=>this.gl.scissor(e,t,n,a))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(this.program==null)throw new Error("No GPU program is currently set.")}};function VZ(e){let t=0;for(;t<e.length&&e[t]();++t);return t-1}var{addImpl:UZ,bincountImpl:O_,bincountReduceImpl:GZ,castImpl:HZ,ceilImpl:jZ,concatImpl:qZ,equalImpl:KZ,expImpl:XZ,expm1Impl:YZ,floorImpl:ZZ,gatherNdImpl:JZ,gatherV2Impl:QZ,greaterImpl:e7,greaterEqualImpl:t7,lessImpl:n7,lessEqualImpl:a7,linSpaceImpl:r7,logImpl:s7,maxImpl:i7,maximumImpl:o7,minimumImpl:l7,multiplyImpl:u7,negImpl:p7,notEqualImpl:c7,prodImpl:d7,raggedGatherImpl:h7,raggedRangeImpl:m7,raggedTensorToTensorImpl:f7,rangeImpl:g7,rsqrtImpl:y7,scatterImpl:b7,sigmoidImpl:x7,simpleAbsImpl:L_,sliceImpl:v7,sparseFillEmptyRowsImpl:w7,sparseReshapeImpl:k7,sparseSegmentReductionImpl:z_,sqrtImpl:I7,stridedSliceImpl:S7,stringNGramsImpl:T7,stringSplitImpl:N7,stringToHashBucketFastImpl:C7,subImpl:_7,tileImpl:E7,topKImpl:A7,transposeImpl:Z0,uniqueImpl:$7}=B2;function W_(e,t){return["x","y","z","w","u","v"].slice(0,t).map(n=>`${e}.${n}`)}function wn(e,t){return t===1?[e]:W_(e,t)}function F7(e,t){if(e===1)return"rc";let n="";for(let a=0;a<e;a++)n+=t[a],a<e-1&&(n+=",");return n}var D7=class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=_n(this.outputShape.length),this.rank===0)this.userCode=`
+    }`;
+  return createVertexShader(gl, vertexShaderSource);
+}
+function createVertexBuffer(gl) {
+  const vertexArray = new Float32Array([-1, 1, 0, 0, 1, -1, -1, 0, 0, 0, 1, 1, 0, 1, 1, 1, -1, 0, 1, 0]);
+  return createStaticVertexBuffer(gl, vertexArray);
+}
+function createIndexBuffer(gl) {
+  const triangleVertexIndices = new Uint16Array([0, 1, 2, 2, 1, 3]);
+  return createStaticIndexBuffer(gl, triangleVertexIndices);
+}
+function createAndConfigureTexture(gl, width, height, internalFormat, textureFormat, textureType) {
+  validateTextureSize(width, height);
+  const texture = createTexture(gl);
+  const tex2d = gl.TEXTURE_2D;
+  callAndCheck(gl, () => gl.bindTexture(tex2d, texture));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_S, gl.CLAMP_TO_EDGE));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_T, gl.CLAMP_TO_EDGE));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MIN_FILTER, gl.NEAREST));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MAG_FILTER, gl.NEAREST));
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    callAndCheck(gl, () => gl.texImage2D(tex2d, 0, internalFormat, width, height, 0, textureFormat, textureType, null));
+  } else {
+    callAndCheck(gl, () => gl.texStorage2D(tex2d, 1, internalFormat, width, height));
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+  return { texture, texShape: [height, width] };
+}
+function getInternalFormatForFloat32MatrixTexture(textureConfig) {
+  return textureConfig.internalFormatFloat;
+}
+function createFloat32MatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat32MatrixTexture(textureConfig), textureConfig.textureFormatFloat, gl.FLOAT);
+}
+function getInternalFormatForFloat16MatrixTexture(textureConfig) {
+  return textureConfig.internalFormatHalfFloat;
+}
+function createFloat16MatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16MatrixTexture(textureConfig), textureConfig.textureFormatFloat, textureConfig.textureTypeHalfFloat);
+}
+function getInternalFormatForUnsignedBytesMatrixTexture(textureConfig) {
+  return textureConfig.downloadTextureFormat;
+}
+function createUnsignedBytesMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForUnsignedBytesMatrixTexture(textureConfig), gl.RGBA, gl.UNSIGNED_BYTE);
+}
+function getInternalFormatForPackedMatrixTexture(textureConfig) {
+  return textureConfig.internalFormatPackedFloat;
+}
+function createPackedMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForPackedMatrixTexture(textureConfig), gl.RGBA, gl.FLOAT);
+}
+function getInternalFormatForFloat16PackedMatrixTexture(textureConfig) {
+  return textureConfig.internalFormatPackedHalfFloat;
+}
+function createFloat16PackedMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16PackedMatrixTexture(textureConfig), gl.RGBA, textureConfig.textureTypeHalfFloat);
+}
+function bindVertexProgramAttributeStreams(gl, program, vertexBuffer) {
+  const posOffset = 0;
+  const uvOffset = 3 * 4;
+  const stride = 3 * 4 + 2 * 4;
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, vertexBuffer));
+  const success = bindVertexBufferToProgramAttribute(gl, program, "clipSpacePos", vertexBuffer, 3, stride, posOffset);
+  return success && bindVertexBufferToProgramAttribute(gl, program, "uv", vertexBuffer, 2, stride, uvOffset);
+}
+function uploadDenseMatrixToTexture(gl, texture, width, height, data, textureConfig) {
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+  let dataForUpload, texelDataType, internalFormat;
+  if (data instanceof Uint8Array) {
+    dataForUpload = new Uint8Array(width * height * 4);
+    texelDataType = gl.UNSIGNED_BYTE;
+    internalFormat = gl.RGBA;
+  } else {
+    dataForUpload = new Float32Array(width * height * 4);
+    texelDataType = gl.FLOAT;
+    internalFormat = textureConfig.internalFormatPackedFloat;
+  }
+  dataForUpload.set(data);
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, width, height, gl.RGBA, texelDataType, dataForUpload));
+  } else {
+    callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, internalFormat, width, height, 0, gl.RGBA, texelDataType, dataForUpload));
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function uploadPixelDataToTexture(gl, texture, pixels) {
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+  if (pixels.data instanceof Uint8Array) {
+    if (env().getNumber("WEBGL_VERSION") === 2) {
+      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, pixels.width, pixels.height, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));
+    } else {
+      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, pixels.width, pixels.height, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));
+    }
+  } else {
+    if (env().getNumber("WEBGL_VERSION") === 2) {
+      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels));
+    } else {
+      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gl.RGBA, gl.UNSIGNED_BYTE, pixels));
+    }
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function createBufferFromOutputTexture(gl2, rows, columns, textureConfig) {
+  const buffer2 = gl2.createBuffer();
+  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2));
+  const bytesPerFloat = 4;
+  const valuesPerTexel = 4;
+  const bufferSizeBytes = bytesPerFloat * valuesPerTexel * rows * columns;
+  callAndCheck(gl2, () => gl2.bufferData(gl2.PIXEL_PACK_BUFFER, bufferSizeBytes, gl2.STREAM_READ));
+  callAndCheck(gl2, () => gl2.readPixels(0, 0, columns, rows, gl2.RGBA, gl2.FLOAT, 0));
+  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null));
+  return buffer2;
+}
+function downloadFloat32MatrixFromBuffer(gl, buffer2, size) {
+  const gl2 = gl;
+  const downloadTarget = new Float32Array(size);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);
+  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);
+  return downloadTarget;
+}
+function downloadByteEncodedFloatMatrixFromOutputTexture(gl, rows, columns, textureConfig) {
+  const [w, h] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  const numChannels = 4;
+  const downloadTarget = new Uint8Array(getUnpackedArraySizeFromMatrixSize(rows * columns, numChannels));
+  callAndCheck(gl, () => gl.readPixels(0, 0, w, h, textureConfig.downloadTextureFormat, gl.UNSIGNED_BYTE, downloadTarget));
+  return new Float32Array(downloadTarget.buffer);
+}
+function downloadPackedMatrixFromBuffer(gl, buffer2, batch, rows, cols, physicalRows, physicalCols, textureConfig) {
+  const gl2 = gl;
+  const downloadTarget = new Float32Array(getPackedRGBAArraySizeFromMatrixShape(physicalRows, physicalCols));
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);
+  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);
+  return downloadTarget;
+}
+function downloadMatrixFromPackedOutputTexture(gl, physicalRows, physicalCols) {
+  const packedRGBA = new Float32Array(physicalRows * physicalCols * 4);
+  callAndCheck(gl, () => gl.readPixels(0, 0, physicalCols, physicalRows, gl.RGBA, gl.FLOAT, packedRGBA));
+  return packedRGBA;
+}
+var GPGPUContext = class {
+  constructor(gl) {
+    this.outputTexture = null;
+    this.program = null;
+    this.disposed = false;
+    this.vertexAttrsAreBound = false;
+    this.itemsToPoll = [];
+    const glVersion = env().getNumber("WEBGL_VERSION");
+    if (gl != null) {
+      this.gl = gl;
+      setWebGLContext(glVersion, gl);
+    } else {
+      this.gl = getWebGLContext(glVersion);
+    }
+    let COLOR_BUFFER_FLOAT = "WEBGL_color_buffer_float";
+    const COLOR_BUFFER_HALF_FLOAT = "EXT_color_buffer_half_float";
+    this.parallelCompilationExtension = this.gl.getExtension("KHR_parallel_shader_compile");
+    if (env().getNumber("WEBGL_VERSION") === 1) {
+      const TEXTURE_FLOAT = "OES_texture_float";
+      const TEXTURE_HALF_FLOAT = "OES_texture_half_float";
+      this.textureFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_FLOAT);
+      if (hasExtension(this.gl, TEXTURE_HALF_FLOAT)) {
+        this.textureHalfFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_HALF_FLOAT);
+      } else if (env().get("WEBGL_FORCE_F16_TEXTURES")) {
+        throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");
+      }
+      this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);
+      if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {
+        this.colorBufferHalfFloatExtension = getExtensionOrThrow(this.gl, COLOR_BUFFER_HALF_FLOAT);
+      } else if (env().get("WEBGL_FORCE_F16_TEXTURES")) {
+        throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");
+      }
+    } else {
+      COLOR_BUFFER_FLOAT = "EXT_color_buffer_float";
+      if (hasExtension(this.gl, COLOR_BUFFER_FLOAT)) {
+        this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);
+      } else if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {
+        this.colorBufferHalfFloatExtension = this.gl.getExtension(COLOR_BUFFER_HALF_FLOAT);
+      } else {
+        throw new Error("GL context does not support color renderable floats");
+      }
+    }
+    this.vertexBuffer = createVertexBuffer(this.gl);
+    this.indexBuffer = createIndexBuffer(this.gl);
+    this.framebuffer = createFramebuffer(this.gl);
+    this.textureConfig = getTextureConfig(this.gl, this.textureHalfFloatExtension);
+  }
+  get debug() {
+    return env().getBool("DEBUG");
+  }
+  dispose() {
+    if (this.disposed) {
+      return;
+    }
+    if (this.program != null) {
+      console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing.");
+    }
+    if (this.outputTexture != null) {
+      console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");
+    }
+    const gl = this.gl;
+    callAndCheck(gl, () => gl.finish());
+    callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));
+    callAndCheck(gl, () => gl.deleteFramebuffer(this.framebuffer));
+    callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, null));
+    callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, null));
+    callAndCheck(gl, () => gl.deleteBuffer(this.indexBuffer));
+    this.disposed = true;
+  }
+  createFloat32MatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat32MatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createFloat16MatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat16MatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createUnsignedBytesMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createUnsignedBytesMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  uploadPixelDataToTexture(texture, pixels) {
+    this.throwIfDisposed();
+    uploadPixelDataToTexture(this.gl, texture, pixels);
+  }
+  uploadDenseMatrixToTexture(texture, width, height, data) {
+    this.throwIfDisposed();
+    uploadDenseMatrixToTexture(this.gl, texture, width, height, data, this.textureConfig);
+  }
+  createFloat16PackedMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat16PackedMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createPackedMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createPackedMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  deleteMatrixTexture(texture) {
+    this.throwIfDisposed();
+    if (this.outputTexture === texture) {
+      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);
+      this.outputTexture = null;
+    }
+    callAndCheck(this.gl, () => this.gl.deleteTexture(texture));
+  }
+  downloadByteEncodedFloatMatrixFromOutputTexture(texture, rows, columns) {
+    return this.downloadMatrixDriver(texture, () => downloadByteEncodedFloatMatrixFromOutputTexture(this.gl, rows, columns, this.textureConfig));
+  }
+  downloadPackedMatrixFromBuffer(buffer2, batch, rows, columns, physicalRows, physicalCols) {
+    return downloadPackedMatrixFromBuffer(this.gl, buffer2, batch, rows, columns, physicalRows, physicalCols, this.textureConfig);
+  }
+  downloadFloat32MatrixFromBuffer(buffer2, size) {
+    return downloadFloat32MatrixFromBuffer(this.gl, buffer2, size);
+  }
+  createBufferFromTexture(texture, rows, columns) {
+    this.bindTextureToFrameBuffer(texture);
+    const result = createBufferFromOutputTexture(this.gl, rows, columns, this.textureConfig);
+    this.unbindTextureToFrameBuffer();
+    return result;
+  }
+  createAndWaitForFence() {
+    const fenceContext = this.createFence(this.gl);
+    return this.pollFence(fenceContext);
+  }
+  createFence(gl) {
+    let query;
+    let isFencePassed;
+    if (env().getBool("WEBGL_FENCE_API_ENABLED")) {
+      const gl2 = gl;
+      const sync = gl2.fenceSync(gl2.SYNC_GPU_COMMANDS_COMPLETE, 0);
+      gl.flush();
+      isFencePassed = () => {
+        const status = gl2.clientWaitSync(sync, 0, 0);
+        return status === gl2.ALREADY_SIGNALED || status === gl2.CONDITION_SATISFIED;
+      };
+      query = sync;
+    } else if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0) {
+      query = this.beginQuery();
+      this.endQuery();
+      isFencePassed = () => this.isQueryAvailable(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"));
+    } else {
+      isFencePassed = () => true;
+    }
+    return { query, isFencePassed };
+  }
+  downloadMatrixFromPackedTexture(texture, physicalRows, physicalCols) {
+    return this.downloadMatrixDriver(texture, () => downloadMatrixFromPackedOutputTexture(this.gl, physicalRows, physicalCols));
+  }
+  createProgram(fragmentShader) {
+    this.throwIfDisposed();
+    const gl = this.gl;
+    if (this.vertexShader == null) {
+      this.vertexShader = createVertexShader2(gl);
+    }
+    const program = createProgram(gl);
+    callAndCheck(gl, () => gl.attachShader(program, this.vertexShader));
+    callAndCheck(gl, () => gl.attachShader(program, fragmentShader));
+    linkProgram(gl, program);
+    if (this.debug) {
+      validateProgram(gl, program);
+    }
+    if (!this.vertexAttrsAreBound) {
+      this.setProgram(program);
+      this.vertexAttrsAreBound = bindVertexProgramAttributeStreams(gl, this.program, this.vertexBuffer);
+    }
+    return program;
+  }
+  deleteProgram(program) {
+    this.throwIfDisposed();
+    if (program === this.program) {
+      this.program = null;
+    }
+    if (program != null) {
+      callAndCheck(this.gl, () => this.gl.deleteProgram(program));
+    }
+  }
+  setProgram(program) {
+    this.throwIfDisposed();
+    this.program = program;
+    if (this.program != null && this.debug) {
+      validateProgram(this.gl, this.program);
+    }
+    callAndCheck(this.gl, () => this.gl.useProgram(program));
+  }
+  getUniformLocation(program, uniformName, shouldThrow = true) {
+    this.throwIfDisposed();
+    if (shouldThrow) {
+      return getProgramUniformLocationOrThrow(this.gl, program, uniformName);
+    } else {
+      return getProgramUniformLocation(this.gl, program, uniformName);
+    }
+  }
+  getAttributeLocation(program, attribute) {
+    this.throwIfDisposed();
+    return callAndCheck(this.gl, () => this.gl.getAttribLocation(program, attribute));
+  }
+  getUniformLocationNoThrow(program, uniformName) {
+    this.throwIfDisposed();
+    return this.gl.getUniformLocation(program, uniformName);
+  }
+  setInputMatrixTexture(inputMatrixTexture, uniformLocation, textureUnit) {
+    this.throwIfDisposed();
+    this.throwIfNoProgram();
+    bindTextureToProgramUniformSampler(this.gl, inputMatrixTexture, uniformLocation, textureUnit);
+  }
+  setOutputMatrixTexture(outputMatrixTexture, rows, columns) {
+    this.setOutputMatrixTextureDriver(outputMatrixTexture, columns, rows);
+  }
+  setOutputPackedMatrixTexture(outputPackedMatrixTexture, rows, columns) {
+    this.throwIfDisposed();
+    const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+    this.setOutputMatrixTextureDriver(outputPackedMatrixTexture, width, height);
+  }
+  setOutputMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {
+    this.setOutputMatrixWriteRegionDriver(startColumn, startRow, numColumns, numRows);
+  }
+  setOutputPackedMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {
+    throw new Error("setOutputPackedMatrixWriteRegion not implemented.");
+  }
+  debugValidate() {
+    if (this.program != null) {
+      validateProgram(this.gl, this.program);
+    }
+    validateFramebuffer(this.gl);
+  }
+  executeProgram() {
+    this.throwIfDisposed();
+    this.throwIfNoProgram();
+    const gl = this.gl;
+    if (this.debug) {
+      this.debugValidate();
+    }
+    callAndCheck(gl, () => gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 0));
+  }
+  blockUntilAllProgramsCompleted() {
+    this.throwIfDisposed();
+    callAndCheck(this.gl, () => this.gl.finish());
+  }
+  getQueryTimerExtension() {
+    if (this.disjointQueryTimerExtension == null) {
+      this.disjointQueryTimerExtension = getExtensionOrThrow(this.gl, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2 ? "EXT_disjoint_timer_query_webgl2" : "EXT_disjoint_timer_query");
+    }
+    return this.disjointQueryTimerExtension;
+  }
+  getQueryTimerExtensionWebGL2() {
+    return this.getQueryTimerExtension();
+  }
+  getQueryTimerExtensionWebGL1() {
+    return this.getQueryTimerExtension();
+  }
+  beginQuery() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2) {
+      const gl2 = this.gl;
+      const ext2 = this.getQueryTimerExtensionWebGL2();
+      const query2 = gl2.createQuery();
+      gl2.beginQuery(ext2.TIME_ELAPSED_EXT, query2);
+      return query2;
+    }
+    const ext = this.getQueryTimerExtensionWebGL1();
+    const query = ext.createQueryEXT();
+    ext.beginQueryEXT(ext.TIME_ELAPSED_EXT, query);
+    return query;
+  }
+  endQuery() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2) {
+      const gl2 = this.gl;
+      const ext2 = this.getQueryTimerExtensionWebGL2();
+      gl2.endQuery(ext2.TIME_ELAPSED_EXT);
+      return;
+    }
+    const ext = this.getQueryTimerExtensionWebGL1();
+    ext.endQueryEXT(ext.TIME_ELAPSED_EXT);
+  }
+  async waitForQueryAndGetTime(query) {
+    await util_exports.repeatedTry(() => this.disposed || this.isQueryAvailable(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")));
+    return this.getQueryTime(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"));
+  }
+  getQueryTime(query, queryTimerVersion) {
+    if (queryTimerVersion === 0) {
+      return null;
+    }
+    if (queryTimerVersion === 2) {
+      const gl2 = this.gl;
+      const timeElapsedNanos = gl2.getQueryParameter(query, gl2.QUERY_RESULT);
+      return timeElapsedNanos / 1e6;
+    } else {
+      const ext = this.getQueryTimerExtensionWebGL1();
+      const timeElapsedNanos = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_EXT);
+      return timeElapsedNanos / 1e6;
+    }
+  }
+  isQueryAvailable(query, queryTimerVersion) {
+    if (queryTimerVersion === 0) {
+      return true;
+    }
+    if (queryTimerVersion === 2) {
+      const gl2 = this.gl;
+      const ext = this.getQueryTimerExtensionWebGL2();
+      const available = gl2.getQueryParameter(query, gl2.QUERY_RESULT_AVAILABLE);
+      if (this.disjoint == null) {
+        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);
+      }
+      return available && !this.disjoint;
+    } else {
+      const ext = this.getQueryTimerExtensionWebGL1();
+      const available = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_AVAILABLE_EXT);
+      if (this.disjoint == null) {
+        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);
+      }
+      return available && !this.disjoint;
+    }
+  }
+  pollFence(fenceContext) {
+    return new Promise((resolve) => {
+      this.addItemToPoll(() => fenceContext.isFencePassed(), () => resolve());
+    });
+  }
+  pollItems() {
+    const index = linearSearchLastTrue(this.itemsToPoll.map((x) => x.isDoneFn));
+    for (let i = 0; i <= index; ++i) {
+      const { resolveFn } = this.itemsToPoll[i];
+      resolveFn();
+    }
+    this.itemsToPoll = this.itemsToPoll.slice(index + 1);
+  }
+  addItemToPoll(isDoneFn, resolveFn) {
+    this.itemsToPoll.push({ isDoneFn, resolveFn });
+    if (this.itemsToPoll.length > 1) {
+      return;
+    }
+    let scheduleFn = void 0;
+    if ("setTimeoutCustom" in env().platform) {
+      scheduleFn = env().platform.setTimeoutCustom.bind(env().platform);
+    }
+    util_exports.repeatedTry(() => {
+      this.pollItems();
+      return this.itemsToPoll.length === 0;
+    }, () => 0, null, scheduleFn);
+  }
+  bindTextureToFrameBuffer(texture) {
+    this.throwIfDisposed();
+    bindColorTextureToFramebuffer(this.gl, texture, this.framebuffer);
+    if (this.debug) {
+      validateFramebuffer(this.gl);
+    }
+  }
+  unbindTextureToFrameBuffer() {
+    if (this.outputTexture != null) {
+      bindColorTextureToFramebuffer(this.gl, this.outputTexture, this.framebuffer);
+      if (this.debug) {
+        validateFramebuffer(this.gl);
+      }
+    } else {
+      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);
+    }
+  }
+  downloadMatrixDriver(texture, downloadAndDecode) {
+    this.bindTextureToFrameBuffer(texture);
+    const result = downloadAndDecode();
+    this.unbindTextureToFrameBuffer();
+    return result;
+  }
+  setOutputMatrixTextureDriver(outputMatrixTextureMaybePacked, width, height) {
+    this.throwIfDisposed();
+    const gl = this.gl;
+    bindColorTextureToFramebuffer(gl, outputMatrixTextureMaybePacked, this.framebuffer);
+    if (this.debug) {
+      validateFramebuffer(gl);
+    }
+    this.outputTexture = outputMatrixTextureMaybePacked;
+    callAndCheck(gl, () => gl.viewport(0, 0, width, height));
+    callAndCheck(gl, () => gl.scissor(0, 0, width, height));
+  }
+  setOutputMatrixWriteRegionDriver(x, y, width, height) {
+    this.throwIfDisposed();
+    callAndCheck(this.gl, () => this.gl.scissor(x, y, width, height));
+  }
+  throwIfDisposed() {
+    if (this.disposed) {
+      throw new Error("Attempted to use disposed GPGPUContext.");
+    }
+  }
+  throwIfNoProgram() {
+    if (this.program == null) {
+      throw new Error("No GPU program is currently set.");
+    }
+  }
+};
+function linearSearchLastTrue(arr) {
+  let i = 0;
+  for (; i < arr.length; ++i) {
+    const isDone = arr[i]();
+    if (!isDone) {
+      break;
+    }
+  }
+  return i - 1;
+}
+var { addImpl: addImplCPU, bincountImpl: bincountImplCPU, bincountReduceImpl: bincountReduceImplCPU, castImpl: castImplCPU, ceilImpl: ceilImplCPU, concatImpl: concatImplCPU, equalImpl: equalImplCPU, expImpl: expImplCPU, expm1Impl: expm1ImplCPU, floorImpl: floorImplCPU, gatherNdImpl: gatherNdImplCPU, gatherV2Impl: gatherV2ImplCPU, greaterImpl: greaterImplCPU, greaterEqualImpl: greaterEqualImplCPU, lessImpl: lessImplCPU, lessEqualImpl: lessEqualImplCPU, linSpaceImpl: linSpaceImplCPU, logImpl: logImplCPU, maxImpl: maxImplCPU, maximumImpl: maximumImplCPU, minimumImpl: minimumImplCPU, multiplyImpl: multiplyImplCPU, negImpl: negImplCPU, notEqualImpl: notEqualImplCPU, prodImpl: prodImplCPU, raggedGatherImpl: raggedGatherImplCPU, raggedRangeImpl: raggedRangeImplCPU, raggedTensorToTensorImpl: raggedTensorToTensorImplCPU, rangeImpl: rangeImplCPU, rsqrtImpl: rsqrtImplCPU, scatterImpl: scatterImplCPU, sigmoidImpl: sigmoidImplCPU, simpleAbsImpl: simpleAbsImplCPU, sliceImpl: sliceImplCPU, sparseFillEmptyRowsImpl: sparseFillEmptyRowsImplCPU, sparseReshapeImpl: sparseReshapeImplCPU, sparseSegmentReductionImpl: sparseSegmentReductionImplCPU, sqrtImpl: sqrtImplCPU, stridedSliceImpl: stridedSliceImplCPU, stringNGramsImpl: stringNGramsImplCPU, stringSplitImpl: stringSplitImplCPU, stringToHashBucketFastImpl: stringToHashBucketFastImplCPU, subImpl: subImplCPU, tileImpl: tileImplCPU, topKImpl: topKImplCPU, transposeImpl: transposeImplCPU, uniqueImpl: uniqueImplCPU } = shared_exports;
+function getVecChannels(name, rank) {
+  return ["x", "y", "z", "w", "u", "v"].slice(0, rank).map((d) => `${name}.${d}`);
+}
+function getChannels(name, rank) {
+  if (rank === 1) {
+    return [name];
+  }
+  return getVecChannels(name, rank);
+}
+function getSourceCoords(rank, dims) {
+  if (rank === 1) {
+    return "rc";
+  }
+  let coords2 = "";
+  for (let i = 0; i < rank; i++) {
+    coords2 += dims[i];
+    if (i < rank - 1) {
+      coords2 += ",";
+    }
+  }
+  return coords2;
+}
+var PackProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    if (this.rank === 0) {
+      this.userCode = `
         void main() {
           setOutput(vec4(getA(), 0., 0., 0.));
         }
-      `;else{let t=wn("rc",this.rank),n=gt(this.rank),a=this.getOutOfBoundsCondition(t),r=this.getSetup(t),s=this.getOutput(t);this.userCode=`
+      `;
+    } else {
+      const channels = getChannels("rc", this.rank);
+      const dtype = getCoordsDataType(this.rank);
+      const outOfBoundsCondition = this.getOutOfBoundsCondition(channels);
+      const setup51 = this.getSetup(channels);
+      const output = this.getOutput(channels);
+      this.userCode = `
         void main() {
-          ${n} rc = getOutputCoords();
+          ${dtype} rc = getOutputCoords();
 
-          if(${a}) {
+          if(${outOfBoundsCondition}) {
             setOutput(vec4(0));
           } else {
-            ${r}
+            ${setup51}
 
-            setOutput(vec4(${s}));
+            setOutput(vec4(${output}));
           }
         }
-      `}}getSourceCoordsArr(e){let t=[];for(let n=0;n<=1;n++)for(let a=0;a<=1;a++){let r=`${n===0?"r":"rp1"}, ${a===0?"c":"cp1"}`;for(let s=2;s<this.rank;s++)r=`${e[e.length-1-s]},`+r;t.push(r)}return t}getOutOfBoundsCondition(e){if(this.rank===1)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let t="";for(let n=this.rank-2;n<this.rank;n++)t+=`${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(t+="||");return t}getSetup(e){if(this.rank===1)return"";let t=e.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],a=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`
-      int r = ${t[0]};
-      int c = ${t[1]};
+      `;
+    }
+  }
+  getSourceCoordsArr(dims) {
+    const coords2 = [];
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        let coord = `${row === 0 ? "r" : "rp1"}, ${col === 0 ? "c" : "cp1"}`;
+        for (let d = 2; d < this.rank; d++) {
+          coord = `${dims[dims.length - 1 - d]},` + coord;
+        }
+        coords2.push(coord);
+      }
+    }
+    return coords2;
+  }
+  getOutOfBoundsCondition(dims) {
+    if (this.rank === 1) {
+      return `rc > ${this.enableShapeUniforms ? "outShape" : this.outputShape[0]}`;
+    }
+    let cond = "";
+    for (let i = this.rank - 2; i < this.rank; i++) {
+      cond += `${dims[i]} >= ${this.enableShapeUniforms ? `outShape[${i}]` : this.outputShape[i]}`;
+      if (i < this.rank - 1) {
+        cond += "||";
+      }
+    }
+    return cond;
+  }
+  getSetup(dims) {
+    if (this.rank === 1) {
+      return "";
+    }
+    const innerDims = dims.slice(-2);
+    const col = this.enableShapeUniforms ? `outShape[${this.rank} - 1]` : this.outputShape[this.rank - 1];
+    const row = this.enableShapeUniforms ? `outShape[${this.rank} - 2]` : this.outputShape[this.rank - 2];
+    return `
+      int r = ${innerDims[0]};
+      int c = ${innerDims[1]};
       int rp1 = r + 1;
       int cp1 = c + 1;
 
-      bool cEdge = cp1 >= ${n};
-      bool rEdge = rp1 >= ${a};
-    `}getOutput(e){let t=this.getSourceCoordsArr(e);return this.rank===1?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${t[0]}),
-            cEdge ? 0. : getA(${t[1]}),
-            rEdge ? 0. : getA(${t[2]}),
-            rEdge || cEdge ? 0. : getA(${t[3]})`}},B_=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length);let n="";for(let a=0;a<4;a++){let r="thisRC = rc;";a%2===1&&(r+="thisRC.z += 1;"),a>1&&(r+="thisRC.y += 1;"),n+=`
-        ${r}
-        ${a>0?"if(thisRC.y < rows && thisRC.z < cols){":""}
+      bool cEdge = cp1 >= ${col};
+      bool rEdge = rp1 >= ${row};
+    `;
+  }
+  getOutput(dims) {
+    const sourceCoords = this.getSourceCoordsArr(dims);
+    if (this.rank === 1) {
+      const outShape = this.enableShapeUniforms ? "outShape" : this.outputShape[0];
+      return `getA(rc), (rc + 1 >= ${outShape} ? 0. : getA(rc + 1)), 0, 0`;
+    }
+    return `getA(${sourceCoords[0]}),
+            cEdge ? 0. : getA(${sourceCoords[1]}),
+            rEdge ? 0. : getA(${sourceCoords[2]}),
+            rEdge || cEdge ? 0. : getA(${sourceCoords[3]})`;
+  }
+};
+var ReshapePackedProgram = class {
+  constructor(outputShape, inputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "inputShape", type: "ivec3" }];
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let mainLoop = ``;
+    for (let i = 0; i < 4; i++) {
+      let thisRC = `thisRC = rc;`;
+      if (i % 2 === 1) {
+        thisRC += `thisRC.z += 1;`;
+      }
+      if (i > 1) {
+        thisRC += `thisRC.y += 1;`;
+      }
+      mainLoop += `
+        ${thisRC}
+        ${i > 0 ? `if(thisRC.y < rows && thisRC.z < cols){` : ""}
           int flatIndex = getFlatIndex(thisRC);
 
           ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);
           vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));
 
-          result[${a}] =
+          result[${i}] =
             getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);
-        ${a>0?"}":""}
-      `}this.userCode=`
-      ${R7(t,this.enableShapeUniforms)}
-      ${this.enableShapeUniforms?G0():U0(e)}
+        ${i > 0 ? "}" : ""}
+      `;
+    }
+    this.userCode = `
+      ${getReshapedInputCoords(inputShape, this.enableShapeUniforms)}
+      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
       void main() {
         ivec3 rc = getOutputCoords();
@@ -1163,21 +51108,224 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         vec4 result = vec4(0.);
 
         ivec3 thisRC;
-        int rows = ${this.enableShapeUniforms?"outShape[1]":e[1]};
-        int cols = ${this.enableShapeUniforms?"outShape[2]":e[2]};
+        int rows = ${this.enableShapeUniforms ? "outShape[1]" : outputShape[1]};
+        int cols = ${this.enableShapeUniforms ? "outShape[2]" : outputShape[2]};
 
-        ${n}
+        ${mainLoop}
 
         setOutput(result);
       }
-    `}};function R7(e,t){return`
+    `;
+  }
+};
+function getReshapedInputCoords(shape, enableShapeUniforms) {
+  const coordsFromIndexSnippet = enableShapeUniforms ? getLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], "inputShape") : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], shape);
+  return `
     ivec3 inputCoordsFromReshapedOutCoords(int index) {
-      ${t?YY(["r","c","d"],"inputShape"):vo(["r","c","d"],e)}
+      ${coordsFromIndexSnippet}
       return ivec3(r, c, d);
     }
-  `}var M7=class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(e,t,n){let a=Kk(t,n),r=Xk(e,a,n);r in this.freeTextures||(this.freeTextures[r]=[]),r in this.usedTextures||(this.usedTextures[r]=[]);let s=qk(e,a,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[r].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=s,this.log();let o=this.freeTextures[r].shift();return this.usedTextures[r].push(o),o}let i;return a===on.PACKED_2X2_FLOAT32?i=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):a===on.PACKED_2X2_FLOAT16?i=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):a===on.UNPACKED_FLOAT32?i=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):a===on.UNPACKED_FLOAT16?i=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):a===on.PACKED_4X1_UNSIGNED_BYTE&&(i=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[r].push(i),this.numUsedTextures++,this._numBytesAllocated+=s,this.log(),i}releaseTexture(e,t,n,a){if(this.freeTextures==null)return;let r=Kk(n,a),s=Xk(t,r,a);s in this.freeTextures||(this.freeTextures[s]=[]);let i=qk(t,r,this.gpgpu.gl,this.gpgpu.textureConfig,a),o=H().get("WEBGL_DELETE_TEXTURE_THRESHOLD");o!==-1&&this._numBytesAllocated>o?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=i):(this.freeTextures[s].push(e),this.numFreeTextures++,this._numBytesFree+=i),this.numUsedTextures--;let l=this.usedTextures[s],u=l.indexOf(e);if(u<0)throw new Error("Cannot release a texture that was never provided by this texture manager");l.splice(u,1),this.log()}log(){if(!this.logEnabled)return;let e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);let t=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let e in this.freeTextures)this.freeTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});for(let e in this.usedTextures)this.usedTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function P7(e,t){let n=e;if(t===n.R32F)return 4;if(t===n.R16F)return 2;if(t===n.RGBA32F||t===e.RGBA)return 16;if(t===n.RGBA16F)return 8;if(t===n.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function qk(e,t,n,a,r){let s=O7(t,a),i;if(r){let[l,u]=Ou(e[0],e[1]);i=l*u}else{let[l,u]=ed(e[0],e[1]);i=l*u}let o=P7(n,s);return i*o}function O7(e,t){switch(e){case on.PACKED_2X2_FLOAT32:return X0(t);case on.PACKED_2X2_FLOAT16:return Y0(t);case on.UNPACKED_FLOAT32:return j0(t);case on.UNPACKED_FLOAT16:return q0(t);case on.PACKED_4X1_UNSIGNED_BYTE:return K0(t);default:throw new Error(`Unknown physical texture type ${e}`)}}function L7(e){return H().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?on.PACKED_2X2_FLOAT32:on.UNPACKED_FLOAT32:e?on.PACKED_2X2_FLOAT16:on.UNPACKED_FLOAT16}function Kk(e,t){if(e===pa.UPLOAD)return on.PACKED_2X2_FLOAT32;if(e===pa.RENDER||e==null)return L7(t);if(e===pa.DOWNLOAD||e===pa.PIXELS)return on.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${e}`)}function Xk(e,t,n){return`${e[0]}_${e[1]}_${t}_${n}`}var Sr=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length),this.userCode=`
+  `;
+}
+var TextureManager = class {
+  constructor(gpgpu) {
+    this.gpgpu = gpgpu;
+    this.numUsedTextures = 0;
+    this.numFreeTextures = 0;
+    this._numBytesAllocated = 0;
+    this._numBytesFree = 0;
+    this.freeTextures = {};
+    this.logEnabled = false;
+    this.usedTextures = {};
+  }
+  acquireTexture(shapeRC, usage, isPacked) {
+    const physicalTexType = getPhysicalFromLogicalTextureType(usage, isPacked);
+    const shapeKey = getKeyFromTextureShape(shapeRC, physicalTexType, isPacked);
+    if (!(shapeKey in this.freeTextures)) {
+      this.freeTextures[shapeKey] = [];
+    }
+    if (!(shapeKey in this.usedTextures)) {
+      this.usedTextures[shapeKey] = [];
+    }
+    const texBytes = computeBytes(shapeRC, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);
+    if (this.freeTextures[shapeKey].length > 0) {
+      this.numFreeTextures--;
+      this.numUsedTextures++;
+      this._numBytesFree -= texBytes;
+      this.log();
+      const newTexture2 = this.freeTextures[shapeKey].shift();
+      this.usedTextures[shapeKey].push(newTexture2);
+      return newTexture2;
+    }
+    let newTexture;
+    if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT32) {
+      newTexture = this.gpgpu.createPackedMatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT16) {
+      newTexture = this.gpgpu.createFloat16PackedMatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT32) {
+      newTexture = this.gpgpu.createFloat32MatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT16) {
+      newTexture = this.gpgpu.createFloat16MatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE) {
+      newTexture = this.gpgpu.createUnsignedBytesMatrixTexture(shapeRC[0], shapeRC[1]);
+    }
+    this.usedTextures[shapeKey].push(newTexture);
+    this.numUsedTextures++;
+    this._numBytesAllocated += texBytes;
+    this.log();
+    return newTexture;
+  }
+  releaseTexture(texture, shape, logicalTexType, isPacked) {
+    if (this.freeTextures == null) {
+      return;
+    }
+    const physicalTexType = getPhysicalFromLogicalTextureType(logicalTexType, isPacked);
+    const shapeKey = getKeyFromTextureShape(shape, physicalTexType, isPacked);
+    if (!(shapeKey in this.freeTextures)) {
+      this.freeTextures[shapeKey] = [];
+    }
+    const texBytes = computeBytes(shape, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);
+    const deleteTexThreshold = env().get("WEBGL_DELETE_TEXTURE_THRESHOLD");
+    if (deleteTexThreshold !== -1 && this._numBytesAllocated > deleteTexThreshold) {
+      this.gpgpu.deleteMatrixTexture(texture.texture);
+      this._numBytesAllocated -= texBytes;
+    } else {
+      this.freeTextures[shapeKey].push(texture);
+      this.numFreeTextures++;
+      this._numBytesFree += texBytes;
+    }
+    this.numUsedTextures--;
+    const texList = this.usedTextures[shapeKey];
+    const texIndex = texList.indexOf(texture);
+    if (texIndex < 0) {
+      throw new Error("Cannot release a texture that was never provided by this texture manager");
+    }
+    texList.splice(texIndex, 1);
+    this.log();
+  }
+  log() {
+    if (!this.logEnabled) {
+      return;
+    }
+    const total = this.numFreeTextures + this.numUsedTextures;
+    console.log("Free/Used", `${this.numFreeTextures} / ${this.numUsedTextures}`, `(${total})`);
+    const freeRatio = this._numBytesFree / this._numBytesAllocated;
+    console.log(`Bytes allocated: ${this._numBytesAllocated}`);
+    console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100 * freeRatio)}%)`);
+  }
+  get numBytesAllocated() {
+    return this._numBytesAllocated;
+  }
+  get numBytesFree() {
+    return this._numBytesFree;
+  }
+  getNumUsedTextures() {
+    return this.numUsedTextures;
+  }
+  getNumFreeTextures() {
+    return this.numFreeTextures;
+  }
+  dispose() {
+    if (this.freeTextures == null) {
+      return;
+    }
+    for (const texShape in this.freeTextures) {
+      this.freeTextures[texShape].forEach((tex) => {
+        this.gpgpu.deleteMatrixTexture(tex.texture);
+      });
+    }
+    for (const texShape in this.usedTextures) {
+      this.usedTextures[texShape].forEach((tex) => {
+        this.gpgpu.deleteMatrixTexture(tex.texture);
+      });
+    }
+    this.freeTextures = null;
+    this.usedTextures = null;
+    this.numUsedTextures = 0;
+    this.numFreeTextures = 0;
+    this._numBytesAllocated = 0;
+    this._numBytesFree = 0;
+  }
+};
+function numBytesForInternalFormat(gl, internalFormat) {
+  const glany = gl;
+  if (internalFormat === glany.R32F) {
+    return 4;
+  } else if (internalFormat === glany.R16F) {
+    return 2;
+  } else if (internalFormat === glany.RGBA32F) {
+    return 16;
+  } else if (internalFormat === gl.RGBA) {
+    return 16;
+  } else if (internalFormat === glany.RGBA16F) {
+    return 8;
+  } else if (internalFormat === glany.RGBA8) {
+    return 4;
+  }
+  throw new Error(`Unknown internal format ${internalFormat}`);
+}
+function computeBytes(shape, physicalTexType, gl, textureConfig, isPacked) {
+  const internalFormat = internalFormatForPhysicalTexType(physicalTexType, textureConfig);
+  let numElements;
+  if (isPacked) {
+    const [packedWidth, packedHeight] = getPackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);
+    numElements = packedWidth * packedHeight;
+  } else {
+    const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);
+    numElements = width * height;
+  }
+  const bytesPerElement2 = numBytesForInternalFormat(gl, internalFormat);
+  return numElements * bytesPerElement2;
+}
+function internalFormatForPhysicalTexType(physicalTexType, textureConfig) {
+  switch (physicalTexType) {
+    case PhysicalTextureType.PACKED_2X2_FLOAT32:
+      return getInternalFormatForPackedMatrixTexture(textureConfig);
+    case PhysicalTextureType.PACKED_2X2_FLOAT16:
+      return getInternalFormatForFloat16PackedMatrixTexture(textureConfig);
+    case PhysicalTextureType.UNPACKED_FLOAT32:
+      return getInternalFormatForFloat32MatrixTexture(textureConfig);
+    case PhysicalTextureType.UNPACKED_FLOAT16:
+      return getInternalFormatForFloat16MatrixTexture(textureConfig);
+    case PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE:
+      return getInternalFormatForUnsignedBytesMatrixTexture(textureConfig);
+    default:
+      throw new Error(`Unknown physical texture type ${physicalTexType}`);
+  }
+}
+function getPhysicalTextureForRendering(isPacked) {
+  if (env().getBool("WEBGL_RENDER_FLOAT32_ENABLED")) {
+    if (isPacked) {
+      return PhysicalTextureType.PACKED_2X2_FLOAT32;
+    }
+    return PhysicalTextureType.UNPACKED_FLOAT32;
+  }
+  if (isPacked) {
+    return PhysicalTextureType.PACKED_2X2_FLOAT16;
+  }
+  return PhysicalTextureType.UNPACKED_FLOAT16;
+}
+function getPhysicalFromLogicalTextureType(logicalTexType, isPacked) {
+  if (logicalTexType === TextureUsage.UPLOAD) {
+    return PhysicalTextureType.PACKED_2X2_FLOAT32;
+  } else if (logicalTexType === TextureUsage.RENDER || logicalTexType == null) {
+    return getPhysicalTextureForRendering(isPacked);
+  } else if (logicalTexType === TextureUsage.DOWNLOAD || logicalTexType === TextureUsage.PIXELS) {
+    return PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE;
+  }
+  throw new Error(`Unknown logical texture type ${logicalTexType}`);
+}
+function getKeyFromTextureShape(shapeRowsCol, physicalTexType, isPacked) {
+  return `${shapeRowsCol[0]}_${shapeRowsCol[1]}_${physicalTexType}_${isPacked}`;
+}
+var UnaryOpProgram = class {
+  constructor(aShape, opSnippet) {
+    this.variableNames = ["A"];
+    this.outputShape = aShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       float unaryOperation(float x) {
-        ${t}
+        ${opSnippet}
       }
 
       void main() {
@@ -1186,11 +51334,23 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}},Da="if (isnan(x)) return x;",z7="return x;",Yk="return abs(x);",W7="return (x >= 0.0) ? x : (exp(x) - 1.0);",B7=Da+`
+    `;
+  }
+};
+var CHECK_NAN_SNIPPET = `if (isnan(x)) return x;`;
+var LINEAR = `return x;`;
+var ABS = `return abs(x);`;
+var ELU2 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;
+var RELU = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : x;
-`,V7=Da+`
+`;
+var RELU6 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,jo="return x;",U7="return 1.0 / (1.0 + exp(-1.0 * x));",G7="return x;",H7=`
+`;
+var CLONE = "return x;";
+var SIGMOID = `return 1.0 / (1.0 + exp(-1.0 * x));`;
+var LINEAR2 = `return x;`;
+var ELU3 = `
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -1199,7 +51359,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,j7=`
+`;
+var RELU2 = `
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1209,7 +51370,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,q7=`
+`;
+var RELU62 = `
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1219,9 +51381,18 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,K7="return 1.0 / (1.0 + exp(-1.0 * x));",js=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length),this.userCode=`
+`;
+var SIGMOID2 = `return 1.0 / (1.0 + exp(-1.0 * x));`;
+var UnaryOpPackedProgram = class {
+  constructor(aShape, opSnippet) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = aShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       vec4 unaryOperation(vec4 x) {
-        ${t}
+        ${opSnippet}
       }
 
       void main() {
@@ -1230,19 +51401,884 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}},X7=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length);let t=e.length,n=wn("rc",t),a=gt(t),r=F7(t,n),s=n.slice(-2),i=t<=1?"rc":`vec2(${s.join(",")})`;this.userCode=`
+    `;
+  }
+};
+var UnpackProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = false;
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const rank = outputShape.length;
+    const channels = getChannels("rc", rank);
+    const dtype = getCoordsDataType(rank);
+    const sourceCoords = getSourceCoords(rank, channels);
+    const innerDims = channels.slice(-2);
+    const coords2 = rank <= 1 ? "rc" : `vec2(${innerDims.join(",")})`;
+    this.userCode = `
       void main() {
-        ${a} rc = getOutputCoords();
-        vec4 packedInput = getA(${r});
+        ${dtype} rc = getOutputCoords();
+        vec4 packedInput = getA(${sourceCoords});
 
-        setOutput(getChannel(packedInput, ${i}));
+        setOutput(getChannel(packedInput, ${coords2}));
       }
-    `}},Y7=cr.whereImpl,Z7=1e-7,J7=1e-4,gb={};function Q7(e){return e in gb||(gb[e]={}),gb[e]}var eJ=H().getNumber("CPU_HANDOFF_SIZE_THRESHOLD"),tJ=600;function nJ(){return H().global.screen==null?1024:H().global.screen.height*H().global.screen.width*window.devicePixelRatio*tJ/1024/1024}var Rf=class extends rc{constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!H().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(e!=null){if(e instanceof ph)t=e;else{let n=qa(H().getNumber("WEBGL_VERSION"),e);t=new ph(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let n=qa(H().getNumber("WEBGL_VERSION"));t=new ph(n),this.binaryCache=Q7(H().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new M7(this.gpgpu),this.numMBBeforeWarning=nJ(),this.texData=new jh(this,Na())}nextDataId(){return Rf.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(e,t,n,a,r,s){let i=this.makeTensorInfo(t,n),o=this.texData.get(i.dataId);o.isPacked=!1,o.texture={texture:e,texShape:[a,r]},o.texShape=[a,r];let l=Mp(t),u=new jk(l,!1,s),p=this.runWebGLProgram(u,[i],n,[[a,r]]);return p.shape=t,o.texture=null,this.disposeIntermediateTensorInfo(i),p.dataId}write(e,t,n){if((H().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||H().getBool("DEBUG"))&&this.checkNumericalProblems(e),n==="complex64"&&e!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let a={id:this.nextDataId()};return this.texData.set(a,{shape:t,dtype:n,values:e,usage:pa.UPLOAD,refCount:1}),a}refCount(e){return this.texData.has(e)?this.texData.get(e).refCount:0}incRef(e){let t=this.texData.get(e);t.refCount++}decRef(e){if(this.texData.has(e)){let t=this.texData.get(e);t.refCount--}}move(e,t,n,a,r){if(H().getBool("DEBUG")&&this.checkNumericalProblems(t),a==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:n,dtype:a,values:t,usage:pa.UPLOAD,refCount:r})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){let t=this.texData.get(e),{values:n,dtype:a,complexTensorInfos:r,slice:s,shape:i,isPacked:o}=t;if(s!=null){let d;o?d=new js(i,jo):d=new Sr(i,jo);let c=this.runWebGLProgram(d,[{dataId:e,shape:i,dtype:a}],a),h=this.readSync(c.dataId);return this.disposeIntermediateTensorInfo(c),h}if(n!=null)return this.convertAndCacheOnCPU(e);if(a==="string")return n;let l=this.activeTimers!=null,u;l&&(u=v.now());let p;if(a==="complex64"){let d=this.readSync(r.real.dataId),c=this.readSync(r.imag.dataId);p=N.mergeRealAndImagArrays(d,c)}else p=this.getValuesFromTexture(e);return l&&(this.downloadWaitMs+=v.now()-u),this.convertAndCacheOnCPU(e,p)}async read(e){if(this.pendingRead.has(e)){let h=this.pendingRead.get(e);return new Promise(m=>h.push(m))}let t=this.texData.get(e),{values:n,shape:a,slice:r,dtype:s,complexTensorInfos:i,isPacked:o}=t;if(r!=null){let h;o?h=new js(a,jo):h=new Sr(a,jo);let m=this.runWebGLProgram(h,[{dataId:e,shape:a,dtype:s}],s),f=this.read(m.dataId);return this.disposeIntermediateTensorInfo(m),f}if(n!=null)return this.convertAndCacheOnCPU(e);if(H().getBool("DEBUG")&&!H().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&H().getNumber("WEBGL_VERSION")===2)throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let l=null,u;if(s!=="complex64"&&H().get("WEBGL_BUFFER_SUPPORTED")){u=this.decode(e);let h=this.texData.get(u.dataId);l=this.gpgpu.createBufferFromTexture(h.texture.texture,...eh(a))}this.pendingRead.set(e,[]),s!=="complex64"&&await this.gpgpu.createAndWaitForFence();let p;if(s==="complex64"){let h=await Promise.all([this.read(i.real.dataId),this.read(i.imag.dataId)]),m=h[0],f=h[1];p=N.mergeRealAndImagArrays(m,f)}else if(l==null)p=this.getValuesFromTexture(e);else{let h=v.sizeFromShape(a);p=this.gpgpu.downloadFloat32MatrixFromBuffer(l,h)}if(u!=null&&this.disposeIntermediateTensorInfo(u),l!=null){let h=this.gpgpu.gl;me(h,()=>h.deleteBuffer(l))}let d=this.convertAndCacheOnCPU(e,p),c=this.pendingRead.get(e);return this.pendingRead.delete(e),c.forEach(h=>h(d)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&Na().removeDataId(e,this),this.pendingDeletes--),d}readToGPU(e,t={}){let n=this.texData.get(e),{values:a,shape:r,slice:s,dtype:i,isPacked:o,texture:l}=n;if(i==="complex64")throw new Error("Does not support reading texture for complex64 dtype.");if(s!=null){let c;o?c=new js(r,jo):c=new Sr(r,jo);let h=this.runWebGLProgram(c,[{dataId:e,shape:r,dtype:i}],i),m=this.readToGPU(h,t);return this.disposeIntermediateTensorInfo(h),m}if(l==null)throw a!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let u=this.decode(e,t.customTexShape),p=Na().makeTensorFromTensorInfo(u),d=this.texData.get(u.dataId);return Object.assign({tensorRef:p},d.texture)}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(a=>v.decodeString(a));return Pe(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Pe(e.shape,e.dtype,t)}checkNumericalProblems(e){if(e!=null)for(let t=0;t<e.length;t++){let n=e[t];if(!HC(n))throw H().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(e){let{shape:t,dtype:n,isPacked:a}=this.texData.get(e),r=v.sizeFromShape(t);if(H().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let d=this.decode(e),c=this.texData.get(d.dataId),h=this.gpgpu.downloadMatrixFromPackedTexture(c.texture.texture,...eh(t)).subarray(0,r);return this.disposeIntermediateTensorInfo(d),h}let s=H().getBool("WEBGL_PACK")&&a===!0,i=s?Mp(t):t,o=s?new zZ(i):new LZ(i),l=this.runWebGLProgram(o,[{shape:i,dtype:n,dataId:e}],"float32"),u=this.texData.get(l.dataId),p=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(u.texture.texture,u.texShape[0],u.texShape[1]).subarray(0,r);return this.disposeIntermediateTensorInfo(l),p}timerAvailable(){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){let t=this.activeTimers,n=[],a=!1;this.programTimersStack==null?(this.programTimersStack=n,a=!0):this.activeTimers.push(n),this.activeTimers=n,e();let r=v.flatten(this.activeTimers.map(o=>o.query)).filter(o=>o!=null),s=v.flatten(this.activeTimers.map(o=>o.name)).filter(o=>o!=null);this.activeTimers=t,a&&(this.programTimersStack=null);let i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let o=await Promise.all(r);i.kernelMs=v.sum(o),i.getExtraProfileInfo=()=>o.map((l,u)=>({name:s[u],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(", ")}else i.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,i})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:v.now(),endMs:null}}endTimer(e){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=v.now(),e)}async getQueryTime(e){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);let t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);let{complexTensorInfos:n}=this.texData.get(e);return n!=null&&(this.disposeData(n.real.dataId,t),this.disposeData(n.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){let{texture:t,dtype:n,texShape:a,usage:r,isPacked:s,slice:i}=this.texData.get(e),o=i&&i.origDataId||e,l=this.dataRefCount.get(o);l>1?this.dataRefCount.set(o,l-1):(this.dataRefCount.delete(o),t!=null&&(this.numBytesInGPU-=this.computeBytes(a,n),this.textureManager.releaseTexture(t,a,r,s)));let u=this.texData.get(e);u.texture=null,u.texShape=null,u.isPacked=!1,u.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=eJ){return H().getBool("WEBGL_CPU_FORWARD")&&e.every(n=>this.texData.get(n.dataId).texture==null&&v.sizeFromShape(n.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){N.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let t=e.dataSync();return Y7(e.shape,t)}packedUnaryOp(e,t,n){let a=new js(e.shape,t),r=this.compileAndRun(a,[e],n);return Na().makeTensorFromTensorInfo(r)}abs(e){if(this.shouldExecuteOnCPU([e])&&e.dtype!=="complex64"){let a=L_(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,a)}if(H().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,Yk,e.dtype);let t=new Sr(e.shape,Yk),n=this.compileAndRun(t,[e]);return Na().makeTensorFromTensorInfo(n)}makeTensorInfo(e,t,n){let a;if(t==="string"&&n!=null&&n.length>0&&v.isString(n[0])){let r=n.map(s=>v.encodeString(s));a=this.write(r,e,t)}else a=this.write(n,e,t);return this.texData.get(a).usage=null,{dataId:a,shape:e,dtype:t}}makeOutput(e,t,n){return Na().makeTensorFromTensorInfo(this.makeTensorInfo(e,t,n),this)}unpackTensor(e){let t=new X7(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){let t=new D7(e.shape),n=!0;return this.runWebGLProgram(t,[e],e.dtype,null,n)}packedReshape(e,t){let n=[pi(e.shape),...ci(e.shape)],a={dtype:e.dtype,shape:n,dataId:e.dataId},r=[pi(t),...ci(t)],s=new B_(r,n),i=!0,o=[n],l=this.runWebGLProgram(s,[a],e.dtype,o,i);return{dataId:l.dataId,shape:t,dtype:l.dtype}}decode(e,t){let n=this.texData.get(e),{isPacked:a,shape:r,dtype:s}=n;if(t!=null){let d=v.sizeFromShape(r),c=t[0]*t[1]*4;v.assert(d<=c,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let i=Mp(r),o;a?o=new OZ(i):o=new PZ(i);let l=!0,u=[t!=null?t:eh(i)],p=this.runWebGLProgram(o,[{shape:i,dtype:s,dataId:e}],s,u,l,t);return{dtype:s,shape:r,dataId:p.dataId}}runWebGLProgram(e,t,n,a,r=!1,s){let i=this.makeTensorInfo(e.outputShape,n),o=this.texData.get(i.dataId);if(e.packedOutput&&(o.isPacked=!0),e.outPackingScheme===Jp.DENSE){let g=s!=null?s:eh(e.outputShape);o.texShape=g.map(y=>y*2)}if(e.outTexUsage!=null&&(o.usage=e.outTexUsage),v.sizeFromShape(i.shape)===0)return o.values=v.getTypedArrayFromDType(i.dtype,0),i;let l=[],u=t.map(g=>{if(g.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let y=this.texData.get(g.dataId);if(y.texture==null){if(!e.packedInputs&&v.sizeFromShape(g.shape)<=H().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:g.shape,texData:null,isUniform:!0,uniformValues:y.values};e.packedInputs&&(y.isPacked=!0,y.shape=g.shape)}if(this.uploadToGPU(g.dataId),!!y.isPacked!=!!e.packedInputs)g=y.isPacked?this.unpackTensor(g):this.packTensor(g),l.push(g),y=this.texData.get(g.dataId);else if(y.isPacked&&!Qp(y.shape,g.shape)){let b=g,x=g.shape;g.shape=y.shape,g=this.packedReshape(g,x),l.push(g),y=this.texData.get(g.dataId),b.shape=x}return{shape:g.shape,texData:y,isUniform:!1}});this.uploadToGPU(i.dataId);let p={shape:i.shape,texData:o,isUniform:!1},d=MZ(e,u,p),c=this.getAndSaveBinary(d,()=>DZ(this.gpgpu,e,u,p)),h=this.activeTimers!=null,m;h&&(m=this.startTimer()),H().get("ENGINE_COMPILE_ONLY")||RZ(this.gpgpu,c,u,p,a),l.forEach(g=>this.disposeIntermediateTensorInfo(g)),h&&(m=this.endTimer(m),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(m)}));let f=H().get("WEBGL_FLUSH_THRESHOLD");if(f>0){let g=v.now();g-this.lastGlFlushTime>f&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=g)}if(!H().getBool("WEBGL_LAZILY_UNPACK")&&o.isPacked&&r===!1){let g=this.unpackTensor(i);return this.disposeIntermediateTensorInfo(i),g}return i}compileAndRun(e,t,n,a,r=!1){return n=n||t[0].dtype,this.runWebGLProgram(e,t,n,a,r)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(H().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!="undefined"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=P(()=>{if(!H().get("WEBGL_RENDER_FLOAT32_ENABLED")){let e=H().getBool("DEBUG");H().set("DEBUG",!1);let t=this.abs(be(1e-8)).dataSync()[0];if(H().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?Z7:J7}uploadToGPU(e){let t=this.texData.get(e),{shape:n,dtype:a,values:r,texture:s,usage:i,isPacked:o}=t;if(s!=null)return;let l=this.activeTimers!=null,u;l&&(u=v.now());let p=t.texShape;if(p==null&&(p=l_(n,o),t.texShape=p),r!=null){let d=Mp(n),c,h=p[1],m=p[0],f=r instanceof Uint8Array||r instanceof Uint8ClampedArray;(o||!f)&&([h,m]=Ou(p[0],p[1])),o?c=new BZ(d,f):c=new jk(d,f);let g=f?[m,h]:p,y=this.makeTensorInfo(g,a),b=this.texData.get(y.dataId);f?b.usage=pa.PIXELS:b.usage=pa.UPLOAD,b.texShape=g,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(y.dataId),h,m,r);let x=[[m,h]],w=!0,I=this.runWebGLProgram(c,[y],a,x,w),T=this.texData.get(I.dataId);t.texShape=T.texShape,t.isPacked=T.isPacked,t.usage=T.usage,H().get("ENGINE_COMPILE_ONLY")?this.disposeData(I.dataId):(t.texture=T.texture,t.values=null,this.texData.delete(I.dataId)),this.disposeIntermediateTensorInfo(y),l&&(this.uploadWaitMs+=v.now()-u)}else{let d=this.acquireTexture(p,i,a,o);t.texture=d}}convertAndCacheOnCPU(e,t){let n=this.texData.get(e),{dtype:a}=n;return this.releaseGPUData(e),t!=null&&(n.values=aJ(t,a)),n.values}acquireTexture(e,t,n,a){if(this.numBytesInGPU+=this.computeBytes(e,n),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let r=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${r} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,a)}computeBytes(e,t){return e[0]*e[1]*v.bytesPerElement(t)}checkCompileCompletion(){for(let[,e]of Object.entries(this.binaryCache))this.checkCompletion_(e)}async checkCompileCompletionAsync(){let e=[];if(this.gpgpu.parallelCompilationExtension){for(let[,t]of Object.entries(this.binaryCache))e.push(this.checkCompletionAsync_(t));return Promise.all(e)}else{for(let[,t]of Object.entries(this.binaryCache)){let n=new Promise(a=>{try{this.checkCompletion_(t),a(!0)}catch(r){throw r}});e.push(n)}return Promise.all(e)}}async checkCompletionAsync_(e){return this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(e):(await Gv(),this.checkCompletionAsync_(e))}checkCompletion_(e){if(this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(e.webGLProgram)),this.gpgpu.gl.getShaderParameter(e.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(V0(e.source,this.gpgpu.gl.getShaderInfoLog(e.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let[,e]of Object.entries(this.binaryCache)){let{uniformLocations:t,customUniformLocations:n,infLoc:a,nanLoc:r,inShapesLocations:s,inTexShapesLocations:i,outShapeLocation:o,outShapeStridesLocation:l,outTexShapeLocation:u}=x_(this.gpgpu,e.program,e.webGLProgram);e.uniformLocations=t,e.customUniformLocations=n,e.infLoc=a,e.nanLoc=r,e.inShapesLocations=s,e.inTexShapesLocations=i,e.outShapeLocation=o,e.outShapeStridesLocation=l,e.outTexShapeLocation=u}}createTensorFromTexture(e,t,n){let{texture:a,height:r,width:s,channels:i}=e,o=Na().backend;if(!o.gpgpu.gl.isTexture(a))throw new Error("The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.");let l=o.writeTexture(a,t,n,r,s,i);return Na().makeTensorFromDataId(l,t,n,o)}};Rf.nextDataId=0;function aJ(e,t){if(t==="float32"||t==="complex64")return e;if(t==="int32"||t==="bool"){let n=t==="int32"?new Int32Array(e.length):new Uint8Array(e.length);for(let a=0;a<n.length;++a)n[a]=Math.round(e[a]);return n}else throw new Error(`Unknown dtype ${t}`)}var rJ="4.0.0";function V_(){H().set("WEBGL_FORCE_F16_TEXTURES",!0)}Tc.isBrowser()&&Em("webgl",()=>new Rf,2);var sJ={forceHalfFloat:V_},J0=`
+    `;
+  }
+};
+var whereImpl3 = kernel_impls_exports.whereImpl;
+var EPSILON_FLOAT322 = 1e-7;
+var EPSILON_FLOAT162 = 1e-4;
+var binaryCaches = {};
+function getBinaryCache(webGLVersion) {
+  if (webGLVersion in binaryCaches) {
+    return binaryCaches[webGLVersion];
+  }
+  binaryCaches[webGLVersion] = {};
+  return binaryCaches[webGLVersion];
+}
+var CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber("CPU_HANDOFF_SIZE_THRESHOLD");
+var BEFORE_PAGING_CONSTANT = 600;
+function numMBBeforeWarning() {
+  if (env().global.screen == null) {
+    return 1024;
+  }
+  return env().global.screen.height * env().global.screen.width * window.devicePixelRatio * BEFORE_PAGING_CONSTANT / 1024 / 1024;
+}
+var MathBackendWebGL = class extends KernelBackend {
+  constructor(gpuResource) {
+    super();
+    this.pendingRead = /* @__PURE__ */ new WeakMap();
+    this.pendingDisposal = /* @__PURE__ */ new WeakSet();
+    this.dataRefCount = /* @__PURE__ */ new WeakMap();
+    this.numBytesInGPU = 0;
+    this.uploadWaitMs = 0;
+    this.downloadWaitMs = 0;
+    this.lastGlFlushTime = 0;
+    this.warnedAboutMemory = false;
+    this.pendingDeletes = 0;
+    this.disposed = false;
+    if (!env().getBool("HAS_WEBGL")) {
+      throw new Error("WebGL is not supported on this device");
+    }
+    let newGPGPU;
+    if (gpuResource != null) {
+      if (gpuResource instanceof GPGPUContext) {
+        newGPGPU = gpuResource;
+      } else {
+        const gl = getWebGLContext(env().getNumber("WEBGL_VERSION"), gpuResource);
+        newGPGPU = new GPGPUContext(gl);
+      }
+      this.binaryCache = {};
+      this.gpgpuCreatedLocally = false;
+    } else {
+      const gl = getWebGLContext(env().getNumber("WEBGL_VERSION"));
+      newGPGPU = new GPGPUContext(gl);
+      this.binaryCache = getBinaryCache(env().getNumber("WEBGL_VERSION"));
+      this.gpgpuCreatedLocally = true;
+    }
+    this.gpgpu = newGPGPU;
+    this.canvas = this.gpgpu.gl.canvas;
+    this.textureManager = new TextureManager(this.gpgpu);
+    this.numMBBeforeWarning = numMBBeforeWarning();
+    this.texData = new DataStorage(this, engine());
+  }
+  nextDataId() {
+    return MathBackendWebGL.nextDataId++;
+  }
+  numDataIds() {
+    return this.texData.numDataIds() - this.pendingDeletes;
+  }
+  writeTexture(texture, shape, dtype, texHeight, texWidth, channels) {
+    const input2 = this.makeTensorInfo(shape, dtype);
+    const inData = this.texData.get(input2.dataId);
+    inData.isPacked = false;
+    inData.texture = { texture, texShape: [texHeight, texWidth] };
+    inData.texShape = [texHeight, texWidth];
+    const shapeAs3D = getShapeAs3D(shape);
+    const program = new EncodeMatrixProgram(shapeAs3D, false, channels);
+    const output = this.runWebGLProgram(program, [input2], dtype, [[texHeight, texWidth]]);
+    output.shape = shape;
+    inData.texture = null;
+    this.disposeIntermediateTensorInfo(input2);
+    return output.dataId;
+  }
+  write(values, shape, dtype) {
+    if (env().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS") || env().getBool("DEBUG")) {
+      this.checkNumericalProblems(values);
+    }
+    if (dtype === "complex64" && values != null) {
+      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);
+    }
+    const dataId = { id: this.nextDataId() };
+    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount: 1 });
+    return dataId;
+  }
+  refCount(dataId) {
+    if (this.texData.has(dataId)) {
+      const tensorData = this.texData.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const texData = this.texData.get(dataId);
+    texData.refCount++;
+  }
+  decRef(dataId) {
+    if (this.texData.has(dataId)) {
+      const texData = this.texData.get(dataId);
+      texData.refCount--;
+    }
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    if (env().getBool("DEBUG")) {
+      this.checkNumericalProblems(values);
+    }
+    if (dtype === "complex64") {
+      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);
+    }
+    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount });
+  }
+  disposeIntermediateTensorInfo(tensorInfo) {
+    this.disposeData(tensorInfo.dataId);
+  }
+  readSync(dataId) {
+    const texData = this.texData.get(dataId);
+    const { values, dtype, complexTensorInfos, slice: slice5, shape, isPacked } = texData;
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const data = this.readSync(res.dataId);
+      this.disposeIntermediateTensorInfo(res);
+      return data;
+    }
+    if (values != null) {
+      return this.convertAndCacheOnCPU(dataId);
+    }
+    if (dtype === "string") {
+      return values;
+    }
+    const shouldTimeProgram = this.activeTimers != null;
+    let start;
+    if (shouldTimeProgram) {
+      start = util_exports.now();
+    }
+    let result;
+    if (dtype === "complex64") {
+      const realValues = this.readSync(complexTensorInfos.real.dataId);
+      const imagValues = this.readSync(complexTensorInfos.imag.dataId);
+      result = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    } else {
+      result = this.getValuesFromTexture(dataId);
+    }
+    if (shouldTimeProgram) {
+      this.downloadWaitMs += util_exports.now() - start;
+    }
+    return this.convertAndCacheOnCPU(dataId, result);
+  }
+  async read(dataId) {
+    if (this.pendingRead.has(dataId)) {
+      const subscribers2 = this.pendingRead.get(dataId);
+      return new Promise((resolve) => subscribers2.push(resolve));
+    }
+    const texData = this.texData.get(dataId);
+    const { values, shape, slice: slice5, dtype, complexTensorInfos, isPacked } = texData;
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const data = this.read(res.dataId);
+      this.disposeIntermediateTensorInfo(res);
+      return data;
+    }
+    if (values != null) {
+      return this.convertAndCacheOnCPU(dataId);
+    }
+    if (env().getBool("DEBUG")) {
+      if (!env().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED") && env().getNumber("WEBGL_VERSION") === 2) {
+        throw new Error(`tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.`);
+      }
+    }
+    let buffer2 = null;
+    let tmpDownloadTarget;
+    if (dtype !== "complex64" && env().get("WEBGL_BUFFER_SUPPORTED")) {
+      tmpDownloadTarget = this.decode(dataId);
+      const tmpData = this.texData.get(tmpDownloadTarget.dataId);
+      buffer2 = this.gpgpu.createBufferFromTexture(tmpData.texture.texture, ...getDenseTexShape(shape));
+    }
+    this.pendingRead.set(dataId, []);
+    if (dtype !== "complex64") {
+      await this.gpgpu.createAndWaitForFence();
+    }
+    let vals;
+    if (dtype === "complex64") {
+      const ps = await Promise.all([
+        this.read(complexTensorInfos.real.dataId),
+        this.read(complexTensorInfos.imag.dataId)
+      ]);
+      const realValues = ps[0];
+      const imagValues = ps[1];
+      vals = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    } else if (buffer2 == null) {
+      vals = this.getValuesFromTexture(dataId);
+    } else {
+      const size = util_exports.sizeFromShape(shape);
+      vals = this.gpgpu.downloadFloat32MatrixFromBuffer(buffer2, size);
+    }
+    if (tmpDownloadTarget != null) {
+      this.disposeIntermediateTensorInfo(tmpDownloadTarget);
+    }
+    if (buffer2 != null) {
+      const gl = this.gpgpu.gl;
+      callAndCheck(gl, () => gl.deleteBuffer(buffer2));
+    }
+    const dTypeVals = this.convertAndCacheOnCPU(dataId, vals);
+    const subscribers = this.pendingRead.get(dataId);
+    this.pendingRead.delete(dataId);
+    subscribers.forEach((resolve) => resolve(dTypeVals));
+    if (this.pendingDisposal.has(dataId)) {
+      this.pendingDisposal.delete(dataId);
+      if (this.disposeData(dataId)) {
+        engine().removeDataId(dataId, this);
+      }
+      this.pendingDeletes--;
+    }
+    return dTypeVals;
+  }
+  readToGPU(dataId, options = {}) {
+    const texData = this.texData.get(dataId);
+    const { values, shape, slice: slice5, dtype, isPacked, texture } = texData;
+    if (dtype === "complex64") {
+      throw new Error("Does not support reading texture for complex64 dtype.");
+    }
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const gpuResouorce = this.readToGPU(res, options);
+      this.disposeIntermediateTensorInfo(res);
+      return gpuResouorce;
+    }
+    if (texture == null) {
+      if (values != null) {
+        throw new Error("Data is not on GPU but on CPU.");
+      } else {
+        throw new Error("There is no data on GPU or CPU.");
+      }
+    }
+    const tmpTarget = this.decode(dataId, options.customTexShape);
+    const tensorRef = engine().makeTensorFromTensorInfo(tmpTarget);
+    const tmpData = this.texData.get(tmpTarget.dataId);
+    return Object.assign({ tensorRef }, tmpData.texture);
+  }
+  bufferSync(t) {
+    const data = this.readSync(t.dataId);
+    if (t.dtype === "string") {
+      try {
+        const strings = data.map((d) => util_exports.decodeString(d));
+        return buffer(t.shape, t.dtype, strings);
+      } catch (_a) {
+        throw new Error("Failed to decode encoded string bytes into utf-8");
+      }
+    }
+    return buffer(t.shape, t.dtype, data);
+  }
+  checkNumericalProblems(values) {
+    if (values == null) {
+      return;
+    }
+    for (let i = 0; i < values.length; i++) {
+      const num = values[i];
+      if (!canBeRepresented(num)) {
+        if (env().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")) {
+          throw Error(`The value ${num} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`);
+        }
+        throw Error(`The value ${num} cannot be represented on this device.`);
+      }
+    }
+  }
+  getValuesFromTexture(dataId) {
+    const { shape, dtype, isPacked } = this.texData.get(dataId);
+    const size = util_exports.sizeFromShape(shape);
+    if (env().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")) {
+      const tmpTarget = this.decode(dataId);
+      const tmpData2 = this.texData.get(tmpTarget.dataId);
+      const vals2 = this.gpgpu.downloadMatrixFromPackedTexture(tmpData2.texture.texture, ...getDenseTexShape(shape)).subarray(0, size);
+      this.disposeIntermediateTensorInfo(tmpTarget);
+      return vals2;
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK") && isPacked === true;
+    const outputShape = shouldUsePackedProgram ? getShapeAs3D(shape) : shape;
+    const program = shouldUsePackedProgram ? new EncodeFloatPackedProgram(outputShape) : new EncodeFloatProgram(outputShape);
+    const output = this.runWebGLProgram(program, [{ shape: outputShape, dtype, dataId }], "float32");
+    const tmpData = this.texData.get(output.dataId);
+    const vals = this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(tmpData.texture.texture, tmpData.texShape[0], tmpData.texShape[1]).subarray(0, size);
+    this.disposeIntermediateTensorInfo(output);
+    return vals;
+  }
+  timerAvailable() {
+    return env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0;
+  }
+  time(f) {
+    const oldActiveTimers = this.activeTimers;
+    const newActiveTimers = [];
+    let outerMostTime = false;
+    if (this.programTimersStack == null) {
+      this.programTimersStack = newActiveTimers;
+      outerMostTime = true;
+    } else {
+      this.activeTimers.push(newActiveTimers);
+    }
+    this.activeTimers = newActiveTimers;
+    f();
+    const flattenedActiveTimerQueries = util_exports.flatten(this.activeTimers.map((d) => d.query)).filter((d) => d != null);
+    const flattenedActiveTimerNames = util_exports.flatten(this.activeTimers.map((d) => d.name)).filter((d) => d != null);
+    this.activeTimers = oldActiveTimers;
+    if (outerMostTime) {
+      this.programTimersStack = null;
+    }
+    const res = {
+      uploadWaitMs: this.uploadWaitMs,
+      downloadWaitMs: this.downloadWaitMs,
+      kernelMs: null,
+      wallMs: null
+    };
+    return (async () => {
+      if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+        const kernelMs = await Promise.all(flattenedActiveTimerQueries);
+        res["kernelMs"] = util_exports.sum(kernelMs);
+        res["getExtraProfileInfo"] = () => kernelMs.map((d, i) => ({ name: flattenedActiveTimerNames[i], ms: d })).map((d) => `${d.name}: ${d.ms}`).join(", ");
+      } else {
+        res["kernelMs"] = {
+          error: "WebGL query timers are not supported in this environment."
+        };
+      }
+      this.uploadWaitMs = 0;
+      this.downloadWaitMs = 0;
+      return res;
+    })();
+  }
+  memory() {
+    return {
+      unreliable: false,
+      numBytesInGPU: this.numBytesInGPU,
+      numBytesInGPUAllocated: this.textureManager.numBytesAllocated,
+      numBytesInGPUFree: this.textureManager.numBytesFree
+    };
+  }
+  startTimer() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      return this.gpgpu.beginQuery();
+    }
+    return { startMs: util_exports.now(), endMs: null };
+  }
+  endTimer(query) {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      this.gpgpu.endQuery();
+      return query;
+    }
+    query.endMs = util_exports.now();
+    return query;
+  }
+  async getQueryTime(query) {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      return this.gpgpu.waitForQueryAndGetTime(query);
+    }
+    const timerQuery = query;
+    return timerQuery.endMs - timerQuery.startMs;
+  }
+  disposeData(dataId, force = false) {
+    if (this.pendingDisposal.has(dataId)) {
+      return false;
+    }
+    if (!this.texData.has(dataId)) {
+      return true;
+    }
+    if (force) {
+      this.texData.get(dataId).refCount = 0;
+    } else {
+      this.texData.get(dataId).refCount--;
+    }
+    if (!force && this.texData.get(dataId).refCount > 0) {
+      return false;
+    }
+    if (this.pendingRead.has(dataId)) {
+      this.pendingDisposal.add(dataId);
+      this.pendingDeletes++;
+      return false;
+    }
+    this.releaseGPUData(dataId);
+    const { complexTensorInfos } = this.texData.get(dataId);
+    if (complexTensorInfos != null) {
+      this.disposeData(complexTensorInfos.real.dataId, force);
+      this.disposeData(complexTensorInfos.imag.dataId, force);
+    }
+    this.texData.delete(dataId);
+    return true;
+  }
+  releaseGPUData(dataId) {
+    const { texture, dtype, texShape, usage, isPacked, slice: slice5 } = this.texData.get(dataId);
+    const key = slice5 && slice5.origDataId || dataId;
+    const refCount = this.dataRefCount.get(key);
+    if (refCount > 1) {
+      this.dataRefCount.set(key, refCount - 1);
+    } else {
+      this.dataRefCount.delete(key);
+      if (texture != null) {
+        this.numBytesInGPU -= this.computeBytes(texShape, dtype);
+        this.textureManager.releaseTexture(texture, texShape, usage, isPacked);
+      }
+    }
+    const texData = this.texData.get(dataId);
+    texData.texture = null;
+    texData.texShape = null;
+    texData.isPacked = false;
+    texData.slice = null;
+  }
+  getTexture(dataId) {
+    this.uploadToGPU(dataId);
+    return this.texData.get(dataId).texture.texture;
+  }
+  getDataInfo(dataId) {
+    return this.texData.get(dataId);
+  }
+  shouldExecuteOnCPU(inputs, sizeThreshold = CPU_HANDOFF_SIZE_THRESHOLD) {
+    return env().getBool("WEBGL_CPU_FORWARD") && inputs.every((input2) => this.texData.get(input2.dataId).texture == null && util_exports.sizeFromShape(input2.shape) < sizeThreshold);
+  }
+  getGPGPUContext() {
+    return this.gpgpu;
+  }
+  where(condition) {
+    backend_util_exports.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");
+    const condVals = condition.dataSync();
+    return whereImpl3(condition.shape, condVals);
+  }
+  packedUnaryOp(x, op2, dtype) {
+    const program = new UnaryOpPackedProgram(x.shape, op2);
+    const outInfo = this.compileAndRun(program, [x], dtype);
+    return engine().makeTensorFromTensorInfo(outInfo);
+  }
+  abs(x) {
+    if (this.shouldExecuteOnCPU([x]) && x.dtype !== "complex64") {
+      const outValues = simpleAbsImplCPU(this.texData.get(x.dataId).values);
+      return this.makeOutput(x.shape, x.dtype, outValues);
+    }
+    if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+      return this.packedUnaryOp(x, ABS, x.dtype);
+    }
+    const program = new UnaryOpProgram(x.shape, ABS);
+    const outInfo = this.compileAndRun(program, [x]);
+    return engine().makeTensorFromTensorInfo(outInfo);
+  }
+  makeTensorInfo(shape, dtype, values) {
+    let dataId;
+    if (dtype === "string" && values != null && values.length > 0 && util_exports.isString(values[0])) {
+      const encodedValues = values.map((d) => util_exports.encodeString(d));
+      dataId = this.write(encodedValues, shape, dtype);
+    } else {
+      dataId = this.write(values, shape, dtype);
+    }
+    this.texData.get(dataId).usage = null;
+    return { dataId, shape, dtype };
+  }
+  makeOutput(shape, dtype, values) {
+    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);
+  }
+  unpackTensor(input2) {
+    const program = new UnpackProgram(input2.shape);
+    return this.runWebGLProgram(program, [input2], input2.dtype);
+  }
+  packTensor(input2) {
+    const program = new PackProgram(input2.shape);
+    const preventEagerUnpackingOutput = true;
+    return this.runWebGLProgram(program, [input2], input2.dtype, null, preventEagerUnpackingOutput);
+  }
+  packedReshape(input2, afterShape) {
+    const input3DShape = [
+      getBatchDim(input2.shape),
+      ...getRowsCols(input2.shape)
+    ];
+    const input3D = {
+      dtype: input2.dtype,
+      shape: input3DShape,
+      dataId: input2.dataId
+    };
+    const afterShapeAs3D = [
+      getBatchDim(afterShape),
+      ...getRowsCols(afterShape)
+    ];
+    const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);
+    const preventEagerUnpackingOfOutput = true;
+    const customValues = [input3DShape];
+    const output = this.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);
+    return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };
+  }
+  decode(dataId, customTexShape) {
+    const texData = this.texData.get(dataId);
+    const { isPacked, shape, dtype } = texData;
+    if (customTexShape != null) {
+      const size = util_exports.sizeFromShape(shape);
+      const texSize = customTexShape[0] * customTexShape[1] * 4;
+      util_exports.assert(size <= texSize, () => "customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.");
+    }
+    const shapeAs3D = getShapeAs3D(shape);
+    let program;
+    if (isPacked) {
+      program = new DecodeMatrixPackedProgram(shapeAs3D);
+    } else {
+      program = new DecodeMatrixProgram(shapeAs3D);
+    }
+    const preventEagerUnpackingOfOutput = true;
+    const customValues = [customTexShape != null ? customTexShape : getDenseTexShape(shapeAs3D)];
+    const out = this.runWebGLProgram(program, [{ shape: shapeAs3D, dtype, dataId }], dtype, customValues, preventEagerUnpackingOfOutput, customTexShape);
+    return { dtype, shape, dataId: out.dataId };
+  }
+  runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false, customTexShape) {
+    const output = this.makeTensorInfo(program.outputShape, outputDtype);
+    const outData = this.texData.get(output.dataId);
+    if (program.packedOutput) {
+      outData.isPacked = true;
+    }
+    if (program.outPackingScheme === PackingScheme.DENSE) {
+      const texelShape = customTexShape != null ? customTexShape : getDenseTexShape(program.outputShape);
+      outData.texShape = texelShape.map((d) => d * 2);
+    }
+    if (program.outTexUsage != null) {
+      outData.usage = program.outTexUsage;
+    }
+    if (util_exports.sizeFromShape(output.shape) === 0) {
+      outData.values = util_exports.getTypedArrayFromDType(output.dtype, 0);
+      return output;
+    }
+    const dataToDispose = [];
+    const inputsData = inputs.map((input2) => {
+      if (input2.dtype === "complex64") {
+        throw new Error(`GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.`);
+      }
+      let texData = this.texData.get(input2.dataId);
+      if (texData.texture == null) {
+        if (!program.packedInputs && util_exports.sizeFromShape(input2.shape) <= env().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM")) {
+          return {
+            shape: input2.shape,
+            texData: null,
+            isUniform: true,
+            uniformValues: texData.values
+          };
+        }
+        if (program.packedInputs) {
+          texData.isPacked = true;
+          texData.shape = input2.shape;
+        }
+      }
+      this.uploadToGPU(input2.dataId);
+      if (!!texData.isPacked !== !!program.packedInputs) {
+        input2 = texData.isPacked ? this.unpackTensor(input2) : this.packTensor(input2);
+        dataToDispose.push(input2);
+        texData = this.texData.get(input2.dataId);
+      } else if (texData.isPacked && !isReshapeFree(texData.shape, input2.shape)) {
+        const savedInput = input2;
+        const targetShape = input2.shape;
+        input2.shape = texData.shape;
+        input2 = this.packedReshape(input2, targetShape);
+        dataToDispose.push(input2);
+        texData = this.texData.get(input2.dataId);
+        savedInput.shape = targetShape;
+      }
+      return { shape: input2.shape, texData, isUniform: false };
+    });
+    this.uploadToGPU(output.dataId);
+    const outputData = { shape: output.shape, texData: outData, isUniform: false };
+    const key = makeShaderKey(program, inputsData, outputData);
+    const binary = this.getAndSaveBinary(key, () => {
+      return compileProgram(this.gpgpu, program, inputsData, outputData);
+    });
+    const shouldTimeProgram = this.activeTimers != null;
+    let query;
+    if (shouldTimeProgram) {
+      query = this.startTimer();
+    }
+    if (!env().get("ENGINE_COMPILE_ONLY")) {
+      runProgram(this.gpgpu, binary, inputsData, outputData, customUniformValues);
+    }
+    dataToDispose.forEach((info) => this.disposeIntermediateTensorInfo(info));
+    if (shouldTimeProgram) {
+      query = this.endTimer(query);
+      this.activeTimers.push({ name: program.constructor.name, query: this.getQueryTime(query) });
+    }
+    const glFlushThreshold = env().get("WEBGL_FLUSH_THRESHOLD");
+    if (glFlushThreshold > 0) {
+      const time2 = util_exports.now();
+      if (time2 - this.lastGlFlushTime > glFlushThreshold) {
+        this.gpgpu.gl.flush();
+        this.lastGlFlushTime = time2;
+      }
+    }
+    if (!env().getBool("WEBGL_LAZILY_UNPACK") && outData.isPacked && preventEagerUnpackingOfOutput === false) {
+      const unpacked = this.unpackTensor(output);
+      this.disposeIntermediateTensorInfo(output);
+      return unpacked;
+    }
+    return output;
+  }
+  compileAndRun(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false) {
+    outputDtype = outputDtype || inputs[0].dtype;
+    const outInfo = this.runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput);
+    return outInfo;
+  }
+  getAndSaveBinary(key, getBinary) {
+    if (!(key in this.binaryCache)) {
+      this.binaryCache[key] = getBinary();
+    }
+    return this.binaryCache[key];
+  }
+  getTextureManager() {
+    return this.textureManager;
+  }
+  dispose() {
+    if (this.disposed) {
+      return;
+    }
+    if (!env().getBool("IS_TEST")) {
+      const allKeys = Object.keys(this.binaryCache);
+      allKeys.forEach((key) => {
+        this.gpgpu.deleteProgram(this.binaryCache[key].webGLProgram);
+        delete this.binaryCache[key];
+      });
+    }
+    this.textureManager.dispose();
+    if (this.canvas != null && (typeof HTMLCanvasElement !== "undefined" && this.canvas instanceof HTMLCanvasElement)) {
+      this.canvas.remove();
+    } else {
+      this.canvas = null;
+    }
+    if (this.gpgpuCreatedLocally) {
+      this.gpgpu.program = null;
+      this.gpgpu.dispose();
+    }
+    this.disposed = true;
+  }
+  floatPrecision() {
+    if (this.floatPrecisionValue == null) {
+      this.floatPrecisionValue = tidy(() => {
+        if (!env().get("WEBGL_RENDER_FLOAT32_ENABLED")) {
+          const debugFlag = env().getBool("DEBUG");
+          env().set("DEBUG", false);
+          const underflowCheckValue = this.abs(scalar(1e-8)).dataSync()[0];
+          env().set("DEBUG", debugFlag);
+          if (underflowCheckValue > 0) {
+            return 32;
+          }
+        }
+        return 16;
+      });
+    }
+    return this.floatPrecisionValue;
+  }
+  epsilon() {
+    return this.floatPrecision() === 32 ? EPSILON_FLOAT322 : EPSILON_FLOAT162;
+  }
+  uploadToGPU(dataId) {
+    const texData = this.texData.get(dataId);
+    const { shape, dtype, values, texture, usage, isPacked } = texData;
+    if (texture != null) {
+      return;
+    }
+    const shouldTimeProgram = this.activeTimers != null;
+    let start;
+    if (shouldTimeProgram) {
+      start = util_exports.now();
+    }
+    let texShape = texData.texShape;
+    if (texShape == null) {
+      texShape = getTextureShapeFromLogicalShape(shape, isPacked);
+      texData.texShape = texShape;
+    }
+    if (values != null) {
+      const shapeAs3D = getShapeAs3D(shape);
+      let program;
+      let width = texShape[1], height = texShape[0];
+      const isByteArray = values instanceof Uint8Array || values instanceof Uint8ClampedArray;
+      if (isPacked || !isByteArray) {
+        [width, height] = getPackedMatrixTextureShapeWidthHeight(texShape[0], texShape[1]);
+      }
+      if (isPacked) {
+        program = new EncodeMatrixPackedProgram(shapeAs3D, isByteArray);
+      } else {
+        program = new EncodeMatrixProgram(shapeAs3D, isByteArray);
+      }
+      const tempDenseInputTexShape = isByteArray ? [height, width] : texShape;
+      const tempDenseInputHandle = this.makeTensorInfo(tempDenseInputTexShape, dtype);
+      const tempDenseInputTexData = this.texData.get(tempDenseInputHandle.dataId);
+      if (isByteArray) {
+        tempDenseInputTexData.usage = TextureUsage.PIXELS;
+      } else {
+        tempDenseInputTexData.usage = TextureUsage.UPLOAD;
+      }
+      tempDenseInputTexData.texShape = tempDenseInputTexShape;
+      this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(tempDenseInputHandle.dataId), width, height, values);
+      const customValues = [[height, width]];
+      const preventEagerUnpacking = true;
+      const encodedOutputTarget = this.runWebGLProgram(program, [tempDenseInputHandle], dtype, customValues, preventEagerUnpacking);
+      const outputTexData = this.texData.get(encodedOutputTarget.dataId);
+      texData.texShape = outputTexData.texShape;
+      texData.isPacked = outputTexData.isPacked;
+      texData.usage = outputTexData.usage;
+      if (!env().get("ENGINE_COMPILE_ONLY")) {
+        texData.texture = outputTexData.texture;
+        texData.values = null;
+        this.texData.delete(encodedOutputTarget.dataId);
+      } else {
+        this.disposeData(encodedOutputTarget.dataId);
+      }
+      this.disposeIntermediateTensorInfo(tempDenseInputHandle);
+      if (shouldTimeProgram) {
+        this.uploadWaitMs += util_exports.now() - start;
+      }
+    } else {
+      const newTexture = this.acquireTexture(texShape, usage, dtype, isPacked);
+      texData.texture = newTexture;
+    }
+  }
+  convertAndCacheOnCPU(dataId, float32Values) {
+    const texData = this.texData.get(dataId);
+    const { dtype } = texData;
+    this.releaseGPUData(dataId);
+    if (float32Values != null) {
+      texData.values = float32ToTypedArray(float32Values, dtype);
+    }
+    return texData.values;
+  }
+  acquireTexture(texShape, texType, dtype, isPacked) {
+    this.numBytesInGPU += this.computeBytes(texShape, dtype);
+    if (!this.warnedAboutMemory && this.numBytesInGPU > this.numMBBeforeWarning * 1024 * 1024) {
+      const mb = (this.numBytesInGPU / 1024 / 1024).toFixed(2);
+      this.warnedAboutMemory = true;
+      console.warn(`High memory usage in GPU: ${mb} MB, most likely due to a memory leak`);
+    }
+    return this.textureManager.acquireTexture(texShape, texType, isPacked);
+  }
+  computeBytes(shape, dtype) {
+    return shape[0] * shape[1] * util_exports.bytesPerElement(dtype);
+  }
+  checkCompileCompletion() {
+    for (const [, binary] of Object.entries(this.binaryCache)) {
+      this.checkCompletion_(binary);
+    }
+  }
+  async checkCompileCompletionAsync() {
+    const ps = [];
+    if (this.gpgpu.parallelCompilationExtension) {
+      for (const [, binary] of Object.entries(this.binaryCache)) {
+        ps.push(this.checkCompletionAsync_(binary));
+      }
+      return Promise.all(ps);
+    } else {
+      for (const [, binary] of Object.entries(this.binaryCache)) {
+        const p2 = new Promise((resolve) => {
+          try {
+            this.checkCompletion_(binary);
+            resolve(true);
+          } catch (error) {
+            throw error;
+          }
+        });
+        ps.push(p2);
+      }
+      return Promise.all(ps);
+    }
+  }
+  async checkCompletionAsync_(binary) {
+    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)) {
+      return this.checkCompletion_(binary);
+    } else {
+      await nextFrame();
+      return this.checkCompletionAsync_(binary);
+    }
+  }
+  checkCompletion_(binary) {
+    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.gl.LINK_STATUS) === false) {
+      console.log(this.gpgpu.gl.getProgramInfoLog(binary.webGLProgram));
+      if (this.gpgpu.gl.getShaderParameter(binary.fragmentShader, this.gpgpu.gl.COMPILE_STATUS) === false) {
+        logShaderSourceAndInfoLog(binary.source, this.gpgpu.gl.getShaderInfoLog(binary.fragmentShader));
+        throw new Error("Failed to compile fragment shader.");
+      }
+      throw new Error("Failed to link vertex and fragment shaders.");
+    }
+    return true;
+  }
+  getUniformLocations() {
+    for (const [, binary] of Object.entries(this.binaryCache)) {
+      const { uniformLocations, customUniformLocations, infLoc, nanLoc, inShapesLocations, inTexShapesLocations, outShapeLocation, outShapeStridesLocation, outTexShapeLocation } = getUniformLocations(this.gpgpu, binary.program, binary.webGLProgram);
+      binary.uniformLocations = uniformLocations;
+      binary.customUniformLocations = customUniformLocations;
+      binary.infLoc = infLoc;
+      binary.nanLoc = nanLoc;
+      binary.inShapesLocations = inShapesLocations;
+      binary.inTexShapesLocations = inTexShapesLocations;
+      binary.outShapeLocation = outShapeLocation;
+      binary.outShapeStridesLocation = outShapeStridesLocation;
+      binary.outTexShapeLocation = outTexShapeLocation;
+    }
+  }
+  createTensorFromTexture(values, shape, dtype) {
+    const { texture, height, width, channels } = values;
+    const backend2 = engine().backend;
+    if (!backend2.gpgpu.gl.isTexture(texture)) {
+      throw new Error(`The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.`);
+    }
+    const dataId = backend2.writeTexture(texture, shape, dtype, height, width, channels);
+    return engine().makeTensorFromDataId(dataId, shape, dtype, backend2);
+  }
+};
+MathBackendWebGL.nextDataId = 0;
+function float32ToTypedArray(a, dtype) {
+  if (dtype === "float32" || dtype === "complex64") {
+    return a;
+  } else if (dtype === "int32" || dtype === "bool") {
+    const result = dtype === "int32" ? new Int32Array(a.length) : new Uint8Array(a.length);
+    for (let i = 0; i < result.length; ++i) {
+      result[i] = Math.round(a[i]);
+    }
+    return result;
+  } else {
+    throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+var version6 = "4.0.0";
+function forceHalfFloat() {
+  env().set("WEBGL_FORCE_F16_TEXTURES", true);
+}
+if (device_util_exports.isBrowser()) {
+  registerBackend("webgl", () => new MathBackendWebGL(), 2);
+}
+var webgl = { forceHalfFloat };
+var CHECK_NAN_SNIPPET2 = `
   if (isnan(a)) return a;
   if (isnan(b)) return b;
-`,xl=class{constructor(e,t,n){this.variableNames=["A","B"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.enableShapeUniforms=_n(this.outputShape.length),this.userCode=`
+`;
+var BinaryOpProgram = class {
+  constructor(op2, aShape, bShape) {
+    this.variableNames = ["A", "B"];
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       float binaryOperation(float a, float b) {
-        ${e}
+        ${op2}
       }
 
       void main() {
@@ -1250,44 +52286,80 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float b = getBAtOutCoords();
         setOutput(binaryOperation(a, b));
       }
-    `}},nd=`
+    `;
+  }
+};
+var CHECK_NAN_SNIPPET_PACKED = `
   result.r = isNaN.r ? NAN : result.r;
   result.g = isNaN.g ? NAN : result.g;
   result.b = isNaN.b ? NAN : result.b;
   result.a = isNaN.a ? NAN : result.a;
-`,ad=class{constructor(e,t,n,a=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=N.assertAndGetBroadcastShape(t,n);let r=this.outputShape.length;this.enableShapeUniforms=_n(r);let s="";if(a)if(r===0||v.sizeFromShape(this.outputShape)===1)s=`
+`;
+var BinaryOpPackedProgram = class {
+  constructor(op2, aShape, bShape, checkOutOfBounds = false) {
+    this.variableNames = ["A", "B"];
+    this.supportsBroadcasting = true;
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const rank = this.outputShape.length;
+    this.enableShapeUniforms = useShapeUniforms(rank);
+    let checkOutOfBoundsString = "";
+    if (checkOutOfBounds) {
+      if (rank === 0 || util_exports.sizeFromShape(this.outputShape) === 1) {
+        checkOutOfBoundsString = `
           result.y = 0.;
           result.z = 0.;
           result.w = 0.;
-        `;else if(s=`
-          ${gt(r)} coords = getOutputCoords();
-        `,r===1)this.enableShapeUniforms?s+=`
+        `;
+      } else {
+        const dtype = getCoordsDataType(rank);
+        checkOutOfBoundsString = `
+          ${dtype} coords = getOutputCoords();
+        `;
+        if (rank === 1) {
+          if (this.enableShapeUniforms) {
+            checkOutOfBoundsString += `
             result.y = (coords + 1) >= outShape ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `:s+=`
+          `;
+          } else {
+            checkOutOfBoundsString += `
             result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `;else{let i=wn("coords",r);this.enableShapeUniforms?s+=`
+          `;
+          }
+        } else {
+          const channels = getChannels("coords", rank);
+          if (this.enableShapeUniforms) {
+            checkOutOfBoundsString += `
             bool nextRowOutOfBounds =
-              (${i[r-2]} + 1) >= outShape[${r} - 2];
+              (${channels[rank - 2]} + 1) >= outShape[${rank} - 2];
             bool nextColOutOfBounds =
-              (${i[r-1]} + 1) >= outShape[${r} - 1];
+              (${channels[rank - 1]} + 1) >= outShape[${rank} - 1];
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
-          `:s+=`
+          `;
+          } else {
+            checkOutOfBoundsString += `
             bool nextRowOutOfBounds =
-              (${i[r-2]} + 1) >= ${this.outputShape[r-2]};
+              (${channels[rank - 2]} + 1) >= ${this.outputShape[rank - 2]};
             bool nextColOutOfBounds =
-              (${i[r-1]} + 1) >= ${this.outputShape[r-1]};
+              (${channels[rank - 1]} + 1) >= ${this.outputShape[rank - 1]};
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
-          `}this.userCode=`
+          `;
+          }
+        }
+      }
+    }
+    this.userCode = `
       vec4 binaryOperation(vec4 a, vec4 b) {
-        ${e}
+        ${op2}
       }
 
       void main() {
@@ -1295,41 +52367,253 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         vec4 b = getBAtOutCoords();
 
         vec4 result = binaryOperation(a, b);
-        ${s}
+        ${checkOutOfBoundsString}
 
         setOutput(result);
       }
-    `}};function na(e){let{inputs:t,backend:n}=e,{x:a}=t;return n.incRef(a.dataId),{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}var iJ={kernelName:Di,backendName:"webgl",kernelFunc:na};function Is(e){let{inputs:t,backend:n}=e,{real:a,imag:r}=t,s=n.makeTensorInfo(a.shape,"complex64"),i=n.texData.get(s.dataId),o=na({inputs:{x:a},backend:n}),l=na({inputs:{x:r},backend:n});return i.complexTensorInfos={real:o,imag:l},s}var oJ={kernelName:Qh,backendName:"webgl",kernelFunc:Is},U_="return (a < 0.) ? b * a : a;",G_=`
+    `;
+  }
+};
+function identity3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  backend2.incRef(x.dataId);
+  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };
+}
+var identityConfig2 = {
+  kernelName: Identity,
+  backendName: "webgl",
+  kernelFunc: identity3
+};
+function complex3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { real: real4, imag: imag4 } = inputs;
+  const complexInfo = backend2.makeTensorInfo(real4.shape, "complex64");
+  const complex4 = backend2.texData.get(complexInfo.dataId);
+  const realTensorInfo = identity3({ inputs: { x: real4 }, backend: backend2 });
+  const imagTensorInfo = identity3({ inputs: { x: imag4 }, backend: backend2 });
+  complex4.complexTensorInfos = { real: realTensorInfo, imag: imagTensorInfo };
+  return complexInfo;
+}
+var complexConfig2 = {
+  kernelName: Complex,
+  backendName: "webgl",
+  kernelFunc: complex3
+};
+var LEAKYRELU = `return (a < 0.) ? b * a : a;`;
+var LEAKYRELU_PACKED = `
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function lJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{alpha:s}=a,i=n.makeTensorInfo([],"float32",v.createScalarValue(s,"float32")),o=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ad(G_,r.shape,i.shape):new xl(U_,r.shape,i.shape),l=n.runWebGLProgram(o,[r,i],"float32");return n.disposeIntermediateTensorInfo(i),l}var uJ={kernelName:Ri,backendName:"webgl",kernelFunc:lJ},H_="return (a < 0.) ? b * a : a;",j_=`
+`;
+function leakyRelu3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { alpha } = attrs;
+  const $alpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(alpha, "float32"));
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(LEAKYRELU_PACKED, x.shape, $alpha.shape) : new BinaryOpProgram(LEAKYRELU, x.shape, $alpha.shape);
+  const result = backend2.runWebGLProgram(program, [x, $alpha], "float32");
+  backend2.disposeIntermediateTensorInfo($alpha);
+  return result;
+}
+var leakyReluConfig2 = {
+  kernelName: LeakyRelu,
+  backendName: "webgl",
+  kernelFunc: leakyRelu3
+};
+var PRELU = `return (a < 0.) ? b * a : a;`;
+var PRELU_PACKED = `
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function pJ(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t,s=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ad(j_,a.shape,r.shape):new xl(H_,a.shape,r.shape);return n.runWebGLProgram(s,[a,r],"float32")}var cJ={kernelName:qi,backendName:"webgl",kernelFunc:pJ},Uu="if (isnan(x)) return x;";function Ye({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:n,dtype:a}){return({inputs:r,backend:s})=>{let{x:i}=r,o=s,l=a||i.dtype;if(o.shouldExecuteOnCPU([i])&&n!=null){let d=o.texData.get(i.dataId),c=n(d.values,l);return o.makeTensorInfo(i.shape,l,c)}let u=H().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&t!=null,p;return u?p=new js(i.shape,t):p=new Sr(i.shape,e),o.runWebGLProgram(p,[i],l)}}function pn({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:n=!1,supportsComplex:a=!1,cpuKernelImpl:r,dtype:s}){return({inputs:i,backend:o})=>{let{a:l,b:u}=i,p=o;if(a&&l.dtype==="complex64"){let m=p.texData.get(l.dataId),f=p.texData.get(u.dataId),[g,y]=[[m.complexTensorInfos.real,f.complexTensorInfos.real],[m.complexTensorInfos.imag,f.complexTensorInfos.imag]].map(x=>{let[w,I]=x,T={dataId:w.dataId,dtype:w.dtype,shape:l.shape},C={dataId:I.dataId,dtype:I.dtype,shape:u.shape},E=new xl(e,l.shape,u.shape);return p.runWebGLProgram(E,[T,C],ma(w.dtype,I.dtype))}),b=Is({inputs:{real:g,imag:y},backend:p});return p.disposeIntermediateTensorInfo(g),p.disposeIntermediateTensorInfo(y),b}let d=s||ma(l.dtype,u.dtype);if((l.dtype==="string"||u.dtype==="string"||p.shouldExecuteOnCPU([l,u]))&&r!=null){let m=p.texData.get(l.dataId).values,f=p.texData.get(u.dataId).values,g=l.dtype==="string"?N.fromUint8ToStringArray(m):m,y=l.dtype==="string"?N.fromUint8ToStringArray(f):f,[b,x]=r(l.shape,u.shape,g,y,d),w=p.makeTensorInfo(x,d),I=p.texData.get(w.dataId);return I.values=b,w}let c=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&t!=null,h;return c?h=new ad(t,l.shape,u.shape,n):h=new xl(e,l.shape,u.shape),p.runWebGLProgram(h,[l,u],d)}}function ec(e,t=!1){if(e==="linear")return t?G7:z7;if(e==="relu")return t?j7:B7;if(e==="elu")return t?H7:W7;if(e==="relu6")return t?q7:V7;if(e==="prelu")return t?j_:H_;if(e==="leakyrelu")return t?G_:U_;if(e==="sigmoid")return t?K7:U7;throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var q_=class{constructor(e,t,n,a=!1,r=!1,s=!1,i=null,o=!1,l=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=_n(this.outputShape.length);let u=a?e[1]:e[2],p=Math.ceil(u/2),d=a?"i * 2, rc.y":"rc.y, i * 2",c=r?"rc.z, i * 2":"i * 2, rc.z",h=a?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],m=r?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],f="",g="";i&&(o?f=`vec4 activation(vec4 a) {
+`;
+function prelu4(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(PRELU_PACKED, x.shape, alpha.shape) : new BinaryOpProgram(PRELU, x.shape, alpha.shape);
+  return backend2.runWebGLProgram(program, [x, alpha], "float32");
+}
+var preluConfig2 = {
+  kernelName: Prelu,
+  backendName: "webgl",
+  kernelFunc: prelu4
+};
+var CHECK_NAN_SNIPPET_UNARY = `if (isnan(x)) return x;`;
+function unaryKernelFunc2({ opSnippet, packedOpSnippet, cpuKernelImpl, dtype }) {
+  return ({ inputs, backend: backend2 }) => {
+    const { x } = inputs;
+    const webglBackend = backend2;
+    const $dtype = dtype || x.dtype;
+    if (webglBackend.shouldExecuteOnCPU([x]) && cpuKernelImpl != null) {
+      const xData = webglBackend.texData.get(x.dataId);
+      const outValues = cpuKernelImpl(xData.values, $dtype);
+      return webglBackend.makeTensorInfo(x.shape, $dtype, outValues);
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK_UNARY_OPERATIONS") && packedOpSnippet != null;
+    let program;
+    if (shouldUsePackedProgram) {
+      program = new UnaryOpPackedProgram(x.shape, packedOpSnippet);
+    } else {
+      program = new UnaryOpProgram(x.shape, opSnippet);
+    }
+    return webglBackend.runWebGLProgram(program, [x], $dtype);
+  };
+}
+function binaryKernelFunc2({ opSnippet, packedOpSnippet, checkOutOfBounds = false, supportsComplex = false, cpuKernelImpl, dtype }) {
+  return ({ inputs, backend: backend2 }) => {
+    const { a, b } = inputs;
+    const webglBackend = backend2;
+    if (supportsComplex && a.dtype === "complex64") {
+      const aData = webglBackend.texData.get(a.dataId);
+      const bData = webglBackend.texData.get(b.dataId);
+      const [real4, imag4] = [
+        [aData.complexTensorInfos.real, bData.complexTensorInfos.real],
+        [aData.complexTensorInfos.imag, bData.complexTensorInfos.imag]
+      ].map((complexParts) => {
+        const [aPart, bPart] = complexParts;
+        const aHandle = {
+          dataId: aPart.dataId,
+          dtype: aPart.dtype,
+          shape: a.shape
+        };
+        const bHandle = {
+          dataId: bPart.dataId,
+          dtype: bPart.dtype,
+          shape: b.shape
+        };
+        const program2 = new BinaryOpProgram(opSnippet, a.shape, b.shape);
+        return webglBackend.runWebGLProgram(program2, [aHandle, bHandle], upcastType(aPart.dtype, bPart.dtype));
+      });
+      const complexOutput = complex3({ inputs: { real: real4, imag: imag4 }, backend: webglBackend });
+      webglBackend.disposeIntermediateTensorInfo(real4);
+      webglBackend.disposeIntermediateTensorInfo(imag4);
+      return complexOutput;
+    }
+    const $dtype = dtype || upcastType(a.dtype, b.dtype);
+    if ((a.dtype === "string" || b.dtype === "string" || webglBackend.shouldExecuteOnCPU([a, b])) && cpuKernelImpl != null) {
+      const aVals = webglBackend.texData.get(a.dataId).values;
+      const bVals = webglBackend.texData.get(b.dataId).values;
+      const decodedAVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;
+      const decodedBVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;
+      const [outValues, outShape] = cpuKernelImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);
+      const out = webglBackend.makeTensorInfo(outShape, $dtype);
+      const outData = webglBackend.texData.get(out.dataId);
+      outData.values = outValues;
+      return out;
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") && packedOpSnippet != null;
+    let program;
+    if (shouldUsePackedProgram) {
+      program = new BinaryOpPackedProgram(packedOpSnippet, a.shape, b.shape, checkOutOfBounds);
+    } else {
+      program = new BinaryOpProgram(opSnippet, a.shape, b.shape);
+    }
+    return webglBackend.runWebGLProgram(program, [a, b], $dtype);
+  };
+}
+function mapActivationToShaderProgram(activation2, packed = false) {
+  if (activation2 === "linear") {
+    if (packed) {
+      return LINEAR2;
+    }
+    return LINEAR;
+  } else if (activation2 === "relu") {
+    if (packed) {
+      return RELU2;
+    }
+    return RELU;
+  } else if (activation2 === "elu") {
+    if (packed) {
+      return ELU3;
+    }
+    return ELU2;
+  } else if (activation2 === "relu6") {
+    if (packed) {
+      return RELU62;
+    }
+    return RELU6;
+  } else if (activation2 === "prelu") {
+    if (packed) {
+      return PRELU_PACKED;
+    }
+    return PRELU;
+  } else if (activation2 === "leakyrelu") {
+    if (packed) {
+      return LEAKYRELU_PACKED;
+    }
+    return LEAKYRELU;
+  } else if (activation2 === "sigmoid") {
+    if (packed) {
+      return SIGMOID2;
+    }
+    return SIGMOID;
+  }
+  throw new Error(`Activation ${activation2} has not been implemented for the WebGL backend.`);
+}
+var MatMulPackedProgram = class {
+  constructor(aShape, bShape, outputShape, transposeA = false, transposeB = false, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyreluActivation = false) {
+    this.variableNames = ["matrixA", "matrixB"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const sharedDim = transposeA ? aShape[1] : aShape[2];
+    const sharedDimensionPacked = Math.ceil(sharedDim / 2);
+    const aSample = transposeA ? "i * 2, rc.y" : "rc.y, i * 2";
+    const bSample = transposeB ? "rc.z, i * 2" : "i * 2, rc.z";
+    const aSwizzle = transposeA ? ["a.xxyy", "a.zzww"] : ["a.xxzz", "a.yyww"];
+    const bSwizzle = transposeB ? ["b.xzxz", "b.ywyw"] : ["b.xyxy", "b.zwzw"];
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
-          ${i}
-        }`:l?f=`vec4 activation(vec4 a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
-          ${i}
-        }`:f=`vec4 activation(vec4 x) {
-          ${i}
-        }`,g="result = activation(result);");let y=s?"result += getBiasAtOutCoords();":"";s&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),l&&this.variableNames.push("leakyreluAlpha");let b="rc.x",x="rc.x";e[0]<t[0]?b=`int(min(float(rc.x), ${e[0]-1}.))`:t[0]<e[0]&&(x=`int(min(float(rc.x), ${t[0]-1}.))`),this.userCode=`
-      ${f}
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+          ${activation2}
+        }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyreluActivation) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    let batchASnippet = "rc.x";
+    let batchBSnippet = "rc.x";
+    if (aShape[0] < bShape[0]) {
+      batchASnippet = `int(min(float(rc.x), ${aShape[0] - 1}.))`;
+    } else if (bShape[0] < aShape[0]) {
+      batchBSnippet = `int(min(float(rc.x), ${bShape[0] - 1}.))`;
+    }
+    this.userCode = `
+      ${activationSnippet}
       // Don't use uniform for sharedDimensionPacked for performance.
-      const float sharedDimension = ${p}.0;
+      const float sharedDimension = ${sharedDimensionPacked}.0;
 
       vec4 dot2x2ARowBCol(ivec3 rc) {
         vec4 result = vec4(0);
-        for (int i = 0; i < ${p}; i++) {
-          int batchA = ${b};
-          int batchB = ${x};
-          vec4 a = getMatrixA(batchA, ${d});
-          vec4 b = getMatrixB(batchB, ${c});
+        for (int i = 0; i < ${sharedDimensionPacked}; i++) {
+          int batchA = ${batchASnippet};
+          int batchB = ${batchBSnippet};
+          vec4 a = getMatrixA(batchA, ${aSample});
+          vec4 b = getMatrixB(batchB, ${bSample});
 
           // These swizzled products need to be separately added.
           // See: https://github.com/tensorflow/tfjs/issues/1735
-          result += (${h[0]} * ${m[0]});
-          result += (${h[1]} * ${m[1]});
+          result += (${aSwizzle[0]} * ${bSwizzle[0]});
+          result += (${aSwizzle[1]} * ${bSwizzle[1]});
         }
         return result;
       }
@@ -1338,16 +52622,27 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec3 rc = getOutputCoords();
         vec4 result = dot2x2ARowBCol(rc);
 
-        ${y}
+        ${addBiasSnippet}
 
-        ${g}
+        ${applyActivationSnippet}
 
         setOutput(result);
       }
-    `}},Zk={REAL:"return areal * breal - aimag * bimag;",IMAG:"return areal * bimag + aimag * breal;"},Jk=class{constructor(e,t,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.userCode=`
+    `;
+  }
+};
+var COMPLEX_MULTIPLY = {
+  REAL: "return areal * breal - aimag * bimag;",
+  IMAG: "return areal * bimag + aimag * breal;"
+};
+var BinaryOpComplexProgram = class {
+  constructor(op2, aShape, bShape) {
+    this.variableNames = ["AReal", "AImag", "BReal", "BImag"];
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    this.userCode = `
       float binaryOpComplex(
           float areal, float aimag, float breal, float bimag) {
-        ${e}
+        ${op2}
       }
 
       void main() {
@@ -1357,15 +52652,136 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float bimag = getBImagAtOutCoords();
         setOutput(binaryOpComplex(areal, aimag, breal, bimag));
       }
-    `}},Qk="return a * b;";function Q0(e){let{inputs:t,backend:n}=e,{a,b:r}=t,s=N.upcastType(a.dtype,r.dtype);if(a.dtype==="complex64"){let o=n.texData.get(a.dataId),l=n.texData.get(r.dataId),u=new Jk(Zk.REAL,a.shape,r.shape),p=new Jk(Zk.IMAG,a.shape,r.shape),d=[{dataId:o.complexTensorInfos.real.dataId,dtype:o.complexTensorInfos.real.dtype,shape:a.shape},{dataId:o.complexTensorInfos.imag.dataId,dtype:o.complexTensorInfos.imag.dtype,shape:a.shape},{dataId:l.complexTensorInfos.real.dataId,dtype:l.complexTensorInfos.real.dtype,shape:r.shape},{dataId:l.complexTensorInfos.imag.dataId,dtype:l.complexTensorInfos.imag.dtype,shape:r.shape}],c=n.runWebGLProgram(u,d,"float32"),h=n.runWebGLProgram(p,d,"float32"),m=Is({inputs:{real:c,imag:h},backend:n});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),m}if(n.shouldExecuteOnCPU([a,r])){let o=n.texData.get(a.dataId),l=n.texData.get(r.dataId),[u,p]=u7(a.shape,r.shape,o.values,l.values,s),d=n.makeTensorInfo(p,s),c=n.texData.get(d.dataId);return c.values=u,d}let i;return H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?i=new ad(Qk,a.shape,r.shape):i=new xl(Qk,a.shape,r.shape),n.runWebGLProgram(i,[a,r],s)}var dJ={kernelName:Ui,backendName:"webgl",kernelFunc:Q0};function hJ(e,t,n){let a=[pi(e.shape),...ci(e.shape)],r={dtype:e.dtype,shape:a,dataId:e.dataId},s=[pi(t),...ci(t)],i=new B_(s,a),o=!0,l=[a],u=n.runWebGLProgram(i,[r],e.dtype,l,o);return{dataId:u.dataId,shape:t,dtype:u.dtype}}function de(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{shape:s}=a,i=n,o=v.sizeFromShape(r.shape),l=v.inferFromImplicitShape(s,o),u=v.sizeFromShape(l);v.assert(o===u,()=>`The new shape (${l}) has ${u} elements and the old shape (${r.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);let p=i.texData.get(r.dataId);return p.isPacked&&!Qp(r.shape,l)&&!(p.texture!==null&&Qp(p.shape,l))?hJ(r,l,i):(i.incRef(r.dataId),{dataId:r.dataId,shape:l,dtype:r.dtype})}var mJ={kernelName:lu,backendName:"webgl",kernelFunc:de},eI=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:a,inSize:r,outSize:s}=e;this.outputShape=[a,s];let i=Math.floor(n/4)*4,o=n%4,l="sumValue += dot(values, ones);";if(t!=null){let p=1/t;l=`sumValue += dot(values * ${v.isInt(p)?p.toPrecision(2):p}, ones);`}let u="";r%n>0&&(u=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+    `;
+  }
+};
+var MUL = "return a * b;";
+function multiply3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { a, b } = inputs;
+  const dtype = backend_util_exports.upcastType(a.dtype, b.dtype);
+  if (a.dtype === "complex64") {
+    const aData = backend2.texData.get(a.dataId);
+    const bData = backend2.texData.get(b.dataId);
+    const realProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.REAL, a.shape, b.shape);
+    const imagProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.IMAG, a.shape, b.shape);
+    const inputs2 = [
+      {
+        dataId: aData.complexTensorInfos.real.dataId,
+        dtype: aData.complexTensorInfos.real.dtype,
+        shape: a.shape
+      },
+      {
+        dataId: aData.complexTensorInfos.imag.dataId,
+        dtype: aData.complexTensorInfos.imag.dtype,
+        shape: a.shape
+      },
+      {
+        dataId: bData.complexTensorInfos.real.dataId,
+        dtype: bData.complexTensorInfos.real.dtype,
+        shape: b.shape
+      },
+      {
+        dataId: bData.complexTensorInfos.imag.dataId,
+        dtype: bData.complexTensorInfos.imag.dtype,
+        shape: b.shape
+      }
+    ];
+    const realPart = backend2.runWebGLProgram(realProgram, inputs2, "float32");
+    const imagPart = backend2.runWebGLProgram(imagProgram, inputs2, "float32");
+    const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    return complexOutput;
+  }
+  if (backend2.shouldExecuteOnCPU([a, b])) {
+    const aData = backend2.texData.get(a.dataId);
+    const bData = backend2.texData.get(b.dataId);
+    const [outValues, outShape] = multiplyImplCPU(a.shape, b.shape, aData.values, bData.values, dtype);
+    const out = backend2.makeTensorInfo(outShape, dtype);
+    const outData = backend2.texData.get(out.dataId);
+    outData.values = outValues;
+    return out;
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_BINARY_OPERATIONS")) {
+    program = new BinaryOpPackedProgram(MUL, a.shape, b.shape);
+  } else {
+    program = new BinaryOpProgram(MUL, a.shape, b.shape);
+  }
+  return backend2.runWebGLProgram(program, [a, b], dtype);
+}
+var multiplyConfig2 = {
+  kernelName: Multiply,
+  backendName: "webgl",
+  kernelFunc: multiply3
+};
+function packedReshape(input2, afterShape, backend2) {
+  const input3DShape = [
+    getBatchDim(input2.shape),
+    ...getRowsCols(input2.shape)
+  ];
+  const input3D = {
+    dtype: input2.dtype,
+    shape: input3DShape,
+    dataId: input2.dataId
+  };
+  const afterShapeAs3D = [
+    getBatchDim(afterShape),
+    ...getRowsCols(afterShape)
+  ];
+  const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);
+  const preventEagerUnpackingOfOutput = true;
+  const customValues = [input3DShape];
+  const output = backend2.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);
+  return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };
+}
+function reshape4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const webglBackend = backend2;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  const $xSize = util_exports.sizeFromShape($shape);
+  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);
+  const xTexData = webglBackend.texData.get(x.dataId);
+  if (xTexData.isPacked && !isReshapeFree(x.shape, $shape) && !(xTexData.texture !== null && isReshapeFree(xTexData.shape, $shape))) {
+    return packedReshape(x, $shape, webglBackend);
+  }
+  webglBackend.incRef(x.dataId);
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig2 = {
+  kernelName: Reshape,
+  backendName: "webgl",
+  kernelFunc: reshape4
+};
+var MeanProgram = class {
+  constructor(reduceInfo, divisor) {
+    this.variableNames = ["x"];
+    const { windowSize, batchSize, inSize, outSize } = reduceInfo;
+    this.outputShape = [batchSize, outSize];
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    let updateSnippet = `sumValue += dot(values, ones);`;
+    if (divisor != null) {
+      const denominator = 1 / divisor;
+      updateSnippet = `sumValue += dot(values * ${util_exports.isInt(denominator) ? denominator.toPrecision(2) : denominator}, ones);`;
+    }
+    let checkOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return 0.0;
         }
-      `),this.userCode=`
+      `;
+    }
+    this.userCode = `
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float getValue(int batch, int inIdx) {
-        ${u}
+        ${checkOutOfBounds}
         return getX(batch, inIdx);
       }
 
@@ -1373,11 +52789,11 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${n};
+        int inOffset = outIdx * ${windowSize};
 
         float sumValue = 0.0;
 
-        for (int i = 0; i < ${i}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
           vec4 values = vec4(
             getValue(batch, inIdx),
@@ -1386,64 +52802,110 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             getValue(batch, inIdx + 3)
           );
 
-          ${l}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${i};
-        if (${o===1}) {
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
           vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);
 
-          ${l}
-        } else if (${o===2}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1), 0.0, 0.0);
 
-          ${l}
-        } else if (${o===3}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2), 0.0);
 
-          ${l}
+          ${updateSnippet}
         }
         setOutput(sumValue);
       }
-    `}},fJ=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:a,inSize:r,outSize:s}=e;this.outputShape=[a,s];let i="0.0",o="";t==="prod"?i="1.0":t==="min"?(i="1.0 / 1e-20",o="min"):t==="max"&&(i="-1.0 / 1e-20",o="max");let l=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="sum"?l="sumValue":t==="prod"?l="prodValue":t==="all"?l="allValue":t==="any"&&(l="anyValue");let u=Math.floor(n/4)*4,p=n%4,d=`
-      if (${t==="sum"}) {
+    `;
+  }
+};
+var ReduceProgram = class {
+  constructor(reduceInfo, reduceType) {
+    this.variableNames = ["x"];
+    const { windowSize, batchSize, inSize, outSize } = reduceInfo;
+    this.outputShape = [batchSize, outSize];
+    let initializationValue = "0.0";
+    let compareOp = ``;
+    if (reduceType === "prod") {
+      initializationValue = "1.0";
+    } else if (reduceType === "min") {
+      initializationValue = "1.0 / 1e-20";
+      compareOp = `min`;
+    } else if (reduceType === "max") {
+      initializationValue = "-1.0 / 1e-20";
+      compareOp = `max`;
+    }
+    let returnValue = `${reduceType}(${reduceType}(${reduceType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (reduceType === "sum") {
+      returnValue = `sumValue`;
+    } else if (reduceType === "prod") {
+      returnValue = `prodValue`;
+    } else if (reduceType === "all") {
+      returnValue = `allValue`;
+    } else if (reduceType === "any") {
+      returnValue = `anyValue`;
+    }
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    let updateSnippet = `
+      if (${reduceType === "sum"}) {
         sumValue += dot(values, ones);
-      } else if (${t==="prod"}) {
+      } else if (${reduceType === "prod"}) {
         vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);
         prodValue *= tmp[0] * tmp[1];
       } else {
-        minMaxValue = ${o}(values, minMaxValue);
-        if (${t==="min"} || ${t==="max"}) {
-          minMaxValue = ${o}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
+        if (${reduceType === "min"} || ${reduceType === "max"}) {
+          minMaxValue = ${compareOp}(values, minMaxValue);
           bvec4 isNaN = isnan(values);
           if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {
             minMaxValue = vec4(NAN);
           }
         }
       }
-    `,c="vec4";t==="all"?(i="1.0",d=`
+    `;
+    let vecType = `vec4`;
+    if (reduceType === "all") {
+      initializationValue = "1.0";
+      updateSnippet = `
         bool reducedAllValue = all(values);
         float floatedReducedAllValue = float(reducedAllValue);
         allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);
-      `,c="bvec4"):t==="any"&&(i="0.0",d=`
+      `;
+      vecType = `bvec4`;
+    } else if (reduceType === "any") {
+      initializationValue = "0.0";
+      updateSnippet = `
         bool reducedAnyValue = any(values);
         float floatedReducedAnyValue = float(reducedAnyValue);
         anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);
-      `,c="bvec4");let h="";r%n>0&&(h=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+      `;
+      vecType = `bvec4`;
+    }
+    let checkOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return initializationValue;
         }
-      `),this.userCode=`
-      const float initializationValue = ${i};
+      `;
+    }
+    this.userCode = `
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float getValue(int batch, int inIdx) {
-        ${h}
+        ${checkOutOfBounds}
         return getX(batch, inIdx);
       }
 
@@ -1451,174 +52913,673 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${n};
+        int inOffset = outIdx * ${windowSize};
 
-        vec4 minMaxValue = vec4(${i});
+        vec4 minMaxValue = vec4(${initializationValue});
         float prodValue = 1.0;
         float sumValue = 0.0;
         float allValue = 1.0;
         float anyValue = 0.0;
 
-        for (int i = 0; i < ${u}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
-          ${c} values = ${c}(
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             getValue(batch, inIdx + 3)
           );
 
-          ${d}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${u};
-        if (${p===1}) {
-          ${c} values = ${c}(
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             initializationValue,
             initializationValue,
             initializationValue
           );
 
-          ${d}
-        } else if (${p===2}) {
-          ${c} values = ${c}(
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             initializationValue,
             initializationValue
           );
 
-          ${d}
-        } else if (${p===3}) {
-          ${c} values = ${c}(
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             initializationValue
           );
 
-          ${d}
+          ${updateSnippet}
         }
-        setOutput(${l});
+        setOutput(${returnValue});
       }
-    `}};function gJ(e){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let n=t.length?t[t.length-1].outSize:e[1],a=N.computeOptimalWindowSize(n);t.push({inSize:n,windowSize:a,outSize:Math.ceil(n/a)})}return t}function ko(e,t,n,a){let r=gJ(e.shape),s=e;for(let i=0;i<r.length;i++){let{inSize:o,windowSize:l,outSize:u}=r[i],p,d;n==="mean"?p=i===0?new eI({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},o):new eI({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u}):p=new fJ({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},n),d=s,s=a.runWebGLProgram(p,[s],t),d.dataId!==e.dataId&&a.disposeIntermediateTensorInfo(d)}return s}var yJ=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let s=0;s<n.length;s++)n[s]=e[t[s]];this.outputShape=n,this.rank=n.length;let a=gt(this.rank),r=bJ(t);this.userCode=`
-    void main() {
-      ${a} resRC = getOutputCoords();
-      setOutput(getA(${r}));
+    `;
+  }
+};
+function getReductionStages(inShape) {
+  const stages = [];
+  while (stages.length === 0 || stages[stages.length - 1].outSize !== 1) {
+    const outSize = stages.length ? stages[stages.length - 1].outSize : inShape[1];
+    const windowSize = backend_util_exports.computeOptimalWindowSize(outSize);
+    stages.push({
+      inSize: outSize,
+      windowSize,
+      outSize: Math.ceil(outSize / windowSize)
+    });
+  }
+  return stages;
+}
+function reduce(x, dtype, reductionType, backend2) {
+  const reductionStages = getReductionStages(x.shape);
+  let result = x;
+  for (let i = 0; i < reductionStages.length; i++) {
+    const { inSize, windowSize, outSize } = reductionStages[i];
+    let program;
+    let previousResult;
+    if (reductionType === "mean") {
+      program = i === 0 ? new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, inSize) : new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize });
+    } else {
+      program = new ReduceProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, reductionType);
     }
-    `}};function bJ(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],a=new Array(t);for(let r=0;r<e.length;r++)a[e[r]]=n[r];return a.join()}var xJ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(e.length);for(let u=0;u<n.length;u++)n[u]=e[t[u]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let a=gt(this.rank),r=W_("rc",this.rank),s=new Array(this.rank);for(let u=0;u<t.length;u++)s[t[u]]=r[u];let i=`vec2(${s.slice(-2).join()})`,o=`++${r[this.rank-1]} < ${n[this.rank-1]}`,l=`getChannel(getA(${s.join()}), ${i})`;this.userCode=`
+    previousResult = result;
+    result = backend2.runWebGLProgram(program, [result], dtype);
+    if (previousResult.dataId !== x.dataId) {
+      backend2.disposeIntermediateTensorInfo(previousResult);
+    }
+  }
+  return result;
+}
+var TransposeProgram = class {
+  constructor(aShape, newDim) {
+    this.variableNames = ["A"];
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[newDim[i]];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const switched = getSwitchedCoords(newDim);
+    this.userCode = `
     void main() {
-      ${a} rc = getOutputCoords();
+      ${dtype} resRC = getOutputCoords();
+      setOutput(getA(${switched}));
+    }
+    `;
+  }
+};
+function getSwitchedCoords(newDim) {
+  const rank = newDim.length;
+  if (rank > 6) {
+    throw Error(`Transpose for rank ${rank} is not yet supported`);
+  }
+  const originalOrder = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u", "resRC.v"];
+  const switchedCoords = new Array(rank);
+  for (let i = 0; i < newDim.length; i++) {
+    switchedCoords[newDim[i]] = originalOrder[i];
+  }
+  return switchedCoords.join();
+}
+var TransposePackedProgram = class {
+  constructor(aShape, newDim) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[newDim[i]];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    if (this.rank > 6) {
+      throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);
+    }
+    const dtype = getCoordsDataType(this.rank);
+    const outputOrder = getVecChannels("rc", this.rank);
+    const switchedOrder = new Array(this.rank);
+    for (let i = 0; i < newDim.length; i++) {
+      switchedOrder[newDim[i]] = outputOrder[i];
+    }
+    const innerDims = `vec2(${switchedOrder.slice(-2).join()})`;
+    const nextColumn = `++${outputOrder[this.rank - 1]} < ${outputShape[this.rank - 1]}`;
+    const getc = `getChannel(getA(${switchedOrder.join()}), ${innerDims})`;
+    this.userCode = `
+    void main() {
+      ${dtype} rc = getOutputCoords();
       vec4 result = vec4(0.);
-      result[0] = ${l};
-      if(${o}) {
-        result[1] = ${l};
+      result[0] = ${getc};
+      if(${nextColumn}) {
+        result[1] = ${getc};
       }
-      --${r[this.rank-1]};
-      if(++${r[this.rank-2]} < ${n[this.rank-2]}) {
-        result[2] = ${l};
-        if(${o}) {
-          result[3] = ${l};
+      --${outputOrder[this.rank - 1]};
+      if(++${outputOrder[this.rank - 2]} < ${outputShape[this.rank - 2]}) {
+        result[2] = ${getc};
+        if(${nextColumn}) {
+          result[3] = ${getc};
         }
       }
       setOutput(result);
     }
-    `}};function Mf(e,t,n){let a=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new xJ(e.shape,t):new yJ(e.shape,t);return n.runWebGLProgram(a,[e],e.dtype)}function vJ(e,t,n,a){let r=t,s=e.shape.length,i=v.parseAxisParam(r,e.shape),o=i,l=N.getAxesPermutation(o,s),u=l!=null,p=e;u&&(p=Mf(e,l,a),o=N.getInnerMostAxes(o.length,s)),N.assertAxesAreInnerMostDims("sum",o,s);let[d,c]=N.computeOutAndReduceShapes(p.shape,o),h=d;n&&(h=N.expandShapeToKeepDim(d,i));let m=v.sizeFromShape(c),f=v.sizeFromShape(e.shape)/m,g=de({inputs:{x:p},attrs:{shape:[f,m]},backend:a}),y=_m(e.dtype),b=ko(g,y,"sum",a),x=de({inputs:{x:b},attrs:{shape:h},backend:a});return a.disposeIntermediateTensorInfo(g),a.disposeIntermediateTensorInfo(b),u&&a.disposeIntermediateTensorInfo(p),x}function Pf(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;return vJ(r,s,i,n)}var wJ={kernelName:so,backendName:"webgl",kernelFunc:Pf};function In(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{perm:s}=a,i=n,o=r.shape.length,l=new Array(o);for(let p=0;p<l.length;p++)l[p]=r.shape[s[p]];let u;if(i.shouldExecuteOnCPU([r])){let p=i.texData.get(r.dataId).values,d=Z0(p,r.shape,r.dtype,s,l);u=i.makeTensorInfo(l,r.dtype);let c=i.texData.get(u.dataId);c.values=d}else u=Mf(r,s,i);return u}var kJ={kernelName:Tr,backendName:"webgl",kernelFunc:In},K_=1e3;function Bh({a:e,b:t,transposeA:n,transposeB:a,backend:r,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:l=null}){let u=e.shape.length,p=t.shape.length,d=n?e.shape[u-2]:e.shape[u-1],c=a?t.shape[p-1]:t.shape[p-2],h=n?e.shape[u-1]:e.shape[u-2],m=a?t.shape[p-2]:t.shape[p-1],f=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=v.sizeFromShape(f),b=v.sizeFromShape(g),x=Tu.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([h,m]);v.assert(d===c,()=>`Error in matMul: inner shapes (${d}) and (${c}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${a} must match.`);let w=n?[y,d,h]:[y,h,d],I=a?[b,m,c]:[b,c,m],T=de({inputs:{x:e},backend:r,attrs:{shape:w}}),C=de({inputs:{x:t},backend:r,attrs:{shape:I}}),E=[T,C],A=Math.max(y,b),R=n?T.shape[1]:T.shape[2],F=s!=null,S=i!=null,M=l==="leakyrelu",B=l!=null?ec(l,!0):null,U=F||S||M||B!=null,G;if((h===1||m===1)&&R>K_&&U===!1){let K=T,Z=C;n&&(K=In({inputs:{x:T},backend:r,attrs:{perm:[0,2,1]}}),E.push(K)),a&&(Z=In({inputs:{x:C},backend:r,attrs:{perm:[0,2,1]}}),E.push(Z));let Q=m!==1,ee=m===1,ae=K;Q&&(ae=de({inputs:{x:K},backend:r,attrs:{shape:[A,R,1]}}),E.push(ae));let te=m===1?2:1,le=Z;ee&&(le=de({inputs:{x:Z},backend:r,attrs:{shape:[A,1,R]}}),E.push(le));let ie=Q0({inputs:{a:ae,b:le},backend:r});G=Pf({inputs:{x:ie},backend:r,attrs:{axis:te,keepDims:!0}}),E.push(ie)}else{let K=ma(e.dtype,t.dtype),Z=new q_(w,I,[A,h,m],n,a,F,B,S,M),Q=[T,C];if(s!=null&&Q.push(s),S&&Q.push(i),M){let ee=r.makeTensorInfo([],"float32",v.createScalarValue(o,"float32"));Q.push(ee),E.push(ee)}G=r.runWebGLProgram(Z,Q,K)}let q=de({inputs:{x:G},backend:r,attrs:{shape:x}});E.push(G);for(let K of E)r.disposeIntermediateTensorInfo(K);return q}function IJ(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a;return Bh({a:r,b:s,transposeA:l,transposeB:u,backend:n,bias:i,preluActivationWeights:o,leakyreluAlpha:d,activation:p})}var SJ={kernelName:Js,backendName:"webgl",kernelFunc:IJ},tI="return abs(x);";function TJ(e){let{inputs:t,backend:n}=e,{x:a}=t;if(n.shouldExecuteOnCPU([a])&&a.dtype!=="complex64"){let s=n.texData.get(a.dataId),i=L_(s.values);return n.makeTensorInfo(a.shape,a.dtype,i)}let r;return H().getBool("WEBGL_PACK_UNARY_OPERATIONS")?r=new js(a.shape,tI):r=new Sr(a.shape,tI),n.runWebGLProgram(r,[a],a.dtype)}var NJ={kernelName:wl,backendName:"webgl",kernelFunc:TJ},CJ=Da+`
+    `;
+  }
+};
+function transposeImpl2(x, perm, backend2) {
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new TransposePackedProgram(x.shape, perm) : new TransposeProgram(x.shape, perm);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+function sumImpl(x, axis, keepDims, backend2) {
+  const reductionIndices = axis;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  const sumInputIsTransposed = permutedAxes != null;
+  let sumInput = x;
+  if (sumInputIsTransposed) {
+    sumInput = transposeImpl2(x, permutedAxes, backend2);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", axes, xRank);
+  const [sumOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(sumInput.shape, axes);
+  let outShape = sumOutShape;
+  if (keepDims) {
+    outShape = backend_util_exports.expandShapeToKeepDim(sumOutShape, origAxes);
+  }
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x: sumInput }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const outType = sumOutType(x.dtype);
+  const reduced = reduce(reshapedInput, outType, "sum", backend2);
+  const out = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (sumInputIsTransposed) {
+    backend2.disposeIntermediateTensorInfo(sumInput);
+  }
+  return out;
+}
+function sum4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  return sumImpl(x, axis, keepDims, backend2);
+}
+var sumConfig2 = {
+  kernelName: Sum,
+  backendName: "webgl",
+  kernelFunc: sum4
+};
+function transpose3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { perm } = attrs;
+  const webglBackend = backend2;
+  const xRank = x.shape.length;
+  const newShape = new Array(xRank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[perm[i]];
+  }
+  let out;
+  if (webglBackend.shouldExecuteOnCPU([x])) {
+    const xTexData = webglBackend.texData.get(x.dataId);
+    const values = xTexData.values;
+    const outValues = transposeImplCPU(values, x.shape, x.dtype, perm, newShape);
+    out = webglBackend.makeTensorInfo(newShape, x.dtype);
+    const outData = webglBackend.texData.get(out.dataId);
+    outData.values = outValues;
+  } else {
+    out = transposeImpl2(x, perm, webglBackend);
+  }
+  return out;
+}
+var transposeConfig2 = {
+  kernelName: Transpose,
+  backendName: "webgl",
+  kernelFunc: transpose3
+};
+var MATMUL_SHARED_DIM_THRESHOLD = 1e3;
+function batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape4({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape4({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const intermediates = [a3d, b3d];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const fusedActivation = activation2 != null ? mapActivationToShaderProgram(activation2, true) : null;
+  const containsFusedOps = hasBias || hasPreluActivationWeights || hasLeakyreluAlpha || fusedActivation != null;
+  let out;
+  if ((outerShapeA === 1 || outerShapeB === 1) && sharedDim > MATMUL_SHARED_DIM_THRESHOLD && containsFusedOps === false) {
+    let aVec = a3d;
+    let bVec = b3d;
+    if (transposeA) {
+      aVec = transpose3({ inputs: { x: a3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });
+      intermediates.push(aVec);
+    }
+    if (transposeB) {
+      bVec = transpose3({ inputs: { x: b3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });
+      intermediates.push(bVec);
+    }
+    const shouldReshapeA = outerShapeB !== 1;
+    const shouldReshapeB = outerShapeB === 1;
+    let aVec3d = aVec;
+    if (shouldReshapeA) {
+      aVec3d = reshape4({
+        inputs: { x: aVec },
+        backend: backend2,
+        attrs: { shape: [batchDim, sharedDim, 1] }
+      });
+      intermediates.push(aVec3d);
+    }
+    const axis = outerShapeB === 1 ? 2 : 1;
+    let bVec3d = bVec;
+    if (shouldReshapeB) {
+      bVec3d = reshape4({
+        inputs: { x: bVec },
+        backend: backend2,
+        attrs: { shape: [batchDim, 1, sharedDim] }
+      });
+      intermediates.push(bVec3d);
+    }
+    const product = multiply3({ inputs: { a: aVec3d, b: bVec3d }, backend: backend2 });
+    out = sum4({ inputs: { x: product }, backend: backend2, attrs: { axis, keepDims: true } });
+    intermediates.push(product);
+  } else {
+    const dtype = upcastType(a.dtype, b.dtype);
+    const program = new MatMulPackedProgram(a3dShape, b3dShape, [batchDim, outerShapeA, outerShapeB], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const inputs = [a3d, b3d];
+    if (bias != null) {
+      inputs.push(bias);
+    }
+    if (hasPreluActivationWeights) {
+      inputs.push(preluActivationWeights);
+    }
+    if (hasLeakyreluAlpha) {
+      const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+      inputs.push($leakyreluAlpha);
+      intermediates.push($leakyreluAlpha);
+    }
+    out = backend2.runWebGLProgram(program, inputs, dtype);
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: outShape } });
+  intermediates.push(out);
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return outReshaped;
+}
+function _fusedMatMul2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  return batchMatMulImpl({
+    a,
+    b,
+    transposeA,
+    transposeB,
+    backend: backend2,
+    bias,
+    preluActivationWeights,
+    leakyreluAlpha,
+    activation: activation2
+  });
+}
+var _fusedMatMulConfig2 = {
+  kernelName: _FusedMatMul,
+  backendName: "webgl",
+  kernelFunc: _fusedMatMul2
+};
+var ABS2 = `return abs(x);`;
+function abs3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (backend2.shouldExecuteOnCPU([x]) && x.dtype !== "complex64") {
+    const xData = backend2.texData.get(x.dataId);
+    const outValues = simpleAbsImplCPU(xData.values);
+    return backend2.makeTensorInfo(x.shape, x.dtype, outValues);
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+    program = new UnaryOpPackedProgram(x.shape, ABS2);
+  } else {
+    program = new UnaryOpProgram(x.shape, ABS2);
+  }
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var absConfig2 = {
+  kernelName: Abs,
+  backendName: "webgl",
+  kernelFunc: abs3
+};
+var ACOS = CHECK_NAN_SNIPPET + `
   if (abs(x) > 1.) {
     return NAN;
   }
   return acos(x);
-`,_J=Ye({opSnippet:CJ}),EJ={kernelName:kl,backendName:"webgl",kernelFunc:_J},AJ=Da+`
+`;
+var acos3 = unaryKernelFunc2({ opSnippet: ACOS });
+var acosConfig2 = {
+  kernelName: Acos,
+  backendName: "webgl",
+  kernelFunc: acos3
+};
+var ACOSH = CHECK_NAN_SNIPPET + `
   if (x < 1.0) return NAN;
-return log(x + sqrt(x * x - 1.0));`,$J=Ye({opSnippet:AJ}),FJ={kernelName:Il,backendName:"webgl",kernelFunc:$J},nI="return a + b;",DJ=pn({opSnippet:nI,packedOpSnippet:nI,supportsComplex:!0,cpuKernelImpl:UZ}),RJ={kernelName:cs,backendName:"webgl",kernelFunc:DJ},MJ=class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((r,s)=>`T${s}`);let n=[];this.variableNames.forEach(r=>{n.push(`float v${r} = get${r}AtOutCoords();`)});let a=this.variableNames.map(r=>`v${r}`).join(" + ");this.userCode=`
+return log(x + sqrt(x * x - 1.0));`;
+var acosh3 = unaryKernelFunc2({ opSnippet: ACOSH });
+var acoshConfig2 = {
+  kernelName: Acosh,
+  backendName: "webgl",
+  kernelFunc: acosh3
+};
+var ADD = "return a + b;";
+var addKernelFunc = binaryKernelFunc2({
+  opSnippet: ADD,
+  packedOpSnippet: ADD,
+  supportsComplex: true,
+  cpuKernelImpl: addImplCPU
+});
+var addConfig2 = {
+  kernelName: Add,
+  backendName: "webgl",
+  kernelFunc: addKernelFunc
+};
+var AddNProgram = class {
+  constructor(outputShape, shapes) {
+    this.outputShape = [];
+    this.outputShape = outputShape;
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const snippets = [];
+    this.variableNames.forEach((variable2) => {
+      snippets.push(`float v${variable2} = get${variable2}AtOutCoords();`);
+    });
+    const operation = this.variableNames.map((variable2) => {
+      return `v${variable2}`;
+    }).join(" + ");
+    this.userCode = `
       void main() {
-        ${n.join(`
-        `)}
+        ${snippets.join("\n        ")}
 
-        float result = ${a};
+        float result = ${operation};
         setOutput(result);
       }
-    `}},PJ=class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((r,s)=>`T${s}`);let n=[];this.variableNames.forEach(r=>{n.push(`vec4 v${r} = get${r}AtOutCoords();`)});let a=this.variableNames.map(r=>`v${r}`).join(" + ");this.userCode=`
+    `;
+  }
+};
+var AddNPackedProgram = class {
+  constructor(outputShape, shapes) {
+    this.outputShape = [];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const snippets = [];
+    this.variableNames.forEach((variable2) => {
+      snippets.push(`vec4 v${variable2} = get${variable2}AtOutCoords();`);
+    });
+    const operation = this.variableNames.map((variable2) => {
+      return `v${variable2}`;
+    }).join(" + ");
+    this.userCode = `
       void main() {
-        ${n.join(`
-        `)}
+        ${snippets.join("\n        ")}
 
-        vec4 result = ${a};
+        vec4 result = ${operation};
         setOutput(result);
       }
-    `}};function ch(e){let{inputs:t,backend:n}=e,a=t;if(a.length===1)return na({inputs:{x:a[0]},backend:n});if(a.length>H().get("WEBGL_MAX_TEXTURES_IN_SHADER")){let o=Math.floor(a.length/2),l=ch({inputs:a.slice(0,o),backend:n}),u=ch({inputs:a.slice(o),backend:n});return ch({inputs:[l,u],backend:n})}let r=a.map(o=>o.dtype).reduce((o,l)=>ma(o,l)),s=a.map(o=>o.shape),i=H().getBool("WEBGL_PACK")?new PJ(a[0].shape,s):new MJ(a[0].shape,s);return n.runWebGLProgram(i,a,r)}var OJ={kernelName:mi,backendName:"webgl",kernelFunc:ch};function LJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=In({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,o)),N.assertAxesAreInnerMostDims("all",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=ko(f,f.dtype,"all",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var zJ={kernelName:Sl,backendName:"webgl",kernelFunc:LJ};function WJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=In({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,o)),N.assertAxesAreInnerMostDims("any",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=ko(f,f.dtype,"any",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var BJ={kernelName:Tl,backendName:"webgl",kernelFunc:WJ},VJ=class{constructor(e,t,n){this.variableNames=["A"];let{windowSize:a,batchSize:r,outSize:s}=e;n||this.variableNames.push("bestIndicesA"),this.outputShape=[r,s];let i=t==="max"?">":"<",o=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`
+    `;
+  }
+};
+function addN3(args) {
+  const { inputs, backend: backend2 } = args;
+  const tensors = inputs;
+  if (tensors.length === 1) {
+    return identity3({ inputs: { x: tensors[0] }, backend: backend2 });
+  }
+  if (tensors.length > env().get("WEBGL_MAX_TEXTURES_IN_SHADER")) {
+    const midIndex = Math.floor(tensors.length / 2);
+    const leftSide = addN3({ inputs: tensors.slice(0, midIndex), backend: backend2 });
+    const rightSide = addN3({ inputs: tensors.slice(midIndex), backend: backend2 });
+    return addN3({ inputs: [leftSide, rightSide], backend: backend2 });
+  }
+  const dtype = tensors.map((t) => t.dtype).reduce((d1, d2) => upcastType(d1, d2));
+  const shapes = tensors.map((t) => t.shape);
+  const usePackedOp = env().getBool("WEBGL_PACK");
+  const program = usePackedOp ? new AddNPackedProgram(tensors[0].shape, shapes) : new AddNProgram(tensors[0].shape, shapes);
+  return backend2.runWebGLProgram(program, tensors, dtype);
+}
+var addNConfig2 = {
+  kernelName: AddN,
+  backendName: "webgl",
+  kernelFunc: addN3
+};
+function all3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "all", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var allConfig2 = {
+  kernelName: All,
+  backendName: "webgl",
+  kernelFunc: all3
+};
+function any3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "any", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var anyConfig2 = {
+  kernelName: Any,
+  backendName: "webgl",
+  kernelFunc: any3
+};
+var ArgMinMaxProgram = class {
+  constructor(reduceInfo, op2, firstPass) {
+    this.variableNames = ["A"];
+    const { windowSize, batchSize, outSize } = reduceInfo;
+    if (!firstPass) {
+      this.variableNames.push("bestIndicesA");
+    }
+    this.outputShape = [batchSize, outSize];
+    const compOp = op2 === "max" ? ">" : "<";
+    const indexSnippet = firstPass ? "inOffset + i;" : "round(getBestIndicesA(batch, inOffset + i));";
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${a};
+        int inOffset = outIdx * ${windowSize};
 
         int bestIndex = inOffset;
         float bestValue = getA(batch, bestIndex);
 
-        for (int i = 0; i < ${a}; i++) {
-          int inIdx = ${o};
+        for (int i = 0; i < ${windowSize}; i++) {
+          int inIdx = ${indexSnippet};
           float candidate = getA(batch, inIdx);
-          if (candidate ${i} bestValue) {
+          if (candidate ${compOp} bestValue) {
             bestValue = candidate;
             bestIndex = inIdx;
           }
         }
         setOutput(float(bestIndex));
       }
-    `}},UJ=class{constructor(e,t,n,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,v.assert(e.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let r=e[e.length-1],s=Math.ceil(r/t);this.outputShape=e.slice(0,-1),s>1&&this.outputShape.push(s),a||this.variableNames.push("bestIndicesA");let i=this.outputShape,o=i.length,l=gt(o),u=wn("coords",o),p,d;if(s===1){d=o+1;let C=gt(d);p=`
-        ${C} sourceLocR = ${C}(${u.join()}, 0);
-        ++${u[o-1]};
-        ${C} sourceLocG = ${C}(${u.join()}, 0);
-        ++${u[o-2]};
-        ${C} sourceLocA = ${C}(${u.join()}, 0);
-        --${u[o-1]};
-        ${C} sourceLocB = ${C}(${u.join()}, 0);
-        --${u[o-2]};`}else d=o,p=`
-        ${l} sourceLocR = coords;
-        ++${u[o-1]};
-        ${l} sourceLocG = coords;
-        ++${u[o-2]};
-        ${l} sourceLocA = coords;
-        --${u[o-1]};
-        ${l} sourceLocB = coords;
-        --${u[o-2]};`;let c=["x","y","z","w","u","v"].slice(0,d),h="."+c[d-1],m=c.map(C=>"int "+C),f=wn("sourceLocR",d-1).concat("inIdx.r"),g=wn("sourceLocG",d-1).concat("inIdx.g"),y=wn("sourceLocB",d-1).concat("inIdx.b"),b=wn("sourceLocA",d-1).concat("inIdx.a"),x=n==="max"?"greaterThan":"lessThan",w=a?"":`
-          inIdx = round(vec4(getBestIndicesAChannel(${f.join()}),
-                             getBestIndicesAChannel(${g.join()}),
-                             getBestIndicesAChannel(${y.join()}),
-                             getBestIndicesAChannel(${b.join()})));`,I=`vec4(
-            getAChannel(${f.join()}),
-            hasNextCol ? getAChannel(${g.join()}) : 0.,
-            hasNextRow ? getAChannel(${y.join()}) : 0.,
-            hasNextRow && hasNextCol ? getAChannel(${b.join()}) : 0.)`,T=a?"":`
-      float getBestIndicesAChannel(${m.join()}) {
-        return getChannel(getBestIndicesA(${c.join()}),
-                                          vec2(${c.slice(-2).join()}));
-      }`;this.userCode=`
-      float getAChannel(${m.join()}) {
-        return getChannel(getA(${c.join()}),
-                               vec2(${c.slice(-2).join()}));
+    `;
+  }
+};
+var ArgMinMaxPackedProgram = class {
+  constructor(shape, windowSize, op2, firstPass) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    util_exports.assert(shape.length > 2, () => `Packed arg${op2.charAt(0).toUpperCase() + op2.slice(1)} supports only inputs with rank above 2.`);
+    const inSize = shape[shape.length - 1];
+    const outSize = Math.ceil(inSize / windowSize);
+    this.outputShape = shape.slice(0, -1);
+    if (outSize > 1) {
+      this.outputShape.push(outSize);
+    }
+    if (!firstPass) {
+      this.variableNames.push("bestIndicesA");
+    }
+    const outShape = this.outputShape;
+    const rank = outShape.length;
+    const dtype = getCoordsDataType(rank);
+    const coords2 = getChannels("coords", rank);
+    let sourceLocSetup;
+    let sourceRank;
+    if (outSize === 1) {
+      sourceRank = rank + 1;
+      const sourceLocDType = getCoordsDataType(sourceRank);
+      sourceLocSetup = `
+        ${sourceLocDType} sourceLocR = ${sourceLocDType}(${coords2.join()}, 0);
+        ++${coords2[rank - 1]};
+        ${sourceLocDType} sourceLocG = ${sourceLocDType}(${coords2.join()}, 0);
+        ++${coords2[rank - 2]};
+        ${sourceLocDType} sourceLocA = ${sourceLocDType}(${coords2.join()}, 0);
+        --${coords2[rank - 1]};
+        ${sourceLocDType} sourceLocB = ${sourceLocDType}(${coords2.join()}, 0);
+        --${coords2[rank - 2]};`;
+    } else {
+      sourceRank = rank;
+      sourceLocSetup = `
+        ${dtype} sourceLocR = coords;
+        ++${coords2[rank - 1]};
+        ${dtype} sourceLocG = coords;
+        ++${coords2[rank - 2]};
+        ${dtype} sourceLocA = coords;
+        --${coords2[rank - 1]};
+        ${dtype} sourceLocB = coords;
+        --${coords2[rank - 2]};`;
+    }
+    const channels = ["x", "y", "z", "w", "u", "v"].slice(0, sourceRank);
+    const inChannel = "." + channels[sourceRank - 1];
+    const intChannels = channels.map((x) => "int " + x);
+    const srcRCoords = getChannels("sourceLocR", sourceRank - 1).concat("inIdx.r");
+    const srcGCoords = getChannels("sourceLocG", sourceRank - 1).concat("inIdx.g");
+    const srcBCoords = getChannels("sourceLocB", sourceRank - 1).concat("inIdx.b");
+    const srcACoords = getChannels("sourceLocA", sourceRank - 1).concat("inIdx.a");
+    const compOp = op2 === "max" ? "greaterThan" : "lessThan";
+    const fetchCandidateIdx = firstPass ? "" : `
+          inIdx = round(vec4(getBestIndicesAChannel(${srcRCoords.join()}),
+                             getBestIndicesAChannel(${srcGCoords.join()}),
+                             getBestIndicesAChannel(${srcBCoords.join()}),
+                             getBestIndicesAChannel(${srcACoords.join()})));`;
+    const fetchValue = `vec4(
+            getAChannel(${srcRCoords.join()}),
+            hasNextCol ? getAChannel(${srcGCoords.join()}) : 0.,
+            hasNextRow ? getAChannel(${srcBCoords.join()}) : 0.,
+            hasNextRow && hasNextCol ? getAChannel(${srcACoords.join()}) : 0.)`;
+    const getBestIndicesAChannelSnippet = firstPass ? "" : `
+      float getBestIndicesAChannel(${intChannels.join()}) {
+        return getChannel(getBestIndicesA(${channels.join()}),
+                                          vec2(${channels.slice(-2).join()}));
+      }`;
+    this.userCode = `
+      float getAChannel(${intChannels.join()}) {
+        return getChannel(getA(${channels.join()}),
+                               vec2(${channels.slice(-2).join()}));
       }
-      ${T}
+      ${getBestIndicesAChannelSnippet}
       void main() {
-        ${l} coords = getOutputCoords();
-        bool hasNextCol = ${u[o-1]} < ${i[o-1]-1};
-        bool hasNextRow = ${u[o-2]} < ${i[o-2]-1};
-        ${p}
-        ivec4 srcIdx = ivec4(sourceLocR${h}, sourceLocG${h},
-          sourceLocB${h}, sourceLocA${h}) * ${t};
+        ${dtype} coords = getOutputCoords();
+        bool hasNextCol = ${coords2[rank - 1]} < ${outShape[rank - 1] - 1};
+        bool hasNextRow = ${coords2[rank - 2]} < ${outShape[rank - 2] - 1};
+        ${sourceLocSetup}
+        ivec4 srcIdx = ivec4(sourceLocR${inChannel}, sourceLocG${inChannel},
+          sourceLocB${inChannel}, sourceLocA${inChannel}) * ${windowSize};
         ivec4 inIdx = srcIdx;
         vec4 bestIndex = vec4(inIdx);
-        vec4 bestValue = ${I};
+        vec4 bestValue = ${fetchValue};
 
-        for (int i = 0; i < ${t}; i++) {
+        for (int i = 0; i < ${windowSize}; i++) {
           inIdx = srcIdx;
-          ${w}
-          vec4 candidate = ${I};
+          ${fetchCandidateIdx}
+          vec4 candidate = ${fetchValue};
           bvec4 nan = isnan(candidate);
           bvec4 replace = bvec4(
-            vec4(${x}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
+            vec4(${compOp}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
 
           bestValue = vec4(replace.x  ? candidate.x : bestValue.x,
                            replace.y  ? candidate.y : bestValue.y,
@@ -1629,27 +53590,197 @@ return log(x + sqrt(x * x - 1.0));`,$J=Ye({opSnippet:AJ}),FJ={kernelName:Il,back
         }
         setOutput(bestIndex);
       }
-    `}};function X_(e,t,n,a=null){let r=t.shape[0],s=t.shape[1];a!=null&&(r=a.shape[0],s=a.shape[1]);let i=N.computeOptimalWindowSize(s),o={windowSize:i,inSize:s,batchSize:r,outSize:Math.ceil(s/i)},l=new VJ(o,n,a==null),u=[t];a!=null&&u.push(a);let p=e.runWebGLProgram(l,u,"int32");if(p.shape[1]===1)return p;let d=X_(e,t,n,p);return e.disposeIntermediateTensorInfo(p),d}function Y_(e,t,n,a=null){let r=a!=null?a.shape:t.shape,s=r[r.length-1],i=N.computeOptimalWindowSize(s),o=new UJ(r,i,n,a==null),l=a==null?[t]:[t,a],u=e.runWebGLProgram(o,l,"int32");if(u.shape.length===t.shape.length){let p=Y_(e,t,n,u);return e.disposeIntermediateTensorInfo(u),p}return u}function Z_(e,t,n,a){let r=[n];if(N.assertAxesAreInnerMostDims("arg"+a.charAt(0).toUpperCase()+a.slice(1),r,t.shape.length),!H().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let s=[],i=e.texData.get(t.dataId),o=i!==null&&i.isPacked,l=t;o&&(l=e.unpackTensor(t),s.push(l));let[u,p]=N.computeOutAndReduceShapes(l.shape,r),d=v.sizeFromShape(p),c=de({inputs:{x:l},backend:e,attrs:{shape:[-1,d]}});s.push(c);let h=X_(e,c,a);s.push(h);let m=de({inputs:{x:h},backend:e,attrs:{shape:u}});return s.forEach(f=>e.disposeIntermediateTensorInfo(f)),m}return Y_(e,t,a)}function GJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a,i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=In({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),N.assertAxesAreInnerMostDims("argMax",[i[0]],l.shape.length);let p=Z_(n,l,i[0],"max");return u.forEach(d=>n.disposeIntermediateTensorInfo(d)),p}var HJ={kernelName:fi,backendName:"webgl",kernelFunc:GJ};function jJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a,i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=In({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),N.assertAxesAreInnerMostDims("argMin",[i[0]],l.shape.length);let p=Z_(n,l,i[0],"min");return u.forEach(d=>n.disposeIntermediateTensorInfo(d)),p}var qJ={kernelName:sc,backendName:"webgl",kernelFunc:jJ},KJ=Da+`
+    `;
+  }
+};
+function argReduce(backend2, x, reduceType, bestIndicesA = null) {
+  let batchSize = x.shape[0];
+  let inSize = x.shape[1];
+  if (bestIndicesA != null) {
+    batchSize = bestIndicesA.shape[0];
+    inSize = bestIndicesA.shape[1];
+  }
+  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);
+  const reduceInfo = { windowSize, inSize, batchSize, outSize: Math.ceil(inSize / windowSize) };
+  const program = new ArgMinMaxProgram(reduceInfo, reduceType, bestIndicesA == null);
+  const inputs = [x];
+  if (bestIndicesA != null) {
+    inputs.push(bestIndicesA);
+  }
+  const output = backend2.runWebGLProgram(program, inputs, "int32");
+  if (output.shape[1] === 1) {
+    return output;
+  }
+  const result = argReduce(backend2, x, reduceType, output);
+  backend2.disposeIntermediateTensorInfo(output);
+  return result;
+}
+function argReducePacked(backend2, x, reduceType, bestIndicesA = null) {
+  const inShape = bestIndicesA != null ? bestIndicesA.shape : x.shape;
+  const inSize = inShape[inShape.length - 1];
+  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);
+  const program = new ArgMinMaxPackedProgram(inShape, windowSize, reduceType, bestIndicesA == null);
+  const inputs = bestIndicesA == null ? [x] : [x, bestIndicesA];
+  const output = backend2.runWebGLProgram(program, inputs, "int32");
+  if (output.shape.length === x.shape.length) {
+    const result = argReducePacked(backend2, x, reduceType, output);
+    backend2.disposeIntermediateTensorInfo(output);
+    return result;
+  }
+  return output;
+}
+function argMinMaxReduce(backend2, x, axis, reduceType) {
+  const axes = [axis];
+  backend_util_exports.assertAxesAreInnerMostDims("arg" + reduceType.charAt(0).toUpperCase() + reduceType.slice(1), axes, x.shape.length);
+  if (!env().getBool("WEBGL_PACK_REDUCE") || x.shape.length <= 2) {
+    const intermediateTensorInfos = [];
+    const xtexData = backend2.texData.get(x.dataId);
+    const xIsPacked = xtexData !== null && xtexData.isPacked;
+    let xUnPacked = x;
+    if (xIsPacked) {
+      xUnPacked = backend2.unpackTensor(x);
+      intermediateTensorInfos.push(xUnPacked);
+    }
+    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xUnPacked.shape, axes);
+    const inSize = util_exports.sizeFromShape(reduceShape);
+    const a2D = reshape4({ inputs: { x: xUnPacked }, backend: backend2, attrs: { shape: [-1, inSize] } });
+    intermediateTensorInfos.push(a2D);
+    const reduced = argReduce(backend2, a2D, reduceType);
+    intermediateTensorInfos.push(reduced);
+    const reshaped = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+    intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return reshaped;
+  }
+  return argReducePacked(backend2, x, reduceType);
+}
+function argMax3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("argMax", [axes[0]], $x.shape.length);
+  const out = argMinMaxReduce(backend2, $x, axes[0], "max");
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return out;
+}
+var argMaxConfig2 = {
+  kernelName: ArgMax,
+  backendName: "webgl",
+  kernelFunc: argMax3
+};
+function argMin3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("argMin", [axes[0]], $x.shape.length);
+  const out = argMinMaxReduce(backend2, $x, axes[0], "min");
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return out;
+}
+var argMinConfig2 = {
+  kernelName: ArgMin,
+  backendName: "webgl",
+  kernelFunc: argMin3
+};
+var ASIN = CHECK_NAN_SNIPPET + `
   if (abs(x) > 1.) {
     return NAN;
   }
   return asin(x);
-`,XJ=Ye({opSnippet:KJ}),YJ={kernelName:Nl,backendName:"webgl",kernelFunc:XJ},ZJ=Da+"return log(x + sqrt(x * x + 1.0));",JJ=Ye({opSnippet:ZJ}),QJ={kernelName:Cl,backendName:"webgl",kernelFunc:JJ},e9=Da+`
+`;
+var asin3 = unaryKernelFunc2({ opSnippet: ASIN });
+var asinConfig2 = {
+  kernelName: Asin,
+  backendName: "webgl",
+  kernelFunc: asin3
+};
+var ASINH = CHECK_NAN_SNIPPET + `return log(x + sqrt(x * x + 1.0));`;
+var asinh3 = unaryKernelFunc2({ opSnippet: ASINH });
+var asinhConfig2 = {
+  kernelName: Asinh,
+  backendName: "webgl",
+  kernelFunc: asinh3
+};
+var ATAN = CHECK_NAN_SNIPPET + `
   return atan(x);
-`,t9=Ye({opSnippet:e9}),n9={kernelName:_l,backendName:"webgl",kernelFunc:t9},a9=J0+`
+`;
+var atan4 = unaryKernelFunc2({ opSnippet: ATAN });
+var atanConfig2 = {
+  kernelName: Atan,
+  backendName: "webgl",
+  kernelFunc: atan4
+};
+var ATAN2 = CHECK_NAN_SNIPPET2 + `
   return atan(a, b);
-`,r9=`
+`;
+var ATAN2_PACKED = `
   vec4 result = atan(a, b);
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+nd+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,s9=pn({opSnippet:a9,packedOpSnippet:r9}),i9={kernelName:Al,backendName:"webgl",kernelFunc:s9},o9=Da+`
+`;
+var atan23 = binaryKernelFunc2({ opSnippet: ATAN2, packedOpSnippet: ATAN2_PACKED });
+var atan2Config2 = {
+  kernelName: Atan2,
+  backendName: "webgl",
+  kernelFunc: atan23
+};
+var ATANH = CHECK_NAN_SNIPPET + `
   if ((x < -1.0) || (x > 1.0)) return NAN;
-return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelName:El,backendName:"webgl",kernelFunc:l9},tc=class{constructor(e,t,n,a=!1,r=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideHeight,o=e.strideWidth,l=e.dilationHeight,u=e.dilationWidth,p=e.effectiveFilterHeight,d=e.effectiveFilterWidth,c=e.padInfo.top,h=e.padInfo.left;this.outputShape=e.outShape;let m=t==="avg",f=`((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`,y="0.0";if(m||(y="-1.0 / 1e-20"),n){let C=">=";this.userCode=`
-        const ivec2 strides = ivec2(${i}, ${o});
-        const ivec2 pads = ivec2(${c}, ${h});
+return (log(1.0 + x) - log(1.0 - x)) / 2.0;`;
+var atanh3 = unaryKernelFunc2({ opSnippet: ATANH });
+var atanhConfig2 = {
+  kernelName: Atanh,
+  backendName: "webgl",
+  kernelFunc: atanh3
+};
+var Pool2DProgram = class {
+  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {
+    this.variableNames = ["x"];
+    if (poolType === "avg" && computePositions) {
+      throw new Error("Cannot compute positions for average pool.");
+    }
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.outputShape = convInfo.outShape;
+    const isAvgPool = poolType === "avg";
+    const batchFlattenPositionStr = `((batch  * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;
+    const flattenPositionStr = `(xR * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;
+    let initializationValue = "0.0";
+    if (!isAvgPool) {
+      initializationValue = "-1.0 / 1e-20";
+    }
+    if (computePositions) {
+      const compareOp2 = ">=";
+      this.userCode = `
+        const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+        const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
         void main() {
           ivec4 coords = getOutputCoords();
@@ -1667,19 +53798,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           int minMaxPosition = 0;
           float avgValue = 0.0;
 
-          for (int wR = 0; wR < ${p};
-              wR += ${l}) {
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
             int xR = xRCorner + wR;
 
-            if (xR < 0 || xR >= ${e.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${d};
-                wC += ${u}) {
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
               int xC = xCCorner + wC;
 
-              if (xC < 0 || xC >= ${e.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
@@ -1689,31 +53820,42 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
               // use the current value.
               float currMinMaxValue = mix(
                   value, minMaxValue, minMaxValueFound);
-              if (value ${C} currMinMaxValue) {
+              if (value ${compareOp2} currMinMaxValue) {
                 minMaxValue = value;
                 minMaxValueFound = 1.0;
-                minMaxPosition = ${a?r?f:g:`wR * ${d} + wC`};
+                minMaxPosition = ${flattenPositions ? includeBatchInIndex ? batchFlattenPositionStr : flattenPositionStr : `wR * ${effectiveFilterWidth} + wC`};
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let b="max",x=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(x="avgValue / count");let w=Math.floor(s/4)*4,I=s%4,T=`
-      if (${m}) {
+      `;
+      return;
+    }
+    const compareOp = "max";
+    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (poolType === "avg") {
+      returnValue = `avgValue / count`;
+    }
+    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;
+    const filterWidthVec4Remainder = filterWidth % 4;
+    const updateSnippet = `
+      if (${isAvgPool}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${b}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
       }
-    `;this.userCode=`
-      const ivec2 strides = ivec2(${i}, ${o});
-      const ivec2 pads = ivec2(${c}, ${h});
-      const float initializationValue = ${y};
+    `;
+    this.userCode = `
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
 
       float getValue(int batch, int xR, int xC, int d) {
-        if (xC < 0 || xC >= ${e.inWidth}) {
+        if (xC < 0 || xC >= ${convInfo.inWidth}) {
           return initializationValue;
         }
         count += 1.0;
@@ -1731,33 +53873,33 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         // max/min x(?, ?, d) to get y(yR, yC, d).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${y});
+        vec4 minMaxValue = vec4(${initializationValue});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wR = 0; wR < ${p};
-            wR += ${l}) {
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
           int xR = xRCorner + wR;
 
-          if (xR < 0 || xR >= ${e.inHeight}) {
+          if (xR < 0 || xR >= ${convInfo.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${w}; wC += 4) {
-            int xC = xCCorner + wC * ${u};
+          for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {
+            int xC = xCCorner + wC * ${dilationWidth};
 
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${u}, d),
-              getValue(batch, xR, xC + 2 * ${u}, d),
-              getValue(batch, xR, xC + 3 * ${u}, d)
+              getValue(batch, xR, xC + ${dilationWidth}, d),
+              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),
+              getValue(batch, xR, xC + 3 * ${dilationWidth}, d)
             );
 
-            ${T}
+            ${updateSnippet}
           }
 
-          int xC = xCCorner + ${w};
-          if (${I===1}) {
+          int xC = xCCorner + ${filterWidthNearestVec4};
+          if (${filterWidthVec4Remainder === 1}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
               initializationValue,
@@ -1765,33 +53907,63 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
               initializationValue
             );
 
-            ${T}
-          } else if (${I===2}) {
+            ${updateSnippet}
+          } else if (${filterWidthVec4Remainder === 2}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${u}, d),
+              getValue(batch, xR, xC + ${dilationWidth}, d),
               initializationValue,
               initializationValue
             );
 
-            ${T}
-          } else if (${I===3}) {
+            ${updateSnippet}
+          } else if (${filterWidthVec4Remainder === 3}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${u}, d),
-              getValue(batch, xR, xC + 2 * ${u}, d),
+              getValue(batch, xR, xC + ${dilationWidth}, d),
+              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),
               initializationValue
             );
 
-            ${T}
+            ${updateSnippet}
           }
         }
-        setOutput(${x});
+        setOutput(${returnValue});
       }
-    `}},e1=class{constructor(e,t,n,a=!1,r=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideDepth,o=e.strideHeight,l=e.strideWidth,u=e.dilationDepth,p=e.dilationHeight,d=e.dilationWidth,c=e.effectiveFilterDepth,h=e.effectiveFilterHeight,m=e.effectiveFilterWidth,f=e.padInfo.front,g=e.padInfo.top,y=e.padInfo.left;this.outputShape=e.outShape;let b=t==="avg",x="0.0";if(b||(x="-1.0 / 1e-20"),n){let A=">=";this.userCode=`
+    `;
+  }
+};
+var Pool3DProgram = class {
+  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {
+    this.variableNames = ["x"];
+    if (poolType === "avg" && computePositions) {
+      throw new Error("Cannot compute positions for average pool.");
+    }
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.outputShape = convInfo.outShape;
+    const isAvgPool = poolType === "avg";
+    let initializationValue = "0.0";
+    if (!isAvgPool) {
+      initializationValue = "-1.0 / 1e-20";
+    }
+    if (computePositions) {
+      const compareOp2 = ">=";
+      this.userCode = `
         const ivec3 strides =
-            ivec3(${i}, ${o}, ${l});
-        const ivec3 pads = ivec3(${f}, ${g}, ${y});
+            ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+        const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
         void main() {
           ivec5 coords = getOutputCoords();
@@ -1809,27 +53981,27 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           float minMaxValueFound = 0.0;
           int minMaxPosition = 0;
 
-          for (int wD = 0; wD < ${c};
-              wD += ${u}) {
+          for (int wD = 0; wD < ${effectiveFilterDepth};
+              wD += ${dilationDepth}) {
             int xD = xDCorner + wD;
 
-            if (xD < 0 || xD >= ${e.inDepth}) {
+            if (xD < 0 || xD >= ${convInfo.inDepth}) {
               continue;
             }
 
-            for (int wR = 0; wR < ${h};
-                wR += ${p}) {
+            for (int wR = 0; wR < ${effectiveFilterHeight};
+                wR += ${dilationHeight}) {
               int xR = xRCorner + wR;
 
-              if (xR < 0 || xR >= ${e.inHeight}) {
+              if (xR < 0 || xR >= ${convInfo.inHeight}) {
                 continue;
               }
 
-              for (int wC = 0; wC < ${m};
-                  wC += ${d}) {
+              for (int wC = 0; wC < ${effectiveFilterWidth};
+                  wC += ${dilationWidth}) {
                 int xC = xCCorner + wC;
 
-                if (xC < 0 || xC >= ${e.inWidth}) {
+                if (xC < 0 || xC >= ${convInfo.inWidth}) {
                   continue;
                 }
 
@@ -1839,34 +54011,45 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                 // use the current value.
                 float currMinMaxValue = mix(
                     value, minMaxValue, minMaxValueFound);
-                if (value ${A} currMinMaxValue) {
+                if (value ${compareOp2} currMinMaxValue) {
                   minMaxValue = value;
                   minMaxValueFound = 1.0;
-                  minMaxPosition = ${a?r?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${h} * ${m} +
-                      wR * ${m} + wC`};
+                  minMaxPosition = ${flattenPositions ? includeBatchInIndex ? `(((batch * ${convInfo.inDepth} + xD) * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `((xD * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +
+                      wR * ${effectiveFilterWidth} + wC`};
                 }
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let w="max",I=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(I="avgValue / count");let T=Math.floor(s/4)*4,C=s%4,E=`
-      if (${b}) {
+      `;
+      return;
+    }
+    const compareOp = "max";
+    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (poolType === "avg") {
+      returnValue = `avgValue / count`;
+    }
+    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;
+    const filterWidthVec4Remainder = filterWidth % 4;
+    const updateSnippet = `
+      if (${isAvgPool}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${w}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
       }
-    `;this.userCode=`
+    `;
+    this.userCode = `
       const ivec3 strides =
-        ivec3(${i}, ${o}, ${l});
-      const ivec3 pads = ivec3(${f}, ${g}, ${y});
-      const float initializationValue = ${x};
+        ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
 
       float getValue(int batch, int xD, int xR, int xC, int ch) {
-        if (xC < 0 || xC >= ${e.inWidth}) {
+        if (xC < 0 || xC >= ${convInfo.inWidth}) {
           return initializationValue;
         }
         count += 1.0;
@@ -1885,41 +54068,41 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${x});
+        vec4 minMaxValue = vec4(${initializationValue});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wD = 0; wD < ${c};
-            wD += ${u}) {
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+            wD += ${dilationDepth}) {
           int xD = xDCorner + wD;
 
-          if (xD < 0 || xD >= ${e.inDepth}) {
+          if (xD < 0 || xD >= ${convInfo.inDepth}) {
             continue;
           }
 
-          for (int wR = 0; wR < ${h};
-            wR += ${p}) {
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
             int xR = xRCorner + wR;
 
-            if (xR < 0 || xR >= ${e.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${T}; wC += 4) {
-              int xC = xCCorner + wC * ${d};
+            for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {
+              int xC = xCCorner + wC * ${dilationWidth};
 
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${d}, ch),
-                getValue(batch, xD, xR, xC + 3 * ${d}, ch)
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 3 * ${dilationWidth}, ch)
               );
 
-              ${E}
+              ${updateSnippet}
             }
 
-            int xC = xCCorner + ${T};
-            if (${C===1}) {
+            int xC = xCCorner + ${filterWidthNearestVec4};
+            if (${filterWidthVec4Remainder === 1}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
                 initializationValue,
@@ -1927,33 +54110,84 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                 initializationValue
               );
 
-              ${E}
-            } else if (${C===2}) {
+              ${updateSnippet}
+            } else if (${filterWidthVec4Remainder === 2}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
                 initializationValue,
                 initializationValue
               );
 
-              ${E}
-            } else if (${C===3}) {
+              ${updateSnippet}
+            } else if (${filterWidthVec4Remainder === 3}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${d}, ch),
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),
                 initializationValue
               );
 
-              ${E}
+              ${updateSnippet}
             }
           }
-          setOutput(${I});
+          setOutput(${returnValue});
         }
       }
-    `}};function p9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;Lu(r,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l);if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))return na({inputs:{x:r},backend:n});let d=new tc(p,"avg",!1);return n.runWebGLProgram(d,[r],"float32")}var c9={kernelName:gi,backendName:"webgl",kernelFunc:p9};function d9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a,p=[1,1,1],d=N.computePool3DInfo(r.shape,s,i,p,o,l,u),c=new e1(d,"avg",!1);return n.runWebGLProgram(c,[r],"float32")}var h9={kernelName:ic,backendName:"webgl",kernelFunc:d9},m9=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterHeight,l=e.effectiveFilterWidth,u=o-1-e.padInfo.top,p=l-1-e.padInfo.left,d=1/(t*n);this.userCode=`
-      const ivec2 pads = ivec2(${u}, ${p});
-      const float avgMultiplier = float(${d});
+    `;
+  }
+};
+function avgPool3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex2(x, "avgPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const avgPoolProgram = new Pool2DProgram(convInfo, "avg", false);
+  return backend2.runWebGLProgram(avgPoolProgram, [x], "float32");
+}
+var avgPoolConfig2 = {
+  kernelName: AvgPool,
+  backendName: "webgl",
+  kernelFunc: avgPool3
+};
+function avgPool3D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);
+  const avgPoolProgram = new Pool3DProgram(convInfo, "avg", false);
+  return backend2.runWebGLProgram(avgPoolProgram, [x], "float32");
+}
+var avgPool3DConfig2 = {
+  kernelName: AvgPool3D,
+  backendName: "webgl",
+  kernelFunc: avgPool3D2
+};
+var AvgPool2DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const avgMultiplier = 1 / (filterHeight * filterWidth);
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
+      const float avgMultiplier = float(${avgMultiplier});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -1967,20 +54201,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${o};
-            wR += ${s}) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${l};
-            wC+= ${i}) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+          for (int wC = 0; wC < ${effectiveFilterWidth};
+            wC+= ${dilationWidth}) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
@@ -1993,9 +54227,32 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},f9=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,a=e.filterWidth,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,p=e.effectiveFilterDepth,d=e.effectiveFilterHeight,c=e.effectiveFilterWidth,h=p-1-e.padInfo.front,m=d-1-e.padInfo.top,f=c-1-e.padInfo.left,g=1/(t*n*a);this.userCode=`
-      const ivec3 pads = ivec3(${h}, ${m}, ${f});
-      const float avgMultiplier = float(${g});
+    `;
+  }
+};
+var AvgPool3DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy"];
+    this.outputShape = convInfo.inShape;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
+      const float avgMultiplier = float(${avgMultiplier});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2012,30 +54269,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${p};
-            wD += ${o}) {
-          float dyD = float(dyDCorner + wD) / ${r}.0;
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+            wD += ${dilationDepth}) {
+          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;
 
-          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
+          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {
             continue;
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${d};
-              wR += ${l}) {
-            float dyR = float(dyRCorner + wR) / ${s}.0;
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
                 fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            for (int wC = 0; wC < ${c};
-                wC += ${u}) {
-              float dyC = float(dyCCorner + wC) / ${i}.0;
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
@@ -2049,69 +54306,427 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function g9(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:p}=a,d=[1,1,1],c=N.computePool3DInfo(i.shape,o,l,d,u,p),h=new f9(c);return n.runWebGLProgram(h,[r],i.dtype)}var y9={kernelName:Yh,backendName:"webgl",kernelFunc:g9};function b9(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s;Lu([r,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=a,p=N.computePool2DInfo(i.shape,o,l,1,u),d=new m9(p);return n.runWebGLProgram(d,[r],i.dtype)}var x9={kernelName:Xh,backendName:"webgl",kernelFunc:b9};function v9(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;return Bh({a:r,b:s,transposeA:i,transposeB:o,backend:n})}var w9={kernelName:yi,backendName:"webgl",kernelFunc:v9},k9=class{constructor(e,t,n,a,r,s){this.outputShape=[],this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let i="0.0";a!=null&&(N.assertAndGetBroadcastShape(e,a),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="1.0";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function avgPool3DGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  const avgPoolBackpropProgram = new AvgPool3DBackpropProgram(convInfo);
+  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);
+}
+var avgPool3DGradConfig3 = {
+  kernelName: AvgPool3DGrad,
+  backendName: "webgl",
+  kernelFunc: avgPool3DGrad2
+};
+function avgPoolGrad3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  assertNotComplex2([dy, input2], "avgPoolGrad");
+  const { filterSize, strides, pad: pad3 } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);
+  const avgPoolBackpropProgram = new AvgPool2DBackpropProgram(convInfo);
+  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);
+}
+var avgPoolGradConfig3 = {
+  kernelName: AvgPoolGrad,
+  backendName: "webgl",
+  kernelFunc: avgPoolGrad3
+};
+function batchMatMul2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  return batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2 });
+}
+var batchMatMulConfig2 = {
+  kernelName: BatchMatMul,
+  backendName: "webgl",
+  kernelFunc: batchMatMul2
+};
+var BatchNormProgram = class {
+  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {
+    this.outputShape = [];
+    this.variableNames = ["x", "mean", "variance"];
+    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);
+    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);
+    let offsetSnippet = "0.0";
+    if (offsetShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);
+      this.variableNames.push("offset");
+      offsetSnippet = "getOffsetAtOutCoords()";
+    }
+    let scaleSnippet = "1.0";
+    if (scaleShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);
+      this.variableNames.push("scale");
+      scaleSnippet = "getScaleAtOutCoords()";
+    }
+    this.outputShape = xShape;
+    this.userCode = `
       void main() {
         float x = getXAtOutCoords();
         float mean = getMeanAtOutCoords();
         float variance = getVarianceAtOutCoords();
-        float offset = ${i};
-        float scale = ${o};
-        float inv = scale * inversesqrt(variance + float(${s}));
+        float offset = ${offsetSnippet};
+        float scale = ${scaleSnippet};
+        float inv = scale * inversesqrt(variance + float(${varianceEpsilon}));
         setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));
       }
-    `}},I9=class{constructor(e,t,n,a,r,s){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let i="vec4(0.0)";a!=null&&(N.assertAndGetBroadcastShape(e,a),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="vec4(1.0)";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+var BatchNormPackedProgram = class {
+  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.variableNames = ["x", "mean", "variance"];
+    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);
+    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);
+    let offsetSnippet = "vec4(0.0)";
+    if (offsetShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);
+      this.variableNames.push("offset");
+      offsetSnippet = "getOffsetAtOutCoords()";
+    }
+    let scaleSnippet = "vec4(1.0)";
+    if (scaleShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);
+      this.variableNames.push("scale");
+      scaleSnippet = "getScaleAtOutCoords()";
+    }
+    this.outputShape = xShape;
+    this.userCode = `
       void main() {
-        vec4 offset = ${i};
-        vec4 scale = ${o};
+        vec4 offset = ${offsetSnippet};
+        vec4 scale = ${scaleSnippet};
 
         vec4 x = getXAtOutCoords();
         vec4 mean = getMeanAtOutCoords();
         vec4 variance = getVarianceAtOutCoords();
 
-        vec4 inv = scale * inversesqrt(variance + vec4(${s}));
+        vec4 inv = scale * inversesqrt(variance + vec4(${varianceEpsilon}));
 
         setOutput((x - mean) * inv + offset);
       }
-    `}},S9=({inputs:e,backend:t,attrs:n})=>{let{x:a,mean:r,variance:s,offset:i,scale:o}=e;v.assert(r.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||r.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(o==null||r.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:l}=n;l==null&&(l=.001);let u=[a,r,s],p=null;i!=null&&(p=i.shape,u.push(i));let d=null;o!=null&&(d=o.shape,u.push(o));let c=H().getBool("WEBGL_PACK_NORMALIZATION")?new I9(a.shape,r.shape,s.shape,p,d,l):new k9(a.shape,r.shape,s.shape,p,d,l);return t.runWebGLProgram(c,u,u[0].dtype)},T9={kernelName:$i,backendName:"webgl",kernelFunc:S9},N9=class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;let t=gt(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let n=C9(this.rank),a,r=e.map((s,i)=>`sourceLoc.${px[i]} = start[${i}] + coords.${px[i]};`);a=`
-        ${t} sourceLoc;
-        ${t} coords = getOutputCoords();
-        ${r.join(`
-`)}
-      `,this.userCode=`
+    `;
+  }
+};
+var batchNorm3 = ({ inputs, backend: backend2, attrs }) => {
+  const { x, mean: mean4, variance, offset, scale: scale22 } = inputs;
+  util_exports.assert(mean4.shape.length === variance.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  util_exports.assert(scale22 == null || mean4.shape.length === scale22.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  let { varianceEpsilon } = attrs;
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const finalInputs = [x, mean4, variance];
+  let offsetShape = null;
+  if (offset != null) {
+    offsetShape = offset.shape;
+    finalInputs.push(offset);
+  }
+  let scaleShape = null;
+  if (scale22 != null) {
+    scaleShape = scale22.shape;
+    finalInputs.push(scale22);
+  }
+  const program = env().getBool("WEBGL_PACK_NORMALIZATION") ? new BatchNormPackedProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon) : new BatchNormProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon);
+  const output = backend2.runWebGLProgram(program, finalInputs, finalInputs[0].dtype);
+  return output;
+};
+var batchNormConfig2 = {
+  kernelName: FusedBatchNorm,
+  backendName: "webgl",
+  kernelFunc: batchNorm3
+};
+var SliceProgram = class {
+  constructor(destSize) {
+    this.variableNames = ["source"];
+    this.outputShape = destSize;
+    this.rank = destSize.length;
+    const dtype = getCoordsDataType(this.rank);
+    this.customUniforms = [{ name: "start", arrayIndex: this.rank, type: "int" }];
+    const sourceCoords = getCoords(this.rank);
+    let body;
+    const coordSum = destSize.map((_, i) => {
+      return `sourceLoc.${coords[i]} = start[${i}] + coords.${coords[i]};`;
+    });
+    body = `
+        ${dtype} sourceLoc;
+        ${dtype} coords = getOutputCoords();
+        ${coordSum.join("\n")}
+      `;
+    this.userCode = `
       void main() {
-        ${a}
-        setOutput(getSource(${n}));
+        ${body}
+        setOutput(getSource(${sourceCoords}));
       }
-    `}},px=["x","y","z","w","u","v"];function C9(e){if(e===1)return"sourceLoc";if(e<=6)return px.slice(0,e).map(t=>"sourceLoc."+t).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}var _9=class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let t=gt(this.rank),n=wn("coords",this.rank),a=wn("sourceLoc",this.rank),r=this.rank===1?"sourceLoc":`vec2(${a.slice(-2).join()})`,s=`getChannel(getSource(${a.join()}), ${r})`,i=`
-      result.x = ${s};
-      if (++${n[this.rank-1]} < ${e[this.rank-1]}) {
-        ++${a[this.rank-1]};
-        result.y = ${s};
-        --${a[this.rank-1]};
+    `;
+  }
+};
+var coords = ["x", "y", "z", "w", "u", "v"];
+function getCoords(rank) {
+  if (rank === 1) {
+    return "sourceLoc";
+  } else if (rank <= 6) {
+    return coords.slice(0, rank).map((x) => "sourceLoc." + x).join(",");
+  } else {
+    throw Error(`Slicing for rank ${rank} is not yet supported`);
+  }
+}
+var SlicePackedProgram = class {
+  constructor(destSize) {
+    this.variableNames = ["source"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = destSize;
+    this.rank = destSize.length;
+    this.customUniforms = [{ name: "start", arrayIndex: this.rank, type: "int" }];
+    const dtype = getCoordsDataType(this.rank);
+    const coords2 = getChannels("coords", this.rank);
+    const sourceLoc = getChannels("sourceLoc", this.rank);
+    const innerDims = this.rank === 1 ? "sourceLoc" : `vec2(${sourceLoc.slice(-2).join()})`;
+    const getChannel = `getChannel(getSource(${sourceLoc.join()}), ${innerDims})`;
+    const upperRow = `
+      result.x = ${getChannel};
+      if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {
+        ++${sourceLoc[this.rank - 1]};
+        result.y = ${getChannel};
+        --${sourceLoc[this.rank - 1]};
       }
-    `,o=this.rank===1?"":`
-      --${n[this.rank-1]};
-      if (++${n[this.rank-2]} < ${e[this.rank-2]}) {
-        ++${a[this.rank-2]};
-        result.z = ${s};
-        if (++${n[this.rank-1]} < ${e[this.rank-1]}) {
-          ++${a[this.rank-1]};
-          result.w = ${s};
+    `;
+    const lowerRow = this.rank === 1 ? "" : `
+      --${coords2[this.rank - 1]};
+      if (++${coords2[this.rank - 2]} < ${destSize[this.rank - 2]}) {
+        ++${sourceLoc[this.rank - 2]};
+        result.z = ${getChannel};
+        if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {
+          ++${sourceLoc[this.rank - 1]};
+          result.w = ${getChannel};
         }
       }
-    `,l=this.rank<=4?`sourceLoc = coords +
-            ${t}(${e.map((u,p)=>`start[${p}]`).join()});`:e.map((u,p)=>`${a[p]} = ${n[p]} + start[${p}];`).join(`
-`);this.userCode=`
+    `;
+    const sourceLocSetup = this.rank <= 4 ? `sourceLoc = coords +
+            ${dtype}(${destSize.map((_, i) => `start[${i}]`).join()});` : destSize.map((_, i) => `${sourceLoc[i]} = ${coords2[i]} + start[${i}];`).join("\n");
+    this.userCode = `
       void main() {
-        ${t} coords = getOutputCoords();
-        ${t} sourceLoc;
-        ${l}
+        ${dtype} coords = getOutputCoords();
+        ${dtype} sourceLoc;
+        ${sourceLocSetup}
         vec4 result = vec4(0.);
-        ${i}
-        ${o}
+        ${upperRow}
+        ${lowerRow}
         setOutput(result);
       }
-    `}};function E9(e,t,n,a){let r=a.texData.get(e.dataId),s=a.makeTensorInfo(n,e.dtype),i=a.texData.get(s.dataId);Object.assign(i,r),i.refCount=1,i.shape=n,i.dtype=e.dtype;let o=jt.computeFlatOffset(t,v.computeStrides(e.shape));r.slice&&(o+=r.slice.flatOffset),i.slice={flatOffset:o,origDataId:r.slice&&r.slice.origDataId||e.dataId};let l=a.dataRefCount.get(i.slice.origDataId)||1;return a.dataRefCount.set(i.slice.origDataId,l+1),s}function Gu(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,size:i}=a,[o,l]=jt.parseSliceParams(r,s,i);if(jt.assertParamsValid(r,o,l),v.sizeFromShape(l)===0)return n.makeTensorInfo(l,r.dtype,[]);if(n.shouldExecuteOnCPU([r])||r.dtype==="string"){let d=n.texData.get(r.dataId),c=v7(d.values,o,l,r.shape,r.dtype);return n.makeTensorInfo(l,r.dtype,c)}let{isPacked:u}=n.texData.get(r.dataId),p=jt.isSliceContinous(r.shape,o,l);if(u||!p){let d=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new _9(l):new N9(l),c=[o];return n.runWebGLProgram(d,[r],r.dtype,c)}return n.uploadToGPU(r.dataId),E9(r,o,l,n)}var A9={kernelName:du,backendName:"webgl",kernelFunc:Gu},$9=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a;v.assert(r.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((b,x)=>b*x),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=[],m=de({inputs:{x:r},backend:n,attrs:{shape:l}}),f=In({inputs:{x:m},backend:n,attrs:{perm:u}}),g=de({inputs:{x:f},backend:n,attrs:{shape:p}}),y=Gu({inputs:{x:g},backend:n,attrs:{begin:d,size:c}});return h.push(m),h.push(f),h.push(g),h.forEach(b=>n.disposeIntermediateTensorInfo(b)),y},F9={kernelName:$l,backendName:"webgl",kernelFunc:$9};function D9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i}=a,o=n.readSync(r.dataId),l=n.readSync(s.dataId),u=O_(o,l,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,u)}var R9={kernelName:Zh,backendName:"webgl",kernelFunc:D9};function M9(e){let{inputs:t,backend:n}=e,{s0:a,s1:r}=t,s=n.readSync(a.dataId),i=n.readSync(r.dataId),o=N.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var P9={kernelName:Jh,backendName:"webgl",kernelFunc:M9},O9="return float(a != b);",J_=pn({opSnippet:O9,cpuKernelImpl:c7,dtype:"bool"}),L9={kernelName:tu,backendName:"webgl",kernelFunc:J_};function rd(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.texData.get(a.dataId);return na({inputs:{x:r.complexTensorInfos.real},backend:n})}var z9={kernelName:wm,backendName:"webgl",kernelFunc:rd},W9="return float(int(x));";function B9(e,t){let n=new Sr(e.shape,W9),a=t.runWebGLProgram(n,[e],"int32");return{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}function cx(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dtype:s}=a;if(s==="complex64"){if(r.dtype==="complex64")return na({inputs:{x:r},backend:n});let i=It(r.shape),o=cx({inputs:{x:r},backend:n,attrs:{dtype:"float32"}}),l=Is({inputs:{real:o,imag:i},backend:n});return i.dispose(),n.disposeIntermediateTensorInfo(o),l}if(r.dtype==="complex64"){let i=rd({inputs:{input:r},backend:n}),o=cx({inputs:{x:i},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(i),o}if(!v.hasEncodingLoss(r.dtype,s)){let i=na({inputs:{x:r},backend:n});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(n.shouldExecuteOnCPU([r])){let i=n.texData.get(r.dataId).values,[o,l,u]=HZ(i,r.shape,r.dtype,s);return n.makeTensorInfo(o,l,u)}if(s==="int32")return B9(r,n);if(s==="bool"){let i=n.makeTensorInfo([],"bool",v.getTypedArrayFromDType("bool",1)),o=J_({inputs:{a:r,b:i},backend:n});return n.disposeIntermediateTensorInfo(i),o}throw new Error(`Error in Cast: failed to cast ${r.dtype} to ${s}`)}var V9={kernelName:bi,backendName:"webgl",kernelFunc:cx},aI="return ceil(x);",U9=Ye({opSnippet:aI,packedOpSnippet:aI,cpuKernelImpl:jZ}),G9={kernelName:xi,backendName:"webgl",kernelFunc:U9},H9=class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function shallowSlice(x, begin, size, backend2) {
+  const xTexData = backend2.texData.get(x.dataId);
+  const t = backend2.makeTensorInfo(size, x.dtype);
+  const newTexData = backend2.texData.get(t.dataId);
+  Object.assign(newTexData, xTexData);
+  newTexData.refCount = 1;
+  newTexData.shape = size;
+  newTexData.dtype = x.dtype;
+  let flatOffset = slice_util_exports.computeFlatOffset(begin, util_exports.computeStrides(x.shape));
+  if (xTexData.slice) {
+    flatOffset += xTexData.slice.flatOffset;
+  }
+  newTexData.slice = {
+    flatOffset,
+    origDataId: xTexData.slice && xTexData.slice.origDataId || x.dataId
+  };
+  const refCount = backend2.dataRefCount.get(newTexData.slice.origDataId) || 1;
+  backend2.dataRefCount.set(newTexData.slice.origDataId, refCount + 1);
+  return t;
+}
+function slice3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, size } = attrs;
+  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);
+  slice_util_exports.assertParamsValid(x, $begin, $size);
+  if (util_exports.sizeFromShape($size) === 0) {
+    return backend2.makeTensorInfo($size, x.dtype, []);
+  }
+  if (backend2.shouldExecuteOnCPU([x]) || x.dtype === "string") {
+    const xTexData = backend2.texData.get(x.dataId);
+    const outValues = sliceImplCPU(xTexData.values, $begin, $size, x.shape, x.dtype);
+    return backend2.makeTensorInfo($size, x.dtype, outValues);
+  }
+  const { isPacked } = backend2.texData.get(x.dataId);
+  const isContinous = slice_util_exports.isSliceContinous(x.shape, $begin, $size);
+  if (isPacked || !isContinous) {
+    const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new SlicePackedProgram($size) : new SliceProgram($size);
+    const customValues = [$begin];
+    return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+  }
+  backend2.uploadToGPU(x.dataId);
+  return shallowSlice(x, $begin, $size, backend2);
+}
+var sliceConfig2 = {
+  kernelName: Slice,
+  backendName: "webgl",
+  kernelFunc: slice3
+};
+var batchToSpaceND3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  util_exports.assert(x.shape.length <= 4, () => "batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const toDispose = [];
+  const reshapedIntermediate = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const transposedIntermediate = transpose3({ inputs: { x: reshapedIntermediate }, backend: backend2, attrs: { perm: permuted } });
+  const reshapedIntermediate2 = reshape4({
+    inputs: { x: transposedIntermediate },
+    backend: backend2,
+    attrs: { shape: reshapedPermuted }
+  });
+  const sliced = slice3({
+    inputs: { x: reshapedIntermediate2 },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  toDispose.push(reshapedIntermediate);
+  toDispose.push(transposedIntermediate);
+  toDispose.push(reshapedIntermediate2);
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return sliced;
+};
+var batchToSpaceNDConfig2 = {
+  kernelName: BatchToSpaceND,
+  backendName: "webgl",
+  kernelFunc: batchToSpaceND3
+};
+function bincount3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size } = attrs;
+  const xVals = backend2.readSync(x.dataId);
+  const weightsVals = backend2.readSync(weights.dataId);
+  const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);
+  return backend2.makeTensorInfo([size], weights.dtype, outVals);
+}
+var bincountConfig2 = {
+  kernelName: Bincount,
+  backendName: "webgl",
+  kernelFunc: bincount3
+};
+function broadcastArgs3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { s0, s1 } = inputs;
+  const s0Vals = backend2.readSync(s0.dataId);
+  const s1Vals = backend2.readSync(s1.dataId);
+  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));
+  return backend2.makeTensorInfo([broadcastShape.length], "int32", Int32Array.from(broadcastShape));
+}
+var broadcastArgsConfig2 = {
+  kernelName: BroadcastArgs,
+  backendName: "webgl",
+  kernelFunc: broadcastArgs3
+};
+var NOT_EQUAL = `return float(a != b);`;
+var notEqual3 = binaryKernelFunc2({ opSnippet: NOT_EQUAL, cpuKernelImpl: notEqualImplCPU, dtype: "bool" });
+var notEqualConfig2 = {
+  kernelName: NotEqual,
+  backendName: "webgl",
+  kernelFunc: notEqual3
+};
+function real3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputData = backend2.texData.get(input2.dataId);
+  return identity3({ inputs: { x: inputData.complexTensorInfos.real }, backend: backend2 });
+}
+var realConfig2 = {
+  kernelName: Real,
+  backendName: "webgl",
+  kernelFunc: real3
+};
+var TO_INT = `return float(int(x));`;
+function int(input2, backend2) {
+  const program = new UnaryOpProgram(input2.shape, TO_INT);
+  const output = backend2.runWebGLProgram(program, [input2], "int32");
+  return { dataId: output.dataId, shape: output.shape, dtype: output.dtype };
+}
+function cast4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dtype } = attrs;
+  if (dtype === "complex64") {
+    if (x.dtype === "complex64") {
+      return identity3({ inputs: { x }, backend: backend2 });
+    }
+    const zerosTensor = zeros(x.shape);
+    const floatX = cast4({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+    const result = complex3({ inputs: { real: floatX, imag: zerosTensor }, backend: backend2 });
+    zerosTensor.dispose();
+    backend2.disposeIntermediateTensorInfo(floatX);
+    return result;
+  }
+  if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const result = cast4({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    return result;
+  }
+  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {
+    const result = identity3({ inputs: { x }, backend: backend2 });
+    return { dataId: result.dataId, shape: result.shape, dtype };
+  }
+  if (backend2.shouldExecuteOnCPU([x])) {
+    const values = backend2.texData.get(x.dataId).values;
+    const [resultShape, resultType, resultData] = castImplCPU(values, x.shape, x.dtype, dtype);
+    return backend2.makeTensorInfo(resultShape, resultType, resultData);
+  }
+  if (dtype === "int32") {
+    return int(x, backend2);
+  }
+  if (dtype === "bool") {
+    const zerosTensorInfo = backend2.makeTensorInfo([], "bool", util_exports.getTypedArrayFromDType("bool", 1));
+    const binaryInputs = { a: x, b: zerosTensorInfo };
+    const result = notEqual3({ inputs: binaryInputs, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);
+    return result;
+  }
+  throw new Error(`Error in Cast: failed to cast ${x.dtype} to ${dtype}`);
+}
+var castConfig2 = {
+  kernelName: Cast,
+  backendName: "webgl",
+  kernelFunc: cast4
+};
+var CEIL = `return ceil(x);`;
+var ceil3 = unaryKernelFunc2({ opSnippet: CEIL, packedOpSnippet: CEIL, cpuKernelImpl: ceilImplCPU });
+var ceilConfig2 = {
+  kernelName: Ceil,
+  backendName: "webgl",
+  kernelFunc: ceil3
+};
+var ClipProgram = class {
+  constructor(aShape) {
+    this.variableNames = ["A"];
+    this.customUniforms = [
+      { name: "minVal", type: "float" },
+      { name: "maxVal", type: "float" }
+    ];
+    this.outputShape = aShape;
+    this.userCode = `
 
       void main() {
         float value = getAAtOutCoords();
@@ -2122,7 +54737,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         setOutput(clamp(value, minVal, maxVal));
       }
-    `}},j9=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+var ClipPackedProgram = class {
+  constructor(aShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "minVal", type: "float" },
+      { name: "maxVal", type: "float" }
+    ];
+    this.outputShape = aShape;
+    this.userCode = `
       void main() {
         vec4 value = getAAtOutCoords();
 
@@ -2133,7 +54761,32 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         setOutput(clamp(value, vec4(minVal), vec4(maxVal)));
       }
-    `}};function q9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{clipValueMin:s,clipValueMax:i}=a,o;H().getBool("WEBGL_PACK_CLIP")?o=new j9(r.shape):o=new H9(r.shape);let l=[[s],[i]];return n.runWebGLProgram(o,[r],r.dtype,l)}var K9={kernelName:ds,backendName:"webgl",kernelFunc:q9},X9=class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function clipByValue3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { clipValueMin, clipValueMax } = attrs;
+  let program;
+  if (env().getBool("WEBGL_PACK_CLIP")) {
+    program = new ClipPackedProgram(x.shape);
+  } else {
+    program = new ClipProgram(x.shape);
+  }
+  const customValues = [[clipValueMin], [clipValueMax]];
+  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+}
+var clipByValueConfig2 = {
+  kernelName: ClipByValue,
+  backendName: "webgl",
+  kernelFunc: clipByValue3
+};
+var ComplexAbsProgram = class {
+  constructor(shape) {
+    this.variableNames = ["real", "imag"];
+    this.outputShape = shape;
+    this.userCode = `
       void main() {
         float re = abs(getRealAtOutCoords());
         float im = abs(getImagAtOutCoords());
@@ -2146,96 +54799,324 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))
         );
       }
-    `}};function rI(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}function Y9(e){let{inputs:t,backend:n}=e,{x:a}=t,r=n.texData.get(a.dataId),s=new X9(a.shape),i=[rI(a,r.complexTensorInfos.real),rI(a,r.complexTensorInfos.imag)];return n.runWebGLProgram(s,i,i[0].dtype)}var Z9={kernelName:oc,backendName:"webgl",kernelFunc:Y9},J9=class{constructor(e){this.outputShape=[],this.outputShape=N.computeOutShape(e,1),this.variableNames=e.map((s,i)=>`T${i}`);let t=new Array(e.length-1);t[0]=e[0][1];for(let s=1;s<t.length;s++)t[s]=t[s-1]+e[s][1];let n=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let s=1;s<t.length;s++){let i=t[s-1];n.push(`else if (yC < ${t[s]}) setOutput(getT${s}(yR, yC-${i}));`)}let a=t.length,r=t[t.length-1];n.push(`else setOutput(getT${a}(yR, yC-${r}));`),this.userCode=`
+    `;
+  }
+};
+function makeComplexComponentTensorInfo(complexTensor, complexPart) {
+  return {
+    dataId: complexPart.dataId,
+    dtype: complexPart.dtype,
+    shape: complexTensor.shape
+  };
+}
+function complexAbs2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const xData = backend2.texData.get(x.dataId);
+  const program = new ComplexAbsProgram(x.shape);
+  const programInputs = [
+    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.real),
+    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.imag)
+  ];
+  return backend2.runWebGLProgram(program, programInputs, programInputs[0].dtype);
+}
+var complexAbsConfig2 = {
+  kernelName: ComplexAbs,
+  backendName: "webgl",
+  kernelFunc: complexAbs2
+};
+var ConcatProgram = class {
+  constructor(shapes) {
+    this.outputShape = [];
+    this.outputShape = backend_util_exports.computeOutShape(shapes, 1);
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const offsets = new Array(shapes.length - 1);
+    offsets[0] = shapes[0][1];
+    for (let i = 1; i < offsets.length; i++) {
+      offsets[i] = offsets[i - 1] + shapes[i][1];
+    }
+    const snippets = [`if (yC < ${offsets[0]}) setOutput(getT0(yR, yC));`];
+    for (let i = 1; i < offsets.length; i++) {
+      const shift = offsets[i - 1];
+      snippets.push(`else if (yC < ${offsets[i]}) setOutput(getT${i}(yR, yC-${shift}));`);
+    }
+    const lastIndex = offsets.length;
+    const lastShift = offsets[offsets.length - 1];
+    snippets.push(`else setOutput(getT${lastIndex}(yR, yC-${lastShift}));`);
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int yR = coords.x;
         int yC = coords.y;
 
-        ${n.join(`
-        `)}
+        ${snippets.join("\n        ")}
       }
-    `}},Q9=class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=N.computeOutShape(e,t);let n=this.outputShape,a=n.length,r=gt(a),s=wn("coords",a),i=["x","y","z","w","u","v"].slice(0,a);this.variableNames=e.map((m,f)=>`T${f}`);let o=new Array(e.length-1);o[0]=e[0][t];for(let m=1;m<o.length;m++)o[m]=o[m-1]+e[m][t];let l=i[t],u=i.slice(-2),p=i.join(),d=`if (${l} < ${o[0]}) {
+    `;
+  }
+};
+var ConcatPackedProgram = class {
+  constructor(shapes, axis) {
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    this.outputShape = backend_util_exports.computeOutShape(shapes, axis);
+    const shape = this.outputShape;
+    const rank = shape.length;
+    const dtype = getCoordsDataType(rank);
+    const coords2 = getChannels("coords", rank);
+    const channels = ["x", "y", "z", "w", "u", "v"].slice(0, rank);
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const offsets = new Array(shapes.length - 1);
+    offsets[0] = shapes[0][axis];
+    for (let i = 1; i < offsets.length; i++) {
+      offsets[i] = offsets[i - 1] + shapes[i][axis];
+    }
+    const channel = channels[axis];
+    const lastChannels = channels.slice(-2);
+    const allChannels = channels.join();
+    let getValueSnippet = `if (${channel} < ${offsets[0]}) {
         return getChannel(
-            getT0(${p}), vec2(${u.join()}));
-        }`;for(let m=1;m<o.length;m++){let f=o[m-1];d+=`
-        if (${l} < ${o[m]}  && ${l} >= ${o[m-1]}) {
+            getT0(${allChannels}), vec2(${lastChannels.join()}));
+        }`;
+    for (let i = 1; i < offsets.length; i++) {
+      const shift2 = offsets[i - 1];
+      getValueSnippet += `
+        if (${channel} < ${offsets[i]}  && ${channel} >= ${offsets[i - 1]}) {
           return getChannel(
-            getT${m}(${nh(i,l,f)}),
-            vec2(${nh(u,l,f)}));
-        }`}let c=o.length,h=o[o.length-1];d+=`
+            getT${i}(${shiftedChannels(channels, channel, shift2)}),
+            vec2(${shiftedChannels(lastChannels, channel, shift2)}));
+        }`;
+    }
+    const lastIndex = offsets.length;
+    const shift = offsets[offsets.length - 1];
+    getValueSnippet += `
         return getChannel(
-          getT${c}(${nh(i,l,h)}),
-          vec2(${nh(u,l,h)}));`,this.userCode=`
-      float getValue(${i.map(m=>"int "+m)}) {
-        ${d}
+          getT${lastIndex}(${shiftedChannels(channels, channel, shift)}),
+          vec2(${shiftedChannels(lastChannels, channel, shift)}));`;
+    this.userCode = `
+      float getValue(${channels.map((x) => "int " + x)}) {
+        ${getValueSnippet}
       }
 
       void main() {
-        ${r} coords = getOutputCoords();
-        vec4 result = vec4(getValue(${s}), 0., 0., 0.);
+        ${dtype} coords = getOutputCoords();
+        vec4 result = vec4(getValue(${coords2}), 0., 0., 0.);
 
-        ${s[a-1]} = ${s[a-1]} + 1;
-        if (${s[a-1]} < ${n[a-1]}) {
-          result.g = getValue(${s});
+        ${coords2[rank - 1]} = ${coords2[rank - 1]} + 1;
+        if (${coords2[rank - 1]} < ${shape[rank - 1]}) {
+          result.g = getValue(${coords2});
         }
 
-        ${s[a-2]} = ${s[a-2]} + 1;
-        if (${s[a-2]} < ${n[a-2]}) {
-          result.a = getValue(${s});
+        ${coords2[rank - 2]} = ${coords2[rank - 2]} + 1;
+        if (${coords2[rank - 2]} < ${shape[rank - 2]}) {
+          result.a = getValue(${coords2});
         }
 
-        ${s[a-1]} = ${s[a-1]} - 1;
-        if (${s[a-2]} < ${n[a-2]} &&
-            ${s[a-1]} < ${n[a-1]}) {
-          result.b = getValue(${s});
+        ${coords2[rank - 1]} = ${coords2[rank - 1]} - 1;
+        if (${coords2[rank - 2]} < ${shape[rank - 2]} &&
+            ${coords2[rank - 1]} < ${shape[rank - 1]}) {
+          result.b = getValue(${coords2});
         }
         setOutput(result);
       }
-    `}};function nh(e,t,n){let a=e.indexOf(t);return e.map((r,s)=>s===a?`${r} - ${n}`:r).join()}function Of(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.texData.get(a.dataId);return na({inputs:{x:r.complexTensorInfos.imag},backend:n})}var eQ={kernelName:cm,backendName:"webgl",kernelFunc:Of};function Pp(e,t,n){let a=e[0].dtype;if(a==="complex64"){let d=e.map(g=>rd({inputs:{input:g},backend:n})),c=e.map(g=>Of({inputs:{input:g},backend:n})),h=Pp(d,t,n),m=Pp(c,t,n),f=Is({inputs:{real:h,imag:m},backend:n});return d.forEach(g=>n.disposeIntermediateTensorInfo(g)),c.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}let r=n.shouldExecuteOnCPU(e);if(a==="string"&&(r=!0),r){let d=e.map(b=>{let x=v.sizeFromShape(b.shape.slice(t));return de({inputs:{x:b},backend:n,attrs:{shape:[-1,x]}})}),c=d.map(b=>({vals:n.readSync(b.dataId),shape:b.shape})),h=N.computeOutShape(d.map(b=>b.shape),1),m=d[0].shape[0]===1,f=qZ(c,h,a,m),g=N.computeOutShape(e.map(b=>b.shape),t),y=n.makeTensorInfo(g,a,f);return d.forEach(b=>n.disposeIntermediateTensorInfo(b)),y}let s=H().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");if(e.length>s){let d=[];for(let h=0;h<e.length;h+=s){let m=e.slice(h,h+s);d.push(Pp(m,t,n))}let c=Pp(d,t,n);for(let h of d)n.disposeIntermediateTensorInfo(h);return c}if(H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&e[0].shape.length>1){let d=new Q9(e.map(c=>c.shape),t);return n.runWebGLProgram(d,e,a)}let{tensors2D:i,outShape:o}=tQ(e,t,n),l=new J9(i.map(d=>d.shape)),u=n.runWebGLProgram(l,i,a);i.forEach(d=>n.disposeIntermediateTensorInfo(d));let p=de({inputs:{x:u},attrs:{shape:o},backend:n});return n.disposeIntermediateTensorInfo(u),p}function tQ(e,t,n){let a=N.computeOutShape(e.map(r=>r.shape),t);return{tensors2D:e.map(r=>de({inputs:{x:r},attrs:{shape:[-1,v.sizeFromShape(r.shape.slice(t))]},backend:n})),outShape:a}}function Q_(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a,s=v.parseAxisParam(r,t[0].shape)[0],i=t.map(u=>u.shape);N.assertParamsConsistent(i,s);let o=N.computeOutShape(t.map(u=>u.shape),s);if(v.sizeFromShape(o)===0)return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(u=>v.sizeFromShape(u.shape)>0);return l.length===1?na({inputs:{x:l[0]},backend:n}):Pp(l,s,n)}var nQ={kernelName:Fl,backendName:"webgl",kernelFunc:Q_},eE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;let s=e.padInfo.top,i=e.padInfo.left,o=e.strideHeight,l=e.strideWidth,u=e.dilationHeight,p=e.dilationWidth,d=e.filterHeight,c=e.filterWidth,h=Math.floor(e.inChannels/4)*4,m=e.inChannels%4,f=e.dataFormat==="channelsLast",g=f?1:2,y=f?2:3,b=f?3:1,x="",w="";n&&(a?x=`float activation(float a) {
+    `;
+  }
+};
+function shiftedChannels(channels, channel, shift) {
+  const channelIdx = channels.indexOf(channel);
+  const res = channels.map((c, idx) => {
+    if (idx === channelIdx) {
+      return `${c} - ${shift}`;
+    } else {
+      return c;
+    }
+  });
+  return res.join();
+}
+function imag3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputData = backend2.texData.get(input2.dataId);
+  return identity3({ inputs: { x: inputData.complexTensorInfos.imag }, backend: backend2 });
+}
+var imagConfig2 = {
+  kernelName: Imag,
+  backendName: "webgl",
+  kernelFunc: imag3
+};
+function concatImpl2(inputs, axis, backend2) {
+  const dtype = inputs[0].dtype;
+  if (dtype === "complex64") {
+    const reals = inputs.map((t) => real3({ inputs: { input: t }, backend: backend2 }));
+    const imags = inputs.map((t) => imag3({ inputs: { input: t }, backend: backend2 }));
+    const realConcated = concatImpl2(reals, axis, backend2);
+    const imagConcated = concatImpl2(imags, axis, backend2);
+    const result2 = complex3({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });
+    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));
+    backend2.disposeIntermediateTensorInfo(realConcated);
+    backend2.disposeIntermediateTensorInfo(imagConcated);
+    return result2;
+  }
+  let runOnCpu = backend2.shouldExecuteOnCPU(inputs);
+  if (dtype === "string") {
+    runOnCpu = true;
+  }
+  if (runOnCpu) {
+    const tensors2D2 = inputs.map((t) => {
+      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));
+      const shape = [-1, innerSize];
+      return reshape4({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+    });
+    const inputsValShapes = tensors2D2.map((t) => {
+      return { vals: backend2.readSync(t.dataId), shape: t.shape };
+    });
+    const outShape2 = backend_util_exports.computeOutShape(tensors2D2.map((t) => t.shape), 1);
+    const simplyConcat = tensors2D2[0].shape[0] === 1;
+    const outVals = concatImplCPU(inputsValShapes, outShape2, dtype, simplyConcat);
+    const finalOutShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+    const outInfo = backend2.makeTensorInfo(finalOutShape, dtype, outVals);
+    tensors2D2.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return outInfo;
+  }
+  const maxTexturesInShader = env().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");
+  if (inputs.length > maxTexturesInShader) {
+    const reducedInputs = [];
+    for (let i = 0; i < inputs.length; i += maxTexturesInShader) {
+      const subArray = inputs.slice(i, i + maxTexturesInShader);
+      reducedInputs.push(concatImpl2(subArray, axis, backend2));
+    }
+    const result2 = concatImpl2(reducedInputs, axis, backend2);
+    for (const i of reducedInputs) {
+      backend2.disposeIntermediateTensorInfo(i);
+    }
+    return result2;
+  }
+  if (env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") && inputs[0].shape.length > 1) {
+    const program2 = new ConcatPackedProgram(inputs.map((t) => t.shape), axis);
+    return backend2.runWebGLProgram(program2, inputs, dtype);
+  }
+  const { tensors2D, outShape } = computeTensors2D(inputs, axis, backend2);
+  const program = new ConcatProgram(tensors2D.map((t) => t.shape));
+  const result = backend2.runWebGLProgram(program, tensors2D, dtype);
+  tensors2D.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+  const reshapedResult = reshape4({ inputs: { x: result }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(result);
+  return reshapedResult;
+}
+function computeTensors2D(inputs, axis, backend2) {
+  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+  const tensors2D = inputs.map((x) => reshape4({
+    inputs: { x },
+    attrs: { shape: [-1, util_exports.sizeFromShape(x.shape.slice(axis))] },
+    backend: backend2
+  }));
+  return { tensors2D, outShape };
+}
+function concat3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, $axis);
+  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);
+  }
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity3({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  return concatImpl2($inputs, $axis, backend2);
+}
+var concatConfig2 = {
+  kernelName: Concat,
+  backendName: "webgl",
+  kernelFunc: concat3
+};
+var Conv2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivationWeights = false, hasLeakyreluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;
+    const inputDepthVec4Remainder = convInfo.inChannels % 4;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const channelDim = isChannelsLast ? 3 : 1;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivationWeights) {
+        activationSnippet = `float activation(float a) {
           float b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:r?x=`float activation(float a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyreluAlpha) {
+        activationSnippet = `float activation(float a) {
           float b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:x=`
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `
           float activation(float x) {
-            ${n}
+            ${activation2}
           }
-        `,w="result = activation(result);");let I=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${x}
+        `;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivationWeights) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyreluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
 
-      const ivec2 strides = ivec2(${o}, ${l});
-      const ivec2 pads = ivec2(${s}, ${i});
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d2 = coords[${b}];
+        int d2 = coords[${channelDim}];
 
         ivec2 xRCCorner =
-            ivec2(coords[${g}], coords[${y}]) * strides - pads;
+            ivec2(coords[${rowDim}], coords[${colDim}]) * strides - pads;
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${d}; wR++) {
-          int xR = xRCorner + wR * ${u};
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          int xR = xRCorner + wR * ${dilationHeight};
 
-          if (xR < 0 || xR >= ${e.inHeight}) {
+          if (xR < 0 || xR >= ${convInfo.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${c}; wC++) {
-            int xC = xCCorner + wC * ${p};
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            int xC = xCCorner + wC * ${dilationWidth};
 
-            if (xC < 0 || xC >= ${e.inWidth}) {
+            if (xC < 0 || xC >= ${convInfo.inWidth}) {
               continue;
             }
 
-            for (int d1 = 0; d1 < ${h}; d1 += 4) {
+            for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {
               vec4 wValues = vec4(
                 getW(wR, wC, d1, d2),
                 getW(wR, wC, d1 + 1, d2),
@@ -2243,7 +55124,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                 getW(wR, wC, d1 + 3, d2)
               );
 
-              if (${f}) {
+              if (${isChannelsLast}) {
                 vec4 xValues = vec4(
                   getX(batch, xR, xC, d1),
                   getX(batch, xR, xC, d1 + 1),
@@ -2262,57 +55143,57 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
               }
             }
 
-            if (${m===1}) {
+            if (${inputDepthVec4Remainder === 1}) {
 
-              if (${f}) {
+              if (${isChannelsLast}) {
                 dotProd +=
-                    getX(batch, xR, xC, ${h}) *
-                    getW(wR, wC, ${h}, d2);
+                    getX(batch, xR, xC, ${inputDepthNearestVec4}) *
+                    getW(wR, wC, ${inputDepthNearestVec4}, d2);
               } else {
                 dotProd +=
-                    getX(batch, ${h}, xR, xC) *
-                    getW(wR, wC, ${h}, d2);
+                    getX(batch, ${inputDepthNearestVec4}, xR, xC) *
+                    getW(wR, wC, ${inputDepthNearestVec4}, d2);
               }
 
-            } else if (${m===2}) {
+            } else if (${inputDepthVec4Remainder === 2}) {
               vec2 wValues = vec2(
-                getW(wR, wC, ${h}, d2),
-                getW(wR, wC, ${h} + 1, d2)
+                getW(wR, wC, ${inputDepthNearestVec4}, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2)
               );
 
-              if (${f}) {
+              if (${isChannelsLast}) {
                 vec2 xValues = vec2(
-                  getX(batch, xR, xC, ${h}),
-                  getX(batch, xR, xC, ${h} + 1)
+                  getX(batch, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1)
                 );
                 dotProd += dot(xValues, wValues);
               } else {
                 vec2 xValues = vec2(
-                  getX(batch, ${h}, xR, xC),
-                  getX(batch, ${h} + 1, xR, xC)
+                  getX(batch, ${inputDepthNearestVec4}, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC)
                 );
                 dotProd += dot(xValues, wValues);
               }
 
-            } else if (${m===3}) {
+            } else if (${inputDepthVec4Remainder === 3}) {
               vec3 wValues = vec3(
-                getW(wR, wC, ${h}, d2),
-                getW(wR, wC, ${h} + 1, d2),
-                getW(wR, wC, ${h} + 2, d2)
+                getW(wR, wC, ${inputDepthNearestVec4}, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 2, d2)
               );
 
-              if (${f}) {
+              if (${isChannelsLast}) {
                 vec3 xValues = vec3(
-                  getX(batch, xR, xC, ${h}),
-                  getX(batch, xR, xC, ${h} + 1),
-                  getX(batch, xR, xC, ${h} + 2)
+                  getX(batch, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 2)
                 );
                 dotProd += dot(xValues, wValues);
               } else {
                 vec3 xValues = vec3(
-                  getX(batch, ${h}, xR, xC),
-                  getX(batch, ${h} + 1, xR, xC),
-                  getX(batch, ${h} + 2, xR, xC)
+                  getX(batch, ${inputDepthNearestVec4}, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 2, xR, xC)
                 );
                 dotProd += dot(xValues, wValues);
               }
@@ -2322,13 +55203,34 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
 
         float result = dotProd;
-        ${I}
-        ${w}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}},aQ=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let t=e.padInfo.front,n=e.padInfo.top,a=e.padInfo.left,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,p=e.filterDepth,d=e.filterHeight,c=e.filterWidth,h=Math.floor(e.inChannels/4)*4,m=e.inChannels%4;this.userCode=`
-      const ivec3 strides = ivec3(${r}, ${s}, ${i});
-      const ivec3 pads = ivec3(${t}, ${n}, ${a});
+    `;
+  }
+};
+var Conv3DProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;
+    const inputDepthVec4Remainder = convInfo.inChannels % 4;
+    this.userCode = `
+      const ivec3 strides = ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2344,28 +55246,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // y(yF, yR, yC, d2). ? = to be determined. : = across all
         // values in that axis.
         float dotProd = 0.0;
-        for (int wF = 0; wF < ${p}; wF++) {
-          int xF = xFCorner + wF * ${o};
+        for (int wF = 0; wF < ${filterDepth}; wF++) {
+          int xF = xFCorner + wF * ${dilationDepth};
 
-          if (xF < 0 || xF >= ${e.inDepth}) {
+          if (xF < 0 || xF >= ${convInfo.inDepth}) {
             continue;
           }
 
-          for (int wR = 0; wR < ${d}; wR++) {
-            int xR = xRCorner + wR * ${l};
+          for (int wR = 0; wR < ${filterHeight}; wR++) {
+            int xR = xRCorner + wR * ${dilationHeight};
 
-            if (xR < 0 || xR >= ${e.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${c}; wC++) {
-              int xC = xCCorner + wC * ${u};
+            for (int wC = 0; wC < ${filterWidth}; wC++) {
+              int xC = xCCorner + wC * ${dilationWidth};
 
-              if (xC < 0 || xC >= ${e.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
-              for (int d1 = 0; d1 < ${h}; d1 += 4) {
+              for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {
                 vec4 xValues = vec4(
                   getX(batch, xF, xR, xC, d1),
                   getX(batch, xF, xR, xC, d1 + 1),
@@ -2382,30 +55284,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                 dotProd += dot(xValues, wValues);
               }
 
-              if (${m===1}) {
+              if (${inputDepthVec4Remainder === 1}) {
                 dotProd +=
-                  getX(batch, xF, xR, xC, ${h}) *
-                  getW(wF, wR, wC, ${h}, d2);
-              } else if (${m===2}) {
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}) *
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2);
+              } else if (${inputDepthVec4Remainder === 2}) {
                 vec2 xValues = vec2(
-                  getX(batch, xF, xR, xC, ${h}),
-                  getX(batch, xF, xR, xC, ${h} + 1)
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1)
                 );
                 vec2 wValues = vec2(
-                  getW(wF, wR, wC, ${h}, d2),
-                  getW(wF, wR, wC, ${h} + 1, d2)
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2)
                 );
                 dotProd += dot(xValues, wValues);
-              } else if (${m===3}) {
+              } else if (${inputDepthVec4Remainder === 3}) {
                 vec3 xValues = vec3(
-                  getX(batch, xF, xR, xC, ${h}),
-                  getX(batch, xF, xR, xC, ${h} + 1),
-                  getX(batch, xF, xR, xC, ${h} + 2)
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 2)
                 );
                 vec3 wValues = vec3(
-                  getW(wF, wR, wC, ${h}, d2),
-                  getW(wF, wR, wC, ${h} + 1, d2),
-                  getW(wF, wR, wC, ${h} + 2, d2)
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 2, d2)
                 );
                 dotProd += dot(xValues, wValues);
               }
@@ -2414,41 +55316,82 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},tE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=_n(this.outputShape.length);let s=e.padInfo.left,i=e.strideWidth,o=e.dilationWidth,l=e.filterHeight,u=e.filterWidth,p=u,d=`
+    `;
+  }
+};
+var Conv2DPackedProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const padLeft = convInfo.padInfo.left;
+    const strideWidth = convInfo.strideWidth;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const texelsAcross = filterWidth;
+    let mainLoop = `
        int xR; int xC; int xCOffset;
-       vec4 wTexel; vec4 previous; vec4 final;`;for(let f=0;f<u;f++)d+=`
-           vec4 xTexelC${f*2};
-           int xTexelC${f*2}Ready;
-           vec4 xTexelC${f*2+1};
-           int xTexelC${f*2+1}Ready;
-           vec4 xC${f};`;d+=`
-     for (int r = 0; r < ${l}; r++) {
-      for (int d1 = 0; d1 < ${e.inChannels}; d1 += 2) {
-       `;for(let f=0;f<u;f++)d+=`
-           xTexelC${f*2} = vec4(0.0);
-           xTexelC${f*2}Ready = 0;
-           xTexelC${f*2+1} = vec4(0.0);
-           xTexelC${f*2+1}Ready = 0;
-           xC${f} = vec4(0.0);`;d+=`
+       vec4 wTexel; vec4 previous; vec4 final;`;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+           vec4 xTexelC${c * 2};
+           int xTexelC${c * 2}Ready;
+           vec4 xTexelC${c * 2 + 1};
+           int xTexelC${c * 2 + 1}Ready;
+           vec4 xC${c};`;
+    }
+    mainLoop += `
+     for (int r = 0; r < ${filterHeight}; r++) {
+      for (int d1 = 0; d1 < ${convInfo.inChannels}; d1 += 2) {
+       `;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+           xTexelC${c * 2} = vec4(0.0);
+           xTexelC${c * 2}Ready = 0;
+           xTexelC${c * 2 + 1} = vec4(0.0);
+           xTexelC${c * 2 + 1}Ready = 0;
+           xC${c} = vec4(0.0);`;
+    }
+    mainLoop += `
          xR = xRCorner + r * dilations[0];
          if (xR >=0 && xR < inDims[0]) {
-       `;for(let f=0;f<(p+1)/2;f++){let g=f*2;if(d+=`
-           xC = xCCorner + ${g*o};
-           `,i===1){if(g<u&&(s%2===1?(d+=`
+       `;
+    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {
+      const colIndex = texelC * 2;
+      mainLoop += `
+           xC = xCCorner + ${colIndex * dilationWidth};
+           `;
+      if (strideWidth === 1) {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                  xCOffset = xC + 1;
-                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g}Ready == 0) {
-                   xTexelC${g} = getX(batch, xR, xCOffset, d1);
+                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
 
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${g}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${g}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
-               `,o===1&&g>0?d+=`
-                 xC${g} = vec4(xTexelC${g-2}.zw, xTexelC${g}.xy);
-                 `:d+=`
+               `;
+            if (dilationWidth === 1 && colIndex > 0) {
+              mainLoop += `
+                 xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);
+                 `;
+            } else {
+              mainLoop += `
                    xCOffset = xC + 1 - 2;
 
                    if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -2460,137 +55403,206 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                        previous.zw = vec2(0.0);
                      }
 
-                     xC${g} = vec4(previous.zw, xTexelC${g}.xy);
+                     xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);
                    } else {
-                     xC${g} = vec4(0.0, 0.0, xTexelC${g}.xy);
+                     xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);
                    }
-                   `):d+=`
-                 if (xC >= 0 && xC < inDims[1] && xTexelC${g}Ready == 0) {
-                   xTexelC${g} = getX(batch, xR, xC, d1);
+                   `;
+            }
+          } else {
+            mainLoop += `
+                 if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
-                     xTexelC${g}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${g}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
-                 xC${g} = xTexelC${g};
-                 `,g+1<u)){let y=s%2===0?v.nearestLargerEven(o):o;o%2===0&&s%2===1||o%2!==0&&s%2!==1?(d+=`
-                   xCOffset = xC + imod(pads[1], 2) + ${y};
+                 xC${colIndex} = xTexelC${colIndex};
+                 `;
+          }
+          if (colIndex + 1 < filterWidth) {
+            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;
+            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {
+              mainLoop += `
+                   xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};
 
-                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g+1}Ready == 0) {
-                     xTexelC${g+1} = getX(batch, xR, xCOffset, d1);
+                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                     xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
 
                      // Need to manually clear unused channels in case
                      // we're reading from recycled texture.
                      if (xCOffset + 1 >= inDims[1]) {
-                       xTexelC${g+1}.zw = vec2(0.0);
+                       xTexelC${colIndex + 1}.zw = vec2(0.0);
                      }
-                     xTexelC${g+1}Ready = 1;
+                     xTexelC${colIndex + 1}Ready = 1;
                    }
-                   `,o>1?d+=`
+                   `;
+              if (dilationWidth > 1) {
+                mainLoop += `
                      xCOffset -= 2;
                      if (xCOffset >= 0 && xCOffset < inDims[1]) {
                       previous = getX(batch, xR, xCOffset, d1);
-                      xC${g+1} = vec4(previous.zw, xTexelC${g+1}.xy);
+                      xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);
                      } else {
-                      xC${g+1} = vec4(0.0, 0.0, xTexelC${g+1}.xy);
+                      xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);
                      }
-                     `:d+=`
-                     xC${g+1} = vec4(xTexelC${g}.zw, xTexelC${g+1}.xy);
-                     `):y===1?d+=`
-                     xC${g+1} = xTexelC${g};
-                     `:d+=`
-                     xCOffset = xC + ${y};
+                     `;
+              } else {
+                mainLoop += `
+                     xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);
+                     `;
+              }
+            } else {
+              if (nextTexelOffset === 1) {
+                mainLoop += `
+                     xC${colIndex + 1} = xTexelC${colIndex};
+                     `;
+              } else {
+                mainLoop += `
+                     xCOffset = xC + ${nextTexelOffset};
 
-                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g+1}Ready == 0) {
-                       xTexelC${g+1} = getX(batch, xR, xCOffset, d1);
+                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                       xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                        if (xCOffset + 1 >= inDims[1]) {
-                         xTexelC${g+1}.zw = vec2(0.0);
+                         xTexelC${colIndex + 1}.zw = vec2(0.0);
                        }
-                       xTexelC${g+1}Ready = 1;
+                       xTexelC${colIndex + 1}Ready = 1;
                      }
 
-                     xC${g+1} = xTexelC${g+1};
-                     `}}else g<u&&(s%2===1?(d+=`
+                     xC${colIndex + 1} = xTexelC${colIndex + 1};
+                     `;
+              }
+            }
+          }
+        }
+      } else {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                  xCOffset = xC + 1 - strides[1];
-                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g}Ready == 0) {
-                   xTexelC${g} = getX(batch, xR, xCOffset, d1);
+                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${g}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${g}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
-                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${g+1}Ready == 0) {
-                   xTexelC${g+1} = getX(batch, xR, xC + 1, d1);
+                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                   xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xC + 2 >= inDims[1]) {
-                     xTexelC${g+1}.zw = vec2(0.0);
+                     xTexelC${colIndex + 1}.zw = vec2(0.0);
                    }
-                   xTexelC${g+1}Ready = 1;
+                   xTexelC${colIndex + 1}Ready = 1;
                  }
 
-                 xC${g} = vec4(xTexelC${g}.zw, xTexelC${g+1}.zw);
-               `,g+1<u&&(d+=`
+                 xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+               `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
                    final = vec4(0.0);
                    xCOffset = xC + 1 + strides[1];
                    if(xCOffset >= 0 && xCOffset < inDims[1]) {
                      final = getX(batch, xR, xCOffset, d1);
                    }
-                   xC${g+1} = vec4(xTexelC${g+1}.xy, final.xy);
-                 `)):(d+=`
-                 if(xC >= 0 && xC < inDims[1] && xTexelC${g}Ready == 0) {
-                   xTexelC${g} = getX(batch, xR, xC, d1);
+                   xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);
+                 `;
+            }
+          } else {
+            mainLoop += `
+                 if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
-                     xTexelC${g}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${g}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
                  xCOffset = xC + strides[1];
-                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g+1}Ready == 0) {
-                   xTexelC${g+1} = getX(batch, xR, xCOffset, d1);
+                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                   xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${g+1}.zw = vec2(0.);
+                     xTexelC${colIndex + 1}.zw = vec2(0.);
                    }
-                   xTexelC${g+1}Ready = 1;
+                   xTexelC${colIndex + 1}Ready = 1;
                  }
 
-                 xC${g} = vec4(
-                   xTexelC${g}.xy, xTexelC${g+1}.xy);
-               `,g+1<u&&(d+=`
-                   xC${g+1} = vec4(xTexelC${g}.zw, xTexelC${g+1}.zw);
-                 `)));g<u&&(d+=`
-             wTexel = getW(r, ${g}, d1, d2);
-             dotProd += xC${g}.xxzz * vec4(wTexel.xy, wTexel.xy);
-             if(d1 + 1 < ${e.inChannels}) {
-               dotProd += xC${g}.yyww * vec4(wTexel.zw, wTexel.zw);
+                 xC${colIndex} = vec4(
+                   xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);
+               `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
+                   xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+                 `;
+            }
+          }
+        }
+      }
+      if (colIndex < filterWidth) {
+        mainLoop += `
+             wTexel = getW(r, ${colIndex}, d1, d2);
+             dotProd += xC${colIndex}.xxzz * vec4(wTexel.xy, wTexel.xy);
+             if(d1 + 1 < ${convInfo.inChannels}) {
+               dotProd += xC${colIndex}.yyww * vec4(wTexel.zw, wTexel.zw);
              }
-           `,g+1<u&&(d+=`
-               wTexel = getW(r, ${g+1}, d1, d2);
-               dotProd += xC${g+1}.xxzz * vec4(wTexel.xy, wTexel.xy);
-               if(d1 + 1 < ${e.inChannels}) {
-                 dotProd += xC${g+1}.yyww * vec4(wTexel.zw, wTexel.zw);
+           `;
+        if (colIndex + 1 < filterWidth) {
+          mainLoop += `
+               wTexel = getW(r, ${colIndex + 1}, d1, d2);
+               dotProd += xC${colIndex + 1}.xxzz * vec4(wTexel.xy, wTexel.xy);
+               if(d1 + 1 < ${convInfo.inChannels}) {
+                 dotProd += xC${colIndex + 1}.yyww * vec4(wTexel.zw, wTexel.zw);
                }
-             `))}d+=`
+             `;
+        }
+      }
+    }
+    mainLoop += `
      }
-   `,d+=`
+   `;
+    mainLoop += `
      }
-   `,d+=`
+   `;
+    mainLoop += `
      }
-   `;let c="",h="";n&&(a?c=`vec4 activation(vec4 a) {
+   `;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
            vec4 b = getPreluActivationWeightsAtOutCoords();
-           ${n}
-         }`:r?c=`vec4 activation(vec4 a) {
+           ${activation2}
+         }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `vec4 activation(vec4 a) {
            vec4 b = getLeakyreluAlphaAtOutCoords();
-           ${n}
-         }`:c=`vec4 activation(vec4 x) {
-           ${n}
-         }`,h="result = activation(result);");let m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-       ${c}
+           ${activation2}
+         }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+           ${activation2}
+         }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+       ${activationSnippet}
 
        void main() {
          ivec4 coords = getOutputCoords();
@@ -2603,22 +55615,50 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
          //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
          vec4 dotProd = vec4(0.000000000000001);
 
-         ${d}
+         ${mainLoop}
 
          vec4 result = dotProd - vec4(0.000000000000001);
-         ${m}
-         ${h}
+         ${addBiasSnippet}
+         ${applyActivationSnippet}
          setOutput(result);
        }
-     `}},rQ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec4"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=_n(this.outputShape.length);let{dataFormat:n}=t,a=Cn(),r=n==="channelsLast",s=r?1:2,i=r?2:3,o=this.enableShapeUniforms?"if(blockIndex < outShape[2] && pos < outShape[1]) {":`if(blockIndex < ${e[2]} && pos < ${e[1]}) {`,l="";for(let u=0;u<=1;u++)for(let p=0;p<=1;p++)l+=`
-          blockIndex = rc.z + ${p};
-          pos = rc.y + ${u};
+     `;
+  }
+};
+var Im2ColPackedProgram = class {
+  constructor(outputShape, convInfo) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "inputShape", type: "ivec4" },
+      { name: "pad", type: "ivec2" },
+      { name: "stride", type: "ivec2" },
+      { name: "dilation", type: "ivec2" },
+      { name: "inChannels", type: "int" },
+      { name: "itemsPerBlockRow", type: "int" },
+      { name: "outWidth", type: "int" }
+    ];
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const { dataFormat } = convInfo;
+    const glsl = getGlslDifferences();
+    const isChannelsLast = dataFormat === "channelsLast";
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const boundsCheckingSnippet = this.enableShapeUniforms ? "if(blockIndex < outShape[2] && pos < outShape[1]) {" : `if(blockIndex < ${outputShape[2]} && pos < ${outputShape[1]}) {`;
+    let unrolled = ``;
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        unrolled += `
+          blockIndex = rc.z + ${col};
+          pos = rc.y + ${row};
 
-          ${o}
+          ${boundsCheckingSnippet}
             offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];
             d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);
 
-            if(d0 < inputShape[${s}] && d0 >= 0) {
+            if(d0 < inputShape[${rowDim}] && d0 >= 0) {
               // Use custom imod instead mod. On Intel GPU, mod may generate
               // unexpected value.
               // https://github.com/tensorflow/tfjs/issues/5447
@@ -2626,25 +55666,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
               d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /
                   inChannels);
 
-              if(d1 < inputShape[${i}] && d1 >= 0) {
+              if(d1 < inputShape[${colDim}] && d1 >= 0) {
 
                 ch = imod(pos, inChannels);
 
-                if (${r}) {
+                if (${isChannelsLast}) {
                   innerDims = vec2(d1, ch);
-                  result[${u*2+p}] = getChannel(
+                  result[${row * 2 + col}] = getChannel(
                     getA(rc.x, d0, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 } else {
                   innerDims = vec2(d0, d1);
-                  result[${u*2+p}] = getChannel(
+                  result[${row * 2 + col}] = getChannel(
                     getA(rc.x, ch, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 }
               }
             }
           }
-        `;this.userCode=`
+        `;
+      }
+    }
+    this.userCode = `
       void main() {
         ivec3 rc = getOutputCoords();
 
@@ -2653,11 +55696,247 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         int blockIndex, pos, offsetY, d0, offsetX, d1, ch;
         vec2 innerDims;
 
-        ${l}
+        ${unrolled}
 
-        ${a.output} = result;
+        ${glsl.output} = result;
       }
-    `}};function Vh(e,t){let n=e.length;return n>=3?t?[...e.slice(0,-3),e[n-3]*e[n-2],e[n-1]]:[...e.slice(0,-3),e[n-3],e[n-2]*e[n-1]]:!t&&n===1&&e[0]>1?[e[0],1]:null}function nE({x:e,filter:t,convInfo:n,backend:a,bias:r=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l=e.shape,u=a.texData.get(e.dataId),p=n.inChannels,d=l[0]*l[1]*l[2],c=n.outChannels,h=n.dataFormat==="channelsLast",m=!1,f=!1,g,y=[];if(s!=null){let b=Vh(s.shape,h);b!=null&&(s=de({inputs:{x:s},backend:a,attrs:{shape:b}}),y.push(s))}if(r!=null){let b=Vh(r.shape,h);b!=null&&(r=de({inputs:{x:r},backend:a,attrs:{shape:b}}),y.push(r))}if(!((d===1||c===1)&&p>K_)&&u.isPacked&&h&&u.texture!=null&&l[2]%2!==0&&v.arraysEqual(u.shape.slice(-3),l.slice(-3))){let b=l[0]*l[1]*(l[2]+1),x={dataId:e.dataId,shape:[1,b,n.inChannels],dtype:e.dtype},w=u.shape;u.shape=u.shape.slice(),u.shape[u.shape.length-2]++,v.assert(Qp(u.shape,x.shape),()=>`packed reshape ${u.shape} to ${x.shape} isn't free`);let I=de({inputs:{x:t},backend:a,attrs:{shape:[1,n.inChannels,n.outChannels]}});y.push(I);let T=Bh({a:x,b:I,backend:a,transposeA:m,transposeB:f,bias:r,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),C=a.texData.get(T.dataId);v.assert(C.isPacked,()=>"batchMatMul result is expected to be packed"),u.shape=w,C.shape=n.outShape,g=na({inputs:{x:T},backend:a}),g.shape=n.outShape,y.push(T)}else{let b=n.outHeight*n.outWidth,x=de({inputs:{x:e},backend:a,attrs:{shape:h?[n.batchSize,b,n.inChannels]:[n.batchSize,n.inChannels,b]}}),w=de({inputs:{x:t},backend:a,attrs:{shape:[1,n.inChannels,n.outChannels]}}),I=Bh({a:h?x:w,b:h?w:x,transposeA:!h,transposeB:f,backend:a,bias:r,activation:o,preluActivationWeights:s,leakyreluAlpha:i});g=de({inputs:{x:I},backend:a,attrs:{shape:n.outShape}}),y.push(x),y.push(w),y.push(I)}for(let b of y)a.disposeIntermediateTensorInfo(b);return g}function aE({x:e,filter:t,convInfo:n,backend:a,bias:r=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let{filterWidth:l,filterHeight:u,inChannels:p,outWidth:d,outHeight:c,dataFormat:h}=n,m=h==="channelsLast",f=l*u*p,g=c*d,y=[n.batchSize,f,g],b=!0,x=!1,w=[];if(s!=null){let K=Vh(s.shape,m);K!=null&&(s=de({inputs:{x:s},backend:a,attrs:{shape:K}}),w.push(s))}if(r!=null){let K=Vh(r.shape,m);K!=null&&(r=de({inputs:{x:r},backend:a,attrs:{shape:K}}),w.push(r))}let I=de({inputs:{x:t},backend:a,attrs:{shape:[1,f,v.sizeFromShape(t.shape)/f]}});w.push(I);let T=new rQ(y,n),C=[e.shape,[n.padInfo.top,n.padInfo.left],[n.strideHeight,n.strideWidth],[n.dilationHeight,n.dilationWidth],[n.inChannels],[n.filterWidth*n.inChannels],[n.outWidth]],E=a.runWebGLProgram(T,[e],"float32",C),A=de({inputs:{x:E},backend:a,attrs:{shape:y}});w.push(E),w.push(A);let R=r!=null,F=s!=null,S=o==="leakyrelu",M=o?ec(o,!0):null,B=new q_(m?A.shape:I.shape,m?I.shape:A.shape,m?[n.batchSize,g,n.outChannels]:[n.batchSize,n.outChannels,g],b,x,R,M,F,S),U=m?[A,I]:[I,A];if(r&&U.push(r),F&&U.push(s),S){let K=a.makeTensorInfo([],"float32",v.createScalarValue(i,"float32"));U.push(K),w.push(K)}let G=a.runWebGLProgram(B,U,"float32"),q=de({inputs:{x:G},backend:a,attrs:{shape:n.outShape}});w.push(G);for(let K of w)a.disposeIntermediateTensorInfo(K);return q}function sQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:p}=a,d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,s.shape,i,u,o,p,!1,d),h;if(c.filterHeight===1&&c.filterWidth===1&&c.dilationHeight===1&&c.dilationWidth===1&&c.strideHeight===1&&c.strideWidth===1&&(c.padInfo.type==="SAME"||c.padInfo.type==="VALID"))h=nE({x:r,filter:s,convInfo:c,backend:n});else if(c.strideWidth<=2&&d==="channelsLast"&&H().getBool("WEBGL_EXP_CONV")){let f=new tE(c),g=[[c.padInfo.top,c.padInfo.left],[c.strideHeight,c.strideWidth],[c.dilationHeight,c.dilationWidth],[c.inHeight,c.inWidth]];h=n.runWebGLProgram(f,[r,s],"float32",g)}else if(H().getBool("WEBGL_CONV_IM2COL"))h=aE({x:r,filter:s,convInfo:c,backend:n});else{let f=new eE(c);h=n.runWebGLProgram(f,[r,s],"float32")}let m=de({inputs:{x:h},backend:n,attrs:{shape:c.outShape}});return n.disposeIntermediateTensorInfo(h),m}var iQ={kernelName:vi,backendName:"webgl",kernelFunc:sQ},oQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,a=e.padInfo.top,r=e.padInfo.left,s=e.dataFormat==="channelsLast";this.userCode=`
+    `;
+  }
+};
+function getShapeForBatchMatMul(shape, isChannelsLast) {
+  const length = shape.length;
+  if (length >= 3) {
+    return isChannelsLast ? [
+      ...shape.slice(0, -3),
+      shape[length - 3] * shape[length - 2],
+      shape[length - 1]
+    ] : [
+      ...shape.slice(0, -3),
+      shape[length - 3],
+      shape[length - 2] * shape[length - 1]
+    ];
+  } else if (!isChannelsLast && length === 1 && shape[0] > 1) {
+    return [shape[0], 1];
+  } else {
+    return null;
+  }
+}
+function conv2dByMatMul({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const xShape = x.shape;
+  const xTexData = backend2.texData.get(x.dataId);
+  const sharedMatMulDim = convInfo.inChannels;
+  const outerShapeX = xShape[0] * xShape[1] * xShape[2];
+  const outerShapeFilter = convInfo.outChannels;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const transposeA = false;
+  const transposeB = false;
+  let out;
+  const intermediates = [];
+  if (preluActivationWeights != null) {
+    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);
+    if (targetShape != null) {
+      preluActivationWeights = reshape4({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: targetShape }
+      });
+      intermediates.push(preluActivationWeights);
+    }
+  }
+  if (bias != null) {
+    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);
+    if (targetShape != null) {
+      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });
+      intermediates.push(bias);
+    }
+  }
+  const batchMatMulWillBeUnpacked = (outerShapeX === 1 || outerShapeFilter === 1) && sharedMatMulDim > MATMUL_SHARED_DIM_THRESHOLD;
+  const canOptimize = !batchMatMulWillBeUnpacked && xTexData.isPacked && isChannelsLast && xTexData.texture != null && xShape[2] % 2 !== 0 && util_exports.arraysEqual(xTexData.shape.slice(-3), xShape.slice(-3));
+  if (canOptimize) {
+    const targetShape = xShape[0] * xShape[1] * (xShape[2] + 1);
+    const xReshaped = {
+      dataId: x.dataId,
+      shape: [1, targetShape, convInfo.inChannels],
+      dtype: x.dtype
+    };
+    const originalXTexDataShape = xTexData.shape;
+    xTexData.shape = xTexData.shape.slice();
+    xTexData.shape[xTexData.shape.length - 2]++;
+    util_exports.assert(isReshapeFree(xTexData.shape, xReshaped.shape), () => `packed reshape ${xTexData.shape} to ${xReshaped.shape} isn't free`);
+    const filterReshaped = reshape4({
+      inputs: { x: filter },
+      backend: backend2,
+      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }
+    });
+    intermediates.push(filterReshaped);
+    const pointwiseConv = batchMatMulImpl({
+      a: xReshaped,
+      b: filterReshaped,
+      backend: backend2,
+      transposeA,
+      transposeB,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+    const pointwiseConvTexData = backend2.texData.get(pointwiseConv.dataId);
+    util_exports.assert(pointwiseConvTexData.isPacked, () => "batchMatMul result is expected to be packed");
+    xTexData.shape = originalXTexDataShape;
+    pointwiseConvTexData.shape = convInfo.outShape;
+    out = identity3({ inputs: { x: pointwiseConv }, backend: backend2 });
+    out.shape = convInfo.outShape;
+    intermediates.push(pointwiseConv);
+  } else {
+    const numCols = convInfo.outHeight * convInfo.outWidth;
+    const xReshaped = reshape4({
+      inputs: { x },
+      backend: backend2,
+      attrs: {
+        shape: isChannelsLast ? [convInfo.batchSize, numCols, convInfo.inChannels] : [convInfo.batchSize, convInfo.inChannels, numCols]
+      }
+    });
+    const filterReshaped = reshape4({
+      inputs: { x: filter },
+      backend: backend2,
+      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }
+    });
+    const result = batchMatMulImpl({
+      a: isChannelsLast ? xReshaped : filterReshaped,
+      b: isChannelsLast ? filterReshaped : xReshaped,
+      transposeA: !isChannelsLast,
+      transposeB,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+    out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: convInfo.outShape } });
+    intermediates.push(xReshaped);
+    intermediates.push(filterReshaped);
+    intermediates.push(result);
+  }
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return out;
+}
+function conv2dWithIm2Row({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const { filterWidth, filterHeight, inChannels, outWidth, outHeight, dataFormat } = convInfo;
+  const isChannelsLast = dataFormat === "channelsLast";
+  const sharedDim = filterWidth * filterHeight * inChannels;
+  const numCols = outHeight * outWidth;
+  const x2ColShape = [convInfo.batchSize, sharedDim, numCols];
+  const transposeA = true;
+  const transposeB = false;
+  const intermediates = [];
+  if (preluActivationWeights != null) {
+    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);
+    if (targetShape != null) {
+      preluActivationWeights = reshape4({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: targetShape }
+      });
+      intermediates.push(preluActivationWeights);
+    }
+  }
+  if (bias != null) {
+    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);
+    if (targetShape != null) {
+      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });
+      intermediates.push(bias);
+    }
+  }
+  const w2Row = reshape4({
+    inputs: { x: filter },
+    backend: backend2,
+    attrs: { shape: [1, sharedDim, util_exports.sizeFromShape(filter.shape) / sharedDim] }
+  });
+  intermediates.push(w2Row);
+  const im2ColProgram = new Im2ColPackedProgram(x2ColShape, convInfo);
+  const customValues = [
+    x.shape,
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inChannels],
+    [convInfo.filterWidth * convInfo.inChannels],
+    [convInfo.outWidth]
+  ];
+  const im2Col = backend2.runWebGLProgram(im2ColProgram, [x], "float32", customValues);
+  const im2ColReshaped = reshape4({ inputs: { x: im2Col }, backend: backend2, attrs: { shape: x2ColShape } });
+  intermediates.push(im2Col);
+  intermediates.push(im2ColReshaped);
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;
+  const matmulProgram = new MatMulPackedProgram(isChannelsLast ? im2ColReshaped.shape : w2Row.shape, isChannelsLast ? w2Row.shape : im2ColReshaped.shape, isChannelsLast ? [convInfo.batchSize, numCols, convInfo.outChannels] : [convInfo.batchSize, convInfo.outChannels, numCols], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  const inputs = isChannelsLast ? [im2ColReshaped, w2Row] : [w2Row, im2ColReshaped];
+  if (bias) {
+    inputs.push(bias);
+  }
+  if (hasPreluActivationWeights) {
+    inputs.push(preluActivationWeights);
+  }
+  if (hasLeakyreluAlpha) {
+    const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+    inputs.push($leakyreluAlpha);
+    intermediates.push($leakyreluAlpha);
+  }
+  const product = backend2.runWebGLProgram(matmulProgram, inputs, "float32");
+  const out = reshape4({ inputs: { x: product }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  intermediates.push(product);
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return out;
+}
+function conv2d4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  let out;
+  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === "SAME" || convInfo.padInfo.type === "VALID")) {
+    out = conv2dByMatMul({ x, filter, convInfo, backend: backend2 });
+  } else if (convInfo.strideWidth <= 2 && $dataFormat === "channelsLast" && env().getBool("WEBGL_EXP_CONV")) {
+    const program = new Conv2DPackedProgram(convInfo);
+    const customValues = [
+      [convInfo.padInfo.top, convInfo.padInfo.left],
+      [convInfo.strideHeight, convInfo.strideWidth],
+      [convInfo.dilationHeight, convInfo.dilationWidth],
+      [convInfo.inHeight, convInfo.inWidth]
+    ];
+    out = backend2.runWebGLProgram(program, [x, filter], "float32", customValues);
+  } else if (env().getBool("WEBGL_CONV_IM2COL")) {
+    out = conv2dWithIm2Row({ x, filter, convInfo, backend: backend2 });
+  } else {
+    const program = new Conv2DProgram(convInfo);
+    out = backend2.runWebGLProgram(program, [x, filter], "float32");
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  backend2.disposeIntermediateTensorInfo(out);
+  return outReshaped;
+}
+var conv2DConfig2 = {
+  kernelName: Conv2D,
+  backendName: "webgl",
+  kernelFunc: conv2d4
+};
+var Conv2DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
@@ -2669,22 +55948,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int b = 0; b < ${e.batchSize}; b++) {
-          for (int yR = 0; yR < ${e.outHeight}; yR++) {
-            int xR = wR + yR * ${t} - ${a};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+            int xR = wR + yR * ${strideHeight} - ${padTop};
 
-            if (xR < 0 || xR >= ${e.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int yC = 0; yC < ${e.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${r};
+            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+              int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-              if (xC < 0 || xC >= ${e.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
-              if (${s}) {
+              if (${isChannelsLast}) {
                 float dyValue = getDy(b, yR, yC, d2);
                 float xValue = getX(b, xR, xC, d1);
                 dotProd += (xValue * dyValue);
@@ -2699,45 +55978,62 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},lQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=e.dataFormat==="channelsLast",i=t-1-e.padInfo.top,o=n-1-e.padInfo.left,l=s?1:2,u=s?2:3,p=s?3:1;this.userCode=`
-      const ivec2 pads = ivec2(${i}, ${o});
+    `;
+  }
+};
+var Conv2DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const channelDim = isChannelsLast ? 3 : 1;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d1 = coords[${p}];
+        int d1 = coords[${channelDim}];
 
-        ivec2 dyCorner = ivec2(coords[${l}], coords[${u}]) - pads;
+        ivec2 dyCorner = ivec2(coords[${rowDim}], coords[${colDim}]) - pads;
         int dyRCorner = dyCorner.x;
         int dyCCorner = dyCorner.y;
 
         // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${t}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          int wRPerm = ${t} - 1 - wR;
+          int wRPerm = ${filterHeight} - 1 - wR;
 
-          for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
-            int wCPerm = ${n} - 1 - wC;
+            int wCPerm = ${filterWidth} - 1 - wC;
 
-            for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
+            for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {
 
-              if (${s}) {
+              if (${isChannelsLast}) {
                 float xValue = getDy(batch, idyR, idyC, d2);
                 float wValue = getW(wRPerm, wCPerm, d1, d2);
                 dotProd += xValue * wValue;
@@ -2752,7 +56048,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},uQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideDepth,n=e.strideHeight,a=e.strideWidth,r=e.padInfo.front,s=e.padInfo.top,i=e.padInfo.left;this.userCode=`
+    `;
+  }
+};
+var Conv3DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.userCode = `
       void main() {
         ivec5 coords = getOutputCoords();
         int wF = coords.x;
@@ -2763,25 +56072,25 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         float dotProd = 0.0;
 
-        for (int b = 0; b < ${e.batchSize}; b++) {
-          for (int yF = 0; yF < ${e.outDepth}; yF++) {
-            int xF = wF + yF * ${t} - ${r};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yF = 0; yF < ${convInfo.outDepth}; yF++) {
+            int xF = wF + yF * ${strideDepth} - ${padFront};
 
-            if (xF < 0 || xF >= ${e.inDepth}) {
+            if (xF < 0 || xF >= ${convInfo.inDepth}) {
               continue;
             }
 
-            for (int yR = 0; yR < ${e.outHeight}; yR++) {
-              int xR = wR + yR * ${n} - ${s};
+            for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+              int xR = wR + yR * ${strideHeight} - ${padTop};
 
-              if (xR < 0 || xR >= ${e.inHeight}) {
+              if (xR < 0 || xR >= ${convInfo.inHeight}) {
                 continue;
               }
 
-              for (int yC = 0; yC < ${e.outWidth}; yC++) {
-                int xC = wC + yC * ${a} - ${i};
+              for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+                int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-                if (xC < 0 || xC >= ${e.inWidth}) {
+                if (xC < 0 || xC >= ${convInfo.inWidth}) {
                   continue;
                 }
 
@@ -2794,8 +56103,24 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},pQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,a=e.filterWidth,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=t-1-e.padInfo.front,l=n-1-e.padInfo.top,u=a-1-e.padInfo.left;this.userCode=`
-      const ivec3 pads = ivec3(${o}, ${l}, ${u});
+    `;
+  }
+};
+var Conv3DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padFront = filterDepth - 1 - convInfo.padInfo.front;
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2809,39 +56134,39 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         int dyCCorner = dyCorner.z;
 
         float dotProd = 0.0;
-        for (int wF = 0; wF < ${t}; wF++) {
-          float dyF = float(dyFCorner + wF) / ${r}.0;
+        for (int wF = 0; wF < ${filterDepth}; wF++) {
+          float dyF = float(dyFCorner + wF) / ${strideDepth}.0;
 
-          if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {
+          if (dyF < 0.0 || dyF >= ${convInfo.outDepth}.0 || fract(dyF) > 0.0) {
             continue;
           }
           int idyF = int(dyF);
 
-          int wFPerm = ${t} - 1 - wF;
+          int wFPerm = ${filterDepth} - 1 - wF;
 
-          for (int wR = 0; wR < ${n}; wR++) {
-            float dyR = float(dyRCorner + wR) / ${s}.0;
+          for (int wR = 0; wR < ${filterHeight}; wR++) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
               fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            int wRPerm = ${n} - 1 - wR;
+            int wRPerm = ${filterHeight} - 1 - wR;
 
-            for (int wC = 0; wC < ${a}; wC++) {
-              float dyC = float(dyCCorner + wC) / ${i}.0;
+            for (int wC = 0; wC < ${filterWidth}; wC++) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
               int idyC = int(dyC);
 
-              int wCPerm = ${a} - 1 - wC;
+              int wCPerm = ${filterWidth} - 1 - wC;
 
-              for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
+              for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {
                 float xValue = getDy(batch, idyF, idyR, idyC, d2);
                 float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);
                 dotProd += xValue * wValue;
@@ -2851,14 +56176,126 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function cQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:p}=a,d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,p,i,1,o,u,!1,d),h=new oQ(c);return n.runWebGLProgram(h,[r,s],"float32")}var dQ={kernelName:em,backendName:"webgl",kernelFunc:cQ};function hQ(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:p}=a,d=N.convertConv2DDataFormat(u),c=N.computeConv2DInfo(i,s.shape,o,1,l,p,!1,d),h=new lQ(c);return n.runWebGLProgram(h,[r,s],"float32")}var mQ={kernelName:wi,backendName:"webgl",kernelFunc:hQ};function fQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a,u=N.computeConv3DInfo(r.shape,s.shape,i,l,o),p=new aQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var gQ={kernelName:lc,backendName:"webgl",kernelFunc:fQ};function yQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,filterShape:l}=a,u=N.computeConv3DInfo(r.shape,l,i,1,o),p=new uQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var bQ={kernelName:tm,backendName:"webgl",kernelFunc:yQ};function xQ(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{pad:i,strides:o,inputShape:l}=a,u=N.computeConv3DInfo(l,s.shape,o,1,i),p=new pQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var vQ={kernelName:nm,backendName:"webgl",kernelFunc:xQ},wQ=Uu+`
+    `;
+  }
+};
+function conv2DBackpropFilter3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const program = new Conv2DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var conv2DBackpropFilterConfig2 = {
+  kernelName: Conv2DBackpropFilter,
+  backendName: "webgl",
+  kernelFunc: conv2DBackpropFilter3
+};
+function conv2DBackpropInput3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const program = new Conv2DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var conv2DBackpropInputConfig2 = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "webgl",
+  kernelFunc: conv2DBackpropInput3
+};
+function conv3D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);
+  const program = new Conv3DProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, filter], "float32");
+}
+var conv3DConfig2 = {
+  kernelName: Conv3D,
+  backendName: "webgl",
+  kernelFunc: conv3D2
+};
+function conv3DBackpropFilterV22(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, filterShape } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);
+  const program = new Conv3DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var conv3DBackpropFilterV2Config2 = {
+  kernelName: Conv3DBackpropFilterV2,
+  backendName: "webgl",
+  kernelFunc: conv3DBackpropFilterV22
+};
+function conv3DBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { pad: pad3, strides, inputShape } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);
+  const program = new Conv3DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var conv3DBackpropInputConfig = {
+  kernelName: Conv3DBackpropInputV2,
+  backendName: "webgl",
+  kernelFunc: conv3DBackpropInput2
+};
+var COS = CHECK_NAN_SNIPPET_UNARY + `
   return cos(x);
-`,kQ=Ye({opSnippet:wQ}),IQ={kernelName:ki,backendName:"webgl",kernelFunc:kQ},SQ=`
+`;
+var cos3 = unaryKernelFunc2({ opSnippet: COS });
+var cosConfig2 = {
+  kernelName: Cos,
+  backendName: "webgl",
+  kernelFunc: cos3
+};
+var COSH = `
   float e2x = exp(-x);
   return (e2x + 1.0 / e2x) / 2.0;
-`,TQ=Ye({opSnippet:SQ}),NQ={kernelName:Ii,backendName:"webgl",kernelFunc:TQ},CQ=class{constructor(e,t,n,a,r){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[s,i,o,l]=e,[u]=t,[p,d]=n;this.outputShape=[u,p,d,l];let c=a==="bilinear"?1:0,[h,m]=[`${i-1}.0`,`${o-1}.0`],[f,g,y]=p>1?[`${(i-1)/(p-1)}`,"(y2-y1) * height_ratio",`y1*${h} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${h}`],[b,x,w]=d>1?[`${(o-1)/(d-1)}`,"(x2-x1) * width_ratio",`x1*${m} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${m}`];this.userCode=`
-      const float height_ratio = float(${f});
-      const float width_ratio = float(${b});
+`;
+var cosh3 = unaryKernelFunc2({ opSnippet: COSH });
+var coshConfig2 = {
+  kernelName: Cosh,
+  backendName: "webgl",
+  kernelFunc: cosh3
+};
+var CropAndResizeProgram = class {
+  constructor(imageShape, boxShape, cropSize, method, extrapolationValue) {
+    this.variableNames = ["Image", "Boxes", "BoxInd"];
+    this.outputShape = [];
+    const [batch, imageHeight, imageWidth, depth] = imageShape;
+    const [numBoxes] = boxShape;
+    const [cropHeight, cropWidth] = cropSize;
+    this.outputShape = [numBoxes, cropHeight, cropWidth, depth];
+    const methodId = method === "bilinear" ? 1 : 0;
+    const [inputHeightFloat, inputWidthFloat] = [`${imageHeight - 1}.0`, `${imageWidth - 1}.0`];
+    const [heightRatio, heightScale, inY] = cropHeight > 1 ? [
+      `${(imageHeight - 1) / (cropHeight - 1)}`,
+      "(y2-y1) * height_ratio",
+      `y1*${inputHeightFloat} + float(y)*(height_scale)`
+    ] : [
+      "0.0",
+      "0.0",
+      `0.5 * (y1+y2) * ${inputHeightFloat}`
+    ];
+    const [widthRatio, widthScale, inX] = cropWidth > 1 ? [
+      `${(imageWidth - 1) / (cropWidth - 1)}`,
+      "(x2-x1) * width_ratio",
+      `x1*${inputWidthFloat} + float(x)*(width_scale)`
+    ] : [
+      "0.0",
+      "0.0",
+      `0.5 * (x1+x2) * ${inputWidthFloat}`
+    ];
+    this.userCode = `
+      const float height_ratio = float(${heightRatio});
+      const float width_ratio = float(${widthRatio});
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -2874,26 +56311,26 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         // get image in batch index
         int bInd = round(getBoxInd(b));
-        if(bInd < 0 || bInd >= ${s}) {
+        if(bInd < 0 || bInd >= ${batch}) {
           return;
         }
 
-        float height_scale = ${g};
-        float width_scale = ${x};
+        float height_scale = ${heightScale};
+        float width_scale = ${widthScale};
 
-        float in_y = ${y};
-        if( in_y < 0.0 || in_y > ${h} ) {
-          setOutput(float(${r}));
+        float in_y = ${inY};
+        if( in_y < 0.0 || in_y > ${inputHeightFloat} ) {
+          setOutput(float(${extrapolationValue}));
           return;
         }
-        float in_x = ${w};
-        if( in_x < 0.0 || in_x > ${m} ) {
-          setOutput(float(${r}));
+        float in_x = ${inX};
+        if( in_x < 0.0 || in_x > ${inputWidthFloat} ) {
+          setOutput(float(${extrapolationValue}));
           return;
         }
 
         vec2 sourceFracIndexCR = vec2(in_x,in_y);
-        if(${c} == 1) {
+        if(${methodId} == 1) {
           // Compute the four integer indices.
           ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);
           ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));
@@ -2917,20 +56354,174 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           setOutput(newValue);
         }
       }
-    `}},_Q=e=>{let{inputs:t,backend:n,attrs:a}=e,{image:r,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=a,p=new CQ(r.shape,s.shape,o,l,u);return n.runWebGLProgram(p,[r,s,i],"float32")},EQ={kernelName:Rl,backendName:"webgl",kernelFunc:_Q},nc;(function(e){e.Prod="*",e.Sum="+"})(nc||(nc={}));var sI=class{constructor(e,t,n,a){this.op=e,this.outputShape=t,this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}];let r=this.outputShape.length,s=this.op===nc.Prod?"1.0":"0.0",i=n?s:`getX(${iI(r,"coords",this.op)})`,o=this.outputShape[this.outputShape.length-1],l="",u="";n?(l=a?`end != ${o-1}`:"end != 0",u=a?"end + 1":"end - 1"):(l=a?`end + pow2 < ${o}`:"end >= pow2",u=a?"end + pow2":"end - pow2"),this.userCode=`
+    `;
+  }
+};
+var cropAndResize3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, boxes, boxInd } = inputs;
+  const { cropSize, method, extrapolationValue } = attrs;
+  const program = new CropAndResizeProgram(image2.shape, boxes.shape, cropSize, method, extrapolationValue);
+  return backend2.runWebGLProgram(program, [image2, boxes, boxInd], "float32");
+};
+var cropAndResizeConfig2 = {
+  kernelName: CropAndResize,
+  backendName: "webgl",
+  kernelFunc: cropAndResize3
+};
+var CumOpType;
+(function(CumOpType2) {
+  CumOpType2["Prod"] = "*";
+  CumOpType2["Sum"] = "+";
+})(CumOpType || (CumOpType = {}));
+var CumProgram = class {
+  constructor(op2, outputShape, exclusive, reverse5) {
+    this.op = op2;
+    this.outputShape = outputShape;
+    this.variableNames = ["x"];
+    this.customUniforms = [{ name: "index", type: "float" }];
+    const rank = this.outputShape.length;
+    const initVal = this.op === CumOpType.Prod ? "1.0" : "0.0";
+    const val = exclusive ? initVal : `getX(${getCoords2(rank, "coords", this.op)})`;
+    const length = this.outputShape[this.outputShape.length - 1];
+    let condition = "";
+    let idxString = "";
+    if (exclusive) {
+      condition = reverse5 ? `end != ${length - 1}` : "end != 0";
+      idxString = reverse5 ? "end + 1" : "end - 1";
+    } else {
+      condition = reverse5 ? `end + pow2 < ${length}` : "end >= pow2";
+      idxString = reverse5 ? "end + pow2" : "end - pow2";
+    }
+    this.userCode = `
       void main() {
-        ${gt(r)} coords = getOutputCoords();
-        int end = ${oI(r,"coords",this.op)};
-        float val = ${i};
+        ${getCoordsDataType(rank)} coords = getOutputCoords();
+        int end = ${getFinalCoord(rank, "coords", this.op)};
+        float val = ${val};
         int pow2 = int(pow(2.0, index));
-        if (${l}) {
-          int idx = ${u};
-          ${oI(r,"coords",this.op)} = idx;
-          val ${this.op}= getX(${iI(r,"coords",this.op)});
+        if (${condition}) {
+          int idx = ${idxString};
+          ${getFinalCoord(rank, "coords", this.op)} = idx;
+          val ${this.op}= getX(${getCoords2(rank, "coords", this.op)});
         }
         setOutput(val);
       }
-    `}};function iI(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.x, ${t}.y`;if(e===3)return`${t}.x, ${t}.y, ${t}.z`;if(e===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function oI(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.y`;if(e===3)return`${t}.z`;if(e===4)return`${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function rE(e,t,n,a,r,s){let i=t.shape.length,o=N.getAxesPermutation([a],i),l=t;o!=null&&(l=In({inputs:{x:t},backend:n,attrs:{perm:o}}));let u=N.getInnerMostAxes(1,i)[0];if(u!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${a}`);let p=l.shape[u],d=na({inputs:{x:l},backend:n});for(let c=0;c<=Math.ceil(Math.log2(p))-1;c++){let h=new sI(e,l.shape,!1,s),m=[[c]],f=d;d=n.runWebGLProgram(h,[d],d.dtype,m),n.disposeIntermediateTensorInfo(f)}if(r){let c=new sI(e,l.shape,r,s),h=d;d=n.runWebGLProgram(c,[d],d.dtype),n.disposeIntermediateTensorInfo(h)}if(o!=null){let c=N.getUndoAxesPermutation(o),h=In({inputs:{x:d},backend:n,attrs:{perm:c}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(l),h}return d}function AQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;return rE(nc.Prod,r,n,s,i,o)}var $Q={kernelName:Dl,backendName:"webgl",kernelFunc:AQ};function FQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;return rE(nc.Sum,r,n,s,i,o)}var DQ={kernelName:Si,backendName:"webgl",kernelFunc:FQ};function RQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i,binaryOutput:o}=a;if(r.shape.length===1){let l=n.readSync(r.dataId),u=n.readSync(s.dataId),p=O_(l,u,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,p)}else if(r.shape.length===2){let l=n.bufferSync(r),u=n.bufferSync(s),p=GZ(l,u,i,o);return n.makeTensorInfo(p.shape,s.dtype,p.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${r.shape.length}.`)}var MQ={kernelName:am,backendName:"webgl",kernelFunc:RQ},PQ=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=n,this.userCode=`
+    `;
+  }
+};
+function getCoords2(rank, name, op2) {
+  if (rank === 1) {
+    return `${name}`;
+  } else if (rank === 2) {
+    return `${name}.x, ${name}.y`;
+  } else if (rank === 3) {
+    return `${name}.x, ${name}.y, ${name}.z`;
+  } else if (rank === 4) {
+    return `${name}.x, ${name}.y, ${name}.z, ${name}.w`;
+  } else {
+    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);
+  }
+}
+function getFinalCoord(rank, name, op2) {
+  if (rank === 1) {
+    return `${name}`;
+  } else if (rank === 2) {
+    return `${name}.y`;
+  } else if (rank === 3) {
+    return `${name}.z`;
+  } else if (rank === 4) {
+    return `${name}.w`;
+  } else {
+    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);
+  }
+}
+function cumImpl(op2, x, backend2, axis, exclusive, reverse5) {
+  const xRank = x.shape.length;
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  if (permutedAxis !== xRank - 1) {
+    throw new Error(`WebGL cumprod shader expects an inner-most axis=${x.shape.length - 1} but got axis=${axis}`);
+  }
+  const size = permutedX.shape[permutedAxis];
+  let result = identity3({ inputs: { x: permutedX }, backend: backend2 });
+  for (let i = 0; i <= Math.ceil(Math.log2(size)) - 1; i++) {
+    const program = new CumProgram(op2, permutedX.shape, false, reverse5);
+    const customValues = [[i]];
+    const prevResult = result;
+    result = backend2.runWebGLProgram(program, [result], result.dtype, customValues);
+    backend2.disposeIntermediateTensorInfo(prevResult);
+  }
+  if (exclusive) {
+    const program = new CumProgram(op2, permutedX.shape, exclusive, reverse5);
+    const prevResult = result;
+    result = backend2.runWebGLProgram(program, [result], result.dtype);
+    backend2.disposeIntermediateTensorInfo(prevResult);
+  }
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo(permutedX);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+function cumprod3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  return cumImpl(CumOpType.Prod, x, backend2, axis, exclusive, reverse5);
+}
+var cumprodConfig2 = {
+  kernelName: Cumprod,
+  backendName: "webgl",
+  kernelFunc: cumprod3
+};
+function cumsum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  return cumImpl(CumOpType.Sum, x, backend2, axis, exclusive, reverse5);
+}
+var cumsumConfig2 = {
+  kernelName: Cumsum,
+  backendName: "webgl",
+  kernelFunc: cumsum3
+};
+function denseBincount3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size, binaryOutput } = attrs;
+  if (x.shape.length === 1) {
+    const xVals = backend2.readSync(x.dataId);
+    const weightsVals = backend2.readSync(weights.dataId);
+    const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);
+    return backend2.makeTensorInfo([size], weights.dtype, outVals);
+  } else if (x.shape.length === 2) {
+    const xBuf = backend2.bufferSync(x);
+    const weightsBuf = backend2.bufferSync(weights);
+    const outBuf = bincountReduceImplCPU(xBuf, weightsBuf, size, binaryOutput);
+    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);
+  }
+  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);
+}
+var denseBincountConfig2 = {
+  kernelName: DenseBincount,
+  backendName: "webgl",
+  kernelFunc: denseBincount3
+};
+var DepthToSpaceProgram = class {
+  constructor(outputShape, blockSize, dataFormat) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    this.outputShape = outputShape;
+    this.blockSize = blockSize;
+    this.dataFormat = dataFormat;
+    this.userCode = `
     void main() {
       ivec4 coords = getOutputCoords();
       int b = coords[0];
@@ -2938,37 +56529,130 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
       int w = ${this.getWidthCoordString()};
       int d = ${this.getDepthCoordString()};
 
-      int in_h = h / ${t};
-      int offset_h = imod(h, ${t});
-      int in_w = w / ${t};
-      int offset_w = imod(w, ${t});
-      int offset_d = (offset_h * ${t} + offset_w) *
+      int in_h = h / ${blockSize};
+      int offset_h = imod(h, ${blockSize});
+      int in_w = w / ${blockSize};
+      int offset_w = imod(w, ${blockSize});
+      int offset_d = (offset_h * ${blockSize} + offset_w) *
         ${this.getOutputDepthSize()};
       int in_d = d + offset_d;
 
       float result = ${this.getInputSamplingString()};
       setOutput(result);
     }
-  `}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function OQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockSize:s,dataFormat:i}=a,o=r.shape[0],l=i==="NHWC"?r.shape[1]:r.shape[2],u=i==="NHWC"?r.shape[2]:r.shape[3],p=i==="NHWC"?r.shape[3]:r.shape[1],d=l*s,c=u*s,h=p/(s*s),m=i==="NHWC"?[o,d,c,h]:[o,h,d,c],f=new PQ(m,s,i);return n.runWebGLProgram(f,[r],r.dtype)}var LQ={kernelName:Ml,backendName:"webgl",kernelFunc:OQ},sE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=_n(this.outputShape.length);let s=e.filterHeight,i=e.filterWidth,o=e.outChannels/e.inChannels,l="",u="";n&&(a?l=`float activation(float a) {
+  `;
+  }
+  getHeightCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[1]`;
+    } else {
+      return `coords[2]`;
+    }
+  }
+  getWidthCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[2]`;
+    } else {
+      return `coords[3]`;
+    }
+  }
+  getDepthCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[3]`;
+    } else {
+      return `coords[1]`;
+    }
+  }
+  getOutputDepthSize() {
+    if (this.dataFormat === "NHWC") {
+      return this.outputShape[3];
+    } else {
+      return this.outputShape[1];
+    }
+  }
+  getInputSamplingString() {
+    if (this.dataFormat === "NHWC") {
+      return `getX(b, in_h, in_w, in_d)`;
+    } else {
+      return `getX(b, in_d, in_h, in_w)`;
+    }
+  }
+};
+function depthToSpace3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  const batchSize = x.shape[0];
+  const inputHeight = dataFormat === "NHWC" ? x.shape[1] : x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? x.shape[2] : x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? x.shape[3] : x.shape[1];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const outputShape = dataFormat === "NHWC" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];
+  const program = new DepthToSpaceProgram(outputShape, blockSize, dataFormat);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var depthToSpaceConfig2 = {
+  kernelName: DepthToSpace,
+  backendName: "webgl",
+  kernelFunc: depthToSpace3
+};
+var DepthwiseConv2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `float activation(float a) {
           float b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:r?l=`float activation(float a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `float activation(float a) {
           float b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:l=`
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `
           float activation(float x) {
-            ${n}
+            ${activation2}
           }
-        `,u="result = activation(result);");let p=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${l}
+        `;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${o};
-        int q = d2 - d1 * ${o};
+        int d1 = d2 / ${channelMul};
+        int q = d2 - d1 * ${channelMul};
 
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
@@ -2977,14 +56661,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
         // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.
-        for (int wR = 0; wR < ${s}; wR++) {
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
           int xR = xRCorner + wR * dilations[0];
 
           if (xR < 0 || xR >= inDims[0]) {
             continue;
           }
 
-          for (int wC = 0; wC < ${i}; wC++) {
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
             int xC = xCCorner + wC * dilations[1];
 
             if (xC < 0 || xC >= inDims[1]) {
@@ -2998,44 +56682,86 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
 
         float result = dotProd;
-        ${p}
-        ${u}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}},iE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=_n(this.outputShape.length);let s=e.outChannels/e.inChannels,i=e.padInfo.left,o=e.strideWidth,l=e.dilationWidth,u=e.filterHeight,p=e.filterWidth,d=p,c=`
+    `;
+  }
+};
+var DepthwiseConvPacked2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    const padLeft = convInfo.padInfo.left;
+    const strideWidth = convInfo.strideWidth;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const texelsAcross = filterWidth;
+    let mainLoop = `
       int xR; int xC; int xCOffset;
-      vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<p;g++)c+=`
-          vec4 xTexelC${g*2};
-          int xTexelC${g*2}Ready;
-          vec4 xTexelC${g*2+1};
-          int xTexelC${g*2+1}Ready;
-          vec4 xC${g};`;c+=`
-    for (int r = 0; r < ${u}; r++) {
-      `;for(let g=0;g<p;g++)c+=`
-          xTexelC${g*2} = vec4(0.0);
-          xTexelC${g*2}Ready = 0;
-          xTexelC${g*2+1} = vec4(0.0);
-          xTexelC${g*2+1}Ready = 0;
-          xC${g} = vec4(0.0);`;c+=`
+      vec4 wTexel; vec4 previous; vec4 final;`;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+          vec4 xTexelC${c * 2};
+          int xTexelC${c * 2}Ready;
+          vec4 xTexelC${c * 2 + 1};
+          int xTexelC${c * 2 + 1}Ready;
+          vec4 xC${c};`;
+    }
+    mainLoop += `
+    for (int r = 0; r < ${filterHeight}; r++) {
+      `;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+          xTexelC${c * 2} = vec4(0.0);
+          xTexelC${c * 2}Ready = 0;
+          xTexelC${c * 2 + 1} = vec4(0.0);
+          xTexelC${c * 2 + 1}Ready = 0;
+          xC${c} = vec4(0.0);`;
+    }
+    mainLoop += `
         xR = xRCorner + r * dilations[0];
         if (xR >=0 && xR < inDims[0]) {
-      `;for(let g=0;g<(d+1)/2;g++){let y=g*2;if(c+=`
-          xC = xCCorner + ${y*l};
-          `,o===1){if(y<p&&(i%2===1?(c+=`
+      `;
+    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {
+      const colIndex = texelC * 2;
+      mainLoop += `
+          xC = xCCorner + ${colIndex * dilationWidth};
+          `;
+      if (strideWidth === 1) {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                 xCOffset = xC + 1;
-                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xCOffset, d1);
+                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
 
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
-              `,l===1&&y>0?c+=`
-                xC${y} = vec4(xTexelC${y-2}.zw, xTexelC${y}.xy);
-                `:c+=`
+              `;
+            if (dilationWidth === 1 && colIndex > 0) {
+              mainLoop += `
+                xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);
+                `;
+            } else {
+              mainLoop += `
                   xCOffset = xC + 1 - 2;
 
                   if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -3047,174 +56773,283 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
                       previous.zw = vec2(0.0);
                     }
 
-                    xC${y} = vec4(previous.zw, xTexelC${y}.xy);
+                    xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);
                   } else {
-                    xC${y} = vec4(0.0, 0.0, xTexelC${y}.xy);
+                    xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);
                   }
-                  `):c+=`
-                if (xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xC, d1);
+                  `;
+            }
+          } else {
+            mainLoop += `
+                if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
-                xC${y} = xTexelC${y};
-                `,y+1<p)){let b=i%2===0?v.nearestLargerEven(l):l;l%2===0&&i%2===1||l%2!==0&&i%2!==1?(c+=`
-                  xCOffset = xC + imod(pads[1], 2) + ${b};
+                xC${colIndex} = xTexelC${colIndex};
+                `;
+          }
+          if (colIndex + 1 < filterWidth) {
+            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;
+            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {
+              mainLoop += `
+                  xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};
 
-                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                    xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                    xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
 
                     // Need to manually clear unused channels in case
                     // we're reading from recycled texture.
                     if (xCOffset + 1 >= inDims[1]) {
-                      xTexelC${y+1}.zw = vec2(0.0);
+                      xTexelC${colIndex + 1}.zw = vec2(0.0);
                     }
-                    xTexelC${y+1}Ready = 1;
+                    xTexelC${colIndex + 1}Ready = 1;
                   }
-                  `,l>1?c+=`
+                  `;
+              if (dilationWidth > 1) {
+                mainLoop += `
                     xCOffset -= 2;
                     if (xCOffset >= 0 && xCOffset < inDims[1]) {
                      previous = getX(batch, xR, xCOffset, d1);
-                     xC${y+1} = vec4(previous.zw, xTexelC${y+1}.xy);
+                     xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);
                     } else {
-                     xC${y+1} = vec4(0.0, 0.0, xTexelC${y+1}.xy);
+                     xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);
                     }
-                    `:c+=`
-                    xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.xy);
-                    `):b===1?c+=`
-                    xC${y+1} = xTexelC${y};
-                    `:c+=`
-                    xCOffset = xC + ${b};
+                    `;
+              } else {
+                mainLoop += `
+                    xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);
+                    `;
+              }
+            } else {
+              if (nextTexelOffset === 1) {
+                mainLoop += `
+                    xC${colIndex + 1} = xTexelC${colIndex};
+                    `;
+              } else {
+                mainLoop += `
+                    xCOffset = xC + ${nextTexelOffset};
 
-                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                      xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                      xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                       if (xCOffset + 1 >= inDims[1]) {
-                        xTexelC${y+1}.zw = vec2(0.0);
+                        xTexelC${colIndex + 1}.zw = vec2(0.0);
                       }
-                      xTexelC${y+1}Ready = 1;
+                      xTexelC${colIndex + 1}Ready = 1;
                     }
 
-                    xC${y+1} = xTexelC${y+1};
-                    `}}else y<p&&(i%2===1?(c+=`
+                    xC${colIndex + 1} = xTexelC${colIndex + 1};
+                    `;
+              }
+            }
+          }
+        }
+      } else {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                 xCOffset = xC + 1 - strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
-                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${y+1}Ready == 0) {
-                  xTexelC${y+1} = getX(batch, xR, xC + 1, d1);
+                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                  xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xC + 2 >= inDims[1]) {
-                    xTexelC${y+1}.zw = vec2(0.0);
+                    xTexelC${colIndex + 1}.zw = vec2(0.0);
                   }
-                  xTexelC${y+1}Ready = 1;
+                  xTexelC${colIndex + 1}Ready = 1;
                 }
 
-                xC${y} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
-              `,y+1<p&&(c+=`
+                xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+              `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
                   final = vec4(0.0);
                   xCOffset = xC + 1 + strides[1];
                   if(xCOffset >= 0 && xCOffset < inDims[1]) {
                     final = getX(batch, xR, xCOffset, d1);
                   }
-                  xC${y+1} = vec4(xTexelC${y+1}.xy, final.xy);
-                `)):(c+=`
-                if(xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xC, d1);
+                  xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);
+                `;
+            }
+          } else {
+            mainLoop += `
+                if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
                 xCOffset = xC + strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                  xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                  xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y+1}.zw = vec2(0.);
+                    xTexelC${colIndex + 1}.zw = vec2(0.);
                   }
-                  xTexelC${y+1}Ready = 1;
+                  xTexelC${colIndex + 1}Ready = 1;
                 }
 
-                xC${y} = vec4(
-                  xTexelC${y}.xy, xTexelC${y+1}.xy);
-              `,y+1<p&&(c+=`
-                  xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
-                `)));y<p&&(c+=`
-            wTexel = getW(r, ${y}, d1, q);
-            dotProd += xC${y} * vec4(wTexel.xz, wTexel.xz);
-          `,y+1<p&&(c+=`
-              wTexel = getW(r, ${y+1}, d1, q);
-              dotProd += xC${y+1} * vec4(wTexel.xz, wTexel.xz);
-            `))}c+=`
-    }
-  `,c+=`
+                xC${colIndex} = vec4(
+                  xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);
+              `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
+                  xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+                `;
+            }
+          }
+        }
       }
-    `;let h="",m="";n&&(a?h=`vec4 activation(vec4 a) {
+      if (colIndex < filterWidth) {
+        mainLoop += `
+            wTexel = getW(r, ${colIndex}, d1, q);
+            dotProd += xC${colIndex} * vec4(wTexel.xz, wTexel.xz);
+          `;
+        if (colIndex + 1 < filterWidth) {
+          mainLoop += `
+              wTexel = getW(r, ${colIndex + 1}, d1, q);
+              dotProd += xC${colIndex + 1} * vec4(wTexel.xz, wTexel.xz);
+            `;
+        }
+      }
+    }
+    mainLoop += `
+    }
+  `;
+    mainLoop += `
+      }
+    `;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:r?h=`vec4 activation(vec4 a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:h=`vec4 activation(vec4 x) {
-          ${n}
-        }`,m="result = activation(result);");let f=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${h}
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+          ${activation2}
+        }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${s};
-        int q = d2 - d1 * ${s};
+        int d1 = d2 / ${channelMul};
+        int q = d2 - d1 * ${channelMul};
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
         vec4 dotProd = vec4(0.000000000000001);
 
-        ${c}
+        ${mainLoop}
 
         vec4 result = dotProd - vec4(0.000000000000001);
-        ${f}
-        ${m}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}};function zQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=a,p=l;p==null&&(p=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(i,p),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);let d=N.computeConv2DInfo(r.shape,s.shape,i,p,o,u,!0),c;H().getBool("WEBGL_PACK_DEPTHWISECONV")&&d.strideWidth<=2&&d.outChannels/d.inChannels===1?c=new iE(d):c=new sE(d);let h=[[d.padInfo.top,d.padInfo.left],[d.strideHeight,d.strideWidth],[d.dilationHeight,d.dilationWidth],[d.inHeight,d.inWidth]];return n.runWebGLProgram(c,[r,s],"float32",h)}var WQ={kernelName:Ti,backendName:"webgl",kernelFunc:zQ},BQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,a=e.padInfo.top,r=e.padInfo.left,s=e.outChannels/e.inChannels;this.userCode=`
+    `;
+  }
+};
+function depthwiseConv2dNative2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  let program;
+  if (env().getBool("WEBGL_PACK_DEPTHWISECONV") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1) {
+    program = new DepthwiseConvPacked2DProgram(convInfo);
+  } else {
+    program = new DepthwiseConv2DProgram(convInfo);
+  }
+  const customValues = [
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inHeight, convInfo.inWidth]
+  ];
+  return backend2.runWebGLProgram(program, [x, filter], "float32", customValues);
+}
+var depthwiseConv2dNativeConfig2 = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNative2
+};
+var DepthwiseConv2DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
         int wC = coords.y;
         int d1 = coords.z;
         int dm = coords.w;
-        int d2 = d1 * ${s} + dm;
+        int d2 = d1 * ${channelMul} + dm;
 
         float dotProd = 0.0;
 
         // TO DO: Vec4 over the batch size
-        for (int b = 0; b < ${e.batchSize}; b++) {
-          for (int yR = 0; yR < ${e.outHeight}; yR++) {
-            int xR = wR + yR * ${t} - ${a};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+            int xR = wR + yR * ${strideHeight} - ${padTop};
 
-            if (xR < 0 || xR >= ${e.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int yC = 0; yC < ${e.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${r};
+            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+              int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-              if (xC < 0 || xC >= ${e.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
@@ -3226,8 +57061,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},VQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=t-1-e.padInfo.top,i=n-1-e.padInfo.left,o=e.outChannels/e.inChannels;this.userCode=`
-      const ivec2 pads = ivec2(${s}, ${i});
+    `;
+  }
+};
+var DepthwiseConv2DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3239,30 +57088,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         float dotProd = 0.0;
 
-        for (int wR = 0; wR < ${t}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          int wRPerm = ${t} - 1 - wR;
+          int wRPerm = ${filterHeight} - 1 - wR;
 
-          for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
-            int wCPerm = ${n} - 1 - wC;
+            int wCPerm = ${filterWidth} - 1 - wC;
 
             // TO DO: Vec4 over the channelMul
-            for (int dm = 0; dm < ${o}; dm++) {
-              int d2 = d1 * ${o} + dm;
+            for (int dm = 0; dm < ${channelMul}; dm++) {
+              int d2 = d1 * ${channelMul} + dm;
               float xValue = getDy(batch, idyR, idyC, d2);
               float wValue = getW(wRPerm, wCPerm, d1, dm);
               dotProd += xValue * wValue;
@@ -3271,15 +57120,75 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function UQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:p}=a,d=N.computeConv2DInfo(r.shape,p,i,o,l,u,!0),c=new BQ(d);return n.runWebGLProgram(c,[r,s],"float32")}var GQ={kernelName:rm,backendName:"webgl",kernelFunc:UQ};function HQ(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:p}=a,d=N.computeConv2DInfo(p,s.shape,i,o,l,u,!0),c=new VQ(d);return n.runWebGLProgram(c,[r,s],"float32")}var jQ={kernelName:sm,backendName:"webgl",kernelFunc:HQ},qQ=class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode=`
+    `;
+  }
+};
+function depthwiseConv2dNativeBackpropFilter3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);
+  const program = new DepthwiseConv2DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var depthwiseConv2dNativeBackpropFilterConfig2 = {
+  kernelName: DepthwiseConv2dNativeBackpropFilter,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNativeBackpropFilter3
+};
+function depthwiseConv2dNativeBackpropInput3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const program = new DepthwiseConv2DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var depthwiseConv2dNativeBackpropInputConfig2 = {
+  kernelName: DepthwiseConv2dNativeBackpropInput,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNativeBackpropInput3
+};
+var DiagProgram = class {
+  constructor(size) {
+    this.variableNames = ["X"];
+    this.outputShape = [size, size];
+    this.userCode = `
       void main() {
           ivec2 coords = getOutputCoords();
           float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;
           setOutput(val);
       }
-    `}};function KQ(e){let{inputs:t,backend:n}=e,{x:a}=t,r=[...a.shape,...a.shape],s=v.sizeFromShape(a.shape),i=de({inputs:{x:a},backend:n,attrs:{shape:[s]}}),o=new qQ(s),l=n.runWebGLProgram(o,[i],i.dtype),u=de({inputs:{x:l},backend:n,attrs:{shape:r}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(l),u}var XQ={kernelName:im,backendName:"webgl",kernelFunc:KQ},YQ=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let{inHeight:t,inWidth:n,padInfo:a,strideHeight:r,strideWidth:s,filterHeight:i,filterWidth:o,dilationHeight:l,dilationWidth:u}=e,{top:p,left:d}=a;this.userCode=`
-      const ivec2 strides = ivec2(${r}, ${s});
-      const ivec2 pads = ivec2(${p}, ${d});
+    `;
+  }
+};
+function diag3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const outShape = [...x.shape, ...x.shape];
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const flat = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [xSize] } });
+  const program = new DiagProgram(xSize);
+  const res = backend2.runWebGLProgram(program, [flat], flat.dtype);
+  const out = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outShape } });
+  backend2.disposeIntermediateTensorInfo(flat);
+  backend2.disposeIntermediateTensorInfo(res);
+  return out;
+}
+var diagConfig2 = {
+  kernelName: Diag,
+  backendName: "webgl",
+  kernelFunc: diag3
+};
+var Dilation2DProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const { inHeight, inWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth } = convInfo;
+    const { top: padTop, left: padLeft } = padInfo;
+    this.userCode = `
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
       const float neg_infinity = -3.4e38;
 
       void main() {
@@ -3292,14 +57201,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         int wBeg = outTopLeftCorner.y;
 
         float curVal = neg_infinity;
-        for (int h = 0; h < ${i}; h++) {
-          int hIn = hBeg + h * ${l};
+        for (int h = 0; h < ${filterHeight}; h++) {
+          int hIn = hBeg + h * ${dilationHeight};
 
-          if (hIn >= 0 && hIn < ${t}) {
-            for (int w = 0; w < ${o}; w++) {
-              int wIn = wBeg + w * ${u};
+          if (hIn >= 0 && hIn < ${inHeight}) {
+            for (int w = 0; w < ${filterWidth}; w++) {
+              int wIn = wBeg + w * ${dilationWidth};
 
-              if (wIn >= 0 && wIn < ${n}) {
+              if (wIn >= 0 && wIn < ${inWidth}) {
                 float xVal = getX(batch, hIn, wIn, d1);
                 float wVal = getW(h, w, d1);
 
@@ -3315,7 +57224,92 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         float result = curVal;
         setOutput(result);
       }
-    `}};function ZQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a,u=N.computeDilation2DInfo(r.shape,s.shape,i,o,"NHWC",l),p,d=new YQ(u);p=n.runWebGLProgram(d,[r,s],"float32");let c=de({inputs:{x:p},backend:n,attrs:{shape:u.outShape}});return n.disposeIntermediateTensorInfo(p),c}var JQ={kernelName:uc,backendName:"webgl",kernelFunc:ZQ};function QQ(e){let{inputs:t,backend:n,attrs:a}=e,{equation:r}=a,s=t,{allDims:i,summedDims:o,idDims:l}=N.decodeEinsumEquation(r,s.length);N.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:p}=N.getEinsumComputePath(o,l),d=p.length,c=null,h=i.length,m=[];for(let f=0;f<d;++f){for(let g of p[f]){let{permutationIndices:y,expandDims:b}=N.getEinsumPermutation(h,l[g]),x;N.isIdentityPermutation(y)?x=s[g]:(x=In({inputs:{x:s[g]},backend:n,attrs:{perm:y}}),m.push(x));let w=x.shape.slice();for(let I=0;I<b.length;++I)w.splice(b[I],0,1);v.arraysEqual(x.shape,w)||(x=de({inputs:{x},backend:n,attrs:{shape:w}}),m.push(x)),c===null?c=x:(c=Q0({inputs:{a:x,b:c},backend:n}),m.push(c))}f<d-1&&(u[f]>=0&&(c=Pf({inputs:{x:c},backend:n,attrs:{axis:u[f]-(i.length-h),keepDims:!1}}),m.push(c)),h--)}for(let f of m)f!==c&&n.disposeIntermediateTensorInfo(f);return c}var eee={kernelName:om,backendName:"webgl",kernelFunc:QQ},tee="return (x >= 0.0) ? x : (exp(x) - 1.0);",nee=`
+    `;
+  }
+};
+function dilation2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  const convInfo = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+  let out;
+  const program = new Dilation2DProgram(convInfo);
+  out = backend2.runWebGLProgram(program, [x, filter], "float32");
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  backend2.disposeIntermediateTensorInfo(out);
+  return outReshaped;
+}
+var dilation2DConfig2 = {
+  kernelName: Dilation2D,
+  backendName: "webgl",
+  kernelFunc: dilation2D
+};
+function einsum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { equation } = attrs;
+  const tensors = inputs;
+  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);
+  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);
+  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);
+  const nSteps = steps.length;
+  let out = null;
+  let numDimsRemaining = allDims.length;
+  const tensorsToDispose = [];
+  for (let i = 0; i < nSteps; ++i) {
+    for (const idTerm of steps[i]) {
+      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);
+      let x;
+      if (backend_util_exports.isIdentityPermutation(perm)) {
+        x = tensors[idTerm];
+      } else {
+        x = transpose3({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });
+        tensorsToDispose.push(x);
+      }
+      const targetShape = x.shape.slice();
+      for (let k = 0; k < dimsToExpand.length; ++k) {
+        targetShape.splice(dimsToExpand[k], 0, 1);
+      }
+      if (!util_exports.arraysEqual(x.shape, targetShape)) {
+        x = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });
+        tensorsToDispose.push(x);
+      }
+      if (out === null) {
+        out = x;
+      } else {
+        out = multiply3({ inputs: { a: x, b: out }, backend: backend2 });
+        tensorsToDispose.push(out);
+      }
+    }
+    if (i < nSteps - 1) {
+      if (path[i] >= 0) {
+        out = sum4({
+          inputs: { x: out },
+          backend: backend2,
+          attrs: {
+            axis: path[i] - (allDims.length - numDimsRemaining),
+            keepDims: false
+          }
+        });
+        tensorsToDispose.push(out);
+      }
+      numDimsRemaining--;
+    }
+  }
+  for (const tensorInfo of tensorsToDispose) {
+    if (tensorInfo === out) {
+      continue;
+    }
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+  return out;
+}
+var einsumConfig2 = {
+  kernelName: Einsum,
+  backendName: "webgl",
+  kernelFunc: einsum3
+};
+var ELU4 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;
+var ELU_PACKED = `
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -3324,29 +57318,70 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,aee=Ye({opSnippet:tee,packedOpSnippet:nee}),ree={kernelName:Ci,backendName:"webgl",kernelFunc:aee},see="return (b >= 1.0) ? a : a * (b + 1.0);",iee=`
+`;
+var elu5 = unaryKernelFunc2({ opSnippet: ELU4, packedOpSnippet: ELU_PACKED });
+var eluConfig2 = {
+  kernelName: Elu,
+  backendName: "webgl",
+  kernelFunc: elu5
+};
+var ELU_DER = `return (b >= 1.0) ? a : a * (b + 1.0);`;
+var ELU_DER_PACKED = `
   vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));
   return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));
-`,oee=e=>{let{inputs:t,backend:n}=e,{dy:a,y:r}=t,s=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ad(iee,a.shape,r.shape):new xl(see,a.shape,r.shape);return n.runWebGLProgram(s,[a,r],a.dtype)},lee={kernelName:lm,backendName:"webgl",kernelFunc:oee},uee=`
+`;
+var eluGrad2 = (args) => {
+  const { inputs, backend: backend2 } = args;
+  const { dy, y } = inputs;
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(ELU_DER_PACKED, dy.shape, y.shape) : new BinaryOpProgram(ELU_DER, dy.shape, y.shape);
+  return backend2.runWebGLProgram(program, [dy, y], dy.dtype);
+};
+var eluGradConfig3 = {
+  kernelName: EluGrad,
+  backendName: "webgl",
+  kernelFunc: eluGrad2
+};
+var PACKED_EQUAL = `
   return vec4(equal(a, b));
-`,pee="return float(a == b);",cee=pn({opSnippet:pee,packedOpSnippet:uee,dtype:"bool",cpuKernelImpl:KZ}),dee={kernelName:Ol,backendName:"webgl",kernelFunc:cee},hee=`
+`;
+var EQUAL = `return float(a == b);`;
+var equal3 = binaryKernelFunc2({
+  opSnippet: EQUAL,
+  packedOpSnippet: PACKED_EQUAL,
+  dtype: "bool",
+  cpuKernelImpl: equalImplCPU
+});
+var equalConfig2 = {
+  kernelName: Equal,
+  backendName: "webgl",
+  kernelFunc: equal3
+};
+var ERF = `
   // Error function is calculated approximately with elementary function.
   // See "Handbook of Mathematical Functions with Formulas,
   // Graphs, and Mathematical Tables", Abramowitz and Stegun.
-  float p = ${N.ERF_P};
-  float a1 = ${N.ERF_A1};
-  float a2 = ${N.ERF_A2};
-  float a3 = ${N.ERF_A3};
-  float a4 = ${N.ERF_A4};
-  float a5 = ${N.ERF_A5};
+  float p = ${backend_util_exports.ERF_P};
+  float a1 = ${backend_util_exports.ERF_A1};
+  float a2 = ${backend_util_exports.ERF_A2};
+  float a3 = ${backend_util_exports.ERF_A3};
+  float a4 = ${backend_util_exports.ERF_A4};
+  float a5 = ${backend_util_exports.ERF_A5};
 
   float sign = sign(x);
   x = abs(x);
   float t = 1.0 / (1.0 + p * x);
   return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));
-`,mee=Ye({opSnippet:hee}),fee={kernelName:Pl,backendName:"webgl",kernelFunc:mee},gee=Uu+`
+`;
+var erf3 = unaryKernelFunc2({ opSnippet: ERF });
+var erfConfig2 = {
+  kernelName: Erf,
+  backendName: "webgl",
+  kernelFunc: erf3
+};
+var EXP = CHECK_NAN_SNIPPET_UNARY + `
   return exp(x);
-`,yee=`
+`;
+var EXP_PACKED = `
   vec4 result = exp(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : result.r;
@@ -3355,21 +57390,74 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,oE=Ye({opSnippet:gee,packedOpSnippet:yee,cpuKernelImpl:XZ,dtype:"float32"}),bee={kernelName:_i,backendName:"webgl",kernelFunc:oE};function dx(e){let{inputs:t,attrs:n,backend:a}=e,{dim:r}=n,{input:s}=t,i=s.shape.length,o=s.shape.slice(),l=r;return r<0&&(v.assert(-(i+1)<=r,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+r+1),o.splice(l,0,1),de({inputs:{x:s},backend:a,attrs:{shape:o}})}var xee={kernelName:Ll,backendName:"webgl",kernelFunc:dx},lI="return exp(x) - 1.0;",vee=Ye({opSnippet:lI,packedOpSnippet:lI,cpuKernelImpl:YZ}),wee={kernelName:zl,backendName:"webgl",kernelFunc:vee},uI=class{constructor(e,t,n){this.variableNames=["real","imag"];let a=t[1];this.outputShape=t;let r=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,s=n?`${a}.0`:"1.0",i;if(e==="real")i="return real * expR - imag * expI;";else if(e==="imag")i="return real * expI + imag * expR;";else throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);this.userCode=`
-      const float exponentMultiplier = ${r};
+`;
+var exp3 = unaryKernelFunc2({
+  opSnippet: EXP,
+  packedOpSnippet: EXP_PACKED,
+  cpuKernelImpl: expImplCPU,
+  dtype: "float32"
+});
+var expConfig2 = {
+  kernelName: Exp,
+  backendName: "webgl",
+  kernelFunc: exp3
+};
+function expandDims4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { dim } = attrs;
+  const { input: input2 } = inputs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape4({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig2 = {
+  kernelName: ExpandDims,
+  backendName: "webgl",
+  kernelFunc: expandDims4
+};
+var EXPM1 = `return exp(x) - 1.0;`;
+var expm13 = unaryKernelFunc2({ opSnippet: EXPM1, packedOpSnippet: EXPM1, cpuKernelImpl: expm1ImplCPU });
+var expm1Config2 = {
+  kernelName: Expm1,
+  backendName: "webgl",
+  kernelFunc: expm13
+};
+var FFTProgram = class {
+  constructor(component, inputShape, inverse) {
+    this.variableNames = ["real", "imag"];
+    const innerDim = inputShape[1];
+    this.outputShape = inputShape;
+    const exponentMultiplierSnippet = inverse ? `2.0 * ${Math.PI}` : `-2.0 * ${Math.PI}`;
+    const resultDenominator = inverse ? `${innerDim}.0` : "1.0";
+    let opString;
+    if (component === "real") {
+      opString = "return real * expR - imag * expI;";
+    } else if (component === "imag") {
+      opString = "return real * expI + imag * expR;";
+    } else {
+      throw new Error(`FFT component must be either "real" or "imag", got ${component}.`);
+    }
+    this.userCode = `
+      const float exponentMultiplier = ${exponentMultiplierSnippet};
 
       float unaryOpComplex(float real, float expR, float imag, float expI) {
-        ${i}
+        ${opString}
       }
 
       float mulMatDFT(int batch, int index) {
-        float indexRatio = float(index) / float(${a});
+        float indexRatio = float(index) / float(${innerDim});
         float exponentMultiplierTimesIndexRatio =
             exponentMultiplier * indexRatio;
 
         float result = 0.0;
 
-        for (int i = 0; i < ${a}; i++) {
+        for (int i = 0; i < ${innerDim}; i++) {
           // x = (-2|2 * PI / N) * index * i;
           float x = exponentMultiplierTimesIndexRatio * float(i);
           float expR = cos(x);
@@ -3378,7 +57466,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           float imag = getImag(batch, i);
 
           result +=
-              unaryOpComplex(real, expR, imag, expI) / ${s};
+              unaryOpComplex(real, expR, imag, expI) / ${resultDenominator};
         }
 
         return result;
@@ -3388,26 +57476,126 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         ivec2 coords = getOutputCoords();
         setOutput(mulMatDFT(coords[0], coords[1]));
       }
-    `}};function lE(e,t,n){let a=n.texData.get(e.dataId),r=v.sizeFromShape(e.shape),s=e.shape[e.shape.length-1],i=r/s,o=de({inputs:{x:e},backend:n,attrs:{shape:[i,s]}}),l=o.shape,u=new uI("real",l,t),p=new uI("imag",l,t),d=[{dataId:a.complexTensorInfos.real.dataId,dtype:a.complexTensorInfos.real.dtype,shape:l},{dataId:a.complexTensorInfos.imag.dataId,dtype:a.complexTensorInfos.imag.dtype,shape:l}],c=n.runWebGLProgram(u,d,"float32"),h=n.runWebGLProgram(p,d,"float32"),m=Is({inputs:{real:c,imag:h},backend:n});n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h);let f=de({inputs:{x:m},backend:n,attrs:{shape:e.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(m),f}function kee(e){let{inputs:t,backend:n}=e,{input:a}=t;return lE(a,!1,n)}var Iee={kernelName:um,backendName:"webgl",kernelFunc:kee},See=class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function fftImpl2(x, inverse, backend2) {
+  const xData = backend2.texData.get(x.dataId);
+  const inputSize = util_exports.sizeFromShape(x.shape);
+  const innerDimensionSize = x.shape[x.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [batch, innerDimensionSize] } });
+  const xShape = input2D.shape;
+  const realProgram = new FFTProgram("real", xShape, inverse);
+  const imagProgram = new FFTProgram("imag", xShape, inverse);
+  const inputs = [
+    {
+      dataId: xData.complexTensorInfos.real.dataId,
+      dtype: xData.complexTensorInfos.real.dtype,
+      shape: xShape
+    },
+    {
+      dataId: xData.complexTensorInfos.imag.dataId,
+      dtype: xData.complexTensorInfos.imag.dtype,
+      shape: xShape
+    }
+  ];
+  const realPart = backend2.runWebGLProgram(realProgram, inputs, "float32");
+  const imagPart = backend2.runWebGLProgram(imagProgram, inputs, "float32");
+  const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(realPart);
+  backend2.disposeIntermediateTensorInfo(imagPart);
+  const complexOutputReshaped = reshape4({ inputs: { x: complexOutput }, backend: backend2, attrs: { shape: x.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(complexOutput);
+  return complexOutputReshaped;
+}
+function fft3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  return fftImpl2(input2, false, backend2);
+}
+var fftConfig2 = {
+  kernelName: FFT,
+  backendName: "webgl",
+  kernelFunc: fft3
+};
+var FillProgram = class {
+  constructor(shape, value) {
+    this.outputShape = [];
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.variableNames = ["x"];
+    this.outputShape = shape;
+    this.userCode = `
       void main() {
         // Input can be obtained from uniform value.
         setOutput(value);
       }
-    `}};function sd(e){let{backend:t,attrs:n}=e,{shape:a,value:r}=n,{dtype:s}=n;if(s=s||v.inferDtype(r),s==="string"){let i=v.getArrayFromDType(s,v.sizeFromShape(a));return i.fill(r),t.makeTensorInfo(a,s,i)}else{let i=new See(a,r),o=[[r]];return t.runWebGLProgram(i,[],s,o)}}var Tee={kernelName:pc,backendName:"webgl",kernelFunc:sd},Nee=class{constructor(e){this.variableNames=["Image"],this.outputShape=[];let t=e[2];this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function fill3(args) {
+  const { backend: backend2, attrs } = args;
+  const { shape, value } = attrs;
+  let { dtype } = attrs;
+  dtype = dtype || util_exports.inferDtype(value);
+  if (dtype === "string") {
+    const values = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(shape));
+    values.fill(value);
+    return backend2.makeTensorInfo(shape, dtype, values);
+  } else {
+    const program = new FillProgram(shape, value);
+    const customValues = [[value]];
+    return backend2.runWebGLProgram(program, [], dtype, customValues);
+  }
+}
+var fillConfig2 = {
+  kernelName: Fill,
+  backendName: "webgl",
+  kernelFunc: fill3
+};
+var FlipLeftRightProgram = class {
+  constructor(imageShape) {
+    this.variableNames = ["Image"];
+    this.outputShape = [];
+    const imageWidth = imageShape[2];
+    this.outputShape = imageShape;
+    this.userCode = `
         void main() {
           ivec4 coords = getOutputCoords();
           int x = coords[2];
 
-          int coordX = ${t} - x - 1;
+          int coordX = ${imageWidth} - x - 1;
           float outputValue;
-          if(coordX >= 0 && coordX < ${t}) {
+          if(coordX >= 0 && coordX < ${imageWidth}) {
             outputValue = getImage(coords[0], coords[1], coordX, coords[3]);
           } else {
             outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);
           }
           setOutput(outputValue);
         }
-    `}},Cee={kernelName:Wl,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{let{image:n}=e,a=t,r=new Nee(n.shape);return a.runWebGLProgram(r,[n],n.dtype)}},pI="return floor(x);",_ee=Ye({opSnippet:pI,packedOpSnippet:pI,cpuKernelImpl:ZZ}),Eee={kernelName:Ei,backendName:"webgl",kernelFunc:_ee},Aee=`
+    `;
+  }
+};
+var flipLeftRightConfig2 = {
+  kernelName: FlipLeftRight,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const webglBackend = backend2;
+    const program = new FlipLeftRightProgram(image2.shape);
+    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype);
+    return output;
+  }
+};
+var FLOOR = `return floor(x);`;
+var floor3 = unaryKernelFunc2({ opSnippet: FLOOR, packedOpSnippet: FLOOR, cpuKernelImpl: floorImplCPU });
+var floorConfig2 = {
+  kernelName: Floor,
+  backendName: "webgl",
+  kernelFunc: floor3
+};
+var INT_DIV = `
   float s = sign(a) * sign(b);
   int ia = round(a);
   int ib = round(b);
@@ -3417,7 +57605,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
   } else {
     return NAN;
   }
-`,$ee=`
+`;
+var INT_DIV_PACKED = `
   ivec4 ia = round(a);
   ivec4 ib = round(b);
   bvec4 cond = notEqual(ib, ivec4(0));
@@ -3438,15 +57627,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
     result[3] = idiv(ia[3], ib[3], s[3]);
   }
   return vec4(result);
-`,Fee=pn({opSnippet:Aee,packedOpSnippet:$ee,dtype:"int32"}),Dee={kernelName:Ai,backendName:"webgl",kernelFunc:Fee},Ree=class{constructor(e){this.variableNames=["A"];let t=Cn(),[n,a]=e;this.outputShape=e,this.userCode=`
+`;
+var floorDiv3 = binaryKernelFunc2({ opSnippet: INT_DIV, packedOpSnippet: INT_DIV_PACKED, dtype: "int32" });
+var floorDivConfig2 = {
+  kernelName: FloorDiv,
+  backendName: "webgl",
+  kernelFunc: floorDiv3
+};
+var FromPixelsProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    const glsl = getGlslDifferences();
+    const [height, width] = outputShape;
+    this.outputShape = outputShape;
+    this.userCode = `
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
         int texC = coords[1];
         int depth = coords[2];
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${a}.0, ${n}.0);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${width}.0, ${height}.0);
 
-        vec4 values = ${t.texture2D}(A, uv);
+        vec4 values = ${glsl.texture2D}(A, uv);
         float value;
         if (depth == 0) {
           value = values.r;
@@ -3460,7 +57662,18 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         setOutput(floor(value * 255.0 + 0.5));
       }
-    `}},Mee=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let t=Cn(),[n,a]=e;this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+var FromPixelsPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    const glsl = getGlslDifferences();
+    const [height, width] = outputShape;
+    this.outputShape = outputShape;
+    this.userCode = `
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
@@ -3475,8 +57688,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
             depth = coords[2] + col;
 
             vec2 uv = (vec2(texC, texR) + halfCR) /
-                       vec2(${a}.0, ${n}.0);
-            vec4 values = ${t.texture2D}(A, uv);
+                       vec2(${width}.0, ${height}.0);
+            vec4 values = ${glsl.texture2D}(A, uv);
             float value;
             if (depth == 0) {
               value = values.r;
@@ -3492,41 +57705,442 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
           }
         }
 
-        ${t.output} = result;
+        ${glsl.output} = result;
       }
-    `}},Pee={kernelName:yh,backendName:"webgl",kernelFunc:Oee},qo,yb=H().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");function Oee(e){let{inputs:t,backend:n,attrs:a}=e,{pixels:r}=t,{numChannels:s}=a,i=typeof HTMLVideoElement!="undefined"&&r instanceof HTMLVideoElement,o=typeof HTMLImageElement!="undefined"&&r instanceof HTMLImageElement,[l,u]=i?[r.videoWidth,r.videoHeight]:[r.width,r.height],p=[u,l],d=[u,l,s];if(o||i){let f=H().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");(qo==null||f!==yb)&&(yb=f,qo=document.createElement("canvas").getContext("2d",{willReadFrequently:yb})),qo.canvas.width=l,qo.canvas.height=u,qo.drawImage(r,0,0,l,u),r=qo.canvas}let c=n.makeTensorInfo(p,"int32");n.texData.get(c.dataId).usage=pa.PIXELS,n.gpgpu.uploadPixelDataToTexture(n.getTexture(c.dataId),r);let h=H().getBool("WEBGL_PACK")?new Mee(d):new Ree(d),m=n.runWebGLProgram(h,[c],"int32");return n.disposeData(c.dataId),m}function Lee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=N.convertConv2DDataFormat(p),g=N.computeConv2DInfo(r.shape,s.shape,l,d,u,c,!1,f),y,b=[],x=i!=null,w=o!=null,I=h==="leakyrelu",T=()=>{let E=[r,s],A=(R,F)=>{if(F==="NCHW"&&R.shape.length===1&&R.shape[0]!==1){let S=de({inputs:{x:R},backend:n,attrs:{shape:[R.shape[0],1,1]}});return b.push(S),S}return R};if(x&&E.push(A(i,p)),w&&E.push(A(o,p)),I){let R=n.makeTensorInfo([],"float32",v.createScalarValue(m,"float32"));E.push(R),b.push(R)}return E};if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type==="SAME"||g.padInfo.type==="VALID"))y=nE({x:r,filter:s,convInfo:g,backend:n,bias:i,activation:h,preluActivationWeights:o,leakyreluAlpha:m});else if(g.strideWidth<=2&&f==="channelsLast"&&H().getBool("WEBGL_EXP_CONV")){let E=h?ec(h,!0):null,A=new tE(g,x,E,w,I),R=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],F=T();y=n.runWebGLProgram(A,F,"float32",R)}else if(H().getBool("WEBGL_CONV_IM2COL"))y=aE({x:r,filter:s,convInfo:g,backend:n,bias:i,activation:h,preluActivationWeights:o,leakyreluAlpha:m});else{let E=h?ec(h,!1):null,A=new eE(g,x,E,w,I),R=T();y=n.runWebGLProgram(A,R,"float32")}let C=de({inputs:{x:y},backend:n,attrs:{shape:g.outShape}});return b.push(y),b.forEach(E=>n.disposeIntermediateTensorInfo(E)),C}var zee={kernelName:Qs,backendName:"webgl",kernelFunc:Lee};function Wee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dimRoundingMode:d,activation:c,leakyreluAlpha:h}=a,m=[],f=p;f==null&&(f=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(l,f),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${l} and dilations '${f}'`);let g=N.computeConv2DInfo(r.shape,s.shape,l,f,u,d,!0),y=H().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,b=c?ec(c,y):null,x=[r,s],w=i!=null,I=o!=null,T=c==="leakyrelu";if(w&&x.push(i),I&&x.push(o),T){let R=n.makeTensorInfo([],"float32",v.createScalarValue(h,"float32"));x.push(R),m.push(R)}let C;y?C=new iE(g,w,b,I,T):C=new sE(g,w,b,I,T);let E=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],A=n.runWebGLProgram(C,x,"float32",E);return m.forEach(R=>n.disposeIntermediateTensorInfo(R)),A}var Bee={kernelName:ei,backendName:"webgl",kernelFunc:Wee},Vee=class{constructor(e,t,n,a){this.sliceDim=e,this.strides=t,this.paramsShape=a,this.variableNames=["x","indices"],this.outputShape=n;let r=gt(n.length),s=`
-    int index;`;for(let i=0;i<this.sliceDim;i++)s+=`
-          index = round(getIndices(coords[0], ${i}));
+    `;
+  }
+};
+var fromPixelsConfig = {
+  kernelName: FromPixels,
+  backendName: "webgl",
+  kernelFunc: fromPixels2
+};
+var fromPixels2DContext2;
+var willReadFrequently = env().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");
+function fromPixels2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  let { pixels } = inputs;
+  const { numChannels } = attrs;
+  const isVideo = typeof HTMLVideoElement !== "undefined" && pixels instanceof HTMLVideoElement;
+  const isImage = typeof HTMLImageElement !== "undefined" && pixels instanceof HTMLImageElement;
+  const [width, height] = isVideo ? [
+    pixels.videoWidth,
+    pixels.videoHeight
+  ] : [pixels.width, pixels.height];
+  const texShape = [height, width];
+  const outShape = [height, width, numChannels];
+  if (isImage || isVideo) {
+    const newWillReadFrequently = env().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");
+    if (fromPixels2DContext2 == null || newWillReadFrequently !== willReadFrequently) {
+      willReadFrequently = newWillReadFrequently;
+      fromPixels2DContext2 = document.createElement("canvas").getContext("2d", { willReadFrequently });
+    }
+    fromPixels2DContext2.canvas.width = width;
+    fromPixels2DContext2.canvas.height = height;
+    fromPixels2DContext2.drawImage(pixels, 0, 0, width, height);
+    pixels = fromPixels2DContext2.canvas;
+  }
+  const tempPixelHandle = backend2.makeTensorInfo(texShape, "int32");
+  backend2.texData.get(tempPixelHandle.dataId).usage = TextureUsage.PIXELS;
+  backend2.gpgpu.uploadPixelDataToTexture(backend2.getTexture(tempPixelHandle.dataId), pixels);
+  const program = env().getBool("WEBGL_PACK") ? new FromPixelsPackedProgram(outShape) : new FromPixelsProgram(outShape);
+  const res = backend2.runWebGLProgram(program, [tempPixelHandle], "int32");
+  backend2.disposeData(tempPixelHandle.dataId);
+  return res;
+}
+function fusedConv2d(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  let out;
+  const intermediates = [];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const prepareInputs = () => {
+    const inputs2 = [x, filter];
+    const alignInputWithDataFormat = (input2, dataFormat2) => {
+      if (dataFormat2 === "NCHW" && input2.shape.length === 1 && input2.shape[0] !== 1) {
+        const alignedInput = reshape4({
+          inputs: { x: input2 },
+          backend: backend2,
+          attrs: { shape: [input2.shape[0], 1, 1] }
+        });
+        intermediates.push(alignedInput);
+        return alignedInput;
+      }
+      return input2;
+    };
+    if (hasBias) {
+      inputs2.push(alignInputWithDataFormat(bias, dataFormat));
+    }
+    if (hasPreluActivationWeights) {
+      inputs2.push(alignInputWithDataFormat(preluActivationWeights, dataFormat));
+    }
+    if (hasLeakyreluAlpha) {
+      const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+      inputs2.push($leakyreluAlpha);
+      intermediates.push($leakyreluAlpha);
+    }
+    return inputs2;
+  };
+  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === "SAME" || convInfo.padInfo.type === "VALID")) {
+    out = conv2dByMatMul({
+      x,
+      filter,
+      convInfo,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+  } else if (convInfo.strideWidth <= 2 && $dataFormat === "channelsLast" && env().getBool("WEBGL_EXP_CONV")) {
+    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;
+    const program = new Conv2DPackedProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const customValues = [
+      [convInfo.padInfo.top, convInfo.padInfo.left],
+      [convInfo.strideHeight, convInfo.strideWidth],
+      [convInfo.dilationHeight, convInfo.dilationWidth],
+      [convInfo.inHeight, convInfo.inWidth]
+    ];
+    const inputs2 = prepareInputs();
+    out = backend2.runWebGLProgram(program, inputs2, "float32", customValues);
+  } else if (env().getBool("WEBGL_CONV_IM2COL")) {
+    out = conv2dWithIm2Row({
+      x,
+      filter,
+      convInfo,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+  } else {
+    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, false) : null;
+    const program = new Conv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const inputs2 = prepareInputs();
+    out = backend2.runWebGLProgram(program, inputs2, "float32");
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  intermediates.push(out);
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return outReshaped;
+}
+var fusedConv2DConfig2 = {
+  kernelName: FusedConv2D,
+  backendName: "webgl",
+  kernelFunc: fusedConv2d
+};
+function fusedDepthwiseConv2D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const intermediates = [];
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const shouldPackDepthwiseConv = env().getBool("WEBGL_PACK_DEPTHWISECONV") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1;
+  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, shouldPackDepthwiseConv) : null;
+  const programInputs = [x, filter];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  if (hasBias) {
+    programInputs.push(bias);
+  }
+  if (hasPreluActivationWeights) {
+    programInputs.push(preluActivationWeights);
+  }
+  if (hasLeakyreluAlpha) {
+    const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+    programInputs.push($leakyreluAlpha);
+    intermediates.push($leakyreluAlpha);
+  }
+  let program;
+  if (shouldPackDepthwiseConv) {
+    program = new DepthwiseConvPacked2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  } else {
+    program = new DepthwiseConv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  }
+  const customValues = [
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inHeight, convInfo.inWidth]
+  ];
+  const result = backend2.runWebGLProgram(program, programInputs, "float32", customValues);
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var fusedDepthwiseConv2DConfig2 = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "webgl",
+  kernelFunc: fusedDepthwiseConv2D2
+};
+var GatherNDProgram = class {
+  constructor(sliceDim, strides, shape, paramsShape) {
+    this.sliceDim = sliceDim;
+    this.strides = strides;
+    this.paramsShape = paramsShape;
+    this.variableNames = ["x", "indices"];
+    this.outputShape = shape;
+    const dtype = getCoordsDataType(shape.length);
+    let mainLoop = `
+    int index;`;
+    for (let j = 0; j < this.sliceDim; j++) {
+      mainLoop += `
+          index = round(getIndices(coords[0], ${j}));
           out_of_bounds = out_of_bounds || index < 0;
-          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[i]};
-          flattenIndex += index * ${this.strides[i]};`;this.userCode=`
+          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[j]};
+          flattenIndex += index * ${this.strides[j]};`;
+    }
+    this.userCode = `
          void main() {
-          ${r} coords = getOutputCoords();
+          ${dtype} coords = getOutputCoords();
           int flattenIndex = 0;
           bool out_of_bounds = false;
 
-          ${s}
+          ${mainLoop}
 
           setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));
         }
-      `}};function Uee(e){let{inputs:t,backend:n}=e,{params:a,indices:r}=t,s=r.shape,i=s[s.length-1],o=v.sizeFromShape(a.shape),[l,u,p,d]=N.prepareAndValidate(a,r),c=de({inputs:{x:r},backend:n,attrs:{shape:[u,i]}}),h=de({inputs:{x:a},backend:n,attrs:{shape:[v.sizeFromShape(a.shape)/p,p]}});if(n.shouldExecuteOnCPU([a,r])||a.dtype==="string"){let y=n.readSync(r.dataId),b=n.bufferSync(a),x=JZ(y,b,a.dtype,u,i,p,d,a.shape,o);return n.makeTensorInfo(l,a.dtype,x.values)}let m=new Vee(i,d,[u,p],a.shape),f=n.runWebGLProgram(m,[h,c],h.dtype),g=de({inputs:{x:f},backend:n,attrs:{shape:l}});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),g}var Gee={kernelName:Vl,backendName:"webgl",kernelFunc:Uee},Hee=class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;let n=gt(this.rank),a=jee(e,2);this.userCode=`
+      `;
+  }
+};
+function gatherNd2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { params, indices } = inputs;
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const paramsSize = util_exports.sizeFromShape(params.shape);
+  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);
+  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numSlices, sliceRank] } });
+  const flattenX = reshape4({
+    inputs: { x: params },
+    backend: backend2,
+    attrs: { shape: [util_exports.sizeFromShape(params.shape) / sliceSize, sliceSize] }
+  });
+  if (backend2.shouldExecuteOnCPU([params, indices]) || params.dtype === "string") {
+    const indicesData = backend2.readSync(indices.dataId);
+    const paramsBuf = backend2.bufferSync(params);
+    const outValue = gatherNdImplCPU(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);
+    return backend2.makeTensorInfo(resultShape, params.dtype, outValue.values);
+  }
+  const program = new GatherNDProgram(sliceRank, strides, [numSlices, sliceSize], params.shape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices], flattenX.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: resultShape } });
+  backend2.disposeIntermediateTensorInfo(flattenIndices);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(res);
+  return reshaped;
+}
+var gatherNdConfig2 = {
+  kernelName: GatherNd,
+  backendName: "webgl",
+  kernelFunc: gatherNd2
+};
+var GatherProgram = class {
+  constructor(aShape, outputShape) {
+    this.variableNames = ["A", "indices"];
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const sourceCoords = getSourceCoords2(aShape, 2);
+    this.userCode = `
       void main() {
-        ${n} resRC = getOutputCoords();
+        ${dtype} resRC = getOutputCoords();
         int index = int(getIndices(resRC.x, resRC.z));
-        float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;
-        setOutput(inBounds * getA(${a}));
+        float inBounds = (index >= 0) && (index < ${aShape[2]}) ? 1.0 : 0.0;
+        setOutput(inBounds * getA(${sourceCoords}));
       }
-    `}};function jee(e,t){let n=["resRC.x","resRC.y","resRC.z","resRC.w"],a=[];for(let r=0;r<e.length;r++)r===2?a.push("index"):a.push(`${n[r]}`);return a.join()}function uE(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,indices:s}=t,{axis:i,batchDims:o}=a,l=v.parseAxisParam(i,r.shape)[0];if(H().get("DEBUG")){let b=n.readSync(s.dataId),x=r.shape[l];for(let w=0;w<b.length;++w){let I=b[w];v.assert(I<=x-1&&I>=0,()=>`GatherV2: the index value ${I} is not in [0, ${x-1}]`)}}let u=N.segment_util.collectGatherOpShapeInfo(r,s,l,o),p=v.sizeFromShape(s.shape),d=[],c=de({inputs:{x:r},backend:n,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),h=de({inputs:{x:s},backend:n,attrs:{shape:[u.batchSize,p/u.batchSize]}});d.push(c),d.push(h);let m=[u.batchSize,u.outerSize,p/u.batchSize,u.sliceSize];if(n.shouldExecuteOnCPU([r,s])||r.dtype==="string"){let b=n.bufferSync(h),x=n.bufferSync(c),w=QZ(x,b,m);return d.forEach(I=>n.disposeIntermediateTensorInfo(I)),n.makeTensorInfo(u.outputShape,w.dtype,w.values)}let f=new Hee(c.shape,m),g=n.runWebGLProgram(f,[c,h],c.dtype);d.push(g);let y=de({inputs:{x:g},backend:n,attrs:{shape:u.outputShape}});return d.forEach(b=>n.disposeIntermediateTensorInfo(b)),y}var qee={kernelName:Bl,backendName:"webgl",kernelFunc:uE},Kee="return float(a > b);",Xee=`
+    `;
+  }
+};
+function getSourceCoords2(aShape, axis) {
+  const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"];
+  const sourceCoords = [];
+  for (let i = 0; i < aShape.length; i++) {
+    if (i === 2) {
+      sourceCoords.push("index");
+    } else {
+      sourceCoords.push(`${currentCoords[i]}`);
+    }
+  }
+  return sourceCoords.join();
+}
+function gatherV22(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  if (env().get("DEBUG")) {
+    const indicesVals = backend2.readSync(indices.dataId);
+    const axisDim = x.shape[parsedAxis];
+    for (let i = 0; i < indicesVals.length; ++i) {
+      const index = indicesVals[i];
+      util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+    }
+  }
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const toDispose = [];
+  const flattenX = reshape4({
+    inputs: { x },
+    backend: backend2,
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    }
+  });
+  const flattenIndex = reshape4({
+    inputs: { x: indices },
+    backend: backend2,
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }
+  });
+  toDispose.push(flattenX);
+  toDispose.push(flattenIndex);
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  if (backend2.shouldExecuteOnCPU([x, indices]) || x.dtype === "string") {
+    const indicesBuf = backend2.bufferSync(flattenIndex);
+    const xBuf = backend2.bufferSync(flattenX);
+    const outBuf = gatherV2ImplCPU(xBuf, indicesBuf, flattenOutputShape);
+    toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);
+  }
+  const program = new GatherProgram(flattenX.shape, flattenOutputShape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndex], flattenX.dtype);
+  toDispose.push(res);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: shapeInfo.outputShape } });
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return reshaped;
+}
+var gatherV2Config2 = {
+  kernelName: GatherV2,
+  backendName: "webgl",
+  kernelFunc: gatherV22
+};
+var GREATER = `return float(a > b);`;
+var GREATER_PACKED = `
   return vec4(greaterThan(a, b));
-`,Yee=pn({opSnippet:Kee,packedOpSnippet:Xee,cpuKernelImpl:e7,dtype:"bool"}),Zee={kernelName:Ul,backendName:"webgl",kernelFunc:Yee},Jee="return float(a >= b);",Qee=`
+`;
+var greater4 = binaryKernelFunc2({
+  opSnippet: GREATER,
+  packedOpSnippet: GREATER_PACKED,
+  cpuKernelImpl: greaterImplCPU,
+  dtype: "bool"
+});
+var greaterConfig2 = {
+  kernelName: Greater,
+  backendName: "webgl",
+  kernelFunc: greater4
+};
+var GREATER_EQUAL = `return float(a >= b);`;
+var GREATER_EQUAL_PACKED = `
   return vec4(greaterThanEqual(a, b));
-`,ete=pn({opSnippet:Jee,packedOpSnippet:Qee,dtype:"bool",cpuKernelImpl:t7}),tte={kernelName:Fi,backendName:"webgl",kernelFunc:ete};function nte(e){let{inputs:t,backend:n}=e,{input:a}=t;return lE(a,!0,n)}var ate={kernelName:pm,backendName:"webgl",kernelFunc:nte},rte="return float(!isnan(x) && !isinf(x));",ste=Ye({opSnippet:rte,dtype:"bool"}),ite={kernelName:Gl,backendName:"webgl",kernelFunc:ste},ote="return float(isinf(x));",lte=Ye({opSnippet:ote,dtype:"bool"}),ute={kernelName:Hl,backendName:"webgl",kernelFunc:lte},pte="return float(isnan(x));",cte=Ye({opSnippet:pte,dtype:"bool"}),dte={kernelName:jl,backendName:"webgl",kernelFunc:cte},hte="return float(a < b);",mte=`
+`;
+var greaterEqual3 = binaryKernelFunc2({
+  opSnippet: GREATER_EQUAL,
+  packedOpSnippet: GREATER_EQUAL_PACKED,
+  dtype: "bool",
+  cpuKernelImpl: greaterEqualImplCPU
+});
+var greaterEqualConfig2 = {
+  kernelName: GreaterEqual,
+  backendName: "webgl",
+  kernelFunc: greaterEqual3
+};
+function ifft3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  return fftImpl2(input2, true, backend2);
+}
+var ifftConfig2 = {
+  kernelName: IFFT,
+  backendName: "webgl",
+  kernelFunc: ifft3
+};
+var IS_FINITE = `return float(!isnan(x) && !isinf(x));`;
+var isFinite4 = unaryKernelFunc2({ opSnippet: IS_FINITE, dtype: "bool" });
+var isFiniteConfig2 = {
+  kernelName: IsFinite,
+  backendName: "webgl",
+  kernelFunc: isFinite4
+};
+var IS_INF = `return float(isinf(x));`;
+var isInf3 = unaryKernelFunc2({ opSnippet: IS_INF, dtype: "bool" });
+var isInfConfig2 = {
+  kernelName: IsInf,
+  backendName: "webgl",
+  kernelFunc: isInf3
+};
+var IS_NAN = `return float(isnan(x));`;
+var isNaN4 = unaryKernelFunc2({ opSnippet: IS_NAN, dtype: "bool" });
+var isNaNConfig2 = {
+  kernelName: IsNan,
+  backendName: "webgl",
+  kernelFunc: isNaN4
+};
+var LESS = `return float(a < b);`;
+var LESS_PACKED = `
   return vec4(lessThan(a, b));
-`,fte=pn({opSnippet:hte,packedOpSnippet:mte,cpuKernelImpl:n7,dtype:"bool"}),gte={kernelName:ql,backendName:"webgl",kernelFunc:fte},yte="return float(a <= b);",bte=`
+`;
+var less4 = binaryKernelFunc2({
+  opSnippet: LESS,
+  packedOpSnippet: LESS_PACKED,
+  cpuKernelImpl: lessImplCPU,
+  dtype: "bool"
+});
+var lessConfig2 = {
+  kernelName: Less,
+  backendName: "webgl",
+  kernelFunc: less4
+};
+var LESS_EQUAL = `return float(a <= b);`;
+var LESS_EQUAL_PACKED = `
   return vec4(lessThanEqual(a, b));
-`,xte=pn({opSnippet:yte,packedOpSnippet:bte,cpuKernelImpl:a7,dtype:"bool"}),vte={kernelName:Kl,backendName:"webgl",kernelFunc:xte};function wte(e){let{backend:t,attrs:n}=e,{start:a,stop:r,num:s}=n,i=r7(a,r,s);return t.makeTensorInfo([i.length],"float32",i)}var kte={kernelName:dm,backendName:"webgl",kernelFunc:wte},Ite=Uu+`
+`;
+var lessEqual3 = binaryKernelFunc2({
+  opSnippet: LESS_EQUAL,
+  packedOpSnippet: LESS_EQUAL_PACKED,
+  cpuKernelImpl: lessEqualImplCPU,
+  dtype: "bool"
+});
+var lessEqualConfig2 = {
+  kernelName: LessEqual,
+  backendName: "webgl",
+  kernelFunc: lessEqual3
+};
+function linSpace2(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, num } = attrs;
+  const outVals = linSpaceImplCPU(start, stop, num);
+  return backend2.makeTensorInfo([outVals.length], "float32", outVals);
+}
+var linSpaceConfig2 = {
+  kernelName: LinSpace,
+  backendName: "webgl",
+  kernelFunc: linSpace2
+};
+var LOG = CHECK_NAN_SNIPPET_UNARY + `
   return x < 0.0 ? 0./0. : log(x);
-`,Ste=`
+`;
+var LOG_PACKED = `
   vec4 result = log(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);
@@ -3534,18 +58148,75 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
   result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);
   result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);
   return result;
-`,Tte=Ye({opSnippet:Ite,packedOpSnippet:Ste,cpuKernelImpl:s7}),Nte={kernelName:Mi,backendName:"webgl",kernelFunc:Tte},Cte=Uu+`
+`;
+var log4 = unaryKernelFunc2({ opSnippet: LOG, packedOpSnippet: LOG_PACKED, cpuKernelImpl: logImplCPU });
+var logConfig2 = {
+  kernelName: Log,
+  backendName: "webgl",
+  kernelFunc: log4
+};
+var LOG1P = CHECK_NAN_SNIPPET_UNARY + `
   return log(1.0 + x);
-`,_te=Ye({opSnippet:Cte}),Ete={kernelName:Xl,backendName:"webgl",kernelFunc:_te},Ate="return float(a >= 1.0 && b >= 1.0);",$te=`
+`;
+var log1p3 = unaryKernelFunc2({ opSnippet: LOG1P });
+var log1pConfig2 = {
+  kernelName: Log1p,
+  backendName: "webgl",
+  kernelFunc: log1p3
+};
+var LOGICAL_AND = `return float(a >= 1.0 && b >= 1.0);`;
+var LOGICAL_AND_PACKED = `
   return vec4(
     vec4(greaterThanEqual(a, vec4(1.0))) *
     vec4(greaterThanEqual(b, vec4(1.0))));
-`,Fte=pn({opSnippet:Ate,packedOpSnippet:$te,dtype:"bool"}),Dte={kernelName:Yl,backendName:"webgl",kernelFunc:Fte},Rte="return float(!(x >= 1.0));",Mte=Ye({opSnippet:Rte}),Pte={kernelName:Zl,backendName:"webgl",kernelFunc:Mte},Ote="return float(a >= 1.0 || b >= 1.0);",Lte=`
+`;
+var logicalAnd3 = binaryKernelFunc2({
+  opSnippet: LOGICAL_AND,
+  packedOpSnippet: LOGICAL_AND_PACKED,
+  dtype: "bool"
+});
+var logicalAndConfig2 = {
+  kernelName: LogicalAnd,
+  backendName: "webgl",
+  kernelFunc: logicalAnd3
+};
+var LOGICAL_NOT = `return float(!(x >= 1.0));`;
+var logicalNot3 = unaryKernelFunc2({ opSnippet: LOGICAL_NOT });
+var logicalNotConfig2 = {
+  kernelName: LogicalNot,
+  backendName: "webgl",
+  kernelFunc: logicalNot3
+};
+var LOGICAL_OR = `return float(a >= 1.0 || b >= 1.0);`;
+var LOGICAL_OR_PACKED = `
   return min(
     vec4(greaterThanEqual(a, vec4(1.0))) +
     vec4(greaterThanEqual(b, vec4(1.0))),
     vec4(1.0));
-`,zte=pn({opSnippet:Ote,packedOpSnippet:Lte,dtype:"bool"}),Wte={kernelName:Jl,backendName:"webgl",kernelFunc:zte},Bte=class{constructor(e,t,n,a,r){this.variableNames=["x"],this.outputShape=[];let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${a}) * sum`;r===.5?o=`inversesqrt(${l})`:r===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${r}));`,this.userCode=`
+`;
+var logicalOr3 = binaryKernelFunc2({ opSnippet: LOGICAL_OR, packedOpSnippet: LOGICAL_OR_PACKED, dtype: "bool" });
+var logicalOrConfig2 = {
+  kernelName: LogicalOr,
+  backendName: "webgl",
+  kernelFunc: logicalOr3
+};
+var LRNProgram = class {
+  constructor(xShape, radius, bias, alpha, beta) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    const rad = radius;
+    const maxD = xShape[3] - 1;
+    this.outputShape = xShape;
+    let powOperator;
+    const basis = `float(${bias}) + float(${alpha}) * sum`;
+    if (beta === 0.5) {
+      powOperator = `inversesqrt(${basis})`;
+    } else if (beta === 1) {
+      powOperator = `1.0/(${basis})`;
+    } else {
+      powOperator = `exp(log(${basis}) * float(-${beta}));`;
+    }
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3554,17 +58225,38 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         int d = coords[3];
         float x = getX(b, r, c, d);
         float sum = 0.0;
-        for (int j = -${s}; j <= ${s}; j++) {
+        for (int j = -${rad}; j <= ${rad}; j++) {
           int idx = d + j;
-          if (idx >= 0 && idx <=  ${i}) {
+          if (idx >= 0 && idx <=  ${maxD}) {
             float z = getX(b, r, c, idx);
             sum += z * z;
           }
         }
-        float val = x * ${o};
+        float val = x * ${powOperator};
         setOutput(val);
       }
-    `}},Vte=class{constructor(e,t,n,a,r){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${a}) * sum`;r===.5?o=`inversesqrt(${l})`:r===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${r}));`,this.userCode=`
+    `;
+  }
+};
+var LRNPackedProgram = class {
+  constructor(xShape, radius, bias, alpha, beta) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const rad = radius;
+    const maxD = xShape[3] - 1;
+    this.outputShape = xShape;
+    let powOperator;
+    const basis = `float(${bias}) + float(${alpha}) * sum`;
+    if (beta === 0.5) {
+      powOperator = `inversesqrt(${basis})`;
+    } else if (beta === 1) {
+      powOperator = `1.0/(${basis})`;
+    } else {
+      powOperator = `exp(log(${basis}) * float(-${beta}));`;
+    }
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords.x;
@@ -3588,7 +58280,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
             getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0
         );
 
-        int firstChannel = d - ${s};
+        int firstChannel = d - ${rad};
         vec2 cache = vec2(0.);
         if(firstChannel >= 0){
           vec4 firstChannelFrag = getX(b, r, c, firstChannel);
@@ -3599,10 +58291,10 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
 
         ivec2 depth = ivec2(d, d + 1);
-        for (int j = - ${s}; j <= ${s}; j++) {
+        for (int j = - ${rad}; j <= ${rad}; j++) {
           ivec2 idx = depth + j;
           bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));
-          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));
+          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${maxD}));
 
           bool depthInRange = aboveLowerBound.x && belowUpperBound.x;
           bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;
@@ -3623,10 +58315,35 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
             sum += z * z;
           }
         }
-        vec4 result = xAtOutputCoords * ${o};
+        vec4 result = xAtOutputCoords * ${powOperator};
         setOutput(result);
       }
-    `}},Ute=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=a,u=H().getBool("WEBGL_PACK_NORMALIZATION")?new Vte(r.shape,s,i,o,l):new Bte(r.shape,s,i,o,l);return n.runWebGLProgram(u,[r],r.dtype)},Gte={kernelName:cc,backendName:"webgl",kernelFunc:Ute},Hte=class{constructor(e,t,n,a,r){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=n,this.alpha=a,this.beta=r,this.userCode=`
+    `;
+  }
+};
+var lrn = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  const program = env().getBool("WEBGL_PACK_NORMALIZATION") ? new LRNPackedProgram(x.shape, depthRadius, bias, alpha, beta) : new LRNProgram(x.shape, depthRadius, bias, alpha, beta);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+};
+var LRNConfig2 = {
+  kernelName: LRN,
+  backendName: "webgl",
+  kernelFunc: lrn
+};
+var LRNGradProgram = class {
+  constructor(inputShape, depthRadius, bias, alpha, beta) {
+    this.variableNames = ["inputImage", "outputImage", "dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    this.depth = inputShape[3];
+    this.depthRadius = depthRadius;
+    this.bias = bias;
+    this.alpha = alpha;
+    this.beta = beta;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3635,9 +58352,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
 
         float result = 0.0;
         for (int d = 0; d < ${this.depth}; ++d) {
-          int depthBegin = int(max(0.0, float(d - ${t})));
+          int depthBegin = int(max(0.0, float(d - ${depthRadius})));
           int depthEnd = int(min(float(${this.depth}),
-              float(d + ${t} + 1)));
+              float(d + ${depthRadius} + 1)));
 
           const int MIN_DEPTH_BEGIN = 0;
           const int MAX_DEPTH_END = ${this.depth};
@@ -3655,19 +58372,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
             }
           }
 
-          norm = float(${a}) * norm + float(${n});
+          norm = float(${alpha}) * norm + float(${bias});
 
           for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){
             if (k < depthBegin){
               continue;
             }
             else if (k >= depthBegin && k < depthEnd){
-              float dyi = -2.0 * float(${a})
-                * float(${r})
+              float dyi = -2.0 * float(${alpha})
+                * float(${beta})
                 * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)
                 / norm;
               if (k == d) {
-                dyi += pow(norm, -1.0 * ${r});
+                dyi += pow(norm, -1.0 * ${beta});
               }
               if (k == coords[3]) {
                 dyi *= getDy(b, r, c, d);
@@ -3681,17 +58398,155 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
       }
       setOutput(result);
       }
-    `}},jte=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:p}=a,d=new Hte(r.shape,o,l,u,p);return n.runWebGLProgram(d,[r,s,i],r.dtype)},qte={kernelName:hm,backendName:"webgl",kernelFunc:jte};function Kte(e,t,n,a){let r=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/r,i=de({inputs:{x:e},attrs:{shape:[s,r]},backend:a}),o=ko(i,e.dtype,"max",a),l=de({inputs:{x:o},attrs:{shape:n},backend:a});return a.disposeIntermediateTensorInfo(i),a.disposeIntermediateTensorInfo(o),l}function pE(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reductionIndices:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=p!=null,c=n.shouldExecuteOnCPU([r]),h=r;if(d){if(c){let b=n.texData.get(h.dataId).values,x=new Array(o);for(let T=0;T<x.length;T++)x[T]=r.shape[p[T]];let w=Z0(b,r.shape,r.dtype,p,x);h=n.makeTensorInfo(x,r.dtype);let I=n.texData.get(h.dataId);I.values=w}else h=Mf(r,p,n);u=N.getInnerMostAxes(u.length,o)}N.assertAxesAreInnerMostDims("max",u,o);let[m,f]=N.computeOutAndReduceShapes(h.shape,u),g=m;i&&(g=N.expandShapeToKeepDim(m,l));let y;if(c){let b=n.texData.get(h.dataId).values,x=i7(b,v.sizeFromShape(f),g,r.dtype);y=n.makeTensorInfo(g,r.dtype);let w=n.texData.get(y.dataId);w.values=x}else y=Kte(h,f,g,n);return d&&n.disposeIntermediateTensorInfo(h),y}var Xte={kernelName:Pi,backendName:"webgl",kernelFunc:pE},Yte=J0+`
+    `;
+  }
+};
+var lrnGrad = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, y, dy } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  const program = new LRNGradProgram(x.shape, depthRadius, bias, alpha, beta);
+  return backend2.runWebGLProgram(program, [x, y, dy], x.dtype);
+};
+var LRNGradConfig2 = {
+  kernelName: LRNGrad,
+  backendName: "webgl",
+  kernelFunc: lrnGrad
+};
+function maxImpl2(x, reduceShape, outShape, backend2) {
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const reduced = reduce(reshapedInput, x.dtype, "max", backend2);
+  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  return reshapedOutput;
+}
+function max4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reductionIndices, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  const maxInputIsTransposed = permutedAxes != null;
+  const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);
+  let maxInput = x;
+  if (maxInputIsTransposed) {
+    if (shouldExecuteOnCPU) {
+      const xTexData = backend2.texData.get(maxInput.dataId);
+      const values = xTexData.values;
+      const newShape = new Array(xRank);
+      for (let i = 0; i < newShape.length; i++) {
+        newShape[i] = x.shape[permutedAxes[i]];
+      }
+      const maxInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);
+      maxInput = backend2.makeTensorInfo(newShape, x.dtype);
+      const maxInputData = backend2.texData.get(maxInput.dataId);
+      maxInputData.values = maxInputValues;
+    } else {
+      maxInput = transposeImpl2(x, permutedAxes, backend2);
+    }
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, xRank);
+  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(maxInput.shape, axes);
+  let outShape = maxOutShape;
+  if (keepDims) {
+    outShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);
+  }
+  let out;
+  if (shouldExecuteOnCPU) {
+    const xTexData = backend2.texData.get(maxInput.dataId);
+    const values = xTexData.values;
+    const outValues = maxImplCPU(values, util_exports.sizeFromShape(reduceShape), outShape, x.dtype);
+    out = backend2.makeTensorInfo(outShape, x.dtype);
+    const outData = backend2.texData.get(out.dataId);
+    outData.values = outValues;
+  } else {
+    out = maxImpl2(maxInput, reduceShape, outShape, backend2);
+  }
+  if (maxInputIsTransposed) {
+    backend2.disposeIntermediateTensorInfo(maxInput);
+  }
+  return out;
+}
+var maxConfig2 = {
+  kernelName: Max,
+  backendName: "webgl",
+  kernelFunc: max4
+};
+var MAXIMUM = CHECK_NAN_SNIPPET2 + `
   return max(a, b);
-`,Zte=`
+`;
+var MAXIMUM_PACKED = `
   vec4 result = vec4(max(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+nd+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Jte=pn({opSnippet:Yte,packedOpSnippet:Zte,cpuKernelImpl:o7}),Qte={kernelName:Oi,backendName:"webgl",kernelFunc:Jte};function ene(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;Lu(r,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l);if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))return na({inputs:{x:r},backend:n});let d=new tc(p,"max",!1);return n.runWebGLProgram(d,[r],r.dtype)}var tne={kernelName:Li,backendName:"webgl",kernelFunc:ene};function nne(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dataFormat:l,dimRoundingMode:u}=a,p=[1,1,1],d=N.computePool3DInfo(r.shape,s,i,p,o,u,l),c=new e1(d,"max",!1);return n.runWebGLProgram(c,[r],r.dtype)}var ane={kernelName:dc,backendName:"webgl",kernelFunc:nne},rne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideHeight,n=e.strideWidth,a=e.dilationHeight,r=e.effectiveFilterHeight,s=e.effectiveFilterWidth,i=r-1-e.padInfo.top,o=s-1-e.padInfo.left,l=r*s-1;this.userCode=`
-      const ivec2 pads = ivec2(${i}, ${o});
+`;
+var maximum4 = binaryKernelFunc2({
+  opSnippet: MAXIMUM,
+  packedOpSnippet: MAXIMUM_PACKED,
+  cpuKernelImpl: maximumImplCPU
+});
+var maximumConfig2 = {
+  kernelName: Maximum,
+  backendName: "webgl",
+  kernelFunc: maximum4
+};
+function maxPool3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex2(x, "maxPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const maxPoolProgram = new Pool2DProgram(convInfo, "max", false);
+  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);
+}
+var maxPoolConfig2 = {
+  kernelName: MaxPool,
+  backendName: "webgl",
+  kernelFunc: maxPool3
+};
+function maxPool3d2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);
+  const maxPoolProgram = new Pool3DProgram(convInfo, "max", false);
+  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);
+}
+var maxPool3DConfig2 = {
+  kernelName: MaxPool3D,
+  backendName: "webgl",
+  kernelFunc: maxPool3d2
+};
+var MaxPool2DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "maxPos"];
+    this.outputShape = convInfo.inShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const lastIndex = effectiveFilterHeight * effectiveFilterWidth - 1;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3705,30 +58560,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${r};
-          wR += ${a}) {
-          float dyR = float(dyRCorner + wR) / ${t}.0;
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+          wR += ${dilationHeight}) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${s}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${n}.0;
+          for (int wC = 0; wC < ${effectiveFilterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
             float dyValue = getDy(b, idyR, idyC, d);
-            int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));
+            int maxPosValue = ${lastIndex} - int(getMaxPos(b, idyR, idyC, d));
 
             // Get the current value, check it against the value from the
             // position matrix.
-            int curPosValue = wR * ${s} + wC;
+            int curPosValue = wR * ${effectiveFilterWidth} + wC;
             float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
             dotProd += dyValue * mask;
@@ -3736,8 +58591,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},sne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideDepth,n=e.strideHeight,a=e.strideWidth,r=e.dilationDepth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterDepth,l=e.effectiveFilterHeight,u=e.effectiveFilterWidth,p=o-1-e.padInfo.front,d=l-1-e.padInfo.top,c=u-1-e.padInfo.left,h=o*l*u-1;this.userCode=`
-      const ivec3 pads = ivec3(${p}, ${d}, ${c});
+    `;
+  }
+};
+var MaxPool3DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "maxPos"];
+    this.outputShape = convInfo.inShape;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const lastIndex = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1;
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -3754,44 +58629,44 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${o};
-           wD += ${r}) {
-          float dyD = float(dyDCorner + wD) / ${t}.0;
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+           wD += ${dilationDepth}) {
+          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;
 
-          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
+          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {
             continue;
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${l};
-              wR += ${s}) {
-            float dyR = float(dyRCorner + wR) / ${n}.0;
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
                 fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            for (int wC = 0; wC < ${u};
-                wC += ${i}) {
-              float dyC = float(dyCCorner + wC) / ${a}.0;
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
               int idyC = int(dyC);
 
               float dyValue = getDy(batch, idyD, idyR, idyC, ch);
-              int maxPosValue = ${h} -
+              int maxPosValue = ${lastIndex} -
                   int(getMaxPos(batch, idyD, idyR, idyC, ch));
 
               // Get the current value, check it against the value from the
               // position matrix.
               int curPosValue =
-                  wD * ${l} * ${u} +
-                  wR * ${u} + wC;
+                  wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +
+                  wR * ${effectiveFilterWidth} + wC;
               float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
               dotProd += dyValue * mask;
@@ -3800,107 +58675,348 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function ine(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:p}=a,d=[1,1,1],c=N.computePool3DInfo(i.shape,o,l,d,u,p),h=new e1(c,"max",!0),m=n.runWebGLProgram(h,[i],i.dtype),f=new sne(c),g=n.runWebGLProgram(f,[r,m],i.dtype);return n.disposeIntermediateTensorInfo(m),g}var one={kernelName:fm,backendName:"webgl",kernelFunc:ine};function lne(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s,output:i}=t,o=s;Lu([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:p,dimRoundingMode:d}=a,c=N.computePool2DInfo(o.shape,l,u,1,p,d),h=!0,m=new tc(c,"max",h),f=n.runWebGLProgram(m,[o],o.dtype),g=new rne(c),y=n.runWebGLProgram(g,[r,f],o.dtype);return n.disposeIntermediateTensorInfo(f),y}var une={kernelName:mm,backendName:"webgl",kernelFunc:lne};function pne(e,t,n,a){let r=new tc(n,"max",!1),s=a.runWebGLProgram(r,[e],"float32");r=new tc(n,"max",!0,!0,t);let i=a.runWebGLProgram(r,[e],"float32");return[s,i]}var cne={kernelName:gm,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{filterSize:r,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;v.assert(a.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${a.shape.length}.`);let u=[1,1];v.assert(N.eitherStridesOrDilationsAreOne(s,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${u}'`);let p=N.computePool2DInfo(a.shape,r,s,u,i),[d,c]=pne(a,o,p,l);return[d,c]}};function dne(e,t,n,a){let r=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/r,i=de({inputs:{x:e},attrs:{shape:[s,r]},backend:a}),o=ko(i,"float32","mean",a),l=de({inputs:{x:o},attrs:{shape:n},backend:a});return a.disposeIntermediateTensorInfo(i),a.disposeIntermediateTensorInfo(o),l}var hne={kernelName:zi,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{keepDims:r,axis:s}=t,i=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,p=N.getAxesPermutation(u,o),d=p!=null,c=i.shouldExecuteOnCPU([a]),h=[],m=a;if(d){if(c){let x=i.texData.get(m.dataId).values,w=new Array(o);for(let C=0;C<w.length;C++)w[C]=a.shape[p[C]];let I=Z0(x,a.shape,a.dtype,p,w);m=i.makeTensorInfo(w,a.dtype);let T=i.texData.get(m.dataId);T.values=I}else m=Mf(a,p,i);h.push(m),u=N.getInnerMostAxes(u.length,o)}N.assertAxesAreInnerMostDims("sum",u,o);let[f,g]=N.computeOutAndReduceShapes(m.shape,u),y=f;r&&(y=N.expandShapeToKeepDim(f,l));let b=dne(m,g,y,i);for(let x of h)i.disposeIntermediateTensorInfo(x);return b}};function mne(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=In({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,r.shape.length)),N.assertAxesAreInnerMostDims("min",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=ko(f,f.dtype,"min",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var fne={kernelName:Wi,backendName:"webgl",kernelFunc:mne},gne=J0+`
+    `;
+  }
+};
+function maxPool3DGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  const maxPool3dPositionsProgram = new Pool3DProgram(convInfo, "max", true);
+  const maxPool3dPositions2 = backend2.runWebGLProgram(maxPool3dPositionsProgram, [x], x.dtype);
+  const maxPoolBackpropProgram = new MaxPool3DBackpropProgram(convInfo);
+  const result = backend2.runWebGLProgram(maxPoolBackpropProgram, [dy, maxPool3dPositions2], x.dtype);
+  backend2.disposeIntermediateTensorInfo(maxPool3dPositions2);
+  return result;
+}
+var maxPool3DGradConfig3 = {
+  kernelName: MaxPool3DGrad,
+  backendName: "webgl",
+  kernelFunc: maxPool3DGrad2
+};
+function maxPoolGrad3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2, output } = inputs;
+  const x = input2;
+  assertNotComplex2([input2, output], "maxPoolGrad");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const getPositions = true;
+  const maxPoolPositionsProgram = new Pool2DProgram(convInfo, "max", getPositions);
+  const maxPoolPositions2 = backend2.runWebGLProgram(maxPoolPositionsProgram, [x], x.dtype);
+  const maxPoolBackPropProgram = new MaxPool2DBackpropProgram(convInfo);
+  const result = backend2.runWebGLProgram(maxPoolBackPropProgram, [dy, maxPoolPositions2], x.dtype);
+  backend2.disposeIntermediateTensorInfo(maxPoolPositions2);
+  return result;
+}
+var maxPoolGradConfig3 = {
+  kernelName: MaxPoolGrad,
+  backendName: "webgl",
+  kernelFunc: maxPoolGrad3
+};
+function maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, backend2) {
+  let program = new Pool2DProgram(convInfo, "max", false);
+  const poolOutput = backend2.runWebGLProgram(program, [x], "float32");
+  program = new Pool2DProgram(convInfo, "max", true, true, includeBatchInIndex);
+  const indexOutput = backend2.runWebGLProgram(program, [x], "float32");
+  return [poolOutput, indexOutput];
+}
+var maxPoolWithArgmaxConfig2 = {
+  kernelName: MaxPoolWithArgmax,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;
+    const webglBackend = backend2;
+    util_exports.assert(x.shape.length === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x.shape.length}.`);
+    const dilations = [1, 1];
+    util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3);
+    const [result, indexes] = maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, webglBackend);
+    return [result, indexes];
+  }
+};
+function meanImpl(x, reduceShape, outShape, backend2) {
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const reduced = reduce(reshapedInput, "float32", "mean", backend2);
+  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  return reshapedOutput;
+}
+var meanConfig2 = {
+  kernelName: Mean,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { keepDims, axis } = attrs;
+    const webglBackend = backend2;
+    const xRank = x.shape.length;
+    const origAxes = util_exports.parseAxisParam(axis, x.shape);
+    let axes = origAxes;
+    const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+    const meanInputIsTransposed = permutedAxes != null;
+    const shouldExecuteOnCPU = webglBackend.shouldExecuteOnCPU([x]);
+    const intermediates = [];
+    let meanInput = x;
+    if (meanInputIsTransposed) {
+      if (shouldExecuteOnCPU) {
+        const xTexData = webglBackend.texData.get(meanInput.dataId);
+        const values = xTexData.values;
+        const newShape = new Array(xRank);
+        for (let i = 0; i < newShape.length; i++) {
+          newShape[i] = x.shape[permutedAxes[i]];
+        }
+        const meanInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);
+        meanInput = webglBackend.makeTensorInfo(newShape, x.dtype);
+        const meanInputData = webglBackend.texData.get(meanInput.dataId);
+        meanInputData.values = meanInputValues;
+      } else {
+        meanInput = transposeImpl2(x, permutedAxes, webglBackend);
+      }
+      intermediates.push(meanInput);
+      axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    }
+    backend_util_exports.assertAxesAreInnerMostDims("sum", axes, xRank);
+    const [meanOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(meanInput.shape, axes);
+    let outShape = meanOutShape;
+    if (keepDims) {
+      outShape = backend_util_exports.expandShapeToKeepDim(meanOutShape, origAxes);
+    }
+    const out = meanImpl(meanInput, reduceShape, outShape, webglBackend);
+    for (const i of intermediates) {
+      webglBackend.disposeIntermediateTensorInfo(i);
+    }
+    return out;
+  }
+};
+function min4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "min", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var minConfig2 = {
+  kernelName: Min,
+  backendName: "webgl",
+  kernelFunc: min4
+};
+var MINIMUM = CHECK_NAN_SNIPPET2 + `
   return min(a, b);
-`,yne=`
+`;
+var MINIMUM_PACKED = `
   vec4 result = vec4(min(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+nd+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,bne=pn({opSnippet:gne,packedOpSnippet:yne,cpuKernelImpl:l7}),xne={kernelName:Bi,backendName:"webgl",kernelFunc:bne},vne=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=t.map((u,p)=>u[0]+e[p]+u[1]);let a=e.length,r=gt(a),s=t.map(u=>u[0]).join(","),i=t.map((u,p)=>u[0]+e[p]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,a),l=n==="reflect"?0:1;if(a===1){this.userCode=`
-        int start = ${s};
-        int end = ${i};
+`;
+var minimum4 = binaryKernelFunc2({
+  opSnippet: MINIMUM,
+  packedOpSnippet: MINIMUM_PACKED,
+  cpuKernelImpl: minimumImplCPU
+});
+var minimumConfig2 = {
+  kernelName: Minimum,
+  backendName: "webgl",
+  kernelFunc: minimum4
+};
+var MirrorPadProgram = class {
+  constructor(xShape, paddings, mode) {
+    this.variableNames = ["x"];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const unpackedCoords = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, rank);
+    const offset = mode === "reflect" ? 0 : 1;
+    if (rank === 1) {
+      this.userCode = `
+        int start = ${start};
+        int end = ${end};
 
         void main() {
           int outC = getOutputCoords();
           if (outC < start) {
-            outC = start * 2 - outC - ${l};
+            outC = start * 2 - outC - ${offset};
           } else if(outC >= end) {
-            outC = (end - 1) * 2 - outC + ${l};
+            outC = (end - 1) * 2 - outC + ${offset};
           }
           setOutput(getX(outC - start));
         }
-      `;return}this.userCode=`
-      ${r} start = ${r}(${s});
-      ${r} end = ${r}(${i});
+      `;
+      return;
+    }
+    this.userCode = `
+      ${dtype} start = ${dtype}(${start});
+      ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${r} outC = getOutputCoords();
-        for (int i = 0; i < ${a}; i++) {
+        ${dtype} outC = getOutputCoords();
+        for (int i = 0; i < ${rank}; i++) {
           if (outC[i] < start[i]) {
-            outC[i] = start[i] * 2 - outC[i] - ${l};
+            outC[i] = start[i] * 2 - outC[i] - ${offset};
           } else if(outC[i] >= end[i]) {
-            outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};
+            outC[i] = (end[i] - 1) * 2 - outC[i] + ${offset};
           }
         }
-        ${r} coords = outC - start;
-        setOutput(getX(${o}));
+        ${dtype} coords = outC - start;
+        setOutput(getX(${unpackedCoords}));
       }
-    `}},wne=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((h,m)=>h[0]+e[m]+h[1]);let a=e.length,r=gt(a),s=t.map(h=>h[0]).join(","),i=t.map((h,m)=>h[0]+e[m]).join(","),o=wn("rc",a),l=wn("source",a),u=`${o[a-1]} < ${this.outputShape[a-1]}`,p=a===1?"source":`vec2(${l.slice(-2).join()})`,d=n==="reflect"?0:1,c="";if(a===1){let h=`
-        ${r} source = rc;
+    `;
+  }
+};
+var MirrorPadPackedProgram = class {
+  constructor(xShape, paddings, mode) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const coords2 = getChannels("rc", rank);
+    const source = getChannels("source", rank);
+    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;
+    const innerDims = rank === 1 ? "source" : `vec2(${source.slice(-2).join()})`;
+    const offset = mode === "reflect" ? 0 : 1;
+    let mainLoop = "";
+    if (rank === 1) {
+      const padSetup = `
+        ${dtype} source = rc;
         if (source < start) {
-          source = start * 2 - source - ${d};
+          source = start * 2 - source - ${offset};
         } else if (source >= end) {
-          source = (end - 1) * 2 - source + ${d};
+          source = (end - 1) * 2 - source + ${offset};
         }
         source -= start;
-      `;c=`
-        ${r} rc = outputLoc;
-        ${h}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${o[a-1]} += 1;
-        if(${u}) {
-          ${h}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+      `;
+      mainLoop = `
+        ${dtype} rc = outputLoc;
+        ${padSetup}
+        result[0] = getChannel(getX(${source.join()}), ${innerDims});
+        ${coords2[rank - 1]} += 1;
+        if(${cLimit}) {
+          ${padSetup}
+          result[1] = getChannel(getX(${source.join()}), ${innerDims});
         }
-      `}else{let h=`
-        ${r} source = rc;
-        ${r} lt = ${r}(lessThan(source, start));
-        ${r} gte = ${r}(greaterThanEqual(source, end));
-        ${r} orig = 1 - (lt + gte);
+      `;
+    } else {
+      const padSetup = `
+        ${dtype} source = rc;
+        ${dtype} lt = ${dtype}(lessThan(source, start));
+        ${dtype} gte = ${dtype}(greaterThanEqual(source, end));
+        ${dtype} orig = 1 - (lt + gte);
         source = orig * source +
-                lt * (start * 2 - source - ${d}) +
-                gte * ((end - 1) * 2 - source + ${d});
+                lt * (start * 2 - source - ${offset}) +
+                gte * ((end - 1) * 2 - source + ${offset});
         source -= start;
-      `;c=`
-        ${r} rc = outputLoc;
-        ${h}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${o[a-1]} += 1;
-        if(${u}) {
-          ${h}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+      `;
+      mainLoop = `
+        ${dtype} rc = outputLoc;
+        ${padSetup}
+        result[0] = getChannel(getX(${source.join()}), ${innerDims});
+        ${coords2[rank - 1]} += 1;
+        if(${cLimit}) {
+          ${padSetup}
+          result[1] = getChannel(getX(${source.join()}), ${innerDims});
         }
         rc = outputLoc;
-        ${o[a-2]} += 1;
-        if(${o[a-2]} < ${this.outputShape[a-2]}) {
-          ${h}
-          result[2] = getChannel(getX(${l.join()}), ${p});
-          ${o[a-1]} += 1;
-          if(${u}) {
-            ${h}
-            result[3] = getChannel(getX(${l.join()}), ${p});
+        ${coords2[rank - 2]} += 1;
+        if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {
+          ${padSetup}
+          result[2] = getChannel(getX(${source.join()}), ${innerDims});
+          ${coords2[rank - 1]} += 1;
+          if(${cLimit}) {
+            ${padSetup}
+            result[3] = getChannel(getX(${source.join()}), ${innerDims});
           }
         }
-      `}this.userCode=`
-      const ${r} start = ${r}(${s});
-      const ${r} end = ${r}(${i});
+      `;
+    }
+    this.userCode = `
+      const ${dtype} start = ${dtype}(${start});
+      const ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${r} outputLoc = getOutputCoords();
+        ${dtype} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
-        ${c}
+        ${mainLoop}
         setOutput(result);
       }
-    `}},kne=({inputs:e,backend:t,attrs:n})=>{let{x:a}=e,{paddings:r,mode:s}=n,i=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new wne(a.shape,r,s):new vne(a.shape,r,s);return t.runWebGLProgram(i,[a],a.dtype)},Ine={kernelName:Vi,backendName:"webgl",kernelFunc:kne},Sne=`if (b == 0.0) return NAN;
-  return mod(a, b);`,Tne=`
+    `;
+  }
+};
+var mirrorPadKernelFunc = ({ inputs, backend: backend2, attrs }) => {
+  const { x } = inputs;
+  const { paddings, mode } = attrs;
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new MirrorPadPackedProgram(x.shape, paddings, mode) : new MirrorPadProgram(x.shape, paddings, mode);
+  const output = backend2.runWebGLProgram(program, [x], x.dtype);
+  return output;
+};
+var mirrorPadConfig2 = {
+  kernelName: MirrorPad,
+  backendName: "webgl",
+  kernelFunc: mirrorPadKernelFunc
+};
+var MOD = `if (b == 0.0) return NAN;
+  return mod(a, b);`;
+var MOD_PACKED = `
   vec4 result = mod(a, b);
   bvec4 isNaN = equal(b, vec4(0.0));
-  `+nd+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Nne=pn({opSnippet:Sne,packedOpSnippet:Tne}),Cne={kernelName:Ql,backendName:"webgl",kernelFunc:Nne},_ne=class{constructor(e,t,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,n],this.userCode=`
+`;
+var mod3 = binaryKernelFunc2({
+  opSnippet: MOD,
+  packedOpSnippet: MOD_PACKED
+});
+var modConfig2 = {
+  kernelName: Mod,
+  backendName: "webgl",
+  kernelFunc: mod3
+};
+var MultinomialProgram = class {
+  constructor(batchSize, numOutcomes, numSamples) {
+    this.variableNames = ["probs"];
+    this.customUniforms = [{ name: "seed", type: "float" }];
+    this.outputShape = [batchSize, numSamples];
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
@@ -3908,7 +59024,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         float r = random(seed);
         float cdf = 0.0;
 
-        for (int i = 0; i < ${t-1}; i++) {
+        for (int i = 0; i < ${numOutcomes - 1}; i++) {
           cdf += getProbs(batch, i);
 
           if (r < cdf) {
@@ -3918,13 +59034,17 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,l9=Ye({opSnippet:o9}),u9={kernelNam
         }
 
         // If no other event happened, last event happened.
-        setOutput(float(${t-1}));
+        setOutput(float(${numOutcomes - 1}));
       }
-    `}},Ene=`
+    `;
+  }
+};
+var DIV = `
 if (a == b) {
   return 1.0;
 };
-return a / b;`,Ane=`
+return a / b;`;
+var DIV_PACKED = `
   // vec4 one = vec4(equal(a, b));
   // return one + (vec4(1.0) - one) * a / b;
   vec4 result = a / b;
@@ -3942,9 +59062,79 @@ return a / b;`,Ane=`
   }
 
   return result;
-`,cE=pn({opSnippet:Ene,packedOpSnippet:Ane,checkOutOfBounds:!0}),$ne={kernelName:Ni,backendName:"webgl",kernelFunc:cE},cI="return a - b;",dE=pn({opSnippet:cI,packedOpSnippet:cI,supportsComplex:!0,cpuKernelImpl:_7}),Fne={kernelName:lo,backendName:"webgl",kernelFunc:dE};function hE(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{dim:s}=a,i=v.parseAxisParam([s],r.shape),o=pE({inputs:{x:r},backend:n,attrs:{reductionIndices:i,keepDims:!1}}),l=N.expandShapeToKeepDim(o.shape,i),u=de({inputs:{x:o},backend:n,attrs:{shape:l}}),p=dE({inputs:{a:r,b:u},backend:n}),d=oE({inputs:{x:p},backend:n}),c=Pf({inputs:{x:d},backend:n,attrs:{axis:i,keepDims:!1}}),h=de({inputs:{x:c},backend:n,attrs:{shape:l}}),m=cE({inputs:{a:d,b:h},backend:n});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),m}var Dne={kernelName:io,backendName:"webgl",kernelFunc:hE};function Rne(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{numSamples:s,seed:i,normalized:o}=a,l=o?r:hE({inputs:{logits:r},backend:n,attrs:{dim:r.shape.length-1}}),u=l.shape[0],p=l.shape[1],d=new _ne(u,p,s),c=[[i]],h=n.runWebGLProgram(d,[l],"int32",c);return o||n.disposeIntermediateTensorInfo(l),h}var Mne={kernelName:ym,backendName:"webgl",kernelFunc:Rne},Pne=Da+`
+`;
+var realDiv = binaryKernelFunc2({ opSnippet: DIV, packedOpSnippet: DIV_PACKED, checkOutOfBounds: true });
+var realDivConfig2 = {
+  kernelName: RealDiv,
+  backendName: "webgl",
+  kernelFunc: realDiv
+};
+var SUB = "return a - b;";
+var sub3 = binaryKernelFunc2({
+  opSnippet: SUB,
+  packedOpSnippet: SUB,
+  supportsComplex: true,
+  cpuKernelImpl: subImplCPU
+});
+var subConfig2 = {
+  kernelName: Sub,
+  backendName: "webgl",
+  kernelFunc: sub3
+};
+function softmax4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { dim } = attrs;
+  const axes = util_exports.parseAxisParam([dim], logits.shape);
+  const maxLogit = max4({
+    inputs: { x: logits },
+    backend: backend2,
+    attrs: { reductionIndices: axes, keepDims: false }
+  });
+  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);
+  const maxLogitsReshaped = reshape4({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });
+  const a = sub3({ inputs: { a: logits, b: maxLogitsReshaped }, backend: backend2 });
+  const b = exp3({ inputs: { x: a }, backend: backend2 });
+  const sumExp = sum4({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });
+  const sumExpReshaped = reshape4({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });
+  const res = realDiv({ inputs: { a: b, b: sumExpReshaped }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(maxLogit);
+  backend2.disposeIntermediateTensorInfo(maxLogitsReshaped);
+  backend2.disposeIntermediateTensorInfo(a);
+  backend2.disposeIntermediateTensorInfo(b);
+  backend2.disposeIntermediateTensorInfo(sumExp);
+  backend2.disposeIntermediateTensorInfo(sumExpReshaped);
+  return res;
+}
+var softmaxConfig2 = {
+  kernelName: Softmax,
+  backendName: "webgl",
+  kernelFunc: softmax4
+};
+function multinomial3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { numSamples, seed, normalized } = attrs;
+  const probs = normalized ? logits : softmax4({ inputs: { logits }, backend: backend2, attrs: { dim: logits.shape.length - 1 } });
+  const batchSize = probs.shape[0];
+  const numOutcomes = probs.shape[1];
+  const program = new MultinomialProgram(batchSize, numOutcomes, numSamples);
+  const customValues = [[seed]];
+  const res = backend2.runWebGLProgram(program, [probs], "int32", customValues);
+  if (!normalized) {
+    backend2.disposeIntermediateTensorInfo(probs);
+  }
+  return res;
+}
+var multinomialConfig2 = {
+  kernelName: Multinomial,
+  backendName: "webgl",
+  kernelFunc: multinomial3
+};
+var NEG = CHECK_NAN_SNIPPET + `
   return -x;
-`,One=`
+`;
+var NEG_PACKED = `
   vec4 result = -x;
   bvec4 isNaN = isnan(x);
 
@@ -3954,16 +59144,215 @@ return a / b;`,Ane=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`;function Lne(e){let{inputs:t,backend:n}=e,{x:a}=t;if(n.shouldExecuteOnCPU([a])){let s=n.texData.get(a.dataId),[i,o]=p7(s.values,a.shape,a.dtype);return n.makeTensorInfo(o,a.dtype,i)}let r;return H().getBool("WEBGL_PACK_UNARY_OPERATIONS")?r=new js(a.shape,One):r=new Sr(a.shape,Pne),n.runWebGLProgram(r,[a],a.dtype)}var zne={kernelName:eu,backendName:"webgl",kernelFunc:Lne},Wne=cr.nonMaxSuppressionV3Impl;function Bne(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=a,u=n.readSync(r.dataId),p=n.readSync(s.dataId),{selectedIndices:d}=Wne(u,p,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}var Vne={kernelName:nu,backendName:"webgl",kernelFunc:Bne},Une=cr.nonMaxSuppressionV4Impl;function Gne(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=a,p=n.readSync(r.dataId),d=n.readSync(s.dataId),{selectedIndices:c,validOutputs:h}=Une(p,d,i,o,l,u);return[n.makeTensorInfo([c.length],"int32",new Int32Array(c)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var Hne={kernelName:au,backendName:"webgl",kernelFunc:Gne},jne=cr.nonMaxSuppressionV5Impl;function qne(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=a,p=n.readSync(r.dataId),d=n.readSync(s.dataId),c=i,h=o,m=l,f=u,{selectedIndices:g,selectedScores:y}=jne(p,d,c,h,m,f);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var Kne={kernelName:ru,backendName:"webgl",kernelFunc:qne},Xne=class{constructor(e,t,n,a){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`
+`;
+function neg3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (backend2.shouldExecuteOnCPU([x])) {
+    const xData = backend2.texData.get(x.dataId);
+    const [outValues, newShape] = negImplCPU(xData.values, x.shape, x.dtype);
+    return backend2.makeTensorInfo(newShape, x.dtype, outValues);
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+    program = new UnaryOpPackedProgram(x.shape, NEG_PACKED);
+  } else {
+    program = new UnaryOpProgram(x.shape, NEG);
+  }
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var negConfig2 = {
+  kernelName: Neg,
+  backendName: "webgl",
+  kernelFunc: neg3
+};
+var nonMaxSuppressionV3Impl3 = kernel_impls_exports.nonMaxSuppressionV3Impl;
+function nonMaxSuppressionV32(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const { selectedIndices } = nonMaxSuppressionV3Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  return backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices));
+}
+var nonMaxSuppressionV3Config2 = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV32
+};
+var nonMaxSuppressionV4Impl3 = kernel_impls_exports.nonMaxSuppressionV4Impl;
+function nonMaxSuppressionV42(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([], "int32", new Int32Array([validOutputs]))
+  ];
+}
+var nonMaxSuppressionV4Config2 = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV42
+};
+var nonMaxSuppressionV5Impl3 = kernel_impls_exports.nonMaxSuppressionV5Impl;
+function nonMaxSuppressionV52(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const maxOutputSizeVal = maxOutputSize;
+  const iouThresholdVal = iouThreshold;
+  const scoreThresholdVal = scoreThreshold;
+  const softNmsSigmaVal = softNmsSigma;
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl3(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([selectedScores.length], "float32", new Float32Array(selectedScores))
+  ];
+}
+var nonMaxSuppressionV5Config2 = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV52
+};
+var OneHotProgram = class {
+  constructor(numIndices, depth, onValue, offValue) {
+    this.variableNames = ["indices"];
+    this.outputShape = [numIndices, depth];
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int index = round(getIndices(coords.x));
-        setOutput(mix(float(${a}), float(${n}),
+        setOutput(mix(float(${offValue}), float(${onValue}),
                       float(index == coords.y)));
       }
-    `}},Yne=e=>{let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a,u=v.sizeFromShape(r.shape),p=new Xne(u,i,o,l),d=de({inputs:{x:r},backend:n,attrs:{shape:[u]}}),c=n.runWebGLProgram(p,[d],s);n.disposeIntermediateTensorInfo(d);let h=[...r.shape,i],m=de({inputs:{x:c},backend:n,attrs:{shape:h}});return n.disposeIntermediateTensorInfo(c),m},Zne={kernelName:Gi,backendName:"webgl",kernelFunc:Yne};function Uh(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="complex64"){let r=rd({inputs:{input:a},backend:n}),s=Uh({inputs:{x:r},backend:n}),i=Of({inputs:{input:a},backend:n}),o=Uh({inputs:{x:i},backend:n}),l=Is({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return sd({attrs:{shape:a.shape,dtype:a.dtype,value:a.dtype==="string"?"":0},backend:n})}var Jne={kernelName:Iu,backendName:"webgl",kernelFunc:Uh};function mE(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(a.dtype==="complex64"){let r=rd({inputs:{input:a},backend:n}),s=mE({inputs:{x:r},backend:n}),i=Of({inputs:{input:a},backend:n}),o=Uh({inputs:{x:i},backend:n}),l=Is({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return sd({attrs:{shape:a.shape,dtype:a.dtype,value:1},backend:n})}var Qne={kernelName:su,backendName:"webgl",kernelFunc:mE};function eae(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return dx({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=dx({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=Q_({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeIntermediateTensorInfo(p)),u}var tae={kernelName:iu,backendName:"webgl",kernelFunc:eae},nae=class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((l,u)=>l[0]+e[u]+l[1]);let a=e.length,r=gt(a),s=t.map(l=>l[0]).join(","),i=t.map((l,u)=>l[0]+e[u]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,a);if(a===1){this.userCode=`
-        int start = ${s};
-        int end = ${i};
+    `;
+  }
+};
+var oneHot3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const program = new OneHotProgram(indicesSize, depth, onValue, offValue);
+  const reshaped = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [indicesSize] } });
+  const result = backend2.runWebGLProgram(program, [reshaped], dtype);
+  backend2.disposeIntermediateTensorInfo(reshaped);
+  const outShape = [...indices.shape, depth];
+  const out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: outShape } });
+  backend2.disposeIntermediateTensorInfo(result);
+  return out;
+};
+var oneHotConfig2 = {
+  kernelName: OneHot,
+  backendName: "webgl",
+  kernelFunc: oneHot3
+};
+function zerosLike3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const r = zerosLike3({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill3({
+      attrs: {
+        shape: x.shape,
+        dtype: x.dtype,
+        value: x.dtype === "string" ? "" : 0
+      },
+      backend: backend2
+    });
+  }
+}
+var zerosLikeConfig2 = {
+  kernelName: ZerosLike,
+  backendName: "webgl",
+  kernelFunc: zerosLike3
+};
+function onesLike3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("onesLike is not supported under string dtype");
+  } else if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const r = onesLike3({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill3({ attrs: { shape: x.shape, dtype: x.dtype, value: 1 }, backend: backend2 });
+  }
+}
+var onesLikeConfig2 = {
+  kernelName: OnesLike,
+  backendName: "webgl",
+  kernelFunc: onesLike3
+};
+function pack2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims4({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims4({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat3({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var packConfig2 = {
+  kernelName: Pack,
+  backendName: "webgl",
+  kernelFunc: pack2
+};
+var PadProgram = class {
+  constructor(xShape, paddings, constantValue) {
+    this.variableNames = ["x"];
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const type = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const unpackedCoords = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, rank);
+    if (rank === 1) {
+      this.userCode = `
+        int start = ${start};
+        int end = ${end};
 
         void main() {
           int outC = getOutputCoords();
@@ -3973,44 +59362,100 @@ return a / b;`,Ane=`
             setOutput(getX(outC - start));
           }
         }
-      `;return}this.userCode=`
-      ${r} start = ${r}(${s});
-      ${r} end = ${r}(${i});
+      `;
+      return;
+    }
+    this.userCode = `
+      ${type} start = ${type}(${start});
+      ${type} end = ${type}(${end});
 
       void main() {
-        ${r} outC = getOutputCoords();
+        ${type} outC = getOutputCoords();
         if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {
           setOutput(value);
         } else {
-          ${r} coords = outC - start;
-          setOutput(getX(${o}));
+          ${type} coords = outC - start;
+          setOutput(getX(${unpackedCoords}));
         }
       }
-    `}},aae=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((m,f)=>m[0]+e[f]+m[1]);let a=e.length,r=gt(a),s=t.map(m=>m[0]).join(","),i=t.map((m,f)=>m[0]+e[f]).join(","),o=wn("rc",a),l=wn("source",a),u=`${o[a-1]} < ${this.outputShape[a-1]}`,p=a===1?"source":`vec2(${l.slice(-2).join()})`,d=[`${r} rc = outputLoc;`,`${o[a-1]} += 1;
-       if(${u}) {
-      `,a===1?"":`}
+    `;
+  }
+};
+var PadPackedProgram = class {
+  constructor(xShape, paddings, constantValue) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const coords2 = getChannels("rc", rank);
+    const source = getChannels("source", rank);
+    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;
+    const innerDims = rank === 1 ? "source" : `vec2(${source.slice(-2).join()})`;
+    const componentSetup = [
+      `${dtype} rc = outputLoc;`,
+      `${coords2[rank - 1]} += 1;
+       if(${cLimit}) {
+      `,
+      rank === 1 ? "" : `}
        rc = outputLoc;
-       ${o[a-2]} += 1;
-       if(${o[a-2]} < ${this.outputShape[a-2]}) {`,a===1?"":`  ${o[a-1]} += 1;
-         if(${u}) {`],c=a===1?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",h="";for(let m=0,f=a===1?2:4;m<f;m++)h+=`
-        ${d[m]}
-        if (${c}) {
-          result[${m}] = float(value);
+       ${coords2[rank - 2]} += 1;
+       if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {`,
+      rank === 1 ? "" : `  ${coords2[rank - 1]} += 1;
+         if(${cLimit}) {`
+    ];
+    const paddingArea = rank === 1 ? "rc < start || rc >= end" : "any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))";
+    let mainLoop = "";
+    for (let i = 0, j = rank === 1 ? 2 : 4; i < j; i++) {
+      mainLoop += `
+        ${componentSetup[i]}
+        if (${paddingArea}) {
+          result[${i}] = float(value);
         } else {
-          ${r} source = rc - start;
-          result[${m}] = getChannel(getX(${l.join()}), ${p});
+          ${dtype} source = rc - start;
+          result[${i}] = getChannel(getX(${source.join()}), ${innerDims});
         }
-      `;h+=a===1?"} ":"}}",this.userCode=`
-      const ${r} start = ${r}(${s});
-      const ${r} end = ${r}(${i});
+      `;
+    }
+    mainLoop += rank === 1 ? `} ` : `}}`;
+    this.userCode = `
+      const ${dtype} start = ${dtype}(${start});
+      const ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${r} outputLoc = getOutputCoords();
+        ${dtype} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
-        ${h}
+        ${mainLoop}
         setOutput(result);
       }
-    `}},fE=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,constantValue:i}=a;if(v.sizeFromShape(r.shape)===0){let u=s.map((p,d)=>p[0]+r.shape[d]+p[1]);return sd({backend:n,attrs:{shape:u,value:i,dtype:r.dtype}})}let o=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new aae(r.shape,s,i):new nae(r.shape,s,i),l=[[i]];return n.runWebGLProgram(o,[r],r.dtype,l)},rae={kernelName:Hi,backendName:"webgl",kernelFunc:fE},sae=`
+    `;
+  }
+};
+var padV22 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, constantValue } = attrs;
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    const outputShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+    return fill3({
+      backend: backend2,
+      attrs: { shape: outputShape, value: constantValue, dtype: x.dtype }
+    });
+  }
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new PadPackedProgram(x.shape, paddings, constantValue) : new PadProgram(x.shape, paddings, constantValue);
+  const customValues = [[constantValue]];
+  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+};
+var padV2Config2 = {
+  kernelName: PadV2,
+  backendName: "webgl",
+  kernelFunc: padV22
+};
+var POW = `
   if(a < 0.0 && floor(b) < b){
     return NAN;
   }
@@ -4019,7 +59464,8 @@ return a / b;`,Ane=`
   }
   return (round(mod(b, 2.0)) != 1) ?
       pow(abs(a), b) : sign(a) * pow(abs(a), b);
-`,iae=`
+`;
+var POW_PACKED = `
   // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.
   vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));
   vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);
@@ -4035,11 +59481,132 @@ return a / b;`,Ane=`
   bvec4 isNaN1 = lessThan(a, vec4(0.0));
   bvec4 isNaN2 = lessThan(floor(b), b);
   bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);
-  `+nd+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,oae=pn({opSnippet:sae,packedOpSnippet:iae}),lae={kernelName:ji,backendName:"webgl",kernelFunc:oae};function uae(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=[],u=v.parseAxisParam(s,r.shape),p=u,d=N.getAxesPermutation(p,o),c=r;d!=null&&(c=In({inputs:{x:r},backend:n,attrs:{perm:d}}),p=N.getInnerMostAxes(p.length,o),l.push(c)),N.assertAxesAreInnerMostDims("prod",p,o);let h;if(n.shouldExecuteOnCPU([c])){let m=n.texData.get(c.dataId).values,{outVals:f,outShape:g,outDtype:y}=d7(c.shape,c.dtype,m,p);h=n.makeTensorInfo(g,y,f)}else{let[m,f]=N.computeOutAndReduceShapes(c.shape,p),g=v.sizeFromShape(f),y=de({inputs:{x:c},backend:n,attrs:{shape:[-1,g]}}),b=_m(r.dtype),x=ko(y,b,"prod",n);h=de({inputs:{x},backend:n,attrs:{shape:m}}),l.push(y),l.push(x)}if(i){l.push(h);let m=N.expandShapeToKeepDim(h.shape,u);h=de({inputs:{x:h},backend:n,attrs:{shape:m}})}return l.forEach(m=>n.disposeIntermediateTensorInfo(m)),h}var pae={kernelName:Ki,backendName:"webgl",kernelFunc:uae};function cae(e){let{inputs:t,backend:n,attrs:a}=e,{paramsNestedSplits:r,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=a,l=r.map(y=>n.readSync(y.dataId)),u=r.map(y=>y.shape),p=n.readSync(s.dataId),d=n.readSync(i.dataId),[c,h,m]=h7(l,u,p,s.shape,s.dtype,d,i.shape,o),f=c.map(y=>n.makeTensorInfo([y.length],"int32",y)),g=n.makeTensorInfo(m,s.dtype,h);return f.concat([g])}var dae={kernelName:bm,backendName:"webgl",kernelFunc:cae};function hae(e){let{inputs:t,backend:n}=e,{starts:a,limits:r,deltas:s}=t,i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=m7(i,a.shape,a.dtype,o,r.shape,l,s.shape),d=n.makeTensorInfo([u.length],"int32",u),c=n.makeTensorInfo([p.length],a.dtype,p);return[d,c]}var mae={kernelName:xm,backendName:"webgl",kernelFunc:hae};function fae(e){let{inputs:t,backend:n,attrs:a}=e,{shape:r,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=a,u=n.readSync(r.dataId),p=n.readSync(s.dataId),d=n.readSync(i.dataId),c=o.map(g=>n.readSync(g.dataId)),h=o.map(g=>g.shape),[m,f]=f7(u,r.shape,p,s.shape,s.dtype,d,i.shape,c,h,l);return n.makeTensorInfo(m,s.dtype,f)}var gae={kernelName:vm,backendName:"webgl",kernelFunc:fae},gE=e=>{let{backend:t,attrs:n}=e,{start:a,stop:r,step:s,dtype:i}=n,o=g7(a,r,s,i);return t.makeTensorInfo([o.length],i,o)},yae={kernelName:hc,backendName:"webgl",kernelFunc:gE},bae="return 1.0 / x;",xae=Ye({opSnippet:bae}),vae={kernelName:ou,backendName:"webgl",kernelFunc:xae},wae=Da+`
+`;
+var pow3 = binaryKernelFunc2({ opSnippet: POW, packedOpSnippet: POW_PACKED });
+var powConfig2 = {
+  kernelName: Pow,
+  backendName: "webgl",
+  kernelFunc: pow3
+};
+function prod3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const toDispose = [];
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    toDispose.push(permutedX);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("prod", axes, xRank);
+  let res;
+  if (backend2.shouldExecuteOnCPU([permutedX])) {
+    const xVals = backend2.texData.get(permutedX.dataId).values;
+    const { outVals, outShape, outDtype } = prodImplCPU(permutedX.shape, permutedX.dtype, xVals, axes);
+    res = backend2.makeTensorInfo(outShape, outDtype, outVals);
+  } else {
+    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+    const inSize = util_exports.sizeFromShape(reduceShape);
+    const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+    const outputDType = sumOutType(x.dtype);
+    const reduced = reduce(a2D, outputDType, "prod", backend2);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+    toDispose.push(a2D);
+    toDispose.push(reduced);
+  }
+  if (keepDims) {
+    toDispose.push(res);
+    const newShape = backend_util_exports.expandShapeToKeepDim(res.shape, origAxes);
+    res = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: newShape } });
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return res;
+}
+var prodConfig2 = {
+  kernelName: Prod,
+  backendName: "webgl",
+  kernelFunc: prod3
+};
+function raggedGather3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;
+  const { outputRaggedRank } = attrs;
+  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.readSync(t.dataId));
+  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);
+  const $paramsDenseValues = backend2.readSync(paramsDenseValues.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImplCPU($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);
+  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], "int32", splits));
+  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);
+  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);
+}
+var raggedGatherConfig2 = {
+  kernelName: RaggedGather,
+  backendName: "webgl",
+  kernelFunc: raggedGather3
+};
+function raggedRange3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { starts, limits, deltas } = inputs;
+  const $starts = backend2.readSync(starts.dataId);
+  const $limits = backend2.readSync(limits.dataId);
+  const $deltas = backend2.readSync(deltas.dataId);
+  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImplCPU($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);
+  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], "int32", rtNestedSplitsData);
+  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);
+  return [rtNestedSplits, rtDenseValues];
+}
+var raggedRangeConfig2 = {
+  kernelName: RaggedRange,
+  backendName: "webgl",
+  kernelFunc: raggedRange3
+};
+function raggedTensorToTensor3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { shape, values, defaultValue, rowPartitionTensors } = inputs;
+  const { rowPartitionTypes } = attrs;
+  const $shape = backend2.readSync(shape.dataId);
+  const $values = backend2.readSync(values.dataId);
+  const $defaultValue = backend2.readSync(defaultValue.dataId);
+  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.readSync(t.dataId));
+  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);
+  const [outputShape, output] = raggedTensorToTensorImplCPU($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);
+  return backend2.makeTensorInfo(outputShape, values.dtype, output);
+}
+var raggedTensorToTensorConfig2 = {
+  kernelName: RaggedTensorToTensor,
+  backendName: "webgl",
+  kernelFunc: raggedTensorToTensor3
+};
+var range4 = (args) => {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, step: step5, dtype } = attrs;
+  const values = rangeImplCPU(start, stop, step5, dtype);
+  return backend2.makeTensorInfo([values.length], dtype, values);
+};
+var rangeConfig2 = {
+  kernelName: Range,
+  backendName: "webgl",
+  kernelFunc: range4
+};
+var RECIPROCAL = `return 1.0 / x;`;
+var reciprocal3 = unaryKernelFunc2({ opSnippet: RECIPROCAL });
+var reciprocalConfig2 = {
+  kernelName: Reciprocal,
+  backendName: "webgl",
+  kernelFunc: reciprocal3
+};
+var RELU3 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : x;
-`,kae=`
+`;
+var RELU_PACKED = `
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4049,9 +59616,17 @@ return a / b;`,Ane=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Iae=Ye({opSnippet:wae,packedOpSnippet:kae}),Sae={kernelName:Xi,backendName:"webgl",kernelFunc:Iae},Tae=Da+`
+`;
+var relu3 = unaryKernelFunc2({ opSnippet: RELU3, packedOpSnippet: RELU_PACKED });
+var reluConfig2 = {
+  kernelName: Relu,
+  backendName: "webgl",
+  kernelFunc: relu3
+};
+var RELU63 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,Nae=`
+`;
+var RELU6_PACKED = `
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4061,11 +59636,38 @@ return a / b;`,Ane=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Cae=Ye({opSnippet:Tae,packedOpSnippet:Nae}),_ae={kernelName:Ji,backendName:"webgl",kernelFunc:Cae},Eae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d;r?d="(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":d="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+`;
+var relu63 = unaryKernelFunc2({ opSnippet: RELU63, packedOpSnippet: RELU6_PACKED });
+var relu6Config2 = {
+  kernelName: Relu6,
+  backendName: "webgl",
+  kernelFunc: relu63
+};
+var ResizeBilinearProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)`;
+    } else {
+      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4074,7 +59676,7 @@ return a / b;`,Ane=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${d};
+        vec2 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the four integer indices.
         ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));
@@ -4094,13 +59696,38 @@ return a / b;`,Ane=`
 
         setOutput(newValue);
       }
-    `}},Aae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d;r?d="(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":d="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+var ResizeBilinearPackedProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)`;
+    } else {
+      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]},
-          ${u[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
-                                     ${o}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,
+                                     ${oldWidth}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4114,7 +59741,7 @@ return a / b;`,Ane=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${d};
+        vec3 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the four integer indices.
         ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));
@@ -4122,8 +59749,8 @@ return a / b;`,Ane=`
           min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
-        bool hasNextRow = coords.z < ${n-1};
+        bool hasNextCol = d < ${depth - 1};
+        bool hasNextRow = coords.z < ${newWidth - 1};
 
         // In parallel, construct four corners for all four components in
         // packed 2x2 cell.
@@ -4171,7 +59798,44 @@ return a / b;`,Ane=`
 
         setOutput(newValue);
       }
-    `}};function $ae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,p=H().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Aae(r.shape,l,u,s,i):new Eae(r.shape,l,u,s,i);return n.runWebGLProgram(p,[r],"float32")}var Fae={kernelName:Zi,backendName:"webgl",kernelFunc:$ae},Dae=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,a,r]=t,[,s,i]=e,o=[n&&s>1?a-1:a,n&&i>1?r-1:r],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],p=o[1]/l[1],d=1/u,c=1/p,h=Math.ceil(d)*2+2,m=Math.ceil(c)*2+2;this.userCode=`
+    `;
+  }
+};
+function resizeBilinear3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const program = env().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new ResizeBilinearPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeBilinearProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);
+  return backend2.runWebGLProgram(program, [images], "float32");
+}
+var resizeBilinearConfig2 = {
+  kernelName: ResizeBilinear,
+  backendName: "webgl",
+  kernelFunc: resizeBilinear3
+};
+var ResizeBilinearBackpropProgram = class {
+  constructor(dyShape, inputShape, alignCorners) {
+    this.variableNames = ["dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    const [, xHeight, xWidth] = inputShape;
+    const [, yHeight, yWidth] = dyShape;
+    const effectiveXSize = [
+      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+    ];
+    const effectiveYSize = [
+      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+    ];
+    const heightScale = effectiveXSize[0] / effectiveYSize[0];
+    const widthScale = effectiveXSize[1] / effectiveYSize[1];
+    const invHeightScale = 1 / heightScale;
+    const invWidthScale = 1 / widthScale;
+    const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+    const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4181,14 +59845,14 @@ return a / b;`,Ane=`
 
         float accumulator = 0.0;
 
-        const float heightScale = float(${u});
-        const float widthScale = float(${p});
+        const float heightScale = float(${heightScale});
+        const float widthScale = float(${widthScale});
 
-        const float invHeightScale = float(${d});
-        const float invWidthScale = float(${c});
+        const float invHeightScale = float(${invHeightScale});
+        const float invWidthScale = float(${invWidthScale});
 
-        const int winHeight = int(${h});
-        const int winWidth = int(${m});
+        const int winHeight = int(${winHeight});
+        const int winWidth = int(${winWidth});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4202,7 +59866,7 @@ return a / b;`,Ane=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${s}) {
+          if (dyR < 0 || dyR >= ${yHeight}) {
             continue;
           }
 
@@ -4210,19 +59874,19 @@ return a / b;`,Ane=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${i}) {
+            if (dyC < 0 || dyC >= ${yWidth}) {
               continue;
             }
 
             float dxR = float(dyR) * heightScale;
             int topDxRIndex = int(floor(dxR));
-            int bottomDxRIndex = int(min(ceil(dxR), ${a-1}.0));
+            int bottomDxRIndex = int(min(ceil(dxR), ${xHeight - 1}.0));
             float dxRLerp = dxR - float(topDxRIndex);
             float inverseDxRLerp = 1.0 - dxRLerp;
 
             float dxC = float(dyC) * widthScale;
             int leftDxCIndex = int(floor(dxC));
-            int rightDxCIndex = int(min(ceil(dxC), ${r-1}.0));
+            int rightDxCIndex = int(min(ceil(dxC), ${xWidth - 1}.0));
             float dxCLerp = dxC - float(leftDxCIndex);
             float inverseDxCLerp = 1.0 - dxCLerp;
 
@@ -4252,11 +59916,47 @@ return a / b;`,Ane=`
 
         setOutput(accumulator);
       }
-    `}};function Rae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a,o=new Dae(s.shape,r.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}var Mae={kernelName:Im,backendName:"webgl",kernelFunc:Rae},Pae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d=a?"0.5":"0.0",c;r?c="max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":c="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+function resizeBilinearGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  const program = new ResizeBilinearBackpropProgram(dy.shape, images.shape, alignCorners);
+  return backend2.runWebGLProgram(program, [dy], dy.dtype);
+}
+var resizeBilinearGradConfig3 = {
+  kernelName: ResizeBilinearGrad,
+  backendName: "webgl",
+  kernelFunc: resizeBilinearGrad2
+};
+var ResizeNearestNeighborProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    const roundBase = alignCorners ? "0.5" : "0.0";
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))`;
+    } else {
+      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4265,22 +59965,48 @@ return a / b;`,Ane=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${c};
+        vec2 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the coordinators of nearest neighbor point.
         ivec2 sourceNearestRC = ivec2(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${d})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));
         float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);
 
         setOutput(newValue);
       }
-    `}},Oae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d=a?"0.5":"0.0",c;r?c="max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":c="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+var ResizeNearestNeighborPackedProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    const roundBase = alignCorners ? "0.5" : "0.0";
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))`;
+    } else {
+      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]},
-          ${u[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
-                                     ${o}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,
+                                     ${oldWidth}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4294,15 +60020,15 @@ return a / b;`,Ane=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${c};
+        vec3 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the coordinators of nearest neighbor point.
         ivec3 sourceNearestRC = ivec3(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${d})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
-        bool hasNextRow = coords.z < ${n-1};
+        bool hasNextCol = d < ${depth - 1};
+        bool hasNextRow = coords.z < ${newWidth - 1};
 
         vec4 newValue = vec4(
           getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),
@@ -4315,7 +60041,44 @@ return a / b;`,Ane=`
 
         setOutput(newValue);
       }
-    `}};function Lae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,p=H().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Oae(r.shape,l,u,s,i):new Pae(r.shape,l,u,s,i);return n.runWebGLProgram(p,[r],r.dtype)}var zae={kernelName:Yi,backendName:"webgl",kernelFunc:Lae},Wae=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,a,r]=t,[,s,i]=e,o=[n&&s>1?a-1:a,n&&i>1?r-1:r],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],p=o[1]/l[1],d=1/u,c=1/p,h=Math.ceil(d)*2+2,m=Math.ceil(c)*2+2;this.userCode=`
+    `;
+  }
+};
+function resizeNearestNeighbor3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const program = env().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new ResizeNearestNeighborPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeNearestNeighborProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);
+  return backend2.runWebGLProgram(program, [images], images.dtype);
+}
+var resizeNearestNeighborConfig2 = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "webgl",
+  kernelFunc: resizeNearestNeighbor3
+};
+var ResizeNearestNeigborBackpropProgram = class {
+  constructor(dyShape, inputShape, alignCorners) {
+    this.variableNames = ["dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    const [, xHeight, xWidth] = inputShape;
+    const [, yHeight, yWidth] = dyShape;
+    const effectiveXSize = [
+      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+    ];
+    const effectiveYSize = [
+      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+    ];
+    const heightScale = effectiveXSize[0] / effectiveYSize[0];
+    const widthScale = effectiveXSize[1] / effectiveYSize[1];
+    const invHeightScale = 1 / heightScale;
+    const invWidthScale = 1 / widthScale;
+    const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+    const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4325,14 +60088,14 @@ return a / b;`,Ane=`
 
         float accumulator = 0.0;
 
-        const float heightScale = float(${u});
-        const float widthScale = float(${p});
+        const float heightScale = float(${heightScale});
+        const float widthScale = float(${widthScale});
 
-        const float invHeightScale = float(${d});
-        const float invWidthScale = float(${c});
+        const float invHeightScale = float(${invHeightScale});
+        const float invWidthScale = float(${invWidthScale});
 
-        const int winHeight = int(${h});
-        const int winWidth = int(${m});
+        const int winHeight = int(${winHeight});
+        const int winWidth = int(${winWidth});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4346,7 +60109,7 @@ return a / b;`,Ane=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${s}) {
+          if (dyR < 0 || dyR >= ${yHeight}) {
             continue;
           }
 
@@ -4354,26 +60117,26 @@ return a / b;`,Ane=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${i}) {
+            if (dyC < 0 || dyC >= ${yWidth}) {
               continue;
             }
 
             float sourceFracRow =
-              float(${o[0]}) *
-                (float(dyR) / float(${l[0]}));
+              float(${effectiveXSize[0]}) *
+                (float(dyR) / float(${effectiveYSize[0]}));
 
             float sourceFracCol =
-                float(${o[1]}) *
-                  (float(dyC) / float(${l[1]}));
+                float(${effectiveXSize[1]}) *
+                  (float(dyC) / float(${effectiveYSize[1]}));
 
             int sourceNearestRow = int(min(
-                float(int(${a}) - 1),
-                ${n} ? float(round(sourceFracRow)) :
+                float(int(${xHeight}) - 1),
+                ${alignCorners} ? float(round(sourceFracRow)) :
                                   float(floor(sourceFracRow))));
 
             int sourceNearestCol = int(min(
-                float(int(${r}) - 1),
-                ${n} ? float(round(sourceFracCol)) :
+                float(int(${xWidth}) - 1),
+                ${alignCorners} ? float(round(sourceFracCol)) :
                                   float(floor(sourceFracCol))));
 
             if (r == sourceNearestRow && c == sourceNearestCol) {
@@ -4385,47 +60148,166 @@ return a / b;`,Ane=`
 
         setOutput(accumulator);
       }
-    `}};function Bae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a,o=new Wae(s.shape,r.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}var Vae={kernelName:km,backendName:"webgl",kernelFunc:Bae},Uae=class{constructor(e,t){this.variableNames=["x"];let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=e,n===1){this.userCode=`
+    `;
+  }
+};
+function resizeNearestNeighborGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  const program = new ResizeNearestNeigborBackpropProgram(dy.shape, images.shape, alignCorners);
+  return backend2.runWebGLProgram(program, [dy], dy.dtype);
+}
+var resizeNearestNeighborGradConfig3 = {
+  kernelName: ResizeNearestNeighborGrad,
+  backendName: "webgl",
+  kernelFunc: resizeNearestNeighborGrad2
+};
+var ReverseProgram = class {
+  constructor(xShape, axis) {
+    this.variableNames = ["x"];
+    const rank = xShape.length;
+    if (rank > 4) {
+      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);
+    }
+    this.outputShape = xShape;
+    if (rank === 1) {
+      this.userCode = `
         void main() {
           int coord = getOutputCoords();
-          setOutput(getX(${e[0]} - coord - 1));
+          setOutput(getX(${xShape[0]} - coord - 1));
         }
-      `;return}let a=i=>t.indexOf(i)!==-1&&e[i]!==1?`${e[i]} - coords[${i}] - 1`:`coords[${i}]`,r=e.map((i,o)=>a(o)).join(","),s=gt(n);this.userCode=`
-      void main() {
-        ${s} coords = getOutputCoords();
-        setOutput(getX(${r}));
+      `;
+      return;
+    }
+    const getInCoord = (i) => {
+      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {
+        return `${xShape[i]} - coords[${i}] - 1`;
       }
-    `}},Gae=class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=e;let a=wn("rc",n),r=`${a[n-1]} + 1 < ${this.outputShape[n-1]}`,s=`${a[n-2]} + 1 < ${this.outputShape[n-2]}`,i=gt(n);n===1?this.userCode=`
+      return `coords[${i}]`;
+    };
+    const inCoords = xShape.map((_, i) => getInCoord(i)).join(",");
+    const type = getCoordsDataType(rank);
+    this.userCode = `
+      void main() {
+        ${type} coords = getOutputCoords();
+        setOutput(getX(${inCoords}));
+      }
+    `;
+  }
+};
+var ReversePackedProgram = class {
+  constructor(xShape, axis) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const rank = xShape.length;
+    if (rank > 4) {
+      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);
+    }
+    this.outputShape = xShape;
+    const channels = getChannels("rc", rank);
+    const nextColumn = `${channels[rank - 1]} + 1 < ${this.outputShape[rank - 1]}`;
+    const nextRow = `${channels[rank - 2]} + 1 < ${this.outputShape[rank - 2]}`;
+    const type = getCoordsDataType(rank);
+    if (rank === 1) {
+      this.userCode = `
         void main(){
           int rc = getOutputCoords();
           vec4 result = vec4(0.);
-          result.r = getChannel(getX(${e[0]} - rc - 1),
-            ${e[0]} - rc - 1);
-          if(${r}){
-              result.g = getChannel(getX(${e[0]} - (rc  + 1) - 1),
-                ${e[0]} - (rc  + 1) - 1);
+          result.r = getChannel(getX(${xShape[0]} - rc - 1),
+            ${xShape[0]} - rc - 1);
+          if(${nextColumn}){
+              result.g = getChannel(getX(${xShape[0]} - (rc  + 1) - 1),
+                ${xShape[0]} - (rc  + 1) - 1);
           }
           setOutput(result);
         }
-      `:this.userCode=`
+      `;
+    } else {
+      this.userCode = `
         void main() {
-          ${i} rc = getOutputCoords();
+          ${type} rc = getOutputCoords();
           vec4 result = vec4(0.);
-          result.r = ${o(a.slice())};
-          if(${r}){
-            result.g = ${l(a.slice())};
+          result.r = ${getR(channels.slice())};
+          if(${nextColumn}){
+            result.g = ${getG(channels.slice())};
           }
-          if(${s}) {
-            result.b = ${u(a.slice())};
-            if(${r}) {
-              result.a = ${p(a.slice())};
+          if(${nextRow}) {
+            result.b = ${getB(channels.slice())};
+            if(${nextColumn}) {
+              result.a = ${getA(channels.slice())};
             }
           }
           setOutput(result);
         }
-    `;function o(h){return d(h)}function l(h){return h[n-1]="("+h[n-1]+" + 1)",d(h)}function u(h){return h[n-2]="("+h[n-2]+" + 1)",d(h)}function p(h){return h[n-1]="("+h[n-1]+" + 1)",h[n-2]="("+h[n-2]+" + 1)",d(h)}function d(h){let m=e.map((y,b)=>c(b,h)),f=m.join(","),g=m.slice(-2).join(",");return`getChannel(getX(${f}), vec2(${g}))`}function c(h,m){return t.indexOf(h)!==-1&&e[h]!==1?`${e[h]} - ${m[h]} - 1`:`${m[h]}`}}};function Hae(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a,i=r.shape.length,o=v.parseAxisParam(s,r.shape);if(i===0)return na({inputs:{x:r},backend:n});let l=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new Gae(r.shape,o):new Uae(r.shape,o);return n.runWebGLProgram(l,[r],r.dtype)}var jae={kernelName:Qi,backendName:"webgl",kernelFunc:Hae},qae=class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let n=e[1],a=e[2];this.outputShape=e;let r="";typeof t=="number"?r=`float outputValue = ${t.toFixed(2)};`:r=`
-        vec3 fill = vec3(${t.join(",")});
-        float outputValue = fill[coords[3]];`,this.userCode=`
+    `;
+    }
+    function getR(channels2) {
+      return getChannel(channels2);
+    }
+    function getG(channels2) {
+      channels2[rank - 1] = "(" + channels2[rank - 1] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getB(channels2) {
+      channels2[rank - 2] = "(" + channels2[rank - 2] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getA(channels2) {
+      channels2[rank - 1] = "(" + channels2[rank - 1] + ` + 1)`;
+      channels2[rank - 2] = "(" + channels2[rank - 2] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getChannel(channels2) {
+      const inCoordsArray = xShape.map((_, i) => getInCoord(i, channels2));
+      const inCoords = inCoordsArray.join(",");
+      const innerDims = inCoordsArray.slice(-2).join(",");
+      return `getChannel(getX(${inCoords}), vec2(${innerDims}))`;
+    }
+    function getInCoord(i, channels1) {
+      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {
+        return `${xShape[i]} - ${channels1[i]} - 1`;
+      } else {
+        return `${channels1[i]}`;
+      }
+    }
+  }
+};
+function reverse3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  const xRank = x.shape.length;
+  const $dims = util_exports.parseAxisParam(dims, x.shape);
+  if (xRank === 0) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new ReversePackedProgram(x.shape, $dims) : new ReverseProgram(x.shape, $dims);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var reverseConfig2 = {
+  kernelName: Reverse,
+  backendName: "webgl",
+  kernelFunc: reverse3
+};
+var RotateProgram = class {
+  constructor(imageShape, fillValue) {
+    this.variableNames = ["Image"];
+    this.outputShape = [];
+    this.customUniforms = [{ name: "params", type: "vec4" }];
+    const imageHeight = imageShape[1];
+    const imageWidth = imageShape[2];
+    this.outputShape = imageShape;
+    let fillSnippet = "";
+    if (typeof fillValue === "number") {
+      fillSnippet = `float outputValue = ${fillValue.toFixed(2)};`;
+    } else {
+      fillSnippet = `
+        vec3 fill = vec3(${fillValue.join(",")});
+        float outputValue = fill[coords[3]];`;
+    }
+    this.userCode = `
         void main() {
           ivec4 coords = getOutputCoords();
           int x = coords[2];
@@ -4436,13 +60318,30 @@ return a / b;`,Ane=`
             (float(y) - params[1]) * params[3];
           int coordX = int(round(coordXFloat + params[0]));
           int coordY = int(round(coordYFloat + params[1]));
-          ${r}
-          if(coordX >= 0 && coordX < ${a} && coordY >= 0 && coordY < ${n}) {
+          ${fillSnippet}
+          if(coordX >= 0 && coordX < ${imageWidth} && coordY >= 0 && coordY < ${imageHeight}) {
             outputValue = getImage(coords[0], coordY, coordX, coords[3]);
           }
           setOutput(outputValue);
         }
-    `}},Kae={kernelName:Su,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,{radians:r,fillValue:s,center:i}=t,o=n,l=new qae(a.shape,s),[u,p]=N.getImageCenter(i,a.shape[1],a.shape[2]),d=[[u,p,Math.sin(r),Math.cos(r)]];return o.runWebGLProgram(l,[a],a.dtype,d)}},Xae=`
+    `;
+  }
+};
+var rotateWithOffsetConfig2 = {
+  kernelName: RotateWithOffset,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const { radians, fillValue, center } = attrs;
+    const webglBackend = backend2;
+    const program = new RotateProgram(image2.shape, fillValue);
+    const [centerX, centerY] = backend_util_exports.getImageCenter(center, image2.shape[1], image2.shape[2]);
+    const customValues = [[centerX, centerY, Math.sin(radians), Math.cos(radians)]];
+    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype, customValues);
+    return output;
+  }
+};
+var ROUND = `
   // OpenGL ES does not support round function.
   // The algorithm is based on banker's rounding.
   float base = floor(x);
@@ -4457,34 +60356,107 @@ return a / b;`,Ane=`
       return base + 1.0;
     }
   }
-`,Yae=Ye({opSnippet:Xae}),Zae={kernelName:eo,backendName:"webgl",kernelFunc:Yae},Jae="return inversesqrt(x);",Qae=Ye({opSnippet:Jae,cpuKernelImpl:y7}),ere={kernelName:to,backendName:"webgl",kernelFunc:Qae},yE=class{constructor(e,t,n,a,r,s,i=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=s;let o=gt(r.length),l=gt(s.length),u="";n===1?u="i":n===2&&(u="i, j");let p=`getIndices(${u})`,d="";a===1?d="i":a===2&&(d="i, coords[1]");let c=`getUpdates(${d})`,h=t>1?"strides[j]":"strides";this.userCode=`
-        ${o} strides = ${o}(${r});
+`;
+var round4 = unaryKernelFunc2({ opSnippet: ROUND });
+var roundConfig2 = {
+  kernelName: Round,
+  backendName: "webgl",
+  kernelFunc: round4
+};
+var RSQRT = `return inversesqrt(x);`;
+var rsqrt3 = unaryKernelFunc2({ opSnippet: RSQRT, cpuKernelImpl: rsqrtImplCPU });
+var rsqrtConfig2 = {
+  kernelName: Rsqrt,
+  backendName: "webgl",
+  kernelFunc: rsqrt3
+};
+var ScatterProgram = class {
+  constructor(updateSize, sliceDim, indicesRank, updatesRank, strides, shape, summingDupeIndex = true) {
+    this.variableNames = ["updates", "indices", "defaultValue"];
+    this.outputShape = shape;
+    const stridesType = getCoordsDataType(strides.length);
+    const dtype = getCoordsDataType(shape.length);
+    let indicesString = "";
+    if (indicesRank === 1) {
+      indicesString = "i";
+    } else if (indicesRank === 2) {
+      indicesString = "i, j";
+    }
+    const indicesSnippet = `getIndices(${indicesString})`;
+    let updatesString = "";
+    if (updatesRank === 1) {
+      updatesString = "i";
+    } else if (updatesRank === 2) {
+      updatesString = "i, coords[1]";
+    }
+    const updatesSnippet = `getUpdates(${updatesString})`;
+    const strideString = sliceDim > 1 ? "strides[j]" : "strides";
+    this.userCode = `
+        ${stridesType} strides = ${stridesType}(${strides});
 
         void main() {
-          ${l} coords = getOutputCoords();
+          ${dtype} coords = getOutputCoords();
           float sum = 0.0;
           bool found = false;
-          for (int i = 0; i < ${e}; i++) {
+          for (int i = 0; i < ${updateSize}; i++) {
             int flattenedIndex = 0;
-            for (int j = 0; j < ${t}; j++) {
-              int index = round(${p});
-              flattenedIndex += index * ${h};
+            for (int j = 0; j < ${sliceDim}; j++) {
+              int index = round(${indicesSnippet});
+              flattenedIndex += index * ${strideString};
             }
             if (flattenedIndex == coords[0]) {
-              sum += ${c};
+              sum += ${updatesSnippet};
               found = true;
             }
           }
           setOutput(mix(getDefaultValue(), sum, float(found)));
         }
-      `}};function tre(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r,updates:s}=t,{shape:i}=a,{sliceRank:o,numUpdates:l,sliceSize:u,strides:p,outputSize:d}=N.calculateShapes(s,r,i),c=[d/u,u];if(d===0)return n.makeTensorInfo(i,r.dtype);let h=de({inputs:{x:r},backend:n,attrs:{shape:[l,o]}}),m=de({inputs:{x:s},backend:n,attrs:{shape:[l,u]}}),f=n.makeTensorInfo([],"float32",new Float32Array([0])),g=new yE(l,o,h.shape.length,m.shape.length,p,c),y=n.runWebGLProgram(g,[m,h,f],m.dtype),b=de({inputs:{x:y},backend:n,attrs:{shape:i}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(f),b}var nre={kernelName:uu,backendName:"webgl",kernelFunc:tre},are=class{constructor(e,t,n,a){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[e,n];let r="while (left < right) {",s=`for (int i = 0; i < ${Math.ceil(Math.log2(t+1))}; ++i) { if (left >= right) break;`,i=H().getNumber("WEBGL_VERSION")===2?r:s,o=a==="left"?"<":"<=";this.userCode=`
+      `;
+  }
+};
+function scatterNd2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);
+  const flattenShape = [outputSize / sliceSize, sliceSize];
+  if (outputSize === 0) {
+    return backend2.makeTensorInfo(shape, indices.dtype);
+  }
+  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numUpdates, sliceRank] } });
+  const flattenX = reshape4({ inputs: { x: updates }, backend: backend2, attrs: { shape: [numUpdates, sliceSize] } });
+  const defaultValue = backend2.makeTensorInfo([], "float32", new Float32Array([0]));
+  const program = new ScatterProgram(numUpdates, sliceRank, flattenIndices.shape.length, flattenX.shape.length, strides, flattenShape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices, defaultValue], flattenX.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape } });
+  backend2.disposeIntermediateTensorInfo(flattenIndices);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(res);
+  backend2.disposeIntermediateTensorInfo(defaultValue);
+  return reshaped;
+}
+var scatterNdConfig2 = {
+  kernelName: ScatterNd,
+  backendName: "webgl",
+  kernelFunc: scatterNd2
+};
+var SearchSortedProgram = class {
+  constructor(batchSize, numInputs, numValues, side) {
+    this.variableNames = ["sortedSequence", "values"];
+    this.customUniforms = [{ name: "numInputs", type: "int" }];
+    this.outputShape = [batchSize, numValues];
+    const webGL2LoopHead = "while (left < right) {";
+    const webGL1LoopHead = `for (int i = 0; i < ${Math.ceil(Math.log2(numInputs + 1))}; ++i) { if (left >= right) break;`;
+    const loopHead = env().getNumber("WEBGL_VERSION") === 2 ? webGL2LoopHead : webGL1LoopHead;
+    const boundComparator = side === "left" ? "<" : "<=";
+    this.userCode = `
        int findBound(int batch, float value) {
          int left = 0;
          int right = numInputs;
          int mid;
-         ${i}
+         ${loopHead}
            mid = (left + right) / 2;
-           if (getSortedSequence(batch, mid) ${o} value) {
+           if (getSortedSequence(batch, mid) ${boundComparator} value) {
              left = mid + 1;
            } else {
              right = mid;
@@ -4502,25 +60474,89 @@ return a / b;`,Ane=`
 
          setOutput(float(findBound(batch, value)));
        }
-     `}};function rre(e){let{inputs:t,backend:n,attrs:a}=e,{sortedSequence:r,values:s}=t,{side:i}=a,o=new are(r.shape[0],r.shape[1],s.shape[1],i),l=[[r.shape[1]]];return n.runWebGLProgram(o,[r,s],"int32",l)}var sre={kernelName:Sm,backendName:"webgl",kernelFunc:rre},ire=class{constructor(e,t,n){this.variableNames=["c","a","b"],this.outputShape=t;let a,r;if(n>4)throw Error(`Where for rank ${n} is not yet supported`);if(n===1)r="resRC",a="resRC";else{let i=["resRC.x","resRC.y","resRC.z","resRC.w"],o=[],l=[];for(let u=0;u<t.length;u++)l.push(`${i[u]}`),u<e&&o.push(`${i[u]}`);a=o.join(),r=l.join()}let s=gt(n);this.userCode=`
-      void main() {
-        ${s} resRC = getOutputCoords();
-        float cVal = getC(${a});
-        if (cVal >= 1.0) {
-          setOutput(getA(${r}));
-        } else {
-          setOutput(getB(${r}));
+     `;
+  }
+};
+function searchSorted3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sortedSequence, values } = inputs;
+  const { side } = attrs;
+  const program = new SearchSortedProgram(sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);
+  const customValues = [[sortedSequence.shape[1]]];
+  return backend2.runWebGLProgram(program, [sortedSequence, values], "int32", customValues);
+}
+var searchSortedConfig2 = {
+  kernelName: SearchSorted,
+  backendName: "webgl",
+  kernelFunc: searchSorted3
+};
+var SelectProgram = class {
+  constructor(cRank, shape, rank) {
+    this.variableNames = ["c", "a", "b"];
+    this.outputShape = shape;
+    let cCoords;
+    let abCoords;
+    if (rank > 4) {
+      throw Error(`Where for rank ${rank} is not yet supported`);
+    }
+    if (rank === 1) {
+      abCoords = `resRC`;
+      cCoords = `resRC`;
+    } else {
+      const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"];
+      const cCoordVars = [];
+      const abCoordVars = [];
+      for (let i = 0; i < shape.length; i++) {
+        abCoordVars.push(`${currentCoords[i]}`);
+        if (i < cRank) {
+          cCoordVars.push(`${currentCoords[i]}`);
         }
       }
-    `}};function ore(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t,i=new ire(a.shape.length,r.shape,r.shape.length);return n.runWebGLProgram(i,[a,r,s],ma(r.dtype,s.dtype))}var lre={kernelName:pu,backendName:"webgl",kernelFunc:ore},ure=`
+      cCoords = cCoordVars.join();
+      abCoords = abCoordVars.join();
+    }
+    const dtype = getCoordsDataType(rank);
+    this.userCode = `
+      void main() {
+        ${dtype} resRC = getOutputCoords();
+        float cVal = getC(${cCoords});
+        if (cVal >= 1.0) {
+          setOutput(getA(${abCoords}));
+        } else {
+          setOutput(getB(${abCoords}));
+        }
+      }
+    `;
+  }
+};
+function select3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  const program = new SelectProgram(condition.shape.length, t.shape, t.shape.length);
+  return backend2.runWebGLProgram(program, [condition, t, e], upcastType(t.dtype, e.dtype));
+}
+var selectConfig2 = {
+  kernelName: Select,
+  backendName: "webgl",
+  kernelFunc: select3
+};
+var SELU = `
   // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.
   // see: https://arxiv.org/abs/1706.02515
-  float scaleAlpha = ${N.SELU_SCALEALPHA};
-  float scale = ${N.SELU_SCALE};
+  float scaleAlpha = ${backend_util_exports.SELU_SCALEALPHA};
+  float scale = ${backend_util_exports.SELU_SCALE};
   return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);
-`,pre=Ye({opSnippet:ure}),cre={kernelName:cu,backendName:"webgl",kernelFunc:pre},dre=Uu+`
+`;
+var selu3 = unaryKernelFunc2({ opSnippet: SELU });
+var seluConfig2 = {
+  kernelName: Selu,
+  backendName: "webgl",
+  kernelFunc: selu3
+};
+var SIGMOID3 = CHECK_NAN_SNIPPET_UNARY + `
   return 1.0 / (1.0 + exp(-1.0 * x));
-`,hre=`
+`;
+var SIGMOID_PACKED = `
   vec4 result = 1.0 / (1.0 + exp(-1.0 * x));
   bvec4 isNaN = isnan(x);
 
@@ -4530,15 +60566,47 @@ return a / b;`,Ane=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,mre=Ye({opSnippet:dre,packedOpSnippet:hre,cpuKernelImpl:x7}),fre={kernelName:ao,backendName:"webgl",kernelFunc:mre},gre=`
+`;
+var sigmoid3 = unaryKernelFunc2({
+  opSnippet: SIGMOID3,
+  packedOpSnippet: SIGMOID_PACKED,
+  cpuKernelImpl: sigmoidImplCPU
+});
+var sigmoidConfig2 = {
+  kernelName: Sigmoid,
+  backendName: "webgl",
+  kernelFunc: sigmoid3
+};
+var SIGN = `
   if (isnan(x)) { return 0.0; }
   return sign(x);
-`,yre=Ye({opSnippet:gre}),bre={kernelName:mu,backendName:"webgl",kernelFunc:yre},xre=Uu+`
+`;
+var sign3 = unaryKernelFunc2({ opSnippet: SIGN });
+var signConfig2 = {
+  kernelName: Sign,
+  backendName: "webgl",
+  kernelFunc: sign3
+};
+var SIN = CHECK_NAN_SNIPPET_UNARY + `
   return sin(x);
-`,vre=Ye({opSnippet:xre}),wre={kernelName:no,backendName:"webgl",kernelFunc:vre},kre=`
+`;
+var sin3 = unaryKernelFunc2({ opSnippet: SIN });
+var sinConfig2 = {
+  kernelName: Sin,
+  backendName: "webgl",
+  kernelFunc: sin3
+};
+var SINH = `
   float e2x = exp(x);
   return (e2x - 1.0 / e2x) / 2.0;
-`,Ire=Ye({opSnippet:kre}),Sre={kernelName:hu,backendName:"webgl",kernelFunc:Ire},Tre=`
+`;
+var sinh3 = unaryKernelFunc2({ opSnippet: SINH });
+var sinhConfig2 = {
+  kernelName: Sinh,
+  backendName: "webgl",
+  kernelFunc: sinh3
+};
+var SOFTPLUS = `
   float epsilon = 1.1920928955078125e-7;
   float threshold = log(epsilon) + 2.0;
 
@@ -4558,33 +60626,453 @@ return a / b;`,Ane=`
     result = log(exp_x + 1.0);
   }
   return result;
-`,Nre=Ye({opSnippet:Tre}),Cre={kernelName:fu,backendName:"webgl",kernelFunc:Nre},_re=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a;v.assert(r.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((y,b)=>y*b),l=[[0,0]];l.push(...i);for(let y=1+s.length;y<r.shape.length;++y)l.push([0,0]);let u=[],p=fE({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),d=N.getReshaped(p.shape,s,o,!1),c=N.getPermuted(d.length,s.length,!1),h=N.getReshapedPermuted(p.shape,s,o,!1),m=de({inputs:{x:p},backend:n,attrs:{shape:d}}),f=In({inputs:{x:m},backend:n,attrs:{perm:c}}),g=de({inputs:{x:f},backend:n,attrs:{shape:h}});return u.push(p),u.push(m),u.push(f),u.forEach(y=>n.disposeIntermediateTensorInfo(y)),g},Ere={kernelName:gu,backendName:"webgl",kernelFunc:_re};function Are(e){let{inputs:t,backend:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-         ${s.shape}`);if(a.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-         ${a.shape}`);if(r.shape.length!==1)throw new Error(`Values must be a vector, saw:
-         ${r.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let o=n.readSync(a.dataId),l=n.readSync(r.dataId),u=n.readSync(s.dataId),p=n.readSync(i.dataId)[0],[d,c,h,m,f]=w7(o,a.shape,a.dtype,l,r.dtype,u,p);return[n.makeTensorInfo(c,a.dtype,d),n.makeTensorInfo([c[0]],r.dtype,h),n.makeTensorInfo([m.length],"bool",new Uint8Array(m.map(g=>Number(g)))),n.makeTensorInfo([f.length],a.dtype,new Int32Array(f))]}var $re={kernelName:mc,backendName:"webgl",kernelFunc:Are};function Fre(e){let{inputs:t,backend:n}=e,{inputIndices:a,inputShape:r,newShape:s}=t;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.readSync(r.dataId)),o=n.readSync(a.dataId),l=Array.from(n.readSync(s.dataId)),[u,p,d]=k7(o,a.shape,a.dtype,i,l);return[n.makeTensorInfo(p,a.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}var Dre={kernelName:bu,backendName:"webgl",kernelFunc:Fre};function Rre(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-              ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-              ${s.shape}`);let i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=z_(i,a.shape,a.dtype,o,l,!0);return n.makeTensorInfo(p,a.dtype,u)}var Mre={kernelName:fc,backendName:"webgl",kernelFunc:Rre};function Pre(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-             ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-             ${s.shape}`);let i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=z_(i,a.shape,a.dtype,o,l);return n.makeTensorInfo(p,a.dtype,u)}var Ore={kernelName:gc,backendName:"webgl",kernelFunc:Pre};function Lre(e){let{inputs:t,backend:n,attrs:a}=e,{sparseIndices:r,sparseValues:s,defaultValue:i}=t,{outputShape:o}=a,{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=N.calculateShapes(s,r,o),h=!1;if(s.dtype==="string"){let y=n.bufferSync(r),b=n.bufferSync(s),x=v.decodeString(n.readSync(i.dataId)[0]),w=b7(y,b,o,c,p,u,l,d,x,h);return n.makeTensorInfo(o,w.dtype,w.values)}let m=new yE(u,l,r.shape.length,s.shape.length,d,[c,1],h),f=n.runWebGLProgram(m,[s,r,i],s.dtype),g=de({inputs:{x:f},backend:n,attrs:{shape:o}});return n.disposeIntermediateTensorInfo(f),g}var zre={kernelName:Tm,backendName:"webgl",kernelFunc:Lre};function Wre(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=a,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=r.shape.length,p=new Array(u).fill(0),d=r.shape.slice();return l.map(c=>{let h=[...d];h[o]=c;let m=Gu({inputs:{x:r},backend:n,attrs:{begin:p,size:h}});return p[o]+=c,m})}var Bre={kernelName:yu,backendName:"webgl",kernelFunc:Wre},dI="return sqrt(x);",Vre=Ye({opSnippet:dI,packedOpSnippet:dI,cpuKernelImpl:I7}),Ure={kernelName:ro,backendName:"webgl",kernelFunc:Vre},Gre="return x * x;",Hre=Ye({opSnippet:Gre}),jre={kernelName:yc,backendName:"webgl",kernelFunc:Hre},hI="return (a - b) * (a - b);",qre=pn({opSnippet:hI,packedOpSnippet:hI}),Kre={kernelName:oo,backendName:"webgl",kernelFunc:qre};function Xre({inputs:e,attrs:t,backend:n}){let{x:a}=e,r=Da+`
-    return x > 0.0 ? 1.0 : float(${t.alpha});
-  `,s=new Sr(a.shape,r);return n.runWebGLProgram(s,[a],a.dtype)}var Yre={kernelName:ms,backendName:"webgl",kernelFunc:Xre},Zre=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=n;let a=n.length,r=gt(n.length),s=gt(n.length),i="";if(a===1)i="coords * strides + begin";else{let o=0;i=n.map((l,u)=>(o++,n.length===1?`coords * strides[${u}] + begin[${u}]`:`coords[${o-1}] * strides[${u}] + begin[${u}]`)).join(",")}this.userCode=`
-      ${r} begin = ${r}(${e});
-      ${r} strides = ${r}(${t});
+`;
+var softplus3 = unaryKernelFunc2({ opSnippet: SOFTPLUS });
+var softplusConfig2 = {
+  kernelName: Softplus,
+  backendName: "webgl",
+  kernelFunc: softplus3
+};
+var spaceToBatchND3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  util_exports.assert(x.shape.length <= 4, () => "spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const toDispose = [];
+  const paddedX = padV22({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapedPaddedX = reshape4({ inputs: { x: paddedX }, backend: backend2, attrs: { shape: reshapedPaddedShape } });
+  const paddedXT = transpose3({
+    inputs: { x: reshapedPaddedX },
+    backend: backend2,
+    attrs: { perm: permutedReshapedPaddedPermutation }
+  });
+  const result = reshape4({ inputs: { x: paddedXT }, backend: backend2, attrs: { shape: flattenShape } });
+  toDispose.push(paddedX);
+  toDispose.push(reshapedPaddedX);
+  toDispose.push(paddedXT);
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+};
+var spaceToBatchNDConfig2 = {
+  kernelName: SpaceToBatchND,
+  backendName: "webgl",
+  kernelFunc: spaceToBatchND3
+};
+function sparseFillEmptyRows3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  if (denseShape.shape.length !== 1) {
+    throw new Error(`Dense shape must be a vector, saw:
+         ${denseShape.shape}`);
+  }
+  if (indices.shape.length !== 2) {
+    throw new Error(`Indices must be a matrix, saw:
+         ${indices.shape}`);
+  }
+  if (values.shape.length !== 1) {
+    throw new Error(`Values must be a vector, saw:
+         ${values.shape}`);
+  }
+  if (defaultValue.shape.length !== 0) {
+    throw new Error(`Default value must be a scalar, saw:
+        ${defaultValue.shape}`);
+  }
+  const $indices = backend2.readSync(indices.dataId);
+  const $values = backend2.readSync(values.dataId);
+  const $denseShape = backend2.readSync(denseShape.dataId);
+  const $defaultValue = backend2.readSync(defaultValue.dataId)[0];
+  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImplCPU($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);
+  return [
+    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),
+    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),
+    backend2.makeTensorInfo([emptyRowIndicator.length], "bool", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),
+    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))
+  ];
+}
+var sparseFillEmptyRowsConfig2 = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "webgl",
+  kernelFunc: sparseFillEmptyRows3
+};
+function sparseReshape3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const $inputShape = Array.from(backend2.readSync(inputShape.dataId));
+  const $inputIndices = backend2.readSync(inputIndices.dataId);
+  const targetShape = Array.from(backend2.readSync(newShape.dataId));
+  const [newIndices, indicesShape, outputShape] = sparseReshapeImplCPU($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);
+  return [
+    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),
+    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))
+  ];
+}
+var sparseReshapeConfig2 = {
+  kernelName: SparseReshape,
+  backendName: "webgl",
+  kernelFunc: sparseReshape3
+};
+function sparseSegmentMean3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+              ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+              ${segmentIds.shape}`);
+  }
+  const $data = backend2.readSync(data.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const $segmentIds = backend2.readSync(segmentIds.dataId);
+  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds, true);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentMeanConfig2 = {
+  kernelName: SparseSegmentMean,
+  backendName: "webgl",
+  kernelFunc: sparseSegmentMean3
+};
+function sparseSegmentSum3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+             ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+             ${segmentIds.shape}`);
+  }
+  const $data = backend2.readSync(data.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const $segmentIds = backend2.readSync(segmentIds.dataId);
+  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentSumConfig2 = {
+  kernelName: SparseSegmentSum,
+  backendName: "webgl",
+  kernelFunc: sparseSegmentSum3
+};
+function sparseToDense3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sparseIndices, sparseValues, defaultValue } = inputs;
+  const { outputShape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);
+  const sumDupeIndices = false;
+  if (sparseValues.dtype === "string") {
+    const indicesBuf = backend2.bufferSync(sparseIndices);
+    const updatesBuf = backend2.bufferSync(sparseValues);
+    const $defaultValue = util_exports.decodeString(backend2.readSync(defaultValue.dataId)[0]);
+    const outBuf = scatterImplCPU(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+    return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);
+  }
+  const program = new ScatterProgram(numUpdates, sliceRank, sparseIndices.shape.length, sparseValues.shape.length, strides, [outputSize, 1], sumDupeIndices);
+  const res = backend2.runWebGLProgram(program, [sparseValues, sparseIndices, defaultValue], sparseValues.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outputShape } });
+  backend2.disposeIntermediateTensorInfo(res);
+  return reshaped;
+}
+var sparseToDenseConfig2 = {
+  kernelName: SparseToDense,
+  backendName: "webgl",
+  kernelFunc: sparseToDense3
+};
+function splitV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const xRank = x.shape.length;
+  const begin = new Array(xRank).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const sliceSize = [...size];
+    sliceSize[$axis] = s;
+    const sliceT = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });
+    begin[$axis] += s;
+    return sliceT;
+  });
+}
+var splitVConfig2 = {
+  kernelName: SplitV,
+  backendName: "webgl",
+  kernelFunc: splitV2
+};
+var SQRT = `return sqrt(x);`;
+var sqrt3 = unaryKernelFunc2({ opSnippet: SQRT, packedOpSnippet: SQRT, cpuKernelImpl: sqrtImplCPU });
+var sqrtConfig2 = {
+  kernelName: Sqrt,
+  backendName: "webgl",
+  kernelFunc: sqrt3
+};
+var SQUARE = `return x * x;`;
+var square3 = unaryKernelFunc2({ opSnippet: SQUARE });
+var squareConfig2 = {
+  kernelName: Square,
+  backendName: "webgl",
+  kernelFunc: square3
+};
+var SQUARED_DIFFERENCE = "return (a - b) * (a - b);";
+var squaredDifference3 = binaryKernelFunc2({ opSnippet: SQUARED_DIFFERENCE, packedOpSnippet: SQUARED_DIFFERENCE });
+var squaredDifferenceConfig2 = {
+  kernelName: SquaredDifference,
+  backendName: "webgl",
+  kernelFunc: squaredDifference3
+};
+function step3({ inputs, attrs, backend: backend2 }) {
+  const { x } = inputs;
+  const opSnippet = CHECK_NAN_SNIPPET + `
+    return x > 0.0 ? 1.0 : float(${attrs.alpha});
+  `;
+  const program = new UnaryOpProgram(x.shape, opSnippet);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var stepConfig2 = {
+  kernelName: Step,
+  backendName: "webgl",
+  kernelFunc: step3
+};
+var StridedSliceProgram = class {
+  constructor(begin, strides, size) {
+    this.variableNames = ["x"];
+    this.outputShape = size;
+    const rank = size.length;
+    const inputDtype = getCoordsDataType(size.length);
+    const dtype = getCoordsDataType(size.length);
+    let newCoords = "";
+    if (rank === 1) {
+      newCoords = "coords * strides + begin";
+    } else {
+      let outputAxis = 0;
+      newCoords = size.map((_, i) => {
+        outputAxis++;
+        return size.length === 1 ? `coords * strides[${i}] + begin[${i}]` : `coords[${outputAxis - 1}] * strides[${i}] + begin[${i}]`;
+      }).join(",");
+    }
+    this.userCode = `
+      ${inputDtype} begin = ${inputDtype}(${begin});
+      ${inputDtype} strides = ${inputDtype}(${strides});
 
       void main() {
-        ${s} coords = getOutputCoords();
-        setOutput(getX(${i}));
+        ${dtype} coords = getOutputCoords();
+        setOutput(getX(${newCoords}));
       }
-    `}};function Jre(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a,{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=de({inputs:{x:r},backend:n,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let C=jt.computeOutShape(b,x,w),E=Gu({inputs:{x:r},backend:n,attrs:{begin:b,size:C}});I=de({inputs:{x:E},backend:n,attrs:{shape:m}}),n.disposeIntermediateTensorInfo(E)}else if(n.shouldExecuteOnCPU([r])){let C=n.readSync(r.dataId),E=Pe(r.shape,r.dtype,C),A=S7(h,E,w,b);I=n.makeTensorInfo(m,r.dtype,A.values)}else{let C=new Zre(b,w,h);I=n.runWebGLProgram(C,[r],r.dtype)}let T=de({inputs:{x:I},backend:n,attrs:{shape:m}});return n.disposeIntermediateTensorInfo(I),T}var Qre={kernelName:xu,backendName:"webgl",kernelFunc:Jre};function ese(e){let{inputs:t,backend:n,attrs:a}=e,{separator:r,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=a,{data:p,dataSplits:d}=t,c=n.readSync(p.dataId),h=n.readSync(d.dataId),[m,f]=T7(c,h,r,s,i,o,l,u);return[n.makeTensorInfo([m.length],"string",m),n.makeTensorInfo(d.shape,"int32",f)]}var tse={kernelName:bc,backendName:"webgl",kernelFunc:ese};function nse(e){let{inputs:t,backend:n,attrs:a}=e,{skipEmpty:r}=a,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.readSync(s.dataId),l=n.readSync(i.dataId)[0],[u,p,d]=N7(o,l,r),c=p.length;return[n.makeTensorInfo([c,2],"int32",u),n.makeTensorInfo([c],"string",p),n.makeTensorInfo([2],"int32",new Int32Array(d))]}var ase={kernelName:xc,backendName:"webgl",kernelFunc:nse};function rse(e){let{inputs:t,backend:n,attrs:a}=e,{numBuckets:r}=a,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(r<=0)throw new Error("Number of buckets must be at least 1");let i=n.readSync(s.dataId),o=C7(i,r);return n.makeTensorInfo(s.shape,"int32",o)}var sse={kernelName:vc,backendName:"webgl",kernelFunc:rse},ise="return tan(x);",ose=Ye({opSnippet:ise}),lse={kernelName:uo,backendName:"webgl",kernelFunc:ose},use=`
+    `;
+  }
+};
+function stridedSlice3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeIntermediateTensorInfo(sliced);
+  } else {
+    const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);
+    if (shouldExecuteOnCPU) {
+      const values = backend2.readSync(x.dataId);
+      const xBuf = buffer(x.shape, x.dtype, values);
+      const resultValues = stridedSliceImplCPU(finalShapeSparse, xBuf, $strides, $begin);
+      result = backend2.makeTensorInfo(finalShape, x.dtype, resultValues.values);
+    } else {
+      const program = new StridedSliceProgram($begin, $strides, finalShapeSparse);
+      result = backend2.runWebGLProgram(program, [x], x.dtype);
+    }
+  }
+  const resultReshaped = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: finalShape } });
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var stridedSliceConfig2 = {
+  kernelName: StridedSlice,
+  backendName: "webgl",
+  kernelFunc: stridedSlice3
+};
+function stringNGrams3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const { data, dataSplits } = inputs;
+  const $data = backend2.readSync(data.dataId);
+  const $dataSplits = backend2.readSync(dataSplits.dataId);
+  const [nGrams, nGramsSplits] = stringNGramsImplCPU($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  return [
+    backend2.makeTensorInfo([nGrams.length], "string", nGrams),
+    backend2.makeTensorInfo(dataSplits.shape, "int32", nGramsSplits)
+  ];
+}
+var stringNGramsConfig2 = {
+  kernelName: StringNGrams,
+  backendName: "webgl",
+  kernelFunc: stringNGrams3
+};
+function stringSplit3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { skipEmpty } = attrs;
+  const { input: input2, delimiter } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (input2.shape.length !== 1) {
+    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);
+  }
+  if (delimiter.shape.length !== 0) {
+    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);
+  }
+  const $input = backend2.readSync(input2.dataId);
+  const $delimiter = backend2.readSync(delimiter.dataId)[0];
+  const [indices, values, shape] = stringSplitImplCPU($input, $delimiter, skipEmpty);
+  const outputSize = values.length;
+  return [
+    backend2.makeTensorInfo([outputSize, 2], "int32", indices),
+    backend2.makeTensorInfo([outputSize], "string", values),
+    backend2.makeTensorInfo([2], "int32", new Int32Array(shape))
+  ];
+}
+var stringSplitConfig2 = {
+  kernelName: StringSplit,
+  backendName: "webgl",
+  kernelFunc: stringSplit3
+};
+function stringToHashBucketFast3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { numBuckets } = attrs;
+  const { input: input2 } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const $input = backend2.readSync(input2.dataId);
+  const output = stringToHashBucketFastImplCPU($input, numBuckets);
+  return backend2.makeTensorInfo(input2.shape, "int32", output);
+}
+var stringToHashBucketFastConfig2 = {
+  kernelName: StringToHashBucketFast,
+  backendName: "webgl",
+  kernelFunc: stringToHashBucketFast3
+};
+var TAN = `return tan(x);`;
+var tan3 = unaryKernelFunc2({ opSnippet: TAN });
+var tanConfig2 = {
+  kernelName: Tan,
+  backendName: "webgl",
+  kernelFunc: tan3
+};
+var TANH = `
   float e2x = exp(-2.0 * abs(x));
   return sign(x) * (1.0 - e2x) / (1.0 + e2x);
-`,pse=Ye({opSnippet:use}),cse={kernelName:po,backendName:"webgl",kernelFunc:pse},dse=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let s=0;s<n.length;s++)n[s]=e[s]*t[s];this.outputShape=n,this.rank=n.length;let a=gt(this.rank),r=hse(e);this.userCode=`
+`;
+var tanh4 = unaryKernelFunc2({ opSnippet: TANH });
+var tanhConfig2 = {
+  kernelName: Tanh,
+  backendName: "webgl",
+  kernelFunc: tanh4
+};
+var TileProgram = class {
+  constructor(aShape, reps) {
+    this.variableNames = ["A"];
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[i] * reps[i];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const sourceCoords = getSourceCoords3(aShape);
+    this.userCode = `
       void main() {
-        ${a} resRC = getOutputCoords();
-        setOutput(getA(${r}));
+        ${dtype} resRC = getOutputCoords();
+        setOutput(getA(${sourceCoords}));
       }
-    `}};function hse(e){let t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${e[0]})`;let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],a=[];for(let r=0;r<e.length;r++)a.push(`imod(${n[r]}, ${e[r]})`);return a.join()}function bE(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reps:s}=a;if(r.dtype==="string"||r.shape.length>5){let o=n.readSync(r.dataId),l=r.dtype==="string"?o.map(d=>v.decodeString(d)):o,u=Pe(r.shape,r.dtype,l),p=E7(u,s);return n.makeTensorInfo(p.shape,p.dtype,p.values)}let i=new dse(r.shape,s);return n.runWebGLProgram(i,[r],r.dtype)}var mse={kernelName:hs,backendName:"webgl",kernelFunc:bE},fse=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode=`
+    `;
+  }
+};
+function getSourceCoords3(aShape) {
+  const rank = aShape.length;
+  if (rank > 5) {
+    throw Error(`Tile for rank ${rank} is not yet supported`);
+  }
+  if (rank === 1) {
+    return `imod(resRC, ${aShape[0]})`;
+  }
+  const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u"];
+  const sourceCoords = [];
+  for (let i = 0; i < aShape.length; i++) {
+    sourceCoords.push(`imod(${currentCoords[i]}, ${aShape[i]})`);
+  }
+  return sourceCoords.join();
+}
+function tile4(params) {
+  const { inputs, backend: backend2, attrs } = params;
+  const { x } = inputs;
+  const { reps } = attrs;
+  if (x.dtype === "string" || x.shape.length > 5) {
+    const data = backend2.readSync(x.dataId);
+    const value = x.dtype === "string" ? data.map((d) => util_exports.decodeString(d)) : data;
+    const buf = buffer(x.shape, x.dtype, value);
+    const outBuf = tileImplCPU(buf, reps);
+    return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+  }
+  const program = new TileProgram(x.shape, reps);
+  const output = backend2.runWebGLProgram(program, [x], x.dtype);
+  return output;
+}
+var tileConfig2 = {
+  kernelName: Tile,
+  backendName: "webgl",
+  kernelFunc: tile4
+};
+var SwapProgram = class {
+  constructor(shape) {
+    this.variableNames = ["x", "indices"];
+    this.customUniforms = [
+      { name: "n", type: "int" },
+      { name: "firstPass", type: "int" },
+      { name: "negativeInf", type: "float" },
+      { name: "dir", type: "int" },
+      { name: "inc", type: "int" }
+    ];
+    this.outputShape = shape;
+    this.userCode = `
        void main() {
          ivec2 coords = getOutputCoords();
          int batch = coords[0];
@@ -4624,7 +61112,19 @@ return a / b;`,Ane=`
             setOutput(float(i1));
          }
        }
-     `}},gse=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode=`
+     `;
+  }
+};
+var MergeProgram = class {
+  constructor(shape) {
+    this.variableNames = ["x", "indices"];
+    this.customUniforms = [
+      { name: "n", type: "int" },
+      { name: "firstPass", type: "int" },
+      { name: "k", type: "int" }
+    ];
+    this.outputShape = shape;
+    this.userCode = `
     void main() {
          // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...
          ivec2 coords = getOutputCoords();
@@ -4658,10 +61158,139 @@ return a / b;`,Ane=`
 
          setOutput(x0 >= x1 ? float(i0) : float(i1));
        }
-     `}};function Os(e,t){t!==null&&e.disposeIntermediateTensorInfo(t)}function mI(e){let t=1;for(;t<e;)t*=2;return t}function yse(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{k:s,sorted:i}=a,o=H().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),l=H().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),u=r.shape,p=u[u.length-1];if(n.shouldExecuteOnCPU([r])||p<o||s>l){let A=n.readSync(r.dataId),[R,F]=A7(A,u,r.dtype,s,i);return[n.makeTensorInfo(R.shape,R.dtype,R.values),n.makeTensorInfo(F.shape,F.dtype,F.values)]}if(s===0)return u[u.length-1]=0,[n.makeTensorInfo(u,r.dtype,[]),n.makeTensorInfo(u,"int32",[])];if(p===1)return[r,sd({attrs:{shape:u,dtype:"int32",value:0},backend:n})];let d=n.texData.get(r.dataId),c=d!==null&&d.isPacked,h=c?n.unpackTensor(r):r,m=v.sizeFromShape(u)/p,f=de({inputs:{x:h},attrs:{shape:[m,p]},backend:n});c&&Os(n,h);let g=mI(s),y=mI(p),b=null,x=()=>b===null?[f,f]:[f,b],w=(A,R,F)=>{let S=x(),M=new fse(F),B=[[p],[b===null?1:0],[Number.NEGATIVE_INFINITY],[A],[R]],U=b;b=n.runWebGLProgram(M,S,"int32",B),Os(n,U)};for(let A=1;A<g;A*=2){let R=A*2;for(let F=A;F>=1;F/=2)w(R,F,[m,y])}for(let A=y;A>g;A/=2){let R=x(),F=new gse([m,A/2]),S=[[p],[b===null?1:0],[g]],M=b;b=n.runWebGLProgram(F,R,"int32",S),Os(n,M);let B=g/2,U=B*2;for(let G=B;G>=1;G/=2)w(U,G,b.shape)}let I=b;b=Gu({inputs:{x:b},backend:n,attrs:{begin:0,size:[m,s]}}),Os(n,I);let T=uE({inputs:{x:f,indices:b},backend:n,attrs:{axis:1,batchDims:1}});Os(n,f);let C=u.slice(0,-1);C.push(s),I=b,b=de({inputs:{x:b},attrs:{shape:C},backend:n}),Os(n,I);let E=T;return T=de({inputs:{x:T},attrs:{shape:C},backend:n}),Os(n,E),[T,b]}var bse={kernelName:vu,backendName:"webgl",kernelFunc:yse},xse=class{constructor(e,t,n,a,r,s){this.variableNames=["Image","Transforms"],this.outputShape=s;let i=n==="nearest"?1:2,o;switch(a){case"constant":o=1;break;case"reflect":o=2;break;case"wrap":o=3;break;case"nearest":o=4;break;default:o=1;break}this.userCode=`
+     `;
+  }
+};
+function disposeIntermediateTensorInfoOrNull(backend2, tensorInfo) {
+  if (tensorInfo !== null) {
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+}
+function roundUpToPow2(num) {
+  let pow22 = 1;
+  while (pow22 < num) {
+    pow22 *= 2;
+  }
+  return pow22;
+}
+function topK2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  const TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD");
+  const TOPK_K_CPU_HANDOFF_THRESHOLD = env().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD");
+  const xShape = x.shape;
+  const lastDim = xShape[xShape.length - 1];
+  if (backend2.shouldExecuteOnCPU([x]) || lastDim < TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD || k > TOPK_K_CPU_HANDOFF_THRESHOLD) {
+    const xVals = backend2.readSync(x.dataId);
+    const [allTopKVals, allTopKIndices] = topKImplCPU(xVals, xShape, x.dtype, k, sorted);
+    return [
+      backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),
+      backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)
+    ];
+  }
+  if (k === 0) {
+    xShape[xShape.length - 1] = 0;
+    return [
+      backend2.makeTensorInfo(xShape, x.dtype, []),
+      backend2.makeTensorInfo(xShape, "int32", [])
+    ];
+  }
+  if (lastDim === 1) {
+    return [
+      x,
+      fill3({ attrs: { shape: xShape, dtype: "int32", value: 0 }, backend: backend2 })
+    ];
+  }
+  const xtexData = backend2.texData.get(x.dataId);
+  const xIsPacked = xtexData !== null && xtexData.isPacked;
+  const xUnPacked = xIsPacked ? backend2.unpackTensor(x) : x;
+  const xSize = util_exports.sizeFromShape(xShape);
+  const batch = xSize / lastDim;
+  const x2D = reshape4({ inputs: { x: xUnPacked }, attrs: { shape: [batch, lastDim] }, backend: backend2 });
+  if (xIsPacked) {
+    disposeIntermediateTensorInfoOrNull(backend2, xUnPacked);
+  }
+  const kPow2 = roundUpToPow2(k);
+  const lastDimPow2 = roundUpToPow2(lastDim);
+  let indices = null;
+  const getInputs = () => indices === null ? [x2D, x2D] : [x2D, indices];
+  const runSwap = (dir, inc, shape) => {
+    const inputs2 = getInputs();
+    const program = new SwapProgram(shape);
+    const fistPass = indices === null ? 1 : 0;
+    const customValues = [[lastDim], [fistPass], [Number.NEGATIVE_INFINITY], [dir], [inc]];
+    const prevIndices2 = indices;
+    indices = backend2.runWebGLProgram(program, inputs2, "int32", customValues);
+    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);
+  };
+  for (let len = 1; len < kPow2; len *= 2) {
+    const dir = len * 2;
+    for (let inc = len; inc >= 1; inc /= 2) {
+      runSwap(dir, inc, [batch, lastDimPow2]);
+    }
+  }
+  for (let indicesSize = lastDimPow2; indicesSize > kPow2; indicesSize /= 2) {
+    const inputs2 = getInputs();
+    const mergeProgram = new MergeProgram([batch, indicesSize / 2]);
+    const firstPass = indices === null ? 1 : 0;
+    const customValues = [[lastDim], [firstPass], [kPow2]];
+    const prevIndices2 = indices;
+    indices = backend2.runWebGLProgram(mergeProgram, inputs2, "int32", customValues);
+    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);
+    const len = kPow2 / 2;
+    const dir = len * 2;
+    for (let inc = len; inc >= 1; inc /= 2) {
+      runSwap(dir, inc, indices.shape);
+    }
+  }
+  let prevIndices = indices;
+  indices = slice3({ inputs: { x: indices }, backend: backend2, attrs: { begin: 0, size: [batch, k] } });
+  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);
+  let values = gatherV22({ inputs: { x: x2D, indices }, backend: backend2, attrs: { axis: 1, batchDims: 1 } });
+  disposeIntermediateTensorInfoOrNull(backend2, x2D);
+  const newShape = xShape.slice(0, -1);
+  newShape.push(k);
+  prevIndices = indices;
+  indices = reshape4({ inputs: { x: indices }, attrs: { shape: newShape }, backend: backend2 });
+  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);
+  const prevValues = values;
+  values = reshape4({ inputs: { x: values }, attrs: { shape: newShape }, backend: backend2 });
+  disposeIntermediateTensorInfoOrNull(backend2, prevValues);
+  return [values, indices];
+}
+var topKConfig2 = {
+  kernelName: TopK,
+  backendName: "webgl",
+  kernelFunc: topK2
+};
+var TransformProgram = class {
+  constructor(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape) {
+    this.variableNames = ["Image", "Transforms"];
+    this.outputShape = outShape;
+    const interpolationModeId = interpolation === "nearest" ? 1 : 2;
+    let fillModeId;
+    switch (fillMode) {
+      case "constant":
+        fillModeId = 1;
+        break;
+      case "reflect":
+        fillModeId = 2;
+        break;
+      case "wrap":
+        fillModeId = 3;
+        break;
+      case "nearest":
+        fillModeId = 4;
+        break;
+      default:
+        fillModeId = 1;
+        break;
+    }
+    this.userCode = `
             float mapCoord(float outCoord, float len) {
               float inCoord = outCoord;
-              if(${o} == 2) {
+              if(${fillModeId} == 2) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4685,7 +61314,7 @@ return a / b;`,Ane=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${o} == 3) {
+              } else if (${fillModeId} == 3) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4702,7 +61331,7 @@ return a / b;`,Ane=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${o} == 4) {
+              } else if (${fillModeId} == 4) {
                 return clamp(outCoord, 0.0, len - 1.0);
               } else {
                 return outCoord;
@@ -4712,10 +61341,10 @@ return a / b;`,Ane=`
             float readWithFillValue(int batch, int coordY, int coordX,
               int channel) {
               float outputValue;
-              if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {
+              if (0 <= coordY && coordY < ${imageHeight} && 0 <= coordX && coordX < ${imageWidth}) {
                   outputValue = getImage(batch, coordY, coordX, channel);
               } else {
-                outputValue = float(${r});
+                outputValue = float(${fillValue});
               }
               return outputValue;
             }
@@ -4739,14 +61368,14 @@ return a / b;`,Ane=`
               float c2 = getTransforms(batch, 7);
               float projection = c1 * xf + c2 * yf + 1.0;
               if (projection == 0.0) {
-                outputValue = float(${r});
+                outputValue = float(${fillValue});
               } else {
                 float inX = (a1 * xf + a2 * yf + a3) / projection;
                 float inY = (b1 * xf + b2 * yf + b3) / projection;
-                float mapX = mapCoord(inX, float(${t}));
-                float mapY = mapCoord(inY, float(${e}));
+                float mapX = mapCoord(inX, float(${imageWidth}));
+                float mapY = mapCoord(inY, float(${imageHeight}));
 
-                if (${i} == 1) {
+                if (${interpolationModeId} == 1) {
                   int coordY = int(round(mapY));
                   int coordX = int(round(mapX));
                   outputValue = readWithFillValue(batch, coordY, coordX,
@@ -4770,26 +61399,126 @@ return a / b;`,Ane=`
               }
               setOutput(outputValue);
             }
-        `}};function vse(e){let{inputs:t,backend:n,attrs:a}=e,{image:r,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=a,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=new xse(d,c,i,o,l,g);return n.runWebGLProgram(y,[r,s],"float32")}var wse={kernelName:wu,backendName:"webgl",kernelFunc:vse};function kse(e){let{inputs:t,attrs:n,backend:a}=e,{axis:r}=n,{x:s}=t;Lu(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let i=a.readSync(s.dataId),{outputValues:o,outputShape:l,indices:u}=$7(i,r,s.shape,s.dtype);return[a.makeTensorInfo(l,s.dtype,o),a.makeTensorInfo([u.length],"int32",u)]}var Ise={kernelName:Nm,backendName:"webgl",kernelFunc:kse};function Sse(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r,o=i.shape.length,l=r.shape[s],u=new Array(o-1),p=0;for(let f=0;f<o;f++)f!==s&&(u[p++]=i.shape[f]);let d=[],c=new Array(o).fill(0),h=i.shape.slice();h[s]=1;let m=new Array(l);for(let f=0;f<m.length;f++){c[s]=f;let g=Gu({inputs:{x:i},backend:n,attrs:{begin:c,size:h}}),y=de({inputs:{x:g},backend:n,attrs:{shape:u}});m[f]=y,d.push(g)}return d.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var Tse={kernelName:ku,backendName:"webgl",kernelFunc:Sse},Nse=class{constructor(e,t){this.variableNames=["x","segmentIds"];let n=e.windowSize,a=e.batchSize,r=e.inSize,s=e.numSegments,i=s*Math.ceil(r/n);this.outputShape=[a,i];let o="0.0",l="sumValue",u=Math.floor(n/4)*4,p=n%4,d=`
+        `;
+  }
+};
+function transform3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [
+    batch,
+    outHeight,
+    outWidth,
+    numChannels
+  ];
+  const program = new TransformProgram(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape);
+  return backend2.runWebGLProgram(program, [image2, transforms], "float32");
+}
+var transformConfig2 = {
+  kernelName: Transform,
+  backendName: "webgl",
+  kernelFunc: transform3
+};
+function unique4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  assertNotComplex2(x, "unique");
+  console.warn("WARNING: ", "UI might be locked temporarily as data is being downloaded");
+  const values = backend2.readSync(x.dataId);
+  const { outputValues, outputShape, indices } = uniqueImplCPU(values, axis, x.shape, x.dtype);
+  return [
+    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),
+    backend2.makeTensorInfo([indices.length], "int32", indices)
+  ];
+}
+var uniqueConfig2 = {
+  kernelName: Unique,
+  backendName: "webgl",
+  kernelFunc: unique4
+};
+function unpack2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const x = value;
+  const xRank = x.shape.length;
+  const num = value.shape[axis];
+  const outShape = new Array(xRank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < xRank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = x.shape[i];
+    }
+  }
+  const toDispose = [];
+  const begin = new Array(xRank).fill(0);
+  const size = x.shape.slice();
+  size[axis] = 1;
+  const res = new Array(num);
+  for (let i = 0; i < res.length; i++) {
+    begin[axis] = i;
+    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size } });
+    const reshaped = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: outShape } });
+    res[i] = reshaped;
+    toDispose.push(sliced);
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return res;
+}
+var unpackConfig2 = {
+  kernelName: Unpack,
+  backendName: "webgl",
+  kernelFunc: unpack2
+};
+var SegmentOpProgram = class {
+  constructor(segOpInfo, segOpType) {
+    this.variableNames = ["x", "segmentIds"];
+    const windowSize = segOpInfo.windowSize;
+    const batchSize = segOpInfo.batchSize;
+    const inSize = segOpInfo.inSize;
+    const numSegments = segOpInfo.numSegments;
+    const outSize = numSegments * Math.ceil(inSize / windowSize);
+    this.outputShape = [batchSize, outSize];
+    const initializationValue = "0.0";
+    const returnValue = `sumValue`;
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    const updateSnippet = `
         sumValue += dot(values, segFilter);
-    `,c="";r%n>0&&(c=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+    `;
+    let checkValueOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkValueOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return initializationValue;
         }
-      `);let h="";r%n>0&&(h=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+      `;
+    }
+    let checkSegmentIdOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkSegmentIdOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return -1.0;
         }
-      `),this.userCode=`
-      const float initializationValue = ${o};
+      `;
+    }
+    this.userCode = `
+      const float initializationValue = ${initializationValue};
 
       float getValue(int batch, int inIdx) {
-        ${c}
+        ${checkValueOutOfBounds}
         return getX(batch, inIdx);
       }
 
       float getSegmentIdAtIndex(int inIdx) {
-        ${h}
+        ${checkSegmentIdOutOfBounds}
         return getSegmentIds(inIdx);
       }
 
@@ -4798,12 +61527,12 @@ return a / b;`,Ane=`
         int batch = coords[0];
         int outIdx = coords[1];
         int inOffset = int(floor(float(outIdx) / float(
-          ${s})) * float(${n}));
-        int currentSeg = int(mod(float(outIdx), float(${s})));
+          ${numSegments})) * float(${windowSize}));
+        int currentSeg = int(mod(float(outIdx), float(${numSegments})));
 
         float sumValue = 0.0;
 
-        for (int i = 0; i < ${u}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
           vec4 values = vec4(
             getValue(batch, inIdx),
@@ -4819,11 +61548,11 @@ return a / b;`,Ane=`
             int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0
           );
 
-          ${d}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${u};
-        if (${p===1}) {
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             initializationValue,
@@ -4840,8 +61569,8 @@ return a / b;`,Ane=`
             0
           );
 
-          ${d}
-        } else if (${p===2}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4856,8 +61585,8 @@ return a / b;`,Ane=`
               0
           );
 
-          ${d}
-        } else if (${p===3}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4872,11 +61601,8430 @@ return a / b;`,Ane=`
             0
           );
 
-          ${d}
+          ${updateSnippet}
         }
-        setOutput(${l});
+        setOutput(${returnValue});
       }
-    `}};function Cse(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,segmentIds:s}=t,{numSegments:i}=a,o=r.shape.length,l=[],u=0,p=N.getAxesPermutation([u],o),d=r;p!=null&&(d=In({inputs:{x:r},backend:n,attrs:{perm:p}}),l.push(d),u=N.getInnerMostAxes(1,o)[0]);let c=N.segment_util.computeOutShape(d.shape,u,i),h=v.sizeFromShape([d.shape[u]]),m=de({inputs:{x:d},backend:n,attrs:{shape:[-1,h]}});l.push(m);let f=_m(r.dtype),g=(w,I,T,C,E)=>{let A=w.shape[0],R=w.shape[1],F=N.segment_util.segOpComputeOptimalWindowSize(R,E),S={windowSize:F,inSize:R,batchSize:A,numSegments:E},M=new Nse(S,I),B=n.compileAndRun(M,[w,T],C);if(l.push(B),B.shape[1]===E)return B;let U=gE({backend:n,attrs:{start:0,stop:E,step:1,dtype:"float32"}}),G=bE({inputs:{x:U},backend:n,attrs:{reps:[R/F]}});return l.push(U),l.push(G),g(B,I,G,C,E)},y=g(m,"unsortedSegmentSum",s,f,i),b=de({inputs:{x:y},backend:n,attrs:{shape:c}}),x=b;if(p!=null){l.push(b);let w=N.getUndoAxesPermutation(p);x=In({inputs:{x},backend:n,attrs:{perm:w}})}return l.forEach(w=>n.disposeIntermediateTensorInfo(w)),x}var _se={kernelName:wc,backendName:"webgl",kernelFunc:Cse},Ese=[SJ,NJ,EJ,FJ,RJ,OJ,zJ,BJ,HJ,qJ,YJ,QJ,n9,i9,u9,c9,h9,y9,x9,w9,T9,F9,R9,P9,V9,G9,K9,oJ,Z9,nQ,iQ,dQ,mQ,gQ,bQ,vQ,IQ,NQ,EQ,$Q,DQ,MQ,LQ,WQ,GQ,jQ,XQ,JQ,eee,ree,lee,dee,fee,bee,xee,wee,Iee,Tee,Cee,Eee,Dee,Pee,zee,Bee,Gee,qee,Zee,tte,iJ,ate,eQ,ite,ute,dte,uJ,gte,vte,kte,Nte,Ete,Dte,Pte,Wte,Gte,qte,Xte,Qte,tne,ane,one,une,cne,hne,fne,xne,Ine,Cne,Mne,dJ,zne,Vne,Hne,Kne,L9,Zne,Qne,tae,rae,lae,cJ,pae,dae,mae,gae,yae,z9,$ne,vae,Sae,_ae,mJ,Fae,Mae,zae,Vae,jae,Kae,Zae,ere,nre,sre,lre,cre,fre,bre,wre,Sre,A9,Dne,Cre,Ere,$re,Dre,Mre,Ore,zre,Bre,Ure,jre,Kre,Yre,Qre,tse,ase,sse,Fne,wJ,lse,cse,mse,bse,wse,kJ,Ise,Tse,_se,Jne];for(let e of Ese)kc(e);var Et;(function(e){e[e.float32=0]="float32",e[e.int32=1]="int32",e[e.bool=2]="bool",e[e.string=3]="string",e[e.complex64=4]="complex64"})(Et||(Et={}));var ac;(function(e){e[e.linear=0]="linear",e[e.relu=1]="relu",e[e.relu6=2]="relu6",e[e.prelu=3]="prelu",e[e.leakyrelu=4]="leakyrelu",e[e.sigmoid=5]="sigmoid",e[e.elu=6]="elu"})(ac||(ac={}));var xE;function Ase(e){xE=e.wasm.cwrap(Js,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])}function $se(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t;if(r.dtype!=="float32"||s.dtype!=="float32")throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a,c=n.dataIdMap.get(r.dataId).id,h=n.dataIdMap.get(s.dataId).id,m=0;if(i!=null){let E=n.dataIdMap.get(i.dataId);if(E.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${E.shape.length}.`);m=E.id}let f=o==null?0:n.dataIdMap.get(o.dataId).id,g=ac[p];if(g==null)throw new Error(`${p} activation not yet supported for FusedConv2D in the wasm backend.`);let y=l?r.shape[2]:r.shape[1],b=u?s.shape[1]:s.shape[2],x=Tu.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)),w=n.makeOutput([...x,y,b],r.dtype),I=n.dataIdMap.get(w.dataId).id,T=new Uint8Array(new Int32Array(r.shape).buffer),C=new Uint8Array(new Int32Array(s.shape).buffer);return xE(c,T,r.shape.length,h,C,s.shape.length,l,u,g,m,f,d||0,I),w}var Fse={kernelName:Js,backendName:"wasm",setupFunc:Ase,kernelFunc:$se};function an(e,t){let n;function a(s){n=s.wasm.cwrap(e,null,["number","number","number"])}function r(s){let{backend:i,inputs:{x:o}}=s,l=i.dataIdMap.get(o.dataId).id,u=i.makeOutput(o.shape,t||o.dtype),p=i.dataIdMap.get(u.dataId).id;return v.sizeFromShape(u.shape)===0||n(l,Et[o.dtype],p),u}return{kernelName:e,backendName:"wasm",setupFunc:a,kernelFunc:r}}var Dse=an(wl);function cn(e,t,n){let a;function r(i){a=i.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])}function s(i){let{backend:o,inputs:l}=i,{a:u,b:p}=l,d=o.dataIdMap.get(u.dataId).id,c=o.dataIdMap.get(p.dataId).id,h=n!=null?n:u.dtype,m=N.assertAndGetBroadcastShape(u.shape,p.shape),f=o.makeOutput(m,h);if(v.sizeFromShape(m)===0)return f;let g=new Uint8Array(new Int32Array(u.shape).buffer),y=new Uint8Array(new Int32Array(p.shape).buffer),b=o.dataIdMap.get(f.dataId).id;return a(d,g,u.shape.length,c,y,p.shape.length,Et[u.dtype],b),f}return{kernelName:e,backendName:"wasm",setupFunc:r,kernelFunc:s}}var Rse=!0,Mse=cn(cs,Rse),vE;function Pse(e){vE=e.wasm.cwrap(mi,null,["array","number","number","number"])}function Ose(e){let{inputs:t,backend:n}=e,a=n.makeOutput(t[0].shape,t[0].dtype);if(v.sizeFromShape(a.shape)===0)return a;let r=t.map(o=>n.dataIdMap.get(o.dataId).id),s=new Uint8Array(new Int32Array(r).buffer),i=n.dataIdMap.get(a.dataId).id;return vE(s,r.length,Et[a.dtype],i),a}var Lse={kernelName:mi,backendName:"wasm",setupFunc:Pse,kernelFunc:Ose};function Lf(e){let{inputs:{x:t},backend:n}=e;if(t.dtype==="string")return kn(n.readSync(t.dataId),t.shape,t.dtype);let a=n.makeOutput(t.shape,t.dtype),r=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(a).set(r),a}var zse={kernelName:Di,backendName:"wasm",kernelFunc:Lf},wE;function Wse(e){wE=e.wasm.cwrap(Tr,null,["number","array","number","number","number","array","number"])}function ls(e){let{inputs:t,backend:n,attrs:a}=e,[r,s]=Vse(t.x.shape,a.perm),i=!0;for(let m=0;m<s.length;m++)s[m]!==m&&(i=!1);let o=Bse(t.x.shape,a.perm),l={dataId:t.x.dataId,shape:r,dtype:t.x.dtype};if(i){let m=Lf({inputs:t,backend:n});return m.shape=o,m}let u=n.makeOutput(o,l.dtype),p=n.dataIdMap.get(l.dataId).id,d=n.dataIdMap.get(u.dataId).id,c=new Uint8Array(new Int32Array(s).buffer),h=new Uint8Array(new Int32Array(l.shape).buffer);return wE(p,h,l.shape.length,Et[l.dtype],d,c,s.length),u}function Bse(e,t){let n=new Array(e.length);for(let a=0;a<n.length;a++)n[a]=e[t[a]];return n}function Vse(e,t){let n=[],a=[];for(let r=0;r<e.length;++r)e[r]!==1&&n.push(e[r]),e[t[r]]!==1&&a.push(t[r]);for(let r=0;r<a.length;++r){let s=-1;for(let i=0;i<a.length;++i)a[i]>=r&&(s===-1||a[s]>a[i])&&(s=i);a[s]=r}return[n,a]}var Use={kernelName:Tr,backendName:"wasm",kernelFunc:ls,setupFunc:Wse};function Ss(e,t,n){let a=e.shape,r=e.shape.length,s=v.parseAxisParam(t,a),i=s,o=N.getAxesPermutation(i,r),l=null,u=!1;if(o!=null){let p=new Array(r);for(let c=0;c<p.length;c++)p[c]=a[o[c]];i=N.getInnerMostAxes(i.length,r),l=ls({inputs:{x:e},attrs:{perm:o},backend:n});let d=n.dataIdMap.get(e.dataId).id;n.dataIdMap.get(l.dataId).id!==d&&(u=!0)}return{transposed:l,originalAxes:s,axes:i,inputWasTransposed:u}}var kE;function Gse(e){kE=e.wasm.cwrap(Sl,null,["number, number, number"])}function Hse(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Ss(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("all",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;kE(o,g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var jse={kernelName:Sl,backendName:"wasm",setupFunc:Gse,kernelFunc:Hse},IE;function qse(e){IE=e.wasm.cwrap(Tl,null,["number, number, number"])}function Kse(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Ss(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("any",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;IE(o,g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var Xse={kernelName:Tl,backendName:"wasm",setupFunc:qse,kernelFunc:Kse},SE;function Yse(e){SE=e.wasm.cwrap(fi,null,["number","number","number","number","number"])}function Zse(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r}=a,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=i,l=s,{transposed:u,axes:p,inputWasTransposed:d}=Ss(s,r,t);if(d){let y=t.dataIdMap.get(u.dataId).id;y!==i&&(l=u,o=y)}let c=l.shape.slice(0,-1),h=t.makeOutput(c,"int32"),m=t.dataIdMap.get(h.dataId).id,f=v.sizeFromShape(h.shape),g=l.shape[p[0]];return SE(o,Et[l.dtype],f,g,m),d&&t.disposeData(u.dataId),h}var Jse={kernelName:fi,backendName:"wasm",kernelFunc:Zse,setupFunc:Yse},TE;function Qse(e){TE=e.wasm.cwrap(gi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function eie(e){let{inputs:t,attrs:n,backend:a}=e,r=t.x,s=a.dataIdMap.get(r.dataId).id,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,p=N.computePool2DInfo(r.shape,i,o,1,l,u),d=p.filterHeight,c=p.filterWidth,h=p.padInfo.top,m=p.padInfo.right,f=p.padInfo.bottom,g=p.padInfo.left,y=p.strideHeight,b=p.strideWidth,x=p.inChannels;if(p.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${p.dataFormat}'. Please use 'channelsLast'.`);if(p.dilationWidth!==1||p.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${p.dilationHeight}, ${p.dilationWidth}].`);let w=a.makeOutput(p.outShape,"float32"),I=a.dataIdMap.get(w.dataId).id;return TE(s,r.shape[0],r.shape[1],r.shape[2],d,c,h,m,f,g,y,b,x,I),w}var tie={kernelName:gi,backendName:"wasm",setupFunc:Qse,kernelFunc:eie};function Wn(e){let{inputs:t,attrs:n}=e,{x:a}=t,{shape:r}=n,s=v.sizeFromShape(a.shape),i=v.inferFromImplicitShape(r,s);return v.assert(s===v.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${a.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(a.dataId),{dataId:a.dataId,shape:i,dtype:a.dtype}}var nie={kernelName:lu,backendName:"wasm",kernelFunc:Wn},NE;function aie(e){NE=e.wasm.cwrap(yi,null,["number","array","number","number","array","number","number","number","number"])}function rie(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;if(r.dtype!=="float32"||s.dtype!=="float32")throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let l=r.shape.length,u=s.shape.length,p=i?r.shape[l-2]:r.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],c=i?r.shape[l-1]:r.shape[l-2],h=o?s.shape[u-2]:s.shape[u-1],m=r.shape.slice(0,-2),f=s.shape.slice(0,-2),g=v.sizeFromShape(m),y=v.sizeFromShape(f),b=Tu.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)).concat([c,h]);v.assert(p===d,()=>`Error in matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${r.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,p,c]:[g,c,p],w=o?[y,h,d]:[y,d,h],I=Wn({inputs:{x:r},backend:n,attrs:{shape:x}}),T=Wn({inputs:{x:s},backend:n,attrs:{shape:w}}),C=n.dataIdMap.get(I.dataId).id,E=n.dataIdMap.get(T.dataId).id,A=i?I.shape[2]:I.shape[1],R=o?T.shape[1]:T.shape[2],F=Math.max(g,y),S=n.makeOutput([F,A,R],I.dtype),M=n.dataIdMap.get(S.dataId).id,B=new Uint8Array(new Int32Array(I.shape).buffer),U=new Uint8Array(new Int32Array(T.shape).buffer);return NE(C,B,I.shape.length,E,U,T.shape.length,i,o,M),n.disposeData(I.dataId),n.disposeData(T.dataId),S.shape=b,S}var sie={kernelName:yi,backendName:"wasm",setupFunc:aie,kernelFunc:rie};function di(e){let{inputs:{x:t},attrs:{begin:n,size:a},backend:r}=e,[s,i]=jt.parseSliceParams(t,n,a),o=jt.isSliceContinous(t.shape,s,i),l=r.readSync(t.dataId),u=r.makeOutput(i,t.dtype),p=v.computeStrides(t.shape),d=r.dataIdMap.get(u.dataId);if(o){let m=jt.computeFlatOffset(s,p);return t.dtype==="string"?d.stringBytes=l.slice(m,m+v.sizeFromShape(i)):r.typedArrayFromHeap(u).set(l.subarray(m,m+v.sizeFromShape(i))),u}if(t.dtype==="string"){let m=Oh(l,s,i,t.shape,t.dtype);return d.stringBytes=m,u}let c=r.typedArrayFromHeap(u),h=t.shape.length;if(h===2)iie(l,p[0],c,s,i);else if(h===3)oie(l,p[0],p[1],c,s,i);else if(h===4)lie(l,p[0],p[1],p[2],c,s,i);else{let m=Oh(l,s,i,t.shape,t.dtype);c.set(m)}return u}function iie(e,t,n,a,r){let s=0,i=a[0],o=a[1],l=i+r[0];for(let u=i;u<l;u++){let p=u*t+o;n.set(e.subarray(p,p+r[1]),s),s+=r[1]}}function oie(e,t,n,a,r,s){let i=0,o=r[0],l=r[1],u=r[2],p=o+s[0],d=l+s[1];for(let c=o;c<p;c++)for(let h=l;h<d;h++){let m=c*t+h*n+u;a.set(e.subarray(m,m+s[2]),i),i+=s[2]}}function lie(e,t,n,a,r,s,i){let o=0,l=s[0],u=s[1],p=s[2],d=l+i[0],c=u+i[1],h=p+i[2],m=s[3];for(let f=l;f<d;f++)for(let g=u;g<c;g++)for(let y=p;y<h;y++){let b=f*t+g*n+y*a+m;r.set(e.subarray(b,b+i[3]),o),o+=i[3]}}var uie={kernelName:du,backendName:"wasm",kernelFunc:di};function pie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a,o=s.reduce((y,b)=>y*b),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=Wn({inputs:{x:r},backend:n,attrs:{shape:l}}),m=ls({inputs:{x:h},backend:n,attrs:{perm:u}}),f=Wn({inputs:{x:m},backend:n,attrs:{shape:p}}),g=di({inputs:{x:f},backend:n,attrs:{begin:d,size:c}});return n.disposeData(h.dataId),n.disposeData(m.dataId),n.disposeData(h.dataId),g}var cie={kernelName:$l,backendName:"wasm",kernelFunc:pie};function Hu(e){let{inputs:{x:t},attrs:{dtype:n},backend:a}=e,r=a.makeOutput(t.shape,n),s=a.typedArrayFromHeap(t);return a.typedArrayFromHeap(r).set(s),r}var die={kernelName:bi,backendName:"wasm",kernelFunc:Hu},hie=an(xi),CE;function mie(e){CE=e.wasm.cwrap(ds,null,["number","number","number","number"])}function fie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{clipValueMin:s,clipValueMax:i}=a,o=n.dataIdMap.get(r.dataId).id,l=n.makeOutput(r.shape,r.dtype),u=n.dataIdMap.get(l.dataId).id;return CE(o,s,i,u),l}var gie={kernelName:ds,backendName:"wasm",setupFunc:mie,kernelFunc:fie};function _E(e){let{inputs:t,backend:n}=e,a=v.parseAxisParam(e.attrs.axis,t[0].shape)[0],r=t.map(h=>h.shape);N.assertParamsConsistent(r,a);let s=N.computeOutShape(t.map(h=>h.shape),a),i=t.filter(h=>v.sizeFromShape(h.shape)>0);if(i.length===1)return Lf({inputs:{x:i[0]},backend:n});let o=n.makeOutput(s,t[0].dtype);if(v.sizeFromShape(s)===0)return o;if(i[0].dtype==="string"){let h=i.map(x=>{let w=v.sizeFromShape(x.shape.slice(a));return Wn({inputs:{x},backend:n,attrs:{shape:[-1,w]}})}),m=h.map(x=>({vals:n.readSync(x.dataId),shape:x.shape}));s=N.computeOutShape(h.map(x=>x.shape),1);let f=h[0].shape[0]===1,g=E0(m,s,t[0].dtype,f),y=N.computeOutShape(i.map(x=>x.shape),a);o.shape=y;let b=n.dataIdMap.get(o.dataId);return b.stringBytes=N.fromStringArrayToUint8(g),h.forEach(x=>n.disposeData(x.dataId)),o}let l=v.sizeFromShape(i[0].shape.slice(0,a)),u=0,p=i.map(h=>{let m=v.sizeFromShape(h.shape.slice(a));return u+=m,m}),d=i.map(h=>n.typedArrayFromHeap(h)),c=n.typedArrayFromHeap(o);for(let h=0;h<l;h++){let m=h*u;for(let f=0;f<d.length;f++){let g=p[f],y=h*g,b=d[f].subarray(y,y+g);c.set(b,m),m+=g}}return o}var yie={kernelName:Fl,backendName:"wasm",kernelFunc:_E},EE;function bie(e){EE=e.wasm.cwrap(vi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function xie(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s}=t,i=a.dataIdMap.get(r.dataId).id,o=a.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:p,dimRoundingMode:d,dataFormat:c}=n,h=N.convertConv2DDataFormat(c),m=N.computeConv2DInfo(r.shape,s.shape,l,u,p,d,!1,h),f=m.filterHeight,g=m.filterWidth,y=m.padInfo.top,b=m.padInfo.right,x=m.padInfo.bottom,w=m.padInfo.left,I=m.dilationHeight,T=m.dilationWidth,C=m.strideHeight,E=m.strideWidth,A=m.inChannels,R=m.outChannels,F=m.padInfo.type==="SAME"?1:0;if(m.dataFormat!=="channelsLast")throw new Error(`wasm backend Conv2D does not support dataFormat:'${m.dataFormat}'. Please use 'channelsLast'.`);let S=a.makeOutput(m.outShape,"float32"),M=a.dataIdMap.get(S.dataId).id;return EE(i,r.shape[0],r.shape[1],r.shape[2],o,f,g,y,b,x,w,F,I,T,C,E,A,R,M),S}var vie={kernelName:vi,backendName:"wasm",setupFunc:bie,kernelFunc:xie},AE;function wie(e){AE=e.wasm.cwrap(wi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function kie(e){let{backend:t,inputs:n,attrs:a}=e,{dy:r,filter:s}=n,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,inputShape:p}=a,d=1,c=N.convertConv2DDataFormat(l),h=N.computeConv2DInfo(p,s.shape,i,d,o,u,!1,c),{batchSize:m,filterHeight:f,filterWidth:g,inChannels:y,inHeight:b,inWidth:x,outChannels:w,outHeight:I,outWidth:T,strideHeight:C,strideWidth:E}=h,A=f-1-h.padInfo.top,R=g-1-h.padInfo.left,F=h.dataFormat==="channelsLast",S=v.computeStrides(h.inShape),M=v.computeStrides(r.shape),[B,U,G]=v.computeStrides(s.shape),q=S[0],K=F?S[1]:S[2],Z=F?S[2]:1,Q=F?1:S[1],ee=M[0],ae=F?M[1]:M[2],te=F?M[2]:1,le=F?1:M[1],ie=t.makeOutput(h.inShape,"float32"),ye=t.dataIdMap.get(ie.dataId).id,ue=t.dataIdMap.get(r.dataId).id,xe=t.dataIdMap.get(s.dataId).id;return AE(ue,xe,m,f,g,b,x,y,I,T,w,C,E,A,R,B,U,G,q,K,Z,Q,ee,ae,te,le,ye),ie}var Iie={kernelName:wi,backendName:"wasm",setupFunc:wie,kernelFunc:kie},Sie=an(ki),Tie=an(Ii),hx;(function(e){e[e.bilinear=0]="bilinear",e[e.nearest=1]="nearest"})(hx||(hx={}));var $E;function Nie(e){$E=e.wasm.cwrap(Rl,null,["number","number","number","number","array","number","number","number","number","number"])}function Cie(e){let{backend:t,inputs:n,attrs:a}=e,{method:r,extrapolationValue:s,cropSize:i}=a,{image:o,boxes:l,boxInd:u}=n,p=l.shape[0],[d,c]=i,h=[p,d,c,o.shape[3]],m=t.dataIdMap.get(o.dataId),f;o.dtype!=="float32"&&(f=Hu({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),m=t.dataIdMap.get(f.dataId));let g=m.id,y=t.dataIdMap.get(l.dataId).id,b=t.dataIdMap.get(u.dataId).id,x=t.makeOutput(h,"float32"),w=t.dataIdMap.get(x.dataId).id,I=new Uint8Array(new Int32Array(o.shape).buffer);return $E(g,y,b,p,I,d,c,hx[r],s,w),f!=null&&t.disposeData(f.dataId),x}var _ie={kernelName:Rl,backendName:"wasm",setupFunc:Nie,kernelFunc:Cie},FE;function Eie(e){FE=e.wasm.cwrap(Dl,null,["number","number","number","number","number","number"])}function Aie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a,l=r.shape.length;v.assert(r.dtype==="float32"||r.dtype==="int32",()=>`cumprod does not support ${r.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([s],l),p=r;u!==null&&(p=ls({inputs:{x:r},attrs:{perm:u},backend:n}));let d=N.getInnerMostAxes(1,l)[0];N.assertAxesAreInnerMostDims("cumprod",[d],l);let c=n.makeOutput(p.shape,p.dtype),h=p.shape[d],m=n.dataIdMap.get(p.dataId).id,f=n.dataIdMap.get(c.dataId).id;FE(m,i?1:0,o?1:0,h,f,Et[r.dtype]);let g=c;if(u!==null){let y=N.getUndoAxesPermutation(u);g=ls({inputs:{x:c},attrs:{perm:y},backend:n}),n.disposeData(p.dataId),n.disposeData(c.dataId)}return g}var $ie={kernelName:Dl,backendName:"wasm",setupFunc:Eie,kernelFunc:Aie},DE;function Fie(e){DE=e.wasm.cwrap(Si,null,["number","number","number","number","number","number"])}function Die(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a,l=r.shape.length;v.assert(r.dtype==="float32"||r.dtype==="int32",()=>`cumsum does not support ${r.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([s],l),p=r;u!==null&&(p=ls({inputs:{x:r},attrs:{perm:u},backend:n}));let d=N.getInnerMostAxes(1,l)[0];N.assertAxesAreInnerMostDims("cumsum",[d],l);let c=n.makeOutput(p.shape,p.dtype),h=p.shape[d],m=n.dataIdMap.get(p.dataId).id,f=n.dataIdMap.get(c.dataId).id;DE(m,i?1:0,o?1:0,h,f,Et[r.dtype]);let g=c;if(u!==null){let y=N.getUndoAxesPermutation(u);g=ls({inputs:{x:c},attrs:{perm:y},backend:n}),n.disposeData(p.dataId),n.disposeData(c.dataId)}return g}var Rie={kernelName:Si,backendName:"wasm",setupFunc:Fie,kernelFunc:Die},RE;function Mie(e){RE=e.wasm.cwrap(Ml,null,["number","number","number","array","number","array","array","number","number"])}function Pie(e){let{backend:t,inputs:n,attrs:a}=e,{x:r}=n,{blockSize:s,dataFormat:i}=a,o=r.shape[0],l=i==="NHWC"?r.shape[1]:r.shape[2],u=i==="NHWC"?r.shape[2]:r.shape[3],p=i==="NHWC"?r.shape[3]:r.shape[1],d=l*s,c=u*s,h=p/(s*s),m=i==="NHWC"?[o,d,c,h]:[o,h,d,c],f=t.makeOutput(m,"float32"),g=t.dataIdMap.get(r.dataId).id,y=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),b=new Uint8Array(new Int32Array(m).buffer),x=new Uint8Array(new Int32Array(v.computeStrides(m)).buffer),w=t.dataIdMap.get(f.dataId).id;return RE(g,s,i==="NHWC"?1:0,y,r.shape.length-1,b,x,m.length,w),f}var Oie={kernelName:Ml,backendName:"wasm",setupFunc:Mie,kernelFunc:Pie},ME;function Lie(e){ME=e.wasm.cwrap(Ti,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function zie(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s}=t,i=a.dataIdMap.get(r.dataId).id,o=a.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:p,dimRoundingMode:d}=n,c=u==null?[1,1]:u,h=N.computeConv2DInfo(r.shape,s.shape,l,c,p,d,!0),m=h.filterHeight,f=h.filterWidth,g=h.padInfo.top,y=h.padInfo.right,b=h.padInfo.bottom,x=h.padInfo.left,w=h.dilationHeight,I=h.dilationWidth,T=h.strideHeight,C=h.strideWidth,E=h.inChannels,A=h.outChannels,R=h.padInfo.type==="SAME"?1:0;if(h.dataFormat!=="channelsLast")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${h.dataFormat}'. Please use 'channelsLast'.`);let F=a.makeOutput(h.outShape,"float32"),S=a.dataIdMap.get(F.dataId).id;return ME(i,r.shape[0],r.shape[1],r.shape[2],o,m,f,g,y,b,x,R,w,I,T,C,E,A,S),F}var Wie={kernelName:Ti,backendName:"wasm",setupFunc:Lie,kernelFunc:zie},Bie=an(Ci),Vie=!1,Uie=cn(Ol,Vie,"bool"),Gie=an(_i,"float32");function mx(e){let{inputs:t,attrs:n,backend:a}=e,{input:r}=t,{dim:s}=n,i=r.shape.length,o=r.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),Wn({inputs:{x:r},backend:a,attrs:{shape:o}})}var Hie={kernelName:Ll,backendName:"wasm",kernelFunc:mx};function PE(e){let{attrs:{shape:t,value:n,dtype:a},backend:r}=e,s=r.makeOutput(t,a);return r.typedArrayFromHeap(s).fill(n),s}var jie={kernelName:pc,backendName:"wasm",kernelFunc:PE},OE;function qie(e){OE=e.wasm.cwrap(Wl,null,["number","number","number","number","number","number"])}function Kie(e){let{inputs:t,backend:n}=e,{image:a}=t,r=n.makeOutput(a.shape,a.dtype),s=n.dataIdMap.get(a.dataId).id,i=n.dataIdMap.get(r.dataId).id,[o,l,u,p]=a.shape;return OE(s,o,l,u,p,i),r}var Xie={kernelName:Wl,backendName:"wasm",kernelFunc:Kie,setupFunc:qie},Yie=an(Ei),Zie=!1,Jie=cn(Ai,Zie),LE;function Qie(e){LE=e.wasm.cwrap($i,null,["number","number","number","number","number","number","number"])}function eoe(e){let{backend:t,inputs:n,attrs:a}=e,{varianceEpsilon:r}=a,{x:s,mean:i,variance:o,offset:l,scale:u}=n,p=t.dataIdMap.get(s.dataId).id,d=t.dataIdMap.get(i.dataId).id,c=t.dataIdMap.get(o.dataId).id,h=l!=null?t.dataIdMap.get(l.dataId).id:0,m=u!=null?t.dataIdMap.get(u.dataId).id:0,f=t.makeOutput(s.shape,s.dtype);if(v.sizeFromShape(s.shape)===0)return f;let g=t.dataIdMap.get(f.dataId).id;return LE(p,d,c,h,m,r,g),f}var toe={kernelName:$i,backendName:"wasm",setupFunc:Qie,kernelFunc:eoe},zE;function noe(e){zE=e.wasm.cwrap(Qs,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function aoe(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dataFormat:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=n,f=N.computeConv2DInfo(r.shape,s.shape,l,p,u,c),g=ac[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedConv2D in the wasm backend.`);let y=a.dataIdMap.get(r.dataId).id,b=a.dataIdMap.get(s.dataId).id,x=f.outChannels,w=0;if(i!=null){let te=a.dataIdMap.get(i.dataId);if(te.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==x)throw new Error(`FusedConv2D bias shape (${te.shape}) does not match the number of output channels (${x})`);w=te.id}let I=f.filterHeight,T=f.filterWidth,C=f.padInfo.top,E=f.padInfo.right,A=f.padInfo.bottom,R=f.padInfo.left,F=f.dilationHeight,S=f.dilationWidth,M=f.strideHeight,B=f.strideWidth,U=f.inChannels,G=f.padInfo.type==="SAME"?1:0,q=f.batchSize,K=f.inHeight,Z=f.inWidth;if(d!=="NHWC")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let Q=a.makeOutput(f.outShape,"float32"),ee=a.dataIdMap.get(Q.dataId).id,ae=o==null?0:a.dataIdMap.get(o.dataId).id;return zE(y,q,K,Z,b,I,T,w,C,E,A,R,G,F,S,M,B,U,x,g,ae,m||0,ee),Q}var roe={kernelName:Qs,backendName:"wasm",setupFunc:noe,kernelFunc:aoe},WE;function soe(e){WE=e.wasm.cwrap(ei,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function ioe(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dataFormat:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=n,f=N.computeConv2DInfo(r.shape,s.shape,l,p,u,c,!0),g=ac[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let y=a.dataIdMap.get(r.dataId).id,b=a.dataIdMap.get(s.dataId).id,x=f.outChannels,w=0;if(i!=null){let te=a.dataIdMap.get(i.dataId);if(te.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==x)throw new Error(`FusedDepthwiseConv2D bias shape (${te.shape}) does not match the number of output channels (${x})`);w=te.id}let I=f.filterHeight,T=f.filterWidth,C=f.padInfo.top,E=f.padInfo.right,A=f.padInfo.bottom,R=f.padInfo.left,F=f.dilationHeight,S=f.dilationWidth,M=f.strideHeight,B=f.strideWidth,U=f.inChannels,G=f.padInfo.type==="SAME"?1:0,q=f.batchSize,K=f.inHeight,Z=f.inWidth;if(d!=="NHWC")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let Q=a.makeOutput(f.outShape,"float32"),ee=a.dataIdMap.get(Q.dataId).id,ae=o==null?0:a.dataIdMap.get(o.dataId).id;return WE(y,q,K,Z,b,I,T,w,C,E,A,R,G,F,S,M,B,U,x,g,ae,m||0,ee),Q}var ooe={kernelName:ei,backendName:"wasm",setupFunc:soe,kernelFunc:ioe},BE;function loe(e){BE=e.wasm.cwrap(Vl,null,["number","number","number","number","number","number","array","number"])}function uoe(e){let{backend:t,inputs:n}=e,{params:a,indices:r}=n,[s,i,o,l]=Mx.prepareAndValidate(a,r),u=t.makeOutput(s,a.dtype);if(i===0)return u;let p=r.shape,d=p[p.length-1],c=t.dataIdMap.get(a.dataId).id,h=t.dataIdMap.get(r.dataId).id,m=new Uint8Array(new Int32Array(l).buffer),f=t.dataIdMap.get(u.dataId).id;return BE(c,Et[a.dtype],h,i,d,o,m,f),u}var poe={kernelName:Vl,backendName:"wasm",setupFunc:loe,kernelFunc:uoe},VE;function coe(e){VE=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])}function doe(e){let{backend:t,inputs:n,attrs:a}=e,{x:r,indices:s}=n,{axis:i,batchDims:o}=a,l=v.parseAxisParam(i,r.shape)[0],u=t.readSync(s.dataId),p=r.shape[l];for(let C=0;C<u.length;++C){let E=u[C];v.assert(E<=p-1&&E>=0,()=>`GatherV2: the index value ${E} is not in [0, ${p-1}]`)}let d=N.segment_util.collectGatherOpShapeInfo(r,s,l,o),c=Wn({inputs:{x:r},attrs:{shape:[d.batchSize,d.outerSize,d.dimSize,d.sliceSize]},backend:t}),h=v.sizeFromShape(s.shape),m=Wn({inputs:{x:s},attrs:{shape:[d.batchSize,h/d.batchSize]},backend:t}),f=[d.batchSize,d.outerSize,h/d.batchSize,d.sliceSize],g=t.makeOutput(f,r.dtype);if(v.sizeFromShape(r.shape)===0)return g;let y=c.shape.length-1,b=t.dataIdMap.get(c.dataId).id,x=t.dataIdMap.get(m.dataId).id,w=t.dataIdMap.get(g.dataId).id,I=new Uint8Array(new Int32Array(v.computeStrides(c.shape)).buffer),T=new Uint8Array(new Int32Array(v.computeStrides(f)).buffer);return VE(b,Et[r.dtype],I,y,x,d.batchSize,T,w),t.disposeData(c.dataId),t.disposeData(m.dataId),g.shape=d.outputShape,g}var hoe={kernelName:Bl,backendName:"wasm",setupFunc:coe,kernelFunc:doe},moe=!1,foe=cn(Ul,moe,"bool"),goe=!1,yoe=cn(Fi,goe,"bool"),UE;function boe(e){UE=e.wasm.cwrap(Ri,null,["number","number","number","number"])}function xoe(e){let{inputs:{x:t},attrs:{alpha:n},backend:a}=e,r=a.dataIdMap.get(t.dataId).id,s=a.makeOutput(t.shape,"float32");if(v.sizeFromShape(t.shape)!==0){let i=a.dataIdMap.get(s.dataId).id;UE(r,Et[t.dtype],n,i)}return s}var voe={kernelName:Ri,backendName:"wasm",setupFunc:boe,kernelFunc:xoe},woe=!1,koe=cn(ql,woe,"bool"),Ioe=!1,Soe=cn(Kl,Ioe,"bool"),Toe=an(Mi),Noe=!1,Coe=cn(Yl,Noe,"bool"),_oe=an(Zl),Eoe=!1,Aoe=cn(Jl,Eoe,"bool"),$oe=!1,Foe=cn(MI,$oe,"bool"),GE;function Doe(e){GE=e.wasm.cwrap(Pi,null,["number","number","number","number"])}function Roe(e){let{backend:t,inputs:n,attrs:a}=e,{reductionIndices:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Ss(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("max",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;GE(o,Et[i.dtype],g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var Moe={kernelName:Pi,backendName:"wasm",setupFunc:Doe,kernelFunc:Roe},Poe=!1,Ooe=cn(Oi,Poe),HE;function Loe(e){HE=e.wasm.cwrap(Li,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function zoe(e){let{inputs:t,attrs:n,backend:a}=e,r=t.x,s=a.dataIdMap.get(r.dataId).id;v.assert(r.dtype==="float32",()=>`Error in MaxPool: only float32 input is supported. Got ${r.dtype}.`);let{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,p=N.computePool2DInfo(r.shape,i,o,1,l,u),d=p.filterHeight,c=p.filterWidth,h=p.padInfo.top,m=p.padInfo.right,f=p.padInfo.bottom,g=p.padInfo.left,y=p.dilationHeight,b=p.dilationWidth,x=p.strideHeight,w=p.strideWidth,I=p.inChannels,T=p.outChannels;if(p.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${p.dataFormat}'. Please use 'channelsLast'.`);let C=a.makeOutput(p.outShape,"float32"),E=a.dataIdMap.get(C.dataId).id;return HE(s,r.shape[0],r.shape[1],r.shape[2],d,c,h,m,f,g,y,b,x,w,I,T,E),C}var Woe={kernelName:Li,backendName:"wasm",setupFunc:Loe,kernelFunc:zoe},jE;function Boe(e){jE=e.wasm.cwrap(zi,null,["number, number, number"])}function Voe(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Ss(i,r,t),m=d;if(h){let w=t.dataIdMap.get(p.dataId).id;w!==o&&(u=p,l=w,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("mean",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=u;u.dtype!=="float32"&&(b=Hu({backend:t,inputs:{x:u},attrs:{dtype:"float32"}}),l=t.dataIdMap.get(b.dataId).id);let x=t.makeOutput(f,"float32");if(v.sizeFromShape(u.shape)!==0){let w=t.dataIdMap.get(x.dataId).id;jE(l,y,w)}if(h&&t.disposeData(p.dataId),s){let w=N.expandShapeToKeepDim(x.shape,c);x.shape=w}return u.dtype!=="float32"&&t.disposeData(b.dataId),x}var Uoe={kernelName:zi,backendName:"wasm",setupFunc:Boe,kernelFunc:Voe},qE;function Goe(e){qE=e.wasm.cwrap(Wi,null,["number","number","number","number"])}function Hoe(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Ss(i,r,t);if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x)}let m=u.shape.length;N.assertAxesAreInnerMostDims("min",d,m);let[f,g]=N.computeOutAndReduceShapes(u.shape,d),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;qE(l,Et[i.dtype],y,x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var joe={kernelName:Wi,backendName:"wasm",setupFunc:Goe,kernelFunc:Hoe},qoe=!1,Koe=cn(Bi,qoe),fx;(function(e){e[e.reflect=0]="reflect",e[e.symmetric=1]="symmetric"})(fx||(fx={}));var KE;function Xoe(e){KE=e.wasm.cwrap(Vi,null,["number","array","number","number","array","array","number","number"])}function Yoe(e){let{inputs:{x:t},backend:n,attrs:{paddings:a,mode:r}}=e,s=a.map((m,f)=>m[0]+t.shape[f]+m[1]),i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),p=a.map(m=>m[0]),d=a.map(m=>m[1]),c=new Uint8Array(new Int32Array(p).buffer),h=new Uint8Array(new Int32Array(d).buffer);return KE(i,u,t.shape.length,Et[t.dtype],c,h,fx[r],l),o}var Zoe={kernelName:Vi,backendName:"wasm",kernelFunc:Yoe,setupFunc:Xoe},Joe=!0,Qoe=cn(Ui,Joe),ele=an(eu);function t1(e,t){let n=new Int32Array(e.wasm.HEAPU8.buffer,t,4),a=n[0],r=n[1],s=n[2],i=n[3];return e.wasm._free(t),{pSelectedIndices:a,selectedSize:r,pSelectedScores:s,pValidOutputs:i}}var XE;function tle(e){XE=e.wasm.cwrap(nu,"number",["number","number","number","number","number"])}function nle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i}=a,{boxes:o,scores:l}=n,u=t.dataIdMap.get(o.dataId).id,p=t.dataIdMap.get(l.dataId).id,d=XE(u,p,s,r,i),{pSelectedIndices:c,selectedSize:h,pSelectedScores:m,pValidOutputs:f}=t1(t,d);return t.wasm._free(m),t.wasm._free(f),t.makeOutput([h],"int32",c)}var ale={kernelName:nu,backendName:"wasm",setupFunc:tle,kernelFunc:nle},YE;function rle(e){YE=e.wasm.cwrap(au,"number",["number","number","number","number","number","bool"])}function sle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:o}=a,{boxes:l,scores:u}=n,p=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,c=YE(p,d,s,r,i,o),{pSelectedIndices:h,selectedSize:m,pSelectedScores:f,pValidOutputs:g}=t1(t,c);t.wasm._free(f);let y=t.makeOutput([m],"int32",h),b=t.makeOutput([],"int32",g);return[y,b]}var ile={kernelName:au,backendName:"wasm",setupFunc:rle,kernelFunc:sle},ZE;function ole(e){ZE=e.wasm.cwrap(ru,"number",["number","number","number","number","number","number"])}function lle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i,softNmsSigma:o}=a,{boxes:l,scores:u}=n,p=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,c=ZE(p,d,s,r,i,o),{pSelectedIndices:h,selectedSize:m,pSelectedScores:f,pValidOutputs:g}=t1(t,c);t.wasm._free(g);let y=t.makeOutput([m],"int32",h),b=t.makeOutput([m],"float32",f);return[y,b]}var ule={kernelName:ru,backendName:"wasm",setupFunc:ole,kernelFunc:lle},ple=!1,cle=cn(tu,ple,"bool"),JE;function dle(e){JE=e.wasm.cwrap(Gi,null,["number","number","number","number","number"])}function hle(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a,u=n.makeOutput([...r.shape,i],s),p=n.dataIdMap.get(u.dataId).id,d=n.dataIdMap.get(r.dataId).id;return JE(d,i,o,l,p),u}var mle={kernelName:Gi,backendName:"wasm",setupFunc:dle,kernelFunc:hle};function fle(e){let{inputs:{x:t},backend:n}=e,a=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(a).fill(1),a}var gle={kernelName:su,backendName:"wasm",kernelFunc:fle};function yle(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return mx({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=mx({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=_E({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeData(p.dataId)),u}var ble={kernelName:iu,backendName:"wasm",kernelFunc:yle},QE;function xle(e){QE=e.wasm.cwrap(Hi,null,["number","array","number","number","array","array","number","number"])}function vle(e){let{inputs:{x:t},backend:n,attrs:{paddings:a,constantValue:r}}=e,s=a.map((m,f)=>m[0]+t.shape[f]+m[1]);if(v.sizeFromShape(t.shape)===0)return PE({backend:n,attrs:{shape:s,value:r,dtype:t.dtype}});let i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),p=a.map(m=>m[0]),d=a.map(m=>m[1]),c=new Uint8Array(new Int32Array(p).buffer),h=new Uint8Array(new Int32Array(d).buffer);return QE(i,u,t.shape.length,Et[t.dtype],c,h,r,l),o}var eA={kernelName:Hi,backendName:"wasm",kernelFunc:vle,setupFunc:xle},wle=!1,kle=cn(ji,wle),tA;function Ile(e){tA=e.wasm.cwrap(qi,null,["number","number","number"])}function Sle(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t,s=n.dataIdMap.get(a.dataId).id,i=n.dataIdMap.get(r.dataId).id,o=s,l=a,u=l;l.dtype!=="float32"&&(u=Hu({backend:n,inputs:{x:a},attrs:{dtype:"float32"}}),o=n.dataIdMap.get(u.dataId).id);let p=n.makeOutput(a.shape,"float32"),d=n.dataIdMap.get(p.dataId).id;return tA(o,i,d),l.dtype!=="float32"&&n.disposeData(u.dataId),p}var Tle={kernelName:qi,backendName:"wasm",setupFunc:Ile,kernelFunc:Sle},nA;function Nle(e){nA=e.wasm.cwrap(Ki,null,["number","number","number","number"])}function Cle(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Ss(i,r,t),m=d;if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("prod",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;nA(l,y,Et[b.dtype],x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var _le={kernelName:Ki,backendName:"wasm",setupFunc:Nle,kernelFunc:Cle},Ele=e=>{let{backend:t,attrs:n}=e,{start:a,stop:r,step:s,dtype:i}=n,o=F0(a,r,s,i),l=t.makeOutput([o.length],i);return t.typedArrayFromHeap(l).set(o),l},Ale={kernelName:hc,backendName:"wasm",kernelFunc:Ele},$le=!0,Fle=cn(Ni,$le),Dle=an(Xi),Rle=an(Ji),aA;function Mle(e){aA=e.wasm.cwrap(Zi,null,["number","number","number","number","number","number","number","number","number","number"])}function Ple(e){let{backend:t,inputs:n,attrs:a}=e,{images:r}=n,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,[p,d,c,h]=r.shape,m=[p,l,u,h],f=t.dataIdMap.get(r.dataId),g;f.dtype!=="float32"&&(g=Hu({backend:t,inputs:{x:r},attrs:{dtype:"float32"}}),f=t.dataIdMap.get(g.dataId));let y=f.id,b=t.makeOutput(m,"float32");if(v.sizeFromShape(r.shape)===0)return b;let x=t.dataIdMap.get(b.dataId).id;return aA(y,p,d,c,h,l,u,s?1:0,i?1:0,x),g!=null&&t.disposeData(g.dataId),b}var Ole={kernelName:Zi,backendName:"wasm",setupFunc:Mle,kernelFunc:Ple},rA;function Lle(e){rA=e.wasm.cwrap(Yi,null,["number","number","number","number","number","number","number","number","number","number"])}function zle(e){let{backend:t,inputs:n,attrs:a}=e,{images:r}=n,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,[p,d,c,h]=r.shape,m=[p,l,u,h],f=t.makeOutput(m,"float32");if(v.sizeFromShape(r.shape)===0)return f;let g=t.dataIdMap.get(r.dataId),y;g.dtype!=="float32"&&(y=Hu({backend:t,inputs:{x:r},attrs:{dtype:"float32"}}),g=t.dataIdMap.get(y.dataId));let b=g.id,x=t.dataIdMap.get(f.dataId).id;return rA(b,p,d,c,h,l,u,s?1:0,i?1:0,x),y!=null&&t.disposeData(y.dataId),f}var Wle={kernelName:Yi,backendName:"wasm",setupFunc:Lle,kernelFunc:zle},sA;function Ble(e){sA=e.wasm.cwrap(Qi,null,["number","array","number","array","number","number"])}function Vle(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a,i=v.parseAxisParam(s,r.shape);if(r.shape.length===0)return Lf({inputs:{x:r},backend:n});let o=n.makeOutput(r.shape,r.dtype),l=n.dataIdMap.get(r.dataId).id,u=n.dataIdMap.get(o.dataId).id,p=new Uint8Array(new Int32Array(i).buffer),d=new Uint8Array(new Int32Array(r.shape).buffer);sA(l,p,i.length,d,r.shape.length,u);let c=Wn({inputs:{x:o},attrs:{shape:r.shape},backend:n});return n.disposeData(o.dataId),c}var Ule={kernelName:Qi,backendName:"wasm",kernelFunc:Vle,setupFunc:Ble},iA;function Gle(e){iA=e.wasm.cwrap(Su,null,["number","number","number","number","number","number","number","number","array","number","number"])}function Hle(e){let{inputs:t,backend:n,attrs:a}=e,{image:r}=t,{radians:s,fillValue:i,center:o}=a,l=n.makeOutput(r.shape,r.dtype),u=n.dataIdMap.get(r.dataId).id,p=n.dataIdMap.get(l.dataId).id,[d,c,h,m]=r.shape,[f,g]=N.getImageCenter(o,c,h),y=i===0,b=255,x=typeof i=="number"?[i,i,i,y?0:b]:[...i,b],w=new Uint8Array(new Int32Array(x).buffer);return iA(u,d,c,h,m,s,f,g,w,x.length,p),l}var jle={kernelName:Su,backendName:"wasm",kernelFunc:Hle,setupFunc:Gle},qle=an(eo),Kle=an(to),oA;function Xle(e){oA=e.wasm.cwrap(uu,null,["number","number","number","number","number","number","array","number","number"])}function Yle(e){let{backend:t,inputs:n,attrs:a}=e,{indices:r,updates:s}=n,{shape:i}=a,o=t.makeOutput(i,s.dtype);if(v.sizeFromShape(i)===0)return o;let{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=Px.calculateShapes(s,r,i),h=t.dataIdMap.get(r.dataId).id,m=t.dataIdMap.get(s.dataId).id,f=new Uint8Array(new Int32Array(d).buffer),g=t.dataIdMap.get(o.dataId).id;return oA(h,m,Et[s.dtype],l,u,p,f,c,g),o}var Zle={kernelName:uu,backendName:"wasm",setupFunc:Xle,kernelFunc:Yle},lA;function Jle(e){lA=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}function Qle(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t,i=n.dataIdMap.get(a.dataId).id,o=n.dataIdMap.get(r.dataId).id,l=n.dataIdMap.get(s.dataId).id,u=n.makeOutput(r.shape,r.dtype),p=n.dataIdMap.get(u.dataId).id,d=a.shape.length,c=r.shape.length,h=d===0||d>1||c===1?1:v.sizeFromShape(r.shape.slice(1));return lA(i,o,l,h,p),u}var eue={kernelName:pu,backendName:"wasm",kernelFunc:Qle,setupFunc:Jle},uA;function tue(e){uA=e.wasm.cwrap(ao,null,["number","number"])}function nue(e){let{backend:t,inputs:{x:n}}=e,a=t.dataIdMap.get(n.dataId).id,r=t.makeOutput(n.shape,n.dtype),s=t.dataIdMap.get(r.dataId).id;return v.sizeFromShape(r.shape)===0||uA(a,s),r}var aue={kernelName:"Sigmoid",backendName:"wasm",setupFunc:tue,kernelFunc:nue},rue=an(no),pA;function sue(e){pA=e.wasm.cwrap(io,null,["number","number","number","number"])}function iue(e){let{backend:t,inputs:{logits:n},attrs:{dim:a}}=e,r=t.dataIdMap.get(n.dataId).id,s=t.makeOutput(n.shape,n.dtype),i=t.dataIdMap.get(s.dataId).id,o=n.shape[a],l=v.sizeFromShape(n.shape)/o;return v.sizeFromShape(s.shape)===0||pA(r,i,o,l),s}var oue={kernelName:io,backendName:"wasm",setupFunc:sue,kernelFunc:iue};function lue(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a,o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<r.shape.length;++g)l.push([0,0]);let u=eA.kernelFunc({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),p=N.getReshaped(u.shape,s,o,!1),d=N.getPermuted(p.length,s.length,!1),c=N.getReshapedPermuted(u.shape,s,o,!1),h=Wn({inputs:{x:u},backend:n,attrs:{shape:p}}),m=ls({inputs:{x:h},backend:n,attrs:{perm:d}}),f=Wn({inputs:{x:m},backend:n,attrs:{shape:c}});return n.disposeData(u.dataId),n.disposeData(h.dataId),n.disposeData(m.dataId),f}var uue={kernelName:gu,backendName:"wasm",kernelFunc:lue},cA;function pue(e){cA=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])}function cue(e){let{backend:t,inputs:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=n,o=a.shape[0],l=a.shape[1],u=t.readSync(s.dataId)[0],p=[o+u,l],d=t.dataIdMap.get(a.dataId).id,c=t.dataIdMap.get(r.dataId).id,h=t.dataIdMap.get(i.dataId).id,m=t.makeOutput(p,a.dtype),f=t.dataIdMap.get(m.dataId).id,g=t.makeOutput(p.slice(0,1),r.dtype),y=t.dataIdMap.get(g.dataId).id,b=t.makeOutput([u],"bool"),x=t.dataIdMap.get(b.dataId).id,w=t.makeOutput([o],a.dtype),I=t.dataIdMap.get(w.dataId).id,T=t.makeOutput([4],"int32"),C=t.dataIdMap.get(T.dataId).id,E=cA(d,c,Et[r.dtype],o,u,l,h,f,y,x,I,C),A=t.readSync(T.dataId),R;switch(A[0]){case 1:{R=N.getSparseFillEmptyRowsIndicesDenseShapeMismatch(A[1]);break}case 2:{R=N.getSparseFillEmptyRowsNegativeIndexErrorMessage(A[1],A[2]);break}case 3:R=N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(A[1],A[2],A[3]);break;default:R=""}if(t.disposeData(T.dataId),R)throw t.disposeData(m.dataId),t.disposeData(g.dataId),t.disposeData(b.dataId),t.disposeData(w.dataId),new Error(R);let F=m,S=g;return E!==p[0]&&(F=di({inputs:{x:m},attrs:{begin:0,size:[E,l]},backend:t}),S=di({inputs:{x:g},attrs:{begin:0,size:E},backend:t}),t.disposeData(m.dataId),t.disposeData(g.dataId)),[F,S,b,w]}var due={kernelName:mc,backendName:"wasm",setupFunc:pue,kernelFunc:cue},dA;function hue(e){dA=e.wasm.cwrap(bu,null,["number","number","number","number","number","number","number"])}function mue(e){let{backend:t,inputs:n}=e,{inputIndices:a,inputShape:r,newShape:s}=n;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(a.dataId).id,o=t.dataIdMap.get(r.dataId).id,l=t.dataIdMap.get(s.dataId).id,u=a.shape[0],p=v.sizeFromShape(s.shape),d=t.makeOutput([u,p],a.dtype),c=t.dataIdMap.get(d.dataId).id,h=t.makeOutput([p],s.dtype),m=t.dataIdMap.get(h.dataId).id,f=t.makeOutput([3],"int32"),g=t.dataIdMap.get(f.dataId).id;dA(i,o,l,u,c,m,g);let y=t.readSync(f.dataId),b;switch(y[0]){case 0:{b=N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(y[1],y[2]);break}case 1:{b=N.getSparseReshapeNegativeOutputDimErrorMessage(y[1],y[2]);break}case 2:b=N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let x=Array.from(t.readSync(r.dataId)),w=Array.from(t.readSync(h.dataId));b=N.getSparseReshapeInputOutputMultipleErrorMessage(x,w);break}case 4:{let x=Array.from(t.readSync(r.dataId)),w=Array.from(t.readSync(h.dataId));b=N.getSparseReshapeInputOutputMismatchErrorMessage(x,w);break}default:b=""}if(t.disposeData(f.dataId),b)throw t.disposeData(d.dataId),t.disposeData(h.dataId),new Error(b);return[d,h]}var fue={kernelName:bu,backendName:"wasm",setupFunc:hue,kernelFunc:mue},hA;function mA(e){hA=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function fA(e,t){let{backend:n,inputs:a}=e,{data:r,indices:s,segmentIds:i}=a,o=s.shape[0],l=n.readSync(i.dataId,o-1,o)[0],u=o>0?l+1:0;if(u<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let p=r.shape.slice();p[0]=u;let d=n.dataIdMap.get(r.dataId).id,c=n.dataIdMap.get(s.dataId).id,h=n.dataIdMap.get(i.dataId).id,m=n.makeOutput(p,r.dtype),f=n.dataIdMap.get(m.dataId).id,g=n.makeOutput([4],"int32"),y=n.dataIdMap.get(g.dataId).id;hA(d,Et[r.dtype],r.shape[0],c,h,f,y,t,0);let b=n.readSync(g.dataId),x;switch(b[0]){case 0:{x=N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{x=N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:x=N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b[1],b[2]);break;case 3:x=N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(b[1],b[2],b[3]);break;default:x=""}if(n.disposeData(g.dataId),x)throw n.disposeData(m.dataId),new Error(x);return m}function gue(e){return fA(e,!0)}var yue={kernelName:fc,backendName:"wasm",setupFunc:mA,kernelFunc:gue};function bue(e){return fA(e,!1)}var xue={kernelName:gc,backendName:"wasm",setupFunc:mA,kernelFunc:bue};function vue(e){let{inputs:t,attrs:n,backend:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=n,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=new Array(r.shape.length).fill(0),p=r.shape.slice();return l.map(d=>{let c=[...p];c[o]=d;let h=di({inputs:{x:r},attrs:{begin:u,size:c},backend:a});return u[o]+=d,h})}var wue={kernelName:yu,backendName:"wasm",kernelFunc:vue},kue=an(ro),Iue=an(yc),Sue=!0,Tue=cn(oo,Sue),gA;function Nue(e){gA=e.wasm.cwrap(ms,null,["number","number","number","number"])}function Cue(e){let{backend:t,inputs:n,attrs:a}=e,{alpha:r}=a,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=t.makeOutput(s.shape,s.dtype),l=t.dataIdMap.get(o.dataId).id;return gA(i,r,Et[s.dtype],l),o}var _ue={kernelName:ms,backendName:"wasm",setupFunc:Nue,kernelFunc:Cue},yA;function Eue(e){yA=e.wasm.cwrap(xu,null,["number","array","number","array","array","array","array","array","number","number"])}function Aue(e){let{backend:t,inputs:n,attrs:a}=e,{x:r}=n,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a,{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=Wn({inputs:{x:r},backend:t,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let T=jt.computeOutShape(b,x,w),C=di({inputs:{x:r},backend:t,attrs:{begin:b,size:T}});I=Wn({inputs:{x:C},backend:t,attrs:{shape:m}}),t.disposeData(C.dataId)}else{let T=t.makeOutput(h,"float32"),C=t.dataIdMap.get(r.dataId).id,E=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),A=new Uint8Array(new Int32Array(b).buffer),R=new Uint8Array(new Int32Array(x).buffer),F=new Uint8Array(new Int32Array(w).buffer),S=new Uint8Array(new Int32Array(h).buffer),M=new Uint8Array(new Int32Array(v.computeStrides(h)).buffer),B=t.dataIdMap.get(T.dataId).id;yA(C,E,r.shape.length,A,R,F,S,M,h.length,B),I=Wn({inputs:{x:T},backend:t,attrs:{shape:m}}),t.disposeData(T.dataId)}return I}var $ue={kernelName:xu,backendName:"wasm",setupFunc:Eue,kernelFunc:Aue};function Fue(e){let{backend:t,inputs:n,attrs:a}=e,{data:r,dataSplits:s}=n,{separator:i,nGramWidths:o,leftPad:l,rightPad:u,padWidth:p,preserveShortSequences:d}=a,c=t.readSync(r.dataId),h=t.readSync(s.dataId),[m,f]=R0(c,h,i,o,l,u,p,d),g=t.makeOutput([m.length],"string"),y=t.dataIdMap.get(g.dataId);y.stringBytes=m;let b=t.makeOutput(s.shape,"int32");return t.typedArrayFromHeap(b).set(f),[g,b]}var Due={kernelName:bc,backendName:"wasm",kernelFunc:Fue};function Rue(e){let{backend:t,inputs:n,attrs:a}=e,{input:r,delimiter:s}=n,{skipEmpty:i}=a,o=t.readSync(r.dataId),l=t.readSync(s.dataId),[u,p,d]=M0(o,l[0],i),c=p.length,h=t.makeOutput([c,2],"int32");t.typedArrayFromHeap(h).set(u);let m=t.makeOutput([c],"string"),f=t.dataIdMap.get(m.dataId);f.stringBytes=p;let g=t.makeOutput([2],"int32");return t.typedArrayFromHeap(g).set(d),[h,m,g]}var Mue={kernelName:xc,backendName:"wasm",kernelFunc:Rue};function Pue(e){let{backend:t,inputs:n,attrs:a}=e,{input:r}=n,{numBuckets:s}=a,i=t.readSync(r.dataId),o=P0(i,s),l=t.makeOutput(r.shape,"int32");return t.typedArrayFromHeap(l).set(o),l}var Oue={kernelName:vc,backendName:"wasm",kernelFunc:Pue},Lue=!0,zue=cn(lo,Lue),bA;function Wue(e){bA=e.wasm.cwrap(so,null,["number","number","number","number"])}function Bue(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Ss(i,r,t),m=d;if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("sum",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;bA(l,y,Et[b.dtype],x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var Vue={kernelName:so,backendName:"wasm",setupFunc:Wue,kernelFunc:Bue},Uue=an(uo),Gue=an(po),xA;function Hue(e){xA=e.wasm.cwrap(hs,null,["number","array","number","array","number","number"])}function jue(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,s=n.dataIdMap.get(r.dataId).id,{reps:i}=a,o=new Array(r.shape.length);for(let c=0;c<o.length;c++)o[c]=r.shape[c]*i[c];let l=new Uint8Array(new Int32Array(r.shape).buffer),u=new Uint8Array(new Int32Array(o).buffer),p=n.makeOutput(o,r.dtype),d=n.dataIdMap.get(p.dataId).id;return xA(s,l,r.shape.length,u,o.length,Et[p.dtype],d),p}var que={kernelName:hs,backendName:"wasm",setupFunc:Hue,kernelFunc:jue},vA;function Kue(e){vA=e.wasm.cwrap(vu,null,["number","array","number","number","number","bool","number","number"])}var Xue=({inputs:e,backend:t,attrs:n})=>{let{x:a}=e,{k:r,sorted:s}=n,i=t.dataIdMap.get(a.dataId).id,o=new Uint8Array(new Int32Array(a.shape).buffer),l=a.shape.slice();l[l.length-1]=r;let u=t.makeOutput(l,a.dtype),p=t.dataIdMap.get(u.dataId).id,d=t.makeOutput(l,"int32"),c=t.dataIdMap.get(d.dataId).id;return vA(i,o,a.shape.length,Et[a.dtype],r,s,p,c),[u,d]},Yue={kernelName:vu,backendName:"wasm",setupFunc:Kue,kernelFunc:Xue},wA;function Zue(e){wA=e.wasm.cwrap(wu,null,["number","number","bool","number","number","number","number","number","number","array","number","array","number","number","number","number","number"])}function Jue(e){let{backend:t,inputs:n,attrs:a}=e,{image:r,transforms:s}=n,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=a,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),b=new Uint8Array(new Int32Array(v.computeStrides(g)).buffer),x=t.makeOutput(g,r.dtype),w=t.dataIdMap.get(x.dataId).id,I=t.dataIdMap.get(r.dataId).id,T=t.dataIdMap.get(s.dataId).id,C=i==="nearest"?1:2,E;switch(o){case"constant":E=1;break;case"reflect":E=2;break;case"wrap":E=3;break;case"nearest":E=4;break;default:E=1;break}return wA(I,T,s.shape[0]>1,p,m,f,h,c,d,y,r.shape.length-1,b,g.length-1,C,E,l,w),x}var Que={kernelName:wu,backendName:"wasm",setupFunc:Zue,kernelFunc:Jue};function epe(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r.shape[s],o=r.shape.length,l=new Array(o-1),u=0;for(let h=0;h<o;h++)h!==s&&(l[u++]=r.shape[h]);let p=new Array(i),d=new Array(o).fill(0),c=r.shape.slice();c[s]=1;for(let h=0;h<p.length;h++)d[s]=h,p[h]=di({inputs:{x:r},attrs:{begin:d,size:c},backend:n});return p.map(({dataId:h,dtype:m})=>({dataId:h,dtype:m,shape:l}))}var tpe={kernelName:ku,backendName:"wasm",kernelFunc:epe};function npe(e){let{inputs:{x:t},backend:n}=e,a=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(a).fill(0),a}var ape={kernelName:Iu,backendName:"wasm",kernelFunc:npe},rpe=[Fse,Dse,Mse,Lse,jse,Xse,Jse,tie,sie,cie,die,hie,gie,yie,vie,Iie,Sie,Tie,_ie,$ie,Rie,Oie,Wie,Bie,Uie,Gie,Hie,jie,Xie,Yie,Jie,toe,roe,ooe,poe,hoe,foe,yoe,zse,voe,koe,Soe,Toe,Coe,_oe,Aoe,Foe,Moe,Ooe,Woe,Uoe,joe,Koe,Zoe,Qoe,ele,ale,ile,ule,cle,mle,gle,ble,eA,kle,Tle,_le,Ale,Fle,Dle,Rle,nie,Ole,Wle,Ule,jle,qle,Kle,Zle,eue,aue,rue,uie,oue,uue,due,fue,yue,xue,wue,kue,Iue,Tue,_ue,$ue,Due,Mue,Oue,zue,Vue,Uue,Gue,que,Yue,Que,Use,tpe,ape];for(let e of rpe)kc(e);var gx=H();gx.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(e){return!1}});gx.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(gx.get("IS_NODE"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var fI=us(SF()),spe=us(TF()),gI=us(NF()),yI=fI.default||fI,ipe=gI.default||gI,kA=class extends rc{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(IA),yx=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new jh(this,Na())}write(e,t,n){let a={id:this.dataIdNextNumber++};return this.move(a,e,t,n,1),a}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}move(e,t,n,a,r){let s=this.dataIdNextNumber++;if(a==="string"){let u=t;this.dataIdMap.set(e,{id:s,stringBytes:u,shape:n,dtype:a,memoryOffset:null,refCount:r});return}let i=v.sizeFromShape(n),o=i*v.bytesPerElement(a),l=this.wasm._malloc(o);this.dataIdMap.set(e,{id:s,memoryOffset:l,shape:n,dtype:a,refCount:r}),this.wasm.tfjs.registerTensor(s,i,l),t!=null&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,o),l)}async read(e){return this.readSync(e)}readSync(e,t,n){let{memoryOffset:a,dtype:r,shape:s,stringBytes:i}=this.dataIdMap.get(e);if(r==="string")return(t==null||t===0)&&(n==null||n>=i.length)?i:i.slice(t,n);t=t||0,n=n||v.sizeFromShape(s);let o=v.bytesPerElement(r),l=this.wasm.HEAPU8.slice(a+t*o,a+n*o);return upe(l.buffer,r)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){let n=this.dataIdMap.get(e);if(n.refCount--,!t&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(e)}return!0}refCount(e){return this.dataIdMap.has(e)?this.dataIdMap.get(e).refCount:0}incRef(e){let t=this.dataIdMap.get(e);t!=null&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,n){let a;if(n==null)a=this.write(null,e,t);else{let r=this.dataIdNextNumber++;a={id:r},this.dataIdMap.set(a,{id:r,memoryOffset:n,shape:e,dtype:t,refCount:1});let s=v.sizeFromShape(e);this.wasm.tfjs.registerTensor(r,s,n)}return{dataId:a,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:n}){let a=this.wasm.HEAPU8.buffer,{memoryOffset:r}=this.dataIdMap.get(n),s=v.sizeFromShape(e);switch(t){case"float32":return new Float32Array(a,r,s);case"int32":return new Int32Array(a,r,s);case"bool":return new Uint8Array(a,r,s);default:throw new Error(`Unknown dtype ${t}`)}}};function ope(e){return(t,n)=>(v.fetch(e,{credentials:"same-origin"}).then(a=>{a.ok||t.env.a(`failed to load wasm binary file at '${e}'`),a.arrayBuffer().then(r=>{WebAssembly.instantiate(r,t).then(s=>{n(s.instance,s.module)})})}),{})}function bI(e,t,n){if(Gh!=null)return Gh;let a="tfjs-backend-wasm.wasm";return e&&t?a="tfjs-backend-wasm-threaded-simd.wasm":e&&(a="tfjs-backend-wasm-simd.wasm"),Wp!=null&&Wp[a]!=null?Wp[a]:n+a}async function lpe(){let[e,t]=await Promise.all([H().getAsync("WASM_HAS_SIMD_SUPPORT"),H().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((n,a)=>{let r={};r.locateFile=(o,l)=>{if(o.endsWith(".worker.js")){let u=spe.wasmWorkerContents.replace(/\n/g,"\\n"),p=new Blob([u],{type:"application/javascript"});return URL.createObjectURL(p)}return o.endsWith(".wasm")?bI(e,t,Op!=null?Op:l):l+o},n1&&(r.instantiateWasm=ope(bI(e,t,Op!=null?Op:"")));let s=!1;r.onAbort=()=>{s||Bp||(Bp=!0,a({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"}))};let i;t&&e&&Gh==null?(r.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+yI.toString()],{type:"text/javascript"}),i=yI(r)):i=ipe(r),i.then(o=>{s=!0,Bp=!1;let l=null;o.tfjs={init:o.cwrap("init",null,[]),initWithThreadsCount:o.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:o.cwrap("get_threads_count","number",[]),registerTensor:o.cwrap("register_tensor",null,["number","number","number"]),disposeData:o.cwrap("dispose_data",l,["number"]),dispose:o.cwrap("dispose",l,[])},n({wasm:o})}).catch(a)})}function upe(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error(`Unknown dtype ${t}`)}}var ppe=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],Gh=null,Op=null,Wp={},Bp=!1,n1=!1;function cpe(e,t=!1){if(Rx("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),Bp)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");Gh=e,n1=t}function dpe(e,t=!1){if(Bp)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if(typeof e=="string")Op=e;else{Wp=e;let n=ppe.filter(a=>Wp[a]==null);if(n.length>0)throw new Error(`There were no entries found for the following binaries: ${n.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}n1=t}var IA=-1,yx=-1;function hpe(e){IA=e}function mpe(){if(yx===-1)throw new Error("WASM backend not initialized.");return yx}var fpe="4.0.0",gpe=2;Em("wasm",async()=>{let{wasm:e}=await lpe();return new kA(e)},gpe);var xI="4.0.0",ype="4.0.0",bpe="4.0.0",xpe="4.0.0",vpe="4.0.0",wpe={tfjs:xI,"tfjs-core":xI,"tfjs-converter":ype,"tfjs-backend-cpu":bpe,"tfjs-backend-webgl":xpe,"tfjs-backend-wasm":vpe};var GA={};ab(GA,{AnchorPosition:()=>c1,DrawBox:()=>ud,DrawBoxOptions:()=>Uf,DrawFaceLandmarks:()=>tg,DrawFaceLandmarksOptions:()=>eg,DrawTextField:()=>Or,DrawTextFieldOptions:()=>Ju,drawContour:()=>Dr,drawDetections:()=>_pe,drawFaceExpressions:()=>Epe,drawFaceLandmarks:()=>$pe});function Dr(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:a,y:r},s)=>{let i=t[s];e.moveTo(i.x,i.y),e.lineTo(a,r)}),n){let a=t[t.length-1],r=t[0];if(!a||!r)return;e.moveTo(a.x,a.y),e.lineTo(r.x,r.y)}e.stroke()}var SA={};ab(SA,{computeReshapedDimensions:()=>s1,getCenterPoint:()=>To,isDimensions:()=>Wf,isEven:()=>zf,isFloat:()=>r1,isTensor:()=>Io,isTensor1D:()=>kpe,isTensor2D:()=>a1,isTensor3D:()=>Rr,isTensor4D:()=>xa,isValidNumber:()=>Ya,isValidProbablitiy:()=>ju,range:()=>hr,round:()=>So});var En=class{constructor(t,n){if(!Ya(t)||!Ya(n))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:n})}`);this._width=t,this._height=n}get width(){return this._width}get height(){return this._height}reverse(){return new En(1/this.width,1/this.height)}};function Io(e,t){return e instanceof Te&&e.shape.length===t}function kpe(e){return Io(e,1)}function a1(e){return Io(e,2)}function Rr(e){return Io(e,3)}function xa(e){return Io(e,4)}function r1(e){return e%1!==0}function zf(e){return e%2===0}function So(e,t=2){let n=10**t;return Math.floor(e*n)/n}function Wf(e){return e&&e.width&&e.height}function s1({width:e,height:t},n){let a=n/Math.max(t,e);return new En(Math.round(e*a),Math.round(t*a))}function To(e){return e.reduce((t,n)=>t.add(n),new Me(0,0)).div(new Me(e.length,e.length))}function hr(e,t,n){return Array(e).fill(0).map((a,r)=>t+r*n)}function Ya(e){return!!e&&e!==1/0&&e!==-1/0&&!Number.isNaN(e)||e===0}function ju(e){return Ya(e)&&e>=0&&e<=1}var Me=class{constructor(t,n){this._x=t,this._y=n}get x(){return this._x}get y(){return this._y}add(t){return new Me(this.x+t.x,this.y+t.y)}sub(t){return new Me(this.x-t.x,this.y-t.y)}mul(t){return new Me(this.x*t.x,this.y*t.y)}div(t){return new Me(this.x/t.x,this.y/t.y)}abs(){return new Me(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new Me(Math.floor(this.x),Math.floor(this.y))}};var ut=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(Ya)}static assertIsValidBox(t,n,a=!1){if(!ut.isRect(t))throw new Error(`${n} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!a&&(t.width<0||t.height<0))throw new Error(`${n} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,n=!0){let a=t||{},r=[a.left,a.top,a.right,a.bottom].every(Ya),s=[a.x,a.y,a.width,a.height].every(Ya);if(!s&&!r)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(a)}`);let[i,o,l,u]=s?[a.x,a.y,a.width,a.height]:[a.left,a.top,a.right-a.left,a.bottom-a.top];ut.assertIsValidBox({x:i,y:o,width:l,height:u},"Box.constructor",n),this._x=i,this._y=o,this._width=l,this._height=u}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new Me(this.left,this.top)}get topRight(){return new Me(this.right,this.top)}get bottomLeft(){return new Me(this.left,this.bottom)}get bottomRight(){return new Me(this.right,this.bottom)}round(){let[t,n,a,r]=[this.x,this.y,this.width,this.height].map(s=>Math.round(s));return new ut({x:t,y:n,width:a,height:r})}floor(){let[t,n,a,r]=[this.x,this.y,this.width,this.height].map(s=>Math.floor(s));return new ut({x:t,y:n,width:a,height:r})}toSquare(){let{x:t,y:n,width:a,height:r}=this,s=Math.abs(a-r);return a<r&&(t-=s/2,a+=s),r<a&&(n-=s/2,r+=s),new ut({x:t,y:n,width:a,height:r})}rescale(t){let n=Wf(t)?t.width:t,a=Wf(t)?t.height:t;return new ut({x:this.x*n,y:this.y*a,width:this.width*n,height:this.height*a})}pad(t,n){let[a,r,s,i]=[this.x-t/2,this.y-n/2,this.width+t,this.height+n];return new ut({x:a,y:r,width:s,height:i})}clipAtImageBorders(t,n){let{x:a,y:r,right:s,bottom:i}=this,o=Math.max(a,0),l=Math.max(r,0),u=s-o,p=i-l,d=Math.min(u,t-o),c=Math.min(p,n-l);return new ut({x:o,y:l,width:d,height:c}).floor()}shift(t,n){let{width:a,height:r}=this,s=this.x+t,i=this.y+n;return new ut({x:s,y:i,width:a,height:r})}padAtBorders(t,n){let a=this.width+1,r=this.height+1,s=1,i=1,o=a,l=r,u=this.left,p=this.top,d=this.right,c=this.bottom;return d>n&&(o=-d+n+a,d=n),c>t&&(l=-c+t+r,c=t),u<1&&(l=2-u,u=1),p<1&&(l=2-p,p=1),{dy:i,edy:l,dx:s,edx:o,y:p,ey:c,x:u,ex:d,w:a,h:r}}calibrate(t){return new ut({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var qu=class extends ut{constructor(t,n,a,r,s=!1){super({left:t,top:n,right:a,bottom:r},s)}};var Ts=class{constructor(t,n,a,r,s){this._imageDims=new En(s.width,s.height),this._score=t,this._classScore=n,this._className=a,this._box=new ut(r).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new ut(this._box).rescale(this.imageDims.reverse())}forSize(t,n){return new Ts(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:n})}};var xt=class extends Ts{constructor(t,n,a){super(t,t,"",n,a)}forSize(t,n){let{score:a,relativeBox:r,imageDims:s}=super.forSize(t,n);return new xt(a,r,s)}};function TA(e,t,n=!0){let a=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left)),r=Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top)),s=a*r;return n?s/(e.area+t.area-s):s/Math.min(e.area,t.area)}function NA(e){let t=e.map(o=>o.x),n=e.map(o=>o.y),a=t.reduce((o,l)=>l<o?l:o,1/0),r=n.reduce((o,l)=>l<o?l:o,1/0),s=t.reduce((o,l)=>o<l?l:o,0),i=n.reduce((o,l)=>o<l?l:o,0);return new qu(a,r,s,i)}function CA(e,t,n,a=!0){let r=t.map((i,o)=>({score:i,boxIndex:o})).sort((i,o)=>i.score-o.score).map(i=>i.boxIndex),s=[];for(;r.length>0;){let i=r.pop();s.push(i);let o=r,l=[];for(let u=0;u<o.length;u++){let p=o[u],d=e[i],c=e[p];l.push(TA(d,c,a))}r=r.filter((u,p)=>l[p]<=n)}return s}function mr(e,t){return P(()=>{let[n,a,r]=t,s=gn([...e.shape.slice(0,3),1],n,"float32"),i=gn([...e.shape.slice(0,3),1],a,"float32"),o=gn([...e.shape.slice(0,3),1],r,"float32"),l=Ze([s,i,o],3);return pe(e,l)})}function _A(e,t=!1){return P(()=>{let[n,a]=e.shape.slice(1);if(n===a)return e;let r=Math.abs(n-a),s=Math.round(r*(t?.5:1)),i=n>a?2:1,o=c=>{let h=e.shape.slice();return h[i]=c,gn(h,0,"float32")},l=o(s),u=r-l.shape[i],d=[t&&u?o(u):null,e,l].filter(c=>!!c).map(c=>oe(c,"float32"));return Ze(d,i)})}function Iye(e){let t=e.slice();for(let n=t.length-1;n>0;n--){let a=Math.floor(Math.random()*(n+1)),r=t[n];t[n]=t[a],t[a]=r}return t}function Bf(e){return 1/(1+Math.exp(-e))}function Tye(e){return Math.log(e/(1-e))}var Ku=class extends ut{constructor(t,n,a,r,s=!1){super({x:t,y:n,width:a,height:r},s)}};var Ipe=.5,Spe=.43,Tpe=.45,va=class{constructor(t,n,a=new Me(0,0)){let{width:r,height:s}=n;this._imgDims=new En(r,s),this._shift=a,this._positions=t.map(i=>i.mul(new Me(r,s)).add(a))}get shift(){return new Me(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new Me(this.imageWidth,this.imageHeight)))}forSize(t,n){return new this.constructor(this.relativePositions,{width:t,height:n})}shiftBy(t,n){return new this.constructor(this.relativePositions,this._imgDims,new Me(t,n))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,n={}){if(t){let s=t instanceof xt?t.box.floor():new ut(t);return this.shiftBy(s.x,s.y).align(null,n)}let{useDlibAlignment:a,minBoxPadding:r}={useDlibAlignment:!1,minBoxPadding:.2,...n};return a?this.alignDlib():this.alignMinBbox(r)}alignDlib(){let t=this.getRefPointsForAlignment(),[n,a,r]=t,s=d=>r.sub(d).magnitude(),i=(s(n)+s(a))/2,o=Math.floor(i/Tpe),l=To(t),u=Math.floor(Math.max(0,l.x-Ipe*o)),p=Math.floor(Math.max(0,l.y-Spe*o));return new Ku(u,p,Math.min(o,this.imageWidth+u),Math.min(o,this.imageHeight+p))}alignMinBbox(t){let n=NA(this.positions);return n.pad(n.width*t,n.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var EA=class extends va{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],To([t[3],t[4]])]}};var Xu=class extends va{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(To)}};var id=class{constructor(t,n){this._label=t,this._distance=n}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${So(this.distance)})`:""}`}};var od=class extends ut{constructor(n,a){super(n);this._label=a}static assertIsValidLabeledBox(n,a){if(ut.assertIsValidBox(n,a),!Ya(n.label))throw new Error(`${a} - expected property label (${n.label}) to be a number`)}get label(){return this._label}};var Mr=class{constructor(t,n){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(n)||n.some(a=>!(a instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=n}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let n=t.descriptors.map(a=>new Float32Array(a));return new Mr(t.label,n)}};var AA=class extends od{constructor(n,a,r,s){super(n,a);this._score=r,this._classScore=s}static assertIsValidPredictedBox(n,a){if(od.assertIsValidLabeledBox(n,a),!ju(n.score)||!ju(n.classScore))throw new Error(`${a} - expected properties score (${n.score}) and (${n.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function Pr(e){return e.detection instanceof xt}function Yu(e,t){return{...e,...{detection:t}}}function i1(){let e=window.fetch;if(!e)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:e,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function ld(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function Vf(e){let t="";if(!e&&ld())try{e=Gr("fs")}catch(a){t=a.toString()}return{readFile:e?a=>new Promise((r,s)=>{e.readFile(a,(i,o)=>i?s(i):r(o))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function o1(){let e=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,n=global.Video||global.HTMLVideoElement,a=()=>{if(e)return new e;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},r=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},s=()=>{if(n)return new n;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},i=global.fetch,o=Vf();return{Canvas:e||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:a,createImageElement:r,createVideoElement:s,fetch:i,...o}}function l1(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var rn;function Npe(){if(!rn)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return rn}function u1(e){rn=e}function p1(){return l1()?u1(i1()):ld()?u1(o1()):null}function Cpe(e){if(rn||p1(),!rn)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=rn.Canvas,Image:n=rn.Image}=e;rn.Canvas=t,rn.Image=n,rn.createCanvasElement=e.createCanvasElement||(()=>new t),rn.createImageElement=e.createImageElement||(()=>new n),rn.ImageData=e.ImageData||rn.ImageData,rn.Video=e.Video||rn.Video,rn.fetch=e.fetch||rn.fetch,rn.readFile=e.readFile||rn.readFile}var et={getEnv:Npe,setEnv:u1,initialize:p1,createBrowserEnv:i1,createFileSystem:Vf,createNodejsEnv:o1,monkeyPatch:Cpe,isBrowser:l1,isNodejs:ld};p1();function Zu(e){return!et.isNodejs()&&typeof e=="string"?document.getElementById(e):e}function aa(e){let{Canvas:t,CanvasRenderingContext2D:n}=et.getEnv();if(e instanceof n)return e;let a=Zu(e);if(!(a instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let r=a.getContext("2d");if(!r)throw new Error("resolveContext2d - canvas 2d context is null");return r}var c1=(r=>(r.TOP_LEFT="TOP_LEFT",r.TOP_RIGHT="TOP_RIGHT",r.BOTTOM_LEFT="BOTTOM_LEFT",r.BOTTOM_RIGHT="BOTTOM_RIGHT",r))(c1||{}),Ju=class{constructor(t={}){let{anchorPosition:n,backgroundColor:a,fontColor:r,fontSize:s,fontStyle:i,padding:o}=t;this.anchorPosition=n||"TOP_LEFT",this.backgroundColor=a||"rgba(0, 0, 0, 0.5)",this.fontColor=r||"rgba(255, 255, 255, 1)",this.fontSize=s||14,this.fontStyle=i||"Georgia",this.padding=o||4}},Or=class{constructor(t,n,a={}){this.text=typeof t=="string"?[t]:t instanceof Or?t.text:t,this.anchor=n,this.options=new Ju(a)}measureWidth(t){let{padding:n}=this.options;return this.text.map(a=>t.measureText(a).width).reduce((a,r)=>a<r?r:a,0)+2*n}measureHeight(){let{fontSize:t,padding:n}=this.options;return this.text.length*t+2*n}getUpperLeft(t,n){let{anchorPosition:a}=this.options,r=a==="BOTTOM_RIGHT"||a==="TOP_RIGHT",s=a==="BOTTOM_LEFT"||a==="BOTTOM_RIGHT",i=this.measureWidth(t),o=this.measureHeight(),l=r?this.anchor.x-i:this.anchor.x,u=s?this.anchor.y-o:this.anchor.y;if(n){let{width:p,height:d}=n,c=Math.max(Math.min(l,p-i),0),h=Math.max(Math.min(u,d-o),0);return{x:c,y:h}}return{x:l,y:u}}draw(t){let n=Zu(t),a=aa(n),{backgroundColor:r,fontColor:s,fontSize:i,fontStyle:o,padding:l}=this.options;a.font=`${i}px ${o}`;let u=this.measureWidth(a),p=this.measureHeight();a.fillStyle=r;let d=this.getUpperLeft(a,n);a.fillRect(d.x,d.y,u,p),a.fillStyle=s,this.text.forEach((c,h)=>{let m=l+d.x,f=l+d.y+(h+1)*i;a.fillText(c,m,f)})}};var Uf=class{constructor(t={}){let{boxColor:n,lineWidth:a,label:r,drawLabelOptions:s}=t;this.boxColor=n||"rgba(0, 0, 255, 1)",this.lineWidth=a||2,this.label=r;let i={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new Ju({...i,...s})}},ud=class{constructor(t,n={}){this.box=new ut(t),this.options=new Uf(n)}draw(t){let n=aa(t),{boxColor:a,lineWidth:r}=this.options,{x:s,y:i,width:o,height:l}=this.box;n.strokeStyle=a,n.lineWidth=r,n.strokeRect(s,i,o,l);let{label:u}=this.options;u&&new Or([u],{x:s-r/2,y:i},this.options.drawLabelOptions).draw(t)}};function _pe(e,t){(Array.isArray(t)?t:[t]).forEach(a=>{let r=a instanceof xt?a.score:Pr(a)?a.detection.score:void 0,s=a instanceof xt?a.box:Pr(a)?a.detection.box:new ut(a),i=r?`${So(r)}`:void 0;new ud(s,{label:i}).draw(e)})}function Gf(e){let{Image:t,Video:n}=et.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}function $A(e){return new Promise((t,n)=>{(e instanceof et.getEnv().Canvas||Gf(e))&&t(null);function a(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",r),s.currentTarget.removeEventListener("error",a),n(s))}function r(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",r),s.currentTarget.removeEventListener("error",a),t(s))}e.addEventListener("load",r),e.addEventListener("error",a)})}function FA(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToImage - expected buf to be of type: Blob"));let a=new FileReader;a.onload=()=>{typeof a.result!="string"&&n(new Error("bufferToImage - expected reader.result to be a string, in onload"));let r=et.getEnv().createImageElement();r.onload=()=>t(r),r.onerror=n,r.src=a.result},a.onerror=n,a.readAsDataURL(e)})}function Qu(e){let{Image:t,Video:n}=et.getEnv();return e instanceof t?new En(e.naturalWidth,e.naturalHeight):e instanceof n?new En(e.videoWidth,e.videoHeight):new En(e.width,e.height)}function ep({width:e,height:t}){let{createCanvasElement:n}=et.getEnv(),a=n();return a.width=e,a.height=t,a}function Hf(e,t){let{ImageData:n}=et.getEnv();if(!(e instanceof n)&&!Gf(e))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:a,height:r}=t||Qu(e),s=ep({width:a,height:r});return e instanceof n?aa(s).putImageData(e,0,0):aa(s).drawImage(e,0,0,a,r),s}async function DA(e,t){let n=t||et.getEnv().createCanvasElement(),[a,r,s]=e.shape.slice(xa(e)?1:0),i=P(()=>e.as3D(a,r,s).toInt());return await co.toPixels(i,n),i.dispose(),n}function d1(e){let{Image:t,Canvas:n,Video:a}=et.getEnv();return e instanceof t||e instanceof n||e instanceof a}function RA(e,t,n=!1){let{Image:a,Canvas:r}=et.getEnv();if(!(e instanceof a||e instanceof r))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return ep({width:1,height:1});let s=Qu(e),i=t/Math.max(s.height,s.width),o=i*s.width,l=i*s.height,u=ep({width:t,height:t}),p=e instanceof r?e:Hf(e),d=Math.abs(o-l)/2,c=n&&o<l?d:0,h=n&&l<o?d:0;return p.width>0&&p.height>0&&aa(u).drawImage(p,c,h,o,l),u}var Lr=class{constructor(t,n=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=n,this._batchSize=t.length,t.forEach((a,r)=>{if(Rr(a)){this._imageTensors[r]=a,this._inputDimensions[r]=a.shape;return}if(xa(a)){let i=a.shape[0];if(i!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${i} passed, but not supported in input array`);this._imageTensors[r]=a,this._inputDimensions[r]=a.shape.slice(1);return}let s=a instanceof et.getEnv().Canvas?a:Hf(a);this._canvases[r]=s,this._inputDimensions[r]=[s.height,s.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return hr(this.batchSize,0,1).map((t,n)=>this.getReshapedInputDimensions(n))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let n=this.getInputWidth(t),a=this.getInputHeight(t);return s1({width:n,height:a},this.inputSize)}toBatchTensor(t,n=!0){return this._inputSize=t,P(()=>{let a=hr(this.batchSize,0,1).map(s=>{let i=this.getInput(s);if(i instanceof Te){let o=xa(i)?i:Zt(i);return o=_A(o,n),(o.shape[1]!==t||o.shape[2]!==t)&&(o=za.resizeBilinear(o,[t,t],!1,!1)),o.as3D(t,t,3)}if(i instanceof et.getEnv().Canvas)return co.fromPixels(RA(i,t,n));throw new Error(`toBatchTensor - at batchIdx ${s}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${i}`)});return Ft(a.map(s=>oe(s,"float32"))).as4D(this.batchSize,t,t,3)})}};async function vt(e){if(e instanceof Lr)return e;let t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");let n=r=>Array.isArray(e)?` at input index ${r}:`:"",a=t.map(Zu);return a.forEach((r,s)=>{if(!d1(r)&&!Rr(r)&&!xa(r))throw typeof t[s]=="string"?new Error(`toNetInput -${n(s)} string passed, but could not resolve HTMLElement for element id ${t[s]}`):new Error(`toNetInput -${n(s)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(xa(r)){let i=r.shape[0];if(i!==1)throw new Error(`toNetInput -${n(s)} tf.Tensor4D with batchSize ${i} passed, but not supported in input array`)}}),await Promise.all(a.map(r=>d1(r)&&$A(r))),new Lr(a,Array.isArray(e))}async function pd(e,t){let{Canvas:n}=et.getEnv(),a=e;if(!(e instanceof n)){let i=await vt(e);if(i.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let o=i.getInput(0);a=o instanceof n?o:await DA(o)}let r=aa(a);return t.map(i=>i instanceof xt?i.forSize(a.width,a.height).box.floor():i).map(i=>i.clipAtImageBorders(a.width,a.height)).map(({x:i,y:o,width:l,height:u})=>{let p=ep({width:l,height:u});return l>0&&u>0&&aa(p).putImageData(r.getImageData(i,o,l,u),0,0),p})}async function cd(e,t){if(!Rr(e)&&!xa(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(xa(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return P(()=>{let[n,a,r]=e.shape.slice(xa(e)?1:0);return t.map(o=>o instanceof xt?o.forSize(a,n).box:o).map(o=>o.clipAtImageBorders(a,n)).filter(o=>o.width>0&&o.height>0).map(({x:o,y:l,width:u,height:p})=>mo(e.as3D(n,a,r),[l,o,0],[p,u,r]))})}async function Ns(e,t){let{fetch:n}=et.getEnv(),a=await n(e,t);if(!(a.status<400))throw new Error(`failed to fetch: (${a.status}) ${a.statusText}, from url: ${a.url}`);return a}async function pve(e){let t=await Ns(e),n=await t.blob();if(!n.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${n.type}, for url: ${t.url}`);return FA(n)}async function MA(e){return(await Ns(e)).json()}async function fve(e){return new Float32Array(await(await Ns(e)).arrayBuffer())}function PA(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToVideo - expected buf to be of type: Blob"));let a=et.getEnv().createVideoElement();a.oncanplay=()=>t(a),a.onerror=n,a.playsInline=!0,a.muted=!0,a.src=URL.createObjectURL(e),a.play()})}async function wve(e){let t=await Ns(e),n=await t.blob();if(!n.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${n.type}, for url: ${t.url}`);return PA(n)}function jf(e,t){let n=`${t}-weights_manifest.json`;if(!e)return{modelBaseUri:"",manifestUri:n};if(e==="/")return{modelBaseUri:"/",manifestUri:`/${n}`};let a=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"";e=e.replace(a,"");let r=e.split("/").filter(o=>o),s=e.endsWith(".json")?r[r.length-1]:n,i=a+(e.endsWith(".json")?r.slice(0,r.length-1):r).join("/");return i=e.startsWith("/")?`/${i}`:i,{modelBaseUri:i,manifestUri:i==="/"?`/${s}`:`${i}/${s}`}}async function OA(e,t){let{manifestUri:n,modelBaseUri:a}=jf(e,t),r=await MA(n);return Ut.loadWeights(r,a)}function _ve(e,t,n=!1){let{width:a,height:r}=n?Qu(t):t;return e.width=a,e.height=r,{width:a,height:r}}var dn=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:n,objProp:a}=this.traversePropertyPath(t);return n[a]}reassignParamFromPath(t,n){let{obj:a,objProp:r}=this.traversePropertyPath(t);a[r].dispose(),a[r]=n}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof ts)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof ts))}variable(){this.getFrozenParams().forEach(({path:t,tensor:n})=>{this.reassignParamFromPath(t,n.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:n})=>{let a=kn(n.dataSync());n.dispose(),this.reassignParamFromPath(t,a)})}dispose(t=!0){this.getParamList().forEach(n=>{if(t&&n.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${n.path}`);n.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,n)=>t.concat(n)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let n=await OA(t,this.getDefaultModelName());this.loadFromWeightMap(n)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:n}=et.getEnv(),{manifestUri:a,modelBaseUri:r}=jf(t,this.getDefaultModelName()),s=u=>Promise.all(u.map(p=>n(p).then(d=>d.buffer))),i=Ut.weightsLoaderFactory(s),o=JSON.parse((await n(a)).toString()),l=await i(o,r);this.loadFromWeightMap(l)}loadFromWeightMap(t){let{paramMappings:n,params:a}=this.extractParamsFromWeightMap(t);this._paramMappings=n,this._params=a}extractWeights(t){let{paramMappings:n,params:a}=this.extractParams(t);this._paramMappings=n,this._params=a}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let n=t.split("/").reduce((s,i)=>{if(!s.nextObj.hasOwnProperty(i))throw new Error(`traversePropertyPath - object does not have property ${i}, for path ${t}`);return{obj:s.nextObj,objProp:i,nextObj:s.nextObj[i]}},{nextObj:this.params}),{obj:a,objProp:r}=n;if(!a||!r||!(a[r]instanceof Te))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:a,objProp:r}}};function Gn(e,t,n){return P(()=>{let a=vs(e,t.depthwise_filter,t.pointwise_filter,n,"same");return a=Y(a,t.bias),a})}function qf(e,t,n=!1){return P(()=>{let a=Xe(n?Y($t(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):Gn(e,t.conv0,[2,2])),r=Gn(a,t.conv1,[1,1]),s=Xe(Y(a,r)),i=Gn(s,t.conv2,[1,1]);return Xe(Y(a,Y(r,i)))})}function dd(e,t,n=!1,a=!0){return P(()=>{let r=Xe(n?Y($t(e,t.conv0.filters,a?[2,2]:[1,1],"same"),t.conv0.bias):Gn(e,t.conv0,a?[2,2]:[1,1])),s=Gn(r,t.conv1,[1,1]),i=Xe(Y(r,s)),o=Gn(i,t.conv2,[1,1]),l=Xe(Y(r,Y(s,o))),u=Gn(l,t.conv3,[1,1]);return Xe(Y(r,Y(s,Y(o,u))))})}function No(e,t,n="same",a=!1){return P(()=>{let r=Y($t(e,t.filters,[1,1],n),t.bias);return a?Xe(r):r})}function An(e,t){Object.keys(e).forEach(n=>{t.some(a=>a.originalPath===n)||e[n].dispose()})}function tp(e,t){return(n,a,r,s)=>{let i=$a(e(n*a*r*r),[r,r,n,a]),o=Ke(e(a));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/bias`}),{filters:i,bias:o}}}function Kf(e,t){return(n,a,r)=>{let s=_a(e(n*a),[n,a]),i=Ke(e(a));return t.push({paramPath:`${r}/weights`},{paramPath:`${r}/bias`}),{weights:s,bias:i}}}var hd=class{constructor(t,n,a){this.depthwise_filter=t;this.pointwise_filter=n;this.bias=a}};function np(e,t){return(n,a,r)=>{let s=$a(e(9*n),[3,3,n,1]),i=$a(e(n*a),[1,1,n,a]),o=Ke(e(a));return t.push({paramPath:`${r}/depthwise_filter`},{paramPath:`${r}/pointwise_filter`},{paramPath:`${r}/bias`}),new hd(s,i,o)}}function ap(e){return t=>{let n=e(`${t}/depthwise_filter`,4),a=e(`${t}/pointwise_filter`,4),r=e(`${t}/bias`,1);return new hd(n,a,r)}}function ra(e,t){return(n,a,r)=>{let s=e[n];if(!Io(s,a))throw new Error(`expected weightMap[${n}] to be a Tensor${a}D, instead have ${s}`);return t.push({originalPath:n,paramPath:r||n}),s}}function $n(e){let t=e;function n(r){let s=t.slice(0,r);return t=t.slice(r),s}function a(){return t}return{extractWeights:n,getRemainingWeights:a}}function Xf(e,t){let n=tp(e,t),a=np(e,t);function r(i,o,l,u=!1){let p=u?n(i,o,3,`${l}/conv0`):a(i,o,`${l}/conv0`),d=a(o,o,`${l}/conv1`),c=a(o,o,`${l}/conv2`);return{conv0:p,conv1:d,conv2:c}}function s(i,o,l,u=!1){let{conv0:p,conv1:d,conv2:c}=r(i,o,l,u),h=a(o,o,`${l}/conv3`);return{conv0:p,conv1:d,conv2:c,conv3:h}}return{extractDenseBlock3Params:r,extractDenseBlock4Params:s}}function LA(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractDenseBlock4Params:r}=Xf(n,t),s=r(3,32,"dense0",!0),i=r(32,64,"dense1"),o=r(64,128,"dense2"),l=r(128,256,"dense3");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o,dense3:l}}}function Yf(e){return t=>{let n=e(`${t}/filters`,4),a=e(`${t}/bias`,1);return{filters:n,bias:a}}}function Zf(e,t){let n=ra(e,t),a=Yf(n),r=ap(n);function s(o,l=!1){let u=l?a(`${o}/conv0`):r(`${o}/conv0`),p=r(`${o}/conv1`),d=r(`${o}/conv2`);return{conv0:u,conv1:p,conv2:d}}function i(o,l=!1){let u=l?a(`${o}/conv0`):r(`${o}/conv0`),p=r(`${o}/conv1`),d=r(`${o}/conv2`),c=r(`${o}/conv3`);return{conv0:u,conv1:p,conv2:d,conv3:c}}return{extractDenseBlock3Params:s,extractDenseBlock4Params:i}}function zA(e){let t=[],{extractDenseBlock4Params:n}=Zf(e,t),a={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return An(e,t),{params:a,paramMappings:t}}var rp=class extends dn{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceFeatureExtractor - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(112,!0),"float32"),s=mr(a,[122.782,117.001,104.298]).div(255),i=dd(s,n.dense0,!0);return i=dd(i,n.dense1),i=dd(i,n.dense2),i=dd(i,n.dense3),i=ga(i,[7,7],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await vt(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return zA(t)}extractParams(t){return LA(t)}};function md(e,t){return P(()=>Y(Fe(e,t.weights),t.bias))}function WA(e,t,n){let a=[],{extractWeights:r,getRemainingWeights:s}=$n(e),o=Kf(r,a)(t,n,"fc");if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{paramMappings:a,params:{fc:o}}}function BA(e){let t=[],n=ra(e,t);function a(s){let i=n(`${s}/weights`,2),o=n(`${s}/bias`,1);return{weights:i,bias:o}}let r={fc:a("fc")};return An(e,t),{params:r,paramMappings:t}}function Jf(e){let t={},n={};return Object.keys(e).forEach(a=>{let r=a.startsWith("fc")?n:t;r[a]=e[a]}),{featureExtractorMap:t,classifierMap:n}}var sp=class extends dn{constructor(n,a){super(n);this._faceFeatureExtractor=a}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:a}=this;if(!a)throw new Error(`${this._name} - load model before inference`);return P(()=>{let r=n instanceof Lr?this.faceFeatureExtractor.forwardInput(n):n;return md(r.as2D(r.shape[0],-1),a.fc)})}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:a,paramMappings:r}=this.extractClassifierParams(n);this._params=a,this._paramMappings=r}extractClassifierParams(n){return WA(n,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(n){let{featureExtractorMap:a,classifierMap:r}=Jf(n);return this.faceFeatureExtractor.loadFromWeightMap(a),BA(r)}extractParams(n){let a=this.getClassifierChannelsIn(),r=this.getClassifierChannelsOut(),s=r*a+r,i=n.slice(0,n.length-s),o=n.slice(n.length-s);return this.faceFeatureExtractor.extractWeights(i),this.extractClassifierParams(o)}};var VA=["neutral","happy","sad","angry","fearful","disgusted","surprised"],Cs=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);VA.forEach((n,a)=>{this[n]=t[a]})}asSortedArray(){return VA.map(t=>({expression:t,probability:this[t]})).sort((t,n)=>n.probability-t.probability)}};var Qf=class extends sp{constructor(t=new rp){super("FaceExpressionNet",t)}forwardInput(t){return P(()=>Ka(this.runNet(t)))}async forward(t){return this.forwardInput(await vt(t))}async predictExpressions(t){let n=await vt(t),a=await this.forwardInput(n),r=await Promise.all(ct(a).map(async i=>{let o=i.dataSync();return i.dispose(),o}));a.dispose();let s=r.map(i=>new Cs(i));return n.isBatchInput?s:s[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function UA(e){return e.expressions instanceof Cs}function h1(e,t){return{...e,...{expressions:t}}}function Epe(e,t,n=.1,a){(Array.isArray(t)?t:[t]).forEach(s=>{let i=s instanceof Cs?s:UA(s)?s.expressions:void 0;if(!i)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let l=i.asSortedArray().filter(d=>d.probability>n),u=Pr(s)?s.detection.box.bottomLeft:a||new Me(0,0);new Or(l.map(d=>`${d.expression} (${So(d.probability)})`),u).draw(e)})}function ip(e){return Pr(e)&&e.landmarks instanceof va&&e.unshiftedLandmarks instanceof va&&e.alignedRect instanceof xt}function Ape(e){let t=l=>l*180/Math.PI,n=(l,u)=>Math.sqrt((l._x-u._x)**2+(l._y-u._y)**2),a={roll:void 0,pitch:void 0,yaw:void 0},r=(l,u,p)=>{let d=Math.floor(l._x-u._x),c=Math.floor(u._x-p._x);return d-c},s=(l,u)=>{let p=Math.hypot(u._x-l._x,u._y-l._y),d=u._y-l._y,c=Math.asin(d/p),h=t(c),m=Math.floor(90-h),f=u._x-l._x<0?-1:1;return m*f},i=(l,u,p)=>{let d=n(l,p),c={_x:(l._x+p._x)/2,_y:(l._y+p._y)/2},h=n(u,c),m=Math.atan(h/d),f=Math.floor(t(m)),g=c._y-u._y<0?-1:1;return f*g};if(!e||!e._positions||e._positions.length!==68)return a;let o=e._positions;return a.roll=s(o[27],o[66]),a.pitch=i(o[14],o[30],o[2]),a.yaw=r(o[14],o[33],o[2]),a}function fd(e,t){let{box:n}=e.detection,a=t.shiftBy(n.x,n.y),r=a.align(),{imageDims:s}=e.detection,i=new xt(e.detection.score,r.rescale(s.reverse()),s),o=Ape(t);return{...e,...{landmarks:a,unshiftedLandmarks:t,alignedRect:i,angle:o}}}var eg=class{constructor(t={}){let{drawLines:n=!0,drawPoints:a=!0,lineWidth:r,lineColor:s,pointSize:i,pointColor:o}=t;this.drawLines=n,this.drawPoints=a,this.lineWidth=r||1,this.pointSize=i||2,this.lineColor=s||"rgba(0, 255, 255, 1)",this.pointColor=o||"rgba(255, 0, 255, 1)"}},tg=class{constructor(t,n={}){this.faceLandmarks=t,this.options=new eg(n)}draw(t){let n=aa(t),{drawLines:a,drawPoints:r,lineWidth:s,lineColor:i,pointSize:o,pointColor:l}=this.options;if(a&&this.faceLandmarks instanceof Xu&&(n.strokeStyle=i,n.lineWidth=s,Dr(n,this.faceLandmarks.getJawOutline()),Dr(n,this.faceLandmarks.getLeftEyeBrow()),Dr(n,this.faceLandmarks.getRightEyeBrow()),Dr(n,this.faceLandmarks.getNose()),Dr(n,this.faceLandmarks.getLeftEye(),!0),Dr(n,this.faceLandmarks.getRightEye(),!0),Dr(n,this.faceLandmarks.getMouth(),!0)),r){n.strokeStyle=l,n.fillStyle=l;let u=p=>{n.beginPath(),n.arc(p.x,p.y,o,0,2*Math.PI),n.fill()};this.faceLandmarks.positions.forEach(u)}}};function $pe(e,t){(Array.isArray(t)?t:[t]).forEach(a=>{let r=a instanceof va?a:ip(a)?a.landmarks:void 0;if(!r)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new tg(r).draw(e)})}var HA="1.7.6";function Rpe(e,t){let n=tp(e,t),a=np(e,t);function r(i,o,l){let u=a(i,o,`${l}/separable_conv0`),p=a(o,o,`${l}/separable_conv1`),d=n(i,o,1,`${l}/expansion_conv`);return{separable_conv0:u,separable_conv1:p,expansion_conv:d}}function s(i,o){let l=a(i,i,`${o}/separable_conv0`),u=a(i,i,`${o}/separable_conv1`),p=a(i,i,`${o}/separable_conv2`);return{separable_conv0:l,separable_conv1:u,separable_conv2:p}}return{extractConvParams:n,extractSeparableConvParams:a,extractReductionBlockParams:r,extractMainBlockParams:s}}function jA(e,t){let n=[],{extractWeights:a,getRemainingWeights:r}=$n(e),{extractConvParams:s,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:l}=Rpe(a,n),u=s(3,32,3,"entry_flow/conv_in"),p=o(32,64,"entry_flow/reduction_block_0"),d=o(64,128,"entry_flow/reduction_block_1"),c={conv_in:u,reduction_block_0:p,reduction_block_1:d},h={};hr(t,0,1).forEach(y=>{h[`main_block_${y}`]=l(128,`middle_flow/main_block_${y}`)});let m=o(128,256,"exit_flow/reduction_block"),f=i(256,512,"exit_flow/separable_conv"),g={reduction_block:m,separable_conv:f};if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:n,params:{entry_flow:c,middle_flow:h,exit_flow:g}}}function Mpe(e,t){let n=ra(e,t),a=Yf(n),r=ap(n);function s(o){let l=r(`${o}/separable_conv0`),u=r(`${o}/separable_conv1`),p=a(`${o}/expansion_conv`);return{separable_conv0:l,separable_conv1:u,expansion_conv:p}}function i(o){let l=r(`${o}/separable_conv0`),u=r(`${o}/separable_conv1`),p=r(`${o}/separable_conv2`);return{separable_conv0:l,separable_conv1:u,separable_conv2:p}}return{extractConvParams:a,extractSeparableConvParams:r,extractReductionBlockParams:s,extractMainBlockParams:i}}function qA(e,t){let n=[],{extractConvParams:a,extractSeparableConvParams:r,extractReductionBlockParams:s,extractMainBlockParams:i}=Mpe(e,n),o=a("entry_flow/conv_in"),l=s("entry_flow/reduction_block_0"),u=s("entry_flow/reduction_block_1"),p={conv_in:o,reduction_block_0:l,reduction_block_1:u},d={};hr(t,0,1).forEach(f=>{d[`main_block_${f}`]=i(`middle_flow/main_block_${f}`)});let c=s("exit_flow/reduction_block"),h=r("exit_flow/separable_conv"),m={reduction_block:c,separable_conv:h};return An(e,n),{params:{entry_flow:p,middle_flow:d,exit_flow:m},paramMappings:n}}function KA(e,t,n){return Y($t(e,t.filters,n,"same"),t.bias)}function m1(e,t,n=!0){let a=n?Xe(e):e;return a=Gn(a,t.separable_conv0,[1,1]),a=Gn(Xe(a),t.separable_conv1,[1,1]),a=Dt(a,[3,3],[2,2],"same"),a=Y(a,KA(e,t.expansion_conv,[2,2])),a}function Ppe(e,t){let n=Gn(Xe(e),t.separable_conv0,[1,1]);return n=Gn(Xe(n),t.separable_conv1,[1,1]),n=Gn(Xe(n),t.separable_conv2,[1,1]),n=Y(n,e),n}var ng=class extends dn{constructor(n){super("TinyXception");this._numMainBlocks=n}forwardInput(n){let{params:a}=this;if(!a)throw new Error("TinyXception - load model before inference");return P(()=>{let r=oe(n.toBatchTensor(112,!0),"float32"),i=mr(r,[122.782,117.001,104.298]).div(255),o=Xe(KA(i,a.entry_flow.conv_in,[2,2]));return o=m1(o,a.entry_flow.reduction_block_0,!1),o=m1(o,a.entry_flow.reduction_block_1),hr(this._numMainBlocks,0,1).forEach(l=>{o=Ppe(o,a.middle_flow[`main_block_${l}`])}),o=m1(o,a.exit_flow.reduction_block),o=Xe(Gn(o,a.exit_flow.separable_conv,[1,1])),o})}async forward(n){return this.forwardInput(await vt(n))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(n){return qA(n,this._numMainBlocks)}extractParams(n){return jA(n,this._numMainBlocks)}};function XA(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),r=Kf(n,t),s=r(512,1,"fc/age"),i=r(512,2,"fc/gender");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{fc:{age:s,gender:i}}}}function YA(e){let t=[],n=ra(e,t);function a(s){let i=n(`${s}/weights`,2),o=n(`${s}/bias`,1);return{weights:i,bias:o}}let r={fc:{age:a("fc/age"),gender:a("fc/gender")}};return An(e,t),{params:r,paramMappings:t}}var f1=(n=>(n.FEMALE="female",n.MALE="male",n))(f1||{});var ag=class extends dn{constructor(n=new ng(2)){super("AgeGenderNet");this._faceFeatureExtractor=n}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:a}=this;if(!a)throw new Error(`${this._name} - load model before inference`);return P(()=>{let r=n instanceof Lr?this.faceFeatureExtractor.forwardInput(n):n,s=ga(r,[7,7],[2,2],"valid").as2D(r.shape[0],-1),i=md(s,a.fc.age).as1D(),o=md(s,a.fc.gender);return{age:i,gender:o}})}forwardInput(n){return P(()=>{let{age:a,gender:r}=this.runNet(n);return{age:a,gender:Ka(r)}})}async forward(n){return this.forwardInput(await vt(n))}async predictAgeAndGender(n){let a=await vt(n),r=await this.forwardInput(a),s=ct(r.age),i=ct(r.gender),o=s.map((u,p)=>({ageTensor:u,genderTensor:i[p]})),l=await Promise.all(o.map(async({ageTensor:u,genderTensor:p})=>{let d=u.dataSync()[0],c=p.dataSync()[0],h=c>.5,m=h?"male":"female",f=h?c:1-c;return u.dispose(),p.dispose(),{age:d,gender:m,genderProbability:f}}));return r.age.dispose(),r.gender.dispose(),a.isBatchInput?l:l[0]}getDefaultModelName(){return"age_gender_model"}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:a,paramMappings:r}=this.extractClassifierParams(n);this._params=a,this._paramMappings=r}extractClassifierParams(n){return XA(n)}extractParamsFromWeightMap(n){let{featureExtractorMap:a,classifierMap:r}=Jf(n);return this.faceFeatureExtractor.loadFromWeightMap(a),YA(r)}extractParams(n){let r=n.slice(0,n.length-1539),s=n.slice(n.length-1539);return this.faceFeatureExtractor.extractWeights(r),this.extractClassifierParams(s)}};var op=class extends sp{postProcess(t,n,a){let r=a.map(({width:i,height:o})=>{let l=n/Math.max(o,i);return{width:i*l,height:o*l}}),s=r.length;return P(()=>{let i=(d,c)=>Ft([gn([68],d,"float32"),gn([68],c,"float32")],1).as2D(1,136).as1D(),o=(d,c)=>{let{width:h,height:m}=r[d];return c(h,m)?Math.abs(h-m)/2:0},l=d=>o(d,(c,h)=>c<h),u=d=>o(d,(c,h)=>h<c);return t.mul(gn([s,136],n,"float32")).sub(Ft(Array.from(Array(s),(d,c)=>i(l(c),u(c))))).div(Ft(Array.from(Array(s),(d,c)=>i(r[c].width,r[c].height))))})}forwardInput(t){return P(()=>{let n=this.runNet(t);return this.postProcess(n,t.inputSize,t.inputDimensions.map(([a,r])=>({height:a,width:r})))})}async forward(t){return this.forwardInput(await vt(t))}async detectLandmarks(t){let n=await vt(t),a=P(()=>ct(this.forwardInput(n))),r=await Promise.all(a.map(async(s,i)=>{let o=Array.from(s.dataSync()),l=o.filter((p,d)=>zf(d)),u=o.filter((p,d)=>!zf(d));return new Xu(Array(68).fill(0).map((p,d)=>new Me(l[d],u[d])),{height:n.getInputHeight(i),width:n.getInputWidth(i)})}));return a.forEach(s=>s.dispose()),n.isBatchInput?r:r[0]}getClassifierChannelsOut(){return 136}};var lp=class extends op{constructor(t=new rp){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};function ZA(e){let t=[],{extractDenseBlock3Params:n}=Zf(e,t),a={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return An(e,t),{params:a,paramMappings:t}}function JA(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractDenseBlock3Params:r}=Xf(n,t),s=r(3,32,"dense0",!0),i=r(32,64,"dense1"),o=r(64,128,"dense2");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o}}}var rg=class extends dn{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("TinyFaceFeatureExtractor - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(112,!0),"float32"),s=mr(a,[122.782,117.001,104.298]).div(255),i=qf(s,n.dense0,!0);return i=qf(i,n.dense1),i=qf(i,n.dense2),i=ga(i,[14,14],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await vt(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return ZA(t)}extractParams(t){return JA(t)}};var sg=class extends op{constructor(t=new rg){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var QA=class extends lp{};function e$(e,t){return Y(z(e,t.weights),t.biases)}function g1(e,t,n,a,r="same"){let{filters:s,bias:i}=t.conv,o=$t(e,s,n,r);return o=Y(o,i),o=e$(o,t.scale),a?Xe(o):o}function t$(e,t){return g1(e,t,[1,1],!0)}function y1(e,t){return g1(e,t,[1,1],!1)}function ig(e,t){return g1(e,t,[2,2],!0,"valid")}function Ope(e,t){function n(o,l,u){let p=e(o),d=p.length/(l*u*u);if(r1(d))throw new Error(`depth has to be an integer: ${d}, weights.length: ${p.length}, numFilters: ${l}, filterSize: ${u}`);return P(()=>Ee($a(p,[l,d,u,u]),[2,3,1,0]))}function a(o,l,u,p){let d=n(o,l,u),c=Ke(e(l));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/bias`}),{filters:d,bias:c}}function r(o,l){let u=Ke(e(o)),p=Ke(e(o));return t.push({paramPath:`${l}/weights`},{paramPath:`${l}/biases`}),{weights:u,biases:p}}function s(o,l,u,p){let d=a(o,l,u,`${p}/conv`),c=r(l,`${p}/scale`);return{conv:d,scale:c}}function i(o,l,u,p,d=!1){let c=s((d?.5:1)*o,l,u,`${p}/conv1`),h=s(o,l,u,`${p}/conv2`);return{conv1:c,conv2:h}}return{extractConvLayerParams:s,extractResidualLayerParams:i}}function n$(e){let{extractWeights:t,getRemainingWeights:n}=$n(e),a=[],{extractConvLayerParams:r,extractResidualLayerParams:s}=Ope(t,a),i=r(4704,32,7,"conv32_down"),o=s(9216,32,3,"conv32_1"),l=s(9216,32,3,"conv32_2"),u=s(9216,32,3,"conv32_3"),p=s(36864,64,3,"conv64_down",!0),d=s(36864,64,3,"conv64_1"),c=s(36864,64,3,"conv64_2"),h=s(36864,64,3,"conv64_3"),m=s(147456,128,3,"conv128_down",!0),f=s(147456,128,3,"conv128_1"),g=s(147456,128,3,"conv128_2"),y=s(589824,256,3,"conv256_down",!0),b=s(589824,256,3,"conv256_1"),x=s(589824,256,3,"conv256_2"),w=s(589824,256,3,"conv256_down_out"),I=P(()=>Ee(_a(t(256*128),[128,256]),[1,0]));if(a.push({paramPath:"fc"}),n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{params:{conv32_down:i,conv32_1:o,conv32_2:l,conv32_3:u,conv64_down:p,conv64_1:d,conv64_2:c,conv64_3:h,conv128_down:m,conv128_1:f,conv128_2:g,conv256_down:y,conv256_1:b,conv256_2:x,conv256_down_out:w,fc:I},paramMappings:a}}function Lpe(e,t){let n=ra(e,t);function a(i){let o=n(`${i}/scale/weights`,1),l=n(`${i}/scale/biases`,1);return{weights:o,biases:l}}function r(i){let o=n(`${i}/conv/filters`,4),l=n(`${i}/conv/bias`,1),u=a(i);return{conv:{filters:o,bias:l},scale:u}}function s(i){return{conv1:r(`${i}/conv1`),conv2:r(`${i}/conv2`)}}return{extractConvLayerParams:r,extractResidualLayerParams:s}}function a$(e){let t=[],{extractConvLayerParams:n,extractResidualLayerParams:a}=Lpe(e,t),r=n("conv32_down"),s=a("conv32_1"),i=a("conv32_2"),o=a("conv32_3"),l=a("conv64_down"),u=a("conv64_1"),p=a("conv64_2"),d=a("conv64_3"),c=a("conv128_down"),h=a("conv128_1"),m=a("conv128_2"),f=a("conv256_down"),g=a("conv256_1"),y=a("conv256_2"),b=a("conv256_down_out"),{fc:x}=e;if(t.push({originalPath:"fc",paramPath:"fc"}),!a1(x))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${x}`);let w={conv32_down:r,conv32_1:s,conv32_2:i,conv32_3:o,conv64_down:l,conv64_1:u,conv64_2:p,conv64_3:d,conv128_down:c,conv128_1:h,conv128_2:m,conv256_down:f,conv256_1:g,conv256_2:y,conv256_down_out:b,fc:x};return An(e,t),{params:w,paramMappings:t}}function Za(e,t){let n=t$(e,t.conv1);return n=y1(n,t.conv2),n=Y(n,e),n=Xe(n),n}function gd(e,t){let n=ig(e,t.conv1);n=y1(n,t.conv2);let a=ga(e,2,2,"valid"),r=It(a.shape),s=a.shape[3]!==n.shape[3];if(a.shape[1]!==n.shape[1]||a.shape[2]!==n.shape[2]){let o=[...n.shape];o[1]=1;let l=It(o);n=Ze([n,l],1);let u=[...n.shape];u[2]=1;let p=It(u);n=Ze([n,p],2)}return a=s?Ze([a,r],3):a,n=Y(a,n),n=Xe(n),n}var up=class extends dn{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceRecognitionNet - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(150,!0),"float32"),s=mr(a,[122.782,117.001,104.298]).div(255),i=ig(s,n.conv32_down);i=Dt(i,3,2,"valid"),i=Za(i,n.conv32_1),i=Za(i,n.conv32_2),i=Za(i,n.conv32_3),i=gd(i,n.conv64_down),i=Za(i,n.conv64_1),i=Za(i,n.conv64_2),i=Za(i,n.conv64_3),i=gd(i,n.conv128_down),i=Za(i,n.conv128_1),i=Za(i,n.conv128_2),i=gd(i,n.conv256_down),i=Za(i,n.conv256_1),i=Za(i,n.conv256_2),i=gd(i,n.conv256_down_out);let o=i.mean([1,2]);return Fe(o,n.fc)})}async forward(t){return this.forwardInput(await vt(t))}async computeFaceDescriptor(t){var s;if((s=t==null?void 0:t.shape)!=null&&s.some(i=>i<=0))return new Float32Array(128);let n=await vt(t),a=P(()=>ct(this.forwardInput(n))),r=await Promise.all(a.map(i=>i.data()));return a.forEach(i=>i.dispose()),n.isBatchInput?r:r[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return a$(t)}extractParams(t){return n$(t)}};function eIe(e){let t=new up;return t.extractWeights(e),t}function b1(e,t){return{...e,...{descriptor:t}}}function rIe(e){return typeof e.age=="number"}function x1(e,t){return{...e,...{age:t}}}function lIe(e){return(e.gender==="male"||e.gender==="female")&&ju(e.genderProbability)}function v1(e,t,n){return{...e,...{gender:t,genderProbability:n}}}function zpe(e,t){function n(l,u){let p=$a(e(9*l),[3,3,l,1]),d=Ke(e(l)),c=Ke(e(l)),h=Ke(e(l)),m=Ke(e(l));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/batch_norm_scale`},{paramPath:`${u}/batch_norm_offset`},{paramPath:`${u}/batch_norm_mean`},{paramPath:`${u}/batch_norm_variance`}),{filters:p,batch_norm_scale:d,batch_norm_offset:c,batch_norm_mean:h,batch_norm_variance:m}}function a(l,u,p,d,c){let h=$a(e(l*u*p*p),[p,p,l,u]),m=Ke(e(u));return t.push({paramPath:`${d}/filters`},{paramPath:`${d}/${c?"batch_norm_offset":"bias"}`}),{filters:h,bias:m}}function r(l,u,p,d){let{filters:c,bias:h}=a(l,u,p,d,!0);return{filters:c,batch_norm_offset:h}}function s(l,u,p){let d=n(l,`${p}/depthwise_conv`),c=r(l,u,1,`${p}/pointwise_conv`);return{depthwise_conv:d,pointwise_conv:c}}function i(){let l=r(3,32,3,"mobilenetv1/conv_0"),u=s(32,64,"mobilenetv1/conv_1"),p=s(64,128,"mobilenetv1/conv_2"),d=s(128,128,"mobilenetv1/conv_3"),c=s(128,256,"mobilenetv1/conv_4"),h=s(256,256,"mobilenetv1/conv_5"),m=s(256,512,"mobilenetv1/conv_6"),f=s(512,512,"mobilenetv1/conv_7"),g=s(512,512,"mobilenetv1/conv_8"),y=s(512,512,"mobilenetv1/conv_9"),b=s(512,512,"mobilenetv1/conv_10"),x=s(512,512,"mobilenetv1/conv_11"),w=s(512,1024,"mobilenetv1/conv_12"),I=s(1024,1024,"mobilenetv1/conv_13");return{conv_0:l,conv_1:u,conv_2:p,conv_3:d,conv_4:c,conv_5:h,conv_6:m,conv_7:f,conv_8:g,conv_9:y,conv_10:b,conv_11:x,conv_12:w,conv_13:I}}function o(){let l=r(1024,256,1,"prediction_layer/conv_0"),u=r(256,512,3,"prediction_layer/conv_1"),p=r(512,128,1,"prediction_layer/conv_2"),d=r(128,256,3,"prediction_layer/conv_3"),c=r(256,128,1,"prediction_layer/conv_4"),h=r(128,256,3,"prediction_layer/conv_5"),m=r(256,64,1,"prediction_layer/conv_6"),f=r(64,128,3,"prediction_layer/conv_7"),g=a(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),y=a(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),b=a(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),x=a(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),w=a(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),I=a(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),T=a(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),C=a(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),E=a(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),A=a(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),R=a(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),F=a(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:l,conv_1:u,conv_2:p,conv_3:d,conv_4:c,conv_5:h,conv_6:m,conv_7:f,box_predictor_0:{box_encoding_predictor:g,class_predictor:y},box_predictor_1:{box_encoding_predictor:b,class_predictor:x},box_predictor_2:{box_encoding_predictor:w,class_predictor:I},box_predictor_3:{box_encoding_predictor:T,class_predictor:C},box_predictor_4:{box_encoding_predictor:E,class_predictor:A},box_predictor_5:{box_encoding_predictor:R,class_predictor:F}}}return{extractMobilenetV1Params:i,extractPredictionLayerParams:o}}function r$(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractMobilenetV1Params:r,extractPredictionLayerParams:s}=zpe(n,t),i=r(),o=s(),u={extra_dim:_c(n(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:{mobilenetv1:i,prediction_layer:o,output_layer:u},paramMappings:t}}function Wpe(e,t){let n=ra(e,t);function a(u,p,d){let c=n(`${u}/Conv2d_${p}_pointwise/weights`,4,`${d}/filters`),h=n(`${u}/Conv2d_${p}_pointwise/convolution_bn_offset`,1,`${d}/batch_norm_offset`);return{filters:c,batch_norm_offset:h}}function r(u){let p=`mobilenetv1/conv_${u}`,d=`MobilenetV1/Conv2d_${u}_depthwise`,c=`${p}/depthwise_conv`,h=`${p}/pointwise_conv`,m=n(`${d}/depthwise_weights`,4,`${c}/filters`),f=n(`${d}/BatchNorm/gamma`,1,`${c}/batch_norm_scale`),g=n(`${d}/BatchNorm/beta`,1,`${c}/batch_norm_offset`),y=n(`${d}/BatchNorm/moving_mean`,1,`${c}/batch_norm_mean`),b=n(`${d}/BatchNorm/moving_variance`,1,`${c}/batch_norm_variance`);return{depthwise_conv:{filters:m,batch_norm_scale:f,batch_norm_offset:g,batch_norm_mean:y,batch_norm_variance:b},pointwise_conv:a("MobilenetV1",u,h)}}function s(){return{conv_0:a("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:r(1),conv_2:r(2),conv_3:r(3),conv_4:r(4),conv_5:r(5),conv_6:r(6),conv_7:r(7),conv_8:r(8),conv_9:r(9),conv_10:r(10),conv_11:r(11),conv_12:r(12),conv_13:r(13)}}function i(u,p){let d=n(`${u}/weights`,4,`${p}/filters`),c=n(`${u}/biases`,1,`${p}/bias`);return{filters:d,bias:c}}function o(u){let p=i(`Prediction/BoxPredictor_${u}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${u}/box_encoding_predictor`),d=i(`Prediction/BoxPredictor_${u}/ClassPredictor`,`prediction_layer/box_predictor_${u}/class_predictor`);return{box_encoding_predictor:p,class_predictor:d}}function l(){return{conv_0:a("Prediction",0,"prediction_layer/conv_0"),conv_1:a("Prediction",1,"prediction_layer/conv_1"),conv_2:a("Prediction",2,"prediction_layer/conv_2"),conv_3:a("Prediction",3,"prediction_layer/conv_3"),conv_4:a("Prediction",4,"prediction_layer/conv_4"),conv_5:a("Prediction",5,"prediction_layer/conv_5"),conv_6:a("Prediction",6,"prediction_layer/conv_6"),conv_7:a("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:o(0),box_predictor_1:o(1),box_predictor_2:o(2),box_predictor_3:o(3),box_predictor_4:o(4),box_predictor_5:o(5)}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:l}}function s$(e){let t=[],{extractMobilenetV1Params:n,extractPredictionLayerParams:a}=Wpe(e,t),r=e["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!Rr(r))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${r}`);let s={mobilenetv1:n(),prediction_layer:a(),output_layer:{extra_dim:r}};return An(e,t),{params:s,paramMappings:t}}function Ra(e,t,n){return P(()=>{let a=$t(e,t.filters,n,"same");return a=Y(a,t.batch_norm_offset),en(a,0,6)})}var Bpe=.0010000000474974513;function Vpe(e,t,n){return P(()=>{let a=bs(e,t.filters,n,"same");return a=ys(a,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,Bpe),en(a,0,6)})}function Upe(e){return[2,4,6,12].some(t=>t===e)?[2,2]:[1,1]}function i$(e,t){return P(()=>{let n,a=Ra(e,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((s,i)=>{let o=i+1,l=Upe(o);a=Vpe(a,s.depthwise_conv,l),a=Ra(a,s.pointwise_conv,[1,1]),o===11&&(n=a)}),n===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:a,conv11:n}})}function Gpe(e,t,n){let a=e.arraySync(),r=Math.min(a[t][0],a[t][2]),s=Math.min(a[t][1],a[t][3]),i=Math.max(a[t][0],a[t][2]),o=Math.max(a[t][1],a[t][3]),l=Math.min(a[n][0],a[n][2]),u=Math.min(a[n][1],a[n][3]),p=Math.max(a[n][0],a[n][2]),d=Math.max(a[n][1],a[n][3]),c=(i-r)*(o-s),h=(p-l)*(d-u);if(c<=0||h<=0)return 0;let m=Math.max(r,l),f=Math.max(s,u),g=Math.min(i,p),y=Math.min(o,d),b=Math.max(g-m,0)*Math.max(y-f,0);return b/(c+h-b)}function o$(e,t,n,a,r){let s=e.shape[0],i=Math.min(n,s),o=t.map((p,d)=>({score:p,boxIndex:d})).filter(p=>p.score>r).sort((p,d)=>d.score-p.score),l=p=>p<=a?1:0,u=[];return o.forEach(p=>{if(u.length>=i)return;let d=p.score;for(let c=u.length-1;c>=0;--c){let h=Gpe(e,p.boxIndex,u[c]);if(h!==0&&(p.score*=l(h),p.score<=r))break}d===p.score&&u.push(p.boxIndex)}),u}function Hpe(e){let t=ct(Ee(e,[1,0])),n=[pe(t[2],t[0]),pe(t[3],t[1])],a=[Y(t[0],he(n[0],2)),Y(t[1],he(n[1],2))];return{sizes:n,centers:a}}function jpe(e,t){let{sizes:n,centers:a}=Hpe(e),r=ct(Ee(t,[1,0])),s=he(z(fn(he(r[2],5)),n[0]),2),i=Y(z(he(r[0],10),n[0]),a[0]),o=he(z(fn(he(r[3],5)),n[1]),2),l=Y(z(he(r[1],10),n[1]),a[1]);return Ee(Ft([pe(i,s),pe(l,o),Y(i,s),Y(l,o)]),[1,0])}function l$(e,t,n){return P(()=>{let a=e.shape[0],r=jpe(W(Ln(n.extra_dim,[a,1,1]),[-1,4]),W(e,[-1,4]));r=W(r,[a,r.shape[0]/a,4]);let s=da(Be(t,[0,0,1],[-1,-1,-1])),i=Be(s,[0,0,0],[-1,-1,1]);i=W(i,[a,i.shape[1]]);let o=ct(r),l=ct(i);return{boxes:o,scores:l}})}function Co(e,t){return P(()=>{let n=e.shape[0],a=W(No(e,t.box_encoding_predictor),[n,-1,1,4]),r=W(No(e,t.class_predictor),[n,-1,3]);return{boxPredictionEncoding:a,classPrediction:r}})}function u$(e,t,n){return P(()=>{let a=Ra(e,n.conv_0,[1,1]),r=Ra(a,n.conv_1,[2,2]),s=Ra(r,n.conv_2,[1,1]),i=Ra(s,n.conv_3,[2,2]),o=Ra(i,n.conv_4,[1,1]),l=Ra(o,n.conv_5,[2,2]),u=Ra(l,n.conv_6,[1,1]),p=Ra(u,n.conv_7,[2,2]),d=Co(t,n.box_predictor_0),c=Co(e,n.box_predictor_1),h=Co(r,n.box_predictor_2),m=Co(i,n.box_predictor_3),f=Co(l,n.box_predictor_4),g=Co(p,n.box_predictor_5),y=Ze([d.boxPredictionEncoding,c.boxPredictionEncoding,h.boxPredictionEncoding,m.boxPredictionEncoding,f.boxPredictionEncoding,g.boxPredictionEncoding],1),b=Ze([d.classPrediction,c.classPrediction,h.classPrediction,m.classPrediction,f.classPrediction,g.classPrediction],1);return{boxPredictions:y,classPredictions:b}})}var Ma=class{constructor({minConfidence:t,maxResults:n}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=n||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var _o=class extends dn{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("SsdMobilenetv1 - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(512,!1),"float32"),r=pe(he(a,127.5),1),s=i$(r,n.mobilenetv1),{boxPredictions:i,classPredictions:o}=u$(s.out,s.conv11,n.prediction_layer);return l$(i,o,n.output_layer)})}async forward(t){return this.forwardInput(await vt(t))}async locateFaces(t,n={}){let{maxResults:a,minConfidence:r}=new Ma(n),s=await vt(t),{boxes:i,scores:o}=this.forwardInput(s),l=i[0],u=o[0];for(let x=1;x<i.length;x++)i[x].dispose(),o[x].dispose();let p=Array.from(u.dataSync()),c=o$(l,p,a,.5,r),h=s.getReshapedInputDimensions(0),m=s.inputSize,f=m/h.width,g=m/h.height,y=l.arraySync(),b=c.map(x=>{let[w,I]=[Math.max(0,y[x][0]),Math.min(1,y[x][2])].map(E=>E*g),[T,C]=[Math.max(0,y[x][1]),Math.min(1,y[x][3])].map(E=>E*f);return new xt(p[x],new Ku(T,w,C-T,I-w),{height:s.getInputHeight(0),width:s.getInputWidth(0)})});return l.dispose(),u.dispose(),b}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return s$(t)}extractParams(t){return r$(t)}};function qpe(e){let t=new _o;return t.extractWeights(e),t}function eSe(e){return qpe(e)}var p$=class extends _o{};var c$=.4,d$=[new Me(.738768,.874946),new Me(2.42204,2.65704),new Me(4.30971,7.04493),new Me(10.246,4.59428),new Me(12.6868,11.8741)],h$=[new Me(1.603231,2.094468),new Me(6.041143,7.080126),new Me(2.882459,3.518061),new Me(4.266906,5.178857),new Me(9.041765,10.66308)],m$=[117.001,114.697,97.404],f$="tiny_yolov2_model",g$="tiny_yolov2_separable_conv_model";var og=e=>typeof e=="number";function y$(e){if(!e)throw new Error(`invalid config: ${e}`);if(typeof e.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${e.withSeparableConvs}`);if(!og(e.iouThreshold)||e.iouThreshold<0||e.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${e.iouThreshold}`);if(!Array.isArray(e.classes)||!e.classes.length||!e.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(e.classes)}`);if(!Array.isArray(e.anchors)||!e.anchors.length||!e.anchors.map(t=>t||{}).every(t=>og(t.x)&&og(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(e.anchors)}`);if(e.meanRgb&&(!Array.isArray(e.meanRgb)||e.meanRgb.length!==3||!e.meanRgb.every(og)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(e.meanRgb)}`)}function pp(e){return P(()=>{let t=z(e,be(.10000000149011612));return Y(Xe(pe(e,t)),t)})}function zr(e,t){return P(()=>{let n=ya(e,[[0,0],[1,1],[1,1],[0,0]]);return n=$t(n,t.conv.filters,[1,1],"valid"),n=pe(n,t.bn.sub),n=z(n,t.bn.truediv),n=Y(n,t.conv.bias),pp(n)})}function Wr(e,t){return P(()=>{let n=ya(e,[[0,0],[1,1],[1,1],[0,0]]);return n=vs(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=Y(n,t.bias),pp(n)})}function Kpe(e,t){let n=tp(e,t);function a(i,o){let l=Ke(e(i)),u=Ke(e(i));return t.push({paramPath:`${o}/sub`},{paramPath:`${o}/truediv`}),{sub:l,truediv:u}}function r(i,o,l){let u=n(i,o,3,`${l}/conv`),p=a(o,`${l}/bn`);return{conv:u,bn:p}}let s=np(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:r,extractSeparableConvParams:s}}function b$(e,t,n,a){let{extractWeights:r,getRemainingWeights:s}=$n(e),i=[],{extractConvParams:o,extractConvWithBatchNormParams:l,extractSeparableConvParams:u}=Kpe(r,i),p;if(t.withSeparableConvs){let[d,c,h,m,f,g,y,b,x]=a,w=t.isFirstLayerConv2d?o(d,c,3,"conv0"):u(d,c,"conv0"),I=u(c,h,"conv1"),T=u(h,m,"conv2"),C=u(m,f,"conv3"),E=u(f,g,"conv4"),A=u(g,y,"conv5"),R=b?u(y,b,"conv6"):void 0,F=x?u(b,x,"conv7"):void 0,S=o(x||b||y,5*n,1,"conv8");p={conv0:w,conv1:I,conv2:T,conv3:C,conv4:E,conv5:A,conv6:R,conv7:F,conv8:S}}else{let[d,c,h,m,f,g,y,b,x]=a,w=l(d,c,"conv0"),I=l(c,h,"conv1"),T=l(h,m,"conv2"),C=l(m,f,"conv3"),E=l(f,g,"conv4"),A=l(g,y,"conv5"),R=l(y,b,"conv6"),F=l(b,x,"conv7"),S=o(x,5*n,1,"conv8");p={conv0:w,conv1:I,conv2:T,conv3:C,conv4:E,conv5:A,conv6:R,conv7:F,conv8:S}}if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{params:p,paramMappings:i}}function Xpe(e,t){let n=ra(e,t);function a(o){let l=n(`${o}/sub`,1),u=n(`${o}/truediv`,1);return{sub:l,truediv:u}}function r(o){let l=n(`${o}/filters`,4),u=n(`${o}/bias`,1);return{filters:l,bias:u}}function s(o){let l=r(`${o}/conv`),u=a(`${o}/bn`);return{conv:l,bn:u}}let i=ap(n);return{extractConvParams:r,extractConvWithBatchNormParams:s,extractSeparableConvParams:i}}function x$(e,t){let n=[],{extractConvParams:a,extractConvWithBatchNormParams:r,extractSeparableConvParams:s}=Xpe(e,n),i;if(t.withSeparableConvs){let o=t.filterSizes&&t.filterSizes.length||9;i={conv0:t.isFirstLayerConv2d?a("conv0"):s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:o>7?s("conv6"):void 0,conv7:o>8?s("conv7"):void 0,conv8:a("conv8")}}else i={conv0:r("conv0"),conv1:r("conv1"),conv2:r("conv2"),conv3:r("conv3"),conv4:r("conv4"),conv5:r("conv5"),conv6:r("conv6"),conv7:r("conv7"),conv8:a("conv8")};return An(e,n),{params:i,paramMappings:n}}var fr=class{constructor({inputSize:t,scoreThreshold:n}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=n||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var w1=class extends dn{constructor(n){super("TinyYolov2");y$(n),this._config=n}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(n,a){let r=zr(n,a.conv0);return r=Dt(r,[2,2],[2,2],"same"),r=zr(r,a.conv1),r=Dt(r,[2,2],[2,2],"same"),r=zr(r,a.conv2),r=Dt(r,[2,2],[2,2],"same"),r=zr(r,a.conv3),r=Dt(r,[2,2],[2,2],"same"),r=zr(r,a.conv4),r=Dt(r,[2,2],[2,2],"same"),r=zr(r,a.conv5),r=Dt(r,[2,2],[1,1],"same"),r=zr(r,a.conv6),r=zr(r,a.conv7),No(r,a.conv8,"valid",!1)}runMobilenet(n,a){let r=this.config.isFirstLayerConv2d?pp(No(n,a.conv0,"valid",!1)):Wr(n,a.conv0);return r=Dt(r,[2,2],[2,2],"same"),r=Wr(r,a.conv1),r=Dt(r,[2,2],[2,2],"same"),r=Wr(r,a.conv2),r=Dt(r,[2,2],[2,2],"same"),r=Wr(r,a.conv3),r=Dt(r,[2,2],[2,2],"same"),r=Wr(r,a.conv4),r=Dt(r,[2,2],[2,2],"same"),r=Wr(r,a.conv5),r=Dt(r,[2,2],[1,1],"same"),r=a.conv6?Wr(r,a.conv6):r,r=a.conv7?Wr(r,a.conv7):r,No(r,a.conv8,"valid",!1)}forwardInput(n,a){let{params:r}=this;if(!r)throw new Error("TinyYolov2 - load model before inference");return P(()=>{let s=oe(n.toBatchTensor(a,!1),"float32");return s=this.config.meanRgb?mr(s,this.config.meanRgb):s,s=s.div(255),this.config.withSeparableConvs?this.runMobilenet(s,r):this.runTinyYolov2(s,r)})}async forward(n,a){return this.forwardInput(await vt(n),a)}async detect(n,a={}){let{inputSize:r,scoreThreshold:s}=new fr(a),i=await vt(n),o=await this.forwardInput(i,r),l=P(()=>ct(o)[0].expandDims()),u={width:i.getInputWidth(0),height:i.getInputHeight(0)},p=await this.extractBoxes(l,i.getReshapedInputDimensions(0),s);o.dispose(),l.dispose();let d=p.map(y=>y.box),c=p.map(y=>y.score),h=p.map(y=>y.classScore),m=p.map(y=>this.config.classes[y.label]);return CA(d.map(y=>y.rescale(r)),c,this.config.iouThreshold,!0).map(y=>new Ts(c[y],h[y],m[y],d[y],u))}getDefaultModelName(){return""}extractParamsFromWeightMap(n){return x$(n,this.config)}extractParams(n){let a=this.config.filterSizes||w1.DEFAULT_FILTER_SIZES,r=a?a.length:void 0;if(r!==7&&r!==8&&r!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${r} filterSizes in config`);return b$(n,this.config,this.boxEncodingSize,a)}async extractBoxes(n,a,r){let{width:s,height:i}=a,o=Math.max(s,i),l=o/s,u=o/i,p=n.shape[1],d=this.config.anchors.length,[c,h,m]=P(()=>{let b=n.reshape([p,p,d,this.boxEncodingSize]),x=b.slice([0,0,0,0],[p,p,d,4]),w=b.slice([0,0,0,4],[p,p,d,1]),I=this.withClassScores?Ka(b.slice([0,0,0,5],[p,p,d,this.config.classes.length]),3):be(0);return[x,w,I]}),f=[],g=await h.array(),y=await c.array();for(let b=0;b<p;b++)for(let x=0;x<p;x++)for(let w=0;w<d;w++){let I=Bf(g[b][x][w][0]);if(!r||I>r){let T=(x+Bf(y[b][x][w][0]))/p*l,C=(b+Bf(y[b][x][w][1]))/p*u,E=Math.exp(y[b][x][w][2])*this.config.anchors[w].x/p*l,A=Math.exp(y[b][x][w][3])*this.config.anchors[w].y/p*u,R=T-E/2,F=C-A/2,S={row:b,col:x,anchor:w},{classScore:M,label:B}=this.withClassScores?await this.extractPredictedClass(m,S):{classScore:1,label:0};f.push({box:new qu(R,F,R+E,F+A),score:I,classScore:I*M,label:B,...S})}}return c.dispose(),h.dispose(),m.dispose(),f}async extractPredictedClass(n,a){let{row:r,col:s,anchor:i}=a,o=await n.array();return Array(this.config.classes.length).fill(0).map((l,u)=>o[r][s][i][u]).map((l,u)=>({classScore:l,label:u})).reduce((l,u)=>l.classScore>u.classScore?l:u)}},Eo=w1;Eo.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var cp=class extends Eo{constructor(t=!0){let n={withSeparableConvs:t,iouThreshold:c$,classes:["face"],...t?{anchors:h$,meanRgb:m$}:{anchors:d$,withClassScores:!0}};super(n)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(r=>new xt(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?g$:f$}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function HSe(e,t=!0){let n=new cp(t);return n.extractWeights(e),n}var lg=class extends fr{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var Pa=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function Ao(e,t,n,a,r=({alignedRect:s})=>s){let s=e.map(l=>ip(l)?r(l):l.detection),i=a||(t instanceof Te?await cd(t,s):await pd(t,s)),o=await n(i);return i.forEach(l=>l instanceof Te&&l.dispose()),o}async function dp(e,t,n,a,r){return Ao([e],t,async s=>n(s[0]),a,r)}var v$=.4,w$=[new Me(1.603231,2.094468),new Me(6.041143,7.080126),new Me(2.882459,3.518061),new Me(4.266906,5.178857),new Me(9.041765,10.66308)],k$=[117.001,114.697,97.404];var hp=class extends Eo{constructor(){let t={withSeparableConvs:!0,iouThreshold:v$,classes:["face"],anchors:w$,meanRgb:k$,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(r=>new xt(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var tt={ssdMobilenetv1:new _o,tinyFaceDetector:new hp,tinyYolov2:new cp,faceLandmark68Net:new lp,faceLandmark68TinyNet:new sg,faceRecognitionNet:new up,faceExpressionNet:new Qf,ageGenderNet:new ag},Ype=(e,t)=>tt.ssdMobilenetv1.locateFaces(e,t),wTe=(e,t)=>tt.tinyFaceDetector.locateFaces(e,t),kTe=(e,t)=>tt.tinyYolov2.locateFaces(e,t),Zpe=e=>tt.faceLandmark68Net.detectLandmarks(e),ITe=e=>tt.faceLandmark68TinyNet.detectLandmarks(e),STe=e=>tt.faceRecognitionNet.computeFaceDescriptor(e),TTe=e=>tt.faceExpressionNet.predictExpressions(e),NTe=e=>tt.ageGenderNet.predictAgeAndGender(e),Jpe=e=>tt.ssdMobilenetv1.load(e),CTe=e=>tt.tinyFaceDetector.load(e),_Te=e=>tt.tinyYolov2.load(e),ETe=e=>tt.faceLandmark68Net.load(e),ATe=e=>tt.faceLandmark68TinyNet.load(e),$Te=e=>tt.faceRecognitionNet.load(e),FTe=e=>tt.faceExpressionNet.load(e),DTe=e=>tt.ageGenderNet.load(e),RTe=Jpe,MTe=Ype,PTe=Zpe;var ug=class extends Pa{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.extractedFaces=r}},$o=class extends ug{async run(){let t=await this.parentTask,n=await Ao(t,this.input,async a=>Promise.all(a.map(r=>tt.faceExpressionNet.predictExpressions(r))),this.extractedFaces);return t.map((a,r)=>h1(a,n[r]))}withAgeAndGender(){return new Do(this,this.input)}},Fo=class extends ug{async run(){let t=await this.parentTask;if(!t)return;let n=await dp(t,this.input,a=>tt.faceExpressionNet.predictExpressions(a),this.extractedFaces);return h1(t,n)}withAgeAndGender(){return new Ro(this,this.input)}},_s=class extends $o{withAgeAndGender(){return new As(this,this.input)}withFaceDescriptors(){return new Fs(this,this.input)}},Es=class extends Fo{withAgeAndGender(){return new $s(this,this.input)}withFaceDescriptor(){return new Ds(this,this.input)}};var pg=class extends Pa{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.extractedFaces=r}},Do=class extends pg{async run(){let t=await this.parentTask,n=await Ao(t,this.input,async a=>Promise.all(a.map(r=>tt.ageGenderNet.predictAgeAndGender(r))),this.extractedFaces);return t.map((a,r)=>{let{age:s,gender:i,genderProbability:o}=n[r];return x1(v1(a,i,o),s)})}withFaceExpressions(){return new $o(this,this.input)}},Ro=class extends pg{async run(){let t=await this.parentTask;if(!t)return;let{age:n,gender:a,genderProbability:r}=await dp(t,this.input,s=>tt.ageGenderNet.predictAgeAndGender(s),this.extractedFaces);return x1(v1(t,a,r),n)}withFaceExpressions(){return new Fo(this,this.input)}},As=class extends Do{withFaceExpressions(){return new _s(this,this.input)}withFaceDescriptors(){return new Fs(this,this.input)}},$s=class extends Ro{withFaceExpressions(){return new Es(this,this.input)}withFaceDescriptor(){return new Ds(this,this.input)}};var cg=class extends Pa{constructor(n,a){super();this.parentTask=n;this.input=a}},Fs=class extends cg{async run(){let t=await this.parentTask;return(await Ao(t,this.input,a=>Promise.all(a.map(r=>tt.faceRecognitionNet.computeFaceDescriptor(r))),null,a=>a.landmarks.align(null,{useDlibAlignment:!0}))).map((a,r)=>b1(t[r],a))}withFaceExpressions(){return new _s(this,this.input)}withAgeAndGender(){return new As(this,this.input)}},Ds=class extends cg{async run(){let t=await this.parentTask;if(!t)return;let n=await dp(t,this.input,a=>tt.faceRecognitionNet.computeFaceDescriptor(a),null,a=>a.landmarks.align(null,{useDlibAlignment:!0}));return b1(t,n)}withFaceExpressions(){return new Es(this,this.input)}withAgeAndGender(){return new $s(this,this.input)}};var dg=class extends Pa{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.useTinyLandmarkNet=r}get landmarkNet(){return this.useTinyLandmarkNet?tt.faceLandmark68TinyNet:tt.faceLandmark68Net}},hg=class extends dg{async run(){let t=await this.parentTask,n=t.map(i=>i.detection),a=this.input instanceof Te?await cd(this.input,n):await pd(this.input,n),r=await Promise.all(a.map(i=>this.landmarkNet.detectLandmarks(i)));return a.forEach(i=>i instanceof Te&&i.dispose()),t.filter((i,o)=>r[o]).map((i,o)=>fd(i,r[o]))}withFaceExpressions(){return new _s(this,this.input)}withAgeAndGender(){return new As(this,this.input)}withFaceDescriptors(){return new Fs(this,this.input)}},mg=class extends dg{async run(){let t=await this.parentTask;if(!t)return;let{detection:n}=t,a=this.input instanceof Te?await cd(this.input,[n]):await pd(this.input,[n]),r=await this.landmarkNet.detectLandmarks(a[0]);return a.forEach(s=>s instanceof Te&&s.dispose()),fd(t,r)}withFaceExpressions(){return new Es(this,this.input)}withAgeAndGender(){return new $s(this,this.input)}withFaceDescriptor(){return new Ds(this,this.input)}};var fg=class extends Pa{constructor(n,a=new Ma){super();this.input=n;this.options=a}},yd=class extends fg{async run(){let{input:t,options:n}=this,a;if(n instanceof lg)a=tt.tinyFaceDetector.locateFaces(t,n);else if(n instanceof Ma)a=tt.ssdMobilenetv1.locateFaces(t,n);else if(n instanceof fr)a=tt.tinyYolov2.locateFaces(t,n);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return a}runAndExtendWithFaceDetections(){return new Promise((t,n)=>{this.run().then(a=>t(a.map(r=>Yu({},r)))).catch(a=>n(a))})}withFaceLandmarks(t=!1){return new hg(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new $o(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new Do(this.runAndExtendWithFaceDetections(),this.input)}},gg=class extends fg{async run(){let t=await new yd(this.input,this.options),n=t[0];return t.forEach(a=>{a.score>n.score&&(n=a)}),n}runAndExtendWithFaceDetection(){return new Promise(async t=>{let n=await this.run();t(n?Yu({},n):void 0)})}withFaceLandmarks(t=!1){return new mg(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new Fo(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new Ro(this.runAndExtendWithFaceDetection(),this.input)}};function DNe(e,t=new Ma){return new gg(e,t)}function k1(e,t=new Ma){return new yd(e,t)}async function Qpe(e,t){return k1(e,new Ma(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function zNe(e,t={}){return k1(e,new fr(t)).withFaceLandmarks().withFaceDescriptors()}var WNe=Qpe;function I$(e,t){if(e.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let n=Array.from(e),a=Array.from(t);return Math.sqrt(n.map((r,s)=>r-a[s]).reduce((r,s)=>r+s*s,0))}var yg=class{constructor(t,n=.6){this._distanceThreshold=n;let a=Array.isArray(t)?t:[t];if(!a.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let r=1,s=()=>`person ${r++}`;this._labeledDescriptors=a.map(i=>{if(i instanceof Mr)return i;if(i instanceof Float32Array)return new Mr(s(),[i]);if(i.descriptor&&i.descriptor instanceof Float32Array)return new Mr(s(),[i.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,n){return n.map(a=>I$(a,t)).reduce((a,r)=>a+r,0)/(n.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:n,label:a})=>new id(a,this.computeMeanDistance(t,n))).reduce((n,a)=>n.distance<a.distance?n:a)}findBestMatch(t){let n=this.matchDescriptor(t);return n.distance<this._distanceThreshold?n:new id("unknown",n.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let n=t.labeledDescriptors.map(a=>Mr.fromJSON(a));return new yg(n,t.distanceThreshold)}};function r2e(e){let t=new hp;return t.extractWeights(e),t}function ece(e,t){let{width:n,height:a}=new En(t.width,t.height);if(n<=0||a<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:n,height:a})}`);if(Array.isArray(e))return e.map(r=>ece(r,{width:n,height:a}));if(ip(e)){let r=e.detection.forSize(n,a),s=e.unshiftedLandmarks.forSize(r.box.width,r.box.height);return fd(Yu(e,r),s)}return Pr(e)?Yu(e,e.detection.forSize(n,a)):e instanceof va||e instanceof xt?e.forSize(n,a):e}var f2e=HA;export{ag as AgeGenderNet,qu as BoundingBox,ut as Box,Pa as ComposableTask,Fs as ComputeAllFaceDescriptorsTask,cg as ComputeFaceDescriptorsTaskBase,Ds as ComputeSingleFaceDescriptorTask,hg as DetectAllFaceLandmarksTask,yd as DetectAllFacesTask,dg as DetectFaceLandmarksTaskBase,fg as DetectFacesTaskBase,mg as DetectSingleFaceLandmarksTask,gg as DetectSingleFaceTask,En as Dimensions,VA as FACE_EXPRESSION_LABELS,xt as FaceDetection,p$ as FaceDetectionNet,Qf as FaceExpressionNet,Cs as FaceExpressions,lp as FaceLandmark68Net,sg as FaceLandmark68TinyNet,QA as FaceLandmarkNet,va as FaceLandmarks,EA as FaceLandmarks5,Xu as FaceLandmarks68,id as FaceMatch,yg as FaceMatcher,up as FaceRecognitionNet,f1 as Gender,od as LabeledBox,Mr as LabeledFaceDescriptors,Lr as NetInput,dn as NeuralNetwork,Ts as ObjectDetection,Me as Point,AA as PredictedBox,Ku as Rect,_o as SsdMobilenetv1,Ma as SsdMobilenetv1Options,hp as TinyFaceDetector,lg as TinyFaceDetectorOptions,cp as TinyYolov2,fr as TinyYolov2Options,WNe as allFaces,Qpe as allFacesSsdMobilenetv1,zNe as allFacesTinyYolov2,$A as awaitMediaLoaded,FA as bufferToImage,STe as computeFaceDescriptor,ep as createCanvas,Hf as createCanvasFromMedia,eSe as createFaceDetectionNet,eIe as createFaceRecognitionNet,qpe as createSsdMobilenetv1,r2e as createTinyFaceDetector,HSe as createTinyYolov2,k1 as detectAllFaces,Zpe as detectFaceLandmarks,ITe as detectFaceLandmarksTiny,PTe as detectLandmarks,DNe as detectSingleFace,GA as draw,et as env,I$ as euclideanDistance,x1 as extendWithAge,b1 as extendWithFaceDescriptor,Yu as extendWithFaceDetection,h1 as extendWithFaceExpressions,fd as extendWithFaceLandmarks,v1 as extendWithGender,cd as extractFaceTensors,pd as extractFaces,pve as fetchImage,MA as fetchJson,fve as fetchNetWeights,Ns as fetchOrThrow,wve as fetchVideo,aa as getContext2dOrThrow,Qu as getMediaDimensions,DA as imageTensorToCanvas,RA as imageToSquare,Tye as inverseSigmoid,TA as iou,d1 as isMediaElement,Gf as isMediaLoaded,rIe as isWithAge,Pr as isWithFaceDetection,UA as isWithFaceExpressions,ip as isWithFaceLandmarks,lIe as isWithGender,DTe as loadAgeGenderModel,RTe as loadFaceDetectionModel,FTe as loadFaceExpressionModel,ETe as loadFaceLandmarkModel,ATe as loadFaceLandmarkTinyModel,$Te as loadFaceRecognitionModel,Jpe as loadSsdMobilenetv1Model,CTe as loadTinyFaceDetectorModel,_Te as loadTinyYolov2Model,OA as loadWeightMap,MTe as locateFaces,_ve as matchDimensions,NA as minBbox,tt as nets,CA as nonMaxSuppression,mr as normalize,_A as padToSquare,NTe as predictAgeAndGender,TTe as recognizeFaceExpressions,ece as resizeResults,Zu as resolveInput,Iye as shuffleArray,Bf as sigmoid,Ype as ssdMobilenetv1,Oe as tf,wTe as tinyFaceDetector,kTe as tinyYolov2,vt as toNetInput,SA as utils,y$ as validateConfig,f2e as version};
+    `;
+  }
+};
+function unsortedSegmentSum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, segmentIds } = inputs;
+  const { numSegments } = attrs;
+  const xRank = x.shape.length;
+  const toDispose = [];
+  let axis = 0;
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+    toDispose.push(permutedX);
+    axis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  }
+  const outShape = backend_util_exports.segment_util.computeOutShape(permutedX.shape, axis, numSegments);
+  const inSize = util_exports.sizeFromShape([permutedX.shape[axis]]);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  toDispose.push(a2D);
+  const outputDType = sumOutType(x.dtype);
+  const segOpCompute = (x2, segOpType, segmentIds2, dtype, numSegments2) => {
+    const batchSize = x2.shape[0];
+    const inSize2 = x2.shape[1];
+    const windowSize = backend_util_exports.segment_util.segOpComputeOptimalWindowSize(inSize2, numSegments2);
+    const segOpInfo = { windowSize, inSize: inSize2, batchSize, numSegments: numSegments2 };
+    const program = new SegmentOpProgram(segOpInfo, segOpType);
+    const output = backend2.compileAndRun(program, [x2, segmentIds2], dtype);
+    toDispose.push(output);
+    if (output.shape[1] === numSegments2) {
+      return output;
+    }
+    const rangeInfo = range4({
+      backend: backend2,
+      attrs: { start: 0, stop: numSegments2, step: 1, dtype: "float32" }
+    });
+    const tileInfo = tile4({
+      inputs: { x: rangeInfo },
+      backend: backend2,
+      attrs: { reps: [inSize2 / windowSize] }
+    });
+    toDispose.push(rangeInfo);
+    toDispose.push(tileInfo);
+    const result2 = segOpCompute(output, segOpType, tileInfo, dtype, numSegments2);
+    return result2;
+  };
+  const segOpResult = segOpCompute(a2D, "unsortedSegmentSum", segmentIds, outputDType, numSegments);
+  const reshaped = reshape4({ inputs: { x: segOpResult }, backend: backend2, attrs: { shape: outShape } });
+  let result = reshaped;
+  if (permutation != null) {
+    toDispose.push(reshaped);
+    const perm = backend_util_exports.getUndoAxesPermutation(permutation);
+    result = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm } });
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var unsortedSegmentSumConfig2 = {
+  kernelName: UnsortedSegmentSum,
+  backendName: "webgl",
+  kernelFunc: unsortedSegmentSum3
+};
+var kernelConfigs2 = [
+  _fusedMatMulConfig2,
+  absConfig2,
+  acosConfig2,
+  acoshConfig2,
+  addConfig2,
+  addNConfig2,
+  allConfig2,
+  anyConfig2,
+  argMaxConfig2,
+  argMinConfig2,
+  asinConfig2,
+  asinhConfig2,
+  atanConfig2,
+  atan2Config2,
+  atanhConfig2,
+  avgPoolConfig2,
+  avgPool3DConfig2,
+  avgPool3DGradConfig3,
+  avgPoolGradConfig3,
+  batchMatMulConfig2,
+  batchNormConfig2,
+  batchToSpaceNDConfig2,
+  bincountConfig2,
+  broadcastArgsConfig2,
+  castConfig2,
+  ceilConfig2,
+  clipByValueConfig2,
+  complexConfig2,
+  complexAbsConfig2,
+  concatConfig2,
+  conv2DConfig2,
+  conv2DBackpropFilterConfig2,
+  conv2DBackpropInputConfig2,
+  conv3DConfig2,
+  conv3DBackpropFilterV2Config2,
+  conv3DBackpropInputConfig,
+  cosConfig2,
+  coshConfig2,
+  cropAndResizeConfig2,
+  cumprodConfig2,
+  cumsumConfig2,
+  denseBincountConfig2,
+  depthToSpaceConfig2,
+  depthwiseConv2dNativeConfig2,
+  depthwiseConv2dNativeBackpropFilterConfig2,
+  depthwiseConv2dNativeBackpropInputConfig2,
+  diagConfig2,
+  dilation2DConfig2,
+  einsumConfig2,
+  eluConfig2,
+  eluGradConfig3,
+  equalConfig2,
+  erfConfig2,
+  expConfig2,
+  expandDimsConfig2,
+  expm1Config2,
+  fftConfig2,
+  fillConfig2,
+  flipLeftRightConfig2,
+  floorConfig2,
+  floorDivConfig2,
+  fromPixelsConfig,
+  fusedConv2DConfig2,
+  fusedDepthwiseConv2DConfig2,
+  gatherNdConfig2,
+  gatherV2Config2,
+  greaterConfig2,
+  greaterEqualConfig2,
+  identityConfig2,
+  ifftConfig2,
+  imagConfig2,
+  isFiniteConfig2,
+  isInfConfig2,
+  isNaNConfig2,
+  leakyReluConfig2,
+  lessConfig2,
+  lessEqualConfig2,
+  linSpaceConfig2,
+  logConfig2,
+  log1pConfig2,
+  logicalAndConfig2,
+  logicalNotConfig2,
+  logicalOrConfig2,
+  LRNConfig2,
+  LRNGradConfig2,
+  maxConfig2,
+  maximumConfig2,
+  maxPoolConfig2,
+  maxPool3DConfig2,
+  maxPool3DGradConfig3,
+  maxPoolGradConfig3,
+  maxPoolWithArgmaxConfig2,
+  meanConfig2,
+  minConfig2,
+  minimumConfig2,
+  mirrorPadConfig2,
+  modConfig2,
+  multinomialConfig2,
+  multiplyConfig2,
+  negConfig2,
+  nonMaxSuppressionV3Config2,
+  nonMaxSuppressionV4Config2,
+  nonMaxSuppressionV5Config2,
+  notEqualConfig2,
+  oneHotConfig2,
+  onesLikeConfig2,
+  packConfig2,
+  padV2Config2,
+  powConfig2,
+  preluConfig2,
+  prodConfig2,
+  raggedGatherConfig2,
+  raggedRangeConfig2,
+  raggedTensorToTensorConfig2,
+  rangeConfig2,
+  realConfig2,
+  realDivConfig2,
+  reciprocalConfig2,
+  reluConfig2,
+  relu6Config2,
+  reshapeConfig2,
+  resizeBilinearConfig2,
+  resizeBilinearGradConfig3,
+  resizeNearestNeighborConfig2,
+  resizeNearestNeighborGradConfig3,
+  reverseConfig2,
+  rotateWithOffsetConfig2,
+  roundConfig2,
+  rsqrtConfig2,
+  scatterNdConfig2,
+  searchSortedConfig2,
+  selectConfig2,
+  seluConfig2,
+  sigmoidConfig2,
+  signConfig2,
+  sinConfig2,
+  sinhConfig2,
+  sliceConfig2,
+  softmaxConfig2,
+  softplusConfig2,
+  spaceToBatchNDConfig2,
+  sparseFillEmptyRowsConfig2,
+  sparseReshapeConfig2,
+  sparseSegmentMeanConfig2,
+  sparseSegmentSumConfig2,
+  sparseToDenseConfig2,
+  splitVConfig2,
+  sqrtConfig2,
+  squareConfig2,
+  squaredDifferenceConfig2,
+  stepConfig2,
+  stridedSliceConfig2,
+  stringNGramsConfig2,
+  stringSplitConfig2,
+  stringToHashBucketFastConfig2,
+  subConfig2,
+  sumConfig2,
+  tanConfig2,
+  tanhConfig2,
+  tileConfig2,
+  topKConfig2,
+  transformConfig2,
+  transposeConfig2,
+  uniqueConfig2,
+  unpackConfig2,
+  unsortedSegmentSumConfig2,
+  zerosLikeConfig2
+];
+for (const kernelConfig of kernelConfigs2) {
+  registerKernel(kernelConfig);
+}
+var CppDType;
+(function(CppDType2) {
+  CppDType2[CppDType2["float32"] = 0] = "float32";
+  CppDType2[CppDType2["int32"] = 1] = "int32";
+  CppDType2[CppDType2["bool"] = 2] = "bool";
+  CppDType2[CppDType2["string"] = 3] = "string";
+  CppDType2[CppDType2["complex64"] = 4] = "complex64";
+})(CppDType || (CppDType = {}));
+var FusableActivation;
+(function(FusableActivation2) {
+  FusableActivation2[FusableActivation2["linear"] = 0] = "linear";
+  FusableActivation2[FusableActivation2["relu"] = 1] = "relu";
+  FusableActivation2[FusableActivation2["relu6"] = 2] = "relu6";
+  FusableActivation2[FusableActivation2["prelu"] = 3] = "prelu";
+  FusableActivation2[FusableActivation2["leakyrelu"] = 4] = "leakyrelu";
+  FusableActivation2[FusableActivation2["sigmoid"] = 5] = "sigmoid";
+  FusableActivation2[FusableActivation2["elu"] = 6] = "elu";
+})(FusableActivation || (FusableActivation = {}));
+var wasmFusedMatMul;
+function setup(backend2) {
+  wasmFusedMatMul = backend2.wasm.cwrap(_FusedMatMul, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedBatchMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  if (a.dtype !== "float32" || b.dtype !== "float32") {
+    throw new Error(`_FusedMatMul for non non-float32 tensors not yet supported.`);
+  }
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  const aId = backend2.dataIdMap.get(a.dataId).id;
+  const bId = backend2.dataIdMap.get(b.dataId).id;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    biasId = biasData.id;
+  }
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);
+  }
+  const leftDim = transposeA ? a.shape[2] : a.shape[1];
+  const rightDim = transposeB ? b.shape[1] : b.shape[2];
+  const batchDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const out = backend2.makeOutput([...batchDims, leftDim, rightDim], a.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);
+  const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);
+  wasmFusedMatMul(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, transposeA, transposeB, fusedActivation, biasId, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var _fusedMatMulConfig3 = {
+  kernelName: _FusedMatMul,
+  backendName: "wasm",
+  setupFunc: setup,
+  kernelFunc: fusedBatchMatMul
+};
+function createUnaryKernelConfig(kernelName, outType) {
+  let wasmFunc9;
+  function setupFunc3(backend2) {
+    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [
+      "number",
+      "number",
+      "number"
+    ]);
+  }
+  function kernelFunc3(args) {
+    const { backend: backend2, inputs: { x } } = args;
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const out = backend2.makeOutput(x.shape, outType || x.dtype);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    if (util_exports.sizeFromShape(out.shape) === 0) {
+      return out;
+    }
+    wasmFunc9(xId, CppDType[x.dtype], outId);
+    return out;
+  }
+  return { kernelName, backendName: "wasm", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };
+}
+var absConfig3 = createUnaryKernelConfig(Abs);
+function createBinaryKernelConfig(kernelName, supportsFullBroadcast19, dtype) {
+  let wasmFunc9;
+  function setupFunc3(backend2) {
+    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [
+      "number",
+      "array",
+      "number",
+      "number",
+      "array",
+      "number",
+      "number",
+      "number"
+    ]);
+  }
+  function kernelFunc3(args) {
+    const { backend: backend2, inputs } = args;
+    const { a, b } = inputs;
+    const aId = backend2.dataIdMap.get(a.dataId).id;
+    const bId = backend2.dataIdMap.get(b.dataId).id;
+    const outputType = dtype != null ? dtype : a.dtype;
+    const newShape = backend_util_exports.assertAndGetBroadcastShape(a.shape, b.shape);
+    const out = backend2.makeOutput(newShape, outputType);
+    if (util_exports.sizeFromShape(newShape) === 0) {
+      return out;
+    }
+    const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);
+    const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    const kernelFunc4 = () => wasmFunc9(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, CppDType[a.dtype], outId);
+    kernelFunc4();
+    return out;
+  }
+  return { kernelName, backendName: "wasm", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };
+}
+var supportsFullBroadcast = true;
+var addConfig3 = createBinaryKernelConfig(Add, supportsFullBroadcast);
+var wasmFunc;
+function setupFunc(backend2) {
+  wasmFunc = backend2.wasm.cwrap(AddN, null, [
+    "array",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function addn(args) {
+  const { inputs, backend: backend2 } = args;
+  const out = backend2.makeOutput(inputs[0].shape, inputs[0].dtype);
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  const inputIds = inputs.map((x) => backend2.dataIdMap.get(x.dataId).id);
+  const inputIdsBytes = new Uint8Array(new Int32Array(inputIds).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmFunc(inputIdsBytes, inputIds.length, CppDType[out.dtype], outId);
+  return out;
+}
+var addNConfig3 = {
+  kernelName: AddN,
+  backendName: "wasm",
+  setupFunc,
+  kernelFunc: addn
+};
+function identity4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  if (x.dtype === "string") {
+    return tensor(backend2.readSync(x.dataId), x.shape, x.dtype);
+  }
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const inVals = backend2.typedArrayFromHeap(x);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(inVals);
+  return out;
+}
+var identityConfig3 = {
+  kernelName: Identity,
+  backendName: "wasm",
+  kernelFunc: identity4
+};
+var wasmTranspose;
+function setup2(backend2) {
+  wasmTranspose = backend2.wasm.cwrap(Transpose, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function transpose4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const [reducedShape, perm] = removeOneSizeDims(inputs.x.shape, attrs.perm);
+  let permIsNoOp = true;
+  for (let i = 0; i < perm.length; i++) {
+    if (perm[i] !== i) {
+      permIsNoOp = false;
+    }
+  }
+  const outShape = computeOutShape4(inputs.x.shape, attrs.perm);
+  const x = {
+    dataId: inputs.x.dataId,
+    shape: reducedShape,
+    dtype: inputs.x.dtype
+  };
+  if (permIsNoOp) {
+    const cloned = identity4({ inputs, backend: backend2 });
+    cloned.shape = outShape;
+    return cloned;
+  }
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const permBytes = new Uint8Array(new Int32Array(perm).buffer);
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  wasmTranspose(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], outId, permBytes, perm.length);
+  return out;
+}
+function computeOutShape4(inShape, perm) {
+  const outShape = new Array(inShape.length);
+  for (let i = 0; i < outShape.length; i++) {
+    outShape[i] = inShape[perm[i]];
+  }
+  return outShape;
+}
+function removeOneSizeDims(shape, perm) {
+  const newShape = [];
+  const newPerm = [];
+  for (let i = 0; i < shape.length; ++i) {
+    if (shape[i] !== 1) {
+      newShape.push(shape[i]);
+    }
+    if (shape[perm[i]] !== 1) {
+      newPerm.push(perm[i]);
+    }
+  }
+  for (let i = 0; i < newPerm.length; ++i) {
+    let minValIdx = -1;
+    for (let j = 0; j < newPerm.length; ++j) {
+      if (newPerm[j] >= i && (minValIdx === -1 || newPerm[minValIdx] > newPerm[j])) {
+        minValIdx = j;
+      }
+    }
+    newPerm[minValIdx] = i;
+  }
+  return [newShape, newPerm];
+}
+var transposeConfig3 = {
+  kernelName: Transpose,
+  backendName: "wasm",
+  kernelFunc: transpose4,
+  setupFunc: setup2
+};
+function permuteAxesAndTranspose(x, axis, backend2) {
+  const xShape = x.shape;
+  const xRank = x.shape.length;
+  const originalAxes = util_exports.parseAxisParam(axis, xShape);
+  let axes = originalAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let xTransposed = null;
+  let inputWasTransposed = false;
+  if (permutedAxes != null) {
+    const newShape = new Array(xRank);
+    for (let i = 0; i < newShape.length; i++) {
+      newShape[i] = xShape[permutedAxes[i]];
+    }
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    xTransposed = transpose4({ inputs: { x }, attrs: { perm: permutedAxes }, backend: backend2 });
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const transposedId = backend2.dataIdMap.get(xTransposed.dataId).id;
+    if (transposedId !== xId) {
+      inputWasTransposed = true;
+    }
+  }
+  return { transposed: xTransposed, originalAxes, axes, inputWasTransposed };
+}
+var wasmAll;
+function setup3(backend2) {
+  wasmAll = backend2.wasm.cwrap(All, null, ["number, number, number"]);
+}
+function all4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmAll(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var allConfig3 = {
+  kernelName: All,
+  backendName: "wasm",
+  setupFunc: setup3,
+  kernelFunc: all4
+};
+var wasmAny;
+function setup4(backend2) {
+  wasmAny = backend2.wasm.cwrap(Any, null, ["number, number, number"]);
+}
+function any4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmAny(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var anyConfig3 = {
+  kernelName: Any,
+  backendName: "wasm",
+  setupFunc: setup4,
+  kernelFunc: any4
+};
+var wasmFunc2;
+function setup5(backend2) {
+  wasmFunc2 = backend2.wasm.cwrap(ArgMax, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function argmax(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+    }
+  }
+  const outShape = input2.shape.slice(0, -1);
+  const out = backend2.makeOutput(outShape, "int32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const outerSize = util_exports.sizeFromShape(out.shape);
+  const innerSize = input2.shape[axes[0]];
+  wasmFunc2(inputId, CppDType[input2.dtype], outerSize, innerSize, outId);
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  return out;
+}
+var argMaxConfig3 = {
+  kernelName: ArgMax,
+  backendName: "wasm",
+  kernelFunc: argmax,
+  setupFunc: setup5
+};
+var wasmAvgPool;
+function setup6(backend2) {
+  wasmAvgPool = backend2.wasm.cwrap(AvgPool, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function avgPool4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const x = inputs.x;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const channels = convInfo.inChannels;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  if (convInfo.dilationWidth !== 1 || convInfo.dilationHeight !== 1) {
+    throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${convInfo.dilationHeight}, ${convInfo.dilationWidth}].`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmAvgPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, strideHeight, strideWidth, channels, outId);
+  return out;
+}
+var avgPoolConfig3 = {
+  kernelName: AvgPool,
+  backendName: "wasm",
+  setupFunc: setup6,
+  kernelFunc: avgPool4
+};
+function reshape5(args) {
+  const { inputs, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  util_exports.assert(xSize === util_exports.sizeFromShape($shape), () => `new shape: ${$shape}, old shape: ${x.shape}. New shape and old shape must have the same number of elements.`);
+  args.backend.incRef(x.dataId);
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig3 = {
+  kernelName: Reshape,
+  backendName: "wasm",
+  kernelFunc: reshape5
+};
+var wasmBatchMatMul;
+function setup7(backend2) {
+  wasmBatchMatMul = backend2.wasm.cwrap(BatchMatMul, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function batchMatMul3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  if (a.dtype !== "float32" || b.dtype !== "float32") {
+    throw new Error(`BatchMatMul for non non-float32 tensors not yet supported.`);
+  }
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape5({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape5({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const a3dId = backend2.dataIdMap.get(a3d.dataId).id;
+  const b3dId = backend2.dataIdMap.get(b3d.dataId).id;
+  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];
+  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const out = backend2.makeOutput([batchDim, leftDim, rightDim], a3d.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const aShapeBytes = new Uint8Array(new Int32Array(a3d.shape).buffer);
+  const bShapeBytes = new Uint8Array(new Int32Array(b3d.shape).buffer);
+  wasmBatchMatMul(a3dId, aShapeBytes, a3d.shape.length, b3dId, bShapeBytes, b3d.shape.length, transposeA, transposeB, outId);
+  backend2.disposeData(a3d.dataId);
+  backend2.disposeData(b3d.dataId);
+  out.shape = outShape;
+  return out;
+}
+var batchMatMulConfig3 = {
+  kernelName: BatchMatMul,
+  backendName: "wasm",
+  setupFunc: setup7,
+  kernelFunc: batchMatMul3
+};
+function slice4(args) {
+  const { inputs: { x }, attrs: { begin, size }, backend: backend2 } = args;
+  const [begin_, size_] = slice_util_exports.parseSliceParams(x, begin, size);
+  const isContinous = slice_util_exports.isSliceContinous(x.shape, begin_, size_);
+  const xVals = backend2.readSync(x.dataId);
+  const out = backend2.makeOutput(size_, x.dtype);
+  const xStrides = util_exports.computeStrides(x.shape);
+  const outData = backend2.dataIdMap.get(out.dataId);
+  if (isContinous) {
+    const flatOffset = slice_util_exports.computeFlatOffset(begin_, xStrides);
+    if (x.dtype === "string") {
+      outData.stringBytes = xVals.slice(flatOffset, flatOffset + util_exports.sizeFromShape(size_));
+    } else {
+      const outVals2 = backend2.typedArrayFromHeap(out);
+      outVals2.set(xVals.subarray(flatOffset, flatOffset + util_exports.sizeFromShape(size_)));
+    }
+    return out;
+  }
+  if (x.dtype === "string") {
+    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);
+    outData.stringBytes = res;
+    return out;
+  }
+  const outVals = backend2.typedArrayFromHeap(out);
+  const rank = x.shape.length;
+  if (rank === 2) {
+    slice2d2(xVals, xStrides[0], outVals, begin_, size_);
+  } else if (rank === 3) {
+    slice3d2(xVals, xStrides[0], xStrides[1], outVals, begin_, size_);
+  } else if (rank === 4) {
+    slice4d2(xVals, xStrides[0], xStrides[1], xStrides[2], outVals, begin_, size_);
+  } else {
+    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);
+    outVals.set(res);
+  }
+  return out;
+}
+function slice2d2(xVals, xStride, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const endI = beginI + size[0];
+  for (let i = beginI; i < endI; i++) {
+    const xOffset = i * xStride + beginJ;
+    outVals.set(xVals.subarray(xOffset, xOffset + size[1]), outOffset);
+    outOffset += size[1];
+  }
+}
+function slice3d2(xVals, xStride1, xStride2, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const beginK = begin[2];
+  const endI = beginI + size[0];
+  const endJ = beginJ + size[1];
+  for (let i = beginI; i < endI; i++) {
+    for (let j = beginJ; j < endJ; j++) {
+      const xOffset = i * xStride1 + j * xStride2 + beginK;
+      outVals.set(xVals.subarray(xOffset, xOffset + size[2]), outOffset);
+      outOffset += size[2];
+    }
+  }
+}
+function slice4d2(xVals, xStride1, xStride2, xStride3, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const beginK = begin[2];
+  const endI = beginI + size[0];
+  const endJ = beginJ + size[1];
+  const endK = beginK + size[2];
+  const beginL = begin[3];
+  for (let i = beginI; i < endI; i++) {
+    for (let j = beginJ; j < endJ; j++) {
+      for (let k = beginK; k < endK; k++) {
+        const xOffset = i * xStride1 + j * xStride2 + k * xStride3 + beginL;
+        outVals.set(xVals.subarray(xOffset, xOffset + size[3]), outOffset);
+        outOffset += size[3];
+      }
+    }
+  }
+}
+var sliceConfig3 = {
+  kernelName: Slice,
+  backendName: "wasm",
+  kernelFunc: slice4
+};
+function batchToSpaceND4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const xReshaped = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const xTransposed = transpose4({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });
+  const xTransposedReshaped = reshape5({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });
+  const result = slice4({
+    inputs: { x: xTransposedReshaped },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  backend2.disposeData(xReshaped.dataId);
+  backend2.disposeData(xTransposed.dataId);
+  backend2.disposeData(xReshaped.dataId);
+  return result;
+}
+var batchToSpaceNDConfig3 = {
+  kernelName: BatchToSpaceND,
+  backendName: "wasm",
+  kernelFunc: batchToSpaceND4
+};
+function cast5(args) {
+  const { inputs: { x }, attrs: { dtype }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, dtype);
+  const inVals = backend2.typedArrayFromHeap(x);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(inVals);
+  return out;
+}
+var castConfig3 = {
+  kernelName: Cast,
+  backendName: "wasm",
+  kernelFunc: cast5
+};
+var ceilConfig3 = createUnaryKernelConfig(Ceil);
+var wasmClip;
+function setup8(backend2) {
+  wasmClip = backend2.wasm.cwrap(ClipByValue, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function clip(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { clipValueMin, clipValueMax } = attrs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmClip(xId, clipValueMin, clipValueMax, outId);
+  return out;
+}
+var clipByValueConfig3 = {
+  kernelName: ClipByValue,
+  backendName: "wasm",
+  setupFunc: setup8,
+  kernelFunc: clip
+};
+function concat4(args) {
+  const { inputs, backend: backend2 } = args;
+  const axis = util_exports.parseAxisParam(args.attrs.axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, axis);
+  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity4({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  const out = backend2.makeOutput(outShape, inputs[0].dtype);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return out;
+  }
+  if ($inputs[0].dtype === "string") {
+    const inputs2D = $inputs.map((t) => {
+      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));
+      const shape = [-1, innerSize];
+      return reshape5({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+    });
+    const inputsValShapes = inputs2D.map((t) => {
+      return { vals: backend2.readSync(t.dataId), shape: t.shape };
+    });
+    outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);
+    const simplyConcat = inputs2D[0].shape[0] === 1;
+    const outVals2 = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);
+    const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), axis);
+    out.shape = finalOutShape;
+    const outData = backend2.dataIdMap.get(out.dataId);
+    outData.stringBytes = backend_util_exports.fromStringArrayToUint8(outVals2);
+    inputs2D.forEach((t) => backend2.disposeData(t.dataId));
+    return out;
+  }
+  const batchDim = util_exports.sizeFromShape($inputs[0].shape.slice(0, axis));
+  let sumInnerDims = 0;
+  const innerDims = $inputs.map((input2) => {
+    const innerDim = util_exports.sizeFromShape(input2.shape.slice(axis));
+    sumInnerDims += innerDim;
+    return innerDim;
+  });
+  const inVals = $inputs.map((input2) => backend2.typedArrayFromHeap(input2));
+  const outVals = backend2.typedArrayFromHeap(out);
+  for (let b = 0; b < batchDim; b++) {
+    let outOffset = b * sumInnerDims;
+    for (let i = 0; i < inVals.length; i++) {
+      const innerDim = innerDims[i];
+      const inOffset = b * innerDim;
+      const vals = inVals[i].subarray(inOffset, inOffset + innerDim);
+      outVals.set(vals, outOffset);
+      outOffset += innerDim;
+    }
+  }
+  return out;
+}
+var concatConfig3 = {
+  kernelName: Concat,
+  backendName: "wasm",
+  kernelFunc: concat4
+};
+var wasmConv2d;
+function setup9(backend2) {
+  wasmConv2d = backend2.wasm.cwrap(Conv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function conv2d5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const { strides, dilations, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend Conv2D does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var conv2DConfig3 = {
+  kernelName: Conv2D,
+  backendName: "wasm",
+  setupFunc: setup9,
+  kernelFunc: conv2d5
+};
+var wasmConv2DBackpropInput;
+function setup10(backend2) {
+  wasmConv2DBackpropInput = backend2.wasm.cwrap(Conv2DBackpropInput, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function conv2DBackpropInput4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape } = attrs;
+  const dilations = 1;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const dxStrides = util_exports.computeStrides(convInfo.inShape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const [fltS0, fltS1, fltS2] = util_exports.computeStrides(filter.shape);
+  const xBatchStride = dxStrides[0];
+  const xRowStride = isChannelsLast ? dxStrides[1] : dxStrides[2];
+  const xColStride = isChannelsLast ? dxStrides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : dxStrides[1];
+  const yBatchStride = dyStrides[0];
+  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];
+  const yColStride = isChannelsLast ? dyStrides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];
+  const out = backend2.makeOutput(convInfo.inShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const dyId = backend2.dataIdMap.get(dy.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  wasmConv2DBackpropInput(dyId, filterId, batchSize, filterHeight, filterWidth, inHeight, inWidth, inChannels, outHeight, outWidth, outChannels, strideHeight, strideWidth, topPad, leftPad, fltS0, fltS1, fltS2, xBatchStride, xRowStride, xColStride, xChannelStride, yBatchStride, yRowStride, yColStride, yChannelStride, outId);
+  return out;
+}
+var conv2DBackpropInputConfig3 = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "wasm",
+  setupFunc: setup10,
+  kernelFunc: conv2DBackpropInput4
+};
+var cosConfig3 = createUnaryKernelConfig(Cos);
+var coshConfig3 = createUnaryKernelConfig(Cosh);
+var InterpolationMethod;
+(function(InterpolationMethod2) {
+  InterpolationMethod2[InterpolationMethod2["bilinear"] = 0] = "bilinear";
+  InterpolationMethod2[InterpolationMethod2["nearest"] = 1] = "nearest";
+})(InterpolationMethod || (InterpolationMethod = {}));
+var wasmCropAndResize;
+function setup11(backend2) {
+  wasmCropAndResize = backend2.wasm.cwrap(CropAndResize, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cropAndResize4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { method, extrapolationValue, cropSize } = attrs;
+  const { image: image2, boxes, boxInd } = inputs;
+  const numBoxes = boxes.shape[0];
+  const [cropHeight, cropWidth] = cropSize;
+  const outShape = [numBoxes, cropHeight, cropWidth, image2.shape[3]];
+  let imagesData = backend2.dataIdMap.get(image2.dataId);
+  let castedData;
+  if (image2.dtype !== "float32") {
+    castedData = cast5({ backend: backend2, inputs: { x: image2 }, attrs: { dtype: "float32" } });
+    imagesData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const imagesId = imagesData.id;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const boxIndId = backend2.dataIdMap.get(boxInd.dataId).id;
+  const out = backend2.makeOutput(outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const imagesShapeBytes = new Uint8Array(new Int32Array(image2.shape).buffer);
+  wasmCropAndResize(imagesId, boxesId, boxIndId, numBoxes, imagesShapeBytes, cropHeight, cropWidth, InterpolationMethod[method], extrapolationValue, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var cropAndResizeConfig3 = {
+  kernelName: CropAndResize,
+  backendName: "wasm",
+  setupFunc: setup11,
+  kernelFunc: cropAndResize4
+};
+var wasmCumprod;
+function setup12(backend2) {
+  wasmCumprod = backend2.wasm.cwrap(Cumprod, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cumprod4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  const xRank = x.shape.length;
+  util_exports.assert(x.dtype === "float32" || x.dtype === "int32", () => `cumprod does not support ${x.dtype} tensors in the WASM backend`);
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation !== null) {
+    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  backend_util_exports.assertAxesAreInnerMostDims("cumprod", [permutedAxis], xRank);
+  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);
+  const finalDim = permutedX.shape[permutedAxis];
+  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;
+  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;
+  wasmCumprod(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);
+  let out = permutedOut;
+  if (permutation !== null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });
+    backend2.disposeData(permutedX.dataId);
+    backend2.disposeData(permutedOut.dataId);
+  }
+  return out;
+}
+var cumprodConfig3 = {
+  kernelName: Cumprod,
+  backendName: "wasm",
+  setupFunc: setup12,
+  kernelFunc: cumprod4
+};
+var wasmCumsum;
+function setup13(backend2) {
+  wasmCumsum = backend2.wasm.cwrap(Cumsum, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cumsum4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  const xRank = x.shape.length;
+  util_exports.assert(x.dtype === "float32" || x.dtype === "int32", () => `cumsum does not support ${x.dtype} tensors in the WASM backend`);
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation !== null) {
+    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  backend_util_exports.assertAxesAreInnerMostDims("cumsum", [permutedAxis], xRank);
+  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);
+  const finalDim = permutedX.shape[permutedAxis];
+  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;
+  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;
+  wasmCumsum(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);
+  let out = permutedOut;
+  if (permutation !== null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });
+    backend2.disposeData(permutedX.dataId);
+    backend2.disposeData(permutedOut.dataId);
+  }
+  return out;
+}
+var cumsumConfig3 = {
+  kernelName: Cumsum,
+  backendName: "wasm",
+  setupFunc: setup13,
+  kernelFunc: cumsum4
+};
+var wasmDepthToSpace;
+function setup14(backend2) {
+  wasmDepthToSpace = backend2.wasm.cwrap(DepthToSpace, null, [
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function depthToSpace4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  const batchSize = x.shape[0];
+  const inputHeight = dataFormat === "NHWC" ? x.shape[1] : x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? x.shape[2] : x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? x.shape[3] : x.shape[1];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const outputShape = dataFormat === "NHWC" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];
+  const out = backend2.makeOutput(outputShape, "float32");
+  const xData = backend2.dataIdMap.get(x.dataId);
+  const xId = xData.id;
+  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);
+  const outputShapeBytes = new Uint8Array(new Int32Array(outputShape).buffer);
+  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(outputShape)).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const channelsLast = dataFormat === "NHWC" ? 1 : 0;
+  wasmDepthToSpace(xId, blockSize, channelsLast, xStridesBytes, x.shape.length - 1, outputShapeBytes, outStridesBytes, outputShape.length, outId);
+  return out;
+}
+var depthToSpaceConfig3 = {
+  kernelName: DepthToSpace,
+  backendName: "wasm",
+  setupFunc: setup14,
+  kernelFunc: depthToSpace4
+};
+var wasmDepthwiseConv2d;
+function setup15(backend2) {
+  wasmDepthwiseConv2d = backend2.wasm.cwrap(DepthwiseConv2dNative, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function depthwiseConv2d5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const { strides, dilations, pad: pad3, dimRoundingMode } = attrs;
+  const $dilations = dilations == null ? [1, 1] : dilations;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmDepthwiseConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var depthwiseConv2dNativeConfig3 = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "wasm",
+  setupFunc: setup15,
+  kernelFunc: depthwiseConv2d5
+};
+var eluConfig3 = createUnaryKernelConfig(Elu);
+var supportsFullBroadcast2 = false;
+var equalConfig3 = createBinaryKernelConfig(Equal, supportsFullBroadcast2, "bool");
+var expConfig3 = createUnaryKernelConfig(Exp, "float32");
+function expandDims5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const { dim } = attrs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape5({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig3 = {
+  kernelName: ExpandDims,
+  backendName: "wasm",
+  kernelFunc: expandDims5
+};
+function fill4(args) {
+  const { attrs: { shape, value, dtype }, backend: backend2 } = args;
+  const out = backend2.makeOutput(shape, dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(value);
+  return out;
+}
+var fillConfig3 = {
+  kernelName: Fill,
+  backendName: "wasm",
+  kernelFunc: fill4
+};
+var wasmFlipLeftRight;
+function setup16(backend2) {
+  wasmFlipLeftRight = backend2.wasm.cwrap(FlipLeftRight, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function flipLeftRight2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { image: image2 } = inputs;
+  const out = backend2.makeOutput(image2.shape, image2.dtype);
+  const imageId = backend2.dataIdMap.get(image2.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  wasmFlipLeftRight(imageId, batch, imageHeight, imageWidth, numChannels, outId);
+  return out;
+}
+var flipLeftRightConfig3 = {
+  kernelName: FlipLeftRight,
+  backendName: "wasm",
+  kernelFunc: flipLeftRight2,
+  setupFunc: setup16
+};
+var floorConfig3 = createUnaryKernelConfig(Floor);
+var supportsFullBroadcast3 = false;
+var floorDivConfig3 = createBinaryKernelConfig(FloorDiv, supportsFullBroadcast3);
+var wasmBatchNorm;
+function setup17(backend2) {
+  wasmBatchNorm = backend2.wasm.cwrap(FusedBatchNorm, null, ["number", "number", "number", "number", "number", "number", "number"]);
+}
+function fusedBatchNorm(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { varianceEpsilon } = attrs;
+  const { x, mean: mean4, variance, offset, scale: scale22 } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const meanId = backend2.dataIdMap.get(mean4.dataId).id;
+  const varianceId = backend2.dataIdMap.get(variance.dataId).id;
+  const offsetId = offset != null ? backend2.dataIdMap.get(offset.dataId).id : 0;
+  const scaleId = scale22 != null ? backend2.dataIdMap.get(scale22.dataId).id : 0;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return out;
+  }
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmBatchNorm(xId, meanId, varianceId, offsetId, scaleId, varianceEpsilon, outId);
+  return out;
+}
+var fusedBatchNormConfig = {
+  kernelName: FusedBatchNorm,
+  backendName: "wasm",
+  setupFunc: setup17,
+  kernelFunc: fusedBatchNorm
+};
+var wasmFusedConv2d;
+function setup18(backend2) {
+  wasmFusedConv2d = backend2.wasm.cwrap(FusedConv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedConv2d2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode);
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const outputChannels = convInfo.outChannels;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    if (biasData.shape[0] !== outputChannels) {
+      throw new Error(`FusedConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);
+    }
+    biasId = biasData.id;
+  }
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  const batchSize = convInfo.batchSize;
+  const inHeight = convInfo.inHeight;
+  const inWidth = convInfo.inWidth;
+  if (dataFormat !== "NHWC") {
+    throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  wasmFusedConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var fusedConv2DConfig3 = {
+  kernelName: FusedConv2D,
+  backendName: "wasm",
+  setupFunc: setup18,
+  kernelFunc: fusedConv2d2
+};
+var wasmFusedDepthwiseConv2d;
+function setup19(backend2) {
+  wasmFusedDepthwiseConv2d = backend2.wasm.cwrap(FusedDepthwiseConv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedDepthwiseConv2d(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const outputChannels = convInfo.outChannels;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    if (biasData.shape[0] !== outputChannels) {
+      throw new Error(`FusedDepthwiseConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);
+    }
+    biasId = biasData.id;
+  }
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  const batchSize = convInfo.batchSize;
+  const inHeight = convInfo.inHeight;
+  const inWidth = convInfo.inWidth;
+  if (dataFormat !== "NHWC") {
+    throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  wasmFusedDepthwiseConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var fusedDepthwiseConv2DConfig3 = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "wasm",
+  setupFunc: setup19,
+  kernelFunc: fusedDepthwiseConv2d
+};
+var wasmGatherNd;
+function setup20(backend2) {
+  wasmGatherNd = backend2.wasm.cwrap(GatherNd, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function gatherNd3(args) {
+  const { backend: backend2, inputs } = args;
+  const { params, indices } = inputs;
+  const [resultShape, numSlices, sliceSize, strides] = gather_nd_util_exports.prepareAndValidate(params, indices);
+  const out = backend2.makeOutput(resultShape, params.dtype);
+  if (numSlices === 0) {
+    return out;
+  }
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const xData = backend2.dataIdMap.get(params.dataId);
+  const xId = xData.id;
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmGatherNd(xId, CppDType[params.dtype], indicesId, numSlices, sliceRank, sliceSize, stridesBytes, outId);
+  return out;
+}
+var gatherNdConfig3 = {
+  kernelName: GatherNd,
+  backendName: "wasm",
+  setupFunc: setup20,
+  kernelFunc: gatherNd3
+};
+var wasmGather;
+function setup21(backend2) {
+  wasmGather = backend2.wasm.cwrap("Gather", null, [
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function gatherV23(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const indicesVals = backend2.readSync(indices.dataId);
+  const axisDim = x.shape[parsedAxis];
+  for (let i = 0; i < indicesVals.length; ++i) {
+    const index = indicesVals[i];
+    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+  }
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);
+  const flattenX = reshape5({
+    inputs: { x },
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    },
+    backend: backend2
+  });
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const flattenIndex = reshape5({
+    inputs: { x: indices },
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] },
+    backend: backend2
+  });
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  const out = backend2.makeOutput(flattenOutputShape, x.dtype);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return out;
+  }
+  const stridesSize = flattenX.shape.length - 1;
+  const xData = backend2.dataIdMap.get(flattenX.dataId);
+  const xId = xData.id;
+  const indicesData = backend2.dataIdMap.get(flattenIndex.dataId);
+  const indicesId = indicesData.id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenX.shape)).buffer);
+  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenOutputShape)).buffer);
+  wasmGather(xId, CppDType[x.dtype], xStridesBytes, stridesSize, indicesId, shapeInfo.batchSize, outStridesBytes, outId);
+  backend2.disposeData(flattenX.dataId);
+  backend2.disposeData(flattenIndex.dataId);
+  out.shape = shapeInfo.outputShape;
+  return out;
+}
+var gatherV2Config3 = {
+  kernelName: GatherV2,
+  backendName: "wasm",
+  setupFunc: setup21,
+  kernelFunc: gatherV23
+};
+var supportsFullBroadcast4 = false;
+var greaterConfig3 = createBinaryKernelConfig(Greater, supportsFullBroadcast4, "bool");
+var supportsFullBroadcast5 = false;
+var greaterEqualConfig3 = createBinaryKernelConfig(GreaterEqual, supportsFullBroadcast5, "bool");
+var wasmFunc3;
+function setupFunc2(backend2) {
+  wasmFunc3 = backend2.wasm.cwrap(LeakyRelu, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function leakyRelu4(args) {
+  const { inputs: { x }, attrs: { alpha }, backend: backend2 } = args;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, "float32");
+  if (util_exports.sizeFromShape(x.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmFunc3(xId, CppDType[x.dtype], alpha, outId);
+  }
+  return out;
+}
+var leakyReluConfig3 = {
+  kernelName: LeakyRelu,
+  backendName: "wasm",
+  setupFunc: setupFunc2,
+  kernelFunc: leakyRelu4
+};
+var supportsFullBroadcast6 = false;
+var lessConfig3 = createBinaryKernelConfig(Less, supportsFullBroadcast6, "bool");
+var supportsFullBroadcast7 = false;
+var lessEqualConfig3 = createBinaryKernelConfig(LessEqual, supportsFullBroadcast7, "bool");
+var logConfig3 = createUnaryKernelConfig(Log);
+var supportsFullBroadcast8 = false;
+var logicalAndConfig3 = createBinaryKernelConfig(LogicalAnd, supportsFullBroadcast8, "bool");
+var logicalNotConfig3 = createUnaryKernelConfig(LogicalNot);
+var supportsFullBroadcast9 = false;
+var logicalOrConfig3 = createBinaryKernelConfig(LogicalOr, supportsFullBroadcast9, "bool");
+var supportsFullBroadcast10 = false;
+var logicalXorConfig = createBinaryKernelConfig(LogicalXor, supportsFullBroadcast10, "bool");
+var wasmMax;
+function setup22(backend2) {
+  wasmMax = backend2.wasm.cwrap(Max, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function max5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { reductionIndices: axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMax(inputId, CppDType[x.dtype], reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var maxConfig3 = {
+  kernelName: Max,
+  backendName: "wasm",
+  setupFunc: setup22,
+  kernelFunc: max5
+};
+var supportsFullBroadcast11 = false;
+var maximumConfig3 = createBinaryKernelConfig(Maximum, supportsFullBroadcast11);
+var wasmMaxPool;
+function setup23(backend2) {
+  wasmMaxPool = backend2.wasm.cwrap(MaxPool, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function maxPool4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const x = inputs.x;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  util_exports.assert(x.dtype === "float32", () => `Error in MaxPool: only float32 input is supported. Got ${x.dtype}.`);
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmMaxPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var maxPoolConfig3 = {
+  kernelName: MaxPool,
+  backendName: "wasm",
+  setupFunc: setup23,
+  kernelFunc: maxPool4
+};
+var wasmMean;
+function setup24(backend2) {
+  wasmMean = backend2.wasm.cwrap(Mean, null, ["number, number, number"]);
+}
+function mean3(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("mean", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  let castedInput = input2;
+  if (input2.dtype !== "float32") {
+    castedInput = cast5({ backend: backend2, inputs: { x: input2 }, attrs: { dtype: "float32" } });
+    inputId = backend2.dataIdMap.get(castedInput.dataId).id;
+  }
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMean(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  if (input2.dtype !== "float32") {
+    backend2.disposeData(castedInput.dataId);
+  }
+  return out;
+}
+var meanConfig3 = {
+  kernelName: Mean,
+  backendName: "wasm",
+  setupFunc: setup24,
+  kernelFunc: mean3
+};
+var wasmMin;
+function setup25(backend2) {
+  wasmMin = backend2.wasm.cwrap(Min, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function min5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+    }
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMin(inputId, CppDType[x.dtype], reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var minConfig3 = {
+  kernelName: Min,
+  backendName: "wasm",
+  setupFunc: setup25,
+  kernelFunc: min5
+};
+var supportsFullBroadcast12 = false;
+var minimumConfig3 = createBinaryKernelConfig(Minimum, supportsFullBroadcast12);
+var MirrorPaddingMode;
+(function(MirrorPaddingMode2) {
+  MirrorPaddingMode2[MirrorPaddingMode2["reflect"] = 0] = "reflect";
+  MirrorPaddingMode2[MirrorPaddingMode2["symmetric"] = 1] = "symmetric";
+})(MirrorPaddingMode || (MirrorPaddingMode = {}));
+var wasmMirrorPad;
+function setup26(backend2) {
+  wasmMirrorPad = backend2.wasm.cwrap(MirrorPad, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function mirrorPad3(args) {
+  const { inputs: { x }, backend: backend2, attrs: { paddings, mode } } = args;
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);
+  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);
+  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);
+  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);
+  wasmMirrorPad(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, MirrorPaddingMode[mode], outId);
+  return out;
+}
+var mirrorPadConfig3 = {
+  kernelName: MirrorPad,
+  backendName: "wasm",
+  kernelFunc: mirrorPad3,
+  setupFunc: setup26
+};
+var supportsFullBroadcast13 = true;
+var multiplyConfig3 = createBinaryKernelConfig(Multiply, supportsFullBroadcast13);
+var negConfig3 = createUnaryKernelConfig(Neg);
+function parseResultStruct(backend2, resOffset) {
+  const result = new Int32Array(backend2.wasm.HEAPU8.buffer, resOffset, 4);
+  const pSelectedIndices = result[0];
+  const selectedSize = result[1];
+  const pSelectedScores = result[2];
+  const pValidOutputs = result[3];
+  backend2.wasm._free(resOffset);
+  return { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs };
+}
+var wasmFunc4;
+function setup27(backend2) {
+  wasmFunc4 = backend2.wasm.cwrap(
+    NonMaxSuppressionV3,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number"
+    ]
+  );
+}
+function kernelFunc(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc4(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pSelectedScores);
+  backend2.wasm._free(pValidOutputs);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  return selectedIndicesTensor;
+}
+var nonMaxSuppressionV3Config3 = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "wasm",
+  setupFunc: setup27,
+  kernelFunc
+};
+var wasmFunc5;
+function setup28(backend2) {
+  wasmFunc5 = backend2.wasm.cwrap(
+    NonMaxSuppressionV4,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number",
+      "bool"
+    ]
+  );
+}
+function nonMaxSuppressionV43(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold, padToMaxOutputSize } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc5(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pSelectedScores);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  const validOutputsTensor = backend2.makeOutput([], "int32", pValidOutputs);
+  return [selectedIndicesTensor, validOutputsTensor];
+}
+var nonMaxSuppressionV4Config3 = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "wasm",
+  setupFunc: setup28,
+  kernelFunc: nonMaxSuppressionV43
+};
+var wasmFunc6;
+function setup29(backend2) {
+  wasmFunc6 = backend2.wasm.cwrap(
+    NonMaxSuppressionV5,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number",
+      "number"
+    ]
+  );
+}
+function kernelFunc2(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold, softNmsSigma } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc6(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pValidOutputs);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  const selectedScoresTensor = backend2.makeOutput([selectedSize], "float32", pSelectedScores);
+  return [selectedIndicesTensor, selectedScoresTensor];
+}
+var nonMaxSuppressionV5Config3 = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "wasm",
+  setupFunc: setup29,
+  kernelFunc: kernelFunc2
+};
+var supportsFullBroadcast14 = false;
+var notEqualConfig3 = createBinaryKernelConfig(NotEqual, supportsFullBroadcast14, "bool");
+var wasmOneHot;
+function setup30(backend2) {
+  wasmOneHot = backend2.wasm.cwrap(OneHot, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function oneHot4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  const out = backend2.makeOutput([...indices.shape, depth], dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  wasmOneHot(indicesId, depth, onValue, offValue, outId);
+  return out;
+}
+var oneHotConfig3 = {
+  kernelName: OneHot,
+  backendName: "wasm",
+  setupFunc: setup30,
+  kernelFunc: oneHot4
+};
+function onesLike4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(1);
+  return out;
+}
+var onesLikeConfig3 = {
+  kernelName: OnesLike,
+  backendName: "wasm",
+  kernelFunc: onesLike4
+};
+function pack3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims5({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims5({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat4({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeData(t.dataId));
+  return result;
+}
+var packConfig3 = {
+  kernelName: Pack,
+  backendName: "wasm",
+  kernelFunc: pack3
+};
+var wasmPadV2;
+function setup31(backend2) {
+  wasmPadV2 = backend2.wasm.cwrap(PadV2, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function pad2(args) {
+  const { inputs: { x }, backend: backend2, attrs: { paddings, constantValue } } = args;
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return fill4({
+      backend: backend2,
+      attrs: { shape: outShape, value: constantValue, dtype: x.dtype }
+    });
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const outTensorData = backend2.dataIdMap.get(out.dataId);
+  const outId = outTensorData.id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);
+  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);
+  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);
+  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);
+  wasmPadV2(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, constantValue, outId);
+  return out;
+}
+var padV2Config3 = {
+  kernelName: PadV2,
+  backendName: "wasm",
+  kernelFunc: pad2,
+  setupFunc: setup31
+};
+var supportsFullBroadcast15 = false;
+var powConfig3 = createBinaryKernelConfig(Pow, supportsFullBroadcast15);
+var wasmPrelu;
+function setup32(backend2) {
+  wasmPrelu = backend2.wasm.cwrap(Prelu, null, [
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function prelu5(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const weightsId = backend2.dataIdMap.get(alpha.dataId).id;
+  let inputId = xId;
+  const input2 = x;
+  let castedInput = input2;
+  if (input2.dtype !== "float32") {
+    castedInput = cast5({ backend: backend2, inputs: { x }, attrs: { dtype: "float32" } });
+    inputId = backend2.dataIdMap.get(castedInput.dataId).id;
+  }
+  const out = backend2.makeOutput(x.shape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmPrelu(inputId, weightsId, outId);
+  if (input2.dtype !== "float32") {
+    backend2.disposeData(castedInput.dataId);
+  }
+  return out;
+}
+var preluConfig3 = {
+  kernelName: Prelu,
+  backendName: "wasm",
+  setupFunc: setup32,
+  kernelFunc: prelu5
+};
+var wasmProd;
+function setup33(backend2) {
+  wasmProd = backend2.wasm.cwrap(Prod, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function prod4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("prod", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmProd(inputId, reduceSize, CppDType[out.dtype], outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var prodConfig3 = {
+  kernelName: Prod,
+  backendName: "wasm",
+  setupFunc: setup33,
+  kernelFunc: prod4
+};
+var range5 = (args) => {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, step: step5, dtype } = attrs;
+  const values = rangeImpl(start, stop, step5, dtype);
+  const out = backend2.makeOutput([values.length], dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(values);
+  return out;
+};
+var rangeConfig3 = {
+  kernelName: Range,
+  backendName: "wasm",
+  kernelFunc: range5
+};
+var supportsFullBroadcast16 = true;
+var realDivConfig3 = createBinaryKernelConfig(RealDiv, supportsFullBroadcast16);
+var reluConfig3 = createUnaryKernelConfig(Relu);
+var relu6Config3 = createUnaryKernelConfig(Relu6);
+var wasmResizeBilinear;
+function setup34(backend2) {
+  wasmResizeBilinear = backend2.wasm.cwrap(ResizeBilinear, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function resizeBilinear4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const outShape = [batch, newHeight, newWidth, numChannels];
+  let xData = backend2.dataIdMap.get(images.dataId);
+  let castedData;
+  if (xData.dtype !== "float32") {
+    castedData = cast5({ backend: backend2, inputs: { x: images }, attrs: { dtype: "float32" } });
+    xData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const xId = xData.id;
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(images.shape) === 0) {
+    return out;
+  }
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmResizeBilinear(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var resizeBilinearConfig3 = {
+  kernelName: ResizeBilinear,
+  backendName: "wasm",
+  setupFunc: setup34,
+  kernelFunc: resizeBilinear4
+};
+var wasmResizeNearestNeighbor;
+function setup35(backend2) {
+  wasmResizeNearestNeighbor = backend2.wasm.cwrap(ResizeNearestNeighbor, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function resizeNearestNeighbor4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const outShape = [batch, newHeight, newWidth, numChannels];
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(images.shape) === 0) {
+    return out;
+  }
+  let xData = backend2.dataIdMap.get(images.dataId);
+  let castedData;
+  if (xData.dtype !== "float32") {
+    castedData = cast5({
+      backend: backend2,
+      inputs: { x: images },
+      attrs: { dtype: "float32" }
+    });
+    xData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const xId = xData.id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmResizeNearestNeighbor(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var resizeNearestNeighborConfig3 = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "wasm",
+  setupFunc: setup35,
+  kernelFunc: resizeNearestNeighbor4
+};
+var wasmReverse;
+function setup36(backend2) {
+  wasmReverse = backend2.wasm.cwrap(Reverse, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function reverse4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  const axes = util_exports.parseAxisParam(dims, x.shape);
+  if (x.shape.length === 0) {
+    return identity4({ inputs: { x }, backend: backend2 });
+  }
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const axesBytes = new Uint8Array(new Int32Array(axes).buffer);
+  const outShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  wasmReverse(xId, axesBytes, axes.length, outShapeBytes, x.shape.length, outId);
+  const reshaped = reshape5({ inputs: { x: out }, attrs: { shape: x.shape }, backend: backend2 });
+  backend2.disposeData(out.dataId);
+  return reshaped;
+}
+var reverseConfig3 = {
+  kernelName: Reverse,
+  backendName: "wasm",
+  kernelFunc: reverse4,
+  setupFunc: setup36
+};
+var wasmRotate;
+function setup37(backend2) {
+  wasmRotate = backend2.wasm.cwrap(RotateWithOffset, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function rotateWithOffset2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2 } = inputs;
+  const { radians, fillValue, center } = attrs;
+  const out = backend2.makeOutput(image2.shape, image2.dtype);
+  const imageId = backend2.dataIdMap.get(image2.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);
+  const fillIsBlack = fillValue === 0;
+  const fullOpacityValue = 255;
+  const fillValues2 = typeof fillValue === "number" ? [fillValue, fillValue, fillValue, fillIsBlack ? 0 : fullOpacityValue] : [...fillValue, fullOpacityValue];
+  const fillBytes = new Uint8Array(new Int32Array(fillValues2).buffer);
+  wasmRotate(imageId, batch, imageHeight, imageWidth, numChannels, radians, centerX, centerY, fillBytes, fillValues2.length, outId);
+  return out;
+}
+var rotateWithOffsetConfig3 = {
+  kernelName: RotateWithOffset,
+  backendName: "wasm",
+  kernelFunc: rotateWithOffset2,
+  setupFunc: setup37
+};
+var roundConfig3 = createUnaryKernelConfig(Round);
+var rsqrtConfig3 = createUnaryKernelConfig(Rsqrt);
+var wasmScatterNd;
+function setup38(backend2) {
+  wasmScatterNd = backend2.wasm.cwrap(ScatterNd, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function scatterNd3(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const out = backend2.makeOutput(shape, updates.dtype);
+  if (util_exports.sizeFromShape(shape) === 0) {
+    return out;
+  }
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = scatter_nd_util_exports.calculateShapes(updates, indices, shape);
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  const updatesData = backend2.dataIdMap.get(updates.dataId);
+  const updatesId = updatesData.id;
+  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmScatterNd(indicesId, updatesId, CppDType[updates.dtype], sliceRank, numUpdates, sliceSize, stridesBytes, outputSize, outId);
+  return out;
+}
+var scatterNdConfig3 = {
+  kernelName: ScatterNd,
+  backendName: "wasm",
+  setupFunc: setup38,
+  kernelFunc: scatterNd3
+};
+var wasmSelect;
+function setup39(backend2) {
+  wasmSelect = backend2.wasm.cwrap("SelectV2", null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function select4(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  const conditionId = backend2.dataIdMap.get(condition.dataId).id;
+  const tId = backend2.dataIdMap.get(t.dataId).id;
+  const eId = backend2.dataIdMap.get(e.dataId).id;
+  const out = backend2.makeOutput(t.shape, t.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const cRank = condition.shape.length;
+  const tRank = t.shape.length;
+  const offset = cRank === 0 || cRank > 1 || tRank === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));
+  wasmSelect(conditionId, tId, eId, offset, outId);
+  return out;
+}
+var selectConfig3 = {
+  kernelName: Select,
+  backendName: "wasm",
+  kernelFunc: select4,
+  setupFunc: setup39
+};
+var wasmFunc7;
+function setup40(backend2) {
+  wasmFunc7 = backend2.wasm.cwrap(Sigmoid, null, ["number", "number"]);
+}
+function sigmoid4(args) {
+  const { backend: backend2, inputs: { x } } = args;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  wasmFunc7(xId, outId);
+  return out;
+}
+var sigmoidConfig3 = {
+  kernelName: "Sigmoid",
+  backendName: "wasm",
+  setupFunc: setup40,
+  kernelFunc: sigmoid4
+};
+var sinConfig3 = createUnaryKernelConfig(Sin);
+var wasmFunc8;
+function setup41(backend2) {
+  wasmFunc8 = backend2.wasm.cwrap(Softmax, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function softmax5(args) {
+  const { backend: backend2, inputs: { logits }, attrs: { dim } } = args;
+  const xId = backend2.dataIdMap.get(logits.dataId).id;
+  const out = backend2.makeOutput(logits.shape, logits.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const channels = logits.shape[dim];
+  const batch = util_exports.sizeFromShape(logits.shape) / channels;
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  wasmFunc8(xId, outId, channels, batch);
+  return out;
+}
+var softmaxConfig3 = {
+  kernelName: Softmax,
+  backendName: "wasm",
+  setupFunc: setup41,
+  kernelFunc: softmax5
+};
+function spaceToBatchND4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  const prod5 = util_exports.sizeFromShape(blockShape);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const paddedX = padV2Config3.kernelFunc({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapeInputs = { x: paddedX };
+  const reshapeAttrs = { shape: reshapedPaddedShape };
+  const paddedXReshaped = reshape5({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });
+  const transposeInputs = { x: paddedXReshaped };
+  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };
+  const paddedXT = transpose4({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });
+  const resultReshapeInputs = { x: paddedXT };
+  const resultReshapeAttrs = { shape: flattenShape };
+  const result = reshape5({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });
+  backend2.disposeData(paddedX.dataId);
+  backend2.disposeData(paddedXReshaped.dataId);
+  backend2.disposeData(paddedXT.dataId);
+  return result;
+}
+var spaceToBatchNDConfig3 = {
+  kernelName: SpaceToBatchND,
+  backendName: "wasm",
+  kernelFunc: spaceToBatchND4
+};
+var wasmSparseFillEmptyRows;
+function setup42(backend2) {
+  wasmSparseFillEmptyRows = backend2.wasm.cwrap("SparseFillEmptyRows", "number", [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseFillEmptyRows4(args) {
+  const { backend: backend2, inputs } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  const indicesCount = indices.shape[0];
+  const rank = indices.shape[1];
+  const denseRows = backend2.readSync(denseShape.dataId)[0];
+  const maxOutputIndicesShape = [indicesCount + denseRows, rank];
+  const indicesId = backend2.dataIdMap.get(indices.dataId).id;
+  const valuesId = backend2.dataIdMap.get(values.dataId).id;
+  const defaultValueId = backend2.dataIdMap.get(defaultValue.dataId).id;
+  const outputIndices = backend2.makeOutput(maxOutputIndicesShape, indices.dtype);
+  const outputIndicesId = backend2.dataIdMap.get(outputIndices.dataId).id;
+  const outputValues = backend2.makeOutput(maxOutputIndicesShape.slice(0, 1), values.dtype);
+  const outputValuesId = backend2.dataIdMap.get(outputValues.dataId).id;
+  const emptyRowIndicator = backend2.makeOutput([denseRows], "bool");
+  const emptyRowIndicatorId = backend2.dataIdMap.get(emptyRowIndicator.dataId).id;
+  const reverseIndexMap = backend2.makeOutput([indicesCount], indices.dtype);
+  const reverseIndexMapId = backend2.dataIdMap.get(reverseIndexMap.dataId).id;
+  const exceptionValues = backend2.makeOutput([4], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  const outputRows = wasmSparseFillEmptyRows(indicesId, valuesId, CppDType[values.dtype], indicesCount, denseRows, rank, defaultValueId, outputIndicesId, outputValuesId, emptyRowIndicatorId, reverseIndexMapId, exceptionValuesId);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(exceptionValuesArray[1]);
+      break;
+    }
+    case 2: {
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 3:
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);
+      break;
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(outputIndices.dataId);
+    backend2.disposeData(outputValues.dataId);
+    backend2.disposeData(emptyRowIndicator.dataId);
+    backend2.disposeData(reverseIndexMap.dataId);
+    throw new Error(exceptionMessage);
+  }
+  let resizedIndices = outputIndices;
+  let resizedValues = outputValues;
+  if (outputRows !== maxOutputIndicesShape[0]) {
+    resizedIndices = slice4({
+      inputs: { x: outputIndices },
+      attrs: { begin: 0, size: [outputRows, rank] },
+      backend: backend2
+    });
+    resizedValues = slice4({
+      inputs: { x: outputValues },
+      attrs: { begin: 0, size: outputRows },
+      backend: backend2
+    });
+    backend2.disposeData(outputIndices.dataId);
+    backend2.disposeData(outputValues.dataId);
+  }
+  return [resizedIndices, resizedValues, emptyRowIndicator, reverseIndexMap];
+}
+var sparseFillEmptyRowsConfig3 = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "wasm",
+  setupFunc: setup42,
+  kernelFunc: sparseFillEmptyRows4
+};
+var wasmSparseReshape;
+function setup43(backend2) {
+  wasmSparseReshape = backend2.wasm.cwrap(SparseReshape, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseReshape4(args) {
+  const { backend: backend2, inputs } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape
+        ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape
+        ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const inputIndicesId = backend2.dataIdMap.get(inputIndices.dataId).id;
+  const inputShapeId = backend2.dataIdMap.get(inputShape.dataId).id;
+  const newShapeId = backend2.dataIdMap.get(newShape.dataId).id;
+  const nnz = inputIndices.shape[0];
+  const outputRank = util_exports.sizeFromShape(newShape.shape);
+  const newIndices = backend2.makeOutput([nnz, outputRank], inputIndices.dtype);
+  const newIndicesId = backend2.dataIdMap.get(newIndices.dataId).id;
+  const outputShape = backend2.makeOutput([outputRank], newShape.dtype);
+  const outputShapeId = backend2.dataIdMap.get(outputShape.dataId).id;
+  const exceptionValues = backend2.makeOutput([3], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  wasmSparseReshape(inputIndicesId, inputShapeId, newShapeId, nnz, newIndicesId, outputShapeId, exceptionValuesId);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 0: {
+      exceptionMessage = backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 2:
+      exceptionMessage = backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();
+      break;
+    case 3: {
+      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));
+      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShapeValues, outputShapeValues);
+      break;
+    }
+    case 4: {
+      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));
+      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShapeValues, outputShapeValues);
+      break;
+    }
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(newIndices.dataId);
+    backend2.disposeData(outputShape.dataId);
+    throw new Error(exceptionMessage);
+  }
+  return [newIndices, outputShape];
+}
+var sparseReshapeConfig3 = {
+  kernelName: SparseReshape,
+  backendName: "wasm",
+  setupFunc: setup43,
+  kernelFunc: sparseReshape4
+};
+var wasmSparseSegmentReduction;
+function setup44(backend2) {
+  wasmSparseSegmentReduction = backend2.wasm.cwrap("SparseSegmentReduction", null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseSegmentReduction(args, isMean) {
+  const { backend: backend2, inputs } = args;
+  const { data, indices, segmentIds } = inputs;
+  const numIndices = indices.shape[0];
+  const segmentIdsBack = backend2.readSync(segmentIds.dataId, numIndices - 1, numIndices)[0];
+  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIdsBack + 1 : 0;
+  const outputRows = lastSegmentIdPlusOne;
+  if (outputRows < 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  const outputShape = data.shape.slice();
+  outputShape[0] = outputRows;
+  const dataId = backend2.dataIdMap.get(data.dataId).id;
+  const indicesId = backend2.dataIdMap.get(indices.dataId).id;
+  const segmentIdsId = backend2.dataIdMap.get(segmentIds.dataId).id;
+  const output = backend2.makeOutput(outputShape, data.dtype);
+  const outputId = backend2.dataIdMap.get(output.dataId).id;
+  const exceptionValues = backend2.makeOutput([4], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  wasmSparseSegmentReduction(dataId, CppDType[data.dtype], data.shape[0], indicesId, segmentIdsId, outputId, exceptionValuesId, isMean, 0);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 0: {
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();
+      break;
+    }
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();
+      break;
+    }
+    case 2:
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    case 3:
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);
+      break;
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(output.dataId);
+    throw new Error(exceptionMessage);
+  }
+  return output;
+}
+function sparseSegmentMean4(args) {
+  return sparseSegmentReduction(args, true);
+}
+var sparseSegmentMeanConfig3 = {
+  kernelName: SparseSegmentMean,
+  backendName: "wasm",
+  setupFunc: setup44,
+  kernelFunc: sparseSegmentMean4
+};
+function sparseSegmentSum4(args) {
+  return sparseSegmentReduction(args, false);
+}
+var sparseSegmentSumConfig3 = {
+  kernelName: SparseSegmentSum,
+  backendName: "wasm",
+  setupFunc: setup44,
+  kernelFunc: sparseSegmentSum4
+};
+function splitV3(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const begin = new Array(x.shape.length).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const xSliceSize = [...size];
+    xSliceSize[$axis] = s;
+    const xSlice = slice4({ inputs: { x }, attrs: { begin, size: xSliceSize }, backend: backend2 });
+    begin[$axis] += s;
+    return xSlice;
+  });
+}
+var splitVConfig3 = {
+  kernelName: SplitV,
+  backendName: "wasm",
+  kernelFunc: splitV3
+};
+var sqrtConfig3 = createUnaryKernelConfig(Sqrt);
+var squareConfig3 = createUnaryKernelConfig(Square);
+var supportsFullBroadcast17 = true;
+var squaredDifferenceConfig3 = createBinaryKernelConfig(SquaredDifference, supportsFullBroadcast17);
+var wasmStep;
+function setup45(backend2) {
+  wasmStep = backend2.wasm.cwrap(Step, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function step4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { alpha } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmStep(xId, alpha, CppDType[x.dtype], outId);
+  return out;
+}
+var stepConfig3 = {
+  kernelName: Step,
+  backendName: "wasm",
+  setupFunc: setup45,
+  kernelFunc: step4
+};
+var wasmStridedSlice;
+function setup46(backend2) {
+  wasmStridedSlice = backend2.wasm.cwrap(StridedSlice, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "array",
+    "array",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function stridedSlice4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice4({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape5({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeData(sliced.dataId);
+  } else {
+    const out = backend2.makeOutput(finalShapeSparse, "float32");
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);
+    const beginBytes = new Uint8Array(new Int32Array($begin).buffer);
+    const endBytes = new Uint8Array(new Int32Array($end).buffer);
+    const stridesBytes = new Uint8Array(new Int32Array($strides).buffer);
+    const outputShapeBytes = new Uint8Array(new Int32Array(finalShapeSparse).buffer);
+    const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(finalShapeSparse)).buffer);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmStridedSlice(xId, xStridesBytes, x.shape.length, beginBytes, endBytes, stridesBytes, outputShapeBytes, outStridesBytes, finalShapeSparse.length, outId);
+    result = reshape5({ inputs: { x: out }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeData(out.dataId);
+  }
+  return result;
+}
+var stridedSliceConfig3 = {
+  kernelName: StridedSlice,
+  backendName: "wasm",
+  setupFunc: setup46,
+  kernelFunc: stridedSlice4
+};
+function stringNGrams4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { data, dataSplits } = inputs;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const $data = backend2.readSync(data.dataId);
+  const $dataSplits = backend2.readSync(dataSplits.dataId);
+  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  const nGramsOut = backend2.makeOutput([nGrams.length], "string");
+  const nGramsOutData = backend2.dataIdMap.get(nGramsOut.dataId);
+  nGramsOutData.stringBytes = nGrams;
+  const nGramsSplitsOut = backend2.makeOutput(dataSplits.shape, "int32");
+  const nGramsSplitsOutVals = backend2.typedArrayFromHeap(nGramsSplitsOut);
+  nGramsSplitsOutVals.set(nGramsSplits);
+  return [nGramsOut, nGramsSplitsOut];
+}
+var stringNGramsConfig3 = {
+  kernelName: StringNGrams,
+  backendName: "wasm",
+  kernelFunc: stringNGrams4
+};
+function stringSplit4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { input: input2, delimiter } = inputs;
+  const { skipEmpty } = attrs;
+  const inputVals = backend2.readSync(input2.dataId);
+  const delimiterVals = backend2.readSync(delimiter.dataId);
+  const [indices, values, shape] = stringSplitImpl(inputVals, delimiterVals[0], skipEmpty);
+  const outputSize = values.length;
+  const indicesOut = backend2.makeOutput([outputSize, 2], "int32");
+  const indicesOutVals = backend2.typedArrayFromHeap(indicesOut);
+  indicesOutVals.set(indices);
+  const valuesOut = backend2.makeOutput([outputSize], "string");
+  const valuesOutData = backend2.dataIdMap.get(valuesOut.dataId);
+  valuesOutData.stringBytes = values;
+  const shapeOut = backend2.makeOutput([2], "int32");
+  const shapeOutVals = backend2.typedArrayFromHeap(shapeOut);
+  shapeOutVals.set(shape);
+  return [indicesOut, valuesOut, shapeOut];
+}
+var stringSplitConfig3 = {
+  kernelName: StringSplit,
+  backendName: "wasm",
+  kernelFunc: stringSplit4
+};
+function stringToHashBucketFast4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { input: input2 } = inputs;
+  const { numBuckets } = attrs;
+  const inputVals = backend2.readSync(input2.dataId);
+  const values = stringToHashBucketFastImpl(inputVals, numBuckets);
+  const out = backend2.makeOutput(input2.shape, "int32");
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(values);
+  return out;
+}
+var stringToHashBucketFastConfig3 = {
+  kernelName: StringToHashBucketFast,
+  backendName: "wasm",
+  kernelFunc: stringToHashBucketFast4
+};
+var supportsFullBroadcast18 = true;
+var subConfig3 = createBinaryKernelConfig(Sub, supportsFullBroadcast18);
+var wasmSum;
+function setup47(backend2) {
+  wasmSum = backend2.wasm.cwrap(Sum, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sum5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmSum(inputId, reduceSize, CppDType[out.dtype], outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var sumConfig3 = {
+  kernelName: Sum,
+  backendName: "wasm",
+  setupFunc: setup47,
+  kernelFunc: sum5
+};
+var tanConfig3 = createUnaryKernelConfig(Tan);
+var tanhConfig3 = createUnaryKernelConfig(Tanh);
+var wasmTile;
+function setup48(backend2) {
+  wasmTile = backend2.wasm.cwrap(Tile, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function tile5(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const { reps } = attrs;
+  const newShape = new Array(x.shape.length);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[i] * reps[i];
+  }
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const newShapeBytes = new Uint8Array(new Int32Array(newShape).buffer);
+  const out = backend2.makeOutput(newShape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmTile(xId, xShapeBytes, x.shape.length, newShapeBytes, newShape.length, CppDType[out.dtype], outId);
+  return out;
+}
+var tileConfig3 = {
+  kernelName: Tile,
+  backendName: "wasm",
+  setupFunc: setup48,
+  kernelFunc: tile5
+};
+var wasmTopK;
+function setup49(backend2) {
+  wasmTopK = backend2.wasm.cwrap(TopK, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "bool",
+    "number",
+    "number"
+  ]);
+}
+var topk2 = ({ inputs, backend: backend2, attrs }) => {
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const outputShape = x.shape.slice();
+  outputShape[outputShape.length - 1] = k;
+  const outValues = backend2.makeOutput(outputShape, x.dtype);
+  const outValuesId = backend2.dataIdMap.get(outValues.dataId).id;
+  const outIndices = backend2.makeOutput(outputShape, "int32");
+  const outIndicesId = backend2.dataIdMap.get(outIndices.dataId).id;
+  wasmTopK(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], k, sorted, outValuesId, outIndicesId);
+  return [outValues, outIndices];
+};
+var topKConfig3 = {
+  kernelName: TopK,
+  backendName: "wasm",
+  setupFunc: setup49,
+  kernelFunc: topk2
+};
+var wasmTransform;
+function setup50(backend2) {
+  wasmTransform = backend2.wasm.cwrap(Transform, null, [
+    "number",
+    "number",
+    "bool",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function transform4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [
+    batch,
+    outHeight,
+    outWidth,
+    numChannels
+  ];
+  const inputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(image2.shape)).buffer);
+  const outputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(outShape)).buffer);
+  const out = backend2.makeOutput(outShape, image2.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const imageData = backend2.dataIdMap.get(image2.dataId);
+  const imageId = imageData.id;
+  const transformsData = backend2.dataIdMap.get(transforms.dataId);
+  const transformsId = transformsData.id;
+  const interpolationModeId = interpolation === "nearest" ? 1 : 2;
+  let fillModeId;
+  switch (fillMode) {
+    case "constant":
+      fillModeId = 1;
+      break;
+    case "reflect":
+      fillModeId = 2;
+      break;
+    case "wrap":
+      fillModeId = 3;
+      break;
+    case "nearest":
+      fillModeId = 4;
+      break;
+    default:
+      fillModeId = 1;
+      break;
+  }
+  wasmTransform(imageId, transformsId, transforms.shape[0] > 1, batch, outHeight, outWidth, numChannels, imageWidth, imageHeight, inputStrides, image2.shape.length - 1, outputStrides, outShape.length - 1, interpolationModeId, fillModeId, fillValue, outId);
+  return out;
+}
+var transformConfig3 = {
+  kernelName: Transform,
+  backendName: "wasm",
+  setupFunc: setup50,
+  kernelFunc: transform4
+};
+function unpack3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const numOutputs = value.shape[axis];
+  const rank = value.shape.length;
+  const outShape = new Array(rank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < rank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = value.shape[i];
+    }
+  }
+  const outs = new Array(numOutputs);
+  const begin = new Array(rank).fill(0);
+  const size = value.shape.slice();
+  size[axis] = 1;
+  for (let i = 0; i < outs.length; i++) {
+    begin[axis] = i;
+    outs[i] = slice4({ inputs: { x: value }, attrs: { begin, size }, backend: backend2 });
+  }
+  return outs.map(({ dataId, dtype }) => ({ dataId, dtype, shape: outShape }));
+}
+var unpackConfig3 = {
+  kernelName: Unpack,
+  backendName: "wasm",
+  kernelFunc: unpack3
+};
+function zerosLike4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(0);
+  return out;
+}
+var zerosLikeConfig3 = {
+  kernelName: ZerosLike,
+  backendName: "wasm",
+  kernelFunc: zerosLike4
+};
+var kernelConfigs3 = [
+  _fusedMatMulConfig3,
+  absConfig3,
+  addConfig3,
+  addNConfig3,
+  allConfig3,
+  anyConfig3,
+  argMaxConfig3,
+  avgPoolConfig3,
+  batchMatMulConfig3,
+  batchToSpaceNDConfig3,
+  castConfig3,
+  ceilConfig3,
+  clipByValueConfig3,
+  concatConfig3,
+  conv2DConfig3,
+  conv2DBackpropInputConfig3,
+  cosConfig3,
+  coshConfig3,
+  cropAndResizeConfig3,
+  cumprodConfig3,
+  cumsumConfig3,
+  depthToSpaceConfig3,
+  depthwiseConv2dNativeConfig3,
+  eluConfig3,
+  equalConfig3,
+  expConfig3,
+  expandDimsConfig3,
+  fillConfig3,
+  flipLeftRightConfig3,
+  floorConfig3,
+  floorDivConfig3,
+  fusedBatchNormConfig,
+  fusedConv2DConfig3,
+  fusedDepthwiseConv2DConfig3,
+  gatherNdConfig3,
+  gatherV2Config3,
+  greaterConfig3,
+  greaterEqualConfig3,
+  identityConfig3,
+  leakyReluConfig3,
+  lessConfig3,
+  lessEqualConfig3,
+  logConfig3,
+  logicalAndConfig3,
+  logicalNotConfig3,
+  logicalOrConfig3,
+  logicalXorConfig,
+  maxConfig3,
+  maximumConfig3,
+  maxPoolConfig3,
+  meanConfig3,
+  minConfig3,
+  minimumConfig3,
+  mirrorPadConfig3,
+  multiplyConfig3,
+  negConfig3,
+  nonMaxSuppressionV3Config3,
+  nonMaxSuppressionV4Config3,
+  nonMaxSuppressionV5Config3,
+  notEqualConfig3,
+  oneHotConfig3,
+  onesLikeConfig3,
+  packConfig3,
+  padV2Config3,
+  powConfig3,
+  preluConfig3,
+  prodConfig3,
+  rangeConfig3,
+  realDivConfig3,
+  reluConfig3,
+  relu6Config3,
+  reshapeConfig3,
+  resizeBilinearConfig3,
+  resizeNearestNeighborConfig3,
+  reverseConfig3,
+  rotateWithOffsetConfig3,
+  roundConfig3,
+  rsqrtConfig3,
+  scatterNdConfig3,
+  selectConfig3,
+  sigmoidConfig3,
+  sinConfig3,
+  sliceConfig3,
+  softmaxConfig3,
+  spaceToBatchNDConfig3,
+  sparseFillEmptyRowsConfig3,
+  sparseReshapeConfig3,
+  sparseSegmentMeanConfig3,
+  sparseSegmentSumConfig3,
+  splitVConfig3,
+  sqrtConfig3,
+  squareConfig3,
+  squaredDifferenceConfig3,
+  stepConfig3,
+  stridedSliceConfig3,
+  stringNGramsConfig3,
+  stringSplitConfig3,
+  stringToHashBucketFastConfig3,
+  subConfig3,
+  sumConfig3,
+  tanConfig3,
+  tanhConfig3,
+  tileConfig3,
+  topKConfig3,
+  transformConfig3,
+  transposeConfig3,
+  unpackConfig3,
+  zerosLikeConfig3
+];
+for (const kernelConfig of kernelConfigs3) {
+  registerKernel(kernelConfig);
+}
+var ENV6 = env();
+ENV6.registerFlag("WASM_HAS_SIMD_SUPPORT", async () => {
+  try {
+    return WebAssembly.validate(new Uint8Array([
+      0,
+      97,
+      115,
+      109,
+      1,
+      0,
+      0,
+      0,
+      1,
+      4,
+      1,
+      96,
+      0,
+      0,
+      3,
+      2,
+      1,
+      0,
+      10,
+      9,
+      1,
+      7,
+      0,
+      65,
+      0,
+      253,
+      15,
+      26,
+      11
+    ]));
+  } catch (e) {
+    return false;
+  }
+});
+ENV6.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT", async () => {
+  if (ENV6.get("IS_NODE")) {
+    return false;
+  }
+  try {
+    new MessageChannel().port1.postMessage(new SharedArrayBuffer(1));
+    return WebAssembly.validate(new Uint8Array([
+      0,
+      97,
+      115,
+      109,
+      1,
+      0,
+      0,
+      0,
+      1,
+      4,
+      1,
+      96,
+      0,
+      0,
+      3,
+      2,
+      1,
+      0,
+      5,
+      4,
+      1,
+      3,
+      1,
+      1,
+      10,
+      11,
+      1,
+      9,
+      0,
+      65,
+      0,
+      254,
+      16,
+      2,
+      0,
+      26,
+      11
+    ]));
+  } catch (e) {
+    return false;
+  }
+});
+var wasmFactoryThreadedSimd_import = __toESM(require_tfjs_backend_wasm_threaded_simd());
+var import_tfjs_backend_wasm_threaded_simd_worker = __toESM(require_tfjs_backend_wasm_threaded_simd_worker());
+var wasmFactory_import = __toESM(require_tfjs_backend_wasm());
+var wasmFactoryThreadedSimd = wasmFactoryThreadedSimd_import.default || wasmFactoryThreadedSimd_import;
+var wasmFactory = wasmFactory_import.default || wasmFactory_import;
+var BackendWasm = class extends KernelBackend {
+  constructor(wasm) {
+    super();
+    this.wasm = wasm;
+    this.dataIdNextNumber = 1;
+    this.wasm.tfjs.initWithThreadsCount(threadsCount);
+    actualThreadsCount = this.wasm.tfjs.getThreadsCount();
+    this.dataIdMap = new DataStorage(this, engine());
+  }
+  write(values, shape, dtype) {
+    const dataId = { id: this.dataIdNextNumber++ };
+    this.move(dataId, values, shape, dtype, 1);
+    return dataId;
+  }
+  numDataIds() {
+    return this.dataIdMap.numDataIds();
+  }
+  async time(f) {
+    const start = util_exports.now();
+    f();
+    const kernelMs = util_exports.now() - start;
+    return { kernelMs };
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    const id = this.dataIdNextNumber++;
+    if (dtype === "string") {
+      const stringBytes = values;
+      this.dataIdMap.set(dataId, { id, stringBytes, shape, dtype, memoryOffset: null, refCount });
+      return;
+    }
+    const size = util_exports.sizeFromShape(shape);
+    const numBytes = size * util_exports.bytesPerElement(dtype);
+    const memoryOffset = this.wasm._malloc(numBytes);
+    this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount });
+    this.wasm.tfjs.registerTensor(id, size, memoryOffset);
+    if (values != null) {
+      this.wasm.HEAPU8.set(new Uint8Array(values.buffer, values.byteOffset, numBytes), memoryOffset);
+    }
+  }
+  async read(dataId) {
+    return this.readSync(dataId);
+  }
+  readSync(dataId, start, end) {
+    const { memoryOffset, dtype, shape, stringBytes } = this.dataIdMap.get(dataId);
+    if (dtype === "string") {
+      if ((start == null || start === 0) && (end == null || end >= stringBytes.length)) {
+        return stringBytes;
+      }
+      return stringBytes.slice(start, end);
+    }
+    start = start || 0;
+    end = end || util_exports.sizeFromShape(shape);
+    const bytesPerElement2 = util_exports.bytesPerElement(dtype);
+    const bytes = this.wasm.HEAPU8.slice(memoryOffset + start * bytesPerElement2, memoryOffset + end * bytesPerElement2);
+    return typedArrayFromBuffer(bytes.buffer, dtype);
+  }
+  disposeData(dataId, force = false) {
+    if (this.dataIdMap.has(dataId)) {
+      const data = this.dataIdMap.get(dataId);
+      data.refCount--;
+      if (!force && data.refCount > 0) {
+        return false;
+      }
+      this.wasm._free(data.memoryOffset);
+      this.wasm.tfjs.disposeData(data.id);
+      this.dataIdMap.delete(dataId);
+    }
+    return true;
+  }
+  refCount(dataId) {
+    if (this.dataIdMap.has(dataId)) {
+      const tensorData = this.dataIdMap.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const data = this.dataIdMap.get(dataId);
+    if (data != null) {
+      data.refCount++;
+    }
+  }
+  floatPrecision() {
+    return 32;
+  }
+  getMemoryOffset(dataId) {
+    return this.dataIdMap.get(dataId).memoryOffset;
+  }
+  dispose() {
+    this.wasm.tfjs.dispose();
+    if ("PThread" in this.wasm) {
+      this.wasm.PThread.terminateAllThreads();
+    }
+    this.wasm = null;
+  }
+  memory() {
+    return { unreliable: false };
+  }
+  makeOutput(shape, dtype, memoryOffset) {
+    let dataId;
+    if (memoryOffset == null) {
+      dataId = this.write(null, shape, dtype);
+    } else {
+      const id = this.dataIdNextNumber++;
+      dataId = { id };
+      this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount: 1 });
+      const size = util_exports.sizeFromShape(shape);
+      this.wasm.tfjs.registerTensor(id, size, memoryOffset);
+    }
+    return { dataId, shape, dtype };
+  }
+  typedArrayFromHeap({ shape, dtype, dataId }) {
+    const buffer2 = this.wasm.HEAPU8.buffer;
+    const { memoryOffset } = this.dataIdMap.get(dataId);
+    const size = util_exports.sizeFromShape(shape);
+    switch (dtype) {
+      case "float32":
+        return new Float32Array(buffer2, memoryOffset, size);
+      case "int32":
+        return new Int32Array(buffer2, memoryOffset, size);
+      case "bool":
+        return new Uint8Array(buffer2, memoryOffset, size);
+      default:
+        throw new Error(`Unknown dtype ${dtype}`);
+    }
+  }
+};
+function createInstantiateWasmFunc(path) {
+  return (imports, callback) => {
+    util_exports.fetch(path, { credentials: "same-origin" }).then((response) => {
+      if (!response["ok"]) {
+        imports.env.a(`failed to load wasm binary file at '${path}'`);
+      }
+      response.arrayBuffer().then((binary) => {
+        WebAssembly.instantiate(binary, imports).then((output) => {
+          callback(output.instance, output.module);
+        });
+      });
+    });
+    return {};
+  };
+}
+function getPathToWasmBinary(simdSupported, threadsSupported, wasmModuleFolder) {
+  if (wasmPath != null) {
+    return wasmPath;
+  }
+  let path = "tfjs-backend-wasm.wasm";
+  if (simdSupported && threadsSupported) {
+    path = "tfjs-backend-wasm-threaded-simd.wasm";
+  } else if (simdSupported) {
+    path = "tfjs-backend-wasm-simd.wasm";
+  }
+  if (wasmFileMap != null) {
+    if (wasmFileMap[path] != null) {
+      return wasmFileMap[path];
+    }
+  }
+  return wasmModuleFolder + path;
+}
+async function init() {
+  const [simdSupported, threadsSupported] = await Promise.all([
+    env().getAsync("WASM_HAS_SIMD_SUPPORT"),
+    env().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")
+  ]);
+  return new Promise((resolve, reject) => {
+    const factoryConfig = {};
+    factoryConfig.locateFile = (path, prefix) => {
+      if (path.endsWith(".worker.js")) {
+        const response = import_tfjs_backend_wasm_threaded_simd_worker.wasmWorkerContents.replace(/\n/g, "\\n");
+        const blob = new Blob([response], { type: "application/javascript" });
+        return URL.createObjectURL(blob);
+      }
+      if (path.endsWith(".wasm")) {
+        return getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : prefix);
+      }
+      return prefix + path;
+    };
+    if (customFetch) {
+      factoryConfig.instantiateWasm = createInstantiateWasmFunc(getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : ""));
+    }
+    let initialized = false;
+    factoryConfig.onAbort = () => {
+      if (initialized) {
+        return;
+      }
+      if (initAborted) {
+        return;
+      }
+      initAborted = true;
+      const rejectMsg = "Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers";
+      reject({ message: rejectMsg });
+    };
+    let wasm;
+    if (threadsSupported && simdSupported && wasmPath == null) {
+      factoryConfig.mainScriptUrlOrBlob = new Blob([`var WasmBackendModuleThreadedSimd = ` + wasmFactoryThreadedSimd.toString()], { type: "text/javascript" });
+      wasm = wasmFactoryThreadedSimd(factoryConfig);
+    } else {
+      wasm = wasmFactory(factoryConfig);
+    }
+    wasm.then((module) => {
+      initialized = true;
+      initAborted = false;
+      const voidReturnType = null;
+      module.tfjs = {
+        init: module.cwrap("init", null, []),
+        initWithThreadsCount: module.cwrap("init_with_threads_count", null, ["number"]),
+        getThreadsCount: module.cwrap("get_threads_count", "number", []),
+        registerTensor: module.cwrap("register_tensor", null, [
+          "number",
+          "number",
+          "number"
+        ]),
+        disposeData: module.cwrap("dispose_data", voidReturnType, ["number"]),
+        dispose: module.cwrap("dispose", voidReturnType, [])
+      };
+      resolve({ wasm: module });
+    }).catch(reject);
+  });
+}
+function typedArrayFromBuffer(buffer2, dtype) {
+  switch (dtype) {
+    case "float32":
+      return new Float32Array(buffer2);
+    case "int32":
+      return new Int32Array(buffer2);
+    case "bool":
+      return new Uint8Array(buffer2);
+    default:
+      throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+var wasmBinaryNames = [
+  "tfjs-backend-wasm.wasm",
+  "tfjs-backend-wasm-simd.wasm",
+  "tfjs-backend-wasm-threaded-simd.wasm"
+];
+var wasmPath = null;
+var wasmPathPrefix = null;
+var wasmFileMap = {};
+var initAborted = false;
+var customFetch = false;
+function setWasmPath(path, usePlatformFetch = false) {
+  deprecationWarn("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release.");
+  if (initAborted) {
+    throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");
+  }
+  wasmPath = path;
+  customFetch = usePlatformFetch;
+}
+function setWasmPaths(prefixOrFileMap, usePlatformFetch = false) {
+  if (initAborted) {
+    throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");
+  }
+  if (typeof prefixOrFileMap === "string") {
+    wasmPathPrefix = prefixOrFileMap;
+  } else {
+    wasmFileMap = prefixOrFileMap;
+    const missingPaths = wasmBinaryNames.filter((name) => wasmFileMap[name] == null);
+    if (missingPaths.length > 0) {
+      throw new Error(`There were no entries found for the following binaries: ${missingPaths.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`);
+    }
+  }
+  customFetch = usePlatformFetch;
+}
+var threadsCount = -1;
+var actualThreadsCount = -1;
+function setThreadsCount(numThreads) {
+  threadsCount = numThreads;
+}
+function getThreadsCount() {
+  if (actualThreadsCount === -1) {
+    throw new Error(`WASM backend not initialized.`);
+  }
+  return actualThreadsCount;
+}
+var version8 = "4.0.0";
+var WASM_PRIORITY = 2;
+registerBackend("wasm", async () => {
+  const { wasm } = await init();
+  return new BackendWasm(wasm);
+}, WASM_PRIORITY);
+var version9 = "4.0.0";
+var version22 = "4.0.0";
+var version32 = "4.0.0";
+var version42 = "4.0.0";
+var version52 = "4.0.0";
+var version62 = {
+  tfjs: version9,
+  "tfjs-core": version9,
+  "tfjs-converter": version22,
+  "tfjs-backend-cpu": version32,
+  "tfjs-backend-webgl": version42,
+  "tfjs-backend-wasm": version52
+};
+
+// src/draw/index.ts
+var draw_exports = {};
+__export(draw_exports, {
+  AnchorPosition: () => AnchorPosition,
+  DrawBox: () => DrawBox,
+  DrawBoxOptions: () => DrawBoxOptions,
+  DrawFaceLandmarks: () => DrawFaceLandmarks,
+  DrawFaceLandmarksOptions: () => DrawFaceLandmarksOptions,
+  DrawTextField: () => DrawTextField,
+  DrawTextFieldOptions: () => DrawTextFieldOptions,
+  drawContour: () => drawContour,
+  drawDetections: () => drawDetections,
+  drawFaceExpressions: () => drawFaceExpressions,
+  drawFaceLandmarks: () => drawFaceLandmarks
+});
+
+// src/draw/drawContour.ts
+function drawContour(ctx, points, isClosed = false) {
+  ctx.beginPath();
+  points.slice(1).forEach(({ x, y }, prevIdx) => {
+    const from = points[prevIdx];
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(x, y);
+  });
+  if (isClosed) {
+    const from = points[points.length - 1];
+    const to = points[0];
+    if (!from || !to) {
+      return;
+    }
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(to.x, to.y);
+  }
+  ctx.stroke();
+}
+
+// src/utils/index.ts
+var utils_exports = {};
+__export(utils_exports, {
+  computeReshapedDimensions: () => computeReshapedDimensions,
+  getCenterPoint: () => getCenterPoint,
+  isDimensions: () => isDimensions,
+  isEven: () => isEven2,
+  isFloat: () => isFloat,
+  isTensor: () => isTensor,
+  isTensor1D: () => isTensor1D,
+  isTensor2D: () => isTensor2D,
+  isTensor3D: () => isTensor3D,
+  isTensor4D: () => isTensor4D,
+  isValidNumber: () => isValidNumber,
+  isValidProbablitiy: () => isValidProbablitiy,
+  range: () => range6,
+  round: () => round5
+});
+
+// src/classes/Dimensions.ts
+var Dimensions = class {
+  constructor(width, height) {
+    if (!isValidNumber(width) || !isValidNumber(height)) {
+      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);
+    }
+    this._width = width;
+    this._height = height;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  reverse() {
+    return new Dimensions(1 / this.width, 1 / this.height);
+  }
+};
+
+// src/utils/index.ts
+function isTensor(tensor2, dim) {
+  return tensor2 instanceof Tensor && tensor2.shape.length === dim;
+}
+function isTensor1D(tensor2) {
+  return isTensor(tensor2, 1);
+}
+function isTensor2D(tensor2) {
+  return isTensor(tensor2, 2);
+}
+function isTensor3D(tensor2) {
+  return isTensor(tensor2, 3);
+}
+function isTensor4D(tensor2) {
+  return isTensor(tensor2, 4);
+}
+function isFloat(num) {
+  return num % 1 !== 0;
+}
+function isEven2(num) {
+  return num % 2 === 0;
+}
+function round5(num, prec = 2) {
+  const f = 10 ** prec;
+  return Math.floor(num * f) / f;
+}
+function isDimensions(obj) {
+  return obj && obj.width && obj.height;
+}
+function computeReshapedDimensions({ width, height }, inputSize) {
+  const scale3 = inputSize / Math.max(height, width);
+  return new Dimensions(Math.round(width * scale3), Math.round(height * scale3));
+}
+function getCenterPoint(pts) {
+  return pts.reduce((sum6, pt) => sum6.add(pt), new Point(0, 0)).div(new Point(pts.length, pts.length));
+}
+function range6(num, start, step5) {
+  return Array(num).fill(0).map((_, i) => start + i * step5);
+}
+function isValidNumber(num) {
+  return !!num && num !== Infinity && num !== -Infinity && !Number.isNaN(num) || num === 0;
+}
+function isValidProbablitiy(num) {
+  return isValidNumber(num) && num >= 0 && num <= 1;
+}
+
+// src/classes/Point.ts
+var Point = class {
+  constructor(x, y) {
+    this._x = x;
+    this._y = y;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  add(pt) {
+    return new Point(this.x + pt.x, this.y + pt.y);
+  }
+  sub(pt) {
+    return new Point(this.x - pt.x, this.y - pt.y);
+  }
+  mul(pt) {
+    return new Point(this.x * pt.x, this.y * pt.y);
+  }
+  div(pt) {
+    return new Point(this.x / pt.x, this.y / pt.y);
+  }
+  abs() {
+    return new Point(Math.abs(this.x), Math.abs(this.y));
+  }
+  magnitude() {
+    return Math.sqrt(this.x ** 2 + this.y ** 2);
+  }
+  floor() {
+    return new Point(Math.floor(this.x), Math.floor(this.y));
+  }
+};
+
+// src/classes/Box.ts
+var Box = class {
+  static isRect(rect) {
+    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);
+  }
+  static assertIsValidBox(box, callee, allowNegativeDimensions = false) {
+    if (!Box.isRect(box)) {
+      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);
+    }
+    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {
+      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);
+    }
+  }
+  constructor(_box, allowNegativeDimensions = true) {
+    const box = _box || {};
+    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);
+    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);
+    if (!isRect && !isBbox) {
+      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);
+    }
+    const [x, y, width, height] = isRect ? [box.x, box.y, box.width, box.height] : [box.left, box.top, box.right - box.left, box.bottom - box.top];
+    Box.assertIsValidBox({
+      x,
+      y,
+      width,
+      height
+    }, "Box.constructor", allowNegativeDimensions);
+    this._x = x;
+    this._y = y;
+    this._width = width;
+    this._height = height;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  get left() {
+    return this.x;
+  }
+  get top() {
+    return this.y;
+  }
+  get right() {
+    return this.x + this.width;
+  }
+  get bottom() {
+    return this.y + this.height;
+  }
+  get area() {
+    return this.width * this.height;
+  }
+  get topLeft() {
+    return new Point(this.left, this.top);
+  }
+  get topRight() {
+    return new Point(this.right, this.top);
+  }
+  get bottomLeft() {
+    return new Point(this.left, this.bottom);
+  }
+  get bottomRight() {
+    return new Point(this.right, this.bottom);
+  }
+  round() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.round(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  floor() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.floor(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  toSquare() {
+    let {
+      x,
+      y,
+      width,
+      height
+    } = this;
+    const diff = Math.abs(width - height);
+    if (width < height) {
+      x -= diff / 2;
+      width += diff;
+    }
+    if (height < width) {
+      y -= diff / 2;
+      height += diff;
+    }
+    return new Box({ x, y, width, height });
+  }
+  rescale(s) {
+    const scaleX = isDimensions(s) ? s.width : s;
+    const scaleY = isDimensions(s) ? s.height : s;
+    return new Box({
+      x: this.x * scaleX,
+      y: this.y * scaleY,
+      width: this.width * scaleX,
+      height: this.height * scaleY
+    });
+  }
+  pad(padX, padY) {
+    const [x, y, width, height] = [
+      this.x - padX / 2,
+      this.y - padY / 2,
+      this.width + padX,
+      this.height + padY
+    ];
+    return new Box({ x, y, width, height });
+  }
+  clipAtImageBorders(imgWidth, imgHeight) {
+    const { x, y, right, bottom } = this;
+    const clippedX = Math.max(x, 0);
+    const clippedY = Math.max(y, 0);
+    const newWidth = right - clippedX;
+    const newHeight = bottom - clippedY;
+    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);
+    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);
+    return new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight }).floor();
+  }
+  shift(sx, sy) {
+    const { width, height } = this;
+    const x = this.x + sx;
+    const y = this.y + sy;
+    return new Box({ x, y, width, height });
+  }
+  padAtBorders(imageHeight, imageWidth) {
+    const w = this.width + 1;
+    const h = this.height + 1;
+    const dx = 1;
+    const dy = 1;
+    let edx = w;
+    let edy = h;
+    let x = this.left;
+    let y = this.top;
+    let ex = this.right;
+    let ey = this.bottom;
+    if (ex > imageWidth) {
+      edx = -ex + imageWidth + w;
+      ex = imageWidth;
+    }
+    if (ey > imageHeight) {
+      edy = -ey + imageHeight + h;
+      ey = imageHeight;
+    }
+    if (x < 1) {
+      edy = 2 - x;
+      x = 1;
+    }
+    if (y < 1) {
+      edy = 2 - y;
+      y = 1;
+    }
+    return { dy, edy, dx, edx, y, ey, x, ex, w, h };
+  }
+  calibrate(region) {
+    return new Box({
+      left: this.left + region.left * this.width,
+      top: this.top + region.top * this.height,
+      right: this.right + region.right * this.width,
+      bottom: this.bottom + region.bottom * this.height
+    }).toSquare().round();
+  }
+};
+
+// src/classes/BoundingBox.ts
+var BoundingBox = class extends Box {
+  constructor(left, top, right, bottom, allowNegativeDimensions = false) {
+    super({ left, top, right, bottom }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/ObjectDetection.ts
+var ObjectDetection = class {
+  constructor(score, classScore, className, relativeBox, imageDims) {
+    this._imageDims = new Dimensions(imageDims.width, imageDims.height);
+    this._score = score;
+    this._classScore = classScore;
+    this._className = className;
+    this._box = new Box(relativeBox).rescale(this._imageDims);
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+  get className() {
+    return this._className;
+  }
+  get box() {
+    return this._box;
+  }
+  get imageDims() {
+    return this._imageDims;
+  }
+  get imageWidth() {
+    return this.imageDims.width;
+  }
+  get imageHeight() {
+    return this.imageDims.height;
+  }
+  get relativeBox() {
+    return new Box(this._box).rescale(this.imageDims.reverse());
+  }
+  forSize(width, height) {
+    return new ObjectDetection(
+      this.score,
+      this.classScore,
+      this.className,
+      this.relativeBox,
+      { width, height }
+    );
+  }
+};
+
+// src/classes/FaceDetection.ts
+var FaceDetection = class extends ObjectDetection {
+  constructor(score, relativeBox, imageDims) {
+    super(score, score, "", relativeBox, imageDims);
+  }
+  forSize(width, height) {
+    const { score, relativeBox, imageDims } = super.forSize(width, height);
+    return new FaceDetection(score, relativeBox, imageDims);
+  }
+};
+
+// src/ops/iou.ts
+function iou(box1, box2, isIOU = true) {
+  const width = Math.max(0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));
+  const height = Math.max(0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));
+  const interSection = width * height;
+  return isIOU ? interSection / (box1.area + box2.area - interSection) : interSection / Math.min(box1.area, box2.area);
+}
+
+// src/ops/minBbox.ts
+function minBbox(pts) {
+  const xs = pts.map((pt) => pt.x);
+  const ys = pts.map((pt) => pt.y);
+  const minX = xs.reduce((min6, x) => x < min6 ? x : min6, Infinity);
+  const minY = ys.reduce((min6, y) => y < min6 ? y : min6, Infinity);
+  const maxX = xs.reduce((max6, x) => max6 < x ? x : max6, 0);
+  const maxY = ys.reduce((max6, y) => max6 < y ? y : max6, 0);
+  return new BoundingBox(minX, minY, maxX, maxY);
+}
+
+// src/ops/nonMaxSuppression.ts
+function nonMaxSuppression2(boxes, scores, iouThreshold, isIOU = true) {
+  let indicesSortedByScore = scores.map((score, boxIndex) => ({ score, boxIndex })).sort((c1, c2) => c1.score - c2.score).map((c) => c.boxIndex);
+  const pick = [];
+  while (indicesSortedByScore.length > 0) {
+    const curr = indicesSortedByScore.pop();
+    pick.push(curr);
+    const indices = indicesSortedByScore;
+    const outputs = [];
+    for (let i = 0; i < indices.length; i++) {
+      const idx = indices[i];
+      const currBox = boxes[curr];
+      const idxBox = boxes[idx];
+      outputs.push(iou(currBox, idxBox, isIOU));
+    }
+    indicesSortedByScore = indicesSortedByScore.filter(
+      (_, j) => outputs[j] <= iouThreshold
+    );
+  }
+  return pick;
+}
+
+// src/ops/normalize.ts
+function normalize(x, meanRgb) {
+  return tidy(() => {
+    const [r, g, b] = meanRgb;
+    const avg_r = fill([...x.shape.slice(0, 3), 1], r, "float32");
+    const avg_g = fill([...x.shape.slice(0, 3), 1], g, "float32");
+    const avg_b = fill([...x.shape.slice(0, 3), 1], b, "float32");
+    const avg_rgb = concat([avg_r, avg_g, avg_b], 3);
+    return sub(x, avg_rgb);
+  });
+}
+
+// src/ops/padToSquare.ts
+function padToSquare(imgTensor, isCenterImage = false) {
+  return tidy(() => {
+    const [height, width] = imgTensor.shape.slice(1);
+    if (height === width)
+      return imgTensor;
+    const dimDiff = Math.abs(height - width);
+    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));
+    const paddingAxis = height > width ? 2 : 1;
+    const createPaddingTensor = (paddingAmountLocal) => {
+      const paddingTensorShape = imgTensor.shape.slice();
+      paddingTensorShape[paddingAxis] = paddingAmountLocal;
+      return fill(paddingTensorShape, 0, "float32");
+    };
+    const paddingTensorAppend = createPaddingTensor(paddingAmount);
+    const remainingPaddingAmount = dimDiff - paddingTensorAppend.shape[paddingAxis];
+    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;
+    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend].filter((t) => !!t).map((t) => cast(t, "float32"));
+    return concat(tensorsToStack, paddingAxis);
+  });
+}
+
+// src/ops/shuffleArray.ts
+function shuffleArray(inputArray) {
+  const array2 = inputArray.slice();
+  for (let i = array2.length - 1; i > 0; i--) {
+    const j = Math.floor(Math.random() * (i + 1));
+    const x = array2[i];
+    array2[i] = array2[j];
+    array2[j] = x;
+  }
+  return array2;
+}
+
+// src/ops/index.ts
+function sigmoid5(x) {
+  return 1 / (1 + Math.exp(-x));
+}
+function inverseSigmoid(x) {
+  return Math.log(x / (1 - x));
+}
+
+// src/classes/Rect.ts
+var Rect = class extends Box {
+  constructor(x, y, width, height, allowNegativeDimensions = false) {
+    super({ x, y, width, height }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/FaceLandmarks.ts
+var relX = 0.5;
+var relY = 0.43;
+var relScale = 0.45;
+var FaceLandmarks = class {
+  constructor(relativeFaceLandmarkPositions, imgDims, shift = new Point(0, 0)) {
+    const { width, height } = imgDims;
+    this._imgDims = new Dimensions(width, height);
+    this._shift = shift;
+    this._positions = relativeFaceLandmarkPositions.map(
+      (pt) => pt.mul(new Point(width, height)).add(shift)
+    );
+  }
+  get shift() {
+    return new Point(this._shift.x, this._shift.y);
+  }
+  get imageWidth() {
+    return this._imgDims.width;
+  }
+  get imageHeight() {
+    return this._imgDims.height;
+  }
+  get positions() {
+    return this._positions;
+  }
+  get relativePositions() {
+    return this._positions.map(
+      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight))
+    );
+  }
+  forSize(width, height) {
+    return new this.constructor(
+      this.relativePositions,
+      { width, height }
+    );
+  }
+  shiftBy(x, y) {
+    return new this.constructor(
+      this.relativePositions,
+      this._imgDims,
+      new Point(x, y)
+    );
+  }
+  shiftByPoint(pt) {
+    return this.shiftBy(pt.x, pt.y);
+  }
+  align(detection, options = {}) {
+    if (detection) {
+      const box = detection instanceof FaceDetection ? detection.box.floor() : new Box(detection);
+      return this.shiftBy(box.x, box.y).align(null, options);
+    }
+    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };
+    if (useDlibAlignment) {
+      return this.alignDlib();
+    }
+    return this.alignMinBbox(minBoxPadding);
+  }
+  alignDlib() {
+    const centers = this.getRefPointsForAlignment();
+    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;
+    const distToMouth = (pt) => mouthCenter.sub(pt).magnitude();
+    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;
+    const size = Math.floor(eyeToMouthDist / relScale);
+    const refPoint = getCenterPoint(centers);
+    const x = Math.floor(Math.max(0, refPoint.x - relX * size));
+    const y = Math.floor(Math.max(0, refPoint.y - relY * size));
+    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));
+  }
+  alignMinBbox(padding) {
+    const box = minBbox(this.positions);
+    return box.pad(box.width * padding, box.height * padding);
+  }
+  getRefPointsForAlignment() {
+    throw new Error("getRefPointsForAlignment not implemented by base class");
+  }
+};
+
+// src/classes/FaceLandmarks5.ts
+var FaceLandmarks5 = class extends FaceLandmarks {
+  getRefPointsForAlignment() {
+    const pts = this.positions;
+    return [
+      pts[0],
+      pts[1],
+      getCenterPoint([pts[3], pts[4]])
+    ];
+  }
+};
+
+// src/classes/FaceLandmarks68.ts
+var FaceLandmarks68 = class extends FaceLandmarks {
+  getJawOutline() {
+    return this.positions.slice(0, 17);
+  }
+  getLeftEyeBrow() {
+    return this.positions.slice(17, 22);
+  }
+  getRightEyeBrow() {
+    return this.positions.slice(22, 27);
+  }
+  getNose() {
+    return this.positions.slice(27, 36);
+  }
+  getLeftEye() {
+    return this.positions.slice(36, 42);
+  }
+  getRightEye() {
+    return this.positions.slice(42, 48);
+  }
+  getMouth() {
+    return this.positions.slice(48, 68);
+  }
+  getRefPointsForAlignment() {
+    return [
+      this.getLeftEye(),
+      this.getRightEye(),
+      this.getMouth()
+    ].map(getCenterPoint);
+  }
+};
+
+// src/classes/FaceMatch.ts
+var FaceMatch = class {
+  constructor(label, distance) {
+    this._label = label;
+    this._distance = distance;
+  }
+  get label() {
+    return this._label;
+  }
+  get distance() {
+    return this._distance;
+  }
+  toString(withDistance = true) {
+    return `${this.label}${withDistance ? ` (${round5(this.distance)})` : ""}`;
+  }
+};
+
+// src/classes/LabeledBox.ts
+var LabeledBox = class extends Box {
+  constructor(box, label) {
+    super(box);
+    this._label = label;
+  }
+  static assertIsValidLabeledBox(box, callee) {
+    Box.assertIsValidBox(box, callee);
+    if (!isValidNumber(box.label)) {
+      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);
+    }
+  }
+  get label() {
+    return this._label;
+  }
+};
+
+// src/classes/LabeledFaceDescriptors.ts
+var LabeledFaceDescriptors = class {
+  constructor(label, descriptors) {
+    if (!(typeof label === "string")) {
+      throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");
+    }
+    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {
+      throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");
+    }
+    this._label = label;
+    this._descriptors = descriptors;
+  }
+  get label() {
+    return this._label;
+  }
+  get descriptors() {
+    return this._descriptors;
+  }
+  toJSON() {
+    return {
+      label: this.label,
+      descriptors: this.descriptors.map((d) => Array.from(d))
+    };
+  }
+  static fromJSON(json20) {
+    const descriptors = json20.descriptors.map((d) => new Float32Array(d));
+    return new LabeledFaceDescriptors(json20.label, descriptors);
+  }
+};
+
+// src/classes/PredictedBox.ts
+var PredictedBox = class extends LabeledBox {
+  constructor(box, label, score, classScore) {
+    super(box, label);
+    this._score = score;
+    this._classScore = classScore;
+  }
+  static assertIsValidPredictedBox(box, callee) {
+    LabeledBox.assertIsValidLabeledBox(box, callee);
+    if (!isValidProbablitiy(box.score) || !isValidProbablitiy(box.classScore)) {
+      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);
+    }
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+};
+
+// src/factories/WithFaceDetection.ts
+function isWithFaceDetection(obj) {
+  return obj.detection instanceof FaceDetection;
+}
+function extendWithFaceDetection(sourceObj, detection) {
+  const extension = { detection };
+  return { ...sourceObj, ...extension };
+}
+
+// src/env/createBrowserEnv.ts
+function createBrowserEnv() {
+  const fetch4 = window.fetch;
+  if (!fetch4)
+    throw new Error("fetch - missing fetch implementation for browser environment");
+  const readFile = () => {
+    throw new Error("readFile - filesystem not available for browser environment");
+  };
+  return {
+    Canvas: HTMLCanvasElement,
+    CanvasRenderingContext2D,
+    Image: HTMLImageElement,
+    ImageData,
+    Video: HTMLVideoElement,
+    createCanvasElement: () => document.createElement("canvas"),
+    createImageElement: () => document.createElement("img"),
+    createVideoElement: () => document.createElement("video"),
+    fetch: fetch4,
+    readFile
+  };
+}
+
+// src/env/isNodejs.ts
+function isNodejs() {
+  return typeof global === "object" && typeof process !== "undefined" && process.versions != null && process.versions.node != null;
+}
+
+// src/env/createFileSystem.ts
+function createFileSystem(fs) {
+  let requireFsError = "";
+  if (!fs && isNodejs()) {
+    try {
+      fs = __require("fs");
+    } catch (err) {
+      requireFsError = err.toString();
+    }
+  }
+  const readFile = fs ? (filePath) => new Promise((resolve, reject) => {
+    fs.readFile(filePath, (err, buffer2) => err ? reject(err) : resolve(buffer2));
+  }) : () => {
+    throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`);
+  };
+  return { readFile };
+}
+
+// src/env/createNodejsEnv.ts
+function createNodejsEnv() {
+  const Canvas = global["Canvas"] || global.HTMLCanvasElement;
+  const Image = global.Image || global.HTMLImageElement;
+  const Video = global["Video"] || global.HTMLVideoElement;
+  const createCanvasElement = () => {
+    if (Canvas)
+      return new Canvas();
+    throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment");
+  };
+  const createImageElement = () => {
+    if (Image)
+      return new Image();
+    throw new Error("createImageElement - missing Image implementation for nodejs environment");
+  };
+  const createVideoElement2 = () => {
+    if (Video)
+      return new Video();
+    throw new Error("createVideoElement - missing Video implementation for nodejs environment");
+  };
+  const fetch4 = global.fetch;
+  const fileSystem = createFileSystem();
+  return {
+    Canvas: Canvas || class {
+    },
+    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {
+    },
+    Image: Image || class {
+    },
+    ImageData: global.ImageData || class {
+    },
+    Video: global.HTMLVideoElement || class {
+    },
+    createCanvasElement,
+    createImageElement,
+    createVideoElement: createVideoElement2,
+    fetch: fetch4,
+    ...fileSystem
+  };
+}
+
+// src/env/isBrowser.ts
+function isBrowser2() {
+  return typeof window === "object" && typeof document !== "undefined" && typeof HTMLImageElement !== "undefined" && typeof HTMLCanvasElement !== "undefined" && typeof HTMLVideoElement !== "undefined" && typeof ImageData !== "undefined" && typeof CanvasRenderingContext2D !== "undefined";
+}
+
+// src/env/index.ts
+var environment;
+function getEnv() {
+  if (!environment) {
+    throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  return environment;
+}
+function setEnv(env3) {
+  environment = env3;
+}
+function initialize() {
+  if (isBrowser2())
+    return setEnv(createBrowserEnv());
+  if (isNodejs())
+    return setEnv(createNodejsEnv());
+  return null;
+}
+function monkeyPatch(env3) {
+  if (!environment) {
+    initialize();
+  }
+  if (!environment) {
+    throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  const { Canvas = environment.Canvas, Image = environment.Image } = env3;
+  environment.Canvas = Canvas;
+  environment.Image = Image;
+  environment.createCanvasElement = env3.createCanvasElement || (() => new Canvas());
+  environment.createImageElement = env3.createImageElement || (() => new Image());
+  environment.ImageData = env3.ImageData || environment.ImageData;
+  environment.Video = env3.Video || environment.Video;
+  environment.fetch = env3.fetch || environment.fetch;
+  environment.readFile = env3.readFile || environment.readFile;
+}
+var env2 = {
+  getEnv,
+  setEnv,
+  initialize,
+  createBrowserEnv,
+  createFileSystem,
+  createNodejsEnv,
+  monkeyPatch,
+  isBrowser: isBrowser2,
+  isNodejs
+};
+initialize();
+
+// src/dom/resolveInput.ts
+function resolveInput(arg) {
+  if (!env2.isNodejs() && typeof arg === "string") {
+    return document.getElementById(arg);
+  }
+  return arg;
+}
+
+// src/dom/getContext2dOrThrow.ts
+function getContext2dOrThrow(canvasArg) {
+  const { Canvas, CanvasRenderingContext2D: CanvasRenderingContext2D2 } = env2.getEnv();
+  if (canvasArg instanceof CanvasRenderingContext2D2) {
+    return canvasArg;
+  }
+  const canvas = resolveInput(canvasArg);
+  if (!(canvas instanceof Canvas)) {
+    throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+  }
+  const ctx = canvas.getContext("2d");
+  if (!ctx) {
+    throw new Error("resolveContext2d - canvas 2d context is null");
+  }
+  return ctx;
+}
+
+// src/draw/DrawTextField.ts
+var AnchorPosition = /* @__PURE__ */ ((AnchorPosition2) => {
+  AnchorPosition2["TOP_LEFT"] = "TOP_LEFT";
+  AnchorPosition2["TOP_RIGHT"] = "TOP_RIGHT";
+  AnchorPosition2["BOTTOM_LEFT"] = "BOTTOM_LEFT";
+  AnchorPosition2["BOTTOM_RIGHT"] = "BOTTOM_RIGHT";
+  return AnchorPosition2;
+})(AnchorPosition || {});
+var DrawTextFieldOptions = class {
+  constructor(options = {}) {
+    const {
+      anchorPosition,
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = options;
+    this.anchorPosition = anchorPosition || "TOP_LEFT" /* TOP_LEFT */;
+    this.backgroundColor = backgroundColor || "rgba(0, 0, 0, 0.5)";
+    this.fontColor = fontColor || "rgba(255, 255, 255, 1)";
+    this.fontSize = fontSize || 14;
+    this.fontStyle = fontStyle || "Georgia";
+    this.padding = padding || 4;
+  }
+};
+var DrawTextField = class {
+  constructor(text, anchor, options = {}) {
+    this.text = typeof text === "string" ? [text] : text instanceof DrawTextField ? text.text : text;
+    this.anchor = anchor;
+    this.options = new DrawTextFieldOptions(options);
+  }
+  measureWidth(ctx) {
+    const { padding } = this.options;
+    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => w0 < w1 ? w1 : w0, 0) + 2 * padding;
+  }
+  measureHeight() {
+    const { fontSize, padding } = this.options;
+    return this.text.length * fontSize + 2 * padding;
+  }
+  getUpperLeft(ctx, canvasDims) {
+    const { anchorPosition } = this.options;
+    const isShiftLeft = anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */ || anchorPosition === "TOP_RIGHT" /* TOP_RIGHT */;
+    const isShiftTop = anchorPosition === "BOTTOM_LEFT" /* BOTTOM_LEFT */ || anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */;
+    const textFieldWidth = this.measureWidth(ctx);
+    const textFieldHeight = this.measureHeight();
+    const x = isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x;
+    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;
+    if (canvasDims) {
+      const { width, height } = canvasDims;
+      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);
+      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);
+      return { x: newX, y: newY };
+    }
+    return { x, y };
+  }
+  draw(canvasArg) {
+    const canvas = resolveInput(canvasArg);
+    const ctx = getContext2dOrThrow(canvas);
+    const {
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = this.options;
+    ctx.font = `${fontSize}px ${fontStyle}`;
+    const maxTextWidth = this.measureWidth(ctx);
+    const textHeight = this.measureHeight();
+    ctx.fillStyle = backgroundColor;
+    const upperLeft = this.getUpperLeft(ctx, canvas);
+    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);
+    ctx.fillStyle = fontColor;
+    this.text.forEach((textLine, i) => {
+      const x = padding + upperLeft.x;
+      const y = padding + upperLeft.y + (i + 1) * fontSize;
+      ctx.fillText(textLine, x, y);
+    });
+  }
+};
+
+// src/draw/DrawBox.ts
+var DrawBoxOptions = class {
+  constructor(options = {}) {
+    const {
+      boxColor,
+      lineWidth,
+      label,
+      drawLabelOptions
+    } = options;
+    this.boxColor = boxColor || "rgba(0, 0, 255, 1)";
+    this.lineWidth = lineWidth || 2;
+    this.label = label;
+    const defaultDrawLabelOptions = {
+      anchorPosition: "BOTTOM_LEFT" /* BOTTOM_LEFT */,
+      backgroundColor: this.boxColor
+    };
+    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });
+  }
+};
+var DrawBox = class {
+  constructor(box, options = {}) {
+    this.box = new Box(box);
+    this.options = new DrawBoxOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const { boxColor, lineWidth } = this.options;
+    const {
+      x,
+      y,
+      width,
+      height
+    } = this.box;
+    ctx.strokeStyle = boxColor;
+    ctx.lineWidth = lineWidth;
+    ctx.strokeRect(x, y, width, height);
+    const { label } = this.options;
+    if (label) {
+      new DrawTextField([label], { x: x - lineWidth / 2, y }, this.options.drawLabelOptions).draw(canvasArg);
+    }
+  }
+};
+
+// src/draw/drawDetections.ts
+function drawDetections(canvasArg, detections) {
+  const detectionsArray = Array.isArray(detections) ? detections : [detections];
+  detectionsArray.forEach((det) => {
+    const score = det instanceof FaceDetection ? det.score : isWithFaceDetection(det) ? det.detection.score : void 0;
+    const box = det instanceof FaceDetection ? det.box : isWithFaceDetection(det) ? det.detection.box : new Box(det);
+    const label = score ? `${round5(score)}` : void 0;
+    new DrawBox(box, { label }).draw(canvasArg);
+  });
+}
+
+// src/dom/isMediaLoaded.ts
+function isMediaLoaded(media) {
+  const { Image, Video } = env2.getEnv();
+  return media instanceof Image && media.complete || media instanceof Video && media.readyState >= 3;
+}
+
+// src/dom/awaitMediaLoaded.ts
+function awaitMediaLoaded(media) {
+  return new Promise((resolve, reject) => {
+    if (media instanceof env2.getEnv().Canvas || isMediaLoaded(media))
+      resolve(null);
+    function onError(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      reject(e);
+    }
+    function onLoad(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      resolve(e);
+    }
+    media.addEventListener("load", onLoad);
+    media.addEventListener("error", onError);
+  });
+}
+
+// src/dom/bufferToImage.ts
+function bufferToImage(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToImage - expected buf to be of type: Blob"));
+    const reader = new FileReader();
+    reader.onload = () => {
+      if (typeof reader.result !== "string")
+        reject(new Error("bufferToImage - expected reader.result to be a string, in onload"));
+      const img = env2.getEnv().createImageElement();
+      img.onload = () => resolve(img);
+      img.onerror = reject;
+      img.src = reader.result;
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(buf);
+  });
+}
+
+// src/dom/getMediaDimensions.ts
+function getMediaDimensions(input2) {
+  const { Image, Video } = env2.getEnv();
+  if (input2 instanceof Image) {
+    return new Dimensions(input2.naturalWidth, input2.naturalHeight);
+  }
+  if (input2 instanceof Video) {
+    return new Dimensions(input2.videoWidth, input2.videoHeight);
+  }
+  return new Dimensions(input2.width, input2.height);
+}
+
+// src/dom/createCanvas.ts
+function createCanvas2({ width, height }) {
+  const { createCanvasElement } = env2.getEnv();
+  const canvas = createCanvasElement();
+  canvas.width = width;
+  canvas.height = height;
+  return canvas;
+}
+function createCanvasFromMedia(media, dims) {
+  const { ImageData: ImageData2 } = env2.getEnv();
+  if (!(media instanceof ImageData2) && !isMediaLoaded(media)) {
+    throw new Error("createCanvasFromMedia - media has not finished loading yet");
+  }
+  const { width, height } = dims || getMediaDimensions(media);
+  const canvas = createCanvas2({ width, height });
+  if (media instanceof ImageData2) {
+    getContext2dOrThrow(canvas).putImageData(media, 0, 0);
+  } else {
+    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);
+  }
+  return canvas;
+}
+
+// src/dom/imageTensorToCanvas.ts
+async function imageTensorToCanvas(imgTensor, canvas) {
+  const targetCanvas = canvas || env2.getEnv().createCanvasElement();
+  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);
+  const imgTensor3D = tidy(() => imgTensor.as3D(height, width, numChannels).toInt());
+  await browser_exports.toPixels(imgTensor3D, targetCanvas);
+  imgTensor3D.dispose();
+  return targetCanvas;
+}
+
+// src/dom/isMediaElement.ts
+function isMediaElement(input2) {
+  const { Image, Canvas, Video } = env2.getEnv();
+  return input2 instanceof Image || input2 instanceof Canvas || input2 instanceof Video;
+}
+
+// src/dom/imageToSquare.ts
+function imageToSquare(input2, inputSize, centerImage = false) {
+  const { Image, Canvas } = env2.getEnv();
+  if (!(input2 instanceof Image || input2 instanceof Canvas)) {
+    throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+  }
+  if (inputSize <= 0)
+    return createCanvas2({ width: 1, height: 1 });
+  const dims = getMediaDimensions(input2);
+  const scale3 = inputSize / Math.max(dims.height, dims.width);
+  const width = scale3 * dims.width;
+  const height = scale3 * dims.height;
+  const targetCanvas = createCanvas2({ width: inputSize, height: inputSize });
+  const inputCanvas = input2 instanceof Canvas ? input2 : createCanvasFromMedia(input2);
+  const offset = Math.abs(width - height) / 2;
+  const dx = centerImage && width < height ? offset : 0;
+  const dy = centerImage && height < width ? offset : 0;
+  if (inputCanvas.width > 0 && inputCanvas.height > 0)
+    getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);
+  return targetCanvas;
+}
+
+// src/dom/NetInput.ts
+var NetInput = class {
+  constructor(inputs, treatAsBatchInput = false) {
+    this._imageTensors = [];
+    this._canvases = [];
+    this._treatAsBatchInput = false;
+    this._inputDimensions = [];
+    this._inputSize = 0;
+    if (!Array.isArray(inputs)) {
+      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);
+    }
+    this._treatAsBatchInput = treatAsBatchInput;
+    this._batchSize = inputs.length;
+    inputs.forEach((input2, idx) => {
+      if (isTensor3D(input2)) {
+        this._imageTensors[idx] = input2;
+        this._inputDimensions[idx] = input2.shape;
+        return;
+      }
+      if (isTensor4D(input2)) {
+        const batchSize = input2.shape[0];
+        if (batchSize !== 1) {
+          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+        }
+        this._imageTensors[idx] = input2;
+        this._inputDimensions[idx] = input2.shape.slice(1);
+        return;
+      }
+      const canvas = input2 instanceof env2.getEnv().Canvas ? input2 : createCanvasFromMedia(input2);
+      this._canvases[idx] = canvas;
+      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];
+    });
+  }
+  get imageTensors() {
+    return this._imageTensors;
+  }
+  get canvases() {
+    return this._canvases;
+  }
+  get isBatchInput() {
+    return this.batchSize > 1 || this._treatAsBatchInput;
+  }
+  get batchSize() {
+    return this._batchSize;
+  }
+  get inputDimensions() {
+    return this._inputDimensions;
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get reshapedInputDimensions() {
+    return range6(this.batchSize, 0, 1).map(
+      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx)
+    );
+  }
+  getInput(batchIdx) {
+    return this.canvases[batchIdx] || this.imageTensors[batchIdx];
+  }
+  getInputDimensions(batchIdx) {
+    return this._inputDimensions[batchIdx];
+  }
+  getInputHeight(batchIdx) {
+    return this._inputDimensions[batchIdx][0];
+  }
+  getInputWidth(batchIdx) {
+    return this._inputDimensions[batchIdx][1];
+  }
+  getReshapedInputDimensions(batchIdx) {
+    if (typeof this.inputSize !== "number") {
+      throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+    }
+    const width = this.getInputWidth(batchIdx);
+    const height = this.getInputHeight(batchIdx);
+    return computeReshapedDimensions({ width, height }, this.inputSize);
+  }
+  toBatchTensor(inputSize, isCenterInputs = true) {
+    this._inputSize = inputSize;
+    return tidy(() => {
+      const inputTensors = range6(this.batchSize, 0, 1).map((batchIdx) => {
+        const input2 = this.getInput(batchIdx);
+        if (input2 instanceof Tensor) {
+          let imgTensor = isTensor4D(input2) ? input2 : expandDims(input2);
+          imgTensor = padToSquare(imgTensor, isCenterInputs);
+          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {
+            imgTensor = image.resizeBilinear(imgTensor, [inputSize, inputSize], false, false);
+          }
+          return imgTensor.as3D(inputSize, inputSize, 3);
+        }
+        if (input2 instanceof env2.getEnv().Canvas) {
+          return browser_exports.fromPixels(imageToSquare(input2, inputSize, isCenterInputs));
+        }
+        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input2}`);
+      });
+      const batchTensor = stack(inputTensors.map((t) => cast(t, "float32"))).as4D(this.batchSize, inputSize, inputSize, 3);
+      return batchTensor;
+    });
+  }
+};
+
+// src/dom/toNetInput.ts
+async function toNetInput(inputs) {
+  if (inputs instanceof NetInput)
+    return inputs;
+  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];
+  if (!inputArgArray.length)
+    throw new Error("toNetInput - empty array passed as input");
+  const getIdxHint = (idx) => Array.isArray(inputs) ? ` at input index ${idx}:` : "";
+  const inputArray = inputArgArray.map(resolveInput);
+  inputArray.forEach((input2, i) => {
+    if (!isMediaElement(input2) && !isTensor3D(input2) && !isTensor4D(input2)) {
+      if (typeof inputArgArray[i] === "string")
+        throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);
+      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);
+    }
+    if (isTensor4D(input2)) {
+      const batchSize = input2.shape[0];
+      if (batchSize !== 1)
+        throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+    }
+  });
+  await Promise.all(inputArray.map((input2) => isMediaElement(input2) && awaitMediaLoaded(input2)));
+  return new NetInput(inputArray, Array.isArray(inputs));
+}
+
+// src/dom/extractFaces.ts
+async function extractFaces(input2, detections) {
+  const { Canvas } = env2.getEnv();
+  let canvas = input2;
+  if (!(input2 instanceof Canvas)) {
+    const netInput = await toNetInput(input2);
+    if (netInput.batchSize > 1)
+      throw new Error("extractFaces - batchSize > 1 not supported");
+    const tensorOrCanvas = netInput.getInput(0);
+    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);
+  }
+  const ctx = getContext2dOrThrow(canvas);
+  const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det).map((box) => box.clipAtImageBorders(canvas.width, canvas.height));
+  return boxes.map(({ x, y, width, height }) => {
+    const faceImg = createCanvas2({ width, height });
+    if (width > 0 && height > 0)
+      getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);
+    return faceImg;
+  });
+}
+
+// src/dom/extractFaceTensors.ts
+async function extractFaceTensors(imageTensor, detections) {
+  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {
+    throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+  }
+  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {
+    throw new Error("extractFaceTensors - batchSize > 1 not supported");
+  }
+  return tidy(() => {
+    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);
+    const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det).map((box) => box.clipAtImageBorders(imgWidth, imgHeight));
+    const faceTensors = boxes.filter((box) => box.width > 0 && box.height > 0).map(({ x, y, width, height }) => slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));
+    return faceTensors;
+  });
+}
+
+// src/dom/fetchOrThrow.ts
+async function fetchOrThrow(url, init2) {
+  const { fetch: fetch4 } = env2.getEnv();
+  const res = await fetch4(url, init2);
+  if (!(res.status < 400)) {
+    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);
+  }
+  return res;
+}
+
+// src/dom/fetchImage.ts
+async function fetchImage(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("image/")) {
+    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToImage(blob);
+}
+
+// src/dom/fetchJson.ts
+async function fetchJson(uri) {
+  return (await fetchOrThrow(uri)).json();
+}
+
+// src/dom/fetchNetWeights.ts
+async function fetchNetWeights(uri) {
+  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());
+}
+
+// src/dom/bufferToVideo.ts
+function bufferToVideo(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToVideo - expected buf to be of type: Blob"));
+    const video = env2.getEnv().createVideoElement();
+    video.oncanplay = () => resolve(video);
+    video.onerror = reject;
+    video.playsInline = true;
+    video.muted = true;
+    video.src = URL.createObjectURL(buf);
+    video.play();
+  });
+}
+
+// src/dom/fetchVideo.ts
+async function fetchVideo(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("video/")) {
+    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToVideo(blob);
+}
+
+// src/common/getModelUris.ts
+function getModelUris(uri, defaultModelName) {
+  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;
+  if (!uri) {
+    return {
+      modelBaseUri: "",
+      manifestUri: defaultManifestFilename
+    };
+  }
+  if (uri === "/") {
+    return {
+      modelBaseUri: "/",
+      manifestUri: `/${defaultManifestFilename}`
+    };
+  }
+  const protocol = uri.startsWith("http://") ? "http://" : uri.startsWith("https://") ? "https://" : "";
+  uri = uri.replace(protocol, "");
+  const parts = uri.split("/").filter((s) => s);
+  const manifestFile = uri.endsWith(".json") ? parts[parts.length - 1] : defaultManifestFilename;
+  let modelBaseUri = protocol + (uri.endsWith(".json") ? parts.slice(0, parts.length - 1) : parts).join("/");
+  modelBaseUri = uri.startsWith("/") ? `/${modelBaseUri}` : modelBaseUri;
+  return {
+    modelBaseUri,
+    manifestUri: modelBaseUri === "/" ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`
+  };
+}
+
+// src/dom/loadWeightMap.ts
+async function loadWeightMap(uri, defaultModelName) {
+  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);
+  const manifest = await fetchJson(manifestUri);
+  return io_exports.loadWeights(manifest, modelBaseUri);
+}
+
+// src/dom/matchDimensions.ts
+function matchDimensions(input2, reference, useMediaDimensions = false) {
+  const { width, height } = useMediaDimensions ? getMediaDimensions(reference) : reference;
+  input2.width = width;
+  input2.height = height;
+  return { width, height };
+}
+
+// src/NeuralNetwork.ts
+var NeuralNetwork = class {
+  constructor(name) {
+    this._params = void 0;
+    this._paramMappings = [];
+    this._name = name;
+  }
+  get params() {
+    return this._params;
+  }
+  get paramMappings() {
+    return this._paramMappings;
+  }
+  get isLoaded() {
+    return !!this.params;
+  }
+  getParamFromPath(paramPath) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    return obj[objProp];
+  }
+  reassignParamFromPath(paramPath, tensor2) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    obj[objProp].dispose();
+    obj[objProp] = tensor2;
+  }
+  getParamList() {
+    return this._paramMappings.map(({ paramPath }) => ({
+      path: paramPath,
+      tensor: this.getParamFromPath(paramPath)
+    }));
+  }
+  getTrainableParams() {
+    return this.getParamList().filter((param) => param.tensor instanceof Variable);
+  }
+  getFrozenParams() {
+    return this.getParamList().filter((param) => !(param.tensor instanceof Variable));
+  }
+  variable() {
+    this.getFrozenParams().forEach(({ path, tensor: tensor2 }) => {
+      this.reassignParamFromPath(path, tensor2.variable());
+    });
+  }
+  freeze() {
+    this.getTrainableParams().forEach(({ path, tensor: variable2 }) => {
+      const tensor2 = tensor(variable2.dataSync());
+      variable2.dispose();
+      this.reassignParamFromPath(path, tensor2);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.getParamList().forEach((param) => {
+      if (throwOnRedispose && param.tensor.isDisposed) {
+        throw new Error(`param tensor has already been disposed for path ${param.path}`);
+      }
+      param.tensor.dispose();
+    });
+    this._params = void 0;
+  }
+  serializeParams() {
+    return new Float32Array(
+      this.getParamList().map(({ tensor: tensor2 }) => Array.from(tensor2.dataSync())).reduce((flat, arr) => flat.concat(arr))
+    );
+  }
+  async load(weightsOrUrl) {
+    if (weightsOrUrl instanceof Float32Array) {
+      this.extractWeights(weightsOrUrl);
+      return;
+    }
+    await this.loadFromUri(weightsOrUrl);
+  }
+  async loadFromUri(uri) {
+    if (uri && typeof uri !== "string") {
+      throw new Error(`${this._name}.loadFromUri - expected model uri`);
+    }
+    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());
+    this.loadFromWeightMap(weightMap);
+  }
+  async loadFromDisk(filePath) {
+    if (filePath && typeof filePath !== "string") {
+      throw new Error(`${this._name}.loadFromDisk - expected model file path`);
+    }
+    const { readFile } = env2.getEnv();
+    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());
+    const fetchWeightsFromDisk = (filePaths) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));
+    const loadWeights2 = io_exports.weightsLoaderFactory(fetchWeightsFromDisk);
+    const manifest = JSON.parse((await readFile(manifestUri)).toString());
+    const weightMap = await loadWeights2(manifest, modelBaseUri);
+    this.loadFromWeightMap(weightMap);
+  }
+  loadFromWeightMap(weightMap) {
+    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  extractWeights(weights) {
+    const { paramMappings, params } = this.extractParams(weights);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  traversePropertyPath(paramPath) {
+    if (!this.params) {
+      throw new Error("traversePropertyPath - model has no loaded params");
+    }
+    const result = paramPath.split("/").reduce((res, objProp2) => {
+      if (!res.nextObj.hasOwnProperty(objProp2)) {
+        throw new Error(`traversePropertyPath - object does not have property ${objProp2}, for path ${paramPath}`);
+      }
+      return { obj: res.nextObj, objProp: objProp2, nextObj: res.nextObj[objProp2] };
+    }, { nextObj: this.params });
+    const { obj, objProp } = result;
+    if (!obj || !objProp || !(obj[objProp] instanceof Tensor)) {
+      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);
+    }
+    return { obj, objProp };
+  }
+};
+
+// src/common/depthwiseSeparableConv.ts
+function depthwiseSeparableConv(x, params, stride) {
+  return tidy(() => {
+    let out = separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, "same");
+    out = add2(out, params.bias);
+    return out;
+  });
+}
+
+// src/faceFeatureExtractor/denseBlock.ts
+function denseBlock3(x, denseBlockParams, isFirstLayer = false) {
+  return tidy(() => {
+    const out1 = relu(
+      isFirstLayer ? add2(
+        conv2d(x, denseBlockParams.conv0.filters, [2, 2], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, [2, 2])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = relu(add2(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    return relu(add2(out1, add2(out2, out3)));
+  });
+}
+function denseBlock4(x, denseBlockParams, isFirstLayer = false, isScaleDown = true) {
+  return tidy(() => {
+    const out1 = relu(
+      isFirstLayer ? add2(
+        conv2d(x, denseBlockParams.conv0.filters, isScaleDown ? [2, 2] : [1, 1], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, isScaleDown ? [2, 2] : [1, 1])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = relu(add2(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    const in4 = relu(add2(out1, add2(out2, out3)));
+    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);
+    return relu(add2(out1, add2(out2, add2(out3, out4))));
+  });
+}
+
+// src/common/convLayer.ts
+function convLayer(x, params, padding = "same", withRelu = false) {
+  return tidy(() => {
+    const out = add2(
+      conv2d(x, params.filters, [1, 1], padding),
+      params.bias
+    );
+    return withRelu ? relu(out) : out;
+  });
+}
+
+// src/common/disposeUnusedWeightTensors.ts
+function disposeUnusedWeightTensors(weightMap, paramMappings) {
+  Object.keys(weightMap).forEach((path) => {
+    if (!paramMappings.some((pm) => pm.originalPath === path)) {
+      weightMap[path].dispose();
+    }
+  });
+}
+
+// src/common/extractConvParamsFactory.ts
+function extractConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, filterSize, mappedPrefix) => {
+    const filters = tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  };
+}
+
+// src/common/extractFCParamsFactory.ts
+function extractFCParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const fc_weights = tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);
+    const fc_bias = tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return {
+      weights: fc_weights,
+      bias: fc_bias
+    };
+  };
+}
+
+// src/common/types.ts
+var SeparableConvParams = class {
+  constructor(depthwise_filter, pointwise_filter, bias) {
+    this.depthwise_filter = depthwise_filter;
+    this.pointwise_filter = pointwise_filter;
+    this.bias = bias;
+  }
+};
+
+// src/common/extractSeparableConvParamsFactory.ts
+function extractSeparableConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const depthwise_filter = tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);
+    const pointwise_filter = tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);
+    const bias = tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/depthwise_filter` },
+      { paramPath: `${mappedPrefix}/pointwise_filter` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+function loadSeparableConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const depthwise_filter = extractWeightEntry(`${prefix}/depthwise_filter`, 4);
+    const pointwise_filter = extractWeightEntry(`${prefix}/pointwise_filter`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+
+// src/common/extractWeightEntryFactory.ts
+function extractWeightEntryFactory(weightMap, paramMappings) {
+  return (originalPath, paramRank, mappedPath) => {
+    const tensor2 = weightMap[originalPath];
+    if (!isTensor(tensor2, paramRank)) {
+      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor2}`);
+    }
+    paramMappings.push(
+      { originalPath, paramPath: mappedPath || originalPath }
+    );
+    return tensor2;
+  };
+}
+
+// src/common/extractWeightsFactory.ts
+function extractWeightsFactory(weights) {
+  let remainingWeights = weights;
+  function extractWeights(numWeights) {
+    const ret = remainingWeights.slice(0, numWeights);
+    remainingWeights = remainingWeights.slice(numWeights);
+    return ret;
+  }
+  function getRemainingWeights() {
+    return remainingWeights;
+  }
+  return {
+    extractWeights,
+    getRemainingWeights
+  };
+}
+
+// src/faceFeatureExtractor/extractorsFactory.ts
+function extractorsFactory(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`) : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);
+    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);
+    const conv22 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const { conv0, conv1, conv2: conv22 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);
+    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParams.ts
+function extractParams(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock4Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock4Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock4Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock4Params(64, 128, "dense2");
+  const dense3 = extractDenseBlock4Params(128, 256, "dense3");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: {
+      dense0,
+      dense1,
+      dense2,
+      dense3
+    }
+  };
+}
+
+// src/common/loadConvParamsFactory.ts
+function loadConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  };
+}
+
+// src/faceFeatureExtractor/loadParamsFactory.ts
+function loadParamsFactory(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractDenseBlock3Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock4Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock4Params("dense0", true),
+    dense1: extractDenseBlock4Params("dense1"),
+    dense2: extractDenseBlock4Params("dense2"),
+    dense3: extractDenseBlock4Params("dense3")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var FaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("FaceFeatureExtractor");
+  }
+  forwardInput(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceFeatureExtractor - load model before inference");
+    }
+    return tidy(() => {
+      const batchTensor = cast(input2.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock4(normalized, params.dense0, true);
+      out = denseBlock4(out, params.dense1);
+      out = denseBlock4(out, params.dense2);
+      out = denseBlock4(out, params.dense3);
+      out = avgPool(out, [7, 7], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams(weights);
+  }
+};
+
+// src/common/fullyConnectedLayer.ts
+function fullyConnectedLayer(x, params) {
+  return tidy(() => add2(
+    matMul(x, params.weights),
+    params.bias
+  ));
+}
+
+// src/faceProcessor/extractParams.ts
+function extractParams2(weights, channelsIn, channelsOut) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const fc = extractFCParams(channelsIn, channelsOut, "fc");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc }
+  };
+}
+
+// src/faceProcessor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap2(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: extractFcParams("fc")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceProcessor/util.ts
+function seperateWeightMaps(weightMap) {
+  const featureExtractorMap = {};
+  const classifierMap = {};
+  Object.keys(weightMap).forEach((key) => {
+    const map = key.startsWith("fc") ? classifierMap : featureExtractorMap;
+    map[key] = weightMap[key];
+  });
+  return { featureExtractorMap, classifierMap };
+}
+
+// src/faceProcessor/FaceProcessor.ts
+var FaceProcessor = class extends NeuralNetwork {
+  constructor(_name, faceFeatureExtractor) {
+    super(_name);
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tidy(() => {
+      const bottleneckFeatures = input2 instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input2) : input2;
+      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams2(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap2(classifierMap);
+  }
+  extractParams(weights) {
+    const cIn = this.getClassifierChannelsIn();
+    const cOut = this.getClassifierChannelsOut();
+    const classifierWeightSize = cOut * cIn + cOut;
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressions.ts
+var FACE_EXPRESSION_LABELS = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"];
+var FaceExpressions = class {
+  constructor(probabilities) {
+    this.neutral = 0;
+    this.happy = 0;
+    this.sad = 0;
+    this.angry = 0;
+    this.fearful = 0;
+    this.disgusted = 0;
+    this.surprised = 0;
+    if (probabilities.length !== 7) {
+      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);
+    }
+    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {
+      this[expression] = probabilities[idx];
+    });
+  }
+  asSortedArray() {
+    return FACE_EXPRESSION_LABELS.map((expression) => ({ expression, probability: this[expression] })).sort((e0, e1) => e1.probability - e0.probability);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var FaceExpressionNet = class extends FaceProcessor {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceExpressionNet", faceFeatureExtractor);
+  }
+  forwardInput(input2) {
+    return tidy(() => softmax(this.runNet(input2)));
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  async predictExpressions(input2) {
+    const netInput = await toNetInput(input2);
+    const out = await this.forwardInput(netInput);
+    const probabilitesByBatch = await Promise.all(unstack(out).map(async (t) => {
+      const data = t.dataSync();
+      t.dispose();
+      return data;
+    }));
+    out.dispose();
+    const predictionsByBatch = probabilitesByBatch.map((probabilites) => new FaceExpressions(probabilites));
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_expression_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+  getClassifierChannelsOut() {
+    return 7;
+  }
+};
+
+// src/factories/WithFaceExpressions.ts
+function isWithFaceExpressions(obj) {
+  return obj.expressions instanceof FaceExpressions;
+}
+function extendWithFaceExpressions(sourceObj, expressions) {
+  const extension = { expressions };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/drawFaceExpressions.ts
+function drawFaceExpressions(canvasArg, faceExpressions, minConfidence = 0.1, textFieldAnchor) {
+  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];
+  faceExpressionsArray.forEach((e) => {
+    const expr = e instanceof FaceExpressions ? e : isWithFaceExpressions(e) ? e.expressions : void 0;
+    if (!expr) {
+      throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+    }
+    const sorted = expr.asSortedArray();
+    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);
+    const anchor = isWithFaceDetection(e) ? e.detection.box.bottomLeft : textFieldAnchor || new Point(0, 0);
+    const drawTextField = new DrawTextField(
+      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round5(exprLocal.probability)})`),
+      anchor
+    );
+    drawTextField.draw(canvasArg);
+  });
+}
+
+// src/factories/WithFaceLandmarks.ts
+function isWithFaceLandmarks(obj) {
+  return isWithFaceDetection(obj) && obj["landmarks"] instanceof FaceLandmarks && obj["unshiftedLandmarks"] instanceof FaceLandmarks && obj["alignedRect"] instanceof FaceDetection;
+}
+function calculateFaceAngle(mesh) {
+  const degrees = (radians) => radians * 180 / Math.PI;
+  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);
+  const angle = {
+    roll: void 0,
+    pitch: void 0,
+    yaw: void 0
+  };
+  const calcYaw = (leftPoint, midPoint, rightPoint) => {
+    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);
+    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);
+    return leftToMidpoint - rightToMidpoint;
+  };
+  const calcRoll = (lever, pivot) => {
+    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);
+    const opposite = pivot._y - lever._y;
+    const angleInRadians = Math.asin(opposite / hypotenuse);
+    const angleInDegrees = degrees(angleInRadians);
+    const normalizeAngle = Math.floor(90 - angleInDegrees);
+    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;
+    const result = normalizeAngle * tiltDirection;
+    return result;
+  };
+  const calcPitch = (leftPoint, midPoint, rightPoint) => {
+    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);
+    const baseCoords = {
+      _x: (leftPoint._x + rightPoint._x) / 2,
+      _y: (leftPoint._y + rightPoint._y) / 2
+    };
+    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);
+    const angleInRadians = Math.atan(midToBaseLength / base);
+    const angleInDegrees = Math.floor(degrees(angleInRadians));
+    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;
+    const result = angleInDegrees * direction;
+    return result;
+  };
+  if (!mesh || !mesh._positions || mesh._positions.length !== 68)
+    return angle;
+  const pt = mesh._positions;
+  angle.roll = calcRoll(pt[27], pt[66]);
+  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);
+  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);
+  return angle;
+}
+function extendWithFaceLandmarks(sourceObj, unshiftedLandmarks) {
+  const { box: shift } = sourceObj.detection;
+  const landmarks = unshiftedLandmarks.shiftBy(shift.x, shift.y);
+  const rect = landmarks.align();
+  const { imageDims } = sourceObj.detection;
+  const alignedRect = new FaceDetection(
+    sourceObj.detection.score,
+    rect.rescale(imageDims.reverse()),
+    imageDims
+  );
+  const angle = calculateFaceAngle(unshiftedLandmarks);
+  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/DrawFaceLandmarks.ts
+var DrawFaceLandmarksOptions = class {
+  constructor(options = {}) {
+    const {
+      drawLines = true,
+      drawPoints = true,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = options;
+    this.drawLines = drawLines;
+    this.drawPoints = drawPoints;
+    this.lineWidth = lineWidth || 1;
+    this.pointSize = pointSize || 2;
+    this.lineColor = lineColor || "rgba(0, 255, 255, 1)";
+    this.pointColor = pointColor || "rgba(255, 0, 255, 1)";
+  }
+};
+var DrawFaceLandmarks = class {
+  constructor(faceLandmarks, options = {}) {
+    this.faceLandmarks = faceLandmarks;
+    this.options = new DrawFaceLandmarksOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const {
+      drawLines,
+      drawPoints,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = this.options;
+    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {
+      ctx.strokeStyle = lineColor;
+      ctx.lineWidth = lineWidth;
+      drawContour(ctx, this.faceLandmarks.getJawOutline());
+      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getNose());
+      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);
+      drawContour(ctx, this.faceLandmarks.getRightEye(), true);
+      drawContour(ctx, this.faceLandmarks.getMouth(), true);
+    }
+    if (drawPoints) {
+      ctx.strokeStyle = pointColor;
+      ctx.fillStyle = pointColor;
+      const drawPoint = (pt) => {
+        ctx.beginPath();
+        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);
+        ctx.fill();
+      };
+      this.faceLandmarks.positions.forEach(drawPoint);
+    }
+  }
+};
+function drawFaceLandmarks(canvasArg, faceLandmarks) {
+  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];
+  faceLandmarksArray.forEach((f) => {
+    const landmarks = f instanceof FaceLandmarks ? f : isWithFaceLandmarks(f) ? f.landmarks : void 0;
+    if (!landmarks) {
+      throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+    }
+    new DrawFaceLandmarks(landmarks).draw(canvasArg);
+  });
+}
+
+// package.json
+var version5 = "1.7.6";
+
+// src/xception/extractParams.ts
+function extractorsFactory2(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractReductionBlockParams(channelsIn, channelsOut, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(channels, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParams3(weights, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = extractorsFactory2(extractWeights, paramMappings);
+  const entry_flow_conv_in = extractConvParams(3, 32, 3, "entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, "entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, "entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range6(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, "exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, "exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { entry_flow, middle_flow, exit_flow }
+  };
+}
+
+// src/xception/extractParamsFromWeightMap.ts
+function loadParamsFactory2(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractReductionBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParamsFromWeightMap3(weightMap, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = loadParamsFactory2(weightMap, paramMappings);
+  const entry_flow_conv_in = extractConvParams("entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams("entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams("entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range6(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams("exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams("exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };
+}
+
+// src/xception/TinyXception.ts
+function conv(x, params, stride) {
+  return add2(conv2d(x, params.filters, stride, "same"), params.bias);
+}
+function reductionBlock(x, params, isActivateInput = true) {
+  let out = isActivateInput ? relu(x) : x;
+  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(relu(out), params.separable_conv1, [1, 1]);
+  out = maxPool(out, [3, 3], [2, 2], "same");
+  out = add2(out, conv(x, params.expansion_conv, [2, 2]));
+  return out;
+}
+function mainBlock(x, params) {
+  let out = depthwiseSeparableConv(relu(x), params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(relu(out), params.separable_conv1, [1, 1]);
+  out = depthwiseSeparableConv(relu(out), params.separable_conv2, [1, 1]);
+  out = add2(out, x);
+  return out;
+}
+var TinyXception = class extends NeuralNetwork {
+  constructor(numMainBlocks) {
+    super("TinyXception");
+    this._numMainBlocks = numMainBlocks;
+  }
+  forwardInput(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyXception - load model before inference");
+    }
+    return tidy(() => {
+      const batchTensor = cast(input2.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));
+      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);
+      out = reductionBlock(out, params.entry_flow.reduction_block_1);
+      range6(this._numMainBlocks, 0, 1).forEach((idx) => {
+        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);
+      });
+      out = reductionBlock(out, params.exit_flow.reduction_block);
+      out = relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));
+      return out;
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  getDefaultModelName() {
+    return "tiny_xception_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap3(weightMap, this._numMainBlocks);
+  }
+  extractParams(weights) {
+    return extractParams3(weights, this._numMainBlocks);
+  }
+};
+
+// src/ageGenderNet/extractParams.ts
+function extractParams4(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const age = extractFCParams(512, 1, "fc/age");
+  const gender = extractFCParams(512, 2, "fc/gender");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc: { age, gender } }
+  };
+}
+
+// src/ageGenderNet/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap4(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: {
+      age: extractFcParams("fc/age"),
+      gender: extractFcParams("fc/gender")
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ageGenderNet/types.ts
+var Gender = /* @__PURE__ */ ((Gender2) => {
+  Gender2["FEMALE"] = "female";
+  Gender2["MALE"] = "male";
+  return Gender2;
+})(Gender || {});
+
+// src/ageGenderNet/AgeGenderNet.ts
+var AgeGenderNet = class extends NeuralNetwork {
+  constructor(faceFeatureExtractor = new TinyXception(2)) {
+    super("AgeGenderNet");
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tidy(() => {
+      const bottleneckFeatures = input2 instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input2) : input2;
+      const pooled = avgPool(bottleneckFeatures, [7, 7], [2, 2], "valid").as2D(bottleneckFeatures.shape[0], -1);
+      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();
+      const gender = fullyConnectedLayer(pooled, params.fc.gender);
+      return { age, gender };
+    });
+  }
+  forwardInput(input2) {
+    return tidy(() => {
+      const { age, gender } = this.runNet(input2);
+      return { age, gender: softmax(gender) };
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  async predictAgeAndGender(input2) {
+    const netInput = await toNetInput(input2);
+    const out = await this.forwardInput(netInput);
+    const ages = unstack(out.age);
+    const genders = unstack(out.gender);
+    const ageAndGenderTensors = ages.map((ageTensor, i) => ({
+      ageTensor,
+      genderTensor: genders[i]
+    }));
+    const predictionsByBatch = await Promise.all(
+      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {
+        const age = ageTensor.dataSync()[0];
+        const probMale = genderTensor.dataSync()[0];
+        const isMale = probMale > 0.5;
+        const gender = isMale ? "male" /* MALE */ : "female" /* FEMALE */;
+        const genderProbability = isMale ? probMale : 1 - probMale;
+        ageTensor.dispose();
+        genderTensor.dispose();
+        return { age, gender, genderProbability };
+      })
+    );
+    out.age.dispose();
+    out.gender.dispose();
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "age_gender_model";
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams4(weights);
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap4(classifierMap);
+  }
+  extractParams(weights) {
+    const classifierWeightSize = 512 * 1 + 1 + (512 * 2 + 2);
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68NetBase.ts
+var FaceLandmark68NetBase = class extends FaceProcessor {
+  postProcess(output, inputSize, originalDimensions) {
+    const inputDimensions = originalDimensions.map(({ width, height }) => {
+      const scale3 = inputSize / Math.max(height, width);
+      return {
+        width: width * scale3,
+        height: height * scale3
+      };
+    });
+    const batchSize = inputDimensions.length;
+    return tidy(() => {
+      const createInterleavedTensor = (fillX, fillY) => stack([fill([68], fillX, "float32"), fill([68], fillY, "float32")], 1).as2D(1, 136).as1D();
+      const getPadding2 = (batchIdx, cond) => {
+        const { width, height } = inputDimensions[batchIdx];
+        return cond(width, height) ? Math.abs(width - height) / 2 : 0;
+      };
+      const getPaddingX = (batchIdx) => getPadding2(batchIdx, (w, h) => w < h);
+      const getPaddingY = (batchIdx) => getPadding2(batchIdx, (w, h) => h < w);
+      const landmarkTensors = output.mul(fill([batchSize, 136], inputSize, "float32")).sub(stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        getPaddingX(batchIdx),
+        getPaddingY(batchIdx)
+      )))).div(stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        inputDimensions[batchIdx].width,
+        inputDimensions[batchIdx].height
+      ))));
+      return landmarkTensors;
+    });
+  }
+  forwardInput(input2) {
+    return tidy(() => {
+      const out = this.runNet(input2);
+      return this.postProcess(
+        out,
+        input2.inputSize,
+        input2.inputDimensions.map(([height, width]) => ({ height, width }))
+      );
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  async detectLandmarks(input2) {
+    const netInput = await toNetInput(input2);
+    const landmarkTensors = tidy(
+      () => unstack(this.forwardInput(netInput))
+    );
+    const landmarksForBatch = await Promise.all(landmarkTensors.map(
+      async (landmarkTensor, batchIdx) => {
+        const landmarksArray = Array.from(landmarkTensor.dataSync());
+        const xCoords = landmarksArray.filter((_, i) => isEven2(i));
+        const yCoords = landmarksArray.filter((_, i) => !isEven2(i));
+        return new FaceLandmarks68(
+          Array(68).fill(0).map((_, i) => new Point(xCoords[i], yCoords[i])),
+          {
+            height: netInput.getInputHeight(batchIdx),
+            width: netInput.getInputWidth(batchIdx)
+          }
+        );
+      }
+    ));
+    landmarkTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? landmarksForBatch : landmarksForBatch[0];
+  }
+  getClassifierChannelsOut() {
+    return 136;
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68Net.ts
+var FaceLandmark68Net = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceLandmark68Net", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+};
+
+// src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts
+function extractParamsFromWeightMapTiny(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock3Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock3Params("dense0", true),
+    dense1: extractDenseBlock3Params("dense1"),
+    dense2: extractDenseBlock3Params("dense2")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/extractParamsTiny.ts
+function extractParamsTiny(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock3Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock3Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock3Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock3Params(64, 128, "dense2");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { dense0, dense1, dense2 }
+  };
+}
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var TinyFaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("TinyFaceFeatureExtractor");
+  }
+  forwardInput(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyFaceFeatureExtractor - load model before inference");
+    }
+    return tidy(() => {
+      const batchTensor = cast(input2.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock3(normalized, params.dense0, true);
+      out = denseBlock3(out, params.dense1);
+      out = denseBlock3(out, params.dense2);
+      out = avgPool(out, [14, 14], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_tiny_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMapTiny(weightMap);
+  }
+  extractParams(weights) {
+    return extractParamsTiny(weights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68TinyNet.ts
+var FaceLandmark68TinyNet = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new TinyFaceFeatureExtractor()) {
+    super("FaceLandmark68TinyNet", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_tiny_model";
+  }
+  getClassifierChannelsIn() {
+    return 128;
+  }
+};
+
+// src/faceLandmarkNet/index.ts
+var FaceLandmarkNet = class extends FaceLandmark68Net {
+};
+
+// src/faceRecognitionNet/scaleLayer.ts
+function scale2(x, params) {
+  return add2(mul(x, params.weights), params.biases);
+}
+
+// src/faceRecognitionNet/convLayer.ts
+function convLayer2(x, params, strides, withRelu, padding = "same") {
+  const { filters, bias } = params.conv;
+  let out = conv2d(x, filters, strides, padding);
+  out = add2(out, bias);
+  out = scale2(out, params.scale);
+  return withRelu ? relu(out) : out;
+}
+function conv2(x, params) {
+  return convLayer2(x, params, [1, 1], true);
+}
+function convNoRelu(x, params) {
+  return convLayer2(x, params, [1, 1], false);
+}
+function convDown(x, params) {
+  return convLayer2(x, params, [2, 2], true, "valid");
+}
+
+// src/faceRecognitionNet/extractParams.ts
+function extractorsFactory3(extractWeights, paramMappings) {
+  function extractFilterValues(numFilterValues, numFilters, filterSize) {
+    const weights = extractWeights(numFilterValues);
+    const depth = weights.length / (numFilters * filterSize * filterSize);
+    if (isFloat(depth)) {
+      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);
+    }
+    return tidy(
+      () => transpose(
+        tensor4d(weights, [numFilters, depth, filterSize, filterSize]),
+        [2, 3, 1, 0]
+      )
+    );
+  }
+  function extractConvParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);
+    const bias = tensor1d(extractWeights(numFilters));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  }
+  function extractScaleLayerParams(numWeights, mappedPrefix) {
+    const weights = tensor1d(extractWeights(numWeights));
+    const biases = tensor1d(extractWeights(numWeights));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/biases` }
+    );
+    return {
+      weights,
+      biases
+    };
+  }
+  function extractConvLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const conv3 = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);
+    const scale3 = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);
+    return { conv: conv3, scale: scale3 };
+  }
+  function extractResidualLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix, isDown = false) {
+    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);
+    const conv22 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);
+    return { conv1, conv2: conv22 };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParams5(weights) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory3(extractWeights, paramMappings);
+  const conv32_down = extractConvLayerParams(4704, 32, 7, "conv32_down");
+  const conv32_1 = extractResidualLayerParams(9216, 32, 3, "conv32_1");
+  const conv32_2 = extractResidualLayerParams(9216, 32, 3, "conv32_2");
+  const conv32_3 = extractResidualLayerParams(9216, 32, 3, "conv32_3");
+  const conv64_down = extractResidualLayerParams(36864, 64, 3, "conv64_down", true);
+  const conv64_1 = extractResidualLayerParams(36864, 64, 3, "conv64_1");
+  const conv64_2 = extractResidualLayerParams(36864, 64, 3, "conv64_2");
+  const conv64_3 = extractResidualLayerParams(36864, 64, 3, "conv64_3");
+  const conv128_down = extractResidualLayerParams(147456, 128, 3, "conv128_down", true);
+  const conv128_1 = extractResidualLayerParams(147456, 128, 3, "conv128_1");
+  const conv128_2 = extractResidualLayerParams(147456, 128, 3, "conv128_2");
+  const conv256_down = extractResidualLayerParams(589824, 256, 3, "conv256_down", true);
+  const conv256_1 = extractResidualLayerParams(589824, 256, 3, "conv256_1");
+  const conv256_2 = extractResidualLayerParams(589824, 256, 3, "conv256_2");
+  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, "conv256_down_out");
+  const fc = tidy(
+    () => transpose(tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0])
+  );
+  paramMappings.push({ paramPath: "fc" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/extractParamsFromWeightMap.ts
+function extractorsFactory4(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractScaleLayerParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);
+    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);
+    return { weights, biases };
+  }
+  function extractConvLayerParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);
+    const scale3 = extractScaleLayerParams(prefix);
+    return { conv: { filters, bias }, scale: scale3 };
+  }
+  function extractResidualLayerParams(prefix) {
+    return {
+      conv1: extractConvLayerParams(`${prefix}/conv1`),
+      conv2: extractConvLayerParams(`${prefix}/conv2`)
+    };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParamsFromWeightMap5(weightMap) {
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory4(weightMap, paramMappings);
+  const conv32_down = extractConvLayerParams("conv32_down");
+  const conv32_1 = extractResidualLayerParams("conv32_1");
+  const conv32_2 = extractResidualLayerParams("conv32_2");
+  const conv32_3 = extractResidualLayerParams("conv32_3");
+  const conv64_down = extractResidualLayerParams("conv64_down");
+  const conv64_1 = extractResidualLayerParams("conv64_1");
+  const conv64_2 = extractResidualLayerParams("conv64_2");
+  const conv64_3 = extractResidualLayerParams("conv64_3");
+  const conv128_down = extractResidualLayerParams("conv128_down");
+  const conv128_1 = extractResidualLayerParams("conv128_1");
+  const conv128_2 = extractResidualLayerParams("conv128_2");
+  const conv256_down = extractResidualLayerParams("conv256_down");
+  const conv256_1 = extractResidualLayerParams("conv256_1");
+  const conv256_2 = extractResidualLayerParams("conv256_2");
+  const conv256_down_out = extractResidualLayerParams("conv256_down_out");
+  const { fc } = weightMap;
+  paramMappings.push({ originalPath: "fc", paramPath: "fc" });
+  if (!isTensor2D(fc)) {
+    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/residualLayer.ts
+function residual(x, params) {
+  let out = conv2(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  out = add2(out, x);
+  out = relu(out);
+  return out;
+}
+function residualDown(x, params) {
+  let out = convDown(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  let pooled = avgPool(x, 2, 2, "valid");
+  const zeros4 = zeros(pooled.shape);
+  const isPad = pooled.shape[3] !== out.shape[3];
+  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];
+  if (isAdjustShape) {
+    const padShapeX = [...out.shape];
+    padShapeX[1] = 1;
+    const zerosW = zeros(padShapeX);
+    out = concat([out, zerosW], 1);
+    const padShapeY = [...out.shape];
+    padShapeY[2] = 1;
+    const zerosH = zeros(padShapeY);
+    out = concat([out, zerosH], 2);
+  }
+  pooled = isPad ? concat([pooled, zeros4], 3) : pooled;
+  out = add2(pooled, out);
+  out = relu(out);
+  return out;
+}
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var FaceRecognitionNet = class extends NeuralNetwork {
+  constructor() {
+    super("FaceRecognitionNet");
+  }
+  forwardInput(input2) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceRecognitionNet - load model before inference");
+    }
+    return tidy(() => {
+      const batchTensor = cast(input2.toBatchTensor(150, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = convDown(normalized, params.conv32_down);
+      out = maxPool(out, 3, 2, "valid");
+      out = residual(out, params.conv32_1);
+      out = residual(out, params.conv32_2);
+      out = residual(out, params.conv32_3);
+      out = residualDown(out, params.conv64_down);
+      out = residual(out, params.conv64_1);
+      out = residual(out, params.conv64_2);
+      out = residual(out, params.conv64_3);
+      out = residualDown(out, params.conv128_down);
+      out = residual(out, params.conv128_1);
+      out = residual(out, params.conv128_2);
+      out = residualDown(out, params.conv256_down);
+      out = residual(out, params.conv256_1);
+      out = residual(out, params.conv256_2);
+      out = residualDown(out, params.conv256_down_out);
+      const globalAvg = out.mean([1, 2]);
+      const fullyConnected = matMul(globalAvg, params.fc);
+      return fullyConnected;
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  async computeFaceDescriptor(input2) {
+    var _a;
+    if ((_a = input2 == null ? void 0 : input2.shape) == null ? void 0 : _a.some((dim) => dim <= 0))
+      return new Float32Array(128);
+    const netInput = await toNetInput(input2);
+    const faceDescriptorTensors = tidy(() => unstack(this.forwardInput(netInput)));
+    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data()));
+    faceDescriptorTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_recognition_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap5(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams5(weights);
+  }
+};
+
+// src/faceRecognitionNet/index.ts
+function createFaceRecognitionNet(weights) {
+  const net = new FaceRecognitionNet();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/factories/WithFaceDescriptor.ts
+function extendWithFaceDescriptor(sourceObj, descriptor) {
+  const extension = { descriptor };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithAge.ts
+function isWithAge(obj) {
+  return typeof obj.age === "number";
+}
+function extendWithAge(sourceObj, age) {
+  const extension = { age };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithGender.ts
+function isWithGender(obj) {
+  return (obj.gender === "male" /* MALE */ || obj.gender === "female" /* FEMALE */) && isValidProbablitiy(obj.genderProbability);
+}
+function extendWithGender(sourceObj, gender, genderProbability) {
+  const extension = { gender, genderProbability };
+  return { ...sourceObj, ...extension };
+}
+
+// src/ssdMobilenetv1/extractParams.ts
+function extractorsFactory5(extractWeights, paramMappings) {
+  function extractDepthwiseConvParams(numChannels, mappedPrefix) {
+    const filters = tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);
+    const batch_norm_scale = tensor1d(extractWeights(numChannels));
+    const batch_norm_offset = tensor1d(extractWeights(numChannels));
+    const batch_norm_mean = tensor1d(extractWeights(numChannels));
+    const batch_norm_variance = tensor1d(extractWeights(numChannels));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/batch_norm_scale` },
+      { paramPath: `${mappedPrefix}/batch_norm_offset` },
+      { paramPath: `${mappedPrefix}/batch_norm_mean` },
+      { paramPath: `${mappedPrefix}/batch_norm_variance` }
+    );
+    return {
+      filters,
+      batch_norm_scale,
+      batch_norm_offset,
+      batch_norm_mean,
+      batch_norm_variance
+    };
+  }
+  function extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, isPointwiseConv) {
+    const filters = tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/${isPointwiseConv ? "batch_norm_offset" : "bias"}` }
+    );
+    return { filters, bias };
+  }
+  function extractPointwiseConvParams(channelsIn, channelsOut, filterSize, mappedPrefix) {
+    const {
+      filters,
+      bias
+    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);
+    return {
+      filters,
+      batch_norm_offset: bias
+    };
+  }
+  function extractConvPairParams(channelsIn, channelsOut, mappedPrefix) {
+    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);
+    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);
+    return { depthwise_conv, pointwise_conv };
+  }
+  function extractMobilenetV1Params() {
+    const conv_0 = extractPointwiseConvParams(3, 32, 3, "mobilenetv1/conv_0");
+    const conv_1 = extractConvPairParams(32, 64, "mobilenetv1/conv_1");
+    const conv_2 = extractConvPairParams(64, 128, "mobilenetv1/conv_2");
+    const conv_3 = extractConvPairParams(128, 128, "mobilenetv1/conv_3");
+    const conv_4 = extractConvPairParams(128, 256, "mobilenetv1/conv_4");
+    const conv_5 = extractConvPairParams(256, 256, "mobilenetv1/conv_5");
+    const conv_6 = extractConvPairParams(256, 512, "mobilenetv1/conv_6");
+    const conv_7 = extractConvPairParams(512, 512, "mobilenetv1/conv_7");
+    const conv_8 = extractConvPairParams(512, 512, "mobilenetv1/conv_8");
+    const conv_9 = extractConvPairParams(512, 512, "mobilenetv1/conv_9");
+    const conv_10 = extractConvPairParams(512, 512, "mobilenetv1/conv_10");
+    const conv_11 = extractConvPairParams(512, 512, "mobilenetv1/conv_11");
+    const conv_12 = extractConvPairParams(512, 1024, "mobilenetv1/conv_12");
+    const conv_13 = extractConvPairParams(1024, 1024, "mobilenetv1/conv_13");
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      conv_8,
+      conv_9,
+      conv_10,
+      conv_11,
+      conv_12,
+      conv_13
+    };
+  }
+  function extractPredictionLayerParams() {
+    const conv_0 = extractPointwiseConvParams(1024, 256, 1, "prediction_layer/conv_0");
+    const conv_1 = extractPointwiseConvParams(256, 512, 3, "prediction_layer/conv_1");
+    const conv_2 = extractPointwiseConvParams(512, 128, 1, "prediction_layer/conv_2");
+    const conv_3 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_3");
+    const conv_4 = extractPointwiseConvParams(256, 128, 1, "prediction_layer/conv_4");
+    const conv_5 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_5");
+    const conv_6 = extractPointwiseConvParams(256, 64, 1, "prediction_layer/conv_6");
+    const conv_7 = extractPointwiseConvParams(64, 128, 3, "prediction_layer/conv_7");
+    const box_encoding_0_predictor = extractConvParams(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor");
+    const class_predictor_0 = extractConvParams(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor");
+    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor");
+    const class_predictor_1 = extractConvParams(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor");
+    const box_encoding_2_predictor = extractConvParams(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor");
+    const class_predictor_2 = extractConvParams(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor");
+    const box_encoding_3_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor");
+    const class_predictor_3 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor");
+    const box_encoding_4_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor");
+    const class_predictor_4 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor");
+    const box_encoding_5_predictor = extractConvParams(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor");
+    const class_predictor_5 = extractConvParams(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor");
+    const box_predictor_0 = {
+      box_encoding_predictor: box_encoding_0_predictor,
+      class_predictor: class_predictor_0
+    };
+    const box_predictor_1 = {
+      box_encoding_predictor: box_encoding_1_predictor,
+      class_predictor: class_predictor_1
+    };
+    const box_predictor_2 = {
+      box_encoding_predictor: box_encoding_2_predictor,
+      class_predictor: class_predictor_2
+    };
+    const box_predictor_3 = {
+      box_encoding_predictor: box_encoding_3_predictor,
+      class_predictor: class_predictor_3
+    };
+    const box_predictor_4 = {
+      box_encoding_predictor: box_encoding_4_predictor,
+      class_predictor: class_predictor_4
+    };
+    const box_predictor_5 = {
+      box_encoding_predictor: box_encoding_5_predictor,
+      class_predictor: class_predictor_5
+    };
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      box_predictor_0,
+      box_predictor_1,
+      box_predictor_2,
+      box_predictor_3,
+      box_predictor_4,
+      box_predictor_5
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParams6(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory5(extractWeights, paramMappings);
+  const mobilenetv1 = extractMobilenetV1Params();
+  const prediction_layer = extractPredictionLayerParams();
+  const extra_dim = tensor3d(
+    extractWeights(5118 * 4),
+    [1, 5118, 4]
+  );
+  const output_layer = {
+    extra_dim
+  };
+  paramMappings.push({ paramPath: "output_layer/extra_dim" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    params: {
+      mobilenetv1,
+      prediction_layer,
+      output_layer
+    },
+    paramMappings
+  };
+}
+
+// src/ssdMobilenetv1/extractParamsFromWeightMap.ts
+function extractorsFactory6(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractPointwiseConvParams(prefix, idx, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);
+    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);
+    return { filters, batch_norm_offset };
+  }
+  function extractConvPairParams(idx) {
+    const mappedPrefix = `mobilenetv1/conv_${idx}`;
+    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;
+    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;
+    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;
+    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);
+    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);
+    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);
+    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);
+    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);
+    return {
+      depthwise_conv: {
+        filters,
+        batch_norm_scale,
+        batch_norm_offset,
+        batch_norm_mean,
+        batch_norm_variance
+      },
+      pointwise_conv: extractPointwiseConvParams("MobilenetV1", idx, mappedPrefixPointwiseConv)
+    };
+  }
+  function extractMobilenetV1Params() {
+    return {
+      conv_0: extractPointwiseConvParams("MobilenetV1", 0, "mobilenetv1/conv_0"),
+      conv_1: extractConvPairParams(1),
+      conv_2: extractConvPairParams(2),
+      conv_3: extractConvPairParams(3),
+      conv_4: extractConvPairParams(4),
+      conv_5: extractConvPairParams(5),
+      conv_6: extractConvPairParams(6),
+      conv_7: extractConvPairParams(7),
+      conv_8: extractConvPairParams(8),
+      conv_9: extractConvPairParams(9),
+      conv_10: extractConvPairParams(10),
+      conv_11: extractConvPairParams(11),
+      conv_12: extractConvPairParams(12),
+      conv_13: extractConvPairParams(13)
+    };
+  }
+  function extractConvParams(prefix, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);
+    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);
+    return { filters, bias };
+  }
+  function extractBoxPredictorParams(idx) {
+    const box_encoding_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,
+      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`
+    );
+    const class_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/ClassPredictor`,
+      `prediction_layer/box_predictor_${idx}/class_predictor`
+    );
+    return { box_encoding_predictor, class_predictor };
+  }
+  function extractPredictionLayerParams() {
+    return {
+      conv_0: extractPointwiseConvParams("Prediction", 0, "prediction_layer/conv_0"),
+      conv_1: extractPointwiseConvParams("Prediction", 1, "prediction_layer/conv_1"),
+      conv_2: extractPointwiseConvParams("Prediction", 2, "prediction_layer/conv_2"),
+      conv_3: extractPointwiseConvParams("Prediction", 3, "prediction_layer/conv_3"),
+      conv_4: extractPointwiseConvParams("Prediction", 4, "prediction_layer/conv_4"),
+      conv_5: extractPointwiseConvParams("Prediction", 5, "prediction_layer/conv_5"),
+      conv_6: extractPointwiseConvParams("Prediction", 6, "prediction_layer/conv_6"),
+      conv_7: extractPointwiseConvParams("Prediction", 7, "prediction_layer/conv_7"),
+      box_predictor_0: extractBoxPredictorParams(0),
+      box_predictor_1: extractBoxPredictorParams(1),
+      box_predictor_2: extractBoxPredictorParams(2),
+      box_predictor_3: extractBoxPredictorParams(3),
+      box_predictor_4: extractBoxPredictorParams(4),
+      box_predictor_5: extractBoxPredictorParams(5)
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParamsFromWeightMap6(weightMap) {
+  const paramMappings = [];
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory6(weightMap, paramMappings);
+  const extra_dim = weightMap["Output/extra_dim"];
+  paramMappings.push({ originalPath: "Output/extra_dim", paramPath: "output_layer/extra_dim" });
+  if (!isTensor3D(extra_dim)) {
+    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);
+  }
+  const params = {
+    mobilenetv1: extractMobilenetV1Params(),
+    prediction_layer: extractPredictionLayerParams(),
+    output_layer: {
+      extra_dim
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ssdMobilenetv1/pointwiseConvLayer.ts
+function pointwiseConvLayer(x, params, strides) {
+  return tidy(() => {
+    let out = conv2d(x, params.filters, strides, "same");
+    out = add2(out, params.batch_norm_offset);
+    return clipByValue(out, 0, 6);
+  });
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var epsilon3 = 0.0010000000474974513;
+function depthwiseConvLayer(x, params, strides) {
+  return tidy(() => {
+    let out = depthwiseConv2d(x, params.filters, strides, "same");
+    out = batchNorm(
+      out,
+      params.batch_norm_mean,
+      params.batch_norm_variance,
+      params.batch_norm_offset,
+      params.batch_norm_scale,
+      epsilon3
+    );
+    return clipByValue(out, 0, 6);
+  });
+}
+function getStridesForLayerIdx(layerIdx) {
+  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];
+}
+function mobileNetV1(x, params) {
+  return tidy(() => {
+    let conv11;
+    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
+    const convPairParams = [
+      params.conv_1,
+      params.conv_2,
+      params.conv_3,
+      params.conv_4,
+      params.conv_5,
+      params.conv_6,
+      params.conv_7,
+      params.conv_8,
+      params.conv_9,
+      params.conv_10,
+      params.conv_11,
+      params.conv_12,
+      params.conv_13
+    ];
+    convPairParams.forEach((param, i) => {
+      const layerIdx = i + 1;
+      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
+      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
+      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
+      if (layerIdx === 11)
+        conv11 = out;
+    });
+    if (conv11 === null) {
+      throw new Error("mobileNetV1 - output of conv layer 11 is null");
+    }
+    return {
+      out,
+      conv11
+    };
+  });
+}
+
+// src/ssdMobilenetv1/nonMaxSuppression.ts
+function IOU(boxes, i, j) {
+  const boxesData = boxes.arraySync();
+  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);
+  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);
+  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);
+  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);
+  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);
+  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);
+  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);
+  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0)
+    return 0;
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function nonMaxSuppression3(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  const numBoxes = boxes.shape[0];
+  const outputSize = Math.min(maxOutputSize, numBoxes);
+  const candidates = scores.map((score, boxIndex) => ({ score, boxIndex })).filter((c) => c.score > scoreThreshold).sort((c1, c2) => c2.score - c1.score);
+  const suppressFunc = (x) => x <= iouThreshold ? 1 : 0;
+  const selected = [];
+  candidates.forEach((c) => {
+    if (selected.length >= outputSize)
+      return;
+    const originalScore = c.score;
+    for (let j = selected.length - 1; j >= 0; --j) {
+      const iou2 = IOU(boxes, c.boxIndex, selected[j]);
+      if (iou2 === 0)
+        continue;
+      c.score *= suppressFunc(iou2);
+      if (c.score <= scoreThreshold)
+        break;
+    }
+    if (originalScore === c.score) {
+      selected.push(c.boxIndex);
+    }
+  });
+  return selected;
+}
+
+// src/ssdMobilenetv1/outputLayer.ts
+function getCenterCoordinatesAndSizesLayer(x) {
+  const vec = unstack(transpose(x, [1, 0]));
+  const sizes = [
+    sub(vec[2], vec[0]),
+    sub(vec[3], vec[1])
+  ];
+  const centers = [
+    add2(vec[0], div(sizes[0], 2)),
+    add2(vec[1], div(sizes[1], 2))
+  ];
+  return { sizes, centers };
+}
+function decodeBoxesLayer(x0, x1) {
+  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);
+  const vec = unstack(transpose(x1, [1, 0]));
+  const div0_out = div(mul(exp(div(vec[2], 5)), sizes[0]), 2);
+  const add0_out = add2(mul(div(vec[0], 10), sizes[0]), centers[0]);
+  const div1_out = div(mul(exp(div(vec[3], 5)), sizes[1]), 2);
+  const add1_out = add2(mul(div(vec[1], 10), sizes[1]), centers[1]);
+  return transpose(
+    stack([
+      sub(add0_out, div0_out),
+      sub(add1_out, div1_out),
+      add2(add0_out, div0_out),
+      add2(add1_out, div1_out)
+    ]),
+    [1, 0]
+  );
+}
+function outputLayer(boxPredictions, classPredictions, params) {
+  return tidy(() => {
+    const batchSize = boxPredictions.shape[0];
+    let boxes = decodeBoxesLayer(
+      reshape(tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]),
+      reshape(boxPredictions, [-1, 4])
+    );
+    boxes = reshape(boxes, [batchSize, boxes.shape[0] / batchSize, 4]);
+    const scoresAndClasses = sigmoid(slice(classPredictions, [0, 0, 1], [-1, -1, -1]));
+    let scores = slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]);
+    scores = reshape(scores, [batchSize, scores.shape[1]]);
+    const boxesByBatch = unstack(boxes);
+    const scoresByBatch = unstack(scores);
+    return { boxes: boxesByBatch, scores: scoresByBatch };
+  });
+}
+
+// src/ssdMobilenetv1/boxPredictionLayer.ts
+function boxPredictionLayer(x, params) {
+  return tidy(() => {
+    const batchSize = x.shape[0];
+    const boxPredictionEncoding = reshape(
+      convLayer(x, params.box_encoding_predictor),
+      [batchSize, -1, 1, 4]
+    );
+    const classPrediction = reshape(
+      convLayer(x, params.class_predictor),
+      [batchSize, -1, 3]
+    );
+    return { boxPredictionEncoding, classPrediction };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+function predictionLayer(x, conv11, params) {
+  return tidy(() => {
+    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);
+    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);
+    const conv22 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);
+    const conv3 = pointwiseConvLayer(conv22, params.conv_3, [2, 2]);
+    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);
+    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);
+    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);
+    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);
+    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);
+    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);
+    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);
+    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);
+    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);
+    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);
+    const boxPredictions = concat([
+      boxPrediction0.boxPredictionEncoding,
+      boxPrediction1.boxPredictionEncoding,
+      boxPrediction2.boxPredictionEncoding,
+      boxPrediction3.boxPredictionEncoding,
+      boxPrediction4.boxPredictionEncoding,
+      boxPrediction5.boxPredictionEncoding
+    ], 1);
+    const classPredictions = concat([
+      boxPrediction0.classPrediction,
+      boxPrediction1.classPrediction,
+      boxPrediction2.classPrediction,
+      boxPrediction3.classPrediction,
+      boxPrediction4.classPrediction,
+      boxPrediction5.classPrediction
+    ], 1);
+    return {
+      boxPredictions,
+      classPredictions
+    };
+  });
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1Options.ts
+var SsdMobilenetv1Options = class {
+  constructor({ minConfidence, maxResults } = {}) {
+    this._name = "SsdMobilenetv1Options";
+    this._minConfidence = minConfidence || 0.5;
+    this._maxResults = maxResults || 100;
+    if (typeof this._minConfidence !== "number" || this._minConfidence <= 0 || this._minConfidence >= 1) {
+      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);
+    }
+    if (typeof this._maxResults !== "number") {
+      throw new Error(`${this._name} - expected maxResults to be a number`);
+    }
+  }
+  get minConfidence() {
+    return this._minConfidence;
+  }
+  get maxResults() {
+    return this._maxResults;
+  }
+};
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var SsdMobilenetv1 = class extends NeuralNetwork {
+  constructor() {
+    super("SsdMobilenetv1");
+  }
+  forwardInput(input2) {
+    const { params } = this;
+    if (!params)
+      throw new Error("SsdMobilenetv1 - load model before inference");
+    return tidy(() => {
+      const batchTensor = cast(input2.toBatchTensor(512, false), "float32");
+      const x = sub(div(batchTensor, 127.5), 1);
+      const features = mobileNetV1(x, params.mobilenetv1);
+      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);
+      return outputLayer(boxPredictions, classPredictions, params.output_layer);
+    });
+  }
+  async forward(input2) {
+    return this.forwardInput(await toNetInput(input2));
+  }
+  async locateFaces(input2, options = {}) {
+    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);
+    const netInput = await toNetInput(input2);
+    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);
+    const boxes = _boxes[0];
+    const scores = _scores[0];
+    for (let i = 1; i < _boxes.length; i++) {
+      _boxes[i].dispose();
+      _scores[i].dispose();
+    }
+    const scoresData = Array.from(scores.dataSync());
+    const iouThreshold = 0.5;
+    const indices = nonMaxSuppression3(boxes, scoresData, maxResults, iouThreshold, minConfidence);
+    const reshapedDims = netInput.getReshapedInputDimensions(0);
+    const inputSize = netInput.inputSize;
+    const padX = inputSize / reshapedDims.width;
+    const padY = inputSize / reshapedDims.height;
+    const boxesData = boxes.arraySync();
+    const results = indices.map((idx) => {
+      const [top, bottom] = [
+        Math.max(0, boxesData[idx][0]),
+        Math.min(1, boxesData[idx][2])
+      ].map((val) => val * padY);
+      const [left, right] = [
+        Math.max(0, boxesData[idx][1]),
+        Math.min(1, boxesData[idx][3])
+      ].map((val) => val * padX);
+      return new FaceDetection(
+        scoresData[idx],
+        new Rect(left, top, right - left, bottom - top),
+        { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) }
+      );
+    });
+    boxes.dispose();
+    scores.dispose();
+    return results;
+  }
+  getDefaultModelName() {
+    return "ssd_mobilenetv1_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap6(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams6(weights);
+  }
+};
+
+// src/ssdMobilenetv1/index.ts
+function createSsdMobilenetv1(weights) {
+  const net = new SsdMobilenetv1();
+  net.extractWeights(weights);
+  return net;
+}
+function createFaceDetectionNet(weights) {
+  return createSsdMobilenetv1(weights);
+}
+var FaceDetectionNet = class extends SsdMobilenetv1 {
+};
+
+// src/tinyYolov2/const.ts
+var IOU_THRESHOLD = 0.4;
+var BOX_ANCHORS = [
+  new Point(0.738768, 0.874946),
+  new Point(2.42204, 2.65704),
+  new Point(4.30971, 7.04493),
+  new Point(10.246, 4.59428),
+  new Point(12.6868, 11.8741)
+];
+var BOX_ANCHORS_SEPARABLE = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB_SEPARABLE = [117.001, 114.697, 97.404];
+var DEFAULT_MODEL_NAME2 = "tiny_yolov2_model";
+var DEFAULT_MODEL_NAME_SEPARABLE_CONV = "tiny_yolov2_separable_conv_model";
+
+// src/tinyYolov2/config.ts
+var isNumber2 = (arg) => typeof arg === "number";
+function validateConfig(config) {
+  if (!config) {
+    throw new Error(`invalid config: ${config}`);
+  }
+  if (typeof config.withSeparableConvs !== "boolean") {
+    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);
+  }
+  if (!isNumber2(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1) {
+    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);
+  }
+  if (!Array.isArray(config.classes) || !config.classes.length || !config.classes.every((c) => typeof c === "string")) {
+    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);
+  }
+  if (!Array.isArray(config.anchors) || !config.anchors.length || !config.anchors.map((a) => a || {}).every((a) => isNumber2(a.x) && isNumber2(a.y))) {
+    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);
+  }
+  if (config.meanRgb && (!Array.isArray(config.meanRgb) || config.meanRgb.length !== 3 || !config.meanRgb.every(isNumber2))) {
+    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);
+  }
+}
+
+// src/tinyYolov2/leaky.ts
+function leaky(x) {
+  return tidy(() => {
+    const min6 = mul(x, scalar(0.10000000149011612));
+    return add2(relu(sub(x, min6)), min6);
+  });
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+function convWithBatchNorm(x, params) {
+  return tidy(() => {
+    let out = pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = conv2d(out, params.conv.filters, [1, 1], "valid");
+    out = sub(out, params.bn.sub);
+    out = mul(out, params.bn.truediv);
+    out = add2(out, params.conv.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/depthwiseSeparableConv.ts
+function depthwiseSeparableConv2(x, params) {
+  return tidy(() => {
+    let out = pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], "valid");
+    out = add2(out, params.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/extractParams.ts
+function extractorsFactory7(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  function extractBatchNormParams(size, mappedPrefix) {
+    const sub4 = tensor1d(extractWeights(size));
+    const truediv = tensor1d(extractWeights(size));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/sub` },
+      { paramPath: `${mappedPrefix}/truediv` }
+    );
+    return { sub: sub4, truediv };
+  }
+  function extractConvWithBatchNormParams(channelsIn, channelsOut, mappedPrefix) {
+    const conv3 = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);
+    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParams7(weights, config, boxEncodingSize, filterSizes) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory7(extractWeights, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = config.isFirstLayerConv2d ? extractConvParams(s0, s1, 3, "conv0") : extractSeparableConvParams(s0, s1, "conv0");
+    const conv1 = extractSeparableConvParams(s1, s2, "conv1");
+    const conv22 = extractSeparableConvParams(s2, s3, "conv2");
+    const conv3 = extractSeparableConvParams(s3, s4, "conv3");
+    const conv4 = extractSeparableConvParams(s4, s5, "conv4");
+    const conv5 = extractSeparableConvParams(s5, s6, "conv5");
+    const conv6 = s7 ? extractSeparableConvParams(s6, s7, "conv6") : void 0;
+    const conv7 = s8 ? extractSeparableConvParams(s7, s8, "conv7") : void 0;
+    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  } else {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = extractConvWithBatchNormParams(s0, s1, "conv0");
+    const conv1 = extractConvWithBatchNormParams(s1, s2, "conv1");
+    const conv22 = extractConvWithBatchNormParams(s2, s3, "conv2");
+    const conv3 = extractConvWithBatchNormParams(s3, s4, "conv3");
+    const conv4 = extractConvWithBatchNormParams(s4, s5, "conv4");
+    const conv5 = extractConvWithBatchNormParams(s5, s6, "conv5");
+    const conv6 = extractConvWithBatchNormParams(s6, s7, "conv6");
+    const conv7 = extractConvWithBatchNormParams(s7, s8, "conv7");
+    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  }
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/extractParamsFromWeightMap.ts
+function extractorsFactory8(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractBatchNormParams(prefix) {
+    const sub4 = extractWeightEntry(`${prefix}/sub`, 1);
+    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);
+    return { sub: sub4, truediv };
+  }
+  function extractConvParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  }
+  function extractConvWithBatchNormParams(prefix) {
+    const conv3 = extractConvParams(`${prefix}/conv`);
+    const bn = extractBatchNormParams(`${prefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParamsFromWeightMap7(weightMap, config) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory8(weightMap, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const numFilters = config.filterSizes && config.filterSizes.length || 9;
+    params = {
+      conv0: config.isFirstLayerConv2d ? extractConvParams("conv0") : extractSeparableConvParams("conv0"),
+      conv1: extractSeparableConvParams("conv1"),
+      conv2: extractSeparableConvParams("conv2"),
+      conv3: extractSeparableConvParams("conv3"),
+      conv4: extractSeparableConvParams("conv4"),
+      conv5: extractSeparableConvParams("conv5"),
+      conv6: numFilters > 7 ? extractSeparableConvParams("conv6") : void 0,
+      conv7: numFilters > 8 ? extractSeparableConvParams("conv7") : void 0,
+      conv8: extractConvParams("conv8")
+    };
+  } else {
+    params = {
+      conv0: extractConvWithBatchNormParams("conv0"),
+      conv1: extractConvWithBatchNormParams("conv1"),
+      conv2: extractConvWithBatchNormParams("conv2"),
+      conv3: extractConvWithBatchNormParams("conv3"),
+      conv4: extractConvWithBatchNormParams("conv4"),
+      conv5: extractConvWithBatchNormParams("conv5"),
+      conv6: extractConvWithBatchNormParams("conv6"),
+      conv7: extractConvWithBatchNormParams("conv7"),
+      conv8: extractConvParams("conv8")
+    };
+  }
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/TinyYolov2Options.ts
+var TinyYolov2Options = class {
+  constructor({ inputSize, scoreThreshold } = {}) {
+    this._name = "TinyYolov2Options";
+    this._inputSize = inputSize || 416;
+    this._scoreThreshold = scoreThreshold || 0.5;
+    if (typeof this._inputSize !== "number" || this._inputSize % 32 !== 0) {
+      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);
+    }
+    if (typeof this._scoreThreshold !== "number" || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {
+      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);
+    }
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get scoreThreshold() {
+    return this._scoreThreshold;
+  }
+};
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var _TinyYolov2Base = class extends NeuralNetwork {
+  constructor(config) {
+    super("TinyYolov2");
+    validateConfig(config);
+    this._config = config;
+  }
+  get config() {
+    return this._config;
+  }
+  get withClassScores() {
+    return this.config.withClassScores || this.config.classes.length > 1;
+  }
+  get boxEncodingSize() {
+    return 5 + (this.withClassScores ? this.config.classes.length : 0);
+  }
+  runTinyYolov2(x, params) {
+    let out = convWithBatchNorm(x, params.conv0);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv1);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv2);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv3);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv4);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv5);
+    out = maxPool(out, [2, 2], [1, 1], "same");
+    out = convWithBatchNorm(out, params.conv6);
+    out = convWithBatchNorm(out, params.conv7);
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  runMobilenet(x, params) {
+    let out = this.config.isFirstLayerConv2d ? leaky(convLayer(x, params.conv0, "valid", false)) : depthwiseSeparableConv2(x, params.conv0);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv1);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv2);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv3);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv4);
+    out = maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv5);
+    out = maxPool(out, [2, 2], [1, 1], "same");
+    out = params.conv6 ? depthwiseSeparableConv2(out, params.conv6) : out;
+    out = params.conv7 ? depthwiseSeparableConv2(out, params.conv7) : out;
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  forwardInput(input2, inputSize) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyYolov2 - load model before inference");
+    }
+    return tidy(() => {
+      let batchTensor = cast(input2.toBatchTensor(inputSize, false), "float32");
+      batchTensor = this.config.meanRgb ? normalize(batchTensor, this.config.meanRgb) : batchTensor;
+      batchTensor = batchTensor.div(255);
+      return this.config.withSeparableConvs ? this.runMobilenet(batchTensor, params) : this.runTinyYolov2(batchTensor, params);
+    });
+  }
+  async forward(input2, inputSize) {
+    return this.forwardInput(await toNetInput(input2), inputSize);
+  }
+  async detect(input2, forwardParams = {}) {
+    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);
+    const netInput = await toNetInput(input2);
+    const out = await this.forwardInput(netInput, inputSize);
+    const out0 = tidy(() => unstack(out)[0].expandDims());
+    const inputDimensions = {
+      width: netInput.getInputWidth(0),
+      height: netInput.getInputHeight(0)
+    };
+    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);
+    out.dispose();
+    out0.dispose();
+    const boxes = results.map((res) => res.box);
+    const scores = results.map((res) => res.score);
+    const classScores = results.map((res) => res.classScore);
+    const classNames = results.map((res) => this.config.classes[res.label]);
+    const indices = nonMaxSuppression2(
+      boxes.map((box) => box.rescale(inputSize)),
+      scores,
+      this.config.iouThreshold,
+      true
+    );
+    const detections = indices.map((idx) => new ObjectDetection(
+      scores[idx],
+      classScores[idx],
+      classNames[idx],
+      boxes[idx],
+      inputDimensions
+    ));
+    return detections;
+  }
+  getDefaultModelName() {
+    return "";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap7(weightMap, this.config);
+  }
+  extractParams(weights) {
+    const filterSizes = this.config.filterSizes || _TinyYolov2Base.DEFAULT_FILTER_SIZES;
+    const numFilters = filterSizes ? filterSizes.length : void 0;
+    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {
+      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);
+    }
+    return extractParams7(weights, this.config, this.boxEncodingSize, filterSizes);
+  }
+  async extractBoxes(outputTensor, inputBlobDimensions, scoreThreshold) {
+    const { width, height } = inputBlobDimensions;
+    const inputSize = Math.max(width, height);
+    const correctionFactorX = inputSize / width;
+    const correctionFactorY = inputSize / height;
+    const numCells = outputTensor.shape[1];
+    const numBoxes = this.config.anchors.length;
+    const [boxesTensor, scoresTensor, classScoresTensor] = tidy(() => {
+      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);
+      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);
+      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);
+      const classScores = this.withClassScores ? softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3) : scalar(0);
+      return [boxes, scores, classScores];
+    });
+    const results = [];
+    const scoresData = await scoresTensor.array();
+    const boxesData = await boxesTensor.array();
+    for (let row = 0; row < numCells; row++) {
+      for (let col = 0; col < numCells; col++) {
+        for (let anchor = 0; anchor < numBoxes; anchor++) {
+          const score = sigmoid5(scoresData[row][col][anchor][0]);
+          if (!scoreThreshold || score > scoreThreshold) {
+            const ctX = (col + sigmoid5(boxesData[row][col][anchor][0])) / numCells * correctionFactorX;
+            const ctY = (row + sigmoid5(boxesData[row][col][anchor][1])) / numCells * correctionFactorY;
+            const widthLocal = Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x / numCells * correctionFactorX;
+            const heightLocal = Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y / numCells * correctionFactorY;
+            const x = ctX - widthLocal / 2;
+            const y = ctY - heightLocal / 2;
+            const pos = { row, col, anchor };
+            const { classScore, label } = this.withClassScores ? await this.extractPredictedClass(classScoresTensor, pos) : { classScore: 1, label: 0 };
+            results.push({
+              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),
+              score,
+              classScore: score * classScore,
+              label,
+              ...pos
+            });
+          }
+        }
+      }
+    }
+    boxesTensor.dispose();
+    scoresTensor.dispose();
+    classScoresTensor.dispose();
+    return results;
+  }
+  async extractPredictedClass(classesTensor, pos) {
+    const { row, col, anchor } = pos;
+    const classesData = await classesTensor.array();
+    return Array(this.config.classes.length).fill(0).map((_, i) => classesData[row][col][anchor][i]).map((classScore, label) => ({
+      classScore,
+      label
+    })).reduce((max6, curr) => max6.classScore > curr.classScore ? max6 : curr);
+  }
+};
+var TinyYolov2Base = _TinyYolov2Base;
+TinyYolov2Base.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+
+// src/tinyYolov2/TinyYolov2.ts
+var TinyYolov2 = class extends TinyYolov2Base {
+  constructor(withSeparableConvs = true) {
+    const config = {
+      withSeparableConvs,
+      iouThreshold: IOU_THRESHOLD,
+      classes: ["face"],
+      ...withSeparableConvs ? {
+        anchors: BOX_ANCHORS_SEPARABLE,
+        meanRgb: MEAN_RGB_SEPARABLE
+      } : {
+        anchors: BOX_ANCHORS,
+        withClassScores: true
+      }
+    };
+    super(config);
+  }
+  get withSeparableConvs() {
+    return this.config.withSeparableConvs;
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input2, forwardParams) {
+    const objectDetections = await this.detect(input2, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME2;
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/tinyYolov2/index.ts
+function createTinyYolov2(weights, withSeparableConvs = true) {
+  const net = new TinyYolov2(withSeparableConvs);
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/tinyFaceDetector/TinyFaceDetectorOptions.ts
+var TinyFaceDetectorOptions = class extends TinyYolov2Options {
+  constructor() {
+    super(...arguments);
+    this._name = "TinyFaceDetectorOptions";
+  }
+};
+
+// src/globalApi/ComposableTask.ts
+var ComposableTask = class {
+  async then(onfulfilled) {
+    return onfulfilled(await this.run());
+  }
+  async run() {
+    throw new Error("ComposableTask - run is not implemented");
+  }
+};
+
+// src/globalApi/extractFacesAndComputeResults.ts
+async function extractAllFacesAndComputeResults(parentResults, input2, computeResults, extractedFaces, getRectForAlignment = ({ alignedRect }) => alignedRect) {
+  const faceBoxes = parentResults.map((parentResult) => isWithFaceLandmarks(parentResult) ? getRectForAlignment(parentResult) : parentResult.detection);
+  const faces = extractedFaces || (input2 instanceof Tensor ? await extractFaceTensors(input2, faceBoxes) : await extractFaces(input2, faceBoxes));
+  const results = await computeResults(faces);
+  faces.forEach((f) => f instanceof Tensor && f.dispose());
+  return results;
+}
+async function extractSingleFaceAndComputeResult(parentResult, input2, computeResult, extractedFaces, getRectForAlignment) {
+  return extractAllFacesAndComputeResults(
+    [parentResult],
+    input2,
+    async (faces) => computeResult(faces[0]),
+    extractedFaces,
+    getRectForAlignment
+  );
+}
+
+// src/tinyFaceDetector/const.ts
+var IOU_THRESHOLD2 = 0.4;
+var BOX_ANCHORS2 = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB = [117.001, 114.697, 97.404];
+
+// src/tinyFaceDetector/TinyFaceDetector.ts
+var TinyFaceDetector = class extends TinyYolov2Base {
+  constructor() {
+    const config = {
+      withSeparableConvs: true,
+      iouThreshold: IOU_THRESHOLD2,
+      classes: ["face"],
+      anchors: BOX_ANCHORS2,
+      meanRgb: MEAN_RGB,
+      isFirstLayerConv2d: true,
+      filterSizes: [3, 16, 32, 64, 128, 256, 512]
+    };
+    super(config);
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input2, forwardParams) {
+    const objectDetections = await this.detect(input2, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return "tiny_face_detector_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/globalApi/nets.ts
+var nets = {
+  ssdMobilenetv1: new SsdMobilenetv1(),
+  tinyFaceDetector: new TinyFaceDetector(),
+  tinyYolov2: new TinyYolov2(),
+  faceLandmark68Net: new FaceLandmark68Net(),
+  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),
+  faceRecognitionNet: new FaceRecognitionNet(),
+  faceExpressionNet: new FaceExpressionNet(),
+  ageGenderNet: new AgeGenderNet()
+};
+var ssdMobilenetv1 = (input2, options) => nets.ssdMobilenetv1.locateFaces(input2, options);
+var tinyFaceDetector = (input2, options) => nets.tinyFaceDetector.locateFaces(input2, options);
+var tinyYolov2 = (input2, options) => nets.tinyYolov2.locateFaces(input2, options);
+var detectFaceLandmarks = (input2) => nets.faceLandmark68Net.detectLandmarks(input2);
+var detectFaceLandmarksTiny = (input2) => nets.faceLandmark68TinyNet.detectLandmarks(input2);
+var computeFaceDescriptor = (input2) => nets.faceRecognitionNet.computeFaceDescriptor(input2);
+var recognizeFaceExpressions = (input2) => nets.faceExpressionNet.predictExpressions(input2);
+var predictAgeAndGender = (input2) => nets.ageGenderNet.predictAgeAndGender(input2);
+var loadSsdMobilenetv1Model = (url) => nets.ssdMobilenetv1.load(url);
+var loadTinyFaceDetectorModel = (url) => nets.tinyFaceDetector.load(url);
+var loadTinyYolov2Model = (url) => nets.tinyYolov2.load(url);
+var loadFaceLandmarkModel = (url) => nets.faceLandmark68Net.load(url);
+var loadFaceLandmarkTinyModel = (url) => nets.faceLandmark68TinyNet.load(url);
+var loadFaceRecognitionModel = (url) => nets.faceRecognitionNet.load(url);
+var loadFaceExpressionModel = (url) => nets.faceExpressionNet.load(url);
+var loadAgeGenderModel = (url) => nets.ageGenderNet.load(url);
+var loadFaceDetectionModel = loadSsdMobilenetv1Model;
+var locateFaces = ssdMobilenetv1;
+var detectLandmarks = detectFaceLandmarks;
+
+// src/globalApi/PredictFaceExpressionsTask.ts
+var PredictFaceExpressionsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input2, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input2;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const faceExpressionsByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(
+        faces.map((face) => nets.faceExpressionNet.predictExpressions(face))
+      ),
+      this.extractedFaces
+    );
+    return parentResults.map(
+      (parentResult, i) => extendWithFaceExpressions(parentResult, faceExpressionsByFace[i])
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const faceExpressions = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceExpressionNet.predictExpressions(face),
+      this.extractedFaces
+    );
+    return extendWithFaceExpressions(parentResult, faceExpressions);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictAllFaceExpressionsWithFaceAlignmentTask = class extends PredictAllFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsWithFaceAlignmentTask = class extends PredictSingleFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/PredictAgeAndGenderTask.ts
+var PredictAgeAndGenderTaskBase = class extends ComposableTask {
+  constructor(parentTask, input2, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input2;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const ageAndGenderByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face))),
+      this.extractedFaces
+    );
+    return parentResults.map((parentResult, i) => {
+      const { age, gender, genderProbability } = ageAndGenderByFace[i];
+      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+    });
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.ageGenderNet.predictAgeAndGender(face),
+      this.extractedFaces
+    );
+    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictAllAgeAndGenderWithFaceAlignmentTask = class extends PredictAllAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderWithFaceAlignmentTask = class extends PredictSingleAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/ComputeFaceDescriptorsTasks.ts
+var ComputeFaceDescriptorsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input2) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input2;
+  }
+};
+var ComputeAllFaceDescriptorsTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const descriptors = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face))),
+      null,
+      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return descriptors.map((descriptor, i) => extendWithFaceDescriptor(parentResults[i], descriptor));
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+var ComputeSingleFaceDescriptorTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const descriptor = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face),
+      null,
+      (parentResult2) => parentResult2.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return extendWithFaceDescriptor(parentResult, descriptor);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var DetectFaceLandmarksTaskBase = class extends ComposableTask {
+  constructor(parentTask, input2, useTinyLandmarkNet) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input2;
+    this.useTinyLandmarkNet = useTinyLandmarkNet;
+  }
+  get landmarkNet() {
+    return this.useTinyLandmarkNet ? nets.faceLandmark68TinyNet : nets.faceLandmark68Net;
+  }
+};
+var DetectAllFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const detections = parentResults.map((res) => res.detection);
+    const faces = this.input instanceof Tensor ? await extractFaceTensors(this.input, detections) : await extractFaces(this.input, detections);
+    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face)));
+    faces.forEach((f) => f instanceof Tensor && f.dispose());
+    const result = parentResults.filter((_parentResult, i) => faceLandmarksByFace[i]).map((parentResult, i) => extendWithFaceLandmarks(parentResult, faceLandmarksByFace[i]));
+    return result;
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var DetectSingleFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const { detection } = parentResult;
+    const faces = this.input instanceof Tensor ? await extractFaceTensors(this.input, [detection]) : await extractFaces(this.input, [detection]);
+    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]);
+    faces.forEach((f) => f instanceof Tensor && f.dispose());
+    return extendWithFaceLandmarks(parentResult, landmarks);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFacesTasks.ts
+var DetectFacesTaskBase = class extends ComposableTask {
+  constructor(input2, options = new SsdMobilenetv1Options()) {
+    super();
+    this.input = input2;
+    this.options = options;
+  }
+};
+var DetectAllFacesTask = class extends DetectFacesTaskBase {
+  async run() {
+    const { input: input2, options } = this;
+    let result;
+    if (options instanceof TinyFaceDetectorOptions)
+      result = nets.tinyFaceDetector.locateFaces(input2, options);
+    else if (options instanceof SsdMobilenetv1Options)
+      result = nets.ssdMobilenetv1.locateFaces(input2, options);
+    else if (options instanceof TinyYolov2Options)
+      result = nets.tinyYolov2.locateFaces(input2, options);
+    else
+      throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+    return result;
+  }
+  runAndExtendWithFaceDetections() {
+    return new Promise((resolve, reject) => {
+      this.run().then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection)))).catch((err) => reject(err));
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectAllFaceLandmarksTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+};
+var DetectSingleFaceTask = class extends DetectFacesTaskBase {
+  async run() {
+    const faceDetections = await new DetectAllFacesTask(this.input, this.options);
+    let faceDetectionWithHighestScore = faceDetections[0];
+    faceDetections.forEach((faceDetection) => {
+      if (faceDetection.score > faceDetectionWithHighestScore.score)
+        faceDetectionWithHighestScore = faceDetection;
+    });
+    return faceDetectionWithHighestScore;
+  }
+  runAndExtendWithFaceDetection() {
+    return new Promise(async (resolve) => {
+      const detection = await this.run();
+      resolve(detection ? extendWithFaceDetection({}, detection) : void 0);
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectSingleFaceLandmarksTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+};
+
+// src/globalApi/detectFaces.ts
+function detectSingleFace(input2, options = new SsdMobilenetv1Options()) {
+  return new DetectSingleFaceTask(input2, options);
+}
+function detectAllFaces(input2, options = new SsdMobilenetv1Options()) {
+  return new DetectAllFacesTask(input2, options);
+}
+
+// src/globalApi/allFaces.ts
+async function allFacesSsdMobilenetv1(input2, minConfidence) {
+  return detectAllFaces(input2, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {})).withFaceLandmarks().withFaceDescriptors();
+}
+async function allFacesTinyYolov2(input2, forwardParams = {}) {
+  return detectAllFaces(input2, new TinyYolov2Options(forwardParams)).withFaceLandmarks().withFaceDescriptors();
+}
+var allFaces = allFacesSsdMobilenetv1;
+
+// src/euclideanDistance.ts
+function euclideanDistance(arr1, arr2) {
+  if (arr1.length !== arr2.length)
+    throw new Error("euclideanDistance: arr1.length !== arr2.length");
+  const desc1 = Array.from(arr1);
+  const desc2 = Array.from(arr2);
+  return Math.sqrt(
+    desc1.map((val, i) => val - desc2[i]).reduce((res, diff) => res + diff * diff, 0)
+  );
+}
+
+// src/globalApi/FaceMatcher.ts
+var FaceMatcher = class {
+  constructor(inputs, distanceThreshold = 0.6) {
+    this._distanceThreshold = distanceThreshold;
+    const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+    if (!inputArray.length)
+      throw new Error("FaceRecognizer.constructor - expected atleast one input");
+    let count2 = 1;
+    const createUniqueLabel = () => `person ${count2++}`;
+    this._labeledDescriptors = inputArray.map((desc) => {
+      if (desc instanceof LabeledFaceDescriptors)
+        return desc;
+      if (desc instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);
+      if (desc.descriptor && desc.descriptor instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);
+      throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>");
+    });
+  }
+  get labeledDescriptors() {
+    return this._labeledDescriptors;
+  }
+  get distanceThreshold() {
+    return this._distanceThreshold;
+  }
+  computeMeanDistance(queryDescriptor, descriptors) {
+    return descriptors.map((d) => euclideanDistance(d, queryDescriptor)).reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);
+  }
+  matchDescriptor(queryDescriptor) {
+    return this.labeledDescriptors.map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors))).reduce((best, curr) => best.distance < curr.distance ? best : curr);
+  }
+  findBestMatch(queryDescriptor) {
+    const bestMatch = this.matchDescriptor(queryDescriptor);
+    return bestMatch.distance < this._distanceThreshold ? bestMatch : new FaceMatch("unknown", bestMatch.distance);
+  }
+  toJSON() {
+    return {
+      distanceThreshold: this._distanceThreshold,
+      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON())
+    };
+  }
+  static fromJSON(json20) {
+    const labeledDescriptors = json20.labeledDescriptors.map((ld) => LabeledFaceDescriptors.fromJSON(ld));
+    return new FaceMatcher(labeledDescriptors, json20.distanceThreshold);
+  }
+};
+
+// src/tinyFaceDetector/index.ts
+function createTinyFaceDetector(weights) {
+  const net = new TinyFaceDetector();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/resizeResults.ts
+function resizeResults(results, dimensions) {
+  const { width, height } = new Dimensions(dimensions.width, dimensions.height);
+  if (width <= 0 || height <= 0) {
+    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);
+  }
+  if (Array.isArray(results)) {
+    return results.map((obj) => resizeResults(obj, { width, height }));
+  }
+  if (isWithFaceLandmarks(results)) {
+    const resizedDetection = results.detection.forSize(width, height);
+    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);
+    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);
+  }
+  if (isWithFaceDetection(results)) {
+    return extendWithFaceDetection(results, results.detection.forSize(width, height));
+  }
+  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {
+    return results.forSize(width, height);
+  }
+  return results;
+}
+
+// src/index.ts
+var version7 = version5;
+export {
+  AgeGenderNet,
+  BoundingBox,
+  Box,
+  ComposableTask,
+  ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask,
+  DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask,
+  Dimensions,
+  FACE_EXPRESSION_LABELS,
+  FaceDetection,
+  FaceDetectionNet,
+  FaceExpressionNet,
+  FaceExpressions,
+  FaceLandmark68Net,
+  FaceLandmark68TinyNet,
+  FaceLandmarkNet,
+  FaceLandmarks,
+  FaceLandmarks5,
+  FaceLandmarks68,
+  FaceMatch,
+  FaceMatcher,
+  FaceRecognitionNet,
+  Gender,
+  LabeledBox,
+  LabeledFaceDescriptors,
+  NetInput,
+  NeuralNetwork,
+  ObjectDetection,
+  Point,
+  PredictedBox,
+  Rect,
+  SsdMobilenetv1,
+  SsdMobilenetv1Options,
+  TinyFaceDetector,
+  TinyFaceDetectorOptions,
+  TinyYolov2,
+  TinyYolov2Options,
+  allFaces,
+  allFacesSsdMobilenetv1,
+  allFacesTinyYolov2,
+  awaitMediaLoaded,
+  bufferToImage,
+  computeFaceDescriptor,
+  createCanvas2 as createCanvas,
+  createCanvasFromMedia,
+  createFaceDetectionNet,
+  createFaceRecognitionNet,
+  createSsdMobilenetv1,
+  createTinyFaceDetector,
+  createTinyYolov2,
+  detectAllFaces,
+  detectFaceLandmarks,
+  detectFaceLandmarksTiny,
+  detectLandmarks,
+  detectSingleFace,
+  draw_exports as draw,
+  env2 as env,
+  euclideanDistance,
+  extendWithAge,
+  extendWithFaceDescriptor,
+  extendWithFaceDetection,
+  extendWithFaceExpressions,
+  extendWithFaceLandmarks,
+  extendWithGender,
+  extractFaceTensors,
+  extractFaces,
+  fetchImage,
+  fetchJson,
+  fetchNetWeights,
+  fetchOrThrow,
+  fetchVideo,
+  getContext2dOrThrow,
+  getMediaDimensions,
+  imageTensorToCanvas,
+  imageToSquare,
+  inverseSigmoid,
+  iou,
+  isMediaElement,
+  isMediaLoaded,
+  isWithAge,
+  isWithFaceDetection,
+  isWithFaceExpressions,
+  isWithFaceLandmarks,
+  isWithGender,
+  loadAgeGenderModel,
+  loadFaceDetectionModel,
+  loadFaceExpressionModel,
+  loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel,
+  loadTinyYolov2Model,
+  loadWeightMap,
+  locateFaces,
+  matchDimensions,
+  minBbox,
+  nets,
+  nonMaxSuppression2 as nonMaxSuppression,
+  normalize,
+  padToSquare,
+  predictAgeAndGender,
+  recognizeFaceExpressions,
+  resizeResults,
+  resolveInput,
+  shuffleArray,
+  sigmoid5 as sigmoid,
+  ssdMobilenetv1,
+  tfjs_esm_exports as tf,
+  tinyFaceDetector,
+  tinyYolov2,
+  toNetInput,
+  utils_exports as utils,
+  validateConfig,
+  version7 as version
+};
 //# sourceMappingURL=face-api.esm.js.map
diff --git a/dist/face-api.esm.js.map b/dist/face-api.esm.js.map
index ae66c16..45851f9 100644
--- a/dist/face-api.esm.js.map
+++ b/dist/face-api.esm.js.map
@@ -1,7 +1,7 @@
 {
   "version": 3,
   "sources": ["tfjs.esm.js", "../src/draw/index.ts", "../src/draw/drawContour.ts", "../src/utils/index.ts", "../src/classes/Dimensions.ts", "../src/classes/Point.ts", "../src/classes/Box.ts", "../src/classes/BoundingBox.ts", "../src/classes/ObjectDetection.ts", "../src/classes/FaceDetection.ts", "../src/ops/iou.ts", "../src/ops/minBbox.ts", "../src/ops/nonMaxSuppression.ts", "../src/ops/normalize.ts", "../src/ops/padToSquare.ts", "../src/ops/shuffleArray.ts", "../src/ops/index.ts", "../src/classes/Rect.ts", "../src/classes/FaceLandmarks.ts", "../src/classes/FaceLandmarks5.ts", "../src/classes/FaceLandmarks68.ts", "../src/classes/FaceMatch.ts", "../src/classes/LabeledBox.ts", "../src/classes/LabeledFaceDescriptors.ts", "../src/classes/PredictedBox.ts", "../src/factories/WithFaceDetection.ts", "../src/env/createBrowserEnv.ts", "../src/env/isNodejs.ts", "../src/env/createFileSystem.ts", "../src/env/createNodejsEnv.ts", "../src/env/isBrowser.ts", "../src/env/index.ts", "../src/dom/resolveInput.ts", "../src/dom/getContext2dOrThrow.ts", "../src/draw/DrawTextField.ts", "../src/draw/DrawBox.ts", "../src/draw/drawDetections.ts", "../src/dom/isMediaLoaded.ts", "../src/dom/awaitMediaLoaded.ts", "../src/dom/bufferToImage.ts", "../src/dom/getMediaDimensions.ts", "../src/dom/createCanvas.ts", "../src/dom/imageTensorToCanvas.ts", "../src/dom/isMediaElement.ts", "../src/dom/imageToSquare.ts", "../src/dom/NetInput.ts", "../src/dom/toNetInput.ts", "../src/dom/extractFaces.ts", "../src/dom/extractFaceTensors.ts", "../src/dom/fetchOrThrow.ts", "../src/dom/fetchImage.ts", "../src/dom/fetchJson.ts", "../src/dom/fetchNetWeights.ts", "../src/dom/bufferToVideo.ts", "../src/dom/fetchVideo.ts", "../src/common/getModelUris.ts", "../src/dom/loadWeightMap.ts", "../src/dom/matchDimensions.ts", "../src/NeuralNetwork.ts", "../src/common/depthwiseSeparableConv.ts", "../src/faceFeatureExtractor/denseBlock.ts", "../src/common/convLayer.ts", "../src/common/disposeUnusedWeightTensors.ts", "../src/common/extractConvParamsFactory.ts", "../src/common/extractFCParamsFactory.ts", "../src/common/types.ts", "../src/common/extractSeparableConvParamsFactory.ts", "../src/common/extractWeightEntryFactory.ts", "../src/common/extractWeightsFactory.ts", "../src/faceFeatureExtractor/extractorsFactory.ts", "../src/faceFeatureExtractor/extractParams.ts", "../src/common/loadConvParamsFactory.ts", "../src/faceFeatureExtractor/loadParamsFactory.ts", "../src/faceFeatureExtractor/extractParamsFromWeightMap.ts", "../src/faceFeatureExtractor/FaceFeatureExtractor.ts", "../src/common/fullyConnectedLayer.ts", "../src/faceProcessor/extractParams.ts", "../src/faceProcessor/extractParamsFromWeightMap.ts", "../src/faceProcessor/util.ts", "../src/faceProcessor/FaceProcessor.ts", "../src/faceExpressionNet/FaceExpressions.ts", "../src/faceExpressionNet/FaceExpressionNet.ts", "../src/factories/WithFaceExpressions.ts", "../src/draw/drawFaceExpressions.ts", "../src/factories/WithFaceLandmarks.ts", "../src/draw/DrawFaceLandmarks.ts", "../src/xception/extractParams.ts", "../src/xception/extractParamsFromWeightMap.ts", "../src/xception/TinyXception.ts", "../src/ageGenderNet/extractParams.ts", "../src/ageGenderNet/extractParamsFromWeightMap.ts", "../src/ageGenderNet/types.ts", "../src/ageGenderNet/AgeGenderNet.ts", "../src/faceLandmarkNet/FaceLandmark68NetBase.ts", "../src/faceLandmarkNet/FaceLandmark68Net.ts", "../src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts", "../src/faceFeatureExtractor/extractParamsTiny.ts", "../src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts", "../src/faceLandmarkNet/FaceLandmark68TinyNet.ts", "../src/faceLandmarkNet/index.ts", "../src/faceRecognitionNet/scaleLayer.ts", "../src/faceRecognitionNet/convLayer.ts", "../src/faceRecognitionNet/extractParams.ts", "../src/faceRecognitionNet/extractParamsFromWeightMap.ts", "../src/faceRecognitionNet/residualLayer.ts", "../src/faceRecognitionNet/FaceRecognitionNet.ts", "../src/faceRecognitionNet/index.ts", "../src/factories/WithFaceDescriptor.ts", "../src/factories/WithAge.ts", "../src/factories/WithGender.ts", "../src/ssdMobilenetv1/extractParams.ts", "../src/ssdMobilenetv1/extractParamsFromWeightMap.ts", "../src/ssdMobilenetv1/pointwiseConvLayer.ts", "../src/ssdMobilenetv1/mobileNetV1.ts", "../src/ssdMobilenetv1/nonMaxSuppression.ts", "../src/ssdMobilenetv1/outputLayer.ts", "../src/ssdMobilenetv1/boxPredictionLayer.ts", "../src/ssdMobilenetv1/predictionLayer.ts", "../src/ssdMobilenetv1/SsdMobilenetv1Options.ts", "../src/ssdMobilenetv1/SsdMobilenetv1.ts", "../src/ssdMobilenetv1/index.ts", "../src/tinyYolov2/const.ts", "../src/tinyYolov2/config.ts", "../src/tinyYolov2/leaky.ts", "../src/tinyYolov2/convWithBatchNorm.ts", "../src/tinyYolov2/depthwiseSeparableConv.ts", "../src/tinyYolov2/extractParams.ts", "../src/tinyYolov2/extractParamsFromWeightMap.ts", "../src/tinyYolov2/TinyYolov2Options.ts", "../src/tinyYolov2/TinyYolov2Base.ts", "../src/tinyYolov2/TinyYolov2.ts", "../src/tinyYolov2/index.ts", "../src/tinyFaceDetector/TinyFaceDetectorOptions.ts", "../src/globalApi/ComposableTask.ts", "../src/globalApi/extractFacesAndComputeResults.ts", "../src/tinyFaceDetector/const.ts", "../src/tinyFaceDetector/TinyFaceDetector.ts", "../src/globalApi/nets.ts", "../src/globalApi/PredictFaceExpressionsTask.ts", "../src/globalApi/PredictAgeAndGenderTask.ts", "../src/globalApi/ComputeFaceDescriptorsTasks.ts", "../src/globalApi/DetectFaceLandmarksTasks.ts", "../src/globalApi/DetectFacesTasks.ts", "../src/globalApi/detectFaces.ts", "../src/globalApi/allFaces.ts", "../src/euclideanDistance.ts", "../src/globalApi/FaceMatcher.ts", "../src/tinyFaceDetector/index.ts", "../src/resizeResults.ts", "../src/index.ts"],
-  "sourcesContent": ["/*\n  Face-API\n  homepage: <https://github.com/vladmandic/face-api>\n  author: <https://github.com/vladmandic>'\n*/\n\nvar DU=Object.create;var QS=Object.defineProperty;var RU=Object.getOwnPropertyDescriptor;var FU=Object.getOwnPropertyNames;var OU=Object.getPrototypeOf,PU=Object.prototype.hasOwnProperty;var Pg=(r=>typeof require!=\"undefined\"?require:typeof Proxy!=\"undefined\"?new Proxy(r,{get:(t,e)=>(typeof require!=\"undefined\"?require:t)[e]}):r)(function(r){if(typeof require!=\"undefined\")return require.apply(this,arguments);throw new Error('Dynamic require of \"'+r+'\" is not supported')});var gr=(r,t)=>()=>(t||r((t={exports:{}}).exports,t),t.exports),Wt=(r,t)=>{for(var e in t)QS(r,e,{get:t[e],enumerable:!0})},LU=(r,t,e,n)=>{if(t&&typeof t==\"object\"||typeof t==\"function\")for(let o of FU(t))!PU.call(r,o)&&o!==e&&QS(r,o,{get:()=>t[o],enumerable:!(n=RU(t,o))||n.enumerable});return r};var Tl=(r,t,e)=>(e=r!=null?DU(OU(r)):{},LU(t||!r||!r.__esModule?QS(e,\"default\",{value:r,enumerable:!0}):e,r));var T1=gr((Elt,N1)=>{N1.exports=Ue;var mo=null;try{mo=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(r){}function Ue(r,t,e){this.low=r|0,this.high=t|0,this.unsigned=!!e}Ue.prototype.__isLong__;Object.defineProperty(Ue.prototype,\"__isLong__\",{value:!0});function Fn(r){return(r&&r.__isLong__)===!0}Ue.isLong=Fn;var g1={},x1={};function zu(r,t){var e,n,o;return t?(r>>>=0,(o=0<=r&&r<256)&&(n=x1[r],n)?n:(e=He(r,(r|0)<0?-1:0,!0),o&&(x1[r]=e),e)):(r|=0,(o=-128<=r&&r<128)&&(n=g1[r],n)?n:(e=He(r,r<0?-1:0,!1),o&&(g1[r]=e),e))}Ue.fromInt=zu;function fo(r,t){if(isNaN(r))return t?Mu:ho;if(t){if(r<0)return Mu;if(r>=C1)return v1}else{if(r<=-b1)return Rn;if(r+1>=b1)return S1}return r<0?fo(-r,t).neg():He(r%Vp|0,r/Vp|0,t)}Ue.fromNumber=fo;function He(r,t,e){return new Ue(r,t,e)}Ue.fromBits=He;var Bg=Math.pow;function m0(r,t,e){if(r.length===0)throw Error(\"empty string\");if(r===\"NaN\"||r===\"Infinity\"||r===\"+Infinity\"||r===\"-Infinity\")return ho;if(typeof t==\"number\"?(e=t,t=!1):t=!!t,e=e||10,e<2||36<e)throw RangeError(\"radix\");var n;if((n=r.indexOf(\"-\"))>0)throw Error(\"interior hyphen\");if(n===0)return m0(r.substring(1),t,e).neg();for(var o=fo(Bg(e,8)),s=ho,i=0;i<r.length;i+=8){var a=Math.min(8,r.length-i),u=parseInt(r.substring(i,i+a),e);if(a<8){var l=fo(Bg(e,a));s=s.mul(l).add(fo(u))}else s=s.mul(o),s=s.add(fo(u))}return s.unsigned=t,s}Ue.fromString=m0;function Ls(r,t){return typeof r==\"number\"?fo(r,t):typeof r==\"string\"?m0(r,t):He(r.low,r.high,typeof t==\"boolean\"?t:r.unsigned)}Ue.fromValue=Ls;var y1=1<<16,s4=1<<24,Vp=y1*y1,C1=Vp*Vp,b1=C1/2,w1=zu(s4),ho=zu(0);Ue.ZERO=ho;var Mu=zu(0,!0);Ue.UZERO=Mu;var Bp=zu(1);Ue.ONE=Bp;var I1=zu(1,!0);Ue.UONE=I1;var p0=zu(-1);Ue.NEG_ONE=p0;var S1=He(-1,2147483647,!1);Ue.MAX_VALUE=S1;var v1=He(-1,-1,!0);Ue.MAX_UNSIGNED_VALUE=v1;var Rn=He(0,-2147483648,!1);Ue.MIN_VALUE=Rn;var gt=Ue.prototype;gt.toInt=function(){return this.unsigned?this.low>>>0:this.low};gt.toNumber=function(){return this.unsigned?(this.high>>>0)*Vp+(this.low>>>0):this.high*Vp+(this.low>>>0)};gt.toString=function(t){if(t=t||10,t<2||36<t)throw RangeError(\"radix\");if(this.isZero())return\"0\";if(this.isNegative())if(this.eq(Rn)){var e=fo(t),n=this.div(e),o=n.mul(e).sub(this);return n.toString(t)+o.toInt().toString(t)}else return\"-\"+this.neg().toString(t);for(var s=fo(Bg(t,6),this.unsigned),i=this,a=\"\";;){var u=i.div(s),l=i.sub(u.mul(s)).toInt()>>>0,c=l.toString(t);if(i=u,i.isZero())return c+a;for(;c.length<6;)c=\"0\"+c;a=\"\"+c+a}};gt.getHighBits=function(){return this.high};gt.getHighBitsUnsigned=function(){return this.high>>>0};gt.getLowBits=function(){return this.low};gt.getLowBitsUnsigned=function(){return this.low>>>0};gt.getNumBitsAbs=function(){if(this.isNegative())return this.eq(Rn)?64:this.neg().getNumBitsAbs();for(var t=this.high!=0?this.high:this.low,e=31;e>0&&(t&1<<e)==0;e--);return this.high!=0?e+33:e+1};gt.isZero=function(){return this.high===0&&this.low===0};gt.eqz=gt.isZero;gt.isNegative=function(){return!this.unsigned&&this.high<0};gt.isPositive=function(){return this.unsigned||this.high>=0};gt.isOdd=function(){return(this.low&1)===1};gt.isEven=function(){return(this.low&1)===0};gt.equals=function(t){return Fn(t)||(t=Ls(t)),this.unsigned!==t.unsigned&&this.high>>>31===1&&t.high>>>31===1?!1:this.high===t.high&&this.low===t.low};gt.eq=gt.equals;gt.notEquals=function(t){return!this.eq(t)};gt.neq=gt.notEquals;gt.ne=gt.notEquals;gt.lessThan=function(t){return this.comp(t)<0};gt.lt=gt.lessThan;gt.lessThanOrEqual=function(t){return this.comp(t)<=0};gt.lte=gt.lessThanOrEqual;gt.le=gt.lessThanOrEqual;gt.greaterThan=function(t){return this.comp(t)>0};gt.gt=gt.greaterThan;gt.greaterThanOrEqual=function(t){return this.comp(t)>=0};gt.gte=gt.greaterThanOrEqual;gt.ge=gt.greaterThanOrEqual;gt.compare=function(t){if(Fn(t)||(t=Ls(t)),this.eq(t))return 0;var e=this.isNegative(),n=t.isNegative();return e&&!n?-1:!e&&n?1:this.unsigned?t.high>>>0>this.high>>>0||t.high===this.high&&t.low>>>0>this.low>>>0?-1:1:this.sub(t).isNegative()?-1:1};gt.comp=gt.compare;gt.negate=function(){return!this.unsigned&&this.eq(Rn)?Rn:this.not().add(Bp)};gt.neg=gt.negate;gt.add=function(t){Fn(t)||(t=Ls(t));var e=this.high>>>16,n=this.high&65535,o=this.low>>>16,s=this.low&65535,i=t.high>>>16,a=t.high&65535,u=t.low>>>16,l=t.low&65535,c=0,p=0,m=0,f=0;return f+=s+l,m+=f>>>16,f&=65535,m+=o+u,p+=m>>>16,m&=65535,p+=n+a,c+=p>>>16,p&=65535,c+=e+i,c&=65535,He(m<<16|f,c<<16|p,this.unsigned)};gt.subtract=function(t){return Fn(t)||(t=Ls(t)),this.add(t.neg())};gt.sub=gt.subtract;gt.multiply=function(t){if(this.isZero())return ho;if(Fn(t)||(t=Ls(t)),mo){var e=mo.mul(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}if(t.isZero())return ho;if(this.eq(Rn))return t.isOdd()?Rn:ho;if(t.eq(Rn))return this.isOdd()?Rn:ho;if(this.isNegative())return t.isNegative()?this.neg().mul(t.neg()):this.neg().mul(t).neg();if(t.isNegative())return this.mul(t.neg()).neg();if(this.lt(w1)&&t.lt(w1))return fo(this.toNumber()*t.toNumber(),this.unsigned);var n=this.high>>>16,o=this.high&65535,s=this.low>>>16,i=this.low&65535,a=t.high>>>16,u=t.high&65535,l=t.low>>>16,c=t.low&65535,p=0,m=0,f=0,d=0;return d+=i*c,f+=d>>>16,d&=65535,f+=s*c,m+=f>>>16,f&=65535,f+=i*l,m+=f>>>16,f&=65535,m+=o*c,p+=m>>>16,m&=65535,m+=s*l,p+=m>>>16,m&=65535,m+=i*u,p+=m>>>16,m&=65535,p+=n*c+o*l+s*u+i*a,p&=65535,He(f<<16|d,p<<16|m,this.unsigned)};gt.mul=gt.multiply;gt.divide=function(t){if(Fn(t)||(t=Ls(t)),t.isZero())throw Error(\"division by zero\");if(mo){if(!this.unsigned&&this.high===-2147483648&&t.low===-1&&t.high===-1)return this;var e=(this.unsigned?mo.div_u:mo.div_s)(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?Mu:ho;var n,o,s;if(this.unsigned){if(t.unsigned||(t=t.toUnsigned()),t.gt(this))return Mu;if(t.gt(this.shru(1)))return I1;s=Mu}else{if(this.eq(Rn)){if(t.eq(Bp)||t.eq(p0))return Rn;if(t.eq(Rn))return Bp;var i=this.shr(1);return n=i.div(t).shl(1),n.eq(ho)?t.isNegative()?Bp:p0:(o=this.sub(t.mul(n)),s=n.add(o.div(t)),s)}else if(t.eq(Rn))return this.unsigned?Mu:ho;if(this.isNegative())return t.isNegative()?this.neg().div(t.neg()):this.neg().div(t).neg();if(t.isNegative())return this.div(t.neg()).neg();s=ho}for(o=this;o.gte(t);){n=Math.max(1,Math.floor(o.toNumber()/t.toNumber()));for(var a=Math.ceil(Math.log(n)/Math.LN2),u=a<=48?1:Bg(2,a-48),l=fo(n),c=l.mul(t);c.isNegative()||c.gt(o);)n-=u,l=fo(n,this.unsigned),c=l.mul(t);l.isZero()&&(l=Bp),s=s.add(l),o=o.sub(c)}return s};gt.div=gt.divide;gt.modulo=function(t){if(Fn(t)||(t=Ls(t)),mo){var e=(this.unsigned?mo.rem_u:mo.rem_s)(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}return this.sub(this.div(t).mul(t))};gt.mod=gt.modulo;gt.rem=gt.modulo;gt.not=function(){return He(~this.low,~this.high,this.unsigned)};gt.and=function(t){return Fn(t)||(t=Ls(t)),He(this.low&t.low,this.high&t.high,this.unsigned)};gt.or=function(t){return Fn(t)||(t=Ls(t)),He(this.low|t.low,this.high|t.high,this.unsigned)};gt.xor=function(t){return Fn(t)||(t=Ls(t)),He(this.low^t.low,this.high^t.high,this.unsigned)};gt.shiftLeft=function(t){return Fn(t)&&(t=t.toInt()),(t&=63)===0?this:t<32?He(this.low<<t,this.high<<t|this.low>>>32-t,this.unsigned):He(0,this.low<<t-32,this.unsigned)};gt.shl=gt.shiftLeft;gt.shiftRight=function(t){return Fn(t)&&(t=t.toInt()),(t&=63)===0?this:t<32?He(this.low>>>t|this.high<<32-t,this.high>>t,this.unsigned):He(this.high>>t-32,this.high>=0?0:-1,this.unsigned)};gt.shr=gt.shiftRight;gt.shiftRightUnsigned=function(t){if(Fn(t)&&(t=t.toInt()),t&=63,t===0)return this;var e=this.high;if(t<32){var n=this.low;return He(n>>>t|e<<32-t,e>>>t,this.unsigned)}else return t===32?He(e,0,this.unsigned):He(e>>>t-32,0,this.unsigned)};gt.shru=gt.shiftRightUnsigned;gt.shr_u=gt.shiftRightUnsigned;gt.toSigned=function(){return this.unsigned?He(this.low,this.high,!1):this};gt.toUnsigned=function(){return this.unsigned?this:He(this.low,this.high,!0)};gt.toBytes=function(t){return t?this.toBytesLE():this.toBytesBE()};gt.toBytesLE=function(){var t=this.high,e=this.low;return[e&255,e>>>8&255,e>>>16&255,e>>>24,t&255,t>>>8&255,t>>>16&255,t>>>24]};gt.toBytesBE=function(){var t=this.high,e=this.low;return[t>>>24,t>>>16&255,t>>>8&255,t&255,e>>>24,e>>>16&255,e>>>8&255,e&255]};Ue.fromBytes=function(t,e,n){return n?Ue.fromBytesLE(t,e):Ue.fromBytesBE(t,e)};Ue.fromBytesLE=function(t,e){return new Ue(t[0]|t[1]<<8|t[2]<<16|t[3]<<24,t[4]|t[5]<<8|t[6]<<16|t[7]<<24,e)};Ue.fromBytesBE=function(t,e){return new Ue(t[4]<<24|t[5]<<16|t[6]<<8|t[7],t[0]<<24|t[1]<<16|t[2]<<8|t[3],e)}});var iE=gr(()=>{});var aE=gr(()=>{});var u_=gr((l_,rv)=>{(function(r,t,e){function n(a){var u=this,l=i();u.next=function(){var c=2091639*u.s0+u.c*23283064365386963e-26;return u.s0=u.s1,u.s1=u.s2,u.s2=c-(u.c=c|0)},u.c=1,u.s0=l(\" \"),u.s1=l(\" \"),u.s2=l(\" \"),u.s0-=l(a),u.s0<0&&(u.s0+=1),u.s1-=l(a),u.s1<0&&(u.s1+=1),u.s2-=l(a),u.s2<0&&(u.s2+=1),l=null}function o(a,u){return u.c=a.c,u.s0=a.s0,u.s1=a.s1,u.s2=a.s2,u}function s(a,u){var l=new n(a),c=u&&u.state,p=l.next;return p.int32=function(){return l.next()*4294967296|0},p.double=function(){return p()+(p()*2097152|0)*11102230246251565e-32},p.quick=p,c&&(typeof c==\"object\"&&o(c,l),p.state=function(){return o(l,{})}),p}function i(){var a=4022871197,u=function(l){l=String(l);for(var c=0;c<l.length;c++){a+=l.charCodeAt(c);var p=.02519603282416938*a;a=p>>>0,p-=a,p*=a,a=p>>>0,p-=a,a+=p*4294967296}return(a>>>0)*23283064365386963e-26};return u}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.alea=s})(l_,typeof rv==\"object\"&&rv,typeof define==\"function\"&&define)});var p_=gr((c_,nv)=>{(function(r,t,e){function n(i){var a=this,u=\"\";a.x=0,a.y=0,a.z=0,a.w=0,a.next=function(){var c=a.x^a.x<<11;return a.x=a.y,a.y=a.z,a.z=a.w,a.w^=a.w>>>19^c^c>>>8},i===(i|0)?a.x=i:u+=i;for(var l=0;l<u.length+64;l++)a.x^=u.charCodeAt(l)|0,a.next()}function o(i,a){return a.x=i.x,a.y=i.y,a.z=i.z,a.w=i.w,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l==\"object\"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xor128=s})(c_,typeof nv==\"object\"&&nv,typeof define==\"function\"&&define)});var f_=gr((m_,ov)=>{(function(r,t,e){function n(i){var a=this,u=\"\";a.next=function(){var c=a.x^a.x>>>2;return a.x=a.y,a.y=a.z,a.z=a.w,a.w=a.v,(a.d=a.d+362437|0)+(a.v=a.v^a.v<<4^(c^c<<1))|0},a.x=0,a.y=0,a.z=0,a.w=0,a.v=0,i===(i|0)?a.x=i:u+=i;for(var l=0;l<u.length+64;l++)a.x^=u.charCodeAt(l)|0,l==u.length&&(a.d=a.x<<10^a.x>>>4),a.next()}function o(i,a){return a.x=i.x,a.y=i.y,a.z=i.z,a.w=i.w,a.v=i.v,a.d=i.d,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l==\"object\"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xorwow=s})(m_,typeof ov==\"object\"&&ov,typeof define==\"function\"&&define)});var h_=gr((d_,sv)=>{(function(r,t,e){function n(i){var a=this;a.next=function(){var l=a.x,c=a.i,p,m,f;return p=l[c],p^=p>>>7,m=p^p<<24,p=l[c+1&7],m^=p^p>>>10,p=l[c+3&7],m^=p^p>>>3,p=l[c+4&7],m^=p^p<<7,p=l[c+7&7],p=p^p<<13,m^=p^p<<9,l[c]=m,a.i=c+1&7,m};function u(l,c){var p,m,f=[];if(c===(c|0))m=f[0]=c;else for(c=\"\"+c,p=0;p<c.length;++p)f[p&7]=f[p&7]<<15^c.charCodeAt(p)+f[p+1&7]<<13;for(;f.length<8;)f.push(0);for(p=0;p<8&&f[p]===0;++p);for(p==8?m=f[7]=-1:m=f[p],l.x=f,l.i=0,p=256;p>0;--p)l.next()}u(a,i)}function o(i,a){return a.x=i.x.slice(),a.i=i.i,a}function s(i,a){i==null&&(i=+new Date);var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(l.x&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xorshift7=s})(d_,typeof sv==\"object\"&&sv,typeof define==\"function\"&&define)});var x_=gr((g_,iv)=>{(function(r,t,e){function n(i){var a=this;a.next=function(){var l=a.w,c=a.X,p=a.i,m,f;return a.w=l=l+1640531527|0,f=c[p+34&127],m=c[p=p+1&127],f^=f<<13,m^=m<<17,f^=f>>>15,m^=m>>>12,f=c[p]=f^m,a.i=p,f+(l^l>>>16)|0};function u(l,c){var p,m,f,d,h,g=[],x=128;for(c===(c|0)?(m=c,c=null):(c=c+\"\\0\",m=0,x=Math.max(x,c.length)),f=0,d=-32;d<x;++d)c&&(m^=c.charCodeAt((d+32)%c.length)),d===0&&(h=m),m^=m<<10,m^=m>>>15,m^=m<<4,m^=m>>>13,d>=0&&(h=h+1640531527|0,p=g[d&127]^=m+h,f=p==0?f+1:0);for(f>=128&&(g[(c&&c.length||0)&127]=-1),f=127,d=4*128;d>0;--d)m=g[f+34&127],p=g[f=f+1&127],m^=m<<13,p^=p<<17,m^=m>>>15,p^=p>>>12,g[f]=m^p;l.w=h,l.X=g,l.i=f}u(a,i)}function o(i,a){return a.i=i.i,a.w=i.w,a.X=i.X.slice(),a}function s(i,a){i==null&&(i=+new Date);var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(l.X&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xor4096=s})(g_,typeof iv==\"object\"&&iv,typeof define==\"function\"&&define)});var b_=gr((y_,av)=>{(function(r,t,e){function n(i){var a=this,u=\"\";a.next=function(){var c=a.b,p=a.c,m=a.d,f=a.a;return c=c<<25^c>>>7^p,p=p-m|0,m=m<<24^m>>>8^f,f=f-c|0,a.b=c=c<<20^c>>>12^p,a.c=p=p-m|0,a.d=m<<16^p>>>16^f,a.a=f-c|0},a.a=0,a.b=0,a.c=-1640531527,a.d=1367130551,i===Math.floor(i)?(a.a=i/4294967296|0,a.b=i|0):u+=i;for(var l=0;l<u.length+20;l++)a.b^=u.charCodeAt(l)|0,a.next()}function o(i,a){return a.a=i.a,a.b=i.b,a.c=i.c,a.d=i.d,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l==\"object\"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.tychei=s})(y_,typeof av==\"object\"&&av,typeof define==\"function\"&&define)});var w_=gr(()=>{});var I_=gr((C_,Yx)=>{(function(r,t,e){var n=256,o=6,s=52,i=\"random\",a=e.pow(n,o),u=e.pow(2,s),l=u*2,c=n-1,p;function m(w,C,N){var _=[];C=C==!0?{entropy:!0}:C||{};var A=g(h(C.entropy?[w,b(t)]:w==null?x():w,3),_),$=new f(_),F=function(){for(var P=$.g(o),V=a,G=0;P<u;)P=(P+G)*n,V*=n,G=$.g(1);for(;P>=l;)P/=2,V/=2,G>>>=1;return(P+G)/V};return F.int32=function(){return $.g(4)|0},F.quick=function(){return $.g(4)/4294967296},F.double=F,g(b($.S),t),(C.pass||N||function(P,V,G,W){return W&&(W.S&&d(W,$),P.state=function(){return d($,{})}),G?(e[i]=P,V):P})(F,A,\"global\"in C?C.global:this==e,C.state)}function f(w){var C,N=w.length,_=this,A=0,$=_.i=_.j=0,F=_.S=[];for(N||(w=[N++]);A<n;)F[A]=A++;for(A=0;A<n;A++)F[A]=F[$=c&$+w[A%N]+(C=F[A])],F[$]=C;(_.g=function(P){for(var V,G=0,W=_.i,q=_.j,H=_.S;P--;)V=H[W=c&W+1],G=G*n+H[c&(H[W]=H[q=c&q+V])+(H[q]=V)];return _.i=W,_.j=q,G})(n)}function d(w,C){return C.i=w.i,C.j=w.j,C.S=w.S.slice(),C}function h(w,C){var N=[],_=typeof w,A;if(C&&_==\"object\")for(A in w)try{N.push(h(w[A],C-1))}catch($){}return N.length?N:_==\"string\"?w:w+\"\\0\"}function g(w,C){for(var N=w+\"\",_,A=0;A<N.length;)C[c&A]=c&(_^=C[c&A]*19)+N.charCodeAt(A++);return b(C)}function x(){try{var w;return p&&(w=p.randomBytes)?w=w(n):(w=new Uint8Array(n),(r.crypto||r.msCrypto).getRandomValues(w)),b(w)}catch(_){var C=r.navigator,N=C&&C.plugins;return[+new Date,r,N,r.screen,b(t)]}}function b(w){return String.fromCharCode.apply(0,w)}if(g(e.random(),t),typeof Yx==\"object\"&&Yx.exports){Yx.exports=m;try{p=w_()}catch(w){}}else typeof define==\"function\"&&define.amd?define(function(){return m}):e[\"seed\"+i]=m})(typeof self!=\"undefined\"?self:C_,[],Math)});var fh=gr((fNt,S_)=>{var lj=u_(),uj=p_(),cj=f_(),pj=h_(),mj=x_(),fj=b_(),Ju=I_();Ju.alea=lj;Ju.xor128=uj;Ju.xorwow=cj;Ju.xorshift7=pj;Ju.xor4096=mj;Ju.tychei=fj;S_.exports=Ju});var gN=gr(()=>{});var Zb=gr(()=>{});var ig=gr(()=>{});var BW=gr(()=>{});var VW=gr(()=>{});var GW=gr(()=>{});var WW=gr((fI,Ok)=>{var Fk=(()=>{var r=typeof document!=\"undefined\"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!=\"undefined\"&&(r=r||__filename),function(t){t=t||{};function e(){return it.buffer!=jt&&qe(it.buffer),ke}function n(){return it.buffer!=jt&&qe(it.buffer),fe}function o(){return it.buffer!=jt&&qe(it.buffer),Ae}function s(){return it.buffer!=jt&&qe(it.buffer),_n}function i(){return it.buffer!=jt&&qe(it.buffer),or}function a(){return it.buffer!=jt&&qe(it.buffer),Hn}function u(){return it.buffer!=jt&&qe(it.buffer),Lr}var l=typeof t!=\"undefined\"?t:{},c,p;l.ready=new Promise(function(L,U){c=L,p=U});var m;typeof process!=\"undefined\"&&process.listeners&&(m={uncaughtException:process.listeners(\"uncaughtException\"),unhandledRejection:process.listeners(\"unhandledRejection\")});var f=Object.assign({},l),d=[],h=\"./this.program\",g=(L,U)=>{throw U},x=typeof window==\"object\",b=typeof importScripts==\"function\",w=typeof process==\"object\"&&typeof process.versions==\"object\"&&typeof process.versions.node==\"string\",C=l.ENVIRONMENT_IS_PTHREAD||!1,N=\"\";function _(L){return l.locateFile?l.locateFile(L,N):N+L}var A,$,F,P;function V(L){if(L instanceof Ru)return;Y(\"exiting due to exception: \"+L)}if(w){b?N=ig().dirname(N)+\"/\":N=__dirname+\"/\";var G,W;typeof Pg==\"function\"&&(G=Zb(),W=ig()),A=(U,ut)=>(U=W.normalize(U),G.readFileSync(U,ut?void 0:\"utf8\")),F=U=>{var ut=A(U,!0);return ut.buffer||(ut=new Uint8Array(ut)),ut},$=(U,ut,xt)=>{U=W.normalize(U),G.readFile(U,function(Dt,ie){Dt?xt(Dt):ut(ie.buffer)})},process.argv.length>1&&(h=process.argv[1].replace(/\\\\/g,\"/\")),d=process.argv.slice(2),process.on(\"uncaughtException\",function(U){if(!(U instanceof Ru))throw U}),process.on(\"unhandledRejection\",function(U){throw U}),g=(U,ut)=>{if(Mo())throw process.exitCode=U,ut;V(ut),process.exit(U)},l.inspect=function(){return\"[Emscripten Module object]\"};let L;try{L=BW()}catch(U){throw console.error('The \"worker_threads\" module is not supported in this node.js build - perhaps a newer version is needed?'),U}global.Worker=L.Worker}else(x||b)&&(b?N=self.location.href:typeof document!=\"undefined\"&&document.currentScript&&(N=document.currentScript.src),typeof r!=\"undefined\"&&r&&(N=r),N.indexOf(\"blob:\")!==0?N=N.substr(0,N.replace(/[?#].*/,\"\").lastIndexOf(\"/\")+1):N=\"\",w||(A=L=>{var U=new XMLHttpRequest;return U.open(\"GET\",L,!1),U.send(null),U.responseText},b&&(F=L=>{var U=new XMLHttpRequest;return U.open(\"GET\",L,!1),U.responseType=\"arraybuffer\",U.send(null),new Uint8Array(U.response)}),$=(L,U,ut)=>{var xt=new XMLHttpRequest;xt.open(\"GET\",L,!0),xt.responseType=\"arraybuffer\",xt.onload=()=>{if(xt.status==200||xt.status==0&&xt.response){U(xt.response);return}ut()},xt.onerror=ut,xt.send(null)}),P=L=>document.title=L);w&&typeof performance==\"undefined\"&&(global.performance=VW().performance);var q=console.log.bind(console),H=console.warn.bind(console);w&&(q=L=>G.writeSync(1,L+`\n`),H=L=>G.writeSync(2,L+`\n`));var j=l.print||q,Y=l.printErr||H;Object.assign(l,f),f=null,l.arguments&&(d=l.arguments),l.thisProgram&&(h=l.thisProgram),l.quit&&(g=l.quit);var Z=4,et=Atomics.load,rt=Atomics.store,ot=Atomics.compareExchange,at;l.wasmBinary&&(at=l.wasmBinary);var nt=l.noExitRuntime||!0;typeof WebAssembly!=\"object\"&&Yc(\"no native wasm support detected\");var it,dt,ht=!1,bt;function Et(L,U){L||Yc(U)}var At=typeof TextDecoder!=\"undefined\"?new TextDecoder(\"utf8\"):void 0;function Vt(L,U,ut){for(var xt=U+ut,Dt=U;L[Dt]&&!(Dt>=xt);)++Dt;if(Dt-U>16&&L.buffer&&At)return At.decode(L.buffer instanceof SharedArrayBuffer?L.slice(U,Dt):L.subarray(U,Dt));for(var ie=\"\";U<Dt;){var Gt=L[U++];if(!(Gt&128)){ie+=String.fromCharCode(Gt);continue}var Xt=L[U++]&63;if((Gt&224)==192){ie+=String.fromCharCode((Gt&31)<<6|Xt);continue}var hr=L[U++]&63;if((Gt&240)==224?Gt=(Gt&15)<<12|Xt<<6|hr:Gt=(Gt&7)<<18|Xt<<12|hr<<6|L[U++]&63,Gt<65536)ie+=String.fromCharCode(Gt);else{var Xn=Gt-65536;ie+=String.fromCharCode(55296|Xn>>10,56320|Xn&1023)}}return ie}function Zt(L,U){return L?Vt(n(),L,U):\"\"}function ce(L,U,ut,xt){if(!(xt>0))return 0;for(var Dt=ut,ie=ut+xt-1,Gt=0;Gt<L.length;++Gt){var Xt=L.charCodeAt(Gt);if(Xt>=55296&&Xt<=57343){var hr=L.charCodeAt(++Gt);Xt=65536+((Xt&1023)<<10)|hr&1023}if(Xt<=127){if(ut>=ie)break;U[ut++]=Xt}else if(Xt<=2047){if(ut+1>=ie)break;U[ut++]=192|Xt>>6,U[ut++]=128|Xt&63}else if(Xt<=65535){if(ut+2>=ie)break;U[ut++]=224|Xt>>12,U[ut++]=128|Xt>>6&63,U[ut++]=128|Xt&63}else{if(ut+3>=ie)break;U[ut++]=240|Xt>>18,U[ut++]=128|Xt>>12&63,U[ut++]=128|Xt>>6&63,U[ut++]=128|Xt&63}}return U[ut]=0,ut-Dt}function he(L,U,ut){return ce(L,n(),U,ut)}var jt,ke,fe,Ae,We,_n,or,Hn,Lr;C&&(jt=l.buffer);function qe(L){jt=L,l.HEAP8=ke=new Int8Array(L),l.HEAP16=Ae=new Int16Array(L),l.HEAP32=_n=new Int32Array(L),l.HEAPU8=fe=new Uint8Array(L),l.HEAPU16=We=new Uint16Array(L),l.HEAPU32=or=new Uint32Array(L),l.HEAPF32=Hn=new Float32Array(L),l.HEAPF64=Lr=new Float64Array(L)}var Mr=l.INITIAL_MEMORY||16777216;if(C)it=l.wasmMemory,jt=l.buffer;else if(l.wasmMemory)it=l.wasmMemory;else if(it=new WebAssembly.Memory({initial:Mr/65536,maximum:32768,shared:!0}),!(it.buffer instanceof SharedArrayBuffer))throw Y(\"requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag\"),w&&console.log(\"(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)\"),Error(\"bad memory\");it&&(jt=it.buffer),Mr=jt.byteLength,qe(jt);var zr,qn=[],Kn=[],Xr=[],Zi=!1;function Mo(){return nt}function Il(){if(l.preRun)for(typeof l.preRun==\"function\"&&(l.preRun=[l.preRun]);l.preRun.length;)Rd(l.preRun.shift());Md(qn)}function Tr(){Zi=!0,!C&&Md(Kn)}function Ji(){if(!C){if(l.postRun)for(typeof l.postRun==\"function\"&&(l.postRun=[l.postRun]);l.postRun.length;)Gk(l.postRun.shift());Md(Xr)}}function Rd(L){qn.unshift(L)}function Fd(L){Kn.unshift(L)}function Gk(L){Xr.unshift(L)}var Sl=0,Xc=null,Qi=null;function Wk(L){Sl++,l.monitorRunDependencies&&l.monitorRunDependencies(Sl)}function Uk(L){if(Sl--,l.monitorRunDependencies&&l.monitorRunDependencies(Sl),Sl==0&&(Xc!==null&&(clearInterval(Xc),Xc=null),Qi)){var U=Qi;Qi=null,U()}}function Yc(L){C?postMessage({cmd:\"onAbort\",arg:L}):l.onAbort&&l.onAbort(L),L=\"Aborted(\"+L+\")\",Y(L),ht=!0,bt=1,L+=\". Build with -sASSERTIONS for more info.\";var U=new WebAssembly.RuntimeError(L);throw p(U),U}var gI=\"data:application/octet-stream;base64,\";function pg(L){return L.startsWith(gI)}function Od(L){return L.startsWith(\"file://\")}var en;en=\"tfjs-backend-wasm-threaded-simd.wasm\",pg(en)||(en=_(en));function mg(L){try{if(L==en&&at)return new Uint8Array(at);if(F)return F(L);throw\"both async and sync fetching of the wasm failed\"}catch(U){Yc(U)}}function xI(){if(!at&&(x||b)){if(typeof fetch==\"function\"&&!Od(en))return fetch(en,{credentials:\"same-origin\"}).then(function(L){if(!L.ok)throw\"failed to load wasm binary file at '\"+en+\"'\";return L.arrayBuffer()}).catch(function(){return mg(en)});if($)return new Promise(function(L,U){$(en,function(ut){L(new Uint8Array(ut))},U)})}return Promise.resolve().then(function(){return mg(en)})}function yI(){var L={env:vg,wasi_snapshot_preview1:vg};function U(Gt,Xt){var hr=Gt.exports;if(l.asm=hr,EI(l.asm._emscripten_tls_init),zr=l.asm.__indirect_function_table,Fd(l.asm.__wasm_call_ctors),dt=Xt,!C){var Xn=Kt.unusedWorkers.length;Kt.unusedWorkers.forEach(function(ea){Kt.loadWasmModuleToWorker(ea,function(){--Xn||Uk(\"wasm-instantiate\")})})}}C||Wk(\"wasm-instantiate\");function ut(Gt){U(Gt.instance,Gt.module)}function xt(Gt){return xI().then(function(Xt){return WebAssembly.instantiate(Xt,L)}).then(function(Xt){return Xt}).then(Gt,function(Xt){Y(\"failed to asynchronously prepare wasm: \"+Xt),Yc(Xt)})}function Dt(){return!at&&typeof WebAssembly.instantiateStreaming==\"function\"&&!pg(en)&&!Od(en)&&!w&&typeof fetch==\"function\"?fetch(en,{credentials:\"same-origin\"}).then(function(Gt){var Xt=WebAssembly.instantiateStreaming(Gt,L);return Xt.then(ut,function(hr){return Y(\"wasm streaming compile failed: \"+hr),Y(\"falling back to ArrayBuffer instantiation\"),xt(ut)})}):xt(ut)}if(l.instantiateWasm)try{var ie=l.instantiateWasm(L,U);return ie}catch(Gt){Y(\"Module.instantiateWasm callback failed with error: \"+Gt),p(Gt)}return Dt().catch(p),{}}var bI,Hk,wI={};function Ru(L){this.name=\"ExitStatus\",this.message=\"Program terminated with exit(\"+L+\")\",this.status=L}function CI(L){var U=Kt.pthreads[L];delete Kt.pthreads[L],U.terminate(),YS(L),Kt.runningWorkers.splice(Kt.runningWorkers.indexOf(U),1),U.pthread_ptr=0}function II(L){var U=Kt.pthreads[L];U.postMessage({cmd:\"cancel\"})}function Pd(L){var U=Kt.pthreads[L];Et(U),Kt.returnWorkerToPool(U)}function fg(L){var U=Kt.getNewWorker();if(!U)return 6;Kt.runningWorkers.push(U),Kt.pthreads[L.pthread_ptr]=U,U.pthread_ptr=L.pthread_ptr;var ut={cmd:\"run\",start_routine:L.startRoutine,arg:L.arg,pthread_ptr:L.pthread_ptr};return U.runPthread=()=>{ut.time=performance.now(),U.postMessage(ut,L.transferList)},U.loaded&&(U.runPthread(),delete U.runPthread),0}var dg={varargs:void 0,get:function(){dg.varargs+=4;var L=s()[dg.varargs-4>>2];return L},getStr:function(L){var U=Zt(L);return U}};function Ld(L){if(C)return vl(1,1,L);bt=L,Mo()||(Kt.terminateAllThreads(),l.onExit&&l.onExit(L),ht=!0),g(L,new Ru(L))}function qk(L,U){if(bt=L,!U&&C)throw gg(L),\"unwind\";Ld(L)}var hg=qk;function SI(L){if(L instanceof Ru||L==\"unwind\")return bt;g(1,L)}var Kt={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){C?Kt.initWorker():Kt.initMainThread()},initMainThread:function(){for(var L=8;L--;)Kt.allocateUnusedWorker()},initWorker:function(){nt=!1},setExitStatus:function(L){bt=L},terminateAllThreads:function(){for(var L of Object.values(Kt.pthreads))Kt.returnWorkerToPool(L);for(var L of Kt.unusedWorkers)L.terminate();Kt.unusedWorkers=[]},returnWorkerToPool:function(L){var U=L.pthread_ptr;delete Kt.pthreads[U],Kt.unusedWorkers.push(L),Kt.runningWorkers.splice(Kt.runningWorkers.indexOf(L),1),L.pthread_ptr=0,YS(U)},receiveObjectTransfer:function(L){},threadInitTLS:function(){Kt.tlsInitFunctions.forEach(L=>L())},loadWasmModuleToWorker:function(L,U){L.onmessage=ut=>{var xt=ut.data,Dt=xt.cmd;if(L.pthread_ptr&&(Kt.currentProxiedOperationCallerThread=L.pthread_ptr),xt.targetThread&&xt.targetThread!=Ag()){var ie=Kt.pthreads[xt.targetThread];ie?ie.postMessage(xt,xt.transferList):Y('Internal error! Worker sent a message \"'+Dt+'\" to target pthread '+xt.targetThread+\", but that thread no longer exists!\"),Kt.currentProxiedOperationCallerThread=void 0;return}Dt===\"processProxyingQueue\"?zd(xt.queue):Dt===\"spawnThread\"?fg(xt):Dt===\"cleanupThread\"?Pd(xt.thread):Dt===\"killThread\"?CI(xt.thread):Dt===\"cancelThread\"?II(xt.thread):Dt===\"loaded\"?(L.loaded=!0,U&&U(L),L.runPthread&&(L.runPthread(),delete L.runPthread)):Dt===\"print\"?j(\"Thread \"+xt.threadId+\": \"+xt.text):Dt===\"printErr\"?Y(\"Thread \"+xt.threadId+\": \"+xt.text):Dt===\"alert\"?alert(\"Thread \"+xt.threadId+\": \"+xt.text):xt.target===\"setimmediate\"?L.postMessage(xt):Dt===\"onAbort\"?l.onAbort&&l.onAbort(xt.arg):Dt&&Y(\"worker sent an unknown command \"+Dt),Kt.currentProxiedOperationCallerThread=void 0},L.onerror=ut=>{var xt=\"worker sent an error!\";throw Y(xt+\" \"+ut.filename+\":\"+ut.lineno+\": \"+ut.message),ut},w&&(L.on(\"message\",function(ut){L.onmessage({data:ut})}),L.on(\"error\",function(ut){L.onerror(ut)}),L.on(\"detachedExit\",function(){})),L.postMessage({cmd:\"load\",urlOrBlob:l.mainScriptUrlOrBlob||r,wasmMemory:it,wasmModule:dt})},allocateUnusedWorker:function(){var L=_(\"tfjs-backend-wasm-threaded-simd.worker.js\");Kt.unusedWorkers.push(new Worker(L))},getNewWorker:function(){return Kt.unusedWorkers.length==0&&(Kt.allocateUnusedWorker(),Kt.loadWasmModuleToWorker(Kt.unusedWorkers[0])),Kt.unusedWorkers.pop()}};l.PThread=Kt;function Md(L){for(;L.length>0;)L.shift()(l)}function vI(L){var U=ZS(),ut=L();return $g(U),ut}function Kk(L){return L}function jk(L){var U=/\\b_Z[\\w\\d_]+/g;return L.replace(U,function(ut){var xt=ut;return ut===xt?ut:xt+\" [\"+ut+\"]\"})}function NI(){var L=Ag(),U=s()[L+44>>2],ut=s()[L+48>>2],xt=U-ut;e1(U,xt),$g(U)}l.establishStackSpace=NI;function gg(L){if(C)return vl(2,0,L);try{hg(L)}catch(U){SI(U)}}var Zc=[];function TI(L){var U=Zc[L];return U||(L>=Zc.length&&(Zc.length=L+1),Zc[L]=U=zr.get(L)),U}function kI(L,U){var ut=TI(L)(U);Mo()?Kt.setExitStatus(ut):t1(ut)}l.invokeEntryPoint=kI;function Xk(){var L=new Error;if(!L.stack){try{throw new Error}catch(U){L=U}if(!L.stack)return\"(no stack trace available)\"}return L.stack.toString()}function EI(L){Kt.tlsInitFunctions.push(L)}function _I(L,U){e().set(L,U)}function AI(L){Zk(L,!b,1,!x),Kt.threadInitTLS()}function $I(L){C?postMessage({cmd:\"cleanupThread\",thread:L}):Pd(L)}function xg(L,U,ut,xt){return C?vl(3,1,L,U,ut,xt):yg(L,U,ut,xt)}function yg(L,U,ut,xt){if(typeof SharedArrayBuffer==\"undefined\")return Y(\"Current environment does not support SharedArrayBuffer, pthreads are not available!\"),6;var Dt=[],ie=0;if(C&&(Dt.length===0||ie))return xg(L,U,ut,xt);if(ie)return ie;var Gt={startRoutine:ut,pthread_ptr:L,arg:xt,transferList:Dt};return C?(Gt.cmd=\"spawnThread\",postMessage(Gt,Dt),0):fg(Gt)}function DI(){return 2097152}var RI=!0;function FI(){return RI}function zd(L){Atomics.store(s(),L>>2,1),Ag()&&Qk(L),Atomics.compareExchange(s(),L>>2,1,0)}l.executeNotifiedProxyingQueue=zd;function OI(L,U,ut,xt){if(L==U)setTimeout(()=>zd(xt));else if(C)postMessage({targetThread:L,cmd:\"processProxyingQueue\",queue:xt});else{var Dt=Kt.pthreads[L];if(!Dt)return;Dt.postMessage({cmd:\"processProxyingQueue\",queue:xt})}return 1}function PI(L,U,ut){return-1}function LI(){Yc(\"\")}function Fu(L){Fu.shown||(Fu.shown={}),Fu.shown[L]||(Fu.shown[L]=1,w&&(L=\"warning: \"+L),Y(L))}function MI(){w||b||Fu(\"Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread\")}function zI(){return Date.now()}function bg(){return 2147483648}function BI(){return bg()}var Jc;w?Jc=()=>{var L=process.hrtime();return L[0]*1e3+L[1]/1e6}:C?Jc=()=>performance.now()-l.__performance_now_clock_drift:Jc=()=>performance.now();function VI(L,U,ut){n().copyWithin(L,U,U+ut)}function GI(){return w?GW().cpus().length:navigator.hardwareConcurrency}function vl(L,U){var ut=arguments.length-2,xt=arguments;return vI(()=>{for(var Dt=ut,ie=Dg(Dt*8),Gt=ie>>3,Xt=0;Xt<ut;Xt++){var hr=xt[2+Xt];u()[Gt+Xt]=hr}return Jk(L,Dt,ie,U)})}var Bd=[];function WI(L,U,ut){Bd.length=U;for(var xt=ut>>3,Dt=0;Dt<U;Dt++)Bd[Dt]=u()[xt+Dt];var ie=L<0,Gt=ie?wI[-L-1]:ZI[L];return Gt.apply(null,Bd)}function UI(L){try{return it.grow(L-jt.byteLength+65535>>>16),qe(it.buffer),1}catch(U){}}function HI(L){var U=n().length;if(L=L>>>0,L<=U)return!1;var ut=bg();if(L>ut)return!1;let xt=(hr,Xn)=>hr+(Xn-hr%Xn)%Xn;for(var Dt=1;Dt<=4;Dt*=2){var ie=U*(1+.2/Dt);ie=Math.min(ie,L+100663296);var Gt=Math.min(ut,xt(Math.max(L,ie),65536)),Xt=UI(Gt);if(Xt)return!0}return!1}function qI(){throw\"unwind\"}function wg(L){return C?vl(4,1,L):52}function Cg(L,U,ut,xt,Dt){return C?vl(5,1,L,U,ut,xt,Dt):70}var KI=[null,[],[]];function jI(L,U){var ut=KI[L];U===0||U===10?((L===1?j:Y)(Vt(ut,0)),ut.length=0):ut.push(U)}function Ig(L,U,ut,xt){if(C)return vl(6,1,L,U,ut,xt);for(var Dt=0,ie=0;ie<ut;ie++){var Gt=i()[U>>2],Xt=i()[U+4>>2];U+=8;for(var hr=0;hr<Xt;hr++)jI(L,n()[Gt+hr]);Dt+=Xt}return i()[xt>>2]=Dt,0}function Sg(L){var U=l[\"_\"+L];return U}function XI(L,U,ut,xt,Dt){var ie={string:An=>{var rp=0;if(An!=null&&An!==0){var o1=(An.length<<2)+1;rp=Dg(o1),he(An,rp,o1)}return rp},array:An=>{var rp=Dg(An.length);return _I(An,rp),rp}};function Gt(An){return U===\"string\"?Zt(An):U===\"boolean\"?Boolean(An):An}var Xt=Sg(L),hr=[],Xn=0;if(xt)for(var ea=0;ea<xt.length;ea++){var n1=ie[ut[ea]];n1?(Xn===0&&(Xn=ZS()),hr[ea]=n1(xt[ea])):hr[ea]=xt[ea]}var JS=Xt.apply(null,hr);function $U(An){return Xn!==0&&$g(Xn),Gt(An)}return JS=$U(JS),JS}function YI(L,U,ut,xt){ut=ut||[];var Dt=ut.every(Gt=>Gt===\"number\"||Gt===\"boolean\"),ie=U!==\"string\";return ie&&Dt&&!xt?Sg(L):function(){return XI(L,U,ut,arguments,xt)}}Kt.init();var ZI=[null,Ld,gg,xg,wg,Cg,Ig],vg={__emscripten_init_main_thread_js:AI,__emscripten_thread_cleanup:$I,__pthread_create_js:yg,_emscripten_default_pthread_stack_size:DI,_emscripten_get_now_is_monotonic:FI,_emscripten_notify_task_queue:OI,_emscripten_set_offscreencanvas_size:PI,abort:LI,emscripten_check_blocking_allowed:MI,emscripten_date_now:zI,emscripten_get_heap_max:BI,emscripten_get_now:Jc,emscripten_memcpy_big:VI,emscripten_num_logical_cores:GI,emscripten_receive_on_main_thread_js:WI,emscripten_resize_heap:HI,emscripten_unwind_to_js_event_loop:qI,exit:hg,fd_close:wg,fd_seek:Cg,fd_write:Ig,memory:it||l.wasmMemory},Yk=yI(),JI=l.___wasm_call_ctors=function(){return(JI=l.___wasm_call_ctors=l.asm.__wasm_call_ctors).apply(null,arguments)},QI=l._init=function(){return(QI=l._init=l.asm.init).apply(null,arguments)},tS=l._init_with_threads_count=function(){return(tS=l._init_with_threads_count=l.asm.init_with_threads_count).apply(null,arguments)},eS=l._get_threads_count=function(){return(eS=l._get_threads_count=l.asm.get_threads_count).apply(null,arguments)},rS=l._register_tensor=function(){return(rS=l._register_tensor=l.asm.register_tensor).apply(null,arguments)},nS=l._dispose_data=function(){return(nS=l._dispose_data=l.asm.dispose_data).apply(null,arguments)},oS=l._dispose=function(){return(oS=l._dispose=l.asm.dispose).apply(null,arguments)},sS=l._Abs=function(){return(sS=l._Abs=l.asm.Abs).apply(null,arguments)},iS=l._Add=function(){return(iS=l._Add=l.asm.Add).apply(null,arguments)},aS=l._AddN=function(){return(aS=l._AddN=l.asm.AddN).apply(null,arguments)},lS=l._All=function(){return(lS=l._All=l.asm.All).apply(null,arguments)},uS=l._Any=function(){return(uS=l._Any=l.asm.Any).apply(null,arguments)},cS=l._ArgMax=function(){return(cS=l._ArgMax=l.asm.ArgMax).apply(null,arguments)},pS=l._AvgPool=function(){return(pS=l._AvgPool=l.asm.AvgPool).apply(null,arguments)},mS=l._BatchMatMul=function(){return(mS=l._BatchMatMul=l.asm.BatchMatMul).apply(null,arguments)},fS=l._Ceil=function(){return(fS=l._Ceil=l.asm.Ceil).apply(null,arguments)},dS=l._ClipByValue=function(){return(dS=l._ClipByValue=l.asm.ClipByValue).apply(null,arguments)},hS=l._Conv2D=function(){return(hS=l._Conv2D=l.asm.Conv2D).apply(null,arguments)},gS=l._Conv2DBackpropInput=function(){return(gS=l._Conv2DBackpropInput=l.asm.Conv2DBackpropInput).apply(null,arguments)},xS=l._Cos=function(){return(xS=l._Cos=l.asm.Cos).apply(null,arguments)},yS=l._Cosh=function(){return(yS=l._Cosh=l.asm.Cosh).apply(null,arguments)},bS=l._CropAndResize=function(){return(bS=l._CropAndResize=l.asm.CropAndResize).apply(null,arguments)},wS=l._Cumprod=function(){return(wS=l._Cumprod=l.asm.Cumprod).apply(null,arguments)},CS=l._Cumsum=function(){return(CS=l._Cumsum=l.asm.Cumsum).apply(null,arguments)},IS=l._DepthToSpace=function(){return(IS=l._DepthToSpace=l.asm.DepthToSpace).apply(null,arguments)},SS=l._DepthwiseConv2dNative=function(){return(SS=l._DepthwiseConv2dNative=l.asm.DepthwiseConv2dNative).apply(null,arguments)},vS=l._Elu=function(){return(vS=l._Elu=l.asm.Elu).apply(null,arguments)},NS=l._Equal=function(){return(NS=l._Equal=l.asm.Equal).apply(null,arguments)},TS=l._Exp=function(){return(TS=l._Exp=l.asm.Exp).apply(null,arguments)},kS=l._FlipLeftRight=function(){return(kS=l._FlipLeftRight=l.asm.FlipLeftRight).apply(null,arguments)},ES=l._Floor=function(){return(ES=l._Floor=l.asm.Floor).apply(null,arguments)},_S=l._FloorDiv=function(){return(_S=l._FloorDiv=l.asm.FloorDiv).apply(null,arguments)},AS=l._FusedBatchNorm=function(){return(AS=l._FusedBatchNorm=l.asm.FusedBatchNorm).apply(null,arguments)},$S=l._FusedConv2D=function(){return($S=l._FusedConv2D=l.asm.FusedConv2D).apply(null,arguments)},DS=l._FusedDepthwiseConv2D=function(){return(DS=l._FusedDepthwiseConv2D=l.asm.FusedDepthwiseConv2D).apply(null,arguments)},RS=l._Gather=function(){return(RS=l._Gather=l.asm.Gather).apply(null,arguments)},FS=l._GatherNd=function(){return(FS=l._GatherNd=l.asm.GatherNd).apply(null,arguments)},OS=l._Greater=function(){return(OS=l._Greater=l.asm.Greater).apply(null,arguments)},PS=l._GreaterEqual=function(){return(PS=l._GreaterEqual=l.asm.GreaterEqual).apply(null,arguments)},LS=l._LeakyRelu=function(){return(LS=l._LeakyRelu=l.asm.LeakyRelu).apply(null,arguments)},MS=l._Less=function(){return(MS=l._Less=l.asm.Less).apply(null,arguments)},zS=l._LessEqual=function(){return(zS=l._LessEqual=l.asm.LessEqual).apply(null,arguments)},BS=l._Log=function(){return(BS=l._Log=l.asm.Log).apply(null,arguments)},VS=l._LogicalAnd=function(){return(VS=l._LogicalAnd=l.asm.LogicalAnd).apply(null,arguments)},GS=l._LogicalNot=function(){return(GS=l._LogicalNot=l.asm.LogicalNot).apply(null,arguments)},WS=l._LogicalOr=function(){return(WS=l._LogicalOr=l.asm.LogicalOr).apply(null,arguments)},US=l._LogicalXor=function(){return(US=l._LogicalXor=l.asm.LogicalXor).apply(null,arguments)},HS=l._Max=function(){return(HS=l._Max=l.asm.Max).apply(null,arguments)},Ng=l._MaxPool=function(){return(Ng=l._MaxPool=l.asm.MaxPool).apply(null,arguments)},Tg=l._Maximum=function(){return(Tg=l._Maximum=l.asm.Maximum).apply(null,arguments)},Vd=l._Mean=function(){return(Vd=l._Mean=l.asm.Mean).apply(null,arguments)},qS=l._Min=function(){return(qS=l._Min=l.asm.Min).apply(null,arguments)},KS=l._Minimum=function(){return(KS=l._Minimum=l.asm.Minimum).apply(null,arguments)},Qc=l._MirrorPad=function(){return(Qc=l._MirrorPad=l.asm.MirrorPad).apply(null,arguments)},kg=l._Multiply=function(){return(kg=l._Multiply=l.asm.Multiply).apply(null,arguments)},tp=l._Neg=function(){return(tp=l._Neg=l.asm.Neg).apply(null,arguments)},ep=l._NonMaxSuppressionV3=function(){return(ep=l._NonMaxSuppressionV3=l.asm.NonMaxSuppressionV3).apply(null,arguments)},jS=l._NonMaxSuppressionV4=function(){return(jS=l._NonMaxSuppressionV4=l.asm.NonMaxSuppressionV4).apply(null,arguments)},K=l._NonMaxSuppressionV5=function(){return(K=l._NonMaxSuppressionV5=l.asm.NonMaxSuppressionV5).apply(null,arguments)},lt=l._NotEqual=function(){return(lt=l._NotEqual=l.asm.NotEqual).apply(null,arguments)},_t=l._OneHot=function(){return(_t=l._OneHot=l.asm.OneHot).apply(null,arguments)},re=l._PadV2=function(){return(re=l._PadV2=l.asm.PadV2).apply(null,arguments)},Ke=l._Pow=function(){return(Ke=l._Pow=l.asm.Pow).apply(null,arguments)},je=l._Prelu=function(){return(je=l._Prelu=l.asm.Prelu).apply(null,arguments)},ee=l._Prod=function(){return(ee=l._Prod=l.asm.Prod).apply(null,arguments)},te=l._RealDiv=function(){return(te=l._RealDiv=l.asm.RealDiv).apply(null,arguments)},dr=l._Relu=function(){return(dr=l._Relu=l.asm.Relu).apply(null,arguments)},jn=l._Relu6=function(){return(jn=l._Relu6=l.asm.Relu6).apply(null,arguments)},ta=l._ResizeBilinear=function(){return(ta=l._ResizeBilinear=l.asm.ResizeBilinear).apply(null,arguments)},Eg=l._ResizeNearestNeighbor=function(){return(Eg=l._ResizeNearestNeighbor=l.asm.ResizeNearestNeighbor).apply(null,arguments)},Gd=l._Reverse=function(){return(Gd=l._Reverse=l.asm.Reverse).apply(null,arguments)},XS=l._RotateWithOffset=function(){return(XS=l._RotateWithOffset=l.asm.RotateWithOffset).apply(null,arguments)},rn=l._Round=function(){return(rn=l._Round=l.asm.Round).apply(null,arguments)},Nl=l._Rsqrt=function(){return(Nl=l._Rsqrt=l.asm.Rsqrt).apply(null,arguments)},_g=l._ScatterNd=function(){return(_g=l._ScatterNd=l.asm.ScatterNd).apply(null,arguments)},QW=l._SelectV2=function(){return(QW=l._SelectV2=l.asm.SelectV2).apply(null,arguments)},tU=l._Sigmoid=function(){return(tU=l._Sigmoid=l.asm.Sigmoid).apply(null,arguments)},eU=l._Sin=function(){return(eU=l._Sin=l.asm.Sin).apply(null,arguments)},rU=l._Softmax=function(){return(rU=l._Softmax=l.asm.Softmax).apply(null,arguments)},nU=l._SparseFillEmptyRows=function(){return(nU=l._SparseFillEmptyRows=l.asm.SparseFillEmptyRows).apply(null,arguments)},oU=l._SparseReshape=function(){return(oU=l._SparseReshape=l.asm.SparseReshape).apply(null,arguments)},sU=l._SparseSegmentReduction=function(){return(sU=l._SparseSegmentReduction=l.asm.SparseSegmentReduction).apply(null,arguments)},iU=l._Sqrt=function(){return(iU=l._Sqrt=l.asm.Sqrt).apply(null,arguments)},aU=l._Square=function(){return(aU=l._Square=l.asm.Square).apply(null,arguments)},lU=l._SquaredDifference=function(){return(lU=l._SquaredDifference=l.asm.SquaredDifference).apply(null,arguments)},uU=l._Step=function(){return(uU=l._Step=l.asm.Step).apply(null,arguments)},cU=l._StridedSlice=function(){return(cU=l._StridedSlice=l.asm.StridedSlice).apply(null,arguments)},pU=l._Sub=function(){return(pU=l._Sub=l.asm.Sub).apply(null,arguments)},mU=l._Sum=function(){return(mU=l._Sum=l.asm.Sum).apply(null,arguments)},fU=l._Tan=function(){return(fU=l._Tan=l.asm.Tan).apply(null,arguments)},dU=l._Tanh=function(){return(dU=l._Tanh=l.asm.Tanh).apply(null,arguments)},hU=l._Tile=function(){return(hU=l._Tile=l.asm.Tile).apply(null,arguments)},gU=l._TopK=function(){return(gU=l._TopK=l.asm.TopK).apply(null,arguments)},xU=l._Transform=function(){return(xU=l._Transform=l.asm.Transform).apply(null,arguments)},yU=l._Transpose=function(){return(yU=l._Transpose=l.asm.Transpose).apply(null,arguments)},bU=l.__FusedMatMul=function(){return(bU=l.__FusedMatMul=l.asm._FusedMatMul).apply(null,arguments)},wU=l._malloc=function(){return(wU=l._malloc=l.asm.malloc).apply(null,arguments)},CU=l._free=function(){return(CU=l._free=l.asm.free).apply(null,arguments)},IU=l.__emscripten_tls_init=function(){return(IU=l.__emscripten_tls_init=l.asm._emscripten_tls_init).apply(null,arguments)},Ag=l._pthread_self=function(){return(Ag=l._pthread_self=l.asm.pthread_self).apply(null,arguments)},SU=l.___errno_location=function(){return(SU=l.___errno_location=l.asm.__errno_location).apply(null,arguments)},Zk=l.__emscripten_thread_init=function(){return(Zk=l.__emscripten_thread_init=l.asm._emscripten_thread_init).apply(null,arguments)},vU=l.__emscripten_thread_crashed=function(){return(vU=l.__emscripten_thread_crashed=l.asm._emscripten_thread_crashed).apply(null,arguments)},NU=l._emscripten_main_thread_process_queued_calls=function(){return(NU=l._emscripten_main_thread_process_queued_calls=l.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},TU=l._emscripten_main_browser_thread_id=function(){return(TU=l._emscripten_main_browser_thread_id=l.asm.emscripten_main_browser_thread_id).apply(null,arguments)},Jk=l._emscripten_run_in_main_runtime_thread_js=function(){return(Jk=l._emscripten_run_in_main_runtime_thread_js=l.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},kU=l._emscripten_dispatch_to_thread_=function(){return(kU=l._emscripten_dispatch_to_thread_=l.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},Qk=l.__emscripten_proxy_execute_task_queue=function(){return(Qk=l.__emscripten_proxy_execute_task_queue=l.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)},YS=l.__emscripten_thread_free_data=function(){return(YS=l.__emscripten_thread_free_data=l.asm._emscripten_thread_free_data).apply(null,arguments)},t1=l.__emscripten_thread_exit=function(){return(t1=l.__emscripten_thread_exit=l.asm._emscripten_thread_exit).apply(null,arguments)},e1=l._emscripten_stack_set_limits=function(){return(e1=l._emscripten_stack_set_limits=l.asm.emscripten_stack_set_limits).apply(null,arguments)},ZS=l.stackSave=function(){return(ZS=l.stackSave=l.asm.stackSave).apply(null,arguments)},$g=l.stackRestore=function(){return($g=l.stackRestore=l.asm.stackRestore).apply(null,arguments)},Dg=l.stackAlloc=function(){return(Dg=l.stackAlloc=l.asm.stackAlloc).apply(null,arguments)},EU=l.dynCall_iijjiiii=function(){return(EU=l.dynCall_iijjiiii=l.asm.dynCall_iijjiiii).apply(null,arguments)},_U=l.dynCall_jiji=function(){return(_U=l.dynCall_jiji=l.asm.dynCall_jiji).apply(null,arguments)};l.keepRuntimeAlive=Mo,l.wasmMemory=it,l.cwrap=YI,l.ExitStatus=Ru,l.PThread=Kt;var Rg;Qi=function L(){Rg||r1(),Rg||(Qi=L)};function r1(L){if(L=L||d,Sl>0)return;if(C){c(l),Tr(),postMessage({cmd:\"loaded\"});return}if(Il(),Sl>0)return;function U(){Rg||(Rg=!0,l.calledRun=!0,!ht&&(Tr(),c(l),l.onRuntimeInitialized&&l.onRuntimeInitialized(),Ji()))}l.setStatus?(l.setStatus(\"Running...\"),setTimeout(function(){setTimeout(function(){l.setStatus(\"\")},1),U()},1)):U()}if(l.preInit)for(typeof l.preInit==\"function\"&&(l.preInit=[l.preInit]);l.preInit.length>0;)l.preInit.pop()();r1();var Fg;m&&(Fg={uncaughtException:process.listeners(\"uncaughtException\").filter(function(L){return!m.uncaughtException.indexOf(L)>-1}),unhandledRejection:process.listeners(\"unhandledRejection\").filter(function(L){return!m.unhandledRejection.indexOf(L)>-1})});var Og;if(typeof WasmBackendModule!=\"undefined\")Og=WasmBackendModule;else if(typeof t!=\"undefined\")Og=t;else throw new Error(\"Could not find wasm module in post.js\");if(Fg){var AU=Og._dispose;Og._dispose=function(){AU(),Fg.uncaughtException.forEach(function(L){process.removeListener(\"uncaughtException\",L)}),Fg.unhandledRejection.forEach(function(L){process.removeListener(\"unhandledRejection\",L)})}}return t.ready}})();typeof fI==\"object\"&&typeof Ok==\"object\"?Ok.exports=Fk:typeof define==\"function\"&&define.amd?define([],function(){return Fk}):typeof fI==\"object\"&&(fI.WasmBackendModuleThreadedSimd=Fk)});var HW=gr((G5e,UW)=>{UW.exports.wasmWorkerContents=`\"use strict\";var Module={};var ENVIRONMENT_IS_NODE=typeof process==\"object\"&&typeof process.versions==\"object\"&&typeof process.versions.node==\"string\";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require(\"worker_threads\");var parentPort=nodeWorkerThreads.parentPort;parentPort.on(\"message\",data=>onmessage({data:data}));var fs=require(\"fs\");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,\"utf8\"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(\" \");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+\"\n\");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(\" \");postMessage({cmd:\"alert\",text:text,threadId:Module[\"_pthread_self\"]()})}var err=threadPrintErr;self.alert=threadAlert;Module[\"instantiateWasm\"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module[\"wasmModule\"],info);receiveInstance(instance);Module[\"wasmModule\"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd===\"load\"){Module[\"wasmModule\"]=e.data.wasmModule;Module[\"wasmMemory\"]=e.data.wasmMemory;Module[\"buffer\"]=Module[\"wasmMemory\"].buffer;Module[\"ENVIRONMENT_IS_PTHREAD\"]=true;if(typeof e.data.urlOrBlob==\"string\"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd===\"run\"){Module[\"__performance_now_clock_drift\"]=performance.now()-e.data.time;Module[\"__emscripten_thread_init\"](e.data.pthread_ptr,0,0,1);Module[\"establishStackSpace\"]();Module[\"PThread\"].receiveObjectTransfer(e.data);Module[\"PThread\"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module[\"executeNotifiedProxyingQueue\"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module[\"invokeEntryPoint\"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!=\"unwind\"){if(ex instanceof Module[\"ExitStatus\"]){if(Module[\"keepRuntimeAlive\"]()){}else{Module[\"__emscripten_thread_exit\"](ex.status)}}else{throw ex}}}}else if(e.data.cmd===\"cancel\"){if(Module[\"_pthread_self\"]()){Module[\"__emscripten_thread_exit\"](-1)}}else if(e.data.target===\"setimmediate\"){}else if(e.data.cmd===\"processProxyingQueue\"){if(initializedJS){Module[\"executeNotifiedProxyingQueue\"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err(\"worker.js received unknown command \"+e.data.cmd);err(e.data)}}catch(ex){if(Module[\"__emscripten_thread_crashed\"]){Module[\"__emscripten_thread_crashed\"]()}throw ex}};`});var qW=gr((dI,Lk)=>{var Pk=(()=>{var r=typeof document!=\"undefined\"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!=\"undefined\"&&(r=r||__filename),function(t){t=t||{};var e=typeof t!=\"undefined\"?t:{},n,o;e.ready=new Promise(function(K,lt){n=K,o=lt});var s;typeof process!=\"undefined\"&&process.listeners&&(s={uncaughtException:process.listeners(\"uncaughtException\"),unhandledRejection:process.listeners(\"unhandledRejection\")});var i=Object.assign({},e),a=[],u=\"./this.program\",l=(K,lt)=>{throw lt},c=typeof window==\"object\",p=typeof importScripts==\"function\",m=typeof process==\"object\"&&typeof process.versions==\"object\"&&typeof process.versions.node==\"string\",f=\"\";function d(K){return e.locateFile?e.locateFile(K,f):f+K}var h,g,x,b;function w(K){if(K instanceof Xc)return;A(\"exiting due to exception: \"+K)}if(m){p?f=ig().dirname(f)+\"/\":f=__dirname+\"/\";var C,N;typeof Pg==\"function\"&&(C=Zb(),N=ig()),h=(K,lt)=>(K=N.normalize(K),C.readFileSync(K,lt?void 0:\"utf8\")),x=K=>{var lt=h(K,!0);return lt.buffer||(lt=new Uint8Array(lt)),lt},g=(K,lt,_t)=>{K=N.normalize(K),C.readFile(K,function(re,Ke){re?_t(re):lt(Ke.buffer)})},process.argv.length>1&&(u=process.argv[1].replace(/\\\\/g,\"/\")),a=process.argv.slice(2),process.on(\"uncaughtException\",function(K){if(!(K instanceof Xc))throw K}),process.on(\"unhandledRejection\",function(K){throw K}),l=(K,lt)=>{if(fe())throw process.exitCode=K,lt;w(lt),process.exit(K)},e.inspect=function(){return\"[Emscripten Module object]\"}}else(c||p)&&(p?f=self.location.href:typeof document!=\"undefined\"&&document.currentScript&&(f=document.currentScript.src),r&&(f=r),f.indexOf(\"blob:\")!==0?f=f.substr(0,f.replace(/[?#].*/,\"\").lastIndexOf(\"/\")+1):f=\"\",h=K=>{var lt=new XMLHttpRequest;return lt.open(\"GET\",K,!1),lt.send(null),lt.responseText},p&&(x=K=>{var lt=new XMLHttpRequest;return lt.open(\"GET\",K,!1),lt.responseType=\"arraybuffer\",lt.send(null),new Uint8Array(lt.response)}),g=(K,lt,_t)=>{var re=new XMLHttpRequest;re.open(\"GET\",K,!0),re.responseType=\"arraybuffer\",re.onload=()=>{if(re.status==200||re.status==0&&re.response){lt(re.response);return}_t()},re.onerror=_t,re.send(null)},b=K=>document.title=K);var _=e.print||console.log.bind(console),A=e.printErr||console.warn.bind(console);Object.assign(e,i),i=null,e.arguments&&(a=e.arguments),e.thisProgram&&(u=e.thisProgram),e.quit&&(l=e.quit);var $=4,F;e.wasmBinary&&(F=e.wasmBinary);var P=e.noExitRuntime||!0;typeof WebAssembly!=\"object\"&&Xr(\"no native wasm support detected\");var V,G=!1,W;function q(K,lt){K||Xr(lt)}var H=typeof TextDecoder!=\"undefined\"?new TextDecoder(\"utf8\"):void 0;function j(K,lt,_t){for(var re=lt+_t,Ke=lt;K[Ke]&&!(Ke>=re);)++Ke;if(Ke-lt>16&&K.buffer&&H)return H.decode(K.subarray(lt,Ke));for(var je=\"\";lt<Ke;){var ee=K[lt++];if(!(ee&128)){je+=String.fromCharCode(ee);continue}var te=K[lt++]&63;if((ee&224)==192){je+=String.fromCharCode((ee&31)<<6|te);continue}var dr=K[lt++]&63;if((ee&240)==224?ee=(ee&15)<<12|te<<6|dr:ee=(ee&7)<<18|te<<12|dr<<6|K[lt++]&63,ee<65536)je+=String.fromCharCode(ee);else{var jn=ee-65536;je+=String.fromCharCode(55296|jn>>10,56320|jn&1023)}}return je}function Y(K,lt){return K?j(at,K,lt):\"\"}function Z(K,lt,_t,re){if(!(re>0))return 0;for(var Ke=_t,je=_t+re-1,ee=0;ee<K.length;++ee){var te=K.charCodeAt(ee);if(te>=55296&&te<=57343){var dr=K.charCodeAt(++ee);te=65536+((te&1023)<<10)|dr&1023}if(te<=127){if(_t>=je)break;lt[_t++]=te}else if(te<=2047){if(_t+1>=je)break;lt[_t++]=192|te>>6,lt[_t++]=128|te&63}else if(te<=65535){if(_t+2>=je)break;lt[_t++]=224|te>>12,lt[_t++]=128|te>>6&63,lt[_t++]=128|te&63}else{if(_t+3>=je)break;lt[_t++]=240|te>>18,lt[_t++]=128|te>>12&63,lt[_t++]=128|te>>6&63,lt[_t++]=128|te&63}}return lt[_t]=0,_t-Ke}function et(K,lt,_t){return Z(K,at,lt,_t)}var rt,ot,at,nt,it,dt,ht,bt,Et;function At(K){rt=K,e.HEAP8=ot=new Int8Array(K),e.HEAP16=nt=new Int16Array(K),e.HEAP32=dt=new Int32Array(K),e.HEAPU8=at=new Uint8Array(K),e.HEAPU16=it=new Uint16Array(K),e.HEAPU32=ht=new Uint32Array(K),e.HEAPF32=bt=new Float32Array(K),e.HEAPF64=Et=new Float64Array(K)}var Vt=e.INITIAL_MEMORY||16777216,Zt,ce=[],he=[],jt=[],ke=!1;function fe(){return P}function Ae(){if(e.preRun)for(typeof e.preRun==\"function\"&&(e.preRun=[e.preRun]);e.preRun.length;)or(e.preRun.shift());Qi(ce)}function We(){ke=!0,Qi(he)}function _n(){if(e.postRun)for(typeof e.postRun==\"function\"&&(e.postRun=[e.postRun]);e.postRun.length;)Lr(e.postRun.shift());Qi(jt)}function or(K){ce.unshift(K)}function Hn(K){he.unshift(K)}function Lr(K){jt.unshift(K)}var qe=0,Mr=null,zr=null;function qn(K){qe++,e.monitorRunDependencies&&e.monitorRunDependencies(qe)}function Kn(K){if(qe--,e.monitorRunDependencies&&e.monitorRunDependencies(qe),qe==0&&(Mr!==null&&(clearInterval(Mr),Mr=null),zr)){var lt=zr;zr=null,lt()}}function Xr(K){e.onAbort&&e.onAbort(K),K=\"Aborted(\"+K+\")\",A(K),G=!0,W=1,K+=\". Build with -sASSERTIONS for more info.\";var lt=new WebAssembly.RuntimeError(K);throw o(lt),lt}var Zi=\"data:application/octet-stream;base64,\";function Mo(K){return K.startsWith(Zi)}function Il(K){return K.startsWith(\"file://\")}var Tr;Tr=\"tfjs-backend-wasm.wasm\",Mo(Tr)||(Tr=d(Tr));function Ji(K){try{if(K==Tr&&F)return new Uint8Array(F);if(x)return x(K);throw\"both async and sync fetching of the wasm failed\"}catch(lt){Xr(lt)}}function Rd(){if(!F&&(c||p)){if(typeof fetch==\"function\"&&!Il(Tr))return fetch(Tr,{credentials:\"same-origin\"}).then(function(K){if(!K.ok)throw\"failed to load wasm binary file at '\"+Tr+\"'\";return K.arrayBuffer()}).catch(function(){return Ji(Tr)});if(g)return new Promise(function(K,lt){g(Tr,function(_t){K(new Uint8Array(_t))},lt)})}return Promise.resolve().then(function(){return Ji(Tr)})}function Fd(){var K={env:Ld,wasi_snapshot_preview1:Ld};function lt(ee,te){var dr=ee.exports;e.asm=dr,V=e.asm.memory,At(V.buffer),Zt=e.asm.__indirect_function_table,Hn(e.asm.__wasm_call_ctors),Kn(\"wasm-instantiate\")}qn(\"wasm-instantiate\");function _t(ee){lt(ee.instance)}function re(ee){return Rd().then(function(te){return WebAssembly.instantiate(te,K)}).then(function(te){return te}).then(ee,function(te){A(\"failed to asynchronously prepare wasm: \"+te),Xr(te)})}function Ke(){return!F&&typeof WebAssembly.instantiateStreaming==\"function\"&&!Mo(Tr)&&!Il(Tr)&&!m&&typeof fetch==\"function\"?fetch(Tr,{credentials:\"same-origin\"}).then(function(ee){var te=WebAssembly.instantiateStreaming(ee,K);return te.then(_t,function(dr){return A(\"wasm streaming compile failed: \"+dr),A(\"falling back to ArrayBuffer instantiation\"),re(_t)})}):re(_t)}if(e.instantiateWasm)try{var je=e.instantiateWasm(K,lt);return je}catch(ee){A(\"Module.instantiateWasm callback failed with error: \"+ee),o(ee)}return Ke().catch(o),{}}var Gk,Sl;function Xc(K){this.name=\"ExitStatus\",this.message=\"Program terminated with exit(\"+K+\")\",this.status=K}function Qi(K){for(;K.length>0;)K.shift()(e)}function Wk(K){return K}function Uk(K){var lt=/\\b_Z[\\w\\d_]+/g;return K.replace(lt,function(_t){var re=_t;return _t===re?_t:re+\" [\"+_t+\"]\"})}function Yc(){var K=new Error;if(!K.stack){try{throw new Error}catch(lt){K=lt}if(!K.stack)return\"(no stack trace available)\"}return K.stack.toString()}function gI(K,lt){ot.set(K,lt)}function pg(){Xr(\"\")}function Od(){return 2147483648}function en(){return Od()}function mg(K,lt,_t){at.copyWithin(K,lt,lt+_t)}function xI(K){try{return V.grow(K-rt.byteLength+65535>>>16),At(V.buffer),1}catch(lt){}}function yI(K){var lt=at.length;K=K>>>0;var _t=Od();if(K>_t)return!1;let re=(dr,jn)=>dr+(jn-dr%jn)%jn;for(var Ke=1;Ke<=4;Ke*=2){var je=lt*(1+.2/Ke);je=Math.min(je,K+100663296);var ee=Math.min(_t,re(Math.max(K,je),65536)),te=xI(ee);if(te)return!0}return!1}var bI={varargs:void 0,get:function(){bI.varargs+=4;var K=dt[bI.varargs-4>>2];return K},getStr:function(K){var lt=Y(K);return lt}};function Hk(K){return 52}function wI(K,lt,_t,re,Ke){return 70}var Ru=[null,[],[]];function CI(K,lt){var _t=Ru[K];lt===0||lt===10?((K===1?_:A)(j(_t,0)),_t.length=0):_t.push(lt)}function II(K,lt,_t,re){for(var Ke=0,je=0;je<_t;je++){var ee=ht[lt>>2],te=ht[lt+4>>2];lt+=8;for(var dr=0;dr<te;dr++)CI(K,at[ee+dr]);Ke+=te}return ht[re>>2]=Ke,0}function Pd(K){var lt=e[\"_\"+K];return lt}function fg(K,lt,_t,re,Ke){var je={string:rn=>{var Nl=0;if(rn!=null&&rn!==0){var _g=(rn.length<<2)+1;Nl=Vd(_g),et(rn,Nl,_g)}return Nl},array:rn=>{var Nl=Vd(rn.length);return gI(rn,Nl),Nl}};function ee(rn){return lt===\"string\"?Y(rn):lt===\"boolean\"?Boolean(rn):rn}var te=Pd(K),dr=[],jn=0;if(re)for(var ta=0;ta<re.length;ta++){var Eg=je[_t[ta]];Eg?(jn===0&&(jn=Ng()),dr[ta]=Eg(re[ta])):dr[ta]=re[ta]}var Gd=te.apply(null,dr);function XS(rn){return jn!==0&&Tg(jn),ee(rn)}return Gd=XS(Gd),Gd}function dg(K,lt,_t,re){_t=_t||[];var Ke=_t.every(ee=>ee===\"number\"||ee===\"boolean\"),je=lt!==\"string\";return je&&Ke&&!re?Pd(K):function(){return fg(K,lt,_t,arguments,re)}}var Ld={abort:pg,emscripten_get_heap_max:en,emscripten_memcpy_big:mg,emscripten_resize_heap:yI,fd_close:Hk,fd_seek:wI,fd_write:II},qk=Fd(),hg=e.___wasm_call_ctors=function(){return(hg=e.___wasm_call_ctors=e.asm.__wasm_call_ctors).apply(null,arguments)},SI=e._init=function(){return(SI=e._init=e.asm.init).apply(null,arguments)},Kt=e._init_with_threads_count=function(){return(Kt=e._init_with_threads_count=e.asm.init_with_threads_count).apply(null,arguments)},Md=e._get_threads_count=function(){return(Md=e._get_threads_count=e.asm.get_threads_count).apply(null,arguments)},vI=e._register_tensor=function(){return(vI=e._register_tensor=e.asm.register_tensor).apply(null,arguments)},Kk=e._dispose_data=function(){return(Kk=e._dispose_data=e.asm.dispose_data).apply(null,arguments)},jk=e._dispose=function(){return(jk=e._dispose=e.asm.dispose).apply(null,arguments)},NI=e._Abs=function(){return(NI=e._Abs=e.asm.Abs).apply(null,arguments)},gg=e._Add=function(){return(gg=e._Add=e.asm.Add).apply(null,arguments)},Zc=e._AddN=function(){return(Zc=e._AddN=e.asm.AddN).apply(null,arguments)},TI=e._All=function(){return(TI=e._All=e.asm.All).apply(null,arguments)},kI=e._Any=function(){return(kI=e._Any=e.asm.Any).apply(null,arguments)},Xk=e._ArgMax=function(){return(Xk=e._ArgMax=e.asm.ArgMax).apply(null,arguments)},EI=e._AvgPool=function(){return(EI=e._AvgPool=e.asm.AvgPool).apply(null,arguments)},_I=e._BatchMatMul=function(){return(_I=e._BatchMatMul=e.asm.BatchMatMul).apply(null,arguments)},AI=e._Ceil=function(){return(AI=e._Ceil=e.asm.Ceil).apply(null,arguments)},$I=e._ClipByValue=function(){return($I=e._ClipByValue=e.asm.ClipByValue).apply(null,arguments)},xg=e._Conv2D=function(){return(xg=e._Conv2D=e.asm.Conv2D).apply(null,arguments)},yg=e._Conv2DBackpropInput=function(){return(yg=e._Conv2DBackpropInput=e.asm.Conv2DBackpropInput).apply(null,arguments)},DI=e._Cos=function(){return(DI=e._Cos=e.asm.Cos).apply(null,arguments)},RI=e._Cosh=function(){return(RI=e._Cosh=e.asm.Cosh).apply(null,arguments)},FI=e._CropAndResize=function(){return(FI=e._CropAndResize=e.asm.CropAndResize).apply(null,arguments)},zd=e._Cumprod=function(){return(zd=e._Cumprod=e.asm.Cumprod).apply(null,arguments)},OI=e._Cumsum=function(){return(OI=e._Cumsum=e.asm.Cumsum).apply(null,arguments)},PI=e._DepthToSpace=function(){return(PI=e._DepthToSpace=e.asm.DepthToSpace).apply(null,arguments)},LI=e._DepthwiseConv2dNative=function(){return(LI=e._DepthwiseConv2dNative=e.asm.DepthwiseConv2dNative).apply(null,arguments)},Fu=e._Elu=function(){return(Fu=e._Elu=e.asm.Elu).apply(null,arguments)},MI=e._Equal=function(){return(MI=e._Equal=e.asm.Equal).apply(null,arguments)},zI=e._Exp=function(){return(zI=e._Exp=e.asm.Exp).apply(null,arguments)},bg=e._FlipLeftRight=function(){return(bg=e._FlipLeftRight=e.asm.FlipLeftRight).apply(null,arguments)},BI=e._Floor=function(){return(BI=e._Floor=e.asm.Floor).apply(null,arguments)},Jc=e._FloorDiv=function(){return(Jc=e._FloorDiv=e.asm.FloorDiv).apply(null,arguments)},VI=e._FusedBatchNorm=function(){return(VI=e._FusedBatchNorm=e.asm.FusedBatchNorm).apply(null,arguments)},GI=e._FusedConv2D=function(){return(GI=e._FusedConv2D=e.asm.FusedConv2D).apply(null,arguments)},vl=e._FusedDepthwiseConv2D=function(){return(vl=e._FusedDepthwiseConv2D=e.asm.FusedDepthwiseConv2D).apply(null,arguments)},Bd=e._Gather=function(){return(Bd=e._Gather=e.asm.Gather).apply(null,arguments)},WI=e._GatherNd=function(){return(WI=e._GatherNd=e.asm.GatherNd).apply(null,arguments)},UI=e._Greater=function(){return(UI=e._Greater=e.asm.Greater).apply(null,arguments)},HI=e._GreaterEqual=function(){return(HI=e._GreaterEqual=e.asm.GreaterEqual).apply(null,arguments)},qI=e._LeakyRelu=function(){return(qI=e._LeakyRelu=e.asm.LeakyRelu).apply(null,arguments)},wg=e._Less=function(){return(wg=e._Less=e.asm.Less).apply(null,arguments)},Cg=e._LessEqual=function(){return(Cg=e._LessEqual=e.asm.LessEqual).apply(null,arguments)},KI=e._Log=function(){return(KI=e._Log=e.asm.Log).apply(null,arguments)},jI=e._LogicalAnd=function(){return(jI=e._LogicalAnd=e.asm.LogicalAnd).apply(null,arguments)},Ig=e._LogicalNot=function(){return(Ig=e._LogicalNot=e.asm.LogicalNot).apply(null,arguments)},Sg=e._LogicalOr=function(){return(Sg=e._LogicalOr=e.asm.LogicalOr).apply(null,arguments)},XI=e._LogicalXor=function(){return(XI=e._LogicalXor=e.asm.LogicalXor).apply(null,arguments)},YI=e._Max=function(){return(YI=e._Max=e.asm.Max).apply(null,arguments)},ZI=e._MaxPool=function(){return(ZI=e._MaxPool=e.asm.MaxPool).apply(null,arguments)},vg=e._Maximum=function(){return(vg=e._Maximum=e.asm.Maximum).apply(null,arguments)},Yk=e._Mean=function(){return(Yk=e._Mean=e.asm.Mean).apply(null,arguments)},JI=e._Min=function(){return(JI=e._Min=e.asm.Min).apply(null,arguments)},QI=e._Minimum=function(){return(QI=e._Minimum=e.asm.Minimum).apply(null,arguments)},tS=e._MirrorPad=function(){return(tS=e._MirrorPad=e.asm.MirrorPad).apply(null,arguments)},eS=e._Multiply=function(){return(eS=e._Multiply=e.asm.Multiply).apply(null,arguments)},rS=e._Neg=function(){return(rS=e._Neg=e.asm.Neg).apply(null,arguments)},nS=e._NonMaxSuppressionV3=function(){return(nS=e._NonMaxSuppressionV3=e.asm.NonMaxSuppressionV3).apply(null,arguments)},oS=e._NonMaxSuppressionV4=function(){return(oS=e._NonMaxSuppressionV4=e.asm.NonMaxSuppressionV4).apply(null,arguments)},sS=e._NonMaxSuppressionV5=function(){return(sS=e._NonMaxSuppressionV5=e.asm.NonMaxSuppressionV5).apply(null,arguments)},iS=e._NotEqual=function(){return(iS=e._NotEqual=e.asm.NotEqual).apply(null,arguments)},aS=e._OneHot=function(){return(aS=e._OneHot=e.asm.OneHot).apply(null,arguments)},lS=e._PadV2=function(){return(lS=e._PadV2=e.asm.PadV2).apply(null,arguments)},uS=e._Pow=function(){return(uS=e._Pow=e.asm.Pow).apply(null,arguments)},cS=e._Prelu=function(){return(cS=e._Prelu=e.asm.Prelu).apply(null,arguments)},pS=e._Prod=function(){return(pS=e._Prod=e.asm.Prod).apply(null,arguments)},mS=e._RealDiv=function(){return(mS=e._RealDiv=e.asm.RealDiv).apply(null,arguments)},fS=e._Relu=function(){return(fS=e._Relu=e.asm.Relu).apply(null,arguments)},dS=e._Relu6=function(){return(dS=e._Relu6=e.asm.Relu6).apply(null,arguments)},hS=e._ResizeBilinear=function(){return(hS=e._ResizeBilinear=e.asm.ResizeBilinear).apply(null,arguments)},gS=e._ResizeNearestNeighbor=function(){return(gS=e._ResizeNearestNeighbor=e.asm.ResizeNearestNeighbor).apply(null,arguments)},xS=e._Reverse=function(){return(xS=e._Reverse=e.asm.Reverse).apply(null,arguments)},yS=e._RotateWithOffset=function(){return(yS=e._RotateWithOffset=e.asm.RotateWithOffset).apply(null,arguments)},bS=e._Round=function(){return(bS=e._Round=e.asm.Round).apply(null,arguments)},wS=e._Rsqrt=function(){return(wS=e._Rsqrt=e.asm.Rsqrt).apply(null,arguments)},CS=e._ScatterNd=function(){return(CS=e._ScatterNd=e.asm.ScatterNd).apply(null,arguments)},IS=e._SelectV2=function(){return(IS=e._SelectV2=e.asm.SelectV2).apply(null,arguments)},SS=e._Sigmoid=function(){return(SS=e._Sigmoid=e.asm.Sigmoid).apply(null,arguments)},vS=e._Sin=function(){return(vS=e._Sin=e.asm.Sin).apply(null,arguments)},NS=e._Softmax=function(){return(NS=e._Softmax=e.asm.Softmax).apply(null,arguments)},TS=e._SparseFillEmptyRows=function(){return(TS=e._SparseFillEmptyRows=e.asm.SparseFillEmptyRows).apply(null,arguments)},kS=e._SparseReshape=function(){return(kS=e._SparseReshape=e.asm.SparseReshape).apply(null,arguments)},ES=e._SparseSegmentReduction=function(){return(ES=e._SparseSegmentReduction=e.asm.SparseSegmentReduction).apply(null,arguments)},_S=e._Sqrt=function(){return(_S=e._Sqrt=e.asm.Sqrt).apply(null,arguments)},AS=e._Square=function(){return(AS=e._Square=e.asm.Square).apply(null,arguments)},$S=e._SquaredDifference=function(){return($S=e._SquaredDifference=e.asm.SquaredDifference).apply(null,arguments)},DS=e._Step=function(){return(DS=e._Step=e.asm.Step).apply(null,arguments)},RS=e._StridedSlice=function(){return(RS=e._StridedSlice=e.asm.StridedSlice).apply(null,arguments)},FS=e._Sub=function(){return(FS=e._Sub=e.asm.Sub).apply(null,arguments)},OS=e._Sum=function(){return(OS=e._Sum=e.asm.Sum).apply(null,arguments)},PS=e._Tan=function(){return(PS=e._Tan=e.asm.Tan).apply(null,arguments)},LS=e._Tanh=function(){return(LS=e._Tanh=e.asm.Tanh).apply(null,arguments)},MS=e._Tile=function(){return(MS=e._Tile=e.asm.Tile).apply(null,arguments)},zS=e._TopK=function(){return(zS=e._TopK=e.asm.TopK).apply(null,arguments)},BS=e._Transform=function(){return(BS=e._Transform=e.asm.Transform).apply(null,arguments)},VS=e._Transpose=function(){return(VS=e._Transpose=e.asm.Transpose).apply(null,arguments)},GS=e.__FusedMatMul=function(){return(GS=e.__FusedMatMul=e.asm._FusedMatMul).apply(null,arguments)},WS=e._malloc=function(){return(WS=e._malloc=e.asm.malloc).apply(null,arguments)},US=e._free=function(){return(US=e._free=e.asm.free).apply(null,arguments)},HS=e.___errno_location=function(){return(HS=e.___errno_location=e.asm.__errno_location).apply(null,arguments)},Ng=e.stackSave=function(){return(Ng=e.stackSave=e.asm.stackSave).apply(null,arguments)},Tg=e.stackRestore=function(){return(Tg=e.stackRestore=e.asm.stackRestore).apply(null,arguments)},Vd=e.stackAlloc=function(){return(Vd=e.stackAlloc=e.asm.stackAlloc).apply(null,arguments)},qS=e.dynCall_iijjiiii=function(){return(qS=e.dynCall_iijjiiii=e.asm.dynCall_iijjiiii).apply(null,arguments)},KS=e.dynCall_jiji=function(){return(KS=e.dynCall_jiji=e.asm.dynCall_jiji).apply(null,arguments)};e.cwrap=dg;var Qc;zr=function K(){Qc||kg(),Qc||(zr=K)};function kg(K){if(K=K||a,qe>0||(Ae(),qe>0))return;function lt(){Qc||(Qc=!0,e.calledRun=!0,!G&&(We(),n(e),e.onRuntimeInitialized&&e.onRuntimeInitialized(),_n()))}e.setStatus?(e.setStatus(\"Running...\"),setTimeout(function(){setTimeout(function(){e.setStatus(\"\")},1),lt()},1)):lt()}if(e.preInit)for(typeof e.preInit==\"function\"&&(e.preInit=[e.preInit]);e.preInit.length>0;)e.preInit.pop()();kg();var tp;s&&(tp={uncaughtException:process.listeners(\"uncaughtException\").filter(function(K){return!s.uncaughtException.indexOf(K)>-1}),unhandledRejection:process.listeners(\"unhandledRejection\").filter(function(K){return!s.unhandledRejection.indexOf(K)>-1})});var ep;if(typeof t!=\"undefined\")ep=t;else if(typeof WasmBackendModuleThreadedSimd!=\"undefined\")ep=WasmBackendModuleThreadedSimd;else throw new Error(\"Could not find wasm module in post.js\");if(tp){var jS=ep._dispose;ep._dispose=function(){jS(),tp.uncaughtException.forEach(function(K){process.removeListener(\"uncaughtException\",K)}),tp.unhandledRejection.forEach(function(K){process.removeListener(\"unhandledRejection\",K)})}}return t.ready}})();typeof dI==\"object\"&&typeof Lk==\"object\"?Lk.exports=Pk:typeof define==\"function\"&&define.amd?define([],function(){return Pk}):typeof dI==\"object\"&&(dI.WasmBackendModule=Pk)});var ra=class{constructor(t,e){this.backend=t,this.dataMover=e,this.data=new WeakMap,this.dataIdsCount=0}get(t){return this.data.has(t)||this.dataMover.moveData(this.backend,t),this.data.get(t)}set(t,e){this.dataIdsCount++,this.data.set(t,e)}has(t){return this.data.has(t)}delete(t){return this.dataIdsCount--,this.data.delete(t)}numDataIds(){return this.dataIdsCount}},zo=class{refCount(t){return $n(\"refCount\")}incRef(t){return $n(\"incRef\")}timerAvailable(){return!0}time(t){return $n(\"time\")}read(t){return $n(\"read\")}readSync(t){return $n(\"readSync\")}readToGPU(t,e){return $n(\"readToGPU\")}numDataIds(){return $n(\"numDataIds\")}disposeData(t,e){return $n(\"disposeData\")}write(t,e,n){return $n(\"write\")}move(t,e,n,o,s){return $n(\"move\")}createTensorFromTexture(t,e,n){return $n(\"createTensorFromTexture\")}memory(){return $n(\"memory\")}floatPrecision(){return $n(\"floatPrecision\")}epsilon(){return this.floatPrecision()===32?1e-7:1e-4}dispose(){return $n(\"dispose\")}};function $n(r){throw new Error(`'${r}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function s1(r){let t=r.length,e=0;for(;t>0;)e=Math.random()*t|0,t--,Lg(r,t,e)}function MU(r,t){if(r.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${r.length}Second array length was ${t.length}`);let e=r.length,n=0;for(;e>0;)n=Math.random()*e|0,e--,Lg(r,e,n),Lg(t,e,n)}function np(r,t,e){return Math.max(r,Math.min(t,e))}function zU(r){return r%2===0?r:r+1}function Lg(r,t,e){let n=r[t];r[t]=r[e],r[e]=n}function BU(r){let t=0;for(let e=0;e<r.length;e++)t+=r[e];return t}function VU(r,t){let e=Math.random();return t*e+(1-e)*r}function GU(r,t){let e=0;for(let n=0;n<r.length;n++){let o=Number(r[n])-Number(t[n]);e+=o*o}return e}function E(r,t){if(!r)throw new Error(typeof t==\"string\"?t:t())}function $e(r,t,e=\"\"){E(Dn(r,t),()=>e+` Shapes ${r} and ${t} must match`)}function Yn(r){E(r!=null,()=>\"The input to the tensor constructor must be a non-null value.\")}function Bo(r,t=[],e=!1){if(t==null&&(t=[]),Array.isArray(r)||xr(r)&&!e)for(let n=0;n<r.length;++n)Bo(r[n],t,e);else t.push(r);return t}function Jt(r){if(r.length===0)return 1;let t=r[0];for(let e=1;e<r.length;e++)t*=r[e];return t}function WU(r){return r.length===0}function Dn(r,t){if(r===t)return!0;if(r==null||t==null||r.length!==t.length)return!1;for(let e=0;e<r.length;e++)if(r[e]!==t[e])return!1;return!0}function na(r){return r%1===0}function UU(r){if(Math.tanh!=null)return Math.tanh(r);if(r===1/0)return 1;if(r===-1/0)return-1;{let t=Math.exp(2*r);return(t-1)/(t+1)}}function HU(r){let t=Math.ceil(Math.sqrt(r));return[t,Math.ceil(r/t)]}function qU(r){let t=new Uint32Array(r);for(let e=0;e<r;++e)t[e]=e;return s1(t),t}function Pu(r,t){return t<=r.length?r:r+\" \".repeat(t-r.length)}function KU(r,t=o=>0,e,n){return new Promise((o,s)=>{let i=0,a=()=>{if(r()){o();return}i++;let u=t(i);if(e!=null&&i>=e){s();return}n!=null?n(a,u):setTimeout(a,u)};a()})}function jU(r,t){let e=1,n=-1;for(let s=0;s<r.length;++s)if(r[s]>=0)e*=r[s];else if(r[s]===-1){if(n!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${n} and dim ${s}`);n=s}else if(r[s]<0)throw Error(`Shapes can not be < 0. Found ${r[s]} at dim ${s}`);if(n===-1){if(t>0&&t!==e)throw Error(`Size(${t}) must match the product of shape ${r}`);return r}if(e===0)throw Error(`Cannot infer the missing size in [${r}] when there are 0 elements`);if(t%e!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${e}`);let o=r.slice();return o[n]=t/e,o}function lr(r,t){let e=t.length;return r=r==null?t.map((n,o)=>o):[].concat(r),E(r.every(n=>n>=-e&&n<e),()=>`All values in axis param must be in range [-${e}, ${e}) but got axis ${r}`),E(r.every(n=>na(n)),()=>`All values in axis param must be integers but got axis ${r}`),r.map(n=>n<0?e+n:n)}function t0(r,t){let e=[],n=[],o=t!=null&&Array.isArray(t)&&t.length===0,s=t==null||o?null:lr(t,r).sort(),i=0;for(let a=0;a<r.length;++a){if(s!=null){if(s[i]===a&&r[a]!==1)throw new Error(`Can't squeeze axis ${a} since its dim '${r[a]}' is not 1`);(s[i]==null||s[i]>a)&&r[a]===1&&(e.push(r[a]),n.push(a)),s[i]<=a&&i++}r[a]!==1&&(e.push(r[a]),n.push(a))}return{newShape:e,keptDims:n}}function e0(r,t){let e=null;if(r==null||r===\"float32\")e=new Float32Array(t);else if(r===\"int32\")e=new Int32Array(t);else if(r===\"bool\")e=new Uint8Array(t);else throw new Error(`Unknown data type ${r}`);return e}function r0(r,t){let e=null;if(r==null||r===\"float32\")e=new Float32Array(t);else if(r===\"int32\")e=new Int32Array(t);else if(r===\"bool\")e=new Uint8Array(t);else if(r===\"string\")e=new Array(t);else throw new Error(`Unknown data type ${r}`);return e}function n0(r,t){for(let e=0;e<r.length;e++){let n=r[e];if(isNaN(n)||!isFinite(n))throw Error(`A tensor of type ${t} being uploaded contains ${n}.`)}}function o0(r){return r===\"bool\"||r===\"complex64\"||r===\"float32\"||r===\"int32\"||r===\"string\"}function XU(r,t){return!(t===\"complex64\"||t===\"float32\"&&r!==\"complex64\"||t===\"int32\"&&r!==\"float32\"&&r!==\"complex64\"||t===\"bool\"&&r===\"bool\")}function xr(r){return r instanceof Float32Array||r instanceof Int32Array||r instanceof Uint8Array||r instanceof Uint8ClampedArray}function Mg(r){if(r===\"float32\"||r===\"int32\")return 4;if(r===\"complex64\")return 8;if(r===\"bool\")return 1;throw new Error(`Unknown dtype ${r}`)}function s0(r){if(r==null)return 0;let t=0;return r.forEach(e=>t+=e.length),t}function Vo(r){return typeof r==\"string\"||r instanceof String}function i1(r){return typeof r==\"boolean\"}function a1(r){return typeof r==\"number\"}function op(r){return Array.isArray(r)?op(r[0]):r instanceof Float32Array?\"float32\":r instanceof Int32Array||r instanceof Uint8Array||r instanceof Uint8ClampedArray?\"int32\":a1(r)?\"float32\":Vo(r)?\"string\":i1(r)?\"bool\":\"float32\"}function oi(r){return!!(r&&r.constructor&&r.call&&r.apply)}function sp(r,t){for(let e=t;e<r;++e)if(r%e===0)return e;return r}function si(r){let t=r.length;if(t<2)return[];let e=new Array(t-1);e[t-2]=r[t-1];for(let n=t-3;n>=0;--n)e[n]=e[n+1]*r[n+1];return e}function l1(r,t,e,n=!1){let o=new Array;if(t.length===1){let s=t[0]*(n?2:1);for(let i=0;i<s;i++)o[i]=e[r+i]}else{let s=t[0],i=t.slice(1),a=i.reduce((u,l)=>u*l)*(n?2:1);for(let u=0;u<s;u++)o[u]=l1(r+u*a,i,e,n)}return o}function Ou(r,t,e=!1){if(r.length===0)return t[0];let n=r.reduce((o,s)=>o*s)*(e?2:1);if(n===0)return[];if(n!==t.length)throw new Error(`[${r}] does not match the input size ${t.length}${e?\" for a complex tensor\":\"\"}.`);return l1(0,r,t,e)}function Wd(r,t){let e=ip(r,t);for(let n=0;n<e.length;n++)e[n]=1;return e}function ip(r,t){if(t==null||t===\"float32\"||t===\"complex64\")return new Float32Array(r);if(t===\"int32\")return new Int32Array(r);if(t===\"bool\")return new Uint8Array(r);throw new Error(`Unknown data type ${t}`)}function YU(r,t){let e=r.reduce((n,o)=>n*o,1);if(t==null||t===\"float32\")return Ou(r,new Float32Array(e));if(t===\"int32\")return Ou(r,new Int32Array(e));if(t===\"bool\")return Ou(r,new Uint8Array(e));throw new Error(`Unknown data type ${t}`)}function Ud(r){r.forEach(t=>{E(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${r}].`)})}function ZU(r,t,e){if(t===0)return 0;if(t===1)return r[0];let n=r[r.length-1];for(let o=0;o<r.length-1;++o)n+=e[o]*r[o];return n}function JU(r,t,e){if(t===0)return[];if(t===1)return[r];let n=new Array(t);for(let o=0;o<n.length-1;++o)n[o]=Math.floor(r/e[o]),r-=n[o]*e[o];return n[n.length-1]=r,n}function Hd(r){return r&&r.then&&typeof r.then==\"function\"}var u1=\"tfjsflags\",qd=class{constructor(t){this.global=t,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=t4,this.populateURLFlags()}setPlatform(t,e){this.platform!=null&&(z().getBool(\"IS_TEST\")||z().getBool(\"PROD\")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${t}.`)),this.platformName=t,this.platform=e}registerFlag(t,e,n){if(this.flagRegistry[t]={evaluationFn:e,setHook:n},this.urlFlags[t]!=null){let o=this.urlFlags[t];z().getBool(\"IS_TEST\")||z().getBool(\"PROD\")||console.warn(`Setting feature override from URL ${t}: ${o}.`),this.set(t,o)}}async getAsync(t){return t in this.flags?this.flags[t]:(this.flags[t]=await this.evaluateFlag(t),this.flags[t])}get(t){if(t in this.flags)return this.flags[t];let e=this.evaluateFlag(t);if(Hd(e))throw new Error(`Flag ${t} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[t]=e,this.flags[t]}getNumber(t){return this.get(t)}getBool(t){return this.get(t)}getFlags(){return this.flags}get features(){return this.flags}set(t,e){if(this.flagRegistry[t]==null)throw new Error(`Cannot set flag ${t} as it has not been registered.`);this.flags[t]=e,this.flagRegistry[t].setHook!=null&&this.flagRegistry[t].setHook(e)}evaluateFlag(t){if(this.flagRegistry[t]==null)throw new Error(`Cannot evaluate flag '${t}': no evaluation function found.`);return this.flagRegistry[t].evaluationFn()}setFlags(t){this.flags=Object.assign({},t)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global==\"undefined\"||typeof this.global.location==\"undefined\"||typeof this.global.location.search==\"undefined\")return;let t=this.getQueryParams(this.global.location.search);u1 in t&&t[u1].split(\",\").forEach(n=>{let[o,s]=n.split(\":\");this.urlFlags[o]=r4(o,s)})}};function t4(r){let t={};return r.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(e,...n)=>(e4(t,n[0],n[1]),n.join(\"=\"))),t}function e4(r,t,e){r[decodeURIComponent(t)]=decodeURIComponent(e||\"\")}function r4(r,t){if(t=t.toLowerCase(),t===\"true\"||t===\"false\")return t===\"true\";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${r}.`)}function z(){return i0}var i0=null;function c1(r){i0=r}var a0;function l0(){if(a0==null){let r;if(typeof window!=\"undefined\")r=window;else if(typeof global!=\"undefined\")r=global;else if(typeof process!=\"undefined\")r=process;else if(typeof self!=\"undefined\")r=self;else throw new Error(\"Could not find a global object\");a0=r}return a0}function n4(){let r=l0();return r._tfGlobals==null&&(r._tfGlobals=new Map),r._tfGlobals}function Kd(r,t){let e=n4();if(e.has(r))return e.get(r);{let n=t();return e.set(r,n),e.get(r)}}var ii=\"Abs\",oa=\"Acos\",sa=\"Acosh\",Zn=\"Add\",Go=\"AddN\",ia=\"All\",aa=\"Any\",Wo=\"ArgMax\",kl=\"ArgMin\",la=\"Asin\",ua=\"Asinh\",ca=\"Atan\",pa=\"Atanh\",ma=\"Atan2\",Uo=\"AvgPool\",ap=\"AvgPoolGrad\",El=\"AvgPool3D\",lp=\"AvgPool3DGrad\",Ho=\"BatchMatMul\",ai=\"BatchToSpaceND\",up=\"Bincount\",p1=\"BroadcastTo\",cp=\"BroadcastArgs\",lo=\"Cast\",qo=\"Ceil\",uo=\"ClipByValue\",pp=\"Complex\",_l=\"ComplexAbs\",li=\"Concat\",Ko=\"Conv2D\",mp=\"Conv2DBackpropFilter\",jo=\"Conv2DBackpropInput\",Al=\"Conv3D\",fp=\"Conv3DBackpropFilterV2\",dp=\"Conv3DBackpropInputV2\",Xo=\"Cos\",Yo=\"Cosh\",fa=\"Cumprod\",Zo=\"Cumsum\",da=\"CropAndResize\",hp=\"DenseBincount\",ha=\"DepthToSpace\",Jo=\"DepthwiseConv2dNative\",gp=\"DepthwiseConv2dNativeBackpropFilter\",xp=\"DepthwiseConv2dNativeBackpropInput\",yp=\"Diag\",$l=\"Dilation2D\",jd=\"Dilation2DBackpropInput\",Xd=\"Dilation2DBackpropFilter\",Qo=\"RealDiv\",bp=\"Einsum\",ts=\"Elu\",wp=\"EluGrad\",ga=\"Erf\",xa=\"Equal\",es=\"Exp\",ui=\"ExpandDims\",ya=\"Expm1\",Cp=\"FFT\",Dl=\"Fill\",ba=\"FlipLeftRight\",rs=\"Floor\",ns=\"FloorDiv\",os=\"FusedBatchNorm\",ci=\"GatherV2\",wa=\"GatherNd\",Ca=\"Greater\",ss=\"GreaterEqual\",co=\"Identity\",Ip=\"IFFT\",Sp=\"Imag\",Ia=\"IsFinite\",Sa=\"IsInf\",va=\"IsNan\",is=\"LeakyRelu\",Na=\"Less\",Ta=\"LessEqual\",vp=\"LinSpace\",as=\"Log\",ka=\"Log1p\",Ea=\"LogicalAnd\",_a=\"LogicalNot\",Aa=\"LogicalOr\",m1=\"LogicalXor\",f1=\"LogSoftmax\",xlt=\"LowerBound\",Rl=\"LRN\",Np=\"LRNGrad\",ls=\"Max\",us=\"Maximum\",cs=\"MaxPool\",Tp=\"MaxPoolGrad\",Fl=\"MaxPool3D\",kp=\"MaxPool3DGrad\",Ep=\"MaxPoolWithArgmax\",ps=\"Mean\",ms=\"Min\",fs=\"Minimum\",ds=\"MirrorPad\",$a=\"Mod\",_p=\"Multinomial\",hs=\"Multiply\",pi=\"Neg\",Da=\"NotEqual\",Ra=\"NonMaxSuppressionV3\",Fa=\"NonMaxSuppressionV4\",Oa=\"NonMaxSuppressionV5\",mi=\"OnesLike\",gs=\"OneHot\",fi=\"Pack\",xs=\"PadV2\",ylt=\"Pool\",ys=\"Pow\",bs=\"Prelu\",ws=\"Prod\",Ap=\"RaggedGather\",$p=\"RaggedRange\",Dp=\"RaggedTensorToTensor\",Ol=\"Range\",Rp=\"Real\",Pa=\"Reciprocal\",Cs=\"Relu\",di=\"Reshape\",Is=\"ResizeNearestNeighbor\",Fp=\"ResizeNearestNeighborGrad\",Ss=\"ResizeBilinear\",Op=\"ResizeBilinearGrad\",vs=\"Relu6\",Ns=\"Reverse\",Ts=\"Round\",ks=\"Rsqrt\",La=\"ScatterNd\",Pp=\"SearchSorted\",hi=\"Select\",Ma=\"Selu\",gi=\"Slice\",Es=\"Sin\",za=\"Sinh\",Ba=\"Sign\",_s=\"Sigmoid\",Va=\"Softplus\",As=\"Sqrt\",$s=\"Sum\",xi=\"SpaceToBatchND\",yi=\"SplitV\",Ds=\"Softmax\",Pl=\"SparseFillEmptyRows\",Ga=\"SparseReshape\",Ll=\"SparseSegmentMean\",Ml=\"SparseSegmentSum\",Lp=\"SparseToDense\",Rs=\"SquaredDifference\",zl=\"Square\",Wa=\"StridedSlice\",Bl=\"StringNGrams\",Vl=\"StringSplit\",Gl=\"StringToHashBucketFast\",Fs=\"Sub\",Os=\"Tan\",Ps=\"Tanh\",Jn=\"Tile\",Ua=\"TopK\",Ha=\"Transform\",Qn=\"Transpose\",Mp=\"Unique\",bi=\"Unpack\",Wl=\"UnsortedSegmentSum\",blt=\"UpperBound\",wi=\"ZerosLike\",po=\"Step\",Yd=\"FromPixels\",qa=\"RotateWithOffset\",Ci=\"_FusedMatMul\",Ii=\"FusedConv2D\",Si=\"FusedDepthwiseConv2D\";function vi(...r){z().getBool(\"IS_TEST\")||z().getBool(\"PROD\")||console.warn(...r)}function o4(...r){z().getBool(\"IS_TEST\")||z().getBool(\"PROD\")||console.log(...r)}var zp=Kd(\"kernelRegistry\",()=>new Map),Zd=Kd(\"gradRegistry\",()=>new Map);function Jd(r,t){let e=c0(r,t);return zp.get(e)}function u0(r){return Zd.get(r)}function zg(r){let t=zp.entries(),e=[];for(;;){let{done:n,value:o}=t.next();if(n)break;let[s,i]=o,[a]=s.split(\"_\");a===r&&e.push(i)}return e}function Lu(r){let{kernelName:t,backendName:e}=r,n=c0(t,e);zp.has(n)&&vi(`The kernel '${t}' for backend '${e}' is already registered`),zp.set(n,r)}function h1(r){let{kernelName:t}=r;Zd.has(t)&&z().getBool(\"DEBUG\")&&vi(`Overriding the gradient for '${t}'`),Zd.set(t,r)}function vlt(r,t){let e=c0(r,t);if(!zp.has(e))throw new Error(`The kernel '${r}' for backend '${t}' is not registered`);zp.delete(e)}function Nlt(r){if(!Zd.has(r))throw new Error(`The gradient '${r}' for backend is not registered`);Zd.delete(r)}function Tlt(r,t){zg(r).forEach(n=>{let o=Object.assign({},n,{backendName:t});Lu(o)})}function c0(r,t){return`${t}_${r}`}var y={};Wt(y,{arraysEqual:()=>Dn,assert:()=>E,assertNonNegativeIntegerDimensions:()=>Ud,assertNonNull:()=>Yn,assertShapesMatch:()=>$e,bytesFromStringArray:()=>s0,bytesPerElement:()=>Mg,checkConversionForErrors:()=>n0,clamp:()=>np,computeStrides:()=>si,createScalarValue:()=>p4,createShuffledIndices:()=>qU,decodeString:()=>Wp,distSquared:()=>GU,encodeString:()=>Hl,fetch:()=>f4,fingerPrint64:()=>c4,flatten:()=>Bo,getArrayFromDType:()=>r0,getTypedArrayFromDType:()=>e0,hasEncodingLoss:()=>XU,hexToLong:()=>Qd,indexToLoc:()=>JU,inferDtype:()=>op,inferFromImplicitShape:()=>jU,isBoolean:()=>i1,isFunction:()=>oi,isInt:()=>na,isNumber:()=>a1,isPromise:()=>Hd,isScalarShape:()=>WU,isString:()=>Vo,isTypedArray:()=>xr,isValidDtype:()=>o0,locToIndex:()=>ZU,makeOnesTypedArray:()=>Wd,makeZerosNestedTypedArray:()=>YU,makeZerosTypedArray:()=>ip,nearestDivisor:()=>sp,nearestLargerEven:()=>zU,now:()=>Gu,parseAxisParam:()=>lr,randUniform:()=>VU,repeatedTry:()=>KU,rightPad:()=>Pu,shuffle:()=>s1,shuffleCombo:()=>MU,sizeFromShape:()=>Jt,sizeToSquarishShape:()=>HU,squeezeShape:()=>t0,sum:()=>BU,swap:()=>Lg,tanh:()=>UU,toNestedArray:()=>Ou,toTypedArray:()=>Gp});var d0=Tl(T1());var Vu=d0.default||d0;function Qd(r){return Vu.fromString(r,!0,16)}var E1=Qd(\"c3a5c85c97cb3127\"),Bu=Qd(\"b492b66fbe98f273\"),nn=Qd(\"9ae16a3b2f90404f\");function f0(r){return r.xor(r.shru(47))}function _1(r,t,e){let n=r.slice(t,t+e);return Vu.fromBytes(Array.from(n),!0,!0)}function Pe(r,t){return _1(r,t,8)}function k1(r,t){return _1(r,t,4)}function kr(r,t){return t===0?r:r.shru(t).or(r.shl(64-t))}function Ul(r,t,e=Qd(\"9ddfea08eb382d69\")){let n=r.xor(t).mul(e);n=n.xor(n.shru(47));let o=t.xor(n).mul(e);return o=o.xor(o.shru(47)),o=o.mul(e),o}function i4(r,t,e,n,o,s){o=o.add(r),s=kr(s.add(o).add(n),21);let i=o;return o=o.add(t),o=o.add(e),s=s.add(kr(o,44)),[o.add(n),s.add(i)]}function Vg(r,t,e,n){return i4(Pe(r,t),Pe(r,t+8),Pe(r,t+16),Pe(r,t+24),e,n)}function a4(r,t=r.length){if(t>=8){let e=nn.add(t*2),n=Pe(r,0).add(nn),o=Pe(r,t-8),s=kr(o,37).mul(e).add(n),i=kr(n,25).add(o).mul(e);return Ul(s,i,e)}if(t>=4){let e=nn.add(t*2),n=k1(r,0);return Ul(n.shl(3).add(t),k1(r,t-4),e)}if(t>0){let e=r[0],n=r[t>>1],o=r[t-1],s=e+(n<<8),i=t+(o<<2);return f0(nn.mul(s).xor(E1.mul(i))).mul(nn)}return nn}function l4(r,t=r.length){let e=nn.add(t*2),n=Pe(r,0).mul(Bu),o=Pe(r,8),s=Pe(r,t-8).mul(e),i=Pe(r,t-16).mul(nn);return Ul(kr(n.add(o),43).add(kr(s,30)).add(i),n.add(kr(o.add(nn),18)).add(s),e)}function u4(r,t=r.length){let e=nn.add(t*2),n=Pe(r,0).mul(nn),o=Pe(r,8),s=Pe(r,t-8).mul(e),i=Pe(r,t-16).mul(nn),a=kr(n.add(o),43).add(kr(s,30)).add(i),u=Ul(a,n.add(kr(o.add(nn),18)).add(s),e),l=Pe(r,16).mul(e),c=Pe(r,24),p=a.add(Pe(r,t-32)).mul(e),m=u.add(Pe(r,t-24)).mul(e);return Ul(kr(l.add(c),43).add(kr(p,30)).add(m),l.add(kr(c.add(n),18)).add(p),e)}function c4(r,t=r.length){let e=Vu.fromNumber(81,!0);if(t<=32)return t<=16?a4(r,t):l4(r,t);if(t<=64)return u4(r,t);let n=e,o=e.mul(Bu).add(113),s=f0(o.mul(nn).add(113)).mul(nn),i=[Vu.UZERO,Vu.UZERO],a=[Vu.UZERO,Vu.UZERO];n=n.mul(nn).add(Pe(r,0));let u=0,l=(t-1>>6)*64,c=l+(t-1&63)-63;do n=kr(n.add(o).add(i[0]).add(Pe(r,u+8)),37).mul(Bu),o=kr(o.add(i[1]).add(Pe(r,u+48)),42).mul(Bu),n=n.xor(a[1]),o=o.add(i[0]).add(Pe(r,u+40)),s=kr(s.add(a[0]),33).mul(Bu),i=Vg(r,u,i[1].mul(Bu),n.add(a[0])),a=Vg(r,u+32,s.add(a[1]),o.add(Pe(r,u+16))),[s,n]=[n,s],u+=64;while(u!==l);let p=Bu.add(s.and(255).shl(1));return u=c,a[0]=a[0].add(t-1&63),i[0]=i[0].add(a[0]),a[0]=a[0].add(i[0]),n=kr(n.add(o).add(i[0]).add(Pe(r,u+8)),37).mul(p),o=kr(o.add(i[1]).add(Pe(r,u+48)),42).mul(p),n=n.xor(a[1].mul(9)),o=o.add(i[0].mul(9).add(Pe(r,u+40))),s=kr(s.add(a[0]),33).mul(p),i=Vg(r,u,i[1].mul(p),n.add(a[0])),a=Vg(r,u+32,s.add(a[1]),o.add(Pe(r,u+16))),[s,n]=[n,s],Ul(Ul(i[0],a[0],p).add(f0(o).mul(E1)).add(s),Ul(i[1],a[1],p).add(n),p)}function p4(r,t){return t===\"string\"?Hl(r):Gp([r],t)}function m4(r,t){return r instanceof Float32Array&&t===\"float32\"||r instanceof Int32Array&&t===\"int32\"||r instanceof Uint8Array&&t===\"bool\"}function Gp(r,t){if(t===\"string\")throw new Error(\"Cannot convert a string[] to a TypedArray\");if(Array.isArray(r)&&(r=Bo(r)),z().getBool(\"DEBUG\")&&n0(r,t),m4(r,t))return r;if(t==null||t===\"float32\"||t===\"complex64\")return new Float32Array(r);if(t===\"int32\")return new Int32Array(r);if(t===\"bool\"){let e=new Uint8Array(r.length);for(let n=0;n<e.length;++n)Math.round(r[n])!==0&&(e[n]=1);return e}else throw new Error(`Unknown data type ${t}`)}function Gu(){return z().platform.now()}function f4(r,t){return z().platform.fetch(r,t)}function Hl(r,t=\"utf-8\"){return t=t||\"utf-8\",z().platform.encode(r,t)}function Wp(r,t=\"utf-8\"){return t=t||\"utf-8\",z().platform.decode(r,t)}var Gg=class{constructor(t,e){this.backendTimer=t,this.logger=e,e==null&&(this.logger=new h0)}profileKernel(t,e,n){let o,s=()=>{o=n()},i,a=Gu();if(this.backendTimer.timerAvailable())i=this.backendTimer.time(s);else{s();for(let l of o)l.dataSync();i=Promise.resolve({kernelMs:Gu()-a})}if(z().getBool(\"CHECK_COMPUTATION_FOR_ERRORS\"))for(let l=0;l<o.length;l++){let c=o[l];c.data().then(p=>{d4(p,c.dtype,t)})}return{kernelName:t,outputs:o,inputs:e,timeMs:i.then(l=>l.kernelMs),extraInfo:i.then(l=>l.getExtraProfileInfo!=null?l.getExtraProfileInfo():\"\")}}logKernelProfile(t){let{kernelName:e,outputs:n,timeMs:o,inputs:s,extraInfo:i}=t;n.forEach(a=>{Promise.all([a.data(),o,i]).then(u=>{this.logger.logKernelProfile(e,a,u[0],u[1],s,u[2])})})}};function d4(r,t,e){if(t!==\"float32\")return!1;for(let n=0;n<r.length;n++){let o=r[n];if(isNaN(o)||!isFinite(o))return console.warn(`Found ${o} in the result of '${e}'`),!0}return!1}var h0=class{logKernelProfile(t,e,n,o,s,i){let a=typeof o==\"number\"?Pu(`${o}ms`,9):o.error,u=Pu(t,25),l=e.rank,c=e.size,p=Pu(e.shape.toString(),14),m=\"\";for(let f in s){let d=s[f];if(d!=null){let h=d.shape||e.shape,g=h.length;m+=`${f}: ${g}D ${g>0?h:\"\"} `}}console.log(`%c${u}\t%c${a}\t%c${l}D ${p}\t%c${c}\t%c${m}\t%c${i}`,\"font-weight:bold\",\"color:red\",\"color:blue\",\"color: orange\",\"color: green\",\"color: steelblue\")}};function A1(r,t,e){let n={},o={};for(let u=0;u<t.length;u++)n[t[u].id]=!0;for(let u=0;u<r.length;u++){let l=r[u],c=l.inputs;for(let p in c){let m=c[p],f=!1;for(let d=0;d<t.length;d++)if(n[m.id]){l.outputs.forEach(h=>n[h.id]=!0),f=!0,o[l.id]=!0;break}if(f)break}}let s={};s[e.id]=!0;let i={};for(let u=r.length-1;u>=0;u--){let l=r[u],c=l.inputs;for(let p=0;p<l.outputs.length;p++)if(s[l.outputs[p].id]){for(let m in c)s[c[m].id]=!0,i[l.id]=!0;break}}let a=[];for(let u=0;u<r.length;u++){let l=r[u];if(o[l.id]&&i[l.id]){let c={};for(let m in l.inputs){let f=l.inputs[m];n[f.id]&&(c[m]=f)}let p=Object.assign({},l);p.inputs=c,p.outputs=l.outputs,a.push(p)}}return a}function $1(r,t,e,n){for(let o=t.length-1;o>=0;o--){let s=t[o],i=[];if(s.outputs.forEach(u=>{let l=r[u.id];l!=null?i.push(l):i.push(null)}),s.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let a=s.gradient(i);for(let u in s.inputs){if(!(u in a))throw new Error(`Cannot backprop through input ${u}. Available gradients found: ${Object.keys(a)}.`);let l=e(()=>a[u]());if(l.dtype!==\"float32\")throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${u} must have 'float32' dtype, but has '${l.dtype}'`);let c=s.inputs[u];if(!Dn(l.shape,c.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${u}' has shape '${l.shape}', which does not match the shape of the input '${c.shape}'`);if(r[c.id]==null)r[c.id]=l;else{let p=r[c.id];r[c.id]=n(p,l),p.dispose()}}}}var D1=20,th=3,g0=7;function R1(r,t,e,n){let o=si(t),s=h4(r,t,e,o),i=t.length,a=Wg(r,t,e,o,s),u=[\"Tensor\"];return n&&(u.push(`  dtype: ${e}`),u.push(`  rank: ${i}`),u.push(`  shape: [${t}]`),u.push(\"  values:\")),u.push(a.map(l=>\"    \"+l).join(`\n`)),u.join(`\n`)}function h4(r,t,e,n){let o=Jt(t),s=n[n.length-1],i=new Array(s).fill(0),a=t.length,u=e===\"complex64\"?rh(r):r;if(a>1)for(let l=0;l<o/s;l++){let c=l*s;for(let p=0;p<s;p++)i[p]=Math.max(i[p],eh(u[c+p],0,e).length)}return i}function eh(r,t,e){let n;return Array.isArray(r)?n=`${parseFloat(r[0].toFixed(g0))} + ${parseFloat(r[1].toFixed(g0))}j`:Vo(r)?n=`'${r}'`:e===\"bool\"?n=F1(r):n=parseFloat(r.toFixed(g0)).toString(),Pu(n,t)}function F1(r){return r===0?\"false\":\"true\"}function Wg(r,t,e,n,o,s=!0){let i=e===\"complex64\"?2:1,a=t[0],u=t.length;if(u===0){if(e===\"complex64\"){let h=rh(r);return[eh(h[0],0,e)]}return e===\"bool\"?[F1(r[0])]:[r[0].toString()]}if(u===1){if(a>D1){let g=th*i,x=Array.from(r.slice(0,g)),b=Array.from(r.slice((a-th)*i,a*i));return e===\"complex64\"&&(x=rh(x),b=rh(b)),[\"[\"+x.map((w,C)=>eh(w,o[C],e)).join(\", \")+\", ..., \"+b.map((w,C)=>eh(w,o[a-th+C],e)).join(\", \")+\"]\"]}let h=e===\"complex64\"?rh(r):Array.from(r);return[\"[\"+h.map((g,x)=>eh(g,o[x],e)).join(\", \")+\"]\"]}let l=t.slice(1),c=n.slice(1),p=n[0]*i,m=[];if(a>D1){for(let h=0;h<th;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,!1))}m.push(\"...\");for(let h=a-th;h<a;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,h===a-1))}}else for(let h=0;h<a;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,h===a-1))}let f=u===2?\",\":\"\";m[0]=\"[\"+m[0]+f;for(let h=1;h<m.length-1;h++)m[h]=\" \"+m[h]+f;let d=`,\n`;for(let h=2;h<u;h++)d+=`\n`;return m[m.length-1]=\" \"+m[m.length-1]+\"]\"+(s?\"\":d),m}function rh(r){let t=[];for(let e=0;e<r.length;e+=2)t.push([r[e],r[e+1]]);return t}var pe=class{constructor(t,e,n){if(this.dtype=e,this.shape=t.slice(),this.size=Jt(t),n!=null){let o=n.length;E(o===this.size,()=>`Length of values '${o}' does not match the size inferred by the shape '${this.size}'.`)}if(e===\"complex64\")throw new Error(\"complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).\");this.values=n||r0(e,this.size),this.strides=si(t)}set(t,...e){e.length===0&&(e=[0]),E(e.length===this.rank,()=>`The number of provided coordinates (${e.length}) must match the rank (${this.rank})`);let n=this.locToIndex(e);this.values[n]=t}get(...t){t.length===0&&(t=[0]);let e=0;for(let o of t){if(o<0||o>=this.shape[e]){let s=`Requested out of range element at ${t}.   Buffer shape=${this.shape}`;throw new Error(s)}e++}let n=t[t.length-1];for(let o=0;o<t.length-1;++o)n+=this.strides[o]*t[o];return this.values[n]}locToIndex(t){if(this.rank===0)return 0;if(this.rank===1)return t[0];let e=t[t.length-1];for(let n=0;n<t.length-1;++n)e+=this.strides[n]*t[n];return e}indexToLoc(t){if(this.rank===0)return[];if(this.rank===1)return[t];let e=new Array(this.shape.length);for(let n=0;n<e.length-1;++n)e[n]=Math.floor(t/this.strides[n]),t-=e[n]*this.strides[n];return e[e.length-1]=t,e}get rank(){return this.shape.length}toTensor(){return Ms().makeTensor(this.values,this.shape,this.dtype)}},Ms=null,Up=null,g4=null;function O1(r){Ms=r}function P1(r){Up=r}function L1(r){g4=r}var Ft=class{constructor(t,e,n,o){this.kept=!1,this.isDisposedInternal=!1,this.shape=t.slice(),this.dtype=e||\"float32\",this.size=Jt(t),this.strides=si(t),this.dataId=n,this.id=o,this.rankType=this.rank<5?this.rank.toString():\"higher\"}get rank(){return this.shape.length}async buffer(){let t=await this.data();return Up.buffer(this.shape,this.dtype,t)}bufferSync(){return Up.buffer(this.shape,this.dtype,this.dataSync())}async array(){let t=await this.data();return Ou(this.shape,t,this.dtype===\"complex64\")}arraySync(){return Ou(this.shape,this.dataSync(),this.dtype===\"complex64\")}async data(){this.throwIfDisposed();let t=Ms().read(this.dataId);if(this.dtype===\"string\"){let e=await t;try{return e.map(n=>Wp(n))}catch(n){throw new Error(\"Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().\")}}return t}dataToGPU(t){return this.throwIfDisposed(),Ms().readToGPU(this.dataId,t)}dataSync(){this.throwIfDisposed();let t=Ms().readSync(this.dataId);if(this.dtype===\"string\")try{return t.map(e=>Wp(e))}catch(e){throw new Error(\"Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().\")}return t}async bytes(){this.throwIfDisposed();let t=await Ms().read(this.dataId);return this.dtype===\"string\"?t:new Uint8Array(t.buffer)}dispose(){this.isDisposed||(Ms().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error(\"Tensor is disposed.\")}print(t=!1){return Up.print(this,t)}clone(){return this.throwIfDisposed(),Up.clone(this)}toString(t=!1){let e=this.dataSync();return R1(e,this.shape,this.dtype,t)}cast(t){return this.throwIfDisposed(),Up.cast(this,t)}variable(t=!0,e,n){return this.throwIfDisposed(),Ms().makeVariable(this,t,e,n)}};Object.defineProperty(Ft,Symbol.hasInstance,{value:r=>!!r&&r.data!=null&&r.dataSync!=null&&r.throwIfDisposed!=null});function O(){return Kd(\"Tensor\",()=>Ft)}O();var Ka=class extends Ft{constructor(t,e,n,o){super(t.shape,t.dtype,t.dataId,o),this.trainable=e,this.name=n}assign(t){if(t.dtype!==this.dtype)throw new Error(`dtype of the new value (${t.dtype}) and previous value (${this.dtype}) must match`);if(!Dn(t.shape,this.shape))throw new Error(`shape of the new value (${t.shape}) and previous value (${this.shape}) must match`);Ms().disposeTensor(this),this.dataId=t.dataId,Ms().incRef(this,null)}dispose(){Ms().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(Ka,Symbol.hasInstance,{value:r=>r instanceof Ft&&r.assign!=null&&r.assign instanceof Function});var go={};Wt(go,{assertTypesMatch:()=>I0,getTensorsInContainer:()=>nh,isTensorInList:()=>y4,makeTypesMatch:()=>Ut});var x0;(function(r){r.R0=\"R0\",r.R1=\"R1\",r.R2=\"R2\",r.R3=\"R3\",r.R4=\"R4\",r.R5=\"R5\",r.R6=\"R6\"})(x0||(x0={}));var y0;(function(r){r.float32=\"float32\",r.int32=\"int32\",r.bool=\"int32\",r.complex64=\"complex64\"})(y0||(y0={}));var b0;(function(r){r.float32=\"float32\",r.int32=\"int32\",r.bool=\"bool\",r.complex64=\"complex64\"})(b0||(b0={}));var w0;(function(r){r.float32=\"float32\",r.int32=\"float32\",r.bool=\"float32\",r.complex64=\"complex64\"})(w0||(w0={}));var C0;(function(r){r.float32=\"complex64\",r.int32=\"complex64\",r.bool=\"complex64\",r.complex64=\"complex64\"})(C0||(C0={}));var x4={float32:w0,int32:y0,bool:b0,complex64:C0};function sr(r,t){if(r===\"string\"||t===\"string\"){if(r===\"string\"&&t===\"string\")return\"string\";throw new Error(`Can not upcast ${r} with ${t}`)}return x4[r][t]}function Wu(r){return sr(r,\"int32\")}function Ut(r,t){if(r.dtype===t.dtype)return[r,t];let e=sr(r.dtype,t.dtype);return[r.cast(e),t.cast(e)]}function I0(r,t){E(r.dtype===t.dtype,()=>`The dtypes of the first(${r.dtype}) and second(${t.dtype}) input must match`)}function y4(r,t){return t.some(e=>e.id===r.id)}function nh(r){let t=[];return M1(r,t,new Set),t}function M1(r,t,e){if(r==null)return;if(r instanceof Ft){t.push(r);return}if(!b4(r))return;let n=r;for(let o in n){let s=n[o];e.has(s)||(e.add(s),M1(s,t,e))}}function b4(r){return Array.isArray(r)||typeof r==\"object\"}function S0(r){return r.kernelName!=null}var Ug=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(t=>t.name)))}}}dispose(){for(let t in this.registeredVariables)this.registeredVariables[t].dispose()}},ql=class{constructor(t){this.ENV=t,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new Ug}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let t=this.getSortedBackends();for(let e=0;e<t.length;e++){let n=t[e];if(await this.initializeBackend(n).success){await this.setBackend(n);return}}throw new Error(\"Could not initialize any backends, all backend initializations failed.\")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:t,asyncInit:e}=this.initializeBackendsAndReturnBest();if(e)throw new Error(`The highest priority backend '${t}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(t)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(t){if(!(t in this.registry))if(t in this.registryFactory){let{asyncInit:e}=this.initializeBackend(t);if(e)return null}else return null;return this.registry[t]}findBackendFactory(t){return t in this.registryFactory?this.registryFactory[t].factory:null}registerBackend(t,e,n=1){return t in this.registryFactory?(vi(`${t} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[t]={factory:e,priority:n},!0)}async setBackend(t){if(this.registryFactory[t]==null)throw new Error(`Backend name '${t}' not found in registry`);if(this.backendName=t,this.registry[t]==null){this.backendInstance=null;let{success:e,asyncInit:n}=this.initializeBackend(t);if(!(n?await e:e))return!1}return this.backendInstance=this.registry[t],this.setupRegisteredKernels(),this.profiler=new Gg(this.backendInstance),!0}setupRegisteredKernels(){zg(this.backendName).forEach(e=>{e.setupFunc!=null&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(t){zg(t).forEach(n=>{n.disposeFunc!=null&&n.disposeFunc(this.registry[t])})}initializeBackend(t){let e=this.registryFactory[t];if(e==null)throw new Error(`Cannot initialize backend ${t}, no registration found.`);try{let n=e.factory();if(n&&!(n instanceof zo)&&typeof n.then==\"function\"){let o=++this.pendingBackendInitId,s=n.then(i=>o<this.pendingBackendInitId?!1:(this.registry[t]=i,this.pendingBackendInit=null,!0)).catch(i=>(o<this.pendingBackendInitId||(this.pendingBackendInit=null,vi(`Initialization of backend ${t} failed`),vi(i.stack||i.message)),!1));return this.pendingBackendInit=s,{success:s,asyncInit:!0}}else return this.registry[t]=n,{success:!0,asyncInit:!1}}catch(n){return vi(`Initialization of backend ${t} failed`),vi(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(t){if(!(t in this.registryFactory))throw new Error(`${t} backend not found in registry`);this.backendName===t&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,t in this.registry&&(this.disposeRegisteredKernels(t),this.registry[t].dispose(),delete this.registry[t]),delete this.registryFactory[t],this.backendName===t&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error(\"No backend found in registry.\");return Object.keys(this.registryFactory).sort((t,e)=>this.registryFactory[e].priority-this.registryFactory[t].priority)}initializeBackendsAndReturnBest(){let t=this.getSortedBackends();for(let e=0;e<t.length;e++){let n=t[e],{success:o,asyncInit:s}=this.initializeBackend(n);if(s||o)return{name:n,asyncInit:s}}throw new Error(\"Could not initialize any backends, all backend initializations failed.\")}moveData(t,e){let n=this.state.tensorInfo.get(e),o=n.backend,s=this.readSync(e),i=o.refCount(e);o.disposeData(e,!0),n.backend=t,t.move(e,s,n.shape,n.dtype,i),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(t,e){let n=null;if(e==null){if(typeof t!=\"function\")throw new Error(\"Please provide a function to tidy()\");e=t}else{if(typeof t!=\"string\"&&!(t instanceof String))throw new Error(\"When calling with two arguments, the first argument to tidy() must be a string\");if(typeof e!=\"function\")throw new Error(\"When calling with two arguments, the 2nd argument to tidy() must be a function\");n=t}let o;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(o),()=>(o=e(),o instanceof Promise&&console.error(\"Cannot return a Promise inside of tidy.\"),o))}scopedRun(t,e,n){t();try{let o=n();return e(),o}catch(o){throw e(),o}}nextTensorId(){return ql.nextTensorId++}nextVariableId(){return ql.nextVariableId++}clone(t){let e=k.runKernel(co,{x:t}),n={x:t},o=i=>({x:()=>{let a=\"float32\",u={x:i},l={dtype:a};return k.runKernel(lo,u,l)}}),s=[];return this.addTapeNode(this.state.activeScope.name,n,[e],o,s,{}),e}runKernel(t,e,n){if(this.backendName==null&&this.backend,!(Jd(t,this.backendName)!=null))throw new Error(`Kernel '${t}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:t,inputs:e,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool(\"IS_TEST\")}checkKernelForMemLeak(t,e,n){let o=this.backend.numDataIds(),s=0;n.forEach(u=>{s+=u.dtype===\"complex64\"?3:1});let i=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],a=o-e-s-i;if(a>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${a} data ids) after running '${t}'`)}runKernelFunc(t){let e,n=[],o=this.isTapeOn(),s=this.state.numBytes,i=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let a;this.backendName==null&&this.backend;let u,l=S0(t)?t.kernelName:this.state.activeScope!=null?this.state.activeScope.name:\"\";if(S0(t)){let{kernelName:d,inputs:h,attrs:g}=t;this.backendName==null&&this.backend;let x=Jd(d,this.backendName);E(x!=null,()=>`Cannot find registered kernel '${d}' for backend '${this.backendName}'`),a=()=>{let b=this.backend.numDataIds();u=x.kernelFunc({inputs:h,attrs:g,backend:this.backend});let w=Array.isArray(u)?u:[u];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(d,b,w);let C=w.map(N=>N.rank!=null?N:this.makeTensorFromTensorInfo(N));if(o){let N=this.getTensorsForGradient(d,h,C);n=this.saveTensorsForBackwardMode(N)}return C}}else{let{forwardFunc:d}=t,h=g=>{!o||(n=g.map(x=>this.keep(this.clone(x))))};a=()=>{let g=this.backend.numDataIds();u=this.tidy(()=>d(this.backend,h));let x=Array.isArray(u)?u:[u];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,g,x),x}}let{inputs:c,attrs:p}=t,m=S0(t)?null:t.backwardsFunc,f;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool(\"DEBUG\")&&!this.state.profiling?e=a():(f=this.profiler.profileKernel(l,c,()=>a()),this.ENV.getBool(\"DEBUG\")&&this.profiler.logKernelProfile(f),e=f.outputs)}),o&&this.addTapeNode(l,c,e,m,n,p),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-s,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-i,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(c).map(d=>c[d]!=null?c[d].shape:null),outputShapes:e.map(d=>d.shape),kernelTimeMs:f.timeMs,extraInfo:f.extraInfo}),Array.isArray(u)?e:e[0]}saveTensorsForBackwardMode(t){return t.map(n=>this.keep(this.clone(n)))}getTensorsForGradient(t,e,n){let o=u0(t);if(o!=null){let s=o.inputsToSave||[],i=o.outputsToSave||[],a;o.saveAllInputs?(E(Array.isArray(e),()=>\"saveAllInputs is true, expected inputs to be an array.\"),a=Object.keys(e).map(l=>e[l])):a=s.map(l=>e[l]);let u=n.filter((l,c)=>i[c]);return a.concat(u)}return[]}makeTensor(t,e,n,o){if(t==null)throw new Error(\"Values passed to engine.makeTensor() are null\");n=n||\"float32\",o=o||this.backend;let s=t;n===\"string\"&&Vo(t[0])&&(s=t.map(u=>Hl(u)));let i=o.write(s,e,n),a=new Ft(e,n,i,this.nextTensorId());if(this.trackTensor(a,o),n===\"string\"){let u=this.state.tensorInfo.get(i),l=s0(s);this.state.numBytes+=l-u.bytes,u.bytes=l}return a}makeTensorFromDataId(t,e,n,o){n=n||\"float32\";let s={dataId:t,shape:e,dtype:n};return this.makeTensorFromTensorInfo(s,o)}makeTensorFromTensorInfo(t,e){let{dataId:n,shape:o,dtype:s}=t,i=new Ft(o,s,n,this.nextTensorId());return this.trackTensor(i,e),i}makeVariable(t,e=!0,n,o){n=n||this.nextVariableId().toString(),o!=null&&o!==t.dtype&&(t=t.cast(o));let s=new Ka(t,e,n,this.nextTensorId());if(this.state.registeredVariables[s.name]!=null)throw new Error(`Variable with name ${s.name} was already registered`);return this.state.registeredVariables[s.name]=s,this.incRef(s,this.backend),s}trackTensor(t,e){this.state.numTensors++,t.dtype===\"string\"&&this.state.numStringTensors++;let n=0;t.dtype!==\"complex64\"&&t.dtype!==\"string\"&&(n=t.size*Mg(t.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(t.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(t.dataId,{backend:e||this.backend,dtype:t.dtype,shape:t.shape,bytes:n})),t instanceof Ka||this.track(t)}incRef(t,e){this.trackTensor(t,e),this.backend.incRef(t.dataId)}removeDataId(t,e){this.state.tensorInfo.has(t)&&this.state.tensorInfo.get(t).backend===e&&(this.state.tensorInfo.delete(t),this.state.numDataBuffers--)}disposeTensor(t){if(!this.state.tensorInfo.has(t.dataId))return;let e=this.state.tensorInfo.get(t.dataId);if(this.state.numTensors--,t.dtype===\"string\"&&(this.state.numStringTensors--,this.state.numBytes-=e.bytes),t.dtype!==\"complex64\"&&t.dtype!==\"string\"){let n=t.size*Mg(t.dtype);this.state.numBytes-=n}e.backend.disposeData(t.dataId)&&this.removeDataId(t.dataId,e.backend)}disposeVariables(){for(let t in this.state.registeredVariables){let e=this.state.registeredVariables[t];this.disposeVariable(e)}}disposeVariable(t){this.disposeTensor(t),this.state.registeredVariables[t.name]!=null&&delete this.state.registeredVariables[t.name]}memory(){let t=this.backend.memory();return t.numTensors=this.state.numTensors,t.numDataBuffers=this.state.numDataBuffers,t.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(t.unreliable=!0,t.reasons==null&&(t.reasons=[]),t.reasons.push(\"Memory usage by string tensors is approximate (2 bytes per character)\")),t}async profile(t){this.state.profiling=!0;let e=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await t(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(o=>o.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-e,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let o of this.state.activeProfile.kernels)o.kernelTimeMs=await o.kernelTimeMs,o.extraInfo=await o.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(t,e,n,o,s,i){let a={id:this.state.nextTapeNodeId++,kernelName:t,inputs:e,outputs:n,saved:s},u=u0(t);u!=null&&(o=u.gradFunc),o!=null&&(a.gradient=l=>(l=l.map((c,p)=>{if(c==null){let m=n[p],f=ip(m.size,m.dtype);return this.makeTensor(f,m.shape,m.dtype)}return c}),o(l.length>1?l:l[0],s,i))),this.state.activeTape.push(a)}keep(t){return t.kept=!0,t}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(t){let e={track:[],name:\"unnamed scope\",id:this.state.nextScopeId++};t&&(e.name=t),this.state.scopeStack.push(e),this.state.activeScope=e}endScope(t){let e=nh(t),n=new Set(e.map(s=>s.id));for(let s=0;s<this.state.activeScope.track.length;s++){let i=this.state.activeScope.track[s];!i.kept&&!n.has(i.id)&&i.dispose()}let o=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],e.forEach(s=>{!s.kept&&s.scopeId===o.id&&this.track(s)})}gradients(t,e,n,o=!1){if(E(e.length>0,()=>\"gradients() received an empty list of xs.\"),n!=null&&n.dtype!==\"float32\")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let s=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy(\"forward\",t));E(s instanceof Ft,()=>\"The result y returned by f() must be a tensor.\");let i=A1(this.state.activeTape,e,s);if(!o&&i.length===0&&e.length>0)throw new Error(\"Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.\");return this.tidy(\"backward\",()=>{let a={};a[s.id]=n==null?w4(s.shape):n,$1(a,i,l=>this.tidy(l),C4);let u=e.map(l=>a[l.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(l=>{for(let c of l.saved)c.dispose()}),this.state.activeTape=null),{value:s,grads:u}})}customGrad(t){return E(oi(t),()=>\"The f passed in customGrad(f) must be a function.\"),(...e)=>{E(e.every(a=>a instanceof Ft),()=>\"The args passed in customGrad(f)(x1, x2,...) must all be tensors\");let n,o={};e.forEach((a,u)=>{o[u]=a});let s=(a,u)=>(n=t(...e,u),E(n.value instanceof Ft,()=>\"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor\"),E(oi(n.gradFunc),()=>\"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function.\"),n.value),i=(a,u)=>{let l=n.gradFunc(a,u),c=Array.isArray(l)?l:[l];E(c.length===e.length,()=>\"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...).\"),E(c.every(m=>m instanceof Ft),()=>\"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.\");let p={};return c.forEach((m,f)=>{p[f]=()=>m}),p};return this.runKernelFunc({forwardFunc:s,backwardsFunc:i,inputs:o})}}readSync(t){return this.state.tensorInfo.get(t).backend.readSync(t)}read(t){return this.state.tensorInfo.get(t).backend.read(t)}readToGPU(t,e){return this.state.tensorInfo.get(t).backend.readToGPU(t,e)}async time(t){let e=Gu(),n=await this.backend.time(t);return n.wallMs=Gu()-e,n}track(t){return this.state.activeScope!=null&&(t.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(t)),t}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new Ug;for(let t in this.registry)this.disposeRegisteredKernels(t),this.registry[t].dispose(),delete this.registry[t];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};ql.nextTensorId=0;ql.nextVariableId=0;function w4(r){let t=Wd(Jt(r),\"float32\");return k.makeTensor(t,r,\"float32\")}function v0(){let r=l0();if(r._tfengine==null){let t=new qd(r);r._tfengine=new ql(t)}return c1(r._tfengine.ENV),O1(()=>r._tfengine),r._tfengine}var k=v0();function C4(r,t){let e={a:r,b:t};return k.runKernel(Zn,e)}var Kl={};Wt(Kl,{isBrowser:()=>T0,isMobile:()=>v4,mockIsMobile:()=>S4});function I4(){return typeof navigator!=\"undefined\"&&navigator!=null}var N0;function S4(r){N0=r}function v4(r){if(N0!==void 0)return N0;if(r||I4()){if(r||(r=navigator),r.product===\"ReactNative\")return!0;let t=r.userAgent||r.vendor||(typeof window!=\"undefined\"?window.opera:\"\");if(!t){let e=r;return e.userAgentData&&e.userAgentData.mobile}return/(android|bb\\d+|meego).+mobile|avantgo|bada\\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\\-(n|u)|c55\\/|capi|ccwa|cdm\\-|cell|chtm|cldc|cmd\\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\\-s|devi|dica|dmob|do(c|p)o|ds(12|\\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\\-|_)|g1 u|g560|gene|gf\\-5|g\\-mo|go(\\.w|od)|gr(ad|un)|haie|hcit|hd\\-(m|p|t)|hei\\-|hi(pt|ta)|hp( i|ip)|hs\\-c|ht(c(\\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\\-(20|go|ma)|i230|iac( |\\-|\\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\\/)|klon|kpt |kwc\\-|kyo(c|k)|le(no|xi)|lg( g|\\/(k|l|u)|50|54|\\-[a-w])|libw|lynx|m1\\-w|m3ga|m50\\/|ma(te|ui|xo)|mc(01|21|ca)|m\\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\\-2|po(ck|rt|se)|prox|psio|pt\\-g|qa\\-a|qc(07|12|21|32|60|\\-[2-7]|i\\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\\-|oo|p\\-)|sdk\\/|se(c(\\-|0|1)|47|mc|nd|ri)|sgh\\-|shar|sie(\\-|m)|sk\\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\\-|v\\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\\-|tdg\\-|tel(i|m)|tim\\-|t\\-mo|to(pl|sh)|ts(70|m\\-|m3|m5)|tx\\-9|up(\\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\\-|your|zeto|zte\\-/i.test(t.substr(0,4))}return!1}function T0(){return typeof window!=\"undefined\"&&window.document!=null||typeof WorkerGlobalScope!=\"undefined\"}var On=z();On.registerFlag(\"DEBUG\",()=>!1,r=>{r&&console.warn(\"Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.\")});On.registerFlag(\"IS_BROWSER\",()=>T0());On.registerFlag(\"IS_NODE\",()=>typeof process!=\"undefined\"&&typeof process.versions!=\"undefined\"&&typeof process.versions.node!=\"undefined\");On.registerFlag(\"IS_CHROME\",()=>typeof navigator!=\"undefined\"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));On.registerFlag(\"PROD\",()=>!1);On.registerFlag(\"TENSORLIKE_CHECK_SHAPE_CONSISTENCY\",()=>On.getBool(\"DEBUG\"));On.registerFlag(\"DEPRECATION_WARNINGS_ENABLED\",()=>!0);On.registerFlag(\"IS_TEST\",()=>!1);On.registerFlag(\"CHECK_COMPUTATION_FOR_ERRORS\",()=>!0);On.registerFlag(\"WRAP_TO_IMAGEBITMAP\",()=>!1);On.registerFlag(\"ENGINE_COMPILE_ONLY\",()=>!1);On.registerFlag(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\",()=>!1);On.registerFlag(\"USE_SETTIMEOUTCUSTOM\",()=>!1);function Br(r,t){let e=r;if(xr(r))return t===\"string\"?[]:[r.length];if(typeof r==\"object\"&&\"texture\"in r){let o=r.channels||\"RGBA\";return[r.height,r.width*o.length]}if(!Array.isArray(r))return[];let n=[];for(;Array.isArray(e)||xr(e)&&t!==\"string\";)n.push(e.length),e=e[0];return Array.isArray(r)&&z().getBool(\"TENSORLIKE_CHECK_SHAPE_CONSISTENCY\")&&B1(r,n,[]),n}function B1(r,t,e){if(e=e||[],!Array.isArray(r)&&!xr(r)){E(t.length===0,()=>`Element arr[${e.join(\"][\")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}E(t.length>0,()=>`Element arr[${e.join(\"][\")}] should be a primitive, but is an array of ${r.length} elements`),E(r.length===t[0],()=>`Element arr[${e.join(\"][\")}] should have ${t[0]} elements, but has ${r.length} elements`);let n=t.slice(1);for(let o=0;o<r.length;++o)B1(r[o],n,e.concat(o))}function z1(r,t,e,n){if(r!==\"string_or_numeric\"){if(r==null)throw new Error(\"Expected dtype cannot be null.\");if(r!==\"numeric\"&&r!==t||r===\"numeric\"&&t===\"string\")throw new Error(`Argument '${e}' passed to '${n}' must be ${r} tensor, but got ${t} tensor`)}}function I(r,t,e,n=\"numeric\"){if(r instanceof Ft)return z1(n,r.dtype,t,e),r;let o=op(r);if(o!==\"string\"&&[\"bool\",\"int32\",\"float32\"].indexOf(n)>=0&&(o=n),z1(n,o,t,e),r==null||!xr(r)&&!Array.isArray(r)&&typeof r!=\"number\"&&typeof r!=\"boolean\"&&typeof r!=\"string\"){let u=r==null?\"null\":r.constructor.name;throw new Error(`Argument '${t}' passed to '${e}' must be a Tensor or TensorLike, but got '${u}'`)}let s=Br(r,o);!xr(r)&&!Array.isArray(r)&&(r=[r]);let a=o!==\"string\"?Gp(r,o):Bo(r,[],!0);return k.makeTensor(a,s,o)}function ja(r,t,e,n=\"numeric\"){if(!Array.isArray(r))throw new Error(`Argument ${t} passed to ${e} must be a \\`Tensor[]\\` or \\`TensorLike[]\\``);return r.map((s,i)=>I(s,`${t}[${i}]`,e,n))}var k0=\"__op\";function T(r){let t=Object.keys(r);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let e=t[0],n=r[e];e.endsWith(\"_\")&&(e=e.substring(0,e.length-1)),e=e+k0;let o=(...s)=>{k.startScope(e);try{let i=n(...s);return Hd(i)&&console.error(\"Cannot return a Promise inside of tidy.\"),k.endScope(i),i}catch(i){throw k.endScope(null),i}};return Object.defineProperty(o,\"name\",{value:e,configurable:!0}),o}function N4(r,t){let e=I(r,\"real\",\"complex\"),n=I(t,\"imag\",\"complex\");$e(e.shape,n.shape,`real and imag shapes, ${e.shape} and ${n.shape}, must match in call to tf.complex().`);let o={real:e,imag:n};return k.runKernel(pp,o)}var wn=T({complex_:N4});function on(r,t,e,n){if(n==null&&(n=op(r)),n===\"complex64\")throw new Error(\"Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).\");if(typeof r==\"object\"&&\"texture\"in r){if(n!==\"float32\"&&n!==\"int32\")throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${n}.`);return r.channels=r.channels||\"RGBA\",k.backend.createTensorFromTexture(r,t||e,n)}if(!xr(r)&&!Array.isArray(r)&&typeof r!=\"number\"&&typeof r!=\"boolean\"&&typeof r!=\"string\")throw new Error(\"values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray\");if(t!=null){Ud(t);let o=Jt(t),s=Jt(e);E(o===s,()=>`Based on the provided shape, [${t}], the tensor should have ${o} values but has ${s}`);for(let i=0;i<e.length;++i){let a=e[i],u=i===e.length-1?a!==Jt(t.slice(i)):!0;E(e[i]===t[i]||!u,()=>`Error creating a new Tensor. Inferred shape (${e}) does not match the provided shape (${t}). `)}}return!xr(r)&&!Array.isArray(r)&&(r=[r]),t=t||e,r=n!==\"string\"?Gp(r,n):Bo(r,[],!0),k.makeTensor(r,t,n)}function ur(r,t,e){let n=Br(r,e);return on(r,t,n,e)}var oh={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8};var Hg=4;async function G1(r,t){let e=[],n=[],o=Array.isArray(r)?r.map(i=>i.name):Object.keys(r);for(let i=0;i<o.length;++i){let a=o[i],u=Array.isArray(r)?r[i].tensor:r[a];if(u.dtype!==\"float32\"&&u.dtype!==\"int32\"&&u.dtype!==\"bool\"&&u.dtype!==\"string\"&&u.dtype!==\"complex64\")throw new Error(`Unsupported dtype in weight '${a}': ${u.dtype}`);let l={name:a,shape:u.shape,dtype:u.dtype};if(u.dtype===\"string\"){let c=new Promise(async p=>{let m=await u.bytes(),f=m.reduce((g,x)=>g+x.length,0)+Hg*m.length,d=new Uint8Array(f),h=0;for(let g=0;g<m.length;g++){let x=m[g],b=new Uint8Array(new Uint32Array([x.length]).buffer);d.set(b,h),h+=Hg,d.set(x,h),h+=x.length}p(d)});n.push(c)}else n.push(u.data());t!=null&&(l.group=t),e.push(l)}let s=await Promise.all(n);return{data:T4(s),specs:e}}function qg(r,t){let e={},n,o=0;for(let s of t){let i=s.name,a=s.dtype,u=s.shape,l=Jt(u),c;if(\"quantization\"in s){let p=s.quantization;if(p.dtype===\"uint8\"||p.dtype===\"uint16\"){if(!(\"min\"in p&&\"scale\"in p))throw new Error(`Weight ${s.name} with quantization ${p.dtype} doesn't have corresponding metadata min and scale.`)}else if(p.dtype===\"float16\"){if(a!==\"float32\")throw new Error(`Weight ${s.name} is quantized with ${p.dtype} which only supports weights of type float32 not ${a}.`)}else throw new Error(`Weight ${s.name} has unknown quantization dtype ${p.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let m=oh[p.dtype],f=r.slice(o,o+l*m),d=p.dtype===\"uint8\"?new Uint8Array(f):new Uint16Array(f);if(a===\"float32\")if(p.dtype===\"uint8\"||p.dtype===\"uint16\"){c=new Float32Array(d.length);for(let h=0;h<d.length;h++){let g=d[h];c[h]=g*p.scale+p.min}}else if(p.dtype===\"float16\")n===void 0&&(n=A4()),c=n(d);else throw new Error(`Unsupported quantization type ${p.dtype} for weight type float32.`);else if(a===\"int32\"){if(p.dtype!==\"uint8\"&&p.dtype!==\"uint16\")throw new Error(`Unsupported quantization type ${p.dtype} for weight type int32.`);c=new Int32Array(d.length);for(let h=0;h<d.length;h++){let g=d[h];c[h]=Math.round(g*p.scale+p.min)}}else throw new Error(`Unsupported dtype in weight '${i}': ${a}`);o+=l*m}else if(a===\"string\"){let p=Jt(s.shape);c=[];for(let m=0;m<p;m++){let f=new Uint32Array(r.slice(o,o+Hg))[0];o+=Hg;let d=new Uint8Array(r.slice(o,o+f));c.push(d),o+=f}}else{let p=oh[a],m=r.slice(o,o+l*p);if(a===\"float32\")c=new Float32Array(m);else if(a===\"int32\")c=new Int32Array(m);else if(a===\"bool\")c=new Uint8Array(m);else if(a===\"complex64\"){c=new Float32Array(m);let f=new Float32Array(c.length/2),d=new Float32Array(c.length/2);for(let x=0;x<f.length;x++)f[x]=c[x*2],d[x]=c[x*2+1];let h=ur(f,u,\"float32\"),g=ur(d,u,\"float32\");e[i]=wn(h,g),h.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${i}': ${a}`);o+=l*p}a!==\"complex64\"&&(e[i]=ur(c,u,a))}return e}function T4(r){if(r===null)throw new Error(`Invalid input value: ${JSON.stringify(r)}`);let t=0,e=[];r.forEach(s=>{if(t+=s.byteLength,e.push(s.byteLength===s.buffer.byteLength?s:new s.constructor(s)),!(s instanceof Float32Array||s instanceof Int32Array||s instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${s.constructor.name}`)});let n=new Uint8Array(t),o=0;return e.forEach(s=>{n.set(new Uint8Array(s.buffer),o),o+=s.byteLength}),n.buffer}var E0=typeof Buffer!=\"undefined\"&&(typeof Blob==\"undefined\"||typeof atob==\"undefined\"||typeof btoa==\"undefined\");function V1(r){return E0?Buffer.byteLength(r):new Blob([r]).size}function W1(r){if(E0)return Buffer.from(r).toString(\"base64\");let t=new Uint8Array(r),e=\"\";for(let n=0,o=t.length;n<o;n++)e+=String.fromCharCode(t[n]);return btoa(e)}function U1(r){if(E0){let n=Buffer.from(r,\"base64\");return n.buffer.slice(n.byteOffset,n.byteOffset+n.byteLength)}let t=atob(r),e=new Uint8Array(t.length);for(let n=0;n<t.length;++n)e.set([t.charCodeAt(n)],n);return e.buffer}function Hp(r){if(r.length===1)return r[0];let t=0;r.forEach(o=>{t+=o.byteLength});let e=new Uint8Array(t),n=0;return r.forEach(o=>{e.set(new Uint8Array(o),n),n+=o.byteLength}),e.buffer}function _0(r){let t=\"/\";for(r=r.trim();r.endsWith(t);)r=r.slice(0,r.length-1);let e=r.split(t);return e[e.length-1]}function Kg(r,t){let e={modelTopology:r.modelTopology,format:r.format,generatedBy:r.generatedBy,convertedBy:r.convertedBy,weightsManifest:t};return r.signature!=null&&(e.signature=r.signature),r.userDefinedMetadata!=null&&(e.userDefinedMetadata=r.userDefinedMetadata),r.modelInitializer!=null&&(e.modelInitializer=r.modelInitializer),r.initializerSignature!=null&&(e.initializerSignature=r.initializerSignature),r.trainingConfig!=null&&(e.trainingConfig=r.trainingConfig),e}function A0(r,t,e){let n={modelTopology:r.modelTopology,format:r.format,generatedBy:r.generatedBy,convertedBy:r.convertedBy};if(r.trainingConfig!=null&&(n.trainingConfig=r.trainingConfig),r.weightsManifest!=null){if(!t)throw new Error(\"modelJSON has weightsManifest but weightSpecs is null\");if(!e)throw new Error(\"modelJSON has weightsManifest but weightData is null\");n.weightSpecs=t,n.weightData=e}return r.signature!=null&&(n.signature=r.signature),r.userDefinedMetadata!=null&&(n.userDefinedMetadata=r.userDefinedMetadata),r.modelInitializer!=null&&(n.modelInitializer=r.modelInitializer),r.initializerSignature!=null&&(n.initializerSignature=r.initializerSignature),n}async function qp(r,t){let e,n;return r.weightsManifest!=null&&([e,n]=await t(r.weightsManifest)),A0(r,e,n)}function Ni(r){if(r.modelTopology instanceof ArrayBuffer)throw new Error(\"Expected JSON model topology, received ArrayBuffer.\");return{dateSaved:new Date,modelTopologyType:\"JSON\",modelTopologyBytes:r.modelTopology==null?0:V1(JSON.stringify(r.modelTopology)),weightSpecsBytes:r.weightSpecs==null?0:V1(JSON.stringify(r.weightSpecs)),weightDataBytes:r.weightData==null?0:r.weightData.byteLength}}function jg(r){let t=[];for(let e of r)t.push(...e.weights);return t}function k4(){let r=e=>{let n=e<<13,o=0;for(;(n&8388608)===0;)o-=8388608,n<<=1;return n&=-8388609,o+=947912704,n|o},t=new Uint32Array(2048);t[0]=0;for(let e=1;e<1024;e++)t[e]=r(e);for(let e=1024;e<2048;e++)t[e]=939524096+(e-1024<<13);return t}function E4(){let r=new Uint32Array(64);r[0]=0,r[31]=1199570944,r[32]=2147483648,r[63]=3347054592;for(let t=1;t<31;t++)r[t]=t<<23;for(let t=33;t<63;t++)r[t]=2147483648+(t-32<<23);return r}function _4(){let r=new Uint32Array(64);for(let t=0;t<64;t++)r[t]=1024;return r[0]=r[32]=0,r}function A4(){let r=k4(),t=E4(),e=_4();return n=>{let o=new ArrayBuffer(4*n.length),s=new Uint32Array(o);for(let i=0;i<n.length;i++){let a=n[i],u=r[e[a>>10]+(a&1023)]+t[a>>10];s[i]=u}return new Float32Array(o)}}var Ce=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return Ce.instance==null&&(Ce.instance=new Ce),Ce.instance}static registerSaveRouter(t){Ce.getInstance().saveRouters.push(t)}static registerLoadRouter(t){Ce.getInstance().loadRouters.push(t)}static getSaveHandlers(t){return Ce.getHandlers(t,\"save\")}static getLoadHandlers(t,e){return Ce.getHandlers(t,\"load\",e)}static getHandlers(t,e,n){let o=[];return(e===\"load\"?Ce.getInstance().loadRouters:Ce.getInstance().saveRouters).forEach(i=>{let a=i(t,n);a!==null&&o.push(a)}),o}},H1=r=>Ce.registerSaveRouter(r),q1=r=>Ce.registerLoadRouter(r),K1=r=>Ce.getSaveHandlers(r),j1=(r,t)=>Ce.getLoadHandlers(r,t);var $0=\"tensorflowjs\",D0=1,Uu=\"models_store\",jl=\"model_info_store\";function X1(){if(!z().getBool(\"IS_BROWSER\"))throw new Error(\"Failed to obtain IndexedDB factory because the current environmentis not a web browser.\");let r=typeof window==\"undefined\"?self:window,t=r.indexedDB||r.mozIndexedDB||r.webkitIndexedDB||r.msIndexedDB||r.shimIndexedDB;if(t==null)throw new Error(\"The current browser does not appear to support IndexedDB.\");return t}function R0(r){let t=r.result;t.createObjectStore(Uu,{keyPath:\"modelPath\"}),t.createObjectStore(jl,{keyPath:\"modelPath\"})}var Ti=class{constructor(t){if(this.indexedDB=X1(),t==null||!t)throw new Error(\"For IndexedDB, modelPath must not be null, undefined or empty.\");this.modelPath=t}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error(\"BrowserLocalStorage.save() does not support saving model topology in binary formats yet.\");return this.databaseAction(this.modelPath,t)}async load(){return this.databaseAction(this.modelPath)}databaseAction(t,e){return new Promise((n,o)=>{let s=this.indexedDB.open($0,D0);s.onupgradeneeded=()=>R0(s),s.onsuccess=()=>{let i=s.result;if(e==null){let a=i.transaction(Uu,\"readonly\"),l=a.objectStore(Uu).get(this.modelPath);l.onsuccess=()=>{if(l.result==null)return i.close(),o(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));n(l.result.modelArtifacts)},l.onerror=c=>(i.close(),o(l.error)),a.oncomplete=()=>i.close()}else{let a=Ni(e),u=i.transaction(jl,\"readwrite\"),l=u.objectStore(jl),c=l.put({modelPath:this.modelPath,modelArtifactsInfo:a}),p;c.onsuccess=()=>{p=i.transaction(Uu,\"readwrite\");let f=p.objectStore(Uu).put({modelPath:this.modelPath,modelArtifacts:e,modelArtifactsInfo:a});f.onsuccess=()=>n({modelArtifactsInfo:a}),f.onerror=d=>{l=u.objectStore(jl);let h=l.delete(this.modelPath);h.onsuccess=()=>(i.close(),o(f.error)),h.onerror=g=>(i.close(),o(f.error))}},c.onerror=m=>(i.close(),o(c.error)),u.oncomplete=()=>{p==null?i.close():p.oncomplete=()=>i.close()}}},s.onerror=i=>o(s.error)})}};Ti.URL_SCHEME=\"indexeddb://\";var Y1=r=>z().getBool(\"IS_BROWSER\")&&!Array.isArray(r)&&r.startsWith(Ti.URL_SCHEME)?$4(r.slice(Ti.URL_SCHEME.length)):null;Ce.registerSaveRouter(Y1);Ce.registerLoadRouter(Y1);function $4(r){return new Ti(r)}function D4(r){return r.startsWith(Ti.URL_SCHEME)?r.slice(Ti.URL_SCHEME.length):r}var Xg=class{constructor(){this.indexedDB=X1()}async listModels(){return new Promise((t,e)=>{let n=this.indexedDB.open($0,D0);n.onupgradeneeded=()=>R0(n),n.onsuccess=()=>{let o=n.result,s=o.transaction(jl,\"readonly\"),a=s.objectStore(jl).getAll();a.onsuccess=()=>{let u={};for(let l of a.result)u[l.modelPath]=l.modelArtifactsInfo;t(u)},a.onerror=u=>(o.close(),e(a.error)),s.oncomplete=()=>o.close()},n.onerror=o=>e(n.error)})}async removeModel(t){return t=D4(t),new Promise((e,n)=>{let o=this.indexedDB.open($0,D0);o.onupgradeneeded=()=>R0(o),o.onsuccess=()=>{let s=o.result,i=s.transaction(jl,\"readwrite\"),a=i.objectStore(jl),u=a.get(t),l;u.onsuccess=()=>{if(u.result==null)return s.close(),n(new Error(`Cannot find model with path '${t}' in IndexedDB.`));{let c=a.delete(t),p=()=>{l=s.transaction(Uu,\"readwrite\");let f=l.objectStore(Uu).delete(t);f.onsuccess=()=>e(u.result.modelArtifactsInfo),f.onerror=d=>n(u.error)};c.onsuccess=p,c.onerror=m=>(p(),s.close(),n(u.error))}},u.onerror=c=>(s.close(),n(u.error)),i.oncomplete=()=>{l==null?s.close():l.oncomplete=()=>s.close()}},o.onerror=s=>n(o.error)})}};var Xa=\"/\",Kp=\"tensorflowjs_models\",Z1=\"info\",R4=\"model_topology\",F4=\"weight_specs\",O4=\"weight_data\",P4=\"model_metadata\";function J1(r){return{info:[Kp,r,Z1].join(Xa),topology:[Kp,r,R4].join(Xa),weightSpecs:[Kp,r,F4].join(Xa),weightData:[Kp,r,O4].join(Xa),modelMetadata:[Kp,r,P4].join(Xa)}}function Q1(r){for(let t of Object.values(r))window.localStorage.removeItem(t)}function L4(r){let t=r.split(Xa);if(t.length<3)throw new Error(`Invalid key format: ${r}`);return t.slice(1,t.length-1).join(Xa)}function M4(r){return r.startsWith(ki.URL_SCHEME)?r.slice(ki.URL_SCHEME.length):r}var ki=class{constructor(t){if(!z().getBool(\"IS_BROWSER\")||typeof window==\"undefined\"||typeof window.localStorage==\"undefined\")throw new Error(\"The current environment does not support local storage.\");if(this.LS=window.localStorage,t==null||!t)throw new Error(\"For local storage, modelPath must not be null, undefined or empty.\");this.modelPath=t,this.keys=J1(this.modelPath)}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error(\"BrowserLocalStorage.save() does not support saving model topology in binary formats yet.\");{let e=JSON.stringify(t.modelTopology),n=JSON.stringify(t.weightSpecs),o=Ni(t);try{this.LS.setItem(this.keys.info,JSON.stringify(o)),this.LS.setItem(this.keys.topology,e),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,W1(t.weightData));let s={format:t.format,generatedBy:t.generatedBy,convertedBy:t.convertedBy,signature:t.signature!=null?t.signature:void 0,userDefinedMetadata:t.userDefinedMetadata!=null?t.userDefinedMetadata:void 0,modelInitializer:t.modelInitializer!=null?t.modelInitializer:void 0,initializerSignature:t.initializerSignature!=null?t.initializerSignature:void 0,trainingConfig:t.trainingConfig!=null?t.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(s)),{modelArtifactsInfo:o}}catch(s){throw Q1(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${o.modelTopologyBytes}, weightSpecsBytes=${o.weightSpecsBytes}, weightDataBytes=${o.weightDataBytes}.`)}}}async load(){let t=JSON.parse(this.LS.getItem(this.keys.info));if(t==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(t.modelTopologyType!==\"JSON\")throw new Error(\"BrowserLocalStorage does not support loading non-JSON model topology yet.\");let e={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(n==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);e.modelTopology=n;let o=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(o==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);e.weightSpecs=o;let s=this.LS.getItem(this.keys.modelMetadata);if(s!=null){let a=JSON.parse(s);e.format=a.format,e.generatedBy=a.generatedBy,e.convertedBy=a.convertedBy,a.signature!=null&&(e.signature=a.signature),a.userDefinedMetadata!=null&&(e.userDefinedMetadata=a.userDefinedMetadata),a.modelInitializer!=null&&(e.modelInitializer=a.modelInitializer),a.initializerSignature!=null&&(e.initializerSignature=a.initializerSignature),a.trainingConfig!=null&&(e.trainingConfig=a.trainingConfig)}let i=this.LS.getItem(this.keys.weightData);if(i==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return e.weightData=U1(i),e}};ki.URL_SCHEME=\"localstorage://\";var tE=r=>z().getBool(\"IS_BROWSER\")&&!Array.isArray(r)&&r.startsWith(ki.URL_SCHEME)?z4(r.slice(ki.URL_SCHEME.length)):null;Ce.registerSaveRouter(tE);Ce.registerLoadRouter(tE);function z4(r){return new ki(r)}var Yg=class{constructor(){E(z().getBool(\"IS_BROWSER\"),()=>\"Current environment is not a web browser\"),E(typeof window==\"undefined\"||typeof window.localStorage!=\"undefined\",()=>\"Current browser does not appear to support localStorage\"),this.LS=window.localStorage}async listModels(){let t={},e=Kp+Xa,n=Xa+Z1;for(let o=0;o<this.LS.length;++o){let s=this.LS.key(o);if(s.startsWith(e)&&s.endsWith(n)){let i=L4(s);t[i]=JSON.parse(this.LS.getItem(s))}}return t}async removeModel(t){t=M4(t);let e=J1(t);if(this.LS.getItem(e.info)==null)throw new Error(`Cannot find model at path '${t}'`);let n=JSON.parse(this.LS.getItem(e.info));return Q1(e),n}};var jp=\"://\",Er=class{constructor(){this.managers={}}static getInstance(){return Er.instance==null&&(Er.instance=new Er),Er.instance}static registerManager(t,e){E(t!=null,()=>\"scheme must not be undefined or null.\"),t.endsWith(jp)&&(t=t.slice(0,t.indexOf(jp))),E(t.length>0,()=>\"scheme must not be an empty string.\");let n=Er.getInstance();E(n.managers[t]==null,()=>`A model store manager is already registered for scheme '${t}'.`),n.managers[t]=e}static getManager(t){let e=Er.getInstance().managers[t];if(e==null)throw new Error(`Cannot find model manager for scheme '${t}'`);return e}static getSchemes(){return Object.keys(Er.getInstance().managers)}};function Zg(r){if(r.indexOf(jp)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Er.getSchemes().join(\",\")}`);return{scheme:r.split(jp)[0],path:r.split(jp)[1]}}async function eE(r,t,e=!1){E(r!==t,()=>`Old path and new path are the same: '${r}'`);let n=Ce.getLoadHandlers(r);E(n.length>0,()=>`Copying failed because no load handler is found for source URL ${r}.`),E(n.length<2,()=>`Copying failed because more than one (${n.length}) load handlers for source URL ${r}.`);let o=n[0],s=Ce.getSaveHandlers(t);E(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),E(s.length<2,()=>`Copying failed because more than one (${n.length}) save handlers for destination URL ${t}.`);let i=s[0],a=Zg(r).scheme,u=Zg(r).path,l=a===Zg(r).scheme,c=await o.load();e&&l&&await Er.getManager(a).removeModel(u);let p=await i.save(c);return e&&!l&&await Er.getManager(a).removeModel(u),p.modelArtifactsInfo}async function rE(){let r=Er.getSchemes(),t={};for(let e of r){let n=await Er.getManager(e).listModels();for(let o in n){let s=e+jp+o;t[s]=n[o]}}return t}async function nE(r){let t=Zg(r);return Er.getManager(t.scheme).removeModel(t.path)}async function oE(r,t){return eE(r,t,!1)}async function sE(r,t){return eE(r,t,!0)}var F0=class{constructor(){this.messageName=\"setTimeoutCustom\",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(t,e){return fetch(t,e)}now(){return performance.now()}encode(t,e){if(e!==\"utf-8\"&&e!==\"utf8\")throw new Error(`Browser's encoder only supports utf-8, but got ${e}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(t)}decode(t,e){return new TextDecoder(e).decode(t)}setTimeoutCustom(t,e){if(typeof window==\"undefined\"||!z().getBool(\"USE_SETTIMEOUTCUSTOM\")){setTimeout(t,e);return}this.functionRefs.push(t),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},\"*\")},e),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener(\"message\",n=>{if(n.source===window&&n.data.name===this.messageName){n.stopPropagation();let o=this.functionRefs[n.data.index];o(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0)}},!0))}};if(z().get(\"IS_BROWSER\")){z().setPlatform(\"browser\",new F0);try{Er.registerManager(ki.URL_SCHEME,new Yg)}catch(r){}try{Er.registerManager(Ti.URL_SCHEME,new Xg)}catch(r){}}var B4={importFetch:()=>iE()},O0;var P0=class{constructor(){this.util=aE(),this.textEncoder=new this.util.TextEncoder}fetch(t,e){return z().global.fetch!=null?z().global.fetch(t,e):(O0==null&&(O0=B4.importFetch()),O0(t,e))}now(){let t=process.hrtime();return t[0]*1e3+t[1]/1e6}encode(t,e){if(e!==\"utf-8\"&&e!==\"utf8\")throw new Error(`Node built-in encoder only supports utf-8, but got ${e}`);return this.textEncoder.encode(t)}decode(t,e){return t.length===0?\"\":new this.util.TextDecoder(e).decode(t)}};z().get(\"IS_NODE\")&&!z().get(\"IS_BROWSER\")&&z().setPlatform(\"node\",new P0);function wt(r,t=\"float32\",e){return t=t||\"float32\",Ud(r),new pe(r,t,e)}function V4(r,t){let e=I(r,\"x\",\"cast\");if(!o0(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t===\"string\"&&e.dtype!==\"string\"||t!==\"string\"&&e.dtype===\"string\")throw new Error(\"Only strings can be casted to strings\");let n={x:e},o={dtype:t};return k.runKernel(lo,n,o)}var J=T({cast_:V4});function G4(r){let e={x:I(r,\"x\",\"clone\",\"string_or_numeric\")};return k.runKernel(co,e)}var sn=T({clone_:G4});function Jg(r,t=!1){console.log(r.toString(t))}v0();var W4={buffer:wt,cast:J,clone:sn,print:Jg};P1(W4);var _r={};Wt(_r,{browserFiles:()=>uE,browserHTTPRequest:()=>mE,concatenateArrayBuffers:()=>Hp,copyModel:()=>oE,decodeWeights:()=>qg,encodeWeights:()=>G1,fromMemory:()=>fE,fromMemorySync:()=>G0,getLoadHandlers:()=>j1,getModelArtifactsForJSON:()=>qp,getModelArtifactsForJSONSync:()=>A0,getModelArtifactsInfoForJSON:()=>Ni,getSaveHandlers:()=>K1,getWeightSpecs:()=>jg,http:()=>tx,isHTTPScheme:()=>Qg,listModels:()=>rE,loadWeights:()=>cE,moveModel:()=>sE,registerLoadRouter:()=>q1,registerSaveRouter:()=>H1,removeModel:()=>nE,weightsLoaderFactory:()=>B0,withSaveHandler:()=>dE,withSaveHandlerSync:()=>hE});var U4=\"model\",H4=\".json\",q4=\".weights.bin\";function lE(r){return new Promise(t=>setTimeout(t)).then(r)}var Ya=class{constructor(t){if(!z().getBool(\"IS_BROWSER\"))throw new Error(\"browserDownloads() cannot proceed because the current environment is not a browser.\");t.startsWith(Ya.URL_SCHEME)&&(t=t.slice(Ya.URL_SCHEME.length)),(t==null||t.length===0)&&(t=U4),this.modelJsonFileName=t+H4,this.weightDataFileName=t+q4}async save(t){if(typeof document==\"undefined\")throw new Error(\"Browser downloads are not supported in this environment since `document` is not present\");let e=window.URL.createObjectURL(new Blob([t.weightData],{type:\"application/octet-stream\"}));if(t.modelTopology instanceof ArrayBuffer)throw new Error(\"BrowserDownloads.save() does not support saving model topology in binary formats yet.\");{let n=[{paths:[\"./\"+this.weightDataFileName],weights:t.weightSpecs}],o=Kg(t,n),s=window.URL.createObjectURL(new Blob([JSON.stringify(o)],{type:\"application/json\"})),i=this.modelJsonAnchor==null?document.createElement(\"a\"):this.modelJsonAnchor;if(i.download=this.modelJsonFileName,i.href=s,await lE(()=>i.dispatchEvent(new MouseEvent(\"click\"))),t.weightData!=null){let a=this.weightDataAnchor==null?document.createElement(\"a\"):this.weightDataAnchor;a.download=this.weightDataFileName,a.href=e,await lE(()=>a.dispatchEvent(new MouseEvent(\"click\")))}return{modelArtifactsInfo:Ni(t)}}}};Ya.URL_SCHEME=\"downloads://\";var L0=class{constructor(t){if(t==null||t.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${t}`);this.jsonFile=t[0],this.weightsFiles=t.slice(1)}async load(){return new Promise((t,e)=>{let n=new FileReader;n.onload=o=>{let s=JSON.parse(o.target.result),i=s.modelTopology;if(i==null){e(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(s.weightsManifest==null){e(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){t({modelTopology:i});return}let u=qp(s,l=>this.loadWeights(l));t(u)},n.onerror=o=>e(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(t){let e=[],n=[];for(let i of t)e.push(...i.weights),n.push(...i.paths);let o=this.checkManifestAndWeightFiles(t),s=n.map(i=>this.loadWeightsFile(i,o[i]));return Promise.all(s).then(i=>[e,Hp(i)])}loadWeightsFile(t,e){return new Promise((n,o)=>{let s=new FileReader;s.onload=i=>{let a=i.target.result;n(a)},s.onerror=i=>o(`Failed to weights data from file of path '${t}'.`),s.readAsArrayBuffer(e)})}checkManifestAndWeightFiles(t){let e=[],n=this.weightsFiles.map(s=>_0(s.name)),o={};for(let s of t)s.paths.forEach(i=>{let a=_0(i);if(e.indexOf(a)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${a}'`);if(e.push(a),n.indexOf(a)===-1)throw new Error(`Weight file with basename '${a}' is not provided.`);o[i]=this.weightsFiles[n.indexOf(a)]});if(e.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${e.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return o}},K4=r=>z().getBool(\"IS_BROWSER\")&&!Array.isArray(r)&&r.startsWith(Ya.URL_SCHEME)?j4(r.slice(Ya.URL_SCHEME.length)):null;Ce.registerSaveRouter(K4);function j4(r=\"model\"){return new Ya(r)}function uE(r){return new L0(r)}function M0(r,t,e,n){i(r),e=e==null?0:e,n=n==null?1:n,a(e,n);let o=0,s=u=>(u.then(l=>{let c=e+ ++o/r.length*(n-e);return t(c),l}),u);function i(u){E(u!=null&&Array.isArray(u)&&u.length>0,()=>\"promises must be a none empty array\")}function a(u,l){E(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${u}`),E(l>=0&&l<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${l}`),E(l>=u,()=>`startFraction must be no more than endFraction, but got startFraction ${u} and endFraction ${l}`)}return Promise.all(r.map(s))}async function z0(r,t){t==null&&(t={});let e=t.fetchFunc==null?z().platform.fetch:t.fetchFunc,n=r.map(p=>e(p,t.requestInit,{isBinary:!0})),o=0,s=.5,a=(t.onProgress==null?await Promise.all(n):await M0(n,t.onProgress,o,s)).map(p=>p.arrayBuffer()),u=.5,l=1;return t.onProgress==null?await Promise.all(a):await M0(a,t.onProgress,u,l)}async function cE(r,t=\"\",e,n){return B0(i=>z0(i,{requestInit:n}))(r,t,e)}function B0(r){return async(t,e=\"\",n)=>{let o=t.map(()=>!1),s={},i=n!=null?n.map(()=>!1):[],a=[];if(t.forEach((f,d)=>{let h=0;f.weights.forEach(g=>{let x=\"quantization\"in g?g.quantization.dtype:g.dtype,b=oh[x]*Jt(g.shape),w=()=>{o[d]=!0,s[d]==null&&(s[d]=[]),s[d].push({manifestEntry:g,groupOffset:h,sizeBytes:b})};n!=null?n.forEach((C,N)=>{C===g.name&&(w(),i[N]=!0)}):w(),a.push(g.name),h+=b})}),!i.every(f=>f)){let f=n.filter((d,h)=>!i[h]);throw new Error(`Could not find weights in manifest with names: ${f.join(\", \")}. \nManifest JSON has weights with names: ${a.join(\", \")}.`)}let u=o.reduce((f,d,h)=>(d&&f.push(h),f),[]),l=[];u.forEach(f=>{t[f].paths.forEach(d=>{let h=e+(e.endsWith(\"/\")?\"\":\"/\")+d;l.push(h)})});let c=await r(l),p={},m=0;return u.forEach(f=>{let d=t[f].paths.length,h=0;for(let C=0;C<d;C++)h+=c[m+C].byteLength;let g=new ArrayBuffer(h),x=new Uint8Array(g),b=0;for(let C=0;C<d;C++){let N=new Uint8Array(c[m+C]);x.set(N,b),b+=N.byteLength}s[f].forEach(C=>{let N=g.slice(C.groupOffset,C.groupOffset+C.sizeBytes),_=qg(N,[C.manifestEntry]);for(let A in _)p[A]=_[A]}),m+=d}),p}}var X4=\"application/octet-stream\",Y4=\"application/json\",sh=class{constructor(t,e){if(this.DEFAULT_METHOD=\"POST\",e==null&&(e={}),this.weightPathPrefix=e.weightPathPrefix,this.onProgress=e.onProgress,this.weightUrlConverter=e.weightUrlConverter,e.fetchFunc!=null?(E(typeof e.fetchFunc==\"function\",()=>\"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)\"),this.fetch=e.fetchFunc):this.fetch=z().platform.fetch,E(t!=null&&t.length>0,()=>\"URL path for http must not be null, undefined or empty.\"),Array.isArray(t)&&E(t.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${t.length}).`),this.path=t,e.requestInit!=null&&e.requestInit.body!=null)throw new Error(\"requestInit is expected to have no pre-existing body, but has one.\");this.requestInit=e.requestInit||{}}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error(\"BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.\");let e=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);e.body=new FormData;let n=[{paths:[\"./model.weights.bin\"],weights:t.weightSpecs}],o=Kg(t,n);e.body.append(\"model.json\",new Blob([JSON.stringify(o)],{type:Y4}),\"model.json\"),t.weightData!=null&&e.body.append(\"model.weights.bin\",new Blob([t.weightData],{type:X4}),\"model.weights.bin\");let s=await this.fetch(this.path,e);if(s.ok)return{modelArtifactsInfo:Ni(t),responses:[s]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${s.status}.`)}async load(){let t=await this.fetch(this.path,this.requestInit);if(!t.ok)throw new Error(`Request to ${this.path} failed with status code ${t.status}. Please verify this URL points to the model JSON of the model to load.`);let e;try{e=await t.json()}catch(s){let i=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(\".pb\")?i+=\" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.\":i+=\" Please make sure the server is serving valid JSON for this request.\",new Error(i)}let n=e.modelTopology,o=e.weightsManifest;if(n==null&&o==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return qp(e,s=>this.loadWeights(s))}async loadWeights(t){let e=Array.isArray(this.path)?this.path[1]:this.path,[n,o]=Z4(e),s=this.weightPathPrefix||n,i=jg(t),a=[],u=[];for(let c of t)for(let p of c.paths)this.weightUrlConverter!=null?u.push(this.weightUrlConverter(p)):a.push(s+p+o);this.weightUrlConverter&&a.push(...await Promise.all(u));let l=await z0(a,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[i,Hp(l)]}};sh.URL_SCHEME_REGEX=/^https?:\\/\\//;function Z4(r){let t=r.lastIndexOf(\"/\"),e=r.lastIndexOf(\"?\"),n=r.substring(0,t),o=e>t?r.substring(e):\"\";return[n+\"/\",o]}function Qg(r){return r.match(sh.URL_SCHEME_REGEX)!=null}var pE=(r,t)=>{if(typeof fetch==\"undefined\"&&(t==null||t.fetchFunc==null))return null;{let e=!0;if(Array.isArray(r)?e=r.every(n=>Qg(n)):e=Qg(r),e)return tx(r,t)}return null};Ce.registerSaveRouter(pE);Ce.registerLoadRouter(pE);function tx(r,t){return new sh(r,t)}function mE(r,t){return tx(r,t)}var ih=class{constructor(t){this.modelArtifacts=t}load(){return this.modelArtifacts}},ex=class{constructor(t){this.saveHandler=t}save(t){return this.saveHandler(t)}},V0=class{constructor(t){t.load&&(this.load=()=>Promise.resolve(t.load())),t.save&&(this.save=e=>Promise.resolve(t.save(e)))}};function fE(r,t,e,n){let o=arguments;return new V0(G0(...o))}function G0(r,t,e,n){return arguments.length===1?r.modelTopology!=null||r.weightSpecs!=null?new ih(r):(console.warn(\"Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.\"),new ih({modelTopology:r})):(console.warn(\"Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.\"),new ih({modelTopology:r,weightSpecs:t,weightData:e,trainingConfig:n}))}function dE(r){return new ex(r)}function hE(r){return new ex(r)}var yE={};Wt(yE,{confusionMatrix:()=>xE});function J4(r,t,e=!1,n=!1){let o=I(r,\"a\",\"matMul\"),s=I(t,\"b\",\"matMul\");[o,s]=Ut(o,s);let i={a:o,b:s},a={transposeA:e,transposeB:n};return k.runKernel(Ho,i,a)}var Lt=T({matMul_:J4});function Q4(r,t,e=1,n=0,o=\"int32\"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let i={indices:I(r,\"indices\",\"oneHot\",\"int32\")},a={dtype:o,depth:t,onValue:e,offValue:n};return k.runKernel(gs,i,a)}var Ei=T({oneHot_:Q4});function dpt(){z().set(\"PROD\",!0)}function hpt(){z().set(\"DEBUG\",!0)}function gpt(){z().set(\"DEPRECATION_WARNINGS_ENABLED\",!1),console.warn(\"TensorFlow.js deprecation warnings have been disabled.\")}function W0(r){z().getBool(\"DEPRECATION_WARNINGS_ENABLED\")&&console.warn(r+\" You can disable deprecation warnings with tf.disableDeprecationWarnings().\")}L1(W0);function xpt(){k.disposeVariables()}function Pn(){return k}function ah(){return k.memory()}function ypt(r){return k.profile(r)}function B(r,t){return k.tidy(r,t)}function vt(r){nh(r).forEach(e=>e.dispose())}function De(r){return k.keep(r)}function bpt(r){return k.time(r)}function tH(r){return k.setBackend(r)}function wpt(){return k.ready()}function Cpt(){return k.backendName}function Ipt(r){k.removeBackend(r)}function Spt(r){return k.findBackend(r)}function vpt(r){return k.findBackendFactory(r)}function Xp(r,t,e=1){return k.registerBackend(r,t,e)}function gE(){return k.backend}function Npt(r,t){z().setPlatform(r,t)}function eH(r){let e={input:I(r,\"input\",\"imag\")};return k.runKernel(Sp,e)}var Xl=T({imag_:eH});function rH(r){let e={x:I(r,\"x\",\"neg\")};return k.runKernel(pi,e)}var Ht=T({neg_:rH});function nH(r){let e={input:I(r,\"input\",\"real\")};return k.runKernel(Rp,e)}var Za=T({real_:nH});function oH(r,t,e){let n=I(r,\"x\",\"transpose\");if(t==null&&(t=n.shape.map((i,a)=>a).reverse()),E(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of perm ${t}.`),t.forEach(i=>{E(i>=0&&i<n.rank,()=>`All entries in 'perm' must be between 0 and ${n.rank-1} but got ${t}`)}),n.rank<=1)return n.clone();let o={x:n},s={perm:t};return n.dtype===\"complex64\"?B(()=>{let i=Za(n),a=Xl(n);return i=k.runKernel(Qn,{x:i},s),a=k.runKernel(Qn,{x:a},s),e&&(a=Ht(a)),wn(i,a)}):k.runKernel(Qn,o,s)}var Ot=T({transpose_:oH});function sH(r,t,e){let n=I(r,\"labels\",\"confusionMatrix\"),o=I(t,\"predictions\",\"confusionMatrix\");E(e==null||e>0&&Number.isInteger(e),()=>`If provided, numClasses must be a positive integer, but got ${e}`),E(n.rank===1,()=>`Expected the rank of labels to be 1, but got ${n.rank}`),E(o.rank===1,()=>`Expected the rank of predictions to be 1, but got ${o.rank}`),E(n.shape[0]===o.shape[0],()=>`Mismatch in the number of examples: ${n.shape[0]} vs. ${o.shape[0]}. Labels and predictions should have the same number of elements.`),E(e>0&&Number.isInteger(e),()=>`numClasses is required to be a positive integer, but got ${e}`);let s=Ei(J(n,\"int32\"),e),i=Ei(J(o,\"int32\"),e),a=Ot(s),u=Lt(a,i);return J(u,\"int32\")}var xE=T({confusionMatrix_:sH});var Vr={};Wt(Vr,{assertAndGetBroadcastShape:()=>Pt,getBroadcastDims:()=>bE,getReductionAxes:()=>ge});function bE(r,t){let e=r.length,n=[];for(let o=0;o<e;o++){let s=e-1-o,i=r[s]||1;(t[t.length-1-o]||1)>1&&i===1&&n.unshift(s)}return n}function ge(r,t){let e=[];for(let n=0;n<t.length;n++){let o=r[r.length-n-1],s=t.length-n-1,i=t[s];(o==null||o===1&&i>1)&&e.unshift(s)}return e}function Pt(r,t){let e=[],n=Math.max(r.length,t.length);for(let o=0;o<n;o++){let s=r[r.length-o-1];s==null&&(s=1);let i=t[t.length-o-1];if(i==null&&(i=1),s===1)e.unshift(i);else if(i===1)e.unshift(s);else if(s!==i){let a=`Operands could not be broadcast together with shapes ${r} and ${t}.`;throw Error(a)}else e.unshift(s)}return e}var nx={};Wt(nx,{fromPixels:()=>mH,fromPixelsAsync:()=>cH,toPixels:()=>pH});function rx(r,t,e){if(Yn(r),t!=null&&t.length!==3)throw new Error(\"tensor3d() requires shape to have three numbers\");let n=Br(r,e);if(n.length!==3&&n.length!==1)throw new Error(\"tensor3d() requires values to be number[][][] or flat/TypedArray\");if(n.length===1&&t==null)throw new Error(\"tensor3d() requires shape to be provided when `values` are a flat array\");return on(r,t,n,e)}var Hu;function wE(r,t=3){if(t>4)throw new Error(\"Cannot construct Tensor with more than 4 channels from pixels.\");if(r==null)throw new Error(\"pixels passed to tf.browser.fromPixels() can not be null\");let e=!1,n=!1,o=!1,s=!1,i=!1,a=!1;if(r.data instanceof Uint8Array)e=!0;else if(typeof ImageData!=\"undefined\"&&r instanceof ImageData)n=!0;else if(typeof HTMLVideoElement!=\"undefined\"&&r instanceof HTMLVideoElement)o=!0;else if(typeof HTMLImageElement!=\"undefined\"&&r instanceof HTMLImageElement)s=!0;else if(r.getContext!=null)i=!0;else if(typeof ImageBitmap!=\"undefined\"&&r instanceof ImageBitmap)a=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${r.constructor.name}`);if(Jd(Yd,k.backendName)!=null){let d={pixels:r},h={numChannels:t};return k.runKernel(Yd,d,h)}let[l,c]=o?[r.videoWidth,r.videoHeight]:[r.width,r.height],p;if(i)p=r.getContext(\"2d\").getImageData(0,0,l,c).data;else if(n||e)p=r.data;else if(s||o||a){if(Hu==null)if(typeof document==\"undefined\")if(typeof OffscreenCanvas!=\"undefined\"&&typeof OffscreenCanvasRenderingContext2D!=\"undefined\")Hu=new OffscreenCanvas(1,1).getContext(\"2d\");else throw new Error(\"Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.\");else Hu=document.createElement(\"canvas\").getContext(\"2d\",{willReadFrequently:!0});Hu.canvas.width=l,Hu.canvas.height=c,Hu.drawImage(r,0,0,l,c),p=Hu.getImageData(0,0,l,c).data}let m;if(t===4)m=new Int32Array(p);else{let d=l*c;m=new Int32Array(d*t);for(let h=0;h<d;h++)for(let g=0;g<t;++g)m[h*t+g]=p[h*4+g]}return rx(m,[c,l,t],\"int32\")}function iH(r){return r!=null&&r.data instanceof Uint8Array}function aH(){return typeof window!=\"undefined\"&&typeof ImageBitmap!=\"undefined\"&&window.hasOwnProperty(\"createImageBitmap\")}function lH(r){return r!=null&&r.width!==0&&r.height!==0}function uH(r){return aH()&&!(r instanceof ImageBitmap)&&lH(r)&&!iH(r)}async function cH(r,t=3){let e=null;if(z().getBool(\"WRAP_TO_IMAGEBITMAP\")&&uH(r)){let n;try{n=await createImageBitmap(r,{premultiplyAlpha:\"none\"})}catch(o){n=null}n!=null&&n.width===r.width&&n.height===r.height?e=n:e=r}else e=r;return wE(e,t)}async function pH(r,t){let e=I(r,\"img\",\"toPixels\");if(!(r instanceof Ft)){let l=e;e=J(l,\"int32\"),l.dispose()}if(e.rank!==2&&e.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${e.rank}.`);let[n,o]=e.shape.slice(0,2),s=e.rank===2?1:e.shape[2];if(s>4||s===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${s}`);if(e.dtype!==\"float32\"&&e.dtype!==\"int32\")throw new Error(`Unsupported type for toPixels: ${e.dtype}. Please use float32 or int32 tensors.`);let i=await e.data(),a=e.dtype===\"float32\"?255:1,u=new Uint8ClampedArray(o*n*4);for(let l=0;l<n*o;++l){let c=[0,0,0,255];for(let m=0;m<s;m++){let f=i[l*s+m];if(e.dtype===\"float32\"){if(f<0||f>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${f}.`)}else if(e.dtype===\"int32\"&&(f<0||f>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${f}.`);s===1?(c[0]=f*a,c[1]=f*a,c[2]=f*a):c[m]=f*a}let p=l*4;u[p+0]=Math.round(c[0]),u[p+1]=Math.round(c[1]),u[p+2]=Math.round(c[2]),u[p+3]=Math.round(c[3])}if(t!=null){t.width=o,t.height=n;let l=t.getContext(\"2d\"),c=new ImageData(u,o,n);l.putImageData(c,0,0)}return e!==r&&e.dispose(),u}var mH=T({fromPixels_:wE});var ox={};Wt(ox,{prepareAndValidate:()=>CE});function CE(r,t){let e=r.shape.length,n=t.shape.length;if(e<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${e}.`);if(n<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${n}.`);if(t.dtype!==\"int32\")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[n-1]>e)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[n-1]} vs. ${e}`);if(Jt(r.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${r.shape}.`);let o=t.shape,s=o[o.length-1],i=1;for(let p=0;p<o.length-1;++p)i*=o[p];let a=r.shape,u=o.slice();u.pop();let l=1;for(let p=s;p<e;++p)l*=a[p],u.push(a[p]);let c=[...si(r.shape).map(p=>p/l),1].slice(0,s);return[u,i,l,c]}var lh={};Wt(lh,{calculateShapes:()=>IE,validateInput:()=>sx,validateUpdateShape:()=>U0});function U0(r,t,e){let n=t.rank>1?t.shape[t.rank-1]:1,o=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${e.shape}, indices.shape: ${t.shape}, shape: ${r}, sliceDim: ${n}, and batchDim: ${o}.`;if(e.rank<o)throw new Error(s+` update.rank < ${o}. `);if(r.length<n+(e.rank-o))throw new Error(s+` Output shape length < ${n+(e.rank-o)}`);if(e.rank!==o+r.length-n)throw new Error(s+` update.rank != ${o+r.length-n}`);for(let i=0;i<o;++i)if(e.shape[i]!==t.shape[i])throw new Error(s+` updates.shape[${i}] (${e.shape[i]}) != indices.shape[${i}] (${t.shape[i]}).`);for(let i=0;i<e.rank-o;++i)if(e.shape[i+o]!==r[i+n])throw new Error(s+` updates.shape[${i+o}] (${e.shape[i+o]}) != shape[${i+o}] (${r[i+o]})`)}function sx(r,t,e){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(r.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${r.rank}.`);if(t.dtype!==\"int32\")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(e.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${e}`);if(e.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(r.size===0)throw new Error(`Updates specified for empty output. updates shape: ${r.shape}`)}U0(e,t,r)}function IE(r,t,e){let n=t.shape.length,o=n>1?t.shape[n-1]:1,s=e.length,i=1;for(let p=o;p<s;++p)i*=e[p];let a=o<1?1:o,u=Jt(t.shape)/a,l=[...si(e.slice(0,o)),1],c=Jt(e);return{sliceRank:o,numUpdates:u,sliceSize:i,strides:l,outputSize:c}}var Le={};Wt(Le,{assertParamsValid:()=>dH,computeFlatOffset:()=>bH,computeOutShape:()=>gH,getNormalizedAxes:()=>xH,isSliceContinous:()=>yH,maskToAxes:()=>hH,parseSliceParams:()=>q0,sliceInfo:()=>wH,startForAxis:()=>AE,startIndicesWithElidedDims:()=>kE,stopForAxis:()=>$E,stopIndicesWithElidedDims:()=>EE,stridesForAxis:()=>_E,stridesWithElidedDims:()=>vE});var H0=-2,fH=-1;function dH(r,t,e){let n=r.shape.length;E(n===t.length,()=>`Error in slice${n}D: Length of begin ${t} must match the rank of the array (${n}).`),E(n===e.length,()=>`Error in slice${n}D: Length of size ${e} must match the rank of the array (${n}).`);for(let o=0;o<n;++o)E(t[o]+e[o]<=r.shape[o],()=>`Error in slice${n}D: begin[${o}] + size[${o}] (${t[o]+e[o]}) would overflow input.shape[${o}] (${r.shape[o]})`)}function hH(r){let t=[],e=0;for(;r>0;)r&1&&t.push(e),r/=2,e++;return t}function gH(r,t,e){let n=[];for(let o=0;o<r.length;o++)n[o]=Math.ceil((t[o]-r[o])/e[o]);return n}function vE(r,t,e,n){let o=[...r];for(let s=o.length;s<n.length;s++)o.push(1);for(let s=0;s<e;s++)s===0?o[t]=1:(o.splice(t,0,1),o.pop());return o}function NE(r,t,e){return e<=r?e:e-(t-1)}function TE(r,t){let e=[];for(let n=0;n<r;n++)e.push(t+n);return e}function xH(r,t,e,n,o,s,i,a,u){let l=r.length,c=new Array(l),p=new Array(l),m=new Array(l);if(t.length&&e>0){let f=t[0],d=e+1;c=kE(i,f,d,n,r),p=EE(a,f,d,o,r),m=vE(s,f,d,r)}else for(let f=0;f<l;f++)c[f]=AE(i,n,s,r,f,u),p[f]=$E(a,o,s,r,f,u),m[f]=_E(s,f,u);return{begin:c,end:p,strides:m}}function kE(r,t,e,n,o){let s=[...o],i=TE(e,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=0;else{let u=NE(t,e,a),l=n[u];r&1<<u&&(l=0),s[a]=l}return s}function EE(r,t,e,n,o){let s=[...o],i=TE(e,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=Number.MAX_SAFE_INTEGER;else{let u=NE(t,e,a),l=n[u];r&1<<u&&(l=Number.MAX_SAFE_INTEGER),s[a]=l}for(let a=0;a<s.length;a++){let u=o[a];s[a]<0&&(s[a]+=u),s[a]=np(0,s[a],o[a])}return s}function _E(r,t,e){let n=r[t];return(e&1<<t||n==null)&&(n=1),n}function AE(r,t,e,n,o,s){let i=t[o],a=e[o]||1;(r&1<<o||s&1<<o||i==null)&&(a>0?i=Number.MIN_SAFE_INTEGER:i=Number.MAX_SAFE_INTEGER);let u=n[o];return i<0&&(i+=u),i=np(0,i,u-1),i}function $E(r,t,e,n,o,s){let i=t[o],a=e[o]||1;(r&1<<o||s&1<<o||i==null)&&(a>0?i=Number.MAX_SAFE_INTEGER:i=Number.MIN_SAFE_INTEGER);let u=n[o];return i<0&&(i+=u),a>0?i=np(0,i,u):i=np(-1,i,u-1),i}function yH(r,t,e){let n=e.length;for(let o=0;o<e.length;o++)if(e[o]>1){n=o;break}for(let o=n+1;o<e.length;o++)if(t[o]>0||e[o]!==r[o])return!1;return!0}function bH(r,t){let e=r.length>0?r[r.length-1]:1;for(let n=0;n<r.length-1;n++)e+=r[n]*t[n];return e}function q0(r,t,e){let n,o=r.shape.length;typeof t==\"number\"?n=[t,...new Array(o-1).fill(0)]:t.length<o?n=t.concat(new Array(o-t.length).fill(0)):n=t.slice(),n.forEach(i=>{E(i!==-1,()=>\"slice() does not support negative begin indexing.\")});let s;return e==null?s=new Array(o).fill(-1):typeof e==\"number\"?s=[e,...new Array(o-1).fill(-1)]:e.length<o?s=e.concat(new Array(o-e.length).fill(-1)):s=e,s=s.map((i,a)=>i>=0?i:(E(i===-1,()=>`Negative size values should be exactly -1 but got ${i} for the slice() size at index ${a}.`),r.shape[a]-n[a])),[n,s]}function wH(r,t,e,n,o,s,i,a,u){let l;if(n==null?(l=new Array(t.length),l.fill(1)):l=n,i!=null&&(i&i-1)!==0)throw new Error(\"Multiple ellipses in slice is not allowed.\");let c=!1,p={dims:l.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:e.slice(),strides:l.slice(),beginMask:o,endMask:s,ellipsisMask:i,newAxisMask:a,shrinkAxisMask:u};for(let w=0;w<p.dims;w++)c&&(1<<w&a)!==0&&p.numAddAxisAfterEllipsis++,1<<w&i&&(c=!0);c||(p.ellipsisMask|=1<<p.dims,p.dims++);let m={dims:r.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};CH(p,m);let f=!0,d=!0,h=!0,g=[],x=[];for(let w=0;w<r.length;++w){if(m.strides[w]===0)throw Error(`strides[${w}] must be non-zero`);let C=!!(m.shrinkAxisMask&1<<w),N=r[w];if(N===-1){g.push(C?1:-1);continue}let _=[m.beginMask&1<<w,m.endMask&1<<w],A=[m.strides[w]>0?0:-1,m.strides[w]>0?N:N-1];if(C&&m.strides[w]<=0)throw Error(\"only stride 1 allowed on non-range indexing.\");h=h&&m.strides[w]===1;let $=!!(m.beginMask&1<<w&&m.endMask&1<<w);if(m.beginValid&&m.endValid){if(C){let G=m.begin[w]<0?N+m.begin[w]:m.begin[w];if(m.begin[w]=G,m.end[w]=m.begin[w]+1,G<0||G>=N)throw Error(`slice index ${m.begin[w]} of dimension ${w} out of bounds.`)}else m.begin[w]=SE(m.begin[w],0,m.strides[w],N,_,A),m.end[w]=SE(m.end[w],1,m.strides[w],N,_,A);let V=m.strides[w]===1&&m.begin[w]===0&&m.end[w]===N;f=f&&V,d=d&&(w===0&&m.strides[w]===1||V)}else f=f&&m.strides[w]===1&&$,d=d&&(w===0&&m.strides[w]===1||$);let F,P=!1;if(m.beginValid&&m.endValid?(F=m.end[w]-m.begin[w],P=!0):C?(F=1,P=!0):$&&N>=0&&(m.strides[w]<0?F=-N:F=N,P=!0),P){let V;F===0||F<0!=m.strides[w]<0?V=0:V=Math.trunc(F/m.strides[w])+(F%m.strides[w]!==0?1:0),g.push(V)}else g.push(-1)}for(let w=0;w<m.finalShapeGatherIndices.length;++w){let C=m.finalShapeGatherIndices[w];C>=0?x.push(g[C]):C===H0&&x.push(1)}return{finalShapeSparse:x.filter((w,C)=>m.finalShapeGatherIndices[C]!==H0),finalShape:x,isIdentity:f,sliceDim0:d,isSimpleSlice:h,begin:m.begin,end:m.end,strides:m.strides}}function CH(r,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let e=0;t.beginValid=r.begin!=null,t.endValid=r.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let n=0;n<r.dims;n++)if(1<<n&r.ellipsisMask){let o=Math.min(t.dims-(r.dims-n)+1+r.numAddAxisAfterEllipsis,t.dims);for(;e<o;e++)t.begin[e]=0,t.end[e]=0,t.strides[e]=1,t.beginMask|=1<<e,t.endMask|=1<<e,t.finalShapeGatherIndices.push(e),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[e]=n}else if(1<<n&r.newAxisMask)t.finalShapeGatherIndices.push(H0),t.finalShapeGatherIndicesSparse.push(-1);else{if(e===t.begin.length)throw Error(`Index out of range using input dim ${e}; input has only ${t.dims} dims, ${t.begin.length}.`);r.begin!=null&&(t.begin[e]=r.begin[n]),r.end!=null&&(t.end[e]=r.end[n]),t.strides[e]=r.strides[n],r.beginMask&1<<n&&(t.beginMask|=1<<e),r.endMask&1<<n&&(t.endMask|=1<<e),r.shrinkAxisMask&1<<n?(t.finalShapeGatherIndices.push(fH),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<e):(t.finalShapeGatherIndices.push(e),t.finalShapeGatherIndicesSparse.push(n)),t.inputShapeGatherIndicesSparse[e]=n,e++}}function SE(r,t,e,n,o,s){if(o[t])return e>0?s[t]:s[t+1&1];{let i=r<0?n+r:r;return i<s[0]?s[0]:i>s[1]?s[1]:i}}var Q={};Wt(Q,{Serializable:()=>uh,SerializationMap:()=>_i,registerClass:()=>Cn});var uh=class{getClassName(){return this.constructor.className}static fromConfig(t,e){return new t(e)}},_i=class{constructor(){this.classNameMap={}}static getMap(){return _i.instance==null&&(_i.instance=new _i),_i.instance}static register(t){_i.getMap().classNameMap[t.className]=[t,t.fromConfig]}};function Cn(r){E(r.className!=null,()=>\"Class being registered does not have the static className property defined.\"),E(typeof r.className==\"string\",()=>\"className is required to be a string, but got type \"+typeof r.className),E(r.className.length>0,()=>\"Class being registered has an empty-string as its className, which is disallowed.\"),_i.register(r)}var OE={};Wt(OE,{TEST_EPSILON_FLOAT16:()=>DE,createVideoElement:()=>EH,encodeStrings:()=>FE,expectArrayBuffersEqual:()=>kH,expectArraysClose:()=>SH,expectArraysEqual:()=>NH,expectNumbersClose:()=>RE,expectPromiseToFail:()=>vH,expectValuesInRange:()=>TH,play:()=>_H,testEpsilon:()=>ix});var IH=.001,DE=.1;function SH(r,t,e){return e==null&&(e=ix()),K0(r,t,(n,o)=>j0(n,o,e))}function ix(){return k.backend.floatPrecision()===32?IH:DE}function K0(r,t,e){let n=!0;if((xr(r)||xr(t))&&(n=!1),xr(r)&&xr(t)&&(n=!0),n){let i=r.constructor.name,a=t.constructor.name;if(i!==a)throw new Error(`Arrays are of different type. Actual: ${i}. Expected: ${a}`)}if(Array.isArray(r)&&Array.isArray(t)){let i=Br(r),a=Br(t);if(!Dn(i,a))throw new Error(`Arrays have different shapes. Actual: [${i}]. Expected: [${a}]`)}let o=xr(r)?r:Bo(r),s=xr(t)?t:Bo(t);if(o.length!==s.length)throw new Error(`Arrays have different lengths actual: ${o.length} vs expected: ${s.length}.\nActual:   ${o}.\nExpected: ${s}.`);for(let i=0;i<s.length;++i){let a=o[i],u=s[i];if(!e(a,u))throw new Error(`Arrays differ: actual[${i}] = ${a}, expected[${i}] = ${u}.\nActual:   ${o}.\nExpected: ${s}.`)}typeof expect!=\"undefined\"&&expect().nothing()}function vH(r,t){r().then(()=>t.fail(),()=>t()),typeof expect!=\"undefined\"&&expect().nothing()}function NH(r,t){let e=typeof t==\"string\"||typeof t==\"number\"||typeof t==\"boolean\"?[t]:t;return Vo(r)||Vo(r[0])||Vo(t)||Vo(t[0])?K0(r,e,(n,o)=>n==o):K0(r,t,(n,o)=>j0(n,o,0))}function RE(r,t,e){if(e==null&&(e=ix()),!j0(r,t,e))throw new Error(`Numbers differ: actual === ${r}, expected === ${t}`);typeof expect!=\"undefined\"&&expect().nothing()}function j0(r,t,e){return!isFinite(r)&&!isFinite(t)?!0:!(isNaN(r)||isNaN(t)||Math.abs(r-t)>e)}function TH(r,t,e){for(let n=0;n<r.length;n++)if(r[n]<t||r[n]>e)throw new Error(`Value out of range:${r[n]} low: ${t}, high: ${e}`)}function kH(r,t){let e=new Float32Array(r),n=new Float32Array(t);if(e.length!==n.length)throw new Error(`Expected ArrayBuffer to be of length ${n.length}, but it was ${e.length}`);for(let o=0;o<n.length;o++)if(e[o]!==n[o])throw new Error(`Expected ArrayBuffer value at ${o} to be ${n[o]} but got ${e[o]} instead`)}function FE(r){for(let t=0;t<r.length;t++){let e=r[t];Array.isArray(e)?FE(e):r[t]=Hl(e)}return r}function EH(r){let t=document.createElement(\"video\");return\"playsInline\"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position=\"fixed\",t.style.left=\"0px\",t.style.top=\"0px\",t.preload=\"auto\",t.appendChild(r),new Promise(e=>{t.addEventListener(\"loadeddata\",n=>e(t)),t.load()})}async function _H(r){await r.play(),\"requestVideoFrameCallback\"in r&&await new Promise(t=>{r.requestVideoFrameCallback(t)})}var PE=\"4.0.0\";function AH(r,t){let e=I(r,\"a\",\"add\"),n=I(t,\"b\",\"add\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(Zn,o)}var X=T({add_:AH});function $H(r,t){let e=I(r,\"a\",\"floorDiv\"),n=I(t,\"b\",\"floorDiv\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ns,o)}var Yp=T({floorDiv_:$H});function DH(r,t){let e=I(r,\"a\",\"div\"),n=I(t,\"b\",\"div\");if([e,n]=Ut(e,n),e.dtype===\"int32\"&&n.dtype===\"int32\")return Yp(e,n);let o={a:e,b:n},s={};return k.runKernel(Qo,o,s)}var pt=T({div_:DH});function RH(r,t){let e=I(r,\"a\",\"mul\"),n=I(t,\"b\",\"mul\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(hs,o)}var D=T({mul_:RH});function FH(r){let t=I(r,\"x\",\"abs\");if(t.dtype===\"complex64\"){let e={x:t};return k.runKernel(_l,e)}else{let e={x:t};return k.runKernel(ii,e)}}var Ee=T({abs_:FH});function OH(r){let e={x:I(r,\"x\",\"acos\")};return k.runKernel(oa,e)}var ax=T({acos_:OH});function PH(r){let e={x:I(r,\"x\",\"acosh\")};return k.runKernel(sa,e)}var lx=T({acosh_:PH});function LH(r){E(Array.isArray(r),()=>\"The argument passed to tf.addN() must be a list of tensors\"),E(r.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${r.length}`);let t=r.map((o,s)=>I(o,`tensors${s}`,\"addN\")),e=t[0];t.forEach(o=>{if(o.dtype!==e.dtype)throw new Error(\"All tensors passed to tf.addN() must have the same dtype\")}),t.forEach(o=>{if(!Dn(o.shape,e.shape))throw new Error(\"All tensors passed to tf.addN() must have the same shape\")});let n=t;return k.runKernel(Go,n)}var LE=T({addN_:LH});function MH(r,t=null,e=!1){let o={x:I(r,\"x\",\"all\",\"bool\")},s={axis:t,keepDims:e};return k.runKernel(ia,o,s)}var Zp=T({all_:MH});function zH(r,t=null,e=!1){let o={x:I(r,\"x\",\"any\",\"bool\")},s={axis:t,keepDims:e};return k.runKernel(aa,o,s)}var qu=T({any_:zH});function BH(r,t=0){let n={x:I(r,\"x\",\"argMax\")},o={axis:t};return k.runKernel(Wo,n,o)}var Ai=T({argMax_:BH});function VH(r,t=0){let n={x:I(r,\"x\",\"argMin\")},o={axis:t};return k.runKernel(kl,n,o)}var ux=T({argMin_:VH});function GH(r){let e={x:I(r,\"x\",\"asin\")};return k.runKernel(la,e)}var cx=T({asin_:GH});function WH(r){let e={x:I(r,\"x\",\"asinh\")};return k.runKernel(ua,e)}var px=T({asinh_:WH});function UH(r){let e={x:I(r,\"x\",\"atan\")};return k.runKernel(ca,e)}var mx=T({atan_:UH});function HH(r,t){let e=I(r,\"a\",\"atan2\"),n=I(t,\"b\",\"atan2\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ma,o)}var fx=T({atan2_:HH});function qH(r){let e={x:I(r,\"x\",\"atanh\")};return k.runKernel(pa,e)}var dx=T({atanh_:qH});function KH(r,t,e,n,o=\"NHWC\",s){let i=r[3],a=[...t,i],u=zE(o);return ju(r,a,e,s,n,null,null,u)}function Y0(r,t,e,n,o,s,i=\"channelsLast\"){let[a,u]=hx(t),l;if(i===\"channelsLast\")l=[a,u,r[3],r[3]];else if(i===\"channelsFirst\")l=[a,u,r[1],r[1]];else throw new Error(`Unknown dataFormat ${i}`);return ju(r,l,e,n,o,s,!1,i)}function jH(r,t,e,n,o,s,i=\"NDHWC\"){let[a,u,l]=X0(t),c,p;if(i===\"NDHWC\")p=\"channelsLast\",c=[a,u,l,r[4],r[4]];else if(i===\"NCDHW\")p=\"channelsFirst\",c=[a,u,l,r[1],r[1]];else throw new Error(`Unknown dataFormat ${i}`);return ME(r,c,e,n,o,!1,p,s)}function ju(r,t,e,n,o,s,i=!1,a=\"channelsLast\"){let[u,l,c,p]=[-1,-1,-1,-1];if(a===\"channelsLast\")[u,l,c,p]=r;else if(a===\"channelsFirst\")[u,p,l,c]=r;else throw new Error(`Unknown dataFormat ${a}`);let[m,f,,d]=t,[h,g]=hx(e),[x,b]=hx(n),w=Jp(m,x),C=Jp(f,b),{padInfo:N,outHeight:_,outWidth:A}=ZH(o,l,c,h,g,w,C,s,a),$=i?d*p:d,F;return a===\"channelsFirst\"?F=[u,$,_,A]:a===\"channelsLast\"&&(F=[u,_,A,$]),{batchSize:u,dataFormat:a,inHeight:l,inWidth:c,inChannels:p,outHeight:_,outWidth:A,outChannels:$,padInfo:N,strideHeight:h,strideWidth:g,filterHeight:m,filterWidth:f,effectiveFilterHeight:w,effectiveFilterWidth:C,dilationHeight:x,dilationWidth:b,inShape:r,outShape:F,filterShape:t}}function ME(r,t,e,n,o,s=!1,i=\"channelsLast\",a){let[u,l,c,p,m]=[-1,-1,-1,-1,-1];if(i===\"channelsLast\")[u,l,c,p,m]=r;else if(i===\"channelsFirst\")[u,m,l,c,p]=r;else throw new Error(`Unknown dataFormat ${i}`);let[f,d,h,,g]=t,[x,b,w]=X0(e),[C,N,_]=X0(n),A=Jp(f,C),$=Jp(d,N),F=Jp(h,_),{padInfo:P,outDepth:V,outHeight:G,outWidth:W}=JH(o,l,c,p,x,b,w,A,$,F,a),q=s?g*m:g,H;return i===\"channelsFirst\"?H=[u,q,V,G,W]:i===\"channelsLast\"&&(H=[u,V,G,W,q]),{batchSize:u,dataFormat:i,inDepth:l,inHeight:c,inWidth:p,inChannels:m,outDepth:V,outHeight:G,outWidth:W,outChannels:q,padInfo:P,strideDepth:x,strideHeight:b,strideWidth:w,filterDepth:f,filterHeight:d,filterWidth:h,effectiveFilterDepth:A,effectiveFilterHeight:$,effectiveFilterWidth:F,dilationDepth:C,dilationHeight:N,dilationWidth:_,inShape:r,outShape:H,filterShape:t}}function XH(r,t,e,n,o){n==null&&(n=Z0(r,t,e));let s=r[0],i=r[1],a=Ku((s-t+2*n)/e+1,o),u=Ku((i-t+2*n)/e+1,o);return[a,u]}function YH(r,t,e,n,o,s){o==null&&(o=Z0(r,t,n));let i=r[0],a=r[1],u=r[2],l=Ku((i-t+2*o)/n+1,s),c=Ku((a-t+2*o)/n+1,s),p=Ku((u-t+2*o)/n+1,s);return[l,c,p,e]}function Z0(r,t,e,n=1){let o=Jp(t,n);return Math.floor((r[0]*(e-1)-e+o)/2)}function hx(r){return typeof r==\"number\"?[r,r,r]:r.length===2?[r[0],r[1],1]:r}function X0(r){return typeof r==\"number\"?[r,r,r]:r}function Jp(r,t){return t<=1?r:r+(r-1)*(t-1)}function ZH(r,t,e,n,o,s,i,a,u){let l,c,p;if(typeof r==\"number\"){l={top:r,bottom:r,left:r,right:r,type:r===0?\"VALID\":\"NUMBER\"};let f=XH([t,e],s,n,r,a);c=f[0],p=f[1]}else if(r===\"same\"){c=Math.ceil(t/n),p=Math.ceil(e/o);let m=Math.max(0,(c-1)*n+s-t),f=Math.max(0,(p-1)*o+i-e),d=Math.floor(m/2),h=m-d,g=Math.floor(f/2),x=f-g;l={top:d,bottom:h,left:g,right:x,type:\"SAME\"}}else if(r===\"valid\")l={top:0,bottom:0,left:0,right:0,type:\"VALID\"},c=Math.ceil((t-s+1)/n),p=Math.ceil((e-i+1)/o);else if(typeof r==\"object\"){let m=u===\"channelsLast\"?r[1][0]:r[2][0],f=u===\"channelsLast\"?r[1][1]:r[2][1],d=u===\"channelsLast\"?r[2][0]:r[3][0],h=u===\"channelsLast\"?r[2][1]:r[3][1];l={top:m,bottom:f,left:d,right:h,type:m===0&&f===0&&d===0&&h===0?\"VALID\":\"EXPLICIT\"},c=Ku((t-s+m+f)/n+1,a),p=Ku((e-i+d+h)/o+1,a)}else throw Error(`Unknown padding parameter: ${r}`);return{padInfo:l,outHeight:c,outWidth:p}}function JH(r,t,e,n,o,s,i,a,u,l,c){let p,m,f,d;if(typeof r==\"number\"){p={top:r,bottom:r,left:r,right:r,front:r,back:r,type:r===0?\"VALID\":\"NUMBER\"};let g=YH([t,e,n,1],a,1,o,r,c);m=g[0],f=g[1],d=g[2]}else if(r===\"same\"){m=Math.ceil(t/o),f=Math.ceil(e/s),d=Math.ceil(n/i);let h=(m-1)*o+a-t,g=(f-1)*s+u-e,x=(d-1)*i+l-n,b=Math.floor(h/2),w=h-b,C=Math.floor(g/2),N=g-C,_=Math.floor(x/2),A=x-_;p={top:C,bottom:N,left:_,right:A,front:b,back:w,type:\"SAME\"}}else if(r===\"valid\")p={top:0,bottom:0,left:0,right:0,front:0,back:0,type:\"VALID\"},m=Math.ceil((t-a+1)/o),f=Math.ceil((e-u+1)/s),d=Math.ceil((n-l+1)/i);else throw Error(`Unknown padding parameter: ${r}`);return{padInfo:p,outDepth:m,outHeight:f,outWidth:d}}function Ku(r,t){if(!t)return Math.trunc(r);switch(t){case\"round\":return Math.round(r);case\"ceil\":return Math.ceil(r);case\"floor\":return Math.floor(r);default:throw new Error(`Unknown roundingMode ${t}`)}}function to(r){let[t,e,n]=hx(r);return t===1&&e===1&&n===1}function Ar(r,t){return to(r)||to(t)}function zE(r){if(r===\"NHWC\")return\"channelsLast\";if(r===\"NCHW\")return\"channelsFirst\";throw new Error(`Unknown dataFormat ${r}`)}function Ie(r,t,e){if(e!=null){if(typeof t==\"string\")throw Error(`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${t}.`);if(typeof t==\"number\")E(na(t),()=>`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${t}.`);else if(typeof t==\"object\")t.forEach(n=>{n.forEach(o=>{E(na(o),()=>`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${o}.`)})});else throw Error(`Error in ${r}: Unknown padding parameter: ${t}`)}}function QH(r,t){let n={x:I(r,\"x\",\"reshape\",\"string_or_numeric\")},o={shape:t};return k.runKernel(di,n,o)}var R=T({reshape_:QH});function tq(r,t,e,n,o){let s=I(r,\"x\",\"avgPool\",\"float32\"),i=1;E(Ar(e,i),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${e} and dilations '${i}'`);let a=s,u=!1;s.rank===3&&(u=!0,a=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(a.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${a.rank}.`),Ie(\"avgPool\",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o},p=k.runKernel(Uo,l,c);return p=J(p,s.dtype),u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var Yl=T({avgPool_:tq});function eq(r,t,e,n,o,s=\"NDHWC\"){let i=I(r,\"x\",\"avgPool3d\",\"float32\"),a=i,u=!1;i.rank===4&&(u=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(a.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${a.rank}.`),E(s===\"NDHWC\",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Ie(\"avgPool3d\",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o,dataFormat:s},p=k.runKernel(El,l,c);return p=J(p,a.dtype),u?R(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var gx=T({avgPool3d_:eq});function rq(r,t=0){E(r.length>=1,()=>\"Pass at least one tensor to concat\");let e=ja(r,\"tensors\",\"concat\",\"string_or_numeric\");if(e[0].dtype===\"complex64\"&&e.forEach(s=>{if(s.dtype!==\"complex64\")throw new Error(`Cannot concatenate complex64 tensors with a tensor\n          with dtype ${s.dtype}. `)}),e.length===1)return sn(e[0]);let n=e,o={axis:t};return k.runKernel(li,n,o)}var ne=T({concat_:rq});function nq(r){let e={x:I(r,\"x\",\"sigmoid\",\"float32\")};return k.runKernel(_s,e)}var Yr=T({sigmoid_:nq});function oq(r,t,e){let n=I(r,\"x\",\"slice\",\"string_or_numeric\");if(n.rank===0)throw new Error(\"Slicing scalar is not possible\");let o={x:n},s={begin:t,size:e};return k.runKernel(gi,o,s)}var Rt=T({slice_:oq});function sq(r){let e={x:I(r,\"x\",\"tanh\",\"float32\")};return k.runKernel(Ps,e)}var $i=T({tanh_:sq});function iq(r,t,e,n,o,s){let i=I(r,\"forgetBias\",\"basicLSTMCell\"),a=I(t,\"lstmKernel\",\"basicLSTMCell\"),u=I(e,\"lstmBias\",\"basicLSTMCell\"),l=I(n,\"data\",\"basicLSTMCell\"),c=I(o,\"c\",\"basicLSTMCell\"),p=I(s,\"h\",\"basicLSTMCell\"),m=ne([l,p],1),f=Lt(m,a),d=X(f,u),h=d.shape[0],g=d.shape[1]/4,x=[h,g],b=Rt(d,[0,0],x),w=Rt(d,[0,g],x),C=Rt(d,[0,g*2],x),N=Rt(d,[0,g*3],x),_=X(D(Yr(b),$i(w)),D(c,Yr(X(i,C)))),A=D($i(_),Yr(N));return[_,A]}var BE=T({basicLSTMCell_:iq});function aq(r,t,e){let n=I(r,\"x\",\"batchToSpaceND\"),o=t.reduce((a,u)=>a*u);E(n.rank>=1+t.length,()=>`input rank is ${n.rank} but should be > than blockShape.length ${t.length}`),E(e.length===t.length,()=>`crops.length is ${e.length} but should be equal to blockShape.length  ${t.length}`),E(n.shape[0]%o===0,()=>`input tensor batch is ${n.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(\" * \")} === ${o}`);let s={x:n},i={blockShape:t,crops:e};return k.runKernel(ai,s,i)}var Zl=T({batchToSpaceND_:aq});function VE(r){let t;return r.rank===0||r.rank===1?t=R(r,[1,1,1,r.size]):r.rank===2?t=R(r,[1,1,r.shape[0],r.shape[1]]):r.rank===3?t=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]):t=r,t}function lq(r,t,e,n,o,s){s==null&&(s=.001);let i=I(r,\"x\",\"batchNorm\"),a=I(t,\"mean\",\"batchNorm\"),u=I(e,\"variance\",\"batchNorm\"),l;o!=null&&(l=I(o,\"scale\",\"batchNorm\"));let c;n!=null&&(c=I(n,\"offset\",\"batchNorm\")),E(a.rank===u.rank,()=>\"Batch normalization gradient requires mean and variance to have equal ranks.\"),E(c==null||a.rank===c.rank,()=>\"Batch normalization gradient requires mean and offset to have equal ranks.\"),E(l==null||a.rank===l.rank,()=>\"Batch normalization gradient requires mean and scale to have equal ranks.\");let m={x:VE(i),scale:l,offset:c,mean:a,variance:u},f={varianceEpsilon:s},d=k.runKernel(os,m,f);return R(d,i.shape)}var Di=T({batchNorm_:lq});function uq(r,t,e,n,o,s){let i=I(r,\"x\",\"batchNorm\"),a=I(t,\"mean\",\"batchNorm\"),u=I(e,\"variance\",\"batchNorm\"),l;o!=null&&(l=I(o,\"scale\",\"batchNorm\"));let c;return n!=null&&(c=I(n,\"offset\",\"batchNorm\")),E(i.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),E(a.rank===2||a.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${a.rank}.`),E(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===2||c.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var xx=T({batchNorm2d_:uq});function cq(r,t,e,n,o,s){let i=I(r,\"x\",\"batchNorm\"),a=I(t,\"mean\",\"batchNorm\"),u=I(e,\"variance\",\"batchNorm\"),l;o!=null&&(l=I(o,\"scale\",\"batchNorm\"));let c;return n!=null&&(c=I(n,\"offset\",\"batchNorm\")),E(i.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),E(a.rank===3||a.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${a.rank}.`),E(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===3||c.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var yx=T({batchNorm3d_:cq});function pq(r,t,e,n,o,s){let i=I(r,\"x\",\"batchNorm\"),a=I(t,\"mean\",\"batchNorm\"),u=I(e,\"variance\",\"batchNorm\"),l;o!=null&&(l=I(o,\"scale\",\"batchNorm\"));let c;return n!=null&&(c=I(n,\"offset\",\"batchNorm\")),E(i.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),E(a.rank===4||a.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${a.rank}.`),E(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===4||c.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var bx=T({batchNorm4d_:pq});function mq(r,t,e){let n=I(r,\"x\",\"bincount\"),o=I(t,\"weights\",\"bincount\");E(n.dtype===\"int32\",()=>`Error in bincount: input dtype must be int32, but got ${n.dtype}`),E(e>=0,()=>`size must be non-negative, but got ${e}.`),E(o.size===n.size||o.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${n.shape}, weights shape: ${o.shape}.`);let s={x:n,weights:o},i={size:e};return k.runKernel(up,s,i)}var wx=T({bincount_:mq});function fq(r,t){let e=I(r,\"s0\",\"broadcastArgs\",\"int32\"),n=I(t,\"s1\",\"broadcastArgs\",\"int32\");if(e.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${e.rank}`);if(n.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${n.rank}`);let o={s0:e,s1:n};return k.runKernel(cp,o)}var GE=T({broadcastArgs_:fq});function dq(r,t){let e=I(r,\"broadcastTo\",\"x\"),n=e.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<e.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${e.rank}.`);if(t.length>e.rank){let l=e.shape.slice();for(;l.length<t.length;)l.unshift(1);e=R(e,l)}let o=e.shape,s=Array.from(t);for(let l=t.length-1;l>=0;l--)if(o[l]===t[l])s[l]=1;else if(e.shape[l]!==1)throw new Error(`broadcastTo(): [${n}] cannot be broadcast to [${t}].`);if(s.map((l,c)=>l>1?c:-1).filter(l=>l>=0).length===0)return sn(e);let a={x:e},u={reps:s};return k.runKernel(Jn,a,u)}var Ri=T({broadcastTo_:dq});function hq(r){let e={x:I(r,\"x\",\"ceil\",\"float32\")};return k.runKernel(qo,e)}var Cx=T({ceil_:hq});function xo(r,t,e){let n={shape:r,value:t,dtype:e};return k.runKernel(Dl,{},n)}function gq(r,t,e){let n=I(r,\"x\",\"clipByValue\");if(E(t<=e,()=>`Error in clip: min (${t}) must be less than or equal to max (${e}).`),t===e)return xo(n.shape,t,n.dtype);let o={x:n},s={clipValueMin:t,clipValueMax:e};return k.runKernel(uo,o,s)}var Cr=T({clipByValue_:gq});function xq(r){return ne(r,0)}var Ix=T({concat1d_:xq});function yq(r,t){return ne(r,t)}var Sx=T({concat2d_:yq});function bq(r,t){return ne(r,t)}var vx=T({concat3d_:bq});function wq(r,t){return ne(r,t)}var Nx=T({concat4d_:wq});function Cq(r,t,e,n,o=\"NHWC\",s=[1,1],i){let a=I(r,\"x\",\"conv2d\",\"float32\"),u=I(t,\"filter\",\"conv2d\",\"float32\"),l=a,c=!1;a.rank===3&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(l.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${l.rank}.`),E(u.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${u.rank}.`),Ie(\"conv2d\",n,i);let p=o===\"NHWC\"?l.shape[3]:l.shape[1];E(p===u.shape[2],()=>`Error in conv2d: depth of input (${p}) must match input depth for filter ${u.shape[2]}.`),E(Ar(e,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`);let m={x:l,filter:u},f={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i},d=k.runKernel(Ko,m,f);return c?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var In=T({conv2d_:Cq});function Iq(r,t,e,n,o=\"NWC\",s=1,i){let a=I(r,\"x\",\"conv1d\"),u=I(t,\"filter\",\"conv1d\"),l=a,c=!1;a.rank===2&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1]])),E(l.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${l.rank}.`),E(u.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${u.rank}.`),Ie(\"conv1d\",n,i),E(l.shape[2]===u.shape[1],()=>`Error in conv1d: depth of input (${l.shape[2]}) must match input depth for filter ${u.shape[1]}.`),E(Ar(e,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${e} and dilation '${s}'`),E(o===\"NWC\",()=>`Error in conv1d: got dataFormat of ${o} but only NWC is currently supported.`);let p=R(u,[1,u.shape[0],u.shape[1],u.shape[2]]),m=R(l,[l.shape[0],1,l.shape[1],l.shape[2]]),g=In(m,p,[1,e],n,\"NHWC\",[1,s],i);return c?R(g,[g.shape[2],g.shape[3]]):R(g,[g.shape[0],g.shape[2],g.shape[3]])}var Qp=T({conv1d_:Iq});function Sq(r,t,e,n,o,s=\"NHWC\",i){E(r.length===t.rank,()=>`Length of inShape (${r.length}) and rank of dy (${t.rank}) must match`);let a=r,u=t,l=!1;t.rank===3&&(l=!0,u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]),a=[1,r[0],r[1],r[2]]),E(a.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${a.length}.`),E(u.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${u.rank}`),E(e.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${e.rank}`);let c=s===\"NHWC\"?a[3]:a[1],p=s===\"NHWC\"?u.shape[3]:u.shape[1];E(c===e.shape[2],()=>`Error in conv2dDerInput: depth of input (${c}) must match input depth for filter ${e.shape[2]}.`),E(p===e.shape[3],()=>`Error in conv2dDerInput: depth of output (${p}) must match output depth for filter ${e.shape[3]}.`),Ie(\"conv2dDerInput\",o,i);let m={dy:u,filter:e},f={strides:n,pad:o,dataFormat:s,dimRoundingMode:i,inputShape:a},d=k.runKernel(jo,m,f);return l?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var tm=T({conv2DBackpropInput_:Sq});function vq(r,t,e,n,o,s){let i=I(r,\"x\",\"conv2dTranspose\"),a=I(t,\"filter\",\"conv2dTranspose\");return tm(e,i,a,n,o,\"NHWC\",s)}var em=T({conv2dTranspose_:vq});function Nq(r,t,e,n,o=\"NDHWC\",s=[1,1,1]){let i=I(r,\"x\",\"conv3d\"),a=I(t,\"filter\",\"conv3d\"),u=i,l=!1;i.rank===4&&(l=!0,u=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(u.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${u.rank}.`),E(a.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${a.rank}.`),E(u.shape[4]===a.shape[3],()=>`Error in conv3d: depth of input (${u.shape[4]}) must match input depth for filter ${a.shape[3]}.`),E(Ar(e,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`),E(o===\"NDHWC\",()=>`Error in conv3d: got dataFormat of ${o} but only NDHWC is currently supported.`);let c={x:u,filter:a},p={strides:e,pad:n,dataFormat:o,dilations:s},m=k.runKernel(Al,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3],m.shape[4]]):m}var Tx=T({conv3d_:Nq});function Tq(r,t,e,n,o){E(r.length===t.rank,()=>`Length of inShape (${r.length}) and rank of dy (${t.rank}) must match`);let s=r,i=t,a=!1;t.rank===4&&(a=!0,i=R(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,r[0],r[1],r[2],r[3]]);let u=s[4],l=i.shape[4];E(s.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),E(i.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),E(e.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${e.rank}`),E(u===e.shape[3],()=>`Error in conv3dDerInput: depth of input (${u}) must match input depth for filter ${e.shape[3]}.`),E(l===e.shape[4],()=>`Error in conv3dDerInput: depth of output (${l}) must match output depth for filter ${e.shape[4]}.`);let c={dy:i,filter:e},p={pad:o,strides:n,inputShape:s},m=k.runKernel(dp,c,p);return a?R(m,[m.shape[1],m.shape[2],m.shape[3],m.shape[4]]):m}var kx=T({conv3DBackpropInput_:Tq});function kq(r,t,e,n,o){let s=I(r,\"x\",\"conv3dTranspose\"),i=I(t,\"filter\",\"conv3dTranspose\");return kx(e,s,i,n,o)}var Ex=T({conv3dTranspose_:kq});function Eq(r){let e={x:I(r,\"x\",\"cos\",\"float32\")};return k.runKernel(Xo,e)}var Jl=T({cos_:Eq});function _q(r){let e={x:I(r,\"x\",\"cosh\",\"float32\")};return k.runKernel(Yo,e)}var rm=T({cosh_:_q});function Aq(r,t=0,e=!1,n=!1){let s={x:I(r,\"x\",\"cumprod\")},i={axis:t,exclusive:e,reverse:n};return k.runKernel(fa,s,i)}var Xu=T({cumprod_:Aq});function $q(r,t=0,e=!1,n=!1){let s={x:I(r,\"x\",\"cumsum\")},i={axis:t,exclusive:e,reverse:n};return k.runKernel(Zo,s,i)}var nm=T({cumsum_:$q});function Dq(r,t,e,n=!1){let o=I(r,\"x\",\"denseBincount\"),s=I(t,\"weights\",\"denseBincount\");E(o.dtype===\"int32\",()=>`Error in denseBincount: input dtype must be int32, but got ${o.dtype}`),E(o.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${o.rank}.`),E(e>=0,()=>`size must be non-negative, but got ${e}.`),E(s.size===o.size||s.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${o.shape}, weights shape: ${s.shape}.`);let i={x:o,weights:s},a={size:e,binaryOutput:n};return k.runKernel(hp,i,a)}var ch=T({denseBincount_:Dq});function Rq(r,t,e=\"NHWC\"){let n=I(r,\"x\",\"depthToSpace\",\"float32\"),o=e===\"NHWC\"?n.shape[1]:n.shape[2],s=e===\"NHWC\"?n.shape[2]:n.shape[3],i=e===\"NHWC\"?n.shape[3]:n.shape[1];E(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),E(o*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${o} and ${t}  for depthToSpace with input shape\n    ${n.shape}`),E(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${s} and ${t} for depthToSpace with input shape\n        ${n.shape}`),E(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${n.shape}`);let a={x:n},u={blockSize:t,dataFormat:e};return k.runKernel(ha,a,u)}var _x=T({depthToSpace_:Rq});function Fq(r,t,e,n,o=\"NHWC\",s=[1,1],i){let a=I(r,\"x\",\"depthwiseConv2d\",\"float32\"),u=I(t,\"filter\",\"depthwiseConv2d\",\"float32\"),l=a,c=!1;a.rank===3&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(l.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${l.rank}.`),E(u.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${u.rank}.`);let p=o===\"NHWC\"?l.shape[3]:l.shape[1];E(p===u.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${p}) must match the inChannels dimension in filter ${u.shape[2]}.`),Ie(\"depthwiseConv2d\",n,i);let m={x:l,filter:u},f={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i},d=k.runKernel(Jo,m,f);return c?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var Fi=T({depthwiseConv2d_:Fq});function Oq(r){let e={x:I(r,\"x\",\"diag\")};return k.runKernel(yp,e)}var WE=T({diag_:Oq});function Pq(r,t,e,n,o=[1,1],s=\"NHWC\"){let i=I(r,\"x\",\"dilation2d\"),a=I(t,\"filter\",\"dilation2d\");E(i.rank===3||i.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),E(a.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${a.rank}.`),E(s===\"NHWC\",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let u=i,l=!1;i.rank===3&&(u=R(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=!0);let c={x:u,filter:a},p={strides:e,pad:n,dilations:o},m=k.runKernel($l,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var Ax=T({dilation2d_:Pq});function Lq(r,t){let e=I(r,\"a\",\"equal\",\"string_or_numeric\"),n=I(t,\"b\",\"equal\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(xa,o)}var $r=T({equal_:Lq});function Mq(r,t,e){let n=I(t,\"a\",\"where\"),o=I(e,\"b\",\"where\"),s=I(r,\"condition\",\"where\",\"bool\"),i=Pt(Pt(s.shape,n.shape),o.shape),a=Ri(s,i),u=Ri(n,i),l=Ri(o,i),c={condition:a,t:u,e:l};return k.runKernel(hi,c)}var _e=T({where_:Mq});function zq(r){let e={x:I(r,\"x\",\"zerosLike\")};return k.runKernel(wi,e)}var It=T({zerosLike_:zq});function Bq(r,t){let e=I(r,\"a\",\"div\"),n=I(t,\"b\",\"div\");[e,n]=Ut(e,n);let o=pt(e,n),s=It(o),i=$r(n,s);return _e(i,s,o)}var $x=T({divNoNan_:Bq});function Vq(r,t){let e=I(r,\"t1\",\"dot\"),n=I(t,\"t2\",\"dot\");E((e.rank===1||e.rank===2)&&(n.rank===1||n.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${e.rank} and ${n.rank}.`);let o=e.rank===1?e.size:e.shape[1],s=n.rank===1?n.size:n.shape[0];if(E(o===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${o} and ${s}.`),e.rank===1&&n.rank===1){let i=R(e,[1,-1]),a=R(n,[-1,1]),u=Lt(i,a);return R(u,[])}else if(e.rank===1&&n.rank===2){let i=R(e,[1,-1]),a=R(n,[n.shape[0],n.shape[1]]),u=Lt(i,a);return R(u,[u.size])}else if(e.rank===2&&n.rank===1){let i=R(n,[-1,1]),a=Lt(e,i);return R(a,[a.size])}else{let i=R(n,[n.shape[0],n.shape[1]]);return Lt(e,i)}}var Dx=T({dot_:Vq});function Gq(r,...t){let e=t.map((o,s)=>I(o,`tensors${s}`,\"einsum\")),n={equation:r};return k.runKernel(bp,e,n)}var UE=T({einsum_:Gq});function Wq(r){let e={x:I(r,\"x\",\"elu\",\"float32\")};return k.runKernel(ts,e)}var Oi=T({elu_:Wq});function Uq(r){let t=I(r,\"x\",\"erf\");E(t.dtype===\"int32\"||t.dtype===\"float32\",()=>\"Input dtype must be `int32` or `float32`.\"),t.dtype===\"int32\"&&(t=J(t,\"float32\"));let e={x:t};return k.runKernel(ga,e)}var Rx=T({erf_:Uq});function J0(r,t){for(let e=0;e<r.length;++e)if(r[r.length-e-1]!==t-1-e)return!1;return!0}function HE(r,t,e){let n=r.length+t.length,o=[],s=0,i=0;for(let a=0;a<n;a++)e.indexOf(a)===-1?o.push(r[s++]):o.push(t[i++]);return o}function Q0(r,t){let e=[],n=r.length;for(let s=0;s<n;s++)t.indexOf(s)===-1&&e.push(r[s]);let o=t.map(s=>r[s]);return[e,o]}function yo(r,t){let e=t.map(n=>1);return HE(r,e,t)}function Hq(r,t,e){E(J0(t,e),()=>`${r} supports only inner-most axes for now. Got axes ${t} and rank-${e} input.`)}function tv(r,t){if(J0(r,t))return null;let e=[];for(let n=0;n<t;++n)r.indexOf(n)===-1&&e.push(n);return r.forEach(n=>e.push(n)),e}function ph(r){return r.map((t,e)=>[e,t]).sort((t,e)=>t[1]-e[1]).map(t=>t[0])}function qq(r,t){let e=[];for(let n=t-r;n<t;++n)e.push(n);return e}function Kq(r,t=null,e=!1){let o={x:I(r,\"x\",\"max\")},s={reductionIndices:t,keepDims:e};return k.runKernel(ls,o,s)}var Ir=T({max_:Kq});function jq(r,t=null,e=!1){let o={x:I(r,\"x\",\"min\")},s={axis:t,keepDims:e};return k.runKernel(ms,o,s)}var Ja=T({min_:jq});function Xq(r,t){let e=I(r,\"base\",\"pow\"),n=I(t,\"exp\",\"pow\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ys,o)}var an=T({pow_:Xq});function mt(r,t){if((xr(r)&&t!==\"string\"||Array.isArray(r))&&t!==\"complex64\")throw new Error(\"Error creating a new Scalar: value must be a primitive (number|boolean|string)\");if(t===\"string\"&&xr(r)&&!(r instanceof Uint8Array))throw new Error(\"When making a scalar from encoded string, the value must be `Uint8Array`.\");return on(r,[],[],t)}function Yq(r){let e={x:I(r,\"x\",\"sqrt\",\"float32\")};return k.runKernel(As,e)}var Se=T({sqrt_:Yq});function Zq(r){let t=I(r,\"x\",\"square\"),e={};return k.runKernel(\"Square\",{x:t},e)}var Mt=T({square_:Zq});function Jq(r,t=null,e=!1){let n=I(r,\"x\",\"sum\");n.dtype===\"bool\"&&(n=J(n,\"int32\"));let o={x:n},s={axis:t,keepDims:e};return k.runKernel($s,o,s)}var ft=T({sum_:Jq});function Qq(r,t=\"euclidean\",e=null,n=!1){r=I(r,\"x\",\"norm\");let o=qE(r,t,e),s=o.shape;if(n){let i=lr(e,r.shape);s=yo(o.shape,i)}return R(o,s)}function qE(r,t,e=null){if(r.rank===0)return Ee(r);if(r.rank!==1&&e===null)return qE(R(r,[-1]),t,e);if(r.rank===1||typeof e==\"number\"||Array.isArray(e)&&e.length===1){if(t===1)return ft(Ee(r),e);if(t===1/0)return Ir(Ee(r),e);if(t===-1/0)return Ja(Ee(r),e);if(t===\"euclidean\"||t===2)return Se(ft(an(Ee(r),mt(2,\"int32\")),e));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(e)&&e.length===2){if(t===1)return Ir(ft(Ee(r),e[0]),e[1]-1);if(t===1/0)return Ir(ft(Ee(r),e[1]),e[0]);if(t===-1/0)return Ja(ft(Ee(r),e[1]),e[0]);if(t===\"fro\"||t===\"euclidean\")return Se(ft(Mt(r),e));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${e}`)}var Qa=T({norm_:Qq});function tK(r,t=null,e=!1){return Qa(r,\"euclidean\",t,e)}var Fx=T({euclideanNorm_:tK});function eK(r){let e={x:I(r,\"x\",\"exp\")};return k.runKernel(es,e)}var er=T({exp_:eK});function rK(r,t=0){let e=I(r,\"x\",\"expandDims\",\"string_or_numeric\");E(t<=e.rank,()=>\"Axis must be <= rank of the tensor\");let n={input:e},o={dim:t};return k.runKernel(ui,n,o)}var rr=T({expandDims_:rK});function nK(r){let e={x:I(r,\"x\",\"expm1\")};return k.runKernel(ya,e)}var Ox=T({expm1_:nK});function oK(r,t){let e=I(r,\"x\",\"tile\",\"string_or_numeric\");E(e.rank===t.length,()=>`Error in transpose: rank of input ${e.rank} must match length of reps ${t}.`);let n={x:e},o={reps:t};return k.runKernel(Jn,n,o)}var Dr=T({tile_:oK});function sK(r,t,e,n=\"float32\"){t==null&&(t=r);let o=wt([r,t],n),s=r<=t?r:t;for(let a=0;a<s;++a)o.set(1,a,a);let i=R(o.toTensor(),[r,t]);if(e==null)return i;if(e.length===1)return Dr(rr(i,0),[e[0],1,1]);if(e.length===2)return Dr(rr(rr(i,0),0),[e[0],e[1],1,1]);if(e.length===3)return Dr(rr(rr(rr(i,0),0),0),[e[0],e[1],e[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${e.length}D.`)}var Yu=T({eye_:sK});function iK(r){let e={x:I(r,\"x\",\"floor\",\"float32\")};return k.runKernel(rs,e)}var Pi=T({floor_:iK});function aK(r,t,e=0,n=0){let o=I(r,\"x\",\"gather\"),s=I(t,\"indices\",\"gather\",\"int32\"),i={x:o,indices:s},a={axis:e,batchDims:n};return k.runKernel(ci,i,a)}var Li=T({gather_:aK});function lK(r,t){let e=I(r,\"a\",\"greater\",\"string_or_numeric\"),n=I(t,\"b\",\"greater\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ca,o)}var Re=T({greater_:lK});function uK(r,t){let e=I(r,\"a\",\"greaterEqual\",\"string_or_numeric\"),n=I(t,\"b\",\"greaterEqual\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(ss,o)}var ln=T({greaterEqual_:uK});function cK(r){let e={x:I(r,\"x\",\"isFinite\")};return k.runKernel(Ia,e)}var Px=T({isFinite_:cK});function pK(r){let e={x:I(r,\"x\",\"isInf\")};return k.runKernel(Sa,e)}var Lx=T({isInf_:pK});function mK(r){let e={x:I(r,\"x\",\"isNaN\")};return k.runKernel(va,e)}var Mx=T({isNaN_:mK});function fK(r,t=.2){let n={x:I(r,\"x\",\"leakyRelu\")},o={alpha:t};return k.runKernel(is,n,o)}var Ql=T({leakyRelu_:fK});function dK(r,t){let e=I(r,\"a\",\"less\",\"string_or_numeric\"),n=I(t,\"b\",\"less\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Na,o)}var om=T({less_:dK});function hK(r,t){let e=I(r,\"a\",\"lessEqual\",\"string_or_numeric\"),n=I(t,\"b\",\"lessEqual\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ta,o)}var Ln=T({lessEqual_:hK});function KE(r,t,e){if(e<=0)throw new Error(\"The number of values should be positive.\");let n={start:r,stop:t,num:e};return k.runKernel(vp,{},n)}function gK(r,t=5,e=1,n=1,o=.5){let s=I(r,\"x\",\"localResponseNormalization\");E(s.rank===4||s.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got\n               rank ${s.rank}.`),E(na(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,a=!1;s.rank===3&&(a=!0,i=R(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let u={x:i},l={depthRadius:t,bias:e,alpha:n,beta:o},c=k.runKernel(Rl,u,l);return a?R(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var zx=T({localResponseNormalization_:gK});function xK(r){let e={x:I(r,\"x\",\"log\",\"float32\")};return k.runKernel(as,e)}var Sr=T({log_:xK});function yK(r){let e={x:I(r,\"x\",\"log1p\")};return k.runKernel(ka,e)}var tu=T({log1p_:yK});function bK(r){return E(oi(r),()=>\"The f passed in grad(f) must be a function\"),(t,e)=>{let n=I(t,\"x\",\"tf.grad\",\"string_or_numeric\"),o=e!=null?I(e,\"dy\",\"tf.grad\"):null;return k.tidy(()=>{let{value:s,grads:i}=k.gradients(()=>r(n),[n],o);return o!=null&&$e(s.shape,o.shape,\"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)\"),Vx(i),i[0]})}}function wK(r){return E(oi(r),()=>\"The f passed in grads(f) must be a function\"),(t,e)=>{E(Array.isArray(t),()=>\"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s\");let n=ja(t,\"args\",\"tf.grads\",\"string_or_numeric\"),o=e!=null?I(e,\"dy\",\"tf.grads\"):null;return k.tidy(()=>{let{value:s,grads:i}=k.gradients(()=>r(...n),n,o);return o!=null&&$e(s.shape,o.shape,\"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])\"),Vx(i),i})}}function CK(r){return E(oi(r),()=>\"The f passed in valueAndGrad(f) must be a function\"),(t,e)=>{E(t instanceof Ft,()=>\"The x passed in valueAndGrad(f)(x) must be a tensor\"),E(e==null||e instanceof Ft,()=>\"The dy passed in valueAndGrad(f)(x, dy) must be a tensor\");let{grads:n,value:o}=k.gradients(()=>r(t),[t],e);return Vx(n),{grad:n[0],value:o}}}function IK(r){return E(oi(r),()=>\"The f passed in valueAndGrads(f) must be a function\"),(t,e)=>{E(Array.isArray(t)&&t.every(o=>o instanceof Ft),()=>\"The args passed in valueAndGrads(f)(args) must be array of tensors\"),E(e==null||e instanceof Ft,()=>\"The dy passed in valueAndGrads(f)(args, dy) must be a tensor\");let n=k.gradients(()=>r(...t),t,e);return e!=null&&$e(n.value.shape,e.shape,\"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])\"),Vx(n.grads),n}}function Bx(r,t){E(oi(r),()=>\"The f passed in variableGrads(f) must be a function\"),E(t==null||Array.isArray(t)&&t.every(l=>l instanceof Ka),()=>\"The varList passed in variableGrads(f, varList) must be an array of variables\");let e=t!=null;if(!e){t=[];for(let l in k.registeredVariables)t.push(k.registeredVariables[l])}let n=e?t.filter(l=>!l.trainable):null,o=t.length;t=t.filter(l=>l.trainable),E(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${o} variables is trainable.`);let s=!0,{value:i,grads:a}=k.gradients(r,t,null,s);E(a.some(l=>l!=null),()=>\"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize().\"),E(i.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${i.rank} tensor`);let u={};return t.forEach((l,c)=>{a[c]!=null&&(u[l.name]=a[c])}),n!=null&&n.forEach(l=>u[l.name]=null),{value:i,grads:u}}function un(r){return k.customGrad(r)}function Vx(r){if(r.filter(e=>e==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that\n    the f you passed encloses all operations that lead from x to y.`)}function SK(r){let e={x:I(r,\"x\",\"softplus\")};return k.runKernel(Va,e)}var zs=T({softplus_:SK});function vK(r){let t=I(r,\"x\",\"logSigmoid\");return un(n=>({value:Ht(zs(Ht(n))),gradFunc:i=>D(i,Yr(Ht(n)))}))(t)}var Gx=T({logSigmoid_:vK});function NK(r,t){let e=I(r,\"a\",\"sub\"),n=I(t,\"b\",\"sub\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(Fs,o)}var ct=T({sub_:NK});function TK(r,t=-1){let e=I(r,\"logits\",\"logSoftmax\");if(t===-1&&(t=e.rank-1),t!==e.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${e.rank} and axis was ${t}`);return un((o,s)=>{let a=Ir(o,t,!0),u=ct(o,a),l=ct(J(u,\"float32\"),Sr(ft(er(u),t,!0)));return s([l]),{value:l,gradFunc:(p,m)=>{let[f]=m,d=!0,h=er(f);return ct(p,D(ft(p,t,d),h))}}})(e)}var sm=T({logSoftmax_:TK});function kK(r,t=null,e=!1){let n=I(r,\"x\",\"logSumExp\"),o=lr(t,n.shape),s=Ir(n,o,!0),i=ct(n,s),a=er(i),u=ft(a,o),l=Sr(u),c=X(R(s,l.shape),l);if(e){let p=yo(c.shape,o);return R(c,p)}return c}var im=T({logSumExp_:kK});function EK(r,t){let e=I(r,\"a\",\"logicalAnd\",\"bool\"),n=I(t,\"b\",\"logicalAnd\",\"bool\");Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ea,o)}var Rr=T({logicalAnd_:EK});function _K(r){let e={x:I(r,\"x\",\"logicalNot\",\"bool\")};return k.runKernel(_a,e)}var eu=T({logicalNot_:_K});function AK(r,t){let e=I(r,\"a\",\"logicalOr\",\"bool\"),n=I(t,\"b\",\"logicalOr\",\"bool\");Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Aa,o)}var am=T({logicalOr_:AK});function $K(r,t){let e=I(r,\"a\",\"logicalXor\",\"bool\"),n=I(t,\"b\",\"logicalXor\",\"bool\");return Pt(e.shape,n.shape),Rr(am(r,t),eu(Rr(r,t)))}var Wx=T({logicalXor_:$K});var Ux=2147483648;function DK(r,t,e=\"left\"){let n=I(r,\"sortedSequence\",\"searchSorted\"),o=I(t,\"values\",\"searchSorted\"),s=n.shape[n.shape.length-1],i=o.shape[o.shape.length-1],a=R(n,[-1,s]),u=R(o,[-1,i]);if(a.rank<2)throw new Error(\"Sorted input argument must be at least 2-dimensional\");if(a.shape[0]!==u.shape[0])throw new Error(\"Leading dimension of 'sortedSequence' and 'values' must match.\");if(Jt(u.shape)>=Ux)throw new Error(`values tensor size must less than ${Ux}`);if(a.shape[1]>=Ux)throw new Error(`trailing dim_size must less than ${Ux} for int32 output type, was ${a.shape[1]}`);let l={sortedSequence:a,values:u},c={side:e};return k.runKernel(Pp,l,c)}var mh=T({searchSorted_:DK});function jE(r,t){return mh(r,t,\"left\")}function RK(r,t,e,n,o){let s=I(r,\"x\",\"maxPool\"),i=1,a=s,u=!1;s.rank===3&&(u=!0,a=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(a.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${a.rank}.`),E(Ar(e,i),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${e} and dilations '${i}'`),Ie(\"maxPool\",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o},p=k.runKernel(cs,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var ru=T({maxPool_:RK});function FK(r,t=[1,1,1],e,n,o,s=\"NDHWC\"){let i=I(r,\"x\",\"maxPool3d\"),a=i,u=!1;i.rank===4&&(u=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(a.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${a.rank}.`),E(s===\"NDHWC\",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Ie(\"maxPool3d\",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o,dataFormat:s},p=k.runKernel(Fl,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var Hx=T({maxPool3d_:FK});function OK(r,t,e,n,o=!1){let i={x:I(r,\"x\",\"maxPoolWithArgmax\")},a={filterSize:t,strides:e,pad:n,includeBatchInIndex:o},u=k.runKernel(Ep,i,a);return{result:u[0],indexes:u[1]}}var XE=T({maxPoolWithArgmax_:OK});function PK(r,t){let e=I(r,\"a\",\"maximum\"),n=I(t,\"b\",\"maximum\");[e,n]=Ut(e,n),e.dtype===\"bool\"&&(e=J(e,\"int32\"),n=J(n,\"int32\")),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(us,o)}var Sn=T({maximum_:PK});function LK(r,t=null,e=!1){let o={x:I(r,\"x\",\"mean\")},s={axis:t,keepDims:e};return k.runKernel(ps,o,s)}var ve=T({mean_:LK});function Ne(r,t=\"float32\"){if(t===\"complex64\"){let n=Ne(r,\"float32\"),o=Ne(r,\"float32\");return wn(n,o)}let e=ip(Jt(r),t);return k.makeTensor(e,r,t)}function cr(r,t=\"float32\"){if(t===\"complex64\"){let n=cr(r,\"float32\"),o=Ne(r,\"float32\");return wn(n,o)}let e=Wd(Jt(r),t);return k.makeTensor(e,r,t)}function YE(r,t,{indexing:e=\"xy\"}={}){if(e!==\"xy\"&&e!==\"ij\")throw new TypeError(`${e} is not a valid third argument to meshgrid`);if(r===void 0)return[];let n=I(r,\"x\",\"meshgrid\",r instanceof Ft?r.dtype:\"float32\");if(t===void 0)return[n];let o=I(t,\"y\",\"meshgrid\",t instanceof Ft?t.dtype:\"float32\"),s=Jt(n.shape),i=Jt(o.shape);return e===\"xy\"?(n=R(n,[1,-1]),o=R(o,[-1,1]),[Lt(cr([i,1],n.dtype),n),Lt(o,cr([1,s],o.dtype))]):(n=R(n,[-1,1]),o=R(o,[1,-1]),[Lt(n,cr([1,i],n.dtype)),Lt(cr([s,1],o.dtype),o)])}function MK(r,t){let e=I(r,\"a\",\"minimum\"),n=I(t,\"b\",\"minimum\");[e,n]=Ut(e,n),e.dtype===\"bool\"&&(e=J(e,\"int32\"),n=J(n,\"int32\")),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(fs,o)}var Mi=T({minimum_:MK});function zK(r,t,e){E(e===\"reflect\"||e===\"symmetric\",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${e}.`);let n=I(r,\"x\",\"mirrorPad\");if(n.rank===0)throw new Error(\"mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad\");E(t.length===n.rank,()=>`Padding doesn't match input. Must be ${n.rank}. Got ${t.length}.`);let o=e===\"reflect\"?1:0;for(let a=0;a<n.rank;a++)E(t[a].length===2,()=>\"Invalid number of paddings. Must be length of 2 each.\"),E(t[a][0]>=0&&t[a][0]<=n.shape[a]-o&&t[a][1]>=0&&t[a][1]<=n.shape[a]-o,()=>`Padding in dimension ${a} cannot be greater than or equal to ${n.shape[a]-o} or less than 0 for input of shape ${n.shape}`);let s={paddings:t,mode:e},i={x:n};return k.runKernel(ds,i,s)}var qx=T({mirrorPad_:zK});function BK(r,t){let e=I(r,\"a\",\"mod\"),n=I(t,\"b\",\"mod\");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel($a,o)}var Kx=T({mod_:BK});function VK(r,t=null,e=!1){r=I(r,\"x\",\"moments\");let n=lr(t,r.shape),o=ve(r,n,e),s=o.shape;e||(s=yo(o.shape,n));let i=Mt(ct(J(r,\"float32\"),R(o,s))),a=ve(i,n,e);return{mean:o,variance:a}}var Zu=T({moments_:VK});function GK(r,t,e,n){let o=I(t,\"data\",\"multiRNNCell\"),s=ja(e,\"c\",\"multiRNNCell\"),i=ja(n,\"h\",\"multiRNNCell\"),a=o,u=[];for(let p=0;p<r.length;p++){let m=r[p](a,s[p],i[p]);u.push(m[0]),u.push(m[1]),a=m[1]}let l=[],c=[];for(let p=0;p<u.length;p+=2)l.push(u[p]),c.push(u[p+1]);return[l,c]}var ZE=T({multiRNNCell_:GK});function WK(r,t,e,n=!1){let o=I(r,\"logits\",\"multinomial\"),s=o.size,i=o.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);e=e||Math.random();let u={logits:i===1?R(o,[1,-1]):o},l={numSamples:t,seed:e,normalized:n},c=k.runKernel(_p,u,l);return i===1?R(c,[c.size]):c}var JE=T({multinomial_:WK});function UK(r,t){let e=I(r,\"a\",\"notEqual\",\"string_or_numeric\"),n=I(t,\"b\",\"notEqual\",\"string_or_numeric\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Da,o)}var Bs=T({notEqual_:UK});function HK(r){let e={x:I(r,\"x\",\"onesLike\")};return k.runKernel(mi,e)}var yr=T({onesLike_:HK});function qK(r,t){let e=I(r,\"v1\",\"outerProduct\"),n=I(t,\"v2\",\"outerProduct\");E(e.rank===1&&n.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${e.rank} and ${n.rank}.`);let o=R(e,[-1,1]),s=R(n,[1,-1]);return Lt(o,s)}var QE=T({outerProduct_:qK});function KK(r,t,e=0){let n=I(r,\"x\",\"pad\");if(n.rank===0)throw new Error(\"pad(scalar) is not defined. Pass non-scalar to pad\");let o={paddings:t,constantValue:e},s={x:n};return k.runKernel(xs,s,o)}var cn=T({pad_:KK});function jK(r,t,e=0){return E(t.length===2,()=>\"Invalid number of paddings. Must be length of 2.\"),cn(r,[t],e)}var t_=T({pad1d_:jK});function XK(r,t,e=0){return E(t.length===2&&t[0].length===2&&t[1].length===2,()=>\"Invalid number of paddings. Must be length of 2 each.\"),cn(r,t,e)}var e_=T({pad2d_:XK});function YK(r,t,e=0){return E(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>\"Invalid number of paddings. Must be length of 2 each.\"),cn(r,t,e)}var r_=T({pad3d_:YK});function ZK(r,t,e=0){return E(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>\"Invalid number of paddings. Must be length of 2 each.\"),cn(r,t,e)}var n_=T({pad4d_:ZK});function JK(r,t,e){let n=I(r,\"x\",\"spaceToBatchND\");E(n.rank>=1+t.length,()=>`input rank ${n.rank} should be > than [blockShape] ${t.length}`),E(e.length===t.length,()=>`paddings.shape[0] ${e.length} must be equal to [blockShape] ${t.length}`),E(n.shape.reduce((i,a,u)=>u>0&&u<=t.length?i&&(a+e[u-1][0]+e[u-1][1])%t[u-1]===0:i,!0),()=>`input spatial dimensions ${n.shape.slice(1)} with paddings ${e.toString()} must be divisible by blockShapes ${t.toString()}`);let o={x:n},s={blockShape:t,paddings:e};return k.runKernel(xi,o,s)}var nu=T({spaceToBatchND_:JK});function QK(r,t,e,n,o,s,i){o==null&&(o=[1,1]),s==null&&(s=1),n===0&&(n=\"valid\");let a=I(r,\"x\",\"maxPool\"),u=a,l=!1;a.rank===3&&(l=!0,u=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(Ar(s,o),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${o}'`);let c=Y0(u.shape,t,s,o,n),p=[c.dilationHeight,c.dilationWidth],m;n===\"same\"?m=ej([c.filterHeight,c.filterWidth],p):m=[[0,0],[0,0]];let f=p[0]===1&&p[1]===1,[d,h]=tj([c.inHeight,c.inWidth],p,m),g=f?n:\"valid\",x=f?u:nu(u,p,d),w=(e===\"avg\"?()=>Yl(x,t,s,g,i):()=>ru(x,t,s,g,i))(),C=f?w:Zl(w,p,h);return l?R(C,[C.shape[1],C.shape[2],C.shape[3]]):C}function tj(r,t,e){let n=e.map(c=>c[0]),o=e.map(c=>c[1]),s=r.concat(n,o),i=t.map((c,p)=>(c-s[p]%c)%c),a=o.map((c,p)=>c+i[p]),u=t.map((c,p)=>[n[p],a[p]]),l=t.map((c,p)=>[0,i[p]]);return[u,l]}function ej(r,t){let n=r.map((i,a)=>i+(i-1)*(t[a]-1)).map(i=>i-1),o=n.map(i=>Math.floor(i/2)),s=n.map((i,a)=>i-o[a]);return n.map((i,a)=>[o[a],s[a]])}var jx=T({pool_:QK});function rj(r,t){let e=I(r,\"x\",\"prelu\"),n=I(t,\"alpha\",\"prelu\"),o={x:e,alpha:n};return k.runKernel(bs,o)}var ou=T({prelu_:rj});function nj(r,t=null,e=!1){let n=I(r,\"x\",\"prod\");n.dtype===\"bool\"&&(n=J(n,\"int32\"));let o={x:n},s={axis:t,keepDims:e};return k.runKernel(ws,o,s)}var Xx=T({prod_:nj});function oj(r,t,e,n){let o=r.map((c,p)=>I(c,`tensors${p}`,\"raggedGather\",\"int32\")),s=I(t,\"paramsDenseValues\",\"raggedGather\"),i=I(e,\"indices\",\"raggedGather\",\"int32\"),a={paramsNestedSplits:o,paramsDenseValues:s,indices:i},u={outputRaggedRank:n},l=k.runKernel(Ap,a,u);return{outputNestedSplits:l.slice(0,l.length-1),outputDenseValues:l[l.length-1]}}var o_=T({raggedGather_:oj});function sj(r,t,e){let n=I(r,\"starts\",\"raggedRange\"),o=I(t,\"limits\",\"raggedRange\",n.dtype),s=I(e,\"deltas\",\"raggedRange\",n.dtype),i={starts:n,limits:o,deltas:s},a=k.runKernel($p,i);return{rtNestedSplits:a[0],rtDenseValues:a[1]}}var s_=T({raggedRange_:sj});function ij(r,t,e,n,o){let s=I(r,\"shape\",\"raggedTensorToTensor\",\"int32\"),i=I(t,\"values\",\"raggedTensorToTensor\"),a=I(e,\"defaultValue\",\"raggedTensorToTensor\",i.dtype),u=n.map((p,m)=>I(p,`tensors${m}`,\"raggedTensorToTensor\",\"int32\")),l={shape:s,values:i,defaultValue:a,rowPartitionTensors:u},c={rowPartitionTypes:o};return k.runKernel(Dp,l,c)}var i_=T({raggedTensorToTensor_:ij});function aj(r,t,e){let n=Jt(r),o=null;if(e==null||e===\"float32\")o=new Float32Array(n);else if(e===\"int32\")o=new Int32Array(n);else if(e===\"bool\")o=new Uint8Array(n);else throw new Error(`Unknown data type ${e}`);for(let s=0;s<n;s++)o[s]=t();return k.makeTensor(o,r,e)}var a_=T({rand_:aj});var Qx=Tl(fh());var Qu=class{constructor(t,e,n,o,s){this.mean=t,this.stdDev=e,this.dtype=n,this.nextVal=NaN,this.truncated=o,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let i=s||Math.random();this.random=Qx.alea(i.toString())}nextValue(){if(!isNaN(this.nextVal)){let o=this.nextVal;return this.nextVal=NaN,o}let t,e,n=!1;for(;!n;){let o,s,i;do o=2*this.random()-1,s=2*this.random()-1,i=o*o+s*s;while(i>=1||i===0);let a=Math.sqrt(-2*Math.log(i)/i);t=this.mean+this.stdDev*o*a,e=this.mean+this.stdDev*s*a,(!this.truncated||this.isValidTruncated(t))&&(n=!0)}return(!this.truncated||this.isValidTruncated(e))&&(this.nextVal=this.convertValue(e)),this.convertValue(t)}convertValue(t){return this.dtype==null||this.dtype===\"float32\"?t:Math.round(t)}isValidTruncated(t){return t<=this.upper&&t>=this.lower}},Zx=class{constructor(t,e,n,o){this.alpha=t,this.beta=1/e,this.dtype=n;let s=o||Math.random();this.randu=Qx.alea(s.toString()),this.randn=new Qu(0,1,n,!1,this.randu()),t<1?this.d=t+2/3:this.d=t-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let t,e,n,o,s,i;for(;;){do o=this.randn.nextValue(),i=1+this.c*o;while(i<=0);if(i*=i*i,t=o*o,e=1-.331*t*t,n=.5*t+this.d*(1-i+Math.log(i)),s=this.randu(),s<e||Math.log(s)<n)break}return i=1/this.beta*this.d*i,this.alpha<1&&(i*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(i)}convertValue(t){return this.dtype===\"float32\"?t:Math.round(t)}},Jx=class{constructor(t=0,e=1,n,o){if(this.canReturnFloat=()=>this.dtype==null||this.dtype===\"float32\",this.min=t,this.range=e-t,this.dtype=n,o==null&&(o=Math.random()),typeof o==\"number\"&&(o=o.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${t} - ${e} <= 1 and dtype is not float`);this.random=Qx.alea(o)}convertValue(t){return this.canReturnFloat()?t:Math.round(t)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function dj(r,t,e=1,n=\"float32\",o){if(e==null&&(e=1),n==null&&(n=\"float32\"),n!==\"float32\"&&n!==\"int32\")throw new Error(`Unsupported data type ${n}`);let s=new Zx(t,e,n,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var v_=T({randomGamma_:dj});function hj(r,t=0,e=1,n,o){if(n!=null&&n===\"bool\")throw new Error(`Unsupported data type ${n}`);let s=new Qu(t,e,n,!1,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var tc=T({randomNormal_:hj});function gj(r,t,e){if(t!=null&&t===\"bool\")throw new Error(`Unsupported data type ${t}`);return tc(r,0,1,t,e)}var N_=T({randomStandardNormal_:gj});function xj(r,t=0,e=1,n=\"float32\",o){let s=wt(r,n),i=new Jx(t,e,null,o);for(let a=0;a<s.values.length;a++)s.values[a]=i.nextValue();return s.toTensor()}var zi=T({randomUniform_:xj});function su(r,t,e=1,n=\"float32\"){if(e===0)throw new Error(\"Cannot have a step of zero\");let o={start:r,stop:t,step:e,dtype:n};return k.runKernel(Ol,{},o)}function yj(r){let e={x:I(r,\"x\",\"reciprocal\")};return k.runKernel(Pa,e)}var ty=T({reciprocal_:yj});function bj(r){let e={x:I(r,\"x\",\"relu\")};return k.runKernel(Cs,e)}var Fr=T({relu_:bj});function wj(r){let e={x:I(r,\"x\",\"relu6\")};return k.runKernel(vs,e)}var lm=T({relu6_:wj});function Cj(r,t){let n={x:I(r,\"x\",\"reverse\")},o={dims:t};return k.runKernel(Ns,n,o)}var pr=T({reverse_:Cj});function Ij(r){let t=I(r,\"x\",\"reverse\");return E(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),pr(t,0)}var T_=T({reverse1d_:Ij});function Sj(r,t){let e=I(r,\"x\",\"reverse\");return E(e.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${e.rank}.`),pr(e,t)}var k_=T({reverse2d_:Sj});function vj(r,t){let e=I(r,\"x\",\"reverse\");return E(e.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${e.rank}.`),pr(e,t)}var E_=T({reverse3d_:vj});function Nj(r,t){let e=I(r,\"x\",\"reverse\");return E(e.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${e.rank}.`),pr(e,t)}var __=T({reverse4d_:Nj});function Tj(r){let e={x:I(r,\"x\",\"round\")};return k.runKernel(Ts,e)}var um=T({round_:Tj});function kj(r){let e={x:I(r,\"x\",\"rsqrt\",\"float32\")};return k.runKernel(ks,e)}var cm=T({rsqrt_:kj});function Ej(r){let e={x:I(r,\"x\",\"selu\")};return k.runKernel(Ma,e)}var pm=T({selu_:Ej});function _j(r,t,e,n,o,s=[1,1],i=\"NHWC\"){let a=I(r,\"x\",\"separableConv2d\"),u=I(t,\"depthwiseFilter\",\"separableConv2d\"),l=I(e,\"pointwiseFilter\",\"separableConv2d\"),c=a,p=!1;if(a.rank===3&&(p=!0,c=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),i===\"NCHW\")throw new Error(\"separableConv2d currently does not support dataFormat NCHW; only NHWC is supported\");E(c.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${c.rank}.`),E(u.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${u.rank}.`),E(l.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${u.rank}.`),E(l.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${l.shape[0]}.`),E(l.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${l.shape[1]}.`);let m=u.shape[2],f=u.shape[3];E(l.shape[2]===m*f,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${m*f}, but got ${l.shape[2]}.`);let d=Fi(c,u,n,o,i,s),g=In(d,l,1,\"valid\",i);return p?R(g,[g.shape[1],g.shape[2],g.shape[3]]):g}var mm=T({separableConv2d_:_j});async function Aj(r,t){let e=I(r,\"x\",\"setdiff1d\"),n=I(t,\"y\",\"setdiff1d\");E(e.dtype===n.dtype,()=>`x and y should have the same dtype, but got x (${e.dtype}) and y (${n.dtype}).`),E(e.rank===1,()=>`x should be 1D tensor, but got x (${e.shape}).`),E(n.rank===1,()=>`y should be 1D tensor, but got y (${n.shape}).`);let o=await e.data(),s=await n.data(),i=new Set(s),a=0;for(let c=0;c<o.length;c++)i.has(o[c])||a++;let u=new pe([a],e.dtype),l=new pe([a],\"int32\");for(let c=0,p=0;c<o.length;c++)i.has(o[c])||(u.values[p]=o[c],l.values[p]=c,p++);return[u.toTensor(),l.toTensor()]}var A_=Aj;function $j(r){let e={x:I(r,\"x\",\"sign\")};return k.runKernel(Ba,e)}var ey=T({sign_:$j});function Dj(r){let e={x:I(r,\"x\",\"sin\",\"float32\")};return k.runKernel(Es,e)}var fm=T({sin_:Dj});function Rj(r){let e={x:I(r,\"x\",\"sinh\")};return k.runKernel(za,e)}var dm=T({sinh_:Rj});function Fj(r,t,e){let n=I(r,\"x\",\"slice1d\");return E(n.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${n.rank} tensor`),Rt(n,[t],[e])}var hm=T({slice1d_:Fj});function Oj(r,t,e){let n=I(r,\"x\",\"slice2d\");return E(n.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var dh=T({slice2d_:Oj});function Pj(r,t,e){let n=I(r,\"x\",\"slice3d\");return E(n.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var gm=T({slice3d_:Pj});function Lj(r,t,e){let n=I(r,\"x\",\"slice4d\");return E(n.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var ec=T({slice4d_:Lj});function Mj(r,t=-1){let e=I(r,\"logits\",\"softmax\",\"float32\");if(t===-1&&(t=e.rank-1),t!==e.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${e.rank} and dim was ${t}`);let n={logits:e},o={dim:t};return k.runKernel(Ds,n,o)}var iu=T({softmax_:Mj});function zj(r){E(r.dtype===\"complex64\",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${r.dtype}.`);let t={input:r};return k.runKernel(Cp,t)}var au=T({fft_:zj});function Bj(r){E(r.dtype===\"complex64\",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${r.dtype}.`);let t={input:r};return k.runKernel(Ip,t)}var tl=T({ifft_:Bj});function Vj(r){let t=r.shape[r.shape.length-1],e=r.size/t,n;if(t<=2){let o=R(r,[e,t]);n=tl(o)}else{let o=[e,2*(t-1)],s=R(Za(r),[e,t]),i=R(Xl(r),[e,t]),a=pr(Rt(s,[0,1],[e,t-2]),1),u=D(pr(Rt(i,[0,1],[e,t-2]),1),mt(-1)),l=ne([s,a],1),c=ne([i,u],1),p=R(wn(l,c),[o[0],o[1]]);n=tl(p)}if(n=Za(n),r.rank===3&&r.shape[0]!==0){let o=n,s=r.shape[0];n=R(n,[s,n.shape[0]/s,n.shape[1]]),o.dispose()}return n}var xm=T({irfft_:Vj});function Gj(r,t,e=0){let o={x:I(r,\"x\",\"split\")},s={numOrSizeSplits:t,axis:e};return k.runKernel(yi,o,s)}var mr=T({split_:Gj});function Wj(r,t){E(r.dtype===\"float32\",()=>`The dtype for rfft() must be real value but got ${r.dtype}`);let e=r.shape[r.shape.length-1],n=r.size/e,o;if(t!=null&&t<e){let d=r.shape.map(g=>0),h=r.shape.map(g=>g);h[r.shape.length-1]=t,o=Rt(r,d,h),e=t}else if(t!=null&&t>e){let d=r.shape.map(h=>h);d[r.shape.length-1]=t-e,o=ne([r,Ne(d)],r.shape.length-1),e=t}else o=r;let s=It(o),i=R(wn(o,s),[n,e]),a=au(i),u=Math.floor(e/2)+1,l=Za(a),c=Xl(a),p=mr(l,[u,e-u],l.shape.length-1),m=mr(c,[u,e-u],c.shape.length-1),f=o.shape.slice();return f[o.shape.length-1]=u,R(wn(p[0],m[0]),f)}var lu=T({rfft_:Wj});function Uj(r,t){let e=I(r,\"a\",\"squaredDifference\"),n=I(t,\"b\",\"squaredDifference\");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n},s={};return k.runKernel(Rs,o,s)}var ym=T({squaredDifference_:Uj});function Hj(r,t){let e=I(r,\"x\",\"squeeze\",\"string_or_numeric\");return R(e,t0(e.shape,t).newShape)}var Mn=T({squeeze_:Hj});function qj(r,t=0){let e=ja(r,\"tensors\",\"stack\",\"string_or_numeric\");E(e.length>=1,()=>\"Pass at least one tensor to tf.stack\"),e.length>0&&E(t<=e[0].rank,()=>\"Axis must be <= rank of the tensor\");let n=e,o={axis:t};return k.runKernel(fi,n,o)}var nr=T({stack_:qj});function Kj(r,t=0){let n={x:I(r,\"x\",\"step\")},o={alpha:t};return k.runKernel(po,n,o)}var bo=T({step_:Kj});function jj(r,t,e,n,o=0,s=0,i=0,a=0,u=0){let c={x:I(r,\"x\",\"stridedSlice\",\"string_or_numeric\")},p={begin:t,end:e,strides:n,beginMask:o,endMask:s,ellipsisMask:i,newAxisMask:a,shrinkAxisMask:u};return k.runKernel(Wa,c,p)}var ry=T({stridedSlice_:jj});function Xj(r){let e={x:I(r,\"x\",\"tan\",\"float32\")};return k.runKernel(Os,e)}var ny=T({tan_:Xj});function Me(r,t){Yn(r);let e=Br(r,t);if(e.length!==1)throw new Error(\"tensor1d() requires values to be a flat/TypedArray\");return on(r,null,e,t)}function Vs(r,t,e){if(Yn(r),t!=null&&t.length!==2)throw new Error(\"tensor2d() requires shape to have two numbers\");let n=Br(r,e);if(n.length!==2&&n.length!==1)throw new Error(\"tensor2d() requires values to be number[][] or flat/TypedArray\");if(n.length===1&&t==null)throw new Error(\"tensor2d() requires shape to be provided when `values` are a flat/TypedArray\");return on(r,t,n,e)}function $_(r,t,e){if(Yn(r),t!=null&&t.length!==4)throw new Error(\"tensor4d() requires shape to have four numbers\");let n=Br(r,e);if(n.length!==4&&n.length!==1)throw new Error(\"tensor4d() requires values to be number[][][][] or flat/TypedArray\");if(n.length===1&&t==null)throw new Error(\"tensor4d() requires shape to be provided when `values` are a flat array\");return on(r,t,n,e)}function D_(r,t,e){if(Yn(r),t!=null&&t.length!==5)throw new Error(\"tensor5d() requires shape to have five numbers\");let n=Br(r,e);if(n.length!==5&&n.length!==1)throw new Error(\"tensor5d() requires values to be number[][][][][] or flat/TypedArray\");if(n.length===1&&t==null)throw new Error(\"tensor5d() requires shape to be provided when `values` are a flat array\");return on(r,t,n,e)}function R_(r,t,e){if(Yn(r),t!=null&&t.length!==6)throw new Error(\"tensor6d() requires shape to have six numbers\");let n=Br(r,e);if(n.length!==6&&n.length!==1)throw new Error(\"tensor6d() requires values to be number[][][][][][] or flat/TypedArray\");if(n.length===1&&t==null)throw new Error(\"tensor6d() requires shape to be provided when `values` are a flat array\");return t=t||n,on(r,t,n,e)}function Yj(r,t=1,e=!0){let n=I(r,\"x\",\"topk\");if(n.rank===0)throw new Error(\"topk() expects the input to be of rank 1 or higher\");let o=n.shape[n.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>o)throw new Error(`'k' passed to topk() must be <= the last dimension (${o}) but got ${t}`);let s={x:n},i={k:t,sorted:e},[a,u]=k.runKernel(Ua,s,i);return{values:a,indices:u}}var oy=T({topk_:Yj});function Zj(r,t=0,e=1,n,o){if(n!=null&&n===\"bool\")throw new Error(\"Unsupported data type $ { dtype }\");let s=new Qu(t,e,n,!0,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var bm=T({truncatedNormal_:Zj});function Jj(r,t=0){let e=I(r,\"x\",\"unique\",\"string_or_numeric\");E(e.rank>0,()=>\"The input tensor must be at least 1D\");let n={x:e},o={axis:t},[s,i]=k.runKernel(Mp,n,o);return{values:s,indices:i}}var sy=T({unique_:Jj});function Qj(r,t,e){let n=I(r,\"x\",\"unsortedSegmentSum\"),o=I(t,\"segmentIds\",\"unsortedSegmentSum\",\"int32\");E(na(e),()=>\"numSegments must be of dtype int\");let s={x:n,segmentIds:o},i={numSegments:e};return k.runKernel(Wl,s,i)}var wm=T({unsortedSegmentSum_:Qj});function t6(r,t=0){let e=I(r,\"x\",\"unstack\",\"string_or_numeric\");E(t>=-e.shape.length&&t<e.shape.length,()=>`Axis = ${t} is not in [-${e.shape.length}, ${e.shape.length})`);let n={value:e},o={axis:t};return k.runKernel(bi,n,o)}var vr=T({unstack_:t6});function F_(r,t){return mh(r,t,\"right\")}function iy(r,t=!0,e,n){return k.makeVariable(r,t,e,n)}function ay(r,t){let e=[];for(let s=0;s<t.length;s++)t[s]&&e.push(s);let n=wt(r,\"int32\"),o=wt([e.length,r.length],\"int32\");for(let s=0;s<e.length;s++){let i=n.indexToLoc(e[s]),a=s*r.length;o.values.set(i,a)}return o.toTensor()}async function e6(r){let t=I(r,\"condition\",\"whereAsync\",\"bool\"),e=await t.data(),n=ay(t.shape,e);return r!==t&&t.dispose(),n}var ly=e6;async function r6(r,t,e){let n=I(r,\"tensor\",\"boolMask\"),o=I(t,\"mask\",\"boolMask\",\"bool\"),s=e==null?0:e,i=o.rank,a=n.shape;E(i>0,()=>\"mask cannot be scalar\"),$e(a.slice(s,s+i),o.shape,\"mask's shape must match the first K dimensions of tensor's shape,\");let u=1;for(let h=s;h<s+i;h++)u*=a[h];let l=a.slice(0,s).concat([u],a.slice(s+i)),c=R(n,l),p=R(o,[-1]),m=await ly(p),f=Mn(m,[1]),d=Li(c,f,s);return r!==n&&n.dispose(),t!==o&&o.dispose(),f.dispose(),c.dispose(),p.dispose(),m.dispose(),d}var n6=r6;function o6(r,t,e,n,o=!0){let s=I(r,\"v\",\"movingAverage\"),i=I(t,\"x\",\"movingAverage\"),a=I(e,\"decay\",\"movingAverage\");I0(s,i),E(Dn(s.shape,i.shape),()=>\"Shape mismatch in v and x\");let u=mt(1),l=ct(u,a),c=D(ct(i,s),l);if(o){E(n!=null,()=>\"When using zeroDebias: true, step is required.\");let p=I(n,\"step\",\"movingAverage\");c=pt(c,ct(u,an(a,p)))}return X(s,c)}var s6=T({movingAverage_:o6});function i6(r,t,e){let n=I(r,\"indices\",\"scatterND\",\"int32\"),o=I(t,\"updates\",\"scatterND\");sx(o,n,e);let s={indices:n,updates:o},i={shape:e};return k.runKernel(La,s,i)}var a6=T({scatterND_:i6});function O_(r,t,e,n){if(r.dtype!==\"int32\")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${r.dtype}.`);if(r.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${r.shape}.`);let o=r.rank>0?r.shape[0]:1,s=r.rank>1?r.shape[1]:1;if(e.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${e.length}, should be: ${s}.`);let i=t.size;if(!(t.rank===0||t.rank===1&&i===o))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${o}]`);if(t.dtype!==n.dtype)throw new Error(\"sparseValues.dtype must match defaultValues.dtype\")}function u6(r,t,e,n=0){let o=I(r,\"sparseIndices\",\"sparseToDense\",\"int32\"),s=I(t,\"sparseValues\",\"sparseToDense\",\"string_or_numeric\"),i=I(n,\"defaultValue\",\"sparseToDense\",s.dtype);O_(o,s,e,i);let a={sparseIndices:o,sparseValues:s,defaultValue:i},u={outputShape:e};return k.runKernel(Lp,a,u)}var c6=T({sparseToDense_:u6});function p6(r,t){let e=I(t,\"indices\",\"gatherND\",\"int32\"),o={params:I(r,\"x\",\"gatherND\",\"string_or_numeric\"),indices:e};return k.runKernel(wa,o)}var m6=T({gatherND_:p6});function P_(r,t){if(t==null)return r.shape.slice();if(Dn(r.shape,t))return t;if(r.shape.length===t.length){let e=[];for(let n=0;n<r.shape.length;n++)t[n]==null&&r.shape[n]!=null?e.push(r.shape[n]):e.push(t[n]);return e}return t}function f6(r,t,e,n){let o=I(r,\"x\",\"dropout\");if(E(o.dtype===\"float32\",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${o.dtype} tensor instead.`),E(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return r instanceof Ft?o.clone():o;let s=P_(o,e),i=1-t,a=pt(Pi(X(zi(s,0,1,\"float32\",n),i)),i);return D(o,a)}var lv=T({dropout_:f6});function uv(r){return Math.floor(Math.pow(2,Math.ceil(Math.log(r)/Math.log(2))))}function hh(r,t,e){let n=1-r%2,o=new Float32Array(r);for(let s=0;s<r;++s){let i=2*Math.PI*s/(r+n-1);o[s]=t-e*Math.cos(i)}return Me(o,\"float32\")}async function d6(r,t,e=1){let n=I(r,\"predictions\",\"inTopK\"),o=I(t,\"targets\",\"inTopK\");E(n.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${n.rank}`),E(n.rank-1===o.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${n.rank} and targets rank ${o.rank}`),$e(n.shape.slice(0,n.shape.length-1),o.shape,\"predictions's shape should be align with the targets' shape, except the last dimension.\");let s=n.shape[n.shape.length-1];E(e>0&&e<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${e}`);let i=await n.data(),a=await o.data(),[u,l]=[i.length/s,s],c=e0(\"bool\",u);for(let p=0;p<u;p++){let m=p*l,f=i.subarray(m,m+l),d=[];for(let h=0;h<f.length;h++)d.push({value:f[h],index:h});d.sort((h,g)=>g.value-h.value),c[p]=0;for(let h=0;h<e;h++)if(d[h].index===a[p]){c[p]=1;break}}return r!==n&&n.dispose(),t!==o&&o.dispose(),ur(c,o.shape,\"bool\")}var h6=d6;var uu={};Wt(uu,{conv2d:()=>L_,depthwiseConv2d:()=>M_,matMul:()=>z_});function g6(r,t,e,n,o,s=\"NHWC\",i){let a=r;r.rank===3&&(a=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let u=t;u.rank===3&&(u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]])),E(a.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${a.shape}.`),E(u.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${u.shape}.`),E(e.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${e}.`);let l=s===\"NHWC\"?a.shape[3]:a.shape[1],c=s===\"NHWC\"?u.shape[3]:u.shape[1];E(l===e[2],()=>`Error in conv2dDerFilter: depth of input ${l}) must match input depth in filter (${e[2]}.`),E(c===e[3],()=>`Error in conv2dDerFilter: depth of dy (${c}) must match output depth for filter (${e[3]}).`),Ie(\"conv2dDerFilter\",o,i);let p={x:a,dy:u},m={strides:n,pad:o,dataFormat:s,dimRoundingMode:i,filterShape:e};return k.runKernel(mp,p,m)}var Cm=T({conv2DBackpropFilter_:g6});function rc(r,t,e){if(e==null||e===\"linear\")return r;if(e===\"relu\")return D(r,bo(t));throw new Error(`Cannot compute gradient for fused activation ${e}.`)}function nc(r,t){let e=t,n=ge(r.shape,t.shape);return n.length>0&&(e=ft(e,n)),R(e,r.shape)}function oc(r,t,e,n){if(t===\"linear\")return r;if(t===\"relu\")return Fr(r);if(t===\"elu\")return Oi(r);if(t===\"relu6\")return lm(r);if(t===\"prelu\")return ou(r,e);if(t===\"leakyrelu\")return Ql(r,n);if(t===\"sigmoid\")return Yr(r);throw new Error(`Unknown fused activation ${t}.`)}var sc=(r,t)=>!(r>0)||t===\"linear\";function x6({x:r,filter:t,strides:e,pad:n,dataFormat:o=\"NHWC\",dilations:s=[1,1],dimRoundingMode:i,bias:a,activation:u=\"linear\",preluActivationWeights:l,leakyreluAlpha:c}){if(u=u||\"linear\",sc(k.state.gradientDepth,u)===!1){E(o===\"NHWC\",()=>`Error in fused conv2d: got dataFormat of ${o} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let _=In(r,t,e,n,o,s,i);return a!=null&&(_=X(_,a)),oc(_,u,l,c)}let p=I(r,\"x\",\"conv2d\",\"float32\"),m=I(t,\"filter\",\"conv2d\",\"float32\"),f=p,d=!1;p.rank===3&&(d=!0,f=R(p,[1,p.shape[0],p.shape[1],p.shape[2]])),E(f.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${f.rank}.`),E(m.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${m.rank}.`),Ie(\"fused conv2d\",n,i);let h=o===\"NHWC\"?f.shape[3]:f.shape[1];E(m.shape[2]===h,()=>`Error in conv2d: depth of input (${h}) must match input depth for filter ${m.shape[2]}.`),E(Ar(e,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`);let g=ju(f.shape,m.shape,e,s,n,i),x;a!=null&&(x=I(a,\"bias\",\"fused conv2d\"),[x]=Ut(x,p),o===\"NHWC\"?Pt(g.outShape,x.shape):(E(x.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${x.shape.length}.`),E(x.shape.length===0||x.shape[0]===g.outChannels||x.shape[0]===1,()=>`Error in fused conv2d: bias shape (${x.shape}) is not compatible with the number of output channels (${g.outChannels})`)));let b;if(l!=null){let _=l.shape;if(E(_.length<=1||_.length===3,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${_.length}.`),_.length===1)E(_[0]===1||_[0]===g.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${_}) is not compatible with the number of output channels (${g.outChannels}).`);else if(_.length===3)try{Pt(_,g.outShape)}catch(A){let $=`Error in fused conv2d: PReLU activation weights (${_}) is not compatible with the output shape of the conv2d (${g.outShape}).`;throw Error($)}b=I(l,\"prelu weights\",\"fused conv2d\")}let w=(_,A)=>{E(o===\"NHWC\",()=>`Error in gradient of fused conv2D: got dataFormat of ${o} but only NHWC is currently supported.`);let[$,F,P,V]=A,G=rc(_,P,u);E(to(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let W=tm(F.shape,G,$,e,n),q=Cm(F,G,$.shape,e,n),H=[W,q];if(V!=null){let j=nc(V,G);H.push(j)}return H},C={x:f,filter:m,bias:x,preluActivationWeights:b},N={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};return a==null?un((A,$,F)=>{let P=k.runKernel(Ii,C,N);return F([$,A,P]),d&&(P=R(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:w}})(f,m):un((A,$,F,P)=>{let V=k.runKernel(Ii,C,N);return P([$,A,V,F]),d&&(V=R(V,[V.shape[1],V.shape[2],V.shape[3]])),{value:V,gradFunc:w}})(f,m,x)}var L_=T({fusedConv2d_:x6});function y6(r,t,e,n,o,s=[1,1],i){let a=r;r.rank===3&&(a=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let u=t;u.rank===3&&(u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let l={x:a,dy:u},c={strides:n,pad:o,dimRoundingMode:i,dilations:s,filterShape:e};return k.runKernel(gp,l,c)}var uy=T({depthwiseConv2dNativeBackpropFilter_:y6});function b6(r,t,e,n,o,s=[1,1],i){let a=t,u=!1;t.rank===3&&(u=!0,a=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let l={dy:a,filter:e},c={strides:n,pad:o,dimRoundingMode:i,dilations:s,inputShape:r},p=k.runKernel(xp,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var cy=T({depthwiseConv2dNativeBackpropInput_:b6});function w6({x:r,filter:t,strides:e,pad:n,dataFormat:o=\"NHWC\",dilations:s=[1,1],dimRoundingMode:i,bias:a,activation:u=\"linear\",preluActivationWeights:l,leakyreluAlpha:c}){if(sc(k.state.gradientDepth,u)===!1){let N=Fi(r,t,e,n,o,s,i);return a!=null&&(N=X(N,a)),oc(N,u,l,c)}let p=I(r,\"x\",\"depthwiseConv2d\",\"float32\"),m=I(t,\"filter\",\"depthwiseConv2d\",\"float32\"),f=p,d=!1;p.rank===3&&(d=!0,f=R(p,[1,p.shape[0],p.shape[1],p.shape[2]])),E(f.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${f.rank}.`),E(m.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${m.rank}.`),E(f.shape[3]===m.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${f.shape[3]}) must match the inChannels dimension in filter ${m.shape[2]}.`),s==null&&(s=[1,1]),E(Ar(e,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`),Ie(\"fused depthwiseConv2d\",n,i);let h=ju(f.shape,m.shape,e,s,n,i,!0),g;a!=null&&(g=I(a,\"bias\",\"fused conv2d\"),[g]=Ut(g,p),Pt(h.outShape,g.shape));let x;l!=null&&(x=I(l,\"prelu weights\",\"fused depthwiseConv2d\"));let b=(N,_)=>{E(to(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[A,$,F,P]=_,V=rc(N,F,u),G=cy($.shape,V,A,e,n,s,i),W=uy($,V,A.shape,e,n,s,i);if(P!=null){let q=nc(g,V);return[G,W,q]}return[G,W]},w={x:f,filter:m,bias:g,preluActivationWeights:x},C={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};return a==null?un((_,A,$)=>{let F=k.runKernel(Si,w,C);return $([A,_,F]),d&&(F=R(F,[F.shape[1],F.shape[2],F.shape[3]])),{value:F,gradFunc:b}})(f,m):un((_,A,$,F)=>{let P=k.runKernel(Si,w,C);return F([A,_,P,$]),d&&(P=R(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:b}})(f,m,g)}var M_=T({fusedDepthwiseConv2d_:w6});function C6({a:r,b:t,transposeA:e=!1,transposeB:n=!1,bias:o,activation:s=\"linear\",preluActivationWeights:i,leakyreluAlpha:a=.2}){if(sc(k.state.gradientDepth,s)===!1){let V=Lt(r,t,e,n);return o!=null&&(V=X(V,o)),oc(V,s,i,a)}let u=I(r,\"a\",\"fused matMul\"),l=I(t,\"b\",\"fused matMul\");[u,l]=Ut(u,l);let c=e?u.shape[u.rank-2]:u.shape[u.rank-1],p=n?l.shape[l.rank-1]:l.shape[l.rank-2],m=e?u.shape[u.rank-1]:u.shape[u.rank-2],f=n?l.shape[l.rank-2]:l.shape[l.rank-1],d=u.shape.slice(0,-2),h=l.shape.slice(0,-2),g=Jt(d),x=Jt(h);E(c===p,()=>`Error in fused matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${u.shape} and ${l.shape} and transposeA=${e} and transposeB=${n} must match.`);let w=Pt(u.shape.slice(0,-2),l.shape.slice(0,-2)).concat([m,f]),C=e?R(u,[g,c,m]):R(u,[g,m,c]),N=n?R(l,[x,f,p]):R(l,[x,p,f]),_;o!=null&&(_=I(o,\"bias\",\"fused matMul\"),[_]=Ut(_,u),Pt(w,_.shape));let A;i!=null&&(A=I(i,\"prelu weights\",\"fused matMul\"));let $=(V,G)=>{let[W,q,H,j]=G,Y=rc(R(V,H.shape),H,s),Z,et;if(!e&&!n?(Z=Lt(Y,q,!1,!0),et=Lt(W,Y,!0,!1)):!e&&n?(Z=Lt(Y,q,!1,!1),et=Lt(Y,W,!0,!1)):e&&!n?(Z=Lt(q,Y,!1,!0),et=Lt(W,Y,!1,!1)):(Z=Lt(q,Y,!0,!0),et=Lt(Y,W,!0,!0)),o!=null){let rt=nc(j,Y);return[Z,et,rt]}else return[Z,et]},F={a:C,b:N,bias:_,preluActivationWeights:A},P={transposeA:e,transposeB:n,activation:s,leakyreluAlpha:a};return o==null?un((G,W,q)=>{let H=k.runKernel(Ci,F,P);return q([G,W,H]),{value:R(H,w),gradFunc:$}})(C,N):un((G,W,q,H)=>{let j=k.runKernel(Ci,F,P);return H([G,W,j,q]),{value:R(j,w),gradFunc:$}})(C,N,_)}var z_=T({fusedMatMul_:C6});function I6(r){return hh(r,.54,.46)}var B_=T({hammingWindow_:I6});function S6(r){return hh(r,.5,.5)}var py=T({hannWindow_:S6});function v6(r,t,e,n=!1,o=0){let s=0,i=[];for(;s+t<=r.size;)i.push(Rt(r,s,t)),s+=e;if(n)for(;s<r.size;){let a=s+t-r.size,u=ne([Rt(r,s,t-a),xo([a],o)]);i.push(u),s+=e}return i.length===0?Vs([],[0,t]):R(ne(i),[i.length,t])}var my=T({frame_:v6});function N6(r,t,e,n,o=py){n==null&&(n=uv(t));let s=my(r,t,e),i=D(s,o(t));return lu(i,n)}var V_=T({stft_:N6});function T6(r,t,e,n,o=\"bilinear\",s=0){let i=I(r,\"image\",\"cropAndResize\"),a=I(t,\"boxes\",\"cropAndResize\",\"float32\"),u=I(e,\"boxInd\",\"cropAndResize\",\"int32\"),l=a.shape[0];E(i.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),E(a.rank===2&&a.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${l},4] but had shape ${a.shape}.`),E(u.rank===1&&u.shape[0]===l,()=>`Error in cropAndResize: boxInd must be have size [${l}] but had shape ${a.shape}.`),E(n.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${n.length}.`),E(n[0]>=1&&n[1]>=1,()=>`cropSize must be atleast [1,1], but was ${n}`),E(o===\"bilinear\"||o===\"nearest\",()=>`method must be bilinear or nearest, but was ${o}`);let c={image:i,boxes:a,boxInd:u},p={method:o,extrapolationValue:s,cropSize:n};return k.runKernel(da,c,p)}var G_=T({cropAndResize_:T6});function k6(r){let t=I(r,\"image\",\"flipLeftRight\",\"float32\");E(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let e={image:t};return k.runKernel(ba,e,{})}var W_=T({flipLeftRight_:k6});function E6(r){let t=I(r,\"image\",\"grayscaleToRGB\"),e=t.rank-1,n=t.shape[e];E(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),E(n===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${n}.`);let o=new Array(t.rank);return o.fill(1,0,e),o[e]=3,Dr(t,o)}var U_=T({grayscaleToRGB_:E6});function _6(r,t,e=0,n=.5){let o=I(r,\"image\",\"rotateWithOffset\",\"float32\");E(o.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${o.rank}.`);let s={image:o},i={radians:t,fillValue:e,center:n};return k.runKernel(qa,s,i)}var H_=T({rotateWithOffset_:_6});function wo(r,t,e,n,o,s){n==null&&(n=.5),o==null&&(o=Number.NEGATIVE_INFINITY),s==null&&(s=0);let i=r.shape[0];return e=Math.min(e,i),E(0<=n&&n<=1,()=>`iouThreshold must be in [0, 1], but was '${n}'`),E(r.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${r.rank}'`),E(r.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${r.shape[1]}`),E(t.rank===1,()=>\"scores must be a 1D tensor\"),E(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),E(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:e,iouThreshold:n,scoreThreshold:o,softNmsSigma:s}}function A6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY){let s=I(r,\"boxes\",\"nonMaxSuppression\",\"float32\"),i=I(t,\"scores\",\"nonMaxSuppression\",\"float32\"),a=wo(s,i,e,n,o);e=a.maxOutputSize,n=a.iouThreshold,o=a.scoreThreshold;let u={maxOutputSize:e,iouThreshold:n,scoreThreshold:o};return k.runKernel(Ra,{boxes:s,scores:i},u)}var q_=T({nonMaxSuppression_:A6});function K_(r,t,e){let n=$6(r,t,e),o=n<0?-(n+1):n;r.splice(o,0,t)}function $6(r,t,e){return R6(r,t,e||D6)}function D6(r,t){return r>t?1:r<t?-1:0}function R6(r,t,e){let n=0,o=r.length,s=0,i=!1;for(;n<o;){s=n+(o-n>>>1);let a=e(t,r[s]);a>0?n=s+1:(o=s,i=!a)}return i?n:-n-1}function fy(r,t,e,n,o){return cv(r,t,e,n,o,0)}function dy(r,t,e,n,o,s){return cv(r,t,e,n,o,0,!1,s,!0)}function hy(r,t,e,n,o,s){return cv(r,t,e,n,o,s,!0)}function cv(r,t,e,n,o,s,i=!1,a=!1,u=!1){let l=[];for(let g=0;g<t.length;g++)t[g]>o&&l.push({score:t[g],boxIndex:g,suppressBeginIndex:0});l.sort(j_);let c=s>0?-.5/s:0,p=[],m=[];for(;p.length<e&&l.length>0;){let g=l.pop(),{score:x,boxIndex:b,suppressBeginIndex:w}=g;if(x<o)break;let C=!1;for(let N=p.length-1;N>=w;--N){let _=F6(r,b,p[N]);if(_>=n){C=!0;break}if(g.score=g.score*O6(n,c,_),g.score<=o)break}g.suppressBeginIndex=p.length,C||(g.score===x?(p.push(b),m.push(g.score)):g.score>o&&K_(l,g,j_))}let f=p.length,d=e-f;a&&d>0&&(p.push(...new Array(d).fill(0)),m.push(...new Array(d).fill(0)));let h={selectedIndices:p};return i&&(h.selectedScores=m),u&&(h.validOutputs=f),h}function F6(r,t,e){let n=r.subarray(t*4,t*4+4),o=r.subarray(e*4,e*4+4),s=Math.min(n[0],n[2]),i=Math.min(n[1],n[3]),a=Math.max(n[0],n[2]),u=Math.max(n[1],n[3]),l=Math.min(o[0],o[2]),c=Math.min(o[1],o[3]),p=Math.max(o[0],o[2]),m=Math.max(o[1],o[3]),f=(a-s)*(u-i),d=(p-l)*(m-c);if(f<=0||d<=0)return 0;let h=Math.max(s,l),g=Math.max(i,c),x=Math.min(a,p),b=Math.min(u,m),w=Math.max(x-h,0)*Math.max(b-g,0);return w/(f+d-w)}function O6(r,t,e){let n=Math.exp(t*e*e);return e<=r?n:0}function j_(r,t){return r.score-t.score||r.score===t.score&&t.boxIndex-r.boxIndex}async function P6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY){let s=I(r,\"boxes\",\"nonMaxSuppressionAsync\"),i=I(t,\"scores\",\"nonMaxSuppressionAsync\"),a=wo(s,i,e,n,o);e=a.maxOutputSize,n=a.iouThreshold,o=a.scoreThreshold;let u=await Promise.all([s.data(),i.data()]),l=u[0],c=u[1],{selectedIndices:p}=fy(l,c,e,n,o);return s!==r&&s.dispose(),i!==t&&i.dispose(),Me(p,\"int32\")}var X_=P6;function L6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=0){let i=I(r,\"boxes\",\"nonMaxSuppression\"),a=I(t,\"scores\",\"nonMaxSuppression\"),u=wo(i,a,e,n,o,s);e=u.maxOutputSize,n=u.iouThreshold,o=u.scoreThreshold,s=u.softNmsSigma;let l={boxes:i,scores:a},c={maxOutputSize:e,iouThreshold:n,scoreThreshold:o,softNmsSigma:s},p=k.runKernel(Oa,l,c);return{selectedIndices:p[0],selectedScores:p[1]}}var Y_=T({nonMaxSuppressionWithScore_:L6});async function M6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=0){let i=I(r,\"boxes\",\"nonMaxSuppressionAsync\"),a=I(t,\"scores\",\"nonMaxSuppressionAsync\"),u=wo(i,a,e,n,o,s);e=u.maxOutputSize,n=u.iouThreshold,o=u.scoreThreshold,s=u.softNmsSigma;let l=await Promise.all([i.data(),a.data()]),c=l[0],p=l[1],{selectedIndices:m,selectedScores:f}=hy(c,p,e,n,o,s);return i!==r&&i.dispose(),a!==t&&a.dispose(),{selectedIndices:Me(m,\"int32\"),selectedScores:Me(f)}}var Z_=M6;function z6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=!1){let i=I(r,\"boxes\",\"nonMaxSuppression\"),a=I(t,\"scores\",\"nonMaxSuppression\"),u=wo(i,a,e,n,o,null),l=u.maxOutputSize,c=u.iouThreshold,p=u.scoreThreshold,m={boxes:i,scores:a},f={maxOutputSize:l,iouThreshold:c,scoreThreshold:p,padToMaxOutputSize:s},d=k.runKernel(Fa,m,f);return{selectedIndices:d[0],validOutputs:d[1]}}var J_=T({nonMaxSuppressionPadded_:z6});async function B6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=!1){let i=I(r,\"boxes\",\"nonMaxSuppressionAsync\"),a=I(t,\"scores\",\"nonMaxSuppressionAsync\"),u=wo(i,a,e,n,o,null),l=u.maxOutputSize,c=u.iouThreshold,p=u.scoreThreshold,[m,f]=await Promise.all([i.data(),a.data()]),{selectedIndices:d,validOutputs:h}=dy(m,f,l,c,p,s);return i!==r&&i.dispose(),a!==t&&a.dispose(),{selectedIndices:Me(d,\"int32\"),validOutputs:mt(h,\"int32\")}}var Q_=B6;function V6(r,t,e=!1,n=!1){let o=I(r,\"images\",\"resizeBilinear\");E(o.rank===3||o.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${o.rank}.`),E(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),E(n===!1||e===!1,()=>\"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.\");let s=o,i=!1;o.rank===3&&(i=!0,s=R(o,[1,o.shape[0],o.shape[1],o.shape[2]]));let[]=t,a={images:s},u={alignCorners:e,halfPixelCenters:n,size:t},l=k.runKernel(Ss,a,u);return i?R(l,[l.shape[1],l.shape[2],l.shape[3]]):l}var gy=T({resizeBilinear_:V6});function G6(r,t,e=!1,n=!1){let o=I(r,\"images\",\"resizeNearestNeighbor\");E(o.rank===3||o.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${o.rank}.`),E(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),E(o.dtype===\"float32\"||o.dtype===\"int32\",()=>\"`images` must have `int32` or `float32` as dtype\"),E(n===!1||e===!1,()=>\"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.\");let s=o,i=!1;o.rank===3&&(i=!0,s=R(o,[1,o.shape[0],o.shape[1],o.shape[2]]));let[]=t,a={images:s},u={alignCorners:e,halfPixelCenters:n,size:t},l=k.runKernel(Is,a,u);return i?R(l,[l.shape[1],l.shape[2],l.shape[3]]):l}var xy=T({resizeNearestNeighbor_:G6});function W6(r,t=\"binary\",e=!1,n=.5){let o=I(r,\"image\",\"threshold\"),s=.2989,i=.587,a=.114,u=o.shape[0]*o.shape[1],l=D(Me([n]),255),c,p,m,f;if(E(o.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${o.rank}.`),E(o.shape[2]===3||o.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${o.shape[2]}.`),E(o.dtype===\"int32\"||o.dtype===\"float32\",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${o.dtype}.`),E(t===\"otsu\"||t===\"binary\",()=>`Method must be binary or otsu, but was ${t}`),o.shape[2]===3){[c,p,m]=mr(o,[1,1,1],-1);let g=D(c,s),x=D(p,i),b=D(m,a);f=X(X(g,x),b)}else f=r;if(t===\"otsu\"){let g=wx(J(um(f),\"int32\"),ur([]),256);l=U6(g,u)}let d=e?Ln(f,l):Re(f,l);return J(D(d,255),\"int32\")}function U6(r,t){let e=Me([-1]),n=Me([0]),o=Me([0]),s,i,a,u,l,c;for(let p=0;p<r.size-1;p++){s=Rt(r,0,p+1),i=Rt(r,p+1),l=pt(ft(s),t),c=pt(ft(i),t);let m=ft(D(s,su(0,s.size)));a=pt(m,ft(s));let f=xo(i.shape,s.size),d=X(su(0,i.size),f),h=D(i,d);u=pt(ft(h),ft(i));let g=ct(a,u),x=ct(a,u),b=D(l,c);o=D(D(b,g),x);let w=Re(o,n);n=_e(w,o,n),e=_e(w,Me([p]),e)}return e}var tA=T({threshold_:W6});function H6(r,t,e=\"nearest\",n=\"constant\",o=0,s){let i=I(r,\"image\",\"transform\",\"float32\"),a=I(t,\"transforms\",\"transform\",\"float32\");E(i.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),E(a.rank===2&&(a.shape[0]===i.shape[0]||a.shape[0]===1)&&a.shape[1]===8,()=>\"Error in transform: Input transform should be batch x 8 or 1 x 8\"),E(s==null||s.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let u={image:i,transforms:a},l={interpolation:e,fillMode:n,fillValue:o,outputShape:s};return k.runKernel(Ha,u,l)}var eA=T({transform_:H6});function q6(r,t,e){E(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),E(e%1===0,()=>`bandPart(): numUpper must be an integer, got ${e}.`);let n=I(r,\"a\",\"bandPart\");E(n.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${n.rank}.`);let o=n.shape,[s,i]=n.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(e<=i))throw new Error(`bandPart(): numUpper (${e}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),e<0&&(e=i);let a=R(su(0,s,1,\"int32\"),[-1,1]),u=su(0,i,1,\"int32\"),l=ct(a,u),c=Rr(Ln(l,mt(+t,\"int32\")),ln(l,mt(-e,\"int32\"))),p=Ne([s,i],n.dtype);return R(nr(vr(R(n,[-1,s,i])).map(m=>_e(c,m,p))),o)}var rA=T({bandPart_:q6});function K6(r){let t;if(Array.isArray(r)){t=!1,E(r!=null&&r.length>0,()=>\"Gram-Schmidt process: input must not be null, undefined, or empty\");let o=r[0].shape[0];for(let s=1;s<r.length;++s)E(r[s].shape[0]===o,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${r[s].shape[0]} vs. ${o})`)}else t=!0,r=mr(r,r.shape[0],0).map(o=>Mn(o,[0]));E(r.length<=r[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${r.length}) exceeds number of dimensions (${r[0].shape[0]}).`);let e=[],n=r;for(let o=0;o<r.length;++o)e.push(k.tidy(()=>{let s=n[o];if(o>0)for(let i=0;i<o;++i){let a=D(ft(D(e[i],s)),e[i]);s=ct(s,a)}return pt(s,Qa(s,\"euclidean\"))}));return t?nr(e,0):e}var nA=T({gramSchmidt_:K6});function j6(r,t=!1){if(E(r.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${r.rank}`),r.rank===2)return oA(r,t);{let e=r.shape.slice(0,r.shape.length-2).reduce((u,l)=>u*l),n=vr(R(r,[e,r.shape[r.shape.length-2],r.shape[r.shape.length-1]]),0),o=[],s=[];n.forEach(u=>{let[l,c]=oA(u,t);o.push(l),s.push(c)});let i=R(nr(o,0),r.shape),a=R(nr(s,0),r.shape);return[i,a]}}function oA(r,t=!1){return k.tidy(()=>{E(r.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${r.shape.length}D Tensor.`);let e=r.shape[0],n=r.shape[1],o=Yu(e),s=sn(r),i=Vs([[1]],[1,1]),a=sn(i),u=e>=n?n:e;for(let l=0;l<u;++l){let c=s,p=a,m=o;[a,s,o]=k.tidy(()=>{let f=Rt(s,[l,l],[e-l,1]),d=Qa(f),h=Rt(s,[l,l],[1,1]),g=_e(Re(h,0),Vs([[-1]]),Vs([[1]])),x=ct(h,D(g,d)),b=pt(f,x);b.shape[0]===1?a=sn(i):a=ne([i,Rt(b,[1,0],[b.shape[0]-1,b.shape[1]])],0);let w=Ht(pt(Lt(g,x),d)),C=Rt(s,[l,0],[e-l,n]),N=D(w,a),_=Ot(a);if(l===0)s=ct(C,Lt(N,Lt(_,C)));else{let F=ct(C,Lt(N,Lt(_,C)));s=ne([Rt(s,[0,0],[l,n]),F],0)}let A=Ot(N),$=Rt(o,[0,l],[e,o.shape[1]-l]);if(l===0)o=ct($,Lt(Lt($,a),A));else{let F=ct($,Lt(Lt($,a),A));o=ne([Rt(o,[0,0],[e,l]),F],1)}return[a,s,o]}),vt([c,p,m])}return!t&&e>n&&(o=Rt(o,[0,0],[e,n]),s=Rt(s,[0,0],[n,n])),[o,s]})}var sA=T({qr_:j6});var Xe;(function(r){r[r.NONE=0]=\"NONE\",r[r.MEAN=1]=\"MEAN\",r[r.SUM=2]=\"SUM\",r[r.SUM_BY_NONZERO_WEIGHTS=3]=\"SUM_BY_NONZERO_WEIGHTS\"})(Xe||(Xe={}));function X6(r,t,e=Xe.SUM_BY_NONZERO_WEIGHTS){let n=I(r,\"losses\",\"computeWeightedLoss\"),o=null;t!=null&&(o=I(t,\"weights\",\"computeWeightedLoss\"));let s=o==null?n:D(n,o);if(e===Xe.NONE)return s;if(e===Xe.SUM)return ft(s);if(e===Xe.MEAN){if(o==null)return ve(s);{let i=n.size/o.size,a=pt(ft(s),ft(o));return i>1?pt(a,mt(i)):a}}if(e===Xe.SUM_BY_NONZERO_WEIGHTS){if(o==null)return pt(ft(s),mt(n.size));{let i=D(o,cr(n.shape)),a=J(ft(Bs(i,mt(0))),\"float32\");return pt(ft(s),a)}}throw Error(`Unknown reduction: ${e}`)}var Gr=T({computeWeightedLoss_:X6});function Y6(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,\"labels\",\"absoluteDifference\"),s=I(t,\"predictions\",\"absoluteDifference\"),i=null;e!=null&&(i=I(e,\"weights\",\"absoluteDifference\")),$e(o.shape,s.shape,\"Error in absoluteDifference: \");let a=Ee(ct(o,s));return Gr(a,i,n)}var iA=T({absoluteDifference_:Y6});function Z6(r,t,e,n,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,\"labels\",\"cosineDistance\"),i=I(t,\"predictions\",\"cosineDistance\"),a=null;n!=null&&(a=I(n,\"weights\",\"cosineDistance\")),$e(s.shape,i.shape,\"Error in cosineDistance: \");let u=mt(1),l=ct(u,ft(D(s,i),e,!0));return Gr(l,a,o)}var aA=T({cosineDistance_:Z6});function J6(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,\"labels\",\"hingeLoss\"),s=I(t,\"predictions\",\"hingeLoss\"),i=null;e!=null&&(i=I(e,\"weights\",\"hingeLoss\")),$e(o.shape,s.shape,\"Error in hingeLoss: \");let a=mt(1);o=ct(D(mt(2),o),a);let u=Fr(ct(a,D(o,s)));return Gr(u,i,n)}var lA=T({hingeLoss_:J6});function Q6(r,t,e,n=1,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,\"labels\",\"huberLoss\"),i=I(t,\"predictions\",\"huberLoss\"),a=null;e!=null&&(a=I(e,\"weights\",\"huberLoss\")),$e(s.shape,i.shape,\"Error in huberLoss: \");let u=mt(n),l=Ee(ct(i,s)),c=Mi(l,u),p=ct(l,c),m=X(D(mt(.5),Mt(c)),D(u,p));return Gr(m,a,o)}var uA=T({huberLoss_:Q6});function tX(r,t,e,n=1e-7,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,\"labels\",\"logLoss\"),i=I(t,\"predictions\",\"logLoss\"),a=null;e!=null&&(a=I(e,\"weights\",\"logLoss\")),$e(s.shape,i.shape,\"Error in logLoss: \");let u=mt(1),l=mt(n),c=Ht(D(s,Sr(X(i,l)))),p=D(ct(u,s),Sr(X(ct(u,i),l))),m=ct(c,p);return Gr(m,a,o)}var cA=T({logLoss_:tX});function eX(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,\"labels\",\"meanSquaredError\"),s=I(t,\"predictions\",\"meanSquaredError\"),i=null;e!=null&&(i=I(e,\"weights\",\"meanSquaredError\")),$e(o.shape,s.shape,\"Error in meanSquaredError: \");let a=ym(o,s);return Gr(a,i,n)}var pA=T({meanSquaredError_:eX});function rX(r,t){let e=I(r,\"labels\",\"sigmoidCrossEntropyWithLogits\"),n=I(t,\"logits\",\"sigmoidCrossEntropyWithLogits\");$e(e.shape,n.shape,\"Error in sigmoidCrossEntropyWithLogits: \");let o=Fr(n),s=D(n,e),i=tu(er(Ht(Ee(n))));return X(ct(o,s),i)}function nX(r,t,e,n=0,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,\"multiClassLabels\",\"sigmoidCrossEntropy\"),i=I(t,\"logits\",\"sigmoidCrossEntropy\"),a=null;if(e!=null&&(a=I(e,\"weights\",\"sigmoidCrossEntropy\")),$e(s.shape,i.shape,\"Error in sigmoidCrossEntropy: \"),n>0){let l=mt(n),c=mt(1),p=mt(.5);s=X(D(s,ct(c,l)),D(p,l))}let u=rX(s,i);return Gr(u,a,o)}var mA=T({sigmoidCrossEntropy_:nX});function oX(r,t,e=-1){if(e===-1&&(e=t.rank-1),e!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${e}`);return un((o,s,i)=>{let u=im(s,[e],!0),l=ct(J(s,\"float32\"),u);i([o,l]);let c=Ht(D(l,o));return{value:ft(c,[e]),gradFunc:(f,d)=>{let[h,g]=d,x=yo(f.shape,[e]);return[D(R(f,x),ct(J(h,\"float32\"),er(g))),D(R(f,x),ct(er(g),J(h,\"float32\")))]}}})(r,t)}function sX(r,t,e,n=0,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,\"onehotLabels\",\"softmaxCrossEntropy\"),i=I(t,\"logits\",\"softmaxCrossEntropy\"),a=null;if(e!=null&&(a=I(e,\"weights\",\"softmaxCrossEntropy\")),$e(s.shape,i.shape,\"Error in softmaxCrossEntropy: \"),n>0){let l=mt(n),c=mt(1),p=mt(s.shape[1]);s=X(D(s,ct(c,l)),pt(l,p))}let u=oX(s,i);return Gr(u,a,o)}var fA=T({softmaxCrossEntropy_:sX});function iX(r,t,e,n){let o=I(r,\"indices\",\"sparseFillEmptyRows\",\"int32\"),s=I(t,\"values\",\"sparseFillEmptyRows\"),i=I(e,\"denseShape\",\"sparseFillEmptyRows\",\"int32\"),a=I(n,\"defaultValue\",\"sparseFillEmptyRows\",s.dtype);if(o.rank!==2)throw new Error(`Indices should be Tensor2D but received shape\n        ${o.shape}`);if(s.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(i.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(a.rank!==0)throw new Error(`Default value should be a scalar but received shape ${a.shape}`);let u={indices:o,values:s,denseShape:i,defaultValue:a},l=k.runKernel(Pl,u);return{outputIndices:l[0],outputValues:l[1],emptyRowIndicator:l[2],reverseIndexMap:l[3]}}var dA=T({sparseFillEmptyRows_:iX});function aX(r,t,e){let n=I(r,\"inputIndices\",\"sparseReshape\",\"int32\"),o=I(t,\"inputShape\",\"sparseReshape\",\"int32\"),s=I(e,\"newShape\",\"sparseReshape\",\"int32\");if(n.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape\n        ${n.shape}`);if(o.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${o.shape}`);if(s.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:n,inputShape:o,newShape:s},a=k.runKernel(Ga,i);return{outputIndices:a[0],outputShape:a[1]}}var hA=T({sparseReshape_:aX});function lX(r,t,e){let n=I(r,\"data\",\"sparseSegmentMean\"),o=I(t,\"indices\",\"sparseSegmentMean\",\"int32\"),s=I(e,\"segmentIds\",\"sparseSegmentMean\",\"int32\");if(n.rank<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.rank!==1)throw new Error(`Indices should be Tensor1D but received shape\n          ${o.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape\n          ${s.shape}`);let i={data:n,indices:o,segmentIds:s};return k.runKernel(Ll,i)}var gA=T({sparseSegmentMean_:lX});function uX(r,t,e){let n=I(r,\"data\",\"sparseSegmentSum\"),o=I(t,\"indices\",\"sparseSegmentSum\",\"int32\"),s=I(e,\"segmentIds\",\"sparseSegmentSum\",\"int32\");if(n.rank<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.rank!==1)throw new Error(`Indices should be Tensor1D but received shape\n         ${o.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape\n         ${s.shape}`);let i={data:n,indices:o,segmentIds:s};return k.runKernel(Ml,i)}var xA=T({sparseSegmentSum_:uX});function cX(r,t,e,n,o,s,i,a){let u=I(r,\"data\",\"stringNGrams\",\"string\");if(u.dtype!==\"string\")throw new Error(\"Data must be of datatype string\");if(u.shape.length!==1)throw new Error(`Data must be a vector, saw: ${u.shape}`);let l=I(t,\"dataSplits\",\"stringNGrams\");if(l.dtype!==\"int32\")throw new Error(\"Data splits must be of datatype int32\");let c={separator:e,nGramWidths:n,leftPad:o,rightPad:s,padWidth:i,preserveShortSequences:a},p={data:u,dataSplits:l},m=k.runKernel(Bl,p,c);return{nGrams:m[0],nGramsSplits:m[1]}}var yA=T({stringNGrams_:cX});function pX(r,t,e=!0){let n=I(r,\"input\",\"stringSplit\",\"string\"),o=I(t,\"delimiter\",\"stringSplit\",\"string\");if(n.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${n.shape}`);if(o.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${o.shape}`);let s={skipEmpty:e},i={input:n,delimiter:o},a=k.runKernel(Vl,i,s);return{indices:a[0],values:a[1],shape:a[2]}}var bA=T({stringSplit_:pX});function mX(r,t){let e=I(r,\"input\",\"stringToHashBucketFast\",\"string\"),n={numBuckets:t};if(t<=0)throw new Error(\"Number of buckets must be at least 1\");let o={input:e};return k.runKernel(Gl,o,n)}var wA=T({stringToHashBucketFast_:mX});var fX={fft:au,ifft:tl,rfft:lu,irfft:xm},dX={hammingWindow:B_,hannWindow:py,frame:my,stft:V_},Gs={flipLeftRight:W_,grayscaleToRGB:U_,resizeNearestNeighbor:xy,resizeBilinear:gy,rotateWithOffset:H_,cropAndResize:G_,nonMaxSuppression:q_,nonMaxSuppressionAsync:X_,nonMaxSuppressionWithScore:Y_,nonMaxSuppressionWithScoreAsync:Z_,nonMaxSuppressionPadded:J_,nonMaxSuppressionPaddedAsync:Q_,threshold:tA,transform:eA},pv={bandPart:rA,gramSchmidt:nA,qr:sA},hX={absoluteDifference:iA,computeWeightedLoss:Gr,cosineDistance:aA,hingeLoss:lA,huberLoss:uA,logLoss:cA,meanSquaredError:pA,sigmoidCrossEntropy:mA,softmaxCrossEntropy:fA},gX={sparseFillEmptyRows:dA,sparseReshape:hA,sparseSegmentMean:gA,sparseSegmentSum:xA},xX={stringNGrams:yA,stringSplit:bA,stringToHashBucketFast:wA};var Wr=class extends uh{minimize(t,e=!1,n){let{value:o,grads:s}=this.computeGradients(t,n);if(n!=null){let i=n.map(a=>({name:a.name,tensor:s[a.name]}));this.applyGradients(i)}else this.applyGradients(s);return vt(s),e?o:(o.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(t,e){return Bx(t,e)}dispose(){this.iterations_!=null&&vt(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:\"iter\",tensor:mt(this.iterations_,\"int32\")}}async getWeights(){throw new Error(\"getWeights() is not implemented for this optimizer yet.\")}async setWeights(t){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(t){return this.iterations_=(await t[0].tensor.data())[0],t.slice(1)}};Object.defineProperty(Wr,Symbol.hasInstance,{value:r=>r.minimize!=null&&r.computeGradients!=null&&r.applyGradients!=null});var cu=class extends Wr{constructor(t,e,n=null){super(),this.learningRate=t,this.rho=e,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n],i=!1;this.accumulatedGrads[o]==null&&(this.accumulatedGrads[o]={originalName:`${n}/accum_grad`,variable:B(()=>It(s).variable(i))}),this.accumulatedUpdates[o]==null&&(this.accumulatedUpdates[o]={originalName:`${n}/accum_var`,variable:B(()=>It(s).variable(i))});let a=Array.isArray(t)?t[o].tensor:t[n];if(a==null)return;let u=this.accumulatedGrads[o].variable,l=this.accumulatedUpdates[o].variable;B(()=>{let c=X(D(u,this.rho),D(Mt(a),1-this.rho)),p=D(pt(Se(X(l,this.epsilon)),Se(X(u,this.epsilon))),a),m=X(D(l,this.rho),D(Mt(p),1-this.rho));u.assign(c),l.assign(m);let f=X(D(p,-this.learningRate),s);s.assign(f)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(vt(this.accumulatedGrads.map(t=>t.variable)),vt(this.accumulatedUpdates.map(t=>t.variable)))}async getWeights(){let t=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=t.length/2,n=!1;this.accumulatedGrads=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedUpdates=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(t,e){return new t(e.learningRate,e.rho,e.epsilon)}};cu.className=\"Adadelta\";Cn(cu);var pu=class extends Wr{constructor(t,e=.1){super(),this.learningRate=t,this.initialAccumulatorValue=e,this.accumulatedGrads=[]}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n];this.accumulatedGrads[o]==null&&(this.accumulatedGrads[o]={originalName:`${n}/accumulator`,variable:B(()=>xo(s.shape,this.initialAccumulatorValue).variable(!1))});let i=Array.isArray(t)?t[o].tensor:t[n];if(i==null)return;let a=this.accumulatedGrads[o].variable;B(()=>{let u=X(a,Mt(i));a.assign(u);let l=X(D(pt(i,Se(X(u,k.backend.epsilon()))),-this.learningRate),s);s.assign(l)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&vt(this.accumulatedGrads.map(t=>t.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=!1;this.accumulatedGrads=t.map(n=>({originalName:n.name,variable:n.tensor.variable(e)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(t,e){return new t(e.learningRate,e.initialAccumulatorValue)}};pu.className=\"Adagrad\";Cn(pu);var mu=class extends Wr{constructor(t,e,n,o=null){super(),this.learningRate=t,this.beta1=e,this.beta2=n,this.epsilon=o,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],B(()=>{this.accBeta1=mt(e).variable(),this.accBeta2=mt(n).variable()}),o==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){let e=Array.isArray(t)?t.map(n=>n.name):Object.keys(t);B(()=>{let n=ct(1,this.accBeta1),o=ct(1,this.accBeta2);e.forEach((s,i)=>{let a=k.registeredVariables[s],u=!1;this.accumulatedFirstMoment[i]==null&&(this.accumulatedFirstMoment[i]={originalName:`${s}/m`,variable:B(()=>It(a).variable(u))}),this.accumulatedSecondMoment[i]==null&&(this.accumulatedSecondMoment[i]={originalName:`${s}/v`,variable:B(()=>It(a).variable(u))});let l=Array.isArray(t)?t[i].tensor:t[s];if(l==null)return;let c=this.accumulatedFirstMoment[i].variable,p=this.accumulatedSecondMoment[i].variable,m=X(D(c,this.beta1),D(l,1-this.beta1)),f=X(D(p,this.beta2),D(Mt(l),1-this.beta2)),d=pt(m,n),h=pt(f,o);c.assign(m),p.assign(f);let g=X(D(pt(d,X(Se(h),this.epsilon)),-this.learningRate),a);a.assign(g)}),this.accBeta1.assign(D(this.accBeta1,this.beta1)),this.accBeta2.assign(D(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&vt(this.accumulatedFirstMoment.map(t=>t.variable)),this.accumulatedSecondMoment!=null&&vt(this.accumulatedSecondMoment.map(t=>t.variable))}async getWeights(){let t=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t),B(()=>{this.accBeta1.assign(an(this.beta1,this.iterations_+1)),this.accBeta2.assign(an(this.beta2,this.iterations_+1))});let e=t.length/2,n=!1;this.accumulatedFirstMoment=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedSecondMoment=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(t,e){return new t(e.learningRate,e.beta1,e.beta2,e.epsilon)}};mu.className=\"Adam\";Cn(mu);var fu=class extends Wr{constructor(t,e,n,o=null,s=0){super(),this.learningRate=t,this.beta1=e,this.beta2=n,this.epsilon=o,this.decay=s,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],B(()=>{this.iteration=mt(0).variable(),this.accBeta1=mt(e).variable()}),o==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){let e=Array.isArray(t)?t.map(n=>n.name):Object.keys(t);B(()=>{let n=ct(1,this.accBeta1),o=pt(-this.learningRate,X(D(this.iteration,this.decay),1));e.forEach((s,i)=>{let a=k.registeredVariables[s],u=!1;this.accumulatedFirstMoment[i]==null&&(this.accumulatedFirstMoment[i]={originalName:`${s}/m`,variable:It(a).variable(u)}),this.accumulatedWeightedInfNorm[i]==null&&(this.accumulatedWeightedInfNorm[i]={originalName:`${s}/v`,variable:It(a).variable(u)});let l=Array.isArray(t)?t[i].tensor:t[s];if(l==null)return;let c=this.accumulatedFirstMoment[i].variable,p=this.accumulatedWeightedInfNorm[i].variable,m=X(D(c,this.beta1),D(l,1-this.beta1)),f=D(p,this.beta2),d=Ee(l),h=Sn(f,d);c.assign(m),p.assign(h);let g=X(D(pt(o,n),pt(m,X(h,this.epsilon))),a);a.assign(g)}),this.iteration.assign(X(this.iteration,1)),this.accBeta1.assign(D(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&vt(this.accumulatedFirstMoment.map(t=>t.variable)),this.accumulatedWeightedInfNorm!=null&&vt(this.accumulatedWeightedInfNorm.map(t=>t.variable))}async getWeights(){throw new Error(\"getWeights() is not implemented for Adamax yet.\")}async setWeights(t){throw new Error(\"setWeights() is not implemented for Adamax yet.\")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(t,e){return new t(e.learningRate,e.beta1,e.beta2,e.epsilon,e.decay)}};fu.className=\"Adamax\";Cn(fu);var Bi=class extends Wr{constructor(t){super(),this.learningRate=t,this.setLearningRate(t)}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=Array.isArray(t)?t[o].tensor:t[n];if(s==null)return;let i=k.registeredVariables[n];B(()=>{let a=X(D(this.c,s),i);i.assign(a)})}),this.incrementIterations()}setLearningRate(t){this.learningRate=t,this.c!=null&&this.c.dispose(),this.c=De(mt(-t))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(t){if(t=await this.extractIterations(t),t.length!==0)throw new Error(\"SGD optimizer does not have settable weights.\")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(t,e){return new t(e.learningRate)}};Bi.className=\"SGD\";Cn(Bi);var du=class extends Bi{constructor(t,e,n=!1){super(t),this.learningRate=t,this.momentum=e,this.useNesterov=n,this.accumulations=[],this.m=mt(this.momentum)}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n];this.accumulations[o]==null&&(this.accumulations[o]={originalName:`${n}/momentum`,variable:B(()=>It(s).variable(!1))});let i=this.accumulations[o].variable,a=Array.isArray(t)?t[o].tensor:t[n];a!=null&&B(()=>{let u,l=X(D(this.m,i),a);this.useNesterov?u=X(D(this.c,X(a,D(l,this.m))),s):u=X(D(this.c,l),s),i.assign(l),s.assign(u)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&vt(this.accumulations.map(t=>t.variable))}setMomentum(t){this.momentum=t}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=!1;this.accumulations=t.map(n=>({originalName:n.name,variable:n.tensor.variable(e)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(t,e){return new t(e.learningRate,e.momentum,e.useNesterov)}};du.className=\"Momentum\";Cn(du);var hu=class extends Wr{constructor(t,e=.9,n=0,o=null,s=!1){if(super(),this.learningRate=t,this.decay=e,this.momentum=n,this.epsilon=o,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=s,o==null&&(this.epsilon=k.backend.epsilon()),t==null)throw new Error(\"learningRate for RMSPropOptimizer must be defined.\")}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n],i=!1;this.accumulatedMeanSquares[o]==null&&(this.accumulatedMeanSquares[o]={originalName:`${n}/rms`,variable:B(()=>It(s).variable(i))}),this.accumulatedMoments[o]==null&&(this.accumulatedMoments[o]={originalName:`${n}/momentum`,variable:B(()=>It(s).variable(i))}),this.accumulatedMeanGrads[o]==null&&this.centered&&(this.accumulatedMeanGrads[o]={originalName:`${n}/mg`,variable:B(()=>It(s).variable(i))});let a=Array.isArray(t)?t[o].tensor:t[n];if(a==null)return;let u=this.accumulatedMeanSquares[o].variable,l=this.accumulatedMoments[o].variable;B(()=>{let c=X(D(u,this.decay),D(Mt(a),1-this.decay));if(this.centered){let p=this.accumulatedMeanGrads[o].variable,m=X(D(p,this.decay),D(a,1-this.decay)),f=pt(D(a,this.learningRate),Se(ct(c,X(Mt(m),this.epsilon)))),d=X(D(l,this.momentum),f);u.assign(c),p.assign(m),l.assign(d);let h=ct(s,d);s.assign(h)}else{let p=X(D(u,this.decay),D(Mt(a),1-this.decay)),m=X(D(l,this.momentum),pt(D(a,this.learningRate),Se(X(p,this.epsilon))));u.assign(p),l.assign(m);let f=ct(s,m);s.assign(f)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&vt(this.accumulatedMeanSquares.map(t=>t.variable)),this.accumulatedMeanGrads!=null&&this.centered&&vt(this.accumulatedMeanGrads.map(t=>t.variable)),this.accumulatedMoments!=null&&vt(this.accumulatedMoments.map(t=>t.variable))}async getWeights(){let t=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&t.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=this.centered?t.length/3:t.length/2,n=!1;this.accumulatedMeanSquares=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedMoments=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=t.slice(e*2,e*3).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(t,e){return new t(e.learningRate,e.decay,e.momentum,e.epsilon,e.centered)}};hu.className=\"RMSProp\";Cn(hu);var Ws=class{static sgd(t){return new Bi(t)}static momentum(t,e,n=!1){return new du(t,e,n)}static rmsprop(t,e=.9,n=0,o=null,s=!1){return new hu(t,e,n,o,s)}static adam(t=.001,e=.9,n=.999,o=null){return new mu(t,e,n,o)}static adadelta(t=.001,e=.95,n=null){return new cu(t,e,n)}static adamax(t=.002,e=.9,n=.999,o=null,s=0){return new fu(t,e,n,o,s)}static adagrad(t,e=.1){return new pu(t,e)}};var ic={sgd:Ws.sgd,momentum:Ws.momentum,adadelta:Ws.adadelta,adagrad:Ws.adagrad,rmsprop:Ws.rmsprop,adamax:Ws.adamax,adam:Ws.adam};var yX=(()=>typeof requestAnimationFrame!=\"undefined\"?requestAnimationFrame:typeof setImmediate!=\"undefined\"?setImmediate:r=>r())();function gh(){return new Promise(r=>yX(()=>r()))}var v={};Wt(v,{ERF_A1:()=>RX,ERF_A2:()=>FX,ERF_A3:()=>OX,ERF_A4:()=>PX,ERF_A5:()=>LX,ERF_P:()=>DX,PARALLELIZE_THRESHOLD:()=>yy,RowPartitionType:()=>Vi,SELU_SCALE:()=>fv,SELU_SCALEALPHA:()=>mv,applyActivation:()=>oc,assertAndGetBroadcastShape:()=>Pt,assertAxesAreInnerMostDims:()=>Hq,assertParamsConsistent:()=>bX,assignToTypedArray:()=>WX,axesAreInnerMostDims:()=>J0,calculateShapes:()=>IE,checkEinsumDimSizes:()=>XX,checkPadOnDimRoundingMode:()=>Ie,combineLocations:()=>HE,combineRaggedTensorToTensorShapes:()=>CX,complexWithEvenIndex:()=>BX,complexWithOddIndex:()=>VX,computeConv2DInfo:()=>ju,computeConv3DInfo:()=>ME,computeDefaultPad:()=>Z0,computeDilation2DInfo:()=>KH,computeOptimalWindowSize:()=>NX,computeOutAndReduceShapes:()=>Q0,computeOutShape:()=>wX,computePool2DInfo:()=>Y0,computePool3DInfo:()=>jH,convertConv2DDataFormat:()=>zE,decodeEinsumEquation:()=>KX,eitherStridesOrDilationsAreOne:()=>Ar,expandShapeToKeepDim:()=>yo,exponent:()=>HX,exponents:()=>UX,fromStringArrayToUint8:()=>g5,fromUint8ToStringArray:()=>h5,getAxesPermutation:()=>tv,getBroadcastDims:()=>bE,getComplexWithIndex:()=>GX,getEinsumComputePath:()=>YX,getEinsumPermutation:()=>jX,getFusedBiasGradient:()=>nc,getFusedDyActivation:()=>rc,getImageCenter:()=>TX,getInnerMostAxes:()=>qq,getPermuted:()=>EX,getRaggedRank:()=>SX,getReductionAxes:()=>ge,getReshaped:()=>kX,getReshapedPermuted:()=>_X,getRowPartitionTypesHelper:()=>IX,getSliceBeginCoords:()=>AX,getSliceSize:()=>$X,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>t5,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>e5,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>r5,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>s5,getSparseReshapeInputOutputMismatchErrorMessage:()=>a5,getSparseReshapeInputOutputMultipleErrorMessage:()=>i5,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>n5,getSparseReshapeNegativeOutputDimErrorMessage:()=>o5,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>p5,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>l5,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>u5,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>c5,getUndoAxesPermutation:()=>ph,isIdentityPermutation:()=>ZX,log:()=>o4,mergeRealAndImagArrays:()=>MX,prepareAndValidate:()=>CE,prepareSplitSize:()=>QX,segment_util:()=>hv,shouldFuse:()=>sc,slice_util:()=>Le,splitRealAndImagArrays:()=>zX,tupleValuesAreOne:()=>to,upcastType:()=>sr,validateDefaultValueShape:()=>vX,validateInput:()=>sx,validateUpdateShape:()=>U0,warn:()=>vi});function bX(r,t){let e=r[0].length;r.forEach((o,s)=>{E(o.length===e,()=>`Error in concat${e}D: rank of tensors[${s}] must be the same as the rank of the rest (${e})`)}),E(t>=0&&t<e,()=>`Error in concat${e}D: axis must be between 0 and ${e-1}.`);let n=r[0];r.forEach((o,s)=>{for(let i=0;i<e;i++)E(i===t||o[i]===n[i],()=>`Error in concat${e}D: Shape of tensors[${s}] (${o}) does not match the shape of the rest (${n}) along the non-concatenated axis ${s}.`)})}function wX(r,t){let e=r[0].slice();for(let n=1;n<r.length;n++)e[t]+=r[n][t];return e}var Vi;(function(r){r[r.FIRST_DIM_SIZE=0]=\"FIRST_DIM_SIZE\",r[r.VALUE_ROWIDS=1]=\"VALUE_ROWIDS\",r[r.ROW_LENGTHS=2]=\"ROW_LENGTHS\",r[r.ROW_SPLITS=3]=\"ROW_SPLITS\",r[r.ROW_LIMITS=4]=\"ROW_LIMITS\",r[r.ROW_STARTS=5]=\"ROW_STARTS\"})(Vi||(Vi={}));function CX(r,t,e){let n=new Array;if(e==null&&t==null)return n;if(t==null)for(;n.length<r+e.length;)n.push(-1);else n=t.slice();if(e==null)return n;if(r+e.length!==n.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${r+e.length}, but shape.rank = ${n.length}`);for(let o=1;o<e.length;++o){let s=e[o],i=n[n.length-e.length+o],a=n[i];if(s>=0)if(a>=0){if(a!==s)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${o+r}] = ${s} but shape[${o+r}] = ${a}`)}else n[i]=s}return n}function IX(r){let t={FIRST_DIM_SIZE:Vi.FIRST_DIM_SIZE,VALUE_ROWIDS:Vi.VALUE_ROWIDS,ROW_LENGTHS:Vi.ROW_LENGTHS,ROW_SPLITS:Vi.ROW_SPLITS,ROW_LIMITS:Vi.ROW_LIMITS,ROW_STARTS:Vi.ROW_STARTS},e=[];for(let n of r)if(n in t)e.push(t[n]);else break;return e}function SX(r){return r.length===0?0:r[0]===Vi.FIRST_DIM_SIZE?r.length-1:r.length}function vX(r,t){if(r==null||t==null)return;let e=r.length,n=t.length;if(e>=n)throw new Error(`defaultValue.shape=${r} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${e} must be less than ragged tensor input flatValues.rank = ${n})`);for(let o=0;o<Math.min(e,n-1);++o){let s=r[o],i=t[o+1];if(s>=0&&i>=0&&s!==1&&s!==i)throw new Error(`defaultValue.shape=${r}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${o-r.length}] = ${s} but ragged tensor input.flatValues.shape[${o-r.length}] = ${i}`)}}var yy=30;function NX(r){return r<=yy?r:sp(r,Math.floor(Math.sqrt(r)))}function TX(r,t,e){let n=e*(typeof r==\"number\"?r:r[0]),o=t*(typeof r==\"number\"?r:r[1]);return[n,o]}function kX(r,t,e,n=!0){let o=[];if(n)o=o.concat(t.slice(0)),o.push(r[0]/e),o=o.concat(r.slice(1));else{o=o.concat(r[0]);let s=t.length;for(let i=0;i<s;++i)o=o.concat([r[i+1]/t[i],t[i]]);o=o.concat(r.slice(s+1))}return o}function EX(r,t,e=!0){let n=[];if(e){n.push(t);for(let o=t+1;o<r;++o)o<=2*t?(n.push(o),n.push(o-(t+1))):n.push(o)}else{let o=[],s=[];for(let i=1;i<r;++i)i>=t*2+1||i%2===1?s.push(i):o.push(i);n.push(...o),n.push(0),n.push(...s)}return n}function _X(r,t,e,n=!0){let o=[];n?o.push(r[0]/e):o.push(r[0]*e);for(let s=1;s<r.length;++s)s<=t.length?n?o.push(t[s-1]*r[s]):o.push(r[s]/t[s-1]):o.push(r[s]);return o}function AX(r,t){let e=[0];for(let n=0;n<t;++n)e.push(r[n][0]);return e}function $X(r,t,e){let n=r.slice(0,1);for(let o=0;o<e;++o)n.push(r[o+1]-t[o][0]-t[o][1]);return n}var mv=1.7580993408473768,fv=1.0507009873554805;var DX=.3275911,RX=.254829592,FX=-.284496736,OX=1.421413741,PX=-1.453152027,LX=1.061405429;function MX(r,t){if(r.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${r.length}, imag: ${t.length}.`);let e=new Float32Array(r.length*2);for(let n=0;n<e.length;n+=2)e[n]=r[n/2],e[n+1]=t[n/2];return e}function zX(r){let t=new Float32Array(r.length/2),e=new Float32Array(r.length/2);for(let n=0;n<r.length;n+=2)t[n/2]=r[n],e[n/2]=r[n+1];return{real:t,imag:e}}function BX(r){let t=Math.ceil(r.length/4),e=new Float32Array(t),n=new Float32Array(t);for(let o=0;o<r.length;o+=4)e[Math.floor(o/4)]=r[o],n[Math.floor(o/4)]=r[o+1];return{real:e,imag:n}}function VX(r){let t=Math.floor(r.length/4),e=new Float32Array(t),n=new Float32Array(t);for(let o=2;o<r.length;o+=4)e[Math.floor(o/4)]=r[o],n[Math.floor(o/4)]=r[o+1];return{real:e,imag:n}}function GX(r,t){let e=r[t*2],n=r[t*2+1];return{real:e,imag:n}}function WX(r,t,e,n){r[n*2]=t,r[n*2+1]=e}function UX(r,t){let e=new Float32Array(r/2),n=new Float32Array(r/2);for(let o=0;o<Math.ceil(r/2);o++){let s=(t?2:-2)*Math.PI*(o/r);e[o]=Math.cos(s),n[o]=Math.sin(s)}return{real:e,imag:n}}function HX(r,t,e){let n=(e?2:-2)*Math.PI*(r/t),o=Math.cos(n),s=Math.sin(n);return{real:o,imag:s}}var dv=\"->\",qX=/->/g,CA=\",\",IA=\"...\";function KX(r,t){r=r.replace(/\\s/g,\"\");let e=(r.length-r.replace(qX,\"\").length)/dv.length;if(e<1)throw new Error(\"Equations without an arrow are not supported.\");if(e>1)throw new Error(`Equation must contain exactly one arrow (\"${dv}\").`);let[n,o]=r.split(dv);E(n.indexOf(IA)===-1,()=>`The ellipsis notation (\"${IA}\") is not supported yet.`);let s=n.split(CA),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error(\"Support for more than 2 input tensors is not implemented yet.\");let a=[];for(let m=0;m<o.length;++m){let f=o[m];if(!s.some(d=>d.indexOf(f)!==-1))throw new Error(`Output subscripts contain the label ${f} not present in the input subscripts.`);a.indexOf(f)===-1&&a.push(f)}for(let m=0;m<n.length;++m){let f=n[m];a.indexOf(f)===-1&&f!==CA&&a.push(f)}let u=new Array(s.length);for(let m=0;m<i;++m){if(new Set(s[m].split(\"\")).size!==s[m].length)throw new Error(`Found duplicate axes in input component ${s[m]}. Support for duplicate axes in input is not implemented yet.`);u[m]=[];for(let f=0;f<s[m].length;++f)u[m].push(a.indexOf(s[m][f]))}let l=a.length,c=o.length,p=[];for(let m=c;m<l;++m)p.push(m);return{allDims:a,summedDims:p,idDims:u}}function jX(r,t){let e=new Array(r);e.fill(-1);for(let o=0;o<t.length;++o)e[t[o]]=o;let n=[];for(let o=0;o<r;++o)e[o]===-1&&n.push(o);return e=e.filter(o=>o!==-1),{permutationIndices:e,expandDims:n}}function XX(r,t,e){let n=new Array(r);for(let o=0;o<e.length;++o){let s=e[o].shape;for(let i=0;i<t[o].length;++i)n[t[o][i]]===void 0?n[t[o][i]]=s[i]:E(n[t[o][i]]===s[i],()=>`Expected dimension ${n[t[o][i]]} at axis ${i} of input shaped ${JSON.stringify(s)}, but got dimension ${s[i]}`)}}function YX(r,t){let e=r,n=[],o=0;r.length===0&&e.push(-1),o=r.length+1;for(let i=0;i<o;++i)n.push([]);let s=[];for(let i=0;i<e.length;++i){let a=e[i],u=JX(t,a);for(let l of u)s.indexOf(l)===-1&&(n[i].push(l),s.push(l))}return{path:e,steps:n}}function ZX(r){return r.every((t,e)=>t===e)}function JX(r,t){let e=[];for(let n=0;n<r.length;++n)(r[n].length===0||r[n].indexOf(t)!==-1||t===-1)&&e.push(n);return e}function QX(r,t,e=0){let n=[];if(typeof t==\"number\")E(r.shape[e]%t===0,()=>\"Number of splits must evenly divide the axis.\"),n=new Array(t).fill(r.shape[e]/t);else{let o=t.reduce((i,a)=>(a===-1&&(i+=1),i),0);E(o<=1,()=>\"There should be only one negative value in split array.\");let s=t.indexOf(-1);if(s!==-1){let i=t.reduce((a,u)=>u>0?a+u:a);t[s]=r.shape[e]-i}E(r.shape[e]===t.reduce((i,a)=>i+a),()=>\"The sum of sizes must match the size of the axis dimension.\"),n=t}return n}function t5(r){return`Received SparseTensor with denseShape[0] = 0 but\n  indices.shape[0] = ${r}`}function e5(r,t){return`indices(${r}, 0) is invalid: ${t} < 0`}function r5(r,t,e){return`indices(${r}, 0) is invalid: ${t} >= ${e}`}function n5(r,t){return`only one output dimension may be -1, not both ${r} and ${t}`}function o5(r,t){return`size ${r} must be non-negative, not ${t}`}function s5(){return\"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero\"}function i5(r,t){let e=Jt(r),n=Jt(t);return`Input to reshape is a SparseTensor with ${e}\n  dense values, but the requested shape requires a multiple of ${n}. inputShape=${r} outputShape= ${t}`}function a5(r,t){let e=Jt(r),n=Jt(t);return`Input to reshape is a tensor with ${e} dense values, but the requested shape has ${n}. inputShape=${r} outputShape=${t}`}function l5(){return\"segment ids must be >= 0\"}function u5(){return\"segment ids are not increasing\"}function c5(r,t){return`Segment id ${r} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function p5(r,t,e){return`Bad: indices[${r}] == ${t} out of range [0, ${e})`}var hv={};Wt(hv,{collectGatherOpShapeInfo:()=>d5,computeOutShape:()=>f5,segOpComputeOptimalWindowSize:()=>m5});function m5(r,t){let e=!1,n;for(r<=yy?(n=r,e=!0):n=sp(r,Math.floor(Math.sqrt(r)));!e;)n>t||n===r?e=!0:n=sp(r,n+1);return n}function f5(r,t,e){let n=[],o=r.length;for(let s=0;s<o;s++)s!==t?n.push(r[s]):n.push(e);return n}function d5(r,t,e,n){let o=t.shape.length,s=r.shape.length;if(n!==0&&(n<-o||n>o))throw new Error(`Expect batchDims in the range of [-${o}, ${o}], but got ${n}`);if(n<0&&(n+=o),n>s)throw new Error(`batchDims (${n}) must be less than rank(x) (\n    ${s}).`);if(e<n)throw new Error(`batchDims (${n}) must be less than or equal to axis (${e}).`);for(let p=0;p<n;++p)if(r.shape[p]!==t.shape[p])throw new Error(`x.shape[${p}]: ${r.shape[p]} should be equal to indices.shape[${p}]: ${t.shape[p]}.`);let i=r.shape[e],a=[],u=1,l=1,c=1;for(let p=0;p<n;++p)a.push(r.shape[p]),u*=r.shape[p];for(let p=n;p<e;p++)a.push(r.shape[p]),l*=r.shape[p];for(let p=n;p<o;p++)a.push(t.shape[p]);for(let p=e+1;p<s;p++)a.push(r.shape[p]),c*=r.shape[p];return{batchSize:u,sliceSize:c,outerSize:l,dimSize:i,outputShape:a}}function h5(r){try{return r.map(t=>Wp(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function g5(r){return r.map(t=>Hl(t))}var Ur={};Wt(Ur,{nonMaxSuppressionV3Impl:()=>fy,nonMaxSuppressionV4Impl:()=>dy,nonMaxSuppressionV5Impl:()=>hy,whereImpl:()=>ay});var by={kernelName:ii,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,bo(J(e,\"float32\"),-1))}}};var SA={kernelName:oa,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Mt(J(e,\"float32\")),o=Se(ct(mt(1),n));return Ht(pt(r,o))}}}};var vA={kernelName:sa,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Se(ct(Mt(J(e,\"float32\")),1));return pt(r,n)}}}};var NA={kernelName:Zn,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=r,u=ge(e.shape,o);return u.length>0&&(a=ft(a,u)),R(a,e.shape)},b:()=>{let a=r,u=ge(n.shape,o);return u.length>0&&(a=ft(a,u)),R(a,n.shape)}}}};var TA={kernelName:Go,saveAllInputs:!0,gradFunc:(r,t)=>{let e={};return t.forEach((n,o)=>{e[o]=()=>r.clone()}),e}};var kA={kernelName:Wo,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>It(e)}}};var EA={kernelName:kl,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>It(e)}}};var _A={kernelName:la,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Se(ct(mt(1),Mt(J(e,\"float32\")))))}}};var AA={kernelName:ua,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Se(X(mt(1),Mt(J(e,\"float32\"))));return pt(r,n)}}}};var $A={kernelName:ma,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=X(Mt(e),Mt(n)),u=D(r,pt(n,a)),l=ge(e.shape,o);return l.length>0&&(u=ft(u,l)),R(u,e.shape)},b:()=>{let a=X(Mt(e),Mt(n)),u=Ht(D(r,pt(e,a))),l=ge(n.shape,o);return l.length>0&&(u=ft(u,l)),R(u,n.shape)}}}};var DA={kernelName:ca,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,X(Mt(J(e,\"float32\")),1))}}};var RA={kernelName:pa,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,ct(mt(1),Mt(J(e,\"float32\"))))}}};function x5(r,t,e,n,o,s){let i=I(r,\"dy\",\"avgPool3dGrad\"),a=I(t,\"input\",\"avgPool3dGrad\"),u=i,l=a,c=!1;a.rank===4&&(c=!0,u=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),l=R(a,[1,a.shape[0],a.shape[1],a.shape[2],a.shape[3]])),E(u.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${u.rank}.`),E(l.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${l.rank}.`),Ie(\"avgPool3dGrad\",o,s);let p={dy:u,input:l},m={filterSize:e,strides:n,pad:o,dimRoundingMode:s},f=k.runKernel(lp,p,m);return c?R(f,[f.shape[1],f.shape[2],f.shape[3],f.shape[4]]):f}var FA=T({avgPool3dGrad_:x5});var OA={kernelName:El,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{filterSize:o,strides:s,pad:i,dimRoundingMode:a}=e;return{x:()=>FA(r,n,o,s,i,a)}}};function y5(r,t,e,n,o){let s=I(r,\"dy\",\"avgPoolGrad\"),i=I(t,\"input\",\"avgPoolGrad\");E(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let a=i,u=s,l=!1;i.rank===3&&(l=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(u.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${u.rank}.`),E(a.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${a.rank}.`);let c={dy:u,input:a},p={filterSize:e,strides:n,pad:o},m=k.runKernel(ap,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var PA=T({avgPoolGrad_:y5});var LA={kernelName:Uo,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{filterSize:o,strides:s,pad:i}=e;return{x:()=>PA(r,n,o,s,i)}}};var MA={kernelName:Ho,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t,e)=>{let[n,o]=t,{transposeA:s,transposeB:i}=e;return!s&&!i?{a:()=>Lt(r,o,!1,!0),b:()=>Lt(n,r,!0,!1)}:!s&&i?{a:()=>Lt(r,o,!1,!1),b:()=>Lt(r,n,!0,!1)}:s&&!i?{a:()=>Lt(o,r,!1,!0),b:()=>Lt(n,r,!1,!1)}:{a:()=>Lt(o,r,!0,!0),b:()=>Lt(r,n,!0,!0)}}};var zA={kernelName:ai,gradFunc:(r,t,e)=>{let{blockShape:n,crops:o}=e;return{x:()=>nu(r,n,o)}}};var BA={kernelName:p1,gradFunc:(r,t,e)=>{let n=e,o=n.inputShape,s=n.shape,i=Array.from(s);for(let u=o.length-1;u>=0;u--)if(o[u]===s[u])i[u]=1;else if(o[u]!==1)throw new Error(`broadcastTo(): [${o}] cannot be broadcast to [${s}].`);let a=[];for(let u=0;u<i.length;u++)i[u]>1&&a.push(u);return{x:()=>ft(r,a,!0)}}};var VA={kernelName:lo,gradFunc:r=>({x:()=>r.clone()})};var GA={kernelName:qo,gradFunc:r=>({x:()=>It(r)})};var WA={kernelName:uo,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{clipValueMin:o,clipValueMax:s}=e;return{x:()=>_e(Rr(ln(n,o),Ln(n,s)),r,It(r))}}};var UA={kernelName:_l,inputsToSave:[\"x\"],gradFunc:by.gradFunc};var HA={kernelName:li,saveAllInputs:!0,gradFunc:(r,t,e)=>{let n=t.map(u=>u.shape),{axis:o}=e,s=lr(o,t[0].shape)[0],i=n.map(u=>u[s]);return mr(r,i,s).map(u=>()=>u)}};var qA={kernelName:Ko,inputsToSave:[\"x\",\"filter\"],gradFunc:(r,t,e)=>{let[n,o]=t,{dilations:s,strides:i,pad:a,dataFormat:u}=e;return E(to(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>tm(n.shape,r,o,i,a,u),filter:()=>Cm(n,r,o.shape,i,a,u)}}};var KA={kernelName:jo,inputsToSave:[\"dy\",\"filter\"],gradFunc:(r,t,e)=>{let[n,o]=t,{strides:s,pad:i,dataFormat:a,dimRoundingMode:u}=e;return{dy:()=>In(r,o,s,i,a,1,u),filter:()=>Cm(r,n,o.shape,s,i,a,u)}}};function b5(r,t,e,n,o){let s=r;r.rank===4&&(s=R(r,[1,r.shape[0],r.shape[1],r.shape[2],r.shape[3]]));let i=t;i.rank===4&&(i=R(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),E(s.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),E(i.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),E(e.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${e}.`),E(s.shape[4]===e[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${e[3]}.`),E(i.shape[4]===e[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${e[4]}).`);let a={x:s,dy:i},u={strides:n,pad:o,filterShape:e};return k.runKernel(fp,a,u)}var jA=T({conv3DBackpropFilter_:b5});var XA={kernelName:Al,inputsToSave:[\"x\",\"filter\"],gradFunc:(r,t,e)=>{let{dilations:n,strides:o,pad:s}=e;E(to(n),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${n}'`);let[i,a]=t;return{x:()=>kx(i.shape,r,a,o,s),filter:()=>jA(i,r,a.shape,o,s)}}};var YA={kernelName:Xo,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(Ht(fm(J(e,\"float32\"))),r)}}};var ZA={kernelName:Yo,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(dm(J(e,\"float32\")),r)}}};var JA={kernelName:Zo,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o,exclusive:s,reverse:i}=e;return{x:()=>{let a=tv([o],n.rank),u=nm(r,o,s,!i);return a!=null&&(u=Ot(u,a)),u}}}};var QA={kernelName:Jo,inputsToSave:[\"x\",\"filter\"],gradFunc:(r,t,e)=>{let{dilations:n,strides:o,pad:s,dimRoundingMode:i}=e,a=n==null?[1,1]:n;E(to(a),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${a}'`);let[u,l]=t;return E(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${u.rank}.`),E(l.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${l.rank}.`),E(u.shape[3]===l.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${u.shape[3]}) must match the inChannels dimension in filter ${l.shape[2]}.`),E(Ar(o,a),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${o} and dilations '${a}'.`),Ie(\"depthwiseConv2d\",s,i),{x:()=>cy(u.shape,r,l,o,s,a,i),filter:()=>uy(u,r,l.shape,o,s,a,i)}}};var t2={kernelName:$l,inputsToSave:[\"x\",\"filter\"],gradFunc:(r,t,e)=>{let[n,o]=t,s={x:n,filter:o,dy:r},i={x:n,filter:o,dy:r};return{x:()=>k.runKernel(jd,s,e),filter:()=>k.runKernel(Xd,i,e)}}};var e2={kernelName:ts,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t,n={dy:r,y:e};return{x:()=>k.runKernel(wp,n)}}};var r2={kernelName:ga,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t,n=D(er(Ht(Mt(e))),2/Math.sqrt(Math.PI));return{x:()=>D(r,n)}}};var n2={kernelName:es,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,e)}}};var o2={kernelName:ui,inputsToSave:[\"input\"],gradFunc:(r,t)=>{let[e]=t;return{input:()=>R(r,e.shape)}}};var s2={kernelName:ya,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,er(e))}}};var i2={kernelName:rs,gradFunc:r=>({x:()=>It(r)})};var a2={kernelName:ns,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=pt(r,J(n,\"float32\")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,\"float32\")),u=ge(n.shape,o);u.length>0&&(a=R(ft(a,u),n.shape));let l=Mt(n);return Ht(pt(a,J(l,\"float32\")))}}}};var l2={kernelName:os,inputsToSave:[\"x\",\"mean\",\"variance\",\"scale\"],gradFunc:(r,t,e)=>{let{varianceEpsilon:n}=e,[o,s,i,a]=t,u=a==null?mt(1):a,l=ge(s.shape,o.shape),c=[];if(s.rank===1){for(let C=0;C<o.shape.length-1;++C)c.push(o.shape[C]);c.push(1)}let p=ct(o,s),m=D(r,u),f=cm(X(i,mt(n))),d=D(D(D(f,f),f),mt(-.5));return{x:()=>s.rank===1?R(D(D(r,Dr(R(f,[1,1,1,s.shape[0]]),c)),u),o.shape):R(D(D(r,f),u),o.shape),mean:()=>{let C=D(D(f,mt(-1)),m);return s.rank===1&&(C=ft(C,l)),R(C,s.shape)},variance:()=>{let C=D(D(d,p),m);return s.rank===1&&(C=ft(C,l)),R(C,s.shape)},scale:()=>{let C=D(p,f),N=D(r,C);return s.rank===1&&(N=ft(N,l)),R(N,s.shape)},offset:()=>{let C=r;return s.rank===1&&(C=ft(C,l)),R(C,s.shape)}}}};var p2={kernelName:ci,inputsToSave:[\"x\",\"indices\"],gradFunc:(r,t,e)=>{let[n,o]=t,{axis:s}=e,i=lr(s,n.shape)[0];return{x:()=>{let u=n.shape,l=o.size,c=u.slice(0,i),p=c.length,m=u.slice(s,u.length).slice(1),f=m.length,d=u2(0,p),h=u2(p+1,p+1+f),g=c2([c,[l],m]),x=R(r,g),b=R(o,[l]),w=c2([[p],d,h]),C=Ot(x,w),N=wm(C,b,n.shape[i]),_=ph(w);return N=Ot(N,_),N},indices:()=>o}}};function u2(r,t){let e=[];for(let n=r;n<t;++n)e.push(n);return e}function c2(r){let t=[];for(let e=0;e<r.length;++e)for(let n=0;n<r[e].length;++n)t.push(r[e][n]);return t}var m2={kernelName:ss,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>It(e),b:()=>It(n)}}};var f2={kernelName:co,gradFunc:r=>({x:()=>J(r,\"float32\")})};var d2={kernelName:Ia,gradFunc:r=>({x:()=>It(r)})};var h2={kernelName:Sa,gradFunc:r=>({x:()=>It(r)})};var g2={kernelName:va,gradFunc:r=>({x:()=>It(r)})};var x2={kernelName:is,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{alpha:o}=e,s=Re(n,0);return{x:()=>_e(s,r,D(r,o))}}};var y2={kernelName:ka,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,X(e,1))}}};var b2={kernelName:as,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,J(e,\"float32\"))}}};var w2={kernelName:f1,inputsToSave:[],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e;return{logits:()=>{let i=er(n);return ct(r,D(ft(r,o,!0),i))}}}};function w5(r,t,e,n=5,o=1,s=1,i=.5){let a={x:r,y:t,dy:e},u={depthRadius:n,bias:o,alpha:s,beta:i};return k.runKernel(Np,a,u)}var C2=T({localResponseNormalizationBackprop_:w5});var I2={kernelName:Rl,inputsToSave:[\"x\"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{depthRadius:s,bias:i,alpha:a,beta:u}=e;return{x:()=>C2(n,o,r,s,i,a,u)}}};function wy(r,t,e,n){return t.rank<e.rank&&(t=R(t,yo(t.shape,n))),r.rank<e.rank&&(r=R(r,yo(r.shape,n))),{x:()=>D(r,J($r(e,t),r.dtype))}}var gv={kernelName:ls,inputsToSave:[\"x\"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let n=e,{reductionIndices:o}=n,s=t[0],i=t[1],a=lr(o,s.shape),u=wy(r,i,s,a);return{x:()=>u.x()}}};var S2={kernelName:us,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>D(r,J(ln(e,n),\"float32\")),b:()=>D(r,J(om(e,n),\"float32\"))}}};function C5(r,t,e,n,o,s,i){let a=I(r,\"dy\",\"maxPool3dGrad\"),u=I(t,\"input\",\"maxPool3dGrad\"),l=I(e,\"output\",\"maxPool3dGrad\"),c=a,p=u,m=l,f=!1;u.rank===4&&(f=!0,c=R(a,[1,a.shape[0],a.shape[1],a.shape[2],a.shape[3]]),p=R(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]]),m=R(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]])),E(c.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${c.rank}.`),E(p.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${p.rank}.`),E(m.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${m.rank}.`),Ie(\"maxPool3dGrad\",s,i);let d={dy:c,input:p,output:m},h={filterSize:n,strides:o,pad:s,dimRoundingMode:i},g=k.runKernel(kp,d,h);return f?R(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var v2=T({maxPool3dGrad_:C5});var N2={kernelName:Fl,inputsToSave:[\"x\"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=e;return{x:()=>v2(r,n,o,s,i,a,u)}}};function I5(r,t,e,n,o,s,i){let a=I(r,\"dy\",\"maxPoolGrad\"),u=I(t,\"input\",\"maxPoolGrad\"),l=I(e,\"output\",\"maxPoolGrad\");E(u.rank===a.rank,()=>`Rank of input (${u.rank}) does not match rank of dy (${a.rank})`),E(a.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${a.rank}.`),E(u.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${u.rank}.`),Ie(\"maxPoolGrad\",s,i);let c={dy:a,input:u,output:l},p={filterSize:n,strides:o,pad:s,dimRoundingMode:i};return k.runKernel(Tp,c,p)}var T2=T({maxPoolGrad_:I5});var k2={kernelName:cs,inputsToSave:[\"x\"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{filterSize:s,strides:i,pad:a}=e;return{x:()=>T2(r,n,o,s,i,a)}}};var E2={kernelName:ps,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e,s=lr(o,n.shape),a=Q0(n.shape,s)[1],u=Jt(a);return{x:()=>{let c=n.shape.slice();s.forEach(f=>{c[f]=1});let p=R(r,c);return pt(D(p,cr(n.shape,\"float32\")),u)}}}};var _2={kernelName:ms,inputsToSave:[\"x\"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let n=e,{axis:o}=n,[s,i]=t,a=lr(o,s.shape),u=wy(r,i,s,a);return{x:()=>u.x()}}};var A2={kernelName:fs,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>D(r,J(Ln(e,n),\"float32\")),b:()=>D(r,J(Re(e,n),\"float32\"))}}};var $2={kernelName:ds,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let n=t[0],{paddings:o}=e,s=o.map(i=>i[0]);return{x:()=>Rt(r,s,n.shape)}}};var D2={kernelName:$a,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=ge(e.shape,o);return a.length>0?R(ft(r,a),e.shape):r},b:()=>{let a=D(r,Ht(Pi(pt(e,n)))),u=ge(n.shape,o);return u.length>0?R(ft(a,u),n.shape):a}}}};var R2={kernelName:hs,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=D(r,J(n,\"float32\")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,\"float32\")),u=ge(n.shape,o);return u.length>0?R(ft(a,u),n.shape):a}}}};var F2={kernelName:pi,gradFunc:r=>({x:()=>Ht(r)})};var O2={kernelName:gs,inputsToSave:[\"indices\"],gradFunc:(r,t)=>{let e=t[0];return{indices:()=>Ne(e.shape,\"float32\")}}};var P2={kernelName:mi,gradFunc:r=>({x:()=>It(r)})};var L2={kernelName:fi,saveAllInputs:!0,gradFunc:(r,t,e)=>{let{axis:n}=e;return vr(r,n).map(s=>()=>s)}};var xv={kernelName:xs,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let n=t[0],{paddings:o}=e,s=o.map(i=>i[0]);return{x:()=>Rt(r,s,n.shape)}}};var M2={kernelName:ys,inputsToSave:[\"a\",\"b\"],outputsToSave:[!0],gradFunc:(r,t)=>{let[e,n,o]=t,s=e,i=n,a=Pt(s.shape,i.shape);return{a:()=>{let c=J(i,\"float32\"),p=D(r,D(c,an(s,ct(c,mt(1))))),m=ge(s.shape,a);return m.length>0&&(p=ft(p,m)),R(p,s.shape)},b:()=>{let c=Re(s,0),p=_e(c,Sr(s),It(s)),m=D(r,D(o,p)),f=ge(i.shape,a);return f.length>0&&(m=ft(m,f)),R(m,i.shape)}}}};var z2={kernelName:bs,inputsToSave:[\"x\",\"alpha\"],gradFunc:(r,t)=>{let[e,n]=t,o=Re(e,0);return{x:()=>_e(o,r,D(r,n)),alpha:()=>{let s=_e(o,It(r),D(r,e)),i=ge(n.shape,r.shape);return i.length>0&&(s=ft(s,i)),R(s,n.shape)}}}};function S5(r,t,e){let n=r.shape.slice();n[e]=1;let o=R(t,n),s=Xu(r,e,!0,!1),i=Xu(r,e,!0,!0),a=D(s,i);return D(o,a)}function v5(r,t,e){let n=r.shape.length,o=n-e.length,s=v.getAxesPermutation(e,n),i=r;s!=null&&(i=Ot(r,s));let a=i.shape.slice(),l=a.splice(n-e.length,e.length).reduce((m,f)=>m*f,1);a.push(l);let c=i.reshape(a),p=S5(c,t,o);if(p=p.reshape(i.shape),s!=null){let m=v.getUndoAxesPermutation(s);p=Ot(p,m)}return p}var B2={kernelName:ws,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e,s=[];return o==null?s=n.shape.map((i,a)=>a):typeof o==\"number\"?s=[o]:s=o,{x:()=>v5(n,r,s)}}};var V2={kernelName:Qo,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=pt(r,J(n,\"float32\")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,\"float32\")),u=ge(n.shape,o);u.length>0&&(a=R(ft(a,u),n.shape));let l=Mt(n);return Ht(pt(a,J(l,\"float32\")))}}}};var G2={kernelName:Pa,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Ht(Mt(e)))}}};var W2={kernelName:vs,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t,n=D(Ln(e,6),bo(e));return{x:()=>D(r,J(n,\"float32\"))}}};var U2={kernelName:Cs,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,J(bo(e),\"float32\"))}}};var H2={kernelName:di,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>R(r,e.shape)}}};var q2={kernelName:Ss,inputsToSave:[\"images\"],gradFunc:(r,t,e)=>{let[n]=t,o={dy:r,images:n};return{images:()=>k.runKernel(Op,o,e)}}};var K2={kernelName:Is,inputsToSave:[\"images\"],gradFunc:(r,t,e)=>{let[n]=t,o={dy:r,images:n};return{images:()=>k.runKernel(Fp,o,e)}}};var j2={kernelName:Ns,gradFunc:(r,t,e)=>{let{dims:n}=e,o=lr(n,r.shape);return{x:()=>pr(r,o)}}};var X2={kernelName:Ts,gradFunc:r=>({x:()=>It(r)})};var Y2={kernelName:ks,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>Ht(pt(r,D(an(e,1.5),2)))}}};var Z2={kernelName:hi,inputsToSave:[\"condition\"],gradFunc:(r,t)=>{let[e]=t;return{condition:()=>J(It(e),\"float32\"),t:()=>D(r,J(e,r.dtype)),e:()=>D(r,J(eu(e),r.dtype))}}};var J2={kernelName:Ma,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Re(e,mt(0)),o=mt(mv),s=mt(fv),i=D(r,s),a=D(D(r,o),er(J(e,\"float32\")));return _e(n,i,a)}}}};var Q2={kernelName:_s,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,D(e,ct(mt(1),e)))}}};var t$={kernelName:Ba,gradFunc:r=>({x:()=>It(r)})};var e$={kernelName:Es,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(Jl(J(e,\"float32\")),r)}}};var r$={kernelName:za,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(rm(J(e,\"float32\")),r)}}};var n$={kernelName:gi,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{begin:o,size:s}=e,i=n.shape,[a,u]=q0(n,o,s),l=[];for(let c=0;c<r.rank;c++)l.push([a[c],i[c]-a[c]-u[c]]);return{x:()=>cn(r,l)}}};var o$={kernelName:Ds,outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n]=t,{dim:o}=e,s=!0,i=D(r,n);return{logits:()=>ct(i,D(ft(i,[o],s),n))}}};var s$={kernelName:Va,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,Yr(e))}}};var yv={kernelName:xi,gradFunc:(r,t,e)=>{let{blockShape:n,paddings:o}=e;return{x:()=>Zl(r,n,o)}}};var bv={kernelName:yi,gradFunc:(r,t,e)=>{let{axis:n}=e;return{x:()=>ne(r,n)}}};var i$={kernelName:As,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,D(Se(J(e,\"float32\")),2))}}};var a$={kernelName:zl,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,D(J(e,\"float32\"),2))}}};var l$={kernelName:Rs,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=mt(2);return{a:()=>D(r,D(o,ct(e,n))),b:()=>D(r,D(o,ct(n,e)))}}};var u$={kernelName:po,gradFunc:r=>({x:()=>It(r)})};var c$={kernelName:Fs,inputsToSave:[\"a\",\"b\"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=r,u=ge(e.shape,o);return u.length>0&&(a=ft(a,u)),R(a,e.shape)},b:()=>{let a=r,u=ge(n.shape,o);return u.length>0&&(a=ft(a,u)),R(Ht(a),n.shape)}}}};var p$={kernelName:$s,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,o=n.shape.slice(),{axis:s}=e;lr(s,n.shape).forEach(l=>{o[l]=1});let a=R(r,o),u=D(a,cr(n.shape,\"float32\"));return{x:()=>u}}};var m$={kernelName:Os,inputsToSave:[\"x\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Mt(Jl(e)))}}};var f$={kernelName:Ps,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(ct(mt(1),Mt(e)),r)}}};var d$={kernelName:Jn,inputsToSave:[\"x\"],gradFunc:(r,t,e)=>{let[n]=t,{reps:o}=e;return{x:()=>{let i=It(n);if(n.rank===1)for(let a=0;a<o[0];++a)i=X(i,Rt(r,[a*n.shape[0]],[n.shape[0]]));else if(n.rank===2)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1]],[n.shape[0],n.shape[1]]));else if(n.rank===3)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)for(let l=0;l<o[2];++l)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1],l*n.shape[2]],[n.shape[0],n.shape[1],n.shape[2]]));else if(n.rank===4)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)for(let l=0;l<o[2];++l)for(let c=0;c<o[3];++c)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1],l*n.shape[2],c*n.shape[3]],[n.shape[0],n.shape[1],n.shape[2],n.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${n.rank} tensors yet.`);return i}}}};var h$={kernelName:Qn,gradFunc:(r,t,e)=>{let n=e,{perm:o}=n,s=ph(o);return{x:()=>Ot(r,s)}}};var g$={kernelName:bi,gradFunc:(r,t,e)=>{let n=e,{axis:o}=n;return{value:()=>nr(r,o)}}};var x$={kernelName:Wl,inputsToSave:[\"segmentIds\"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>N5(r,e)}}};function N5(r,t){let e=Sn(t,It(t)),n=Li(r,e),o=ln(t,mt(0,\"int32\")),s=n.rank-o.rank;for(let a=0;a<s;++a)o=rr(o,a+1);o=Rr(o,cr(n.shape,\"bool\"));let i=It(n);return _e(o,n,i)}var y$={kernelName:wi,gradFunc:r=>({x:()=>It(r)})};var T5=[by,SA,vA,NA,TA,kA,EA,_A,AA,$A,DA,RA,OA,LA,MA,zA,BA,VA,GA,WA,UA,HA,KA,qA,XA,YA,ZA,JA,QA,t2,V2,e2,r2,n2,o2,s2,a2,i2,l2,p2,m2,f2,d2,h2,g2,x2,y2,b2,w2,I2,gv,gv,S2,N2,k2,E2,_2,A2,$2,D2,R2,F2,O2,P2,L2,xv,xv,M2,z2,B2,G2,W2,U2,H2,q2,K2,j2,X2,Y2,Z2,J2,Q2,t$,e$,r$,n$,o$,s$,yv,yv,bv,bv,i$,l$,a$,u$,c$,p$,m$,f$,d$,h$,g$,x$,y$];for(let r of T5)h1(r);O().prototype.abs=function(){return this.throwIfDisposed(),Ee(this)};O().prototype.acos=function(){return this.throwIfDisposed(),ax(this)};O().prototype.acosh=function(){return this.throwIfDisposed(),lx(this)};O().prototype.add=function(r){return this.throwIfDisposed(),X(this,r)};O().prototype.all=function(r,t){return this.throwIfDisposed(),Zp(this,r,t)};O().prototype.any=function(r,t){return this.throwIfDisposed(),qu(this,r,t)};O().prototype.argMax=function(r){return this.throwIfDisposed(),Ai(this,r)};O().prototype.argMin=function(r){return this.throwIfDisposed(),ux(this,r)};O().prototype.asScalar=function(){return this.throwIfDisposed(),E(this.size===1,()=>\"The array must have only 1 element.\"),R(this,[])};O().prototype.asType=function(r){return this.throwIfDisposed(),J(this,r)};O().prototype.as1D=function(){return this.throwIfDisposed(),R(this,[this.size])};O().prototype.as2D=function(r,t){return this.throwIfDisposed(),R(this,[r,t])};O().prototype.as3D=function(r,t,e){return this.throwIfDisposed(),R(this,[r,t,e])};O().prototype.as4D=function(r,t,e,n){return this.throwIfDisposed(),R(this,[r,t,e,n])};O().prototype.as5D=function(r,t,e,n,o){return this.throwIfDisposed(),R(this,[r,t,e,n,o])};O().prototype.asin=function(){return this.throwIfDisposed(),cx(this)};O().prototype.asinh=function(){return this.throwIfDisposed(),px(this)};O().prototype.atan=function(){return this.throwIfDisposed(),mx(this)};O().prototype.atan2=function(r){return this.throwIfDisposed(),fx(this,r)};O().prototype.atanh=function(){return this.throwIfDisposed(),dx(this)};O().prototype.avgPool=function(r,t,e,n){return this.throwIfDisposed(),Yl(this,r,t,e,n)};O().prototype.batchToSpaceND=function(r,t){return this.throwIfDisposed(),Zl(this,r,t)};O().prototype.batchNorm=function(r,t,e,n,o){return this.throwIfDisposed(),Di(this,r,t,e,n,o)};O().prototype.broadcastTo=function(r){return this.throwIfDisposed(),Ri(this,r)};O().prototype.cast=function(r){return this.throwIfDisposed(),J(this,r)};O().prototype.ceil=function(){return this.throwIfDisposed(),Cx(this)};O().prototype.clipByValue=function(r,t){return this.throwIfDisposed(),Cr(this,r,t)};O().prototype.concat=function(r,t){return this.throwIfDisposed(),r instanceof Ft&&(r=[r]),ne([this,...r],t)};O().prototype.conv1d=function(r,t,e,n,o,s){return this.throwIfDisposed(),Qp(this,r,t,e,n,o,s)};O().prototype.conv2dTranspose=function(r,t,e,n,o){return this.throwIfDisposed(),em(this,r,t,e,n,o)};O().prototype.conv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),In(this,r,t,e,n,o,s)};O().prototype.cos=function(){return this.throwIfDisposed(),Jl(this)};O().prototype.cosh=function(){return this.throwIfDisposed(),rm(this)};O().prototype.cumprod=function(r,t,e){return this.throwIfDisposed(),Xu(this,r,t,e)};O().prototype.cumsum=function(r,t,e){return this.throwIfDisposed(),nm(this,r,t,e)};O().prototype.depthToSpace=function(r,t){return this.throwIfDisposed(),_x(this,r,t)};O().prototype.depthwiseConv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),Fi(this,r,t,e,n,o,s)};O().prototype.dilation2d=function(r,t,e,n,o){return this.throwIfDisposed(),Ax(this,r,t,e,n,o)};O().prototype.divNoNan=function(r){return this.throwIfDisposed(),$x(this,r)};O().prototype.div=function(r){return this.throwIfDisposed(),pt(this,r)};O().prototype.dot=function(r){return this.throwIfDisposed(),Dx(this,r)};O().prototype.elu=function(){return this.throwIfDisposed(),Oi(this)};O().prototype.equal=function(r){return this.throwIfDisposed(),$r(this,r)};O().prototype.erf=function(){return this.throwIfDisposed(),Rx(this)};O().prototype.euclideanNorm=function(r,t){return this.throwIfDisposed(),Fx(this,r,t)};O().prototype.exp=function(){return this.throwIfDisposed(),er(this)};O().prototype.expandDims=function(r){return this.throwIfDisposed(),rr(this,r)};O().prototype.expm1=function(){return this.throwIfDisposed(),Ox(this)};O().prototype.fft=function(){return this.throwIfDisposed(),au(this)};O().prototype.flatten=function(){return this.throwIfDisposed(),R(this,[this.size])};O().prototype.floor=function(){return this.throwIfDisposed(),Pi(this)};O().prototype.floorDiv=function(r){return this.throwIfDisposed(),Yp(this,r)};O().prototype.gather=function(r,t){return this.throwIfDisposed(),Li(this,r,t)};O().prototype.greaterEqual=function(r){return this.throwIfDisposed(),ln(this,r)};O().prototype.greater=function(r){return this.throwIfDisposed(),Re(this,r)};O().prototype.ifft=function(){return this.throwIfDisposed(),tl(this)};O().prototype.irfft=function(){return this.throwIfDisposed(),xm(this)};O().prototype.isFinite=function(){return this.throwIfDisposed(),Px(this)};O().prototype.isInf=function(){return this.throwIfDisposed(),Lx(this)};O().prototype.isNaN=function(){return this.throwIfDisposed(),Mx(this)};O().prototype.leakyRelu=function(r){return this.throwIfDisposed(),Ql(this,r)};O().prototype.lessEqual=function(r){return this.throwIfDisposed(),Ln(this,r)};O().prototype.less=function(r){return this.throwIfDisposed(),om(this,r)};O().prototype.localResponseNormalization=function(r,t,e,n){return this.throwIfDisposed(),zx(this,r,t,e,n)};O().prototype.logSigmoid=function(){return this.throwIfDisposed(),Gx(this)};O().prototype.logSoftmax=function(r){return this.throwIfDisposed(),sm(this,r)};O().prototype.logSumExp=function(r,t){return this.throwIfDisposed(),im(this,r,t)};O().prototype.log=function(){return this.throwIfDisposed(),Sr(this)};O().prototype.log1p=function(){return this.throwIfDisposed(),tu(this)};O().prototype.logicalAnd=function(r){return this.throwIfDisposed(),Rr(this,r)};O().prototype.logicalNot=function(){return this.throwIfDisposed(),eu(this)};O().prototype.logicalOr=function(r){return this.throwIfDisposed(),am(this,r)};O().prototype.logicalXor=function(r){return this.throwIfDisposed(),Wx(this,r)};O().prototype.matMul=function(r,t,e){return this.throwIfDisposed(),Lt(this,r,t,e)};O().prototype.maxPool=function(r,t,e,n){return this.throwIfDisposed(),ru(this,r,t,e,n)};O().prototype.max=function(r,t){return this.throwIfDisposed(),Ir(this,r,t)};O().prototype.maximum=function(r){return this.throwIfDisposed(),Sn(this,r)};O().prototype.mean=function(r,t){return this.throwIfDisposed(),ve(this,r,t)};O().prototype.min=function(r,t){return this.throwIfDisposed(),Ja(this,r,t)};O().prototype.minimum=function(r){return this.throwIfDisposed(),Mi(this,r)};O().prototype.mirrorPad=function(r,t){return this.throwIfDisposed(),qx(this,r,t)};O().prototype.mod=function(r){return this.throwIfDisposed(),Kx(this,r)};O().prototype.mul=function(r){return this.throwIfDisposed(),D(this,r)};O().prototype.neg=function(){return this.throwIfDisposed(),Ht(this)};O().prototype.norm=function(r,t,e){return this.throwIfDisposed(),Qa(this,r,t,e)};O().prototype.notEqual=function(r){return this.throwIfDisposed(),Bs(this,r)};O().prototype.oneHot=function(r,t=1,e=0){return this.throwIfDisposed(),Ei(this,r,t,e)};O().prototype.onesLike=function(){return this.throwIfDisposed(),yr(this)};O().prototype.pad=function(r,t){return this.throwIfDisposed(),cn(this,r,t)};O().prototype.pool=function(r,t,e,n,o,s){return this.throwIfDisposed(),jx(this,r,t,e,n,o,s)};O().prototype.pow=function(r){return this.throwIfDisposed(),an(this,r)};O().prototype.prelu=function(r){return this.throwIfDisposed(),ou(this,r)};O().prototype.prod=function(r,t){return this.throwIfDisposed(),Xx(this,r,t)};O().prototype.reciprocal=function(){return this.throwIfDisposed(),ty(this)};O().prototype.relu=function(){return this.throwIfDisposed(),Fr(this)};O().prototype.relu6=function(){return this.throwIfDisposed(),lm(this)};O().prototype.reshapeAs=function(r){return this.throwIfDisposed(),R(this,r.shape)};O().prototype.reshape=function(r){return this.throwIfDisposed(),R(this,r)};O().prototype.resizeBilinear=function(r,t,e){return this.throwIfDisposed(),gy(this,r,t,e)};O().prototype.resizeNearestNeighbor=function(r,t,e){return this.throwIfDisposed(),xy(this,r,t,e)};O().prototype.reverse=function(r){return this.throwIfDisposed(),pr(this,r)};O().prototype.rfft=function(){return this.throwIfDisposed(),lu(this)};O().prototype.round=function(){return this.throwIfDisposed(),um(this)};O().prototype.rsqrt=function(){return this.throwIfDisposed(),cm(this)};O().prototype.selu=function(){return this.throwIfDisposed(),pm(this)};O().prototype.separableConv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),mm(this,r,t,e,n,o,s)};O().prototype.sigmoid=function(){return this.throwIfDisposed(),Yr(this)};O().prototype.sign=function(){return this.throwIfDisposed(),ey(this)};O().prototype.sin=function(){return this.throwIfDisposed(),fm(this)};O().prototype.sinh=function(){return this.throwIfDisposed(),dm(this)};O().prototype.slice=function(r,t){return this.throwIfDisposed(),Rt(this,r,t)};O().prototype.softmax=function(r){return this.throwIfDisposed(),iu(this,r)};O().prototype.softplus=function(){return this.throwIfDisposed(),zs(this)};O().prototype.spaceToBatchND=function(r,t){return this.throwIfDisposed(),nu(this,r,t)};O().prototype.split=function(r,t){return this.throwIfDisposed(),mr(this,r,t)};O().prototype.sqrt=function(){return this.throwIfDisposed(),Se(this)};O().prototype.square=function(){return this.throwIfDisposed(),Mt(this)};O().prototype.squaredDifference=function(r){return this.throwIfDisposed(),ym(this,r)};O().prototype.squeeze=function(r){return this.throwIfDisposed(),Mn(this,r)};O().prototype.stack=function(r,t){this.throwIfDisposed();let e=r instanceof Ft?[this,r]:[this,...r];return nr(e,t)};O().prototype.step=function(r){return this.throwIfDisposed(),bo(this,r)};O().prototype.stridedSlice=function(r,t,e,n,o,s,i,a){return this.throwIfDisposed(),ry(this,r,t,e,n,o,s,i,a)};O().prototype.sub=function(r){return this.throwIfDisposed(),ct(this,r)};O().prototype.sum=function(r,t){return this.throwIfDisposed(),ft(this,r,t)};O().prototype.tan=function(){return this.throwIfDisposed(),ny(this)};O().prototype.tanh=function(){return this.throwIfDisposed(),$i(this)};O().prototype.tile=function(r){return this.throwIfDisposed(),Dr(this,r)};O().prototype.toBool=function(){return this.throwIfDisposed(),J(this,\"bool\")};O().prototype.toFloat=function(){return this.throwIfDisposed(),J(this,\"float32\")};O().prototype.toInt=function(){return this.throwIfDisposed(),J(this,\"int32\")};O().prototype.topk=function(r,t){return this.throwIfDisposed(),oy(this,r,t)};O().prototype.transpose=function(r){return this.throwIfDisposed(),Ot(this,r)};O().prototype.unique=function(r){return this.throwIfDisposed(),sy(this,r)};O().prototype.unsortedSegmentSum=function(r,t){return this.throwIfDisposed(),wm(this,r,t)};O().prototype.unstack=function(r){return this.throwIfDisposed(),vr(this,r)};O().prototype.where=function(r,t){return this.throwIfDisposed(),_e(r,this,t)};O().prototype.zerosLike=function(){return this.throwIfDisposed(),It(this)};var vn=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,vn.prototype)}},Hr=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,Hr.prototype)}},M=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,M.prototype)}},St=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,St.prototype)}},Im=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,Im.prototype)}};var xh=class{constructor(t){this.maxEntries=t||100,this.cache=new Map}get(t){let e;return this.cache.has(t)&&(e=this.cache.get(t),this.cache.delete(t),this.cache.set(t,e)),e}put(t,e){if(this.cache.has(t))this.cache.delete(t);else if(this.cache.size>=this.maxEntries){let n=this.cache.keys().next().value;this.cache.delete(n)}this.cache.set(t,e)}getMaxEntries(){return this.maxEntries}setMaxEntries(t){if(t<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${t}.`);if(this.maxEntries>t)for(let e=0;e<this.maxEntries-t;e++){let n=this.cache.keys().next().value;this.cache.delete(n)}this.maxEntries=t}};function Io(r,t){if(Array.isArray(r)){let e=[];for(let n=0;n<t;n++)e=e.concat(r);return e}else{let e=new Array(t);return e.fill(r),e}}function ro(r,t){if(!r)throw new Im(t)}function Cv(r,t){let e=0;for(let n of r)n===t&&e++;return e}function Nr(r){return r.length===1?r[0]:r}function xe(r){return Array.isArray(r)?r:[r]}function So(r){let e=r.replace(/(.)([A-Z][a-z0-9]+)/g,\"$1_$2\").replace(/([a-z])([A-Z])/g,\"$1_$2\").toLowerCase();return e[0]!==\"_\"?e:\"private\"+e}function el(r){return r.length<=1||r.indexOf(\"_\")===-1?r:r.replace(/[_]+(\\w|$)/g,(t,e)=>e.toUpperCase())}var Co={};function Sm(r){if(r==null)return null;let t={};return t.className=r.getClassName(),t.config=r.getConfig(),t}function wv(r){if(!(r==null||typeof r!=\"object\"))if(Array.isArray(r))r.forEach(t=>wv(t));else{let t=Object.keys(r);for(let e of t){let n=r[e];n!=null&&typeof n==\"object\"&&(!Array.isArray(n)&&n.type===\"ndarray\"&&typeof n.value==\"number\"?r[e]=n.value:wv(n))}}}function Gi(r,t={},e={},n=\"object\",o=!1){if(typeof r==\"string\"){let s=r,i;if(s in e)i=e[s];else if(s in Co)i=Co[s];else if(i=t[s],i==null)throw new M(`Unknown ${n}: ${r}. This may be due to one of the following reasons:\n1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return i}else{let s=r;if(s.className==null||s.config==null)throw new M(`${n}: Improper config format: ${JSON.stringify(s)}.\n'className' and 'config' must set.`);let i=s.className,a,u;if(i in e?[a,u]=e[i]:i in Co?[a,u]=Co.className:i in t&&([a,u]=t[i]),a==null)throw new M(`Unknown ${n}: ${i}. This may be due to one of the following reasons:\n1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(u!=null){let l={};for(let f of Object.keys(Co))l[f]=Co[f];for(let f of Object.keys(e))l[f]=e[f];let c=s.config;c.customObjects=l;let p=Object.assign({},Co);for(let f of Object.keys(e))Co[f]=e[f];wv(s.config);let m=u(a,s.config,e,o);return Co=Object.assign({},p),m}else{let l=Object.assign({},Co);for(let p of Object.keys(e))Co[p]=e[p];let c=new a(s.config);return Co=Object.assign({},l),c}}}function k5(r,t){return r<t?-1:r>t?1:0}function yh(r,t){return-1*k5(r,t)}function vo(r){if(r==null)return r;let t=[];for(let e of r)t.indexOf(e)===-1&&t.push(e);return t}function b$(r){if(r==null)throw new M(`Invalid value in obj: ${JSON.stringify(r)}`);for(let t in r)if(r.hasOwnProperty(t))return!1;return!0}function Wi(r,t,e){if(e!=null&&r.indexOf(e)<0)throw new M(`${e} is not a valid ${t}.  Valid values are ${r} or null/undefined.`)}function Cy(r,t,e=0,n=1/0){return ro(e>=0),ro(n>=e),Array.isArray(r)&&r.length>=e&&r.length<=n&&r.every(o=>typeof o===t)}function Ze(r,t){Array.isArray(r)?(y.assert(r.length>0,()=>`${t} is unexpectedly an empty array.`),r.forEach((e,n)=>Ze(e,`element ${n+1} of ${t}`))):y.assert(Number.isInteger(r)&&r>0,()=>`Expected ${t} to be a positive integer, but got ${w$(r)}.`)}function w$(r){return r===null?\"null\":Array.isArray(r)?\"[\"+r.map(t=>w$(t)).join(\",\")+\"]\":typeof r==\"string\"?`\"${r}\"`:`${r}`}function C$(r,t,e){let n=e!=null?e():y.now(),o;return(...i)=>{let a=e!=null?e():y.now();return a-n<t||(n=a,o=r(...i)),o}}function Iy(r){return r===\"relu\"?\"relu\":r===\"linear\"?\"linear\":r===\"elu\"?\"elu\":null}var E5=0;function vy(){return E5++}var Sy={};function gu(r=\"\"){return r in Sy||(Sy[r]=0),Sy[r]+=1,r+Sy[r].toString()}var I$=[\"channelsFirst\",\"channelsLast\"],S$=[\"nearest\",\"bilinear\"],v$=[\"valid\",\"same\",\"causal\"],N$=[\"max\",\"avg\"],T$=[\"sum\",\"mul\",\"concat\",\"ave\"];var vm=new Map;function Fe(r){Wi(I$,\"DataFormat\",r)}function E$(r){Wi(S$,\"InterpolationFormat\",r)}function pn(r){Wi(v$,\"PaddingMode\",r)}function Iv(r){Wi(N$,\"PoolMode\",r)}var bh=[],k$=\"/\";function Hs(r,t){bh.push(r);try{let e=t();return bh.pop(),e}catch(e){throw bh.pop(),e}}function _5(){return bh.length===0?\"\":bh.join(k$)+k$}function Ny(r){if(!_$(r))throw new Error(\"Not a valid tensor name: '\"+r+\"'\");return _5()+r}function Ty(r){if(!_$(r))throw new Error(\"Not a valid tensor name: '\"+r+\"'\");vm.has(r)||vm.set(r,0);let t=vm.get(r);if(vm.set(r,vm.get(r)+1),t>0){let e=`${r}_${t}`;return vm.set(e,1),e}else return r}var A5=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\\._\\/]*$/);function _$(r){return!!r.match(A5)}function A$(r){return r===parseInt(r.toString(),10)}function No(r,t,e){t==null&&(t=0),e==null&&(e=r.length);let n=1;for(let o=t;o<e;++o)n*=r[o];return n}function ac(r){if(r.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let e=0;e<r.length;e++){let n=r[e];n<t&&(t=n)}return t}function qs(r){if(r.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let e=0;e<r.length;e++){let n=r[e];n>t&&(t=n)}return t}function Zr(r,t){if(t<r)throw new M(`end (${t}) < begin (${r}) is forbidden.`);let e=[];for(let n=r;n<t;++n)e.push(n);return e}var Sv;function ir(){return Sv==null&&(Sv=gE().epsilon()),Sv}function mn(){return\"channelsLast\"}function no(r,t){return J(r,t)}function nl(r,t=-1){let e=r.shape.slice();return t<0&&(t=e.length+t+1),e.splice(t,0,1),R(r,e)}function $$(r,t){return B(()=>{if(r.shape.length!==2)throw new M(`repeat() expects a rank-2 tensor, but received a rank-${r.shape.length} tensor.`);let e=nl(r,1);return Ey(e,[1,t,1])})}function D$(r){let t=[No(r.shape)];return R(r,t)}function R$(r){if(r.rank<=1)throw new M(`batchFlatten requires a minimum rank of 2. Got rank: ${r.rank}.`);let t=[r.shape[0],No(r.shape,1)];return R(r,t)}function rl(r,t,e){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:return dh(r,[t,0],[e,r.shape[1]]);case 3:return gm(r,[t,0,0],[e,r.shape[1],r.shape[2]]);case 4:return ec(r,[t,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3]]);case 5:return Rt(r,[t,0,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3],r.shape[4]]);case 6:return Rt(r,[t,0,0,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3],r.shape[4],r.shape[5]]);default:throw new M(`sliceAlongFirstAxis() received an unsupported tensor rank: ${r.rank}`)}})}function vv(r,t,e){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:return dh(r,[0,t],[r.shape[0],e]);case 3:return gm(r,[0,0,t],[r.shape[0],r.shape[1],e]);case 4:return ec(r,[0,0,0,t],[r.shape[0],r.shape[1],r.shape[2],e]);default:throw new M(`sliceAlongLastAxis() received an unsupported tensor rank: ${r.rank}`)}})}function wh(r,t,e,n){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:switch(n){case 1:return rl(r,t,e);case 2:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}case 3:switch(n){case 1:return rl(r,t,e);case 2:return gm(r,[0,t,0],[r.shape[0],e,r.shape[2]]);case 3:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}case 4:switch(n){case 1:return rl(r,t,e);case 2:return ec(r,[0,t,0,0],[r.shape[0],e,r.shape[2],r.shape[3]]);case 3:return ec(r,[0,0,t,0],[r.shape[0],r.shape[1],e,r.shape[3]]);case 4:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}default:throw new M(`sliceAlongLastAxis() received an unsupported tensor rank: ${r.rank}`)}})}function Nm(r,t=-1){let e;return t<0&&(e=r[0].rank,e!==0?t=e:t=0),t===r[0].rank&&(t=-1),ne(r,t)}function Tv(r,t){switch(r.rank){case 1:return Ix([r,t]);case 2:return Sx([r,t],0);case 3:return vx([r,t],0);case 4:return Nx([r,t],0);default:throw new M(`concatAlongFirstAxis() received an unsupported tensor rank: ${r.rank}`)}}function Ey(r,t){if(Array.isArray(t)||(t=[t]),r.rank!==t.length)throw new M(`The length of input n (${t.length}) does not match the number of dimensions in input x (${r.rank})`);return Dr(r,t)}function Tm(r,t=0,e=1,n,o){return tc(r,t,e,n,o)}function To(r,t,e,n){if(r.rank<2||t.rank<2)throw new St(`dot requires both inputs to be rank >= 2 but got x shape = ${r.shape} and y shape = ${t.shape}`);if(t.rank>=3){let o=r.shape.slice(-1)[0],s=t.shape.slice(-2)[0];if(o!==s)throw new St(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${r.shape} and  y shape = ${t.shape}`)}if(r.rank===2&&t.rank===2)return uu.matMul({a:r,b:t,transposeA:!1,transposeB:!1,bias:n?Nv(r.rank,n,mn()):null,activation:e});{let o=r.shape.slice(),s=o.pop();r=R(r,[-1,s]);let i=t.shape.slice(),a=i.pop(),u=i.pop(),l=[...i,a],c=Array.from({length:t.rank},(d,h)=>h===0?t.rank-2:h<=t.rank-2?h-1:h);t=R(Ot(t,c),[u,-1]);let p=[...o,...l],m=!1,f=!1;return R(uu.matMul({a:r,b:t,transposeA:m,transposeB:f,bias:n?Nv(r.rank,n,mn()):null,activation:e}),p)}}function _y(r,t,e){return B(()=>(Array.isArray(t)?t=Me(t,\"int32\"):t=J(t,\"int32\"),Li(r,t,e)))}function lc(r){return D(r,r)}function Nv(r,t,e){let n=t.shape;if(t.rank!==1&&t.rank!==r)throw new M(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${r}`);if(r===5){if(e===\"channelsFirst\")return n.length===1?R(t,[1,n[0],1,1,1]):R(t,[1,n[3],n[0],n[1],n[2]]);if(e===\"channelsLast\")return n.length===1?R(t,[1,1,1,1,n[0]]):R(t,[1].concat(n))}else if(r===4){if(e===\"channelsFirst\")return n.length===1?R(t,[1,n[0],1,1]):R(t,[1,n[2],n[0],n[1]]);if(e===\"channelsLast\")return n.length===1?R(t,[1,1,1,n[0]]):R(t,[1].concat(n))}else if(r===3){if(e===\"channelsFirst\")return n.length===1?R(t,[1,n[0],1]):R(t,[1,n[1],n[0]]);if(e===\"channelsLast\")return n.length===1?R(t,[1,1,n[0]]):R(t,[1].concat(n))}else if(r<3)return t;throw new M(`Unsupported input rank by biasAdd: ${t.rank}`)}function fn(r,t,e){return B(()=>(e==null&&(e=mn()),Fe(e),X(r,Nv(r.rank,t,e))))}function F$(r,t=1){if(t!==1)throw new St(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Oi(r)}function O$(r){return B(()=>pt(r,X(Ee(r),1)))}function Ay(r,t,e,n){return B(()=>lv(r,t,e,n))}function P$(r){return B(()=>{let t=X(.5,D(.2,r));return Cr(t,0,1)})}function xu(r,t,e=!1){return e?r():t()}var L$=[\"fanIn\",\"fanOut\",\"fanAvg\"],M$=[\"normal\",\"uniform\",\"truncatedNormal\"];function $5(r){Wi(L$,\"FanMode\",r)}function D5(r){Wi(M$,\"Distribution\",r)}var dn=class extends Q.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},km=class extends dn{apply(t,e){return Ne(t,e)}};km.className=\"Zeros\";Q.registerClass(km);var yu=class extends dn{apply(t,e){return cr(t,e)}};yu.className=\"Ones\";Q.registerClass(yu);var Em=class extends dn{constructor(t){if(super(),typeof t!=\"object\")throw new M(`Expected argument of type ConstantConfig but got ${t}`);if(t.value===void 0)throw new M(`config must have value set but got ${t}`);this.value=t.value}apply(t,e){return B(()=>D(mt(this.value),cr(t,e)))}getConfig(){return{value:this.value}}};Em.className=\"Constant\";Q.registerClass(Em);var _m=class extends dn{constructor(t){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=t.minval||this.DEFAULT_MINVAL,this.maxval=t.maxval||this.DEFAULT_MAXVAL,this.seed=t.seed}apply(t,e){return zi(t,this.minval,this.maxval,e)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};_m.className=\"RandomUniform\";Q.registerClass(_m);var Am=class extends dn{constructor(t){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=t.mean||this.DEFAULT_MEAN,this.stddev=t.stddev||this.DEFAULT_STDDEV,this.seed=t.seed}apply(t,e){if(e=e||\"float32\",e!==\"float32\"&&e!==\"int32\")throw new St(`randomNormal does not support dType ${e}.`);return Tm(t,this.mean,this.stddev,e,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};Am.className=\"RandomNormal\";Q.registerClass(Am);var $m=class extends dn{constructor(t){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=t.mean||this.DEFAULT_MEAN,this.stddev=t.stddev||this.DEFAULT_STDDEV,this.seed=t.seed}apply(t,e){if(e=e||\"float32\",e!==\"float32\"&&e!==\"int32\")throw new St(`truncatedNormal does not support dType ${e}.`);return bm(t,this.mean,this.stddev,e,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};$m.className=\"TruncatedNormal\";Q.registerClass($m);var Dm=class extends dn{constructor(t){super(),this.gain=t.gain!=null?t.gain:1}apply(t,e){return B(()=>{if(t.length!==2||t[0]!==t[1])throw new M(\"Identity matrix initializer can only be used for 2D square matrices.\");return D(this.gain,Yu(t[0]))})}getConfig(){return{gain:this.gain}}};Dm.className=\"Identity\";Q.registerClass(Dm);function R5(r,t=\"channelsLast\"){let e,n;if(Fe(t),r.length===2)e=r[0],n=r[1];else if([3,4,5].indexOf(r.length)!==-1){if(t===\"channelsFirst\"){let o=No(r,2);e=r[1]*o,n=r[0]*o}else if(t===\"channelsLast\"){let o=No(r,0,r.length-2);e=r[r.length-2]*o,n=r[r.length-1]*o}}else{let o=No(r);e=Math.sqrt(o),n=Math.sqrt(o)}return[e,n]}var qr=class extends dn{constructor(t){if(super(),t.scale<0)throw new M(`scale must be a positive float. Got: ${t.scale}`);this.scale=t.scale==null?1:t.scale,this.mode=t.mode==null?\"fanIn\":t.mode,$5(this.mode),this.distribution=t.distribution==null?\"normal\":t.distribution,D5(this.distribution),this.seed=t.seed}apply(t,e){let n=R5(t),o=n[0],s=n[1],i=this.scale;if(this.mode===\"fanIn\"?i/=Math.max(1,o):this.mode===\"fanOut\"?i/=Math.max(1,s):i/=Math.max(1,(o+s)/2),this.distribution===\"normal\"){let a=Math.sqrt(i);if(e=e||\"float32\",e!==\"float32\"&&e!==\"int32\")throw new St(`${this.getClassName()} does not support dType ${e}.`);return bm(t,0,a,e,this.seed)}else{let a=Math.sqrt(3*i);return zi(t,-a,a,e)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};qr.className=\"VarianceScaling\";Q.registerClass(qr);var uc=class extends qr{constructor(t){super({scale:1,mode:\"fanAvg\",distribution:\"uniform\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};uc.className=\"GlorotUniform\";Q.registerClass(uc);var cc=class extends qr{constructor(t){super({scale:1,mode:\"fanAvg\",distribution:\"normal\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};cc.className=\"GlorotNormal\";Q.registerClass(cc);var pc=class extends qr{constructor(t){super({scale:2,mode:\"fanIn\",distribution:\"normal\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};pc.className=\"HeNormal\";Q.registerClass(pc);var mc=class extends qr{constructor(t){super({scale:2,mode:\"fanIn\",distribution:\"uniform\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};mc.className=\"HeUniform\";Q.registerClass(mc);var fc=class extends qr{constructor(t){super({scale:1,mode:\"fanIn\",distribution:\"normal\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};fc.className=\"LeCunNormal\";Q.registerClass(fc);var dc=class extends qr{constructor(t){super({scale:1,mode:\"fanIn\",distribution:\"uniform\",seed:t==null?null:t.seed})}getClassName(){return qr.className}};dc.className=\"LeCunNormal\";Q.registerClass(dc);var Rm=class extends dn{constructor(t){if(super(),this.DEFAULT_GAIN=1,this.gain=t.gain==null?this.DEFAULT_GAIN:t.gain,this.seed=t.seed,this.seed!=null)throw new St(\"Random seed is not implemented for Orthogonal Initializer yet.\")}apply(t,e){return B(()=>{if(t.length<2)throw new St(\"Shape must be at least 2D.\");t[0]*t[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${t[0]*t[1]}) elements: Slowness may result.`);let n=t[0]>t[1]?[t[1],t[0]]:t,o=Tm(n,0,1,\"float32\"),s=pv.gramSchmidt(o);return t[0]>t[1]&&(s=Ot(s)),D(this.gain,s)})}getConfig(){return{gain:this.gain,seed:this.seed}}};Rm.className=\"Orthogonal\";Q.registerClass(Rm);var z$={constant:\"Constant\",glorotNormal:\"GlorotNormal\",glorotUniform:\"GlorotUniform\",heNormal:\"HeNormal\",heUniform:\"HeUniform\",identity:\"Identity\",leCunNormal:\"LeCunNormal\",leCunUniform:\"LeCunUniform\",ones:\"Ones\",orthogonal:\"Orthogonal\",randomNormal:\"RandomNormal\",randomUniform:\"RandomUniform\",truncatedNormal:\"TruncatedNormal\",varianceScaling:\"VarianceScaling\",zeros:\"Zeros\"};function B$(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,\"initializer\")}function Te(r){return Sm(r)}function de(r){if(typeof r==\"string\"){let t=r in z$?z$[r]:r;if(t===\"GlorotNormal\")return new cc;if(t===\"GlorotUniform\")return new uc;if(t===\"HeNormal\")return new pc;if(t===\"HeUniform\")return new mc;if(t===\"LeCunNormal\")return new fc;if(t===\"LeCunUniform\")return new dc;{let e={};return e.className=t,e.config={},B$(e)}}else return r instanceof dn?r:B$(r)}function $y(r){return Array.isArray(r)&&Array.isArray(r[0])}function Fm(r){return r.length===0?[]:Array.isArray(r[0])?r:[r]}function Nt(r){let t;if(Array.isArray(r)){if(r.length!==1)throw new M(`Expected Tensor length to be 1; got ${r.length}`);t=r[0]}else t=r;return t}function Bt(r){if(Array.isArray(r)&&Array.isArray(r[0])){if(r.length===1)return r=r,r[0];throw new M(`Expected exactly 1 Shape; got ${r.length}`)}else return r}function Om(r){let t=0;for(let e of r)e.shape.length===0?t+=1:t+=e.shape.reduce((n,o)=>n*o);return t}var G$=\"Variable\",Ch=class{constructor(t,e=\"float32\",n=G$,o=!0,s=null){this.dtype=e==null?\"float32\":e,this.shape=t.shape,this.id=vy(),n=n==null?G$:n,this.originalName=Ny(n),this.name=Ty(this.originalName),this.trainable_=o,this.constraint=s,this.val=iy(t,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(t){return this.assertNotDisposed(),O5(this.val,t),this.val.id!==t.id&&(this.val.assign(t),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(t){this.trainable_=t,this.val.trainable=t}};function O5(r,t){if(r.shape.toString()!==t.shape.toString())throw new Error(\"Shape mismatch: \"+JSON.stringify(r.shape)+\" vs. \"+JSON.stringify(t.shape))}function Ih(r){return r.map(t=>t.read())}function Pm(r){r.forEach(t=>{t[0].write(t[1])})}var ye=class{constructor(t){this.dtype=t.dtype,this.shape=t.shape,t.shape!=null?this.ndim=t.shape.length:this.ndim=t.ndim,this.maxNDim=t.maxNDim,this.minNDim=t.minNDim,this.axes=t.axes||{}}},Jr=class{constructor(t,e,n,o,s,i,a){this.dtype=t,this.shape=e,this.sourceLayer=n,this.inputs=o,this.callArgs=s,this.outputTensorIndex=a,this.id=vy(),i!=null&&(this.originalName=Ny(i),this.name=Ty(this.originalName)),this.rank=e.length}},P5=0,ol=class{constructor(t,e){this.callArgs=e,this.id=P5++,this.outboundLayer=t.outboundLayer,this.inboundLayers=t.inboundLayers,this.nodeIndices=t.nodeIndices,this.tensorIndices=t.tensorIndices,this.inputTensors=t.inputTensors,this.outputTensors=t.outputTensors,this.inputMasks=t.inputMasks,this.outputMasks=t.outputMasks,this.inputShapes=t.inputShapes,this.outputShapes=t.outputShapes;for(let n of t.inboundLayers)n!=null&&n.outboundNodes.push(this);t.outboundLayer.inboundNodes.push(this)}getConfig(){let t=[];for(let e of this.inboundLayers)e!=null?t.push(e.name):t.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:t,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},L5=0,$t=class extends Q.Serializable{constructor(t={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=L5++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let e=t.name;if(!e){let n=this.getClassName();e=So(n)+\"_\"+gu(n)}if(this.name=e,this.trainable_=t.trainable==null?!0:t.trainable,t.inputShape!=null||t.batchInputShape!=null){let n;if(t.batchInputShape!=null)n=t.batchInputShape;else if(t.inputShape!=null){let s=null;t.batchSize!=null&&(s=t.batchSize),n=[s].concat(t.inputShape)}this.batchInputShape=n;let o=t.dtype;o==null&&(o=t.inputDType),o==null&&(o=\"float32\"),this.dtype=o}t.weights!=null?this.initialWeights=t.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(t,e){return t.name+\"_ib-\"+e.toString()}getNodeAtIndex(t,e){if(this.inboundNodes.length===0)throw new Hr(`The layer has never been called and thus has no defined ${e}.`);if(this.inboundNodes.length<=t)throw new M(`Asked to get ${e} at node ${t}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[t]}getInputAt(t){return Nr(this.getNodeAtIndex(t,\"input\").inputTensors)}getOutputAt(t){return Nr(this.getNodeAtIndex(t,\"output\").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new vn(`Layer ${this.name} has multiple inbound nodes, hence the notion of \"layer input\" is ill-defined. Use \\`getInputAt(nodeIndex)\\` instead.`);if(this.inboundNodes.length===0)throw new vn(`Layer ${this.name} is not connected, no input to return.`);return Nr(this.getNodeAtIndex(0,\"input\").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new vn(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new vn(`Layer ${this.name} has multiple inbound nodes, hence the notion of \"layer output\" is ill-defined. Use \\`getOutputAt(nodeIndex)\\` instead.`);return Nr(this.getNodeAtIndex(0,\"output\").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(t=>t())}get updates(){return this._updates}get built(){return this._built}set built(t){this._built=t}get trainable(){return this.trainable_}set trainable(t){this._trainableWeights.forEach(e=>e.trainable=t),this.trainable_=t}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(t=>t.trainable):[]}set trainableWeights(t){this._trainableWeights=t}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(t=>!t.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(t){this._nonTrainableWeights=t}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error(\"Cannot call the resetStates() method of a non-stateful Layer object.\")}assertInputCompatibility(t){if(t=xe(t),this.inputSpec==null||this.inputSpec.length===0)return;let e=xe(this.inputSpec);if(t.length!==e.length)throw new M(`Layer ${this.name} expects ${e.length} inputs, but it received ${t.length} input tensors. Input received: ${t}`);for(let n=0;n<t.length;n++){let o=t[n],s=e[n];if(s==null)continue;let i=o.rank;if(s.ndim!=null&&i!==s.ndim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${s.ndim}, found ndim=${i}`);if(s.maxNDim!=null&&i>s.maxNDim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${s.maxNDim}, found ndim=${i}`);if(s.minNDim!=null&&i<s.minNDim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${s.minNDim}, found ndim=${i}.`);if(s.dtype!=null&&o.dtype!==s.dtype)throw new M(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${s.dtype}, found dtype=${o.dtype}.`);if(s.axes){let a=o.shape;for(let u in s.axes){let l=Number(u),c=s.axes[u],p=l>=0?a[l]:a[a.length+l];if(c!=null&&[c,null].indexOf(p)===-1)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected axis ${l} of input shape to have value ${c} but got shape ${a}.`)}}if(s.shape!=null)for(let a=0;a<s.shape.length;++a){let u=s.shape[a],l=o.shape[a];if(u!=null&&l!=null&&u!==l)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected shape=${s.shape}, found shape=${o.shape}.`)}}}call(t,e){return t}invokeCallHook(t,e){this._callHook!=null&&this._callHook(t,e)}setCallHook(t){this._callHook=t}clearCallHook(){this._callHook=null}apply(t,e){e=e||{},this.assertNotDisposed();let n=xe(t),o=!0;for(let i of n)if(!(i instanceof Jr)){o=!1;break}let s=!0;for(let i of n)if(i instanceof Jr){s=!1;break}if(o===s)throw new M(\"Arguments to apply() must be all SymbolicTensors or all Tensors\");return Hs(this.name,()=>{if(!this.built){this.assertInputCompatibility(t);let i=[];for(let a of xe(t))i.push(a.shape);this.build(Nr(i)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&s&&(this._refCount=1)}if(this.assertInputCompatibility(t),s){let i=this.call(t,e),a=xe(i),u=[];for(let l of a)n.indexOf(l)!==-1&&(l=l.clone()),u.push(l);if(i=Nr(u),this.activityRegularizer!=null)throw new St(\"Layer invocation in the presence of activity regularizer(s) is not supported yet.\");return i}else{let i=M5(t),a=this.computeOutputShape(i),u,l=z5(t);if(this.warnOnIncompatibleInputShape(Array.isArray(t)?i[0]:i),a!=null&&a.length>0&&Array.isArray(a[0])?u=a.map((c,p)=>new Jr(l,c,this,xe(t),e,this.name,p)):u=new Jr(l,a,this,xe(t),e,this.name),this.addInboundNode(t,u,null,null,i,a,e),this._refCount++,this.activityRegularizer!=null)throw new St(\"Layer invocation in the presence of activity regularizer(s) is not supported yet.\");return u}})}warnOnIncompatibleInputShape(t){if(this.batchInputShape!=null)if(t.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(t)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let e=!1;this.batchInputShape.forEach((n,o)=>{n!=null&&t[o]!=null&&t[o]!==n&&(e=!0)}),e&&console.warn(`The shape of the input tensor (${JSON.stringify(t)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new vn(`The layer ${this.name} has never been called and thus has no defined output shape.`);let t=[];for(let e of this.inboundNodes){let n=JSON.stringify(e.outputShapes);t.indexOf(n)===-1&&t.push(n)}if(t.length===1){let e=this.inboundNodes[0].outputShapes;return Array.isArray(e)&&Array.isArray(e[0])&&e.length===1?e[0]:e}else throw new vn(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of \"output shape\" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new Hr(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Om(this.weights)}build(t){this.built=!0}getWeights(t=!1){return Ih(t?this.trainableWeights:this.weights)}setWeights(t){B(()=>{let e=this.weights;if(e.length!==t.length)throw new M(`You called setWeights(weights) on layer \"${this.name}\" with a weight list of length ${t.length}, but the layer was expecting ${e.length} weights. Provided weights: ${t}...`);if(e.length===0)return;let n=[],o=Ih(e);for(let s=0;s<o.length;++s){let i=o[s],a=e[s],u=t[s];if(!y.arraysEqual(i.shape,u.shape))throw new M(`Layer weight shape ${i.shape} not compatible with provided weight shape ${u.shape}`);n.push([a,u])}Pm(n)})}addWeight(t,e,n,o,s,i,a,u){if(this._addedWeightNames.indexOf(t)!==-1)throw new M(`Duplicate weight name ${t} for layer ${this.name}`);this._addedWeightNames.push(t),n==null&&(n=\"float32\"),this.fastWeightInitDuringBuild&&(o=u!=null?u():de(\"zeros\"));let l=o.apply(e,n),c=new Ch(l,n,t,i,a);return l.dispose(),s!=null&&this.addLoss(()=>s.apply(c.read())),i==null&&(i=!0),i?this._trainableWeights.push(c):this._nonTrainableWeights.push(c),c}setFastWeightInitDuringBuild(t){this.fastWeightInitDuringBuild=t}addLoss(t){t==null||Array.isArray(t)&&t.length===0||(t=xe(t),this._losses!==void 0&&this._losses!==null&&this.losses.push(...t))}computeOutputShape(t){return t}computeMask(t,e){if(!this.supportsMasking){if(e!=null)if(Array.isArray(e))e.forEach(n=>{if(n!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return e}addInboundNode(t,e,n,o,s,i,a=null){let u=xe(t);e=xe(e),n=xe(n),o=xe(o),s=Fm(s),i=Fm(i);let l=[],c=[],p=[];for(let m of u)l.push(m.sourceLayer),c.push(m.nodeIndex),p.push(m.tensorIndex);new ol({outboundLayer:this,inboundLayers:l,nodeIndices:c,tensorIndices:p,inputTensors:u,outputTensors:e,inputMasks:n,outputMasks:o,inputShapes:s,outputShapes:i},a);for(let m=0;m<e.length;m++)e[m].sourceLayer=this,e[m].nodeIndex=this.inboundNodes.length-1,e[m].tensorIndex=m}getConfig(){let t={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(t.batchInputShape=this.batchInputShape),this.dtype!=null&&(t.dtype=this.dtype),t}disposeWeights(){return this.weights.forEach(t=>t.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let t=0;return--this._refCount===0&&(t=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:t}}};function M5(r){r=xe(r);let t=[];for(let e of r)t.push(e.shape);return Nr(t)}function z5(r){return\"float32\"}function kv(r,t,e){if((t==null||e!=null&&e>0)&&(t=r.sourceLayer,e=r.nodeIndex),t.inboundNodes.length===0)return[r];{let n=t.inboundNodes[e];if(n.inboundLayers.length===0)return n.inputTensors;{let o=[];for(let s=0;s<n.inboundLayers.length;s++){let i=n.inputTensors[s],a=n.inboundLayers[s],u=n.nodeIndices[s],l=kv(i,a,u);for(let c of l)o.indexOf(c)===-1&&o.push(c)}return o}}}var Ks=class extends $t{constructor(t){if(super({dtype:t.dtype,name:t.name!=null?t.name:gu(\"input\").toString()}),t.batchSize==null&&(t.batchSize=null),t.sparse==null&&(t.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=t.sparse,t.inputShape!=null&&t.batchInputShape!=null)throw new M(\"Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.\");let e=t.batchInputShape;if(e==null){if(t.inputShape==null)throw new M(\"An InputLayer should be passed either a `batchInputShape` or an `inputShape`.\");e=[t.batchSize].concat(t.inputShape)}else if(t.batchSize!=null)throw new M(\"Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.\");let n=t.dtype||\"float32\";this.batchInputShape=e,this.dtype=n,this.inputSpec=[{shape:e}];let o=new Jr(this.dtype,this.batchInputShape,this,[],{},this.name);o.nodeIndex=0,o.tensorIndex=0,new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[o],outputTensors:[o],inputMasks:[null],outputMasks:[null],inputShapes:[e],outputShapes:[e]})}apply(t,e){throw new M(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};Ks.className=\"InputLayer\";Q.registerClass(Ks);function Dy(r){if(r.batchShape==null&&r.shape==null)throw new Error(\"Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.\");if(r.batchShape!=null&&r.shape!=null)throw new M(\"Please provide either a `shape` or `batchShape` argument to Input, but not both.\");let t=r.batchShape;r.shape!=null&&t==null&&(t=[null].concat(r.shape));let e=r.dtype;return e==null&&(e=\"float32\"),new Ks({batchInputShape:t,name:r.name,dtype:e,sparse:r.sparse}).inboundNodes[0].outputTensors[0]}function B5(r,t){if(r.dtype==null||r.dtype===t.dtype)return t;try{return J(t,r.dtype)}catch(e){throw new M(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${r.name}' (${r.dtype}).`)}}var ko=class{constructor(t){if(this.id2Value={},this.id2Mask={},this.name2Id={},t instanceof ko)for(let e in t.id2Value)this.id2Value[e]=t.id2Value[e],e in t.id2Mask&&(this.id2Mask[e]=t.id2Mask[e]);else{if(t==null)return;for(let e of t)this.add(e.key,e.value)}}add(t,e,n){if(this.id2Value[t.id]==null)this.id2Value[t.id]=B5(t,e),this.name2Id[t.name]=t.id,n!=null&&(this.id2Mask[t.id]=n);else throw new M(`Duplicate key: name=${t.name}, id=${t.id}`);return this}addFeed(t){this.add(t.key,t.value)}hasKey(t){return this.id2Value[t.id]!=null}names(){return Object.keys(this.name2Id)}getValue(t){if(t instanceof Jr){if(this.id2Value[t.id]==null)throw new M(`Nonexistent key: ${t.name}`);return this.id2Value[t.id]}else{let e=this.name2Id[t];if(e==null)throw new M(`Feed dict has no SymbolicTensor name: ${t}`);return this.id2Value[e]}}getMask(t){if(t instanceof Jr){if(this.id2Value[t.id]==null)throw new M(`Nonexistent key: ${t.name}`);return this.id2Mask[t.id]}else{let e=this.name2Id[t];if(e==null)throw new M(`Feed dict has no SymbolicTensor name: ${t}`);return this.id2Mask[e]}}disposeMasks(){this.id2Mask!=null&&vt(this.id2Mask)}},Ry=new xh,Fy=new xh;function U$(r){Ry!=null&&Ry.setMaxEntries(r),Fy!=null&&Fy.setMaxEntries(r)}function hc(r,t,e,n){let o=e==null?!1:e.training,s=Array.isArray(r),i=s?r:[r],a=i.map(d=>d.name),u=[],l=t.names();for(let d of a)l.indexOf(d)!==-1?u.push(t.getValue(d)):u.push(null);n!=null&&(n.maxNumTensors=-1/0,n.minNumTensors=1/0);let c=a.join(\",\")+\"|\"+t.names().sort().join(\",\"),p=Ry.get(c),m;if(p==null){let d=V5(i,t);p=d.sorted,m=d.recipientCounts,Ry.put(c,p),Fy.put(c,m)}m={},o||Object.assign(m,Fy.get(c));let f=new ko(t);for(let d=0;d<p.length;++d){if(n!=null){let F=ah().numTensors;F>n.maxNumTensors&&(n.maxNumTensors=F),F<n.minNumTensors&&(n.minNumTensors=F)}let h=p[d],g=h.sourceLayer;if(g instanceof Ks)continue;let x=[],b=[],w=[],C=!1;for(let F of h.inputs){let P=f.getValue(F),V=f.getMask(F);x.push(P),b.push(V),V!=null&&(C=!0),o||(m[F.name]--,m[F.name]===0&&!t.hasKey(F)&&a.indexOf(F.name)===-1&&!P.isDisposed&&F.sourceLayer.stateful!==!0&&w.push(P))}C&&(e=e||{},e.mask=b[0]);let N=xe(g.apply(x,e)),_=null;g.supportsMasking&&(_=g.computeMask(x,b));let A=W5(h),$=Array.isArray(A)?A:[A];for(let F=0;F<$.length;++F){f.hasKey($[F])||f.add($[F],N[F],Array.isArray(_)?_[0]:_);let P=a.indexOf($[F].name);P!==-1&&(u[P]=N[F])}o||vt(w)}return f.disposeMasks(),s?u:u[0]}function V5(r,t){y.assert(r!=null&&r.length>0,()=>\"Expected at least one fetch, got none\");let e=[],n={};if(r.length===1){let o=W$(r[0],t);e=o.sorted,n=o.recipientMap}else{let o=new Set;for(let s of r){let{sorted:i,recipientMap:a}=W$(s,t);for(let u of i)o.has(u.name)||(e.push(u),o.add(u.name));for(let u in a)n[u]==null&&(n[u]=new Set),a[u].forEach(l=>n[u].add(l))}}return{sorted:e,recipientCounts:G5(n)}}function G5(r){let t={};for(let e in r)t[e]=r[e].size;return t}function W$(r,t){let e=new Set,n=[],o={};for(let a of t.names())e.add(a);let s=[],i=[];for(s.push(r);s.length>0;){let a=s[s.length-1];if(e.has(a.name)){s.pop();continue}let u=i[i.length-1]===s.length-1;if(a.inputs.length===0||u)s.pop(),n.push(a),e.add(a.name),u&&i.pop();else{i.push(s.length-1);for(let l of a.inputs)o[l.name]==null&&(o[l.name]=new Set),o[l.name].add(a.name),!e.has(l.name)&&s.push(l)}}return{sorted:n,recipientMap:o}}function W5(r){let t;if(r.sourceLayer.inboundNodes.length===1)t=r.sourceLayer.output;else{let e=null;for(let n=0;n<r.sourceLayer.inboundNodes.length;++n)for(let o of r.sourceLayer.inboundNodes[n].outputTensors)if(o.id===r.id){e=n;break}t=r.sourceLayer.getOutputAt(e)}return t}var U5=z();U5.registerFlag(\"TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES\",()=>100,U$);var K$={};Wt(K$,{maxNorm:()=>H5,minMaxNorm:()=>j5,nonNeg:()=>K5,unitNorm:()=>q5});function Ev(r,t){return B(()=>Se(ft(D(r,r),t,!0)))}var gc=class extends Q.Serializable{getConfig(){return{}}},Lm=class extends gc{constructor(t){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=t.maxValue!=null?t.maxValue:this.defaultMaxValue,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>{let e=Ev(t,this.axis),n=Cr(e,0,this.maxValue);return D(t,pt(n,X(ir(),e)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};Lm.className=\"MaxNorm\";Q.registerClass(Lm);var Mm=class extends gc{constructor(t){super(),this.defaultAxis=0,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>pt(t,X(ir(),Ev(t,this.axis))))}getConfig(){return{axis:this.axis}}};Mm.className=\"UnitNorm\";Q.registerClass(Mm);var zm=class extends gc{apply(t){return Fr(t)}};zm.className=\"NonNeg\";Q.registerClass(zm);var Bm=class extends gc{constructor(t){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=t.minValue!=null?t.minValue:this.defaultMinValue,this.maxValue=t.maxValue!=null?t.maxValue:this.defaultMaxValue,this.rate=t.rate!=null?t.rate:this.defaultRate,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>{let e=Ev(t,this.axis),n=X(D(this.rate,Cr(e,this.minValue,this.maxValue)),D(1-this.rate,e));return D(t,pt(n,X(ir(),e)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};Bm.className=\"MinMaxNorm\";Q.registerClass(Bm);var H$={maxNorm:\"MaxNorm\",minMaxNorm:\"MinMaxNorm\",nonNeg:\"NonNeg\",unitNorm:\"UnitNorm\"};function ze(r){return Sm(r)}function q$(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,\"constraint\")}function Be(r){if(r==null)return null;if(typeof r==\"string\"){let e={className:r in H$?H$[r]:r,config:{}};return q$(e)}else return r instanceof gc?r:q$(r)}function H5(r){return new Lm(r)}function q5(r){return new Mm(r)}function K5(){return new zm}function j5(r){return new Bm(r)}var j$={};Wt(j$,{constant:()=>Z5,glorotNormal:()=>o8,glorotUniform:()=>n8,heNormal:()=>s8,heUniform:()=>i8,identity:()=>e8,leCunNormal:()=>a8,leCunUniform:()=>l8,ones:()=>Y5,orthogonal:()=>u8,randomNormal:()=>Q5,randomUniform:()=>J5,truncatedNormal:()=>t8,varianceScaling:()=>r8,zeros:()=>X5});function X5(){return new km}function Y5(){return new yu}function Z5(r){return new Em(r)}function J5(r){return new _m(r)}function Q5(r){return new Am(r)}function t8(r){return new $m(r)}function e8(r){return new Dm(r)}function r8(r){return new qr(r)}function n8(r){return new uc(r)}function o8(r){return new cc(r)}function s8(r){return new pc(r)}function i8(r){return new mc(r)}function a8(r){return new fc(r)}function l8(r){return new dc(r)}function u8(r){return new Rm(r)}var ED={};Wt(ED,{Layer:()=>$t,RNN:()=>Tn,RNNCell:()=>ll,activation:()=>NY,add:()=>FY,alphaDropout:()=>xZ,average:()=>OY,averagePooling1d:()=>Gv,averagePooling2d:()=>Wv,averagePooling3d:()=>Uv,avgPool1d:()=>UY,avgPool2d:()=>qY,avgPool3d:()=>jY,avgPooling1d:()=>HY,avgPooling2d:()=>KY,avgPooling3d:()=>XY,batchNormalization:()=>VY,bidirectional:()=>uZ,categoryEncoding:()=>CZ,concatenate:()=>PY,conv1d:()=>gY,conv2d:()=>xY,conv2dTranspose:()=>yY,conv3d:()=>bY,conv3dTranspose:()=>wY,convLstm2d:()=>sZ,convLstm2dCell:()=>iZ,cropping2D:()=>IY,dense:()=>TY,depthwiseConv2d:()=>vY,dot:()=>BY,dropout:()=>kY,elu:()=>cY,embedding:()=>RY,flatten:()=>_Y,gaussianDropout:()=>gZ,gaussianNoise:()=>hZ,globalAveragePooling1d:()=>YY,globalAveragePooling2d:()=>ZY,globalMaxPool1d:()=>pZ,globalMaxPool2d:()=>mZ,globalMaxPooling1d:()=>vD,globalMaxPooling2d:()=>ND,gru:()=>QY,gruCell:()=>tZ,input:()=>Pv,inputLayer:()=>uY,layerNormalization:()=>GY,leakyReLU:()=>mY,lstm:()=>eZ,lstmCell:()=>rZ,masking:()=>yZ,maxPool1d:()=>fZ,maxPool2d:()=>dZ,maxPooling1d:()=>TD,maxPooling2d:()=>kD,maxPooling3d:()=>JY,maximum:()=>LY,minimum:()=>MY,multiply:()=>zY,permute:()=>DY,prelu:()=>fY,reLU:()=>pY,repeatVector:()=>AY,rescaling:()=>bZ,reshape:()=>$Y,resizing:()=>wZ,rnn:()=>aZ,separableConv2d:()=>CY,simpleRNN:()=>nZ,simpleRNNCell:()=>oZ,softmax:()=>dY,spatialDropout1d:()=>EY,stackedRNNCells:()=>lZ,thresholdedReLU:()=>hY,timeDistributed:()=>cZ,upSampling2d:()=>SY,zeroPadding2d:()=>WY});async function Ui(r){if(r==null)return;let t=[],e=[],n=[];for(let o in r){let s=r[o];if(typeof s!=\"number\"){let i=s;t.push(i.data()),e.push(o),n.push(i)}}if(t.length>0){let o=await Promise.all(t);for(let s=0;s<o.length;++s)r[e[s]]=o[s][0];vt(n)}}function Oy(r){if(r!=null)for(let t in r){let e=r[t];typeof e!=\"number\"&&e.dispose()}}var X$;(function(r){r[r.SILENT=0]=\"SILENT\",r[r.VERBOSE=1]=\"VERBOSE\"})(X$||(X$={}));var c8=125,sl=class{constructor(){this.validationData=null}setParams(t){this.params=t}async onEpochBegin(t,e){}async onEpochEnd(t,e){}async onBatchBegin(t,e){}async onBatchEnd(t,e){}async onTrainBegin(t){}async onTrainEnd(t){}setModel(t){}},Py=class{constructor(t,e=10){t==null&&(t=[]),this.callbacks=t,this.queueLength=e}append(t){this.callbacks.push(t)}setParams(t){for(let e of this.callbacks)e.setParams(t)}setModel(t){for(let e of this.callbacks)e.setModel(t)}async onEpochBegin(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onEpochBegin(t,e)}async onEpochEnd(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onEpochEnd(t,e)}async onBatchBegin(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onBatchBegin(t,e)}async onBatchEnd(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onBatchEnd(t,e)}async onTrainBegin(t){t==null&&(t={});for(let e of this.callbacks)await e.onTrainBegin(t)}async onTrainEnd(t){t==null&&(t={});for(let e of this.callbacks)await e.onTrainEnd(t)}},_v=class extends sl{constructor(){super()}async onEpochBegin(t){this.seen=0,this.totals={}}async onBatchEnd(t,e){e==null&&(e={});let n=e.size==null?0:e.size;this.seen+=n;for(let o in e){let s=e[o];if(typeof s==\"number\")this.totals.hasOwnProperty(o)||(this.totals[o]=0),this.totals[o]=this.totals[o]+s*n;else{let i;o in this.totals?i=this.totals[o]:this.totals[o]=0;let a=B(()=>X(this.totals[o],D(s,n)));this.totals[o]=a,i!=null&&i.dispose()}}}async onEpochEnd(t,e){if(e!=null)for(let n of this.params.metrics)this.totals[n]!=null&&(typeof this.totals[n]==\"number\"?e[n]=this.totals[n]/this.seen:B(()=>{let o=D(pt(1,this.seen),this.totals[n]);e[n]=o,this.totals[n].dispose(),De(e[n])}))}},Ly=class extends sl{async onTrainBegin(t){this.epoch=[],this.history={}}async onEpochEnd(t,e){e==null&&(e={}),this.epoch.push(t);for(let n in e)this.history[n]==null&&(this.history[n]=[]),this.history[n].push(e[n])}async syncData(){let t=[],e=[],n=[];for(let s in this.history){let i=this.history[s];for(let a=0;a<i.length;++a)if(typeof i[a]!=\"number\"){let u=i[a];t.push(u.data()),e.push(s),n.push(a)}}let o=await Promise.all(t);for(let s=0;s<o.length;++s)this.history[e[s]][n[s]].dispose(),this.history[e[s]][n[s]]=o[s][0]}},My=class extends sl{constructor(t,e){if(super(),this.currentEpoch=0,this.nowFunc=t.nowFunc,this.nextFrameFunc=t.nextFrameFunc||gh,this.yieldEvery=e||\"auto\",this.yieldEvery===\"auto\"&&(this.yieldEvery=c8),this.yieldEvery===\"never\"&&t.onYield!=null)throw new Error(\"yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback\");y.isNumber(this.yieldEvery)&&(this.maybeWait=C$(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=t.onTrainBegin,this.trainEnd=t.onTrainEnd,this.epochBegin=t.onEpochBegin,this.epochEnd=t.onEpochEnd,this.batchBegin=t.onBatchBegin,this.batchEnd=t.onBatchEnd,this.yield=t.onYield}async maybeWait(t,e,n){let o=[];this.yield!=null&&(await Ui(n),o.push(this.yield(t,e,n))),o.push(this.nextFrameFunc()),await Promise.all(o)}async onEpochBegin(t,e){this.currentEpoch=t,this.epochBegin!=null&&(await Ui(e),await this.epochBegin(t,e))}async onEpochEnd(t,e){let n=[];this.epochEnd!=null&&(await Ui(e),n.push(this.epochEnd(t,e))),this.yieldEvery===\"epoch\"&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(t,e){this.batchBegin!=null&&(await Ui(e),await this.batchBegin(t,e))}async onBatchEnd(t,e){let n=[];this.batchEnd!=null&&(await Ui(e),n.push(this.batchEnd(t,e))),this.yieldEvery===\"batch\"?n.push(this.nextFrameFunc()):y.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,t,e)),await Promise.all(n)}async onTrainBegin(t){this.trainBegin!=null&&(await Ui(t),await this.trainBegin(t))}async onTrainEnd(t){this.trainEnd!=null&&(await Ui(t),await this.trainEnd(t))}};function zy(r,t){return r==null&&(r={}),r instanceof sl?[r]:Array.isArray(r)&&r[0]instanceof sl?r:xe(r).map(n=>new My(n,t))}var hn=class{constructor(){}static registerCallbackConstructor(t,e){y.assert(t>=0&&Number.isInteger(t),()=>`Verbosity level is expected to be an integer >= 0, but got ${t}`),hn.checkForDuplicate(e),hn.constructors[t]==null&&(hn.constructors[t]=[]),hn.constructors[t].push(e)}static checkForDuplicate(t){for(let e in hn.constructors)hn.constructors[+e].forEach(o=>{if(o===t)throw new M(\"Duplicate callback constructor.\")})}static clear(){hn.constructors={}}static createCallbacks(t){let e=[];for(let n in hn.constructors){let o=+n;t>=o&&e.push(...hn.constructors[o])}return e.map(n=>new n)}};hn.constructors={};function By(r,t,e,n,o,s,i,a,u){let l=new Ly,c=[new _v,...hn.createCallbacks(t)];r!=null&&c.push(...r),c.push(l);let p=new Py(c);return p.setParams({epochs:e,initialEpoch:n,samples:o,steps:s,batchSize:i,verbose:t,doValidation:a,metrics:u}),{callbackList:p,history:l}}function gn(r,t={},e=!1){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,\"layer\",e)}function Sh(r,t){return B(()=>{r.dtype!==\"float32\"&&(r=J(r,\"float32\"));let e=ft(lc(r),t,!0),n=xo(e.shape,ir()),o=Se(Sn(e,n));return pt(r,o)})}function Hi(r,t){return B(()=>ve(lc(ct(t,r)),-1))}function Vm(r,t){return B(()=>ve(Ee(ct(t,r)),-1))}function bu(r,t){return B(()=>{let e=ct(r,t),n=Cr(Ee(r),ir(),Number.MAX_VALUE),o=Ee(pt(e,n));return D(100,ve(o,-1))})}function p8(r,t){return B(()=>{let e=Cr(t,ir(),Number.MAX_VALUE),n=Sr(X(1,e)),o=Cr(r,ir(),Number.MAX_VALUE),s=Sr(X(1,o));return ve(lc(ct(n,s)),-1)})}function m8(r,t){return B(()=>{let e=Sn(0,ct(1,D(r,t)));return ve(lc(e),-1)})}function f8(r,t){return B(()=>{let e=Sn(0,ct(1,D(r,t)));return ve(e,-1)})}function d8(r,t){return B(()=>{let e=ft(D(r,t),-1),n=Ir(D(ct(1,r),t),-1);return Sn(0,X(1,ct(n,e)))})}function h8(r,t){return B(()=>{let e=Math.log(2),n=ct(t,r),o=ct(X(n,zs(D(-2,n))),e);return ve(o,-1)})}function xc(r,t,e=!1){return B(()=>{if(e)t=iu(t);else{let n=ft(t,t.shape.length-1,!0);t=pt(t,n)}return t=Cr(t,ir(),1-ir()),Ht(ft(D(J(r,\"float32\"),Sr(t)),t.shape.length-1))})}function Gm(r,t,e=!1){return B(()=>{let n=J(Pi(D$(r)),\"int32\");t=Cr(t,ir(),1-ir());let o=t.shape,s=R(Ei(n,o[o.length-1]),o);return xc(s,t,e)})}function g8(r,t){if(!y.arraysEqual(r.shape,t.shape))throw new M(`logits and labels must have the same shape, but got shapes ${JSON.stringify(r.shape)} and ${JSON.stringify(t.shape)}`);return B(()=>{let e=Fr(t),n=Ht(Ee(t));return X(ct(e,D(t,r)),tu(er(n)))})}function Wm(r,t){return B(()=>{let e;return e=Cr(t,ir(),1-ir()),e=Sr(pt(e,ct(1,e))),ve(g8(r,e),-1)})}function x8(r,t){return B(()=>{let e=Cr(r,ir(),1),n=Cr(t,ir(),1);return ft(D(r,Sr(pt(e,n))),-1)})}function y8(r,t){return B(()=>{let e=Sr(X(ir(),t));return ve(ct(t,D(r,e)),-1)})}function Nh(r,t){return B(()=>{let e=Sh(r,-1),n=Sh(t,-1),o=D(e,n);return Ht(ft(o,-1))})}var vh={meanSquaredError:Hi,meanAbsoluteError:Vm,meanAbsolutePercentageError:bu,meanSquaredLogarithmicError:p8,squaredHinge:m8,hinge:f8,categoricalHinge:d8,logcosh:h8,categoricalCrossentropy:xc,sparseCategoricalCrossentropy:Gm,binaryCrossentropy:Wm,kullbackLeiblerDivergence:x8,poisson:y8,cosineProximity:Nh};function Vy(r){if(typeof r==\"string\"){if(r in vh)return vh[r];let t=`Unknown loss ${r}`;throw r.toLowerCase().includes(\"softmaxcrossentropy\")&&(t=`Unknown loss ${r}. Use \"categoricalCrossentropy\" as the string name for tf.losses.softmaxCrossEntropy`),new M(t)}else return r}function Th(r,t){return B(()=>{let e=D(.5,yr(t)),n=no(Re(t,e),r.dtype);return ve($r(r,n),-1)})}function kh(r,t){return B(()=>no($r(Ai(r,-1),Ai(t,-1)),\"float32\"))}function Z$(r,t){return B(()=>J(ft(Rr($r(r,1),$r(t,1))),\"float32\"))}function b8(r,t){return B(()=>J(ft(Rr($r(r,1),$r(t,0))),\"float32\"))}function w8(r,t){return B(()=>J(ft(Rr($r(r,0),$r(t,1))),\"float32\"))}function Av(r,t){return B(()=>{let e=Z$(r,t),n=w8(r,t),o=X(e,n);return J(_e(Re(o,0),pt(e,o),0),\"float32\")})}function J$(r,t){return B(()=>{let e=Z$(r,t),n=b8(r,t),o=X(e,n);return J(_e(Re(o,0),pt(e,o),0),\"float32\")})}function Wy(r,t){return Wm(r,t)}function Uy(r,t){return r.rank===t.rank&&(r=Mn(r,[r.rank-1])),t=Ai(t,-1),t.dtype!==r.dtype&&(t=J(t,r.dtype)),J($r(r,t),\"float32\")}var C8=Hi,I8=Hi,S8=Vm,v8=Vm,N8=bu,T8=bu,Eh=xc,k8=Nh,$v=Gm,Gy={binaryAccuracy:Th,categoricalAccuracy:kh,precision:Av,categoricalCrossentropy:Eh,sparseCategoricalCrossentropy:$v,mse:C8,MSE:I8,mae:S8,MAE:v8,mape:N8,MAPE:T8,cosine:k8};function Q$(r){if(typeof r==\"string\"&&r in Gy)return Gy[r];if(typeof r!=\"string\"&&r!=null)return r;throw new M(`Unknown metric ${r}`)}function _h(r){if(ro(r!==null,`Unknown LossOrMetricFn ${r}`),typeof r==\"string\")return r;{let t;for(let e of Object.keys(vh))if(vh[e]===r){t=e;break}if(t!==void 0)return t;for(let e of Object.keys(Gy))if(Gy[e]===r){t=e;break}return t!==void 0?t:r.name}}function eD(r){let t={Adagrad:()=>ic.adagrad(.01),Adadelta:()=>ic.adadelta(1,.95,ir()),Adam:()=>ic.adam(.001,.9,.999,ir()),Adamax:()=>ic.adamax(.002,.9,.999,ir(),0),RMSProp:()=>ic.rmsprop(.001,.9,0,ir()),SGD:()=>ic.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,r in t)return t[r]();throw new M(`Unknown Optimizer ${r}`)}function Rv(r,t,e=!1){if(r==null||typeof r!=\"object\"||Object.getPrototypeOf(r)!==Object.prototype||!Dv(r))throw new Error(\"User-defined metadata is expected to be a JSON object, but is not.\");if(e){let n=JSON.stringify(r);n.length>1048576&&console.warn(`User-defined metadata of model \"${t}\" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${1048576}.`)}}function Dv(r){if(r===null)return!0;if(typeof r==\"object\")if(Object.getPrototypeOf(r)===Object.prototype){let t=Object.keys(r);for(let e of t)if(typeof e!=\"string\"||!Dv(r[e]))return!1;return!0}else if(Array.isArray(r)){for(let t of r)if(!Dv(t))return!1;return!0}else return!1;else{let t=typeof r;return t===\"string\"||t===\"number\"||t===\"boolean\"}}function rD(r,t,e,n=console.log){let o=A8(r),s=[\"Layer (type)\",\"Input Shape\",\"Output shape\",\"Param #\"];o?(t=t||90,e=e||[.32,.61,.89,1]):(t=t||115,e=e||[.24,.48,.7,.8,1]),e[e.length-1]<=1&&(e=e.map(c=>Math.floor(t*c)));let i;if(!o){s.push(\"Receives inputs\"),i=[];for(let c in r.nodesByDepth)i.push(...r.nodesByDepth[c])}n(\"_\".repeat(t)),Hy(s,e,n),n(\"=\".repeat(t));let a=r.layers;for(let c=0;c<a.length;++c)o?$8(a[c],e,n):D8(a[c],e,i,n),n((c===a.length-1?\"=\":\"_\").repeat(t));r.checkTrainableWeightsConsistency();let u=_8(r),l=Om(r.nonTrainableWeights);n(`Total params: ${u+l}`),n(`Trainable params: ${u}`),n(`Non-trainable params: ${l}`),n(\"_\".repeat(t))}function _8(r){let t;return r.collectedTrainableWeights!=null?t=Om(r.collectedTrainableWeights):t=Om(r.trainableWeights),t}function A8(r){let t=!0,e=[],n=[];for(let o in r.nodesByDepth)e.push(r.nodesByDepth[o]);for(let o of e){if(o.length>1||o.length===1&&o[0].inboundLayers.length>1){t=!1;break}n.push(...o)}if(t)for(let o of r.layers){let s=!1;for(let i of o.inboundNodes)if(n.indexOf(i)!==-1)if(s){t=!1;break}else s=!0;if(!t)break}return t}function Hy(r,t,e=console.log){let n=\"\";for(let o=0;o<r.length;++o)o>0&&(n=n.slice(0,n.length-1)+\" \"),n+=r[o],n=n.slice(0,t[o]),n+=\" \".repeat(t[o]-n.length);e(n)}function $8(r,t,e){let n,o;try{o=r.inboundNodes.map(u=>JSON.stringify(u.inputShapes)).join(\",\")}catch(u){o=\"multiple\"}try{n=JSON.stringify(r.outputShape)}catch(u){n=\"multiple\"}let s=r.name,i=r.getClassName(),a=[`${s} (${i})`,o,n,r.countParams().toString()];Hy(a,t,e)}function D8(r,t,e,n){let o,s;try{s=r.inboundNodes.map(p=>JSON.stringify(p.inputShapes)).join(\",\")}catch(p){s=\"multiple\"}try{o=JSON.stringify(r.outputShape)}catch(p){o=\"multiple\"}let i=[];for(let p of r.inboundNodes)if(!(e!=null&&e.length>0&&e.indexOf(p)===-1))for(let m=0;m<p.inboundLayers.length;++m){let f=p.inboundLayers[m].name,d=p.nodeIndices[m],h=p.tensorIndices[m];i.push(`${f}[${d}][${h}]`)}let a=r.name,u=r.getClassName(),l=i.length===0?\"\":i[0],c=[`${a} (${u})`,s,o,r.countParams().toString(),l];Hy(c,t,n);for(let p=1;p<i.length;++p)Hy([\"\",\"\",\"\",\"\",i[p]],t,n)}function nD(r,t,e){return(r===\"inboundNodes\"||r===\"outputLayers\"||r===\"inputLayers\")&&t===0&&typeof e==\"string\"}function yc(r,t){if(r===null)return null;if(typeof r==\"string\")return el(r);if(typeof r==\"number\"||typeof r==\"boolean\")return r;if(r instanceof Array){let e=[],n=r.length;for(let o=0;o<n;++o){let s=r[o];nD(t,o,s)?e.push(s):e.push(yc(s,t))}return e}else{let e={};for(let n of Object.keys(r)){let o=r[n];if(n===\"name\"&&typeof o==\"string\")e[n]=o;else{let s=el(n);e[s]=yc(o,s)}}return e}}function qy(r,t){if(r==null)return null;if(typeof r==\"string\")return So(r);if(typeof r==\"number\"||typeof r==\"boolean\")return r;if(r instanceof Array){let e=[],n=r.length;for(let o=0;o<n;++o){let s=r[o];nD(t,o,s)?e.push(s):e.push(qy(s,t))}return e}else{let e={};for(let n of Object.keys(r)){let o=r[n],s=So(n);(n===\"name\"||n===\"className\")&&typeof o==\"string\"?e[s]=o:e[s]=qy(o,n)}return e}}var Um=\"4.0.0\";var zn=class extends $t{constructor(t){if(super({}),this.containerNodes=new Set,this.name=t.name,this.name==null){let b=this.getClassName().toLowerCase();this.name=gu(b)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(t.inputs)?this.inputs=t.inputs.slice():this.inputs=[t.inputs],Array.isArray(t.outputs)?this.outputs=t.outputs.slice():this.outputs=[t.outputs],vo(this.inputs).length!==this.inputs.length)throw new M(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(b=>b.name)}`);vo(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(b=>b.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let b of this.outputs){let w=b.sourceLayer,C=b.nodeIndex,N=b.tensorIndex;this.outputLayers.push(w),this.outputLayersNodeIndices.push(C),this.outputLayersTensorIndices.push(N)}for(let b of this.inputs){let w=b.sourceLayer,C=b.nodeIndex,N=b.tensorIndex;ro(C===0,\"input layer has >1 nodes\"),ro(N===0,\"input layer has >1 tensors\"),this.inputLayers.push(w),this.inputLayersNodeIndices.push(C),this.inputLayersTensorIndices.push(N)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let b=0;b<this.inputLayers.length;b++){let w=this.inputLayers[b];if(!(w instanceof Ks))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${t.inputs}. Input ${b} (0-based) originates from layer type ${w.getClassName()}.`);this.inputNames.push(w.name),this.feedInputShapes.push(w.batchInputShape),this.feedInputNames.push(w.name)}for(let b of this.outputLayers)this.outputNames.push(b.name);this.internalInputShapes=this.inputs.map(b=>b.shape),this.internalOutputShapes=this.outputs.map(b=>b.shape);let e={},n={},o={},s={},i={},a=[],u=(b,w,C,N,_,A)=>{(N==null||_==null||A==null)&&(N=b.sourceLayer,_=b.nodeIndex,A=b.tensorIndex);let $=N.inboundNodes[_];if(C.indexOf($)!==-1)throw new Hr(`The tensor ${b.name} at layer \"${N.name}\" is part of a cycle.`);if(w.indexOf($)!==-1)return;this.containerNodes.add(zn.nodeKey(N,_)),N.id in i||(i[N.id]=Object.keys(i).length),C.indexOf($)===-1&&C.push($);let F=$.inboundLayers.length;for(let P=0;P<F;P++){let V=$.inputTensors[P],G=$.inboundLayers[P],W=$.nodeIndices[P],q=$.tensorIndices[P];u(V,w,C,G,W,q)}for(w.push($);C.indexOf($)>=0;)C.splice(C.indexOf($),1);a.push($)},l=[],c=[];for(let b of this.outputs)u(b,l,c);let p=a.slice().reverse();for(let b of p){n[b.id]=b,b.id in e||(e[b.id]=0);let w=e[b.id],C=o[b.outboundLayer.id]==null?0:o[b.outboundLayer.id];w=Math.max(w,C),o[b.outboundLayer.id]=w,s[b.outboundLayer.id]=b.outboundLayer,e[b.id]=w;for(let N=0;N<b.inboundLayers.length;N++){let _=b.inboundLayers[N],A=b.nodeIndices[N],$=_.inboundNodes[A],F=e[$.id]==null?0:e[$.id];e[$.id]=Math.max(w+1,F),n[$.id]=$}}let m={};for(let b in e){let w=e[b];w in m||(m[w]=[]),m[w].push(n[b])}let f={};for(let b in o){let w=o[b];w in f||(f[w]=[]),f[w].push(s[b])}let d=Object.keys(f).map(b=>parseInt(b,10)).sort(yh);this.layers=[];for(let b of d){let w=f[b];w.sort((C,N)=>{let _=i[C.id],A=i[N.id];return _<A?-1:_>A?1:0});for(let C of w)C instanceof zn&&this.internalContainerRefs.push(C),this.layers.push(C)}this.layersByDepth=f,d=Object.keys(m).map(b=>parseInt(b,10)).sort(yh);let h=this.inputs.slice(),g=[];for(let b of d)for(let w of m[b]){let C=w.outboundLayer;if(C!=null){for(let N of w.inputTensors)if(h.indexOf(N)===-1)throw new Hr(`Graph disconnected: cannot obtain value for tensor ${N} at layer \"${C.name}\". The following previous layers were accessed without issue: ${g}`);for(let N of w.outputTensors)h.push(N);g.push(C.name)}}this.nodesByDepth=m;let x=this.layers.map(b=>b.name);for(let b of x){let w=x.filter(C=>C===b).length;if(w!==1)throw new Hr(`The name \"${b}\" is used ${w} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(x))}this.outboundNodes=[],this.inboundNodes=[],new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(b=>null),outputMasks:this.outputs.map(b=>null),inputShapes:this.inputs.map(b=>b.shape),outputShapes:this.outputs.map(b=>b.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let t={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let e of this.layers)t.numDisposedVariables+=e.dispose().numDisposedVariables;for(let e of this.internalContainerRefs)t.numDisposedVariables+=e.dispose().numDisposedVariables}return t.refCountAfterDispose=this._refCount,t}get trainable(){return this.trainable_}set trainable(t){this.layers.forEach(e=>{e._trainableWeights.forEach(n=>n.trainable=t)}),this.trainable_=t}get trainableWeights(){if(this._trainableWeights.length>0)throw new M(\"Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.\");if(!this.trainable)return[];let t=[];for(let e of this.layers)t=t.concat(e.trainableWeights);return t}get nonTrainableWeights(){let t=[];for(let e of this.layers)t.push(...e.nonTrainableWeights);if(!this.trainable){let e=[];for(let n of this.layers)e.push(...n.trainableWeights);return e.concat(t)}return t}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(t,e=!0){let n={},o=0;for(let i of this.layers)for(let a of i.weights){if(n[a.originalName]!=null)throw new M(`Duplicate weight name: ${a.originalName}`);n[a.originalName]=a,o++}let s=[];for(let i in t){let a=i;if(n[i]==null){let u=i.split(\"/\");a=u.slice(0,-2).concat([u[u.length-1]]).join(\"/\")}if(n[a]!=null)s.push([n[a],t[i]]);else if(e)throw new M(`Provided weight data has no target variable: ${i}`);delete n[a]}if(e){let i=[];for(let a in n)i.push(a);if(i.length>0)throw new M(`${i.length} of ${o} weights are not set: ${i}`)}Pm(s)}updatedConfig(){let t=this.getConfig(),e={};return e.className=this.getClassName(),e.config=t,e.kerasVersion=`tfjs-layers ${Um}`,e.backend=\"TensorFlow.js\",e}toJSON(t,e=!0){let n=qy(this.updatedConfig());return e?JSON.stringify(n):n}call(t,e){return B(()=>{t=xe(t);let n=new ko;for(let o=0;o<this.inputs.length;++o)n.add(this.inputs[o],t[o]);return hc(this.outputs,n,e)})}computeMask(t,e){return B(()=>{t=xe(t);let n;return e==null?n=Io(null,t.length):n=xe(e),this.runInternalGraph(t,n)[1]})}computeOutputShape(t){let e=Fm(t);if(e.length!==this.inputLayers.length)throw new M(`Invalid inputShape argument ${t}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let a=0;a<e.length;a++){let u=this.inputLayers[a],l=e[a],c=u.name+\"_0_0\";n[c]=l}let o=Object.keys(this.nodesByDepth).map(a=>parseInt(a,10)).sort(yh);if(o.length>1)for(let a of o){let u=this.nodesByDepth[a];for(let l of u){let c=l.outboundLayer;if(this.inputLayers.map(h=>h.id).indexOf(c.id)!==-1)continue;let p=[];for(let h=0;h<l.inboundLayers.length;h++){let g=l.inboundLayers[h],x=l.nodeIndices[h],b=l.tensorIndices[h],w=`${g.name}_${x}_${b}`,C=n[w];p.push(C)}let m=c.computeOutputShape(Nr(p)),f=Fm(m),d=c.inboundNodes.indexOf(l);for(let h=0;h<f.length;h++){let g=`${c.name}_${d}_${h}`;n[g]=f[h]}}}let s=[],i=[];for(let a=0;a<this.outputLayers.length;a++){let u=this.outputLayers[a],l=this.outputLayersNodeIndices[a],c=this.outputLayersTensorIndices[a],p=`${u.name}_${l}_${c}`;i.push(p)}for(let a=0;a<i.length;a++){let u=i[a];ro(u in n),s.push(n[u])}return Nr(s)}runInternalGraph(t,e){e==null&&(e=Io(null,t.length));let n={};for(let u=0;u<this.inputs.length;++u){let l=this.inputs[u],c=t[u],p=e[u];n[l.id]=[c,p]}let o=Object.keys(this.nodesByDepth).map(u=>parseInt(u,10)).sort(yh);for(let u of o){let l=this.nodesByDepth[u];for(let c of l){let p=c.outboundLayer,m=c.inputTensors,f=c.outputTensors,d=new Array;for(let h of m)h.id in n&&d.push(n[h.id]);if(d.length===m.length){let h={},g,x,b,w;if(c.callArgs!=null&&(h=c.callArgs),d.length===1){let[C,N]=d[0];h.mask==null&&(h.mask=N),b=xe(p.call(C,h)),w=xe(p.computeMask(C,N)),g=[C],x=[N]}else g=d.map(C=>C[0]),x=d.map(C=>C[1]),h.mask==null&&(h.mask=x),b=xe(p.call(g,h)),w=xe(p.computeMask(g,x));if(p.activityRegularizer)throw new St(\"LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.\");for(let C=0;C<f.length;++C){let N=f[C],_=b[C],A=w[C];n[N.id]=[_,A]}}}}let s=[],i=[],a=[];for(let u of this.outputs){ro(u.id in n,`Could not compute output ${u.name} : ${u.id}`);let[l,c]=n[u.id];a.push(l.shape),s.push(l),i.push(c)}return[s,i,a]}buildNodeConversionMap(t){let e={},n;for(let o of this.layers){n=o instanceof zn?1:0;for(let s=0;s<o.inboundNodes.length;s++){let i=zn.nodeKey(o,s);this.containerNodes.has(i)&&(e[i]=n,n+=1)}}return e}getLayer(t,e){if(e!=null){if(this.layers.length<=e)throw new M(`Was asked to retrieve layer at index ${e}, but model only has ${this.layers.length} layer(s).`);return this.layers[e]}else if(t==null)throw new M(\"Provide either a layer name or layer index\");for(let n of this.layers)if(n.name===t)return n;throw new M(`No such layer: ${t}`)}calculateLosses(){return B(()=>{let t=[];for(let e of this.layers)for(let n=0;n<e.inboundNodes.length;++n){let o=zn.nodeKey(e,n);this.containerNodes.has(o)&&t.push(...e.calculateLosses())}return t})}getConfig(){let t={name:this.name},e=this.buildNodeConversionMap(this.layers),n=[];for(let i of this.layers){let a=i.getClassName(),u=i.getConfig(),l=[];for(let p=0;p<i.inboundNodes.length;p++){let m=i.inboundNodes[p],f=zn.nodeKey(i,p),d={};if(this.containerNodes.has(f)){if(m.callArgs)try{JSON.stringify(m.callArgs),d=m.callArgs}catch(h){console.warn(`Layer ${i.name} was passed non-serializable keyword arguments: ${m.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),d={}}if(m.inboundLayers.length>0){let h=[];for(let g=0;g<m.inboundLayers.length;g++){let x=m.inboundLayers[g],b=m.nodeIndices[g],w=m.tensorIndices[g],C=zn.nodeKey(x,b),N=e[C];N==null&&(N=0),h.push([x.name,N,w,d])}l.push(h)}}}let c={};c.name=i.name,c.className=a,c.config=u,c.inboundNodes=l,n.push(c)}t.layers=n;let o=[];for(let i=0;i<this.inputLayers.length;i++){let a=this.inputLayers[i],u=this.inputLayersNodeIndices[i],l=zn.nodeKey(a,u);if(!this.containerNodes.has(l))continue;let c=e[l];c==null&&(c=0);let p=this.inputLayersTensorIndices[i];o.push([a.name,c,p])}t.inputLayers=o;let s=[];for(let i=0;i<this.outputLayers.length;i++){let a=this.outputLayers[i],u=this.outputLayersNodeIndices[i],l=zn.nodeKey(a,u);if(!this.containerNodes.has(l))continue;let c=e[l];c==null&&(c=0);let p=this.outputLayersTensorIndices[i];s.push([a.name,c,p])}return t.outputLayers=s,t}static fromConfig(t,e,n={},o=!1){let s={},i={};function a(g,x){g.name in i?i[g.name].push(x):i[g.name]=[x]}function u(g,x){let b=[],w;for(let C of x){let N=C[0],_=C[1],A=C[2];if(w=C[3]==null?{}:C[3],!(N in s)){a(g,x);return}let $=s[N];if($.inboundNodes.length<=_){a(g,x);return}let F=$.inboundNodes[_];b.push(F.outputTensors[A])}b.length>0&&g.apply(Nr(b),w)}function l(g){let x=g.name,b=gn(g,e.customObjects!=null?e.customObjects:{});b.setFastWeightInitDuringBuild(o),s[x]=b,g.inboundNodes.forEach(C=>{if(!(C instanceof Array))throw new M(`Corrupted configuration, expected array for nodeData: ${C}`);a(b,C)})}let c=e.name,p=e.layers;for(let g of p)l(g);for(;!b$(i);)for(let g of p){let x=s[g.name];if(x.name in i){let b=i[x.name];delete i[x.name];for(let w of b)u(x,w)}}let m=[],f=[],d=e.inputLayers;for(let g of d){let x=g[0],b=g[1],w=g[2];ro(x in s);let N=s[x].inboundNodes[b].outputTensors;m.push(N[w])}let h=e.outputLayers;for(let g of h){let x=g[0],b=g[1],w=g[2];ro(x in s);let N=s[x].inboundNodes[b].outputTensors;f.push(N[w])}return new t({inputs:m,outputs:f,name:c})}get stateful(){if(this._stateful)throw new M(\"Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.\");for(let t of this.layers)if(t.stateful)return!0;return!1}resetStates(){B(()=>{this.layers.forEach(t=>{t.stateful&&t.resetStates()})})}};function R8(r,t,e){let n=t.length;if(r==null||Array.isArray(r)&&r.length===0)return t.map(o=>null);if(n===1)return Array.isArray(r)&&r.length===1?r:typeof r==\"object\"&&t[0]in r?[r[t[0]]]:[r];if(Array.isArray(r)){if(r.length!==n)throw new Error(`Provided ${e} is an array of ${r.length} element(s), but the model has ${n} outputs. Make sure a set of weights is provided for each model output.`);return r}else if(typeof r==\"object\"&&Object.keys(r).length>0&&typeof r[Object.keys(r)[0]]==\"object\"){let o=[];return t.forEach(s=>{s in r?o.push(r[s]):o.push(null)}),o}else throw new Error(`The model has multiple (${n}) outputs, so ${e} must be either an array with ${n} elements or an object with ${t} keys. Provided ${e} not understood: ${JSON.stringify(r)}`)}function Ky(r,t){return R8(r,t,\"classWeight\")}async function jy(r,t,e,n){if(t!=null||n!=null)throw new Error(\"Support sampleWeight is not implemented yet\");if(e!=null){let o=B(()=>{if(r.shape.length===1)return sn(r);if(r.shape.length===2){if(r.shape[1]>1)return Ai(r,1);if(r.shape[1]===1)return R(r,[r.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${r.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${r.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),s=Array.from(await o.data());vt(o);let i=[];return s.forEach(a=>{if(e[a]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${a} exists in the data but not in classWeight`);i.push(e[a])}),Me(i,\"float32\")}else return null}function oD(r,t){return D(r,t)}var F8=32;function aD(r,t){let e,n,o=t;e=o.xs,n=o.ys,y.assert(e!=null&&n!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \\`{xs: xVal, ys: yVal}\\`, where the two values may be \\`tf.Tensor\\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let s=sD(\"input\",r.inputNames,e),i=sD(\"output\",r.outputNames,n),a=s[0].shape[0];y.assert(s.length===r.inputs.length,()=>`LayersModel has ${r.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(r.inputNames)})`),y.assert(i.length===r.outputs.length,()=>`LayersModel has ${r.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(r.outputNames)})`);for(let u=0;u<s.length;u++)y.assert(s[u].shape[0]===a,()=>`Batch size mismatch: input ${r.inputNames[u]} has ${s[u].shape[0]}; expected  ${a} based on input ${r.inputNames[0]}.`);for(let u=0;u<i.length;u++)y.assert(i[u].shape[0]===a,()=>`Batch size mismatch: output ${r.outputNames[u]} has ${i[u].shape[0]}; expected  ${a} based on input ${r.inputNames[0]}.`);return{xs:s,ys:i}}function sD(r,t,e){if(e instanceof Ft)return[e];if(Array.isArray(e))return y.assert(e.length===t.length,()=>`Received an array of ${e.length} Tensors, but expected ${t.length} to match the ${r} keys ${t}.`),e;{let n=[];for(let o of t){if(e[o]==null)throw new M(`The feature data generated by the dataset lacks the required ${r} key '${o}'.`);n.push(e[o])}return n}}function O8(r){if(r.length===3)throw new St(\"Validation with sample weights is not implemented yet.\");return{xs:r[0],ys:r[1]}}async function lD(r,t,e){let n=e.batchesPerEpoch!=null;if(y.assert(r.optimizer!=null,()=>\"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig).\"),y.assert(e!=null,()=>\"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call.\"),y.assert(e.epochs!=null&&e.epochs>0&&Number.isInteger(e.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${e.epochs}`),y.assert(!n||e.batchesPerEpoch>0&&Number.isInteger(e.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${e.batchesPerEpoch}`),y.assert(e.validationSplit==null,()=>\"`validationSplit` is not supported by `fitDataset()`. Use validationData instead.\"),r.isTraining)throw new Error(\"Cannot start training because another fit() call is ongoing.\");r.isTraining=!0;try{let o=e.validationData!=null,s,i;if(o)if(iD(e.validationData))y.assert(e.validationBatches==null||e.validationBatches>0&&Number.isInteger(e.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${e.validationBatches}`);else{let g=O8(e.validationData);s=g.xs,i=g.ys}let a=r.makeTrainFunction(),u=r.getDedupedMetricsNames(),l;o?l=u.slice().concat(u.map(g=>\"val_\"+g)):l=u.slice();let c=zy(e.callbacks,e.yieldEvery),p=e.verbose==null?1:e.verbose,{callbackList:m,history:f}=By(c,p,e.epochs,null,null,P8(t,e),null,o,l);m.setModel(r),r.history=f,await m.onTrainBegin(),r.stopTraining_=!1;let d=e.initialEpoch==null?0:e.initialEpoch,h=await t.iterator();for(;d<e.epochs;){let g={};await m.onEpochBegin(d);let x=0,b=0;for(n||(h=await t.iterator());!n||x<e.batchesPerEpoch;){let w=await h.next();if(n&&w.done){console.warn(`You provided \\`batchesPerEpoch\\` as ${e.batchesPerEpoch}, but your dataset iterator ran out of data after ${x} batches; interrupting training. Make sure that your dataset can generate at least \\`batchesPerEpoch * epochs\\` batches (in this case, ${e.batchesPerEpoch*e.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(w.value!=null){let{xs:C,ys:N}=aD(r,w.value),_={};_.batch=b,_.size=C[0].shape[0],await m.onBatchBegin(b,_);let A=[];if(e.classWeight!=null){let P=Ky(e.classWeight,r.outputNames);for(let V=0;V<P.length;++V)A.push(await jy(N[V],null,P[V]))}let $=C.concat(N).concat(A),F=a($);vt($);for(let P=0;P<u.length;++P){let V=u[P],G=F[P];_[V]=G,De(G)}await m.onBatchEnd(b,_),Oy(_),b++,x++}if(n?x>=e.batchesPerEpoch:w.done){if(o){let C;iD(e.validationData)?C=xe(await r.evaluateDataset(e.validationData,{batches:e.validationBatches})):C=xe(r.evaluate(s,i,{batchSize:e.validationBatchSize==null?F8:e.validationBatchSize,verbose:0}));for(let N=0;N<r.metricsNames.length;++N)g[`val_${r.metricsNames[N]}`]=C[N]}break}if(r.stopTraining_)break}if(await m.onEpochEnd(d,g),d++,r.stopTraining_)break}return await m.onTrainEnd(),await r.history.syncData(),r.history}finally{r.isTraining=!1}}function P8(r,t){let e=null;return t.batchesPerEpoch!=null?e=t.batchesPerEpoch:Number.isFinite(r.size)&&(e=r.size),e}function iD(r){return typeof r.iterator==\"function\"}function L8(r){return typeof r.next==\"function\"}async function uD(r,t,e){e=e||{};let n=e.batches!=null,o=r.testFunction,s=[];if(e.verbose>0)throw new St(\"Verbose mode is not implemented yet.\");y.assert(!n||e.batches>0&&Number.isInteger(e.batches),()=>`Test loop expects \\`batches\\` to be a positive integer, but received ${JSON.stringify(e.batches)}`);let i=L8(t)?t:await t.iterator(),a=0,u=0;for(;!n||u<e.batches;){let l=await i.next();if(s=B(()=>{if(l.value){let{xs:c,ys:p}=aD(r,l.value),m=c.concat(p),f=B(()=>o(m));if(vt(m),u===0)for(let h=0;h<f.length;++h)s.push(mt(0));let d=m[0].shape[0];for(let h=0;h<f.length;++h){let g=f[h],x=s[h];s[h]=B(()=>X(s[h],D(d,g))),u>0&&vt(x)}vt(f),a+=d,++u}return s}),l.done){n&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \\`batches\\` batches (in this case, ${e.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let l=0;l<s.length;++l){let c=s[l];s[l]=pt(s[l],a),vt(c)}return Nr(s)}function Xy(r){y.assert(r>0&&Number.isInteger(r),()=>`batchSize is required to be a positive integer, but got ${r}`)}function Hm(r,t,e){return r==null?[null]:Array.isArray(r)?r.map(n=>rl(n,t,e-t)):rl(r,t,e-t)}function Yy(r,t){return B(()=>r==null?null:Array.isArray(r)?r.map(e=>Yy(e,t)):_y(r,t.dtype===\"int32\"?t:J(t,\"int32\")))}function Zy(r,t){let e=[],n=0,o=null;for(;n<r;)o=n+t,o>=r&&(o=r),e.push([n,o]),n=o;return e}async function M8(r,t,e,n,o,s,i,a,u,l,c,p,m,f,d){o==null&&(o=32),s==null&&(s=1),c==null&&(c=!0),m==null&&(m=0);let h=!1;if(u!=null&&l!=null&&(h=!0),d!=null&&(h=!0,f==null))throw new M(\"Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.\");let g=r.checkNumSamples(e,o,f,\"steps_per_epoch\"),x;g!=null&&(x=Zr(0,g)),i==null&&(i=1);let{callbackList:b,history:w}=By(a,i,s,m,g,f,o,h,p);b.setModel(r),r.history=w,await b.onTrainBegin(),r.stopTraining_=!1;for(let C=m;C<s;++C){await b.onEpochBegin(C);let N={};if(f!=null)throw new St(\"stepsPerEpoch mode is not implemented yet.\");{if(c===\"batch\")throw new St(\"batch shuffling is not implemneted yet\");c&&y.shuffle(x);let _=Me(x),A=Zy(g,o);for(let $=0;$<A.length;++$){let F={};if(await b.onBatchBegin($,F),B(()=>{let P=A[$][0],V=A[$][1],G=rl(_,P,V-P);F.batch=$,F.size=V-P;let W=Yy(e,G),q=t(W);for(let H=0;H<n.length;++H){let j=n[H],Y=q[H];F[j]=Y,De(Y)}if($===A.length-1&&h){let H=r.testLoop(u,l,o);for(let j=0;j<n.length;++j){let Y=n[j],Z=H[j];De(Z),N[\"val_\"+Y]=Z}}}),await b.onBatchEnd($,F),Oy(F),r.stopTraining_)break}_.dispose()}if(await b.onEpochEnd(C,N),r.stopTraining_)break}return await b.onTrainEnd(),await r.history.syncData(),r.history}async function cD(r,t,e,n={}){if(r.isTraining)throw new Error(\"Cannot start training because another fit() call is ongoing.\");r.isTraining=!0;let o,s,i,a,u,l,c,p,m;try{let f=n.batchSize==null?32:n.batchSize;Xy(f);let d=!1,h=await r.standardizeUserData(t,e,n.sampleWeight,n.classWeight,d,f);o=h[0],s=h[1],m=h[2];let g=!1,x;if(n.validationData!=null&&n.validationData.length>0){if(g=!0,n.validationData.length===2)u=n.validationData[0],l=n.validationData[1];else throw n.validationData.length===3?new St(\"validationData including sample weights is not supported yet.\"):new M(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${n.validationData} is invalid.`);let F=!0,P=await r.standardizeUserData(u,l,null,null,F,f);c=P[0],p=P[1],x=c.concat(p)}else if(n.validationSplit!=null&&n.validationSplit>0&&n.validationSplit<1){g=!0;let F=Math.floor(o[0].shape[0]*(1-n.validationSplit)),P=o[0].shape[0];c=Hm(o,F,P),i=o,o=Hm(o,0,F),p=Hm(s,F,P),a=s,s=Hm(s,0,F),x=c.concat(p)}else n.validationSteps!=null&&(g=!0);let b=o.concat(s).concat(m);r.checkTrainableWeightsConsistency();let w=r.makeTrainFunction(),C=r.getDedupedMetricsNames(),N,_;g?(r.makeTestFunction(),N=r.testFunction,_=C.slice().concat(C.map(F=>\"val_\"+F))):(N=null,x=[],_=C.slice());let A=zy(n.callbacks,n.yieldEvery);return await M8(r,w,b,C,f,n.epochs,n.verbose,A,N,x,n.shuffle,_,n.initialEpoch,null,null)}finally{r.isTraining=!1,Eo(o,t),Eo(s,e),Eo(i,t),Eo(a,e),Eo(c,u),Eo(p,l),m!=null&&vt(m)}}function Fv(r){let t=[];r instanceof Ft&&(r=[r]);for(let e=0;e<r.length;++e){let n=r[e];if(n.rank===1)t.push(nl(n,1));else{if(n.rank===0)throw new Error(\"Expected tensor to be at least 1D, but received a 0D tensor (scalar).\");t.push(n)}}return t}function Eo(r,t){if(r==null)return;let e=[];if(t instanceof Ft)e.push(t.id);else if(Array.isArray(t))t.forEach(o=>e.push(o.id));else if(t!=null)for(let o in t){let s=t[o];e.push(s.id)}let n=[];if(r instanceof Ft)e.indexOf(r.id)===-1&&n.push(r);else if(Array.isArray(r))r.forEach(o=>{e.indexOf(o.id)===-1&&n.push(o)});else if(r!=null)for(let o in r){let s=r[o];e.indexOf(s.id)===-1&&n.push(s)}n.forEach(o=>{o.isDisposed||o.dispose()})}function z8(r){return r instanceof Ft}function Ov(r){return Array.isArray(r)}function pD(r){return!z8(r)&&!Ov(r)}function mD(r,t,e,n=!0,o=\"\"){if(t==null||t.length===0){if(r!=null){let i=!1;if(Ov(r)&&r.length>0)i=!0;else if(pD(r)){for(let a in r)if(r.hasOwnProperty(a)){i=!0;break}}else i=!0;if(i)throw new M(`Error when checking model ${o} expected no data, but got ${r}`)}return[]}if(r==null)return t.map(i=>null);let s;if(pD(r)){r=r,s=[];for(let i of t){if(r[i]==null)throw new M(`No data provided for \"${i}\". Need data for each key in: ${t}`);s.push(r[i])}}else if(Ov(r)){if(r=r,r.length!==t.length)throw new M(`Error when checking model ${o}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${r}`);s=r}else{if(r=r,t.length>1)throw new M(`The model ${o} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${r.shape}`);s=[r]}if(s=Fv(s),e!=null)for(let i=0;i<t.length;++i){if(e[i]==null)continue;let a=s[i];if(a.shape.length!==e[i].length)throw new M(`Error when checking ${o}: expected ${t[i]} to have ${e[i].length} dimension(s). but got array with shape ${a.shape}`);for(let u=0;u<e[i].length;++u){if(u===0&&!n)continue;let l=a.shape[u],c=e[i][u];if(c!=null&&c>=0&&l!==c)throw new M(`${o} expected a batch of elements where each example has shape [${e[i].slice(1,e[i].length)}] (i.e.,tensor shape [*,${e[i].slice(1,e[i].length)}]) but the ${o} received an input with ${a.shape[0]} examples, each with shape [${a.shape.slice(1,a.shape.length)}] (tensor shape [${a.shape}])`)}}return s}function B8(r,t,e){let n=vo(r.map(s=>s.shape[0]));n.sort();let o=vo(t.map(s=>s.shape[0]));if(o.sort(),n.length>1)throw new M(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(r.map(s=>s.shape))}`);if(o.length>1)throw new M(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(s=>s.shape))}`);if(n.length>0&&o.length>0&&!y.arraysEqual(n,o))throw new M(`Input Tensors should have the same number of samples as target Tensors. Found ${n[0]} input sample(s) and ${o[0]} target sample(s).`)}function V8(r,t,e){let n=[Hi,Wm,xc];for(let o=0;o<r.length;++o){let s=r[o],i=t[o],a=e[o];if(i!=null){if(i===xc&&s.shape[s.shape.length-1]===1)throw new M(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(n.indexOf(i)!==-1){let u=s.shape.slice(1),l=a.slice(1);for(let c=0;c<u.length;++c){let p=u[c],m=l[c];if(m!=null&&p!==m)throw new M(`A target Tensor with shape ${s.shape} was passed for an output of shape ${a}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function fD(r,t,e,n=!0,o=\"\"){let s;if(Array.isArray(r)){if(r.length!==t.length)throw new M(`Error when checking model ${o}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${r.length} Tensors(s).`);s=r}else{if(t.length>1)throw new M(`The model expects ${t.length} ${o} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(r.shape)}.`);s=[r]}if(e!=null)for(let i=0;i<t.length;++i){if(e[i]==null)continue;let a=s[i];if(a.shape.length!==e[i].length)throw new M(`Error when checking ${o}: expected ${t[i]} to have ${e[i].length} dimension(s), but got array with shape ${JSON.stringify(a.shape)}`);for(let u=0;u<e[i].length;++u){if(u===0&&!n)continue;let l=a.shape[u],c=e[i][u];if(c!=null&&c!==l)throw new M(`Error when checking ${o}: expected ${t[i]} to have shape ${JSON.stringify(e[i])} but got array with shape ${JSON.stringify(a.shape)}.`)}}}function G8(r,t){if(r==null||Array.isArray(r)&&r.length===0)return t.map(n=>[]);let e;if(typeof r==\"string\"||typeof r==\"function\")e=[r];else if(Array.isArray(r)||typeof r==\"object\")e=r;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${r}`);if(Array.isArray(e))return t.map(n=>e);{let n=[];for(let o of t){let s=e.hasOwnProperty(o)?e[o]:[];Array.isArray(s)||(s=[s]),n.push(s)}return n}}var W8=\"layers-model\",Bn=class extends zn{constructor(t){super(t),this.isTraining=!1}summary(t,e,n=console.log){if(!this.built)throw new M(\"This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).\");rD(this,t,e,n)}compile(t){if(t.loss==null&&(t.loss=[]),this.loss=t.loss,typeof t.optimizer==\"string\")this.optimizer_=eD(t.optimizer),this.isOptimizerOwned=!0;else{if(!(t.optimizer instanceof Wr))throw new M(\"User-defined optimizer must be an instance of tf.Optimizer.\");this.optimizer_=t.optimizer,this.isOptimizerOwned=!1}let e=[];if(!Array.isArray(t.loss)&&typeof t.loss!=\"string\"&&typeof t.loss!=\"function\"){t.loss=t.loss;for(let i in t.loss)if(this.outputNames.indexOf(i)===-1)throw new M(`Unknown entry in loss dictionary: \"${i}\". Only expected the following keys: ${this.outputNames}`);for(let i of this.outputNames)t.loss[i]==null&&console.warn(`Output \"${i}\" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${i} during training`),e.push(Vy(t.loss[i]))}else if(Array.isArray(t.loss)){if(t.loss.length!==this.outputs.length)throw new M(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${t.loss}.`);e=t.loss.map(a=>Vy(a))}else{let i=Vy(t.loss);this.outputs.forEach(a=>{e.push(i)})}this.lossFunctions=e,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let i=0;i<this.outputs.length;++i){let a=this.internalOutputShapes[i],u=this.outputNames[i];this.feedOutputNames.push(u),this.feedOutputShapes.push(a),this.feedLossFns.push(this.lossFunctions[i])}let n=[];this.metrics=t.metrics,this.metricsNames=[\"loss\"],this.metricsTensors=[],Hs(\"loss\",()=>{for(let i=0;i<this.outputs.length;++i){if(n.indexOf(i)!==-1)continue;let a=this.lossFunctions[i];this.outputs.length>1&&(this.metricsTensors.push([a,i]),this.metricsNames.push(this.outputNames[i]+\"_loss\"))}});let o=G8(t.metrics,this.outputNames),s=(i,a,u)=>{this.outputNames.length>1&&(a=this.outputNames[i]+\"_\"+a),this.metricsNames.push(a),this.metricsTensors.push([u,i])};Hs(\"metric\",()=>{for(let i=0;i<this.outputs.length;++i){if(n.indexOf(i)!==-1)continue;let a=o[i];(l=>{let c=\"\",p,m,f;for(let d of l){if(typeof d==\"string\"&&[\"accuracy\",\"acc\",\"crossentropy\",\"ce\"].indexOf(d)!==-1){let g=this.internalOutputShapes[i];g[g.length-1]===1||this.lossFunctions[i]===Wm?[\"accuracy\",\"acc\"].indexOf(d)!==-1?m=Th:[\"crossentropy\",\"ce\"].indexOf(d)!==-1&&(m=Wy):this.lossFunctions[i]===Gm?[\"accuracy\",\"acc\"].indexOf(d)!==-1?m=Uy:[\"crossentropy\",\"ce\"].indexOf(d)!==-1&&(m=$v):[\"accuracy\",\"acc\"].indexOf(d)!==-1?m=kh:[\"crossentropy\",\"ce\"].indexOf(d)!==-1&&(m=Eh);let x;[\"accuracy\",\"acc\"].indexOf(d)!==-1?x=\"acc\":[\"crossentropy\",\"ce\"].indexOf(d)!==-1&&(x=\"ce\"),f=m,p=c+x}else f=Q$(d),p=c+_h(d);let h;Hs(p,()=>{h=f}),s(i,p,h)}})(a)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn(\"Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?\")}evaluate(t,e,n={}){let o=n.batchSize==null?32:n.batchSize;Xy(o);let s=!0,i=this.standardizeUserDataXY(t,e,s,o);try{let a=i[0].concat(i[1]);this.makeTestFunction();let u=this.testFunction,l=this.testLoop(u,a,o,n.verbose,n.steps);return Nr(l)}finally{Eo(i[0],t),Eo(i[1],e)}}async evaluateDataset(t,e){return this.makeTestFunction(),uD(this,t,e)}checkNumSamples(t,e,n,o=\"steps\"){let s;if(n!=null){if(s=null,e!=null)throw new M(`If ${o} is set, batchSize must be null or undefined.Got batchSize = ${e}`)}else if(t!=null)Array.isArray(t)?s=t[0].shape[0]:s=t.shape[0];else throw new M(`Either the input data should have a defined shape, or ${o} shoud be specified.`);return s}execute(t,e){if(Array.isArray(e)&&e.length===0)throw new M(\"`outputs` is an empty Array, which is not allowed.\");let n=Array.isArray(e),o=n?e:[e],s=this.retrieveSymbolicTensors(o),i=new ko;if(t instanceof Ft&&(t=[t]),Array.isArray(t)){if(t.length!==this.inputs.length)throw new M(`The number of inputs provided (${t.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let u=0;u<this.inputs.length;++u)i.add(this.inputs[u],t[u])}else for(let u of this.inputs){let l=t[u.name];if(l==null)throw new M(`No value is provided for the model's input ${u.name}`);i.add(u,l)}let a=hc(s,i);return n?a:a[0]}retrieveSymbolicTensors(t){let e=Io(null,t.length),n=t.length;for(let o of this.layers){let s=Array.isArray(o.output)?o.output:[o.output],i=s.map(a=>a.name);for(let a=0;a<t.length;++a){let u=i.indexOf(t[a]);if(u!==-1&&(e[a]=s[u],n--),n===0)break}if(n===0)break}if(n>0){let o=[];throw e.forEach((s,i)=>{s==null&&o.push(t[i])}),new M(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(o)}`)}return e}predictLoop(t,e=32,n=!1){return B(()=>{let o=this.checkNumSamples(t);if(n)throw new St(\"Verbose predictLoop() is not implemented yet.\");let s=Zy(o,e),i=this.outputs.map(a=>[]);for(let a=0;a<s.length;++a)B(()=>{let l=s[a][0],c=s[a][1],p=Hm(t,l,c),m=[];if(Array.isArray(p))for(let d=0;d<p.length;++d)m.push({key:this.inputs[d],value:p[d]});else m.push({key:this.inputs[0],value:p});let f=new ko(m);return hc(this.outputs,f)}).forEach((l,c)=>i[c].push(l));return Nr(i.map(a=>ne(a,0)))})}predict(t,e={}){let n=Fv(t);fD(n,this.inputNames,this.feedInputShapes,!1);try{let o=e.batchSize==null?32:e.batchSize;return Xy(o),this.predictLoop(n,o)}finally{Eo(n,t)}}predictOnBatch(t){fD(t,this.inputNames,this.feedInputShapes,!0);let e=(Array.isArray(t)?t[0]:t).shape[0];return this.predictLoop(t,e)}standardizeUserDataXY(t,e,n=!0,o){if(this.optimizer_==null)throw new Hr(\"You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).\");let s=[];for(let i=0;i<this.feedOutputShapes.length;++i){let a=this.feedOutputShapes[i];this.feedLossFns[i]===Gm?s.push(a.slice(0,a.length-1).concat([1])):s.push(a)}if(t=mD(t,this.feedInputNames,this.feedInputShapes,!1,\"input\"),e=mD(e,this.feedOutputNames,s,!1,\"target\"),B8(t,e,null),V8(e,this.feedLossFns,this.feedOutputShapes),this.stateful&&o!=null&&o>0&&t[0].shape[0]%o!==0)throw new M(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${o}. Found: ${t[0].shape[0]} sample(s).`);return[t,e]}async standardizeUserData(t,e,n,o,s=!0,i){let[a,u]=this.standardizeUserDataXY(t,e,s,i);if(n!=null)throw new Error(\"sample weight is not supported yet.\");let l=null;if(o!=null){let c=Ky(o,this.outputNames);l=[];for(let p=0;p<c.length;++p)l.push(await jy(u[p],null,c[p]))}return[a,u,l]}testLoop(t,e,n,o=0,s){return B(()=>{let i=this.checkNumSamples(e,n,s,\"steps\"),a=[];if(o>0)throw new St(\"Verbose mode is not implemented yet.\");if(s!=null)throw new St(\"steps mode in testLoop() is not implemented yet\");{let u=Zy(i,n),l=Me(Zr(0,i));for(let c=0;c<u.length;++c){let p=u[c][0],m=u[c][1],f=rl(l,p,m-p),d=Yy(e,f),h=t(d);if(c===0)for(let g=0;g<h.length;++g)a.push(mt(0));for(let g=0;g<h.length;++g){let x=h[g];a[g]=X(a[g],D(m-p,x))}}for(let c=0;c<a.length;++c)a[c]=pt(a[c],i)}return a})}getDedupedMetricsNames(){let t=this.metricsNames,e=[];for(let n=0;n<t.length;++n){let o=t[n],s=o;Cv(t,o)>1&&(s+=`_${Cv(t.slice(0,n),o)}`),e.push(s)}return e}makeTrainFunction(){return t=>{let e=[],n=t.slice(0,this.inputs.length),o=t.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=t.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),i=[],a=()=>{let p=[];for(let h=0;h<this.inputs.length;++h)p.push({key:this.inputs[h],value:n[h]});let m=new ko(p),f=hc(this.outputs,m,{training:!0}),d;for(let h=0;h<this.lossFunctions.length;++h){let g=this.lossFunctions[h],x=g(o[h],f[h]);s[h]!=null&&(x=oD(x,s[h]));let b=ve(x);e.push(b),h===0?d=x:d=X(d,x)}for(let h=0;h<this.metricsTensors.length;++h){let g;if(this.outputs.length>1&&h<this.outputs.length)g=e[h];else{let x=this.metricsTensors[h][0],b=this.metricsTensors[h][1];g=ve(x(o[b],f[b]))}De(g),i.push(g)}return d=ve(d),this.calculateLosses().forEach(h=>{d=X(d,h)}),d},u=this.collectedTrainableWeights.map(p=>p.read()),l=!0;return[this.optimizer_.minimize(a,l,u)].concat(i)}}makeTestFunction(){this.testFunction=t=>B(()=>{let e=[],n,o=t.slice(0,this.inputs.length),s=t.slice(this.inputs.length,this.inputs.length+this.outputs.length),i=[];for(let l=0;l<this.inputs.length;++l)i.push({key:this.inputs[l],value:o[l]});let a=new ko(i),u=hc(this.outputs,a);for(let l=0;l<this.lossFunctions.length;++l){let c=this.lossFunctions[l],p=ve(c(s[l],u[l]));l===0?n=p:n=X(n,p),e.push(n)}for(let l=0;l<this.metricsTensors.length;++l){let c=this.metricsTensors[l][0],p=this.metricsTensors[l][1],m=ve(c(s[p],u[p]));e.push(m)}return e})}async fit(t,e,n={}){return cD(this,t,e,n)}async fitDataset(t,e){return lD(this,t,e)}async trainOnBatch(t,e){let n=await this.standardizeUserData(t,e),o=n[0],s=n[1],a=this.makeTrainFunction()(o.concat(s)),u=[];for(let l of a){let c=await l.data();u.push(c[0])}return vt(a),Eo(n[0],t),Eo(n[1],e),Nr(u)}getNamedWeights(t){let e=[],n=t!=null&&t.trainableOnly,o=n?this.trainableWeights:this.weights,s=this.getWeights(n);for(let i=0;i<o.length;++i)n&&!o[i].trainable||e.push({name:o[i].originalName,tensor:s[i]});return e}set stopTraining(t){this.stopTraining_=t}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(t){this.optimizer_!==t&&(this.optimizer_=t,this.isOptimizerOwned=!1)}dispose(){let t=super.dispose();if(t.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let e=ah().numTensors;this.optimizer_.dispose(),t.numDisposedVariables+=e-ah().numTensors}return t}getLossIdentifiers(){let t;if(typeof this.loss==\"string\")t=So(this.loss);else if(Array.isArray(this.loss)){for(let e of this.loss)if(typeof e!=\"string\")throw new Error(\"Serialization of non-string loss is not supported.\");t=this.loss.map(e=>So(e))}else{let e=Object.keys(this.loss);t={};let n=this.loss;for(let o of e)if(typeof n[o]==\"string\")t[o]=So(n[o]);else throw new Error(\"Serialization of non-string loss is not supported.\")}return t}getMetricIdentifiers(){if(typeof this.metrics==\"string\"||typeof this.metrics==\"function\")return[So(_h(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(t=>So(_h(t)));{let t={};for(let e in this.metrics)t[e]=So(_h(this.metrics[e]));return t}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(t){if(t.weighted_metrics!=null)throw new Error(\"Loading weight_metrics is not supported yet.\");if(t.loss_weights!=null)throw new Error(\"Loading loss_weights is not supported yet.\");if(t.sample_weight_mode!=null)throw new Error(\"Loading sample_weight_mode is not supported yet.\");let e=yc(t.optimizer_config),n=gn(e),o;if(typeof t.loss==\"string\")o=el(t.loss);else if(Array.isArray(t.loss))o=t.loss.map(i=>el(i));else if(t.loss!=null){o={};for(let i in t.loss)o[i]=el(t.loss[i])}let s;if(Array.isArray(t.metrics))s=t.metrics.map(i=>el(i));else if(t.metrics!=null){s={};for(let i in t.metrics)s[i]=el(t.metrics[i])}this.compile({loss:o,metrics:s,optimizer:n})}async save(t,e){if(typeof t==\"string\"){let l=_r.getSaveHandlers(t);if(l.length===0)throw new M(`Cannot find any save handlers for URL '${t}'`);if(l.length>1)throw new M(`Found more than one (${l.length}) save handlers for URL '${t}'`);t=l[0]}if(t.save==null)throw new M(\"LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.\");let n=await _r.encodeWeights(this.getNamedWeights(e)),o=!1,s=null,a={modelTopology:this.toJSON(s,o),format:W8,generatedBy:`TensorFlow.js tfjs-layers v${Um}`,convertedBy:null};if((e==null?!1:e.includeOptimizer)&&this.optimizer!=null){a.trainingConfig=this.getTrainingConfig();let l=\"optimizer\",{data:c,specs:p}=await _r.encodeWeights(await this.optimizer.getWeights(),l);n.specs.push(...p),n.data=_r.concatenateArrayBuffers([n.data,c])}return this.userDefinedMetadata!=null&&(Rv(this.userDefinedMetadata,this.name,!0),a.userDefinedMetadata=this.userDefinedMetadata),a.weightData=n.data,a.weightSpecs=n.specs,t.save(a)}setUserDefinedMetadata(t){Rv(t,this.name),this.userDefinedMetadata=t}getUserDefinedMetadata(){return this.userDefinedMetadata}};Bn.className=\"Model\";Q.registerClass(Bn);var Jy=class extends Bn{};Jy.className=\"Functional\";Q.registerClass(Jy);async function dD(r,t){\"modelTopology\"in r||(r={modelTopology:r}),r=r;let e=r.modelTopology;e.model_config!=null&&(e=e.model_config);let n=yc(e),o=gn(n,t);if(r.weightsManifest!=null){let s=await _r.loadWeights(r.weightsManifest,r.pathPrefix,o.weights.map(a=>a.originalName)),i={};for(let a of o.weights)i[a.originalName]=s[a.originalName];o.loadWeights(i),vt(s)}return o}async function hD(r,t){if(t==null&&(t={}),typeof r==\"string\"){let e=_r.getLoadHandlers(r,t);if(e.length===0)e.push(_r.browserHTTPRequest(r,t));else if(e.length>1)throw new M(`Found more than one (${e.length}) load handlers for URL '${r}'`);r=e[0]}return U8(r,void 0,t)}async function U8(r,t,e){if(e==null&&(e={}),r.load==null)throw new M(\"Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.\");let n=await r.load(),o=n.modelTopology;o.model_config!=null&&(o=o.model_config);let s=e.strict==null?!0:e.strict,i=n.weightData!=null&&n.weightSpecs!=null&&s,a=gn(yc(o),t,i),u=n.trainingConfig;if(u!=null&&a.loadTrainingConfig(u),n.userDefinedMetadata!=null&&a.setUserDefinedMetadata(n.userDefinedMetadata),n.weightData!=null){if(n.weightSpecs==null)throw new M(\"LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.\");let{modelWeights:l,optimizerWeights:c}=H8(n.weightData,n.weightSpecs);a.loadWeights(l,s),a.optimizer!=null&&c.length>0&&await a.optimizer.setWeights(c),vt(l),vt(c.map(p=>p.tensor))}return a}function H8(r,t){let e=_r.decodeWeights(r,t),n={},o=[];return t.forEach(s=>{s.group===\"optimizer\"?o.push({name:s.name,tensor:e[s.name]}):n[s.name]=e[s.name]}),{modelWeights:n,optimizerWeights:o}}var qi=class extends Bn{constructor(t){if(super({inputs:[],outputs:[]}),t=t||{},this.trainable=!0,this.built=!1,this.name=t.name!=null?t.name:gu(\"sequential_\"),t.layers!=null)for(let e of t.layers)this.add(e)}checkShape(t){if(t.inboundNodes[0].outputTensors[0].shape.some(n=>n<0))throw new M(`Negative dimension size caused by adding layer ${t.name} with input shape [${t.inboundNodes[0].inputTensors[0].shape}]`)}add(t){let e=t instanceof qi||t instanceof Bn,n;if(e){if(n=t,n.outputs.length!==1)throw new M(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");if(n.inputs.length!==1)throw new M(\"All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.\")}if(this.outputs.length===0){if(t.inboundNodes.length===0){if(t.batchInputShape==null)throw new M(\"The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.\");let o=Dy({batchShape:t.batchInputShape,dtype:t.dtype,name:t.name+\"_input\"});t.apply(o)}if(e)this.outputs=n.outputs,this.inputs=n.inputs;else{if(t.inboundNodes.length!==1)throw new M(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${t.name} which has ${t.inboundNodes.length} pre-existing inbound connections.`);if(t.inboundNodes[0].outputTensors.length!==1)throw new M(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");this.checkShape(t),this.outputs=[t.inboundNodes[0].outputTensors[0]],this.inputs=kv(this.outputs[0])}this.inboundNodes=[],new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:Io(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(o=>o.shape),outputShapes:this.outputs[0].shape})}else{let o=t.apply(this.outputs[0]);if(Array.isArray(o))throw new TypeError(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");this.checkShape(t),this.outputs=[o],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(t),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError(\"There are no layers in the model.\");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let t=this.layers.length-1;this.layers[t].outboundNodes=[],this.outputs=[this.layers[t].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(t,e){return this.model==null&&this.build(),this.model.call(t,e)}build(t){if(Bt(t),this.inputs.length===0||this.outputs.length===0)throw new TypeError(\"Sequential model cannot be built: model is empty. Add some layers first.\");this.model=new Bn({inputs:this.inputs,outputs:this.outputs[0],name:this.name+\"_model\"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(t,e,n=console.log){this.built||this.build(),super.summary(t,e,n)}setWeights(t){this.model==null&&this.build(),this.model.setWeights(t)}evaluate(t,e,n={}){if(!this.built)throw new Hr(\"The model needs to be compiled before being used.\");return this.model.evaluate(t,e,n)}async evaluateDataset(t,e){if(!this.built)throw new Hr(\"The model needs to be compiled before being used.\");return this.model.evaluateDataset(t,e)}predict(t,e={}){return this.model==null&&this.build(),this.model.predict(t,e)}predictOnBatch(t){return this.model==null&&this.build(),this.model.predictOnBatch(t)}compile(t){this.build(),this.model.compile(t),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(t){this.model.optimizer=t}async fit(t,e,n={}){if(!this.built)throw new Hr(\"The model needs to be compiled before being used.\");return this.model.fit(t,e,n)}async fitDataset(t,e){if(!this.built)throw new Hr(\"The model needs to be compiled before being used.\");return this.model.fitDataset(t,e)}async trainOnBatch(t,e){return this.model.trainOnBatch(t,e)}static fromConfig(t,e,n={},o=!1){let s,i={};if(e instanceof Array){if(e[0].className==null||e[0].className===\"Merge\")throw new M(\"Legacy serialization format not supported yet.\");s=e}else y.assert(e.layers!=null,()=>\"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field.\"),s=e.layers,delete e.layers,i=e;let a=new t(i);if(!(a instanceof qi))throw new St(`Sequential.fromConfig called on non-Sequential input: ${a}`);for(let u of s){let c=gn(u,void 0,o);o&&c.setFastWeightInitDuringBuild(!0),a.add(c)}return a}set stopTraining(t){if(this.model==null)throw new M(\"Cannot set the stopTraining property of a sequential model before it is compiled.\");this.model.stopTraining=t}get stopTraining(){if(this.model==null)throw new M(\"Cannot get the stopTraining property of a sequential model before it is compiled.\");return this.model.stopTraining}getConfig(){let t=[];for(let e of this.layers){let n={};n.className=e.getClassName(),n.config=e.getConfig(),t.push(n)}return{name:this.name,layers:t}}};qi.className=\"Sequential\";Q.registerClass(qi);function q8(r){return new Bn(r)}function K8(r){return new qi(r)}function Pv(r){return Dy(r)}function j8(r,t){hn.registerCallbackConstructor(r,t)}var Qr=class extends Q.Serializable{getConfig(){return{}}},Qy=class extends Qr{apply(t,e=1){return F$(t,e)}};Qy.className=\"elu\";Q.registerClass(Qy);var tb=class extends Qr{apply(t){return pm(t)}};tb.className=\"selu\";Q.registerClass(tb);var eb=class extends Qr{apply(t){return Fr(t)}};eb.className=\"relu\";Q.registerClass(eb);var rb=class extends Qr{apply(t){return B(()=>Mi(6,Fr(t)))}};rb.className=\"relu6\";Q.registerClass(rb);var nb=class extends Qr{apply(t){return t}};nb.className=\"linear\";Q.registerClass(nb);var ob=class extends Qr{apply(t){return Yr(t)}};ob.className=\"sigmoid\";Q.registerClass(ob);var sb=class extends Qr{apply(t){return P$(t)}};sb.className=\"hardSigmoid\";Q.registerClass(sb);var ib=class extends Qr{apply(t){return zs(t)}};ib.className=\"softplus\";Q.registerClass(ib);var ab=class extends Qr{apply(t){return O$(t)}};ab.className=\"softsign\";Q.registerClass(ab);var lb=class extends Qr{apply(t){return $i(t)}};lb.className=\"tanh\";Q.registerClass(lb);var qm=class extends Qr{apply(t,e=-1){return iu(t,e)}};qm.className=\"softmax\";Q.registerClass(qm);var ub=class extends Qr{apply(t,e=-1){return sm(t,e)}};ub.className=\"logSoftmax\";Q.registerClass(ub);var cb=class extends Qr{apply(t,e=1){return B(()=>D(Yr(D(t,e)),t))}};cb.className=\"swish\";Q.registerClass(cb);var pb=class extends Qr{apply(t){return B(()=>D(t,$i(zs(t))))}};pb.className=\"mish\";Q.registerClass(pb);function js(r){return r.getClassName()}function Lv(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,\"activation\")}function Xs(r){if(r==null){let t={};return t.className=\"linear\",t.config={},Lv(t)}if(typeof r==\"string\"){let t={};return t.className=r,t.config={},Lv(t)}else return r instanceof Qr?r:Lv(r)}function Mv(r){if(r!=null&&typeof r!=\"object\")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${r}`)}var mb=class extends Q.Serializable{},wu=class extends mb{constructor(t){super(),Mv(t),this.l1=t==null||t.l1==null?.01:t.l1,this.l2=t==null||t.l2==null?.01:t.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(t){return B(()=>{let e=Ne([1]);return this.hasL1&&(e=X(e,ft(D(this.l1,Ee(t))))),this.hasL2&&(e=X(e,ft(D(this.l2,lc(t))))),R(e,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(t,e){return new t({l1:e.l1,l2:e.l2})}};wu.className=\"L1L2\";Q.registerClass(wu);function yD(r){return Mv(r),new wu({l1:r!=null?r.l1:null,l2:0})}function bD(r){return Mv(r),new wu({l2:r!=null?r.l2:null,l1:0})}var gD={l1l2:\"L1L2\"};function me(r){return Sm(r)}function xD(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,\"regularizer\")}function be(r){if(r==null)return null;if(typeof r==\"string\"){let e={className:r in gD?gD[r]:r,config:{}};return xD(e)}else return r instanceof mb?r:xD(r)}var Km=class extends $t{constructor(t){super(t==null?{}:t),this.supportsMasking=!0,t!=null&&(this.maxValue=t.maxValue)}call(t,e){t=Nt(t);let n=Fr(t);return this.maxValue!=null&&(n=Cr(n,0,this.maxValue)),n}computeOutputShape(t){return t}getConfig(){let t={maxValue:this.maxValue},e=super.getConfig();return Object.assign(t,e),t}};Km.className=\"ReLU\";Q.registerClass(Km);var jm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_ALPHA=.3,t==null&&(t={}),this.alpha=t.alpha==null?this.DEFAULT_ALPHA:t.alpha}call(t,e){let n=Nt(t);return Ql(n,this.alpha)}computeOutputShape(t){return t}getConfig(){let t={alpha:this.alpha},e=super.getConfig();return Object.assign(t,e),t}};jm.className=\"LeakyReLU\";Q.registerClass(jm);var Xm=class extends $t{constructor(t){if(super(t==null?{}:t),this.DEFAULT_ALPHA_INITIALIZER=\"zeros\",t==null&&(t={}),this.supportsMasking=!0,this.alphaInitializer=de(t.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=be(t.alphaRegularizer),this.alphaConstraint=Be(t.alphaConstraint),t.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(t.sharedAxes))this.sharedAxes=t.sharedAxes;else if(typeof t.sharedAxes==\"number\")this.sharedAxes=[t.sharedAxes];else throw new M(`Expected sharedAxes to be a number or an array of numbers, but got ${t.sharedAxes}`)}build(t){t=Bt(t);let e=t.slice(1);if(this.sharedAxes!=null)for(let o of this.sharedAxes)e[o-1]=1;this.alpha=this.addWeight(\"alpha\",e,\"float32\",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(this.sharedAxes!=null)for(let o=1;o<t.length;++o)n[o]=t[o];this.inputSpec=[new ye({ndim:t.length,axes:n})],this.built=!0}call(t,e){return t=Nt(t),ou(t,this.alpha.read())}getConfig(){let t={alphaInitializer:Te(this.alphaInitializer),alphaRegularizer:me(this.alphaRegularizer),alphaConstraint:ze(this.alphaConstraint),sharedAxes:this.sharedAxes},e=super.getConfig();return Object.assign(t,e),t}};Xm.className=\"PReLU\";Q.registerClass(Xm);var Ym=class extends $t{constructor(t){if(super(t==null?{}:t),this.DEFAULT_ALPHA=1,t==null&&(t={}),t.alpha!=null&&t.alpha!==this.DEFAULT_ALPHA)throw new St(`Non-default alpha value (${t.alpha}) is not supported by the ELU layer yet.`);this.alpha=t.alpha==null?this.DEFAULT_ALPHA:t.alpha}call(t,e){let n=Nt(t);return Oi(n)}computeOutputShape(t){return t}getConfig(){let t={alpha:this.alpha},e=super.getConfig();return Object.assign(t,e),t}};Ym.className=\"ELU\";Q.registerClass(Ym);var Zm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_THETA=1,t==null&&(t={}),this.theta=t.theta==null?this.DEFAULT_THETA:t.theta}call(t,e){let n=Nt(t);return D(n,J(Re(n,this.theta),\"float32\"))}computeOutputShape(t){return t}getConfig(){let t={theta:this.theta},e=super.getConfig();return Object.assign(t,e),t}};Zm.className=\"ThresholdedReLU\";Q.registerClass(Zm);var Jm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_AXIS=1,t==null&&(t={}),this.softmax=new qm().apply,this.axis=t.axis==null?this.DEFAULT_AXIS:t.axis}call(t,e){let n=Nt(t);return this.softmax(n,this.axis)}computeOutputShape(t){return t}getConfig(){let t={axis:this.axis},e=super.getConfig();return Object.assign(t,e),t}};Jm.className=\"Softmax\";Q.registerClass(Jm);function Cu(r,t,e){if(typeof r==\"number\")return Io(r,t);if(r.length!==t)throw new M(`The ${e} argument must be an integer or tuple of ${t} integers. Received: ${r.length} elements.`);for(let n=0;n<t;++n){let o=r[n];if(!A$(o))throw new M(`The ${e} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(r)} including a non-integer number ${o}`)}return r}function Nn(r,t,e,n,o=1){if(r==null)return r;let s=t+(t-1)*(o-1),i;return e===\"same\"?i=r:i=r-s+1,Math.floor((i+n-1)/n)}function Ys(r,t,e,n){if(r==null)return null;if(n===\"valid\")r=r*t+qs([e-t,0]);else if(n===\"same\")r=r*t;else throw new M(`Unsupport padding mode: ${n}.`);return r}function Ah(r,t){return B(()=>(Fe(t),t===\"channelsFirst\"?Ot(r,[0,2,3,1]):r))}function zv(r,t){return B(()=>(Fe(t),t===\"channelsFirst\"?Ot(r,[0,2,3,4,1]):r))}function Y8(r,t,e,n=1,o=\"valid\",s,i=1){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.shape.length!==3)throw new M(`The input of a conv1dWithBias operation should be 3, but is ${r.shape.length} instead.`);if(t.shape.length!==3)throw new M(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(e!=null&&e.shape.length!==1)throw new M(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(s===\"channelsFirst\"&&(r=Ot(r,[0,2,1])),o===\"causal\")throw new St(\"The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.\");let a=Qp(r,t,n,o===\"same\"?\"same\":\"valid\",\"NWC\",i);return e!=null&&(a=fn(a,e)),a})}function wD(r,t,e,n=[1,1],o=\"valid\",s,i,a=null){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.rank!==3&&r.rank!==4)throw new M(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${r.rank}.`);if(t.rank!==3&&t.rank!==4)throw new M(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${r.rank}.`);let u=Ah(r,s);if(o===\"causal\")throw new St(\"The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.\");return u=uu.conv2d({x:u,filter:t,strides:n,pad:o===\"same\"?\"same\":\"valid\",dilations:i,dataFormat:\"NHWC\",bias:e,activation:a}),s===\"channelsFirst\"&&(u=Ot(u,[0,3,1,2])),u})}function Z8(r,t,e,n=[1,1,1],o=\"valid\",s,i){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.rank!==4&&r.rank!==5)throw new M(`conv3dWithBias expects input to be of rank 4 or 5, but received ${r.rank}.`);if(t.rank!==4&&t.rank!==5)throw new M(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${r.rank}.`);let a=zv(r,s);if(o===\"causal\")throw new St(\"The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.\");return a=Tx(a,t,n,o===\"same\"?\"same\":\"valid\",\"NDHWC\",i),e!=null&&(a=fn(a,e)),s===\"channelsFirst\"&&(a=Ot(a,[0,4,1,2,3])),a})}var bc=class extends $t{constructor(t,e){if(super(e),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER=\"glorotNormal\",this.DEFAULT_BIAS_INITIALIZER=\"zeros\",bc.verifyArgs(e),this.rank=t,Ze(this.rank,\"rank\"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new St(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=Cu(e.kernelSize,t,\"kernelSize\"),this.strides=Cu(e.strides==null?1:e.strides,t,\"strides\"),this.padding=e.padding==null?\"valid\":e.padding,pn(this.padding),this.dataFormat=e.dataFormat==null?\"channelsLast\":e.dataFormat,Fe(this.dataFormat),this.activation=Xs(e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.biasInitializer=de(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Be(e.biasConstraint),this.biasRegularizer=be(e.biasRegularizer),this.activityRegularizer=be(e.activityRegularizer),this.dilationRate=Cu(e.dilationRate==null?1:e.dilationRate,t,\"dilationRate\"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new M(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate==\"number\")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new M(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate==\"number\")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new M(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(t){if(ro(\"kernelSize\"in t,\"required key 'kernelSize' not in config\"),typeof t.kernelSize!=\"number\"&&!Cy(t.kernelSize,\"number\",1,3))throw new M(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(t.kernelSize)}.`)}getConfig(){let t={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:js(this.activation),useBias:this.useBias,biasInitializer:Te(this.biasInitializer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),biasConstraint:ze(this.biasConstraint)},e=super.getConfig();return Object.assign(t,e),t}},Iu=class extends bc{constructor(t,e){super(t,e),this.kernel=null,Iu.verifyArgs(e),this.filters=e.filters,Ze(this.filters,\"filters\"),this.kernelInitializer=de(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Be(e.kernelConstraint),this.kernelRegularizer=be(e.kernelRegularizer)}build(t){t=Bt(t);let e=this.dataFormat===\"channelsFirst\"?1:t.length-1;if(t[e]==null)throw new M(`The channel dimension of the input should be defined. Found ${t[e]}`);let n=t[e],o=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight(\"kernel\",o,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight(\"bias\",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[e]:n}}],this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n,o=this.bias==null?null:this.bias.read(),s=Iy(this.activation.getClassName());if(s!=null&&this.rank===2)n=wD(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate,s);else{if(this.rank===1)n=Y8(t,this.kernel.read(),o,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)n=wD(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)n=Z8(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new St(\"convolutions greater than 3D are not implemented yet.\");this.activation!=null&&(n=this.activation.apply(n))}return n})}computeOutputShape(t){t=Bt(t);let e=[],n=this.dataFormat===\"channelsLast\"?t.slice(1,t.length-1):t.slice(2);for(let s=0;s<n.length;++s){let i=Nn(n[s],this.kernelSize[s],this.padding,this.strides[s],typeof this.dilationRate==\"number\"?this.dilationRate:this.dilationRate[s]);e.push(i)}let o=[t[0]];return this.dataFormat===\"channelsLast\"?(o=o.concat(e),o.push(this.filters)):(o.push(this.filters),o=o.concat(e)),o}getConfig(){let t={filters:this.filters,kernelInitializer:Te(this.kernelInitializer),kernelRegularizer:me(this.kernelRegularizer),kernelConstraint:ze(this.kernelConstraint)},e=super.getConfig();return Object.assign(t,e),t}static verifyArgs(t){if(!(\"filters\"in t)||typeof t.filters!=\"number\"||t.filters<1)throw new M(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(t.filters)}`)}},il=class extends Iu{constructor(t){super(2,t),il.verifyArgs(t)}getConfig(){let t=super.getConfig();return delete t.rank,t}static verifyArgs(t){if(typeof t.kernelSize!=\"number\"&&!Cy(t.kernelSize,\"number\",1,2))throw new M(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(t.kernelSize)}.`)}};il.className=\"Conv2D\";Q.registerClass(il);var al=class extends Iu{constructor(t){super(3,t),al.verifyArgs(t)}getConfig(){let t=super.getConfig();return delete t.rank,t}static verifyArgs(t){if(typeof t.kernelSize!=\"number\"&&!(Array.isArray(t.kernelSize)&&(t.kernelSize.length===1||t.kernelSize.length===3)))throw new M(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(t.kernelSize)}.`)}};al.className=\"Conv3D\";Q.registerClass(al);var Qm=class extends il{constructor(t){if(super(t),this.inputSpec=[new ye({ndim:4})],this.padding!==\"same\"&&this.padding!==\"valid\")throw new M(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(t){if(t=Bt(t),t.length!==4)throw new M(\"Input should have rank 4; Received input shape: \"+JSON.stringify(t));let e=this.dataFormat===\"channelsFirst\"?1:t.length-1;if(t[e]==null)throw new M(\"The channel dimension of the inputs should be defined. Found `None`.\");let n=t[e],o=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight(\"kernel\",o,\"float32\",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight(\"bias\",[this.filters],\"float32\",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new ye({ndim:4,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{let n=Nt(t);if(n.shape.length!==4)throw new M(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let o=n.shape,s=o[0],i,a;this.dataFormat===\"channelsFirst\"?(i=2,a=3):(i=1,a=2);let u=o[i],l=o[a],c=this.kernelSize[0],p=this.kernelSize[1],m=this.strides[0],f=this.strides[1],d=Ys(u,m,c,this.padding),h=Ys(l,f,p,this.padding),g=[s,d,h,this.filters];this.dataFormat!==\"channelsLast\"&&(n=Ot(n,[0,2,3,1]));let x=em(n,this.kernel.read(),g,this.strides,this.padding);return this.dataFormat!==\"channelsLast\"&&(x=Ot(x,[0,3,1,2])),this.bias!=null&&(x=fn(x,this.bias.read(),this.dataFormat)),this.activation!=null&&(x=this.activation.apply(x)),x})}computeOutputShape(t){t=Bt(t);let e=t.slice(),n,o,s;this.dataFormat===\"channelsFirst\"?(n=1,o=2,s=3):(n=3,o=1,s=2);let i=this.kernelSize[0],a=this.kernelSize[1],u=this.strides[0],l=this.strides[1];return e[n]=this.filters,e[o]=Ys(e[o],u,i,this.padding),e[s]=Ys(e[s],l,a,this.padding),e}getConfig(){let t=super.getConfig();return delete t.dilationRate,t}};Qm.className=\"Conv2DTranspose\";Q.registerClass(Qm);var tf=class extends al{constructor(t){if(super(t),this.inputSpec=[new ye({ndim:5})],this.padding!==\"same\"&&this.padding!==\"valid\")throw new M(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(t){if(t=Bt(t),t.length!==5)throw new M(\"Input should have rank 5; Received input shape: \"+JSON.stringify(t));let e=this.dataFormat===\"channelsFirst\"?1:t.length-1;if(t[e]==null)throw new M(\"The channel dimension of the inputs should be defined. Found `None`.\");let n=t[e],o=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight(\"kernel\",o,\"float32\",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight(\"bias\",[this.filters],\"float32\",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new ye({ndim:5,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{let n=Nt(t);if(n.shape.length!==5)throw new M(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let o=n.shape,s=o[0],i,a,u;this.dataFormat===\"channelsFirst\"?(u=2,i=3,a=4):(u=1,i=2,a=3);let l=o[u],c=o[i],p=o[a],m=this.kernelSize[0],f=this.kernelSize[1],d=this.kernelSize[2],h=this.strides[0],g=this.strides[1],x=this.strides[2],b=Ys(l,h,m,this.padding),w=Ys(c,g,f,this.padding),C=Ys(p,x,d,this.padding),N=[s,b,w,C,this.filters];this.dataFormat!==\"channelsLast\"&&(n=Ot(n,[0,2,3,4,1]));let _=Ex(n,this.kernel.read(),N,this.strides,this.padding);return this.dataFormat!==\"channelsLast\"&&(_=Ot(_,[0,4,1,2,3])),this.bias!==null&&(_=fn(_,this.bias.read(),this.dataFormat)),this.activation!==null&&(_=this.activation.apply(_)),_})}computeOutputShape(t){t=Bt(t);let e=t.slice(),n,o,s,i;this.dataFormat===\"channelsFirst\"?(n=1,o=2,s=3,i=4):(n=4,o=1,s=2,i=3);let a=this.kernelSize[0],u=this.kernelSize[1],l=this.kernelSize[2],c=this.strides[0],p=this.strides[1],m=this.strides[2];return e[n]=this.filters,e[o]=Ys(e[o],c,a,this.padding),e[s]=Ys(e[s],p,u,this.padding),e[i]=Ys(e[i],m,l,this.padding),e}getConfig(){let t=super.getConfig();return delete t.dilationRate,t}};tf.className=\"Conv3DTranspose\";Q.registerClass(tf);var fb=class extends Iu{constructor(t,e){if(super(t,e),this.DEFAULT_DEPTHWISE_INITIALIZER=\"glorotUniform\",this.DEFAULT_POINTWISE_INITIALIZER=\"glorotUniform\",this.depthwiseKernel=null,this.pointwiseKernel=null,e.filters==null)throw new M(\"The `filters` configuration field is required by SeparableConv, but is unspecified.\");if(e.kernelInitializer!=null||e.kernelRegularizer!=null||e.kernelConstraint!=null)throw new M(\"Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.\");if(e.padding!=null&&e.padding!==\"same\"&&e.padding!==\"valid\")throw new M(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(e.padding)}`);this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=de(e.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=be(e.depthwiseRegularizer),this.depthwiseConstraint=Be(e.depthwiseConstraint),this.pointwiseInitializer=de(e.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=be(e.pointwiseRegularizer),this.pointwiseConstraint=Be(e.pointwiseConstraint)}build(t){if(t=Bt(t),t.length<this.rank+2)throw new M(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(t)}`);let e=this.dataFormat===\"channelsFirst\"?1:t.length-1;if(t[e]==null||t[e]<0)throw new M(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(t[e])}`);let n=t[e],o=this.kernelSize.concat([n,this.depthMultiplier]),s=[];for(let a=0;a<this.rank;++a)s.push(1);s.push(n*this.depthMultiplier,this.filters);let i=!0;this.depthwiseKernel=this.addWeight(\"depthwise_kernel\",o,\"float32\",this.depthwiseInitializer,this.depthwiseRegularizer,i,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight(\"pointwise_kernel\",s,\"float32\",this.pointwiseInitializer,this.pointwiseRegularizer,i,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight(\"bias\",[this.filters],\"float32\",this.biasInitializer,this.biasRegularizer,i,this.biasConstraint):this.bias=null,this.inputSpec=[new ye({ndim:this.rank+2,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n;if(this.rank===1)throw new St(\"1D separable convolution is not implemented yet.\");return this.rank===2&&(this.dataFormat===\"channelsFirst\"&&(t=Ot(t,[0,2,3,1])),n=mm(t,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,\"NHWC\")),this.useBias&&(n=fn(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),this.dataFormat===\"channelsFirst\"&&(n=Ot(n,[0,3,1,2])),n})}getConfig(){let t=super.getConfig();return delete t.rank,delete t.kernelInitializer,delete t.kernelRegularizer,delete t.kernelConstraint,t.depthwiseInitializer=Te(this.depthwiseInitializer),t.pointwiseInitializer=Te(this.pointwiseInitializer),t.depthwiseRegularizer=me(this.depthwiseRegularizer),t.pointwiseRegularizer=me(this.pointwiseRegularizer),t.depthwiseConstraint=ze(this.depthwiseConstraint),t.pointwiseConstraint=ze(this.pointwiseConstraint),t}};fb.className=\"SeparableConv\";var ef=class extends fb{constructor(t){super(2,t)}};ef.className=\"SeparableConv2D\";Q.registerClass(ef);var Su=class extends Iu{constructor(t){super(1,t),Su.verifyArgs(t),this.inputSpec=[{ndim:3}]}getConfig(){let t=super.getConfig();return delete t.rank,delete t.dataFormat,t}static verifyArgs(t){if(typeof t.kernelSize!=\"number\"&&!Cy(t.kernelSize,\"number\",1,1))throw new M(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(t.kernelSize)}.`)}};Su.className=\"Conv1D\";Q.registerClass(Su);var rf=class extends $t{constructor(t){super(t),typeof t.cropping==\"number\"?this.cropping=[[t.cropping,t.cropping],[t.cropping,t.cropping]]:typeof t.cropping[0]==\"number\"?this.cropping=[[t.cropping[0],t.cropping[0]],[t.cropping[1],t.cropping[1]]]:this.cropping=t.cropping,this.dataFormat=t.dataFormat===void 0?\"channelsLast\":t.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(t){return this.dataFormat===\"channelsFirst\"?[t[0],t[1],t[2]-this.cropping[0][0]-this.cropping[0][1],t[3]-this.cropping[1][0]-this.cropping[1][1]]:[t[0],t[1]-this.cropping[0][0]-this.cropping[0][1],t[2]-this.cropping[1][0]-this.cropping[1][1],t[3]]}call(t,e){return B(()=>{if(t=Nt(t),this.dataFormat===\"channelsLast\"){let n=wh(t,this.cropping[0][0],t.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return wh(n,this.cropping[1][0],t.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let n=wh(t,this.cropping[0][0],t.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return wh(n,this.cropping[1][0],t.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let t={cropping:this.cropping,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}};rf.className=\"Cropping2D\";Q.registerClass(rf);var nf=class extends $t{constructor(t){super(t),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=t.size==null?this.DEFAULT_SIZE:t.size,this.dataFormat=t.dataFormat==null?\"channelsLast\":t.dataFormat,Fe(this.dataFormat),this.interpolation=t.interpolation==null?\"nearest\":t.interpolation,E$(this.interpolation)}computeOutputShape(t){if(this.dataFormat===\"channelsFirst\"){let e=t[2]==null?null:this.size[0]*t[2],n=t[3]==null?null:this.size[1]*t[3];return[t[0],t[1],e,n]}else{let e=t[1]==null?null:this.size[0]*t[1],n=t[2]==null?null:this.size[1]*t[2];return[t[0],e,n,t[3]]}}call(t,e){return B(()=>{let n=Nt(t),o=n.shape;if(this.dataFormat===\"channelsFirst\"){n=Ot(n,[0,2,3,1]);let s=this.size[0]*o[2],i=this.size[1]*o[3],a=this.interpolation===\"nearest\"?Gs.resizeNearestNeighbor(n,[s,i]):Gs.resizeBilinear(n,[s,i]);return Ot(a,[0,3,1,2])}else{let s=this.size[0]*o[1],i=this.size[1]*o[2];return this.interpolation===\"nearest\"?Gs.resizeNearestNeighbor(n,[s,i]):Gs.resizeBilinear(n,[s,i])}})}getConfig(){let t={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},e=super.getConfig();return Object.assign(t,e),t}};nf.className=\"UpSampling2D\";Q.registerClass(nf);function J8(r,t,e=[1,1],n=\"valid\",o,s){return B(()=>{o==null&&(o=mn()),Fe(o);let i=Ah(r,o);if(r.rank!==4)throw new M(`Input for depthwiseConv2d is required to be 4-D, but is instead ${r.rank}-D`);if(t.rank!==4)throw new M(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=Fi(i,t,e,n===\"same\"?\"same\":\"valid\",\"NHWC\",s),o===\"channelsFirst\"&&(i=Ot(i,[0,3,1,2])),i})}var of=class extends bc{constructor(t){super(2,t),this.depthwiseKernel=null,this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=de(t.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Be(t.depthwiseConstraint),this.depthwiseRegularizer=be(t.depthwiseRegularizer)}build(t){if(t=Bt(t),t.length<4)throw new M(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(t)}.`);let e=this.dataFormat===\"channelsFirst\"?1:3;if(t[e]==null||t[e]<0)throw new M(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${t[e]}).`);let n=t[e],o=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight(\"depthwise_kernel\",o,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight(\"bias\",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n=J8(t,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(n=fn(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),n})}computeOutputShape(t){t=Bt(t);let e=this.dataFormat===\"channelsFirst\"?t[2]:t[1],n=this.dataFormat===\"channelsFirst\"?t[3]:t[2],o=this.dataFormat===\"channelsFirst\"?t[1]*this.depthMultiplier:t[3]*this.depthMultiplier,s=Nn(e,this.kernelSize[0],this.padding,this.strides[0]),i=Nn(n,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat===\"channelsFirst\"?[t[0],o,s,i]:[t[0],s,i,o]}getConfig(){let t=super.getConfig();return t.depthMultiplier=this.depthMultiplier,t.depthwiseInitializer=Te(this.depthwiseInitializer),t.depthwiseRegularizer=me(this.depthwiseRegularizer),t.depthwiseConstraint=ze(this.depthwiseRegularizer),t}};of.className=\"DepthwiseConv2D\";Q.registerClass(of);function Bv(r,t,e,n){if(Array.isArray(r)){if(t!=null||e!=null)throw new M(\"When inputs is an array, neither initialState or constants should be provided\");n!=null&&(e=r.slice(r.length-n,r.length),r=r.slice(0,r.length-n)),r.length>1&&(t=r.slice(1,r.length)),r=r[0]}function o(s){return s==null||Array.isArray(s)?s:[s]}return t=o(t),e=o(e),{inputs:r,initialState:t,constants:e}}function Vv(r,t,e,n=!1,o,s,i=!1,a=!1){return B(()=>{let u=t.shape.length;if(u<3)throw new M(`Input should be at least 3D, but is ${u}D.`);let l=[1,0].concat(Zr(2,u));if(t=Ot(t,l),s!=null)throw new St(\"The rnn() functoin of the deeplearn.js backend does not support constants yet.\");i&&console.warn(\"Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend.\"),o!=null&&(o=J(J(o,\"bool\"),\"float32\"),o.rank===u-1&&(o=rr(o,-1)),o=Ot(o,l)),n&&(t=pr(t,0),o!=null&&(o=pr(o,0)));let c=[],p,m=e,f=t.shape[0],d=vr(t),h;o!=null&&(h=vr(o));for(let x=0;x<f;++x){let b=d[x],w=B(()=>r(b,m));if(o==null)p=w[0],m=w[1];else{let C=B(()=>{let N=h[x],_=ct(yr(N),N),A=X(D(w[0],N),D(m[0],_)),$=m.map((F,P)=>X(D(w[1][P],N),D(F,_)));return{output:A,newStates:$}});p=C.output,m=C.newStates}a&&c.push(p)}let g;return a&&(g=nr(c,1)),[p,g,m]})}var Tn=class extends $t{constructor(t){super(t);let e;if(t.cell==null)throw new M(\"cell property is missing for the constructor of RNN.\");if(Array.isArray(t.cell)?e=new Ic({cells:t.cell}):e=t.cell,e.stateSize==null)throw new M(\"The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).\");this.cell=e,this.returnSequences=t.returnSequences==null?!1:t.returnSequences,this.returnState=t.returnState==null?!1:t.returnState,this.goBackwards=t.goBackwards==null?!1:t.goBackwards,this._stateful=t.stateful==null?!1:t.stateful,this.unroll=t.unroll==null?!1:t.unroll,this.supportsMasking=!0,this.inputSpec=[new ye({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let t=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return Zr(0,t).map(e=>null)}else return this.states_}setStates(t){this.states_=t}computeOutputShape(t){$y(t)&&(t=t[0]),t=t;let e=this.cell.stateSize;Array.isArray(e)||(e=[e]);let n=e[0],o;if(this.returnSequences?o=[t[0],t[1],n]:o=[t[0],n],this.returnState){let s=[];for(let i of e)s.push([t[0],i]);return[o].concat(s)}else return o}computeMask(t,e){return B(()=>{Array.isArray(e)&&(e=e[0]);let n=this.returnSequences?e:null;if(this.returnState){let o=this.states.map(s=>null);return[n].concat(o)}else return n})}get states(){if(this.states_==null){let t=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,e=[];for(let n=0;n<t;++n)e.push(null);return e}else return this.states_}set states(t){this.states_=t}build(t){if(this.numConstants!=null)throw new St(\"Constants support is not implemented in RNN yet.\");$y(t)&&(t=t[0]),t=t;let n=this.stateful?t[0]:null,o=t.slice(2);this.inputSpec[0]=new ye({shape:[n,null,...o]});let s=[t[0]].concat(t.slice(2));this.cell.build(s);let i;if(Array.isArray(this.cell.stateSize)?i=this.cell.stateSize:i=[this.cell.stateSize],this.stateSpec!=null){if(!y.arraysEqual(this.stateSpec.map(a=>a.shape[a.shape.length-1]),i))throw new M(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=i.map(a=>new ye({shape:[null,a]}));this.stateful&&this.resetStates()}resetStates(t,e=!1){B(()=>{if(!this.stateful)throw new vn(\"Cannot call resetStates() on an RNN Layer that is not stateful.\");let n=this.inputSpec[0].shape[0];if(n==null)throw new M(\"If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \\n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.\");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(o=>Ne([n,o])):this.states_=[Ne([n,this.cell.stateSize])];else if(t==null)vt(this.states_),this.keptStates!=null&&(vt(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(o=>Ne([n,o])):this.states_[0]=Ne([n,this.cell.stateSize]);else{if(Array.isArray(t)||(t=[t]),t.length!==this.states_.length)throw new M(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${t.length} state value(s). Input received: ${t}`);e===!0?this.keptStates.push(this.states_.slice()):vt(this.states_);for(let o=0;o<this.states_.length;++o){let s=t[o],i=Array.isArray(this.cell.stateSize)?this.cell.stateSize[o]:this.cell.stateSize,a=[n,i];if(!y.arraysEqual(s.shape,a))throw new M(`State ${o} is incompatible with layer ${this.name}: expected shape=${a}, received shape=${s.shape}`);this.states_[o]=s}}this.states_=this.states_.map(o=>De(o.clone()))})}apply(t,e){let n=e==null?null:e.initialState,o=e==null?null:e.constants;e==null&&(e={});let s=Bv(t,n,o,this.numConstants);t=s.inputs,n=s.initialState,o=s.constants;let i=[],a=[];if(n!=null){e.initialState=n,i=i.concat(n),this.stateSpec=[];for(let l of n)this.stateSpec.push(new ye({shape:l.shape}));a=a.concat(this.stateSpec)}if(o!=null&&(e.constants=o,i=i.concat(o),this.numConstants=o.length),i[0]instanceof Jr){let l=[t].concat(i),c=this.inputSpec.concat(a),p=this.inputSpec;this.inputSpec=c;let m=super.apply(l,e);return this.inputSpec=p,m}else return super.apply(t,e)}call(t,e){return B(()=>{let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;t=Nt(t),s==null&&(this.stateful?s=this.states_:s=this.getInitialState(t));let i=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(s.length!==i)throw new M(`RNN Layer has ${i} state(s) but was passed ${s.length} initial state(s).`);this.unroll&&console.warn(\"Ignoring unroll = true for RNN layer, due to imperative backend.\");let a={training:o},l=Vv((d,h)=>{let g=this.cell.call([d].concat(h),a);return[g[0],g.slice(1)]},t,s,this.goBackwards,n,null,this.unroll,this.returnSequences),c=l[0],p=l[1],m=l[2];this.stateful&&this.resetStates(m,o);let f=this.returnSequences?p:c;return this.returnState?[f].concat(m):f})}getInitialState(t){return B(()=>{let e=Ne(t.shape);return e=ft(e,[1,2]),e=nl(e),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(n=>n>1?Ey(e,[1,n]):e):this.cell.stateSize>1?[Ey(e,[1,this.cell.stateSize])]:[e]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(t)}getConfig(){let t=super.getConfig(),e={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(e.numConstants=this.numConstants);let n=this.cell.getConfig();return this.getClassName()===Tn.className&&(e.cell={className:this.cell.getClassName(),config:n}),Object.assign(Object.assign(Object.assign({},n),t),e)}static fromConfig(t,e,n={}){let o=e.cell,s=gn(o,n);return new t(Object.assign(e,{cell:s}))}};Tn.className=\"RNN\";Q.registerClass(Tn);var ll=class extends $t{},wc=class extends ll{constructor(t){super(t),this.DEFAULT_ACTIVATION=\"tanh\",this.DEFAULT_KERNEL_INITIALIZER=\"glorotNormal\",this.DEFAULT_RECURRENT_INITIALIZER=\"orthogonal\",this.DEFAULT_BIAS_INITIALIZER=\"zeros\",this.units=t.units,Ze(this.units,\"units\"),this.activation=Xs(t.activation==null?this.DEFAULT_ACTIVATION:t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){t=Bt(t),this.kernel=this.addWeight(\"kernel\",[t[t.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight(\"recurrent_kernel\",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight(\"bias\",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{if(t=t,t.length!==2)throw new M(`SimpleRNNCell expects 2 input Tensors, got ${t.length}.`);let n=t[1];t=t[0];let o=e.training==null?!1:e.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:o,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(n),rate:this.recurrentDropout,training:o,dropoutFunc:this.dropoutFunc}));let s,i=this.dropoutMask,a=this.recurrentDropoutMask;i!=null?s=To(D(t,i),this.kernel.read()):s=To(t,this.kernel.read()),this.bias!=null&&(s=fn(s,this.bias.read())),a!=null&&(n=D(n,a));let u=X(s,To(n,this.recurrentKernel.read()));return this.activation!=null&&(u=this.activation.apply(u)),[u,u]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},t),e)}};wc.className=\"SimpleRNNCell\";Q.registerClass(wc);var sf=class extends Tn{constructor(t){t.cell=new wc(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return new t(e)}};sf.className=\"SimpleRNN\";Q.registerClass(sf);var Cc=class extends ll{constructor(t){if(super(t),this.DEFAULT_ACTIVATION=\"tanh\",this.DEFAULT_RECURRENT_ACTIVATION=\"hardSigmoid\",this.DEFAULT_KERNEL_INITIALIZER=\"glorotNormal\",this.DEFAULT_RECURRENT_INITIALIZER=\"orthogonal\",this.DEFAULT_BIAS_INITIALIZER=\"zeros\",t.resetAfter)throw new M(\"GRUCell does not support reset_after parameter set to true.\");this.units=t.units,Ze(this.units,\"units\"),this.activation=Xs(t.activation===void 0?this.DEFAULT_ACTIVATION:t.activation),this.recurrentActivation=Xs(t.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:t.recurrentActivation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.implementation=t.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){t=Bt(t);let e=t[t.length-1];this.kernel=this.addWeight(\"kernel\",[e,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight(\"recurrent_kernel\",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight(\"bias\",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{if(t=t,t.length!==2)throw new M(`GRUCell expects 2 input Tensors (inputs, h, c), got ${t.length}.`);let n=e.training==null?!1:e.training,o=t[1];t=t[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(o),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,i=this.recurrentDropoutMask,a,u,l;0<this.dropout&&this.dropout<1&&(t=D(t,s[0]));let c=To(t,this.kernel.read());this.useBias&&(c=fn(c,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(o=D(o,i[0]));let p=this.recurrentKernel.read(),[m,f]=mr(p,[2*this.units,this.units],p.rank-1),d=To(o,m),[h,g,x]=mr(c,3,c.rank-1),[b,w]=mr(d,2,d.rank-1);a=this.recurrentActivation.apply(X(h,b)),u=this.recurrentActivation.apply(X(g,w));let C=To(D(u,o),f);l=this.activation.apply(X(x,C));let N=X(D(a,o),D(X(1,Ht(a)),l));return[N,N]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),recurrentActivation:js(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},t),e)}};Cc.className=\"GRUCell\";Q.registerClass(Cc);var af=class extends Tn{constructor(t){t.implementation===0&&console.warn(\"`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.\"),t.cell=new Cc(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return e.implmentation===0&&(e.implementation=1),new t(e)}};af.className=\"GRU\";Q.registerClass(af);var ul=class extends ll{constructor(t){super(t),this.DEFAULT_ACTIVATION=\"tanh\",this.DEFAULT_RECURRENT_ACTIVATION=\"hardSigmoid\",this.DEFAULT_KERNEL_INITIALIZER=\"glorotNormal\",this.DEFAULT_RECURRENT_INITIALIZER=\"orthogonal\",this.DEFAULT_BIAS_INITIALIZER=\"zeros\",this.units=t.units,Ze(this.units,\"units\"),this.activation=Xs(t.activation===void 0?this.DEFAULT_ACTIVATION:t.activation),this.recurrentActivation=Xs(t.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:t.recurrentActivation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=t.unitForgetBias,this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.implementation=t.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){var e;t=Bt(t);let n=t[t.length-1];this.kernel=this.addWeight(\"kernel\",[n,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight(\"recurrent_kernel\",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let o;if(this.useBias){if(this.unitForgetBias){let s=this.biasInitializer,i=this.units;o=new(e=class extends dn{apply(u,l){let c=s.apply([i]),p=new yu().apply([i]),m=s.apply([i*2]);return Tv(Tv(c,p),m)}},e.className=\"CustomInit\",e)}else o=this.biasInitializer;this.bias=this.addWeight(\"bias\",[this.units*4],null,o,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(t,e){return B(()=>{let n=e.training==null?!1:e.training;if(t=t,t.length!==3)throw new M(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${t.length}.`);let o=t[1],s=t[2];t=t[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(o),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let i=this.dropoutMask,a=this.recurrentDropoutMask,u,l,c,p;0<this.dropout&&this.dropout<1&&(t=D(t,i[0]));let m=To(t,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(o=D(o,a[0])),m=X(m,To(o,this.recurrentKernel.read())),this.useBias&&(m=fn(m,this.bias.read()));let[f,d,h,g]=mr(m,4,m.rank-1);u=this.recurrentActivation.apply(f),l=this.recurrentActivation.apply(d),c=X(D(l,s),D(u,this.activation.apply(h))),p=this.recurrentActivation.apply(g);let x=D(p,this.activation.apply(c));return[x,x,c]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),recurrentActivation:js(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},t),e)}};ul.className=\"LSTMCell\";Q.registerClass(ul);var lf=class extends Tn{constructor(t){t.implementation===0&&console.warn(\"`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.\"),t.cell=new ul(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return e.implmentation===0&&(e.implementation=1),new t(e)}};lf.className=\"LSTM\";Q.registerClass(lf);var Ic=class extends ll{constructor(t){super(t),this.cells=t.cells}get stateSize(){let t=[];for(let e of this.cells.slice().reverse())Array.isArray(e.stateSize)?t.push(...e.stateSize):t.push(e.stateSize);return t}call(t,e){return B(()=>{t=t;let n=t.slice(1),o=[];for(let a of this.cells.slice().reverse())Array.isArray(a.stateSize)?o.push(n.splice(0,a.stateSize.length)):o.push(n.splice(0,1));o.reverse();let s=[],i;for(let a=0;a<this.cells.length;++a){let u=this.cells[a];n=o[a],a===0?i=[t[0]].concat(n):i=[i[0]].concat(n),i=u.call(i,e),s.push(i.slice(1))}n=[];for(let a of s.slice().reverse())n.push(...a);return[i[0]].concat(n)})}build(t){$y(t)&&(t=t[0]),t=t;let e;this.cells.forEach((n,o)=>{Hs(`RNNCell_${o}`,()=>{n.build(t),Array.isArray(n.stateSize)?e=n.stateSize[0]:e=n.stateSize,t=[t[0],e]})}),this.built=!0}getConfig(){let t=super.getConfig(),e=s=>({className:s.getClassName(),config:s.getConfig()}),o={cells:this.cells.map(e)};return Object.assign(Object.assign({},t),o)}static fromConfig(t,e,n={}){let o=[];for(let s of e.cells)o.push(gn(s,n));return new t({cells:o})}get trainableWeights(){if(!this.trainable)return[];let t=[];for(let e of this.cells)t.push(...e.trainableWeights);return t}get nonTrainableWeights(){let t=[];for(let e of this.cells)t.push(...e.nonTrainableWeights);if(!this.trainable){let e=[];for(let n of this.cells)e.push(...n.trainableWeights);return e.concat(t)}return t}getWeights(){let t=[];for(let e of this.cells)t.push(...e.weights);return Ih(t)}setWeights(t){let e=[];for(let n of this.cells){let o=n.weights.length,s=t.splice(o);for(let i=0;i<n.weights.length;++i)e.push([n.weights[i],s[i]])}Pm(e)}};Ic.className=\"StackedRNNCells\";Q.registerClass(Ic);function cl(r){let{ones:t,rate:e,training:n=!1,count:o=1,dropoutFunc:s}=r,i=()=>s!=null?s(t(),e):Ay(t(),e),a=()=>xu(i,t,n);return!o||o<=1?De(a().clone()):Array(o).fill(void 0).map(a).map(l=>De(l.clone()))}var Q8=function(r,t){var e={};for(var n in r)Object.prototype.hasOwnProperty.call(r,n)&&t.indexOf(n)<0&&(e[n]=r[n]);if(r!=null&&typeof Object.getOwnPropertySymbols==\"function\")for(var o=0,n=Object.getOwnPropertySymbols(r);o<n.length;o++)t.indexOf(n[o])<0&&Object.prototype.propertyIsEnumerable.call(r,n[o])&&(e[n[o]]=r[n[o]]);return e};var db=class extends Tn{constructor(t){if(t.unroll)throw new St(\"Unrolling is not possible with convolutional RNNs.\");if(Array.isArray(t.cell))throw new St(\"It is not possible at the moment to stack convolutional cells.\");super(t),this.inputSpec=[new ye({ndim:5})]}call(t,e){return B(()=>{if(this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),e&&e.constants)throw new M(\"ConvRNN2D cell does not support constants\");let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}computeOutputShape(t){let e=this.computeSingleOutputShape(t);return this.returnSequences||(e=[e[0],...e.slice(2)]),this.returnState&&(e=[e,...Array(2).fill([t[0],...e.slice(-3)])]),e}getInitialState(t){return B(()=>{let{stateSize:e}=this.cell,n=t.shape,o=this.computeSingleOutputShape(n),s=[o[0],...o.slice(2)],i=Ne(s);return Array.isArray(e)?Array(e.length).fill(i):[i]})}resetStates(t,e=!1){B(()=>{if(!this.stateful)throw new vn(\"Cannot call resetStates() on an RNN Layer that is not stateful.\");let n=this.inputSpec[0].shape,o=this.computeSingleOutputShape(n),s=[o[0],...o.slice(2)];if(n[0]==null)throw new M(\"If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \\n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.\");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Ne(s)):this.states_=[Ne(s)];else if(t==null)vt(this.states_),this.keptStates!=null&&(vt(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Ne(s)):this.states_[0]=Ne(s);else{if(Array.isArray(t)||(t=[t]),t.length!==this.states_.length)throw new M(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${t.length} state value(s). Input received: ${t}`);e?this.keptStates.push(this.states_.slice()):vt(this.states_);for(let a=0;a<this.states_.length;++a){let u=t[a],l=s;if(!y.arraysEqual(u.shape,l))throw new M(`State ${a} is incompatible with layer ${this.name}: expected shape=${l}, received shape=${u.shape}`);this.states_[a]=u}}this.states_=this.states_.map(a=>De(a.clone()))})}computeSingleOutputShape(t){let{dataFormat:e,filters:n,kernelSize:o,padding:s,strides:i,dilationRate:a}=this.cell,u=e===\"channelsFirst\",l=t[u?3:2],c=t[u?4:3],p=Nn(l,o[0],s,i[0],a[0]),m=Nn(c,o[1],s,i[1],a[1]);return[...t.slice(0,2),...u?[n,p,m]:[p,m,n]]}};db.className=\"ConvRNN2D\";var Sc=class extends ul{constructor(t){let{filters:e,kernelSize:n,strides:o,padding:s,dataFormat:i,dilationRate:a}=t;super(Object.assign(Object.assign({},t),{units:e})),this.filters=e,Ze(this.filters,\"filters\"),this.kernelSize=Cu(n,2,\"kernelSize\"),this.kernelSize.forEach(u=>Ze(u,\"kernelSize\")),this.strides=Cu(o||1,2,\"strides\"),this.strides.forEach(u=>Ze(u,\"strides\")),this.padding=s||\"valid\",pn(this.padding),this.dataFormat=i||\"channelsLast\",Fe(this.dataFormat),this.dilationRate=Cu(a||1,2,\"dilationRate\"),this.dilationRate.forEach(u=>Ze(u,\"dilationRate\"))}build(t){var e;t=Bt(t);let n=this.dataFormat===\"channelsFirst\"?1:t.length-1;if(t[n]==null)throw new M(`The channel dimension of the input should be defined. Found ${t[n]}`);let o=t[n],s=4,i=this.kernelSize.concat([o,this.filters*s]);this.kernel=this.addWeight(\"kernel\",i,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let a=this.kernelSize.concat([this.filters,this.filters*s]);if(this.recurrentKernel=this.addWeight(\"recurrent_kernel\",a,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let u;if(this.unitForgetBias){let l=this.biasInitializer,c=this.filters;u=new(e=class extends dn{apply(m,f){let d=l.apply([c]),h=cr([c]),g=l.apply([c*2]);return Nm([d,h,g])}},e.className=\"CustomInit\",e)}else u=this.biasInitializer;this.bias=this.addWeight(\"bias\",[this.filters*s],null,u,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(t,e){return B(()=>{if(t.length!==3)throw new M(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${t.length}.`);let n=e.training||!1,o=t[0],s=t[1],i=t[2],a=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(o),rate:this.dropout,training:n,count:a,dropoutFunc:this.dropoutFunc}));let u=this.dropoutMask,l=(rt,ot,at)=>!ot||!ot[at]?rt:D(ot[at],rt),c=l(o,u,0),p=l(o,u,1),m=l(o,u,2),f=l(o,u,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(s),rate:this.recurrentDropout,training:n,count:a,dropoutFunc:this.dropoutFunc}));let d=this.recurrentDropoutMask,h=l(s,d,0),g=l(s,d,1),x=l(s,d,2),b=l(s,d,3),w=3,[C,N,_,A]=mr(this.kernel.read(),a,w),[$,F,P,V]=this.useBias?mr(this.bias.read(),a):[null,null,null,null];c=this.inputConv(c,C,$,this.padding),p=this.inputConv(p,N,F,this.padding),m=this.inputConv(m,_,P,this.padding),f=this.inputConv(f,A,V,this.padding);let[G,W,q,H]=mr(this.recurrentKernel.read(),a,w);h=this.recurrentConv(h,G),g=this.recurrentConv(g,W),x=this.recurrentConv(x,q),b=this.recurrentConv(b,H);let j=this.recurrentActivation.apply(X(c,h)),Y=this.recurrentActivation.apply(X(p,g)),Z=X(D(Y,i),D(j,this.activation.apply(X(m,x)))),et=D(this.recurrentActivation.apply(X(f,b)),this.activation.apply(Z));return[et,et,Z]})}getConfig(){let t=super.getConfig(),{units:e}=t,n=Q8(t,[\"units\"]),o={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),o)}inputConv(t,e,n,o){let s=In(t,e,this.strides,o||\"valid\",this.dataFormat===\"channelsFirst\"?\"NCHW\":\"NHWC\",this.dilationRate);return n?fn(s,n,this.dataFormat):s}recurrentConv(t,e){return In(t,e,1,\"same\",this.dataFormat===\"channelsFirst\"?\"NCHW\":\"NHWC\")}};Sc.className=\"ConvLSTM2DCell\";Q.registerClass(Sc);var uf=class extends db{constructor(t){let e=new Sc(t);super(Object.assign(Object.assign({},t),{cell:e}))}static fromConfig(t,e){return new t(e)}};uf.className=\"ConvLSTM2D\";Q.registerClass(uf);var vc=class extends $t{constructor(t){super(t),this.rate=Math.max(Math.min(t.rate,1),0),this.noiseShape=t.noiseShape,this.seed=t.seed,this.supportsMasking=!0}getNoiseShape(t){if(this.noiseShape==null)return this.noiseShape;let e=t.shape,n=[];for(let o=0;o<this.noiseShape.length;++o)n.push(this.noiseShape[o]==null?e[o]:this.noiseShape[o]);return n}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);if(0<this.rate&&this.rate<1){let o=e.training==null?!1:e.training,s=this.getNoiseShape(n);return xu(()=>Ay(n,this.rate,s,this.seed),()=>n,o)}return t})}getConfig(){let t={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},e=super.getConfig();return Object.assign(t,e),t}dispose(){return super.dispose()}};vc.className=\"Dropout\";Q.registerClass(vc);var cf=class extends vc{constructor(t){super(t),this.inputSpec=[{ndim:3}]}getNoiseShape(t){let e=t.shape;return[e[0],1,e[2]]}};cf.className=\"SpatialDropout1D\";Q.registerClass(cf);var pf=class extends $t{constructor(t){if(super(t),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER=\"glorotNormal\",this.DEFAULT_BIAS_INITIALIZER=\"zeros\",t.batchInputShape==null&&t.inputShape==null&&t.inputDim!=null){let e=null;t.batchSize!=null&&(e=t.batchSize),this.batchInputShape=[e,t.inputDim]}this.units=t.units,Ze(this.units,\"units\"),this.activation=Xs(t.activation),t.useBias!=null&&(this.useBias=t.useBias),this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Be(t.kernelConstraint),this.biasConstraint=Be(t.biasConstraint),this.kernelRegularizer=be(t.kernelRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.activityRegularizer=be(t.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(t){t=Bt(t);let e=t[t.length-1];this.kernel==null&&(this.kernel=this.addWeight(\"kernel\",[e,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight(\"bias\",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:e}}],this.built=!0}computeOutputShape(t){t=Bt(t);let e=t.slice();return e[e.length-1]=this.units,e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=Iy(this.activation.getClassName()),s;return o!=null?s=To(n,this.kernel.read(),o,this.bias?this.bias.read():null):(s=To(n,this.kernel.read()),this.bias!=null&&(s=fn(s,this.bias.read())),this.activation!=null&&(s=this.activation.apply(s))),s})}getConfig(){let t={units:this.units,activation:js(this.activation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),biasConstraint:ze(this.biasConstraint)},e=super.getConfig();return Object.assign(t,e),t}};pf.className=\"Dense\";Q.registerClass(pf);var mf=class extends $t{constructor(t){t=t||{},super(t),this.inputSpec=[{minNDim:3}],this.dataFormat=t.dataFormat}computeOutputShape(t){t=Bt(t);for(let e of t.slice(1))if(e==null)throw new M(`The shape of the input to \"Flatten\" is not fully defined (got ${t.slice(1)}). Make sure to pass a complete \"input_shape\" or \"batch_input_shape\" argument to the first layer in your model.`);return[t[0],No(t,1)]}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);if(this.dataFormat===\"channelsFirst\"&&n.rank>1){let o=[0];for(let s=2;s<n.rank;++s)o.push(s);o.push(1),n=Ot(n,o)}return R$(n)})}getConfig(){let t={};this.dataFormat!=null&&(t.dataFormat=this.dataFormat);let e=super.getConfig();return Object.assign(t,e),t}};mf.className=\"Flatten\";Q.registerClass(mf);var ff=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.activation=Xs(t.activation)}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return this.activation.apply(n)})}getConfig(){let t={activation:js(this.activation)},e=super.getConfig();return Object.assign(t,e),t}};ff.className=\"Activation\";Q.registerClass(ff);var df=class extends $t{constructor(t){super(t),this.n=t.n,this.inputSpec=[{ndim:2}]}computeOutputShape(t){return[t[0],this.n,t[1]]}call(t,e){return B(()=>(t=Nt(t),$$(t,this.n)))}getConfig(){let t={n:this.n},e=super.getConfig();return Object.assign(t,e),t}};df.className=\"RepeatVector\";Q.registerClass(df);var hf=class extends $t{constructor(t){super(t),this.targetShape=t.targetShape;for(let e=0;e<this.targetShape.length;++e)this.isUnknown(this.targetShape[e])&&(this.targetShape[e]=null)}isUnknown(t){return t<0||t==null}fixUnknownDimension(t,e){let n=\"Total size of new array must be unchanged.\",o=e.slice(),s=1,i=null;for(let u=0;u<o.length;++u){let l=o[u];if(this.isUnknown(l))if(i===null)i=u;else throw new M(\"Can only specifiy one unknown dimension.\");else s*=l}let a=No(t);if(i!==null){if(s===0||a%s!==0)throw new M(n);o[i]=a/s}else if(a!==s)throw new M(n);return o}computeOutputShape(t){let e=!1;for(let n=0;n<t.length;++n)if(this.isUnknown(t[n])){e=!0;break}return e?t.slice(0,1).concat(this.targetShape):t.slice(0,1).concat(this.fixUnknownDimension(t.slice(1),this.targetShape))}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=n.shape,s=o.slice(0,1).concat(this.fixUnknownDimension(o.slice(1),this.targetShape));return R(n,s)})}getConfig(){let t={targetShape:this.targetShape},e=super.getConfig();return Object.assign(t,e),t}};hf.className=\"Reshape\";Q.registerClass(hf);var gf=class extends $t{constructor(t){if(super(t),t.dims==null)throw new Error(\"Required configuration field `dims` is missing during Permute constructor call.\");if(!Array.isArray(t.dims))throw new Error(`Permute constructor requires \\`dims\\` to be an Array, but received ${t.dims} instead.`);let e=Zr(1,t.dims.length+1);if(!y.arraysEqual(t.dims.slice().sort(),e))throw new Error(\"Invalid permutation `dims`: \"+JSON.stringify(t.dims)+\" `dims` must contain consecutive integers starting from 1.\");this.dims=t.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new ye({ndim:this.dims.length+1})]}computeOutputShape(t){t=Bt(t);let e=t.slice();return this.dims.forEach((n,o)=>{e[o+1]=t[n]}),e}call(t,e){return Ot(Nt(t),this.dimsIncludingBatch)}getConfig(){let t={dims:this.dims},e=super.getConfig();return Object.assign(t,e),t}};gf.className=\"Permute\";Q.registerClass(gf);var xf=class extends $t{constructor(t){super(t==null?{}:t),this.supportsMasking=!0,t!=null?this.maskValue=t.maskValue==null?0:t.maskValue:this.maskValue=0}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={maskValue:this.maskValue};return Object.assign(e,t),e}computeMask(t,e){let n=Nt(t),o=-1;return qu(Bs(n,this.maskValue),o)}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=-1,s=!0,i=qu(Bs(n,this.maskValue),o,s);return D(n,J(i,n.dtype))})}};xf.className=\"Masking\";Q.registerClass(xf);var yf=class extends $t{constructor(t){if(super(t),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER=\"randomUniform\",t.batchInputShape==null&&t.inputShape==null){let e=null;t.batchSize!=null&&(e=t.batchSize),t.inputLength==null?this.batchInputShape=[e,null]:this.batchInputShape=[e].concat(xe(t.inputLength))}this.inputDim=t.inputDim,Ze(this.inputDim,\"inputDim\"),this.outputDim=t.outputDim,Ze(this.outputDim,\"outputDim\"),this.embeddingsInitializer=de(t.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=be(t.embeddingsRegularizer),this.activityRegularizer=be(t.activityRegularizer),this.embeddingsConstraint=Be(t.embeddingsConstraint),this.maskZero=t.maskZero,this.supportsMasking=t.maskZero,this.inputLength=t.inputLength}build(t){this.embeddings=this.addWeight(\"embeddings\",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(t){}computeMask(t,e){return B(()=>this.maskZero?(t=Nt(t),Bs(t,It(t))):null)}computeOutputShape(t){if(t=Bt(t),this.inputLength==null)return[...t,this.outputDim];let e=xe(this.inputLength);if(e.length!==t.length-1)throw new M(`\"inputLength\" is ${this.inputLength}, but received input shape has shape ${t}`);{let n=0;for(let o=0;o<e.length;++o){let s=e[o],i=t[o+1];if(s!=null&&i!=null&&s!==i)throw new M(`\"inputLength\" is ${this.inputLength}, but received input shape has shape ${t}`);s==null&&(e[n]=i),n++}}return[t[0],...e,this.outputDim]}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);n.dtype!==\"int32\"&&(n=no(n,\"int32\"));let o=_y(this.embeddings.read(),R(n,[n.size]));return R(o,Bt(this.computeOutputShape(n.shape)))})}getConfig(){let t={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Te(this.embeddingsInitializer),embeddingsRegularizer:me(this.embeddingsRegularizer),activityRegularizer:me(this.activityRegularizer),embeddingsConstraint:ze(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},e=super.getConfig();return Object.assign(t,e),t}};yf.className=\"Embedding\";Q.registerClass(yf);var pl=class extends $t{constructor(t){super(t||{}),this.supportsMasking=!0}mergeFunction(t){throw new St}computeElementwiseOpOutputShape(t,e){if(t==null||e==null)return null;if(t.length<e.length)return this.computeElementwiseOpOutputShape(e,t);if(e.length===0)return t;let n=t.slice(0,t.length-e.length);for(let o=0;o<e.length;++o){let s=t[t.length-e.length+o],i=e[o];if(s==null||i==null||s<0||i<0)n.push(null);else if(s===1)n.push(i);else if(i===1)n.push(s);else{if(s!==i)throw new M(\"Operands could not be broadcast together with shapes \"+JSON.stringify(t)+\" \"+JSON.stringify(e));n.push(s)}}return n}build(t){if(Array.isArray(t)&&!Array.isArray(t[0])&&(t=[Bt(t)]),t=t,t.length<2)throw new M(`A merge layer should be called on an Array of at least 2 inputs. Got ${t.length} input(s).`);let e=[];for(let s of t)s!=null&&s[0]!==null&&e.push(s[0]);if(e=vo(e),e.length>1)throw new M(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(t)}.`);let n=t[0]==null?null:t[0].slice(1);for(let s=1;s<t.length;++s){let i=t[s]==null?null:t[s].slice(1);n=this.computeElementwiseOpOutputShape(n,i)}let o=t.map(s=>s.length);t.indexOf(null)===-1&&vo(o).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(t,e){return B(()=>{if(t=t,this.reshapeRequired){let n=[],o=t.map(s=>s.rank);if(o.indexOf(null)===-1){let s=qs(o);for(let i of t){let a=i.rank;for(let u=0;u<s-a;++u)i=nl(i,1);n.push(i)}return this.mergeFunction(n)}else{let s=!1;for(let u of t){let l=u.rank;if(l==null){let c=u.shape,p=c[0],m=c.slice(1).concat([p]),f=R(u,[p].concat(No(c.slice(1))));f=Ot(f,[1,0]),f=R(f,m),n.push(f),s=!0}else if(l>1){let c=Zr(1,l).concat([0]);n.push(Ot(u,c)),s=!0}else n.push(u)}let i=this.mergeFunction(n),a=i.rank;if(s){if(a==null){let u=i.shape,l=u.length,c=u[l-1],p=[c].concat(u.slice(0,u.length-1));i=R(Ot(R(i,[-1,c]),[1,0]),p)}else if(a>1){let u=[a-1].concat(Zr(0,a-1));i=Ot(i,u)}}return i}}else return this.mergeFunction(t)})}computeOutputShape(t){t=t;let e;t[0]==null?e=null:e=t[0].slice(1);for(let o=1;o<t.length;++o){let s=t[o]==null?null:t[o].slice(1);e=this.computeElementwiseOpOutputShape(e,s)}let n=[];for(let o of t)o!=null&&o[0]!==null&&n.push(o[0]);return n=vo(n),n.length===1?e=n.concat(e):e=[null].concat(e),e}computeMask(t,e){return B(()=>{if(e==null)return null;if(!Array.isArray(e))throw new M(\"`mask` should be an Array\");if(!Array.isArray(t))throw new M(\"`inputs` should be an Array\");if(e.length!==t.length)throw new M(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${t.length} vs ${e.length})`);if(e.every(o=>o==null))return null;e=e.map(o=>o==null?o:rr(o,0));let n=e[0];for(let o=1;o<e.length-1;++o)n=Rr(n,e[o]);return n})}},bf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=X(e,t[n]);return e})}};bf.className=\"Add\";Q.registerClass(bf);var wf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=D(e,t[n]);return e})}};wf.className=\"Multiply\";Q.registerClass(wf);var Cf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=X(e,t[n]);return D(1/t.length,e)})}};Cf.className=\"Average\";Q.registerClass(Cf);var If=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0];for(let n=1;n<t.length;++n)e=Sn(e,t[n]);return e})}};If.className=\"Maximum\";Q.registerClass(If);var Sf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0];for(let n=1;n<t.length;++n)e=Mi(e,t[n]);return e})}};Sf.className=\"Minimum\";Q.registerClass(Sf);var vf=class extends pl{constructor(t){super(t),this.DEFAULT_AXIS=-1,t==null&&(t={}),this.axis=t.axis==null?this.DEFAULT_AXIS:t.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(t){if(!(Array.isArray(t)&&Array.isArray(t[0]))||t.length===1)throw new M(\"A `Concatenate` layer should be called on a list of at least 2 inputs\");t=t;let e=!0;for(let o of t)if(o!=null){e=!1;break}if(e)return;let n=[];for(let o=0;o<t.length;++o){let s=t[o].slice();s.splice(this.axis,1);let i=!1;for(let a of n)if(y.arraysEqual(a,s)){i=!0;break}i||n.push(s)}if(n.length>1)throw new M(\"A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: \"+JSON.stringify(t))}mergeFunction(t){return B(()=>Nm(t,this.axis))}computeOutputShape(t){if(!(Array.isArray(t)&&Array.isArray(t[0])))throw new M(\"A `Concatenate` layer should be called on a list of inputs.\");let e=t,n=e[0].slice(),o=this.axis<0?n.length+this.axis:this.axis;for(let s of e.slice(1)){if(n[o]==null||s[o]==null){n[o]=null;break}n[o]+=s[o]}return n}computeMask(t,e){if(e==null)return null;if(!Array.isArray(e))throw new M(\"`mask` should be an array for Concatenate\");if(!Array.isArray(t))throw new M(\"`inputs` should be an array for Concatenate\");if(e.length!==t.length)throw new M(`Mismatch in the length of mask (${e.length}) and the legnth of inputs (${t.length})`);return B(()=>{let n=!0;if(e.forEach(i=>{if(i!=null){n=!1;return}}),n)return null;let o=[];for(let i=0;i<t.length;++i)e[i]==null?o.push(J(yr(t[i]),\"bool\")):e[i].rank<t[i].rank?o.push(rr(e[i],-1)):o.push(e[i]);let s=ne(o,this.axis);return Zp(s,-1,!1)})}getConfig(){let t={axis:this.axis},e=super.getConfig();return Object.assign(t,e),t}};vf.className=\"Concatenate\";Q.registerClass(vf);function $h(r,t){for(;r<0;)r+=t;return r}function tY(r,t,e){if(r.shape.length>3||t.shape.length>3)throw new St(\"batchDot is not implemented for tensors of 4D or higher rank yet\");if(y.assert(r.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${r.shape.length}`),y.assert(r.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof e==\"number\"&&(e=[e,e]),r.dtype===\"complex64\"||t.dtype===\"complex64\")throw new St(\"batchDot is not implemented for complex64-type Tensors yet.\");let n=r.shape.length,o=t.shape.length;e==null&&(e=[n-1,o-2]);let s=e;return B(()=>{let i;if(n>o){i=n-o;let u=[];for(let l=0;l<i;++l)u.push(1);t=R(t,t.shape.concat(u))}else if(o>n){i=o-n;let u=[];for(let l=0;l<i;++l)u.push(1);r=R(r,r.shape.concat(u))}else i=0;let a;if(r.shape.length===2&&t.shape.length===2)s[0]===s[1]?a=ft(D(r,t),s[0]):a=ft(D(Ot(r,[1,0]),t),s[1]);else{let u=s[0]!==r.shape.length-1,l=s[1]===t.shape.length-1;a=Lt(r,t,u,l)}if(i>0){let u;n>o?u=n+o-3:u=n-1;let l=[];for(let c=u;c<u+i;++c)l.push(c);a=Mn(a,l)}return a.shape.length===1&&(a=rr(a,1)),a})}var Nf=class extends pl{constructor(t){super(t),this.axes=t.axes,this.normalize=t.normalize==null?!1:t.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(t){y.assert(Array.isArray(t)&&t.length===2&&Array.isArray(t[0])&&Array.isArray(t[1]),()=>\"A `Dot` layer should be called on a list of exactly 2 inputs.\");let e=t[0],n=t[1];if(e.length>3||n.length>3)throw new St(\"Dot layer does not support tensors of 4D or higher rank yet.\");let o=this.interpretAxes(e,n);if(e[o[0]]!==n[o[1]])throw new M(`Dimension incompatibility: ${e[o[0]]} !== ${n[o[1]]}`)}mergeFunction(t){if(t.length!==2)throw new M(`A \\`Dot\\` layer must be called on exactly 2 inputs, but received ${t.length} input(s).`);let e=t[0],n=t[1],o;return Array.isArray(this.axes)?o=this.axes.map((s,i)=>$h(s,t[i].shape.length)):o=[$h(this.axes,e.shape.length),$h(this.axes,n.shape.length)],this.normalize&&(e=Sh(e,o[0]),n=Sh(n,o[1])),tY(e,n,o)}interpretAxes(t,e){let n;return Array.isArray(this.axes)?n=this.axes:n=[$h(this.axes,t.length),$h(this.axes,e.length)],n}computeOutputShape(t){y.assert(Array.isArray(t)&&t.length===2&&Array.isArray(t[0])&&Array.isArray(t[1]),()=>\"A `Dot` layer should be called on a list of exactly 2 inputs.\");let e=t[0].slice(),n=t[1].slice();if(e.length>3||n.length>3)throw new St(\"Dot layer does not support tensors of 4D or higher rank yet.\");let o=this.interpretAxes(e,n);e.splice(o[0],1),n.splice(o[1],1),n.splice(0,1);let s=e.concat(n);return s.length===1&&s.push(1),s}computeMask(t,e){return null}getConfig(){let t={axes:this.axes,normalize:this.normalize},e=super.getConfig();return Object.assign(t,e),t}};Nf.className=\"Dot\";Q.registerClass(Nf);var Tf=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.stddev=t.stddev}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={stddev:this.stddev};return Object.assign(e,t),e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return xu(()=>X(Tm(n.shape,0,this.stddev),n),()=>n,e.training||!1)})}};Tf.className=\"GaussianNoise\";Q.registerClass(Tf);var kf=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.rate=t.rate}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={rate:this.rate};return Object.assign(e,t),e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return this.rate>0&&this.rate<1?xu(()=>{let s=Math.sqrt(this.rate/(1-this.rate));return D(n,Tm(n.shape,1,s))},()=>n,e.training||!1):n})}};kf.className=\"GaussianDropout\";Q.registerClass(kf);var Ef=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.rate=t.rate,this.noiseShape=t.noiseShape}_getNoiseShape(t){return this.noiseShape||Nt(t).shape}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={rate:this.rate};return Object.assign(e,t),e}call(t,e){return B(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(t);return xu(()=>{let s=Nt(t),i=1.6732632423543772,a=1.0507009873554805,u=-i*a,l=ln(zi(n),this.rate);l=no(l,\"float32\");let c=((1-this.rate)*(1+this.rate*u**2))**-.5,p=-c*u*this.rate,m=X(D(s,l),D(X(l,-1),u));return X(D(m,c),p)},()=>Nt(t),e.training||!1)}return t})}};Ef.className=\"AlphaDropout\";Q.registerClass(Ef);function Dh(r,t,e,n,o,s=.001){let i;if(r.rank===2)i=xx(r,t,e,n,o,s);else if(r.rank===3)i=yx(r,t,e,n,o,s);else if(r.rank===4)i=bx(r,t,e,n,o,s);else throw new St(`batchNormalization is not implemented for array of rank ${r.rank} yet`);return i}function eY(r,t,e,n,o=.001){return B(()=>{let s=Zu(r,n),i=s.mean,a=s.variance;return[Dh(r,i,a,e,t,o),i,a]})}function rY(r,t,e,n,o=.001){return B(()=>{let s=Zu(r,n),i=s.mean,a=s.variance,u=[];for(let d of Zr(0,r.rank))n.indexOf(d)!==-1?u.push(1):u.push(r.shape[d]);let l=R(i,u),c=R(a,u),p=t==null?null:R(t,u),m=e==null?null:R(e,u);return[Dh(r,l,c,m,p,o),i,a]})}function nY(r,t,e,n,o=.001){return y.arraysEqual(n.slice().sort(),Zr(0,r.rank-1))?eY(r,t,e,n,o):rY(r,t,e,n,o)}var _f=class extends $t{constructor(t){t==null&&(t={}),super(t),this.supportsMasking=!0,this.axis=t.axis==null?-1:t.axis,this.momentum=t.momentum==null?.99:t.momentum,this.epsilon=t.epsilon==null?.001:t.epsilon,this.center=t.center==null?!0:t.center,this.scale=t.scale==null?!0:t.scale,this.betaInitializer=de(t.betaInitializer||\"zeros\"),this.gammaInitializer=de(t.gammaInitializer||\"ones\"),this.movingMeanInitializer=de(t.movingMeanInitializer||\"zeros\"),this.movingVarianceInitializer=de(t.movingVarianceInitializer||\"ones\"),this.betaConstraint=Be(t.betaConstraint),this.gammaConstraint=Be(t.gammaConstraint),this.betaRegularizer=be(t.betaRegularizer),this.gammaRegularizer=be(t.gammaRegularizer)}build(t){t=Bt(t);let e=this.axis>=0?this.axis:this.axis+t.length,n=t[e];if(n==null)throw new M(`Axis ${e} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(t)}.`);this.inputSpec=[new ye({ndim:t.length,axes:{[e]:n}})];let o=[n];this.scale&&(this.gamma=this.addWeight(\"gamma\",o,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight(\"beta\",o,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight(\"moving_mean\",o,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight(\"moving_variance\",o,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(t,e){return B(()=>{let n=e.training==null?!1:e.training,o=Nt(t),s=o.shape,i=s.length,a=Zr(0,i),u=this.axis>=0?this.axis:this.axis+i;a.splice(u,1);let l=Io(1,i);l[u]=s[u];let c=a.slice();c.sort();let p=!y.arraysEqual(c,Zr(0,i).slice(0,i-1)),m=()=>{if(p){let b=R(this.movingMean.read(),l),w=R(this.movingVariance.read(),l),C=this.center?R(this.beta.read(),l):null,N=this.scale?R(this.gamma.read(),l):null;return Dh(o,b,w,C,N,this.epsilon)}else return Dh(o,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!n)return m();let[f,d,h]=nY(o,this.gamma.read(),this.beta.read(),a,this.epsilon),g=(b,w,C)=>{B(()=>{let N=1-C,_=b.read(),A=D(ct(_,w),N);b.write(ct(_,A))})};return(()=>{g(this.movingMean,d,this.momentum),g(this.movingVariance,h,this.momentum)})(),f})}getConfig(){let t={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Te(this.betaInitializer),gammaInitializer:Te(this.gammaInitializer),movingMeanInitializer:Te(this.movingMeanInitializer),movingVarianceInitializer:Te(this.movingVarianceInitializer),betaRegularizer:me(this.betaRegularizer),gammaRegularizer:me(this.gammaRegularizer),betaConstraint:ze(this.betaConstraint),gammaConstraint:ze(this.gammaConstraint)},e=super.getConfig();return Object.assign(t,e),t}};_f.className=\"BatchNormalization\";Q.registerClass(_f);var Af=class extends $t{constructor(t){if(t==null&&(t={}),super(t),this.axis=t.axis==null?-1:t.axis,typeof this.axis==\"number\"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let e of this.axis)if(!Number.isInteger(e))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=t.epsilon==null?.001:t.epsilon,this.center=t.center==null?!0:t.center,this.scale=t.scale==null?!0:t.scale,this.betaInitializer=de(t.betaInitializer||\"zeros\"),this.gammaInitializer=de(t.gammaInitializer||\"ones\"),this.betaRegularizer=be(t.betaRegularizer),this.gammaRegularizer=be(t.gammaRegularizer),this.supportsMasking=!0}build(t){t=Bt(t);let e=t.length;typeof this.axis==\"number\"&&(this.axis=[this.axis]);for(let s=0;s<this.axis.length;++s)this.axis[s]<0&&(this.axis[s]+=e);for(let s of this.axis)if(s<0||s>=e)throw new Error(`Invalid axis: ${s}`);if(this.axis.length!==vo(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(s=>t[s]),o=!0;this.scale?this.gamma=this.addWeight(\"gamma\",n,\"float32\",this.gammaInitializer,this.gammaRegularizer,o):this.gamma=null,this.center?this.beta=this.addWeight(\"beta\",n,\"float32\",this.betaInitializer,this.betaRegularizer,o):this.beta=null,this.built=!0}call(t,e){let n=Nt(t),o=n.shape,s=o.length;return B(()=>{let{mean:a,variance:u}=Zu(n,this.axis,!0),l=Io(1,s);for(let h of this.axis)l[h]=o[h];let c=h=>h!=null&&h.shape.length!==s?R(h,l):h,p=this.scale?c(this.gamma.read()):null,m=this.center?c(this.beta.read()):null,f=[],d=[];for(let h=0;h<s;++h)this.axis.indexOf(h)!==-1?(f.push(o[h]),d.push(1)):(f.push(1),d.push(o[h]));return a=Dr(a,f),u=Dr(u,f),p!=null&&(p=Dr(p,d)),m!=null&&(m=Dr(m,d)),Dh(n,a,u,m,p,this.epsilon)})}getConfig(){let t={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Te(this.betaInitializer),gammaInitializer:Te(this.gammaInitializer),betaRegularizer:me(this.betaRegularizer),gammaRegularizer:me(this.gammaRegularizer)},e=super.getConfig();return Object.assign(t,e),t}};Af.className=\"LayerNormalization\";Q.registerClass(Af);function oY(r,t,e){return B(()=>{if(r.rank!==4)throw new M(`temporalPadding expects input tensor to be 4-D, but received a ${r.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new M(\"spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.\");if(e==null&&(e=mn()),e!==\"channelsLast\"&&e!==\"channelsFirst\")throw new M(`Unknown data format: ${e}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let n;return e===\"channelsFirst\"?n=[[0,0],[0,0],t[0],t[1]]:n=[[0,0],t[0],t[1],[0,0]],cn(r,n)})}var $f=class extends $t{constructor(t){if(t==null&&(t={}),super(t),this.dataFormat=t.dataFormat==null?mn():t.dataFormat,t.padding==null)this.padding=[[1,1],[1,1]];else if(typeof t.padding==\"number\")this.padding=[[t.padding,t.padding],[t.padding,t.padding]];else{if(t.padding=t.padding,t.padding.length!==2)throw new M(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${t.padding.length} array.`);let e,n;if(typeof t.padding[0]==\"number\")e=[t.padding[0],t.padding[0]],n=[t.padding[1],t.padding[1]];else{if(t.padding=t.padding,t.padding[0].length!==2)throw new M(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${t.padding[0].length} array.`);if(e=t.padding[0],t.padding[1].length!==2)throw new M(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${t.padding[1].length} array.`);n=t.padding[1]}this.padding=[e,n]}this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){t=Bt(t);let e,n;return this.dataFormat===\"channelsFirst\"?(t[2]!=null&&t[2]>=0?e=t[2]+this.padding[0][0]+this.padding[0][1]:e=null,t[3]!=null&&t[3]>=0?n=t[3]+this.padding[1][0]+this.padding[1][1]:n=null,[t[0],t[1],e,n]):(t[1]!=null&&t[1]>=0?e=t[1]+this.padding[0][0]+this.padding[0][1]:e=null,t[2]!=null&&t[2]>=0?n=t[2]+this.padding[1][0]+this.padding[1][1]:n=null,[t[0],e,n,t[3]])}call(t,e){return B(()=>oY(Nt(t),this.padding,this.dataFormat))}getConfig(){let t={padding:this.padding,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}};$f.className=\"ZeroPadding2D\";Q.registerClass($f);function wb(r,t,e,n,o,s){return B(()=>{Fe(o),Iv(s),pn(n),e==null&&(e=[1,1]),n==null&&(n=\"valid\"),o==null&&(o=mn()),s==null&&(s=\"max\"),r=Ah(r,o);let i,a=n===\"same\"?\"same\":\"valid\";return s===\"max\"?i=ru(r,t,e,a):i=Yl(r,t,e,a),o===\"channelsFirst\"&&(i=Ot(i,[0,3,1,2])),i})}function CD(r,t,e,n,o,s){return B(()=>{Fe(o),Iv(s),pn(n),e==null&&(e=[1,1,1]),n==null&&(n=\"valid\"),o==null&&(o=mn()),s==null&&(s=\"max\"),r=zv(r,o);let i,a=n===\"same\"?\"same\":\"valid\";return s===\"max\"?i=Hx(r,t,e,a):i=gx(r,t,e,a),o===\"channelsFirst\"&&(i=Ot(i,[0,4,1,2,3])),i})}var hb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=2),super(t),typeof t.poolSize==\"number\")this.poolSize=[t.poolSize];else if(Array.isArray(t.poolSize)&&t.poolSize.length===1&&typeof t.poolSize[0]==\"number\")this.poolSize=t.poolSize;else throw new M(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(t.poolSize)}`);if(Ze(this.poolSize,\"poolSize\"),t.strides==null)this.strides=this.poolSize;else if(typeof t.strides==\"number\")this.strides=[t.strides];else if(Array.isArray(t.strides)&&t.strides.length===1&&typeof t.strides[0]==\"number\")this.strides=t.strides;else throw new M(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(t.strides)}`);Ze(this.strides,\"strides\"),this.padding=t.padding==null?\"valid\":t.padding,pn(this.padding),this.inputSpec=[new ye({ndim:3})]}computeOutputShape(t){t=Bt(t);let e=Nn(t[1],this.poolSize[0],this.padding,this.strides[0]);return[t[0],e,t[2]]}call(t,e){return B(()=>{this.invokeCallHook(t,e),t=nl(Nt(t),2);let n=this.poolingFunction(Nt(t),[this.poolSize[0],1],[this.strides[0],1],this.padding,\"channelsLast\");return Mn(n,[2])})}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides},e=super.getConfig();return Object.assign(t,e),t}},Df=class extends hb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,\"max\")}};Df.className=\"MaxPooling1D\";Q.registerClass(Df);var Rf=class extends hb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,\"avg\")}};Rf.className=\"AveragePooling1D\";Q.registerClass(Rf);var gb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=[2,2]),super(t),this.poolSize=Array.isArray(t.poolSize)?t.poolSize:[t.poolSize,t.poolSize],t.strides==null)this.strides=this.poolSize;else if(Array.isArray(t.strides)){if(t.strides.length!==2)throw new M(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${t.strides.length}.`);this.strides=t.strides}else this.strides=[t.strides,t.strides];Ze(this.poolSize,\"poolSize\"),Ze(this.strides,\"strides\"),this.padding=t.padding==null?\"valid\":t.padding,this.dataFormat=t.dataFormat==null?\"channelsLast\":t.dataFormat,Fe(this.dataFormat),pn(this.padding),this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){t=Bt(t);let e=this.dataFormat===\"channelsFirst\"?t[2]:t[1],n=this.dataFormat===\"channelsFirst\"?t[3]:t[2];return e=Nn(e,this.poolSize[0],this.padding,this.strides[0]),n=Nn(n,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat===\"channelsFirst\"?[t[0],t[1],e,n]:[t[0],e,n,t[3]]}call(t,e){return B(()=>(this.invokeCallHook(t,e),this.poolingFunction(Nt(t),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Ff=class extends gb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,\"max\")}};Ff.className=\"MaxPooling2D\";Q.registerClass(Ff);var Of=class extends gb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,\"avg\")}};Of.className=\"AveragePooling2D\";Q.registerClass(Of);var xb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=[2,2,2]),super(t),this.poolSize=Array.isArray(t.poolSize)?t.poolSize:[t.poolSize,t.poolSize,t.poolSize],t.strides==null)this.strides=this.poolSize;else if(Array.isArray(t.strides)){if(t.strides.length!==3)throw new M(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${t.strides.length}.`);this.strides=t.strides}else this.strides=[t.strides,t.strides,t.strides];Ze(this.poolSize,\"poolSize\"),Ze(this.strides,\"strides\"),this.padding=t.padding==null?\"valid\":t.padding,this.dataFormat=t.dataFormat==null?\"channelsLast\":t.dataFormat,Fe(this.dataFormat),pn(this.padding),this.inputSpec=[new ye({ndim:5})]}computeOutputShape(t){t=Bt(t);let e=this.dataFormat===\"channelsFirst\"?t[2]:t[1],n=this.dataFormat===\"channelsFirst\"?t[3]:t[2],o=this.dataFormat===\"channelsFirst\"?t[4]:t[3];return e=Nn(e,this.poolSize[0],this.padding,this.strides[0]),n=Nn(n,this.poolSize[1],this.padding,this.strides[1]),o=Nn(o,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat===\"channelsFirst\"?[t[0],t[1],e,n,o]:[t[0],e,n,o,t[4]]}call(t,e){return B(()=>(this.invokeCallHook(t,e),this.poolingFunction(Nt(t),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Pf=class extends xb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),CD(t,e,n,o,s,\"max\")}};Pf.className=\"MaxPooling3D\";Q.registerClass(Pf);var Lf=class extends xb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),CD(t,e,n,o,s,\"avg\")}};Lf.className=\"AveragePooling3D\";Q.registerClass(Lf);var yb=class extends $t{constructor(t){super(t),this.inputSpec=[new ye({ndim:3})]}computeOutputShape(t){return[t[0],t[2]]}call(t,e){throw new St}},Mf=class extends yb{constructor(t){super(t||{})}call(t,e){return B(()=>{let n=Nt(t);return ve(n,1)})}};Mf.className=\"GlobalAveragePooling1D\";Q.registerClass(Mf);var zf=class extends yb{constructor(t){super(t||{})}call(t,e){return B(()=>{let n=Nt(t);return Ir(n,1)})}};zf.className=\"GlobalMaxPooling1D\";Q.registerClass(zf);var bb=class extends $t{constructor(t){super(t),this.dataFormat=t.dataFormat==null?\"channelsLast\":t.dataFormat,Fe(this.dataFormat),this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){return t=t,this.dataFormat===\"channelsLast\"?[t[0],t[3]]:[t[0],t[1]]}call(t,e){throw new St}getConfig(){let t={dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Bf=class extends bb{call(t,e){return B(()=>{let n=Nt(t);return this.dataFormat===\"channelsLast\"?ve(n,[1,2]):ve(n,[2,3])})}};Bf.className=\"GlobalAveragePooling2D\";Q.registerClass(Bf);var Vf=class extends bb{call(t,e){return B(()=>{let n=Nt(t);return this.dataFormat===\"channelsLast\"?Ir(n,[1,2]):Ir(n,[2,3])})}};Vf.className=\"GlobalMaxPooling2D\";Q.registerClass(Vf);var Cb=class extends $t{constructor(t){super(t),this.layer=t.layer}build(t){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(t){this.layer!=null&&(this.layer.trainable=t)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(t){this.layer.setWeights(t)}getConfig(){let t={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},e=super.getConfig();return Object.assign(t,e),t}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(t)}static fromConfig(t,e,n={}){let o=e.layer,s=gn(o,n);delete e.layer;let i={layer:s};return Object.assign(i,e),new t(i)}},Gf=class extends Cb{constructor(t){super(t),this.supportsMasking=!0}build(t){if(t=Bt(t),t.length<3)throw new M(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(t)}`);this.inputSpec=[{shape:t}];let e=[t[0]].concat(t.slice(2));this.layer.built||(this.layer.build(e),this.layer.built=!0),super.build(t)}computeOutputShape(t){t=Bt(t);let e=[t[0]].concat(t.slice(2)),n=this.layer.computeOutputShape(e),o=t[1];return[n[0],o].concat(n.slice(1))}call(t,e){return B(()=>(t=Nt(t),Vv((i,a)=>[Nt(this.layer.call(i,e)),[]],t,[],!1,null,null,!1,!0)[1]))}};Gf.className=\"TimeDistributed\";Q.registerClass(Gf);function sY(r){Wi(T$,\"BidirectionalMergeMode\",r)}var iY=\"concat\",Wf=class extends Cb{constructor(t){super(t);let e=t.layer.getConfig(),n={};n.className=t.layer.getClassName(),n.config=e,this.forwardLayer=gn(n),e.goBackwards=e.goBackwards!==!0;let o={};if(o.className=t.layer.getClassName(),o.config=e,this.backwardLayer=gn(o),this.forwardLayer.name=\"forward_\"+this.forwardLayer.name,this.backwardLayer.name=\"backward_\"+this.backwardLayer.name,this.mergeMode=t.mergeMode===void 0?iY:t.mergeMode,sY(this.mergeMode),t.weights)throw new St(\"weights support is not implemented for Bidirectional layer yet.\");this._stateful=t.layer.stateful,this.returnSequences=t.layer.returnSequences,this.returnState=t.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=t.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(t){this._trainable=t,this.forwardLayer!=null&&(this.forwardLayer.trainable=t),this.backwardLayer!=null&&(this.backwardLayer.trainable=t)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(t){let e=t.length,n=Math.floor(e/2);this.forwardLayer.setWeights(t.slice(0,n)),this.backwardLayer.setWeights(t.slice(n))}computeOutputShape(t){let e=this.forwardLayer.computeOutputShape(t);Array.isArray(e)&&Array.isArray(e[0])||(e=[e]),e=e;let n,o,s;return this.returnState&&(s=e.slice(1)),n=e[0],n=n,this.mergeMode===\"concat\"?(n[n.length-1]*=2,o=[n]):this.mergeMode==null?o=[n,n.slice()]:o=[n],this.returnState?this.mergeMode==null?o.concat(s).concat(s.slice()):[n].concat(s).concat(s.slice()):Nr(o)}apply(t,e){let n=e==null?null:e.initialState,o=e==null?null:e.constants;e==null&&(e={});let s=Bv(t,n,o,this.numConstants);if(t=s.inputs,n=s.initialState,o=s.constants,Array.isArray(t)&&(n=t.slice(1),t=t[0]),(n==null||n.length===0)&&o==null)return super.apply(t,e);let i=[],a=[];if(n!=null){let l=n.length;if(l%2>0)throw new M(\"When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.\");e.initialState=n,i.push(...n);let c=n.map(p=>new ye({shape:p.shape}));this.forwardLayer.stateSpec=c.slice(0,l/2),this.backwardLayer.stateSpec=c.slice(l/2),a.push(...c)}if(o!=null)throw new St(\"Support for constants in Bidirectional layers is not implemented yet.\");let u=i[0]instanceof Jr;for(let l of i)if(l instanceof Jr!==u)throw new M(\"The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors\");if(u){let l=[t].concat(i),c=this.inputSpec.concat(a),p=this.inputSpec;this.inputSpec=c;let m=super.apply(l,e);return this.inputSpec=p,m}else return super.apply(t,e)}call(t,e){return B(()=>{let n=e.initialState,o,s;if(n==null)o=this.forwardLayer.call(t,e),s=this.backwardLayer.call(t,e);else{let u=n.slice(0,n.length/2),l=n.slice(n.length/2);o=this.forwardLayer.call(t,Object.assign(e,{initialState:u})),s=this.backwardLayer.call(t,Object.assign(e,{initialState:l}))}let i;this.returnState&&(Array.isArray(o)&&(i=o.slice(1).concat(s.slice(1))),o=o[0],s=s[0]),this.returnSequences&&(s=pr(s,1));let a;return this.mergeMode===\"concat\"?a=Nm([o,s]):this.mergeMode===\"sum\"?a=X(o,s):this.mergeMode===\"ave\"?a=D(.5,X(o,s)):this.mergeMode===\"mul\"?a=D(o,s):this.mergeMode==null&&(a=[o,s]),this.returnState?this.mergeMode==null?a.concat(i):[a].concat(i):a})}resetStates(t){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(t){Hs(this.forwardLayer.name,()=>{this.forwardLayer.build(t)}),Hs(this.backwardLayer.name,()=>{this.backwardLayer.build(t)}),this.built=!0}computeMask(t,e){Array.isArray(e)&&(e=e[0]);let n;if(this.returnSequences?this.mergeMode==null?n=[e,e]:n=e:this.mergeMode==null?n=[null,null]:n=null,this.returnState){let s=this.forwardLayer.states.map(i=>null);return Array.isArray(n)?n.concat(s).concat(s):[n].concat(s).concat(s)}else return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(t),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(t)}getConfig(){let t={mergeMode:this.mergeMode},e=super.getConfig();return Object.assign(t,e),t}static fromConfig(t,e){let n=gn(e.layer);if(delete e.layer,e.numConstants!=null)throw new St(\"Deserialization of a Bidirectional layer with numConstants present is not supported yet.\");let o=e;return o.layer=n,new t(o)}};Wf.className=\"Bidirectional\";Q.registerClass(Wf);var Uf=class extends $t{constructor(t){super(t),this.scale=t.scale,t.offset?this.offset=t.offset:this.offset=0}getConfig(){let t={scale:this.scale,offset:this.offset},e=super.getConfig();return Object.assign(t,e),t}call(t,e){return B(()=>(t=Nt(t),t.dtype!==\"float32\"&&(t=no(t,\"float32\")),X(D(t,this.scale),this.offset)))}};Uf.className=\"Rescaling\";Q.registerClass(Uf);var aY=[\"bilinear\",\"nearest\"],ID=new Set(aY),Hf=class extends $t{constructor(t){if(super(t),this.height=t.height,this.width=t.width,t.interpolation)if(ID.has(t.interpolation))this.interpolation=t.interpolation;else throw new M(`Invalid interpolation parameter: ${t.interpolation} is not implemented`);else this.interpolation=\"bilinear\";this.cropToAspectRatio=Boolean(t.cropToAspectRatio)}computeOutputShape(t){t=Bt(t);let e=t[2];return[this.height,this.width,e]}getConfig(){let t={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},e=super.getConfig();return Object.assign(t,e),t}call(t,e){return B(()=>{let n=[this.height,this.width];if(this.interpolation===\"bilinear\")return Gs.resizeBilinear(t,n,!this.cropToAspectRatio);if(this.interpolation===\"nearest\")return Gs.resizeNearestNeighbor(t,n,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...ID]} are supported`)})}};Hf.className=\"Resizing\";Q.registerClass(Hf);function SD(r,t,e,n){let o=Nt(r);if(o.dtype!==\"int32\"&&(o=no(o,\"int32\")),t===\"int\")return o;let s=o.shape;if(o.rank===0&&(o=rr(o,-1)),t===\"oneHot\"&&o.shape[o.shape.length-1]!==1&&(o=rr(o,-1)),o.rank>2)throw new M(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${s} which would result in output rank ${o.rank}.`);let i=[\"multiHot\",\"oneHot\"].includes(t),a=o,u;if(typeof n!=\"undefined\"&&t===\"count\"?u=ch(a,n,e,i):u=ch(a,[],e,i),t!==\"tfIdf\")return u;if(n)return D(u,n);throw new M(\"When outputMode is 'tfIdf', weights must be provided.\")}var qf=class extends $t{constructor(t){super(t),this.numTokens=t.numTokens,t.outputMode?this.outputMode=t.outputMode:this.outputMode=\"multiHot\"}getConfig(){let t={numTokens:this.numTokens,outputMode:this.outputMode},e=super.getConfig();return Object.assign(t,e),t}computeOutputShape(t){return t=Bt(t),t==null?[this.numTokens]:this.outputMode===\"oneHot\"&&t[t.length-1]!==1?(t.push(this.numTokens),t):(t[t.length-1]=this.numTokens,t)}call(t,e){return B(()=>{t=Nt(t),t.dtype!==\"int32\"&&(t=no(t,\"int32\"));let n;if(typeof e.countWeights!=\"undefined\"){if(this.outputMode!==\"count\")throw new M(`countWeights is not used when outputMode !== count.\n              Received countWeights=${e.countWeights}`);n=Nt(e.countWeights)}let o=Ir(t),s=Ja(t),i=Re(this.numTokens,o).bufferSync().get(0),a=ln(s,0).bufferSync().get(0);if(!(i&&a))throw new M(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return SD(t,this.outputMode,this.numTokens,n)})}};qf.className=\"CategoryEncoding\";Q.registerClass(qf);function uY(r){return new Ks(r)}function cY(r){return new Ym(r)}function pY(r){return new Km(r)}function mY(r){return new jm(r)}function fY(r){return new Xm(r)}function dY(r){return new Jm(r)}function hY(r){return new Zm(r)}function gY(r){return new Su(r)}function xY(r){return new il(r)}function yY(r){return new Qm(r)}function bY(r){return new al(r)}function wY(r){return new tf(r)}function CY(r){return new ef(r)}function IY(r){return new rf(r)}function SY(r){return new nf(r)}function vY(r){return new of(r)}function NY(r){return new ff(r)}function TY(r){return new pf(r)}function kY(r){return new vc(r)}function EY(r){return new cf(r)}function _Y(r){return new mf(r)}function AY(r){return new df(r)}function $Y(r){return new hf(r)}function DY(r){return new gf(r)}function RY(r){return new yf(r)}function FY(r){return new bf(r)}function OY(r){return new Cf(r)}function PY(r){return new vf(r)}function LY(r){return new If(r)}function MY(r){return new Sf(r)}function zY(r){return new wf(r)}function BY(r){return new Nf(r)}function VY(r){return new _f(r)}function GY(r){return new Af(r)}function WY(r){return new $f(r)}function Gv(r){return new Rf(r)}function UY(r){return Gv(r)}function HY(r){return Gv(r)}function Wv(r){return new Of(r)}function qY(r){return Wv(r)}function KY(r){return Wv(r)}function Uv(r){return new Lf(r)}function jY(r){return Uv(r)}function XY(r){return Uv(r)}function YY(r){return new Mf(r)}function ZY(r){return new Bf(r)}function vD(r){return new zf(r)}function ND(r){return new Vf(r)}function TD(r){return new Df(r)}function kD(r){return new Ff(r)}function JY(r){return new Pf(r)}function QY(r){return new af(r)}function tZ(r){return new Cc(r)}function eZ(r){return new lf(r)}function rZ(r){return new ul(r)}function nZ(r){return new sf(r)}function oZ(r){return new wc(r)}function sZ(r){return new uf(r)}function iZ(r){return new Sc(r)}function aZ(r){return new Tn(r)}function lZ(r){return new Ic(r)}function uZ(r){return new Wf(r)}function cZ(r){return new Gf(r)}var pZ=vD,mZ=ND,fZ=TD,dZ=kD;function hZ(r){return new Tf(r)}function gZ(r){return new kf(r)}function xZ(r){return new Ef(r)}function yZ(r){return new xf(r)}function bZ(r){return new Uf(r)}function wZ(r){return new Hf(r)}function CZ(r){return new qf(r)}var _D={};Wt(_D,{MAPE:()=>DZ,MSE:()=>OZ,binaryAccuracy:()=>IZ,binaryCrossentropy:()=>SZ,categoricalAccuracy:()=>NZ,categoricalCrossentropy:()=>TZ,cosineProximity:()=>_Z,mape:()=>RZ,meanAbsoluteError:()=>AZ,meanAbsolutePercentageError:()=>$Z,meanSquaredError:()=>FZ,mse:()=>PZ,precision:()=>kZ,recall:()=>EZ,sparseCategoricalAccuracy:()=>vZ});function IZ(r,t){return Th(r,t)}function SZ(r,t){return Wy(r,t)}function vZ(r,t){return Uy(r,t)}function NZ(r,t){return kh(r,t)}function TZ(r,t){return Eh(r,t)}function kZ(r,t){return Av(r,t)}function EZ(r,t){return J$(r,t)}function _Z(r,t){return Nh(r,t)}function AZ(r,t){return Vm(r,t)}function $Z(r,t){return bu(r,t)}function DZ(r,t){return bu(r,t)}function RZ(r,t){return bu(r,t)}function FZ(r,t){return Hi(r,t)}function OZ(r,t){return Hi(r,t)}function PZ(r,t){return Hi(r,t)}var AD={};Wt(AD,{modelFromJSON:()=>dD});var $D={};Wt($D,{l1:()=>MZ,l1l2:()=>LZ,l2:()=>zZ});function LZ(r){return new wu(r)}function MZ(r){return yD(r)}function zZ(r){return bD(r)}var Sb=class extends sl{constructor(){super(...arguments),this.model=null}setModel(t){if(!(t instanceof Bn))throw new Error(\"model must be a LayersModel, not some other Container\");this.model=t}};function Ib(r,t){return r<t}function DD(r,t){return r>t}var vb=class extends Sb{constructor(t){if(super(),t==null&&(t={}),t.restoreBestWeights)throw new St(\"restoreBestWeights = True is not implemented in EarlyStopping yet.\");this.monitor=t.monitor||\"val_loss\",this.minDelta=Math.abs(t.minDelta||0),this.patience=t.patience||0,this.verbose=t.verbose||0,this.mode=t.mode||\"auto\",this.baseline=t.baseline,[\"auto\",\"min\",\"max\"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode=\"auto\"),this.mode===\"min\"?this.monitorFunc=Ib:this.mode===\"max\"?this.monitorFunc=DD:this.monitor.indexOf(\"acc\")!==-1?this.monitorFunc=DD:this.monitorFunc=Ib,this.monitorFunc===Ib&&(this.minDelta*=-1)}async onTrainBegin(t){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===Ib?1/0:-1/0}async onEpochEnd(t,e){await Ui(e);let n=this.getMonitorValue(e);n!=null&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=t,this.model.stopTraining=!0)))}async onTrainEnd(t){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(t){t==null&&(t={});let e=t[this.monitor];return e==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(t)}`),e}};function BZ(r){return new vb(r)}var VZ={earlyStopping:BZ};var GZ=z();GZ.registerFlag(\"KEEP_INTERMEDIATE_TENSORS\",()=>!1,r=>{r&&console.warn(\"Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.\")});var oo;(function(r){r[r.DT_INVALID=0]=\"DT_INVALID\",r[r.DT_FLOAT=1]=\"DT_FLOAT\",r[r.DT_DOUBLE=2]=\"DT_DOUBLE\",r[r.DT_INT32=3]=\"DT_INT32\",r[r.DT_UINT8=4]=\"DT_UINT8\",r[r.DT_INT16=5]=\"DT_INT16\",r[r.DT_INT8=6]=\"DT_INT8\",r[r.DT_STRING=7]=\"DT_STRING\",r[r.DT_COMPLEX64=8]=\"DT_COMPLEX64\",r[r.DT_INT64=9]=\"DT_INT64\",r[r.DT_BOOL=10]=\"DT_BOOL\",r[r.DT_QINT8=11]=\"DT_QINT8\",r[r.DT_QUINT8=12]=\"DT_QUINT8\",r[r.DT_QINT32=13]=\"DT_QINT32\",r[r.DT_BFLOAT16=14]=\"DT_BFLOAT16\",r[r.DT_QINT16=15]=\"DT_QINT16\",r[r.DT_QUINT16=16]=\"DT_QUINT16\",r[r.DT_UINT16=17]=\"DT_UINT16\",r[r.DT_COMPLEX128=18]=\"DT_COMPLEX128\",r[r.DT_HALF=19]=\"DT_HALF\",r[r.DT_RESOURCE=20]=\"DT_RESOURCE\",r[r.DT_VARIANT=21]=\"DT_VARIANT\",r[r.DT_UINT32=22]=\"DT_UINT32\",r[r.DT_UINT64=23]=\"DT_UINT64\",r[r.DT_FLOAT_REF=101]=\"DT_FLOAT_REF\",r[r.DT_DOUBLE_REF=102]=\"DT_DOUBLE_REF\",r[r.DT_INT32_REF=103]=\"DT_INT32_REF\",r[r.DT_UINT8_REF=104]=\"DT_UINT8_REF\",r[r.DT_INT16_REF=105]=\"DT_INT16_REF\",r[r.DT_INT8_REF=106]=\"DT_INT8_REF\",r[r.DT_STRING_REF=107]=\"DT_STRING_REF\",r[r.DT_COMPLEX64_REF=108]=\"DT_COMPLEX64_REF\",r[r.DT_INT64_REF=109]=\"DT_INT64_REF\",r[r.DT_BOOL_REF=110]=\"DT_BOOL_REF\",r[r.DT_QINT8_REF=111]=\"DT_QINT8_REF\",r[r.DT_QUINT8_REF=112]=\"DT_QUINT8_REF\",r[r.DT_QINT32_REF=113]=\"DT_QINT32_REF\",r[r.DT_BFLOAT16_REF=114]=\"DT_BFLOAT16_REF\",r[r.DT_QINT16_REF=115]=\"DT_QINT16_REF\",r[r.DT_QUINT16_REF=116]=\"DT_QUINT16_REF\",r[r.DT_UINT16_REF=117]=\"DT_UINT16_REF\",r[r.DT_COMPLEX128_REF=118]=\"DT_COMPLEX128_REF\",r[r.DT_HALF_REF=119]=\"DT_HALF_REF\",r[r.DT_RESOURCE_REF=120]=\"DT_RESOURCE_REF\",r[r.DT_VARIANT_REF=121]=\"DT_VARIANT_REF\",r[r.DT_UINT32_REF=122]=\"DT_UINT32_REF\",r[r.DT_UINT64_REF=123]=\"DT_UINT64_REF\"})(oo||(oo={}));var RD;(function(r){let t;(function(e){e[e.LEGACY=0]=\"LEGACY\",e[e.V1=1]=\"V1\",e[e.V2=2]=\"V2\"})(t=r.CheckpointFormatVersion||(r.CheckpointFormatVersion={}))})(RD||(RD={}));var Hv={};function UZ(r,t){let e={tfOpName:r,category:\"custom\",inputs:[],attrs:[],customExecutor:t};Hv[r]=e}function Nb(r){return Hv[r]}function HZ(r){delete Hv[r]}function S(r,t,e,n,o){let s=t.inputParams[r];if(s&&s.inputIndexStart!==void 0){let a=s.inputIndexStart,u=s.inputIndexEnd===0?void 0:s.inputIndexEnd===void 0?a+1:s.inputIndexEnd;if(s.type===\"tensor\")return br(t.inputNames[s.inputIndexStart],e,n,o);if(s.type===\"tensors\")return t.inputNames.slice(a,u).map(m=>br(m,e,n,o));let l=br(t.inputNames.slice(a)[0],e,n,o),c=l.dataSync();return s.type===\"number\"?c[0]:y.toNestedArray(l.shape,c)}let i=t.attrParams[r];return i&&i.value}function br(r,t,e,n){let[o,s]=xn(r);if(n!=null){let a=n.getHashTableHandleByName(o);if(a!=null)return a}let i=e.currentContextIds.find(a=>!!t[Tb(o,a)]);return i!==void 0?t[Tb(o,i)][s]:void 0}function FD(r,t,e){return t[Tb(r,e.currentContextId)]}function _o(r,t){let[e,n,o]=xn(r);return[Tb(e,t&&t.currentContextId),n,o]}function Tb(r,t){return t?`${r}-${t}`:r}function xn(r){let t=r.split(\":\");if(t.length===1)return[r,0,void 0];let e=t[0],n=t.length===3?t[1]:void 0,o=Number(t[t.length-1]);return[e,o,n]}function Rh(r,t,e){let n=S(\"pad\",r,t,e);if(n===\"explicit\"){n=S(\"explicitPaddings\",r,t,e);let o=[[0,0],[0,0],[0,0],[0,0]];for(let s=0;s<4;s++)o[s][0]=n[s*2],o[s][1]=n[s*2+1];return o}return n}function Zs(r){return r.kept?r:sn(r)}var qv={};Wt(qv,{json:()=>qZ});var qZ=[{tfOpName:\"Add\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"AddV2\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"AddN\",category:\"arithmetic\",inputs:[{start:0,end:0,name:\"tensors\",type:\"tensors\"}]},{tfOpName:\"BiasAdd\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0}]},{tfOpName:\"Sub\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"RealDiv\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Div\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"DivNoNan\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"FloorDiv\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Mul\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Maximum\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Minimum\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Pow\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"SquaredDifference\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Mod\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"FloorMod\",category:\"arithmetic\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]}];var Kv={};Wt(Kv,{json:()=>KZ});var KZ=[{tfOpName:\"Abs\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Acos\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Asin\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Atan\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Atan2\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"y\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Ceil\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"ClipByValue\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"clipValueMin\",type:\"number\"},{start:2,name:\"clipValueMax\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Complex\",category:\"basic_math\",inputs:[{start:0,name:\"real\",type:\"tensor\"},{start:1,name:\"imag\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"ComplexAbs\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Cos\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Cosh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Elu\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Exp\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Floor\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Log\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Imag\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"outputType\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Neg\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Real\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"outputType\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Prelu\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"alpha\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Relu\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Relu6\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Selu\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Sigmoid\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Sin\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Sinh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Sqrt\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Rsqrt\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Square\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Tan\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Tanh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Sign\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Round\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Expm1\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Log1p\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Reciprocal\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Softplus\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Asinh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Acosh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Atanh\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Erf\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Prod\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axes\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\",notSupported:!0},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LeakyRelu\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"alpha\",name:\"alpha\",type:\"number\",defaultValue:.2},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"IsNan\",category:\"basic_math\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]}];var jv={};Wt(jv,{json:()=>jZ});var jZ=[{tfOpName:\"EmptyTensorList\",category:\"control\",inputs:[{start:0,name:\"elementShape\",type:\"shape\"},{start:1,name:\"maxNumElements\",type:\"number\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"LoopCond\",category:\"control\",inputs:[{start:0,name:\"pred\",type:\"tensor\"}]},{tfOpName:\"Switch\",category:\"control\",inputs:[{start:0,name:\"data\",type:\"tensor\"},{start:1,name:\"pred\",type:\"tensor\"}]},{tfOpName:\"Merge\",category:\"control\",inputs:[{start:0,end:0,name:\"tensors\",type:\"tensors\"}]},{tfOpName:\"Enter\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"frame_name\",name:\"frameName\",type:\"string\"},{tfName:\"is_constant\",name:\"isConstant\",type:\"bool\"}]},{tfOpName:\"Exit\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"NextIteration\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"TensorArrayV3\",category:\"control\",inputs:[{start:0,name:\"size\",type:\"number\"}],attrs:[{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"element_shape\",name:\"elementShape\",type:\"shape\"},{tfName:\"dynamic_size\",name:\"dynamicSize\",type:\"bool\"},{tfName:\"clear_after_read\",name:\"clearAfterRead\",type:\"bool\"},{tfName:\"identical_element_shapes\",name:\"identicalElementShapes\",type:\"bool\"},{tfName:\"tensor_array_name\",name:\"name\",type:\"string\"}]},{tfOpName:\"TensorArrayWriteV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"index\",type:\"number\"},{start:2,name:\"tensor\",type:\"tensor\"},{start:3,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"TensorArrayReadV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"index\",type:\"number\"},{start:2,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"dtype\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"TensorArrayGatherV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"indices\",type:\"number[]\"},{start:2,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"element_shape\",name:\"elementShape\",type:\"shape\"}]},{tfOpName:\"TensorArrayScatterV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"indices\",type:\"number[]\"},{start:2,name:\"tensor\",type:\"tensor\"},{start:3,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"TensorArrayConcatV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"element_shape_except0\",name:\"elementShapeExcept0\",type:\"shape\",notSupported:!0}]},{tfOpName:\"TensorArraySplitV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"tensor\",type:\"tensor\"},{start:2,name:\"lengths\",type:\"number[]\"},{start:3,name:\"flowIn\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"TensorArraySizeV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"},{start:1,name:\"flowIn\",type:\"number\"}]},{tfOpName:\"TensorArrayCloseV3\",category:\"control\",inputs:[{start:0,name:\"tensorArrayId\",type:\"tensor\"}]},{tfOpName:\"StatelessIf\",category:\"control\",inputs:[{start:0,name:\"cond\",type:\"tensor\"},{start:1,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"then_branch\",name:\"thenBranch\",type:\"func\"},{tfName:\"else_branch\",name:\"elseBranch\",type:\"func\"}]},{tfOpName:\"If\",category:\"control\",inputs:[{start:0,name:\"cond\",type:\"tensor\"},{start:1,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"then_branch\",name:\"thenBranch\",type:\"func\"},{tfName:\"else_branch\",name:\"elseBranch\",type:\"func\"}]},{tfOpName:\"StatelessWhile\",category:\"control\",inputs:[{start:0,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"cond\",name:\"cond\",type:\"func\"},{tfName:\"body\",name:\"body\",type:\"func\"}]},{tfOpName:\"While\",category:\"control\",inputs:[{start:0,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"cond\",name:\"cond\",type:\"func\"},{tfName:\"body\",name:\"body\",type:\"func\"}]},{tfOpName:\"TensorListScatter\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"},{start:1,name:\"indices\",type:\"number[]\"},{start:2,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListScatterV2\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"},{start:1,name:\"indices\",type:\"number[]\"},{start:2,name:\"elementShape\",type:\"shape\"},{start:3,name:\"numElements\",type:\"number\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListGather\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"indices\",type:\"number[]\"},{start:2,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListGetItem\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"index\",type:\"number\"},{start:2,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListSetItem\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"index\",type:\"number\"},{start:2,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListReserve\",category:\"control\",inputs:[{start:0,name:\"elementShape\",type:\"shape\"},{start:1,name:\"numElements\",type:\"number\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListFromTensor\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"},{start:1,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListStack\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"},{tfName:\"num_elements\",name:\"numElements\",type:\"dtype\"}]},{tfOpName:\"TensorListSplit\",category:\"control\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"},{start:1,name:\"elementShape\",type:\"shape\"},{start:2,name:\"lengths\",type:\"number[]\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListConcat\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"}],attrs:[{tfName:\"element_shape\",name:\"elementShape\",type:\"shape\"},{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListConcatV2\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"}],attrs:[{tfName:\"element_shape\",name:\"elementShape\",type:\"shape\"},{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListPopBack\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"elementShape\",type:\"shape\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListPushBack\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"element_dtype\",name:\"elementDType\",type:\"dtype\"}]},{tfOpName:\"TensorListLength\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"}]},{tfOpName:\"TensorListResize\",category:\"control\",inputs:[{start:0,name:\"tensorListId\",type:\"tensor\"},{start:1,name:\"size\",type:\"number\"}]}];var Xv={};Wt(Xv,{json:()=>XZ});var XZ=[{tfOpName:\"AvgPool\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0},{tfName:\"ksize\",name:\"kernelSize\",type:\"number[]\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"MaxPool\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0},{tfName:\"ksize\",name:\"kernelSize\",type:\"number[]\"},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[],notSupported:!0},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"MaxPoolWithArgmax\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"ksize\",name:\"kernelSize\",type:\"number[]\"},{tfName:\"include_batch_in_index\",name:\"includeBatchInIndex\",type:\"bool\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"AvgPool3D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0},{tfName:\"ksize\",name:\"kernelSize\",type:\"number[]\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"MaxPool3D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0},{tfName:\"ksize\",name:\"kernelSize\",type:\"number[]\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Conv1D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"stride\",name:\"stride\",type:\"number\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NWC\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"dilation\",name:\"dilation\",type:\"number\",defaultValue:1}]},{tfOpName:\"Conv2D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"useCudnnOnGpu\",name:\"useCudnnOnGpu\",type:\"bool\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\"}]},{tfOpName:\"_FusedConv2D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"},{start:2,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"num_args\",name:\"numArgs\",type:\"number\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]},{tfName:\"use_cudnn_on_gpu\",name:\"useCudnnOnGpu\",type:\"bool\",defaultValue:!0},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\",defaultValue:[1,1,1,1]},{tfName:\"fused_ops\",name:\"fusedOps\",type:\"string[]\",defaultValue:[]},{tfName:\"epsilon\",name:\"epsilon\",type:\"number\",defaultValue:1e-4},{tfName:\"leakyrelu_alpha\",name:\"leakyreluAlpha\",type:\"number\",defaultValue:.2}]},{tfOpName:\"Conv2DBackpropInput\",category:\"convolution\",inputs:[{start:2,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"},{start:0,name:\"outputShape\",type:\"number[]\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\",notSupported:!0}]},{tfOpName:\"DepthwiseConv2d\",category:\"convolution\",inputs:[{start:0,name:\"input\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\"}]},{tfOpName:\"DepthwiseConv2dNative\",category:\"convolution\",inputs:[{start:0,name:\"input\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\"}]},{tfOpName:\"FusedDepthwiseConv2dNative\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"},{start:2,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"num_args\",name:\"numArgs\",type:\"number\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\",defaultValue:[1,1,1,1]},{tfName:\"fused_ops\",name:\"fusedOps\",type:\"string[]\",defaultValue:[]},{tfName:\"explicit_paddings\",name:\"explicitPaddings\",type:\"number[]\",defaultValue:[]}]},{tfOpName:\"Conv3D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",defaultValue:\"NHWC\"},{tfName:\"dilations\",name:\"dilations\",type:\"number[]\"}]},{tfOpName:\"Dilation2D\",category:\"convolution\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"filter\",type:\"tensor\"}],attrs:[{tfName:\"strides\",name:\"strides\",type:\"number[]\"},{tfName:\"rates\",name:\"dilations\",type:\"number[]\"},{tfName:\"padding\",name:\"pad\",type:\"string\"}]}];var Yv={};Wt(Yv,{json:()=>YZ});var YZ=[{tfOpName:\"Fill\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"},{start:1,name:\"value\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"LinSpace\",category:\"creation\",inputs:[{start:0,name:\"start\",type:\"number\"},{start:1,name:\"stop\",type:\"number\"},{start:2,name:\"num\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"OneHot\",category:\"creation\",inputs:[{start:0,name:\"indices\",type:\"tensor\"},{start:1,name:\"depth\",type:\"number\"},{start:2,name:\"onValue\",type:\"number\",defaultValue:1},{start:3,name:\"offValue\",type:\"number\",defaultValue:0}],attrs:[{tfName:\"axis\",name:\"axis\",type:\"number\",notSupported:!0},{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"Ones\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"OnesLike\",category:\"creation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"RandomStandardNormal\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"}],attrs:[{tfName:\"seed\",name:\"seed\",type:\"number\",defaultValue:0},{tfName:\"seed2\",name:\"seed2\",type:\"number\",defaultValue:0,notSupported:!0},{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"T\",name:\"T\",type:\"number\",notSupported:!0}]},{tfOpName:\"RandomUniform\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"}],attrs:[{tfName:\"minval\",name:\"minval\",type:\"number\",defaultValue:0},{tfName:\"maxval\",name:\"maxval\",type:\"number\",defaultValue:1},{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"seed\",name:\"seed\",type:\"number\",defaultValue:0},{tfName:\"seed2\",name:\"seed2\",type:\"number\",defaultValue:0,notSupported:!0},{tfName:\"T\",name:\"T\",type:\"number\",notSupported:!0}]},{tfOpName:\"Range\",category:\"creation\",inputs:[{start:0,name:\"start\",type:\"number\"},{start:1,name:\"stop\",type:\"number\"},{start:2,name:\"step\",type:\"number\",defaultValue:0}],attrs:[{tfName:\"Tidx\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"TruncatedNormal\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"}],attrs:[{tfName:\"means\",name:\"mean\",type:\"number\",defaultValue:0},{tfName:\"stddev\",name:\"stdDev\",type:\"number\",defaultValue:1},{tfName:\"seed\",name:\"seed\",type:\"number\"},{tfName:\"seed2\",name:\"seed2\",type:\"number\",defaultValue:0,notSupported:!0},{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"},{tfName:\"T\",name:\"T\",type:\"number\",notSupported:!0}]},{tfOpName:\"Zeros\",category:\"creation\",inputs:[{start:0,name:\"shape\",type:\"number[]\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"ZerosLike\",category:\"creation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"Multinomial\",category:\"creation\",inputs:[{start:0,name:\"logits\",type:\"tensor\"},{start:1,name:\"numSamples\",type:\"number\"}],attrs:[{tfName:\"seed\",name:\"seed\",type:\"number\"},{tfName:\"seed2\",name:\"seed2\",type:\"number\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\"},{tfName:\"output_dtype\",name:\"output_dtype\",type:\"dtype\"}]}];var Zv={};Wt(Zv,{json:()=>ZZ});var ZZ=[{tfOpName:\"NonMaxSuppressionV2\",category:\"dynamic\",inputs:[{start:0,name:\"boxes\",type:\"tensor\"},{start:1,name:\"scores\",type:\"tensor\"},{start:2,name:\"maxOutputSize\",type:\"number\"},{start:3,name:\"iouThreshold\",type:\"number\"}]},{tfOpName:\"NonMaxSuppressionV3\",category:\"dynamic\",inputs:[{start:0,name:\"boxes\",type:\"tensor\"},{start:1,name:\"scores\",type:\"tensor\"},{start:2,name:\"maxOutputSize\",type:\"number\"},{start:3,name:\"iouThreshold\",type:\"number\"},{start:4,name:\"scoreThreshold\",type:\"number\"}]},{tfOpName:\"NonMaxSuppressionV4\",category:\"dynamic\",inputs:[{start:0,name:\"boxes\",type:\"tensor\"},{start:1,name:\"scores\",type:\"tensor\"},{start:2,name:\"maxOutputSize\",type:\"number\"},{start:3,name:\"iouThreshold\",type:\"number\"},{start:4,name:\"scoreThreshold\",type:\"number\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0},{tfName:\"T_threshold\",name:\"threshold\",type:\"dtype\",notSupported:!0},{tfName:\"pad_to_max_output_size\",name:\"padToMaxOutputSize\",type:\"bool\"}]},{tfOpName:\"NonMaxSuppressionV5\",category:\"dynamic\",inputs:[{start:0,name:\"boxes\",type:\"tensor\"},{start:1,name:\"scores\",type:\"tensor\"},{start:2,name:\"maxOutputSize\",type:\"number\"},{start:3,name:\"iouThreshold\",type:\"number\"},{start:4,name:\"scoreThreshold\",type:\"number\"},{start:5,name:\"softNmsSigma\",type:\"number\"}]},{tfOpName:\"Where\",category:\"dynamic\",inputs:[{start:0,name:\"condition\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"ListDiff\",category:\"dynamic\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"y\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]}];var Jv={};Wt(Jv,{json:()=>JZ});var JZ=[{tfOpName:\"LowerBound\",category:\"evaluation\",inputs:[{start:0,name:\"sortedSequence\",type:\"tensor\"},{start:1,name:\"values\",type:\"tensor\"}]},{tfOpName:\"TopKV2\",category:\"evaluation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"k\",type:\"number\"}],attrs:[{tfName:\"sorted\",name:\"sorted\",type:\"bool\"}]},{tfOpName:\"UpperBound\",category:\"evaluation\",inputs:[{start:0,name:\"sortedSequence\",type:\"tensor\"},{start:1,name:\"values\",type:\"tensor\"}]},{tfOpName:\"Unique\",category:\"evaluation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"UniqueV2\",category:\"evaluation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}]}];var Qv={};Wt(Qv,{json:()=>QZ});var QZ=[{tfOpName:\"PlaceholderWithDefault\",category:\"graph\",inputs:[{start:0,name:\"default\",type:\"tensor\"}],attrs:[{tfName:\"shape\",name:\"shape\",type:\"shape\"},{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"Placeholder\",category:\"graph\",attrs:[{tfName:\"shape\",name:\"shape\",type:\"shape\"},{tfName:\"dtype\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"Const\",category:\"graph\"},{tfOpName:\"Identity\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"IdentityN\",category:\"graph\",inputs:[{start:0,end:0,name:\"x\",type:\"tensors\"}]},{tfOpName:\"Snapshot\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"Rank\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"Size\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"Shape\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"ShapeN\",category:\"graph\",inputs:[{start:0,end:0,name:\"x\",type:\"tensors\"}]},{tfOpName:\"Print\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"data\",type:\"tensors\"}],attrs:[{tfName:\"message\",name:\"message\",type:\"string\"},{tfName:\"first_n\",name:\"firstN\",type:\"number\",notSupported:!0},{tfName:\"summarize\",name:\"summarize\",type:\"number\",defaultValue:3}]},{tfOpName:\"NoOp\",category:\"graph\",inputs:[]},{tfOpName:\"StopGradient\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"FakeQuantWithMinMaxVars\",category:\"graph\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"min\",name:\"min\",type:\"number\"},{tfName:\"max\",name:\"max\",type:\"number\"}]}];var tN={};Wt(tN,{json:()=>t7});var t7=[{tfOpName:\"HashTable\",category:\"hash_table\",inputs:[],attrs:[{tfName:\"shared_name\",name:\"sharedName\",type:\"string\"},{tfName:\"use_node_name_sharing\",name:\"useNodeNameSharing\",type:\"bool\"},{tfName:\"key_dtype\",name:\"keyDType\",type:\"dtype\"},{tfName:\"value_dtype\",name:\"valueDType\",type:\"dtype\"}]},{tfOpName:\"HashTableV2\",category:\"hash_table\",inputs:[],attrs:[{tfName:\"shared_name\",name:\"sharedName\",type:\"string\"},{tfName:\"use_node_name_sharing\",name:\"useNodeNameSharing\",type:\"bool\"},{tfName:\"key_dtype\",name:\"keyDType\",type:\"dtype\"},{tfName:\"value_dtype\",name:\"valueDType\",type:\"dtype\"}]},{tfOpName:\"LookupTableImport\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"},{start:1,name:\"keys\",type:\"tensor\"},{start:2,name:\"values\",type:\"tensor\"}],attrs:[{tfName:\"Tin\",name:\"tIn\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"tOut\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LookupTableImportV2\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"},{start:1,name:\"keys\",type:\"tensor\"},{start:2,name:\"values\",type:\"tensor\"}],attrs:[{tfName:\"Tin\",name:\"tIn\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"tOut\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LookupTableFind\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"},{start:1,name:\"keys\",type:\"tensor\"},{start:2,name:\"defaultValue\",type:\"tensor\"}],attrs:[{tfName:\"Tin\",name:\"tIn\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"tOut\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LookupTableFindV2\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"},{start:1,name:\"keys\",type:\"tensor\"},{start:2,name:\"defaultValue\",type:\"tensor\"}],attrs:[{tfName:\"Tin\",name:\"tIn\",type:\"dtype\",notSupported:!0},{tfName:\"Tout\",name:\"tOut\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LookupTableSize\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"}]},{tfOpName:\"LookupTableSizeV2\",category:\"hash_table\",inputs:[{start:0,name:\"tableHandle\",type:\"tensor\"}]}];var eN={};Wt(eN,{json:()=>e7});var e7=[{tfOpName:\"ResizeBilinear\",category:\"image\",inputs:[{start:0,name:\"images\",type:\"tensor\"},{start:1,name:\"size\",type:\"number[]\"}],attrs:[{tfName:\"align_corners\",name:\"alignCorners\",type:\"bool\"},{tfName:\"half_pixel_centers\",name:\"halfPixelCenters\",type:\"bool\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"ResizeNearestNeighbor\",category:\"image\",inputs:[{start:0,name:\"images\",type:\"tensor\"},{start:1,name:\"size\",type:\"number[]\"}],attrs:[{tfName:\"align_corners\",name:\"alignCorners\",type:\"bool\"},{tfName:\"half_pixel_centers\",name:\"halfPixelCenters\",type:\"bool\"},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"CropAndResize\",category:\"image\",inputs:[{start:0,name:\"image\",type:\"tensor\"},{start:1,name:\"boxes\",type:\"tensor\"},{start:2,name:\"boxInd\",type:\"tensor\"},{start:3,name:\"cropSize\",type:\"number[]\"}],attrs:[{tfName:\"method\",name:\"method\",type:\"string\"},{tfName:\"extrapolation_value\",name:\"extrapolationValue\",type:\"number\"}]},{tfOpName:\"ImageProjectiveTransformV3\",category:\"image\",inputs:[{start:0,name:\"images\",type:\"tensor\"},{start:1,name:\"transforms\",type:\"tensor\"},{start:2,name:\"outputShape\",type:\"number[]\"},{start:3,name:\"fillValue\",type:\"number\"}],attrs:[{tfName:\"interpolation\",name:\"interpolation\",type:\"string\"},{tfName:\"fill_mode\",name:\"fillMode\",type:\"string\"}]}];var rN={};Wt(rN,{json:()=>r7});var r7=[{tfOpName:\"Equal\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"NotEqual\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Greater\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"GreaterEqual\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Less\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LessEqual\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LogicalAnd\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LogicalNot\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"LogicalOr\",category:\"logical\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Select\",category:\"logical\",inputs:[{start:0,name:\"condition\",type:\"tensor\"},{start:1,name:\"a\",type:\"tensor\"},{start:2,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"SelectV2\",category:\"logical\",inputs:[{start:0,name:\"condition\",type:\"tensor\"},{start:1,name:\"a\",type:\"tensor\"},{start:2,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]}];var nN={};Wt(nN,{json:()=>n7});var n7=[{tfOpName:\"_FusedMatMul\",category:\"matrices\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"},{start:2,end:0,name:\"args\",type:\"tensors\"}],attrs:[{tfName:\"num_args\",name:\"numArgs\",type:\"number\"},{tfName:\"fused_ops\",name:\"fusedOps\",type:\"string[]\",defaultValue:[]},{tfName:\"epsilon\",name:\"epsilon\",type:\"number\",defaultValue:1e-4},{tfName:\"transpose_a\",name:\"transposeA\",type:\"bool\",defaultValue:!1},{tfName:\"transpose_b\",name:\"transposeB\",type:\"bool\",defaultValue:!1},{tfName:\"leakyrelu_alpha\",name:\"leakyreluAlpha\",type:\"number\",defaultValue:.2},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"MatMul\",category:\"matrices\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"transpose_a\",name:\"transposeA\",type:\"bool\",defaultValue:!1},{tfName:\"transpose_b\",name:\"transposeB\",type:\"bool\",defaultValue:!1},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"BatchMatMul\",category:\"matrices\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"adj_x\",name:\"transposeA\",type:\"bool\",defaultValue:!1},{tfName:\"adj_y\",name:\"transposeB\",type:\"bool\",defaultValue:!1},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"BatchMatMulV2\",category:\"matrices\",inputs:[{start:0,name:\"a\",type:\"tensor\"},{start:1,name:\"b\",type:\"tensor\"}],attrs:[{tfName:\"adj_x\",name:\"transposeA\",type:\"bool\",defaultValue:!1},{tfName:\"adj_y\",name:\"transposeB\",type:\"bool\",defaultValue:!1},{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Transpose\",category:\"matrices\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"perm\",type:\"number[]\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"Einsum\",category:\"matrices\",inputs:[{start:0,end:0,name:\"tensors\",type:\"tensors\"}],attrs:[{tfName:\"equation\",name:\"equation\",type:\"string\"},{tfName:\"N\",name:\"n\",type:\"number\",defaultValue:2},{tfName:\"T\",name:\"dtype\",type:\"dtype\"}]}];var oN={};Wt(oN,{json:()=>o7});var o7=[{tfOpName:\"EuclideanNorm\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\",defaultValue:!1}]},{tfOpName:\"FusedBatchNorm\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"scale\",type:\"tensor\"},{start:2,name:\"offset\",type:\"tensor\"},{start:3,name:\"mean\",type:\"tensor\"},{start:4,name:\"variance\",type:\"tensor\"}],attrs:[{tfName:\"epsilon\",name:\"epsilon\",type:\"number\",defaultValue:.001},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0}]},{tfOpName:\"FusedBatchNormV2\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"scale\",type:\"tensor\"},{start:2,name:\"offset\",type:\"tensor\"},{start:3,name:\"mean\",type:\"tensor\"},{start:4,name:\"variance\",type:\"tensor\"}],attrs:[{tfName:\"epsilon\",name:\"epsilon\",type:\"number\",defaultValue:.001},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0}]},{tfOpName:\"FusedBatchNormV3\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"scale\",type:\"tensor\"},{start:2,name:\"offset\",type:\"tensor\"},{start:3,name:\"mean\",type:\"tensor\"},{start:4,name:\"variance\",type:\"tensor\"}],attrs:[{tfName:\"epsilon\",name:\"epsilon\",type:\"number\",defaultValue:.001},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\",notSupported:!0}]},{tfOpName:\"LRN\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"depth_radius\",name:\"radius\",type:\"number\",defaultValue:5},{tfName:\"bias\",name:\"bias\",type:\"number\",defaultValue:1},{tfName:\"alpha\",name:\"alpha\",type:\"number\",defaultValue:1},{tfName:\"beta\",name:\"beta\",type:\"number\",defaultValue:.5}]},{tfOpName:\"Softmax\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"LogSoftmax\",category:\"normalization\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"SparseToDense\",category:\"normalization\",inputs:[{start:0,name:\"sparseIndices\",type:\"tensor\"},{start:1,name:\"outputShape\",type:\"number[]\"},{start:2,name:\"sparseValues\",type:\"tensor\"},{start:3,name:\"defaultValue\",type:\"tensor\"}],attrs:[{tfName:\"validate_indices\",name:\"validateIndices\",type:\"bool\",defaultValue:!0,notSupported:!0}]}];var sN={};Wt(sN,{json:()=>s7});var s7=[{tfOpName:\"Bincount\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"size\",type:\"number\"},{start:2,name:\"weights\",type:\"tensor\"}]},{tfOpName:\"DenseBincount\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"size\",type:\"number\"},{start:2,name:\"weights\",type:\"tensor\"}],attrs:[{tfName:\"binary_output\",name:\"binaryOutput\",type:\"bool\"}]},{tfOpName:\"Max\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"Mean\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"Min\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"Sum\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"All\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"Any\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"ArgMax\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}]},{tfOpName:\"ArgMin\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}]},{tfOpName:\"Prod\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}],attrs:[{tfName:\"keep_dims\",name:\"keepDims\",type:\"bool\"}]},{tfOpName:\"Cumprod\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}],attrs:[{tfName:\"exclusive\",name:\"exclusive\",type:\"bool\"},{tfName:\"reverse\",name:\"reverse\",type:\"bool\"}]},{tfOpName:\"Cumsum\",category:\"reduction\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}],attrs:[{tfName:\"exclusive\",name:\"exclusive\",type:\"bool\"},{tfName:\"reverse\",name:\"reverse\",type:\"bool\"}]}];var iN={};Wt(iN,{json:()=>i7});var i7=[{tfOpName:\"ConcatV2\",category:\"slice_join\",inputs:[{start:0,end:-1,name:\"tensors\",type:\"tensors\"},{start:-1,name:\"axis\",type:\"number\"}],attrs:[{tfName:\"N\",name:\"n\",type:\"number\",defaultValue:2}]},{tfOpName:\"Concat\",category:\"slice_join\",inputs:[{start:1,end:0,name:\"tensors\",type:\"tensors\"},{start:0,name:\"axis\",type:\"number\"}],attrs:[{tfName:\"N\",name:\"n\",type:\"number\",defaultValue:2}]},{tfOpName:\"GatherV2\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"indices\",type:\"tensor\"},{start:2,name:\"axis\",type:\"number\",defaultValue:0}],attrs:[{tfName:\"batch_dims\",name:\"batchDims\",type:\"number\",defaultValue:0}]},{tfOpName:\"Gather\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"indices\",type:\"tensor\"}],attrs:[{tfName:\"validate_indices\",name:\"validateIndices\",type:\"bool\",notSupported:!0}]},{tfOpName:\"Reverse\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"dims\",type:\"bool[]\"}]},{tfOpName:\"ReverseV2\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number[]\"}]},{tfOpName:\"Slice\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"begin\",type:\"number[]\"},{start:2,name:\"size\",type:\"number[]\"}]},{tfOpName:\"StridedSlice\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"begin\",type:\"number[]\"},{start:2,name:\"end\",type:\"number[]\"},{start:3,name:\"strides\",type:\"number[]\"}],attrs:[{tfName:\"begin_mask\",name:\"beginMask\",type:\"number\",defaultValue:0},{tfName:\"end_mask\",name:\"endMask\",type:\"number\",defaultValue:0},{tfName:\"new_axis_mask\",name:\"newAxisMask\",type:\"number\",defaultValue:0},{tfName:\"ellipsis_mask\",name:\"ellipsisMask\",type:\"number\",defaultValue:0},{tfName:\"shrink_axis_mask\",name:\"shrinkAxisMask\",type:\"number\",defaultValue:0}]},{tfOpName:\"Pack\",category:\"slice_join\",inputs:[{start:0,end:0,name:\"tensors\",type:\"tensors\"}],attrs:[{tfName:\"axis\",name:\"axis\",type:\"number\",defaultValue:0}]},{tfOpName:\"Unpack\",category:\"slice_join\",inputs:[{start:0,name:\"tensor\",type:\"tensor\"}],attrs:[{tfName:\"axis\",name:\"axis\",type:\"number\",defaultValue:0},{tfName:\"num\",name:\"num\",type:\"number\",defaultValue:0,notSupported:!0}]},{tfOpName:\"Tile\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"reps\",type:\"number[]\"}]},{tfOpName:\"Split\",category:\"slice_join\",inputs:[{start:0,name:\"axis\",type:\"number\",defaultValue:0},{start:1,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"num_split\",name:\"numOrSizeSplits\",type:\"number\",defaultValue:1}]},{tfOpName:\"SplitV\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"numOrSizeSplits\",type:\"number[]\"},{start:2,name:\"axis\",type:\"number\",defaultValue:0}]},{tfOpName:\"ScatterNd\",category:\"slice_join\",inputs:[{start:0,name:\"indices\",type:\"tensor\"},{start:1,name:\"values\",type:\"tensor\"},{start:2,name:\"shape\",type:\"number[]\"}]},{tfOpName:\"GatherNd\",category:\"slice_join\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"indices\",type:\"tensor\"}]},{tfOpName:\"SparseToDense\",category:\"slice_join\",inputs:[{start:0,name:\"sparseIndices\",type:\"tensor\"},{start:1,name:\"outputShape\",type:\"number[]\"},{start:2,name:\"sparseValues\",type:\"tensor\"},{start:3,name:\"defaultValue\",type:\"tensor\"}],attrs:[{tfName:\"validate_indices\",name:\"validateIndices\",type:\"bool\",defaultValue:!1,notSupported:!0}]}];var aN={};Wt(aN,{json:()=>a7});var a7=[{tfOpName:\"SparseFillEmptyRows\",category:\"sparse\",inputs:[{start:0,name:\"indices\",type:\"tensor\"},{start:1,name:\"values\",type:\"tensor\"},{start:2,name:\"denseShape\",type:\"tensor\"},{start:3,name:\"defaultValue\",type:\"tensor\"}]},{tfOpName:\"SparseReshape\",category:\"sparse\",inputs:[{start:0,name:\"inputIndices\",type:\"tensor\"},{start:1,name:\"inputShape\",type:\"tensor\"},{start:2,name:\"newShape\",type:\"tensor\"}],attrs:[{tfName:\"T\",name:\"dtype\",type:\"dtype\",notSupported:!0}]},{tfOpName:\"SparseSegmentMean\",category:\"sparse\",inputs:[{start:0,name:\"data\",type:\"tensor\"},{start:1,name:\"indices\",type:\"tensor\"},{start:2,name:\"segmentIds\",type:\"tensor\"}]},{tfOpName:\"SparseSegmentSum\",category:\"sparse\",inputs:[{start:0,name:\"data\",type:\"tensor\"},{start:1,name:\"indices\",type:\"tensor\"},{start:2,name:\"segmentIds\",type:\"tensor\"}]}];var lN={};Wt(lN,{json:()=>l7});var l7=[{tfOpName:\"FFT\",category:\"spectral\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"IFFT\",category:\"spectral\",inputs:[{start:0,name:\"x\",type:\"tensor\"}]},{tfOpName:\"RFFT\",category:\"spectral\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"fft_length\",type:\"number\",notSupported:!0}]},{tfOpName:\"IRFFT\",category:\"spectral\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"fft_length\",type:\"number\",notSupported:!0}]}];var uN={};Wt(uN,{json:()=>u7});var u7=[{tfOpName:\"StringNGrams\",category:\"string\",inputs:[{start:0,name:\"data\",type:\"tensor\"},{start:1,name:\"dataSplits\",type:\"tensor\"}],attrs:[{tfName:\"separator\",name:\"separator\",type:\"string\"},{tfName:\"ngram_widths\",name:\"nGramWidths\",type:\"number[]\"},{tfName:\"left_pad\",name:\"leftPad\",type:\"string\"},{tfName:\"right_pad\",name:\"rightPad\",type:\"string\"},{tfName:\"pad_width\",name:\"padWidth\",type:\"number\"},{tfName:\"preserve_short_sequences\",name:\"preserveShortSequences\",type:\"bool\"}],outputs:[\"ngrams\",\"ngrams_splits\"]},{tfOpName:\"StringSplit\",category:\"string\",inputs:[{start:0,name:\"input\",type:\"tensor\"},{start:1,name:\"delimiter\",type:\"tensor\"}],attrs:[{tfName:\"skip_empty\",name:\"skipEmpty\",type:\"bool\"}],outputs:[\"indices\",\"values\",\"shape\"]},{tfOpName:\"StringToHashBucketFast\",category:\"string\",inputs:[{start:0,name:\"input\",type:\"tensor\"}],attrs:[{tfName:\"num_buckets\",name:\"numBuckets\",type:\"number\"}]}];var cN={};Wt(cN,{json:()=>c7});var c7=[{tfOpName:\"Cast\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"SrcT\",name:\"sdtype\",type:\"dtype\",notSupported:!0},{tfName:\"DstT\",name:\"dtype\",type:\"dtype\"}]},{tfOpName:\"ExpandDims\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"axis\",type:\"number\"}]},{tfOpName:\"MirrorPad\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"padding\",type:\"number[]\"}],attrs:[{tfName:\"mode\",name:\"mode\",type:\"string\"}]},{tfOpName:\"Pad\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"padding\",type:\"number[]\"}],attrs:[{tfName:\"constant_value\",name:\"constantValue\",type:\"number\",defaultValue:0}]},{tfOpName:\"PadV2\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"padding\",type:\"number[]\"},{start:2,name:\"constantValue\",type:\"number\",defaultValue:0}]},{tfOpName:\"Reshape\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"shape\",type:\"number[]\"}]},{tfOpName:\"Squeeze\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"axis\",tfDeprecatedName:\"squeeze_dims\",name:\"axis\",type:\"number[]\"}]},{tfOpName:\"SpaceToBatchND\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"blockShape\",type:\"number[]\"},{start:2,name:\"paddings\",type:\"number[]\"}]},{tfOpName:\"BatchToSpaceND\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"blockShape\",type:\"number[]\"},{start:2,name:\"crops\",type:\"number[]\"}]},{tfOpName:\"DepthToSpace\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"}],attrs:[{tfName:\"block_size\",name:\"blockSize\",type:\"number\"},{tfName:\"data_format\",name:\"dataFormat\",type:\"string\"}]},{tfOpName:\"BroadcastTo\",category:\"transformation\",inputs:[{start:0,name:\"x\",type:\"tensor\"},{start:1,name:\"shape\",type:\"number[]\"}],attrs:[]},{tfOpName:\"BroadcastArgs\",category:\"transformation\",inputs:[{start:0,name:\"s0\",type:\"tensor\"},{start:1,name:\"s1\",type:\"tensor\"}],attrs:[]}];var Fh=class{constructor(){let t=[qv,Kv,jv,Xv,Yv,Zv,Jv,Qv,tN,eN,rN,nN,oN,sN,iN,aN,lN,uN,cN],e=[].concat(...t.map(n=>n.json));this.opMappers=e.reduce((n,o)=>(n[o.tfOpName]=o,n),{})}static get Instance(){return this._instance||(this._instance=new this)}transformGraph(t,e={}){let n=t.node,o=[],s=[],i=[],a=n.reduce((h,g)=>(h[g.name]=this.mapNode(g),g.op.startsWith(\"Placeholder\")?o.push(h[g.name]):g.op===\"Const\"?s.push(h[g.name]):(g.input==null||g.input.length===0)&&i.push(h[g.name]),h),{}),u=[],l=[],c={},p={};e!=null&&(c=this.mapSignatureEntries(e.inputs),p=this.mapSignatureEntries(e.outputs));let m=Object.keys(a);m.forEach(h=>{let g=a[h];g.inputNames.forEach((x,b)=>{let[w,,C]=_o(x),N=a[w];if(N.outputs!=null){let _=N.outputs.indexOf(C);if(_!==-1){let A=`${w}:${_}`;g.inputNames[b]=A}}g.inputs.push(N),N.children.push(g)})}),Object.keys(p).length===0?m.forEach(h=>{let g=a[h];g.children.length===0&&l.push(g)}):Object.keys(p).forEach(h=>{let[g]=_o(h),x=a[g];x!=null&&(x.signatureKey=p[h],l.push(x))}),Object.keys(c).length>0?Object.keys(c).forEach(h=>{let[g]=_o(h),x=a[g];x&&(x.signatureKey=c[h],u.push(x))}):u=o;let f={};t.library!=null&&t.library.function!=null&&(f=t.library.function.reduce((h,g)=>(h[g.signature.name]=this.mapFunction(g),h),{}));let d={nodes:a,inputs:u,outputs:l,weights:s,placeholders:o,signature:e,functions:f};return i.length>0&&(d.initNodes=i),d}mapSignatureEntries(t){return Object.keys(t||{}).reduce((e,n)=>(e[t[n].name]=n,e),{})}mapNode(t){let e=Nb(t.op)||this.opMappers[t.op]||{};t.attr==null&&(t.attr={});let n={name:t.name,op:t.op,category:e.category,inputNames:(t.input||[]).map(o=>o.startsWith(\"^\")?o.slice(1):o),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:t.attr,outputs:e.outputs};return e.inputs!=null&&(n.inputParams=e.inputs.reduce((o,s)=>(o[s.name]={type:s.type,inputIndexStart:s.start,inputIndexEnd:s.end},o),{})),e.attrs!=null&&(n.attrParams=e.attrs.reduce((o,s)=>{let i=s.type,a;switch(s.type){case\"string\":a=kb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=kb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"string[]\":a=Fb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Fb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"number\":a=_b(t.attr,s.tfName,s.defaultValue||0),a===void 0&&!!s.tfDeprecatedName&&(a=_b(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"number[]\":a=Rb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Rb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"bool\":a=Eb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Eb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"bool[]\":a=Pb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Pb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"shape\":a=Db(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Db(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"shape[]\":a=Ob(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Ob(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"dtype\":a=Ab(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Ab(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"dtype[]\":a=$b(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=$b(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"func\":a=OD(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=OD(t.attr,s.tfDeprecatedName,s.defaultValue));break;case\"tensor\":case\"tensors\":break;default:throw new Error(`Unsupported param type: ${s.type} for op: ${t.op}`)}return o[s.name]={value:a,type:i},o},{})),n}mapFunction(t){let e=t.nodeDef,n=[],o=[],s={};e!=null&&(s=e.reduce((p,m)=>(p[m.name]=this.mapNode(m),m.op===\"Const\"&&o.push(p[m.name]),p),{}));let i=[],a=[];t.signature.inputArg.forEach(p=>{let[m]=_o(p.name),f={name:m,op:\"Placeholder\",inputs:[],inputNames:[],category:\"graph\",inputParams:{},attrParams:{dtype:{value:pN(p.type),type:\"dtype\"}},children:[]};f.signatureKey=p.name,i.push(f),s[m]=f}),Object.keys(s).forEach(p=>{let m=s[p];m.inputNames.forEach((f,d)=>{let[h,,g]=_o(f),x=s[h];if(x.outputs!=null){let b=x.outputs.indexOf(g);if(b!==-1){let w=`${h}:${b}`;m.inputNames[d]=w}}m.inputs.push(x),x.children.push(m)})});let l=t.ret;t.signature.outputArg.forEach(p=>{let[m,f]=_o(l[p.name]),d=s[m];d!=null&&(d.defaultOutput=f,a.push(d))});let c=this.mapArgsToSignature(t);return{nodes:s,inputs:i,outputs:a,weights:o,placeholders:n,signature:c}}mapArgsToSignature(t){return{methodName:t.signature.name,inputs:t.signature.inputArg.reduce((e,n)=>(e[n.name]=this.mapArgToTensorInfo(n),e),{}),outputs:t.signature.outputArg.reduce((e,n)=>(e[n.name]=this.mapArgToTensorInfo(n,t.ret),e),{})}}mapArgToTensorInfo(t,e){let n=t.name;return e!=null&&(n=e[n]),{name:n,dtype:t.type}}};function p7(r){let t=z().global;if(typeof t.atob!=\"undefined\")return t.atob(r);if(typeof Buffer!=\"undefined\")return new Buffer(r,\"base64\").toString();throw new Error(\"Unable to decode base64 in this environment. Missing built-in atob() or Buffer()\")}function PD(r,t){let e=Array.isArray(r)?String.fromCharCode.apply(null,r):p7(r);return t?e:e.toLowerCase()}function kb(r,t,e,n=!1){let o=r[t];return o!=null?PD(o.s,n):e}function Eb(r,t,e){let n=r[t];return n?n.b:e}function _b(r,t,e){let n=r[t]||{},o=n.i!=null?n.i:n.f!=null?n.f:e;return typeof o==\"number\"?o:parseInt(o,10)}function pN(r){switch(typeof r==\"string\"&&(r=oo[r]),r){case oo.DT_FLOAT:case oo.DT_HALF:return\"float32\";case oo.DT_INT32:case oo.DT_INT64:case oo.DT_INT8:case oo.DT_UINT8:return\"int32\";case oo.DT_BOOL:return\"bool\";case oo.DT_DOUBLE:return\"float32\";case oo.DT_STRING:return\"string\";default:return null}}function OD(r,t,e){let n=r[t];return n&&n.func?n.func.name:e}function Ab(r,t,e){let n=r[t];return n&&n.type?pN(n.type):e}function $b(r,t,e){let n=r[t];return n&&n.list&&n.list.type?n.list.type.map(o=>pN(o)):e}function LD(r){if(!r.unknownRank)return r.dim!=null?r.dim.map(t=>typeof t.size==\"number\"?t.size:parseInt(t.size,10)):[]}function Db(r,t,e){let n=r[t];return n&&n.shape?LD(n.shape):e}function Rb(r,t,e){let n=r[t];return n?((n.list.f&&n.list.f.length?n.list.f:n.list.i)||[]).map(o=>typeof o==\"number\"?o:parseInt(o,10)):e}function Fb(r,t,e,n=!1){let o=r[t];return o&&o.list&&o.list.s?o.list.s.map(s=>PD(s,n)):e}function Ob(r,t,e){let n=r[t];return n&&n.list&&n.list.shape?n.list.shape.map(o=>LD(o)):e}function Pb(r,t,e){let n=r[t];return n&&n.list&&n.list.b?n.list.b:e}var Lb=class{constructor(t,e,n){this.node=t,this.tensorMap=e,this.context=n,this.inputs=[],this.attrs={},this.inputs=t.inputNames.map(o=>this.getInput(o)),t.rawAttrs!=null&&(this.attrs=Object.keys(t.rawAttrs).reduce((o,s)=>(o[s]=this.getAttr(s),o),{}))}getInput(t){return br(t,this.tensorMap,this.context)}getAttr(t,e){let n=this.node.rawAttrs[t];if(n.tensor!=null)return br(t,this.tensorMap,this.context);if(n.i!=null||n.f!=null)return _b(this.node.rawAttrs,t,e);if(n.s!=null)return kb(this.node.rawAttrs,t,e);if(n.b!=null)return Eb(this.node.rawAttrs,t,e);if(n.shape!=null)return Db(this.node.rawAttrs,t,e);if(n.type!=null)return Ab(this.node.rawAttrs,t,e);if(n.list!=null){if(n.list.i!=null||n.list.f!=null)return Rb(this.node.rawAttrs,t,e);if(n.list.s!=null)return Fb(this.node.rawAttrs,t,e);if(n.list.shape!=null)return Ob(this.node.rawAttrs,t,e);if(n.list.b!=null)return Pb(this.node.rawAttrs,t,e);if(n.list.type!=null)return $b(this.node.rawAttrs,t,e)}return e}};var ae={};Wt(ae,{OP_SCOPE_SUFFIX:()=>k0,abs:()=>Ee,acos:()=>ax,acosh:()=>lx,add:()=>X,addN:()=>LE,all:()=>Zp,any:()=>qu,argMax:()=>Ai,argMin:()=>ux,asin:()=>cx,asinh:()=>px,atan:()=>mx,atan2:()=>fx,atanh:()=>dx,avgPool:()=>Yl,avgPool3d:()=>gx,basicLSTMCell:()=>BE,batchNorm:()=>Di,batchNorm2d:()=>xx,batchNorm3d:()=>yx,batchNorm4d:()=>bx,batchToSpaceND:()=>Zl,bincount:()=>wx,booleanMaskAsync:()=>n6,broadcastArgs:()=>GE,broadcastTo:()=>Ri,buffer:()=>wt,cast:()=>J,ceil:()=>Cx,clipByValue:()=>Cr,clone:()=>sn,complex:()=>wn,concat:()=>ne,concat1d:()=>Ix,concat2d:()=>Sx,concat3d:()=>vx,concat4d:()=>Nx,conv1d:()=>Qp,conv2d:()=>In,conv2dTranspose:()=>em,conv3d:()=>Tx,conv3dTranspose:()=>Ex,cos:()=>Jl,cosh:()=>rm,cosineWindow:()=>hh,cumprod:()=>Xu,cumsum:()=>nm,denseBincount:()=>ch,depthToSpace:()=>_x,depthwiseConv2d:()=>Fi,diag:()=>WE,dilation2d:()=>Ax,div:()=>pt,divNoNan:()=>$x,dot:()=>Dx,dropout:()=>lv,einsum:()=>UE,elu:()=>Oi,enclosingPowerOfTwo:()=>uv,equal:()=>$r,erf:()=>Rx,euclideanNorm:()=>Fx,exp:()=>er,expandDims:()=>rr,expm1:()=>Ox,eye:()=>Yu,fft:()=>au,fill:()=>xo,floor:()=>Pi,floorDiv:()=>Yp,fused:()=>uu,gather:()=>Li,gatherND:()=>m6,greater:()=>Re,greaterEqual:()=>ln,ifft:()=>tl,imag:()=>Xl,image:()=>Gs,inTopKAsync:()=>h6,irfft:()=>xm,isFinite:()=>Px,isInf:()=>Lx,isNaN:()=>Mx,leakyRelu:()=>Ql,less:()=>om,lessEqual:()=>Ln,linalg:()=>pv,linspace:()=>KE,localResponseNormalization:()=>zx,log:()=>Sr,log1p:()=>tu,logSigmoid:()=>Gx,logSoftmax:()=>sm,logSumExp:()=>im,logicalAnd:()=>Rr,logicalNot:()=>eu,logicalOr:()=>am,logicalXor:()=>Wx,losses:()=>hX,lowerBound:()=>jE,matMul:()=>Lt,max:()=>Ir,maxPool:()=>ru,maxPool3d:()=>Hx,maxPoolWithArgmax:()=>XE,maximum:()=>Sn,mean:()=>ve,meshgrid:()=>YE,min:()=>Ja,minimum:()=>Mi,mirrorPad:()=>qx,mod:()=>Kx,moments:()=>Zu,movingAverage:()=>s6,mul:()=>D,multiRNNCell:()=>ZE,multinomial:()=>JE,neg:()=>Ht,norm:()=>Qa,notEqual:()=>Bs,oneHot:()=>Ei,ones:()=>cr,onesLike:()=>yr,op:()=>T,outerProduct:()=>QE,pad:()=>cn,pad1d:()=>t_,pad2d:()=>e_,pad3d:()=>r_,pad4d:()=>n_,pool:()=>jx,pow:()=>an,prelu:()=>ou,print:()=>Jg,prod:()=>Xx,raggedGather:()=>o_,raggedRange:()=>s_,raggedTensorToTensor:()=>i_,rand:()=>a_,randomGamma:()=>v_,randomNormal:()=>tc,randomStandardNormal:()=>N_,randomUniform:()=>zi,range:()=>su,real:()=>Za,reciprocal:()=>ty,relu:()=>Fr,relu6:()=>lm,reshape:()=>R,reverse:()=>pr,reverse1d:()=>T_,reverse2d:()=>k_,reverse3d:()=>E_,reverse4d:()=>__,rfft:()=>lu,round:()=>um,rsqrt:()=>cm,scalar:()=>mt,scatterND:()=>a6,searchSorted:()=>mh,selu:()=>pm,separableConv2d:()=>mm,setdiff1dAsync:()=>A_,sigmoid:()=>Yr,sign:()=>ey,signal:()=>dX,sin:()=>fm,sinh:()=>dm,slice:()=>Rt,slice1d:()=>hm,slice2d:()=>dh,slice3d:()=>gm,slice4d:()=>ec,softmax:()=>iu,softplus:()=>zs,spaceToBatchND:()=>nu,sparse:()=>gX,sparseToDense:()=>c6,spectral:()=>fX,split:()=>mr,sqrt:()=>Se,square:()=>Mt,squaredDifference:()=>ym,squeeze:()=>Mn,stack:()=>nr,step:()=>bo,stridedSlice:()=>ry,string:()=>xX,sub:()=>ct,sum:()=>ft,tan:()=>ny,tanh:()=>$i,tensor:()=>ur,tensor1d:()=>Me,tensor2d:()=>Vs,tensor3d:()=>rx,tensor4d:()=>$_,tensor5d:()=>D_,tensor6d:()=>R_,tile:()=>Dr,topk:()=>oy,transpose:()=>Ot,truncatedNormal:()=>bm,unique:()=>sy,unsortedSegmentSum:()=>wm,unstack:()=>vr,upperBound:()=>F_,variable:()=>iy,where:()=>_e,whereAsync:()=>ly,zeros:()=>Ne,zerosLike:()=>It});var MD=(r,t,e,n=ae)=>{switch(r.op){case\"BiasAdd\":case\"AddV2\":case\"Add\":return[n.add(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"AddN\":return[n.addN(S(\"tensors\",r,t,e))];case\"FloorMod\":case\"Mod\":return[n.mod(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Mul\":return[n.mul(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"RealDiv\":case\"Div\":return[n.div(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"DivNoNan\":return[n.divNoNan(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"FloorDiv\":return[n.floorDiv(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Sub\":return[n.sub(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Minimum\":return[n.minimum(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Maximum\":return[n.maximum(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Pow\":return[n.pow(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"SquaredDifference\":return[n.squaredDifference(S(\"a\",r,t,e),S(\"b\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var zD=(r,t,e,n=ae)=>{switch(r.op){case\"Abs\":case\"ComplexAbs\":return[n.abs(S(\"x\",r,t,e))];case\"Acos\":return[n.acos(S(\"x\",r,t,e))];case\"Acosh\":return[n.acosh(S(\"x\",r,t,e))];case\"Asin\":return[n.asin(S(\"x\",r,t,e))];case\"Asinh\":return[n.asinh(S(\"x\",r,t,e))];case\"Atan\":return[n.atan(S(\"x\",r,t,e))];case\"Atan2\":return[n.atan2(S(\"x\",r,t,e),S(\"y\",r,t,e))];case\"Atanh\":return[n.atanh(S(\"x\",r,t,e))];case\"Ceil\":return[n.ceil(S(\"x\",r,t,e))];case\"Complex\":return[n.complex(S(\"real\",r,t,e),S(\"imag\",r,t,e))];case\"Cos\":return[n.cos(S(\"x\",r,t,e))];case\"Cosh\":return[n.cosh(S(\"x\",r,t,e))];case\"Elu\":return[n.elu(S(\"x\",r,t,e))];case\"Erf\":return[n.erf(S(\"x\",r,t,e))];case\"Exp\":return[n.exp(S(\"x\",r,t,e))];case\"Expm1\":return[n.expm1(S(\"x\",r,t,e))];case\"Floor\":return[n.floor(S(\"x\",r,t,e))];case\"Log\":return[n.log(S(\"x\",r,t,e))];case\"Log1p\":return[n.log1p(S(\"x\",r,t,e))];case\"Imag\":return[n.imag(S(\"x\",r,t,e))];case\"Neg\":return[n.neg(S(\"x\",r,t,e))];case\"Reciprocal\":return[n.reciprocal(S(\"x\",r,t,e))];case\"Real\":return[n.real(S(\"x\",r,t,e))];case\"Relu\":return[n.relu(S(\"x\",r,t,e))];case\"Round\":return[n.round(S(\"x\",r,t,e))];case\"Selu\":return[n.selu(S(\"x\",r,t,e))];case\"Sigmoid\":return[n.sigmoid(S(\"x\",r,t,e))];case\"Sin\":return[n.sin(S(\"x\",r,t,e))];case\"Sign\":return[n.sign(S(\"x\",r,t,e))];case\"Sinh\":return[n.sinh(S(\"x\",r,t,e))];case\"Softplus\":return[n.softplus(S(\"x\",r,t,e))];case\"Sqrt\":return[n.sqrt(S(\"x\",r,t,e))];case\"Square\":return[n.square(S(\"x\",r,t,e))];case\"Tanh\":return[n.tanh(S(\"x\",r,t,e))];case\"Tan\":return[n.tan(S(\"x\",r,t,e))];case\"ClipByValue\":return[n.clipByValue(S(\"x\",r,t,e),S(\"clipValueMin\",r,t,e),S(\"clipValueMax\",r,t,e))];case\"Relu6\":return[n.relu6(S(\"x\",r,t,e))];case\"Rsqrt\":return[n.rsqrt(br(r.inputNames[0],t,e))];case\"Prod\":return[n.prod(S(\"x\",r,t,e),S(\"axes\",r,t,e))];case\"LeakyRelu\":return[n.leakyRelu(S(\"x\",r,t,e),S(\"alpha\",r,t,e))];case\"Prelu\":return[n.prelu(S(\"x\",r,t,e),S(\"alpha\",r,t,e))];case\"IsNan\":return[n.isNaN(br(r.inputNames[0],t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function Vn(r,t,e=\"\"){if(!(typeof r==\"number\"||typeof t==\"number\")){y.assert(r.length===t.length,()=>e+` Shapes ${r} and ${t} must match`);for(let n=0;n<r.length;n++){let o=r[n],s=t[n];y.assert(o<0||s<0||o===s,()=>e+` Shapes ${r} and ${t} must match`)}}}function BD(r){return!(typeof r==\"number\"||r.some(t=>t<0))}function Kf(r,t,e){let n=Mb(r,e),o=!BD(n);if(o&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${n}`);if(o&&t.forEach(s=>{n=Mb(s.shape,n)}),!BD(n))throw new Error(`Non-fully-defined elementShape: ${n}`);return n}function Mb(r,t){if(typeof r==\"number\")return t;if(typeof t==\"number\")return r;if(r.length!==t.length)throw new Error(`Incompatible ranks during merge: ${r} vs. ${t}`);let e=[];for(let n=0;n<r.length;++n){let o=r[n],s=t[n];if(o>=0&&s>=0&&o!==s)throw new Error(`Incompatible shape during merge: ${r} vs. ${t}`);e[n]=o>=0?o:s}return e}var zb=class{constructor(t,e,n,o,s,i,a){this.name=t,this.dtype=e,this.maxSize=n,this.elementShape=o,this.identicalElementShapes=s,this.dynamicSize=i,this.clearAfterRead=a,this.tensors=[],this.closed_=!1,this.idTensor=mt(0),De(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(t){this.tensors.forEach(e=>{(t==null||!t.has(e.tensor.id))&&e.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(t<0||t>=this.size())throw new Error(`Tried to read from index ${t}, but array size is: ${this.size()}`);let e=this.tensors[t];if(e.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${t} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(e.cleared=!0),e.read=!0,e.tensor}readMany(t){return t.map(e=>this.read(e))}write(t,e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(t<0||!this.dynamicSize&&t>=this.maxSize)throw new Error(`Tried to write to index ${t}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[t]||{};if(e.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t},\n          because the value dtype is ${e.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=e.shape),Vn(this.elementShape,e.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${t}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t}, because it has already been written.`);n.tensor=e,De(e),n.written=!0,this.tensors[t]=n}writeMany(t,e){if(t.length!==e.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${t.length} is not the same as tensors size: ${e.length}.`);t.forEach((n,o)=>this.write(n,e[o]))}gather(t,e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${e}`);if(t)t=t.slice(0,this.size());else{t=[];for(let o=0;o<this.size();o++)t.push(o)}if(t.length===0)return ur([],[0].concat(this.elementShape));let n=this.readMany(t);return Vn(this.elementShape,n[0].shape,\"TensorArray shape mismatch: \"),nr(n,0)}concat(t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${t}`);if(this.size()===0)return ur([],[0].concat(this.elementShape));let e=[];for(let o=0;o<this.size();o++)e.push(o);let n=this.readMany(e);return Vn(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),ne(n,0)}scatter(t,e){if(e.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${e.dtype}`);if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let n=Math.max(...t);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(t,vr(e,0))}split(t,e){if(e.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${e.dtype}`);let n=0,o=t.map(u=>(n+=u,n));if(n!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${n}, and tensor's shape is: ${e.shape}`);if(!this.dynamicSize&&t.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${t.length}), and the TensorArray is not marked as dynamically resizeable`);let s=n===0?0:e.size/n,i=[];B(()=>{e=R(e,[1,n,s]);for(let u=0;u<t.length;++u){let l=u===0?0:o[u-1],c=[0,l,0],p=[1,t[u],s];i[u]=R(Rt(e,c,p),this.elementShape)}return i});let a=[];for(let u=0;u<t.length;u++)a[u]=u;this.writeMany(a,i)}};var ml=class{constructor(t,e,n,o=-1){this.tensors=t,this.elementShape=e,this.elementDtype=n,t!=null&&t.forEach(s=>{if(n!==s.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${s.dtype}`);Vn(e,s.shape,\"TensorList shape mismatch: \"),De(s)}),this.idTensor=mt(0),this.maxNumElements=o,De(this.idTensor)}get id(){return this.idTensor.id}copy(){return new ml([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(t){this.tensors.forEach(e=>{(t==null||!t.has(e.id))&&e.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(t,e,n=-1){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);if(n!==-1&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);Vn(t,this.elementShape,\"TensorList shape mismatch: \");let o=Kf(this.elementShape,this.tensors,t);return B(()=>{let s=this.tensors.map(i=>R(i,o));return nr(s,0)})}popBack(t,e){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error(\"Trying to pop from an empty list.\");let n=Kf(this.elementShape,this.tensors,t),o=this.tensors.pop();return o.kept=!1,Vn(o.shape,t,\"TensorList shape mismatch: \"),R(o,n)}pushBack(t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(Vn(t.shape,this.elementShape,\"TensorList shape mismatch: \"),this.maxNumElements===this.size())throw new Error(\"Trying to push element into a full list.\");De(t),this.tensors.push(t)}resize(t){if(t<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${t}`);if(this.maxNumElements!==-1&&t>this.maxNumElements)throw new Error(`TensorListResize input size ${t} is greater maxNumElement ${this.maxNumElements}.`);let e=new ml([],this.elementShape,this.elementDtype,this.maxNumElements);e.tensors.length=t;for(let n=0;n<Math.min(this.tensors.length,t);++n)e.tensors[n]=this.tensors[n];return e}getItem(t,e,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(t<0||t>this.tensors.length)throw new Error(`Trying to access element ${t} in a list with ${this.tensors.length} elements.`);if(this.tensors[t]==null)throw new Error(`element at index ${t} is null.`);Vn(this.tensors[t].shape,e,\"TensorList shape mismatch: \");let o=Kf(this.elementShape,this.tensors,e);return R(this.tensors[t],o)}setItem(t,e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(t<0||this.maxNumElements!==-1&&t>=this.maxNumElements)throw new Error(`Trying to set element ${t} in a list with max ${this.maxNumElements} elements.`);Vn(this.elementShape,e.shape,\"TensorList shape mismatch: \"),De(e),this.tensors[t]!=null&&(this.tensors[t].kept=!1),this.tensors[t]=e}gather(t,e,n){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);Vn(this.elementShape,n,\"TensorList shape mismatch: \"),t=t.slice(0,this.size());let o=Kf(this.elementShape,this.tensors,n);return t.length===0?ur([],[0].concat(o)):B(()=>{let s=t.map(i=>R(this.tensors[i],o));return nr(s,0)})}concat(t,e){if(!!t&&t!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${t}`);Vn(this.elementShape,e,\"TensorList shape mismatch: \");let n=Kf(this.elementShape,this.tensors,e);return this.size()===0?ur([],[0].concat(n)):B(()=>{let o=this.tensors.map(s=>R(s,n));return ne(o,0)})}};function VD(r,t,e){let n=r.dtype;if(r.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${r.shape}`);if(r.dtype!==e)throw new Error(`Invalid data types; op elements ${r.dtype}, but list elements ${e}`);let o=r.shape.slice(1);Vn(o,t,\"TensorList shape mismatch: \");let s=vr(r);return new ml(s,t,n)}function GD(r,t,e,n){return new ml([],r,t,n)}function WD(r,t,e,n){if(t.length!==r.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${r.shape[0]}`);let o=Math.max(...t);if(n!=null&&n!==-1&&o>=n)throw new Error(`Max index must be < array size (${o}  vs. ${n})`);let s=new ml([],e,r.dtype,n),i=vr(r,0);return t.forEach((a,u)=>{s.setItem(a,i[u])}),s}function UD(r,t,e){let n=0,o=t.map(c=>(n+=c,n));if(n!==r.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${n}, and tensor's shape is: ${r.shape}`);let s=r.shape.slice(1),i=Mb(s,e),a=n===0?0:r.size/n,u=B(()=>{let c=[];r=R(r,[1,n,a]);for(let p=0;p<t.length;++p){let m=p===0?0:o[p-1],f=[0,m,0],d=[1,t[p],a];c[p]=R(Rt(r,f,d),i)}return r.dispose(),c}),l=new ml([],e,r.dtype,t.length);for(let c=0;c<u.length;c++)l.setItem(c,u[c]);return l}var HD=async(r,t,e)=>{switch(r.op){case\"If\":case\"StatelessIf\":{let n=S(\"thenBranch\",r,t,e),o=S(\"elseBranch\",r,t,e),s=S(\"cond\",r,t,e),i=S(\"args\",r,t,e);return(await s.data())[0]?e.functionMap[n].executeFunctionAsync(i,e.tensorArrayMap,e.tensorListMap):e.functionMap[o].executeFunctionAsync(i,e.tensorArrayMap,e.tensorListMap)}case\"While\":case\"StatelessWhile\":{let n=S(\"body\",r,t,e),o=S(\"cond\",r,t,e),s=S(\"args\",r,t,e),i=await e.functionMap[o].executeFunctionAsync(s,e.tensorArrayMap,e.tensorListMap),a=s.map(c=>c.id),u=await i[0].data();i.forEach(c=>{!c.kept&&a.indexOf(c.id)===-1&&c.dispose()});let l=s;for(;u[0];){let c=l;l=await e.functionMap[n].executeFunctionAsync(l,e.tensorArrayMap,e.tensorListMap);let p=l.map(f=>f.id);c.forEach(f=>{!f.kept&&a.indexOf(f.id)===-1&&p.indexOf(f.id)===-1&&f.dispose()});let m=await e.functionMap[o].executeFunctionAsync(l,e.tensorArrayMap,e.tensorListMap);u=await m[0].data(),m.forEach(f=>{!f.kept&&a.indexOf(f.id)===-1&&p.indexOf(f.id)===-1&&f.dispose()})}return l}case\"LoopCond\":{let n=S(\"pred\",r,t,e);return[Zs(n)]}case\"Switch\":{let n=S(\"pred\",r,t,e),o=S(\"data\",r,t,e);return o.kept||(o=Zs(o)),(await n.data())[0]?[void 0,o]:[o,void 0]}case\"Merge\":{let n=r.inputNames.find(o=>br(o,t,e)!==void 0);if(n){let o=br(n,t,e);return[Zs(o)]}return}case\"Enter\":{let n=S(\"frameName\",r,t,e),o=S(\"tensor\",r,t,e);return e.enterFrame(n),[Zs(o)]}case\"Exit\":{let n=S(\"tensor\",r,t,e);return e.exitFrame(),[Zs(n)]}case\"NextIteration\":{let n=S(\"tensor\",r,t,e);return e.nextIteration(),[Zs(n)]}case\"TensorArrayV3\":{let n=S(\"size\",r,t,e),o=S(\"dtype\",r,t,e),s=S(\"elementShape\",r,t,e),i=S(\"dynamicSize\",r,t,e),a=S(\"clearAfterRead\",r,t,e),u=S(\"identicalElementShapes\",r,t,e),l=S(\"name\",r,t,e),c=new zb(l,o,n,s,u,i,a);return e.addTensorArray(c),[c.idTensor,mt(1)]}case\"TensorArrayWriteV3\":{let n=S(\"tensorArrayId\",r,t,e),o=S(\"index\",r,t,e),s=S(\"tensor\",r,t,e),i=e.getTensorArray(n.id);return i.write(o,s),[i.idTensor]}case\"TensorArrayReadV3\":{let n=S(\"tensorArrayId\",r,t,e),o=S(\"index\",r,t,e);return[e.getTensorArray(n.id).read(o)]}case\"TensorArrayGatherV3\":{let n=S(\"tensorArrayId\",r,t,e),o=S(\"indices\",r,t,e),s=S(\"dtype\",r,t,e);return[e.getTensorArray(n.id).gather(o,s)]}case\"TensorArrayScatterV3\":{let n=S(\"tensorArrayId\",r,t,e),o=S(\"indices\",r,t,e),s=S(\"tensor\",r,t,e),i=e.getTensorArray(n.id);return i.scatter(o,s),[i.idTensor]}case\"TensorArrayConcatV3\":{let n=S(\"tensorArrayId\",r,t,e),o=e.getTensorArray(n.id),s=S(\"dtype\",r,t,e);return[o.concat(s)]}case\"TensorArraySplitV3\":{let n=S(\"tensorArrayId\",r,t,e),o=S(\"tensor\",r,t,e),s=S(\"lengths\",r,t,e),i=e.getTensorArray(n.id);return i.split(s,o),[i.idTensor]}case\"TensorArraySizeV3\":{let n=S(\"tensorArrayId\",r,t,e),o=e.getTensorArray(n.id);return[mt(o.size(),\"int32\")]}case\"TensorArrayCloseV3\":{let n=S(\"tensorArrayId\",r,t,e),o=e.getTensorArray(n.id);return o.clearAndClose(),[o.idTensor]}case\"TensorListSetItem\":{let n=S(\"tensorListId\",r,t,e),o=S(\"index\",r,t,e),s=S(\"tensor\",r,t,e),i=e.getTensorList(n.id);return i.setItem(o,s),[i.idTensor]}case\"TensorListGetItem\":{let n=S(\"tensorListId\",r,t,e),o=S(\"index\",r,t,e),s=S(\"elementShape\",r,t,e),i=S(\"elementDType\",r,t,e);return[e.getTensorList(n.id).getItem(o,s,i)]}case\"TensorListScatterV2\":case\"TensorListScatter\":{let n=S(\"indices\",r,t,e),o=S(\"tensor\",r,t,e),s=S(\"elementShape\",r,t,e),i=S(\"numElements\",r,t,e),a=WD(o,n,s,i);return e.addTensorList(a),[a.idTensor]}case\"TensorListReserve\":case\"EmptyTensorList\":{let n=S(\"elementShape\",r,t,e),o=S(\"elementDType\",r,t,e),s;r.op===\"TensorListReserve\"?s=\"numElements\":s=\"maxNumElements\";let i=S(s,r,t,e),a=r.op===\"TensorListReserve\"?-1:i,u=GD(n,o,i,a);return e.addTensorList(u),[u.idTensor]}case\"TensorListGather\":{let n=S(\"tensorListId\",r,t,e),o=S(\"indices\",r,t,e),s=S(\"elementShape\",r,t,e),i=S(\"elementDType\",r,t,e);return[e.getTensorList(n.id).gather(o,i,s)]}case\"TensorListStack\":{let n=S(\"tensorListId\",r,t,e),o=S(\"elementShape\",r,t,e),s=S(\"elementDType\",r,t,e),i=S(\"numElements\",r,t,e);return[e.getTensorList(n.id).stack(o,s,i)]}case\"TensorListFromTensor\":{let n=S(\"tensor\",r,t,e),o=S(\"elementShape\",r,t,e),s=S(\"elementDType\",r,t,e),i=VD(n,o,s);return e.addTensorList(i),[i.idTensor]}case\"TensorListConcat\":case\"TensorListConcatV2\":{let n=S(\"tensorListId\",r,t,e),o=e.getTensorList(n.id),s=S(\"dtype\",r,t,e),i=S(\"elementShape\",r,t,e);return[o.concat(s,i)]}case\"TensorListPushBack\":{let n=S(\"tensorListId\",r,t,e),o=S(\"tensor\",r,t,e),s=e.getTensorList(n.id);return s.pushBack(o),[s.idTensor]}case\"TensorListPopBack\":{let n=S(\"tensorListId\",r,t,e),o=S(\"elementShape\",r,t,e),s=S(\"elementDType\",r,t,e);return[e.getTensorList(n.id).popBack(o,s)]}case\"TensorListSplit\":{let n=S(\"tensor\",r,t,e),o=S(\"elementShape\",r,t,e),s=S(\"lengths\",r,t,e),i=UD(n,s,o);return e.addTensorList(i),[i.idTensor]}case\"TensorListLength\":{let n=S(\"tensorListId\",r,t,e),o=e.getTensorList(n.id);return[mt(o.size(),\"int32\")]}case\"TensorListResize\":{let n=S(\"tensorListId\",r,t,e),o=S(\"size\",r,t,e),i=e.getTensorList(n.id).resize(o);return e.addTensorList(i),[i.idTensor]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};function qD(r,t,e){let[n,o]=S(\"fusedOps\",r,t,e),s=n===\"biasadd\",i=!s,a=o===\"prelu\",u=n===\"fusedbatchnorm\",l=S(\"numArgs\",r,t,e);if(s){if(a&&l!==2)throw new Error(\"FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.\");if(!a&&s&&l!==1)throw new Error(\"FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.\")}if(u)throw new Error(\"FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported\");let c=S(\"strides\",r,t,e),p=Rh(r,t,e),m=S(\"dataFormat\",r,t,e).toUpperCase(),f=S(\"dilations\",r,t,e),[d,h]=S(\"args\",r,t,e);i&&(h=d,d=void 0);let g=S(\"leakyreluAlpha\",r,t,e);return{stride:c,pad:p,dataFormat:m,dilations:f,biasArg:d,preluArg:h,activationFunc:o,leakyreluAlpha:g}}var KD=(r,t,e,n=ae)=>{switch(r.op){case\"Conv1D\":{let o=S(\"stride\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"dataFormat\",r,t,e).toUpperCase(),a=S(\"dilation\",r,t,e);return[n.conv1d(S(\"x\",r,t,e),S(\"filter\",r,t,e),o,s,i,a)]}case\"Conv2D\":{let o=S(\"strides\",r,t,e),s=Rh(r,t,e),i=S(\"dataFormat\",r,t,e).toUpperCase(),a=S(\"dilations\",r,t,e);return[n.conv2d(S(\"x\",r,t,e),S(\"filter\",r,t,e),[o[1],o[2]],s,i,[a[1],a[2]])]}case\"_FusedConv2D\":{let{stride:o,pad:s,dataFormat:i,dilations:a,biasArg:u,preluArg:l,activationFunc:c,leakyreluAlpha:p}=qD(r,t,e);return[n.fused.conv2d({x:S(\"x\",r,t,e),filter:S(\"filter\",r,t,e),strides:[o[1],o[2]],pad:s,dataFormat:i,dilations:[a[1],a[2]],bias:u,activation:c,preluActivationWeights:l,leakyreluAlpha:p})]}case\"FusedDepthwiseConv2dNative\":{let{stride:o,pad:s,dataFormat:i,dilations:a,biasArg:u,preluArg:l,activationFunc:c,leakyreluAlpha:p}=qD(r,t,e);return[n.fused.depthwiseConv2d({x:S(\"x\",r,t,e),filter:S(\"filter\",r,t,e),strides:[o[1],o[2]],pad:s,dataFormat:i,dilations:[a[1],a[2]],bias:u,activation:c,preluActivationWeights:l,leakyreluAlpha:p})]}case\"Conv2DBackpropInput\":case\"Conv2dTranspose\":{let o=S(\"outputShape\",r,t,e),s=S(\"strides\",r,t,e),i=Rh(r,t,e);return[n.conv2dTranspose(S(\"x\",r,t,e),S(\"filter\",r,t,e),o,[s[1],s[2]],i)]}case\"DepthwiseConv2dNative\":case\"DepthwiseConv2d\":{let o=S(\"strides\",r,t,e),s=Rh(r,t,e),i=S(\"dilations\",r,t,e),a=S(\"dataFormat\",r,t,e).toUpperCase();return[n.depthwiseConv2d(S(\"input\",r,t,e),S(\"filter\",r,t,e),[o[1],o[2]],s,a,[i[1],i[2]])]}case\"Conv3D\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"dataFormat\",r,t,e).toUpperCase(),a=S(\"dilations\",r,t,e);return[n.conv3d(S(\"x\",r,t,e),S(\"filter\",r,t,e),[o[1],o[2],o[3]],s,i,[a[1],a[2],a[3]])]}case\"AvgPool\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"kernelSize\",r,t,e);return[n.avgPool(S(\"x\",r,t,e),[i[1],i[2]],[o[1],o[2]],s)]}case\"MaxPool\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"kernelSize\",r,t,e);return[n.maxPool(S(\"x\",r,t,e),[i[1],i[2]],[o[1],o[2]],s)]}case\"MaxPoolWithArgmax\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"kernelSize\",r,t,e),a=S(\"includeBatchInIndex\",r,t,e),{result:u,indexes:l}=n.maxPoolWithArgmax(S(\"x\",r,t,e),[i[1],i[2]],[o[1],o[2]],s,a);return[u,l]}case\"AvgPool3D\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"kernelSize\",r,t,e);return[n.avgPool3d(S(\"x\",r,t,e),[i[1],i[2],i[3]],[o[1],o[2],o[3]],s)]}case\"MaxPool3D\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"kernelSize\",r,t,e);return[n.maxPool3d(S(\"x\",r,t,e),[i[1],i[2],i[3]],[o[1],o[2],o[3]],s)]}case\"Dilation2D\":{let o=S(\"strides\",r,t,e),s=S(\"pad\",r,t,e),i=S(\"dilations\",r,t,e),a=o[1],u=o[2],l=i[1],c=i[2];return[n.dilation2d(S(\"x\",r,t,e),S(\"filter\",r,t,e),[a,u],s,[l,c],\"NHWC\")]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var jD=(r,t,e,n=ae)=>{switch(r.op){case\"Fill\":{let o=S(\"shape\",r,t,e),s=S(\"dtype\",r,t,e),i=S(\"value\",r,t,e);return[n.fill(o,i,s)]}case\"LinSpace\":{let o=S(\"start\",r,t,e),s=S(\"stop\",r,t,e),i=S(\"num\",r,t,e);return[n.linspace(o,s,i)]}case\"Multinomial\":{let o=S(\"logits\",r,t,e),s=S(\"numSamples\",r,t,e),i=S(\"seed\",r,t,e);return[n.multinomial(o,s,i)]}case\"OneHot\":{let o=S(\"indices\",r,t,e),s=S(\"depth\",r,t,e),i=S(\"onValue\",r,t,e),a=S(\"offValue\",r,t,e),u=S(\"dtype\",r,t,e);return[n.oneHot(o,s,i,a,u)]}case\"Ones\":return[n.ones(S(\"shape\",r,t,e),S(\"dtype\",r,t,e))];case\"OnesLike\":return[n.onesLike(S(\"x\",r,t,e))];case\"RandomStandardNormal\":return[n.randomStandardNormal(S(\"shape\",r,t,e),S(\"dtype\",r,t,e),S(\"seed\",r,t,e))];case\"RandomUniform\":return[n.randomUniform(S(\"shape\",r,t,e),S(\"minval\",r,t,e),S(\"maxval\",r,t,e),S(\"dtype\",r,t,e))];case\"Range\":{let o=S(\"start\",r,t,e),s=S(\"stop\",r,t,e),i=S(\"step\",r,t,e);return[n.range(o,s,i,S(\"dtype\",r,t,e))]}case\"TruncatedNormal\":{let o=S(\"shape\",r,t,e),s=S(\"mean\",r,t,e),i=S(\"stdDev\",r,t,e),a=S(\"seed\",r,t,e);return[n.truncatedNormal(o,s,i,S(\"dtype\",r,t,e),a)]}case\"Zeros\":return[n.zeros(S(\"shape\",r,t,e),S(\"dtype\",r,t,e))];case\"ZerosLike\":return[n.zerosLike(S(\"x\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function mN(r,t,e){let n=S(\"boxes\",r,t,e),o=S(\"scores\",r,t,e),s=S(\"maxOutputSize\",r,t,e),i=S(\"iouThreshold\",r,t,e),a=S(\"scoreThreshold\",r,t,e),u=S(\"softNmsSigma\",r,t,e);return{boxes:n,scores:o,maxOutputSize:s,iouThreshold:i,scoreThreshold:a,softNmsSigma:u}}var XD=async(r,t,e,n,o=ae)=>{switch(r.op){case\"NonMaxSuppressionV5\":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l,softNmsSigma:c}=mN(r,t,e),p=await o.image.nonMaxSuppressionWithScoreAsync(s,i,a,u,l,c);return[p.selectedIndices,p.selectedScores]}case\"NonMaxSuppressionV4\":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l}=mN(r,t,e),c=S(\"padToMaxOutputSize\",r,t,e),p=await o.image.nonMaxSuppressionPaddedAsync(s,i,a,u,l,c);return[p.selectedIndices,p.validOutputs]}case\"NonMaxSuppressionV3\":case\"NonMaxSuppressionV2\":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l}=mN(r,t,e);return[await o.image.nonMaxSuppressionAsync(s,i,a,u,l)]}case\"Where\":{let s=o.cast(S(\"condition\",r,t,e),\"bool\"),i=[await o.whereAsync(s)];return s.dispose(),i}case\"ListDiff\":return o.setdiff1dAsync(S(\"x\",r,t,e),S(\"y\",r,t,e));default:throw TypeError(`Node type ${r.op} is not implemented`)}};var YD=(r,t,e,n=ae)=>{switch(r.op){case\"LowerBound\":{let o=S(\"sortedSequence\",r,t,e),s=S(\"values\",r,t,e);return[n.lowerBound(o,s)]}case\"TopKV2\":{let o=S(\"x\",r,t,e),s=S(\"k\",r,t,e),i=S(\"sorted\",r,t,e),a=n.topk(o,s,i);return[a.values,a.indices]}case\"UpperBound\":{let o=S(\"sortedSequence\",r,t,e),s=S(\"values\",r,t,e);return[n.upperBound(o,s)]}case\"Unique\":{let o=S(\"x\",r,t,e),s=n.unique(o);return[s.values,s.indices]}case\"UniqueV2\":{let o=S(\"x\",r,t,e),s=S(\"axis\",r,t,e),i=n.unique(o,s);return[i.values,i.indices]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var ZD=(r,t,e,n=ae)=>{switch(r.op){case\"Const\":return t[r.name];case\"PlaceholderWithDefault\":let o=S(\"default\",r,t,e);return[br(r.name,t,e)||o];case\"Placeholder\":return[br(r.name,t,e)];case\"Identity\":case\"StopGradient\":case\"FakeQuantWithMinMaxVars\":{let c=S(\"x\",r,t,e);return[Zs(c)]}case\"IdentityN\":return S(\"x\",r,t,e).map(c=>Zs(c));case\"Snapshot\":let s=S(\"x\",r,t,e);return[Zs(s)];case\"Shape\":return[n.tensor1d(S(\"x\",r,t,e).shape,\"int32\")];case\"ShapeN\":return S(\"x\",r,t,e).map(c=>n.tensor1d(c.shape));case\"Size\":return[n.scalar(S(\"x\",r,t,e).size,\"int32\")];case\"Rank\":return[n.scalar(S(\"x\",r,t,e).rank,\"int32\")];case\"NoOp\":return[n.scalar(1)];case\"Print\":let i=S(\"x\",r,t,e),a=S(\"data\",r,t,e),u=S(\"message\",r,t,e),l=S(\"summarize\",r,t,e);console.warn(\"The graph has a tf.print() operation,usually used for debugging, which slows down performance.\"),console.log(u);for(let c=0;c<a.length;c++)console.log(Array.prototype.slice.call(a[c].dataSync()).slice(0,l));return[i];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var Bb=class{constructor(t,e){this.keyDType=t,this.valueDType=e,this.handle=mt(0),this.tensorMap=new Map,De(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(t=>t.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return mt(this.size(),\"int32\")}async import(t,e){this.checkKeyAndValueTensor(t,e);let n=await t.data();return this.tensorMap.forEach(o=>o.dispose()),this.tensorMap.clear(),B(()=>{let o=vr(e),s=n.length,i=o.length;y.assert(s===i,()=>`The number of elements doesn't match, keys has ${s} elements, the values has ${i} elements.`);for(let a=0;a<s;a++){let u=n[a],l=o[a];De(l),this.tensorMap.set(u,l)}return this.handle})}async find(t,e){this.checkKeyAndValueTensor(t,e);let n=await t.data();return B(()=>{let o=[];for(let s=0;s<n.length;s++){let i=n[s],a=this.findWithDefault(i,e);o.push(a)}return nr(o)})}findWithDefault(t,e){let n=this.tensorMap.get(t);return n!=null?n:e}checkKeyAndValueTensor(t,e){if(t.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${t.dtype}`);if(e.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${e.dtype}`)}};var JD=async(r,t,e,n)=>{switch(r.op){case\"HashTable\":case\"HashTableV2\":{let o=n.getHashTableHandleByName(r.name);if(o!=null)return[o];{let s=S(\"keyDType\",r,t,e),i=S(\"valueDType\",r,t,e),a=new Bb(s,i);return n.addHashTable(r.name,a),[a.handle]}}case\"LookupTableImport\":case\"LookupTableImportV2\":{let o=S(\"tableHandle\",r,t,e,n),s=S(\"keys\",r,t,e),i=S(\"values\",r,t,e);return[await n.getHashTableById(o.id).import(s,i)]}case\"LookupTableFind\":case\"LookupTableFindV2\":{let o=S(\"tableHandle\",r,t,e,n),s=S(\"keys\",r,t,e),i=S(\"defaultValue\",r,t,e);return[await n.getHashTableById(o.id).find(s,i)]}case\"LookupTableSize\":case\"LookupTableSizeV2\":{let o=S(\"tableHandle\",r,t,e,n);return[n.getHashTableById(o.id).tensorSize()]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var QD=(r,t,e,n=ae)=>{switch(r.op){case\"ResizeBilinear\":{let o=S(\"images\",r,t,e),s=S(\"size\",r,t,e),i=S(\"alignCorners\",r,t,e),a=S(\"halfPixelCenters\",r,t,e);return[n.image.resizeBilinear(o,[s[0],s[1]],i,a)]}case\"ResizeNearestNeighbor\":{let o=S(\"images\",r,t,e),s=S(\"size\",r,t,e),i=S(\"alignCorners\",r,t,e),a=S(\"halfPixelCenters\",r,t,e);return[n.image.resizeNearestNeighbor(o,[s[0],s[1]],i,a)]}case\"CropAndResize\":{let o=S(\"image\",r,t,e),s=S(\"boxes\",r,t,e),i=S(\"boxInd\",r,t,e),a=S(\"cropSize\",r,t,e),u=S(\"method\",r,t,e),l=S(\"extrapolationValue\",r,t,e);return[n.image.cropAndResize(o,s,i,a,u,l)]}case\"ImageProjectiveTransformV3\":{let o=S(\"images\",r,t,e),s=S(\"transforms\",r,t,e),i=S(\"outputShape\",r,t,e),a=S(\"fillValue\",r,t,e),u=S(\"interpolation\",r,t,e),l=S(\"fillMode\",r,t,e);return[n.image.transform(o,s,u.toLowerCase(),l.toLowerCase(),a,i)]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var tR=(r,t,e,n=ae)=>{switch(r.op){case\"Equal\":return[n.equal(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"NotEqual\":return[n.notEqual(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Greater\":return[n.greater(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"GreaterEqual\":return[n.greaterEqual(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Less\":return[n.less(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"LessEqual\":return[n.lessEqual(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"LogicalAnd\":return[n.logicalAnd(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"LogicalNot\":return[n.logicalNot(S(\"a\",r,t,e))];case\"LogicalOr\":return[n.logicalOr(S(\"a\",r,t,e),S(\"b\",r,t,e))];case\"Select\":case\"SelectV2\":return[n.where(S(\"condition\",r,t,e),S(\"a\",r,t,e),S(\"b\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var eR=(r,t,e,n=ae)=>{switch(r.op){case\"BatchMatMul\":case\"BatchMatMulV2\":case\"MatMul\":return[n.matMul(S(\"a\",r,t,e),S(\"b\",r,t,e),S(\"transposeA\",r,t,e),S(\"transposeB\",r,t,e))];case\"Einsum\":return[n.einsum(S(\"equation\",r,t,e),...S(\"tensors\",r,t,e))];case\"Transpose\":return[n.transpose(S(\"x\",r,t,e),S(\"perm\",r,t,e))];case\"_FusedMatMul\":let[o,s]=S(\"fusedOps\",r,t,e),i=o===\"biasadd\",a=s===\"prelu\",u=S(\"numArgs\",r,t,e),l=S(\"leakyreluAlpha\",r,t,e);if(i){if(a&&u!==2)throw new Error(\"Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.\");if(!a&&u!==1)throw new Error(\"Fused MatMul with BiasAdd must have one extra argument: bias.\")}let[c,p]=S(\"args\",r,t,e);return[n.fused.matMul({a:S(\"a\",r,t,e),b:S(\"b\",r,t,e),transposeA:S(\"transposeA\",r,t,e),transposeB:S(\"transposeB\",r,t,e),bias:c,activation:s,preluActivationWeights:p,leakyreluAlpha:l})];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var rR=(r,t,e,n=ae)=>{switch(r.op){case\"EuclideanNorm\":return[n.euclideanNorm(S(\"x\",r,t,e),S(\"axis\",r,t,e),S(\"keepDims\",r,t,e))];case\"FusedBatchNorm\":case\"FusedBatchNormV2\":return[n.batchNorm(S(\"x\",r,t,e),S(\"mean\",r,t,e),S(\"variance\",r,t,e),S(\"offset\",r,t,e),S(\"scale\",r,t,e),S(\"epsilon\",r,t,e))];case\"FusedBatchNormV3\":return[n.batchNorm(S(\"x\",r,t,e),S(\"mean\",r,t,e),S(\"variance\",r,t,e),S(\"offset\",r,t,e),S(\"scale\",r,t,e),S(\"epsilon\",r,t,e))];case\"LRN\":return[n.localResponseNormalization(S(\"x\",r,t,e),S(\"radius\",r,t,e),S(\"bias\",r,t,e),S(\"alpha\",r,t,e),S(\"beta\",r,t,e))];case\"Softmax\":return[n.softmax(S(\"x\",r,t,e))];case\"LogSoftmax\":return[n.logSoftmax(S(\"x\",r,t,e))];case\"SparseToDense\":return[n.sparseToDense(S(\"sparseIndices\",r,t,e),S(\"outputShape\",r,t,e),S(\"sparseValues\",r,t,e),S(\"defaultValue\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var nR=(r,t,e,n=ae)=>{switch(r.op){case\"Max\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.max(S(\"x\",r,t,e),a,u)]}case\"Mean\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.mean(S(\"x\",r,t,e),a,u)]}case\"Min\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.min(S(\"x\",r,t,e),a,u)]}case\"Sum\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.sum(S(\"x\",r,t,e),a,u)]}case\"All\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.all(S(\"x\",r,t,e),a,u)]}case\"Any\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.any(S(\"x\",r,t,e),a,u)]}case\"ArgMax\":{let a=S(\"axis\",r,t,e);return[n.argMax(S(\"x\",r,t,e),a)]}case\"ArgMin\":{let a=S(\"axis\",r,t,e);return[n.argMin(S(\"x\",r,t,e),a)]}case\"Prod\":{let a=S(\"axis\",r,t,e),u=S(\"keepDims\",r,t,e);return[n.prod(S(\"x\",r,t,e),a,u)]}case\"Cumprod\":{let a=S(\"axis\",r,t,e),u=S(\"exclusive\",r,t,e),l=S(\"reverse\",r,t,e);return[n.cumprod(S(\"x\",r,t,e),a,u,l)]}case\"Cumsum\":{let a=S(\"axis\",r,t,e),u=S(\"exclusive\",r,t,e),l=S(\"reverse\",r,t,e);return[n.cumsum(S(\"x\",r,t,e),a,u,l)]}case\"Bincount\":let o=S(\"x\",r,t,e),s=S(\"weights\",r,t,e),i=S(\"size\",r,t,e);return[n.bincount(o,s,i)];case\"DenseBincount\":{let a=S(\"x\",r,t,e),u=S(\"weights\",r,t,e),l=S(\"size\",r,t,e),c=S(\"binaryOutput\",r,t,e);return[n.denseBincount(a,u,l,c)]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var oR=(r,t,e,n=ae)=>{switch(r.op){case\"ConcatV2\":case\"Concat\":{let o=S(\"n\",r,t,e),s=S(\"axis\",r,t,e),i=S(\"tensors\",r,t,e);return i=i.slice(0,o),[n.concat(i,s)]}case\"Gather\":{let o=S(\"x\",r,t,e),s=S(\"indices\",r,t,e);return[n.gather(o,n.cast(s,\"int32\"),0)]}case\"GatherV2\":{let o=S(\"axis\",r,t,e),s=S(\"batchDims\",r,t,e),i=S(\"x\",r,t,e),a=S(\"indices\",r,t,e);return[n.gather(i,n.cast(a,\"int32\"),o,s)]}case\"Reverse\":{let o=S(\"dims\",r,t,e),s=[];for(let a=0;a<o.length;a++)o[a]&&s.push(a);let i=S(\"x\",r,t,e);return[n.reverse(i,s)]}case\"ReverseV2\":{let o=S(\"axis\",r,t,e),s=S(\"x\",r,t,e);return[n.reverse(s,o)]}case\"Slice\":{let o=S(\"begin\",r,t,e),s=S(\"size\",r,t,e);return[n.slice(S(\"x\",r,t,e),o,s)]}case\"StridedSlice\":{let o=S(\"begin\",r,t,e),s=S(\"end\",r,t,e),i=S(\"strides\",r,t,e),a=S(\"beginMask\",r,t,e),u=S(\"endMask\",r,t,e),l=S(\"ellipsisMask\",r,t,e),c=S(\"newAxisMask\",r,t,e),p=S(\"shrinkAxisMask\",r,t,e),m=S(\"x\",r,t,e);return[n.stridedSlice(m,o,s,i,a,u,l,c,p)]}case\"Pack\":return B(()=>{let o=S(\"axis\",r,t,e),s=S(\"tensors\",r,t,e),i=s[0].shape,a=n.squeeze(s[0]).shape,u=s.map(l=>{let c=y.arraysEqual(l.shape,i);if(!c&&!y.arraysEqual(n.squeeze(l).shape,a))throw new Error(\"the input tensors shape does not match\");return c?l:n.reshape(l,i)});return[n.stack(u,o)]});case\"Unpack\":{let o=S(\"axis\",r,t,e),s=S(\"tensor\",r,t,e);return n.unstack(s,o)}case\"Tile\":{let o=S(\"reps\",r,t,e);return[n.tile(S(\"x\",r,t,e),o)]}case\"Split\":case\"SplitV\":{let o=S(\"axis\",r,t,e),s=S(\"numOrSizeSplits\",r,t,e),i=S(\"x\",r,t,e);return n.split(i,s,o)}case\"ScatterNd\":{let o=S(\"indices\",r,t,e),s=S(\"values\",r,t,e),i=S(\"shape\",r,t,e);return[n.scatterND(o,s,i)]}case\"GatherNd\":{let o=S(\"x\",r,t,e),s=S(\"indices\",r,t,e);return[n.gatherND(o,s)]}case\"SparseToDense\":{let o=S(\"sparseIndices\",r,t,e),s=S(\"outputShape\",r,t,e),i=S(\"sparseValues\",r,t,e),a=S(\"defaultValue\",r,t,e);return[n.sparseToDense(o,i,s,i.dtype===a.dtype?a:n.cast(a,i.dtype))]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var sR=(r,t,e,n=ae)=>{switch(r.op){case\"SparseFillEmptyRows\":{let{outputIndices:o,outputValues:s,emptyRowIndicator:i,reverseIndexMap:a}=n.sparse.sparseFillEmptyRows(S(\"indices\",r,t,e),S(\"values\",r,t,e),S(\"denseShape\",r,t,e),S(\"defaultValue\",r,t,e));return[o,s,i,a]}case\"SparseReshape\":{let{outputIndices:o,outputShape:s}=n.sparse.sparseReshape(S(\"inputIndices\",r,t,e),S(\"inputShape\",r,t,e),S(\"newShape\",r,t,e));return[o,s]}case\"SparseSegmentMean\":return[n.sparse.sparseSegmentMean(S(\"data\",r,t,e),S(\"indices\",r,t,e),S(\"segmentIds\",r,t,e))];case\"SparseSegmentSum\":return[n.sparse.sparseSegmentSum(S(\"data\",r,t,e),S(\"indices\",r,t,e),S(\"segmentIds\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var iR=(r,t,e,n=ae)=>{switch(r.op){case\"FFT\":return[n.fft(S(\"x\",r,t,e))];case\"IFFT\":return[n.ifft(S(\"x\",r,t,e))];case\"RFFT\":return[n.rfft(S(\"x\",r,t,e))];case\"IRFFT\":return[n.irfft(S(\"x\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var aR=(r,t,e,n=ae)=>{switch(r.op){case\"StringNGrams\":{let{nGrams:o,nGramsSplits:s}=n.string.stringNGrams(S(\"data\",r,t,e),S(\"dataSplits\",r,t,e),S(\"separator\",r,t,e),S(\"nGramWidths\",r,t,e),S(\"leftPad\",r,t,e),S(\"rightPad\",r,t,e),S(\"padWidth\",r,t,e),S(\"preserveShortSequences\",r,t,e));return[o,s]}case\"StringSplit\":{let{indices:o,values:s,shape:i}=n.string.stringSplit(S(\"input\",r,t,e),S(\"delimiter\",r,t,e),S(\"skipEmpty\",r,t,e));return[o,s,i]}case\"StringToHashBucketFast\":return[n.string.stringToHashBucketFast(S(\"input\",r,t,e),S(\"numBuckets\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var lR=(r,t,e,n=ae)=>{switch(r.op){case\"Cast\":return[n.cast(S(\"x\",r,t,e),S(\"dtype\",r,t,e))];case\"ExpandDims\":{let o=S(\"axis\",r,t,e);return[n.expandDims(S(\"x\",r,t,e),o)]}case\"Squeeze\":{let o=S(\"axis\",r,t,e);return[n.squeeze(S(\"x\",r,t,e),o)]}case\"Reshape\":return[n.reshape(S(\"x\",r,t,e),S(\"shape\",r,t,e))];case\"MirrorPad\":return[n.mirrorPad(S(\"x\",r,t,e),S(\"padding\",r,t,e),S(\"mode\",r,t,e))];case\"PadV2\":case\"Pad\":return[n.pad(S(\"x\",r,t,e),S(\"padding\",r,t,e),S(\"constantValue\",r,t,e))];case\"SpaceToBatchND\":{let o=S(\"blockShape\",r,t,e),s=S(\"paddings\",r,t,e);return[n.spaceToBatchND(S(\"x\",r,t,e),o,s)]}case\"BatchToSpaceND\":{let o=S(\"blockShape\",r,t,e),s=S(\"crops\",r,t,e);return[n.batchToSpaceND(S(\"x\",r,t,e),o,s)]}case\"DepthToSpace\":{let o=S(\"blockSize\",r,t,e),s=S(\"dataFormat\",r,t,e).toUpperCase();return[n.depthToSpace(S(\"x\",r,t,e),o,s)]}case\"BroadcastTo\":return[n.broadcastTo(S(\"x\",r,t,e),S(\"shape\",r,t,e))];case\"BroadcastArgs\":return[n.broadcastArgs(S(\"s0\",r,t,e),S(\"s1\",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function fN(r,t,e,n,o=B){let s=((i,a,u)=>{switch(i.category){case\"arithmetic\":return o(()=>MD(i,a,u));case\"basic_math\":return o(()=>zD(i,a,u));case\"control\":return HD(i,a,u);case\"convolution\":return o(()=>KD(i,a,u));case\"creation\":return o(()=>jD(i,a,u));case\"dynamic\":return XD(i,a,u);case\"evaluation\":return o(()=>YD(i,a,u));case\"image\":return o(()=>QD(i,a,u));case\"graph\":return o(()=>ZD(i,a,u));case\"logical\":return o(()=>tR(i,a,u));case\"matrices\":return o(()=>eR(i,a,u));case\"normalization\":return o(()=>rR(i,a,u));case\"reduction\":return o(()=>nR(i,a,u));case\"slice_join\":return o(()=>oR(i,a,u));case\"sparse\":return o(()=>sR(i,a,u));case\"spectral\":return o(()=>iR(i,a,u));case\"string\":return o(()=>aR(i,a,u));case\"transformation\":return o(()=>lR(i,a,u));case\"hash_table\":return JD(i,a,u,n);case\"custom\":let l=Nb(i.op);if(l&&l.customExecutor)return l.customExecutor(new Lb(i,a,u));throw TypeError(`Custom op ${i.op} is not registered.`);default:throw TypeError(`Unknown op '${i.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(r,t,e);return y.isPromise(s)?s.then(i=>[].concat(i)):[].concat(s)}var Oh=class{constructor(t={},e={},n={},o={}){this.weightMap=t,this.tensorArrayMap=e,this.tensorListMap=n,this.functionMap=o,this.rootContext={id:0,frameName:\"\",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(t,e){return{id:t,frameName:e,iterationId:0}}set currentContext(t){this.contexts!==t&&(this.contexts=t,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let t=[];for(let e=0;e<this.contexts.length-1;e++){let n=this.contexts.slice(0,this.contexts.length-e);t.push(this.contextIdforContexts(n))}t.push(\"\"),this._currentContextIds=t}contextIdforContexts(t){return t?t.map(e=>e.id===0&&e.iterationId===0?\"\":`${e.frameName}-${e.iterationId}`).join(\"/\"):\"\"}enterFrame(t){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,t)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error(\"Cannot exit frame, the context is empty\")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let t=Object.assign({},this.contexts[this.contexts.length-1]);t.iterationId+=1,t.id=this.lastId,this.contexts.splice(-1,1,t),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error(\"Cannot increase frame iteration, the context is empty\")}getWeight(t){return this.weightMap[t]}addTensorArray(t){this.tensorArrayMap[t.id]=t}getTensorArray(t){return this.tensorArrayMap[t]}addTensorList(t){this.tensorListMap[t.id]=t}getTensorList(t){return this.tensorListMap[t]}dispose(t){for(let e in this.tensorArrayMap)this.tensorArrayMap[e].clearAndClose(t);for(let e in this.tensorListMap)this.tensorListMap[e].clearAndClose(t)}};function dN(r,t,e,n){let o=new Set,s=[],i=null,a=null,u=new Set,l=Object.keys(r).map(m=>xn(m)[0]),c=[];n!=null&&(c=n.map(m=>xn(m.name)[0]));let p=[...t];for(;p.length>0;){let m=p.pop();if((hN(m)||F7(m)||O7(m))&&i==null&&(i=m,a=i.children.map(f=>f.name).filter(f=>o.has(f))),o.add(m.name),e[m.name]==null&&l.indexOf(m.name)===-1&&c.indexOf(m.name)===-1){if(m.inputs.length===0){s.push(m.name);continue}m.inputs.forEach(f=>{u.has(f.name)||(u.add(f.name),p.push(f))})}}return{inputs:r,outputs:t,usedNodes:o,missingInputs:s,dynamicNode:i,syncInputs:a}}function uR(r,t,e){let{usedNodes:n,inputs:o}=e,s=[],i=Object.keys(o).map(c=>xn(c)[0]).map(c=>r.nodes[c]),a=r.initNodes;i.forEach(c=>{n.has(c.name)&&s.push(c)}),r.weights.forEach(c=>{n.has(c.name)&&s.push(c)}),a!=null&&a.forEach(c=>{n.has(c.name)&&s.push(c)});let u=new Set,l=[];for(;s.length>0;){let c=s.pop();u.add(c.name),t[c.name]||l.push(c),c.children.forEach(p=>{!u.has(p.name)&&n.has(p.name)&&p.inputs.every(m=>u.has(m.name))&&s.push(p)})}return l}var $7=[\"Switch\",\"Merge\",\"Enter\",\"Exit\",\"NextIteration\",\"StatelessIf\",\"StatelessWhile\",\"if\",\"While\"],D7=[\"NonMaxSuppressionV2\",\"NonMaxSuppressionV3\",\"NonMaxSuppressionV5\",\"Where\"],R7=[\"HashTable\",\"HashTableV2\",\"LookupTableImport\",\"LookupTableImportV2\",\"LookupTableFind\",\"LookupTableFindV2\",\"LookupTableSize\",\"LookupTableSizeV2\"];function hN(r){return $7.indexOf(r.op)>=0}function F7(r){return D7.indexOf(r.op)>=0}function O7(r){return R7.indexOf(r.op)>=0}var Nc=class{constructor(t,e){this.graph=t,this.parent=e,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=\",\",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=t.outputs,this._inputs=t.inputs,this._initNodes=t.initNodes,this._signature=t.signature,this._functions=t.functions,t.functions!=null&&Object.keys(t.functions).forEach(n=>{this._functionExecutorMap[n]=new Nc(t.functions[n],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(t){let e=Object.keys(t).map(n=>t[n].map(o=>o.id));this._weightIds=[].concat(...e),this._weightMap=t}set resourceManager(t){this._resourceManager=t}get inputs(){return this._inputs.map(t=>({name:t.name,shape:t.attrParams.shape?t.attrParams.shape.value:void 0,dtype:t.attrParams.dtype?t.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(t=>({name:t.name,shape:t.attrParams.shape?t.attrParams.shape.value:void 0,dtype:t.attrParams.dtype?t.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(t=>t.signatureKey||t.name)}get outputNodes(){return this._outputs.map(t=>{let e=t.signatureKey||t.name;return t.defaultOutput?`${e}:${t.defaultOutput}`:e})}get functions(){return Object.keys(this._functions).reduce((t,e)=>(t[e]=this._functions[e].signature,t),{})}getCompilationKey(t,e){let n=t.map(s=>s.name).sort(),o=e.map(s=>s.name).sort();return n.join(this.SEPERATOR)+\"--\"+o.join(this.SEPERATOR)}compile(t,e){let n=dN(t,e,this.weightMap,this._initNodes),{missingInputs:o,dynamicNode:s,syncInputs:i}=n;if(s!=null)throw new Error(`This execution contains the node '${s.name}', which has the dynamic op '${s.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${i}]`);if(o.length>0){let a=e.map(l=>l.name),u=Object.keys(t);throw new Error(`Cannot compute the outputs [${a}] from the provided inputs [${u}]. Missing the following inputs: [${o}]`)}return uR(this.graph,this.weightMap,n)}execute(t,e){t=this.mapInputs(t);let n=Object.keys(t).sort();this.checkInputs(t),this.checkInputShapeAndType(t),e=this.mapOutputs(e),this.checkOutputs(e);let o=n.map(p=>this.graph.nodes[xn(p)[0]]),s=e.map(p=>xn(p)[0]),i=s.map(p=>this.graph.nodes[p]);this.resetIntermediateTensors(),i.length===0&&(i=this._outputs);let a=this.getCompilationKey(o,i),u=this.compiledMap.get(a);u==null&&(u=this.compile(t,i),this.compiledMap.set(a,u));let l={},c={};return B(()=>{let p=new Oh(this.weightMap,l,c,this.functionExecutorMap),m=Object.assign({},this.weightMap);Object.keys(t).forEach(h=>{let[g,x]=xn(h),b=[];b[x]=t[h],m[g]=b});let f=this.getFrozenTensorIds(m),d={};for(let h=0;h<u.length;h++){let g=u[h];if(!m[g.name]){let x=fN(g,m,p,this._resourceManager);if(y.isPromise(x))throw new Error(`The execution of the op '${g.op}' returned a promise. Please use model.executeAsync() instead.`);m[g.name]=x,this.checkTensorForDisposal(g.name,g,m,p,f,s,d)}}return this.parent==null&&p.dispose(f),e.map(h=>br(h,m,p))})}getFrozenTensorIds(t){let e=[].concat.apply([],Object.keys(t).map(n=>t[n]).map(n=>n.map(o=>o.id)));return new Set(e)}checkTensorForDisposal(t,e,n,o,s,i,a){e.category===\"control\"||i.indexOf(t)!==-1||(n[t].forEach(u=>{u!=null&&(a[u.id]=(a[u.id]||0)+e.children.length)}),e.inputs.forEach(u=>{if(u.category!==\"control\"){let l=FD(u.name,n,o);l!=null&&l.forEach(c=>{if(c&&!c.kept&&!s.has(c.id)){let p=a[c.id];if(p===1){if(!this.keepTensorForDebug)c.dispose();else{let[m,f]=_o(e.name,o);this.intermediateTensors[m]?this.intermediateTensors[m][f]=c:(this.intermediateTensors[m]=[],this.intermediateTensors[m][f]=c)}delete a[c.id]}else p!=null&&a[c.id]--}})}}))}async executeAsync(t,e){return this._executeAsync(t,e)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(t=>this.intermediateTensors[t].forEach(e=>e.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(t=>{this.tensorsMap[t].forEach(n=>{n&&!n.kept&&!n.isDisposed&&!this.keepIds.has(n.id)&&n.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let t in this.intermediateTensors)this.intermediateTensors[t].forEach(e=>e.dispose()),delete this.intermediateTensors[t]}async _executeAsync(t,e,n=!1,o={},s={}){n||(t=this.mapInputs(t),this.checkInputs(t),this.checkInputShapeAndType(t),e=this.mapOutputs(e),this.checkOutputs(e));try{this.keepTensorForDebug=z().getBool(\"KEEP_INTERMEDIATE_TENSORS\")}catch(c){console.warn(c.message)}this.resetIntermediateTensors();let i=new Oh(this.weightMap,o,s,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(t,i,e,n);let a=e.map(c=>br(c,this.tensorsMap,i)),u=a.map(c=>c.id),l=Object.keys(t).map(c=>t[c].id);return this.keepIds=new Set([...u,...l,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&i.dispose(this.keepIds),a}async executeFunctionAsync(t,e,n){let o=t.reduce((s,i,a)=>(s[this.inputs[a].name]=i,s),{});return this._executeAsync(o,this.outputNodes,!0,e,n)}async executeWithControlFlow(t,e,n,o){let s=Object.keys(t),i=s.map(w=>this.graph.nodes[xn(w)[0]]),a=n.map(w=>xn(w)[0]),u=a.map(w=>this.graph.nodes[w]);u.length===0&&(u=this._outputs);let{usedNodes:l,missingInputs:c,dynamicNode:p,syncInputs:m}=dN(t,u,this.weightMap,this._initNodes),f=[...i,...this.graph.weights,...this._initNodes||[]].map(w=>({node:w,contexts:e.currentContext})),d=Object.assign({},this.weightMap);Object.keys(t).forEach(w=>{let[C,N]=xn(w),_=[];_[N]=t[w],d[C]=_});let h={},g=this.getFrozenTensorIds(d),x={};for(;f.length>0;){let w=this.processStack(i,f,e,d,x,g,a,h,l);await Promise.all(w)}p==null&&!o&&console.warn(\"This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.\");let b=u.filter(w=>!hN(w)&&!br(w.name,d,e)).map(w=>w.name);if(b.length>0){let w=\"\";throw p!=null&&(w=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${m}]`),new Error(`Cannot compute the outputs [${b}] from the provided inputs [${s}]. Consider providing the following inputs: [${c}]. ${w}`)}return d}processStack(t,e,n,o,s,i,a,u,l){let c=[];for(;e.length>0;){let p=e.pop();n.currentContext=p.contexts;let m=\"\";if(p.node.op===\"Enter\"&&S(\"isConstant\",p.node,o,n)&&([m]=_o(p.node.name,n)),o[p.node.name]==null){let f=fN(p.node,o,n,this._resourceManager);m||([m]=_o(p.node.name,n));let d=n.currentContext;y.isPromise(f)?c.push(f.then(h=>(o[m]=h,n.currentContext=d,this.checkTensorForDisposal(m,p.node,o,n,i,a,u),this.processChildNodes(p.node,e,n,o,s,l),h))):(o[m]=f,this.checkTensorForDisposal(m,p.node,o,n,i,a,u),this.processChildNodes(p.node,e,n,o,s,l))}else this.processChildNodes(p.node,e,n,o,s,l)}return c}processChildNodes(t,e,n,o,s,i){t.children.forEach(a=>{let[u]=_o(a.name,n);s[u]||!i.has(a.name)||(a.op===\"Merge\"?a.inputNames.some(l=>!!br(l,o,n))&&(s[u]=!0,e.push({contexts:n.currentContext,node:a})):a.inputNames.every(l=>!!br(l,o,n))&&(s[u]=!0,e.push({contexts:n.currentContext,node:a})))})}dispose(){Object.keys(this.weightMap).forEach(t=>this.weightMap[t].forEach(e=>e.dispose()))}checkInputShapeAndType(t){Object.keys(t).forEach(e=>{let n=t[e],[o]=xn(e),s=this.graph.nodes[o];if(s.attrParams.shape&&s.attrParams.shape.value){let i=s.attrParams.shape.value,a=i.length===n.shape.length&&n.shape.every((u,l)=>i[l]===-1||i[l]===u);y.assert(a,()=>`The shape of dict['${s.name}'] provided in model.execute(dict) must be [${i}], but was [${n.shape}]`)}s.attrParams.dtype&&s.attrParams.dtype.value&&y.assert(n.dtype===s.attrParams.dtype.value,()=>`The dtype of dict['${s.name}'] provided in model.execute(dict) must be ${s.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(t){let e={};for(let n in t)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[n]!=null){let o=this._signature.inputs[n];e[o.name]=t[n]}else e[n]=t[n];return e}checkInputs(t){let e=Object.keys(t).filter(n=>{let[o]=xn(n);return this.graph.nodes[o]==null});if(e.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${e}] that are not part of graph`)}mapOutputs(t){return t.map(e=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[e]!=null?this._signature.outputs[e].name:e,{})}checkOutputs(t){t.forEach(e=>{let[n]=xn(e);if(!this.graph.nodes[n])throw new Error(`The output '${e}' is not found in the graph`)})}};var Vb=class{constructor(t={},e={}){this.hashTableNameToHandle=t,this.hashTableMap=e}addHashTable(t,e){this.hashTableNameToHandle[t]=e.handle,this.hashTableMap[e.id]=e}getHashTableHandleByName(t){return this.hashTableNameToHandle[t]}getHashTableById(t){return this.hashTableMap[t]}dispose(){for(let t in this.hashTableMap)this.hashTableMap[t].clearAndClose(),delete this.hashTableMap[t];for(let t in this.hashTableNameToHandle)this.hashTableNameToHandle[t].dispose(),delete this.hashTableNameToHandle[t]}};var P7=\"?tfjs-format=file\",L7=\"model.json\",Ph=class{constructor(t,e={},n=_r){this.modelUrl=t,this.loadOptions=e,this.version=\"n/a\",this.io=n,e==null&&(this.loadOptions={}),this.resourceManager=new Vb}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}findIOHandler(){let t=this.modelUrl;if(t.load!=null)this.handler=t;else if(this.loadOptions.requestInit!=null)this.handler=this.io.browserHTTPRequest(t,this.loadOptions);else{let e=this.io.getLoadHandlers(t,this.loadOptions);if(e.length===0)e.push(this.io.browserHTTPRequest(t,this.loadOptions));else if(e.length>1)throw new Error(`Found more than one (${e.length}) load handlers for URL '${[t]}'`);this.handler=e[0]}}load(){if(this.findIOHandler(),this.handler.load==null)throw new Error(\"Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.\");let t=this.handler.load();return y.isPromise(t)?t.then(e=>this.loadSync(e)):this.loadSync(t)}loadSync(t){this.artifacts=t;let e=this.artifacts.modelTopology,n=this.artifacts.signature;if(this.artifacts.userDefinedMetadata!=null){let s=this.artifacts.userDefinedMetadata;s.signature!=null&&(n=s.signature),s.structuredOutputKeys!=null&&(this.structuredOutputKeys=s.structuredOutputKeys)}this.signature=n,this.version=`${e.versions.producer}.${e.versions.minConsumer}`;let o=this.io.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new Nc(Fh.Instance.transformGraph(e,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(o),this.executor.resourceManager=this.resourceManager,t.modelInitializer!=null&&t.modelInitializer.node!=null){let s=Fh.Instance.transformGraph(t.modelInitializer);this.initializer=new Nc(s),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=t.initializerSignature}return!0}async save(t,e){if(typeof t==\"string\"){let n=this.io.getSaveHandlers(t);if(n.length===0)throw new Error(`Cannot find any save handlers for URL '${t}'`);if(n.length>1)throw new Error(`Found more than one (${n.length}) save handlers for URL '${t}'`);t=n[0]}if(t.save==null)throw new Error(\"GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.\");return t.save(this.artifacts)}predict(t,e){let n=this.execute(t,this.outputNodes);if(this.structuredOutputKeys){let o=n instanceof Ft?[n]:n,s={};return o.forEach((i,a)=>s[this.structuredOutputKeys[a]]=i),s}return n}normalizeInputs(t){if(!(t instanceof Ft)&&!Array.isArray(t)){if(this.signature!=null&&this.signature.inputs!=null)for(let o in this.signature.inputs){let s=this.signature.inputs[o];s.resourceId!=null&&(t[o]=this.resourceIdToCapturedInput[s.resourceId])}return t}t=Array.isArray(t)?t:[t];let e=Object.keys(this.resourceIdToCapturedInput).length;if(t.length+e!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-e} non-resource placeholders, while there are ${t.length} input tensors provided.`);let n=0;return this.inputNodes.reduce((o,s)=>{let i=this.signature?this.signature.inputs[s]:null;return i!=null&&i.resourceId!=null?o[s]=this.resourceIdToCapturedInput[i.resourceId]:o[s]=t[n++],o},{})}normalizeOutputs(t){return t=t||this.outputNodes,Array.isArray(t)?t:[t]}executeInitializerGraph(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(t){if(this.resourceIdToCapturedInput={},this.initializerSignature){let e=Object.keys(this.initializerSignature.outputs);for(let n=0;n<e.length;n++){let o=e[n],s=this.initializerSignature.outputs[o];this.resourceIdToCapturedInput[s.resourceId]=t[n]}}}execute(t,e){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),t=this.normalizeInputs(t),e=this.normalizeOutputs(e);let n=this.executor.execute(t,e);return n.length>1?n:n[0]}async executeAsync(t,e){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),t=this.normalizeInputs(t),e=this.normalizeOutputs(e);let n=await this.executor.executeAsync(t,e);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(t){return Object.keys(t).reduce((e,n)=>(e[n]=[t[n]],e),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&vt(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function M7(r,t={},e=_r){if(r==null)throw new Error(\"modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model\");t==null&&(t={}),t.fromTFHub&&typeof r==\"string\"&&(r=B7(r));let n=new Ph(r,t,e);return await n.load(),n}function z7(r){if(r==null)throw new Error(\"modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model\");let t;if(r instanceof Array){let[n,o]=r;if(!n)throw new Error(\"modelJSON must be the first element of the array\");if(!o||!(o instanceof ArrayBuffer))throw new Error(\"An ArrayBuffer of weights must be the second element of the array\");if(!(\"modelTopology\"in n))throw new Error(\"Model JSON is missing 'modelTopology'\");if(!(\"weightsManifest\"in n))throw new Error(\"Model JSON is missing 'weightsManifest'\");let s=_r.getWeightSpecs(n.weightsManifest),i=_r.getModelArtifactsForJSONSync(n,s,o);t=_r.fromMemorySync(i)}else if(\"load\"in r)t=r;else if(\"modelTopology\"in r&&\"weightSpecs\"in r&&\"weightData\"in r)t=_r.fromMemorySync(r);else throw new Error(\"Unknown model format\");let e=new Ph(t);return e.load(),e}function B7(r){return r.endsWith(\"/\")||(r=r+\"/\"),`${r}${L7}${P7}`}var cR=\"4.0.0\";var AR={};Wt(AR,{CSVDataset:()=>Yf,Dataset:()=>Js,FileDataSource:()=>ed,TextLineDataset:()=>Xf,URLDataSource:()=>rd,array:()=>wR,csv:()=>NR,func:()=>TR,generator:()=>kR,microphone:()=>_R,version_data:()=>PN,webcam:()=>ER,zip:()=>CR});var bR=Tl(fh());var gR=Tl(fh());function pR(r,t){return Gb(r,t)}function Gb(r,t,e=new Map,n=new Set){if(r==null)return null;if(typeof Blob==\"function\"&&r instanceof Blob)return r.slice();if(n.has(r))throw new Error(\"Circular references are not supported.\");if(e.has(r))return e.get(r);let o=t(r);if(o.recurse&&o.value!==null)throw new Error(\"A deep map function may not return both a value and recurse=true.\");if(o.recurse)if(vu(r)){let s=Array.isArray(r)?[]:{};n.add(r);for(let i in r){let a=r[i],u=Gb(a,t,e,n);s[i]=u}return n.delete(r),r.__proto__&&(s.__proto__=r.__proto__),s}else throw new Error(`Can't recurse into non-iterable type: ${r}`);else return e.set(r,o.value),o.value}function mR(r,t=xN){return fR(r,t)}function fR(r,t,e=new Set){let n=r[0];if(e.has(n))throw new Error(\"Circular references are not supported.\");let o=t(r);if(o.recurse&&o.value!==null)throw new Error(\"A deep zip function may not return both a value and recurse=true.\");if(o.recurse)if(vu(n)){let s=Array.isArray(n)?[]:{};e.add(n);for(let i in n){let a=r.map(l=>l[i]),u=fR(a,t,e);s[i]=u}return e.delete(n),s}else throw new Error(`Can't recurse into non-iterable type: ${n}`);else return o.value}function xN(r){return r===null?null:vu(r[0])?{value:null,recurse:!0}:{value:r,recurse:!1}}async function Wb(r,t){let e=new Map;Gb(r,t,e);for(let o of Array.from(e.keys())){let s=e.get(o);if(y.isPromise(s)){let i=await s;e.set(o,i)}}return Gb(r,t,e)}function vu(r){let t=!1;if(z().get(\"IS_BROWSER\"))t=r instanceof TextDecoder;else{let{StringDecoder:e}=gN();t=r instanceof e}return r!=null&&!ArrayBuffer.isView(r)&&(Array.isArray(r)||typeof r==\"object\"&&!(r instanceof Ft)&&!(r instanceof Promise)&&!t)}function dR(r){return r==null||V7(r)||Array.isArray(r)||typeof r==\"object\"&&r instanceof Ft||y.isTypedArray(r)}function V7(r){return r===null||typeof r!=\"object\"&&typeof r!=\"function\"}function hR(r){return pR(r,G7)}function G7(r){return r instanceof Ft?{value:r.clone(),recurse:!1}:vu(r)?{value:null,recurse:!0}:{value:r,recurse:!1}}var jf=class{constructor(t){if(this.capacity=t,this.begin=0,this.end=0,t==null)throw new RangeError(\"Can't create a ring buffer of unknown capacity.\");if(t<1)throw new RangeError(\"Can't create ring buffer of capacity < 1.\");this.data=new Array(t),this.doubledCapacity=2*t}wrap(t){for(;t<0;)t+=this.doubledCapacity;return t%this.doubledCapacity}get(t){if(t<0)throw new RangeError(\"Can't get item at a negative index.\");return this.data[t%this.capacity]}set(t,e){if(t<0)throw new RangeError(\"Can't set item at a negative index.\");this.data[t%this.capacity]=e}length(){let t=this.end-this.begin;return t<0&&(t=this.doubledCapacity+t),t}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(t){if(this.isFull())throw new RangeError(\"Ring buffer is full.\");this.set(this.end,t),this.end=this.wrap(this.end+1)}pushAll(t){for(let e of t)this.push(e)}pop(){if(this.isEmpty())throw new RangeError(\"Ring buffer is empty.\");this.end=this.wrap(this.end-1);let t=this.get(this.end);return this.set(this.end,void 0),t}unshift(t){if(this.isFull())throw new RangeError(\"Ring buffer is full.\");this.begin=this.wrap(this.begin-1),this.set(this.begin,t)}shift(){if(this.isEmpty())throw new RangeError(\"Ring buffer is empty.\");let t=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),t}shuffleExcise(t){if(this.isEmpty())throw new RangeError(\"Ring buffer is empty.\");let e=this.wrap(this.begin+t),n=this.get(e);return this.set(e,this.pop()),n}};var Tc=class extends jf{constructor(){super(Tc.INITIAL_CAPACITY)}isFull(){return!1}push(t){super.isFull()&&this.expand(),super.push(t)}unshift(t){super.isFull()&&this.expand(),super.unshift(t)}expand(){let t=this.capacity*2,e=new Array(t),n=this.length();for(let o=0;o<n;o++)e[o]=this.get(this.wrap(this.begin+o));this.data=e,this.capacity=t,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};Tc.INITIAL_CAPACITY=32;function AN(r){return new yN(r)}function Lh(r){return new bN(r)}function xR(r,t){return new Hb(r,t)}function yR(r,t=fl.FAIL){return new EN(r,t)}var Je=class{async toArray(){let t=[],e=await this.next();for(;!e.done;)t.push(e.value),e=await this.next();return t}async toArrayForTest(){let t=this.prefetch(100),e=[],n=await t.next();for(;!n.done;)e.push(n.value),n=await t.next();return e}async resolveFully(){let t=await this.next();for(;!t.done;)t=await this.next()}async resolveWhile(t){let e=await this.next(),n=t(e.value);for(;!e.done&&n;)e=await this.next(),n=t(e.value)}handleErrors(t){return new TN(this,t)}filter(t){return new vN(this,t)}map(t){return new NN(this,t)}mapAsync(t){return new Ub(this,t)}serialMapAsync(t){return new Ub(this,t).serial()}flatmap(t){return new kN(this,t)}async forEachAsync(t){return this.map(t).resolveFully()}async serialForEach(t){return this.serialMapAsync(t).resolveWhile(e=>e===!0)}rowMajorBatch(t,e=!0){return new SN(this,t,e)}columnMajorBatch(t,e=!0,n=xN){return this.rowMajorBatch(t,e).map(s=>mR(s,n))}concatenate(t,e){return new Hb(AN([this,t]),e)}take(t){return t<0||t==null?this:new IN(this,t)}skip(t){return t<0||t==null?this:new CN(this,t)}prefetch(t){return new qb(this,t)}shuffle(t,e){return new _N(this,t,e)}serial(){return new wN(this)}},yN=class extends Je{constructor(t){super(),this.items=t,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let t=this.items[this.trav];return this.trav++,{value:hR(t),done:!1}}},bN=class extends Je{constructor(t){super(),this.nextFn=t}summary(){return\"Function call\"}async next(){try{return this.nextFn()}catch(t){throw t.message=`Error thrown while iterating through a dataset: ${t.message}`,t}}},wN=class extends Je{constructor(t){super(),this.upstream=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},CN=class extends Je{constructor(t,e){super(),this.upstream=t,this.maxCount=e,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let t=await this.upstream.next();if(t.done)return t;vt(t.value)}return this.upstream.next()}},IN=class extends Je{constructor(t,e){super(),this.upstream=t,this.maxCount=e,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},SN=class extends Je{constructor(t,e,n=!0){super(),this.upstream=t,this.batchSize=e,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let t=[];for(;t.length<this.batchSize;){let e=await this.upstream.next();if(e.done)return this.enableSmallLastBatch&&t.length>0?{value:t,done:!1}:{value:null,done:!0};t.push(e.value)}return{value:t,done:!1}}},vN=class extends Je{constructor(t,e){super(),this.upstream=t,this.predicate=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let t=await this.upstream.next();if(t.done||this.predicate(t.value))return t;vt(t.value)}}},NN=class extends Je{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> Map`}async next(){let t=await this.upstream.next();if(t.done)return{value:null,done:!0};let e=go.getTensorsInContainer(t.value),n=this.transform(t.value),o=go.getTensorsInContainer(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return{value:n,done:!1}}},TN=class extends Je{constructor(t,e){super(),this.upstream=t,this.handler=e,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(t){if(!this.handler(t))return{value:null,done:!0}}}},Ub=class extends Je{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let t=await this.upstream.next();if(t.done)return{value:null,done:!0};let e=go.getTensorsInContainer(t.value),n=await this.transform(t.value),o=go.getTensorsInContainer(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return{value:n,done:!1}}},kc=class extends Je{constructor(){super(),this.outputQueue=new Tc,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},kN=class extends kc{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let t=await this.upstream.next();if(t.done)return!1;let e=go.getTensorsInContainer(t.value),n=this.transform(t.value),o=go.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return!0}},Hb=class extends Je{constructor(t,e){super(),this.baseErrorHandler=e,this.lastRead=null,this.iterator=null,this.moreIterators=t}summary(){return\"TODO: fill in upstream of chained summaries -> Chained\"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(t){if(await t,this.iterator==null){let n=await this.moreIterators.next();if(n.done)return{value:null,done:!0};this.iterator=n.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let e=await this.iterator.next();return e.done?(this.iterator=null,this.readFromChain(t)):e}},fl;(function(r){r[r.FAIL=0]=\"FAIL\",r[r.SHORTEST=1]=\"SHORTEST\",r[r.LONGEST=2]=\"LONGEST\"})(fl||(fl={}));var EN=class extends Je{constructor(t,e=fl.FAIL){super(),this.iterators=t,this.mismatchMode=e,this.count=0,this.currentPromise=null}summary(){return\"{TODO: fill in upstream of zip summaries} -> Zip\"}async nextState(t){await t;let e=0,n=0;function o(i){return i instanceof Je?{value:i.next().then(u=>(e++,u.done&&n++,u.value)),recurse:!1}:{value:null,recurse:!0}}let s=await Wb(this.iterators,o);if(e===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case fl.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case fl.SHORTEST:return{value:null,done:!0};case fl.LONGEST:default:}return this.count++,{value:s,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},qb=class extends Je{constructor(t,e){super(),this.upstream=t,this.bufferSize=e,this.buffer=new jf(e)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let t=this.upstream.next();this.buffer.push(t)}}next(){return this.refill(),this.buffer.shift()}},_N=class extends qb{constructor(t,e,n){super(t,e),this.upstream=t,this.windowSize=e,this.upstreamExhausted=!1,this.random=gR.alea(n||y.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(t){return Math.floor(this.random()*t)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let t=this.chooseIndex(),e=await this.buffer.shuffleExcise(t);if(e.done)this.upstreamExhausted=!0;else return this.refill(),e}return{value:null,done:!0}}};var Js=class{constructor(){this.size=null}batch(t,e=!0){let n=this;y.assert(t>0,()=>`batchSize needs to be positive, but it is\n      ${t}`);let o;return this.size===1/0||this.size==null?o=this.size:e?o=Math.ceil(this.size/t):o=Math.floor(this.size/t),kn(async()=>(await n.iterator()).columnMajorBatch(t,e,W7),o)}concatenate(t){let e=this,n;return this.size===1/0||t.size===1/0?n=1/0:this.size!=null&&t.size!=null?n=this.size+t.size:n=null,kn(async()=>(await e.iterator()).concatenate(await t.iterator()),n)}filter(t){let e=this,n;return this.size===1/0?n=1/0:n=null,kn(async()=>(await e.iterator()).filter(o=>B(()=>t(o))),n)}async forEachAsync(t){return(await this.iterator()).forEachAsync(t)}map(t){let e=this;return kn(async()=>(await e.iterator()).map(n=>B(()=>t(n))),this.size)}mapAsync(t){let e=this;return kn(async()=>(await e.iterator()).mapAsync(t),this.size)}prefetch(t){if(t==null)throw new RangeError(\"`Dataset.prefetch()` requires bufferSize to be specified.\");let e=this;return kn(async()=>(await e.iterator()).prefetch(t),this.size)}repeat(t){let e=this,n;return this.size!=null&&t>0?n=this.size*t:t===0?n=0:this.size!=null&&(t===void 0||t<0)?n=1/0:n=null,kn(async()=>{let o=Lh(async()=>({value:await e.iterator(),done:!1}));return xR(o.take(t))},n)}skip(t){let e=this,n;return this.size!=null&&t>=0&&this.size>=t?n=this.size-t:this.size!=null&&(this.size<t||t===void 0||t<0)?n=0:n=null,kn(async()=>(await e.iterator()).skip(t),n)}shuffle(t,e,n=!0){if(t==null||t<0)throw this.size==null?new RangeError(\"`Dataset.shuffle()` requires bufferSize to be specified.\"):new RangeError(`\\`Dataset.shuffle()\\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \\`tf.Tensor\\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let o=this,s=bR.alea(e||y.now().toString());return kn(async()=>{let i=s.int32();return n&&(i+=s.int32()),(await o.iterator()).shuffle(t,i.toString())},this.size)}take(t){let e=this,n;return this.size!=null&&this.size>t?n=t:this.size!=null&&this.size<=t?n=this.size:n=null,kn(async()=>(await e.iterator()).take(t),n)}async toArray(){if(this.size===1/0)throw new Error(\"Can not convert infinite data stream to array.\");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error(\"Can not convert infinite data stream to array.\");return(await this.iterator()).toArrayForTest()}};Js.MAX_BUFFER_SIZE=1e4;function kn(r,t=null){return new class extends Js{constructor(){super(...arguments),this.size=t}async iterator(){return r()}}}function wR(r){return kn(async()=>AN(r),r.length)}function CR(r){if(!vu(r))throw new Error(\"The argument to zip() must be an object or array.\");let t;if(Array.isArray(r))for(let e=0;e<r.length;e++)t=t==null?r[e].size:Math.min(t,r[e].size);else if(r instanceof Object)for(let e in r)t=t==null?r[e].size:Math.min(t,r[e].size);return kn(async()=>{let e=await Wb(r,n=>{if(n instanceof Js)return{value:n.iterator(),recurse:!1};if(vu(n))return{value:null,recurse:!0};throw new Error(\"Leaves of the structure passed to zip() must be Datasets, not primitives.\")});return yR(e,fl.SHORTEST)},t)}function W7(r){if(r===null)return null;let t=r[0];return dR(t)?{value:U7(r),recurse:!1}:{value:null,recurse:!0}}function U7(r){if(r.length===0)throw new Error(\"Can't make a batch of zero elements.\");return r[0]instanceof Ft?nr(r):ur(r)}var Xf=class extends Js{constructor(t){super(),this.input=t}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`\n`).map(o=>(o.endsWith(\"\\r\")&&(o=o.slice(0,-1)),o))}};var Kb='\"',Mh=Symbol(\"out\"),IR=Symbol(\"field\"),jb=Symbol(\"quote\"),$N=Symbol(\"quoteafterquote\"),SR=Symbol(\"quoteinquote\"),Yf=class extends Js{constructor(t,e){super(),this.input=t,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=\",\",this.delimWhitespace=!1,this.base=new Xf(t),e||(e={}),this.hasHeader=e.hasHeader!==!1,this.fullColumnNames=e.columnNames,this.columnConfigs=e.columnConfigs,this.configuredColumnsOnly=e.configuredColumnsOnly,e.delimWhitespace?(y.assert(e.delimiter==null,()=>\"Delimiter should not be provided when delimWhitespace is true.\"),this.delimWhitespace=!0,this.delimiter=\" \"):this.delimiter=e.delimiter?e.delimiter:\",\"}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let t=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!t)throw new Error(\"Column names must be provided if there is no header line.\");this.fullColumnNames&&t&&y.assert(t.length===this.fullColumnNames.length,()=>\"The length of provided columnNames (\"+this.fullColumnNames.length.toString()+\") does not match the length of the header line read from file (\"+t.length.toString()+\").\"),this.fullColumnNames||(this.fullColumnNames=t);let e=this.fullColumnNames.reduce((o,s)=>(o[s]=o[s]+1||1,o),{}),n=Object.keys(e).filter(o=>e[o]>1);if(y.assert(n.length===0,()=>\"Duplicate column names found: \"+n.toString()),this.columnConfigs){for(let o of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(o)===-1)throw new Error('The key \"'+o+'\" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+\").\")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error(\"No data was found for CSV parsing.\");let n=e.value;return this.parseRow(n,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let t=await this.base.iterator();return this.hasHeader&&(t=t.skip(1)),t.map(e=>this.makeDataElement(e))}makeDataElement(t){let e=this.parseRow(t),n={},o={};for(let s=0;s<this.fullColumnNames.length;s++){let i=this.fullColumnNames[s],a=this.columnConfigs?this.columnConfigs[i]:null;if(!(this.configuredColumnsOnly&&!a)){let u=e[s],l=null;if(u===\"\")if(a&&a.default!==void 0)l=a.default;else{if(a&&(a.required||a.isLabel))throw new Error(`Required column ${i} is empty in this line: ${t}`);l=void 0}else{let c=Number(u);if(isNaN(c))a&&a.dtype===\"bool\"?l=this.getBoolean(u):l=u;else if(!a||!a.dtype)l=c;else switch(a.dtype){case\"float32\":l=c;break;case\"int32\":l=Math.floor(c);break;case\"bool\":l=this.getBoolean(u);break;default:l=c}}a&&a.isLabel?o[i]=l:n[i]=l}}return Object.keys(o).length===0?n:{xs:n,ys:o}}getBoolean(t){return t===\"1\"||t.toLowerCase()===\"true\"?1:0}parseRow(t,e=!0){let n=[],o=0,s=t.length,i=Mh;for(let a=0;a<s;a++)switch(i){case Mh:switch(t.charAt(a)){case Kb:o=a+1,i=jb;break;case this.delimiter:if(o=a+1,this.delimiter===\" \"&&this.delimWhitespace)break;n.push(\"\"),i=Mh;break;default:i=IR,o=a;break}break;case IR:switch(t.charAt(a)){case this.delimiter:n.push(t.substring(o,a)),i=Mh,o=a+1;break;default:}break;case jb:switch(t.charAt(a)){case Kb:i=$N;break;default:}break;case $N:switch(t.charAt(a)){case this.delimiter:n.push(t.substring(o,a-1)),i=Mh,o=a+1;break;case Kb:i=jb;break;default:i=SR;break}break;case SR:switch(t.charAt(a)){case Kb:i=jb;break;default:}break;default:}if(i===$N?n.push(t.substring(o,s-1)):n.push(t.substring(o)),e&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}};var Zf=class extends Je{constructor(t){super(),this.microphoneConfig=t,this.isClosed=!1,this.fftSize=t.fftSize||1024;let e=Math.log2(this.fftSize);if(this.fftSize<0||e<4||e>14||!Number.isInteger(e))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=t.numFramesPerSpectrogram||43,this.sampleRateHz=t.sampleRateHz,this.columnTruncateLength=t.columnTruncateLength||this.fftSize,this.audioTrackConstraints=t.audioTrackConstraints,this.smoothingTimeConstant=t.smoothingTimeConstant||0,this.includeSpectrogram=t.includeSpectrogram!==!1,this.includeWaveform=t.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error(\"Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.\")}summary(){return\"microphone\"}static async create(t={}){if(!z().get(\"IS_BROWSER\"))throw new Error(\"microphone API is only supported in browser environment.\");let e=new Zf(t);return await e.start(),e}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(n){throw new Error(`Error thrown while initializing video stream: ${n.message}`)}if(!this.stream)throw new Error(\"Could not obtain audio from microphone.\");let t=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new t,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let e=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,e.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let t,e,n=await this.getAudioData();if(this.includeSpectrogram){let o=this.flattenQueue(n.freqDataQueue);t=this.getTensorFromAudioDataArray(o,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let o=this.flattenQueue(n.timeDataQueue);e=this.getTensorFromAudioDataArray(o,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:t,waveform:e},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let t=[],e=[],n=0;return new Promise(o=>{let s=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&o({freqDataQueue:t,timeDataQueue:e}),t.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),e.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(s),o({freqDataQueue:t,timeDataQueue:e}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error(\"Can not convert infinite audio stream to array.\")}getSampleRate(){return this.sampleRateHz}flattenQueue(t){let e=t[0].length,n=new Float32Array(t.length*e);return t.forEach((o,s)=>n.set(o,s*e)),n}getTensorFromAudioDataArray(t,e){let n=new Float32Array(y.sizeFromShape(e));return n.set(t,n.length-t.length),ur(n,e)}};var Jf=class extends Je{constructor(t,e){if(super(),this.webcamVideoElement=t,this.webcamConfig=e,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Me([0],\"int32\"),this.webcamConfig.centerCrop){let n=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,o=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,s=(1-n)/2,i=(1-o)/2,a=s+n,u=o+i;this.cropBox=Vs([i,s,u,a],[1,4])}else this.cropBox=Vs([0,0,1,1],[1,4])}summary(){return\"webcam\"}static async create(t,e={}){if(!z().get(\"IS_BROWSER\"))throw new Error(\"tf.data.webcam is only supported in browser environment.\");if(!t){if(t=document.createElement(\"video\"),!e.resizeWidth||!e.resizeHeight)throw new Error(\"Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.\");t.width=e.resizeWidth,t.height=e.resizeHeight}let n=new Jf(t,e);return await n.start(),n}async start(){this.webcamConfig.facingMode&&y.assert(this.webcamConfig.facingMode===\"user\"||this.webcamConfig.facingMode===\"environment\",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:\"user\",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(t){throw t.message=`Error thrown while initializing video stream: ${t.message}`,t}if(!this.stream)throw new Error(\"Could not obtain video from webcam.\");try{this.webcamVideoElement.srcObject=this.stream}catch(t){console.log(t),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(t=>{this.webcamVideoElement.onloadedmetadata=()=>{t()}})}async next(){if(this.isClosed)return{value:null,done:!0};let t;try{t=nx.fromPixels(this.webcamVideoElement)}catch(e){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(t),done:!1}}catch(e){throw new Error(`Error thrown cropping the video: ${e.message}`)}finally{t.dispose()}else return{value:t,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(t){return B(()=>{let e=rr(J(t,\"float32\"),0),n;n=Gs.cropAndResize(e,this.cropBox,this.cropBoxInd,this.cropSize,\"bilinear\");let o=n.shape;return R(n,o.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error(\"Can not convert infinite video stream to array.\")}};var Qf=class{};var zh=class extends Je{split(t){return new DN(this,t)}},DN=class extends zh{constructor(t,e){super(),this.upstream=t,this.impl=new RN(t,e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},RN=class extends kc{constructor(t,e){super(),this.upstream=t,this.separator=e,this.carryover=\"\"}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let t=await this.upstream.next();if(t.done)return this.carryover===\"\"?!1:(this.outputQueue.push(this.carryover),this.carryover=\"\",!0);let e=t.value.split(this.separator);e[0]=this.carryover+e[0];for(let n of e.slice(0,-1))this.outputQueue.push(n);return this.carryover=e[e.length-1],!0}};var Xb=class extends Je{decodeUTF8(){return new FN(this)}},FN=class extends zh{constructor(t){super(),this.upstream=t,this.impl=new ON(t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},ON=class extends kc{constructor(t){if(super(),this.upstream=t,z().get(\"IS_BROWSER\"))this.decoder=new TextDecoder(\"utf-8\");else{let{StringDecoder:e}=gN();this.decoder=new e(\"utf8\")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let t=await this.upstream.next(),e;if(t.done)return!1;e=t.value;let n;return z().get(\"IS_BROWSER\")?n=this.decoder.decode(e,{stream:!0}):n=this.decoder.write(Buffer.from(e.buffer)),this.outputQueue.push(n),!0}};var td=class extends Xb{constructor(t,e={}){super(),this.file=t,this.options=e,y.assert(t instanceof Uint8Array||(z().get(\"IS_BROWSER\")?t instanceof File||t instanceof Blob:!1),()=>\"FileChunkIterator only supports File, Blob and Uint8Array right now.\"),this.offset=e.offset||0,this.chunkSize=e.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,n)=>{let o=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,o)));else{let s=new FileReader;s.onload=a=>{let u=s.result;if(u instanceof ArrayBuffer&&(u=new Uint8Array(u)),!(u instanceof Uint8Array))return n(new TypeError(\"FileReader returned unknown type.\"));e(u)},s.onabort=a=>n(new Error(\"Aborted\")),s.onerror=a=>n(new Error(a.type));let i=this.file.slice(this.offset,o);s.readAsArrayBuffer(i)}this.offset=o}),done:!1}}};async function vR(r,t={},e){let n,o;typeof r==\"string\"?n=r:(n=r.url,o=H7(r));let s=await(e||y.fetch)(n,o);if(s.ok){let i=new Uint8Array(await s.arrayBuffer());return new td(i,t)}else throw new Error(s.statusText)}var H7=r=>({method:r.method,headers:r.headers,body:r.body,mode:r.mode,credentials:r.credentials,cache:r.cache,redirect:r.redirect,referrer:r.referrer,integrity:r.integrity});function Yb(r){return typeof r==\"string\"&&r.slice(0,7)===\"file://\"}var ed=class extends Qf{constructor(t,e={}){super(),this.input=t,this.options=e}async iterator(){if(Yb(this.input)&&z().get(\"IS_NODE\")){let t=Zb();this.input=t.readFileSync(this.input.slice(7))}return new td(this.input,this.options)}};var rd=class extends Qf{constructor(t,e={}){super(),this.url=t,this.fileOptions=e}async iterator(){return Yb(this.url)?new ed(this.url,this.fileOptions).iterator():vR(this.url,this.fileOptions)}};function NR(r,t={}){return new Yf(new rd(r),t)}function TR(r){let t=Lh(r);return kn(async()=>t)}function kR(r){return kn(async()=>{let t=await r();return Lh(()=>t.next())})}async function ER(r,t){return Jf.create(r,t)}async function _R(r){return Zf.create(r)}var PN=\"4.0.0\";function tt(r,t){Array.isArray(r)||(r=[r]),r.forEach(e=>{e!=null&&y.assert(e.dtype!==\"complex64\",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var q7=Ur.whereImpl,Nu=class extends zo{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new ra(this,Pn())}nextDataId(){return Nu.nextDataId++}write(t,e,n){this.firstUse&&(this.firstUse=!1,z().get(\"IS_NODE\")&&v.warn(`\n============================\nHi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. \n============================`));let o={id:this.nextDataId()};return this.data.set(o,{values:t,dtype:n,refCount:1}),o}makeTensorInfo(t,e,n){let o;if(e===\"string\"&&n!=null&&n.length>0&&y.isString(n[0])){let s=n.map(i=>y.encodeString(i));o=this.write(s,t,e)}else o=this.write(n,t,e);return{dataId:o,shape:t,dtype:e}}refCount(t){return this.data.has(t)?this.data.get(t).refCount:0}incRef(t){let e=this.data.get(t);e.refCount++}decRef(t){if(this.data.has(t)){let e=this.data.get(t);e.refCount--}}move(t,e,n,o,s){this.data.set(t,{values:e,dtype:o,refCount:s})}numDataIds(){return this.data.numDataIds()}async read(t){return this.readSync(t)}readSync(t){let{dtype:e,complexTensorInfos:n}=this.data.get(t);if(e===\"complex64\"){let o=this.readSync(n.real.dataId),s=this.readSync(n.imag.dataId);return v.mergeRealAndImagArrays(o,s)}return this.data.get(t).values}bufferSync(t){let e=this.readSync(t.dataId);if(t.dtype===\"string\")try{let n=e.map(o=>y.decodeString(o));return wt(t.shape,t.dtype,n)}catch(n){throw new Error(\"Failed to decode encoded string bytes into utf-8\")}return wt(t.shape,t.dtype,e)}makeOutput(t,e,n){return Pn().makeTensorFromTensorInfo(this.makeTensorInfo(e,n,t),this)}disposeData(t,e=!1){if(this.data.has(t)){if(this.data.get(t).refCount--,!e&&this.data.get(t).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(t);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(t)}return!0}disposeIntermediateTensorInfo(t){this.disposeData(t.dataId)}async time(t){let e=y.now();return t(),{kernelMs:y.now()-e}}memory(){return{unreliable:!0,reasons:[\"The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less.\"]}}where(t){tt([t],\"where\");let e=this.readSync(t.dataId);return q7(t.shape,e)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};Nu.nextDataId=0;var mw={};Wt(mw,{addImpl:()=>zN,bincountImpl:()=>sd,bincountReduceImpl:()=>Jb,castImpl:()=>MN,ceilImpl:()=>BN,concatImpl:()=>Ec,equalImpl:()=>VN,expImpl:()=>WN,expm1Impl:()=>HN,floorImpl:()=>qN,gatherNdImpl:()=>Qb,gatherV2Impl:()=>tw,greaterEqualImpl:()=>jN,greaterImpl:()=>KN,lessEqualImpl:()=>YN,lessImpl:()=>XN,linSpaceImpl:()=>ew,logImpl:()=>ZN,maxImpl:()=>rw,maximumImpl:()=>JN,minimumImpl:()=>QN,multiplyImpl:()=>Bh,negImpl:()=>tT,notEqualImpl:()=>eT,prodImpl:()=>rT,raggedGatherImpl:()=>nw,raggedRangeImpl:()=>ow,raggedTensorToTensorImpl:()=>sw,rangeImpl:()=>Ac,rsqrtImpl:()=>nT,scatterImpl:()=>dl,sigmoidImpl:()=>sF,simpleAbsImpl:()=>LN,sliceImpl:()=>$c,sparseFillEmptyRowsImpl:()=>iw,sparseReshapeImpl:()=>aw,sparseSegmentReductionImpl:()=>ld,sqrtImpl:()=>lF,squaredDifferenceImpl:()=>sT,stridedSliceImpl:()=>lw,stringNGramsImpl:()=>Dc,stringSplitImpl:()=>Rc,stringToHashBucketFastImpl:()=>Fc,subImpl:()=>aT,tileImpl:()=>uw,topKImpl:()=>cw,transposeImpl:()=>id,uniqueImpl:()=>pw});function LN(r){let t=new Float32Array(r.length);for(let e=0;e<r.length;++e)t[e]=Math.abs(r[e]);return t}var K7=r=>{let{x:t}=r.inputs,e=r.backend;tt(t,\"abs\");let n=new Float32Array(y.sizeFromShape(t.shape)),o=e.data.get(t.dataId).values;return n=LN(o),e.makeOutput(n,t.shape,t.dtype)},$R={kernelName:ii,backendName:\"cpu\",kernelFunc:K7};function Qt(r){return(t,e,n,o,s)=>{let i=v.assertAndGetBroadcastShape(t,e),a=i.length,u=y.computeStrides(i),l=y.sizeFromShape(i),c=y.getTypedArrayFromDType(s,l),p=t.length,m=e.length,f=y.computeStrides(t),d=y.computeStrides(e),h=v.getBroadcastDims(t,i),g=v.getBroadcastDims(e,i);if(h.length+g.length===0)for(let x=0;x<c.length;++x)c[x]=r(n[x%n.length],o[x%o.length]);else for(let x=0;x<c.length;++x){let b=y.indexToLoc(x,a,u),w=b.slice(-p);h.forEach(A=>w[A]=0);let C=y.locToIndex(w,p,f),N=b.slice(-m);g.forEach(A=>N[A]=0);let _=y.locToIndex(N,m,d);c[x]=r(n[C],o[_])}return[c,i]}}function wr(r){let{inputs:t,backend:e}=r,{real:n,imag:o}=t,s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,a=e.makeTensorInfo(n.shape,\"complex64\"),u=e.data.get(a.dataId);return u.complexTensorInfos={real:e.makeTensorInfo(n.shape,\"float32\",s),imag:e.makeTensorInfo(o.shape,\"float32\",i)},a}var DR={kernelName:pp,backendName:\"cpu\",kernelFunc:wr};function nd(r,t,e=\"float32\"){if(e===\"complex64\"){let o=nd(r,t,\"float32\"),s=nd(r,t,\"float32\");return wr({inputs:{real:o,imag:s},backend:r})}let n=y.makeZerosTypedArray(y.sizeFromShape(t),e);return r.makeTensorInfo(t,e,n)}function Kr(r){let{inputs:t,backend:e}=r,{x:n}=t;return e.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var RR={kernelName:co,backendName:\"cpu\",kernelFunc:Kr};function Ao(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.data.get(n.dataId).complexTensorInfos.real,s=e.data.get(o.dataId).values;return e.makeTensorInfo(o.shape,o.dtype,s)}var FR={kernelName:Rp,backendName:\"cpu\",kernelFunc:Ao};function MN(r,t,e,n){if(n===\"int32\"){let o=Int32Array.from(r);return[t,\"int32\",o]}if(n===\"bool\"){let o=y.toTypedArray([0],e),[s,i]=Qt((a,u)=>a!==u?1:0)(t,[],r,o,\"bool\");return[i,\"bool\",s]}throw new Error(`Error in Cast: failed to cast ${e} to ${n}`)}function $o(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dtype:s}=n;if(s===\"complex64\"){if(o.dtype===\"complex64\")return Kr({inputs:{x:o},backend:e});let c=nd(e,o.shape,o.dtype),p=$o({inputs:{x:o},backend:e,attrs:{dtype:\"float32\"}}),m=wr({inputs:{real:p,imag:c},backend:e});return e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),m}if(o.dtype===\"complex64\"){let c=Ao({inputs:{input:o},backend:e}),p=$o({inputs:{x:c},backend:e,attrs:{dtype:s}});return e.disposeIntermediateTensorInfo(c),p}if(!y.hasEncodingLoss(o.dtype,s)){let c=Kr({inputs:{x:o},backend:e});return{dataId:c.dataId,shape:c.shape,dtype:s}}let i=e.data.get(o.dataId).values,[a,u,l]=MN(i,o.shape,o.dtype,s);return e.makeTensorInfo(a,u,l)}var OR={kernelName:lo,backendName:\"cpu\",kernelFunc:$o};function oe(r,t,e,n){return e==null?({inputs:o,backend:s})=>{let{a:i,b:a}=o,u=s;tt([i,a],r);let l=u.data.get(i.dataId).values,c=u.data.get(a.dataId).values,p=i.dtype===\"string\"?v.fromUint8ToStringArray(l):l,m=i.dtype===\"string\"?v.fromUint8ToStringArray(c):c,f=n||i.dtype,[d,h]=t(i.shape,a.shape,p,m,f);return u.makeTensorInfo(h,f,d)}:({inputs:o,backend:s})=>{let{a:i,b:a}=o,u=s;if(i.dtype===\"complex64\"||a.dtype===\"complex64\"){let l=$o({inputs:{x:i},backend:u,attrs:{dtype:\"complex64\"}}),c=u.data.get(l.dataId),p=c.complexTensorInfos.real,m=c.complexTensorInfos.imag,f=u.data.get(p.dataId).values,d=u.data.get(m.dataId).values,h=$o({inputs:{x:a},backend:u,attrs:{dtype:\"complex64\"}}),g=u.data.get(h.dataId),x=g.complexTensorInfos.real,b=g.complexTensorInfos.imag,w=u.data.get(x.dataId).values,C=u.data.get(b.dataId).values,[N,_,A]=e(i.shape,a.shape,f,d,w,C),$=u.makeTensorInfo(A,\"float32\",N),F=u.makeTensorInfo(A,\"float32\",_),P=wr({inputs:{real:$,imag:F},backend:u});return u.disposeIntermediateTensorInfo(l),u.disposeIntermediateTensorInfo(h),u.disposeIntermediateTensorInfo($),u.disposeIntermediateTensorInfo(F),P}else{let l=u.data.get(i.dataId).values,c=u.data.get(a.dataId).values,p=n||i.dtype,[m,f]=t(i.shape,a.shape,l,c,p);return u.makeTensorInfo(f,p,m)}}}function od(r){return(t,e,n,o,s,i)=>{let a=v.assertAndGetBroadcastShape(t,e),u=y.sizeFromShape(a),l=a.length,c=y.computeStrides(a),p=y.getTypedArrayFromDType(\"float32\",u),m=y.getTypedArrayFromDType(\"float32\",u),f=v.getBroadcastDims(t,a),d=v.getBroadcastDims(e,a),h=v.mergeRealAndImagArrays(n,o),g=v.mergeRealAndImagArrays(s,i),x=t.length,b=y.computeStrides(t),w=e.length,C=y.computeStrides(e);if(f.length+d.length===0)for(let N=0;N<p.length;N++){let _=N%h.length,A=N%g.length,$=r(h[_*2],h[_*2+1],g[A*2],g[A*2+1]);p[N]=$.real,m[N]=$.imag}else for(let N=0;N<p.length;N++){let _=y.indexToLoc(N,l,c),A=_.slice(-x);f.forEach(G=>A[G]=0);let $=y.locToIndex(A,x,b),F=_.slice(-w);d.forEach(G=>F[G]=0);let P=y.locToIndex(F,w,C),V=r(h[$*2],h[$*2+1],g[P*2],g[P*2+1]);p[N]=V.real,m[N]=V.imag}return[p,m,a]}}var zN=Qt((r,t)=>r+t),j7=od((r,t,e,n)=>({real:r+e,imag:t+n})),Ki=oe(Zn,zN,j7),PR={kernelName:Zn,backendName:\"cpu\",kernelFunc:Ki};function sd(r,t,e,n,o){let s=y.sizeFromShape(n),i=y.makeZerosTypedArray(o,e);for(let a=0;a<r.length;a++){let u=r[a];if(u<0)throw new Error(\"Input x must be non-negative!\");u>=o||(s>0?i[u]+=t[a]:i[u]+=1)}return i}function Jb(r,t,e,n=!1){let o=r.shape[0],s=r.shape[1],i=wt([o,e],t.dtype);for(let a=0;a<o;a++)for(let u=0;u<s;u++){let l=r.get(a,u);if(l<0)throw new Error(\"Input x must be non-negative!\");l>=e||(n?i.set(1,a,l):t.size>0?i.set(i.get(a,l)+t.get(a,u),a,l):i.set(i.get(a,l)+1,a,l))}return i}function yn(r){return(t,e,n)=>{let o=y.getTypedArrayFromDType(e,t.length);for(let s=0;s<t.length;++s)o[s]=r(t[s],n);return o}}function kt(r,t,e){return({inputs:n,attrs:o,backend:s})=>{let{x:i}=n;if(tt(i,r),i.dtype===\"string\"||e===\"string\")throw new Error(\"unaryKernelFunc does not support string input/output\");let a=s,u=a.data.get(i.dataId).values,l=y.sizeFromShape(i.shape),c=e||i.dtype,p=y.getArrayFromDType(c,l);for(let m=0;m<l;++m)p[m]=t(u[m],o);return a.makeTensorInfo(i.shape,c,p)}}function Do(r,t,e){return({inputs:n,attrs:o,backend:s})=>{let{x:i}=n;if(tt(i,r),i.dtype===\"string\"||e===\"string\")throw new Error(\"unaryKernelFunc does not support string input/output\");let a=s,u=a.data.get(i.dataId).values,l=e||i.dtype,c=t(u,l,o);return a.makeTensorInfo(i.shape,l,c)}}var BN=yn(r=>Math.ceil(r)),X7=Do(qo,BN),LR={kernelName:qo,backendName:\"cpu\",kernelFunc:X7};function Ec(r,t,e,n){let o=y.getArrayFromDType(e,y.sizeFromShape(t));if(n&&e!==\"string\"){let s=0;r.forEach(i=>{let a=y.sizeFromShape(i.shape);o.set(i.vals,s),s+=a})}else{let s=0;r.forEach(i=>{let a=e===\"string\"?v.fromUint8ToStringArray(i.vals):i.vals,u=0;for(let l=0;l<i.shape[0];++l){let c=l*t[1]+s;for(let p=0;p<i.shape[1];++p)o[c+p]=a[u++]}s+=i.shape[1]})}return o}var VN=Qt((r,t)=>r===t?1:0),GN=oe(xa,VN,null,\"bool\"),MR={kernelName:xa,backendName:\"cpu\",kernelFunc:GN};var WN=yn(r=>Math.exp(r)),UN=Do(es,WN,\"float32\"),zR={kernelName:es,backendName:\"cpu\",kernelFunc:UN};var HN=yn(r=>Math.expm1(r)),Y7=Do(ya,HN),BR={kernelName:ya,backendName:\"cpu\",kernelFunc:Y7};var qN=yn(r=>Math.floor(r)),Z7=Do(rs,qN),VR={kernelName:rs,backendName:\"cpu\",kernelFunc:Z7};function Qb(r,t,e,n,o,s,i,a,u){let l=wt([n,s],e);for(let c=0;c<n;c++){let p=[],m=0;for(let f=0;f<o;f++){let d=r[c*o+f];m+=d*i[f],p.push(d)}if(m<0||m>=u/s)throw new Error(`Invalid indices: ${p} does not index into ${a}`);for(let f=0;f<s;f++)l.values[c*s+f]=t.get(...t.indexToLoc(m*s+f))}return l}function tw(r,t,e){let n=wt(e,r.dtype);for(let o=0;o<n.size;++o){let i=n.indexToLoc(o).slice(),a=i[0],u=i[2],l=t.locToIndex([a,u]);i[2]=t.values[l];let c=r.locToIndex(i);0<=c&&c<r.values.length&&(n.values[o]=r.values[c])}return n}var KN=Qt((r,t)=>r>t?1:0),J7=oe(Ca,KN,null,\"bool\"),GR={kernelName:Ca,backendName:\"cpu\",kernelFunc:J7};var jN=Qt((r,t)=>r>=t?1:0),Q7=oe(ss,jN,null,\"bool\"),WR={kernelName:ss,backendName:\"cpu\",kernelFunc:Q7};var XN=Qt((r,t)=>r<t?1:0),tJ=oe(Na,XN,null,\"bool\"),UR={kernelName:Na,backendName:\"cpu\",kernelFunc:tJ};var YN=Qt((r,t)=>r<=t?1:0),eJ=oe(Ta,YN,null,\"bool\"),HR={kernelName:Ta,backendName:\"cpu\",kernelFunc:eJ};function ew(r,t,e){let n=(t-r)/(e-1),o=y.makeZerosTypedArray(e,\"float32\");o[0]=r;for(let s=1;s<o.length;s++)o[s]=o[s-1]+n;return o}var ZN=yn(r=>Math.log(r)),rJ=Do(as,ZN),qR={kernelName:as,backendName:\"cpu\",kernelFunc:rJ};function rw(r,t,e,n){let o=y.getTypedArrayFromDType(n,y.sizeFromShape(e));for(let s=0;s<o.length;++s){let i=s*t,a=r[i];for(let u=0;u<t;++u){let l=r[i+u];(Number.isNaN(l)||l>a)&&(a=l)}o[s]=a}return o}var JN=Qt((r,t)=>Math.max(r,t)),nJ=oe(us,JN),KR={kernelName:us,backendName:\"cpu\",kernelFunc:nJ};var QN=Qt((r,t)=>Math.min(r,t)),oJ=oe(fs,QN),jR={kernelName:fs,backendName:\"cpu\",kernelFunc:oJ};var Bh=Qt((r,t)=>r*t),sJ=od((r,t,e,n)=>({real:r*e-t*n,imag:r*n+t*e})),_c=oe(hs,Bh,sJ),XR={kernelName:hs,backendName:\"cpu\",kernelFunc:_c};function tT(r,t,e){let n=y.createScalarValue(-1,e);return Bh([],t,n,r,e)}function iJ(r){let{inputs:t,backend:e}=r,{x:n}=t;tt(n,\"neg\");let o=e.data.get(n.dataId).values,[s,i]=tT(o,n.shape,n.dtype);return e.makeTensorInfo(i,n.dtype,s)}var YR={kernelName:pi,backendName:\"cpu\",kernelFunc:iJ};var eT=Qt((r,t)=>r!==t?1:0),aJ=oe(Da,eT,null,\"bool\"),ZR={kernelName:Da,backendName:\"cpu\",kernelFunc:aJ};function id(r,t,e,n,o){let s=t.length,i=y.sizeFromShape(t),a=y.computeStrides(t),u=y.computeStrides(o),l=y.getTypedArrayFromDType(e,y.sizeFromShape(o));for(let c=0;c<i;++c){let p=y.indexToLoc(c,s,a),m=new Array(p.length);for(let d=0;d<m.length;d++)m[d]=p[n[d]];let f=y.locToIndex(m,s,u);l[f]=r[c]}return l}function Ve(r){let{inputs:t,attrs:e,backend:n}=r,{x:o}=t,{perm:s}=e;tt(o,\"transpose\");let i=o.shape.length,a=new Array(i);for(let p=0;p<a.length;p++)a[p]=o.shape[s[p]];let u=n.data.get(o.dataId).values,l=id(u,o.shape,o.dtype,s,a);return{dataId:n.write(l,a,o.dtype),shape:a,dtype:o.dtype}}var JR={kernelName:Qn,backendName:\"cpu\",kernelFunc:Ve};function rT(r,t,e,n){let[o,s]=v.computeOutAndReduceShapes(r,n),i=sr(t,\"int32\"),a=y.makeZerosTypedArray(y.sizeFromShape(o),i),u=y.sizeFromShape(s);for(let l=0;l<a.length;++l){let c=l*u,p=1;for(let m=0;m<u;++m)p*=e[c+m];a[l]=p}return{outVals:a,outShape:o,outDtype:i}}function lJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,\"prod\");let a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=v.getAxesPermutation(u,a),c=u,p=o,m=[];l!=null&&(p=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),m.push(p),c=v.getInnerMostAxes(c.length,a));let f=e.data.get(p.dataId).values,{outVals:d,outShape:h,outDtype:g}=rT(p.shape,p.dtype,f,c),x=h;return i&&(x=v.expandShapeToKeepDim(h,u)),m.forEach(b=>e.disposeIntermediateTensorInfo(b)),e.makeTensorInfo(x,g,d)}var QR={kernelName:ws,backendName:\"cpu\",kernelFunc:lJ};function uJ(r,t,e){r.forEach((n,o)=>{if(n<0||n>=e){let s=y.indexToLoc(o,t.length,y.computeStrides(t)).join(\",\");throw new Error(`indices[${s}] = ${n} is not in [0, ${e})`)}})}function cJ(r,t){for(let e=0;e<r.length;++e){let n=r[e],o=e===r.length-1?t:r[e+1].length;if(n.length===0)throw new Error(\"Ragged splits may not be empty\");if(n[0]<0)throw new Error(\"Ragged splits must be non-negative\");if(n[n.length-1]>o)throw new Error(\"Ragged splits must not point past values\");for(let s=1;s<n.length;++s)if(n[s-1]>n[s])throw new Error(\"Ragged splits must be sorted in ascending order\")}}function pJ(r,t,e,n){let o=[],s=0,i=t.length-1+e.length,a=new Array(i).fill(null).map(()=>[0]);cJ(e,n);let u=1;for(let l=0;l<t.length-1;++l){u*=t[l];let c=t[l+1];for(let p=1;p<u+1;++p)a[l].push(p*c)}for(let l=0;l<r.length;++l){let c=r[l],p=r[l]+1;for(let m=0;m<e.length;++m){let f=e[m],d=m+t.length-1;if(d>=0){let h=a[d],g=h[h.length-1]-f[c];for(let x=c;x<p;++x)a[d].push(f[x+1]+g)}c=f[c],p=f[p]}p!==c&&(o.push([c,p]),s+=p-c)}return{outSplits:a,valueSlices:o,numValues:s}}function mJ(r){let t=[];for(let e=0;e<r.length;++e){let n=r[e].length,o=y.getArrayFromDType(\"int32\",n);t.push(o),r[e].forEach((s,i)=>o[i]=s)}return t}function tF(r,t){let e=r.slice(0,t);for(;e.length<t;)e.push(1);for(let n=t;n<r.length;n++)e[t-1]*=r[n];return e}function fJ(r,t,e,n,o,s){let i=tF(t,2)[1],a=tF(s,2)[1],u=0;for(let l of e)for(let c=l[0];c<l[1];++c){for(let p=0;p<n;++p)o[u*a+p]=r[c*i+p];++u}}function dJ(r,t,e,n,o){let s=t.slice();s[0]=o;let i=y.getArrayFromDType(e,y.sizeFromShape(s)),a=r.length,u=a===0?0:a/t[0];return fJ(r,t,n,u,i,s),[i,s]}function nw(r,t,e,n,o,s,i,a){if(r.length===0)throw new Error(\"paramsNestedSplits must be non empty\");if(t[0].length===0)throw new Error(\"Split tensors must not be scalars\");let u=t[0][0]-1;if(uJ(s,i,u),n.length===0)throw new Error(\"params.rank must be nonzero\");let l=n[0],{outSplits:c,valueSlices:p,numValues:m}=pJ(s,i,r,l),f=mJ(c),d=dJ(e,n,o,p,m);return[f,d[0],d[1]]}var eF=2147483647;function ow(r,t,e,n,o,s,i){if(t.length>1)throw new Error(\"starts must be a scalar or vector\");if(o.length>1)throw new Error(\"limits must be a scalar or vector\");if(i.length>1)throw new Error(\"deltas must be a scalar or vector\");let a=t.length===0,u=o.length===0,l=i.length===0,c=[];a||c.push(t[0]),u||c.push(o[0]),l||c.push(i[0]);for(let g=1;g<c.length;++g)if(c[g]!==c[g-1])throw new Error(\"starts, limits, and deltas must have the same shape\");let p=c.length===0?1:c[0],m=y.getArrayFromDType(\"int32\",p+1);m[0]=0;for(let g=0;g<p;++g){let x=a?r[0]:r[g],b=u?n[0]:n[g],w=l?s[0]:s[g];if(w===0)throw new Error(\"Requires delta != 0\");let C;if(w>0&&b<x||w<0&&b>x)C=0;else if(C=Math.ceil(Math.abs((b-x)/w)),C>eF)throw new Error(`Requires ((limit - start) / delta) <= ${eF}`);m[g+1]=m[g]+C}let f=m[p],d=y.getArrayFromDType(e,f),h=0;for(let g=0;g<p;++g){let x=m[g+1]-m[g],b=a?r[0]:r[g],w=l?s[0]:s[g];for(let C=0;C<x;++C)d[h++]=b,b+=w}return[m,d]}var Ro=v.RowPartitionType,ad=class{constructor(t,e,n,o,s,i,a,u,l,c){this.shape=t,this.shapeShape=e,this.values=n,this.valuesShape=o,this.valuesDType=s,this.defaultValue=i,this.defaultValueShape=a,this.rowPartitionValues=u,this.rowPartitionValuesShapes=l,this.rowPartitionTypes=v.getRowPartitionTypesHelper(c),this.raggedRank=v.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(t){return this.rowPartitionTypes[0]===Ro.FIRST_DIM_SIZE?this.rowPartitionTypes[t+1]:this.rowPartitionTypes[t]}getRowPartitionTensor(t){return this.rowPartitionTypes[0]===Ro.FIRST_DIM_SIZE?this.rowPartitionValues[t+1]:this.rowPartitionValues[t]}getMaxWidth(t){let e=this.getRowPartitionTensor(t-1);switch(this.getRowPartitionTypeByDimension(t-1)){case Ro.VALUE_ROWIDS:return ad.getMaxWidthValueRowID(e);case Ro.ROW_SPLITS:return ad.getMaxWidthRowSplit(e);default:throw new Error(`Cannot handle partition type ${Ro[this.getRowPartitionTypeByDimension(t-1)]}`)}}static getMaxWidthRowSplit(t){let e=t.length;if(e===0||e===1)return 0;let n=0;for(let o=0;o<e-1;++o){let s=t[o+1]-t[o];s>n&&(n=s)}return n}static getMaxWidthValueRowID(t){let e=t.length;if(e===0)return 0;let n=0,o=t[0],s=0;for(let i=1;i<e;++i){let a=t[i];a!==o&&(o=a,s=Math.max(i-n,s),n=i)}return Math.max(e-n,s)}tensorShapeFromTensor(t,e,n=!0){if(e.length===0){if(t[0]===-1)return[];throw new Error(\"The only valid scalar shape tensor is the fully unknown shape specified as -1.\")}return nF(t,n)}calculateOutputSize(t){let e=this.valuesShape,n=this.defaultValueShape;v.validateDefaultValueShape(n,e);let o=this.tensorShapeFromTensor(this.shape,this.shapeShape),i=v.combineRaggedTensorToTensorShapes(this.raggedRank,o,e);i[0]<0&&(i[0]=t);for(let a=1;a<=this.raggedRank;++a)i[a]<0&&(i[a]=this.getMaxWidth(a));return i}calculateFirstParentOutputIndex(t,e,n){let o=Math.min(t,n),s=[],i=0;for(let a=0;a<o;++a,i+=e)s.push(i);for(let a=o;a<t;++a)s.push(-1);return y.assert(s.length===t,()=>\"Final length of result must be equal to firstDimension.\"),s}calculateOutputIndexRowSplit(t,e,n,o){let s=t.length,i=[];for(let a=0;a<s-1;++a){let u=t[a+1]-t[a],l=Math.min(o,u),c=e[a];c===-1&&(l=0);for(let p=0;p<l;++p)i.push(c),c+=n;for(let p=0;p<u-l;++p)i.push(-1)}if(s>0&&i.length!==t[s-1])throw new Error(\"Invalid row split size.\");return i}calculateOutputIndexValueRowID(t,e,n,o){let s=t.length,i=[];if(s===0)return[];let a=0,u=t[0];if(u>=e.length)throw new Error(`Got currentValueRowId=${u}, which is not less than ${e.length}`);let l=e[u];i.push(l);for(let c=1;c<s;++c){let p=t[c];if(p===u)l>=0&&(++a,a<o?l+=n:l=-1);else{if(a=0,u=p,p>=e.length)throw new Error(`Got nextValueRowId=${p} which is not less than ${e.length}`);l=e[p]}i.push(l)}if(i.length!==t.length)throw new Error(\"Invalid row ids.\");return i}calculateOutputIndex(t,e,n,o){let s=this.getRowPartitionTensor(t),i=this.getRowPartitionTypeByDimension(t);switch(i){case Ro.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(s,e,n,o);case Ro.ROW_SPLITS:if(s.length-1>e.length)throw new Error(`Row partition size is greater than output size: ${s.length-1} > ${e.length}`);return this.calculateOutputIndexRowSplit(s,e,n,o);default:throw new Error(`Unsupported partition type: ${Ro[i]}`)}}getFirstDimensionSize(){let t=this.rowPartitionValues[0];if(this.rowPartitionTypes.length===0)throw new Error(\"No row_partition_types given.\");let e=this.rowPartitionTypes[0];switch(e){case Ro.FIRST_DIM_SIZE:return t[0];case Ro.VALUE_ROWIDS:throw new Error(\"Cannot handle VALUE_ROWIDS in first dimension.\");case Ro.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${Ro[e]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error(\"Invalid first partition input. Tensor requires at least one element.\");let e=this.getFirstDimensionSize(),n=this.calculateOutputSize(e),o=new Array(this.raggedRank+1);o[o.length-1]=1;for(let u=o.length-2;u>=0;--u)o[u]=o[u+1]*n[u+1];let s=nF(n,!1),i=y.getArrayFromDType(this.valuesDType,y.sizeFromShape(s));if(o[0]*n[0]>0){let u=this.calculateFirstParentOutputIndex(e,o[0],n[0]);for(let l=1;l<=this.raggedRank;++l)u=this.calculateOutputIndex(l-1,u,o[l],n[l]);this.setOutput(this.raggedRank,u,i,s)}return[s,i]}setOutput(t,e,n,o){if(n.length===0)return;let s=this.values,i=n,a=o.slice();a=a.slice(t+1);let u=y.sizeFromShape(a),l=e.length,c=this.defaultValue;if(c.length!==u&&c.length!==1){let d=this.defaultValueShape;B(()=>{let h=R(c,d);c=Ri(h,a).dataSync()})}let p=0,m=0,f=0;for(let d=0;d<=l;++d){let h=d<l?e[d]:-1;if(h===f){++f;continue}if(m<f){let g=s.subarray(p*u),x=i.subarray(m*u),b=(f-m)*u;rF(x,g,b)}if(d>=l){let g=n.length;h=Math.floor(g/u)}if(h>f)if(this.defaultValue.length===1)i.subarray(f*u,h*u).fill(this.defaultValue[0]),f=h;else for(;h>f;){let g=i.slice(f*u);rF(g,c,u),++f}h<0?(p=d+1,m=f):(p=d,m=f,f=m+1)}}};function rF(r,t,e){for(let n=0;n<e;n++)r[n]=t[n]}function nF(r,t){let e=[];for(let n of r){if(n<0){if(!t)throw new Error(`Dimension ${n} must be >= 0`);if(n<-1)throw new Error(`Dimension ${n} must be >= -1`);n=-1}e.push(n)}return e}function sw(r,t,e,n,o,s,i,a,u,l){return new ad(r,t,e,n,o,s,i,a,u,l).compute()}function Ac(r,t,e,n){let o=r===t,s=r<t&&e<0,i=t<r&&e>1;if(o||s||i)return y.makeZerosTypedArray(0,n);let a=Math.abs(Math.ceil((t-r)/e)),u=y.makeZerosTypedArray(a,n);t<r&&e===1&&(e=-1),u[0]=r;for(let l=1;l<u.length;l++)u[l]=u[l-1]+e;return u}var nT=yn(r=>1/Math.sqrt(r)),hJ=Do(ks,nT),oF={kernelName:ks,backendName:\"cpu\",kernelFunc:hJ};function dl(r,t,e,n,o,s,i,a,u,l){let c=[n/o,o],p=r.values,m=t.values;if(n===0)return wt(e,t.dtype);let f=wt(c,t.dtype);typeof u==\"string\"||typeof u==\"number\"?f.values.fill(u):typeof u==\"boolean\"&&f.values.fill(+u);for(let d=0;d<s;d++){let h=[],g=0;for(let x=0;x<i;x++){let b=p[d*i+x];h.push(b),g+=b*a[x]}if(g<0||g>=n/o)throw new Error(`Invalid indices: ${h} does not index into ${e}`);for(let x=0;x<o;x++)l?f.values[g*o+x]+=m[d*o+x]:f.values[g*o+x]=t.rank===0?m[0]:m[d*o+x]}return f}var sF=yn(r=>1/(1+Math.exp(-r))),oT=kt(_s,r=>1/(1+Math.exp(-r))),iF={kernelName:_s,backendName:\"cpu\",kernelFunc:oT};function $c(r,t,e,n,o){let s=Le.isSliceContinous(n,t,e),i=y.sizeFromShape(e),a=y.computeStrides(n);if(s){let p=Le.computeFlatOffset(t,a);return o===\"string\"?r.slice(p,p+i):r.subarray(p,p+i)}let u=o===\"string\"?v.fromUint8ToStringArray(r):r,l=wt(n,o,u),c=wt(e,o);for(let p=0;p<c.size;++p){let m=c.indexToLoc(p),f=m.map((d,h)=>d+t[h]);c.set(l.get(...f),...m)}return o===\"string\"?v.fromStringArrayToUint8(c.values):c.values}function Fo(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,size:i}=n;tt(o,\"slice\");let[a,u]=Le.parseSliceParams(o,s,i);Le.assertParamsValid(o,a,u);let l=e.data.get(o.dataId).values,c=$c(l,a,u,o.shape,o.dtype);return e.makeTensorInfo(u,o.dtype,c)}var aF={kernelName:gi,backendName:\"cpu\",kernelFunc:Fo};function iw(r,t,e,n,o,s,i){let a=t[0],u=s[0],l=new Array(u),c=new Array(a),p=t[1];if(u===0){if(a!==0)throw new Error(v.getSparseFillEmptyRowsIndicesDenseShapeMismatch(a));let g=y.getArrayFromDType(e,0),x=y.getArrayFromDType(o,0);return[g,[0,p],x,l,c]}let m=!0,f=0,d=new Array(u).fill(0);for(let g=0;g<a;++g){let x=r[g*p];if(x<0)throw new Error(v.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,x));if(x>=u)throw new Error(v.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,x,u));++d[x],m=m&&x>=f,f=x}let h=!0;for(let g=0;g<u;++g){let x=d[g]===0;l[g]=x,h=h&&!x,d[g]=Math.max(d[g],1),g>0&&(d[g]+=d[g-1])}if(h&&m){let g=r,x=n;for(let b=0;b<a;++b)c[b]=b;return[g,[a,p],x,l,c]}else{let g=d[u-1],x=y.getArrayFromDType(e,g*p),b=y.getArrayFromDType(o,g),w=new Array(u).fill(0);for(let C=0;C<a;++C){let N=r[C*p],_=w[N],A=(N===0?0:d[N-1])+_;w[N]++;for(let $=0;$<p;++$)x[A*p+$]=r[C*p+$];b[A]=n[C],c[C]=A}for(let C=0;C<u;++C)if(w[C]===0){let _=C===0?0:d[C-1];x[_*p+0]=C;for(let A=1;A<p;++A)x[_*p+A]=0;b[_]=i}return[x,[g,p],b,l,c]}}function aw(r,t,e,n,o){let s=y.sizeFromShape(n),i=t[0],a=o.length,u=[],l=1,c=-1;for(let g=0;g<a;++g){let x=o[g];if(x===-1){if(c!==-1)throw new Error(v.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(c,g));c=g,u.push(1)}else{if(x<0)throw new Error(v.getSparseReshapeNegativeOutputDimErrorMessage(g,x));l*=x,u.push(x)}}if(c!==-1){if(l<=0)throw new Error(v.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let g=Math.trunc(s/l);if(l*g!==s)throw new Error(v.getSparseReshapeInputOutputMultipleErrorMessage(n,u));u[c]=g}if(y.sizeFromShape(u)!==s)throw new Error(v.getSparseReshapeInputOutputMismatchErrorMessage(n,u));let m=n.length,f=[];if(m>0){f[m-1]=1;for(let g=m-2;g>=0;--g)f[g]=f[g+1]*n[g+1]}let d=[];if(a>0){d[a-1]=1;for(let g=a-2;g>=0;--g)d[g]=d[g+1]*u[g+1]}let h=y.getArrayFromDType(e,i*a);for(let g=0;g<i;++g){let x=0;for(let b=0;b<m;++b)x+=r[g*m+b]*f[b];for(let b=0;b<a;++b)h[g*a+b]=Math.trunc(x/d[b]),x%=d[b]}return[h,[i,a],u]}function ld(r,t,e,n,o,s=!1,i=0){let a=n.length,u=[t[0],r.length/t[0]],l=u[1],p=a>0?o[a-1]+1:0;if(p<0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let m=t.slice();m[0]=p;let f=m.reduce((w,C)=>w*C,1),d=y.getArrayFromDType(e,f);if(a===0)return p>0&&d.fill(i),[d,m];if(p<=0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let h=0,g=1,x=0,b=o[h];for(;;){let w=0;if(g<a){if(w=o[g],b===w){++g;continue}if(b>=w)throw new Error(v.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(b<0||b>=p)throw new Error(v.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b,p));b>x&&d.fill(i,x*l,b*l);for(let C=h;C<g;++C){let N=n[C];if(N<0||N>=u[0])throw new Error(v.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(C,n[C],u[0]));for(let _=0;_<l;_++)d[b*l+_]+=r[N*l+_]}if(s)for(let C=0;C<l;C++)d[b*l+C]/=g-h;if(h=g,++g,x=b+1,b=w,g>a)break}return x<p&&d.fill(i,x*l,p*l),[d,m]}var lF=yn(r=>Math.sqrt(r)),gJ=kt(As,r=>Math.sqrt(r)),uF={kernelName:As,backendName:\"cpu\",kernelFunc:gJ};var sT=Qt((r,t)=>{let e=r-t;return e*e}),xJ=oe(Rs,sT),cF={kernelName:Rs,backendName:\"cpu\",kernelFunc:xJ};function lw(r,t,e,n){let o=wt(r,t.dtype);for(let s=0;s<o.size;s++){let i=o.indexToLoc(s),a=new Array(i.length);for(let u=0;u<a.length;u++)a[u]=i[u]*e[u]+n[u];o.set(t.get(...a),...i)}return o}var iT=class{constructor(t,e,n,o,s,i){this.separator=y.encodeString(t),this.nGramWidths=e,this.leftPad=y.encodeString(n),this.rightPad=y.encodeString(o),this.padWidth=s,this.preserveShort=i}getPadWidth(t){return Math.min(this.padWidth<0?t-1:this.padWidth,t-1)}getNumNGrams(t,e){let n=this.getPadWidth(e);return Math.max(0,t+2*n-e+1)}createNGrams(t,e,n,o,s,i){for(let a=0;a<s;++a){let u=this.getPadWidth(i),l=Math.max(0,u-a),c=Math.max(0,u-(s-(a+1))),p=i-(l+c),m=e+(l>0?0:a-u),f=0;f+=l*this.leftPad.length;for(let b=0;b<p;++b)f+=t[m+b].length;f+=c*this.rightPad.length,f+=(l+c+p-1)*this.separator.length,n[o+a]=new Uint8Array(f);let h=n[o+a],g=0,x=b=>b.forEach(w=>h[g++]=w);for(let b=0;b<l;++b)x(this.leftPad),x(this.separator);for(let b=0;b<p-1;++b)x(t[m+b]),x(this.separator);if(p>0){x(t[m+p-1]);for(let b=0;b<c;++b)x(this.separator),x(this.rightPad)}else{for(let b=0;b<c-1;++b)x(this.rightPad),x(this.separator);x(this.rightPad)}}}compute(t,e){let n=t.length,o=e.length;if(o>0){let u=e[0];if(u!==0)throw new Error(`First split value must be 0, got ${u}`);for(let l=1;l<o;++l){let c=e[l]>=u;if(c=c&&e[l]<=n,!c)throw new Error(`Invalid split value ${e[l]}, must be in [${u}, ${n}]`);u=e[l]}if(u!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${u}`)}let s=o-1,i=y.getArrayFromDType(\"int32\",o);if(n===0||o===0){let u=new Array(n);for(let l=0;l<=s;++l)i[l]=0;return[u,i]}i[0]=0;for(let u=1;u<=s;++u){let l=e[u]-e[u-1],c=0;this.nGramWidths.forEach(p=>{c+=this.getNumNGrams(l,p)}),this.preserveShort&&l>0&&c===0&&(c=1),i[u]=i[u-1]+c}let a=new Array(i[s]);for(let u=0;u<s;++u){let l=e[u],c=i[u];if(this.nGramWidths.forEach(p=>{let m=e[u+1]-e[u],f=this.getNumNGrams(m,p);this.createNGrams(t,l,a,c,f,p),c+=f}),this.preserveShort&&c===i[u]){let p=e[u+1]-e[u];if(p===0)continue;let m=p+2*this.padWidth,f=1;this.createNGrams(t,l,a,c,f,m)}}return[a,i]}};function Dc(r,t,e,n,o,s,i,a){return new iT(e,n,o,s,i,a).compute(r,t)}function yJ(r,t,e,n){if(!r.length)return;if(t.length===0){for(let s=0;s<r.length;++s)n.push(r.subarray(s,s+1));return}if(t.length===1){let s=t[0],i=r.indexOf(s);for(;i!==-1;){let a=r.subarray(0,i);(!e||a.length!==0)&&n.push(a),r=r.subarray(i+1),i=r.indexOf(s)}(!e||r.length!==0)&&n.push(r);return}let o=0;for(let s=0;s<r.length+1;s++)if(s===r.length||t.indexOf(r[s])!==-1){let i=r.subarray(o,s);(!e||i.length!==0)&&n.push(i),o=s+1}}function Rc(r,t,e){let n=r.length,o=[],s=0,i=0,a=new Array(n);for(let m=0;m<n;++m){let f=o.length;yJ(r[m],t,e,o);let d=o.length-f;a[m]=d,s+=d,i=Math.max(i,d)}let u=y.getArrayFromDType(\"int32\",s*2),l=new Array(s),c=[n,i],p=0;for(let m=0;m<n;++m)for(let f=0;f<a[m];++f)u[p*2]=m,u[p*2+1]=f,l[p]=o[p],++p;return[u,l,c]}function Fc(r,t){let e=y.getArrayFromDType(\"int32\",r.length);for(let n=0;n<r.length;++n)e[n]=y.fingerPrint64(r[n]).modulo(t).getLowBitsUnsigned();return e}var aT=Qt((r,t)=>r-t),bJ=od((r,t,e,n)=>({real:r-e,imag:t-n})),Vh=oe(Fs,aT,bJ),pF={kernelName:Fs,backendName:\"cpu\",kernelFunc:Vh};function uw(r,t){let e=new Array(r.rank);for(let o=0;o<e.length;o++)e[o]=r.shape[o]*t[o];let n=wt(e,r.dtype);for(let o=0;o<n.values.length;++o){let s=n.indexToLoc(o),i=new Array(r.rank);for(let u=0;u<i.length;u++)i[u]=s[u]%r.shape[u];let a=r.locToIndex(i);n.values[o]=r.values[a]}return n}var Gh=(r,t)=>{let e=t.value-r.value;return e===0?r.index-t.index:e};function mF(r,t,e=0,n=r.length-1){for(;n>e;){if(n-e>600){let a=n-e+1,u=t-e+1,l=Math.log(a),c=.5*Math.exp(2*l/3),p=.5*Math.sqrt(l*c*(a-c)/a)*Math.sign(u-a/2),m=Math.max(e,Math.floor(t-u*c/a+p)),f=Math.min(n,Math.floor(t+(a-u)*c/a+p));mF(r,t,m,f)}let o=r[t],s=e,i=n;for(y.swap(r,e,t),Gh(r[n],o)>0&&y.swap(r,e,n);s<i;){for(y.swap(r,s,i),s++,i--;Gh(r[s],o)<0;)s=s+1;for(;Gh(r[i],o)>0;)i=i-1}Gh(r[e],o)===0?y.swap(r,e,i):(i=i+1,y.swap(r,i,n)),i<=t&&(e=i+1),t<=i&&(n=i-1)}}function cw(r,t,e,n,o){let s=t[t.length-1],[i,a]=[r.length/s,s],u=y.getTypedArrayFromDType(e,i*n),l=y.getTypedArrayFromDType(\"int32\",i*n);for(let p=0;p<i;p++){let m=p*a,f=r.subarray(m,m+a),d=new Array(f.length);f.forEach((b,w)=>d[w]={value:b,index:w}),n<d.length&&(mF(d,n),d=d.slice(0,n)),o&&d.sort(Gh);let h=p*n,g=u.subarray(h,h+n),x=l.subarray(h,h+n);for(let b=0;b<n;b++)g[b]=d[b].value,x[b]=d[b].index}let c=t.slice();return c[c.length-1]=n,[wt(c,e,u),wt(c,\"int32\",l)]}function pw(r,t,e,n){let o=y.parseAxisParam(t,e)[0],s=[1,e[0],1];for(let d=0;d<o;d++)s[0]*=e[d];s[1]=e[o];for(let d=o+1;d<e.length;d++)s[2]*=e[d];let i={},a=new Int32Array(e[o]),u=new pe(s,n,r),l=[],c=s[0]===1&&s[2]===1;for(let d=0;d<e[o];d++){let h;if(c)h=r[d].toString();else{let g=[];for(let x=0;x<s[0];x++)for(let b=0;b<s[2];b++)g.push(u.get(x,d,b));h=g.join(\",\")}if(i[h]!==void 0)a[d]=i[h];else{let g=Object.keys(i).length;i[h]=g,a[d]=g,l.push(d)}}let p=s.slice();p[1]=Object.keys(i).length;let m=new pe(p,n);l.forEach((d,h)=>{for(let g=0;g<s[0];g++)for(let x=0;x<s[2];x++)m.set(u.get(g,d,x),g,h,x)});let f=e.slice();return f[o]=p[1],{outputValues:m.values,outputShape:f,indices:a}}Xp(\"cpu\",()=>new Nu,1);var lT=kt(ts,r=>r>=0?r:Math.exp(r)-1),fF={kernelName:ts,backendName:\"cpu\",kernelFunc:lT};function uT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{alpha:s}=n;tt([o],\"leakyRelu\");let i=y.sizeFromShape(o.shape),a=e.data.get(o.dataId).values,u=y.getTypedArrayFromDType(\"float32\",i);for(let l=0;l<a.length;l++)u[l]=a[l]<0?s*a[l]:a[l];return e.makeTensorInfo(o.shape,\"float32\",u)}var dF={kernelName:is,backendName:\"cpu\",kernelFunc:uT};var CJ=Qt((r,t)=>r<0?t*r:r);function cT(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t;tt([n,o],\"prelu\");let s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,[a,u]=CJ(n.shape,o.shape,s,i,\"float32\");return e.makeTensorInfo(u,\"float32\",a)}var hF={kernelName:bs,backendName:\"cpu\",kernelFunc:cT};var pT=kt(Cs,r=>Math.max(0,r)),gF={kernelName:Cs,backendName:\"cpu\",kernelFunc:pT};var mT=kt(vs,r=>Math.min(Math.max(0,r),6)),xF={kernelName:vs,backendName:\"cpu\",kernelFunc:mT};function Oc(r,t,e,n,o){if(e===\"linear\")return Kr({inputs:{x:t},backend:r});if(e===\"relu\")return pT({inputs:{x:t},backend:r});if(e===\"elu\")return lT({inputs:{x:t},backend:r});if(e===\"relu6\")return mT({inputs:{x:t},backend:r});if(e===\"prelu\")return cT({inputs:{x:t,alpha:n},backend:r});if(e===\"leakyrelu\")return uT({inputs:{x:t},backend:r,attrs:{alpha:o}});if(e===\"sigmoid\")return oT({inputs:{x:t},backend:r});throw new Error(`Activation ${e} has not been implemented for the CPU backend.`)}function Yt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{shape:s}=n,i=y.sizeFromShape(o.shape),a=y.inferFromImplicitShape(s,i),u=y.sizeFromShape(a);y.assert(i===u,()=>`The new shape (${a}) has ${u} elements and the old shape (${o.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),e.incRef(o.dataId);let l=e.data.get(o.dataId);if(l.complexTensorInfos!=null){let c=l.complexTensorInfos.real,p=l.complexTensorInfos.imag;c.shape=a,p.shape=a}return{dataId:o.dataId,shape:a,dtype:o.dtype}}var yF={kernelName:di,backendName:\"cpu\",kernelFunc:Yt};function fT(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;tt([o,s],\"matMul\");let u=o.shape.length,l=s.shape.length,c=i?o.shape[u-2]:o.shape[u-1],p=a?s.shape[l-1]:s.shape[l-2],m=i?o.shape[u-1]:o.shape[u-2],f=a?s.shape[l-2]:s.shape[l-1],d=o.shape.slice(0,-2),h=s.shape.slice(0,-2),g=y.sizeFromShape(d),x=y.sizeFromShape(h),w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)).concat([m,f]);y.assert(c===p,()=>`Error in matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${o.shape} and ${s.shape} and transposeA=${i} and transposeB=${a} must match.`);let C=i?[g,c,m]:[g,m,c],N=a?[x,f,p]:[x,p,f],_=Yt({inputs:{x:o},backend:e,attrs:{shape:C}}),A=Yt({inputs:{x:s},backend:e,attrs:{shape:N}}),$=i?_.shape[1]:_.shape[2],F=i?_.shape[2]:_.shape[1],P=a?A.shape[1]:A.shape[2],V=Math.max(g,x),G=e.data.get(_.dataId).values,W=e.data.get(A.dataId).values,q=y.computeStrides(_.shape),H=y.computeStrides(A.shape),[j,Y,Z]=i?[q[0],1,q[1]]:[q[0],q[1],1],[et,rt,ot]=a?[1,H[1],H[0]]:[H[1],1,H[0]],at=F*P,nt=wt([V,F,P],_.dtype),it=nt.values,dt=e.blockSize;for(let ht=0;ht<V;ht++)for(let bt=0;bt<F;bt+=dt)for(let Et=0;Et<P;Et+=dt)for(let At=0;At<$;At+=dt){let Vt=Math.min(bt+dt,F),Zt=Math.min(Et+dt,P),ce=Math.min(At+dt,$);for(let he=bt;he<Vt;he++)for(let jt=Et;jt<Zt;jt++){let ke=0;for(let fe=At;fe<ce;fe++){let Ae=Math.min(ht,g-1)*j,We=Math.min(ht,x-1)*ot,_n=G[Ae+he*Y+fe*Z],or=W[fe*et+jt*rt+We];ke+=_n*or}it[ht*at+(he*P+jt)]+=ke}}return e.disposeIntermediateTensorInfo(_),e.disposeIntermediateTensorInfo(A),e.makeTensorInfo(w,nt.dtype,nt.values)}var bF={kernelName:Ho,backendName:\"cpu\",kernelFunc:fT};function IJ(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n,m,f,d,h=[];m=fT({inputs:{a:o,b:s},attrs:{transposeA:u,transposeB:l},backend:e}),i&&(f=Ki({inputs:{a:m,b:i},backend:e}),h.push(m),m=f),c&&(d=Oc(e,m,c,a,p),h.push(m),m=d);for(let x of h)e.disposeIntermediateTensorInfo(x);return m}var wF={kernelName:Ci,backendName:\"cpu\",kernelFunc:IJ};var SJ=kt(oa,r=>Math.acos(r)),CF={kernelName:oa,backendName:\"cpu\",kernelFunc:SJ};var vJ=kt(sa,r=>Math.acosh(r)),IF={kernelName:sa,backendName:\"cpu\",kernelFunc:vJ};function NJ(r){let{inputs:t,backend:e}=r,n=t;tt(t,\"addN\");let o=n.map(a=>e.data.get(a.dataId).values),s=wt(n[0].shape,n[0].dtype),i=s.values;for(let a=0;a<n.length;a++){let u=o[a];for(let l=0;l<i.length;l++)i[l]+=u[l]}return e.makeTensorInfo(s.shape,s.dtype,s.values)}var SF={kernelName:Go,backendName:\"cpu\",kernelFunc:NJ};function TJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,\"all\");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims(\"all\",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];w=w&&N}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var vF={kernelName:ia,backendName:\"cpu\",kernelFunc:TJ};function kJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,\"any\");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims(\"any\",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];w=w||N}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var NF={kernelName:aa,backendName:\"cpu\",kernelFunc:kJ};function EJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n;tt(o,\"argMax\");let i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Ve({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],v.assertAxesAreInnerMostDims(\"argMax\",i,u.shape.length);let[c,p]=v.computeOutAndReduceShapes(u.shape,i),m=y.sizeFromShape(c),f=y.makeZerosTypedArray(m,\"int32\"),d=y.sizeFromShape(p),h=e.data.get(u.dataId).values;for(let g=0;g<f.length;++g){let x=g*d,b=h[x],w=0;for(let C=0;C<d;++C){let N=h[x+C];N>b&&(b=N,w=C)}f[g]=w}return l.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.makeTensorInfo(c,\"int32\",f)}var TF={kernelName:Wo,backendName:\"cpu\",kernelFunc:EJ};function _J(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n;tt(o,\"argMin\");let i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Ve({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],v.assertAxesAreInnerMostDims(\"argMin\",i,u.shape.length);let[c,p]=v.computeOutAndReduceShapes(u.shape,i),m=y.sizeFromShape(c),f=y.makeZerosTypedArray(m,\"int32\"),d=y.sizeFromShape(p),h=e.data.get(u.dataId).values;for(let g=0;g<f.length;++g){let x=g*d,b=h[x],w=0;for(let C=0;C<d;++C){let N=h[x+C];N<b&&(b=N,w=C)}f[g]=w}return l.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.makeTensorInfo(c,\"int32\",f)}var kF={kernelName:kl,backendName:\"cpu\",kernelFunc:_J};var AJ=kt(la,r=>Math.asin(r)),EF={kernelName:la,backendName:\"cpu\",kernelFunc:AJ};var $J=kt(ua,r=>Math.asinh(r)),_F={kernelName:ua,backendName:\"cpu\",kernelFunc:$J};var DJ=kt(ca,r=>Math.atan(r)),AF={kernelName:ca,backendName:\"cpu\",kernelFunc:DJ};var RJ=Qt((r,t)=>Math.atan2(r,t)),FJ=oe(ma,RJ),$F={kernelName:ma,backendName:\"cpu\",kernelFunc:FJ};var OJ=kt(pa,r=>Math.atanh(r)),DF={kernelName:pa,backendName:\"cpu\",kernelFunc:OJ};function ud(r,t,e,n,o,s){let i=o.strideHeight,a=o.strideWidth,u=o.dilationHeight,l=o.dilationWidth,c=o.effectiveFilterHeight,p=o.effectiveFilterWidth,m=o.padInfo.top,f=o.padInfo.left,d=s===\"max\"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,h=wt(o.outShape,e),g=h.values,x=o.outShape[1]*o.outShape[2]*o.outShape[3],b=o.outShape[2]*o.outShape[3],w=o.outShape[3];for(let C=0;C<o.batchSize;++C){let N=C*x,_=C*n[0];for(let A=0;A<o.inChannels;++A)for(let $=0;$<o.outHeight;++$){let F=$*i-m,P=Math.max(0,F),V=Math.min(o.inHeight,c+F),G=N+$*b;for(let W=0;W<o.outWidth;++W){let q=W*a-f,H=Math.max(0,q),j=Math.min(o.inWidth,p+q),Y=d,Z=0,et=0;for(let ot=P;ot<V;ot+=u){let at=_+ot*n[1];for(let nt=H;nt<j;nt+=l){let it=at+nt*n[2],dt=r[it+A];s===\"max\"&&dt>Y?Y=dt:s===\"avg\"&&(Z+=dt,et++)}if(isNaN(Y))break}let rt=G+W*w+A;g[rt]=s===\"avg\"?Z/et:Y}}}return h}function fw(r,t,e,n,o=!1,s=!1){let i=wt(n.outShape,\"int32\"),a=n.strideHeight,u=n.strideWidth,l=n.dilationHeight,c=n.dilationWidth,p=n.effectiveFilterHeight,m=n.effectiveFilterWidth,f=n.padInfo.top,d=n.padInfo.left,h=wt(t,e,r);for(let g=0;g<n.batchSize;++g)for(let x=0;x<n.inChannels;++x)for(let b=0;b<n.outHeight;++b){let w=b*a-f,C=w;for(;C<0;)C+=l;let N=Math.min(n.inHeight,p+w);for(let _=0;_<n.outWidth;++_){let A=_*u-d,$=A;for(;$<0;)$+=c;let F=Math.min(n.inWidth,m+A),P=Number.NEGATIVE_INFINITY,V=-1;for(let G=C;G<N;G+=l){let W=G-w;for(let q=$;q<F;q+=c){let H=q-A,j=h.get(g,G,q,x);j>P&&(P=j,o?V=s?((g*n.inHeight+G)*n.inWidth+q)*n.inChannels+x:(G*n.inWidth+q)*n.inChannels+x:V=W*m+H)}}i.set(V,g,b,_,x)}}return i}function dw(r,t,e,n,o,s){let i=o.strideDepth,a=o.strideHeight,u=o.strideWidth,l=o.dilationDepth,c=o.dilationHeight,p=o.dilationWidth,m=o.effectiveFilterDepth,f=o.effectiveFilterHeight,d=o.effectiveFilterWidth,h=o.padInfo.front,g=o.padInfo.top,x=o.padInfo.left,b=s===\"max\"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,w=wt(o.outShape,e),C=w.values,N=o.outShape[1]*o.outShape[2]*o.outShape[3]*o.outShape[4],_=o.outShape[2]*o.outShape[3]*o.outShape[4],A=o.outShape[3]*o.outShape[4],$=o.outShape[4];for(let F=0;F<o.batchSize;++F){let P=F*N,V=F*n[0];for(let G=0;G<o.inChannels;++G)for(let W=0;W<o.outDepth;++W){let q=W*i-h,H=q;for(;H<0;)H+=l;let j=Math.min(o.inDepth,m+q),Y=P+W*_;for(let Z=0;Z<o.outHeight;++Z){let et=Z*a-g,rt=et;for(;rt<0;)rt+=c;let ot=Math.min(o.inHeight,f+et),at=Y+Z*A;for(let nt=0;nt<o.outWidth;++nt){let it=nt*u-x,dt=it;for(;dt<0;)dt+=p;let ht=Math.min(o.inWidth,d+it),bt=at+nt*$,Et=b,At=0,Vt=0;for(let ce=H;ce<j;ce+=l){let he=V+ce*n[1];for(let jt=rt;jt<ot;jt+=c){let ke=he+jt*n[2];for(let fe=dt;fe<ht;fe+=p){let Ae=ke+fe*n[3],We=r[Ae+G];if(s===\"max\"&&We>Et?Et=We:s===\"avg\"&&(At+=We,Vt++),isNaN(Et))break}if(isNaN(Et))break}if(isNaN(Et))break}let Zt=bt+G;C[Zt]=s===\"avg\"?At/Vt:Et}}}}return w}function RF(r,t){let e=wt(t.outShape,\"int32\"),n=t.strideDepth,o=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,a=t.dilationHeight,u=t.dilationWidth,l=t.effectiveFilterDepth,c=t.effectiveFilterHeight,p=t.effectiveFilterWidth,m=t.padInfo.front,f=t.padInfo.top,d=t.padInfo.left;for(let h=0;h<t.batchSize;++h)for(let g=0;g<t.inChannels;++g)for(let x=0;x<t.outDepth;++x){let b=x*n-m,w=b;for(;w<0;)w+=i;let C=Math.min(t.inDepth,l+b);for(let N=0;N<t.outHeight;++N){let _=N*o-f,A=_;for(;A<0;)A+=a;let $=Math.min(t.inHeight,c+_);for(let F=0;F<t.outWidth;++F){let P=F*s-d,V=P;for(;V<0;)V+=u;let G=Math.min(t.inWidth,p+P),W=Number.NEGATIVE_INFINITY,q=-1;for(let H=w;H<C;H+=i){let j=H-b;for(let Y=A;Y<$;Y+=a){let Z=Y-_;for(let et=V;et<G;et+=u){let rt=et-P,ot=r.get(h,H,Y,et,g);ot>=W&&(W=ot,q=j*c*p+Z*c+rt)}}}e.set(q,h,x,N,F,g)}}}return e}function PJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;tt(o,\"avgPool\");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u),p;if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))p=Kr({inputs:{x:o},backend:e});else{let m=e.data.get(o.dataId).values,f=y.computeStrides(o.shape),d=ud(m,o.shape,o.dtype,f,c,\"avg\");p=e.makeTensorInfo(c.outShape,o.dtype,d.values)}return p}var FF={kernelName:Uo,backendName:\"cpu\",kernelFunc:PJ};function LJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n;tt(o,\"avgPool3d\");let c=v.computePool3DInfo(o.shape,s,i,1,a,u,l),p=e.data.get(o.dataId).values,m=dw(p,o.shape,o.dtype,y.computeStrides(o.shape),c,\"avg\");return e.makeTensorInfo(m.shape,\"float32\",m.values)}var OF={kernelName:El,backendName:\"cpu\",kernelFunc:LJ};function MJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=n;tt([o,s],\"avgPool3DGrad\");let c=v.computePool3DInfo(s.shape,i,a,1,u,l),p=c.strideDepth,m=c.strideHeight,f=c.strideWidth,d=c.filterDepth,h=c.filterHeight,g=c.filterWidth,x=c.dilationDepth,b=c.dilationHeight,w=c.dilationWidth,C=c.effectiveFilterDepth,N=c.effectiveFilterHeight,_=c.effectiveFilterWidth,A=C-1-c.padInfo.front,$=_-1-c.padInfo.left,F=N-1-c.padInfo.top,P=wt(s.shape,\"float32\"),V=1/(d*h*g),G=e.bufferSync(o);for(let W=0;W<c.batchSize;++W)for(let q=0;q<c.inChannels;++q)for(let H=0;H<c.inDepth;++H)for(let j=0;j<c.inHeight;++j)for(let Y=0;Y<c.inWidth;++Y){let Z=H-A,et=j-F,rt=Y-$,ot=0;for(let at=0;at<C;at+=x){let nt=(Z+at)/p;if(!(nt<0||nt>=c.outDepth||Math.floor(nt)!==nt))for(let it=0;it<N;it+=b){let dt=(et+it)/m;if(!(dt<0||dt>=c.outHeight||Math.floor(dt)!==dt))for(let ht=0;ht<_;ht+=w){let bt=(rt+ht)/f;if(bt<0||bt>=c.outWidth||Math.floor(bt)!==bt)continue;ot+=G.get(W,nt,dt,bt,q)}}}P.set(ot*V,W,H,j,Y,q)}return e.makeTensorInfo(P.shape,P.dtype,P.values)}var PF={kernelName:lp,backendName:\"cpu\",kernelFunc:MJ};function zJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s;tt([o,s],\"avgPoolGrad\");let{filterSize:a,strides:u,pad:l}=n,c=v.computePool2DInfo(i.shape,a,u,1,l),p=c.strideHeight,m=c.strideWidth,f=c.filterHeight,d=c.filterWidth,h=c.dilationHeight,g=c.dilationWidth,x=c.effectiveFilterHeight,b=c.effectiveFilterWidth,w=b-1-c.padInfo.left,C=x-1-c.padInfo.top,N=wt(i.shape,\"float32\"),_=1/(f*d),A=e.data.get(o.dataId).values,$=wt(o.shape,\"float32\",A);for(let F=0;F<c.batchSize;++F)for(let P=0;P<c.inChannels;++P)for(let V=0;V<c.inHeight;++V)for(let G=0;G<c.inWidth;++G){let W=V-C,q=G-w,H=0;for(let j=0;j<x;j+=h){let Y=(W+j)/p;if(!(Y<0||Y>=c.outHeight||Math.floor(Y)!==Y))for(let Z=0;Z<b;Z+=g){let et=(q+Z)/m;if(et<0||et>=c.outWidth||Math.floor(et)!==et)continue;H+=$.get(F,Y,et,P)}}N.set(H*_,F,V,G,P)}return e.makeTensorInfo(N.shape,N.dtype,N.values)}var LF={kernelName:ap,backendName:\"cpu\",kernelFunc:zJ};function BJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,scale:s,offset:i,mean:a,variance:u}=t;y.assert(a.shape.length===u.shape.length,()=>\"Batch normalization gradient requires mean and variance to have equal ranks.\"),y.assert(i==null||a.shape.length===i.shape.length,()=>\"Batch normalization gradient requires mean and offset to have equal ranks.\"),y.assert(s==null||a.shape.length===s.shape.length,()=>\"Batch normalization gradient requires mean and scale to have equal ranks.\"),tt([o,a,u,s,i],\"batchNorm\");let{varianceEpsilon:l}=n;l==null&&(l=.001);let c=e.data.get(o.dataId).values,p=e.data.get(a.dataId).values,m=e.data.get(u.dataId).values,f=s?e.data.get(s.dataId).values:new Float32Array([1]),d=i?e.data.get(i.dataId).values:new Float32Array([0]),h=new Float32Array(c.length),g=d.length,x=f.length,b=m.length,w=p.length,C=0,N=0,_=0,A=0;for(let $=0;$<c.length;++$)h[$]=d[C++]+(c[$]-p[N++])*f[_++]/Math.sqrt(m[A++]+l),C>=g&&(C=0),N>=w&&(N=0),_>=x&&(_=0),A>=b&&(A=0);return e.makeTensorInfo(o.shape,o.dtype,h)}var MF={kernelName:os,backendName:\"cpu\",kernelFunc:BJ};function VJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n;tt([o],\"batchToSpaceND\");let a=s.reduce((x,b)=>x*b),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=Yt({inputs:{x:o},backend:e,attrs:{shape:u}}),d=Ve({inputs:{x:f},backend:e,attrs:{perm:l}}),h=Yt({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Fo({inputs:{x:h},backend:e,attrs:{begin:p,size:m}});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),g}var zF={kernelName:ai,backendName:\"cpu\",kernelFunc:VJ};function GJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i}=n,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,l=sd(a,u,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,l)}var BF={kernelName:up,backendName:\"cpu\",kernelFunc:GJ};function WJ(r){let{inputs:t,backend:e}=r,{s0:n,s1:o}=t,s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,a=v.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return e.makeTensorInfo([a.length],\"int32\",Int32Array.from(a))}var VF={kernelName:cp,backendName:\"cpu\",kernelFunc:WJ};var UJ=kt(uo,(r,t)=>{let e=t;return r>e.clipValueMax?e.clipValueMax:r<e.clipValueMin?e.clipValueMin:r}),GF={kernelName:uo,backendName:\"cpu\",kernelFunc:UJ};var HJ=r=>{let{x:t}=r.inputs,e=r.backend,n=new Float32Array(y.sizeFromShape(t.shape)),o=e.data.get(t.dataId),s=o.complexTensorInfos.real,i=o.complexTensorInfos.imag,a=e.data.get(s.dataId).values,u=e.data.get(i.dataId).values;for(let l=0;l<a.length;l++){let c=a[l],p=u[l];n[l]=Math.hypot(c,p)}return e.makeOutput(n,t.shape,\"float32\")},WF={kernelName:_l,backendName:\"cpu\",kernelFunc:HJ};function ji(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.data.get(n.dataId).complexTensorInfos.imag,s=e.data.get(o.dataId).values;return e.makeTensorInfo(o.shape,o.dtype,s)}var UF={kernelName:Sp,backendName:\"cpu\",kernelFunc:ji};function Tu(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n,s=y.parseAxisParam(o,t[0].shape)[0],i=t.map(h=>h.shape);v.assertParamsConsistent(i,s);let a=v.computeOutShape(t.map(h=>h.shape),s);if(y.sizeFromShape(a)===0)return e.makeTensorInfo(a,t[0].dtype,[]);let u=t.filter(h=>y.sizeFromShape(h.shape)>0);if(u.length===1)return Kr({inputs:{x:u[0]},backend:e});if(u[0].dtype===\"complex64\"){let h=u.map(C=>Ao({inputs:{input:C},backend:e})),g=u.map(C=>ji({inputs:{input:C},backend:e})),x=Tu({inputs:h,backend:e,attrs:{axis:s}}),b=Tu({inputs:g,backend:e,attrs:{axis:s}}),w=wr({inputs:{real:x,imag:b},backend:e});return h.forEach(C=>e.disposeIntermediateTensorInfo(C)),g.forEach(C=>e.disposeIntermediateTensorInfo(C)),e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(b),w}let l=u.map(h=>{let g=y.sizeFromShape(h.shape.slice(s));return Yt({inputs:{x:h},backend:e,attrs:{shape:[-1,g]}})}),c=l.map(h=>({vals:e.data.get(h.dataId).values,shape:h.shape}));a=v.computeOutShape(l.map(h=>h.shape),1);let p=l[0].shape[0]===1,m=Ec(c,a,t[0].dtype,p),f=v.computeOutShape(u.map(h=>h.shape),s),d=e.makeTensorInfo(f,t[0].dtype,m);return l.forEach(h=>e.disposeIntermediateTensorInfo(h)),d}var HF={kernelName:li,backendName:\"cpu\",kernelFunc:Tu};function dT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dataFormat:u,dilations:l,dimRoundingMode:c}=n;tt([o,s],\"conv2d\");let p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,s.shape,i,l,a,c,!1,p),f=m.filterHeight,d=m.filterWidth,h=m.dilationHeight,g=m.dilationWidth,x=m.padInfo.left,b=m.padInfo.top,w=m.dataFormat===\"channelsLast\",C=new pe(m.outShape,o.dtype),N=y.computeStrides(o.shape),_=y.computeStrides(s.shape),A=N[0],$=w?N[1]:N[2],F=w?N[2]:1,P=w?1:N[1],V=C.strides[0],G=w?C.strides[1]:C.strides[2],W=w?C.strides[2]:1,q=w?1:C.strides[1],H=e.data.get(o.dataId).values,j=e.data.get(s.dataId).values,Y=C.values;for(let Z=0;Z<m.batchSize;++Z){let et=Z*A,rt=Z*V;for(let ot=0;ot<m.outHeight;++ot){let at=rt+ot*G,nt=ot*m.strideHeight-b;for(let it=0;it<f;++it){let dt=nt+it*h;if(dt<0||dt>=m.inHeight)continue;let ht=it*_[0],bt=et+dt*$;for(let Et=0;Et<m.outWidth;++Et){let At=at+Et*W,Vt=Et*m.strideWidth-x;for(let Zt=0;Zt<d;++Zt){let ce=Vt+Zt*g;if(ce<0||ce>=m.inWidth)continue;let he=ht+Zt*_[1],jt=bt+ce*F,ke=he;for(let fe=0;fe<m.inChannels;++fe){let Ae=H[jt+fe*P];for(let We=0;We<m.outChannels;++We)Y[At+We*q]+=Ae*j[ke+We];ke+=m.outChannels}}}}}}return e.makeTensorInfo(C.shape,C.dtype,Y)}var qF={kernelName:Ko,backendName:\"cpu\",kernelFunc:dT};function qJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,filterShape:c}=n;tt([o,s],\"conv2dBackpropFilter\");let p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,c,i,1,a,l,!1,p),{strideHeight:f,strideWidth:d,filterHeight:h,filterWidth:g}=m,x=m.dataFormat===\"channelsLast\",b=new pe(m.filterShape,\"float32\"),w=m.padInfo.left,C=m.padInfo.top,N=e.data.get(o.dataId).values,_=e.data.get(s.dataId).values,A=new pe(o.shape,o.dtype,N),$=new pe(s.shape,s.dtype,_);for(let F=0;F<h;++F){let P=Math.max(0,Math.ceil((C-F)/f)),V=Math.min(m.outHeight,(m.inHeight+C-F)/f);for(let G=0;G<g;++G){let W=Math.max(0,Math.ceil((w-G)/d)),q=Math.min(m.outWidth,(m.inWidth+w-G)/d);for(let H=0;H<m.inChannels;++H)for(let j=0;j<m.outChannels;++j){let Y=0;for(let Z=0;Z<m.batchSize;++Z)for(let et=P;et<V;++et){let rt=F+et*f-C;for(let ot=W;ot<q;++ot){let at=G+ot*d-w;x?Y+=A.get(Z,rt,at,H)*$.get(Z,et,ot,j):Y+=A.get(Z,H,rt,at)*$.get(Z,j,et,ot)}}b.set(Y,F,G,H,j)}}}return e.makeTensorInfo(b.shape,b.dtype,b.values)}var KF={kernelName:mp,backendName:\"cpu\",kernelFunc:qJ};function KJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{inputShape:i,strides:a,pad:u,dataFormat:l,dimRoundingMode:c}=n;tt([o,s],\"conv2dBackpropInput\");let p=y.computeStrides(s.shape),m=y.computeStrides(o.shape),f=v.convertConv2DDataFormat(l),d=v.computeConv2DInfo(i,s.shape,a,1,u,c,!1,f),h=new pe(d.inShape,\"float32\"),g=h.values,x=e.data.get(o.dataId).values,b=e.data.get(s.dataId).values,[w,C,N]=p,{batchSize:_,filterHeight:A,filterWidth:$,inChannels:F,inHeight:P,inWidth:V,outChannels:G,outHeight:W,outWidth:q,strideHeight:H,strideWidth:j}=d;f=d.dataFormat;let Y=A-1-d.padInfo.top,Z=$-1-d.padInfo.left,et=f===\"channelsLast\",rt=h.strides[0],ot=et?h.strides[1]:h.strides[2],at=et?h.strides[2]:1,nt=et?1:h.strides[1],it=m[0],dt=et?m[1]:m[2],ht=et?m[2]:1,bt=et?1:m[1];for(let Et=0;Et<_;++Et)for(let At=0;At<F;++At)for(let Vt=0;Vt<P;++Vt){let Zt=Vt-Y,ce=Math.max(0,Math.ceil(Zt/H)),he=Math.min(W,(A+Zt)/H);for(let jt=0;jt<V;++jt){let ke=jt-Z,fe=Math.max(0,Math.ceil(ke/j)),Ae=Math.min(q,($+ke)/j),We=0;for(let or=ce;or<he;++or){let Hn=or*H-Zt;for(let Lr=fe;Lr<Ae;++Lr){let qe=Lr*j-ke,Mr=it*Et+dt*or+ht*Lr,zr=w*(A-1-Hn)+C*($-1-qe)+N*At;for(let qn=0;qn<G;++qn){let Kn=x[Mr+bt*qn],Xr=b[zr+qn];We+=Kn*Xr}}}let _n=rt*Et+ot*Vt+at*jt+nt*At;g[_n]=We}}return e.makeTensorInfo(h.shape,h.dtype,h.values)}var jF={kernelName:jo,backendName:\"cpu\",kernelFunc:KJ};function jJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n;tt([o,s],\"conv3d\");let l=v.computeConv3DInfo(o.shape,s.shape,i,u,a),{filterDepth:c,filterHeight:p,filterWidth:m,dilationDepth:f,dilationHeight:d,dilationWidth:h,padInfo:g}=l,x=g.front,b=g.left,w=g.top,C=new pe(l.outShape,o.dtype),N=e.data.get(o.dataId).values,_=e.data.get(s.dataId).values,A=C.values,$=y.computeStrides(o.shape),F=y.computeStrides(s.shape);for(let P=0;P<l.batchSize;++P){let V=P*$[0],G=P*C.strides[0];for(let W=0;W<l.outDepth;++W){let q=G+W*C.strides[1],H=W*l.strideDepth-x;for(let j=0;j<c;++j){let Y=H+j*f;if(Y<0||Y>=l.inDepth)continue;let Z=j*F[0],et=V+Y*$[1];for(let rt=0;rt<l.outHeight;++rt){let ot=q+rt*C.strides[2],at=rt*l.strideHeight-w;for(let nt=0;nt<p;++nt){let it=at+nt*d;if(it<0||it>=l.inHeight)continue;let dt=Z+nt*F[1],ht=et+it*$[2];for(let bt=0;bt<l.outWidth;++bt){let Et=ot+bt*l.outChannels,At=bt*l.strideWidth-b;for(let Vt=0;Vt<m;++Vt){let Zt=At+Vt*h;if(Zt<0||Zt>=l.inWidth)continue;let ce=dt+Vt*F[2],he=ht+Zt*l.inChannels,jt=ce;for(let ke=0;ke<l.inChannels;++ke){let fe=N[he+ke];for(let Ae=0;Ae<l.outChannels;++Ae)A[Et+Ae]+=fe*_[jt+Ae];jt+=l.outChannels}}}}}}}}return e.makeTensorInfo(C.shape,C.dtype,C.values)}var XF={kernelName:Al,backendName:\"cpu\",kernelFunc:jJ};function XJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,filterShape:u}=n;tt([o,s],\"conv3dBackpropFilterV2\");let l=y.computeStrides(o.shape),c=y.computeStrides(s.shape),p=v.computeConv3DInfo(o.shape,u,i,1,a),m=p.strideDepth,f=p.strideHeight,d=p.strideWidth,h=p.filterDepth,g=p.filterHeight,x=p.filterWidth,b=new pe(p.filterShape,\"float32\"),w=b.values,[C,N,_,A]=b.strides,$=e.data.get(s.dataId).values,[F,P,V,G]=c,W=e.data.get(o.dataId).values,[q,H,j,Y]=l,Z=p.padInfo.front,et=p.padInfo.left,rt=p.padInfo.top;for(let ot=0;ot<h;++ot){let at=Math.max(0,Math.ceil((Z-ot)/m)),nt=Math.min(p.outDepth,(p.inDepth+Z-ot)/m),it=ot*C;for(let dt=0;dt<g;++dt){let ht=Math.max(0,Math.ceil((rt-dt)/f)),bt=Math.min(p.outHeight,(p.inHeight+rt-dt)/f),Et=dt*N+it;for(let At=0;At<x;++At){let Vt=Math.max(0,Math.ceil((et-At)/d)),Zt=Math.min(p.outWidth,(p.inWidth+et-At)/d),ce=At*_+Et;for(let he=0;he<p.inChannels;++he){let jt=he*A+ce;for(let ke=0;ke<p.outChannels;++ke){let fe=0;for(let Ae=0;Ae<p.batchSize;++Ae){let We=Ae*q,_n=Ae*F;for(let or=at;or<nt;++or){let Lr=(ot+or*m-Z)*H+We,qe=or*P+_n;for(let Mr=ht;Mr<bt;++Mr){let qn=(dt+Mr*f-rt)*j+Lr,Kn=Mr*V+qe;for(let Xr=Vt;Xr<Zt;++Xr){let Mo=(At+Xr*d-et)*Y+qn,Il=Xr*G+Kn;fe+=W[Mo+he]*$[Il+ke]}}}}w[jt+ke]=fe}}}}}return e.makeTensorInfo(b.shape,b.dtype,b.values)}var YF={kernelName:fp,backendName:\"cpu\",kernelFunc:XJ};function YJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{pad:i,strides:a,inputShape:u}=n;tt([o],\"conv3dBackpropInputV2\");let l=y.computeStrides(o.shape),c=y.computeStrides(s.shape),p=v.computeConv3DInfo(u,s.shape,a,1,i),m=new pe(p.inShape,\"float32\"),f=m.values,[d,h,g,x]=m.strides,b=e.data.get(o.dataId).values,[w,C,N,_]=l,A=e.data.get(s.dataId).values,[$,F,P,V]=c,{batchSize:G,filterDepth:W,filterHeight:q,filterWidth:H,inChannels:j,inDepth:Y,inHeight:Z,inWidth:et,outChannels:rt,outDepth:ot,outHeight:at,outWidth:nt,strideDepth:it,strideHeight:dt,strideWidth:ht}=p,bt=W-1-p.padInfo.front,Et=q-1-p.padInfo.top,At=H-1-p.padInfo.left;for(let Vt=0;Vt<G;++Vt)for(let Zt=0;Zt<j;++Zt)for(let ce=0;ce<Y;++ce){let he=ce-bt,jt=Math.max(0,Math.ceil(he/it)),ke=Math.min(ot,(W+he)/it);for(let fe=0;fe<Z;++fe){let Ae=fe-Et,We=Math.max(0,Math.ceil(Ae/dt)),_n=Math.min(at,(q+Ae)/dt);for(let or=0;or<et;++or){let Hn=or-At,Lr=Math.max(0,Math.ceil(Hn/ht)),qe=Math.min(nt,(H+Hn)/ht),Mr=0;for(let zr=jt;zr<ke;++zr){let qn=zr*it-he;for(let Kn=We;Kn<_n;++Kn){let Xr=Kn*dt-Ae;for(let Zi=Lr;Zi<qe;++Zi){let Mo=Zi*ht-Hn,Il=w*Vt+C*zr+N*Kn+_*Zi,Tr=$*(W-1-qn)+F*(q-1-Xr)+P*(H-1-Mo)+V*Zt;for(let Ji=0;Ji<rt;++Ji){let Rd=b[Il+Ji],Fd=A[Tr+Ji];Mr+=Rd*Fd}}}}f[d*Vt+h*ce+g*fe+x*or+Zt]=Mr}}}return e.makeTensorInfo(m.shape,m.dtype,m.values)}var ZF={kernelName:dp,backendName:\"cpu\",kernelFunc:YJ};var ZJ=kt(Xo,r=>Math.cos(r)),JF={kernelName:Xo,backendName:\"cpu\",kernelFunc:ZJ};var JJ=kt(Yo,r=>Math.cosh(r)),QF={kernelName:Yo,backendName:\"cpu\",kernelFunc:JJ};function QJ(r){let{inputs:t,backend:e,attrs:n}=r,{image:o,boxes:s,boxInd:i}=t,{cropSize:a,method:u,extrapolationValue:l}=n,[c,p,m,f]=o.shape,d=s.shape[0],[h,g]=a,x=wt([d,h,g,f],\"float32\"),b=e.data.get(s.dataId).values,w=e.data.get(i.dataId).values,C=e.data.get(o.dataId).values,N=y.computeStrides(o.shape),_=y.computeStrides(x.shape);for(let A=0;A<d;A++){let $=A*4,F=b[$],P=b[$+1],V=b[$+2],G=b[$+3],W=w[A];if(W>=c)continue;let q=h>1?(V-F)*(p-1)/(h-1):0,H=g>1?(G-P)*(m-1)/(g-1):0;for(let j=0;j<h;j++){let Y=h>1?F*(p-1)+j*q:.5*(F+V)*(p-1);if(Y<0||Y>p-1){for(let Z=0;Z<g;Z++)for(let et=0;et<f;et++){let rt=et+Z*_[2]+j*_[1]+A*_[0];x.values[rt]=l}continue}if(u===\"bilinear\"){let Z=Math.floor(Y),et=Math.ceil(Y),rt=Y-Z;for(let ot=0;ot<g;ot++){let at=g>1?P*(m-1)+ot*H:.5*(P+G)*(m-1);if(at<0||at>m-1){for(let ht=0;ht<f;ht++){let bt=ht+ot*_[2]+j*_[1]+A*_[0];x.values[bt]=l}continue}let nt=Math.floor(at),it=Math.ceil(at),dt=at-nt;for(let ht=0;ht<f;ht++){let bt=ht+nt*N[2]+Z*N[1]+W*N[0],Et=C[bt];bt=ht+it*N[2]+Z*N[1]+W*N[0];let At=C[bt];bt=ht+nt*N[2]+et*N[1]+W*N[0];let Vt=C[bt];bt=ht+it*N[2]+et*N[1]+W*N[0];let Zt=C[bt],ce=Et+(At-Et)*dt,he=Vt+(Zt-Vt)*dt;bt=ht+ot*_[2]+j*_[1]+A*_[0],x.values[bt]=ce+(he-ce)*rt}}}else for(let Z=0;Z<g;++Z){let et=g>1?P*(m-1)+Z*H:.5*(P+G)*(m-1);if(et<0||et>m-1){for(let at=0;at<f;at++){let nt=at+Z*_[2]+j*_[1]+A*_[0];x.values[nt]=l}continue}let rt=Math.round(et),ot=Math.round(Y);for(let at=0;at<f;at++){let nt=at+rt*N[2]+ot*N[1]+W*N[0],it=at+Z*_[2]+j*_[1]+A*_[0];x.values[it]=C[nt]}}}}return e.makeTensorInfo(x.shape,x.dtype,x.values)}var tO={kernelName:da,backendName:\"cpu\",kernelFunc:QJ};function tQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;tt(o,\"cumprod\");let u=v.getAxesPermutation([s],o.shape.length),l=o;u!=null&&(l=Ve({inputs:{x:o},backend:e,attrs:{perm:u}}));let c=v.getInnerMostAxes(1,o.shape.length)[0];if(c!==l.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);let p=sr(l.dtype,\"int32\"),m=y.makeOnesTypedArray(y.sizeFromShape(l.shape),p),f=e.data.get(l.dataId).values,d=l.shape[l.shape.length-1],h=a?(x,b)=>x+d-b-1:(x,b)=>x+b;for(let x=0;x<f.length;x+=d)for(let b=0;b<d;b++){let w=h(x,b);if(b===0)m[w]=i?1:f[w];else{let C=h(x,b-1);m[w]=i?f[C]*m[C]:f[w]*m[C]}}let g=e.makeTensorInfo(l.shape,p,m);if(u!=null){let x=v.getUndoAxesPermutation(u),b=Ve({inputs:{x:g},backend:e,attrs:{perm:x}});return e.disposeIntermediateTensorInfo(g),e.disposeIntermediateTensorInfo(l),b}return g}var eO={kernelName:fa,backendName:\"cpu\",kernelFunc:tQ};function eQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;tt(o,\"cumsum\");let u=v.getAxesPermutation([s],o.shape.length),l=o;u!=null&&(l=Ve({inputs:{x:o},backend:e,attrs:{perm:u}}));let c=v.getInnerMostAxes(1,o.shape.length)[0];if(c!==l.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);let p=sr(l.dtype,\"int32\"),m=y.makeZerosTypedArray(y.sizeFromShape(l.shape),p),f=e.data.get(l.dataId).values,d=l.shape[l.shape.length-1],h=a?(x,b)=>x+d-b-1:(x,b)=>x+b;for(let x=0;x<f.length;x+=d)for(let b=0;b<d;b++){let w=h(x,b);if(b===0)m[w]=i?0:f[w];else{let C=h(x,b-1);m[w]=i?f[C]+m[C]:f[w]+m[C]}}let g=e.makeTensorInfo(l.shape,p,m);if(u!=null){let x=v.getUndoAxesPermutation(u),b=Ve({inputs:{x:g},backend:e,attrs:{perm:x}});return e.disposeIntermediateTensorInfo(g),e.disposeIntermediateTensorInfo(l),b}return g}var rO={kernelName:Zo,backendName:\"cpu\",kernelFunc:eQ};function rQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i,binaryOutput:a}=n;if(o.shape.length===1){let u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=sd(u,l,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,c)}else if(o.shape.length===2){let u=e.bufferSync(o),l=e.bufferSync(s),c=Jb(u,l,i,a);return e.makeTensorInfo(c.shape,s.dtype,c.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${o.shape.length}.`)}var nO={kernelName:hp,backendName:\"cpu\",kernelFunc:rQ};function nQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockSize:s,dataFormat:i}=n;y.assert(i===\"NHWC\",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let a=o.shape[0],u=o.shape[1],l=o.shape[2],c=o.shape[3],p=u*s,m=l*s,f=c/(s*s),d=e.data.get(o.dataId).values,h=new Float32Array(a*p*m*f),g=0;for(let x=0;x<a;++x)for(let b=0;b<p;++b){let w=Math.floor(b/s),C=b%s;for(let N=0;N<m;++N){let _=Math.floor(N/s),A=N%s,$=(C*s+A)*f;for(let F=0;F<f;++F){let V=F+$+c*(_+l*(w+u*x));h[g++]=d[V]}}}return e.makeTensorInfo([a,p,m,f],o.dtype,h)}var oO={kernelName:ha,backendName:\"cpu\",kernelFunc:nQ};function hT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u,dimRoundingMode:l}=n;tt([o,s],\"depthwiseConv2DNative\");let c=y.computeStrides(o.shape),p=y.computeStrides(s.shape),m=u;m==null&&(m=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(i,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${m}'`);let f=v.computeConv2DInfo(o.shape,s.shape,i,m,a,l,!0),{filterHeight:d,filterWidth:h,dilationHeight:g,dilationWidth:x,padInfo:b}=f,w=b.left,C=b.top,N=f.outChannels/f.inChannels,_=new pe(f.outShape,o.dtype),A=e.data.get(o.dataId).values,$=e.data.get(s.dataId).values,F=_.values;for(let P=0;P<f.batchSize;++P){let V=P*c[0],G=P*_.strides[0];for(let W=0;W<f.outHeight;++W){let q=G+W*_.strides[1],H=W*f.strideHeight-C;for(let j=0;j<d;++j){let Y=H+j*g;if(Y<0||Y>=f.inHeight)continue;let Z=j*p[0],et=V+Y*c[1];for(let rt=0;rt<f.outWidth;++rt){let ot=q+rt*_.strides[2],at=rt*f.strideWidth-w;for(let nt=0;nt<h;++nt){let it=at+nt*x;if(it<0||it>=f.inWidth)continue;let dt=Z+nt*p[1],ht=et+it*f.inChannels,bt=ot,Et=dt;for(let At=0;At<f.inChannels;++At){let Vt=A[ht+At];for(let Zt=0;Zt<N;++Zt)F[bt+Zt]+=Vt*$[Et+Zt];bt+=N,Et+=N}}}}}}return e.makeTensorInfo(_.shape,_.dtype,_.values)}var sO={kernelName:Jo,backendName:\"cpu\",kernelFunc:hT};function oQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,filterShape:c}=n;tt([o,s],\"depthwiseConv2dNativeBackpropFilter\");let p=v.computeConv2DInfo(o.shape,c,i,a,u,l,!0),{strideHeight:m,strideWidth:f,filterHeight:d,filterWidth:h}=p,g=new pe(p.filterShape,\"float32\"),x=p.padInfo.left,b=p.padInfo.top,w=p.outChannels/p.inChannels,C=e.data.get(o.dataId).values,N=new pe(o.shape,o.dtype,C),_=e.data.get(s.dataId).values,A=new pe(s.shape,s.dtype,_);for(let $=0;$<d;++$){let F=Math.max(0,Math.ceil((b-$)/m)),P=Math.min(p.outHeight,(p.inHeight+b-$)/m);for(let V=0;V<h;++V){let G=Math.max(0,Math.ceil((x-V)/f)),W=Math.min(p.outWidth,(p.inWidth+x-V)/f);for(let q=0;q<p.outChannels;++q){let H=Math.trunc(q/w),j=q%w,Y=0;for(let Z=0;Z<p.batchSize;++Z)for(let et=F;et<P;++et){let rt=$+et*m-b;for(let ot=G;ot<W;++ot){let at=V+ot*f-x;Y+=N.get(Z,rt,at,H)*A.get(Z,et,ot,q)}}g.set(Y,$,V,H,j)}}}return e.makeTensorInfo(g.shape,g.dtype,g.values)}var iO={kernelName:gp,backendName:\"cpu\",kernelFunc:oQ};function sQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,inputShape:c}=n;tt([o,s],\"depthwiseConv2DNativeBackpropInput\");let p=y.computeStrides(o.shape),m=y.computeStrides(s.shape),f=v.computeConv2DInfo(c,s.shape,i,a,u,l,!0),d=new pe(f.inShape,\"float32\"),h=d.values,[g,x,b]=d.strides,w=e.data.get(o.dataId).values,[C,N,_]=p,A=e.data.get(s.dataId).values,[$,F,P]=m,{batchSize:V,filterHeight:G,filterWidth:W,inChannels:q,inHeight:H,inWidth:j,outChannels:Y,outHeight:Z,outWidth:et,strideHeight:rt,strideWidth:ot}=f,at=G-1-f.padInfo.top,nt=W-1-f.padInfo.left,it=Y/q;for(let dt=0;dt<V;++dt)for(let ht=0;ht<q;++ht)for(let bt=0;bt<H;++bt){let Et=bt-at,At=Math.max(0,Math.ceil(Et/rt)),Vt=Math.min(Z,(G+Et)/rt);for(let Zt=0;Zt<j;++Zt){let ce=Zt-nt,he=Math.max(0,Math.ceil(ce/ot)),jt=Math.min(et,(W+ce)/ot),ke=0;for(let fe=At;fe<Vt;++fe){let Ae=fe*rt-Et;for(let We=he;We<jt;++We){let _n=We*ot-ce,or=C*dt+N*fe+_*We,Hn=$*(G-1-Ae)+F*(W-1-_n)+P*ht;for(let Lr=0;Lr<it;++Lr){let qe=ht*it+Lr,Mr=w[or+qe],zr=A[Hn+Lr];ke+=Mr*zr}}}h[g*dt+x*bt+b*Zt+ht]=ke}}return e.makeTensorInfo(d.shape,d.dtype,d.values)}var aO={kernelName:xp,backendName:\"cpu\",kernelFunc:sQ};function iQ(r){let{inputs:t,backend:e}=r,{x:n}=t,o=y.sizeFromShape(n.shape),s=e.data.get(n.dataId).values,i=wt([o,o],n.dtype),a=i.values;for(let l=0;l<s.length;l++)a[l*o+l]=s[l];let u=[...n.shape,...n.shape];return e.makeTensorInfo(u,i.dtype,i.values)}var lO={kernelName:yp,backendName:\"cpu\",kernelFunc:iQ};var uO={kernelName:$l,backendName:\"cpu\",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o}=r,{strides:s,pad:i,dilations:a}=e,u=t,l=u.data.get(n.dataId).values,c=n.shape.length,p=u.data.get(o.dataId).values,m=o.shape.length,{batchSize:f,inHeight:d,inWidth:h,inChannels:g,outHeight:x,outWidth:b,padInfo:w,strideHeight:C,strideWidth:N,filterHeight:_,filterWidth:A,dilationHeight:$,dilationWidth:F,outShape:P}=v.computeDilation2DInfo(n.shape,o.shape,s,i,\"NHWC\",a),V=y.sizeFromShape(P),G=P.length,W=y.getArrayFromDType(n.dtype,V);for(let H=0;H<f;++H)for(let j=0;j<x;++j){let Y=j*C-w.top;for(let Z=0;Z<b;++Z){let et=Z*N-w.left;for(let rt=0;rt<g;++rt){let ot=Number.MIN_SAFE_INTEGER;for(let nt=0;nt<_;++nt){let it=Y+nt*$;if(it>=0&&it<d)for(let dt=0;dt<A;++dt){let ht=et+dt*F;if(ht>=0&&ht<h){let bt=y.locToIndex([H,it,ht,rt],c,y.computeStrides(n.shape)),Et=y.locToIndex([nt,dt,rt],m,y.computeStrides(o.shape)),At=l[bt]+p[Et];At>ot&&(ot=At)}}}let at=y.locToIndex([H,j,Z,rt],G,y.computeStrides(P));W[at]=ot}}}return{dataId:u.write(y.toTypedArray(W,n.dtype),P,n.dtype),shape:P,dtype:n.dtype}}};var cO={kernelName:Xd,backendName:\"cpu\",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o,dy:s}=r,{strides:i,pad:a,dilations:u}=e,l=t,c=y.toNestedArray(n.shape,l.data.get(n.dataId).values),p=y.toNestedArray(o.shape,l.data.get(o.dataId).values),{batchSize:m,inHeight:f,inWidth:d,inChannels:h,outHeight:g,outWidth:x,padInfo:b,strideHeight:w,strideWidth:C,filterHeight:N,filterWidth:_,dilationHeight:A,dilationWidth:$,outShape:F}=v.computeDilation2DInfo(n.shape,o.shape,i,a,\"NHWC\",u);y.assert(s.rank===F.length,()=>`Error in ${Xd}, dy must have the same rank as output ${F.length}, but got ${s.rank}`);let P=y.toNestedArray(F,l.data.get(s.dataId).values),V=y.makeZerosNestedTypedArray(o.shape,o.dtype);for(let W=0;W<m;++W)for(let q=0;q<g;++q){let H=q*w-b.top;for(let j=0;j<x;++j){let Y=j*C-b.left;for(let Z=0;Z<h;++Z){let et=Number.MIN_SAFE_INTEGER,rt=0,ot=0;for(let at=0;at<N;++at){let nt=H+at*A;if(nt>=0&&nt<f)for(let it=0;it<_;++it){let dt=Y+it*$;if(dt>=0&&dt<d){let ht=c[W][nt][dt][Z]+p[at][it][Z];ht>et&&(et=ht,rt=at,ot=it)}}}V[rt][ot][Z]+=P[W][q][j][Z]}}}return{dataId:l.write(y.toTypedArray(V,n.dtype),o.shape,o.dtype),shape:o.shape,dtype:o.dtype}}};var pO={kernelName:jd,backendName:\"cpu\",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o,dy:s}=r,{strides:i,pad:a,dilations:u}=e,l=t,c=y.toNestedArray(n.shape,l.data.get(n.dataId).values),p=y.toNestedArray(o.shape,l.data.get(o.dataId).values),{batchSize:m,inHeight:f,inWidth:d,inChannels:h,outHeight:g,outWidth:x,padInfo:b,strideHeight:w,strideWidth:C,filterHeight:N,filterWidth:_,dilationHeight:A,dilationWidth:$,outShape:F}=v.computeDilation2DInfo(n.shape,o.shape,i,a,\"NHWC\",u);y.assert(s.rank===F.length,()=>`Error in ${jd}, dy must have the same rank as output ${F.length}, but got ${s.rank}`);let P=y.toNestedArray(F,l.data.get(s.dataId).values),V=y.makeZerosNestedTypedArray(n.shape,n.dtype);for(let W=0;W<m;++W)for(let q=0;q<g;++q){let H=q*w-b.top;for(let j=0;j<x;++j){let Y=j*C-b.left;for(let Z=0;Z<h;++Z){let et=Number.MIN_SAFE_INTEGER,rt=H<0?0:H,ot=Y<0?0:Y;for(let at=0;at<N;++at){let nt=H+at*A;if(nt>=0&&nt<f)for(let it=0;it<_;++it){let dt=Y+it*$;if(dt>=0&&dt<d){let ht=c[W][nt][dt][Z]+p[at][it][Z];ht>et&&(et=ht,rt=nt,ot=dt)}}}V[W][rt][ot][Z]+=P[W][q][j][Z]}}}return{dataId:l.write(y.toTypedArray(V,n.dtype),n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};function hl(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,\"sum\");let a;o.dtype===\"bool\"?a=$o({inputs:{x:o},backend:e,attrs:{dtype:\"int32\"}}):a=Kr({inputs:{x:o},backend:e});let u=a.shape.length,l=y.parseAxisParam(s,a.shape),c=v.getAxesPermutation(l,u),p=l,m=a;c!=null&&(m=Ve({inputs:{x:a},backend:e,attrs:{perm:c}}),p=v.getInnerMostAxes(p.length,u)),v.assertAxesAreInnerMostDims(\"sum\",p,m.shape.length);let[f,d]=v.computeOutAndReduceShapes(m.shape,p),h=v.upcastType(m.dtype,\"int32\"),g=nd(e,f,h),x=y.sizeFromShape(d),b=e.data.get(g.dataId).values,w=e.data.get(m.dataId).values;for(let C=0;C<b.length;++C){let N=C*x,_=0;for(let A=0;A<x;++A)_+=w[N+A];b[C]=_}if(i){let C=v.expandShapeToKeepDim(g.shape,l),N=g;g=Yt({inputs:{x:g},backend:e,attrs:{shape:C}}),e.disposeIntermediateTensorInfo(N)}return e.disposeIntermediateTensorInfo(a),c!=null&&e.disposeIntermediateTensorInfo(m),g}var mO={kernelName:$s,backendName:\"cpu\",kernelFunc:hl};function aQ(r){let{inputs:t,backend:e,attrs:n}=r,{equation:o}=n,s=t,{allDims:i,summedDims:a,idDims:u}=v.decodeEinsumEquation(o,s.length);v.checkEinsumDimSizes(i.length,u,s);let{path:l,steps:c}=v.getEinsumComputePath(a,u),p=c.length,m=null,f=i.length,d=[];for(let h=0;h<p;++h){for(let g of c[h]){let{permutationIndices:x,expandDims:b}=v.getEinsumPermutation(f,u[g]),w;v.isIdentityPermutation(x)?w=s[g]:(w=Ve({inputs:{x:s[g]},backend:e,attrs:{perm:x}}),d.push(w));let C=w.shape.slice();for(let N=0;N<b.length;++N)C.splice(b[N],0,1);y.arraysEqual(w.shape,C)||(w=Yt({inputs:{x:w},backend:e,attrs:{shape:C}}),d.push(w)),m===null?m=w:(m=_c({inputs:{a:w,b:m},backend:e}),d.push(m))}h<p-1&&(l[h]>=0&&(m=hl({inputs:{x:m},backend:e,attrs:{axis:l[h]-(i.length-f),keepDims:!1}}),d.push(m)),f--)}for(let h of d)h!==m&&e.disposeIntermediateTensorInfo(h);return m}var fO={kernelName:bp,backendName:\"cpu\",kernelFunc:aQ};function lQ(r){let{inputs:t,backend:e}=r,{dy:n,y:o}=t;tt([n,o],\"eluGrad\");let s=new Float32Array(y.sizeFromShape(o.shape)),i=e.data.get(o.dataId).values,a=e.data.get(n.dataId).values;for(let u=0;u<i.length;++u){let l=i[u];l>=1?s[u]=a[u]:s[u]=a[u]*(l+1)}return e.makeTensorInfo(o.shape,\"float32\",s)}var dO={kernelName:wp,backendName:\"cpu\",kernelFunc:lQ};var uQ=v.ERF_P,cQ=v.ERF_A1,pQ=v.ERF_A2,mQ=v.ERF_A3,fQ=v.ERF_A4,dQ=v.ERF_A5,hQ=kt(ga,r=>{let t=Math.sign(r),e=Math.abs(r),n=1/(1+uQ*e);return t*(1-((((dQ*n+fQ)*n+mQ)*n+pQ)*n+cQ)*n*Math.exp(-e*e))}),hO={kernelName:ga,backendName:\"cpu\",kernelFunc:hQ};function cd(r){let{inputs:t,backend:e,attrs:n}=r,{input:o}=t,{dim:s}=n,i=o.shape.length,a=o.shape.slice(),u=s;return s<0&&(y.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),a.splice(u,0,1),Yt({inputs:{x:o},backend:e,attrs:{shape:a}})}var gO={kernelName:ui,backendName:\"cpu\",kernelFunc:cd};var gQ=Qt((r,t)=>r/t),Wh=oe(Qo,gQ),Uh={kernelName:Qo,backendName:\"cpu\",kernelFunc:Wh};function hw(r,t,e){let n=r.shape,o=n[0],s=n[1],i=e.data.get(r.dataId),a=i.complexTensorInfos.real,u=i.complexTensorInfos.imag,l=[o,s],c=y.sizeFromShape(l),p=y.getTypedArrayFromDType(\"float32\",c),m=y.getTypedArrayFromDType(\"float32\",c);for(let g=0;g<o;g++){let x=Fo({inputs:{x:a},backend:e,attrs:{begin:[g,0],size:[1,s]}}),b=Fo({inputs:{x:u},backend:e,attrs:{begin:[g,0],size:[1,s]}}),w=wr({inputs:{real:x,imag:b},backend:e}),{real:C,imag:N}=xQ(w,t,e),_=v.mergeRealAndImagArrays(C,N);for(let A=0;A<s;A++){let $=v.getComplexWithIndex(_,A);p[g*s+A]=$.real,m[g*s+A]=$.imag}e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(b),e.disposeIntermediateTensorInfo(w)}let f=e.makeTensorInfo(l,\"float32\",p),d=e.makeTensorInfo(l,\"float32\",m),h=wr({inputs:{real:f,imag:d},backend:e});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),h}function xQ(r,t,e){let n=y.sizeFromShape(r.shape),o=e.data.get(r.dataId),s=e.data.get(o.complexTensorInfos.real.dataId).values,i=e.data.get(o.complexTensorInfos.imag.dataId).values;if(yQ(n)){let a=gT(s,i,n,t,e),u=[r.shape[0],r.shape[1]];if(t){let l=e.makeTensorInfo(u,\"float32\",a.real),c=e.makeTensorInfo(u,\"float32\",a.imag),p=e.makeTensorInfo([],\"float32\",y.createScalarValue(n,\"float32\")),m=Kr({inputs:{x:p},backend:e}),f=Uh.kernelFunc({inputs:{a:l,b:p},backend:e}),d=Uh.kernelFunc({inputs:{a:c,b:m},backend:e}),h=e.data.get(f.dataId).values,g=e.data.get(d.dataId).values;return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),{real:h,imag:g}}return a}else{let a=v.mergeRealAndImagArrays(s,i),u=bQ(a,n,t);return v.splitRealAndImagArrays(u)}}function yQ(r){return(r&r-1)===0}function gT(r,t,e,n,o){if(e===1)return{real:r,imag:t};let s=v.mergeRealAndImagArrays(r,t),i=e/2,a=v.complexWithEvenIndex(s),u=a.real,l=a.imag,c=[u.length],p=o.makeTensorInfo(c,\"float32\",u),m=o.makeTensorInfo(c,\"float32\",l),f=wr({inputs:{real:p,imag:m},backend:o}),d=v.complexWithOddIndex(s),h=d.real,g=d.imag,x=[h.length],b=o.makeTensorInfo(x,\"float32\",h),w=o.makeTensorInfo(x,\"float32\",g),C=wr({inputs:{real:b,imag:w},backend:o}),N=gT(u,l,i,n,o),_=N.real,A=N.imag,$=[_.length],F=o.makeTensorInfo($,\"float32\",_),P=o.makeTensorInfo($,\"float32\",A),V=wr({inputs:{real:F,imag:P},backend:o}),G=gT(h,g,i,n,o),W=G.real,q=G.imag,H=[W.length],j=o.makeTensorInfo(H,\"float32\",W),Y=o.makeTensorInfo(H,\"float32\",q),Z=wr({inputs:{real:j,imag:Y},backend:o}),et=v.exponents(e,n),rt=[et.real.length],ot=o.makeTensorInfo(rt,\"float32\",et.real),at=o.makeTensorInfo(rt,\"float32\",et.imag),nt=wr({inputs:{real:ot,imag:at},backend:o}),it=_c({inputs:{a:nt,b:Z},backend:o}),dt=Ki({inputs:{a:V,b:it},backend:o}),ht=Vh({inputs:{a:V,b:it},backend:o}),bt=Ao({inputs:{input:dt},backend:o}),Et=Ao({inputs:{input:ht},backend:o}),At=ji({inputs:{input:dt},backend:o}),Vt=ji({inputs:{input:ht},backend:o}),Zt=Tu({inputs:[bt,Et],backend:o,attrs:{axis:0}}),ce=Tu({inputs:[At,Vt],backend:o,attrs:{axis:0}}),he=o.data.get(Zt.dataId).values,jt=o.data.get(ce.dataId).values;return o.disposeIntermediateTensorInfo(p),o.disposeIntermediateTensorInfo(m),o.disposeIntermediateTensorInfo(f),o.disposeIntermediateTensorInfo(b),o.disposeIntermediateTensorInfo(w),o.disposeIntermediateTensorInfo(C),o.disposeIntermediateTensorInfo(F),o.disposeIntermediateTensorInfo(P),o.disposeIntermediateTensorInfo(V),o.disposeIntermediateTensorInfo(j),o.disposeIntermediateTensorInfo(Y),o.disposeIntermediateTensorInfo(Z),o.disposeIntermediateTensorInfo(ot),o.disposeIntermediateTensorInfo(at),o.disposeIntermediateTensorInfo(nt),o.disposeIntermediateTensorInfo(it),o.disposeIntermediateTensorInfo(dt),o.disposeIntermediateTensorInfo(ht),o.disposeIntermediateTensorInfo(bt),o.disposeIntermediateTensorInfo(At),o.disposeIntermediateTensorInfo(Et),o.disposeIntermediateTensorInfo(Vt),o.disposeIntermediateTensorInfo(Zt),o.disposeIntermediateTensorInfo(ce),{real:he,imag:jt}}function bQ(r,t,e){let n=new Float32Array(t*2);for(let o=0;o<t;o++){let s=0,i=0;for(let a=0;a<t;a++){let u=v.exponent(o*a,t,e),l=v.getComplexWithIndex(r,a);s+=l.real*u.real-l.imag*u.imag,i+=l.real*u.imag+l.imag*u.real}e&&(s/=t,i/=t),v.assignToTypedArray(n,s,i,o)}return n}function wQ(r){let{inputs:t,backend:e}=r,{input:n}=t,o=y.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=o/s,a=Yt({inputs:{x:n},backend:e,attrs:{shape:[i,s]}}),u=hw(a,!1,e),l=Yt({inputs:{x:u},backend:e,attrs:{shape:n.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(u),l}var xO={kernelName:Cp,backendName:\"cpu\",kernelFunc:wQ};function Hh(r){let{backend:t,attrs:e}=r,{shape:n,value:o,dtype:s}=e,i=s||y.inferDtype(o),a=y.getArrayFromDType(i,y.sizeFromShape(n));return CQ(a,o,i),t.makeTensorInfo(n,i,a)}var yO={kernelName:Dl,backendName:\"cpu\",kernelFunc:Hh};function CQ(r,t,e){r.fill(t)}var bO={kernelName:ba,backendName:\"cpu\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,o=e,s=y.getTypedArrayFromDType(n.dtype,y.sizeFromShape(n.shape)),[i,a,u,l]=n.shape,c=o.data.get(n.dataId).values;for(let m=0;m<i;m++){let f=m*u*a*l;for(let d=0;d<a;d++){let h=d*(u*l);for(let g=0;g<u;g++){let x=g*l;for(let b=0;b<l;b++){let w=Math.round(u-g-1),C=f+h+x+b,N=c[C];if(w>=0&&w<u){let _=w*l,A=f+h+_+b;N=c[A]}s[C]=N}}}}return{dataId:o.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};var IQ=Qt((r,t)=>Math.floor(r/t)),SQ=oe(ns,IQ,null,\"int32\"),wO={kernelName:ns,backendName:\"cpu\",kernelFunc:SQ};function vQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=dT({inputs:{x:o,filter:s},backend:e,attrs:{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m}});if(i){let g=h;if(c===\"NCHW\"&&i.shape.length===1&&i.shape[0]!==1){let x=Yt({inputs:{x:i},backend:e,attrs:{shape:[i.shape[0],1,1]}});h=Ki({inputs:{a:h,b:x},backend:e}),e.disposeIntermediateTensorInfo(x)}else h=Ki({inputs:{a:h,b:i},backend:e});e.disposeIntermediateTensorInfo(g)}if(f){let g=h;if(c===\"NCHW\"&&f===\"prelu\"&&a.shape.length===1&&a.shape[0]!==1){let x=Yt({inputs:{x:a},backend:e,attrs:{shape:[a.shape[0],1,1]}});h=Oc(e,h,f,x,d),e.disposeIntermediateTensorInfo(x)}else h=Oc(e,h,f,a,d);e.disposeIntermediateTensorInfo(g)}return h}var CO={kernelName:Ii,backendName:\"cpu\",kernelFunc:vQ};function NQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=hT({inputs:{x:o,filter:s},backend:e,attrs:{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m}});if(i){let g=h;h=Ki({inputs:{a:h,b:i},backend:e}),e.disposeIntermediateTensorInfo(g)}if(f){let g=h;h=Oc(e,h,f,a,d),e.disposeIntermediateTensorInfo(g)}return h}var IO={kernelName:Si,backendName:\"cpu\",kernelFunc:NQ};function TQ(r){let{inputs:t,backend:e}=r,{params:n,indices:o}=t,s=y.sizeFromShape(n.shape),i=o.shape,a=i[i.length-1],[u,l,c,p]=v.prepareAndValidate(n,o);if(l===0)return e.makeTensorInfo(u,n.dtype,[]);let m=e.data.get(o.dataId).values,f=e.bufferSync(n),d=Qb(m,f,n.dtype,l,a,c,p,n.shape,s);return e.makeTensorInfo(u,n.dtype,d.values)}var SO={kernelName:wa,backendName:\"cpu\",kernelFunc:TQ};function kQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,indices:s}=t,{axis:i,batchDims:a}=n;tt([o,s],\"gatherV2\");let u=y.parseAxisParam(i,o.shape)[0],l=e.data.get(s.dataId).values,c=o.shape[u];for(let C=0;C<l.length;++C){let N=l[C];y.assert(N<=c-1&&N>=0,()=>`GatherV2: the index value ${N} is not in [0, ${c-1}]`)}let p=a;a==null&&(p=0);let m=y.sizeFromShape(s.shape),f=v.segment_util.collectGatherOpShapeInfo(o,s,u,p),d=Yt({inputs:{x:o},backend:e,attrs:{shape:[f.batchSize,f.outerSize,f.dimSize,f.sliceSize]}}),h=Yt({inputs:{x:s},backend:e,attrs:{shape:[f.batchSize,m/f.batchSize]}}),g=[f.batchSize,f.outerSize,m/f.batchSize,f.sliceSize],x=e.bufferSync(h),b=e.bufferSync(d),w=tw(b,x,g);return e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),e.makeTensorInfo(f.outputShape,w.dtype,w.values)}var vO={kernelName:ci,backendName:\"cpu\",kernelFunc:kQ};function EQ(r){let{inputs:t,backend:e}=r,{input:n}=t,o=y.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=o/s,a=Yt({inputs:{x:n},backend:e,attrs:{shape:[i,s]}}),u=hw(a,!0,e),l=Yt({inputs:{x:u},backend:e,attrs:{shape:n.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(u),l}var NO={kernelName:Ip,backendName:\"cpu\",kernelFunc:EQ};var _Q=kt(Ia,r=>Number.isFinite(r)?1:0,\"bool\"),TO={kernelName:Ia,backendName:\"cpu\",kernelFunc:_Q};var AQ=kt(Sa,r=>Math.abs(r)===1/0?1:0,\"bool\"),kO={kernelName:Sa,backendName:\"cpu\",kernelFunc:AQ};var $Q=kt(va,r=>Number.isNaN(r)?1:0,\"bool\"),EO={kernelName:va,backendName:\"cpu\",kernelFunc:$Q};function DQ(r){let{backend:t,attrs:e}=r,{start:n,stop:o,num:s}=e,i=ew(n,o,s);return t.makeTensorInfo([i.length],\"float32\",i)}var _O={kernelName:vp,backendName:\"cpu\",kernelFunc:DQ};var RQ=kt(ka,r=>Math.log1p(r)),AO={kernelName:ka,backendName:\"cpu\",kernelFunc:RQ};var FQ=Qt((r,t)=>r&&t),OQ=oe(Ea,FQ,null,\"bool\"),$O={kernelName:Ea,backendName:\"cpu\",kernelFunc:OQ};var PQ=kt(_a,r=>r?0:1,\"bool\"),DO={kernelName:_a,backendName:\"cpu\",kernelFunc:PQ};var LQ=Qt((r,t)=>r||t),MQ=oe(Aa,LQ,null,\"bool\"),RO={kernelName:Aa,backendName:\"cpu\",kernelFunc:MQ};function zQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{depthRadius:s,bias:i,alpha:a,beta:u}=n;tt(o,\"LRN\");let l=o.shape[3],c=l-1,p=e.data.get(o.dataId).values,m=y.sizeFromShape(o.shape),f=new Float32Array(m);function d(h){let g=h%l,x=h-g+Math.max(0,g-s),b=h-g+Math.min(g+s,c),w=0;for(;x<=b;x++){let C=p[x];w+=C*C}return w}for(let h=0;h<m;h++){let g=d(h),x=p[h]*Math.pow(i+a*g,-u);f[h]=x}return e.makeTensorInfo(o.shape,o.dtype,f)}var FO={kernelName:Rl,backendName:\"cpu\",kernelFunc:zQ};function BQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,y:s,dy:i}=t,{depthRadius:a,bias:u,alpha:l,beta:c}=n;tt(i,\"LRNGrad\");let p=y.sizeFromShape(i.shape),m=i.shape[3],f=e.data.get(i.dataId).values,d=e.data.get(o.dataId).values,h=e.data.get(s.dataId).values,g=new Float32Array(p),x=p;for(let b=0;b<x;b++){let w=b%m,C=b-w+Math.max(0,w-a),N=b-w+Math.min(m,w+a+1),_=0;for(let A=C;A<N;A++)_+=Math.pow(d[A],2);_=l*_+u;for(let A=C;A<N;A++){let $=-2*l*c*d[A]*h[b]/_;b===A&&($+=Math.pow(_,-c)),$*=f[b],g[A]+=$}}return e.makeTensorInfo(i.shape,o.dtype,g)}var OO={kernelName:Np,backendName:\"cpu\",kernelFunc:BQ};function xT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reductionIndices:s,keepDims:i}=n,a=e,u=o.shape,l=u.length,c=y.parseAxisParam(s,u),p=c,m=v.getAxesPermutation(p,l),f=a.data.get(o.dataId).values;if(m!=null){let C=new Array(l);for(let N=0;N<C.length;N++)C[N]=u[m[N]];f=id(f,u,o.dtype,m,C),p=v.getInnerMostAxes(p.length,l),u=C}tt(o,\"max\"),v.assertAxesAreInnerMostDims(\"max\",p,l);let[d,h]=v.computeOutAndReduceShapes(u,p),g=y.sizeFromShape(h),x=rw(f,g,d,o.dtype),b=a.write(x,d,o.dtype),w=d;return i&&(w=v.expandShapeToKeepDim(d,c)),{dataId:b,shape:w,dtype:o.dtype}}var PO={kernelName:ls,backendName:\"cpu\",kernelFunc:xT};function VQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;tt(o,\"maxPool\");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u),p;if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))p=Kr({inputs:{x:o},backend:e});else{let m=e.data.get(o.dataId).values,f=y.computeStrides(o.shape),d=ud(m,o.shape,o.dtype,f,c,\"max\");p=e.makeTensorInfo(c.outShape,o.dtype,d.values)}return p}var LO={kernelName:cs,backendName:\"cpu\",kernelFunc:VQ};function GQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n;tt(o,\"maxPool3d\");let c=v.computePool3DInfo(o.shape,s,i,1,a,u,l),p=e.data.get(o.dataId).values,m=dw(p,o.shape,o.dtype,y.computeStrides(o.shape),c,\"max\");return e.makeTensorInfo(m.shape,\"float32\",m.values)}var MO={kernelName:Fl,backendName:\"cpu\",kernelFunc:GQ};function WQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=n;tt([o,s],\"maxPool3DGrad\");let c=v.computePool3DInfo(s.shape,i,a,1,u,l),p=e.bufferSync(s),m=RF(p,c),f=c.strideDepth,d=c.strideHeight,h=c.strideWidth,g=c.dilationDepth,x=c.dilationHeight,b=c.dilationWidth,w=c.effectiveFilterDepth,C=c.effectiveFilterHeight,N=c.effectiveFilterWidth,_=w-1-c.padInfo.front,A=N-1-c.padInfo.left,$=C-1-c.padInfo.top,F=wt(s.shape,\"float32\"),P=e.bufferSync(o);for(let V=0;V<c.batchSize;++V)for(let G=0;G<c.inChannels;++G)for(let W=0;W<c.inDepth;++W)for(let q=0;q<c.inHeight;++q)for(let H=0;H<c.inWidth;++H){let j=W-_,Y=q-$,Z=H-A,et=0;for(let rt=0;rt<w;rt+=g){let ot=(j+rt)/f;if(!(ot<0||ot>=c.outDepth||Math.floor(ot)!==ot))for(let at=0;at<C;at+=x){let nt=(Y+at)/d;if(!(nt<0||nt>=c.outHeight||Math.floor(nt)!==nt))for(let it=0;it<N;it+=b){let dt=(Z+it)/h;if(dt<0||dt>=c.outWidth||Math.floor(dt)!==dt)continue;let ht=w*C*N-1-m.get(V,ot,nt,dt,G),bt=rt*C*N+at*N+it,Et=ht===bt?1:0;if(Et===0)continue;et+=P.get(V,ot,nt,dt,G)*Et}}}F.set(et,V,W,q,H,G)}return e.makeTensorInfo(F.shape,F.dtype,F.values)}var zO={kernelName:kp,backendName:\"cpu\",kernelFunc:WQ};function UQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s,output:i}=t,a=s;tt([s,i],\"maxPoolGrad\");let{filterSize:u,strides:l,pad:c,dimRoundingMode:p}=n,m=v.computePool2DInfo(a.shape,u,l,1,c,p),f=e.data.get(a.dataId).values,d=wt(m.outShape,a.dtype,fw(f,a.shape,a.dtype,m).values),h=m.strideHeight,g=m.strideWidth,x=m.dilationHeight,b=m.dilationWidth,w=m.effectiveFilterHeight,C=m.effectiveFilterWidth,N=C-1-m.padInfo.left,_=w-1-m.padInfo.top,A=wt(a.shape,\"float32\"),$=e.data.get(o.dataId).values,F=wt(o.shape,\"float32\",$);for(let P=0;P<m.batchSize;++P)for(let V=0;V<m.inChannels;++V)for(let G=0;G<m.inHeight;++G)for(let W=0;W<m.inWidth;++W){let q=G-_,H=W-N,j=0;for(let Y=0;Y<w;Y+=x){let Z=(q+Y)/h;if(!(Z<0||Z>=m.outHeight||Math.floor(Z)!==Z))for(let et=0;et<C;et+=b){let rt=(H+et)/g;if(rt<0||rt>=m.outWidth||Math.floor(rt)!==rt)continue;let ot=w*C-1-d.get(P,Z,rt,V),at=Y*C+et,nt=ot===at?1:0;if(nt===0)continue;j+=F.get(P,Z,rt,V)*nt}}A.set(j,P,G,W,V)}return e.makeTensorInfo(A.shape,A.dtype,A.values)}var BO={kernelName:Tp,backendName:\"cpu\",kernelFunc:UQ};function VO(r,t,e,n,o){let s=y.computeStrides(t),i=ud(r,t,e,s,o,\"max\"),a=fw(r,t,e,o,!0,n);return[i.values,a.values]}var GO={kernelName:Ep,backendName:\"cpu\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{filterSize:o,strides:s,pad:i,includeBatchInIndex:a}=t,u=e;tt(n,\"MaxPoolWithArgmax\");let l=u.data.get(n.dataId).values,c=v.computePool2DInfo(n.shape,o,s,[1,1],i),[p,m]=VO(l,n.shape,n.dtype,a,c),f=u.write(p,c.outShape,n.dtype),d=u.write(m,c.outShape,n.dtype);return[{dataId:f,shape:c.outShape,dtype:n.dtype},{dataId:d,shape:c.outShape,dtype:\"int32\"}]}};function HQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=y.parseAxisParam(s,o.shape),l=v.computeOutAndReduceShapes(o.shape,a)[1],c=y.sizeFromShape(l),p=[],m=e.makeTensorInfo([],\"float32\",new Float32Array([c]));p.push(m);let f=$o({inputs:{x:o},backend:e,attrs:{dtype:\"float32\"}});p.push(f);let d=Wh({inputs:{a:f,b:m},backend:e});p.push(d);let h=hl({inputs:{x:d},backend:e,attrs:{axis:s,keepDims:i}});return p.forEach(g=>e.disposeIntermediateTensorInfo(g)),h}var WO={kernelName:ps,backendName:\"cpu\",kernelFunc:HQ};function qQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,\"min\");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims(\"min\",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];(Number.isNaN(N)||N<w)&&(w=N)}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var UO={kernelName:ms,backendName:\"cpu\",kernelFunc:qQ};function KQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,mode:i}=n;tt(o,\"mirrorPad\");let a=s.map((w,C)=>w[0]+o.shape[C]+w[1]),u=s.map(w=>w[0]),l=s.map((w,C)=>w[0]+o.shape[C]),c=i===\"reflect\"?0:1,p=e.data.get(o.dataId).values,m=o.shape.length,f=y.computeStrides(o.shape),d=y.sizeFromShape(a),h=a.length,g=y.computeStrides(a),x=y.getTypedArrayFromDType(o.dtype,d);for(let w=0;w<d;w++){let C=y.indexToLoc(w,h,g);for(let _=0;_<h;_++)C[_]<u[_]?C[_]=u[_]*2-C[_]-c:C[_]>=l[_]&&(C[_]=(l[_]-1)*2-C[_]+c);C=C.map((_,A)=>_-u[A]);let N=y.locToIndex(C,m,f);x[w]=p[N]}return{dataId:e.write(x,a,o.dtype),shape:a,dtype:o.dtype}}var HO={kernelName:ds,backendName:\"cpu\",kernelFunc:KQ};var jQ=Qt((r,t)=>{let e=r%t;return r<0&&t<0||r>=0&&t>=0?e:(e+t)%t}),XQ=oe($a,jQ),qO={kernelName:$a,backendName:\"cpu\",kernelFunc:XQ};var jO=Tl(fh());function yT(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{dim:s}=n,i=o.shape.length,a=s;if(a===-1&&(a=i-1),a!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${a}`);let u=y.parseAxisParam([a],o.shape),l=xT({inputs:{x:o},backend:e,attrs:{reductionIndices:u,keepDims:!1}}),c=v.expandShapeToKeepDim(l.shape,u),p=Yt({inputs:{x:l},backend:e,attrs:{shape:c}}),m=Vh({inputs:{a:o,b:p},backend:e}),f=UN({inputs:{x:m},backend:e}),d=hl({inputs:{x:f},backend:e,attrs:{axis:u,keepDims:!1}}),h=Yt({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Wh({inputs:{a:f,b:h},backend:e});return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),g}var KO={kernelName:Ds,backendName:\"cpu\",kernelFunc:yT};function YQ(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{numSamples:s,seed:i,normalized:a}=n;tt(o,\"multinomial\");let u=a?o:yT({inputs:{logits:o},backend:e,attrs:{dim:-1}}),l=u.shape[0],c=u.shape[1],p=e.data.get(u.dataId).values,m=[l,s],f=y.makeZerosTypedArray(y.sizeFromShape(m),\"int32\");for(let d=0;d<l;++d){let h=d*c,g=new Float32Array(c-1);g[0]=p[h];for(let w=1;w<g.length;++w)g[w]=g[w-1]+p[h+w];let x=jO.alea(i.toString()),b=d*s;for(let w=0;w<s;++w){let C=x();f[b+w]=g.length;for(let N=0;N<g.length;N++)if(C<g[N]){f[b+w]=N;break}}}return a||e.disposeIntermediateTensorInfo(u),e.makeTensorInfo(m,\"int32\",f)}var XO={kernelName:_p,backendName:\"cpu\",kernelFunc:YQ};var ZQ=Ur.nonMaxSuppressionV3Impl;function JQ(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u}=n;tt(o,\"NonMaxSuppression\");let l=e.data.get(o.dataId).values,c=e.data.get(s.dataId).values,{selectedIndices:p}=ZQ(l,c,i,a,u);return e.makeTensorInfo([p.length],\"int32\",new Int32Array(p))}var YO={kernelName:Ra,backendName:\"cpu\",kernelFunc:JQ};var QQ=Ur.nonMaxSuppressionV4Impl;function t9(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,padToMaxOutputSize:l}=n;tt(o,\"NonMaxSuppressionPadded\");let c=e.data.get(o.dataId).values,p=e.data.get(s.dataId).values,{selectedIndices:m,validOutputs:f}=QQ(c,p,i,a,u,l);return[e.makeTensorInfo([m.length],\"int32\",new Int32Array(m)),e.makeTensorInfo([],\"int32\",new Int32Array([f]))]}var ZO={kernelName:Fa,backendName:\"cpu\",kernelFunc:t9};var e9=Ur.nonMaxSuppressionV5Impl;function r9(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,softNmsSigma:l}=n;tt(o,\"NonMaxSuppressionWithScore\");let c=e.data.get(o.dataId).values,p=e.data.get(s.dataId).values,m=i,f=a,d=u,h=l,{selectedIndices:g,selectedScores:x}=e9(c,p,m,f,d,h);return[e.makeTensorInfo([g.length],\"int32\",new Int32Array(g)),e.makeTensorInfo([x.length],\"float32\",new Float32Array(x))]}var JO={kernelName:Oa,backendName:\"cpu\",kernelFunc:r9};function n9(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n;tt(o,\"oneHot\");let l=y.sizeFromShape(o.shape),c=new Float32Array(l*i);c.fill(u);let p=e.data.get(o.dataId).values;for(let m=0;m<l;++m)p[m]>=0&&p[m]<i&&(c[m*i+p[m]]=a);return e.makeTensorInfo([...o.shape,i],s,c)}var QO={kernelName:gs,backendName:\"cpu\",kernelFunc:n9};function qh(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype===\"string\")throw new Error(\"zerosLike is not supported for string tensors\");if(n.dtype===\"complex64\"){let o=Ao({inputs:{input:n},backend:e}),s=qh({inputs:{x:o},backend:e}),i=ji({inputs:{input:n},backend:e}),a=qh({inputs:{x:i},backend:e}),u=wr({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Hh({backend:e,attrs:{shape:n.shape,value:0,dtype:n.dtype}})}var tP={kernelName:wi,backendName:\"cpu\",kernelFunc:qh};function eP(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype===\"string\")throw new Error(\"onesLike is not supported for string tensors\");if(n.dtype===\"complex64\"){let o=Ao({inputs:{input:n},backend:e}),s=eP({inputs:{x:o},backend:e}),i=ji({inputs:{input:n},backend:e}),a=qh({inputs:{x:i},backend:e}),u=wr({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Hh({backend:e,attrs:{shape:n.shape,value:1,dtype:n.dtype}})}var rP={kernelName:mi,backendName:\"cpu\",kernelFunc:eP};function bT(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return cd({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,\"All tensors passed to stack must have matching shapes\"),y.assert(i===c.dtype,()=>\"All tensors passed to stack must have matching dtypes\")});let a=[],u=t.map(c=>{let p=cd({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=Tu({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeIntermediateTensorInfo(c)),l}var nP={kernelName:fi,backendName:\"cpu\",kernelFunc:bT};function o9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,constantValue:i}=n;tt(o,\"pad\");let a=s.map((b,w)=>b[0]+o.shape[w]+b[1]),u=s.map(b=>b[0]),l=e.data.get(o.dataId).values,c=y.sizeFromShape(o.shape),p=o.shape.length,m=y.computeStrides(o.shape),f=y.sizeFromShape(a),d=a.length,h=y.computeStrides(a),g=y.getTypedArrayFromDType(o.dtype,f);i!==0&&g.fill(i);for(let b=0;b<c;b++){let C=y.indexToLoc(b,p,m).map((_,A)=>_+u[A]),N=y.locToIndex(C,d,h);g[N]=l[b]}return{dataId:e.write(g,a,o.dtype),shape:a,dtype:o.dtype}}var gw={kernelName:xs,backendName:\"cpu\",kernelFunc:o9};var s9=Qt((r,t)=>Math.pow(r,t)),i9=oe(ys,s9),oP={kernelName:ys,backendName:\"cpu\",kernelFunc:i9};function a9(r){let{inputs:t,backend:e,attrs:n}=r,{paramsNestedSplits:o,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:a}=n,u=o.map(x=>e.data.get(x.dataId).values),l=o.map(x=>x.shape),c=e.data.get(s.dataId).values,p=e.data.get(i.dataId).values,[m,f,d]=nw(u,l,c,s.shape,s.dtype,p,i.shape,a),h=m.map(x=>e.makeTensorInfo([x.length],\"int32\",x)),g=e.makeTensorInfo(d,s.dtype,f);return h.concat([g])}var sP={kernelName:Ap,backendName:\"cpu\",kernelFunc:a9};function l9(r){let{inputs:t,backend:e}=r,{starts:n,limits:o,deltas:s}=t,i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ow(i,n.shape,n.dtype,a,o.shape,u,s.shape),p=e.makeTensorInfo([l.length],\"int32\",l),m=e.makeTensorInfo([c.length],n.dtype,c);return[p,m]}var iP={kernelName:$p,backendName:\"cpu\",kernelFunc:l9};function u9(r){let{inputs:t,backend:e,attrs:n}=r,{shape:o,values:s,defaultValue:i,rowPartitionTensors:a}=t,{rowPartitionTypes:u}=n,l=e.data.get(o.dataId).values,c=e.data.get(s.dataId).values,p=e.data.get(i.dataId).values,m=a.map(g=>e.data.get(g.dataId).values),f=a.map(g=>g.shape),[d,h]=sw(l,o.shape,c,s.shape,s.dtype,p,i.shape,m,f,u);return e.makeTensorInfo(d,s.dtype,h)}var aP={kernelName:Dp,backendName:\"cpu\",kernelFunc:u9};function c9(r){let{backend:t,attrs:e}=r,{start:n,stop:o,dtype:s,step:i}=e,a=Ac(n,o,i,s);return t.makeTensorInfo([a.length],s,a)}var lP={kernelName:Ol,backendName:\"cpu\",kernelFunc:c9};var p9=kt(Pa,r=>1/r),uP={kernelName:Pa,backendName:\"cpu\",kernelFunc:p9};function m9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n;tt(o,\"resizeBilinear\");let u=y.computeStrides(o.shape),[l,c]=a,[p,m,f,d]=o.shape,h=e.data.get(o.dataId).values,g=new Float32Array(y.sizeFromShape([p,l,c,d])),x=[s&&l>1?m-1:m,s&&c>1?f-1:f],b=[s&&l>1?l-1:l,s&&c>1?c-1:c],w=0,C=x[0]/b[0],N=x[1]/b[1];for(let _=0;_<p;_++)for(let A=0;A<l;A++){let $;i?$=C*(A+.5)-.5:$=C*A;let F=Math.max(0,Math.floor($)),P=$-F,V=Math.min(m-1,Math.ceil($)),G=_*u[0]+F*u[1],W=_*u[0]+V*u[1];for(let q=0;q<c;q++){let H;i?H=N*(q+.5)-.5:H=N*q;let j=Math.max(0,Math.floor(H)),Y=H-j,Z=Math.min(f-1,Math.ceil(H)),et=G+j*u[2],rt=W+j*u[2],ot=G+Z*u[2],at=W+Z*u[2];for(let nt=0;nt<d;nt++){let it=h[et+nt],dt=h[rt+nt],ht=h[ot+nt],bt=h[at+nt],Et=it+(ht-it)*Y,At=dt+(bt-dt)*Y,Vt=Et+(At-Et)*P;g[w++]=Vt}}}return e.makeTensorInfo([p,l,c,d],\"float32\",g)}var cP={kernelName:Ss,backendName:\"cpu\",kernelFunc:m9};function f9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n;tt([s,o],\"resizeBilinearGrad\");let a=y.computeStrides(o.shape),[u,l,c,p]=o.shape,[,m,f]=s.shape,d=new Float32Array(u*l*c*p),h=[i&&m>1?l-1:l,i&&f>1?c-1:c],g=[i&&m>1?m-1:m,i&&f>1?f-1:f],x=h[0]/g[0],b=h[1]/g[1],w=e.data.get(s.dataId).values,C=0;for(let N=0;N<u;N++){let _=N*a[0];for(let A=0;A<m;A++){let $=A*x,F=Math.floor($),P=Math.min(Math.ceil($),l-1),V=_+F*a[1],G=_+P*a[1],W=$-F,q=1-W;for(let H=0;H<f;H++){let j=H*b,Y=Math.floor(j),Z=Math.min(Math.ceil(j),c-1),et=j-Y,rt=1-et,ot=V+Y*a[2],at=V+Z*a[2],nt=G+Y*a[2],it=G+Z*a[2],dt=q*rt,ht=q*et,bt=W*rt,Et=W*et;for(let At=0;At<p;At++){let Vt=w[C++];d[ot+At]+=Vt*dt,d[at+At]+=Vt*ht,d[nt+At]+=Vt*bt,d[it+At]+=Vt*Et}}}}return e.makeTensorInfo([u,c,l,p],\"float32\",d)}var pP={kernelName:Op,backendName:\"cpu\",kernelFunc:f9};function d9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n;tt(o,\"resizeNearestNeighbor\");let u=y.computeStrides(o.shape),[l,c]=a,[p,m,f,d]=o.shape,h=e.data.get(o.dataId).values,g=new Float32Array(p*l*c*d),x=[s&&l>1?m-1:m,s&&c>1?f-1:f],b=[s&&l>1?l-1:l,s&&c>1?c-1:c],w=x[0]/b[0],C=x[1]/b[1],N=0;for(let _=0;_<p;_++){let A=_*u[0];for(let $=0;$<l;$++){let F=i?w*($+.5):w*$,P=Math.min(m-1,s?Math.round(F):Math.floor(F));i&&(P=Math.max(0,P));let V=A+P*u[1];for(let G=0;G<c;G++){let W=i?C*(G+.5):C*G,q=Math.min(f-1,s?Math.round(W):Math.floor(W));i&&(q=Math.max(0,q));let H=V+q*u[2];for(let j=0;j<d;j++){let Y=h[H+j];g[N++]=Y}}}}return e.makeTensorInfo([p,l,c,d],o.dtype,g)}var mP={kernelName:Is,backendName:\"cpu\",kernelFunc:d9};function h9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n;tt([s,o],\"resizeNearestNeighborGrad\");let a=y.computeStrides(o.shape),u=y.computeStrides(s.shape),[l,c,p,m]=o.shape,[,f,d]=s.shape,h=new Float32Array(l*c*p*m),g=e.data.get(s.dataId).values,x=[i&&f>1?c-1:c,i&&d>1?p-1:p],b=[i&&f>1?f-1:f,i&&d>1?d-1:d],w=x[0]/b[0],C=x[1]/b[1],N=1/w,_=1/C,A=Math.ceil(N)*2+2,$=Math.ceil(_)*2+2;for(let F=0;F<l;F++){let P=F*a[0];for(let V=0;V<c;V++){let G=P+V*a[1],W=Math.floor(V*N),q=Math.floor(W-A/2);for(let H=0;H<p;H++){let j=G+H*a[2],Y=Math.floor(H*_),Z=Math.floor(Y-$/2);for(let et=0;et<m;et++){let rt=0;for(let ot=0;ot<A;ot++){let at=ot+q;if(at<0||at>=f)continue;let nt=P+at*u[1],it=at*w,dt=Math.min(c-1,i?Math.round(it):Math.floor(it));if(V===dt)for(let ht=0;ht<$;ht++){let bt=ht+Z;if(bt<0||bt>=d)continue;let Et=nt+bt*u[2],At=bt*C,Vt=Math.min(p-1,i?Math.round(At):Math.floor(At));H===Vt&&(rt+=g[Et+et])}}h[j+et]=rt}}}}return e.makeTensorInfo(o.shape,o.dtype,h)}var fP={kernelName:Fp,backendName:\"cpu\",kernelFunc:h9};function g9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n;tt(o,\"reverse\");let i=o.shape.length,a=y.parseAxisParam(s,o.shape);if(i===0)return Kr({inputs:{x:o},backend:e});let u=new pe(o.shape,o.dtype),l=e.bufferSync(o);for(let c=0;c<u.size;c++){let p=u.indexToLoc(c),m=p.slice();a.forEach(f=>m[f]=o.shape[f]-1-m[f]),u.set(l.get(...m),...p)}return e.makeTensorInfo(u.shape,u.dtype,u.values)}var dP={kernelName:Ns,backendName:\"cpu\",kernelFunc:g9};var hP={kernelName:qa,backendName:\"cpu\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,{radians:o,fillValue:s,center:i}=t,a=e,u=y.getTypedArrayFromDType(n.dtype,y.sizeFromShape(n.shape)),[l,c,p,m]=n.shape,[f,d]=v.getImageCenter(i,c,p),h=255,g=Math.sin(o),x=Math.cos(o),b=a.data.get(n.dataId).values;for(let C=0;C<l;C++){let N=C*p*c*m;for(let _=0;_<c;_++){let A=_*(p*m);for(let $=0;$<p;$++){let F=$*m;for(let P=0;P<m;P++){let V=[l,_,$,P],G=V[2],W=V[1],q=(G-f)*x-(W-d)*g,H=(G-f)*g+(W-d)*x;q=Math.round(q+f),H=Math.round(H+d);let j=s;if(typeof s!=\"number\"&&(P===3?j=h:j=s[P]),q>=0&&q<p&&H>=0&&H<c){let Z=H*(p*m),et=q*m,rt=N+Z+et+P;j=b[rt]}let Y=N+A+F+P;u[Y]=j}}}}return{dataId:a.write(u,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};var x9=kt(Ts,r=>{let t=Math.floor(r);return r-t<.5?Math.floor(r):r-t>.5?Math.ceil(r):t%2===0?t:t+1}),gP={kernelName:Ts,backendName:\"cpu\",kernelFunc:x9};function y9(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o,updates:s}=t,{shape:i}=n,{sliceRank:a,numUpdates:u,sliceSize:l,strides:c,outputSize:p}=v.calculateShapes(s,o,i),m=!0,f=e.bufferSync(o),d=e.bufferSync(s),h=dl(f,d,i,p,l,u,a,c,0,m);return e.makeTensorInfo(i,h.dtype,h.values)}var xP={kernelName:La,backendName:\"cpu\",kernelFunc:y9};function b9(r,t){let e=0,n=r.length,o=0;for(;e<n;)o=Math.floor((e+n)/2),r[o]<t?e=o+1:n=o;return n}function w9(r,t){let e=0,n=r.length,o=0;for(;e<n;)o=Math.floor((e+n)/2),r[o]<=t?e=o+1:n=o;return n}function yP(r,t,e,n,o,s){let i=y.getArrayFromDType(\"int32\",e*o);for(let a=0;a<e;++a){let u=r.slice(a*n,(a+1)*n),l=a*o;for(let c=0;c<o;++c)i[l+c]=s===\"left\"?b9(u,t[c+l]):w9(u,t[c+l])}return i}function C9(r){let{inputs:t,backend:e,attrs:n}=r,{sortedSequence:o,values:s}=t,{side:i}=n,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,l=yP(a,u,o.shape[0],o.shape[1],s.shape[1],i);return e.makeTensorInfo(s.shape,\"int32\",l)}var bP={kernelName:Pp,backendName:\"cpu\",kernelFunc:C9};function I9(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t;tt([n,o,s],\"select\");let i=n.shape.length,a=e.data.get(n.dataId).values,u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=sr(o.dtype,s.dtype),p=y.makeZerosTypedArray(y.sizeFromShape(o.shape),c),m=0,f=i===0||i>1||o.shape.length===1?1:y.sizeFromShape(o.shape.slice(1));for(let d=0;d<a.length;d++)for(let h=0;h<f;h++)a[d]===1?p[m++]=u[d]:p[m++]=l[d];return e.makeTensorInfo(o.shape,c,p)}var wP={kernelName:hi,backendName:\"cpu\",kernelFunc:I9};var S9=v.SELU_SCALEALPHA,v9=v.SELU_SCALE,N9=kt(Ma,r=>r>=0?v9*r:S9*(Math.exp(r)-1)),CP={kernelName:Ma,backendName:\"cpu\",kernelFunc:N9};var T9=kt(Ba,r=>r<0?-1:r>0?1:0),IP={kernelName:Ba,backendName:\"cpu\",kernelFunc:T9};var k9=kt(Es,r=>Math.sin(r)),SP={kernelName:Es,backendName:\"cpu\",kernelFunc:k9};var E9=kt(za,r=>Math.sinh(r)),vP={kernelName:za,backendName:\"cpu\",kernelFunc:E9};var _9=11920928955078125e-23,NP=Math.log(_9)+2,A9=kt(Va,r=>{let t=r>-NP,e=r<NP,n=Math.exp(r),o;return e?o=n:t?o=r:o=Math.log(1+n),o}),TP={kernelName:Va,backendName:\"cpu\",kernelFunc:A9};function $9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n;tt([o],\"spaceToBatchND\");let a=y.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let _=1+s.length;_<o.shape.length;++_)u.push([0,0]);let l=gw.kernelFunc({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),c=v.getReshaped(l.shape,s,a,!1),p=v.getPermuted(c.length,s.length,!1),m=v.getReshapedPermuted(l.shape,s,a,!1),h=Yt({inputs:{x:l},backend:e,attrs:{shape:c}}),b=Ve({inputs:{x:h},backend:e,attrs:{perm:p}}),N=Yt({inputs:{x:b},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(b),N}var kP={kernelName:xi,backendName:\"cpu\",kernelFunc:$9};function D9(r){let{inputs:t,backend:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:\n        ${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:\n        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Values must be a vector, saw:\n        ${o.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:\n        ${i.shape}`);let a=e.data.get(n.dataId).values,u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=e.data.get(i.dataId).values[0],[p,m,f,d,h]=iw(a,n.shape,n.dtype,u,o.dtype,l,c);return[e.makeTensorInfo(m,n.dtype,p),e.makeTensorInfo([m[0]],o.dtype,f),e.makeTensorInfo([d.length],\"bool\",new Uint8Array(d.map(g=>Number(g)))),e.makeTensorInfo([h.length],n.dtype,new Int32Array(h))]}var EP={kernelName:Pl,backendName:\"cpu\",kernelFunc:D9};function R9(r){let{inputs:t,backend:e}=r,{inputIndices:n,inputShape:o,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape\n        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape\n        ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(e.data.get(o.dataId).values),a=e.data.get(n.dataId).values,u=Array.from(e.data.get(s.dataId).values),[l,c,p]=aw(a,n.shape,n.dtype,i,u);return[e.makeTensorInfo(c,n.dtype,l),e.makeTensorInfo([p.length],s.dtype,new Int32Array(p))]}var _P={kernelName:Ga,backendName:\"cpu\",kernelFunc:R9};function F9(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape\n          ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape\n          ${s.shape}`);if(o.shape[0]!==s.shape[0])throw new Error(\"segmentIds and indices should have same size.\");let i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ld(i,n.shape,n.dtype,a,u,!0);return e.makeTensorInfo(c,n.dtype,l)}var AP={kernelName:Ll,backendName:\"cpu\",kernelFunc:F9};function O9(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape\n         ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape\n         ${s.shape}`);if(o.shape[0]!==s.shape[0])throw new Error(\"segmentIds and indices should have same size.\");let i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ld(i,n.shape,n.dtype,a,u);return e.makeTensorInfo(c,n.dtype,l)}var $P={kernelName:Ml,backendName:\"cpu\",kernelFunc:O9};function P9(r){let{inputs:t,backend:e,attrs:n}=r,{sparseIndices:o,sparseValues:s,defaultValue:i}=t,{outputShape:a}=n,{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=v.calculateShapes(s,o,a),f=!1,d=e.bufferSync(o),h;switch(s.dtype){case\"bool\":{let g=e.bufferSync(s),x=Boolean(e.data.get(i.dataId).values[0]);h=dl(d,g,a,m,c,l,u,p,x,f);break}case\"float32\":{let g=e.bufferSync(s),x=e.data.get(i.dataId).values[0];h=dl(d,g,a,m,c,l,u,p,x,f);break}case\"int32\":{let g=e.bufferSync(s),x=e.data.get(i.dataId).values[0];h=dl(d,g,a,m,c,l,u,p,x,f);break}case\"string\":{let g=e.bufferSync(s),x=y.decodeString(e.data.get(i.dataId).values[0]);h=dl(d,g,a,m,c,l,u,p,x,f);break}default:throw new Error(`Unsupported type ${s.dtype}`)}return e.makeTensorInfo(a,h.dtype,h.values)}var DP={kernelName:Lp,backendName:\"cpu\",kernelFunc:P9};function L9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=n,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=new Array(o.shape.length).fill(0),c=o.shape.slice();return u.map(p=>{let m=[...c];m[a]=p;let f=Fo({inputs:{x:o},backend:e,attrs:{begin:l,size:m}});return l[a]+=p,f})}var RP={kernelName:yi,backendName:\"cpu\",kernelFunc:L9};var FP={kernelName:zl,backendName:\"cpu\",kernelFunc:({inputs:r,backend:t})=>{let{x:e}=r,n=t;tt(e,\"square\");let o=n.data.get(e.dataId).values,s=new Float32Array(o.length);for(let a=0;a<o.length;++a){let u=o[a];s[a]=u*u}return{dataId:n.write(s,e.shape,e.dtype),shape:e.shape,dtype:e.dtype}}};var M9=kt(po,(r,t)=>{let e=t;return isNaN(r)?NaN:r>0?1:e.alpha}),OP={kernelName:po,backendName:\"cpu\",kernelFunc:M9};function z9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n;tt(o,\"stridedSlice\");let{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=Yt({inputs:{x:o},backend:e,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let _=Le.computeOutShape(b,w,C),A=Fo({inputs:{x:o},backend:e,attrs:{begin:b,size:_}});N=Yt({inputs:{x:A},backend:e,attrs:{shape:d}}),e.disposeIntermediateTensorInfo(A)}else{let _=e.bufferSync(o),A=lw(f,_,C,b);N=e.makeTensorInfo(d,A.dtype,A.values)}return N}var PP={kernelName:Wa,backendName:\"cpu\",kernelFunc:z9};function B9(r){let{inputs:t,backend:e,attrs:n}=r,{separator:o,nGramWidths:s,leftPad:i,rightPad:a,padWidth:u,preserveShortSequences:l}=n,{data:c,dataSplits:p}=t,m=e.data.get(c.dataId).values,f=e.data.get(p.dataId).values,[d,h]=Dc(m,f,o,s,i,a,u,l);return[e.makeTensorInfo([d.length],\"string\",d),e.makeTensorInfo(p.shape,\"int32\",h)]}var LP={kernelName:Bl,backendName:\"cpu\",kernelFunc:B9};function V9(r){let{inputs:t,backend:e,attrs:n}=r,{skipEmpty:o}=n,{input:s,delimiter:i}=t;if(s.dtype!==\"string\")throw new Error(\"Input must be of datatype string\");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let a=e.data.get(s.dataId).values,u=e.data.get(i.dataId).values[0],[l,c,p]=Rc(a,u,o),m=c.length;return[e.makeTensorInfo([m,2],\"int32\",l),e.makeTensorInfo([m],\"string\",c),e.makeTensorInfo([2],\"int32\",new Int32Array(p))]}var MP={kernelName:Vl,backendName:\"cpu\",kernelFunc:V9};function G9(r){let{inputs:t,backend:e,attrs:n}=r,{numBuckets:o}=n,{input:s}=t;if(s.dtype!==\"string\")throw new Error(\"Input must be of datatype string\");if(o<=0)throw new Error(\"Number of buckets must be at least 1\");let i=e.data.get(s.dataId).values,a=Fc(i,o);return e.makeTensorInfo(s.shape,\"int32\",a)}var zP={kernelName:Gl,backendName:\"cpu\",kernelFunc:G9};var W9=kt(Os,r=>Math.tan(r)),BP={kernelName:Os,backendName:\"cpu\",kernelFunc:W9};var U9=kt(Ps,r=>Math.tanh(r)),VP={kernelName:Ps,backendName:\"cpu\",kernelFunc:U9};function H9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reps:s}=n;tt(o,\"tile\");let i=uw(e.bufferSync(o),s);return e.makeTensorInfo(i.shape,i.dtype,i.values)}var GP={kernelName:Jn,backendName:\"cpu\",kernelFunc:H9};function q9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{k:s,sorted:i}=n;tt(o,\"topk\");let a=e.data.get(o.dataId).values,[u,l]=cw(a,o.shape,o.dtype,s,i);return[e.makeTensorInfo(u.shape,u.dtype,u.values),e.makeTensorInfo(l.shape,l.dtype,l.values)]}var WP={kernelName:Ua,backendName:\"cpu\",kernelFunc:q9};function K9(r){let{inputs:t,attrs:e,backend:n}=r,{image:o,transforms:s}=t,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=e,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=y.computeStrides(o.shape),b=x[0],w=x[1],C=x[2],N=y.computeStrides(g),_=N[0],A=N[1],$=N[2],F=y.getTypedArrayFromDType(o.dtype,y.sizeFromShape(g));F.fill(u);let P=n.data.get(o.dataId).values,V=n.data.get(s.dataId).values;for(let W=0;W<c;++W){let q=s.shape[0]===1?V:V.subarray(W*8,W*8+8);for(let H=0;H<d;++H)for(let j=0;j<h;++j)for(let Y=0;Y<f;++Y){let Z,et=q[6]*j+q[7]*H+1;if(et===0)continue;let rt=(q[0]*j+q[1]*H+q[2])/et,ot=(q[3]*j+q[4]*H+q[5])/et,at=UP(rt,m,a),nt=UP(ot,p,a);switch(i){case\"nearest\":Z=J9(P,p,m,b,w,C,W,nt,at,Y,u);break;case\"bilinear\":Z=Q9(P,p,m,b,w,C,W,nt,at,Y,u);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}let it=W*_+H*A+j*$+Y;F[it]=Z}return n.makeTensorInfo(g,o.dtype,F)}return{dataId:n.write(F,g,o.dtype),shape:o.shape,dtype:o.dtype}}var HP={kernelName:Ha,backendName:\"cpu\",kernelFunc:K9};function UP(r,t,e){switch(e){case\"reflect\":return j9(r,t);case\"wrap\":return X9(r,t);case\"nearest\":return Z9(r,t);case\"constant\":default:return Y9(r,t)}}function j9(r,t){let e=r;if(e<0)if(t<=1)e=0;else{let n=2*t;e<n&&(e=n*Math.trunc(-e/n)+e),e=e<-t?e+n:-e-1}else if(e>t-1)if(t<=1)e=0;else{let n=2*t;e-=n*Math.trunc(e/n),e>=t&&(e=n-e-1)}return y.clamp(0,e,t-1)}function X9(r,t){let e=r;if(e<0)if(t<=1)e=0;else{let n=t-1;e+=t*(Math.trunc(-e/n)+1)}else if(e>t-1)if(t<=1)e=0;else{let n=t-1;e-=t*Math.trunc(e/n)}return y.clamp(0,e,t-1)}function Y9(r,t){return r}function Z9(r,t){return y.clamp(0,r,t-1)}function Kh(r,t,e,n,o,s,i,a,u,l,c){let p=i*n+a*o+u*s+l;return 0<=a&&a<t&&0<=u&&u<e?r[p]:c}function J9(r,t,e,n,o,s,i,a,u,l,c){let p=Math.round(a),m=Math.round(u);return Kh(r,t,e,n,o,s,i,p,m,l,c)}function Q9(r,t,e,n,o,s,i,a,u,l,c){let p=Math.floor(a),m=Math.floor(u),f=p+1,d=m+1,h=(d-u)*Kh(r,t,e,n,o,s,i,p,m,l,c)+(u-m)*Kh(r,t,e,n,o,s,i,p,d,l,c),g=(d-u)*Kh(r,t,e,n,o,s,i,f,m,l,c)+(u-m)*Kh(r,t,e,n,o,s,i,f,d,l,c);return(f-a)*h+(a-p)*g}function ttt(r){let{inputs:t,attrs:e,backend:n}=r,{axis:o}=e,{x:s}=t;tt(s,\"unique\");let i=n.data.get(s.dataId).values,{outputValues:a,outputShape:u,indices:l}=pw(i,o,s.shape,s.dtype);return[n.makeTensorInfo(u,s.dtype,a),n.makeTensorInfo([l.length],\"int32\",l)]}var qP={kernelName:Mp,backendName:\"cpu\",kernelFunc:ttt};function ett(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o.shape.length,a=o.shape[s],u=new Array(i-1),l=0;for(let f=0;f<i;f++)f!==s&&(u[l++]=o.shape[f]);let c=new Array(i).fill(0),p=o.shape.slice();p[s]=1;let m=new Array(a);for(let f=0;f<m.length;f++){c[s]=f;let d=Fo({inputs:{x:o},backend:e,attrs:{begin:c,size:p}});m[f]=Yt({inputs:{x:d},backend:e,attrs:{shape:u}}),e.disposeIntermediateTensorInfo(d)}return m}var KP={kernelName:bi,backendName:\"cpu\",kernelFunc:ett};function rtt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,segmentIds:s}=t,{numSegments:i}=n;tt(o,\"unsortedSegmentSum\");let a=o.shape.length,u=s.shape.length,l=[],c=[],p=a-u,m=s;for(let d=0;d<p;++d){let h=cd({inputs:{input:m},backend:e,attrs:{dim:d+1}});m=h,c.push(h)}for(let d=0;d<i;++d){let h=y.createScalarValue(d,\"int32\"),g=e.makeTensorInfo([],\"int32\",h),x=GN({inputs:{a:g,b:m},backend:e}),b=$o({inputs:{x},backend:e,attrs:{dtype:\"float32\"}}),w=_c({inputs:{a:b,b:o},backend:e}),C=hl({inputs:{x:w},backend:e,attrs:{axis:0,keepDims:!1}});l.push(C),c.push(g),c.push(x),c.push(b),c.push(w),c.push(C)}let f=bT({inputs:l,backend:e,attrs:{axis:0}});return c.forEach(d=>e.disposeIntermediateTensorInfo(d)),f}var jP={kernelName:Wl,backendName:\"cpu\",kernelFunc:rtt};var ntt=[wF,$R,CF,IF,PR,SF,vF,NF,TF,kF,EF,_F,AF,$F,DF,FF,OF,PF,LF,bF,MF,zF,BF,VF,OR,LR,GF,DR,WF,HF,qF,KF,jF,XF,YF,ZF,JF,QF,tO,eO,rO,nO,oO,sO,iO,aO,lO,uO,cO,pO,fO,fF,dO,MR,hO,zR,gO,BR,xO,yO,bO,VR,wO,CO,IO,SO,vO,GR,WR,RR,NO,UF,TO,kO,EO,dF,UR,HR,_O,qR,AO,$O,DO,RO,FO,OO,PO,KR,LO,MO,zO,BO,GO,WO,UO,jR,HO,qO,XO,XR,YR,YO,ZO,JO,ZR,QO,rP,nP,gw,oP,hF,QR,sP,iP,aP,lP,FR,Uh,uP,gF,xF,yF,cP,pP,mP,fP,dP,hP,gP,oF,xP,bP,wP,CP,iF,IP,SP,vP,aF,KO,TP,kP,EP,_P,AP,$P,DP,RP,uF,FP,cF,OP,PP,LP,MP,zP,pF,mO,BP,VP,GP,WP,HP,JR,qP,KP,jP,tP];for(let r of ntt)Lu(r);var dd={};Wt(dd,{assertNotComplex:()=>Qs,bindCanvasToFramebuffer:()=>mtt,bindColorTextureToFramebuffer:()=>Zh,bindTextureToProgramUniformSampler:()=>OT,bindTextureUnit:()=>JP,bindVertexBufferToProgramAttribute:()=>Iw,callAndCheck:()=>yt,canBeRepresented:()=>ST,createFragmentShader:()=>NT,createFramebuffer:()=>DT,createProgram:()=>TT,createStaticIndexBuffer:()=>_T,createStaticVertexBuffer:()=>ET,createTexture:()=>AT,createVertexShader:()=>vT,getBatchDim:()=>xl,getExtensionOrThrow:()=>pd,getFramebufferErrorMessage:()=>QP,getMaxTexturesInShader:()=>MT,getNumChannels:()=>ctt,getProgramUniformLocation:()=>FT,getProgramUniformLocationOrThrow:()=>RT,getRowsCols:()=>yl,getShapeAs3D:()=>fd,getTextureShapeFromLogicalShape:()=>PT,getWebGLDisjointQueryTimerVersion:()=>zT,getWebGLErrorMessage:()=>ZP,getWebGLMaxTextureSize:()=>LT,hasExtension:()=>Wn,isCapableOfRenderingToFloatTexture:()=>BT,isDownloadFloatTextureEnabled:()=>VT,isReshapeFree:()=>Eu,isWebGLFenceEnabled:()=>GT,isWebGLVersionEnabled:()=>vw,linkProgram:()=>kT,logShaderSourceAndInfoLog:()=>Cw,resetMaxTextureSize:()=>ftt,resetMaxTexturesInShader:()=>dtt,unbindColorTextureFromFramebuffer:()=>Sw,unbindTextureUnit:()=>ptt,validateFramebuffer:()=>md,validateProgram:()=>Yh,validateTextureSize:()=>$T});var Pc={},xw={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function wT(r,t){Pc[r]=t}function Gn(r,t){if(!(r in Pc)||t!=null){let n=stt(r,t);if(n!==null)Pc[r]=n;else return console.log(\"Could not get context for WebGL version\",r),null}let e=Pc[r];return e==null||e.isContextLost()?(delete Pc[r],Gn(r)):(e.disable(e.DEPTH_TEST),e.disable(e.STENCIL_TEST),e.disable(e.BLEND),e.disable(e.DITHER),e.disable(e.POLYGON_OFFSET_FILL),e.disable(e.SAMPLE_COVERAGE),e.enable(e.SCISSOR_TEST),e.enable(e.CULL_FACE),e.cullFace(e.BACK),Pc[r])}function ott(r){if(typeof OffscreenCanvas!=\"undefined\"&&r===2)return new OffscreenCanvas(300,150);if(typeof document!=\"undefined\")return document.createElement(\"canvas\");throw new Error(\"Cannot create a canvas in this context\")}function stt(r,t){if(r!==1&&r!==2)throw new Error(\"Cannot get WebGL rendering context, WebGL is disabled.\");let e=t==null?ott(r):t;return e.addEventListener(\"webglcontextlost\",n=>{n.preventDefault(),delete Pc[r]},!1),z().getBool(\"SOFTWARE_WEBGL_ENABLED\")&&(xw.failIfMajorPerformanceCaveat=!1),r===1?e.getContext(\"webgl\",xw)||e.getContext(\"experimental-webgl\",xw):e.getContext(\"webgl2\",xw)}var ku;(function(r){r[r.DENSE=0]=\"DENSE\",r[r.SHARED_BATCH=1]=\"SHARED_BATCH\"})(ku||(ku={}));var jr;(function(r){r[r.RENDER=0]=\"RENDER\",r[r.UPLOAD=1]=\"UPLOAD\",r[r.PIXELS=2]=\"PIXELS\",r[r.DOWNLOAD=3]=\"DOWNLOAD\"})(jr||(jr={}));var Pr;(function(r){r[r.UNPACKED_FLOAT16=0]=\"UNPACKED_FLOAT16\",r[r.UNPACKED_FLOAT32=1]=\"UNPACKED_FLOAT32\",r[r.PACKED_4X1_UNSIGNED_BYTE=2]=\"PACKED_4X1_UNSIGNED_BYTE\",r[r.PACKED_2X2_FLOAT32=3]=\"PACKED_2X2_FLOAT32\",r[r.PACKED_2X2_FLOAT16=4]=\"PACKED_2X2_FLOAT16\"})(Pr||(Pr={}));function Lc(r,t){return[t,r]}function XP(r,t){return r*t}function jh(r){let t=y.sizeFromShape(r),e=Math.ceil(t/4);return y.sizeToSquarishShape(e)}function Xi(r,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(r/2))]}function YP(r,t){let[e,n]=Xi(r,t);return e*n*4}function Xh(r,t){let e=r,n,o,s,i,a,u,l,c,p,m;return z().getNumber(\"WEBGL_VERSION\")===2?(n=e.R32F,o=e.R16F,s=e.RGBA16F,i=e.RGBA32F,a=e.RED,l=4,c=1,p=e.HALF_FLOAT,m=e.FLOAT,u=e.RGBA8):(n=r.RGBA,o=r.RGBA,s=r.RGBA,i=e.RGBA,a=r.RGBA,l=4,c=4,p=t!=null?t.HALF_FLOAT_OES:null,m=r.FLOAT,u=r.RGBA),{internalFormatFloat:n,internalFormatHalfFloat:o,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:a,downloadTextureFormat:u,downloadUnpackNumChannels:l,defaultNumChannels:c,textureTypeHalfFloat:p,textureTypeFloat:m}}function yt(r,t){let e=t();return z().getBool(\"DEBUG\")&&itt(r),e}function itt(r){let t=r.getError();if(t!==r.NO_ERROR)throw new Error(\"WebGL Error: \"+ZP(r,t))}var att=596e-10,ltt=65504;function ST(r){return!!(z().getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\")||r===0||att<Math.abs(r)&&Math.abs(r)<ltt)}function ZP(r,t){switch(t){case r.NO_ERROR:return\"NO_ERROR\";case r.INVALID_ENUM:return\"INVALID_ENUM\";case r.INVALID_VALUE:return\"INVALID_VALUE\";case r.INVALID_OPERATION:return\"INVALID_OPERATION\";case r.INVALID_FRAMEBUFFER_OPERATION:return\"INVALID_FRAMEBUFFER_OPERATION\";case r.OUT_OF_MEMORY:return\"OUT_OF_MEMORY\";case r.CONTEXT_LOST_WEBGL:return\"CONTEXT_LOST_WEBGL\";default:return`Unknown error code ${t}`}}function pd(r,t){return gl(r,()=>r.getExtension(t),'Extension \"'+t+'\" not supported on this browser.')}function vT(r,t){let e=gl(r,()=>r.createShader(r.VERTEX_SHADER),\"Unable to create vertex WebGLShader.\");if(yt(r,()=>r.shaderSource(e,t)),yt(r,()=>r.compileShader(e)),r.getShaderParameter(e,r.COMPILE_STATUS)===!1)throw console.log(r.getShaderInfoLog(e)),new Error(\"Failed to compile vertex shader.\");return e}function NT(r,t){let e=gl(r,()=>r.createShader(r.FRAGMENT_SHADER),\"Unable to create fragment WebGLShader.\");if(yt(r,()=>r.shaderSource(e,t)),yt(r,()=>r.compileShader(e)),z().get(\"ENGINE_COMPILE_ONLY\"))return e;if(r.getShaderParameter(e,r.COMPILE_STATUS)===!1)throw Cw(t,r.getShaderInfoLog(e)),new Error(\"Failed to compile fragment shader.\");return e}var utt=/ERROR: [0-9]+:([0-9]+):/g;function Cw(r,t){let e=utt.exec(t);if(e==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(r);return}let n=+e[1],o=r.split(`\n`),s=o.length.toString().length+2,i=o.map((p,m)=>y.rightPad((m+1).toString(),s)+p),a=0;for(let p=0;p<i.length;p++)a=Math.max(i[p].length,a);let u=i.slice(0,n-1),l=i.slice(n-1,n),c=i.slice(n);console.log(u.join(`\n`)),console.log(t.split(`\n`)[0]),console.log(`%c ${y.rightPad(l[0],a)}`,\"border:1px solid red; background-color:#e3d2d2; color:#a61717\"),console.log(c.join(`\n`))}function TT(r){return gl(r,()=>r.createProgram(),\"Unable to create WebGLProgram.\")}function kT(r,t){if(yt(r,()=>r.linkProgram(t)),!z().get(\"ENGINE_COMPILE_ONLY\")&&r.getProgramParameter(t,r.LINK_STATUS)===!1)throw console.log(r.getProgramInfoLog(t)),new Error(\"Failed to link vertex and fragment shaders.\")}function Yh(r,t){if(yt(r,()=>r.validateProgram(t)),r.getProgramParameter(t,r.VALIDATE_STATUS)===!1)throw console.log(r.getProgramInfoLog(t)),new Error(\"Shader program validation failed.\")}function ET(r,t){let e=gl(r,()=>r.createBuffer(),\"Unable to create WebGLBuffer\");return yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,e)),yt(r,()=>r.bufferData(r.ARRAY_BUFFER,t,r.STATIC_DRAW)),e}function _T(r,t){let e=gl(r,()=>r.createBuffer(),\"Unable to create WebGLBuffer\");return yt(r,()=>r.bindBuffer(r.ELEMENT_ARRAY_BUFFER,e)),yt(r,()=>r.bufferData(r.ELEMENT_ARRAY_BUFFER,t,r.STATIC_DRAW)),e}function ctt(){return z().getNumber(\"WEBGL_VERSION\")===2?1:4}function AT(r){return gl(r,()=>r.createTexture(),\"Unable to create WebGLTexture.\")}function $T(r,t){let e=z().getNumber(\"WEBGL_MAX_TEXTURE_SIZE\");if(r<=0||t<=0){let n=`[${r}x${t}]`;throw new Error(\"Requested texture size \"+n+\" is invalid.\")}if(r>e||t>e){let n=`[${r}x${t}]`,o=`[${e}x${e}]`;throw new Error(\"Requested texture size \"+n+\" greater than WebGL maximum on this browser / GPU \"+o+\".\")}}function DT(r){return gl(r,()=>r.createFramebuffer(),\"Unable to create WebGLFramebuffer.\")}function Iw(r,t,e,n,o,s,i){let a=r.getAttribLocation(t,e);return a===-1?!1:(yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,n)),yt(r,()=>r.vertexAttribPointer(a,o,r.FLOAT,!1,s,i)),yt(r,()=>r.enableVertexAttribArray(a)),!0)}function JP(r,t,e){tL(r,e),yt(r,()=>r.activeTexture(r.TEXTURE0+e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,t))}function ptt(r,t){tL(r,t),yt(r,()=>r.activeTexture(r.TEXTURE0+t)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function RT(r,t,e){return gl(r,()=>r.getUniformLocation(t,e),'uniform \"'+e+'\" not present in program.')}function FT(r,t,e){return r.getUniformLocation(t,e)}function OT(r,t,e,n){yt(r,()=>JP(r,t,n)),yt(r,()=>r.uniform1i(e,n))}function mtt(r){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,null)),yt(r,()=>r.viewport(0,0,r.canvas.width,r.canvas.height)),yt(r,()=>r.scissor(0,0,r.canvas.width,r.canvas.height))}function Zh(r,t,e){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,e)),yt(r,()=>r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,t,0))}function Sw(r,t){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,t)),yt(r,()=>r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,null,0))}function md(r){let t=r.checkFramebufferStatus(r.FRAMEBUFFER);if(t!==r.FRAMEBUFFER_COMPLETE)throw new Error(\"Error binding framebuffer: \"+QP(r,t))}function QP(r,t){switch(t){case r.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return\"FRAMEBUFFER_INCOMPLETE_ATTACHMENT\";case r.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return\"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT\";case r.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return\"FRAMEBUFFER_INCOMPLETE_DIMENSIONS\";case r.FRAMEBUFFER_UNSUPPORTED:return\"FRAMEBUFFER_UNSUPPORTED\";default:return`unknown error ${t}`}}function gl(r,t,e){let n=yt(r,()=>t());if(n==null)throw new Error(e);return n}function tL(r,t){let e=r.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,n=t+r.TEXTURE0;if(n<r.TEXTURE0||n>e){let o=`[gl.TEXTURE0, gl.TEXTURE${e}]`;throw new Error(`textureUnit must be in ${o}.`)}}function xl(r,t=2){return y.sizeFromShape(r.slice(0,r.length-t))}function yl(r){if(r.length===0)throw Error(\"Cannot get rows and columns of an empty shape array.\");return[r.length>1?r[r.length-2]:1,r[r.length-1]]}function fd(r){let t=[1,1,1];return r.length===0||r.length===1&&r[0]===1||(t=[xl(r),...yl(r)]),t}function PT(r,t=!1){let e=z().getNumber(\"WEBGL_MAX_TEXTURE_SIZE\"),n=z().getNumber(\"WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE\");n===1/0&&z().getBool(\"WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE\")&&(n=e/2),t&&(e=e*2,n=n*2,r=r.map((a,u)=>u>=r.length-2?y.nearestLargerEven(r[u]):r[u]),r.length===1&&(r=[2,r[0]])),r.length!==2&&(r=y.squeezeShape(r).newShape);let o=y.sizeFromShape(r),s=null;r.length<=1&&o<=e?s=[1,o]:r.length===2&&r[0]<=e&&r[1]<=e?s=r:r.length===3&&r[0]*r[1]<=e&&r[2]<=e?s=[r[0]*r[1],r[2]]:r.length===3&&r[0]<=e&&r[1]*r[2]<=e?s=[r[0],r[1]*r[2]]:r.length===4&&r[0]*r[1]*r[2]<=e&&r[3]<=e?s=[r[0]*r[1]*r[2],r[3]]:r.length===4&&r[0]<=e&&r[1]*r[2]*r[3]<=e&&(s=[r[0],r[1]*r[2]*r[3]]);let i=s!=null&&Math.max(...s)>n&&Math.min(...s)<=(t?2:1)&&Math.min(...s)>0;if(s==null||i)if(t){let a=xl(r),u=2,l=2;r.length&&([u,l]=yl(r)),o=a*(u/2)*(l/2),s=y.sizeToSquarishShape(o).map(c=>c*2)}else s=y.sizeToSquarishShape(o);return s}function yw(r){return r%2===0}function Eu(r,t){if(r=r.slice(-2),t=t.slice(-2),y.arraysEqual(r,t)||!r.length||!t.length||r[0]===0||r[1]===0||t[0]===0||t[1]===0)return!0;if(r.length!==t.length){let e=r.slice(-1)[0],n=t.slice(-1)[0];if(e===n||yw(e)&&yw(n)&&(r[0]===1||t[0]===1))return!0}return r[1]===t[1]&&yw(r[0])&&yw(t[0])}var bw,ww;function LT(r){if(bw==null){let t=Gn(r);bw=t.getParameter(t.MAX_TEXTURE_SIZE)}return bw}function ftt(){bw=null}function dtt(){ww=null}function MT(r){if(ww==null){let t=Gn(r);ww=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,ww)}function zT(r){if(r===0)return 0;let t,e=Gn(r);return Wn(e,\"EXT_disjoint_timer_query_webgl2\")&&r===2?t=2:Wn(e,\"EXT_disjoint_timer_query\")?t=1:t=0,t}function Wn(r,t){return r.getExtension(t)!=null}function vw(r){try{if(Gn(r)!=null)return!0}catch(t){return console.log(\"Error when getting WebGL context: \",t),!1}return!1}function BT(r){if(r===0)return!1;let t=Gn(r);if(r===1){if(!Wn(t,\"OES_texture_float\"))return!1}else if(!Wn(t,\"EXT_color_buffer_float\"))return!1;return IT(t)}function VT(r){if(r===0)return!1;let t=Gn(r);if(r===1){if(!Wn(t,\"OES_texture_float\")||!Wn(t,\"WEBGL_color_buffer_float\"))return!1}else{if(Wn(t,\"EXT_color_buffer_float\"))return IT(t);let n=\"EXT_color_buffer_half_float\";if(Wn(t,n)){let o=t.getExtension(n);return htt(t,o)}return!1}return IT(t)}function IT(r){let t=Xh(r),e=r.createTexture();r.bindTexture(r.TEXTURE_2D,e);let n=1,o=1;r.texImage2D(r.TEXTURE_2D,0,t.internalFormatFloat,n,o,0,t.textureFormatFloat,t.textureTypeFloat,null);let s=r.createFramebuffer();r.bindFramebuffer(r.FRAMEBUFFER,s),r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,e,0);let i=r.checkFramebufferStatus(r.FRAMEBUFFER)===r.FRAMEBUFFER_COMPLETE;return r.bindTexture(r.TEXTURE_2D,null),r.bindFramebuffer(r.FRAMEBUFFER,null),r.deleteTexture(e),r.deleteFramebuffer(s),i}function htt(r,t){let e=Xh(r,t),n=r.createTexture();r.bindTexture(r.TEXTURE_2D,n);let o=1,s=1;r.texImage2D(r.TEXTURE_2D,0,e.internalFormatHalfFloat,o,s,0,e.textureFormatFloat,e.textureTypeHalfFloat,null);let i=r.createFramebuffer();r.bindFramebuffer(r.FRAMEBUFFER,i),r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,n,0);let a=r.checkFramebufferStatus(r.FRAMEBUFFER)===r.FRAMEBUFFER_COMPLETE;return r.bindTexture(r.TEXTURE_2D,null),r.bindFramebuffer(r.FRAMEBUFFER,null),r.deleteTexture(n),r.deleteFramebuffer(i),a}function GT(r){return r!==2?!1:Gn(r).fenceSync!=null}function Qs(r,t){Array.isArray(r)||(r=[r]),r.forEach(e=>{e!=null&&y.assert(e.dtype!==\"complex64\",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var Tt=z();Tt.registerFlag(\"HAS_WEBGL\",()=>Tt.getNumber(\"WEBGL_VERSION\")>0);Tt.registerFlag(\"WEBGL_VERSION\",()=>vw(2)?2:vw(1)?1:0);Tt.registerFlag(\"WEBGL_CHECK_NUMERICAL_PROBLEMS\",()=>!1);Tt.registerFlag(\"WEBGL_BUFFER_SUPPORTED\",()=>Tt.get(\"WEBGL_VERSION\")===2);Tt.registerFlag(\"WEBGL_CPU_FORWARD\",()=>!0);Tt.registerFlag(\"WEBGL_FORCE_F16_TEXTURES\",()=>!1);Tt.registerFlag(\"WEBGL_PACK\",()=>Tt.getBool(\"HAS_WEBGL\"));Tt.registerFlag(\"WEBGL_PACK_NORMALIZATION\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_CLIP\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_DEPTHWISECONV\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_BINARY_OPERATIONS\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_UNARY_OPERATIONS\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_ARRAY_OPERATIONS\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_IMAGE_OPERATIONS\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_PACK_REDUCE\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_LAZILY_UNPACK\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_CONV_IM2COL\",()=>Tt.getBool(\"WEBGL_PACK\"));Tt.registerFlag(\"WEBGL_MAX_TEXTURE_SIZE\",()=>LT(Tt.getNumber(\"WEBGL_VERSION\")));Tt.registerFlag(\"WEBGL_MAX_TEXTURES_IN_SHADER\",()=>MT(Tt.getNumber(\"WEBGL_VERSION\")));Tt.registerFlag(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\",()=>{let r=Tt.getNumber(\"WEBGL_VERSION\");return r===0?0:zT(r)});Tt.registerFlag(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\",()=>Tt.getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")>0&&!Kl.isMobile());Tt.registerFlag(\"WEBGL_RENDER_FLOAT32_CAPABLE\",()=>BT(Tt.getNumber(\"WEBGL_VERSION\")));Tt.registerFlag(\"WEBGL_RENDER_FLOAT32_ENABLED\",()=>Tt.getBool(\"WEBGL_FORCE_F16_TEXTURES\")?!1:Tt.getBool(\"WEBGL_RENDER_FLOAT32_CAPABLE\"));Tt.registerFlag(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\",()=>VT(Tt.getNumber(\"WEBGL_VERSION\")));Tt.registerFlag(\"WEBGL_FENCE_API_ENABLED\",()=>GT(Tt.getNumber(\"WEBGL_VERSION\")));Tt.registerFlag(\"WEBGL_SIZE_UPLOAD_UNIFORM\",()=>Tt.getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\")?4:0);Tt.registerFlag(\"WEBGL_DELETE_TEXTURE_THRESHOLD\",()=>-1,r=>{if(r<0&&r!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${r}.`)});Tt.registerFlag(\"WEBGL_FLUSH_THRESHOLD\",()=>Kl.isMobile()?1:-1,r=>{if(r<0&&r!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${r}.`)});Tt.registerFlag(\"CPU_HANDOFF_SIZE_THRESHOLD\",()=>128);Tt.registerFlag(\"WEBGL_USE_SHAPES_UNIFORMS\",()=>!1);Tt.registerFlag(\"TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD\",()=>1e5);Tt.registerFlag(\"TOPK_K_CPU_HANDOFF_THRESHOLD\",()=>128);Tt.registerFlag(\"WEBGL_EXP_CONV\",()=>!1);Tt.registerFlag(\"SOFTWARE_WEBGL_ENABLED\",()=>Tt.getBool(\"IS_TEST\"));Tt.registerFlag(\"WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE\",()=>1/0);Tt.registerFlag(\"WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE\",()=>!1);Tt.registerFlag(\"WEBGL2_ISNAN_CUSTOM\",()=>!1);function Ge(){let r,t,e,n,o,s,i,a,u,l;return z().getNumber(\"WEBGL_VERSION\")===2?(r=\"#version 300 es\",t=\"in\",e=\"out\",n=\"in\",o=\"texture\",s=\"outputColor\",i=\"out vec4 outputColor;\",a=z().getBool(\"WEBGL2_ISNAN_CUSTOM\")?`\n      bool isnan_custom(float val) {\n        uint floatToUint = floatBitsToUint(val);\n        return (floatToUint & 0x7fffffffu) > 0x7f800000u;\n      }\n\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan_custom(val.x),\n          isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));\n      }\n\n      #define isnan(value) isnan_custom(value)\n    `:\"\",u=\"\",l=`\n      #define round(value) newRound(value)\n      int newRound(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 newRound(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    `):(r=\"\",t=\"attribute\",e=\"varying\",n=\"varying\",o=\"texture2D\",s=\"gl_FragColor\",i=\"\",a=`\n      #define isnan(value) isnan_custom(value)\n      bool isnan_custom(float val) {\n        return (val > 0. || val < 1. || val == 0.) ? false : true;\n      }\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));\n      }\n    `,u=`\n      uniform float INFINITY;\n\n      bool isinf(float val) {\n        return abs(val) == INFINITY;\n      }\n      bvec4 isinf(vec4 val) {\n        return equal(abs(val), vec4(INFINITY));\n      }\n    `,l=`\n      int round(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 round(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    `),{version:r,attribute:t,varyingVs:e,varyingFs:n,texture2D:o,output:s,defineOutput:i,defineSpecialNaN:a,defineSpecialInf:u,defineRound:l}}function ti(r,t,e=\"index\"){let n=y.computeStrides(t);return n.map((o,s)=>{let i=`int ${r[s]} = ${e} / ${o}`,a=s===n.length-1?`int ${r[s+1]} = ${e} - ${r[s]} * ${o}`:`index -= ${r[s]} * ${o}`;return`${i}; ${a};`}).join(\"\")}function Mc(r,t,e=\"index\"){let n=y.computeStrides(t);return n.map((o,s)=>{let i=`int ${r[s]} = ${e} / outShapeStrides[${s}]`,a=s===n.length-1?`int ${r[s+1]} = ${e} - ${r[s]} * outShapeStrides[${s}]`:`index -= ${r[s]} * outShapeStrides[${s}]`;return`${i}; ${a};`}).join(\"\")}function gtt(r,t){let e=r.length,n=r.map(s=>`${t}[${s}]`),o=new Array(e-1);o[e-2]=n[e-1];for(let s=e-3;s>=0;--s)o[s]=`(${o[s+1]} * ${n[s+1]})`;return o}function eL(r,t,e=\"index\"){let n=r.map((s,i)=>i),o=gtt(n,t);return o.map((s,i)=>{let a=`int ${r[i]} = ${e} / ${o[i]}`,u=i===o.length-1?`int ${r[i+1]} = ${e} - ${r[i]} * ${o[i]}`:`index -= ${r[i]} * ${o[i]}`;return`${a}; ${u};`}).join(\"\")}function hd(r){let t=y.computeStrides(r).map(e=>e.toString());return`\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;\n  }\n`}function gd(){return`\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n`}var Nw=`\n  const float FLOAT_MAX = 1.70141184e38;\n  const float FLOAT_MIN = 1.17549435e-38;\n\n  lowp vec4 encode_float(highp float v) {\n    if (isnan(v)) {\n      return vec4(255, 255, 255, 255);\n    }\n\n    highp float av = abs(v);\n\n    if(av < FLOAT_MIN) {\n      return vec4(0.0, 0.0, 0.0, 0.0);\n    } else if(v > FLOAT_MAX) {\n      return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;\n    } else if(v < -FLOAT_MAX) {\n      return vec4(0.0, 0.0,  128.0, 255.0) / 255.0;\n    }\n\n    highp vec4 c = vec4(0,0,0,0);\n\n    highp float e = floor(log2(av));\n    highp float m = exp2(fract(log2(av))) - 1.0;\n\n    c[2] = floor(128.0 * m);\n    m -= c[2] / 128.0;\n    c[1] = floor(32768.0 * m);\n    m -= c[1] / 32768.0;\n    c[0] = floor(8388608.0 * m);\n\n    highp float ebias = e + 127.0;\n    c[3] = floor(ebias / 2.0);\n    ebias -= c[3] * 2.0;\n    c[2] += floor(ebias) * 128.0;\n\n    c[3] += 128.0 * step(0.0, -v);\n\n    return c / 255.0;\n  }\n`;var{getBroadcastDims:rL}=v;function nL(r,t,e){let n=[];if(r.forEach(f=>{let d=y.sizeFromShape(f.shapeInfo.logicalShape);if(f.shapeInfo.isUniform?n.push(`uniform float ${f.name}${d>1?`[${d}]`:\"\"};`):(n.push(`uniform sampler2D ${f.name};`),n.push(`uniform int offset${f.name};`)),e.enableShapeUniforms){let{uniformShape:h}=Tw(e.packedInputs,f.shapeInfo.logicalShape,f.shapeInfo.texShape);switch(h.length){case 1:n.push(`uniform int ${f.name}Shape;`);break;case 2:n.push(`uniform ivec2 ${f.name}Shape;`);break;case 3:n.push(`uniform ivec3 ${f.name}Shape;`);break;case 4:n.push(`uniform ivec4 ${f.name}Shape;`);break;default:break}n.push(`uniform ivec2 ${f.name}TexShape;`)}}),e.enableShapeUniforms){switch(t.logicalShape.length){case 1:n.push(\"uniform int outShape;\");break;case 2:n.push(\"uniform ivec2 outShape;\"),n.push(\"uniform int outShapeStrides;\");break;case 3:n.push(\"uniform ivec3 outShape;\"),n.push(\"uniform ivec2 outShapeStrides;\");break;case 4:n.push(\"uniform ivec4 outShape;\"),n.push(\"uniform ivec3 outShapeStrides;\");break;default:break}n.push(\"uniform ivec2 outTexShape;\")}e.customUniforms&&e.customUniforms.forEach(f=>{n.push(`uniform ${f.type} ${f.name}${f.arrayIndex?`[${f.arrayIndex}]`:\"\"};`)});let o=n.join(`\n`),s=r.map(f=>xtt(f,t,e.packedInputs,e.enableShapeUniforms)).join(`\n`),i=t.texShape,a=Ge(),u=wtt(a),l,c,p=Stt(a);return t.isPacked?(l=ytt(t.logicalShape,i,e.enableShapeUniforms),c=Itt(a)):(l=btt(t.logicalShape,i,e.enableShapeUniforms),c=Ctt(a)),e.packedInputs&&(p+=ktt),[p,u,c,o,l,s,e.userCode].join(`\n`)}function yd(r,t=!1){let e=r.shapeInfo.logicalShape;switch(e.length){case 0:return ztt(r,t);case 1:return Vtt(r,t);case 2:return Wtt(r,t);case 3:return Htt(r,t);case 4:return Ktt(r,t);case 5:return jtt(r);case 6:return Xtt(r);default:throw new Error(`${e.length}-D input sampling is not yet supported`)}}function oL(r,t){switch(r.shapeInfo.logicalShape.length){case 0:return Mtt(r);case 1:return Btt(r,t);case 2:return Gtt(r,t);case 3:return Utt(r,t);default:return qtt(r,t)}}function xtt(r,t,e=!1,n){let o=\"\";e?o+=oL(r,n):o+=yd(r,n);let s=r.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(e?o+=Ytt(r,t):o+=Ztt(r,t)),o}function ytt(r,t,e){switch(r.length){case 0:return sL();case 1:return Ett(r,t,e);case 2:return Ptt(r,t,e);case 3:return Att(r,t,e);default:return Dtt(r,t,e)}}function btt(r,t,e){switch(r.length){case 0:return sL();case 1:return _tt(r,t,e);case 2:return Ltt(r,t,e);case 3:return $tt(r,t,e);case 4:return Rtt(r,t,e);case 5:return Ftt(r,t);case 6:return Ott(r,t);default:throw new Error(`${r.length}-D output sampling is not yet supported`)}}function wtt(r){return`\n    float sampleTexture(sampler2D textureSampler, vec2 uv) {\n      return ${r.texture2D}(textureSampler, uv).r;\n    }\n  `}function Ctt(r){return`\n    void setOutput(float val) {\n      ${r.output} = vec4(val, 0, 0, 0);\n    }\n  `}function Itt(r){return`\n    void setOutput(vec4 val) {\n      ${r.output} = val;\n    }\n  `}function Stt(r){return`${r.version}\n    precision highp float;\n    precision highp int;\n    precision highp sampler2D;\n    ${r.varyingFs} vec2 resultUV;\n    ${r.defineOutput}\n    const vec2 halfCR = vec2(0.5, 0.5);\n\n    struct ivec5\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n    };\n\n    struct ivec6\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n      int v;\n    };\n\n    uniform float NAN;\n    ${r.defineSpecialNaN}\n    ${r.defineSpecialInf}\n    ${r.defineRound}\n\n    int imod(int x, int y) {\n      return x - y * (x / y);\n    }\n\n    int idiv(int a, int b, float sign) {\n      int res = a / b;\n      int mod = imod(a, b);\n      if (sign < 0. && mod != 0) {\n        res -= 1;\n      }\n      return res;\n    }\n\n    //Based on the work of Dave Hoskins\n    //https://www.shadertoy.com/view/4djSRW\n    #define HASHSCALE1 443.8975\n    float random(float seed){\n      vec2 p = resultUV * seed;\n      vec3 p3  = fract(vec3(p.xyx) * HASHSCALE1);\n      p3 += dot(p3, p3.yzx + 19.19);\n      return fract((p3.x + p3.y) * p3.z);\n    }\n\n    ${vtt}\n    ${Ntt}\n    ${Ttt}\n  `}var vtt=`\nvec2 uvFromFlat(int texNumR, int texNumC, int index) {\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\nvec2 packedUVfrom1D(int texNumR, int texNumC, int index) {\n  int texelIndex = index / 2;\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`,Ntt=`\nvec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,\n  int texNumC, int row, int col) {\n  int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`,Ttt=`\nvec2 packedUVfrom3D(int texNumR, int texNumC,\n    int texelsInBatch, int texelsInLogicalRow, int b,\n    int row, int col) {\n  int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`,ktt=`\n  float getChannel(vec4 frag, vec2 innerDims) {\n    vec2 modCoord = mod(innerDims, 2.);\n    return modCoord.x == 0. ?\n      (modCoord.y == 0. ? frag.r : frag.g) :\n      (modCoord.y == 0. ? frag.b : frag.a);\n  }\n  float getChannel(vec4 frag, int dim) {\n    float modCoord = mod(float(dim), 2.);\n    return modCoord == 0. ? frag.r : frag.g;\n  }\n`;function sL(){return`\n    int getOutputCoords() {\n      return 0;\n    }\n  `}function Ett(r,t,e){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return n[0]===1?e?`\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));\n      }\n    `:`\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ${n[1]}.0);\n      }\n    `:n[1]===1?e?`\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));\n      }\n    `:`\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ${n[0]}.0);\n      }\n    `:e?`\n    int getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);\n    }\n  `:`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${n[0]}, ${n[1]}));\n      return 2 * (resTexRC.x * ${n[1]} + resTexRC.y);\n    }\n  `}function _tt(r,t,e){return t[0]===1?e?`\n      int getOutputCoords() {\n        return int(resultUV.x * float(outTexShape[1]));\n      }\n    `:`\n      int getOutputCoords() {\n        return int(resultUV.x * ${t[1]}.0);\n      }\n    `:t[1]===1?e?`\n      int getOutputCoords() {\n        return int(resultUV.y * float(outTexShape[0]));\n      }\n    `:`\n      int getOutputCoords() {\n        return int(resultUV.y * ${t[0]}.0);\n      }\n    `:e?`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      return resTexRC.x * outTexShape[1] + resTexRC.y;\n    }\n  `:`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      return resTexRC.x * ${t[1]} + resTexRC.y;\n    }\n  `}function Att(r,t,e){if(e)return`\n    ivec3 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],o=Math.ceil(r[2]/2),s=o*Math.ceil(r[1]/2);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${n[0]}, ${n[1]}));\n      int index = resTexRC.x * ${n[1]} + resTexRC.y;\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${o});\n      int c = imod(index, ${o}) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `}function $tt(r,t,e){if(e)return`\n  ivec3 getOutputCoords() {\n    ivec2 resTexRC = ivec2(resultUV.yx *\n                           vec2(outTexShape[0], outTexShape[1]));\n    int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n    ${Mc([\"r\",\"c\",\"d\"],r)}\n    return ivec3(r, c, d);\n  }\n`;let n=ti([\"r\",\"c\",\"d\"],r);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${n}\n      return ivec3(r, c, d);\n    }\n  `}function Dtt(r,t,e){if(e)return`\n    ivec4 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatchN = texelsInBatch * outShape[1];\n\n      int b2 = index / texelsInBatchN;\n      index -= b2 * texelsInBatchN;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec4(b2, b, r, c);\n    }\n  `;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],o=Math.ceil(r[r.length-1]/2),s=o*Math.ceil(r[r.length-2]/2),i=s,a=\"\",u=\"b, r, c\";for(let l=2;l<r.length-1;l++)i*=r[r.length-l-1],a=`\n      int b${l} = index / ${i};\n      index -= b${l} * ${i};\n    `+a,u=`b${l}, `+u;return`\n    ivec${r.length} getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${n[0]}, ${n[1]}));\n      int index = resTexRC.x * ${n[1]} + resTexRC.y;\n\n      ${a}\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${o});\n      int c = imod(index, ${o}) * 2;\n\n      return ivec${r.length}(${u});\n    }\n  `}function Rtt(r,t,e){if(e)return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      ${Mc([\"r\",\"c\",\"d\",\"d2\"],r)}\n      return ivec4(r, c, d, d2);\n    }\n  `;let n=ti([\"r\",\"c\",\"d\",\"d2\"],r);return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${n}\n      return ivec4(r, c, d, d2);\n    }\n  `}function Ftt(r,t){let e=ti([\"r\",\"c\",\"d\",\"d2\",\"d3\"],r);return`\n    ivec5 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},\n                             ${t[1]}));\n\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${e}\n\n      ivec5 outShape = ivec5(r, c, d, d2, d3);\n      return outShape;\n    }\n  `}function Ott(r,t){let e=ti([\"r\",\"c\",\"d\",\"d2\",\"d3\",\"d4\"],r);return`\n    ivec6 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${e}\n\n      ivec6 result = ivec6(r, c, d, d2, d3, d4);\n      return result;\n    }\n  `}function Ptt(r,t,e){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(y.arraysEqual(r,t))return e?`\n      ivec2 getOutputCoords() {\n        ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n        return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));\n      }\n    `:`\n      ivec2 getOutputCoords() {\n        return 2 * ivec2(resultUV.yx * vec2(${n[0]}, ${n[1]}));\n      }\n    `;let o=Math.ceil(r[1]/2);return e?`\n    ivec2 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec2(r, c);\n    }\n  `:`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${n[0]}, ${n[1]}));\n\n      int index = resTexRC.x * ${n[1]} + resTexRC.y;\n      int r = 2 * (index / ${o});\n      int c = imod(index, ${o}) * 2;\n\n      return ivec2(r, c);\n    }\n  `}function Ltt(r,t,e){return y.arraysEqual(r,t)?e?`\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));\n      }\n    `:`\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));\n      }\n    `:r[1]===1?e?`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `:`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `:r[0]===1?e?`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `:`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `:e?`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      int r = index / outShape[1];\n      int c = index - r * outShape[1];\n      return ivec2(r, c);\n    }\n  `:`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      int r = index / ${r[1]};\n      int c = index - r * ${r[1]};\n      return ivec2(r, c);\n    }\n  `}function zc(r){return`offset${r}`}function Mtt(r){let t=r.name,e=\"get\"+t.charAt(0).toUpperCase()+t.slice(1),n=Ge();return`\n    vec4 ${e}() {\n      return ${n.texture2D}(${t}, halfCR);\n    }\n  `}function ztt(r,t){let e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1);if(r.shapeInfo.isUniform)return`float ${n}() {return ${e};}`;let[o,s]=r.shapeInfo.texShape;if(o===1&&s===1)return`\n      float ${n}() {\n        return sampleTexture(${e}, halfCR);\n      }\n    `;let i=zc(e);if(t)return`\n    float ${n}() {\n      vec2 uv = uvFromFlat(${e}TexShape[0], ${e}TexShape[1], ${i});\n      return sampleTexture(${e}, uv);\n    }\n  `;let[a,u]=r.shapeInfo.texShape;return`\n    float ${n}() {\n      vec2 uv = uvFromFlat(${a}, ${u}, ${i});\n      return sampleTexture(${e}, uv);\n    }\n  `}function Btt(r,t){let e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1),o=r.shapeInfo.texShape,s=Ge();if(t)return`\n    vec4 ${n}(int index) {\n      ivec2 packedTexShape = ivec2(ceil(float(${e}TexShape[0]) / 2.0), ceil(float(${e}TexShape[1]) / 2.0));\n      vec2 uv = packedUVfrom1D(\n        packedTexShape[0], packedTexShape[1], index);\n      return ${s.texture2D}(${e}, uv);\n    }\n  `;let i=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)];return`\n    vec4 ${n}(int index) {\n      vec2 uv = packedUVfrom1D(\n        ${i[0]}, ${i[1]}, index);\n      return ${s.texture2D}(${e}, uv);\n    }\n  `}function Vtt(r,t){let e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1);if(r.shapeInfo.isUniform)return`\n      float ${n}(int index) {\n        ${bd(r)}\n      }\n    `;let o=r.shapeInfo.texShape,s=o[0],i=o[1];if(i===1&&s===1)return`\n      float ${n}(int index) {\n        return sampleTexture(${e}, halfCR);\n      }\n    `;let a=zc(e);return i===1?t?`\n      float ${n}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${a}) + 0.5) / float(${e}TexShape[0]));\n        return sampleTexture(${e}, uv);\n      }\n    `:`\n      float ${n}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${a}) + 0.5) / ${s}.0);\n        return sampleTexture(${e}, uv);\n      }\n    `:s===1?t?`\n      float ${n}(int index) {\n        vec2 uv = vec2((float(index + ${a}) + 0.5) / float(${e}TexShape[1]), 0.5);\n        return sampleTexture(${e}, uv);\n      }\n    `:`\n      float ${n}(int index) {\n        vec2 uv = vec2((float(index + ${a}) + 0.5) / ${i}.0, 0.5);\n        return sampleTexture(${e}, uv);\n      }\n    `:t?`\n    float ${n}(int index) {\n      vec2 uv = uvFromFlat(${e}TexShape[0], ${e}TexShape[1], index + ${a});\n      return sampleTexture(${e}, uv);\n    }\n  `:`\n    float ${n}(int index) {\n      vec2 uv = uvFromFlat(${s}, ${i}, index + ${a});\n      return sampleTexture(${e}, uv);\n    }\n  `}function Gtt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o=\"get\"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape,i=s[0],a=s[1],u=Ge();if(s!=null&&y.arraysEqual(e,s))return t?`\n      vec4 ${o}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);\n\n        return ${u.texture2D}(${n}, uv);\n      }\n    `:`\n      vec4 ${o}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}.0, ${i}.0);\n\n        return ${u.texture2D}(${n}, uv);\n      }\n    `;if(t)return`\n    vec4 ${o}(int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${n}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);\n      return ${u.texture2D}(${n}, uv);\n    }\n  `;let l=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],c=Math.ceil(e[1]/2);return`\n    vec4 ${o}(int row, int col) {\n      vec2 uv = packedUVfrom2D(${c}, ${l[0]}, ${l[1]}, row, col);\n      return ${u.texture2D}(${n}, uv);\n    }\n  `}function Wtt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o=\"get\"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape;if(s!=null&&y.arraysEqual(e,s)){if(t)return`\n      float ${o}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);\n        return sampleTexture(${n}, uv);\n      }\n    `;let m=s[0],f=s[1];return`\n    float ${o}(int row, int col) {\n      vec2 uv = (vec2(col, row) + halfCR) / vec2(${f}.0, ${m}.0);\n      return sampleTexture(${n}, uv);\n    }\n  `}let{newShape:i,keptDims:a}=y.squeezeShape(e),u=i;if(u.length<e.length){let m=wd(r,u),f=[\"row\",\"col\"];return`\n      ${yd(m,t)}\n      float ${o}(int row, int col) {\n        return ${o}(${Cd(f,a)});\n      }\n    `}if(r.shapeInfo.isUniform)return`\n      float ${o}(int row, int col) {\n        int index = round(dot(vec2(row, col), vec2(${e[1]}, 1)));\n        ${bd(r)}\n      }\n    `;let l=s[0],c=s[1],p=zc(n);return c===1?t?`\n      float ${o}(int row, int col) {\n        float index = dot(vec3(row, col, ${p}), vec3(${n}Shape[1], 1, 1));\n        vec2 uv = vec2(0.5, (index + 0.5) / float(${n}TexShape[0]));\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n    float ${o}(int row, int col) {\n      float index = dot(vec3(row, col, ${p}), vec3(${e[1]}, 1, 1));\n      vec2 uv = vec2(0.5, (index + 0.5) / ${l}.0);\n      return sampleTexture(${n}, uv);\n    }\n  `:l===1?t?`\n      float ${o}(int row, int col) {\n        float index = dot(vec3(row, col, ${p}), vec3(${n}Shape[1], 1, 1));\n        vec2 uv = vec2((index + 0.5) / float(${n}TexShape[1]), 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n    float ${o}(int row, int col) {\n      float index = dot(vec3(row, col, ${p}), vec3(${e[1]}, 1, 1));\n      vec2 uv = vec2((index + 0.5) / ${c}.0, 0.5);\n      return sampleTexture(${n}, uv);\n    }\n  `:t?`\n      float ${o}(int row, int col) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${n}Shape[1] + col + ${p};\n        vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n  float ${o}(int row, int col) {\n    // Explicitly use integer operations as dot() only works on floats.\n    int index = row * ${e[1]} + col + ${p};\n    vec2 uv = uvFromFlat(${l}, ${c}, index);\n    return sampleTexture(${n}, uv);\n  }\n`}function Utt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o=\"get\"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(e[0]===1){let m=e.slice(1),f=[1,2],d=wd(r,m),h=[\"b\",\"row\",\"col\"];return`\n        ${oL(d,t)}\n        vec4 ${o}(int b, int row, int col) {\n          return ${o}(${Cd(h,f)});\n        }\n      `}let a=Ge();if(t)return`\n    vec4 ${o}(int b, int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${n}Shape[2]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom3D(\n        packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);\n      return ${a.texture2D}(${n}, uv);\n    }\n  `;let u=i[0],l=i[1],c=Math.ceil(e[2]/2),p=c*Math.ceil(e[1]/2);return`\n    vec4 ${o}(int b, int row, int col) {\n      vec2 uv = packedUVfrom3D(\n        ${u}, ${l}, ${p}, ${c}, b, row, col);\n      return ${a.texture2D}(${n}, uv);\n    }\n  `}function Htt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o=\"get\"+n.charAt(0).toUpperCase()+n.slice(1),s=e[1]*e[2],i=e[2],{newShape:a,keptDims:u}=y.squeezeShape(e),l=a;if(l.length<e.length){let h=wd(r,l),g=[\"row\",\"col\",\"depth\"];return`\n        ${yd(h,t)}\n        float ${o}(int row, int col, int depth) {\n          return ${o}(${Cd(g,u)});\n        }\n      `}if(r.shapeInfo.isUniform)return`\n      float ${o}(int row, int col, int depth) {\n        int index = round(dot(vec3(row, col, depth),\n                          vec3(${s}, ${i}, 1)));\n        ${bd(r)}\n      }\n    `;let c=r.shapeInfo.texShape,p=c[0],m=c[1],f=r.shapeInfo.flatOffset;if(m===s&&f==null)return t?`\n      float ${o}(int row, int col, int depth) {\n        int stride1 = ${n}Shape[2];\n        float texR = float(row);\n        float texC = dot(vec2(col, depth), vec2(stride1, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${n}TexShape[1], ${n}TexShape[0]);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n        float ${o}(int row, int col, int depth) {\n          float texR = float(row);\n          float texC = dot(vec2(col, depth), vec2(${i}, 1));\n          vec2 uv = (vec2(texC, texR) + halfCR) /\n                     vec2(${m}.0, ${p}.0);\n          return sampleTexture(${n}, uv);\n        }\n      `;if(m===i&&f==null)return t?`\n      float ${o}(int row, int col, int depth) {\n        float texR = dot(vec2(row, col), vec2(${n}Shape[1], 1));\n        float texC = float(depth);\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n    float ${o}(int row, int col, int depth) {\n      float texR = dot(vec2(row, col), vec2(${e[1]}, 1));\n      float texC = float(depth);\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${m}.0, ${p}.0);\n      return sampleTexture(${n}, uv);\n    }\n  `;let d=zc(n);return t?`\n    float ${o}(int row, int col, int depth) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int stride0 = ${n}Shape[1] * ${n}Shape[2];\n      int stride1 = ${n}Shape[2];\n      int index = row * stride0 + col * stride1 + depth + ${d};\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);\n      return sampleTexture(${n}, uv);\n    }\n    `:`\n      float ${o}(int row, int col, int depth) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${s} + col * ${i} + depth + ${d};\n        vec2 uv = uvFromFlat(${p}, ${m}, index);\n        return sampleTexture(${n}, uv);\n      }\n  `}function qtt(r,t){let e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1),o=Ge();if(t)return`\n    vec4 ${n}(int b2, int b, int row, int col) {\n      int valuesPerRow = int(ceil(float(${e}Shape[3]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${e}Shape[2]) / 2.0));\n      int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);\n      texelsInBatch *= ${e}Shape[1];\n      index = b2 * texelsInBatch + index;\n      ivec2 packedTexShape = ivec2(ceil(float(${e}TexShape[0]) / 2.0), ceil(float(${e}TexShape[1]) / 2.0));\n      int texR = index / packedTexShape[1];\n      int texC = index - texR * packedTexShape[1];\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${o.texture2D}(${e}, uv);\n    }\n  `;let s=r.shapeInfo.logicalShape,i=s.length,a=r.shapeInfo.texShape,u=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)],l=u[0],c=u[1],p=Math.ceil(s[i-1]/2),m=p*Math.ceil(s[i-2]/2),f=\"int b, int row, int col\",d=`b * ${m} + (row / 2) * ${p} + (col / 2)`;for(let h=2;h<i-1;h++)f=`int b${h}, `+f,m*=s[i-h-1],d=`b${h} * ${m} + `+d;return`\n    vec4 ${n}(${f}) {\n      int index = ${d};\n      int texR = index / ${c};\n      int texC = index - texR * ${c};\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}, ${l});\n      return ${o.texture2D}(${e}, uv);\n    }\n  `}function Ktt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o=\"get\"+n.charAt(0).toUpperCase()+n.slice(1),s=e[3],i=e[2]*s,a=e[1]*i,{newShape:u,keptDims:l}=y.squeezeShape(e);if(u.length<e.length){let b=wd(r,u),w=[\"row\",\"col\",\"depth\",\"depth2\"];return`\n      ${yd(b,t)}\n      float ${o}(int row, int col, int depth, int depth2) {\n        return ${o}(${Cd(w,l)});\n      }\n    `}if(r.shapeInfo.isUniform)return`\n      float ${o}(int row, int col, int depth, int depth2) {\n        int index = round(dot(vec4(row, col, depth, depth2),\n                          vec4(${a}, ${i}, ${s}, 1)));\n        ${bd(r)}\n      }\n    `;let c=r.shapeInfo.flatOffset,p=r.shapeInfo.texShape,m=p[0],f=p[1],d=`int stride2 = ${n}Shape[3];`,h=`int stride1 = ${n}Shape[2] * stride2;`,g=`int stride0 = ${n}Shape[1] * stride1;`;if(f===a&&c==null)return t?`\n      float ${o}(int row, int col, int depth, int depth2) {\n        ${d}\n        ${h}\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(stride1, stride2, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${n}TexShape[1], ${n}TexShape[0]);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${o}(int row, int col, int depth, int depth2) {\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(${i}, ${s}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${f}.0, ${m}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(f===s&&c==null)return t?`\n      float ${o}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${n}Shape[1] * ${n}Shape[2], ${n}Shape[2], 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${n}TexShape[1], ${n}TexShape[0]);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${o}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${e[1]*e[2]}, ${e[2]}, 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${f}.0, ${m}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;let x=zc(n);return t?`\n    float ${o}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      ${d}\n      ${h}\n      ${g}\n      int index = row * stride0 + col * stride1 +\n          depth * stride2 + depth2;\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${x});\n      return sampleTexture(${n}, uv);\n    }\n  `:`\n    float ${o}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${a} + col * ${i} +\n          depth * ${s} + depth2;\n      vec2 uv = uvFromFlat(${m}, ${f}, index + ${x});\n      return sampleTexture(${n}, uv);\n    }\n  `}function jtt(r){let t=r.shapeInfo.logicalShape,e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1),o=t[4],s=t[3]*o,i=t[2]*s,a=t[1]*i,{newShape:u,keptDims:l}=y.squeezeShape(t);if(u.length<t.length){let h=wd(r,u),g=[\"row\",\"col\",\"depth\",\"depth2\",\"depth3\"];return`\n      ${yd(h)}\n      float ${n}(int row, int col, int depth, int depth2, int depth3) {\n        return ${n}(${Cd(g,l)});\n      }\n    `}if(r.shapeInfo.isUniform)return`\n      float ${n}(int row, int col, int depth, int depth2, int depth3) {\n        float index = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${a}, ${i}, ${s}, ${o})) +\n          depth3;\n        ${bd(r)}\n      }\n    `;let c=r.shapeInfo.flatOffset,p=r.shapeInfo.texShape,m=p[0],f=p[1];if(f===a&&c==null)return`\n      float ${n}(int row, int col, int depth, int depth2, int depth3) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n                         vec4(${i}, ${s}, ${o}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${f}.0, ${m}.0);\n        return sampleTexture(${e}, uv);\n      }\n    `;if(f===o&&c==null)return`\n      float ${n}(int row, int col, int depth, int depth2, int depth3) {\n        float texR = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]},\n               ${t[2]*t[3]}, ${t[3]}, 1));\n        int texC = depth3;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${f}.0, ${m}.0);\n        return sampleTexture(${e}, uv);\n      }\n    `;let d=zc(e);return`\n    float ${n}(int row, int col, int depth, int depth2, int depth3) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${a} + col * ${i} + depth * ${s} +\n          depth2 * ${o} + depth3 + ${d};\n      vec2 uv = uvFromFlat(${m}, ${f}, index);\n      return sampleTexture(${e}, uv);\n    }\n  `}function Xtt(r){let t=r.shapeInfo.logicalShape,e=r.name,n=\"get\"+e.charAt(0).toUpperCase()+e.slice(1),{newShape:o,keptDims:s}=y.squeezeShape(t);if(o.length<t.length){let g=wd(r,o),x=[\"row\",\"col\",\"depth\",\"depth2\",\"depth3\",\"depth4\"];return`\n      ${yd(g)}\n      float ${n}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        return ${n}(${Cd(x,s)});\n      }\n    `}let i=t[5],a=t[4]*i,u=t[3]*a,l=t[2]*u,c=t[1]*l;if(r.shapeInfo.isUniform)return`\n      float ${n}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n        int index = round(dot(\n          vec4(row, col, depth, depth2),\n          vec4(${c}, ${l}, ${u}, ${a})) +\n          dot(\n            vec2(depth3, depth4),\n            vec2(${i}, 1)));\n        ${bd(r)}\n      }\n    `;let p=r.shapeInfo.flatOffset,m=r.shapeInfo.texShape,f=m[0],d=m[1];if(d===c&&p==null)return`\n      float ${n}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n          vec4(${l}, ${u}, ${a}, ${i})) +\n               float(depth4);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${d}.0, ${f}.0);\n        return sampleTexture(${e}, uv);\n      }\n    `;if(d===i&&p==null)return`\n      float ${n}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        float texR = dot(vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]*t[4]},\n               ${t[2]*t[3]*t[4]},\n               ${t[3]*t[4]},\n               ${t[4]})) + float(depth3);\n        int texC = depth4;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${d}.0, ${f}.0);\n        return sampleTexture(${e}, uv);\n      }\n    `;let h=zc(e);return`\n    float ${n}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${c} + col * ${l} + depth * ${u} +\n          depth2 * ${a} + depth3 * ${i} + depth4 + ${h};\n      vec2 uv = uvFromFlat(${f}, ${d}, index);\n      return sampleTexture(${e}, uv);\n    }\n  `}function bd(r){let t=r.name,e=y.sizeFromShape(r.shapeInfo.logicalShape);return e<2?`return ${t};`:`\n    for (int i = 0; i < ${e}; i++) {\n      if (i == index) {\n        return ${t}[i];\n      }\n    }\n  `}function Ytt(r,t){let e=r.name,n=e.charAt(0).toUpperCase()+e.slice(1),o=\"get\"+n+\"AtOutCoords\",s=r.shapeInfo.logicalShape.length,i=t.logicalShape.length,a=rL(r.shapeInfo.logicalShape,t.logicalShape),u=zt(i),l=i-s,c,p=[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"];s===0?c=\"\":i<2&&a.length>=1?c=\"coords = 0;\":c=a.map(b=>`coords.${p[b+l]} = 0;`).join(`\n`);let m=\"\";i<2&&s>0?m=\"coords\":m=r.shapeInfo.logicalShape.map((b,w)=>`coords.${p[w+l]}`).join(\", \");let f=\"return outputValue;\",h=y.sizeFromShape(r.shapeInfo.logicalShape)===1,x=y.sizeFromShape(t.logicalShape)===1;if(s===1&&!h&&!x)f=`\n      return vec4(outputValue.xy, outputValue.xy);\n    `;else if(h&&!x)i===1?f=`\n        return vec4(outputValue.x, outputValue.x, 0., 0.);\n      `:f=`\n        return vec4(outputValue.x);\n      `;else if(a.length){let b=s-2,w=s-1;a.indexOf(b)>-1&&a.indexOf(w)>-1?f=\"return vec4(outputValue.x);\":a.indexOf(b)>-1?f=\"return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);\":a.indexOf(w)>-1&&(f=\"return vec4(outputValue.xx, outputValue.zz);\")}return`\n    vec4 ${o}() {\n      ${u} coords = getOutputCoords();\n      ${c}\n      vec4 outputValue = get${n}(${m});\n      ${f}\n    }\n  `}function Ztt(r,t){let e=r.name,n=e.charAt(0).toUpperCase()+e.slice(1),o=\"get\"+n+\"AtOutCoords\",s=t.texShape,i=r.shapeInfo.texShape,a=r.shapeInfo.logicalShape.length,u=t.logicalShape.length;if(!r.shapeInfo.isUniform&&a===u&&r.shapeInfo.flatOffset==null&&y.arraysEqual(i,s))return`\n      float ${o}() {\n        return sampleTexture(${e}, resultUV);\n      }\n    `;let l=zt(u),c=rL(r.shapeInfo.logicalShape,t.logicalShape),p=u-a,m,f=[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"];a===0?m=\"\":u<2&&c.length>=1?m=\"coords = 0;\":m=c.map(h=>`coords.${f[h+p]} = 0;`).join(`\n`);let d=\"\";return u<2&&a>0?d=\"coords\":d=r.shapeInfo.logicalShape.map((h,g)=>`coords.${f[g+p]}`).join(\", \"),`\n    float ${o}() {\n      ${l} coords = getOutputCoords();\n      ${m}\n      return get${n}(${d});\n    }\n  `}function zt(r){if(r<=1)return\"int\";if(r===2)return\"ivec2\";if(r===3)return\"ivec3\";if(r===4)return\"ivec4\";if(r===5)return\"ivec5\";if(r===6)return\"ivec6\";throw Error(`GPU for rank ${r} is not yet supported`)}function Tw(r,t,e){let{newShape:n,keptDims:o}=y.squeezeShape(t),s=t.length,i=r&&s===3&&t[0]===1,a=i?t.slice(1):n,u=!r&&s>1&&!y.arraysEqual(t,e)&&n.length<s||i;return{useSqueezeShape:u,uniformShape:u?a:t,keptDims:o}}function wd(r,t){let e=JSON.parse(JSON.stringify(r));return e.shapeInfo.logicalShape=t,e}function Cd(r,t){return t.map(e=>r[e]).join(\", \")}function aL(r,t,e,n){let o=e.map((c,p)=>{let m={logicalShape:c.shape,texShape:c.isUniform?null:c.texData.texShape,isUniform:c.isUniform,isPacked:c.isUniform?!1:c.texData.isPacked,flatOffset:null};return c.texData!=null&&c.texData.slice!=null&&c.texData.slice.flatOffset>0&&(m.flatOffset=c.texData.slice.flatOffset),{name:t.variableNames[p],shapeInfo:m}}),s=o.map(c=>c.shapeInfo),i={logicalShape:n.shape,texShape:n.texData.texShape,isUniform:!1,isPacked:n.texData.isPacked,flatOffset:null},a=nL(o,i,t),u=NT(r.gl,a),l=r.createProgram(u);return z().get(\"ENGINE_COMPILE_ONLY\")?{program:t,fragmentShader:u,source:a,webGLProgram:l,inShapeInfos:s,outShapeInfo:i,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:Object.assign({program:t,fragmentShader:u,source:a,webGLProgram:l,inShapeInfos:s,outShapeInfo:i},WT(r,t,l))}function WT(r,t,e){let n={},o={},s={},i=[],a,u,l,c=null,p=null;p=r.getUniformLocation(e,\"NAN\",!1),z().getNumber(\"WEBGL_VERSION\")===1&&(c=r.getUniformLocation(e,\"INFINITY\",!1));let m=!1;for(let f=0;f<t.variableNames.length;f++){let d=t.variableNames[f];n[d]=r.getUniformLocation(e,d,m),n[`offset${d}`]=r.getUniformLocation(e,`offset${d}`,m),t.enableShapeUniforms&&(o[`${d}Shape`]=r.getUniformLocation(e,`${d}Shape`,m),s[`${d}TexShape`]=r.getUniformLocation(e,`${d}TexShape`,m))}return t.enableShapeUniforms&&(a=r.getUniformLocation(e,\"outShape\",m),l=r.getUniformLocation(e,\"outShapeStrides\",m),u=r.getUniformLocation(e,\"outTexShape\",m)),t.customUniforms&&t.customUniforms.forEach((f,d)=>{i[d]=r.getUniformLocation(e,f.name,m)}),{uniformLocations:n,customUniformLocations:i,infLoc:c,nanLoc:p,inShapesLocations:o,inTexShapesLocations:s,outShapeLocation:a,outShapeStridesLocation:l,outTexShapeLocation:u}}function iL(r,t){if(r.length!==t.length)throw Error(`Binary was compiled with ${r.length} inputs, but was executed with ${t.length} inputs`);r.forEach((e,n)=>{let o=e.logicalShape,s=t[n],i=s.shape;if(!y.arraysEqual(o,i))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${o} and ${i} must match`);if(e.isUniform&&s.isUniform)return;let a=e.texShape,u=s.isUniform?null:s.texData.texShape;if(!y.arraysEqual(a,u))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${a} and ${u} must match`)})}function lL(r,t,e,n,o){t.program.enableShapeUniforms||(iL(t.inShapeInfos,e),iL([t.outShapeInfo],[n]));let s=n.texData.texture,i=n.texData.texShape;n.texData.isPacked?r.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):r.setOutputMatrixTexture(s.texture,i[0],i[1]),r.setProgram(t.webGLProgram),z().getNumber(\"WEBGL_VERSION\")===1&&t.infLoc!==null&&r.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&r.gl.uniform1f(t.nanLoc,NaN),e.forEach((u,l)=>{let c=t.program.variableNames[l],p=t.uniformLocations[c],m=t.uniformLocations[`offset${c}`],f=t.inShapesLocations[`${c}Shape`],d=t.inTexShapesLocations[`${c}TexShape`];if(f){let{uniformShape:h}=Tw(t.program.packedInputs,u.shape,u.texData.texShape);switch(h.length){case 1:r.gl.uniform1iv(f,new Int32Array(h));break;case 2:r.gl.uniform2iv(f,new Int32Array(h));break;case 3:r.gl.uniform3iv(f,new Int32Array(h));break;case 4:r.gl.uniform4iv(f,new Int32Array(h));break;default:break}}if(d&&r.gl.uniform2i(d,u.texData.texShape[0],u.texData.texShape[1]),p!=null){if(u.isUniform){if(y.sizeFromShape(u.shape)<2)r.gl.uniform1f(p,u.uniformValues[0]);else{let h=u.uniformValues;h instanceof Float32Array||(h=new Float32Array(h)),r.gl.uniform1fv(p,h)}return}u.texData.slice!=null&&m!=null&&r.gl.uniform1i(m,u.texData.slice.flatOffset),r.setInputMatrixTexture(u.texData.texture.texture,p,l)}});let a=t.outShapeLocation;if(a)switch(n.shape.length){case 1:r.gl.uniform1iv(a,new Int32Array(n.shape));break;case 2:r.gl.uniform2iv(a,new Int32Array(n.shape));break;case 3:r.gl.uniform3iv(a,new Int32Array(n.shape));break;case 4:r.gl.uniform4iv(a,new Int32Array(n.shape));break;default:break}if(t.outShapeStridesLocation){let u=y.computeStrides(n.shape);switch(n.shape.length){case 2:r.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(u));break;case 3:r.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(u));break;case 4:r.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(u));break;default:break}}t.outTexShapeLocation&&r.gl.uniform2i(t.outTexShapeLocation,n.texData.texShape[0],n.texData.texShape[1]),t.program.customUniforms&&o&&t.program.customUniforms.forEach((u,l)=>{let c=t.customUniformLocations[l],p=o[l];if(u.type===\"float\")r.gl.uniform1fv(c,p);else if(u.type===\"vec2\")r.gl.uniform2fv(c,p);else if(u.type===\"vec3\")r.gl.uniform3fv(c,p);else if(u.type===\"vec4\")r.gl.uniform4fv(c,p);else if(u.type===\"int\")r.gl.uniform1iv(c,p);else if(u.type===\"ivec2\")r.gl.uniform2iv(c,p);else if(u.type===\"ivec3\")r.gl.uniform3iv(c,p);else if(u.type===\"ivec4\")r.gl.uniform4iv(c,p);else throw Error(`uniform type ${u.type} is not supported yet.`)}),r.executeProgram()}function uL(r,t,e){let n=\"\";t.concat(e).forEach(i=>{let a=i.texData!=null&&i.texData.slice!=null&&i.texData.slice.flatOffset>0;if(r.enableShapeUniforms&&!i.isUniform){let u=i.texData.texShape,{useSqueezeShape:l,uniformShape:c,keptDims:p}=Tw(r.packedInputs,i.shape,u),m=\"\",f=\"\",d=\"\";if(c.length===1&&r.packedInputs){let N=[Math.ceil(u[0]/2),Math.ceil(u[1]/2)];m=`${N[0]>1}_${N[1]>1}`}else if(c.length===2&&!r.packedInputs)f=`${c[0]>1}_${c[1]>1}`;else if(c.length>2&&!r.packedInputs){let N=y.computeStrides(c);d=`${N[0]===u[1]}_${N[N.length-1]===u[1]}`}let h=i.shape.length,g=c.length===2&&y.arraysEqual(i.shape,u),x=y.sizeFromShape(i.shape)===1,b=v.getBroadcastDims(i.shape,e.shape),w=!r.packedInputs&&h===e.shape.length&&y.arraysEqual(u,e.texData.texShape),C=r.packedInputs||c.length>2?\"\":`${u[0]>1}_${u[1]>1}`;n+=`${h}_${w}_${l?p:\"\"}_${c.length}_${x}_${b}_${g}_${m}_${f}_${d}_${C}_${a}`}else{let u=i.isUniform?\"uniform\":i.texData.texShape;n+=`${i.shape}_${u}_${a}`}});let o=r.userCode,s=r.constructor.name;return s+=\"_\"+n+\"_\"+o+`${z().getNumber(\"WEBGL_VERSION\")}`,s}function we(r){return z().getBool(\"WEBGL_USE_SHAPES_UNIFORMS\")&&r<=4}var kw=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=ku.DENSE,this.customUniforms=[{name:\"texShape\",type:\"ivec2\"}];let e=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?Mc([\"r\",\"c\",\"d\"],t):ti([\"r\",\"c\",\"d\"],t)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getA(rc.x, rc.y, rc.z);\n        }\n\n        ${e.output} = result;\n      }\n    `}};var Ew=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=ku.DENSE,this.customUniforms=[{name:\"texShape\",type:\"ivec2\"}];let e=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?Mc([\"r\",\"c\",\"d\"],t):ti([\"r\",\"c\",\"d\"],t)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));\n        }\n\n        ${e.output} = result;\n      }\n    `}};var _w=class{constructor(t){this.variableNames=[\"A\"],this.outTexUsage=jr.DOWNLOAD;let e=Ge();this.outputShape=t,this.userCode=`\n      ${Nw}\n\n      void main() {\n        float x = getAAtOutCoords();\n        ${e.output} = encode_float(x);\n      }\n    `}};var Aw=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=jr.DOWNLOAD;let e=Ge();this.outputShape=t,this.userCode=`\n      ${Nw}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));\n        ${e.output} = encode_float(x);\n      }\n    `}};var tet={R:0,G:1,B:2,A:3},Jh=class{constructor(t,e=!1,n=\"RGBA\"){this.variableNames=[\"A\"],this.customUniforms=[{name:\"texShape\",type:\"ivec2\"}];let o=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let s=\"result\";e&&(s=\"floor(result * 255. + 0.5)\");let i=\"\";for(let a=0;a<n.length;a++){let u=n[a];i+=`\n          if(offset == ${a}) {\n            result = values[${tet[u]}];\n          }`}this.userCode=`\n      ${this.enableShapeUniforms?gd():hd(t)}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int flatIndex = getFlatIndex(coords);\n        float result = 0.;\n        int offset = imod(flatIndex, ${n.length});\n\n        flatIndex = idiv(flatIndex, ${n.length}, 1.);\n\n        int r = flatIndex / texShape[1];\n        if (r < texShape[0]) {\n          int c = imod(flatIndex, texShape[1]);\n          vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n          vec4 values = ${o.texture2D}(A, uv);\n          ${i}\n        }\n        ${o.output} = vec4(${s}, 0., 0., 0.);\n      }\n    `}};var $w=class{constructor(t,e=!1){this.variableNames=[\"A\"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:\"texShape\",type:\"ivec2\"}];let n=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let o=\"\",s=\"result\";e&&(s=\"floor(result * 255. + 0.5)\");for(let i=0;i<=1;i++)for(let a=0;a<=1;a++){let u=i*2+a;o+=`\n          localCoords = coords;\n          if(localCoords[2] + ${a} < ${this.enableShapeUniforms?\"outShape[2]\":`${t[2]}`}) {\n          localCoords[2] += ${a};\n          if (localCoords[1] + ${i} < ${this.enableShapeUniforms?\"outShape[1]\":`${t[1]}`}) {\n            localCoords[1] += ${i};\n\n            flatIndex = getFlatIndex(localCoords);\n            offset = imod(flatIndex, 4);\n\n            flatIndex = idiv(flatIndex, 4, 1.);\n\n            int r = flatIndex / texShape[1];\n            int c = imod(flatIndex, texShape[1]);\n            vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n            values = ${n.texture2D}(A, uv);\n\n            if (offset == 0) {\n              result[${u}] = values[0];\n            } else if (offset == 1) {\n              result[${u}] = values[1];\n            } else if (offset == 2) {\n              result[${u}] = values[2];\n            } else {\n              result[${u}] = values[3];\n            }\n          }\n        }\n        `}this.userCode=`\n        ${this.enableShapeUniforms?gd():hd(t)}\n\n        void main() {\n          ivec3 coords = getOutputCoords();\n\n          vec4 result = vec4(0.);\n          int flatIndex, r, c, offset;\n          ivec3 localCoords;\n          vec2 uv;\n          vec4 values;\n\n          ${o}\n\n          ${n.output} = ${s};\n        }\n    `}};var ik={};Wt(ik,{bindVertexProgramAttributeStreams:()=>JT,createBufferFromOutputTexture:()=>ek,createFloat16MatrixTexture:()=>jT,createFloat16PackedMatrixTexture:()=>ZT,createFloat32MatrixTexture:()=>KT,createIndexBuffer:()=>qT,createPackedMatrixTexture:()=>YT,createUnsignedBytesMatrixTexture:()=>XT,createVertexBuffer:()=>HT,createVertexShader:()=>UT,downloadByteEncodedFloatMatrixFromOutputTexture:()=>nk,downloadFloat32MatrixFromBuffer:()=>rk,downloadMatrixFromPackedOutputTexture:()=>sk,downloadPackedMatrixFromBuffer:()=>ok,getInternalFormatForFloat16MatrixTexture:()=>Rw,getInternalFormatForFloat16PackedMatrixTexture:()=>Pw,getInternalFormatForFloat32MatrixTexture:()=>Dw,getInternalFormatForPackedMatrixTexture:()=>Ow,getInternalFormatForUnsignedBytesMatrixTexture:()=>Fw,uploadDenseMatrixToTexture:()=>QT,uploadPixelDataToTexture:()=>tk});function UT(r){let t=Ge(),e=`${t.version}\n    precision highp float;\n    ${t.attribute} vec3 clipSpacePos;\n    ${t.attribute} vec2 uv;\n    ${t.varyingVs} vec2 resultUV;\n\n    void main() {\n      gl_Position = vec4(clipSpacePos, 1);\n      resultUV = uv;\n    }`;return vT(r,e)}function HT(r){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return ET(r,t)}function qT(r){let t=new Uint16Array([0,1,2,2,1,3]);return _T(r,t)}function Qh(r,t,e,n,o,s){$T(t,e);let i=AT(r),a=r.TEXTURE_2D;return yt(r,()=>r.bindTexture(a,i)),yt(r,()=>r.texParameteri(a,r.TEXTURE_WRAP_S,r.CLAMP_TO_EDGE)),yt(r,()=>r.texParameteri(a,r.TEXTURE_WRAP_T,r.CLAMP_TO_EDGE)),yt(r,()=>r.texParameteri(a,r.TEXTURE_MIN_FILTER,r.NEAREST)),yt(r,()=>r.texParameteri(a,r.TEXTURE_MAG_FILTER,r.NEAREST)),z().getNumber(\"WEBGL_VERSION\")===1?yt(r,()=>r.texImage2D(a,0,n,t,e,0,o,s,null)):yt(r,()=>r.texStorage2D(a,1,n,t,e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null)),{texture:i,texShape:[e,t]}}function Dw(r){return r.internalFormatFloat}function KT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Dw(n),n.textureFormatFloat,r.FLOAT)}function Rw(r){return r.internalFormatHalfFloat}function jT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Rw(n),n.textureFormatFloat,n.textureTypeHalfFloat)}function Fw(r){return r.downloadTextureFormat}function XT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Fw(n),r.RGBA,r.UNSIGNED_BYTE)}function Ow(r){return r.internalFormatPackedFloat}function YT(r,t,e,n){let[o,s]=Xi(t,e);return Qh(r,o,s,Ow(n),r.RGBA,r.FLOAT)}function Pw(r){return r.internalFormatPackedHalfFloat}function ZT(r,t,e,n){let[o,s]=Xi(t,e);return Qh(r,o,s,Pw(n),r.RGBA,n.textureTypeHalfFloat)}function JT(r,t,e){return yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,e)),Iw(r,t,\"clipSpacePos\",e,3,20,0)&&Iw(r,t,\"uv\",e,2,20,12)}function QT(r,t,e,n,o,s){yt(r,()=>r.bindTexture(r.TEXTURE_2D,t));let i,a,u;o instanceof Uint8Array?(i=new Uint8Array(e*n*4),a=r.UNSIGNED_BYTE,u=r.RGBA):(i=new Float32Array(e*n*4),a=r.FLOAT,u=s.internalFormatPackedFloat),i.set(o),z().getNumber(\"WEBGL_VERSION\")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,e,n,r.RGBA,a,i)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,u,e,n,0,r.RGBA,a,i)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function tk(r,t,e){yt(r,()=>r.bindTexture(r.TEXTURE_2D,t)),e.data instanceof Uint8Array?z().getNumber(\"WEBGL_VERSION\")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,e.width,e.height,r.RGBA,r.UNSIGNED_BYTE,e.data)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,r.RGBA,e.width,e.height,0,r.RGBA,r.UNSIGNED_BYTE,e.data)):z().getNumber(\"WEBGL_VERSION\")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,r.RGBA,r.UNSIGNED_BYTE,e)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,r.RGBA,r.RGBA,r.UNSIGNED_BYTE,e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function ek(r,t,e,n){let o=r.createBuffer();yt(r,()=>r.bindBuffer(r.PIXEL_PACK_BUFFER,o));let a=4*4*t*e;return yt(r,()=>r.bufferData(r.PIXEL_PACK_BUFFER,a,r.STREAM_READ)),yt(r,()=>r.readPixels(0,0,e,t,r.RGBA,r.FLOAT,0)),yt(r,()=>r.bindBuffer(r.PIXEL_PACK_BUFFER,null)),o}function rk(r,t,e){let n=r,o=new Float32Array(e);return n.bindBuffer(n.PIXEL_PACK_BUFFER,t),n.getBufferSubData(n.PIXEL_PACK_BUFFER,0,o),n.bindBuffer(n.PIXEL_PACK_BUFFER,null),o}function nk(r,t,e,n){let[o,s]=Lc(t,e),i=4,a=new Uint8Array(XP(t*e,i));return yt(r,()=>r.readPixels(0,0,o,s,n.downloadTextureFormat,r.UNSIGNED_BYTE,a)),new Float32Array(a.buffer)}function ok(r,t,e,n,o,s,i,a){let u=r,l=new Float32Array(YP(s,i));return u.bindBuffer(u.PIXEL_PACK_BUFFER,t),u.getBufferSubData(u.PIXEL_PACK_BUFFER,0,l),u.bindBuffer(u.PIXEL_PACK_BUFFER,null),l}function sk(r,t,e){let n=new Float32Array(t*e*4);return yt(r,()=>r.readPixels(0,0,e,t,r.RGBA,r.FLOAT,n)),n}var Bc=class{constructor(t){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let e=z().getNumber(\"WEBGL_VERSION\");t!=null?(this.gl=t,wT(e,t)):this.gl=Gn(e);let n=\"WEBGL_color_buffer_float\",o=\"EXT_color_buffer_half_float\";if(this.parallelCompilationExtension=this.gl.getExtension(\"KHR_parallel_shader_compile\"),z().getNumber(\"WEBGL_VERSION\")===1){let s=\"OES_texture_float\",i=\"OES_texture_half_float\";if(this.textureFloatExtension=pd(this.gl,s),Wn(this.gl,i))this.textureHalfFloatExtension=pd(this.gl,i);else if(z().get(\"WEBGL_FORCE_F16_TEXTURES\"))throw new Error(\"GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.\");if(this.colorBufferFloatExtension=this.gl.getExtension(n),Wn(this.gl,o))this.colorBufferHalfFloatExtension=pd(this.gl,o);else if(z().get(\"WEBGL_FORCE_F16_TEXTURES\"))throw new Error(\"GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.\")}else if(n=\"EXT_color_buffer_float\",Wn(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else if(Wn(this.gl,o))this.colorBufferHalfFloatExtension=this.gl.getExtension(o);else throw new Error(\"GL context does not support color renderable floats\");this.vertexBuffer=HT(this.gl),this.indexBuffer=qT(this.gl),this.framebuffer=DT(this.gl),this.textureConfig=Xh(this.gl,this.textureHalfFloatExtension)}get debug(){return z().getBool(\"DEBUG\")}dispose(){if(this.disposed)return;this.program!=null&&console.warn(\"Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing.\"),this.outputTexture!=null&&console.warn(\"Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.\");let t=this.gl;yt(t,()=>t.finish()),yt(t,()=>t.bindFramebuffer(t.FRAMEBUFFER,null)),yt(t,()=>t.deleteFramebuffer(this.framebuffer)),yt(t,()=>t.bindBuffer(t.ARRAY_BUFFER,null)),yt(t,()=>t.bindBuffer(t.ELEMENT_ARRAY_BUFFER,null)),yt(t,()=>t.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(t,e){return this.throwIfDisposed(),KT(this.gl,t,e,this.textureConfig)}createFloat16MatrixTexture(t,e){return this.throwIfDisposed(),jT(this.gl,t,e,this.textureConfig)}createUnsignedBytesMatrixTexture(t,e){return this.throwIfDisposed(),XT(this.gl,t,e,this.textureConfig)}uploadPixelDataToTexture(t,e){this.throwIfDisposed(),tk(this.gl,t,e)}uploadDenseMatrixToTexture(t,e,n,o){this.throwIfDisposed(),QT(this.gl,t,e,n,o,this.textureConfig)}createFloat16PackedMatrixTexture(t,e){return this.throwIfDisposed(),ZT(this.gl,t,e,this.textureConfig)}createPackedMatrixTexture(t,e){return this.throwIfDisposed(),YT(this.gl,t,e,this.textureConfig)}deleteMatrixTexture(t){this.throwIfDisposed(),this.outputTexture===t&&(Sw(this.gl,this.framebuffer),this.outputTexture=null),yt(this.gl,()=>this.gl.deleteTexture(t))}downloadByteEncodedFloatMatrixFromOutputTexture(t,e,n){return this.downloadMatrixDriver(t,()=>nk(this.gl,e,n,this.textureConfig))}downloadPackedMatrixFromBuffer(t,e,n,o,s,i){return ok(this.gl,t,e,n,o,s,i,this.textureConfig)}downloadFloat32MatrixFromBuffer(t,e){return rk(this.gl,t,e)}createBufferFromTexture(t,e,n){this.bindTextureToFrameBuffer(t);let o=ek(this.gl,e,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),o}createAndWaitForFence(){let t=this.createFence(this.gl);return this.pollFence(t)}createFence(t){let e,n;if(z().getBool(\"WEBGL_FENCE_API_ENABLED\")){let o=t,s=o.fenceSync(o.SYNC_GPU_COMMANDS_COMPLETE,0);t.flush(),n=()=>{let i=o.clientWaitSync(s,0,0);return i===o.ALREADY_SIGNALED||i===o.CONDITION_SATISFIED},e=s}else z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")>0?(e=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(e,z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\"))):n=()=>!0;return{query:e,isFencePassed:n}}downloadMatrixFromPackedTexture(t,e,n){return this.downloadMatrixDriver(t,()=>sk(this.gl,e,n))}createProgram(t){this.throwIfDisposed();let e=this.gl;this.vertexShader==null&&(this.vertexShader=UT(e));let n=TT(e);return yt(e,()=>e.attachShader(n,this.vertexShader)),yt(e,()=>e.attachShader(n,t)),kT(e,n),this.debug&&Yh(e,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=JT(e,this.program,this.vertexBuffer)),n}deleteProgram(t){this.throwIfDisposed(),t===this.program&&(this.program=null),t!=null&&yt(this.gl,()=>this.gl.deleteProgram(t))}setProgram(t){this.throwIfDisposed(),this.program=t,this.program!=null&&this.debug&&Yh(this.gl,this.program),yt(this.gl,()=>this.gl.useProgram(t))}getUniformLocation(t,e,n=!0){return this.throwIfDisposed(),n?RT(this.gl,t,e):FT(this.gl,t,e)}getAttributeLocation(t,e){return this.throwIfDisposed(),yt(this.gl,()=>this.gl.getAttribLocation(t,e))}getUniformLocationNoThrow(t,e){return this.throwIfDisposed(),this.gl.getUniformLocation(t,e)}setInputMatrixTexture(t,e,n){this.throwIfDisposed(),this.throwIfNoProgram(),OT(this.gl,t,e,n)}setOutputMatrixTexture(t,e,n){this.setOutputMatrixTextureDriver(t,n,e)}setOutputPackedMatrixTexture(t,e,n){this.throwIfDisposed();let[o,s]=Xi(e,n);this.setOutputMatrixTextureDriver(t,o,s)}setOutputMatrixWriteRegion(t,e,n,o){this.setOutputMatrixWriteRegionDriver(n,t,o,e)}setOutputPackedMatrixWriteRegion(t,e,n,o){throw new Error(\"setOutputPackedMatrixWriteRegion not implemented.\")}debugValidate(){this.program!=null&&Yh(this.gl,this.program),md(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let t=this.gl;this.debug&&this.debugValidate(),yt(t,()=>t.drawElements(t.TRIANGLES,6,t.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),yt(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=pd(this.gl,z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")===2?\"EXT_disjoint_timer_query_webgl2\":\"EXT_disjoint_timer_query\")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")===2){let n=this.gl,o=this.getQueryTimerExtensionWebGL2(),s=n.createQuery();return n.beginQuery(o.TIME_ELAPSED_EXT,s),s}let t=this.getQueryTimerExtensionWebGL1(),e=t.createQueryEXT();return t.beginQueryEXT(t.TIME_ELAPSED_EXT,e),e}endQuery(){if(z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")===2){let e=this.gl,n=this.getQueryTimerExtensionWebGL2();e.endQuery(n.TIME_ELAPSED_EXT);return}let t=this.getQueryTimerExtensionWebGL1();t.endQueryEXT(t.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(t){return await y.repeatedTry(()=>this.disposed||this.isQueryAvailable(t,z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\"))),this.getQueryTime(t,z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\"))}getQueryTime(t,e){if(e===0)return null;if(e===2){let n=this.gl;return n.getQueryParameter(t,n.QUERY_RESULT)/1e6}else{let n=this.getQueryTimerExtensionWebGL1();return n.getQueryObjectEXT(t,n.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(t,e){if(e===0)return!0;if(e===2){let n=this.gl,o=this.getQueryTimerExtensionWebGL2(),s=n.getQueryParameter(t,n.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(o.GPU_DISJOINT_EXT)),s&&!this.disjoint}else{let n=this.getQueryTimerExtensionWebGL1(),o=n.getQueryObjectEXT(t,n.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),o&&!this.disjoint}}pollFence(t){return new Promise(e=>{this.addItemToPoll(()=>t.isFencePassed(),()=>e())})}pollItems(){let t=eet(this.itemsToPoll.map(e=>e.isDoneFn));for(let e=0;e<=t;++e){let{resolveFn:n}=this.itemsToPoll[e];n()}this.itemsToPoll=this.itemsToPoll.slice(t+1)}addItemToPoll(t,e){if(this.itemsToPoll.push({isDoneFn:t,resolveFn:e}),this.itemsToPoll.length>1)return;let n;\"setTimeoutCustom\"in z().platform&&(n=z().platform.setTimeoutCustom.bind(z().platform)),y.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0),()=>0,null,n)}bindTextureToFrameBuffer(t){this.throwIfDisposed(),Zh(this.gl,t,this.framebuffer),this.debug&&md(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(Zh(this.gl,this.outputTexture,this.framebuffer),this.debug&&md(this.gl)):Sw(this.gl,this.framebuffer)}downloadMatrixDriver(t,e){this.bindTextureToFrameBuffer(t);let n=e();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(t,e,n){this.throwIfDisposed();let o=this.gl;Zh(o,t,this.framebuffer),this.debug&&md(o),this.outputTexture=t,yt(o,()=>o.viewport(0,0,e,n)),yt(o,()=>o.scissor(0,0,e,n))}setOutputMatrixWriteRegionDriver(t,e,n,o){this.throwIfDisposed(),yt(this.gl,()=>this.gl.scissor(t,e,n,o))}throwIfDisposed(){if(this.disposed)throw new Error(\"Attempted to use disposed GPGPUContext.\")}throwIfNoProgram(){if(this.program==null)throw new Error(\"No GPU program is currently set.\")}};function eet(r){let t=0;for(;t<r.length&&r[t]();++t);return t-1}var{addImpl:cL,bincountImpl:Lw,bincountReduceImpl:pL,castImpl:mL,ceilImpl:fL,concatImpl:dL,equalImpl:hL,expImpl:gL,expm1Impl:xL,floorImpl:yL,gatherNdImpl:bL,gatherV2Impl:wL,greaterImpl:CL,greaterEqualImpl:IL,lessImpl:SL,lessEqualImpl:vL,linSpaceImpl:NL,logImpl:TL,maxImpl:kL,maximumImpl:EL,minimumImpl:_L,multiplyImpl:AL,negImpl:$L,notEqualImpl:DL,prodImpl:RL,raggedGatherImpl:FL,raggedRangeImpl:OL,raggedTensorToTensorImpl:PL,rangeImpl:LL,rsqrtImpl:ML,scatterImpl:zL,sigmoidImpl:BL,simpleAbsImpl:Mw,sliceImpl:VL,sparseFillEmptyRowsImpl:GL,sparseReshapeImpl:WL,sparseSegmentReductionImpl:zw,sqrtImpl:UL,stridedSliceImpl:HL,stringNGramsImpl:qL,stringSplitImpl:KL,stringToHashBucketFastImpl:jL,subImpl:XL,tileImpl:YL,topKImpl:ZL,transposeImpl:Vc,uniqueImpl:JL}=mw;function ak(r,t){return[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"].slice(0,t).map(e=>`${r}.${e}`)}function Qe(r,t){return t===1?[r]:ak(r,t)}function QL(r,t){if(r===1)return\"rc\";let e=\"\";for(let n=0;n<r;n++)e+=t[n],n<r-1&&(e+=\",\");return e}var Bw=class{constructor(t){if(this.variableNames=[\"A\"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=t,this.rank=t.length,this.enableShapeUniforms=we(this.outputShape.length),this.rank===0)this.userCode=`\n        void main() {\n          setOutput(vec4(getA(), 0., 0., 0.));\n        }\n      `;else{let e=Qe(\"rc\",this.rank),n=zt(this.rank),o=this.getOutOfBoundsCondition(e),s=this.getSetup(e),i=this.getOutput(e);this.userCode=`\n        void main() {\n          ${n} rc = getOutputCoords();\n\n          if(${o}) {\n            setOutput(vec4(0));\n          } else {\n            ${s}\n\n            setOutput(vec4(${i}));\n          }\n        }\n      `}}getSourceCoordsArr(t){let e=[];for(let n=0;n<=1;n++)for(let o=0;o<=1;o++){let s=`${n===0?\"r\":\"rp1\"}, ${o===0?\"c\":\"cp1\"}`;for(let i=2;i<this.rank;i++)s=`${t[t.length-1-i]},`+s;e.push(s)}return e}getOutOfBoundsCondition(t){if(this.rank===1)return`rc > ${this.enableShapeUniforms?\"outShape\":this.outputShape[0]}`;let e=\"\";for(let n=this.rank-2;n<this.rank;n++)e+=`${t[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(e+=\"||\");return e}getSetup(t){if(this.rank===1)return\"\";let e=t.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],o=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`\n      int r = ${e[0]};\n      int c = ${e[1]};\n      int rp1 = r + 1;\n      int cp1 = c + 1;\n\n      bool cEdge = cp1 >= ${n};\n      bool rEdge = rp1 >= ${o};\n    `}getOutput(t){let e=this.getSourceCoordsArr(t);return this.rank===1?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?\"outShape\":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${e[0]}),\n            cEdge ? 0. : getA(${e[1]}),\n            rEdge ? 0. : getA(${e[2]}),\n            rEdge || cEdge ? 0. : getA(${e[3]})`}};var Id=class{constructor(t,e){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"inputShape\",type:\"ivec3\"}],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let n=\"\";for(let o=0;o<4;o++){let s=\"thisRC = rc;\";o%2===1&&(s+=\"thisRC.z += 1;\"),o>1&&(s+=\"thisRC.y += 1;\"),n+=`\n        ${s}\n        ${o>0?\"if(thisRC.y < rows && thisRC.z < cols){\":\"\"}\n          int flatIndex = getFlatIndex(thisRC);\n\n          ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);\n          vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));\n\n          result[${o}] =\n            getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);\n        ${o>0?\"}\":\"\"}\n      `}this.userCode=`\n      ${ret(e,this.enableShapeUniforms)}\n      ${this.enableShapeUniforms?gd():hd(t)}\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0.);\n\n        ivec3 thisRC;\n        int rows = ${this.enableShapeUniforms?\"outShape[1]\":t[1]};\n        int cols = ${this.enableShapeUniforms?\"outShape[2]\":t[2]};\n\n        ${n}\n\n        setOutput(result);\n      }\n    `}};function ret(r,t){return`\n    ivec3 inputCoordsFromReshapedOutCoords(int index) {\n      ${t?eL([\"r\",\"c\",\"d\"],\"inputShape\"):ti([\"r\",\"c\",\"d\"],r)}\n      return ivec3(r, c, d);\n    }\n  `}var Vw=class{constructor(t){this.gpgpu=t,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(t,e,n){let o=eM(e,n),s=rM(t,o,n);s in this.freeTextures||(this.freeTextures[s]=[]),s in this.usedTextures||(this.usedTextures[s]=[]);let i=tM(t,o,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[s].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=i,this.log();let u=this.freeTextures[s].shift();return this.usedTextures[s].push(u),u}let a;return o===Pr.PACKED_2X2_FLOAT32?a=this.gpgpu.createPackedMatrixTexture(t[0],t[1]):o===Pr.PACKED_2X2_FLOAT16?a=this.gpgpu.createFloat16PackedMatrixTexture(t[0],t[1]):o===Pr.UNPACKED_FLOAT32?a=this.gpgpu.createFloat32MatrixTexture(t[0],t[1]):o===Pr.UNPACKED_FLOAT16?a=this.gpgpu.createFloat16MatrixTexture(t[0],t[1]):o===Pr.PACKED_4X1_UNSIGNED_BYTE&&(a=this.gpgpu.createUnsignedBytesMatrixTexture(t[0],t[1])),this.usedTextures[s].push(a),this.numUsedTextures++,this._numBytesAllocated+=i,this.log(),a}releaseTexture(t,e,n,o){if(this.freeTextures==null)return;let s=eM(n,o),i=rM(e,s,o);i in this.freeTextures||(this.freeTextures[i]=[]);let a=tM(e,s,this.gpgpu.gl,this.gpgpu.textureConfig,o),u=z().get(\"WEBGL_DELETE_TEXTURE_THRESHOLD\");u!==-1&&this._numBytesAllocated>u?(this.gpgpu.deleteMatrixTexture(t.texture),this._numBytesAllocated-=a):(this.freeTextures[i].push(t),this.numFreeTextures++,this._numBytesFree+=a),this.numUsedTextures--;let l=this.usedTextures[i],c=l.indexOf(t);if(c<0)throw new Error(\"Cannot release a texture that was never provided by this texture manager\");l.splice(c,1),this.log()}log(){if(!this.logEnabled)return;let t=this.numFreeTextures+this.numUsedTextures;console.log(\"Free/Used\",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${t})`);let e=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*e)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let t in this.freeTextures)this.freeTextures[t].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});for(let t in this.usedTextures)this.usedTextures[t].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function net(r,t){let e=r;if(t===e.R32F)return 4;if(t===e.R16F)return 2;if(t===e.RGBA32F)return 16;if(t===r.RGBA)return 16;if(t===e.RGBA16F)return 8;if(t===e.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function tM(r,t,e,n,o){let s=oet(t,n),i;if(o){let[u,l]=Xi(r[0],r[1]);i=u*l}else{let[u,l]=Lc(r[0],r[1]);i=u*l}let a=net(e,s);return i*a}function oet(r,t){switch(r){case Pr.PACKED_2X2_FLOAT32:return Ow(t);case Pr.PACKED_2X2_FLOAT16:return Pw(t);case Pr.UNPACKED_FLOAT32:return Dw(t);case Pr.UNPACKED_FLOAT16:return Rw(t);case Pr.PACKED_4X1_UNSIGNED_BYTE:return Fw(t);default:throw new Error(`Unknown physical texture type ${r}`)}}function set(r){return z().getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\")?r?Pr.PACKED_2X2_FLOAT32:Pr.UNPACKED_FLOAT32:r?Pr.PACKED_2X2_FLOAT16:Pr.UNPACKED_FLOAT16}function eM(r,t){if(r===jr.UPLOAD)return Pr.PACKED_2X2_FLOAT32;if(r===jr.RENDER||r==null)return set(t);if(r===jr.DOWNLOAD||r===jr.PIXELS)return Pr.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${r}`)}function rM(r,t,e){return`${r[0]}_${r[1]}_${t}_${e}`}var tn=class{constructor(t,e){this.variableNames=[\"A\"],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`\n      float unaryOperation(float x) {\n        ${e}\n      }\n\n      void main() {\n        float x = getAAtOutCoords();\n        float y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}},fr=\"if (isnan(x)) return x;\",nM=\"return x;\",lk=\"return abs(x);\";var oM=\"return (x >= 0.0) ? x : (exp(x) - 1.0);\",sM=fr+`\n  return (x < 0.0) ? 0.0 : x;\n`,iM=fr+`\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n`,Gc=\"return x;\",aM=\"return 1.0 / (1.0 + exp(-1.0 * x));\";var uM=\"return x;\",cM=`\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n`,pM=`\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,mM=`\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,fM=\"return 1.0 / (1.0 + exp(-1.0 * x));\",so=class{constructor(t,e){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`\n      vec4 unaryOperation(vec4 x) {\n        ${e}\n      }\n\n      void main() {\n        vec4 x = getAAtOutCoords();\n        vec4 y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}};var Gw=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let e=t.length,n=Qe(\"rc\",e),o=zt(e),s=QL(e,n),i=n.slice(-2),a=e<=1?\"rc\":`vec2(${i.join(\",\")})`;this.userCode=`\n      void main() {\n        ${o} rc = getOutputCoords();\n        vec4 packedInput = getA(${s});\n\n        setOutput(getChannel(packedInput, ${a}));\n      }\n    `}};var aet=Ur.whereImpl,uet=1e-7,cet=1e-4,Ww={};function pet(r){return r in Ww||(Ww[r]={}),Ww[r]}var met=z().getNumber(\"CPU_HANDOFF_SIZE_THRESHOLD\"),fet=600;function det(){return z().global.screen==null?1024:z().global.screen.height*z().global.screen.width*window.devicePixelRatio*fet/1024/1024}var _u=class extends zo{constructor(t){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!z().getBool(\"HAS_WEBGL\"))throw new Error(\"WebGL is not supported on this device\");let e;if(t!=null){if(t instanceof Bc)e=t;else{let n=Gn(z().getNumber(\"WEBGL_VERSION\"),t);e=new Bc(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let n=Gn(z().getNumber(\"WEBGL_VERSION\"));e=new Bc(n),this.binaryCache=pet(z().getNumber(\"WEBGL_VERSION\")),this.gpgpuCreatedLocally=!0}this.gpgpu=e,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new Vw(this.gpgpu),this.numMBBeforeWarning=det(),this.texData=new ra(this,Pn())}nextDataId(){return _u.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(t,e,n,o,s,i){let a=this.makeTensorInfo(e,n),u=this.texData.get(a.dataId);u.isPacked=!1,u.texture={texture:t,texShape:[o,s]},u.texShape=[o,s];let l=fd(e),c=new Jh(l,!1,i),p=this.runWebGLProgram(c,[a],n,[[o,s]]);return p.shape=e,u.texture=null,this.disposeIntermediateTensorInfo(a),p.dataId}write(t,e,n){if((z().getBool(\"WEBGL_CHECK_NUMERICAL_PROBLEMS\")||z().getBool(\"DEBUG\"))&&this.checkNumericalProblems(t),n===\"complex64\"&&t!=null)throw new Error(\"Cannot write to a complex64 dtype. Please use tf.complex(real, imag).\");let o={id:this.nextDataId()};return this.texData.set(o,{shape:e,dtype:n,values:t,usage:jr.UPLOAD,refCount:1}),o}refCount(t){return this.texData.has(t)?this.texData.get(t).refCount:0}incRef(t){let e=this.texData.get(t);e.refCount++}decRef(t){if(this.texData.has(t)){let e=this.texData.get(t);e.refCount--}}move(t,e,n,o,s){if(z().getBool(\"DEBUG\")&&this.checkNumericalProblems(e),o===\"complex64\")throw new Error(\"Cannot write to a complex64 dtype. Please use tf.complex(real, imag).\");this.texData.set(t,{shape:n,dtype:o,values:e,usage:jr.UPLOAD,refCount:s})}disposeIntermediateTensorInfo(t){this.disposeData(t.dataId)}readSync(t){let e=this.texData.get(t),{values:n,dtype:o,complexTensorInfos:s,slice:i,shape:a,isPacked:u}=e;if(i!=null){let m;u?m=new so(a,Gc):m=new tn(a,Gc);let f=this.runWebGLProgram(m,[{dataId:t,shape:a,dtype:o}],o),d=this.readSync(f.dataId);return this.disposeIntermediateTensorInfo(f),d}if(n!=null)return this.convertAndCacheOnCPU(t);if(o===\"string\")return n;let l=this.activeTimers!=null,c;l&&(c=y.now());let p;if(o===\"complex64\"){let m=this.readSync(s.real.dataId),f=this.readSync(s.imag.dataId);p=v.mergeRealAndImagArrays(m,f)}else p=this.getValuesFromTexture(t);return l&&(this.downloadWaitMs+=y.now()-c),this.convertAndCacheOnCPU(t,p)}async read(t){if(this.pendingRead.has(t)){let d=this.pendingRead.get(t);return new Promise(h=>d.push(h))}let e=this.texData.get(t),{values:n,shape:o,slice:s,dtype:i,complexTensorInfos:a,isPacked:u}=e;if(s!=null){let d;u?d=new so(o,Gc):d=new tn(o,Gc);let h=this.runWebGLProgram(d,[{dataId:t,shape:o,dtype:i}],i),g=this.read(h.dataId);return this.disposeIntermediateTensorInfo(h),g}if(n!=null)return this.convertAndCacheOnCPU(t);if(z().getBool(\"DEBUG\")&&!z().getBool(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\")&&z().getNumber(\"WEBGL_VERSION\")===2)throw new Error(\"tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.\");let l=null,c;if(i!==\"complex64\"&&z().get(\"WEBGL_BUFFER_SUPPORTED\")){c=this.decode(t);let d=this.texData.get(c.dataId);l=this.gpgpu.createBufferFromTexture(d.texture.texture,...jh(o))}this.pendingRead.set(t,[]),i!==\"complex64\"&&await this.gpgpu.createAndWaitForFence();let p;if(i===\"complex64\"){let d=await Promise.all([this.read(a.real.dataId),this.read(a.imag.dataId)]),h=d[0],g=d[1];p=v.mergeRealAndImagArrays(h,g)}else if(l==null)p=this.getValuesFromTexture(t);else{let d=y.sizeFromShape(o);p=this.gpgpu.downloadFloat32MatrixFromBuffer(l,d)}if(c!=null&&this.disposeIntermediateTensorInfo(c),l!=null){let d=this.gpgpu.gl;yt(d,()=>d.deleteBuffer(l))}let m=this.convertAndCacheOnCPU(t,p),f=this.pendingRead.get(t);return this.pendingRead.delete(t),f.forEach(d=>d(m)),this.pendingDisposal.has(t)&&(this.pendingDisposal.delete(t),this.disposeData(t)&&Pn().removeDataId(t,this),this.pendingDeletes--),m}readToGPU(t,e={}){let n=this.texData.get(t),{values:o,shape:s,slice:i,dtype:a,isPacked:u,texture:l}=n;if(a===\"complex64\")throw new Error(\"Does not support reading texture for complex64 dtype.\");if(i!=null){let f;u?f=new so(s,Gc):f=new tn(s,Gc);let d=this.runWebGLProgram(f,[{dataId:t,shape:s,dtype:a}],a),h=this.readToGPU(d,e);return this.disposeIntermediateTensorInfo(d),h}if(l==null)throw o!=null?new Error(\"Data is not on GPU but on CPU.\"):new Error(\"There is no data on GPU or CPU.\");let c=this.decode(t,e.customTexShape),p=Pn().makeTensorFromTensorInfo(c),m=this.texData.get(c.dataId);return Object.assign({tensorRef:p},m.texture)}bufferSync(t){let e=this.readSync(t.dataId);if(t.dtype===\"string\")try{let n=e.map(o=>y.decodeString(o));return wt(t.shape,t.dtype,n)}catch(n){throw new Error(\"Failed to decode encoded string bytes into utf-8\")}return wt(t.shape,t.dtype,e)}checkNumericalProblems(t){if(t!=null)for(let e=0;e<t.length;e++){let n=t[e];if(!ST(n))throw z().getBool(\"WEBGL_RENDER_FLOAT32_CAPABLE\")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(t){let{shape:e,dtype:n,isPacked:o}=this.texData.get(t),s=y.sizeFromShape(e);if(z().getBool(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\")){let m=this.decode(t),f=this.texData.get(m.dataId),d=this.gpgpu.downloadMatrixFromPackedTexture(f.texture.texture,...jh(e)).subarray(0,s);return this.disposeIntermediateTensorInfo(m),d}let i=z().getBool(\"WEBGL_PACK\")&&o===!0,a=i?fd(e):e,u=i?new Aw(a):new _w(a),l=this.runWebGLProgram(u,[{shape:a,dtype:n,dataId:t}],\"float32\"),c=this.texData.get(l.dataId),p=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(c.texture.texture,c.texShape[0],c.texShape[1]).subarray(0,s);return this.disposeIntermediateTensorInfo(l),p}timerAvailable(){return z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\")>0}time(t){let e=this.activeTimers,n=[],o=!1;this.programTimersStack==null?(this.programTimersStack=n,o=!0):this.activeTimers.push(n),this.activeTimers=n,t();let s=y.flatten(this.activeTimers.map(u=>u.query)).filter(u=>u!=null),i=y.flatten(this.activeTimers.map(u=>u.name)).filter(u=>u!=null);this.activeTimers=e,o&&(this.programTimersStack=null);let a={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\")>0){let u=await Promise.all(s);a.kernelMs=y.sum(u),a.getExtraProfileInfo=()=>u.map((l,c)=>({name:i[c],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(\", \")}else a.kernelMs={error:\"WebGL query timers are not supported in this environment.\"};return this.uploadWaitMs=0,this.downloadWaitMs=0,a})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\")>0?this.gpgpu.beginQuery():{startMs:y.now(),endMs:null}}endTimer(t){return z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\")>0?(this.gpgpu.endQuery(),t):(t.endMs=y.now(),t)}async getQueryTime(t){if(z().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\")>0)return this.gpgpu.waitForQueryAndGetTime(t);let e=t;return e.endMs-e.startMs}disposeData(t,e=!1){if(this.pendingDisposal.has(t))return!1;if(!this.texData.has(t))return!0;if(e?this.texData.get(t).refCount=0:this.texData.get(t).refCount--,!e&&this.texData.get(t).refCount>0)return!1;if(this.pendingRead.has(t))return this.pendingDisposal.add(t),this.pendingDeletes++,!1;this.releaseGPUData(t);let{complexTensorInfos:n}=this.texData.get(t);return n!=null&&(this.disposeData(n.real.dataId,e),this.disposeData(n.imag.dataId,e)),this.texData.delete(t),!0}releaseGPUData(t){let{texture:e,dtype:n,texShape:o,usage:s,isPacked:i,slice:a}=this.texData.get(t),u=a&&a.origDataId||t,l=this.dataRefCount.get(u);l>1?this.dataRefCount.set(u,l-1):(this.dataRefCount.delete(u),e!=null&&(this.numBytesInGPU-=this.computeBytes(o,n),this.textureManager.releaseTexture(e,o,s,i)));let c=this.texData.get(t);c.texture=null,c.texShape=null,c.isPacked=!1,c.slice=null}getTexture(t){return this.uploadToGPU(t),this.texData.get(t).texture.texture}getDataInfo(t){return this.texData.get(t)}shouldExecuteOnCPU(t,e=met){return z().getBool(\"WEBGL_CPU_FORWARD\")&&t.every(n=>this.texData.get(n.dataId).texture==null&&y.sizeFromShape(n.shape)<e)}getGPGPUContext(){return this.gpgpu}where(t){v.warn(\"tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead\");let e=t.dataSync();return aet(t.shape,e)}packedUnaryOp(t,e,n){let o=new so(t.shape,e),s=this.compileAndRun(o,[t],n);return Pn().makeTensorFromTensorInfo(s)}abs(t){if(this.shouldExecuteOnCPU([t])&&t.dtype!==\"complex64\"){let o=Mw(this.texData.get(t.dataId).values);return this.makeOutput(t.shape,t.dtype,o)}if(z().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\"))return this.packedUnaryOp(t,lk,t.dtype);let e=new tn(t.shape,lk),n=this.compileAndRun(e,[t]);return Pn().makeTensorFromTensorInfo(n)}makeTensorInfo(t,e,n){let o;if(e===\"string\"&&n!=null&&n.length>0&&y.isString(n[0])){let s=n.map(i=>y.encodeString(i));o=this.write(s,t,e)}else o=this.write(n,t,e);return this.texData.get(o).usage=null,{dataId:o,shape:t,dtype:e}}makeOutput(t,e,n){return Pn().makeTensorFromTensorInfo(this.makeTensorInfo(t,e,n),this)}unpackTensor(t){let e=new Gw(t.shape);return this.runWebGLProgram(e,[t],t.dtype)}packTensor(t){let e=new Bw(t.shape),n=!0;return this.runWebGLProgram(e,[t],t.dtype,null,n)}packedReshape(t,e){let n=[xl(t.shape),...yl(t.shape)],o={dtype:t.dtype,shape:n,dataId:t.dataId},s=[xl(e),...yl(e)],i=new Id(s,n),a=!0,u=[n],l=this.runWebGLProgram(i,[o],t.dtype,u,a);return{dataId:l.dataId,shape:e,dtype:l.dtype}}decode(t,e){let n=this.texData.get(t),{isPacked:o,shape:s,dtype:i}=n;if(e!=null){let m=y.sizeFromShape(s),f=e[0]*e[1]*4;y.assert(m<=f,()=>\"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.\")}let a=fd(s),u;o?u=new Ew(a):u=new kw(a);let l=!0,c=[e!=null?e:jh(a)],p=this.runWebGLProgram(u,[{shape:a,dtype:i,dataId:t}],i,c,l,e);return{dtype:i,shape:s,dataId:p.dataId}}runWebGLProgram(t,e,n,o,s=!1,i){let a=this.makeTensorInfo(t.outputShape,n),u=this.texData.get(a.dataId);if(t.packedOutput&&(u.isPacked=!0),t.outPackingScheme===ku.DENSE){let x=i!=null?i:jh(t.outputShape);u.texShape=x.map(b=>b*2)}if(t.outTexUsage!=null&&(u.usage=t.outTexUsage),y.sizeFromShape(a.shape)===0)return u.values=y.getTypedArrayFromDType(a.dtype,0),a;let l=[],c=e.map(x=>{if(x.dtype===\"complex64\")throw new Error(\"GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.\");let b=this.texData.get(x.dataId);if(b.texture==null){if(!t.packedInputs&&y.sizeFromShape(x.shape)<=z().getNumber(\"WEBGL_SIZE_UPLOAD_UNIFORM\"))return{shape:x.shape,texData:null,isUniform:!0,uniformValues:b.values};t.packedInputs&&(b.isPacked=!0,b.shape=x.shape)}if(this.uploadToGPU(x.dataId),!!b.isPacked!=!!t.packedInputs)x=b.isPacked?this.unpackTensor(x):this.packTensor(x),l.push(x),b=this.texData.get(x.dataId);else if(b.isPacked&&!Eu(b.shape,x.shape)){let w=x,C=x.shape;x.shape=b.shape,x=this.packedReshape(x,C),l.push(x),b=this.texData.get(x.dataId),w.shape=C}return{shape:x.shape,texData:b,isUniform:!1}});this.uploadToGPU(a.dataId);let p={shape:a.shape,texData:u,isUniform:!1},m=uL(t,c,p),f=this.getAndSaveBinary(m,()=>aL(this.gpgpu,t,c,p)),d=this.activeTimers!=null,h;d&&(h=this.startTimer()),z().get(\"ENGINE_COMPILE_ONLY\")||lL(this.gpgpu,f,c,p,o),l.forEach(x=>this.disposeIntermediateTensorInfo(x)),d&&(h=this.endTimer(h),this.activeTimers.push({name:t.constructor.name,query:this.getQueryTime(h)}));let g=z().get(\"WEBGL_FLUSH_THRESHOLD\");if(g>0){let x=y.now();x-this.lastGlFlushTime>g&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=x)}if(!z().getBool(\"WEBGL_LAZILY_UNPACK\")&&u.isPacked&&s===!1){let x=this.unpackTensor(a);return this.disposeIntermediateTensorInfo(a),x}return a}compileAndRun(t,e,n,o,s=!1){return n=n||e[0].dtype,this.runWebGLProgram(t,e,n,o,s)}getAndSaveBinary(t,e){return t in this.binaryCache||(this.binaryCache[t]=e()),this.binaryCache[t]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(z().getBool(\"IS_TEST\")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!=\"undefined\"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=B(()=>{if(!z().get(\"WEBGL_RENDER_FLOAT32_ENABLED\")){let t=z().getBool(\"DEBUG\");z().set(\"DEBUG\",!1);let e=this.abs(mt(1e-8)).dataSync()[0];if(z().set(\"DEBUG\",t),e>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?uet:cet}uploadToGPU(t){let e=this.texData.get(t),{shape:n,dtype:o,values:s,texture:i,usage:a,isPacked:u}=e;if(i!=null)return;let l=this.activeTimers!=null,c;l&&(c=y.now());let p=e.texShape;if(p==null&&(p=PT(n,u),e.texShape=p),s!=null){let m=fd(n),f,d=p[1],h=p[0],g=s instanceof Uint8Array||s instanceof Uint8ClampedArray;(u||!g)&&([d,h]=Xi(p[0],p[1])),u?f=new $w(m,g):f=new Jh(m,g);let x=g?[h,d]:p,b=this.makeTensorInfo(x,o),w=this.texData.get(b.dataId);g?w.usage=jr.PIXELS:w.usage=jr.UPLOAD,w.texShape=x,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(b.dataId),d,h,s);let C=[[h,d]],N=!0,_=this.runWebGLProgram(f,[b],o,C,N),A=this.texData.get(_.dataId);e.texShape=A.texShape,e.isPacked=A.isPacked,e.usage=A.usage,z().get(\"ENGINE_COMPILE_ONLY\")?this.disposeData(_.dataId):(e.texture=A.texture,e.values=null,this.texData.delete(_.dataId)),this.disposeIntermediateTensorInfo(b),l&&(this.uploadWaitMs+=y.now()-c)}else{let m=this.acquireTexture(p,a,o,u);e.texture=m}}convertAndCacheOnCPU(t,e){let n=this.texData.get(t),{dtype:o}=n;return this.releaseGPUData(t),e!=null&&(n.values=het(e,o)),n.values}acquireTexture(t,e,n,o){if(this.numBytesInGPU+=this.computeBytes(t,n),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let s=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${s} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(t,e,o)}computeBytes(t,e){return t[0]*t[1]*y.bytesPerElement(e)}checkCompileCompletion(){for(let[,t]of Object.entries(this.binaryCache))this.checkCompletion_(t)}async checkCompileCompletionAsync(){let t=[];if(this.gpgpu.parallelCompilationExtension){for(let[,e]of Object.entries(this.binaryCache))t.push(this.checkCompletionAsync_(e));return Promise.all(t)}else{for(let[,e]of Object.entries(this.binaryCache)){let n=new Promise(o=>{try{this.checkCompletion_(e),o(!0)}catch(s){throw s}});t.push(n)}return Promise.all(t)}}async checkCompletionAsync_(t){return this.gpgpu.gl.getProgramParameter(t.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(t):(await gh(),this.checkCompletionAsync_(t))}checkCompletion_(t){if(this.gpgpu.gl.getProgramParameter(t.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(t.webGLProgram)),this.gpgpu.gl.getShaderParameter(t.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(Cw(t.source,this.gpgpu.gl.getShaderInfoLog(t.fragmentShader)),new Error(\"Failed to compile fragment shader.\")):new Error(\"Failed to link vertex and fragment shaders.\");return!0}getUniformLocations(){for(let[,t]of Object.entries(this.binaryCache)){let{uniformLocations:e,customUniformLocations:n,infLoc:o,nanLoc:s,inShapesLocations:i,inTexShapesLocations:a,outShapeLocation:u,outShapeStridesLocation:l,outTexShapeLocation:c}=WT(this.gpgpu,t.program,t.webGLProgram);t.uniformLocations=e,t.customUniformLocations=n,t.infLoc=o,t.nanLoc=s,t.inShapesLocations=i,t.inTexShapesLocations=a,t.outShapeLocation=u,t.outShapeStridesLocation=l,t.outTexShapeLocation=c}}createTensorFromTexture(t,e,n){let{texture:o,height:s,width:i,channels:a}=t,u=Pn().backend;if(!u.gpgpu.gl.isTexture(o))throw new Error(\"The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.\");let l=u.writeTexture(o,e,n,s,i,a);return Pn().makeTensorFromDataId(l,e,n,u)}};_u.nextDataId=0;function het(r,t){if(t===\"float32\"||t===\"complex64\")return r;if(t===\"int32\"||t===\"bool\"){let e=t===\"int32\"?new Int32Array(r.length):new Uint8Array(r.length);for(let n=0;n<e.length;++n)e[n]=Math.round(r[n]);return e}else throw new Error(`Unknown dtype ${t}`)}var dM=\"4.0.0\";function hM(){z().set(\"WEBGL_FORCE_F16_TEXTURES\",!0)}Kl.isBrowser()&&Xp(\"webgl\",()=>new _u,2);var Zke={forceHalfFloat:hM};var Sd=`\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n`;var io=class{constructor(t,e,n){this.variableNames=[\"A\",\"B\"],this.outputShape=v.assertAndGetBroadcastShape(e,n),this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`\n      float binaryOperation(float a, float b) {\n        ${t}\n      }\n\n      void main() {\n        float a = getAAtOutCoords();\n        float b = getBAtOutCoords();\n        setOutput(binaryOperation(a, b));\n      }\n    `}};var Yi=`\n  result.r = isNaN.r ? NAN : result.r;\n  result.g = isNaN.g ? NAN : result.g;\n  result.b = isNaN.b ? NAN : result.b;\n  result.a = isNaN.a ? NAN : result.a;\n`;var Oo=class{constructor(t,e,n,o=!1){this.variableNames=[\"A\",\"B\"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=v.assertAndGetBroadcastShape(e,n);let s=this.outputShape.length;this.enableShapeUniforms=we(s);let i=\"\";if(o)if(s===0||y.sizeFromShape(this.outputShape)===1)i=`\n          result.y = 0.;\n          result.z = 0.;\n          result.w = 0.;\n        `;else if(i=`\n          ${zt(s)} coords = getOutputCoords();\n        `,s===1)this.enableShapeUniforms?i+=`\n            result.y = (coords + 1) >= outShape ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `:i+=`\n            result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;else{let u=Qe(\"coords\",s);this.enableShapeUniforms?i+=`\n            bool nextRowOutOfBounds =\n              (${u[s-2]} + 1) >= outShape[${s} - 2];\n            bool nextColOutOfBounds =\n              (${u[s-1]} + 1) >= outShape[${s} - 1];\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `:i+=`\n            bool nextRowOutOfBounds =\n              (${u[s-2]} + 1) >= ${this.outputShape[s-2]};\n            bool nextColOutOfBounds =\n              (${u[s-1]} + 1) >= ${this.outputShape[s-1]};\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `}this.userCode=`\n      vec4 binaryOperation(vec4 a, vec4 b) {\n        ${t}\n      }\n\n      void main() {\n        vec4 a = getAAtOutCoords();\n        vec4 b = getBAtOutCoords();\n\n        vec4 result = binaryOperation(a, b);\n        ${i}\n\n        setOutput(result);\n      }\n    `}};function tr(r){let{inputs:t,backend:e}=r,{x:n}=t;return e.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var gM={kernelName:co,backendName:\"webgl\",kernelFunc:tr};function En(r){let{inputs:t,backend:e}=r,{real:n,imag:o}=t,s=e.makeTensorInfo(n.shape,\"complex64\"),i=e.texData.get(s.dataId),a=tr({inputs:{x:n},backend:e}),u=tr({inputs:{x:o},backend:e});return i.complexTensorInfos={real:a,imag:u},s}var xM={kernelName:pp,backendName:\"webgl\",kernelFunc:En};var uk=\"return (a < 0.) ? b * a : a;\",ck=`\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n`;function get(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{alpha:s}=n,i=e.makeTensorInfo([],\"float32\",y.createScalarValue(s,\"float32\")),a=z().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")?new Oo(ck,o.shape,i.shape):new io(uk,o.shape,i.shape),u=e.runWebGLProgram(a,[o,i],\"float32\");return e.disposeIntermediateTensorInfo(i),u}var yM={kernelName:is,backendName:\"webgl\",kernelFunc:get};var pk=\"return (a < 0.) ? b * a : a;\",mk=`\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n`;function xet(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t,s=z().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")?new Oo(mk,n.shape,o.shape):new io(pk,n.shape,o.shape);return e.runWebGLProgram(s,[n,o],\"float32\")}var bM={kernelName:bs,backendName:\"webgl\",kernelFunc:xet};var Po=\"if (isnan(x)) return x;\";function Ct({opSnippet:r,packedOpSnippet:t,cpuKernelImpl:e,dtype:n}){return({inputs:o,backend:s})=>{let{x:i}=o,a=s,u=n||i.dtype;if(a.shouldExecuteOnCPU([i])&&e!=null){let p=a.texData.get(i.dataId),m=e(p.values,u);return a.makeTensorInfo(i.shape,u,m)}let l=z().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")&&t!=null,c;return l?c=new so(i.shape,t):c=new tn(i.shape,r),a.runWebGLProgram(c,[i],u)}}function le({opSnippet:r,packedOpSnippet:t,checkOutOfBounds:e=!1,supportsComplex:n=!1,cpuKernelImpl:o,dtype:s}){return({inputs:i,backend:a})=>{let{a:u,b:l}=i,c=a;if(n&&u.dtype===\"complex64\"){let d=c.texData.get(u.dataId),h=c.texData.get(l.dataId),[g,x]=[[d.complexTensorInfos.real,h.complexTensorInfos.real],[d.complexTensorInfos.imag,h.complexTensorInfos.imag]].map(w=>{let[C,N]=w,_={dataId:C.dataId,dtype:C.dtype,shape:u.shape},A={dataId:N.dataId,dtype:N.dtype,shape:l.shape},$=new io(r,u.shape,l.shape);return c.runWebGLProgram($,[_,A],sr(C.dtype,N.dtype))}),b=En({inputs:{real:g,imag:x},backend:c});return c.disposeIntermediateTensorInfo(g),c.disposeIntermediateTensorInfo(x),b}let p=s||sr(u.dtype,l.dtype);if((u.dtype===\"string\"||l.dtype===\"string\"||c.shouldExecuteOnCPU([u,l]))&&o!=null){let d=c.texData.get(u.dataId).values,h=c.texData.get(l.dataId).values,g=u.dtype===\"string\"?v.fromUint8ToStringArray(d):d,x=u.dtype===\"string\"?v.fromUint8ToStringArray(h):h,[b,w]=o(u.shape,l.shape,g,x,p),C=c.makeTensorInfo(w,p),N=c.texData.get(C.dataId);return N.values=b,C}let m=z().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")&&t!=null,f;return m?f=new Oo(t,u.shape,l.shape,e):f=new io(r,u.shape,l.shape),c.runWebGLProgram(f,[u,l],p)}}function bl(r,t=!1){if(r===\"linear\")return t?uM:nM;if(r===\"relu\")return t?pM:sM;if(r===\"elu\")return t?cM:oM;if(r===\"relu6\")return t?mM:iM;if(r===\"prelu\")return t?mk:pk;if(r===\"leakyrelu\")return t?ck:uk;if(r===\"sigmoid\")return t?fM:aM;throw new Error(`Activation ${r} has not been implemented for the WebGL backend.`)}var vd=class{constructor(t,e,n,o=!1,s=!1,i=!1,a=null,u=!1,l=!1){this.variableNames=[\"matrixA\",\"matrixB\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=we(this.outputShape.length);let c=o?t[1]:t[2],p=Math.ceil(c/2),m=o?\"i * 2, rc.y\":\"rc.y, i * 2\",f=s?\"rc.z, i * 2\":\"i * 2, rc.z\",d=o?[\"a.xxyy\",\"a.zzww\"]:[\"a.xxzz\",\"a.yyww\"],h=s?[\"b.xzxz\",\"b.ywyw\"]:[\"b.xyxy\",\"b.zwzw\"],g=\"\",x=\"\";a&&(u?g=`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${a}\n        }`:l?g=`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${a}\n        }`:g=`vec4 activation(vec4 x) {\n          ${a}\n        }`,x=\"result = activation(result);\");let b=i?\"result += getBiasAtOutCoords();\":\"\";i&&this.variableNames.push(\"bias\"),u&&this.variableNames.push(\"preluActivationWeights\"),l&&this.variableNames.push(\"leakyreluAlpha\");let w=\"rc.x\",C=\"rc.x\";t[0]<e[0]?w=`int(min(float(rc.x), ${t[0]-1}.))`:e[0]<t[0]&&(C=`int(min(float(rc.x), ${e[0]-1}.))`),this.userCode=`\n      ${g}\n      // Don't use uniform for sharedDimensionPacked for performance.\n      const float sharedDimension = ${p}.0;\n\n      vec4 dot2x2ARowBCol(ivec3 rc) {\n        vec4 result = vec4(0);\n        for (int i = 0; i < ${p}; i++) {\n          int batchA = ${w};\n          int batchB = ${C};\n          vec4 a = getMatrixA(batchA, ${m});\n          vec4 b = getMatrixB(batchB, ${f});\n\n          // These swizzled products need to be separately added.\n          // See: https://github.com/tensorflow/tfjs/issues/1735\n          result += (${d[0]} * ${h[0]});\n          result += (${d[1]} * ${h[1]});\n        }\n        return result;\n      }\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n        vec4 result = dot2x2ARowBCol(rc);\n\n        ${b}\n\n        ${x}\n\n        setOutput(result);\n      }\n    `}};var fk={REAL:\"return areal * breal - aimag * bimag;\",IMAG:\"return areal * bimag + aimag * breal;\"},tg=class{constructor(t,e,n){this.variableNames=[\"AReal\",\"AImag\",\"BReal\",\"BImag\"],this.outputShape=v.assertAndGetBroadcastShape(e,n),this.userCode=`\n      float binaryOpComplex(\n          float areal, float aimag, float breal, float bimag) {\n        ${t}\n      }\n\n      void main() {\n        float areal = getARealAtOutCoords();\n        float aimag = getAImagAtOutCoords();\n        float breal = getBRealAtOutCoords();\n        float bimag = getBImagAtOutCoords();\n        setOutput(binaryOpComplex(areal, aimag, breal, bimag));\n      }\n    `}};var wM=\"return a * b;\";function eg(r){let{inputs:t,backend:e}=r,{a:n,b:o}=t,s=v.upcastType(n.dtype,o.dtype);if(n.dtype===\"complex64\"){let a=e.texData.get(n.dataId),u=e.texData.get(o.dataId),l=new tg(fk.REAL,n.shape,o.shape),c=new tg(fk.IMAG,n.shape,o.shape),p=[{dataId:a.complexTensorInfos.real.dataId,dtype:a.complexTensorInfos.real.dtype,shape:n.shape},{dataId:a.complexTensorInfos.imag.dataId,dtype:a.complexTensorInfos.imag.dtype,shape:n.shape},{dataId:u.complexTensorInfos.real.dataId,dtype:u.complexTensorInfos.real.dtype,shape:o.shape},{dataId:u.complexTensorInfos.imag.dataId,dtype:u.complexTensorInfos.imag.dtype,shape:o.shape}],m=e.runWebGLProgram(l,p,\"float32\"),f=e.runWebGLProgram(c,p,\"float32\"),d=En({inputs:{real:m,imag:f},backend:e});return e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),d}if(e.shouldExecuteOnCPU([n,o])){let a=e.texData.get(n.dataId),u=e.texData.get(o.dataId),[l,c]=AL(n.shape,o.shape,a.values,u.values,s),p=e.makeTensorInfo(c,s),m=e.texData.get(p.dataId);return m.values=l,p}let i;return z().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")?i=new Oo(wM,n.shape,o.shape):i=new io(wM,n.shape,o.shape),e.runWebGLProgram(i,[n,o],s)}var CM={kernelName:hs,backendName:\"webgl\",kernelFunc:eg};function IM(r,t,e){let n=[xl(r.shape),...yl(r.shape)],o={dtype:r.dtype,shape:n,dataId:r.dataId},s=[xl(t),...yl(t)],i=new Id(s,n),a=!0,u=[n],l=e.runWebGLProgram(i,[o],r.dtype,u,a);return{dataId:l.dataId,shape:t,dtype:l.dtype}}function st(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{shape:s}=n,i=e,a=y.sizeFromShape(o.shape),u=y.inferFromImplicitShape(s,a),l=y.sizeFromShape(u);y.assert(a===l,()=>`The new shape (${u}) has ${l} elements and the old shape (${o.shape}) has ${a} elements. The new shape and old shape must have the same number of elements.`);let c=i.texData.get(o.dataId);return c.isPacked&&!Eu(o.shape,u)&&!(c.texture!==null&&Eu(c.shape,u))?IM(o,u,i):(i.incRef(o.dataId),{dataId:o.dataId,shape:u,dtype:o.dtype})}var SM={kernelName:di,backendName:\"webgl\",kernelFunc:st};var rg=class{constructor(t,e){this.variableNames=[\"x\"];let{windowSize:n,batchSize:o,inSize:s,outSize:i}=t;this.outputShape=[o,i];let a=Math.floor(n/4)*4,u=n%4,l=\"sumValue += dot(values, ones);\";if(e!=null){let p=1/e;l=`sumValue += dot(values * ${y.isInt(p)?p.toPrecision(2):p}, ones);`}let c=\"\";s%n>0&&(c=`\n        if (inIdx < 0 || inIdx >= ${s}) {\n          return 0.0;\n        }\n      `),this.userCode=`\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${c}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${a}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${l}\n        }\n\n        int inIdx = inOffset + ${a};\n        if (${u===1}) {\n          vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);\n\n          ${l}\n        } else if (${u===2}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1), 0.0, 0.0);\n\n          ${l}\n        } else if (${u===3}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2), 0.0);\n\n          ${l}\n        }\n        setOutput(sumValue);\n      }\n    `}};var Uw=class{constructor(t,e){this.variableNames=[\"x\"];let{windowSize:n,batchSize:o,inSize:s,outSize:i}=t;this.outputShape=[o,i];let a=\"0.0\",u=\"\";e===\"prod\"?a=\"1.0\":e===\"min\"?(a=\"1.0 / 1e-20\",u=\"min\"):e===\"max\"&&(a=\"-1.0 / 1e-20\",u=\"max\");let l=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e===\"sum\"?l=\"sumValue\":e===\"prod\"?l=\"prodValue\":e===\"all\"?l=\"allValue\":e===\"any\"&&(l=\"anyValue\");let c=Math.floor(n/4)*4,p=n%4,m=`\n      if (${e===\"sum\"}) {\n        sumValue += dot(values, ones);\n      } else if (${e===\"prod\"}) {\n        vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);\n        prodValue *= tmp[0] * tmp[1];\n      } else {\n        minMaxValue = ${u}(values, minMaxValue);\n        if (${e===\"min\"} || ${e===\"max\"}) {\n          minMaxValue = ${u}(values, minMaxValue);\n          bvec4 isNaN = isnan(values);\n          if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {\n            minMaxValue = vec4(NAN);\n          }\n        }\n      }\n    `,f=\"vec4\";e===\"all\"?(a=\"1.0\",m=`\n        bool reducedAllValue = all(values);\n        float floatedReducedAllValue = float(reducedAllValue);\n        allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);\n      `,f=\"bvec4\"):e===\"any\"&&(a=\"0.0\",m=`\n        bool reducedAnyValue = any(values);\n        float floatedReducedAnyValue = float(reducedAnyValue);\n        anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);\n      `,f=\"bvec4\");let d=\"\";s%n>0&&(d=`\n        if (inIdx < 0 || inIdx >= ${s}) {\n          return initializationValue;\n        }\n      `),this.userCode=`\n      const float initializationValue = ${a};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${d}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        vec4 minMaxValue = vec4(${a});\n        float prodValue = 1.0;\n        float sumValue = 0.0;\n        float allValue = 1.0;\n        float anyValue = 0.0;\n\n        for (int i = 0; i < ${c}; i += 4) {\n          int inIdx = inOffset + i;\n          ${f} values = ${f}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${m}\n        }\n\n        int inIdx = inOffset + ${c};\n        if (${p===1}) {\n          ${f} values = ${f}(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          ${m}\n        } else if (${p===2}) {\n          ${f} values = ${f}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          ${m}\n        } else if (${p===3}) {\n          ${f} values = ${f}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          ${m}\n        }\n        setOutput(${l});\n      }\n    `}};function bet(r){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let e=t.length?t[t.length-1].outSize:r[1],n=v.computeOptimalWindowSize(e);t.push({inSize:e,windowSize:n,outSize:Math.ceil(e/n)})}return t}function Un(r,t,e,n){let o=bet(r.shape),s=r;for(let i=0;i<o.length;i++){let{inSize:a,windowSize:u,outSize:l}=o[i],c,p;e===\"mean\"?c=i===0?new rg({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l},a):new rg({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l}):c=new Uw({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l},e),p=s,s=n.runWebGLProgram(c,[s],t),p.dataId!==r.dataId&&n.disposeIntermediateTensorInfo(p)}return s}var Hw=class{constructor(t,e){this.variableNames=[\"A\"];let n=new Array(t.length);for(let i=0;i<n.length;i++)n[i]=t[e[i]];this.outputShape=n,this.rank=n.length;let o=zt(this.rank),s=wet(e);this.userCode=`\n    void main() {\n      ${o} resRC = getOutputCoords();\n      setOutput(getA(${s}));\n    }\n    `}};function wet(r){let t=r.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let e=[\"resRC.x\",\"resRC.y\",\"resRC.z\",\"resRC.w\",\"resRC.u\",\"resRC.v\"],n=new Array(t);for(let o=0;o<r.length;o++)n[r[o]]=e[o];return n.join()}var qw=class{constructor(t,e){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(t.length);for(let c=0;c<n.length;c++)n[c]=t[e[c]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let o=zt(this.rank),s=ak(\"rc\",this.rank),i=new Array(this.rank);for(let c=0;c<e.length;c++)i[e[c]]=s[c];let a=`vec2(${i.slice(-2).join()})`,u=`++${s[this.rank-1]} < ${n[this.rank-1]}`,l=`getChannel(getA(${i.join()}), ${a})`;this.userCode=`\n    void main() {\n      ${o} rc = getOutputCoords();\n      vec4 result = vec4(0.);\n      result[0] = ${l};\n      if(${u}) {\n        result[1] = ${l};\n      }\n      --${s[this.rank-1]};\n      if(++${s[this.rank-2]} < ${n[this.rank-2]}) {\n        result[2] = ${l};\n        if(${u}) {\n          result[3] = ${l};\n        }\n      }\n      setOutput(result);\n    }\n    `}};function Au(r,t,e){let n=z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")?new qw(r.shape,t):new Hw(r.shape,t);return e.runWebGLProgram(n,[r],r.dtype)}function vM(r,t,e,n){let o=t,s=r.shape.length,i=y.parseAxisParam(o,r.shape),a=i,u=v.getAxesPermutation(a,s),l=u!=null,c=r;l&&(c=Au(r,u,n),a=v.getInnerMostAxes(a.length,s)),v.assertAxesAreInnerMostDims(\"sum\",a,s);let[p,m]=v.computeOutAndReduceShapes(c.shape,a),f=p;e&&(f=v.expandShapeToKeepDim(p,i));let d=y.sizeFromShape(m),g=y.sizeFromShape(r.shape)/d,x=st({inputs:{x:c},attrs:{shape:[g,d]},backend:n}),b=Wu(r.dtype),w=Un(x,b,\"sum\",n),C=st({inputs:{x:w},attrs:{shape:f},backend:n});return n.disposeIntermediateTensorInfo(x),n.disposeIntermediateTensorInfo(w),l&&n.disposeIntermediateTensorInfo(c),C}function Wc(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;return vM(o,s,i,e)}var NM={kernelName:$s,backendName:\"webgl\",kernelFunc:Wc};function Oe(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{perm:s}=n,i=e,a=o.shape.length,u=new Array(a);for(let c=0;c<u.length;c++)u[c]=o.shape[s[c]];let l;if(i.shouldExecuteOnCPU([o])){let p=i.texData.get(o.dataId).values,m=Vc(p,o.shape,o.dtype,s,u);l=i.makeTensorInfo(u,o.dtype);let f=i.texData.get(l.dataId);f.values=m}else l=Au(o,s,i);return l}var TM={kernelName:Qn,backendName:\"webgl\",kernelFunc:Oe};var dk=1e3;function Uc({a:r,b:t,transposeA:e,transposeB:n,backend:o,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:a=0,activation:u=null}){let l=r.shape.length,c=t.shape.length,p=e?r.shape[l-2]:r.shape[l-1],m=n?t.shape[c-1]:t.shape[c-2],f=e?r.shape[l-1]:r.shape[l-2],d=n?t.shape[c-2]:t.shape[c-1],h=r.shape.slice(0,-2),g=t.shape.slice(0,-2),x=y.sizeFromShape(h),b=y.sizeFromShape(g),C=Vr.assertAndGetBroadcastShape(r.shape.slice(0,-2),t.shape.slice(0,-2)).concat([f,d]);y.assert(p===m,()=>`Error in matMul: inner shapes (${p}) and (${m}) of Tensors with shapes ${r.shape} and ${t.shape} and transposeA=${e} and transposeB=${n} must match.`);let N=e?[x,p,f]:[x,f,p],_=n?[b,d,m]:[b,m,d],A=st({inputs:{x:r},backend:o,attrs:{shape:N}}),$=st({inputs:{x:t},backend:o,attrs:{shape:_}}),F=[A,$],P=Math.max(x,b),V=e?A.shape[1]:A.shape[2],G=s!=null,W=i!=null,q=u===\"leakyrelu\",H=u!=null?bl(u,!0):null,j=G||W||q||H!=null,Y;if((f===1||d===1)&&V>dk&&j===!1){let et=A,rt=$;e&&(et=Oe({inputs:{x:A},backend:o,attrs:{perm:[0,2,1]}}),F.push(et)),n&&(rt=Oe({inputs:{x:$},backend:o,attrs:{perm:[0,2,1]}}),F.push(rt));let ot=d!==1,at=d===1,nt=et;ot&&(nt=st({inputs:{x:et},backend:o,attrs:{shape:[P,V,1]}}),F.push(nt));let it=d===1?2:1,dt=rt;at&&(dt=st({inputs:{x:rt},backend:o,attrs:{shape:[P,1,V]}}),F.push(dt));let ht=eg({inputs:{a:nt,b:dt},backend:o});Y=Wc({inputs:{x:ht},backend:o,attrs:{axis:it,keepDims:!0}}),F.push(ht)}else{let et=sr(r.dtype,t.dtype),rt=new vd(N,_,[P,f,d],e,n,G,H,W,q),ot=[A,$];if(s!=null&&ot.push(s),W&&ot.push(i),q){let at=o.makeTensorInfo([],\"float32\",y.createScalarValue(a,\"float32\"));ot.push(at),F.push(at)}Y=o.runWebGLProgram(rt,ot,et)}let Z=st({inputs:{x:Y},backend:o,attrs:{shape:C}});F.push(Y);for(let et of F)o.disposeIntermediateTensorInfo(et);return Z}function Cet(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n;return Uc({a:o,b:s,transposeA:u,transposeB:l,backend:e,bias:i,preluActivationWeights:a,leakyreluAlpha:p,activation:c})}var kM={kernelName:Ci,backendName:\"webgl\",kernelFunc:Cet};var EM=\"return abs(x);\";function Iet(r){let{inputs:t,backend:e}=r,{x:n}=t;if(e.shouldExecuteOnCPU([n])&&n.dtype!==\"complex64\"){let s=e.texData.get(n.dataId),i=Mw(s.values);return e.makeTensorInfo(n.shape,n.dtype,i)}let o;return z().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")?o=new so(n.shape,EM):o=new tn(n.shape,EM),e.runWebGLProgram(o,[n],n.dtype)}var _M={kernelName:ii,backendName:\"webgl\",kernelFunc:Iet};var vet=fr+`\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return acos(x);\n`,Net=Ct({opSnippet:vet}),AM={kernelName:oa,backendName:\"webgl\",kernelFunc:Net};var Tet=fr+`\n  if (x < 1.0) return NAN;\nreturn log(x + sqrt(x * x - 1.0));`,ket=Ct({opSnippet:Tet}),$M={kernelName:sa,backendName:\"webgl\",kernelFunc:ket};var DM=\"return a + b;\",Eet=le({opSnippet:DM,packedOpSnippet:DM,supportsComplex:!0,cpuKernelImpl:cL}),RM={kernelName:Zn,backendName:\"webgl\",kernelFunc:Eet};var Kw=class{constructor(t,e){this.outputShape=[],this.outputShape=t,this.variableNames=e.map((s,i)=>`T${i}`);let n=[];this.variableNames.forEach(s=>{n.push(`float v${s} = get${s}AtOutCoords();`)});let o=this.variableNames.map(s=>`v${s}`).join(\" + \");this.userCode=`\n      void main() {\n        ${n.join(`\n        `)}\n\n        float result = ${o};\n        setOutput(result);\n      }\n    `}};var jw=class{constructor(t,e){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.variableNames=e.map((s,i)=>`T${i}`);let n=[];this.variableNames.forEach(s=>{n.push(`vec4 v${s} = get${s}AtOutCoords();`)});let o=this.variableNames.map(s=>`v${s}`).join(\" + \");this.userCode=`\n      void main() {\n        ${n.join(`\n        `)}\n\n        vec4 result = ${o};\n        setOutput(result);\n      }\n    `}};function Xw(r){let{inputs:t,backend:e}=r,n=t;if(n.length===1)return tr({inputs:{x:n[0]},backend:e});if(n.length>z().get(\"WEBGL_MAX_TEXTURES_IN_SHADER\")){let u=Math.floor(n.length/2),l=Xw({inputs:n.slice(0,u),backend:e}),c=Xw({inputs:n.slice(u),backend:e});return Xw({inputs:[l,c],backend:e})}let o=n.map(u=>u.dtype).reduce((u,l)=>sr(u,l)),s=n.map(u=>u.shape),a=z().getBool(\"WEBGL_PACK\")?new jw(n[0].shape,s):new Kw(n[0].shape,s);return e.runWebGLProgram(a,n,o)}var FM={kernelName:Go,backendName:\"webgl\",kernelFunc:Xw};function _et(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,a)),v.assertAxesAreInnerMostDims(\"all\",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,\"all\",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var OM={kernelName:ia,backendName:\"webgl\",kernelFunc:_et};function Aet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,a)),v.assertAxesAreInnerMostDims(\"any\",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,\"any\",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var PM={kernelName:aa,backendName:\"webgl\",kernelFunc:Aet};var Yw=class{constructor(t,e,n){this.variableNames=[\"A\"];let{windowSize:o,batchSize:s,outSize:i}=t;n||this.variableNames.push(\"bestIndicesA\"),this.outputShape=[s,i];let a=e===\"max\"?\">\":\"<\",u=n?\"inOffset + i;\":\"round(getBestIndicesA(batch, inOffset + i));\";this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${o};\n\n        int bestIndex = inOffset;\n        float bestValue = getA(batch, bestIndex);\n\n        for (int i = 0; i < ${o}; i++) {\n          int inIdx = ${u};\n          float candidate = getA(batch, inIdx);\n          if (candidate ${a} bestValue) {\n            bestValue = candidate;\n            bestIndex = inIdx;\n          }\n        }\n        setOutput(float(bestIndex));\n      }\n    `}};var Zw=class{constructor(t,e,n,o){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,y.assert(t.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let s=t[t.length-1],i=Math.ceil(s/e);this.outputShape=t.slice(0,-1),i>1&&this.outputShape.push(i),o||this.variableNames.push(\"bestIndicesA\");let a=this.outputShape,u=a.length,l=zt(u),c=Qe(\"coords\",u),p,m;if(i===1){m=u+1;let $=zt(m);p=`\n        ${$} sourceLocR = ${$}(${c.join()}, 0);\n        ++${c[u-1]};\n        ${$} sourceLocG = ${$}(${c.join()}, 0);\n        ++${c[u-2]};\n        ${$} sourceLocA = ${$}(${c.join()}, 0);\n        --${c[u-1]};\n        ${$} sourceLocB = ${$}(${c.join()}, 0);\n        --${c[u-2]};`}else m=u,p=`\n        ${l} sourceLocR = coords;\n        ++${c[u-1]};\n        ${l} sourceLocG = coords;\n        ++${c[u-2]};\n        ${l} sourceLocA = coords;\n        --${c[u-1]};\n        ${l} sourceLocB = coords;\n        --${c[u-2]};`;let f=[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"].slice(0,m),d=\".\"+f[m-1],h=f.map($=>\"int \"+$),g=Qe(\"sourceLocR\",m-1).concat(\"inIdx.r\"),x=Qe(\"sourceLocG\",m-1).concat(\"inIdx.g\"),b=Qe(\"sourceLocB\",m-1).concat(\"inIdx.b\"),w=Qe(\"sourceLocA\",m-1).concat(\"inIdx.a\"),C=n===\"max\"?\"greaterThan\":\"lessThan\",N=o?\"\":`\n          inIdx = round(vec4(getBestIndicesAChannel(${g.join()}),\n                             getBestIndicesAChannel(${x.join()}),\n                             getBestIndicesAChannel(${b.join()}),\n                             getBestIndicesAChannel(${w.join()})));`,_=`vec4(\n            getAChannel(${g.join()}),\n            hasNextCol ? getAChannel(${x.join()}) : 0.,\n            hasNextRow ? getAChannel(${b.join()}) : 0.,\n            hasNextRow && hasNextCol ? getAChannel(${w.join()}) : 0.)`,A=o?\"\":`\n      float getBestIndicesAChannel(${h.join()}) {\n        return getChannel(getBestIndicesA(${f.join()}),\n                                          vec2(${f.slice(-2).join()}));\n      }`;this.userCode=`\n      float getAChannel(${h.join()}) {\n        return getChannel(getA(${f.join()}),\n                               vec2(${f.slice(-2).join()}));\n      }\n      ${A}\n      void main() {\n        ${l} coords = getOutputCoords();\n        bool hasNextCol = ${c[u-1]} < ${a[u-1]-1};\n        bool hasNextRow = ${c[u-2]} < ${a[u-2]-1};\n        ${p}\n        ivec4 srcIdx = ivec4(sourceLocR${d}, sourceLocG${d},\n          sourceLocB${d}, sourceLocA${d}) * ${e};\n        ivec4 inIdx = srcIdx;\n        vec4 bestIndex = vec4(inIdx);\n        vec4 bestValue = ${_};\n\n        for (int i = 0; i < ${e}; i++) {\n          inIdx = srcIdx;\n          ${N}\n          vec4 candidate = ${_};\n          bvec4 nan = isnan(candidate);\n          bvec4 replace = bvec4(\n            vec4(${C}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));\n\n          bestValue = vec4(replace.x  ? candidate.x : bestValue.x,\n                           replace.y  ? candidate.y : bestValue.y,\n                           replace.z  ? candidate.z : bestValue.z,\n                           replace.w  ? candidate.w : bestValue.w);\n          bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));\n          srcIdx++;\n        }\n        setOutput(bestIndex);\n      }\n    `}};function LM(r,t,e,n=null){let o=t.shape[0],s=t.shape[1];n!=null&&(o=n.shape[0],s=n.shape[1]);let i=v.computeOptimalWindowSize(s),a={windowSize:i,inSize:s,batchSize:o,outSize:Math.ceil(s/i)},u=new Yw(a,e,n==null),l=[t];n!=null&&l.push(n);let c=r.runWebGLProgram(u,l,\"int32\");if(c.shape[1]===1)return c;let p=LM(r,t,e,c);return r.disposeIntermediateTensorInfo(c),p}function MM(r,t,e,n=null){let o=n!=null?n.shape:t.shape,s=o[o.length-1],i=v.computeOptimalWindowSize(s),a=new Zw(o,i,e,n==null),u=n==null?[t]:[t,n],l=r.runWebGLProgram(a,u,\"int32\");if(l.shape.length===t.shape.length){let c=MM(r,t,e,l);return r.disposeIntermediateTensorInfo(l),c}return l}function Jw(r,t,e,n){let o=[e];if(v.assertAxesAreInnerMostDims(\"arg\"+n.charAt(0).toUpperCase()+n.slice(1),o,t.shape.length),!z().getBool(\"WEBGL_PACK_REDUCE\")||t.shape.length<=2){let s=[],i=r.texData.get(t.dataId),a=i!==null&&i.isPacked,u=t;a&&(u=r.unpackTensor(t),s.push(u));let[l,c]=v.computeOutAndReduceShapes(u.shape,o),p=y.sizeFromShape(c),m=st({inputs:{x:u},backend:r,attrs:{shape:[-1,p]}});s.push(m);let f=LM(r,m,n);s.push(f);let d=st({inputs:{x:f},backend:r,attrs:{shape:l}});return s.forEach(h=>r.disposeIntermediateTensorInfo(h)),d}return MM(r,t,n)}function $et(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n,i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Oe({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),v.assertAxesAreInnerMostDims(\"argMax\",[i[0]],u.shape.length);let c=Jw(e,u,i[0],\"max\");return l.forEach(p=>e.disposeIntermediateTensorInfo(p)),c}var zM={kernelName:Wo,backendName:\"webgl\",kernelFunc:$et};function Det(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n,i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Oe({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),v.assertAxesAreInnerMostDims(\"argMin\",[i[0]],u.shape.length);let c=Jw(e,u,i[0],\"min\");return l.forEach(p=>e.disposeIntermediateTensorInfo(p)),c}var BM={kernelName:kl,backendName:\"webgl\",kernelFunc:Det};var Ret=fr+`\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return asin(x);\n`,Fet=Ct({opSnippet:Ret}),VM={kernelName:la,backendName:\"webgl\",kernelFunc:Fet};var Oet=fr+\"return log(x + sqrt(x * x + 1.0));\",Pet=Ct({opSnippet:Oet}),GM={kernelName:ua,backendName:\"webgl\",kernelFunc:Pet};var Let=fr+`\n  return atan(x);\n`,Met=Ct({opSnippet:Let}),WM={kernelName:ca,backendName:\"webgl\",kernelFunc:Met};var zet=Sd+`\n  return atan(a, b);\n`,Bet=`\n  vec4 result = atan(a, b);\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  `+Yi+`\n  return result;\n`,Vet=le({opSnippet:zet,packedOpSnippet:Bet}),UM={kernelName:ma,backendName:\"webgl\",kernelFunc:Vet};var Get=fr+`\n  if ((x < -1.0) || (x > 1.0)) return NAN;\nreturn (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelName:pa,backendName:\"webgl\",kernelFunc:Wet};var ei=class{constructor(t,e,n,o=!1,s=!1){if(this.variableNames=[\"x\"],e===\"avg\"&&n)throw new Error(\"Cannot compute positions for average pool.\");let i=t.filterWidth,a=t.strideHeight,u=t.strideWidth,l=t.dilationHeight,c=t.dilationWidth,p=t.effectiveFilterHeight,m=t.effectiveFilterWidth,f=t.padInfo.top,d=t.padInfo.left;this.outputShape=t.outShape;let h=e===\"avg\",g=`((batch  * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + d`,x=`(xR * ${t.inWidth} + xC) * ${t.inChannels} + d`,b=\"0.0\";if(h||(b=\"-1.0 / 1e-20\"),n){let $=\">=\";this.userCode=`\n        const ivec2 strides = ivec2(${a}, ${u});\n        const ivec2 pads = ivec2(${f}, ${d});\n\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int batch = coords[0];\n          int d = coords[3];\n\n          ivec2 xRCCorner = coords.yz * strides - pads;\n          int xRCorner = xRCCorner.x;\n          int xCCorner = xRCCorner.y;\n\n          // max/min x(?, ?, d) to get y(yR, yC, d).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n          float avgValue = 0.0;\n\n          for (int wR = 0; wR < ${p};\n              wR += ${l}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${t.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${m};\n                wC += ${c}) {\n              int xC = xCCorner + wC;\n\n              if (xC < 0 || xC >= ${t.inWidth}) {\n                continue;\n              }\n\n              float value = getX(batch, xR, xC, d);\n\n              // If a min / max value has already been found, use it. If not,\n              // use the current value.\n              float currMinMaxValue = mix(\n                  value, minMaxValue, minMaxValueFound);\n              if (value ${$} currMinMaxValue) {\n                minMaxValue = value;\n                minMaxValueFound = 1.0;\n                minMaxPosition = ${o?s?g:x:`wR * ${m} + wC`};\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `;return}let w=\"max\",C=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e===\"avg\"&&(C=\"avgValue / count\");let N=Math.floor(i/4)*4,_=i%4,A=`\n      if (${h}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = ${w}(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec2 strides = ivec2(${a}, ${u});\n      const ivec2 pads = ivec2(${f}, ${d});\n      const float initializationValue = ${b};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xR, int xC, int d) {\n        if (xC < 0 || xC >= ${t.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xR, xC, d);\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d = coords[3];\n\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // max/min x(?, ?, d) to get y(yR, yC, d).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${b});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wR = 0; wR < ${p};\n            wR += ${l}) {\n          int xR = xRCorner + wR;\n\n          if (xR < 0 || xR >= ${t.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${N}; wC += 4) {\n            int xC = xCCorner + wC * ${c};\n\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${c}, d),\n              getValue(batch, xR, xC + 2 * ${c}, d),\n              getValue(batch, xR, xC + 3 * ${c}, d)\n            );\n\n            ${A}\n          }\n\n          int xC = xCCorner + ${N};\n          if (${_===1}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              initializationValue,\n              initializationValue,\n              initializationValue\n            );\n\n            ${A}\n          } else if (${_===2}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${c}, d),\n              initializationValue,\n              initializationValue\n            );\n\n            ${A}\n          } else if (${_===3}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${c}, d),\n              getValue(batch, xR, xC + 2 * ${c}, d),\n              initializationValue\n            );\n\n            ${A}\n          }\n        }\n        setOutput(${C});\n      }\n    `}},$u=class{constructor(t,e,n,o=!1,s=!1){if(this.variableNames=[\"x\"],e===\"avg\"&&n)throw new Error(\"Cannot compute positions for average pool.\");let i=t.filterWidth,a=t.strideDepth,u=t.strideHeight,l=t.strideWidth,c=t.dilationDepth,p=t.dilationHeight,m=t.dilationWidth,f=t.effectiveFilterDepth,d=t.effectiveFilterHeight,h=t.effectiveFilterWidth,g=t.padInfo.front,x=t.padInfo.top,b=t.padInfo.left;this.outputShape=t.outShape;let w=e===\"avg\",C=\"0.0\";if(w||(C=\"-1.0 / 1e-20\"),n){let P=\">=\";this.userCode=`\n        const ivec3 strides =\n            ivec3(${a}, ${u}, ${l});\n        const ivec3 pads = ivec3(${g}, ${x}, ${b});\n\n        void main() {\n          ivec5 coords = getOutputCoords();\n          int batch = coords.x;\n          int ch = coords.u;\n\n          ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n          int xDCorner = xCorner.x;\n          int xRCorner = xCorner.y;\n          int xCCorner = xCorner.z;\n\n          // max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n\n          for (int wD = 0; wD < ${f};\n              wD += ${c}) {\n            int xD = xDCorner + wD;\n\n            if (xD < 0 || xD >= ${t.inDepth}) {\n              continue;\n            }\n\n            for (int wR = 0; wR < ${d};\n                wR += ${p}) {\n              int xR = xRCorner + wR;\n\n              if (xR < 0 || xR >= ${t.inHeight}) {\n                continue;\n              }\n\n              for (int wC = 0; wC < ${h};\n                  wC += ${m}) {\n                int xC = xCCorner + wC;\n\n                if (xC < 0 || xC >= ${t.inWidth}) {\n                  continue;\n                }\n\n                float value = getX(batch, xD, xR, xC, ch);\n\n                // If a min / max value has already been found, use it. If not,\n                // use the current value.\n                float currMinMaxValue = mix(\n                    value, minMaxValue, minMaxValueFound);\n                if (value ${P} currMinMaxValue) {\n                  minMaxValue = value;\n                  minMaxValueFound = 1.0;\n                  minMaxPosition = ${o?s?`(((batch * ${t.inDepth} + xD) * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + ch`:`((xD * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + ch`:`wD * ${d} * ${h} +\n                      wR * ${h} + wC`};\n                }\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `;return}let N=\"max\",_=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e===\"avg\"&&(_=\"avgValue / count\");let A=Math.floor(i/4)*4,$=i%4,F=`\n      if (${w}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = ${N}(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec3 strides =\n        ivec3(${a}, ${u}, ${l});\n      const ivec3 pads = ivec3(${g}, ${x}, ${b});\n      const float initializationValue = ${C};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xD, int xR, int xC, int ch) {\n        if (xC < 0 || xC >= ${t.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xD, xR, xC, ch);\n      }\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xDCorner = xCorner.x;\n        int xRCorner = xCorner.y;\n        int xCCorner = xCorner.z;\n\n        // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${C});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wD = 0; wD < ${f};\n            wD += ${c}) {\n          int xD = xDCorner + wD;\n\n          if (xD < 0 || xD >= ${t.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${d};\n            wR += ${p}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${t.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${A}; wC += 4) {\n              int xC = xCCorner + wC * ${m};\n\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${m}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${m}, ch),\n                getValue(batch, xD, xR, xC + 3 * ${m}, ch)\n              );\n\n              ${F}\n            }\n\n            int xC = xCCorner + ${A};\n            if (${$===1}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                initializationValue,\n                initializationValue,\n                initializationValue\n              );\n\n              ${F}\n            } else if (${$===2}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${m}, ch),\n                initializationValue,\n                initializationValue\n              );\n\n              ${F}\n            } else if (${$===3}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${m}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${m}, ch),\n                initializationValue\n              );\n\n              ${F}\n            }\n          }\n          setOutput(${_});\n        }\n      }\n    `}};function Uet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;Qs(o,\"avgPool\");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u);if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))return tr({inputs:{x:o},backend:e});let p=new ei(c,\"avg\",!1);return e.runWebGLProgram(p,[o],\"float32\")}var qM={kernelName:Uo,backendName:\"webgl\",kernelFunc:Uet};function Het(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n,c=[1,1,1],p=v.computePool3DInfo(o.shape,s,i,c,a,u,l),m=new $u(p,\"avg\",!1);return e.runWebGLProgram(m,[o],\"float32\")}var KM={kernelName:El,backendName:\"webgl\",kernelFunc:Het};var Qw=class{constructor(t){this.variableNames=[\"dy\"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=t.dilationHeight,a=t.dilationWidth,u=t.effectiveFilterHeight,l=t.effectiveFilterWidth,c=u-1-t.padInfo.top,p=l-1-t.padInfo.left,m=1/(e*n);this.userCode=`\n      const ivec2 pads = ivec2(${c}, ${p});\n      const float avgMultiplier = float(${m});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${u};\n            wR += ${i}) {\n          float dyR = float(dyRCorner + wR) / ${o}.0;\n\n          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${l};\n            wC+= ${a}) {\n            float dyC = float(dyCCorner + wC) / ${s}.0;\n\n            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n\n            dotProd += dyValue * avgMultiplier;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},tC=class{constructor(t){this.variableNames=[\"dy\"],this.outputShape=t.inShape;let e=t.filterDepth,n=t.filterHeight,o=t.filterWidth,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=t.dilationDepth,l=t.dilationHeight,c=t.dilationWidth,p=t.effectiveFilterDepth,m=t.effectiveFilterHeight,f=t.effectiveFilterWidth,d=p-1-t.padInfo.front,h=m-1-t.padInfo.top,g=f-1-t.padInfo.left,x=1/(e*n*o);this.userCode=`\n      const ivec3 pads = ivec3(${d}, ${h}, ${g});\n      const float avgMultiplier = float(${x});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${p};\n            wD += ${u}) {\n          float dyD = float(dyDCorner + wD) / ${s}.0;\n\n          if (dyD < 0.0 || dyD >= ${t.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${m};\n              wR += ${l}) {\n            float dyR = float(dyRCorner + wR) / ${i}.0;\n\n            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${f};\n                wC += ${c}) {\n              float dyC = float(dyCCorner + wC) / ${a}.0;\n\n              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n\n              dotProd += dyValue * avgMultiplier;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};function qet(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s,{filterSize:a,strides:u,pad:l,dimRoundingMode:c}=n,p=[1,1,1],m=v.computePool3DInfo(i.shape,a,u,p,l,c),f=new tC(m);return e.runWebGLProgram(f,[o],i.dtype)}var jM={kernelName:lp,backendName:\"webgl\",kernelFunc:qet};function Ket(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s;Qs([o,s],\"avgPoolGrad\");let{filterSize:a,strides:u,pad:l}=n,c=v.computePool2DInfo(i.shape,a,u,1,l),p=new Qw(c);return e.runWebGLProgram(p,[o],i.dtype)}var XM={kernelName:ap,backendName:\"webgl\",kernelFunc:Ket};function jet(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;return Uc({a:o,b:s,transposeA:i,transposeB:a,backend:e})}var YM={kernelName:Ho,backendName:\"webgl\",kernelFunc:jet};var eC=class{constructor(t,e,n,o,s,i){this.outputShape=[],this.variableNames=[\"x\",\"mean\",\"variance\"],v.assertAndGetBroadcastShape(t,e),v.assertAndGetBroadcastShape(t,n);let a=\"0.0\";o!=null&&(v.assertAndGetBroadcastShape(t,o),this.variableNames.push(\"offset\"),a=\"getOffsetAtOutCoords()\");let u=\"1.0\";s!=null&&(v.assertAndGetBroadcastShape(t,s),this.variableNames.push(\"scale\"),u=\"getScaleAtOutCoords()\"),this.outputShape=t,this.userCode=`\n      void main() {\n        float x = getXAtOutCoords();\n        float mean = getMeanAtOutCoords();\n        float variance = getVarianceAtOutCoords();\n        float offset = ${a};\n        float scale = ${u};\n        float inv = scale * inversesqrt(variance + float(${i}));\n        setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));\n      }\n    `}};var rC=class{constructor(t,e,n,o,s,i){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=[\"x\",\"mean\",\"variance\"],v.assertAndGetBroadcastShape(t,e),v.assertAndGetBroadcastShape(t,n);let a=\"vec4(0.0)\";o!=null&&(v.assertAndGetBroadcastShape(t,o),this.variableNames.push(\"offset\"),a=\"getOffsetAtOutCoords()\");let u=\"vec4(1.0)\";s!=null&&(v.assertAndGetBroadcastShape(t,s),this.variableNames.push(\"scale\"),u=\"getScaleAtOutCoords()\"),this.outputShape=t,this.userCode=`\n      void main() {\n        vec4 offset = ${a};\n        vec4 scale = ${u};\n\n        vec4 x = getXAtOutCoords();\n        vec4 mean = getMeanAtOutCoords();\n        vec4 variance = getVarianceAtOutCoords();\n\n        vec4 inv = scale * inversesqrt(variance + vec4(${i}));\n\n        setOutput((x - mean) * inv + offset);\n      }\n    `}};var Xet=({inputs:r,backend:t,attrs:e})=>{let{x:n,mean:o,variance:s,offset:i,scale:a}=r;y.assert(o.shape.length===s.shape.length,()=>\"Batch normalization gradient requires mean and variance to have equal ranks.\"),y.assert(i==null||o.shape.length===i.shape.length,()=>\"Batch normalization gradient requires mean and offset to have equal ranks.\"),y.assert(a==null||o.shape.length===a.shape.length,()=>\"Batch normalization gradient requires mean and scale to have equal ranks.\");let{varianceEpsilon:u}=e;u==null&&(u=.001);let l=[n,o,s],c=null;i!=null&&(c=i.shape,l.push(i));let p=null;a!=null&&(p=a.shape,l.push(a));let m=z().getBool(\"WEBGL_PACK_NORMALIZATION\")?new rC(n.shape,o.shape,s.shape,c,p,u):new eC(n.shape,o.shape,s.shape,c,p,u);return t.runWebGLProgram(m,l,l[0].dtype)},ZM={kernelName:os,backendName:\"webgl\",kernelFunc:Xet};var nC=class{constructor(t){this.variableNames=[\"source\"],this.outputShape=t,this.rank=t.length;let e=zt(this.rank);this.customUniforms=[{name:\"start\",arrayIndex:this.rank,type:\"int\"}];let n=Yet(this.rank),o,s=t.map((i,a)=>`sourceLoc.${hk[a]} = start[${a}] + coords.${hk[a]};`);o=`\n        ${e} sourceLoc;\n        ${e} coords = getOutputCoords();\n        ${s.join(`\n`)}\n      `,this.userCode=`\n      void main() {\n        ${o}\n        setOutput(getSource(${n}));\n      }\n    `}},hk=[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"];function Yet(r){if(r===1)return\"sourceLoc\";if(r<=6)return hk.slice(0,r).map(t=>\"sourceLoc.\"+t).join(\",\");throw Error(`Slicing for rank ${r} is not yet supported`)}var oC=class{constructor(t){this.variableNames=[\"source\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.rank=t.length,this.customUniforms=[{name:\"start\",arrayIndex:this.rank,type:\"int\"}];let e=zt(this.rank),n=Qe(\"coords\",this.rank),o=Qe(\"sourceLoc\",this.rank),s=this.rank===1?\"sourceLoc\":`vec2(${o.slice(-2).join()})`,i=`getChannel(getSource(${o.join()}), ${s})`,a=`\n      result.x = ${i};\n      if (++${n[this.rank-1]} < ${t[this.rank-1]}) {\n        ++${o[this.rank-1]};\n        result.y = ${i};\n        --${o[this.rank-1]};\n      }\n    `,u=this.rank===1?\"\":`\n      --${n[this.rank-1]};\n      if (++${n[this.rank-2]} < ${t[this.rank-2]}) {\n        ++${o[this.rank-2]};\n        result.z = ${i};\n        if (++${n[this.rank-1]} < ${t[this.rank-1]}) {\n          ++${o[this.rank-1]};\n          result.w = ${i};\n        }\n      }\n    `,l=this.rank<=4?`sourceLoc = coords +\n            ${e}(${t.map((c,p)=>`start[${p}]`).join()});`:t.map((c,p)=>`${o[p]} = ${n[p]} + start[${p}];`).join(`\n`);this.userCode=`\n      void main() {\n        ${e} coords = getOutputCoords();\n        ${e} sourceLoc;\n        ${l}\n        vec4 result = vec4(0.);\n        ${a}\n        ${u}\n        setOutput(result);\n      }\n    `}};function Zet(r,t,e,n){let o=n.texData.get(r.dataId),s=n.makeTensorInfo(e,r.dtype),i=n.texData.get(s.dataId);Object.assign(i,o),i.refCount=1,i.shape=e,i.dtype=r.dtype;let a=Le.computeFlatOffset(t,y.computeStrides(r.shape));o.slice&&(a+=o.slice.flatOffset),i.slice={flatOffset:a,origDataId:o.slice&&o.slice.origDataId||r.dataId};let u=n.dataRefCount.get(i.slice.origDataId)||1;return n.dataRefCount.set(i.slice.origDataId,u+1),s}function ri(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,size:i}=n,[a,u]=Le.parseSliceParams(o,s,i);if(Le.assertParamsValid(o,a,u),y.sizeFromShape(u)===0)return e.makeTensorInfo(u,o.dtype,[]);if(e.shouldExecuteOnCPU([o])||o.dtype===\"string\"){let p=e.texData.get(o.dataId),m=VL(p.values,a,u,o.shape,o.dtype);return e.makeTensorInfo(u,o.dtype,m)}let{isPacked:l}=e.texData.get(o.dataId),c=Le.isSliceContinous(o.shape,a,u);if(l||!c){let p=z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")?new oC(u):new nC(u),m=[a];return e.runWebGLProgram(p,[o],o.dtype,m)}return e.uploadToGPU(o.dataId),Zet(o,a,u,e)}var JM={kernelName:gi,backendName:\"webgl\",kernelFunc:ri};var Jet=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n;y.assert(o.shape.length<=4,()=>\"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet\");let a=s.reduce((b,w)=>b*w),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=[],d=st({inputs:{x:o},backend:e,attrs:{shape:u}}),h=Oe({inputs:{x:d},backend:e,attrs:{perm:l}}),g=st({inputs:{x:h},backend:e,attrs:{shape:c}}),x=ri({inputs:{x:g},backend:e,attrs:{begin:p,size:m}});return f.push(d),f.push(h),f.push(g),f.forEach(b=>e.disposeIntermediateTensorInfo(b)),x},QM={kernelName:ai,backendName:\"webgl\",kernelFunc:Jet};function Qet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i}=n,a=e.readSync(o.dataId),u=e.readSync(s.dataId),l=Lw(a,u,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,l)}var tz={kernelName:up,backendName:\"webgl\",kernelFunc:Qet};function trt(r){let{inputs:t,backend:e}=r,{s0:n,s1:o}=t,s=e.readSync(n.dataId),i=e.readSync(o.dataId),a=v.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return e.makeTensorInfo([a.length],\"int32\",Int32Array.from(a))}var ez={kernelName:cp,backendName:\"webgl\",kernelFunc:trt};var ert=\"return float(a != b);\",gk=le({opSnippet:ert,cpuKernelImpl:DL,dtype:\"bool\"}),rz={kernelName:Da,backendName:\"webgl\",kernelFunc:gk};function wl(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.texData.get(n.dataId);return tr({inputs:{x:o.complexTensorInfos.real},backend:e})}var nz={kernelName:Rp,backendName:\"webgl\",kernelFunc:wl};var rrt=\"return float(int(x));\";function oz(r,t){let e=new tn(r.shape,rrt),n=t.runWebGLProgram(e,[r],\"int32\");return{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}function xk(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dtype:s}=n;if(s===\"complex64\"){if(o.dtype===\"complex64\")return tr({inputs:{x:o},backend:e});let i=Ne(o.shape),a=xk({inputs:{x:o},backend:e,attrs:{dtype:\"float32\"}}),u=En({inputs:{real:a,imag:i},backend:e});return i.dispose(),e.disposeIntermediateTensorInfo(a),u}if(o.dtype===\"complex64\"){let i=wl({inputs:{input:o},backend:e}),a=xk({inputs:{x:i},backend:e,attrs:{dtype:s}});return e.disposeIntermediateTensorInfo(i),a}if(!y.hasEncodingLoss(o.dtype,s)){let i=tr({inputs:{x:o},backend:e});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(e.shouldExecuteOnCPU([o])){let i=e.texData.get(o.dataId).values,[a,u,l]=mL(i,o.shape,o.dtype,s);return e.makeTensorInfo(a,u,l)}if(s===\"int32\")return oz(o,e);if(s===\"bool\"){let i=e.makeTensorInfo([],\"bool\",y.getTypedArrayFromDType(\"bool\",1)),u=gk({inputs:{a:o,b:i},backend:e});return e.disposeIntermediateTensorInfo(i),u}throw new Error(`Error in Cast: failed to cast ${o.dtype} to ${s}`)}var sz={kernelName:lo,backendName:\"webgl\",kernelFunc:xk};var iz=\"return ceil(x);\",nrt=Ct({opSnippet:iz,packedOpSnippet:iz,cpuKernelImpl:fL}),az={kernelName:qo,backendName:\"webgl\",kernelFunc:nrt};var sC=class{constructor(t){this.variableNames=[\"A\"],this.customUniforms=[{name:\"minVal\",type:\"float\"},{name:\"maxVal\",type:\"float\"}],this.outputShape=t,this.userCode=`\n\n      void main() {\n        float value = getAAtOutCoords();\n        if (isnan(value)) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, minVal, maxVal));\n      }\n    `}};var iC=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"minVal\",type:\"float\"},{name:\"maxVal\",type:\"float\"}],this.outputShape=t,this.userCode=`\n      void main() {\n        vec4 value = getAAtOutCoords();\n\n        if (any(isnan(value))) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, vec4(minVal), vec4(maxVal)));\n      }\n    `}};function ort(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{clipValueMin:s,clipValueMax:i}=n,a;z().getBool(\"WEBGL_PACK_CLIP\")?a=new iC(o.shape):a=new sC(o.shape);let u=[[s],[i]];return e.runWebGLProgram(a,[o],o.dtype,u)}var lz={kernelName:uo,backendName:\"webgl\",kernelFunc:ort};var aC=class{constructor(t){this.variableNames=[\"real\",\"imag\"],this.outputShape=t,this.userCode=`\n      void main() {\n        float re = abs(getRealAtOutCoords());\n        float im = abs(getImagAtOutCoords());\n        float mx = max(re, im);\n\n        // sadly the length function in glsl is not underflow-safe\n        // (at least not on Intel GPUs). So the safe solution is\n        // to ensure underflow-safety in all cases.\n        setOutput(\n          mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))\n        );\n      }\n    `}};function uz(r,t){return{dataId:t.dataId,dtype:t.dtype,shape:r.shape}}function srt(r){let{inputs:t,backend:e}=r,{x:n}=t,o=e.texData.get(n.dataId),s=new aC(n.shape),i=[uz(n,o.complexTensorInfos.real),uz(n,o.complexTensorInfos.imag)];return e.runWebGLProgram(s,i,i[0].dtype)}var cz={kernelName:_l,backendName:\"webgl\",kernelFunc:srt};var lC=class{constructor(t){this.outputShape=[],this.outputShape=v.computeOutShape(t,1),this.variableNames=t.map((i,a)=>`T${a}`);let e=new Array(t.length-1);e[0]=t[0][1];for(let i=1;i<e.length;i++)e[i]=e[i-1]+t[i][1];let n=[`if (yC < ${e[0]}) setOutput(getT0(yR, yC));`];for(let i=1;i<e.length;i++){let a=e[i-1];n.push(`else if (yC < ${e[i]}) setOutput(getT${i}(yR, yC-${a}));`)}let o=e.length,s=e[e.length-1];n.push(`else setOutput(getT${o}(yR, yC-${s}));`),this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int yR = coords.x;\n        int yC = coords.y;\n\n        ${n.join(`\n        `)}\n      }\n    `}};var cC=class{constructor(t,e){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=v.computeOutShape(t,e);let n=this.outputShape,o=n.length,s=zt(o),i=Qe(\"coords\",o),a=[\"x\",\"y\",\"z\",\"w\",\"u\",\"v\"].slice(0,o);this.variableNames=t.map((h,g)=>`T${g}`);let u=new Array(t.length-1);u[0]=t[0][e];for(let h=1;h<u.length;h++)u[h]=u[h-1]+t[h][e];let l=a[e],c=a.slice(-2),p=a.join(),m=`if (${l} < ${u[0]}) {\n        return getChannel(\n            getT0(${p}), vec2(${c.join()}));\n        }`;for(let h=1;h<u.length;h++){let g=u[h-1];m+=`\n        if (${l} < ${u[h]}  && ${l} >= ${u[h-1]}) {\n          return getChannel(\n            getT${h}(${uC(a,l,g)}),\n            vec2(${uC(c,l,g)}));\n        }`}let f=u.length,d=u[u.length-1];m+=`\n        return getChannel(\n          getT${f}(${uC(a,l,d)}),\n          vec2(${uC(c,l,d)}));`,this.userCode=`\n      float getValue(${a.map(h=>\"int \"+h)}) {\n        ${m}\n      }\n\n      void main() {\n        ${s} coords = getOutputCoords();\n        vec4 result = vec4(getValue(${i}), 0., 0., 0.);\n\n        ${i[o-1]} = ${i[o-1]} + 1;\n        if (${i[o-1]} < ${n[o-1]}) {\n          result.g = getValue(${i});\n        }\n\n        ${i[o-2]} = ${i[o-2]} + 1;\n        if (${i[o-2]} < ${n[o-2]}) {\n          result.a = getValue(${i});\n        }\n\n        ${i[o-1]} = ${i[o-1]} - 1;\n        if (${i[o-2]} < ${n[o-2]} &&\n            ${i[o-1]} < ${n[o-1]}) {\n          result.b = getValue(${i});\n        }\n        setOutput(result);\n      }\n    `}};function uC(r,t,e){let n=r.indexOf(t);return r.map((s,i)=>i===n?`${s} - ${e}`:s).join()}function Hc(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.texData.get(n.dataId);return tr({inputs:{x:o.complexTensorInfos.imag},backend:e})}var pz={kernelName:Sp,backendName:\"webgl\",kernelFunc:Hc};function Nd(r,t,e){let n=r[0].dtype;if(n===\"complex64\"){let p=r.map(g=>wl({inputs:{input:g},backend:e})),m=r.map(g=>Hc({inputs:{input:g},backend:e})),f=Nd(p,t,e),d=Nd(m,t,e),h=En({inputs:{real:f,imag:d},backend:e});return p.forEach(g=>e.disposeIntermediateTensorInfo(g)),m.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),h}let o=e.shouldExecuteOnCPU(r);if(n===\"string\"&&(o=!0),o){let p=r.map(b=>{let w=y.sizeFromShape(b.shape.slice(t));return st({inputs:{x:b},backend:e,attrs:{shape:[-1,w]}})}),m=p.map(b=>({vals:e.readSync(b.dataId),shape:b.shape})),f=v.computeOutShape(p.map(b=>b.shape),1),d=p[0].shape[0]===1,h=dL(m,f,n,d),g=v.computeOutShape(r.map(b=>b.shape),t),x=e.makeTensorInfo(g,n,h);return p.forEach(b=>e.disposeIntermediateTensorInfo(b)),x}let s=z().getNumber(\"WEBGL_MAX_TEXTURES_IN_SHADER\");if(r.length>s){let p=[];for(let f=0;f<r.length;f+=s){let d=r.slice(f,f+s);p.push(Nd(d,t,e))}let m=Nd(p,t,e);for(let f of p)e.disposeIntermediateTensorInfo(f);return m}if(z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")&&r[0].shape.length>1){let p=new cC(r.map(m=>m.shape),t);return e.runWebGLProgram(p,r,n)}let{tensors2D:i,outShape:a}=irt(r,t,e),u=new lC(i.map(p=>p.shape)),l=e.runWebGLProgram(u,i,n);i.forEach(p=>e.disposeIntermediateTensorInfo(p));let c=st({inputs:{x:l},attrs:{shape:a},backend:e});return e.disposeIntermediateTensorInfo(l),c}function irt(r,t,e){let n=v.computeOutShape(r.map(s=>s.shape),t);return{tensors2D:r.map(s=>st({inputs:{x:s},attrs:{shape:[-1,y.sizeFromShape(s.shape.slice(t))]},backend:e})),outShape:n}}function yk(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n,s=y.parseAxisParam(o,t[0].shape)[0],i=t.map(l=>l.shape);v.assertParamsConsistent(i,s);let a=v.computeOutShape(t.map(l=>l.shape),s);if(y.sizeFromShape(a)===0)return e.makeTensorInfo(a,t[0].dtype,[]);let u=t.filter(l=>y.sizeFromShape(l.shape)>0);return u.length===1?tr({inputs:{x:u[0]},backend:e}):Nd(u,s,e)}var mz={kernelName:li,backendName:\"webgl\",kernelFunc:yk};var Td=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=[\"x\",\"W\"],this.outputShape=t.outShape;let i=t.padInfo.top,a=t.padInfo.left,u=t.strideHeight,l=t.strideWidth,c=t.dilationHeight,p=t.dilationWidth,m=t.filterHeight,f=t.filterWidth,d=Math.floor(t.inChannels/4)*4,h=t.inChannels%4,g=t.dataFormat===\"channelsLast\",x=g?1:2,b=g?2:3,w=g?3:1,C=\"\",N=\"\";n&&(o?C=`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:s?C=`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:C=`\n          float activation(float x) {\n            ${n}\n          }\n        `,N=\"result = activation(result);\");let _=e?\"result += getBiasAtOutCoords();\":\"\";e&&this.variableNames.push(\"bias\"),o&&this.variableNames.push(\"preluActivationWeights\"),s&&this.variableNames.push(\"leakyreluAlpha\"),this.userCode=`\n      ${C}\n\n      const ivec2 strides = ivec2(${u}, ${l});\n      const ivec2 pads = ivec2(${i}, ${a});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d2 = coords[${w}];\n\n        ivec2 xRCCorner =\n            ivec2(coords[${x}], coords[${b}]) * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${m}; wR++) {\n          int xR = xRCorner + wR * ${c};\n\n          if (xR < 0 || xR >= ${t.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${f}; wC++) {\n            int xC = xCCorner + wC * ${p};\n\n            if (xC < 0 || xC >= ${t.inWidth}) {\n              continue;\n            }\n\n            for (int d1 = 0; d1 < ${d}; d1 += 4) {\n              vec4 wValues = vec4(\n                getW(wR, wC, d1, d2),\n                getW(wR, wC, d1 + 1, d2),\n                getW(wR, wC, d1 + 2, d2),\n                getW(wR, wC, d1 + 3, d2)\n              );\n\n              if (${g}) {\n                vec4 xValues = vec4(\n                  getX(batch, xR, xC, d1),\n                  getX(batch, xR, xC, d1 + 1),\n                  getX(batch, xR, xC, d1 + 2),\n                  getX(batch, xR, xC, d1 + 3)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec4 xValues = vec4(\n                  getX(batch, d1, xR, xC),\n                  getX(batch, d1 + 1, xR, xC),\n                  getX(batch, d1 + 2, xR, xC),\n                  getX(batch, d1 + 3, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n\n            if (${h===1}) {\n\n              if (${g}) {\n                dotProd +=\n                    getX(batch, xR, xC, ${d}) *\n                    getW(wR, wC, ${d}, d2);\n              } else {\n                dotProd +=\n                    getX(batch, ${d}, xR, xC) *\n                    getW(wR, wC, ${d}, d2);\n              }\n\n            } else if (${h===2}) {\n              vec2 wValues = vec2(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2)\n              );\n\n              if (${g}) {\n                vec2 xValues = vec2(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec2 xValues = vec2(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            } else if (${h===3}) {\n              vec3 wValues = vec3(\n                getW(wR, wC, ${d}, d2),\n                getW(wR, wC, ${d} + 1, d2),\n                getW(wR, wC, ${d} + 2, d2)\n              );\n\n              if (${g}) {\n                vec3 xValues = vec3(\n                  getX(batch, xR, xC, ${d}),\n                  getX(batch, xR, xC, ${d} + 1),\n                  getX(batch, xR, xC, ${d} + 2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec3 xValues = vec3(\n                  getX(batch, ${d}, xR, xC),\n                  getX(batch, ${d} + 1, xR, xC),\n                  getX(batch, ${d} + 2, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            }\n          }\n        }\n\n        float result = dotProd;\n        ${_}\n        ${N}\n        setOutput(result);\n      }\n    `}},pC=class{constructor(t){this.variableNames=[\"x\",\"W\"],this.outputShape=t.outShape;let e=t.padInfo.front,n=t.padInfo.top,o=t.padInfo.left,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=t.dilationDepth,l=t.dilationHeight,c=t.dilationWidth,p=t.filterDepth,m=t.filterHeight,f=t.filterWidth,d=Math.floor(t.inChannels/4)*4,h=t.inChannels%4;this.userCode=`\n      const ivec3 strides = ivec3(${s}, ${i}, ${a});\n      const ivec3 pads = ivec3(${e}, ${n}, ${o});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d2 = coords.u;\n\n        ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xFCorner = xFRCCorner.x;\n        int xRCorner = xFRCCorner.y;\n        int xCCorner = xFRCCorner.z;\n\n        // Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get\n        // y(yF, yR, yC, d2). ? = to be determined. : = across all\n        // values in that axis.\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${p}; wF++) {\n          int xF = xFCorner + wF * ${u};\n\n          if (xF < 0 || xF >= ${t.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${m}; wR++) {\n            int xR = xRCorner + wR * ${l};\n\n            if (xR < 0 || xR >= ${t.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${f}; wC++) {\n              int xC = xCCorner + wC * ${c};\n\n              if (xC < 0 || xC >= ${t.inWidth}) {\n                continue;\n              }\n\n              for (int d1 = 0; d1 < ${d}; d1 += 4) {\n                vec4 xValues = vec4(\n                  getX(batch, xF, xR, xC, d1),\n                  getX(batch, xF, xR, xC, d1 + 1),\n                  getX(batch, xF, xR, xC, d1 + 2),\n                  getX(batch, xF, xR, xC, d1 + 3)\n                );\n                vec4 wValues = vec4(\n                  getW(wF, wR, wC, d1, d2),\n                  getW(wF, wR, wC, d1 + 1, d2),\n                  getW(wF, wR, wC, d1 + 2, d2),\n                  getW(wF, wR, wC, d1 + 3, d2)\n                );\n\n                dotProd += dot(xValues, wValues);\n              }\n\n              if (${h===1}) {\n                dotProd +=\n                  getX(batch, xF, xR, xC, ${d}) *\n                  getW(wF, wR, wC, ${d}, d2);\n              } else if (${h===2}) {\n                vec2 xValues = vec2(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1)\n                );\n                vec2 wValues = vec2(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else if (${h===3}) {\n                vec3 xValues = vec3(\n                  getX(batch, xF, xR, xC, ${d}),\n                  getX(batch, xF, xR, xC, ${d} + 1),\n                  getX(batch, xF, xR, xC, ${d} + 2)\n                );\n                vec3 wValues = vec3(\n                  getW(wF, wR, wC, ${d}, d2),\n                  getW(wF, wR, wC, ${d} + 1, d2),\n                  getW(wF, wR, wC, ${d} + 2, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};var kd=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=[\"x\",\"W\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"pads\",type:\"ivec2\"},{name:\"strides\",type:\"ivec2\"},{name:\"dilations\",type:\"ivec2\"},{name:\"inDims\",type:\"ivec2\"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.padInfo.left,a=t.strideWidth,u=t.dilationWidth,l=t.filterHeight,c=t.filterWidth,p=c,m=`\n       int xR; int xC; int xCOffset;\n       vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<c;g++)m+=`\n           vec4 xTexelC${g*2};\n           int xTexelC${g*2}Ready;\n           vec4 xTexelC${g*2+1};\n           int xTexelC${g*2+1}Ready;\n           vec4 xC${g};`;m+=`\n     for (int r = 0; r < ${l}; r++) {\n      for (int d1 = 0; d1 < ${t.inChannels}; d1 += 2) {\n       `;for(let g=0;g<c;g++)m+=`\n           xTexelC${g*2} = vec4(0.0);\n           xTexelC${g*2}Ready = 0;\n           xTexelC${g*2+1} = vec4(0.0);\n           xTexelC${g*2+1}Ready = 0;\n           xC${g} = vec4(0.0);`;m+=`\n         xR = xRCorner + r * dilations[0];\n         if (xR >=0 && xR < inDims[0]) {\n       `;for(let g=0;g<(p+1)/2;g++){let x=g*2;if(m+=`\n           xC = xCCorner + ${x*u};\n           `,a===1){if(x<c&&(i%2===1?(m+=`\n                 xCOffset = xC + 1;\n                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x}Ready == 0) {\n                   xTexelC${x} = getX(batch, xR, xCOffset, d1);\n\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${x}.zw = vec2(0.0);\n                   }\n                   xTexelC${x}Ready = 1;\n                 }\n               `,u===1&&x>0?m+=`\n                 xC${x} = vec4(xTexelC${x-2}.zw, xTexelC${x}.xy);\n                 `:m+=`\n                   xCOffset = xC + 1 - 2;\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       previous.zw = vec2(0.0);\n                     }\n\n                     xC${x} = vec4(previous.zw, xTexelC${x}.xy);\n                   } else {\n                     xC${x} = vec4(0.0, 0.0, xTexelC${x}.xy);\n                   }\n                   `):m+=`\n                 if (xC >= 0 && xC < inDims[1] && xTexelC${x}Ready == 0) {\n                   xTexelC${x} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${x}.zw = vec2(0.0);\n                   }\n                   xTexelC${x}Ready = 1;\n                 }\n\n                 xC${x} = xTexelC${x};\n                 `,x+1<c)){let b=i%2===0?y.nearestLargerEven(u):u;u%2===0&&i%2===1||u%2!==0&&i%2!==1?(m+=`\n                   xCOffset = xC + imod(pads[1], 2) + ${b};\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {\n                     xTexelC${x+1} = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       xTexelC${x+1}.zw = vec2(0.0);\n                     }\n                     xTexelC${x+1}Ready = 1;\n                   }\n                   `,u>1?m+=`\n                     xCOffset -= 2;\n                     if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                      previous = getX(batch, xR, xCOffset, d1);\n                      xC${x+1} = vec4(previous.zw, xTexelC${x+1}.xy);\n                     } else {\n                      xC${x+1} = vec4(0.0, 0.0, xTexelC${x+1}.xy);\n                     }\n                     `:m+=`\n                     xC${x+1} = vec4(xTexelC${x}.zw, xTexelC${x+1}.xy);\n                     `):b===1?m+=`\n                     xC${x+1} = xTexelC${x};\n                     `:m+=`\n                     xCOffset = xC + ${b};\n\n                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {\n                       xTexelC${x+1} = getX(batch, xR, xCOffset, d1);\n                       if (xCOffset + 1 >= inDims[1]) {\n                         xTexelC${x+1}.zw = vec2(0.0);\n                       }\n                       xTexelC${x+1}Ready = 1;\n                     }\n\n                     xC${x+1} = xTexelC${x+1};\n                     `}}else x<c&&(i%2===1?(m+=`\n                 xCOffset = xC + 1 - strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x}Ready == 0) {\n                   xTexelC${x} = getX(batch, xR, xCOffset, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${x}.zw = vec2(0.0);\n                   }\n                   xTexelC${x}Ready = 1;\n                 }\n\n                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${x+1}Ready == 0) {\n                   xTexelC${x+1} = getX(batch, xR, xC + 1, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xC + 2 >= inDims[1]) {\n                     xTexelC${x+1}.zw = vec2(0.0);\n                   }\n                   xTexelC${x+1}Ready = 1;\n                 }\n\n                 xC${x} = vec4(xTexelC${x}.zw, xTexelC${x+1}.zw);\n               `,x+1<c&&(m+=`\n                   final = vec4(0.0);\n                   xCOffset = xC + 1 + strides[1];\n                   if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                     final = getX(batch, xR, xCOffset, d1);\n                   }\n                   xC${x+1} = vec4(xTexelC${x+1}.xy, final.xy);\n                 `)):(m+=`\n                 if(xC >= 0 && xC < inDims[1] && xTexelC${x}Ready == 0) {\n                   xTexelC${x} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${x}.zw = vec2(0.0);\n                   }\n                   xTexelC${x}Ready = 1;\n                 }\n\n                 xCOffset = xC + strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {\n                   xTexelC${x+1} = getX(batch, xR, xCOffset, d1);\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${x+1}.zw = vec2(0.);\n                   }\n                   xTexelC${x+1}Ready = 1;\n                 }\n\n                 xC${x} = vec4(\n                   xTexelC${x}.xy, xTexelC${x+1}.xy);\n               `,x+1<c&&(m+=`\n                   xC${x+1} = vec4(xTexelC${x}.zw, xTexelC${x+1}.zw);\n                 `)));x<c&&(m+=`\n             wTexel = getW(r, ${x}, d1, d2);\n             dotProd += xC${x}.xxzz * vec4(wTexel.xy, wTexel.xy);\n             if(d1 + 1 < ${t.inChannels}) {\n               dotProd += xC${x}.yyww * vec4(wTexel.zw, wTexel.zw);\n             }\n           `,x+1<c&&(m+=`\n               wTexel = getW(r, ${x+1}, d1, d2);\n               dotProd += xC${x+1}.xxzz * vec4(wTexel.xy, wTexel.xy);\n               if(d1 + 1 < ${t.inChannels}) {\n                 dotProd += xC${x+1}.yyww * vec4(wTexel.zw, wTexel.zw);\n               }\n             `))}m+=`\n     }\n   `,m+=`\n     }\n   `,m+=`\n     }\n   `;let f=\"\",d=\"\";n&&(o?f=`vec4 activation(vec4 a) {\n           vec4 b = getPreluActivationWeightsAtOutCoords();\n           ${n}\n         }`:s?f=`vec4 activation(vec4 a) {\n           vec4 b = getLeakyreluAlphaAtOutCoords();\n           ${n}\n         }`:f=`vec4 activation(vec4 x) {\n           ${n}\n         }`,d=\"result = activation(result);\");let h=e?\"result += getBiasAtOutCoords();\":\"\";e&&this.variableNames.push(\"bias\"),o&&this.variableNames.push(\"preluActivationWeights\"),s&&this.variableNames.push(\"leakyreluAlpha\"),this.userCode=`\n       ${f}\n\n       void main() {\n         ivec4 coords = getOutputCoords();\n         int batch = coords.x;\n         ivec2 xRCCorner = coords.yz * strides - pads;\n         int d2 = coords.w;\n         int xRCorner = xRCCorner.x;\n         int xCCorner = xRCCorner.y;\n\n         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n         vec4 dotProd = vec4(0.000000000000001);\n\n         ${m}\n\n         vec4 result = dotProd - vec4(0.000000000000001);\n         ${h}\n         ${d}\n         setOutput(result);\n       }\n     `}};var mC=class{constructor(t,e){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"inputShape\",type:\"ivec4\"},{name:\"pad\",type:\"ivec2\"},{name:\"stride\",type:\"ivec2\"},{name:\"dilation\",type:\"ivec2\"},{name:\"inChannels\",type:\"int\"},{name:\"itemsPerBlockRow\",type:\"int\"},{name:\"outWidth\",type:\"int\"}],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let{dataFormat:n}=e,o=Ge(),s=n===\"channelsLast\",i=s?1:2,a=s?2:3,u=this.enableShapeUniforms?\"if(blockIndex < outShape[2] && pos < outShape[1]) {\":`if(blockIndex < ${t[2]} && pos < ${t[1]}) {`,l=\"\";for(let c=0;c<=1;c++)for(let p=0;p<=1;p++)l+=`\n          blockIndex = rc.z + ${p};\n          pos = rc.y + ${c};\n\n          ${u}\n            offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];\n            d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);\n\n            if(d0 < inputShape[${i}] && d0 >= 0) {\n              // Use custom imod instead mod. On Intel GPU, mod may generate\n              // unexpected value.\n              // https://github.com/tensorflow/tfjs/issues/5447\n              offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];\n              d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /\n                  inChannels);\n\n              if(d1 < inputShape[${a}] && d1 >= 0) {\n\n                ch = imod(pos, inChannels);\n\n                if (${s}) {\n                  innerDims = vec2(d1, ch);\n                  result[${c*2+p}] = getChannel(\n                    getA(rc.x, d0, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                } else {\n                  innerDims = vec2(d0, d1);\n                  result[${c*2+p}] = getChannel(\n                    getA(rc.x, ch, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                }\n              }\n            }\n          }\n        `;this.userCode=`\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0);\n\n        int blockIndex, pos, offsetY, d0, offsetX, d1, ch;\n        vec2 innerDims;\n\n        ${l}\n\n        ${o.output} = result;\n      }\n    `}};function fC(r,t){let e=r.length;return e>=3?t?[...r.slice(0,-3),r[e-3]*r[e-2],r[e-1]]:[...r.slice(0,-3),r[e-3],r[e-2]*r[e-1]]:!t&&e===1&&r[0]>1?[r[0],1]:null}function dC({x:r,filter:t,convInfo:e,backend:n,bias:o=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:a=null}){let u=r.shape,l=n.texData.get(r.dataId),c=e.inChannels,p=u[0]*u[1]*u[2],m=e.outChannels,f=e.dataFormat===\"channelsLast\",d=!1,h=!1,g,x=[];if(s!=null){let C=fC(s.shape,f);C!=null&&(s=st({inputs:{x:s},backend:n,attrs:{shape:C}}),x.push(s))}if(o!=null){let C=fC(o.shape,f);C!=null&&(o=st({inputs:{x:o},backend:n,attrs:{shape:C}}),x.push(o))}if(!((p===1||m===1)&&c>dk)&&l.isPacked&&f&&l.texture!=null&&u[2]%2!==0&&y.arraysEqual(l.shape.slice(-3),u.slice(-3))){let C=u[0]*u[1]*(u[2]+1),N={dataId:r.dataId,shape:[1,C,e.inChannels],dtype:r.dtype},_=l.shape;l.shape=l.shape.slice(),l.shape[l.shape.length-2]++,y.assert(Eu(l.shape,N.shape),()=>`packed reshape ${l.shape} to ${N.shape} isn't free`);let A=st({inputs:{x:t},backend:n,attrs:{shape:[1,e.inChannels,e.outChannels]}});x.push(A);let $=Uc({a:N,b:A,backend:n,transposeA:d,transposeB:h,bias:o,activation:a,preluActivationWeights:s,leakyreluAlpha:i}),F=n.texData.get($.dataId);y.assert(F.isPacked,()=>\"batchMatMul result is expected to be packed\"),l.shape=_,F.shape=e.outShape,g=tr({inputs:{x:$},backend:n}),g.shape=e.outShape,x.push($)}else{let C=e.outHeight*e.outWidth,N=st({inputs:{x:r},backend:n,attrs:{shape:f?[e.batchSize,C,e.inChannels]:[e.batchSize,e.inChannels,C]}}),_=st({inputs:{x:t},backend:n,attrs:{shape:[1,e.inChannels,e.outChannels]}}),A=Uc({a:f?N:_,b:f?_:N,transposeA:!f,transposeB:h,backend:n,bias:o,activation:a,preluActivationWeights:s,leakyreluAlpha:i});g=st({inputs:{x:A},backend:n,attrs:{shape:e.outShape}}),x.push(N),x.push(_),x.push(A)}for(let C of x)n.disposeIntermediateTensorInfo(C);return g}function hC({x:r,filter:t,convInfo:e,backend:n,bias:o=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:a=null}){let{filterWidth:u,filterHeight:l,inChannels:c,outWidth:p,outHeight:m,dataFormat:f}=e,d=f===\"channelsLast\",h=u*l*c,g=m*p,x=[e.batchSize,h,g],b=!0,w=!1,C=[];if(s!=null){let Z=fC(s.shape,d);Z!=null&&(s=st({inputs:{x:s},backend:n,attrs:{shape:Z}}),C.push(s))}if(o!=null){let Z=fC(o.shape,d);Z!=null&&(o=st({inputs:{x:o},backend:n,attrs:{shape:Z}}),C.push(o))}let N=st({inputs:{x:t},backend:n,attrs:{shape:[1,h,y.sizeFromShape(t.shape)/h]}});C.push(N);let _=new mC(x,e),A=[r.shape,[e.padInfo.top,e.padInfo.left],[e.strideHeight,e.strideWidth],[e.dilationHeight,e.dilationWidth],[e.inChannels],[e.filterWidth*e.inChannels],[e.outWidth]],$=n.runWebGLProgram(_,[r],\"float32\",A),F=st({inputs:{x:$},backend:n,attrs:{shape:x}});C.push($),C.push(F);let P=o!=null,V=s!=null,G=a===\"leakyrelu\",W=a?bl(a,!0):null,q=new vd(d?F.shape:N.shape,d?N.shape:F.shape,d?[e.batchSize,g,e.outChannels]:[e.batchSize,e.outChannels,g],b,w,P,W,V,G),H=d?[F,N]:[N,F];if(o&&H.push(o),V&&H.push(s),G){let Z=n.makeTensorInfo([],\"float32\",y.createScalarValue(i,\"float32\"));H.push(Z),C.push(Z)}let j=n.runWebGLProgram(q,H,\"float32\"),Y=st({inputs:{x:j},backend:n,attrs:{shape:e.outShape}});C.push(j);for(let Z of C)n.disposeIntermediateTensorInfo(Z);return Y}function art(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dataFormat:u,dilations:l,dimRoundingMode:c}=n,p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,s.shape,i,l,a,c,!1,p),f;if(m.filterHeight===1&&m.filterWidth===1&&m.dilationHeight===1&&m.dilationWidth===1&&m.strideHeight===1&&m.strideWidth===1&&(m.padInfo.type===\"SAME\"||m.padInfo.type===\"VALID\"))f=dC({x:o,filter:s,convInfo:m,backend:e});else if(m.strideWidth<=2&&p===\"channelsLast\"&&z().getBool(\"WEBGL_EXP_CONV\")){let h=new kd(m),g=[[m.padInfo.top,m.padInfo.left],[m.strideHeight,m.strideWidth],[m.dilationHeight,m.dilationWidth],[m.inHeight,m.inWidth]];f=e.runWebGLProgram(h,[o,s],\"float32\",g)}else if(z().getBool(\"WEBGL_CONV_IM2COL\"))f=hC({x:o,filter:s,convInfo:m,backend:e});else{let h=new Td(m);f=e.runWebGLProgram(h,[o,s],\"float32\")}let d=st({inputs:{x:f},backend:e,attrs:{shape:m.outShape}});return e.disposeIntermediateTensorInfo(f),d}var fz={kernelName:Ko,backendName:\"webgl\",kernelFunc:art};var gC=class{constructor(t){this.variableNames=[\"x\",\"dy\"],this.outputShape=t.filterShape;let e=t.strideHeight,n=t.strideWidth,o=t.padInfo.top,s=t.padInfo.left,i=t.dataFormat===\"channelsLast\";this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int d2 = coords.w;\n\n        // Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${t.batchSize}; b++) {\n          for (int yR = 0; yR < ${t.outHeight}; yR++) {\n            int xR = wR + yR * ${e} - ${o};\n\n            if (xR < 0 || xR >= ${t.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${t.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${s};\n\n              if (xC < 0 || xC >= ${t.inWidth}) {\n                continue;\n              }\n\n              if (${i}) {\n                float dyValue = getDy(b, yR, yC, d2);\n                float xValue = getX(b, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              } else {\n                float dyValue = getDy(b, d2, yR, yC);\n                float xValue = getX(b, d1, xR, xC);\n                dotProd += (xValue * dyValue);\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},xC=class{constructor(t){this.variableNames=[\"dy\",\"W\"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=t.dataFormat===\"channelsLast\",a=e-1-t.padInfo.top,u=n-1-t.padInfo.left,l=i?1:2,c=i?2:3,p=i?3:1;this.userCode=`\n      const ivec2 pads = ivec2(${a}, ${u});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[${p}];\n\n        ivec2 dyCorner = ivec2(coords[${l}], coords[${c}]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${e}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${o}.0;\n\n          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${e} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${s}.0;\n\n            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            for (int d2 = 0; d2 < ${t.outChannels}; d2++) {\n\n              if (${i}) {\n                float xValue = getDy(batch, idyR, idyC, d2);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              } else {\n                float xValue = getDy(batch, d2, idyR, idyC);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},yC=class{constructor(t){this.variableNames=[\"x\",\"dy\"],this.outputShape=t.filterShape;let e=t.strideDepth,n=t.strideHeight,o=t.strideWidth,s=t.padInfo.front,i=t.padInfo.top,a=t.padInfo.left;this.userCode=`\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int wF = coords.x;\n        int wR = coords.y;\n        int wC = coords.z;\n        int d1 = coords.w;\n        int d2 = coords.u;\n\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${t.batchSize}; b++) {\n          for (int yF = 0; yF < ${t.outDepth}; yF++) {\n            int xF = wF + yF * ${e} - ${s};\n\n            if (xF < 0 || xF >= ${t.inDepth}) {\n              continue;\n            }\n\n            for (int yR = 0; yR < ${t.outHeight}; yR++) {\n              int xR = wR + yR * ${n} - ${i};\n\n              if (xR < 0 || xR >= ${t.inHeight}) {\n                continue;\n              }\n\n              for (int yC = 0; yC < ${t.outWidth}; yC++) {\n                int xC = wC + yC * ${o} - ${a};\n\n                if (xC < 0 || xC >= ${t.inWidth}) {\n                  continue;\n                }\n\n                float dyValue = getDy(b, yF, yR, yC, d2);\n                float xValue = getX(b, xF, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},bC=class{constructor(t){this.variableNames=[\"dy\",\"W\"],this.outputShape=t.inShape;let e=t.filterDepth,n=t.filterHeight,o=t.filterWidth,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=e-1-t.padInfo.front,l=n-1-t.padInfo.top,c=o-1-t.padInfo.left;this.userCode=`\n      const ivec3 pads = ivec3(${u}, ${l}, ${c});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.u;\n\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyFCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${e}; wF++) {\n          float dyF = float(dyFCorner + wF) / ${s}.0;\n\n          if (dyF < 0.0 || dyF >= ${t.outDepth}.0 || fract(dyF) > 0.0) {\n            continue;\n          }\n          int idyF = int(dyF);\n\n          int wFPerm = ${e} - 1 - wF;\n\n          for (int wR = 0; wR < ${n}; wR++) {\n            float dyR = float(dyRCorner + wR) / ${i}.0;\n\n            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||\n              fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            int wRPerm = ${n} - 1 - wR;\n\n            for (int wC = 0; wC < ${o}; wC++) {\n              float dyC = float(dyCCorner + wC) / ${a}.0;\n\n              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              int wCPerm = ${o} - 1 - wC;\n\n              for (int d2 = 0; d2 < ${t.outChannels}; d2++) {\n                float xValue = getDy(batch, idyF, idyR, idyC, d2);\n                float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};function lrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,filterShape:c}=n,p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,c,i,1,a,l,!1,p),f=new gC(m);return e.runWebGLProgram(f,[o,s],\"float32\")}var dz={kernelName:mp,backendName:\"webgl\",kernelFunc:lrt};function urt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{inputShape:i,strides:a,pad:u,dataFormat:l,dimRoundingMode:c}=n,p=v.convertConv2DDataFormat(l),m=v.computeConv2DInfo(i,s.shape,a,1,u,c,!1,p),f=new xC(m);return e.runWebGLProgram(f,[o,s],\"float32\")}var hz={kernelName:jo,backendName:\"webgl\",kernelFunc:urt};function crt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n,l=v.computeConv3DInfo(o.shape,s.shape,i,u,a),c=new pC(l);return e.runWebGLProgram(c,[o,s],\"float32\")}var gz={kernelName:Al,backendName:\"webgl\",kernelFunc:crt};function prt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,filterShape:u}=n,l=v.computeConv3DInfo(o.shape,u,i,1,a),c=new yC(l);return e.runWebGLProgram(c,[o,s],\"float32\")}var xz={kernelName:fp,backendName:\"webgl\",kernelFunc:prt};function mrt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{pad:i,strides:a,inputShape:u}=n,l=v.computeConv3DInfo(u,s.shape,a,1,i),c=new bC(l);return e.runWebGLProgram(c,[o,s],\"float32\")}var yz={kernelName:dp,backendName:\"webgl\",kernelFunc:mrt};var frt=Po+`\n  return cos(x);\n`,drt=Ct({opSnippet:frt}),bz={kernelName:Xo,backendName:\"webgl\",kernelFunc:drt};var hrt=`\n  float e2x = exp(-x);\n  return (e2x + 1.0 / e2x) / 2.0;\n`,grt=Ct({opSnippet:hrt}),wz={kernelName:Yo,backendName:\"webgl\",kernelFunc:grt};var wC=class{constructor(t,e,n,o,s){this.variableNames=[\"Image\",\"Boxes\",\"BoxInd\"],this.outputShape=[];let[i,a,u,l]=t,[c]=e,[p,m]=n;this.outputShape=[c,p,m,l];let f=o===\"bilinear\"?1:0,[d,h]=[`${a-1}.0`,`${u-1}.0`],[g,x,b]=p>1?[`${(a-1)/(p-1)}`,\"(y2-y1) * height_ratio\",`y1*${d} + float(y)*(height_scale)`]:[\"0.0\",\"0.0\",`0.5 * (y1+y2) * ${d}`],[w,C,N]=m>1?[`${(u-1)/(m-1)}`,\"(x2-x1) * width_ratio\",`x1*${h} + float(x)*(width_scale)`]:[\"0.0\",\"0.0\",`0.5 * (x1+x2) * ${h}`];this.userCode=`\n      const float height_ratio = float(${g});\n      const float width_ratio = float(${w});\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int y = coords[1];\n        int x = coords[2];\n        int d = coords[3];\n\n        // get box vals\n        float y1 = getBoxes(b,0);\n        float x1 = getBoxes(b,1);\n        float y2 = getBoxes(b,2);\n        float x2 = getBoxes(b,3);\n\n        // get image in batch index\n        int bInd = round(getBoxInd(b));\n        if(bInd < 0 || bInd >= ${i}) {\n          return;\n        }\n\n        float height_scale = ${x};\n        float width_scale = ${C};\n\n        float in_y = ${b};\n        if( in_y < 0.0 || in_y > ${d} ) {\n          setOutput(float(${s}));\n          return;\n        }\n        float in_x = ${N};\n        if( in_x < 0.0 || in_x > ${h} ) {\n          setOutput(float(${s}));\n          return;\n        }\n\n        vec2 sourceFracIndexCR = vec2(in_x,in_y);\n        if(${f} == 1) {\n          // Compute the four integer indices.\n          ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);\n          ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));\n\n          float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);\n          float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);\n          float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);\n          float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);\n\n          vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);\n\n          float top = topLeft + (topRight - topLeft) * fracCR.x;\n          float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;\n          float newValue = top + (bottom - top) * fracCR.y;\n          setOutput(newValue);\n        } else {\n          // Compute the coordinators of nearest neighbor point.\n          ivec2 sourceNearestCR = ivec2(floor(\n            sourceFracIndexCR + vec2(0.5,0.5)));\n          float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);\n          setOutput(newValue);\n        }\n      }\n    `}};var xrt=r=>{let{inputs:t,backend:e,attrs:n}=r,{image:o,boxes:s,boxInd:i}=t,{cropSize:a,method:u,extrapolationValue:l}=n,c=new wC(o.shape,s.shape,a,u,l);return e.runWebGLProgram(c,[o,s,i],\"float32\")},Cz={kernelName:da,backendName:\"webgl\",kernelFunc:xrt};var qc;(function(r){r.Prod=\"*\",r.Sum=\"+\"})(qc||(qc={}));var ng=class{constructor(t,e,n,o){this.op=t,this.outputShape=e,this.variableNames=[\"x\"],this.customUniforms=[{name:\"index\",type:\"float\"}];let s=this.outputShape.length,i=this.op===qc.Prod?\"1.0\":\"0.0\",a=n?i:`getX(${Iz(s,\"coords\",this.op)})`,u=this.outputShape[this.outputShape.length-1],l=\"\",c=\"\";n?(l=o?`end != ${u-1}`:\"end != 0\",c=o?\"end + 1\":\"end - 1\"):(l=o?`end + pow2 < ${u}`:\"end >= pow2\",c=o?\"end + pow2\":\"end - pow2\"),this.userCode=`\n      void main() {\n        ${zt(s)} coords = getOutputCoords();\n        int end = ${Sz(s,\"coords\",this.op)};\n        float val = ${a};\n        int pow2 = int(pow(2.0, index));\n        if (${l}) {\n          int idx = ${c};\n          ${Sz(s,\"coords\",this.op)} = idx;\n          val ${this.op}= getX(${Iz(s,\"coords\",this.op)});\n        }\n        setOutput(val);\n      }\n    `}};function Iz(r,t,e){if(r===1)return`${t}`;if(r===2)return`${t}.x, ${t}.y`;if(r===3)return`${t}.x, ${t}.y, ${t}.z`;if(r===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${e} for rank ${r} is not yet supported`)}function Sz(r,t,e){if(r===1)return`${t}`;if(r===2)return`${t}.y`;if(r===3)return`${t}.z`;if(r===4)return`${t}.w`;throw new Error(`Cumulative ${e} for rank ${r} is not yet supported`)}function CC(r,t,e,n,o,s){let i=t.shape.length,a=v.getAxesPermutation([n],i),u=t;a!=null&&(u=Oe({inputs:{x:t},backend:e,attrs:{perm:a}}));let l=v.getInnerMostAxes(1,i)[0];if(l!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${n}`);let c=u.shape[l],p=tr({inputs:{x:u},backend:e});for(let m=0;m<=Math.ceil(Math.log2(c))-1;m++){let f=new ng(r,u.shape,!1,s),d=[[m]],h=p;p=e.runWebGLProgram(f,[p],p.dtype,d),e.disposeIntermediateTensorInfo(h)}if(o){let m=new ng(r,u.shape,o,s),f=p;p=e.runWebGLProgram(m,[p],p.dtype),e.disposeIntermediateTensorInfo(f)}if(a!=null){let m=v.getUndoAxesPermutation(a),f=Oe({inputs:{x:p},backend:e,attrs:{perm:m}});return e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(u),f}return p}function yrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;return CC(qc.Prod,o,e,s,i,a)}var vz={kernelName:fa,backendName:\"webgl\",kernelFunc:yrt};function brt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;return CC(qc.Sum,o,e,s,i,a)}var Nz={kernelName:Zo,backendName:\"webgl\",kernelFunc:brt};function wrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i,binaryOutput:a}=n;if(o.shape.length===1){let u=e.readSync(o.dataId),l=e.readSync(s.dataId),c=Lw(u,l,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,c)}else if(o.shape.length===2){let u=e.bufferSync(o),l=e.bufferSync(s),c=pL(u,l,i,a);return e.makeTensorInfo(c.shape,s.dtype,c.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${o.shape.length}.`)}var Tz={kernelName:hp,backendName:\"webgl\",kernelFunc:wrt};var IC=class{constructor(t,e,n){this.variableNames=[\"x\"],this.outputShape=[],this.outputShape=t,this.blockSize=e,this.dataFormat=n,this.userCode=`\n    void main() {\n      ivec4 coords = getOutputCoords();\n      int b = coords[0];\n      int h = ${this.getHeightCoordString()};\n      int w = ${this.getWidthCoordString()};\n      int d = ${this.getDepthCoordString()};\n\n      int in_h = h / ${e};\n      int offset_h = imod(h, ${e});\n      int in_w = w / ${e};\n      int offset_w = imod(w, ${e});\n      int offset_d = (offset_h * ${e} + offset_w) *\n        ${this.getOutputDepthSize()};\n      int in_d = d + offset_d;\n\n      float result = ${this.getInputSamplingString()};\n      setOutput(result);\n    }\n  `}getHeightCoordString(){return this.dataFormat===\"NHWC\"?\"coords[1]\":\"coords[2]\"}getWidthCoordString(){return this.dataFormat===\"NHWC\"?\"coords[2]\":\"coords[3]\"}getDepthCoordString(){return this.dataFormat===\"NHWC\"?\"coords[3]\":\"coords[1]\"}getOutputDepthSize(){return this.dataFormat===\"NHWC\"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat===\"NHWC\"?\"getX(b, in_h, in_w, in_d)\":\"getX(b, in_d, in_h, in_w)\"}};function Crt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockSize:s,dataFormat:i}=n,a=o.shape[0],u=i===\"NHWC\"?o.shape[1]:o.shape[2],l=i===\"NHWC\"?o.shape[2]:o.shape[3],c=i===\"NHWC\"?o.shape[3]:o.shape[1],p=u*s,m=l*s,f=c/(s*s),d=i===\"NHWC\"?[a,p,m,f]:[a,f,p,m],h=new IC(d,s,i);return e.runWebGLProgram(h,[o],o.dtype)}var kz={kernelName:ha,backendName:\"webgl\",kernelFunc:Crt};var Ed=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=[\"x\",\"W\"],this.customUniforms=[{name:\"pads\",type:\"ivec2\"},{name:\"strides\",type:\"ivec2\"},{name:\"dilations\",type:\"ivec2\"},{name:\"inDims\",type:\"ivec2\"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.filterHeight,a=t.filterWidth,u=t.outChannels/t.inChannels,l=\"\",c=\"\";n&&(o?l=`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:s?l=`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:l=`\n          float activation(float x) {\n            ${n}\n          }\n        `,c=\"result = activation(result);\");let p=e?\"result += getBiasAtOutCoords();\":\"\";e&&this.variableNames.push(\"bias\"),o&&this.variableNames.push(\"preluActivationWeights\"),s&&this.variableNames.push(\"leakyreluAlpha\"),this.userCode=`\n      ${l}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${u};\n        int q = d2 - d1 * ${u};\n\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.\n        for (int wR = 0; wR < ${i}; wR++) {\n          int xR = xRCorner + wR * dilations[0];\n\n          if (xR < 0 || xR >= inDims[0]) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${a}; wC++) {\n            int xC = xCCorner + wC * dilations[1];\n\n            if (xC < 0 || xC >= inDims[1]) {\n              continue;\n            }\n\n            float xVal = getX(batch, xR, xC, d1);\n            float wVal = getW(wR, wC, d1, q);\n            dotProd += xVal * wVal;\n          }\n        }\n\n        float result = dotProd;\n        ${p}\n        ${c}\n        setOutput(result);\n      }\n    `}};var _d=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=[\"x\",\"W\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"pads\",type:\"ivec2\"},{name:\"strides\",type:\"ivec2\"},{name:\"dilations\",type:\"ivec2\"},{name:\"inDims\",type:\"ivec2\"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.outChannels/t.inChannels,a=t.padInfo.left,u=t.strideWidth,l=t.dilationWidth,c=t.filterHeight,p=t.filterWidth,m=p,f=`\n      int xR; int xC; int xCOffset;\n      vec4 wTexel; vec4 previous; vec4 final;`;for(let x=0;x<p;x++)f+=`\n          vec4 xTexelC${x*2};\n          int xTexelC${x*2}Ready;\n          vec4 xTexelC${x*2+1};\n          int xTexelC${x*2+1}Ready;\n          vec4 xC${x};`;f+=`\n    for (int r = 0; r < ${c}; r++) {\n      `;for(let x=0;x<p;x++)f+=`\n          xTexelC${x*2} = vec4(0.0);\n          xTexelC${x*2}Ready = 0;\n          xTexelC${x*2+1} = vec4(0.0);\n          xTexelC${x*2+1}Ready = 0;\n          xC${x} = vec4(0.0);`;f+=`\n        xR = xRCorner + r * dilations[0];\n        if (xR >=0 && xR < inDims[0]) {\n      `;for(let x=0;x<(m+1)/2;x++){let b=x*2;if(f+=`\n          xC = xCCorner + ${b*l};\n          `,u===1){if(b<p&&(a%2===1?(f+=`\n                xCOffset = xC + 1;\n                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {\n                  xTexelC${b} = getX(batch, xR, xCOffset, d1);\n\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${b}.zw = vec2(0.0);\n                  }\n                  xTexelC${b}Ready = 1;\n                }\n              `,l===1&&b>0?f+=`\n                xC${b} = vec4(xTexelC${b-2}.zw, xTexelC${b}.xy);\n                `:f+=`\n                  xCOffset = xC + 1 - 2;\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                    previous = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      previous.zw = vec2(0.0);\n                    }\n\n                    xC${b} = vec4(previous.zw, xTexelC${b}.xy);\n                  } else {\n                    xC${b} = vec4(0.0, 0.0, xTexelC${b}.xy);\n                  }\n                  `):f+=`\n                if (xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {\n                  xTexelC${b} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${b}.zw = vec2(0.0);\n                  }\n                  xTexelC${b}Ready = 1;\n                }\n\n                xC${b} = xTexelC${b};\n                `,b+1<p)){let w=a%2===0?y.nearestLargerEven(l):l;l%2===0&&a%2===1||l%2!==0&&a%2!==1?(f+=`\n                  xCOffset = xC + imod(pads[1], 2) + ${w};\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {\n                    xTexelC${b+1} = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      xTexelC${b+1}.zw = vec2(0.0);\n                    }\n                    xTexelC${b+1}Ready = 1;\n                  }\n                  `,l>1?f+=`\n                    xCOffset -= 2;\n                    if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n                     xC${b+1} = vec4(previous.zw, xTexelC${b+1}.xy);\n                    } else {\n                     xC${b+1} = vec4(0.0, 0.0, xTexelC${b+1}.xy);\n                    }\n                    `:f+=`\n                    xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.xy);\n                    `):w===1?f+=`\n                    xC${b+1} = xTexelC${b};\n                    `:f+=`\n                    xCOffset = xC + ${w};\n\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {\n                      xTexelC${b+1} = getX(batch, xR, xCOffset, d1);\n                      if (xCOffset + 1 >= inDims[1]) {\n                        xTexelC${b+1}.zw = vec2(0.0);\n                      }\n                      xTexelC${b+1}Ready = 1;\n                    }\n\n                    xC${b+1} = xTexelC${b+1};\n                    `}}else b<p&&(a%2===1?(f+=`\n                xCOffset = xC + 1 - strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {\n                  xTexelC${b} = getX(batch, xR, xCOffset, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${b}.zw = vec2(0.0);\n                  }\n                  xTexelC${b}Ready = 1;\n                }\n\n                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${b+1}Ready == 0) {\n                  xTexelC${b+1} = getX(batch, xR, xC + 1, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xC + 2 >= inDims[1]) {\n                    xTexelC${b+1}.zw = vec2(0.0);\n                  }\n                  xTexelC${b+1}Ready = 1;\n                }\n\n                xC${b} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);\n              `,b+1<p&&(f+=`\n                  final = vec4(0.0);\n                  xCOffset = xC + 1 + strides[1];\n                  if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                    final = getX(batch, xR, xCOffset, d1);\n                  }\n                  xC${b+1} = vec4(xTexelC${b+1}.xy, final.xy);\n                `)):(f+=`\n                if(xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {\n                  xTexelC${b} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${b}.zw = vec2(0.0);\n                  }\n                  xTexelC${b}Ready = 1;\n                }\n\n                xCOffset = xC + strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {\n                  xTexelC${b+1} = getX(batch, xR, xCOffset, d1);\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${b+1}.zw = vec2(0.);\n                  }\n                  xTexelC${b+1}Ready = 1;\n                }\n\n                xC${b} = vec4(\n                  xTexelC${b}.xy, xTexelC${b+1}.xy);\n              `,b+1<p&&(f+=`\n                  xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);\n                `)));b<p&&(f+=`\n            wTexel = getW(r, ${b}, d1, q);\n            dotProd += xC${b} * vec4(wTexel.xz, wTexel.xz);\n          `,b+1<p&&(f+=`\n              wTexel = getW(r, ${b+1}, d1, q);\n              dotProd += xC${b+1} * vec4(wTexel.xz, wTexel.xz);\n            `))}f+=`\n    }\n  `,f+=`\n      }\n    `;let d=\"\",h=\"\";n&&(o?d=`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:s?d=`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:d=`vec4 activation(vec4 x) {\n          ${n}\n        }`,h=\"result = activation(result);\");let g=e?\"result += getBiasAtOutCoords();\":\"\";e&&this.variableNames.push(\"bias\"),o&&this.variableNames.push(\"preluActivationWeights\"),s&&this.variableNames.push(\"leakyreluAlpha\"),this.userCode=`\n      ${d}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${i};\n        int q = d2 - d1 * ${i};\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n        vec4 dotProd = vec4(0.000000000000001);\n\n        ${f}\n\n        vec4 result = dotProd - vec4(0.000000000000001);\n        ${g}\n        ${h}\n        setOutput(result);\n      }\n    `}};function Irt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u,dimRoundingMode:l}=n,c=u;c==null&&(c=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(i,c),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);let p=v.computeConv2DInfo(o.shape,s.shape,i,c,a,l,!0),m;z().getBool(\"WEBGL_PACK_DEPTHWISECONV\")&&p.strideWidth<=2&&p.outChannels/p.inChannels===1?m=new _d(p):m=new Ed(p);let f=[[p.padInfo.top,p.padInfo.left],[p.strideHeight,p.strideWidth],[p.dilationHeight,p.dilationWidth],[p.inHeight,p.inWidth]];return e.runWebGLProgram(m,[o,s],\"float32\",f)}var Ez={kernelName:Jo,backendName:\"webgl\",kernelFunc:Irt};var SC=class{constructor(t){this.variableNames=[\"x\",\"dy\"],this.outputShape=t.filterShape;let e=t.strideHeight,n=t.strideWidth,o=t.padInfo.top,s=t.padInfo.left,i=t.outChannels/t.inChannels;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int dm = coords.w;\n        int d2 = d1 * ${i} + dm;\n\n        float dotProd = 0.0;\n\n        // TO DO: Vec4 over the batch size\n        for (int b = 0; b < ${t.batchSize}; b++) {\n          for (int yR = 0; yR < ${t.outHeight}; yR++) {\n            int xR = wR + yR * ${e} - ${o};\n\n            if (xR < 0 || xR >= ${t.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${t.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${s};\n\n              if (xC < 0 || xC >= ${t.inWidth}) {\n                continue;\n              }\n\n              float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},vC=class{constructor(t){this.variableNames=[\"dy\",\"W\"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=e-1-t.padInfo.top,a=n-1-t.padInfo.left,u=t.outChannels/t.inChannels;this.userCode=`\n      const ivec2 pads = ivec2(${i}, ${a});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n        ivec2 dyCorner = coords.yz - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        float dotProd = 0.0;\n\n        for (int wR = 0; wR < ${e}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${o}.0;\n\n          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${e} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${s}.0;\n\n            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            // TO DO: Vec4 over the channelMul\n            for (int dm = 0; dm < ${u}; dm++) {\n              int d2 = d1 * ${u} + dm;\n              float xValue = getDy(batch, idyR, idyC, d2);\n              float wValue = getW(wRPerm, wCPerm, d1, dm);\n              dotProd += xValue * wValue;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};function Srt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,filterShape:c}=n,p=v.computeConv2DInfo(o.shape,c,i,a,u,l,!0),m=new SC(p);return e.runWebGLProgram(m,[o,s],\"float32\")}var _z={kernelName:gp,backendName:\"webgl\",kernelFunc:Srt};function vrt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,inputShape:c}=n,p=v.computeConv2DInfo(c,s.shape,i,a,u,l,!0),m=new vC(p);return e.runWebGLProgram(m,[o,s],\"float32\")}var Az={kernelName:xp,backendName:\"webgl\",kernelFunc:vrt};var NC=class{constructor(t){this.variableNames=[\"X\"],this.outputShape=[t,t],this.userCode=`\n      void main() {\n          ivec2 coords = getOutputCoords();\n          float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;\n          setOutput(val);\n      }\n    `}};function Nrt(r){let{inputs:t,backend:e}=r,{x:n}=t,o=[...n.shape,...n.shape],s=y.sizeFromShape(n.shape),i=st({inputs:{x:n},backend:e,attrs:{shape:[s]}}),a=new NC(s),u=e.runWebGLProgram(a,[i],i.dtype),l=st({inputs:{x:u},backend:e,attrs:{shape:o}});return e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(u),l}var $z={kernelName:yp,backendName:\"webgl\",kernelFunc:Nrt};var TC=class{constructor(t){this.variableNames=[\"x\",\"W\"],this.outputShape=t.outShape;let{inHeight:e,inWidth:n,padInfo:o,strideHeight:s,strideWidth:i,filterHeight:a,filterWidth:u,dilationHeight:l,dilationWidth:c}=t,{top:p,left:m}=o;this.userCode=`\n      const ivec2 strides = ivec2(${s}, ${i});\n      const ivec2 pads = ivec2(${p}, ${m});\n      const float neg_infinity = -3.4e38;\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.w;\n        ivec2 outTopLeftCorner =\n            coords.yz * strides - pads;\n        int hBeg = outTopLeftCorner.x;\n        int wBeg = outTopLeftCorner.y;\n\n        float curVal = neg_infinity;\n        for (int h = 0; h < ${a}; h++) {\n          int hIn = hBeg + h * ${l};\n\n          if (hIn >= 0 && hIn < ${e}) {\n            for (int w = 0; w < ${u}; w++) {\n              int wIn = wBeg + w * ${c};\n\n              if (wIn >= 0 && wIn < ${n}) {\n                float xVal = getX(batch, hIn, wIn, d1);\n                float wVal = getW(h, w, d1);\n\n                float val = xVal + wVal;\n                if (val > curVal) {\n                  curVal = val;\n                }\n              }\n            }\n          }\n        }\n\n        float result = curVal;\n        setOutput(result);\n      }\n    `}};function Trt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n,l=v.computeDilation2DInfo(o.shape,s.shape,i,a,\"NHWC\",u),c,p=new TC(l);c=e.runWebGLProgram(p,[o,s],\"float32\");let m=st({inputs:{x:c},backend:e,attrs:{shape:l.outShape}});return e.disposeIntermediateTensorInfo(c),m}var Dz={kernelName:$l,backendName:\"webgl\",kernelFunc:Trt};function krt(r){let{inputs:t,backend:e,attrs:n}=r,{equation:o}=n,s=t,{allDims:i,summedDims:a,idDims:u}=v.decodeEinsumEquation(o,s.length);v.checkEinsumDimSizes(i.length,u,s);let{path:l,steps:c}=v.getEinsumComputePath(a,u),p=c.length,m=null,f=i.length,d=[];for(let h=0;h<p;++h){for(let g of c[h]){let{permutationIndices:x,expandDims:b}=v.getEinsumPermutation(f,u[g]),w;v.isIdentityPermutation(x)?w=s[g]:(w=Oe({inputs:{x:s[g]},backend:e,attrs:{perm:x}}),d.push(w));let C=w.shape.slice();for(let N=0;N<b.length;++N)C.splice(b[N],0,1);y.arraysEqual(w.shape,C)||(w=st({inputs:{x:w},backend:e,attrs:{shape:C}}),d.push(w)),m===null?m=w:(m=eg({inputs:{a:w,b:m},backend:e}),d.push(m))}h<p-1&&(l[h]>=0&&(m=Wc({inputs:{x:m},backend:e,attrs:{axis:l[h]-(i.length-f),keepDims:!1}}),d.push(m)),f--)}for(let h of d)h!==m&&e.disposeIntermediateTensorInfo(h);return m}var Rz={kernelName:bp,backendName:\"webgl\",kernelFunc:krt};var Ert=\"return (x >= 0.0) ? x : (exp(x) - 1.0);\",_rt=`\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n`,Art=Ct({opSnippet:Ert,packedOpSnippet:_rt}),Fz={kernelName:ts,backendName:\"webgl\",kernelFunc:Art};var $rt=\"return (b >= 1.0) ? a : a * (b + 1.0);\",Drt=`\n  vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));\n  return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));\n`,Rrt=r=>{let{inputs:t,backend:e}=r,{dy:n,y:o}=t,s=z().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")?new Oo(Drt,n.shape,o.shape):new io($rt,n.shape,o.shape);return e.runWebGLProgram(s,[n,o],n.dtype)},Oz={kernelName:wp,backendName:\"webgl\",kernelFunc:Rrt};var Frt=`\n  return vec4(equal(a, b));\n`,Ort=\"return float(a == b);\",Prt=le({opSnippet:Ort,packedOpSnippet:Frt,dtype:\"bool\",cpuKernelImpl:hL}),Pz={kernelName:xa,backendName:\"webgl\",kernelFunc:Prt};var Lrt=`\n  // Error function is calculated approximately with elementary function.\n  // See \"Handbook of Mathematical Functions with Formulas,\n  // Graphs, and Mathematical Tables\", Abramowitz and Stegun.\n  float p = ${v.ERF_P};\n  float a1 = ${v.ERF_A1};\n  float a2 = ${v.ERF_A2};\n  float a3 = ${v.ERF_A3};\n  float a4 = ${v.ERF_A4};\n  float a5 = ${v.ERF_A5};\n\n  float sign = sign(x);\n  x = abs(x);\n  float t = 1.0 / (1.0 + p * x);\n  return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));\n`,Mrt=Ct({opSnippet:Lrt}),Lz={kernelName:ga,backendName:\"webgl\",kernelFunc:Mrt};var zrt=Po+`\n  return exp(x);\n`,Brt=`\n  vec4 result = exp(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,bk=Ct({opSnippet:zrt,packedOpSnippet:Brt,cpuKernelImpl:gL,dtype:\"float32\"}),Mz={kernelName:es,backendName:\"webgl\",kernelFunc:bk};function kC(r){let{inputs:t,attrs:e,backend:n}=r,{dim:o}=e,{input:s}=t,i=s.shape.length,a=s.shape.slice(),u=o;return o<0&&(y.assert(-(i+1)<=o,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+o+1),a.splice(u,0,1),st({inputs:{x:s},backend:n,attrs:{shape:a}})}var zz={kernelName:ui,backendName:\"webgl\",kernelFunc:kC};var Bz=\"return exp(x) - 1.0;\",Vrt=Ct({opSnippet:Bz,packedOpSnippet:Bz,cpuKernelImpl:xL}),Vz={kernelName:ya,backendName:\"webgl\",kernelFunc:Vrt};var og=class{constructor(t,e,n){this.variableNames=[\"real\",\"imag\"];let o=e[1];this.outputShape=e;let s=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,i=n?`${o}.0`:\"1.0\",a;if(t===\"real\")a=\"return real * expR - imag * expI;\";else if(t===\"imag\")a=\"return real * expI + imag * expR;\";else throw new Error(`FFT component must be either \"real\" or \"imag\", got ${t}.`);this.userCode=`\n      const float exponentMultiplier = ${s};\n\n      float unaryOpComplex(float real, float expR, float imag, float expI) {\n        ${a}\n      }\n\n      float mulMatDFT(int batch, int index) {\n        float indexRatio = float(index) / float(${o});\n        float exponentMultiplierTimesIndexRatio =\n            exponentMultiplier * indexRatio;\n\n        float result = 0.0;\n\n        for (int i = 0; i < ${o}; i++) {\n          // x = (-2|2 * PI / N) * index * i;\n          float x = exponentMultiplierTimesIndexRatio * float(i);\n          float expR = cos(x);\n          float expI = sin(x);\n          float real = getReal(batch, i);\n          float imag = getImag(batch, i);\n\n          result +=\n              unaryOpComplex(real, expR, imag, expI) / ${i};\n        }\n\n        return result;\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        setOutput(mulMatDFT(coords[0], coords[1]));\n      }\n    `}};function EC(r,t,e){let n=e.texData.get(r.dataId),o=y.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=o/s,a=st({inputs:{x:r},backend:e,attrs:{shape:[i,s]}}),u=a.shape,l=new og(\"real\",u,t),c=new og(\"imag\",u,t),p=[{dataId:n.complexTensorInfos.real.dataId,dtype:n.complexTensorInfos.real.dtype,shape:u},{dataId:n.complexTensorInfos.imag.dataId,dtype:n.complexTensorInfos.imag.dtype,shape:u}],m=e.runWebGLProgram(l,p,\"float32\"),f=e.runWebGLProgram(c,p,\"float32\"),d=En({inputs:{real:m,imag:f},backend:e});e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f);let h=st({inputs:{x:d},backend:e,attrs:{shape:r.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(d),h}function Grt(r){let{inputs:t,backend:e}=r,{input:n}=t;return EC(n,!1,e)}var Gz={kernelName:Cp,backendName:\"webgl\",kernelFunc:Grt};var _C=class{constructor(t,e){this.outputShape=[],this.customUniforms=[{name:\"value\",type:\"float\"}],this.variableNames=[\"x\"],this.outputShape=t,this.userCode=`\n      void main() {\n        // Input can be obtained from uniform value.\n        setOutput(value);\n      }\n    `}};function Cl(r){let{backend:t,attrs:e}=r,{shape:n,value:o}=e,{dtype:s}=e;if(s=s||y.inferDtype(o),s===\"string\"){let i=y.getArrayFromDType(s,y.sizeFromShape(n));return i.fill(o),t.makeTensorInfo(n,s,i)}else{let i=new _C(n,o),a=[[o]];return t.runWebGLProgram(i,[],s,a)}}var Wz={kernelName:Dl,backendName:\"webgl\",kernelFunc:Cl};var AC=class{constructor(t){this.variableNames=[\"Image\"],this.outputShape=[];let e=t[2];this.outputShape=t,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n\n          int coordX = ${e} - x - 1;\n          float outputValue;\n          if(coordX >= 0 && coordX < ${e}) {\n            outputValue = getImage(coords[0], coords[1], coordX, coords[3]);\n          } else {\n            outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}};var Uz={kernelName:ba,backendName:\"webgl\",kernelFunc:({inputs:r,backend:t})=>{let{image:e}=r,n=t,o=new AC(e.shape);return n.runWebGLProgram(o,[e],e.dtype)}};var Hz=\"return floor(x);\",Wrt=Ct({opSnippet:Hz,packedOpSnippet:Hz,cpuKernelImpl:yL}),qz={kernelName:rs,backendName:\"webgl\",kernelFunc:Wrt};var Urt=`\n  float s = sign(a) * sign(b);\n  int ia = round(a);\n  int ib = round(b);\n  if (ib != 0) {\n    // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n    return float(idiv(ia, ib, s));\n  } else {\n    return NAN;\n  }\n`,Hrt=`\n  ivec4 ia = round(a);\n  ivec4 ib = round(b);\n  bvec4 cond = notEqual(ib, ivec4(0));\n  ivec4 result = ivec4(0);\n  vec4 s = sign(a) * sign(b);\n\n  // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n  if (cond[0]) {\n    result[0] = idiv(ia[0], ib[0], s[0]);\n  }\n  if (cond[1]) {\n    result[1] = idiv(ia[1], ib[1], s[1]);\n  }\n  if (cond[2]) {\n    result[2] = idiv(ia[2], ib[2], s[2]);\n  }\n  if (cond[3]) {\n    result[3] = idiv(ia[3], ib[3], s[3]);\n  }\n  return vec4(result);\n`,qrt=le({opSnippet:Urt,packedOpSnippet:Hrt,dtype:\"int32\"}),Kz={kernelName:ns,backendName:\"webgl\",kernelFunc:qrt};var $C=class{constructor(t){this.variableNames=[\"A\"];let e=Ge(),[n,o]=t;this.outputShape=t,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${o}.0, ${n}.0);\n\n        vec4 values = ${e.texture2D}(A, uv);\n        float value;\n        if (depth == 0) {\n          value = values.r;\n        } else if (depth == 1) {\n          value = values.g;\n        } else if (depth == 2) {\n          value = values.b;\n        } else if (depth == 3) {\n          value = values.a;\n        }\n\n        setOutput(floor(value * 255.0 + 0.5));\n      }\n    `}};var DC=class{constructor(t){this.variableNames=[\"A\"],this.packedInputs=!1,this.packedOutput=!0;let e=Ge(),[n,o]=t;this.outputShape=t,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n\n        vec4 result = vec4(0.);\n\n        for(int row=0; row<=1; row++) {\n          for(int col=0; col<=1; col++) {\n            texC = coords[1] + row;\n            depth = coords[2] + col;\n\n            vec2 uv = (vec2(texC, texR) + halfCR) /\n                       vec2(${o}.0, ${n}.0);\n            vec4 values = ${e.texture2D}(A, uv);\n            float value;\n            if (depth == 0) {\n              value = values.r;\n            } else if (depth == 1) {\n              value = values.g;\n            } else if (depth == 2) {\n              value = values.b;\n            } else if (depth == 3) {\n              value = values.a;\n            }\n\n            result[row * 2 + col] = floor(value * 255.0 + 0.5);\n          }\n        }\n\n        ${e.output} = result;\n      }\n    `}};var jz={kernelName:Yd,backendName:\"webgl\",kernelFunc:Krt},Ad,wk=z().getBool(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\");function Krt(r){let{inputs:t,backend:e,attrs:n}=r,{pixels:o}=t,{numChannels:s}=n,i=typeof HTMLVideoElement!=\"undefined\"&&o instanceof HTMLVideoElement,a=typeof HTMLImageElement!=\"undefined\"&&o instanceof HTMLImageElement,[u,l]=i?[o.videoWidth,o.videoHeight]:[o.width,o.height],c=[l,u],p=[l,u,s];if(a||i){let h=z().getBool(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\");(Ad==null||h!==wk)&&(wk=h,Ad=document.createElement(\"canvas\").getContext(\"2d\",{willReadFrequently:wk})),Ad.canvas.width=u,Ad.canvas.height=l,Ad.drawImage(o,0,0,u,l),o=Ad.canvas}let m=e.makeTensorInfo(c,\"int32\");e.texData.get(m.dataId).usage=jr.PIXELS,e.gpgpu.uploadPixelDataToTexture(e.getTexture(m.dataId),o);let f=z().getBool(\"WEBGL_PACK\")?new DC(p):new $C(p),d=e.runWebGLProgram(f,[m],\"int32\");return e.disposeData(m.dataId),d}function jrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=v.convertConv2DDataFormat(c),g=v.computeConv2DInfo(o.shape,s.shape,u,p,l,m,!1,h),x,b=[],w=i!=null,C=a!=null,N=f===\"leakyrelu\",_=()=>{let $=[o,s],F=(P,V)=>{if(V===\"NCHW\"&&P.shape.length===1&&P.shape[0]!==1){let G=st({inputs:{x:P},backend:e,attrs:{shape:[P.shape[0],1,1]}});return b.push(G),G}return P};if(w&&$.push(F(i,c)),C&&$.push(F(a,c)),N){let P=e.makeTensorInfo([],\"float32\",y.createScalarValue(d,\"float32\"));$.push(P),b.push(P)}return $};if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type===\"SAME\"||g.padInfo.type===\"VALID\"))x=dC({x:o,filter:s,convInfo:g,backend:e,bias:i,activation:f,preluActivationWeights:a,leakyreluAlpha:d});else if(g.strideWidth<=2&&h===\"channelsLast\"&&z().getBool(\"WEBGL_EXP_CONV\")){let $=f?bl(f,!0):null,F=new kd(g,w,$,C,N),P=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],V=_();x=e.runWebGLProgram(F,V,\"float32\",P)}else if(z().getBool(\"WEBGL_CONV_IM2COL\"))x=hC({x:o,filter:s,convInfo:g,backend:e,bias:i,activation:f,preluActivationWeights:a,leakyreluAlpha:d});else{let $=f?bl(f,!1):null,F=new Td(g,w,$,C,N),P=_();x=e.runWebGLProgram(F,P,\"float32\")}let A=st({inputs:{x},backend:e,attrs:{shape:g.outShape}});return b.push(x),b.forEach($=>e.disposeIntermediateTensorInfo($)),A}var Xz={kernelName:Ii,backendName:\"webgl\",kernelFunc:jrt};function Xrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dimRoundingMode:p,activation:m,leakyreluAlpha:f}=n,d=[],h=c;h==null&&(h=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(u,h),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${u} and dilations '${h}'`);let g=v.computeConv2DInfo(o.shape,s.shape,u,h,l,p,!0),x=z().getBool(\"WEBGL_PACK_DEPTHWISECONV\")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,b=m?bl(m,x):null,w=[o,s],C=i!=null,N=a!=null,_=m===\"leakyrelu\";if(C&&w.push(i),N&&w.push(a),_){let P=e.makeTensorInfo([],\"float32\",y.createScalarValue(f,\"float32\"));w.push(P),d.push(P)}let A;x?A=new _d(g,C,b,N,_):A=new Ed(g,C,b,N,_);let $=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],F=e.runWebGLProgram(A,w,\"float32\",$);return d.forEach(P=>e.disposeIntermediateTensorInfo(P)),F}var Yz={kernelName:Si,backendName:\"webgl\",kernelFunc:Xrt};var RC=class{constructor(t,e,n,o){this.sliceDim=t,this.strides=e,this.paramsShape=o,this.variableNames=[\"x\",\"indices\"],this.outputShape=n;let s=zt(n.length),i=`\n    int index;`;for(let a=0;a<this.sliceDim;a++)i+=`\n          index = round(getIndices(coords[0], ${a}));\n          out_of_bounds = out_of_bounds || index < 0;\n          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[a]};\n          flattenIndex += index * ${this.strides[a]};`;this.userCode=`\n         void main() {\n          ${s} coords = getOutputCoords();\n          int flattenIndex = 0;\n          bool out_of_bounds = false;\n\n          ${i}\n\n          setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));\n        }\n      `}};function Yrt(r){let{inputs:t,backend:e}=r,{params:n,indices:o}=t,s=o.shape,i=s[s.length-1],a=y.sizeFromShape(n.shape),[u,l,c,p]=v.prepareAndValidate(n,o),m=st({inputs:{x:o},backend:e,attrs:{shape:[l,i]}}),f=st({inputs:{x:n},backend:e,attrs:{shape:[y.sizeFromShape(n.shape)/c,c]}});if(e.shouldExecuteOnCPU([n,o])||n.dtype===\"string\"){let x=e.readSync(o.dataId),b=e.bufferSync(n),w=bL(x,b,n.dtype,l,i,c,p,n.shape,a);return e.makeTensorInfo(u,n.dtype,w.values)}let d=new RC(i,p,[l,c],n.shape),h=e.runWebGLProgram(d,[f,m],f.dtype),g=st({inputs:{x:h},backend:e,attrs:{shape:u}});return e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(h),g}var Zz={kernelName:wa,backendName:\"webgl\",kernelFunc:Yrt};var FC=class{constructor(t,e){this.variableNames=[\"A\",\"indices\"],this.outputShape=e,this.rank=e.length;let n=zt(this.rank),o=Zrt(t,2);this.userCode=`\n      void main() {\n        ${n} resRC = getOutputCoords();\n        int index = int(getIndices(resRC.x, resRC.z));\n        float inBounds = (index >= 0) && (index < ${t[2]}) ? 1.0 : 0.0;\n        setOutput(inBounds * getA(${o}));\n      }\n    `}};function Zrt(r,t){let e=[\"resRC.x\",\"resRC.y\",\"resRC.z\",\"resRC.w\"],n=[];for(let o=0;o<r.length;o++)o===2?n.push(\"index\"):n.push(`${e[o]}`);return n.join()}function Ck(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,indices:s}=t,{axis:i,batchDims:a}=n,u=y.parseAxisParam(i,o.shape)[0];if(z().get(\"DEBUG\")){let b=e.readSync(s.dataId),w=o.shape[u];for(let C=0;C<b.length;++C){let N=b[C];y.assert(N<=w-1&&N>=0,()=>`GatherV2: the index value ${N} is not in [0, ${w-1}]`)}}let l=v.segment_util.collectGatherOpShapeInfo(o,s,u,a),c=y.sizeFromShape(s.shape),p=[],m=st({inputs:{x:o},backend:e,attrs:{shape:[l.batchSize,l.outerSize,l.dimSize,l.sliceSize]}}),f=st({inputs:{x:s},backend:e,attrs:{shape:[l.batchSize,c/l.batchSize]}});p.push(m),p.push(f);let d=[l.batchSize,l.outerSize,c/l.batchSize,l.sliceSize];if(e.shouldExecuteOnCPU([o,s])||o.dtype===\"string\"){let b=e.bufferSync(f),w=e.bufferSync(m),C=wL(w,b,d);return p.forEach(N=>e.disposeIntermediateTensorInfo(N)),e.makeTensorInfo(l.outputShape,C.dtype,C.values)}let h=new FC(m.shape,d),g=e.runWebGLProgram(h,[m,f],m.dtype);p.push(g);let x=st({inputs:{x:g},backend:e,attrs:{shape:l.outputShape}});return p.forEach(b=>e.disposeIntermediateTensorInfo(b)),x}var Jz={kernelName:ci,backendName:\"webgl\",kernelFunc:Ck};var Jrt=\"return float(a > b);\",Qrt=`\n  return vec4(greaterThan(a, b));\n`,tnt=le({opSnippet:Jrt,packedOpSnippet:Qrt,cpuKernelImpl:CL,dtype:\"bool\"}),Qz={kernelName:Ca,backendName:\"webgl\",kernelFunc:tnt};var ent=\"return float(a >= b);\",rnt=`\n  return vec4(greaterThanEqual(a, b));\n`,nnt=le({opSnippet:ent,packedOpSnippet:rnt,dtype:\"bool\",cpuKernelImpl:IL}),t3={kernelName:ss,backendName:\"webgl\",kernelFunc:nnt};function ont(r){let{inputs:t,backend:e}=r,{input:n}=t;return EC(n,!0,e)}var e3={kernelName:Ip,backendName:\"webgl\",kernelFunc:ont};var snt=\"return float(!isnan(x) && !isinf(x));\",int=Ct({opSnippet:snt,dtype:\"bool\"}),r3={kernelName:Ia,backendName:\"webgl\",kernelFunc:int};var ant=\"return float(isinf(x));\",lnt=Ct({opSnippet:ant,dtype:\"bool\"}),n3={kernelName:Sa,backendName:\"webgl\",kernelFunc:lnt};var unt=\"return float(isnan(x));\",cnt=Ct({opSnippet:unt,dtype:\"bool\"}),o3={kernelName:va,backendName:\"webgl\",kernelFunc:cnt};var pnt=\"return float(a < b);\",mnt=`\n  return vec4(lessThan(a, b));\n`,fnt=le({opSnippet:pnt,packedOpSnippet:mnt,cpuKernelImpl:SL,dtype:\"bool\"}),s3={kernelName:Na,backendName:\"webgl\",kernelFunc:fnt};var dnt=\"return float(a <= b);\",hnt=`\n  return vec4(lessThanEqual(a, b));\n`,gnt=le({opSnippet:dnt,packedOpSnippet:hnt,cpuKernelImpl:vL,dtype:\"bool\"}),i3={kernelName:Ta,backendName:\"webgl\",kernelFunc:gnt};function xnt(r){let{backend:t,attrs:e}=r,{start:n,stop:o,num:s}=e,i=NL(n,o,s);return t.makeTensorInfo([i.length],\"float32\",i)}var a3={kernelName:vp,backendName:\"webgl\",kernelFunc:xnt};var ynt=Po+`\n  return x < 0.0 ? 0./0. : log(x);\n`,bnt=`\n  vec4 result = log(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);\n  result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);\n  result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);\n  result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);\n  return result;\n`,wnt=Ct({opSnippet:ynt,packedOpSnippet:bnt,cpuKernelImpl:TL}),l3={kernelName:as,backendName:\"webgl\",kernelFunc:wnt};var Cnt=Po+`\n  return log(1.0 + x);\n`,Int=Ct({opSnippet:Cnt}),u3={kernelName:ka,backendName:\"webgl\",kernelFunc:Int};var Snt=\"return float(a >= 1.0 && b >= 1.0);\",vnt=`\n  return vec4(\n    vec4(greaterThanEqual(a, vec4(1.0))) *\n    vec4(greaterThanEqual(b, vec4(1.0))));\n`,Nnt=le({opSnippet:Snt,packedOpSnippet:vnt,dtype:\"bool\"}),c3={kernelName:Ea,backendName:\"webgl\",kernelFunc:Nnt};var Tnt=\"return float(!(x >= 1.0));\",knt=Ct({opSnippet:Tnt}),p3={kernelName:_a,backendName:\"webgl\",kernelFunc:knt};var Ent=\"return float(a >= 1.0 || b >= 1.0);\",_nt=`\n  return min(\n    vec4(greaterThanEqual(a, vec4(1.0))) +\n    vec4(greaterThanEqual(b, vec4(1.0))),\n    vec4(1.0));\n`,Ant=le({opSnippet:Ent,packedOpSnippet:_nt,dtype:\"bool\"}),m3={kernelName:Aa,backendName:\"webgl\",kernelFunc:Ant};var OC=class{constructor(t,e,n,o,s){this.variableNames=[\"x\"],this.outputShape=[];let i=e,a=t[3]-1;this.outputShape=t;let u,l=`float(${n}) + float(${o}) * sum`;s===.5?u=`inversesqrt(${l})`:s===1?u=`1.0/(${l})`:u=`exp(log(${l}) * float(-${s}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n        int d = coords[3];\n        float x = getX(b, r, c, d);\n        float sum = 0.0;\n        for (int j = -${i}; j <= ${i}; j++) {\n          int idx = d + j;\n          if (idx >= 0 && idx <=  ${a}) {\n            float z = getX(b, r, c, idx);\n            sum += z * z;\n          }\n        }\n        float val = x * ${u};\n        setOutput(val);\n      }\n    `}};var PC=class{constructor(t,e,n,o,s){this.variableNames=[\"x\"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let i=e,a=t[3]-1;this.outputShape=t;let u,l=`float(${n}) + float(${o}) * sum`;s===.5?u=`inversesqrt(${l})`:s===1?u=`1.0/(${l})`:u=`exp(log(${l}) * float(-${s}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords.x;\n        int r = coords.y;\n        int c = coords.z;\n        int d = coords.w;\n\n        bool hasNextCol = d < ${this.outputShape[3]};\n        bool hasNextRow = c < ${this.outputShape[2]};\n\n        vec4 sum = vec4(0.);\n        vec4 xFragAtOutputCoords = getX(b, r, c, d);\n\n        vec4 xAtOutputCoords = vec4(\n          getChannel(xFragAtOutputCoords, vec2(c, d)),\n          hasNextCol ?\n            getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,\n          hasNextRow ?\n            getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0\n        );\n\n        int firstChannel = d - ${i};\n        vec2 cache = vec2(0.);\n        if(firstChannel >= 0){\n          vec4 firstChannelFrag = getX(b, r, c, firstChannel);\n          cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));\n            if(hasNextRow){\n              cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));\n            }\n        }\n\n        ivec2 depth = ivec2(d, d + 1);\n        for (int j = - ${i}; j <= ${i}; j++) {\n          ivec2 idx = depth + j;\n          bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));\n          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${a}));\n\n          bool depthInRange = aboveLowerBound.x && belowUpperBound.x;\n          bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;\n\n          if(depthInRange || depthPlusOneInRange){\n            vec4 z = vec4(0.);\n            vec4 xFragAtCurrentDepth;\n            z.xz = cache.xy;\n            if(depthPlusOneInRange && hasNextCol){\n              xFragAtCurrentDepth = idx.y != d ?\n                getX(b, r, c, idx.y) : xFragAtOutputCoords;\n              z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));\n              if(hasNextRow){\n                z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));\n              }\n            }\n            cache.xy = z.yw;\n            sum += z * z;\n          }\n        }\n        vec4 result = xAtOutputCoords * ${u};\n        setOutput(result);\n      }\n    `}};var $nt=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{depthRadius:s,bias:i,alpha:a,beta:u}=n,l=z().getBool(\"WEBGL_PACK_NORMALIZATION\")?new PC(o.shape,s,i,a,u):new OC(o.shape,s,i,a,u);return e.runWebGLProgram(l,[o],o.dtype)},f3={kernelName:Rl,backendName:\"webgl\",kernelFunc:$nt};var LC=class{constructor(t,e,n,o,s){this.variableNames=[\"inputImage\",\"outputImage\",\"dy\"],this.outputShape=[],this.outputShape=t,this.depth=t[3],this.depthRadius=e,this.bias=n,this.alpha=o,this.beta=s,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n\n        float result = 0.0;\n        for (int d = 0; d < ${this.depth}; ++d) {\n          int depthBegin = int(max(0.0, float(d - ${e})));\n          int depthEnd = int(min(float(${this.depth}),\n              float(d + ${e} + 1)));\n\n          const int MIN_DEPTH_BEGIN = 0;\n          const int MAX_DEPTH_END = ${this.depth};\n\n          float norm = 0.0;\n          for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd) {\n              norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);\n            }\n            else {\n              break;\n            }\n          }\n\n          norm = float(${o}) * norm + float(${n});\n\n          for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd){\n              float dyi = -2.0 * float(${o})\n                * float(${s})\n                * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)\n                / norm;\n              if (k == d) {\n                dyi += pow(norm, -1.0 * ${s});\n              }\n              if (k == coords[3]) {\n                dyi *= getDy(b, r, c, d);\n                result += dyi;\n              }\n            }\n            else {\n              break;\n            }\n          }\n      }\n      setOutput(result);\n      }\n    `}};var Dnt=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o,y:s,dy:i}=t,{depthRadius:a,bias:u,alpha:l,beta:c}=n,p=new LC(o.shape,a,u,l,c);return e.runWebGLProgram(p,[o,s,i],o.dtype)},d3={kernelName:Np,backendName:\"webgl\",kernelFunc:Dnt};function h3(r,t,e,n){let o=y.sizeFromShape(t),i=y.sizeFromShape(r.shape)/o,a=st({inputs:{x:r},attrs:{shape:[i,o]},backend:n}),u=Un(a,r.dtype,\"max\",n),l=st({inputs:{x:u},attrs:{shape:e},backend:n});return n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(u),l}function Ik(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reductionIndices:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=c!=null,m=e.shouldExecuteOnCPU([o]),f=o;if(p){if(m){let w=e.texData.get(f.dataId).values,C=new Array(a);for(let A=0;A<C.length;A++)C[A]=o.shape[c[A]];let N=Vc(w,o.shape,o.dtype,c,C);f=e.makeTensorInfo(C,o.dtype);let _=e.texData.get(f.dataId);_.values=N}else f=Au(o,c,e);l=v.getInnerMostAxes(l.length,a)}v.assertAxesAreInnerMostDims(\"max\",l,a);let[d,h]=v.computeOutAndReduceShapes(f.shape,l),g=d;i&&(g=v.expandShapeToKeepDim(d,u));let x;if(m){let w=e.texData.get(f.dataId).values,C=kL(w,y.sizeFromShape(h),g,o.dtype);x=e.makeTensorInfo(g,o.dtype);let N=e.texData.get(x.dataId);N.values=C}else x=h3(f,h,g,e);return p&&e.disposeIntermediateTensorInfo(f),x}var g3={kernelName:ls,backendName:\"webgl\",kernelFunc:Ik};var Rnt=Sd+`\n  return max(a, b);\n`,Fnt=`\n  vec4 result = vec4(max(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  `+Yi+`\n  return result;\n`,Ont=le({opSnippet:Rnt,packedOpSnippet:Fnt,cpuKernelImpl:EL}),x3={kernelName:us,backendName:\"webgl\",kernelFunc:Ont};function Pnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;Qs(o,\"maxPool\");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u);if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))return tr({inputs:{x:o},backend:e});let p=new ei(c,\"max\",!1);return e.runWebGLProgram(p,[o],o.dtype)}var y3={kernelName:cs,backendName:\"webgl\",kernelFunc:Pnt};function Lnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dataFormat:u,dimRoundingMode:l}=n,c=[1,1,1],p=v.computePool3DInfo(o.shape,s,i,c,a,l,u),m=new $u(p,\"max\",!1);return e.runWebGLProgram(m,[o],o.dtype)}var b3={kernelName:Fl,backendName:\"webgl\",kernelFunc:Lnt};var MC=class{constructor(t){this.variableNames=[\"dy\",\"maxPos\"],this.outputShape=t.inShape;let e=t.strideHeight,n=t.strideWidth,o=t.dilationHeight,s=t.effectiveFilterHeight,i=t.effectiveFilterWidth,a=s-1-t.padInfo.top,u=i-1-t.padInfo.left,l=s*i-1;this.userCode=`\n      const ivec2 pads = ivec2(${a}, ${u});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${s};\n          wR += ${o}) {\n          float dyR = float(dyRCorner + wR) / ${e}.0;\n\n          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${i}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${n}.0;\n\n            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n            int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));\n\n            // Get the current value, check it against the value from the\n            // position matrix.\n            int curPosValue = wR * ${i} + wC;\n            float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n            dotProd += dyValue * mask;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},zC=class{constructor(t){this.variableNames=[\"dy\",\"maxPos\"],this.outputShape=t.inShape;let e=t.strideDepth,n=t.strideHeight,o=t.strideWidth,s=t.dilationDepth,i=t.dilationHeight,a=t.dilationWidth,u=t.effectiveFilterDepth,l=t.effectiveFilterHeight,c=t.effectiveFilterWidth,p=u-1-t.padInfo.front,m=l-1-t.padInfo.top,f=c-1-t.padInfo.left,d=u*l*c-1;this.userCode=`\n      const ivec3 pads = ivec3(${p}, ${m}, ${f});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${u};\n           wD += ${s}) {\n          float dyD = float(dyDCorner + wD) / ${e}.0;\n\n          if (dyD < 0.0 || dyD >= ${t.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${l};\n              wR += ${i}) {\n            float dyR = float(dyRCorner + wR) / ${n}.0;\n\n            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${c};\n                wC += ${a}) {\n              float dyC = float(dyCCorner + wC) / ${o}.0;\n\n              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n              int maxPosValue = ${d} -\n                  int(getMaxPos(batch, idyD, idyR, idyC, ch));\n\n              // Get the current value, check it against the value from the\n              // position matrix.\n              int curPosValue =\n                  wD * ${l} * ${c} +\n                  wR * ${c} + wC;\n              float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n              dotProd += dyValue * mask;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};function Mnt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s,{filterSize:a,strides:u,pad:l,dimRoundingMode:c}=n,p=[1,1,1],m=v.computePool3DInfo(i.shape,a,u,p,l,c),f=new $u(m,\"max\",!0),d=e.runWebGLProgram(f,[i],i.dtype),h=new zC(m),g=e.runWebGLProgram(h,[o,d],i.dtype);return e.disposeIntermediateTensorInfo(d),g}var w3={kernelName:kp,backendName:\"webgl\",kernelFunc:Mnt};function znt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s,output:i}=t,a=s;Qs([s,i],\"maxPoolGrad\");let{filterSize:u,strides:l,pad:c,dimRoundingMode:p}=n,m=v.computePool2DInfo(a.shape,u,l,1,c,p),f=!0,d=new ei(m,\"max\",f),h=e.runWebGLProgram(d,[a],a.dtype),g=new MC(m),x=e.runWebGLProgram(g,[o,h],a.dtype);return e.disposeIntermediateTensorInfo(h),x}var C3={kernelName:Tp,backendName:\"webgl\",kernelFunc:znt};function I3(r,t,e,n){let o=new ei(e,\"max\",!1),s=n.runWebGLProgram(o,[r],\"float32\");o=new ei(e,\"max\",!0,!0,t);let i=n.runWebGLProgram(o,[r],\"float32\");return[s,i]}var S3={kernelName:Ep,backendName:\"webgl\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{filterSize:o,strides:s,pad:i,includeBatchInIndex:a}=t,u=e;y.assert(n.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${n.shape.length}.`);let l=[1,1];y.assert(v.eitherStridesOrDilationsAreOne(s,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${l}'`);let c=v.computePool2DInfo(n.shape,o,s,l,i),[p,m]=I3(n,a,c,u);return[p,m]}};function v3(r,t,e,n){let o=y.sizeFromShape(t),i=y.sizeFromShape(r.shape)/o,a=st({inputs:{x:r},attrs:{shape:[i,o]},backend:n}),u=Un(a,\"float32\",\"mean\",n),l=st({inputs:{x:u},attrs:{shape:e},backend:n});return n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(u),l}var N3={kernelName:ps,backendName:\"webgl\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{keepDims:o,axis:s}=t,i=e,a=n.shape.length,u=y.parseAxisParam(s,n.shape),l=u,c=v.getAxesPermutation(l,a),p=c!=null,m=i.shouldExecuteOnCPU([n]),f=[],d=n;if(p){if(m){let C=i.texData.get(d.dataId).values,N=new Array(a);for(let $=0;$<N.length;$++)N[$]=n.shape[c[$]];let _=Vc(C,n.shape,n.dtype,c,N);d=i.makeTensorInfo(N,n.dtype);let A=i.texData.get(d.dataId);A.values=_}else d=Au(n,c,i);f.push(d),l=v.getInnerMostAxes(l.length,a)}v.assertAxesAreInnerMostDims(\"sum\",l,a);let[h,g]=v.computeOutAndReduceShapes(d.shape,l),x=h;o&&(x=v.expandShapeToKeepDim(h,u));let b=v3(d,g,x,i);for(let w of f)i.disposeIntermediateTensorInfo(w);return b}};function Bnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,o.shape.length)),v.assertAxesAreInnerMostDims(\"min\",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,\"min\",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var T3={kernelName:ms,backendName:\"webgl\",kernelFunc:Bnt};var Vnt=Sd+`\n  return min(a, b);\n`,Gnt=`\n  vec4 result = vec4(min(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  `+Yi+`\n  return result;\n`,Wnt=le({opSnippet:Vnt,packedOpSnippet:Gnt,cpuKernelImpl:_L}),k3={kernelName:fs,backendName:\"webgl\",kernelFunc:Wnt};var BC=class{constructor(t,e,n){this.variableNames=[\"x\"],this.outputShape=e.map((c,p)=>c[0]+t[p]+c[1]);let o=t.length,s=zt(o),i=e.map(c=>c[0]).join(\",\"),a=e.map((c,p)=>c[0]+t[p]).join(\",\"),u=[\"coords[0]\",\"coords[1]\",\"coords[2]\",\"coords[3]\"].slice(0,o),l=n===\"reflect\"?0:1;if(o===1){this.userCode=`\n        int start = ${i};\n        int end = ${a};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start) {\n            outC = start * 2 - outC - ${l};\n          } else if(outC >= end) {\n            outC = (end - 1) * 2 - outC + ${l};\n          }\n          setOutput(getX(outC - start));\n        }\n      `;return}this.userCode=`\n      ${s} start = ${s}(${i});\n      ${s} end = ${s}(${a});\n\n      void main() {\n        ${s} outC = getOutputCoords();\n        for (int i = 0; i < ${o}; i++) {\n          if (outC[i] < start[i]) {\n            outC[i] = start[i] * 2 - outC[i] - ${l};\n          } else if(outC[i] >= end[i]) {\n            outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};\n          }\n        }\n        ${s} coords = outC - start;\n        setOutput(getX(${u}));\n      }\n    `}};var VC=class{constructor(t,e,n){this.variableNames=[\"x\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e.map((d,h)=>d[0]+t[h]+d[1]);let o=t.length,s=zt(o),i=e.map(d=>d[0]).join(\",\"),a=e.map((d,h)=>d[0]+t[h]).join(\",\"),u=Qe(\"rc\",o),l=Qe(\"source\",o),c=`${u[o-1]} < ${this.outputShape[o-1]}`,p=o===1?\"source\":`vec2(${l.slice(-2).join()})`,m=n===\"reflect\"?0:1,f=\"\";if(o===1){let d=`\n        ${s} source = rc;\n        if (source < start) {\n          source = start * 2 - source - ${m};\n        } else if (source >= end) {\n          source = (end - 1) * 2 - source + ${m};\n        }\n        source -= start;\n      `;f=`\n        ${s} rc = outputLoc;\n        ${d}\n        result[0] = getChannel(getX(${l.join()}), ${p});\n        ${u[o-1]} += 1;\n        if(${c}) {\n          ${d}\n          result[1] = getChannel(getX(${l.join()}), ${p});\n        }\n      `}else{let d=`\n        ${s} source = rc;\n        ${s} lt = ${s}(lessThan(source, start));\n        ${s} gte = ${s}(greaterThanEqual(source, end));\n        ${s} orig = 1 - (lt + gte);\n        source = orig * source +\n                lt * (start * 2 - source - ${m}) +\n                gte * ((end - 1) * 2 - source + ${m});\n        source -= start;\n      `;f=`\n        ${s} rc = outputLoc;\n        ${d}\n        result[0] = getChannel(getX(${l.join()}), ${p});\n        ${u[o-1]} += 1;\n        if(${c}) {\n          ${d}\n          result[1] = getChannel(getX(${l.join()}), ${p});\n        }\n        rc = outputLoc;\n        ${u[o-2]} += 1;\n        if(${u[o-2]} < ${this.outputShape[o-2]}) {\n          ${d}\n          result[2] = getChannel(getX(${l.join()}), ${p});\n          ${u[o-1]} += 1;\n          if(${c}) {\n            ${d}\n            result[3] = getChannel(getX(${l.join()}), ${p});\n          }\n        }\n      `}this.userCode=`\n      const ${s} start = ${s}(${i});\n      const ${s} end = ${s}(${a});\n\n      void main() {\n        ${s} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${f}\n        setOutput(result);\n      }\n    `}};var Unt=({inputs:r,backend:t,attrs:e})=>{let{x:n}=r,{paddings:o,mode:s}=e,i=z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")?new VC(n.shape,o,s):new BC(n.shape,o,s);return t.runWebGLProgram(i,[n],n.dtype)},E3={kernelName:ds,backendName:\"webgl\",kernelFunc:Unt};var Hnt=`if (b == 0.0) return NAN;\n  return mod(a, b);`,qnt=`\n  vec4 result = mod(a, b);\n  bvec4 isNaN = equal(b, vec4(0.0));\n  `+Yi+`\n  return result;\n`,Knt=le({opSnippet:Hnt,packedOpSnippet:qnt}),_3={kernelName:$a,backendName:\"webgl\",kernelFunc:Knt};var GC=class{constructor(t,e,n){this.variableNames=[\"probs\"],this.customUniforms=[{name:\"seed\",type:\"float\"}],this.outputShape=[t,n],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n\n        float r = random(seed);\n        float cdf = 0.0;\n\n        for (int i = 0; i < ${e-1}; i++) {\n          cdf += getProbs(batch, i);\n\n          if (r < cdf) {\n            setOutput(float(i));\n            return;\n          }\n        }\n\n        // If no other event happened, last event happened.\n        setOutput(float(${e-1}));\n      }\n    `}};var jnt=`\nif (a == b) {\n  return 1.0;\n};\nreturn a / b;`,Xnt=`\n  // vec4 one = vec4(equal(a, b));\n  // return one + (vec4(1.0) - one) * a / b;\n  vec4 result = a / b;\n  if(a.x == b.x) {\n    result.x = 1.;\n  }\n  if(a.y == b.y) {\n    result.y = 1.;\n  }\n  if(a.z == b.z) {\n    result.z = 1.;\n  }\n  if(a.w == b.w) {\n    result.w = 1.;\n  }\n\n  return result;\n`,Sk=le({opSnippet:jnt,packedOpSnippet:Xnt,checkOutOfBounds:!0}),A3={kernelName:Qo,backendName:\"webgl\",kernelFunc:Sk};var $3=\"return a - b;\",vk=le({opSnippet:$3,packedOpSnippet:$3,supportsComplex:!0,cpuKernelImpl:XL}),D3={kernelName:Fs,backendName:\"webgl\",kernelFunc:vk};function Nk(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{dim:s}=n,i=y.parseAxisParam([s],o.shape),a=Ik({inputs:{x:o},backend:e,attrs:{reductionIndices:i,keepDims:!1}}),u=v.expandShapeToKeepDim(a.shape,i),l=st({inputs:{x:a},backend:e,attrs:{shape:u}}),c=vk({inputs:{a:o,b:l},backend:e}),p=bk({inputs:{x:c},backend:e}),m=Wc({inputs:{x:p},backend:e,attrs:{axis:i,keepDims:!1}}),f=st({inputs:{x:m},backend:e,attrs:{shape:u}}),d=Sk({inputs:{a:p,b:f},backend:e});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),d}var R3={kernelName:Ds,backendName:\"webgl\",kernelFunc:Nk};function Ynt(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{numSamples:s,seed:i,normalized:a}=n,u=a?o:Nk({inputs:{logits:o},backend:e,attrs:{dim:o.shape.length-1}}),l=u.shape[0],c=u.shape[1],p=new GC(l,c,s),m=[[i]],f=e.runWebGLProgram(p,[u],\"int32\",m);return a||e.disposeIntermediateTensorInfo(u),f}var F3={kernelName:_p,backendName:\"webgl\",kernelFunc:Ynt};var Znt=fr+`\n  return -x;\n`,Jnt=`\n  vec4 result = -x;\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;function Qnt(r){let{inputs:t,backend:e}=r,{x:n}=t;if(e.shouldExecuteOnCPU([n])){let s=e.texData.get(n.dataId),[i,a]=$L(s.values,n.shape,n.dtype);return e.makeTensorInfo(a,n.dtype,i)}let o;return z().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")?o=new so(n.shape,Jnt):o=new tn(n.shape,Znt),e.runWebGLProgram(o,[n],n.dtype)}var O3={kernelName:pi,backendName:\"webgl\",kernelFunc:Qnt};var tot=Ur.nonMaxSuppressionV3Impl;function eot(r){v.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u}=n,l=e.readSync(o.dataId),c=e.readSync(s.dataId),{selectedIndices:p}=tot(l,c,i,a,u);return e.makeTensorInfo([p.length],\"int32\",new Int32Array(p))}var P3={kernelName:Ra,backendName:\"webgl\",kernelFunc:eot};var rot=Ur.nonMaxSuppressionV4Impl;function not(r){v.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,padToMaxOutputSize:l}=n,c=e.readSync(o.dataId),p=e.readSync(s.dataId),{selectedIndices:m,validOutputs:f}=rot(c,p,i,a,u,l);return[e.makeTensorInfo([m.length],\"int32\",new Int32Array(m)),e.makeTensorInfo([],\"int32\",new Int32Array([f]))]}var L3={kernelName:Fa,backendName:\"webgl\",kernelFunc:not};var oot=Ur.nonMaxSuppressionV5Impl;function sot(r){v.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,softNmsSigma:l}=n,c=e.readSync(o.dataId),p=e.readSync(s.dataId),m=i,f=a,d=u,h=l,{selectedIndices:g,selectedScores:x}=oot(c,p,m,f,d,h);return[e.makeTensorInfo([g.length],\"int32\",new Int32Array(g)),e.makeTensorInfo([x.length],\"float32\",new Float32Array(x))]}var M3={kernelName:Oa,backendName:\"webgl\",kernelFunc:sot};var WC=class{constructor(t,e,n,o){this.variableNames=[\"indices\"],this.outputShape=[t,e],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int index = round(getIndices(coords.x));\n        setOutput(mix(float(${o}), float(${n}),\n                      float(index == coords.y)));\n      }\n    `}};var iot=r=>{let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n,l=y.sizeFromShape(o.shape),c=new WC(l,i,a,u),p=st({inputs:{x:o},backend:e,attrs:{shape:[l]}}),m=e.runWebGLProgram(c,[p],s);e.disposeIntermediateTensorInfo(p);let f=[...o.shape,i],d=st({inputs:{x:m},backend:e,attrs:{shape:f}});return e.disposeIntermediateTensorInfo(m),d},z3={kernelName:gs,backendName:\"webgl\",kernelFunc:iot};function sg(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype===\"complex64\"){let o=wl({inputs:{input:n},backend:e}),s=sg({inputs:{x:o},backend:e}),i=Hc({inputs:{input:n},backend:e}),a=sg({inputs:{x:i},backend:e}),u=En({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Cl({attrs:{shape:n.shape,dtype:n.dtype,value:n.dtype===\"string\"?\"\":0},backend:e})}var B3={kernelName:wi,backendName:\"webgl\",kernelFunc:sg};function V3(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype===\"string\")throw new Error(\"onesLike is not supported under string dtype\");if(n.dtype===\"complex64\"){let o=wl({inputs:{input:n},backend:e}),s=V3({inputs:{x:o},backend:e}),i=Hc({inputs:{input:n},backend:e}),a=sg({inputs:{x:i},backend:e}),u=En({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Cl({attrs:{shape:n.shape,dtype:n.dtype,value:1},backend:e})}var G3={kernelName:mi,backendName:\"webgl\",kernelFunc:V3};function aot(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return kC({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,\"All tensors passed to stack must have matching shapes\"),y.assert(i===c.dtype,()=>\"All tensors passed to stack must have matching dtypes\")});let a=[],u=t.map(c=>{let p=kC({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=yk({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeIntermediateTensorInfo(c)),l}var W3={kernelName:fi,backendName:\"webgl\",kernelFunc:aot};var UC=class{constructor(t,e,n){this.variableNames=[\"x\"],this.customUniforms=[{name:\"value\",type:\"float\"}],this.outputShape=e.map((l,c)=>l[0]+t[c]+l[1]);let o=t.length,s=zt(o),i=e.map(l=>l[0]).join(\",\"),a=e.map((l,c)=>l[0]+t[c]).join(\",\"),u=[\"coords[0]\",\"coords[1]\",\"coords[2]\",\"coords[3]\"].slice(0,o);if(o===1){this.userCode=`\n        int start = ${i};\n        int end = ${a};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start || outC >= end) {\n            setOutput(value);\n          } else {\n            setOutput(getX(outC - start));\n          }\n        }\n      `;return}this.userCode=`\n      ${s} start = ${s}(${i});\n      ${s} end = ${s}(${a});\n\n      void main() {\n        ${s} outC = getOutputCoords();\n        if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {\n          setOutput(value);\n        } else {\n          ${s} coords = outC - start;\n          setOutput(getX(${u}));\n        }\n      }\n    `}};var HC=class{constructor(t,e,n){this.variableNames=[\"x\"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:\"value\",type:\"float\"}],this.outputShape=e.map((h,g)=>h[0]+t[g]+h[1]);let o=t.length,s=zt(o),i=e.map(h=>h[0]).join(\",\"),a=e.map((h,g)=>h[0]+t[g]).join(\",\"),u=Qe(\"rc\",o),l=Qe(\"source\",o),c=`${u[o-1]} < ${this.outputShape[o-1]}`,p=o===1?\"source\":`vec2(${l.slice(-2).join()})`,m=[`${s} rc = outputLoc;`,`${u[o-1]} += 1;\n       if(${c}) {\n      `,o===1?\"\":`}\n       rc = outputLoc;\n       ${u[o-2]} += 1;\n       if(${u[o-2]} < ${this.outputShape[o-2]}) {`,o===1?\"\":`  ${u[o-1]} += 1;\n         if(${c}) {`],f=o===1?\"rc < start || rc >= end\":\"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))\",d=\"\";for(let h=0,g=o===1?2:4;h<g;h++)d+=`\n        ${m[h]}\n        if (${f}) {\n          result[${h}] = float(value);\n        } else {\n          ${s} source = rc - start;\n          result[${h}] = getChannel(getX(${l.join()}), ${p});\n        }\n      `;d+=o===1?\"} \":\"}}\",this.userCode=`\n      const ${s} start = ${s}(${i});\n      const ${s} end = ${s}(${a});\n\n      void main() {\n        ${s} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${d}\n        setOutput(result);\n      }\n    `}};var Tk=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,constantValue:i}=n;if(y.sizeFromShape(o.shape)===0){let l=s.map((c,p)=>c[0]+o.shape[p]+c[1]);return Cl({backend:e,attrs:{shape:l,value:i,dtype:o.dtype}})}let a=z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")?new HC(o.shape,s,i):new UC(o.shape,s,i),u=[[i]];return e.runWebGLProgram(a,[o],o.dtype,u)},U3={kernelName:xs,backendName:\"webgl\",kernelFunc:Tk};var lot=`\n  if(a < 0.0 && floor(b) < b){\n    return NAN;\n  }\n  if (b == 0.0) {\n    return 1.0;\n  }\n  return (round(mod(b, 2.0)) != 1) ?\n      pow(abs(a), b) : sign(a) * pow(abs(a), b);\n`,uot=`\n  // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.\n  vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));\n  vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);\n  vec4 result = multiplier * pow(abs(a), b);\n\n  // Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS\n  bvec4 isExpZero = equal(b, vec4(0.0));\n  result.r = isExpZero.r ? 1.0 : result.r;\n  result.g = isExpZero.g ? 1.0 : result.g;\n  result.b = isExpZero.b ? 1.0 : result.b;\n  result.a = isExpZero.a ? 1.0 : result.a;\n\n  bvec4 isNaN1 = lessThan(a, vec4(0.0));\n  bvec4 isNaN2 = lessThan(floor(b), b);\n  bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);\n  `+Yi+`\n  return result;\n`,cot=le({opSnippet:lot,packedOpSnippet:uot}),H3={kernelName:ys,backendName:\"webgl\",kernelFunc:cot};function pot(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=[],l=y.parseAxisParam(s,o.shape),c=l,p=v.getAxesPermutation(c,a),m=o;p!=null&&(m=Oe({inputs:{x:o},backend:e,attrs:{perm:p}}),c=v.getInnerMostAxes(c.length,a),u.push(m)),v.assertAxesAreInnerMostDims(\"prod\",c,a);let f;if(e.shouldExecuteOnCPU([m])){let d=e.texData.get(m.dataId).values,{outVals:h,outShape:g,outDtype:x}=RL(m.shape,m.dtype,d,c);f=e.makeTensorInfo(g,x,h)}else{let[d,h]=v.computeOutAndReduceShapes(m.shape,c),g=y.sizeFromShape(h),x=st({inputs:{x:m},backend:e,attrs:{shape:[-1,g]}}),b=Wu(o.dtype),w=Un(x,b,\"prod\",e);f=st({inputs:{x:w},backend:e,attrs:{shape:d}}),u.push(x),u.push(w)}if(i){u.push(f);let d=v.expandShapeToKeepDim(f.shape,l);f=st({inputs:{x:f},backend:e,attrs:{shape:d}})}return u.forEach(d=>e.disposeIntermediateTensorInfo(d)),f}var q3={kernelName:ws,backendName:\"webgl\",kernelFunc:pot};function mot(r){let{inputs:t,backend:e,attrs:n}=r,{paramsNestedSplits:o,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:a}=n,u=o.map(x=>e.readSync(x.dataId)),l=o.map(x=>x.shape),c=e.readSync(s.dataId),p=e.readSync(i.dataId),[m,f,d]=FL(u,l,c,s.shape,s.dtype,p,i.shape,a),h=m.map(x=>e.makeTensorInfo([x.length],\"int32\",x)),g=e.makeTensorInfo(d,s.dtype,f);return h.concat([g])}var K3={kernelName:Ap,backendName:\"webgl\",kernelFunc:mot};function fot(r){let{inputs:t,backend:e}=r,{starts:n,limits:o,deltas:s}=t,i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=OL(i,n.shape,n.dtype,a,o.shape,u,s.shape),p=e.makeTensorInfo([l.length],\"int32\",l),m=e.makeTensorInfo([c.length],n.dtype,c);return[p,m]}var j3={kernelName:$p,backendName:\"webgl\",kernelFunc:fot};function dot(r){let{inputs:t,backend:e,attrs:n}=r,{shape:o,values:s,defaultValue:i,rowPartitionTensors:a}=t,{rowPartitionTypes:u}=n,l=e.readSync(o.dataId),c=e.readSync(s.dataId),p=e.readSync(i.dataId),m=a.map(g=>e.readSync(g.dataId)),f=a.map(g=>g.shape),[d,h]=PL(l,o.shape,c,s.shape,s.dtype,p,i.shape,m,f,u);return e.makeTensorInfo(d,s.dtype,h)}var X3={kernelName:Dp,backendName:\"webgl\",kernelFunc:dot};var kk=r=>{let{backend:t,attrs:e}=r,{start:n,stop:o,step:s,dtype:i}=e,a=LL(n,o,s,i);return t.makeTensorInfo([a.length],i,a)},Y3={kernelName:Ol,backendName:\"webgl\",kernelFunc:kk};var hot=\"return 1.0 / x;\",got=Ct({opSnippet:hot}),Z3={kernelName:Pa,backendName:\"webgl\",kernelFunc:got};var xot=fr+`\n  return (x < 0.0) ? 0.0 : x;\n`,yot=`\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,bot=Ct({opSnippet:xot,packedOpSnippet:yot}),J3={kernelName:Cs,backendName:\"webgl\",kernelFunc:bot};var wot=fr+`\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n`,Cot=`\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,Iot=Ct({opSnippet:wot,packedOpSnippet:Cot}),Q3={kernelName:vs,backendName:\"webgl\",kernelFunc:Iot};var qC=class{constructor(t,e,n,o,s){this.variableNames=[\"A\"],this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m;s?m=\"(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)\":m=\"vec2(yRC) * effectiveInputOverOutputRatioRC\",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${c[0]/p[0]},\n          ${c[1]/p[1]});\n      const vec2 inputShapeRC = vec2(${a}.0, ${u}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${m};\n\n        // Compute the four integer indices.\n        ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));\n        ivec2 sourceCeilRC = ivec2(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);\n        float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);\n        float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);\n        float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);\n\n        vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);\n\n        float top = topLeft + (topRight - topLeft) * fracRC.y;\n        float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;\n        float newValue = top + (bottom - top) * fracRC.x;\n\n        setOutput(newValue);\n      }\n    `}};var KC=class{constructor(t,e,n,o,s){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m;s?m=\"(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)\":m=\"vec3(yRC) * effectiveInputOverOutputRatioRC\",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${c[0]/p[0]},\n          ${c[1]/p[1]},\n          ${c[1]/p[1]});\n      const vec3 inputShapeRC = vec3(${a}.0, ${u}.0,\n                                     ${u}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${m};\n\n        // Compute the four integer indices.\n        ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));\n        ivec3 sourceCeilRC = ivec3(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${l-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        // In parallel, construct four corners for all four components in\n        // packed 2x2 cell.\n        vec4 topLeft = vec4(\n          getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 bottomLeft = vec4(\n          getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 topRight = vec4(\n          getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec4 bottomRight = vec4(\n          getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);\n\n        vec4 top = mix(topLeft, topRight, fracRC.yyzz);\n        vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);\n        vec4 newValue = mix(top, bottom, fracRC.x);\n\n        setOutput(newValue);\n      }\n    `}};function Sot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,c=z().getBool(\"WEBGL_PACK_IMAGE_OPERATIONS\")?new KC(o.shape,u,l,s,i):new qC(o.shape,u,l,s,i);return e.runWebGLProgram(c,[o],\"float32\")}var tB={kernelName:Ss,backendName:\"webgl\",kernelFunc:Sot};var jC=class{constructor(t,e,n){this.variableNames=[\"dy\"],this.outputShape=[],this.outputShape=e;let[,o,s]=e,[,i,a]=t,u=[n&&i>1?o-1:o,n&&a>1?s-1:s],l=[n&&i>1?i-1:i,n&&a>1?a-1:a],c=u[0]/l[0],p=u[1]/l[1],m=1/c,f=1/p,d=Math.ceil(m)*2+2,h=Math.ceil(f)*2+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${c});\n        const float widthScale = float(${p});\n\n        const float invHeightScale = float(${m});\n        const float invWidthScale = float(${f});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${h});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(startRLerp - float(winHeight / 2));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(startCLerp - float(winWidth / 2));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${i}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${a}) {\n              continue;\n            }\n\n            float dxR = float(dyR) * heightScale;\n            int topDxRIndex = int(floor(dxR));\n            int bottomDxRIndex = int(min(ceil(dxR), ${o-1}.0));\n            float dxRLerp = dxR - float(topDxRIndex);\n            float inverseDxRLerp = 1.0 - dxRLerp;\n\n            float dxC = float(dyC) * widthScale;\n            int leftDxCIndex = int(floor(dxC));\n            int rightDxCIndex = int(min(ceil(dxC), ${s-1}.0));\n            float dxCLerp = dxC - float(leftDxCIndex);\n            float inverseDxCLerp = 1.0 - dxCLerp;\n\n            if (r == topDxRIndex && c == leftDxCIndex) {\n              // topLeft\n              accumulator +=\n                getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;\n            }\n\n            if (r == topDxRIndex && c == rightDxCIndex) {\n              // topRight\n              accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == leftDxCIndex) {\n              // bottomLeft\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == rightDxCIndex) {\n              // bottomRight\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}};function vot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n,a=new jC(s.shape,o.shape,i);return e.runWebGLProgram(a,[s],s.dtype)}var eB={kernelName:Op,backendName:\"webgl\",kernelFunc:vot};var XC=class{constructor(t,e,n,o,s){this.variableNames=[\"A\"],this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m=o?\"0.5\":\"0.0\",f;s?f=\"max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))\":f=\"vec2(yRC) * effectiveInputOverOutputRatioRC\",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${c[0]/p[0]},\n          ${c[1]/p[1]});\n      const vec2 inputShapeRC = vec2(${a}.0, ${u}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${f};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec2 sourceNearestRC = ivec2(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${m})));\n        float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);\n\n        setOutput(newValue);\n      }\n    `}};var YC=class{constructor(t,e,n,o,s){this.variableNames=[\"A\"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m=o?\"0.5\":\"0.0\",f;s?f=\"max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))\":f=\"vec3(yRC) * effectiveInputOverOutputRatioRC\",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${c[0]/p[0]},\n          ${c[1]/p[1]},\n          ${c[1]/p[1]});\n      const vec3 inputShapeRC = vec3(${a}.0, ${u}.0,\n                                     ${u}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${f};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec3 sourceNearestRC = ivec3(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${m})));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${l-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        vec4 newValue = vec4(\n          getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),\n          hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);\n\n        setOutput(newValue);\n      }\n    `}};function Not(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,c=z().getBool(\"WEBGL_PACK_IMAGE_OPERATIONS\")?new YC(o.shape,u,l,s,i):new XC(o.shape,u,l,s,i);return e.runWebGLProgram(c,[o],o.dtype)}var rB={kernelName:Is,backendName:\"webgl\",kernelFunc:Not};var ZC=class{constructor(t,e,n){this.variableNames=[\"dy\"],this.outputShape=[],this.outputShape=e;let[,o,s]=e,[,i,a]=t,u=[n&&i>1?o-1:o,n&&a>1?s-1:s],l=[n&&i>1?i-1:i,n&&a>1?a-1:a],c=u[0]/l[0],p=u[1]/l[1],m=1/c,f=1/p,d=Math.ceil(m)*2+2,h=Math.ceil(f)*2+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${c});\n        const float widthScale = float(${p});\n\n        const float invHeightScale = float(${m});\n        const float invWidthScale = float(${f});\n\n        const int winHeight = int(${d});\n        const int winWidth = int(${h});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(floor(startRLerp - float(winHeight / 2)));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(floor(startCLerp - float(winWidth / 2)));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${i}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${a}) {\n              continue;\n            }\n\n            float sourceFracRow =\n              float(${u[0]}) *\n                (float(dyR) / float(${l[0]}));\n\n            float sourceFracCol =\n                float(${u[1]}) *\n                  (float(dyC) / float(${l[1]}));\n\n            int sourceNearestRow = int(min(\n                float(int(${o}) - 1),\n                ${n} ? float(round(sourceFracRow)) :\n                                  float(floor(sourceFracRow))));\n\n            int sourceNearestCol = int(min(\n                float(int(${s}) - 1),\n                ${n} ? float(round(sourceFracCol)) :\n                                  float(floor(sourceFracCol))));\n\n            if (r == sourceNearestRow && c == sourceNearestCol) {\n              accumulator += getDy(b, dyR, dyC, d);\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}};function Tot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n,a=new ZC(s.shape,o.shape,i);return e.runWebGLProgram(a,[s],s.dtype)}var nB={kernelName:Fp,backendName:\"webgl\",kernelFunc:Tot};var JC=class{constructor(t,e){this.variableNames=[\"x\"];let n=t.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=t,n===1){this.userCode=`\n        void main() {\n          int coord = getOutputCoords();\n          setOutput(getX(${t[0]} - coord - 1));\n        }\n      `;return}let o=a=>e.indexOf(a)!==-1&&t[a]!==1?`${t[a]} - coords[${a}] - 1`:`coords[${a}]`,s=t.map((a,u)=>o(u)).join(\",\"),i=zt(n);this.userCode=`\n      void main() {\n        ${i} coords = getOutputCoords();\n        setOutput(getX(${s}));\n      }\n    `}};var QC=class{constructor(t,e){this.variableNames=[\"x\"],this.packedInputs=!0,this.packedOutput=!0;let n=t.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=t;let o=Qe(\"rc\",n),s=`${o[n-1]} + 1 < ${this.outputShape[n-1]}`,i=`${o[n-2]} + 1 < ${this.outputShape[n-2]}`,a=zt(n);n===1?this.userCode=`\n        void main(){\n          int rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = getChannel(getX(${t[0]} - rc - 1),\n            ${t[0]} - rc - 1);\n          if(${s}){\n              result.g = getChannel(getX(${t[0]} - (rc  + 1) - 1),\n                ${t[0]} - (rc  + 1) - 1);\n          }\n          setOutput(result);\n        }\n      `:this.userCode=`\n        void main() {\n          ${a} rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = ${u(o.slice())};\n          if(${s}){\n            result.g = ${l(o.slice())};\n          }\n          if(${i}) {\n            result.b = ${c(o.slice())};\n            if(${s}) {\n              result.a = ${p(o.slice())};\n            }\n          }\n          setOutput(result);\n        }\n    `;function u(d){return m(d)}function l(d){return d[n-1]=\"(\"+d[n-1]+\" + 1)\",m(d)}function c(d){return d[n-2]=\"(\"+d[n-2]+\" + 1)\",m(d)}function p(d){return d[n-1]=\"(\"+d[n-1]+\" + 1)\",d[n-2]=\"(\"+d[n-2]+\" + 1)\",m(d)}function m(d){let h=t.map((b,w)=>f(w,d)),g=h.join(\",\"),x=h.slice(-2).join(\",\");return`getChannel(getX(${g}), vec2(${x}))`}function f(d,h){return e.indexOf(d)!==-1&&t[d]!==1?`${t[d]} - ${h[d]} - 1`:`${h[d]}`}}};function kot(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n,i=o.shape.length,a=y.parseAxisParam(s,o.shape);if(i===0)return tr({inputs:{x:o},backend:e});let u=z().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\")?new QC(o.shape,a):new JC(o.shape,a);return e.runWebGLProgram(u,[o],o.dtype)}var oB={kernelName:Ns,backendName:\"webgl\",kernelFunc:kot};var tI=class{constructor(t,e){this.variableNames=[\"Image\"],this.outputShape=[],this.customUniforms=[{name:\"params\",type:\"vec4\"}];let n=t[1],o=t[2];this.outputShape=t;let s=\"\";typeof e==\"number\"?s=`float outputValue = ${e.toFixed(2)};`:s=`\n        vec3 fill = vec3(${e.join(\",\")});\n        float outputValue = fill[coords[3]];`,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n          int y = coords[1];\n          float coordXFloat = (float(x) - params[0]) * params[3] -\n            (float(y) - params[1]) * params[2];\n          float coordYFloat = (float(x) - params[0]) * params[2] +\n            (float(y) - params[1]) * params[3];\n          int coordX = int(round(coordXFloat + params[0]));\n          int coordY = int(round(coordYFloat + params[1]));\n          ${s}\n          if(coordX >= 0 && coordX < ${o} && coordY >= 0 && coordY < ${n}) {\n            outputValue = getImage(coords[0], coordY, coordX, coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}};var sB={kernelName:qa,backendName:\"webgl\",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,{radians:o,fillValue:s,center:i}=t,a=e,u=new tI(n.shape,s),[l,c]=v.getImageCenter(i,n.shape[1],n.shape[2]),p=[[l,c,Math.sin(o),Math.cos(o)]];return a.runWebGLProgram(u,[n],n.dtype,p)}};var Eot=`\n  // OpenGL ES does not support round function.\n  // The algorithm is based on banker's rounding.\n  float base = floor(x);\n  if ((x - base) < 0.5) {\n    return floor(x);\n  } else if ((x - base) > 0.5) {\n    return ceil(x);\n  } else {\n    if (mod(base, 2.0) == 0.0) {\n      return base;\n    } else {\n      return base + 1.0;\n    }\n  }\n`,_ot=Ct({opSnippet:Eot}),iB={kernelName:Ts,backendName:\"webgl\",kernelFunc:_ot};var Aot=\"return inversesqrt(x);\",$ot=Ct({opSnippet:Aot,cpuKernelImpl:ML}),aB={kernelName:ks,backendName:\"webgl\",kernelFunc:$ot};var $d=class{constructor(t,e,n,o,s,i,a=!0){this.variableNames=[\"updates\",\"indices\",\"defaultValue\"],this.outputShape=i;let u=zt(s.length),l=zt(i.length),c=\"\";n===1?c=\"i\":n===2&&(c=\"i, j\");let p=`getIndices(${c})`,m=\"\";o===1?m=\"i\":o===2&&(m=\"i, coords[1]\");let f=`getUpdates(${m})`,d=e>1?\"strides[j]\":\"strides\";this.userCode=`\n        ${u} strides = ${u}(${s});\n\n        void main() {\n          ${l} coords = getOutputCoords();\n          float sum = 0.0;\n          bool found = false;\n          for (int i = 0; i < ${t}; i++) {\n            int flattenedIndex = 0;\n            for (int j = 0; j < ${e}; j++) {\n              int index = round(${p});\n              flattenedIndex += index * ${d};\n            }\n            if (flattenedIndex == coords[0]) {\n              sum += ${f};\n              found = true;\n            }\n          }\n          setOutput(mix(getDefaultValue(), sum, float(found)));\n        }\n      `}};function Dot(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o,updates:s}=t,{shape:i}=n,{sliceRank:a,numUpdates:u,sliceSize:l,strides:c,outputSize:p}=v.calculateShapes(s,o,i),m=[p/l,l];if(p===0)return e.makeTensorInfo(i,o.dtype);let f=st({inputs:{x:o},backend:e,attrs:{shape:[u,a]}}),d=st({inputs:{x:s},backend:e,attrs:{shape:[u,l]}}),h=e.makeTensorInfo([],\"float32\",new Float32Array([0])),g=new $d(u,a,f.shape.length,d.shape.length,c,m),x=e.runWebGLProgram(g,[d,f,h],d.dtype),b=st({inputs:{x},backend:e,attrs:{shape:i}});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(h),b}var lB={kernelName:La,backendName:\"webgl\",kernelFunc:Dot};var eI=class{constructor(t,e,n,o){this.variableNames=[\"sortedSequence\",\"values\"],this.customUniforms=[{name:\"numInputs\",type:\"int\"}],this.outputShape=[t,n];let s=\"while (left < right) {\",i=`for (int i = 0; i < ${Math.ceil(Math.log2(e+1))}; ++i) { if (left >= right) break;`,a=z().getNumber(\"WEBGL_VERSION\")===2?s:i,u=o===\"left\"?\"<\":\"<=\";this.userCode=`\n       int findBound(int batch, float value) {\n         int left = 0;\n         int right = numInputs;\n         int mid;\n         ${a}\n           mid = (left + right) / 2;\n           if (getSortedSequence(batch, mid) ${u} value) {\n             left = mid + 1;\n           } else {\n             right = mid;\n           }\n         }\n         return right;\n       }\n\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int valueIndex = coords[1];\n\n         float value = getValues(batch, valueIndex);\n\n         setOutput(float(findBound(batch, value)));\n       }\n     `}};function Rot(r){let{inputs:t,backend:e,attrs:n}=r,{sortedSequence:o,values:s}=t,{side:i}=n,a=new eI(o.shape[0],o.shape[1],s.shape[1],i),u=[[o.shape[1]]];return e.runWebGLProgram(a,[o,s],\"int32\",u)}var uB={kernelName:Pp,backendName:\"webgl\",kernelFunc:Rot};var rI=class{constructor(t,e,n){this.variableNames=[\"c\",\"a\",\"b\"],this.outputShape=e;let o,s;if(n>4)throw Error(`Where for rank ${n} is not yet supported`);if(n===1)s=\"resRC\",o=\"resRC\";else{let a=[\"resRC.x\",\"resRC.y\",\"resRC.z\",\"resRC.w\"],u=[],l=[];for(let c=0;c<e.length;c++)l.push(`${a[c]}`),c<t&&u.push(`${a[c]}`);o=u.join(),s=l.join()}let i=zt(n);this.userCode=`\n      void main() {\n        ${i} resRC = getOutputCoords();\n        float cVal = getC(${o});\n        if (cVal >= 1.0) {\n          setOutput(getA(${s}));\n        } else {\n          setOutput(getB(${s}));\n        }\n      }\n    `}};function Fot(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t,i=new rI(n.shape.length,o.shape,o.shape.length);return e.runWebGLProgram(i,[n,o,s],sr(o.dtype,s.dtype))}var cB={kernelName:hi,backendName:\"webgl\",kernelFunc:Fot};var Oot=`\n  // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.\n  // see: https://arxiv.org/abs/1706.02515\n  float scaleAlpha = ${v.SELU_SCALEALPHA};\n  float scale = ${v.SELU_SCALE};\n  return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);\n`,Pot=Ct({opSnippet:Oot}),pB={kernelName:Ma,backendName:\"webgl\",kernelFunc:Pot};var Lot=Po+`\n  return 1.0 / (1.0 + exp(-1.0 * x));\n`,Mot=`\n  vec4 result = 1.0 / (1.0 + exp(-1.0 * x));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`,zot=Ct({opSnippet:Lot,packedOpSnippet:Mot,cpuKernelImpl:BL}),mB={kernelName:_s,backendName:\"webgl\",kernelFunc:zot};var Bot=`\n  if (isnan(x)) { return 0.0; }\n  return sign(x);\n`,Vot=Ct({opSnippet:Bot}),fB={kernelName:Ba,backendName:\"webgl\",kernelFunc:Vot};var Got=Po+`\n  return sin(x);\n`,Wot=Ct({opSnippet:Got}),dB={kernelName:Es,backendName:\"webgl\",kernelFunc:Wot};var Uot=`\n  float e2x = exp(x);\n  return (e2x - 1.0 / e2x) / 2.0;\n`,Hot=Ct({opSnippet:Uot}),hB={kernelName:za,backendName:\"webgl\",kernelFunc:Hot};var qot=`\n  float epsilon = 1.1920928955078125e-7;\n  float threshold = log(epsilon) + 2.0;\n\n  bool too_large = x > -threshold;\n  bool too_small = x < threshold;\n\n  float result;\n  float exp_x = exp(x);\n\n  if (too_large){\n    result = x;\n  }\n  else if (too_small){\n    result = exp_x;\n  }\n  else{\n    result = log(exp_x + 1.0);\n  }\n  return result;\n`,Kot=Ct({opSnippet:qot}),gB={kernelName:Va,backendName:\"webgl\",kernelFunc:Kot};var jot=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n;y.assert(o.shape.length<=4,()=>\"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet\");let a=s.reduce((x,b)=>x*b),u=[[0,0]];u.push(...i);for(let x=1+s.length;x<o.shape.length;++x)u.push([0,0]);let l=[],c=Tk({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),p=v.getReshaped(c.shape,s,a,!1),m=v.getPermuted(p.length,s.length,!1),f=v.getReshapedPermuted(c.shape,s,a,!1),d=st({inputs:{x:c},backend:e,attrs:{shape:p}}),h=Oe({inputs:{x:d},backend:e,attrs:{perm:m}}),g=st({inputs:{x:h},backend:e,attrs:{shape:f}});return l.push(c),l.push(d),l.push(h),l.forEach(x=>e.disposeIntermediateTensorInfo(x)),g},xB={kernelName:xi,backendName:\"webgl\",kernelFunc:jot};function Xot(r){let{inputs:t,backend:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:\n         ${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:\n         ${n.shape}`);if(o.shape.length!==1)throw new Error(`Values must be a vector, saw:\n         ${o.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:\n        ${i.shape}`);let a=e.readSync(n.dataId),u=e.readSync(o.dataId),l=e.readSync(s.dataId),c=e.readSync(i.dataId)[0],[p,m,f,d,h]=GL(a,n.shape,n.dtype,u,o.dtype,l,c);return[e.makeTensorInfo(m,n.dtype,p),e.makeTensorInfo([m[0]],o.dtype,f),e.makeTensorInfo([d.length],\"bool\",new Uint8Array(d.map(g=>Number(g)))),e.makeTensorInfo([h.length],n.dtype,new Int32Array(h))]}var yB={kernelName:Pl,backendName:\"webgl\",kernelFunc:Xot};function Yot(r){let{inputs:t,backend:e}=r,{inputIndices:n,inputShape:o,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(e.readSync(o.dataId)),a=e.readSync(n.dataId),u=Array.from(e.readSync(s.dataId)),[l,c,p]=WL(a,n.shape,n.dtype,i,u);return[e.makeTensorInfo(c,n.dtype,l),e.makeTensorInfo([p.length],s.dtype,new Int32Array(p))]}var bB={kernelName:Ga,backendName:\"webgl\",kernelFunc:Yot};function Zot(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape\n              ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape\n              ${s.shape}`);let i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=zw(i,n.shape,n.dtype,a,u,!0);return e.makeTensorInfo(c,n.dtype,l)}var wB={kernelName:Ll,backendName:\"webgl\",kernelFunc:Zot};function Jot(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error(\"Data should be at least 1 dimensional but received scalar\");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape\n             ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape\n             ${s.shape}`);let i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=zw(i,n.shape,n.dtype,a,u);return e.makeTensorInfo(c,n.dtype,l)}var CB={kernelName:Ml,backendName:\"webgl\",kernelFunc:Jot};function Qot(r){let{inputs:t,backend:e,attrs:n}=r,{sparseIndices:o,sparseValues:s,defaultValue:i}=t,{outputShape:a}=n,{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=v.calculateShapes(s,o,a),f=!1;if(s.dtype===\"string\"){let x=e.bufferSync(o),b=e.bufferSync(s),w=y.decodeString(e.readSync(i.dataId)[0]),C=zL(x,b,a,m,c,l,u,p,w,f);return e.makeTensorInfo(a,C.dtype,C.values)}let d=new $d(l,u,o.shape.length,s.shape.length,p,[m,1],f),h=e.runWebGLProgram(d,[s,o,i],s.dtype),g=st({inputs:{x:h},backend:e,attrs:{shape:a}});return e.disposeIntermediateTensorInfo(h),g}var IB={kernelName:Lp,backendName:\"webgl\",kernelFunc:Qot};function tst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=n,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=o.shape.length,c=new Array(l).fill(0),p=o.shape.slice();return u.map(m=>{let f=[...p];f[a]=m;let d=ri({inputs:{x:o},backend:e,attrs:{begin:c,size:f}});return c[a]+=m,d})}var SB={kernelName:yi,backendName:\"webgl\",kernelFunc:tst};var vB=\"return sqrt(x);\",est=Ct({opSnippet:vB,packedOpSnippet:vB,cpuKernelImpl:UL}),NB={kernelName:As,backendName:\"webgl\",kernelFunc:est};var rst=\"return x * x;\",nst=Ct({opSnippet:rst}),TB={kernelName:zl,backendName:\"webgl\",kernelFunc:nst};var kB=\"return (a - b) * (a - b);\",ost=le({opSnippet:kB,packedOpSnippet:kB}),EB={kernelName:Rs,backendName:\"webgl\",kernelFunc:ost};function sst({inputs:r,attrs:t,backend:e}){let{x:n}=r,o=fr+`\n    return x > 0.0 ? 1.0 : float(${t.alpha});\n  `,s=new tn(n.shape,o);return e.runWebGLProgram(s,[n],n.dtype)}var _B={kernelName:po,backendName:\"webgl\",kernelFunc:sst};var nI=class{constructor(t,e,n){this.variableNames=[\"x\"],this.outputShape=n;let o=n.length,s=zt(n.length),i=zt(n.length),a=\"\";if(o===1)a=\"coords * strides + begin\";else{let u=0;a=n.map((l,c)=>(u++,n.length===1?`coords * strides[${c}] + begin[${c}]`:`coords[${u-1}] * strides[${c}] + begin[${c}]`)).join(\",\")}this.userCode=`\n      ${s} begin = ${s}(${t});\n      ${s} strides = ${s}(${e});\n\n      void main() {\n        ${i} coords = getOutputCoords();\n        setOutput(getX(${a}));\n      }\n    `}};function ist(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n,{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=st({inputs:{x:o},backend:e,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let A=Le.computeOutShape(b,w,C),$=ri({inputs:{x:o},backend:e,attrs:{begin:b,size:A}});N=st({inputs:{x:$},backend:e,attrs:{shape:d}}),e.disposeIntermediateTensorInfo($)}else if(e.shouldExecuteOnCPU([o])){let $=e.readSync(o.dataId),F=wt(o.shape,o.dtype,$),P=HL(f,F,C,b);N=e.makeTensorInfo(d,o.dtype,P.values)}else{let $=new nI(b,C,f);N=e.runWebGLProgram($,[o],o.dtype)}let _=st({inputs:{x:N},backend:e,attrs:{shape:d}});return e.disposeIntermediateTensorInfo(N),_}var AB={kernelName:Wa,backendName:\"webgl\",kernelFunc:ist};function ast(r){let{inputs:t,backend:e,attrs:n}=r,{separator:o,nGramWidths:s,leftPad:i,rightPad:a,padWidth:u,preserveShortSequences:l}=n,{data:c,dataSplits:p}=t,m=e.readSync(c.dataId),f=e.readSync(p.dataId),[d,h]=qL(m,f,o,s,i,a,u,l);return[e.makeTensorInfo([d.length],\"string\",d),e.makeTensorInfo(p.shape,\"int32\",h)]}var $B={kernelName:Bl,backendName:\"webgl\",kernelFunc:ast};function lst(r){let{inputs:t,backend:e,attrs:n}=r,{skipEmpty:o}=n,{input:s,delimiter:i}=t;if(s.dtype!==\"string\")throw new Error(\"Input must be of datatype string\");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let a=e.readSync(s.dataId),u=e.readSync(i.dataId)[0],[l,c,p]=KL(a,u,o),m=c.length;return[e.makeTensorInfo([m,2],\"int32\",l),e.makeTensorInfo([m],\"string\",c),e.makeTensorInfo([2],\"int32\",new Int32Array(p))]}var DB={kernelName:Vl,backendName:\"webgl\",kernelFunc:lst};function ust(r){let{inputs:t,backend:e,attrs:n}=r,{numBuckets:o}=n,{input:s}=t;if(s.dtype!==\"string\")throw new Error(\"Input must be of datatype string\");if(o<=0)throw new Error(\"Number of buckets must be at least 1\");let i=e.readSync(s.dataId),a=jL(i,o);return e.makeTensorInfo(s.shape,\"int32\",a)}var RB={kernelName:Gl,backendName:\"webgl\",kernelFunc:ust};var cst=\"return tan(x);\",pst=Ct({opSnippet:cst}),FB={kernelName:Os,backendName:\"webgl\",kernelFunc:pst};var mst=`\n  float e2x = exp(-2.0 * abs(x));\n  return sign(x) * (1.0 - e2x) / (1.0 + e2x);\n`,fst=Ct({opSnippet:mst}),OB={kernelName:Ps,backendName:\"webgl\",kernelFunc:fst};var oI=class{constructor(t,e){this.variableNames=[\"A\"];let n=new Array(t.length);for(let i=0;i<n.length;i++)n[i]=t[i]*e[i];this.outputShape=n,this.rank=n.length;let o=zt(this.rank),s=dst(t);this.userCode=`\n      void main() {\n        ${o} resRC = getOutputCoords();\n        setOutput(getA(${s}));\n      }\n    `}};function dst(r){let t=r.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${r[0]})`;let e=[\"resRC.x\",\"resRC.y\",\"resRC.z\",\"resRC.w\",\"resRC.u\"],n=[];for(let o=0;o<r.length;o++)n.push(`imod(${e[o]}, ${r[o]})`);return n.join()}function Ek(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reps:s}=n;if(o.dtype===\"string\"||o.shape.length>5){let u=e.readSync(o.dataId),l=o.dtype===\"string\"?u.map(m=>y.decodeString(m)):u,c=wt(o.shape,o.dtype,l),p=YL(c,s);return e.makeTensorInfo(p.shape,p.dtype,p.values)}let i=new oI(o.shape,s);return e.runWebGLProgram(i,[o],o.dtype)}var PB={kernelName:Jn,backendName:\"webgl\",kernelFunc:Ek};var sI=class{constructor(t){this.variableNames=[\"x\",\"indices\"],this.customUniforms=[{name:\"n\",type:\"int\"},{name:\"firstPass\",type:\"int\"},{name:\"negativeInf\",type:\"float\"},{name:\"dir\",type:\"int\"},{name:\"inc\",type:\"int\"}],this.outputShape=t,this.userCode=`\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // We compare elements pair-wise within a group of size 2 * inc.\n         // The comparing rule for each group alternates between ascending\n         // and descending. Within each group, we compare each pair at\n         // positions i and i+inc. To decide whether an element at position i\n         // is x0 or x1, we mod it by 2 * inc, if the result is smaller than\n         // inc, it is in the first half of the group, we denote it as x0,\n         // otherwise we denote it as x1.\n         // For example, as shown in the Bitonic top K paper referenced above,\n         // Figure5(a) shows that element[1] is in the\n         // second half of the group when group size is 2, but it is in the\n         // first half of the group when group size is 4.\n\n         bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;\n         int i = isFirstInPair ? elemIdx : elemIdx - inc;\n\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));\n         float x0 = i0 < n ? getX(batch, i0) : negativeInf;\n         float x1 = i1 < n ? getX(batch, i1) : negativeInf;\n\n         // Denotes which direction indices are in (ascending or descending).\n         bool reverse = imod(elemIdx, 2 * dir) >= dir;\n         bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);\n         if (reverse == isGreater) { // Elements in opposite order of direction\n           int iTemp = i0;\n           i0 = i1;\n           i1 = iTemp;\n         }\n         if (isFirstInPair) {\n            setOutput(float(i0));\n         } else {\n            setOutput(float(i1));\n         }\n       }\n     `}},iI=class{constructor(t){this.variableNames=[\"x\",\"indices\"],this.customUniforms=[{name:\"n\",type:\"int\"},{name:\"firstPass\",type:\"int\"},{name:\"k\",type:\"int\"}],this.outputShape=t,this.userCode=`\n    void main() {\n         // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // The output size is half of the previous size.\n         // If the previous sequence is | | | | _ _ _ _  | | | |  _ _ _ _ (k=4),\n         // we only need to output the indices at positions |, the indices at\n         // positions _ can be thrown away, see Figure5(b) After Phase 2\n         // (Merge phase) in the Bitonic Top K paper referenced above.\n         // For example, the paper shows we only need to output the orange bars.\n         // The output sequence should look like this | | | | | | | |.\n         // Because the sequence is halved, to map the output index back\n         // to the previous sequence to find the corresponding value,\n         // we need to double the index. When we double the index,\n         // we basically interpolate a position, so 2i looks like\n         // | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position\n         // of each 2k positions by - elemIdx % k. E.g. for output at\n         // index 4,5,6,7, we want to get the corresponding element at\n         // original index 8,9,10,11, for output at index 8,9,10,11,\n         // we want to get the corresponding element at original index\n         // 16,17,18,19, so on and so forth.\n\n         int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));\n\n         float x0 = getX(batch, i0);\n         float x1 = i1 < n ? getX(batch, i1) : x0;\n\n         setOutput(x0 >= x1 ? float(i0) : float(i1));\n       }\n     `}};function Kc(r,t){t!==null&&r.disposeIntermediateTensorInfo(t)}function LB(r){let t=1;for(;t<r;)t*=2;return t}function hst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{k:s,sorted:i}=n,a=z().getNumber(\"TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD\"),u=z().getNumber(\"TOPK_K_CPU_HANDOFF_THRESHOLD\"),l=o.shape,c=l[l.length-1];if(e.shouldExecuteOnCPU([o])||c<a||s>u){let P=e.readSync(o.dataId),[V,G]=ZL(P,l,o.dtype,s,i);return[e.makeTensorInfo(V.shape,V.dtype,V.values),e.makeTensorInfo(G.shape,G.dtype,G.values)]}if(s===0)return l[l.length-1]=0,[e.makeTensorInfo(l,o.dtype,[]),e.makeTensorInfo(l,\"int32\",[])];if(c===1)return[o,Cl({attrs:{shape:l,dtype:\"int32\",value:0},backend:e})];let p=e.texData.get(o.dataId),m=p!==null&&p.isPacked,f=m?e.unpackTensor(o):o,h=y.sizeFromShape(l)/c,g=st({inputs:{x:f},attrs:{shape:[h,c]},backend:e});m&&Kc(e,f);let x=LB(s),b=LB(c),w=null,C=()=>w===null?[g,g]:[g,w],N=(P,V,G)=>{let W=C(),q=new sI(G),j=[[c],[w===null?1:0],[Number.NEGATIVE_INFINITY],[P],[V]],Y=w;w=e.runWebGLProgram(q,W,\"int32\",j),Kc(e,Y)};for(let P=1;P<x;P*=2){let V=P*2;for(let G=P;G>=1;G/=2)N(V,G,[h,b])}for(let P=b;P>x;P/=2){let V=C(),G=new iI([h,P/2]),q=[[c],[w===null?1:0],[x]],H=w;w=e.runWebGLProgram(G,V,\"int32\",q),Kc(e,H);let j=x/2,Y=j*2;for(let Z=j;Z>=1;Z/=2)N(Y,Z,w.shape)}let _=w;w=ri({inputs:{x:w},backend:e,attrs:{begin:0,size:[h,s]}}),Kc(e,_);let A=Ck({inputs:{x:g,indices:w},backend:e,attrs:{axis:1,batchDims:1}});Kc(e,g);let $=l.slice(0,-1);$.push(s),_=w,w=st({inputs:{x:w},attrs:{shape:$},backend:e}),Kc(e,_);let F=A;return A=st({inputs:{x:A},attrs:{shape:$},backend:e}),Kc(e,F),[A,w]}var MB={kernelName:Ua,backendName:\"webgl\",kernelFunc:hst};var aI=class{constructor(t,e,n,o,s,i){this.variableNames=[\"Image\",\"Transforms\"],this.outputShape=i;let a=n===\"nearest\"?1:2,u;switch(o){case\"constant\":u=1;break;case\"reflect\":u=2;break;case\"wrap\":u=3;break;case\"nearest\":u=4;break;default:u=1;break}this.userCode=`\n            float mapCoord(float outCoord, float len) {\n              float inCoord = outCoord;\n              if(${u} == 2) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    if (inCoord < sz2) {\n                      inCoord = sz2 * float(int(float(-inCoord / sz2))) +\n                      inCoord;\n                    }\n                    inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    inCoord -= sz2 * float(int(float(inCoord / sz2)));\n                    if (inCoord >= len) {\n                      inCoord = sz2 - inCoord - 1.0;\n                    }\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${u} == 3) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord -= len * float(int(float(inCoord / sz)));\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${u} == 4) {\n                return clamp(outCoord, 0.0, len - 1.0);\n              } else {\n                return outCoord;\n              }\n            }\n\n            float readWithFillValue(int batch, int coordY, int coordX,\n              int channel) {\n              float outputValue;\n              if (0 <= coordY && coordY < ${t} && 0 <= coordX && coordX < ${e}) {\n                  outputValue = getImage(batch, coordY, coordX, channel);\n              } else {\n                outputValue = float(${s});\n              }\n              return outputValue;\n            }\n\n            void main() {\n              ivec4 coords = getOutputCoords();\n              float outputValue;\n              int batch = coords[0];\n              int x = coords[2];\n              int y = coords[1];\n              int channel = coords[3];\n              float xf = float(x);\n              float yf = float(y);\n              float a1 = getTransforms(batch, 0);\n              float a2 = getTransforms(batch, 1);\n              float a3 = getTransforms(batch, 2);\n              float b1 = getTransforms(batch, 3);\n              float b2 = getTransforms(batch, 4);\n              float b3 = getTransforms(batch, 5);\n              float c1 = getTransforms(batch, 6);\n              float c2 = getTransforms(batch, 7);\n              float projection = c1 * xf + c2 * yf + 1.0;\n              if (projection == 0.0) {\n                outputValue = float(${s});\n              } else {\n                float inX = (a1 * xf + a2 * yf + a3) / projection;\n                float inY = (b1 * xf + b2 * yf + b3) / projection;\n                float mapX = mapCoord(inX, float(${e}));\n                float mapY = mapCoord(inY, float(${t}));\n\n                if (${a} == 1) {\n                  int coordY = int(round(mapY));\n                  int coordX = int(round(mapX));\n                  outputValue = readWithFillValue(batch, coordY, coordX,\n                    channel);\n                } else {\n                  float yFloor = floor(mapY);\n                  float xFloor = floor(mapX);\n                  float yCeil = yFloor + 1.0;\n                  float xCeil = xFloor + 1.0;\n                  float valueYFloor = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yFloor), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yFloor), int(xCeil), channel);\n                  float valueYCeil = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yCeil), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yCeil), int(xCeil), channel);\n                  outputValue = (yCeil - mapY) * valueYFloor +\n                  (mapY - yFloor) * valueYCeil;\n                }\n              }\n              setOutput(outputValue);\n            }\n        `}};function gst(r){let{inputs:t,backend:e,attrs:n}=r,{image:o,transforms:s}=t,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=n,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=new aI(p,m,i,a,u,g);return e.runWebGLProgram(x,[o,s],\"float32\")}var zB={kernelName:Ha,backendName:\"webgl\",kernelFunc:gst};function xst(r){let{inputs:t,attrs:e,backend:n}=r,{axis:o}=e,{x:s}=t;Qs(s,\"unique\"),console.warn(\"WARNING: \",\"UI might be locked temporarily as data is being downloaded\");let i=n.readSync(s.dataId),{outputValues:a,outputShape:u,indices:l}=JL(i,o,s.shape,s.dtype);return[n.makeTensorInfo(u,s.dtype,a),n.makeTensorInfo([l.length],\"int32\",l)]}var BB={kernelName:Mp,backendName:\"webgl\",kernelFunc:xst};function yst(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o,a=i.shape.length,u=o.shape[s],l=new Array(a-1),c=0;for(let h=0;h<a;h++)h!==s&&(l[c++]=i.shape[h]);let p=[],m=new Array(a).fill(0),f=i.shape.slice();f[s]=1;let d=new Array(u);for(let h=0;h<d.length;h++){m[s]=h;let g=ri({inputs:{x:i},backend:e,attrs:{begin:m,size:f}}),x=st({inputs:{x:g},backend:e,attrs:{shape:l}});d[h]=x,p.push(g)}return p.forEach(h=>e.disposeIntermediateTensorInfo(h)),d}var VB={kernelName:bi,backendName:\"webgl\",kernelFunc:yst};var lI=class{constructor(t,e){this.variableNames=[\"x\",\"segmentIds\"];let n=t.windowSize,o=t.batchSize,s=t.inSize,i=t.numSegments,a=i*Math.ceil(s/n);this.outputShape=[o,a];let u=\"0.0\",l=\"sumValue\",c=Math.floor(n/4)*4,p=n%4,m=`\n        sumValue += dot(values, segFilter);\n    `,f=\"\";s%n>0&&(f=`\n        if (inIdx < 0 || inIdx >= ${s}) {\n          return initializationValue;\n        }\n      `);let d=\"\";s%n>0&&(d=`\n        if (inIdx < 0 || inIdx >= ${s}) {\n          return -1.0;\n        }\n      `),this.userCode=`\n      const float initializationValue = ${u};\n\n      float getValue(int batch, int inIdx) {\n        ${f}\n        return getX(batch, inIdx);\n      }\n\n      float getSegmentIdAtIndex(int inIdx) {\n        ${d}\n        return getSegmentIds(inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = int(floor(float(outIdx) / float(\n          ${i})) * float(${n}));\n        int currentSeg = int(mod(float(outIdx), float(${i})));\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${c}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0\n          );\n\n          ${m}\n        }\n\n        int inIdx = inOffset + ${c};\n        if (${p===1}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          int inIdxSeg = int(getSegmentIdAtIndex(inIdx));\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            0,\n            0,\n            0\n          );\n\n          ${m}\n        } else if (${p===2}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n              0,\n              0\n          );\n\n          ${m}\n        } else if (${p===3}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            0\n          );\n\n          ${m}\n        }\n        setOutput(${l});\n      }\n    `}};function bst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,segmentIds:s}=t,{numSegments:i}=n,a=o.shape.length,u=[],l=0,c=v.getAxesPermutation([l],a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),u.push(p),l=v.getInnerMostAxes(1,a)[0]);let m=v.segment_util.computeOutShape(p.shape,l,i),f=y.sizeFromShape([p.shape[l]]),d=st({inputs:{x:p},backend:e,attrs:{shape:[-1,f]}});u.push(d);let h=Wu(o.dtype),g=(C,N,_,A,$)=>{let F=C.shape[0],P=C.shape[1],V=v.segment_util.segOpComputeOptimalWindowSize(P,$),G={windowSize:V,inSize:P,batchSize:F,numSegments:$},W=new lI(G,N),q=e.compileAndRun(W,[C,_],A);if(u.push(q),q.shape[1]===$)return q;let H=kk({backend:e,attrs:{start:0,stop:$,step:1,dtype:\"float32\"}}),j=Ek({inputs:{x:H},backend:e,attrs:{reps:[P/V]}});return u.push(H),u.push(j),g(q,N,j,A,$)},x=g(d,\"unsortedSegmentSum\",s,h,i),b=st({inputs:{x},backend:e,attrs:{shape:m}}),w=b;if(c!=null){u.push(b);let C=v.getUndoAxesPermutation(c);w=Oe({inputs:{x:w},backend:e,attrs:{perm:C}})}return u.forEach(C=>e.disposeIntermediateTensorInfo(C)),w}var GB={kernelName:Wl,backendName:\"webgl\",kernelFunc:bst};var wst=[kM,_M,AM,$M,RM,FM,OM,PM,zM,BM,VM,GM,WM,UM,HM,qM,KM,jM,XM,YM,ZM,QM,tz,ez,sz,az,lz,xM,cz,mz,fz,dz,hz,gz,xz,yz,bz,wz,Cz,vz,Nz,Tz,kz,Ez,_z,Az,$z,Dz,Rz,Fz,Oz,Pz,Lz,Mz,zz,Vz,Gz,Wz,Uz,qz,Kz,jz,Xz,Yz,Zz,Jz,Qz,t3,gM,e3,pz,r3,n3,o3,yM,s3,i3,a3,l3,u3,c3,p3,m3,f3,d3,g3,x3,y3,b3,w3,C3,S3,N3,T3,k3,E3,_3,F3,CM,O3,P3,L3,M3,rz,z3,G3,W3,U3,H3,bM,q3,K3,j3,X3,Y3,nz,A3,Z3,J3,Q3,SM,tB,eB,rB,nB,oB,sB,iB,aB,lB,uB,cB,pB,mB,fB,dB,hB,JM,R3,gB,xB,yB,bB,wB,CB,IB,SB,NB,TB,EB,_B,AB,$B,DB,RB,D3,NM,FB,OB,PB,MB,zB,TM,BB,VB,GB,B3];for(let r of wst)Lu(r);var qt;(function(r){r[r.float32=0]=\"float32\",r[r.int32=1]=\"int32\",r[r.bool=2]=\"bool\",r[r.string=3]=\"string\",r[r.complex64=4]=\"complex64\"})(qt||(qt={}));var Du;(function(r){r[r.linear=0]=\"linear\",r[r.relu=1]=\"relu\",r[r.relu6=2]=\"relu6\",r[r.prelu=3]=\"prelu\",r[r.leakyrelu=4]=\"leakyrelu\",r[r.sigmoid=5]=\"sigmoid\",r[r.elu=6]=\"elu\"})(Du||(Du={}));var WB;function Cst(r){WB=r.wasm.cwrap(Ci,null,[\"number\",\"array\",\"number\",\"number\",\"array\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Ist(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t;if(o.dtype!==\"float32\"||s.dtype!==\"float32\")throw new Error(\"_FusedMatMul for non non-float32 tensors not yet supported.\");let{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n,m=e.dataIdMap.get(o.dataId).id,f=e.dataIdMap.get(s.dataId).id,d=0;if(i!=null){let $=e.dataIdMap.get(i.dataId);if($.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${$.shape.length}.`);d=$.id}let h=a==null?0:e.dataIdMap.get(a.dataId).id,g=Du[c];if(g==null)throw new Error(`${c} activation not yet supported for FusedConv2D in the wasm backend.`);let x=u?o.shape[2]:o.shape[1],b=l?s.shape[1]:s.shape[2],w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)),C=e.makeOutput([...w,x,b],o.dtype),N=e.dataIdMap.get(C.dataId).id,_=new Uint8Array(new Int32Array(o.shape).buffer),A=new Uint8Array(new Int32Array(s.shape).buffer);return WB(m,_,o.shape.length,f,A,s.shape.length,u,l,g,d,h,p||0,N),C}var UB={kernelName:Ci,backendName:\"wasm\",setupFunc:Cst,kernelFunc:Ist};function se(r,t){let e;function n(s){e=s.wasm.cwrap(r,null,[\"number\",\"number\",\"number\"])}function o(s){let{backend:i,inputs:{x:a}}=s,u=i.dataIdMap.get(a.dataId).id,l=i.makeOutput(a.shape,t||a.dtype),c=i.dataIdMap.get(l.dataId).id;return y.sizeFromShape(l.shape)===0||e(u,qt[a.dtype],c),l}return{kernelName:r,backendName:\"wasm\",setupFunc:n,kernelFunc:o}}var HB=se(ii);function ue(r,t,e){let n;function o(i){n=i.wasm.cwrap(r,null,[\"number\",\"array\",\"number\",\"number\",\"array\",\"number\",\"number\",\"number\"])}function s(i){let{backend:a,inputs:u}=i,{a:l,b:c}=u,p=a.dataIdMap.get(l.dataId).id,m=a.dataIdMap.get(c.dataId).id,f=e!=null?e:l.dtype,d=v.assertAndGetBroadcastShape(l.shape,c.shape),h=a.makeOutput(d,f);if(y.sizeFromShape(d)===0)return h;let g=new Uint8Array(new Int32Array(l.shape).buffer),x=new Uint8Array(new Int32Array(c.shape).buffer),b=a.dataIdMap.get(h.dataId).id;return(()=>n(p,g,l.shape.length,m,x,c.shape.length,qt[l.dtype],b))(),h}return{kernelName:r,backendName:\"wasm\",setupFunc:o,kernelFunc:s}}var Sst=!0,qB=ue(Zn,Sst);var KB;function vst(r){KB=r.wasm.cwrap(Go,null,[\"array\",\"number\",\"number\",\"number\"])}function Nst(r){let{inputs:t,backend:e}=r,n=e.makeOutput(t[0].shape,t[0].dtype);if(y.sizeFromShape(n.shape)===0)return n;let o=t.map(a=>e.dataIdMap.get(a.dataId).id),s=new Uint8Array(new Int32Array(o).buffer),i=e.dataIdMap.get(n.dataId).id;return KB(s,o.length,qt[n.dtype],i),n}var jB={kernelName:Go,backendName:\"wasm\",setupFunc:vst,kernelFunc:Nst};function jc(r){let{inputs:{x:t},backend:e}=r;if(t.dtype===\"string\")return ur(e.readSync(t.dataId),t.shape,t.dtype);let n=e.makeOutput(t.shape,t.dtype),o=e.typedArrayFromHeap(t);return e.typedArrayFromHeap(n).set(o),n}var XB={kernelName:co,backendName:\"wasm\",kernelFunc:jc};var YB;function Tst(r){YB=r.wasm.cwrap(Qn,null,[\"number\",\"array\",\"number\",\"number\",\"number\",\"array\",\"number\"])}function ao(r){let{inputs:t,backend:e,attrs:n}=r,[o,s]=Est(t.x.shape,n.perm),i=!0;for(let d=0;d<s.length;d++)s[d]!==d&&(i=!1);let a=kst(t.x.shape,n.perm),u={dataId:t.x.dataId,shape:o,dtype:t.x.dtype};if(i){let d=jc({inputs:t,backend:e});return d.shape=a,d}let l=e.makeOutput(a,u.dtype),c=e.dataIdMap.get(u.dataId).id,p=e.dataIdMap.get(l.dataId).id,m=new Uint8Array(new Int32Array(s).buffer),f=new Uint8Array(new Int32Array(u.shape).buffer);return YB(c,f,u.shape.length,qt[u.dtype],p,m,s.length),l}function kst(r,t){let e=new Array(r.length);for(let n=0;n<e.length;n++)e[n]=r[t[n]];return e}function Est(r,t){let e=[],n=[];for(let o=0;o<r.length;++o)r[o]!==1&&e.push(r[o]),r[t[o]]!==1&&n.push(t[o]);for(let o=0;o<n.length;++o){let s=-1;for(let i=0;i<n.length;++i)n[i]>=o&&(s===-1||n[s]>n[i])&&(s=i);n[s]=o}return[e,n]}var ZB={kernelName:Qn,backendName:\"wasm\",kernelFunc:ao,setupFunc:Tst};function bn(r,t,e){let n=r.shape,o=r.shape.length,s=y.parseAxisParam(t,n),i=s,a=v.getAxesPermutation(i,o),u=null,l=!1;if(a!=null){let c=new Array(o);for(let f=0;f<c.length;f++)c[f]=n[a[f]];i=v.getInnerMostAxes(i.length,o),u=ao({inputs:{x:r},attrs:{perm:a},backend:e});let p=e.dataIdMap.get(r.dataId).id;e.dataIdMap.get(u.dataId).id!==p&&(l=!0)}return{transposed:u,originalAxes:s,axes:i,inputWasTransposed:l}}var JB;function _st(r){JB=r.wasm.cwrap(ia,null,[\"number, number, number\"])}function Ast(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims(\"all\",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;JB(u,x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var QB={kernelName:ia,backendName:\"wasm\",setupFunc:_st,kernelFunc:Ast};var tV;function $st(r){tV=r.wasm.cwrap(aa,null,[\"number, number, number\"])}function Dst(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims(\"any\",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;tV(u,x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var eV={kernelName:aa,backendName:\"wasm\",setupFunc:$st,kernelFunc:Dst};var rV;function Rst(r){rV=r.wasm.cwrap(Wo,null,[\"number\",\"number\",\"number\",\"number\",\"number\"])}function Fst(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o}=n,{x:s}=e,i=t.dataIdMap.get(s.dataId).id,a=i,u=s,{transposed:l,axes:c,inputWasTransposed:p}=bn(s,o,t);if(p){let x=t.dataIdMap.get(l.dataId).id;x!==i&&(u=l,a=x)}let m=u.shape.slice(0,-1),f=t.makeOutput(m,\"int32\"),d=t.dataIdMap.get(f.dataId).id,h=y.sizeFromShape(f.shape),g=u.shape[c[0]];return rV(a,qt[u.dtype],h,g,d),p&&t.disposeData(l.dataId),f}var nV={kernelName:Wo,backendName:\"wasm\",kernelFunc:Fst,setupFunc:Rst};var oV;function Ost(r){oV=r.wasm.cwrap(Uo,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Pst(r){let{inputs:t,attrs:e,backend:n}=r,o=t.x,s=n.dataIdMap.get(o.dataId).id,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=e,c=v.computePool2DInfo(o.shape,i,a,1,u,l),p=c.filterHeight,m=c.filterWidth,f=c.padInfo.top,d=c.padInfo.right,h=c.padInfo.bottom,g=c.padInfo.left,x=c.strideHeight,b=c.strideWidth,w=c.inChannels;if(c.dataFormat!==\"channelsLast\")throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);if(c.dilationWidth!==1||c.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${c.dilationHeight}, ${c.dilationWidth}].`);let C=n.makeOutput(c.outShape,\"float32\"),N=n.dataIdMap.get(C.dataId).id;return oV(s,o.shape[0],o.shape[1],o.shape[2],p,m,f,d,h,g,x,b,w,N),C}var sV={kernelName:Uo,backendName:\"wasm\",setupFunc:Ost,kernelFunc:Pst};function ar(r){let{inputs:t,attrs:e}=r,{x:n}=t,{shape:o}=e,s=y.sizeFromShape(n.shape),i=y.inferFromImplicitShape(o,s);return y.assert(s===y.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${n.shape}. New shape and old shape must have the same number of elements.`),r.backend.incRef(n.dataId),{dataId:n.dataId,shape:i,dtype:n.dtype}}var iV={kernelName:di,backendName:\"wasm\",kernelFunc:ar};var aV;function Lst(r){aV=r.wasm.cwrap(Ho,null,[\"number\",\"array\",\"number\",\"number\",\"array\",\"number\",\"number\",\"number\",\"number\"])}function Mst(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;if(o.dtype!==\"float32\"||s.dtype!==\"float32\")throw new Error(\"BatchMatMul for non non-float32 tensors not yet supported.\");let u=o.shape.length,l=s.shape.length,c=i?o.shape[u-2]:o.shape[u-1],p=a?s.shape[l-1]:s.shape[l-2],m=i?o.shape[u-1]:o.shape[u-2],f=a?s.shape[l-2]:s.shape[l-1],d=o.shape.slice(0,-2),h=s.shape.slice(0,-2),g=y.sizeFromShape(d),x=y.sizeFromShape(h),w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)).concat([m,f]);y.assert(c===p,()=>`Error in matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${o.shape} and ${s.shape} and transposeA=${i} and transposeB=${a} must match.`);let C=i?[g,c,m]:[g,m,c],N=a?[x,f,p]:[x,p,f],_=ar({inputs:{x:o},backend:e,attrs:{shape:C}}),A=ar({inputs:{x:s},backend:e,attrs:{shape:N}}),$=e.dataIdMap.get(_.dataId).id,F=e.dataIdMap.get(A.dataId).id,P=i?_.shape[2]:_.shape[1],V=a?A.shape[1]:A.shape[2],G=Math.max(g,x),W=e.makeOutput([G,P,V],_.dtype),q=e.dataIdMap.get(W.dataId).id,H=new Uint8Array(new Int32Array(_.shape).buffer),j=new Uint8Array(new Int32Array(A.shape).buffer);return aV($,H,_.shape.length,F,j,A.shape.length,i,a,q),e.disposeData(_.dataId),e.disposeData(A.dataId),W.shape=w,W}var lV={kernelName:Ho,backendName:\"wasm\",setupFunc:Lst,kernelFunc:Mst};function Lo(r){let{inputs:{x:t},attrs:{begin:e,size:n},backend:o}=r,[s,i]=Le.parseSliceParams(t,e,n),a=Le.isSliceContinous(t.shape,s,i),u=o.readSync(t.dataId),l=o.makeOutput(i,t.dtype),c=y.computeStrides(t.shape),p=o.dataIdMap.get(l.dataId);if(a){let d=Le.computeFlatOffset(s,c);return t.dtype===\"string\"?p.stringBytes=u.slice(d,d+y.sizeFromShape(i)):o.typedArrayFromHeap(l).set(u.subarray(d,d+y.sizeFromShape(i))),l}if(t.dtype===\"string\"){let d=$c(u,s,i,t.shape,t.dtype);return p.stringBytes=d,l}let m=o.typedArrayFromHeap(l),f=t.shape.length;if(f===2)zst(u,c[0],m,s,i);else if(f===3)Bst(u,c[0],c[1],m,s,i);else if(f===4)Vst(u,c[0],c[1],c[2],m,s,i);else{let d=$c(u,s,i,t.shape,t.dtype);m.set(d)}return l}function zst(r,t,e,n,o){let s=0,i=n[0],a=n[1],u=i+o[0];for(let l=i;l<u;l++){let c=l*t+a;e.set(r.subarray(c,c+o[1]),s),s+=o[1]}}function Bst(r,t,e,n,o,s){let i=0,a=o[0],u=o[1],l=o[2],c=a+s[0],p=u+s[1];for(let m=a;m<c;m++)for(let f=u;f<p;f++){let d=m*t+f*e+l;n.set(r.subarray(d,d+s[2]),i),i+=s[2]}}function Vst(r,t,e,n,o,s,i){let a=0,u=s[0],l=s[1],c=s[2],p=u+i[0],m=l+i[1],f=c+i[2],d=s[3];for(let h=u;h<p;h++)for(let g=l;g<m;g++)for(let x=c;x<f;x++){let b=h*t+g*e+x*n+d;o.set(r.subarray(b,b+i[3]),a),a+=i[3]}}var uV={kernelName:gi,backendName:\"wasm\",kernelFunc:Lo};function Gst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n,a=s.reduce((x,b)=>x*b),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=ar({inputs:{x:o},backend:e,attrs:{shape:u}}),d=ao({inputs:{x:f},backend:e,attrs:{perm:l}}),h=ar({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Lo({inputs:{x:h},backend:e,attrs:{begin:p,size:m}});return e.disposeData(f.dataId),e.disposeData(d.dataId),e.disposeData(f.dataId),g}var cV={kernelName:ai,backendName:\"wasm\",kernelFunc:Gst};function ni(r){let{inputs:{x:t},attrs:{dtype:e},backend:n}=r,o=n.makeOutput(t.shape,e),s=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(o).set(s),o}var pV={kernelName:lo,backendName:\"wasm\",kernelFunc:ni};var mV=se(qo);var fV;function Wst(r){fV=r.wasm.cwrap(uo,null,[\"number\",\"number\",\"number\",\"number\"])}function Ust(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{clipValueMin:s,clipValueMax:i}=n,a=e.dataIdMap.get(o.dataId).id,u=e.makeOutput(o.shape,o.dtype),l=e.dataIdMap.get(u.dataId).id;return fV(a,s,i,l),u}var dV={kernelName:uo,backendName:\"wasm\",setupFunc:Wst,kernelFunc:Ust};function _k(r){let{inputs:t,backend:e}=r,n=y.parseAxisParam(r.attrs.axis,t[0].shape)[0],o=t.map(f=>f.shape);v.assertParamsConsistent(o,n);let s=v.computeOutShape(t.map(f=>f.shape),n),i=t.filter(f=>y.sizeFromShape(f.shape)>0);if(i.length===1)return jc({inputs:{x:i[0]},backend:e});let a=e.makeOutput(s,t[0].dtype);if(y.sizeFromShape(s)===0)return a;if(i[0].dtype===\"string\"){let f=i.map(w=>{let C=y.sizeFromShape(w.shape.slice(n));return ar({inputs:{x:w},backend:e,attrs:{shape:[-1,C]}})}),d=f.map(w=>({vals:e.readSync(w.dataId),shape:w.shape}));s=v.computeOutShape(f.map(w=>w.shape),1);let h=f[0].shape[0]===1,g=Ec(d,s,t[0].dtype,h),x=v.computeOutShape(i.map(w=>w.shape),n);a.shape=x;let b=e.dataIdMap.get(a.dataId);return b.stringBytes=v.fromStringArrayToUint8(g),f.forEach(w=>e.disposeData(w.dataId)),a}let u=y.sizeFromShape(i[0].shape.slice(0,n)),l=0,c=i.map(f=>{let d=y.sizeFromShape(f.shape.slice(n));return l+=d,d}),p=i.map(f=>e.typedArrayFromHeap(f)),m=e.typedArrayFromHeap(a);for(let f=0;f<u;f++){let d=f*l;for(let h=0;h<p.length;h++){let g=c[h],x=f*g,b=p[h].subarray(x,x+g);m.set(b,d),d+=g}}return a}var hV={kernelName:li,backendName:\"wasm\",kernelFunc:_k};var gV;function Hst(r){gV=r.wasm.cwrap(Ko,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function qst(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s}=t,i=n.dataIdMap.get(o.dataId).id,a=n.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:p,dataFormat:m}=e,f=v.convertConv2DDataFormat(m),d=v.computeConv2DInfo(o.shape,s.shape,u,l,c,p,!1,f),h=d.filterHeight,g=d.filterWidth,x=d.padInfo.top,b=d.padInfo.right,w=d.padInfo.bottom,C=d.padInfo.left,N=d.dilationHeight,_=d.dilationWidth,A=d.strideHeight,$=d.strideWidth,F=d.inChannels,P=d.outChannels,V=d.padInfo.type===\"SAME\"?1:0;if(d.dataFormat!==\"channelsLast\")throw new Error(`wasm backend Conv2D does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);let G=n.makeOutput(d.outShape,\"float32\"),W=n.dataIdMap.get(G.dataId).id;return gV(i,o.shape[0],o.shape[1],o.shape[2],a,h,g,x,b,w,C,V,N,_,A,$,F,P,W),G}var xV={kernelName:Ko,backendName:\"wasm\",setupFunc:Hst,kernelFunc:qst};var yV;function Kst(r){yV=r.wasm.cwrap(jo,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function jst(r){let{backend:t,inputs:e,attrs:n}=r,{dy:o,filter:s}=e,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,inputShape:c}=n,p=1,m=v.convertConv2DDataFormat(u),f=v.computeConv2DInfo(c,s.shape,i,p,a,l,!1,m),{batchSize:d,filterHeight:h,filterWidth:g,inChannels:x,inHeight:b,inWidth:w,outChannels:C,outHeight:N,outWidth:_,strideHeight:A,strideWidth:$}=f,F=h-1-f.padInfo.top,P=g-1-f.padInfo.left,V=f.dataFormat===\"channelsLast\",G=y.computeStrides(f.inShape),W=y.computeStrides(o.shape),[q,H,j]=y.computeStrides(s.shape),Y=G[0],Z=V?G[1]:G[2],et=V?G[2]:1,rt=V?1:G[1],ot=W[0],at=V?W[1]:W[2],nt=V?W[2]:1,it=V?1:W[1],dt=t.makeOutput(f.inShape,\"float32\"),ht=t.dataIdMap.get(dt.dataId).id,bt=t.dataIdMap.get(o.dataId).id,Et=t.dataIdMap.get(s.dataId).id;return yV(bt,Et,d,h,g,b,w,x,N,_,C,A,$,F,P,q,H,j,Y,Z,et,rt,ot,at,nt,it,ht),dt}var bV={kernelName:jo,backendName:\"wasm\",setupFunc:Kst,kernelFunc:jst};var wV=se(Xo);var CV=se(Yo);var Ak;(function(r){r[r.bilinear=0]=\"bilinear\",r[r.nearest=1]=\"nearest\"})(Ak||(Ak={}));var IV;function Xst(r){IV=r.wasm.cwrap(da,null,[\"number\",\"number\",\"number\",\"number\",\"array\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Yst(r){let{backend:t,inputs:e,attrs:n}=r,{method:o,extrapolationValue:s,cropSize:i}=n,{image:a,boxes:u,boxInd:l}=e,c=u.shape[0],[p,m]=i,f=[c,p,m,a.shape[3]],d=t.dataIdMap.get(a.dataId),h;a.dtype!==\"float32\"&&(h=ni({backend:t,inputs:{x:a},attrs:{dtype:\"float32\"}}),d=t.dataIdMap.get(h.dataId));let g=d.id,x=t.dataIdMap.get(u.dataId).id,b=t.dataIdMap.get(l.dataId).id,w=t.makeOutput(f,\"float32\"),C=t.dataIdMap.get(w.dataId).id,N=new Uint8Array(new Int32Array(a.shape).buffer);return IV(g,x,b,c,N,p,m,Ak[o],s,C),h!=null&&t.disposeData(h.dataId),w}var SV={kernelName:da,backendName:\"wasm\",setupFunc:Xst,kernelFunc:Yst};var vV;function Zst(r){vV=r.wasm.cwrap(fa,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Jst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n,u=o.shape.length;y.assert(o.dtype===\"float32\"||o.dtype===\"int32\",()=>`cumprod does not support ${o.dtype} tensors in the WASM backend`);let l=v.getAxesPermutation([s],u),c=o;l!==null&&(c=ao({inputs:{x:o},attrs:{perm:l},backend:e}));let p=v.getInnerMostAxes(1,u)[0];v.assertAxesAreInnerMostDims(\"cumprod\",[p],u);let m=e.makeOutput(c.shape,c.dtype),f=c.shape[p],d=e.dataIdMap.get(c.dataId).id,h=e.dataIdMap.get(m.dataId).id;vV(d,i?1:0,a?1:0,f,h,qt[o.dtype]);let g=m;if(l!==null){let x=v.getUndoAxesPermutation(l);g=ao({inputs:{x:m},attrs:{perm:x},backend:e}),e.disposeData(c.dataId),e.disposeData(m.dataId)}return g}var NV={kernelName:fa,backendName:\"wasm\",setupFunc:Zst,kernelFunc:Jst};var TV;function Qst(r){TV=r.wasm.cwrap(Zo,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function tit(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n,u=o.shape.length;y.assert(o.dtype===\"float32\"||o.dtype===\"int32\",()=>`cumsum does not support ${o.dtype} tensors in the WASM backend`);let l=v.getAxesPermutation([s],u),c=o;l!==null&&(c=ao({inputs:{x:o},attrs:{perm:l},backend:e}));let p=v.getInnerMostAxes(1,u)[0];v.assertAxesAreInnerMostDims(\"cumsum\",[p],u);let m=e.makeOutput(c.shape,c.dtype),f=c.shape[p],d=e.dataIdMap.get(c.dataId).id,h=e.dataIdMap.get(m.dataId).id;TV(d,i?1:0,a?1:0,f,h,qt[o.dtype]);let g=m;if(l!==null){let x=v.getUndoAxesPermutation(l);g=ao({inputs:{x:m},attrs:{perm:x},backend:e}),e.disposeData(c.dataId),e.disposeData(m.dataId)}return g}var kV={kernelName:Zo,backendName:\"wasm\",setupFunc:Qst,kernelFunc:tit};var EV;function eit(r){EV=r.wasm.cwrap(ha,null,[\"number\",\"number\",\"number\",\"array\",\"number\",\"array\",\"array\",\"number\",\"number\"])}function rit(r){let{backend:t,inputs:e,attrs:n}=r,{x:o}=e,{blockSize:s,dataFormat:i}=n,a=o.shape[0],u=i===\"NHWC\"?o.shape[1]:o.shape[2],l=i===\"NHWC\"?o.shape[2]:o.shape[3],c=i===\"NHWC\"?o.shape[3]:o.shape[1],p=u*s,m=l*s,f=c/(s*s),d=i===\"NHWC\"?[a,p,m,f]:[a,f,p,m],h=t.makeOutput(d,\"float32\"),x=t.dataIdMap.get(o.dataId).id,b=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),w=new Uint8Array(new Int32Array(d).buffer),C=new Uint8Array(new Int32Array(y.computeStrides(d)).buffer),N=t.dataIdMap.get(h.dataId).id;return EV(x,s,i===\"NHWC\"?1:0,b,o.shape.length-1,w,C,d.length,N),h}var _V={kernelName:ha,backendName:\"wasm\",setupFunc:eit,kernelFunc:rit};var AV;function nit(r){AV=r.wasm.cwrap(Jo,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function oit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s}=t,i=n.dataIdMap.get(o.dataId).id,a=n.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:p}=e,m=l==null?[1,1]:l,f=v.computeConv2DInfo(o.shape,s.shape,u,m,c,p,!0),d=f.filterHeight,h=f.filterWidth,g=f.padInfo.top,x=f.padInfo.right,b=f.padInfo.bottom,w=f.padInfo.left,C=f.dilationHeight,N=f.dilationWidth,_=f.strideHeight,A=f.strideWidth,$=f.inChannels,F=f.outChannels,P=f.padInfo.type===\"SAME\"?1:0;if(f.dataFormat!==\"channelsLast\")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);let V=n.makeOutput(f.outShape,\"float32\"),G=n.dataIdMap.get(V.dataId).id;return AV(i,o.shape[0],o.shape[1],o.shape[2],a,d,h,g,x,b,w,P,C,N,_,A,$,F,G),V}var $V={kernelName:Jo,backendName:\"wasm\",setupFunc:nit,kernelFunc:oit};var DV=se(ts);var sit=!1,RV=ue(xa,sit,\"bool\");var FV=se(es,\"float32\");function uI(r){let{inputs:t,attrs:e,backend:n}=r,{input:o}=t,{dim:s}=e,i=o.shape.length,a=o.shape.slice(),u=s;return s<0&&(y.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),a.splice(u,0,1),ar({inputs:{x:o},backend:n,attrs:{shape:a}})}var OV={kernelName:ui,backendName:\"wasm\",kernelFunc:uI};function $k(r){let{attrs:{shape:t,value:e,dtype:n},backend:o}=r,s=o.makeOutput(t,n);return o.typedArrayFromHeap(s).fill(e),s}var PV={kernelName:Dl,backendName:\"wasm\",kernelFunc:$k};var LV;function iit(r){LV=r.wasm.cwrap(ba,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function ait(r){let{inputs:t,backend:e}=r,{image:n}=t,o=e.makeOutput(n.shape,n.dtype),s=e.dataIdMap.get(n.dataId).id,i=e.dataIdMap.get(o.dataId).id,[a,u,l,c]=n.shape;return LV(s,a,u,l,c,i),o}var MV={kernelName:ba,backendName:\"wasm\",kernelFunc:ait,setupFunc:iit};var zV=se(rs);var lit=!1,BV=ue(ns,lit);var VV;function uit(r){VV=r.wasm.cwrap(os,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function cit(r){let{backend:t,inputs:e,attrs:n}=r,{varianceEpsilon:o}=n,{x:s,mean:i,variance:a,offset:u,scale:l}=e,c=t.dataIdMap.get(s.dataId).id,p=t.dataIdMap.get(i.dataId).id,m=t.dataIdMap.get(a.dataId).id,f=u!=null?t.dataIdMap.get(u.dataId).id:0,d=l!=null?t.dataIdMap.get(l.dataId).id:0,h=t.makeOutput(s.shape,s.dtype);if(y.sizeFromShape(s.shape)===0)return h;let g=t.dataIdMap.get(h.dataId).id;return VV(c,p,m,f,d,o,g),h}var GV={kernelName:os,backendName:\"wasm\",setupFunc:uit,kernelFunc:cit};var WV;function pit(r){WV=r.wasm.cwrap(Ii,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function mit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dataFormat:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=e,h=v.computeConv2DInfo(o.shape,s.shape,u,c,l,m),g=Du[f];if(g==null)throw new Error(`${f} activation not yet supported for FusedConv2D in the wasm backend.`);let x=n.dataIdMap.get(o.dataId).id,b=n.dataIdMap.get(s.dataId).id,w=h.outChannels,C=0;if(i!=null){let nt=n.dataIdMap.get(i.dataId);if(nt.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${nt.shape.length}.`);if(nt.shape[0]!==w)throw new Error(`FusedConv2D bias shape (${nt.shape}) does not match the number of output channels (${w})`);C=nt.id}let N=h.filterHeight,_=h.filterWidth,A=h.padInfo.top,$=h.padInfo.right,F=h.padInfo.bottom,P=h.padInfo.left,V=h.dilationHeight,G=h.dilationWidth,W=h.strideHeight,q=h.strideWidth,H=h.inChannels,j=h.padInfo.type===\"SAME\"?1:0,Y=h.batchSize,Z=h.inHeight,et=h.inWidth;if(p!==\"NHWC\")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let rt=n.makeOutput(h.outShape,\"float32\"),ot=n.dataIdMap.get(rt.dataId).id,at=a==null?0:n.dataIdMap.get(a.dataId).id;return WV(x,Y,Z,et,b,N,_,C,A,$,F,P,j,V,G,W,q,H,w,g,at,d||0,ot),rt}var UV={kernelName:Ii,backendName:\"wasm\",setupFunc:pit,kernelFunc:mit};var HV;function fit(r){HV=r.wasm.cwrap(Si,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function dit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dataFormat:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=e,h=v.computeConv2DInfo(o.shape,s.shape,u,c,l,m,!0),g=Du[f];if(g==null)throw new Error(`${f} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let x=n.dataIdMap.get(o.dataId).id,b=n.dataIdMap.get(s.dataId).id,w=h.outChannels,C=0;if(i!=null){let nt=n.dataIdMap.get(i.dataId);if(nt.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${nt.shape.length}.`);if(nt.shape[0]!==w)throw new Error(`FusedDepthwiseConv2D bias shape (${nt.shape}) does not match the number of output channels (${w})`);C=nt.id}let N=h.filterHeight,_=h.filterWidth,A=h.padInfo.top,$=h.padInfo.right,F=h.padInfo.bottom,P=h.padInfo.left,V=h.dilationHeight,G=h.dilationWidth,W=h.strideHeight,q=h.strideWidth,H=h.inChannels,j=h.padInfo.type===\"SAME\"?1:0,Y=h.batchSize,Z=h.inHeight,et=h.inWidth;if(p!==\"NHWC\")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let rt=n.makeOutput(h.outShape,\"float32\"),ot=n.dataIdMap.get(rt.dataId).id,at=a==null?0:n.dataIdMap.get(a.dataId).id;return HV(x,Y,Z,et,b,N,_,C,A,$,F,P,j,V,G,W,q,H,w,g,at,d||0,ot),rt}var qV={kernelName:Si,backendName:\"wasm\",setupFunc:fit,kernelFunc:dit};var KV;function hit(r){KV=r.wasm.cwrap(wa,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"array\",\"number\"])}function git(r){let{backend:t,inputs:e}=r,{params:n,indices:o}=e,[s,i,a,u]=ox.prepareAndValidate(n,o),l=t.makeOutput(s,n.dtype);if(i===0)return l;let c=o.shape,p=c[c.length-1],f=t.dataIdMap.get(n.dataId).id,h=t.dataIdMap.get(o.dataId).id,g=new Uint8Array(new Int32Array(u).buffer),x=t.dataIdMap.get(l.dataId).id;return KV(f,qt[n.dtype],h,i,p,a,g,x),l}var jV={kernelName:wa,backendName:\"wasm\",setupFunc:hit,kernelFunc:git};var XV;function xit(r){XV=r.wasm.cwrap(\"Gather\",null,[\"number\",\"number\",\"array\",\"number\",\"number\",\"number\",\"array\",\"number\"])}function yit(r){let{backend:t,inputs:e,attrs:n}=r,{x:o,indices:s}=e,{axis:i,batchDims:a}=n,u=y.parseAxisParam(i,o.shape)[0],l=t.readSync(s.dataId),c=o.shape[u];for(let F=0;F<l.length;++F){let P=l[F];y.assert(P<=c-1&&P>=0,()=>`GatherV2: the index value ${P} is not in [0, ${c-1}]`)}let p=v.segment_util.collectGatherOpShapeInfo(o,s,u,a),m=ar({inputs:{x:o},attrs:{shape:[p.batchSize,p.outerSize,p.dimSize,p.sliceSize]},backend:t}),f=y.sizeFromShape(s.shape),d=ar({inputs:{x:s},attrs:{shape:[p.batchSize,f/p.batchSize]},backend:t}),h=[p.batchSize,p.outerSize,f/p.batchSize,p.sliceSize],g=t.makeOutput(h,o.dtype);if(y.sizeFromShape(o.shape)===0)return g;let x=m.shape.length-1,w=t.dataIdMap.get(m.dataId).id,N=t.dataIdMap.get(d.dataId).id,_=t.dataIdMap.get(g.dataId).id,A=new Uint8Array(new Int32Array(y.computeStrides(m.shape)).buffer),$=new Uint8Array(new Int32Array(y.computeStrides(h)).buffer);return XV(w,qt[o.dtype],A,x,N,p.batchSize,$,_),t.disposeData(m.dataId),t.disposeData(d.dataId),g.shape=p.outputShape,g}var YV={kernelName:ci,backendName:\"wasm\",setupFunc:xit,kernelFunc:yit};var bit=!1,ZV=ue(Ca,bit,\"bool\");var wit=!1,JV=ue(ss,wit,\"bool\");var QV;function Cit(r){QV=r.wasm.cwrap(is,null,[\"number\",\"number\",\"number\",\"number\"])}function Iit(r){let{inputs:{x:t},attrs:{alpha:e},backend:n}=r,o=n.dataIdMap.get(t.dataId).id,s=n.makeOutput(t.shape,\"float32\");if(y.sizeFromShape(t.shape)!==0){let i=n.dataIdMap.get(s.dataId).id;QV(o,qt[t.dtype],e,i)}return s}var tG={kernelName:is,backendName:\"wasm\",setupFunc:Cit,kernelFunc:Iit};var Sit=!1,eG=ue(Na,Sit,\"bool\");var vit=!1,rG=ue(Ta,vit,\"bool\");var nG=se(as);var Nit=!1,oG=ue(Ea,Nit,\"bool\");var sG=se(_a);var Tit=!1,iG=ue(Aa,Tit,\"bool\");var kit=!1,aG=ue(m1,kit,\"bool\");var lG;function Eit(r){lG=r.wasm.cwrap(ls,null,[\"number\",\"number\",\"number\",\"number\"])}function _it(r){let{backend:t,inputs:e,attrs:n}=r,{reductionIndices:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims(\"max\",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;lG(u,qt[i.dtype],x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var uG={kernelName:ls,backendName:\"wasm\",setupFunc:Eit,kernelFunc:_it};var Ait=!1,cG=ue(us,Ait);var pG;function $it(r){pG=r.wasm.cwrap(cs,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Dit(r){let{inputs:t,attrs:e,backend:n}=r,o=t.x,s=n.dataIdMap.get(o.dataId).id;y.assert(o.dtype===\"float32\",()=>`Error in MaxPool: only float32 input is supported. Got ${o.dtype}.`);let{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=e,c=v.computePool2DInfo(o.shape,i,a,1,u,l),p=c.filterHeight,m=c.filterWidth,f=c.padInfo.top,d=c.padInfo.right,h=c.padInfo.bottom,g=c.padInfo.left,x=c.dilationHeight,b=c.dilationWidth,w=c.strideHeight,C=c.strideWidth,N=c.inChannels,_=c.outChannels;if(c.dataFormat!==\"channelsLast\")throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);let A=n.makeOutput(c.outShape,\"float32\"),$=n.dataIdMap.get(A.dataId).id;return pG(s,o.shape[0],o.shape[1],o.shape[2],p,m,f,d,h,g,x,b,w,C,N,_,$),A}var mG={kernelName:cs,backendName:\"wasm\",setupFunc:$it,kernelFunc:Dit};var fG;function Rit(r){fG=r.wasm.cwrap(ps,null,[\"number, number, number\"])}function Fit(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let C=t.dataIdMap.get(c.dataId).id;C!==a&&(l=c,u=C,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims(\"mean\",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=l;l.dtype!==\"float32\"&&(b=ni({backend:t,inputs:{x:l},attrs:{dtype:\"float32\"}}),u=t.dataIdMap.get(b.dataId).id);let w=t.makeOutput(h,\"float32\");if(y.sizeFromShape(l.shape)!==0){let C=t.dataIdMap.get(w.dataId).id;fG(u,x,C)}if(f&&t.disposeData(c.dataId),s){let C=v.expandShapeToKeepDim(w.shape,m);w.shape=C}return l.dtype!==\"float32\"&&t.disposeData(b.dataId),w}var dG={kernelName:ps,backendName:\"wasm\",setupFunc:Rit,kernelFunc:Fit};var hG;function Oit(r){hG=r.wasm.cwrap(ms,null,[\"number\",\"number\",\"number\",\"number\"])}function Pit(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w)}let d=l.shape.length;v.assertAxesAreInnerMostDims(\"min\",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;hG(u,qt[i.dtype],x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var gG={kernelName:ms,backendName:\"wasm\",setupFunc:Oit,kernelFunc:Pit};var Lit=!1,xG=ue(fs,Lit);var Dk;(function(r){r[r.reflect=0]=\"reflect\",r[r.symmetric=1]=\"symmetric\"})(Dk||(Dk={}));var yG;function Mit(r){yG=r.wasm.cwrap(ds,null,[\"number\",\"array\",\"number\",\"number\",\"array\",\"array\",\"number\",\"number\"])}function zit(r){let{inputs:{x:t},backend:e,attrs:{paddings:n,mode:o}}=r,s=n.map((d,h)=>d[0]+t.shape[h]+d[1]),i=e.dataIdMap.get(t.dataId).id,a=e.makeOutput(s,t.dtype),u=e.dataIdMap.get(a.dataId).id,l=new Uint8Array(new Int32Array(t.shape).buffer),c=n.map(d=>d[0]),p=n.map(d=>d[1]),m=new Uint8Array(new Int32Array(c).buffer),f=new Uint8Array(new Int32Array(p).buffer);return yG(i,l,t.shape.length,qt[t.dtype],m,f,Dk[o],u),a}var bG={kernelName:ds,backendName:\"wasm\",kernelFunc:zit,setupFunc:Mit};var Bit=!0,wG=ue(hs,Bit);var CG=se(pi);function Dd(r,t){let e=new Int32Array(r.wasm.HEAPU8.buffer,t,4),n=e[0],o=e[1],s=e[2],i=e[3];return r.wasm._free(t),{pSelectedIndices:n,selectedSize:o,pSelectedScores:s,pValidOutputs:i}}var IG;function Vit(r){IG=r.wasm.cwrap(Ra,\"number\",[\"number\",\"number\",\"number\",\"number\",\"number\"])}function Git(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i}=n,{boxes:a,scores:u}=e,l=t.dataIdMap.get(a.dataId).id,c=t.dataIdMap.get(u.dataId).id,p=IG(l,c,s,o,i),{pSelectedIndices:m,selectedSize:f,pSelectedScores:d,pValidOutputs:h}=Dd(t,p);return t.wasm._free(d),t.wasm._free(h),t.makeOutput([f],\"int32\",m)}var SG={kernelName:Ra,backendName:\"wasm\",setupFunc:Vit,kernelFunc:Git};var vG;function Wit(r){vG=r.wasm.cwrap(Fa,\"number\",[\"number\",\"number\",\"number\",\"number\",\"number\",\"bool\"])}function Uit(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:a}=n,{boxes:u,scores:l}=e,c=t.dataIdMap.get(u.dataId).id,p=t.dataIdMap.get(l.dataId).id,m=vG(c,p,s,o,i,a),{pSelectedIndices:f,selectedSize:d,pSelectedScores:h,pValidOutputs:g}=Dd(t,m);t.wasm._free(h);let x=t.makeOutput([d],\"int32\",f),b=t.makeOutput([],\"int32\",g);return[x,b]}var NG={kernelName:Fa,backendName:\"wasm\",setupFunc:Wit,kernelFunc:Uit};var TG;function Hit(r){TG=r.wasm.cwrap(Oa,\"number\",[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function qit(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i,softNmsSigma:a}=n,{boxes:u,scores:l}=e,c=t.dataIdMap.get(u.dataId).id,p=t.dataIdMap.get(l.dataId).id,m=TG(c,p,s,o,i,a),{pSelectedIndices:f,selectedSize:d,pSelectedScores:h,pValidOutputs:g}=Dd(t,m);t.wasm._free(g);let x=t.makeOutput([d],\"int32\",f),b=t.makeOutput([d],\"float32\",h);return[x,b]}var kG={kernelName:Oa,backendName:\"wasm\",setupFunc:Hit,kernelFunc:qit};var Kit=!1,EG=ue(Da,Kit,\"bool\");var _G;function jit(r){_G=r.wasm.cwrap(gs,null,[\"number\",\"number\",\"number\",\"number\",\"number\"])}function Xit(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n,l=e.makeOutput([...o.shape,i],s),c=e.dataIdMap.get(l.dataId).id,m=e.dataIdMap.get(o.dataId).id;return _G(m,i,a,u,c),l}var AG={kernelName:gs,backendName:\"wasm\",setupFunc:jit,kernelFunc:Xit};function Yit(r){let{inputs:{x:t},backend:e}=r,n=e.makeOutput(t.shape,t.dtype);return e.typedArrayFromHeap(n).fill(1),n}var $G={kernelName:mi,backendName:\"wasm\",kernelFunc:Yit};function Zit(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return uI({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,\"All tensors passed to stack must have matching shapes\"),y.assert(i===c.dtype,()=>\"All tensors passed to stack must have matching dtypes\")});let a=[],u=t.map(c=>{let p=uI({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=_k({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeData(c.dataId)),l}var DG={kernelName:fi,backendName:\"wasm\",kernelFunc:Zit};var RG;function Jit(r){RG=r.wasm.cwrap(xs,null,[\"number\",\"array\",\"number\",\"number\",\"array\",\"array\",\"number\",\"number\"])}function Qit(r){let{inputs:{x:t},backend:e,attrs:{paddings:n,constantValue:o}}=r,s=n.map((h,g)=>h[0]+t.shape[g]+h[1]);if(y.sizeFromShape(t.shape)===0)return $k({backend:e,attrs:{shape:s,value:o,dtype:t.dtype}});let i=e.dataIdMap.get(t.dataId).id,a=e.makeOutput(s,t.dtype),l=e.dataIdMap.get(a.dataId).id,c=new Uint8Array(new Int32Array(t.shape).buffer),p=n.map(h=>h[0]),m=n.map(h=>h[1]),f=new Uint8Array(new Int32Array(p).buffer),d=new Uint8Array(new Int32Array(m).buffer);return RG(i,c,t.shape.length,qt[t.dtype],f,d,o,l),a}var cI={kernelName:xs,backendName:\"wasm\",kernelFunc:Qit,setupFunc:Jit};var tat=!1,FG=ue(ys,tat);var OG;function eat(r){OG=r.wasm.cwrap(bs,null,[\"number\",\"number\",\"number\"])}function rat(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t,s=e.dataIdMap.get(n.dataId).id,i=e.dataIdMap.get(o.dataId).id,a=s,u=n,l=u;u.dtype!==\"float32\"&&(l=ni({backend:e,inputs:{x:n},attrs:{dtype:\"float32\"}}),a=e.dataIdMap.get(l.dataId).id);let c=e.makeOutput(n.shape,\"float32\"),p=e.dataIdMap.get(c.dataId).id;return OG(a,i,p),u.dtype!==\"float32\"&&e.disposeData(l.dataId),c}var PG={kernelName:bs,backendName:\"wasm\",setupFunc:eat,kernelFunc:rat};var LG;function nat(r){LG=r.wasm.cwrap(ws,null,[\"number\",\"number\",\"number\",\"number\"])}function oat(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims(\"prod\",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;LG(u,x,qt[b.dtype],w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var MG={kernelName:ws,backendName:\"wasm\",setupFunc:nat,kernelFunc:oat};var sat=r=>{let{backend:t,attrs:e}=r,{start:n,stop:o,step:s,dtype:i}=e,a=Ac(n,o,s,i),u=t.makeOutput([a.length],i);return t.typedArrayFromHeap(u).set(a),u},zG={kernelName:Ol,backendName:\"wasm\",kernelFunc:sat};var iat=!0,BG=ue(Qo,iat);var VG=se(Cs);var GG=se(vs);var WG;function aat(r){WG=r.wasm.cwrap(Ss,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function lat(r){let{backend:t,inputs:e,attrs:n}=r,{images:o}=e,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,[c,p,m,f]=o.shape,d=[c,u,l,f],h=t.dataIdMap.get(o.dataId),g;h.dtype!==\"float32\"&&(g=ni({backend:t,inputs:{x:o},attrs:{dtype:\"float32\"}}),h=t.dataIdMap.get(g.dataId));let x=h.id,b=t.makeOutput(d,\"float32\");if(y.sizeFromShape(o.shape)===0)return b;let w=t.dataIdMap.get(b.dataId).id;return WG(x,c,p,m,f,u,l,s?1:0,i?1:0,w),g!=null&&t.disposeData(g.dataId),b}var UG={kernelName:Ss,backendName:\"wasm\",setupFunc:aat,kernelFunc:lat};var HG;function uat(r){HG=r.wasm.cwrap(Is,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function cat(r){let{backend:t,inputs:e,attrs:n}=r,{images:o}=e,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,[c,p,m,f]=o.shape,d=[c,u,l,f],h=t.makeOutput(d,\"float32\");if(y.sizeFromShape(o.shape)===0)return h;let g=t.dataIdMap.get(o.dataId),x;g.dtype!==\"float32\"&&(x=ni({backend:t,inputs:{x:o},attrs:{dtype:\"float32\"}}),g=t.dataIdMap.get(x.dataId));let b=g.id,w=t.dataIdMap.get(h.dataId).id;return HG(b,c,p,m,f,u,l,s?1:0,i?1:0,w),x!=null&&t.disposeData(x.dataId),h}var qG={kernelName:Is,backendName:\"wasm\",setupFunc:uat,kernelFunc:cat};var KG;function pat(r){KG=r.wasm.cwrap(Ns,null,[\"number\",\"array\",\"number\",\"array\",\"number\",\"number\"])}function mat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n,i=y.parseAxisParam(s,o.shape);if(o.shape.length===0)return jc({inputs:{x:o},backend:e});let a=e.makeOutput(o.shape,o.dtype),u=e.dataIdMap.get(o.dataId).id,l=e.dataIdMap.get(a.dataId).id,c=new Uint8Array(new Int32Array(i).buffer),p=new Uint8Array(new Int32Array(o.shape).buffer);KG(u,c,i.length,p,o.shape.length,l);let m=ar({inputs:{x:a},attrs:{shape:o.shape},backend:e});return e.disposeData(a.dataId),m}var jG={kernelName:Ns,backendName:\"wasm\",kernelFunc:mat,setupFunc:pat};var XG;function fat(r){XG=r.wasm.cwrap(qa,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"array\",\"number\",\"number\"])}function dat(r){let{inputs:t,backend:e,attrs:n}=r,{image:o}=t,{radians:s,fillValue:i,center:a}=n,u=e.makeOutput(o.shape,o.dtype),l=e.dataIdMap.get(o.dataId).id,c=e.dataIdMap.get(u.dataId).id,[p,m,f,d]=o.shape,[h,g]=v.getImageCenter(a,m,f),x=i===0,b=255,w=typeof i==\"number\"?[i,i,i,x?0:b]:[...i,b],C=new Uint8Array(new Int32Array(w).buffer);return XG(l,p,m,f,d,s,h,g,C,w.length,c),u}var YG={kernelName:qa,backendName:\"wasm\",kernelFunc:dat,setupFunc:fat};var ZG=se(Ts);var JG=se(ks);var QG;function hat(r){QG=r.wasm.cwrap(La,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"array\",\"number\",\"number\"])}function gat(r){let{backend:t,inputs:e,attrs:n}=r,{indices:o,updates:s}=e,{shape:i}=n,a=t.makeOutput(i,s.dtype);if(y.sizeFromShape(i)===0)return a;let{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=lh.calculateShapes(s,o,i),d=t.dataIdMap.get(o.dataId).id,g=t.dataIdMap.get(s.dataId).id,x=new Uint8Array(new Int32Array(p).buffer),b=t.dataIdMap.get(a.dataId).id;return QG(d,g,qt[s.dtype],u,l,c,x,m,b),a}var tW={kernelName:La,backendName:\"wasm\",setupFunc:hat,kernelFunc:gat};var eW;function xat(r){eW=r.wasm.cwrap(\"SelectV2\",null,[\"number\",\"number\",\"number\",\"number\",\"number\"])}function yat(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t,i=e.dataIdMap.get(n.dataId).id,a=e.dataIdMap.get(o.dataId).id,u=e.dataIdMap.get(s.dataId).id,l=e.makeOutput(o.shape,o.dtype),c=e.dataIdMap.get(l.dataId).id,p=n.shape.length,m=o.shape.length,f=p===0||p>1||m===1?1:y.sizeFromShape(o.shape.slice(1));return eW(i,a,u,f,c),l}var rW={kernelName:hi,backendName:\"wasm\",kernelFunc:yat,setupFunc:xat};var nW;function bat(r){nW=r.wasm.cwrap(_s,null,[\"number\",\"number\"])}function wat(r){let{backend:t,inputs:{x:e}}=r,n=t.dataIdMap.get(e.dataId).id,o=t.makeOutput(e.shape,e.dtype),s=t.dataIdMap.get(o.dataId).id;return y.sizeFromShape(o.shape)===0||nW(n,s),o}var oW={kernelName:\"Sigmoid\",backendName:\"wasm\",setupFunc:bat,kernelFunc:wat};var sW=se(Es);var iW;function Cat(r){iW=r.wasm.cwrap(Ds,null,[\"number\",\"number\",\"number\",\"number\"])}function Iat(r){let{backend:t,inputs:{logits:e},attrs:{dim:n}}=r,o=t.dataIdMap.get(e.dataId).id,s=t.makeOutput(e.shape,e.dtype),i=t.dataIdMap.get(s.dataId).id,a=e.shape[n],u=y.sizeFromShape(e.shape)/a;return y.sizeFromShape(s.shape)===0||iW(o,i,a,u),s}var aW={kernelName:Ds,backendName:\"wasm\",setupFunc:Cat,kernelFunc:Iat};function Sat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n,a=y.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let _=1+s.length;_<o.shape.length;++_)u.push([0,0]);let l=cI.kernelFunc({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),c=v.getReshaped(l.shape,s,a,!1),p=v.getPermuted(c.length,s.length,!1),m=v.getReshapedPermuted(l.shape,s,a,!1),h=ar({inputs:{x:l},backend:e,attrs:{shape:c}}),b=ao({inputs:{x:h},backend:e,attrs:{perm:p}}),N=ar({inputs:{x:b},backend:e,attrs:{shape:m}});return e.disposeData(l.dataId),e.disposeData(h.dataId),e.disposeData(b.dataId),N}var lW={kernelName:xi,backendName:\"wasm\",kernelFunc:Sat};var uW;function vat(r){uW=r.wasm.cwrap(\"SparseFillEmptyRows\",\"number\",[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Nat(r){let{backend:t,inputs:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=e,a=n.shape[0],u=n.shape[1],l=t.readSync(s.dataId)[0],c=[a+l,u],p=t.dataIdMap.get(n.dataId).id,m=t.dataIdMap.get(o.dataId).id,f=t.dataIdMap.get(i.dataId).id,d=t.makeOutput(c,n.dtype),h=t.dataIdMap.get(d.dataId).id,g=t.makeOutput(c.slice(0,1),o.dtype),x=t.dataIdMap.get(g.dataId).id,b=t.makeOutput([l],\"bool\"),w=t.dataIdMap.get(b.dataId).id,C=t.makeOutput([a],n.dtype),N=t.dataIdMap.get(C.dataId).id,_=t.makeOutput([4],\"int32\"),A=t.dataIdMap.get(_.dataId).id,$=uW(p,m,qt[o.dtype],a,l,u,f,h,x,w,N,A),F=t.readSync(_.dataId),P;switch(F[0]){case 1:{P=v.getSparseFillEmptyRowsIndicesDenseShapeMismatch(F[1]);break}case 2:{P=v.getSparseFillEmptyRowsNegativeIndexErrorMessage(F[1],F[2]);break}case 3:P=v.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(F[1],F[2],F[3]);break;default:P=\"\"}if(t.disposeData(_.dataId),P)throw t.disposeData(d.dataId),t.disposeData(g.dataId),t.disposeData(b.dataId),t.disposeData(C.dataId),new Error(P);let V=d,G=g;return $!==c[0]&&(V=Lo({inputs:{x:d},attrs:{begin:0,size:[$,u]},backend:t}),G=Lo({inputs:{x:g},attrs:{begin:0,size:$},backend:t}),t.disposeData(d.dataId),t.disposeData(g.dataId)),[V,G,b,C]}var cW={kernelName:Pl,backendName:\"wasm\",setupFunc:vat,kernelFunc:Nat};var pW;function Tat(r){pW=r.wasm.cwrap(Ga,null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function kat(r){let{backend:t,inputs:e}=r,{inputIndices:n,inputShape:o,newShape:s}=e;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape\n        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape\n        ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(n.dataId).id,a=t.dataIdMap.get(o.dataId).id,u=t.dataIdMap.get(s.dataId).id,l=n.shape[0],c=y.sizeFromShape(s.shape),p=t.makeOutput([l,c],n.dtype),m=t.dataIdMap.get(p.dataId).id,f=t.makeOutput([c],s.dtype),d=t.dataIdMap.get(f.dataId).id,h=t.makeOutput([3],\"int32\"),g=t.dataIdMap.get(h.dataId).id;pW(i,a,u,l,m,d,g);let x=t.readSync(h.dataId),b;switch(x[0]){case 0:{b=v.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(x[1],x[2]);break}case 1:{b=v.getSparseReshapeNegativeOutputDimErrorMessage(x[1],x[2]);break}case 2:b=v.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let w=Array.from(t.readSync(o.dataId)),C=Array.from(t.readSync(f.dataId));b=v.getSparseReshapeInputOutputMultipleErrorMessage(w,C);break}case 4:{let w=Array.from(t.readSync(o.dataId)),C=Array.from(t.readSync(f.dataId));b=v.getSparseReshapeInputOutputMismatchErrorMessage(w,C);break}default:b=\"\"}if(t.disposeData(h.dataId),b)throw t.disposeData(p.dataId),t.disposeData(f.dataId),new Error(b);return[p,f]}var mW={kernelName:Ga,backendName:\"wasm\",setupFunc:Tat,kernelFunc:kat};var fW;function pI(r){fW=r.wasm.cwrap(\"SparseSegmentReduction\",null,[\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function mI(r,t){let{backend:e,inputs:n}=r,{data:o,indices:s,segmentIds:i}=n,a=s.shape[0],u=e.readSync(i.dataId,a-1,a)[0],c=a>0?u+1:0;if(c<0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let p=o.shape.slice();p[0]=c;let m=e.dataIdMap.get(o.dataId).id,f=e.dataIdMap.get(s.dataId).id,d=e.dataIdMap.get(i.dataId).id,h=e.makeOutput(p,o.dtype),g=e.dataIdMap.get(h.dataId).id,x=e.makeOutput([4],\"int32\"),b=e.dataIdMap.get(x.dataId).id;fW(m,qt[o.dtype],o.shape[0],f,d,g,b,t,0);let w=e.readSync(x.dataId),C;switch(w[0]){case 0:{C=v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{C=v.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:C=v.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(w[1],w[2]);break;case 3:C=v.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(w[1],w[2],w[3]);break;default:C=\"\"}if(e.disposeData(x.dataId),C)throw e.disposeData(h.dataId),new Error(C);return h}function Eat(r){return mI(r,!0)}var dW={kernelName:Ll,backendName:\"wasm\",setupFunc:pI,kernelFunc:Eat};function _at(r){return mI(r,!1)}var hW={kernelName:Ml,backendName:\"wasm\",setupFunc:pI,kernelFunc:_at};function Aat(r){let{inputs:t,attrs:e,backend:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=e,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=new Array(o.shape.length).fill(0),c=o.shape.slice();return u.map(p=>{let m=[...c];m[a]=p;let f=Lo({inputs:{x:o},attrs:{begin:l,size:m},backend:n});return l[a]+=p,f})}var gW={kernelName:yi,backendName:\"wasm\",kernelFunc:Aat};var xW=se(As);var yW=se(zl);var $at=!0,bW=ue(Rs,$at);var wW;function Dat(r){wW=r.wasm.cwrap(po,null,[\"number\",\"number\",\"number\",\"number\"])}function Rat(r){let{backend:t,inputs:e,attrs:n}=r,{alpha:o}=n,{x:s}=e,i=t.dataIdMap.get(s.dataId).id,a=t.makeOutput(s.shape,s.dtype),u=t.dataIdMap.get(a.dataId).id;return wW(i,o,qt[s.dtype],u),a}var CW={kernelName:po,backendName:\"wasm\",setupFunc:Dat,kernelFunc:Rat};var IW;function Fat(r){IW=r.wasm.cwrap(Wa,null,[\"number\",\"array\",\"number\",\"array\",\"array\",\"array\",\"array\",\"array\",\"number\",\"number\"])}function Oat(r){let{backend:t,inputs:e,attrs:n}=r,{x:o}=e,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n,{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=ar({inputs:{x:o},backend:t,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let _=Le.computeOutShape(b,w,C),A=Lo({inputs:{x:o},backend:t,attrs:{begin:b,size:_}});N=ar({inputs:{x:A},backend:t,attrs:{shape:d}}),t.disposeData(A.dataId)}else{let _=t.makeOutput(f,\"float32\"),A=t.dataIdMap.get(o.dataId).id,$=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),F=new Uint8Array(new Int32Array(b).buffer),P=new Uint8Array(new Int32Array(w).buffer),V=new Uint8Array(new Int32Array(C).buffer),G=new Uint8Array(new Int32Array(f).buffer),W=new Uint8Array(new Int32Array(y.computeStrides(f)).buffer),q=t.dataIdMap.get(_.dataId).id;IW(A,$,o.shape.length,F,P,V,G,W,f.length,q),N=ar({inputs:{x:_},backend:t,attrs:{shape:d}}),t.disposeData(_.dataId)}return N}var SW={kernelName:Wa,backendName:\"wasm\",setupFunc:Fat,kernelFunc:Oat};function Pat(r){let{backend:t,inputs:e,attrs:n}=r,{data:o,dataSplits:s}=e,{separator:i,nGramWidths:a,leftPad:u,rightPad:l,padWidth:c,preserveShortSequences:p}=n,m=t.readSync(o.dataId),f=t.readSync(s.dataId),[d,h]=Dc(m,f,i,a,u,l,c,p),g=t.makeOutput([d.length],\"string\"),x=t.dataIdMap.get(g.dataId);x.stringBytes=d;let b=t.makeOutput(s.shape,\"int32\");return t.typedArrayFromHeap(b).set(h),[g,b]}var vW={kernelName:Bl,backendName:\"wasm\",kernelFunc:Pat};function Lat(r){let{backend:t,inputs:e,attrs:n}=r,{input:o,delimiter:s}=e,{skipEmpty:i}=n,a=t.readSync(o.dataId),u=t.readSync(s.dataId),[l,c,p]=Rc(a,u[0],i),m=c.length,f=t.makeOutput([m,2],\"int32\");t.typedArrayFromHeap(f).set(l);let h=t.makeOutput([m],\"string\"),g=t.dataIdMap.get(h.dataId);g.stringBytes=c;let x=t.makeOutput([2],\"int32\");return t.typedArrayFromHeap(x).set(p),[f,h,x]}var NW={kernelName:Vl,backendName:\"wasm\",kernelFunc:Lat};function Mat(r){let{backend:t,inputs:e,attrs:n}=r,{input:o}=e,{numBuckets:s}=n,i=t.readSync(o.dataId),a=Fc(i,s),u=t.makeOutput(o.shape,\"int32\");return t.typedArrayFromHeap(u).set(a),u}var TW={kernelName:Gl,backendName:\"wasm\",kernelFunc:Mat};var zat=!0,kW=ue(Fs,zat);var EW;function Bat(r){EW=r.wasm.cwrap($s,null,[\"number\",\"number\",\"number\",\"number\"])}function Vat(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims(\"sum\",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;EW(u,x,qt[b.dtype],w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var _W={kernelName:$s,backendName:\"wasm\",setupFunc:Bat,kernelFunc:Vat};var AW=se(Os);var $W=se(Ps);var DW;function Gat(r){DW=r.wasm.cwrap(Jn,null,[\"number\",\"array\",\"number\",\"array\",\"number\",\"number\"])}function Wat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,s=e.dataIdMap.get(o.dataId).id,{reps:i}=n,a=new Array(o.shape.length);for(let m=0;m<a.length;m++)a[m]=o.shape[m]*i[m];let u=new Uint8Array(new Int32Array(o.shape).buffer),l=new Uint8Array(new Int32Array(a).buffer),c=e.makeOutput(a,o.dtype),p=e.dataIdMap.get(c.dataId).id;return DW(s,u,o.shape.length,l,a.length,qt[c.dtype],p),c}var RW={kernelName:Jn,backendName:\"wasm\",setupFunc:Gat,kernelFunc:Wat};var FW;function Uat(r){FW=r.wasm.cwrap(Ua,null,[\"number\",\"array\",\"number\",\"number\",\"number\",\"bool\",\"number\",\"number\"])}var Hat=({inputs:r,backend:t,attrs:e})=>{let{x:n}=r,{k:o,sorted:s}=e,i=t.dataIdMap.get(n.dataId).id,a=new Uint8Array(new Int32Array(n.shape).buffer),u=n.shape.slice();u[u.length-1]=o;let l=t.makeOutput(u,n.dtype),c=t.dataIdMap.get(l.dataId).id,p=t.makeOutput(u,\"int32\"),m=t.dataIdMap.get(p.dataId).id;return FW(i,a,n.shape.length,qt[n.dtype],o,s,c,m),[l,p]},OW={kernelName:Ua,backendName:\"wasm\",setupFunc:Uat,kernelFunc:Hat};var PW;function qat(r){PW=r.wasm.cwrap(Ha,null,[\"number\",\"number\",\"bool\",\"number\",\"number\",\"number\",\"number\",\"number\",\"number\",\"array\",\"number\",\"array\",\"number\",\"number\",\"number\",\"number\",\"number\"])}function Kat(r){let{backend:t,inputs:e,attrs:n}=r,{image:o,transforms:s}=e,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=n,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),b=new Uint8Array(new Int32Array(y.computeStrides(g)).buffer),w=t.makeOutput(g,o.dtype),C=t.dataIdMap.get(w.dataId).id,_=t.dataIdMap.get(o.dataId).id,$=t.dataIdMap.get(s.dataId).id,F=i===\"nearest\"?1:2,P;switch(a){case\"constant\":P=1;break;case\"reflect\":P=2;break;case\"wrap\":P=3;break;case\"nearest\":P=4;break;default:P=1;break}return PW(_,$,s.shape[0]>1,c,d,h,f,m,p,x,o.shape.length-1,b,g.length-1,F,P,u,C),w}var LW={kernelName:Ha,backendName:\"wasm\",setupFunc:qat,kernelFunc:Kat};function jat(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o.shape[s],a=o.shape.length,u=new Array(a-1),l=0;for(let f=0;f<a;f++)f!==s&&(u[l++]=o.shape[f]);let c=new Array(i),p=new Array(a).fill(0),m=o.shape.slice();m[s]=1;for(let f=0;f<c.length;f++)p[s]=f,c[f]=Lo({inputs:{x:o},attrs:{begin:p,size:m},backend:e});return c.map(({dataId:f,dtype:d})=>({dataId:f,dtype:d,shape:u}))}var MW={kernelName:bi,backendName:\"wasm\",kernelFunc:jat};function Xat(r){let{inputs:{x:t},backend:e}=r,n=e.makeOutput(t.shape,t.dtype);return e.typedArrayFromHeap(n).fill(0),n}var zW={kernelName:wi,backendName:\"wasm\",kernelFunc:Xat};var Yat=[UB,HB,qB,jB,QB,eV,nV,sV,lV,cV,pV,mV,dV,hV,xV,bV,wV,CV,SV,NV,kV,_V,$V,DV,RV,FV,OV,PV,MV,zV,BV,GV,UV,qV,jV,YV,ZV,JV,XB,tG,eG,rG,nG,oG,sG,iG,aG,uG,cG,mG,dG,gG,xG,bG,wG,CG,SG,NG,kG,EG,AG,$G,DG,cI,FG,PG,MG,zG,BG,VG,GG,iV,UG,qG,jG,YG,ZG,JG,tW,rW,oW,sW,uV,aW,lW,cW,mW,dW,hW,gW,xW,yW,bW,CW,SW,vW,NW,TW,kW,_W,AW,$W,RW,OW,LW,ZB,MW,zW];for(let r of Yat)Lu(r);var Rk=z();Rk.registerFlag(\"WASM_HAS_SIMD_SUPPORT\",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(r){return!1}});Rk.registerFlag(\"WASM_HAS_MULTITHREAD_SUPPORT\",async()=>{if(Rk.get(\"IS_NODE\"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(r){return!1}});var zk=Tl(WW()),XW=Tl(HW()),Bk=Tl(qW());var KW=zk.default||zk,Zat=Bk.default||Bk,cg=class extends zo{constructor(t){super(),this.wasm=t,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(ZW),Mk=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new ra(this,Pn())}write(t,e,n){let o={id:this.dataIdNextNumber++};return this.move(o,t,e,n,1),o}numDataIds(){return this.dataIdMap.numDataIds()}async time(t){let e=y.now();return t(),{kernelMs:y.now()-e}}move(t,e,n,o,s){let i=this.dataIdNextNumber++;if(o===\"string\"){let c=e;this.dataIdMap.set(t,{id:i,stringBytes:c,shape:n,dtype:o,memoryOffset:null,refCount:s});return}let a=y.sizeFromShape(n),u=a*y.bytesPerElement(o),l=this.wasm._malloc(u);this.dataIdMap.set(t,{id:i,memoryOffset:l,shape:n,dtype:o,refCount:s}),this.wasm.tfjs.registerTensor(i,a,l),e!=null&&this.wasm.HEAPU8.set(new Uint8Array(e.buffer,e.byteOffset,u),l)}async read(t){return this.readSync(t)}readSync(t,e,n){let{memoryOffset:o,dtype:s,shape:i,stringBytes:a}=this.dataIdMap.get(t);if(s===\"string\")return(e==null||e===0)&&(n==null||n>=a.length)?a:a.slice(e,n);e=e||0,n=n||y.sizeFromShape(i);let u=y.bytesPerElement(s),l=this.wasm.HEAPU8.slice(o+e*u,o+n*u);return Qat(l.buffer,s)}disposeData(t,e=!1){if(this.dataIdMap.has(t)){let n=this.dataIdMap.get(t);if(n.refCount--,!e&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(t)}return!0}refCount(t){return this.dataIdMap.has(t)?this.dataIdMap.get(t).refCount:0}incRef(t){let e=this.dataIdMap.get(t);e!=null&&e.refCount++}floatPrecision(){return 32}getMemoryOffset(t){return this.dataIdMap.get(t).memoryOffset}dispose(){this.wasm.tfjs.dispose(),\"PThread\"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(t,e,n){let o;if(n==null)o=this.write(null,t,e);else{let s=this.dataIdNextNumber++;o={id:s},this.dataIdMap.set(o,{id:s,memoryOffset:n,shape:t,dtype:e,refCount:1});let i=y.sizeFromShape(t);this.wasm.tfjs.registerTensor(s,i,n)}return{dataId:o,shape:t,dtype:e}}typedArrayFromHeap({shape:t,dtype:e,dataId:n}){let o=this.wasm.HEAPU8.buffer,{memoryOffset:s}=this.dataIdMap.get(n),i=y.sizeFromShape(t);switch(e){case\"float32\":return new Float32Array(o,s,i);case\"int32\":return new Int32Array(o,s,i);case\"bool\":return new Uint8Array(o,s,i);default:throw new Error(`Unknown dtype ${e}`)}}};function Jat(r){return(t,e)=>(y.fetch(r,{credentials:\"same-origin\"}).then(n=>{n.ok||t.env.a(`failed to load wasm binary file at '${r}'`),n.arrayBuffer().then(o=>{WebAssembly.instantiate(o,t).then(s=>{e(s.instance,s.module)})})}),{})}function jW(r,t,e){if(hI!=null)return hI;let n=\"tfjs-backend-wasm.wasm\";return r&&t?n=\"tfjs-backend-wasm-threaded-simd.wasm\":r&&(n=\"tfjs-backend-wasm-simd.wasm\"),lg!=null&&lg[n]!=null?lg[n]:e+n}async function YW(){let[r,t]=await Promise.all([z().getAsync(\"WASM_HAS_SIMD_SUPPORT\"),z().getAsync(\"WASM_HAS_MULTITHREAD_SUPPORT\")]);return new Promise((e,n)=>{let o={};o.locateFile=(a,u)=>{if(a.endsWith(\".worker.js\")){let l=XW.wasmWorkerContents.replace(/\\n/g,\"\\\\n\"),c=new Blob([l],{type:\"application/javascript\"});return URL.createObjectURL(c)}return a.endsWith(\".wasm\")?jW(r,t,ag!=null?ag:u):u+a},Vk&&(o.instantiateWasm=Jat(jW(r,t,ag!=null?ag:\"\")));let s=!1;o.onAbort=()=>{if(s||ug)return;ug=!0,n({message:\"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers\"})};let i;t&&r&&hI==null?(o.mainScriptUrlOrBlob=new Blob([\"var WasmBackendModuleThreadedSimd = \"+KW.toString()],{type:\"text/javascript\"}),i=KW(o)):i=Zat(o),i.then(a=>{s=!0,ug=!1;let u=null;a.tfjs={init:a.cwrap(\"init\",null,[]),initWithThreadsCount:a.cwrap(\"init_with_threads_count\",null,[\"number\"]),getThreadsCount:a.cwrap(\"get_threads_count\",\"number\",[]),registerTensor:a.cwrap(\"register_tensor\",null,[\"number\",\"number\",\"number\"]),disposeData:a.cwrap(\"dispose_data\",u,[\"number\"]),dispose:a.cwrap(\"dispose\",u,[])},e({wasm:a})}).catch(n)})}function Qat(r,t){switch(t){case\"float32\":return new Float32Array(r);case\"int32\":return new Int32Array(r);case\"bool\":return new Uint8Array(r);default:throw new Error(`Unknown dtype ${t}`)}}var tlt=[\"tfjs-backend-wasm.wasm\",\"tfjs-backend-wasm-simd.wasm\",\"tfjs-backend-wasm-threaded-simd.wasm\"],hI=null,ag=null,lg={},ug=!1,Vk=!1;function elt(r,t=!1){if(W0(\"setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release.\"),ug)throw new Error(\"The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`\");hI=r,Vk=t}function rlt(r,t=!1){if(ug)throw new Error(\"The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`\");if(typeof r==\"string\")ag=r;else{lg=r;let e=tlt.filter(n=>lg[n]==null);if(e.length>0)throw new Error(`There were no entries found for the following binaries: ${e.join(\",\")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}Vk=t}var ZW=-1,Mk=-1;function nlt(r){ZW=r}function olt(){if(Mk===-1)throw new Error(\"WASM backend not initialized.\");return Mk}var slt=\"4.0.0\";var ilt=2;Xp(\"wasm\",async()=>{let{wasm:r}=await YW();return new cg(r)},ilt);var JW=\"4.0.0\",alt=\"4.0.0\",llt=\"4.0.0\",ult=\"4.0.0\",clt=\"4.0.0\",plt={tfjs:JW,\"tfjs-core\":JW,\"tfjs-converter\":alt,\"tfjs-backend-cpu\":llt,\"tfjs-backend-webgl\":ult,\"tfjs-backend-wasm\":clt};export{ii as Abs,oa as Acos,sa as Acosh,cu as AdadeltaOptimizer,pu as AdagradOptimizer,mu as AdamOptimizer,fu as AdamaxOptimizer,Zn as Add,Go as AddN,ia as All,aa as Any,Wo as ArgMax,kl as ArgMin,la as Asin,ua as Asinh,ca as Atan,ma as Atan2,pa as Atanh,Uo as AvgPool,El as AvgPool3D,lp as AvgPool3DGrad,ap as AvgPoolGrad,cg as BackendWasm,Ho as BatchMatMul,ai as BatchToSpaceND,up as Bincount,cp as BroadcastArgs,p1 as BroadcastTo,Sb as Callback,Py as CallbackList,lo as Cast,qo as Ceil,uo as ClipByValue,pp as Complex,_l as ComplexAbs,li as Concat,Ko as Conv2D,mp as Conv2DBackpropFilter,jo as Conv2DBackpropInput,Al as Conv3D,fp as Conv3DBackpropFilterV2,dp as Conv3DBackpropInputV2,Xo as Cos,Yo as Cosh,da as CropAndResize,fa as Cumprod,Zo as Cumsum,My as CustomCallback,ra as DataStorage,hp as DenseBincount,ha as DepthToSpace,Jo as DepthwiseConv2dNative,gp as DepthwiseConv2dNativeBackpropFilter,xp as DepthwiseConv2dNativeBackpropInput,yp as Diag,$l as Dilation2D,Xd as Dilation2DBackpropFilter,jd as Dilation2DBackpropInput,i0 as ENV,vb as EarlyStopping,bp as Einsum,ts as Elu,wp as EluGrad,qd as Environment,xa as Equal,ga as Erf,es as Exp,ui as ExpandDims,ya as Expm1,Cp as FFT,Dl as Fill,ba as FlipLeftRight,rs as Floor,ns as FloorDiv,Yd as FromPixels,os as FusedBatchNorm,Ii as FusedConv2D,Si as FusedDepthwiseConv2D,Bc as GPGPUContext,wa as GatherNd,ci as GatherV2,Ph as GraphModel,Ca as Greater,ss as GreaterEqual,Ly as History,Ip as IFFT,co as Identity,Sp as Imag,ye as InputSpec,Ia as IsFinite,Sa as IsInf,va as IsNan,zo as KernelBackend,Rl as LRN,Np as LRNGrad,Ch as LayerVariable,Bn as LayersModel,is as LeakyRelu,Na as Less,Ta as LessEqual,vp as LinSpace,as as Log,ka as Log1p,f1 as LogSoftmax,Ea as LogicalAnd,_a as LogicalNot,Aa as LogicalOr,m1 as LogicalXor,xlt as LowerBound,_u as MathBackendWebGL,ls as Max,cs as MaxPool,Fl as MaxPool3D,kp as MaxPool3DGrad,Tp as MaxPoolGrad,Ep as MaxPoolWithArgmax,us as Maximum,ps as Mean,ms as Min,fs as Minimum,ds as MirrorPad,$a as Mod,du as MomentumOptimizer,_p as Multinomial,hs as Multiply,pi as Neg,Ra as NonMaxSuppressionV3,Fa as NonMaxSuppressionV4,Oa as NonMaxSuppressionV5,Da as NotEqual,k0 as OP_SCOPE_SUFFIX,gs as OneHot,mi as OnesLike,Wr as Optimizer,Ws as OptimizerConstructors,fi as Pack,xs as PadV2,ylt as Pool,ys as Pow,bs as Prelu,ws as Prod,hu as RMSPropOptimizer,Tn as RNN,Ap as RaggedGather,$p as RaggedRange,Dp as RaggedTensorToTensor,Ol as Range,x0 as Rank,Rp as Real,Qo as RealDiv,Pa as Reciprocal,Xe as Reduction,Cs as Relu,vs as Relu6,di as Reshape,Ss as ResizeBilinear,Op as ResizeBilinearGrad,Is as ResizeNearestNeighbor,Fp as ResizeNearestNeighborGrad,Ns as Reverse,qa as RotateWithOffset,Ts as Round,ks as Rsqrt,Bi as SGDOptimizer,La as ScatterNd,Pp as SearchSorted,hi as Select,Ma as Selu,qi as Sequential,_s as Sigmoid,Ba as Sign,Es as Sin,za as Sinh,gi as Slice,Ds as Softmax,Va as Softplus,xi as SpaceToBatchND,Pl as SparseFillEmptyRows,Ga as SparseReshape,Ll as SparseSegmentMean,Ml as SparseSegmentSum,Lp as SparseToDense,yi as SplitV,As as Sqrt,zl as Square,Rs as SquaredDifference,po as Step,Wa as StridedSlice,Bl as StringNGrams,Vl as StringSplit,Gl as StringToHashBucketFast,Fs as Sub,$s as Sum,Jr as SymbolicTensor,Os as Tan,Ps as Tanh,Ft as Tensor,pe as TensorBuffer,Jn as Tile,Ua as TopK,Ha as Transform,Qn as Transpose,Mp as Unique,bi as Unpack,Wl as UnsortedSegmentSum,blt as UpperBound,Ka as Variable,wi as ZerosLike,Ci as _FusedMatMul,Ee as abs,ax as acos,lx as acosh,X as add,LE as addN,Zp as all,qu as any,Ai as argMax,ux as argMin,cx as asin,px as asinh,mx as atan,fx as atan2,dx as atanh,Yl as avgPool,gx as avgPool3d,gE as backend,v as backend_util,BE as basicLSTMCell,Di as batchNorm,xx as batchNorm2d,yx as batchNorm3d,bx as batchNorm4d,Zl as batchToSpaceND,wx as bincount,n6 as booleanMaskAsync,GE as broadcastArgs,Ri as broadcastTo,Vr as broadcast_util,nx as browser,wt as buffer,VZ as callbacks,J as cast,Cx as ceil,Cr as clipByValue,sn as clone,wn as complex,ne as concat,Ix as concat1d,Sx as concat2d,vx as concat3d,Nx as concat4d,K$ as constraints,Qp as conv1d,In as conv2d,em as conv2dTranspose,Tx as conv3d,Ex as conv3dTranspose,Tlt as copyRegisteredKernels,Jl as cos,rm as cosh,hh as cosineWindow,Xu as cumprod,nm as cumsum,un as customGrad,AR as data,ch as denseBincount,W0 as deprecationWarn,_x as depthToSpace,Fi as depthwiseConv2d,HZ as deregisterOp,Kl as device_util,WE as diag,Ax as dilation2d,gpt as disableDeprecationWarnings,vt as dispose,xpt as disposeVariables,pt as div,$x as divNoNan,Dx as dot,lv as dropout,UE as einsum,Oi as elu,hpt as enableDebugMode,dpt as enableProdMode,uv as enclosingPowerOfTwo,Pn as engine,z as env,$r as equal,Rx as erf,Fx as euclideanNorm,er as exp,rr as expandDims,Ox as expm1,Yu as eye,au as fft,xo as fill,Spt as findBackend,vpt as findBackendFactory,Pi as floor,Yp as floorDiv,hM as forceHalfFloat,uu as fused,Li as gather,m6 as gatherND,ox as gather_util,Cpt as getBackend,u0 as getGradient,Jd as getKernel,zg as getKernelsForBackend,olt as getThreadsCount,ik as gpgpu_util,bK as grad,wK as grads,Re as greater,ln as greaterEqual,tl as ifft,Xl as imag,Gs as image,h6 as inTopKAsync,j$ as initializers,Pv as input,_r as io,xm as irfft,Px as isFinite,Lx as isInf,Mx as isNaN,De as keep,Ur as kernel_impls,ED as layers,Ql as leakyRelu,om as less,Ln as lessEqual,pv as linalg,KE as linspace,M7 as loadGraphModel,z7 as loadGraphModelSync,hD as loadLayersModel,zx as localResponseNormalization,Sr as log,tu as log1p,Gx as logSigmoid,sm as logSoftmax,im as logSumExp,Rr as logicalAnd,eu as logicalNot,am as logicalOr,Wx as logicalXor,hX as losses,jE as lowerBound,Lt as matMul,yE as math,Ir as max,ru as maxPool,Hx as maxPool3d,XE as maxPoolWithArgmax,Sn as maximum,ve as mean,ah as memory,YE as meshgrid,_D as metrics,Ja as min,Mi as minimum,qx as mirrorPad,Kx as mod,q8 as model,AD as models,Zu as moments,s6 as movingAverage,D as mul,ZE as multiRNNCell,JE as multinomial,Ht as neg,gh as nextFrame,Qa as norm,Bs as notEqual,Ei as oneHot,cr as ones,yr as onesLike,T as op,QE as outerProduct,cn as pad,t_ as pad1d,e_ as pad2d,r_ as pad3d,n_ as pad4d,jx as pool,an as pow,ou as prelu,Jg as print,Xx as prod,ypt as profile,o_ as raggedGather,s_ as raggedRange,i_ as raggedTensorToTensor,a_ as rand,v_ as randomGamma,tc as randomNormal,N_ as randomStandardNormal,zi as randomUniform,su as range,wpt as ready,Za as real,ty as reciprocal,Xp as registerBackend,j8 as registerCallbackConstructor,h1 as registerGradient,Lu as registerKernel,UZ as registerOp,$D as regularizers,Fr as relu,lm as relu6,Ipt as removeBackend,R as reshape,pr as reverse,T_ as reverse1d,k_ as reverse2d,E_ as reverse3d,__ as reverse4d,lu as rfft,um as round,cm as rsqrt,mt as scalar,a6 as scatterND,lh as scatter_util,mh as searchSorted,pm as selu,mm as separableConv2d,K8 as sequential,Q as serialization,tH as setBackend,Npt as setPlatform,nlt as setThreadsCount,elt as setWasmPath,rlt as setWasmPaths,wT as setWebGLContext,A_ as setdiff1dAsync,Yr as sigmoid,ey as sign,dX as signal,fm as sin,dm as sinh,Rt as slice,hm as slice1d,dh as slice2d,gm as slice3d,ec as slice4d,Le as slice_util,iu as softmax,zs as softplus,nu as spaceToBatchND,gX as sparse,c6 as sparseToDense,fX as spectral,mr as split,Se as sqrt,Mt as square,ym as squaredDifference,Mn as squeeze,nr as stack,bo as step,ry as stridedSlice,xX as string,ct as sub,ft as sum,Wu as sumOutType,ny as tan,$i as tanh,ur as tensor,Me as tensor1d,Vs as tensor2d,rx as tensor3d,$_ as tensor4d,D_ as tensor5d,R_ as tensor6d,go as tensor_util,OE as test_util,B as tidy,Dr as tile,bpt as time,oy as topk,ic as train,Ot as transpose,bm as truncatedNormal,sy as unique,Nlt as unregisterGradient,vlt as unregisterKernel,wm as unsortedSegmentSum,vr as unstack,sr as upcastType,F_ as upperBound,y as util,CK as valueAndGrad,IK as valueAndGrads,iy as variable,Bx as variableGrads,plt as version,cR as version_converter,PE as version_core,Um as version_layers,slt as version_wasm,dM as version_webgl,Zke as webgl,dd as webgl_util,_e as where,ly as whereAsync,Ne as zeros,It as zerosLike};\n", "export * from './drawContour';\nexport * from './drawDetections';\nexport * from './drawFaceExpressions';\nexport * from './DrawBox';\nexport * from './DrawFaceLandmarks';\nexport * from './DrawTextField';\n", "import { Point } from '../classes/index';\n\nexport function drawContour(\n  ctx: CanvasRenderingContext2D,\n  points: Point[],\n  isClosed = false,\n) {\n  ctx.beginPath();\n\n  points.slice(1).forEach(({ x, y }, prevIdx) => {\n    const from = points[prevIdx];\n    ctx.moveTo(from.x, from.y);\n    ctx.lineTo(x, y);\n  });\n\n  if (isClosed) {\n    const from = points[points.length - 1];\n    const to = points[0];\n    if (!from || !to) {\n      return;\n    }\n\n    ctx.moveTo(from.x, from.y);\n    ctx.lineTo(to.x, to.y);\n  }\n\n  ctx.stroke();\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Point } from '../classes/index';\nimport { Dimensions, IDimensions } from '../classes/Dimensions';\n\nexport function isTensor(tensor: any, dim: number) {\n  return tensor instanceof tf.Tensor && tensor.shape.length === dim;\n}\n\nexport function isTensor1D(tensor: any): tensor is tf.Tensor1D {\n  return isTensor(tensor, 1);\n}\n\nexport function isTensor2D(tensor: any): tensor is tf.Tensor2D {\n  return isTensor(tensor, 2);\n}\n\nexport function isTensor3D(tensor: any): tensor is tf.Tensor3D {\n  return isTensor(tensor, 3);\n}\n\nexport function isTensor4D(tensor: any): tensor is tf.Tensor4D {\n  return isTensor(tensor, 4);\n}\n\nexport function isFloat(num: number) {\n  return num % 1 !== 0;\n}\n\nexport function isEven(num: number) {\n  return num % 2 === 0;\n}\n\nexport function round(num: number, prec = 2) {\n  const f = 10 ** prec;\n  return Math.floor(num * f) / f;\n}\n\nexport function isDimensions(obj: any): boolean {\n  return obj && obj.width && obj.height;\n}\n\nexport function computeReshapedDimensions({ width, height }: IDimensions, inputSize: number) {\n  const scale = inputSize / Math.max(height, width);\n  return new Dimensions(Math.round(width * scale), Math.round(height * scale));\n}\n\nexport function getCenterPoint(pts: Point[]): Point {\n  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0))\n    .div(new Point(pts.length, pts.length));\n}\n\nexport function range(num: number, start: number, step: number): number[] {\n  return Array(num).fill(0).map((_, i) => start + (i * step));\n}\n\nexport function isValidNumber(num: any) {\n  return !!num && (num !== Infinity) && (num !== -Infinity) && !Number.isNaN(num) || num === 0;\n}\n\nexport function isValidProbablitiy(num: any) {\n  return isValidNumber(num) && num >= 0 && num <= 1.0;\n}\n", "import { isValidNumber } from '../utils/index';\n\nexport interface IDimensions {\n  width: number\n  height: number\n}\n\nexport class Dimensions implements IDimensions {\n  private _width: number;\n\n  private _height: number;\n\n  constructor(width: number, height: number) {\n    if (!isValidNumber(width) || !isValidNumber(height)) {\n      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);\n    }\n\n    this._width = width;\n    this._height = height;\n  }\n\n  public get width(): number { return this._width; }\n\n  public get height(): number { return this._height; }\n\n  public reverse(): Dimensions {\n    return new Dimensions(1 / this.width, 1 / this.height);\n  }\n}\n", "export interface IPoint {\n  x: number\n  y: number\n}\n\nexport class Point implements IPoint {\n  private _x: number;\n\n  private _y: number;\n\n  constructor(x: number, y: number) {\n    this._x = x;\n    this._y = y;\n  }\n\n  get x(): number { return this._x; }\n\n  get y(): number { return this._y; }\n\n  public add(pt: IPoint): Point {\n    return new Point(this.x + pt.x, this.y + pt.y);\n  }\n\n  public sub(pt: IPoint): Point {\n    return new Point(this.x - pt.x, this.y - pt.y);\n  }\n\n  public mul(pt: IPoint): Point {\n    return new Point(this.x * pt.x, this.y * pt.y);\n  }\n\n  public div(pt: IPoint): Point {\n    return new Point(this.x / pt.x, this.y / pt.y);\n  }\n\n  public abs(): Point {\n    return new Point(Math.abs(this.x), Math.abs(this.y));\n  }\n\n  public magnitude(): number {\n    return Math.sqrt((this.x ** 2) + (this.y ** 2));\n  }\n\n  public floor(): Point {\n    return new Point(Math.floor(this.x), Math.floor(this.y));\n  }\n}\n", "import { isDimensions, isValidNumber } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { IDimensions } from './Dimensions';\nimport { Point } from './Point';\nimport { IRect } from './Rect';\n\nexport class Box<BoxType = any> implements IBoundingBox, IRect {\n  public static isRect(rect: any): boolean {\n    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);\n  }\n\n  public static assertIsValidBox(box: any, callee: string, allowNegativeDimensions = false) {\n    if (!Box.isRect(box)) {\n      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);\n    }\n\n    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {\n      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);\n    }\n  }\n\n  private _x: number;\n\n  private _y: number;\n\n  private _width: number;\n\n  private _height: number;\n\n  constructor(_box: IBoundingBox | IRect, allowNegativeDimensions = true) {\n    const box = (_box || {}) as any;\n\n    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);\n    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);\n\n    if (!isRect && !isBbox) {\n      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);\n    }\n\n    const [x, y, width, height] = isRect\n      ? [box.x, box.y, box.width, box.height]\n      : [box.left, box.top, box.right - box.left, box.bottom - box.top];\n\n    Box.assertIsValidBox({\n      x, y, width, height,\n    }, 'Box.constructor', allowNegativeDimensions);\n\n    this._x = x;\n    this._y = y;\n    this._width = width;\n    this._height = height;\n  }\n\n  public get x(): number { return this._x; }\n\n  public get y(): number { return this._y; }\n\n  public get width(): number { return this._width; }\n\n  public get height(): number { return this._height; }\n\n  public get left(): number { return this.x; }\n\n  public get top(): number { return this.y; }\n\n  public get right(): number { return this.x + this.width; }\n\n  public get bottom(): number { return this.y + this.height; }\n\n  public get area(): number { return this.width * this.height; }\n\n  public get topLeft(): Point { return new Point(this.left, this.top); }\n\n  public get topRight(): Point { return new Point(this.right, this.top); }\n\n  public get bottomLeft(): Point { return new Point(this.left, this.bottom); }\n\n  public get bottomRight(): Point { return new Point(this.right, this.bottom); }\n\n  public round(): Box<BoxType> {\n    const [x, y, width, height] = [this.x, this.y, this.width, this.height]\n      .map((val) => Math.round(val));\n    return new Box({\n      x, y, width, height,\n    });\n  }\n\n  public floor(): Box<BoxType> {\n    const [x, y, width, height] = [this.x, this.y, this.width, this.height]\n      .map((val) => Math.floor(val));\n    return new Box({\n      x, y, width, height,\n    });\n  }\n\n  public toSquare(): Box<BoxType> {\n    let {\n      x, y, width, height,\n    } = this;\n    const diff = Math.abs(width - height);\n    if (width < height) {\n      x -= (diff / 2);\n      width += diff;\n    }\n    if (height < width) {\n      y -= (diff / 2);\n      height += diff;\n    }\n\n    return new Box({ x, y, width, height });\n  }\n\n  public rescale(s: IDimensions | number): Box<BoxType> {\n    const scaleX = isDimensions(s) ? (s as IDimensions).width : s as number;\n    const scaleY = isDimensions(s) ? (s as IDimensions).height : s as number;\n    return new Box({\n      x: this.x * scaleX,\n      y: this.y * scaleY,\n      width: this.width * scaleX,\n      height: this.height * scaleY,\n    });\n  }\n\n  public pad(padX: number, padY: number): Box<BoxType> {\n    const [x, y, width, height] = [\n      this.x - (padX / 2),\n      this.y - (padY / 2),\n      this.width + padX,\n      this.height + padY,\n    ];\n    return new Box({ x, y, width, height });\n  }\n\n  public clipAtImageBorders(imgWidth: number, imgHeight: number): Box<BoxType> {\n    const { x, y, right, bottom } = this;\n    const clippedX = Math.max(x, 0);\n    const clippedY = Math.max(y, 0);\n\n    const newWidth = right - clippedX;\n    const newHeight = bottom - clippedY;\n    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);\n    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);\n\n    return (new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight })).floor();\n  }\n\n  public shift(sx: number, sy: number): Box<BoxType> {\n    const { width, height } = this;\n    const x = this.x + sx;\n    const y = this.y + sy;\n\n    return new Box({ x, y, width, height });\n  }\n\n  public padAtBorders(imageHeight: number, imageWidth: number) {\n    const w = this.width + 1;\n    const h = this.height + 1;\n\n    const dx = 1;\n    const dy = 1;\n    let edx = w;\n    let edy = h;\n\n    let x = this.left;\n    let y = this.top;\n    let ex = this.right;\n    let ey = this.bottom;\n\n    if (ex > imageWidth) {\n      edx = -ex + imageWidth + w;\n      ex = imageWidth;\n    }\n    if (ey > imageHeight) {\n      edy = -ey + imageHeight + h;\n      ey = imageHeight;\n    }\n    if (x < 1) {\n      edy = 2 - x;\n      x = 1;\n    }\n    if (y < 1) {\n      edy = 2 - y;\n      y = 1;\n    }\n\n    return { dy, edy, dx, edx, y, ey, x, ex, w, h };\n  }\n\n  public calibrate(region: Box) {\n    return new Box({\n      left: this.left + (region.left * this.width),\n      top: this.top + (region.top * this.height),\n      right: this.right + (region.right * this.width),\n      bottom: this.bottom + (region.bottom * this.height),\n    }).toSquare().round();\n  }\n}\n", "import { Box } from './Box';\n\nexport interface IBoundingBox {\n  left: number\n  top: number\n  right: number\n  bottom: number\n}\n\nexport class BoundingBox extends Box implements IBoundingBox {\n  constructor(left: number, top: number, right: number, bottom: number, allowNegativeDimensions = false) {\n    super({ left, top, right, bottom }, allowNegativeDimensions);\n  }\n}\n", "import { Box } from './Box';\nimport { Dimensions, IDimensions } from './Dimensions';\nimport { IRect, Rect } from './Rect';\n\nexport class ObjectDetection {\n  private _score: number;\n\n  private _classScore: number;\n\n  private _className: string;\n\n  private _box: Rect;\n\n  private _imageDims: Dimensions;\n\n  constructor(\n    score: number,\n    classScore: number,\n    className: string,\n    relativeBox: IRect,\n    imageDims: IDimensions,\n  ) {\n    this._imageDims = new Dimensions(imageDims.width, imageDims.height);\n    this._score = score;\n    this._classScore = classScore;\n    this._className = className;\n    this._box = new Box(relativeBox).rescale(this._imageDims);\n  }\n\n  public get score(): number { return this._score; }\n\n  public get classScore(): number { return this._classScore; }\n\n  public get className(): string { return this._className; }\n\n  public get box(): Box { return this._box; }\n\n  public get imageDims(): Dimensions { return this._imageDims; }\n\n  public get imageWidth(): number { return this.imageDims.width; }\n\n  public get imageHeight(): number { return this.imageDims.height; }\n\n  public get relativeBox(): Box { return new Box(this._box).rescale(this.imageDims.reverse()); }\n\n  public forSize(width: number, height: number): ObjectDetection {\n    return new ObjectDetection(\n      this.score,\n      this.classScore,\n      this.className,\n      this.relativeBox,\n      { width, height },\n    );\n  }\n}\n", "import { Box } from './Box';\nimport { IDimensions } from './Dimensions';\nimport { ObjectDetection } from './ObjectDetection';\nimport { Rect } from './Rect';\n\nexport interface IFaceDetecion {\n  score: number\n  box: Box\n}\n\nexport class FaceDetection extends ObjectDetection implements IFaceDetecion {\n  constructor(\n    score: number,\n    relativeBox: Rect,\n    imageDims: IDimensions,\n  ) {\n    super(score, score, '', relativeBox, imageDims);\n  }\n\n  public override forSize(width: number, height: number): FaceDetection {\n    const { score, relativeBox, imageDims } = super.forSize(width, height);\n    return new FaceDetection(score, relativeBox, imageDims);\n  }\n}\n", "import { Box } from '../classes/Box';\n\nexport function iou(box1: Box, box2: Box, isIOU = true) {\n  const width = Math.max(0.0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));\n  const height = Math.max(0.0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));\n  const interSection = width * height;\n\n  return isIOU\n    ? interSection / (box1.area + box2.area - interSection)\n    : interSection / Math.min(box1.area, box2.area);\n}\n", "import { BoundingBox, IPoint } from '../classes/index';\n\nexport function minBbox(pts: IPoint[]): BoundingBox {\n  const xs = pts.map((pt) => pt.x);\n  const ys = pts.map((pt) => pt.y);\n  const minX = xs.reduce((min, x) => (x < min ? x : min), Infinity);\n  const minY = ys.reduce((min, y) => (y < min ? y : min), Infinity);\n  const maxX = xs.reduce((max, x) => (max < x ? x : max), 0);\n  const maxY = ys.reduce((max, y) => (max < y ? y : max), 0);\n\n  return new BoundingBox(minX, minY, maxX, maxY);\n}\n", "import { Box } from '../classes/Box';\nimport { iou } from './iou';\n\nexport function nonMaxSuppression(\n  boxes: Box[],\n  scores: number[],\n  iouThreshold: number,\n  isIOU = true,\n): number[] {\n  let indicesSortedByScore = scores\n    .map((score, boxIndex) => ({ score, boxIndex }))\n    .sort((c1, c2) => c1.score - c2.score)\n    .map((c) => c.boxIndex);\n\n  const pick: number[] = [];\n\n  while (indicesSortedByScore.length > 0) {\n    const curr = indicesSortedByScore.pop() as number;\n    pick.push(curr);\n\n    const indices = indicesSortedByScore;\n\n    const outputs: number[] = [];\n    for (let i = 0; i < indices.length; i++) {\n      const idx = indices[i];\n\n      const currBox = boxes[curr];\n      const idxBox = boxes[idx];\n\n      outputs.push(iou(currBox, idxBox, isIOU));\n    }\n\n    indicesSortedByScore = indicesSortedByScore.filter(\n      (_, j) => outputs[j] <= iouThreshold,\n    );\n  }\n\n  return pick;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function normalize(x: tf.Tensor4D, meanRgb: number[]): tf.Tensor4D {\n  return tf.tidy(() => {\n    const [r, g, b] = meanRgb;\n    const avg_r = tf.fill([...x.shape.slice(0, 3), 1], r, 'float32');\n    const avg_g = tf.fill([...x.shape.slice(0, 3), 1], g, 'float32');\n    const avg_b = tf.fill([...x.shape.slice(0, 3), 1], b, 'float32');\n    const avg_rgb = tf.concat([avg_r, avg_g, avg_b], 3);\n\n    return tf.sub(x, avg_rgb);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\n/**\n * Pads the smaller dimension of an image tensor with zeros, such that width === height.\n *\n * @param imgTensor The image tensor.\n * @param isCenterImage (optional, default: false) If true, add an equal amount of padding on\n * both sides of the minor dimension oof the image.\n * @returns The padded tensor with width === height.\n */\nexport function padToSquare(imgTensor: tf.Tensor4D, isCenterImage = false): tf.Tensor4D {\n  return tf.tidy(() => {\n    const [height, width] = imgTensor.shape.slice(1);\n    if (height === width) return imgTensor;\n    const dimDiff = Math.abs(height - width);\n    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));\n    const paddingAxis = height > width ? 2 : 1;\n    const createPaddingTensor = (paddingAmountLocal: number): tf.Tensor => {\n      const paddingTensorShape = imgTensor.shape.slice();\n      paddingTensorShape[paddingAxis] = paddingAmountLocal;\n      return tf.fill(paddingTensorShape, 0, 'float32');\n    };\n    const paddingTensorAppend = createPaddingTensor(paddingAmount);\n    const remainingPaddingAmount = dimDiff - (paddingTensorAppend.shape[paddingAxis] as number);\n    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;\n    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend]\n      .filter((t) => !!t)\n      .map((t) => tf.cast(t as tf.Tensor4D, 'float32')) as tf.Tensor4D[];\n    return tf.concat(tensorsToStack, paddingAxis);\n  });\n}\n", "export function shuffleArray(inputArray: any[]) {\n  const array = inputArray.slice();\n  for (let i = array.length - 1; i > 0; i--) {\n    const j = Math.floor(Math.random() * (i + 1));\n    const x = array[i];\n    array[i] = array[j];\n    array[j] = x;\n  }\n  return array;\n}\n", "export * from './iou';\nexport * from './minBbox';\nexport * from './nonMaxSuppression';\nexport * from './normalize';\nexport * from './padToSquare';\nexport * from './shuffleArray';\n\nexport function sigmoid(x: number) {\n  return 1 / (1 + Math.exp(-x));\n}\n\nexport function inverseSigmoid(x: number) {\n  return Math.log(x / (1 - x));\n}\n", "import { Box } from './Box';\n\nexport interface IRect {\n  x: number\n  y: number\n  width: number\n  height: number\n}\n\nexport class Rect extends Box implements IRect {\n  constructor(x: number, y: number, width: number, height: number, allowNegativeDimensions = false) {\n    super({ x, y, width, height }, allowNegativeDimensions);\n  }\n}\n", "import { minBbox } from '../ops/index';\nimport { getCenterPoint } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { Box } from './Box';\nimport { Dimensions, IDimensions } from './Dimensions';\nimport { FaceDetection } from './FaceDetection';\nimport { Point } from './Point';\nimport { IRect, Rect } from './Rect';\n\n// face alignment constants\nconst relX = 0.5;\nconst relY = 0.43;\nconst relScale = 0.45;\n\nexport interface IFaceLandmarks {\n  positions: Point[]\n  shift: Point\n}\n\nexport class FaceLandmarks implements IFaceLandmarks {\n  protected _shift: Point;\n\n  protected _positions: Point[];\n\n  protected _imgDims: Dimensions;\n\n  constructor(\n    relativeFaceLandmarkPositions: Point[],\n    imgDims: IDimensions,\n    shift: Point = new Point(0, 0),\n  ) {\n    const { width, height } = imgDims;\n    this._imgDims = new Dimensions(width, height);\n    this._shift = shift;\n    this._positions = relativeFaceLandmarkPositions.map(\n      (pt) => pt.mul(new Point(width, height)).add(shift),\n    );\n  }\n\n  public get shift(): Point { return new Point(this._shift.x, this._shift.y); }\n\n  public get imageWidth(): number { return this._imgDims.width; }\n\n  public get imageHeight(): number { return this._imgDims.height; }\n\n  public get positions(): Point[] { return this._positions; }\n\n  public get relativePositions(): Point[] {\n    return this._positions.map(\n      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight)),\n    );\n  }\n\n  public forSize<T extends FaceLandmarks>(width: number, height: number): T {\n    return new (this.constructor as any)(\n      this.relativePositions,\n      { width, height },\n    );\n  }\n\n  public shiftBy<T extends FaceLandmarks>(x: number, y: number): T {\n    return new (this.constructor as any)(\n      this.relativePositions,\n      this._imgDims,\n      new Point(x, y),\n    );\n  }\n\n  public shiftByPoint<T extends FaceLandmarks>(pt: Point): T {\n    return this.shiftBy(pt.x, pt.y);\n  }\n\n  /**\n   * Aligns the face landmarks after face detection from the relative positions of the faces\n   * bounding box, or it's current shift. This function should be used to align the face images\n   * after face detection has been performed, before they are passed to the face recognition net.\n   * This will make the computed face descriptor more accurate.\n   *\n   * @param detection (optional) The bounding box of the face or the face detection result. If\n   * no argument was passed the position of the face landmarks are assumed to be relative to\n   * it's current shift.\n   * @returns The bounding box of the aligned face.\n   */\n  public align(\n    detection?: FaceDetection | IRect | IBoundingBox | null,\n    options: { useDlibAlignment?: boolean, minBoxPadding?: number } = { },\n  ): Box {\n    if (detection) {\n      const box = detection instanceof FaceDetection\n        ? detection.box.floor()\n        : new Box(detection);\n\n      return this.shiftBy(box.x, box.y).align(null, options);\n    }\n\n    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };\n\n    if (useDlibAlignment) {\n      return this.alignDlib();\n    }\n\n    return this.alignMinBbox(minBoxPadding);\n  }\n\n  private alignDlib(): Box {\n    const centers = this.getRefPointsForAlignment();\n\n    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;\n    const distToMouth = (pt: Point) => mouthCenter.sub(pt).magnitude();\n    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;\n\n    const size = Math.floor(eyeToMouthDist / relScale);\n\n    const refPoint = getCenterPoint(centers);\n    // TODO: pad in case rectangle is out of image bounds\n    const x = Math.floor(Math.max(0, refPoint.x - (relX * size)));\n    const y = Math.floor(Math.max(0, refPoint.y - (relY * size)));\n\n    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));\n  }\n\n  private alignMinBbox(padding: number): Box {\n    const box = minBbox(this.positions);\n    return box.pad(box.width * padding, box.height * padding);\n  }\n\n  protected getRefPointsForAlignment(): Point[] {\n    throw new Error('getRefPointsForAlignment not implemented by base class');\n  }\n}\n", "import { getCenterPoint } from '../utils/index';\nimport { FaceLandmarks } from './FaceLandmarks';\nimport { Point } from './Point';\n\nexport class FaceLandmarks5 extends FaceLandmarks {\n  protected override getRefPointsForAlignment(): Point[] {\n    const pts = this.positions;\n    return [\n      pts[0],\n      pts[1],\n      getCenterPoint([pts[3], pts[4]]),\n    ];\n  }\n}\n", "import { getCenterPoint } from '../utils/index';\nimport { FaceLandmarks } from './FaceLandmarks';\nimport { Point } from './Point';\n\nexport class FaceLandmarks68 extends FaceLandmarks {\n  public getJawOutline(): Point[] {\n    return this.positions.slice(0, 17);\n  }\n\n  public getLeftEyeBrow(): Point[] {\n    return this.positions.slice(17, 22);\n  }\n\n  public getRightEyeBrow(): Point[] {\n    return this.positions.slice(22, 27);\n  }\n\n  public getNose(): Point[] {\n    return this.positions.slice(27, 36);\n  }\n\n  public getLeftEye(): Point[] {\n    return this.positions.slice(36, 42);\n  }\n\n  public getRightEye(): Point[] {\n    return this.positions.slice(42, 48);\n  }\n\n  public getMouth(): Point[] {\n    return this.positions.slice(48, 68);\n  }\n\n  protected override getRefPointsForAlignment(): Point[] {\n    return [\n      this.getLeftEye(),\n      this.getRightEye(),\n      this.getMouth(),\n    ].map(getCenterPoint);\n  }\n}\n", "import { round } from '../utils/index';\n\nexport interface IFaceMatch {\n  label: string\n  distance: number\n}\n\nexport class FaceMatch implements IFaceMatch {\n  private _label: string;\n  private _distance: number;\n\n  constructor(label: string, distance: number) {\n    this._label = label;\n    this._distance = distance;\n  }\n\n  public get label(): string { return this._label; }\n\n  public get distance(): number { return this._distance; }\n\n  public toString(withDistance = true): string {\n    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ''}`;\n  }\n}\n", "import { isValidNumber } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { Box } from './Box';\nimport { IRect } from './Rect';\n\nexport class LabeledBox extends Box {\n  public static assertIsValidLabeledBox(box: any, callee: string) {\n    Box.assertIsValidBox(box, callee);\n    if (!isValidNumber(box.label)) {\n      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);\n    }\n  }\n\n  private _label: number;\n\n  constructor(box: IBoundingBox | IRect | any, label: number) {\n    super(box);\n    this._label = label;\n  }\n\n  public get label(): number { return this._label; }\n}\n", "export class LabeledFaceDescriptors {\n  private _label: string;\n\n  private _descriptors: Float32Array[];\n\n  constructor(label: string, descriptors: Float32Array[]) {\n    if (!(typeof label === 'string')) {\n      throw new Error('LabeledFaceDescriptors - constructor expected label to be a string');\n    }\n\n    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {\n      throw new Error('LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array');\n    }\n\n    this._label = label;\n    this._descriptors = descriptors;\n  }\n\n  public get label(): string { return this._label; }\n\n  public get descriptors(): Float32Array[] { return this._descriptors; }\n\n  public toJSON(): any {\n    return {\n      label: this.label,\n      descriptors: this.descriptors.map((d) => Array.from(d)),\n    };\n  }\n\n  public static fromJSON(json: any): LabeledFaceDescriptors {\n    const descriptors = json.descriptors.map((d: any) => new Float32Array(d));\n    return new LabeledFaceDescriptors(json.label, descriptors);\n  }\n}\n", "import { isValidProbablitiy } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { LabeledBox } from './LabeledBox';\nimport { IRect } from './Rect';\n\nexport class PredictedBox extends LabeledBox {\n  public static assertIsValidPredictedBox(box: any, callee: string) {\n    LabeledBox.assertIsValidLabeledBox(box, callee);\n\n    if (\n      !isValidProbablitiy(box.score)\n      || !isValidProbablitiy(box.classScore)\n    ) {\n      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);\n    }\n  }\n\n  private _score: number;\n\n  private _classScore: number;\n\n  constructor(box: IBoundingBox | IRect | any, label: number, score: number, classScore: number) {\n    super(box, label);\n    this._score = score;\n    this._classScore = classScore;\n  }\n\n  public get score(): number { return this._score; }\n\n  public get classScore(): number { return this._classScore; }\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\n\nexport type WithFaceDetection<TSource> = TSource & {\n  detection: FaceDetection\n}\n\nexport function isWithFaceDetection(obj: any): obj is WithFaceDetection<{}> {\n  return obj.detection instanceof FaceDetection;\n}\n\nexport function extendWithFaceDetection<TSource>(sourceObj: TSource, detection: FaceDetection): WithFaceDetection<TSource> {\n  const extension = { detection };\n  return { ...sourceObj, ...extension };\n}\n", "import { Environment } from './types';\n\nexport function createBrowserEnv(): Environment {\n  const fetch = window.fetch;\n  if (!fetch) throw new Error('fetch - missing fetch implementation for browser environment');\n\n  const readFile = () => {\n    throw new Error('readFile - filesystem not available for browser environment');\n  };\n\n  return {\n    Canvas: HTMLCanvasElement,\n    CanvasRenderingContext2D,\n    Image: HTMLImageElement,\n    ImageData,\n    Video: HTMLVideoElement,\n    createCanvasElement: () => document.createElement('canvas'),\n    createImageElement: () => document.createElement('img'),\n    createVideoElement: () => document.createElement('video'),\n    fetch,\n    readFile,\n  };\n}\n", "export function isNodejs(): boolean {\n  return typeof global === 'object'\n    && typeof process !== 'undefined'\n    && process.versions != null\n    && process.versions.node != null;\n}\n", "import { FileSystem } from './types';\nimport { isNodejs } from './isNodejs';\n\nexport function createFileSystem(fs?: any): FileSystem {\n  let requireFsError = '';\n  if (!fs && isNodejs()) {\n    try {\n      // eslint-disable-next-line global-require\n      fs = require('fs');\n    } catch (err) {\n      requireFsError = (err as any).toString();\n    }\n  }\n\n  const readFile = fs\n    ? (filePath: string) => new Promise((resolve, reject) => { fs.readFile(filePath, (err: any, buffer) => (err ? reject(err) : resolve(buffer))); })\n    : () => { throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`); };\n  return { readFile };\n}\n", "/* eslint-disable max-classes-per-file */\nimport { createFileSystem } from './createFileSystem';\nimport { Environment } from './types';\n\nexport function createNodejsEnv(): Environment {\n  // eslint-disable-next-line dot-notation\n  const Canvas = global['Canvas'] || global.HTMLCanvasElement;\n  const Image = global.Image || global.HTMLImageElement;\n  // eslint-disable-next-line dot-notation\n  const Video = global['Video'] || global.HTMLVideoElement;\n\n  const createCanvasElement = () => {\n    if (Canvas) return new Canvas();\n    throw new Error('createCanvasElement - missing Canvas implementation for nodejs environment');\n  };\n\n  const createImageElement = () => {\n    if (Image) return new Image();\n    throw new Error('createImageElement - missing Image implementation for nodejs environment');\n  };\n\n  const createVideoElement = () => {\n    if (Video) return new Video();\n    throw new Error('createVideoElement - missing Video implementation for nodejs environment');\n  };\n\n  const fetch = global.fetch;\n  // if (!fetch) throw new Error('fetch - missing fetch implementation for nodejs environment');\n\n  const fileSystem = createFileSystem();\n\n  return {\n    Canvas: Canvas || class {},\n    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {},\n    Image: Image || class {},\n    ImageData: global.ImageData || class {},\n    Video: global.HTMLVideoElement || class {},\n    createCanvasElement,\n    createImageElement,\n    createVideoElement,\n    fetch,\n    ...fileSystem,\n  };\n}\n", "export function isBrowser(): boolean {\n  return typeof window === 'object'\n    && typeof document !== 'undefined'\n    && typeof HTMLImageElement !== 'undefined'\n    && typeof HTMLCanvasElement !== 'undefined'\n    && typeof HTMLVideoElement !== 'undefined'\n    && typeof ImageData !== 'undefined'\n    && typeof CanvasRenderingContext2D !== 'undefined';\n}\n", "import { createBrowserEnv } from './createBrowserEnv';\nimport { createFileSystem } from './createFileSystem';\nimport { createNodejsEnv } from './createNodejsEnv';\nimport { isBrowser } from './isBrowser';\nimport { isNodejs } from './isNodejs';\nimport { Environment } from './types';\n\nlet environment: Environment | null;\n\nfunction getEnv(): Environment {\n  if (!environment) {\n    throw new Error('getEnv - environment is not defined, check isNodejs() and isBrowser()');\n  }\n  return environment;\n}\n\nfunction setEnv(env: Environment) {\n  environment = env;\n}\n\nfunction initialize() {\n  // check for isBrowser() first to prevent electron renderer process\n  // to be initialized with wrong environment due to isNodejs() returning true\n  if (isBrowser()) return setEnv(createBrowserEnv());\n  if (isNodejs()) return setEnv(createNodejsEnv());\n  return null;\n}\n\nfunction monkeyPatch(env: Partial<Environment>) {\n  if (!environment) {\n    initialize();\n  }\n\n  if (!environment) {\n    throw new Error('monkeyPatch - environment is not defined, check isNodejs() and isBrowser()');\n  }\n\n  const { Canvas = environment.Canvas, Image = environment.Image } = env;\n  environment.Canvas = Canvas;\n  environment.Image = Image;\n  environment.createCanvasElement = env.createCanvasElement || (() => new Canvas());\n  environment.createImageElement = env.createImageElement || (() => new Image());\n\n  environment.ImageData = env.ImageData || environment.ImageData;\n  environment.Video = env.Video || environment.Video;\n  environment.fetch = env.fetch || environment.fetch;\n  environment.readFile = env.readFile || environment.readFile;\n}\n\nexport const env = {\n  getEnv,\n  setEnv,\n  initialize,\n  createBrowserEnv,\n  createFileSystem,\n  createNodejsEnv,\n  monkeyPatch,\n  isBrowser,\n  isNodejs,\n};\n\ninitialize();\n\nexport * from './types';\n", "import { env } from '../env/index';\n\nexport function resolveInput(arg: string | any) {\n  if (!env.isNodejs() && typeof arg === 'string') {\n    return document.getElementById(arg);\n  }\n  return arg;\n}\n", "import { env } from '../env/index';\nimport { resolveInput } from './resolveInput';\n\nexport function getContext2dOrThrow(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D): CanvasRenderingContext2D {\n  const { Canvas, CanvasRenderingContext2D } = env.getEnv();\n\n  if (canvasArg instanceof CanvasRenderingContext2D) {\n    return canvasArg;\n  }\n\n  const canvas = resolveInput(canvasArg);\n\n  if (!(canvas instanceof Canvas)) {\n    throw new Error('resolveContext2d - expected canvas to be of instance of Canvas');\n  }\n\n  const ctx = canvas.getContext('2d');\n  if (!ctx) {\n    throw new Error('resolveContext2d - canvas 2d context is null');\n  }\n\n  return ctx;\n}\n", "/* eslint-disable max-classes-per-file */\nimport { IDimensions, IPoint } from '../classes/index';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { resolveInput } from '../dom/resolveInput';\n\n// eslint-disable-next-line no-shadow\nexport enum AnchorPosition {\n  // eslint-disable-next-line no-unused-vars\n  TOP_LEFT = 'TOP_LEFT',\n  // eslint-disable-next-line no-unused-vars\n  TOP_RIGHT = 'TOP_RIGHT',\n  // eslint-disable-next-line no-unused-vars\n  BOTTOM_LEFT = 'BOTTOM_LEFT',\n  // eslint-disable-next-line no-unused-vars\n  BOTTOM_RIGHT = 'BOTTOM_RIGHT'\n}\n\nexport interface IDrawTextFieldOptions {\n  anchorPosition?: AnchorPosition\n  backgroundColor?: string\n  fontColor?: string\n  fontSize?: number\n  fontStyle?: string\n  padding?: number\n}\n\nexport class DrawTextFieldOptions implements IDrawTextFieldOptions {\n  public anchorPosition: AnchorPosition;\n\n  public backgroundColor: string;\n\n  public fontColor: string;\n\n  public fontSize: number;\n\n  public fontStyle: string;\n\n  public padding: number;\n\n  constructor(options: IDrawTextFieldOptions = {}) {\n    const {\n      anchorPosition, backgroundColor, fontColor, fontSize, fontStyle, padding,\n    } = options;\n    this.anchorPosition = anchorPosition || AnchorPosition.TOP_LEFT;\n    this.backgroundColor = backgroundColor || 'rgba(0, 0, 0, 0.5)';\n    this.fontColor = fontColor || 'rgba(255, 255, 255, 1)';\n    this.fontSize = fontSize || 14;\n    this.fontStyle = fontStyle || 'Georgia';\n    this.padding = padding || 4;\n  }\n}\n\nexport class DrawTextField {\n  public text: string[];\n\n  public anchor : IPoint;\n\n  public options: DrawTextFieldOptions;\n\n  constructor(\n    text: string | string[] | DrawTextField,\n    anchor: IPoint,\n    options: IDrawTextFieldOptions = {},\n  ) {\n    // eslint-disable-next-line no-nested-ternary\n    this.text = typeof text === 'string'\n      ? [text]\n      : (text instanceof DrawTextField ? text.text : text);\n    this.anchor = anchor;\n    this.options = new DrawTextFieldOptions(options);\n  }\n\n  measureWidth(ctx: CanvasRenderingContext2D): number {\n    const { padding } = this.options;\n    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => (w0 < w1 ? w1 : w0), 0) + (2 * padding);\n  }\n\n  measureHeight(): number {\n    const { fontSize, padding } = this.options;\n    return this.text.length * fontSize + (2 * padding);\n  }\n\n  getUpperLeft(ctx: CanvasRenderingContext2D, canvasDims?: IDimensions): IPoint {\n    const { anchorPosition } = this.options;\n    const isShiftLeft = anchorPosition === AnchorPosition.BOTTOM_RIGHT || anchorPosition === AnchorPosition.TOP_RIGHT;\n    const isShiftTop = anchorPosition === AnchorPosition.BOTTOM_LEFT || anchorPosition === AnchorPosition.BOTTOM_RIGHT;\n\n    const textFieldWidth = this.measureWidth(ctx);\n    const textFieldHeight = this.measureHeight();\n    const x = (isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x);\n    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;\n\n    // adjust anchor if text box exceeds canvas borders\n    if (canvasDims) {\n      const { width, height } = canvasDims;\n      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);\n      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);\n      return { x: newX, y: newY };\n    }\n    return { x, y };\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const canvas = resolveInput(canvasArg);\n    const ctx = getContext2dOrThrow(canvas);\n\n    const {\n      backgroundColor, fontColor, fontSize, fontStyle, padding,\n    } = this.options;\n\n    ctx.font = `${fontSize}px ${fontStyle}`;\n    const maxTextWidth = this.measureWidth(ctx);\n    const textHeight = this.measureHeight();\n\n    ctx.fillStyle = backgroundColor;\n    const upperLeft = this.getUpperLeft(ctx, canvas);\n    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);\n\n    ctx.fillStyle = fontColor;\n    this.text.forEach((textLine, i) => {\n      const x = padding + upperLeft.x;\n      const y = padding + upperLeft.y + ((i + 1) * fontSize);\n      ctx.fillText(textLine, x, y);\n    });\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { Box, IBoundingBox, IRect } from '../classes/index';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { AnchorPosition, DrawTextField, DrawTextFieldOptions, IDrawTextFieldOptions } from './DrawTextField';\n\nexport interface IDrawBoxOptions {\n  boxColor?: string\n  lineWidth?: number\n  drawLabelOptions?: IDrawTextFieldOptions\n  label?: string\n}\n\nexport class DrawBoxOptions {\n  public boxColor: string;\n\n  public lineWidth: number;\n\n  public drawLabelOptions: DrawTextFieldOptions;\n\n  public label?: string;\n\n  constructor(options: IDrawBoxOptions = {}) {\n    const {\n      boxColor, lineWidth, label, drawLabelOptions,\n    } = options;\n    this.boxColor = boxColor || 'rgba(0, 0, 255, 1)';\n    this.lineWidth = lineWidth || 2;\n    this.label = label;\n\n    const defaultDrawLabelOptions = {\n      anchorPosition: AnchorPosition.BOTTOM_LEFT,\n      backgroundColor: this.boxColor,\n    };\n    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });\n  }\n}\n\nexport class DrawBox {\n  public box: Box;\n\n  public options: DrawBoxOptions;\n\n  constructor(\n    box: IBoundingBox | IRect,\n    options: IDrawBoxOptions = {},\n  ) {\n    this.box = new Box(box);\n    this.options = new DrawBoxOptions(options);\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const ctx = getContext2dOrThrow(canvasArg);\n\n    const { boxColor, lineWidth } = this.options;\n\n    const {\n      x, y, width, height,\n    } = this.box;\n    ctx.strokeStyle = boxColor;\n    ctx.lineWidth = lineWidth;\n    ctx.strokeRect(x, y, width, height);\n\n    const { label } = this.options;\n    if (label) {\n      new DrawTextField([label], { x: x - (lineWidth / 2), y }, this.options.drawLabelOptions).draw(canvasArg);\n    }\n  }\n}\n", "import { Box, IBoundingBox, IRect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { isWithFaceDetection, WithFaceDetection } from '../factories/WithFaceDetection';\nimport { round } from '../utils/index';\nimport { DrawBox } from './DrawBox';\n\nexport type TDrawDetectionsInput = IRect | IBoundingBox | FaceDetection | WithFaceDetection<{}>\n\nexport function drawDetections(\n  canvasArg: string | HTMLCanvasElement,\n  detections: TDrawDetectionsInput | Array<TDrawDetectionsInput>,\n) {\n  const detectionsArray = Array.isArray(detections) ? detections : [detections];\n\n  detectionsArray.forEach((det) => {\n    // eslint-disable-next-line no-nested-ternary\n    const score = det instanceof FaceDetection\n      ? det.score\n      : (isWithFaceDetection(det) ? det.detection.score : undefined);\n\n    // eslint-disable-next-line no-nested-ternary\n    const box = det instanceof FaceDetection\n      ? det.box\n      : (isWithFaceDetection(det) ? det.detection.box : new Box(det));\n\n    const label = score ? `${round(score)}` : undefined;\n    new DrawBox(box, { label }).draw(canvasArg);\n  });\n}\n", "import { env } from '../env/index';\n\nexport function isMediaLoaded(media: HTMLImageElement | HTMLVideoElement) : boolean {\n  const { Image, Video } = env.getEnv();\n\n  return (media instanceof Image && media.complete)\n    || (media instanceof Video && media.readyState >= 3);\n}\n", "import { env } from '../env/index';\nimport { isMediaLoaded } from './isMediaLoaded';\n\nexport function awaitMediaLoaded(media: HTMLImageElement | HTMLVideoElement | HTMLCanvasElement) {\n  // eslint-disable-next-line consistent-return\n  return new Promise((resolve, reject) => {\n    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media)) resolve(null);\n\n    function onError(e: Event) {\n      if (!e.currentTarget) return;\n      // eslint-disable-next-line no-use-before-define\n      e.currentTarget.removeEventListener('load', onLoad);\n      e.currentTarget.removeEventListener('error', onError);\n      reject(e);\n    }\n\n    function onLoad(e: Event) {\n      if (!e.currentTarget) return;\n      e.currentTarget.removeEventListener('load', onLoad);\n      e.currentTarget.removeEventListener('error', onError);\n      resolve(e);\n    }\n\n    media.addEventListener('load', onLoad);\n    media.addEventListener('error', onError);\n  });\n}\n", "import { env } from '../env/index';\n\nexport function bufferToImage(buf: Blob): Promise<HTMLImageElement> {\n  return new Promise((resolve, reject) => {\n    if (!(buf instanceof Blob)) reject(new Error('bufferToImage - expected buf to be of type: Blob'));\n    const reader = new FileReader();\n    reader.onload = () => {\n      if (typeof reader.result !== 'string') reject(new Error('bufferToImage - expected reader.result to be a string, in onload'));\n      const img = env.getEnv().createImageElement();\n      img.onload = () => resolve(img);\n      img.onerror = reject;\n      img.src = reader.result as string;\n    };\n    reader.onerror = reject;\n    reader.readAsDataURL(buf);\n  });\n}\n", "import { Dimensions, IDimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\n\nexport function getMediaDimensions(input: HTMLImageElement | HTMLCanvasElement | HTMLVideoElement | IDimensions): Dimensions {\n  const { Image, Video } = env.getEnv();\n\n  if (input instanceof Image) {\n    return new Dimensions(input.naturalWidth, input.naturalHeight);\n  }\n  if (input instanceof Video) {\n    return new Dimensions(input.videoWidth, input.videoHeight);\n  }\n  return new Dimensions(input.width, input.height);\n}\n", "import { IDimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { getMediaDimensions } from './getMediaDimensions';\nimport { isMediaLoaded } from './isMediaLoaded';\n\nexport function createCanvas({ width, height }: IDimensions): HTMLCanvasElement {\n  const { createCanvasElement } = env.getEnv();\n  const canvas = createCanvasElement();\n  canvas.width = width;\n  canvas.height = height;\n  return canvas;\n}\n\nexport function createCanvasFromMedia(media: HTMLImageElement | HTMLVideoElement | ImageData, dims?: IDimensions): HTMLCanvasElement {\n  const { ImageData } = env.getEnv();\n\n  if (!(media instanceof ImageData) && !isMediaLoaded(media)) {\n    throw new Error('createCanvasFromMedia - media has not finished loading yet');\n  }\n\n  const { width, height } = dims || getMediaDimensions(media);\n  const canvas = createCanvas({ width, height });\n\n  if (media instanceof ImageData) {\n    getContext2dOrThrow(canvas).putImageData(media, 0, 0);\n  } else {\n    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);\n  }\n  return canvas;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { env } from '../env/index';\nimport { isTensor4D } from '../utils/index';\n\nexport async function imageTensorToCanvas(\n  imgTensor: tf.Tensor,\n  canvas?: HTMLCanvasElement,\n): Promise<HTMLCanvasElement> {\n  const targetCanvas = canvas || env.getEnv().createCanvasElement();\n\n  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);\n  const imgTensor3D = tf.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());\n  await tf['browser'].toPixels(imgTensor3D, targetCanvas);\n\n  imgTensor3D.dispose();\n\n  return targetCanvas;\n}\n", "import { env } from '../env/index';\n\nexport function isMediaElement(input: any) {\n  const { Image, Canvas, Video } = env.getEnv();\n\n  return input instanceof Image\n    || input instanceof Canvas\n    || input instanceof Video;\n}\n", "import { env } from '../env/index';\nimport { createCanvas, createCanvasFromMedia } from './createCanvas';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { getMediaDimensions } from './getMediaDimensions';\n\nexport function imageToSquare(input: HTMLImageElement | HTMLCanvasElement, inputSize: number, centerImage = false) {\n  const { Image, Canvas } = env.getEnv();\n\n  if (!(input instanceof Image || input instanceof Canvas)) {\n    throw new Error('imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement');\n  }\n\n  if (inputSize <= 0) return createCanvas({ width: 1, height: 1 });\n  const dims = getMediaDimensions(input);\n  const scale = inputSize / Math.max(dims.height, dims.width);\n  const width = scale * dims.width;\n  const height = scale * dims.height;\n\n  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });\n  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);\n\n  const offset = Math.abs(width - height) / 2;\n  const dx = centerImage && width < height ? offset : 0;\n  const dy = centerImage && height < width ? offset : 0;\n  if (inputCanvas.width > 0 && inputCanvas.height > 0) getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);\n\n  return targetCanvas;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Dimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\nimport { padToSquare } from '../ops/padToSquare';\nimport { computeReshapedDimensions, isTensor3D, isTensor4D, range } from '../utils/index';\nimport { createCanvasFromMedia } from './createCanvas';\nimport { imageToSquare } from './imageToSquare';\nimport { TResolvedNetInput } from './types';\n\nexport class NetInput {\n  private _imageTensors: Array<tf.Tensor3D | tf.Tensor4D> = [];\n\n  private _canvases: HTMLCanvasElement[] = [];\n\n  private _batchSize: number;\n\n  private _treatAsBatchInput = false;\n\n  private _inputDimensions: number[][] = [];\n\n  private _inputSize = 0;\n\n  constructor(inputs: Array<TResolvedNetInput>, treatAsBatchInput = false) {\n    if (!Array.isArray(inputs)) {\n      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);\n    }\n\n    this._treatAsBatchInput = treatAsBatchInput;\n    this._batchSize = inputs.length;\n\n    inputs.forEach((input, idx) => {\n      if (isTensor3D(input)) {\n        this._imageTensors[idx] = input;\n        this._inputDimensions[idx] = input.shape;\n        return;\n      }\n\n      if (isTensor4D(input)) {\n        const batchSize = (input as any).shape[0];\n        if (batchSize !== 1) {\n          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);\n        }\n\n        this._imageTensors[idx] = input;\n        this._inputDimensions[idx] = (input as any).shape.slice(1);\n        return;\n      }\n\n      // @ts-ignore\n      const canvas = (input as any) instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);\n      this._canvases[idx] = canvas as HTMLCanvasElement;\n      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];\n    });\n  }\n\n  public get imageTensors(): Array<tf.Tensor3D | tf.Tensor4D> {\n    return this._imageTensors;\n  }\n\n  public get canvases(): HTMLCanvasElement[] {\n    return this._canvases;\n  }\n\n  public get isBatchInput(): boolean {\n    return this.batchSize > 1 || this._treatAsBatchInput;\n  }\n\n  public get batchSize(): number {\n    return this._batchSize;\n  }\n\n  public get inputDimensions(): number[][] {\n    return this._inputDimensions;\n  }\n\n  public get inputSize(): number | undefined {\n    return this._inputSize;\n  }\n\n  public get reshapedInputDimensions(): Dimensions[] {\n    return range(this.batchSize, 0, 1).map(\n      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx),\n    );\n  }\n\n  public getInput(batchIdx: number): tf.Tensor3D | tf.Tensor4D | HTMLCanvasElement {\n    return this.canvases[batchIdx] || this.imageTensors[batchIdx];\n  }\n\n  public getInputDimensions(batchIdx: number): number[] {\n    return this._inputDimensions[batchIdx];\n  }\n\n  public getInputHeight(batchIdx: number): number {\n    return this._inputDimensions[batchIdx][0];\n  }\n\n  public getInputWidth(batchIdx: number): number {\n    return this._inputDimensions[batchIdx][1];\n  }\n\n  public getReshapedInputDimensions(batchIdx: number): Dimensions {\n    if (typeof this.inputSize !== 'number') {\n      throw new Error('getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet');\n    }\n\n    const width = this.getInputWidth(batchIdx);\n    const height = this.getInputHeight(batchIdx);\n    return computeReshapedDimensions({ width, height }, this.inputSize);\n  }\n\n  /**\n   * Create a batch tensor from all input canvases and tensors\n   * with size [batchSize, inputSize, inputSize, 3].\n   *\n   * @param inputSize Height and width of the tensor.\n   * @param isCenterImage (optional, default: false) If true, add an equal amount of padding on\n   * both sides of the minor dimension oof the image.\n   * @returns The batch tensor.\n   */\n  public toBatchTensor(inputSize: number, isCenterInputs = true): tf.Tensor4D {\n    this._inputSize = inputSize;\n\n    return tf.tidy(() => {\n      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {\n        const input = this.getInput(batchIdx);\n\n        if (input instanceof tf.Tensor) {\n          let imgTensor = isTensor4D(input) ? input : tf.expandDims(input);\n          imgTensor = padToSquare(imgTensor as tf.Tensor4D, isCenterInputs);\n\n          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {\n            imgTensor = tf['image'].resizeBilinear(imgTensor as tf.Tensor4D, [inputSize, inputSize], false, false);\n          }\n\n          return imgTensor.as3D(inputSize, inputSize, 3);\n        }\n\n        if (input instanceof env.getEnv().Canvas) {\n          return tf['browser'].fromPixels(imageToSquare(input, inputSize, isCenterInputs));\n        }\n\n        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);\n      });\n\n      const batchTensor = tf.stack(inputTensors.map((t) => tf.cast(t, 'float32'))).as4D(this.batchSize, inputSize, inputSize, 3);\n      // const batchTensor = tf.stack(inputTensors.map((t) => tf.cast(t, 'float32'))) as tf.Tensor4D;\n\n      return batchTensor;\n    });\n  }\n}\n", "import { isTensor3D, isTensor4D } from '../utils/index';\nimport { awaitMediaLoaded } from './awaitMediaLoaded';\nimport { isMediaElement } from './isMediaElement';\nimport { NetInput } from './NetInput';\nimport { resolveInput } from './resolveInput';\nimport { TNetInput } from './types';\n\n/**\n * Validates the input to make sure, they are valid net inputs and awaits all media elements\n * to be finished loading.\n *\n * @param input The input, which can be a media element or an array of different media elements.\n * @returns A NetInput instance, which can be passed into one of the neural networks.\n */\nexport async function toNetInput(inputs: TNetInput): Promise<NetInput> {\n  if (inputs instanceof NetInput) return inputs;\n  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];\n  if (!inputArgArray.length) throw new Error('toNetInput - empty array passed as input');\n  const getIdxHint = (idx: number) => (Array.isArray(inputs) ? ` at input index ${idx}:` : '');\n  const inputArray = inputArgArray.map(resolveInput);\n  inputArray.forEach((input, i) => {\n    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {\n      if (typeof inputArgArray[i] === 'string') throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);\n      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);\n    }\n    if (isTensor4D(input)) {\n      // if tf.Tensor4D is passed in the input array, the batch size has to be 1\n      const batchSize = input.shape[0];\n      if (batchSize !== 1) throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);\n    }\n  });\n  // wait for all media elements being loaded\n  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));\n  return new NetInput(inputArray, Array.isArray(inputs));\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\nimport { Rect } from '../classes/Rect';\nimport { env } from '../env/index';\nimport { createCanvas } from './createCanvas';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { imageTensorToCanvas } from './imageTensorToCanvas';\nimport { toNetInput } from './toNetInput';\nimport { TNetInput } from './types';\n\n/**\n * Extracts the image regions containing the detected faces.\n *\n * @param input The image that face detection has been performed on.\n * @param detections The face detection results or face bounding boxes for that image.\n * @returns The Canvases of the corresponding image region for each detected face.\n */\nexport async function extractFaces(input: TNetInput, detections: Array<FaceDetection | Rect>): Promise<HTMLCanvasElement[]> {\n  const { Canvas } = env.getEnv();\n  let canvas = input as HTMLCanvasElement;\n  if (!(input instanceof Canvas)) {\n    const netInput = await toNetInput(input);\n    if (netInput.batchSize > 1) throw new Error('extractFaces - batchSize > 1 not supported');\n    const tensorOrCanvas = netInput.getInput(0);\n    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);\n  }\n  const ctx = getContext2dOrThrow(canvas);\n  const boxes = detections\n    .map((det) => (det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det))\n    .map((box) => box.clipAtImageBorders(canvas.width, canvas.height));\n  return boxes.map(({ x, y, width, height }) => {\n    const faceImg = createCanvas({ width, height });\n    if (width > 0 && height > 0) getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);\n    return faceImg;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Rect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { isTensor3D, isTensor4D } from '../utils/index';\n\n/**\n * Extracts the tensors of the image regions containing the detected faces.\n * Useful if you want to compute the face descriptors for the face images.\n * Using this method is faster then extracting a canvas for each face and\n * converting them to tensors individually.\n *\n * @param imageTensor The image tensor that face detection has been performed on.\n * @param detections The face detection results or face bounding boxes for that image.\n * @returns Tensors of the corresponding image region for each detected face.\n */\nexport async function extractFaceTensors(imageTensor: tf.Tensor3D | tf.Tensor4D, detections: Array<FaceDetection | Rect>): Promise<tf.Tensor3D[]> {\n  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {\n    throw new Error('extractFaceTensors - expected image tensor to be 3D or 4D');\n  }\n\n  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {\n    throw new Error('extractFaceTensors - batchSize > 1 not supported');\n  }\n\n  return tf.tidy(() => {\n    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);\n    const boxes = detections.map((det) => (det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det))\n      .map((box) => box.clipAtImageBorders(imgWidth, imgHeight));\n    const faceTensors = boxes\n      .filter((box) => box.width > 0 && box.height > 0)\n      .map(({ x, y, width, height }) => tf.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));\n    return faceTensors;\n  });\n}\n", "import { env } from '../env/index';\n\nexport async function fetchOrThrow(\n  url: string,\n  // eslint-disable-next-line no-undef\n  init?: RequestInit,\n): Promise<Response> {\n  const { fetch } = env.getEnv();\n  const res = await fetch(url, init);\n  if (!(res.status < 400)) {\n    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);\n  }\n  return res;\n}\n", "import { bufferToImage } from './bufferToImage';\nimport { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchImage(uri: string): Promise<HTMLImageElement> {\n  const res = await fetchOrThrow(uri);\n  const blob = await (res).blob();\n\n  if (!blob.type.startsWith('image/')) {\n    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);\n  }\n  return bufferToImage(blob);\n}\n", "import { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchJson<T>(uri: string): Promise<T> {\n  return (await fetchOrThrow(uri)).json();\n}\n", "import { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchNetWeights(uri: string): Promise<Float32Array> {\n  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());\n}\n", "import { env } from '../env/index';\n\nexport function bufferToVideo(buf: Blob): Promise<HTMLVideoElement> {\n  return new Promise((resolve, reject) => {\n    if (!(buf instanceof Blob)) reject(new Error('bufferToVideo - expected buf to be of type: Blob'));\n\n    const video = env.getEnv().createVideoElement();\n    video.oncanplay = () => resolve(video);\n    video.onerror = reject;\n    video.playsInline = true;\n    video.muted = true;\n    video.src = URL.createObjectURL(buf);\n    video.play();\n  });\n}\n", "import { bufferToVideo } from './bufferToVideo';\nimport { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchVideo(uri: string): Promise<HTMLVideoElement> {\n  const res = await fetchOrThrow(uri);\n  const blob = await (res).blob();\n\n  if (!blob.type.startsWith('video/')) {\n    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);\n  }\n  return bufferToVideo(blob);\n}\n", "export function getModelUris(uri: string | undefined, defaultModelName: string) {\n  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;\n\n  if (!uri) {\n    return {\n      modelBaseUri: '',\n      manifestUri: defaultManifestFilename,\n    };\n  }\n\n  if (uri === '/') {\n    return {\n      modelBaseUri: '/',\n      manifestUri: `/${defaultManifestFilename}`,\n    };\n  }\n  // eslint-disable-next-line no-nested-ternary\n  const protocol = uri.startsWith('http://') ? 'http://' : uri.startsWith('https://') ? 'https://' : '';\n  uri = uri.replace(protocol, '');\n\n  const parts = uri.split('/').filter((s) => s);\n\n  const manifestFile = uri.endsWith('.json')\n    ? parts[parts.length - 1]\n    : defaultManifestFilename;\n\n  let modelBaseUri = protocol + (uri.endsWith('.json') ? parts.slice(0, parts.length - 1) : parts).join('/');\n  modelBaseUri = uri.startsWith('/') ? `/${modelBaseUri}` : modelBaseUri;\n\n  return {\n    modelBaseUri,\n    manifestUri: modelBaseUri === '/' ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { getModelUris } from '../common/getModelUris';\nimport { fetchJson } from './fetchJson';\n\nexport async function loadWeightMap(\n  uri: string | undefined,\n  defaultModelName: string,\n): Promise<tf.NamedTensorMap> {\n  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);\n  // @ts-ignore\n  const manifest = await fetchJson<tf.io.WeightsManifestConfig>(manifestUri);\n  // if (manifest['weightsManifest']) manifest = manifest['weightsManifest'];\n  return tf['io'].loadWeights(manifest, modelBaseUri);\n}\n", "import { IDimensions } from '../classes/index';\nimport { getMediaDimensions } from './getMediaDimensions';\n\nexport function matchDimensions(input: IDimensions, reference: IDimensions, useMediaDimensions = false) {\n  const { width, height } = useMediaDimensions\n    ? getMediaDimensions(reference)\n    : reference;\n  input.width = width;\n  input.height = height;\n  return { width, height };\n}\n", "import * as tf from '../dist/tfjs.esm';\n\nimport { ParamMapping } from './common/index';\nimport { getModelUris } from './common/getModelUris';\nimport { loadWeightMap } from './dom/index';\nimport { env } from './env/index';\n\nexport abstract class NeuralNetwork<TNetParams> {\n  constructor(name: string) {\n    this._name = name;\n  }\n\n  protected _params: TNetParams | undefined = undefined;\n\n  protected _paramMappings: ParamMapping[] = [];\n\n  public _name: any;\n\n  public get params(): TNetParams | undefined { return this._params; }\n\n  public get paramMappings(): ParamMapping[] { return this._paramMappings; }\n\n  public get isLoaded(): boolean { return !!this.params; }\n\n  public getParamFromPath(paramPath: string): tf.Tensor {\n    const { obj, objProp } = this.traversePropertyPath(paramPath);\n    return obj[objProp];\n  }\n\n  public reassignParamFromPath(paramPath: string, tensor: tf.Tensor) {\n    const { obj, objProp } = this.traversePropertyPath(paramPath);\n    obj[objProp].dispose();\n    obj[objProp] = tensor;\n  }\n\n  public getParamList() {\n    return this._paramMappings.map(({ paramPath }) => ({\n      path: paramPath,\n      tensor: this.getParamFromPath(paramPath),\n    }));\n  }\n\n  public getTrainableParams() {\n    return this.getParamList().filter((param) => param.tensor instanceof tf.Variable);\n  }\n\n  public getFrozenParams() {\n    return this.getParamList().filter((param) => !(param.tensor instanceof tf.Variable));\n  }\n\n  public variable() {\n    this.getFrozenParams().forEach(({ path, tensor }) => {\n      this.reassignParamFromPath(path, tensor.variable());\n    });\n  }\n\n  public freeze() {\n    this.getTrainableParams().forEach(({ path, tensor: variable }) => {\n      const tensor = tf.tensor(variable.dataSync());\n      variable.dispose();\n      this.reassignParamFromPath(path, tensor);\n    });\n  }\n\n  public dispose(throwOnRedispose = true) {\n    this.getParamList().forEach((param) => {\n      if (throwOnRedispose && param.tensor.isDisposed) {\n        throw new Error(`param tensor has already been disposed for path ${param.path}`);\n      }\n      param.tensor.dispose();\n    });\n    this._params = undefined;\n  }\n\n  public serializeParams(): Float32Array {\n    return new Float32Array(\n      this.getParamList()\n        .map(({ tensor }) => Array.from(tensor.dataSync()) as number[])\n        .reduce((flat, arr) => flat.concat(arr)),\n    );\n  }\n\n  public async load(weightsOrUrl: Float32Array | string | undefined): Promise<void> {\n    if (weightsOrUrl instanceof Float32Array) {\n      this.extractWeights(weightsOrUrl);\n      return;\n    }\n    await this.loadFromUri(weightsOrUrl);\n  }\n\n  public async loadFromUri(uri: string | undefined) {\n    if (uri && typeof uri !== 'string') {\n      throw new Error(`${this._name}.loadFromUri - expected model uri`);\n    }\n    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());\n    this.loadFromWeightMap(weightMap);\n  }\n\n  public async loadFromDisk(filePath: string | undefined) {\n    if (filePath && typeof filePath !== 'string') {\n      throw new Error(`${this._name}.loadFromDisk - expected model file path`);\n    }\n    const { readFile } = env.getEnv();\n    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());\n    const fetchWeightsFromDisk = (filePaths: string[]) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));\n    const loadWeights = tf['io'].weightsLoaderFactory(fetchWeightsFromDisk);\n    const manifest = JSON.parse((await readFile(manifestUri)).toString());\n    const weightMap = await loadWeights(manifest, modelBaseUri);\n    this.loadFromWeightMap(weightMap);\n  }\n\n  public loadFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);\n    this._paramMappings = paramMappings;\n    this._params = params;\n  }\n\n  public extractWeights(weights: Float32Array) {\n    const { paramMappings, params } = this.extractParams(weights);\n    this._paramMappings = paramMappings;\n    this._params = params;\n  }\n\n  private traversePropertyPath(paramPath: string) {\n    if (!this.params) {\n      throw new Error('traversePropertyPath - model has no loaded params');\n    }\n\n    const result = paramPath.split('/').reduce((res: { nextObj: any, obj?: any, objProp?: string }, objProp) => {\n      // eslint-disable-next-line no-prototype-builtins\n      if (!res.nextObj.hasOwnProperty(objProp)) {\n        throw new Error(`traversePropertyPath - object does not have property ${objProp}, for path ${paramPath}`);\n      }\n      return { obj: res.nextObj, objProp, nextObj: res.nextObj[objProp] };\n    }, { nextObj: this.params });\n\n    const { obj, objProp } = result;\n    if (!obj || !objProp || !(obj[objProp] instanceof tf.Tensor)) {\n      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);\n    }\n\n    return { obj, objProp };\n  }\n\n  protected abstract getDefaultModelName(): string\n\n  // eslint-disable-next-line no-unused-vars\n  protected abstract extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TNetParams, paramMappings: ParamMapping[] }\n\n  // eslint-disable-next-line no-unused-vars\n  protected abstract extractParams(weights: Float32Array): { params: TNetParams, paramMappings: ParamMapping[] }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { SeparableConvParams } from './types';\n\nexport function depthwiseSeparableConv(\n  x: tf.Tensor4D,\n  params: SeparableConvParams,\n  stride: [number, number],\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, 'same');\n    out = tf.add(out, params.bias);\n    return out;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, SeparableConvParams } from '../common/index';\nimport { depthwiseSeparableConv } from '../common/depthwiseSeparableConv';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function denseBlock3(\n  x: tf.Tensor4D,\n  denseBlockParams: DenseBlock3Params,\n  isFirstLayer = false,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out1 = tf.relu(\n      isFirstLayer\n        ? tf.add(\n          tf.conv2d(x, (denseBlockParams.conv0 as ConvParams).filters, [2, 2], 'same'),\n          denseBlockParams.conv0.bias,\n        )\n        : depthwiseSeparableConv(x, denseBlockParams.conv0 as SeparableConvParams, [2, 2]),\n    ) as tf.Tensor4D;\n    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);\n\n    const in3 = tf.relu(tf.add(out1, out2)) as tf.Tensor4D;\n    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);\n\n    return tf.relu(tf.add(out1, tf.add(out2, out3))) as tf.Tensor4D;\n  });\n}\n\nexport function denseBlock4(\n  x: tf.Tensor4D,\n  denseBlockParams: DenseBlock4Params,\n  isFirstLayer = false,\n  isScaleDown = true,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out1 = tf.relu(\n      isFirstLayer\n        ? tf.add(\n          tf.conv2d(x, (denseBlockParams.conv0 as ConvParams).filters, isScaleDown ? [2, 2] : [1, 1], 'same'),\n          denseBlockParams.conv0.bias,\n        )\n        : depthwiseSeparableConv(x, denseBlockParams.conv0 as SeparableConvParams, isScaleDown ? [2, 2] : [1, 1]),\n    ) as tf.Tensor4D;\n    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);\n\n    const in3 = tf.relu(tf.add(out1, out2)) as tf.Tensor4D;\n    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);\n\n    const in4 = tf.relu(tf.add(out1, tf.add(out2, out3))) as tf.Tensor4D;\n    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);\n\n    return tf.relu(tf.add(out1, tf.add(out2, tf.add(out3, out4)))) as tf.Tensor4D;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from './types';\n\nexport function convLayer(\n  x: tf.Tensor4D,\n  params: ConvParams,\n  padding: 'valid' | 'same' = 'same',\n  withRelu = false,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out = tf.add(\n      tf.conv2d(x, params.filters, [1, 1], padding),\n      params.bias,\n    ) as tf.Tensor4D;\n\n    return withRelu ? tf.relu(out) : out;\n  });\n}\n", "import { ParamMapping } from './types';\n\nexport function disposeUnusedWeightTensors(weightMap: any, paramMappings: ParamMapping[]) {\n  Object.keys(weightMap).forEach((path) => {\n    if (!paramMappings.some((pm) => pm.originalPath === path)) {\n      weightMap[path].dispose();\n    }\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, ExtractWeightsFunction, ParamMapping } from './types';\n\nexport function extractConvParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvParams => {\n    const filters = tf.tensor4d(\n      extractWeights(channelsIn * channelsOut * filterSize * filterSize),\n      [filterSize, filterSize, channelsIn, channelsOut],\n    );\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return { filters, bias };\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, FCParams, ParamMapping } from './types';\n\nexport function extractFCParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (\n    channelsIn: number,\n    channelsOut: number,\n    mappedPrefix: string,\n  ): FCParams => {\n    const fc_weights = tf.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);\n    const fc_bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/weights` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return {\n      weights: fc_weights,\n      bias: fc_bias,\n    };\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\n// eslint-disable-next-line no-unused-vars\nexport type ExtractWeightsFunction = (numWeights: number) => Float32Array\n\nexport type ParamMapping = {\n  originalPath?: string\n  paramPath: string\n}\n\nexport type ConvParams = {\n  filters: tf.Tensor4D\n  bias: tf.Tensor1D\n}\n\nexport type FCParams = {\n  weights: tf.Tensor2D\n  bias: tf.Tensor1D\n}\n\nexport class SeparableConvParams {\n  // eslint-disable-next-line no-useless-constructor\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    public depthwise_filter: tf.Tensor4D,\n    // eslint-disable-next-line no-unused-vars\n    public pointwise_filter: tf.Tensor4D,\n    // eslint-disable-next-line no-unused-vars\n    public bias: tf.Tensor1D,\n  // eslint-disable-next-line no-empty-function\n  ) {}\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, ParamMapping, SeparableConvParams } from './types';\n\nexport function extractSeparableConvParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (channelsIn: number, channelsOut: number, mappedPrefix: string): SeparableConvParams => {\n    const depthwise_filter = tf.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);\n    const pointwise_filter = tf.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/depthwise_filter` },\n      { paramPath: `${mappedPrefix}/pointwise_filter` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return new SeparableConvParams(\n      depthwise_filter,\n      pointwise_filter,\n      bias,\n    );\n  };\n}\n\nexport function loadSeparableConvParamsFactory(\n  // eslint-disable-next-line no-unused-vars\n  extractWeightEntry: <T>(originalPath: string, paramRank: number) => T,\n) {\n  return (prefix: string): SeparableConvParams => {\n    const depthwise_filter = extractWeightEntry<tf.Tensor4D>(`${prefix}/depthwise_filter`, 4);\n    const pointwise_filter = extractWeightEntry<tf.Tensor4D>(`${prefix}/pointwise_filter`, 4);\n    const bias = extractWeightEntry<tf.Tensor1D>(`${prefix}/bias`, 1);\n\n    return new SeparableConvParams(\n      depthwise_filter,\n      pointwise_filter,\n      bias,\n    );\n  };\n}\n", "import { isTensor } from '../utils/index';\nimport { ParamMapping } from './types';\n\nexport function extractWeightEntryFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  return (originalPath: string, paramRank: number, mappedPath?: string) => {\n    const tensor = weightMap[originalPath];\n\n    if (!isTensor(tensor, paramRank)) {\n      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor}`);\n    }\n\n    paramMappings.push(\n      { originalPath, paramPath: mappedPath || originalPath },\n    );\n\n    return tensor;\n  };\n}\n", "export function extractWeightsFactory(weights: Float32Array) {\n  let remainingWeights = weights;\n\n  function extractWeights(numWeights: number): Float32Array {\n    const ret = remainingWeights.slice(0, numWeights);\n    remainingWeights = remainingWeights.slice(numWeights);\n    return ret;\n  }\n\n  function getRemainingWeights(): Float32Array {\n    return remainingWeights;\n  }\n\n  return {\n    extractWeights,\n    getRemainingWeights,\n  };\n}\n", "import { extractConvParamsFactory, extractSeparableConvParamsFactory, ExtractWeightsFunction, ParamMapping } from '../common/index';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  function extractDenseBlock3Params(channelsIn: number, channelsOut: number, mappedPrefix: string, isFirstLayer = false): DenseBlock3Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`)\n      : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);\n    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);\n    const conv2 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);\n\n    return { conv0, conv1, conv2 };\n  }\n\n  function extractDenseBlock4Params(channelsIn: number, channelsOut: number, mappedPrefix: string, isFirstLayer = false): DenseBlock4Params {\n    const { conv0, conv1, conv2 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);\n    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);\n\n    return {\n      conv0, conv1, conv2, conv3,\n    };\n  }\n\n  return {\n    extractDenseBlock3Params,\n    extractDenseBlock4Params,\n  };\n}\n", "import { extractWeightsFactory, ParamMapping } from '../common/index';\nimport { extractorsFactory } from './extractorsFactory';\nimport { FaceFeatureExtractorParams } from './types';\n\nexport function extractParams(weights: Float32Array): { params: FaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractDenseBlock4Params,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const dense0 = extractDenseBlock4Params(3, 32, 'dense0', true);\n  const dense1 = extractDenseBlock4Params(32, 64, 'dense1');\n  const dense2 = extractDenseBlock4Params(64, 128, 'dense2');\n  const dense3 = extractDenseBlock4Params(128, 256, 'dense3');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: {\n      dense0, dense1, dense2, dense3,\n    },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from './types';\n\n// eslint-disable-next-line no-unused-vars\nexport function loadConvParamsFactory(extractWeightEntry: <T>(originalPath: string, paramRank: number) => T) {\n  return (prefix: string): ConvParams => {\n    const filters = extractWeightEntry<tf.Tensor4D>(`${prefix}/filters`, 4);\n    const bias = extractWeightEntry<tf.Tensor1D>(`${prefix}/bias`, 1);\n\n    return { filters, bias };\n  };\n}\n", "import { extractWeightEntryFactory, loadSeparableConvParamsFactory, ParamMapping } from '../common/index';\nimport { loadConvParamsFactory } from '../common/loadConvParamsFactory';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function loadParamsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  const extractConvParams = loadConvParamsFactory(extractWeightEntry);\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n\n  function extractDenseBlock3Params(prefix: string, isFirstLayer = false): DenseBlock3Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(`${prefix}/conv0`)\n      : extractSeparableConvParams(`${prefix}/conv0`);\n    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);\n    const conv2 = extractSeparableConvParams(`${prefix}/conv2`);\n\n    return { conv0, conv1, conv2 };\n  }\n\n  function extractDenseBlock4Params(prefix: string, isFirstLayer = false): DenseBlock4Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(`${prefix}/conv0`)\n      : extractSeparableConvParams(`${prefix}/conv0`);\n    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);\n    const conv2 = extractSeparableConvParams(`${prefix}/conv2`);\n    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);\n\n    return {\n      conv0, conv1, conv2, conv3,\n    };\n  }\n\n  return {\n    extractDenseBlock3Params,\n    extractDenseBlock4Params,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, ParamMapping } from '../common/index';\nimport { loadParamsFactory } from './loadParamsFactory';\nimport { FaceFeatureExtractorParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: FaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractDenseBlock4Params,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const params = {\n    dense0: extractDenseBlock4Params('dense0', true),\n    dense1: extractDenseBlock4Params('dense1'),\n    dense2: extractDenseBlock4Params('dense2'),\n    dense3: extractDenseBlock4Params('dense3'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { denseBlock4 } from './denseBlock';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { FaceFeatureExtractorParams, IFaceFeatureExtractor } from './types';\n\nexport class FaceFeatureExtractor extends NeuralNetwork<FaceFeatureExtractorParams> implements IFaceFeatureExtractor<FaceFeatureExtractorParams> {\n  constructor() {\n    super('FaceFeatureExtractor');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('FaceFeatureExtractor - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = denseBlock4(normalized, params.dense0, true);\n      out = denseBlock4(out, params.dense1);\n      out = denseBlock4(out, params.dense2);\n      out = denseBlock4(out, params.dense3);\n      out = tf.avgPool(out, [7, 7], [2, 2], 'valid');\n\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_feature_extractor_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FCParams } from './types';\n\nexport function fullyConnectedLayer(\n  x: tf.Tensor2D,\n  params: FCParams,\n): tf.Tensor2D {\n  return tf.tidy(() => tf.add(\n    tf.matMul(x, params.weights),\n    params.bias,\n  ));\n}\n", "import { extractFCParamsFactory, extractWeightsFactory, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParams(weights: Float32Array, channelsIn: number, channelsOut: number): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);\n\n  const fc = extractFCParams(channelsIn, channelsOut, 'fc');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { fc },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, FCParams, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractFcParams(prefix: string): FCParams {\n    const weights = extractWeightEntry(`${prefix}/weights`, 2);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { weights, bias };\n  }\n\n  const params = {\n    fc: extractFcParams('fc'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function seperateWeightMaps(weightMap: tf.NamedTensorMap) {\n  const featureExtractorMap: tf.NamedTensorMap = {};\n  const classifierMap: tf.NamedTensorMap = {};\n\n  Object.keys(weightMap).forEach((key) => {\n    const map = key.startsWith('fc') ? classifierMap : featureExtractorMap;\n    map[key] = weightMap[key];\n  });\n\n  return { featureExtractorMap, classifierMap };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { fullyConnectedLayer } from '../common/fullyConnectedLayer';\nimport { NetInput } from '../dom/index';\nimport { FaceFeatureExtractorParams, IFaceFeatureExtractor, TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { NetParams } from './types';\nimport { seperateWeightMaps } from './util';\n\nexport abstract class FaceProcessor<\n  TExtractorParams extends FaceFeatureExtractorParams | TinyFaceFeatureExtractorParams\n>\n  extends NeuralNetwork<NetParams> {\n  protected _faceFeatureExtractor: IFaceFeatureExtractor<TExtractorParams>;\n\n  constructor(_name: string, faceFeatureExtractor: IFaceFeatureExtractor<TExtractorParams>) {\n    super(_name);\n    this._faceFeatureExtractor = faceFeatureExtractor;\n  }\n\n  public get faceFeatureExtractor(): IFaceFeatureExtractor<TExtractorParams> {\n    return this._faceFeatureExtractor;\n  }\n\n  protected abstract override getDefaultModelName(): string\n\n  protected abstract getClassifierChannelsIn(): number\n\n  protected abstract getClassifierChannelsOut(): number\n\n  public runNet(input: NetInput | tf.Tensor4D): tf.Tensor2D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error(`${this._name} - load model before inference`);\n    }\n\n    return tf.tidy(() => {\n      const bottleneckFeatures = input instanceof NetInput\n        ? this.faceFeatureExtractor.forwardInput(input)\n        : input;\n      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);\n    });\n  }\n\n  public override dispose(throwOnRedispose = true) {\n    this.faceFeatureExtractor.dispose(throwOnRedispose);\n    super.dispose(throwOnRedispose);\n  }\n\n  public loadClassifierParams(weights: Float32Array) {\n    const { params, paramMappings } = this.extractClassifierParams(weights);\n    this._params = params;\n    this._paramMappings = paramMappings;\n  }\n\n  public extractClassifierParams(weights: Float32Array) {\n    return extractParams(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);\n\n    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);\n\n    return extractParamsFromWeightMap(classifierMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const cIn = this.getClassifierChannelsIn();\n    const cOut = this.getClassifierChannelsOut();\n    const classifierWeightSize = (cOut * cIn) + cOut;\n\n    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);\n    const classifierWeights = weights.slice(weights.length - classifierWeightSize);\n\n    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);\n    return this.extractClassifierParams(classifierWeights);\n  }\n}\n", "export const FACE_EXPRESSION_LABELS = ['neutral', 'happy', 'sad', 'angry', 'fearful', 'disgusted', 'surprised'];\n\nexport class FaceExpressions {\n  public neutral = 0;\n  public happy = 0;\n  public sad = 0;\n  public angry = 0;\n  public fearful = 0;\n  public disgusted = 0;\n  public surprised = 0;\n\n  constructor(probabilities: number[] | Float32Array) {\n    if (probabilities.length !== 7) {\n      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);\n    }\n\n    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {\n      this[expression] = probabilities[idx];\n    });\n  }\n\n  asSortedArray() {\n    return FACE_EXPRESSION_LABELS\n      .map((expression) => ({ expression, probability: this[expression] as number }))\n      .sort((e0, e1) => e1.probability - e0.probability);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { FaceFeatureExtractor } from '../faceFeatureExtractor/FaceFeatureExtractor';\nimport { FaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceProcessor } from '../faceProcessor/FaceProcessor';\nimport { FaceExpressions } from './FaceExpressions';\n\nexport class FaceExpressionNet extends FaceProcessor<FaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: FaceFeatureExtractor = new FaceFeatureExtractor()) {\n    super('FaceExpressionNet', faceFeatureExtractor);\n  }\n\n  public forwardInput(input: NetInput | tf.Tensor4D): tf.Tensor2D {\n    return tf.tidy(() => tf.softmax(this.runNet(input)));\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async predictExpressions(input: TNetInput) {\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput);\n    const probabilitesByBatch = await Promise.all(tf.unstack(out).map(async (t) => {\n      const data = t.dataSync();\n      t.dispose();\n      return data;\n    }));\n    out.dispose();\n\n    const predictionsByBatch = probabilitesByBatch\n      .map((probabilites) => new FaceExpressions(probabilites as Float32Array));\n\n    return netInput.isBatchInput\n      ? predictionsByBatch\n      : predictionsByBatch[0];\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_expression_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 256;\n  }\n\n  protected getClassifierChannelsOut(): number {\n    return 7;\n  }\n}\n", "import { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\n\nexport type WithFaceExpressions<TSource> = TSource & { expressions: FaceExpressions }\n\nexport function isWithFaceExpressions(obj: any): obj is WithFaceExpressions<{}> {\n  return obj.expressions instanceof FaceExpressions;\n}\n\nexport function extendWithFaceExpressions<TSource>(sourceObj: TSource, expressions: FaceExpressions): WithFaceExpressions<TSource> {\n  const extension = { expressions };\n  return { ...sourceObj, ...extension };\n}\n", "import { IPoint, Point } from '../classes/index';\nimport { FaceExpressions } from '../faceExpressionNet/index';\nimport { isWithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceExpressions, WithFaceExpressions } from '../factories/WithFaceExpressions';\nimport { round } from '../utils/index';\nimport { DrawTextField } from './DrawTextField';\n\nexport type DrawFaceExpressionsInput = FaceExpressions | WithFaceExpressions<{}>\n\nexport function drawFaceExpressions(canvasArg: string | HTMLCanvasElement, faceExpressions: DrawFaceExpressionsInput | Array<DrawFaceExpressionsInput>, minConfidence = 0.1, textFieldAnchor?: IPoint) {\n  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];\n\n  faceExpressionsArray.forEach((e) => {\n    // eslint-disable-next-line no-nested-ternary\n    const expr = e instanceof FaceExpressions\n      ? e\n      : (isWithFaceExpressions(e) ? e.expressions : undefined);\n    if (!expr) {\n      throw new Error('drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof');\n    }\n\n    const sorted = expr.asSortedArray();\n    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);\n\n    const anchor = isWithFaceDetection(e)\n      ? e.detection.box.bottomLeft\n      : (textFieldAnchor || new Point(0, 0));\n\n    const drawTextField = new DrawTextField(\n      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),\n      anchor,\n    );\n    drawTextField.draw(canvasArg);\n  });\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\nimport { FaceLandmarks } from '../classes/FaceLandmarks';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { isWithFaceDetection, WithFaceDetection } from './WithFaceDetection';\n\nexport type WithFaceLandmarks<\n  TSource extends WithFaceDetection<{}>,\n  TFaceLandmarks extends FaceLandmarks = FaceLandmarks68\n> = TSource & {\n  landmarks: TFaceLandmarks;\n  unshiftedLandmarks: TFaceLandmarks;\n  alignedRect: FaceDetection;\n  angle: {\n    roll: number | undefined;\n    pitch: number | undefined;\n    yaw: number | undefined;\n  };\n};\n\nexport function isWithFaceLandmarks(\n  obj: any,\n): obj is WithFaceLandmarks<WithFaceDetection<{}>, FaceLandmarks> {\n  return (\n    isWithFaceDetection(obj)\n    // eslint-disable-next-line dot-notation\n    && obj['landmarks'] instanceof FaceLandmarks\n    // eslint-disable-next-line dot-notation\n    && obj['unshiftedLandmarks'] instanceof FaceLandmarks\n    // eslint-disable-next-line dot-notation\n    && obj['alignedRect'] instanceof FaceDetection\n  );\n}\n\nfunction calculateFaceAngle(mesh) {\n  // Helper to convert radians to degrees\n  // eslint-disable-next-line no-unused-vars, @typescript-eslint/no-unused-vars\n  const degrees = (radians) => (radians * 180) / Math.PI;\n  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);\n\n  const angle = {\n    roll: <number | undefined>undefined,\n    pitch: <number | undefined>undefined,\n    yaw: <number | undefined>undefined,\n  };\n\n  const calcYaw = (leftPoint, midPoint, rightPoint) => {\n    // Calc x-distance from left side of the face (\"ear\") to facial midpoint (\"nose\")\n    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);\n    // Calc x-distance from facial midpoint (\"nose\") to the right side of the face (\"ear\")\n    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);\n    // Difference in distances coincidentally approximates to angles\n    return leftToMidpoint - rightToMidpoint;\n  };\n\n  const calcRoll = (lever, pivot) => {\n    // When rolling, the head seems to pivot from the nose/lips/chin area.\n    // So, we'll choose any two points from the facial midline, where the first point should be the pivot, and the other \"lever\"\n    // Plan/Execution: get the hypotenuse & opposite sides of a 90deg triangle ==> Calculate angle in radians\n    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);\n    const opposite = pivot._y - lever._y;\n    const angleInRadians = Math.asin(opposite / hypotenuse);\n    const angleInDegrees = degrees(angleInRadians);\n    const normalizeAngle = Math.floor(90 - angleInDegrees);\n    // If lever more to the left of the pivot, then we're tilting left\n    // \"-\" is negative direction. \"+\", or absence of a sign is positive direction\n    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;\n    const result = normalizeAngle * tiltDirection;\n    return result;\n  };\n\n  const calcPitch = (leftPoint, midPoint, rightPoint) => {\n    // Theory: While pitching, the nose is the most salient point --> That's what we'll use to make a trianle.\n    // The \"base\" is between point that don't move when we pitch our head (i.e. an imaginary line running ear to ear through the nose).\n    // Executuin: Get the opposite & adjacent lengths of the triangle from the ear's perspective. Use it to get angle.\n\n    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);\n    // adjecent is base/2 technically.\n    const baseCoords = {\n      _x: (leftPoint._x + rightPoint._x) / 2,\n      _y: (leftPoint._y + rightPoint._y) / 2,\n    };\n    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);\n    const angleInRadians = Math.atan(midToBaseLength / base);\n    const angleInDegrees = Math.floor(degrees(angleInRadians));\n    // Account for directionality.\n    // pitch forwards (_i.e. tilting your head forwards) is positive (or no sign); backward is negative.\n    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;\n    const result = angleInDegrees * direction;\n    return result;\n  };\n\n  if (!mesh || !mesh._positions || mesh._positions.length !== 68) return angle;\n  const pt = mesh._positions;\n  angle.roll = calcRoll(pt[27], pt[66]);\n  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);\n  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);\n  return angle;\n}\n\nexport function extendWithFaceLandmarks<TSource extends WithFaceDetection<{}>, TFaceLandmarks extends FaceLandmarks = FaceLandmarks68>(\n  sourceObj: TSource,\n  unshiftedLandmarks: TFaceLandmarks,\n): WithFaceLandmarks<TSource, TFaceLandmarks> {\n  const { box: shift } = sourceObj.detection;\n  const landmarks = unshiftedLandmarks.shiftBy<TFaceLandmarks>(shift.x, shift.y);\n  const rect = landmarks.align();\n  const { imageDims } = sourceObj.detection;\n  const alignedRect = new FaceDetection(\n    sourceObj.detection.score,\n    rect.rescale(imageDims.reverse()),\n    imageDims,\n  );\n  const angle = calculateFaceAngle(unshiftedLandmarks);\n  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };\n  return { ...sourceObj, ...extension };\n}\n", "/* eslint-disable max-classes-per-file */\nimport { IPoint } from '../classes/index';\nimport { FaceLandmarks } from '../classes/FaceLandmarks';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { drawContour } from './drawContour';\n\nexport interface IDrawFaceLandmarksOptions {\n  drawLines?: boolean\n  drawPoints?: boolean\n  lineWidth?: number\n  pointSize?: number\n  lineColor?: string\n  pointColor?: string\n}\n\nexport class DrawFaceLandmarksOptions {\n  public drawLines: boolean;\n\n  public drawPoints: boolean;\n\n  public lineWidth: number;\n\n  public pointSize: number;\n\n  public lineColor: string;\n\n  public pointColor: string;\n\n  constructor(options: IDrawFaceLandmarksOptions = {}) {\n    const {\n      drawLines = true, drawPoints = true, lineWidth, lineColor, pointSize, pointColor,\n    } = options;\n    this.drawLines = drawLines;\n    this.drawPoints = drawPoints;\n    this.lineWidth = lineWidth || 1;\n    this.pointSize = pointSize || 2;\n    this.lineColor = lineColor || 'rgba(0, 255, 255, 1)';\n    this.pointColor = pointColor || 'rgba(255, 0, 255, 1)';\n  }\n}\n\nexport class DrawFaceLandmarks {\n  public faceLandmarks: FaceLandmarks;\n\n  public options: DrawFaceLandmarksOptions;\n\n  constructor(\n    faceLandmarks: FaceLandmarks,\n    options: IDrawFaceLandmarksOptions = {},\n  ) {\n    this.faceLandmarks = faceLandmarks;\n    this.options = new DrawFaceLandmarksOptions(options);\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const ctx = getContext2dOrThrow(canvasArg);\n\n    const {\n      drawLines, drawPoints, lineWidth, lineColor, pointSize, pointColor,\n    } = this.options;\n\n    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {\n      ctx.strokeStyle = lineColor;\n      ctx.lineWidth = lineWidth;\n      drawContour(ctx, this.faceLandmarks.getJawOutline());\n      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());\n      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());\n      drawContour(ctx, this.faceLandmarks.getNose());\n      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);\n      drawContour(ctx, this.faceLandmarks.getRightEye(), true);\n      drawContour(ctx, this.faceLandmarks.getMouth(), true);\n    }\n\n    if (drawPoints) {\n      ctx.strokeStyle = pointColor;\n      ctx.fillStyle = pointColor;\n\n      const drawPoint = (pt: IPoint) => {\n        ctx.beginPath();\n        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);\n        ctx.fill();\n      };\n      this.faceLandmarks.positions.forEach(drawPoint);\n    }\n  }\n}\n\nexport type DrawFaceLandmarksInput = FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>>\n\nexport function drawFaceLandmarks(\n  canvasArg: string | HTMLCanvasElement,\n  faceLandmarks: DrawFaceLandmarksInput | Array<DrawFaceLandmarksInput>,\n) {\n  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];\n  faceLandmarksArray.forEach((f) => {\n    // eslint-disable-next-line no-nested-ternary\n    const landmarks = f instanceof FaceLandmarks\n      ? f\n      : (isWithFaceLandmarks(f) ? f.landmarks : undefined);\n    if (!landmarks) {\n      throw new Error('drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof');\n    }\n\n    new DrawFaceLandmarks(landmarks).draw(canvasArg);\n  });\n}\n", "import { extractConvParamsFactory, extractSeparableConvParamsFactory, extractWeightsFactory } from '../common/index';\nimport { ExtractWeightsFunction, ParamMapping } from '../common/types';\nimport { range } from '../utils/index';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  function extractReductionBlockParams(channelsIn: number, channelsOut: number, mappedPrefix: string): ReductionBlockParams {\n    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);\n    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);\n\n    return { separable_conv0, separable_conv1, expansion_conv };\n  }\n\n  function extractMainBlockParams(channels: number, mappedPrefix: string): MainBlockParams {\n    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);\n    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);\n\n    return { separable_conv0, separable_conv1, separable_conv2 };\n  }\n\n  return {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array, numMainBlocks: number): { params: TinyXceptionParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const entry_flow_conv_in = extractConvParams(3, 32, 3, 'entry_flow/conv_in');\n  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, 'entry_flow/reduction_block_0');\n  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, 'entry_flow/reduction_block_1');\n\n  const entry_flow = {\n    conv_in: entry_flow_conv_in,\n    reduction_block_0: entry_flow_reduction_block_0,\n    reduction_block_1: entry_flow_reduction_block_1,\n  };\n\n  const middle_flow = {};\n  range(numMainBlocks, 0, 1).forEach((idx) => {\n    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);\n  });\n\n  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, 'exit_flow/reduction_block');\n  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, 'exit_flow/separable_conv');\n\n  const exit_flow = {\n    reduction_block: exit_flow_reduction_block,\n    separable_conv: exit_flow_separable_conv,\n  };\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { entry_flow, middle_flow, exit_flow },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, loadSeparableConvParamsFactory, ParamMapping } from '../common/index';\nimport { loadConvParamsFactory } from '../common/loadConvParamsFactory';\nimport { range } from '../utils/index';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction loadParamsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  const extractConvParams = loadConvParamsFactory(extractWeightEntry);\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n\n  function extractReductionBlockParams(mappedPrefix: string): ReductionBlockParams {\n    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);\n    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);\n\n    return { separable_conv0, separable_conv1, expansion_conv };\n  }\n\n  function extractMainBlockParams(mappedPrefix: string): MainBlockParams {\n    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);\n    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);\n\n    return { separable_conv0, separable_conv1, separable_conv2 };\n  }\n\n  return {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n  numMainBlocks: number,\n): { params: TinyXceptionParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const entry_flow_conv_in = extractConvParams('entry_flow/conv_in');\n  const entry_flow_reduction_block_0 = extractReductionBlockParams('entry_flow/reduction_block_0');\n  const entry_flow_reduction_block_1 = extractReductionBlockParams('entry_flow/reduction_block_1');\n\n  const entry_flow = {\n    conv_in: entry_flow_conv_in,\n    reduction_block_0: entry_flow_reduction_block_0,\n    reduction_block_1: entry_flow_reduction_block_1,\n  };\n\n  const middle_flow = {};\n  range(numMainBlocks, 0, 1).forEach((idx) => {\n    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);\n  });\n\n  const exit_flow_reduction_block = extractReductionBlockParams('exit_flow/reduction_block');\n  const exit_flow_separable_conv = extractSeparableConvParams('exit_flow/separable_conv');\n\n  const exit_flow = {\n    reduction_block: exit_flow_reduction_block,\n    separable_conv: exit_flow_separable_conv,\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, depthwiseSeparableConv } from '../common/index';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { range } from '../utils/index';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction conv(x: tf.Tensor4D, params: ConvParams, stride: [number, number]): tf.Tensor4D {\n  return tf.add(tf.conv2d(x, params.filters, stride, 'same'), params.bias);\n}\n\nfunction reductionBlock(x: tf.Tensor4D, params: ReductionBlockParams, isActivateInput = true): tf.Tensor4D {\n  let out = isActivateInput ? tf.relu(x) : x;\n  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv1, [1, 1]);\n  out = tf.maxPool(out, [3, 3], [2, 2], 'same');\n  out = tf.add(out, conv(x, params.expansion_conv, [2, 2]));\n  return out;\n}\n\nfunction mainBlock(x: tf.Tensor4D, params: MainBlockParams): tf.Tensor4D {\n  let out = depthwiseSeparableConv(tf.relu(x), params.separable_conv0, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv1, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv2, [1, 1]);\n  out = tf.add(out, x);\n  return out;\n}\n\nexport class TinyXception extends NeuralNetwork<TinyXceptionParams> {\n  private _numMainBlocks: number;\n\n  constructor(numMainBlocks: number) {\n    super('TinyXception');\n    this._numMainBlocks = numMainBlocks;\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n    if (!params) {\n      throw new Error('TinyXception - load model before inference');\n    }\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n      let out = tf.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));\n      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);\n      out = reductionBlock(out, params.entry_flow.reduction_block_1);\n      range(this._numMainBlocks, 0, 1).forEach((idx) => {\n        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);\n      });\n      out = reductionBlock(out, params.exit_flow.reduction_block);\n      out = tf.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'tiny_xception_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap, this._numMainBlocks);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights, this._numMainBlocks);\n  }\n}\n", "import { extractFCParamsFactory, extractWeightsFactory, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);\n\n  const age = extractFCParams(512, 1, 'fc/age');\n  const gender = extractFCParams(512, 2, 'fc/gender');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { fc: { age, gender } },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, FCParams, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractFcParams(prefix: string): FCParams {\n    const weights = extractWeightEntry(`${prefix}/weights`, 2);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { weights, bias };\n  }\n\n  const params = {\n    fc: {\n      age: extractFcParams('fc/age'),\n      gender: extractFcParams('fc/gender'),\n    },\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FCParams } from '../common/index';\n\n// eslint-disable-next-line no-shadow\nexport enum Gender {\n  // eslint-disable-next-line no-unused-vars\n  FEMALE = 'female',\n  // eslint-disable-next-line no-unused-vars\n  MALE = 'male'\n}\n\nexport type AgeAndGenderPrediction = {\n  age: number\n  gender: Gender\n  genderProbability: number\n}\n\nexport type NetOutput = { age: tf.Tensor1D, gender: tf.Tensor2D }\n\nexport type NetParams = {\n  fc: {\n    age: FCParams\n    gender: FCParams\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm.js';\nimport { fullyConnectedLayer } from '../common/fullyConnectedLayer';\nimport { seperateWeightMaps } from '../faceProcessor/util';\nimport { TinyXception } from '../xception/TinyXception';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { AgeAndGenderPrediction, Gender, NetOutput, NetParams } from './types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\n\nexport class AgeGenderNet extends NeuralNetwork<NetParams> {\n  private _faceFeatureExtractor: TinyXception;\n\n  constructor(faceFeatureExtractor: TinyXception = new TinyXception(2)) {\n    super('AgeGenderNet');\n    this._faceFeatureExtractor = faceFeatureExtractor;\n  }\n\n  public get faceFeatureExtractor(): TinyXception {\n    return this._faceFeatureExtractor;\n  }\n\n  public runNet(input: NetInput | tf.Tensor4D): NetOutput {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error(`${this._name} - load model before inference`);\n    }\n\n    return tf.tidy(() => {\n      const bottleneckFeatures = input instanceof NetInput\n        ? this.faceFeatureExtractor.forwardInput(input)\n        : input;\n\n      const pooled = tf.avgPool(bottleneckFeatures, [7, 7], [2, 2], 'valid').as2D(bottleneckFeatures.shape[0], -1);\n      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();\n      const gender = fullyConnectedLayer(pooled, params.fc.gender);\n      return { age, gender };\n    });\n  }\n\n  public forwardInput(input: NetInput | tf.Tensor4D): NetOutput {\n    return tf.tidy(() => {\n      const { age, gender } = this.runNet(input);\n      return { age, gender: tf.softmax(gender) };\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<NetOutput> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async predictAgeAndGender(input: TNetInput): Promise<AgeAndGenderPrediction | AgeAndGenderPrediction[]> {\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput);\n\n    const ages = tf.unstack(out.age);\n    const genders = tf.unstack(out.gender);\n    const ageAndGenderTensors = ages.map((ageTensor, i) => ({\n      ageTensor,\n      genderTensor: genders[i],\n    }));\n\n    const predictionsByBatch = await Promise.all(\n      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {\n        const age = (ageTensor.dataSync())[0];\n        const probMale = (genderTensor.dataSync())[0];\n        const isMale = probMale > 0.5;\n        const gender = isMale ? Gender.MALE : Gender.FEMALE;\n        const genderProbability = isMale ? probMale : (1 - probMale);\n\n        ageTensor.dispose();\n        genderTensor.dispose();\n        return { age, gender, genderProbability };\n      }),\n    );\n    out.age.dispose();\n    out.gender.dispose();\n\n    return netInput.isBatchInput ? predictionsByBatch as AgeAndGenderPrediction[] : predictionsByBatch[0] as AgeAndGenderPrediction;\n  }\n\n  protected getDefaultModelName(): string {\n    return 'age_gender_model';\n  }\n\n  public override dispose(throwOnRedispose = true) {\n    this.faceFeatureExtractor.dispose(throwOnRedispose);\n    super.dispose(throwOnRedispose);\n  }\n\n  public loadClassifierParams(weights: Float32Array) {\n    const { params, paramMappings } = this.extractClassifierParams(weights);\n    this._params = params;\n    this._paramMappings = paramMappings;\n  }\n\n  public extractClassifierParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);\n\n    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);\n\n    return extractParamsFromWeightMap(classifierMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const classifierWeightSize = (512 * 1 + 1) + (512 * 2 + 2);\n\n    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);\n    const classifierWeights = weights.slice(weights.length - classifierWeightSize);\n\n    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);\n    return this.extractClassifierParams(classifierWeights);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { IDimensions, Point } from '../classes/index';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { FaceFeatureExtractorParams, TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceProcessor } from '../faceProcessor/FaceProcessor';\nimport { isEven } from '../utils/index';\n\nexport abstract class FaceLandmark68NetBase<\n  TExtractorParams extends FaceFeatureExtractorParams | TinyFaceFeatureExtractorParams\n>\n  extends FaceProcessor<TExtractorParams> {\n  public postProcess(output: tf.Tensor2D, inputSize: number, originalDimensions: IDimensions[]): tf.Tensor2D {\n    const inputDimensions = originalDimensions.map(({ width, height }) => {\n      const scale = inputSize / Math.max(height, width);\n      return {\n        width: width * scale,\n        height: height * scale,\n      };\n    });\n\n    const batchSize = inputDimensions.length;\n\n    return tf.tidy(() => {\n      const createInterleavedTensor = (fillX: number, fillY: number) => tf.stack([tf.fill([68], fillX, 'float32'), tf.fill([68], fillY, 'float32')], 1).as2D(1, 136).as1D();\n\n      // eslint-disable-next-line no-unused-vars\n      const getPadding = (batchIdx: number, cond: (w: number, h: number) => boolean): number => {\n        const { width, height } = inputDimensions[batchIdx];\n        return cond(width, height) ? Math.abs(width - height) / 2 : 0;\n      };\n\n      const getPaddingX = (batchIdx: number) => getPadding(batchIdx, (w, h) => w < h);\n      const getPaddingY = (batchIdx: number) => getPadding(batchIdx, (w, h) => h < w);\n\n      const landmarkTensors = output\n        .mul(tf.fill([batchSize, 136], inputSize, 'float32'))\n        .sub(tf.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(\n          getPaddingX(batchIdx),\n          getPaddingY(batchIdx),\n        ))))\n        .div(tf.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(\n          inputDimensions[batchIdx].width,\n          inputDimensions[batchIdx].height,\n        ))));\n\n      return landmarkTensors as tf.Tensor2D;\n    });\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor2D {\n    return tf.tidy(() => {\n      const out = this.runNet(input);\n      return this.postProcess(\n        out,\n        input.inputSize as number,\n        input.inputDimensions.map(([height, width]) => ({ height, width })),\n      );\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async detectLandmarks(input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> {\n    const netInput = await toNetInput(input);\n    const landmarkTensors = tf.tidy(\n      () => tf.unstack(this.forwardInput(netInput)),\n    );\n\n    const landmarksForBatch = await Promise.all(landmarkTensors.map(\n      async (landmarkTensor, batchIdx) => {\n        const landmarksArray = Array.from(landmarkTensor.dataSync());\n        const xCoords = landmarksArray.filter((_, i) => isEven(i));\n        const yCoords = landmarksArray.filter((_, i) => !isEven(i));\n\n        return new FaceLandmarks68(\n          Array(68).fill(0).map((_, i) => new Point(xCoords[i] as number, yCoords[i] as number)),\n          {\n            height: netInput.getInputHeight(batchIdx),\n            width: netInput.getInputWidth(batchIdx),\n          },\n        );\n      },\n    ));\n\n    landmarkTensors.forEach((t) => t.dispose());\n\n    return netInput.isBatchInput ? landmarksForBatch as FaceLandmarks68[] : landmarksForBatch[0] as FaceLandmarks68;\n  }\n\n  protected getClassifierChannelsOut(): number {\n    return 136;\n  }\n}\n", "import { FaceFeatureExtractor } from '../faceFeatureExtractor/FaceFeatureExtractor';\nimport { FaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceLandmark68NetBase } from './FaceLandmark68NetBase';\n\nexport class FaceLandmark68Net extends FaceLandmark68NetBase<FaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: FaceFeatureExtractor = new FaceFeatureExtractor()) {\n    super('FaceLandmark68Net', faceFeatureExtractor);\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_landmark_68_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 256;\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, ParamMapping } from '../common/index';\nimport { loadParamsFactory } from './loadParamsFactory';\nimport { TinyFaceFeatureExtractorParams } from './types';\n\nexport function extractParamsFromWeightMapTiny(\n  weightMap: tf.NamedTensorMap,\n): { params: TinyFaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractDenseBlock3Params,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const params = {\n    dense0: extractDenseBlock3Params('dense0', true),\n    dense1: extractDenseBlock3Params('dense1'),\n    dense2: extractDenseBlock3Params('dense2'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import { extractWeightsFactory, ParamMapping } from '../common/index';\nimport { extractorsFactory } from './extractorsFactory';\nimport { TinyFaceFeatureExtractorParams } from './types';\n\nexport function extractParamsTiny(weights: Float32Array): { params: TinyFaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractDenseBlock3Params,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const dense0 = extractDenseBlock3Params(3, 32, 'dense0', true);\n  const dense1 = extractDenseBlock3Params(32, 64, 'dense1');\n  const dense2 = extractDenseBlock3Params(64, 128, 'dense2');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { dense0, dense1, dense2 },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { denseBlock3 } from './denseBlock';\nimport { extractParamsFromWeightMapTiny } from './extractParamsFromWeightMapTiny';\nimport { extractParamsTiny } from './extractParamsTiny';\nimport { IFaceFeatureExtractor, TinyFaceFeatureExtractorParams } from './types';\n\nexport class TinyFaceFeatureExtractor extends NeuralNetwork<TinyFaceFeatureExtractorParams> implements IFaceFeatureExtractor<TinyFaceFeatureExtractorParams> {\n  constructor() {\n    super('TinyFaceFeatureExtractor');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('TinyFaceFeatureExtractor - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = denseBlock3(normalized, params.dense0, true);\n      out = denseBlock3(out, params.dense1);\n      out = denseBlock3(out, params.dense2);\n      out = tf.avgPool(out, [14, 14], [2, 2], 'valid');\n\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_feature_extractor_tiny_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMapTiny(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParamsTiny(weights);\n  }\n}\n", "import { TinyFaceFeatureExtractor } from '../faceFeatureExtractor/TinyFaceFeatureExtractor';\nimport { TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceLandmark68NetBase } from './FaceLandmark68NetBase';\n\nexport class FaceLandmark68TinyNet extends FaceLandmark68NetBase<TinyFaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: TinyFaceFeatureExtractor = new TinyFaceFeatureExtractor()) {\n    super('FaceLandmark68TinyNet', faceFeatureExtractor);\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_landmark_68_tiny_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 128;\n  }\n}\n", "import { FaceLandmark68Net } from './FaceLandmark68Net';\n\nexport * from './FaceLandmark68Net';\nexport * from './FaceLandmark68TinyNet';\nexport class FaceLandmarkNet extends FaceLandmark68Net {}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ScaleLayerParams } from './types';\n\nexport function scale(x: tf.Tensor4D, params: ScaleLayerParams): tf.Tensor4D {\n  return tf.add(tf.mul(x, params.weights), params.biases);\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { scale } from './scaleLayer';\nimport { ConvLayerParams } from './types';\n\nfunction convLayer(\n  x: tf.Tensor4D,\n  params: ConvLayerParams,\n  strides: [number, number],\n  withRelu: boolean,\n  padding: 'valid' | 'same' = 'same',\n): tf.Tensor4D {\n  const { filters, bias } = params.conv;\n\n  let out = tf.conv2d(x, filters, strides, padding);\n  out = tf.add(out, bias);\n  out = scale(out, params.scale);\n  return withRelu ? tf.relu(out) : out;\n}\n\nexport function conv(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [1, 1], true);\n}\n\nexport function convNoRelu(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [1, 1], false);\n}\n\nexport function convDown(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [2, 2], true, 'valid');\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, extractWeightsFactory, ExtractWeightsFunction, ParamMapping } from '../common/index';\nimport { isFloat } from '../utils/index';\nimport { ConvLayerParams, NetParams, ResidualLayerParams, ScaleLayerParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  function extractFilterValues(numFilterValues: number, numFilters: number, filterSize: number): tf.Tensor4D {\n    const weights = extractWeights(numFilterValues);\n    const depth = weights.length / (numFilters * filterSize * filterSize);\n\n    if (isFloat(depth)) {\n      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);\n    }\n\n    return tf.tidy(\n      () => tf.transpose(\n        tf.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),\n        [2, 3, 1, 0],\n      ),\n    );\n  }\n\n  function extractConvParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvParams {\n    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);\n    const bias = tf.tensor1d(extractWeights(numFilters));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return { filters, bias };\n  }\n\n  function extractScaleLayerParams(numWeights: number, mappedPrefix: string): ScaleLayerParams {\n    const weights = tf.tensor1d(extractWeights(numWeights));\n    const biases = tf.tensor1d(extractWeights(numWeights));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/weights` },\n      { paramPath: `${mappedPrefix}/biases` },\n    );\n\n    return {\n      weights,\n      biases,\n    };\n  }\n\n  function extractConvLayerParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvLayerParams {\n    const conv = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);\n    const scale = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);\n\n    return { conv, scale };\n  }\n\n  function extractResidualLayerParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n    isDown = false,\n  ): ResidualLayerParams {\n    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);\n    const conv2 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);\n\n    return { conv1, conv2 };\n  }\n\n  return {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const conv32_down = extractConvLayerParams(4704, 32, 7, 'conv32_down');\n  const conv32_1 = extractResidualLayerParams(9216, 32, 3, 'conv32_1');\n  const conv32_2 = extractResidualLayerParams(9216, 32, 3, 'conv32_2');\n  const conv32_3 = extractResidualLayerParams(9216, 32, 3, 'conv32_3');\n\n  const conv64_down = extractResidualLayerParams(36864, 64, 3, 'conv64_down', true);\n  const conv64_1 = extractResidualLayerParams(36864, 64, 3, 'conv64_1');\n  const conv64_2 = extractResidualLayerParams(36864, 64, 3, 'conv64_2');\n  const conv64_3 = extractResidualLayerParams(36864, 64, 3, 'conv64_3');\n\n  const conv128_down = extractResidualLayerParams(147456, 128, 3, 'conv128_down', true);\n  const conv128_1 = extractResidualLayerParams(147456, 128, 3, 'conv128_1');\n  const conv128_2 = extractResidualLayerParams(147456, 128, 3, 'conv128_2');\n\n  const conv256_down = extractResidualLayerParams(589824, 256, 3, 'conv256_down', true);\n  const conv256_1 = extractResidualLayerParams(589824, 256, 3, 'conv256_1');\n  const conv256_2 = extractResidualLayerParams(589824, 256, 3, 'conv256_2');\n  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, 'conv256_down_out');\n\n  const fc = tf.tidy(\n    () => tf.transpose(tf.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0]),\n  );\n  paramMappings.push({ paramPath: 'fc' });\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  const params = {\n    conv32_down,\n    conv32_1,\n    conv32_2,\n    conv32_3,\n    conv64_down,\n    conv64_1,\n    conv64_2,\n    conv64_3,\n    conv128_down,\n    conv128_1,\n    conv128_2,\n    conv256_down,\n    conv256_1,\n    conv256_2,\n    conv256_down_out,\n    fc,\n  };\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, ParamMapping } from '../common/index';\nimport { isTensor2D } from '../utils/index';\nimport { ConvLayerParams, NetParams, ResidualLayerParams, ScaleLayerParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractScaleLayerParams(prefix: string): ScaleLayerParams {\n    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);\n    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);\n\n    return { weights, biases };\n  }\n\n  function extractConvLayerParams(prefix: string): ConvLayerParams {\n    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);\n    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);\n    const scale = extractScaleLayerParams(prefix);\n\n    return { conv: { filters, bias }, scale };\n  }\n\n  function extractResidualLayerParams(prefix: string): ResidualLayerParams {\n    return {\n      conv1: extractConvLayerParams(`${prefix}/conv1`),\n      conv2: extractConvLayerParams(`${prefix}/conv2`),\n    };\n  }\n\n  return {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  } = extractorsFactory(weightMap, paramMappings);\n\n  const conv32_down = extractConvLayerParams('conv32_down');\n  const conv32_1 = extractResidualLayerParams('conv32_1');\n  const conv32_2 = extractResidualLayerParams('conv32_2');\n  const conv32_3 = extractResidualLayerParams('conv32_3');\n\n  const conv64_down = extractResidualLayerParams('conv64_down');\n  const conv64_1 = extractResidualLayerParams('conv64_1');\n  const conv64_2 = extractResidualLayerParams('conv64_2');\n  const conv64_3 = extractResidualLayerParams('conv64_3');\n\n  const conv128_down = extractResidualLayerParams('conv128_down');\n  const conv128_1 = extractResidualLayerParams('conv128_1');\n  const conv128_2 = extractResidualLayerParams('conv128_2');\n\n  const conv256_down = extractResidualLayerParams('conv256_down');\n  const conv256_1 = extractResidualLayerParams('conv256_1');\n  const conv256_2 = extractResidualLayerParams('conv256_2');\n  const conv256_down_out = extractResidualLayerParams('conv256_down_out');\n\n  const { fc } = weightMap;\n  paramMappings.push({ originalPath: 'fc', paramPath: 'fc' });\n\n  if (!isTensor2D(fc)) {\n    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);\n  }\n\n  const params = {\n    conv32_down,\n    conv32_1,\n    conv32_2,\n    conv32_3,\n    conv64_down,\n    conv64_1,\n    conv64_2,\n    conv64_3,\n    conv128_down,\n    conv128_1,\n    conv128_2,\n    conv256_down,\n    conv256_1,\n    conv256_2,\n    conv256_down_out,\n    fc,\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { conv, convDown, convNoRelu } from './convLayer';\nimport { ResidualLayerParams } from './types';\n\nexport function residual(x: tf.Tensor4D, params: ResidualLayerParams): tf.Tensor4D {\n  let out = conv(x, params.conv1);\n  out = convNoRelu(out, params.conv2);\n  out = tf.add(out, x);\n  out = tf.relu(out);\n  return out;\n}\n\nexport function residualDown(x: tf.Tensor4D, params: ResidualLayerParams): tf.Tensor4D {\n  let out = convDown(x, params.conv1);\n  out = convNoRelu(out, params.conv2);\n\n  let pooled = tf.avgPool(x, 2, 2, 'valid') as tf.Tensor4D;\n  const zeros = tf.zeros<tf.Rank.R4>(pooled.shape);\n  const isPad = pooled.shape[3] !== out.shape[3];\n  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];\n\n  if (isAdjustShape) {\n    const padShapeX = [...out.shape] as [number, number, number, number];\n    padShapeX[1] = 1;\n    const zerosW = tf.zeros<tf.Rank.R4>(padShapeX);\n    out = tf.concat([out, zerosW], 1);\n\n    const padShapeY = [...out.shape] as [number, number, number, number];\n    padShapeY[2] = 1;\n    const zerosH = tf.zeros<tf.Rank.R4>(padShapeY);\n    out = tf.concat([out, zerosH], 2);\n  }\n\n  pooled = isPad ? tf.concat([pooled, zeros], 3) : pooled;\n  out = tf.add(pooled, out) as tf.Tensor4D;\n\n  out = tf.relu(out);\n  return out;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { convDown } from './convLayer';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { residual, residualDown } from './residualLayer';\nimport { NetParams } from './types';\n\nexport class FaceRecognitionNet extends NeuralNetwork<NetParams> {\n  constructor() {\n    super('FaceRecognitionNet');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor2D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('FaceRecognitionNet - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(150, true), 'float32');\n\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = convDown(normalized, params.conv32_down);\n      out = tf.maxPool(out, 3, 2, 'valid');\n\n      out = residual(out, params.conv32_1);\n      out = residual(out, params.conv32_2);\n      out = residual(out, params.conv32_3);\n\n      out = residualDown(out, params.conv64_down);\n      out = residual(out, params.conv64_1);\n      out = residual(out, params.conv64_2);\n      out = residual(out, params.conv64_3);\n\n      out = residualDown(out, params.conv128_down);\n      out = residual(out, params.conv128_1);\n      out = residual(out, params.conv128_2);\n\n      out = residualDown(out, params.conv256_down);\n      out = residual(out, params.conv256_1);\n      out = residual(out, params.conv256_2);\n      out = residualDown(out, params.conv256_down_out);\n\n      const globalAvg = out.mean([1, 2]) as tf.Tensor2D;\n      const fullyConnected = tf.matMul(globalAvg, params.fc);\n\n      return fullyConnected as tf.Tensor2D;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async computeFaceDescriptor(input: TNetInput): Promise<Float32Array|Float32Array[]> {\n    // @ts-ignore\n    if (input?.shape?.some((dim) => dim <= 0)) return new Float32Array(128);\n    const netInput = await toNetInput(input);\n    const faceDescriptorTensors = tf.tidy(() => tf.unstack(this.forwardInput(netInput)));\n    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data())) as Float32Array[];\n    faceDescriptorTensors.forEach((t) => t.dispose());\n    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_recognition_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import { FaceRecognitionNet } from './FaceRecognitionNet';\n\nexport * from './FaceRecognitionNet';\n\nexport function createFaceRecognitionNet(weights: Float32Array) {\n  const net = new FaceRecognitionNet();\n  net.extractWeights(weights);\n  return net;\n}\n", "export type WithFaceDescriptor<TSource> = TSource & {\n  descriptor: Float32Array\n}\n\nexport function extendWithFaceDescriptor<\n  TSource\n>(\n  sourceObj: TSource,\n  descriptor: Float32Array,\n): WithFaceDescriptor<TSource> {\n  const extension = { descriptor };\n  return { ...sourceObj, ...extension };\n}\n", "export type WithAge<TSource> = TSource & {\n  age: number\n}\n\nexport function isWithAge(obj: any): obj is WithAge<{}> {\n  return typeof obj.age === 'number';\n}\n\nexport function extendWithAge<\n  TSource\n>(\n  sourceObj: TSource,\n  age: number,\n): WithAge<TSource> {\n  const extension = { age };\n  return { ...sourceObj, ...extension };\n}\n", "import { Gender } from '../ageGenderNet/types';\nimport { isValidProbablitiy } from '../utils/index';\n\nexport type WithGender<TSource> = TSource & {\n  gender: Gender\n  genderProbability: number\n}\n\nexport function isWithGender(obj: any): obj is WithGender<{}> {\n  return (obj.gender === Gender.MALE || obj.gender === Gender.FEMALE)\n    && isValidProbablitiy(obj.genderProbability);\n}\n\nexport function extendWithGender<\n  TSource\n>(\n  sourceObj: TSource,\n  gender: Gender,\n  genderProbability: number,\n): WithGender<TSource> {\n  const extension = { gender, genderProbability };\n  return { ...sourceObj, ...extension };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, ParamMapping, ConvParams, extractWeightsFactory } from '../common/index';\nimport { MobileNetV1, NetParams, PointwiseConvParams, PredictionLayerParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  function extractDepthwiseConvParams(numChannels: number, mappedPrefix: string): MobileNetV1.DepthwiseConvParams {\n    const filters = tf.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);\n    const batch_norm_scale = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_offset = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_mean = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_variance = tf.tensor1d(extractWeights(numChannels));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/batch_norm_scale` },\n      { paramPath: `${mappedPrefix}/batch_norm_offset` },\n      { paramPath: `${mappedPrefix}/batch_norm_mean` },\n      { paramPath: `${mappedPrefix}/batch_norm_variance` },\n    );\n\n    return {\n      filters,\n      batch_norm_scale,\n      batch_norm_offset,\n      batch_norm_mean,\n      batch_norm_variance,\n    };\n  }\n\n  function extractConvParams(\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n    isPointwiseConv?: boolean,\n  ): ConvParams {\n    const filters = tf.tensor4d(\n      extractWeights(channelsIn * channelsOut * filterSize * filterSize),\n      [filterSize, filterSize, channelsIn, channelsOut],\n    );\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/${isPointwiseConv ? 'batch_norm_offset' : 'bias'}` },\n    );\n\n    return { filters, bias };\n  }\n\n  function extractPointwiseConvParams(\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): PointwiseConvParams {\n    const {\n      filters,\n      bias,\n    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);\n\n    return {\n      filters,\n      batch_norm_offset: bias,\n    };\n  }\n\n  function extractConvPairParams(\n    channelsIn: number,\n    channelsOut: number,\n    mappedPrefix: string,\n  ): MobileNetV1.ConvPairParams {\n    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);\n    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);\n\n    return { depthwise_conv, pointwise_conv };\n  }\n\n  function extractMobilenetV1Params(): MobileNetV1.Params {\n    const conv_0 = extractPointwiseConvParams(3, 32, 3, 'mobilenetv1/conv_0');\n    const conv_1 = extractConvPairParams(32, 64, 'mobilenetv1/conv_1');\n    const conv_2 = extractConvPairParams(64, 128, 'mobilenetv1/conv_2');\n    const conv_3 = extractConvPairParams(128, 128, 'mobilenetv1/conv_3');\n    const conv_4 = extractConvPairParams(128, 256, 'mobilenetv1/conv_4');\n    const conv_5 = extractConvPairParams(256, 256, 'mobilenetv1/conv_5');\n    const conv_6 = extractConvPairParams(256, 512, 'mobilenetv1/conv_6');\n    const conv_7 = extractConvPairParams(512, 512, 'mobilenetv1/conv_7');\n    const conv_8 = extractConvPairParams(512, 512, 'mobilenetv1/conv_8');\n    const conv_9 = extractConvPairParams(512, 512, 'mobilenetv1/conv_9');\n    const conv_10 = extractConvPairParams(512, 512, 'mobilenetv1/conv_10');\n    const conv_11 = extractConvPairParams(512, 512, 'mobilenetv1/conv_11');\n    const conv_12 = extractConvPairParams(512, 1024, 'mobilenetv1/conv_12');\n    const conv_13 = extractConvPairParams(1024, 1024, 'mobilenetv1/conv_13');\n    return {\n      conv_0,\n      conv_1,\n      conv_2,\n      conv_3,\n      conv_4,\n      conv_5,\n      conv_6,\n      conv_7,\n      conv_8,\n      conv_9,\n      conv_10,\n      conv_11,\n      conv_12,\n      conv_13,\n    };\n  }\n\n  function extractPredictionLayerParams(): PredictionLayerParams {\n    const conv_0 = extractPointwiseConvParams(1024, 256, 1, 'prediction_layer/conv_0');\n    const conv_1 = extractPointwiseConvParams(256, 512, 3, 'prediction_layer/conv_1');\n    const conv_2 = extractPointwiseConvParams(512, 128, 1, 'prediction_layer/conv_2');\n    const conv_3 = extractPointwiseConvParams(128, 256, 3, 'prediction_layer/conv_3');\n    const conv_4 = extractPointwiseConvParams(256, 128, 1, 'prediction_layer/conv_4');\n    const conv_5 = extractPointwiseConvParams(128, 256, 3, 'prediction_layer/conv_5');\n    const conv_6 = extractPointwiseConvParams(256, 64, 1, 'prediction_layer/conv_6');\n    const conv_7 = extractPointwiseConvParams(64, 128, 3, 'prediction_layer/conv_7');\n    const box_encoding_0_predictor = extractConvParams(512, 12, 1, 'prediction_layer/box_predictor_0/box_encoding_predictor');\n    const class_predictor_0 = extractConvParams(512, 9, 1, 'prediction_layer/box_predictor_0/class_predictor');\n    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, 'prediction_layer/box_predictor_1/box_encoding_predictor');\n    const class_predictor_1 = extractConvParams(1024, 18, 1, 'prediction_layer/box_predictor_1/class_predictor');\n    const box_encoding_2_predictor = extractConvParams(512, 24, 1, 'prediction_layer/box_predictor_2/box_encoding_predictor');\n    const class_predictor_2 = extractConvParams(512, 18, 1, 'prediction_layer/box_predictor_2/class_predictor');\n    const box_encoding_3_predictor = extractConvParams(256, 24, 1, 'prediction_layer/box_predictor_3/box_encoding_predictor');\n    const class_predictor_3 = extractConvParams(256, 18, 1, 'prediction_layer/box_predictor_3/class_predictor');\n    const box_encoding_4_predictor = extractConvParams(256, 24, 1, 'prediction_layer/box_predictor_4/box_encoding_predictor');\n    const class_predictor_4 = extractConvParams(256, 18, 1, 'prediction_layer/box_predictor_4/class_predictor');\n    const box_encoding_5_predictor = extractConvParams(128, 24, 1, 'prediction_layer/box_predictor_5/box_encoding_predictor');\n    const class_predictor_5 = extractConvParams(128, 18, 1, 'prediction_layer/box_predictor_5/class_predictor');\n\n    const box_predictor_0 = {\n      box_encoding_predictor: box_encoding_0_predictor,\n      class_predictor: class_predictor_0,\n    };\n    const box_predictor_1 = {\n      box_encoding_predictor: box_encoding_1_predictor,\n      class_predictor: class_predictor_1,\n    };\n    const box_predictor_2 = {\n      box_encoding_predictor: box_encoding_2_predictor,\n      class_predictor: class_predictor_2,\n    };\n    const box_predictor_3 = {\n      box_encoding_predictor: box_encoding_3_predictor,\n      class_predictor: class_predictor_3,\n    };\n    const box_predictor_4 = {\n      box_encoding_predictor: box_encoding_4_predictor,\n      class_predictor: class_predictor_4,\n    };\n    const box_predictor_5 = {\n      box_encoding_predictor: box_encoding_5_predictor,\n      class_predictor: class_predictor_5,\n    };\n    return {\n      conv_0,\n      conv_1,\n      conv_2,\n      conv_3,\n      conv_4,\n      conv_5,\n      conv_6,\n      conv_7,\n      box_predictor_0,\n      box_predictor_1,\n      box_predictor_2,\n      box_predictor_3,\n      box_predictor_4,\n      box_predictor_5,\n    };\n  }\n\n  return {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n  const {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n  const mobilenetv1 = extractMobilenetV1Params();\n  const prediction_layer = extractPredictionLayerParams();\n  const extra_dim = tf.tensor3d(\n    extractWeights(5118 * 4),\n    [1, 5118, 4],\n  );\n  const output_layer = {\n    extra_dim,\n  };\n  paramMappings.push({ paramPath: 'output_layer/extra_dim' });\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    params: {\n      mobilenetv1,\n      prediction_layer,\n      output_layer,\n    },\n    paramMappings,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, disposeUnusedWeightTensors, extractWeightEntryFactory, ParamMapping } from '../common/index';\nimport { isTensor3D } from '../utils/index';\nimport { BoxPredictionParams, MobileNetV1, NetParams, PointwiseConvParams, PredictionLayerParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractPointwiseConvParams(prefix: string, idx: number, mappedPrefix: string): PointwiseConvParams {\n    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);\n    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);\n    return { filters, batch_norm_offset };\n  }\n\n  function extractConvPairParams(idx: number): MobileNetV1.ConvPairParams {\n    const mappedPrefix = `mobilenetv1/conv_${idx}`;\n    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;\n    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;\n    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;\n\n    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);\n    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);\n    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);\n    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);\n    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);\n\n    return {\n      depthwise_conv: {\n        filters,\n        batch_norm_scale,\n        batch_norm_offset,\n        batch_norm_mean,\n        batch_norm_variance,\n      },\n      pointwise_conv: extractPointwiseConvParams('MobilenetV1', idx, mappedPrefixPointwiseConv),\n    };\n  }\n\n  function extractMobilenetV1Params(): MobileNetV1.Params {\n    return {\n      conv_0: extractPointwiseConvParams('MobilenetV1', 0, 'mobilenetv1/conv_0'),\n      conv_1: extractConvPairParams(1),\n      conv_2: extractConvPairParams(2),\n      conv_3: extractConvPairParams(3),\n      conv_4: extractConvPairParams(4),\n      conv_5: extractConvPairParams(5),\n      conv_6: extractConvPairParams(6),\n      conv_7: extractConvPairParams(7),\n      conv_8: extractConvPairParams(8),\n      conv_9: extractConvPairParams(9),\n      conv_10: extractConvPairParams(10),\n      conv_11: extractConvPairParams(11),\n      conv_12: extractConvPairParams(12),\n      conv_13: extractConvPairParams(13),\n    };\n  }\n\n  function extractConvParams(prefix: string, mappedPrefix: string): ConvParams {\n    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);\n    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);\n    return { filters, bias };\n  }\n\n  function extractBoxPredictorParams(idx: number): BoxPredictionParams {\n    const box_encoding_predictor = extractConvParams(\n      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,\n      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`,\n    );\n    const class_predictor = extractConvParams(\n      `Prediction/BoxPredictor_${idx}/ClassPredictor`,\n      `prediction_layer/box_predictor_${idx}/class_predictor`,\n    );\n    return { box_encoding_predictor, class_predictor };\n  }\n\n  function extractPredictionLayerParams(): PredictionLayerParams {\n    return {\n      conv_0: extractPointwiseConvParams('Prediction', 0, 'prediction_layer/conv_0'),\n      conv_1: extractPointwiseConvParams('Prediction', 1, 'prediction_layer/conv_1'),\n      conv_2: extractPointwiseConvParams('Prediction', 2, 'prediction_layer/conv_2'),\n      conv_3: extractPointwiseConvParams('Prediction', 3, 'prediction_layer/conv_3'),\n      conv_4: extractPointwiseConvParams('Prediction', 4, 'prediction_layer/conv_4'),\n      conv_5: extractPointwiseConvParams('Prediction', 5, 'prediction_layer/conv_5'),\n      conv_6: extractPointwiseConvParams('Prediction', 6, 'prediction_layer/conv_6'),\n      conv_7: extractPointwiseConvParams('Prediction', 7, 'prediction_layer/conv_7'),\n      box_predictor_0: extractBoxPredictorParams(0),\n      box_predictor_1: extractBoxPredictorParams(1),\n      box_predictor_2: extractBoxPredictorParams(2),\n      box_predictor_3: extractBoxPredictorParams(3),\n      box_predictor_4: extractBoxPredictorParams(4),\n      box_predictor_5: extractBoxPredictorParams(5),\n    };\n  }\n\n  return {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  } = extractorsFactory(weightMap, paramMappings);\n  const extra_dim = weightMap['Output/extra_dim'];\n  paramMappings.push({ originalPath: 'Output/extra_dim', paramPath: 'output_layer/extra_dim' });\n  if (!isTensor3D(extra_dim)) {\n    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);\n  }\n\n  const params = {\n    mobilenetv1: extractMobilenetV1Params(),\n    prediction_layer: extractPredictionLayerParams(),\n    output_layer: {\n      extra_dim,\n    },\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { PointwiseConvParams } from './types';\n\nexport function pointwiseConvLayer(x: tf.Tensor4D, params: PointwiseConvParams, strides: [number, number]) {\n  return tf.tidy(() => {\n    let out = tf.conv2d(x, params.filters, strides, 'same');\n    out = tf.add(out, params.batch_norm_offset);\n    return tf.clipByValue(out, 0, 6);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { pointwiseConvLayer } from './pointwiseConvLayer';\nimport { MobileNetV1 } from './types';\n\nconst epsilon = 0.0010000000474974513;\n\nfunction depthwiseConvLayer(x: tf.Tensor4D, params: MobileNetV1.DepthwiseConvParams, strides: [number, number]) {\n  return tf.tidy(() => {\n    let out = tf.depthwiseConv2d(x, params.filters, strides, 'same');\n    out = tf.batchNorm<tf.Rank.R4>(\n      out,\n      params.batch_norm_mean,\n      params.batch_norm_variance,\n      params.batch_norm_offset,\n      params.batch_norm_scale,\n      epsilon,\n    );\n    return tf.clipByValue(out, 0, 6);\n  });\n}\n\nfunction getStridesForLayerIdx(layerIdx: number): [number, number] {\n  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];\n}\n\nexport function mobileNetV1(x: tf.Tensor4D, params: MobileNetV1.Params) {\n  return tf.tidy(() => {\n    let conv11;\n    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);\n\n    const convPairParams = [\n      params.conv_1,\n      params.conv_2,\n      params.conv_3,\n      params.conv_4,\n      params.conv_5,\n      params.conv_6,\n      params.conv_7,\n      params.conv_8,\n      params.conv_9,\n      params.conv_10,\n      params.conv_11,\n      params.conv_12,\n      params.conv_13,\n    ];\n\n    convPairParams.forEach((param, i) => {\n      const layerIdx = i + 1;\n      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);\n      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);\n      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);\n      if (layerIdx === 11) conv11 = out;\n    });\n\n    if (conv11 === null) {\n      throw new Error('mobileNetV1 - output of conv layer 11 is null');\n    }\n\n    return {\n      out,\n      conv11: conv11 as any,\n    };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nfunction IOU(boxes: tf.Tensor2D, i: number, j: number) {\n  const boxesData = boxes.arraySync();\n  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);\n  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);\n  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);\n  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);\n  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);\n  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);\n  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);\n  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);\n  const areaI = (ymaxI - yminI) * (xmaxI - xminI);\n  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);\n  if (areaI <= 0 || areaJ <= 0) return 0.0;\n  const intersectionYmin = Math.max(yminI, yminJ);\n  const intersectionXmin = Math.max(xminI, xminJ);\n  const intersectionYmax = Math.min(ymaxI, ymaxJ);\n  const intersectionXmax = Math.min(xmaxI, xmaxJ);\n  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0.0) * Math.max(intersectionXmax - intersectionXmin, 0.0);\n  return intersectionArea / (areaI + areaJ - intersectionArea);\n}\n\nexport function nonMaxSuppression(\n  boxes: tf.Tensor2D,\n  scores: number[],\n  maxOutputSize: number,\n  iouThreshold: number,\n  scoreThreshold: number,\n): number[] {\n  const numBoxes = boxes.shape[0];\n  const outputSize = Math.min(maxOutputSize, numBoxes);\n\n  const candidates = scores\n    .map((score, boxIndex) => ({ score, boxIndex }))\n    .filter((c) => c.score > scoreThreshold)\n    .sort((c1, c2) => c2.score - c1.score);\n\n  const suppressFunc = (x: number) => (x <= iouThreshold ? 1 : 0);\n  const selected: number[] = [];\n\n  candidates.forEach((c) => {\n    if (selected.length >= outputSize) return;\n    const originalScore = c.score;\n    for (let j = selected.length - 1; j >= 0; --j) {\n      const iou = IOU(boxes, c.boxIndex, selected[j]);\n      if (iou === 0.0) continue;\n      c.score *= suppressFunc(iou);\n      if (c.score <= scoreThreshold) break;\n    }\n    if (originalScore === c.score) {\n      selected.push(c.boxIndex);\n    }\n  });\n  return selected;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { OutputLayerParams } from './types';\n\nfunction getCenterCoordinatesAndSizesLayer(x: tf.Tensor2D) {\n  const vec = tf.unstack(tf.transpose(x, [1, 0]));\n\n  const sizes = [\n    tf.sub(vec[2], vec[0]),\n    tf.sub(vec[3], vec[1]),\n  ];\n  const centers = [\n    tf.add(vec[0], tf.div(sizes[0], 2)),\n    tf.add(vec[1], tf.div(sizes[1], 2)),\n  ];\n  return { sizes, centers };\n}\n\nfunction decodeBoxesLayer(x0: tf.Tensor2D, x1: tf.Tensor2D) {\n  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);\n\n  const vec = tf.unstack(tf.transpose(x1, [1, 0]));\n  const div0_out = tf.div(tf.mul(tf.exp(tf.div(vec[2], 5)), sizes[0]), 2);\n  const add0_out = tf.add(tf.mul(tf.div(vec[0], 10), sizes[0]), centers[0]);\n  const div1_out = tf.div(tf.mul(tf.exp(tf.div(vec[3], 5)), sizes[1]), 2);\n  const add1_out = tf.add(tf.mul(tf.div(vec[1], 10), sizes[1]), centers[1]);\n\n  return tf.transpose(\n    tf.stack([\n      tf.sub(add0_out, div0_out),\n      tf.sub(add1_out, div1_out),\n      tf.add(add0_out, div0_out),\n      tf.add(add1_out, div1_out),\n    ]),\n    [1, 0],\n  );\n}\n\nexport function outputLayer(boxPredictions: tf.Tensor4D, classPredictions: tf.Tensor4D, params: OutputLayerParams) {\n  return tf.tidy(() => {\n    const batchSize = boxPredictions.shape[0];\n\n    let boxes = decodeBoxesLayer(\n      tf.reshape(tf.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]) as tf.Tensor2D,\n      tf.reshape(boxPredictions, [-1, 4]) as tf.Tensor2D,\n    );\n    boxes = tf.reshape(boxes, [batchSize, (boxes.shape[0] / batchSize), 4]);\n\n    const scoresAndClasses = tf.sigmoid(tf.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));\n    let scores = tf.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]) as tf.Tensor;\n\n    scores = tf.reshape(scores, [batchSize, scores.shape[1] as number]);\n\n    const boxesByBatch = tf.unstack(boxes) as tf.Tensor2D[];\n    const scoresByBatch = tf.unstack(scores) as tf.Tensor1D[];\n\n    return { boxes: boxesByBatch, scores: scoresByBatch };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { convLayer } from '../common/index';\nimport { BoxPredictionParams } from './types';\n\nexport function boxPredictionLayer(\n  x: tf.Tensor4D,\n  params: BoxPredictionParams,\n) {\n  return tf.tidy(() => {\n    const batchSize = x.shape[0];\n    const boxPredictionEncoding = tf.reshape(\n      convLayer(x, params.box_encoding_predictor),\n      [batchSize, -1, 1, 4],\n    );\n    const classPrediction = tf.reshape(\n      convLayer(x, params.class_predictor),\n      [batchSize, -1, 3],\n    );\n    return { boxPredictionEncoding, classPrediction };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { boxPredictionLayer } from './boxPredictionLayer';\nimport { pointwiseConvLayer } from './pointwiseConvLayer';\nimport { PredictionLayerParams } from './types';\n\nexport function predictionLayer(\n  x: tf.Tensor4D,\n  conv11: tf.Tensor4D,\n  params: PredictionLayerParams,\n) {\n  return tf.tidy(() => {\n    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);\n    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);\n    const conv2 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);\n    const conv3 = pointwiseConvLayer(conv2, params.conv_3, [2, 2]);\n    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);\n    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);\n    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);\n    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);\n\n    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);\n    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);\n    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);\n    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);\n    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);\n    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);\n\n    const boxPredictions = tf.concat([\n      boxPrediction0.boxPredictionEncoding,\n      boxPrediction1.boxPredictionEncoding,\n      boxPrediction2.boxPredictionEncoding,\n      boxPrediction3.boxPredictionEncoding,\n      boxPrediction4.boxPredictionEncoding,\n      boxPrediction5.boxPredictionEncoding,\n    ], 1) as tf.Tensor4D;\n\n    const classPredictions = tf.concat([\n      boxPrediction0.classPrediction,\n      boxPrediction1.classPrediction,\n      boxPrediction2.classPrediction,\n      boxPrediction3.classPrediction,\n      boxPrediction4.classPrediction,\n      boxPrediction5.classPrediction,\n    ], 1) as tf.Tensor4D;\n\n    return {\n      boxPredictions,\n      classPredictions,\n    };\n  });\n}\n", "export interface ISsdMobilenetv1Options {\n  minConfidence?: number\n  maxResults?: number\n}\n\nexport class SsdMobilenetv1Options {\n  protected _name = 'SsdMobilenetv1Options';\n\n  private _minConfidence: number;\n\n  private _maxResults: number;\n\n  constructor({ minConfidence, maxResults }: ISsdMobilenetv1Options = {}) {\n    this._minConfidence = minConfidence || 0.5;\n    this._maxResults = maxResults || 100;\n\n    if (typeof this._minConfidence !== 'number' || this._minConfidence <= 0 || this._minConfidence >= 1) {\n      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);\n    }\n\n    if (typeof this._maxResults !== 'number') {\n      throw new Error(`${this._name} - expected maxResults to be a number`);\n    }\n  }\n\n  get minConfidence(): number { return this._minConfidence; }\n\n  get maxResults(): number { return this._maxResults; }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Rect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { mobileNetV1 } from './mobileNetV1';\nimport { nonMaxSuppression } from './nonMaxSuppression';\nimport { outputLayer } from './outputLayer';\nimport { predictionLayer } from './predictionLayer';\nimport { ISsdMobilenetv1Options, SsdMobilenetv1Options } from './SsdMobilenetv1Options';\nimport { NetParams } from './types';\n\nexport class SsdMobilenetv1 extends NeuralNetwork<NetParams> {\n  constructor() {\n    super('SsdMobilenetv1');\n  }\n\n  public forwardInput(input: NetInput) {\n    const { params } = this;\n    if (!params) throw new Error('SsdMobilenetv1 - load model before inference');\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(512, false), 'float32');\n      const x = tf.sub(tf.div(batchTensor, 127.5), 1) as tf.Tensor4D; // input is normalized -1..1\n      const features = mobileNetV1(x, params.mobilenetv1);\n      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);\n      return outputLayer(boxPredictions, classPredictions, params.output_layer);\n    });\n  }\n\n  public async forward(input: TNetInput) {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async locateFaces(input: TNetInput, options: ISsdMobilenetv1Options = {}): Promise<FaceDetection[]> {\n    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);\n    const netInput = await toNetInput(input);\n    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);\n    const boxes = _boxes[0];\n    const scores = _scores[0];\n    for (let i = 1; i < _boxes.length; i++) {\n      _boxes[i].dispose();\n      _scores[i].dispose();\n    }\n    const scoresData = Array.from(scores.dataSync());\n    const iouThreshold = 0.5;\n    const indices = nonMaxSuppression(boxes, scoresData as number[], maxResults, iouThreshold, minConfidence);\n    const reshapedDims = netInput.getReshapedInputDimensions(0);\n    const inputSize = netInput.inputSize as number;\n    const padX = inputSize / reshapedDims.width;\n    const padY = inputSize / reshapedDims.height;\n    const boxesData = boxes.arraySync();\n    const results = indices\n      .map((idx) => {\n        const [top, bottom] = [\n          Math.max(0, boxesData[idx][0]),\n          Math.min(1.0, boxesData[idx][2]),\n        ].map((val) => val * padY);\n        const [left, right] = [\n          Math.max(0, boxesData[idx][1]),\n          Math.min(1.0, boxesData[idx][3]),\n        ].map((val) => val * padX);\n        return new FaceDetection(\n          scoresData[idx] as number,\n          new Rect(left, top, right - left, bottom - top),\n          { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) },\n        );\n      });\n    boxes.dispose();\n    scores.dispose();\n    return results;\n  }\n\n  protected getDefaultModelName(): string {\n    return 'ssd_mobilenetv1_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import { SsdMobilenetv1 } from './SsdMobilenetv1';\n\nexport * from './SsdMobilenetv1';\nexport * from './SsdMobilenetv1Options';\n\nexport function createSsdMobilenetv1(weights: Float32Array) {\n  const net = new SsdMobilenetv1();\n  net.extractWeights(weights);\n  return net;\n}\n\nexport function createFaceDetectionNet(weights: Float32Array) {\n  return createSsdMobilenetv1(weights);\n}\n\n// alias for backward compatibily\nexport class FaceDetectionNet extends SsdMobilenetv1 {}\n", "import { Point } from '../classes/index';\n\nexport const IOU_THRESHOLD = 0.4;\n\nexport const BOX_ANCHORS = [\n  new Point(0.738768, 0.874946),\n  new Point(2.42204, 2.65704),\n  new Point(4.30971, 7.04493),\n  new Point(10.246, 4.59428),\n  new Point(12.6868, 11.8741),\n];\n\nexport const BOX_ANCHORS_SEPARABLE = [\n  new Point(1.603231, 2.094468),\n  new Point(6.041143, 7.080126),\n  new Point(2.882459, 3.518061),\n  new Point(4.266906, 5.178857),\n  new Point(9.041765, 10.66308),\n];\n\nexport const MEAN_RGB_SEPARABLE: [number, number, number] = [117.001, 114.697, 97.404];\n\nexport const DEFAULT_MODEL_NAME = 'tiny_yolov2_model';\nexport const DEFAULT_MODEL_NAME_SEPARABLE_CONV = 'tiny_yolov2_separable_conv_model';\n", "import { Point } from '../classes/Point';\n\nexport type TinyYolov2Config = {\n  withSeparableConvs: boolean\n  iouThreshold: number\n  anchors: Point[]\n  classes: string[]\n  meanRgb?: [number, number, number]\n  withClassScores?: boolean,\n  filterSizes?: number[]\n  isFirstLayerConv2d?: boolean\n}\n\nconst isNumber = (arg: any) => typeof arg === 'number';\n\nexport function validateConfig(config: any) {\n  if (!config) {\n    throw new Error(`invalid config: ${config}`);\n  }\n\n  if (typeof config.withSeparableConvs !== 'boolean') {\n    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);\n  }\n\n  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1.0) {\n    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);\n  }\n\n  if (\n    !Array.isArray(config.classes)\n    || !config.classes.length\n    || !config.classes.every((c: any) => typeof c === 'string')\n  ) {\n    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);\n  }\n\n  if (\n    !Array.isArray(config.anchors)\n    || !config.anchors.length\n    || !config.anchors.map((a: any) => a || {}).every((a: any) => isNumber(a.x) && isNumber(a.y))\n  ) {\n    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);\n  }\n\n  if (config.meanRgb && (\n    !Array.isArray(config.meanRgb)\n    || config.meanRgb.length !== 3\n    || !config.meanRgb.every(isNumber)\n  )) {\n    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function leaky(x: tf.Tensor4D): tf.Tensor4D {\n  return tf.tidy(() => {\n    const min = tf.mul(x, tf.scalar(0.10000000149011612));\n    return tf.add(tf.relu(tf.sub(x, min)), min);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { leaky } from './leaky';\nimport { ConvWithBatchNorm } from './types';\n\nexport function convWithBatchNorm(x: tf.Tensor4D, params: ConvWithBatchNorm): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]) as tf.Tensor4D;\n    out = tf.conv2d(out, params.conv.filters, [1, 1], 'valid');\n    out = tf.sub(out, params.bn.sub);\n    out = tf.mul(out, params.bn.truediv);\n    out = tf.add(out, params.conv.bias);\n    return leaky(out);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { SeparableConvParams } from '../common/types';\nimport { leaky } from './leaky';\n\nexport function depthwiseSeparableConv(x: tf.Tensor4D, params: SeparableConvParams): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]) as tf.Tensor4D;\n    out = tf.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], 'valid');\n    out = tf.add(out, params.bias);\n    return leaky(out);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { extractConvParamsFactory } from '../common/index';\nimport { extractSeparableConvParamsFactory } from '../common/extractSeparableConvParamsFactory';\nimport { extractWeightsFactory } from '../common/extractWeightsFactory';\nimport { ExtractWeightsFunction, ParamMapping } from '../common/types';\nimport { TinyYolov2Config } from './config';\nimport { BatchNorm, ConvWithBatchNorm, TinyYolov2NetParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n\n  function extractBatchNormParams(size: number, mappedPrefix: string): BatchNorm {\n    const sub = tf.tensor1d(extractWeights(size));\n    const truediv = tf.tensor1d(extractWeights(size));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/sub` },\n      { paramPath: `${mappedPrefix}/truediv` },\n    );\n    return { sub, truediv };\n  }\n\n  function extractConvWithBatchNormParams(channelsIn: number, channelsOut: number, mappedPrefix: string): ConvWithBatchNorm {\n    const conv = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);\n    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);\n    return { conv, bn };\n  }\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  return {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  };\n}\n\nexport function extractParams(\n  weights: Float32Array,\n  config: TinyYolov2Config,\n  boxEncodingSize: number,\n  filterSizes: number[],\n): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n  let params: TinyYolov2NetParams;\n\n  if (config.withSeparableConvs) {\n    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;\n    const conv0 = config.isFirstLayerConv2d\n      ? extractConvParams(s0, s1, 3, 'conv0')\n      : extractSeparableConvParams(s0, s1, 'conv0');\n    const conv1 = extractSeparableConvParams(s1, s2, 'conv1');\n    const conv2 = extractSeparableConvParams(s2, s3, 'conv2');\n    const conv3 = extractSeparableConvParams(s3, s4, 'conv3');\n    const conv4 = extractSeparableConvParams(s4, s5, 'conv4');\n    const conv5 = extractSeparableConvParams(s5, s6, 'conv5');\n    const conv6 = s7 ? extractSeparableConvParams(s6, s7, 'conv6') : undefined;\n    const conv7 = s8 ? extractSeparableConvParams(s7, s8, 'conv7') : undefined;\n    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, 'conv8');\n    params = {\n      conv0, conv1, conv2, conv3, conv4, conv5, conv6, conv7, conv8,\n    };\n  } else {\n    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;\n    const conv0 = extractConvWithBatchNormParams(s0, s1, 'conv0');\n    const conv1 = extractConvWithBatchNormParams(s1, s2, 'conv1');\n    const conv2 = extractConvWithBatchNormParams(s2, s3, 'conv2');\n    const conv3 = extractConvWithBatchNormParams(s3, s4, 'conv3');\n    const conv4 = extractConvWithBatchNormParams(s4, s5, 'conv4');\n    const conv5 = extractConvWithBatchNormParams(s5, s6, 'conv5');\n    const conv6 = extractConvWithBatchNormParams(s6, s7, 'conv6');\n    const conv7 = extractConvWithBatchNormParams(s7, s8, 'conv7');\n    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, 'conv8');\n    params = {\n      conv0, conv1, conv2, conv3, conv4, conv5, conv6, conv7, conv8,\n    };\n  }\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from '../common/index';\nimport { disposeUnusedWeightTensors } from '../common/disposeUnusedWeightTensors';\nimport { loadSeparableConvParamsFactory } from '../common/extractSeparableConvParamsFactory';\nimport { extractWeightEntryFactory } from '../common/extractWeightEntryFactory';\nimport { ParamMapping } from '../common/types';\nimport { TinyYolov2Config } from './config';\nimport { BatchNorm, ConvWithBatchNorm, TinyYolov2NetParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractBatchNormParams(prefix: string): BatchNorm {\n    const sub = extractWeightEntry(`${prefix}/sub`, 1);\n    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);\n    return { sub, truediv };\n  }\n\n  function extractConvParams(prefix: string): ConvParams {\n    const filters = extractWeightEntry(`${prefix}/filters`, 4);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { filters, bias };\n  }\n\n  function extractConvWithBatchNormParams(prefix: string): ConvWithBatchNorm {\n    const conv = extractConvParams(`${prefix}/conv`);\n    const bn = extractBatchNormParams(`${prefix}/bn`);\n    return { conv, bn };\n  }\n\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n  return {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n  config: TinyYolov2Config,\n): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  } = extractorsFactory(weightMap, paramMappings);\n\n  let params: TinyYolov2NetParams;\n\n  if (config.withSeparableConvs) {\n    // eslint-disable-next-line no-mixed-operators\n    const numFilters = (config.filterSizes && config.filterSizes.length || 9);\n    params = {\n      conv0: config.isFirstLayerConv2d ? extractConvParams('conv0') : extractSeparableConvParams('conv0'),\n      conv1: extractSeparableConvParams('conv1'),\n      conv2: extractSeparableConvParams('conv2'),\n      conv3: extractSeparableConvParams('conv3'),\n      conv4: extractSeparableConvParams('conv4'),\n      conv5: extractSeparableConvParams('conv5'),\n      conv6: numFilters > 7 ? extractSeparableConvParams('conv6') : undefined,\n      conv7: numFilters > 8 ? extractSeparableConvParams('conv7') : undefined,\n      conv8: extractConvParams('conv8'),\n    };\n  } else {\n    params = {\n      conv0: extractConvWithBatchNormParams('conv0'),\n      conv1: extractConvWithBatchNormParams('conv1'),\n      conv2: extractConvWithBatchNormParams('conv2'),\n      conv3: extractConvWithBatchNormParams('conv3'),\n      conv4: extractConvWithBatchNormParams('conv4'),\n      conv5: extractConvWithBatchNormParams('conv5'),\n      conv6: extractConvWithBatchNormParams('conv6'),\n      conv7: extractConvWithBatchNormParams('conv7'),\n      conv8: extractConvParams('conv8'),\n    };\n  }\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n  return { params, paramMappings };\n}\n", "export interface ITinyYolov2Options {\n  inputSize?: number\n  scoreThreshold?: number\n}\n\nexport class TinyYolov2Options {\n  protected _name = 'TinyYolov2Options';\n\n  private _inputSize: number;\n\n  private _scoreThreshold: number;\n\n  constructor({ inputSize, scoreThreshold }: ITinyYolov2Options = {}) {\n    this._inputSize = inputSize || 416;\n    this._scoreThreshold = scoreThreshold || 0.5;\n\n    if (typeof this._inputSize !== 'number' || this._inputSize % 32 !== 0) {\n      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);\n    }\n\n    if (typeof this._scoreThreshold !== 'number' || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {\n      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);\n    }\n  }\n\n  get inputSize(): number { return this._inputSize; }\n\n  get scoreThreshold(): number { return this._scoreThreshold; }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { BoundingBox } from '../classes/BoundingBox';\nimport { Dimensions } from '../classes/Dimensions';\nimport { ObjectDetection } from '../classes/ObjectDetection';\nimport { convLayer } from '../common/index';\nimport { ConvParams, SeparableConvParams } from '../common/types';\nimport { toNetInput } from '../dom/index';\nimport { NetInput } from '../dom/NetInput';\nimport { TNetInput } from '../dom/types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { sigmoid } from '../ops/index';\nimport { nonMaxSuppression } from '../ops/nonMaxSuppression';\nimport { normalize } from '../ops/normalize';\nimport { TinyYolov2Config, validateConfig } from './config';\nimport { convWithBatchNorm } from './convWithBatchNorm';\nimport { depthwiseSeparableConv } from './depthwiseSeparableConv';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { leaky } from './leaky';\nimport { ITinyYolov2Options, TinyYolov2Options } from './TinyYolov2Options';\nimport { DefaultTinyYolov2NetParams, MobilenetParams, TinyYolov2NetParams } from './types';\n\nexport class TinyYolov2Base extends NeuralNetwork<TinyYolov2NetParams> {\n  public static DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];\n\n  private _config: TinyYolov2Config;\n\n  constructor(config: TinyYolov2Config) {\n    super('TinyYolov2');\n    validateConfig(config);\n    this._config = config;\n  }\n\n  public get config(): TinyYolov2Config {\n    return this._config;\n  }\n\n  public get withClassScores(): boolean {\n    return this.config.withClassScores || this.config.classes.length > 1;\n  }\n\n  public get boxEncodingSize(): number {\n    return 5 + (this.withClassScores ? this.config.classes.length : 0);\n  }\n\n  public runTinyYolov2(x: tf.Tensor4D, params: DefaultTinyYolov2NetParams): tf.Tensor4D {\n    let out = convWithBatchNorm(x, params.conv0);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv1);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv2);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv3);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv4);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv5);\n    out = tf.maxPool(out, [2, 2], [1, 1], 'same');\n    out = convWithBatchNorm(out, params.conv6);\n    out = convWithBatchNorm(out, params.conv7);\n    return convLayer(out, params.conv8, 'valid', false);\n  }\n\n  public runMobilenet(x: tf.Tensor4D, params: MobilenetParams): tf.Tensor4D {\n    let out = this.config.isFirstLayerConv2d\n      ? leaky(convLayer(x, params.conv0 as ConvParams, 'valid', false))\n      : depthwiseSeparableConv(x, params.conv0 as SeparableConvParams);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv1);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv2);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv3);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv4);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv5);\n    out = tf.maxPool(out, [2, 2], [1, 1], 'same');\n    out = params.conv6 ? depthwiseSeparableConv(out, params.conv6) : out;\n    out = params.conv7 ? depthwiseSeparableConv(out, params.conv7) : out;\n    return convLayer(out, params.conv8, 'valid', false);\n  }\n\n  public forwardInput(input: NetInput, inputSize: number): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('TinyYolov2 - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      let batchTensor = tf.cast(input.toBatchTensor(inputSize, false), 'float32');\n      batchTensor = this.config.meanRgb\n        ? normalize(batchTensor, this.config.meanRgb)\n        : batchTensor;\n      batchTensor = batchTensor.div(255) as tf.Tensor4D;\n      return this.config.withSeparableConvs\n        ? this.runMobilenet(batchTensor, params as MobilenetParams)\n        : this.runTinyYolov2(batchTensor, params as DefaultTinyYolov2NetParams);\n    });\n  }\n\n  public async forward(input: TNetInput, inputSize: number): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input), inputSize);\n  }\n\n  public async detect(input: TNetInput, forwardParams: ITinyYolov2Options = {}): Promise<ObjectDetection[]> {\n    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput, inputSize);\n    const out0 = tf.tidy(() => tf.unstack(out)[0].expandDims()) as tf.Tensor4D;\n    const inputDimensions = {\n      width: netInput.getInputWidth(0),\n      height: netInput.getInputHeight(0),\n    };\n\n    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);\n    out.dispose();\n    out0.dispose();\n\n    const boxes = results.map((res) => res.box);\n    const scores = results.map((res) => res.score);\n    const classScores = results.map((res) => res.classScore);\n    const classNames = results.map((res) => this.config.classes[res.label]);\n\n    const indices = nonMaxSuppression(\n      boxes.map((box) => box.rescale(inputSize)),\n      scores,\n      this.config.iouThreshold,\n      true,\n    );\n\n    const detections = indices.map((idx) => new ObjectDetection(\n      scores[idx],\n      classScores[idx],\n      classNames[idx],\n      boxes[idx],\n      inputDimensions,\n    ));\n    return detections;\n  }\n\n  protected getDefaultModelName(): string {\n    return '';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap, this.config);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const filterSizes = this.config.filterSizes || TinyYolov2Base.DEFAULT_FILTER_SIZES;\n\n    const numFilters = filterSizes ? filterSizes.length : undefined;\n    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {\n      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);\n    }\n    return extractParams(weights, this.config, this.boxEncodingSize, filterSizes);\n  }\n\n  protected async extractBoxes(\n    outputTensor: tf.Tensor4D,\n    inputBlobDimensions: Dimensions,\n    scoreThreshold?: number,\n  ) {\n    const { width, height } = inputBlobDimensions;\n    const inputSize = Math.max(width, height);\n    const correctionFactorX = inputSize / width;\n    const correctionFactorY = inputSize / height;\n\n    const numCells = outputTensor.shape[1];\n    const numBoxes = this.config.anchors.length;\n\n    const [boxesTensor, scoresTensor, classScoresTensor] = tf.tidy(() => {\n      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);\n\n      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);\n      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);\n      const classScores = this.withClassScores\n        ? tf.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3)\n        : tf.scalar(0);\n      return [boxes, scores, classScores];\n    });\n\n    const results = [] as any;\n    const scoresData = await scoresTensor.array();\n    const boxesData = await boxesTensor.array();\n    for (let row = 0; row < numCells; row++) {\n      for (let col = 0; col < numCells; col++) {\n        for (let anchor = 0; anchor < numBoxes; anchor++) {\n          const score = sigmoid(scoresData[row][col][anchor][0]);\n          if (!scoreThreshold || score > scoreThreshold) {\n            const ctX = ((col + sigmoid(boxesData[row][col][anchor][0])) / numCells) * correctionFactorX;\n            const ctY = ((row + sigmoid(boxesData[row][col][anchor][1])) / numCells) * correctionFactorY;\n            const widthLocal = ((Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x) / numCells) * correctionFactorX;\n            const heightLocal = ((Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y) / numCells) * correctionFactorY;\n            const x = (ctX - (widthLocal / 2));\n            const y = (ctY - (heightLocal / 2));\n            const pos = { row, col, anchor };\n            const { classScore, label } = this.withClassScores\n              ? await this.extractPredictedClass(classScoresTensor as tf.Tensor4D, pos)\n              : { classScore: 1, label: 0 };\n            results.push({\n              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),\n              score,\n              classScore: score * classScore,\n              label,\n              ...pos,\n            });\n          }\n        }\n      }\n    }\n\n    boxesTensor.dispose();\n    scoresTensor.dispose();\n    classScoresTensor.dispose();\n    return results;\n  }\n\n  private async extractPredictedClass(classesTensor: tf.Tensor4D, pos: { row: number, col: number, anchor: number }) {\n    const { row, col, anchor } = pos;\n    const classesData = await classesTensor.array();\n    return Array(this.config.classes.length).fill(0)\n      .map((_, i) => classesData[row][col][anchor][i])\n      .map((classScore, label) => ({\n        classScore,\n        label,\n      }))\n      .reduce((max, curr) => (max.classScore > curr.classScore ? max : curr));\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection, Point } from '../classes/index';\nimport { ParamMapping } from '../common/types';\nimport { TNetInput } from '../dom/types';\nimport {\n  BOX_ANCHORS,\n  BOX_ANCHORS_SEPARABLE,\n  DEFAULT_MODEL_NAME,\n  DEFAULT_MODEL_NAME_SEPARABLE_CONV,\n  IOU_THRESHOLD,\n  MEAN_RGB_SEPARABLE,\n} from './const';\nimport { TinyYolov2Base } from './TinyYolov2Base';\nimport { ITinyYolov2Options } from './TinyYolov2Options';\nimport { TinyYolov2NetParams } from './types';\n\nexport class TinyYolov2 extends TinyYolov2Base {\n  constructor(withSeparableConvs = true) {\n    const config = {\n      withSeparableConvs,\n      iouThreshold: IOU_THRESHOLD,\n      classes: ['face'],\n      ...(withSeparableConvs\n        ? {\n          anchors: BOX_ANCHORS_SEPARABLE,\n          meanRgb: MEAN_RGB_SEPARABLE,\n        }\n        : {\n          anchors: BOX_ANCHORS,\n          withClassScores: true,\n        }),\n    };\n\n    super(config);\n  }\n\n  public get withSeparableConvs(): boolean {\n    return this.config.withSeparableConvs;\n  }\n\n  public get anchors(): Point[] {\n    return this.config.anchors;\n  }\n\n  public async locateFaces(input: TNetInput, forwardParams: ITinyYolov2Options): Promise<FaceDetection[]> {\n    const objectDetections = await this.detect(input, forwardParams);\n    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));\n  }\n\n  protected override getDefaultModelName(): string {\n    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;\n  }\n\n  protected override extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n    return super.extractParamsFromWeightMap(weightMap);\n  }\n}\n", "import { TinyYolov2 } from './TinyYolov2';\n\nexport * from './TinyYolov2Options';\nexport * from './config';\nexport * from './types';\nexport { TinyYolov2 };\n\nexport function createTinyYolov2(weights: Float32Array, withSeparableConvs = true) {\n  const net = new TinyYolov2(withSeparableConvs);\n  net.extractWeights(weights);\n  return net;\n}\n", "import { ITinyYolov2Options, TinyYolov2Options } from '../tinyYolov2/index';\n\nexport type ITinyFaceDetectorOptions = ITinyYolov2Options\n\nexport class TinyFaceDetectorOptions extends TinyYolov2Options {\n  protected override _name = 'TinyFaceDetectorOptions';\n}\n", "export class ComposableTask<T> {\n  // eslint-disable-next-line no-unused-vars\n  public async then(onfulfilled: (value: T) => T | PromiseLike<T>): Promise<T> {\n    return onfulfilled(await this.run());\n  }\n\n  public async run(): Promise<T> {\n    throw new Error('ComposableTask - run is not implemented');\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { extractFaces, extractFaceTensors, TNetInput } from '../dom/index';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\n\nexport async function extractAllFacesAndComputeResults<TSource extends WithFaceDetection<{}>, TResult>(\n  parentResults: TSource[],\n  input: TNetInput,\n  // eslint-disable-next-line no-unused-vars\n  computeResults: (faces: Array<HTMLCanvasElement | tf.Tensor3D>) => Promise<TResult>,\n  extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D> | null,\n  // eslint-disable-next-line no-unused-vars\n  getRectForAlignment: (parentResult: WithFaceLandmarks<TSource, any>) => FaceDetection = ({ alignedRect }) => alignedRect,\n) {\n  const faceBoxes = parentResults.map((parentResult) => (isWithFaceLandmarks(parentResult)\n    ? getRectForAlignment(parentResult)\n    : parentResult.detection));\n  const faces: Array<HTMLCanvasElement | tf.Tensor3D> = extractedFaces || (\n    input instanceof tf.Tensor\n      ? await extractFaceTensors(input, faceBoxes)\n      : await extractFaces(input, faceBoxes)\n  );\n  const results = await computeResults(faces);\n  faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n  return results;\n}\n\nexport async function extractSingleFaceAndComputeResult<TSource extends WithFaceDetection<{}>, TResult>(\n  parentResult: TSource,\n  input: TNetInput,\n  // eslint-disable-next-line no-unused-vars\n  computeResult: (face: HTMLCanvasElement | tf.Tensor3D) => Promise<TResult>,\n  extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D> | null,\n  // eslint-disable-next-line no-unused-vars\n  getRectForAlignment?: (parentResultLocal: WithFaceLandmarks<TSource, any>) => FaceDetection,\n) {\n  return extractAllFacesAndComputeResults<TSource, TResult>(\n    [parentResult],\n    input,\n    async (faces) => computeResult(faces[0]),\n    extractedFaces,\n    getRectForAlignment,\n  );\n}\n", "import { Point } from '../classes/index';\n\nexport const IOU_THRESHOLD = 0.4;\n\nexport const BOX_ANCHORS = [\n  new Point(1.603231, 2.094468),\n  new Point(6.041143, 7.080126),\n  new Point(2.882459, 3.518061),\n  new Point(4.266906, 5.178857),\n  new Point(9.041765, 10.66308),\n];\n\nexport const MEAN_RGB: [number, number, number] = [117.001, 114.697, 97.404];\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection, Point } from '../classes/index';\nimport { ParamMapping } from '../common/index';\nimport { TNetInput } from '../dom/index';\nimport { ITinyYolov2Options } from '../tinyYolov2/index';\nimport { TinyYolov2Base } from '../tinyYolov2/TinyYolov2Base';\nimport { TinyYolov2NetParams } from '../tinyYolov2/types';\nimport { BOX_ANCHORS, IOU_THRESHOLD, MEAN_RGB } from './const';\n\nexport class TinyFaceDetector extends TinyYolov2Base {\n  constructor() {\n    const config = {\n      withSeparableConvs: true,\n      iouThreshold: IOU_THRESHOLD,\n      classes: ['face'],\n      anchors: BOX_ANCHORS,\n      meanRgb: MEAN_RGB,\n      isFirstLayerConv2d: true,\n      filterSizes: [3, 16, 32, 64, 128, 256, 512],\n    };\n\n    super(config);\n  }\n\n  public get anchors(): Point[] {\n    return this.config.anchors;\n  }\n\n  public async locateFaces(input: TNetInput, forwardParams: ITinyYolov2Options): Promise<FaceDetection[]> {\n    const objectDetections = await this.detect(input, forwardParams);\n    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));\n  }\n\n  protected override getDefaultModelName(): string {\n    return 'tiny_face_detector_model';\n  }\n\n  protected override extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n    return super.extractParamsFromWeightMap(weightMap);\n  }\n}\n", "import { AgeGenderNet } from '../ageGenderNet/AgeGenderNet';\nimport { AgeAndGenderPrediction } from '../ageGenderNet/types';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { TNetInput } from '../dom/index';\nimport { FaceExpressionNet } from '../faceExpressionNet/FaceExpressionNet';\nimport { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\nimport { FaceLandmark68Net } from '../faceLandmarkNet/FaceLandmark68Net';\nimport { FaceLandmark68TinyNet } from '../faceLandmarkNet/FaceLandmark68TinyNet';\nimport { FaceRecognitionNet } from '../faceRecognitionNet/FaceRecognitionNet';\nimport { SsdMobilenetv1 } from '../ssdMobilenetv1/SsdMobilenetv1';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { TinyFaceDetector } from '../tinyFaceDetector/TinyFaceDetector';\nimport { TinyFaceDetectorOptions } from '../tinyFaceDetector/TinyFaceDetectorOptions';\nimport { ITinyYolov2Options, TinyYolov2 } from '../tinyYolov2/index';\n\nexport const nets = {\n  ssdMobilenetv1: new SsdMobilenetv1(),\n  tinyFaceDetector: new TinyFaceDetector(),\n  tinyYolov2: new TinyYolov2(),\n  faceLandmark68Net: new FaceLandmark68Net(),\n  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),\n  faceRecognitionNet: new FaceRecognitionNet(),\n  faceExpressionNet: new FaceExpressionNet(),\n  ageGenderNet: new AgeGenderNet(),\n};\n\n/**\n * Attempts to detect all faces in an image using SSD Mobilenetv1 Network.\n *\n * @param input The input image.\n * @param options (optional, default: see SsdMobilenetv1Options constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const ssdMobilenetv1 = (input: TNetInput, options: SsdMobilenetv1Options): Promise<FaceDetection[]> => nets.ssdMobilenetv1.locateFaces(input, options);\n\n/**\n * Attempts to detect all faces in an image using the Tiny Face Detector.\n *\n * @param input The input image.\n * @param options (optional, default: see TinyFaceDetectorOptions constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const tinyFaceDetector = (input: TNetInput, options: TinyFaceDetectorOptions): Promise<FaceDetection[]> => nets.tinyFaceDetector.locateFaces(input, options);\n\n/**\n * Attempts to detect all faces in an image using the Tiny Yolov2 Network.\n *\n * @param input The input image.\n * @param options (optional, default: see TinyYolov2Options constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const tinyYolov2 = (input: TNetInput, options: ITinyYolov2Options): Promise<FaceDetection[]> => nets.tinyYolov2.locateFaces(input, options);\n\n/**\n * Detects the 68 point face landmark positions of the face shown in an image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns 68 point face landmarks or array thereof in case of batch input.\n */\nexport const detectFaceLandmarks = (input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> => nets.faceLandmark68Net.detectLandmarks(input);\n\n/**\n * Detects the 68 point face landmark positions of the face shown in an image\n * using a tinier version of the 68 point face landmark model, which is slightly\n * faster at inference, but also slightly less accurate.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns 68 point face landmarks or array thereof in case of batch input.\n */\nexport const detectFaceLandmarksTiny = (input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> => nets.faceLandmark68TinyNet.detectLandmarks(input);\n\n/**\n * Computes a 128 entry vector (face descriptor / face embeddings) from the face shown in an image,\n * which uniquely represents the features of that persons face. The computed face descriptor can\n * be used to measure the similarity between faces, by computing the euclidean distance of two\n * face descriptors.\n *\n * @param inputs The face image extracted from the aligned bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Face descriptor with 128 entries or array thereof in case of batch input.\n */\nexport const computeFaceDescriptor = (input: TNetInput): Promise<Float32Array | Float32Array[]> => nets.faceRecognitionNet.computeFaceDescriptor(input);\n\n/**\n * Recognizes the facial expressions from a face image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Facial expressions with corresponding probabilities or array thereof in case of batch input.\n */\nexport const recognizeFaceExpressions = (input: TNetInput): Promise<FaceExpressions | FaceExpressions[]> => nets.faceExpressionNet.predictExpressions(input);\n\n/**\n * Predicts age and gender from a face image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Predictions with age, gender and gender probability or array thereof in case of batch input.\n */\nexport const predictAgeAndGender = (input: TNetInput): Promise<AgeAndGenderPrediction | AgeAndGenderPrediction[]> => nets.ageGenderNet.predictAgeAndGender(input);\n\nexport const loadSsdMobilenetv1Model = (url: string) => nets.ssdMobilenetv1.load(url);\nexport const loadTinyFaceDetectorModel = (url: string) => nets.tinyFaceDetector.load(url);\nexport const loadTinyYolov2Model = (url: string) => nets.tinyYolov2.load(url);\nexport const loadFaceLandmarkModel = (url: string) => nets.faceLandmark68Net.load(url);\nexport const loadFaceLandmarkTinyModel = (url: string) => nets.faceLandmark68TinyNet.load(url);\nexport const loadFaceRecognitionModel = (url: string) => nets.faceRecognitionNet.load(url);\nexport const loadFaceExpressionModel = (url: string) => nets.faceExpressionNet.load(url);\nexport const loadAgeGenderModel = (url: string) => nets.ageGenderNet.load(url);\n\n// backward compatibility\nexport const loadFaceDetectionModel = loadSsdMobilenetv1Model;\nexport const locateFaces = ssdMobilenetv1;\nexport const detectLandmarks = detectFaceLandmarks;\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { TNetInput } from '../dom/index';\nimport { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { extendWithFaceExpressions, WithFaceExpressions } from '../factories/WithFaceExpressions';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderTask, PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\n\nexport class PredictFaceExpressionsTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D>,\n  ) {\n    super();\n  }\n}\n\nexport class PredictAllFaceExpressionsTask<TSource extends WithFaceDetection<{}>> extends PredictFaceExpressionsTaskBase<WithFaceExpressions<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceExpressions<TSource>[]> {\n    const parentResults = await this.parentTask;\n\n    const faceExpressionsByFace = await extractAllFacesAndComputeResults<TSource, FaceExpressions[]>(\n      parentResults,\n      this.input,\n      async (faces) => Promise.all(\n        faces.map((face) => nets.faceExpressionNet.predictExpressions(face) as Promise<FaceExpressions>),\n      ),\n      this.extractedFaces,\n    );\n\n    return parentResults.map(\n      (parentResult, i) => extendWithFaceExpressions<TSource>(parentResult, faceExpressionsByFace[i]),\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderTask(this, this.input);\n  }\n}\n\nexport class PredictSingleFaceExpressionsTask<TSource extends WithFaceDetection<{}>> extends PredictFaceExpressionsTaskBase<WithFaceExpressions<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceExpressions<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) {\n      return undefined;\n    }\n\n    const faceExpressions = await extractSingleFaceAndComputeResult<TSource, FaceExpressions>(\n      parentResult,\n      this.input,\n      (face) => nets.faceExpressionNet.predictExpressions(face) as Promise<FaceExpressions>,\n      this.extractedFaces,\n    );\n\n    return extendWithFaceExpressions(parentResult, faceExpressions);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderTask(this, this.input);\n  }\n}\n\nexport class PredictAllFaceExpressionsWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictAllFaceExpressionsTask<TSource> {\n  override withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleFaceExpressionsWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictSingleFaceExpressionsTask<TSource> {\n  override withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { AgeAndGenderPrediction } from '../ageGenderNet/types';\nimport { TNetInput } from '../dom/index';\nimport { extendWithAge, WithAge } from '../factories/WithAge';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { extendWithGender, WithGender } from '../factories/WithGender';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllFaceExpressionsTask, PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class PredictAgeAndGenderTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D>,\n  ) {\n    super();\n  }\n}\n\nexport class PredictAllAgeAndGenderTask<TSource extends WithFaceDetection<{}>> extends PredictAgeAndGenderTaskBase<WithAge<WithGender<TSource>>[], TSource[]> {\n  public override async run(): Promise<WithAge<WithGender<TSource>>[]> {\n    const parentResults = await this.parentTask;\n    const ageAndGenderByFace = await extractAllFacesAndComputeResults<TSource, AgeAndGenderPrediction[]>(\n      parentResults,\n      this.input,\n      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face) as Promise<AgeAndGenderPrediction>)),\n      this.extractedFaces,\n    );\n    return parentResults.map((parentResult, i) => {\n      const { age, gender, genderProbability } = ageAndGenderByFace[i];\n      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);\n    });\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleAgeAndGenderTask<TSource extends WithFaceDetection<{}>> extends PredictAgeAndGenderTaskBase<WithAge<WithGender<TSource>> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithAge<WithGender<TSource>> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) return undefined;\n    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult<TSource, AgeAndGenderPrediction>(\n      parentResult,\n      this.input,\n      (face) => nets.ageGenderNet.predictAgeAndGender(face) as Promise<AgeAndGenderPrediction>,\n      this.extractedFaces,\n    );\n    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsTask(this, this.input);\n  }\n}\n\nexport class PredictAllAgeAndGenderWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictAllAgeAndGenderTask<TSource> {\n  override withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleAgeAndGenderWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictSingleAgeAndGenderTask<TSource> {\n  override withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { TNetInput } from '../dom/index';\nimport { extendWithFaceDescriptor, WithFaceDescriptor } from '../factories/WithFaceDescriptor';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class ComputeFaceDescriptorsTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n  ) {\n    super();\n  }\n}\n\nexport class ComputeAllFaceDescriptorsTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends ComputeFaceDescriptorsTaskBase<WithFaceDescriptor<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceDescriptor<TSource>[]> {\n    const parentResults = await this.parentTask;\n    const descriptors = await extractAllFacesAndComputeResults<TSource, Float32Array[]>(\n      parentResults,\n      this.input,\n      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face) as Promise<Float32Array>)),\n      null,\n      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true }),\n    );\n    return descriptors.map((descriptor, i) => extendWithFaceDescriptor<TSource>(parentResults[i], descriptor));\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n}\n\nexport class ComputeSingleFaceDescriptorTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends ComputeFaceDescriptorsTaskBase<WithFaceDescriptor<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceDescriptor<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) return undefined;\n    const descriptor = await extractSingleFaceAndComputeResult<TSource, Float32Array>(\n      parentResult,\n      this.input,\n      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face) as Promise<Float32Array>,\n      null,\n      // eslint-disable-next-line no-shadow, @typescript-eslint/no-shadow\n      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true }),\n    );\n    return extendWithFaceDescriptor(parentResult, descriptor);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { extractFaces, extractFaceTensors, TNetInput } from '../dom/index';\nimport { FaceLandmark68Net } from '../faceLandmarkNet/FaceLandmark68Net';\nimport { FaceLandmark68TinyNet } from '../faceLandmarkNet/FaceLandmark68TinyNet';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { extendWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class DetectFaceLandmarksTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected useTinyLandmarkNet: boolean,\n  ) {\n    super();\n  }\n\n  protected get landmarkNet(): FaceLandmark68Net | FaceLandmark68TinyNet {\n    return this.useTinyLandmarkNet\n      ? nets.faceLandmark68TinyNet\n      : nets.faceLandmark68Net;\n  }\n}\n\nexport class DetectAllFaceLandmarksTask<TSource extends WithFaceDetection<{}>> extends DetectFaceLandmarksTaskBase<WithFaceLandmarks<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceLandmarks<TSource>[]> {\n    const parentResults = await this.parentTask;\n    const detections = parentResults.map((res) => res.detection);\n    const faces: Array<HTMLCanvasElement | tf.Tensor3D> = this.input instanceof tf.Tensor\n      ? await extractFaceTensors(this.input, detections)\n      : await extractFaces(this.input, detections);\n    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face))) as FaceLandmarks68[];\n    faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n    const result = parentResults\n      .filter((_parentResult, i) => faceLandmarksByFace[i])\n      .map((parentResult, i) => extendWithFaceLandmarks<TSource>(parentResult, faceLandmarksByFace[i]));\n    return result;\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class DetectSingleFaceLandmarksTask<TSource extends WithFaceDetection<{}>> extends DetectFaceLandmarksTaskBase<WithFaceLandmarks<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceLandmarks<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) {\n      return undefined;\n    }\n    const { detection } = parentResult;\n    const faces: Array<HTMLCanvasElement | tf.Tensor3D> = this.input instanceof tf.Tensor\n      ? await extractFaceTensors(this.input, [detection])\n      : await extractFaces(this.input, [detection]);\n    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]) as FaceLandmarks68;\n    faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n    return extendWithFaceLandmarks<TSource>(parentResult, landmarks);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { TNetInput } from '../dom/index';\nimport { extendWithFaceDetection, WithFaceDetection } from '../factories/WithFaceDetection';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { TinyFaceDetectorOptions } from '../tinyFaceDetector/TinyFaceDetectorOptions';\nimport { TinyYolov2Options } from '../tinyYolov2/index';\nimport { ComposableTask } from './ComposableTask';\nimport { DetectAllFaceLandmarksTask, DetectSingleFaceLandmarksTask } from './DetectFaceLandmarksTasks';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderTask, PredictSingleAgeAndGenderTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsTask, PredictSingleFaceExpressionsTask } from './PredictFaceExpressionsTask';\nimport { FaceDetectionOptions } from './types';\n\nexport class DetectFacesTaskBase<TReturn> extends ComposableTask<TReturn> {\n  // eslint-disable-next-line no-unused-vars\n  constructor(protected input: TNetInput, protected options: FaceDetectionOptions = new SsdMobilenetv1Options()) {\n    super();\n  }\n}\n\nexport class DetectAllFacesTask extends DetectFacesTaskBase<FaceDetection[]> {\n  public override async run(): Promise<FaceDetection[]> {\n    const { input, options } = this;\n    let result;\n    if (options instanceof TinyFaceDetectorOptions) result = nets.tinyFaceDetector.locateFaces(input, options);\n    else if (options instanceof SsdMobilenetv1Options) result = nets.ssdMobilenetv1.locateFaces(input, options);\n    else if (options instanceof TinyYolov2Options) result = nets.tinyYolov2.locateFaces(input, options);\n    else throw new Error('detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options');\n    return result;\n  }\n\n  private runAndExtendWithFaceDetections(): Promise<WithFaceDetection<{}>[]> {\n    return new Promise<WithFaceDetection<{}>[]>((resolve, reject) => {\n      this.run()\n        .then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection))))\n        .catch((err) => reject(err));\n    });\n  }\n\n  withFaceLandmarks(useTinyLandmarkNet = false) {\n    return new DetectAllFaceLandmarksTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n      useTinyLandmarkNet,\n    );\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n    );\n  }\n}\n\nexport class DetectSingleFaceTask extends DetectFacesTaskBase<FaceDetection | undefined> {\n  public override async run(): Promise<FaceDetection | undefined> {\n    const faceDetections = await new DetectAllFacesTask(this.input, this.options);\n    let faceDetectionWithHighestScore = faceDetections[0];\n    faceDetections.forEach((faceDetection) => {\n      if (faceDetection.score > faceDetectionWithHighestScore.score) faceDetectionWithHighestScore = faceDetection;\n    });\n    return faceDetectionWithHighestScore;\n  }\n\n  private runAndExtendWithFaceDetection(): Promise<WithFaceDetection<{}> | undefined> {\n    // eslint-disable-next-line no-async-promise-executor\n    return new Promise<WithFaceDetection<{}> | undefined>(async (resolve) => {\n      const detection = await this.run();\n      resolve(detection ? extendWithFaceDetection<{}>({}, detection) : undefined);\n    });\n  }\n\n  withFaceLandmarks(useTinyLandmarkNet = false) {\n    return new DetectSingleFaceLandmarksTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n      useTinyLandmarkNet,\n    );\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n    );\n  }\n}\n", "import { TNetInput } from '../dom/index';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { DetectAllFacesTask, DetectSingleFaceTask } from './DetectFacesTasks';\nimport { FaceDetectionOptions } from './types';\n\nexport function detectSingleFace(input: TNetInput, options: FaceDetectionOptions = new SsdMobilenetv1Options()): DetectSingleFaceTask {\n  return new DetectSingleFaceTask(input, options);\n}\n\nexport function detectAllFaces(input: TNetInput, options: FaceDetectionOptions = new SsdMobilenetv1Options()): DetectAllFacesTask {\n  return new DetectAllFacesTask(input, options);\n}\n", "import { TNetInput } from '../dom/index';\nimport { WithFaceDescriptor, WithFaceDetection, WithFaceLandmarks } from '../factories/index';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/index';\nimport { ITinyYolov2Options, TinyYolov2Options } from '../tinyYolov2/index';\nimport { detectAllFaces } from './detectFaces';\n\nexport async function allFacesSsdMobilenetv1(input: TNetInput, minConfidence?: number): Promise<WithFaceDescriptor<WithFaceLandmarks<WithFaceDetection<{}>>>[]> {\n  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {}))\n    .withFaceLandmarks()\n    .withFaceDescriptors();\n}\n\nexport async function allFacesTinyYolov2(input: TNetInput, forwardParams: ITinyYolov2Options = {}): Promise<WithFaceDescriptor<WithFaceLandmarks<WithFaceDetection<{}>>>[]> {\n  return detectAllFaces(input, new TinyYolov2Options(forwardParams))\n    .withFaceLandmarks()\n    .withFaceDescriptors();\n}\n\nexport const allFaces = allFacesSsdMobilenetv1;\n", "export function euclideanDistance(arr1: number[] | Float32Array, arr2: number[] | Float32Array) {\n  if (arr1.length !== arr2.length) throw new Error('euclideanDistance: arr1.length !== arr2.length');\n  const desc1 = Array.from(arr1);\n  const desc2 = Array.from(arr2);\n  return Math.sqrt(\n    desc1\n      .map((val, i) => val - desc2[i])\n      .reduce((res, diff) => res + (diff * diff), 0),\n  );\n}\n", "import { FaceMatch } from '../classes/FaceMatch';\nimport { LabeledFaceDescriptors } from '../classes/LabeledFaceDescriptors';\nimport { euclideanDistance } from '../euclideanDistance';\nimport { WithFaceDescriptor } from '../factories/index';\n\nexport class FaceMatcher {\n  private _labeledDescriptors: LabeledFaceDescriptors[];\n  private _distanceThreshold: number;\n\n  constructor(inputs: LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>, distanceThreshold = 0.6) {\n    this._distanceThreshold = distanceThreshold;\n    const inputArray = Array.isArray(inputs) ? inputs : [inputs];\n    if (!inputArray.length) throw new Error('FaceRecognizer.constructor - expected atleast one input');\n    let count = 1;\n    const createUniqueLabel = () => `person ${count++}`;\n    this._labeledDescriptors = inputArray.map((desc) => {\n      if (desc instanceof LabeledFaceDescriptors) return desc;\n      if (desc instanceof Float32Array) return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);\n      if (desc.descriptor && desc.descriptor instanceof Float32Array) return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);\n      throw new Error('FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>');\n    });\n  }\n\n  public get labeledDescriptors(): LabeledFaceDescriptors[] { return this._labeledDescriptors; }\n\n  public get distanceThreshold(): number { return this._distanceThreshold; }\n\n  public computeMeanDistance(queryDescriptor: Float32Array, descriptors: Float32Array[]): number {\n    return descriptors\n      .map((d) => euclideanDistance(d, queryDescriptor))\n      .reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);\n  }\n\n  public matchDescriptor(queryDescriptor: Float32Array): FaceMatch {\n    return this.labeledDescriptors\n      .map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors)))\n      .reduce((best, curr) => (best.distance < curr.distance ? best : curr));\n  }\n\n  public findBestMatch(queryDescriptor: Float32Array): FaceMatch {\n    const bestMatch = this.matchDescriptor(queryDescriptor);\n    return (bestMatch.distance < this._distanceThreshold) ? bestMatch : new FaceMatch('unknown', bestMatch.distance);\n  }\n\n  public toJSON(): any {\n    return {\n      distanceThreshold: this._distanceThreshold,\n      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON()),\n    };\n  }\n\n  public static fromJSON(json: any): FaceMatcher {\n    const labeledDescriptors = json.labeledDescriptors.map((ld: any) => LabeledFaceDescriptors.fromJSON(ld));\n    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);\n  }\n}\n", "import { TinyFaceDetector } from './TinyFaceDetector';\n\nexport * from './TinyFaceDetector';\nexport * from './TinyFaceDetectorOptions';\n\nexport function createTinyFaceDetector(weights: Float32Array) {\n  const net = new TinyFaceDetector();\n  net.extractWeights(weights);\n  return net;\n}\n", "import { Dimensions, IDimensions } from './classes/index';\nimport { FaceDetection } from './classes/FaceDetection';\nimport { FaceLandmarks } from './classes/FaceLandmarks';\nimport { extendWithFaceDetection, isWithFaceDetection } from './factories/WithFaceDetection';\nimport { extendWithFaceLandmarks, isWithFaceLandmarks } from './factories/WithFaceLandmarks';\n\nexport function resizeResults<T>(results: T, dimensions: IDimensions): T {\n  const { width, height } = new Dimensions(dimensions.width, dimensions.height);\n\n  if (width <= 0 || height <= 0) {\n    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);\n  }\n\n  if (Array.isArray(results)) {\n    // return results.map(obj => resizeResults(obj, { width, height })) as any as T\n    return (results as Array<any>).map((obj) => resizeResults(obj, { width, height } as IDimensions)) as any as T;\n  }\n\n  if (isWithFaceLandmarks(results)) {\n    const resizedDetection = results.detection.forSize(width, height);\n    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);\n    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);\n  }\n\n  if (isWithFaceDetection(results)) {\n    return extendWithFaceDetection(results, results.detection.forSize(width, height));\n  }\n\n  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {\n    return (results as any).forSize(width, height);\n  }\n\n  return results;\n}\n", "import * as tf from '../dist/tfjs.esm';\nimport * as draw from './draw/index';\nimport * as utils from './utils/index';\nimport * as pkg from '../package.json';\n\nexport { tf, draw, utils };\n\nexport * from './ageGenderNet/index';\nexport * from './classes/index';\nexport * from './dom/index';\nexport * from './env/index';\nexport * from './faceExpressionNet/index';\nexport * from './faceLandmarkNet/index';\nexport * from './faceRecognitionNet/index';\nexport * from './factories/index';\nexport * from './globalApi/index';\nexport * from './ops/index';\nexport * from './ssdMobilenetv1/index';\nexport * from './tinyFaceDetector/index';\nexport * from './tinyYolov2/index';\nexport * from './euclideanDistance';\nexport * from './NeuralNetwork';\nexport * from './resizeResults';\n\nexport const version = pkg.version as string;\n\n// set webgl defaults\n// if (browser) tf.ENV.set('WEBGL_USE_SHAPES_UNIFORMS', true);\n"],
-  "mappings": ";;;;;;+XAAA,IAAAA,GAAA,GAAAC,GAAAD,GAAA,SAAAE,GAAA,SAAAC,GAAA,UAAAC,GAAA,sBAAAC,GAAA,qBAAAC,GAAA,kBAAAC,GAAA,oBAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,WAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,kBAAAC,GAAA,gBAAAC,GAAA,gBAAAC,GAAA,gBAAAC,GAAA,mBAAAC,GAAA,aAAAC,GAAA,kBAAAC,GAAA,gBAAAC,GAAA,aAAAC,GAAA,iBAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,gBAAAC,GAAA,YAAAC,GAAA,eAAAC,GAAA,WAAAC,GAAA,WAAAC,GAAA,yBAAAC,GAAA,wBAAAC,GAAA,WAAAC,GAAA,2BAAAC,GAAA,0BAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,kBAAAC,GAAA,YAAAC,GAAA,WAAAC,GAAA,mBAAAC,GAAA,gBAAAC,GAAA,kBAAAC,GAAA,iBAAAC,GAAA,0BAAAC,GAAA,wCAAAC,GAAA,uCAAAC,GAAA,SAAAC,GAAA,eAAAC,GAAA,6BAAAC,GAAA,4BAAAC,GAAA,QAAAC,GAAA,kBAAAC,GAAA,WAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,gBAAAC,GAAA,UAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,eAAAC,GAAA,UAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,kBAAAC,GAAA,UAAAC,GAAA,aAAAC,GAAA,eAAAC,GAAA,mBAAAC,GAAA,gBAAAC,GAAA,yBAAAC,GAAA,iBAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,eAAAC,GAAA,YAAAC,GAAA,iBAAAC,GAAA,YAAAC,GAAA,SAAAC,GAAA,aAAAC,GAAA,SAAAC,GAAA,cAAAC,GAAA,aAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,kBAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,kBAAAC,GAAA,gBAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,cAAAC,GAAA,aAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,cAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,qBAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,kBAAAC,GAAA,gBAAAC,GAAA,sBAAAC,GAAA,YAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,QAAAC,GAAA,sBAAAC,GAAA,gBAAAC,GAAA,aAAAC,GAAA,QAAAC,GAAA,wBAAAC,GAAA,wBAAAC,GAAA,wBAAAC,GAAA,aAAAC,GAAA,oBAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,cAAAC,GAAA,0BAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,qBAAAC,GAAA,QAAAC,GAAA,iBAAAC,GAAA,gBAAAC,GAAA,yBAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,YAAAC,GAAA,eAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,mBAAAC,GAAA,uBAAAC,GAAA,0BAAAC,GAAA,8BAAAC,GAAA,YAAAC,GAAA,qBAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,iBAAAC,GAAA,cAAAC,GAAA,iBAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,eAAAC,GAAA,YAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,aAAAC,GAAA,mBAAAC,GAAA,wBAAAC,GAAA,kBAAAC,GAAA,sBAAAC,GAAA,qBAAAC,GAAA,kBAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,sBAAAC,GAAA,SAAAC,GAAA,iBAAAC,GAAA,iBAAAC,GAAA,gBAAAC,GAAA,2BAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,mBAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,iBAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,cAAAC,GAAA,cAAAC,GAAA,WAAAC,GAAA,WAAAC,GAAA,uBAAAC,GAAA,eAAAC,GAAA,aAAAC,GAAA,cAAAC,GAAA,iBAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,QAAAC,EAAA,SAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,WAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,YAAAC,GAAA,iBAAAC,EAAA,kBAAAC,GAAA,cAAAC,GAAA,gBAAAC,GAAA,gBAAAC,GAAA,gBAAAC,GAAA,mBAAAC,GAAA,aAAAC,GAAA,qBAAAC,GAAA,kBAAAC,GAAA,gBAAAC,GAAA,mBAAAC,GAAA,YAAAC,GAAA,WAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,gBAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,gBAAAC,GAAA,WAAAC,GAAA,WAAAC,GAAA,oBAAAC,GAAA,WAAAC,GAAA,oBAAAC,GAAA,0BAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,iBAAAC,GAAA,YAAAC,GAAA,WAAAC,GAAA,eAAAC,GAAA,SAAAC,GAAA,kBAAAC,GAAA,oBAAAC,GAAA,iBAAAC,GAAA,oBAAAC,GAAA,iBAAAC,GAAA,gBAAAC,GAAA,SAAAC,GAAA,eAAAC,GAAA,+BAAAC,GAAA,YAAAC,GAAA,qBAAAC,GAAA,QAAAC,GAAA,aAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,WAAAC,GAAA,QAAAC,GAAA,oBAAAC,GAAA,mBAAAC,GAAA,wBAAAC,GAAA,WAAAC,GAAA,QAAAC,EAAA,UAAAC,GAAA,QAAAC,GAAA,kBAAAC,GAAA,QAAAC,GAAA,eAAAC,GAAA,UAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,gBAAAC,GAAA,uBAAAC,GAAA,UAAAC,GAAA,aAAAC,GAAA,mBAAAC,GAAA,UAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,gBAAAC,GAAA,eAAAC,GAAA,gBAAAC,GAAA,cAAAC,GAAA,yBAAAC,GAAA,oBAAAC,IAAA,eAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,iBAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,gBAAAC,GAAA,iBAAAC,GAAA,UAAAC,GAAA,OAAAC,GAAA,UAAAC,GAAA,aAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,iBAAAC,GAAA,WAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,cAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,mBAAAC,GAAA,uBAAAC,GAAA,oBAAAC,GAAA,+BAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,cAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,cAAAC,GAAA,eAAAC,GAAA,WAAAC,GAAA,eAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,sBAAAC,GAAA,YAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,YAAAC,GAAA,QAAAC,GAAA,YAAAC,GAAA,cAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,WAAAC,GAAA,YAAAC,GAAA,kBAAAC,GAAA,QAAAC,EAAA,iBAAAC,GAAA,gBAAAC,GAAA,QAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,aAAAC,GAAA,WAAAC,GAAA,SAAAC,GAAA,aAAAC,GAAA,OAAAC,EAAA,iBAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,QAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,YAAAC,GAAA,iBAAAC,GAAA,gBAAAC,GAAA,yBAAAC,GAAA,SAAAC,GAAA,gBAAAC,GAAA,iBAAAC,GAAA,yBAAAC,GAAA,kBAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,eAAAC,GAAA,oBAAAC,GAAA,gCAAAC,GAAA,qBAAAC,GAAA,mBAAAC,GAAA,eAAAC,GAAA,iBAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,kBAAAC,GAAA,YAAAC,EAAA,YAAAC,GAAA,cAAAC,GAAA,cAAAC,GAAA,cAAAC,GAAA,cAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,UAAAC,GAAA,WAAAC,GAAA,cAAAC,GAAA,iBAAAC,GAAA,iBAAAC,GAAA,SAAAC,GAAA,oBAAAC,GAAA,eAAAC,GAAA,kBAAAC,GAAA,eAAAC,GAAA,gBAAAC,GAAA,oBAAAC,IAAA,gBAAAC,IAAA,iBAAAC,IAAA,oBAAAC,GAAA,mBAAAC,GAAA,YAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,YAAAC,GAAA,YAAAC,GAAA,YAAAC,GAAA,YAAAC,GAAA,eAAAC,GAAA,YAAAC,GAAA,aAAAC,GAAA,mBAAAC,GAAA,WAAAC,GAAA,kBAAAC,GAAA,aAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,sBAAAC,GAAA,YAAAC,GAAA,UAAAC,GAAA,SAAAC,GAAA,iBAAAC,GAAA,WAAAC,GAAA,QAAAC,GAAA,QAAAC,GAAA,eAAAC,GAAA,QAAAC,GAAA,SAAAC,GAAA,WAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,aAAAC,GAAA,gBAAAC,GAAA,cAAAC,GAAA,SAAAC,EAAA,SAAAC,GAAA,SAAAC,GAAA,SAAAC,GAAA,UAAAC,GAAA,cAAAC,GAAA,oBAAAC,GAAA,WAAAC,GAAA,uBAAAC,GAAA,qBAAAC,GAAA,uBAAAC,GAAA,YAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,SAAAC,EAAA,iBAAAC,GAAA,kBAAAC,GAAA,aAAAC,GAAA,kBAAAC,GAAA,YAAAC,IAAA,sBAAAC,GAAA,iBAAAC,GAAA,mBAAAC,GAAA,iBAAAC,IAAA,kBAAAC,GAAA,UAAAC,GAAA,eAAAC,GAAA,UAAAC,GAAA,eAAAC,GAAA,UAAAC,GAAA,cAAAC,KAMA,IAAIC,GAAG,OAAO,OAAWC,GAAG,OAAO,eAAmBC,GAAG,OAAO,yBAA6BC,GAAG,OAAO,oBAAwBC,GAAG,OAAO,eAAeC,GAAG,OAAO,UAAU,eAAmBC,IAAIC,GAAG,OAAOC,IAAS,YAAYA,GAAQ,OAAO,OAAO,YAAY,IAAI,MAAMD,EAAE,CAAC,IAAI,CAAC,EAAEE,KAAK,OAAOD,IAAS,YAAYA,GAAQ,GAAGC,EAAE,CAAC,EAAEF,GAAG,SAASA,EAAE,CAAC,GAAG,OAAOC,IAAS,YAAY,OAAOA,GAAQ,MAAM,KAAK,SAAS,EAAE,MAAM,IAAI,MAAM,uBAAuBD,EAAE,oBAAoB,CAAC,CAAC,EAAMG,GAAG,CAACH,EAAE,IAAI,KAAK,GAAGA,GAAG,EAAE,CAAC,QAAQ,CAAC,CAAC,GAAG,QAAQ,CAAC,EAAE,EAAE,SAASI,GAAG,CAACJ,EAAE,IAAI,CAAC,QAAQE,KAAK,EAAER,GAAGM,EAAEE,EAAE,CAAC,IAAI,EAAEA,GAAG,WAAW,EAAE,CAAC,CAAC,EAAEG,GAAG,CAACL,EAAE,EAAEE,EAAEI,IAAI,CAAC,GAAG,GAAG,OAAO,GAAG,UAAU,OAAO,GAAG,WAAW,QAAQC,KAAKX,GAAG,CAAC,EAAE,CAACE,GAAG,KAAKE,EAAEO,CAAC,GAAGA,IAAIL,GAAGR,GAAGM,EAAEO,EAAE,CAAC,IAAI,IAAI,EAAEA,GAAG,WAAW,EAAED,EAAEX,GAAG,EAAEY,CAAC,IAAID,EAAE,UAAU,CAAC,EAAE,OAAON,CAAC,EAAMQ,GAAG,CAACR,EAAE,EAAEE,KAAKA,EAAEF,GAAG,KAAKP,GAAGI,GAAGG,CAAC,CAAC,EAAE,CAAC,EAAEK,GAAG,GAAG,CAACL,GAAG,CAACA,EAAE,WAAWN,GAAGQ,EAAE,UAAU,CAAC,MAAMF,EAAE,WAAW,EAAE,CAAC,EAAEE,EAAEF,CAAC,GAAOS,GAAGN,GAAG,CAACO,EAAIC,IAAK,CAACA,EAAG,QAAQC,EAAG,IAAIC,EAAG,KAAK,GAAG,CAACA,EAAG,IAAI,YAAY,SAAS,IAAI,YAAY,OAAO,IAAI,WAAW,CAAC,EAAE,GAAG,IAAI,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,IAAI,IAAI,IAAI,EAAE,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,IAAI,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,IAAI,IAAI,IAAI,EAAE,EAAE,EAAE,IAAI,IAAI,IAAI,GAAG,IAAI,EAAE,EAAE,EAAE,IAAI,IAAI,IAAI,GAAG,IAAI,EAAE,EAAE,EAAE,IAAI,IAAI,IAAI,GAAG,IAAI,EAAE,EAAE,EAAE,IAAI,IAAI,IAAI,GAAG,IAAI,EAAE,EAAE,EAAE,IAAI,IAAI,IAAI,GAAG,IAAI,IAAI,IAAI,IAAI,EAAE,EAAE,GAAG,IAAI,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,IAAI,GAAG,EAAE,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,GAAG,EAAE,IAAI,GAAG,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,IAAI,GAAG,EAAE,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,GAAG,EAAE,IAAI,GAAG,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,IAAI,GAAG,EAAE,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,GAAG,EAAE,IAAI,GAAG,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,IAAI,GAAG,EAAE,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,GAAG,EAAE,IAAI,GAAG,GAAG,EAAE,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,GAAG,EAAE,IAAI,GAAG,GAAG,IAAI,IAAI,IAAI,GAAG,EAAE,GAAG,GAAG,IAAI,IAAI,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,OAAO,OAAOb,EAAN,CAAS,CAAC,SAASY,EAAGZ,EAAEc,EAAEZ,EAAE,CAAC,KAAK,IAAIF,EAAE,EAAE,KAAK,KAAKc,EAAE,EAAE,KAAK,SAAS,CAAC,CAACZ,CAAC,CAACU,EAAG,UAAU,WAAW,OAAO,eAAeA,EAAG,UAAU,aAAa,CAAC,MAAM,EAAE,CAAC,EAAE,SAASG,EAAGf,EAAE,CAAC,OAAOA,GAAGA,EAAE,cAAc,EAAE,CAACY,EAAG,OAAOG,EAAG,IAAIC,EAAG,CAAC,EAAEC,EAAG,CAAC,EAAE,SAASC,EAAGlB,EAAEc,EAAE,CAAC,IAAIZ,EAAEI,EAAEC,EAAE,OAAOO,GAAGd,KAAK,GAAGO,EAAE,GAAGP,GAAGA,EAAE,OAAOM,EAAEW,EAAGjB,GAAGM,GAAGA,GAAGJ,EAAEiB,EAAGnB,GAAGA,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,EAAEO,IAAIU,EAAGjB,GAAGE,GAAGA,KAAKF,GAAG,GAAGO,EAAE,MAAMP,GAAGA,EAAE,OAAOM,EAAEU,EAAGhB,GAAGM,GAAGA,GAAGJ,EAAEiB,EAAGnB,EAAEA,EAAE,EAAE,GAAG,EAAE,EAAE,EAAEO,IAAIS,EAAGhB,GAAGE,GAAGA,GAAG,CAACU,EAAG,QAAQM,EAAG,SAASE,EAAGpB,EAAEc,EAAE,CAAC,GAAG,MAAMd,CAAC,EAAE,OAAOc,EAAEO,EAAGC,EAAG,GAAGR,EAAE,CAAC,GAAGd,EAAE,EAAE,OAAOqB,EAAG,GAAGrB,GAAGuB,EAAG,OAAOC,CAAE,KAAK,CAAC,GAAGxB,GAAG,CAACyB,EAAG,OAAOC,EAAG,GAAG1B,EAAE,GAAGyB,EAAG,OAAOE,CAAE,CAAC,OAAO3B,EAAE,EAAEoB,EAAG,CAACpB,EAAEc,CAAC,EAAE,IAAI,EAAEK,EAAGnB,EAAE4B,EAAG,EAAE5B,EAAE4B,EAAG,EAAEd,CAAC,CAAC,CAACF,EAAG,WAAWQ,EAAG,SAASD,EAAGnB,EAAEc,EAAEZ,EAAE,CAAC,OAAO,IAAIU,EAAGZ,EAAEc,EAAEZ,CAAC,CAAC,CAACU,EAAG,SAASO,EAAG,IAAIU,EAAG,KAAK,IAAI,SAASC,EAAG9B,EAAEc,EAAEZ,EAAE,CAAC,GAAGF,EAAE,SAAS,EAAE,MAAM,MAAM,cAAc,EAAE,GAAGA,IAAI,OAAOA,IAAI,YAAYA,IAAI,aAAaA,IAAI,YAAY,OAAOsB,EAAG,GAAG,OAAOR,GAAG,UAAUZ,EAAEY,EAAEA,EAAE,IAAIA,EAAE,CAAC,CAACA,EAAEZ,EAAEA,GAAG,GAAGA,EAAE,GAAG,GAAGA,EAAE,MAAM,WAAW,OAAO,EAAE,IAAII,EAAE,IAAIA,EAAEN,EAAE,QAAQ,GAAG,GAAG,EAAE,MAAM,MAAM,iBAAiB,EAAE,GAAGM,IAAI,EAAE,OAAOwB,EAAG9B,EAAE,UAAU,CAAC,EAAEc,EAAEZ,CAAC,EAAE,IAAI,EAAE,QAAQK,EAAEa,EAAGS,EAAG3B,EAAE,CAAC,CAAC,EAAE6B,EAAET,EAAGU,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,GAAG,EAAE,CAAC,IAAIC,EAAE,KAAK,IAAI,EAAEjC,EAAE,OAAOgC,CAAC,EAAEE,EAAE,SAASlC,EAAE,UAAUgC,EAAEA,EAAEC,CAAC,EAAE/B,CAAC,EAAE,GAAG+B,EAAE,EAAE,CAAC,IAAIE,GAAEf,EAAGS,EAAG3B,EAAE+B,CAAC,CAAC,EAAEF,EAAEA,EAAE,IAAII,EAAC,EAAE,IAAIf,EAAGc,CAAC,CAAC,CAAC,MAAMH,EAAEA,EAAE,IAAIxB,CAAC,EAAEwB,EAAEA,EAAE,IAAIX,EAAGc,CAAC,CAAC,CAAC,CAAC,OAAOH,EAAE,SAASjB,EAAEiB,CAAC,CAACnB,EAAG,WAAWkB,EAAG,SAASM,EAAGpC,EAAEc,EAAE,CAAC,OAAO,OAAOd,GAAG,SAASoB,EAAGpB,EAAEc,CAAC,EAAE,OAAOd,GAAG,SAAS8B,EAAG9B,EAAEc,CAAC,EAAEK,EAAGnB,EAAE,IAAIA,EAAE,KAAK,OAAOc,GAAG,UAAUA,EAAEd,EAAE,QAAQ,CAAC,CAACY,EAAG,UAAUwB,EAAG,IAAIC,EAAG,GAAG,GAAGC,EAAG,GAAG,GAAGV,EAAGS,EAAGA,EAAGd,EAAGK,EAAGA,EAAGH,EAAGF,EAAG,EAAEgB,EAAGrB,EAAGoB,CAAE,EAAEhB,EAAGJ,EAAG,CAAC,EAAEN,EAAG,KAAKU,EAAG,IAAID,EAAGH,EAAG,EAAE,EAAE,EAAEN,EAAG,MAAMS,EAAG,IAAImB,EAAGtB,EAAG,CAAC,EAAEN,EAAG,IAAI4B,EAAG,IAAIC,EAAGvB,EAAG,EAAE,EAAE,EAAEN,EAAG,KAAK6B,EAAG,IAAIC,EAAGxB,EAAG,EAAE,EAAEN,EAAG,QAAQ8B,EAAG,IAAIf,EAAGR,EAAG,GAAG,WAAW,EAAE,EAAEP,EAAG,UAAUe,EAAG,IAAIH,EAAGL,EAAG,GAAG,GAAG,EAAE,EAAEP,EAAG,mBAAmBY,EAAG,IAAIE,EAAGP,EAAG,EAAE,YAAY,EAAE,EAAEP,EAAG,UAAUc,EAAG,IAAIiB,EAAG/B,EAAG,UAAU+B,EAAG,MAAM,UAAU,CAAC,OAAO,KAAK,SAAS,KAAK,MAAM,EAAE,KAAK,GAAG,EAAEA,EAAG,SAAS,UAAU,CAAC,OAAO,KAAK,UAAU,KAAK,OAAO,GAAGf,GAAI,KAAK,MAAM,GAAG,KAAK,KAAKA,GAAI,KAAK,MAAM,EAAE,EAAEe,EAAG,SAAS,SAAS7B,EAAE,CAAC,GAAGA,EAAEA,GAAG,GAAGA,EAAE,GAAG,GAAGA,EAAE,MAAM,WAAW,OAAO,EAAE,GAAG,KAAK,OAAO,EAAE,MAAM,IAAI,GAAG,KAAK,WAAW,EAAE,GAAG,KAAK,GAAGY,CAAE,EAAE,CAAC,IAAIxB,EAAEkB,EAAGN,CAAC,EAAER,EAAE,KAAK,IAAIJ,CAAC,EAAEK,EAAED,EAAE,IAAIJ,CAAC,EAAE,IAAI,IAAI,EAAE,OAAOI,EAAE,SAASQ,CAAC,EAAEP,EAAE,MAAM,EAAE,SAASO,CAAC,CAAC,KAAM,OAAM,IAAI,KAAK,IAAI,EAAE,SAASA,CAAC,EAAE,QAAQiB,EAAEX,EAAGS,EAAGf,EAAE,CAAC,EAAE,KAAK,QAAQ,EAAEkB,EAAE,KAAKC,EAAE,KAAK,CAAC,IAAIC,EAAEF,EAAE,IAAID,CAAC,EAAEI,EAAEH,EAAE,IAAIE,EAAE,IAAIH,CAAC,CAAC,EAAE,MAAM,IAAI,EAAEa,GAAET,EAAE,SAASrB,CAAC,EAAE,GAAGkB,EAAEE,EAAEF,EAAE,OAAO,EAAE,OAAOY,GAAEX,EAAE,KAAKW,GAAE,OAAO,GAAGA,GAAE,IAAIA,GAAEX,EAAE,GAAGW,GAAEX,CAAC,CAAC,EAAEU,EAAG,YAAY,UAAU,CAAC,OAAO,KAAK,IAAI,EAAEA,EAAG,oBAAoB,UAAU,CAAC,OAAO,KAAK,OAAO,CAAC,EAAEA,EAAG,WAAW,UAAU,CAAC,OAAO,KAAK,GAAG,EAAEA,EAAG,mBAAmB,UAAU,CAAC,OAAO,KAAK,MAAM,CAAC,EAAEA,EAAG,cAAc,UAAU,CAAC,GAAG,KAAK,WAAW,EAAE,OAAO,KAAK,GAAGjB,CAAE,EAAE,GAAG,KAAK,IAAI,EAAE,cAAc,EAAE,QAAQZ,EAAE,KAAK,MAAM,EAAE,KAAK,KAAK,KAAK,IAAIZ,EAAE,GAAGA,EAAE,IAAIY,EAAE,GAAGZ,IAAI,EAAEA,IAAI,CAAC,OAAO,KAAK,MAAM,EAAEA,EAAE,GAAGA,EAAE,CAAC,EAAEyC,EAAG,OAAO,UAAU,CAAC,OAAO,KAAK,OAAO,GAAG,KAAK,MAAM,CAAC,EAAEA,EAAG,IAAIA,EAAG,OAAOA,EAAG,WAAW,UAAU,CAAC,MAAM,CAAC,KAAK,UAAU,KAAK,KAAK,CAAC,EAAEA,EAAG,WAAW,UAAU,CAAC,OAAO,KAAK,UAAU,KAAK,MAAM,CAAC,EAAEA,EAAG,MAAM,UAAU,CAAC,OAAO,KAAK,IAAI,KAAK,CAAC,EAAEA,EAAG,OAAO,UAAU,CAAC,OAAO,KAAK,IAAI,KAAK,CAAC,EAAEA,EAAG,OAAO,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAG,KAAK,WAAWA,EAAE,UAAU,KAAK,OAAO,KAAK,GAAGA,EAAE,OAAO,KAAK,EAAE,GAAG,KAAK,OAAOA,EAAE,MAAM,KAAK,MAAMA,EAAE,GAAG,EAAE6B,EAAG,GAAGA,EAAG,OAAOA,EAAG,UAAU,SAAS7B,EAAE,CAAC,MAAM,CAAC,KAAK,GAAGA,CAAC,CAAC,EAAE6B,EAAG,IAAIA,EAAG,UAAUA,EAAG,GAAGA,EAAG,UAAUA,EAAG,SAAS,SAAS7B,EAAE,CAAC,OAAO,KAAK,KAAKA,CAAC,EAAE,CAAC,EAAE6B,EAAG,GAAGA,EAAG,SAASA,EAAG,gBAAgB,SAAS7B,EAAE,CAAC,OAAO,KAAK,KAAKA,CAAC,GAAG,CAAC,EAAE6B,EAAG,IAAIA,EAAG,gBAAgBA,EAAG,GAAGA,EAAG,gBAAgBA,EAAG,YAAY,SAAS7B,EAAE,CAAC,OAAO,KAAK,KAAKA,CAAC,EAAE,CAAC,EAAE6B,EAAG,GAAGA,EAAG,YAAYA,EAAG,mBAAmB,SAAS7B,EAAE,CAAC,OAAO,KAAK,KAAKA,CAAC,GAAG,CAAC,EAAE6B,EAAG,IAAIA,EAAG,mBAAmBA,EAAG,GAAGA,EAAG,mBAAmBA,EAAG,QAAQ,SAAS7B,EAAE,CAAC,GAAGC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAG,KAAK,GAAGA,CAAC,EAAE,MAAO,GAAE,IAAIZ,EAAE,KAAK,WAAW,EAAEI,EAAEQ,EAAE,WAAW,EAAE,OAAOZ,GAAG,CAACI,EAAE,GAAG,CAACJ,GAAGI,EAAE,EAAE,KAAK,SAASQ,EAAE,OAAO,EAAE,KAAK,OAAO,GAAGA,EAAE,OAAO,KAAK,MAAMA,EAAE,MAAM,EAAE,KAAK,MAAM,EAAE,GAAG,EAAE,KAAK,IAAIA,CAAC,EAAE,WAAW,EAAE,GAAG,CAAC,EAAE6B,EAAG,KAAKA,EAAG,QAAQA,EAAG,OAAO,UAAU,CAAC,MAAM,CAAC,KAAK,UAAU,KAAK,GAAGjB,CAAE,EAAEA,EAAG,KAAK,IAAI,EAAE,IAAIc,CAAE,CAAC,EAAEG,EAAG,IAAIA,EAAG,OAAOA,EAAG,IAAI,SAAS7B,EAAE,CAACC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAG,IAAIZ,EAAE,KAAK,OAAO,GAAGI,EAAE,KAAK,KAAK,MAAMC,EAAE,KAAK,MAAM,GAAGwB,EAAE,KAAK,IAAI,MAAMC,EAAElB,EAAE,OAAO,GAAGmB,EAAEnB,EAAE,KAAK,MAAMoB,EAAEpB,EAAE,MAAM,GAAGqB,EAAErB,EAAE,IAAI,MAAM8B,GAAE,EAAEC,GAAE,EAAEC,GAAE,EAAEC,GAAE,EAAE,OAAOA,IAAGhB,EAAEI,EAAEW,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAGvC,EAAE2B,EAAEW,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAGvC,EAAE2B,EAAEW,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAG1C,EAAE8B,EAAEY,IAAG,MAAMzB,EAAG2B,IAAG,GAAGC,GAAEH,IAAG,GAAGC,GAAE,KAAK,QAAQ,CAAC,EAAEF,EAAG,SAAS,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAG,KAAK,IAAIA,EAAE,IAAI,CAAC,CAAC,EAAE6B,EAAG,IAAIA,EAAG,SAASA,EAAG,SAAS,SAAS7B,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,OAAOQ,EAAG,GAAGP,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGD,EAAG,CAAC,IAAIX,EAAEW,EAAG,IAAI,KAAK,IAAI,KAAK,KAAKC,EAAE,IAAIA,EAAE,IAAI,EAAE,OAAOK,EAAGjB,EAAEW,EAAG,SAAS,EAAE,KAAK,QAAQ,CAAC,CAAC,GAAGC,EAAE,OAAO,EAAE,OAAOQ,EAAG,GAAG,KAAK,GAAGI,CAAE,EAAE,OAAOZ,EAAE,MAAM,EAAEY,EAAGJ,EAAG,GAAGR,EAAE,GAAGY,CAAE,EAAE,OAAO,KAAK,MAAM,EAAEA,EAAGJ,EAAG,GAAG,KAAK,WAAW,EAAE,OAAOR,EAAE,WAAW,EAAE,KAAK,IAAI,EAAE,IAAIA,EAAE,IAAI,CAAC,EAAE,KAAK,IAAI,EAAE,IAAIA,CAAC,EAAE,IAAI,EAAE,GAAGA,EAAE,WAAW,EAAE,OAAO,KAAK,IAAIA,EAAE,IAAI,CAAC,EAAE,IAAI,EAAE,GAAG,KAAK,GAAGyB,CAAE,GAAGzB,EAAE,GAAGyB,CAAE,EAAE,OAAOnB,EAAG,KAAK,SAAS,EAAEN,EAAE,SAAS,EAAE,KAAK,QAAQ,EAAE,IAAIR,EAAE,KAAK,OAAO,GAAGC,EAAE,KAAK,KAAK,MAAMwB,EAAE,KAAK,MAAM,GAAGC,EAAE,KAAK,IAAI,MAAMC,EAAEnB,EAAE,OAAO,GAAGoB,EAAEpB,EAAE,KAAK,MAAMqB,EAAErB,EAAE,MAAM,GAAG8B,GAAE9B,EAAE,IAAI,MAAM+B,GAAE,EAAEC,GAAE,EAAEC,GAAE,EAAEC,GAAE,EAAE,OAAOA,IAAGhB,EAAEY,GAAEG,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAGhB,EAAEa,GAAEE,IAAGC,KAAI,GAAGA,IAAG,MAAMA,IAAGf,EAAEG,EAAEW,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAGvC,EAAEqC,GAAEC,IAAGC,KAAI,GAAGA,IAAG,MAAMA,IAAGf,EAAEI,EAAEU,IAAGC,KAAI,GAAGA,IAAG,MAAMA,IAAGd,EAAEE,EAAEW,IAAGC,KAAI,GAAGA,IAAG,MAAMD,IAAGvC,EAAEsC,GAAErC,EAAE4B,EAAEJ,EAAEG,EAAEF,EAAEC,EAAEY,IAAG,MAAM1B,EAAG4B,IAAG,GAAGC,GAAEH,IAAG,GAAGC,GAAE,KAAK,QAAQ,CAAC,EAAEH,EAAG,IAAIA,EAAG,SAASA,EAAG,OAAO,SAAS7B,EAAE,CAAC,GAAGC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGA,EAAE,OAAO,EAAE,MAAM,MAAM,kBAAkB,EAAE,GAAGD,EAAG,CAAC,GAAG,CAAC,KAAK,UAAU,KAAK,OAAO,aAAaC,EAAE,MAAM,IAAIA,EAAE,OAAO,GAAG,OAAO,KAAK,IAAIZ,GAAG,KAAK,SAASW,EAAG,MAAMA,EAAG,OAAO,KAAK,IAAI,KAAK,KAAKC,EAAE,IAAIA,EAAE,IAAI,EAAE,OAAOK,EAAGjB,EAAEW,EAAG,SAAS,EAAE,KAAK,QAAQ,CAAC,CAAC,GAAG,KAAK,OAAO,EAAE,OAAO,KAAK,SAASQ,EAAGC,EAAG,IAAIhB,EAAEC,EAAEwB,EAAE,GAAG,KAAK,SAAS,CAAC,GAAGjB,EAAE,WAAWA,EAAEA,EAAE,WAAW,GAAGA,EAAE,GAAG,IAAI,EAAE,OAAOO,EAAG,GAAGP,EAAE,GAAG,KAAK,KAAK,CAAC,CAAC,EAAE,OAAO2B,EAAGV,EAAEV,CAAE,KAAK,CAAC,GAAG,KAAK,GAAGK,CAAE,EAAE,CAAC,GAAGZ,EAAE,GAAG0B,CAAE,GAAG1B,EAAE,GAAG4B,CAAE,EAAE,OAAOhB,EAAG,GAAGZ,EAAE,GAAGY,CAAE,EAAE,OAAOc,EAAG,IAAIR,EAAE,KAAK,IAAI,CAAC,EAAE,OAAO1B,EAAE0B,EAAE,IAAIlB,CAAC,EAAE,IAAI,CAAC,EAAER,EAAE,GAAGgB,CAAE,EAAER,EAAE,WAAW,EAAE0B,EAAGE,GAAInC,EAAE,KAAK,IAAIO,EAAE,IAAIR,CAAC,CAAC,EAAEyB,EAAEzB,EAAE,IAAIC,EAAE,IAAIO,CAAC,CAAC,EAAEiB,EAAE,SAASjB,EAAE,GAAGY,CAAE,EAAE,OAAO,KAAK,SAASL,EAAGC,EAAG,GAAG,KAAK,WAAW,EAAE,OAAOR,EAAE,WAAW,EAAE,KAAK,IAAI,EAAE,IAAIA,EAAE,IAAI,CAAC,EAAE,KAAK,IAAI,EAAE,IAAIA,CAAC,EAAE,IAAI,EAAE,GAAGA,EAAE,WAAW,EAAE,OAAO,KAAK,IAAIA,EAAE,IAAI,CAAC,EAAE,IAAI,EAAEiB,EAAET,CAAE,CAAC,IAAIf,EAAE,KAAKA,EAAE,IAAIO,CAAC,GAAG,CAACR,EAAE,KAAK,IAAI,EAAE,KAAK,MAAMC,EAAE,SAAS,EAAEO,EAAE,SAAS,CAAC,CAAC,EAAE,QAAQmB,EAAE,KAAK,KAAK,KAAK,IAAI3B,CAAC,EAAE,KAAK,GAAG,EAAE4B,EAAED,GAAG,GAAG,EAAEJ,EAAG,EAAEI,EAAE,EAAE,EAAEE,EAAEf,EAAGd,CAAC,EAAEsC,GAAET,EAAE,IAAIrB,CAAC,EAAE8B,GAAE,WAAW,GAAGA,GAAE,GAAGrC,CAAC,GAAGD,GAAG4B,EAAEC,EAAEf,EAAGd,EAAE,KAAK,QAAQ,EAAEsC,GAAET,EAAE,IAAIrB,CAAC,EAAEqB,EAAE,OAAO,IAAIA,EAAEK,GAAIT,EAAEA,EAAE,IAAII,CAAC,EAAE5B,EAAEA,EAAE,IAAIqC,EAAC,CAAC,CAAC,OAAOb,CAAC,EAAEY,EAAG,IAAIA,EAAG,OAAOA,EAAG,OAAO,SAAS7B,EAAE,CAAC,GAAGC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGD,EAAG,CAAC,IAAIX,GAAG,KAAK,SAASW,EAAG,MAAMA,EAAG,OAAO,KAAK,IAAI,KAAK,KAAKC,EAAE,IAAIA,EAAE,IAAI,EAAE,OAAOK,EAAGjB,EAAEW,EAAG,SAAS,EAAE,KAAK,QAAQ,CAAC,CAAC,OAAO,KAAK,IAAI,KAAK,IAAIC,CAAC,EAAE,IAAIA,CAAC,CAAC,CAAC,EAAE6B,EAAG,IAAIA,EAAG,OAAOA,EAAG,IAAIA,EAAG,OAAOA,EAAG,IAAI,UAAU,CAAC,OAAOxB,EAAG,CAAC,KAAK,IAAI,CAAC,KAAK,KAAK,KAAK,QAAQ,CAAC,EAAEwB,EAAG,IAAI,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGK,EAAG,KAAK,IAAIL,EAAE,IAAI,KAAK,KAAKA,EAAE,KAAK,KAAK,QAAQ,CAAC,EAAE6B,EAAG,GAAG,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGK,EAAG,KAAK,IAAIL,EAAE,IAAI,KAAK,KAAKA,EAAE,KAAK,KAAK,QAAQ,CAAC,EAAE6B,EAAG,IAAI,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEsB,EAAGtB,CAAC,GAAGK,EAAG,KAAK,IAAIL,EAAE,IAAI,KAAK,KAAKA,EAAE,KAAK,KAAK,QAAQ,CAAC,EAAE6B,EAAG,UAAU,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEA,EAAE,MAAM,IAAIA,GAAG,MAAM,EAAE,KAAKA,EAAE,GAAGK,EAAG,KAAK,KAAKL,EAAE,KAAK,MAAMA,EAAE,KAAK,MAAM,GAAGA,EAAE,KAAK,QAAQ,EAAEK,EAAG,EAAE,KAAK,KAAKL,EAAE,GAAG,KAAK,QAAQ,CAAC,EAAE6B,EAAG,IAAIA,EAAG,UAAUA,EAAG,WAAW,SAAS7B,EAAE,CAAC,OAAOC,EAAGD,CAAC,IAAIA,EAAEA,EAAE,MAAM,IAAIA,GAAG,MAAM,EAAE,KAAKA,EAAE,GAAGK,EAAG,KAAK,MAAML,EAAE,KAAK,MAAM,GAAGA,EAAE,KAAK,MAAMA,EAAE,KAAK,QAAQ,EAAEK,EAAG,KAAK,MAAML,EAAE,GAAG,KAAK,MAAM,EAAE,EAAE,GAAG,KAAK,QAAQ,CAAC,EAAE6B,EAAG,IAAIA,EAAG,WAAWA,EAAG,mBAAmB,SAAS7B,EAAE,CAAC,GAAGC,EAAGD,CAAC,IAAIA,EAAEA,EAAE,MAAM,GAAGA,GAAG,GAAGA,IAAI,EAAE,OAAO,KAAK,IAAIZ,EAAE,KAAK,KAAK,GAAGY,EAAE,GAAG,CAAC,IAAIR,EAAE,KAAK,IAAI,OAAOa,EAAGb,IAAIQ,EAAEZ,GAAG,GAAGY,EAAEZ,IAAIY,EAAE,KAAK,QAAQ,CAAC,KAAM,QAAOA,IAAI,GAAGK,EAAGjB,EAAE,EAAE,KAAK,QAAQ,EAAEiB,EAAGjB,IAAIY,EAAE,GAAG,EAAE,KAAK,QAAQ,CAAC,EAAE6B,EAAG,KAAKA,EAAG,mBAAmBA,EAAG,MAAMA,EAAG,mBAAmBA,EAAG,SAAS,UAAU,CAAC,OAAO,KAAK,SAASxB,EAAG,KAAK,IAAI,KAAK,KAAK,EAAE,EAAE,IAAI,EAAEwB,EAAG,WAAW,UAAU,CAAC,OAAO,KAAK,SAAS,KAAKxB,EAAG,KAAK,IAAI,KAAK,KAAK,EAAE,CAAC,EAAEwB,EAAG,QAAQ,SAAS7B,EAAE,CAAC,OAAOA,EAAE,KAAK,UAAU,EAAE,KAAK,UAAU,CAAC,EAAE6B,EAAG,UAAU,UAAU,CAAC,IAAI7B,EAAE,KAAK,KAAKZ,EAAE,KAAK,IAAI,MAAM,CAACA,EAAE,IAAIA,IAAI,EAAE,IAAIA,IAAI,GAAG,IAAIA,IAAI,GAAGY,EAAE,IAAIA,IAAI,EAAE,IAAIA,IAAI,GAAG,IAAIA,IAAI,EAAE,CAAC,EAAE6B,EAAG,UAAU,UAAU,CAAC,IAAI7B,EAAE,KAAK,KAAKZ,EAAE,KAAK,IAAI,MAAM,CAACY,IAAI,GAAGA,IAAI,GAAG,IAAIA,IAAI,EAAE,IAAIA,EAAE,IAAIZ,IAAI,GAAGA,IAAI,GAAG,IAAIA,IAAI,EAAE,IAAIA,EAAE,GAAG,CAAC,EAAEU,EAAG,UAAU,SAASE,EAAEZ,EAAEI,EAAE,CAAC,OAAOA,EAAEM,EAAG,YAAYE,EAAEZ,CAAC,EAAEU,EAAG,YAAYE,EAAEZ,CAAC,CAAC,EAAEU,EAAG,YAAY,SAASE,EAAEZ,EAAE,CAAC,OAAO,IAAIU,EAAGE,EAAE,GAAGA,EAAE,IAAI,EAAEA,EAAE,IAAI,GAAGA,EAAE,IAAI,GAAGA,EAAE,GAAGA,EAAE,IAAI,EAAEA,EAAE,IAAI,GAAGA,EAAE,IAAI,GAAGZ,CAAC,CAAC,EAAEU,EAAG,YAAY,SAASE,EAAEZ,EAAE,CAAC,OAAO,IAAIU,EAAGE,EAAE,IAAI,GAAGA,EAAE,IAAI,GAAGA,EAAE,IAAI,EAAEA,EAAE,GAAGA,EAAE,IAAI,GAAGA,EAAE,IAAI,GAAGA,EAAE,IAAI,EAAEA,EAAE,GAAGZ,CAAC,CAAC,CAAC,CAAC,EAAM+C,GAAG9C,GAAG,IAAI,CAAC,CAAC,EAAM+C,GAAG/C,GAAG,IAAI,CAAC,CAAC,EAAMgD,GAAGhD,GAAG,CAACiD,EAAGC,IAAK,EAAE,SAASrD,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE2B,EAAE,CAAC,IAAIC,EAAE,KAAKC,EAAEH,EAAE,EAAEE,EAAE,KAAK,UAAU,CAAC,IAAI,EAAE,QAAQA,EAAE,GAAGA,EAAE,EAAE,sBAAsB,OAAOA,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAG,GAAGA,EAAE,EAAE,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,GAAGC,EAAE,GAAG,EAAED,EAAE,GAAGC,EAAE,GAAG,EAAED,EAAE,GAAGC,EAAE,GAAG,EAAED,EAAE,IAAIC,EAAEF,CAAC,EAAEC,EAAE,GAAG,IAAIA,EAAE,IAAI,GAAGA,EAAE,IAAIC,EAAEF,CAAC,EAAEC,EAAE,GAAG,IAAIA,EAAE,IAAI,GAAGA,EAAE,IAAIC,EAAEF,CAAC,EAAEC,EAAE,GAAG,IAAIA,EAAE,IAAI,GAAGC,EAAE,IAAI,CAAC,SAAS5B,EAAE0B,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,GAAGD,EAAE,GAAGC,EAAE,GAAGD,EAAE,GAAGC,EAAE,GAAGD,EAAE,GAAGC,CAAC,CAAC,SAASH,EAAEE,EAAEC,EAAE,CAAC,IAAIC,EAAE,IAAI7B,EAAE2B,CAAC,EAAE,EAAEC,GAAGA,EAAE,MAAMW,EAAEV,EAAE,KAAK,OAAOU,EAAE,MAAM,UAAU,CAAC,OAAOV,EAAE,KAAK,EAAE,WAAW,CAAC,EAAEU,EAAE,OAAO,UAAU,CAAC,OAAOA,EAAE,GAAGA,EAAE,EAAE,QAAQ,GAAG,qBAAqB,EAAEA,EAAE,MAAMA,EAAE,IAAI,OAAO,GAAG,UAAUtC,EAAE,EAAE4B,CAAC,EAAEU,EAAE,MAAM,UAAU,CAAC,OAAOtC,EAAE4B,EAAE,CAAC,CAAC,CAAC,GAAGU,CAAC,CAAC,SAASb,GAAG,CAAC,IAAIC,EAAE,WAAWC,EAAE,SAASC,EAAE,CAACA,EAAE,OAAOA,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,IAAI,CAACF,GAAGE,EAAE,WAAW,CAAC,EAAE,IAAIU,EAAE,mBAAmBZ,EAAEA,EAAEY,IAAI,EAAEA,GAAGZ,EAAEY,GAAGZ,EAAEA,EAAEY,IAAI,EAAEA,GAAGZ,EAAEA,GAAGY,EAAE,UAAU,CAAC,OAAOZ,IAAI,GAAG,qBAAqB,EAAE,OAAOC,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,KAAKA,CAAC,GAAGqB,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMC,GAAGnD,GAAG,CAACoD,EAAGC,IAAK,EAAE,SAASxD,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE0B,EAAE,CAAC,IAAIC,EAAE,KAAKC,EAAE,GAAGD,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,KAAK,UAAU,CAAC,IAAI,EAAEA,EAAE,EAAEA,EAAE,GAAG,GAAG,OAAOA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGA,EAAE,IAAI,GAAG,EAAE,IAAI,CAAC,EAAED,KAAKA,EAAE,GAAGC,EAAE,EAAED,EAAEE,GAAGF,EAAE,QAAQG,EAAE,EAAEA,EAAED,EAAE,OAAO,GAAGC,IAAIF,EAAE,GAAGC,EAAE,WAAWC,CAAC,EAAE,EAAEF,EAAE,KAAK,CAAC,CAAC,SAAS1B,EAAEyB,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,CAAC,CAAC,SAASF,EAAEC,EAAEC,EAAE,CAAC,IAAIC,EAAE,IAAI5B,EAAE0B,CAAC,EAAEG,EAAEF,GAAGA,EAAE,MAAM,EAAE,UAAU,CAAC,OAAOC,EAAE,KAAK,IAAI,GAAG,UAAU,EAAE,OAAO,EAAE,OAAO,UAAU,CAAC,EAAG,KAAIW,EAAEX,EAAE,KAAK,IAAI,GAAG,GAAGA,EAAE,KAAK,IAAI,GAAG,WAAW,GAAGW,EAAE,IAAI,GAAG,UAAU,IAAI,GAAG,OAAO,CAAC,EAAE,EAAE,MAAMX,EAAE,KAAK,EAAE,MAAM,EAAEC,IAAI,OAAOA,GAAG,UAAU5B,EAAE4B,EAAED,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAO3B,EAAE2B,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,OAAOA,CAAC,GAAGwB,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMC,GAAGtD,GAAG,CAACuD,EAAGC,IAAK,EAAE,SAAS3D,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE0B,EAAE,CAAC,IAAIC,EAAE,KAAKC,EAAE,GAAGD,EAAE,KAAK,UAAU,CAAC,IAAI,EAAEA,EAAE,EAAEA,EAAE,IAAI,EAAE,OAAOA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGA,EAAE,EAAEA,EAAE,EAAE,OAAO,IAAIA,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAG,GAAG,EAAE,GAAG,IAAI,CAAC,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAED,KAAKA,EAAE,GAAGC,EAAE,EAAED,EAAEE,GAAGF,EAAE,QAAQG,EAAE,EAAEA,EAAED,EAAE,OAAO,GAAGC,IAAIF,EAAE,GAAGC,EAAE,WAAWC,CAAC,EAAE,EAAEA,GAAGD,EAAE,SAASD,EAAE,EAAEA,EAAE,GAAG,GAAGA,EAAE,IAAI,GAAGA,EAAE,KAAK,CAAC,CAAC,SAAS1B,EAAEyB,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,CAAC,CAAC,SAASF,EAAEC,EAAEC,EAAE,CAAC,IAAIC,EAAE,IAAI5B,EAAE0B,CAAC,EAAEG,EAAEF,GAAGA,EAAE,MAAM,EAAE,UAAU,CAAC,OAAOC,EAAE,KAAK,IAAI,GAAG,UAAU,EAAE,OAAO,EAAE,OAAO,UAAU,CAAC,EAAG,KAAIW,EAAEX,EAAE,KAAK,IAAI,GAAG,GAAGA,EAAE,KAAK,IAAI,GAAG,WAAW,GAAGW,EAAE,IAAI,GAAG,UAAU,IAAI,GAAG,OAAO,CAAC,EAAE,EAAE,MAAMX,EAAE,KAAK,EAAE,MAAM,EAAEC,IAAI,OAAOA,GAAG,UAAU5B,EAAE4B,EAAED,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAO3B,EAAE2B,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,OAAOA,CAAC,GAAG2B,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMC,GAAGzD,GAAG,CAAC0D,EAAGC,IAAK,EAAE,SAAS9D,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE0B,EAAE,CAAC,IAAIC,EAAE,KAAKA,EAAE,KAAK,UAAU,CAAC,IAAIE,EAAEF,EAAE,EAAE,EAAEA,EAAE,EAAEY,EAAE,EAAE,EAAE,OAAOA,EAAEV,EAAE,GAAGU,GAAGA,IAAI,EAAE,EAAEA,EAAEA,GAAG,GAAGA,EAAEV,EAAE,EAAE,EAAE,GAAG,GAAGU,EAAEA,IAAI,GAAGA,EAAEV,EAAE,EAAE,EAAE,GAAG,GAAGU,EAAEA,IAAI,EAAEA,EAAEV,EAAE,EAAE,EAAE,GAAG,GAAGU,EAAEA,GAAG,EAAEA,EAAEV,EAAE,EAAE,EAAE,GAAGU,EAAEA,EAAEA,GAAG,GAAG,GAAGA,EAAEA,GAAG,EAAEV,EAAE,GAAG,EAAEF,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,SAASC,EAAEC,EAAE,EAAE,CAAC,IAAIU,EAAE,EAAE,EAAE,CAAC,EAAE,GAAG,KAAK,EAAE,GAAG,EAAE,EAAE,GAAG,MAAO,KAAI,EAAE,GAAG,EAAEA,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,EAAEA,EAAE,GAAG,EAAEA,EAAE,IAAI,GAAG,EAAE,WAAWA,CAAC,EAAE,EAAEA,EAAE,EAAE,IAAI,GAAG,KAAK,EAAE,OAAO,GAAG,EAAE,KAAK,CAAC,EAAE,IAAIA,EAAE,EAAEA,EAAE,GAAG,EAAEA,KAAK,EAAE,EAAEA,EAAE,CAAC,IAAIA,GAAG,EAAE,EAAE,EAAE,GAAG,GAAG,EAAE,EAAEA,GAAGV,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEU,EAAE,IAAIA,EAAE,EAAE,EAAEA,EAAEV,EAAE,KAAK,CAAC,CAACD,EAAED,EAAED,CAAC,CAAC,CAAC,SAASzB,EAAEyB,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAE,MAAM,EAAEC,EAAE,EAAED,EAAE,EAAEC,CAAC,CAAC,SAASF,EAAEC,EAAEC,EAAE,CAACD,GAAG,OAAOA,EAAE,CAAC,IAAI,MAAM,IAAIE,EAAE,IAAI5B,EAAE0B,CAAC,EAAEG,EAAEF,GAAGA,EAAE,MAAM,EAAE,UAAU,CAAC,OAAOC,EAAE,KAAK,IAAI,GAAG,UAAU,EAAE,OAAO,EAAE,OAAO,UAAU,CAAC,EAAG,KAAIW,EAAEX,EAAE,KAAK,IAAI,GAAG,GAAGA,EAAE,KAAK,IAAI,GAAG,WAAW,GAAGW,EAAE,IAAI,GAAG,UAAU,IAAI,GAAG,OAAO,CAAC,EAAE,EAAE,MAAMX,EAAE,KAAK,EAAE,MAAM,EAAEC,IAAIA,EAAE,GAAG5B,EAAE4B,EAAED,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAO3B,EAAE2B,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,UAAUA,CAAC,GAAG8B,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMC,GAAG5D,GAAG,CAAC6D,EAAGC,IAAK,EAAE,SAASjE,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE0B,EAAE,CAAC,IAAIC,EAAE,KAAKA,EAAE,KAAK,UAAU,CAAC,IAAIE,EAAEF,EAAE,EAAE,EAAEA,EAAE,EAAEY,EAAEZ,EAAE,EAAE,EAAE,EAAE,OAAOA,EAAE,EAAEE,EAAEA,EAAE,WAAW,EAAE,EAAE,EAAEU,EAAE,GAAG,KAAK,EAAE,EAAEA,EAAEA,EAAE,EAAE,KAAK,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,EAAE,EAAEA,GAAG,EAAE,EAAEZ,EAAE,EAAEY,EAAE,GAAGV,EAAEA,IAAI,IAAI,CAAC,EAAE,SAASD,EAAEC,EAAE,EAAE,CAAC,IAAIU,EAAE,EAAE,EAAEG,EAAEkB,EAAEC,EAAE,CAAC,EAAE,EAAE,IAAI,IAAI,KAAK,EAAE,IAAI,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,KAAK,EAAE,EAAE,EAAE,KAAK,IAAI,EAAE,EAAE,MAAM,GAAG,EAAE,EAAEnB,EAAE,IAAIA,EAAE,EAAE,EAAEA,EAAE,IAAI,GAAG,EAAE,YAAYA,EAAE,IAAI,EAAE,MAAM,GAAGA,IAAI,IAAIkB,EAAE,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,GAAG,GAAG,GAAG,EAAE,GAAG,IAAI,GAAGlB,GAAG,IAAIkB,EAAEA,EAAE,WAAW,EAAErB,EAAEsB,EAAEnB,EAAE,MAAM,EAAEkB,EAAE,EAAErB,GAAG,EAAE,EAAE,EAAE,GAAG,IAAI,GAAG,MAAMsB,GAAG,GAAG,EAAE,QAAQ,GAAG,KAAK,IAAI,EAAE,IAAInB,EAAE,EAAE,IAAIA,EAAE,EAAE,EAAEA,EAAE,EAAEmB,EAAE,EAAE,GAAG,KAAKtB,EAAEsB,EAAE,EAAE,EAAE,EAAE,KAAK,GAAG,GAAG,GAAGtB,GAAGA,GAAG,GAAG,GAAG,IAAI,GAAGA,GAAGA,IAAI,GAAGsB,EAAE,GAAG,EAAEtB,EAAEV,EAAE,EAAE+B,EAAE/B,EAAE,EAAEgC,EAAEhC,EAAE,EAAE,CAAC,CAACD,EAAED,EAAED,CAAC,CAAC,CAAC,SAASzB,EAAEyB,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAE,MAAM,EAAEC,CAAC,CAAC,SAASF,EAAEC,EAAEC,EAAE,CAACD,GAAG,OAAOA,EAAE,CAAC,IAAI,MAAM,IAAIE,EAAE,IAAI5B,EAAE0B,CAAC,EAAEG,EAAEF,GAAGA,EAAE,MAAM,EAAE,UAAU,CAAC,OAAOC,EAAE,KAAK,IAAI,GAAG,UAAU,EAAE,OAAO,EAAE,OAAO,UAAU,CAAC,EAAG,KAAIW,EAAEX,EAAE,KAAK,IAAI,GAAG,GAAGA,EAAE,KAAK,IAAI,GAAG,WAAW,GAAGW,EAAE,IAAI,GAAG,UAAU,IAAI,GAAG,OAAO,CAAC,EAAE,EAAE,MAAMX,EAAE,KAAK,EAAE,MAAM,EAAEC,IAAIA,EAAE,GAAG5B,EAAE4B,EAAED,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAO3B,EAAE2B,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,QAAQA,CAAC,GAAGiC,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMG,GAAGjE,GAAG,CAACkE,EAAGC,IAAK,EAAE,SAAStE,EAAEc,EAAEZ,EAAE,CAAC,SAASI,EAAE0B,EAAE,CAAC,IAAIC,EAAE,KAAKC,EAAE,GAAGD,EAAE,KAAK,UAAU,CAAC,IAAI,EAAEA,EAAE,EAAEY,EAAEZ,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAE,GAAG,GAAG,IAAI,EAAEY,EAAEA,EAAEA,EAAE,EAAE,EAAE,EAAE,GAAG,GAAG,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,EAAEZ,EAAE,EAAE,EAAE,GAAG,GAAG,IAAI,GAAGY,EAAEZ,EAAE,EAAEY,EAAEA,EAAE,EAAE,EAAEZ,EAAE,EAAE,GAAG,GAAGY,IAAI,GAAG,EAAEZ,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,EAAE,YAAYA,EAAE,EAAE,WAAWD,IAAI,KAAK,MAAMA,CAAC,GAAGC,EAAE,EAAED,EAAE,WAAW,EAAEC,EAAE,EAAED,EAAE,GAAGE,GAAGF,EAAE,QAAQG,EAAE,EAAEA,EAAED,EAAE,OAAO,GAAGC,IAAIF,EAAE,GAAGC,EAAE,WAAWC,CAAC,EAAE,EAAEF,EAAE,KAAK,CAAC,CAAC,SAAS1B,EAAEyB,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,CAAC,CAAC,SAASF,EAAEC,EAAEC,EAAE,CAAC,IAAIC,EAAE,IAAI5B,EAAE0B,CAAC,EAAEG,EAAEF,GAAGA,EAAE,MAAM,EAAE,UAAU,CAAC,OAAOC,EAAE,KAAK,IAAI,GAAG,UAAU,EAAE,OAAO,EAAE,OAAO,UAAU,CAAC,EAAG,KAAIW,EAAEX,EAAE,KAAK,IAAI,GAAG,GAAGA,EAAE,KAAK,IAAI,GAAG,WAAW,GAAGW,EAAE,IAAI,GAAG,UAAU,IAAI,GAAG,OAAO,CAAC,EAAE,EAAE,MAAMX,EAAE,KAAK,EAAE,MAAM,EAAEC,IAAI,OAAOA,GAAG,UAAU5B,EAAE4B,EAAED,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAO3B,EAAE2B,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,CAACpB,GAAGA,EAAE,QAAQA,EAAE,QAAQiB,EAAE7B,GAAGA,EAAE,IAAIA,EAAE,UAAU,CAAC,OAAO6B,CAAC,CAAC,EAAE,KAAK,OAAOA,CAAC,GAAGsC,EAAG,OAAOC,GAAI,UAAUA,EAAG,OAAO,QAAQ,YAAY,MAAM,CAAC,CAAC,EAAMC,GAAGpE,GAAG,IAAI,CAAC,CAAC,EAAMqE,GAAGrE,GAAG,CAACsE,EAAGC,IAAK,EAAE,SAAS1E,EAAEc,EAAEZ,EAAE,CAAC,IAAII,EAAE,IAAIC,EAAE,EAAEwB,EAAE,GAAGC,EAAE,SAASC,EAAE/B,EAAE,IAAII,EAAEC,CAAC,EAAE2B,EAAEhC,EAAE,IAAI,EAAE6B,CAAC,EAAEI,EAAED,EAAE,EAAE,EAAE5B,EAAE,EAAEuC,EAAE,SAAS,EAAE8B,EAAEC,EAAEC,EAAE,CAAC,IAAIC,EAAE,CAAC,EAAEF,EAAEA,GAAG,GAAG,CAAC,QAAQ,EAAE,EAAEA,GAAG,CAAC,EAAE,IAAI,EAAET,EAAED,EAAEU,EAAE,QAAQ,CAACD,EAAEI,EAAEjE,CAAC,CAAC,EAAE6D,GAAG,KAAK,EAAE,EAAEA,EAAE,CAAC,EAAEG,CAAC,EAAEE,EAAE,IAAI,EAAEF,CAAC,EAAE,EAAE,UAAU,CAAC,QAAQG,EAAED,EAAE,EAAEzE,CAAC,EAAE2E,EAAEjD,EAAEkD,EAAE,EAAEF,EAAE/C,GAAG+C,GAAGA,EAAEE,GAAG7E,EAAE4E,GAAG5E,EAAE6E,EAAEH,EAAE,EAAE,CAAC,EAAE,KAAKC,GAAG9C,GAAG8C,GAAG,EAAEC,GAAG,EAAEC,KAAK,EAAE,OAAOF,EAAEE,GAAGD,CAAC,EAAE,OAAO,EAAE,MAAM,UAAU,CAAC,OAAOF,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,MAAM,UAAU,CAAC,OAAOA,EAAE,EAAE,CAAC,EAAE,UAAU,EAAE,EAAE,OAAO,EAAEb,EAAEY,EAAEC,EAAE,CAAC,EAAElE,CAAC,GAAG8D,EAAE,MAAMC,GAAG,SAASI,EAAEC,EAAEC,EAAEC,EAAE,CAAC,OAAOA,IAAIA,EAAE,GAAGpC,EAAEoC,EAAEJ,CAAC,EAAEC,EAAE,MAAM,UAAU,CAAC,OAAOjC,EAAEgC,EAAE,CAAC,CAAC,CAAC,GAAGG,GAAGjF,EAAE8B,GAAGiD,EAAEC,GAAGD,CAAC,GAAG,EAAE,EAAE,WAAWL,EAAEA,EAAE,OAAO,MAAM1E,EAAE0E,EAAE,KAAK,CAAC,CAAC,SAAS,EAAED,EAAE,CAAC,IAAIC,EAAEC,EAAEF,EAAE,OAAOG,EAAE,KAAK,EAAE,EAAEE,EAAEF,EAAE,EAAEA,EAAE,EAAE,EAAE,EAAEA,EAAE,EAAE,CAAC,EAAE,IAAID,IAAIF,EAAE,CAACE,GAAG,GAAG,EAAEvE,GAAG,EAAE,GAAG,IAAI,IAAI,EAAE,EAAE,EAAEA,EAAE,IAAI,EAAE,GAAG,EAAE0E,EAAE,EAAEA,EAAEL,EAAE,EAAEE,IAAID,EAAE,EAAE,KAAK,EAAEI,GAAGJ,GAAGE,EAAE,EAAE,SAASG,EAAE,CAAC,QAAQC,EAAEC,EAAE,EAAEC,EAAEN,EAAE,EAAEO,EAAEP,EAAE,EAAEQ,EAAER,EAAE,EAAEG,KAAKC,EAAEI,EAAEF,EAAE,EAAEA,EAAE,GAAGD,EAAEA,EAAE7E,EAAEgF,EAAE,GAAGA,EAAEF,GAAGE,EAAED,EAAE,EAAEA,EAAEH,KAAKI,EAAED,GAAGH,IAAI,OAAOJ,EAAE,EAAEM,EAAEN,EAAE,EAAEO,EAAEF,CAAC,GAAG7E,CAAC,CAAC,CAAC,SAAS0C,EAAE2B,EAAEC,EAAE,CAAC,OAAOA,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAEC,EAAE,EAAED,EAAE,EAAE,MAAM,EAAEC,CAAC,CAAC,SAASV,EAAES,EAAEC,EAAE,CAAC,IAAIC,EAAE,CAAC,EAAEC,EAAE,OAAOH,EAAE,EAAE,GAAGC,GAAGE,GAAG,SAAS,IAAI,KAAKH,EAAE,GAAG,CAACE,EAAE,KAAKX,EAAES,EAAE,GAAGC,EAAE,CAAC,CAAC,CAAC,OAAOI,EAAN,CAAS,CAAC,OAAOH,EAAE,OAAOA,EAAEC,GAAG,SAASH,EAAEA,EAAE,IAAI,CAAC,SAASR,EAAEQ,EAAEC,EAAE,CAAC,QAAQC,EAAEF,EAAE,GAAGG,EAAE,EAAE,EAAE,EAAED,EAAE,QAAQD,EAAE,EAAE,GAAG,GAAGE,GAAGF,EAAE,EAAE,GAAG,IAAIC,EAAE,WAAW,GAAG,EAAE,OAAOE,EAAEH,CAAC,CAAC,CAAC,SAAS,GAAG,CAAC,GAAG,CAAC,IAAID,EAAE,OAAO9B,IAAI8B,EAAE9B,EAAE,aAAa8B,EAAEA,EAAErE,CAAC,GAAGqE,EAAE,IAAI,WAAWrE,CAAC,GAAGN,EAAE,QAAQA,EAAE,UAAU,gBAAgB2E,CAAC,GAAGI,EAAEJ,CAAC,CAAC,OAAOG,EAAN,CAAS,IAAIF,EAAE5E,EAAE,UAAU6E,EAAED,GAAGA,EAAE,QAAQ,MAAM,CAAC,CAAC,IAAI,KAAK5E,EAAE6E,EAAE7E,EAAE,OAAO+E,EAAEjE,CAAC,CAAC,CAAC,CAAC,CAAC,SAASiE,EAAEJ,EAAE,CAAC,OAAO,OAAO,aAAa,MAAM,EAAEA,CAAC,CAAC,CAAC,GAAGR,EAAEjE,EAAE,OAAO,EAAEY,CAAC,EAAE,OAAO4D,GAAI,UAAUA,EAAG,QAAQ,CAACA,EAAG,QAAQ,EAAE,GAAG,CAAC7B,EAAE0B,GAAG,CAAC,OAAOI,EAAN,CAAS,CAAC,MAAM,OAAO,QAAQ,YAAY,OAAO,IAAI,OAAO,UAAU,CAAC,OAAO,CAAC,CAAC,EAAEzE,EAAE,OAAO8B,GAAG,CAAC,GAAG,OAAO,MAAM,YAAY,KAAKyC,EAAG,CAAC,EAAE,IAAI,CAAC,CAAC,EAAMc,GAAGpF,GAAG,CAACqF,EAAIC,IAAK,CAAC,IAAIC,EAAGvC,GAAG,EAAEwC,EAAGrC,GAAG,EAAEsC,EAAGnC,GAAG,EAAEoC,EAAGjC,GAAG,EAAEkC,EAAG/B,GAAG,EAAEgC,EAAG3B,GAAG,EAAE4B,EAAGxB,GAAG,EAAEwB,EAAG,KAAKN,EAAGM,EAAG,OAAOL,EAAGK,EAAG,OAAOJ,EAAGI,EAAG,UAAUH,EAAGG,EAAG,QAAQF,EAAGE,EAAG,OAAOD,EAAGN,EAAG,QAAQO,CAAE,CAAC,EAAMC,GAAG9F,GAAG,IAAI,CAAC,CAAC,EAAM+F,GAAG/F,GAAG,IAAI,CAAC,CAAC,EAAMgG,GAAGhG,GAAG,IAAI,CAAC,CAAC,EAAMiG,GAAGjG,GAAG,IAAI,CAAC,CAAC,EAAMkG,GAAGlG,GAAG,IAAI,CAAC,CAAC,EAAMmG,GAAGnG,GAAG,IAAI,CAAC,CAAC,EAAMoG,GAAGpG,GAAG,CAACqG,EAAGC,IAAK,CAAC,IAAIC,GAAI,IAAI,CAAC,IAAI1G,EAAE,OAAO,UAAU,aAAa,SAAS,cAAc,SAAS,cAAc,IAAI,OAAO,OAAO,OAAO,YAAY,cAAcA,EAAEA,GAAG,YAAY,SAASc,EAAE,CAACA,EAAEA,GAAG,CAAC,EAAE,SAASZ,GAAG,CAAC,OAAOyG,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEG,EAAE,CAAC,SAASxG,GAAG,CAAC,OAAOqG,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEI,EAAE,CAAC,SAAS,GAAG,CAAC,OAAOJ,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEK,EAAE,CAAC,SAASjF,GAAG,CAAC,OAAO4E,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEM,EAAE,CAAC,SAASjF,GAAG,CAAC,OAAO2E,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEO,EAAE,CAAC,SAASjF,GAAG,CAAC,OAAO0E,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAEQ,EAAE,CAAC,SAASjF,GAAG,CAAC,OAAOyE,GAAG,QAAQC,IAAIC,GAAGF,GAAG,MAAM,EAAES,EAAE,CAAC,IAAIjF,EAAE,OAAOrB,GAAG,YAAYA,EAAE,CAAC,EAAE8B,EAAEC,EAAEV,EAAE,MAAM,IAAI,QAAQ,SAASkF,EAAEC,EAAE,CAAC1E,EAAEyE,EAAExE,EAAEyE,CAAC,CAAC,EAAE,IAAIxE,EAAE,OAAO,SAAS,aAAa,QAAQ,YAAYA,EAAE,CAAC,kBAAkB,QAAQ,UAAU,mBAAmB,EAAE,mBAAmB,QAAQ,UAAU,oBAAoB,CAAC,GAAG,IAAIC,EAAE,OAAO,OAAO,CAAC,EAAEZ,CAAC,EAAEa,EAAE,CAAC,EAAEkB,EAAE,iBAAiBC,EAAE,CAACkD,EAAEC,IAAI,CAAC,MAAMA,CAAC,EAAEC,EAAE,OAAO,QAAQ,SAASxC,EAAE,OAAO,eAAe,WAAWJ,EAAE,OAAO,SAAS,UAAU,OAAO,QAAQ,UAAU,UAAU,OAAO,QAAQ,SAAS,MAAM,SAAS,EAAExC,EAAE,wBAAwB,GAAG0C,EAAE,GAAG,SAASC,EAAEuC,EAAE,CAAC,OAAOlF,EAAE,WAAWA,EAAE,WAAWkF,EAAExC,CAAC,EAAEA,EAAEwC,CAAC,CAAC,IAAIG,EAAExC,EAAEyC,EAAExC,EAAE,SAASC,EAAEmC,EAAE,CAAIA,aAAaK,IAAUC,EAAE,6BAA6BN,CAAC,CAAC,CAAC,GAAG1C,EAAE,CAACI,EAAEF,EAAEsB,GAAG,EAAE,QAAQtB,CAAC,EAAE,IAAIA,EAAE,UAAU,IAAI,IAAIM,EAAEC,EAAE,OAAOrF,IAAI,aAAaoF,EAAEe,GAAG,EAAEd,EAAEe,GAAG,GAAGqB,EAAE,CAACF,EAAEM,MAAMN,EAAElC,EAAE,UAAUkC,CAAC,EAAEnC,EAAE,aAAamC,EAAEM,GAAG,OAAO,MAAM,GAAGH,EAAEH,GAAG,CAAC,IAAIM,GAAGJ,EAAEF,EAAE,EAAE,EAAE,OAAOM,GAAG,SAASA,GAAG,IAAI,WAAWA,EAAE,GAAGA,EAAE,EAAE5C,EAAE,CAACsC,EAAEM,GAAGC,KAAK,CAACP,EAAElC,EAAE,UAAUkC,CAAC,EAAEnC,EAAE,SAASmC,EAAE,SAASQ,GAAGC,GAAG,CAACD,GAAGD,GAAGC,EAAE,EAAEF,GAAGG,GAAG,MAAM,CAAC,CAAC,CAAC,EAAE,QAAQ,KAAK,OAAO,IAAI7D,EAAE,QAAQ,KAAK,GAAG,QAAQ,MAAM,GAAG,GAAGlB,EAAE,QAAQ,KAAK,MAAM,CAAC,EAAE,QAAQ,GAAG,oBAAoB,SAASsE,EAAE,CAAC,GAAG,EAAEA,aAAaI,IAAI,MAAMJ,CAAC,CAAC,EAAE,QAAQ,GAAG,qBAAqB,SAASA,EAAE,CAAC,MAAMA,CAAC,CAAC,EAAEnD,EAAE,CAACmD,EAAEM,KAAK,CAAC,GAAGI,GAAG,EAAE,MAAM,QAAQ,SAASV,EAAEM,GAAG1C,EAAE0C,EAAE,EAAE,QAAQ,KAAKN,CAAC,CAAC,EAAEnF,EAAE,QAAQ,UAAU,CAAC,MAAM,4BAA4B,EAAE,IAAIkF,EAAE,GAAG,CAACA,EAAEjB,GAAG,CAAC,OAAOkB,EAAN,CAAS,MAAM,QAAQ,MAAM,yGAAyG,EAAEA,CAAC,CAAC,OAAO,OAAOD,EAAE,MAAM,MAAME,GAAGxC,KAAKA,EAAEF,EAAE,KAAK,SAAS,KAAK,OAAO,UAAU,aAAa,SAAS,gBAAgBA,EAAE,SAAS,cAAc,KAAK,OAAO7E,GAAG,aAAaA,IAAI6E,EAAE7E,GAAG6E,EAAE,QAAQ,OAAO,IAAI,EAAEA,EAAEA,EAAE,OAAO,EAAEA,EAAE,QAAQ,SAAS,EAAE,EAAE,YAAY,GAAG,EAAE,CAAC,EAAEA,EAAE,GAAGF,IAAI6C,EAAEH,GAAG,CAAC,IAAIC,EAAE,IAAI,eAAe,OAAOA,EAAE,KAAK,MAAMD,EAAE,EAAE,EAAEC,EAAE,KAAK,IAAI,EAAEA,EAAE,YAAY,EAAEvC,IAAI0C,EAAEJ,GAAG,CAAC,IAAIC,EAAE,IAAI,eAAe,OAAOA,EAAE,KAAK,MAAMD,EAAE,EAAE,EAAEC,EAAE,aAAa,cAAcA,EAAE,KAAK,IAAI,EAAE,IAAI,WAAWA,EAAE,QAAQ,CAAC,GAAGtC,EAAE,CAACqC,EAAEC,EAAEM,KAAK,CAAC,IAAIC,GAAG,IAAI,eAAeA,GAAG,KAAK,MAAMR,EAAE,EAAE,EAAEQ,GAAG,aAAa,cAAcA,GAAG,OAAO,IAAI,CAAC,GAAGA,GAAG,QAAQ,KAAKA,GAAG,QAAQ,GAAGA,GAAG,SAAS,CAACP,EAAEO,GAAG,QAAQ,EAAE,MAAM,CAACD,GAAG,CAAC,EAAEC,GAAG,QAAQD,GAAGC,GAAG,KAAK,IAAI,CAAC,GAAG5C,EAAEoC,GAAG,SAAS,MAAMA,GAAG1C,GAAG,OAAO,aAAa,cAAc,OAAO,YAAY0B,GAAG,EAAE,aAAa,IAAI,EAAE,QAAQ,IAAI,KAAK,OAAO,EAAEf,EAAE,QAAQ,KAAK,KAAK,OAAO,EAAEX,IAAI,EAAE0C,GAAGlC,EAAE,UAAU,EAAEkC,EAAE;AAAA,CAC94oB,EAAE/B,EAAE+B,GAAGlC,EAAE,UAAU,EAAEkC,EAAE;AAAA,CACvB,GAAG,IAAIY,EAAE9F,EAAE,OAAO,EAAEwF,EAAExF,EAAE,UAAUmD,EAAE,OAAO,OAAOnD,EAAEY,CAAC,EAAEA,EAAE,KAAKZ,EAAE,YAAYa,EAAEb,EAAE,WAAWA,EAAE,cAAc+B,EAAE/B,EAAE,aAAaA,EAAE,OAAOgC,EAAEhC,EAAE,MAAM,IAAI+F,GAAE,EAAEC,GAAG,QAAQ,KAAKC,GAAG,QAAQ,MAAMC,GAAG,QAAQ,gBAAgBC,GAAGnG,EAAE,aAAamG,GAAGnG,EAAE,YAAY,IAAIoG,GAAGpG,EAAE,eAAe,GAAG,OAAO,aAAa,UAAUqG,GAAG,iCAAiC,EAAE,IAAI7B,GAAG8B,GAAGC,GAAG,GAAGC,GAAG,SAASC,GAAGvB,EAAEC,EAAE,CAACD,GAAGmB,GAAGlB,CAAC,CAAC,CAAC,IAAIuB,GAAG,OAAO,aAAa,YAAY,IAAI,YAAY,MAAM,EAAE,OAAO,SAASC,GAAGzB,EAAEC,EAAEM,GAAG,CAAC,QAAQC,GAAGP,EAAEM,GAAGE,GAAGR,EAAED,EAAES,KAAK,EAAEA,IAAID,KAAK,EAAEC,GAAG,GAAGA,GAAGR,EAAE,IAAID,EAAE,QAAQwB,GAAG,OAAOA,GAAG,OAAOxB,EAAE,kBAAkB,kBAAkBA,EAAE,MAAMC,EAAEQ,EAAE,EAAET,EAAE,SAASC,EAAEQ,EAAE,CAAC,EAAE,QAAQC,GAAG,GAAGT,EAAEQ,IAAI,CAAC,IAAIiB,GAAG1B,EAAEC,KAAK,GAAG,EAAEyB,GAAG,KAAK,CAAChB,IAAI,OAAO,aAAagB,EAAE,EAAE,QAAQ,CAAC,IAAIC,GAAG3B,EAAEC,KAAK,GAAG,IAAIyB,GAAG,MAAM,IAAI,CAAChB,IAAI,OAAO,cAAcgB,GAAG,KAAK,EAAEC,EAAE,EAAE,QAAQ,CAAC,IAAIC,GAAG5B,EAAEC,KAAK,GAAG,IAAIyB,GAAG,MAAM,IAAIA,IAAIA,GAAG,KAAK,GAAGC,IAAI,EAAEC,GAAGF,IAAIA,GAAG,IAAI,GAAGC,IAAI,GAAGC,IAAI,EAAE5B,EAAEC,KAAK,GAAGyB,GAAG,MAAMhB,IAAI,OAAO,aAAagB,EAAE,MAAM,CAAC,IAAIG,GAAGH,GAAG,MAAMhB,IAAI,OAAO,aAAa,MAAMmB,IAAI,GAAG,MAAMA,GAAG,IAAI,CAAC,CAAC,CAAC,OAAOnB,EAAE,CAAC,SAASoB,GAAG9B,EAAEC,EAAE,CAAC,OAAOD,EAAEyB,GAAGxI,EAAE,EAAE+G,EAAEC,CAAC,EAAE,EAAE,CAAC,SAAS8B,GAAG/B,EAAEC,EAAEM,GAAGC,GAAG,CAAC,GAAG,EAAEA,GAAG,GAAG,MAAO,GAAE,QAAQC,GAAGF,GAAGG,GAAGH,GAAGC,GAAG,EAAEkB,GAAG,EAAEA,GAAG1B,EAAE,OAAO,EAAE0B,GAAG,CAAC,IAAIC,GAAG3B,EAAE,WAAW0B,EAAE,EAAE,GAAGC,IAAI,OAAOA,IAAI,MAAM,CAAC,IAAIC,GAAG5B,EAAE,WAAW,EAAE0B,EAAE,EAAEC,GAAG,QAAQA,GAAG,OAAO,IAAIC,GAAG,IAAI,CAAC,GAAGD,IAAI,IAAI,CAAC,GAAGpB,IAAIG,GAAG,MAAMT,EAAEM,MAAMoB,EAAE,SAASA,IAAI,KAAK,CAAC,GAAGpB,GAAG,GAAGG,GAAG,MAAMT,EAAEM,MAAM,IAAIoB,IAAI,EAAE1B,EAAEM,MAAM,IAAIoB,GAAG,EAAE,SAASA,IAAI,MAAM,CAAC,GAAGpB,GAAG,GAAGG,GAAG,MAAMT,EAAEM,MAAM,IAAIoB,IAAI,GAAG1B,EAAEM,MAAM,IAAIoB,IAAI,EAAE,GAAG1B,EAAEM,MAAM,IAAIoB,GAAG,EAAE,KAAK,CAAC,GAAGpB,GAAG,GAAGG,GAAG,MAAMT,EAAEM,MAAM,IAAIoB,IAAI,GAAG1B,EAAEM,MAAM,IAAIoB,IAAI,GAAG,GAAG1B,EAAEM,MAAM,IAAIoB,IAAI,EAAE,GAAG1B,EAAEM,MAAM,IAAIoB,GAAG,EAAE,CAAC,CAAC,OAAO1B,EAAEM,IAAI,EAAEA,GAAGE,EAAE,CAAC,SAASuB,GAAGhC,EAAEC,EAAEM,GAAG,CAAC,OAAOwB,GAAG/B,EAAE/G,EAAE,EAAEgH,EAAEM,EAAE,CAAC,CAAC,IAAIhB,GAAGE,GAAGC,GAAGC,GAAGsC,GAAGrC,GAAGC,GAAGC,GAAGC,GAAG,IAAIR,GAAGzE,EAAE,QAAQ,SAAS0E,GAAGQ,EAAE,CAACT,GAAGS,EAAElF,EAAE,MAAM2E,GAAG,IAAI,UAAUO,CAAC,EAAElF,EAAE,OAAO6E,GAAG,IAAI,WAAWK,CAAC,EAAElF,EAAE,OAAO8E,GAAG,IAAI,WAAWI,CAAC,EAAElF,EAAE,OAAO4E,GAAG,IAAI,WAAWM,CAAC,EAAElF,EAAE,QAAQmH,GAAG,IAAI,YAAYjC,CAAC,EAAElF,EAAE,QAAQ+E,GAAG,IAAI,YAAYG,CAAC,EAAElF,EAAE,QAAQgF,GAAG,IAAI,aAAaE,CAAC,EAAElF,EAAE,QAAQiF,GAAG,IAAI,aAAaC,CAAC,CAAC,CAAC,IAAIkC,GAAGpH,EAAE,gBAAgB,SAAS,GAAG,EAAEwE,GAAGxE,EAAE,WAAWyE,GAAGzE,EAAE,eAAeA,EAAE,WAAWwE,GAAGxE,EAAE,mBAAmBwE,GAAG,IAAI,YAAY,OAAO,CAAC,QAAQ4C,GAAG,MAAM,QAAQ,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE5C,GAAG,kBAAkB,mBAAmB,MAAMgB,EAAE,6NAA6N,EAAEhD,GAAG,QAAQ,IAAI,mHAAmH,EAAE,MAAM,YAAY,EAAEgC,KAAKC,GAAGD,GAAG,QAAQ4C,GAAG3C,GAAG,WAAWC,GAAGD,EAAE,EAAE,IAAI4C,GAAGC,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEC,GAAG,GAAG,SAAS5B,IAAI,CAAC,OAAOO,EAAE,CAAC,SAASsB,IAAI,CAAC,GAAG1H,EAAE,OAAO,IAAI,OAAOA,EAAE,QAAQ,aAAaA,EAAE,OAAO,CAACA,EAAE,MAAM,GAAGA,EAAE,OAAO,QAAQ2H,GAAG3H,EAAE,OAAO,MAAM,CAAC,EAAE4H,GAAGN,EAAE,CAAC,CAAC,SAASO,IAAI,CAACJ,GAAG,GAAG,CAAC,GAAGG,GAAGL,EAAE,CAAC,CAAC,SAASO,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG9H,EAAE,QAAQ,IAAI,OAAOA,EAAE,SAAS,aAAaA,EAAE,QAAQ,CAACA,EAAE,OAAO,GAAGA,EAAE,QAAQ,QAAQ+H,GAAG/H,EAAE,QAAQ,MAAM,CAAC,EAAE4H,GAAGJ,EAAE,CAAC,CAAC,CAAC,SAASG,GAAGzC,EAAE,CAACoC,GAAG,QAAQpC,CAAC,CAAC,CAAC,SAAS8C,GAAG9C,EAAE,CAACqC,GAAG,QAAQrC,CAAC,CAAC,CAAC,SAAS6C,GAAG7C,EAAE,CAACsC,GAAG,QAAQtC,CAAC,CAAC,CAAC,IAAI+C,GAAG,EAAEC,GAAG,KAAKC,GAAG,KAAK,SAASC,GAAGlD,EAAE,CAAC+C,KAAKjI,EAAE,wBAAwBA,EAAE,uBAAuBiI,EAAE,CAAC,CAAC,SAASI,GAAGnD,EAAE,CAAC,GAAG+C,KAAKjI,EAAE,wBAAwBA,EAAE,uBAAuBiI,EAAE,EAAEA,IAAI,IAAIC,KAAK,OAAO,cAAcA,EAAE,EAAEA,GAAG,MAAMC,IAAI,CAAC,IAAIhD,EAAEgD,GAAGA,GAAG,KAAKhD,EAAE,CAAC,CAAC,CAAC,SAASkB,GAAGnB,EAAE,CAAC,EAAE,YAAY,CAAC,IAAI,UAAU,IAAIA,CAAC,CAAC,EAAElF,EAAE,SAASA,EAAE,QAAQkF,CAAC,EAAEA,EAAE,WAAWA,EAAE,IAAIM,EAAEN,CAAC,EAAEqB,GAAG,GAAGC,GAAG,EAAEtB,GAAG,2CAA2C,IAAIC,EAAE,IAAI,YAAY,aAAaD,CAAC,EAAE,MAAMxE,EAAEyE,CAAC,EAAEA,CAAC,CAAC,IAAImD,GAAG,wCAAwC,SAASC,GAAGrD,EAAE,CAAC,OAAOA,EAAE,WAAWoD,EAAE,CAAC,CAAC,SAASE,GAAGtD,EAAE,CAAC,OAAOA,EAAE,WAAW,SAAS,CAAC,CAAC,IAAIuD,GAAGA,GAAG,uCAAuCF,GAAGE,EAAE,IAAIA,GAAG9F,EAAE8F,EAAE,GAAG,SAASC,GAAGxD,EAAE,CAAC,GAAG,CAAC,GAAGA,GAAGuD,IAAItC,GAAG,OAAO,IAAI,WAAWA,EAAE,EAAE,GAAGb,EAAE,OAAOA,EAAEJ,CAAC,EAAE,KAAK,iDAAiD,OAAOC,EAAN,CAASkB,GAAGlB,CAAC,CAAC,CAAC,CAAC,SAASwD,IAAI,CAAC,GAAG,CAACxC,KAAKf,GAAGxC,GAAG,CAAC,GAAG,OAAO,OAAO,YAAY,CAAC4F,GAAGC,EAAE,EAAE,OAAO,MAAMA,GAAG,CAAC,YAAY,aAAa,CAAC,EAAE,KAAK,SAASvD,EAAE,CAAC,GAAG,CAACA,EAAE,GAAG,KAAK,uCAAuCuD,GAAG,IAAI,OAAOvD,EAAE,YAAY,CAAC,CAAC,EAAE,MAAM,UAAU,CAAC,OAAOwD,GAAGD,EAAE,CAAC,CAAC,EAAE,GAAG5F,EAAE,OAAO,IAAI,QAAQ,SAASqC,EAAEC,EAAE,CAACtC,EAAE4F,GAAG,SAAShD,GAAG,CAACP,EAAE,IAAI,WAAWO,EAAE,CAAC,CAAC,EAAEN,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,QAAQ,QAAQ,EAAE,KAAK,UAAU,CAAC,OAAOuD,GAAGD,EAAE,CAAC,CAAC,CAAC,CAAC,SAASG,IAAI,CAAC,IAAI1D,EAAE,CAAC,IAAI2D,GAAG,uBAAuBA,EAAE,EAAE,SAAS1D,EAAEyB,GAAGC,GAAG,CAAC,IAAIC,GAAGF,GAAG,QAAQ,GAAG5G,EAAE,IAAI8G,GAAGgC,GAAG9I,EAAE,IAAI,oBAAoB,EAAEqH,GAAGrH,EAAE,IAAI,0BAA0BgI,GAAGhI,EAAE,IAAI,iBAAiB,EAAEsG,GAAGO,GAAG,CAAC,EAAE,CAAC,IAAIE,GAAGgC,GAAG,cAAc,OAAOA,GAAG,cAAc,QAAQ,SAASC,GAAG,CAACD,GAAG,uBAAuBC,GAAG,UAAU,CAAC,EAAEjC,IAAIsB,GAAG,kBAAkB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAGD,GAAG,kBAAkB,EAAE,SAAS3C,GAAGmB,GAAG,CAACzB,EAAEyB,GAAG,SAASA,GAAG,MAAM,CAAC,CAAC,SAASlB,GAAGkB,GAAG,CAAC,OAAO+B,GAAG,EAAE,KAAK,SAAS9B,GAAG,CAAC,OAAO,YAAY,YAAYA,GAAG3B,CAAC,CAAC,CAAC,EAAE,KAAK,SAAS2B,GAAG,CAAC,OAAOA,EAAE,CAAC,EAAE,KAAKD,GAAG,SAASC,GAAG,CAACrB,EAAE,0CAA0CqB,EAAE,EAAER,GAAGQ,EAAE,CAAC,CAAC,CAAC,CAAC,SAASlB,IAAI,CAAC,MAAM,CAACQ,IAAI,OAAO,YAAY,sBAAsB,YAAY,CAACoC,GAAGE,EAAE,GAAG,CAACD,GAAGC,EAAE,GAAG,CAACjG,GAAG,OAAO,OAAO,WAAW,MAAMiG,GAAG,CAAC,YAAY,aAAa,CAAC,EAAE,KAAK,SAAS7B,GAAG,CAAC,IAAIC,GAAG,YAAY,qBAAqBD,GAAG1B,CAAC,EAAE,OAAO2B,GAAG,KAAKpB,GAAG,SAASqB,GAAG,CAAC,OAAOtB,EAAE,kCAAkCsB,EAAE,EAAEtB,EAAE,2CAA2C,EAAEE,GAAGD,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEC,GAAGD,EAAE,CAAC,CAAC,GAAGzF,EAAE,gBAAgB,GAAG,CAAC,IAAI4F,GAAG5F,EAAE,gBAAgBkF,EAAEC,CAAC,EAAE,OAAOS,EAAE,OAAOgB,GAAN,CAAUpB,EAAE,sDAAsDoB,EAAE,EAAElG,EAAEkG,EAAE,CAAC,CAAC,OAAOjB,GAAG,EAAE,MAAMjF,CAAC,EAAE,CAAC,CAAC,CAAC,IAAIuI,GAAGC,GAAGC,GAAG,CAAC,EAAE,SAAS5D,GAAGL,EAAE,CAAC,KAAK,KAAK,aAAa,KAAK,QAAQ,gCAAgCA,EAAE,IAAI,KAAK,OAAOA,CAAC,CAAC,SAASkE,GAAGlE,EAAE,CAAC,IAAIC,EAAE4D,GAAG,SAAS7D,GAAG,OAAO6D,GAAG,SAAS7D,GAAGC,EAAE,UAAU,EAAEkE,GAAGnE,CAAC,EAAE6D,GAAG,eAAe,OAAOA,GAAG,eAAe,QAAQ5D,CAAC,EAAE,CAAC,EAAEA,EAAE,YAAY,CAAC,CAAC,SAASmE,GAAGpE,EAAE,CAAC,IAAIC,EAAE4D,GAAG,SAAS7D,GAAGC,EAAE,YAAY,CAAC,IAAI,QAAQ,CAAC,CAAC,CAAC,SAASoE,GAAGrE,EAAE,CAAC,IAAIC,EAAE4D,GAAG,SAAS7D,GAAGuB,GAAGtB,CAAC,EAAE4D,GAAG,mBAAmB5D,CAAC,CAAC,CAAC,SAASqE,GAAGtE,EAAE,CAAC,IAAIC,EAAE4D,GAAG,aAAa,EAAE,GAAG,CAAC5D,EAAE,MAAO,GAAE4D,GAAG,eAAe,KAAK5D,CAAC,EAAE4D,GAAG,SAAS7D,EAAE,aAAaC,EAAEA,EAAE,YAAYD,EAAE,YAAY,IAAIO,GAAG,CAAC,IAAI,MAAM,cAAcP,EAAE,aAAa,IAAIA,EAAE,IAAI,YAAYA,EAAE,WAAW,EAAE,OAAOC,EAAE,WAAW,IAAI,CAACM,GAAG,KAAK,YAAY,IAAI,EAAEN,EAAE,YAAYM,GAAGP,EAAE,YAAY,CAAC,EAAEC,EAAE,SAASA,EAAE,WAAW,EAAE,OAAOA,EAAE,YAAY,CAAC,CAAC,IAAIsE,GAAG,CAAC,QAAQ,OAAO,IAAI,UAAU,CAACA,GAAG,SAAS,EAAE,IAAIvE,EAAEtF,EAAE,EAAE6J,GAAG,QAAQ,GAAG,GAAG,OAAOvE,CAAC,EAAE,OAAO,SAASA,EAAE,CAAC,IAAIC,EAAE6B,GAAG9B,CAAC,EAAE,OAAOC,CAAC,CAAC,EAAE,SAASuE,GAAGxE,EAAE,CAAC,GAAG,EAAE,OAAOyE,GAAG,EAAE,EAAEzE,CAAC,EAAEsB,GAAGtB,EAAEW,GAAG,IAAIkD,GAAG,oBAAoB,EAAE/I,EAAE,QAAQA,EAAE,OAAOkF,CAAC,EAAEqB,GAAG,IAAIvE,EAAEkD,EAAE,IAAIK,GAAGL,CAAC,CAAC,CAAC,CAAC,SAAS0E,GAAG1E,EAAEC,EAAE,CAAC,GAAGqB,GAAGtB,EAAE,CAACC,GAAG,EAAE,MAAM0E,GAAG3E,CAAC,EAAE,SAASwE,GAAGxE,CAAC,CAAC,CAAC,IAAI4E,GAAGF,GAAG,SAASG,GAAG7E,EAAE,CAAC,GAAGA,aAAaK,IAAIL,GAAG,SAAS,OAAOsB,GAAGxE,EAAE,EAAEkD,CAAC,CAAC,CAAC,IAAI6D,GAAG,CAAC,cAAc,CAAC,EAAE,eAAe,CAAC,EAAE,iBAAiB,CAAC,EAAE,SAAS,CAAC,EAAE,KAAK,UAAU,CAAC,EAAEA,GAAG,WAAW,EAAEA,GAAG,eAAe,CAAC,EAAE,eAAe,UAAU,CAAC,QAAQ7D,EAAE,EAAEA,KAAK6D,GAAG,qBAAqB,CAAC,EAAE,WAAW,UAAU,CAAC3C,GAAG,EAAE,EAAE,cAAc,SAASlB,EAAE,CAACsB,GAAGtB,CAAC,EAAE,oBAAoB,UAAU,CAAC,QAAQA,KAAK,OAAO,OAAO6D,GAAG,QAAQ,EAAEA,GAAG,mBAAmB7D,CAAC,EAAE,QAAQA,KAAK6D,GAAG,cAAc7D,EAAE,UAAU,EAAE6D,GAAG,cAAc,CAAC,CAAC,EAAE,mBAAmB,SAAS7D,EAAE,CAAC,IAAIC,EAAED,EAAE,YAAY,OAAO6D,GAAG,SAAS5D,GAAG4D,GAAG,cAAc,KAAK7D,CAAC,EAAE6D,GAAG,eAAe,OAAOA,GAAG,eAAe,QAAQ7D,CAAC,EAAE,CAAC,EAAEA,EAAE,YAAY,EAAEmE,GAAGlE,CAAC,CAAC,EAAE,sBAAsB,SAASD,EAAE,CAAC,EAAE,cAAc,UAAU,CAAC6D,GAAG,iBAAiB,QAAQ7D,GAAGA,EAAE,CAAC,CAAC,EAAE,uBAAuB,SAASA,EAAEC,EAAE,CAACD,EAAE,UAAUO,IAAI,CAAC,IAAIC,GAAGD,GAAG,KAAKE,GAAGD,GAAG,IAAI,GAAGR,EAAE,cAAc6D,GAAG,oCAAoC7D,EAAE,aAAaQ,GAAG,cAAcA,GAAG,cAAcsE,GAAG,EAAE,CAAC,IAAIpE,GAAGmD,GAAG,SAASrD,GAAG,cAAcE,GAAGA,GAAG,YAAYF,GAAGA,GAAG,YAAY,EAAEF,EAAE,0CAA0CG,GAAG,uBAAuBD,GAAG,aAAa,qCAAqC,EAAEqD,GAAG,oCAAoC,OAAO,MAAM,CAACpD,KAAK,uBAAuBsE,GAAGvE,GAAG,KAAK,EAAEC,KAAK,cAAc6D,GAAG9D,EAAE,EAAEC,KAAK,gBAAgB4D,GAAG7D,GAAG,MAAM,EAAEC,KAAK,aAAayD,GAAG1D,GAAG,MAAM,EAAEC,KAAK,eAAe2D,GAAG5D,GAAG,MAAM,EAAEC,KAAK,UAAUT,EAAE,OAAO,GAAGC,GAAGA,EAAED,CAAC,EAAEA,EAAE,aAAaA,EAAE,WAAW,EAAE,OAAOA,EAAE,aAAaS,KAAK,QAAQG,EAAE,UAAUJ,GAAG,SAAS,KAAKA,GAAG,IAAI,EAAEC,KAAK,WAAWH,EAAE,UAAUE,GAAG,SAAS,KAAKA,GAAG,IAAI,EAAEC,KAAK,QAAQ,MAAM,UAAUD,GAAG,SAAS,KAAKA,GAAG,IAAI,EAAEA,GAAG,SAAS,eAAeR,EAAE,YAAYQ,EAAE,EAAEC,KAAK,UAAU3F,EAAE,SAASA,EAAE,QAAQ0F,GAAG,GAAG,EAAEC,IAAIH,EAAE,kCAAkCG,EAAE,EAAEoD,GAAG,oCAAoC,MAAM,EAAE7D,EAAE,QAAQO,IAAI,CAAC,IAAIC,GAAG,wBAAwB,MAAMF,EAAEE,GAAG,IAAID,GAAG,SAAS,IAAIA,GAAG,OAAO,KAAKA,GAAG,OAAO,EAAEA,EAAE,EAAEjD,IAAI0C,EAAE,GAAG,UAAU,SAASO,GAAG,CAACP,EAAE,UAAU,CAAC,KAAKO,EAAE,CAAC,CAAC,CAAC,EAAEP,EAAE,GAAG,QAAQ,SAASO,GAAG,CAACP,EAAE,QAAQO,EAAE,CAAC,CAAC,EAAEP,EAAE,GAAG,eAAe,UAAU,CAAC,CAAC,GAAGA,EAAE,YAAY,CAAC,IAAI,OAAO,UAAUlF,EAAE,qBAAqBnC,EAAE,WAAW2G,GAAG,WAAW8B,EAAE,CAAC,CAAC,EAAE,qBAAqB,UAAU,CAAC,IAAIpB,EAAEvC,EAAE,2CAA2C,EAAEoG,GAAG,cAAc,KAAK,IAAI,OAAO7D,CAAC,CAAC,CAAC,EAAE,aAAa,UAAU,CAAC,OAAO6D,GAAG,cAAc,QAAQ,IAAIA,GAAG,qBAAqB,EAAEA,GAAG,uBAAuBA,GAAG,cAAc,EAAE,GAAGA,GAAG,cAAc,IAAI,CAAC,CAAC,EAAE/I,EAAE,QAAQ+I,GAAG,SAASnB,GAAG1C,EAAE,CAAC,KAAKA,EAAE,OAAO,GAAGA,EAAE,MAAM,EAAElF,CAAC,CAAC,CAAC,SAASkK,GAAGhF,EAAE,CAAC,IAAIC,EAAEgF,GAAG,EAAE1E,GAAGP,EAAE,EAAE,OAAOkF,GAAGjF,CAAC,EAAEM,EAAE,CAAC,SAAS4E,GAAGnF,EAAE,CAAC,OAAOA,CAAC,CAAC,SAASoF,GAAGpF,EAAE,CAAC,IAAIC,EAAE,gBAAgB,OAAOD,EAAE,QAAQC,EAAE,SAASM,GAAG,CAAC,IAAIC,GAAGD,GAAG,OAAOA,KAAKC,GAAGD,GAAGC,GAAG,KAAKD,GAAG,GAAG,CAAC,CAAC,CAAC,SAAS8E,IAAI,CAAC,IAAIrF,EAAE8E,GAAG,EAAE7E,EAAEvF,EAAE,EAAEsF,EAAE,IAAI,GAAGO,GAAG7F,EAAE,EAAEsF,EAAE,IAAI,GAAGQ,GAAGP,EAAEM,GAAG+E,GAAGrF,EAAEO,EAAE,EAAE0E,GAAGjF,CAAC,CAAC,CAACnF,EAAE,oBAAoBuK,GAAG,SAASV,GAAG3E,EAAE,CAAC,GAAG,EAAE,OAAOyE,GAAG,EAAE,EAAEzE,CAAC,EAAE,GAAG,CAAC4E,GAAG5E,CAAC,CAAC,OAAOC,EAAN,CAAS4E,GAAG5E,CAAC,CAAC,CAAC,CAAC,IAAIsF,GAAG,CAAC,EAAE,SAASC,GAAGxF,EAAE,CAAC,IAAIC,EAAEsF,GAAGvF,GAAG,OAAOC,IAAID,GAAGuF,GAAG,SAASA,GAAG,OAAOvF,EAAE,GAAGuF,GAAGvF,GAAGC,EAAEkC,GAAG,IAAInC,CAAC,GAAGC,CAAC,CAAC,SAASwF,GAAGzF,EAAEC,EAAE,CAAC,IAAIM,GAAGiF,GAAGxF,CAAC,EAAEC,CAAC,EAAEU,GAAG,EAAEkD,GAAG,cAActD,EAAE,EAAEmF,GAAGnF,EAAE,CAAC,CAACzF,EAAE,iBAAiB2K,GAAG,SAASE,IAAI,CAAC,IAAI3F,EAAE,IAAI,MAAM,GAAG,CAACA,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,IAAI,KAAK,OAAOC,EAAN,CAASD,EAAEC,CAAC,CAAC,GAAG,CAACD,EAAE,MAAM,MAAM,4BAA4B,CAAC,OAAOA,EAAE,MAAM,SAAS,CAAC,CAAC,SAAS4D,GAAG5D,EAAE,CAAC6D,GAAG,iBAAiB,KAAK7D,CAAC,CAAC,CAAC,SAAS4F,GAAG5F,EAAEC,EAAE,CAACpH,EAAE,EAAE,IAAImH,EAAEC,CAAC,CAAC,CAAC,SAAS4F,GAAG7F,EAAE,CAAC8F,GAAG9F,EAAE,CAACtC,EAAE,EAAE,CAACwC,CAAC,EAAE2D,GAAG,cAAc,CAAC,CAAC,SAASkC,GAAG/F,EAAE,CAAC,EAAE,YAAY,CAAC,IAAI,gBAAgB,OAAOA,CAAC,CAAC,EAAEqE,GAAGrE,CAAC,CAAC,CAAC,SAASgG,GAAGhG,EAAEC,EAAEM,GAAGC,GAAG,CAAC,OAAO,EAAEiE,GAAG,EAAE,EAAEzE,EAAEC,EAAEM,GAAGC,EAAE,EAAEyF,GAAGjG,EAAEC,EAAEM,GAAGC,EAAE,CAAC,CAAC,SAASyF,GAAGjG,EAAEC,EAAEM,GAAGC,GAAG,CAAC,GAAG,OAAO,mBAAmB,YAAY,OAAOF,EAAE,qFAAqF,EAAE,EAAE,IAAIG,GAAG,CAAC,EAAEC,GAAG,EAAE,GAAG,IAAID,GAAG,SAAS,GAAGC,IAAI,OAAOsF,GAAGhG,EAAEC,EAAEM,GAAGC,EAAE,EAAE,GAAGE,GAAG,OAAOA,GAAG,IAAIgB,GAAG,CAAC,aAAanB,GAAG,YAAYP,EAAE,IAAIQ,GAAG,aAAaC,EAAE,EAAE,OAAO,GAAGiB,GAAG,IAAI,cAAc,YAAYA,GAAGjB,EAAE,EAAE,GAAG6D,GAAG5C,EAAE,CAAC,CAAC,SAASwE,IAAI,CAAC,MAAO,QAAO,CAAC,IAAIC,GAAG,GAAG,SAASC,IAAI,CAAC,OAAOD,EAAE,CAAC,SAASpB,GAAG/E,EAAE,CAAC,QAAQ,MAAMtF,EAAE,EAAEsF,GAAG,EAAE,CAAC,EAAE8E,GAAG,GAAGuB,GAAGrG,CAAC,EAAE,QAAQ,gBAAgBtF,EAAE,EAAEsF,GAAG,EAAE,EAAE,CAAC,CAAC,CAAClF,EAAE,6BAA6BiK,GAAG,SAASuB,GAAGtG,EAAEC,EAAEM,GAAGC,GAAG,CAAC,GAAGR,GAAGC,EAAE,WAAW,IAAI8E,GAAGvE,EAAE,CAAC,UAAU,EAAE,YAAY,CAAC,aAAaR,EAAE,IAAI,uBAAuB,MAAMQ,EAAE,CAAC,MAAM,CAAC,IAAIC,GAAGoD,GAAG,SAAS7D,GAAG,GAAG,CAACS,GAAG,OAAOA,GAAG,YAAY,CAAC,IAAI,uBAAuB,MAAMD,EAAE,CAAC,CAAC,CAAC,MAAO,EAAC,CAAC,SAAS+F,GAAGvG,EAAEC,EAAEM,GAAG,CAAC,MAAM,EAAE,CAAC,SAASiG,IAAI,CAACrF,GAAG,EAAE,CAAC,CAAC,SAASsF,GAAGzG,EAAE,CAACyG,GAAG,QAAQA,GAAG,MAAM,CAAC,GAAGA,GAAG,MAAMzG,KAAKyG,GAAG,MAAMzG,GAAG,EAAE1C,IAAI0C,EAAE,YAAYA,GAAGM,EAAEN,CAAC,EAAE,CAAC,SAAS0G,IAAI,CAACpJ,GAAGI,GAAG+I,GAAG,0IAA0I,CAAC,CAAC,SAASE,IAAI,CAAC,OAAO,KAAK,IAAI,CAAC,CAAC,SAASC,IAAI,CAAC,MAAO,WAAU,CAAC,SAASC,IAAI,CAAC,OAAOD,GAAG,CAAC,CAAC,IAAIE,GAAGxJ,EAAEwJ,GAAG,IAAI,CAAC,IAAI9G,EAAE,QAAQ,OAAO,EAAE,OAAOA,EAAE,GAAG,IAAIA,EAAE,GAAG,GAAG,EAAE,EAAE8G,GAAG,IAAI,YAAY,IAAI,EAAEhM,EAAE,8BAA8BgM,GAAG,IAAI,YAAY,IAAI,EAAE,SAASC,GAAG/G,EAAEC,EAAEM,GAAG,CAACtH,EAAE,EAAE,WAAW+G,EAAEC,EAAEA,EAAEM,EAAE,CAAC,CAAC,SAASyG,IAAI,CAAC,OAAO1J,EAAE2B,GAAG,EAAE,KAAK,EAAE,OAAO,UAAU,mBAAmB,CAAC,SAASwF,GAAGzE,EAAEC,EAAE,CAAC,IAAIM,GAAG,UAAU,OAAO,EAAEC,GAAG,UAAU,OAAOwE,GAAG,IAAI,CAAC,QAAQvE,GAAGF,GAAGG,GAAGuG,GAAGxG,GAAG,CAAC,EAAEiB,GAAGhB,IAAI,EAAEiB,GAAG,EAAEA,GAAGpB,GAAGoB,KAAK,CAAC,IAAIC,GAAGpB,GAAG,EAAEmB,IAAI9G,EAAE,EAAE6G,GAAGC,IAAIC,EAAE,CAAC,OAAOsF,GAAGlH,EAAES,GAAGC,GAAGT,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIkH,GAAG,CAAC,EAAE,SAASC,GAAGpH,EAAEC,EAAEM,GAAG,CAAC4G,GAAG,OAAOlH,EAAE,QAAQO,GAAGD,IAAI,EAAEE,GAAG,EAAEA,GAAGR,EAAEQ,KAAK0G,GAAG1G,IAAI5F,EAAE,EAAE2F,GAAGC,IAAI,IAAIC,GAAGV,EAAE,EAAE0B,GAAGhB,GAAGuD,GAAG,CAACjE,EAAE,GAAGqH,GAAGrH,GAAG,OAAO0B,GAAG,MAAM,KAAKyF,EAAE,CAAC,CAAC,SAASG,GAAGtH,EAAE,CAAC,GAAG,CAAC,OAAOV,GAAG,KAAKU,EAAET,GAAG,WAAW,QAAQ,EAAE,EAAEC,GAAGF,GAAG,MAAM,EAAE,CAAC,OAAOW,EAAN,CAAS,CAAC,CAAC,SAASsH,GAAGvH,EAAE,CAAC,IAAIC,EAAEhH,EAAE,EAAE,OAAO,GAAG+G,EAAEA,IAAI,EAAEA,GAAGC,EAAE,MAAM,GAAG,IAAIM,GAAGqG,GAAG,EAAE,GAAG5G,EAAEO,GAAG,MAAM,GAAG,IAAIC,GAAG,CAACoB,GAAGC,KAAKD,IAAIC,GAAGD,GAAGC,IAAIA,GAAG,QAAQpB,GAAG,EAAEA,IAAI,EAAEA,IAAI,EAAE,CAAC,IAAIC,GAAGT,GAAG,EAAE,GAAGQ,IAAIC,GAAG,KAAK,IAAIA,GAAGV,EAAE,SAAS,EAAE,IAAI0B,GAAG,KAAK,IAAInB,GAAGC,GAAG,KAAK,IAAIR,EAAEU,EAAE,EAAE,KAAK,CAAC,EAAEiB,GAAG2F,GAAG5F,EAAE,EAAE,GAAGC,GAAG,MAAM,EAAE,CAAC,MAAM,EAAE,CAAC,SAAS6F,IAAI,CAAC,KAAK,QAAQ,CAAC,SAASC,GAAGzH,EAAE,CAAC,OAAO,EAAEyE,GAAG,EAAE,EAAEzE,CAAC,EAAE,EAAE,CAAC,SAAS0H,GAAG1H,EAAEC,EAAEM,GAAGC,GAAGC,GAAG,CAAC,OAAO,EAAEgE,GAAG,EAAE,EAAEzE,EAAEC,EAAEM,GAAGC,GAAGC,EAAE,EAAE,EAAE,CAAC,IAAIkH,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC,EAAE,SAASC,GAAG5H,EAAEC,EAAE,CAAC,IAAIM,GAAGoH,GAAG3H,GAAGC,IAAI,GAAGA,IAAI,KAAKD,IAAI,EAAEY,EAAEN,GAAGmB,GAAGlB,GAAG,CAAC,CAAC,EAAEA,GAAG,OAAO,GAAGA,GAAG,KAAKN,CAAC,CAAC,CAAC,SAAS4H,GAAG7H,EAAEC,EAAEM,GAAGC,GAAG,CAAC,GAAG,EAAE,OAAOiE,GAAG,EAAE,EAAEzE,EAAEC,EAAEM,GAAGC,EAAE,EAAE,QAAQC,GAAG,EAAEC,GAAG,EAAEA,GAAGH,GAAGG,KAAK,CAAC,IAAIgB,GAAG/G,EAAE,EAAEsF,GAAG,GAAG0B,GAAGhH,EAAE,EAAEsF,EAAE,GAAG,GAAGA,GAAG,EAAE,QAAQ2B,GAAG,EAAEA,GAAGD,GAAGC,KAAKgG,GAAG5H,EAAE/G,EAAE,EAAEyI,GAAGE,GAAG,EAAEnB,IAAIkB,EAAE,CAAC,OAAOhH,EAAE,EAAE6F,IAAI,GAAGC,GAAG,CAAC,CAAC,SAASqH,GAAG9H,EAAE,CAAC,IAAIC,EAAEnF,EAAE,IAAIkF,GAAG,OAAOC,CAAC,CAAC,SAAS8H,GAAG/H,EAAEC,EAAEM,GAAGC,GAAGC,GAAG,CAAC,IAAIC,GAAG,CAAC,OAAOsH,IAAI,CAAC,IAAIC,GAAG,EAAE,GAAGD,IAAI,MAAMA,KAAK,EAAE,CAAC,IAAIE,IAAIF,GAAG,QAAQ,GAAG,EAAEC,GAAGhB,GAAGiB,EAAE,EAAElG,GAAGgG,GAAGC,GAAGC,EAAE,CAAC,CAAC,OAAOD,EAAE,EAAE,MAAMD,IAAI,CAAC,IAAIC,GAAGhB,GAAGe,GAAG,MAAM,EAAE,OAAOpC,GAAGoC,GAAGC,EAAE,EAAEA,EAAE,CAAC,EAAE,SAASvG,GAAGsG,GAAG,CAAC,OAAO/H,IAAI,SAAS6B,GAAGkG,EAAE,EAAE/H,IAAI,UAAU,QAAQ+H,EAAE,EAAEA,EAAE,CAAC,IAAIrG,GAAGmG,GAAG9H,CAAC,EAAE4B,GAAG,CAAC,EAAEC,GAAG,EAAE,GAAGrB,GAAG,QAAQsD,GAAG,EAAEA,GAAGtD,GAAG,OAAOsD,KAAK,CAAC,IAAIqE,GAAGzH,GAAGH,GAAGuD,KAAKqE,IAAItG,KAAK,IAAIA,GAAGoD,GAAG,GAAGrD,GAAGkC,IAAIqE,GAAG3H,GAAGsD,GAAG,GAAGlC,GAAGkC,IAAItD,GAAGsD,GAAG,CAAC,IAAIsE,GAAGzG,GAAG,MAAM,KAAKC,EAAE,EAAE,SAASyG,GAAGL,GAAG,CAAC,OAAOnG,KAAK,GAAGqD,GAAGrD,EAAE,EAAEH,GAAGsG,EAAE,CAAC,CAAC,OAAOI,GAAGC,GAAGD,EAAE,EAAEA,EAAE,CAAC,SAASE,GAAGtI,EAAEC,EAAEM,GAAGC,GAAG,CAACD,GAAGA,IAAI,CAAC,EAAE,IAAIE,GAAGF,GAAG,MAAMmB,IAAIA,KAAK,UAAUA,KAAK,SAAS,EAAEhB,GAAGT,IAAI,SAAS,OAAOS,IAAID,IAAI,CAACD,GAAGsH,GAAG9H,CAAC,EAAE,UAAU,CAAC,OAAO+H,GAAG/H,EAAEC,EAAEM,GAAG,UAAUC,EAAE,CAAC,CAAC,CAACqD,GAAG,KAAK,EAAE,IAAIwD,GAAG,CAAC,KAAK7C,GAAGG,GAAGqB,GAAGyB,GAAGC,GAAGG,EAAE,EAAElE,GAAG,CAAC,iCAAiCkC,GAAG,4BAA4BE,GAAG,oBAAoBE,GAAG,uCAAuCC,GAAG,iCAAiCE,GAAG,8BAA8BE,GAAG,qCAAqCC,GAAG,MAAMC,GAAG,kCAAkCE,GAAG,oBAAoBC,GAAG,wBAAwBE,GAAG,mBAAmBC,GAAG,sBAAsBC,GAAG,6BAA6BC,GAAG,qCAAqCI,GAAG,uBAAuBG,GAAG,mCAAmCC,GAAG,KAAK5C,GAAG,SAAS6C,GAAG,QAAQC,GAAG,SAASG,GAAG,OAAOvI,IAAIxE,EAAE,UAAU,EAAEyN,GAAG7E,GAAG,EAAE8E,GAAG1N,EAAE,mBAAmB,UAAU,CAAC,OAAO0N,GAAG1N,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAE2N,GAAG3N,EAAE,MAAM,UAAU,CAAC,OAAO2N,GAAG3N,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE4N,GAAG5N,EAAE,yBAAyB,UAAU,CAAC,OAAO4N,GAAG5N,EAAE,yBAAyBA,EAAE,IAAI,yBAAyB,MAAM,KAAK,SAAS,CAAC,EAAE6N,GAAG7N,EAAE,mBAAmB,UAAU,CAAC,OAAO6N,GAAG7N,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAE8N,GAAG9N,EAAE,iBAAiB,UAAU,CAAC,OAAO8N,GAAG9N,EAAE,iBAAiBA,EAAE,IAAI,iBAAiB,MAAM,KAAK,SAAS,CAAC,EAAE+N,GAAG/N,EAAE,cAAc,UAAU,CAAC,OAAO+N,GAAG/N,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEgO,GAAGhO,EAAE,SAAS,UAAU,CAAC,OAAOgO,GAAGhO,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEiO,GAAGjO,EAAE,KAAK,UAAU,CAAC,OAAOiO,GAAGjO,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEkO,GAAGlO,EAAE,KAAK,UAAU,CAAC,OAAOkO,GAAGlO,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEmO,GAAGnO,EAAE,MAAM,UAAU,CAAC,OAAOmO,GAAGnO,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEoO,GAAGpO,EAAE,KAAK,UAAU,CAAC,OAAOoO,GAAGpO,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEqO,GAAGrO,EAAE,KAAK,UAAU,CAAC,OAAOqO,GAAGrO,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEsO,GAAGtO,EAAE,QAAQ,UAAU,CAAC,OAAOsO,GAAGtO,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEuO,GAAGvO,EAAE,SAAS,UAAU,CAAC,OAAOuO,GAAGvO,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEwO,GAAGxO,EAAE,aAAa,UAAU,CAAC,OAAOwO,GAAGxO,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAEyO,GAAGzO,EAAE,MAAM,UAAU,CAAC,OAAOyO,GAAGzO,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE0O,GAAG1O,EAAE,aAAa,UAAU,CAAC,OAAO0O,GAAG1O,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAE2O,GAAG3O,EAAE,QAAQ,UAAU,CAAC,OAAO2O,GAAG3O,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE4O,GAAG5O,EAAE,qBAAqB,UAAU,CAAC,OAAO4O,GAAG5O,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAE6O,GAAG7O,EAAE,KAAK,UAAU,CAAC,OAAO6O,GAAG7O,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE8O,GAAG9O,EAAE,MAAM,UAAU,CAAC,OAAO8O,GAAG9O,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE+O,GAAG/O,EAAE,eAAe,UAAU,CAAC,OAAO+O,GAAG/O,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAEgP,GAAGhP,EAAE,SAAS,UAAU,CAAC,OAAOgP,GAAGhP,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEiP,GAAGjP,EAAE,QAAQ,UAAU,CAAC,OAAOiP,GAAGjP,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEkP,GAAGlP,EAAE,cAAc,UAAU,CAAC,OAAOkP,GAAGlP,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEmP,GAAGnP,EAAE,uBAAuB,UAAU,CAAC,OAAOmP,GAAGnP,EAAE,uBAAuBA,EAAE,IAAI,uBAAuB,MAAM,KAAK,SAAS,CAAC,EAAEoP,GAAGpP,EAAE,KAAK,UAAU,CAAC,OAAOoP,GAAGpP,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEqP,GAAGrP,EAAE,OAAO,UAAU,CAAC,OAAOqP,GAAGrP,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEsP,GAAGtP,EAAE,KAAK,UAAU,CAAC,OAAOsP,GAAGtP,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEuP,GAAGvP,EAAE,eAAe,UAAU,CAAC,OAAOuP,GAAGvP,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAEwP,GAAGxP,EAAE,OAAO,UAAU,CAAC,OAAOwP,GAAGxP,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEyP,GAAGzP,EAAE,UAAU,UAAU,CAAC,OAAOyP,GAAGzP,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAE0P,GAAG1P,EAAE,gBAAgB,UAAU,CAAC,OAAO0P,GAAG1P,EAAE,gBAAgBA,EAAE,IAAI,gBAAgB,MAAM,KAAK,SAAS,CAAC,EAAE2P,GAAG3P,EAAE,aAAa,UAAU,CAAC,OAAO2P,GAAG3P,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAE4P,GAAG5P,EAAE,sBAAsB,UAAU,CAAC,OAAO4P,GAAG5P,EAAE,sBAAsBA,EAAE,IAAI,sBAAsB,MAAM,KAAK,SAAS,CAAC,EAAE6P,GAAG7P,EAAE,QAAQ,UAAU,CAAC,OAAO6P,GAAG7P,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE8P,GAAG9P,EAAE,UAAU,UAAU,CAAC,OAAO8P,GAAG9P,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAE+P,GAAG/P,EAAE,SAAS,UAAU,CAAC,OAAO+P,GAAG/P,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEgQ,GAAGhQ,EAAE,cAAc,UAAU,CAAC,OAAOgQ,GAAGhQ,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEiQ,GAAGjQ,EAAE,WAAW,UAAU,CAAC,OAAOiQ,GAAGjQ,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEkQ,GAAGlQ,EAAE,MAAM,UAAU,CAAC,OAAOkQ,GAAGlQ,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEmQ,GAAGnQ,EAAE,WAAW,UAAU,CAAC,OAAOmQ,GAAGnQ,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEoQ,GAAGpQ,EAAE,KAAK,UAAU,CAAC,OAAOoQ,GAAGpQ,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEqQ,GAAGrQ,EAAE,YAAY,UAAU,CAAC,OAAOqQ,GAAGrQ,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEsQ,GAAGtQ,EAAE,YAAY,UAAU,CAAC,OAAOsQ,GAAGtQ,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEuQ,GAAGvQ,EAAE,WAAW,UAAU,CAAC,OAAOuQ,GAAGvQ,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEwQ,GAAGxQ,EAAE,YAAY,UAAU,CAAC,OAAOwQ,GAAGxQ,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEyQ,GAAGzQ,EAAE,KAAK,UAAU,CAAC,OAAOyQ,GAAGzQ,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE0Q,GAAG1Q,EAAE,SAAS,UAAU,CAAC,OAAO0Q,GAAG1Q,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE2Q,GAAG3Q,EAAE,SAAS,UAAU,CAAC,OAAO2Q,GAAG3Q,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE4Q,GAAG5Q,EAAE,MAAM,UAAU,CAAC,OAAO4Q,GAAG5Q,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE6Q,GAAG7Q,EAAE,KAAK,UAAU,CAAC,OAAO6Q,GAAG7Q,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE8Q,GAAG9Q,EAAE,SAAS,UAAU,CAAC,OAAO8Q,GAAG9Q,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE+Q,GAAG/Q,EAAE,WAAW,UAAU,CAAC,OAAO+Q,GAAG/Q,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEgR,GAAGhR,EAAE,UAAU,UAAU,CAAC,OAAOgR,GAAGhR,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEiR,GAAGjR,EAAE,KAAK,UAAU,CAAC,OAAOiR,GAAGjR,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEkR,GAAGlR,EAAE,qBAAqB,UAAU,CAAC,OAAOkR,GAAGlR,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEmR,GAAGnR,EAAE,qBAAqB,UAAU,CAAC,OAAOmR,GAAGnR,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEoR,EAAEpR,EAAE,qBAAqB,UAAU,CAAC,OAAOoR,EAAEpR,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEqR,GAAGrR,EAAE,UAAU,UAAU,CAAC,OAAOqR,GAAGrR,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEsR,GAAGtR,EAAE,QAAQ,UAAU,CAAC,OAAOsR,GAAGtR,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEuR,GAAGvR,EAAE,OAAO,UAAU,CAAC,OAAOuR,GAAGvR,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEwR,GAAGxR,EAAE,KAAK,UAAU,CAAC,OAAOwR,GAAGxR,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEyR,GAAGzR,EAAE,OAAO,UAAU,CAAC,OAAOyR,GAAGzR,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAE0R,GAAG1R,EAAE,MAAM,UAAU,CAAC,OAAO0R,GAAG1R,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE2R,GAAG3R,EAAE,SAAS,UAAU,CAAC,OAAO2R,GAAG3R,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE4R,GAAG5R,EAAE,MAAM,UAAU,CAAC,OAAO4R,GAAG5R,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE6R,GAAG7R,EAAE,OAAO,UAAU,CAAC,OAAO6R,GAAG7R,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAE8R,GAAG9R,EAAE,gBAAgB,UAAU,CAAC,OAAO8R,GAAG9R,EAAE,gBAAgBA,EAAE,IAAI,gBAAgB,MAAM,KAAK,SAAS,CAAC,EAAE+R,GAAG/R,EAAE,uBAAuB,UAAU,CAAC,OAAO+R,GAAG/R,EAAE,uBAAuBA,EAAE,IAAI,uBAAuB,MAAM,KAAK,SAAS,CAAC,EAAEgS,GAAGhS,EAAE,SAAS,UAAU,CAAC,OAAOgS,GAAGhS,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEiS,GAAGjS,EAAE,kBAAkB,UAAU,CAAC,OAAOiS,GAAGjS,EAAE,kBAAkBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAEkS,GAAGlS,EAAE,OAAO,UAAU,CAAC,OAAOkS,GAAGlS,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEmS,GAAGnS,EAAE,OAAO,UAAU,CAAC,OAAOmS,GAAGnS,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEoS,GAAGpS,EAAE,WAAW,UAAU,CAAC,OAAOoS,GAAGpS,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEqS,GAAGrS,EAAE,UAAU,UAAU,CAAC,OAAOqS,GAAGrS,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEsS,GAAGtS,EAAE,SAAS,UAAU,CAAC,OAAOsS,GAAGtS,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEuS,GAAGvS,EAAE,KAAK,UAAU,CAAC,OAAOuS,GAAGvS,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEwS,GAAGxS,EAAE,SAAS,UAAU,CAAC,OAAOwS,GAAGxS,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEyS,GAAGzS,EAAE,qBAAqB,UAAU,CAAC,OAAOyS,GAAGzS,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAE0S,GAAG1S,EAAE,eAAe,UAAU,CAAC,OAAO0S,GAAG1S,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAE2S,GAAG3S,EAAE,wBAAwB,UAAU,CAAC,OAAO2S,GAAG3S,EAAE,wBAAwBA,EAAE,IAAI,wBAAwB,MAAM,KAAK,SAAS,CAAC,EAAE4S,GAAG5S,EAAE,MAAM,UAAU,CAAC,OAAO4S,GAAG5S,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE6S,GAAG7S,EAAE,QAAQ,UAAU,CAAC,OAAO6S,GAAG7S,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE8S,GAAG9S,EAAE,mBAAmB,UAAU,CAAC,OAAO8S,GAAG9S,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAE+S,GAAG/S,EAAE,MAAM,UAAU,CAAC,OAAO+S,GAAG/S,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEgT,GAAGhT,EAAE,cAAc,UAAU,CAAC,OAAOgT,GAAGhT,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEiT,GAAGjT,EAAE,KAAK,UAAU,CAAC,OAAOiT,GAAGjT,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEkT,GAAGlT,EAAE,KAAK,UAAU,CAAC,OAAOkT,GAAGlT,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEmT,GAAGnT,EAAE,KAAK,UAAU,CAAC,OAAOmT,GAAGnT,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEoT,GAAGpT,EAAE,MAAM,UAAU,CAAC,OAAOoT,GAAGpT,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEqT,GAAGrT,EAAE,MAAM,UAAU,CAAC,OAAOqT,GAAGrT,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEsT,GAAGtT,EAAE,MAAM,UAAU,CAAC,OAAOsT,GAAGtT,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEuT,GAAGvT,EAAE,WAAW,UAAU,CAAC,OAAOuT,GAAGvT,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEwT,GAAGxT,EAAE,WAAW,UAAU,CAAC,OAAOwT,GAAGxT,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEyT,GAAGzT,EAAE,cAAc,UAAU,CAAC,OAAOyT,GAAGzT,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE0T,GAAG1T,EAAE,QAAQ,UAAU,CAAC,OAAO0T,GAAG1T,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE2T,GAAG3T,EAAE,MAAM,UAAU,CAAC,OAAO2T,GAAG3T,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE4T,GAAG5T,EAAE,sBAAsB,UAAU,CAAC,OAAO4T,GAAG5T,EAAE,sBAAsBA,EAAE,IAAI,sBAAsB,MAAM,KAAK,SAAS,CAAC,EAAEgK,GAAGhK,EAAE,cAAc,UAAU,CAAC,OAAOgK,GAAGhK,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE6T,GAAG7T,EAAE,kBAAkB,UAAU,CAAC,OAAO6T,GAAG7T,EAAE,kBAAkBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAEgL,GAAGhL,EAAE,yBAAyB,UAAU,CAAC,OAAOgL,GAAGhL,EAAE,yBAAyBA,EAAE,IAAI,yBAAyB,MAAM,KAAK,SAAS,CAAC,EAAE8T,GAAG9T,EAAE,4BAA4B,UAAU,CAAC,OAAO8T,GAAG9T,EAAE,4BAA4BA,EAAE,IAAI,4BAA4B,MAAM,KAAK,SAAS,CAAC,EAAE+T,GAAG/T,EAAE,6CAA6C,UAAU,CAAC,OAAO+T,GAAG/T,EAAE,6CAA6CA,EAAE,IAAI,6CAA6C,MAAM,KAAK,SAAS,CAAC,EAAEgU,GAAGhU,EAAE,mCAAmC,UAAU,CAAC,OAAOgU,GAAGhU,EAAE,mCAAmCA,EAAE,IAAI,mCAAmC,MAAM,KAAK,SAAS,CAAC,EAAEoM,GAAGpM,EAAE,0CAA0C,UAAU,CAAC,OAAOoM,GAAGpM,EAAE,0CAA0CA,EAAE,IAAI,0CAA0C,MAAM,KAAK,SAAS,CAAC,EAAEiU,GAAGjU,EAAE,gCAAgC,UAAU,CAAC,OAAOiU,GAAGjU,EAAE,gCAAgCA,EAAE,IAAI,gCAAgC,MAAM,KAAK,SAAS,CAAC,EAAEuL,GAAGvL,EAAE,sCAAsC,UAAU,CAAC,OAAOuL,GAAGvL,EAAE,sCAAsCA,EAAE,IAAI,sCAAsC,MAAM,KAAK,SAAS,CAAC,EAAEqJ,GAAGrJ,EAAE,8BAA8B,UAAU,CAAC,OAAOqJ,GAAGrJ,EAAE,8BAA8BA,EAAE,IAAI,8BAA8B,MAAM,KAAK,SAAS,CAAC,EAAE4K,GAAG5K,EAAE,yBAAyB,UAAU,CAAC,OAAO4K,GAAG5K,EAAE,yBAAyBA,EAAE,IAAI,yBAAyB,MAAM,KAAK,SAAS,CAAC,EAAEwK,GAAGxK,EAAE,6BAA6B,UAAU,CAAC,OAAOwK,GAAGxK,EAAE,6BAA6BA,EAAE,IAAI,6BAA6B,MAAM,KAAK,SAAS,CAAC,EAAEmK,GAAGnK,EAAE,UAAU,UAAU,CAAC,OAAOmK,GAAGnK,EAAE,UAAUA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEoK,GAAGpK,EAAE,aAAa,UAAU,CAAC,OAAOoK,GAAGpK,EAAE,aAAaA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEmM,GAAGnM,EAAE,WAAW,UAAU,CAAC,OAAOmM,GAAGnM,EAAE,WAAWA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEkU,GAAGlU,EAAE,iBAAiB,UAAU,CAAC,OAAOkU,GAAGlU,EAAE,iBAAiBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAEmU,GAAGnU,EAAE,aAAa,UAAU,CAAC,OAAOmU,GAAGnU,EAAE,aAAaA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEA,EAAE,iBAAiB6F,GAAG7F,EAAE,WAAWwE,GAAGxE,EAAE,MAAMwN,GAAGxN,EAAE,WAAWuF,GAAGvF,EAAE,QAAQ+I,GAAG,IAAIqL,GAAGjM,GAAG,SAASjD,GAAG,CAACkP,IAAIC,GAAG,EAAED,KAAKjM,GAAGjD,EAAE,EAAE,SAASmP,GAAGnP,EAAE,CAAC,GAAGA,EAAEA,GAAGrE,EAAEoH,GAAG,EAAE,OAAO,GAAG,EAAE,CAACxH,EAAET,CAAC,EAAE6H,GAAG,EAAE,YAAY,CAAC,IAAI,QAAQ,CAAC,EAAE,MAAM,CAAC,GAAGH,GAAG,EAAEO,GAAG,EAAE,OAAO,SAAS9C,GAAG,CAACiP,KAAKA,GAAG,GAAGpU,EAAE,UAAU,GAAG,CAACuG,KAAKsB,GAAG,EAAEpH,EAAET,CAAC,EAAEA,EAAE,sBAAsBA,EAAE,qBAAqB,EAAE8H,GAAG,GAAG,CAAC9H,EAAE,WAAWA,EAAE,UAAU,YAAY,EAAE,WAAW,UAAU,CAAC,WAAW,UAAU,CAACA,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC,EAAEmF,EAAE,CAAC,EAAE,CAAC,GAAGA,EAAE,CAAC,CAAC,GAAGnF,EAAE,QAAQ,IAAI,OAAOA,EAAE,SAAS,aAAaA,EAAE,QAAQ,CAACA,EAAE,OAAO,GAAGA,EAAE,QAAQ,OAAO,GAAGA,EAAE,QAAQ,IAAI,EAAE,EAAEqU,GAAG,EAAE,IAAIC,GAAG3T,IAAI2T,GAAG,CAAC,kBAAkB,QAAQ,UAAU,mBAAmB,EAAE,OAAO,SAASpP,EAAE,CAAC,MAAM,CAACvE,EAAE,kBAAkB,QAAQuE,CAAC,EAAE,EAAE,CAAC,EAAE,mBAAmB,QAAQ,UAAU,oBAAoB,EAAE,OAAO,SAASA,EAAE,CAAC,MAAM,CAACvE,EAAE,mBAAmB,QAAQuE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,IAAIqP,GAAG,GAAG,OAAO,mBAAmB,YAAYA,GAAG,0BAA0B,OAAO5V,GAAG,YAAY4V,GAAG5V,MAAO,OAAM,IAAI,MAAM,uCAAuC,EAAE,GAAG2V,GAAG,CAAC,IAAIE,GAAGD,GAAG,SAASA,GAAG,SAAS,UAAU,CAACC,GAAG,EAAEF,GAAG,kBAAkB,QAAQ,SAASpP,EAAE,CAAC,QAAQ,eAAe,oBAAoBA,CAAC,CAAC,CAAC,EAAEoP,GAAG,mBAAmB,QAAQ,SAASpP,EAAE,CAAC,QAAQ,eAAe,qBAAqBA,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOvG,EAAE,KAAK,CAAC,GAAG,EAAE,OAAO0F,GAAI,UAAU,OAAOC,GAAI,SAASA,EAAG,QAAQC,EAAG,OAAO,QAAQ,YAAY,OAAO,IAAI,OAAO,CAAC,EAAE,UAAU,CAAC,OAAOA,CAAE,CAAC,EAAE,OAAOF,GAAI,WAAWA,EAAG,8BAA8BE,EAAG,CAAC,EAAMkQ,GAAGzW,GAAG,CAAC0W,EAAIC,IAAK,CAACA,EAAG,QAAQ,mBAAmB;AAAA,okEACrnvB,CAAC,EAAMC,GAAG5W,GAAG,CAAC6W,EAAGC,IAAK,CAAC,IAAIC,GAAI,IAAI,CAAC,IAAIlX,EAAE,OAAO,UAAU,aAAa,SAAS,cAAc,SAAS,cAAc,IAAI,OAAO,OAAO,OAAO,YAAY,cAAcA,EAAEA,GAAG,YAAY,SAASc,EAAE,CAACA,EAAEA,GAAG,CAAC,EAAE,IAAIZ,EAAE,OAAOY,GAAG,YAAYA,EAAE,CAAC,EAAER,EAAE,EAAEJ,EAAE,MAAM,IAAI,QAAQ,SAASqT,EAAEC,GAAG,CAAClT,EAAEiT,EAAE,EAAEC,EAAE,CAAC,EAAE,IAAIzR,EAAE,OAAO,SAAS,aAAa,QAAQ,YAAYA,EAAE,CAAC,kBAAkB,QAAQ,UAAU,mBAAmB,EAAE,mBAAmB,QAAQ,UAAU,oBAAoB,CAAC,GAAG,IAAIC,EAAE,OAAO,OAAO,CAAC,EAAE9B,CAAC,EAAE+B,EAAE,CAAC,EAAEC,EAAE,iBAAiBC,EAAE,CAACoR,EAAEC,KAAK,CAAC,MAAMA,EAAE,EAAE5Q,EAAE,OAAO,QAAQ,SAASC,EAAE,OAAO,eAAe,WAAWC,EAAE,OAAO,SAAS,UAAU,OAAO,QAAQ,UAAU,UAAU,OAAO,QAAQ,SAAS,MAAM,SAASC,EAAE,GAAG,SAASC,EAAEuQ,EAAE,CAAC,OAAOrT,EAAE,WAAWA,EAAE,WAAWqT,EAAExQ,CAAC,EAAEA,EAAEwQ,CAAC,CAAC,IAAIrP,EAAEC,EAAEoD,EAAExC,EAAE,SAASJ,EAAE4O,EAAE,CAAIA,aAAalJ,IAAU7C,EAAE,6BAA6B+L,CAAC,CAAC,CAAC,GAAGzQ,EAAE,CAACD,EAAEE,EAAEoD,GAAG,EAAE,QAAQpD,CAAC,EAAE,IAAIA,EAAE,UAAU,IAAI,IAAI,EAAE8B,EAAE,OAAO9E,IAAI,aAAa,EAAEmG,GAAG,EAAErB,EAAEsB,GAAG,GAAGjC,EAAE,CAACqP,EAAEC,MAAMD,EAAE1O,EAAE,UAAU0O,CAAC,EAAE,EAAE,aAAaA,EAAEC,GAAG,OAAO,MAAM,GAAGjM,EAAEgM,GAAG,CAAC,IAAIC,GAAGtP,EAAEqP,EAAE,EAAE,EAAE,OAAOC,GAAG,SAASA,GAAG,IAAI,WAAWA,EAAE,GAAGA,EAAE,EAAErP,EAAE,CAACoP,EAAEC,GAAGC,KAAK,CAACF,EAAE1O,EAAE,UAAU0O,CAAC,EAAE,EAAE,SAASA,EAAE,SAASG,GAAGC,GAAG,CAACD,GAAGD,GAAGC,EAAE,EAAEF,GAAGG,GAAG,MAAM,CAAC,CAAC,CAAC,EAAE,QAAQ,KAAK,OAAO,IAAIzR,EAAE,QAAQ,KAAK,GAAG,QAAQ,MAAM,GAAG,GAAGD,EAAE,QAAQ,KAAK,MAAM,CAAC,EAAE,QAAQ,GAAG,oBAAoB,SAASsR,EAAE,CAAC,GAAG,EAAEA,aAAalJ,IAAI,MAAMkJ,CAAC,CAAC,EAAE,QAAQ,GAAG,qBAAqB,SAASA,EAAE,CAAC,MAAMA,CAAC,CAAC,EAAEpR,EAAE,CAACoR,EAAEC,KAAK,CAAC,GAAGzM,GAAG,EAAE,MAAM,QAAQ,SAASwM,EAAEC,GAAG7O,EAAE6O,EAAE,EAAE,QAAQ,KAAKD,CAAC,CAAC,EAAErT,EAAE,QAAQ,UAAU,CAAC,MAAM,4BAA4B,CAAC,MAAM0C,GAAGC,KAAKA,EAAEE,EAAE,KAAK,SAAS,KAAK,OAAO,UAAU,aAAa,SAAS,gBAAgBA,EAAE,SAAS,cAAc,KAAK/C,IAAI+C,EAAE/C,GAAG+C,EAAE,QAAQ,OAAO,IAAI,EAAEA,EAAEA,EAAE,OAAO,EAAEA,EAAE,QAAQ,SAAS,EAAE,EAAE,YAAY,GAAG,EAAE,CAAC,EAAEA,EAAE,GAAGmB,EAAEqP,GAAG,CAAC,IAAIC,GAAG,IAAI,eAAe,OAAOA,GAAG,KAAK,MAAMD,EAAE,EAAE,EAAEC,GAAG,KAAK,IAAI,EAAEA,GAAG,YAAY,EAAE3Q,IAAI0E,EAAEgM,GAAG,CAAC,IAAIC,GAAG,IAAI,eAAe,OAAOA,GAAG,KAAK,MAAMD,EAAE,EAAE,EAAEC,GAAG,aAAa,cAAcA,GAAG,KAAK,IAAI,EAAE,IAAI,WAAWA,GAAG,QAAQ,CAAC,GAAGrP,EAAE,CAACoP,EAAEC,GAAGC,KAAK,CAAC,IAAIC,GAAG,IAAI,eAAeA,GAAG,KAAK,MAAMH,EAAE,EAAE,EAAEG,GAAG,aAAa,cAAcA,GAAG,OAAO,IAAI,CAAC,GAAGA,GAAG,QAAQ,KAAKA,GAAG,QAAQ,GAAGA,GAAG,SAAS,CAACF,GAAGE,GAAG,QAAQ,EAAE,MAAM,CAACD,GAAG,CAAC,EAAEC,GAAG,QAAQD,GAAGC,GAAG,KAAK,IAAI,CAAC,EAAE3O,EAAEwO,GAAG,SAAS,MAAMA,GAAG,IAAIzO,EAAE5E,EAAE,OAAO,QAAQ,IAAI,KAAK,OAAO,EAAEsH,EAAEtH,EAAE,UAAU,QAAQ,KAAK,KAAK,OAAO,EAAE,OAAO,OAAOA,EAAE8B,CAAC,EAAEA,EAAE,KAAK9B,EAAE,YAAY+B,EAAE/B,EAAE,WAAWA,EAAE,cAAcgC,EAAEhC,EAAE,aAAaA,EAAE,OAAOiC,EAAEjC,EAAE,MAAM,IAAI8E,EAAE,EAAEyC,EAAEvH,EAAE,aAAauH,EAAEvH,EAAE,YAAY,IAAI+E,EAAE/E,EAAE,eAAe,GAAG,OAAO,aAAa,UAAUyJ,GAAG,iCAAiC,EAAE,IAAIzE,EAAEC,EAAE,GAAGC,EAAE,SAAS,EAAEmO,EAAEC,GAAG,CAACD,GAAG5J,GAAG6J,EAAE,CAAC,CAAC,IAAIlO,EAAE,OAAO,aAAa,YAAY,IAAI,YAAY,MAAM,EAAE,OAAO,SAAS2C,EAAEsL,EAAEC,GAAGC,GAAG,CAAC,QAAQC,GAAGF,GAAGC,GAAGE,GAAGH,GAAGD,EAAEI,KAAK,EAAEA,IAAID,KAAK,EAAEC,GAAG,GAAGA,GAAGH,GAAG,IAAID,EAAE,QAAQjO,EAAE,OAAOA,EAAE,OAAOiO,EAAE,SAASC,GAAGG,EAAE,CAAC,EAAE,QAAQC,GAAG,GAAGJ,GAAGG,IAAI,CAAC,IAAIE,GAAGN,EAAEC,MAAM,GAAG,EAAEK,GAAG,KAAK,CAACD,IAAI,OAAO,aAAaC,EAAE,EAAE,QAAQ,CAAC,IAAIC,GAAGP,EAAEC,MAAM,GAAG,IAAIK,GAAG,MAAM,IAAI,CAACD,IAAI,OAAO,cAAcC,GAAG,KAAK,EAAEC,EAAE,EAAE,QAAQ,CAAC,IAAIC,GAAGR,EAAEC,MAAM,GAAG,IAAIK,GAAG,MAAM,IAAIA,IAAIA,GAAG,KAAK,GAAGC,IAAI,EAAEC,GAAGF,IAAIA,GAAG,IAAI,GAAGC,IAAI,GAAGC,IAAI,EAAER,EAAEC,MAAM,GAAGK,GAAG,MAAMD,IAAI,OAAO,aAAaC,EAAE,MAAM,CAAC,IAAIG,GAAGH,GAAG,MAAMD,IAAI,OAAO,aAAa,MAAMI,IAAI,GAAG,MAAMA,GAAG,IAAI,CAAC,CAAC,CAAC,OAAOJ,EAAE,CAAC,SAASjM,EAAE4L,EAAEC,GAAG,CAAC,OAAOD,EAAEtL,EAAEK,GAAGiL,EAAEC,EAAE,EAAE,EAAE,CAAC,SAAStL,GAAEqL,EAAEC,GAAGC,GAAGC,GAAG,CAAC,GAAG,EAAEA,GAAG,GAAG,MAAO,GAAE,QAAQC,GAAGF,GAAGG,GAAGH,GAAGC,GAAG,EAAEG,GAAG,EAAEA,GAAGN,EAAE,OAAO,EAAEM,GAAG,CAAC,IAAIC,GAAGP,EAAE,WAAWM,EAAE,EAAE,GAAGC,IAAI,OAAOA,IAAI,MAAM,CAAC,IAAIC,GAAGR,EAAE,WAAW,EAAEM,EAAE,EAAEC,GAAG,QAAQA,GAAG,OAAO,IAAIC,GAAG,IAAI,CAAC,GAAGD,IAAI,IAAI,CAAC,GAAGL,IAAIG,GAAG,MAAMJ,GAAGC,MAAMK,EAAE,SAASA,IAAI,KAAK,CAAC,GAAGL,GAAG,GAAGG,GAAG,MAAMJ,GAAGC,MAAM,IAAIK,IAAI,EAAEN,GAAGC,MAAM,IAAIK,GAAG,EAAE,SAASA,IAAI,MAAM,CAAC,GAAGL,GAAG,GAAGG,GAAG,MAAMJ,GAAGC,MAAM,IAAIK,IAAI,GAAGN,GAAGC,MAAM,IAAIK,IAAI,EAAE,GAAGN,GAAGC,MAAM,IAAIK,GAAG,EAAE,KAAK,CAAC,GAAGL,GAAG,GAAGG,GAAG,MAAMJ,GAAGC,MAAM,IAAIK,IAAI,GAAGN,GAAGC,MAAM,IAAIK,IAAI,GAAG,GAAGN,GAAGC,MAAM,IAAIK,IAAI,EAAE,GAAGN,GAAGC,MAAM,IAAIK,GAAG,EAAE,CAAC,CAAC,OAAON,GAAGC,IAAI,EAAEA,GAAGE,EAAE,CAAC,SAASxL,GAAGoL,EAAEC,GAAGC,GAAG,CAAC,OAAOvL,GAAEqL,EAAEjL,GAAGkL,GAAGC,EAAE,CAAC,CAAC,IAAIrL,GAAGC,GAAGC,GAAGC,GAAG5B,GAAG8B,GAAGC,GAAGC,GAAGC,GAAG,SAASC,GAAG0K,EAAE,CAACnL,GAAGmL,EAAErT,EAAE,MAAMmI,GAAG,IAAI,UAAUkL,CAAC,EAAErT,EAAE,OAAOqI,GAAG,IAAI,WAAWgL,CAAC,EAAErT,EAAE,OAAOuI,GAAG,IAAI,WAAW8K,CAAC,EAAErT,EAAE,OAAOoI,GAAG,IAAI,WAAWiL,CAAC,EAAErT,EAAE,QAAQyG,GAAG,IAAI,YAAY4M,CAAC,EAAErT,EAAE,QAAQwI,GAAG,IAAI,YAAY6K,CAAC,EAAErT,EAAE,QAAQyI,GAAG,IAAI,aAAa4K,CAAC,EAAErT,EAAE,QAAQ0I,GAAG,IAAI,aAAa2K,CAAC,CAAC,CAAC,IAAIzK,GAAG5I,EAAE,gBAAgB,SAASiJ,GAAGC,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEzC,GAAG,CAAC,EAAEE,GAAG,GAAG,SAASC,IAAI,CAAC,OAAO9B,CAAC,CAAC,SAAS+B,IAAI,CAAC,GAAG9G,EAAE,OAAO,IAAI,OAAOA,EAAE,QAAQ,aAAaA,EAAE,OAAO,CAACA,EAAE,MAAM,GAAGA,EAAE,OAAO,QAAQgH,GAAGhH,EAAE,OAAO,MAAM,CAAC,EAAEoK,GAAGlB,EAAE,CAAC,CAAC,SAASE,IAAI,CAACxC,GAAG,GAAGwD,GAAGjB,EAAE,CAAC,CAAC,SAASpC,IAAI,CAAC,GAAG/G,EAAE,QAAQ,IAAI,OAAOA,EAAE,SAAS,aAAaA,EAAE,QAAQ,CAACA,EAAE,OAAO,GAAGA,EAAE,QAAQ,QAAQkH,GAAGlH,EAAE,QAAQ,MAAM,CAAC,EAAEoK,GAAG1D,EAAE,CAAC,CAAC,SAASM,GAAGqM,EAAE,CAACnK,GAAG,QAAQmK,CAAC,CAAC,CAAC,SAASpM,GAAGoM,EAAE,CAAClK,GAAG,QAAQkK,CAAC,CAAC,CAAC,SAASnM,GAAGmM,EAAE,CAAC3M,GAAG,QAAQ2M,CAAC,CAAC,CAAC,IAAI1M,GAAG,EAAE0C,GAAG,KAAKC,GAAG,KAAK,SAASC,GAAG8J,EAAE,CAAC1M,KAAK3G,EAAE,wBAAwBA,EAAE,uBAAuB2G,EAAE,CAAC,CAAC,SAAS6C,GAAG6J,EAAE,CAAC,GAAG1M,KAAK3G,EAAE,wBAAwBA,EAAE,uBAAuB2G,EAAE,EAAEA,IAAI,IAAI0C,KAAK,OAAO,cAAcA,EAAE,EAAEA,GAAG,MAAMC,IAAI,CAAC,IAAIgK,GAAGhK,GAAGA,GAAG,KAAKgK,GAAG,CAAC,CAAC,CAAC,SAAS7J,GAAG4J,EAAE,CAACrT,EAAE,SAASA,EAAE,QAAQqT,CAAC,EAAEA,EAAE,WAAWA,EAAE,IAAI/L,EAAE+L,CAAC,EAAEpO,EAAE,GAAGC,EAAE,EAAEmO,GAAG,2CAA2C,IAAIC,GAAG,IAAI,YAAY,aAAaD,CAAC,EAAE,MAAM,EAAEC,EAAE,EAAEA,EAAE,CAAC,IAAI5J,GAAG,wCAAwC,SAAS5B,GAAGuL,EAAE,CAAC,OAAOA,EAAE,WAAW3J,EAAE,CAAC,CAAC,SAASC,GAAG0J,EAAE,CAAC,OAAOA,EAAE,WAAW,SAAS,CAAC,CAAC,IAAIvJ,GAAGA,GAAG,yBAAyBhC,GAAGgC,EAAE,IAAIA,GAAGhH,EAAEgH,EAAE,GAAG,SAASC,GAAGsJ,EAAE,CAAC,GAAG,CAAC,GAAGA,GAAGvJ,IAAIvC,EAAE,OAAO,IAAI,WAAWA,CAAC,EAAE,GAAGF,EAAE,OAAOA,EAAEgM,CAAC,EAAE,KAAK,iDAAiD,OAAOC,GAAN,CAAU7J,GAAG6J,EAAE,CAAC,CAAC,CAAC,SAAS1J,IAAI,CAAC,GAAG,CAACrC,IAAI7E,GAAGC,GAAG,CAAC,GAAG,OAAO,OAAO,YAAY,CAACgH,GAAGG,EAAE,EAAE,OAAO,MAAMA,GAAG,CAAC,YAAY,aAAa,CAAC,EAAE,KAAK,SAASuJ,EAAE,CAAC,GAAG,CAACA,EAAE,GAAG,KAAK,uCAAuCvJ,GAAG,IAAI,OAAOuJ,EAAE,YAAY,CAAC,CAAC,EAAE,MAAM,UAAU,CAAC,OAAOtJ,GAAGD,EAAE,CAAC,CAAC,EAAE,GAAG7F,EAAE,OAAO,IAAI,QAAQ,SAASoP,EAAEC,GAAG,CAACrP,EAAE6F,GAAG,SAASyJ,GAAG,CAACF,EAAE,IAAI,WAAWE,EAAE,CAAC,CAAC,EAAED,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO,QAAQ,QAAQ,EAAE,KAAK,UAAU,CAAC,OAAOvJ,GAAGD,EAAE,CAAC,CAAC,CAAC,CAAC,SAASG,IAAI,CAAC,IAAIoJ,EAAE,CAAC,IAAI1H,GAAG,uBAAuBA,EAAE,EAAE,SAAS2H,GAAGK,GAAGC,GAAG,CAAC,IAAIC,GAAGF,GAAG,QAAQ3T,EAAE,IAAI6T,GAAG7O,EAAEhF,EAAE,IAAI,OAAO2I,GAAG3D,EAAE,MAAM,EAAEiE,GAAGjJ,EAAE,IAAI,0BAA0BiH,GAAGjH,EAAE,IAAI,iBAAiB,EAAEwJ,GAAG,kBAAkB,CAAC,CAACD,GAAG,kBAAkB,EAAE,SAASgK,GAAGI,GAAG,CAACL,GAAGK,GAAG,QAAQ,CAAC,CAAC,SAASH,GAAGG,GAAG,CAAC,OAAO/J,GAAG,EAAE,KAAK,SAASgK,GAAG,CAAC,OAAO,YAAY,YAAYA,GAAGP,CAAC,CAAC,CAAC,EAAE,KAAK,SAASO,GAAG,CAAC,OAAOA,EAAE,CAAC,EAAE,KAAKD,GAAG,SAASC,GAAG,CAACtM,EAAE,0CAA0CsM,EAAE,EAAEnK,GAAGmK,EAAE,CAAC,CAAC,CAAC,CAAC,SAASH,IAAI,CAAC,MAAM,CAAClM,GAAG,OAAO,YAAY,sBAAsB,YAAY,CAACO,GAAGgC,EAAE,GAAG,CAACH,GAAGG,EAAE,GAAG,CAAClH,GAAG,OAAO,OAAO,WAAW,MAAMkH,GAAG,CAAC,YAAY,aAAa,CAAC,EAAE,KAAK,SAAS6J,GAAG,CAAC,IAAIC,GAAG,YAAY,qBAAqBD,GAAGN,CAAC,EAAE,OAAOO,GAAG,KAAKL,GAAG,SAASM,GAAG,CAAC,OAAOvM,EAAE,kCAAkCuM,EAAE,EAAEvM,EAAE,2CAA2C,EAAEkM,GAAGD,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEC,GAAGD,EAAE,CAAC,CAAC,GAAGvT,EAAE,gBAAgB,GAAG,CAAC,IAAI0T,GAAG1T,EAAE,gBAAgBqT,EAAEC,EAAE,EAAE,OAAOI,EAAE,OAAOC,GAAN,CAAUrM,EAAE,sDAAsDqM,EAAE,EAAE,EAAEA,EAAE,CAAC,CAAC,OAAOF,GAAG,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,IAAIzJ,GAAGE,GAAG,SAASC,GAAGkJ,EAAE,CAAC,KAAK,KAAK,aAAa,KAAK,QAAQ,gCAAgCA,EAAE,IAAI,KAAK,OAAOA,CAAC,CAAC,SAASjJ,GAAGiJ,EAAE,CAAC,KAAKA,EAAE,OAAO,GAAGA,EAAE,MAAM,EAAErT,CAAC,CAAC,CAAC,SAASqK,GAAGgJ,EAAE,CAAC,OAAOA,CAAC,CAAC,SAAS/I,GAAG+I,EAAE,CAAC,IAAIC,GAAG,gBAAgB,OAAOD,EAAE,QAAQC,GAAG,SAASC,GAAG,CAAC,IAAIC,GAAGD,GAAG,OAAOA,KAAKC,GAAGD,GAAGC,GAAG,KAAKD,GAAG,GAAG,CAAC,CAAC,CAAC,SAASjL,IAAI,CAAC,IAAI+K,EAAE,IAAI,MAAM,GAAG,CAACA,EAAE,MAAM,CAAC,GAAG,CAAC,MAAM,IAAI,KAAK,OAAOC,GAAN,CAAUD,EAAEC,EAAE,CAAC,GAAG,CAACD,EAAE,MAAM,MAAM,4BAA4B,CAAC,OAAOA,EAAE,MAAM,SAAS,CAAC,CAAC,SAAS9I,GAAG8I,EAAEC,GAAG,CAACnL,GAAG,IAAIkL,EAAEC,EAAE,CAAC,CAAC,SAAS9I,IAAI,CAACf,GAAG,EAAE,CAAC,CAAC,SAASgB,IAAI,CAAC,MAAO,WAAU,CAAC,SAASC,IAAI,CAAC,OAAOD,GAAG,CAAC,CAAC,SAASE,GAAG0I,EAAEC,GAAGC,GAAG,CAACnL,GAAG,WAAWiL,EAAEC,GAAGA,GAAGC,EAAE,CAAC,CAAC,SAAS3I,GAAGyI,EAAE,CAAC,GAAG,CAAC,OAAOrO,EAAE,KAAKqO,EAAEnL,GAAG,WAAW,QAAQ,EAAE,EAAES,GAAG3D,EAAE,MAAM,EAAE,CAAC,OAAOsO,GAAN,CAAU,CAAC,CAAC,SAASzI,GAAGwI,EAAE,CAAC,IAAIC,GAAGlL,GAAG,OAAOiL,EAAEA,IAAI,EAAE,IAAIE,GAAG9I,GAAG,EAAE,GAAG4I,EAAEE,GAAG,MAAM,GAAG,IAAIC,GAAG,CAACK,GAAGC,KAAKD,IAAIC,GAAGD,GAAGC,IAAIA,GAAG,QAAQL,GAAG,EAAEA,IAAI,EAAEA,IAAI,EAAE,CAAC,IAAIC,GAAGJ,IAAI,EAAE,GAAGG,IAAIC,GAAG,KAAK,IAAIA,GAAGL,EAAE,SAAS,EAAE,IAAIM,GAAG,KAAK,IAAIJ,GAAGC,GAAG,KAAK,IAAIH,EAAEK,EAAE,EAAE,KAAK,CAAC,EAAEE,GAAGhJ,GAAG+I,EAAE,EAAE,GAAGC,GAAG,MAAM,EAAE,CAAC,MAAM,EAAE,CAAC,IAAI1I,GAAG,CAAC,QAAQ,OAAO,IAAI,UAAU,CAACA,GAAG,SAAS,EAAE,IAAImI,EAAE9K,GAAG2C,GAAG,QAAQ,GAAG,GAAG,OAAOmI,CAAC,EAAE,OAAO,SAASA,EAAE,CAAC,IAAIC,GAAG7L,EAAE4L,CAAC,EAAE,OAAOC,EAAE,CAAC,EAAE,SAASnI,GAAGkI,EAAE,CAAC,MAAO,GAAE,CAAC,SAASjI,GAAGiI,EAAEC,GAAGC,GAAGC,GAAGC,GAAG,CAAC,MAAO,GAAE,CAAC,IAAIjM,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC,EAAE,SAAS6D,GAAGgI,EAAEC,GAAG,CAAC,IAAIC,GAAG/L,GAAG6L,GAAGC,KAAK,GAAGA,KAAK,KAAKD,IAAI,EAAEzO,EAAE0C,GAAGS,EAAEwL,GAAG,CAAC,CAAC,EAAEA,GAAG,OAAO,GAAGA,GAAG,KAAKD,EAAE,CAAC,CAAC,SAAS/H,GAAG8H,EAAEC,GAAGC,GAAGC,GAAG,CAAC,QAAQC,GAAG,EAAEC,GAAG,EAAEA,GAAGH,GAAGG,KAAK,CAAC,IAAIC,GAAGnL,GAAG8K,IAAI,GAAGM,GAAGpL,GAAG8K,GAAG,GAAG,GAAGA,IAAI,EAAE,QAAQO,GAAG,EAAEA,GAAGD,GAAGC,KAAKxI,GAAGgI,EAAEjL,GAAGuL,GAAGE,GAAG,EAAEJ,IAAIG,EAAE,CAAC,OAAOpL,GAAGgL,IAAI,GAAGC,GAAG,CAAC,CAAC,SAASjI,GAAG6H,EAAE,CAAC,IAAIC,GAAGtT,EAAE,IAAIqT,GAAG,OAAOC,EAAE,CAAC,SAAS7H,GAAG4H,EAAEC,GAAGC,GAAGC,GAAGC,GAAG,CAAC,IAAIC,GAAG,CAAC,OAAOS,IAAI,CAAC,IAAIC,GAAG,EAAE,GAAGD,IAAI,MAAMA,KAAK,EAAE,CAAC,IAAIE,IAAIF,GAAG,QAAQ,GAAG,EAAEC,GAAGvB,GAAGwB,EAAE,EAAEpM,GAAGkM,GAAGC,GAAGC,EAAE,CAAC,CAAC,OAAOD,EAAE,EAAE,MAAMD,IAAI,CAAC,IAAIC,GAAGvB,GAAGsB,GAAG,MAAM,EAAE,OAAO5J,GAAG4J,GAAGC,EAAE,EAAEA,EAAE,CAAC,EAAE,SAAST,GAAGQ,GAAG,CAAC,OAAOb,KAAK,SAAS7L,EAAE0M,EAAE,EAAEb,KAAK,UAAU,QAAQa,EAAE,EAAEA,EAAE,CAAC,IAAIP,GAAGpI,GAAG6H,CAAC,EAAEQ,GAAG,CAAC,EAAEC,GAAG,EAAE,GAAGN,GAAG,QAAQO,GAAG,EAAEA,GAAGP,GAAG,OAAOO,KAAK,CAAC,IAAIC,GAAGN,GAAGH,GAAGQ,KAAKC,IAAIF,KAAK,IAAIA,GAAGnB,GAAG,GAAGkB,GAAGE,IAAIC,GAAGR,GAAGO,GAAG,GAAGF,GAAGE,IAAIP,GAAGO,GAAG,CAAC,IAAIE,GAAGL,GAAG,MAAM,KAAKC,EAAE,EAAE,SAASK,GAAGC,GAAG,CAAC,OAAOL,KAAK,GAAGlB,GAAGkB,EAAE,EAAEH,GAAGQ,EAAE,CAAC,CAAC,OAAOF,GAAGC,GAAGD,EAAE,EAAEA,EAAE,CAAC,SAASvI,GAAG2H,EAAEC,GAAGC,GAAGC,GAAG,CAACD,GAAGA,IAAI,CAAC,EAAE,IAAIE,GAAGF,GAAG,MAAMI,IAAIA,KAAK,UAAUA,KAAK,SAAS,EAAED,GAAGJ,KAAK,SAAS,OAAOI,IAAID,IAAI,CAACD,GAAGhI,GAAG6H,CAAC,EAAE,UAAU,CAAC,OAAO5H,GAAG4H,EAAEC,GAAGC,GAAG,UAAUC,EAAE,CAAC,CAAC,CAAC,IAAI7H,GAAG,CAAC,MAAMnB,GAAG,wBAAwBE,GAAG,sBAAsBC,GAAG,uBAAuBE,GAAG,SAASM,GAAG,QAAQC,GAAG,SAASG,EAAE,EAAEM,GAAG5B,GAAG,EAAE8B,GAAG/L,EAAE,mBAAmB,UAAU,CAAC,OAAO+L,GAAG/L,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAEgM,GAAGhM,EAAE,MAAM,UAAU,CAAC,OAAOgM,GAAGhM,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEgL,GAAGhL,EAAE,yBAAyB,UAAU,CAAC,OAAOgL,GAAGhL,EAAE,yBAAyBA,EAAE,IAAI,yBAAyB,MAAM,KAAK,SAAS,CAAC,EAAE6J,GAAG7J,EAAE,mBAAmB,UAAU,CAAC,OAAO6J,GAAG7J,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAEmM,GAAGnM,EAAE,iBAAiB,UAAU,CAAC,OAAOmM,GAAGnM,EAAE,iBAAiBA,EAAE,IAAI,iBAAiB,MAAM,KAAK,SAAS,CAAC,EAAEsM,GAAGtM,EAAE,cAAc,UAAU,CAAC,OAAOsM,GAAGtM,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEuM,GAAGvM,EAAE,SAAS,UAAU,CAAC,OAAOuM,GAAGvM,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEwM,GAAGxM,EAAE,KAAK,UAAU,CAAC,OAAOwM,GAAGxM,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE8L,GAAG9L,EAAE,KAAK,UAAU,CAAC,OAAO8L,GAAG9L,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE0M,GAAG1M,EAAE,MAAM,UAAU,CAAC,OAAO0M,GAAG1M,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE2M,GAAG3M,EAAE,KAAK,UAAU,CAAC,OAAO2M,GAAG3M,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE4M,GAAG5M,EAAE,KAAK,UAAU,CAAC,OAAO4M,GAAG5M,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE8M,GAAG9M,EAAE,QAAQ,UAAU,CAAC,OAAO8M,GAAG9M,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE+K,GAAG/K,EAAE,SAAS,UAAU,CAAC,OAAO+K,GAAG/K,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE+M,GAAG/M,EAAE,aAAa,UAAU,CAAC,OAAO+M,GAAG/M,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAEgN,GAAGhN,EAAE,MAAM,UAAU,CAAC,OAAOgN,GAAGhN,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEkN,GAAGlN,EAAE,aAAa,UAAU,CAAC,OAAOkN,GAAGlN,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAEmN,GAAGnN,EAAE,QAAQ,UAAU,CAAC,OAAOmN,GAAGnN,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEoN,GAAGpN,EAAE,qBAAqB,UAAU,CAAC,OAAOoN,GAAGpN,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEqN,GAAGrN,EAAE,KAAK,UAAU,CAAC,OAAOqN,GAAGrN,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEsN,GAAGtN,EAAE,MAAM,UAAU,CAAC,OAAOsN,GAAGtN,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEuN,GAAGvN,EAAE,eAAe,UAAU,CAAC,OAAOuN,GAAGvN,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAEkM,GAAGlM,EAAE,SAAS,UAAU,CAAC,OAAOkM,GAAGlM,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEyN,GAAGzN,EAAE,QAAQ,UAAU,CAAC,OAAOyN,GAAGzN,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE0N,GAAG1N,EAAE,cAAc,UAAU,CAAC,OAAO0N,GAAG1N,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE2N,GAAG3N,EAAE,uBAAuB,UAAU,CAAC,OAAO2N,GAAG3N,EAAE,uBAAuBA,EAAE,IAAI,uBAAuB,MAAM,KAAK,SAAS,CAAC,EAAE4N,GAAG5N,EAAE,KAAK,UAAU,CAAC,OAAO4N,GAAG5N,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE6N,GAAG7N,EAAE,OAAO,UAAU,CAAC,OAAO6N,GAAG7N,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAE8N,GAAG9N,EAAE,KAAK,UAAU,CAAC,OAAO8N,GAAG9N,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE+N,GAAG/N,EAAE,eAAe,UAAU,CAAC,OAAO+N,GAAG/N,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAEgO,GAAGhO,EAAE,OAAO,UAAU,CAAC,OAAOgO,GAAGhO,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEiO,GAAGjO,EAAE,UAAU,UAAU,CAAC,OAAOiO,GAAGjO,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEkO,GAAGlO,EAAE,gBAAgB,UAAU,CAAC,OAAOkO,GAAGlO,EAAE,gBAAgBA,EAAE,IAAI,gBAAgB,MAAM,KAAK,SAAS,CAAC,EAAEmO,GAAGnO,EAAE,aAAa,UAAU,CAAC,OAAOmO,GAAGnO,EAAE,aAAaA,EAAE,IAAI,aAAa,MAAM,KAAK,SAAS,CAAC,EAAE4L,GAAG5L,EAAE,sBAAsB,UAAU,CAAC,OAAO4L,GAAG5L,EAAE,sBAAsBA,EAAE,IAAI,sBAAsB,MAAM,KAAK,SAAS,CAAC,EAAEsO,GAAGtO,EAAE,QAAQ,UAAU,CAAC,OAAOsO,GAAGtO,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEuO,GAAGvO,EAAE,UAAU,UAAU,CAAC,OAAOuO,GAAGvO,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEyO,GAAGzO,EAAE,SAAS,UAAU,CAAC,OAAOyO,GAAGzO,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE0O,GAAG1O,EAAE,cAAc,UAAU,CAAC,OAAO0O,GAAG1O,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE2O,GAAG3O,EAAE,WAAW,UAAU,CAAC,OAAO2O,GAAG3O,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAE4O,GAAG5O,EAAE,MAAM,UAAU,CAAC,OAAO4O,GAAG5O,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE6O,GAAG7O,EAAE,WAAW,UAAU,CAAC,OAAO6O,GAAG7O,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAE8O,GAAG9O,EAAE,KAAK,UAAU,CAAC,OAAO8O,GAAG9O,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE+O,GAAG/O,EAAE,YAAY,UAAU,CAAC,OAAO+O,GAAG/O,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEgP,GAAGhP,EAAE,YAAY,UAAU,CAAC,OAAOgP,GAAGhP,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEiP,GAAGjP,EAAE,WAAW,UAAU,CAAC,OAAOiP,GAAGjP,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEkP,GAAGlP,EAAE,YAAY,UAAU,CAAC,OAAOkP,GAAGlP,EAAE,YAAYA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAEyP,GAAGzP,EAAE,KAAK,UAAU,CAAC,OAAOyP,GAAGzP,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEwO,GAAGxO,EAAE,SAAS,UAAU,CAAC,OAAOwO,GAAGxO,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE8K,GAAG9K,EAAE,SAAS,UAAU,CAAC,OAAO8K,GAAG9K,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE0P,GAAG1P,EAAE,MAAM,UAAU,CAAC,OAAO0P,GAAG1P,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE2P,GAAG3P,EAAE,KAAK,UAAU,CAAC,OAAO2P,GAAG3P,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAE4P,GAAG5P,EAAE,SAAS,UAAU,CAAC,OAAO4P,GAAG5P,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE6P,GAAG7P,EAAE,WAAW,UAAU,CAAC,OAAO6P,GAAG7P,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAE8P,GAAG9P,EAAE,UAAU,UAAU,CAAC,OAAO8P,GAAG9P,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAE+P,GAAG/P,EAAE,KAAK,UAAU,CAAC,OAAO+P,GAAG/P,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEgQ,GAAGhQ,EAAE,qBAAqB,UAAU,CAAC,OAAOgQ,GAAGhQ,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEiQ,GAAGjQ,EAAE,qBAAqB,UAAU,CAAC,OAAOiQ,GAAGjQ,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEkQ,GAAGlQ,EAAE,qBAAqB,UAAU,CAAC,OAAOkQ,GAAGlQ,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEmQ,GAAGnQ,EAAE,UAAU,UAAU,CAAC,OAAOmQ,GAAGnQ,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEoQ,GAAGpQ,EAAE,QAAQ,UAAU,CAAC,OAAOoQ,GAAGpQ,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEqQ,GAAGrQ,EAAE,OAAO,UAAU,CAAC,OAAOqQ,GAAGrQ,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEsQ,GAAGtQ,EAAE,KAAK,UAAU,CAAC,OAAOsQ,GAAGtQ,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEuQ,GAAGvQ,EAAE,OAAO,UAAU,CAAC,OAAOuQ,GAAGvQ,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEwQ,GAAGxQ,EAAE,MAAM,UAAU,CAAC,OAAOwQ,GAAGxQ,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEyQ,GAAGzQ,EAAE,SAAS,UAAU,CAAC,OAAOyQ,GAAGzQ,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE0Q,GAAG1Q,EAAE,MAAM,UAAU,CAAC,OAAO0Q,GAAG1Q,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE2Q,GAAG3Q,EAAE,OAAO,UAAU,CAAC,OAAO2Q,GAAG3Q,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAE4Q,GAAG5Q,EAAE,gBAAgB,UAAU,CAAC,OAAO4Q,GAAG5Q,EAAE,gBAAgBA,EAAE,IAAI,gBAAgB,MAAM,KAAK,SAAS,CAAC,EAAE6Q,GAAG7Q,EAAE,uBAAuB,UAAU,CAAC,OAAO6Q,GAAG7Q,EAAE,uBAAuBA,EAAE,IAAI,uBAAuB,MAAM,KAAK,SAAS,CAAC,EAAE8Q,GAAG9Q,EAAE,SAAS,UAAU,CAAC,OAAO8Q,GAAG9Q,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAE+Q,GAAG/Q,EAAE,kBAAkB,UAAU,CAAC,OAAO+Q,GAAG/Q,EAAE,kBAAkBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAEgR,GAAGhR,EAAE,OAAO,UAAU,CAAC,OAAOgR,GAAGhR,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEiR,GAAGjR,EAAE,OAAO,UAAU,CAAC,OAAOiR,GAAGjR,EAAE,OAAOA,EAAE,IAAI,OAAO,MAAM,KAAK,SAAS,CAAC,EAAEkR,GAAGlR,EAAE,WAAW,UAAU,CAAC,OAAOkR,GAAGlR,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEmR,GAAGnR,EAAE,UAAU,UAAU,CAAC,OAAOmR,GAAGnR,EAAE,UAAUA,EAAE,IAAI,UAAU,MAAM,KAAK,SAAS,CAAC,EAAEoR,GAAGpR,EAAE,SAAS,UAAU,CAAC,OAAOoR,GAAGpR,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEqR,GAAGrR,EAAE,KAAK,UAAU,CAAC,OAAOqR,GAAGrR,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEsR,GAAGtR,EAAE,SAAS,UAAU,CAAC,OAAOsR,GAAGtR,EAAE,SAASA,EAAE,IAAI,SAAS,MAAM,KAAK,SAAS,CAAC,EAAEuR,GAAGvR,EAAE,qBAAqB,UAAU,CAAC,OAAOuR,GAAGvR,EAAE,qBAAqBA,EAAE,IAAI,qBAAqB,MAAM,KAAK,SAAS,CAAC,EAAEwR,GAAGxR,EAAE,eAAe,UAAU,CAAC,OAAOwR,GAAGxR,EAAE,eAAeA,EAAE,IAAI,eAAe,MAAM,KAAK,SAAS,CAAC,EAAEyR,GAAGzR,EAAE,wBAAwB,UAAU,CAAC,OAAOyR,GAAGzR,EAAE,wBAAwBA,EAAE,IAAI,wBAAwB,MAAM,KAAK,SAAS,CAAC,EAAE0R,GAAG1R,EAAE,MAAM,UAAU,CAAC,OAAO0R,GAAG1R,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE2R,GAAG3R,EAAE,QAAQ,UAAU,CAAC,OAAO2R,GAAG3R,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAE4R,GAAG5R,EAAE,mBAAmB,UAAU,CAAC,OAAO4R,GAAG5R,EAAE,mBAAmBA,EAAE,IAAI,mBAAmB,MAAM,KAAK,SAAS,CAAC,EAAE6R,GAAG7R,EAAE,MAAM,UAAU,CAAC,OAAO6R,GAAG7R,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE8R,GAAG9R,EAAE,cAAc,UAAU,CAAC,OAAO8R,GAAG9R,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE+R,GAAG/R,EAAE,KAAK,UAAU,CAAC,OAAO+R,GAAG/R,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEgS,GAAGhS,EAAE,KAAK,UAAU,CAAC,OAAOgS,GAAGhS,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEiS,GAAGjS,EAAE,KAAK,UAAU,CAAC,OAAOiS,GAAGjS,EAAE,KAAKA,EAAE,IAAI,KAAK,MAAM,KAAK,SAAS,CAAC,EAAEkS,GAAGlS,EAAE,MAAM,UAAU,CAAC,OAAOkS,GAAGlS,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEmS,GAAGnS,EAAE,MAAM,UAAU,CAAC,OAAOmS,GAAGnS,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEoS,GAAGpS,EAAE,MAAM,UAAU,CAAC,OAAOoS,GAAGpS,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAEqS,GAAGrS,EAAE,WAAW,UAAU,CAAC,OAAOqS,GAAGrS,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEsS,GAAGtS,EAAE,WAAW,UAAU,CAAC,OAAOsS,GAAGtS,EAAE,WAAWA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAEuS,GAAGvS,EAAE,cAAc,UAAU,CAAC,OAAOuS,GAAGvS,EAAE,cAAcA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEwS,GAAGxS,EAAE,QAAQ,UAAU,CAAC,OAAOwS,GAAGxS,EAAE,QAAQA,EAAE,IAAI,QAAQ,MAAM,KAAK,SAAS,CAAC,EAAEyS,GAAGzS,EAAE,MAAM,UAAU,CAAC,OAAOyS,GAAGzS,EAAE,MAAMA,EAAE,IAAI,MAAM,MAAM,KAAK,SAAS,CAAC,EAAE0S,GAAG1S,EAAE,kBAAkB,UAAU,CAAC,OAAO0S,GAAG1S,EAAE,kBAAkBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAE2S,GAAG3S,EAAE,UAAU,UAAU,CAAC,OAAO2S,GAAG3S,EAAE,UAAUA,EAAE,IAAI,WAAW,MAAM,KAAK,SAAS,CAAC,EAAE4S,GAAG5S,EAAE,aAAa,UAAU,CAAC,OAAO4S,GAAG5S,EAAE,aAAaA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAE6S,GAAG7S,EAAE,WAAW,UAAU,CAAC,OAAO6S,GAAG7S,EAAE,WAAWA,EAAE,IAAI,YAAY,MAAM,KAAK,SAAS,CAAC,EAAE8S,GAAG9S,EAAE,iBAAiB,UAAU,CAAC,OAAO8S,GAAG9S,EAAE,iBAAiBA,EAAE,IAAI,kBAAkB,MAAM,KAAK,SAAS,CAAC,EAAE+S,GAAG/S,EAAE,aAAa,UAAU,CAAC,OAAO+S,GAAG/S,EAAE,aAAaA,EAAE,IAAI,cAAc,MAAM,KAAK,SAAS,CAAC,EAAEA,EAAE,MAAM0L,GAAG,IAAIsH,GAAG1J,GAAG,SAAS+J,GAAG,CAACL,IAAIC,GAAG,EAAED,KAAK1J,GAAG+J,EAAE,EAAE,SAASJ,GAAGI,EAAE,CAAC,GAAGA,EAAEA,GAAGtR,EAAE4E,GAAG,IAAIG,GAAG,EAAEH,GAAG,GAAG,OAAO,SAAS2M,IAAI,CAACN,KAAKA,GAAG,GAAGhT,EAAE,UAAU,GAAG,CAACiF,IAAImE,GAAG,EAAEhJ,EAAEJ,CAAC,EAAEA,EAAE,sBAAsBA,EAAE,qBAAqB,EAAE+G,GAAG,GAAG,CAAC/G,EAAE,WAAWA,EAAE,UAAU,YAAY,EAAE,WAAW,UAAU,CAAC,WAAW,UAAU,CAACA,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC,EAAEsT,GAAG,CAAC,EAAE,CAAC,GAAGA,GAAG,CAAC,CAAC,GAAGtT,EAAE,QAAQ,IAAI,OAAOA,EAAE,SAAS,aAAaA,EAAE,QAAQ,CAACA,EAAE,OAAO,GAAGA,EAAE,QAAQ,OAAO,GAAGA,EAAE,QAAQ,IAAI,EAAE,EAAEiT,GAAG,EAAE,IAAIC,GAAGrR,IAAIqR,GAAG,CAAC,kBAAkB,QAAQ,UAAU,mBAAmB,EAAE,OAAO,SAASG,EAAE,CAAC,MAAM,CAACxR,EAAE,kBAAkB,QAAQwR,CAAC,EAAE,EAAE,CAAC,EAAE,mBAAmB,QAAQ,UAAU,oBAAoB,EAAE,OAAO,SAASA,EAAE,CAAC,MAAM,CAACxR,EAAE,mBAAmB,QAAQwR,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,IAAIF,GAAG,GAAG,OAAOvS,GAAG,YAAYuS,GAAGvS,UAAU,OAAO,+BAA+B,YAAYuS,GAAG,kCAAmC,OAAM,IAAI,MAAM,uCAAuC,EAAE,GAAGD,GAAG,CAAC,IAAIE,GAAGD,GAAG,SAASA,GAAG,SAAS,UAAU,CAACC,GAAG,EAAEF,GAAG,kBAAkB,QAAQ,SAASG,EAAE,CAAC,QAAQ,eAAe,oBAAoBA,CAAC,CAAC,CAAC,EAAEH,GAAG,mBAAmB,QAAQ,SAASG,EAAE,CAAC,QAAQ,eAAe,qBAAqBA,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOzS,EAAE,KAAK,CAAC,GAAG,EAAE,OAAOkW,GAAI,UAAU,OAAOC,GAAI,SAASA,EAAG,QAAQC,EAAG,OAAO,QAAQ,YAAY,OAAO,IAAI,OAAO,CAAC,EAAE,UAAU,CAAC,OAAOA,CAAE,CAAC,EAAE,OAAOF,GAAI,WAAWA,EAAG,kBAAkBE,EAAG,CAAC,EAAMC,GAAG,KAAK,CAAC,YAAYrW,EAAEZ,EAAE,CAAC,KAAK,QAAQY,EAAE,KAAK,UAAUZ,EAAE,KAAK,KAAK,IAAI,QAAQ,KAAK,aAAa,CAAC,CAAC,IAAIY,EAAE,CAAC,OAAO,KAAK,KAAK,IAAIA,CAAC,GAAG,KAAK,UAAU,SAAS,KAAK,QAAQA,CAAC,EAAE,KAAK,KAAK,IAAIA,CAAC,CAAC,CAAC,IAAIA,EAAEZ,EAAE,CAAC,KAAK,eAAe,KAAK,KAAK,IAAIY,EAAEZ,CAAC,CAAC,CAAC,IAAIY,EAAE,CAAC,OAAO,KAAK,KAAK,IAAIA,CAAC,CAAC,CAAC,OAAOA,EAAE,CAAC,OAAO,KAAK,eAAe,KAAK,KAAK,OAAOA,CAAC,CAAC,CAAC,YAAY,CAAC,OAAO,KAAK,YAAY,CAAC,EAAEsW,GAAG,KAAK,CAAC,SAAStW,EAAE,CAAC,OAAOuW,GAAG,UAAU,CAAC,CAAC,OAAOvW,EAAE,CAAC,OAAOuW,GAAG,QAAQ,CAAC,CAAC,gBAAgB,CAAC,MAAM,EAAE,CAAC,KAAKvW,EAAE,CAAC,OAAOuW,GAAG,MAAM,CAAC,CAAC,KAAKvW,EAAE,CAAC,OAAOuW,GAAG,MAAM,CAAC,CAAC,SAASvW,EAAE,CAAC,OAAOuW,GAAG,UAAU,CAAC,CAAC,UAAUvW,EAAEZ,EAAE,CAAC,OAAOmX,GAAG,WAAW,CAAC,CAAC,YAAY,CAAC,OAAOA,GAAG,YAAY,CAAC,CAAC,YAAYvW,EAAEZ,EAAE,CAAC,OAAOmX,GAAG,aAAa,CAAC,CAAC,MAAMvW,EAAEZ,EAAE,EAAE,CAAC,OAAOmX,GAAG,OAAO,CAAC,CAAC,KAAKvW,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOsV,GAAG,MAAM,CAAC,CAAC,wBAAwBvW,EAAEZ,EAAE,EAAE,CAAC,OAAOmX,GAAG,yBAAyB,CAAC,CAAC,QAAQ,CAAC,OAAOA,GAAG,QAAQ,CAAC,CAAC,gBAAgB,CAAC,OAAOA,GAAG,gBAAgB,CAAC,CAAC,SAAS,CAAC,OAAO,KAAK,eAAe,IAAI,GAAG,KAAK,IAAI,CAAC,SAAS,CAAC,OAAOA,GAAG,SAAS,CAAC,CAAC,EAAE,SAASA,GAAGrX,EAAE,CAAC,MAAM,IAAI,MAAM,IAAIA,2HAA2H,CAAC,CAAC,SAASsX,GAAGtX,EAAE,CAAC,IAAI,EAAEA,EAAE,OAAOE,EAAE,EAAE,KAAK,EAAE,GAAGA,EAAE,KAAK,OAAO,EAAE,EAAE,EAAE,IAAIqX,GAAGvX,EAAE,EAAEE,CAAC,CAAC,CAAC,SAASsX,GAAGxX,EAAE,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,MAAM,IAAI,MAAM,yEAAyEA,EAAE,iCAAiC,EAAE,QAAQ,EAAE,IAAIE,EAAEF,EAAE,OAAOM,EAAE,EAAE,KAAKJ,EAAE,GAAGI,EAAE,KAAK,OAAO,EAAEJ,EAAE,EAAEA,IAAIqX,GAAGvX,EAAEE,EAAEI,CAAC,EAAEiX,GAAG,EAAErX,EAAEI,CAAC,CAAC,CAAC,SAASmX,GAAGzX,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,IAAIF,EAAE,KAAK,IAAI,EAAEE,CAAC,CAAC,CAAC,CAAC,SAASwX,GAAG1X,EAAE,CAAC,OAAOA,EAAE,IAAI,EAAEA,EAAEA,EAAE,CAAC,CAAC,SAASuX,GAAGvX,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAGA,EAAE,GAAGA,EAAEE,GAAGF,EAAEE,GAAGI,CAAC,CAAC,SAASqX,GAAG3X,EAAE,CAAC,IAAI,EAAE,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,GAAGF,EAAEE,GAAG,OAAO,CAAC,CAAC,SAAS0X,GAAG5X,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,OAAO,EAAE,OAAO,EAAEA,GAAG,EAAEA,GAAGF,CAAC,CAAC,SAAS6X,GAAG7X,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAI,CAAC,IAAIC,EAAE,OAAOP,EAAEM,EAAE,EAAE,OAAO,EAAEA,EAAE,EAAEJ,GAAGK,EAAEA,CAAC,CAAC,OAAOL,CAAC,CAAC,SAAS4X,EAAE9X,EAAE,EAAE,CAAC,GAAG,CAACA,EAAE,MAAM,IAAI,MAAM,OAAO,GAAG,SAAS,EAAE,EAAE,CAAC,CAAC,CAAC,SAAS+X,GAAG/X,EAAE,EAAEE,EAAE,GAAG,CAAC4X,EAAEE,GAAGhY,EAAE,CAAC,EAAE,IAAIE,EAAE,WAAWF,SAAS,cAAc,CAAC,CAAC,SAASiY,GAAGjY,EAAE,CAAC8X,EAAE9X,GAAG,KAAK,IAAI,+DAA+D,CAAC,CAAC,SAASkY,GAAGlY,EAAE,EAAE,CAAC,EAAEE,EAAE,GAAG,CAAC,GAAG,GAAG,OAAO,EAAE,CAAC,GAAG,MAAM,QAAQF,CAAC,GAAGmY,GAAGnY,CAAC,GAAG,CAACE,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAO,EAAEM,EAAE4X,GAAGlY,EAAEM,GAAG,EAAEJ,CAAC,OAAO,EAAE,KAAKF,CAAC,EAAE,OAAO,CAAC,CAAC,SAASoY,GAAGpY,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAO,GAAE,IAAI,EAAEA,EAAE,GAAG,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,GAAGF,EAAEE,GAAG,OAAO,CAAC,CAAC,SAASmY,GAAGrY,EAAE,CAAC,OAAOA,EAAE,SAAS,CAAC,CAAC,SAASgY,GAAGhY,EAAE,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,GAAG,GAAGA,GAAG,MAAM,GAAG,MAAMA,EAAE,SAAS,EAAE,OAAO,MAAM,GAAG,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,GAAGF,EAAEE,KAAK,EAAEA,GAAG,MAAM,GAAG,MAAM,EAAE,CAAC,SAASoY,GAAGtY,EAAE,CAAC,OAAOA,EAAE,IAAI,CAAC,CAAC,SAASuY,GAAGvY,EAAE,CAAC,GAAG,KAAK,MAAM,KAAK,OAAO,KAAK,KAAKA,CAAC,EAAE,GAAGA,IAAI,EAAE,EAAE,MAAO,GAAE,GAAGA,IAAI,GAAG,EAAE,MAAM,GAAG,CAAC,IAAI,EAAE,KAAK,IAAI,EAAEA,CAAC,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,CAAC,CAAC,SAASwY,GAAGxY,EAAE,CAAC,IAAI,EAAE,KAAK,KAAK,KAAK,KAAKA,CAAC,CAAC,EAAE,MAAM,CAAC,EAAE,KAAK,KAAKA,EAAE,CAAC,CAAC,CAAC,CAAC,SAASyY,GAAGzY,EAAE,CAAC,IAAI,EAAE,IAAI,YAAYA,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,EAAEE,EAAE,EAAEA,GAAGA,EAAE,OAAOoX,GAAG,CAAC,EAAE,CAAC,CAAC,SAASoB,GAAG1Y,EAAE,EAAE,CAAC,OAAO,GAAGA,EAAE,OAAOA,EAAEA,EAAE,IAAI,OAAO,EAAEA,EAAE,MAAM,CAAC,CAAC,SAAS2Y,GAAG3Y,EAAE,EAAEO,GAAG,EAAEL,EAAEI,EAAE,CAAC,OAAO,IAAI,QAAQ,CAACC,EAAE,IAAI,CAAC,IAAI,EAAE,EAAE0B,EAAE,IAAI,CAAC,GAAGjC,EAAE,EAAE,CAACO,EAAE,EAAE,MAAM,CAAC,IAAI,IAAI2B,EAAE,EAAE,CAAC,EAAE,GAAGhC,GAAG,MAAM,GAAGA,EAAE,CAAC,EAAE,EAAE,MAAM,CAACI,GAAG,KAAKA,EAAE2B,EAAEC,CAAC,EAAE,WAAWD,EAAEC,CAAC,CAAC,EAAED,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS2W,GAAG5Y,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAEI,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAEN,EAAE,OAAO,EAAE,EAAE,GAAGA,EAAE,IAAI,EAAEE,GAAGF,EAAE,WAAWA,EAAE,KAAK,GAAG,CAAC,GAAGM,IAAI,GAAG,MAAM,MAAM,yDAAyDA,aAAa,GAAG,EAAEA,EAAE,CAAC,SAASN,EAAE,GAAG,EAAE,MAAM,MAAM,gCAAgCA,EAAE,aAAa,GAAG,EAAE,GAAGM,IAAI,GAAG,CAAC,GAAG,EAAE,GAAG,IAAIJ,EAAE,MAAM,MAAM,QAAQ,sCAAsCF,GAAG,EAAE,OAAOA,CAAC,CAAC,GAAGE,IAAI,EAAE,MAAM,MAAM,qCAAqCF,8BAA8B,EAAE,GAAG,EAAEE,IAAI,EAAE,MAAM,MAAM,wDAAwD,OAAOA,GAAG,EAAE,IAAIK,EAAEP,EAAE,MAAM,EAAE,OAAOO,EAAED,GAAG,EAAEJ,EAAEK,CAAC,CAAC,SAASsY,GAAG7Y,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAE,OAAO,OAAOF,EAAEA,GAAG,KAAK,EAAE,IAAI,CAACM,EAAEC,IAAIA,CAAC,EAAE,CAAC,EAAE,OAAOP,CAAC,EAAE8X,EAAE9X,EAAE,MAAMM,GAAGA,GAAG,CAACJ,GAAGI,EAAEJ,CAAC,EAAE,IAAI,+CAA+CA,MAAMA,mBAAmBF,GAAG,EAAE8X,EAAE9X,EAAE,MAAMM,GAAGgY,GAAGhY,CAAC,CAAC,EAAE,IAAI,0DAA0DN,GAAG,EAAEA,EAAE,IAAIM,GAAGA,EAAE,EAAEJ,EAAEI,EAAEA,CAAC,CAAC,CAAC,SAASwY,GAAG9Y,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAEC,EAAE,GAAG,MAAM,MAAM,QAAQ,CAAC,GAAG,EAAE,SAAS,EAAE,EAAE,GAAG,MAAMA,EAAE,KAAKsY,GAAG,EAAE7Y,CAAC,EAAE,KAAK,EAAE,EAAE,EAAE,QAAQiC,EAAE,EAAEA,EAAEjC,EAAE,OAAO,EAAEiC,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,GAAG,EAAE,KAAKA,GAAGjC,EAAEiC,KAAK,EAAE,MAAM,IAAI,MAAM,sBAAsBA,oBAAoBjC,EAAEiC,cAAc,GAAG,EAAE,IAAI,MAAM,EAAE,GAAGA,IAAIjC,EAAEiC,KAAK,IAAI/B,EAAE,KAAKF,EAAEiC,EAAE,EAAE3B,EAAE,KAAK2B,CAAC,GAAG,EAAE,IAAIA,GAAG,GAAG,CAACjC,EAAEiC,KAAK,IAAI/B,EAAE,KAAKF,EAAEiC,EAAE,EAAE3B,EAAE,KAAK2B,CAAC,EAAE,CAAC,MAAM,CAAC,SAAS/B,EAAE,SAASI,CAAC,CAAC,CAAC,SAASyY,GAAG/Y,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,GAAGF,GAAG,MAAMA,IAAI,UAAUE,EAAE,IAAI,aAAa,CAAC,UAAUF,IAAI,QAAQE,EAAE,IAAI,WAAW,CAAC,UAAUF,IAAI,OAAOE,EAAE,IAAI,WAAW,CAAC,MAAO,OAAM,IAAI,MAAM,qBAAqBF,GAAG,EAAE,OAAOE,CAAC,CAAC,SAAS8Y,GAAGhZ,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,GAAGF,GAAG,MAAMA,IAAI,UAAUE,EAAE,IAAI,aAAa,CAAC,UAAUF,IAAI,QAAQE,EAAE,IAAI,WAAW,CAAC,UAAUF,IAAI,OAAOE,EAAE,IAAI,WAAW,CAAC,UAAUF,IAAI,SAASE,EAAE,IAAI,MAAM,CAAC,MAAO,OAAM,IAAI,MAAM,qBAAqBF,GAAG,EAAE,OAAOE,CAAC,CAAC,SAAS+Y,GAAGjZ,EAAE,EAAE,CAAC,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,CAAC,IAAII,EAAEN,EAAEE,GAAG,GAAG,MAAMI,CAAC,GAAG,CAAC,SAASA,CAAC,EAAE,MAAM,MAAM,oBAAoB,6BAA6BA,IAAI,CAAC,CAAC,CAAC,SAAS4Y,GAAGlZ,EAAE,CAAC,OAAOA,IAAI,QAAQA,IAAI,aAAaA,IAAI,WAAWA,IAAI,SAASA,IAAI,QAAQ,CAAC,SAASmZ,GAAGnZ,EAAE,EAAE,CAAC,MAAM,EAAE,IAAI,aAAa,IAAI,WAAWA,IAAI,aAAa,IAAI,SAASA,IAAI,WAAWA,IAAI,aAAa,IAAI,QAAQA,IAAI,OAAO,CAAC,SAASmY,GAAGnY,EAAE,CAAC,OAAOA,aAAa,cAAcA,aAAa,YAAYA,aAAa,YAAYA,aAAa,iBAAiB,CAAC,SAASoZ,GAAGpZ,EAAE,CAAC,GAAGA,IAAI,WAAWA,IAAI,QAAQ,MAAO,GAAE,GAAGA,IAAI,YAAY,MAAO,GAAE,GAAGA,IAAI,OAAO,MAAO,GAAE,MAAM,IAAI,MAAM,iBAAiBA,GAAG,CAAC,CAAC,SAASqZ,GAAGrZ,EAAE,CAAC,GAAGA,GAAG,KAAK,MAAO,GAAE,IAAI,EAAE,EAAE,OAAOA,EAAE,QAAQE,GAAG,GAAGA,EAAE,MAAM,EAAE,CAAC,CAAC,SAASoZ,GAAGtZ,EAAE,CAAC,OAAO,OAAOA,GAAG,UAAUA,aAAa,MAAM,CAAC,SAASuZ,GAAGvZ,EAAE,CAAC,OAAO,OAAOA,GAAG,SAAS,CAAC,SAASwZ,GAAGxZ,EAAE,CAAC,OAAO,OAAOA,GAAG,QAAQ,CAAC,SAASyZ,GAAGzZ,EAAE,CAAC,OAAO,MAAM,QAAQA,CAAC,EAAEyZ,GAAGzZ,EAAE,EAAE,EAAEA,aAAa,aAAa,UAAUA,aAAa,YAAYA,aAAa,YAAYA,aAAa,kBAAkB,QAAQwZ,GAAGxZ,CAAC,EAAE,UAAUsZ,GAAGtZ,CAAC,EAAE,SAASuZ,GAAGvZ,CAAC,EAAE,OAAO,SAAS,CAAC,SAAS0Z,GAAG1Z,EAAE,CAAC,MAAM,CAAC,EAAEA,GAAGA,EAAE,aAAaA,EAAE,MAAMA,EAAE,MAAM,CAAC,SAAS2Z,GAAG3Z,EAAE,EAAE,CAAC,QAAQE,EAAE,EAAEA,EAAEF,EAAE,EAAEE,EAAE,GAAGF,EAAEE,IAAI,EAAE,OAAOA,EAAE,OAAOF,CAAC,CAAC,SAAS4Z,GAAG5Z,EAAE,CAAC,IAAI,EAAEA,EAAE,OAAO,GAAG,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIE,EAAE,IAAI,MAAM,EAAE,CAAC,EAAEA,EAAE,EAAE,GAAGF,EAAE,EAAE,GAAG,QAAQM,EAAE,EAAE,EAAEA,GAAG,EAAE,EAAEA,EAAEJ,EAAEI,GAAGJ,EAAEI,EAAE,GAAGN,EAAEM,EAAE,GAAG,OAAOJ,CAAC,CAAC,SAAS2Z,GAAG7Z,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAE,IAAI,MAAM,GAAG,EAAE,SAAS,EAAE,CAAC,IAAI,EAAE,EAAE,IAAID,EAAE,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAE,EAAE,IAAIC,EAAE,GAAGL,EAAEF,EAAE,EAAE,KAAK,CAAC,IAAI,EAAE,EAAE,GAAG,EAAE,EAAE,MAAM,CAAC,EAAEiC,EAAE,EAAE,OAAO,CAACC,EAAEC,IAAID,EAAEC,CAAC,GAAG7B,EAAE,EAAE,GAAG,QAAQ4B,EAAE,EAAEA,EAAE,EAAEA,IAAI3B,EAAE2B,GAAG2X,GAAG7Z,EAAEkC,EAAED,EAAE,EAAE/B,EAAEI,CAAC,CAAC,CAAC,OAAOC,CAAC,CAAC,SAASuZ,GAAG9Z,EAAE,EAAEE,EAAE,GAAG,CAAC,GAAGF,EAAE,SAAS,EAAE,OAAO,EAAE,GAAG,IAAIM,EAAEN,EAAE,OAAO,CAACO,EAAE,IAAIA,EAAE,CAAC,GAAGL,EAAE,EAAE,GAAG,GAAGI,IAAI,EAAE,MAAM,CAAC,EAAE,GAAGA,IAAI,EAAE,OAAO,MAAM,IAAI,MAAM,IAAIN,oCAAoC,EAAE,SAASE,EAAE,wBAAwB,KAAK,EAAE,OAAO2Z,GAAG,EAAE7Z,EAAE,EAAEE,CAAC,CAAC,CAAC,SAAS6Z,GAAG/Z,EAAE,EAAE,CAAC,IAAIE,EAAE8Z,GAAGha,EAAE,CAAC,EAAE,QAAQM,EAAE,EAAEA,EAAEJ,EAAE,OAAOI,IAAIJ,EAAEI,GAAG,EAAE,OAAOJ,CAAC,CAAC,SAAS8Z,GAAGha,EAAE,EAAE,CAAC,GAAG,GAAG,MAAM,IAAI,WAAW,IAAI,YAAY,OAAO,IAAI,aAAaA,CAAC,EAAE,GAAG,IAAI,QAAQ,OAAO,IAAI,WAAWA,CAAC,EAAE,GAAG,IAAI,OAAO,OAAO,IAAI,WAAWA,CAAC,EAAE,MAAM,IAAI,MAAM,qBAAqB,GAAG,CAAC,CAAC,SAASia,GAAGja,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAO,CAACM,EAAEC,IAAID,EAAEC,EAAE,CAAC,EAAE,GAAG,GAAG,MAAM,IAAI,UAAU,OAAOuZ,GAAG9Z,EAAE,IAAI,aAAaE,CAAC,CAAC,EAAE,GAAG,IAAI,QAAQ,OAAO4Z,GAAG9Z,EAAE,IAAI,WAAWE,CAAC,CAAC,EAAE,GAAG,IAAI,OAAO,OAAO4Z,GAAG9Z,EAAE,IAAI,WAAWE,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,qBAAqB,GAAG,CAAC,CAAC,SAASga,GAAGla,EAAE,CAACA,EAAE,QAAQ,GAAG,CAAC8X,EAAE,OAAO,UAAU,CAAC,GAAG,GAAG,EAAE,IAAI,0EAA0E9X,KAAK,CAAC,CAAC,CAAC,CAAC,SAASma,GAAGna,EAAE,EAAEE,EAAE,CAAC,GAAG,IAAI,EAAE,MAAO,GAAE,GAAG,IAAI,EAAE,OAAOF,EAAE,GAAG,IAAIM,EAAEN,EAAEA,EAAE,OAAO,GAAG,QAAQO,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAE,EAAEO,EAAED,GAAGJ,EAAEK,GAAGP,EAAEO,GAAG,OAAOD,CAAC,CAAC,SAAS8Z,GAAGpa,EAAE,EAAEE,EAAE,CAAC,GAAG,IAAI,EAAE,MAAM,CAAC,EAAE,GAAG,IAAI,EAAE,MAAM,CAACF,CAAC,EAAE,IAAIM,EAAE,IAAI,MAAM,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAE,EAAEC,EAAED,EAAEC,GAAG,KAAK,MAAMP,EAAEE,EAAEK,EAAE,EAAEP,GAAGM,EAAEC,GAAGL,EAAEK,GAAG,OAAOD,EAAEA,EAAE,OAAO,GAAGN,EAAEM,CAAC,CAAC,SAAS+Z,GAAGra,EAAE,CAAC,OAAOA,GAAGA,EAAE,MAAM,OAAOA,EAAE,MAAM,UAAU,CAAC,IAAIsa,GAAG,YAAYC,GAAG,KAAK,CAAC,YAAYzZ,EAAE,CAAC,KAAK,OAAOA,EAAE,KAAK,MAAM,CAAC,EAAE,KAAK,aAAa,CAAC,EAAE,KAAK,SAAS,CAAC,EAAE,KAAK,eAAe0Z,GAAG,KAAK,iBAAiB,CAAC,CAAC,YAAY1Z,EAAEZ,EAAE,CAAC,KAAK,UAAU,OAAOua,EAAE,EAAE,QAAQ,SAAS,GAAGA,EAAE,EAAE,QAAQ,MAAM,GAAG,QAAQ,KAAK,YAAY,KAAK,oEAAoE3Z,IAAI,GAAG,KAAK,aAAaA,EAAE,KAAK,SAASZ,CAAC,CAAC,aAAaY,EAAEZ,EAAE,EAAE,CAAC,GAAG,KAAK,aAAaY,GAAG,CAAC,aAAaZ,EAAE,QAAQ,CAAC,EAAE,KAAK,SAASY,IAAI,KAAK,CAAC,IAAIP,EAAE,KAAK,SAASO,GAAG2Z,EAAE,EAAE,QAAQ,SAAS,GAAGA,EAAE,EAAE,QAAQ,MAAM,GAAG,QAAQ,KAAK,qCAAqC3Z,MAAMP,IAAI,EAAE,KAAK,IAAIO,EAAEP,CAAC,CAAC,CAAC,CAAC,MAAM,SAASO,EAAE,CAAC,OAAOA,KAAK,KAAK,MAAM,KAAK,MAAMA,IAAI,KAAK,MAAMA,GAAG,MAAM,KAAK,aAAaA,CAAC,EAAE,KAAK,MAAMA,GAAG,CAAC,IAAIA,EAAE,CAAC,GAAGA,KAAK,KAAK,MAAM,OAAO,KAAK,MAAMA,GAAG,IAAIZ,EAAE,KAAK,aAAaY,CAAC,EAAE,GAAGuZ,GAAGna,CAAC,EAAE,MAAM,IAAI,MAAM,QAAQY,qEAAqE,EAAE,OAAO,KAAK,MAAMA,GAAGZ,EAAE,KAAK,MAAMY,EAAE,CAAC,UAAUA,EAAE,CAAC,OAAO,KAAK,IAAIA,CAAC,CAAC,CAAC,QAAQA,EAAE,CAAC,OAAO,KAAK,IAAIA,CAAC,CAAC,CAAC,UAAU,CAAC,OAAO,KAAK,KAAK,CAAC,IAAI,UAAU,CAAC,OAAO,KAAK,KAAK,CAAC,IAAIA,EAAEZ,EAAE,CAAC,GAAG,KAAK,aAAaY,IAAI,KAAK,MAAM,IAAI,MAAM,mBAAmBA,kCAAkC,EAAE,KAAK,MAAMA,GAAGZ,EAAE,KAAK,aAAaY,GAAG,SAAS,MAAM,KAAK,aAAaA,GAAG,QAAQZ,CAAC,CAAC,CAAC,aAAaY,EAAE,CAAC,GAAG,KAAK,aAAaA,IAAI,KAAK,MAAM,IAAI,MAAM,yBAAyBA,mCAAmC,EAAE,OAAO,KAAK,aAAaA,GAAG,aAAa,CAAC,CAAC,SAASA,EAAE,CAAC,KAAK,MAAM,OAAO,OAAO,CAAC,EAAEA,CAAC,CAAC,CAAC,OAAO,CAAC,KAAK,MAAM,CAAC,EAAE,KAAK,SAAS,CAAC,EAAE,KAAK,iBAAiB,CAAC,CAAC,kBAAkB,CAAC,GAAG,OAAO,KAAK,QAAQ,aAAa,OAAO,KAAK,OAAO,UAAU,aAAa,OAAO,KAAK,OAAO,SAAS,QAAQ,YAAY,OAAO,IAAIA,EAAE,KAAK,eAAe,KAAK,OAAO,SAAS,MAAM,EAAEwZ,MAAMxZ,GAAGA,EAAEwZ,IAAI,MAAM,GAAG,EAAE,QAAQha,GAAG,CAAC,GAAG,CAACC,EAAEwB,CAAC,EAAEzB,EAAE,MAAM,GAAG,EAAE,KAAK,SAASC,GAAGma,GAAGna,EAAEwB,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASyY,GAAGxa,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,OAAOA,EAAE,QAAQ,8BAA8B,CAACE,KAAKI,KAAKqa,GAAG,EAAEra,EAAE,GAAGA,EAAE,EAAE,EAAEA,EAAE,KAAK,GAAG,EAAE,EAAE,CAAC,CAAC,SAASqa,GAAG3a,EAAE,EAAEE,EAAE,CAACF,EAAE,mBAAmB,CAAC,GAAG,mBAAmBE,GAAG,EAAE,CAAC,CAAC,SAASwa,GAAG1a,EAAE,EAAE,CAAC,GAAG,EAAE,EAAE,YAAY,EAAE,IAAI,QAAQ,IAAI,QAAQ,OAAO,IAAI,OAAO,GAAG,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,MAAM,IAAI,MAAM,oCAAoC,cAAcA,IAAI,CAAC,CAAC,SAASya,GAAG,CAAC,OAAOG,EAAE,CAAC,IAAIA,GAAG,KAAK,SAASC,GAAG7a,EAAE,CAAC4a,GAAG5a,CAAC,CAAC,IAAI8a,GAAG,SAASC,IAAI,CAAC,GAAGD,IAAI,KAAK,CAAC,IAAI9a,EAAE,GAAG,OAAO,QAAQ,YAAYA,EAAE,eAAe,OAAO,QAAQ,YAAYA,EAAE,eAAe,OAAO,SAAS,YAAYA,EAAE,gBAAgB,OAAO,MAAM,YAAYA,EAAE,SAAU,OAAM,IAAI,MAAM,gCAAgC,EAAE8a,GAAG9a,CAAC,CAAC,OAAO8a,EAAE,CAAC,SAASE,IAAI,CAAC,IAAIhb,EAAE+a,GAAG,EAAE,OAAO/a,EAAE,YAAY,OAAOA,EAAE,WAAW,IAAI,KAAKA,EAAE,UAAU,CAAC,SAASib,GAAGjb,EAAE,EAAE,CAAC,IAAIE,EAAE8a,GAAG,EAAE,GAAG9a,EAAE,IAAIF,CAAC,EAAE,OAAOE,EAAE,IAAIF,CAAC,EAAE,CAAC,IAAIM,EAAE,EAAE,EAAE,OAAOJ,EAAE,IAAIF,EAAEM,CAAC,EAAEJ,EAAE,IAAIF,CAAC,CAAC,CAAC,CAAC,IAAIkb,GAAG,MAAMC,GAAG,OAAOC,GAAG,QAAQC,GAAG,MAAMC,GAAG,OAAOC,GAAG,MAAMC,GAAG,MAAMC,GAAG,SAASC,GAAG,SAASC,GAAG,OAAOC,GAAG,QAAQC,GAAG,OAAOC,GAAG,QAAQC,GAAG,QAAQC,GAAG,UAAUC,GAAG,cAAcC,GAAG,YAAYC,GAAG,gBAAgBC,GAAG,cAAcC,GAAG,iBAAiBC,GAAG,WAAWC,GAAG,cAAcC,GAAG,gBAAgBC,GAAG,OAAOC,GAAG,OAAOC,GAAG,cAAcC,GAAG,UAAUC,GAAG,aAAaC,GAAG,SAASC,GAAG,SAASC,GAAG,uBAAuBC,GAAG,sBAAsBC,GAAG,SAASC,GAAG,yBAAyBC,GAAG,wBAAwBC,GAAG,MAAMC,GAAG,OAAOC,GAAG,UAAUC,GAAG,SAASC,GAAG,gBAAgBC,GAAG,gBAAgBC,GAAG,eAAeC,GAAG,wBAAwBC,GAAG,sCAAsCC,GAAG,qCAAqCC,GAAG,OAAOC,GAAG,aAAaC,GAAG,0BAA0BC,GAAG,2BAA2BC,GAAG,UAAUC,GAAG,SAASC,GAAG,MAAMC,GAAG,UAAUC,GAAG,MAAMC,GAAG,QAAQC,GAAG,MAAMC,GAAG,aAAaC,GAAG,QAAQC,GAAG,MAAMC,GAAG,OAAOC,GAAG,gBAAgBC,GAAG,QAAQC,GAAG,WAAWC,GAAG,iBAAiBC,GAAG,WAAWC,GAAG,WAAWC,GAAG,UAAUC,GAAG,eAAeC,GAAG,WAAWC,GAAG,OAAOC,GAAG,OAAOC,GAAG,WAAWC,GAAG,QAAQC,GAAG,QAAQC,GAAG,YAAYC,GAAG,OAAOC,GAAG,YAAYC,GAAG,WAAWC,GAAG,MAAMC,GAAG,QAAQC,GAAG,aAAaC,GAAG,aAAaC,GAAG,YAAYC,GAAG,aAAaC,GAAG,aAAaC,GAAI,aAAaC,GAAG,MAAMC,GAAG,UAAUC,GAAG,MAAMC,GAAG,UAAUC,GAAG,UAAUC,GAAG,cAAcC,GAAG,YAAYC,GAAG,gBAAgBC,GAAG,oBAAoBC,GAAG,OAAOC,GAAG,MAAMC,GAAG,UAAUC,GAAG,YAAYC,GAAG,MAAMC,GAAG,cAAcC,GAAG,WAAWC,GAAG,MAAMC,GAAG,WAAWC,GAAG,sBAAsBC,GAAG,sBAAsBC,GAAG,sBAAsBC,GAAG,WAAWC,GAAG,SAASC,GAAG,OAAOC,GAAG,QAAQC,GAAI,OAAOC,GAAG,MAAMC,GAAG,QAAQC,GAAG,OAAOC,GAAG,eAAeC,GAAG,cAAcC,GAAG,uBAAuBC,GAAG,QAAQC,GAAG,OAAOC,GAAG,aAAaC,GAAG,OAAOC,GAAG,UAAUC,GAAG,wBAAwBC,GAAG,4BAA4BC,GAAG,iBAAiBC,GAAG,qBAAqBC,GAAG,QAAQC,GAAG,UAAUC,GAAG,QAAQC,GAAG,QAAQC,GAAG,YAAYC,GAAG,eAAeC,GAAG,SAASC,GAAG,OAAOC,GAAG,QAAQC,GAAG,MAAMC,GAAG,OAAOC,GAAG,OAAOC,GAAG,UAAUC,GAAG,WAAWC,GAAG,OAAOC,GAAG,MAAMC,GAAG,iBAAiBC,GAAG,SAASC,GAAG,UAAUC,GAAG,sBAAsBC,GAAG,gBAAgBC,GAAG,oBAAoBC,GAAG,mBAAmBC,GAAG,gBAAgBC,GAAG,oBAAoBC,GAAG,SAASC,GAAG,eAAeC,GAAG,eAAeC,GAAG,cAAcC,GAAG,yBAAyBC,GAAG,MAAMC,GAAG,MAAMC,GAAG,OAAOC,GAAG,OAAOC,GAAG,OAAOC,GAAG,YAAYC,GAAG,YAAYC,GAAG,SAASC,GAAG,SAASC,GAAG,qBAAqBC,GAAI,aAAaC,GAAG,YAAYC,GAAG,OAAOC,GAAG,aAAaC,GAAG,mBAAmBC,GAAG,eAAeC,GAAG,cAAcC,GAAG,uBAAuB,SAASC,MAAMjmB,EAAE,CAACya,EAAE,EAAE,QAAQ,SAAS,GAAGA,EAAE,EAAE,QAAQ,MAAM,GAAG,QAAQ,KAAK,GAAGza,CAAC,CAAC,CAAC,SAASkmB,MAAMlmB,EAAE,CAACya,EAAE,EAAE,QAAQ,SAAS,GAAGA,EAAE,EAAE,QAAQ,MAAM,GAAG,QAAQ,IAAI,GAAGza,CAAC,CAAC,CAAC,IAAImmB,GAAGlL,GAAG,iBAAiB,IAAI,IAAI,GAAG,EAAEmL,GAAGnL,GAAG,eAAe,IAAI,IAAI,GAAG,EAAE,SAASoL,GAAGrmB,EAAE,EAAE,CAAC,IAAIE,EAAEomB,GAAGtmB,EAAE,CAAC,EAAE,OAAOmmB,GAAG,IAAIjmB,CAAC,CAAC,CAAC,SAASqmB,GAAGvmB,EAAE,CAAC,OAAOomB,GAAG,IAAIpmB,CAAC,CAAC,CAAC,SAASwmB,GAAGxmB,EAAE,CAAC,IAAI,EAAEmmB,GAAG,QAAQ,EAAEjmB,EAAE,CAAC,EAAE,OAAO,CAAC,GAAG,CAAC,KAAKI,EAAE,MAAMC,CAAC,EAAE,EAAE,KAAK,EAAE,GAAGD,EAAE,MAAM,GAAG,CAAC,EAAE,CAAC,EAAEC,EAAE,CAAC0B,CAAC,EAAE,EAAE,MAAM,GAAG,EAAEA,IAAIjC,GAAGE,EAAE,KAAK,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,SAASumB,GAAGzmB,EAAE,CAAC,GAAG,CAAC,WAAW,EAAE,YAAYE,CAAC,EAAEF,EAAEM,EAAEgmB,GAAG,EAAEpmB,CAAC,EAAEimB,GAAG,IAAI7lB,CAAC,GAAG2lB,GAAG,eAAe,mBAAmB/lB,0BAA0B,EAAEimB,GAAG,IAAI7lB,EAAEN,CAAC,CAAC,CAAC,SAAS0mB,GAAG1mB,EAAE,CAAC,GAAG,CAAC,WAAW,CAAC,EAAEA,EAAEomB,GAAG,IAAI,CAAC,GAAG3L,EAAE,EAAE,QAAQ,OAAO,GAAGwL,GAAG,gCAAgC,IAAI,EAAEG,GAAG,IAAI,EAAEpmB,CAAC,CAAC,CAAC,SAAS2mB,GAAI3mB,EAAE,EAAE,CAAC,IAAIE,EAAEomB,GAAGtmB,EAAE,CAAC,EAAE,GAAG,CAACmmB,GAAG,IAAIjmB,CAAC,EAAE,MAAM,IAAI,MAAM,eAAeF,mBAAmB,sBAAsB,EAAEmmB,GAAG,OAAOjmB,CAAC,CAAC,CAAC,SAAS0mB,GAAI5mB,EAAE,CAAC,GAAG,CAAComB,GAAG,IAAIpmB,CAAC,EAAE,MAAM,IAAI,MAAM,iBAAiBA,kCAAkC,EAAEomB,GAAG,OAAOpmB,CAAC,CAAC,CAAC,SAAS6mB,GAAI7mB,EAAE,EAAE,CAACwmB,GAAGxmB,CAAC,EAAE,QAAQ,GAAG,CAAC,IAAIO,EAAE,OAAO,OAAO,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,CAAC,EAAEkmB,GAAGlmB,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS+lB,GAAGtmB,EAAE,EAAE,CAAC,MAAM,GAAG,KAAKA,GAAG,CAAC,IAAI8mB,EAAE,CAAC,EAAE1mB,GAAG0mB,EAAE,CAAC,YAAY,IAAI9O,GAAG,OAAO,IAAIF,EAAE,mCAAmC,IAAIoC,GAAG,cAAc,IAAIjC,GAAG,kBAAkB,IAAIF,GAAG,qBAAqB,IAAIsB,GAAG,gBAAgB,IAAID,GAAG,yBAAyB,IAAIH,GAAG,MAAM,IAAIxB,GAAG,eAAe,IAAImC,GAAG,kBAAkB,IAAImN,GAAG,sBAAsB,IAAItO,GAAG,aAAa,IAAIuO,GAAG,YAAY,IAAInP,GAAG,aAAa,IAAIoP,GAAG,MAAM,IAAIC,GAAG,cAAc,IAAIC,GAAG,QAAQ,IAAIjP,GAAG,kBAAkB,IAAIc,GAAG,uBAAuB,IAAID,GAAG,gBAAgB,IAAII,GAAG,UAAU,IAAIiO,GAAG,WAAW,IAAIhN,GAAG,WAAW,IAAIX,GAAG,uBAAuB,IAAIb,GAAG,UAAU,IAAIW,GAAG,WAAW,IAAIG,GAAG,MAAM,IAAIpB,GAAG,SAAS,IAAIkB,GAAG,UAAU,IAAIa,GAAG,cAAc,IAAIhC,GAAG,SAAS,IAAIiB,GAAG,aAAa,IAAInB,GAAG,aAAa,IAAIe,GAAG,WAAW,IAAIiB,GAAG,mBAAmB,IAAIJ,GAAG,0BAA0B,IAAIE,GAAG,oBAAoB,IAAID,GAAG,eAAe,IAAIL,GAAG,kBAAkB,IAAIjC,GAAG,IAAI,IAAI2P,GAAG,eAAe,IAAIxO,GAAG,YAAY,IAAIjB,GAAG,YAAY,IAAIe,GAAG,SAAS,IAAID,GAAG,QAAQ,IAAIpB,GAAG,aAAa,IAAIE,GAAG,cAAc,IAAIY,GAAG,oBAAoB,IAAII,GAAG,aAAa,IAAIM,GAAG,IAAI,IAAInB,GAAG,KAAK,IAAIJ,GAAG,KAAK,IAAIgB,GAAG,cAAc,IAAIuB,GAAG,aAAa,IAAIwN,EAAE,CAAC,EAAE,IAAIC,GAAG/mB,GAAGC,GAAG,CAAC,EAAM+mB,GAAGD,GAAG,SAASA,GAAG,SAASH,GAAGpnB,EAAE,CAAC,OAAOwnB,GAAG,WAAWxnB,EAAE,GAAG,EAAE,CAAC,CAAC,IAAIynB,GAAGL,GAAG,kBAAkB,EAAEM,GAAGN,GAAG,kBAAkB,EAAEO,GAAGP,GAAG,kBAAkB,EAAE,SAASQ,GAAG5nB,EAAE,CAAC,OAAOA,EAAE,IAAIA,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS6nB,GAAG7nB,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,EAAE,EAAEE,CAAC,EAAE,OAAOsnB,GAAG,UAAU,MAAM,KAAKlnB,CAAC,EAAE,GAAG,EAAE,CAAC,CAAC,SAASwnB,GAAG9nB,EAAE,EAAE,CAAC,OAAO6nB,GAAG7nB,EAAE,EAAE,CAAC,CAAC,CAAC,SAAS+nB,GAAG/nB,EAAE,EAAE,CAAC,OAAO6nB,GAAG7nB,EAAE,EAAE,CAAC,CAAC,CAAC,SAASgoB,GAAGhoB,EAAE,EAAE,CAAC,OAAO,IAAI,EAAEA,EAAEA,EAAE,KAAK,CAAC,EAAE,GAAGA,EAAE,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC,SAASioB,GAAGjoB,EAAE,EAAEE,EAAEknB,GAAG,kBAAkB,EAAE,CAAC,IAAI9mB,EAAEN,EAAE,IAAI,CAAC,EAAE,IAAIE,CAAC,EAAEI,EAAEA,EAAE,IAAIA,EAAE,KAAK,EAAE,CAAC,EAAE,IAAIC,EAAE,EAAE,IAAID,CAAC,EAAE,IAAIJ,CAAC,EAAE,OAAOK,EAAEA,EAAE,IAAIA,EAAE,KAAK,EAAE,CAAC,EAAEA,EAAEA,EAAE,IAAIL,CAAC,EAAEK,CAAC,CAAC,SAAS2nB,GAAGloB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAACA,EAAEA,EAAE,IAAIP,CAAC,EAAE,EAAEgoB,GAAG,EAAE,IAAIznB,CAAC,EAAE,IAAID,CAAC,EAAE,EAAE,EAAE,IAAI,EAAEC,EAAE,OAAOA,EAAEA,EAAE,IAAI,CAAC,EAAEA,EAAEA,EAAE,IAAIL,CAAC,EAAE,EAAE,EAAE,IAAI8nB,GAAGznB,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,IAAID,CAAC,EAAE,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,SAAS6nB,GAAGnoB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO4nB,GAAGJ,GAAG9nB,EAAE,CAAC,EAAE8nB,GAAG9nB,EAAE,EAAE,CAAC,EAAE8nB,GAAG9nB,EAAE,EAAE,EAAE,EAAE8nB,GAAG9nB,EAAE,EAAE,EAAE,EAAEE,EAAEI,CAAC,CAAC,CAAC,SAAS8nB,GAAGpoB,EAAE,EAAEA,EAAE,OAAO,CAAC,GAAG,GAAG,EAAE,CAAC,IAAIE,EAAEynB,GAAG,IAAI,EAAE,CAAC,EAAErnB,EAAEwnB,GAAG9nB,EAAE,CAAC,EAAE,IAAI2nB,EAAE,EAAEpnB,EAAEunB,GAAG9nB,EAAE,EAAE,CAAC,EAAE,EAAEgoB,GAAGznB,EAAE,EAAE,EAAE,IAAIL,CAAC,EAAE,IAAII,CAAC,EAAE,EAAE0nB,GAAG1nB,EAAE,EAAE,EAAE,IAAIC,CAAC,EAAE,IAAIL,CAAC,EAAE,OAAO+nB,GAAG,EAAE,EAAE/nB,CAAC,CAAC,CAAC,GAAG,GAAG,EAAE,CAAC,IAAIA,EAAEynB,GAAG,IAAI,EAAE,CAAC,EAAErnB,EAAEynB,GAAG/nB,EAAE,CAAC,EAAE,OAAOioB,GAAG3nB,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC,EAAEynB,GAAG/nB,EAAE,EAAE,CAAC,EAAEE,CAAC,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,IAAIA,EAAEF,EAAE,GAAGM,EAAEN,EAAE,GAAG,GAAGO,EAAEP,EAAE,EAAE,GAAG,EAAEE,GAAGI,GAAG,GAAG,EAAE,GAAGC,GAAG,GAAG,OAAOqnB,GAAGD,GAAG,IAAI,CAAC,EAAE,IAAIF,GAAG,IAAI,CAAC,CAAC,CAAC,EAAE,IAAIE,EAAE,CAAC,CAAC,OAAOA,EAAE,CAAC,SAASU,GAAGroB,EAAE,EAAEA,EAAE,OAAO,CAAC,IAAIE,EAAEynB,GAAG,IAAI,EAAE,CAAC,EAAErnB,EAAEwnB,GAAG9nB,EAAE,CAAC,EAAE,IAAI0nB,EAAE,EAAEnnB,EAAEunB,GAAG9nB,EAAE,CAAC,EAAE,EAAE8nB,GAAG9nB,EAAE,EAAE,CAAC,EAAE,IAAIE,CAAC,EAAE,EAAE4nB,GAAG9nB,EAAE,EAAE,EAAE,EAAE,IAAI2nB,EAAE,EAAE,OAAOM,GAAGD,GAAG1nB,EAAE,IAAIC,CAAC,EAAE,EAAE,EAAE,IAAIynB,GAAG,EAAE,EAAE,CAAC,EAAE,IAAI,CAAC,EAAE1nB,EAAE,IAAI0nB,GAAGznB,EAAE,IAAIonB,EAAE,EAAE,EAAE,CAAC,EAAE,IAAI,CAAC,EAAEznB,CAAC,CAAC,CAAC,SAASooB,GAAGtoB,EAAE,EAAEA,EAAE,OAAO,CAAC,IAAIE,EAAEynB,GAAG,IAAI,EAAE,CAAC,EAAErnB,EAAEwnB,GAAG9nB,EAAE,CAAC,EAAE,IAAI2nB,EAAE,EAAEpnB,EAAEunB,GAAG9nB,EAAE,CAAC,EAAE,EAAE8nB,GAAG9nB,EAAE,EAAE,CAAC,EAAE,IAAIE,CAAC,EAAE,EAAE4nB,GAAG9nB,EAAE,EAAE,EAAE,EAAE,IAAI2nB,EAAE,EAAE1lB,EAAE+lB,GAAG1nB,EAAE,IAAIC,CAAC,EAAE,EAAE,EAAE,IAAIynB,GAAG,EAAE,EAAE,CAAC,EAAE,IAAI,CAAC,EAAE9lB,EAAE+lB,GAAGhmB,EAAE3B,EAAE,IAAI0nB,GAAGznB,EAAE,IAAIonB,EAAE,EAAE,EAAE,CAAC,EAAE,IAAI,CAAC,EAAEznB,CAAC,EAAEiC,EAAE2lB,GAAG9nB,EAAE,EAAE,EAAE,IAAIE,CAAC,EAAE0C,EAAEklB,GAAG9nB,EAAE,EAAE,EAAE6C,EAAEZ,EAAE,IAAI6lB,GAAG9nB,EAAE,EAAE,EAAE,CAAC,EAAE,IAAIE,CAAC,EAAE4C,EAAEZ,EAAE,IAAI4lB,GAAG9nB,EAAE,EAAE,EAAE,CAAC,EAAE,IAAIE,CAAC,EAAE,OAAO+nB,GAAGD,GAAG7lB,EAAE,IAAIS,CAAC,EAAE,EAAE,EAAE,IAAIolB,GAAGnlB,EAAE,EAAE,CAAC,EAAE,IAAIC,CAAC,EAAEX,EAAE,IAAI6lB,GAAGplB,EAAE,IAAItC,CAAC,EAAE,EAAE,CAAC,EAAE,IAAIuC,CAAC,EAAE3C,CAAC,CAAC,CAAC,SAASinB,GAAGnnB,EAAE,EAAEA,EAAE,OAAO,CAAC,IAAIE,EAAEsnB,GAAG,WAAW,GAAG,EAAE,EAAE,GAAG,GAAG,GAAG,OAAO,GAAG,GAAGY,GAAGpoB,EAAE,CAAC,EAAEqoB,GAAGroB,EAAE,CAAC,EAAE,GAAG,GAAG,GAAG,OAAOsoB,GAAGtoB,EAAE,CAAC,EAAE,IAAIM,EAAEJ,EAAEK,EAAEL,EAAE,IAAIwnB,EAAE,EAAE,IAAI,GAAG,EAAE,EAAEE,GAAGrnB,EAAE,IAAIonB,EAAE,EAAE,IAAI,GAAG,CAAC,EAAE,IAAIA,EAAE,EAAE,EAAE,CAACH,GAAG,MAAMA,GAAG,KAAK,EAAEvlB,EAAE,CAACulB,GAAG,MAAMA,GAAG,KAAK,EAAElnB,EAAEA,EAAE,IAAIqnB,EAAE,EAAE,IAAIG,GAAG9nB,EAAE,CAAC,CAAC,EAAE,IAAIkC,EAAE,EAAEC,GAAG,EAAE,GAAG,GAAG,GAAGS,EAAET,GAAG,EAAE,EAAE,IAAI,GAAG,GAAG7B,EAAE0nB,GAAG1nB,EAAE,IAAIC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,IAAIwlB,EAAE,EAAEnnB,EAAEynB,GAAGznB,EAAE,IAAI,EAAE,EAAE,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,IAAIwlB,EAAE,EAAEpnB,EAAEA,EAAE,IAAI2B,EAAE,EAAE,EAAE1B,EAAEA,EAAE,IAAI,EAAE,EAAE,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,EAAE,EAAE8lB,GAAG,EAAE,IAAI/lB,EAAE,EAAE,EAAE,EAAE,EAAE,IAAIylB,EAAE,EAAE,EAAES,GAAGnoB,EAAEkC,EAAE,EAAE,GAAG,IAAIwlB,EAAE,EAAEpnB,EAAE,IAAI2B,EAAE,EAAE,CAAC,EAAEA,EAAEkmB,GAAGnoB,EAAEkC,EAAE,GAAG,EAAE,IAAID,EAAE,EAAE,EAAE1B,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE5B,CAAC,EAAE,CAACA,EAAE,CAAC,EAAE4B,GAAG,SAASA,IAAIC,GAAG,IAAIU,EAAE6kB,GAAG,IAAI,EAAE,IAAI,GAAG,EAAE,IAAI,CAAC,CAAC,EAAE,OAAOxlB,EAAEU,EAAEX,EAAE,GAAGA,EAAE,GAAG,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,IAAIA,EAAE,EAAE,EAAEA,EAAE,GAAGA,EAAE,GAAG,IAAI,EAAE,EAAE,EAAE3B,EAAE0nB,GAAG1nB,EAAE,IAAIC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,IAAIW,CAAC,EAAEtC,EAAEynB,GAAGznB,EAAE,IAAI,EAAE,EAAE,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,IAAIW,CAAC,EAAEvC,EAAEA,EAAE,IAAI2B,EAAE,GAAG,IAAI,CAAC,CAAC,EAAE1B,EAAEA,EAAE,IAAI,EAAE,GAAG,IAAI,CAAC,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,CAAC,EAAE,EAAE8lB,GAAG,EAAE,IAAI/lB,EAAE,EAAE,EAAE,EAAE,EAAE,IAAIY,CAAC,EAAE,EAAEslB,GAAGnoB,EAAEkC,EAAE,EAAE,GAAG,IAAIW,CAAC,EAAEvC,EAAE,IAAI2B,EAAE,EAAE,CAAC,EAAEA,EAAEkmB,GAAGnoB,EAAEkC,EAAE,GAAG,EAAE,IAAID,EAAE,EAAE,EAAE1B,EAAE,IAAIunB,GAAG9nB,EAAEkC,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE5B,CAAC,EAAE,CAACA,EAAE,CAAC,EAAE2nB,GAAGA,GAAG,EAAE,GAAGhmB,EAAE,GAAGY,CAAC,EAAE,IAAI+kB,GAAGrnB,CAAC,EAAE,IAAIknB,EAAE,CAAC,EAAE,IAAI,CAAC,EAAEQ,GAAG,EAAE,GAAGhmB,EAAE,GAAGY,CAAC,EAAE,IAAIvC,CAAC,EAAEuC,CAAC,CAAC,CAAC,SAASkkB,GAAG/mB,EAAE,EAAE,CAAC,OAAO,IAAI,SAASinB,GAAGjnB,CAAC,EAAEsnB,GAAG,CAACtnB,CAAC,EAAE,CAAC,CAAC,CAAC,SAASuoB,GAAGvoB,EAAE,EAAE,CAAC,OAAOA,aAAa,cAAc,IAAI,WAAWA,aAAa,YAAY,IAAI,SAASA,aAAa,YAAY,IAAI,MAAM,CAAC,SAASsnB,GAAGtnB,EAAE,EAAE,CAAC,GAAG,IAAI,SAAS,MAAM,IAAI,MAAM,2CAA2C,EAAE,GAAG,MAAM,QAAQA,CAAC,IAAIA,EAAEkY,GAAGlY,CAAC,GAAGya,EAAE,EAAE,QAAQ,OAAO,GAAGxB,GAAGjZ,EAAE,CAAC,EAAEuoB,GAAGvoB,EAAE,CAAC,EAAE,OAAOA,EAAE,GAAG,GAAG,MAAM,IAAI,WAAW,IAAI,YAAY,OAAO,IAAI,aAAaA,CAAC,EAAE,GAAG,IAAI,QAAQ,OAAO,IAAI,WAAWA,CAAC,EAAE,GAAG,IAAI,OAAO,CAAC,IAAIE,EAAE,IAAI,WAAWF,EAAE,MAAM,EAAE,QAAQM,EAAE,EAAEA,EAAEJ,EAAE,OAAO,EAAEI,EAAE,KAAK,MAAMN,EAAEM,EAAE,IAAI,IAAIJ,EAAEI,GAAG,GAAG,OAAOJ,CAAC,KAAM,OAAM,IAAI,MAAM,qBAAqB,GAAG,CAAC,CAAC,SAASmnB,IAAI,CAAC,OAAO5M,EAAE,EAAE,SAAS,IAAI,CAAC,CAAC,SAASyM,GAAGlnB,EAAE,EAAE,CAAC,OAAOya,EAAE,EAAE,SAAS,MAAMza,EAAE,CAAC,CAAC,CAAC,SAASinB,GAAGjnB,EAAE,EAAE,QAAQ,CAAC,OAAO,EAAE,GAAG,QAAQya,EAAE,EAAE,SAAS,OAAOza,EAAE,CAAC,CAAC,CAAC,SAASgnB,GAAGhnB,EAAE,EAAE,QAAQ,CAAC,OAAO,EAAE,GAAG,QAAQya,EAAE,EAAE,SAAS,OAAOza,EAAE,CAAC,CAAC,CAAC,IAAIwoB,GAAG,KAAK,CAAC,YAAY1nB,EAAEZ,EAAE,CAAC,KAAK,aAAaY,EAAE,KAAK,OAAOZ,EAAEA,GAAG,OAAO,KAAK,OAAO,IAAIuoB,GAAG,CAAC,cAAc3nB,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAEwB,EAAE,IAAI,CAACxB,EAAE,EAAE,CAAC,EAAEyB,EAAEC,EAAEolB,GAAG,EAAE,GAAG,KAAK,aAAa,eAAe,EAAErlB,EAAE,KAAK,aAAa,KAAKD,CAAC,MAAM,CAACA,EAAE,EAAE,QAAQI,KAAK5B,EAAE4B,EAAE,SAAS,EAAEH,EAAE,QAAQ,QAAQ,CAAC,SAASqlB,GAAG,EAAEplB,CAAC,CAAC,CAAC,CAAC,GAAGwY,EAAE,EAAE,QAAQ,8BAA8B,EAAE,QAAQtY,EAAE,EAAEA,EAAE5B,EAAE,OAAO4B,IAAI,CAAC,IAAIS,EAAErC,EAAE4B,GAAGS,EAAE,KAAK,EAAE,KAAKC,GAAG,CAAC6lB,GAAG7lB,EAAED,EAAE,MAAM9B,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,WAAWA,EAAE,QAAQP,EAAE,OAAOL,EAAE,OAAO8B,EAAE,KAAKG,GAAGA,EAAE,QAAQ,EAAE,UAAUH,EAAE,KAAKG,GAAGA,EAAE,qBAAqB,KAAKA,EAAE,oBAAoB,EAAE,EAAE,CAAC,CAAC,CAAC,iBAAiBrB,EAAE,CAAC,GAAG,CAAC,WAAWZ,EAAE,QAAQ,EAAE,OAAOK,EAAE,OAAOwB,EAAE,UAAUC,CAAC,EAAElB,EAAE,EAAE,QAAQmB,GAAG,CAAC,QAAQ,IAAI,CAACA,EAAE,KAAK,EAAE1B,EAAEyB,CAAC,CAAC,EAAE,KAAKE,GAAG,CAAC,KAAK,OAAO,iBAAiBhC,EAAE+B,EAAEC,EAAE,GAAGA,EAAE,GAAGH,EAAEG,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASwmB,GAAG1oB,EAAE,EAAEE,EAAE,CAAC,GAAG,IAAI,UAAU,MAAM,GAAG,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAI,CAAC,IAAIC,EAAEP,EAAEM,GAAG,GAAG,MAAMC,CAAC,GAAG,CAAC,SAASA,CAAC,EAAE,OAAO,QAAQ,KAAK,SAASA,uBAAuBL,IAAI,EAAE,EAAE,CAAC,MAAM,EAAE,CAAC,IAAIuoB,GAAG,KAAK,CAAC,iBAAiB3nB,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,IAAIC,EAAE,OAAO1B,GAAG,SAASmY,GAAG,GAAGnY,MAAM,CAAC,EAAEA,EAAE,MAAM2B,EAAEwW,GAAG5X,EAAE,EAAE,EAAE,EAAEZ,EAAE,KAAK0C,EAAE1C,EAAE,KAAK,EAAEwY,GAAGxY,EAAE,MAAM,SAAS,EAAE,EAAE,EAAE4C,EAAE,GAAG,QAAQC,KAAKhB,EAAE,CAAC,IAAIiB,EAAEjB,EAAEgB,GAAG,GAAGC,GAAG,KAAK,CAAC,IAAIkB,EAAElB,EAAE,OAAO9C,EAAE,MAAMiE,EAAED,EAAE,OAAOpB,GAAG,GAAGC,MAAMoB,MAAMA,EAAE,EAAED,EAAE,KAAK,CAAC,CAAC,QAAQ,IAAI,KAAKhC,OAAOD,OAAO,MAAM,OAAOW,OAAOE,OAAOd,IAAI,mBAAmB,YAAY,aAAa,gBAAgB,eAAe,kBAAkB,CAAC,CAAC,EAAE,SAAS2mB,GAAG3oB,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQ2B,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI5B,EAAE,EAAE4B,GAAG,IAAI,GAAG,QAAQA,EAAE,EAAEA,EAAElC,EAAE,OAAOkC,IAAI,CAAC,IAAIC,EAAEnC,EAAEkC,GAAGU,EAAET,EAAE,OAAO,QAAQU,KAAKD,EAAE,CAAC,IAAIE,EAAEF,EAAEC,GAAGE,EAAE,GAAG,QAAQC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,GAAG1C,EAAEwC,EAAE,IAAI,CAACX,EAAE,QAAQ,QAAQ+B,GAAG5D,EAAE4D,EAAE,IAAI,EAAE,EAAEnB,EAAE,GAAGxC,EAAE4B,EAAE,IAAI,GAAG,KAAK,CAAC,GAAGY,EAAE,KAAK,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,EAAE7C,EAAE,IAAI,GAAG,IAAI,EAAE,CAAC,EAAE,QAAQgC,EAAElC,EAAE,OAAO,EAAEkC,GAAG,EAAEA,IAAI,CAAC,IAAIC,EAAEnC,EAAEkC,GAAGU,EAAET,EAAE,OAAO,QAAQU,EAAE,EAAEA,EAAEV,EAAE,QAAQ,OAAOU,IAAI,GAAG,EAAEV,EAAE,QAAQU,GAAG,IAAI,CAAC,QAAQC,KAAKF,EAAE,EAAEA,EAAEE,GAAG,IAAI,GAAG,EAAEX,EAAE,IAAI,GAAG,KAAK,CAAC,CAAC,IAAIF,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAElC,EAAE,OAAOkC,IAAI,CAAC,IAAIC,EAAEnC,EAAEkC,GAAG,GAAG3B,EAAE4B,EAAE,KAAK,EAAEA,EAAE,IAAI,CAAC,IAAIS,EAAE,CAAC,EAAE,QAAQE,KAAKX,EAAE,OAAO,CAAC,IAAIY,EAAEZ,EAAE,OAAOW,GAAGxC,EAAEyC,EAAE,MAAMH,EAAEE,GAAGC,EAAE,CAAC,IAAIF,EAAE,OAAO,OAAO,CAAC,EAAEV,CAAC,EAAEU,EAAE,OAAOD,EAAEC,EAAE,QAAQV,EAAE,QAAQF,EAAE,KAAKY,CAAC,CAAC,CAAC,CAAC,OAAOZ,CAAC,CAAC,SAAS2mB,GAAG5oB,EAAE,EAAEE,EAAEI,EAAE,CAAC,QAAQC,EAAE,EAAE,OAAO,EAAEA,GAAG,EAAEA,IAAI,CAAC,IAAI,EAAE,EAAEA,GAAG,EAAE,CAAC,EAAE,GAAG,EAAE,QAAQ,QAAQ2B,GAAG,CAAC,IAAIC,EAAEnC,EAAEkC,EAAE,IAAIC,GAAG,KAAK,EAAE,KAAKA,CAAC,EAAE,EAAE,KAAK,IAAI,CAAC,CAAC,EAAE,EAAE,UAAU,KAAK,MAAM,IAAI,MAAM,4DAA4D,EAAE,aAAa,EAAE,IAAIF,EAAE,EAAE,SAAS,CAAC,EAAE,QAAQC,KAAK,EAAE,OAAO,CAAC,GAAG,EAAEA,KAAKD,GAAG,MAAM,IAAI,MAAM,iCAAiCC,iCAAiC,OAAO,KAAKD,CAAC,IAAI,EAAE,IAAIE,EAAEjC,EAAE,IAAI+B,EAAEC,GAAG,CAAC,EAAE,GAAGC,EAAE,QAAQ,UAAU,MAAM,IAAI,MAAM,4BAA4B,EAAE,qCAAqCD,yCAAyCC,EAAE,QAAQ,EAAE,IAAIS,EAAE,EAAE,OAAOV,GAAG,GAAG,CAAC8V,GAAG7V,EAAE,MAAMS,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,4BAA4B,EAAE,sCAAsCV,iBAAiBC,EAAE,wDAAwDS,EAAE,QAAQ,EAAE,GAAG5C,EAAE4C,EAAE,KAAK,KAAK5C,EAAE4C,EAAE,IAAIT,MAAM,CAAC,IAAIU,EAAE7C,EAAE4C,EAAE,IAAI5C,EAAE4C,EAAE,IAAItC,EAAEuC,EAAEV,CAAC,EAAEU,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIgmB,GAAG,GAAGC,GAAG,EAAEC,GAAG,EAAE,SAASC,GAAGhpB,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEqZ,GAAG,CAAC,EAAE,EAAEqP,GAAGjpB,EAAE,EAAEE,EAAEK,CAAC,EAAE,EAAE,EAAE,OAAO0B,EAAEinB,GAAGlpB,EAAE,EAAEE,EAAEK,EAAE,CAAC,EAAE2B,EAAE,CAAC,QAAQ,EAAE,OAAO5B,IAAI4B,EAAE,KAAK,YAAYhC,GAAG,EAAEgC,EAAE,KAAK,WAAW,GAAG,EAAEA,EAAE,KAAK,aAAa,IAAI,EAAEA,EAAE,KAAK,WAAW,GAAGA,EAAE,KAAKD,EAAE,IAAIE,GAAG,OAAOA,CAAC,EAAE,KAAK;AAAA,CACr50C,CAAC,EAAED,EAAE,KAAK;AAAA,CACV,CAAC,CAAC,SAAS+mB,GAAGjpB,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE6X,GAAG,CAAC,EAAE,EAAE9X,EAAEA,EAAE,OAAO,GAAG,EAAE,IAAI,MAAM,CAAC,EAAE,KAAK,CAAC,EAAE2B,EAAE,EAAE,OAAOC,EAAEhC,IAAI,YAAYipB,GAAGnpB,CAAC,EAAEA,EAAE,GAAGiC,EAAE,EAAE,QAAQE,EAAE,EAAEA,EAAE5B,EAAE,EAAE4B,IAAI,CAAC,IAAIS,EAAET,EAAE,EAAE,QAAQU,EAAE,EAAEA,EAAE,EAAEA,IAAI,EAAEA,GAAG,KAAK,IAAI,EAAEA,GAAGumB,GAAGlnB,EAAEU,EAAEC,GAAG,EAAE3C,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,SAASkpB,GAAGppB,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,OAAO,MAAM,QAAQN,CAAC,EAAEM,EAAE,GAAG,WAAWN,EAAE,GAAG,QAAQ+oB,EAAE,CAAC,OAAO,WAAW/oB,EAAE,GAAG,QAAQ+oB,EAAE,CAAC,KAAKzP,GAAGtZ,CAAC,EAAEM,EAAE,IAAIN,KAAKE,IAAI,OAAOI,EAAE+oB,GAAGrpB,CAAC,EAAEM,EAAE,WAAWN,EAAE,QAAQ+oB,EAAE,CAAC,EAAE,SAAS,EAAErQ,GAAGpY,EAAE,CAAC,CAAC,CAAC,SAAS+oB,GAAGrpB,EAAE,CAAC,OAAOA,IAAI,EAAE,QAAQ,MAAM,CAAC,SAASkpB,GAAGlpB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,GAAG,CAAC,IAAI,EAAEL,IAAI,YAAY,EAAE,EAAE+B,EAAE,EAAE,GAAGC,EAAE,EAAE,OAAO,GAAGA,IAAI,EAAE,CAAC,GAAGhC,IAAI,YAAY,CAAC,IAAIgE,EAAEilB,GAAGnpB,CAAC,EAAE,MAAM,CAACopB,GAAGllB,EAAE,GAAG,EAAEhE,CAAC,CAAC,CAAC,CAAC,OAAOA,IAAI,OAAO,CAACmpB,GAAGrpB,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,GAAG,SAAS,CAAC,CAAC,CAAC,GAAGkC,IAAI,EAAE,CAAC,GAAGD,EAAE4mB,GAAG,CAAC,IAAI,EAAEC,GAAG,EAAEvhB,EAAE,MAAM,KAAKvH,EAAE,MAAM,EAAE,CAAC,CAAC,EAAE,EAAE,MAAM,KAAKA,EAAE,OAAOiC,EAAE6mB,IAAI,EAAE7mB,EAAE,CAAC,CAAC,EAAE,OAAO/B,IAAI,cAAcqH,EAAE4hB,GAAG5hB,CAAC,EAAE,EAAE4hB,GAAG,CAAC,GAAG,CAAC,IAAI5hB,EAAE,IAAI,CAAC5C,EAAEC,IAAIwkB,GAAGzkB,EAAEpE,EAAEqE,GAAG1E,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE,UAAU,EAAE,IAAI,CAACyE,EAAEC,IAAIwkB,GAAGzkB,EAAEpE,EAAE0B,EAAE6mB,GAAGlkB,GAAG1E,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE,GAAG,CAAC,CAAC,IAAIgE,EAAEhE,IAAI,YAAYipB,GAAGnpB,CAAC,EAAE,MAAM,KAAKA,CAAC,EAAE,MAAM,CAAC,IAAIkE,EAAE,IAAI,CAAC,EAAEqD,IAAI6hB,GAAG,EAAE7oB,EAAEgH,GAAGrH,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE,GAAG,CAAC,CAAC,IAAIiC,EAAE,EAAE,MAAM,CAAC,EAAES,EAAEtC,EAAE,MAAM,CAAC,EAAEuC,EAAEvC,EAAE,GAAG,EAAEwC,EAAE,CAAC,EAAE,GAAGb,EAAE4mB,GAAG,CAAC,QAAQ3kB,EAAE,EAAEA,EAAE4kB,GAAG5kB,IAAI,CAAC,IAAI,EAAEA,EAAErB,EAAE0E,EAAE,EAAE1E,EAAEC,EAAE,KAAK,GAAGomB,GAAGlpB,EAAE,MAAM,EAAEuH,CAAC,EAAEpF,EAAEjC,EAAE0C,EAAErC,EAAE,EAAE,CAAC,CAAC,CAACuC,EAAE,KAAK,KAAK,EAAE,QAAQoB,EAAEjC,EAAE6mB,GAAG5kB,EAAEjC,EAAEiC,IAAI,CAAC,IAAI,EAAEA,EAAErB,EAAE0E,EAAE,EAAE1E,EAAEC,EAAE,KAAK,GAAGomB,GAAGlpB,EAAE,MAAM,EAAEuH,CAAC,EAAEpF,EAAEjC,EAAE0C,EAAErC,EAAE2D,IAAIjC,EAAE,CAAC,CAAC,CAAC,CAAC,KAAM,SAAQiC,EAAE,EAAEA,EAAEjC,EAAEiC,IAAI,CAAC,IAAI,EAAEA,EAAErB,EAAE0E,EAAE,EAAE1E,EAAEC,EAAE,KAAK,GAAGomB,GAAGlpB,EAAE,MAAM,EAAEuH,CAAC,EAAEpF,EAAEjC,EAAE0C,EAAErC,EAAE2D,IAAIjC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIc,EAAEb,IAAI,EAAE,IAAI,GAAGY,EAAE,GAAG,IAAIA,EAAE,GAAGC,EAAE,QAAQmB,EAAE,EAAEA,EAAEpB,EAAE,OAAO,EAAEoB,IAAIpB,EAAEoB,GAAG,IAAIpB,EAAEoB,GAAGnB,EAAE,IAAIC,EAAE;AAAA,EACt2C,QAAQkB,EAAE,EAAEA,EAAEhC,EAAEgC,IAAIlB,GAAG;AAAA,EACvB,OAAOF,EAAEA,EAAE,OAAO,GAAG,IAAIA,EAAEA,EAAE,OAAO,GAAG,KAAK,EAAE,GAAGE,GAAGF,CAAC,CAAC,SAASqmB,GAAGnpB,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,GAAG,EAAE,EAAE,KAAK,CAACF,EAAEE,GAAGF,EAAEE,EAAE,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,IAAIopB,GAAG,KAAK,CAAC,YAAYxoB,EAAEZ,EAAE,EAAE,CAAC,GAAG,KAAK,MAAMA,EAAE,KAAK,MAAMY,EAAE,MAAM,EAAE,KAAK,KAAKsX,GAAGtX,CAAC,EAAE,GAAG,KAAK,CAAC,IAAIP,EAAE,EAAE,OAAOuX,EAAEvX,IAAI,KAAK,KAAK,IAAI,qBAAqBA,qDAAqD,KAAK,QAAQ,CAAC,CAAC,GAAGL,IAAI,YAAY,MAAM,IAAI,MAAM,4JAA4J,EAAE,KAAK,OAAO,GAAG8Y,GAAG9Y,EAAE,KAAK,IAAI,EAAE,KAAK,QAAQ0Z,GAAG9Y,CAAC,CAAC,CAAC,IAAIA,KAAKZ,EAAE,CAACA,EAAE,SAAS,IAAIA,EAAE,CAAC,CAAC,GAAG4X,EAAE5X,EAAE,SAAS,KAAK,KAAK,IAAI,uCAAuCA,EAAE,gCAAgC,KAAK,OAAO,EAAE,IAAI,EAAE,KAAK,WAAWA,CAAC,EAAE,KAAK,OAAO,GAAGY,CAAC,CAAC,OAAOA,EAAE,CAACA,EAAE,SAAS,IAAIA,EAAE,CAAC,CAAC,GAAG,IAAIZ,EAAE,EAAE,QAAQK,KAAKO,EAAE,CAAC,GAAGP,EAAE,GAAGA,GAAG,KAAK,MAAML,GAAG,CAAC,IAAI6B,EAAE,qCAAqCjB,qBAAqB,KAAK,QAAQ,MAAM,IAAI,MAAMiB,CAAC,CAAC,CAAC7B,GAAG,CAAC,IAAI,EAAEY,EAAEA,EAAE,OAAO,GAAG,QAAQP,EAAE,EAAEA,EAAEO,EAAE,OAAO,EAAE,EAAEP,EAAE,GAAG,KAAK,QAAQA,GAAGO,EAAEP,GAAG,OAAO,KAAK,OAAO,EAAE,CAAC,WAAWO,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAO,GAAE,GAAG,KAAK,OAAO,EAAE,OAAOA,EAAE,GAAG,IAAIZ,EAAEY,EAAEA,EAAE,OAAO,GAAG,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAE,EAAEZ,GAAG,KAAK,QAAQ,GAAGY,EAAE,GAAG,OAAOZ,CAAC,CAAC,WAAWY,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAM,CAAC,EAAE,GAAG,KAAK,OAAO,EAAE,MAAM,CAACA,CAAC,EAAE,IAAIZ,EAAE,IAAI,MAAM,KAAK,MAAM,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAE,EAAEA,EAAE,GAAG,KAAK,MAAMY,EAAE,KAAK,QAAQ,EAAE,EAAEA,GAAGZ,EAAE,GAAG,KAAK,QAAQ,GAAG,OAAOA,EAAEA,EAAE,OAAO,GAAGY,EAAEZ,CAAC,CAAC,IAAI,MAAM,CAAC,OAAO,KAAK,MAAM,MAAM,CAAC,UAAU,CAAC,OAAOqpB,GAAG,EAAE,WAAW,KAAK,OAAO,KAAK,MAAM,KAAK,KAAK,CAAC,CAAC,EAAEA,GAAG,KAAKC,GAAG,KAAKC,GAAG,KAAK,SAASC,GAAG1pB,EAAE,CAACupB,GAAGvpB,CAAC,CAAC,SAAS2pB,GAAG3pB,EAAE,CAACwpB,GAAGxpB,CAAC,CAAC,SAAS4pB,GAAG5pB,EAAE,CAACypB,GAAGzpB,CAAC,CAAC,IAAI6pB,GAAG,KAAK,CAAC,YAAY/oB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,KAAK,GAAG,KAAK,mBAAmB,GAAG,KAAK,MAAMO,EAAE,MAAM,EAAE,KAAK,MAAMZ,GAAG,UAAU,KAAK,KAAKkY,GAAGtX,CAAC,EAAE,KAAK,QAAQ8Y,GAAG9Y,CAAC,EAAE,KAAK,OAAO,EAAE,KAAK,GAAGP,EAAE,KAAK,SAAS,KAAK,KAAK,EAAE,KAAK,KAAK,SAAS,EAAE,QAAQ,CAAC,IAAI,MAAM,CAAC,OAAO,KAAK,MAAM,MAAM,CAAC,MAAM,QAAQ,CAAC,IAAIO,EAAE,MAAM,KAAK,KAAK,EAAE,OAAO0oB,GAAG,OAAO,KAAK,MAAM,KAAK,MAAM1oB,CAAC,CAAC,CAAC,YAAY,CAAC,OAAO0oB,GAAG,OAAO,KAAK,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,MAAM,OAAO,CAAC,IAAI1oB,EAAE,MAAM,KAAK,KAAK,EAAE,OAAOgZ,GAAG,KAAK,MAAMhZ,EAAE,KAAK,QAAQ,WAAW,CAAC,CAAC,WAAW,CAAC,OAAOgZ,GAAG,KAAK,MAAM,KAAK,SAAS,EAAE,KAAK,QAAQ,WAAW,CAAC,CAAC,MAAM,MAAM,CAAC,KAAK,gBAAgB,EAAE,IAAIhZ,EAAEyoB,GAAG,EAAE,KAAK,KAAK,MAAM,EAAE,GAAG,KAAK,QAAQ,SAAS,CAAC,IAAIrpB,EAAE,MAAMY,EAAE,GAAG,CAAC,OAAOZ,EAAE,IAAI,GAAG8mB,GAAG,CAAC,CAAC,CAAC,OAAO,EAAN,CAAS,MAAM,IAAI,MAAM,+FAA+F,CAAC,CAAC,CAAC,OAAOlmB,CAAC,CAAC,UAAUA,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEyoB,GAAG,EAAE,UAAU,KAAK,OAAOzoB,CAAC,CAAC,CAAC,UAAU,CAAC,KAAK,gBAAgB,EAAE,IAAIA,EAAEyoB,GAAG,EAAE,SAAS,KAAK,MAAM,EAAE,GAAG,KAAK,QAAQ,SAAS,GAAG,CAAC,OAAOzoB,EAAE,IAAIZ,GAAG8mB,GAAG9mB,CAAC,CAAC,CAAC,OAAOA,EAAN,CAAS,MAAM,IAAI,MAAM,+FAA+F,CAAC,CAAC,OAAOY,CAAC,CAAC,MAAM,OAAO,CAAC,KAAK,gBAAgB,EAAE,IAAIA,EAAE,MAAMyoB,GAAG,EAAE,KAAK,KAAK,MAAM,EAAE,OAAO,KAAK,QAAQ,SAASzoB,EAAE,IAAI,WAAWA,EAAE,MAAM,CAAC,CAAC,SAAS,CAAC,KAAK,aAAayoB,GAAG,EAAE,cAAc,IAAI,EAAE,KAAK,mBAAmB,GAAG,CAAC,IAAI,YAAY,CAAC,OAAO,KAAK,kBAAkB,CAAC,iBAAiB,CAAC,GAAG,KAAK,WAAW,MAAM,IAAI,MAAM,qBAAqB,CAAC,CAAC,MAAMzoB,EAAE,GAAG,CAAC,OAAO0oB,GAAG,MAAM,KAAK1oB,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,KAAK,gBAAgB,EAAE0oB,GAAG,MAAM,IAAI,CAAC,CAAC,SAAS1oB,EAAE,GAAG,CAAC,IAAIZ,EAAE,KAAK,SAAS,EAAE,OAAO8oB,GAAG9oB,EAAE,KAAK,MAAM,KAAK,MAAMY,CAAC,CAAC,CAAC,KAAKA,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0oB,GAAG,KAAK,KAAK1oB,CAAC,CAAC,CAAC,SAASA,EAAE,GAAGZ,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEqpB,GAAG,EAAE,aAAa,KAAKzoB,EAAEZ,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,eAAe2pB,GAAG,OAAO,YAAY,CAAC,MAAM7pB,GAAG,CAAC,CAACA,GAAGA,EAAE,MAAM,MAAMA,EAAE,UAAU,MAAMA,EAAE,iBAAiB,IAAI,CAAC,EAAE,SAAS8pB,GAAG,CAAC,OAAO7O,GAAG,SAAS,IAAI4O,EAAE,CAAC,CAACC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY/oB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,MAAMO,EAAE,MAAMA,EAAE,MAAMA,EAAE,OAAOP,CAAC,EAAE,KAAK,UAAUL,EAAE,KAAK,KAAK,CAAC,CAAC,OAAOY,EAAE,CAAC,GAAGA,EAAE,QAAQ,KAAK,MAAM,MAAM,IAAI,MAAM,2BAA2BA,EAAE,8BAA8B,KAAK,mBAAmB,EAAE,GAAG,CAACkX,GAAGlX,EAAE,MAAM,KAAK,KAAK,EAAE,MAAM,IAAI,MAAM,2BAA2BA,EAAE,8BAA8B,KAAK,mBAAmB,EAAEyoB,GAAG,EAAE,cAAc,IAAI,EAAE,KAAK,OAAOzoB,EAAE,OAAOyoB,GAAG,EAAE,OAAO,KAAK,IAAI,CAAC,CAAC,SAAS,CAACA,GAAG,EAAE,gBAAgB,IAAI,EAAE,KAAK,mBAAmB,EAAE,CAAC,EAAE,OAAO,eAAeQ,GAAG,OAAO,YAAY,CAAC,MAAM/pB,GAAGA,aAAa6pB,IAAI7pB,EAAE,QAAQ,MAAMA,EAAE,kBAAkB,QAAQ,CAAC,EAAE,IAAIgqB,GAAG,CAAC,EAAE5pB,GAAG4pB,GAAG,CAAC,iBAAiB,IAAIC,GAAG,sBAAsB,IAAIC,GAAG,eAAe,IAAIC,GAAG,eAAe,IAAIC,EAAE,CAAC,EAAE,IAAIC,IAAI,SAASrqB,EAAE,CAACA,EAAE,GAAG,KAAKA,EAAE,GAAG,KAAKA,EAAE,GAAG,KAAKA,EAAE,GAAG,KAAKA,EAAE,GAAG,KAAKA,EAAE,GAAG,KAAKA,EAAE,GAAG,IAAI,GAAGqqB,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAAStqB,EAAE,CAACA,EAAE,QAAQ,UAAUA,EAAE,MAAM,QAAQA,EAAE,KAAK,QAAQA,EAAE,UAAU,WAAW,GAAGsqB,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASvqB,EAAE,CAACA,EAAE,QAAQ,UAAUA,EAAE,MAAM,QAAQA,EAAE,KAAK,OAAOA,EAAE,UAAU,WAAW,GAAGuqB,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASxqB,EAAE,CAACA,EAAE,QAAQ,UAAUA,EAAE,MAAM,UAAUA,EAAE,KAAK,UAAUA,EAAE,UAAU,WAAW,GAAGwqB,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASzqB,EAAE,CAACA,EAAE,QAAQ,YAAYA,EAAE,MAAM,YAAYA,EAAE,KAAK,YAAYA,EAAE,UAAU,WAAW,GAAGyqB,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,CAAC,QAAQF,GAAG,MAAMF,GAAG,KAAKC,GAAG,UAAUE,EAAE,EAAE,SAASE,GAAG3qB,EAAE,EAAE,CAAC,GAAGA,IAAI,UAAU,IAAI,SAAS,CAAC,GAAGA,IAAI,UAAU,IAAI,SAAS,MAAM,SAAS,MAAM,IAAI,MAAM,kBAAkBA,UAAU,GAAG,CAAC,CAAC,OAAO0qB,GAAG1qB,GAAG,EAAE,CAAC,SAAS4qB,GAAG5qB,EAAE,CAAC,OAAO2qB,GAAG3qB,EAAE,OAAO,CAAC,CAAC,SAASoqB,GAAGpqB,EAAE,EAAE,CAAC,GAAGA,EAAE,QAAQ,EAAE,MAAM,MAAM,CAACA,EAAE,CAAC,EAAE,IAAIE,EAAEyqB,GAAG3qB,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAACA,EAAE,KAAKE,CAAC,EAAE,EAAE,KAAKA,CAAC,CAAC,CAAC,CAAC,SAAS+pB,GAAGjqB,EAAE,EAAE,CAAC8X,EAAE9X,EAAE,QAAQ,EAAE,MAAM,IAAI,2BAA2BA,EAAE,qBAAqB,EAAE,yBAAyB,CAAC,CAAC,SAASmqB,GAAGnqB,EAAE,EAAE,CAAC,OAAO,EAAE,KAAKE,GAAGA,EAAE,KAAKF,EAAE,EAAE,CAAC,CAAC,SAASkqB,GAAGlqB,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,OAAO6qB,GAAG7qB,EAAE,EAAE,IAAI,GAAG,EAAE,CAAC,CAAC,SAAS6qB,GAAG7qB,EAAE,EAAEE,EAAE,CAAC,GAAGF,GAAG,KAAK,OAAO,GAAGA,aAAa6pB,GAAG,CAAC,EAAE,KAAK7pB,CAAC,EAAE,MAAM,CAAC,GAAG,CAAC8qB,GAAG9qB,CAAC,EAAE,OAAO,IAAIM,EAAEN,EAAE,QAAQO,KAAKD,EAAE,CAAC,IAAI,EAAEA,EAAEC,GAAGL,EAAE,IAAI,CAAC,IAAIA,EAAE,IAAI,CAAC,EAAE2qB,GAAG,EAAE,EAAE3qB,CAAC,EAAE,CAAC,CAAC,SAAS4qB,GAAG9qB,EAAE,CAAC,OAAO,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,QAAQ,CAAC,SAAS+qB,GAAG/qB,EAAE,CAAC,OAAOA,EAAE,YAAY,IAAI,CAAC,IAAIgrB,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,oBAAoB,CAAC,EAAE,KAAK,eAAe,EAAE,KAAK,SAAS,EAAE,KAAK,WAAW,EAAE,KAAK,iBAAiB,EAAE,KAAK,eAAe,EAAE,KAAK,cAAc,EAAE,KAAK,YAAY,EAAE,KAAK,WAAW,CAAC,EAAE,KAAK,kBAAkB,CAAC,EAAE,KAAK,YAAY,EAAE,KAAK,WAAW,IAAI,QAAQ,KAAK,UAAU,GAAG,KAAK,cAAc,CAAC,SAAS,EAAE,WAAW,EAAE,UAAU,EAAE,QAAQ,CAAC,EAAE,OAAO,KAAK,IAAI,aAAa,CAAC,OAAO,MAAM,KAAK,IAAI,IAAI,KAAK,QAAQ,IAAIlqB,GAAGA,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,QAAQA,KAAK,KAAK,oBAAoB,KAAK,oBAAoBA,GAAG,QAAQ,CAAC,CAAC,EAAEmqB,GAAG,KAAK,CAAC,YAAYnqB,EAAE,CAAC,KAAK,IAAIA,EAAE,KAAK,SAAS,CAAC,EAAE,KAAK,gBAAgB,CAAC,EAAE,KAAK,qBAAqB,EAAE,KAAK,MAAM,IAAIkqB,EAAE,CAAC,MAAM,OAAO,CAAC,GAAG,KAAK,oBAAoB,KAAK,OAAO,KAAK,mBAAmB,KAAK,IAAI,CAAC,CAAC,EAAE,GAAG,KAAK,iBAAiB,KAAK,OAAO,IAAIlqB,EAAE,KAAK,kBAAkB,EAAE,QAAQZ,EAAE,EAAEA,EAAEY,EAAE,OAAOZ,IAAI,CAAC,IAAI,EAAEY,EAAEZ,GAAG,GAAG,MAAM,KAAK,kBAAkB,CAAC,EAAE,QAAQ,CAAC,MAAM,KAAK,WAAW,CAAC,EAAE,MAAM,CAAC,CAAC,MAAM,IAAI,MAAM,wEAAwE,CAAC,CAAC,IAAI,SAAS,CAAC,GAAG,KAAK,oBAAoB,KAAK,MAAM,IAAI,MAAM,YAAY,KAAK,gIAAgI,EAAE,GAAG,KAAK,iBAAiB,KAAK,CAAC,GAAG,CAAC,KAAKY,EAAE,UAAUZ,CAAC,EAAE,KAAK,gCAAgC,EAAE,GAAGA,EAAE,MAAM,IAAI,MAAM,iCAAiCY,sHAAsH,EAAE,KAAK,WAAWA,CAAC,CAAC,CAAC,OAAO,KAAK,eAAe,CAAC,cAAc,CAAC,OAAO,OAAO,KAAK,KAAK,eAAe,CAAC,CAAC,YAAYA,EAAE,CAAC,GAAG,EAAEA,KAAK,KAAK,UAAU,GAAGA,KAAK,KAAK,gBAAgB,CAAC,GAAG,CAAC,UAAUZ,CAAC,EAAE,KAAK,kBAAkBY,CAAC,EAAE,GAAGZ,EAAE,OAAO,IAAI,KAAM,QAAO,KAAK,OAAO,KAAK,SAASY,EAAE,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,KAAK,KAAK,gBAAgB,KAAK,gBAAgBA,GAAG,QAAQ,IAAI,CAAC,gBAAgBA,EAAEZ,EAAE,EAAE,EAAE,CAAC,OAAOY,KAAK,KAAK,iBAAiBmlB,GAAG,GAAGnlB,qEAAqE,EAAE,KAAK,KAAK,gBAAgBA,GAAG,CAAC,QAAQZ,EAAE,SAAS,CAAC,EAAE,GAAG,CAAC,MAAM,WAAWY,EAAE,CAAC,GAAG,KAAK,gBAAgBA,IAAI,KAAK,MAAM,IAAI,MAAM,iBAAiBA,0BAA0B,EAAE,GAAG,KAAK,YAAYA,EAAE,KAAK,SAASA,IAAI,KAAK,CAAC,KAAK,gBAAgB,KAAK,GAAG,CAAC,QAAQZ,EAAE,UAAU,CAAC,EAAE,KAAK,kBAAkBY,CAAC,EAAE,GAAG,EAAE,EAAE,MAAMZ,EAAEA,GAAG,MAAM,EAAE,CAAC,OAAO,KAAK,gBAAgB,KAAK,SAASY,GAAG,KAAK,uBAAuB,EAAE,KAAK,SAAS,IAAI0nB,GAAG,KAAK,eAAe,EAAE,EAAE,CAAC,wBAAwB,CAAChC,GAAG,KAAK,WAAW,EAAE,QAAQ,GAAG,CAAC,EAAE,WAAW,MAAM,EAAE,UAAU,KAAK,eAAe,CAAC,CAAC,CAAC,CAAC,yBAAyB1lB,EAAE,CAAC0lB,GAAG1lB,CAAC,EAAE,QAAQR,GAAG,CAACA,EAAE,aAAa,MAAMA,EAAE,YAAY,KAAK,SAASQ,EAAE,CAAC,CAAC,CAAC,CAAC,kBAAkBA,EAAE,CAAC,IAAIZ,EAAE,KAAK,gBAAgBY,GAAG,GAAGZ,GAAG,KAAK,MAAM,IAAI,MAAM,6BAA6BY,2BAA2B,EAAE,GAAG,CAAC,IAAI,EAAEZ,EAAE,QAAQ,EAAE,GAAG,GAAG,EAAE,aAAakX,KAAK,OAAO,EAAE,MAAM,WAAW,CAAC,IAAI7W,EAAE,EAAE,KAAK,qBAAqBwB,EAAE,EAAE,KAAKC,GAAGzB,EAAE,KAAK,qBAAqB,IAAI,KAAK,SAASO,GAAGkB,EAAE,KAAK,mBAAmB,KAAK,GAAG,EAAE,MAAMA,IAAIzB,EAAE,KAAK,uBAAuB,KAAK,mBAAmB,KAAK0lB,GAAG,6BAA6BnlB,UAAU,EAAEmlB,GAAGjkB,EAAE,OAAOA,EAAE,OAAO,GAAG,GAAG,EAAE,OAAO,KAAK,mBAAmBD,EAAE,CAAC,QAAQA,EAAE,UAAU,EAAE,CAAC,KAAM,QAAO,KAAK,SAASjB,GAAG,EAAE,CAAC,QAAQ,GAAG,UAAU,EAAE,CAAC,OAAO,EAAN,CAAS,OAAOmlB,GAAG,6BAA6BnlB,UAAU,EAAEmlB,GAAG,EAAE,OAAO,EAAE,OAAO,EAAE,CAAC,QAAQ,GAAG,UAAU,EAAE,CAAC,CAAC,CAAC,cAAcnlB,EAAE,CAAC,GAAG,EAAEA,KAAK,KAAK,iBAAiB,MAAM,IAAI,MAAM,GAAGA,iCAAiC,EAAE,KAAK,cAAcA,GAAG,KAAK,oBAAoB,MAAM,KAAK,uBAAuBA,KAAK,KAAK,WAAW,KAAK,yBAAyBA,CAAC,EAAE,KAAK,SAASA,GAAG,QAAQ,EAAE,OAAO,KAAK,SAASA,IAAI,OAAO,KAAK,gBAAgBA,GAAG,KAAK,cAAcA,IAAI,KAAK,mBAAmB,KAAK,KAAK,YAAY,KAAK,KAAK,gBAAgB,KAAK,CAAC,mBAAmB,CAAC,GAAG,OAAO,KAAK,KAAK,eAAe,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,+BAA+B,EAAE,OAAO,OAAO,KAAK,KAAK,eAAe,EAAE,KAAK,CAACA,EAAEZ,IAAI,KAAK,gBAAgBA,GAAG,SAAS,KAAK,gBAAgBY,GAAG,QAAQ,CAAC,CAAC,iCAAiC,CAAC,IAAIA,EAAE,KAAK,kBAAkB,EAAE,QAAQZ,EAAE,EAAEA,EAAEY,EAAE,OAAOZ,IAAI,CAAC,IAAI,EAAEY,EAAEZ,GAAG,CAAC,QAAQK,EAAE,UAAUwB,CAAC,EAAE,KAAK,kBAAkB,CAAC,EAAE,GAAGA,GAAGxB,EAAE,MAAM,CAAC,KAAK,EAAE,UAAUwB,CAAC,CAAC,CAAC,MAAM,IAAI,MAAM,wEAAwE,CAAC,CAAC,SAASjB,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,MAAM,WAAW,IAAIA,CAAC,EAAEK,EAAE,EAAE,QAAQwB,EAAE,KAAK,SAAS7B,CAAC,EAAE8B,EAAEzB,EAAE,SAASL,CAAC,EAAEK,EAAE,YAAYL,EAAE,EAAE,EAAE,EAAE,QAAQY,EAAEA,EAAE,KAAKZ,EAAE6B,EAAE,EAAE,MAAM,EAAE,MAAMC,CAAC,EAAE,KAAK,uBAAuB,GAAG,KAAK,MAAM,kBAAkB,KAAK,MAAM,kBAAkB,OAAO,IAAI,CAAC,KAAKlB,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,GAAGA,GAAG,KAAK,CAAC,GAAG,OAAOY,GAAG,WAAW,MAAM,IAAI,MAAM,qCAAqC,EAAEZ,EAAEY,CAAC,KAAK,CAAC,GAAG,OAAOA,GAAG,UAAU,EAAEA,aAAa,QAAQ,MAAM,IAAI,MAAM,gFAAgF,EAAE,GAAG,OAAOZ,GAAG,WAAW,MAAM,IAAI,MAAM,gFAAgF,EAAE,EAAEY,CAAC,CAAC,IAAIP,EAAE,OAAO,KAAK,UAAU,IAAI,KAAK,WAAW,CAAC,EAAE,IAAI,KAAK,SAASA,CAAC,EAAE,KAAKA,EAAEL,EAAE,EAAEK,aAAa,SAAS,QAAQ,MAAM,yCAAyC,EAAEA,EAAE,CAAC,CAAC,UAAUO,EAAEZ,EAAE,EAAE,CAACY,EAAE,EAAE,GAAG,CAAC,IAAIP,EAAE,EAAE,EAAE,OAAOL,EAAE,EAAEK,CAAC,OAAOA,EAAN,CAAS,MAAML,EAAE,EAAEK,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO0qB,GAAG,cAAc,CAAC,gBAAgB,CAAC,OAAOA,GAAG,gBAAgB,CAAC,MAAMnqB,EAAE,CAAC,IAAIZ,EAAEgrB,EAAE,UAAU5L,GAAG,CAAC,EAAExe,CAAC,CAAC,EAAE,EAAE,CAAC,EAAEA,CAAC,EAAEP,EAAEyB,IAAI,CAAC,EAAE,IAAI,CAAC,IAAIC,EAAE,UAAUC,EAAE,CAAC,EAAEF,CAAC,EAAE,EAAE,CAAC,MAAMC,CAAC,EAAE,OAAOipB,EAAE,UAAUzO,GAAGva,EAAE,CAAC,CAAC,CAAC,GAAGH,EAAE,CAAC,EAAE,OAAO,KAAK,YAAY,KAAK,MAAM,YAAY,KAAK,EAAE,CAAC7B,CAAC,EAAEK,EAAEwB,EAAE,CAAC,CAAC,EAAE7B,CAAC,CAAC,UAAUY,EAAEZ,EAAE,EAAE,CAAC,GAAG,KAAK,aAAa,MAAM,KAAK,QAAUmmB,GAAGvlB,EAAE,KAAK,WAAW,GAAG,KAAM,MAAM,IAAI,MAAM,WAAWA,kCAAkC,KAAK,cAAc,EAAE,OAAO,KAAK,cAAc,CAAC,WAAWA,EAAE,OAAOZ,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,OAAO,KAAK,IAAI,QAAQ,SAAS,CAAC,CAAC,sBAAsBY,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,KAAK,QAAQ,WAAW,EAAEwB,EAAE,EAAE,EAAE,QAAQG,GAAG,CAACH,GAAGG,EAAE,QAAQ,YAAY,EAAE,CAAC,CAAC,EAAE,IAAIF,EAAE,KAAK,MAAM,kBAAkB,KAAK,MAAM,kBAAkB,OAAO,GAAGC,EAAE1B,EAAEL,EAAE6B,EAAEC,EAAE,GAAGC,EAAE,EAAE,MAAM,IAAI,MAAM,YAAY,KAAK,6CAA6CA,8BAA8BnB,IAAI,CAAC,CAAC,cAAcA,EAAE,CAAC,IAAIZ,EAAE,EAAE,CAAC,EAAEK,EAAE,KAAK,SAAS,EAAEwB,EAAE,KAAK,MAAM,SAASC,EAAE,KAAK,MAAM,WAAW,KAAK,uBAAuB,GAAG,KAAK,MAAM,kBAAkB,KAAK,CAAC,EAAE,IAAIC,EAAE,KAAK,aAAa,MAAM,KAAK,QAAQ,IAAIC,EAAE,EAAE6oB,GAAGjqB,CAAC,EAAEA,EAAE,WAAW,KAAK,MAAM,aAAa,KAAK,KAAK,MAAM,YAAY,KAAK,GAAG,GAAGiqB,GAAGjqB,CAAC,EAAE,CAAC,GAAG,CAAC,WAAWkC,EAAE,OAAOkB,EAAE,MAAMC,CAAC,EAAErD,EAAE,KAAK,aAAa,MAAM,KAAK,QAAQ,IAAIyG,EAAE8e,GAAGrjB,EAAE,KAAK,WAAW,EAAE8U,EAAEvQ,GAAG,KAAK,IAAI,kCAAkCvE,mBAAmB,KAAK,cAAc,EAAEf,EAAE,IAAI,CAAC,IAAI8C,EAAE,KAAK,QAAQ,WAAW,EAAE7C,EAAEqF,EAAE,WAAW,CAAC,OAAOrD,EAAE,MAAMC,EAAE,QAAQ,KAAK,OAAO,CAAC,EAAE,IAAIQ,EAAE,MAAM,QAAQzC,CAAC,EAAEA,EAAE,CAACA,CAAC,EAAE,KAAK,uBAAuB,GAAG,KAAK,sBAAsBc,EAAE+B,EAAEJ,CAAC,EAAE,IAAIC,EAAED,EAAE,IAAIE,GAAGA,EAAE,MAAM,KAAKA,EAAE,KAAK,yBAAyBA,CAAC,CAAC,EAAE,GAAGtE,EAAE,CAAC,IAAIsE,EAAE,KAAK,sBAAsB7B,EAAEkB,EAAEU,CAAC,EAAE,EAAE,KAAK,2BAA2BC,CAAC,CAAC,CAAC,OAAOD,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,YAAY5B,CAAC,EAAElC,EAAEoD,EAAEC,GAAG,CAAC,CAAC5D,IAAI,EAAE4D,EAAE,IAAIoD,GAAG,KAAK,KAAK,KAAK,MAAMA,CAAC,CAAC,CAAC,EAAE,EAAEtF,EAAE,IAAI,CAAC,IAAIkC,EAAE,KAAK,QAAQ,WAAW,EAAEjC,EAAE,KAAK,KAAK,IAAIc,EAAE,KAAK,QAAQkB,CAAC,CAAC,EAAE,IAAIqD,EAAE,MAAM,QAAQrF,CAAC,EAAEA,EAAE,CAACA,CAAC,EAAE,OAAO,KAAK,uBAAuB,GAAG,KAAK,sBAAsB,EAAEiC,EAAEoD,CAAC,EAAEA,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO3E,EAAE,MAAM,CAAC,EAAE9B,EAAEgC,EAAEioB,GAAGjqB,CAAC,EAAE,KAAKA,EAAE,cAAciC,EAAE,OAAO,KAAK,UAAU,IAAI,KAAK,MAAM,cAAc,IAAI,KAAK,MAAM,cAAc,IAAI,CAAC,CAAC,KAAK,IAAI,QAAQ,OAAO,GAAG,CAAC,KAAK,MAAM,UAAU7C,EAAE+B,EAAE,GAAGc,EAAE,KAAK,SAAS,cAAc,EAAEH,EAAE,IAAIX,EAAE,CAAC,EAAE,KAAK,IAAI,QAAQ,OAAO,GAAG,KAAK,SAAS,iBAAiBc,CAAC,EAAE7C,EAAE6C,EAAE,QAAQ,CAAC,EAAExC,GAAG,KAAK,YAAY,EAAEqC,EAAE1C,EAAE4C,EAAE,EAAE,CAAC,EAAE,KAAK,MAAM,WAAW,KAAK,MAAM,cAAc,QAAQ,KAAK,CAAC,KAAK,EAAE,WAAW,KAAK,MAAM,SAASf,EAAE,mBAAmB,KAAK,MAAM,SAAS,aAAa,KAAK,MAAM,WAAWC,EAAE,qBAAqB,KAAK,MAAM,WAAW,YAAY,OAAO,KAAKY,CAAC,EAAE,IAAII,GAAGJ,EAAEI,IAAI,KAAKJ,EAAEI,GAAG,MAAM,IAAI,EAAE,aAAa9C,EAAE,IAAI8C,GAAGA,EAAE,KAAK,EAAE,aAAaD,EAAE,OAAO,UAAUA,EAAE,SAAS,CAAC,EAAE,MAAM,QAAQb,CAAC,EAAEhC,EAAEA,EAAE,EAAE,CAAC,2BAA2BY,EAAE,CAAC,OAAOA,EAAE,IAAIR,GAAG,KAAK,KAAK,KAAK,MAAMA,CAAC,CAAC,CAAC,CAAC,CAAC,sBAAsBQ,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAEgmB,GAAGzlB,CAAC,EAAE,GAAGP,GAAG,KAAK,CAAC,IAAIwB,EAAExB,EAAE,cAAc,CAAC,EAAEyB,EAAEzB,EAAE,eAAe,CAAC,EAAE0B,EAAE1B,EAAE,eAAeuX,EAAE,MAAM,QAAQ5X,CAAC,EAAE,IAAI,wDAAwD,EAAE+B,EAAE,OAAO,KAAK/B,CAAC,EAAE,IAAI,GAAGA,EAAE,EAAE,GAAG+B,EAAEF,EAAE,IAAI,GAAG7B,EAAE,EAAE,EAAE,IAAIgC,EAAE,EAAE,OAAO,CAAC,EAAEU,IAAIZ,EAAEY,EAAE,EAAE,OAAOX,EAAE,OAAOC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,WAAWpB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,GAAGO,GAAG,KAAK,MAAM,IAAI,MAAM,+CAA+C,EAAE,EAAE,GAAG,UAAUP,EAAEA,GAAG,KAAK,QAAQ,IAAIwB,EAAEjB,EAAE,IAAI,UAAUwY,GAAGxY,EAAE,EAAE,IAAIiB,EAAEjB,EAAE,IAAIoB,GAAG+kB,GAAG/kB,CAAC,CAAC,GAAG,IAAIF,EAAEzB,EAAE,MAAMwB,EAAE7B,EAAE,CAAC,EAAE+B,EAAE,IAAI4nB,GAAG3pB,EAAE,EAAE8B,EAAE,KAAK,aAAa,CAAC,EAAE,GAAG,KAAK,YAAYC,EAAE1B,CAAC,EAAE,IAAI,SAAS,CAAC,IAAI2B,EAAE,KAAK,MAAM,WAAW,IAAIF,CAAC,EAAE,EAAEqX,GAAGtX,CAAC,EAAE,KAAK,MAAM,UAAU,EAAEG,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,OAAOD,CAAC,CAAC,qBAAqBnB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,EAAE,GAAG,UAAU,IAAIwB,EAAE,CAAC,OAAOjB,EAAE,MAAMZ,EAAE,MAAM,CAAC,EAAE,OAAO,KAAK,yBAAyB6B,EAAExB,CAAC,CAAC,CAAC,yBAAyBO,EAAEZ,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAMK,EAAE,MAAMwB,CAAC,EAAEjB,EAAEkB,EAAE,IAAI6nB,GAAGtpB,EAAEwB,EAAE,EAAE,KAAK,aAAa,CAAC,EAAE,OAAO,KAAK,YAAYC,EAAE9B,CAAC,EAAE8B,CAAC,CAAC,aAAalB,EAAEZ,EAAE,GAAG,EAAEK,EAAE,CAAC,EAAE,GAAG,KAAK,eAAe,EAAE,SAAS,EAAEA,GAAG,MAAMA,IAAIO,EAAE,QAAQA,EAAEA,EAAE,KAAKP,CAAC,GAAG,IAAIwB,EAAE,IAAIgoB,GAAGjpB,EAAEZ,EAAE,EAAE,KAAK,aAAa,CAAC,EAAE,GAAG,KAAK,MAAM,oBAAoB6B,EAAE,OAAO,KAAK,MAAM,IAAI,MAAM,sBAAsBA,EAAE,6BAA6B,EAAE,OAAO,KAAK,MAAM,oBAAoBA,EAAE,MAAMA,EAAE,KAAK,OAAOA,EAAE,KAAK,OAAO,EAAEA,CAAC,CAAC,YAAYjB,EAAEZ,EAAE,CAAC,KAAK,MAAM,aAAaY,EAAE,QAAQ,UAAU,KAAK,MAAM,mBAAmB,IAAI,EAAE,EAAEA,EAAE,QAAQ,aAAaA,EAAE,QAAQ,WAAW,EAAEA,EAAE,KAAKsY,GAAGtY,EAAE,KAAK,GAAG,KAAK,MAAM,UAAU,EAAE,KAAK,MAAM,WAAW,IAAIA,EAAE,MAAM,IAAI,KAAK,MAAM,iBAAiB,KAAK,MAAM,WAAW,IAAIA,EAAE,OAAO,CAAC,QAAQZ,GAAG,KAAK,QAAQ,MAAMY,EAAE,MAAM,MAAMA,EAAE,MAAM,MAAM,CAAC,CAAC,GAAGA,aAAaipB,IAAI,KAAK,MAAMjpB,CAAC,CAAC,CAAC,OAAOA,EAAEZ,EAAE,CAAC,KAAK,YAAYY,EAAEZ,CAAC,EAAE,KAAK,QAAQ,OAAOY,EAAE,MAAM,CAAC,CAAC,aAAaA,EAAEZ,EAAE,CAAC,KAAK,MAAM,WAAW,IAAIY,CAAC,GAAG,KAAK,MAAM,WAAW,IAAIA,CAAC,EAAE,UAAUZ,IAAI,KAAK,MAAM,WAAW,OAAOY,CAAC,EAAE,KAAK,MAAM,iBAAiB,CAAC,cAAcA,EAAE,CAAC,GAAG,CAAC,KAAK,MAAM,WAAW,IAAIA,EAAE,MAAM,EAAE,OAAO,IAAIZ,EAAE,KAAK,MAAM,WAAW,IAAIY,EAAE,MAAM,EAAE,GAAG,KAAK,MAAM,aAAaA,EAAE,QAAQ,WAAW,KAAK,MAAM,mBAAmB,KAAK,MAAM,UAAUZ,EAAE,OAAOY,EAAE,QAAQ,aAAaA,EAAE,QAAQ,SAAS,CAAC,IAAI,EAAEA,EAAE,KAAKsY,GAAGtY,EAAE,KAAK,EAAE,KAAK,MAAM,UAAU,CAAC,CAACZ,EAAE,QAAQ,YAAYY,EAAE,MAAM,GAAG,KAAK,aAAaA,EAAE,OAAOZ,EAAE,OAAO,CAAC,CAAC,kBAAkB,CAAC,QAAQY,KAAK,KAAK,MAAM,oBAAoB,CAAC,IAAIZ,EAAE,KAAK,MAAM,oBAAoBY,GAAG,KAAK,gBAAgBZ,CAAC,CAAC,CAAC,CAAC,gBAAgBY,EAAE,CAAC,KAAK,cAAcA,CAAC,EAAE,KAAK,MAAM,oBAAoBA,EAAE,OAAO,MAAM,OAAO,KAAK,MAAM,oBAAoBA,EAAE,KAAK,CAAC,QAAQ,CAAC,IAAIA,EAAE,KAAK,QAAQ,OAAO,EAAE,OAAOA,EAAE,WAAW,KAAK,MAAM,WAAWA,EAAE,eAAe,KAAK,MAAM,eAAeA,EAAE,SAAS,KAAK,MAAM,SAAS,KAAK,MAAM,iBAAiB,IAAIA,EAAE,WAAW,GAAGA,EAAE,SAAS,OAAOA,EAAE,QAAQ,CAAC,GAAGA,EAAE,QAAQ,KAAK,uEAAuE,GAAGA,CAAC,CAAC,MAAM,QAAQA,EAAE,CAAC,KAAK,MAAM,UAAU,GAAG,IAAIZ,EAAE,KAAK,MAAM,SAAS,EAAE,KAAK,MAAM,WAAW,KAAK,MAAM,cAAc,QAAQ,CAAC,EAAE,KAAK,MAAM,cAAc,OAAO,MAAMY,EAAE,EAAE,KAAK,MAAM,UAAU,GAAG,KAAK,MAAM,cAAc,UAAU,KAAK,IAAI,GAAG,KAAK,MAAM,cAAc,QAAQ,IAAIP,GAAGA,EAAE,kBAAkB,CAAC,EAAE,KAAK,MAAM,cAAc,SAAS,KAAK,MAAM,SAASL,EAAE,KAAK,MAAM,cAAc,WAAW,KAAK,MAAM,WAAW,EAAE,QAAQK,KAAK,KAAK,MAAM,cAAc,QAAQA,EAAE,aAAa,MAAMA,EAAE,aAAaA,EAAE,UAAU,MAAMA,EAAE,UAAU,OAAO,KAAK,MAAM,aAAa,CAAC,UAAU,CAAC,OAAO,KAAK,MAAM,cAAc,GAAG,KAAK,MAAM,cAAc,CAAC,CAAC,YAAYO,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,IAAIC,EAAE,CAAC,GAAG,KAAK,MAAM,iBAAiB,WAAWnB,EAAE,OAAOZ,EAAE,QAAQ,EAAE,MAAM6B,CAAC,EAAEG,EAAEqkB,GAAGzlB,CAAC,EAAEoB,GAAG,OAAO3B,EAAE2B,EAAE,UAAU3B,GAAG,OAAO0B,EAAE,SAAS,IAAI,EAAE,EAAE,IAAI,CAACW,EAAE,IAAI,CAAC,GAAGA,GAAG,KAAK,CAAC,IAAIE,EAAE,EAAE,GAAGC,EAAEiX,GAAGlX,EAAE,KAAKA,EAAE,KAAK,EAAE,OAAO,KAAK,WAAWC,EAAED,EAAE,MAAMA,EAAE,KAAK,CAAC,CAAC,OAAOF,CAAC,CAAC,EAAErC,EAAE,EAAE,OAAO,EAAE,EAAE,EAAE,GAAGwB,EAAEC,CAAC,IAAI,KAAK,MAAM,WAAW,KAAKC,CAAC,CAAC,CAAC,KAAKnB,EAAE,CAAC,OAAOA,EAAE,KAAK,GAAGA,CAAC,CAAC,WAAW,CAAC,KAAK,MAAM,gBAAgB,IAAI,KAAK,MAAM,WAAW,CAAC,GAAG,KAAK,MAAM,eAAe,CAAC,SAAS,CAAC,KAAK,MAAM,eAAe,CAAC,WAAWA,EAAE,CAAC,IAAIZ,EAAE,CAAC,MAAM,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,MAAM,aAAa,EAAEY,IAAIZ,EAAE,KAAKY,GAAG,KAAK,MAAM,WAAW,KAAKZ,CAAC,EAAE,KAAK,MAAM,YAAYA,CAAC,CAAC,SAASY,EAAE,CAAC,IAAIZ,EAAEgqB,GAAGppB,CAAC,EAAE,EAAE,IAAI,IAAIZ,EAAE,IAAI6B,GAAGA,EAAE,EAAE,CAAC,EAAE,QAAQA,EAAE,EAAEA,EAAE,KAAK,MAAM,YAAY,MAAM,OAAOA,IAAI,CAAC,IAAIC,EAAE,KAAK,MAAM,YAAY,MAAMD,GAAG,CAACC,EAAE,MAAM,CAAC,EAAE,IAAIA,EAAE,EAAE,GAAGA,EAAE,QAAQ,CAAC,CAAC,IAAIzB,EAAE,KAAK,MAAM,WAAW,IAAI,EAAE,KAAK,MAAM,YAAY,KAAK,MAAM,WAAW,SAAS,EAAE,KAAK,KAAK,MAAM,WAAW,KAAK,MAAM,WAAW,OAAO,GAAGL,EAAE,QAAQ6B,GAAG,CAAC,CAACA,EAAE,MAAMA,EAAE,UAAUxB,EAAE,IAAI,KAAK,MAAMwB,CAAC,CAAC,CAAC,CAAC,CAAC,UAAUjB,EAAEZ,EAAE,EAAEK,EAAE,GAAG,CAAC,GAAGuX,EAAE5X,EAAE,OAAO,EAAE,IAAI,2CAA2C,EAAE,GAAG,MAAM,EAAE,QAAQ,UAAU,MAAM,IAAI,MAAM,0CAA0C,EAAE,QAAQ,EAAE,IAAI6B,EAAE,KAAK,UAAU,IAAI,KAAK,UAAU,EAAE,IAAI,KAAK,QAAQ,EAAE,IAAI,KAAK,KAAK,UAAUjB,CAAC,CAAC,EAAEgX,EAAE/V,aAAa8nB,GAAG,IAAI,gDAAgD,EAAE,IAAI7nB,EAAE2mB,GAAG,KAAK,MAAM,WAAWzoB,EAAE6B,CAAC,EAAE,GAAG,CAACxB,GAAGyB,EAAE,SAAS,GAAG9B,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,qIAAqI,EAAE,OAAO,KAAK,KAAK,WAAW,IAAI,CAAC,IAAI+B,EAAE,CAAC,EAAEA,EAAEF,EAAE,IAAI,GAAG,KAAKopB,GAAGppB,EAAE,KAAK,EAAE,EAAE6mB,GAAG3mB,EAAED,EAAE,GAAG,KAAK,KAAK,CAAC,EAAEopB,EAAE,EAAE,IAAIlpB,EAAEhC,EAAE,IAAI,GAAG+B,EAAE,EAAE,GAAG,EAAE,OAAO,KAAK,MAAM,gBAAgB,IAAI,KAAK,MAAM,WAAW,QAAQ,GAAG,CAAC,QAAQW,KAAK,EAAE,MAAMA,EAAE,QAAQ,CAAC,CAAC,EAAE,KAAK,MAAM,WAAW,MAAM,CAAC,MAAMb,EAAE,MAAMG,CAAC,CAAC,CAAC,CAAC,CAAC,WAAWpB,EAAE,CAAC,OAAOgX,EAAE4B,GAAG5Y,CAAC,EAAE,IAAI,mDAAmD,EAAE,IAAIZ,IAAI,CAAC4X,EAAE5X,EAAE,MAAM+B,GAAGA,aAAa4nB,EAAE,EAAE,IAAI,kEAAkE,EAAE,IAAI,EAAEtpB,EAAE,CAAC,EAAEL,EAAE,QAAQ,CAAC+B,EAAEC,IAAI,CAAC3B,EAAE2B,GAAGD,CAAC,CAAC,EAAE,IAAIF,EAAE,CAACE,EAAEC,KAAK,EAAEpB,EAAE,GAAGZ,EAAEgC,CAAC,EAAE4V,EAAE,EAAE,iBAAiB+R,GAAG,IAAI,4FAA4F,EAAE/R,EAAE4B,GAAG,EAAE,QAAQ,EAAE,IAAI,kGAAkG,EAAE,EAAE,OAAO1X,EAAE,CAACC,EAAEC,IAAI,CAAC,IAAI,EAAE,EAAE,SAASD,EAAEC,CAAC,EAAEU,EAAE,MAAM,QAAQ,CAAC,EAAE,EAAE,CAAC,CAAC,EAAEkV,EAAElV,EAAE,SAAS1C,EAAE,OAAO,IAAI,qKAAqK,EAAE4X,EAAElV,EAAE,MAAME,GAAGA,aAAa+mB,EAAE,EAAE,IAAI,sIAAsI,EAAE,IAAI,EAAE,CAAC,EAAE,OAAOjnB,EAAE,QAAQ,CAACE,EAAEC,IAAI,CAAC,EAAEA,GAAG,IAAID,CAAC,CAAC,EAAE,CAAC,EAAE,OAAO,KAAK,cAAc,CAAC,YAAYf,EAAE,cAAcC,EAAE,OAAOzB,CAAC,CAAC,CAAC,CAAC,CAAC,SAASO,EAAE,CAAC,OAAO,KAAK,MAAM,WAAW,IAAIA,CAAC,EAAE,QAAQ,SAASA,CAAC,CAAC,CAAC,KAAKA,EAAE,CAAC,OAAO,KAAK,MAAM,WAAW,IAAIA,CAAC,EAAE,QAAQ,KAAKA,CAAC,CAAC,CAAC,UAAUA,EAAEZ,EAAE,CAAC,OAAO,KAAK,MAAM,WAAW,IAAIY,CAAC,EAAE,QAAQ,UAAUA,EAAEZ,CAAC,CAAC,CAAC,MAAM,KAAKY,EAAE,CAAC,IAAIZ,EAAEmnB,GAAG,EAAE,EAAE,MAAM,KAAK,QAAQ,KAAKvmB,CAAC,EAAE,OAAO,EAAE,OAAOumB,GAAG,EAAEnnB,EAAE,CAAC,CAAC,MAAMY,EAAE,CAAC,OAAO,KAAK,MAAM,aAAa,OAAOA,EAAE,QAAQ,KAAK,MAAM,YAAY,GAAG,KAAK,MAAM,YAAY,MAAM,KAAKA,CAAC,GAAGA,CAAC,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,MAAM,mBAAmB,CAAC,OAAO,CAAC,KAAK,uBAAuB,KAAK,MAAM,QAAQ,EAAE,KAAK,IAAI,MAAM,EAAE,KAAK,MAAM,IAAIkqB,GAAG,QAAQlqB,KAAK,KAAK,SAAS,KAAK,yBAAyBA,CAAC,EAAE,KAAK,SAASA,GAAG,QAAQ,EAAE,OAAO,KAAK,SAASA,GAAG,KAAK,YAAY,KAAK,KAAK,gBAAgB,KAAK,KAAK,mBAAmB,IAAI,CAAC,EAAEmqB,GAAG,aAAa,EAAEA,GAAG,eAAe,EAAE,SAASE,GAAGnrB,EAAE,CAAC,IAAI,EAAE+Z,GAAG3B,GAAGpY,CAAC,EAAE,SAAS,EAAE,OAAOkrB,EAAE,WAAW,EAAElrB,EAAE,SAAS,CAAC,CAAC,SAASqrB,IAAI,CAAC,IAAIrrB,EAAE+a,GAAG,EAAE,GAAG/a,EAAE,WAAW,KAAK,CAAC,IAAI,EAAE,IAAIua,GAAGva,CAAC,EAAEA,EAAE,UAAU,IAAIirB,GAAG,CAAC,CAAC,CAAC,OAAOpQ,GAAG7a,EAAE,UAAU,GAAG,EAAE0pB,GAAG,IAAI1pB,EAAE,SAAS,EAAEA,EAAE,SAAS,CAAC,IAAIkrB,EAAEG,GAAG,EAAE,SAASD,GAAGprB,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEF,EAAE,EAAE,CAAC,EAAE,OAAOkrB,EAAE,UAAU7P,GAAGnb,CAAC,CAAC,CAAC,IAAIorB,GAAG,CAAC,EAAElrB,GAAGkrB,GAAG,CAAC,UAAU,IAAIC,GAAG,SAAS,IAAIC,GAAG,aAAa,IAAIC,EAAE,CAAC,EAAE,SAASC,IAAI,CAAC,OAAO,OAAO,WAAW,aAAa,WAAW,IAAI,CAAC,IAAIC,GAAG,SAASF,GAAGzrB,EAAE,CAAC2rB,GAAG3rB,CAAC,CAAC,SAASwrB,GAAGxrB,EAAE,CAAC,GAAG2rB,KAAK,OAAO,OAAOA,GAAG,GAAG3rB,GAAG0rB,GAAG,EAAE,CAAC,GAAG1rB,IAAIA,EAAE,WAAWA,EAAE,UAAU,cAAc,MAAM,GAAG,IAAI,EAAEA,EAAE,WAAWA,EAAE,SAAS,OAAO,QAAQ,YAAY,OAAO,MAAM,IAAI,GAAG,CAAC,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAOE,EAAE,eAAeA,EAAE,cAAc,MAAM,CAAC,MAAM,2TAA2T,KAAK,CAAC,GAAG,0kDAA0kD,KAAK,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,SAASqrB,IAAI,CAAC,OAAO,OAAO,QAAQ,aAAa,OAAO,UAAU,MAAM,OAAO,mBAAmB,WAAW,CAAC,IAAIK,GAAGnR,EAAE,EAAEmR,GAAG,aAAa,QAAQ,IAAI,GAAG5rB,GAAG,CAACA,GAAG,QAAQ,KAAK,6IAA6I,CAAC,CAAC,EAAE4rB,GAAG,aAAa,aAAa,IAAIL,GAAG,CAAC,EAAEK,GAAG,aAAa,UAAU,IAAI,OAAO,SAAS,aAAa,OAAO,QAAQ,UAAU,aAAa,OAAO,QAAQ,SAAS,MAAM,WAAW,EAAEA,GAAG,aAAa,YAAY,IAAI,OAAO,WAAW,aAAa,WAAW,MAAM,UAAU,WAAW,MAAM,SAAS,KAAK,UAAU,SAAS,GAAG,aAAa,KAAK,UAAU,MAAM,CAAC,EAAEA,GAAG,aAAa,OAAO,IAAI,EAAE,EAAEA,GAAG,aAAa,qCAAqC,IAAIA,GAAG,QAAQ,OAAO,CAAC,EAAEA,GAAG,aAAa,+BAA+B,IAAI,EAAE,EAAEA,GAAG,aAAa,UAAU,IAAI,EAAE,EAAEA,GAAG,aAAa,+BAA+B,IAAI,EAAE,EAAEA,GAAG,aAAa,sBAAsB,IAAI,EAAE,EAAEA,GAAG,aAAa,sBAAsB,IAAI,EAAE,EAAEA,GAAG,aAAa,wCAAwC,IAAI,EAAE,EAAEA,GAAG,aAAa,uBAAuB,IAAI,EAAE,EAAE,SAASC,GAAG7rB,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAGmY,GAAGnY,CAAC,EAAE,OAAO,IAAI,SAAS,CAAC,EAAE,CAACA,EAAE,MAAM,EAAE,GAAG,OAAOA,GAAG,UAAU,YAAYA,EAAE,CAAC,IAAIO,EAAEP,EAAE,UAAU,OAAO,MAAM,CAACA,EAAE,OAAOA,EAAE,MAAMO,EAAE,MAAM,CAAC,CAAC,GAAG,CAAC,MAAM,QAAQP,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIM,EAAE,CAAC,EAAE,KAAK,MAAM,QAAQJ,CAAC,GAAGiY,GAAGjY,CAAC,GAAG,IAAI,UAAUI,EAAE,KAAKJ,EAAE,MAAM,EAAEA,EAAEA,EAAE,GAAG,OAAO,MAAM,QAAQF,CAAC,GAAGya,EAAE,EAAE,QAAQ,oCAAoC,GAAGqR,GAAG9rB,EAAEM,EAAE,CAAC,CAAC,EAAEA,CAAC,CAAC,SAASwrB,GAAG9rB,EAAE,EAAEE,EAAE,CAAC,GAAGA,EAAEA,GAAG,CAAC,EAAE,CAAC,MAAM,QAAQF,CAAC,GAAG,CAACmY,GAAGnY,CAAC,EAAE,CAAC8X,EAAE,EAAE,SAAS,EAAE,IAAI,eAAe5X,EAAE,KAAK,IAAI,2DAA2D,EAAE,aAAa,EAAE,MAAM,CAAC4X,EAAE,EAAE,OAAO,EAAE,IAAI,eAAe5X,EAAE,KAAK,IAAI,gDAAgDF,EAAE,iBAAiB,EAAE8X,EAAE9X,EAAE,SAAS,EAAE,GAAG,IAAI,eAAeE,EAAE,KAAK,IAAI,kBAAkB,EAAE,wBAAwBF,EAAE,iBAAiB,EAAE,IAAIM,EAAE,EAAE,MAAM,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAEO,EAAEurB,GAAG9rB,EAAEO,GAAGD,EAAEJ,EAAE,OAAOK,CAAC,CAAC,CAAC,CAAC,SAASwrB,GAAG/rB,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGN,IAAI,oBAAoB,CAAC,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,gCAAgC,EAAE,GAAGA,IAAI,WAAWA,IAAI,GAAGA,IAAI,WAAW,IAAI,SAAS,MAAM,IAAI,MAAM,aAAaE,iBAAiBI,cAAcN,qBAAqB,UAAU,CAAC,CAAC,CAAC,SAASgsB,EAAEhsB,EAAE,EAAEE,EAAEI,EAAE,UAAU,CAAC,GAAGN,aAAa6pB,GAAG,OAAOkC,GAAGzrB,EAAEN,EAAE,MAAM,EAAEE,CAAC,EAAEF,EAAE,IAAIO,EAAEkZ,GAAGzZ,CAAC,EAAE,GAAGO,IAAI,UAAU,CAAC,OAAO,QAAQ,SAAS,EAAE,QAAQD,CAAC,GAAG,IAAIC,EAAED,GAAGyrB,GAAGzrB,EAAEC,EAAE,EAAEL,CAAC,EAAEF,GAAG,MAAM,CAACmY,GAAGnY,CAAC,GAAG,CAAC,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,UAAU,OAAOA,GAAG,WAAW,OAAOA,GAAG,SAAS,CAAC,IAAIkC,EAAElC,GAAG,KAAK,OAAOA,EAAE,YAAY,KAAK,MAAM,IAAI,MAAM,aAAa,iBAAiBE,+CAA+CgC,IAAI,CAAC,CAAC,IAAI,EAAE2pB,GAAG7rB,EAAEO,CAAC,EAAE,CAAC4X,GAAGnY,CAAC,GAAG,CAAC,MAAM,QAAQA,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAG,IAAIiC,EAAE1B,IAAI,SAAS+mB,GAAGtnB,EAAEO,CAAC,EAAE2X,GAAGlY,EAAE,CAAC,EAAE,EAAE,EAAE,OAAOkrB,EAAE,WAAWjpB,EAAE,EAAE1B,CAAC,CAAC,CAAC,SAAS0rB,GAAGjsB,EAAE,EAAEE,EAAEI,EAAE,UAAU,CAAC,GAAG,CAAC,MAAM,QAAQN,CAAC,EAAE,MAAM,IAAI,MAAM,YAAY,eAAeE,8CAA8C,EAAE,OAAOF,EAAE,IAAI,CAAC+B,EAAEC,IAAIgqB,EAAEjqB,EAAE,GAAG,KAAKC,KAAK9B,EAAEI,CAAC,CAAC,CAAC,CAAC,IAAI4rB,GAAG,OAAO,SAASC,EAAEnsB,EAAE,CAAC,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE,GAAG,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,yGAAyG,EAAE,cAAc,EAAE,IAAIE,EAAE,EAAE,GAAGI,EAAEN,EAAEE,GAAGA,EAAE,SAAS,GAAG,IAAIA,EAAEA,EAAE,UAAU,EAAEA,EAAE,OAAO,CAAC,GAAGA,EAAEA,EAAEgsB,GAAG,IAAI3rB,EAAE,IAAI,IAAI,CAAC2qB,EAAE,WAAWhrB,CAAC,EAAE,GAAG,CAAC,IAAI,EAAEI,EAAE,GAAG,CAAC,EAAE,OAAO+Z,GAAG,CAAC,GAAG,QAAQ,MAAM,yCAAyC,EAAE6Q,EAAE,SAAS,CAAC,EAAE,CAAC,OAAO,EAAN,CAAS,MAAMA,EAAE,SAAS,IAAI,EAAE,CAAC,CAAC,EAAE,OAAO,OAAO,eAAe3qB,EAAE,OAAO,CAAC,MAAML,EAAE,aAAa,EAAE,CAAC,EAAEK,CAAC,CAAC,SAAS6rB,GAAGpsB,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,OAAO,SAAS,EAAEM,EAAE0rB,EAAE,EAAE,OAAO,SAAS,EAAEjU,GAAG7X,EAAE,MAAMI,EAAE,MAAM,yBAAyBJ,EAAE,aAAaI,EAAE,4CAA4C,EAAE,IAAIC,EAAE,CAAC,KAAKL,EAAE,KAAKI,CAAC,EAAE,OAAO4qB,EAAE,UAAUtO,GAAGrc,CAAC,CAAC,CAAC,IAAI8rB,GAAGF,EAAE,CAAC,SAASC,EAAE,CAAC,EAAE,SAASE,GAAGtsB,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGA,GAAG,OAAOA,EAAEmZ,GAAGzZ,CAAC,GAAGM,IAAI,YAAY,MAAM,IAAI,MAAM,kFAAkF,EAAE,GAAG,OAAON,GAAG,UAAU,YAAYA,EAAE,CAAC,GAAGM,IAAI,WAAWA,IAAI,QAAQ,MAAM,IAAI,MAAM,0FAA0FA,IAAI,EAAE,OAAON,EAAE,SAASA,EAAE,UAAU,OAAOkrB,EAAE,QAAQ,wBAAwBlrB,EAAE,GAAGE,EAAEI,CAAC,CAAC,CAAC,GAAG,CAAC6X,GAAGnY,CAAC,GAAG,CAAC,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,UAAU,OAAOA,GAAG,WAAW,OAAOA,GAAG,SAAS,MAAM,IAAI,MAAM,0HAA0H,EAAE,GAAG,GAAG,KAAK,CAACka,GAAG,CAAC,EAAE,IAAI3Z,EAAE6X,GAAG,CAAC,EAAE,EAAEA,GAAGlY,CAAC,EAAE4X,EAAEvX,IAAI,EAAE,IAAI,iCAAiC,8BAA8BA,oBAAoB,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAEL,EAAE,OAAO,EAAE,EAAE,CAAC,IAAI+B,EAAE/B,EAAE,GAAGgC,EAAE,IAAIhC,EAAE,OAAO,EAAE+B,IAAImW,GAAG,EAAE,MAAM,CAAC,CAAC,EAAE,GAAGN,EAAE5X,EAAE,KAAK,EAAE,IAAI,CAACgC,EAAE,IAAI,gDAAgDhC,yCAAyC,MAAM,CAAC,CAAC,CAAC,MAAM,CAACiY,GAAGnY,CAAC,GAAG,CAAC,MAAM,QAAQA,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAG,EAAE,GAAGE,EAAEF,EAAEM,IAAI,SAASgnB,GAAGtnB,EAAEM,CAAC,EAAE4X,GAAGlY,EAAE,CAAC,EAAE,EAAE,EAAEkrB,EAAE,WAAWlrB,EAAE,EAAEM,CAAC,CAAC,CAAC,SAASisB,GAAGvsB,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,OAAOosB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,IAAIssB,GAAG,CAAC,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,KAAK,EAAE,UAAU,CAAC,EAAMC,GAAG,EAAE,eAAeC,GAAG1sB,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAEC,EAAE,MAAM,QAAQP,CAAC,EAAEA,EAAE,IAAI,GAAG,EAAE,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEO,EAAE,OAAO,EAAE,EAAE,CAAC,IAAI0B,EAAE1B,EAAE,GAAG2B,EAAE,MAAM,QAAQlC,CAAC,EAAEA,EAAE,GAAG,OAAOA,EAAEiC,GAAG,GAAGC,EAAE,QAAQ,WAAWA,EAAE,QAAQ,SAASA,EAAE,QAAQ,QAAQA,EAAE,QAAQ,UAAUA,EAAE,QAAQ,YAAY,MAAM,IAAI,MAAM,gCAAgCD,OAAOC,EAAE,OAAO,EAAE,IAAIC,EAAE,CAAC,KAAKF,EAAE,MAAMC,EAAE,MAAM,MAAMA,EAAE,KAAK,EAAE,GAAGA,EAAE,QAAQ,SAAS,CAAC,IAAIU,EAAE,IAAI,QAAQ,MAAMC,GAAG,CAAC,IAAIC,EAAE,MAAMZ,EAAE,MAAM,EAAEa,EAAED,EAAE,OAAO,CAAC,EAAEyE,IAAI,EAAEA,EAAE,OAAO,CAAC,EAAEklB,GAAG3pB,EAAE,OAAOE,EAAE,IAAI,WAAWD,CAAC,EAAEmB,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAEpB,EAAE,OAAO,IAAI,CAAC,IAAIyE,EAAEzE,EAAE,GAAG,EAAE,IAAI,WAAW,IAAI,YAAY,CAACyE,EAAE,MAAM,CAAC,EAAE,MAAM,EAAEvE,EAAE,IAAI,EAAEkB,CAAC,EAAEA,GAAGuoB,GAAGzpB,EAAE,IAAIuE,EAAErD,CAAC,EAAEA,GAAGqD,EAAE,MAAM,CAAC1E,EAAEG,CAAC,CAAC,CAAC,EAAE1C,EAAE,KAAKsC,CAAC,CAAC,MAAMtC,EAAE,KAAK4B,EAAE,KAAK,CAAC,EAAE,GAAG,OAAOC,EAAE,MAAM,GAAGjC,EAAE,KAAKiC,CAAC,CAAC,CAAC,IAAI,EAAE,MAAM,QAAQ,IAAI7B,CAAC,EAAE,MAAM,CAAC,KAAKqsB,GAAG,CAAC,EAAE,MAAMzsB,CAAC,CAAC,CAAC,SAAS0sB,GAAG5sB,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAEC,EAAE,EAAE,QAAQ,KAAK,EAAE,CAAC,IAAI,EAAE,EAAE,KAAK0B,EAAE,EAAE,MAAMC,EAAE,EAAE,MAAMC,EAAEiW,GAAGlW,CAAC,EAAEU,EAAE,GAAG,iBAAiB,EAAE,CAAC,IAAIC,EAAE,EAAE,aAAa,GAAGA,EAAE,QAAQ,SAASA,EAAE,QAAQ,UAAU,GAAG,EAAE,QAAQA,GAAG,UAAUA,GAAG,MAAM,IAAI,MAAM,UAAU,EAAE,0BAA0BA,EAAE,0DAA0D,UAAUA,EAAE,QAAQ,WAAW,GAAGZ,IAAI,UAAU,MAAM,IAAI,MAAM,UAAU,EAAE,0BAA0BY,EAAE,yDAAyDZ,IAAI,MAAO,OAAM,IAAI,MAAM,UAAU,EAAE,uCAAuCY,EAAE,6EAA6E,EAAE,IAAIC,EAAE0pB,GAAG3pB,EAAE,OAAOE,EAAE/C,EAAE,MAAMO,EAAEA,EAAE4B,EAAEW,CAAC,EAAEE,EAAEH,EAAE,QAAQ,QAAQ,IAAI,WAAWE,CAAC,EAAE,IAAI,YAAYA,CAAC,EAAE,GAAGd,IAAI,UAAU,GAAGY,EAAE,QAAQ,SAASA,EAAE,QAAQ,SAAS,CAACD,EAAE,IAAI,aAAaI,EAAE,MAAM,EAAE,QAAQkB,EAAE,EAAEA,EAAElB,EAAE,OAAOkB,IAAI,CAAC,IAAI,EAAElB,EAAEkB,GAAGtB,EAAEsB,GAAG,EAAErB,EAAE,MAAMA,EAAE,GAAG,CAAC,SAASA,EAAE,QAAQ,UAAUvC,IAAI,SAASA,EAAEusB,GAAG,GAAGjqB,EAAEtC,EAAE0C,CAAC,MAAO,OAAM,IAAI,MAAM,iCAAiCH,EAAE,gCAAgC,UAAUZ,IAAI,QAAQ,CAAC,GAAGY,EAAE,QAAQ,SAASA,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,iCAAiCA,EAAE,8BAA8B,EAAED,EAAE,IAAI,WAAWI,EAAE,MAAM,EAAE,QAAQkB,EAAE,EAAEA,EAAElB,EAAE,OAAOkB,IAAI,CAAC,IAAI,EAAElB,EAAEkB,GAAGtB,EAAEsB,GAAG,KAAK,MAAM,EAAErB,EAAE,MAAMA,EAAE,GAAG,CAAC,CAAC,KAAM,OAAM,IAAI,MAAM,gCAAgC,OAAOZ,GAAG,EAAE1B,GAAG4B,EAAEW,CAAC,SAASb,IAAI,SAAS,CAAC,IAAIY,EAAEuV,GAAG,EAAE,KAAK,EAAExV,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAED,EAAEC,IAAI,CAAC,IAAIC,EAAE,IAAI,YAAY/C,EAAE,MAAMO,EAAEA,EAAEksB,EAAE,CAAC,EAAE,GAAGlsB,GAAGksB,GAAG,IAAIzpB,EAAE,IAAI,WAAWhD,EAAE,MAAMO,EAAEA,EAAEwC,CAAC,CAAC,EAAEH,EAAE,KAAKI,CAAC,EAAEzC,GAAGwC,CAAC,CAAC,KAAK,CAAC,IAAIF,EAAE2pB,GAAGvqB,GAAGa,EAAE9C,EAAE,MAAMO,EAAEA,EAAE4B,EAAEU,CAAC,EAAE,GAAGZ,IAAI,UAAUW,EAAE,IAAI,aAAaE,CAAC,UAAUb,IAAI,QAAQW,EAAE,IAAI,WAAWE,CAAC,UAAUb,IAAI,OAAOW,EAAE,IAAI,WAAWE,CAAC,UAAUb,IAAI,YAAY,CAACW,EAAE,IAAI,aAAaE,CAAC,EAAE,IAAIC,EAAE,IAAI,aAAaH,EAAE,OAAO,CAAC,EAAEI,EAAE,IAAI,aAAaJ,EAAE,OAAO,CAAC,EAAE,QAAQ2E,EAAE,EAAEA,EAAExE,EAAE,OAAOwE,IAAIxE,EAAEwE,GAAG3E,EAAE2E,EAAE,GAAGvE,EAAEuE,GAAG3E,EAAE2E,EAAE,EAAE,GAAG,IAAIrD,EAAEqoB,GAAGxpB,EAAEb,EAAE,SAAS,EAAE,EAAEqqB,GAAGvpB,EAAEd,EAAE,SAAS,EAAEhC,EAAE,GAAGmsB,GAAGnoB,EAAE,CAAC,EAAEA,EAAE,QAAQ,EAAE,EAAE,QAAQ,CAAC,KAAM,OAAM,IAAI,MAAM,gCAAgC,OAAOjC,GAAG,EAAE1B,GAAG4B,EAAEU,CAAC,CAACZ,IAAI,cAAc/B,EAAE,GAAGqsB,GAAG3pB,EAAEV,EAAED,CAAC,EAAE,CAAC,OAAO/B,CAAC,CAAC,SAASysB,GAAG3sB,EAAE,CAAC,GAAGA,IAAI,KAAK,MAAM,IAAI,MAAM,wBAAwB,KAAK,UAAUA,CAAC,GAAG,EAAE,IAAI,EAAE,EAAEE,EAAE,CAAC,EAAEF,EAAE,QAAQ,GAAG,CAAC,GAAG,GAAG,EAAE,WAAWE,EAAE,KAAK,EAAE,aAAa,EAAE,OAAO,WAAW,EAAE,IAAI,EAAE,YAAY,CAAC,CAAC,EAAE,EAAE,aAAa,cAAc,aAAa,YAAY,aAAa,YAAY,MAAM,IAAI,MAAM,mCAAmC,EAAE,YAAY,MAAM,CAAC,CAAC,EAAE,IAAII,EAAE,IAAI,WAAW,CAAC,EAAEC,EAAE,EAAE,OAAOL,EAAE,QAAQ,GAAG,CAACI,EAAE,IAAI,IAAI,WAAW,EAAE,MAAM,EAAEC,CAAC,EAAEA,GAAG,EAAE,UAAU,CAAC,EAAED,EAAE,MAAM,CAAC,IAAIwsB,GAAG,OAAO,QAAQ,cAAc,OAAO,MAAM,aAAa,OAAO,MAAM,aAAa,OAAO,MAAM,aAAa,SAASC,GAAG/sB,EAAE,CAAC,OAAO8sB,GAAG,OAAO,WAAW9sB,CAAC,EAAE,IAAI,KAAK,CAACA,CAAC,CAAC,EAAE,IAAI,CAAC,SAASgtB,GAAGhtB,EAAE,CAAC,GAAG8sB,GAAG,OAAO,OAAO,KAAK9sB,CAAC,EAAE,SAAS,QAAQ,EAAE,IAAI,EAAE,IAAI,WAAWA,CAAC,EAAEE,EAAE,GAAG,QAAQI,EAAE,EAAEC,EAAE,EAAE,OAAOD,EAAEC,EAAED,IAAIJ,GAAG,OAAO,aAAa,EAAEI,EAAE,EAAE,OAAO,KAAKJ,CAAC,CAAC,CAAC,SAAS+sB,GAAGjtB,EAAE,CAAC,GAAG8sB,GAAG,CAAC,IAAIxsB,EAAE,OAAO,KAAKN,EAAE,QAAQ,EAAE,OAAOM,EAAE,OAAO,MAAMA,EAAE,WAAWA,EAAE,WAAWA,EAAE,UAAU,CAAC,CAAC,IAAI,EAAE,KAAKN,CAAC,EAAEE,EAAE,IAAI,WAAW,EAAE,MAAM,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAEJ,EAAE,IAAI,CAAC,EAAE,WAAWI,CAAC,CAAC,EAAEA,CAAC,EAAE,OAAOJ,EAAE,MAAM,CAAC,SAASgtB,GAAGltB,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAOA,EAAE,GAAG,IAAI,EAAE,EAAEA,EAAE,QAAQO,GAAG,CAAC,GAAGA,EAAE,UAAU,CAAC,EAAE,IAAIL,EAAE,IAAI,WAAW,CAAC,EAAEI,EAAE,EAAE,OAAON,EAAE,QAAQO,GAAG,CAACL,EAAE,IAAI,IAAI,WAAWK,CAAC,EAAED,CAAC,EAAEA,GAAGC,EAAE,UAAU,CAAC,EAAEL,EAAE,MAAM,CAAC,SAASitB,GAAGntB,EAAE,CAAC,IAAI,EAAE,IAAI,IAAIA,EAAEA,EAAE,KAAK,EAAEA,EAAE,SAAS,CAAC,GAAGA,EAAEA,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,EAAE,IAAIE,EAAEF,EAAE,MAAM,CAAC,EAAE,OAAOE,EAAEA,EAAE,OAAO,EAAE,CAAC,SAASktB,GAAGptB,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,cAAcF,EAAE,cAAc,OAAOA,EAAE,OAAO,YAAYA,EAAE,YAAY,YAAYA,EAAE,YAAY,gBAAgB,CAAC,EAAE,OAAOA,EAAE,WAAW,OAAOE,EAAE,UAAUF,EAAE,WAAWA,EAAE,qBAAqB,OAAOE,EAAE,oBAAoBF,EAAE,qBAAqBA,EAAE,kBAAkB,OAAOE,EAAE,iBAAiBF,EAAE,kBAAkBA,EAAE,sBAAsB,OAAOE,EAAE,qBAAqBF,EAAE,sBAAsBA,EAAE,gBAAgB,OAAOE,EAAE,eAAeF,EAAE,gBAAgBE,CAAC,CAAC,SAASmtB,GAAGrtB,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,cAAcN,EAAE,cAAc,OAAOA,EAAE,OAAO,YAAYA,EAAE,YAAY,YAAYA,EAAE,WAAW,EAAE,GAAGA,EAAE,gBAAgB,OAAOM,EAAE,eAAeN,EAAE,gBAAgBA,EAAE,iBAAiB,KAAK,CAAC,GAAG,CAAC,EAAE,MAAM,IAAI,MAAM,uDAAuD,EAAE,GAAG,CAACE,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAEI,EAAE,YAAY,EAAEA,EAAE,WAAWJ,CAAC,CAAC,OAAOF,EAAE,WAAW,OAAOM,EAAE,UAAUN,EAAE,WAAWA,EAAE,qBAAqB,OAAOM,EAAE,oBAAoBN,EAAE,qBAAqBA,EAAE,kBAAkB,OAAOM,EAAE,iBAAiBN,EAAE,kBAAkBA,EAAE,sBAAsB,OAAOM,EAAE,qBAAqBN,EAAE,sBAAsBM,CAAC,CAAC,eAAegtB,GAAGttB,EAAE,EAAE,CAAC,IAAIE,EAAEI,EAAE,OAAON,EAAE,iBAAiB,OAAO,CAACE,EAAEI,CAAC,EAAE,MAAM,EAAEN,EAAE,eAAe,GAAGqtB,GAAGrtB,EAAEE,EAAEI,CAAC,CAAC,CAAC,SAASitB,GAAGvtB,EAAE,CAAC,GAAGA,EAAE,yBAAyB,YAAY,MAAM,IAAI,MAAM,qDAAqD,EAAE,MAAM,CAAC,UAAU,IAAI,KAAK,kBAAkB,OAAO,mBAAmBA,EAAE,eAAe,KAAK,EAAE+sB,GAAG,KAAK,UAAU/sB,EAAE,aAAa,CAAC,EAAE,iBAAiBA,EAAE,aAAa,KAAK,EAAE+sB,GAAG,KAAK,UAAU/sB,EAAE,WAAW,CAAC,EAAE,gBAAgBA,EAAE,YAAY,KAAK,EAAEA,EAAE,WAAW,UAAU,CAAC,CAAC,SAASwtB,GAAGxtB,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQE,KAAKF,EAAE,EAAE,KAAK,GAAGE,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC,SAASutB,IAAI,CAAC,IAAIztB,EAAEE,GAAG,CAAC,IAAII,EAAEJ,GAAG,GAAGK,EAAE,EAAE,MAAMD,EAAE,WAAW,GAAGC,GAAG,QAAQD,IAAI,EAAE,OAAOA,GAAG,SAASC,GAAG,UAAUD,EAAEC,CAAC,EAAE,EAAE,IAAI,YAAY,IAAI,EAAE,EAAE,GAAG,EAAE,QAAQL,EAAE,EAAEA,EAAE,KAAKA,IAAI,EAAEA,GAAGF,EAAEE,CAAC,EAAE,QAAQA,EAAE,KAAKA,EAAE,KAAKA,IAAI,EAAEA,GAAG,WAAWA,EAAE,MAAM,IAAI,OAAO,CAAC,CAAC,SAASwtB,IAAI,CAAC,IAAI1tB,EAAE,IAAI,YAAY,EAAE,EAAEA,EAAE,GAAG,EAAEA,EAAE,IAAI,WAAWA,EAAE,IAAI,WAAWA,EAAE,IAAI,WAAW,QAAQ,EAAE,EAAE,EAAE,GAAG,IAAIA,EAAE,GAAG,GAAG,GAAG,QAAQ,EAAE,GAAG,EAAE,GAAG,IAAIA,EAAE,GAAG,YAAY,EAAE,IAAI,IAAI,OAAOA,CAAC,CAAC,SAAS2tB,IAAI,CAAC,IAAI3tB,EAAE,IAAI,YAAY,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAE,GAAG,IAAIA,EAAE,GAAG,KAAK,OAAOA,EAAE,GAAGA,EAAE,IAAI,EAAEA,CAAC,CAAC,SAAS6sB,IAAI,CAAC,IAAI7sB,EAAEytB,GAAG,EAAE,EAAEC,GAAG,EAAExtB,EAAEytB,GAAG,EAAE,OAAOrtB,GAAG,CAAC,IAAIC,EAAE,IAAI,YAAY,EAAED,EAAE,MAAM,EAAE,EAAE,IAAI,YAAYC,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAED,EAAE,OAAO,IAAI,CAAC,IAAI2B,EAAE3B,EAAE,GAAG4B,EAAElC,EAAEE,EAAE+B,GAAG,KAAKA,EAAE,OAAO,EAAEA,GAAG,IAAI,EAAE,GAAGC,CAAC,CAAC,OAAO,IAAI,aAAa3B,CAAC,CAAC,CAAC,CAAC,IAAIqtB,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,YAAY,CAAC,CAAC,CAAC,OAAO,aAAa,CAAC,OAAOA,GAAG,UAAU,OAAOA,GAAG,SAAS,IAAIA,IAAIA,GAAG,QAAQ,CAAC,OAAO,mBAAmB9sB,EAAE,CAAC8sB,GAAG,YAAY,EAAE,YAAY,KAAK9sB,CAAC,CAAC,CAAC,OAAO,mBAAmBA,EAAE,CAAC8sB,GAAG,YAAY,EAAE,YAAY,KAAK9sB,CAAC,CAAC,CAAC,OAAO,gBAAgBA,EAAE,CAAC,OAAO8sB,GAAG,YAAY9sB,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgBA,EAAEZ,EAAE,CAAC,OAAO0tB,GAAG,YAAY9sB,EAAE,OAAOZ,CAAC,CAAC,CAAC,OAAO,YAAYY,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,CAAC,EAAE,OAAOL,IAAI,OAAO0tB,GAAG,YAAY,EAAE,YAAYA,GAAG,YAAY,EAAE,aAAa,QAAQ5rB,GAAG,CAAC,IAAIC,EAAED,EAAElB,EAAE,CAAC,EAAEmB,IAAI,MAAM1B,EAAE,KAAK0B,CAAC,CAAC,CAAC,EAAE1B,CAAC,CAAC,EAAEstB,GAAG7tB,GAAG4tB,GAAG,mBAAmB5tB,CAAC,EAAE8tB,GAAG9tB,GAAG4tB,GAAG,mBAAmB5tB,CAAC,EAAE+tB,GAAG/tB,GAAG4tB,GAAG,gBAAgB5tB,CAAC,EAAEguB,GAAG,CAAChuB,EAAE,IAAI4tB,GAAG,gBAAgB5tB,EAAE,CAAC,EAAMiuB,GAAG,eAAeC,GAAG,EAAEC,GAAG,eAAeC,GAAG,mBAAmB,SAASC,IAAI,CAAC,GAAG,CAAC5T,EAAE,EAAE,QAAQ,YAAY,EAAE,MAAM,IAAI,MAAM,yFAAyF,EAAE,IAAIza,EAAE,OAAO,QAAQ,YAAY,KAAK,OAAO,EAAEA,EAAE,WAAWA,EAAE,cAAcA,EAAE,iBAAiBA,EAAE,aAAaA,EAAE,cAAc,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,2DAA2D,EAAE,OAAO,CAAC,CAAC,SAASsuB,GAAGtuB,EAAE,CAAC,IAAI,EAAEA,EAAE,OAAO,EAAE,kBAAkBmuB,GAAG,CAAC,QAAQ,WAAW,CAAC,EAAE,EAAE,kBAAkBC,GAAG,CAAC,QAAQ,WAAW,CAAC,CAAC,CAAC,IAAIG,GAAG,KAAK,CAAC,YAAYztB,EAAE,CAAC,GAAG,KAAK,UAAUutB,GAAG,EAAEvtB,GAAG,MAAM,CAACA,EAAE,MAAM,IAAI,MAAM,gEAAgE,EAAE,KAAK,UAAUA,CAAC,CAAC,MAAM,KAAKA,EAAE,CAAC,GAAGA,EAAE,yBAAyB,YAAY,MAAM,IAAI,MAAM,0FAA0F,EAAE,OAAO,KAAK,eAAe,KAAK,UAAUA,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,eAAe,KAAK,SAAS,CAAC,CAAC,eAAeA,EAAEZ,EAAE,CAAC,OAAO,IAAI,QAAQ,CAAC,EAAEK,IAAI,CAAC,IAAIwB,EAAE,KAAK,UAAU,KAAKksB,GAAGC,EAAE,EAAEnsB,EAAE,gBAAgB,IAAIusB,GAAGvsB,CAAC,EAAEA,EAAE,UAAU,IAAI,CAAC,IAAIC,EAAED,EAAE,OAAO,GAAG7B,GAAG,KAAK,CAAC,IAAI+B,EAAED,EAAE,YAAYmsB,GAAG,UAAU,EAAEhsB,EAAEF,EAAE,YAAYksB,EAAE,EAAE,IAAI,KAAK,SAAS,EAAEhsB,EAAE,UAAU,IAAI,CAAC,GAAGA,EAAE,QAAQ,KAAK,OAAOH,EAAE,MAAM,EAAEzB,EAAE,IAAI,MAAM,gCAAgC,KAAK,0BAA0B,CAAC,EAAE,EAAE4B,EAAE,OAAO,cAAc,CAAC,EAAEA,EAAE,QAAQS,IAAIZ,EAAE,MAAM,EAAEzB,EAAE4B,EAAE,KAAK,GAAGF,EAAE,WAAW,IAAID,EAAE,MAAM,CAAC,KAAK,CAAC,IAAIC,EAAEsrB,GAAGrtB,CAAC,EAAEgC,EAAEF,EAAE,YAAYosB,GAAG,WAAW,EAAE,EAAElsB,EAAE,YAAYksB,EAAE,EAAExrB,EAAE,EAAE,IAAI,CAAC,UAAU,KAAK,UAAU,mBAAmBX,CAAC,CAAC,EAAE,EAAEW,EAAE,UAAU,IAAI,CAAC,EAAEZ,EAAE,YAAYmsB,GAAG,WAAW,EAAE,IAAIprB,EAAE,EAAE,YAAYorB,EAAE,EAAE,IAAI,CAAC,UAAU,KAAK,UAAU,eAAejuB,EAAE,mBAAmB+B,CAAC,CAAC,EAAEc,EAAE,UAAU,IAAI,EAAE,CAAC,mBAAmBd,CAAC,CAAC,EAAEc,EAAE,QAAQC,GAAG,CAAC,EAAEd,EAAE,YAAYksB,EAAE,EAAE,IAAI,EAAE,EAAE,OAAO,KAAK,SAAS,EAAE,EAAE,UAAU,KAAKpsB,EAAE,MAAM,EAAEzB,EAAEwC,EAAE,KAAK,GAAG,EAAE,QAAQoB,IAAInC,EAAE,MAAM,EAAEzB,EAAEwC,EAAE,KAAK,EAAE,CAAC,EAAEH,EAAE,QAAQE,IAAId,EAAE,MAAM,EAAEzB,EAAEqC,EAAE,KAAK,GAAGV,EAAE,WAAW,IAAI,CAAC,GAAG,KAAKF,EAAE,MAAM,EAAE,EAAE,WAAW,IAAIA,EAAE,MAAM,CAAC,CAAC,CAAC,EAAED,EAAE,QAAQC,GAAGzB,EAAEwB,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAEwsB,GAAG,WAAW,eAAe,IAAIC,GAAGxuB,GAAGya,EAAE,EAAE,QAAQ,YAAY,GAAG,CAAC,MAAM,QAAQza,CAAC,GAAGA,EAAE,WAAWuuB,GAAG,UAAU,EAAEE,GAAGzuB,EAAE,MAAMuuB,GAAG,WAAW,MAAM,CAAC,EAAE,KAAKX,GAAG,mBAAmBY,EAAE,EAAEZ,GAAG,mBAAmBY,EAAE,EAAE,SAASC,GAAGzuB,EAAE,CAAC,OAAO,IAAIuuB,GAAGvuB,CAAC,CAAC,CAAC,SAAS0uB,GAAG1uB,EAAE,CAAC,OAAOA,EAAE,WAAWuuB,GAAG,UAAU,EAAEvuB,EAAE,MAAMuuB,GAAG,WAAW,MAAM,EAAEvuB,CAAC,CAAC,IAAI2uB,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,UAAUN,GAAG,CAAC,CAAC,MAAM,YAAY,CAAC,OAAO,IAAI,QAAQ,CAACvtB,EAAEZ,IAAI,CAAC,IAAI,EAAE,KAAK,UAAU,KAAK+tB,GAAGC,EAAE,EAAE,EAAE,gBAAgB,IAAII,GAAG,CAAC,EAAE,EAAE,UAAU,IAAI,CAAC,IAAI/tB,EAAE,EAAE,OAAOwB,EAAExB,EAAE,YAAY6tB,GAAG,UAAU,EAAEnsB,EAAEF,EAAE,YAAYqsB,EAAE,EAAE,OAAO,EAAEnsB,EAAE,UAAU,IAAI,CAAC,IAAIC,EAAE,CAAC,EAAE,QAAQC,KAAKF,EAAE,OAAOC,EAAEC,EAAE,WAAWA,EAAE,mBAAmBrB,EAAEoB,CAAC,CAAC,EAAED,EAAE,QAAQC,IAAI3B,EAAE,MAAM,EAAEL,EAAE+B,EAAE,KAAK,GAAGF,EAAE,WAAW,IAAIxB,EAAE,MAAM,CAAC,EAAE,EAAE,QAAQA,GAAGL,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,MAAM,YAAYY,EAAE,CAAC,OAAOA,EAAE4tB,GAAG5tB,CAAC,EAAE,IAAI,QAAQ,CAACZ,EAAE,IAAI,CAAC,IAAIK,EAAE,KAAK,UAAU,KAAK0tB,GAAGC,EAAE,EAAE3tB,EAAE,gBAAgB,IAAI+tB,GAAG/tB,CAAC,EAAEA,EAAE,UAAU,IAAI,CAAC,IAAIwB,EAAExB,EAAE,OAAOyB,EAAED,EAAE,YAAYqsB,GAAG,WAAW,EAAEnsB,EAAED,EAAE,YAAYosB,EAAE,EAAElsB,EAAED,EAAE,IAAInB,CAAC,EAAE,EAAEoB,EAAE,UAAU,IAAI,CAAC,GAAGA,EAAE,QAAQ,KAAK,OAAOH,EAAE,MAAM,EAAE,EAAE,IAAI,MAAM,gCAAgCjB,kBAAkB,CAAC,EAAE,CAAC,IAAI8B,EAAEX,EAAE,OAAOnB,CAAC,EAAE,EAAE,IAAI,CAAC,EAAEiB,EAAE,YAAYosB,GAAG,WAAW,EAAE,IAAIprB,EAAE,EAAE,YAAYorB,EAAE,EAAE,OAAOrtB,CAAC,EAAEiC,EAAE,UAAU,IAAI7C,EAAEgC,EAAE,OAAO,kBAAkB,EAAEa,EAAE,QAAQC,GAAG,EAAEd,EAAE,KAAK,CAAC,EAAEU,EAAE,UAAU,EAAEA,EAAE,QAAQE,IAAI,EAAE,EAAEf,EAAE,MAAM,EAAE,EAAEG,EAAE,KAAK,EAAE,CAAC,EAAEA,EAAE,QAAQU,IAAIb,EAAE,MAAM,EAAE,EAAEG,EAAE,KAAK,GAAGF,EAAE,WAAW,IAAI,CAAC,GAAG,KAAKD,EAAE,MAAM,EAAE,EAAE,WAAW,IAAIA,EAAE,MAAM,CAAC,CAAC,EAAExB,EAAE,QAAQwB,GAAG,EAAExB,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAMquB,GAAG,IAAIC,GAAG,sBAAsBC,GAAG,OAAOC,GAAG,iBAAiBC,GAAG,eAAeC,GAAG,cAAcC,GAAG,iBAAiB,SAASC,GAAGnvB,EAAE,CAAC,MAAM,CAAC,KAAK,CAAC6uB,GAAG7uB,EAAE8uB,EAAE,EAAE,KAAKF,EAAE,EAAE,SAAS,CAACC,GAAG7uB,EAAE+uB,EAAE,EAAE,KAAKH,EAAE,EAAE,YAAY,CAACC,GAAG7uB,EAAEgvB,EAAE,EAAE,KAAKJ,EAAE,EAAE,WAAW,CAACC,GAAG7uB,EAAEivB,EAAE,EAAE,KAAKL,EAAE,EAAE,cAAc,CAACC,GAAG7uB,EAAEkvB,EAAE,EAAE,KAAKN,EAAE,CAAC,CAAC,CAAC,SAASQ,GAAGpvB,EAAE,CAAC,QAAQ,KAAK,OAAO,OAAOA,CAAC,EAAE,OAAO,aAAa,WAAW,CAAC,CAAC,CAAC,SAASqvB,GAAGrvB,EAAE,CAAC,IAAI,EAAEA,EAAE,MAAM4uB,EAAE,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,uBAAuB5uB,GAAG,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE,OAAO,CAAC,EAAE,KAAK4uB,EAAE,CAAC,CAAC,SAASU,GAAGtvB,EAAE,CAAC,OAAOA,EAAE,WAAWuvB,GAAG,UAAU,EAAEvvB,EAAE,MAAMuvB,GAAG,WAAW,MAAM,EAAEvvB,CAAC,CAAC,IAAIuvB,GAAG,KAAK,CAAC,YAAYzuB,EAAE,CAAC,GAAG,CAAC2Z,EAAE,EAAE,QAAQ,YAAY,GAAG,OAAO,QAAQ,aAAa,OAAO,OAAO,cAAc,YAAY,MAAM,IAAI,MAAM,yDAAyD,EAAE,GAAG,KAAK,GAAG,OAAO,aAAa3Z,GAAG,MAAM,CAACA,EAAE,MAAM,IAAI,MAAM,oEAAoE,EAAE,KAAK,UAAUA,EAAE,KAAK,KAAKquB,GAAG,KAAK,SAAS,CAAC,CAAC,MAAM,KAAKruB,EAAE,CAAC,GAAGA,EAAE,yBAAyB,YAAY,MAAM,IAAI,MAAM,0FAA0F,EAAE,CAAC,IAAIZ,EAAE,KAAK,UAAUY,EAAE,aAAa,EAAE,EAAE,KAAK,UAAUA,EAAE,WAAW,EAAEP,EAAEgtB,GAAGzsB,CAAC,EAAE,GAAG,CAAC,KAAK,GAAG,QAAQ,KAAK,KAAK,KAAK,KAAK,UAAUP,CAAC,CAAC,EAAE,KAAK,GAAG,QAAQ,KAAK,KAAK,SAASL,CAAC,EAAE,KAAK,GAAG,QAAQ,KAAK,KAAK,YAAY,CAAC,EAAE,KAAK,GAAG,QAAQ,KAAK,KAAK,WAAW8sB,GAAGlsB,EAAE,UAAU,CAAC,EAAE,IAAIiB,EAAE,CAAC,OAAOjB,EAAE,OAAO,YAAYA,EAAE,YAAY,YAAYA,EAAE,YAAY,UAAUA,EAAE,WAAW,KAAKA,EAAE,UAAU,OAAO,oBAAoBA,EAAE,qBAAqB,KAAKA,EAAE,oBAAoB,OAAO,iBAAiBA,EAAE,kBAAkB,KAAKA,EAAE,iBAAiB,OAAO,qBAAqBA,EAAE,sBAAsB,KAAKA,EAAE,qBAAqB,OAAO,eAAeA,EAAE,gBAAgB,KAAKA,EAAE,eAAe,MAAM,EAAE,OAAO,KAAK,GAAG,QAAQ,KAAK,KAAK,cAAc,KAAK,UAAUiB,CAAC,CAAC,EAAE,CAAC,mBAAmBxB,CAAC,CAAC,OAAOwB,EAAN,CAAS,MAAMqtB,GAAG,KAAK,IAAI,EAAE,IAAI,MAAM,yBAAyB,KAAK,kHAAkH7uB,EAAE,wCAAwCA,EAAE,qCAAqCA,EAAE,kBAAkB,CAAC,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,IAAIO,EAAE,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,IAAI,CAAC,EAAE,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,kDAAkD,KAAK,YAAY,EAAE,GAAGA,EAAE,oBAAoB,OAAO,MAAM,IAAI,MAAM,2EAA2E,EAAE,IAAIZ,EAAE,CAAC,EAAE,EAAE,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,QAAQ,CAAC,EAAE,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,4CAA4C,KAAK,wBAAwB,EAAEA,EAAE,cAAc,EAAE,IAAIK,EAAE,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,WAAW,CAAC,EAAE,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,gDAAgD,KAAK,yBAAyB,EAAEL,EAAE,YAAYK,EAAE,IAAIwB,EAAE,KAAK,GAAG,QAAQ,KAAK,KAAK,aAAa,EAAE,GAAGA,GAAG,KAAK,CAAC,IAAIE,EAAE,KAAK,MAAMF,CAAC,EAAE7B,EAAE,OAAO+B,EAAE,OAAO/B,EAAE,YAAY+B,EAAE,YAAY/B,EAAE,YAAY+B,EAAE,YAAYA,EAAE,WAAW,OAAO/B,EAAE,UAAU+B,EAAE,WAAWA,EAAE,qBAAqB,OAAO/B,EAAE,oBAAoB+B,EAAE,qBAAqBA,EAAE,kBAAkB,OAAO/B,EAAE,iBAAiB+B,EAAE,kBAAkBA,EAAE,sBAAsB,OAAO/B,EAAE,qBAAqB+B,EAAE,sBAAsBA,EAAE,gBAAgB,OAAO/B,EAAE,eAAe+B,EAAE,eAAe,CAAC,IAAID,EAAE,KAAK,GAAG,QAAQ,KAAK,KAAK,UAAU,EAAE,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,wDAAwD,KAAK,yBAAyB,EAAE,OAAO9B,EAAE,WAAW+sB,GAAGjrB,CAAC,EAAE9B,CAAC,CAAC,EAAEqvB,GAAG,WAAW,kBAAkB,IAAIC,GAAGxvB,GAAGya,EAAE,EAAE,QAAQ,YAAY,GAAG,CAAC,MAAM,QAAQza,CAAC,GAAGA,EAAE,WAAWuvB,GAAG,UAAU,EAAEE,GAAGzvB,EAAE,MAAMuvB,GAAG,WAAW,MAAM,CAAC,EAAE,KAAK3B,GAAG,mBAAmB4B,EAAE,EAAE5B,GAAG,mBAAmB4B,EAAE,EAAE,SAASC,GAAGzvB,EAAE,CAAC,OAAO,IAAIuvB,GAAGvvB,CAAC,CAAC,CAAC,IAAI0vB,GAAG,KAAK,CAAC,aAAa,CAAC5X,EAAE2C,EAAE,EAAE,QAAQ,YAAY,EAAE,IAAI,0CAA0C,EAAE3C,EAAE,OAAO,QAAQ,aAAa,OAAO,OAAO,cAAc,YAAY,IAAI,yDAAyD,EAAE,KAAK,GAAG,OAAO,YAAY,CAAC,MAAM,YAAY,CAAC,IAAIhX,EAAE,CAAC,EAAEZ,EAAE2uB,GAAGD,GAAG,EAAEA,GAAGE,GAAG,QAAQvuB,EAAE,EAAEA,EAAE,KAAK,GAAG,OAAO,EAAEA,EAAE,CAAC,IAAIwB,EAAE,KAAK,GAAG,IAAIxB,CAAC,EAAE,GAAGwB,EAAE,WAAW7B,CAAC,GAAG6B,EAAE,SAAS,CAAC,EAAE,CAAC,IAAIC,EAAEqtB,GAAGttB,CAAC,EAAEjB,EAAEkB,GAAG,KAAK,MAAM,KAAK,GAAG,QAAQD,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOjB,CAAC,CAAC,MAAM,YAAYA,EAAE,CAACA,EAAEwuB,GAAGxuB,CAAC,EAAE,IAAIZ,EAAEivB,GAAGruB,CAAC,EAAE,GAAG,KAAK,GAAG,QAAQZ,EAAE,IAAI,GAAG,KAAK,MAAM,IAAI,MAAM,8BAA8BY,IAAI,EAAE,IAAI,EAAE,KAAK,MAAM,KAAK,GAAG,QAAQZ,EAAE,IAAI,CAAC,EAAE,OAAOkvB,GAAGlvB,CAAC,EAAE,CAAC,CAAC,EAAMyvB,GAAG,MAAMC,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,SAAS,CAAC,CAAC,CAAC,OAAO,aAAa,CAAC,OAAOA,GAAG,UAAU,OAAOA,GAAG,SAAS,IAAIA,IAAIA,GAAG,QAAQ,CAAC,OAAO,gBAAgB9uB,EAAEZ,EAAE,CAAC4X,EAAEhX,GAAG,KAAK,IAAI,uCAAuC,EAAEA,EAAE,SAAS6uB,EAAE,IAAI7uB,EAAEA,EAAE,MAAM,EAAEA,EAAE,QAAQ6uB,EAAE,CAAC,GAAG7X,EAAEhX,EAAE,OAAO,EAAE,IAAI,qCAAqC,EAAE,IAAI,EAAE8uB,GAAG,YAAY,EAAE9X,EAAE,EAAE,SAAShX,IAAI,KAAK,IAAI,2DAA2DA,KAAK,EAAE,EAAE,SAASA,GAAGZ,CAAC,CAAC,OAAO,WAAWY,EAAE,CAAC,IAAIZ,EAAE0vB,GAAG,YAAY,EAAE,SAAS9uB,GAAG,GAAGZ,GAAG,KAAK,MAAM,IAAI,MAAM,yCAAyCY,IAAI,EAAE,OAAOZ,CAAC,CAAC,OAAO,YAAY,CAAC,OAAO,OAAO,KAAK0vB,GAAG,YAAY,EAAE,QAAQ,CAAC,CAAC,EAAE,SAASC,GAAG7vB,EAAE,CAAC,GAAGA,EAAE,QAAQ2vB,EAAE,IAAI,GAAG,MAAM,IAAI,MAAM,6EAA6EC,GAAG,WAAW,EAAE,KAAK,GAAG,GAAG,EAAE,MAAM,CAAC,OAAO5vB,EAAE,MAAM2vB,EAAE,EAAE,GAAG,KAAK3vB,EAAE,MAAM2vB,EAAE,EAAE,EAAE,CAAC,CAAC,eAAeG,GAAG9vB,EAAE,EAAEE,EAAE,GAAG,CAAC4X,EAAE9X,IAAI,EAAE,IAAI,wCAAwCA,IAAI,EAAE,IAAIM,EAAEstB,GAAG,gBAAgB5tB,CAAC,EAAE8X,EAAExX,EAAE,OAAO,EAAE,IAAI,kEAAkEN,IAAI,EAAE8X,EAAExX,EAAE,OAAO,EAAE,IAAI,yCAAyCA,EAAE,wCAAwCN,IAAI,EAAE,IAAIO,EAAED,EAAE,GAAG,EAAEstB,GAAG,gBAAgB,CAAC,EAAE9V,EAAE,EAAE,OAAO,EAAE,IAAI,uEAAuE,IAAI,EAAEA,EAAE,EAAE,OAAO,EAAE,IAAI,yCAAyCxX,EAAE,6CAA6C,IAAI,EAAE,IAAI,EAAE,EAAE,GAAG2B,EAAE4tB,GAAG7vB,CAAC,EAAE,OAAOkC,EAAE2tB,GAAG7vB,CAAC,EAAE,KAAKmC,EAAEF,IAAI4tB,GAAG7vB,CAAC,EAAE,OAAO4C,EAAE,MAAMrC,EAAE,KAAK,EAAEL,GAAGiC,GAAG,MAAMytB,GAAG,WAAW3tB,CAAC,EAAE,YAAYC,CAAC,EAAE,IAAIW,EAAE,MAAM,EAAE,KAAKD,CAAC,EAAE,OAAO1C,GAAG,CAACiC,GAAG,MAAMytB,GAAG,WAAW3tB,CAAC,EAAE,YAAYC,CAAC,EAAEW,EAAE,kBAAkB,CAAC,eAAektB,IAAI,CAAC,IAAI/vB,EAAE4vB,GAAG,WAAW,EAAE,EAAE,CAAC,EAAE,QAAQ1vB,KAAKF,EAAE,CAAC,IAAIM,EAAE,MAAMsvB,GAAG,WAAW1vB,CAAC,EAAE,WAAW,EAAE,QAAQK,KAAKD,EAAE,CAAC,IAAI,EAAEJ,EAAEyvB,GAAGpvB,EAAE,EAAE,GAAGD,EAAEC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,eAAeyvB,GAAGhwB,EAAE,CAAC,IAAI,EAAE6vB,GAAG7vB,CAAC,EAAE,OAAO4vB,GAAG,WAAW,EAAE,MAAM,EAAE,YAAY,EAAE,IAAI,CAAC,CAAC,eAAeK,GAAGjwB,EAAE,EAAE,CAAC,OAAO8vB,GAAG9vB,EAAE,EAAE,EAAE,CAAC,CAAC,eAAekwB,GAAGlwB,EAAE,EAAE,CAAC,OAAO8vB,GAAG9vB,EAAE,EAAE,EAAE,CAAC,CAAC,IAAImwB,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,YAAY,mBAAmB,KAAK,aAAa,CAAC,EAAE,KAAK,oBAAoB,EAAE,KAAK,iBAAiB,EAAE,CAAC,MAAMrvB,EAAEZ,EAAE,CAAC,OAAO,MAAMY,EAAEZ,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,YAAY,IAAI,CAAC,CAAC,OAAOY,EAAEZ,EAAE,CAAC,GAAGA,IAAI,SAASA,IAAI,OAAO,MAAM,IAAI,MAAM,kDAAkDA,GAAG,EAAE,OAAO,KAAK,aAAa,OAAO,KAAK,YAAY,IAAI,aAAa,KAAK,YAAY,OAAOY,CAAC,CAAC,CAAC,OAAOA,EAAEZ,EAAE,CAAC,OAAO,IAAI,YAAYA,CAAC,EAAE,OAAOY,CAAC,CAAC,CAAC,iBAAiBA,EAAEZ,EAAE,CAAC,GAAG,OAAO,QAAQ,aAAa,CAACua,EAAE,EAAE,QAAQ,sBAAsB,EAAE,CAAC,WAAW3Z,EAAEZ,CAAC,EAAE,MAAM,CAAC,KAAK,aAAa,KAAKY,CAAC,EAAE,WAAW,IAAI,CAAC,OAAO,YAAY,CAAC,KAAK,KAAK,YAAY,MAAM,KAAK,aAAa,OAAO,CAAC,EAAE,GAAG,CAAC,EAAEZ,CAAC,EAAE,KAAK,mBAAmB,KAAK,iBAAiB,GAAG,OAAO,iBAAiB,UAAU,GAAG,CAAC,GAAG,EAAE,SAAS,QAAQ,EAAE,KAAK,OAAO,KAAK,YAAY,CAAC,EAAE,gBAAgB,EAAE,IAAIK,EAAE,KAAK,aAAa,EAAE,KAAK,OAAOA,EAAE,EAAE,KAAK,sBAAsB,KAAK,sBAAsB,KAAK,aAAa,SAAS,KAAK,aAAa,CAAC,EAAE,KAAK,oBAAoB,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,GAAGka,EAAE,EAAE,IAAI,YAAY,EAAE,CAACA,EAAE,EAAE,YAAY,UAAU,IAAI0V,EAAE,EAAE,GAAG,CAACP,GAAG,gBAAgBL,GAAG,WAAW,IAAIG,EAAE,CAAC,OAAO1vB,EAAN,CAAS,CAAC,GAAG,CAAC4vB,GAAG,gBAAgBrB,GAAG,WAAW,IAAII,EAAE,CAAC,OAAO3uB,EAAN,CAAS,CAAC,CAAC,IAAIowB,GAAG,CAAC,YAAY,IAAIntB,GAAG,CAAC,EAAEotB,GAAOC,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,KAAKptB,GAAG,EAAE,KAAK,YAAY,IAAI,KAAK,KAAK,WAAW,CAAC,MAAMpC,EAAEZ,EAAE,CAAC,OAAOua,EAAE,EAAE,OAAO,OAAO,KAAKA,EAAE,EAAE,OAAO,MAAM3Z,EAAEZ,CAAC,GAAGmwB,IAAI,OAAOA,GAAGD,GAAG,YAAY,GAAGC,GAAGvvB,EAAEZ,CAAC,EAAE,CAAC,KAAK,CAAC,IAAIY,EAAE,QAAQ,OAAO,EAAE,OAAOA,EAAE,GAAG,IAAIA,EAAE,GAAG,GAAG,CAAC,OAAOA,EAAEZ,EAAE,CAAC,GAAGA,IAAI,SAASA,IAAI,OAAO,MAAM,IAAI,MAAM,sDAAsDA,GAAG,EAAE,OAAO,KAAK,YAAY,OAAOY,CAAC,CAAC,CAAC,OAAOA,EAAEZ,EAAE,CAAC,OAAOY,EAAE,SAAS,EAAE,GAAG,IAAI,KAAK,KAAK,YAAYZ,CAAC,EAAE,OAAOY,CAAC,CAAC,CAAC,EAAE2Z,EAAE,EAAE,IAAI,SAAS,GAAG,CAACA,EAAE,EAAE,IAAI,YAAY,GAAGA,EAAE,EAAE,YAAY,OAAO,IAAI6V,EAAE,EAAE,SAASC,GAAGvwB,EAAE,EAAE,UAAUE,EAAE,CAAC,OAAO,EAAE,GAAG,UAAUga,GAAGla,CAAC,EAAE,IAAIspB,GAAGtpB,EAAE,EAAEE,CAAC,CAAC,CAAC,SAASswB,GAAGxwB,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,EAAE,GAAG,CAACkZ,GAAG,CAAC,EAAE,MAAM,IAAI,MAAM,mCAAmC,GAAG,EAAE,GAAG,IAAI,UAAUhZ,EAAE,QAAQ,UAAU,IAAI,UAAUA,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,uCAAuC,EAAE,IAAII,EAAE,CAAC,EAAEJ,CAAC,EAAEK,EAAE,CAAC,MAAM,CAAC,EAAE,OAAO2qB,EAAE,UAAUzO,GAAGnc,EAAEC,CAAC,CAAC,CAAC,IAAIkwB,GAAEtE,EAAE,CAAC,MAAMqE,EAAE,CAAC,EAAE,SAASE,GAAG1wB,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,QAAQ,mBAAmB,CAAC,EAAE,OAAOkrB,EAAE,UAAU5L,GAAGpf,CAAC,CAAC,CAAC,IAAIywB,GAAGxE,EAAE,CAAC,OAAOuE,EAAE,CAAC,EAAE,SAASE,GAAG5wB,EAAE,EAAE,GAAG,CAAC,QAAQ,IAAIA,EAAE,SAAS,CAAC,CAAC,CAAC,CAACqrB,GAAG,EAAE,IAAIwF,GAAG,CAAC,OAAON,GAAG,KAAKE,GAAE,MAAME,GAAG,MAAMC,EAAE,EAAEjH,GAAGkH,EAAE,EAAE,IAAIC,GAAG,CAAC,EAAE1wB,GAAG0wB,GAAG,CAAC,aAAa,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,wBAAwB,IAAI9D,GAAG,UAAU,IAAI+C,GAAG,cAAc,IAAIrD,GAAG,cAAc,IAAIF,GAAG,WAAW,IAAIuE,GAAG,eAAe,IAAIC,GAAG,gBAAgB,IAAIlD,GAAG,yBAAyB,IAAIV,GAAG,6BAA6B,IAAID,GAAG,6BAA6B,IAAIE,GAAG,gBAAgB,IAAIQ,GAAG,eAAe,IAAIP,GAAG,KAAK,IAAI2D,GAAG,aAAa,IAAIC,GAAG,WAAW,IAAIrB,GAAG,YAAY,IAAIsB,GAAG,UAAU,IAAInB,GAAG,mBAAmB,IAAIpC,GAAG,mBAAmB,IAAID,GAAG,YAAY,IAAImC,GAAG,qBAAqB,IAAIsB,GAAG,gBAAgB,IAAIC,GAAG,oBAAoB,IAAIC,EAAE,CAAC,EAAE,IAAIC,GAAG,QAAQC,GAAG,QAAQC,GAAG,eAAe,SAASC,GAAG5xB,EAAE,CAAC,OAAO,IAAI,QAAQ,GAAG,WAAW,CAAC,CAAC,EAAE,KAAKA,CAAC,CAAC,CAAC,IAAI6xB,GAAG,KAAK,CAAC,YAAY/wB,EAAE,CAAC,GAAG,CAAC2Z,EAAE,EAAE,QAAQ,YAAY,EAAE,MAAM,IAAI,MAAM,qFAAqF,EAAE3Z,EAAE,WAAW+wB,GAAG,UAAU,IAAI/wB,EAAEA,EAAE,MAAM+wB,GAAG,WAAW,MAAM,IAAI/wB,GAAG,MAAMA,EAAE,SAAS,KAAKA,EAAE2wB,IAAI,KAAK,kBAAkB3wB,EAAE4wB,GAAG,KAAK,mBAAmB5wB,EAAE6wB,EAAE,CAAC,MAAM,KAAK7wB,EAAE,CAAC,GAAG,OAAO,UAAU,YAAY,MAAM,IAAI,MAAM,yFAAyF,EAAE,IAAIZ,EAAE,OAAO,IAAI,gBAAgB,IAAI,KAAK,CAACY,EAAE,UAAU,EAAE,CAAC,KAAK,0BAA0B,CAAC,CAAC,EAAE,GAAGA,EAAE,yBAAyB,YAAY,MAAM,IAAI,MAAM,uFAAuF,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,KAAK,KAAK,kBAAkB,EAAE,QAAQA,EAAE,WAAW,CAAC,EAAEP,EAAE6sB,GAAGtsB,EAAE,CAAC,EAAEiB,EAAE,OAAO,IAAI,gBAAgB,IAAI,KAAK,CAAC,KAAK,UAAUxB,CAAC,CAAC,EAAE,CAAC,KAAK,kBAAkB,CAAC,CAAC,EAAEyB,EAAE,KAAK,iBAAiB,KAAK,SAAS,cAAc,GAAG,EAAE,KAAK,gBAAgB,GAAGA,EAAE,SAAS,KAAK,kBAAkBA,EAAE,KAAKD,EAAE,MAAM6vB,GAAG,IAAI5vB,EAAE,cAAc,IAAI,WAAW,OAAO,CAAC,CAAC,EAAElB,EAAE,YAAY,KAAK,CAAC,IAAImB,EAAE,KAAK,kBAAkB,KAAK,SAAS,cAAc,GAAG,EAAE,KAAK,iBAAiBA,EAAE,SAAS,KAAK,mBAAmBA,EAAE,KAAK/B,EAAE,MAAM0xB,GAAG,IAAI3vB,EAAE,cAAc,IAAI,WAAW,OAAO,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,mBAAmBsrB,GAAGzsB,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE+wB,GAAG,WAAW,eAAe,IAAIC,GAAG,KAAK,CAAC,YAAYhxB,EAAE,CAAC,GAAGA,GAAG,MAAMA,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,wEAAwEA,GAAG,EAAE,KAAK,SAASA,EAAE,GAAG,KAAK,aAAaA,EAAE,MAAM,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,IAAI,QAAQ,CAACA,EAAEZ,IAAI,CAAC,IAAI,EAAE,IAAI,WAAW,EAAE,OAAOK,GAAG,CAAC,IAAIwB,EAAE,KAAK,MAAMxB,EAAE,OAAO,MAAM,EAAEyB,EAAED,EAAE,cAAc,GAAGC,GAAG,KAAK,CAAC9B,EAAE,IAAI,MAAM,4CAA4C,KAAK,SAAS,MAAM,CAAC,EAAE,MAAM,CAAC,GAAG6B,EAAE,iBAAiB,KAAK,CAAC7B,EAAE,IAAI,MAAM,6CAA6C,KAAK,SAAS,MAAM,CAAC,EAAE,MAAM,CAAC,GAAG,KAAK,aAAa,SAAS,EAAE,CAACY,EAAE,CAAC,cAAckB,CAAC,CAAC,EAAE,MAAM,CAAC,IAAIE,EAAEorB,GAAGvrB,EAAEI,GAAG,KAAK,YAAYA,CAAC,CAAC,EAAErB,EAAEoB,CAAC,CAAC,EAAE,EAAE,QAAQ3B,GAAGL,EAAE,sEAAsE,KAAK,SAAS,2EAA2E,EAAE,EAAE,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,YAAYY,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQ8B,KAAKlB,EAAEZ,EAAE,KAAK,GAAG8B,EAAE,OAAO,EAAE,EAAE,KAAK,GAAGA,EAAE,KAAK,EAAE,IAAIzB,EAAE,KAAK,4BAA4BO,CAAC,EAAEiB,EAAE,EAAE,IAAIC,GAAG,KAAK,gBAAgBA,EAAEzB,EAAEyB,EAAE,CAAC,EAAE,OAAO,QAAQ,IAAID,CAAC,EAAE,KAAKC,GAAG,CAAC9B,EAAEgtB,GAAGlrB,CAAC,CAAC,CAAC,CAAC,CAAC,gBAAgBlB,EAAEZ,EAAE,CAAC,OAAO,IAAI,QAAQ,CAAC,EAAEK,IAAI,CAAC,IAAIwB,EAAE,IAAI,WAAWA,EAAE,OAAOC,GAAG,CAAC,IAAIC,EAAED,EAAE,OAAO,OAAO,EAAEC,CAAC,CAAC,EAAEF,EAAE,QAAQC,GAAGzB,EAAE,6CAA6CO,KAAK,EAAEiB,EAAE,kBAAkB7B,CAAC,CAAC,CAAC,CAAC,CAAC,4BAA4BY,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,EAAE,KAAK,aAAa,IAAI6B,GAAGorB,GAAGprB,EAAE,IAAI,CAAC,EAAExB,EAAE,CAAC,EAAE,QAAQwB,KAAKjB,EAAEiB,EAAE,MAAM,QAAQC,GAAG,CAAC,IAAIC,EAAEkrB,GAAGnrB,CAAC,EAAE,GAAG9B,EAAE,QAAQ+B,CAAC,IAAI,GAAG,MAAM,IAAI,MAAM,uDAAuDA,IAAI,EAAE,GAAG/B,EAAE,KAAK+B,CAAC,EAAE,EAAE,QAAQA,CAAC,IAAI,GAAG,MAAM,IAAI,MAAM,8BAA8BA,qBAAqB,EAAE1B,EAAEyB,GAAG,KAAK,aAAa,EAAE,QAAQC,CAAC,EAAE,CAAC,EAAE,GAAG/B,EAAE,SAAS,KAAK,aAAa,OAAO,MAAM,IAAI,MAAM,wDAAwDA,EAAE,oDAAoD,KAAK,aAAa,UAAU,EAAE,OAAOK,CAAC,CAAC,EAAEwxB,GAAG/xB,GAAGya,EAAE,EAAE,QAAQ,YAAY,GAAG,CAAC,MAAM,QAAQza,CAAC,GAAGA,EAAE,WAAW6xB,GAAG,UAAU,EAAEG,GAAGhyB,EAAE,MAAM6xB,GAAG,WAAW,MAAM,CAAC,EAAE,KAAKjE,GAAG,mBAAmBmE,EAAE,EAAE,SAASC,GAAGhyB,EAAE,QAAQ,CAAC,OAAO,IAAI6xB,GAAG7xB,CAAC,CAAC,CAAC,SAAS+wB,GAAG/wB,EAAE,CAAC,OAAO,IAAI8xB,GAAG9xB,CAAC,CAAC,CAAC,SAASiyB,GAAGjyB,EAAE,EAAEE,EAAEI,EAAE,CAAC,EAAEN,CAAC,EAAEE,EAAEA,GAAG,KAAK,EAAEA,EAAEI,EAAEA,GAAG,KAAK,EAAEA,EAAE2B,EAAE/B,EAAEI,CAAC,EAAE,IAAIC,EAAE,EAAE,EAAE2B,IAAIA,EAAE,KAAKC,GAAG,CAAC,IAAIS,EAAE1C,GAAG,EAAEK,EAAEP,EAAE,QAAQM,EAAEJ,GAAG,OAAO,EAAE0C,CAAC,EAAET,CAAC,CAAC,EAAED,GAAG,SAAS,EAAEA,EAAE,CAAC4V,EAAE5V,GAAG,MAAM,MAAM,QAAQA,CAAC,GAAGA,EAAE,OAAO,EAAE,IAAI,qCAAqC,CAAC,CAAC,SAASD,EAAEC,EAAEC,EAAE,CAAC2V,EAAE5V,GAAG,GAAGA,GAAG,EAAE,IAAI,oEAAoEA,GAAG,EAAE4V,EAAE3V,GAAG,GAAGA,GAAG,EAAE,IAAI,kEAAkEA,GAAG,EAAE2V,EAAE3V,GAAGD,EAAE,IAAI,yEAAyEA,qBAAqBC,GAAG,CAAC,CAAC,OAAO,QAAQ,IAAInC,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,eAAekyB,GAAGlyB,EAAE,EAAE,CAAC,GAAG,OAAO,EAAE,CAAC,GAAG,IAAIE,EAAE,EAAE,WAAW,KAAKua,EAAE,EAAE,SAAS,MAAM,EAAE,UAAUna,EAAEN,EAAE,IAAI6C,GAAG3C,EAAE2C,EAAE,EAAE,YAAY,CAAC,SAAS,EAAE,CAAC,CAAC,EAAEtC,EAAE,EAAE,EAAE,GAAG0B,GAAG,EAAE,YAAY,KAAK,MAAM,QAAQ,IAAI3B,CAAC,EAAE,MAAM2xB,GAAG3xB,EAAE,EAAE,WAAWC,EAAE,CAAC,GAAG,IAAIsC,GAAGA,EAAE,YAAY,CAAC,EAAEX,EAAE,GAAG,EAAE,EAAE,OAAO,EAAE,YAAY,KAAK,MAAM,QAAQ,IAAID,CAAC,EAAE,MAAMgwB,GAAGhwB,EAAE,EAAE,WAAWC,EAAE,CAAC,CAAC,CAAC,eAAemvB,GAAGrxB,EAAE,EAAE,GAAGE,EAAEI,EAAE,CAAC,OAAOgxB,GAAGtvB,GAAGkwB,GAAGlwB,EAAE,CAAC,YAAY1B,CAAC,CAAC,CAAC,EAAEN,EAAE,EAAEE,CAAC,CAAC,CAAC,SAASoxB,GAAGtxB,EAAE,CAAC,MAAO,OAAM,EAAEE,EAAE,GAAGI,IAAI,CAAC,IAAIC,EAAE,EAAE,IAAI,IAAI,EAAE,EAAE,EAAE,CAAC,EAAE,EAAED,GAAG,KAAKA,EAAE,IAAI,IAAI,EAAE,EAAE,CAAC,EAAE2B,EAAE,CAAC,EAAE,GAAG,EAAE,QAAQ,CAACc,EAAEC,IAAI,CAAC,IAAIkB,EAAE,EAAEnB,EAAE,QAAQ,QAAQ,GAAG,CAAC,IAAIwE,EAAE,iBAAiB,EAAE,EAAE,aAAa,MAAM,EAAE,MAAM,EAAEilB,GAAGjlB,GAAG6Q,GAAG,EAAE,KAAK,EAAEzT,EAAE,IAAI,CAACpE,EAAEyC,GAAG,GAAG,EAAEA,IAAI,OAAO,EAAEA,GAAG,CAAC,GAAG,EAAEA,GAAG,KAAK,CAAC,cAAc,EAAE,YAAYkB,EAAE,UAAU,CAAC,CAAC,CAAC,EAAE5D,GAAG,KAAKA,EAAE,QAAQ,CAACsE,EAAEC,IAAI,CAACD,IAAI,EAAE,OAAOD,EAAE,EAAE,EAAEE,GAAG,GAAG,CAAC,EAAEF,EAAE,EAAE1C,EAAE,KAAK,EAAE,IAAI,EAAEiC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,MAAMnB,GAAGA,CAAC,EAAE,CAAC,IAAIA,EAAEzC,EAAE,OAAO,CAAC0C,EAAEkB,IAAI,CAAC,EAAEA,EAAE,EAAE,MAAM,IAAI,MAAM,kDAAkDnB,EAAE,KAAK,IAAI;AAAA,wCACrmoDd,EAAE,KAAK,IAAI,IAAI,CAAC,CAAC,IAAIC,EAAE3B,EAAE,OAAO,CAACwC,EAAEC,EAAEkB,KAAKlB,GAAGD,EAAE,KAAKmB,CAAC,EAAEnB,GAAG,CAAC,CAAC,EAAEZ,EAAE,CAAC,EAAED,EAAE,QAAQa,GAAG,CAAC,EAAEA,GAAG,MAAM,QAAQC,GAAG,CAAC,IAAIkB,EAAEhE,GAAGA,EAAE,SAAS,GAAG,EAAE,GAAG,KAAK8C,EAAEb,EAAE,KAAK+B,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,IAAItB,EAAE,MAAM5C,EAAEmC,CAAC,EAAEU,EAAE,CAAC,EAAEC,EAAE,EAAE,OAAOZ,EAAE,QAAQa,GAAG,CAAC,IAAIC,EAAE,EAAED,GAAG,MAAM,OAAOmB,EAAE,EAAE,QAAQU,EAAE,EAAEA,EAAE5B,EAAE4B,IAAIV,GAAGtB,EAAEE,EAAE8B,GAAG,WAAW,IAAI,EAAE,IAAI,YAAYV,CAAC,EAAEqD,EAAE,IAAI,WAAW,CAAC,EAAE,EAAE,EAAE,QAAQ3C,EAAE,EAAEA,EAAE5B,EAAE4B,IAAI,CAAC,IAAIC,EAAE,IAAI,WAAWjC,EAAEE,EAAE8B,EAAE,EAAE2C,EAAE,IAAI1C,EAAE,CAAC,EAAE,GAAGA,EAAE,UAAU,CAAC,EAAE9B,GAAG,QAAQ6B,GAAG,CAAC,IAAIC,EAAE,EAAE,MAAMD,EAAE,YAAYA,EAAE,YAAYA,EAAE,SAAS,EAAEE,EAAE8nB,GAAG/nB,EAAE,CAACD,EAAE,aAAa,CAAC,EAAE,QAAQ4C,KAAK1C,EAAEjC,EAAE2E,GAAG1C,EAAE0C,EAAE,CAAC,EAAE1E,GAAGE,CAAC,CAAC,EAAEH,CAAC,CAAC,CAAC,IAAIsvB,GAAG,2BAA2BC,GAAG,mBAAmBC,GAAG,KAAK,CAAC,YAAYvxB,EAAEZ,EAAE,CAAC,GAAG,KAAK,eAAe,OAAOA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,iBAAiBA,EAAE,iBAAiB,KAAK,WAAWA,EAAE,WAAW,KAAK,mBAAmBA,EAAE,mBAAmBA,EAAE,WAAW,MAAM4X,EAAE,OAAO5X,EAAE,WAAW,WAAW,IAAI,6HAA6H,EAAE,KAAK,MAAMA,EAAE,WAAW,KAAK,MAAMua,EAAE,EAAE,SAAS,MAAM3C,EAAEhX,GAAG,MAAMA,EAAE,OAAO,EAAE,IAAI,yDAAyD,EAAE,MAAM,QAAQA,CAAC,GAAGgX,EAAEhX,EAAE,SAAS,EAAE,IAAI,iEAAiEA,EAAE,UAAU,EAAE,KAAK,KAAKA,EAAEZ,EAAE,aAAa,MAAMA,EAAE,YAAY,MAAM,KAAK,MAAM,IAAI,MAAM,oEAAoE,EAAE,KAAK,YAAYA,EAAE,aAAa,CAAC,CAAC,CAAC,MAAM,KAAKY,EAAE,CAAC,GAAGA,EAAE,yBAAyB,YAAY,MAAM,IAAI,MAAM,yFAAyF,EAAE,IAAIZ,EAAE,OAAO,OAAO,CAAC,OAAO,KAAK,cAAc,EAAE,KAAK,WAAW,EAAEA,EAAE,KAAK,IAAI,SAAS,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,qBAAqB,EAAE,QAAQY,EAAE,WAAW,CAAC,EAAEP,EAAE6sB,GAAGtsB,EAAE,CAAC,EAAEZ,EAAE,KAAK,OAAO,aAAa,IAAI,KAAK,CAAC,KAAK,UAAUK,CAAC,CAAC,EAAE,CAAC,KAAK6xB,EAAE,CAAC,EAAE,YAAY,EAAEtxB,EAAE,YAAY,MAAMZ,EAAE,KAAK,OAAO,oBAAoB,IAAI,KAAK,CAACY,EAAE,UAAU,EAAE,CAAC,KAAKqxB,EAAE,CAAC,EAAE,mBAAmB,EAAE,IAAIpwB,EAAE,MAAM,KAAK,MAAM,KAAK,KAAK7B,CAAC,EAAE,GAAG6B,EAAE,GAAG,MAAM,CAAC,mBAAmBwrB,GAAGzsB,CAAC,EAAE,UAAU,CAACiB,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,gEAAgEA,EAAE,SAAS,CAAC,CAAC,MAAM,MAAM,CAAC,IAAIjB,EAAE,MAAM,KAAK,MAAM,KAAK,KAAK,KAAK,WAAW,EAAE,GAAG,CAACA,EAAE,GAAG,MAAM,IAAI,MAAM,cAAc,KAAK,gCAAgCA,EAAE,+EAA+E,EAAE,IAAIZ,EAAE,GAAG,CAACA,EAAE,MAAMY,EAAE,KAAK,CAAC,OAAOiB,EAAN,CAAS,IAAIC,EAAE,+CAA+C,KAAK,QAAQ,MAAM,KAAK,KAAK,SAAS,KAAK,EAAEA,GAAG,+UAA+UA,GAAG,uEAAuE,IAAI,MAAMA,CAAC,CAAC,CAAC,IAAI,EAAE9B,EAAE,cAAcK,EAAEL,EAAE,gBAAgB,GAAG,GAAG,MAAMK,GAAG,KAAK,MAAM,IAAI,MAAM,2BAA2B,KAAK,+DAA+D,EAAE,OAAO+sB,GAAGptB,EAAE6B,GAAG,KAAK,YAAYA,CAAC,CAAC,CAAC,CAAC,MAAM,YAAYjB,EAAE,CAAC,IAAIZ,EAAE,MAAM,QAAQ,KAAK,IAAI,EAAE,KAAK,KAAK,GAAG,KAAK,KAAK,CAAC,EAAEK,CAAC,EAAE+xB,GAAGpyB,CAAC,EAAE6B,EAAE,KAAK,kBAAkB,EAAEC,EAAEwrB,GAAG1sB,CAAC,EAAEmB,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQU,KAAK9B,EAAE,QAAQ,KAAK8B,EAAE,MAAM,KAAK,oBAAoB,KAAKV,EAAE,KAAK,KAAK,mBAAmB,CAAC,CAAC,EAAED,EAAE,KAAKF,EAAE,EAAExB,CAAC,EAAE,KAAK,oBAAoB0B,EAAE,KAAK,GAAG,MAAM,QAAQ,IAAIC,CAAC,CAAC,EAAE,IAAI,EAAE,MAAMgwB,GAAGjwB,EAAE,CAAC,YAAY,KAAK,YAAY,UAAU,KAAK,MAAM,WAAW,KAAK,UAAU,CAAC,EAAE,MAAM,CAACD,EAAEkrB,GAAG,CAAC,CAAC,CAAC,CAAC,EAAEmF,GAAG,iBAAiB,eAAe,SAASC,GAAGtyB,EAAE,CAAC,IAAI,EAAEA,EAAE,YAAY,GAAG,EAAEE,EAAEF,EAAE,YAAY,GAAG,EAAEM,EAAEN,EAAE,UAAU,EAAE,CAAC,EAAEO,EAAEL,EAAE,EAAEF,EAAE,UAAUE,CAAC,EAAE,GAAG,MAAM,CAACI,EAAE,IAAIC,CAAC,CAAC,CAAC,SAAS6wB,GAAGpxB,EAAE,CAAC,OAAOA,EAAE,MAAMqyB,GAAG,gBAAgB,GAAG,IAAI,CAAC,IAAIE,GAAG,CAACvyB,EAAE,IAAI,CAAC,GAAG,OAAO,OAAO,cAAc,GAAG,MAAM,EAAE,WAAW,MAAM,OAAO,KAAK,CAAC,IAAIE,EAAE,GAAG,GAAG,MAAM,QAAQF,CAAC,EAAEE,EAAEF,EAAE,MAAMM,GAAG8wB,GAAG9wB,CAAC,CAAC,EAAEJ,EAAEkxB,GAAGpxB,CAAC,EAAEE,EAAE,OAAOixB,GAAGnxB,EAAE,CAAC,CAAC,CAAC,OAAO,IAAI,EAAE4tB,GAAG,mBAAmB2E,EAAE,EAAE3E,GAAG,mBAAmB2E,EAAE,EAAE,SAASpB,GAAGnxB,EAAE,EAAE,CAAC,OAAO,IAAIqyB,GAAGryB,EAAE,CAAC,CAAC,CAAC,SAASgxB,GAAGhxB,EAAE,EAAE,CAAC,OAAOmxB,GAAGnxB,EAAE,CAAC,CAAC,CAAC,IAAIwyB,GAAG,KAAK,CAAC,YAAY1xB,EAAE,CAAC,KAAK,eAAeA,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,cAAc,CAAC,EAAE2xB,GAAG,KAAK,CAAC,YAAY3xB,EAAE,CAAC,KAAK,YAAYA,CAAC,CAAC,KAAKA,EAAE,CAAC,OAAO,KAAK,YAAYA,CAAC,CAAC,CAAC,EAAE4xB,GAAG,KAAK,CAAC,YAAY5xB,EAAE,CAACA,EAAE,OAAO,KAAK,KAAK,IAAI,QAAQ,QAAQA,EAAE,KAAK,CAAC,GAAGA,EAAE,OAAO,KAAK,KAAKZ,GAAG,QAAQ,QAAQY,EAAE,KAAKZ,CAAC,CAAC,EAAE,CAAC,EAAE,SAAS+wB,GAAGjxB,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,UAAU,OAAO,IAAImyB,GAAGxB,GAAG,GAAG3wB,CAAC,CAAC,CAAC,CAAC,SAAS2wB,GAAGlxB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,UAAU,SAAS,EAAEN,EAAE,eAAe,MAAMA,EAAE,aAAa,KAAK,IAAIwyB,GAAGxyB,CAAC,GAAG,QAAQ,KAAK,uNAAuN,EAAE,IAAIwyB,GAAG,CAAC,cAAcxyB,CAAC,CAAC,IAAI,QAAQ,KAAK,uNAAuN,EAAE,IAAIwyB,GAAG,CAAC,cAAcxyB,EAAE,YAAY,EAAE,WAAWE,EAAE,eAAeI,CAAC,CAAC,EAAE,CAAC,SAASixB,GAAGvxB,EAAE,CAAC,OAAO,IAAIyyB,GAAGzyB,CAAC,CAAC,CAAC,SAASwxB,GAAGxxB,EAAE,CAAC,OAAO,IAAIyyB,GAAGzyB,CAAC,CAAC,CAAC,IAAI2yB,GAAG,CAAC,EAAEvyB,GAAGuyB,GAAG,CAAC,gBAAgB,IAAIC,EAAE,CAAC,EAAE,SAASC,GAAG7yB,EAAE,EAAEE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,IAAI,QAAQ,EAAE,EAAEgsB,EAAE,EAAE,IAAI,QAAQ,EAAE,CAACzrB,EAAE,CAAC,EAAE6pB,GAAG7pB,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC,EAAEA,EAAE,EAAE,CAAC,EAAE0B,EAAE,CAAC,WAAW/B,EAAE,WAAWI,CAAC,EAAE,OAAO4qB,EAAE,UAAU9O,GAAG,EAAEna,CAAC,CAAC,CAAC,IAAI6wB,GAAG3G,EAAE,CAAC,QAAQ0G,EAAE,CAAC,EAAE,SAASE,GAAG/yB,EAAE,EAAEE,EAAE,EAAEI,EAAE,EAAEC,EAAE,QAAQ,CAAC,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,iDAAiD,GAAG,EAAE,IAAIyB,EAAE,CAAC,QAAQgqB,EAAEhsB,EAAE,UAAU,SAAS,OAAO,CAAC,EAAEiC,EAAE,CAAC,MAAM1B,EAAE,MAAM,EAAE,QAAQL,EAAE,SAASI,CAAC,EAAE,OAAO4qB,EAAE,UAAUpJ,GAAG9f,EAAEC,CAAC,CAAC,CAAC,IAAI+wB,GAAG7G,EAAE,CAAC,QAAQ4G,EAAE,CAAC,EAAE,SAASE,IAAK,CAACxY,EAAE,EAAE,IAAI,OAAO,EAAE,CAAC,CAAC,SAASyY,IAAK,CAACzY,EAAE,EAAE,IAAI,QAAQ,EAAE,CAAC,CAAC,SAAS0Y,IAAK,CAAC1Y,EAAE,EAAE,IAAI,+BAA+B,EAAE,EAAE,QAAQ,KAAK,wDAAwD,CAAC,CAAC,SAAS2Y,GAAGpzB,EAAE,CAACya,EAAE,EAAE,QAAQ,8BAA8B,GAAG,QAAQ,KAAKza,EAAE,6EAA6E,CAAC,CAAC4pB,GAAGwJ,EAAE,EAAE,SAASC,IAAK,CAACnI,EAAE,iBAAiB,CAAC,CAAC,SAASoI,IAAI,CAAC,OAAOpI,CAAC,CAAC,SAASqI,IAAI,CAAC,OAAOrI,EAAE,OAAO,CAAC,CAAC,SAASsI,GAAIxzB,EAAE,CAAC,OAAOkrB,EAAE,QAAQlrB,CAAC,CAAC,CAAC,SAASyzB,EAAEzzB,EAAE,EAAE,CAAC,OAAOkrB,EAAE,KAAKlrB,EAAE,CAAC,CAAC,CAAC,SAAS0zB,GAAG1zB,EAAE,CAACkqB,GAAGlqB,CAAC,EAAE,QAAQE,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,SAASyzB,GAAG3zB,EAAE,CAAC,OAAOkrB,EAAE,KAAKlrB,CAAC,CAAC,CAAC,SAAS4zB,GAAI5zB,EAAE,CAAC,OAAOkrB,EAAE,KAAKlrB,CAAC,CAAC,CAAC,SAAS6zB,GAAG7zB,EAAE,CAAC,OAAOkrB,EAAE,WAAWlrB,CAAC,CAAC,CAAC,SAAS8zB,IAAK,CAAC,OAAO5I,EAAE,MAAM,CAAC,CAAC,SAAS6I,IAAK,CAAC,OAAO7I,EAAE,WAAW,CAAC,SAAS8I,GAAIh0B,EAAE,CAACkrB,EAAE,cAAclrB,CAAC,CAAC,CAAC,SAASi0B,GAAIj0B,EAAE,CAAC,OAAOkrB,EAAE,YAAYlrB,CAAC,CAAC,CAAC,SAASk0B,GAAIl0B,EAAE,CAAC,OAAOkrB,EAAE,mBAAmBlrB,CAAC,CAAC,CAAC,SAASm0B,GAAGn0B,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAOgrB,EAAE,gBAAgBlrB,EAAE,EAAEE,CAAC,CAAC,CAAC,SAASk0B,IAAI,CAAC,OAAOlJ,EAAE,OAAO,CAAC,SAASmJ,GAAIr0B,EAAE,EAAE,CAACya,EAAE,EAAE,YAAYza,EAAE,CAAC,CAAC,CAAC,SAASs0B,GAAGt0B,EAAE,CAAC,IAAIE,EAAE,CAAC,MAAM8rB,EAAEhsB,EAAE,QAAQ,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAU1L,GAAGtf,CAAC,CAAC,CAAC,IAAIq0B,GAAGpI,EAAE,CAAC,MAAMmI,EAAE,CAAC,EAAE,SAASE,GAAGx0B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,CAAC,EAAE,OAAOkrB,EAAE,UAAU1J,GAAGthB,CAAC,CAAC,CAAC,IAAIu0B,GAAGtI,EAAE,CAAC,KAAKqI,EAAE,CAAC,EAAE,SAASE,GAAG10B,EAAE,CAAC,IAAIE,EAAE,CAAC,MAAM8rB,EAAEhsB,EAAE,QAAQ,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUzI,GAAGviB,CAAC,CAAC,CAAC,IAAIy0B,GAAGxI,EAAE,CAAC,MAAMuI,EAAE,CAAC,EAAE,SAASE,GAAG50B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,WAAW,EAAE,GAAG,GAAG,OAAO,EAAEM,EAAE,MAAM,IAAI,CAAC,EAAE2B,IAAIA,CAAC,EAAE,QAAQ,GAAG6V,EAAExX,EAAE,OAAO,EAAE,OAAO,IAAI,qCAAqCA,EAAE,kCAAkC,IAAI,EAAE,EAAE,QAAQ,GAAG,CAACwX,EAAE,GAAG,GAAG,EAAExX,EAAE,KAAK,IAAI,+CAA+CA,EAAE,KAAK,aAAa,GAAG,CAAC,CAAC,EAAEA,EAAE,MAAM,EAAE,OAAOA,EAAE,MAAM,EAAE,IAAIC,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAE,OAAOA,EAAE,QAAQ,YAAYmzB,EAAE,IAAI,CAAC,IAAI,EAAEkB,GAAGr0B,CAAC,EAAE2B,EAAEsyB,GAAGj0B,CAAC,EAAE,OAAO,EAAE4qB,EAAE,UAAU7F,GAAG,CAAC,EAAE,CAAC,EAAE,CAAC,EAAEpjB,EAAEipB,EAAE,UAAU7F,GAAG,CAAC,EAAEpjB,CAAC,EAAE,CAAC,EAAE/B,IAAI+B,EAAEwyB,GAAGxyB,CAAC,GAAGoqB,GAAG,EAAEpqB,CAAC,CAAC,CAAC,EAAEipB,EAAE,UAAU7F,GAAG9kB,EAAE,CAAC,CAAC,CAAC,IAAIs0B,GAAG1I,EAAE,CAAC,WAAWyI,EAAE,CAAC,EAAE,SAASE,GAAG90B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,SAAS,iBAAiB,EAAEO,EAAEyrB,EAAE,EAAE,cAAc,iBAAiB,EAAElU,EAAE5X,GAAG,MAAMA,EAAE,GAAG,OAAO,UAAUA,CAAC,EAAE,IAAI,+DAA+DA,GAAG,EAAE4X,EAAExX,EAAE,OAAO,EAAE,IAAI,gDAAgDA,EAAE,MAAM,EAAEwX,EAAEvX,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,MAAM,EAAEuX,EAAExX,EAAE,MAAM,KAAKC,EAAE,MAAM,GAAG,IAAI,uCAAuCD,EAAE,MAAM,UAAUC,EAAE,MAAM,qEAAqE,EAAEuX,EAAE5X,EAAE,GAAG,OAAO,UAAUA,CAAC,EAAE,IAAI,4DAA4DA,GAAG,EAAE,IAAI,EAAE8yB,GAAGvC,GAAEnwB,EAAE,OAAO,EAAEJ,CAAC,EAAE,EAAE8yB,GAAGvC,GAAElwB,EAAE,OAAO,EAAEL,CAAC,EAAE+B,EAAE4yB,GAAG,CAAC,EAAE3yB,EAAE4wB,GAAG7wB,EAAE,CAAC,EAAE,OAAOwuB,GAAEvuB,EAAE,OAAO,CAAC,CAAC,IAAI0wB,GAAGzG,EAAE,CAAC,iBAAiB2I,EAAE,CAAC,EAAMC,GAAG,CAAC,EAAE30B,GAAG20B,GAAG,CAAC,2BAA2B,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,iBAAiB,IAAIC,EAAE,CAAC,EAAE,SAASD,GAAGj1B,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAOM,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEL,EAAEK,IAAI,CAAC,IAAI,EAAEL,EAAE,EAAEK,EAAE,EAAEP,EAAE,IAAI,GAAG,EAAE,EAAE,OAAO,EAAEO,IAAI,GAAG,GAAG,IAAI,GAAGD,EAAE,QAAQ,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,SAAS40B,GAAGl1B,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,CAAC,IAAIC,EAAEP,EAAEA,EAAE,OAAOM,EAAE,GAAG,EAAE,EAAE,OAAOA,EAAE,EAAE,EAAE,EAAE,IAAIC,GAAG,MAAMA,IAAI,GAAG,EAAE,IAAIL,EAAE,QAAQ,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,SAAS80B,GAAGh1B,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAE,KAAK,IAAIN,EAAE,OAAO,EAAE,MAAM,EAAE,QAAQO,EAAE,EAAEA,EAAED,EAAEC,IAAI,CAAC,IAAI,EAAEP,EAAEA,EAAE,OAAOO,EAAE,GAAG,GAAG,OAAO,EAAE,GAAG,IAAI,EAAE,EAAE,EAAE,OAAOA,EAAE,GAAG,GAAG,GAAG,OAAO,EAAE,GAAG,IAAI,EAAEL,EAAE,QAAQ,CAAC,UAAU,IAAI,EAAEA,EAAE,QAAQ,CAAC,UAAU,IAAI,EAAE,CAAC,IAAI+B,EAAE,wDAAwDjC,SAAS,KAAK,MAAM,MAAMiC,CAAC,CAAC,MAAM/B,EAAE,QAAQ,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,IAAIi1B,GAAG,CAAC,EAAE/0B,GAAG+0B,GAAG,CAAC,WAAW,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,SAAS,IAAIC,EAAE,CAAC,EAAE,SAASC,GAAGv1B,EAAE,EAAEE,EAAE,CAAC,GAAG+X,GAAGjY,CAAC,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,iDAAiD,EAAE,IAAIM,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,GAAGI,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,kEAAkE,EAAE,GAAGA,EAAE,SAAS,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,yEAAyE,EAAE,OAAOgsB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,IAAIs1B,GAAG,SAASC,GAAGz1B,EAAE,EAAE,EAAE,CAAC,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,gEAAgE,EAAE,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,0DAA0D,EAAE,IAAIE,EAAE,GAAGI,EAAE,GAAGC,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG0B,EAAE,GAAG,GAAGjC,EAAE,gBAAgB,WAAWE,EAAE,WAAW,OAAO,WAAW,aAAaF,aAAa,UAAUM,EAAE,WAAW,OAAO,kBAAkB,aAAaN,aAAa,iBAAiBO,EAAE,WAAW,OAAO,kBAAkB,aAAaP,aAAa,iBAAiB,EAAE,WAAWA,EAAE,YAAY,KAAK,EAAE,WAAW,OAAO,aAAa,aAAaA,aAAa,YAAYiC,EAAE,OAAQ,OAAM,IAAI,MAAM,qPAAqPjC,EAAE,YAAY,MAAM,EAAE,GAAGqmB,GAAGT,GAAGsF,EAAE,WAAW,GAAG,KAAK,CAAC,IAAIloB,EAAE,CAAC,OAAOhD,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,EAAE,OAAOkrB,EAAE,UAAUtF,GAAG5iB,EAAE,CAAC,CAAC,CAAC,GAAG,CAAC,EAAEJ,CAAC,EAAErC,EAAE,CAACP,EAAE,WAAWA,EAAE,WAAW,EAAE,CAACA,EAAE,MAAMA,EAAE,MAAM,EAAE,EAAE,GAAG,EAAE,EAAEA,EAAE,WAAW,IAAI,EAAE,aAAa,EAAE,EAAE,EAAE4C,CAAC,EAAE,aAAatC,GAAGJ,EAAE,EAAEF,EAAE,aAAa,GAAGO,GAAG0B,EAAE,CAAC,GAAGuzB,IAAI,KAAK,GAAG,OAAO,UAAU,YAAY,GAAG,OAAO,iBAAiB,aAAa,OAAO,mCAAmC,YAAYA,GAAG,IAAI,gBAAgB,EAAE,CAAC,EAAE,WAAW,IAAI,MAAO,OAAM,IAAI,MAAM,sGAAsG,OAAOA,GAAG,SAAS,cAAc,QAAQ,EAAE,WAAW,KAAK,CAAC,mBAAmB,EAAE,CAAC,EAAEA,GAAG,OAAO,MAAM,EAAEA,GAAG,OAAO,OAAO5yB,EAAE4yB,GAAG,UAAUx1B,EAAE,EAAE,EAAE,EAAE4C,CAAC,EAAE,EAAE4yB,GAAG,aAAa,EAAE,EAAE,EAAE5yB,CAAC,EAAE,IAAI,CAAC,IAAIE,EAAE,GAAG,IAAI,EAAEA,EAAE,IAAI,WAAW,CAAC,MAAM,CAAC,IAAIE,EAAE,EAAEJ,EAAEE,EAAE,IAAI,WAAWE,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,IAAI,QAAQmB,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAErB,EAAE,EAAE,EAAEqB,GAAG,EAAE,EAAE,EAAEA,EAAE,CAAC,OAAOoxB,GAAGzyB,EAAE,CAACF,EAAE,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,SAAS8yB,GAAG11B,EAAE,CAAC,OAAOA,GAAG,MAAMA,EAAE,gBAAgB,UAAU,CAAC,SAAS21B,IAAI,CAAC,OAAO,OAAO,QAAQ,aAAa,OAAO,aAAa,aAAa,OAAO,eAAe,mBAAmB,CAAC,CAAC,SAASC,GAAG51B,EAAE,CAAC,OAAOA,GAAG,MAAMA,EAAE,QAAQ,GAAGA,EAAE,SAAS,CAAC,CAAC,SAAS61B,GAAG71B,EAAE,CAAC,OAAO21B,GAAG,GAAG,EAAE31B,aAAa,cAAc41B,GAAG51B,CAAC,GAAG,CAAC01B,GAAG11B,CAAC,CAAC,CAAC,eAAeq1B,GAAGr1B,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,GAAGua,EAAE,EAAE,QAAQ,qBAAqB,GAAGob,GAAG71B,CAAC,EAAE,CAAC,IAAIM,EAAE,GAAG,CAACA,EAAE,MAAM,kBAAkBN,EAAE,CAAC,iBAAiB,MAAM,CAAC,CAAC,OAAOO,EAAN,CAASD,EAAE,IAAI,CAACA,GAAG,MAAMA,EAAE,QAAQN,EAAE,OAAOM,EAAE,SAASN,EAAE,OAAOE,EAAEI,EAAEJ,EAAEF,CAAC,MAAME,EAAEF,EAAE,OAAOy1B,GAAGv1B,EAAE,CAAC,CAAC,CAAC,eAAeo1B,GAAGt1B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,MAAM,UAAU,EAAE,GAAG,EAAEA,aAAa6pB,IAAI,CAAC,IAAI1nB,EAAEjC,EAAEA,EAAEuwB,GAAEtuB,EAAE,OAAO,EAAEA,EAAE,QAAQ,CAAC,CAAC,GAAGjC,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,wDAAwDA,EAAE,OAAO,EAAE,GAAG,CAACI,EAAEC,CAAC,EAAEL,EAAE,MAAM,MAAM,EAAE,CAAC,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEA,EAAE,MAAM,GAAG,GAAG,EAAE,GAAG,IAAI,EAAE,MAAM,IAAI,MAAM,0DAA0D,GAAG,EAAE,GAAGA,EAAE,QAAQ,WAAWA,EAAE,QAAQ,QAAQ,MAAM,IAAI,MAAM,kCAAkCA,EAAE,6CAA6C,EAAE,IAAI,EAAE,MAAMA,EAAE,KAAK,EAAE+B,EAAE/B,EAAE,QAAQ,UAAU,IAAI,EAAEgC,EAAE,IAAI,kBAAkB3B,EAAED,EAAE,CAAC,EAAE,QAAQ6B,EAAE,EAAEA,EAAE7B,EAAEC,EAAE,EAAE4B,EAAE,CAAC,IAAIS,EAAE,CAAC,EAAE,EAAE,EAAE,GAAG,EAAE,QAAQE,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIC,EAAE,EAAEZ,EAAE,EAAEW,GAAG,GAAG5C,EAAE,QAAQ,WAAW,GAAG6C,EAAE,GAAGA,EAAE,EAAE,MAAM,IAAI,MAAM,mFAAmFA,IAAI,UAAU7C,EAAE,QAAQ,UAAU6C,EAAE,GAAGA,EAAE,KAAK,MAAM,IAAI,MAAM,mFAAmFA,IAAI,EAAE,IAAI,GAAGH,EAAE,GAAGG,EAAEd,EAAEW,EAAE,GAAGG,EAAEd,EAAEW,EAAE,GAAGG,EAAEd,GAAGW,EAAEE,GAAGC,EAAEd,CAAC,CAAC,IAAIY,EAAEV,EAAE,EAAED,EAAEW,EAAE,GAAG,KAAK,MAAMD,EAAE,EAAE,EAAEV,EAAEW,EAAE,GAAG,KAAK,MAAMD,EAAE,EAAE,EAAEV,EAAEW,EAAE,GAAG,KAAK,MAAMD,EAAE,EAAE,EAAEV,EAAEW,EAAE,GAAG,KAAK,MAAMD,EAAE,EAAE,CAAC,CAAC,GAAG,GAAG,KAAK,CAAC,EAAE,MAAMrC,EAAE,EAAE,OAAOD,EAAE,IAAI6B,EAAE,EAAE,WAAW,IAAI,EAAES,EAAE,IAAI,UAAUV,EAAE3B,EAAED,CAAC,EAAE6B,EAAE,aAAaS,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO1C,IAAIF,GAAGE,EAAE,QAAQ,EAAEgC,CAAC,CAAC,IAAIkzB,GAAGjJ,EAAE,CAAC,YAAYsJ,EAAE,CAAC,EAAMK,GAAG,CAAC,EAAE11B,GAAG01B,GAAG,CAAC,mBAAmB,IAAIC,EAAE,CAAC,EAAE,SAASA,GAAG/1B,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,MAAM,OAAOM,EAAE,EAAE,MAAM,OAAO,GAAGJ,EAAE,EAAE,MAAM,IAAI,MAAM,4EAA4EA,IAAI,EAAE,GAAGI,EAAE,EAAE,MAAM,IAAI,MAAM,8EAA8EA,IAAI,EAAE,GAAG,EAAE,QAAQ,QAAQ,MAAM,IAAI,MAAM,yEAAyE,EAAE,QAAQ,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGJ,EAAE,MAAM,IAAI,MAAM,iEAAiE,EAAE,MAAMI,EAAE,UAAUJ,GAAG,EAAE,GAAGkY,GAAGpY,EAAE,KAAK,IAAI,EAAE,MAAM,IAAI,MAAM,mEAAmEA,EAAE,QAAQ,EAAE,IAAIO,EAAE,EAAE,MAAM,EAAEA,EAAEA,EAAE,OAAO,GAAG,EAAE,EAAE,QAAQsC,EAAE,EAAEA,EAAEtC,EAAE,OAAO,EAAE,EAAEsC,EAAE,GAAGtC,EAAEsC,GAAG,IAAIZ,EAAEjC,EAAE,MAAMkC,EAAE3B,EAAE,MAAM,EAAE2B,EAAE,IAAI,EAAE,IAAIC,EAAE,EAAE,QAAQU,EAAE,EAAEA,EAAE3C,EAAE,EAAE2C,EAAEV,GAAGF,EAAEY,GAAGX,EAAE,KAAKD,EAAEY,EAAE,EAAE,IAAID,EAAE,CAAC,GAAGgX,GAAG5Z,EAAE,KAAK,EAAE,IAAI6C,GAAGA,EAAEV,CAAC,EAAE,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,CAACD,EAAE,EAAEC,EAAES,CAAC,CAAC,CAAC,IAAIozB,GAAG,CAAC,EAAE51B,GAAG41B,GAAG,CAAC,gBAAgB,IAAIC,GAAG,cAAc,IAAIC,GAAG,oBAAoB,IAAIC,EAAE,CAAC,EAAE,SAASA,GAAGn2B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAE,KAAK,EAAE,EAAE,MAAM,EAAE,KAAK,GAAG,EAAEC,EAAE,EAAE,KAAK,EAAE,EAAE,KAAK,EAAE,EAAE,EAAE,6FAA6FL,EAAE,yBAAyB,EAAE,iBAAiBF,gBAAgBM,oBAAoBC,KAAK,GAAGL,EAAE,KAAKK,EAAE,MAAM,IAAI,MAAM,EAAE,kBAAkBA,KAAK,EAAE,GAAGP,EAAE,OAAOM,GAAGJ,EAAE,KAAKK,GAAG,MAAM,IAAI,MAAM,EAAE,0BAA0BD,GAAGJ,EAAE,KAAKK,IAAI,EAAE,GAAGL,EAAE,OAAOK,EAAEP,EAAE,OAAOM,EAAE,MAAM,IAAI,MAAM,EAAE,mBAAmBC,EAAEP,EAAE,OAAOM,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAEC,EAAE,EAAE,EAAE,GAAGL,EAAE,MAAM,KAAK,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,EAAE,kBAAkB,OAAOA,EAAE,MAAM,wBAAwB,OAAO,EAAE,MAAM,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,KAAKK,EAAE,EAAE,EAAE,GAAGL,EAAE,MAAM,EAAEK,KAAKP,EAAE,EAAEM,GAAG,MAAM,IAAI,MAAM,EAAE,kBAAkB,EAAEC,OAAOL,EAAE,MAAM,EAAEK,gBAAgB,EAAEA,OAAOP,EAAE,EAAEO,KAAK,CAAC,CAAC,SAAS21B,GAAGl2B,EAAE,EAAEE,EAAE,CAAC,GAAG,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,+EAA+E,EAAE,OAAO,EAAE,GAAGF,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,+EAA+EA,EAAE,OAAO,EAAE,GAAG,EAAE,QAAQ,QAAQ,MAAM,IAAI,MAAM,0DAA0D,EAAE,OAAO,EAAE,GAAGE,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,6DAA6DA,GAAG,EAAE,GAAGA,EAAE,SAAS,EAAE,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAE,OAAO,EAAE,GAAGF,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,sDAAsDA,EAAE,OAAO,CAAC,CAACm2B,GAAGj2B,EAAE,EAAEF,CAAC,CAAC,CAAC,SAASi2B,GAAGj2B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAE,MAAM,OAAOC,EAAED,EAAE,EAAE,EAAE,MAAMA,EAAE,GAAG,EAAE,EAAEJ,EAAE,OAAO,EAAE,EAAE,QAAQ2C,EAAEtC,EAAEsC,EAAE,EAAE,EAAEA,EAAE,GAAG3C,EAAE2C,GAAG,IAAIZ,EAAE1B,EAAE,EAAE,EAAEA,EAAE2B,EAAEkW,GAAG,EAAE,KAAK,EAAEnW,EAAEE,EAAE,CAAC,GAAGyX,GAAG1Z,EAAE,MAAM,EAAEK,CAAC,CAAC,EAAE,CAAC,EAAEqC,EAAEwV,GAAGlY,CAAC,EAAE,MAAM,CAAC,UAAUK,EAAE,WAAW2B,EAAE,UAAU,EAAE,QAAQC,EAAE,WAAWS,CAAC,CAAC,CAAC,IAAIwzB,GAAG,CAAC,EAAEh2B,GAAGg2B,GAAG,CAAC,kBAAkB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,WAAW,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,UAAU,IAAIC,GAAG,aAAa,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,YAAY,IAAIC,GAAG,0BAA0B,IAAIC,GAAG,eAAe,IAAIC,GAAG,sBAAsB,IAAIC,EAAE,CAAC,EAAE,IAAIC,GAAG,GAAGC,GAAG,GAAG,SAASf,GAAGr2B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,OAAO8X,EAAExX,IAAI,EAAE,OAAO,IAAI,iBAAiBA,uBAAuB,uCAAuCA,KAAK,EAAEwX,EAAExX,IAAIJ,EAAE,OAAO,IAAI,iBAAiBI,sBAAsBJ,uCAAuCI,KAAK,EAAE,QAAQC,EAAE,EAAEA,EAAED,EAAE,EAAEC,EAAEuX,EAAE,EAAEvX,GAAGL,EAAEK,IAAIP,EAAE,MAAMO,GAAG,IAAI,iBAAiBD,aAAaC,aAAaA,OAAO,EAAEA,GAAGL,EAAEK,kCAAkCA,OAAOP,EAAE,MAAMO,KAAK,CAAC,CAAC,SAASm2B,GAAG12B,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEE,EAAE,EAAE,KAAKF,EAAE,GAAGA,EAAE,GAAG,EAAE,KAAKE,CAAC,EAAEF,GAAG,EAAEE,IAAI,OAAO,CAAC,CAAC,SAASq2B,GAAGv2B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,IAAID,EAAEC,GAAG,KAAK,MAAM,EAAEA,GAAGP,EAAEO,IAAIL,EAAEK,EAAE,EAAE,OAAOD,CAAC,CAAC,SAAS42B,GAAGl3B,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,CAAC,GAAGP,CAAC,EAAE,QAAQ,EAAEO,EAAE,OAAO,EAAED,EAAE,OAAO,IAAIC,EAAE,KAAK,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEL,EAAE,IAAI,IAAI,EAAEK,EAAE,GAAG,GAAGA,EAAE,OAAO,EAAE,EAAE,CAAC,EAAEA,EAAE,IAAI,GAAG,OAAOA,CAAC,CAAC,SAAS82B,GAAGr3B,EAAE,EAAEE,EAAE,CAAC,OAAOA,GAAGF,EAAEE,EAAEA,GAAG,EAAE,EAAE,CAAC,SAASo3B,GAAGt3B,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAEM,IAAIJ,EAAE,KAAK,EAAEI,CAAC,EAAE,OAAOJ,CAAC,CAAC,SAASs2B,GAAGx2B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAE,CAAC,IAAIC,EAAEnC,EAAE,OAAO4C,EAAE,IAAI,MAAMT,CAAC,EAAEU,EAAE,IAAI,MAAMV,CAAC,EAAEW,EAAE,IAAI,MAAMX,CAAC,EAAE,GAAG,EAAE,QAAQjC,EAAE,EAAE,CAAC,IAAI6C,EAAE,EAAE,GAAGC,EAAE9C,EAAE,EAAE0C,EAAEk0B,GAAG,EAAE/zB,EAAEC,EAAE1C,EAAEN,CAAC,EAAE6C,EAAEm0B,GAAG/0B,EAAEc,EAAEC,EAAEzC,EAAEP,CAAC,EAAE8C,EAAEo0B,GAAG,EAAEn0B,EAAEC,EAAEhD,CAAC,CAAC,KAAM,SAAQ+C,EAAE,EAAEA,EAAEZ,EAAEY,IAAIH,EAAEG,GAAG8zB,GAAG,EAAEv2B,EAAE,EAAEN,EAAE+C,EAAEb,CAAC,EAAEW,EAAEE,GAAGg0B,GAAG90B,EAAE1B,EAAE,EAAEP,EAAE+C,EAAEb,CAAC,EAAEY,EAAEC,GAAGk0B,GAAG,EAAEl0B,EAAEb,CAAC,EAAE,MAAM,CAAC,MAAMU,EAAE,IAAIC,EAAE,QAAQC,CAAC,CAAC,CAAC,SAASg0B,GAAG92B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,CAAC,GAAGA,CAAC,EAAE,EAAE+2B,GAAGp3B,EAAE,CAAC,EAAE,QAAQ+B,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,GAAG,EAAE,QAAQA,CAAC,EAAE,GAAG,EAAEA,GAAG,MAAM,CAAC,IAAIC,EAAEm1B,GAAG,EAAEn3B,EAAE+B,CAAC,EAAEE,EAAE7B,EAAE4B,GAAGlC,EAAE,GAAGkC,IAAIC,EAAE,GAAG,EAAEF,GAAGE,CAAC,CAAC,OAAO,CAAC,CAAC,SAAS60B,GAAGh3B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,CAAC,GAAGA,CAAC,EAAE,EAAE+2B,GAAGp3B,EAAE,CAAC,EAAE,QAAQ+B,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,GAAG,EAAE,QAAQA,CAAC,EAAE,GAAG,EAAEA,GAAG,OAAO,qBAAqB,CAAC,IAAIC,EAAEm1B,GAAG,EAAEn3B,EAAE+B,CAAC,EAAEE,EAAE7B,EAAE4B,GAAGlC,EAAE,GAAGkC,IAAIC,EAAE,OAAO,kBAAkB,EAAEF,GAAGE,CAAC,CAAC,QAAQF,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,CAAC,IAAIC,EAAE3B,EAAE0B,GAAG,EAAEA,GAAG,IAAI,EAAEA,IAAIC,GAAG,EAAED,GAAGwV,GAAG,EAAE,EAAExV,GAAG1B,EAAE0B,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,SAASg1B,GAAGj3B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOE,EAAE,GAAG,GAAGI,GAAG,QAAQA,EAAE,GAAGA,CAAC,CAAC,SAASu2B,GAAG72B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAE,EAAEA,GAAG0B,EAAE/B,EAAEK,IAAI,GAAGP,EAAE,GAAGO,GAAG,EAAE,GAAGA,GAAG,GAAG,QAAQ0B,EAAE,EAAE,EAAE,OAAO,iBAAiB,EAAE,OAAO,kBAAkB,IAAIC,EAAE5B,EAAEC,GAAG,OAAO,EAAE,IAAI,GAAG2B,GAAG,EAAEuV,GAAG,EAAE,EAAEvV,EAAE,CAAC,EAAE,CAAC,CAAC,SAAS60B,GAAG/2B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAE,EAAEA,GAAG0B,EAAE/B,EAAEK,IAAI,GAAGP,EAAE,GAAGO,GAAG,EAAE,GAAGA,GAAG,GAAG,QAAQ0B,EAAE,EAAE,EAAE,OAAO,iBAAiB,EAAE,OAAO,kBAAkB,IAAIC,EAAE5B,EAAEC,GAAG,OAAO,EAAE,IAAI,GAAG2B,GAAGD,EAAE,EAAE,EAAEwV,GAAG,EAAE,EAAEvV,CAAC,EAAE,EAAEuV,GAAG,GAAG,EAAEvV,EAAE,CAAC,EAAE,CAAC,CAAC,SAASu0B,GAAGz2B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,EAAE,OAAO,QAAQK,EAAE,EAAEA,EAAEL,EAAE,OAAOK,IAAI,GAAGL,EAAEK,GAAG,EAAE,CAACD,EAAEC,EAAE,KAAK,CAAC,QAAQA,EAAED,EAAE,EAAEC,EAAEL,EAAE,OAAOK,IAAI,GAAG,EAAEA,GAAG,GAAGL,EAAEK,KAAKP,EAAEO,GAAG,MAAM,GAAG,MAAM,EAAE,CAAC,SAAS+1B,GAAGt2B,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAO,EAAEA,EAAEA,EAAE,OAAO,GAAG,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,OAAO,EAAEM,IAAIJ,GAAGF,EAAEM,GAAG,EAAEA,GAAG,OAAOJ,CAAC,CAAC,SAASy2B,GAAG32B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEC,EAAEP,EAAE,MAAM,OAAO,OAAO,GAAG,SAASM,EAAE,CAAC,EAAE,GAAG,IAAI,MAAMC,EAAE,CAAC,EAAE,KAAK,CAAC,CAAC,EAAE,EAAE,OAAOA,EAAED,EAAE,EAAE,OAAO,IAAI,MAAMC,EAAE,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,EAAED,EAAE,EAAE,MAAM,EAAEA,EAAE,QAAQ,GAAG,CAACwX,EAAE,IAAI,GAAG,IAAI,mDAAmD,CAAC,CAAC,EAAE,IAAI,EAAE,OAAO5X,GAAG,KAAK,EAAE,IAAI,MAAMK,CAAC,EAAE,KAAK,EAAE,EAAE,OAAOL,GAAG,SAAS,EAAE,CAACA,EAAE,GAAG,IAAI,MAAMK,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,EAAEL,EAAE,OAAOK,EAAE,EAAEL,EAAE,OAAO,IAAI,MAAMK,EAAEL,EAAE,MAAM,EAAE,KAAK,EAAE,CAAC,EAAE,EAAEA,EAAE,EAAE,EAAE,IAAI,CAAC,EAAE+B,IAAI,GAAG,EAAE,GAAG6V,EAAE,IAAI,GAAG,IAAI,qDAAqD,mCAAmC7V,IAAI,EAAEjC,EAAE,MAAMiC,GAAG3B,EAAE2B,GAAG,EAAE,CAAC3B,EAAE,CAAC,CAAC,CAAC,SAASs2B,GAAG52B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAE,CAAC,IAAIC,EAAE,GAAG7B,GAAG,MAAM6B,EAAE,IAAI,MAAM,EAAE,MAAM,EAAEA,EAAE,KAAK,CAAC,GAAGA,EAAE7B,EAAE,GAAG,OAAO,EAAE,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,4CAA4C,EAAE,IAAIsC,EAAE,GAAGC,EAAE,CAAC,KAAKV,EAAE,OAAO,wBAAwB,EAAE,MAAM,EAAE,MAAM,EAAE,IAAIjC,EAAE,MAAM,EAAE,QAAQiC,EAAE,MAAM,EAAE,UAAU5B,EAAE,QAAQ,EAAE,aAAa,EAAE,YAAY0B,EAAE,eAAeC,CAAC,EAAE,QAAQyC,EAAE,EAAEA,EAAE9B,EAAE,KAAK8B,IAAI/B,IAAI,GAAG+B,EAAE1C,KAAK,GAAGY,EAAE,0BAA0B,GAAG8B,EAAE,IAAI/B,EAAE,IAAIA,IAAIC,EAAE,cAAc,GAAGA,EAAE,KAAKA,EAAE,QAAQ,IAAIC,EAAE,CAAC,KAAK9C,EAAE,OAAO,UAAU,EAAE,QAAQ,EAAE,WAAW,GAAG,SAAS,EAAE,EAAEu3B,GAAG10B,EAAEC,CAAC,EAAE,IAAIC,EAAE,GAAGC,EAAE,GAAGkB,EAAE,GAAG,EAAE,CAAC,EAAEqD,EAAE,CAAC,EAAE,QAAQ5C,EAAE,EAAEA,EAAE3E,EAAE,OAAO,EAAE2E,EAAE,CAAC,GAAG7B,EAAE,QAAQ6B,KAAK,EAAE,MAAM,MAAM,WAAWA,qBAAqB,EAAE,IAAIC,EAAE,CAAC,EAAE9B,EAAE,eAAe,GAAG6B,GAAGE,EAAE7E,EAAE2E,GAAG,GAAGE,IAAI,GAAG,CAAC,EAAE,KAAKD,EAAE,EAAE,EAAE,EAAE,QAAQ,CAAC,IAAIE,EAAE,CAAChC,EAAE,UAAU,GAAG6B,EAAE7B,EAAE,QAAQ,GAAG6B,CAAC,EAAE6C,EAAE,CAAC1E,EAAE,QAAQ6B,GAAG,EAAE,EAAE,GAAG7B,EAAE,QAAQ6B,GAAG,EAAEE,EAAEA,EAAE,CAAC,EAAE,GAAGD,GAAG9B,EAAE,QAAQ6B,IAAI,EAAE,MAAM,MAAM,8CAA8C,EAAET,EAAEA,GAAGpB,EAAE,QAAQ6B,KAAK,EAAE,IAAIK,EAAE,CAAC,EAAElC,EAAE,UAAU,GAAG6B,GAAG7B,EAAE,QAAQ,GAAG6B,GAAG,GAAG7B,EAAE,YAAYA,EAAE,SAAS,CAAC,GAAG8B,EAAE,CAAC,IAAIO,EAAErC,EAAE,MAAM6B,GAAG,EAAEE,EAAE/B,EAAE,MAAM6B,GAAG7B,EAAE,MAAM6B,GAAG,GAAG7B,EAAE,MAAM6B,GAAGQ,EAAErC,EAAE,IAAI6B,GAAG7B,EAAE,MAAM6B,GAAG,EAAEQ,EAAE,GAAGA,GAAGN,EAAE,MAAM,MAAM,eAAe/B,EAAE,MAAM6B,mBAAmBA,kBAAkB,CAAC,MAAM7B,EAAE,MAAM6B,GAAG6yB,GAAG10B,EAAE,MAAM6B,GAAG,EAAE7B,EAAE,QAAQ6B,GAAGE,EAAEC,EAAE0C,CAAC,EAAE1E,EAAE,IAAI6B,GAAG6yB,GAAG10B,EAAE,IAAI6B,GAAG,EAAE7B,EAAE,QAAQ6B,GAAGE,EAAEC,EAAE0C,CAAC,EAAE,IAAItC,EAAEpC,EAAE,QAAQ6B,KAAK,GAAG7B,EAAE,MAAM6B,KAAK,GAAG7B,EAAE,IAAI6B,KAAKE,EAAE9B,EAAEA,GAAGmC,EAAElC,EAAEA,IAAI2B,IAAI,GAAG7B,EAAE,QAAQ6B,KAAK,GAAGO,EAAE,MAAMnC,EAAEA,GAAGD,EAAE,QAAQ6B,KAAK,GAAGK,EAAEhC,EAAEA,IAAI2B,IAAI,GAAG7B,EAAE,QAAQ6B,KAAK,GAAGK,GAAG,IAAIyC,EAAExC,EAAE,GAAG,GAAGnC,EAAE,YAAYA,EAAE,UAAU2E,EAAE3E,EAAE,IAAI6B,GAAG7B,EAAE,MAAM6B,GAAGM,EAAE,IAAIL,GAAG6C,EAAE,EAAExC,EAAE,IAAID,GAAGH,GAAG,IAAI/B,EAAE,QAAQ6B,GAAG,EAAE8C,EAAE,CAAC5C,EAAE4C,EAAE5C,EAAEI,EAAE,IAAIA,EAAE,CAAC,IAAIC,EAAEuC,IAAI,GAAGA,EAAE,GAAG3E,EAAE,QAAQ6B,GAAG,EAAEO,EAAE,EAAEA,EAAE,KAAK,MAAMuC,EAAE3E,EAAE,QAAQ6B,EAAE,GAAG8C,EAAE3E,EAAE,QAAQ6B,KAAK,EAAE,EAAE,GAAG,EAAE,KAAKO,CAAC,CAAC,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,QAAQP,EAAE,EAAEA,EAAE7B,EAAE,wBAAwB,OAAO,EAAE6B,EAAE,CAAC,IAAIC,EAAE9B,EAAE,wBAAwB6B,GAAGC,GAAG,EAAE2C,EAAE,KAAK,EAAE3C,EAAE,EAAEA,IAAIuyB,IAAI5vB,EAAE,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,iBAAiBA,EAAE,OAAO,CAAC5C,EAAEC,IAAI9B,EAAE,wBAAwB8B,KAAKuyB,EAAE,EAAE,WAAW5vB,EAAE,WAAWxE,EAAE,UAAUC,EAAE,cAAckB,EAAE,MAAMpB,EAAE,MAAM,IAAIA,EAAE,IAAI,QAAQA,EAAE,OAAO,CAAC,CAAC,SAASy0B,GAAGv3B,EAAE,EAAE,CAAC,EAAE,UAAU,EAAE,EAAE,QAAQ,EAAE,EAAE,eAAe,EAAE,IAAIE,EAAE,EAAE,EAAE,WAAWF,EAAE,OAAO,KAAK,EAAE,SAASA,EAAE,KAAK,KAAK,EAAE,MAAM,IAAI,MAAM,EAAE,IAAI,EAAE,EAAE,IAAI,IAAI,MAAM,EAAE,IAAI,EAAE,EAAE,QAAQ,IAAI,MAAM,EAAE,IAAI,EAAE,EAAE,wBAAwB,CAAC,EAAE,EAAE,8BAA8B,CAAC,EAAE,EAAE,8BAA8B,IAAI,MAAM,EAAE,IAAI,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,KAAKM,IAAI,GAAG,GAAGA,EAAEN,EAAE,aAAa,CAAC,IAAIO,EAAE,KAAK,IAAI,EAAE,MAAMP,EAAE,KAAKM,GAAG,EAAEN,EAAE,wBAAwB,EAAE,IAAI,EAAE,KAAKE,EAAEK,EAAEL,IAAI,EAAE,MAAMA,GAAG,EAAE,EAAE,IAAIA,GAAG,EAAE,EAAE,QAAQA,GAAG,EAAE,EAAE,WAAW,GAAGA,EAAE,EAAE,SAAS,GAAGA,EAAE,EAAE,wBAAwB,KAAKA,CAAC,EAAE,EAAE,8BAA8B,KAAK,EAAE,EAAE,EAAE,8BAA8BA,GAAGI,CAAC,SAAS,GAAGA,EAAEN,EAAE,YAAY,EAAE,wBAAwB,KAAKm3B,EAAE,EAAE,EAAE,8BAA8B,KAAK,EAAE,MAAM,CAAC,GAAGj3B,IAAI,EAAE,MAAM,OAAO,MAAM,MAAM,sCAAsCA,qBAAqB,EAAE,cAAc,EAAE,MAAM,SAAS,EAAEF,EAAE,OAAO,OAAO,EAAE,MAAME,GAAGF,EAAE,MAAMM,IAAIN,EAAE,KAAK,OAAO,EAAE,IAAIE,GAAGF,EAAE,IAAIM,IAAI,EAAE,QAAQJ,GAAGF,EAAE,QAAQM,GAAGN,EAAE,UAAU,GAAGM,IAAI,EAAE,WAAW,GAAGJ,GAAGF,EAAE,QAAQ,GAAGM,IAAI,EAAE,SAAS,GAAGJ,GAAGF,EAAE,eAAe,GAAGM,GAAG,EAAE,wBAAwB,KAAK82B,EAAE,EAAE,EAAE,8BAA8B,KAAK,EAAE,EAAE,EAAE,gBAAgB,GAAGl3B,IAAI,EAAE,wBAAwB,KAAKA,CAAC,EAAE,EAAE,8BAA8B,KAAKI,CAAC,GAAG,EAAE,8BAA8BJ,GAAGI,EAAEJ,GAAG,CAAC,CAAC,SAASs3B,GAAGx3B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,GAAGA,EAAE,GAAG,OAAOL,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,EAAE,GAAG,CAAC,IAAI,EAAEF,EAAE,EAAEM,EAAEN,EAAEA,EAAE,OAAO,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC,IAAIy3B,GAAE,CAAC,EAAEr3B,GAAGq3B,GAAE,CAAC,aAAa,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,cAAc,IAAIC,EAAE,CAAC,EAAE,IAAIF,GAAG,KAAK,CAAC,cAAc,CAAC,OAAO,KAAK,YAAY,SAAS,CAAC,OAAO,WAAW52B,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,CAAC,CAAC,CAAC,EAAEy3B,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,aAAa,CAAC,CAAC,CAAC,OAAO,QAAQ,CAAC,OAAOA,GAAG,UAAU,OAAOA,GAAG,SAAS,IAAIA,IAAIA,GAAG,QAAQ,CAAC,OAAO,SAAS72B,EAAE,CAAC62B,GAAG,OAAO,EAAE,aAAa72B,EAAE,WAAW,CAACA,EAAEA,EAAE,UAAU,CAAC,CAAC,EAAE,SAAS82B,GAAG53B,EAAE,CAAC8X,EAAE9X,EAAE,WAAW,KAAK,IAAI,6EAA6E,EAAE8X,EAAE,OAAO9X,EAAE,WAAW,SAAS,IAAI,sDAAsD,OAAOA,EAAE,SAAS,EAAE8X,EAAE9X,EAAE,UAAU,OAAO,EAAE,IAAI,mFAAmF,EAAE23B,GAAG,SAAS33B,CAAC,CAAC,CAAC,IAAI63B,GAAG,CAAC,EAAEz3B,GAAGy3B,GAAG,CAAC,qBAAqB,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,cAAc,IAAIC,GAAG,wBAAwB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,KAAK,IAAIC,GAAG,YAAY,IAAIC,EAAE,CAAC,EAAE,IAAIC,GAAG,KAAKX,GAAG,GAAG,SAASI,GAAGl4B,EAAE,EAAEE,EAAE,CAAC,OAAOA,GAAG,OAAOA,EAAEs4B,GAAG,GAAGE,GAAG14B,EAAE,EAAE,CAACM,EAAEC,IAAIo4B,GAAGr4B,EAAEC,EAAEL,CAAC,CAAC,CAAC,CAAC,SAASs4B,IAAI,CAAC,OAAOtN,EAAE,QAAQ,eAAe,IAAI,GAAGuN,GAAGX,EAAE,CAAC,SAASY,GAAG14B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,GAAG,IAAI6X,GAAGnY,CAAC,GAAGmY,GAAG,CAAC,KAAK7X,EAAE,IAAI6X,GAAGnY,CAAC,GAAGmY,GAAG,CAAC,IAAI7X,EAAE,IAAIA,EAAE,CAAC,IAAI,EAAEN,EAAE,YAAY,KAAKiC,EAAE,EAAE,YAAY,KAAK,GAAG,IAAIA,EAAE,MAAM,IAAI,MAAM,yCAAyC,gBAAgBA,GAAG,CAAC,CAAC,GAAG,MAAM,QAAQjC,CAAC,GAAG,MAAM,QAAQ,CAAC,EAAE,CAAC,IAAI,EAAE6rB,GAAG7rB,CAAC,EAAEiC,EAAE4pB,GAAG,CAAC,EAAE,GAAG,CAAC7T,GAAG,EAAE/V,CAAC,EAAE,MAAM,IAAI,MAAM,0CAA0C,kBAAkBA,IAAI,CAAC,CAAC,IAAI1B,EAAE4X,GAAGnY,CAAC,EAAEA,EAAEkY,GAAGlY,CAAC,EAAE,EAAEmY,GAAG,CAAC,EAAE,EAAED,GAAG,CAAC,EAAE,GAAG3X,EAAE,SAAS,EAAE,OAAO,MAAM,IAAI,MAAM,yCAAyCA,EAAE,uBAAuB,EAAE;AAAA,YACjquBA;AAAA,YACA,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,CAAC,IAAI0B,EAAE1B,EAAE,GAAG2B,EAAE,EAAE,GAAG,GAAG,CAAChC,EAAE+B,EAAEC,CAAC,EAAE,MAAM,IAAI,MAAM,yBAAyB,QAAQD,eAAe,QAAQC;AAAA,YACvI3B;AAAA,YACA,IAAI,CAAC,CAAC,OAAO,QAAQ,aAAa,OAAO,EAAE,QAAQ,CAAC,CAAC,SAAS83B,GAAGr4B,EAAE,EAAE,CAACA,EAAE,EAAE,KAAK,IAAI,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC,EAAE,OAAO,QAAQ,aAAa,OAAO,EAAE,QAAQ,CAAC,CAAC,SAASm4B,GAAGn4B,EAAE,EAAE,CAAC,IAAIE,EAAE,OAAO,GAAG,UAAU,OAAO,GAAG,UAAU,OAAO,GAAG,UAAU,CAAC,CAAC,EAAE,EAAE,OAAOoZ,GAAGtZ,CAAC,GAAGsZ,GAAGtZ,EAAE,EAAE,GAAGsZ,GAAG,CAAC,GAAGA,GAAG,EAAE,EAAE,EAAEof,GAAG14B,EAAEE,EAAE,CAACI,EAAEC,IAAID,GAAGC,CAAC,EAAEm4B,GAAG14B,EAAE,EAAE,CAACM,EAAEC,IAAIo4B,GAAGr4B,EAAEC,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS63B,GAAGp4B,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,OAAOA,EAAEs4B,GAAG,GAAG,CAACG,GAAG34B,EAAE,EAAEE,CAAC,EAAE,MAAM,IAAI,MAAM,8BAA8BF,mBAAmB,GAAG,EAAE,OAAO,QAAQ,aAAa,OAAO,EAAE,QAAQ,CAAC,CAAC,SAAS24B,GAAG34B,EAAE,EAAEE,EAAE,CAAC,MAAM,CAAC,SAASF,CAAC,GAAG,CAAC,SAAS,CAAC,EAAE,GAAG,EAAE,MAAMA,CAAC,GAAG,MAAM,CAAC,GAAG,KAAK,IAAIA,EAAE,CAAC,EAAEE,EAAE,CAAC,SAASo4B,GAAGt4B,EAAE,EAAEE,EAAE,CAAC,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAI,GAAGN,EAAEM,GAAG,GAAGN,EAAEM,GAAGJ,EAAE,MAAM,IAAI,MAAM,sBAAsBF,EAAEM,WAAW,YAAYJ,GAAG,CAAC,CAAC,SAAS+3B,GAAGj4B,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,aAAaF,CAAC,EAAEM,EAAE,IAAI,aAAa,CAAC,EAAE,GAAGJ,EAAE,SAASI,EAAE,OAAO,MAAM,IAAI,MAAM,wCAAwCA,EAAE,sBAAsBJ,EAAE,QAAQ,EAAE,QAAQK,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAI,GAAGL,EAAEK,KAAKD,EAAEC,GAAG,MAAM,IAAI,MAAM,iCAAiCA,WAAWD,EAAEC,cAAcL,EAAEK,YAAY,CAAC,CAAC,SAASy3B,GAAGh4B,EAAE,CAAC,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,IAAI,CAAC,IAAIE,EAAEF,EAAE,GAAG,MAAM,QAAQE,CAAC,EAAE83B,GAAG93B,CAAC,EAAEF,EAAE,GAAGinB,GAAG/mB,CAAC,CAAC,CAAC,OAAOF,CAAC,CAAC,SAAS+3B,GAAG/3B,EAAE,CAAC,IAAI,EAAE,SAAS,cAAc,OAAO,EAAE,MAAM,gBAAgB,IAAI,EAAE,YAAY,IAAI,EAAE,MAAM,GAAG,EAAE,KAAK,GAAG,EAAE,MAAM,SAAS,QAAQ,EAAE,MAAM,KAAK,MAAM,EAAE,MAAM,IAAI,MAAM,EAAE,QAAQ,OAAO,EAAE,YAAYA,CAAC,EAAE,IAAI,QAAQE,GAAG,CAAC,EAAE,iBAAiB,aAAaI,GAAGJ,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,eAAeq4B,GAAGv4B,EAAE,CAAC,MAAMA,EAAE,KAAK,EAAE,8BAA8BA,GAAG,MAAM,IAAI,QAAQ,GAAG,CAACA,EAAE,0BAA0B,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI44B,GAAG,QAAQ,SAASC,GAAG74B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU7P,GAAG9a,CAAC,CAAC,CAAC,IAAIu4B,EAAE3M,EAAE,CAAC,KAAK0M,EAAE,CAAC,EAAE,SAASE,GAAG/4B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,UAAU,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUlM,GAAGze,CAAC,CAAC,CAAC,IAAIy4B,GAAG7M,EAAE,CAAC,UAAU4M,EAAE,CAAC,EAAE,SAASE,GAAGj5B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,GAAG,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAEJ,EAAE,QAAQ,SAASI,EAAE,QAAQ,QAAQ,OAAO04B,GAAG94B,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,EAAE,CAAC,EAAE,OAAO4qB,EAAE,UAAU/M,GAAG5d,EAAE,CAAC,CAAC,CAAC,IAAI24B,GAAG/M,EAAE,CAAC,KAAK8M,EAAE,CAAC,EAAE,SAASE,GAAGn5B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU3J,GAAGhhB,CAAC,CAAC,CAAC,IAAI64B,EAAEjN,EAAE,CAAC,KAAKgN,EAAE,CAAC,EAAE,SAASE,GAAGr5B,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,IAAI,KAAK,EAAE,GAAG,EAAE,QAAQ,YAAY,CAAC,IAAIE,EAAE,CAAC,EAAE,CAAC,EAAE,OAAOgrB,EAAE,UAAUrO,GAAG3c,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAE,CAAC,EAAE,CAAC,EAAE,OAAOgrB,EAAE,UAAUhQ,GAAGhb,CAAC,CAAC,CAAC,CAAC,IAAIo5B,GAAGnN,EAAE,CAAC,KAAKkN,EAAE,CAAC,EAAE,SAASE,GAAGv5B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAU/P,GAAGjb,CAAC,CAAC,CAAC,IAAIs5B,GAAGrN,EAAE,CAAC,MAAMoN,EAAE,CAAC,EAAE,SAASE,GAAGz5B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAU9P,GAAGlb,CAAC,CAAC,CAAC,IAAIw5B,GAAGvN,EAAE,CAAC,OAAOsN,EAAE,CAAC,EAAE,SAASE,GAAG35B,EAAE,CAAC8X,EAAE,MAAM,QAAQ9X,CAAC,EAAE,IAAI,4DAA4D,EAAE8X,EAAE9X,EAAE,QAAQ,EAAE,IAAI,uDAAuDA,EAAE,QAAQ,EAAE,IAAI,EAAEA,EAAE,IAAI,CAACO,EAAE,IAAIyrB,EAAEzrB,EAAE,UAAU,IAAI,MAAM,CAAC,EAAEL,EAAE,EAAE,GAAG,EAAE,QAAQK,GAAG,CAAC,GAAGA,EAAE,QAAQL,EAAE,MAAM,MAAM,IAAI,MAAM,0DAA0D,CAAC,CAAC,EAAE,EAAE,QAAQK,GAAG,CAAC,GAAG,CAACyX,GAAGzX,EAAE,MAAML,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,0DAA0D,CAAC,CAAC,EAAE,IAAII,EAAE,EAAE,OAAO4qB,EAAE,UAAU5P,GAAGhb,CAAC,CAAC,CAAC,IAAIs5B,GAAGzN,EAAE,CAAC,MAAMwN,EAAE,CAAC,EAAE,SAASE,GAAG75B,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,MAAM,MAAM,CAAC,EAAE+B,EAAE,CAAC,KAAK,EAAE,SAAS7B,CAAC,EAAE,OAAOgrB,EAAE,UAAU3P,GAAGhb,EAAEwB,CAAC,CAAC,CAAC,IAAI+3B,GAAG3N,EAAE,CAAC,KAAK0N,EAAE,CAAC,EAAE,SAASE,GAAG/5B,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,MAAM,MAAM,CAAC,EAAE+B,EAAE,CAAC,KAAK,EAAE,SAAS7B,CAAC,EAAE,OAAOgrB,EAAE,UAAU1P,GAAGjb,EAAEwB,CAAC,CAAC,CAAC,IAAIi4B,GAAG7N,EAAE,CAAC,KAAK4N,EAAE,CAAC,EAAE,SAASE,GAAGj6B,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,QAAQ,CAAC,EAAEO,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUzP,GAAG,EAAElb,CAAC,CAAC,CAAC,IAAI25B,GAAG/N,EAAE,CAAC,QAAQ8N,EAAE,CAAC,EAAE,SAASE,GAAGn6B,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,QAAQ,CAAC,EAAEO,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUxP,GAAG,EAAEnb,CAAC,CAAC,CAAC,IAAI65B,GAAGjO,EAAE,CAAC,QAAQgO,EAAE,CAAC,EAAE,SAASE,GAAGr6B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUvP,GAAGzb,CAAC,CAAC,CAAC,IAAIo6B,GAAGnO,EAAE,CAAC,MAAMkO,EAAE,CAAC,EAAE,SAASE,GAAGv6B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUtP,GAAG1b,CAAC,CAAC,CAAC,IAAIs6B,GAAGrO,EAAE,CAAC,OAAOoO,EAAE,CAAC,EAAE,SAASE,GAAGz6B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUrP,GAAG3b,CAAC,CAAC,CAAC,IAAIw6B,GAAGvO,EAAE,CAAC,MAAMsO,EAAE,CAAC,EAAE,SAASE,GAAG36B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,OAAO,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUnP,GAAGxb,CAAC,CAAC,CAAC,IAAIq6B,GAAGzO,EAAE,CAAC,OAAOwO,EAAE,CAAC,EAAE,SAASE,GAAG76B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUpP,GAAG5b,CAAC,CAAC,CAAC,IAAI46B,GAAG3O,EAAE,CAAC,OAAO0O,EAAE,CAAC,EAAE,SAASE,GAAG/6B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,OAAO,EAAE,CAAC,IAAI,EAAEP,EAAE,GAAGiC,EAAE,CAAC,GAAG,EAAE,CAAC,EAAEC,EAAE84B,GAAGz6B,CAAC,EAAE,OAAO06B,GAAGj7B,EAAEiC,EAAE/B,EAAE,EAAEI,EAAE,KAAK,KAAK4B,CAAC,CAAC,CAAC,SAASg5B,GAAGl7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,eAAe,CAAC,GAAG,CAAC0B,EAAEC,CAAC,EAAEi5B,GAAG,CAAC,EAAEh5B,EAAE,GAAG,IAAI,eAAeA,EAAE,CAACF,EAAEC,EAAElC,EAAE,GAAGA,EAAE,EAAE,UAAU,IAAI,gBAAgBmC,EAAE,CAACF,EAAEC,EAAElC,EAAE,GAAGA,EAAE,EAAE,MAAO,OAAM,IAAI,MAAM,sBAAsB,GAAG,EAAE,OAAOi7B,GAAGj7B,EAAEmC,EAAEjC,EAAEI,EAAEC,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,SAAS66B,GAAGp7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,QAAQ,CAAC,GAAG,CAAC0B,EAAEC,EAAEC,CAAC,EAAEk5B,GAAG,CAAC,EAAEz4B,EAAEC,EAAE,GAAG,IAAI,QAAQA,EAAE,eAAeD,EAAE,CAACX,EAAEC,EAAEC,EAAEnC,EAAE,GAAGA,EAAE,EAAE,UAAU,IAAI,QAAQ6C,EAAE,gBAAgBD,EAAE,CAACX,EAAEC,EAAEC,EAAEnC,EAAE,GAAGA,EAAE,EAAE,MAAO,OAAM,IAAI,MAAM,sBAAsB,GAAG,EAAE,OAAOs7B,GAAGt7B,EAAE4C,EAAE1C,EAAEI,EAAEC,EAAE,GAAGsC,EAAE,CAAC,CAAC,CAAC,SAASo4B,GAAGj7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,GAAG0B,EAAE,eAAe,CAAC,GAAG,CAACC,EAAEC,EAAES,EAAEC,CAAC,EAAE,CAAC,GAAG,GAAG,GAAG,EAAE,EAAE,GAAGZ,IAAI,eAAe,CAACC,EAAEC,EAAES,EAAEC,CAAC,EAAE7C,UAAUiC,IAAI,gBAAgB,CAACC,EAAEW,EAAEV,EAAES,CAAC,EAAE5C,MAAO,OAAM,IAAI,MAAM,sBAAsBiC,GAAG,EAAE,GAAG,CAACa,EAAEC,EAAE,CAACC,CAAC,EAAE,EAAE,CAACkB,EAAE,CAAC,EAAEi3B,GAAGj7B,CAAC,EAAE,CAACqH,EAAE,CAAC,EAAE4zB,GAAG76B,CAAC,EAAEqE,EAAE42B,GAAGz4B,EAAEyE,CAAC,EAAE3C,EAAE22B,GAAGx4B,EAAE,CAAC,EAAE,CAAC,QAAQ8B,EAAE,UAAUC,EAAE,SAAS0C,CAAC,EAAEg0B,GAAGj7B,EAAE4B,EAAES,EAAEsB,EAAE,EAAES,EAAEC,EAAE,EAAE3C,CAAC,EAAE+C,EAAE,EAAEhC,EAAEH,EAAEG,EAAEyE,EAAE,OAAOxF,IAAI,gBAAgBwF,EAAE,CAACvF,EAAE8C,EAAEF,EAAE0C,CAAC,EAAEvF,IAAI,iBAAiBwF,EAAE,CAACvF,EAAE4C,EAAE0C,EAAExC,CAAC,GAAG,CAAC,UAAU9C,EAAE,WAAWD,EAAE,SAASE,EAAE,QAAQS,EAAE,WAAWC,EAAE,UAAUiC,EAAE,SAAS0C,EAAE,YAAYxC,EAAE,QAAQH,EAAE,aAAaX,EAAE,YAAY,EAAE,aAAapB,EAAE,YAAYC,EAAE,sBAAsB4B,EAAE,qBAAqBC,EAAE,eAAe2C,EAAE,cAAc,EAAE,QAAQvH,EAAE,SAASyH,EAAE,YAAY,CAAC,CAAC,CAAC,SAAS6zB,GAAGt7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,GAAG,EAAE,eAAe0B,EAAE,CAAC,GAAG,CAACC,EAAEC,EAAES,EAAEC,EAAEC,CAAC,EAAE,CAAC,GAAG,GAAG,GAAG,GAAG,EAAE,EAAE,GAAG,IAAI,eAAe,CAACZ,EAAEC,EAAES,EAAEC,EAAEC,CAAC,EAAE9C,UAAU,IAAI,gBAAgB,CAACkC,EAAEY,EAAEX,EAAES,EAAEC,CAAC,EAAE7C,MAAO,OAAM,IAAI,MAAM,sBAAsB,GAAG,EAAE,GAAG,CAAC+C,EAAEC,EAAEkB,EAAE,CAAC,CAAC,EAAE,EAAE,CAACqD,EAAE,EAAE5C,CAAC,EAAE02B,GAAGn7B,CAAC,EAAE,CAAC0E,EAAEC,EAAEC,CAAC,EAAEu2B,GAAG/6B,CAAC,EAAEkH,EAAE+zB,GAAGx4B,EAAE6B,CAAC,EAAEI,EAAEu2B,GAAGv4B,EAAE6B,CAAC,EAAE4C,EAAE8zB,GAAGr3B,EAAEY,CAAC,EAAE,CAAC,QAAQG,EAAE,SAASC,EAAE,UAAUC,EAAE,SAASC,CAAC,EAAEq2B,GAAGl7B,EAAE4B,EAAES,EAAEC,EAAE0E,EAAE,EAAE5C,EAAE6C,EAAExC,EAAEyC,EAAExF,CAAC,EAAEoD,EAAE,EAAE,EAAEvC,EAAE,EAAEwC,EAAE,OAAO,IAAI,gBAAgBA,EAAE,CAACpD,EAAEmD,EAAEH,EAAEC,EAAEC,CAAC,EAAE,IAAI,iBAAiBE,EAAE,CAACpD,EAAEgD,EAAEC,EAAEC,EAAEC,CAAC,GAAG,CAAC,UAAUnD,EAAE,WAAW,EAAE,QAAQC,EAAE,SAASS,EAAE,QAAQC,EAAE,WAAWC,EAAE,SAASoC,EAAE,UAAUC,EAAE,SAASC,EAAE,YAAYC,EAAE,QAAQJ,EAAE,YAAYsC,EAAE,aAAa,EAAE,YAAY5C,EAAE,YAAY5B,EAAE,aAAaC,EAAE,YAAYkB,EAAE,qBAAqBsD,EAAE,sBAAsBxC,EAAE,qBAAqByC,EAAE,cAAc7C,EAAE,eAAeC,EAAE,cAAcC,EAAE,QAAQ9E,EAAE,SAASsF,EAAE,YAAY,CAAC,CAAC,CAAC,SAASo2B,GAAG17B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAACD,GAAG,OAAOA,EAAEq7B,GAAG37B,EAAE,EAAEE,CAAC,GAAG,IAAI,EAAEF,EAAE,GAAG,EAAEA,EAAE,GAAGiC,EAAE25B,IAAI,EAAE,EAAE,EAAEt7B,GAAGJ,EAAE,EAAEK,CAAC,EAAE2B,EAAE05B,IAAI,EAAE,EAAE,EAAEt7B,GAAGJ,EAAE,EAAEK,CAAC,EAAE,MAAM,CAAC0B,EAAEC,CAAC,CAAC,CAAC,SAAS25B,GAAG77B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAACA,GAAG,OAAOA,EAAEo7B,GAAG37B,EAAE,EAAEM,CAAC,GAAG,IAAI,EAAEN,EAAE,GAAGiC,EAAEjC,EAAE,GAAGkC,EAAElC,EAAE,GAAGmC,EAAEy5B,IAAI,EAAE,EAAE,EAAEr7B,GAAGD,EAAE,EAAE,CAAC,EAAEsC,EAAEg5B,IAAI35B,EAAE,EAAE,EAAE1B,GAAGD,EAAE,EAAE,CAAC,EAAEuC,EAAE+4B,IAAI15B,EAAE,EAAE,EAAE3B,GAAGD,EAAE,EAAE,CAAC,EAAE,MAAM,CAAC6B,EAAES,EAAEC,EAAE3C,CAAC,CAAC,CAAC,SAASy7B,GAAG37B,EAAE,EAAEE,EAAEI,EAAE,EAAE,CAAC,IAAIC,EAAEg7B,GAAG,EAAEj7B,CAAC,EAAE,OAAO,KAAK,OAAON,EAAE,IAAIE,EAAE,GAAGA,EAAEK,GAAG,CAAC,CAAC,CAAC,SAAS46B,GAAGn7B,EAAE,CAAC,OAAO,OAAOA,GAAG,SAAS,CAACA,EAAEA,EAAEA,CAAC,EAAEA,EAAE,SAAS,EAAE,CAACA,EAAE,GAAGA,EAAE,GAAG,CAAC,EAAEA,CAAC,CAAC,SAASq7B,GAAGr7B,EAAE,CAAC,OAAO,OAAOA,GAAG,SAAS,CAACA,EAAEA,EAAEA,CAAC,EAAEA,CAAC,CAAC,SAASu7B,GAAGv7B,EAAE,EAAE,CAAC,OAAO,GAAG,EAAEA,EAAEA,GAAGA,EAAE,IAAI,EAAE,EAAE,CAAC,SAASw7B,GAAGx7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAE,CAAC,IAAIC,EAAES,EAAEC,EAAE,GAAG,OAAO7C,GAAG,SAAS,CAACmC,EAAE,CAAC,IAAInC,EAAE,OAAOA,EAAE,KAAKA,EAAE,MAAMA,EAAE,KAAKA,IAAI,EAAE,QAAQ,QAAQ,EAAE,IAAI+C,EAAE24B,GAAG,CAAC,EAAEx7B,CAAC,EAAE,EAAEI,EAAEN,EAAEiC,CAAC,EAAEW,EAAEG,EAAE,GAAGF,EAAEE,EAAE,EAAE,SAAS/C,IAAI,OAAO,CAAC4C,EAAE,KAAK,KAAK,EAAEtC,CAAC,EAAEuC,EAAE,KAAK,KAAK3C,EAAEK,CAAC,EAAE,IAAIuC,EAAE,KAAK,IAAI,GAAGF,EAAE,GAAGtC,EAAE,EAAE,CAAC,EAAEyC,EAAE,KAAK,IAAI,GAAGF,EAAE,GAAGtC,EAAE,EAAEL,CAAC,EAAE8C,EAAE,KAAK,MAAMF,EAAE,CAAC,EAAEoB,EAAEpB,EAAEE,EAAE,EAAE,KAAK,MAAMD,EAAE,CAAC,EAAEwE,EAAExE,EAAE,EAAEZ,EAAE,CAAC,IAAIa,EAAE,OAAOkB,EAAE,KAAK,EAAE,MAAMqD,EAAE,KAAK,MAAM,CAAC,SAASvH,IAAI,QAAQmC,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,OAAO,EAAES,EAAE,KAAK,MAAM,EAAE,EAAE,GAAGtC,CAAC,EAAEuC,EAAE,KAAK,MAAM3C,EAAE,EAAE,GAAGK,CAAC,UAAU,OAAOP,GAAG,SAAS,CAAC,IAAI8C,EAAEZ,IAAI,eAAelC,EAAE,GAAG,GAAGA,EAAE,GAAG,GAAG+C,EAAEb,IAAI,eAAelC,EAAE,GAAG,GAAGA,EAAE,GAAG,GAAGgD,EAAEd,IAAI,eAAelC,EAAE,GAAG,GAAGA,EAAE,GAAG,GAAGkE,EAAEhC,IAAI,eAAelC,EAAE,GAAG,GAAGA,EAAE,GAAG,GAAGmC,EAAE,CAAC,IAAIW,EAAE,OAAOC,EAAE,KAAKC,EAAE,MAAMkB,EAAE,KAAKpB,IAAI,GAAGC,IAAI,GAAGC,IAAI,GAAGkB,IAAI,EAAE,QAAQ,UAAU,EAAEtB,EAAEg5B,IAAI,EAAE,EAAE94B,EAAEC,GAAGzC,EAAE,EAAE2B,CAAC,EAAEY,EAAE+4B,IAAI17B,EAAE,EAAE8C,EAAEkB,GAAG3D,EAAE,EAAE0B,CAAC,CAAC,KAAM,OAAM,MAAM,8BAA8BjC,GAAG,EAAE,MAAM,CAAC,QAAQmC,EAAE,UAAUS,EAAE,SAASC,CAAC,CAAC,CAAC,SAAS44B,GAAGz7B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAE,CAAC,IAAIC,EAAEC,EAAEC,EAAEC,EAAE,GAAG,OAAOhD,GAAG,SAAS,CAAC6C,EAAE,CAAC,IAAI7C,EAAE,OAAOA,EAAE,KAAKA,EAAE,MAAMA,EAAE,MAAMA,EAAE,KAAKA,EAAE,KAAKA,IAAI,EAAE,QAAQ,QAAQ,EAAE,IAAImE,EAAE03B,GAAG,CAAC,EAAE37B,EAAEI,EAAE,CAAC,EAAE2B,EAAE,EAAE1B,EAAEP,EAAE4C,CAAC,EAAEE,EAAEqB,EAAE,GAAGpB,EAAEoB,EAAE,GAAGnB,EAAEmB,EAAE,EAAE,SAASnE,IAAI,OAAO,CAAC8C,EAAE,KAAK,KAAK,EAAEvC,CAAC,EAAEwC,EAAE,KAAK,KAAK7C,EAAE,CAAC,EAAE8C,EAAE,KAAK,KAAK1C,EAAE,CAAC,EAAE,IAAI4D,GAAGpB,EAAE,GAAGvC,EAAE0B,EAAE,EAAE,GAAGc,EAAE,GAAG,EAAEb,EAAEhC,EAAEqH,GAAGvE,EAAE,GAAG,EAAEb,EAAE7B,EAAE,EAAE,KAAK,MAAM4D,EAAE,CAAC,EAAES,EAAET,EAAE,EAAEU,EAAE,KAAK,MAAM,EAAE,CAAC,EAAEC,EAAE,EAAED,EAAEE,EAAE,KAAK,MAAMyC,EAAE,CAAC,EAAEC,EAAED,EAAEzC,EAAEjC,EAAE,CAAC,IAAI+B,EAAE,OAAOC,EAAE,KAAKC,EAAE,MAAM0C,EAAE,MAAM,EAAE,KAAK7C,EAAE,KAAK,MAAM,CAAC,SAAS3E,IAAI,QAAQ6C,EAAE,CAAC,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK,OAAO,EAAEC,EAAE,KAAK,MAAM,EAAEb,EAAE,GAAG1B,CAAC,EAAEwC,EAAE,KAAK,MAAM7C,EAAEgC,EAAE,GAAG,CAAC,EAAEc,EAAE,KAAK,MAAM1C,EAAE6B,EAAE,GAAG,CAAC,MAAO,OAAM,MAAM,8BAA8BnC,GAAG,EAAE,MAAM,CAAC,QAAQ6C,EAAE,SAASC,EAAE,UAAUC,EAAE,SAASC,CAAC,CAAC,CAAC,SAAS44B,GAAG57B,EAAE,EAAE,CAAC,GAAG,CAAC,EAAE,OAAO,KAAK,MAAMA,CAAC,EAAE,OAAO,EAAE,CAAC,IAAI,QAAQ,OAAO,KAAK,MAAMA,CAAC,EAAE,IAAI,OAAO,OAAO,KAAK,KAAKA,CAAC,EAAE,IAAI,QAAQ,OAAO,KAAK,MAAMA,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,wBAAwB,GAAG,CAAC,CAAC,CAAC,SAAS87B,GAAG97B,EAAE,CAAC,GAAG,CAAC,EAAEE,EAAEI,CAAC,EAAE66B,GAAGn7B,CAAC,EAAE,OAAO,IAAI,GAAGE,IAAI,GAAGI,IAAI,CAAC,CAAC,SAASy7B,GAAG/7B,EAAE,EAAE,CAAC,OAAO87B,GAAG97B,CAAC,GAAG87B,GAAG,CAAC,CAAC,CAAC,SAASd,GAAGh7B,EAAE,CAAC,GAAGA,IAAI,OAAO,MAAM,eAAe,GAAGA,IAAI,OAAO,MAAM,gBAAgB,MAAM,IAAI,MAAM,sBAAsBA,GAAG,CAAC,CAAC,SAASg8B,GAAGh8B,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,KAAK,CAAC,GAAG,OAAO,GAAG,SAAS,MAAM,MAAM,YAAYF,wDAAwDE,iBAAiB,IAAI,EAAE,GAAG,OAAO,GAAG,SAAS4X,EAAEQ,GAAG,CAAC,EAAE,IAAI,YAAYtY,wDAAwDE,iBAAiB,IAAI,UAAU,OAAO,GAAG,SAAS,EAAE,QAAQI,GAAG,CAACA,EAAE,QAAQC,GAAG,CAACuX,EAAEQ,GAAG/X,CAAC,EAAE,IAAI,YAAYP,wDAAwDE,iBAAiBK,IAAI,CAAC,CAAC,CAAC,CAAC,MAAO,OAAM,MAAM,YAAYP,iCAAiC,GAAG,CAAC,CAAC,CAAC,SAASi8B,GAAGj8B,EAAE,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,UAAU,mBAAmB,CAAC,EAAEO,EAAE,CAAC,MAAM,CAAC,EAAE,OAAO2qB,EAAE,UAAUtI,GAAG,EAAEriB,CAAC,CAAC,CAAC,IAAI27B,EAAE/P,EAAE,CAAC,SAAS8P,EAAE,CAAC,EAAE,SAASE,GAAGn8B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,UAAU,SAAS,EAAE,EAAE,EAAE8X,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,wEAAwEA,oBAAoB,IAAI,EAAE,IAAI+B,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE7V,EAAE,OAAO,EAAE,IAAI,mDAAmDA,EAAE,OAAO,EAAE+5B,GAAG,UAAU17B,EAAEC,CAAC,EAAE,IAAI4B,EAAE,CAAC,EAAEF,CAAC,EAAEW,EAAE,CAAC,WAAW,EAAE,QAAQ1C,EAAE,IAAII,EAAE,gBAAgBC,CAAC,EAAEsC,EAAEqoB,EAAE,UAAUlP,GAAG7Z,EAAES,CAAC,EAAE,OAAOC,EAAE4tB,GAAE5tB,EAAE,EAAE,KAAK,EAAEX,EAAEg6B,EAAEr5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIu5B,GAAGjQ,EAAE,CAAC,SAASgQ,EAAE,CAAC,EAAE,SAASE,GAAGr8B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,QAAQ,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,YAAY,SAAS,EAAEiC,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE7V,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,OAAO,EAAE6V,EAAE,IAAI,QAAQ,IAAI,gFAAgF,GAAG,EAAEkkB,GAAG,YAAY17B,EAAEC,CAAC,EAAE,IAAI4B,EAAE,CAAC,EAAEF,CAAC,EAAEW,EAAE,CAAC,WAAW,EAAE,QAAQ1C,EAAE,IAAII,EAAE,gBAAgBC,EAAE,WAAW,CAAC,EAAEsC,EAAEqoB,EAAE,UAAUhP,GAAG/Z,EAAES,CAAC,EAAE,OAAOC,EAAE4tB,GAAE5tB,EAAEZ,EAAE,KAAK,EAAEC,EAAEg6B,EAAEr5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIy5B,GAAGnQ,EAAE,CAAC,WAAWkQ,EAAE,CAAC,EAAE,SAASE,GAAGv8B,EAAE,EAAE,EAAE,CAAC8X,EAAE9X,EAAE,QAAQ,EAAE,IAAI,oCAAoC,EAAE,IAAIE,EAAE+rB,GAAGjsB,EAAE,UAAU,SAAS,mBAAmB,EAAE,GAAGE,EAAE,GAAG,QAAQ,aAAaA,EAAE,QAAQ,GAAG,CAAC,GAAG,EAAE,QAAQ,YAAY,MAAM,IAAI,MAAM;AAAA,uBAChuU,EAAE,SAAS,CAAC,CAAC,EAAEA,EAAE,SAAS,EAAE,OAAOywB,GAAGzwB,EAAE,EAAE,EAAE,IAAII,EAAEJ,EAAEK,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUpO,GAAGxc,EAAEC,CAAC,CAAC,CAAC,IAAIi8B,GAAGrQ,EAAE,CAAC,QAAQoQ,EAAE,CAAC,EAAE,SAASE,GAAGz8B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUrH,GAAG3jB,CAAC,CAAC,CAAC,IAAIw8B,GAAGvQ,EAAE,CAAC,SAASsQ,EAAE,CAAC,EAAE,SAASE,GAAG38B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,QAAQ,mBAAmB,EAAE,GAAGM,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,gCAAgC,EAAE,IAAIC,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,KAAKJ,CAAC,EAAE,OAAOgrB,EAAE,UAAUzH,GAAGljB,EAAE,CAAC,CAAC,CAAC,IAAIq8B,GAAGzQ,EAAE,CAAC,OAAOwQ,EAAE,CAAC,EAAE,SAASE,GAAG78B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUjG,GAAG/kB,CAAC,CAAC,CAAC,IAAI48B,GAAG3Q,EAAE,CAAC,MAAM0Q,EAAE,CAAC,EAAE,SAASE,GAAG/8B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,aAAa,eAAe,EAAEiC,EAAE+pB,EAAE,EAAE,aAAa,eAAe,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,WAAW,eAAe,EAAEiC,EAAE6pB,EAAE1rB,EAAE,OAAO,eAAe,EAAEsC,EAAEopB,EAAEzrB,EAAE,IAAI,eAAe,EAAEsC,EAAEmpB,EAAE,EAAE,IAAI,eAAe,EAAElpB,EAAE05B,GAAG,CAACr6B,EAAEU,CAAC,EAAE,CAAC,EAAEE,EAAE+vB,GAAGhwB,EAAEb,CAAC,EAAEe,EAAE81B,EAAE/1B,EAAEb,CAAC,EAAEgC,EAAElB,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,GAAG,EAAEuE,EAAE,CAACrD,EAAE,CAAC,EAAE,EAAE04B,GAAG55B,EAAE,CAAC,EAAE,CAAC,EAAEuE,CAAC,EAAE5C,EAAEi4B,GAAG55B,EAAE,CAAC,EAAE,CAAC,EAAEuE,CAAC,EAAE3C,EAAEg4B,GAAG55B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEuE,CAAC,EAAE1C,EAAE+3B,GAAG55B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEuE,CAAC,EAAEzC,EAAEg0B,EAAEM,EAAEsD,GAAG,CAAC,EAAEI,GAAGn4B,CAAC,CAAC,EAAEy0B,EAAEx2B,EAAE85B,GAAG5D,EAAE,EAAEl0B,CAAC,CAAC,CAAC,CAAC,EAAE4C,EAAE4xB,EAAE0D,GAAGh4B,CAAC,EAAE43B,GAAG73B,CAAC,CAAC,EAAE,MAAM,CAACC,EAAE0C,CAAC,CAAC,CAAC,IAAIw1B,GAAG7Q,EAAE,CAAC,eAAe4Q,EAAE,CAAC,EAAE,SAASE,GAAGj9B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,gBAAgB,EAAEO,EAAE,EAAE,OAAO,CAAC0B,EAAEC,IAAID,EAAEC,CAAC,EAAE4V,EAAExX,EAAE,MAAM,EAAE,EAAE,OAAO,IAAI,iBAAiBA,EAAE,+CAA+C,EAAE,QAAQ,EAAEwX,EAAE5X,EAAE,SAAS,EAAE,OAAO,IAAI,mBAAmBA,EAAE,oDAAoD,EAAE,QAAQ,EAAE4X,EAAExX,EAAE,MAAM,GAAGC,IAAI,EAAE,IAAI,yBAAyBD,EAAE,MAAM,wEAAwE,EAAE,KAAK,KAAK,SAASC,GAAG,EAAE,IAAI,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,MAAMJ,CAAC,EAAE,OAAOgrB,EAAE,UAAU7O,GAAG,EAAE,CAAC,CAAC,CAAC,IAAI6gB,GAAG/Q,EAAE,CAAC,gBAAgB8Q,EAAE,CAAC,EAAE,SAASE,GAAGn9B,EAAE,CAAC,IAAI,EAAE,OAAOA,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,EAAEk8B,EAAEl8B,EAAE,CAAC,EAAE,EAAE,EAAEA,EAAE,IAAI,CAAC,EAAEA,EAAE,OAAO,EAAE,EAAEk8B,EAAEl8B,EAAE,CAAC,EAAE,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,EAAE,OAAO,EAAE,EAAEk8B,EAAEl8B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,EAAEA,EAAE,CAAC,CAAC,SAASo9B,GAAGp9B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,GAAG,OAAO,EAAE,MAAM,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,WAAW,EAAEiC,EAAE+pB,EAAE,EAAE,OAAO,WAAW,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,WAAW,WAAW,EAAEiC,EAAE5B,GAAG,OAAO4B,EAAE6pB,EAAEzrB,EAAE,QAAQ,WAAW,GAAG,IAAIqC,EAAEtC,GAAG,OAAOsC,EAAEopB,EAAE1rB,EAAE,SAAS,WAAW,GAAGwX,EAAE7V,EAAE,OAAOC,EAAE,KAAK,IAAI,8EAA8E,EAAE4V,EAAElV,GAAG,MAAMX,EAAE,OAAOW,EAAE,KAAK,IAAI,4EAA4E,EAAEkV,EAAE3V,GAAG,MAAMF,EAAE,OAAOE,EAAE,KAAK,IAAI,2EAA2E,EAAE,IAAIW,EAAE,CAAC,EAAEq6B,GAAG,CAAC,EAAE,MAAMh7B,EAAE,OAAOS,EAAE,KAAKX,EAAE,SAASC,CAAC,EAAEa,EAAE,CAAC,gBAAgB,CAAC,EAAEC,EAAEkoB,EAAE,UAAUjM,GAAGnc,EAAEC,CAAC,EAAE,OAAOm5B,EAAEl5B,EAAE,EAAE,KAAK,CAAC,CAAC,IAAIq6B,GAAGlR,EAAE,CAAC,WAAWiR,EAAE,CAAC,EAAE,SAASE,GAAGt9B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,WAAW,EAAEiC,EAAE+pB,EAAE,EAAE,OAAO,WAAW,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,WAAW,WAAW,EAAEiC,EAAE5B,GAAG,OAAO4B,EAAE6pB,EAAEzrB,EAAE,QAAQ,WAAW,GAAG,IAAIqC,EAAE,OAAOtC,GAAG,OAAOsC,EAAEopB,EAAE1rB,EAAE,SAAS,WAAW,GAAGwX,EAAE,EAAE,OAAO,EAAE,IAAI,uDAAuD,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,oEAAoEA,EAAE,OAAO,EAAE6V,EAAE5V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,wEAAwEA,EAAE,OAAO,EAAEC,GAAG,MAAM2V,EAAE3V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,qEAAqEA,EAAE,OAAO,EAAES,GAAG,MAAMkV,EAAElV,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,sEAAsEA,EAAE,OAAO,EAAEy6B,GAAG,EAAEp7B,EAAEC,EAAEU,EAAET,EAAE,CAAC,CAAC,CAAC,IAAIo7B,GAAGpR,EAAE,CAAC,aAAamR,EAAE,CAAC,EAAE,SAASE,GAAGx9B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,WAAW,EAAEiC,EAAE+pB,EAAE,EAAE,OAAO,WAAW,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,WAAW,WAAW,EAAEiC,EAAE5B,GAAG,OAAO4B,EAAE6pB,EAAEzrB,EAAE,QAAQ,WAAW,GAAG,IAAIqC,EAAE,OAAOtC,GAAG,OAAOsC,EAAEopB,EAAE1rB,EAAE,SAAS,WAAW,GAAGwX,EAAE,EAAE,OAAO,EAAE,IAAI,uDAAuD,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,oEAAoEA,EAAE,OAAO,EAAE6V,EAAE5V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,wEAAwEA,EAAE,OAAO,EAAEC,GAAG,MAAM2V,EAAE3V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,qEAAqEA,EAAE,OAAO,EAAES,GAAG,MAAMkV,EAAElV,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,sEAAsEA,EAAE,OAAO,EAAEy6B,GAAG,EAAEp7B,EAAEC,EAAEU,EAAET,EAAE,CAAC,CAAC,CAAC,IAAIs7B,GAAGtR,EAAE,CAAC,aAAaqR,EAAE,CAAC,EAAE,SAASE,GAAG19B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,WAAW,EAAEiC,EAAE+pB,EAAE,EAAE,OAAO,WAAW,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,WAAW,WAAW,EAAEiC,EAAE5B,GAAG,OAAO4B,EAAE6pB,EAAEzrB,EAAE,QAAQ,WAAW,GAAG,IAAIqC,EAAE,OAAOtC,GAAG,OAAOsC,EAAEopB,EAAE1rB,EAAE,SAAS,WAAW,GAAGwX,EAAE,EAAE,OAAO,EAAE,IAAI,uDAAuD,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,oEAAoEA,EAAE,OAAO,EAAE6V,EAAE5V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,wEAAwEA,EAAE,OAAO,EAAEC,GAAG,MAAM2V,EAAE3V,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,qEAAqEA,EAAE,OAAO,EAAES,GAAG,MAAMkV,EAAElV,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,sEAAsEA,EAAE,OAAO,EAAEy6B,GAAG,EAAEp7B,EAAEC,EAAEU,EAAET,EAAE,CAAC,CAAC,CAAC,IAAIw7B,GAAGxR,EAAE,CAAC,aAAauR,EAAE,CAAC,EAAE,SAASE,GAAG59B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,UAAU,EAAEO,EAAEyrB,EAAE,EAAE,UAAU,UAAU,EAAElU,EAAExX,EAAE,QAAQ,QAAQ,IAAI,yDAAyDA,EAAE,OAAO,EAAEwX,EAAE5X,GAAG,EAAE,IAAI,sCAAsCA,IAAI,EAAE4X,EAAEvX,EAAE,OAAOD,EAAE,MAAMC,EAAE,OAAO,EAAE,IAAI,gGAAgGD,EAAE,yBAAyBC,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAED,EAAE,QAAQC,CAAC,EAAE,EAAE,CAAC,KAAKL,CAAC,EAAE,OAAOgrB,EAAE,UAAU5O,GAAG,EAAE,CAAC,CAAC,CAAC,IAAIuhB,GAAG1R,EAAE,CAAC,UAAUyR,EAAE,CAAC,EAAE,SAASE,GAAG99B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,KAAK,gBAAgB,OAAO,EAAEM,EAAE0rB,EAAE,EAAE,KAAK,gBAAgB,OAAO,EAAE,GAAG9rB,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,oEAAoEA,EAAE,MAAM,EAAE,GAAGI,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,qEAAqEA,EAAE,MAAM,EAAE,IAAIC,EAAE,CAAC,GAAGL,EAAE,GAAGI,CAAC,EAAE,OAAO4qB,EAAE,UAAU1O,GAAGjc,CAAC,CAAC,CAAC,IAAIw9B,GAAG5R,EAAE,CAAC,eAAe2R,EAAE,CAAC,EAAE,SAASE,GAAGh+B,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,cAAc,GAAG,EAAEM,EAAEJ,EAAE,MAAM,GAAG,EAAE,KAAK,GAAG,EAAE,EAAE,IAAI,EAAE,IAAI,CAAC,EAAE,MAAM,IAAI,MAAM,2CAA2C,KAAK,EAAE,GAAG,EAAE,OAAOA,EAAE,KAAK,MAAM,IAAI,MAAM,+BAA+B,EAAE,uBAAuBA,EAAE,OAAO,EAAE,GAAG,EAAE,OAAOA,EAAE,KAAK,CAAC,IAAI,EAAEA,EAAE,MAAM,MAAM,EAAE,KAAK,EAAE,OAAO,EAAE,QAAQ,EAAE,QAAQ,CAAC,EAAEA,EAAEg8B,EAAEh8B,EAAE,CAAC,CAAC,CAAC,IAAIK,EAAEL,EAAE,MAAM,EAAE,MAAM,KAAK,CAAC,EAAE,QAAQ,EAAE,EAAE,OAAO,EAAE,GAAG,EAAE,IAAI,GAAGK,EAAE,KAAK,EAAE,GAAG,EAAE,GAAG,UAAUL,EAAE,MAAM,KAAK,EAAE,MAAM,IAAI,MAAM,mBAAmBI,8BAA8B,KAAK,EAAE,GAAG,EAAE,IAAI,CAAC,EAAEsC,IAAI,EAAE,EAAEA,EAAE,EAAE,EAAE,OAAO,GAAG,GAAG,CAAC,EAAE,SAAS,EAAE,OAAO+tB,GAAGzwB,CAAC,EAAE,IAAI+B,EAAE,CAAC,EAAE/B,CAAC,EAAEgC,EAAE,CAAC,KAAK,CAAC,EAAE,OAAOgpB,EAAE,UAAUhG,GAAGjjB,EAAEC,CAAC,CAAC,CAAC,IAAI+7B,GAAG9R,EAAE,CAAC,aAAa6R,EAAE,CAAC,EAAE,SAASE,GAAGl+B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUxO,GAAGxc,CAAC,CAAC,CAAC,IAAIi+B,GAAGhS,EAAE,CAAC,MAAM+R,EAAE,CAAC,EAAE,SAASE,GAAGp+B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,MAAMN,EAAE,MAAM,EAAE,MAAME,CAAC,EAAE,OAAOgrB,EAAE,UAAUrM,GAAG,CAAC,EAAEve,CAAC,CAAC,CAAC,SAAS+9B,GAAGr+B,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,aAAa,EAAE,GAAG8X,EAAE,GAAG5X,EAAE,IAAI,uBAAuB,yCAAyCA,KAAK,EAAE,IAAIA,EAAE,OAAOk+B,GAAG99B,EAAE,MAAM,EAAEA,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,aAAaJ,CAAC,EAAE,OAAOgrB,EAAE,UAAUvO,GAAGpc,EAAE,CAAC,CAAC,CAAC,IAAI+9B,GAAGnS,EAAE,CAAC,aAAakS,EAAE,CAAC,EAAE,SAASE,GAAGv+B,EAAE,CAAC,OAAOw8B,GAAGx8B,EAAE,CAAC,CAAC,CAAC,IAAIw+B,GAAGrS,EAAE,CAAC,UAAUoS,EAAE,CAAC,EAAE,SAASE,GAAGz+B,EAAE,EAAE,CAAC,OAAOw8B,GAAGx8B,EAAE,CAAC,CAAC,CAAC,IAAI0+B,GAAGvS,EAAE,CAAC,UAAUsS,EAAE,CAAC,EAAE,SAASE,GAAG3+B,EAAE,EAAE,CAAC,OAAOw8B,GAAGx8B,EAAE,CAAC,CAAC,CAAC,IAAI4+B,GAAGzS,EAAE,CAAC,UAAUwS,EAAE,CAAC,EAAE,SAASE,GAAG7+B,EAAE,EAAE,CAAC,OAAOw8B,GAAGx8B,EAAE,CAAC,CAAC,CAAC,IAAI8+B,GAAG3S,EAAE,CAAC,UAAU0S,EAAE,CAAC,EAAE,SAASE,GAAG/+B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,OAAO,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,IAAI,SAAS,SAAS,EAAEkC,EAAE8pB,EAAE,EAAE,SAAS,SAAS,SAAS,EAAE7pB,EAAEF,EAAEW,EAAE,GAAGX,EAAE,OAAO,IAAIW,EAAE,GAAGT,EAAE+5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG6V,EAAE3V,EAAE,OAAO,EAAE,IAAI,uDAAuDA,EAAE,OAAO,EAAE2V,EAAE5V,EAAE,OAAO,EAAE,IAAI,wDAAwDA,EAAE,OAAO,EAAE85B,GAAG,SAAS17B,EAAE,CAAC,EAAE,IAAIuC,EAAEtC,IAAI,OAAO4B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG2V,EAAEjV,IAAIX,EAAE,MAAM,GAAG,IAAI,oCAAoCW,wCAAwCX,EAAE,MAAM,KAAK,EAAE4V,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,uEAAuEA,oBAAoB,IAAI,EAAE,IAAI4C,EAAE,CAAC,EAAEX,EAAE,OAAOD,CAAC,EAAEa,EAAE,CAAC,QAAQ7C,EAAE,IAAII,EAAE,WAAWC,EAAE,UAAU,EAAE,gBAAgB,CAAC,EAAEyC,EAAEkoB,EAAE,UAAUnO,GAAGja,EAAEC,CAAC,EAAE,OAAOH,EAAEs5B,EAAEl5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIg8B,GAAG7S,EAAE,CAAC,QAAQ4S,EAAE,CAAC,EAAE,SAASE,GAAGj/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,MAAM,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,IAAI,QAAQ,EAAEkC,EAAE8pB,EAAE,EAAE,SAAS,QAAQ,EAAE7pB,EAAEF,EAAEW,EAAE,GAAGX,EAAE,OAAO,IAAIW,EAAE,GAAGT,EAAE+5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG6V,EAAE3V,EAAE,OAAO,EAAE,IAAI,uDAAuDA,EAAE,OAAO,EAAE2V,EAAE5V,EAAE,OAAO,EAAE,IAAI,wDAAwDA,EAAE,OAAO,EAAE85B,GAAG,SAAS17B,EAAE,CAAC,EAAEwX,EAAE3V,EAAE,MAAM,KAAKD,EAAE,MAAM,GAAG,IAAI,oCAAoCC,EAAE,MAAM,yCAAyCD,EAAE,MAAM,KAAK,EAAE4V,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,oEAAoEA,mBAAmB,IAAI,EAAE4X,EAAEvX,IAAI,MAAM,IAAI,sCAAsCA,wCAAwC,EAAE,IAAIsC,EAAEq5B,EAAEh6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEY,EAAEo5B,EAAE/5B,EAAE,CAACA,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEgC,EAAE66B,GAAGl8B,EAAED,EAAE,CAAC,EAAE3C,CAAC,EAAEI,EAAE,OAAO,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,OAAOsC,EAAEs5B,EAAE/3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE+3B,EAAE/3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,IAAI+6B,GAAG/S,EAAE,CAAC,QAAQ8S,EAAE,CAAC,EAAE,SAASE,GAAGn/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,OAAO,EAAE,CAACuX,EAAE9X,EAAE,SAAS,EAAE,KAAK,IAAI,sBAAsBA,EAAE,2BAA2B,EAAE,kBAAkB,EAAE,IAAIiC,EAAEjC,EAAEkC,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAEj6B,EAAE,CAAC,EAAEjC,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,GAAG8X,EAAE7V,EAAE,SAAS,EAAE,IAAI,qEAAqEA,EAAE,SAAS,EAAE6V,EAAE5V,EAAE,OAAO,EAAE,IAAI,4DAA4DA,EAAE,MAAM,EAAE4V,EAAE5X,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,MAAM,EAAE,IAAI0C,EAAE,IAAI,OAAOX,EAAE,GAAGA,EAAE,GAAGY,EAAE,IAAI,OAAOX,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG4V,EAAElV,IAAI1C,EAAE,MAAM,GAAG,IAAI,4CAA4C0C,wCAAwC1C,EAAE,MAAM,KAAK,EAAE4X,EAAEjV,IAAI3C,EAAE,MAAM,GAAG,IAAI,6CAA6C2C,yCAAyC3C,EAAE,MAAM,KAAK,EAAE87B,GAAG,iBAAiBz7B,EAAE,CAAC,EAAE,IAAIuC,EAAE,CAAC,GAAGZ,EAAE,OAAOhC,CAAC,EAAE6C,EAAE,CAAC,QAAQzC,EAAE,IAAIC,EAAE,WAAW,EAAE,gBAAgB,EAAE,WAAW0B,CAAC,EAAEe,EAAEkoB,EAAE,UAAUjO,GAAGna,EAAEC,CAAC,EAAE,OAAOZ,EAAE+5B,EAAEl5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIo8B,GAAGjT,EAAE,CAAC,qBAAqBgT,EAAE,CAAC,EAAE,SAASE,GAAGr/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,iBAAiB,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,iBAAiB,EAAE,OAAOoT,GAAGl/B,EAAE,EAAE+B,EAAE3B,EAAEC,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI++B,GAAGnT,EAAE,CAAC,iBAAiBkT,EAAE,CAAC,EAAE,SAASE,GAAGv/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,QAAQ,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,QAAQ,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,QAAQ,EAAE9pB,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE5V,EAAE,OAAO,EAAE,IAAI,uDAAuDA,EAAE,OAAO,EAAE4V,EAAE7V,EAAE,OAAO,EAAE,IAAI,wDAAwDA,EAAE,OAAO,EAAE6V,EAAE5V,EAAE,MAAM,KAAKD,EAAE,MAAM,GAAG,IAAI,oCAAoCC,EAAE,MAAM,yCAAyCD,EAAE,MAAM,KAAK,EAAE6V,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,uEAAuEA,oBAAoB,IAAI,EAAE4X,EAAEvX,IAAI,QAAQ,IAAI,sCAAsCA,0CAA0C,EAAE,IAAIqC,EAAE,CAAC,EAAEV,EAAE,OAAOD,CAAC,EAAEY,EAAE,CAAC,QAAQ3C,EAAE,IAAII,EAAE,WAAWC,EAAE,UAAU,CAAC,EAAEuC,EAAEooB,EAAE,UAAUhO,GAAGta,EAAEC,CAAC,EAAE,OAAOV,EAAE+5B,EAAEp5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI08B,GAAGrT,EAAE,CAAC,QAAQoT,EAAE,CAAC,EAAE,SAASE,GAAGz/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAACuX,EAAE9X,EAAE,SAAS,EAAE,KAAK,IAAI,sBAAsBA,EAAE,2BAA2B,EAAE,kBAAkB,EAAE,IAAI,EAAEA,EAAE,EAAE,EAAEiC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAG,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEl8B,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,GAAG,IAAIkC,EAAE,EAAE,GAAGC,EAAE,EAAE,MAAM,GAAG2V,EAAE,EAAE,SAAS,EAAE,IAAI,qEAAqE,EAAE,SAAS,EAAEA,EAAE,EAAE,OAAO,EAAE,IAAI,4DAA4D,EAAE,MAAM,EAAEA,EAAE5X,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,MAAM,EAAE4X,EAAE5V,IAAIhC,EAAE,MAAM,GAAG,IAAI,4CAA4CgC,wCAAwChC,EAAE,MAAM,KAAK,EAAE4X,EAAE3V,IAAIjC,EAAE,MAAM,GAAG,IAAI,6CAA6CiC,yCAAyCjC,EAAE,MAAM,KAAK,EAAE,IAAI0C,EAAE,CAAC,GAAG,EAAE,OAAO1C,CAAC,EAAE2C,EAAE,CAAC,IAAItC,EAAE,QAAQD,EAAE,WAAW,CAAC,EAAEwC,EAAEooB,EAAE,UAAU9N,GAAGxa,EAAEC,CAAC,EAAE,OAAOZ,EAAEi6B,EAAEp5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI48B,GAAGvT,EAAE,CAAC,qBAAqBsT,EAAE,CAAC,EAAE,SAASE,GAAG3/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,iBAAiB,EAAE,EAAEgsB,EAAE,EAAE,SAAS,iBAAiB,EAAE,OAAO0T,GAAGx/B,EAAE,EAAE,EAAEI,EAAEC,CAAC,CAAC,CAAC,IAAIq/B,GAAGzT,EAAE,CAAC,iBAAiBwT,EAAE,CAAC,EAAE,SAASE,GAAG7/B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAU7N,GAAGnd,CAAC,CAAC,CAAC,IAAI4/B,GAAG3T,EAAE,CAAC,KAAK0T,EAAE,CAAC,EAAE,SAASE,GAAG//B,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAU5N,GAAGpd,CAAC,CAAC,CAAC,IAAI8/B,GAAG7T,EAAE,CAAC,MAAM4T,EAAE,CAAC,EAAE,SAASE,GAAGjgC,EAAE,EAAE,EAAEE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIyB,EAAE,CAAC,EAAEiqB,EAAEhsB,EAAE,IAAI,SAAS,CAAC,EAAEgC,EAAE,CAAC,KAAK,EAAE,UAAU9B,EAAE,QAAQI,CAAC,EAAE,OAAO4qB,EAAE,UAAU3N,GAAGxb,EAAEC,CAAC,CAAC,CAAC,IAAIk+B,GAAG/T,EAAE,CAAC,SAAS8T,EAAE,CAAC,EAAE,SAASE,GAAGngC,EAAE,EAAE,EAAEE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIyB,EAAE,CAAC,EAAEiqB,EAAEhsB,EAAE,IAAI,QAAQ,CAAC,EAAEgC,EAAE,CAAC,KAAK,EAAE,UAAU9B,EAAE,QAAQI,CAAC,EAAE,OAAO4qB,EAAE,UAAU1N,GAAGzb,EAAEC,CAAC,CAAC,CAAC,IAAIo+B,GAAGjU,EAAE,CAAC,QAAQgU,EAAE,CAAC,EAAE,SAASE,GAAGrgC,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,IAAI,eAAe,EAAE,EAAEgsB,EAAE,EAAE,UAAU,eAAe,EAAElU,EAAEvX,EAAE,QAAQ,QAAQ,IAAI,8DAA8DA,EAAE,OAAO,EAAEuX,EAAEvX,EAAE,MAAM,EAAE,IAAI,sEAAsEA,EAAE,OAAO,EAAEuX,EAAE5X,GAAG,EAAE,IAAI,sCAAsCA,IAAI,EAAE4X,EAAE,EAAE,OAAOvX,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,+FAA+FA,EAAE,yBAAyB,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAEA,EAAE,QAAQ,CAAC,EAAE0B,EAAE,CAAC,KAAK/B,EAAE,aAAaI,CAAC,EAAE,OAAO4qB,EAAE,UAAUxN,GAAG,EAAEzb,CAAC,CAAC,CAAC,IAAIq+B,GAAGnU,EAAE,CAAC,eAAekU,EAAE,CAAC,EAAE,SAASE,GAAGvgC,EAAE,EAAEE,EAAE,OAAO,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,eAAe,SAAS,EAAEO,EAAEL,IAAI,OAAOI,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG,EAAEJ,IAAI,OAAOI,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG,EAAEJ,IAAI,OAAOI,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGwX,EAAE,EAAE,EAAE,IAAI,sDAAsD,GAAG,EAAEA,EAAEvX,EAAE,GAAG,EAAE,IAAI;AAAA,MACryZA,SAAS;AAAA,MACTD,EAAE,OAAO,EAAEwX,EAAE,EAAE,GAAG,EAAE,IAAI;AAAA,MACxB,SAAS;AAAA,UACLxX,EAAE,OAAO,EAAEwX,EAAE,GAAG,EAAE,KAAK,EAAE,IAAI,8CAA8C,EAAE,YAAY,uCAAuCxX,EAAE,OAAO,EAAE,IAAI2B,EAAE,CAAC,EAAE3B,CAAC,EAAE4B,EAAE,CAAC,UAAU,EAAE,WAAWhC,CAAC,EAAE,OAAOgrB,EAAE,UAAUvN,GAAG1b,EAAEC,CAAC,CAAC,CAAC,IAAIs+B,GAAGrU,EAAE,CAAC,cAAcoU,EAAE,CAAC,EAAE,SAASE,GAAGzgC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,OAAO,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,IAAI,kBAAkB,SAAS,EAAEkC,EAAE8pB,EAAE,EAAE,SAAS,kBAAkB,SAAS,EAAE7pB,EAAEF,EAAEW,EAAE,GAAGX,EAAE,OAAO,IAAIW,EAAE,GAAGT,EAAE+5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG6V,EAAE3V,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,OAAO,EAAE2V,EAAE5V,EAAE,OAAO,EAAE,IAAI,iEAAiEA,EAAE,OAAO,EAAE,IAAIW,EAAEtC,IAAI,OAAO4B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG2V,EAAEjV,IAAIX,EAAE,MAAM,GAAG,IAAI,uDAAuDW,oDAAoDX,EAAE,MAAM,KAAK,EAAE85B,GAAG,kBAAkB17B,EAAE,CAAC,EAAE,IAAIwC,EAAE,CAAC,EAAEX,EAAE,OAAOD,CAAC,EAAEa,EAAE,CAAC,QAAQ7C,EAAE,IAAII,EAAE,WAAWC,EAAE,UAAU,EAAE,gBAAgB,CAAC,EAAEyC,EAAEkoB,EAAE,UAAUtN,GAAG9a,EAAEC,CAAC,EAAE,OAAOH,EAAEs5B,EAAEl5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI09B,GAAGvU,EAAE,CAAC,iBAAiBsU,EAAE,CAAC,EAAE,SAASE,GAAG3gC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUnN,GAAG7d,CAAC,CAAC,CAAC,IAAI0gC,GAAGzU,EAAE,CAAC,MAAMwU,EAAE,CAAC,EAAE,SAASE,GAAG7gC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,OAAO,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,YAAY,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,YAAY,EAAElU,EAAE,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,IAAI,gEAAgE,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,EAAE,IAAI,4DAA4DA,EAAE,OAAO,EAAE6V,EAAE,IAAI,OAAO,IAAI,gFAAgF,GAAG,EAAE,IAAI5V,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAID,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE/5B,EAAE,IAAI,IAAIS,EAAE,CAAC,EAAEV,EAAE,OAAOD,CAAC,EAAEY,EAAE,CAAC,QAAQ3C,EAAE,IAAII,EAAE,UAAUC,CAAC,EAAEuC,EAAEooB,EAAE,UAAUlN,GAAGpb,EAAEC,CAAC,EAAE,OAAOV,EAAE+5B,EAAEp5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIg+B,GAAG3U,EAAE,CAAC,YAAY0U,EAAE,CAAC,EAAE,SAASE,GAAG/gC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,QAAQ,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,QAAQ,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU1M,GAAGje,CAAC,CAAC,CAAC,IAAIygC,GAAG7U,EAAE,CAAC,OAAO4U,EAAE,CAAC,EAAE,SAASE,GAAGjhC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAE,EAAE,IAAI,OAAO,EAAEzrB,EAAEyrB,EAAE9rB,EAAE,IAAI,OAAO,EAAE,EAAE8rB,EAAEhsB,EAAE,YAAY,QAAQ,MAAM,EAAE,EAAEg1B,GAAGA,GAAG,EAAE,MAAM10B,EAAE,KAAK,EAAEC,EAAE,KAAK,EAAE0B,EAAEg8B,GAAG,EAAE,CAAC,EAAE/7B,EAAE+7B,GAAG39B,EAAE,CAAC,EAAE6B,EAAE87B,GAAG19B,EAAE,CAAC,EAAEqC,EAAE,CAAC,UAAUX,EAAE,EAAEC,EAAE,EAAEC,CAAC,EAAE,OAAO+oB,EAAE,UAAU3H,GAAG3gB,CAAC,CAAC,CAAC,IAAIs+B,GAAG/U,EAAE,CAAC,OAAO8U,EAAE,CAAC,EAAE,SAASE,GAAGnhC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,WAAW,CAAC,EAAE,OAAOkrB,EAAE,UAAUxF,GAAGxlB,CAAC,CAAC,CAAC,IAAIkhC,GAAGjV,EAAE,CAAC,WAAWgV,EAAE,CAAC,EAAE,SAASE,GAAGrhC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE24B,GAAGh5B,EAAEI,CAAC,EAAE,EAAE8gC,GAAG7gC,CAAC,EAAE,EAAEygC,GAAG1gC,EAAE,CAAC,EAAE,OAAO4gC,GAAG,EAAE,EAAE3gC,CAAC,CAAC,CAAC,IAAI+gC,GAAGnV,EAAE,CAAC,UAAUkV,EAAE,CAAC,EAAE,SAASE,GAAGvhC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,KAAK,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,KAAK,KAAK,EAAElU,GAAG5X,EAAE,OAAO,GAAGA,EAAE,OAAO,KAAKI,EAAE,OAAO,GAAGA,EAAE,OAAO,GAAG,IAAI,+DAA+DJ,EAAE,YAAYI,EAAE,OAAO,EAAE,IAAIC,EAAEL,EAAE,OAAO,EAAEA,EAAE,KAAKA,EAAE,MAAM,GAAG,EAAEI,EAAE,OAAO,EAAEA,EAAE,KAAKA,EAAE,MAAM,GAAG,GAAGwX,EAAEvX,IAAI,EAAE,IAAI,gEAAgEA,SAAS,IAAI,EAAEL,EAAE,OAAO,GAAGI,EAAE,OAAO,EAAE,CAAC,IAAI,EAAE47B,EAAEh8B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE+B,EAAEi6B,EAAE57B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE4B,EAAE4wB,GAAG,EAAE7wB,CAAC,EAAE,OAAOi6B,EAAEh6B,EAAE,CAAC,CAAC,CAAC,SAAShC,EAAE,OAAO,GAAGI,EAAE,OAAO,EAAE,CAAC,IAAI,EAAE47B,EAAEh8B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE+B,EAAEi6B,EAAE57B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE4B,EAAE4wB,GAAG,EAAE7wB,CAAC,EAAE,OAAOi6B,EAAEh6B,EAAE,CAACA,EAAE,IAAI,CAAC,CAAC,SAAShC,EAAE,OAAO,GAAGI,EAAE,OAAO,EAAE,CAAC,IAAI,EAAE47B,EAAE57B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE2B,EAAE6wB,GAAG5yB,EAAE,CAAC,EAAE,OAAOg8B,EAAEj6B,EAAE,CAACA,EAAE,IAAI,CAAC,CAAC,KAAK,CAAC,IAAI,EAAEi6B,EAAE57B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,OAAOwyB,GAAG5yB,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIshC,GAAGrV,EAAE,CAAC,KAAKoV,EAAE,CAAC,EAAE,SAASE,GAAGzhC,KAAK,EAAE,CAAC,IAAIE,EAAE,EAAE,IAAI,CAACK,EAAE,IAAIyrB,EAAEzrB,EAAE,UAAU,IAAI,QAAQ,CAAC,EAAED,EAAE,CAAC,SAASN,CAAC,EAAE,OAAOkrB,EAAE,UAAU9M,GAAGle,EAAEI,CAAC,CAAC,CAAC,IAAIohC,GAAGvV,EAAE,CAAC,QAAQsV,EAAE,CAAC,EAAE,SAASE,GAAG3hC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAU7M,GAAGne,CAAC,CAAC,CAAC,IAAI0hC,GAAGzV,EAAE,CAAC,KAAKwV,EAAE,CAAC,EAAE,SAASE,GAAG7hC,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,IAAI,KAAK,EAAE8X,EAAE,EAAE,QAAQ,SAAS,EAAE,QAAQ,UAAU,IAAI,2CAA2C,EAAE,EAAE,QAAQ,UAAU,EAAE2Y,GAAE,EAAE,SAAS,GAAG,IAAIvwB,EAAE,CAAC,EAAE,CAAC,EAAE,OAAOgrB,EAAE,UAAU3M,GAAGre,CAAC,CAAC,CAAC,IAAI4hC,GAAG3V,EAAE,CAAC,KAAK0V,EAAE,CAAC,EAAE,SAASE,GAAG/hC,EAAE,EAAE,CAAC,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,GAAGF,EAAEA,EAAE,OAAOE,EAAE,KAAK,EAAE,EAAEA,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC,SAAS8hC,GAAGhiC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,OAAO,EAAE,OAAOO,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,QAAQ0B,EAAE,EAAEA,EAAE3B,EAAE2B,IAAI/B,EAAE,QAAQ+B,CAAC,IAAI,GAAG1B,EAAE,KAAKP,EAAE,IAAI,EAAEO,EAAE,KAAK,EAAE,IAAI,EAAE,OAAOA,CAAC,CAAC,SAAS0hC,GAAGjiC,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAEN,EAAE,OAAO,QAAQ,EAAE,EAAE,EAAEM,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,IAAIJ,EAAE,KAAKF,EAAE,EAAE,EAAE,IAAIO,EAAE,EAAE,IAAI,GAAGP,EAAE,EAAE,EAAE,MAAM,CAACE,EAAEK,CAAC,CAAC,CAAC,SAAS2hC,GAAGliC,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAE,IAAII,GAAG,CAAC,EAAE,OAAO0hC,GAAGhiC,EAAEE,EAAE,CAAC,CAAC,CAAC,SAASiiC,GAAGniC,EAAE,EAAEE,EAAE,CAAC4X,EAAEiqB,GAAG,EAAE7hC,CAAC,EAAE,IAAI,GAAGF,qDAAqD,cAAcE,UAAU,CAAC,CAAC,SAASkiC,GAAGpiC,EAAE,EAAE,CAAC,GAAG+hC,GAAG/hC,EAAE,CAAC,EAAE,OAAO,KAAK,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAEN,EAAE,QAAQM,CAAC,IAAI,IAAIJ,EAAE,KAAKI,CAAC,EAAE,OAAON,EAAE,QAAQM,GAAGJ,EAAE,KAAKI,CAAC,CAAC,EAAEJ,CAAC,CAAC,SAASmiC,GAAGriC,EAAE,CAAC,OAAOA,EAAE,IAAI,CAAC,EAAEE,IAAI,CAACA,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,EAAEA,IAAI,EAAE,GAAGA,EAAE,EAAE,EAAE,IAAI,GAAG,EAAE,EAAE,CAAC,CAAC,SAASoiC,GAAGtiC,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEN,EAAEM,EAAE,EAAE,EAAEA,EAAEJ,EAAE,KAAKI,CAAC,EAAE,OAAOJ,CAAC,CAAC,SAASqiC,GAAGviC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,KAAK,CAAC,EAAE+B,EAAE,CAAC,iBAAiB,EAAE,SAAS7B,CAAC,EAAE,OAAOgrB,EAAE,UAAUxK,GAAGngB,EAAEwB,CAAC,CAAC,CAAC,IAAIygC,GAAGrW,EAAE,CAAC,KAAKoW,EAAE,CAAC,EAAE,SAASE,GAAGziC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,KAAK,CAAC,EAAE+B,EAAE,CAAC,KAAK,EAAE,SAAS7B,CAAC,EAAE,OAAOgrB,EAAE,UAAUhK,GAAG3gB,EAAEwB,CAAC,CAAC,CAAC,IAAI2gC,GAAGvW,EAAE,CAAC,KAAKsW,EAAE,CAAC,EAAE,SAASE,GAAG3iC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,OAAO,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,MAAM,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUhJ,GAAG3hB,CAAC,CAAC,CAAC,IAAIqiC,GAAGzW,EAAE,CAAC,KAAKwW,EAAE,CAAC,EAAE,SAASE,GAAG7iC,EAAE,EAAE,CAAC,IAAImY,GAAGnY,CAAC,GAAG,IAAI,UAAU,MAAM,QAAQA,CAAC,IAAI,IAAI,YAAY,MAAM,IAAI,MAAM,gFAAgF,EAAE,GAAG,IAAI,UAAUmY,GAAGnY,CAAC,GAAG,EAAEA,aAAa,YAAY,MAAM,IAAI,MAAM,2EAA2E,EAAE,OAAOssB,GAAGtsB,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,SAAS8iC,GAAG9iC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUnH,GAAG7jB,CAAC,CAAC,CAAC,IAAI6iC,GAAG5W,EAAE,CAAC,MAAM2W,EAAE,CAAC,EAAE,SAASE,GAAGhjC,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,IAAI,QAAQ,EAAEE,EAAE,CAAC,EAAE,OAAOgrB,EAAE,UAAU,SAAS,CAAC,EAAE,CAAC,EAAEhrB,CAAC,CAAC,CAAC,IAAI+iC,GAAG9W,EAAE,CAAC,QAAQ6W,EAAE,CAAC,EAAE,SAASE,GAAGljC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE,QAAQ,SAASA,EAAEmwB,GAAEnwB,EAAE,OAAO,GAAG,IAAIC,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAASJ,CAAC,EAAE,OAAOgrB,EAAE,UAAUlH,GAAGzjB,EAAE,CAAC,CAAC,CAAC,IAAI4iC,GAAGhX,EAAE,CAAC,KAAK+W,EAAE,CAAC,EAAE,SAASE,GAAGpjC,EAAE,EAAE,YAAYE,EAAE,KAAKI,EAAE,GAAG,CAACN,EAAEgsB,EAAEhsB,EAAE,IAAI,MAAM,EAAE,IAAIO,EAAE8iC,GAAGrjC,EAAE,EAAEE,CAAC,EAAE,EAAEK,EAAE,MAAM,GAAGD,EAAE,CAAC,IAAI,EAAEuY,GAAG3Y,EAAEF,EAAE,KAAK,EAAE,EAAEkiC,GAAG3hC,EAAE,MAAM,CAAC,CAAC,CAAC,OAAO27B,EAAE37B,EAAE,CAAC,CAAC,CAAC,SAAS8iC,GAAGrjC,EAAE,EAAEE,EAAE,KAAK,CAAC,GAAGF,EAAE,OAAO,EAAE,OAAOs5B,GAAGt5B,CAAC,EAAE,GAAGA,EAAE,OAAO,GAAGE,IAAI,KAAK,OAAOmjC,GAAGnH,EAAEl8B,EAAE,CAAC,EAAE,CAAC,EAAE,EAAEE,CAAC,EAAE,GAAGF,EAAE,OAAO,GAAG,OAAOE,GAAG,UAAU,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,EAAE,CAAC,GAAG,IAAI,EAAE,OAAOijC,GAAG7J,GAAGt5B,CAAC,EAAEE,CAAC,EAAE,GAAG,IAAI,EAAE,EAAE,OAAOsiC,GAAGlJ,GAAGt5B,CAAC,EAAEE,CAAC,EAAE,GAAG,IAAI,GAAG,EAAE,OAAOwiC,GAAGpJ,GAAGt5B,CAAC,EAAEE,CAAC,EAAE,GAAG,IAAI,aAAa,IAAI,EAAE,OAAO6iC,GAAGI,GAAGP,GAAGtJ,GAAGt5B,CAAC,EAAE6iC,GAAG,EAAE,OAAO,CAAC,EAAE3iC,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,qCAAqC,GAAG,CAAC,CAAC,GAAG,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,EAAE,CAAC,GAAG,IAAI,EAAE,OAAOsiC,GAAGW,GAAG7J,GAAGt5B,CAAC,EAAEE,EAAE,EAAE,EAAEA,EAAE,GAAG,CAAC,EAAE,GAAG,IAAI,EAAE,EAAE,OAAOsiC,GAAGW,GAAG7J,GAAGt5B,CAAC,EAAEE,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAE,GAAG,IAAI,GAAG,EAAE,OAAOwiC,GAAGS,GAAG7J,GAAGt5B,CAAC,EAAEE,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAE,GAAG,IAAI,OAAO,IAAI,YAAY,OAAO6iC,GAAGI,GAAGF,GAAGjjC,CAAC,EAAEE,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,qCAAqC,GAAG,CAAC,CAAC,MAAM,IAAI,MAAM,gCAAgCA,GAAG,CAAC,CAAC,IAAIojC,GAAGnX,EAAE,CAAC,MAAMiX,EAAE,CAAC,EAAE,SAASG,GAAGvjC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,OAAOojC,GAAGtjC,EAAE,YAAY,EAAEE,CAAC,CAAC,CAAC,IAAIsjC,GAAGrX,EAAE,CAAC,eAAeoX,EAAE,CAAC,EAAE,SAASE,GAAGzjC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,CAAC,EAAE,OAAOkrB,EAAE,UAAUzM,GAAGve,CAAC,CAAC,CAAC,IAAIwjC,GAAGvX,EAAE,CAAC,KAAKsX,EAAE,CAAC,EAAE,SAASE,GAAG3jC,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,aAAa,mBAAmB,EAAE8X,EAAE,GAAG5X,EAAE,KAAK,IAAI,oCAAoC,EAAE,IAAII,EAAE,CAAC,MAAMJ,CAAC,EAAEK,EAAE,CAAC,IAAI,CAAC,EAAE,OAAO2qB,EAAE,UAAUxM,GAAGpe,EAAEC,CAAC,CAAC,CAAC,IAAIqjC,GAAGzX,EAAE,CAAC,YAAYwX,EAAE,CAAC,EAAE,SAASE,GAAG7jC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUvM,GAAGze,CAAC,CAAC,CAAC,IAAI4jC,GAAG3X,EAAE,CAAC,OAAO0X,EAAE,CAAC,EAAE,SAASE,GAAG/jC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,mBAAmB,EAAE8X,EAAE5X,EAAE,OAAO,EAAE,OAAO,IAAI,qCAAqCA,EAAE,kCAAkC,IAAI,EAAE,IAAII,EAAE,CAAC,EAAEJ,CAAC,EAAEK,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUhG,GAAG5kB,EAAEC,CAAC,CAAC,CAAC,IAAIyjC,GAAG7X,EAAE,CAAC,MAAM4X,EAAE,CAAC,EAAE,SAASE,GAAGjkC,EAAE,EAAEE,EAAEI,EAAE,UAAU,CAAC,GAAG,OAAO,EAAEN,GAAG,IAAIO,EAAEgwB,GAAG,CAACvwB,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAEN,GAAG,EAAEA,EAAE,EAAE,QAAQiC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE1B,EAAE,IAAI,EAAE0B,EAAEA,CAAC,EAAE,IAAI,EAAEi6B,EAAE37B,EAAE,SAAS,EAAE,CAACP,EAAE,CAAC,CAAC,EAAE,GAAGE,GAAG,KAAK,OAAO,EAAE,GAAGA,EAAE,SAAS,EAAE,OAAO8jC,GAAGJ,GAAG,EAAE,CAAC,EAAE,CAAC1jC,EAAE,GAAG,EAAE,CAAC,CAAC,EAAE,GAAGA,EAAE,SAAS,EAAE,OAAO8jC,GAAGJ,GAAGA,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC1jC,EAAE,GAAGA,EAAE,GAAG,EAAE,CAAC,CAAC,EAAE,GAAGA,EAAE,SAAS,EAAE,OAAO8jC,GAAGJ,GAAGA,GAAGA,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC1jC,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAG,EAAE,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,qEAAqEA,EAAE,UAAU,CAAC,CAAC,IAAIgkC,GAAG/X,EAAE,CAAC,KAAK8X,EAAE,CAAC,EAAE,SAASE,GAAGnkC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,QAAQ,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUnM,GAAG7e,CAAC,CAAC,CAAC,IAAIkkC,GAAGjY,EAAE,CAAC,OAAOgY,EAAE,CAAC,EAAE,SAASE,GAAGrkC,EAAE,EAAEE,EAAE,EAAEI,EAAE,EAAE,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,IAAI,QAAQ,EAAE,EAAEgsB,EAAE,EAAE,UAAU,SAAS,OAAO,EAAE,EAAE,CAAC,EAAEzrB,EAAE,QAAQ,CAAC,EAAE0B,EAAE,CAAC,KAAK/B,EAAE,UAAUI,CAAC,EAAE,OAAO4qB,EAAE,UAAUhM,GAAG,EAAEjd,CAAC,CAAC,CAAC,IAAIqiC,GAAGnY,EAAE,CAAC,QAAQkY,EAAE,CAAC,EAAE,SAASE,GAAGvkC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,UAAU,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU9L,GAAG7e,CAAC,CAAC,CAAC,IAAIikC,GAAGrY,EAAE,CAAC,SAASoY,EAAE,CAAC,EAAE,SAASE,GAAGzkC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,eAAe,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,eAAe,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU7L,GAAG9e,CAAC,CAAC,CAAC,IAAImkC,GAAGvY,EAAE,CAAC,cAAcsY,EAAE,CAAC,EAAE,SAASE,GAAG3kC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,CAAC,EAAE,OAAOkrB,EAAE,UAAUzL,GAAGvf,CAAC,CAAC,CAAC,IAAI0kC,GAAGzY,EAAE,CAAC,UAAUwY,EAAE,CAAC,EAAE,SAASE,GAAG7kC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUxL,GAAGxf,CAAC,CAAC,CAAC,IAAI4kC,GAAG3Y,EAAE,CAAC,OAAO0Y,EAAE,CAAC,EAAE,SAASE,GAAG/kC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUvL,GAAGzf,CAAC,CAAC,CAAC,IAAI8kC,GAAG7Y,EAAE,CAAC,OAAO4Y,EAAE,CAAC,EAAE,SAASE,GAAGjlC,EAAE,EAAE,GAAG,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,WAAW,CAAC,EAAEO,EAAE,CAAC,MAAM,CAAC,EAAE,OAAO2qB,EAAE,UAAUtL,GAAG,EAAErf,CAAC,CAAC,CAAC,IAAI2kC,GAAG/Y,EAAE,CAAC,WAAW8Y,EAAE,CAAC,EAAE,SAASE,GAAGnlC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,OAAO,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUrL,GAAGtf,CAAC,CAAC,CAAC,IAAI6kC,GAAGjZ,EAAE,CAAC,MAAMgZ,EAAE,CAAC,EAAE,SAASE,GAAGrlC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,YAAY,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,YAAY,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUpL,GAAGvf,CAAC,CAAC,CAAC,IAAI+kC,GAAGnZ,EAAE,CAAC,WAAWkZ,EAAE,CAAC,EAAE,SAASE,GAAGvlC,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,EAAE,MAAM,IAAI,MAAM,0CAA0C,EAAE,IAAII,EAAE,CAAC,MAAMN,EAAE,KAAK,EAAE,IAAIE,CAAC,EAAE,OAAOgrB,EAAE,UAAUnL,GAAG,CAAC,EAAEzf,CAAC,CAAC,CAAC,SAASklC,GAAGxlC,EAAE,EAAE,EAAEE,EAAE,EAAEI,EAAE,EAAEC,EAAE,GAAG,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,4BAA4B,EAAE8X,EAAE,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,IAAI;AAAA,sBACt8R,EAAE,OAAO,EAAEA,EAAEQ,GAAG,CAAC,EAAE,IAAI,2FAA2F,IAAI,EAAE,IAAI,EAAE,EAAErW,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAG,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAG,IAAIh6B,EAAE,CAAC,EAAE,CAAC,EAAEC,EAAE,CAAC,YAAY,EAAE,KAAKjC,EAAE,MAAMI,EAAE,KAAKC,CAAC,EAAEqC,EAAEsoB,EAAE,UAAU1K,GAAGte,EAAEC,CAAC,EAAE,OAAOF,EAAEi6B,EAAEt5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI6iC,GAAGtZ,EAAE,CAAC,4BAA4BqZ,EAAE,CAAC,EAAE,SAASE,GAAG1lC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUlL,GAAG9f,CAAC,CAAC,CAAC,IAAIylC,GAAGxZ,EAAE,CAAC,KAAKuZ,EAAE,CAAC,EAAE,SAASE,GAAG5lC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUjL,GAAG/f,CAAC,CAAC,CAAC,IAAI2lC,GAAG1Z,EAAE,CAAC,OAAOyZ,EAAE,CAAC,EAAE,SAASE,GAAG9lC,EAAE,CAAC,OAAO8X,EAAE4B,GAAG1Z,CAAC,EAAE,IAAI,4CAA4C,EAAE,CAAC,EAAEE,IAAI,CAAC,IAAII,EAAE0rB,EAAE,EAAE,IAAI,UAAU,mBAAmB,EAAEzrB,EAAEL,GAAG,KAAK8rB,EAAE9rB,EAAE,KAAK,SAAS,EAAE,KAAK,OAAOgrB,EAAE,KAAK,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,EAAEA,EAAE,UAAU,IAAIlrB,EAAEM,CAAC,EAAE,CAACA,CAAC,EAAEC,CAAC,EAAE,OAAOA,GAAG,MAAMwX,GAAG,EAAE,MAAMxX,EAAE,MAAM,gFAAgF,EAAEwlC,GAAG,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASC,GAAGhmC,EAAE,CAAC,OAAO8X,EAAE4B,GAAG1Z,CAAC,EAAE,IAAI,6CAA6C,EAAE,CAAC,EAAEE,IAAI,CAAC4X,EAAE,MAAM,QAAQ,CAAC,EAAE,IAAI,kFAAkF,EAAE,IAAIxX,EAAE2rB,GAAG,EAAE,OAAO,WAAW,mBAAmB,EAAE1rB,EAAEL,GAAG,KAAK8rB,EAAE9rB,EAAE,KAAK,UAAU,EAAE,KAAK,OAAOgrB,EAAE,KAAK,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,EAAEA,EAAE,UAAU,IAAIlrB,EAAE,GAAGM,CAAC,EAAEA,EAAEC,CAAC,EAAE,OAAOA,GAAG,MAAMwX,GAAG,EAAE,MAAMxX,EAAE,MAAM,+FAA+F,EAAEwlC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,SAASE,GAAGjmC,EAAE,CAAC,OAAO8X,EAAE4B,GAAG1Z,CAAC,EAAE,IAAI,oDAAoD,EAAE,CAAC,EAAEE,IAAI,CAAC4X,EAAE,aAAa+R,GAAG,IAAI,qDAAqD,EAAE/R,EAAE5X,GAAG,MAAMA,aAAa2pB,GAAG,IAAI,0DAA0D,EAAE,GAAG,CAAC,MAAMvpB,EAAE,MAAMC,CAAC,EAAE2qB,EAAE,UAAU,IAAIlrB,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEE,CAAC,EAAE,OAAO6lC,GAAGzlC,CAAC,EAAE,CAAC,KAAKA,EAAE,GAAG,MAAMC,CAAC,CAAC,CAAC,CAAC,SAAS2lC,GAAGlmC,EAAE,CAAC,OAAO8X,EAAE4B,GAAG1Z,CAAC,EAAE,IAAI,qDAAqD,EAAE,CAAC,EAAEE,IAAI,CAAC4X,EAAE,MAAM,QAAQ,CAAC,GAAG,EAAE,MAAMvX,GAAGA,aAAaspB,EAAE,EAAE,IAAI,oEAAoE,EAAE/R,EAAE5X,GAAG,MAAMA,aAAa2pB,GAAG,IAAI,8DAA8D,EAAE,IAAIvpB,EAAE4qB,EAAE,UAAU,IAAIlrB,EAAE,GAAG,CAAC,EAAE,EAAEE,CAAC,EAAE,OAAOA,GAAG,MAAM6X,GAAGzX,EAAE,MAAM,MAAMJ,EAAE,MAAM,uGAAuG,EAAE6lC,GAAGzlC,EAAE,KAAK,EAAEA,CAAC,CAAC,CAAC,SAAS6lC,GAAGnmC,EAAE,EAAE,CAAC8X,EAAE4B,GAAG1Z,CAAC,EAAE,IAAI,qDAAqD,EAAE8X,EAAE,GAAG,MAAM,MAAM,QAAQ,CAAC,GAAG,EAAE,MAAM3V,GAAGA,aAAa4nB,EAAE,EAAE,IAAI,+EAA+E,EAAE,IAAI7pB,EAAE,GAAG,KAAK,GAAG,CAACA,EAAE,CAAC,EAAE,CAAC,EAAE,QAAQiC,KAAK+oB,EAAE,oBAAoB,EAAE,KAAKA,EAAE,oBAAoB/oB,EAAE,CAAC,CAAC,IAAI7B,EAAEJ,EAAE,EAAE,OAAOiC,GAAG,CAACA,EAAE,SAAS,EAAE,KAAK5B,EAAE,EAAE,OAAO,EAAE,EAAE,OAAO4B,GAAGA,EAAE,SAAS,EAAE2V,EAAE,EAAE,OAAO,EAAE,IAAI,gGAAgGvX,2BAA2B,EAAE,IAAI,EAAE,GAAG,CAAC,MAAM,EAAE,MAAM0B,CAAC,EAAEipB,EAAE,UAAUlrB,EAAE,EAAE,KAAK,CAAC,EAAE8X,EAAE7V,EAAE,KAAKE,GAAGA,GAAG,IAAI,EAAE,IAAI,8LAA8L,EAAE2V,EAAE,EAAE,OAAO,EAAE,IAAI,iFAAiF,EAAE,aAAa,EAAE,IAAI5V,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ,CAACC,EAAES,IAAI,CAACX,EAAEW,IAAI,OAAOV,EAAEC,EAAE,MAAMF,EAAEW,GAAG,CAAC,EAAEtC,GAAG,MAAMA,EAAE,QAAQ6B,GAAGD,EAAEC,EAAE,MAAM,IAAI,EAAE,CAAC,MAAM,EAAE,MAAMD,CAAC,CAAC,CAAC,SAASkkC,GAAGpmC,EAAE,CAAC,OAAOkrB,EAAE,WAAWlrB,CAAC,CAAC,CAAC,SAAS+lC,GAAG/lC,EAAE,CAAC,GAAGA,EAAE,OAAOE,GAAGA,GAAG,IAAI,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,oEACryG,CAAC,CAAC,SAASmmC,GAAGrmC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,CAAC,EAAE,OAAOkrB,EAAE,UAAUpH,GAAG5jB,CAAC,CAAC,CAAC,IAAIomC,GAAGna,EAAE,CAAC,UAAUka,EAAE,CAAC,EAAE,SAASE,GAAGvmC,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,IAAI,YAAY,EAAE,OAAOomC,GAAG,IAAI,CAAC,MAAM3R,GAAG6R,GAAG7R,GAAG,CAAC,CAAC,CAAC,EAAE,SAASzyB,GAAGo3B,EAAEp3B,EAAE06B,GAAGjI,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI+R,GAAGra,EAAE,CAAC,YAAYoa,EAAE,CAAC,EAAE,SAASE,GAAGzmC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUnG,GAAGxkB,CAAC,CAAC,CAAC,IAAImmC,GAAGva,EAAE,CAAC,KAAKsa,EAAE,CAAC,EAAE,SAASE,GAAG3mC,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,SAAS,YAAY,EAAE,GAAG,IAAI,KAAK,EAAEE,EAAE,KAAK,GAAG,IAAIA,EAAE,KAAK,EAAE,MAAM,MAAM,gFAAgFA,EAAE,qBAAqB,GAAG,EAAE,OAAOkmC,GAAG,CAAC7lC,EAAEwB,IAAI,CAAC,IAAIE,EAAEugC,GAAGjiC,EAAE,EAAE,EAAE,EAAE2B,EAAEwkC,GAAGnmC,EAAE0B,CAAC,EAAEE,EAAEukC,GAAGjW,GAAEvuB,EAAE,SAAS,EAAEyjC,GAAGxC,GAAGO,GAAGxhC,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,OAAOH,EAAE,CAACI,CAAC,CAAC,EAAE,CAAC,MAAMA,EAAE,SAAS,CAACU,EAAEC,IAAI,CAAC,GAAG,CAACC,CAAC,EAAED,EAAE,EAAE,GAAGoB,EAAEw/B,GAAG3gC,CAAC,EAAE,OAAO2jC,GAAG7jC,EAAEu2B,EAAE+J,GAAGtgC,EAAE,EAAE,CAAC,EAAEqB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEhE,CAAC,CAAC,CAAC,IAAI0mC,GAAGza,EAAE,CAAC,YAAYwa,EAAE,CAAC,EAAE,SAASE,GAAG7mC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,WAAW,EAAEO,EAAEsY,GAAG,EAAEvY,EAAE,KAAK,EAAE,EAAEkiC,GAAGliC,EAAEC,EAAE,EAAE,EAAE,EAAEmmC,GAAGpmC,EAAE,CAAC,EAAE2B,EAAEyhC,GAAG,CAAC,EAAExhC,EAAEihC,GAAGlhC,EAAE1B,CAAC,EAAE4B,EAAEwjC,GAAGzjC,CAAC,EAAEU,EAAEk2B,EAAEoD,EAAE,EAAE/5B,EAAE,KAAK,EAAEA,CAAC,EAAE,GAAGjC,EAAE,CAAC,IAAI2C,EAAEq/B,GAAGt/B,EAAE,MAAMrC,CAAC,EAAE,OAAO27B,EAAEt5B,EAAEC,CAAC,CAAC,CAAC,OAAOD,CAAC,CAAC,IAAIkkC,GAAG3a,EAAE,CAAC,WAAW0a,EAAE,CAAC,EAAE,SAASE,GAAG/mC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,aAAa,MAAM,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,aAAa,MAAM,EAAEgJ,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUhL,GAAG3f,CAAC,CAAC,CAAC,IAAIymC,GAAG7a,EAAE,CAAC,YAAY4a,EAAE,CAAC,EAAE,SAASE,GAAGjnC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,aAAa,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAU/K,GAAGjgB,CAAC,CAAC,CAAC,IAAIgnC,GAAG/a,EAAE,CAAC,YAAY8a,EAAE,CAAC,EAAE,SAASE,GAAGnnC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,YAAY,MAAM,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,YAAY,MAAM,EAAEgJ,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU9K,GAAG7f,CAAC,CAAC,CAAC,IAAI6mC,GAAGjb,EAAE,CAAC,WAAWgb,EAAE,CAAC,EAAE,SAASE,GAAGrnC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,aAAa,MAAM,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,aAAa,MAAM,EAAE,OAAOgJ,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE0mC,GAAGI,GAAGpnC,EAAE,CAAC,EAAEknC,GAAGF,GAAGhnC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIsnC,GAAGnb,EAAE,CAAC,YAAYkb,EAAE,CAAC,EAAME,GAAG,WAAW,SAASC,GAAGxnC,EAAE,EAAEE,EAAE,OAAO,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,iBAAiB,cAAc,EAAEO,EAAEyrB,EAAE,EAAE,SAAS,cAAc,EAAE,EAAE1rB,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG,EAAEC,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG0B,EAAEi6B,EAAE57B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE4B,EAAEg6B,EAAE37B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG0B,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAE,GAAGA,EAAE,MAAM,KAAKC,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,gEAAgE,EAAE,GAAGkW,GAAGlW,EAAE,KAAK,GAAGqlC,GAAG,MAAM,IAAI,MAAM,qCAAqCA,IAAI,EAAE,GAAGtlC,EAAE,MAAM,IAAIslC,GAAG,MAAM,IAAI,MAAM,oCAAoCA,iCAAiCtlC,EAAE,MAAM,IAAI,EAAE,IAAIE,EAAE,CAAC,eAAeF,EAAE,OAAOC,CAAC,EAAEU,EAAE,CAAC,KAAK1C,CAAC,EAAE,OAAOgrB,EAAE,UAAU5H,GAAGnhB,EAAES,CAAC,CAAC,CAAC,IAAI6kC,GAAGtb,EAAE,CAAC,cAAcqb,EAAE,CAAC,EAAE,SAASE,GAAG1nC,EAAE,EAAE,CAAC,OAAOynC,GAAGznC,EAAE,EAAE,MAAM,CAAC,CAAC,SAAS2nC,GAAG3nC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,EAAE,EAAEiC,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE7V,EAAE,OAAO,EAAE,IAAI,uDAAuDA,EAAE,OAAO,EAAE6V,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,wEAAwEA,oBAAoB,IAAI,EAAE87B,GAAG,UAAU17B,EAAEC,CAAC,EAAE,IAAI4B,EAAE,CAAC,EAAEF,CAAC,EAAEW,EAAE,CAAC,WAAW,EAAE,QAAQ1C,EAAE,IAAII,EAAE,gBAAgBC,CAAC,EAAEsC,EAAEqoB,EAAE,UAAUtK,GAAGze,EAAES,CAAC,EAAE,OAAOV,EAAEg6B,EAAEr5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI+kC,GAAGzb,EAAE,CAAC,SAASwb,EAAE,CAAC,EAAE,SAASE,GAAG7nC,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEE,EAAEI,EAAEC,EAAE,EAAE,QAAQ,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,WAAW,EAAEiC,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE7V,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,OAAO,EAAE6V,EAAE,IAAI,QAAQ,IAAI,gFAAgF,GAAG,EAAEkkB,GAAG,YAAY17B,EAAEC,CAAC,EAAE,IAAI4B,EAAE,CAAC,EAAEF,CAAC,EAAEW,EAAE,CAAC,WAAW,EAAE,QAAQ1C,EAAE,IAAII,EAAE,gBAAgBC,EAAE,WAAW,CAAC,EAAEsC,EAAEqoB,EAAE,UAAUpK,GAAG3e,EAAES,CAAC,EAAE,OAAOV,EAAEg6B,EAAEr5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIilC,GAAG3b,EAAE,CAAC,WAAW0b,EAAE,CAAC,EAAE,SAASE,GAAG/nC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,GAAG,CAAC,IAAIyB,EAAE,CAAC,EAAEgqB,EAAEhsB,EAAE,IAAI,mBAAmB,CAAC,EAAEiC,EAAE,CAAC,WAAW,EAAE,QAAQ/B,EAAE,IAAII,EAAE,oBAAoBC,CAAC,EAAE2B,EAAEgpB,EAAE,UAAUlK,GAAGhf,EAAEC,CAAC,EAAE,MAAM,CAAC,OAAOC,EAAE,GAAG,QAAQA,EAAE,EAAE,CAAC,CAAC,IAAI8lC,GAAG7b,EAAE,CAAC,mBAAmB4b,EAAE,CAAC,EAAE,SAASE,GAAGjoC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,SAAS,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAEJ,EAAE,QAAQ,SAASA,EAAEuwB,GAAEvwB,EAAE,OAAO,EAAEI,EAAEmwB,GAAEnwB,EAAE,OAAO,GAAG00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUvK,GAAGpgB,CAAC,CAAC,CAAC,IAAI2nC,GAAG/b,EAAE,CAAC,SAAS8b,EAAE,CAAC,EAAE,SAASE,GAAGnoC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE+B,EAAE,CAAC,KAAK,EAAE,SAAS7B,CAAC,EAAE,OAAOgrB,EAAE,UAAUjK,GAAG1gB,EAAEwB,CAAC,CAAC,CAAC,IAAIqmC,GAAGjc,EAAE,CAAC,MAAMgc,EAAE,CAAC,EAAE,SAASE,GAAGroC,EAAE,EAAE,UAAU,CAAC,GAAG,IAAI,YAAY,CAAC,IAAIM,EAAE+nC,GAAGroC,EAAE,SAAS,EAAEO,EAAE8nC,GAAGroC,EAAE,SAAS,EAAE,OAAOqsB,GAAG/rB,EAAEC,CAAC,CAAC,CAAC,IAAIL,EAAE8Z,GAAG5B,GAAGpY,CAAC,EAAE,CAAC,EAAE,OAAOkrB,EAAE,WAAWhrB,EAAEF,EAAE,CAAC,CAAC,CAAC,SAASsoC,GAAGtoC,EAAE,EAAE,UAAU,CAAC,GAAG,IAAI,YAAY,CAAC,IAAIM,EAAEgoC,GAAGtoC,EAAE,SAAS,EAAEO,EAAE8nC,GAAGroC,EAAE,SAAS,EAAE,OAAOqsB,GAAG/rB,EAAEC,CAAC,CAAC,CAAC,IAAIL,EAAE6Z,GAAG3B,GAAGpY,CAAC,EAAE,CAAC,EAAE,OAAOkrB,EAAE,WAAWhrB,EAAEF,EAAE,CAAC,CAAC,CAAC,SAASuoC,GAAGvoC,EAAE,EAAE,CAAC,SAASE,EAAE,IAAI,EAAE,CAAC,EAAE,CAAC,GAAGA,IAAI,MAAMA,IAAI,KAAK,MAAM,IAAI,UAAU,GAAGA,6CAA6C,EAAE,GAAGF,IAAI,OAAO,MAAM,CAAC,EAAE,IAAIM,EAAE0rB,EAAEhsB,EAAE,IAAI,WAAWA,aAAa6pB,GAAG7pB,EAAE,MAAM,SAAS,EAAE,GAAG,IAAI,OAAO,MAAM,CAACM,CAAC,EAAE,IAAIC,EAAEyrB,EAAE,EAAE,IAAI,WAAW,aAAanC,GAAG,EAAE,MAAM,SAAS,EAAE,EAAEzR,GAAG9X,EAAE,KAAK,EAAE,EAAE8X,GAAG7X,EAAE,KAAK,EAAE,OAAOL,IAAI,MAAMI,EAAE47B,EAAE57B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE27B,EAAE37B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE,CAACuyB,GAAGwV,GAAG,CAAC,EAAE,CAAC,EAAEhoC,EAAE,KAAK,EAAEA,CAAC,EAAEwyB,GAAGvyB,EAAE+nC,GAAG,CAAC,EAAE,CAAC,EAAE/nC,EAAE,KAAK,CAAC,CAAC,IAAID,EAAE47B,EAAE57B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAEC,EAAE27B,EAAE37B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,CAACuyB,GAAGxyB,EAAEgoC,GAAG,CAAC,EAAE,CAAC,EAAEhoC,EAAE,KAAK,CAAC,EAAEwyB,GAAGwV,GAAG,CAAC,EAAE,CAAC,EAAE/nC,EAAE,KAAK,EAAEA,CAAC,CAAC,EAAE,CAAC,SAASioC,GAAGxoC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,SAAS,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAEJ,EAAE,QAAQ,SAASA,EAAEuwB,GAAEvwB,EAAE,OAAO,EAAEI,EAAEmwB,GAAEnwB,EAAE,OAAO,GAAG00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU/J,GAAG5gB,CAAC,CAAC,CAAC,IAAIkoC,GAAGtc,EAAE,CAAC,SAASqc,EAAE,CAAC,EAAE,SAASE,GAAG1oC,EAAE,EAAEE,EAAE,CAAC4X,EAAE5X,IAAI,WAAWA,IAAI,YAAY,IAAI,+DAA+DA,IAAI,EAAE,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,WAAW,EAAE,GAAGM,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,gEAAgE,EAAEwX,EAAE,EAAE,SAASxX,EAAE,KAAK,IAAI,wCAAwCA,EAAE,aAAa,EAAE,SAAS,EAAE,IAAIC,EAAEL,IAAI,UAAU,EAAE,EAAE,QAAQ+B,EAAE,EAAEA,EAAE3B,EAAE,KAAK2B,IAAI6V,EAAE,EAAE7V,GAAG,SAAS,EAAE,IAAI,uDAAuD,EAAE6V,EAAE,EAAE7V,GAAG,IAAI,GAAG,EAAEA,GAAG,IAAI3B,EAAE,MAAM2B,GAAG1B,GAAG,EAAE0B,GAAG,IAAI,GAAG,EAAEA,GAAG,IAAI3B,EAAE,MAAM2B,GAAG1B,EAAE,IAAI,wBAAwB0B,wCAAwC3B,EAAE,MAAM2B,GAAG1B,uCAAuCD,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,SAAS,EAAE,KAAKJ,CAAC,EAAE,EAAE,CAAC,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU9J,GAAG,EAAE,CAAC,CAAC,CAAC,IAAIunB,GAAGxc,EAAE,CAAC,WAAWuc,EAAE,CAAC,EAAE,SAASE,GAAG5oC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,KAAK,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,KAAK,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAU7J,GAAG9gB,CAAC,CAAC,CAAC,IAAIsoC,GAAG1c,EAAE,CAAC,KAAKyc,EAAE,CAAC,EAAE,SAASE,GAAG9oC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAACF,EAAEgsB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,IAAIM,EAAEuY,GAAG,EAAE7Y,EAAE,KAAK,EAAEO,EAAE6nC,GAAGpoC,EAAEM,EAAEJ,CAAC,EAAE,EAAEK,EAAE,MAAML,IAAI,EAAEgiC,GAAG3hC,EAAE,MAAMD,CAAC,GAAG,IAAI,EAAE2iC,GAAGyD,GAAGjW,GAAEzwB,EAAE,SAAS,EAAEk8B,EAAE37B,EAAE,CAAC,CAAC,CAAC,EAAE0B,EAAEmmC,GAAG,EAAE9nC,EAAEJ,CAAC,EAAE,MAAM,CAAC,KAAKK,EAAE,SAAS0B,CAAC,CAAC,CAAC,IAAI8mC,GAAG5c,EAAE,CAAC,SAAS2c,EAAE,CAAC,EAAE,SAASE,GAAGhpC,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEyrB,EAAE,EAAE,OAAO,cAAc,EAAE,EAAEC,GAAG/rB,EAAE,IAAI,cAAc,EAAE,EAAE+rB,GAAG3rB,EAAE,IAAI,cAAc,EAAE2B,EAAE1B,EAAE2B,EAAE,CAAC,EAAE,QAAQW,EAAE,EAAEA,EAAE7C,EAAE,OAAO6C,IAAI,CAAC,IAAIC,EAAE9C,EAAE6C,GAAGZ,EAAE,EAAEY,GAAG,EAAEA,EAAE,EAAEX,EAAE,KAAKY,EAAE,EAAE,EAAEZ,EAAE,KAAKY,EAAE,EAAE,EAAEb,EAAEa,EAAE,EAAE,CAAC,IAAIX,EAAE,CAAC,EAAES,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEX,EAAE,OAAOW,GAAG,EAAEV,EAAE,KAAKD,EAAEW,EAAE,EAAED,EAAE,KAAKV,EAAEW,EAAE,EAAE,EAAE,MAAM,CAACV,EAAES,CAAC,CAAC,CAAC,IAAIqmC,GAAG9c,EAAE,CAAC,cAAc6c,EAAE,CAAC,EAAE,SAASE,GAAGlpC,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,SAAS,aAAa,EAAE,EAAEO,EAAE,KAAK,EAAEA,EAAE,KAAK,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,+DAA+D,IAAI,EAAE,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,gDAAgD,GAAG,EAAEL,EAAEA,GAAG,KAAK,OAAO,EAAE,IAAIgC,EAAE,CAAC,OAAO,IAAI,EAAEg6B,EAAE37B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEA,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,KAAKL,EAAE,WAAWI,CAAC,EAAEsC,EAAEsoB,EAAE,UAAU5J,GAAGpf,EAAE,CAAC,EAAE,OAAO,IAAI,EAAEg6B,EAAEt5B,EAAE,CAACA,EAAE,IAAI,CAAC,EAAEA,CAAC,CAAC,IAAIumC,GAAGhd,EAAE,CAAC,aAAa+c,EAAE,CAAC,EAAE,SAASE,GAAGppC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,WAAW,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,WAAW,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUzJ,GAAGlhB,CAAC,CAAC,CAAC,IAAI8oC,GAAGld,EAAE,CAAC,UAAUid,EAAE,CAAC,EAAE,SAASE,GAAGtpC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,CAAC,EAAE,OAAOkrB,EAAE,UAAUrJ,GAAG3hB,CAAC,CAAC,CAAC,IAAIqpC,GAAGpd,EAAE,CAAC,UAAUmd,EAAE,CAAC,EAAE,SAASE,GAAGxpC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,KAAK,cAAc,EAAEM,EAAE0rB,EAAE,EAAE,KAAK,cAAc,EAAElU,EAAE5X,EAAE,OAAO,GAAGI,EAAE,OAAO,EAAE,IAAI,+DAA+DJ,EAAE,YAAYI,EAAE,OAAO,EAAE,IAAIC,EAAE27B,EAAEh8B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE,EAAEg8B,EAAE57B,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,OAAOwyB,GAAGvyB,EAAE,CAAC,CAAC,CAAC,IAAIkpC,GAAGtd,EAAE,CAAC,cAAcqd,EAAE,CAAC,EAAE,SAASE,GAAG1pC,EAAE,EAAEE,EAAE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,KAAK,EAAE,GAAGM,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,oDAAoD,EAAE,IAAIC,EAAE,CAAC,SAAS,EAAE,cAAcL,CAAC,EAAE,EAAE,CAAC,EAAEI,CAAC,EAAE,OAAO4qB,EAAE,UAAUlJ,GAAG,EAAEzhB,CAAC,CAAC,CAAC,IAAIopC,GAAGxd,EAAE,CAAC,KAAKud,EAAE,CAAC,EAAE,SAASE,GAAG5pC,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAO4X,EAAE,EAAE,SAAS,EAAE,IAAI,kDAAkD,EAAE6xB,GAAG3pC,EAAE,CAAC,CAAC,EAAEE,CAAC,CAAC,CAAC,IAAI2pC,GAAG1d,EAAE,CAAC,OAAOyd,EAAE,CAAC,EAAE,SAASE,GAAG9pC,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAO4X,EAAE,EAAE,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,EAAE,IAAI,uDAAuD,EAAE6xB,GAAG3pC,EAAE,EAAEE,CAAC,CAAC,CAAC,IAAI6pC,GAAG5d,EAAE,CAAC,OAAO2d,EAAE,CAAC,EAAE,SAASE,GAAGhqC,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAO4X,EAAE,EAAE,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,EAAE,IAAI,uDAAuD,EAAE6xB,GAAG3pC,EAAE,EAAEE,CAAC,CAAC,CAAC,IAAI+pC,GAAG9d,EAAE,CAAC,OAAO6d,EAAE,CAAC,EAAE,SAASE,GAAGlqC,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAO4X,EAAE,EAAE,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,EAAE,IAAI,uDAAuD,EAAE6xB,GAAG3pC,EAAE,EAAEE,CAAC,CAAC,CAAC,IAAIiqC,GAAGhe,EAAE,CAAC,OAAO+d,EAAE,CAAC,EAAE,SAASE,GAAGpqC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,gBAAgB,EAAE8X,EAAExX,EAAE,MAAM,EAAE,EAAE,OAAO,IAAI,cAAcA,EAAE,sCAAsC,EAAE,QAAQ,EAAEwX,EAAE5X,EAAE,SAAS,EAAE,OAAO,IAAI,qBAAqBA,EAAE,wCAAwC,EAAE,QAAQ,EAAE4X,EAAExX,EAAE,MAAM,OAAO,CAAC,EAAE2B,EAAEC,IAAIA,EAAE,GAAGA,GAAG,EAAE,OAAO,IAAID,EAAE/B,EAAEgC,EAAE,GAAG,GAAGhC,EAAEgC,EAAE,GAAG,IAAI,EAAEA,EAAE,KAAK,EAAE,EAAE,EAAE,EAAE,IAAI,4BAA4B5B,EAAE,MAAM,MAAM,CAAC,mBAAmBJ,EAAE,SAAS,sCAAsC,EAAE,SAAS,GAAG,EAAE,IAAIK,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,SAASJ,CAAC,EAAE,OAAOgrB,EAAE,UAAUjH,GAAG1jB,EAAE,CAAC,CAAC,CAAC,IAAI8pC,GAAGle,EAAE,CAAC,gBAAgBie,EAAE,CAAC,EAAE,SAASE,GAAGtqC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAG,GAAG,OAAO,EAAE,GAAGD,IAAI,IAAIA,EAAE,SAAS,IAAI2B,EAAE+pB,EAAEhsB,EAAE,IAAI,SAAS,EAAEkC,EAAED,EAAEE,EAAE,GAAGF,EAAE,OAAO,IAAIE,EAAE,GAAGD,EAAEg6B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG6V,EAAEikB,GAAG,EAAEx7B,CAAC,EAAE,IAAI,qEAAqE,oBAAoBA,IAAI,EAAE,IAAIqC,EAAEs4B,GAAGh5B,EAAE,MAAM,EAAE,EAAE3B,EAAED,CAAC,EAAEuC,EAAE,CAACD,EAAE,eAAeA,EAAE,aAAa,EAAEE,EAAExC,IAAI,OAAOwC,EAAEynC,GAAG,CAAC3nC,EAAE,aAAaA,EAAE,WAAW,EAAEC,CAAC,EAAEC,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAE,IAAIC,EAAEF,EAAE,KAAK,GAAGA,EAAE,KAAK,EAAE,CAACG,EAAEkB,CAAC,EAAEsmC,GAAG,CAAC5nC,EAAE,SAASA,EAAE,OAAO,EAAEC,EAAEC,CAAC,EAAE,EAAEC,EAAEzC,EAAE,QAAQiH,EAAExE,EAAEb,EAAEmoC,GAAGnoC,EAAEW,EAAEG,CAAC,EAAE2B,GAAGzE,IAAI,MAAM,IAAIk8B,GAAG70B,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,IAAIqgC,GAAGrgC,EAAE,EAAE,EAAE,EAAE,CAAC,GAAG,EAAE3C,EAAE7B,EAAE4B,EAAEu4B,GAAGv4B,EAAE9B,EAAEqB,CAAC,EAAE,OAAO/B,EAAE+5B,EAAEt3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,SAAS4lC,GAAGxqC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,EAAE,IAAI0C,GAAGA,EAAE,EAAE,EAAErC,EAAEL,EAAE,IAAI0C,GAAGA,EAAE,EAAE,EAAE,EAAE5C,EAAE,OAAOM,EAAEC,CAAC,EAAE,EAAE,EAAE,IAAI,CAACqC,EAAEC,KAAKD,EAAE,EAAEC,GAAGD,GAAGA,CAAC,EAAEX,EAAE1B,EAAE,IAAI,CAACqC,EAAEC,IAAID,EAAE,EAAEC,EAAE,EAAEX,EAAE,EAAE,IAAI,CAACU,EAAEC,IAAI,CAACvC,EAAEuC,GAAGZ,EAAEY,EAAE,CAAC,EAAEV,EAAE,EAAE,IAAI,CAACS,EAAEC,IAAI,CAAC,EAAE,EAAEA,EAAE,CAAC,EAAE,MAAM,CAACX,EAAEC,CAAC,CAAC,CAAC,SAASooC,GAAGvqC,EAAE,EAAE,CAAC,IAAI,EAAEA,EAAE,IAAI,CAACgC,EAAEC,IAAID,GAAGA,EAAE,IAAI,EAAEC,GAAG,EAAE,EAAE,IAAID,GAAGA,EAAE,CAAC,EAAEzB,EAAE,EAAE,IAAIyB,GAAG,KAAK,MAAMA,EAAE,CAAC,CAAC,EAAED,EAAE,EAAE,IAAI,CAACC,EAAEC,IAAID,EAAEzB,EAAE0B,EAAE,EAAE,OAAO,EAAE,IAAI,CAACD,EAAEC,IAAI,CAAC1B,EAAE0B,GAAGF,EAAEE,EAAE,CAAC,CAAC,CAAC,IAAIwoC,GAAGte,EAAE,CAAC,MAAMme,EAAE,CAAC,EAAE,SAASI,GAAG1qC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,EAAEM,EAAE0rB,EAAE,EAAE,QAAQ,OAAO,EAAEzrB,EAAE,CAAC,EAAEL,EAAE,MAAMI,CAAC,EAAE,OAAO4qB,EAAE,UAAU/I,GAAG5hB,CAAC,CAAC,CAAC,IAAIoqC,GAAGxe,EAAE,CAAC,OAAOue,EAAE,CAAC,EAAE,SAASE,GAAG5qC,EAAE,EAAE,KAAKE,EAAE,GAAG,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,MAAM,EAAEM,EAAE,QAAQ,SAASA,EAAEmwB,GAAEnwB,EAAE,OAAO,GAAG,IAAIC,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAASJ,CAAC,EAAE,OAAOgrB,EAAE,UAAU9I,GAAG7hB,EAAE,CAAC,CAAC,CAAC,IAAIsqC,GAAG1e,EAAE,CAAC,MAAMye,EAAE,CAAC,EAAE,SAASE,GAAG9qC,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEP,EAAE,IAAI,CAAC4C,EAAEC,IAAImpB,EAAEppB,EAAE,UAAUC,IAAI,eAAe,OAAO,CAAC,EAAE,EAAEmpB,EAAE,EAAE,oBAAoB,cAAc,EAAE,EAAEA,EAAE9rB,EAAE,UAAU,eAAe,OAAO,EAAE+B,EAAE,CAAC,mBAAmB1B,EAAE,kBAAkB,EAAE,QAAQ,CAAC,EAAE2B,EAAE,CAAC,iBAAiB5B,CAAC,EAAE6B,EAAE+oB,EAAE,UAAU7I,GAAGpgB,EAAEC,CAAC,EAAE,MAAM,CAAC,mBAAmBC,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,EAAE,kBAAkBA,EAAEA,EAAE,OAAO,EAAE,CAAC,CAAC,IAAI4oC,GAAG5e,EAAE,CAAC,cAAc2e,EAAE,CAAC,EAAE,SAASE,GAAGhrC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,SAAS,aAAa,EAAEO,EAAEyrB,EAAE,EAAE,SAAS,cAAc1rB,EAAE,KAAK,EAAE,EAAE0rB,EAAE9rB,EAAE,SAAS,cAAcI,EAAE,KAAK,EAAE,EAAE,CAAC,OAAOA,EAAE,OAAOC,EAAE,OAAO,CAAC,EAAE0B,EAAEipB,EAAE,UAAU5I,GAAG,CAAC,EAAE,MAAM,CAAC,eAAergB,EAAE,GAAG,cAAcA,EAAE,EAAE,CAAC,CAAC,IAAIgpC,GAAG9e,EAAE,CAAC,aAAa6e,EAAE,CAAC,EAAE,SAASE,GAAGlrC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,uBAAuB,OAAO,EAAE,EAAEgsB,EAAE,EAAE,SAAS,sBAAsB,EAAE/pB,EAAE+pB,EAAE9rB,EAAE,eAAe,uBAAuB,EAAE,KAAK,EAAEgC,EAAE5B,EAAE,IAAI,CAACuC,EAAEC,IAAIkpB,EAAEnpB,EAAE,UAAUC,IAAI,uBAAuB,OAAO,CAAC,EAAEX,EAAE,CAAC,MAAM,EAAE,OAAO,EAAE,aAAaF,EAAE,oBAAoBC,CAAC,EAAEU,EAAE,CAAC,kBAAkBrC,CAAC,EAAE,OAAO2qB,EAAE,UAAU3I,GAAGpgB,EAAES,CAAC,CAAC,CAAC,IAAIuoC,GAAGhf,EAAE,CAAC,sBAAsB+e,EAAE,CAAC,EAAE,SAASE,GAAGprC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE8X,GAAGpY,CAAC,EAAEO,EAAE,KAAK,GAAGL,GAAG,MAAMA,IAAI,UAAUK,EAAE,IAAI,aAAaD,CAAC,UAAUJ,IAAI,QAAQK,EAAE,IAAI,WAAWD,CAAC,UAAUJ,IAAI,OAAOK,EAAE,IAAI,WAAWD,CAAC,MAAO,OAAM,IAAI,MAAM,qBAAqBJ,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAEI,EAAE,IAAIC,EAAE,GAAG,EAAE,EAAE,OAAO2qB,EAAE,WAAW3qB,EAAEP,EAAEE,CAAC,CAAC,CAAC,IAAImrC,GAAGlf,EAAE,CAAC,MAAMif,EAAE,CAAC,EAAME,GAAG9qC,GAAG+E,GAAG,CAAC,EAAMgmC,GAAG,KAAK,CAAC,YAAYzqC,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,KAAKjB,EAAE,KAAK,OAAOZ,EAAE,KAAK,MAAM,EAAE,KAAK,QAAQ,IAAI,KAAK,UAAUK,EAAE,KAAK,YAAY,KAAK,MAAM,KAAK,KAAK,KAAK,OAAO,EAAE,KAAK,MAAM,KAAK,KAAK,KAAK,OAAO,GAAG,IAAIyB,EAAED,GAAG,KAAK,OAAO,EAAE,KAAK,OAAOupC,GAAG,KAAKtpC,EAAE,SAAS,CAAC,CAAC,CAAC,WAAW,CAAC,GAAG,CAAC,MAAM,KAAK,OAAO,EAAE,CAAC,IAAIzB,EAAE,KAAK,QAAQ,OAAO,KAAK,QAAQ,IAAIA,CAAC,CAAC,IAAIO,EAAEZ,EAAE,EAAE,GAAG,KAAK,CAAC,GAAG,CAAC,IAAIK,EAAEwB,EAAEC,EAAE,GAAGzB,EAAE,EAAE,KAAK,OAAO,EAAE,EAAEwB,EAAE,EAAE,KAAK,OAAO,EAAE,EAAEC,EAAEzB,EAAEA,EAAEwB,EAAEA,QAAQC,GAAG,GAAGA,IAAI,GAAG,IAAIC,EAAE,KAAK,KAAK,GAAG,KAAK,IAAID,CAAC,EAAEA,CAAC,EAAElB,EAAE,KAAK,KAAK,KAAK,OAAOP,EAAE0B,EAAE/B,EAAE,KAAK,KAAK,KAAK,OAAO6B,EAAEE,GAAG,CAAC,KAAK,WAAW,KAAK,iBAAiBnB,CAAC,KAAK,EAAE,GAAG,CAAC,OAAO,CAAC,KAAK,WAAW,KAAK,iBAAiBZ,CAAC,KAAK,KAAK,QAAQ,KAAK,aAAaA,CAAC,GAAG,KAAK,aAAaY,CAAC,CAAC,CAAC,aAAaA,EAAE,CAAC,OAAO,KAAK,OAAO,MAAM,KAAK,QAAQ,UAAUA,EAAE,KAAK,MAAMA,CAAC,CAAC,CAAC,iBAAiBA,EAAE,CAAC,OAAOA,GAAG,KAAK,OAAOA,GAAG,KAAK,KAAK,CAAC,EAAE0qC,GAAG,KAAK,CAAC,YAAY1qC,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,MAAMO,EAAE,KAAK,KAAK,EAAEZ,EAAE,KAAK,MAAM,EAAE,IAAI6B,EAAExB,GAAG,KAAK,OAAO,EAAE,KAAK,MAAM+qC,GAAG,KAAKvpC,EAAE,SAAS,CAAC,EAAE,KAAK,MAAM,IAAIwpC,GAAG,EAAE,EAAE,EAAE,GAAG,KAAK,MAAM,CAAC,EAAEzqC,EAAE,EAAE,KAAK,EAAEA,EAAE,EAAE,EAAE,KAAK,EAAEA,EAAE,EAAE,EAAE,KAAK,EAAE,EAAE,KAAK,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,OAAO,CAAC,GAAGzB,EAAE,KAAK,MAAM,UAAU,EAAEyB,EAAE,EAAE,KAAK,EAAEzB,QAAQyB,GAAG,GAAG,GAAGA,GAAGA,EAAEA,EAAElB,EAAEP,EAAEA,EAAEL,EAAE,EAAE,KAAKY,EAAEA,EAAE,EAAE,GAAGA,EAAE,KAAK,GAAG,EAAEkB,EAAE,KAAK,IAAIA,CAAC,GAAGD,EAAE,KAAK,MAAM,EAAEA,EAAE7B,GAAG,KAAK,IAAI6B,CAAC,EAAE,EAAE,KAAK,CAAC,OAAOC,EAAE,EAAE,KAAK,KAAK,KAAK,EAAEA,EAAE,KAAK,MAAM,IAAIA,GAAG,KAAK,IAAI,KAAK,MAAM,EAAE,EAAE,KAAK,KAAK,GAAG,KAAK,aAAaA,CAAC,CAAC,CAAC,aAAalB,EAAE,CAAC,OAAO,KAAK,QAAQ,UAAUA,EAAE,KAAK,MAAMA,CAAC,CAAC,CAAC,EAAE2qC,GAAG,KAAK,CAAC,YAAY3qC,EAAE,EAAEZ,EAAE,EAAE,EAAEK,EAAE,CAAC,GAAG,KAAK,eAAe,IAAI,KAAK,OAAO,MAAM,KAAK,QAAQ,UAAU,KAAK,IAAIO,EAAE,KAAK,MAAMZ,EAAEY,EAAE,KAAK,MAAM,EAAEP,GAAG,OAAOA,EAAE,KAAK,OAAO,GAAG,OAAOA,GAAG,WAAWA,EAAEA,EAAE,SAAS,GAAG,CAAC,KAAK,eAAe,GAAG,KAAK,OAAO,EAAE,MAAM,IAAI,MAAM,0BAA0BO,OAAOZ,+BAA+B,EAAE,KAAK,OAAOorC,GAAG,KAAK/qC,CAAC,CAAC,CAAC,aAAaO,EAAE,CAAC,OAAO,KAAK,eAAe,EAAEA,EAAE,KAAK,MAAMA,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,KAAK,aAAa,KAAK,IAAI,KAAK,MAAM,KAAK,OAAO,CAAC,CAAC,CAAC,EAAE,SAAS4qC,GAAG1rC,EAAE,EAAEE,EAAE,EAAEI,EAAE,UAAUC,EAAE,CAAC,GAAGL,GAAG,OAAOA,EAAE,GAAGI,GAAG,OAAOA,EAAE,WAAWA,IAAI,WAAWA,IAAI,QAAQ,MAAM,IAAI,MAAM,yBAAyBA,GAAG,EAAE,IAAI,EAAE,IAAIkrC,GAAG,EAAEtrC,EAAEI,EAAEC,CAAC,EAAE,EAAEgwB,GAAGvwB,EAAEM,CAAC,EAAE,QAAQ2B,EAAE,EAAEA,EAAE,EAAE,OAAO,OAAOA,IAAI,EAAE,OAAOA,GAAG,EAAE,UAAU,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC,IAAI0pC,GAAGxf,EAAE,CAAC,aAAauf,EAAE,CAAC,EAAE,SAASE,GAAG5rC,EAAE,EAAE,EAAEE,EAAE,EAAEI,EAAEC,EAAE,CAAC,GAAGD,GAAG,MAAMA,IAAI,OAAO,MAAM,IAAI,MAAM,yBAAyBA,GAAG,EAAE,IAAI,EAAE,IAAIirC,GAAG,EAAErrC,EAAEI,EAAE,GAAGC,CAAC,EAAE,EAAEgwB,GAAGvwB,EAAEM,CAAC,EAAE,QAAQ2B,EAAE,EAAEA,EAAE,EAAE,OAAO,OAAOA,IAAI,EAAE,OAAOA,GAAG,EAAE,UAAU,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC,IAAI4pC,GAAG1f,EAAE,CAAC,cAAcyf,EAAE,CAAC,EAAE,SAASE,GAAG9rC,EAAE,EAAEE,EAAE,CAAC,GAAG,GAAG,MAAM,IAAI,OAAO,MAAM,IAAI,MAAM,yBAAyB,GAAG,EAAE,OAAO2rC,GAAG7rC,EAAE,EAAE,EAAE,EAAEE,CAAC,CAAC,CAAC,IAAI6rC,GAAG5f,EAAE,CAAC,sBAAsB2f,EAAE,CAAC,EAAE,SAASE,GAAGhsC,EAAE,EAAE,EAAEE,EAAE,EAAEI,EAAE,UAAUC,EAAE,CAAC,IAAI,EAAEgwB,GAAGvwB,EAAEM,CAAC,EAAE,EAAE,IAAImrC,GAAG,EAAEvrC,EAAE,KAAKK,CAAC,EAAE,QAAQ0B,EAAE,EAAEA,EAAE,EAAE,OAAO,OAAOA,IAAI,EAAE,OAAOA,GAAG,EAAE,UAAU,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC,IAAIgqC,GAAG9f,EAAE,CAAC,eAAe6f,EAAE,CAAC,EAAE,SAASE,GAAGlsC,EAAE,EAAEE,EAAE,EAAEI,EAAE,UAAU,CAAC,GAAGJ,IAAI,EAAE,MAAM,IAAI,MAAM,4BAA4B,EAAE,IAAIK,EAAE,CAAC,MAAMP,EAAE,KAAK,EAAE,KAAKE,EAAE,MAAMI,CAAC,EAAE,OAAO4qB,EAAE,UAAU1I,GAAG,CAAC,EAAEjiB,CAAC,CAAC,CAAC,SAAS4rC,GAAGnsC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,YAAY,CAAC,EAAE,OAAOkrB,EAAE,UAAUxI,GAAGxiB,CAAC,CAAC,CAAC,IAAIksC,GAAGjgB,EAAE,CAAC,YAAYggB,EAAE,CAAC,EAAE,SAASE,GAAGrsC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUvI,GAAGziB,CAAC,CAAC,CAAC,IAAIosC,GAAGngB,EAAE,CAAC,MAAMkgB,EAAE,CAAC,EAAE,SAASE,GAAGvsC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAUjI,GAAG/iB,CAAC,CAAC,CAAC,IAAIssC,GAAGrgB,EAAE,CAAC,OAAOogB,EAAE,CAAC,EAAE,SAASE,GAAGzsC,EAAE,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,SAAS,CAAC,EAAEO,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUhI,GAAG,EAAE3iB,CAAC,CAAC,CAAC,IAAImsC,GAAGvgB,EAAE,CAAC,SAASsgB,EAAE,CAAC,EAAE,SAASE,GAAG3sC,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAE,EAAE,OAAO,EAAE,IAAI,qDAAqD,EAAE,OAAO,EAAE40B,GAAG,EAAE,CAAC,CAAC,CAAC,IAAIE,GAAGzgB,EAAE,CAAC,WAAWwgB,EAAE,CAAC,EAAE,SAASE,GAAG7sC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAE5X,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,OAAO,EAAEwsC,GAAGxsC,EAAE,CAAC,CAAC,CAAC,IAAI4sC,GAAG3gB,EAAE,CAAC,WAAW0gB,EAAE,CAAC,EAAE,SAASE,GAAG/sC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAE5X,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,OAAO,EAAEwsC,GAAGxsC,EAAE,CAAC,CAAC,CAAC,IAAI8sC,GAAG7gB,EAAE,CAAC,WAAW4gB,EAAE,CAAC,EAAE,SAASE,GAAGjtC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAE5X,EAAE,OAAO,EAAE,IAAI,qDAAqDA,EAAE,OAAO,EAAEwsC,GAAGxsC,EAAE,CAAC,CAAC,CAAC,IAAIgtC,GAAG/gB,EAAE,CAAC,WAAW8gB,EAAE,CAAC,EAAE,SAASE,GAAGntC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE,OAAOkrB,EAAE,UAAU/H,GAAGjjB,CAAC,CAAC,CAAC,IAAIktC,GAAGjhB,EAAE,CAAC,OAAOghB,EAAE,CAAC,EAAE,SAASE,GAAGrtC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,QAAQ,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAU9H,GAAGljB,CAAC,CAAC,CAAC,IAAIotC,GAAGnhB,EAAE,CAAC,OAAOkhB,EAAE,CAAC,EAAE,SAASE,GAAGvtC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAU1H,GAAGtjB,CAAC,CAAC,CAAC,IAAIstC,GAAGrhB,EAAE,CAAC,MAAMohB,EAAE,CAAC,EAAE,SAASE,GAAGztC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,OAAO,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,IAAI,iBAAiB,EAAEkC,EAAE8pB,EAAE,EAAE,kBAAkB,iBAAiB,EAAE7pB,EAAE6pB,EAAE9rB,EAAE,kBAAkB,iBAAiB,EAAE0C,EAAEX,EAAEY,EAAE,GAAG,GAAGZ,EAAE,OAAO,IAAIY,EAAE,GAAGD,EAAEs5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI,OAAO,MAAM,IAAI,MAAM,oFAAoF,EAAE6V,EAAElV,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,OAAO,EAAEkV,EAAE5V,EAAE,OAAO,EAAE,IAAI,2EAA2EA,EAAE,OAAO,EAAE4V,EAAE3V,EAAE,OAAO,EAAE,IAAI,2EAA2ED,EAAE,OAAO,EAAE4V,EAAE3V,EAAE,MAAM,KAAK,EAAE,IAAI,yFAAyFA,EAAE,MAAM,KAAK,EAAE2V,EAAE3V,EAAE,MAAM,KAAK,EAAE,IAAI,yFAAyFA,EAAE,MAAM,KAAK,EAAE,IAAIW,EAAEZ,EAAE,MAAM,GAAGa,EAAEb,EAAE,MAAM,GAAG4V,EAAE3V,EAAE,MAAM,KAAKW,EAAEC,EAAE,IAAI,6EAA6ED,EAAEC,cAAcZ,EAAE,MAAM,KAAK,EAAE,IAAIa,EAAE09B,GAAG99B,EAAEV,EAAE5B,EAAEC,EAAE,EAAE,CAAC,EAAE4D,EAAE66B,GAAGh8B,EAAEb,EAAE,EAAE,QAAQ,CAAC,EAAE,OAAOU,EAAEq5B,EAAE/3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIupC,GAAGvhB,EAAE,CAAC,iBAAiBshB,EAAE,CAAC,EAAE,eAAeE,GAAG3tC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,WAAW,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,WAAW,EAAElU,EAAE5X,EAAE,QAAQI,EAAE,MAAM,IAAI,kDAAkDJ,EAAE,iBAAiBI,EAAE,SAAS,EAAEwX,EAAE5X,EAAE,OAAO,EAAE,IAAI,qCAAqCA,EAAE,SAAS,EAAE4X,EAAExX,EAAE,OAAO,EAAE,IAAI,qCAAqCA,EAAE,SAAS,EAAE,IAAIC,EAAE,MAAML,EAAE,KAAK,EAAE,EAAE,MAAMI,EAAE,KAAK,EAAE,EAAE,IAAI,IAAI,CAAC,EAAE2B,EAAE,EAAE,QAAQW,EAAE,EAAEA,EAAErC,EAAE,OAAOqC,IAAI,EAAE,IAAIrC,EAAEqC,EAAE,GAAGX,IAAI,IAAIC,EAAE,IAAIonB,GAAG,CAACrnB,CAAC,EAAE/B,EAAE,KAAK,EAAEiC,EAAE,IAAImnB,GAAG,CAACrnB,CAAC,EAAE,OAAO,EAAE,QAAQW,EAAE,EAAEC,EAAE,EAAED,EAAErC,EAAE,OAAOqC,IAAI,EAAE,IAAIrC,EAAEqC,EAAE,IAAIV,EAAE,OAAOW,GAAGtC,EAAEqC,GAAGT,EAAE,OAAOU,GAAGD,EAAEC,KAAK,MAAM,CAACX,EAAE,SAAS,EAAEC,EAAE,SAAS,CAAC,CAAC,CAAC,IAAIyrC,GAAGD,GAAG,SAASE,GAAG7tC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUtH,GAAG1jB,CAAC,CAAC,CAAC,IAAI4tC,GAAG3hB,EAAE,CAAC,MAAM0hB,EAAE,CAAC,EAAE,SAASE,GAAG/tC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUxH,GAAGxjB,CAAC,CAAC,CAAC,IAAI8tC,GAAG7hB,EAAE,CAAC,KAAK4hB,EAAE,CAAC,EAAE,SAASE,GAAGjuC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOkrB,EAAE,UAAUvH,GAAGzjB,CAAC,CAAC,CAAC,IAAIguC,GAAG/hB,EAAE,CAAC,MAAM8hB,EAAE,CAAC,EAAE,SAASE,GAAGnuC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAExX,EAAE,OAAO,EAAE,IAAI,mDAAmDA,EAAE,aAAa,EAAEs8B,GAAGt8B,EAAE,CAAC,CAAC,EAAE,CAACJ,CAAC,CAAC,CAAC,CAAC,IAAIkuC,GAAGjiB,EAAE,CAAC,SAASgiB,EAAE,CAAC,EAAE,SAASE,GAAGruC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAExX,EAAE,OAAO,EAAE,IAAI,mDAAmDA,EAAE,aAAa,EAAEs8B,GAAGt8B,EAAE,EAAEJ,CAAC,CAAC,CAAC,IAAIouC,GAAGniB,EAAE,CAAC,SAASkiB,EAAE,CAAC,EAAE,SAASE,GAAGvuC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAExX,EAAE,OAAO,EAAE,IAAI,mDAAmDA,EAAE,aAAa,EAAEs8B,GAAGt8B,EAAE,EAAEJ,CAAC,CAAC,CAAC,IAAIsuC,GAAGriB,EAAE,CAAC,SAASoiB,EAAE,CAAC,EAAE,SAASE,GAAGzuC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,OAAO8X,EAAExX,EAAE,OAAO,EAAE,IAAI,mDAAmDA,EAAE,aAAa,EAAEs8B,GAAGt8B,EAAE,EAAEJ,CAAC,CAAC,CAAC,IAAIwuC,GAAGviB,EAAE,CAAC,SAASsiB,EAAE,CAAC,EAAE,SAASE,GAAG3uC,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,SAAS,UAAU,SAAS,EAAE,GAAG,IAAI,KAAK,EAAEE,EAAE,KAAK,GAAG,IAAIA,EAAE,KAAK,EAAE,MAAM,MAAM,4EAA4EA,EAAE,oBAAoB,GAAG,EAAE,IAAII,EAAE,CAAC,OAAOJ,CAAC,EAAEK,EAAE,CAAC,IAAI,CAAC,EAAE,OAAO2qB,EAAE,UAAU/G,GAAG7jB,EAAEC,CAAC,CAAC,CAAC,IAAIquC,GAAGziB,EAAE,CAAC,SAASwiB,EAAE,CAAC,EAAE,SAASE,GAAG7uC,EAAE,CAAC8X,EAAE9X,EAAE,QAAQ,YAAY,IAAI,6DAA6DA,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAOkrB,EAAE,UAAUtM,GAAG,CAAC,CAAC,CAAC,IAAIkwB,GAAG3iB,EAAE,CAAC,KAAK0iB,EAAE,CAAC,EAAE,SAASE,GAAG/uC,EAAE,CAAC8X,EAAE9X,EAAE,QAAQ,YAAY,IAAI,8DAA8DA,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAOkrB,EAAE,UAAU3L,GAAG,CAAC,CAAC,CAAC,IAAIyvB,GAAG7iB,EAAE,CAAC,MAAM4iB,EAAE,CAAC,EAAE,SAASE,GAAGjvC,EAAE,CAAC,IAAI,EAAEA,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAGE,EAAEF,EAAE,KAAK,EAAEM,EAAE,GAAG,GAAG,EAAE,CAAC,IAAIC,EAAE27B,EAAEl8B,EAAE,CAACE,EAAE,CAAC,CAAC,EAAEI,EAAE0uC,GAAGzuC,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAE,CAACL,EAAE,GAAG,EAAE,EAAE,EAAE,EAAEg8B,EAAEvH,GAAG30B,CAAC,EAAE,CAACE,EAAE,CAAC,CAAC,EAAE,EAAEg8B,EAAE3H,GAAGv0B,CAAC,EAAE,CAACE,EAAE,CAAC,CAAC,EAAE+B,EAAEyqC,GAAG9P,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC18B,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAEgC,EAAEk3B,EAAEsT,GAAG9P,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC18B,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE2iC,GAAG,EAAE,CAAC,EAAE1gC,EAAEq6B,GAAG,CAAC,EAAEv6B,CAAC,EAAE,CAAC,EAAEW,EAAE45B,GAAG,CAAC,EAAEt6B,CAAC,EAAE,CAAC,EAAEW,EAAEq5B,EAAE7P,GAAGlqB,EAAES,CAAC,EAAE,CAACrC,EAAE,GAAGA,EAAE,EAAE,CAAC,EAAED,EAAE0uC,GAAGnsC,CAAC,CAAC,CAAC,GAAGvC,EAAEq0B,GAAGr0B,CAAC,EAAEN,EAAE,OAAO,GAAGA,EAAE,MAAM,KAAK,EAAE,CAAC,IAAIO,EAAED,EAAE,EAAEN,EAAE,MAAM,GAAGM,EAAE47B,EAAE57B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,EAAE,CAAC,EAAEC,EAAE,QAAQ,CAAC,CAAC,OAAOD,CAAC,CAAC,IAAI4uC,GAAG/iB,EAAE,CAAC,OAAO8iB,EAAE,CAAC,EAAE,SAASE,GAAGnvC,EAAE,EAAEE,EAAE,EAAE,CAAC,IAAIK,EAAE,CAAC,EAAEyrB,EAAEhsB,EAAE,IAAI,OAAO,CAAC,EAAE+B,EAAE,CAAC,gBAAgB,EAAE,KAAK7B,CAAC,EAAE,OAAOgrB,EAAE,UAAUhH,GAAG3jB,EAAEwB,CAAC,CAAC,CAAC,IAAIqtC,GAAGjjB,EAAE,CAAC,OAAOgjB,EAAE,CAAC,EAAE,SAASE,GAAGrvC,EAAE,EAAE,CAAC8X,EAAE9X,EAAE,QAAQ,UAAU,IAAI,mDAAmDA,EAAE,OAAO,EAAE,IAAIE,EAAEF,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAGM,EAAEN,EAAE,KAAKE,EAAEK,EAAE,GAAG,GAAG,MAAM,EAAEL,EAAE,CAAC,IAAI8C,EAAEhD,EAAE,MAAM,IAAI,GAAG,CAAC,EAAEkE,EAAElE,EAAE,MAAM,IAAI,GAAG,CAAC,EAAEkE,EAAElE,EAAE,MAAM,OAAO,GAAG,EAAEO,EAAEq8B,GAAG58B,EAAEgD,EAAEkB,CAAC,EAAEhE,EAAE,CAAC,SAAS,GAAG,MAAM,EAAEA,EAAE,CAAC,IAAI8C,EAAEhD,EAAE,MAAM,IAAIkE,GAAGA,CAAC,EAAElB,EAAEhD,EAAE,MAAM,OAAO,GAAG,EAAEE,EAAEK,EAAEi8B,GAAG,CAACx8B,EAAEqoC,GAAGrlC,CAAC,CAAC,EAAEhD,EAAE,MAAM,OAAO,CAAC,EAAEE,EAAE,CAAC,MAAMK,EAAEP,EAAE,IAAI,EAAEohC,GAAG7gC,CAAC,EAAE,EAAE27B,EAAE7P,GAAG9rB,EAAE,CAAC,EAAE,CAACD,EAAEJ,CAAC,CAAC,EAAE+B,EAAE6sC,GAAG,CAAC,EAAE5sC,EAAE,KAAK,MAAMhC,EAAE,CAAC,EAAE,EAAEiC,EAAEwyB,GAAG1yB,CAAC,EAAEW,EAAE2xB,GAAGtyB,CAAC,EAAEY,EAAEusC,GAAGjtC,EAAE,CAACD,EAAEhC,EAAEgC,CAAC,EAAEC,EAAE,MAAM,OAAO,CAAC,EAAEW,EAAEssC,GAAGxsC,EAAE,CAACV,EAAEhC,EAAEgC,CAAC,EAAEU,EAAE,MAAM,OAAO,CAAC,EAAEG,EAAExC,EAAE,MAAM,MAAM,EAAE,OAAOwC,EAAExC,EAAE,MAAM,OAAO,GAAG2B,EAAEg6B,EAAE7P,GAAGxpB,EAAE,GAAGC,EAAE,EAAE,EAAEC,CAAC,CAAC,CAAC,IAAIusC,GAAGnjB,EAAE,CAAC,MAAMkjB,EAAE,CAAC,EAAE,SAASE,GAAGvvC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,mBAAmB,EAAEM,EAAE0rB,EAAE,EAAE,IAAI,mBAAmB,EAAE,CAAC9rB,EAAEI,CAAC,EAAE8pB,GAAGlqB,EAAEI,CAAC,EAAE00B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAEL,EAAE,EAAEI,CAAC,EAAE,EAAE,CAAC,EAAE,OAAO4qB,EAAE,UAAUzG,GAAGlkB,EAAE,CAAC,CAAC,CAAC,IAAIivC,GAAGrjB,EAAE,CAAC,mBAAmBojB,EAAE,CAAC,EAAE,SAASE,GAAGzvC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,mBAAmB,EAAE,OAAOk8B,EAAEh8B,EAAE4Y,GAAG5Y,EAAE,MAAM,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAIwvC,GAAGvjB,EAAE,CAAC,SAASsjB,EAAE,CAAC,EAAE,SAASE,GAAG3vC,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE+rB,GAAGjsB,EAAE,UAAU,QAAQ,mBAAmB,EAAE8X,EAAE5X,EAAE,QAAQ,EAAE,IAAI,sCAAsC,EAAEA,EAAE,OAAO,GAAG4X,EAAE,GAAG5X,EAAE,GAAG,KAAK,IAAI,oCAAoC,EAAE,IAAII,EAAEJ,EAAEK,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUnJ,GAAGzhB,EAAEC,CAAC,CAAC,CAAC,IAAIqvC,GAAGzjB,EAAE,CAAC,OAAOwjB,EAAE,CAAC,EAAE,SAASE,GAAG7vC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEgsB,EAAEhsB,EAAE,IAAI,MAAM,CAAC,EAAEO,EAAE,CAAC,MAAM,CAAC,EAAE,OAAO2qB,EAAE,UAAUvF,GAAG,EAAEplB,CAAC,CAAC,CAAC,IAAIuvC,GAAG3jB,EAAE,CAAC,MAAM0jB,EAAE,CAAC,EAAE,SAASE,GAAG/vC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE0B,EAAE,EAAEC,EAAE,EAAE,CAAC,IAAIU,EAAE,CAAC,EAAEopB,EAAEhsB,EAAE,IAAI,eAAe,mBAAmB,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,IAAIE,EAAE,QAAQI,EAAE,UAAUC,EAAE,QAAQ,EAAE,aAAa,EAAE,YAAY0B,EAAE,eAAeC,CAAC,EAAE,OAAOgpB,EAAE,UAAUvG,GAAG/hB,EAAE,CAAC,CAAC,CAAC,IAAIotC,GAAG7jB,EAAE,CAAC,cAAc4jB,EAAE,CAAC,EAAE,SAASE,GAAGjwC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE8rB,EAAEhsB,EAAE,IAAI,MAAM,SAAS,CAAC,EAAE,OAAOkrB,EAAE,UAAUlG,GAAG9kB,CAAC,CAAC,CAAC,IAAIgwC,GAAG/jB,EAAE,CAAC,KAAK8jB,EAAE,CAAC,EAAE,SAASE,GAAGnwC,EAAE,EAAE,CAACiY,GAAGjY,CAAC,EAAE,IAAIE,EAAE2rB,GAAG7rB,EAAE,CAAC,EAAE,GAAGE,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,oDAAoD,EAAE,OAAOosB,GAAGtsB,EAAE,KAAKE,EAAE,CAAC,CAAC,CAAC,SAASkwC,GAAGpwC,EAAE,EAAEE,EAAE,CAAC,GAAG+X,GAAGjY,CAAC,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,+CAA+C,EAAE,IAAIM,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,GAAGI,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,gEAAgE,EAAE,GAAGA,EAAE,SAAS,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,8EAA8E,EAAE,OAAOgsB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,SAASmwC,GAAGrwC,EAAE,EAAEE,EAAE,CAAC,GAAG+X,GAAGjY,CAAC,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,gDAAgD,EAAE,IAAIM,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,GAAGI,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,oEAAoE,EAAE,GAAGA,EAAE,SAAS,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,yEAAyE,EAAE,OAAOgsB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,SAASowC,GAAGtwC,EAAE,EAAEE,EAAE,CAAC,GAAG+X,GAAGjY,CAAC,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,gDAAgD,EAAE,IAAIM,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,GAAGI,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,sEAAsE,EAAE,GAAGA,EAAE,SAAS,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,yEAAyE,EAAE,OAAOgsB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,SAASqwC,GAAGvwC,EAAE,EAAEE,EAAE,CAAC,GAAG+X,GAAGjY,CAAC,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,+CAA+C,EAAE,IAAIM,EAAEurB,GAAG7rB,EAAEE,CAAC,EAAE,GAAGI,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,wEAAwE,EAAE,GAAGA,EAAE,SAAS,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,yEAAyE,EAAE,OAAO,EAAE,GAAGA,EAAEgsB,GAAGtsB,EAAE,EAAEM,EAAEJ,CAAC,CAAC,CAAC,SAASswC,GAAGxwC,EAAE,EAAE,EAAEE,EAAE,GAAG,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,MAAM,EAAE,GAAGM,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,oDAAoD,EAAE,IAAIC,EAAED,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,6CAA6C,GAAG,EAAE,GAAG,EAAEC,EAAE,MAAM,IAAI,MAAM,uDAAuDA,cAAc,GAAG,EAAE,IAAI,EAAE,CAAC,EAAED,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,OAAOJ,CAAC,EAAE,CAAC+B,EAAEC,CAAC,EAAEgpB,EAAE,UAAU/F,GAAG,EAAE,CAAC,EAAE,MAAM,CAAC,OAAOljB,EAAE,QAAQC,CAAC,CAAC,CAAC,IAAIuuC,GAAGtkB,EAAE,CAAC,MAAMqkB,EAAE,CAAC,EAAE,SAASE,GAAG1wC,EAAE,EAAE,EAAEE,EAAE,EAAEI,EAAEC,EAAE,CAAC,GAAGD,GAAG,MAAMA,IAAI,OAAO,MAAM,IAAI,MAAM,mCAAmC,EAAE,IAAI,EAAE,IAAIirC,GAAG,EAAErrC,EAAEI,EAAE,GAAGC,CAAC,EAAE,EAAEgwB,GAAGvwB,EAAEM,CAAC,EAAE,QAAQ2B,EAAE,EAAEA,EAAE,EAAE,OAAO,OAAOA,IAAI,EAAE,OAAOA,GAAG,EAAE,UAAU,EAAE,OAAO,EAAE,SAAS,CAAC,CAAC,IAAI0uC,GAAGxkB,EAAE,CAAC,iBAAiBukB,EAAE,CAAC,EAAE,SAASE,GAAG5wC,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,SAAS,mBAAmB,EAAE8X,EAAE5X,EAAE,KAAK,EAAE,IAAI,sCAAsC,EAAE,IAAII,EAAE,CAAC,EAAEJ,CAAC,EAAEK,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE2qB,EAAE,UAAU5F,GAAGhlB,EAAEC,CAAC,EAAE,MAAM,CAAC,OAAO,EAAE,QAAQ,CAAC,CAAC,CAAC,IAAIswC,GAAG1kB,EAAE,CAAC,QAAQykB,EAAE,CAAC,EAAE,SAASE,GAAG9wC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,oBAAoB,EAAEO,EAAEyrB,EAAE,EAAE,aAAa,qBAAqB,OAAO,EAAElU,EAAEQ,GAAGpY,CAAC,EAAE,IAAI,kCAAkC,EAAE,IAAI,EAAE,CAAC,EAAEI,EAAE,WAAWC,CAAC,EAAE,EAAE,CAAC,YAAYL,CAAC,EAAE,OAAOgrB,EAAE,UAAU1F,GAAG,EAAE,CAAC,CAAC,CAAC,IAAIurB,GAAG5kB,EAAE,CAAC,oBAAoB2kB,EAAE,CAAC,EAAE,SAASE,GAAGhxC,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,IAAI,UAAU,mBAAmB,EAAE8X,EAAE,GAAG,CAAC5X,EAAE,MAAM,QAAQ,EAAEA,EAAE,MAAM,OAAO,IAAI,UAAU,iBAAiBA,EAAE,MAAM,WAAWA,EAAE,MAAM,SAAS,EAAE,IAAII,EAAE,CAAC,MAAMJ,CAAC,EAAEK,EAAE,CAAC,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAU3F,GAAGjlB,EAAEC,CAAC,CAAC,CAAC,IAAI0wC,GAAG9kB,EAAE,CAAC,SAAS6kB,EAAE,CAAC,EAAE,SAASE,GAAGlxC,EAAE,EAAE,CAAC,OAAOynC,GAAGznC,EAAE,EAAE,OAAO,CAAC,CAAC,SAASmxC,GAAGnxC,EAAE,EAAE,GAAGE,EAAEI,EAAE,CAAC,OAAO4qB,EAAE,aAAalrB,EAAE,EAAEE,EAAEI,CAAC,CAAC,CAAC,SAAS8wC,GAAGpxC,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,IAAI,EAAE,IAAIA,EAAE,KAAK,CAAC,EAAE,IAAII,EAAEiwB,GAAGvwB,EAAE,OAAO,EAAEO,EAAEgwB,GAAG,CAACrwB,EAAE,OAAOF,EAAE,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,EAAE,EAAEE,EAAE,OAAO,IAAI,CAAC,IAAI,EAAEI,EAAE,WAAWJ,EAAE,EAAE,EAAE+B,EAAE,EAAEjC,EAAE,OAAOO,EAAE,OAAO,IAAI,EAAE0B,CAAC,CAAC,CAAC,OAAO1B,EAAE,SAAS,CAAC,CAAC,eAAe8wC,GAAGrxC,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,YAAY,aAAa,MAAM,EAAEE,EAAE,MAAM,EAAE,KAAK,EAAEI,EAAE8wC,GAAG,EAAE,MAAMlxC,CAAC,EAAE,OAAOF,IAAI,GAAG,EAAE,QAAQ,EAAEM,CAAC,CAAC,IAAIgxC,GAAGD,GAAG,eAAeE,GAAGvxC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,SAAS,UAAU,EAAEO,EAAEyrB,EAAE,EAAE,OAAO,WAAW,MAAM,EAAE,EAAE9rB,GAAG,KAAK,EAAEA,EAAE,EAAEK,EAAE,KAAK0B,EAAE3B,EAAE,MAAMwX,EAAE,EAAE,EAAE,IAAI,uBAAuB,EAAEC,GAAG9V,EAAE,MAAM,EAAE,EAAE,CAAC,EAAE1B,EAAE,MAAM,mEAAmE,EAAE,IAAI2B,EAAE,EAAE,QAAQgC,EAAE,EAAEA,EAAE,EAAE,EAAEA,IAAIhC,GAAGD,EAAEiC,GAAG,IAAI/B,EAAEF,EAAE,MAAM,EAAE,CAAC,EAAE,OAAO,CAACC,CAAC,EAAED,EAAE,MAAM,EAAE,CAAC,CAAC,EAAEW,EAAEs5B,EAAE57B,EAAE6B,CAAC,EAAEU,EAAEq5B,EAAE37B,EAAE,CAAC,EAAE,CAAC,EAAEuC,EAAE,MAAMwuC,GAAGzuC,CAAC,EAAEE,EAAE2sC,GAAG5sC,EAAE,CAAC,CAAC,CAAC,EAAEE,EAAEshC,GAAG1hC,EAAEG,EAAE,CAAC,EAAE,OAAO/C,IAAIM,GAAGA,EAAE,QAAQ,EAAE,IAAIC,GAAGA,EAAE,QAAQ,EAAEwC,EAAE,QAAQ,EAAEH,EAAE,QAAQ,EAAEC,EAAE,QAAQ,EAAEC,EAAE,QAAQ,EAAEE,CAAC,CAAC,IAAIwuC,GAAGD,GAAG,SAASE,GAAGzxC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,GAAG,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,IAAI,eAAe,EAAE,EAAEgsB,EAAE,EAAE,IAAI,eAAe,EAAE/pB,EAAE+pB,EAAE9rB,EAAE,QAAQ,eAAe,EAAE+pB,GAAG,EAAE,CAAC,EAAEnS,EAAEE,GAAG,EAAE,MAAM,EAAE,KAAK,EAAE,IAAI,2BAA2B,EAAE,IAAI9V,EAAE2gC,GAAG,CAAC,EAAE1gC,EAAEukC,GAAGxkC,EAAED,CAAC,EAAEW,EAAEw2B,EAAEsN,GAAG,EAAE,CAAC,EAAEvkC,CAAC,EAAE,GAAG5B,EAAE,CAACuX,EAAExX,GAAG,KAAK,IAAI,gDAAgD,EAAE,IAAIuC,EAAEmpB,EAAE1rB,EAAE,OAAO,eAAe,EAAEsC,EAAEs2B,GAAGt2B,EAAE8jC,GAAGxkC,EAAE0gC,GAAG3gC,EAAEY,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOi2B,EAAE,EAAEl2B,CAAC,CAAC,CAAC,IAAI8uC,GAAGvlB,EAAE,CAAC,eAAeslB,EAAE,CAAC,EAAE,SAASE,GAAG3xC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,UAAU,YAAY,OAAO,EAAEO,EAAEyrB,EAAE,EAAE,UAAU,WAAW,EAAEkK,GAAG31B,EAAED,EAAEJ,CAAC,EAAE,IAAI,EAAE,CAAC,QAAQI,EAAE,QAAQC,CAAC,EAAE,EAAE,CAAC,MAAML,CAAC,EAAE,OAAOgrB,EAAE,UAAU7H,GAAG,EAAE,CAAC,CAAC,CAAC,IAAIuuB,GAAGzlB,EAAE,CAAC,WAAWwlB,EAAE,CAAC,EAAE,SAASE,GAAG7xC,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGN,EAAE,QAAQ,QAAQ,MAAM,IAAI,MAAM,8EAA8EA,EAAE,QAAQ,EAAE,GAAGA,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,sEAAsEA,EAAE,QAAQ,EAAE,IAAIO,EAAEP,EAAE,KAAK,EAAEA,EAAE,MAAM,GAAG,EAAE,EAAEA,EAAE,KAAK,EAAEA,EAAE,MAAM,GAAG,EAAE,GAAGE,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,kDAAkDA,EAAE,sBAAsB,IAAI,EAAE,IAAI,EAAE,EAAE,KAAK,GAAG,EAAE,EAAE,OAAO,GAAG,EAAE,OAAO,GAAG,IAAIK,GAAG,MAAM,IAAI,MAAM,oCAAoC,EAAE,2BAA2BA,IAAI,EAAE,GAAG,EAAE,QAAQD,EAAE,MAAM,MAAM,IAAI,MAAM,mDAAmD,CAAC,CAAC,SAASwxC,GAAG9xC,EAAE,EAAEE,EAAEI,EAAE,EAAE,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,gBAAgB,gBAAgB,OAAO,EAAE,EAAEgsB,EAAE,EAAE,eAAe,gBAAgB,mBAAmB,EAAE,EAAEA,EAAE1rB,EAAE,eAAe,gBAAgB,EAAE,KAAK,EAAEuxC,GAAGtxC,EAAE,EAAEL,EAAE,CAAC,EAAE,IAAI+B,EAAE,CAAC,cAAc1B,EAAE,aAAa,EAAE,aAAa,CAAC,EAAE2B,EAAE,CAAC,YAAYhC,CAAC,EAAE,OAAOgrB,EAAE,UAAU1G,GAAGviB,EAAEC,CAAC,CAAC,CAAC,IAAI6vC,GAAG5lB,EAAE,CAAC,eAAe2lB,EAAE,CAAC,EAAE,SAASE,GAAGhyC,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAE,EAAE,UAAU,WAAW,OAAO,EAAEzrB,EAAE,CAAC,OAAOyrB,EAAEhsB,EAAE,IAAI,WAAW,mBAAmB,EAAE,QAAQE,CAAC,EAAE,OAAOgrB,EAAE,UAAU/L,GAAG5e,CAAC,CAAC,CAAC,IAAI0xC,GAAG9lB,EAAE,CAAC,UAAU6lB,EAAE,CAAC,EAAE,SAASE,GAAGlyC,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,OAAOA,EAAE,MAAM,MAAM,EAAE,GAAGgY,GAAGhY,EAAE,MAAM,CAAC,EAAE,OAAO,EAAE,GAAGA,EAAE,MAAM,SAAS,EAAE,OAAO,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,MAAM,OAAOM,IAAI,EAAEA,IAAI,MAAMN,EAAE,MAAMM,IAAI,KAAKJ,EAAE,KAAKF,EAAE,MAAMM,EAAE,EAAEJ,EAAE,KAAK,EAAEI,EAAE,EAAE,OAAOJ,CAAC,CAAC,OAAO,CAAC,CAAC,SAASiyC,GAAGnyC,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,IAAI,SAAS,EAAE,GAAG8X,EAAEvX,EAAE,QAAQ,UAAU,IAAI,gFAAgFA,EAAE,uBAAuB,EAAEuX,EAAE,GAAG,GAAG,EAAE,EAAE,IAAI,qDAAqD,IAAI,EAAE,IAAI,EAAE,OAAO9X,aAAa6pB,GAAGtpB,EAAE,MAAM,EAAEA,EAAE,IAAI,EAAE2xC,GAAG3xC,EAAEL,CAAC,EAAE,EAAE,EAAE,EAAE+B,EAAEi3B,GAAGkL,GAAGtL,EAAEmT,GAAG,EAAE,EAAE,EAAE,UAAU3rC,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,OAAO84B,EAAE74B,EAAE0B,CAAC,CAAC,CAAC,IAAImwC,GAAGjmB,EAAE,CAAC,SAASgmB,EAAE,CAAC,EAAE,SAASE,GAAGryC,EAAE,CAAC,OAAO,KAAK,MAAM,KAAK,IAAI,EAAE,KAAK,KAAK,KAAK,IAAIA,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAASsyC,GAAGtyC,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAEN,EAAE,EAAEO,EAAE,IAAI,aAAaP,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE,KAAK,GAAG,GAAGA,EAAEM,EAAE,GAAGC,EAAE,GAAG,EAAEL,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC,OAAOiwC,GAAG5vC,EAAE,SAAS,CAAC,CAAC,eAAegyC,GAAGvyC,EAAE,EAAEE,EAAE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,cAAc,QAAQ,EAAEO,EAAEyrB,EAAE,EAAE,UAAU,QAAQ,EAAElU,EAAExX,EAAE,KAAK,EAAE,IAAI,uEAAuEA,EAAE,MAAM,EAAEwX,EAAExX,EAAE,KAAK,IAAIC,EAAE,KAAK,IAAI,mFAAmFD,EAAE,yBAAyBC,EAAE,MAAM,EAAEwX,GAAGzX,EAAE,MAAM,MAAM,EAAEA,EAAE,MAAM,OAAO,CAAC,EAAEC,EAAE,MAAM,yFAAyF,EAAE,IAAI,EAAED,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAGwX,EAAE5X,EAAE,GAAGA,GAAG,EAAE,IAAI,4EAA4E,eAAeA,GAAG,EAAE,IAAI,EAAE,MAAMI,EAAE,KAAK,EAAE2B,EAAE,MAAM1B,EAAE,KAAK,EAAE,CAAC2B,EAAEC,CAAC,EAAE,CAAC,EAAE,OAAO,EAAE,CAAC,EAAES,EAAEmW,GAAG,OAAO7W,CAAC,EAAE,QAAQW,EAAE,EAAEA,EAAEX,EAAEW,IAAI,CAAC,IAAIC,EAAED,EAAEV,EAAEY,EAAE,EAAE,SAASD,EAAEA,EAAEX,CAAC,EAAEa,EAAE,CAAC,EAAE,QAAQkB,EAAE,EAAEA,EAAEnB,EAAE,OAAOmB,IAAIlB,EAAE,KAAK,CAAC,MAAMD,EAAEmB,GAAG,MAAMA,CAAC,CAAC,EAAElB,EAAE,KAAK,CAACkB,EAAE,IAAI,EAAE,MAAMA,EAAE,KAAK,EAAEtB,EAAEC,GAAG,EAAE,QAAQqB,EAAE,EAAEA,EAAEhE,EAAEgE,IAAI,GAAGlB,EAAEkB,GAAG,QAAQjC,EAAEY,GAAG,CAACD,EAAEC,GAAG,EAAE,KAAK,CAAC,CAAC,OAAO7C,IAAIM,GAAGA,EAAE,QAAQ,EAAE,IAAIC,GAAGA,EAAE,QAAQ,EAAEgsB,GAAG3pB,EAAErC,EAAE,MAAM,MAAM,CAAC,CAAC,IAAIiyC,GAAGD,GAAOE,GAAG,CAAC,EAAEryC,GAAGqyC,GAAG,CAAC,OAAO,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,OAAO,IAAIC,EAAE,CAAC,EAAE,SAASC,GAAG7yC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,OAAO,EAAE,CAAC,IAAI0B,EAAEjC,EAAEA,EAAE,OAAO,IAAIiC,EAAEi6B,EAAEl8B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,IAAIkC,EAAE,EAAEA,EAAE,OAAO,IAAIA,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE7V,EAAE,OAAO,EAAE,IAAI,iEAAiEA,EAAE,QAAQ,EAAE6V,EAAE5V,EAAE,OAAO,EAAE,IAAI,8DAA8DA,EAAE,QAAQ,EAAE4V,EAAE5X,EAAE,SAAS,EAAE,IAAI,mEAAmEA,IAAI,EAAE,IAAIiC,EAAE,IAAI,OAAOF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGW,EAAE,IAAI,OAAOV,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG4V,EAAE3V,IAAIjC,EAAE,GAAG,IAAI,4CAA4CiC,wCAAwCjC,EAAE,KAAK,EAAE4X,EAAElV,IAAI1C,EAAE,GAAG,IAAI,0CAA0C0C,0CAA0C1C,EAAE,MAAM,EAAE87B,GAAG,kBAAkBz7B,EAAE,CAAC,EAAE,IAAIsC,EAAE,CAAC,EAAEZ,EAAE,GAAGC,CAAC,EAAEY,EAAE,CAAC,QAAQxC,EAAE,IAAIC,EAAE,WAAW,EAAE,gBAAgB,EAAE,YAAYL,CAAC,EAAE,OAAOgrB,EAAE,UAAUlO,GAAGna,EAAEC,CAAC,CAAC,CAAC,IAAIgwC,GAAG3mB,EAAE,CAAC,sBAAsB0mB,EAAE,CAAC,EAAE,SAASE,GAAG/yC,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,MAAMA,IAAI,SAAS,OAAOF,EAAE,GAAGE,IAAI,OAAO,OAAOk5B,EAAEp5B,EAAE8vC,GAAG,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,gDAAgD5vC,IAAI,CAAC,CAAC,SAAS8yC,GAAGhzC,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAEI,EAAE40B,GAAGl1B,EAAE,MAAM,EAAE,KAAK,EAAE,OAAOM,EAAE,OAAO,IAAIJ,EAAEijC,GAAGjjC,EAAEI,CAAC,GAAG47B,EAAEh8B,EAAEF,EAAE,KAAK,CAAC,CAAC,SAASizC,GAAGjzC,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,IAAI,SAAS,OAAON,EAAE,GAAG,IAAI,OAAO,OAAOssC,GAAGtsC,CAAC,EAAE,GAAG,IAAI,MAAM,OAAO4hC,GAAG5hC,CAAC,EAAE,GAAG,IAAI,QAAQ,OAAOwsC,GAAGxsC,CAAC,EAAE,GAAG,IAAI,QAAQ,OAAO2qC,GAAG3qC,EAAEE,CAAC,EAAE,GAAG,IAAI,YAAY,OAAOglC,GAAGllC,EAAEM,CAAC,EAAE,GAAG,IAAI,UAAU,OAAOo8B,GAAG18B,CAAC,EAAE,MAAM,IAAI,MAAM,4BAA4B,IAAI,CAAC,CAAC,IAAIkzC,GAAG,CAAClzC,EAAE,IAAI,EAAEA,EAAE,IAAI,IAAI,SAAS,SAASmzC,GAAG,CAAC,EAAEnzC,EAAE,OAAO,EAAE,QAAQE,EAAE,IAAII,EAAE,WAAWC,EAAE,OAAO,UAAU,EAAE,CAAC,EAAE,CAAC,EAAE,gBAAgB,EAAE,KAAK0B,EAAE,WAAWC,EAAE,SAAS,uBAAuBC,EAAE,eAAeS,CAAC,EAAE,CAAC,GAAGV,EAAEA,GAAG,SAASgxC,GAAGhoB,EAAE,MAAM,cAAchpB,CAAC,IAAI,GAAG,CAAC4V,EAAEvX,IAAI,OAAO,IAAI,4CAA4CA,8GAA8G,EAAE,IAAIuE,EAAEk6B,GAAGh/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAO0B,GAAG,OAAO6C,EAAEg0B,EAAEh0B,EAAE7C,CAAC,GAAGgxC,GAAGnuC,EAAE5C,EAAEC,EAAES,CAAC,CAAC,CAAC,IAAIC,EAAEmpB,EAAEhsB,EAAE,IAAI,SAAS,SAAS,EAAE8C,EAAEkpB,EAAE,EAAE,SAAS,SAAS,SAAS,EAAEjpB,EAAEF,EAAEG,EAAE,GAAGH,EAAE,OAAO,IAAIG,EAAE,GAAGD,EAAEm5B,EAAEr5B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAGiV,EAAE/U,EAAE,OAAO,EAAE,IAAI,6DAA6DA,EAAE,OAAO,EAAE+U,EAAEhV,EAAE,OAAO,EAAE,IAAI,8DAA8DA,EAAE,OAAO,EAAEk5B,GAAG,eAAe17B,EAAE,CAAC,EAAE,IAAI4D,EAAE3D,IAAI,OAAOwC,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG+U,EAAEhV,EAAE,MAAM,KAAKoB,EAAE,IAAI,oCAAoCA,wCAAwCpB,EAAE,MAAM,KAAK,EAAEgV,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,uEAAuEA,oBAAoB,IAAI,EAAE,IAAI,EAAE+6B,GAAGl4B,EAAE,MAAMD,EAAE,MAAM5C,EAAE,EAAEI,EAAE,CAAC,EAAEiH,EAAEtF,GAAG,OAAOsF,EAAEykB,EAAE/pB,EAAE,OAAO,cAAc,EAAE,CAACsF,CAAC,EAAE6iB,GAAG7iB,EAAE1E,CAAC,EAAEtC,IAAI,OAAOy0B,GAAG,EAAE,SAASztB,EAAE,KAAK,GAAGuQ,EAAEvQ,EAAE,MAAM,QAAQ,EAAE,IAAI,2GAA2GA,EAAE,MAAM,SAAS,EAAEuQ,EAAEvQ,EAAE,MAAM,SAAS,GAAGA,EAAE,MAAM,KAAK,EAAE,aAAaA,EAAE,MAAM,KAAK,EAAE,IAAI,sCAAsCA,EAAE,gEAAgE,EAAE,cAAc,IAAI,IAAI,EAAE,GAAGpF,GAAG,KAAK,CAAC,IAAI2C,EAAE3C,EAAE,MAAM,GAAG2V,EAAEhT,EAAE,QAAQ,GAAGA,EAAE,SAAS,EAAE,IAAI,2HAA2HA,EAAE,SAAS,EAAEA,EAAE,SAAS,EAAEgT,EAAEhT,EAAE,KAAK,GAAGA,EAAE,KAAK,EAAE,YAAY,IAAI,oDAAoDA,4DAA4D,EAAE,eAAe,UAAUA,EAAE,SAAS,EAAE,GAAG,CAACkwB,GAAGlwB,EAAE,EAAE,QAAQ,CAAC,OAAO0C,EAAN,CAAS,IAAIxC,EAAE,oDAAoDF,6DAA6D,EAAE,aAAa,MAAM,MAAME,CAAC,CAAC,CAAC,EAAEgnB,EAAE7pB,EAAE,gBAAgB,cAAc,CAAC,CAAC,IAAIwC,EAAE,CAACG,EAAE0C,IAAI,CAACsQ,EAAEvX,IAAI,OAAO,IAAI,wDAAwDA,yCAAyC,EAAE,GAAG,CAACyE,EAAEyC,EAAExC,EAAEC,CAAC,EAAEsC,EAAErC,EAAE4tC,GAAGjuC,EAAEG,EAAE/C,CAAC,EAAE4V,EAAEgkB,GAAG,CAAC,EAAE,IAAI,uHAAuH,IAAI,EAAE,IAAI12B,EAAEg6B,GAAG33B,EAAE,MAAMtC,EAAEH,EAAE9E,EAAEI,CAAC,EAAE+E,EAAEytC,GAAGrrC,EAAEtC,EAAEH,EAAE,MAAM9E,EAAEI,CAAC,EAAEgF,EAAE,CAACF,EAAEC,CAAC,EAAE,GAAGH,GAAG,KAAK,CAAC,IAAI+C,EAAE+qC,GAAG9tC,EAAEC,CAAC,EAAEG,EAAE,KAAK2C,CAAC,CAAC,CAAC,OAAO3C,CAAC,EAAEV,EAAE,CAAC,EAAE7B,EAAE,OAAOD,EAAE,KAAKyE,EAAE,uBAAuB,CAAC,EAAE1C,EAAE,CAAC,QAAQ3E,EAAE,IAAII,EAAE,WAAWC,EAAE,UAAU,EAAE,gBAAgB,EAAE,WAAW2B,EAAE,eAAeU,CAAC,EAAE,OAAOX,GAAG,KAAKmkC,GAAG,CAAC5+B,EAAExC,EAAEyC,IAAI,CAAC,IAAIxC,EAAEimB,EAAE,UAAUnF,GAAGnhB,EAAEC,CAAC,EAAE,OAAO4C,EAAE,CAACzC,EAAEwC,EAAEvC,CAAC,CAAC,EAAEjC,IAAIiC,EAAEi3B,EAAEj3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,CAAC,MAAMA,EAAE,SAASN,CAAC,CAAC,CAAC,EAAE5B,EAAED,CAAC,EAAEsjC,GAAG,CAAC5+B,EAAExC,EAAEyC,EAAExC,IAAI,CAAC,IAAIC,EAAEgmB,EAAE,UAAUnF,GAAGnhB,EAAEC,CAAC,EAAE,OAAOI,EAAE,CAACD,EAAEwC,EAAEtC,EAAEuC,CAAC,CAAC,EAAEzE,IAAIkC,EAAEg3B,EAAEh3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,CAAC,MAAMA,EAAE,SAASP,CAAC,CAAC,CAAC,EAAE5B,EAAED,EAAEyE,CAAC,CAAC,CAAC,IAAImrC,GAAGvmB,EAAE,CAAC,aAAagnB,EAAE,CAAC,EAAE,SAASC,GAAGpzC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,IAAI0B,EAAEjC,EAAEA,EAAE,OAAO,IAAIiC,EAAEi6B,EAAEl8B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,IAAIkC,EAAE,EAAEA,EAAE,OAAO,IAAIA,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI/5B,EAAE,CAAC,EAAEF,EAAE,GAAGC,CAAC,EAAEU,EAAE,CAAC,QAAQtC,EAAE,IAAIC,EAAE,gBAAgB,EAAE,UAAU,EAAE,YAAYL,CAAC,EAAE,OAAOgrB,EAAE,UAAUrN,GAAG1b,EAAES,CAAC,CAAC,CAAC,IAAIywC,GAAGlnB,EAAE,CAAC,qCAAqCinB,EAAE,CAAC,EAAE,SAASE,GAAGtzC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,IAAI0B,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGD,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI/5B,EAAE,CAAC,GAAGF,EAAE,OAAO/B,CAAC,EAAE0C,EAAE,CAAC,QAAQtC,EAAE,IAAIC,EAAE,gBAAgB,EAAE,UAAU,EAAE,WAAWP,CAAC,EAAE6C,EAAEqoB,EAAE,UAAUpN,GAAG3b,EAAES,CAAC,EAAE,OAAOV,EAAEg6B,EAAEr5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI0wC,GAAGpnB,EAAE,CAAC,oCAAoCmnB,EAAE,CAAC,EAAE,SAASE,GAAG,CAAC,EAAExzC,EAAE,OAAO,EAAE,QAAQE,EAAE,IAAII,EAAE,WAAWC,EAAE,OAAO,UAAU,EAAE,CAAC,EAAE,CAAC,EAAE,gBAAgB,EAAE,KAAK0B,EAAE,WAAWC,EAAE,SAAS,uBAAuBC,EAAE,eAAeS,CAAC,EAAE,CAAC,GAAGswC,GAAGhoB,EAAE,MAAM,cAAchpB,CAAC,IAAI,GAAG,CAAC,IAAI2C,EAAE67B,GAAG1gC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAO0B,GAAG,OAAO4C,EAAEi0B,EAAEj0B,EAAE5C,CAAC,GAAGgxC,GAAGpuC,EAAE3C,EAAEC,EAAES,CAAC,CAAC,CAAC,IAAIC,EAAEmpB,EAAEhsB,EAAE,IAAI,kBAAkB,SAAS,EAAE8C,EAAEkpB,EAAE,EAAE,SAAS,kBAAkB,SAAS,EAAEjpB,EAAEF,EAAEG,EAAE,GAAGH,EAAE,OAAO,IAAIG,EAAE,GAAGD,EAAEm5B,EAAEr5B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAGiV,EAAE/U,EAAE,OAAO,EAAE,IAAI,sEAAsEA,EAAE,OAAO,EAAE+U,EAAEhV,EAAE,OAAO,EAAE,IAAI,uEAAuEA,EAAE,OAAO,EAAEgV,EAAE/U,EAAE,MAAM,KAAKD,EAAE,MAAM,GAAG,IAAI,6DAA6DC,EAAE,MAAM,qDAAqDD,EAAE,MAAM,KAAK,EAAE,GAAG,OAAO,EAAE,CAAC,EAAE,CAAC,GAAGgV,EAAEikB,GAAG77B,EAAE,CAAC,EAAE,IAAI,sFAAsFA,oBAAoB,IAAI,EAAE87B,GAAG,wBAAwB17B,EAAE,CAAC,EAAE,IAAI4D,EAAE+2B,GAAGl4B,EAAE,MAAMD,EAAE,MAAM5C,EAAE,EAAEI,EAAE,EAAE,EAAE,EAAE,EAAE2B,GAAG,OAAO,EAAE+pB,EAAE/pB,EAAE,OAAO,cAAc,EAAE,CAAC,CAAC,EAAEmoB,GAAG,EAAEvnB,CAAC,EAAEmyB,GAAG9wB,EAAE,SAAS,EAAE,KAAK,GAAG,IAAIqD,EAAEpF,GAAG,OAAOoF,EAAEykB,EAAE7pB,EAAE,gBAAgB,uBAAuB,GAAG,IAAI,EAAE,CAAC0C,EAAEC,IAAI,CAACgT,EAAEgkB,GAAG,CAAC,EAAE,IAAI,mHAAmH,IAAI,EAAE,GAAG,CAACt0B,EAAExC,EAAEyC,EAAExC,CAAC,EAAEH,EAAEI,EAAE6tC,GAAGluC,EAAE4C,EAAEvF,CAAC,EAAEiD,EAAEouC,GAAGvuC,EAAE,MAAME,EAAEsC,EAAEtH,EAAEI,EAAE,EAAE,CAAC,EAAE8E,EAAEiuC,GAAGruC,EAAEE,EAAEsC,EAAE,MAAMtH,EAAEI,EAAE,EAAE,CAAC,EAAE,GAAG2E,GAAG,KAAK,CAAC,IAAII,EAAE2tC,GAAG,EAAE9tC,CAAC,EAAE,MAAM,CAACC,EAAEC,EAAEC,CAAC,CAAC,CAAC,MAAM,CAACF,EAAEC,CAAC,CAAC,EAAET,EAAE,CAAC,EAAE5B,EAAE,OAAOD,EAAE,KAAK,EAAE,uBAAuByE,CAAC,EAAE3C,EAAE,CAAC,QAAQ1E,EAAE,IAAII,EAAE,WAAWC,EAAE,UAAU,EAAE,gBAAgB,EAAE,WAAW2B,EAAE,eAAeU,CAAC,EAAE,OAAOX,GAAG,KAAKmkC,GAAG,CAACthC,EAAE0C,EAAExC,IAAI,CAAC,IAAIyC,EAAEyjB,EAAE,UAAUlF,GAAGrhB,EAAEC,CAAC,EAAE,OAAOI,EAAE,CAACwC,EAAE1C,EAAE2C,CAAC,CAAC,EAAEzE,IAAIyE,EAAEy0B,EAAEz0B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,CAAC,MAAMA,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE1E,EAAED,CAAC,EAAEsjC,GAAG,CAACthC,EAAE0C,EAAExC,EAAEyC,IAAI,CAAC,IAAIxC,EAAEimB,EAAE,UAAUlF,GAAGrhB,EAAEC,CAAC,EAAE,OAAO6C,EAAE,CAACD,EAAE1C,EAAEG,EAAED,CAAC,CAAC,EAAEhC,IAAIiC,EAAEi3B,EAAEj3B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,CAAC,MAAMA,EAAE,SAAS,CAAC,CAAC,CAAC,EAAElC,EAAED,EAAE,CAAC,CAAC,CAAC,IAAI6vC,GAAGxmB,EAAE,CAAC,sBAAsBqnB,EAAE,CAAC,EAAE,SAASC,GAAG,CAAC,EAAEzzC,EAAE,EAAE,EAAE,WAAWE,EAAE,GAAG,WAAWI,EAAE,GAAG,KAAKC,EAAE,WAAW,EAAE,SAAS,uBAAuB,EAAE,eAAe0B,EAAE,EAAE,EAAE,CAAC,GAAGixC,GAAGhoB,EAAE,MAAM,cAAc,CAAC,IAAI,GAAG,CAAC,IAAIhmB,EAAE4tB,GAAG9yB,EAAE,EAAEE,EAAEI,CAAC,EAAE,OAAOC,GAAG,OAAO2E,EAAE4zB,EAAE5zB,EAAE3E,CAAC,GAAG0yC,GAAG/tC,EAAE,EAAE,EAAEjD,CAAC,CAAC,CAAC,IAAIC,EAAE8pB,EAAEhsB,EAAE,IAAI,cAAc,EAAEmC,EAAE6pB,EAAE,EAAE,IAAI,cAAc,EAAE,CAAC9pB,EAAEC,CAAC,EAAEioB,GAAGloB,EAAEC,CAAC,EAAE,IAAIS,EAAE1C,EAAEgC,EAAE,MAAMA,EAAE,KAAK,GAAGA,EAAE,MAAMA,EAAE,KAAK,GAAGW,EAAEvC,EAAE6B,EAAE,MAAMA,EAAE,KAAK,GAAGA,EAAE,MAAMA,EAAE,KAAK,GAAGW,EAAE5C,EAAEgC,EAAE,MAAMA,EAAE,KAAK,GAAGA,EAAE,MAAMA,EAAE,KAAK,GAAGa,EAAEzC,EAAE6B,EAAE,MAAMA,EAAE,KAAK,GAAGA,EAAE,MAAMA,EAAE,KAAK,GAAGa,EAAEd,EAAE,MAAM,MAAM,EAAE,EAAE,EAAEgC,EAAE/B,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAEiW,GAAGpV,CAAC,EAAEuE,EAAE6Q,GAAGlU,CAAC,EAAE4T,EAAElV,IAAIC,EAAE,IAAI,wCAAwCD,WAAWC,6BAA6BX,EAAE,aAAaC,EAAE,wBAAwBjC,oBAAoBI,eAAe,EAAE,IAAIqE,EAAEqwB,GAAG9yB,EAAE,MAAM,MAAM,EAAE,EAAE,EAAEC,EAAE,MAAM,MAAM,EAAE,EAAE,CAAC,EAAE,OAAO,CAACW,EAAEC,CAAC,CAAC,EAAE6B,EAAE1E,EAAEg8B,EAAEh6B,EAAE,CAAC,EAAEU,EAAEE,CAAC,CAAC,EAAEo5B,EAAEh6B,EAAE,CAAC,EAAEY,EAAEF,CAAC,CAAC,EAAEiC,EAAEvE,EAAE47B,EAAE/5B,EAAE,CAACoF,EAAExE,EAAEF,CAAC,CAAC,EAAEq5B,EAAE/5B,EAAE,CAACoF,EAAE1E,EAAEE,CAAC,CAAC,EAAE+B,EAAEvE,GAAG,OAAOuE,EAAEknB,EAAEzrB,EAAE,OAAO,cAAc,EAAE,CAACuE,CAAC,EAAEslB,GAAGtlB,EAAE5C,CAAC,EAAE8yB,GAAGrwB,EAAEG,EAAE,KAAK,GAAG,IAAI0C,EAAE,GAAG,OAAOA,EAAEwkB,EAAE,EAAE,gBAAgB,cAAc,GAAG,IAAIhnB,EAAE,CAACE,EAAEC,IAAI,CAAC,GAAG,CAACC,EAAEC,EAAEC,EAAE2C,CAAC,EAAE9C,EAAEwC,EAAEorC,GAAG7W,EAAEh3B,EAAEI,EAAE,KAAK,EAAEA,EAAE,CAAC,EAAE4C,EAAEC,EAAG,GAAG,CAACjI,GAAG,CAACI,GAAG4H,EAAE4qB,GAAGnrB,EAAEtC,EAAE,GAAG,EAAE,EAAE8C,EAAG2qB,GAAG1tB,EAAEuC,EAAE,GAAG,EAAE,GAAG,CAACzH,GAAGI,GAAG4H,EAAE4qB,GAAGnrB,EAAEtC,EAAE,GAAG,EAAE,EAAE8C,EAAG2qB,GAAGnrB,EAAEvC,EAAE,GAAG,EAAE,GAAGlF,GAAG,CAACI,GAAG4H,EAAE4qB,GAAGztB,EAAEsC,EAAE,GAAG,EAAE,EAAEQ,EAAG2qB,GAAG1tB,EAAEuC,EAAE,GAAG,EAAE,IAAIO,EAAE4qB,GAAGztB,EAAEsC,EAAE,GAAG,EAAE,EAAEQ,EAAG2qB,GAAGnrB,EAAEvC,EAAE,GAAG,EAAE,GAAG7E,GAAG,KAAK,CAAC,IAAI6H,EAAG4qC,GAAG/qC,EAAEN,CAAC,EAAE,MAAM,CAACO,EAAEC,EAAGC,CAAE,CAAC,KAAM,OAAM,CAACF,EAAEC,CAAE,CAAC,EAAEV,EAAE,CAAC,EAAE7C,EAAE,EAAEC,EAAE,KAAKC,EAAE,uBAAuB0C,CAAC,EAAEvC,EAAE,CAAC,WAAW/E,EAAE,WAAWI,EAAE,WAAW,EAAE,eAAe2B,CAAC,EAAE,OAAO1B,GAAG,KAAK6lC,GAAG,CAACjhC,EAAEC,EAAEC,IAAI,CAAC,IAAIC,EAAE4lB,EAAE,UAAUpF,GAAGre,EAAExC,CAAC,EAAE,OAAOI,EAAE,CAACF,EAAEC,EAAEE,CAAC,CAAC,EAAE,CAAC,MAAM42B,EAAE52B,EAAEX,CAAC,EAAE,SAASK,CAAC,CAAC,CAAC,EAAEJ,EAAEC,CAAC,EAAEuhC,GAAG,CAACjhC,EAAEC,EAAEC,EAAEC,IAAI,CAAC,IAAI2C,EAAEijB,EAAE,UAAUpF,GAAGre,EAAExC,CAAC,EAAE,OAAOK,EAAE,CAACH,EAAEC,EAAE6C,EAAE5C,CAAC,CAAC,EAAE,CAAC,MAAM62B,EAAEj0B,EAAEtD,CAAC,EAAE,SAASK,CAAC,CAAC,CAAC,EAAEJ,EAAEC,EAAEC,CAAC,CAAC,CAAC,IAAI8tC,GAAGzmB,EAAE,CAAC,aAAasnB,EAAE,CAAC,EAAE,SAASC,GAAG1zC,EAAE,CAAC,OAAOsyC,GAAGtyC,EAAE,IAAI,GAAG,CAAC,CAAC,IAAI2zC,GAAGxnB,EAAE,CAAC,eAAeunB,EAAE,CAAC,EAAE,SAASE,GAAG5zC,EAAE,CAAC,OAAOsyC,GAAGtyC,EAAE,GAAG,EAAE,CAAC,CAAC,IAAI6zC,GAAG1nB,EAAE,CAAC,YAAYynB,EAAE,CAAC,EAAE,SAASE,GAAG9zC,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE,EAAE,CAAC,EAAE,KAAK,EAAE,GAAGP,EAAE,MAAM,EAAE,KAAK48B,GAAG58B,EAAE,EAAE,CAAC,CAAC,EAAE,GAAGE,EAAE,GAAGI,EAAE,KAAK,EAAEN,EAAE,MAAM,CAAC,IAAIiC,EAAE,EAAE,EAAEjC,EAAE,KAAKkC,EAAEs6B,GAAG,CAACI,GAAG58B,EAAE,EAAE,EAAEiC,CAAC,EAAEm8B,GAAG,CAACn8B,CAAC,EAAE1B,CAAC,CAAC,CAAC,EAAE,EAAE,KAAK2B,CAAC,EAAE,GAAGhC,CAAC,CAAC,OAAO,EAAE,SAAS,EAAEkwC,GAAG,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAElU,EAAEM,GAAG,CAAC,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,IAAIuX,GAAG5nB,EAAE,CAAC,OAAO2nB,EAAE,CAAC,EAAE,SAASE,GAAGh0C,EAAE,EAAEE,EAAEI,EAAEC,EAAEszC,GAAG,CAACvzC,GAAG,OAAOA,EAAE+xC,GAAG,CAAC,GAAG,IAAI,EAAE0B,GAAG/zC,EAAE,EAAEE,CAAC,EAAE,EAAEk5B,EAAE,EAAE74B,EAAE,CAAC,CAAC,EAAE,OAAO+uC,GAAG,EAAEhvC,CAAC,CAAC,CAAC,IAAI2zC,GAAG9nB,EAAE,CAAC,MAAM6nB,EAAE,CAAC,EAAE,SAASE,GAAGl0C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,WAAW,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,eAAe,EAAEiC,EAAE+pB,EAAE,EAAE,QAAQ,gBAAgB,SAAS,EAAE9pB,EAAE8pB,EAAE9rB,EAAE,SAAS,gBAAgB,OAAO,EAAEiC,EAAEF,EAAE,MAAM,GAAG6V,EAAE,EAAE,OAAO,EAAE,IAAI,6DAA6D,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,GAAGA,EAAE,MAAM,KAAK,EAAE,IAAI,oDAAoDE,sBAAsBF,EAAE,QAAQ,EAAE6V,EAAE5V,EAAE,OAAO,GAAGA,EAAE,MAAM,KAAKC,EAAE,IAAI,qDAAqDA,oBAAoBF,EAAE,QAAQ,EAAE6V,EAAExX,EAAE,SAAS,EAAE,IAAI,wEAAwEA,EAAE,SAAS,EAAEwX,EAAExX,EAAE,IAAI,GAAGA,EAAE,IAAI,EAAE,IAAI,2CAA2CA,GAAG,EAAEwX,EAAEvX,IAAI,YAAYA,IAAI,UAAU,IAAI,+CAA+CA,GAAG,EAAE,IAAIqC,EAAE,CAAC,MAAM,EAAE,MAAMX,EAAE,OAAOC,CAAC,EAAEW,EAAE,CAAC,OAAOtC,EAAE,mBAAmB,EAAE,SAASD,CAAC,EAAE,OAAO4qB,EAAE,UAAUzN,GAAG7a,EAAEC,CAAC,CAAC,CAAC,IAAIsxC,GAAGhoB,EAAE,CAAC,eAAe+nB,EAAE,CAAC,EAAE,SAASE,GAAGp0C,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,QAAQ,gBAAgB,SAAS,EAAE8X,EAAE,EAAE,OAAO,EAAE,IAAI,6DAA6D,EAAE,OAAO,EAAE,IAAI5X,EAAE,CAAC,MAAM,CAAC,EAAE,OAAOgrB,EAAE,UAAUpM,GAAG5e,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIm0C,GAAGloB,EAAE,CAAC,eAAeioB,EAAE,CAAC,EAAE,SAASE,GAAGt0C,EAAE,CAAC,IAAI,EAAEgsB,EAAEhsB,EAAE,QAAQ,gBAAgB,EAAEE,EAAE,EAAE,KAAK,EAAEI,EAAE,EAAE,MAAMJ,GAAG4X,EAAE,EAAE,MAAM,EAAE,IAAI,yEAAyE,EAAE,OAAO,EAAEA,EAAExX,IAAI,EAAE,IAAI,+FAA+FA,IAAI,EAAE,IAAIC,EAAE,IAAI,MAAM,EAAE,IAAI,EAAE,OAAOA,EAAE,KAAK,EAAE,EAAEL,CAAC,EAAEK,EAAEL,GAAG,EAAE8jC,GAAG,EAAEzjC,CAAC,CAAC,CAAC,IAAIg0C,GAAGpoB,EAAE,CAAC,gBAAgBmoB,EAAE,CAAC,EAAE,SAASE,GAAGx0C,EAAE,EAAEE,EAAE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,QAAQ,mBAAmB,SAAS,EAAE8X,EAAEvX,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,MAAMA,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAUL,EAAE,OAAOI,CAAC,EAAE,OAAO4qB,EAAE,UAAUrF,GAAG,EAAE,CAAC,CAAC,CAAC,IAAI4uB,GAAGtoB,EAAE,CAAC,kBAAkBqoB,EAAE,CAAC,EAAE,SAASE,GAAG10C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAACD,GAAG,OAAOA,EAAE,IAAIC,GAAG,OAAOA,EAAE,OAAO,mBAAmB,GAAG,OAAO,EAAE,GAAG,IAAI,EAAEP,EAAE,MAAM,GAAG,OAAOE,EAAE,KAAK,IAAIA,EAAE,CAAC,EAAE4X,EAAE,GAAGxX,GAAGA,GAAG,EAAE,IAAI,4CAA4CA,IAAI,EAAEwX,EAAE9X,EAAE,OAAO,EAAE,IAAI,+CAA+CA,EAAE,OAAO,EAAE8X,EAAE9X,EAAE,MAAM,KAAK,EAAE,IAAI,oDAAoDA,EAAE,MAAM,IAAI,EAAE8X,EAAE,EAAE,OAAO,EAAE,IAAI,4BAA4B,EAAEA,EAAE,EAAE,MAAM,KAAK,EAAE,IAAI,sDAAsD,cAAc,EAAE,MAAM,IAAI,EAAEA,EAAE,GAAG,GAAG,GAAG,EAAE,IAAI,4CAA4C,IAAI,EAAE,CAAC,cAAc5X,EAAE,aAAaI,EAAE,eAAeC,EAAE,aAAa,CAAC,CAAC,CAAC,SAASo0C,GAAG30C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,oBAAoB,SAAS,EAAE,EAAEgsB,EAAE,EAAE,SAAS,oBAAoB,SAAS,EAAE/pB,EAAEyyC,GAAG,EAAE,EAAEx0C,EAAEI,EAAEC,CAAC,EAAEL,EAAE+B,EAAE,cAAc3B,EAAE2B,EAAE,aAAa1B,EAAE0B,EAAE,eAAe,IAAIC,EAAE,CAAC,cAAchC,EAAE,aAAaI,EAAE,eAAeC,CAAC,EAAE,OAAO2qB,EAAE,UAAUxJ,GAAG,CAAC,MAAM,EAAE,OAAO,CAAC,EAAExf,CAAC,CAAC,CAAC,IAAI0yC,GAAGzoB,EAAE,CAAC,mBAAmBwoB,EAAE,CAAC,EAAE,SAASE,GAAG70C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEw0C,GAAG90C,EAAE,EAAEE,CAAC,EAAEK,EAAED,EAAE,EAAE,EAAEA,EAAE,GAAGA,EAAEN,EAAE,OAAOO,EAAE,EAAE,CAAC,CAAC,CAAC,SAASu0C,GAAG90C,EAAE,EAAEE,EAAE,CAAC,OAAO60C,GAAG/0C,EAAE,EAAEE,GAAG80C,EAAE,CAAC,CAAC,SAASA,GAAGh1C,EAAE,EAAE,CAAC,OAAOA,EAAE,EAAE,EAAEA,EAAE,EAAE,GAAG,CAAC,CAAC,SAAS+0C,GAAG/0C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAEC,EAAEP,EAAE,OAAO,EAAE,EAAE,EAAE,GAAG,KAAKM,EAAEC,GAAG,CAAC,EAAED,GAAGC,EAAED,IAAI,GAAG,IAAI2B,EAAE/B,EAAE,EAAEF,EAAE,EAAE,EAAEiC,EAAE,EAAE3B,EAAE,EAAE,GAAGC,EAAE,EAAE,EAAE,CAAC0B,EAAE,CAAC,OAAO,EAAE3B,EAAE,CAACA,EAAE,CAAC,CAAC,SAAS20C,GAAGj1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,OAAO20C,GAAGl1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,CAAC,SAAS40C,GAAGn1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO20C,GAAGl1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC,SAAS60C,GAAGp1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO20C,GAAGl1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,CAAC,SAAS20C,GAAGl1C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,GAAG0B,EAAE,GAAGC,EAAE,GAAG,CAAC,IAAIC,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,IAAI,EAAE,GAAG5B,GAAG4B,EAAE,KAAK,CAAC,MAAM,EAAE,GAAG,SAAS,EAAE,mBAAmB,CAAC,CAAC,EAAEA,EAAE,KAAKkzC,EAAE,EAAE,IAAIzyC,EAAE,EAAE,EAAE,IAAI,EAAE,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,KAAKD,EAAE,OAAO3C,GAAGiC,EAAE,OAAO,GAAG,CAAC,IAAI,EAAEA,EAAE,IAAI,EAAE,CAAC,MAAMoF,EAAE,SAAS,EAAE,mBAAmB5C,CAAC,EAAE,EAAE,GAAG4C,EAAEhH,EAAE,MAAM,IAAIqE,EAAE,GAAG,QAAQC,EAAEhC,EAAE,OAAO,EAAEgC,GAAGF,EAAE,EAAEE,EAAE,CAAC,IAAIC,EAAEwwC,GAAGt1C,EAAE,EAAE6C,EAAEgC,EAAE,EAAE,GAAGC,GAAGxE,EAAE,CAACsE,EAAE,GAAG,KAAK,CAAC,GAAG,EAAE,MAAM,EAAE,MAAM2wC,GAAGj1C,EAAEsC,EAAEkC,CAAC,EAAE,EAAE,OAAOvE,EAAE,KAAK,CAAC,EAAE,mBAAmBsC,EAAE,OAAO+B,IAAI,EAAE,QAAQ2C,GAAG1E,EAAE,KAAK,CAAC,EAAEC,EAAE,KAAK,EAAE,KAAK,GAAG,EAAE,MAAMvC,GAAGs0C,GAAG1yC,EAAE,EAAEkzC,EAAE,EAAE,CAAC,IAAItyC,EAAEF,EAAE,OAAOG,EAAE9C,EAAE6C,EAAEd,GAAGe,EAAE,IAAIH,EAAE,KAAK,GAAG,IAAI,MAAMG,CAAC,EAAE,KAAK,CAAC,CAAC,EAAEF,EAAE,KAAK,GAAG,IAAI,MAAME,CAAC,EAAE,KAAK,CAAC,CAAC,GAAG,IAAIkB,EAAE,CAAC,gBAAgBrB,CAAC,EAAE,OAAO,IAAIqB,EAAE,eAAepB,GAAGZ,IAAIgC,EAAE,aAAanB,GAAGmB,CAAC,CAAC,SAASoxC,GAAGt1C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,SAAS,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEO,EAAEP,EAAE,SAASE,EAAE,EAAEA,EAAE,EAAE,CAAC,EAAE,EAAE,KAAK,IAAII,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAE,KAAK,IAAIA,EAAE,GAAGA,EAAE,EAAE,EAAE2B,EAAE,KAAK,IAAI3B,EAAE,GAAGA,EAAE,EAAE,EAAE4B,EAAE,KAAK,IAAI5B,EAAE,GAAGA,EAAE,EAAE,EAAE6B,EAAE,KAAK,IAAI5B,EAAE,GAAGA,EAAE,EAAE,EAAEqC,EAAE,KAAK,IAAIrC,EAAE,GAAGA,EAAE,EAAE,EAAEsC,EAAE,KAAK,IAAItC,EAAE,GAAGA,EAAE,EAAE,EAAEuC,EAAE,KAAK,IAAIvC,EAAE,GAAGA,EAAE,EAAE,EAAEwC,GAAGd,EAAE,IAAIC,EAAE,GAAGc,GAAGH,EAAEV,IAAIW,EAAEF,GAAG,GAAGG,GAAG,GAAGC,GAAG,EAAE,MAAO,GAAE,IAAIkB,EAAE,KAAK,IAAI,EAAE/B,CAAC,EAAE,EAAE,KAAK,IAAI,EAAES,CAAC,EAAE2E,EAAE,KAAK,IAAItF,EAAEY,CAAC,EAAE,EAAE,KAAK,IAAIX,EAAEY,CAAC,EAAE6B,EAAE,KAAK,IAAI4C,EAAErD,EAAE,CAAC,EAAE,KAAK,IAAI,EAAE,EAAE,CAAC,EAAE,OAAOS,GAAG5B,EAAEC,EAAE2B,EAAE,CAAC,SAAS4wC,GAAGv1C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,KAAK,IAAI,EAAEJ,EAAEA,CAAC,EAAE,OAAOA,GAAGF,EAAEM,EAAE,CAAC,CAAC,SAAS+0C,GAAGr1C,EAAE,EAAE,CAAC,OAAOA,EAAE,MAAM,EAAE,OAAOA,EAAE,QAAQ,EAAE,OAAO,EAAE,SAASA,EAAE,QAAQ,CAAC,eAAew1C,GAAGx1C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,wBAAwB,EAAE,EAAEgsB,EAAE,EAAE,SAAS,wBAAwB,EAAE/pB,EAAEyyC,GAAG,EAAE,EAAEx0C,EAAEI,EAAEC,CAAC,EAAEL,EAAE+B,EAAE,cAAc3B,EAAE2B,EAAE,aAAa1B,EAAE0B,EAAE,eAAe,IAAIC,EAAE,MAAM,QAAQ,IAAI,CAAC,EAAE,KAAK,EAAE,EAAE,KAAK,CAAC,CAAC,EAAEC,EAAED,EAAE,GAAGU,EAAEV,EAAE,GAAG,CAAC,gBAAgBW,CAAC,EAAEoyC,GAAG9yC,EAAES,EAAE1C,EAAEI,EAAEC,CAAC,EAAE,OAAO,IAAIP,GAAG,EAAE,QAAQ,EAAE,IAAI,GAAG,EAAE,QAAQ,EAAEmwC,GAAGttC,EAAE,OAAO,CAAC,CAAC,IAAI4yC,GAAGD,GAAG,SAASE,GAAG11C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,mBAAmB,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,mBAAmB,EAAE9pB,EAAEwyC,GAAG,EAAEzyC,EAAE/B,EAAEI,EAAEC,EAAE,CAAC,EAAEL,EAAEgC,EAAE,cAAc5B,EAAE4B,EAAE,aAAa3B,EAAE2B,EAAE,eAAe,EAAEA,EAAE,aAAa,IAAIC,EAAE,CAAC,MAAM,EAAE,OAAOF,CAAC,EAAEW,EAAE,CAAC,cAAc1C,EAAE,aAAaI,EAAE,eAAeC,EAAE,aAAa,CAAC,EAAEsC,EAAEqoB,EAAE,UAAUtJ,GAAGzf,EAAES,CAAC,EAAE,MAAM,CAAC,gBAAgBC,EAAE,GAAG,eAAeA,EAAE,EAAE,CAAC,CAAC,IAAI8yC,GAAGxpB,EAAE,CAAC,4BAA4BupB,EAAE,CAAC,EAAE,eAAeE,GAAG51C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,wBAAwB,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,wBAAwB,EAAE9pB,EAAEwyC,GAAG,EAAEzyC,EAAE/B,EAAEI,EAAEC,EAAE,CAAC,EAAEL,EAAEgC,EAAE,cAAc5B,EAAE4B,EAAE,aAAa3B,EAAE2B,EAAE,eAAe,EAAEA,EAAE,aAAa,IAAIC,EAAE,MAAM,QAAQ,IAAI,CAAC,EAAE,KAAK,EAAEF,EAAE,KAAK,CAAC,CAAC,EAAEW,EAAET,EAAE,GAAGU,EAAEV,EAAE,GAAG,CAAC,gBAAgBW,EAAE,eAAeC,CAAC,EAAEqyC,GAAGxyC,EAAEC,EAAE3C,EAAEI,EAAEC,EAAE,CAAC,EAAE,OAAO,IAAIP,GAAG,EAAE,QAAQ,EAAEiC,IAAI,GAAGA,EAAE,QAAQ,EAAE,CAAC,gBAAgBkuC,GAAGrtC,EAAE,OAAO,EAAE,eAAeqtC,GAAGptC,CAAC,CAAC,CAAC,CAAC,IAAI8yC,GAAGD,GAAG,SAASE,GAAG91C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,EAAE,GAAG,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,mBAAmB,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,mBAAmB,EAAE9pB,EAAEwyC,GAAG,EAAEzyC,EAAE/B,EAAEI,EAAEC,EAAE,IAAI,EAAE4B,EAAED,EAAE,cAAcU,EAAEV,EAAE,aAAaW,EAAEX,EAAE,eAAeY,EAAE,CAAC,MAAM,EAAE,OAAOb,CAAC,EAAEc,EAAE,CAAC,cAAcZ,EAAE,aAAaS,EAAE,eAAeC,EAAE,mBAAmB,CAAC,EAAEG,EAAEkoB,EAAE,UAAUvJ,GAAG7e,EAAEC,CAAC,EAAE,MAAM,CAAC,gBAAgBC,EAAE,GAAG,aAAaA,EAAE,EAAE,CAAC,CAAC,IAAI+yC,GAAG5pB,EAAE,CAAC,yBAAyB2pB,EAAE,CAAC,EAAE,eAAeE,GAAGh2C,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,OAAO,kBAAkB,EAAE,GAAG,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,wBAAwB,EAAEiC,EAAE+pB,EAAE,EAAE,SAAS,wBAAwB,EAAE9pB,EAAEwyC,GAAG,EAAEzyC,EAAE/B,EAAEI,EAAEC,EAAE,IAAI,EAAE4B,EAAED,EAAE,cAAcU,EAAEV,EAAE,aAAaW,EAAEX,EAAE,eAAe,CAACY,EAAEC,CAAC,EAAE,MAAM,QAAQ,IAAI,CAAC,EAAE,KAAK,EAAEd,EAAE,KAAK,CAAC,CAAC,EAAE,CAAC,gBAAgBe,EAAE,aAAakB,CAAC,EAAEixC,GAAGryC,EAAEC,EAAEZ,EAAES,EAAEC,EAAE,CAAC,EAAE,OAAO,IAAI7C,GAAG,EAAE,QAAQ,EAAEiC,IAAI,GAAGA,EAAE,QAAQ,EAAE,CAAC,gBAAgBkuC,GAAGntC,EAAE,OAAO,EAAE,aAAa6/B,GAAG3+B,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI+xC,GAAGD,GAAG,SAASE,GAAGl2C,EAAE,EAAEE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,SAAS,gBAAgB,EAAE8X,EAAEvX,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,gEAAgEA,EAAE,OAAO,EAAEuX,EAAE,EAAE,SAAS,EAAE,IAAI,6DAA6D,IAAI,EAAEA,EAAExX,IAAI,IAAIJ,IAAI,GAAG,IAAI,mFAAmF,EAAE,IAAI,EAAEK,EAAE,EAAE,GAAGA,EAAE,OAAO,IAAI,EAAE,GAAG,EAAE27B,EAAE37B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,GAAG,CAAC,EAAE,EAAE0B,EAAE,CAAC,OAAO,CAAC,EAAEC,EAAE,CAAC,aAAahC,EAAE,iBAAiBI,EAAE,KAAK,CAAC,EAAE6B,EAAE+oB,EAAE,UAAUnI,GAAG9gB,EAAEC,CAAC,EAAE,OAAO,EAAEg6B,EAAE/5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIg0C,GAAGhqB,EAAE,CAAC,gBAAgB+pB,EAAE,CAAC,EAAE,SAASE,GAAGp2C,EAAE,EAAEE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,SAAS,uBAAuB,EAAE8X,EAAEvX,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,IAAI,uEAAuEA,EAAE,OAAO,EAAEuX,EAAE,EAAE,SAAS,EAAE,IAAI,oEAAoE,IAAI,EAAEA,EAAEvX,EAAE,QAAQ,WAAWA,EAAE,QAAQ,QAAQ,IAAI,kDAAkD,EAAEuX,EAAExX,IAAI,IAAIJ,IAAI,GAAG,IAAI,0FAA0F,EAAE,IAAI,EAAEK,EAAE,EAAE,GAAGA,EAAE,OAAO,IAAI,EAAE,GAAG,EAAE27B,EAAE37B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,GAAG,CAAC,EAAE,EAAE0B,EAAE,CAAC,OAAO,CAAC,EAAEC,EAAE,CAAC,aAAahC,EAAE,iBAAiBI,EAAE,KAAK,CAAC,EAAE6B,EAAE+oB,EAAE,UAAUrI,GAAG5gB,EAAEC,CAAC,EAAE,OAAO,EAAEg6B,EAAE/5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIk0C,GAAGlqB,EAAE,CAAC,uBAAuBiqB,EAAE,CAAC,EAAE,SAASE,GAAGt2C,EAAE,EAAE,SAASE,EAAE,GAAGI,EAAE,GAAG,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,QAAQ,WAAW,EAAE,EAAE,MAAM,EAAE,KAAKiC,EAAE,KAAKC,EAAE3B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG4B,EAAEi3B,EAAE+W,GAAG,CAAC7vC,CAAC,CAAC,EAAE,GAAG,EAAEsC,EAAEC,EAAEC,EAAEC,EAAE,GAAG+U,EAAEvX,EAAE,OAAO,EAAE,IAAI,yDAAyDA,EAAE,OAAO,EAAEuX,EAAEvX,EAAE,MAAM,KAAK,GAAGA,EAAE,MAAM,KAAK,EAAE,IAAI,0EAA0EA,EAAE,MAAM,KAAK,EAAEuX,EAAEvX,EAAE,QAAQ,SAASA,EAAE,QAAQ,UAAU,IAAI,sEAAsEA,EAAE,QAAQ,EAAEuX,EAAE,IAAI,QAAQ,IAAI,SAAS,IAAI,0CAA0C,GAAG,EAAEvX,EAAE,MAAM,KAAK,EAAE,CAAC,CAACqC,EAAEC,EAAEC,CAAC,EAAEssC,GAAG7uC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,IAAI4D,EAAEi1B,EAAEx2B,EAAE,CAAC,EAAE2E,EAAE6xB,EAAEv2B,EAAE,CAAC,EAAEkC,EAAEq0B,EAAEt2B,EAAEb,CAAC,EAAEc,EAAE+1B,EAAEA,EAAE30B,EAAEoD,CAAC,EAAExC,CAAC,CAAC,MAAMhC,EAAE/C,EAAE,GAAG,IAAI,OAAO,CAAC,IAAImE,EAAE05B,GAAGpN,GAAE2c,GAAGrqC,CAAC,EAAE,OAAO,EAAEwpB,GAAG,CAAC,CAAC,EAAE,GAAG,EAAEpqB,EAAEo0C,GAAGpyC,EAAEjC,CAAC,CAAC,CAAC,IAAIc,EAAE9C,EAAEolC,GAAGviC,EAAEZ,CAAC,EAAEqiC,GAAGzhC,EAAEZ,CAAC,EAAE,OAAOsuB,GAAE2I,EAAEp2B,EAAE,GAAG,EAAE,OAAO,CAAC,CAAC,SAASuzC,GAAGv2C,EAAE,EAAE,CAAC,IAAIE,EAAEiwC,GAAG,CAAC,EAAE,CAAC,EAAE7vC,EAAE6vC,GAAG,CAAC,CAAC,CAAC,EAAE5vC,EAAE4vC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,EAAEluC,EAAEC,EAAEC,EAAES,EAAE,QAAQC,EAAE,EAAEA,EAAE7C,EAAE,KAAK,EAAE6C,IAAI,CAAC,EAAE+5B,GAAG58B,EAAE,EAAE6C,EAAE,CAAC,EAAE,EAAE+5B,GAAG58B,EAAE6C,EAAE,CAAC,EAAEV,EAAE+2B,GAAGiK,GAAG,CAAC,EAAE,CAAC,EAAEvgC,EAAEs2B,GAAGiK,GAAG,CAAC,EAAE,CAAC,EAAE,IAAIrgC,EAAEqgC,GAAG/J,EAAE,EAAE8S,GAAG,EAAE,EAAE,IAAI,CAAC,CAAC,EAAEjqC,EAAEi3B,GAAGp2B,EAAEqgC,GAAG,CAAC,CAAC,EAAE,IAAIpgC,EAAEq7B,GAAG,EAAE,MAAM,EAAE,IAAI,EAAEp7B,EAAE81B,EAAEoT,GAAG,EAAE,EAAE,IAAI,EAAEnpC,CAAC,EAAEmB,EAAEk1B,EAAE,EAAEp2B,CAAC,EAAEd,EAAEg3B,GAAGiK,GAAGj/B,CAAC,EAAEi/B,GAAG,CAAC,CAAC,EAAE,IAAI,EAAEuD,GAAGzkC,EAAEC,CAAC,EAAEqF,EAAEm/B,GAAGzkC,EAAEC,CAAC,EAAE,EAAEk3B,EAAEj3B,EAAES,CAAC,EAAErC,EAAE64B,EAAEA,EAAE,EAAE,CAAC,EAAE7xB,CAAC,EAAE,IAAI5C,EAAE6/B,GAAGjkC,EAAED,CAAC,EAAEA,EAAE4gC,GAAGv8B,EAAEpE,EAAED,CAAC,EAAEJ,EAAEghC,GAAGv8B,EAAEwrC,GAAG,CAACttC,CAAC,CAAC,EAAE3C,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,IAAIs2C,GAAGrqB,EAAE,CAAC,WAAWmqB,EAAE,CAAC,EAAE,SAASG,GAAGz2C,EAAE,EAAEE,EAAE,UAAUI,EAAE,WAAWC,EAAE,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,QAAQ,YAAY,SAAS,EAAEiC,EAAE+pB,EAAE,EAAE,aAAa,YAAY,SAAS,EAAElU,EAAE,EAAE,OAAO,EAAE,IAAI,yDAAyD,EAAE,OAAO,EAAEA,EAAE7V,EAAE,OAAO,IAAIA,EAAE,MAAM,KAAK,EAAE,MAAM,IAAIA,EAAE,MAAM,KAAK,IAAIA,EAAE,MAAM,KAAK,EAAE,IAAI,kEAAkE,EAAE6V,EAAE,GAAG,MAAM,EAAE,SAAS,EAAE,IAAI,4EAA4E,IAAI,EAAE,IAAI5V,EAAE,CAAC,MAAM,EAAE,WAAWD,CAAC,EAAEE,EAAE,CAAC,cAAcjC,EAAE,SAASI,EAAE,UAAUC,EAAE,YAAY,CAAC,EAAE,OAAO2qB,EAAE,UAAU9F,GAAGljB,EAAEC,CAAC,CAAC,CAAC,IAAIu0C,GAAGvqB,EAAE,CAAC,WAAWsqB,EAAE,CAAC,EAAE,SAASE,GAAG32C,EAAE,EAAEE,EAAE,CAAC4X,EAAE,EAAE,IAAI,EAAE,IAAI,gDAAgD,IAAI,EAAEA,EAAE5X,EAAE,IAAI,EAAE,IAAI,gDAAgDA,IAAI,EAAE,IAAII,EAAE0rB,EAAEhsB,EAAE,IAAI,UAAU,EAAE8X,EAAExX,EAAE,MAAM,EAAE,IAAI,4CAA4CA,EAAE,OAAO,EAAE,IAAIC,EAAED,EAAE,MAAM,CAAC,EAAE,CAAC,EAAEA,EAAE,MAAM,MAAM,EAAE,EAAE,GAAG,EAAE,GAAG,GAAG,MAAM,IAAI,MAAM,yBAAyB,mDAAmD,KAAK,EAAE,GAAG,EAAEJ,GAAG,GAAG,MAAM,IAAI,MAAM,yBAAyBA,sDAAsD,KAAK,EAAE,EAAE,IAAI,EAAE,GAAGA,EAAE,IAAIA,EAAE,GAAG,IAAI+B,EAAEi6B,EAAEgQ,GAAG,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,GAAG,CAAC,CAAC,EAAEhqC,EAAEgqC,GAAG,EAAE,EAAE,EAAE,OAAO,EAAE/pC,EAAEukC,GAAGzkC,EAAEC,CAAC,EAAEU,EAAEokC,GAAG1B,GAAGnjC,EAAE0gC,GAAG,CAAC,EAAE,OAAO,CAAC,EAAE6B,GAAGviC,EAAE0gC,GAAG,CAAC3iC,EAAE,OAAO,CAAC,CAAC,EAAE2C,EAAEwlC,GAAG,CAAC,EAAE,CAAC,EAAE/nC,EAAE,KAAK,EAAE,OAAO47B,EAAE0T,GAAGqB,GAAG/U,EAAE57B,EAAE,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,IAAIwC,GAAGo+B,GAAGt+B,EAAEE,EAAED,CAAC,CAAC,CAAC,EAAEtC,CAAC,CAAC,CAAC,IAAIq2C,GAAGzqB,EAAE,CAAC,UAAUwqB,EAAE,CAAC,EAAE,SAASE,GAAG72C,EAAE,CAAC,IAAI,EAAE,GAAG,MAAM,QAAQA,CAAC,EAAE,CAAC,EAAE,GAAG8X,EAAE9X,GAAG,MAAMA,EAAE,OAAO,EAAE,IAAI,mEAAmE,EAAE,IAAIO,EAAEP,EAAE,GAAG,MAAM,GAAG,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAE8X,EAAE9X,EAAE,GAAG,MAAM,KAAKO,EAAE,IAAI,iEAAiEP,EAAE,GAAG,MAAM,UAAUO,IAAI,CAAC,MAAM,EAAE,GAAGP,EAAEovC,GAAGpvC,EAAEA,EAAE,MAAM,GAAG,CAAC,EAAE,IAAIO,GAAGmvC,GAAGnvC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEuX,EAAE9X,EAAE,QAAQA,EAAE,GAAG,MAAM,GAAG,IAAI,oCAAoCA,EAAE,yCAAyCA,EAAE,GAAG,MAAM,MAAM,EAAE,IAAIE,EAAE,CAAC,EAAEI,EAAEN,EAAE,QAAQO,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAEO,EAAEL,EAAE,KAAKgrB,EAAE,KAAK,IAAI,CAAC,IAAI,EAAE5qB,EAAEC,GAAG,GAAGA,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAEm3B,EAAE+J,GAAG/J,EAAEl5B,EAAE,GAAG,CAAC,CAAC,EAAEA,EAAE,EAAE,EAAE,EAAEwmC,GAAG,EAAEzkC,CAAC,CAAC,CAAC,OAAOi3B,GAAG,EAAEoK,GAAG,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,EAAE,OAAO,EAAEsM,GAAG1vC,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAI42C,GAAG3qB,EAAE,CAAC,aAAa0qB,EAAE,CAAC,EAAE,SAASE,GAAG/2C,EAAE,EAAE,GAAG,CAAC,GAAG8X,EAAE9X,EAAE,MAAM,EAAE,IAAI,gEAAgEA,EAAE,MAAM,EAAEA,EAAE,OAAO,EAAE,OAAOg3C,GAAGh3C,EAAE,CAAC,EAAE,CAAC,IAAIE,EAAEF,EAAE,MAAM,MAAM,EAAEA,EAAE,MAAM,OAAO,CAAC,EAAE,OAAO,CAACkC,EAAEC,IAAID,EAAEC,CAAC,EAAE7B,EAAE2wC,GAAG/U,EAAEl8B,EAAE,CAACE,EAAEF,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAGA,EAAE,MAAMA,EAAE,MAAM,OAAO,EAAE,CAAC,EAAE,CAAC,EAAEO,EAAE,CAAC,EAAE,EAAE,CAAC,EAAED,EAAE,QAAQ4B,GAAG,CAAC,GAAG,CAACC,EAAES,CAAC,EAAEo0C,GAAG90C,EAAE,CAAC,EAAE3B,EAAE,KAAK4B,CAAC,EAAE,EAAE,KAAKS,CAAC,CAAC,CAAC,EAAE,IAAI,EAAEs5B,EAAE0T,GAAGrvC,EAAE,CAAC,EAAEP,EAAE,KAAK,EAAEiC,EAAEi6B,EAAE0T,GAAG,EAAE,CAAC,EAAE5vC,EAAE,KAAK,EAAE,MAAM,CAAC,EAAEiC,CAAC,CAAC,CAAC,CAAC,SAAS+0C,GAAGh3C,EAAE,EAAE,GAAG,CAAC,OAAOkrB,EAAE,KAAK,IAAI,CAACpT,EAAE9X,EAAE,MAAM,SAAS,EAAE,IAAI,0CAA0CA,EAAE,MAAM,iBAAiB,EAAE,IAAIE,EAAEF,EAAE,MAAM,GAAGM,EAAEN,EAAE,MAAM,GAAGO,EAAE2jC,GAAGhkC,CAAC,EAAE,EAAEywB,GAAG3wB,CAAC,EAAE,EAAEowC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEnuC,EAAE0uB,GAAG,CAAC,EAAEzuB,EAAEhC,GAAGI,EAAEA,EAAEJ,EAAE,QAAQiC,EAAE,EAAEA,EAAED,EAAE,EAAEC,EAAE,CAAC,IAAIS,EAAE,EAAEC,EAAEZ,EAAEa,EAAEvC,EAAE,CAAC0B,EAAE,EAAE1B,CAAC,EAAE2qB,EAAE,KAAK,IAAI,CAAC,IAAInoB,EAAE65B,GAAG,EAAE,CAACz6B,EAAEA,CAAC,EAAE,CAACjC,EAAEiC,EAAE,CAAC,CAAC,EAAEa,EAAEsgC,GAAGvgC,CAAC,EAAEmB,EAAE04B,GAAG,EAAE,CAACz6B,EAAEA,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAE,EAAE++B,GAAGsD,GAAGtgC,EAAE,CAAC,EAAEksC,GAAG,CAAC,CAAC,EAAE,CAAC,CAAC,EAAEA,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE7oC,EAAEm/B,GAAGxiC,EAAEk1B,EAAE,EAAEp2B,CAAC,CAAC,EAAE,EAAEk2B,GAAGn2B,EAAEwE,CAAC,EAAE,EAAE,MAAM,KAAK,EAAEtF,EAAE0uB,GAAG,CAAC,EAAE1uB,EAAEu6B,GAAG,CAAC,EAAEI,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,MAAM,GAAG,EAAE,EAAE,MAAM,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,IAAIj4B,EAAE8vB,GAAGyE,GAAGpG,GAAG,EAAEvrB,CAAC,EAAEvE,CAAC,CAAC,EAAE4B,EAAEg4B,GAAG,EAAE,CAACz6B,EAAE,CAAC,EAAE,CAACjC,EAAEiC,EAAE7B,CAAC,CAAC,EAAEuE,EAAEu0B,EAAEz0B,EAAE1C,CAAC,EAAE6C,EAAE+vB,GAAG5yB,CAAC,EAAE,GAAGE,IAAI,EAAE,EAAEukC,GAAG9hC,EAAEkuB,GAAGjuB,EAAEiuB,GAAGhuB,EAAEF,CAAC,CAAC,CAAC,MAAM,CAAC,IAAI6C,EAAEi/B,GAAG9hC,EAAEkuB,GAAGjuB,EAAEiuB,GAAGhuB,EAAEF,CAAC,CAAC,CAAC,EAAE,EAAE43B,GAAG,CAACI,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAACz6B,EAAE7B,CAAC,CAAC,EAAEmH,CAAC,EAAE,CAAC,CAAC,CAAC,IAAID,EAAEqtB,GAAGhwB,CAAC,EAAEG,EAAE43B,GAAGr8B,EAAE,CAAC,EAAE4B,CAAC,EAAE,CAACjC,EAAEK,EAAE,MAAM,GAAG4B,CAAC,CAAC,EAAE,GAAGA,IAAI,EAAE5B,EAAEmmC,GAAG1hC,EAAE8tB,GAAGA,GAAG9tB,EAAE/C,CAAC,EAAEuF,CAAC,CAAC,MAAM,CAAC,IAAIC,EAAEi/B,GAAG1hC,EAAE8tB,GAAGA,GAAG9tB,EAAE/C,CAAC,EAAEuF,CAAC,CAAC,EAAEjH,EAAEi8B,GAAG,CAACI,GAAGr8B,EAAE,CAAC,EAAE,CAAC,EAAE,CAACL,EAAEiC,CAAC,CAAC,EAAEsF,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAACxF,EAAE,EAAE1B,CAAC,CAAC,CAAC,EAAEmzB,GAAG,CAAC9wB,EAAEC,EAAEC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG5C,EAAEI,IAAIC,EAAEq8B,GAAGr8B,EAAE,CAAC,EAAE,CAAC,EAAE,CAACL,EAAEI,CAAC,CAAC,EAAE,EAAEs8B,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,CAACt8B,EAAEA,CAAC,CAAC,GAAG,CAACC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI02C,GAAG9qB,EAAE,CAAC,IAAI4qB,EAAE,CAAC,EAAMG,IAAI,SAASl3C,EAAE,CAACA,EAAEA,EAAE,KAAK,GAAG,OAAOA,EAAEA,EAAE,KAAK,GAAG,OAAOA,EAAEA,EAAE,IAAI,GAAG,MAAMA,EAAEA,EAAE,uBAAuB,GAAG,wBAAwB,GAAGk3C,KAAKA,GAAG,CAAC,EAAE,EAAE,SAASC,GAAGn3C,EAAE,EAAEE,EAAEg3C,GAAG,uBAAuB,CAAC,IAAI52C,EAAE0rB,EAAEhsB,EAAE,SAAS,qBAAqB,EAAEO,EAAE,KAAK,GAAG,OAAOA,EAAEyrB,EAAE,EAAE,UAAU,qBAAqB,GAAG,IAAI,EAAEzrB,GAAG,KAAKD,EAAE84B,EAAE94B,EAAEC,CAAC,EAAE,GAAGL,IAAIg3C,GAAG,KAAK,OAAO,EAAE,GAAGh3C,IAAIg3C,GAAG,IAAI,OAAO/T,GAAG,CAAC,EAAE,GAAGjjC,IAAIg3C,GAAG,KAAK,CAAC,GAAG32C,GAAG,KAAK,OAAO6nC,GAAG,CAAC,EAAE,CAAC,IAAI,EAAE9nC,EAAE,KAAKC,EAAE,KAAK0B,EAAEi3B,GAAGiK,GAAG,CAAC,EAAEA,GAAG5iC,CAAC,CAAC,EAAE,OAAO,EAAE,EAAE24B,GAAGj3B,EAAE4gC,GAAG,CAAC,CAAC,EAAE5gC,CAAC,CAAC,CAAC,GAAG/B,IAAIg3C,GAAG,uBAAuB,CAAC,GAAG32C,GAAG,KAAK,OAAO24B,GAAGiK,GAAG,CAAC,EAAEN,GAAGviC,EAAE,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE84B,EAAE74B,EAAE+nC,GAAGhoC,EAAE,KAAK,CAAC,EAAE2B,EAAEwuB,GAAE0S,GAAGkG,GAAG,EAAExG,GAAG,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,OAAO3J,GAAGiK,GAAG,CAAC,EAAElhC,CAAC,CAAC,CAAC,CAAC,MAAM,MAAM,sBAAsB/B,GAAG,CAAC,CAAC,IAAIk3C,GAAGjrB,EAAE,CAAC,qBAAqBgrB,EAAE,CAAC,EAAE,SAASE,GAAGr3C,EAAE,EAAEE,EAAEI,EAAE42C,GAAG,uBAAuB,CAAC,IAAI32C,EAAEyrB,EAAEhsB,EAAE,SAAS,oBAAoB,EAAE,EAAEgsB,EAAE,EAAE,cAAc,oBAAoB,EAAE,EAAE,KAAK9rB,GAAG,OAAO,EAAE8rB,EAAE9rB,EAAE,UAAU,oBAAoB,GAAG6X,GAAGxX,EAAE,MAAM,EAAE,MAAM,+BAA+B,EAAE,IAAI0B,EAAEq3B,GAAGoN,GAAGnmC,EAAE,CAAC,CAAC,EAAE,OAAO62C,GAAGn1C,EAAE,EAAE3B,CAAC,CAAC,CAAC,IAAIg3C,GAAGnrB,EAAE,CAAC,oBAAoBkrB,EAAE,CAAC,EAAE,SAASE,GAAGv3C,EAAE,EAAEE,EAAEI,EAAEC,EAAE22C,GAAG,uBAAuB,CAAC,IAAI,EAAElrB,EAAEhsB,EAAE,SAAS,gBAAgB,EAAE,EAAEgsB,EAAE,EAAE,cAAc,gBAAgB,EAAE/pB,EAAE,KAAK3B,GAAG,OAAO2B,EAAE+pB,EAAE1rB,EAAE,UAAU,gBAAgB,GAAGyX,GAAG,EAAE,MAAM,EAAE,MAAM,2BAA2B,EAAE,IAAI7V,EAAE2gC,GAAG,CAAC,EAAE1gC,EAAEukC,GAAGxkC,EAAEihC,GAAG/J,EAAE,EAAE,CAAC,EAAEl5B,EAAE,EAAE,CAAC,EAAE,OAAOk3C,GAAGj1C,EAAEF,EAAE1B,CAAC,CAAC,CAAC,IAAIi3C,GAAGrrB,EAAE,CAAC,gBAAgBorB,EAAE,CAAC,EAAE,SAASE,GAAGz3C,EAAE,EAAEE,EAAEI,EAAE42C,GAAG,uBAAuB,CAAC,IAAI32C,EAAEyrB,EAAEhsB,EAAE,SAAS,WAAW,EAAE,EAAEgsB,EAAE,EAAE,cAAc,WAAW,EAAE,EAAE,KAAK9rB,GAAG,OAAO,EAAE8rB,EAAE9rB,EAAE,UAAU,WAAW,GAAG6X,GAAGxX,EAAE,MAAM,EAAE,MAAM,sBAAsB,EAAE,IAAI0B,EAAE4gC,GAAG,CAAC,EAAEtiC,EAAEmmC,GAAGtN,EAAEyJ,GAAG,CAAC,EAAEtiC,CAAC,EAAE0B,CAAC,EAAE,IAAIC,EAAEoqC,GAAG5F,GAAGzkC,EAAEm3B,EAAE74B,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO62C,GAAGl1C,EAAE,EAAE5B,CAAC,CAAC,CAAC,IAAIo3C,GAAGvrB,EAAE,CAAC,WAAWsrB,EAAE,CAAC,EAAE,SAASE,GAAG33C,EAAE,EAAEE,EAAEI,EAAE,EAAEC,EAAE22C,GAAG,uBAAuB,CAAC,IAAI,EAAElrB,EAAEhsB,EAAE,SAAS,WAAW,EAAE,EAAEgsB,EAAE,EAAE,cAAc,WAAW,EAAE/pB,EAAE,KAAK/B,GAAG,OAAO+B,EAAE+pB,EAAE9rB,EAAE,UAAU,WAAW,GAAG6X,GAAG,EAAE,MAAM,EAAE,MAAM,sBAAsB,EAAE,IAAI7V,EAAE2gC,GAAGviC,CAAC,EAAE6B,EAAEm3B,GAAGoN,GAAG,EAAE,CAAC,CAAC,EAAE9jC,EAAE6lC,GAAGtmC,EAAED,CAAC,EAAEW,EAAE6jC,GAAGvkC,EAAES,CAAC,EAAEE,EAAEg2B,EAAEM,EAAEyJ,GAAG,EAAE,EAAEI,GAAGrgC,CAAC,CAAC,EAAEw2B,EAAEl3B,EAAEW,CAAC,CAAC,EAAE,OAAOu0C,GAAGt0C,EAAEb,EAAE1B,CAAC,CAAC,CAAC,IAAIq3C,GAAGzrB,EAAE,CAAC,WAAWwrB,EAAE,CAAC,EAAE,SAASE,GAAG73C,EAAE,EAAEE,EAAEI,EAAE,KAAKC,EAAE22C,GAAG,uBAAuB,CAAC,IAAI,EAAElrB,EAAEhsB,EAAE,SAAS,SAAS,EAAE,EAAEgsB,EAAE,EAAE,cAAc,SAAS,EAAE/pB,EAAE,KAAK/B,GAAG,OAAO+B,EAAE+pB,EAAE9rB,EAAE,UAAU,SAAS,GAAG6X,GAAG,EAAE,MAAM,EAAE,MAAM,oBAAoB,EAAE,IAAI7V,EAAE2gC,GAAG,CAAC,EAAE1gC,EAAE0gC,GAAGviC,CAAC,EAAEsC,EAAE6xB,GAAG2E,EAAE,EAAEuM,GAAG7M,EAAE,EAAE32B,CAAC,CAAC,CAAC,CAAC,EAAEU,EAAEu2B,EAAEsN,GAAGxkC,EAAE,CAAC,EAAEyjC,GAAG7M,EAAE4N,GAAGxkC,EAAE,CAAC,EAAEC,CAAC,CAAC,CAAC,EAAEW,EAAE4jC,GAAG9jC,EAAEC,CAAC,EAAE,OAAOu0C,GAAGt0C,EAAEb,EAAE1B,CAAC,CAAC,CAAC,IAAIu3C,GAAG3rB,EAAE,CAAC,SAAS0rB,EAAE,CAAC,EAAE,SAASE,GAAG/3C,EAAE,EAAEE,EAAEI,EAAE42C,GAAG,uBAAuB,CAAC,IAAI32C,EAAEyrB,EAAEhsB,EAAE,SAAS,kBAAkB,EAAE,EAAEgsB,EAAE,EAAE,cAAc,kBAAkB,EAAE,EAAE,KAAK9rB,GAAG,OAAO,EAAE8rB,EAAE9rB,EAAE,UAAU,kBAAkB,GAAG6X,GAAGxX,EAAE,MAAM,EAAE,MAAM,6BAA6B,EAAE,IAAI0B,EAAEutC,GAAGjvC,EAAE,CAAC,EAAE,OAAO62C,GAAGn1C,EAAE,EAAE3B,CAAC,CAAC,CAAC,IAAI03C,GAAG7rB,EAAE,CAAC,kBAAkB4rB,EAAE,CAAC,EAAE,SAASE,GAAGj4C,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,SAAS,+BAA+B,EAAEM,EAAE0rB,EAAE,EAAE,SAAS,+BAA+B,EAAEjU,GAAG7X,EAAE,MAAMI,EAAE,MAAM,0CAA0C,EAAE,IAAIC,EAAE+rC,GAAGhsC,CAAC,EAAE,EAAE84B,EAAE94B,EAAEJ,CAAC,EAAE,EAAE2lC,GAAGnC,GAAGjP,GAAG6E,GAAGh5B,CAAC,CAAC,CAAC,CAAC,EAAE,OAAOw4B,EAAE4N,GAAGnmC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,SAAS23C,GAAGl4C,EAAE,EAAEE,EAAEI,EAAE,EAAEC,EAAE22C,GAAG,uBAAuB,CAAC,IAAI,EAAElrB,EAAEhsB,EAAE,mBAAmB,qBAAqB,EAAE,EAAEgsB,EAAE,EAAE,SAAS,qBAAqB,EAAE/pB,EAAE,KAAK,GAAG/B,GAAG,OAAO+B,EAAE+pB,EAAE9rB,EAAE,UAAU,qBAAqB,GAAG6X,GAAG,EAAE,MAAM,EAAE,MAAM,gCAAgC,EAAEzX,EAAE,EAAE,CAAC,IAAI6B,EAAE0gC,GAAGviC,CAAC,EAAEsC,EAAEigC,GAAG,CAAC,EAAEhgC,EAAEggC,GAAG,EAAE,EAAE,EAAE/J,EAAEM,EAAE,EAAEsN,GAAG9jC,EAAET,CAAC,CAAC,EAAEi3B,EAAEv2B,EAAEV,CAAC,CAAC,CAAC,CAAC,IAAID,EAAE+1C,GAAG,EAAE,CAAC,EAAE,OAAOb,GAAGl1C,EAAED,EAAE1B,CAAC,CAAC,CAAC,IAAI43C,GAAGhsB,EAAE,CAAC,qBAAqB+rB,EAAE,CAAC,EAAE,SAASE,GAAGp4C,EAAE,EAAEE,EAAE,GAAG,CAAC,GAAGA,IAAI,KAAKA,EAAE,EAAE,KAAK,GAAGA,IAAI,EAAE,KAAK,EAAE,MAAM,MAAM,mGAAmG,EAAE,oBAAoBA,GAAG,EAAE,OAAOkmC,GAAG,CAAC7lC,EAAEwB,EAAEC,IAAI,CAAC,IAAIE,EAAE4kC,GAAG/kC,EAAE,CAAC7B,CAAC,EAAE,EAAE,EAAEiC,EAAEukC,GAAGjW,GAAE1uB,EAAE,SAAS,EAAEG,CAAC,EAAEF,EAAE,CAACzB,EAAE4B,CAAC,CAAC,EAAE,IAAIS,EAAE6xB,GAAG2E,EAAEj3B,EAAE5B,CAAC,CAAC,EAAE,MAAM,CAAC,MAAM4iC,GAAGvgC,EAAE,CAAC1C,CAAC,CAAC,EAAE,SAAS,CAAC6C,EAAEC,IAAI,CAAC,GAAG,CAACkB,EAAEC,CAAC,EAAEnB,EAAEuE,EAAE26B,GAAGn/B,EAAE,MAAM,CAAC7C,CAAC,CAAC,EAAE,MAAM,CAACk5B,EAAE8C,EAAEn5B,EAAEwE,CAAC,EAAEm/B,GAAGjW,GAAEvsB,EAAE,SAAS,EAAEw/B,GAAGv/B,CAAC,CAAC,CAAC,EAAEi1B,EAAE8C,EAAEn5B,EAAEwE,CAAC,EAAEm/B,GAAGhD,GAAGv/B,CAAC,EAAEssB,GAAEvsB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAElE,EAAE,CAAC,CAAC,CAAC,SAASq4C,GAAGr4C,EAAE,EAAEE,EAAEI,EAAE,EAAEC,EAAE22C,GAAG,uBAAuB,CAAC,IAAI,EAAElrB,EAAEhsB,EAAE,eAAe,qBAAqB,EAAE,EAAEgsB,EAAE,EAAE,SAAS,qBAAqB,EAAE/pB,EAAE,KAAK,GAAG/B,GAAG,OAAO+B,EAAE+pB,EAAE9rB,EAAE,UAAU,qBAAqB,GAAG6X,GAAG,EAAE,MAAM,EAAE,MAAM,gCAAgC,EAAEzX,EAAE,EAAE,CAAC,IAAI6B,EAAE0gC,GAAGviC,CAAC,EAAEsC,EAAEigC,GAAG,CAAC,EAAEhgC,EAAEggC,GAAG,EAAE,MAAM,EAAE,EAAE,EAAE/J,EAAEM,EAAE,EAAEsN,GAAG9jC,EAAET,CAAC,CAAC,EAAE+2B,GAAG/2B,EAAEU,CAAC,CAAC,CAAC,CAAC,IAAIX,EAAEk2C,GAAG,EAAE,CAAC,EAAE,OAAOhB,GAAGl1C,EAAED,EAAE1B,CAAC,CAAC,CAAC,IAAI+3C,GAAGnsB,EAAE,CAAC,qBAAqBksB,EAAE,CAAC,EAAE,SAASE,GAAGv4C,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEyrB,EAAEhsB,EAAE,UAAU,sBAAsB,OAAO,EAAE,EAAEgsB,EAAE,EAAE,SAAS,qBAAqB,EAAE,EAAEA,EAAE9rB,EAAE,aAAa,sBAAsB,OAAO,EAAE+B,EAAE+pB,EAAE1rB,EAAE,eAAe,sBAAsB,EAAE,KAAK,EAAE,GAAGC,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,UACtnqDA,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,gDAAgD,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,qDAAqD,EAAE,OAAO,EAAE,GAAG0B,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,uDAAuDA,EAAE,OAAO,EAAE,IAAIC,EAAE,CAAC,QAAQ3B,EAAE,OAAO,EAAE,WAAW,EAAE,aAAa0B,CAAC,EAAEE,EAAE+oB,EAAE,UAAU9G,GAAGliB,CAAC,EAAE,MAAM,CAAC,cAAcC,EAAE,GAAG,aAAaA,EAAE,GAAG,kBAAkBA,EAAE,GAAG,gBAAgBA,EAAE,EAAE,CAAC,CAAC,IAAIq2C,GAAGrsB,EAAE,CAAC,qBAAqBosB,EAAE,CAAC,EAAE,SAASE,GAAGz4C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,eAAe,gBAAgB,OAAO,EAAEO,EAAEyrB,EAAE,EAAE,aAAa,gBAAgB,OAAO,EAAE,EAAEA,EAAE9rB,EAAE,WAAW,gBAAgB,OAAO,EAAE,GAAGI,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,UACnqBA,EAAE,OAAO,EAAE,GAAGC,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,qDAAqDA,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,mDAAmD,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,aAAaD,EAAE,WAAWC,EAAE,SAAS,CAAC,EAAE0B,EAAEipB,EAAE,UAAU7G,GAAG,CAAC,EAAE,MAAM,CAAC,cAAcpiB,EAAE,GAAG,YAAYA,EAAE,EAAE,CAAC,CAAC,IAAIy2C,GAAGvsB,EAAE,CAAC,eAAessB,EAAE,CAAC,EAAE,SAASE,GAAG34C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,OAAO,mBAAmB,EAAEO,EAAEyrB,EAAE,EAAE,UAAU,oBAAoB,OAAO,EAAE,EAAEA,EAAE9rB,EAAE,aAAa,oBAAoB,OAAO,EAAE,GAAGI,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,YAC7lBA,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,YACzC,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,KAAKD,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,OAAO2qB,EAAE,UAAU5G,GAAG,CAAC,CAAC,CAAC,IAAIs0B,GAAGzsB,EAAE,CAAC,mBAAmBwsB,EAAE,CAAC,EAAE,SAASE,GAAG74C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,OAAO,kBAAkB,EAAEO,EAAEyrB,EAAE,EAAE,UAAU,mBAAmB,OAAO,EAAE,EAAEA,EAAE9rB,EAAE,aAAa,mBAAmB,OAAO,EAAE,GAAGI,EAAE,KAAK,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,WACvXA,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM;AAAA,WACzC,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,KAAKD,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,OAAO2qB,EAAE,UAAU3G,GAAG,CAAC,CAAC,CAAC,IAAIu0B,GAAG3sB,EAAE,CAAC,kBAAkB0sB,EAAE,CAAC,EAAE,SAASE,GAAG/4C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAE,CAAC,IAAIC,EAAE8pB,EAAEhsB,EAAE,OAAO,eAAe,QAAQ,EAAE,GAAGkC,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,iCAAiC,EAAE,GAAGA,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,+BAA+BA,EAAE,OAAO,EAAE,IAAIC,EAAE6pB,EAAE,EAAE,aAAa,cAAc,EAAE,GAAG7pB,EAAE,QAAQ,QAAQ,MAAM,IAAI,MAAM,uCAAuC,EAAE,IAAIS,EAAE,CAAC,UAAU1C,EAAE,YAAYI,EAAE,QAAQC,EAAE,SAAS,EAAE,SAAS,EAAE,uBAAuB0B,CAAC,EAAEY,EAAE,CAAC,KAAKX,EAAE,WAAWC,CAAC,EAAEW,EAAEooB,EAAE,UAAUtG,GAAG/hB,EAAED,CAAC,EAAE,MAAM,CAAC,OAAOE,EAAE,GAAG,aAAaA,EAAE,EAAE,CAAC,CAAC,IAAIk2C,GAAG7sB,EAAE,CAAC,cAAc4sB,EAAE,CAAC,EAAE,SAASE,GAAGj5C,EAAE,EAAEE,EAAE,GAAG,CAAC,IAAII,EAAE0rB,EAAEhsB,EAAE,QAAQ,cAAc,QAAQ,EAAEO,EAAEyrB,EAAE,EAAE,YAAY,cAAc,QAAQ,EAAE,GAAG1rB,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,+CAA+CA,EAAE,OAAO,EAAE,GAAGC,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,mDAAmDA,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,UAAUL,CAAC,EAAE,EAAE,CAAC,MAAMI,EAAE,UAAUC,CAAC,EAAE0B,EAAEipB,EAAE,UAAUrG,GAAG,EAAE,CAAC,EAAE,MAAM,CAAC,QAAQ5iB,EAAE,GAAG,OAAOA,EAAE,GAAG,MAAMA,EAAE,EAAE,CAAC,CAAC,IAAIi3C,GAAG/sB,EAAE,CAAC,aAAa8sB,EAAE,CAAC,EAAE,SAASE,GAAGn5C,EAAE,EAAE,CAAC,IAAIE,EAAE8rB,EAAEhsB,EAAE,QAAQ,yBAAyB,QAAQ,EAAEM,EAAE,CAAC,WAAW,CAAC,EAAE,GAAG,GAAG,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,IAAIC,EAAE,CAAC,MAAML,CAAC,EAAE,OAAOgrB,EAAE,UAAUpG,GAAGvkB,EAAED,CAAC,CAAC,CAAC,IAAI84C,GAAGjtB,EAAE,CAAC,wBAAwBgtB,EAAE,CAAC,EAAME,GAAG,CAAC,IAAIvK,GAAG,KAAKE,GAAG,KAAKM,GAAG,MAAMJ,EAAE,EAAEoK,GAAG,CAAC,cAAc3F,GAAG,WAAWE,GAAG,MAAME,GAAG,KAAKE,EAAE,EAAEsF,GAAG,CAAC,cAAclF,GAAG,eAAeE,GAAG,sBAAsB8B,GAAG,eAAeF,GAAG,iBAAiB1B,GAAG,cAAcN,GAAG,kBAAkBS,GAAG,uBAAuBa,GAAG,2BAA2BE,GAAG,gCAAgCE,GAAG,wBAAwBE,GAAG,6BAA6BE,GAAG,UAAUO,GAAG,UAAUE,EAAE,EAAE8C,GAAG,CAAC,SAAS5C,GAAG,YAAYE,GAAG,GAAGG,EAAE,EAAEwC,GAAG,CAAC,mBAAmBnC,GAAG,oBAAoBF,GAAG,eAAeI,GAAG,UAAUE,GAAG,UAAUE,GAAG,QAAQE,GAAG,iBAAiBE,GAAG,oBAAoBG,GAAG,oBAAoBG,EAAE,EAAEoB,GAAG,CAAC,oBAAoBlB,GAAG,cAAcE,GAAG,kBAAkBE,GAAG,iBAAiBE,EAAE,EAAEa,GAAG,CAAC,aAAaX,GAAG,YAAYE,GAAG,uBAAuBE,EAAE,EAAMQ,GAAG,cAAcliB,EAAE,CAAC,SAAS52B,EAAEZ,EAAE,GAAG,EAAE,CAAC,GAAG,CAAC,MAAMK,EAAE,MAAMwB,CAAC,EAAE,KAAK,iBAAiBjB,EAAE,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC,IAAIkB,EAAE,EAAE,IAAIC,IAAI,CAAC,KAAKA,EAAE,KAAK,OAAOF,EAAEE,EAAE,KAAK,EAAE,EAAE,KAAK,eAAeD,CAAC,CAAC,MAAM,KAAK,eAAeD,CAAC,EAAE,OAAO2xB,GAAG3xB,CAAC,EAAE7B,EAAEK,GAAGA,EAAE,QAAQ,EAAE,KAAK,CAAC,IAAI,YAAY,CAAC,OAAO,KAAK,aAAa,OAAO,KAAK,YAAY,GAAG,KAAK,WAAW,CAAC,qBAAqB,CAAC,KAAK,YAAY,KAAK,WAAW,CAAC,CAAC,iBAAiBO,EAAEZ,EAAE,CAAC,OAAOimC,GAAGrlC,EAAEZ,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,aAAa,MAAMwzB,GAAG,KAAK,WAAW,CAAC,CAAC,MAAM,gBAAgB,CAAC,OAAO,KAAK,aAAa,OAAO,KAAK,YAAY,GAAG,CAAC,KAAK,OAAO,OAAOmP,GAAG,KAAK,YAAY,OAAO,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,MAAM,IAAI,MAAM,yDAAyD,CAAC,CAAC,MAAM,WAAW/hC,EAAE,CAAC,MAAM,IAAI,MAAM,4DAA4D,KAAK,aAAa,GAAG,CAAC,CAAC,MAAM,kBAAkBA,EAAE,CAAC,OAAO,KAAK,aAAa,MAAMA,EAAE,GAAG,OAAO,KAAK,GAAG,GAAGA,EAAE,MAAM,CAAC,CAAC,CAAC,EAAE,OAAO,eAAe84C,GAAG,OAAO,YAAY,CAAC,MAAM55C,GAAGA,EAAE,UAAU,MAAMA,EAAE,kBAAkB,MAAMA,EAAE,gBAAgB,IAAI,CAAC,EAAE,IAAI65C,GAAG,cAAcD,EAAE,CAAC,YAAY94C,EAAEZ,EAAE,EAAE,KAAK,CAAC,MAAM,EAAE,KAAK,aAAaY,EAAE,KAAK,IAAIZ,EAAE,KAAK,QAAQ,EAAE,KAAK,iBAAiB,CAAC,EAAE,KAAK,mBAAmB,CAAC,EAAE,GAAG,OAAO,KAAK,QAAQgrB,EAAE,QAAQ,QAAQ,EAAE,CAAC,eAAepqB,EAAE,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIR,GAAGA,EAAE,IAAI,EAAE,OAAO,KAAKQ,CAAC,GAAG,QAAQ,CAACR,EAAEC,IAAI,CAAC,IAAIwB,EAAEmpB,EAAE,oBAAoB5qB,GAAG0B,EAAE,GAAG,KAAK,iBAAiBzB,IAAI,OAAO,KAAK,iBAAiBA,GAAG,CAAC,aAAa,GAAGD,eAAe,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,KAAK,mBAAmBzB,IAAI,OAAO,KAAK,mBAAmBA,GAAG,CAAC,aAAa,GAAGD,cAAc,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,IAAIC,EAAE,MAAM,QAAQnB,CAAC,EAAEA,EAAEP,GAAG,OAAOO,EAAER,GAAG,GAAG2B,GAAG,KAAK,OAAO,IAAIC,EAAE,KAAK,iBAAiB3B,GAAG,SAAS4B,EAAE,KAAK,mBAAmB5B,GAAG,SAASkzB,EAAE,IAAI,CAAC,IAAI7wB,EAAEk2B,EAAEM,EAAEl3B,EAAE,KAAK,GAAG,EAAEk3B,EAAE6J,GAAGhhC,CAAC,EAAE,EAAE,KAAK,GAAG,CAAC,EAAEY,EAAEu2B,EAAEF,GAAG6J,GAAGjK,EAAE32B,EAAE,KAAK,OAAO,CAAC,EAAE4gC,GAAGjK,EAAE52B,EAAE,KAAK,OAAO,CAAC,CAAC,EAAED,CAAC,EAAEa,EAAEg2B,EAAEM,EAAEj3B,EAAE,KAAK,GAAG,EAAEi3B,EAAE6J,GAAGpgC,CAAC,EAAE,EAAE,KAAK,GAAG,CAAC,EAAEX,EAAE,OAAOU,CAAC,EAAET,EAAE,OAAOW,CAAC,EAAE,IAAIC,EAAE+1B,EAAEM,EAAEv2B,EAAE,CAAC,KAAK,YAAY,EAAEd,CAAC,EAAEA,EAAE,OAAOgB,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,oBAAoB,OAAO2wB,GAAG,KAAK,iBAAiB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE4yB,GAAG,KAAK,mBAAmB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,CAAC,MAAM,YAAY,CAAC,IAAIA,EAAE,CAAC,GAAG,KAAK,iBAAiB,GAAG,KAAK,kBAAkB,EAAE,MAAM,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAOA,EAAE,IAAIZ,IAAI,CAAC,KAAKA,EAAE,aAAa,OAAOA,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAE,CAACA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAE,IAAIZ,EAAEY,EAAE,OAAO,EAAE,EAAE,GAAG,KAAK,iBAAiBA,EAAE,MAAM,EAAEZ,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,EAAE,KAAK,mBAAmBO,EAAE,MAAMZ,EAAEA,EAAE,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,IAAI,KAAK,IAAI,QAAQ,KAAK,OAAO,CAAC,CAAC,OAAO,WAAWO,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,IAAIA,EAAE,OAAO,CAAC,CAAC,EAAE25C,GAAG,UAAU,WAAWjiB,GAAGiiB,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY94C,EAAEZ,EAAE,GAAG,CAAC,MAAM,EAAE,KAAK,aAAaY,EAAE,KAAK,wBAAwBZ,EAAE,KAAK,iBAAiB,CAAC,CAAC,CAAC,eAAeY,EAAE,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIR,GAAGA,EAAE,IAAI,EAAE,OAAO,KAAKQ,CAAC,GAAG,QAAQ,CAACR,EAAEC,IAAI,CAAC,IAAIwB,EAAEmpB,EAAE,oBAAoB5qB,GAAG,KAAK,iBAAiBC,IAAI,OAAO,KAAK,iBAAiBA,GAAG,CAAC,aAAa,GAAGD,gBAAgB,SAASmzB,EAAE,IAAI2K,GAAGr8B,EAAE,MAAM,KAAK,uBAAuB,EAAE,SAAS,EAAE,CAAC,CAAC,GAAG,IAAIC,EAAE,MAAM,QAAQlB,CAAC,EAAEA,EAAEP,GAAG,OAAOO,EAAER,GAAG,GAAG0B,GAAG,KAAK,OAAO,IAAIC,EAAE,KAAK,iBAAiB1B,GAAG,SAASkzB,EAAE,IAAI,CAAC,IAAIvxB,EAAE42B,EAAE72B,EAAEghC,GAAGjhC,CAAC,CAAC,EAAEC,EAAE,OAAOC,CAAC,EAAE,IAAIC,EAAE22B,EAAEM,EAAEF,GAAGl3B,EAAE+gC,GAAGjK,EAAE52B,EAAEgpB,EAAE,QAAQ,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC,KAAK,YAAY,EAAEnpB,CAAC,EAAEA,EAAE,OAAOI,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,kBAAkB,MAAMuxB,GAAG,KAAK,iBAAiB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,MAAM,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,KAAK,iBAAiB,IAAIA,IAAI,CAAC,KAAKA,EAAE,aAAa,OAAOA,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAE,CAACA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAE,IAAIZ,EAAE,GAAG,KAAK,iBAAiBY,EAAE,IAAI,IAAI,CAAC,aAAa,EAAE,KAAK,SAAS,EAAE,OAAO,SAASZ,CAAC,CAAC,EAAE,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,wBAAwB,KAAK,uBAAuB,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,uBAAuB,CAAC,CAAC,EAAE45C,GAAG,UAAU,UAAUliB,GAAGkiB,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,YAAY94C,EAAEZ,EAAE,EAAEK,EAAE,KAAK,CAAC,MAAM,EAAE,KAAK,aAAaO,EAAE,KAAK,MAAMZ,EAAE,KAAK,MAAM,EAAE,KAAK,QAAQK,EAAE,KAAK,uBAAuB,CAAC,EAAE,KAAK,wBAAwB,CAAC,EAAEkzB,EAAE,IAAI,CAAC,KAAK,SAASoP,GAAG3iC,CAAC,EAAE,SAAS,EAAE,KAAK,SAAS2iC,GAAG,CAAC,EAAE,SAAS,CAAC,CAAC,EAAEtiC,GAAG,OAAO,KAAK,QAAQ2qB,EAAE,QAAQ,QAAQ,EAAE,CAAC,eAAepqB,EAAE,CAAC,IAAIZ,EAAE,MAAM,QAAQY,CAAC,EAAEA,EAAE,IAAI,GAAG,EAAE,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE2yB,EAAE,IAAI,CAAC,IAAI,EAAEiT,GAAG,EAAE,KAAK,QAAQ,EAAEnmC,EAAEmmC,GAAG,EAAE,KAAK,QAAQ,EAAExmC,EAAE,QAAQ,CAAC6B,EAAEC,IAAI,CAAC,IAAIC,EAAEipB,EAAE,oBAAoBnpB,GAAGG,EAAE,GAAG,KAAK,uBAAuBF,IAAI,OAAO,KAAK,uBAAuBA,GAAG,CAAC,aAAa,GAAGD,MAAM,SAAS0xB,EAAE,IAAI2N,GAAGn/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,KAAK,wBAAwBF,IAAI,OAAO,KAAK,wBAAwBA,GAAG,CAAC,aAAa,GAAGD,MAAM,SAAS0xB,EAAE,IAAI2N,GAAGn/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,IAAI,EAAE,MAAM,QAAQpB,CAAC,EAAEA,EAAEkB,GAAG,OAAOlB,EAAEiB,GAAG,GAAG,GAAG,KAAK,OAAO,IAAIa,EAAE,KAAK,uBAAuBZ,GAAG,SAAS,EAAE,KAAK,wBAAwBA,GAAG,SAASc,EAAEg2B,EAAEM,EAAEx2B,EAAE,KAAK,KAAK,EAAEw2B,EAAE,EAAE,EAAE,KAAK,KAAK,CAAC,EAAEr2B,EAAE+1B,EAAEM,EAAE,EAAE,KAAK,KAAK,EAAEA,EAAE6J,GAAG,CAAC,EAAE,EAAE,KAAK,KAAK,CAAC,EAAEjgC,EAAEk2B,GAAGp2B,EAAE,CAAC,EAAEoB,EAAEg1B,GAAGn2B,EAAExC,CAAC,EAAEqC,EAAE,OAAOE,CAAC,EAAE,EAAE,OAAOC,CAAC,EAAE,IAAIoB,EAAE20B,EAAEM,EAAEF,GAAGl2B,EAAE81B,EAAEiK,GAAG7+B,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC,KAAK,YAAY,EAAEjC,CAAC,EAAEA,EAAE,OAAOkC,CAAC,CAAC,CAAC,EAAE,KAAK,SAAS,OAAOi1B,EAAE,KAAK,SAAS,KAAK,KAAK,CAAC,EAAE,KAAK,SAAS,OAAOA,EAAE,KAAK,SAAS,KAAK,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,SAAS,QAAQ,EAAE,KAAK,SAAS,QAAQ,EAAE,KAAK,wBAAwB,MAAM1F,GAAG,KAAK,uBAAuB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,yBAAyB,MAAM4yB,GAAG,KAAK,wBAAwB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,IAAIA,EAAE,CAAC,GAAG,KAAK,uBAAuB,GAAG,KAAK,uBAAuB,EAAE,MAAM,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAOA,EAAE,IAAIZ,IAAI,CAAC,KAAKA,EAAE,aAAa,OAAOA,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAE,CAACA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAE2yB,EAAE,IAAI,CAAC,KAAK,SAAS,OAAOmP,GAAG,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC,EAAE,KAAK,SAAS,OAAOA,GAAG,KAAK,MAAM,KAAK,YAAY,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI1iC,EAAEY,EAAE,OAAO,EAAE,EAAE,GAAG,KAAK,uBAAuBA,EAAE,MAAM,EAAEZ,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,EAAE,KAAK,wBAAwBO,EAAE,MAAMZ,EAAEA,EAAE,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,MAAM,KAAK,MAAM,MAAM,KAAK,MAAM,QAAQ,KAAK,OAAO,CAAC,CAAC,OAAO,WAAWO,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,MAAMA,EAAE,MAAMA,EAAE,OAAO,CAAC,CAAC,EAAE65C,GAAG,UAAU,OAAOniB,GAAGmiB,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAY94C,EAAEZ,EAAE,EAAEK,EAAE,KAAKwB,EAAE,EAAE,CAAC,MAAM,EAAE,KAAK,aAAajB,EAAE,KAAK,MAAMZ,EAAE,KAAK,MAAM,EAAE,KAAK,QAAQK,EAAE,KAAK,MAAMwB,EAAE,KAAK,uBAAuB,CAAC,EAAE,KAAK,2BAA2B,CAAC,EAAE0xB,EAAE,IAAI,CAAC,KAAK,UAAUoP,GAAG,CAAC,EAAE,SAAS,EAAE,KAAK,SAASA,GAAG3iC,CAAC,EAAE,SAAS,CAAC,CAAC,EAAEK,GAAG,OAAO,KAAK,QAAQ2qB,EAAE,QAAQ,QAAQ,EAAE,CAAC,eAAepqB,EAAE,CAAC,IAAIZ,EAAE,MAAM,QAAQY,CAAC,EAAEA,EAAE,IAAI,GAAG,EAAE,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE2yB,EAAE,IAAI,CAAC,IAAI,EAAEiT,GAAG,EAAE,KAAK,QAAQ,EAAEnmC,EAAE24B,GAAG,CAAC,KAAK,aAAaJ,EAAEM,EAAE,KAAK,UAAU,KAAK,KAAK,EAAE,CAAC,CAAC,EAAEl5B,EAAE,QAAQ,CAAC6B,EAAEC,IAAI,CAAC,IAAIC,EAAEipB,EAAE,oBAAoBnpB,GAAGG,EAAE,GAAG,KAAK,uBAAuBF,IAAI,OAAO,KAAK,uBAAuBA,GAAG,CAAC,aAAa,GAAGD,MAAM,SAASq/B,GAAGn/B,CAAC,EAAE,SAASC,CAAC,CAAC,GAAG,KAAK,2BAA2BF,IAAI,OAAO,KAAK,2BAA2BA,GAAG,CAAC,aAAa,GAAGD,MAAM,SAASq/B,GAAGn/B,CAAC,EAAE,SAASC,CAAC,CAAC,GAAG,IAAI,EAAE,MAAM,QAAQpB,CAAC,EAAEA,EAAEkB,GAAG,OAAOlB,EAAEiB,GAAG,GAAG,GAAG,KAAK,OAAO,IAAIa,EAAE,KAAK,uBAAuBZ,GAAG,SAAS,EAAE,KAAK,2BAA2BA,GAAG,SAASc,EAAEg2B,EAAEM,EAAEx2B,EAAE,KAAK,KAAK,EAAEw2B,EAAE,EAAE,EAAE,KAAK,KAAK,CAAC,EAAEr2B,EAAEq2B,EAAE,EAAE,KAAK,KAAK,EAAEp2B,EAAEs2B,GAAG,CAAC,EAAEp1B,EAAEgkC,GAAGnlC,EAAEC,CAAC,EAAEJ,EAAE,OAAOE,CAAC,EAAE,EAAE,OAAOoB,CAAC,EAAE,IAAIC,EAAE20B,EAAEM,EAAEF,GAAG34B,EAAE,CAAC,EAAE24B,GAAGp2B,EAAEg2B,EAAE50B,EAAE,KAAK,OAAO,CAAC,CAAC,EAAEjC,CAAC,EAAEA,EAAE,OAAOkC,CAAC,CAAC,CAAC,EAAE,KAAK,UAAU,OAAO20B,EAAE,KAAK,UAAU,CAAC,CAAC,EAAE,KAAK,SAAS,OAAOM,EAAE,KAAK,SAAS,KAAK,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,SAAS,QAAQ,EAAE,KAAK,UAAU,QAAQ,EAAE,KAAK,wBAAwB,MAAM1F,GAAG,KAAK,uBAAuB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,4BAA4B,MAAM4yB,GAAG,KAAK,2BAA2B,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,MAAM,IAAI,MAAM,iDAAiD,CAAC,CAAC,MAAM,WAAWA,EAAE,CAAC,MAAM,IAAI,MAAM,iDAAiD,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,MAAM,KAAK,MAAM,MAAM,KAAK,MAAM,QAAQ,KAAK,QAAQ,MAAM,KAAK,KAAK,CAAC,CAAC,OAAO,WAAWA,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,MAAMA,EAAE,MAAMA,EAAE,QAAQA,EAAE,KAAK,CAAC,CAAC,EAAE85C,GAAG,UAAU,SAASpiB,GAAGoiB,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,YAAY94C,EAAE,CAAC,MAAM,EAAE,KAAK,aAAaA,EAAE,KAAK,gBAAgBA,CAAC,CAAC,CAAC,eAAeA,EAAE,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIR,GAAGA,EAAE,IAAI,EAAE,OAAO,KAAKQ,CAAC,GAAG,QAAQ,CAACR,EAAEC,IAAI,CAAC,IAAIwB,EAAE,MAAM,QAAQjB,CAAC,EAAEA,EAAEP,GAAG,OAAOO,EAAER,GAAG,GAAGyB,GAAG,KAAK,OAAO,IAAIC,EAAEkpB,EAAE,oBAAoB5qB,GAAGmzB,EAAE,IAAI,CAAC,IAAIxxB,EAAE62B,EAAEM,EAAE,KAAK,EAAEr3B,CAAC,EAAEC,CAAC,EAAEA,EAAE,OAAOC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,gBAAgBnB,EAAE,CAAC,KAAK,aAAaA,EAAE,KAAK,GAAG,MAAM,KAAK,EAAE,QAAQ,EAAE,KAAK,EAAE6yB,GAAGkP,GAAG,CAAC/hC,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAC,MAAM,YAAY,CAAC,MAAM,CAAC,MAAM,KAAK,eAAe,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAE,CAAC,GAAGA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,+CAA+C,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,YAAY,CAAC,CAAC,OAAO,WAAWA,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,YAAY,CAAC,CAAC,EAAE+5C,GAAG,UAAU,MAAMriB,GAAGqiB,EAAE,EAAE,IAAIC,GAAG,cAAcD,EAAE,CAAC,YAAYn5C,EAAEZ,EAAE,EAAE,GAAG,CAAC,MAAMY,CAAC,EAAE,KAAK,aAAaA,EAAE,KAAK,SAASZ,EAAE,KAAK,YAAY,EAAE,KAAK,cAAc,CAAC,EAAE,KAAK,EAAE2iC,GAAG,KAAK,QAAQ,CAAC,CAAC,eAAe/hC,EAAE,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIR,GAAGA,EAAE,IAAI,EAAE,OAAO,KAAKQ,CAAC,GAAG,QAAQ,CAACR,EAAEC,IAAI,CAAC,IAAIwB,EAAEmpB,EAAE,oBAAoB5qB,GAAG,KAAK,cAAcC,IAAI,OAAO,KAAK,cAAcA,GAAG,CAAC,aAAa,GAAGD,aAAa,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,GAAG,IAAIC,EAAE,KAAK,cAAczB,GAAG,SAAS0B,EAAE,MAAM,QAAQnB,CAAC,EAAEA,EAAEP,GAAG,OAAOO,EAAER,GAAG2B,GAAG,MAAMwxB,EAAE,IAAI,CAAC,IAAIvxB,EAAEC,EAAE22B,EAAEM,EAAE,KAAK,EAAEp3B,CAAC,EAAEC,CAAC,EAAE,KAAK,YAAYC,EAAE42B,EAAEM,EAAE,KAAK,EAAEN,EAAE72B,EAAEm3B,EAAEj3B,EAAE,KAAK,CAAC,CAAC,CAAC,EAAEJ,CAAC,EAAEG,EAAE42B,EAAEM,EAAE,KAAK,EAAEj3B,CAAC,EAAEJ,CAAC,EAAEC,EAAE,OAAOG,CAAC,EAAEJ,EAAE,OAAOG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,EAAE,QAAQ,EAAE,KAAK,eAAe,MAAMwxB,GAAG,KAAK,cAAc,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,YAAYA,EAAE,CAAC,KAAK,SAASA,CAAC,CAAC,MAAM,YAAY,CAAC,MAAM,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,KAAK,cAAc,IAAIA,IAAI,CAAC,KAAKA,EAAE,aAAa,OAAOA,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAE,CAACA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAE,IAAIZ,EAAE,GAAG,KAAK,cAAcY,EAAE,IAAI,IAAI,CAAC,aAAa,EAAE,KAAK,SAAS,EAAE,OAAO,SAASZ,CAAC,CAAC,EAAE,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,SAAS,KAAK,SAAS,YAAY,KAAK,WAAW,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,SAASA,EAAE,WAAW,CAAC,CAAC,EAAEg6C,GAAG,UAAU,WAAWtiB,GAAGsiB,EAAE,EAAE,IAAIC,GAAG,cAAcP,EAAE,CAAC,YAAY94C,EAAEZ,EAAE,GAAG,EAAE,EAAEK,EAAE,KAAKwB,EAAE,GAAG,CAAC,GAAG,MAAM,EAAE,KAAK,aAAajB,EAAE,KAAK,MAAMZ,EAAE,KAAK,SAAS,EAAE,KAAK,QAAQK,EAAE,KAAK,uBAAuB,CAAC,EAAE,KAAK,mBAAmB,CAAC,EAAE,KAAK,qBAAqB,CAAC,EAAE,KAAK,SAASwB,EAAExB,GAAG,OAAO,KAAK,QAAQ2qB,EAAE,QAAQ,QAAQ,GAAGpqB,GAAG,KAAK,MAAM,IAAI,MAAM,oDAAoD,CAAC,CAAC,eAAeA,EAAE,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIR,GAAGA,EAAE,IAAI,EAAE,OAAO,KAAKQ,CAAC,GAAG,QAAQ,CAACR,EAAEC,IAAI,CAAC,IAAIwB,EAAEmpB,EAAE,oBAAoB5qB,GAAG0B,EAAE,GAAG,KAAK,uBAAuBzB,IAAI,OAAO,KAAK,uBAAuBA,GAAG,CAAC,aAAa,GAAGD,QAAQ,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,KAAK,mBAAmBzB,IAAI,OAAO,KAAK,mBAAmBA,GAAG,CAAC,aAAa,GAAGD,aAAa,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,KAAK,qBAAqBzB,IAAI,MAAM,KAAK,WAAW,KAAK,qBAAqBA,GAAG,CAAC,aAAa,GAAGD,OAAO,SAASmzB,EAAE,IAAI2N,GAAGr/B,CAAC,EAAE,SAASC,CAAC,CAAC,CAAC,GAAG,IAAIC,EAAE,MAAM,QAAQnB,CAAC,EAAEA,EAAEP,GAAG,OAAOO,EAAER,GAAG,GAAG2B,GAAG,KAAK,OAAO,IAAIC,EAAE,KAAK,uBAAuB3B,GAAG,SAAS4B,EAAE,KAAK,mBAAmB5B,GAAG,SAASkzB,EAAE,IAAI,CAAC,IAAI7wB,EAAEk2B,EAAEM,EAAEl3B,EAAE,KAAK,KAAK,EAAEk3B,EAAE6J,GAAGhhC,CAAC,EAAE,EAAE,KAAK,KAAK,CAAC,EAAE,GAAG,KAAK,SAAS,CAAC,IAAIY,EAAE,KAAK,qBAAqBtC,GAAG,SAASuC,EAAEg2B,EAAEM,EAAEv2B,EAAE,KAAK,KAAK,EAAEu2B,EAAEn3B,EAAE,EAAE,KAAK,KAAK,CAAC,EAAEc,EAAEm2B,GAAGE,EAAEn3B,EAAE,KAAK,YAAY,EAAE8gC,GAAG2D,GAAG9jC,EAAEk2B,EAAEmK,GAAGngC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC,CAAC,EAAEE,EAAE81B,EAAEM,EAAEj3B,EAAE,KAAK,QAAQ,EAAEY,CAAC,EAAEb,EAAE,OAAOU,CAAC,EAAEC,EAAE,OAAOC,CAAC,EAAEX,EAAE,OAAOa,CAAC,EAAE,IAAI,EAAE0jC,GAAG3kC,EAAEiB,CAAC,EAAEjB,EAAE,OAAO,CAAC,CAAC,KAAK,CAAC,IAAIc,EAAEi2B,EAAEM,EAAEl3B,EAAE,KAAK,KAAK,EAAEk3B,EAAE6J,GAAGhhC,CAAC,EAAE,EAAE,KAAK,KAAK,CAAC,EAAEa,EAAEg2B,EAAEM,EAAEj3B,EAAE,KAAK,QAAQ,EAAE+2B,GAAGE,EAAEn3B,EAAE,KAAK,YAAY,EAAE8gC,GAAGjK,EAAEj2B,EAAE,KAAK,OAAO,CAAC,CAAC,CAAC,EAAEX,EAAE,OAAOW,CAAC,EAAEV,EAAE,OAAOW,CAAC,EAAE,IAAIC,EAAE2jC,GAAG3kC,EAAEe,CAAC,EAAEf,EAAE,OAAOgB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,oBAAoB,CAAC,CAAC,SAAS,CAAC,KAAK,wBAAwB,MAAM2wB,GAAG,KAAK,uBAAuB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,sBAAsB,MAAM,KAAK,UAAU4yB,GAAG,KAAK,qBAAqB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,oBAAoB,MAAM4yB,GAAG,KAAK,mBAAmB,IAAI5yB,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,IAAIA,EAAE,CAAC,GAAG,KAAK,uBAAuB,GAAG,KAAK,kBAAkB,EAAE,OAAO,KAAK,UAAUA,EAAE,KAAK,GAAG,KAAK,oBAAoB,EAAE,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAOA,EAAE,IAAIZ,IAAI,CAAC,KAAKA,EAAE,aAAa,OAAOA,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAE,CAACA,EAAE,MAAM,KAAK,kBAAkBA,CAAC,EAAE,IAAIZ,EAAE,KAAK,SAASY,EAAE,OAAO,EAAEA,EAAE,OAAO,EAAE,EAAE,GAAG,KAAK,uBAAuBA,EAAE,MAAM,EAAEZ,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,EAAE,KAAK,mBAAmBO,EAAE,MAAMZ,EAAEA,EAAE,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,EAAE,KAAK,WAAW,KAAK,qBAAqBO,EAAE,MAAMZ,EAAE,EAAEA,EAAE,CAAC,EAAE,IAAIK,IAAI,CAAC,aAAaA,EAAE,KAAK,SAASA,EAAE,OAAO,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,WAAW,CAAC,MAAM,CAAC,aAAa,KAAK,aAAa,MAAM,KAAK,MAAM,SAAS,KAAK,SAAS,QAAQ,KAAK,QAAQ,SAAS,KAAK,QAAQ,CAAC,CAAC,OAAO,WAAWO,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,EAAE,aAAaA,EAAE,MAAMA,EAAE,SAASA,EAAE,QAAQA,EAAE,QAAQ,CAAC,CAAC,EAAEi6C,GAAG,UAAU,UAAUviB,GAAGuiB,EAAE,EAAE,IAAIC,GAAG,KAAK,CAAC,OAAO,IAAIt5C,EAAE,CAAC,OAAO,IAAIm5C,GAAGn5C,CAAC,CAAC,CAAC,OAAO,SAASA,EAAEZ,EAAE,EAAE,GAAG,CAAC,OAAO,IAAIg6C,GAAGp5C,EAAEZ,EAAE,CAAC,CAAC,CAAC,OAAO,QAAQY,EAAEZ,EAAE,GAAG,EAAE,EAAEK,EAAE,KAAKwB,EAAE,GAAG,CAAC,OAAO,IAAIo4C,GAAGr5C,EAAEZ,EAAE,EAAEK,EAAEwB,CAAC,CAAC,CAAC,OAAO,KAAKjB,EAAE,KAAKZ,EAAE,GAAG,EAAE,KAAKK,EAAE,KAAK,CAAC,OAAO,IAAIw5C,GAAGj5C,EAAEZ,EAAE,EAAEK,CAAC,CAAC,CAAC,OAAO,SAASO,EAAE,KAAKZ,EAAE,IAAI,EAAE,KAAK,CAAC,OAAO,IAAI25C,GAAG/4C,EAAEZ,EAAE,CAAC,CAAC,CAAC,OAAO,OAAOY,EAAE,KAAKZ,EAAE,GAAG,EAAE,KAAKK,EAAE,KAAKwB,EAAE,EAAE,CAAC,OAAO,IAAIi4C,GAAGl5C,EAAEZ,EAAE,EAAEK,EAAEwB,CAAC,CAAC,CAAC,OAAO,QAAQjB,EAAEZ,EAAE,GAAG,CAAC,OAAO,IAAI45C,GAAGh5C,EAAEZ,CAAC,CAAC,CAAC,EAAMm6C,GAAG,CAAC,IAAID,GAAG,IAAI,SAASA,GAAG,SAAS,SAASA,GAAG,SAAS,QAAQA,GAAG,QAAQ,QAAQA,GAAG,QAAQ,OAAOA,GAAG,OAAO,KAAKA,GAAG,IAAI,EAAME,IAAI,IAAI,OAAO,uBAAuB,YAAY,sBAAsB,OAAO,cAAc,YAAY,aAAat6C,GAAGA,EAAE,GAAG,EAAE,SAASu6C,IAAI,CAAC,OAAO,IAAI,QAAQv6C,GAAGs6C,GAAG,IAAIt6C,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIw6C,EAAE,CAAC,EAAEp6C,GAAGo6C,EAAE,CAAC,OAAO,IAAIC,GAAG,OAAO,IAAIC,GAAG,OAAO,IAAIC,GAAG,OAAO,IAAIC,GAAG,OAAO,IAAIC,GAAG,MAAM,IAAIC,GAAG,sBAAsB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,WAAW,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,gBAAgB,IAAIjI,GAAG,2BAA2B,IAAIje,GAAG,2BAA2B,IAAImN,GAAG,uBAAuB,IAAIgZ,GAAG,mBAAmB,IAAIC,GAAG,qBAAqB,IAAIrZ,GAAG,gBAAgB,IAAI9L,GAAG,oBAAoB,IAAIolB,GAAG,0BAA0B,IAAIrf,GAAG,iBAAiB,IAAIgG,GAAG,kCAAkC,IAAIsZ,GAAG,qBAAqB,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,kBAAkB,IAAIvgB,GAAG,kBAAkB,IAAIK,GAAG,kBAAkB,IAAIK,GAAG,sBAAsB,IAAIZ,GAAG,yBAAyB,IAAI0gB,GAAG,0BAA0B,IAAIxZ,GAAG,gBAAgB,IAAIyZ,GAAG,kBAAkB,IAAIxgB,GAAG,kBAAkB,IAAIE,GAAG,wBAAwB,IAAIJ,GAAG,qBAAqB,IAAI2gB,GAAG,+BAA+B,IAAI5f,GAAG,qBAAqB,IAAImG,GAAG,SAAS,IAAI0Z,GAAG,UAAU,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,mBAAmB,IAAI3Z,GAAG,iBAAiB,IAAInN,GAAG,oBAAoB,IAAI+mB,GAAG,qBAAqB,IAAIC,GAAG,qBAAqB,IAAIC,GAAG,qBAAqB,IAAIlJ,GAAG,qBAAqB,IAAID,GAAG,eAAe,IAAIoJ,GAAG,iBAAiB,IAAI7Z,GAAG,YAAY,IAAI8Z,GAAG,cAAc,IAAIC,GAAG,iBAAiB,IAAInnB,GAAG,YAAY,IAAIonB,GAAG,oBAAoB,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,aAAa,IAAIC,GAAG,gDAAgD,IAAIC,GAAG,gDAAgD,IAAIC,GAAG,kDAAkD,IAAIC,GAAG,qDAAqD,IAAIC,GAAG,gDAAgD,IAAIC,GAAG,gDAAgD,IAAIC,GAAG,yDAAyD,IAAIC,GAAG,8CAA8C,IAAIC,GAAG,uDAAuD,IAAIC,GAAG,wDAAwD,IAAIC,GAAG,6DAA6D,IAAIC,GAAG,yDAAyD,IAAIC,GAAG,uBAAuB,IAAIjb,GAAG,sBAAsB,IAAIkb,GAAG,IAAI,IAAIr3B,GAAG,uBAAuB,IAAIs3B,GAAG,mBAAmB,IAAIznB,GAAG,iBAAiB,IAAI0nB,GAAG,aAAa,IAAIC,GAAG,WAAW,IAAIxK,GAAG,WAAW,IAAI9c,GAAG,uBAAuB,IAAIunB,GAAG,kBAAkB,IAAI7hB,GAAG,WAAW,IAAInR,GAAG,0BAA0B,IAAIizB,GAAG,cAAc,IAAI1nB,GAAG,oBAAoB,IAAIC,GAAG,KAAK,IAAIlQ,EAAE,CAAC,EAAE,SAASk1B,GAAGn7C,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAG,OAAOA,EAAE,QAAQ,CAACO,EAAE,IAAI,CAACuX,EAAEvX,EAAE,SAASL,EAAE,IAAI,kBAAkBA,uBAAuB,gDAAgDA,IAAI,CAAC,CAAC,EAAE4X,EAAE,GAAG,GAAG,EAAE5X,EAAE,IAAI,kBAAkBA,kCAAkCA,EAAE,IAAI,EAAE,IAAII,EAAEN,EAAE,GAAGA,EAAE,QAAQ,CAACO,EAAE,IAAI,CAAC,QAAQ,EAAE,EAAE,EAAEL,EAAE,IAAI4X,EAAE,IAAI,GAAGvX,EAAE,KAAKD,EAAE,GAAG,IAAI,kBAAkBJ,wBAAwB,OAAOK,4CAA4CD,sCAAsC,IAAI,CAAC,CAAC,CAAC,CAAC,SAASo7C,GAAG17C,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAG,MAAM,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAIJ,EAAE,IAAIF,EAAEM,GAAG,GAAG,OAAOJ,CAAC,CAAC,IAAI86C,IAAI,SAASh7C,EAAE,CAACA,EAAEA,EAAE,eAAe,GAAG,iBAAiBA,EAAEA,EAAE,aAAa,GAAG,eAAeA,EAAEA,EAAE,YAAY,GAAG,cAAcA,EAAEA,EAAE,WAAW,GAAG,aAAaA,EAAEA,EAAE,WAAW,GAAG,aAAaA,EAAEA,EAAE,WAAW,GAAG,YAAY,GAAGg7C,KAAKA,GAAG,CAAC,EAAE,EAAE,SAASM,GAAGt7C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,IAAI,MAAM,GAAGJ,GAAG,MAAM,GAAG,KAAK,OAAOI,EAAE,GAAG,GAAG,KAAK,KAAKA,EAAE,OAAON,EAAEE,EAAE,QAAQI,EAAE,KAAK,EAAE,OAAOA,EAAE,EAAE,MAAM,EAAE,GAAGJ,GAAG,KAAK,OAAOI,EAAE,GAAGN,EAAEE,EAAE,SAASI,EAAE,OAAO,MAAM,IAAI,MAAM,4BAA4B,uCAAuCN,EAAEE,EAAE,4BAA4BI,EAAE,QAAQ,EAAE,QAAQC,EAAE,EAAEA,EAAEL,EAAE,OAAO,EAAEK,EAAE,CAAC,IAAI,EAAEL,EAAEK,GAAG,EAAED,EAAEA,EAAE,OAAOJ,EAAE,OAAOK,GAAG0B,EAAE3B,EAAE,GAAG,GAAG,GAAG,EAAE,GAAG2B,GAAG,GAAG,GAAGA,IAAI,EAAE,MAAM,IAAI,MAAM,4BAA4B,sCAAsC1B,EAAEP,QAAQ,eAAeO,EAAEP,QAAQiC,GAAG,OAAO3B,EAAE,GAAG,CAAC,CAAC,OAAOA,CAAC,CAAC,SAASk8C,GAAGx8C,EAAE,CAAC,IAAI,EAAE,CAAC,eAAeg7C,GAAG,eAAe,aAAaA,GAAG,aAAa,YAAYA,GAAG,YAAY,WAAWA,GAAG,WAAW,WAAWA,GAAG,WAAW,WAAWA,GAAG,UAAU,EAAE96C,EAAE,CAAC,EAAE,QAAQI,KAAKN,EAAE,GAAGM,KAAK,EAAEJ,EAAE,KAAK,EAAEI,EAAE,MAAO,OAAM,OAAOJ,CAAC,CAAC,SAASm8C,GAAGr8C,EAAE,CAAC,OAAOA,EAAE,SAAS,EAAE,EAAEA,EAAE,KAAKg7C,GAAG,eAAeh7C,EAAE,OAAO,EAAEA,EAAE,MAAM,CAAC,SAAS49C,GAAG59C,EAAE,EAAE,CAAC,GAAGA,GAAG,MAAM,GAAG,KAAK,OAAO,IAAIE,EAAEF,EAAE,OAAOM,EAAE,EAAE,OAAO,GAAGJ,GAAGI,EAAE,MAAM,IAAI,MAAM,sBAAsBN,wCAAwC,4CAA4CE,6DAA6DI,IAAI,EAAE,QAAQC,EAAE,EAAEA,EAAE,KAAK,IAAIL,EAAEI,EAAE,CAAC,EAAE,EAAEC,EAAE,CAAC,IAAI,EAAEP,EAAEO,GAAG,EAAE,EAAEA,EAAE,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,GAAG,IAAI,EAAE,MAAM,IAAI,MAAM,sBAAsBP,+CAA+C,0CAA0CO,EAAEP,EAAE,aAAa,8CAA8CO,EAAEP,EAAE,aAAa,GAAG,CAAC,CAAC,CAAC,IAAI+6C,GAAG,GAAG,SAASU,GAAGz7C,EAAE,CAAC,OAAOA,GAAG+6C,GAAG/6C,EAAE2Z,GAAG3Z,EAAE,KAAK,MAAM,KAAK,KAAKA,CAAC,CAAC,CAAC,CAAC,CAAC,SAASm8C,GAAGn8C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,GAAG,OAAOF,GAAG,SAASA,EAAEA,EAAE,IAAIO,EAAE,GAAG,OAAOP,GAAG,SAASA,EAAEA,EAAE,IAAI,MAAM,CAACM,EAAEC,CAAC,CAAC,CAAC,SAAS+7C,GAAGt8C,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAE,CAAC,EAAE,GAAGD,EAAEC,EAAEA,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,EAAEA,EAAE,KAAKP,EAAE,GAAGE,CAAC,EAAEK,EAAEA,EAAE,OAAOP,EAAE,MAAM,CAAC,CAAC,MAAM,CAACO,EAAEA,EAAE,OAAOP,EAAE,EAAE,EAAE,IAAI,EAAE,EAAE,OAAO,QAAQ,EAAE,EAAE,EAAE,EAAE,EAAE,EAAEO,EAAEA,EAAE,OAAO,CAACP,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,CAAC,EAAEO,EAAEA,EAAE,OAAOP,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,CAAC,OAAOO,CAAC,CAAC,SAAS67C,GAAGp8C,EAAE,EAAEE,EAAE,GAAG,CAAC,IAAII,EAAE,CAAC,EAAE,GAAGJ,EAAE,CAACI,EAAE,KAAK,CAAC,EAAE,QAAQC,EAAE,EAAE,EAAEA,EAAEP,EAAE,EAAEO,EAAEA,GAAG,EAAE,GAAGD,EAAE,KAAKC,CAAC,EAAED,EAAE,KAAKC,GAAG,EAAE,EAAE,GAAGD,EAAE,KAAKC,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEP,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,GAAG,EAAE,IAAI,EAAE,EAAE,KAAK,CAAC,EAAEO,EAAE,KAAK,CAAC,EAAED,EAAE,KAAK,GAAGC,CAAC,EAAED,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,SAASi8C,GAAGv8C,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAE,CAAC,EAAED,EAAEC,EAAE,KAAKP,EAAE,GAAGE,CAAC,EAAEK,EAAE,KAAKP,EAAE,GAAGE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEF,EAAE,OAAO,EAAE,EAAE,GAAG,EAAE,OAAOM,EAAEC,EAAE,KAAK,EAAE,EAAE,GAAGP,EAAE,EAAE,EAAEO,EAAE,KAAKP,EAAE,GAAG,EAAE,EAAE,EAAE,EAAEO,EAAE,KAAKP,EAAE,EAAE,EAAE,OAAOO,CAAC,CAAC,SAASk8C,GAAGz8C,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAEJ,EAAE,KAAKF,EAAEM,GAAG,EAAE,EAAE,OAAOJ,CAAC,CAAC,SAASw8C,GAAG18C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,EAAE,CAAC,EAAE,QAAQO,EAAE,EAAEA,EAAEL,EAAE,EAAEK,EAAED,EAAE,KAAKN,EAAEO,EAAE,GAAG,EAAEA,GAAG,GAAG,EAAEA,GAAG,EAAE,EAAE,OAAOD,CAAC,CAAC,IAAI46C,GAAG,mBAAmBD,GAAG,mBAAuBH,GAAG,SAASL,GAAG,WAAWC,GAAG,YAAYC,GAAG,YAAYC,GAAG,aAAaC,GAAG,YAAY,SAAS2C,GAAGx9C,EAAE,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,MAAM,IAAI,MAAM,gEAAgEA,EAAE,iBAAiB,EAAE,SAAS,EAAE,IAAIE,EAAE,IAAI,aAAaF,EAAE,OAAO,CAAC,EAAE,QAAQM,EAAE,EAAEA,EAAEJ,EAAE,OAAOI,GAAG,EAAEJ,EAAEI,GAAGN,EAAEM,EAAE,GAAGJ,EAAEI,EAAE,GAAG,EAAEA,EAAE,GAAG,OAAOJ,CAAC,CAAC,SAASy9C,GAAG39C,EAAE,CAAC,IAAI,EAAE,IAAI,aAAaA,EAAE,OAAO,CAAC,EAAEE,EAAE,IAAI,aAAaF,EAAE,OAAO,CAAC,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,OAAOM,GAAG,EAAE,EAAEA,EAAE,GAAGN,EAAEM,GAAGJ,EAAEI,EAAE,GAAGN,EAAEM,EAAE,GAAG,MAAM,CAAC,KAAK,EAAE,KAAKJ,CAAC,CAAC,CAAC,SAASq7C,GAAGv7C,EAAE,CAAC,IAAI,EAAE,KAAK,KAAKA,EAAE,OAAO,CAAC,EAAEE,EAAE,IAAI,aAAa,CAAC,EAAEI,EAAE,IAAI,aAAa,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,GAAG,EAAEL,EAAE,KAAK,MAAMK,EAAE,CAAC,GAAGP,EAAEO,GAAGD,EAAE,KAAK,MAAMC,EAAE,CAAC,GAAGP,EAAEO,EAAE,GAAG,MAAM,CAAC,KAAKL,EAAE,KAAKI,CAAC,CAAC,CAAC,SAASk7C,GAAGx7C,EAAE,CAAC,IAAI,EAAE,KAAK,MAAMA,EAAE,OAAO,CAAC,EAAEE,EAAE,IAAI,aAAa,CAAC,EAAEI,EAAE,IAAI,aAAa,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,GAAG,EAAEL,EAAE,KAAK,MAAMK,EAAE,CAAC,GAAGP,EAAEO,GAAGD,EAAE,KAAK,MAAMC,EAAE,CAAC,GAAGP,EAAEO,EAAE,GAAG,MAAM,CAAC,KAAKL,EAAE,KAAKI,CAAC,CAAC,CAAC,SAAS07C,GAAGh8C,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,EAAE,GAAGM,EAAEN,EAAE,EAAE,EAAE,GAAG,MAAM,CAAC,KAAKE,EAAE,KAAKI,CAAC,CAAC,CAAC,SAAS86C,GAAGp7C,EAAE,EAAEE,EAAEI,EAAE,CAACN,EAAEM,EAAE,GAAG,EAAEN,EAAEM,EAAE,EAAE,GAAGJ,CAAC,CAAC,SAAS27C,GAAG77C,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,aAAaF,EAAE,CAAC,EAAEM,EAAE,IAAI,aAAaN,EAAE,CAAC,EAAE,QAAQO,EAAE,EAAEA,EAAE,KAAK,KAAKP,EAAE,CAAC,EAAEO,IAAI,CAAC,IAAI,GAAG,EAAE,EAAE,IAAI,KAAK,IAAIA,EAAEP,GAAGE,EAAEK,GAAG,KAAK,IAAI,CAAC,EAAED,EAAEC,GAAG,KAAK,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,KAAKL,EAAE,KAAKI,CAAC,CAAC,CAAC,SAASs7C,GAAG57C,EAAE,EAAEE,EAAE,CAAC,IAAII,GAAGJ,EAAE,EAAE,IAAI,KAAK,IAAIF,EAAE,GAAGO,EAAE,KAAK,IAAID,CAAC,EAAE,EAAE,KAAK,IAAIA,CAAC,EAAE,MAAM,CAAC,KAAKC,EAAE,KAAK,CAAC,CAAC,CAAC,IAAIs9C,GAAG,KAAKC,GAAG,MAAMC,GAAG,IAAIC,GAAG,MAAM,SAASrC,GAAG37C,EAAE,EAAE,CAACA,EAAEA,EAAE,QAAQ,MAAM,EAAE,EAAE,IAAIE,GAAGF,EAAE,OAAOA,EAAE,QAAQ89C,GAAG,EAAE,EAAE,QAAQD,GAAG,OAAO,GAAG39C,EAAE,EAAE,MAAM,IAAI,MAAM,+CAA+C,EAAE,GAAGA,EAAE,EAAE,MAAM,IAAI,MAAM,6CAA6C29C,OAAO,EAAE,GAAG,CAACv9C,EAAEC,CAAC,EAAEP,EAAE,MAAM69C,EAAE,EAAE/lC,EAAExX,EAAE,QAAQ09C,EAAE,IAAI,GAAG,IAAI,2BAA2BA,4BAA4B,EAAE,IAAI,EAAE19C,EAAE,MAAMy9C,EAAE,EAAE,EAAE,EAAE,OAAO,GAAG,IAAI,EAAE,MAAM,IAAI,MAAM,YAAY,6BAA6B,GAAG,EAAE,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,+DAA+D,EAAE,IAAI97C,EAAE,CAAC,EAAE,QAAQa,EAAE,EAAEA,EAAEvC,EAAE,OAAO,EAAEuC,EAAE,CAAC,IAAIC,EAAExC,EAAEuC,GAAG,GAAG,CAAC,EAAE,KAAKE,GAAGA,EAAE,QAAQD,CAAC,IAAI,EAAE,EAAE,MAAM,IAAI,MAAM,uCAAuCA,wCAAwC,EAAEd,EAAE,QAAQc,CAAC,IAAI,IAAId,EAAE,KAAKc,CAAC,CAAC,CAAC,QAAQD,EAAE,EAAEA,EAAExC,EAAE,OAAO,EAAEwC,EAAE,CAAC,IAAIC,EAAEzC,EAAEwC,GAAGb,EAAE,QAAQc,CAAC,IAAI,IAAIA,IAAIg7C,IAAI97C,EAAE,KAAKc,CAAC,CAAC,CAAC,IAAIb,EAAE,IAAI,MAAM,EAAE,MAAM,EAAE,QAAQY,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,GAAG,IAAI,IAAI,EAAEA,GAAG,MAAM,EAAE,CAAC,EAAE,OAAO,EAAEA,GAAG,OAAO,MAAM,IAAI,MAAM,2CAA2C,EAAEA,iEAAiE,EAAEZ,EAAEY,GAAG,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAED,GAAG,OAAO,EAAEC,EAAEb,EAAEY,GAAG,KAAKb,EAAE,QAAQ,EAAEa,GAAGC,EAAE,CAAC,CAAC,CAAC,IAAIZ,EAAEF,EAAE,OAAOW,EAAErC,EAAE,OAAOsC,EAAE,CAAC,EAAE,QAAQC,EAAEF,EAAEE,EAAEX,EAAE,EAAEW,EAAED,EAAE,KAAKC,CAAC,EAAE,MAAM,CAAC,QAAQb,EAAE,WAAWY,EAAE,OAAOX,CAAC,CAAC,CAAC,SAASg6C,GAAGl8C,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,MAAMF,CAAC,EAAEE,EAAE,KAAK,EAAE,EAAE,QAAQK,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAEL,EAAE,EAAEK,IAAIA,EAAE,IAAID,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,EAAEO,EAAEL,EAAEK,KAAK,IAAID,EAAE,KAAKC,CAAC,EAAE,OAAOL,EAAEA,EAAE,OAAOK,GAAGA,IAAI,EAAE,EAAE,CAAC,mBAAmBL,EAAE,WAAWI,CAAC,CAAC,CAAC,SAAS+6C,GAAGr7C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,IAAI,MAAMN,CAAC,EAAE,QAAQO,EAAE,EAAEA,EAAEL,EAAE,OAAO,EAAEK,EAAE,CAAC,IAAI,EAAEL,EAAEK,GAAG,MAAM,QAAQ,EAAE,EAAE,EAAE,EAAEA,GAAG,OAAO,EAAE,EAAED,EAAE,EAAEC,GAAG,MAAM,OAAOD,EAAE,EAAEC,GAAG,IAAI,EAAE,GAAGuX,EAAExX,EAAE,EAAEC,GAAG,MAAM,EAAE,GAAG,IAAI,sBAAsBD,EAAE,EAAEC,GAAG,eAAe,qBAAqB,KAAK,UAAU,CAAC,wBAAwB,EAAE,IAAI,CAAC,CAAC,CAAC,SAAS07C,GAAGj8C,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAEM,EAAE,CAAC,EAAEC,EAAE,EAAEP,EAAE,SAAS,GAAGE,EAAE,KAAK,EAAE,EAAEK,EAAEP,EAAE,OAAO,EAAE,QAAQ,EAAE,EAAE,EAAEO,EAAE,EAAE,EAAED,EAAE,KAAK,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEJ,EAAE,OAAO,EAAE,EAAE,CAAC,IAAI+B,EAAE/B,EAAE,GAAGgC,EAAE+7C,GAAG,EAAEh8C,CAAC,EAAE,QAAQE,KAAKD,EAAE,EAAE,QAAQC,CAAC,IAAI,KAAK7B,EAAE,GAAG,KAAK6B,CAAC,EAAE,EAAE,KAAKA,CAAC,EAAE,CAAC,MAAM,CAAC,KAAKjC,EAAE,MAAMI,CAAC,CAAC,CAAC,SAASi9C,GAAGv9C,EAAE,CAAC,OAAOA,EAAE,MAAM,CAAC,EAAEE,IAAI,IAAIA,CAAC,CAAC,CAAC,SAAS+9C,GAAGj+C,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAO,EAAEM,GAAGN,EAAEM,GAAG,SAAS,GAAGN,EAAEM,GAAG,QAAQ,CAAC,IAAI,IAAI,IAAI,KAAKJ,EAAE,KAAKI,CAAC,EAAE,OAAOJ,CAAC,CAAC,SAASu9C,GAAGz9C,EAAE,EAAEE,EAAE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAE,GAAG,OAAO,GAAG,SAASwX,EAAE9X,EAAE,MAAME,GAAG,IAAI,EAAE,IAAI,+CAA+C,EAAEI,EAAE,IAAI,MAAM,CAAC,EAAE,KAAKN,EAAE,MAAME,GAAG,CAAC,MAAM,CAAC,IAAIK,EAAE,EAAE,OAAO,CAAC,EAAE0B,KAAKA,IAAI,KAAK,GAAG,GAAG,GAAG,CAAC,EAAE6V,EAAEvX,GAAG,EAAE,IAAI,yDAAyD,EAAE,IAAI,EAAE,EAAE,QAAQ,EAAE,EAAE,GAAG,IAAI,GAAG,CAAC,IAAI,EAAE,EAAE,OAAO,CAAC0B,EAAEC,IAAIA,EAAE,EAAED,EAAEC,EAAED,CAAC,EAAE,EAAE,GAAGjC,EAAE,MAAME,GAAG,CAAC,CAAC4X,EAAE9X,EAAE,MAAME,KAAK,EAAE,OAAO,CAAC,EAAE+B,IAAI,EAAEA,CAAC,EAAE,IAAI,6DAA6D,EAAE3B,EAAE,CAAC,CAAC,OAAOA,CAAC,CAAC,SAASq8C,GAAG38C,EAAE,CAAC,MAAM;AAAA,uBACxsxBA,GAAG,CAAC,SAAS48C,GAAG58C,EAAE,EAAE,CAAC,MAAM,WAAWA,qBAAqB,OAAO,CAAC,SAAS68C,GAAG78C,EAAE,EAAEE,EAAE,CAAC,MAAM,WAAWF,qBAAqB,QAAQE,GAAG,CAAC,SAAS+8C,GAAGj9C,EAAE,EAAE,CAAC,MAAM,iDAAiDA,SAAS,GAAG,CAAC,SAASk9C,GAAGl9C,EAAE,EAAE,CAAC,MAAM,QAAQA,+BAA+B,GAAG,CAAC,SAAS88C,IAAI,CAAC,MAAM,+GAA+G,CAAC,SAASE,GAAGh9C,EAAE,EAAE,CAAC,IAAIE,EAAEkY,GAAGpY,CAAC,EAAEM,EAAE8X,GAAG,CAAC,EAAE,MAAM,2CAA2ClY;AAAA,iEAC/cI,iBAAiBN,kBAAkB,GAAG,CAAC,SAAS+8C,GAAG/8C,EAAE,EAAE,CAAC,IAAIE,EAAEkY,GAAGpY,CAAC,EAAEM,EAAE8X,GAAG,CAAC,EAAE,MAAM,qCAAqClY,+CAA+CI,iBAAiBN,iBAAiB,GAAG,CAAC,SAASo9C,IAAI,CAAC,MAAM,0BAA0B,CAAC,SAASC,IAAI,CAAC,MAAM,gCAAgC,CAAC,SAASC,GAAGt9C,EAAE,EAAE,CAAC,MAAM,cAAcA,sBAAsB,sDAAsD,CAAC,SAASm9C,GAAGn9C,EAAE,EAAEE,EAAE,CAAC,MAAM,gBAAgBF,SAAS,sBAAsBE,IAAI,CAAC,IAAIw9C,GAAG,CAAC,EAAEt9C,GAAGs9C,GAAG,CAAC,yBAAyB,IAAIQ,GAAG,gBAAgB,IAAIC,GAAG,8BAA8B,IAAIC,EAAE,CAAC,EAAE,SAASA,GAAGp+C,EAAE,EAAE,CAAC,IAAIE,EAAE,GAAGI,EAAE,IAAIN,GAAG+6C,IAAIz6C,EAAEN,EAAEE,EAAE,IAAII,EAAEqZ,GAAG3Z,EAAE,KAAK,MAAM,KAAK,KAAKA,CAAC,CAAC,CAAC,EAAE,CAACE,GAAGI,EAAE,GAAGA,IAAIN,EAAEE,EAAE,GAAGI,EAAEqZ,GAAG3Z,EAAEM,EAAE,CAAC,EAAE,OAAOA,CAAC,CAAC,SAAS69C,GAAGn+C,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAEC,EAAEP,EAAE,OAAO,QAAQ,EAAE,EAAE,EAAEO,EAAE,IAAI,IAAI,EAAED,EAAE,KAAKN,EAAE,EAAE,EAAEM,EAAE,KAAKJ,CAAC,EAAE,OAAOI,CAAC,CAAC,SAAS49C,GAAGl+C,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,EAAE,MAAM,OAAO,EAAEP,EAAE,MAAM,OAAO,GAAGM,IAAI,IAAIA,EAAE,CAACC,GAAGD,EAAEC,GAAG,MAAM,IAAI,MAAM,sCAAsCA,MAAMA,eAAeD,GAAG,EAAE,GAAGA,EAAE,IAAIA,GAAGC,GAAGD,EAAE,EAAE,MAAM,IAAI,MAAM,cAAcA;AAAA,MACvkC,KAAK,EAAE,GAAGJ,EAAEI,EAAE,MAAM,IAAI,MAAM,cAAcA,0CAA0CJ,KAAK,EAAE,QAAQ2C,EAAE,EAAEA,EAAEvC,EAAE,EAAEuC,EAAE,GAAG7C,EAAE,MAAM6C,KAAK,EAAE,MAAMA,GAAG,MAAM,IAAI,MAAM,WAAWA,OAAO7C,EAAE,MAAM6C,uCAAuCA,OAAO,EAAE,MAAMA,KAAK,EAAE,IAAI,EAAE7C,EAAE,MAAME,GAAG+B,EAAE,CAAC,EAAEC,EAAE,EAAEC,EAAE,EAAES,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAEvC,EAAE,EAAEuC,EAAEZ,EAAE,KAAKjC,EAAE,MAAM6C,EAAE,EAAEX,GAAGlC,EAAE,MAAM6C,GAAG,QAAQA,EAAEvC,EAAEuC,EAAE3C,EAAE2C,IAAIZ,EAAE,KAAKjC,EAAE,MAAM6C,EAAE,EAAEV,GAAGnC,EAAE,MAAM6C,GAAG,QAAQA,EAAEvC,EAAEuC,EAAEtC,EAAEsC,IAAIZ,EAAE,KAAK,EAAE,MAAMY,EAAE,EAAE,QAAQA,EAAE3C,EAAE,EAAE2C,EAAE,EAAEA,IAAIZ,EAAE,KAAKjC,EAAE,MAAM6C,EAAE,EAAED,GAAG5C,EAAE,MAAM6C,GAAG,MAAM,CAAC,UAAUX,EAAE,UAAUU,EAAE,UAAUT,EAAE,QAAQ,EAAE,YAAYF,CAAC,CAAC,CAAC,SAAS85C,GAAG/7C,EAAE,CAAC,GAAG,CAAC,OAAOA,EAAE,IAAI,GAAGgnB,GAAG,CAAC,CAAC,CAAC,OAAO,EAAN,CAAS,MAAM,IAAI,MAAM,4DAA4D,GAAG,CAAC,CAAC,CAAC,SAAS80B,GAAG97C,EAAE,CAAC,OAAOA,EAAE,IAAI,GAAGinB,GAAG,CAAC,CAAC,CAAC,CAAC,IAAIo3B,GAAG,CAAC,EAAEj+C,GAAGi+C,GAAG,CAAC,wBAAwB,IAAIpJ,GAAG,wBAAwB,IAAIE,GAAG,wBAAwB,IAAIC,GAAG,UAAU,IAAIhE,EAAE,CAAC,EAAE,IAAIkN,GAAG,CAAC,WAAWpjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAClb,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAE8vC,GAAGrf,GAAEvwB,EAAE,SAAS,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAMq+C,GAAG,CAAC,WAAWpjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACnb,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAII,EAAE2iC,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,EAAEK,EAAEwiC,GAAG2D,GAAG7D,GAAG,CAAC,EAAEviC,CAAC,CAAC,EAAE,OAAOm0B,GAAGyE,GAAGl5B,EAAEO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMi+C,GAAG,CAAC,WAAWpjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACpb,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAII,EAAEyiC,GAAG2D,GAAGzD,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,EAAE,CAAC,CAAC,EAAE,OAAOg5B,GAAGl5B,EAAEM,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMm+C,GAAG,CAAC,WAAWpjC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACrb,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEjC,EAAEkC,EAAEgzB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO2B,EAAE,OAAO,IAAID,EAAEkhC,GAAGlhC,EAAEC,CAAC,GAAGg6B,EAAEj6B,EAAE/B,EAAE,KAAK,CAAC,EAAE,EAAE,IAAI,CAAC,IAAI+B,EAAEjC,EAAEkC,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE,OAAO2B,EAAE,OAAO,IAAID,EAAEkhC,GAAGlhC,EAAEC,CAAC,GAAGg6B,EAAEj6B,EAAE3B,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAMo+C,GAAG,CAAC,WAAWpjC,GAAG,cAAc,GAAG,SAAS,CAACtb,EAAE,IAAI,CAAC,IAAIE,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ,CAACI,EAAEC,IAAI,CAACL,EAAEK,GAAG,IAAIP,EAAE,MAAM,CAAC,CAAC,EAAEE,CAAC,CAAC,EAAMy+C,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACzb,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIkhC,GAAGlhC,CAAC,CAAC,CAAC,CAAC,EAAM0+C,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC1b,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIkhC,GAAGlhC,CAAC,CAAC,CAAC,CAAC,EAAM2+C,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC3b,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAE+iC,GAAG2D,GAAG7D,GAAG,CAAC,EAAEI,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM4+C,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC5b,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAII,EAAEyiC,GAAGjK,EAAE+J,GAAG,CAAC,EAAEI,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE,OAAOg5B,GAAGl5B,EAAEM,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMy+C,GAAG,CAAC,WAAWhjC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAAC/b,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAE62B,EAAEmK,GAAG/iC,CAAC,EAAE+iC,GAAG3iC,CAAC,CAAC,EAAE4B,EAAEk3B,EAAEp5B,EAAEk5B,GAAG54B,EAAE2B,CAAC,CAAC,EAAEE,EAAE+yB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO4B,EAAE,OAAO,IAAID,EAAEihC,GAAGjhC,EAAEC,CAAC,GAAG+5B,EAAEh6B,EAAEhC,EAAE,KAAK,CAAC,EAAE,EAAE,IAAI,CAAC,IAAI+B,EAAE62B,EAAEmK,GAAG/iC,CAAC,EAAE+iC,GAAG3iC,CAAC,CAAC,EAAE4B,EAAEuyB,GAAG2E,EAAEp5B,EAAEk5B,GAAGh5B,EAAE+B,CAAC,CAAC,CAAC,EAAEE,EAAE+yB,GAAG50B,EAAE,MAAMC,CAAC,EAAE,OAAO4B,EAAE,OAAO,IAAID,EAAEihC,GAAGjhC,EAAEC,CAAC,GAAG+5B,EAAEh6B,EAAE5B,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAM0+C,GAAG,CAAC,WAAWnjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC7b,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAE84B,EAAEmK,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM++C,GAAG,CAAC,WAAWnjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC9b,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAE0mC,GAAG7D,GAAG,CAAC,EAAEI,GAAGxS,GAAEvwB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASg/C,GAAGl/C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,KAAK,eAAe,EAAEiC,EAAE+pB,EAAE,EAAE,QAAQ,eAAe,EAAE9pB,EAAE,EAAEC,EAAEF,EAAEW,EAAE,GAAGX,EAAE,OAAO,IAAIW,EAAE,GAAGV,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE/5B,EAAE+5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG6V,EAAE5V,EAAE,OAAO,EAAE,IAAI,0DAA0DA,EAAE,OAAO,EAAE4V,EAAE3V,EAAE,OAAO,EAAE,IAAI,6DAA6DA,EAAE,OAAO,EAAE65B,GAAG,gBAAgBz7B,EAAE,CAAC,EAAE,IAAIsC,EAAE,CAAC,GAAGX,EAAE,MAAMC,CAAC,EAAEW,EAAE,CAAC,WAAW5C,EAAE,QAAQI,EAAE,IAAIC,EAAE,gBAAgB,CAAC,EAAEwC,EAAEmoB,EAAE,UAAU/O,GAAGtZ,EAAEC,CAAC,EAAE,OAAOF,EAAEs5B,EAAEn5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIo8C,GAAGhzB,EAAE,CAAC,eAAe+yB,EAAE,CAAC,EAAME,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAClc,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,WAAWC,EAAE,QAAQ,EAAE,IAAI,EAAE,gBAAgB0B,CAAC,EAAE/B,EAAE,MAAM,CAAC,EAAE,IAAIi/C,GAAGn/C,EAAEM,EAAEC,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,EAAE,SAASo9C,GAAGr/C,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEyrB,EAAEhsB,EAAE,KAAK,aAAa,EAAE,EAAEgsB,EAAE,EAAE,QAAQ,aAAa,EAAElU,EAAE,EAAE,OAAO,EAAE,KAAK,IAAI,kBAAkB,EAAE,oCAAoC,EAAE,OAAO,EAAE,IAAI7V,EAAE,EAAEC,EAAE,EAAEC,EAAE,GAAG,EAAE,OAAO,IAAIA,EAAE,GAAGF,EAAEi6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAEh6B,EAAEg6B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE5V,EAAE,OAAO,EAAE,IAAI,wDAAwDA,EAAE,OAAO,EAAE4V,EAAE7V,EAAE,OAAO,EAAE,IAAI,2DAA2DA,EAAE,OAAO,EAAE,IAAIW,EAAE,CAAC,GAAGV,EAAE,MAAMD,CAAC,EAAEY,EAAE,CAAC,WAAW3C,EAAE,QAAQI,EAAE,IAAIC,CAAC,EAAEuC,EAAEooB,EAAE,UAAUjP,GAAGrZ,EAAEC,CAAC,EAAE,OAAOV,EAAE+5B,EAAEp5B,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEA,CAAC,CAAC,IAAIw8C,GAAGnzB,EAAE,CAAC,aAAakzB,EAAE,CAAC,EAAME,GAAG,CAAC,WAAWvjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAChc,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,WAAWC,EAAE,QAAQ,EAAE,IAAI,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAIo/C,GAAGt/C,EAAEM,EAAEC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAMi/C,GAAG,CAAC,WAAWpjC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACpc,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,WAAW,CAAC,EAAEL,EAAE,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,IAAI4yB,GAAG9yB,EAAEO,EAAE,GAAG,EAAE,EAAE,EAAE,IAAIuyB,GAAGxyB,EAAEN,EAAE,GAAG,EAAE,CAAC,EAAE,CAAC,GAAG,EAAE,CAAC,EAAE,IAAI8yB,GAAG9yB,EAAEO,EAAE,GAAG,EAAE,EAAE,EAAE,IAAIuyB,GAAG9yB,EAAEM,EAAE,GAAG,EAAE,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,IAAIwyB,GAAGvyB,EAAEP,EAAE,GAAG,EAAE,EAAE,EAAE,IAAI8yB,GAAGxyB,EAAEN,EAAE,GAAG,EAAE,CAAC,EAAE,CAAC,EAAE,IAAI8yB,GAAGvyB,EAAEP,EAAE,GAAG,EAAE,EAAE,EAAE,IAAI8yB,GAAG9yB,EAAEM,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,EAAMm/C,GAAG,CAAC,WAAWpjC,GAAG,SAAS,CAACrc,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,WAAWI,EAAE,MAAMC,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAImqC,GAAGrqC,EAAEM,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAMm/C,GAAG,CAAC,WAAWnjC,GAAG,SAAS,CAACvc,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAEJ,EAAEK,EAAED,EAAE,WAAW,EAAEA,EAAE,MAAM,EAAE,MAAM,KAAK,CAAC,EAAE,QAAQ4B,EAAE3B,EAAE,OAAO,EAAE2B,GAAG,EAAEA,IAAI,GAAG3B,EAAE2B,KAAK,EAAEA,GAAG,EAAEA,GAAG,UAAU3B,EAAE2B,KAAK,EAAE,MAAM,IAAI,MAAM,mBAAmB3B,8BAA8B,KAAK,EAAE,IAAI0B,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,GAAG,GAAGD,EAAE,KAAKC,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIihC,GAAGnjC,EAAEiC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAM09C,GAAG,CAAC,WAAWljC,GAAG,SAASzc,IAAI,CAAC,EAAE,IAAIA,EAAE,MAAM,CAAC,EAAE,EAAM4/C,GAAG,CAAC,WAAWljC,GAAG,SAAS1c,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAM6/C,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC3c,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,aAAaC,EAAE,aAAa,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAIghC,GAAG8F,GAAGtC,GAAGpkC,EAAEC,CAAC,EAAE+kC,GAAGhlC,EAAE,CAAC,CAAC,EAAEN,EAAEohC,GAAGphC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM8/C,GAAG,CAAC,WAAWjjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAASyhC,GAAG,QAAQ,EAAMyB,GAAG,CAAC,WAAWjjC,GAAG,cAAc,GAAG,SAAS,CAAC9c,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAE,EAAE,IAAI4B,GAAGA,EAAE,KAAK,EAAE,CAAC,KAAK3B,CAAC,EAAEL,EAAE,EAAE2Y,GAAGtY,EAAE,EAAE,GAAG,KAAK,EAAE,GAAG,EAAED,EAAE,IAAI4B,GAAGA,EAAE,EAAE,EAAE,OAAOktC,GAAGpvC,EAAE,EAAE,CAAC,EAAE,IAAIkC,GAAG,IAAIA,CAAC,CAAC,CAAC,EAAM89C,GAAG,CAAC,WAAWjjC,GAAG,aAAa,CAAC,IAAI,QAAQ,EAAE,SAAS,CAAC/c,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,UAAU,EAAE,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,CAAC,EAAEhC,EAAE,OAAO4X,EAAEgkB,GAAG,CAAC,EAAE,IAAI,iHAAiH,IAAI,EAAE,CAAC,EAAE,IAAIsD,GAAG9+B,EAAE,MAAMN,EAAEO,EAAE,EAAE0B,EAAEC,CAAC,EAAE,OAAO,IAAI4wC,GAAGxyC,EAAEN,EAAEO,EAAE,MAAM,EAAE0B,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAM+9C,GAAG,CAAC,WAAWhjC,GAAG,aAAa,CAAC,KAAK,QAAQ,EAAE,SAAS,CAACjd,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI,EAAE,WAAW0B,EAAE,gBAAgBC,CAAC,EAAEhC,EAAE,MAAM,CAAC,GAAG,IAAI8+B,GAAGh/B,EAAEO,EAAE,EAAE,EAAE0B,EAAE,EAAEC,CAAC,EAAE,OAAO,IAAI4wC,GAAG9yC,EAAEM,EAAEC,EAAE,MAAM,EAAE,EAAE0B,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASg+C,GAAGlgD,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEP,EAAEA,EAAE,OAAO,IAAI,EAAEk8B,EAAEl8B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI,EAAE,EAAE,EAAE,OAAO,IAAI,EAAEk8B,EAAE,EAAE,CAAC,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,GAAGpkB,EAAE,EAAE,OAAO,EAAE,IAAI,iEAAiE,EAAE,QAAQ,EAAEA,EAAE,EAAE,OAAO,EAAE,IAAI,8DAA8D,EAAE,QAAQ,EAAEA,EAAE5X,EAAE,SAAS,EAAE,IAAI,mEAAmEA,IAAI,EAAE4X,EAAE,EAAE,MAAM,KAAK5X,EAAE,GAAG,IAAI,4CAA4C,EAAE,MAAM,yCAAyCA,EAAE,KAAK,EAAE4X,EAAE,EAAE,MAAM,KAAK5X,EAAE,GAAG,IAAI,0CAA0C,EAAE,MAAM,2CAA2CA,EAAE,MAAM,EAAE,IAAI+B,EAAE,CAAC,EAAE,EAAE,GAAG,CAAC,EAAEC,EAAE,CAAC,QAAQ5B,EAAE,IAAIC,EAAE,YAAYL,CAAC,EAAE,OAAOgrB,EAAE,UAAU/N,GAAGlb,EAAEC,CAAC,CAAC,CAAC,IAAIi+C,GAAGh0B,EAAE,CAAC,sBAAsB+zB,EAAE,CAAC,EAAME,GAAG,CAAC,WAAWljC,GAAG,aAAa,CAAC,IAAI,QAAQ,EAAE,SAAS,CAACld,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,UAAUI,EAAE,QAAQC,EAAE,IAAI,CAAC,EAAEL,EAAE4X,EAAEgkB,GAAGx7B,CAAC,EAAE,IAAI,iHAAiHA,IAAI,EAAE,GAAG,CAAC,EAAE2B,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIy9B,GAAG,EAAE,MAAM1/B,EAAEiC,EAAE1B,EAAE,CAAC,EAAE,OAAO,IAAI4/C,GAAG,EAAEngD,EAAEiC,EAAE,MAAM1B,EAAE,CAAC,CAAC,CAAC,CAAC,EAAM8/C,GAAG,CAAC,WAAWhjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACrd,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAE3E,GAAGuZ,GAAGvd,GAAEvwB,EAAE,SAAS,CAAC,CAAC,EAAEF,CAAC,CAAC,CAAC,CAAC,EAAMsgD,GAAG,CAAC,WAAWhjC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACtd,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAE8U,GAAGzd,GAAEvwB,EAAE,SAAS,CAAC,EAAEF,CAAC,CAAC,CAAC,CAAC,EAAMugD,GAAG,CAAC,WAAW/iC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACxd,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,KAAKC,EAAE,UAAU,EAAE,QAAQ,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI+B,EAAEmgC,GAAG,CAAC7hC,CAAC,EAAED,EAAE,IAAI,EAAE4B,EAAEk+B,GAAGpgC,EAAEO,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO0B,GAAG,OAAOC,EAAE2yB,GAAG3yB,EAAED,CAAC,GAAGC,CAAC,CAAC,CAAC,CAAC,EAAMs+C,GAAG,CAAC,WAAW5iC,GAAG,aAAa,CAAC,IAAI,QAAQ,EAAE,SAAS,CAAC5d,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,UAAUI,EAAE,QAAQC,EAAE,IAAI,EAAE,gBAAgB,CAAC,EAAEL,EAAE+B,EAAE3B,GAAG,KAAK,CAAC,EAAE,CAAC,EAAEA,EAAEwX,EAAEgkB,GAAG75B,CAAC,EAAE,IAAI,mHAAmHA,IAAI,EAAE,GAAG,CAACC,EAAEC,CAAC,EAAE,EAAE,OAAO2V,EAAE5V,EAAE,OAAO,EAAE,IAAI,kFAAkFA,EAAE,OAAO,EAAE4V,EAAE3V,EAAE,OAAO,EAAE,IAAI,mFAAmFA,EAAE,OAAO,EAAE2V,EAAE5V,EAAE,MAAM,KAAKC,EAAE,MAAM,GAAG,IAAI,mEAAmED,EAAE,MAAM,qDAAqDC,EAAE,MAAM,KAAK,EAAE2V,EAAEikB,GAAGx7B,EAAE0B,CAAC,EAAE,IAAI,6FAA6F1B,oBAAoB0B,KAAK,EAAE+5B,GAAG,kBAAkB,EAAE,CAAC,EAAE,CAAC,EAAE,IAAIuX,GAAGrxC,EAAE,MAAMlC,EAAEmC,EAAE5B,EAAE,EAAE0B,EAAE,CAAC,EAAE,OAAO,IAAIoxC,GAAGnxC,EAAElC,EAAEmC,EAAE,MAAM5B,EAAE,EAAE0B,EAAE,CAAC,CAAC,CAAC,CAAC,EAAMw+C,GAAG,CAAC,WAAWziC,GAAG,aAAa,CAAC,IAAI,QAAQ,EAAE,SAAS,CAAChe,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAED,EAAE,OAAOC,EAAE,GAAGP,CAAC,EAAE,EAAE,CAAC,EAAEM,EAAE,OAAOC,EAAE,GAAGP,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIkrB,EAAE,UAAUjN,GAAG,EAAE/d,CAAC,EAAE,OAAO,IAAIgrB,EAAE,UAAUhN,GAAG,EAAEhe,CAAC,CAAC,CAAC,CAAC,EAAMwgD,GAAG,CAAC,WAAWriC,GAAG,cAAc,CAAC,EAAE,EAAE,SAAS,CAACre,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAEI,EAAE,CAAC,GAAGN,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIgrB,EAAE,UAAU5M,GAAGhe,CAAC,CAAC,CAAC,CAAC,EAAMqgD,GAAG,CAAC,WAAWpiC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACve,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAEI,EAAE84B,EAAEsK,GAAGjP,GAAGwO,GAAG/iC,CAAC,CAAC,CAAC,EAAE,EAAE,KAAK,KAAK,KAAK,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEM,CAAC,CAAC,CAAC,CAAC,EAAMsgD,GAAG,CAAC,WAAWniC,GAAG,cAAc,CAAC,EAAE,EAAE,SAAS,CAACze,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEE,CAAC,CAAC,CAAC,CAAC,EAAM2gD,GAAG,CAAC,WAAWniC,GAAG,aAAa,CAAC,OAAO,EAAE,SAAS,CAAC1e,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,MAAM,IAAIg8B,EAAEl8B,EAAEE,EAAE,KAAK,CAAC,CAAC,CAAC,EAAM4gD,GAAG,CAAC,WAAWniC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC3e,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAE0jC,GAAGxjC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM6gD,GAAG,CAAC,WAAWhiC,GAAG,SAAS/e,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMghD,GAAG,CAAC,WAAWhiC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAAChf,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEi3B,GAAGl5B,EAAEywB,GAAEnwB,EAAE,SAAS,CAAC,EAAE4B,EAAEgzB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO2B,EAAE,OAAO,EAAEg6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAEhC,EAAE,KAAK,EAAE+B,CAAC,EAAE,EAAE,IAAI,CAAC,IAAIA,EAAEm3B,EAAEp5B,EAAEywB,GAAEvwB,EAAE,SAAS,CAAC,EAAEgC,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE2B,EAAE,OAAO,IAAID,EAAEi6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAE5B,EAAE,KAAK,GAAG,IAAI6B,EAAE8gC,GAAG3iC,CAAC,EAAE,OAAOm0B,GAAGyE,GAAGj3B,EAAEwuB,GAAEtuB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM8+C,GAAG,CAAC,WAAWhiC,GAAG,aAAa,CAAC,IAAI,OAAO,WAAW,OAAO,EAAE,SAAS,CAACjf,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,gBAAgBI,CAAC,EAAEJ,EAAE,CAACK,EAAE,EAAE,EAAE0B,CAAC,EAAE,EAAEC,EAAED,GAAG,KAAK4gC,GAAG,CAAC,EAAE5gC,EAAEE,EAAE+yB,GAAG,EAAE,MAAM30B,EAAE,KAAK,EAAEqC,EAAE,CAAC,EAAE,GAAG,EAAE,OAAO,EAAE,CAAC,QAAQgC,EAAE,EAAEA,EAAErE,EAAE,MAAM,OAAO,EAAE,EAAEqE,EAAEhC,EAAE,KAAKrC,EAAE,MAAMqE,EAAE,EAAEhC,EAAE,KAAK,CAAC,CAAC,CAAC,IAAIC,EAAE6jC,GAAGnmC,EAAE,CAAC,EAAEuC,EAAEs2B,EAAEp5B,EAAEkC,CAAC,EAAEa,EAAEuqC,GAAGxU,EAAE,EAAE+J,GAAGviC,CAAC,CAAC,CAAC,EAAE0C,EAAEo2B,EAAEA,EAAEA,EAAEr2B,EAAEA,CAAC,EAAEA,CAAC,EAAE8/B,GAAG,GAAG,CAAC,EAAE,MAAM,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE3G,EAAE9C,EAAEA,EAAEp5B,EAAEgkC,GAAG9H,EAAEn5B,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE,CAAC,EAAEH,CAAC,CAAC,EAAEV,CAAC,EAAE3B,EAAE,KAAK,EAAE27B,EAAE9C,EAAEA,EAAEp5B,EAAE+C,CAAC,EAAEb,CAAC,EAAE3B,EAAE,KAAK,EAAE,KAAK,IAAI,CAAC,IAAIqE,EAAEw0B,EAAEA,EAAEr2B,EAAE8/B,GAAG,EAAE,CAAC,EAAE//B,CAAC,EAAE,OAAO,EAAE,OAAO,IAAI8B,EAAEu+B,GAAGv+B,EAAEzC,CAAC,GAAG+5B,EAAEt3B,EAAE,EAAE,KAAK,CAAC,EAAE,SAAS,IAAI,CAAC,IAAIA,EAAEw0B,EAAEA,EAAEp2B,EAAEH,CAAC,EAAEC,CAAC,EAAE,OAAO,EAAE,OAAO,IAAI8B,EAAEu+B,GAAGv+B,EAAEzC,CAAC,GAAG+5B,EAAEt3B,EAAE,EAAE,KAAK,CAAC,EAAE,MAAM,IAAI,CAAC,IAAIA,EAAEw0B,EAAEv2B,EAAEE,CAAC,EAAE8B,EAAEu0B,EAAEp5B,EAAE4E,CAAC,EAAE,OAAO,EAAE,OAAO,IAAIC,EAAEs+B,GAAGt+B,EAAE1C,CAAC,GAAG+5B,EAAEr3B,EAAE,EAAE,KAAK,CAAC,EAAE,OAAO,IAAI,CAAC,IAAID,EAAE5E,EAAE,OAAO,EAAE,OAAO,IAAI4E,EAAEu+B,GAAGv+B,EAAEzC,CAAC,GAAG+5B,EAAEt3B,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAMs8C,GAAG,CAAC,WAAWhiC,GAAG,aAAa,CAAC,IAAI,SAAS,EAAE,SAAS,CAAClf,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAEL,EAAE,EAAE2Y,GAAG,EAAEvY,EAAE,KAAK,EAAE,GAAG,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI4B,EAAE5B,EAAE,MAAM,EAAEC,EAAE,KAAKqC,EAAEV,EAAE,MAAM,EAAE,CAAC,EAAE,EAAEU,EAAE,OAAOE,EAAEZ,EAAE,MAAM,EAAEA,EAAE,MAAM,EAAE,MAAM,CAAC,EAAEa,EAAED,EAAE,OAAOE,EAAEm+C,GAAG,EAAE,CAAC,EAAEj9C,EAAEi9C,GAAG,EAAE,EAAE,EAAE,EAAEp+C,CAAC,EAAEoB,EAAEi9C,GAAG,CAACx+C,EAAE,CAAC,CAAC,EAAEE,CAAC,CAAC,EAAEyE,EAAE20B,EAAEl8B,EAAEmE,CAAC,EAAEY,EAAEm3B,EAAE37B,EAAE,CAAC,CAAC,CAAC,EAAEoE,EAAEy8C,GAAG,CAAC,CAAC,CAAC,EAAEp+C,EAAEkB,CAAC,CAAC,EAAEU,EAAEiwB,GAAGttB,EAAE5C,CAAC,EAAEE,EAAEksC,GAAGnsC,EAAEG,EAAEzE,EAAE,MAAM,EAAE,EAAEwE,EAAEu9B,GAAG19B,CAAC,EAAE,OAAOE,EAAEgwB,GAAGhwB,EAAEC,CAAC,EAAED,CAAC,EAAE,QAAQ,IAAItE,CAAC,CAAC,CAAC,EAAE,SAAS4gD,GAAGnhD,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAEN,EAAEM,EAAE,EAAE,EAAEA,EAAEJ,EAAE,KAAKI,CAAC,EAAE,OAAOJ,CAAC,CAAC,SAASkhD,GAAGphD,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAEE,GAAG,OAAO,EAAEI,EAAE,EAAE,KAAKN,EAAEE,GAAGI,EAAE,EAAE,OAAO,CAAC,CAAC,IAAI+gD,GAAG,CAAC,WAAWhiC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACrf,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI8gC,GAAGlhC,CAAC,EAAE,EAAE,IAAIkhC,GAAG9gC,CAAC,CAAC,CAAC,CAAC,EAAMghD,GAAG,CAAC,WAAWhiC,GAAG,SAAStf,IAAI,CAAC,EAAE,IAAIywB,GAAEzwB,EAAE,SAAS,CAAC,EAAE,EAAMuhD,GAAG,CAAC,WAAW9hC,GAAG,SAASzf,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMwhD,GAAG,CAAC,WAAW9hC,GAAG,SAAS1f,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMyhD,GAAG,CAAC,WAAW9hC,GAAG,SAAS3f,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAM0hD,GAAG,CAAC,WAAW9hC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC5f,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,MAAMC,CAAC,EAAEL,EAAE,EAAEskC,GAAGlkC,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,IAAI4gC,GAAG,EAAElhC,EAAEo5B,EAAEp5B,EAAEO,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMohD,GAAG,CAAC,WAAW1hC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACjgB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAE84B,EAAE54B,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM0hD,GAAG,CAAC,WAAW5hC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAChgB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAEywB,GAAEvwB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,EAAM2hD,GAAG,CAAC,WAAWvhC,GAAG,aAAa,CAAC,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAACtgB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,KAAKC,CAAC,EAAEL,EAAE,MAAM,CAAC,OAAO,IAAI,CAAC,IAAI8B,EAAE0hC,GAAGpjC,CAAC,EAAE,OAAOomC,GAAG1mC,EAAEo5B,EAAE+J,GAAGnjC,EAAEO,EAAE,EAAE,EAAEyB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAAS8/C,GAAG9hD,EAAE,EAAEE,EAAEI,EAAE,EAAEC,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,CAAC,IAAI0B,EAAE,CAAC,EAAEjC,EAAE,EAAE,EAAE,GAAGE,CAAC,EAAEgC,EAAE,CAAC,YAAY5B,EAAE,KAAKC,EAAE,MAAM,EAAE,KAAK,CAAC,EAAE,OAAO2qB,EAAE,UAAUzK,GAAGxe,EAAEC,CAAC,CAAC,CAAC,IAAI6/C,GAAG51B,EAAE,CAAC,oCAAoC21B,EAAE,CAAC,EAAME,GAAG,CAAC,WAAWxhC,GAAG,aAAa,CAAC,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAACxgB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,YAAY,EAAE,KAAK,EAAE,MAAM0B,EAAE,KAAKC,CAAC,EAAEhC,EAAE,MAAM,CAAC,EAAE,IAAI6hD,GAAGzhD,EAAEC,EAAEP,EAAE,EAAE,EAAEiC,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAE,SAAS+/C,GAAGjiD,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,EAAE,KAAKJ,EAAE,OAAO,EAAEg8B,EAAE,EAAEgG,GAAG,EAAE,MAAM5hC,CAAC,CAAC,GAAGN,EAAE,KAAKE,EAAE,OAAOF,EAAEk8B,EAAEl8B,EAAEkiC,GAAGliC,EAAE,MAAMM,CAAC,CAAC,GAAG,CAAC,EAAE,IAAI84B,EAAEp5B,EAAEywB,GAAEuQ,GAAG9gC,EAAE,CAAC,EAAEF,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAIkiD,GAAG,CAAC,WAAWxhC,GAAG,aAAa,CAAC,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAAC1gB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAEJ,EAAE,CAAC,iBAAiBK,CAAC,EAAED,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,GAAG2B,EAAE4W,GAAGtY,EAAE,EAAE,KAAK,EAAE2B,EAAE+/C,GAAGjiD,EAAE,EAAE,EAAEiC,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAMigD,GAAG,CAAC,WAAWxhC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAAC3gB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI84B,EAAEp5B,EAAEywB,GAAEiU,GAAGxkC,EAAEI,CAAC,EAAE,SAAS,CAAC,EAAE,EAAE,IAAI84B,EAAEp5B,EAAEywB,GAAE2U,GAAGllC,EAAEI,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,EAAE,SAAS8hD,GAAGpiD,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,KAAK,eAAe,EAAEkC,EAAE8pB,EAAE,EAAE,QAAQ,eAAe,EAAE7pB,EAAE6pB,EAAE9rB,EAAE,SAAS,eAAe,EAAE0C,EAAEX,EAAEY,EAAEX,EAAEY,EAAEX,EAAEY,EAAE,GAAGb,EAAE,OAAO,IAAIa,EAAE,GAAGH,EAAEs5B,EAAEj6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEY,EAAEq5B,EAAEh6B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAEY,EAAEo5B,EAAE/5B,EAAE,CAAC,EAAEA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,GAAG2V,EAAElV,EAAE,OAAO,EAAE,IAAI,0DAA0DA,EAAE,OAAO,EAAEkV,EAAEjV,EAAE,OAAO,EAAE,IAAI,6DAA6DA,EAAE,OAAO,EAAEiV,EAAEhV,EAAE,OAAO,EAAE,IAAI,8DAA8DA,EAAE,OAAO,EAAEk5B,GAAG,gBAAgB,EAAE,CAAC,EAAE,IAAIh5B,EAAE,CAAC,GAAGJ,EAAE,MAAMC,EAAE,OAAOC,CAAC,EAAEoB,EAAE,CAAC,WAAW5D,EAAE,QAAQC,EAAE,IAAI,EAAE,gBAAgB,CAAC,EAAE,EAAE2qB,EAAE,UAAUnK,GAAG/d,EAAEkB,CAAC,EAAE,OAAOnB,EAAEm5B,EAAE,EAAE,CAAC,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,CAAC,IAAImmB,GAAGl2B,EAAE,CAAC,eAAei2B,EAAE,CAAC,EAAME,GAAG,CAAC,WAAWxhC,GAAG,aAAa,CAAC,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAAC9gB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,CAAC,EAAEhC,EAAE,MAAM,CAAC,EAAE,IAAImiD,GAAGriD,EAAEM,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASqgD,GAAGviD,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAE+pB,EAAEhsB,EAAE,KAAK,aAAa,EAAEkC,EAAE8pB,EAAE,EAAE,QAAQ,aAAa,EAAE7pB,EAAE6pB,EAAE9rB,EAAE,SAAS,aAAa,EAAE4X,EAAE5V,EAAE,OAAOD,EAAE,KAAK,IAAI,kBAAkBC,EAAE,oCAAoCD,EAAE,OAAO,EAAE6V,EAAE7V,EAAE,OAAO,EAAE,IAAI,wDAAwDA,EAAE,OAAO,EAAE6V,EAAE5V,EAAE,OAAO,EAAE,IAAI,2DAA2DA,EAAE,OAAO,EAAE85B,GAAG,cAAc,EAAE,CAAC,EAAE,IAAIp5B,EAAE,CAAC,GAAGX,EAAE,MAAMC,EAAE,OAAOC,CAAC,EAAEU,EAAE,CAAC,WAAWvC,EAAE,QAAQC,EAAE,IAAI,EAAE,gBAAgB,CAAC,EAAE,OAAO2qB,EAAE,UAAUrK,GAAGje,EAAEC,CAAC,CAAC,CAAC,IAAI2/C,GAAGr2B,EAAE,CAAC,aAAao2B,EAAE,CAAC,EAAME,GAAG,CAAC,WAAW7hC,GAAG,aAAa,CAAC,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAAC5gB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,CAAC,EAAE/B,EAAE,MAAM,CAAC,EAAE,IAAIsiD,GAAGxiD,EAAEM,EAAEC,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,EAAMygD,GAAG,CAAC,WAAWzhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACjhB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,KAAKC,CAAC,EAAEL,EAAE,EAAE2Y,GAAGtY,EAAED,EAAE,KAAK,EAAE2B,EAAEggC,GAAG3hC,EAAE,MAAM,CAAC,EAAE,GAAG4B,EAAEkW,GAAGnW,CAAC,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAIW,EAAEtC,EAAE,MAAM,MAAM,EAAE,EAAE,QAAQyC,GAAG,CAACH,EAAEG,GAAG,CAAC,CAAC,EAAE,IAAIF,EAAEq5B,EAAEl8B,EAAE4C,CAAC,EAAE,OAAOs2B,GAAGE,EAAEv2B,EAAEylC,GAAGhoC,EAAE,MAAM,SAAS,CAAC,EAAE4B,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMygD,GAAG,CAAC,WAAWzhC,GAAG,aAAa,CAAC,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAAClhB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAEJ,EAAE,CAAC,KAAKK,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE2B,EAAE4W,GAAGtY,EAAE,EAAE,KAAK,EAAE2B,EAAE+/C,GAAGjiD,EAAE,EAAE,EAAEiC,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAM0gD,GAAG,CAAC,WAAWzhC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACnhB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI84B,EAAEp5B,EAAEywB,GAAE6U,GAAGplC,EAAEI,CAAC,EAAE,SAAS,CAAC,EAAE,EAAE,IAAI84B,EAAEp5B,EAAEywB,GAAE+T,GAAGtkC,EAAEI,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,EAAMuiD,GAAG,CAAC,WAAWzhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACphB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAE,EAAE,GAAG,CAAC,SAASC,CAAC,EAAEL,EAAE,EAAEK,EAAE,IAAI,GAAG,EAAE,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIq8B,GAAG58B,EAAE,EAAEM,EAAE,KAAK,CAAC,CAAC,CAAC,EAAMwiD,GAAG,CAAC,WAAWzhC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACrhB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEizB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO0B,EAAE,OAAO,EAAEi6B,EAAEiH,GAAGnjC,EAAEiC,CAAC,EAAE/B,EAAE,KAAK,EAAEF,CAAC,EAAE,EAAE,IAAI,CAAC,IAAIiC,EAAEm3B,EAAEp5B,EAAEy0B,GAAG2P,GAAGlL,GAAGh5B,EAAEI,CAAC,CAAC,CAAC,CAAC,EAAE4B,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE,OAAO2B,EAAE,OAAO,EAAEg6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAE5B,EAAE,KAAK,EAAE2B,CAAC,CAAC,CAAC,CAAC,EAAM8gD,GAAG,CAAC,WAAWxhC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACvhB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEm3B,EAAEp5B,EAAEywB,GAAEnwB,EAAE,SAAS,CAAC,EAAE4B,EAAEgzB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO2B,EAAE,OAAO,EAAEg6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAEhC,EAAE,KAAK,EAAE+B,CAAC,EAAE,EAAE,IAAI,CAAC,IAAIA,EAAEm3B,EAAEp5B,EAAEywB,GAAEvwB,EAAE,SAAS,CAAC,EAAEgC,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE,OAAO2B,EAAE,OAAO,EAAEg6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAE5B,EAAE,KAAK,EAAE2B,CAAC,CAAC,CAAC,CAAC,EAAM+gD,GAAG,CAAC,WAAWxhC,GAAG,SAASxhB,IAAI,CAAC,EAAE,IAAIy0B,GAAGz0B,CAAC,CAAC,EAAE,EAAMijD,GAAG,CAAC,WAAWnhC,GAAG,aAAa,CAAC,SAAS,EAAE,SAAS,CAAC9hB,EAAE,IAAI,CAAC,IAAIE,EAAE,EAAE,GAAG,MAAM,CAAC,QAAQ,IAAImoC,GAAGnoC,EAAE,MAAM,SAAS,CAAC,CAAC,CAAC,EAAMgjD,GAAG,CAAC,WAAWrhC,GAAG,SAAS7hB,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMmjD,GAAG,CAAC,WAAWphC,GAAG,cAAc,GAAG,SAAS,CAAC/hB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,KAAKI,CAAC,EAAEJ,EAAE,OAAO+wC,GAAGjxC,EAAEM,CAAC,EAAE,IAAIyB,GAAG,IAAIA,CAAC,CAAC,CAAC,EAAMqhD,GAAG,CAAC,WAAWphC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAChiB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAE,EAAE,GAAG,CAAC,SAASC,CAAC,EAAEL,EAAE,EAAEK,EAAE,IAAI,GAAG,EAAE,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIq8B,GAAG58B,EAAE,EAAEM,EAAE,KAAK,CAAC,CAAC,CAAC,EAAM+iD,GAAG,CAAC,WAAWnhC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,CAACliB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,EAAEC,CAAC,EAAE,EAAE,EAAEL,EAAE,EAAEI,EAAE2B,EAAE+yB,GAAG,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAIpyB,EAAE6tB,GAAE,EAAE,SAAS,EAAE5tB,EAAEu2B,EAAEp5B,EAAEo5B,EAAEx2B,EAAEggC,GAAG,EAAE8D,GAAG9jC,EAAEigC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE//B,EAAEoyB,GAAG,EAAE,MAAMjzB,CAAC,EAAE,OAAOa,EAAE,OAAO,IAAID,EAAEsgC,GAAGtgC,EAAEC,CAAC,GAAGo5B,EAAEr5B,EAAE,EAAE,KAAK,CAAC,EAAE,EAAE,IAAI,CAAC,IAAID,EAAE4hC,GAAG,EAAE,CAAC,EAAE3hC,EAAEq+B,GAAGt+B,EAAE+iC,GAAG,CAAC,EAAEvE,GAAG,CAAC,CAAC,EAAEt+B,EAAEs2B,EAAEp5B,EAAEo5B,EAAE74B,EAAEsC,CAAC,CAAC,EAAEE,EAAEmyB,GAAG,EAAE,MAAMjzB,CAAC,EAAE,OAAOc,EAAE,OAAO,IAAID,EAAEqgC,GAAGrgC,EAAEC,CAAC,GAAGm5B,EAAEp5B,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAMwgD,GAAG,CAAC,WAAWnhC,GAAG,aAAa,CAAC,IAAI,OAAO,EAAE,SAAS,CAACniB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEikC,GAAGtkC,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIghC,GAAG3gC,EAAEP,EAAEo5B,EAAEp5B,EAAEM,CAAC,CAAC,EAAE,MAAM,IAAI,CAAC,IAAI,EAAE4gC,GAAG3gC,EAAE6gC,GAAGphC,CAAC,EAAEo5B,EAAEp5B,EAAEE,CAAC,CAAC,EAAE,EAAEg1B,GAAG50B,EAAE,MAAMN,EAAE,KAAK,EAAE,OAAO,EAAE,OAAO,IAAI,EAAEmjC,GAAG,EAAE,CAAC,GAAGjH,EAAE,EAAE57B,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,SAASijD,GAAGvjD,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,MAAM,EAAEM,EAAEJ,GAAG,EAAE,IAAIK,EAAE27B,EAAE,EAAE57B,CAAC,EAAE,EAAE4/B,GAAGlgC,EAAEE,EAAE,GAAG,EAAE,EAAE,EAAEggC,GAAGlgC,EAAEE,EAAE,GAAG,EAAE,EAAE+B,EAAEm3B,EAAE,EAAE,CAAC,EAAE,OAAOA,EAAE74B,EAAE0B,CAAC,CAAC,CAAC,SAASuhD,GAAGxjD,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,OAAOO,EAAED,EAAEJ,EAAE,OAAO,EAAEs6C,EAAE,mBAAmBt6C,EAAEI,CAAC,EAAE,EAAEN,EAAE,GAAG,OAAO,EAAE60B,GAAG70B,EAAE,CAAC,GAAG,IAAIiC,EAAE,EAAE,MAAM,MAAM,EAAE,EAAEA,EAAE,OAAO3B,EAAEJ,EAAE,OAAOA,EAAE,MAAM,EAAE,OAAO,CAAC4C,EAAEC,IAAID,EAAEC,EAAE,CAAC,EAAEd,EAAE,KAAK,CAAC,EAAE,IAAIW,EAAE,EAAE,QAAQX,CAAC,EAAE,EAAEshD,GAAG3gD,EAAE,EAAErC,CAAC,EAAE,GAAG,EAAE,EAAE,QAAQ,EAAE,KAAK,EAAE,GAAG,KAAK,CAAC,IAAIuC,EAAE03C,EAAE,uBAAuB,CAAC,EAAE,EAAE3lB,GAAG,EAAE/xB,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI2gD,GAAG,CAAC,WAAWrhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACpiB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,KAAKC,CAAC,EAAEL,EAAE,EAAE,CAAC,EAAE,OAAOK,GAAG,KAAK,EAAED,EAAE,MAAM,IAAI,CAAC,EAAE2B,IAAIA,CAAC,EAAE,OAAO1B,GAAG,SAAS,EAAE,CAACA,CAAC,EAAE,EAAEA,EAAE,CAAC,EAAE,IAAIijD,GAAGljD,EAAEN,EAAE,CAAC,CAAC,CAAC,CAAC,EAAM0jD,GAAG,CAAC,WAAWvlC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACne,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEi3B,GAAGl5B,EAAEywB,GAAEnwB,EAAE,SAAS,CAAC,EAAE4B,EAAEgzB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO2B,EAAE,OAAO,EAAEg6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAEhC,EAAE,KAAK,EAAE+B,CAAC,EAAE,EAAE,IAAI,CAAC,IAAIA,EAAEm3B,EAAEp5B,EAAEywB,GAAEvwB,EAAE,SAAS,CAAC,EAAEgC,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE2B,EAAE,OAAO,IAAID,EAAEi6B,EAAEiH,GAAGlhC,EAAEC,CAAC,EAAE5B,EAAE,KAAK,GAAG,IAAI6B,EAAE8gC,GAAG3iC,CAAC,EAAE,OAAOm0B,GAAGyE,GAAGj3B,EAAEwuB,GAAEtuB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMwhD,GAAG,CAAC,WAAWjhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC1iB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAEy0B,GAAGwO,GAAG/iC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM0jD,GAAG,CAAC,WAAW3gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACjjB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAEI,EAAE84B,EAAEkM,GAAGplC,EAAE,CAAC,EAAE4vC,GAAG5vC,CAAC,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEywB,GAAEnwB,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,EAAMujD,GAAG,CAAC,WAAWlhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC3iB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEywB,GAAEqf,GAAG5vC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,EAAM4jD,GAAG,CAAC,WAAWlhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC5iB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg8B,EAAEl8B,EAAEE,EAAE,KAAK,CAAC,CAAC,CAAC,EAAM6jD,GAAG,CAAC,WAAWhhC,GAAG,aAAa,CAAC,QAAQ,EAAE,SAAS,CAAC/iB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAEC,EAAE,CAAC,GAAGP,EAAE,OAAOM,CAAC,EAAE,MAAM,CAAC,OAAO,IAAI4qB,EAAE,UAAUlI,GAAGziB,EAAEL,CAAC,CAAC,CAAC,CAAC,EAAM8jD,GAAG,CAAC,WAAWnhC,GAAG,aAAa,CAAC,QAAQ,EAAE,SAAS,CAAC7iB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAEC,EAAE,CAAC,GAAGP,EAAE,OAAOM,CAAC,EAAE,MAAM,CAAC,OAAO,IAAI4qB,EAAE,UAAUpI,GAAGviB,EAAEL,CAAC,CAAC,CAAC,CAAC,EAAM+jD,GAAG,CAAC,WAAW/gC,GAAG,SAAS,CAACljB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,KAAKI,CAAC,EAAEJ,EAAEK,EAAEsY,GAAGvY,EAAEN,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI0sC,GAAG1sC,EAAEO,CAAC,CAAC,CAAC,CAAC,EAAM2jD,GAAG,CAAC,WAAW/gC,GAAG,SAASnjB,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMmkD,GAAG,CAAC,WAAW/gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACpjB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIu0B,GAAGyE,GAAGl5B,EAAEo5B,EAAEwJ,GAAG1iC,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMkkD,GAAG,CAAC,WAAW7gC,GAAG,aAAa,CAAC,WAAW,EAAE,SAAS,CAACvjB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,UAAU,IAAIuwB,GAAE2Q,GAAGlhC,CAAC,EAAE,SAAS,EAAE,EAAE,IAAIk5B,EAAEp5B,EAAEywB,GAAEvwB,EAAEF,EAAE,KAAK,CAAC,EAAE,EAAE,IAAIo5B,EAAEp5B,EAAEywB,GAAEyW,GAAGhnC,CAAC,EAAEF,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAMqkD,GAAG,CAAC,WAAW7gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACxjB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAII,EAAEkkC,GAAGtkC,EAAE2iC,GAAG,CAAC,CAAC,EAAEtiC,EAAEsiC,GAAGqY,EAAE,EAAE,EAAErY,GAAGoY,EAAE,EAAE,EAAE7hB,EAAEp5B,EAAE,CAAC,EAAEiC,EAAEm3B,EAAEA,EAAEp5B,EAAEO,CAAC,EAAEmjC,GAAGjT,GAAEvwB,EAAE,SAAS,CAAC,CAAC,EAAE,OAAOghC,GAAG5gC,EAAE,EAAE2B,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMqiD,GAAG,CAAC,WAAWzgC,GAAG,cAAc,CAAC,EAAE,EAAE,SAAS,CAAC7jB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEo5B,EAAEl5B,EAAEwmC,GAAG7D,GAAG,CAAC,EAAE3iC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMqkD,GAAG,CAAC,WAAW3gC,GAAG,SAAS5jB,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMwkD,GAAG,CAAC,WAAW9gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC1jB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAE0G,GAAGrP,GAAEvwB,EAAE,SAAS,CAAC,EAAEF,CAAC,CAAC,CAAC,CAAC,EAAMykD,GAAG,CAAC,WAAW9gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC3jB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAE4G,GAAGvP,GAAEvwB,EAAE,SAAS,CAAC,EAAEF,CAAC,CAAC,CAAC,CAAC,EAAM0kD,GAAG,CAAC,WAAWjhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACzjB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,MAAMC,EAAE,KAAK,CAAC,EAAEL,EAAE,EAAEI,EAAE,MAAM,CAAC2B,EAAEC,CAAC,EAAEy0B,GAAGr2B,EAAEC,EAAE,CAAC,EAAE4B,EAAE,CAAC,EAAE,QAAQS,EAAE,EAAEA,EAAE5C,EAAE,KAAK4C,IAAIT,EAAE,KAAK,CAACF,EAAEW,GAAG,EAAEA,GAAGX,EAAEW,GAAGV,EAAEU,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,IAAI+mC,GAAG3pC,EAAEmC,CAAC,CAAC,CAAC,CAAC,EAAMwiD,GAAG,CAAC,WAAWxgC,GAAG,cAAc,CAAC,EAAE,EAAE,SAAS,CAACnkB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,IAAIC,CAAC,EAAEL,EAAE,EAAE,GAAG,EAAEk5B,EAAEp5B,EAAEM,CAAC,EAAE,MAAM,CAAC,OAAO,IAAIomC,GAAG,EAAEtN,EAAE+J,GAAG,EAAE,CAAC5iC,CAAC,EAAE,CAAC,EAAED,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMskD,GAAG,CAAC,WAAW9gC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC9jB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAE08B,GAAGx8B,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM2kD,GAAG,CAAC,WAAW5gC,GAAG,SAAS,CAACjkB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,WAAWI,EAAE,SAASC,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAIg9B,GAAGl9B,EAAEM,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAMukD,GAAG,CAAC,WAAW5gC,GAAG,SAAS,CAAClkB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAAC,KAAKI,CAAC,EAAEJ,EAAE,MAAM,CAAC,EAAE,IAAIs8B,GAAGx8B,EAAEM,CAAC,CAAC,CAAC,CAAC,EAAMykD,GAAG,CAAC,WAAWhhC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC/jB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAEo5B,EAAE2J,GAAGtS,GAAEvwB,EAAE,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM8kD,GAAG,CAAC,WAAWtgC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC1kB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEp5B,EAAEo5B,EAAE3I,GAAEvwB,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAM+kD,GAAG,CAAC,WAAWxgC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAACzkB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEsiC,GAAG,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIzJ,EAAEp5B,EAAEo5B,EAAE74B,EAAEmmC,GAAGxmC,EAAEI,CAAC,CAAC,CAAC,EAAE,EAAE,IAAI84B,EAAEp5B,EAAEo5B,EAAE74B,EAAEmmC,GAAGpmC,EAAEJ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMglD,GAAG,CAAC,WAAWv/B,GAAG,SAAS3lB,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAMmlD,GAAG,CAAC,WAAWpgC,GAAG,aAAa,CAAC,IAAI,GAAG,EAAE,SAAS,CAAC/kB,EAAE,IAAI,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAE,EAAEC,EAAEy0B,GAAG90B,EAAE,MAAMI,EAAE,KAAK,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI2B,EAAEjC,EAAEkC,EAAEgzB,GAAGh1B,EAAE,MAAMK,CAAC,EAAE,OAAO2B,EAAE,OAAO,IAAID,EAAEkhC,GAAGlhC,EAAEC,CAAC,GAAGg6B,EAAEj6B,EAAE/B,EAAE,KAAK,CAAC,EAAE,EAAE,IAAI,CAAC,IAAI+B,EAAEjC,EAAEkC,EAAEgzB,GAAG50B,EAAE,MAAMC,CAAC,EAAE,OAAO2B,EAAE,OAAO,IAAID,EAAEkhC,GAAGlhC,EAAEC,CAAC,GAAGg6B,EAAEzH,GAAGxyB,CAAC,EAAE3B,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,EAAM8kD,GAAG,CAAC,WAAWphC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAChkB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAEC,EAAED,EAAE,MAAM,MAAM,EAAE,CAAC,KAAK,CAAC,EAAEJ,EAAE2Y,GAAG,EAAEvY,EAAE,KAAK,EAAE,QAAQ,GAAG,CAACC,EAAE,GAAG,CAAC,CAAC,EAAE,IAAI0B,EAAEi6B,EAAEl8B,EAAEO,CAAC,EAAE2B,EAAEk3B,EAAEn3B,EAAEqmC,GAAGhoC,EAAE,MAAM,SAAS,CAAC,EAAE,MAAM,CAAC,EAAE,IAAI4B,CAAC,CAAC,CAAC,EAAMmjD,GAAG,CAAC,WAAWrgC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAAChlB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIg5B,GAAGl5B,EAAEijC,GAAGnD,GAAG5/B,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAMolD,GAAG,CAAC,WAAWrgC,GAAG,cAAc,CAAC,EAAE,EAAE,SAAS,CAACjlB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIk5B,EAAEsN,GAAG7D,GAAG,CAAC,EAAEI,GAAG/iC,CAAC,CAAC,EAAEF,CAAC,CAAC,CAAC,CAAC,EAAMulD,GAAG,CAAC,WAAWrgC,GAAG,aAAa,CAAC,GAAG,EAAE,SAAS,CAACllB,EAAE,EAAEE,IAAI,CAAC,GAAG,CAACI,CAAC,EAAE,EAAE,CAAC,KAAKC,CAAC,EAAEL,EAAE,MAAM,CAAC,EAAE,IAAI,CAAC,IAAI8B,EAAEo/B,GAAG9gC,CAAC,EAAE,GAAGA,EAAE,OAAO,EAAE,QAAQ2B,EAAE,EAAEA,EAAE1B,EAAE,GAAG,EAAE0B,EAAED,EAAE82B,EAAE92B,EAAE46B,GAAG58B,EAAE,CAACiC,EAAE3B,EAAE,MAAM,EAAE,EAAE,CAACA,EAAE,MAAM,EAAE,CAAC,CAAC,UAAUA,EAAE,OAAO,EAAE,QAAQ2B,EAAE,EAAEA,EAAE1B,EAAE,GAAG,EAAE0B,EAAE,QAAQC,EAAE,EAAEA,EAAE3B,EAAE,GAAG,EAAE2B,EAAEF,EAAE82B,EAAE92B,EAAE46B,GAAG58B,EAAE,CAACiC,EAAE3B,EAAE,MAAM,GAAG4B,EAAE5B,EAAE,MAAM,EAAE,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,CAAC,UAAUA,EAAE,OAAO,EAAE,QAAQ2B,EAAE,EAAEA,EAAE1B,EAAE,GAAG,EAAE0B,EAAE,QAAQC,EAAE,EAAEA,EAAE3B,EAAE,GAAG,EAAE2B,EAAE,QAAQ,EAAE,EAAE,EAAE3B,EAAE,GAAG,EAAE,EAAEyB,EAAE82B,EAAE92B,EAAE46B,GAAG58B,EAAE,CAACiC,EAAE3B,EAAE,MAAM,GAAG4B,EAAE5B,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,EAAE,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,CAAC,UAAUA,EAAE,OAAO,EAAE,QAAQ2B,EAAE,EAAEA,EAAE1B,EAAE,GAAG,EAAE0B,EAAE,QAAQC,EAAE,EAAEA,EAAE3B,EAAE,GAAG,EAAE2B,EAAE,QAAQ,EAAE,EAAE,EAAE3B,EAAE,GAAG,EAAE,EAAE,QAAQqC,EAAE,EAAEA,EAAErC,EAAE,GAAG,EAAEqC,EAAEZ,EAAE82B,EAAE92B,EAAE46B,GAAG58B,EAAE,CAACiC,EAAE3B,EAAE,MAAM,GAAG4B,EAAE5B,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,GAAGsC,EAAEtC,EAAE,MAAM,EAAE,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,CAAC,MAAO,OAAM,IAAI,MAAM,2DAA2DA,EAAE,mBAAmB,EAAE,OAAO0B,CAAC,CAAC,CAAC,CAAC,EAAMwjD,GAAG,CAAC,WAAWngC,GAAG,SAAS,CAACrlB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAEJ,EAAE,CAAC,KAAKK,CAAC,EAAED,EAAE,EAAE+hC,GAAG9hC,CAAC,EAAE,MAAM,CAAC,EAAE,IAAIs0B,GAAG70B,EAAE,CAAC,CAAC,CAAC,CAAC,EAAMylD,GAAG,CAAC,WAAWlgC,GAAG,SAAS,CAACvlB,EAAE,EAAEE,IAAI,CAAC,IAAII,EAAEJ,EAAE,CAAC,KAAKK,CAAC,EAAED,EAAE,MAAM,CAAC,MAAM,IAAIsvC,GAAG5vC,EAAEO,CAAC,CAAC,CAAC,CAAC,EAAMmlD,GAAG,CAAC,WAAWlgC,GAAG,aAAa,CAAC,YAAY,EAAE,SAAS,CAACxlB,EAAE,IAAI,CAAC,GAAG,CAACE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAE,IAAIylD,GAAG3lD,EAAEE,CAAC,CAAC,CAAC,CAAC,EAAE,SAASylD,GAAG3lD,EAAE,EAAE,CAAC,IAAIE,EAAEgoC,GAAG,EAAE9G,GAAG,CAAC,CAAC,EAAE9gC,EAAEgkC,GAAGtkC,EAAEE,CAAC,EAAEK,EAAEmkC,GAAG,EAAE7B,GAAG,EAAE,OAAO,CAAC,EAAE,EAAEviC,EAAE,KAAKC,EAAE,KAAK,QAAQ0B,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE1B,EAAEqjC,GAAGrjC,EAAE0B,EAAE,CAAC,EAAE1B,EAAEymC,GAAGzmC,EAAE+nC,GAAGhoC,EAAE,MAAM,MAAM,CAAC,EAAE,IAAI,EAAE8gC,GAAG9gC,CAAC,EAAE,OAAO4gC,GAAG3gC,EAAED,EAAE,CAAC,CAAC,CAAC,IAAIslD,GAAG,CAAC,WAAWlgC,GAAG,SAAS1lB,IAAI,CAAC,EAAE,IAAIohC,GAAGphC,CAAC,CAAC,EAAE,EAAM6lD,GAAG,CAACvH,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGG,GAAGG,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGE,GAAGD,GAAGI,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGiD,GAAGhD,GAAGC,GAAGC,GAAGC,GAAGC,GAAGE,GAAGD,GAAGE,GAAGC,GAAGG,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGG,GAAGE,GAAGA,GAAGC,GAAGG,GAAGG,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGA,GAAGC,GAAGC,GAAGG,GAAGE,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGA,GAAGC,GAAGA,GAAGC,GAAGE,GAAGD,GAAGE,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGC,GAAGE,EAAE,EAAE,QAAQ5lD,KAAK6lD,GAAGn/B,GAAG1mB,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwP,GAAG,IAAI,CAAC,EAAExP,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE0P,GAAG,IAAI,CAAC,EAAE1P,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE4P,GAAG,IAAI,CAAC,EAAE5P,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE84B,EAAE,KAAK94B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE85B,GAAG,KAAK95B,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEg6B,GAAG,KAAKh6B,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk6B,GAAG,KAAKl6B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEo6B,GAAG,KAAKp6B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,SAAS,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEhS,EAAE,KAAK,OAAO,EAAE,IAAI,qCAAqC,EAAEokB,EAAE,KAAK,CAAC,CAAC,CAAC,EAAEpS,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEywB,GAAE,KAAKzwB,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEoS,EAAE,KAAK,CAAC,KAAK,IAAI,CAAC,CAAC,EAAEpS,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk8B,EAAE,KAAK,CAACl8B,EAAE,CAAC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEg8B,EAAE,KAAK,CAACl8B,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE47B,EAAE,KAAK,CAACl8B,EAAE,EAAEE,EAAEI,CAAC,CAAC,CAAC,EAAEwpB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE27B,EAAE,KAAK,CAACl8B,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwQ,GAAG,IAAI,CAAC,EAAExQ,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE0Q,GAAG,IAAI,CAAC,EAAE1Q,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE4Q,GAAG,IAAI,CAAC,EAAE5Q,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE46B,GAAG,KAAK56B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgR,GAAG,IAAI,CAAC,EAAEhR,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE87B,GAAG,KAAKp8B,EAAE,EAAEE,EAAEI,CAAC,CAAC,EAAEwpB,EAAE,EAAE,UAAU,eAAe,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk9B,GAAG,KAAKl9B,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE88B,GAAG,KAAKr9B,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,YAAY,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEi+B,GAAG,KAAKj+B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEywB,GAAE,KAAKzwB,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEqU,GAAG,IAAI,CAAC,EAAErU,EAAE,EAAE,UAAU,YAAY,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEs+B,GAAG,KAAKt+B,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEA,aAAa6pB,KAAK7pB,EAAE,CAACA,CAAC,GAAGw8B,GAAG,CAAC,KAAK,GAAGx8B,CAAC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE2+B,GAAG,KAAKl/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,gBAAgB,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE++B,GAAG,KAAKt/B,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEy+B,GAAG,KAAKh/B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgW,GAAG,IAAI,CAAC,EAAEhW,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEkW,GAAG,IAAI,CAAC,EAAElW,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEggC,GAAG,KAAKlgC,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEkgC,GAAG,KAAKpgC,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,aAAa,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwgC,GAAG,KAAKxgC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,gBAAgB,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEmgC,GAAG,KAAK1gC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,WAAW,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEugC,GAAG,KAAK9gC,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,SAAS,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEshC,GAAG,KAAKthC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk5B,GAAG,KAAKl5B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwhC,GAAG,KAAKxhC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE8X,GAAG,IAAI,CAAC,EAAE9X,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEghC,GAAG,KAAKhhC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgY,GAAG,IAAI,CAAC,EAAEhY,EAAE,EAAE,UAAU,cAAc,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwjC,GAAG,KAAKxjC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE4Z,GAAG,IAAI,CAAC,EAAE5Z,EAAE,EAAE,UAAU,WAAW,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE4jC,GAAG,KAAK5jC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEga,GAAG,IAAI,CAAC,EAAEha,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEglB,GAAG,IAAI,CAAC,EAAEhlB,EAAE,EAAE,UAAU,QAAQ,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEoS,EAAE,KAAK,CAAC,KAAK,IAAI,CAAC,CAAC,EAAEpS,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEsa,GAAG,IAAI,CAAC,EAAEta,EAAE,EAAE,UAAU,SAAS,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEg5B,GAAG,KAAKh5B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEskC,GAAG,KAAKtkC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,aAAa,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0kC,GAAG,KAAK1kC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwkC,GAAG,KAAKxkC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEklB,GAAG,IAAI,CAAC,EAAEllB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEolB,GAAG,IAAI,CAAC,EAAEplB,EAAE,EAAE,UAAU,SAAS,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE8a,GAAG,IAAI,CAAC,EAAE9a,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgb,GAAG,IAAI,CAAC,EAAEhb,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEkb,GAAG,IAAI,CAAC,EAAElb,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEklC,GAAG,KAAKllC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEslC,GAAG,KAAKtlC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEolC,GAAG,KAAKplC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,2BAA2B,SAAS9pB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEmlC,GAAG,KAAKzlC,EAAE,EAAEE,EAAEI,CAAC,CAAC,EAAEwpB,EAAE,EAAE,UAAU,WAAW,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE0c,GAAG,IAAI,CAAC,EAAE1c,EAAE,EAAE,UAAU,WAAW,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE4mC,GAAG,KAAK5mC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE8mC,GAAG,KAAK9mC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE6b,GAAG,IAAI,CAAC,EAAE7b,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE+b,GAAG,IAAI,CAAC,EAAE/b,EAAE,EAAE,UAAU,WAAW,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEgnC,GAAG,KAAKhnC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,WAAW,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEod,GAAG,IAAI,CAAC,EAAEpd,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEonC,GAAG,KAAKpnC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,WAAW,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEsnC,GAAG,KAAKtnC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE4yB,GAAG,KAAK9yB,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEsnC,GAAG,KAAK5nC,EAAE,EAAEE,EAAEI,CAAC,CAAC,EAAEwpB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwiC,GAAG,KAAKxiC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEkoC,GAAG,KAAKloC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEooC,GAAG,KAAKpoC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0iC,GAAG,KAAK1iC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEyoC,GAAG,KAAKzoC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE2oC,GAAG,KAAK3oC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE6oC,GAAG,KAAK7oC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEo5B,EAAE,KAAKp5B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE2K,GAAG,IAAI,CAAC,EAAE3K,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEojC,GAAG,KAAKtjC,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,SAAS,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEqpC,GAAG,KAAKrpC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,EAAE,EAAEE,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE8yB,GAAG,KAAKhzB,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,SAAS,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEyf,GAAG,IAAI,CAAC,EAAEzf,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE2pC,GAAG,KAAK3pC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEkqC,GAAG,KAAKzqC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE4iC,GAAG,KAAK5iC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE2qC,GAAG,KAAK3qC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE6qC,GAAG,KAAK7qC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,WAAW,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEsiB,GAAG,IAAI,CAAC,EAAEtiB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwiB,GAAG,IAAI,CAAC,EAAExiB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE0iB,GAAG,IAAI,CAAC,EAAE1iB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk8B,EAAE,KAAKl8B,EAAE,KAAK,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEk8B,EAAE,KAAKl8B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,eAAe,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEi2C,GAAG,KAAKn2C,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,sBAAsB,SAAS9pB,EAAE,EAAEE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEm2C,GAAG,KAAKr2C,EAAE,EAAEE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0sC,GAAG,KAAK1sC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwlB,GAAG,IAAI,CAAC,EAAExlB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEsjB,GAAG,IAAI,CAAC,EAAEtjB,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwjB,GAAG,IAAI,CAAC,EAAExjB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE0jB,GAAG,IAAI,CAAC,EAAE1jB,EAAE,EAAE,UAAU,gBAAgB,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEmtC,GAAG,KAAK1tC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,CAAC,EAAEupB,EAAE,EAAE,UAAU,QAAQ,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE4S,GAAG,IAAI,CAAC,EAAE5S,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgkB,GAAG,IAAI,CAAC,EAAEhkB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEkkB,GAAG,IAAI,CAAC,EAAElkB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEokB,GAAG,IAAI,CAAC,EAAEpkB,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE48B,GAAG,KAAK58B,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE4uC,GAAG,KAAK5uC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,SAAS,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEwc,GAAG,IAAI,CAAC,EAAExc,EAAE,EAAE,UAAU,eAAe,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEqqC,GAAG,KAAKrqC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEovC,GAAG,KAAKpvC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEiZ,GAAG,IAAI,CAAC,EAAEjZ,EAAE,EAAE,UAAU,OAAO,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEmZ,GAAG,IAAI,CAAC,EAAEnZ,EAAE,EAAE,UAAU,kBAAkB,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEwvC,GAAG,KAAKxvC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0vC,GAAG,KAAK1vC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,EAAE,CAAC,KAAK,gBAAgB,EAAE,IAAIE,EAAEF,aAAa6pB,GAAG,CAAC,KAAK7pB,CAAC,EAAE,CAAC,KAAK,GAAGA,CAAC,EAAE,OAAO4vC,GAAG1vC,EAAE,CAAC,CAAC,EAAE4pB,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE8vC,GAAG,KAAK9vC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,aAAa,SAAS9pB,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE+tC,GAAG,KAAKhwC,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,CAAC,CAAC,EAAE6nB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE0mC,GAAG,KAAK1mC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEmjC,GAAG,KAAKnjC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,IAAI,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEomB,GAAG,IAAI,CAAC,EAAEpmB,EAAE,EAAE,UAAU,KAAK,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEgT,GAAG,IAAI,CAAC,EAAEhT,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEgkC,GAAG,KAAKhkC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE2G,GAAE,KAAK,MAAM,CAAC,EAAE3G,EAAE,EAAE,UAAU,QAAQ,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE2G,GAAE,KAAK,SAAS,CAAC,EAAE3G,EAAE,EAAE,UAAU,MAAM,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAE2G,GAAE,KAAK,OAAO,CAAC,EAAE3G,EAAE,EAAE,UAAU,KAAK,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEywC,GAAG,KAAKzwC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE60B,GAAG,KAAK70B,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,OAAO,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE6wC,GAAG,KAAK7wC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,mBAAmB,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE+wC,GAAG,KAAK/wC,EAAE,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,QAAQ,SAAS9pB,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEixC,GAAG,KAAKjxC,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,MAAM,SAAS9pB,EAAE,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEkhC,GAAGlhC,EAAE,KAAK,CAAC,CAAC,EAAE8pB,EAAE,EAAE,UAAU,UAAU,UAAU,CAAC,OAAO,KAAK,gBAAgB,EAAEsX,GAAG,IAAI,CAAC,EAAE,IAAI0kB,GAAG,cAAc,KAAK,CAAC,YAAYhlD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAO,eAAe,KAAKglD,GAAG,SAAS,CAAC,CAAC,EAAEC,GAAG,cAAc,KAAK,CAAC,YAAYjlD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAO,eAAe,KAAKilD,GAAG,SAAS,CAAC,CAAC,EAAEC,EAAE,cAAc,KAAK,CAAC,YAAYllD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAO,eAAe,KAAKklD,EAAE,SAAS,CAAC,CAAC,EAAEC,GAAG,cAAc,KAAK,CAAC,YAAYnlD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAO,eAAe,KAAKmlD,GAAG,SAAS,CAAC,CAAC,EAAEC,GAAG,cAAc,KAAK,CAAC,YAAYplD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAO,eAAe,KAAKolD,GAAG,SAAS,CAAC,CAAC,EAAMC,GAAG,KAAK,CAAC,YAAYrlD,EAAE,CAAC,KAAK,WAAWA,GAAG,IAAI,KAAK,MAAM,IAAI,GAAG,CAAC,IAAIA,EAAE,CAAC,IAAIZ,EAAE,OAAO,KAAK,MAAM,IAAIY,CAAC,IAAIZ,EAAE,KAAK,MAAM,IAAIY,CAAC,EAAE,KAAK,MAAM,OAAOA,CAAC,EAAE,KAAK,MAAM,IAAIA,EAAEZ,CAAC,GAAGA,CAAC,CAAC,IAAIY,EAAEZ,EAAE,CAAC,GAAG,KAAK,MAAM,IAAIY,CAAC,EAAE,KAAK,MAAM,OAAOA,CAAC,UAAU,KAAK,MAAM,MAAM,KAAK,WAAW,CAAC,IAAI,EAAE,KAAK,MAAM,KAAK,EAAE,KAAK,EAAE,MAAM,KAAK,MAAM,OAAO,CAAC,CAAC,CAAC,KAAK,MAAM,IAAIA,EAAEZ,CAAC,CAAC,CAAC,eAAe,CAAC,OAAO,KAAK,UAAU,CAAC,cAAcY,EAAE,CAAC,GAAGA,EAAE,EAAE,MAAM,IAAI,MAAM,4DAA4DA,IAAI,EAAE,GAAG,KAAK,WAAWA,EAAE,QAAQZ,EAAE,EAAEA,EAAE,KAAK,WAAWY,EAAEZ,IAAI,CAAC,IAAI,EAAE,KAAK,MAAM,KAAK,EAAE,KAAK,EAAE,MAAM,KAAK,MAAM,OAAO,CAAC,CAAC,CAAC,KAAK,WAAWY,CAAC,CAAC,EAAE,SAASslD,GAAGpmD,EAAE,EAAE,CAAC,GAAG,MAAM,QAAQA,CAAC,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAEA,IAAIJ,EAAEA,EAAE,OAAOF,CAAC,EAAE,OAAOE,CAAC,KAAK,CAAC,IAAIA,EAAE,IAAI,MAAM,CAAC,EAAE,OAAOA,EAAE,KAAKF,CAAC,EAAEE,CAAC,CAAC,CAAC,SAASmmD,GAAGrmD,EAAE,EAAE,CAAC,GAAG,CAACA,EAAE,MAAM,IAAIkmD,GAAG,CAAC,CAAC,CAAC,SAASI,GAAGtmD,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAE,QAAQI,KAAKN,EAAEM,IAAI,GAAGJ,IAAI,OAAOA,CAAC,CAAC,SAASqmD,GAAGvmD,EAAE,CAAC,OAAOA,EAAE,SAAS,EAAEA,EAAE,GAAGA,CAAC,CAAC,SAASwmD,GAAGxmD,EAAE,CAAC,OAAO,MAAM,QAAQA,CAAC,EAAEA,EAAE,CAACA,CAAC,CAAC,CAAC,SAASymD,GAAGzmD,EAAE,CAAC,IAAIE,EAAEF,EAAE,QAAQ,uBAAuB,OAAO,EAAE,QAAQ,kBAAkB,OAAO,EAAE,YAAY,EAAE,OAAOE,EAAE,KAAK,IAAIA,EAAE,UAAUA,CAAC,CAAC,SAASwmD,GAAG1mD,EAAE,CAAC,OAAOA,EAAE,QAAQ,GAAGA,EAAE,QAAQ,GAAG,IAAI,GAAGA,EAAEA,EAAE,QAAQ,cAAc,CAAC,EAAEE,IAAIA,EAAE,YAAY,CAAC,CAAC,CAAC,IAAIymD,GAAG,CAAC,EAAE,SAASC,GAAG5mD,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,KAAK,IAAI,EAAE,CAAC,EAAE,OAAO,EAAE,UAAUA,EAAE,aAAa,EAAE,EAAE,OAAOA,EAAE,UAAU,EAAE,CAAC,CAAC,SAAS6mD,GAAG7mD,EAAE,CAAC,GAAG,EAAEA,GAAG,MAAM,OAAOA,GAAG,UAAU,GAAG,MAAM,QAAQA,CAAC,EAAEA,EAAE,QAAQ,GAAG6mD,GAAG,CAAC,CAAC,MAAM,CAAC,IAAI,EAAE,OAAO,KAAK7mD,CAAC,EAAE,QAAQE,KAAK,EAAE,CAAC,IAAII,EAAEN,EAAEE,GAAGI,GAAG,MAAM,OAAOA,GAAG,WAAW,CAAC,MAAM,QAAQA,CAAC,GAAGA,EAAE,OAAO,WAAW,OAAOA,EAAE,OAAO,SAASN,EAAEE,GAAGI,EAAE,MAAMumD,GAAGvmD,CAAC,EAAE,CAAC,CAAC,CAAC,SAASwmD,GAAG9mD,EAAE,EAAE,CAAC,EAAEE,EAAE,CAAC,EAAEI,EAAE,SAASC,EAAE,GAAG,CAAC,GAAG,OAAOP,GAAG,SAAS,CAAC,IAAI,EAAEA,EAAE,EAAE,GAAG,KAAKE,EAAE,EAAEA,EAAE,WAAW,KAAKymD,GAAG,EAAEA,GAAG,WAAW,EAAE,EAAE,GAAG,GAAG,KAAK,MAAM,IAAIX,EAAE,WAAW1lD,MAAMN;AAAA,SAC7xkCM;AAAA,gBACOA,mGAAmG,EAAE,OAAO,CAAC,KAAK,CAAC,IAAI,EAAEN,EAAE,GAAG,EAAE,WAAW,MAAM,EAAE,QAAQ,KAAK,MAAM,IAAIgmD,EAAE,GAAG1lD,8BAA8B,KAAK,UAAU,CAAC;AAAA,mCAC1M,EAAE,IAAI,EAAE,EAAE,UAAU2B,EAAEC,EAAE,GAAG,KAAKhC,EAAE,CAAC+B,EAAEC,CAAC,EAAEhC,EAAE,GAAG,KAAKymD,GAAG,CAAC1kD,EAAEC,CAAC,EAAEykD,GAAG,UAAU,KAAK,IAAI,CAAC1kD,EAAEC,CAAC,EAAE,EAAE,IAAID,GAAG,KAAK,MAAM,IAAI+jD,EAAE,WAAW1lD,MAAM;AAAA,SAC5JA;AAAA,gBACOA,mGAAmG,EAAE,GAAG4B,GAAG,KAAK,CAAC,IAAIC,EAAE,CAAC,EAAE,QAAQY,KAAK,OAAO,KAAK4jD,EAAE,EAAExkD,EAAEY,GAAG4jD,GAAG5jD,GAAG,QAAQA,KAAK,OAAO,KAAK7C,CAAC,EAAEiC,EAAEY,GAAG7C,EAAE6C,GAAG,IAAIH,EAAE,EAAE,OAAOA,EAAE,cAAcT,EAAE,IAAIU,EAAE,OAAO,OAAO,CAAC,EAAE8jD,EAAE,EAAE,QAAQ5jD,KAAK,OAAO,KAAK7C,CAAC,EAAEymD,GAAG5jD,GAAG7C,EAAE6C,GAAG8jD,GAAG,EAAE,MAAM,EAAE,IAAI/jD,EAAEZ,EAAED,EAAE,EAAE,OAAO/B,EAAEK,CAAC,EAAE,OAAOomD,GAAG,OAAO,OAAO,CAAC,EAAE9jD,CAAC,EAAEC,CAAC,KAAK,CAAC,IAAIX,EAAE,OAAO,OAAO,CAAC,EAAEwkD,EAAE,EAAE,QAAQ9jD,KAAK,OAAO,KAAK3C,CAAC,EAAEymD,GAAG9jD,GAAG3C,EAAE2C,GAAG,IAAID,EAAE,IAAIX,EAAE,EAAE,MAAM,EAAE,OAAO0kD,GAAG,OAAO,OAAO,CAAC,EAAExkD,CAAC,EAAES,CAAC,CAAC,CAAC,CAAC,SAASmkD,GAAG/mD,EAAE,EAAE,CAAC,OAAOA,EAAE,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,CAAC,SAASgnD,GAAGhnD,EAAE,EAAE,CAAC,MAAM,GAAG+mD,GAAG/mD,EAAE,CAAC,CAAC,CAAC,SAASinD,GAAGjnD,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAOA,EAAE,IAAI,EAAE,CAAC,EAAE,QAAQE,KAAKF,EAAE,EAAE,QAAQE,CAAC,IAAI,IAAI,EAAE,KAAKA,CAAC,EAAE,OAAO,CAAC,CAAC,SAASgnD,GAAGlnD,EAAE,CAAC,GAAGA,GAAG,KAAK,MAAM,IAAIgmD,EAAE,yBAAyB,KAAK,UAAUhmD,CAAC,GAAG,EAAE,QAAQ,KAAKA,EAAE,GAAGA,EAAE,eAAe,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC,SAASmnD,GAAGnnD,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,MAAMF,EAAE,QAAQE,CAAC,EAAE,EAAE,MAAM,IAAI8lD,EAAE,GAAG9lD,oBAAoB,wBAAwBF,sBAAsB,CAAC,CAAC,SAASonD,GAAGpnD,EAAE,EAAEE,EAAE,EAAEI,EAAE,EAAE,EAAE,CAAC,OAAO+lD,GAAGnmD,GAAG,CAAC,EAAEmmD,GAAG/lD,GAAGJ,CAAC,EAAE,MAAM,QAAQF,CAAC,GAAGA,EAAE,QAAQE,GAAGF,EAAE,QAAQM,GAAGN,EAAE,MAAMO,GAAG,OAAOA,IAAI,CAAC,CAAC,CAAC,SAAS8mD,GAAGrnD,EAAE,EAAE,CAAC,MAAM,QAAQA,CAAC,GAAG8mB,EAAE,OAAO9mB,EAAE,OAAO,EAAE,IAAI,GAAG,mCAAmC,EAAEA,EAAE,QAAQ,CAACE,EAAEI,IAAI+mD,GAAGnnD,EAAE,WAAWI,EAAE,QAAQ,GAAG,CAAC,GAAGwmB,EAAE,OAAO,OAAO,UAAU9mB,CAAC,GAAGA,EAAE,EAAE,IAAI,YAAY,uCAAuCsnD,GAAGtnD,CAAC,IAAI,CAAC,CAAC,SAASsnD,GAAGtnD,EAAE,CAAC,OAAOA,IAAI,KAAK,OAAO,MAAM,QAAQA,CAAC,EAAE,IAAIA,EAAE,IAAI,GAAGsnD,GAAG,CAAC,CAAC,EAAE,KAAK,GAAG,EAAE,IAAI,OAAOtnD,GAAG,SAAS,IAAIA,KAAK,GAAGA,GAAG,CAAC,SAASunD,GAAGvnD,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,GAAG,KAAKA,EAAE,EAAE4mB,EAAE,IAAI,EAAEvmB,EAAE,MAAM,IAAIyB,IAAI,CAAC,IAAIC,EAAE/B,GAAG,KAAKA,EAAE,EAAE4mB,EAAE,IAAI,EAAE,OAAO7kB,EAAE3B,EAAE,IAAIA,EAAE2B,EAAE1B,EAAEP,EAAE,GAAGgC,CAAC,GAAGzB,CAAC,CAAC,CAAC,SAASinD,GAAGxnD,EAAE,CAAC,OAAOA,IAAI,OAAO,OAAOA,IAAI,SAAS,SAASA,IAAI,MAAM,MAAM,IAAI,CAAC,IAAIynD,GAAG,EAAE,SAASC,IAAI,CAAC,OAAOD,IAAI,CAAC,IAAIE,GAAG,CAAC,EAAE,SAASC,GAAG5nD,EAAE,GAAG,CAAC,OAAOA,KAAK2nD,KAAKA,GAAG3nD,GAAG,GAAG2nD,GAAG3nD,IAAI,EAAEA,EAAE2nD,GAAG3nD,GAAG,SAAS,CAAC,CAAC,IAAI6nD,GAAG,CAAC,gBAAgB,cAAc,EAAEC,GAAG,CAAC,UAAU,UAAU,EAAEC,GAAG,CAAC,QAAQ,OAAO,QAAQ,EAAEC,GAAG,CAAC,MAAM,KAAK,EAAEC,GAAG,CAAC,MAAM,MAAM,SAAS,KAAK,EAAMC,GAAG,IAAI,IAAI,SAASC,GAAGnoD,EAAE,CAACmnD,GAAGU,GAAG,aAAa7nD,CAAC,CAAC,CAAC,SAASooD,GAAGpoD,EAAE,CAACmnD,GAAGW,GAAG,sBAAsB9nD,CAAC,CAAC,CAAC,SAASqoD,GAAGroD,EAAE,CAACmnD,GAAGY,GAAG,cAAc/nD,CAAC,CAAC,CAAC,SAASsoD,GAAGtoD,EAAE,CAACmnD,GAAGa,GAAG,WAAWhoD,CAAC,CAAC,CAAC,IAAIuoD,GAAG,CAAC,EAAEC,GAAG,IAAI,SAASC,GAAGzoD,EAAE,EAAE,CAACuoD,GAAG,KAAKvoD,CAAC,EAAE,GAAG,CAAC,IAAIE,EAAE,EAAE,EAAE,OAAOqoD,GAAG,IAAI,EAAEroD,CAAC,OAAOA,EAAN,CAAS,MAAMqoD,GAAG,IAAI,EAAEroD,CAAC,CAAC,CAAC,SAASwoD,IAAI,CAAC,OAAOH,GAAG,SAAS,EAAE,GAAGA,GAAG,KAAKC,EAAE,EAAEA,EAAE,CAAC,SAASG,GAAG3oD,EAAE,CAAC,GAAG,CAAC4oD,GAAG5oD,CAAC,EAAE,MAAM,IAAI,MAAM,6BAA6BA,EAAE,GAAG,EAAE,OAAO0oD,GAAG,EAAE1oD,CAAC,CAAC,SAAS6oD,GAAG7oD,EAAE,CAAC,GAAG,CAAC4oD,GAAG5oD,CAAC,EAAE,MAAM,IAAI,MAAM,6BAA6BA,EAAE,GAAG,EAAEkoD,GAAG,IAAIloD,CAAC,GAAGkoD,GAAG,IAAIloD,EAAE,CAAC,EAAE,IAAI,EAAEkoD,GAAG,IAAIloD,CAAC,EAAE,GAAGkoD,GAAG,IAAIloD,EAAEkoD,GAAG,IAAIloD,CAAC,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,IAAIE,EAAE,GAAGF,KAAK,IAAI,OAAOkoD,GAAG,IAAIhoD,EAAE,CAAC,EAAEA,CAAC,KAAM,QAAOF,CAAC,CAAC,IAAI8oD,GAAG,IAAI,OAAO,iCAAiC,EAAE,SAASF,GAAG5oD,EAAE,CAAC,MAAM,CAAC,CAACA,EAAE,MAAM8oD,EAAE,CAAC,CAAC,SAASC,GAAG/oD,EAAE,CAAC,OAAOA,IAAI,SAASA,EAAE,SAAS,EAAE,EAAE,CAAC,CAAC,SAASgpD,GAAGhpD,EAAE,EAAEE,EAAE,CAAC,GAAG,OAAO,EAAE,GAAGA,GAAG,OAAOA,EAAEF,EAAE,QAAQ,IAAIM,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAEL,EAAE,EAAEK,EAAED,GAAGN,EAAEO,GAAG,OAAOD,CAAC,CAAC,SAAS2oD,GAAGjpD,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,OAAO,IAAI,IAAI,EAAE,OAAO,kBAAkB,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,CAAC,IAAII,EAAEN,EAAEE,GAAGI,EAAE,IAAI,EAAEA,EAAE,CAAC,OAAO,CAAC,CAAC,SAAS4oD,GAAGlpD,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,OAAO,IAAI,IAAI,EAAE,OAAO,kBAAkB,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,CAAC,IAAII,EAAEN,EAAEE,GAAGI,EAAE,IAAI,EAAEA,EAAE,CAAC,OAAO,CAAC,CAAC,SAAS6oD,GAAGnpD,EAAE,EAAE,CAAC,GAAG,EAAEA,EAAE,MAAM,IAAIgmD,EAAE,QAAQ,eAAehmD,kBAAkB,EAAE,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAEN,EAAEM,EAAE,EAAE,EAAEA,EAAEJ,EAAE,KAAKI,CAAC,EAAE,OAAOJ,CAAC,CAAC,IAAIkpD,GAAG,SAASC,IAAI,CAAC,OAAOD,IAAI,OAAOA,GAAGh1B,GAAG,EAAE,QAAQ,GAAGg1B,EAAE,CAAC,SAASE,IAAI,CAAC,MAAM,cAAc,CAAC,SAASC,GAAGvpD,EAAE,EAAE,CAAC,OAAOywB,GAAEzwB,EAAE,CAAC,CAAC,CAAC,SAASwpD,GAAGxpD,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAEF,EAAE,MAAM,MAAM,EAAE,OAAO,EAAE,IAAI,EAAEE,EAAE,OAAO,EAAE,GAAGA,EAAE,OAAO,EAAE,EAAE,CAAC,EAAEg8B,EAAEl8B,EAAEE,CAAC,CAAC,CAAC,SAASupD,GAAGzpD,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,GAAGzzB,EAAE,MAAM,SAAS,EAAE,MAAM,IAAIgmD,EAAE,yDAAyDhmD,EAAE,MAAM,gBAAgB,EAAE,IAAIE,EAAEspD,GAAGxpD,EAAE,CAAC,EAAE,OAAO0pD,GAAGxpD,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAASypD,GAAG3pD,EAAE,CAAC,IAAI,EAAE,CAACgpD,GAAGhpD,EAAE,KAAK,CAAC,EAAE,OAAOk8B,EAAEl8B,EAAE,CAAC,CAAC,CAAC,SAAS4pD,GAAG5pD,EAAE,CAAC,GAAGA,EAAE,MAAM,EAAE,MAAM,IAAIgmD,EAAE,wDAAwDhmD,EAAE,OAAO,EAAE,IAAI,EAAE,CAACA,EAAE,MAAM,GAAGgpD,GAAGhpD,EAAE,MAAM,CAAC,CAAC,EAAE,OAAOk8B,EAAEl8B,EAAE,CAAC,CAAC,CAAC,SAAS6pD,GAAG7pD,EAAE,EAAEE,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,OAAOzzB,EAAE,KAAK,CAAC,IAAK,GAAE,OAAOouC,GAAGpuC,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOouC,GAAGtuC,EAAE,CAAC,EAAE,CAAC,EAAE,CAACE,EAAEF,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAOwuC,GAAGxuC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,CAACE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO0uC,GAAG1uC,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAACE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO48B,GAAG58B,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,CAACE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO48B,GAAG58B,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,CAACE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,QAAQ,MAAM,IAAIgmD,EAAE,8DAA8DhmD,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS8pD,GAAG9pD,EAAE,EAAEE,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,OAAOzzB,EAAE,KAAK,CAAC,IAAK,GAAE,OAAOouC,GAAGpuC,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOouC,GAAGtuC,EAAE,CAAC,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGE,CAAC,CAAC,EAAE,IAAK,GAAE,OAAOsuC,GAAGxuC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGE,CAAC,CAAC,EAAE,IAAK,GAAE,OAAOwuC,GAAG1uC,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGE,CAAC,CAAC,EAAE,QAAQ,MAAM,IAAI8lD,EAAE,6DAA6DhmD,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS+pD,GAAG/pD,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAOmzB,EAAE,IAAI,CAAC,OAAOzzB,EAAE,KAAK,CAAC,IAAK,GAAE,OAAOouC,GAAGpuC,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOI,EAAE,CAAC,IAAK,GAAE,OAAOupD,GAAG7pD,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAO4pD,GAAG9pD,EAAE,EAAEE,CAAC,EAAE,QAAQ,MAAM,IAAI8lD,EAAE,iDAAiD1lD,GAAG,CAAC,CAAC,IAAK,GAAE,OAAOA,EAAE,CAAC,IAAK,GAAE,OAAOupD,GAAG7pD,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOsuC,GAAGxuC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGE,EAAEF,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO8pD,GAAG9pD,EAAE,EAAEE,CAAC,EAAE,QAAQ,MAAM,IAAI8lD,EAAE,iDAAiD1lD,GAAG,CAAC,CAAC,IAAK,GAAE,OAAOA,EAAE,CAAC,IAAK,GAAE,OAAOupD,GAAG7pD,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOwuC,GAAG1uC,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO0uC,GAAG1uC,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGE,EAAEF,EAAE,MAAM,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO8pD,GAAG9pD,EAAE,EAAEE,CAAC,EAAE,QAAQ,MAAM,IAAI8lD,EAAE,iDAAiD1lD,GAAG,CAAC,CAAC,QAAQ,MAAM,IAAI0lD,EAAE,6DAA6DhmD,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,SAASgqD,GAAGhqD,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAE,OAAO,EAAE,IAAIA,EAAEF,EAAE,GAAG,KAAKE,IAAI,EAAE,EAAEA,EAAE,EAAE,GAAG,IAAIF,EAAE,GAAG,OAAO,EAAE,IAAIw8B,GAAGx8B,EAAE,CAAC,CAAC,CAAC,SAASiqD,GAAGjqD,EAAE,EAAE,CAAC,OAAOA,EAAE,KAAK,CAAC,IAAK,GAAE,OAAOw+B,GAAG,CAACx+B,EAAE,CAAC,CAAC,EAAE,IAAK,GAAE,OAAO0+B,GAAG,CAAC1+B,EAAE,CAAC,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO4+B,GAAG,CAAC5+B,EAAE,CAAC,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO8+B,GAAG,CAAC9+B,EAAE,CAAC,EAAE,CAAC,EAAE,QAAQ,MAAM,IAAIgmD,EAAE,+DAA+DhmD,EAAE,MAAM,CAAC,CAAC,CAAC,SAAS0pD,GAAG1pD,EAAE,EAAE,CAAC,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,GAAGA,EAAE,OAAO,EAAE,OAAO,MAAM,IAAIgmD,EAAE,0BAA0B,EAAE,+DAA+DhmD,EAAE,OAAO,EAAE,OAAOgkC,GAAGhkC,EAAE,CAAC,CAAC,CAAC,SAASkqD,GAAGlqD,EAAE,EAAE,EAAEE,EAAE,EAAEI,EAAEC,EAAE,CAAC,OAAOsrC,GAAG7rC,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,CAAC,SAAS4pD,GAAGnqD,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGN,EAAE,KAAK,GAAG,EAAE,KAAK,EAAE,MAAM,IAAIimD,GAAG,8DAA8DjmD,EAAE,uBAAuB,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,EAAE,CAAC,IAAIO,EAAEP,EAAE,MAAM,MAAM,EAAE,EAAE,GAAG,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,GAAG,GAAGO,IAAI,EAAE,MAAM,IAAI0lD,GAAG,gGAAgGjmD,EAAE,wBAAwB,EAAE,OAAO,CAAC,CAAC,GAAGA,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,OAAOyyC,GAAG,OAAO,CAAC,EAAEzyC,EAAE,EAAE,EAAE,WAAW,GAAG,WAAW,GAAG,KAAKM,EAAE8pD,GAAGpqD,EAAE,KAAKM,EAAEgpD,GAAG,CAAC,EAAE,KAAK,WAAWppD,CAAC,CAAC,EAAE,CAAC,IAAIK,EAAEP,EAAE,MAAM,MAAM,EAAE,EAAEO,EAAE,IAAI,EAAEP,EAAEk8B,EAAEl8B,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,MAAM,MAAM,EAAEiC,EAAE,EAAE,IAAI,EAAEC,EAAE,EAAE,IAAI,EAAEC,EAAE,CAAC,GAAG,EAAEF,CAAC,EAAEW,EAAE,MAAM,KAAK,CAAC,OAAO,EAAE,IAAI,EAAE,CAACI,EAAEkB,IAAIA,IAAI,EAAE,EAAE,KAAK,EAAEA,GAAG,EAAE,KAAK,EAAEA,EAAE,EAAEA,CAAC,EAAE,EAAEg4B,EAAErH,GAAG,EAAEjyB,CAAC,EAAE,CAACV,EAAE,EAAE,CAAC,EAAE,IAAIW,EAAE,CAAC,GAAGtC,EAAE,GAAG4B,CAAC,EAAEW,EAAE,GAAGC,EAAE,GAAG,OAAOm5B,EAAEuW,GAAG,OAAO,CAAC,EAAEzyC,EAAE,EAAE,EAAE,WAAW8C,EAAE,WAAWC,EAAE,KAAKzC,EAAE8pD,GAAGpqD,EAAE,KAAKM,EAAEgpD,GAAG,CAAC,EAAE,KAAK,WAAWppD,CAAC,CAAC,EAAE2C,CAAC,CAAC,CAAC,CAAC,SAASwnD,GAAGrqD,EAAE,EAAEE,EAAE,CAAC,OAAOuzB,EAAE,KAAK,MAAM,QAAQ,CAAC,EAAE,EAAE0c,GAAG,EAAE,OAAO,EAAE,EAAE1f,GAAE,EAAE,OAAO,EAAE6T,GAAGtkC,EAAE,EAAEE,CAAC,EAAE,CAAC,CAAC,SAASoqD,GAAGtqD,EAAE,CAAC,OAAOo5B,EAAEp5B,EAAEA,CAAC,CAAC,CAAC,SAASoqD,GAAGpqD,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAE,MAAM,GAAG,EAAE,OAAO,GAAG,EAAE,OAAON,EAAE,MAAM,IAAIgmD,EAAE,+BAA+B,EAAE,gCAAgChmD,GAAG,EAAE,GAAGA,IAAI,EAAE,CAAC,GAAGE,IAAI,gBAAgB,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,CAAC,EAAE,GAAGJ,IAAI,eAAe,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE57B,EAAE,EAAE,CAAC,EAAE47B,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO57B,CAAC,CAAC,CAAC,SAASN,IAAI,EAAE,CAAC,GAAGE,IAAI,gBAAgB,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAG,EAAE,CAAC,CAAC,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,CAAC,EAAE,GAAGJ,IAAI,eAAe,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE57B,EAAE,EAAE,CAAC,EAAE47B,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO57B,CAAC,CAAC,CAAC,SAASN,IAAI,EAAE,CAAC,GAAGE,IAAI,gBAAgB,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAG,CAAC,CAAC,EAAE47B,EAAE,EAAE,CAAC,EAAE57B,EAAE,GAAGA,EAAE,EAAE,CAAC,EAAE,GAAGJ,IAAI,eAAe,OAAOI,EAAE,SAAS,EAAE47B,EAAE,EAAE,CAAC,EAAE,EAAE57B,EAAE,EAAE,CAAC,EAAE47B,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO57B,CAAC,CAAC,CAAC,SAASN,EAAE,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,sCAAsC,EAAE,MAAM,CAAC,CAAC,SAASuE,GAAGvqD,EAAE,EAAEE,EAAE,CAAC,OAAOuzB,EAAE,KAAKvzB,GAAG,OAAOA,EAAEopD,GAAG,GAAGnB,GAAGjoD,CAAC,EAAE44B,EAAE94B,EAAEoqD,GAAGpqD,EAAE,KAAK,EAAEE,CAAC,CAAC,EAAE,CAAC,CAAC,SAASsqD,GAAGxqD,EAAE,EAAE,EAAE,CAAC,GAAG,IAAI,EAAE,MAAM,IAAIimD,GAAG,0CAA0C,4BAA4B,EAAE,OAAOrkB,GAAG5hC,CAAC,CAAC,CAAC,SAASyqD,GAAGzqD,EAAE,CAAC,OAAOyzB,EAAE,IAAIyF,GAAGl5B,EAAE84B,EAAEQ,GAAGt5B,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS0qD,GAAG1qD,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAOmzB,EAAE,IAAI2e,GAAGpyC,EAAE,EAAEE,EAAEI,CAAC,CAAC,CAAC,CAAC,SAASqqD,GAAG3qD,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAI,EAAEqF,EAAE,GAAGM,EAAE,GAAGp5B,CAAC,CAAC,EAAE,OAAOs+B,GAAG,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,SAASssB,GAAG5qD,EAAE,EAAEE,EAAE,GAAG,CAAC,OAAOA,EAAEF,EAAE,EAAE,EAAE,CAAC,CAAC,IAAI6qD,GAAG,CAAC,QAAQ,SAAS,QAAQ,EAAEC,GAAG,CAAC,SAAS,UAAU,iBAAiB,EAAE,SAASC,GAAG/qD,EAAE,CAACmnD,GAAG0D,GAAG,UAAU7qD,CAAC,CAAC,CAAC,SAASgrD,GAAGhrD,EAAE,CAACmnD,GAAG2D,GAAG,eAAe9qD,CAAC,CAAC,CAAC,IAAIirD,GAAG,cAAcxzB,GAAE,YAAY,CAAC,6BAA6B,CAAC,MAAM,EAAE,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,CAAC,EAAEyzB,GAAG,cAAcD,EAAE,CAAC,MAAMnqD,EAAEZ,EAAE,CAAC,OAAOmoC,GAAGvnC,EAAEZ,CAAC,CAAC,CAAC,EAAEgrD,GAAG,UAAU,QAAQzzB,GAAE,cAAcyzB,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,MAAMnqD,EAAEZ,EAAE,CAAC,OAAOooC,GAAGxnC,EAAEZ,CAAC,CAAC,CAAC,EAAEirD,GAAG,UAAU,OAAO1zB,GAAE,cAAc0zB,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,YAAYnqD,EAAE,CAAC,GAAG,MAAM,EAAE,OAAOA,GAAG,SAAS,MAAM,IAAIklD,EAAE,oDAAoDllD,GAAG,EAAE,GAAGA,EAAE,QAAQ,OAAO,MAAM,IAAIklD,EAAE,sCAAsCllD,GAAG,EAAE,KAAK,MAAMA,EAAE,KAAK,CAAC,MAAMA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI2F,EAAEyJ,GAAG,KAAK,KAAK,EAAEyF,GAAGxnC,EAAEZ,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,MAAM,KAAK,KAAK,CAAC,CAAC,EAAEkrD,GAAG,UAAU,WAAW3zB,GAAE,cAAc2zB,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAYnqD,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,KAAK,eAAe,IAAI,KAAK,OAAOA,EAAE,QAAQ,KAAK,eAAe,KAAK,OAAOA,EAAE,QAAQ,KAAK,eAAe,KAAK,KAAKA,EAAE,IAAI,CAAC,MAAMA,EAAEZ,EAAE,CAAC,OAAO+rC,GAAGnrC,EAAE,KAAK,OAAO,KAAK,OAAOZ,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,OAAO,KAAK,OAAO,OAAO,KAAK,OAAO,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEmrD,GAAG,UAAU,gBAAgB5zB,GAAE,cAAc4zB,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,YAAYnqD,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,EAAE,KAAK,eAAe,IAAI,KAAK,KAAKA,EAAE,MAAM,KAAK,aAAa,KAAK,OAAOA,EAAE,QAAQ,KAAK,eAAe,KAAK,KAAKA,EAAE,IAAI,CAAC,MAAMA,EAAEZ,EAAE,CAAC,GAAGA,EAAEA,GAAG,UAAUA,IAAI,WAAWA,IAAI,QAAQ,MAAM,IAAI+lD,GAAG,uCAAuC/lD,IAAI,EAAE,OAAOgqD,GAAGppD,EAAE,KAAK,KAAK,KAAK,OAAOZ,EAAE,KAAK,IAAI,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,KAAK,KAAK,OAAO,KAAK,OAAO,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEorD,GAAG,UAAU,eAAe7zB,GAAE,cAAc6zB,EAAE,EAAE,IAAIC,GAAG,cAAcN,EAAE,CAAC,YAAYnqD,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,EAAE,KAAK,eAAe,IAAI,KAAK,KAAKA,EAAE,MAAM,KAAK,aAAa,KAAK,OAAOA,EAAE,QAAQ,KAAK,eAAe,KAAK,KAAKA,EAAE,IAAI,CAAC,MAAMA,EAAEZ,EAAE,CAAC,GAAGA,EAAEA,GAAG,UAAUA,IAAI,WAAWA,IAAI,QAAQ,MAAM,IAAI+lD,GAAG,0CAA0C/lD,IAAI,EAAE,OAAOywC,GAAG7vC,EAAE,KAAK,KAAK,KAAK,OAAOZ,EAAE,KAAK,IAAI,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,KAAK,KAAK,OAAO,KAAK,OAAO,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEqrD,GAAG,UAAU,kBAAkB9zB,GAAE,cAAc8zB,EAAE,EAAE,IAAIC,GAAG,cAAcP,EAAE,CAAC,YAAYnqD,EAAE,CAAC,MAAM,EAAE,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK,CAAC,CAAC,MAAMA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAE,SAAS,GAAGA,EAAE,KAAKA,EAAE,GAAG,MAAM,IAAIklD,EAAE,sEAAsE,EAAE,OAAO5sB,EAAE,KAAK,KAAK8K,GAAGpjC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,KAAK,IAAI,CAAC,CAAC,EAAE0qD,GAAG,UAAU,WAAW/zB,GAAE,cAAc+zB,EAAE,EAAE,SAASC,GAAGzrD,EAAE,EAAE,eAAe,CAAC,IAAIE,EAAEI,EAAE,GAAG6nD,GAAG,CAAC,EAAEnoD,EAAE,SAAS,EAAEE,EAAEF,EAAE,GAAGM,EAAEN,EAAE,WAAW,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQA,EAAE,MAAM,IAAI,IAAI,GAAG,IAAI,gBAAgB,CAAC,IAAIO,EAAEyoD,GAAGhpD,EAAE,CAAC,EAAEE,EAAEF,EAAE,GAAGO,EAAED,EAAEN,EAAE,GAAGO,CAAC,SAAS,IAAI,eAAe,CAAC,IAAIA,EAAEyoD,GAAGhpD,EAAE,EAAEA,EAAE,OAAO,CAAC,EAAEE,EAAEF,EAAEA,EAAE,OAAO,GAAGO,EAAED,EAAEN,EAAEA,EAAE,OAAO,GAAGO,CAAC,MAAM,CAAC,IAAIA,EAAEyoD,GAAGhpD,CAAC,EAAEE,EAAE,KAAK,KAAKK,CAAC,EAAED,EAAE,KAAK,KAAKC,CAAC,CAAC,CAAC,MAAM,CAACL,EAAEI,CAAC,CAAC,CAAC,IAAIorD,GAAG,cAAcT,EAAE,CAAC,YAAYnqD,EAAE,CAAC,GAAG,MAAM,EAAEA,EAAE,MAAM,EAAE,MAAM,IAAIklD,EAAE,wCAAwCllD,EAAE,OAAO,EAAE,KAAK,MAAMA,EAAE,OAAO,KAAK,EAAEA,EAAE,MAAM,KAAK,KAAKA,EAAE,MAAM,KAAK,QAAQA,EAAE,KAAKiqD,GAAG,KAAK,IAAI,EAAE,KAAK,aAAajqD,EAAE,cAAc,KAAK,SAASA,EAAE,aAAakqD,GAAG,KAAK,YAAY,EAAE,KAAK,KAAKlqD,EAAE,IAAI,CAAC,MAAMA,EAAEZ,EAAE,CAAC,IAAI,EAAEurD,GAAG3qD,CAAC,EAAEP,EAAE,EAAE,GAAGwB,EAAE,EAAE,GAAGC,EAAE,KAAK,MAAM,GAAG,KAAK,OAAO,QAAQA,GAAG,KAAK,IAAI,EAAEzB,CAAC,EAAE,KAAK,OAAO,SAASyB,GAAG,KAAK,IAAI,EAAED,CAAC,EAAEC,GAAG,KAAK,IAAI,GAAGzB,EAAEwB,GAAG,CAAC,EAAE,KAAK,eAAe,SAAS,CAAC,IAAIE,EAAE,KAAK,KAAKD,CAAC,EAAE,GAAG9B,EAAEA,GAAG,UAAUA,IAAI,WAAWA,IAAI,QAAQ,MAAM,IAAI+lD,GAAG,GAAG,KAAK,aAAa,4BAA4B/lD,IAAI,EAAE,OAAOywC,GAAG7vC,EAAE,EAAEmB,EAAE/B,EAAE,KAAK,IAAI,CAAC,KAAK,CAAC,IAAI+B,EAAE,KAAK,KAAK,EAAED,CAAC,EAAE,OAAOiqC,GAAGnrC,EAAE,CAACmB,EAAEA,EAAE/B,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,MAAM,KAAK,MAAM,KAAK,KAAK,KAAK,aAAa,KAAK,aAAa,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEwrD,GAAG,UAAU,kBAAkBj0B,GAAE,cAAci0B,EAAE,EAAE,IAAIC,GAAG,cAAcD,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,SAAS,aAAa,UAAU,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEC,GAAG,UAAU,gBAAgBl0B,GAAE,cAAck0B,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,SAAS,aAAa,SAAS,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEE,GAAG,UAAU,eAAen0B,GAAE,cAAcm0B,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,QAAQ,aAAa,SAAS,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEG,GAAG,UAAU,WAAWp0B,GAAE,cAAco0B,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,QAAQ,aAAa,UAAU,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEI,GAAG,UAAU,YAAYr0B,GAAE,cAAcq0B,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,QAAQ,aAAa,SAAS,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEK,GAAG,UAAU,cAAct0B,GAAE,cAAcs0B,EAAE,EAAE,IAAIC,GAAG,cAAcN,EAAE,CAAC,YAAY5qD,EAAE,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,QAAQ,aAAa,UAAU,KAAKA,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,CAAC,CAAC,cAAc,CAAC,OAAO4qD,GAAG,SAAS,CAAC,EAAEM,GAAG,UAAU,cAAcv0B,GAAE,cAAcu0B,EAAE,EAAE,IAAIC,GAAG,cAAchB,EAAE,CAAC,YAAYnqD,EAAE,CAAC,GAAG,MAAM,EAAE,KAAK,aAAa,EAAE,KAAK,KAAKA,EAAE,MAAM,KAAK,KAAK,aAAaA,EAAE,KAAK,KAAK,KAAKA,EAAE,KAAK,KAAK,MAAM,KAAK,MAAM,IAAImlD,GAAG,gEAAgE,CAAC,CAAC,MAAMnlD,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAE,OAAO,EAAE,MAAM,IAAImlD,GAAG,4BAA4B,EAAEnlD,EAAE,GAAGA,EAAE,GAAG,KAAK,QAAQ,KAAK,2EAA2EA,EAAE,GAAGA,EAAE,oCAAoC,EAAE,IAAI,EAAEA,EAAE,GAAGA,EAAE,GAAG,CAACA,EAAE,GAAGA,EAAE,EAAE,EAAEA,EAAEP,EAAE2pD,GAAG,EAAE,EAAE,EAAE,SAAS,EAAEnoD,EAAEy3C,GAAG,YAAYj5C,CAAC,EAAE,OAAOO,EAAE,GAAGA,EAAE,KAAKiB,EAAE8yB,GAAG9yB,CAAC,GAAGq3B,EAAE,KAAK,KAAKr3B,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEkqD,GAAG,UAAU,aAAax0B,GAAE,cAAcw0B,EAAE,EAAE,IAAIC,GAAG,CAAC,SAAS,WAAW,aAAa,eAAe,cAAc,gBAAgB,SAAS,WAAW,UAAU,YAAY,SAAS,WAAW,YAAY,cAAc,aAAa,eAAe,KAAK,OAAO,WAAW,aAAa,aAAa,eAAe,cAAc,gBAAgB,gBAAgB,kBAAkB,gBAAgB,kBAAkB,MAAM,OAAO,EAAE,SAASC,GAAGnsD,EAAE,EAAE,CAAC,EAAE,CAAC,OAAO8mD,GAAG9mD,EAAEy3B,GAAE,iBAAiB,OAAO,EAAE,aAAa,EAAE,aAAa,CAAC,CAAC,SAAS20B,GAAGpsD,EAAE,CAAC,OAAO4mD,GAAG5mD,CAAC,CAAC,CAAC,SAASqsD,GAAGrsD,EAAE,CAAC,GAAG,OAAOA,GAAG,SAAS,CAAC,IAAI,EAAEA,KAAKksD,GAAGA,GAAGlsD,GAAGA,EAAE,GAAG,IAAI,eAAe,OAAO,IAAI4rD,GAAG,GAAG,IAAI,gBAAgB,OAAO,IAAID,GAAG,GAAG,IAAI,WAAW,OAAO,IAAIE,GAAG,GAAG,IAAI,YAAY,OAAO,IAAIC,GAAG,GAAG,IAAI,cAAc,OAAO,IAAIC,GAAG,GAAG,IAAI,eAAe,OAAO,IAAIC,GAAG,CAAC,IAAI9rD,EAAE,CAAC,EAAE,OAAOA,EAAE,UAAU,EAAEA,EAAE,OAAO,CAAC,EAAEisD,GAAGjsD,CAAC,CAAC,CAAC,KAAM,QAAOF,aAAairD,GAAGjrD,EAAEmsD,GAAGnsD,CAAC,CAAC,CAAC,SAASssD,GAAGtsD,EAAE,CAAC,OAAO,MAAM,QAAQA,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,CAAC,CAAC,SAASusD,GAAGvsD,EAAE,CAAC,OAAOA,EAAE,SAAS,EAAE,CAAC,EAAE,MAAM,QAAQA,EAAE,EAAE,EAAEA,EAAE,CAACA,CAAC,CAAC,CAAC,SAASwsD,GAAGxsD,EAAE,CAAC,IAAI,EAAE,GAAG,MAAM,QAAQA,CAAC,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAIgmD,EAAE,uCAAuChmD,EAAE,QAAQ,EAAE,EAAEA,EAAE,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,CAAC,SAASysD,GAAGzsD,EAAE,CAAC,GAAG,MAAM,QAAQA,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAOA,EAAEA,EAAEA,EAAE,GAAG,MAAM,IAAIgmD,EAAE,iCAAiChmD,EAAE,QAAQ,CAAC,KAAM,QAAOA,CAAC,CAAC,SAAS0sD,GAAG1sD,EAAE,CAAC,IAAI,EAAE,EAAE,QAAQE,KAAKF,EAAEE,EAAE,MAAM,SAAS,EAAE,GAAG,EAAE,GAAGA,EAAE,MAAM,OAAO,CAACI,EAAEC,IAAID,EAAEC,CAAC,EAAE,OAAO,CAAC,CAAC,IAAIosD,GAAG,WAAWC,GAAG,KAAK,CAAC,YAAY9rD,EAAEZ,EAAE,UAAU,EAAEysD,GAAGpsD,EAAE,GAAGwB,EAAE,KAAK,CAAC,KAAK,MAAM7B,GAAG,KAAK,UAAUA,EAAE,KAAK,MAAMY,EAAE,MAAM,KAAK,GAAG4mD,GAAG,EAAE,EAAE,GAAG,KAAKiF,GAAG,EAAE,KAAK,aAAahE,GAAG,CAAC,EAAE,KAAK,KAAKE,GAAG,KAAK,YAAY,EAAE,KAAK,WAAWtoD,EAAE,KAAK,WAAWwB,EAAE,KAAK,IAAIovC,GAAGrwC,EAAE,KAAK,WAAW,KAAK,KAAK,KAAK,KAAK,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,kBAAkB,EAAE,KAAK,GAAG,CAAC,MAAMA,EAAE,CAAC,OAAO,KAAK,kBAAkB,EAAE+rD,GAAG,KAAK,IAAI/rD,CAAC,EAAE,KAAK,IAAI,KAAKA,EAAE,KAAK,KAAK,IAAI,OAAOA,CAAC,EAAE,KAAK,YAAY,MAAM,KAAK,IAAI,OAAO,KAAK,WAAW,MAAM,KAAK,GAAG,CAAC,GAAG,IAAI,CAAC,SAAS,CAAC,KAAK,kBAAkB,EAAE,KAAK,IAAI,QAAQ,CAAC,CAAC,mBAAmB,CAAC,GAAG,KAAK,IAAI,WAAW,MAAM,IAAI,MAAM,kBAAkB,KAAK,2BAA2B,CAAC,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,WAAWA,EAAE,KAAK,IAAI,UAAUA,CAAC,CAAC,EAAE,SAAS+rD,GAAG7sD,EAAE,EAAE,CAAC,GAAGA,EAAE,MAAM,SAAS,IAAI,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,mBAAmB,KAAK,UAAUA,EAAE,KAAK,EAAE,QAAQ,KAAK,UAAU,EAAE,KAAK,CAAC,CAAC,CAAC,SAAS8sD,GAAG9sD,EAAE,CAAC,OAAOA,EAAE,IAAI,GAAG,EAAE,KAAK,CAAC,CAAC,CAAC,SAAS+sD,GAAG/sD,EAAE,CAACA,EAAE,QAAQ,GAAG,CAAC,EAAE,GAAG,MAAM,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIgtD,GAAG,KAAK,CAAC,YAAYlsD,EAAE,CAAC,KAAK,MAAMA,EAAE,MAAM,KAAK,MAAMA,EAAE,MAAMA,EAAE,OAAO,KAAK,KAAK,KAAKA,EAAE,MAAM,OAAO,KAAK,KAAKA,EAAE,KAAK,KAAK,QAAQA,EAAE,QAAQ,KAAK,QAAQA,EAAE,QAAQ,KAAK,KAAKA,EAAE,MAAM,CAAC,CAAC,CAAC,EAAEmsD,GAAG,KAAK,CAAC,YAAYnsD,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAE,CAAC,KAAK,MAAMnB,EAAE,KAAK,MAAMZ,EAAE,KAAK,YAAY,EAAE,KAAK,OAAOK,EAAE,KAAK,SAASwB,EAAE,KAAK,kBAAkBE,EAAE,KAAK,GAAGylD,GAAG,EAAE1lD,GAAG,OAAO,KAAK,aAAa2mD,GAAG3mD,CAAC,EAAE,KAAK,KAAK6mD,GAAG,KAAK,YAAY,GAAG,KAAK,KAAK3oD,EAAE,MAAM,CAAC,EAAEgtD,GAAG,EAAEC,GAAG,KAAK,CAAC,YAAYrsD,EAAEZ,EAAE,CAAC,KAAK,SAASA,EAAE,KAAK,GAAGgtD,KAAK,KAAK,cAAcpsD,EAAE,cAAc,KAAK,cAAcA,EAAE,cAAc,KAAK,YAAYA,EAAE,YAAY,KAAK,cAAcA,EAAE,cAAc,KAAK,aAAaA,EAAE,aAAa,KAAK,cAAcA,EAAE,cAAc,KAAK,WAAWA,EAAE,WAAW,KAAK,YAAYA,EAAE,YAAY,KAAK,YAAYA,EAAE,YAAY,KAAK,aAAaA,EAAE,aAAa,QAAQ,KAAKA,EAAE,cAAc,GAAG,MAAM,EAAE,cAAc,KAAK,IAAI,EAAEA,EAAE,cAAc,aAAa,KAAK,IAAI,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,cAAcA,GAAG,KAAKY,EAAE,KAAKZ,EAAE,IAAI,EAAEY,EAAE,KAAK,IAAI,EAAE,MAAM,CAAC,cAAc,KAAK,cAAc,KAAK,cAAc,KAAK,KAAK,cAAcA,EAAE,YAAY,KAAK,YAAY,cAAc,KAAK,aAAa,CAAC,CAAC,EAAEssD,GAAG,EAAEC,GAAG,cAAc51B,GAAE,YAAY,CAAC,YAAY32B,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,KAAK,kBAAkB,CAAC,EAAE,KAAK,UAAU,GAAG,KAAK,GAAGssD,KAAK,KAAK,oBAAoB,KAAK,KAAK,UAAU,KAAK,KAAK,gBAAgB,GAAG,KAAK,kBAAkB,CAAC,EAAE,KAAK,qBAAqB,CAAC,EAAE,KAAK,QAAQ,CAAC,EAAE,KAAK,SAAS,CAAC,EAAE,KAAK,OAAO,GAAG,KAAK,aAAa,CAAC,EAAE,KAAK,cAAc,CAAC,EAAE,IAAIltD,EAAEY,EAAE,KAAK,GAAG,CAACZ,EAAE,CAAC,IAAI,EAAE,KAAK,aAAa,EAAEA,EAAEumD,GAAG,CAAC,EAAE,IAAImB,GAAG,CAAC,CAAC,CAAC,GAAG,KAAK,KAAK1nD,EAAE,KAAK,WAAWY,EAAE,WAAW,KAAK,GAAGA,EAAE,UAAUA,EAAE,YAAY,MAAMA,EAAE,iBAAiB,KAAK,CAAC,IAAI,EAAE,GAAGA,EAAE,iBAAiB,KAAK,EAAEA,EAAE,wBAAwBA,EAAE,YAAY,KAAK,CAAC,IAAIiB,EAAE,KAAKjB,EAAE,WAAW,OAAOiB,EAAEjB,EAAE,WAAW,EAAE,CAACiB,CAAC,EAAE,OAAOjB,EAAE,UAAU,CAAC,CAAC,KAAK,gBAAgB,EAAE,IAAIP,EAAEO,EAAE,MAAMP,GAAG,OAAOA,EAAEO,EAAE,YAAYP,GAAG,OAAOA,EAAE,WAAW,KAAK,MAAMA,CAAC,CAACO,EAAE,SAAS,KAAK,KAAK,eAAeA,EAAE,QAAQ,KAAK,eAAe,KAAK,KAAK,UAAU,KAAK,KAAK,0BAA0B,EAAE,CAAC,OAAO,QAAQA,EAAEZ,EAAE,CAAC,OAAOY,EAAE,KAAK,OAAOZ,EAAE,SAAS,CAAC,CAAC,eAAeY,EAAEZ,EAAE,CAAC,GAAG,KAAK,aAAa,SAAS,EAAE,MAAM,IAAI6lD,GAAG,2DAA2D7lD,IAAI,EAAE,GAAG,KAAK,aAAa,QAAQY,EAAE,MAAM,IAAIklD,EAAE,gBAAgB9lD,aAAaY,6BAA6B,KAAK,aAAa,uBAAuB,EAAE,OAAO,KAAK,aAAaA,EAAE,CAAC,WAAWA,EAAE,CAAC,OAAOylD,GAAG,KAAK,eAAezlD,EAAE,OAAO,EAAE,YAAY,CAAC,CAAC,YAAYA,EAAE,CAAC,OAAOylD,GAAG,KAAK,eAAezlD,EAAE,QAAQ,EAAE,aAAa,CAAC,CAAC,IAAI,OAAO,CAAC,GAAG,KAAK,aAAa,OAAO,EAAE,MAAM,IAAIglD,GAAG,SAAS,KAAK,2HAA2H,EAAE,GAAG,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIA,GAAG,SAAS,KAAK,4CAA4C,EAAE,OAAOS,GAAG,KAAK,eAAe,EAAE,OAAO,EAAE,YAAY,CAAC,CAAC,IAAI,QAAQ,CAAC,GAAG,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIT,GAAG,SAAS,KAAK,4BAA4B,EAAE,GAAG,KAAK,aAAa,OAAO,EAAE,MAAM,IAAIA,GAAG,SAAS,KAAK,6HAA6H,EAAE,OAAOS,GAAG,KAAK,eAAe,EAAE,QAAQ,EAAE,aAAa,CAAC,CAAC,IAAI,QAAQ,CAAC,OAAO,KAAK,OAAO,CAAC,iBAAiB,CAAC,OAAO,KAAK,OAAO,IAAIzlD,GAAGA,EAAE,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,QAAQ,CAAC,IAAI,OAAO,CAAC,OAAO,KAAK,MAAM,CAAC,IAAI,MAAMA,EAAE,CAAC,KAAK,OAAOA,CAAC,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,kBAAkB,QAAQZ,GAAGA,EAAE,UAAUY,CAAC,EAAE,KAAK,WAAWA,CAAC,CAAC,IAAI,kBAAkB,CAAC,OAAO,KAAK,WAAW,KAAK,kBAAkB,OAAOA,GAAGA,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,IAAI,iBAAiBA,EAAE,CAAC,KAAK,kBAAkBA,CAAC,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,UAAU,KAAK,kBAAkB,OAAOA,GAAG,CAACA,EAAE,SAAS,EAAE,OAAO,KAAK,oBAAoB,EAAE,KAAK,kBAAkB,OAAO,KAAK,oBAAoB,CAAC,CAAC,IAAI,oBAAoBA,EAAE,CAAC,KAAK,qBAAqBA,CAAC,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,iBAAiB,OAAO,KAAK,mBAAmB,CAAC,CAAC,IAAI,UAAU,CAAC,OAAO,KAAK,SAAS,CAAC,aAAa,CAAC,GAAG,CAAC,KAAK,SAAS,MAAM,IAAI,MAAM,sEAAsE,CAAC,CAAC,yBAAyBA,EAAE,CAAC,GAAGA,EAAE0lD,GAAG1lD,CAAC,EAAE,KAAK,WAAW,MAAM,KAAK,UAAU,SAAS,EAAE,OAAO,IAAIZ,EAAEsmD,GAAG,KAAK,SAAS,EAAE,GAAG1lD,EAAE,SAASZ,EAAE,OAAO,MAAM,IAAI8lD,EAAE,SAAS,KAAK,gBAAgB9lD,EAAE,kCAAkCY,EAAE,yCAAyCA,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,IAAI,CAAC,IAAIP,EAAEO,EAAE,GAAGiB,EAAE7B,EAAE,GAAG,GAAG6B,GAAG,KAAK,SAAS,IAAIC,EAAEzB,EAAE,KAAK,GAAGwB,EAAE,MAAM,MAAMC,IAAID,EAAE,KAAK,MAAM,IAAIikD,EAAE,SAAS,gCAAgC,KAAK,uBAAuBjkD,EAAE,oBAAoBC,GAAG,EAAE,GAAGD,EAAE,SAAS,MAAMC,EAAED,EAAE,QAAQ,MAAM,IAAIikD,EAAE,SAAS,gCAAgC,KAAK,2BAA2BjkD,EAAE,uBAAuBC,GAAG,EAAE,GAAGD,EAAE,SAAS,MAAMC,EAAED,EAAE,QAAQ,MAAM,IAAIikD,EAAE,SAAS,gCAAgC,KAAK,2BAA2BjkD,EAAE,uBAAuBC,IAAI,EAAE,GAAGD,EAAE,OAAO,MAAMxB,EAAE,QAAQwB,EAAE,MAAM,MAAM,IAAIikD,EAAE,SAAS,gCAAgC,KAAK,yBAAyBjkD,EAAE,sBAAsBxB,EAAE,QAAQ,EAAE,GAAGwB,EAAE,KAAK,CAAC,IAAIE,EAAE1B,EAAE,MAAM,QAAQ2B,KAAKH,EAAE,KAAK,CAAC,IAAI,EAAE,OAAOG,CAAC,EAAEU,EAAEb,EAAE,KAAKG,GAAG,EAAE,GAAG,EAAED,EAAE,GAAGA,EAAEA,EAAE,OAAO,GAAG,GAAGW,GAAG,MAAM,CAACA,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,GAAG,MAAM,IAAIojD,EAAE,SAAS,gCAAgC,KAAK,uBAAuB,kCAAkCpjD,mBAAmBX,IAAI,CAAC,CAAC,CAAC,GAAGF,EAAE,OAAO,KAAK,QAAQE,EAAE,EAAEA,EAAEF,EAAE,MAAM,OAAO,EAAEE,EAAE,CAAC,IAAIC,EAAEH,EAAE,MAAME,GAAG,EAAE1B,EAAE,MAAM0B,GAAG,GAAGC,GAAG,MAAM,GAAG,MAAMA,IAAI,EAAE,MAAM,IAAI8jD,EAAE,SAAS,gCAAgC,KAAK,wBAAwBjkD,EAAE,sBAAsBxB,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOY,CAAC,CAAC,eAAeA,EAAEZ,EAAE,CAAC,KAAK,WAAW,MAAM,KAAK,UAAUY,EAAEZ,CAAC,CAAC,CAAC,YAAYY,EAAE,CAAC,KAAK,UAAUA,CAAC,CAAC,eAAe,CAAC,KAAK,UAAU,IAAI,CAAC,MAAMA,EAAEZ,EAAE,CAACA,EAAEA,GAAG,CAAC,EAAE,KAAK,kBAAkB,EAAE,IAAI,EAAEsmD,GAAG1lD,CAAC,EAAEP,EAAE,GAAG,QAAQyB,KAAK,EAAE,GAAG,EAAEA,aAAairD,IAAI,CAAC1sD,EAAE,GAAG,KAAK,CAAC,IAAIwB,EAAE,GAAG,QAAQC,KAAK,EAAE,GAAGA,aAAairD,GAAG,CAAClrD,EAAE,GAAG,KAAK,CAAC,GAAGxB,IAAIwB,EAAE,MAAM,IAAIikD,EAAE,iEAAiE,EAAE,OAAOyC,GAAG,KAAK,KAAK,IAAI,CAAC,GAAG,CAAC,KAAK,MAAM,CAAC,KAAK,yBAAyB3nD,CAAC,EAAE,IAAIkB,EAAE,CAAC,EAAE,QAAQC,KAAKukD,GAAG1lD,CAAC,EAAEkB,EAAE,KAAKC,EAAE,KAAK,EAAE,KAAK,MAAMskD,GAAGvkD,CAAC,CAAC,EAAE,KAAK,MAAM,GAAG,KAAK,gBAAgB,KAAK,WAAW,KAAK,cAAc,EAAE,KAAK,YAAY,MAAMD,IAAI,KAAK,UAAU,EAAE,CAAC,GAAG,KAAK,yBAAyBjB,CAAC,EAAEiB,EAAE,CAAC,IAAIC,EAAE,KAAK,KAAKlB,EAAEZ,CAAC,EAAE+B,EAAEukD,GAAGxkD,CAAC,EAAEE,EAAE,CAAC,EAAE,QAAQ,KAAKD,EAAE,EAAE,QAAQ,CAAC,IAAI,KAAK,EAAE,EAAE,MAAM,GAAGC,EAAE,KAAK,CAAC,EAAE,GAAGF,EAAEukD,GAAGrkD,CAAC,EAAE,KAAK,qBAAqB,KAAK,MAAM,IAAI+jD,GAAG,mFAAmF,EAAE,OAAOjkD,CAAC,KAAK,CAAC,IAAIA,EAAEsrD,GAAGxsD,CAAC,EAAEmB,EAAE,KAAK,mBAAmBD,CAAC,EAAEE,EAAE,EAAEqrD,GAAGzsD,CAAC,EAAE,GAAG,KAAK,6BAA6B,MAAM,QAAQA,CAAC,EAAEkB,EAAE,GAAGA,CAAC,EAAEC,GAAG,MAAMA,EAAE,OAAO,GAAG,MAAM,QAAQA,EAAE,EAAE,EAAEC,EAAED,EAAE,IAAI,CAACW,EAAE,IAAI,IAAIqqD,GAAG,EAAErqD,EAAE,KAAK4jD,GAAG1lD,CAAC,EAAEZ,EAAE,KAAK,KAAK,CAAC,CAAC,EAAEgC,EAAE,IAAI+qD,GAAG,EAAEhrD,EAAE,KAAKukD,GAAG1lD,CAAC,EAAEZ,EAAE,KAAK,IAAI,EAAE,KAAK,eAAeY,EAAEoB,EAAE,KAAK,KAAKF,EAAEC,EAAE/B,CAAC,EAAE,KAAK,YAAY,KAAK,qBAAqB,KAAK,MAAM,IAAI+lD,GAAG,mFAAmF,EAAE,OAAO/jD,CAAC,CAAC,CAAC,CAAC,CAAC,6BAA6BpB,EAAE,CAAC,GAAG,KAAK,iBAAiB,KAAK,GAAGA,EAAE,SAAS,KAAK,gBAAgB,OAAO,QAAQ,KAAK,iDAAiD,KAAK,UAAUA,CAAC,kDAAkD,KAAK,UAAU,KAAK,eAAe,mBAAmB,KAAK,MAAM,MAAM,CAAC,IAAIZ,EAAE,GAAG,KAAK,gBAAgB,QAAQ,CAAC,EAAEK,IAAI,CAAC,GAAG,MAAMO,EAAEP,IAAI,MAAMO,EAAEP,KAAK,IAAIL,EAAE,GAAG,CAAC,EAAEA,GAAG,QAAQ,KAAK,kCAAkC,KAAK,UAAUY,CAAC,8CAA8C,KAAK,SAAS,KAAK,UAAU,KAAK,eAAe,GAAG,CAAC,CAAC,CAAC,IAAI,aAAa,CAAC,GAAG,KAAK,cAAc,MAAM,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIglD,GAAG,aAAa,KAAK,kEAAkE,EAAE,IAAIhlD,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,aAAa,CAAC,IAAI,EAAE,KAAK,UAAUA,EAAE,YAAY,EAAEY,EAAE,QAAQ,CAAC,IAAI,IAAIA,EAAE,KAAK,CAAC,CAAC,CAAC,GAAGA,EAAE,SAAS,EAAE,CAAC,IAAIZ,EAAE,KAAK,aAAa,GAAG,aAAa,OAAO,MAAM,QAAQA,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,GAAGA,EAAE,SAAS,EAAEA,EAAE,GAAGA,CAAC,KAAM,OAAM,IAAI4lD,GAAG,aAAa,KAAK,gIAAgI,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAIC,GAAG,sCAAsC,KAAK,yFAAyF,EAAE,OAAO2G,GAAG,KAAK,OAAO,CAAC,CAAC,MAAM5rD,EAAE,CAAC,KAAK,MAAM,EAAE,CAAC,WAAWA,EAAE,GAAG,CAAC,OAAOgsD,GAAGhsD,EAAE,KAAK,iBAAiB,KAAK,OAAO,CAAC,CAAC,WAAWA,EAAE,CAAC2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAE,KAAK,QAAQ,GAAGA,EAAE,SAASY,EAAE,OAAO,MAAM,IAAIklD,EAAE,4CAA4C,KAAK,sCAAsCllD,EAAE,uCAAuCZ,EAAE,qCAAqCY,MAAM,EAAE,GAAGZ,EAAE,SAAS,EAAE,OAAO,IAAI,EAAE,CAAC,EAAEK,EAAEusD,GAAG5sD,CAAC,EAAE,QAAQ6B,EAAE,EAAEA,EAAExB,EAAE,OAAO,EAAEwB,EAAE,CAAC,IAAIC,EAAEzB,EAAEwB,GAAGE,EAAE/B,EAAE6B,GAAGG,EAAEpB,EAAEiB,GAAG,GAAG,CAAC+kB,EAAE,YAAY9kB,EAAE,MAAME,EAAE,KAAK,EAAE,MAAM,IAAI8jD,EAAE,sBAAsBhkD,EAAE,mDAAmDE,EAAE,OAAO,EAAE,EAAE,KAAK,CAACD,EAAEC,CAAC,CAAC,CAAC,CAAC6qD,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,UAAUjsD,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAEC,EAAE,CAAC,GAAG,KAAK,kBAAkB,QAAQpB,CAAC,IAAI,GAAG,MAAM,IAAIklD,EAAE,yBAAyBllD,eAAe,KAAK,MAAM,EAAE,KAAK,kBAAkB,KAAKA,CAAC,EAAE,GAAG,OAAO,EAAE,WAAW,KAAK,4BAA4BP,EAAE2B,GAAG,KAAKA,EAAE,EAAEmqD,GAAG,OAAO,GAAG,IAAI,EAAE9rD,EAAE,MAAML,EAAE,CAAC,EAAE0C,EAAE,IAAIgqD,GAAG,EAAE,EAAE9rD,EAAEkB,EAAEC,CAAC,EAAE,OAAO,EAAE,QAAQ,EAAEF,GAAG,MAAM,KAAK,QAAQ,IAAIA,EAAE,MAAMa,EAAE,KAAK,CAAC,CAAC,EAAEZ,GAAG,OAAOA,EAAE,IAAIA,EAAE,KAAK,kBAAkB,KAAKY,CAAC,EAAE,KAAK,qBAAqB,KAAKA,CAAC,EAAEA,CAAC,CAAC,6BAA6B9B,EAAE,CAAC,KAAK,0BAA0BA,CAAC,CAAC,QAAQA,EAAE,CAACA,GAAG,MAAM,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,IAAIA,EAAE0lD,GAAG1lD,CAAC,EAAE,KAAK,UAAU,QAAQ,KAAK,UAAU,MAAM,KAAK,OAAO,KAAK,GAAGA,CAAC,EAAE,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,YAAYA,EAAEZ,EAAE,CAAC,GAAG,CAAC,KAAK,gBAAgB,CAAC,GAAGA,GAAG,KAAK,GAAG,MAAM,QAAQA,CAAC,EAAEA,EAAE,QAAQ,GAAG,CAAC,GAAG,GAAG,KAAK,MAAM,IAAI,UAAU,SAAS,KAAK,6DAA6D,CAAC,CAAC,MAAO,OAAM,IAAI,UAAU,SAAS,KAAK,6DAA6D,EAAE,OAAO,IAAI,CAAC,OAAOA,CAAC,CAAC,eAAeY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAE,KAAK,CAAC,IAAIC,EAAEskD,GAAG1lD,CAAC,EAAEZ,EAAEsmD,GAAGtmD,CAAC,EAAE,EAAEsmD,GAAG,CAAC,EAAEjmD,EAAEimD,GAAGjmD,CAAC,EAAEwB,EAAEwqD,GAAGxqD,CAAC,EAAEC,EAAEuqD,GAAGvqD,CAAC,EAAE,IAAI,EAAE,CAAC,EAAEY,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQE,KAAKZ,EAAE,EAAE,KAAKY,EAAE,WAAW,EAAEF,EAAE,KAAKE,EAAE,SAAS,EAAE,EAAE,KAAKA,EAAE,WAAW,EAAE,IAAIqqD,GAAG,CAAC,cAAc,KAAK,cAAc,EAAE,YAAYvqD,EAAE,cAAc,EAAE,aAAaV,EAAE,cAAchC,EAAE,WAAW,EAAE,YAAYK,EAAE,YAAYwB,EAAE,aAAaC,CAAC,EAAEC,CAAC,EAAE,QAAQa,EAAE,EAAEA,EAAE5C,EAAE,OAAO4C,IAAI5C,EAAE4C,GAAG,YAAY,KAAK5C,EAAE4C,GAAG,UAAU,KAAK,aAAa,OAAO,EAAE5C,EAAE4C,GAAG,YAAYA,CAAC,CAAC,WAAW,CAAC,IAAIhC,EAAE,CAAC,KAAK,KAAK,KAAK,UAAU,KAAK,SAAS,EAAE,OAAO,KAAK,iBAAiB,OAAOA,EAAE,gBAAgB,KAAK,iBAAiB,KAAK,OAAO,OAAOA,EAAE,MAAM,KAAK,OAAOA,CAAC,CAAC,gBAAgB,CAAC,OAAO,KAAK,QAAQ,QAAQA,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,QAAQ,MAAM,CAAC,mBAAmB,CAAC,GAAG,KAAK,YAAY,EAAE,MAAM,IAAI,MAAM,UAAU,KAAK,4BAA4B,CAAC,CAAC,SAAS,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI,MAAM,wBAAwB,KAAK,yCAAyC,EAAE,GAAG,KAAK,YAAY,KAAK,MAAM,IAAI,MAAM,wBAAwB,KAAK,wCAAwC,EAAE,KAAK,kBAAkB,EAAE,IAAIA,EAAE,EAAE,MAAM,EAAE,KAAK,YAAY,IAAIA,EAAE,KAAK,eAAe,GAAG,CAAC,qBAAqB,KAAK,UAAU,qBAAqBA,CAAC,CAAC,CAAC,EAAE,SAASwsD,GAAGttD,EAAE,CAACA,EAAEwmD,GAAGxmD,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,QAAQE,KAAKF,EAAE,EAAE,KAAKE,EAAE,KAAK,EAAE,OAAOqmD,GAAG,CAAC,CAAC,CAAC,SAASgH,GAAGvtD,EAAE,CAAC,MAAM,SAAS,CAAC,SAASwtD,GAAGxtD,EAAE,EAAEE,EAAE,CAAC,IAAI,GAAG,MAAMA,GAAG,MAAMA,EAAE,KAAK,EAAEF,EAAE,YAAYE,EAAEF,EAAE,WAAW,EAAE,aAAa,SAAS,EAAE,MAAM,CAACA,CAAC,EAAE,CAAC,IAAIM,EAAE,EAAE,aAAaJ,GAAG,GAAGI,EAAE,cAAc,SAAS,EAAE,OAAOA,EAAE,aAAa,CAAC,IAAIC,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAED,EAAE,cAAc,OAAO,IAAI,CAAC,IAAI,EAAEA,EAAE,aAAa,GAAG2B,EAAE3B,EAAE,cAAc,GAAG4B,EAAE5B,EAAE,YAAY,GAAG6B,EAAEqrD,GAAG,EAAEvrD,EAAEC,CAAC,EAAE,QAAQU,KAAKT,EAAE5B,EAAE,QAAQqC,CAAC,IAAI,IAAIrC,EAAE,KAAKqC,CAAC,CAAC,CAAC,OAAOrC,CAAC,CAAC,CAAC,CAAC,IAAIktD,GAAG,cAAcJ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAM,CAAC,MAAMA,EAAE,MAAM,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK8mD,GAAG,OAAO,EAAE,SAAS,CAAC,CAAC,EAAE9mD,EAAE,WAAW,OAAOA,EAAE,UAAU,MAAMA,EAAE,QAAQ,OAAOA,EAAE,OAAO,IAAI,KAAK,UAAU,GAAG,KAAK,MAAM,GAAG,KAAK,OAAOA,EAAE,OAAOA,EAAE,YAAY,MAAMA,EAAE,iBAAiB,KAAK,MAAM,IAAIklD,EAAE,mGAAmG,EAAE,IAAI9lD,EAAEY,EAAE,gBAAgB,GAAGZ,GAAG,KAAK,CAAC,GAAGY,EAAE,YAAY,KAAK,MAAM,IAAIklD,EAAE,+EAA+E,EAAE9lD,EAAE,CAACY,EAAE,SAAS,EAAE,OAAOA,EAAE,UAAU,CAAC,SAASA,EAAE,WAAW,KAAK,MAAM,IAAIklD,EAAE,uFAAuF,EAAE,IAAI,EAAEllD,EAAE,OAAO,UAAU,KAAK,gBAAgBZ,EAAE,KAAK,MAAM,EAAE,KAAK,UAAU,CAAC,CAAC,MAAMA,CAAC,CAAC,EAAE,IAAIK,EAAE,IAAI0sD,GAAG,KAAK,MAAM,KAAK,gBAAgB,KAAK,CAAC,EAAE,CAAC,EAAE,KAAK,IAAI,EAAE1sD,EAAE,UAAU,EAAEA,EAAE,YAAY,EAAE,IAAI4sD,GAAG,CAAC,cAAc,KAAK,cAAc,CAAC,EAAE,YAAY,CAAC,EAAE,cAAc,CAAC,EAAE,aAAa,CAAC5sD,CAAC,EAAE,cAAc,CAACA,CAAC,EAAE,WAAW,CAAC,IAAI,EAAE,YAAY,CAAC,IAAI,EAAE,YAAY,CAACL,CAAC,EAAE,aAAa,CAACA,CAAC,CAAC,CAAC,CAAC,CAAC,MAAMY,EAAEZ,EAAE,CAAC,MAAM,IAAI8lD,EAAE,6EAA6E,KAAK,MAAM,CAAC,CAAC,SAAS,CAAC,MAAM,CAAC,qBAAqB,KAAK,UAAU,qBAAqB,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,gBAAgB,KAAK,gBAAgB,MAAM,KAAK,MAAM,OAAO,KAAK,OAAO,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEyH,GAAG,UAAU,aAAah2B,GAAE,cAAcg2B,EAAE,EAAE,SAASC,GAAG1tD,EAAE,CAAC,GAAGA,EAAE,YAAY,MAAMA,EAAE,OAAO,KAAK,MAAM,IAAI,MAAM,8HAA8H,EAAE,GAAGA,EAAE,YAAY,MAAMA,EAAE,OAAO,KAAK,MAAM,IAAIgmD,EAAE,kFAAkF,EAAE,IAAI,EAAEhmD,EAAE,WAAWA,EAAE,OAAO,MAAM,GAAG,OAAO,EAAE,CAAC,IAAI,EAAE,OAAOA,EAAE,KAAK,GAAG,IAAIE,EAAEF,EAAE,MAAM,OAAOE,GAAG,OAAOA,EAAE,WAAW,IAAIutD,GAAG,CAAC,gBAAgB,EAAE,KAAKztD,EAAE,KAAK,MAAME,EAAE,OAAOF,EAAE,MAAM,CAAC,EAAE,aAAa,GAAG,cAAc,EAAE,CAAC,SAAS2tD,GAAG3tD,EAAE,EAAE,CAAC,GAAGA,EAAE,OAAO,MAAMA,EAAE,QAAQ,EAAE,MAAM,OAAO,EAAE,GAAG,CAAC,OAAOywB,GAAE,EAAEzwB,EAAE,KAAK,CAAC,OAAOE,EAAN,CAAS,MAAM,IAAI8lD,EAAE,0BAA0B,EAAE,mDAAmDhmD,EAAE,UAAUA,EAAE,SAAS,CAAC,CAAC,CAAC,IAAI4tD,GAAG,KAAK,CAAC,YAAY9sD,EAAE,CAAC,GAAG,KAAK,SAAS,CAAC,EAAE,KAAK,QAAQ,CAAC,EAAE,KAAK,QAAQ,CAAC,EAAEA,aAAa8sD,GAAG,QAAQ1tD,KAAKY,EAAE,SAAS,KAAK,SAASZ,GAAGY,EAAE,SAASZ,GAAGA,KAAKY,EAAE,UAAU,KAAK,QAAQZ,GAAGY,EAAE,QAAQZ,QAAQ,CAAC,GAAGY,GAAG,KAAK,OAAO,QAAQZ,KAAKY,EAAE,KAAK,IAAIZ,EAAE,IAAIA,EAAE,KAAK,CAAC,CAAC,CAAC,IAAIY,EAAEZ,EAAE,EAAE,CAAC,GAAG,KAAK,SAASY,EAAE,KAAK,KAAK,KAAK,SAASA,EAAE,IAAI6sD,GAAG7sD,EAAEZ,CAAC,EAAE,KAAK,QAAQY,EAAE,MAAMA,EAAE,GAAG,GAAG,OAAO,KAAK,QAAQA,EAAE,IAAI,OAAQ,OAAM,IAAIklD,EAAE,uBAAuBllD,EAAE,YAAYA,EAAE,IAAI,EAAE,OAAO,IAAI,CAAC,QAAQA,EAAE,CAAC,KAAK,IAAIA,EAAE,IAAIA,EAAE,KAAK,CAAC,CAAC,OAAOA,EAAE,CAAC,OAAO,KAAK,SAASA,EAAE,KAAK,IAAI,CAAC,OAAO,CAAC,OAAO,OAAO,KAAK,KAAK,OAAO,CAAC,CAAC,SAASA,EAAE,CAAC,GAAGA,aAAamsD,GAAG,CAAC,GAAG,KAAK,SAASnsD,EAAE,KAAK,KAAK,MAAM,IAAIklD,EAAE,oBAAoBllD,EAAE,MAAM,EAAE,OAAO,KAAK,SAASA,EAAE,GAAG,KAAK,CAAC,IAAIZ,EAAE,KAAK,QAAQY,GAAG,GAAGZ,GAAG,KAAK,MAAM,IAAI8lD,EAAE,yCAAyCllD,GAAG,EAAE,OAAO,KAAK,SAASZ,EAAE,CAAC,CAAC,QAAQY,EAAE,CAAC,GAAGA,aAAamsD,GAAG,CAAC,GAAG,KAAK,SAASnsD,EAAE,KAAK,KAAK,MAAM,IAAIklD,EAAE,oBAAoBllD,EAAE,MAAM,EAAE,OAAO,KAAK,QAAQA,EAAE,GAAG,KAAK,CAAC,IAAIZ,EAAE,KAAK,QAAQY,GAAG,GAAGZ,GAAG,KAAK,MAAM,IAAI8lD,EAAE,yCAAyCllD,GAAG,EAAE,OAAO,KAAK,QAAQZ,EAAE,CAAC,CAAC,cAAc,CAAC,KAAK,SAAS,MAAMwzB,GAAG,KAAK,OAAO,CAAC,CAAC,EAAEm6B,GAAG,IAAI1H,GAAG2H,GAAG,IAAI3H,GAAG,SAAS4H,GAAG/tD,EAAE,CAAC6tD,IAAI,MAAMA,GAAG,cAAc7tD,CAAC,EAAE8tD,IAAI,MAAMA,GAAG,cAAc9tD,CAAC,CAAC,CAAC,SAASguD,GAAGhuD,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEL,GAAG,KAAK,GAAGA,EAAE,SAAS,EAAE,MAAM,QAAQF,CAAC,EAAE,EAAE,EAAEA,EAAE,CAACA,CAAC,EAAEiC,EAAE,EAAE,IAAIe,GAAGA,EAAE,IAAI,EAAEd,EAAE,CAAC,EAAEC,EAAE,EAAE,MAAM,EAAE,QAAQa,KAAKf,EAAEE,EAAE,QAAQa,CAAC,IAAI,GAAGd,EAAE,KAAK,EAAE,SAASc,CAAC,CAAC,EAAEd,EAAE,KAAK,IAAI,EAAE5B,GAAG,OAAOA,EAAE,cAAc,GAAG,EAAEA,EAAE,cAAc,EAAE,GAAG,IAAIsC,EAAEX,EAAE,KAAK,GAAG,EAAE,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK,GAAG,EAAEY,EAAEgrD,GAAG,IAAIjrD,CAAC,EAAEE,EAAE,GAAGD,GAAG,KAAK,CAAC,IAAIG,EAAEirD,GAAG,EAAE,CAAC,EAAEprD,EAAEG,EAAE,OAAOF,EAAEE,EAAE,gBAAgB6qD,GAAG,IAAIjrD,EAAEC,CAAC,EAAEirD,GAAG,IAAIlrD,EAAEE,CAAC,CAAC,CAACA,EAAE,CAAC,EAAEvC,GAAG,OAAO,OAAOuC,EAAEgrD,GAAG,IAAIlrD,CAAC,CAAC,EAAE,IAAIG,EAAE,IAAI6qD,GAAG,CAAC,EAAE,QAAQ5qD,EAAE,EAAEA,EAAEH,EAAE,OAAO,EAAEG,EAAE,CAAC,GAAG1C,GAAG,KAAK,CAAC,IAAImH,EAAE8rB,GAAG,EAAE,WAAW9rB,EAAEnH,EAAE,gBAAgBA,EAAE,cAAcmH,GAAGA,EAAEnH,EAAE,gBAAgBA,EAAE,cAAcmH,EAAE,CAAC,IAAIvD,EAAErB,EAAEG,GAAG,EAAEkB,EAAE,YAAY,GAAG,aAAaupD,GAAG,SAAS,IAAIlmD,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE5C,EAAE,CAAC,EAAEC,EAAE,GAAG,QAAQ6C,KAAKvD,EAAE,OAAO,CAAC,IAAIe,EAAElC,EAAE,SAAS0E,CAAC,EAAEvC,EAAEnC,EAAE,QAAQ0E,CAAC,EAAEF,EAAE,KAAKtC,CAAC,EAAE,EAAE,KAAKC,CAAC,EAAEA,GAAG,OAAON,EAAE,IAAIrE,IAAIuC,EAAE2E,EAAE,QAAQ3E,EAAE2E,EAAE,QAAQ,GAAG,CAAC,EAAE,OAAOA,CAAC,GAAGxF,EAAE,QAAQwF,EAAE,IAAI,IAAI,IAAI,CAACxC,EAAE,YAAYwC,EAAE,YAAY,WAAW,IAAI9C,EAAE,KAAKM,CAAC,EAAE,CAACL,IAAI1E,EAAEA,GAAG,CAAC,EAAEA,EAAE,KAAK,EAAE,IAAI,IAAI2E,EAAE2hD,GAAG,EAAE,MAAMj/C,EAAErH,CAAC,CAAC,EAAE4E,EAAE,KAAK,EAAE,kBAAkBA,EAAE,EAAE,YAAYyC,EAAE,CAAC,GAAG,IAAIC,EAAE0mD,GAAGhqD,CAAC,EAAEc,EAAE,MAAM,QAAQwC,CAAC,EAAEA,EAAE,CAACA,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEzC,EAAE,OAAO,EAAEyC,EAAE,CAAC1E,EAAE,OAAOiC,EAAEyC,EAAE,GAAG1E,EAAE,IAAIiC,EAAEyC,GAAG5C,EAAE4C,GAAG,MAAM,QAAQ3C,CAAC,EAAEA,EAAE,GAAGA,CAAC,EAAE,IAAIG,EAAEhD,EAAE,QAAQ+C,EAAEyC,GAAG,IAAI,EAAExC,IAAI,KAAK/C,EAAE+C,GAAGJ,EAAE4C,GAAG,CAAClH,GAAGmzB,GAAG/uB,CAAC,CAAC,CAAC,OAAO5B,EAAE,aAAa,EAAE,EAAEb,EAAEA,EAAE,EAAE,CAAC,SAAS+rD,GAAGjuD,EAAE,EAAE,CAAC8mB,EAAE,OAAO9mB,GAAG,MAAMA,EAAE,OAAO,EAAE,IAAI,uCAAuC,EAAE,IAAIE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAE,GAAGN,EAAE,SAAS,EAAE,CAAC,IAAIO,EAAE4tD,GAAGnuD,EAAE,GAAG,CAAC,EAAEE,EAAEK,EAAE,OAAOD,EAAEC,EAAE,YAAY,KAAK,CAAC,IAAIA,EAAE,IAAI,IAAI,QAAQ,KAAKP,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,aAAaiC,CAAC,EAAEksD,GAAG,EAAE,CAAC,EAAE,QAAQjsD,KAAK,EAAE3B,EAAE,IAAI2B,EAAE,IAAI,IAAIhC,EAAE,KAAKgC,CAAC,EAAE3B,EAAE,IAAI2B,EAAE,IAAI,GAAG,QAAQA,KAAKD,EAAE3B,EAAE4B,IAAI,OAAO5B,EAAE4B,GAAG,IAAI,KAAKD,EAAEC,GAAG,QAAQC,GAAG7B,EAAE4B,GAAG,IAAIC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOjC,EAAE,gBAAgBkuD,GAAG9tD,CAAC,CAAC,CAAC,CAAC,SAAS8tD,GAAGpuD,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQE,KAAKF,EAAE,EAAEE,GAAGF,EAAEE,GAAG,KAAK,OAAO,CAAC,CAAC,SAASiuD,GAAGnuD,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,IAAII,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQ0B,KAAK,EAAE,MAAM,EAAE/B,EAAE,IAAI+B,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,IAAI,EAAE,KAAKjC,CAAC,EAAE,EAAE,OAAO,GAAG,CAAC,IAAIiC,EAAE,EAAE,EAAE,OAAO,GAAG,GAAG/B,EAAE,IAAI+B,EAAE,IAAI,EAAE,CAAC,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAIC,EAAE,EAAE,EAAE,OAAO,KAAK,EAAE,OAAO,EAAE,GAAGD,EAAE,OAAO,SAAS,GAAGC,EAAE,EAAE,IAAI,EAAE5B,EAAE,KAAK2B,CAAC,EAAE/B,EAAE,IAAI+B,EAAE,IAAI,EAAEC,GAAG,EAAE,IAAI,MAAM,CAAC,EAAE,KAAK,EAAE,OAAO,CAAC,EAAE,QAAQC,KAAKF,EAAE,OAAO1B,EAAE4B,EAAE,OAAO,OAAO5B,EAAE4B,EAAE,MAAM,IAAI,KAAK5B,EAAE4B,EAAE,MAAM,IAAIF,EAAE,IAAI,EAAE,CAAC/B,EAAE,IAAIiC,EAAE,IAAI,GAAG,EAAE,KAAKA,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO7B,EAAE,aAAaC,CAAC,CAAC,CAAC,SAAS2tD,GAAGluD,EAAE,CAAC,IAAI,EAAE,GAAGA,EAAE,YAAY,aAAa,SAAS,EAAE,EAAEA,EAAE,YAAY,WAAW,CAAC,IAAIE,EAAE,KAAK,QAAQI,EAAE,EAAEA,EAAEN,EAAE,YAAY,aAAa,OAAO,EAAEM,EAAE,QAAQC,KAAKP,EAAE,YAAY,aAAaM,GAAG,cAAc,GAAGC,EAAE,KAAKP,EAAE,GAAG,CAACE,EAAEI,EAAE,KAAK,CAAC,EAAEN,EAAE,YAAY,YAAYE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAImuD,GAAG5zC,EAAE,EAAE4zC,GAAG,aAAa,qCAAqC,IAAI,IAAIN,EAAE,EAAE,IAAIO,GAAG,CAAC,EAAEluD,GAAGkuD,GAAG,CAAC,QAAQ,IAAIC,GAAG,WAAW,IAAIC,GAAG,OAAO,IAAIC,GAAG,SAAS,IAAIC,EAAE,CAAC,EAAE,SAASC,GAAG3uD,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAIsP,GAAGI,GAAG/J,EAAEp5B,EAAEA,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI4uD,GAAG,cAAcn3B,GAAE,YAAY,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,CAAC,EAAEo3B,GAAG,cAAcD,EAAE,CAAC,YAAY9tD,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,EAAE,KAAK,YAAY,EAAE,KAAK,SAASA,EAAE,UAAU,KAAKA,EAAE,SAAS,KAAK,gBAAgB,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK,KAAK,WAAW,CAAC,MAAMA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEyuD,GAAG7tD,EAAE,KAAK,IAAI,EAAE,EAAEw9B,GAAGp+B,EAAE,EAAE,KAAK,QAAQ,EAAE,OAAOk5B,EAAEt4B,EAAEo4B,GAAG,EAAEJ,EAAEuwB,GAAG,EAAEnpD,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,SAAS,KAAK,SAAS,KAAK,KAAK,IAAI,CAAC,CAAC,EAAE2uD,GAAG,UAAU,UAAUp3B,GAAE,cAAco3B,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY9tD,EAAE,CAAC,MAAM,EAAE,KAAK,YAAY,EAAE,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK,KAAK,WAAW,CAAC,MAAMA,EAAE,CAAC,OAAO2yB,EAAE,IAAIyF,GAAGp4B,EAAEg4B,EAAEuwB,GAAG,EAAEsF,GAAG7tD,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,KAAK,IAAI,CAAC,CAAC,EAAEguD,GAAG,UAAU,WAAWr3B,GAAE,cAAcq3B,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,MAAM9tD,EAAE,CAAC,OAAOwrC,GAAGxrC,CAAC,CAAC,CAAC,EAAEiuD,GAAG,UAAU,SAASt3B,GAAE,cAAcs3B,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAY9tD,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,EAAE,KAAK,YAAY,EAAE,KAAK,YAAY,EAAE,KAAK,SAASA,EAAE,UAAU,KAAKA,EAAE,SAAS,KAAK,gBAAgB,KAAK,SAASA,EAAE,UAAU,KAAKA,EAAE,SAAS,KAAK,gBAAgB,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK,KAAK,YAAY,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK,KAAK,WAAW,CAAC,MAAMA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEyuD,GAAG7tD,EAAE,KAAK,IAAI,EAAE,EAAEg4B,EAAEM,EAAE,KAAK,KAAKkF,GAAGp+B,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAEk5B,EAAE,EAAE,KAAK,KAAKl5B,CAAC,CAAC,EAAE,OAAOk5B,EAAEt4B,EAAEo4B,GAAG,EAAEJ,EAAEuwB,GAAG,EAAEnpD,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,SAAS,KAAK,SAAS,SAAS,KAAK,SAAS,KAAK,KAAK,KAAK,KAAK,KAAK,IAAI,CAAC,CAAC,EAAE8uD,GAAG,UAAU,aAAav3B,GAAE,cAAcu3B,EAAE,EAAE,IAAIC,GAAG,CAAC,QAAQ,UAAU,WAAW,aAAa,OAAO,SAAS,SAAS,UAAU,EAAE,SAASC,GAAGlvD,EAAE,CAAC,OAAO4mD,GAAG5mD,CAAC,CAAC,CAAC,SAASmvD,GAAGnvD,EAAE,EAAE,CAAC,EAAE,CAAC,OAAO8mD,GAAG9mD,EAAEy3B,GAAE,iBAAiB,OAAO,EAAE,aAAa,EAAE,YAAY,CAAC,CAAC,SAAS23B,GAAGpvD,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,KAAK,GAAG,OAAOA,GAAG,SAAS,CAAC,IAAIE,EAAE,CAAC,UAAUF,KAAKivD,GAAGA,GAAGjvD,GAAGA,EAAE,OAAO,CAAC,CAAC,EAAE,OAAOmvD,GAAGjvD,CAAC,CAAC,KAAM,QAAOF,aAAa4uD,GAAG5uD,EAAEmvD,GAAGnvD,CAAC,CAAC,CAAC,SAASuuD,GAAGvuD,EAAE,CAAC,OAAO,IAAI6uD,GAAG7uD,CAAC,CAAC,CAAC,SAAS0uD,GAAG1uD,EAAE,CAAC,OAAO,IAAI8uD,GAAG9uD,CAAC,CAAC,CAAC,SAASyuD,IAAI,CAAC,OAAO,IAAIM,EAAE,CAAC,SAASP,GAAGxuD,EAAE,CAAC,OAAO,IAAIgvD,GAAGhvD,CAAC,CAAC,CAAC,IAAIqvD,GAAG,CAAC,EAAEjvD,GAAGivD,GAAG,CAAC,SAAS,IAAIC,GAAG,aAAa,IAAIC,GAAG,cAAc,IAAIC,GAAG,SAAS,IAAIC,GAAG,UAAU,IAAIC,GAAG,SAAS,IAAIC,GAAG,YAAY,IAAIC,GAAG,aAAa,IAAIC,GAAG,KAAK,IAAIC,GAAG,WAAW,IAAIC,GAAG,aAAa,IAAIC,GAAG,cAAc,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,MAAM,IAAIC,EAAE,CAAC,EAAE,SAASA,IAAI,CAAC,OAAO,IAAIlF,EAAE,CAAC,SAAS4E,IAAI,CAAC,OAAO,IAAI3E,EAAE,CAAC,SAASmE,GAAGtvD,EAAE,CAAC,OAAO,IAAIorD,GAAGprD,CAAC,CAAC,CAAC,SAASiwD,GAAGjwD,EAAE,CAAC,OAAO,IAAIqrD,GAAGrrD,CAAC,CAAC,CAAC,SAASgwD,GAAGhwD,EAAE,CAAC,OAAO,IAAIsrD,GAAGtrD,CAAC,CAAC,CAAC,SAASkwD,GAAGlwD,EAAE,CAAC,OAAO,IAAIurD,GAAGvrD,CAAC,CAAC,CAAC,SAAS2vD,GAAG3vD,EAAE,CAAC,OAAO,IAAIwrD,GAAGxrD,CAAC,CAAC,CAAC,SAASmwD,GAAGnwD,EAAE,CAAC,OAAO,IAAI0rD,GAAG1rD,CAAC,CAAC,CAAC,SAASwvD,GAAGxvD,EAAE,CAAC,OAAO,IAAI2rD,GAAG3rD,CAAC,CAAC,CAAC,SAASuvD,GAAGvvD,EAAE,CAAC,OAAO,IAAI4rD,GAAG5rD,CAAC,CAAC,CAAC,SAASyvD,GAAGzvD,EAAE,CAAC,OAAO,IAAI6rD,GAAG7rD,CAAC,CAAC,CAAC,SAAS0vD,GAAG1vD,EAAE,CAAC,OAAO,IAAI8rD,GAAG9rD,CAAC,CAAC,CAAC,SAAS4vD,GAAG5vD,EAAE,CAAC,OAAO,IAAI+rD,GAAG/rD,CAAC,CAAC,CAAC,SAAS6vD,GAAG7vD,EAAE,CAAC,OAAO,IAAIgsD,GAAGhsD,CAAC,CAAC,CAAC,SAAS+vD,GAAG/vD,EAAE,CAAC,OAAO,IAAIisD,GAAGjsD,CAAC,CAAC,CAAC,IAAIqwD,GAAG,CAAC,EAAEjwD,GAAGiwD,GAAG,CAAC,MAAM,IAAIhD,GAAG,IAAI,IAAIiD,GAAG,QAAQ,IAAIC,GAAG,WAAW,IAAIC,GAAG,IAAI,IAAIC,GAAG,aAAa,IAAIC,GAAG,QAAQ,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,UAAU,IAAIC,GAAG,UAAU,IAAIC,GAAG,UAAU,IAAIC,GAAG,aAAa,IAAIC,GAAG,aAAa,IAAIC,GAAG,aAAa,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,cAAc,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,YAAY,IAAIC,GAAG,OAAO,IAAIC,GAAG,OAAO,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,OAAO,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,WAAW,IAAIC,GAAG,eAAe,IAAIC,GAAG,WAAW,IAAIC,GAAG,MAAM,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,IAAI,IAAIC,GAAG,QAAQ,IAAIC,GAAG,IAAI,IAAIC,GAAG,UAAU,IAAIC,GAAG,QAAQ,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,cAAc,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,IAAI,IAAIC,GAAG,QAAQ,IAAIC,GAAG,MAAM,IAAIC,GAAG,WAAW,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,UAAU,IAAIC,GAAG,KAAK,IAAIC,GAAG,SAAS,IAAIC,GAAG,QAAQ,IAAIC,GAAG,UAAU,IAAIC,GAAG,UAAU,IAAIC,GAAG,aAAa,IAAIC,GAAG,aAAa,IAAIC,GAAG,aAAa,IAAIC,GAAG,QAAQ,IAAIC,GAAG,QAAQ,IAAIC,GAAG,SAAS,IAAIC,GAAG,QAAQ,IAAIC,GAAG,MAAM,IAAIC,GAAG,KAAK,IAAIC,GAAG,aAAa,IAAIC,GAAG,UAAU,IAAIC,GAAG,QAAQ,IAAIC,GAAG,SAAS,IAAIC,GAAG,IAAI,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,UAAU,IAAIC,GAAG,cAAc,IAAIC,GAAG,QAAQ,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,aAAa,IAAIC,GAAG,cAAc,IAAIC,EAAE,CAAC,EAAE,eAAeC,GAAGn1D,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,IAAI,EAAE,CAAC,EAAEE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAE,QAAQC,KAAKP,EAAE,CAAC,IAAI,EAAEA,EAAEO,GAAG,GAAG,OAAO,GAAG,SAAS,CAAC,IAAI,EAAE,EAAE,EAAE,KAAK,EAAE,KAAK,CAAC,EAAEL,EAAE,KAAKK,CAAC,EAAED,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,EAAE,OAAO,EAAE,CAAC,IAAIC,EAAE,MAAM,QAAQ,IAAI,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEP,EAAEE,EAAE,IAAIK,EAAE,GAAG,GAAGmzB,GAAGpzB,CAAC,CAAC,CAAC,CAAC,SAAS80D,GAAGp1D,EAAE,CAAC,GAAGA,GAAG,KAAK,QAAQ,KAAKA,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAG,OAAOE,GAAG,UAAUA,EAAE,QAAQ,CAAC,CAAC,CAAC,IAAIm1D,IAAI,SAASr1D,EAAE,CAACA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,QAAQ,GAAG,SAAS,GAAGq1D,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,IAAIC,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,eAAe,IAAI,CAAC,UAAUz0D,EAAE,CAAC,KAAK,OAAOA,CAAC,CAAC,MAAM,aAAaA,EAAEZ,EAAE,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,CAAC,MAAM,aAAaY,EAAE,CAAC,CAAC,MAAM,WAAWA,EAAE,CAAC,CAAC,SAASA,EAAE,CAAC,CAAC,EAAE00D,GAAG,KAAK,CAAC,YAAY10D,EAAEZ,EAAE,GAAG,CAACY,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,UAAUA,EAAE,KAAK,YAAYZ,CAAC,CAAC,OAAOY,EAAE,CAAC,KAAK,UAAU,KAAKA,CAAC,CAAC,CAAC,UAAUA,EAAE,CAAC,QAAQZ,KAAK,KAAK,UAAUA,EAAE,UAAUY,CAAC,CAAC,CAAC,SAASA,EAAE,CAAC,QAAQZ,KAAK,KAAK,UAAUA,EAAE,SAASY,CAAC,CAAC,CAAC,MAAM,aAAaA,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQ,KAAK,KAAK,UAAU,MAAM,EAAE,aAAaY,EAAEZ,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQ,KAAK,KAAK,UAAU,MAAM,EAAE,WAAWY,EAAEZ,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQ,KAAK,KAAK,UAAU,MAAM,EAAE,aAAaY,EAAEZ,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQ,KAAK,KAAK,UAAU,MAAM,EAAE,WAAWY,EAAEZ,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQZ,KAAK,KAAK,UAAU,MAAMA,EAAE,aAAaY,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,QAAQZ,KAAK,KAAK,UAAU,MAAMA,EAAE,WAAWY,CAAC,CAAC,CAAC,EAAE20D,GAAG,cAAcF,EAAE,CAAC,aAAa,CAAC,MAAM,CAAC,CAAC,MAAM,aAAaz0D,EAAE,CAAC,KAAK,KAAK,EAAE,KAAK,OAAO,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,IAAI,EAAEA,EAAE,MAAM,KAAK,EAAEA,EAAE,KAAK,KAAK,MAAM,EAAE,QAAQK,KAAKL,EAAE,CAAC,IAAI6B,EAAE7B,EAAEK,GAAG,GAAG,OAAOwB,GAAG,SAAS,KAAK,OAAO,eAAexB,CAAC,IAAI,KAAK,OAAOA,GAAG,GAAG,KAAK,OAAOA,GAAG,KAAK,OAAOA,GAAGwB,EAAE,MAAM,CAAC,IAAIC,EAAEzB,KAAK,KAAK,OAAOyB,EAAE,KAAK,OAAOzB,GAAG,KAAK,OAAOA,GAAG,EAAE,IAAI0B,EAAEwxB,EAAE,IAAIqF,EAAE,KAAK,OAAOv4B,GAAG64B,EAAEr3B,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAOxB,GAAG0B,EAAED,GAAG,MAAMA,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,MAAM,WAAWlB,EAAEZ,EAAE,CAAC,GAAGA,GAAG,KAAK,QAAQ,KAAK,KAAK,OAAO,QAAQ,KAAK,OAAO,IAAI,OAAO,OAAO,KAAK,OAAO,IAAI,SAASA,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK,KAAKuzB,EAAE,IAAI,CAAC,IAAIlzB,EAAE64B,EAAEF,GAAG,EAAE,KAAK,IAAI,EAAE,KAAK,OAAO,EAAE,EAAEh5B,EAAE,GAAGK,EAAE,KAAK,OAAO,GAAG,QAAQ,EAAEozB,GAAGzzB,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,EAAEw1D,GAAG,cAAcH,EAAE,CAAC,MAAM,aAAaz0D,EAAE,CAAC,KAAK,MAAM,CAAC,EAAE,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,WAAWA,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,MAAM,KAAKY,CAAC,EAAE,QAAQ,KAAKZ,EAAE,KAAK,QAAQ,IAAI,OAAO,KAAK,QAAQ,GAAG,CAAC,GAAG,KAAK,QAAQ,GAAG,KAAKA,EAAE,EAAE,CAAC,CAAC,MAAM,UAAU,CAAC,IAAIY,EAAE,CAAC,EAAEZ,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQ6B,KAAK,KAAK,QAAQ,CAAC,IAAIC,EAAE,KAAK,QAAQD,GAAG,QAAQE,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAEC,EAAE,GAAG,OAAOD,EAAEC,IAAI,SAAS,CAAC,IAAIC,EAAEF,EAAEC,GAAGnB,EAAE,KAAKoB,EAAE,KAAK,CAAC,EAAEhC,EAAE,KAAK6B,CAAC,EAAE,EAAE,KAAKE,CAAC,CAAC,CAAC,CAAC,IAAI1B,EAAE,MAAM,QAAQ,IAAIO,CAAC,EAAE,QAAQiB,EAAE,EAAEA,EAAExB,EAAE,OAAO,EAAEwB,EAAE,KAAK,QAAQ7B,EAAE6B,IAAI,EAAEA,IAAI,QAAQ,EAAE,KAAK,QAAQ7B,EAAE6B,IAAI,EAAEA,IAAIxB,EAAEwB,GAAG,EAAE,CAAC,EAAE4zD,GAAG,cAAcJ,EAAE,CAAC,YAAYz0D,EAAEZ,EAAE,CAAC,GAAG,MAAM,EAAE,KAAK,aAAa,EAAE,KAAK,QAAQY,EAAE,QAAQ,KAAK,cAAcA,EAAE,eAAey5C,GAAG,KAAK,WAAWr6C,GAAG,OAAO,KAAK,aAAa,SAAS,KAAK,WAAWo1D,IAAI,KAAK,aAAa,SAASx0D,EAAE,SAAS,KAAK,MAAM,IAAI,MAAM,iHAAiH,EAAEgmB,EAAE,SAAS,KAAK,UAAU,IAAI,KAAK,UAAUygC,GAAG,KAAK,UAAU,KAAK,IAAI,EAAE,KAAK,WAAW,KAAK,OAAO,GAAG,KAAK,WAAWzmD,EAAE,aAAa,KAAK,SAASA,EAAE,WAAW,KAAK,WAAWA,EAAE,aAAa,KAAK,SAASA,EAAE,WAAW,KAAK,WAAWA,EAAE,aAAa,KAAK,SAASA,EAAE,WAAW,KAAK,MAAMA,EAAE,OAAO,CAAC,MAAM,UAAUA,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,CAAC,EAAE,KAAK,OAAO,OAAO,MAAM40D,GAAG,CAAC,EAAE50D,EAAE,KAAK,KAAK,MAAMO,EAAEZ,EAAE,CAAC,CAAC,GAAGK,EAAE,KAAK,KAAK,cAAc,CAAC,EAAE,MAAM,QAAQ,IAAIA,CAAC,CAAC,CAAC,MAAM,aAAaO,EAAEZ,EAAE,CAAC,KAAK,aAAaY,EAAE,KAAK,YAAY,OAAO,MAAMq0D,GAAGj1D,CAAC,EAAE,MAAM,KAAK,WAAWY,EAAEZ,CAAC,EAAE,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,UAAU,OAAO,MAAMi1D,GAAGj1D,CAAC,EAAE,EAAE,KAAK,KAAK,SAASY,EAAEZ,CAAC,CAAC,GAAG,KAAK,aAAa,SAAS,EAAE,KAAK,KAAK,cAAc,CAAC,EAAE,MAAM,QAAQ,IAAI,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAAC,KAAK,YAAY,OAAO,MAAMi1D,GAAGj1D,CAAC,EAAE,MAAM,KAAK,WAAWY,EAAEZ,CAAC,EAAE,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,UAAU,OAAO,MAAMi1D,GAAGj1D,CAAC,EAAE,EAAE,KAAK,KAAK,SAASY,EAAEZ,CAAC,CAAC,GAAG,KAAK,aAAa,QAAQ,EAAE,KAAK,KAAK,cAAc,CAAC,EAAE4mB,EAAE,SAAS,KAAK,UAAU,GAAG,EAAE,KAAK,KAAK,UAAU,KAAK,aAAahmB,EAAEZ,CAAC,CAAC,EAAE,MAAM,QAAQ,IAAI,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAE,CAAC,KAAK,YAAY,OAAO,MAAMq0D,GAAGr0D,CAAC,EAAE,MAAM,KAAK,WAAWA,CAAC,EAAE,CAAC,MAAM,WAAWA,EAAE,CAAC,KAAK,UAAU,OAAO,MAAMq0D,GAAGr0D,CAAC,EAAE,MAAM,KAAK,SAASA,CAAC,EAAE,CAAC,EAAE,SAAS80D,GAAG51D,EAAE,EAAE,CAAC,OAAOA,GAAG,OAAOA,EAAE,CAAC,GAAGA,aAAau1D,GAAG,CAACv1D,CAAC,EAAE,MAAM,QAAQA,CAAC,GAAGA,EAAE,aAAau1D,GAAGv1D,EAAEwmD,GAAGxmD,CAAC,EAAE,IAAI,GAAG,IAAI21D,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAIE,GAAG,KAAK,CAAC,aAAa,CAAC,CAAC,OAAO,4BAA4B/0D,EAAEZ,EAAE,CAAC4mB,EAAE,OAAOhmB,GAAG,GAAG,OAAO,UAAUA,CAAC,EAAE,IAAI,8DAA8DA,GAAG,EAAE+0D,GAAG,kBAAkB31D,CAAC,EAAE21D,GAAG,aAAa/0D,IAAI,OAAO+0D,GAAG,aAAa/0D,GAAG,CAAC,GAAG+0D,GAAG,aAAa/0D,GAAG,KAAKZ,CAAC,CAAC,CAAC,OAAO,kBAAkBY,EAAE,CAAC,QAAQZ,KAAK21D,GAAG,aAAaA,GAAG,aAAa,CAAC31D,GAAG,QAAQK,GAAG,CAAC,GAAGA,IAAIO,EAAE,MAAM,IAAIklD,EAAE,iCAAiC,CAAC,CAAC,CAAC,CAAC,OAAO,OAAO,CAAC6P,GAAG,aAAa,CAAC,CAAC,CAAC,OAAO,gBAAgB/0D,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,QAAQ,KAAK21D,GAAG,aAAa,CAAC,IAAIt1D,EAAE,CAAC,EAAEO,GAAGP,GAAGL,EAAE,KAAK,GAAG21D,GAAG,aAAat1D,EAAE,CAAC,CAAC,OAAOL,EAAE,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,EAAE21D,GAAG,aAAa,CAAC,EAAE,SAASC,GAAG91D,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAE,CAAC,IAAIC,EAAE,IAAIuzD,GAAG9yD,EAAE,CAAC,IAAI6yD,GAAG,GAAGI,GAAG,gBAAgB,CAAC,CAAC,EAAE71D,GAAG,MAAM4C,EAAE,KAAK,GAAG5C,CAAC,EAAE4C,EAAE,KAAKT,CAAC,EAAE,IAAIU,EAAE,IAAI2yD,GAAG5yD,CAAC,EAAE,OAAOC,EAAE,UAAU,CAAC,OAAO3C,EAAE,aAAaI,EAAE,QAAQC,EAAE,MAAM,EAAE,UAAU,EAAE,QAAQ,EAAE,aAAa0B,EAAE,QAAQC,CAAC,CAAC,EAAE,CAAC,aAAaW,EAAE,QAAQV,CAAC,CAAC,CAAC,SAAS4zD,GAAG/1D,EAAE,EAAE,CAAC,EAAEE,EAAE,GAAG,CAAC,OAAO4mD,GAAG9mD,EAAEy3B,GAAE,iBAAiB,OAAO,EAAE,aAAa,EAAE,QAAQv3B,CAAC,CAAC,CAAC,SAAS81D,GAAGh2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAACzzB,EAAE,QAAQ,YAAYA,EAAEywB,GAAEzwB,EAAE,SAAS,GAAG,IAAIE,EAAEijC,GAAGmnB,GAAGtqD,CAAC,EAAE,EAAE,EAAE,EAAEM,EAAE89B,GAAGl+B,EAAE,MAAMmpD,GAAG,CAAC,EAAE9oD,EAAEwiC,GAAGmF,GAAGhoC,EAAEI,CAAC,CAAC,EAAE,OAAO44B,GAAGl5B,EAAEO,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS01D,GAAGj2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI2U,GAAGkiB,GAAG5jB,GAAG,EAAE1mC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,SAASk2D,GAAGl2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI2U,GAAG9O,GAAGoN,GAAG,EAAE1mC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,SAASm2D,GAAGn2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEwmC,GAAG1mC,EAAE,CAAC,EAAEM,EAAEg+B,GAAGhF,GAAGt5B,CAAC,EAAEqpD,GAAG,EAAE,OAAO,SAAS,EAAE9oD,EAAE+4B,GAAGJ,GAAGh5B,EAAEI,CAAC,CAAC,EAAE,OAAO84B,EAAE,IAAIgP,GAAG7nC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS61D,GAAGp2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEo+B,GAAG,EAAE+qB,GAAG,EAAE,OAAO,SAAS,EAAE/oD,EAAEqlC,GAAG7M,EAAE,EAAE54B,CAAC,CAAC,EAAEK,EAAE+9B,GAAGt+B,EAAEqpD,GAAG,EAAE,OAAO,SAAS,EAAE,EAAE1jB,GAAG7M,EAAE,EAAEv4B,CAAC,CAAC,EAAE,OAAO6nC,GAAGkiB,GAAG5jB,GAAGpmC,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS+1D,GAAGr2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEgoC,GAAG,EAAExB,GAAG,EAAEtN,EAAEp5B,EAAE,CAAC,CAAC,CAAC,EAAE,OAAOooC,GAAGkiB,GAAGpqD,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASo2D,GAAGt2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEgoC,GAAG,EAAExB,GAAG,EAAEtN,EAAEp5B,EAAE,CAAC,CAAC,CAAC,EAAE,OAAOooC,GAAGloC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASq2D,GAAGv2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEijC,GAAG/J,EAAEp5B,EAAE,CAAC,EAAE,EAAE,EAAEM,EAAEkiC,GAAGpJ,EAAEsN,GAAG,EAAE1mC,CAAC,EAAE,CAAC,EAAE,EAAE,EAAE,OAAOkoC,GAAG,EAAEpP,EAAE,EAAE4N,GAAGpmC,EAAEJ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAASs2D,GAAGx2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAE,KAAK,IAAI,CAAC,EAAEI,EAAEomC,GAAG,EAAE1mC,CAAC,EAAEO,EAAEmmC,GAAG5N,EAAEx4B,EAAEgmC,GAAGlN,EAAE,GAAG94B,CAAC,CAAC,CAAC,EAAEJ,CAAC,EAAE,OAAOkoC,GAAG7nC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASk2D,GAAGz2D,EAAE,EAAEE,EAAE,GAAG,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAGvzB,EAAE,EAAE0uC,GAAG,CAAC,MAAM,CAAC,IAAItuC,EAAE6iC,GAAG,EAAE,EAAE,MAAM,OAAO,EAAE,EAAE,EAAE,EAAEjK,GAAG,EAAE54B,CAAC,CAAC,CAAC,OAAO,EAAEg+B,GAAG,EAAE+qB,GAAG,EAAE,EAAEA,GAAG,CAAC,EAAE50B,GAAG0O,GAAG/J,EAAE3I,GAAEzwB,EAAE,SAAS,EAAE2lC,GAAG,CAAC,CAAC,EAAE,EAAE,MAAM,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS+wB,GAAG12D,EAAE,EAAEE,EAAE,GAAG,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAInzB,EAAEmwB,GAAE2T,GAAGulB,GAAG3pD,CAAC,CAAC,EAAE,OAAO,EAAE,EAAEs+B,GAAG,EAAE+qB,GAAG,EAAE,EAAEA,GAAG,CAAC,EAAE,IAAI9oD,EAAE,EAAE,MAAM,EAAE27B,EAAElJ,GAAG1yB,EAAEC,EAAEA,EAAE,OAAO,EAAE,EAAEA,CAAC,EAAE,OAAOk2D,GAAG,EAAE,EAAEv2D,CAAC,CAAC,CAAC,CAAC,CAAC,SAASy2D,GAAG32D,EAAE,EAAE,CAAC,GAAG,CAAC8mB,EAAE,YAAY9mB,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,IAAIgmD,EAAE,8DAA8D,KAAK,UAAUhmD,EAAE,KAAK,SAAS,KAAK,UAAU,EAAE,KAAK,GAAG,EAAE,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEosC,GAAG,CAAC,EAAEhsC,EAAEm0B,GAAG6E,GAAG,CAAC,CAAC,EAAE,OAAOR,EAAE4N,GAAGxmC,EAAEk5B,EAAE,EAAEp5B,CAAC,CAAC,EAAE6lC,GAAGnC,GAAGpjC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAASs2D,GAAG52D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAE,OAAOA,EAAEo+B,GAAG,EAAE+qB,GAAG,EAAE,EAAEA,GAAG,CAAC,EAAEnpD,EAAEylC,GAAGzM,GAAGh5B,EAAEwmC,GAAG,EAAExmC,CAAC,CAAC,CAAC,EAAEkoC,GAAGuuB,GAAG32D,EAAEE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS22D,GAAG72D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEo+B,GAAGt+B,EAAEqpD,GAAG,EAAE,CAAC,EAAE/oD,EAAEg+B,GAAG,EAAE+qB,GAAG,EAAE,CAAC,EAAE,OAAOlmB,GAAG/J,EAAEp5B,EAAE2lC,GAAGzM,GAAGh5B,EAAEI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASw2D,GAAG92D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEylC,GAAG7M,EAAEuwB,GAAG,EAAE,CAAC,CAAC,EAAE,OAAOjhB,GAAG1B,GAAG,EAAEtN,EAAEp5B,EAAEE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS62D,GAAG/2D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAE81D,GAAGh2D,EAAE,EAAE,EAAEM,EAAE01D,GAAG,EAAE,EAAE,EAAEz1D,EAAE64B,EAAEl5B,EAAEI,CAAC,EAAE,OAAOm0B,GAAG0O,GAAG5iC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIy2D,GAAG,CAAC,iBAAiBf,GAAG,kBAAkBC,GAAG,4BAA4BC,GAAG,4BAA4BC,GAAG,aAAaC,GAAG,MAAMC,GAAG,iBAAiBC,GAAG,QAAQC,GAAG,wBAAwBC,GAAG,8BAA8BC,GAAG,mBAAmBE,GAAG,0BAA0BC,GAAG,QAAQC,GAAG,gBAAgBC,EAAE,EAAE,SAASE,GAAGj3D,EAAE,CAAC,GAAG,OAAOA,GAAG,SAAS,CAAC,GAAGA,KAAKg3D,GAAG,OAAOA,GAAGh3D,GAAG,IAAI,EAAE,gBAAgBA,IAAI,MAAMA,EAAE,YAAY,EAAE,SAAS,qBAAqB,IAAI,EAAE,gBAAgBA,yFAAyF,IAAIgmD,EAAE,CAAC,CAAC,KAAM,QAAOhmD,CAAC,CAAC,SAASk3D,GAAGl3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEk5B,EAAE,GAAGmQ,GAAG,CAAC,CAAC,EAAEjpC,EAAEipD,GAAG/kB,GAAG,EAAEtkC,CAAC,EAAEF,EAAE,KAAK,EAAE,OAAOooC,GAAGpH,GAAGhhC,EAAEM,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS62D,GAAGn3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI81B,GAAGvoB,GAAG9G,GAAGl6B,EAAE,EAAE,EAAEk6B,GAAG,EAAE,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,SAASk9B,GAAGp3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAIhD,GAAE0S,GAAG6D,GAAGhG,GAAGhhC,EAAE,CAAC,EAAEghC,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,SAASq2B,GAAGr3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAIhD,GAAE0S,GAAG6D,GAAGhG,GAAGhhC,EAAE,CAAC,EAAEghC,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,SAASs2B,GAAGt3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAIhD,GAAE0S,GAAG6D,GAAGhG,GAAGhhC,EAAE,CAAC,EAAEghC,GAAG,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,SAASu2B,GAAGv3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEk3D,GAAGp3D,EAAE,CAAC,EAAEM,EAAEg3D,GAAGt3D,EAAE,CAAC,EAAEO,EAAEu4B,EAAE54B,EAAEI,CAAC,EAAE,OAAOmwB,GAAEyQ,GAAGsD,GAAGjkC,EAAE,CAAC,EAAE24B,GAAGh5B,EAAEK,CAAC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,SAASi3D,GAAGx3D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAI,CAAC,IAAIvzB,EAAEk3D,GAAGp3D,EAAE,CAAC,EAAEM,EAAE+2D,GAAGr3D,EAAE,CAAC,EAAEO,EAAEu4B,EAAE54B,EAAEI,CAAC,EAAE,OAAOmwB,GAAEyQ,GAAGsD,GAAGjkC,EAAE,CAAC,EAAE24B,GAAGh5B,EAAEK,CAAC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,SAASk3D,GAAGz3D,EAAE,EAAE,CAAC,OAAO42D,GAAG52D,EAAE,CAAC,CAAC,CAAC,SAAS03D,GAAG13D,EAAE,EAAE,CAAC,OAAOA,EAAE,OAAO,EAAE,OAAOA,EAAE0vC,GAAG1vC,EAAE,CAACA,EAAE,KAAK,CAAC,CAAC,GAAG,EAAEk6B,GAAG,EAAE,EAAE,EAAE,EAAE,QAAQl6B,EAAE,QAAQ,EAAEywB,GAAE,EAAEzwB,EAAE,KAAK,GAAGywB,GAAEuQ,GAAGhhC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI23D,GAAG1B,GAAG2B,GAAG3B,GAAG4B,GAAG3B,GAAG4B,GAAG5B,GAAG6B,GAAG5B,GAAG6B,GAAG7B,GAAG8B,GAAGxB,GAAGyB,GAAGnB,GAAGoB,GAAGzB,GAAG0B,GAAG,CAAC,eAAelB,GAAG,oBAAoBC,GAAG,UAAUI,GAAG,wBAAwBU,GAAG,8BAA8BE,GAAG,IAAIR,GAAG,IAAIC,GAAG,IAAIC,GAAG,IAAIC,GAAG,KAAKC,GAAG,KAAKC,GAAG,OAAOE,EAAE,EAAE,SAASG,GAAGr4D,EAAE,CAAC,GAAG,OAAOA,GAAG,UAAUA,KAAKo4D,GAAG,OAAOA,GAAGp4D,GAAG,GAAG,OAAOA,GAAG,UAAUA,GAAG,KAAK,OAAOA,EAAE,MAAM,IAAIgmD,EAAE,kBAAkBhmD,GAAG,CAAC,CAAC,SAASs4D,GAAGt4D,EAAE,CAAC,GAAGqmD,GAAGrmD,IAAI,KAAK,0BAA0BA,GAAG,EAAE,OAAOA,GAAG,SAAS,OAAOA,EAAE,CAAC,IAAI,EAAE,QAAQE,KAAK,OAAO,KAAK82D,EAAE,EAAE,GAAGA,GAAG92D,KAAKF,EAAE,CAAC,EAAEE,EAAE,KAAK,CAAC,GAAG,IAAI,OAAO,OAAO,EAAE,QAAQA,KAAK,OAAO,KAAKk4D,EAAE,EAAE,GAAGA,GAAGl4D,KAAKF,EAAE,CAAC,EAAEE,EAAE,KAAK,CAAC,OAAO,IAAI,OAAO,EAAEF,EAAE,IAAI,CAAC,CAAC,SAASu4D,GAAGv4D,EAAE,CAAC,IAAI,EAAE,CAAC,QAAQ,IAAIq6C,GAAG,QAAQ,GAAG,EAAE,SAAS,IAAIA,GAAG,SAAS,EAAE,IAAIgP,GAAG,CAAC,EAAE,KAAK,IAAIhP,GAAG,KAAK,KAAK,GAAG,KAAKgP,GAAG,CAAC,EAAE,OAAO,IAAIhP,GAAG,OAAO,KAAK,GAAG,KAAKgP,GAAG,EAAE,CAAC,EAAE,QAAQ,IAAIhP,GAAG,QAAQ,KAAK,GAAG,EAAEgP,GAAG,CAAC,EAAE,IAAI,IAAIhP,GAAG,IAAI,GAAG,CAAC,EAAE,GAAG,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,SAAS,EAAE,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,IAAI,EAAE,IAAIr6C,KAAK,EAAE,OAAO,EAAEA,GAAG,EAAE,MAAM,IAAIgmD,EAAE,qBAAqBhmD,GAAG,CAAC,CAAC,SAASw4D,GAAGx4D,EAAE,EAAEE,EAAE,GAAG,CAAC,GAAGF,GAAG,MAAM,OAAOA,GAAG,UAAU,OAAO,eAAeA,CAAC,IAAI,OAAO,WAAW,CAACy4D,GAAGz4D,CAAC,EAAE,MAAM,IAAI,MAAM,oEAAoE,EAAE,GAAGE,EAAE,CAAC,IAAII,EAAE,KAAK,UAAUN,CAAC,EAAEM,EAAE,OAAO,SAAS,QAAQ,KAAK,mCAAmC,mCAAmCA,EAAE,qJAAqJ,UAAU,CAAC,CAAC,CAAC,SAASm4D,GAAGz4D,EAAE,CAAC,GAAGA,IAAI,KAAK,MAAM,GAAG,GAAG,OAAOA,GAAG,SAAS,GAAG,OAAO,eAAeA,CAAC,IAAI,OAAO,UAAU,CAAC,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE,QAAQE,KAAK,EAAE,GAAG,OAAOA,GAAG,UAAU,CAACu4D,GAAGz4D,EAAEE,EAAE,EAAE,MAAM,GAAG,MAAM,EAAE,SAAS,MAAM,QAAQF,CAAC,EAAE,CAAC,QAAQ,KAAKA,EAAE,GAAG,CAACy4D,GAAG,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,KAAM,OAAM,OAAO,CAAC,IAAI,EAAE,OAAOz4D,EAAE,OAAO,IAAI,UAAU,IAAI,UAAU,IAAI,SAAS,CAAC,CAAC,SAAS04D,GAAG14D,EAAE,EAAEE,EAAEI,EAAE,QAAQ,IAAI,CAAC,IAAIC,EAAEo4D,GAAG34D,CAAC,EAAE,EAAE,CAAC,eAAe,cAAc,eAAe,SAAS,EAAEO,GAAG,EAAE,GAAG,GAAGL,EAAEA,GAAG,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,EAAE,GAAG,IAAIA,EAAEA,GAAG,CAAC,IAAI,IAAI,GAAG,GAAG,CAAC,GAAGA,EAAEA,EAAE,OAAO,IAAI,IAAIA,EAAEA,EAAE,IAAI0C,GAAG,KAAK,MAAM,EAAEA,CAAC,CAAC,GAAG,IAAI,EAAE,GAAG,CAACrC,EAAE,CAAC,EAAE,KAAK,iBAAiB,EAAE,EAAE,CAAC,EAAE,QAAQqC,KAAK5C,EAAE,aAAa,EAAE,KAAK,GAAGA,EAAE,aAAa4C,EAAE,CAAC,CAACtC,EAAE,IAAI,OAAO,CAAC,CAAC,EAAEs4D,GAAG,EAAE14D,EAAEI,CAAC,EAAEA,EAAE,IAAI,OAAO,CAAC,CAAC,EAAE,IAAI2B,EAAEjC,EAAE,OAAO,QAAQ4C,EAAE,EAAEA,EAAEX,EAAE,OAAO,EAAEW,EAAErC,EAAEs4D,GAAG52D,EAAEW,GAAG1C,EAAEI,CAAC,EAAEw4D,GAAG72D,EAAEW,GAAG1C,EAAE,EAAEI,CAAC,EAAEA,GAAGsC,IAAIX,EAAE,OAAO,EAAE,IAAI,KAAK,OAAO,CAAC,CAAC,EAAEjC,EAAE,iCAAiC,EAAE,IAAIkC,EAAE62D,GAAG/4D,CAAC,EAAEmC,EAAEuqD,GAAG1sD,EAAE,mBAAmB,EAAEM,EAAE,iBAAiB4B,EAAEC,GAAG,EAAE7B,EAAE,qBAAqB4B,GAAG,EAAE5B,EAAE,yBAAyB6B,GAAG,EAAE7B,EAAE,IAAI,OAAO,CAAC,CAAC,CAAC,CAAC,SAASy4D,GAAG/4D,EAAE,CAAC,IAAI,EAAE,OAAOA,EAAE,2BAA2B,KAAK,EAAE0sD,GAAG1sD,EAAE,yBAAyB,EAAE,EAAE0sD,GAAG1sD,EAAE,gBAAgB,EAAE,CAAC,CAAC,SAAS24D,GAAG34D,EAAE,CAAC,IAAI,EAAE,GAAGE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAE,QAAQC,KAAKP,EAAE,aAAaE,EAAE,KAAKF,EAAE,aAAaO,EAAE,EAAE,QAAQA,KAAKL,EAAE,CAAC,GAAGK,EAAE,OAAO,GAAGA,EAAE,SAAS,GAAGA,EAAE,GAAG,cAAc,OAAO,EAAE,CAAC,EAAE,GAAG,KAAK,CAACD,EAAE,KAAK,GAAGC,CAAC,CAAC,CAAC,GAAG,EAAE,QAAQA,KAAKP,EAAE,OAAO,CAAC,IAAI,EAAE,GAAG,QAAQ,KAAKO,EAAE,aAAa,GAAGD,EAAE,QAAQ,CAAC,IAAI,GAAG,GAAG,EAAE,CAAC,EAAE,GAAG,KAAK,MAAM,EAAE,GAAG,GAAG,CAAC,EAAE,KAAK,CAAC,OAAO,CAAC,CAAC,SAASs4D,GAAG54D,EAAE,EAAEE,EAAE,QAAQ,IAAI,CAAC,IAAII,EAAE,GAAG,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAEO,EAAEA,EAAE,IAAID,EAAEA,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,EAAE,KAAKA,GAAGN,EAAEO,GAAGD,EAAEA,EAAE,MAAM,EAAE,EAAEC,EAAE,EAAED,GAAG,IAAI,OAAO,EAAEC,GAAGD,EAAE,MAAM,EAAEJ,EAAEI,CAAC,CAAC,CAAC,SAASu4D,GAAG74D,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEC,EAAE,GAAG,CAACA,EAAEP,EAAE,aAAa,IAAIkC,GAAG,KAAK,UAAUA,EAAE,WAAW,CAAC,EAAE,KAAK,GAAG,CAAC,OAAOA,EAAN,CAAS3B,EAAE,UAAU,CAAC,GAAG,CAACD,EAAE,KAAK,UAAUN,EAAE,WAAW,CAAC,OAAOkC,EAAN,CAAS5B,EAAE,UAAU,CAAC,IAAI,EAAEN,EAAE,KAAK,EAAEA,EAAE,aAAa,EAAEiC,EAAE,CAAC,GAAG,MAAM,KAAK1B,EAAED,EAAEN,EAAE,YAAY,EAAE,SAAS,CAAC,EAAE44D,GAAG32D,EAAE,EAAE/B,CAAC,CAAC,CAAC,SAAS44D,GAAG94D,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,EAAE,GAAG,CAAC,EAAEP,EAAE,aAAa,IAAI6C,GAAG,KAAK,UAAUA,EAAE,WAAW,CAAC,EAAE,KAAK,GAAG,CAAC,OAAOA,EAAN,CAAS,EAAE,UAAU,CAAC,GAAG,CAACtC,EAAE,KAAK,UAAUP,EAAE,WAAW,CAAC,OAAO6C,EAAN,CAAStC,EAAE,UAAU,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQsC,KAAK7C,EAAE,aAAa,GAAG,EAAEE,GAAG,MAAMA,EAAE,OAAO,GAAGA,EAAE,QAAQ2C,CAAC,IAAI,IAAI,QAAQC,EAAE,EAAEA,EAAED,EAAE,cAAc,OAAO,EAAEC,EAAE,CAAC,IAAIC,EAAEF,EAAE,cAAcC,GAAG,KAAKE,EAAEH,EAAE,YAAYC,GAAGoB,EAAErB,EAAE,cAAcC,GAAG,EAAE,KAAK,GAAGC,KAAKC,MAAMkB,IAAI,CAAC,CAAC,IAAIjC,EAAEjC,EAAE,KAAKkC,EAAElC,EAAE,aAAa,EAAEmC,EAAE,EAAE,SAAS,EAAE,GAAG,EAAE,GAAGS,EAAE,CAAC,GAAGX,MAAMC,KAAK,EAAE3B,EAAEP,EAAE,YAAY,EAAE,SAAS,EAAEmC,CAAC,EAAEy2D,GAAGh2D,EAAE,EAAEtC,CAAC,EAAE,QAAQuC,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE+1D,GAAG,CAAC,GAAG,GAAG,GAAG,GAAG,EAAE/1D,EAAE,EAAE,EAAEvC,CAAC,CAAC,CAAC,SAAS04D,GAAGh5D,EAAE,EAAEE,EAAE,CAAC,OAAOF,IAAI,gBAAgBA,IAAI,gBAAgBA,IAAI,gBAAgB,IAAI,GAAG,OAAOE,GAAG,QAAQ,CAAC,SAAS+4D,GAAGj5D,EAAE,EAAE,CAAC,GAAGA,IAAI,KAAK,OAAO,KAAK,GAAG,OAAOA,GAAG,SAAS,OAAO0mD,GAAG1mD,CAAC,EAAE,GAAG,OAAOA,GAAG,UAAU,OAAOA,GAAG,UAAU,OAAOA,EAAE,GAAGA,aAAa,MAAM,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAEN,EAAE,OAAO,QAAQO,EAAE,EAAEA,EAAED,EAAE,EAAEC,EAAE,CAAC,IAAI,EAAEP,EAAEO,GAAGy4D,GAAG,EAAEz4D,EAAE,CAAC,EAAEL,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAK+4D,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO/4D,CAAC,KAAK,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQI,KAAK,OAAO,KAAKN,CAAC,EAAE,CAAC,IAAIO,EAAEP,EAAEM,GAAG,GAAGA,IAAI,QAAQ,OAAOC,GAAG,SAASL,EAAEI,GAAGC,MAAM,CAAC,IAAI,EAAEmmD,GAAGpmD,CAAC,EAAEJ,EAAE,GAAG+4D,GAAG14D,EAAE,CAAC,CAAC,CAAC,CAAC,OAAOL,CAAC,CAAC,CAAC,SAASg5D,GAAGl5D,EAAE,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,KAAK,GAAG,OAAOA,GAAG,SAAS,OAAOymD,GAAGzmD,CAAC,EAAE,GAAG,OAAOA,GAAG,UAAU,OAAOA,GAAG,UAAU,OAAOA,EAAE,GAAGA,aAAa,MAAM,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAEN,EAAE,OAAO,QAAQO,EAAE,EAAEA,EAAED,EAAE,EAAEC,EAAE,CAAC,IAAI,EAAEP,EAAEO,GAAGy4D,GAAG,EAAEz4D,EAAE,CAAC,EAAEL,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAKg5D,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,OAAOh5D,CAAC,KAAK,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQI,KAAK,OAAO,KAAKN,CAAC,EAAE,CAAC,IAAIO,EAAEP,EAAEM,GAAG,EAAEmmD,GAAGnmD,CAAC,GAAGA,IAAI,QAAQA,IAAI,cAAc,OAAOC,GAAG,SAASL,EAAE,GAAGK,EAAEL,EAAE,GAAGg5D,GAAG34D,EAAED,CAAC,CAAC,CAAC,OAAOJ,CAAC,CAAC,CAAC,IAAIi5D,GAAG,QAAYC,GAAG,cAAc/L,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAM,CAAC,CAAC,EAAE,KAAK,eAAe,IAAI,IAAI,KAAK,KAAKA,EAAE,KAAK,KAAK,MAAM,KAAK,CAAC,IAAIiE,EAAE,KAAK,aAAa,EAAE,YAAY,EAAE,KAAK,KAAK6iD,GAAG7iD,CAAC,CAAC,CAAC,GAAG,KAAK,gBAAgB,GAAG,KAAK,WAAW,GAAG,MAAM,QAAQjE,EAAE,MAAM,EAAE,KAAK,OAAOA,EAAE,OAAO,MAAM,EAAE,KAAK,OAAO,CAACA,EAAE,MAAM,EAAE,MAAM,QAAQA,EAAE,OAAO,EAAE,KAAK,QAAQA,EAAE,QAAQ,MAAM,EAAE,KAAK,QAAQ,CAACA,EAAE,OAAO,EAAEmmD,GAAG,KAAK,MAAM,EAAE,SAAS,KAAK,OAAO,OAAO,MAAM,IAAIjB,EAAE,mGAAmG,KAAK,OAAO,IAAIjhD,GAAGA,EAAE,IAAI,GAAG,EAAEkiD,GAAG,KAAK,OAAO,EAAE,SAAS,KAAK,QAAQ,QAAQ,QAAQ,KAAK,qGAAqG,KAAK,QAAQ,IAAIliD,GAAGA,EAAE,IAAI,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,uBAAuB,CAAC,EAAE,KAAK,yBAAyB,CAAC,EAAE,KAAK,aAAa,CAAC,EAAE,KAAK,wBAAwB,CAAC,EAAE,KAAK,0BAA0B,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,KAAK,sBAAsB,CAAC,EAAE,QAAQA,KAAK,KAAK,QAAQ,CAAC,IAAIJ,EAAEI,EAAE,YAAYH,EAAEG,EAAE,UAAUF,EAAEE,EAAE,YAAY,KAAK,aAAa,KAAKJ,CAAC,EAAE,KAAK,wBAAwB,KAAKC,CAAC,EAAE,KAAK,0BAA0B,KAAKC,CAAC,CAAC,CAAC,QAAQE,KAAK,KAAK,OAAO,CAAC,IAAIJ,EAAEI,EAAE,YAAYH,EAAEG,EAAE,UAAUF,EAAEE,EAAE,YAAYshD,GAAGzhD,IAAI,EAAE,0BAA0B,EAAEyhD,GAAGxhD,IAAI,EAAE,4BAA4B,EAAE,KAAK,YAAY,KAAKF,CAAC,EAAE,KAAK,uBAAuB,KAAKC,CAAC,EAAE,KAAK,yBAAyB,KAAKC,CAAC,CAAC,CAAC,KAAK,WAAW,CAAC,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,gBAAgB,CAAC,EAAE,KAAK,eAAe,CAAC,EAAE,KAAK,gBAAgB,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAE,KAAK,YAAY,OAAOA,IAAI,CAAC,IAAIJ,EAAE,KAAK,YAAYI,GAAG,GAAG,EAAEJ,aAAa8oD,IAAI,MAAM,IAAI,UAAU,8EAA8E3sD,EAAE,iBAAiBiE,0CAA0CJ,EAAE,aAAa,IAAI,EAAE,KAAK,WAAW,KAAKA,EAAE,IAAI,EAAE,KAAK,gBAAgB,KAAKA,EAAE,eAAe,EAAE,KAAK,eAAe,KAAKA,EAAE,IAAI,CAAC,CAAC,QAAQI,KAAK,KAAK,aAAa,KAAK,YAAY,KAAKA,EAAE,IAAI,EAAE,KAAK,oBAAoB,KAAK,OAAO,IAAIA,GAAGA,EAAE,KAAK,EAAE,KAAK,qBAAqB,KAAK,QAAQ,IAAIA,GAAGA,EAAE,KAAK,EAAE,IAAI7E,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEK,EAAE,CAAC,EAAEwB,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC6C,EAAEJ,EAAEC,EAAEC,EAAEC,EAAE0C,IAAI,EAAE3C,GAAG,MAAMC,GAAG,MAAM0C,GAAG,QAAQ3C,EAAEE,EAAE,YAAYD,EAAEC,EAAE,UAAUyC,EAAEzC,EAAE,aAAa,IAAIC,EAAEH,EAAE,aAAaC,GAAG,GAAGF,EAAE,QAAQI,CAAC,IAAI,GAAG,MAAM,IAAI+gD,GAAG,cAAchhD,EAAE,kBAAkBF,EAAE,2BAA2B,EAAE,GAAGF,EAAE,QAAQK,CAAC,IAAI,GAAG,OAAO,KAAK,eAAe,IAAIo0D,GAAG,QAAQv0D,EAAEC,CAAC,CAAC,EAAED,EAAE,MAAM7C,IAAIA,EAAE6C,EAAE,IAAI,OAAO,KAAK7C,CAAC,EAAE,QAAQ4C,EAAE,QAAQI,CAAC,IAAI,IAAIJ,EAAE,KAAKI,CAAC,EAAE,IAAIyC,EAAEzC,EAAE,cAAc,OAAO,QAAQC,EAAE,EAAEA,EAAEwC,EAAExC,IAAI,CAAC,IAAIC,EAAEF,EAAE,aAAaC,GAAGE,EAAEH,EAAE,cAAcC,GAAGG,EAAEJ,EAAE,YAAYC,GAAGI,EAAEL,EAAE,cAAcC,GAAG/C,EAAEgD,EAAEP,EAAEC,EAAEO,EAAEC,EAAEC,CAAC,CAAC,CAAC,IAAIV,EAAE,KAAKK,CAAC,EAAEJ,EAAE,QAAQI,CAAC,GAAG,GAAGJ,EAAE,OAAOA,EAAE,QAAQI,CAAC,EAAE,CAAC,EAAE/C,EAAE,KAAK+C,CAAC,CAAC,EAAE,EAAE,CAAC,EAAEpC,EAAE,CAAC,EAAE,QAAQmC,KAAK,KAAK,QAAQ7C,EAAE6C,EAAE,EAAEnC,CAAC,EAAE,IAAI,EAAEX,EAAE,MAAM,EAAE,QAAQ,EAAE,QAAQ8C,KAAK,EAAE,CAAC,EAAEA,EAAE,IAAIA,EAAEA,EAAE,MAAM7E,IAAIA,EAAE6E,EAAE,IAAI,GAAG,IAAIJ,EAAEzE,EAAE6E,EAAE,IAAIH,EAAErE,EAAEwE,EAAE,cAAc,KAAK,KAAK,EAAExE,EAAEwE,EAAE,cAAc,IAAIJ,EAAE,KAAK,IAAIA,EAAEC,CAAC,EAAErE,EAAEwE,EAAE,cAAc,IAAIJ,EAAE5C,EAAEgD,EAAE,cAAc,IAAIA,EAAE,cAAc7E,EAAE6E,EAAE,IAAIJ,EAAE,QAAQE,EAAE,EAAEA,EAAEE,EAAE,cAAc,OAAOF,IAAI,CAAC,IAAIC,EAAEC,EAAE,cAAcF,GAAG2C,EAAEzC,EAAE,YAAYF,GAAGG,EAAEF,EAAE,aAAa0C,GAAGC,EAAEvH,EAAE8E,EAAE,KAAK,KAAK,EAAE9E,EAAE8E,EAAE,IAAI9E,EAAE8E,EAAE,IAAI,KAAK,IAAIL,EAAE,EAAE8C,CAAC,EAAE,EAAEzC,EAAE,IAAIA,CAAC,CAAC,CAAC,IAAIlC,EAAE,CAAC,EAAE,QAAQiC,KAAK7E,EAAE,CAAC,IAAIyE,EAAEzE,EAAE6E,GAAGJ,KAAK7B,IAAIA,EAAE6B,GAAG,CAAC,GAAG7B,EAAE6B,GAAG,KAAK,EAAEI,EAAE,CAAC,CAAC,IAAIhC,EAAE,CAAC,EAAE,QAAQgC,KAAKxE,EAAE,CAAC,IAAIoE,EAAEpE,EAAEwE,GAAGJ,KAAK5B,IAAIA,EAAE4B,GAAG,CAAC,GAAG5B,EAAE4B,GAAG,KAAK5C,EAAEgD,EAAE,CAAC,CAAC,IAAI/B,EAAE,OAAO,KAAKD,CAAC,EAAE,IAAIgC,GAAG,SAASA,EAAE,EAAE,CAAC,EAAE,KAAKiiD,EAAE,EAAE,KAAK,OAAO,CAAC,EAAE,QAAQjiD,KAAK/B,EAAE,CAAC,IAAI2B,EAAE5B,EAAEgC,GAAGJ,EAAE,KAAK,CAACC,EAAEC,IAAI,CAAC,IAAIC,EAAE9C,EAAE4C,EAAE,IAAI4C,EAAExF,EAAE6C,EAAE,IAAI,OAAOC,EAAE0C,EAAE,GAAG1C,EAAE0C,EAAE,EAAE,CAAC,CAAC,EAAE,QAAQ5C,KAAKD,EAAEC,aAAaw0D,IAAI,KAAK,sBAAsB,KAAKx0D,CAAC,EAAE,KAAK,OAAO,KAAKA,CAAC,CAAC,CAAC,KAAK,cAAc7B,EAAEC,EAAE,OAAO,KAAKF,CAAC,EAAE,IAAIiC,GAAG,SAASA,EAAE,EAAE,CAAC,EAAE,KAAKiiD,EAAE,EAAE,IAAI9iD,EAAE,KAAK,OAAO,MAAM,EAAEC,EAAE,CAAC,EAAE,QAAQY,KAAK/B,EAAE,QAAQ2B,KAAK7B,EAAEiC,GAAG,CAAC,IAAIH,EAAED,EAAE,cAAc,GAAGC,GAAG,KAAK,CAAC,QAAQC,KAAKF,EAAE,aAAa,GAAGT,EAAE,QAAQW,CAAC,IAAI,GAAG,MAAM,IAAIkhD,GAAG,sDAAsDlhD,eAAeD,EAAE,qEAAqET,GAAG,EAAE,QAAQU,KAAKF,EAAE,cAAcT,EAAE,KAAKW,CAAC,EAAEV,EAAE,KAAKS,EAAE,IAAI,CAAC,CAAC,CAAC,KAAK,aAAa9B,EAAE,IAAIyE,EAAE,KAAK,OAAO,IAAIxC,GAAGA,EAAE,IAAI,EAAE,QAAQA,KAAKwC,EAAE,CAAC,IAAI5C,EAAE4C,EAAE,OAAO3C,GAAGA,IAAIG,CAAC,EAAE,OAAO,GAAGJ,IAAI,EAAE,MAAM,IAAIohD,GAAG,aAAahhD,cAAcJ,wEAAwE,KAAK,UAAU4C,CAAC,CAAC,CAAC,CAAC,KAAK,cAAc,CAAC,EAAE,KAAK,aAAa,CAAC,EAAE,IAAI4lD,GAAG,CAAC,cAAc,KAAK,cAAc,CAAC,EAAE,YAAY,CAAC,EAAE,cAAc,CAAC,EAAE,aAAa,KAAK,OAAO,cAAc,KAAK,QAAQ,WAAW,KAAK,OAAO,IAAIpoD,GAAG,IAAI,EAAE,YAAY,KAAK,QAAQ,IAAIA,GAAG,IAAI,EAAE,YAAY,KAAK,OAAO,IAAIA,GAAGA,EAAE,KAAK,EAAE,aAAa,KAAK,QAAQ,IAAIA,GAAGA,EAAE,KAAK,CAAC,CAAC,EAAE,KAAK,MAAM,GAAG,KAAK,UAAU,CAAC,CAAC,mBAAmB,CAAC,GAAG,KAAK,YAAY,EAAE,MAAM,IAAI,MAAM,cAAc,KAAK,4BAA4B,CAAC,CAAC,SAAS,CAAC,KAAK,kBAAkB,EAAE,IAAIjE,EAAE,CAAC,qBAAqB,KAAK,qBAAqB,CAAC,EAAE,GAAG,EAAE,KAAK,YAAY,EAAE,CAAC,QAAQZ,KAAK,KAAK,OAAOY,EAAE,sBAAsBZ,EAAE,QAAQ,EAAE,qBAAqB,QAAQA,KAAK,KAAK,sBAAsBY,EAAE,sBAAsBZ,EAAE,QAAQ,EAAE,oBAAoB,CAAC,OAAOY,EAAE,qBAAqB,KAAK,UAAUA,CAAC,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,OAAO,QAAQZ,GAAG,CAACA,EAAE,kBAAkB,QAAQ,GAAG,EAAE,UAAUY,CAAC,CAAC,CAAC,EAAE,KAAK,WAAWA,CAAC,CAAC,IAAI,kBAAkB,CAAC,GAAG,KAAK,kBAAkB,OAAO,EAAE,MAAM,IAAIklD,EAAE,sNAAsN,EAAE,GAAG,CAAC,KAAK,UAAU,MAAM,CAAC,EAAE,IAAIllD,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,OAAOY,EAAEA,EAAE,OAAOZ,EAAE,gBAAgB,EAAE,OAAOY,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,OAAOY,EAAE,KAAK,GAAGZ,EAAE,mBAAmB,EAAE,GAAG,CAAC,KAAK,UAAU,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQ,KAAK,KAAK,OAAOA,EAAE,KAAK,GAAG,EAAE,gBAAgB,EAAE,OAAOA,EAAE,OAAOY,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,iBAAiB,OAAO,KAAK,mBAAmB,CAAC,CAAC,YAAYA,EAAEZ,EAAE,GAAG,CAAC,IAAI,EAAE,CAAC,EAAEK,EAAE,EAAE,QAAQyB,KAAK,KAAK,OAAO,QAAQC,KAAKD,EAAE,QAAQ,CAAC,GAAG,EAAEC,EAAE,eAAe,KAAK,MAAM,IAAI+jD,EAAE,0BAA0B/jD,EAAE,cAAc,EAAE,EAAEA,EAAE,cAAcA,EAAE1B,GAAG,CAAC,IAAIwB,EAAE,CAAC,EAAE,QAAQC,KAAKlB,EAAE,CAAC,IAAImB,EAAED,EAAE,GAAG,EAAEA,IAAI,KAAK,CAAC,IAAIE,EAAEF,EAAE,MAAM,GAAG,EAAEC,EAAEC,EAAE,MAAM,EAAE,EAAE,EAAE,OAAO,CAACA,EAAEA,EAAE,OAAO,EAAE,CAAC,EAAE,KAAK,GAAG,CAAC,CAAC,GAAG,EAAED,IAAI,KAAKF,EAAE,KAAK,CAAC,EAAEE,GAAGnB,EAAEkB,EAAE,CAAC,UAAU9B,EAAE,MAAM,IAAI8lD,EAAE,gDAAgDhkD,GAAG,EAAE,OAAO,EAAEC,EAAE,CAAC,GAAG/B,EAAE,CAAC,IAAI8B,EAAE,CAAC,EAAE,QAAQC,KAAK,EAAED,EAAE,KAAKC,CAAC,EAAE,GAAGD,EAAE,OAAO,EAAE,MAAM,IAAIgkD,EAAE,GAAGhkD,EAAE,aAAazB,0BAA0ByB,GAAG,CAAC,CAAC+qD,GAAGhrD,CAAC,CAAC,CAAC,eAAe,CAAC,IAAIjB,EAAE,KAAK,UAAU,EAAEZ,EAAE,CAAC,EAAE,OAAOA,EAAE,UAAU,KAAK,aAAa,EAAEA,EAAE,OAAOY,EAAEZ,EAAE,aAAa,eAAei5D,KAAKj5D,EAAE,QAAQ,gBAAgBA,CAAC,CAAC,OAAOY,EAAEZ,EAAE,GAAG,CAAC,IAAI,EAAEg5D,GAAG,KAAK,cAAc,CAAC,EAAE,OAAOh5D,EAAE,KAAK,UAAU,CAAC,EAAE,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0lD,GAAG1lD,CAAC,EAAE,IAAI,EAAE,IAAI8sD,GAAG,QAAQrtD,EAAE,EAAEA,EAAE,KAAK,OAAO,OAAO,EAAEA,EAAE,EAAE,IAAI,KAAK,OAAOA,GAAGO,EAAEP,EAAE,EAAE,OAAOytD,GAAG,KAAK,QAAQ,EAAE9tD,CAAC,CAAC,CAAC,CAAC,CAAC,YAAYY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0lD,GAAG1lD,CAAC,EAAE,IAAI,EAAE,OAAOZ,GAAG,KAAK,EAAEkmD,GAAG,KAAKtlD,EAAE,MAAM,EAAE,EAAE0lD,GAAGtmD,CAAC,EAAE,KAAK,iBAAiBY,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,IAAIZ,EAAEqsD,GAAGzrD,CAAC,EAAE,GAAGZ,EAAE,SAAS,KAAK,YAAY,OAAO,MAAM,IAAI8lD,EAAE,+BAA+BllD,gBAAgB,KAAK,YAAY,uBAAuB,EAAE,IAAI,EAAE,CAAC,EAAE,QAAQmB,EAAE,EAAEA,EAAE/B,EAAE,OAAO+B,IAAI,CAAC,IAAIC,EAAE,KAAK,YAAYD,GAAG,EAAE/B,EAAE+B,GAAGW,EAAEV,EAAE,KAAK,OAAO,EAAEU,GAAG,CAAC,CAAC,IAAIrC,EAAE,OAAO,KAAK,KAAK,YAAY,EAAE,IAAI0B,GAAG,SAASA,EAAE,EAAE,CAAC,EAAE,KAAK+kD,EAAE,EAAE,GAAGzmD,EAAE,OAAO,EAAE,QAAQ0B,KAAK1B,EAAE,CAAC,IAAI2B,EAAE,KAAK,aAAaD,GAAG,QAAQ,KAAKC,EAAE,CAAC,IAAIU,EAAE,EAAE,cAAc,GAAG,KAAK,YAAY,IAAIsB,GAAGA,EAAE,EAAE,EAAE,QAAQtB,EAAE,EAAE,IAAI,GAAG,SAAS,IAAI,EAAE,CAAC,EAAE,QAAQsB,EAAE,EAAEA,EAAE,EAAE,cAAc,OAAOA,IAAI,CAAC,IAAIC,EAAE,EAAE,cAAcD,GAAGqD,EAAE,EAAE,YAAYrD,GAAGa,EAAE,EAAE,cAAcb,GAAGS,EAAE,GAAGR,EAAE,QAAQoD,KAAKxC,IAAIH,EAAE,EAAED,GAAG,EAAE,KAAKC,CAAC,CAAC,CAAC,IAAI9B,EAAEF,EAAE,mBAAmB2jD,GAAG,CAAC,CAAC,EAAExjD,EAAEwpD,GAAGzpD,CAAC,EAAEE,EAAEJ,EAAE,aAAa,QAAQ,CAAC,EAAE,QAAQsB,EAAE,EAAEA,EAAEnB,EAAE,OAAOmB,IAAI,CAAC,IAAIC,EAAE,GAAGvB,EAAE,QAAQI,KAAKkB,IAAI,EAAEC,GAAGpB,EAAEmB,EAAE,CAAC,CAAC,CAAC,IAAInC,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,KAAK,aAAa,OAAOA,IAAI,CAAC,IAAIC,EAAE,KAAK,aAAaD,GAAG,EAAE,KAAK,wBAAwBA,GAAGW,EAAE,KAAK,0BAA0BX,GAAG,EAAE,GAAGC,EAAE,QAAQ,KAAKU,IAAIZ,EAAE,KAAK,CAAC,CAAC,CAAC,QAAQC,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAI,CAAC,IAAIC,EAAEF,EAAEC,GAAGokD,GAAGnkD,KAAK,CAAC,EAAEH,EAAE,KAAK,EAAEG,EAAE,CAAC,CAAC,OAAOqkD,GAAGxkD,CAAC,CAAC,CAAC,iBAAiBjB,EAAEZ,EAAE,CAACA,GAAG,OAAOA,EAAEkmD,GAAG,KAAKtlD,EAAE,MAAM,GAAG,IAAI,EAAE,CAAC,EAAE,QAAQoB,EAAE,EAAEA,EAAE,KAAK,OAAO,OAAO,EAAEA,EAAE,CAAC,IAAI,EAAE,KAAK,OAAOA,GAAGU,EAAE9B,EAAEoB,GAAG,EAAEhC,EAAEgC,GAAG,EAAE,EAAE,IAAI,CAACU,EAAE,CAAC,CAAC,CAAC,IAAIrC,EAAE,OAAO,KAAK,KAAK,YAAY,EAAE,IAAI2B,GAAG,SAASA,EAAE,EAAE,CAAC,EAAE,KAAK8kD,EAAE,EAAE,QAAQ9kD,KAAK3B,EAAE,CAAC,IAAI,EAAE,KAAK,aAAa2B,GAAG,QAAQU,KAAK,EAAE,CAAC,IAAI,EAAEA,EAAE,cAAcE,EAAEF,EAAE,aAAaG,EAAEH,EAAE,cAAcI,EAAE,IAAI,MAAM,QAAQkB,KAAKpB,EAAEoB,EAAE,MAAM,GAAGlB,EAAE,KAAK,EAAEkB,EAAE,GAAG,EAAE,GAAGlB,EAAE,SAASF,EAAE,OAAO,CAAC,IAAIoB,EAAE,CAAC,EAAEC,EAAEoD,EAAExC,EAAEJ,EAAE,GAAG/B,EAAE,UAAU,OAAOsB,EAAEtB,EAAE,UAAUI,EAAE,SAAS,EAAE,CAAC,GAAG,CAAC4B,EAAEC,CAAC,EAAE7B,EAAE,GAAGkB,EAAE,MAAM,OAAOA,EAAE,KAAKW,GAAGE,EAAEyhD,GAAG,EAAE,KAAK5hD,EAAEV,CAAC,CAAC,EAAES,EAAE6hD,GAAG,EAAE,YAAY5hD,EAAEC,CAAC,CAAC,EAAEV,EAAE,CAACS,CAAC,EAAE2C,EAAE,CAAC1C,CAAC,CAAC,MAAMV,EAAEnB,EAAE,IAAI4B,GAAGA,EAAE,EAAE,EAAE2C,EAAEvE,EAAE,IAAI4B,GAAGA,EAAE,EAAE,EAAEV,EAAE,MAAM,OAAOA,EAAE,KAAKqD,GAAGxC,EAAEyhD,GAAG,EAAE,KAAKriD,EAAED,CAAC,CAAC,EAAES,EAAE6hD,GAAG,EAAE,YAAYriD,EAAEoD,CAAC,CAAC,EAAE,GAAG,EAAE,oBAAoB,MAAM,IAAI0+C,GAAG,uHAAuH,EAAE,QAAQrhD,EAAE,EAAEA,EAAE7B,EAAE,OAAO,EAAE6B,EAAE,CAAC,IAAIC,EAAE9B,EAAE6B,GAAGE,EAAEC,EAAEH,GAAG4C,EAAE7C,EAAEC,GAAG,EAAEC,EAAE,IAAI,CAACC,EAAE0C,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIzF,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQC,KAAK,KAAK,QAAQ,CAACmkD,GAAGnkD,EAAE,MAAM,EAAE,4BAA4BA,EAAE,UAAUA,EAAE,IAAI,EAAE,GAAG,CAAC,EAAEU,CAAC,EAAE,EAAEV,EAAE,IAAID,EAAE,KAAK,EAAE,KAAK,EAAEF,EAAE,KAAK,CAAC,EAAEC,EAAE,KAAKY,CAAC,CAAC,CAAC,MAAM,CAACb,EAAEC,EAAEC,CAAC,CAAC,CAAC,uBAAuBnB,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,EAAE,QAAQK,KAAK,KAAK,OAAO,CAAC,EAAEA,aAAa64D,GAAG,EAAE,EAAE,QAAQr3D,EAAE,EAAEA,EAAExB,EAAE,aAAa,OAAOwB,IAAI,CAAC,IAAIC,EAAEo3D,GAAG,QAAQ74D,EAAEwB,CAAC,EAAE,KAAK,eAAe,IAAIC,CAAC,IAAI9B,EAAE8B,GAAG,EAAE,GAAG,EAAE,CAAC,CAAC,OAAO9B,CAAC,CAAC,SAASY,EAAEZ,EAAE,CAAC,GAAGA,GAAG,KAAK,CAAC,GAAG,KAAK,OAAO,QAAQA,EAAE,MAAM,IAAI8lD,EAAE,wCAAwC9lD,yBAAyB,KAAK,OAAO,kBAAkB,EAAE,OAAO,KAAK,OAAOA,EAAE,SAASY,GAAG,KAAK,MAAM,IAAIklD,EAAE,4CAA4C,EAAE,QAAQ,KAAK,KAAK,OAAO,GAAG,EAAE,OAAOllD,EAAE,OAAO,EAAE,MAAM,IAAIklD,EAAE,kBAAkBllD,GAAG,CAAC,CAAC,iBAAiB,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAI3yB,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,OAAO,QAAQ,EAAE,EAAE,EAAEA,EAAE,aAAa,OAAO,EAAE,EAAE,CAAC,IAAIK,EAAE64D,GAAG,QAAQl5D,EAAE,CAAC,EAAE,KAAK,eAAe,IAAIK,CAAC,GAAGO,EAAE,KAAK,GAAGZ,EAAE,gBAAgB,CAAC,CAAC,CAAC,OAAOY,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,KAAK,KAAK,IAAI,EAAEZ,EAAE,KAAK,uBAAuB,KAAK,MAAM,EAAE,EAAE,CAAC,EAAE,QAAQ8B,KAAK,KAAK,OAAO,CAAC,IAAIC,EAAED,EAAE,aAAa,EAAEE,EAAEF,EAAE,UAAU,EAAE,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,aAAa,OAAO,IAAI,CAAC,IAAIc,EAAEd,EAAE,aAAa,GAAGe,EAAEq2D,GAAG,QAAQp3D,EAAE,CAAC,EAAEgB,EAAE,CAAC,EAAE,GAAG,KAAK,eAAe,IAAID,CAAC,EAAE,CAAC,GAAGD,EAAE,SAAS,GAAG,CAAC,KAAK,UAAUA,EAAE,QAAQ,EAAEE,EAAEF,EAAE,QAAQ,OAAOoB,EAAN,CAAS,QAAQ,KAAK,SAASlC,EAAE,uDAAuDc,EAAE,iHAAiH,EAAEE,EAAE,CAAC,CAAC,CAAC,GAAGF,EAAE,cAAc,OAAO,EAAE,CAAC,IAAIoB,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAErB,EAAE,cAAc,OAAOqB,IAAI,CAAC,IAAIoD,EAAEzE,EAAE,cAAcqB,GAAGY,EAAEjC,EAAE,YAAYqB,GAAGQ,EAAE7B,EAAE,cAAcqB,GAAGS,EAAEw0D,GAAG,QAAQ7xD,EAAExC,CAAC,EAAEF,EAAE3E,EAAE0E,GAAGC,GAAG,OAAOA,EAAE,GAAGX,EAAE,KAAK,CAACqD,EAAE,KAAK1C,EAAEF,EAAE3B,CAAC,CAAC,CAAC,CAAC,EAAE,KAAKkB,CAAC,CAAC,CAAC,CAAC,CAAC,IAAItB,EAAE,CAAC,EAAEA,EAAE,KAAKZ,EAAE,KAAKY,EAAE,UAAUX,EAAEW,EAAE,OAAOV,EAAEU,EAAE,aAAa,EAAE,EAAE,KAAKA,CAAC,CAAC,CAAC9B,EAAE,OAAO,EAAE,IAAIP,EAAE,CAAC,EAAE,QAAQyB,EAAE,EAAEA,EAAE,KAAK,YAAY,OAAOA,IAAI,CAAC,IAAIC,EAAE,KAAK,YAAYD,GAAGE,EAAE,KAAK,uBAAuBF,GAAG,EAAEo3D,GAAG,QAAQn3D,EAAEC,CAAC,EAAE,GAAG,CAAC,KAAK,eAAe,IAAI,CAAC,EAAE,SAAS,IAAIU,EAAE1C,EAAE,GAAG0C,GAAG,OAAOA,EAAE,GAAG,IAAI,EAAE,KAAK,yBAAyBZ,GAAGzB,EAAE,KAAK,CAAC0B,EAAE,KAAKW,EAAE,CAAC,CAAC,CAAC,CAAC9B,EAAE,YAAYP,EAAE,IAAIwB,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,KAAK,aAAa,OAAOA,IAAI,CAAC,IAAIC,EAAE,KAAK,aAAaD,GAAGE,EAAE,KAAK,wBAAwBF,GAAG,EAAEo3D,GAAG,QAAQn3D,EAAEC,CAAC,EAAE,GAAG,CAAC,KAAK,eAAe,IAAI,CAAC,EAAE,SAAS,IAAIU,EAAE1C,EAAE,GAAG0C,GAAG,OAAOA,EAAE,GAAG,IAAI,EAAE,KAAK,0BAA0BZ,GAAGD,EAAE,KAAK,CAACE,EAAE,KAAKW,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO9B,EAAE,aAAaiB,EAAEjB,CAAC,CAAC,OAAO,WAAWA,EAAEZ,EAAE,EAAE,CAAC,EAAEK,EAAE,GAAG,CAAC,IAAIwB,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,SAASC,EAAEkC,EAAEoD,EAAE,CAACpD,EAAE,QAAQnC,EAAEA,EAAEmC,EAAE,MAAM,KAAKoD,CAAC,EAAEvF,EAAEmC,EAAE,MAAM,CAACoD,CAAC,CAAC,CAAC,SAASrF,EAAEiC,EAAEoD,EAAE,CAAC,IAAIxC,EAAE,CAAC,EAAEJ,EAAE,QAAQC,KAAK2C,EAAE,CAAC,IAAI1C,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAG4C,EAAE5C,EAAE,GAAG,GAAGD,EAAEC,EAAE,IAAI,KAAK,CAAC,EAAEA,EAAE,GAAG,EAAEC,KAAK9C,GAAG,CAACE,EAAEkC,EAAEoD,CAAC,EAAE,MAAM,CAAC,IAAIvC,EAAEjD,EAAE8C,GAAG,GAAGG,EAAE,aAAa,QAAQF,EAAE,CAAC7C,EAAEkC,EAAEoD,CAAC,EAAE,MAAM,CAAC,IAAIE,EAAEzC,EAAE,aAAaF,GAAGC,EAAE,KAAK0C,EAAE,cAAcD,EAAE,CAAC,CAACzC,EAAE,OAAO,GAAGZ,EAAE,MAAMoiD,GAAGxhD,CAAC,EAAEJ,CAAC,CAAC,CAAC,SAAS,EAAER,EAAE,CAAC,IAAIoD,EAAEpD,EAAE,KAAKY,EAAEgxD,GAAG5xD,EAAEjE,EAAE,eAAe,KAAKA,EAAE,cAAc,CAAC,CAAC,EAAE6E,EAAE,6BAA6BxE,CAAC,EAAEwB,EAAEwF,GAAGxC,EAAEZ,EAAE,aAAa,QAAQS,GAAG,CAAC,GAAG,EAAEA,aAAa,OAAO,MAAM,IAAIohD,EAAE,yDAAyDphD,GAAG,EAAE3C,EAAE8C,EAAEH,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIhC,EAAE1C,EAAE,KAAK,EAAEA,EAAE,OAAO,QAAQiE,KAAK,EAAE,EAAEA,CAAC,EAAE,KAAK,CAAC+iD,GAAGllD,CAAC,GAAG,QAAQmC,KAAK,EAAE,CAAC,IAAIoD,EAAExF,EAAEoC,EAAE,MAAM,GAAGoD,EAAE,QAAQvF,EAAE,CAAC,IAAI+C,EAAE/C,EAAEuF,EAAE,MAAM,OAAOvF,EAAEuF,EAAE,MAAM,QAAQ5C,KAAKI,EAAE7C,EAAEqF,EAAE5C,CAAC,CAAC,CAAC,CAAC,IAAI7B,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE9C,EAAE,YAAY,QAAQiE,KAAKnB,EAAE,CAAC,IAAIuE,EAAEpD,EAAE,GAAGY,EAAEZ,EAAE,GAAGQ,EAAER,EAAE,GAAGkiD,GAAG9+C,KAAKxF,CAAC,EAAE,IAAI8C,EAAE9C,EAAEwF,GAAG,aAAaxC,GAAG,cAAcjC,EAAE,KAAK+B,EAAEF,EAAE,CAAC,CAAC,IAAIT,EAAEhE,EAAE,aAAa,QAAQiE,KAAKD,EAAE,CAAC,IAAIqD,EAAEpD,EAAE,GAAGY,EAAEZ,EAAE,GAAGQ,EAAER,EAAE,GAAGkiD,GAAG9+C,KAAKxF,CAAC,EAAE,IAAI8C,EAAE9C,EAAEwF,GAAG,aAAaxC,GAAG,cAAchC,EAAE,KAAK8B,EAAEF,EAAE,CAAC,CAAC,OAAO,IAAI7D,EAAE,CAAC,OAAOgC,EAAE,QAAQC,EAAE,KAAKH,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,GAAG,KAAK,UAAU,MAAM,IAAIojD,EAAE,sLAAsL,EAAE,QAAQllD,KAAK,KAAK,OAAO,GAAGA,EAAE,SAAS,MAAM,GAAG,MAAM,EAAE,CAAC,aAAa,CAAC2yB,EAAE,IAAI,CAAC,KAAK,OAAO,QAAQ3yB,GAAG,CAACA,EAAE,UAAUA,EAAE,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAASu4D,GAAGr5D,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAE,OAAO,GAAGN,GAAG,MAAM,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,EAAE,IAAIO,GAAG,IAAI,EAAE,GAAGD,IAAI,EAAE,OAAO,MAAM,QAAQN,CAAC,GAAGA,EAAE,SAAS,EAAEA,EAAE,OAAOA,GAAG,UAAU,EAAE,KAAKA,EAAE,CAACA,EAAE,EAAE,GAAG,EAAE,CAACA,CAAC,EAAE,GAAG,MAAM,QAAQA,CAAC,EAAE,CAAC,GAAGA,EAAE,SAASM,EAAE,MAAM,IAAI,MAAM,YAAYJ,oBAAoBF,EAAE,wCAAwCM,0EAA0E,EAAE,OAAON,CAAC,SAAS,OAAOA,GAAG,UAAU,OAAO,KAAKA,CAAC,EAAE,OAAO,GAAG,OAAOA,EAAE,OAAO,KAAKA,CAAC,EAAE,KAAK,SAAS,CAAC,IAAIO,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ,GAAG,CAAC,KAAKP,EAAEO,EAAE,KAAKP,EAAE,EAAE,EAAEO,EAAE,KAAK,IAAI,CAAC,CAAC,EAAEA,CAAC,KAAM,OAAM,IAAI,MAAM,2BAA2BD,kBAAkBJ,kCAAkCI,gCAAgC,oBAAoBJ,qBAAqB,KAAK,UAAUF,CAAC,GAAG,CAAC,CAAC,SAASs5D,GAAGt5D,EAAE,EAAE,CAAC,OAAOq5D,GAAGr5D,EAAE,EAAE,aAAa,CAAC,CAAC,eAAeu5D,GAAGv5D,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,GAAG,MAAMA,GAAG,KAAK,MAAM,IAAI,MAAM,6CAA6C,EAAE,GAAGJ,GAAG,KAAK,CAAC,IAAIK,EAAEkzB,EAAE,IAAI,CAAC,GAAGzzB,EAAE,MAAM,SAAS,EAAE,OAAO2wB,GAAG3wB,CAAC,EAAE,GAAGA,EAAE,MAAM,SAAS,EAAE,CAAC,GAAGA,EAAE,MAAM,GAAG,EAAE,OAAOk6B,GAAGl6B,EAAE,CAAC,EAAE,GAAGA,EAAE,MAAM,KAAK,EAAE,OAAOk8B,EAAEl8B,EAAE,CAACA,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,IAAI,MAAM,+CAA+CA,EAAE,MAAM,wEAAwE,CAAC,KAAM,OAAM,IAAI,MAAM,yCAAyCA,EAAE,4EAA4E,CAAC,CAAC,EAAE,EAAE,MAAM,KAAK,MAAMO,EAAE,KAAK,CAAC,EAAEmzB,GAAGnzB,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ0B,GAAG,CAAC,GAAG/B,EAAE+B,IAAI,KAAK,MAAM,IAAI,MAAM,wEAAwEA,6CAA6C,EAAE,EAAE,KAAK/B,EAAE+B,EAAE,CAAC,CAAC,EAAEkuC,GAAG,EAAE,SAAS,CAAC,KAAM,QAAO,IAAI,CAAC,SAASqpB,GAAGx5D,EAAE,EAAE,CAAC,OAAOo5B,EAAEp5B,EAAE,CAAC,CAAC,CAAC,IAAIy5D,GAAG,GAAG,SAASC,GAAG15D,EAAE,EAAE,CAAC,IAAIE,EAAEI,EAAEC,EAAE,EAAEL,EAAEK,EAAE,GAAGD,EAAEC,EAAE,GAAGumB,EAAE,OAAO5mB,GAAG,MAAMI,GAAG,KAAK,IAAI,mPAAmP,GAAG,EAAE,IAAI,EAAEq5D,GAAG,QAAQ35D,EAAE,WAAWE,CAAC,EAAE,EAAEy5D,GAAG,SAAS35D,EAAE,YAAYM,CAAC,EAAE2B,EAAE,EAAE,GAAG,MAAM,GAAG6kB,EAAE,OAAO,EAAE,SAAS9mB,EAAE,OAAO,OAAO,IAAI,mBAAmBA,EAAE,OAAO,2CAA2C,EAAE,yCAAyC,KAAK,UAAUA,EAAE,UAAU,IAAI,EAAE8mB,EAAE,OAAO,EAAE,SAAS9mB,EAAE,QAAQ,OAAO,IAAI,mBAAmBA,EAAE,QAAQ,4CAA4C,EAAE,2CAA2C,KAAK,UAAUA,EAAE,WAAW,IAAI,EAAE,QAAQkC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI4kB,EAAE,OAAO,EAAE5kB,GAAG,MAAM,KAAKD,EAAE,IAAI,8BAA8BjC,EAAE,WAAWkC,UAAU,EAAEA,GAAG,MAAM,iBAAiBD,oBAAoBjC,EAAE,WAAW,KAAK,EAAE,QAAQkC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI4kB,EAAE,OAAO,EAAE5kB,GAAG,MAAM,KAAKD,EAAE,IAAI,+BAA+BjC,EAAE,YAAYkC,UAAU,EAAEA,GAAG,MAAM,iBAAiBD,oBAAoBjC,EAAE,WAAW,KAAK,EAAE,MAAM,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC,SAAS25D,GAAG35D,EAAE,EAAEE,EAAE,CAAC,GAAGA,aAAa2pB,GAAG,MAAM,CAAC3pB,CAAC,EAAE,GAAG,MAAM,QAAQA,CAAC,EAAE,OAAO4mB,EAAE,OAAO5mB,EAAE,SAAS,EAAE,OAAO,IAAI,wBAAwBA,EAAE,gCAAgC,EAAE,uBAAuBF,UAAU,IAAI,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAE,QAAQC,KAAK,EAAE,CAAC,GAAGL,EAAEK,IAAI,KAAK,MAAM,IAAIylD,EAAE,gEAAgEhmD,UAAUO,KAAK,EAAED,EAAE,KAAKJ,EAAEK,EAAE,CAAC,CAAC,OAAOD,CAAC,CAAC,CAAC,SAASs5D,GAAG55D,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAIimD,GAAG,wDAAwD,EAAE,MAAM,CAAC,GAAGjmD,EAAE,GAAG,GAAGA,EAAE,EAAE,CAAC,CAAC,eAAe65D,GAAG75D,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,EAAE,iBAAiB,KAAK,GAAG4mB,EAAE,OAAO9mB,EAAE,WAAW,KAAK,IAAI,gGAAgG,EAAE8mB,EAAE,OAAO5mB,GAAG,KAAK,IAAI,+FAA+F,EAAE4mB,EAAE,OAAO5mB,EAAE,QAAQ,MAAMA,EAAE,OAAO,GAAG,OAAO,UAAUA,EAAE,MAAM,EAAE,IAAI,iFAAiFA,EAAE,QAAQ,EAAE4mB,EAAE,OAAO,CAACxmB,GAAGJ,EAAE,gBAAgB,GAAG,OAAO,UAAUA,EAAE,eAAe,EAAE,IAAI,uGAAuGA,EAAE,iBAAiB,EAAE4mB,EAAE,OAAO5mB,EAAE,iBAAiB,KAAK,IAAI,mFAAmF,EAAEF,EAAE,WAAW,MAAM,IAAI,MAAM,8DAA8D,EAAEA,EAAE,WAAW,GAAG,GAAG,CAAC,IAAIO,EAAEL,EAAE,gBAAgB,KAAK,EAAE,EAAE,GAAGK,EAAE,GAAGu5D,GAAG55D,EAAE,cAAc,EAAE4mB,EAAE,OAAO5mB,EAAE,mBAAmB,MAAMA,EAAE,kBAAkB,GAAG,OAAO,UAAUA,EAAE,iBAAiB,EAAE,IAAI,iJAAiJA,EAAE,mBAAmB,MAAM,CAAC,IAAI,EAAE05D,GAAG15D,EAAE,cAAc,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,EAAE,CAAC,IAAI+B,EAAEjC,EAAE,kBAAkB,EAAEkC,EAAElC,EAAE,uBAAuB,EAAEmC,EAAE5B,EAAE4B,EAAED,EAAE,MAAM,EAAE,OAAOA,EAAE,IAAI,GAAG,OAAO,CAAC,CAAC,EAAEC,EAAED,EAAE,MAAM,EAAE,IAAIU,EAAEgzD,GAAG11D,EAAE,UAAUA,EAAE,UAAU,EAAE2C,EAAE3C,EAAE,SAAS,KAAK,EAAEA,EAAE,QAAQ,CAAC,aAAa4C,EAAE,QAAQC,CAAC,EAAE+yD,GAAGlzD,EAAEC,EAAE3C,EAAE,OAAO,KAAK,KAAK65D,GAAG,EAAE75D,CAAC,EAAE,KAAKK,EAAE4B,CAAC,EAAEW,EAAE,SAAS9C,CAAC,EAAEA,EAAE,QAAQ+C,EAAE,MAAMD,EAAE,aAAa,EAAE9C,EAAE,cAAc,GAAG,IAAIgD,EAAE9C,EAAE,cAAc,KAAK,EAAEA,EAAE,aAAagE,EAAE,MAAM,EAAE,SAAS,EAAE,KAAKlB,EAAE9C,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC,EAAE,MAAM4C,EAAE,aAAaE,CAAC,EAAE,IAAIuE,EAAE,EAAE,EAAE,EAAE,IAAIjH,IAAI4D,EAAE,MAAM,EAAE,SAAS,GAAG,CAAC5D,GAAGiH,EAAErH,EAAE,iBAAiB,CAAC,IAAIyE,EAAE,MAAMT,EAAE,KAAK,EAAE,GAAG5D,GAAGqE,EAAE,KAAK,CAAC,QAAQ,KAAK,uCAAuCzE,EAAE,oEAAoEqH,2IAA2IrH,EAAE,gBAAgBA,EAAE,wFAAwF,EAAE,KAAK,CAAC,GAAGyE,EAAE,OAAO,KAAK,CAAC,GAAG,CAAC,GAAGC,EAAE,GAAGC,CAAC,EAAE60D,GAAG15D,EAAE2E,EAAE,KAAK,EAAEG,EAAE,CAAC,EAAEA,EAAE,MAAM,EAAEA,EAAE,KAAKF,EAAE,GAAG,MAAM,GAAG,MAAM9B,EAAE,aAAa,EAAEgC,CAAC,EAAE,IAAI0C,EAAE,CAAC,EAAE,GAAGtH,EAAE,aAAa,KAAK,CAAC,IAAI+E,EAAEq0D,GAAGp5D,EAAE,YAAYF,EAAE,WAAW,EAAE,QAAQkF,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAEC,EAAEsC,EAAE,KAAK,MAAM+xD,GAAG10D,EAAEK,GAAG,KAAKD,EAAEC,EAAE,CAAC,CAAC,CAAC,IAAIF,EAAEJ,EAAE,OAAOC,CAAC,EAAE,OAAO2C,CAAC,EAAEC,EAAExF,EAAE+C,CAAC,EAAE0uB,GAAG1uB,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE/C,EAAE,OAAO,EAAE+C,EAAE,CAAC,IAAIC,EAAEhD,EAAE+C,GAAGE,EAAEsC,EAAExC,GAAGH,EAAEI,GAAGC,EAAEwuB,GAAGxuB,CAAC,CAAC,CAAC,MAAMrC,EAAE,WAAW,EAAEgC,CAAC,EAAEswD,GAAGtwD,CAAC,EAAE,IAAIyC,GAAG,CAAC,GAAGjH,EAAEiH,GAAGrH,EAAE,gBAAgByE,EAAE,KAAK,CAAC,GAAGpE,EAAE,CAAC,IAAIqE,EAAEk1D,GAAG55D,EAAE,cAAc,EAAE0E,EAAE4hD,GAAG,MAAMxmD,EAAE,gBAAgBE,EAAE,eAAe,CAAC,QAAQA,EAAE,iBAAiB,CAAC,CAAC,EAAE0E,EAAE4hD,GAAGxmD,EAAE,SAAS,EAAE,EAAE,CAAC,UAAUE,EAAE,qBAAqB,KAAKu5D,GAAGv5D,EAAE,oBAAoB,QAAQ,CAAC,CAAC,CAAC,EAAE,QAAQ2E,EAAE,EAAEA,EAAE7E,EAAE,aAAa,OAAO,EAAE6E,EAAE,EAAE,OAAO7E,EAAE,aAAa6E,MAAMD,EAAEC,EAAE,CAAC,KAAK,CAAC,GAAG7E,EAAE,cAAc,KAAK,CAAC,GAAG,MAAM8C,EAAE,WAAWE,EAAE,CAAC,EAAEA,IAAIhD,EAAE,cAAc,KAAK,CAAC,OAAO,MAAM8C,EAAE,WAAW,EAAE,MAAM9C,EAAE,QAAQ,SAAS,EAAEA,EAAE,OAAO,QAAC,CAAQA,EAAE,WAAW,EAAE,CAAC,CAAC,SAAS+5D,GAAG/5D,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,OAAO,EAAE,iBAAiB,KAAKA,EAAE,EAAE,gBAAgB,OAAO,SAASF,EAAE,IAAI,IAAIE,EAAEF,EAAE,MAAME,CAAC,CAAC,SAAS45D,GAAG95D,EAAE,CAAC,OAAO,OAAOA,EAAE,UAAU,UAAU,CAAC,SAASg6D,GAAGh6D,EAAE,CAAC,OAAO,OAAOA,EAAE,MAAM,UAAU,CAAC,eAAei6D,GAAGj6D,EAAE,EAAEE,EAAE,CAACA,EAAEA,GAAG,CAAC,EAAE,IAAII,EAAEJ,EAAE,SAAS,KAAKK,EAAEP,EAAE,aAAa,EAAE,CAAC,EAAE,GAAGE,EAAE,QAAQ,EAAE,MAAM,IAAI+lD,GAAG,sCAAsC,EAAEn/B,EAAE,OAAO,CAACxmB,GAAGJ,EAAE,QAAQ,GAAG,OAAO,UAAUA,EAAE,OAAO,EAAE,IAAI,wEAAwE,KAAK,UAAUA,EAAE,OAAO,GAAG,EAAE,IAAI,EAAE85D,GAAG,CAAC,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE/3D,EAAE,EAAEC,EAAE,EAAE,KAAK,CAAC5B,GAAG4B,EAAEhC,EAAE,SAAS,CAAC,IAAIiC,EAAE,MAAM,EAAE,KAAK,EAAE,GAAG,EAAEsxB,EAAE,IAAI,CAAC,GAAGtxB,EAAE,MAAM,CAAC,GAAG,CAAC,GAAGS,EAAE,GAAGC,CAAC,EAAE62D,GAAG15D,EAAEmC,EAAE,KAAK,EAAEW,EAAEF,EAAE,OAAOC,CAAC,EAAEE,EAAE0wB,EAAE,IAAIlzB,EAAEuC,CAAC,CAAC,EAAE,GAAG4wB,GAAG5wB,CAAC,EAAEZ,IAAI,EAAE,QAAQgC,EAAE,EAAEA,EAAEnB,EAAE,OAAO,EAAEmB,EAAE,EAAE,KAAK2+B,GAAG,CAAC,CAAC,EAAE,IAAI7/B,EAAEF,EAAE,GAAG,MAAM,GAAG,QAAQoB,EAAE,EAAEA,EAAEnB,EAAE,OAAO,EAAEmB,EAAE,CAAC,IAAI,EAAEnB,EAAEmB,GAAGqD,EAAE,EAAErD,GAAG,EAAEA,GAAGuvB,EAAE,IAAIqF,EAAE,EAAE50B,GAAGk1B,EAAEp2B,EAAE,CAAC,CAAC,CAAC,EAAEd,EAAE,GAAGwxB,GAAGnsB,CAAC,CAAC,CAACmsB,GAAG3wB,CAAC,EAAEd,GAAGe,EAAE,EAAEd,CAAC,CAAC,OAAO,CAAC,CAAC,EAAEC,EAAE,KAAK,CAAC7B,GAAG,QAAQ,KAAK,gLAAgLJ,EAAE,yFAAyF,EAAE,KAAK,CAAC,CAAC,QAAQiC,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIS,EAAE,EAAET,GAAG,EAAEA,GAAG+2B,GAAG,EAAE/2B,GAAGF,CAAC,EAAEyxB,GAAG9wB,CAAC,CAAC,CAAC,OAAO2jD,GAAG,CAAC,CAAC,CAAC,SAAS2T,GAAGl6D,EAAE,CAAC8mB,EAAE,OAAO9mB,EAAE,GAAG,OAAO,UAAUA,CAAC,EAAE,IAAI,2DAA2DA,GAAG,CAAC,CAAC,SAASm6D,GAAGn6D,EAAE,EAAEE,EAAE,CAAC,OAAOF,GAAG,KAAK,CAAC,IAAI,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIM,GAAGupD,GAAGvpD,EAAE,EAAEJ,EAAE,CAAC,CAAC,EAAE2pD,GAAG7pD,EAAE,EAAEE,EAAE,CAAC,CAAC,CAAC,SAASk6D,GAAGp6D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,IAAIzzB,GAAG,KAAK,KAAK,MAAM,QAAQA,CAAC,EAAEA,EAAE,IAAIE,GAAGk6D,GAAGl6D,EAAE,CAAC,CAAC,EAAEmqD,GAAGrqD,EAAE,EAAE,QAAQ,QAAQ,EAAEywB,GAAE,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,SAAS4pC,GAAGr6D,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAE,EAAEC,EAAE,KAAK,KAAKD,EAAEN,GAAGO,EAAED,EAAE,EAAEC,GAAGP,IAAIO,EAAEP,GAAGE,EAAE,KAAK,CAACI,EAAEC,CAAC,CAAC,EAAED,EAAEC,EAAE,OAAOL,CAAC,CAAC,eAAeo6D,GAAGt6D,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,EAAEC,EAAEC,EAAE,CAACzC,GAAG,OAAOA,EAAE,IAAI,GAAG,OAAO,EAAE,GAAGqC,GAAG,OAAOA,EAAE,IAAIE,GAAG,OAAOA,EAAE,GAAG,IAAIoB,EAAE,GAAG,GAAGhC,GAAG,MAAMC,GAAG,OAAO+B,EAAE,IAAIlB,GAAG,OAAOkB,EAAE,GAAGnB,GAAG,MAAM,MAAM,IAAIijD,EAAE,kGAAkG,EAAE,IAAI,EAAEhmD,EAAE,gBAAgBE,EAAEK,EAAEwC,EAAE,iBAAiB,EAAEwE,EAAE,GAAG,OAAOA,EAAE4hD,GAAG,EAAE,CAAC,GAAG,GAAG,OAAO,EAAE,GAAG,GAAG,CAAC,aAAa,EAAE,QAAQxkD,CAAC,EAAEmxD,GAAG7zD,EAAE,EAAE,EAAEa,EAAE,EAAEC,EAAExC,EAAE2D,EAAErB,CAAC,EAAE,EAAE,SAAS7C,CAAC,EAAEA,EAAE,QAAQ2E,EAAE,MAAM,EAAE,aAAa,EAAE3E,EAAE,cAAc,GAAG,QAAQ4E,EAAE9B,EAAE8B,EAAE,EAAE,EAAEA,EAAE,CAAC,MAAM,EAAE,aAAaA,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAE,GAAG9B,GAAG,KAAK,MAAM,IAAIkjD,GAAG,4CAA4C,EAAE,CAAC,GAAGrjD,IAAI,QAAQ,MAAM,IAAIqjD,GAAG,wCAAwC,EAAErjD,GAAGkkB,EAAE,QAAQvf,CAAC,EAAE,IAAIzC,EAAEqrC,GAAG5oC,CAAC,EAAEC,EAAE6yD,GAAG,EAAE95D,CAAC,EAAE,QAAQyE,EAAE,EAAEA,EAAEwC,EAAE,OAAO,EAAExC,EAAE,CAAC,IAAIyC,EAAE,CAAC,EAAE,GAAG,MAAM,EAAE,aAAazC,EAAEyC,CAAC,EAAEgsB,EAAE,IAAI,CAAC,IAAIxuB,EAAEuC,EAAExC,GAAG,GAAGE,EAAEsC,EAAExC,GAAG,GAAGG,EAAE0kD,GAAG/kD,EAAEG,EAAEC,EAAED,CAAC,EAAEwC,EAAE,MAAMzC,EAAEyC,EAAE,KAAKvC,EAAED,EAAE,IAAIG,EAAEg1D,GAAGl6D,EAAEiF,CAAC,EAAEE,EAAE,EAAED,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEhF,EAAE,OAAO,EAAEgF,EAAE,CAAC,IAAI2C,EAAE3H,EAAEgF,GAAGqC,EAAEtC,EAAEC,GAAGmC,EAAEQ,GAAGN,EAAEgsB,GAAGhsB,CAAC,CAAC,CAAC,GAAG3C,IAAIwC,EAAE,OAAO,GAAGtD,EAAE,CAAC,IAAIoB,EAAEtF,EAAE,SAASkC,EAAEC,EAAE5B,CAAC,EAAE,QAAQ0H,EAAE,EAAEA,EAAE3H,EAAE,OAAO,EAAE2H,EAAE,CAAC,IAAIN,EAAErH,EAAE2H,GAAGC,EAAE5C,EAAE2C,GAAG0rB,GAAGzrB,CAAC,EAAErD,EAAE,OAAO8C,GAAGO,CAAC,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,WAAWlD,EAAEyC,CAAC,EAAE2tD,GAAG3tD,CAAC,EAAEzH,EAAE,cAAc,KAAK,CAAC8E,EAAE,QAAQ,CAAC,CAAC,GAAG,MAAM,EAAE,WAAWF,EAAEC,CAAC,EAAE7E,EAAE,cAAc,KAAK,CAAC,OAAO,MAAM,EAAE,WAAW,EAAE,MAAMA,EAAE,QAAQ,SAAS,EAAEA,EAAE,OAAO,CAAC,eAAeu6D,GAAGv6D,EAAE,EAAEE,EAAEI,EAAE,CAAC,EAAE,CAAC,GAAGN,EAAE,WAAW,MAAM,IAAI,MAAM,8DAA8D,EAAEA,EAAE,WAAW,GAAG,IAAIO,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,EAAE,GAAG,CAAC,IAAIC,EAAEzC,EAAE,WAAW,KAAK,GAAGA,EAAE,UAAU45D,GAAGn3D,CAAC,EAAE,IAAIC,EAAE,GAAGkB,EAAE,MAAMlE,EAAE,oBAAoB,EAAEE,EAAEI,EAAE,aAAaA,EAAE,YAAY0C,EAAED,CAAC,EAAExC,EAAE2D,EAAE,GAAG,EAAEA,EAAE,GAAGpB,EAAEoB,EAAE,GAAG,IAAI,EAAE,GAAGqD,EAAE,GAAGjH,EAAE,gBAAgB,MAAMA,EAAE,eAAe,OAAO,EAAE,CAAC,GAAG,EAAE,GAAGA,EAAE,eAAe,SAAS,EAAE4B,EAAE5B,EAAE,eAAe,GAAG6B,EAAE7B,EAAE,eAAe,OAAQ,OAAMA,EAAE,eAAe,SAAS,EAAE,IAAI2lD,GAAG,+DAA+D,EAAE,IAAID,EAAE,0GAA0G1lD,EAAE,4BAA4B,EAAE,IAAImH,EAAE,GAAGxC,EAAE,MAAMjF,EAAE,oBAAoBkC,EAAEC,EAAE,KAAK,KAAKsF,EAAE1E,CAAC,EAAEH,EAAEqC,EAAE,GAAGpC,EAAEoC,EAAE,GAAGsC,EAAE3E,EAAE,OAAOC,CAAC,CAAC,SAASvC,EAAE,iBAAiB,MAAMA,EAAE,gBAAgB,GAAGA,EAAE,gBAAgB,EAAE,CAAC,EAAE,GAAG,IAAImH,EAAE,KAAK,MAAMlH,EAAE,GAAG,MAAM,IAAI,EAAED,EAAE,gBAAgB,EAAE2E,EAAE1E,EAAE,GAAG,MAAM,GAAGqC,EAAEu3D,GAAG55D,EAAEkH,EAAExC,CAAC,EAAE,EAAE1E,EAAEA,EAAE45D,GAAG55D,EAAE,EAAEkH,CAAC,EAAE5E,EAAEs3D,GAAG,EAAE1yD,EAAExC,CAAC,EAAEhD,EAAE,EAAE,EAAEk4D,GAAG,EAAE,EAAE1yD,CAAC,EAAEF,EAAE3E,EAAE,OAAOC,CAAC,CAAC,MAAMvC,EAAE,iBAAiB,OAAO,EAAE,IAAI,IAAI,EAAEC,EAAE,OAAO,CAAC,EAAE,OAAOuC,CAAC,EAAE9C,EAAE,iCAAiC,EAAE,IAAI2E,EAAE3E,EAAE,kBAAkB,EAAE4E,EAAE5E,EAAE,uBAAuB,EAAE6E,EAAEC,EAAE,GAAG9E,EAAE,iBAAiB,EAAE6E,EAAE7E,EAAE,aAAa8E,EAAEF,EAAE,MAAM,EAAE,OAAOA,EAAE,IAAI6C,GAAG,OAAOA,CAAC,CAAC,IAAI5C,EAAE,KAAK0C,EAAE,CAAC,EAAEzC,EAAEF,EAAE,MAAM,GAAG,IAAI4C,EAAEouD,GAAGt1D,EAAE,UAAUA,EAAE,UAAU,EAAE,OAAO,MAAMg6D,GAAGt6D,EAAE2E,EAAE,EAAEC,EAAE7B,EAAEzC,EAAE,OAAOA,EAAE,QAAQkH,EAAE3C,EAAE0C,EAAEjH,EAAE,QAAQwE,EAAExE,EAAE,aAAa,KAAK,IAAI,CAAC,QAAC,CAAQN,EAAE,WAAW,GAAGw6D,GAAGj6D,EAAE,CAAC,EAAEi6D,GAAG,EAAEt6D,CAAC,EAAEs6D,GAAG,EAAE,CAAC,EAAEA,GAAGv4D,EAAE/B,CAAC,EAAEs6D,GAAG53D,EAAEV,CAAC,EAAEs4D,GAAG33D,EAAEV,CAAC,EAAEW,GAAG,MAAM4wB,GAAG5wB,CAAC,CAAC,CAAC,CAAC,SAAS23D,GAAGz6D,EAAE,CAAC,IAAI,EAAE,CAAC,EAAEA,aAAa6pB,KAAK7pB,EAAE,CAACA,CAAC,GAAG,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAEE,GAAG,GAAGI,EAAE,OAAO,EAAE,EAAE,KAAKkpD,GAAGlpD,EAAE,CAAC,CAAC,MAAM,CAAC,GAAGA,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,uEAAuE,EAAE,EAAE,KAAKA,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,SAASk6D,GAAGx6D,EAAE,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,IAAIE,EAAE,CAAC,EAAE,GAAG,aAAa2pB,GAAG3pB,EAAE,KAAK,EAAE,EAAE,UAAU,MAAM,QAAQ,CAAC,EAAE,EAAE,QAAQK,GAAGL,EAAE,KAAKK,EAAE,EAAE,CAAC,UAAU,GAAG,KAAK,QAAQA,KAAK,EAAE,CAAC,IAAI,EAAE,EAAEA,GAAGL,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,IAAII,EAAE,CAAC,EAAE,GAAGN,aAAa6pB,GAAG3pB,EAAE,QAAQF,EAAE,EAAE,IAAI,IAAIM,EAAE,KAAKN,CAAC,UAAU,MAAM,QAAQA,CAAC,EAAEA,EAAE,QAAQO,GAAG,CAACL,EAAE,QAAQK,EAAE,EAAE,IAAI,IAAID,EAAE,KAAKC,CAAC,CAAC,CAAC,UAAUP,GAAG,KAAK,QAAQO,KAAKP,EAAE,CAAC,IAAI,EAAEA,EAAEO,GAAGL,EAAE,QAAQ,EAAE,EAAE,IAAI,IAAII,EAAE,KAAK,CAAC,CAAC,CAACA,EAAE,QAAQC,GAAG,CAACA,EAAE,YAAYA,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,SAASm6D,GAAG16D,EAAE,CAAC,OAAOA,aAAa6pB,EAAE,CAAC,SAAS8wC,GAAG36D,EAAE,CAAC,OAAO,MAAM,QAAQA,CAAC,CAAC,CAAC,SAAS46D,GAAG56D,EAAE,CAAC,MAAM,CAAC06D,GAAG16D,CAAC,GAAG,CAAC26D,GAAG36D,CAAC,CAAC,CAAC,SAAS66D,GAAG76D,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,GAAG,CAAC,GAAG,GAAG,MAAM,EAAE,SAAS,EAAE,CAAC,GAAGP,GAAG,KAAK,CAAC,IAAI,EAAE,GAAG,GAAG26D,GAAG36D,CAAC,GAAGA,EAAE,OAAO,EAAE,EAAE,WAAW46D,GAAG56D,CAAC,GAAG,QAAQiC,KAAKjC,EAAE,GAAGA,EAAE,eAAeiC,CAAC,EAAE,CAAC,EAAE,GAAG,KAAK,OAAO,EAAE,GAAG,GAAG,EAAE,MAAM,IAAI+jD,EAAE,6BAA6BzlD,+BAA+BP,GAAG,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,GAAGA,GAAG,KAAK,OAAO,EAAE,IAAI,GAAG,IAAI,EAAE,IAAI,EAAE,GAAG46D,GAAG56D,CAAC,EAAE,CAACA,EAAEA,EAAE,EAAE,CAAC,EAAE,QAAQ,KAAK,EAAE,CAAC,GAAGA,EAAE,IAAI,KAAK,MAAM,IAAIgmD,EAAE,yBAAyB,kCAAkC,GAAG,EAAE,EAAE,KAAKhmD,EAAE,EAAE,CAAC,CAAC,SAAS26D,GAAG36D,CAAC,EAAE,CAAC,GAAGA,EAAEA,EAAEA,EAAE,SAAS,EAAE,OAAO,MAAM,IAAIgmD,EAAE,6BAA6BzlD,kHAAkH,EAAE,sEAAsEP,GAAG,EAAE,EAAEA,CAAC,KAAK,CAAC,GAAGA,EAAEA,EAAE,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,aAAazlD,aAAa,EAAE,4EAA4EP,EAAE,OAAO,EAAE,EAAE,CAACA,CAAC,CAAC,CAAC,GAAG,EAAEy6D,GAAG,CAAC,EAAEv6D,GAAG,KAAK,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,CAAC,GAAGA,EAAE,IAAI,KAAK,SAAS,IAAI+B,EAAE,EAAE,GAAG,GAAGA,EAAE,MAAM,SAAS/B,EAAE,GAAG,OAAO,MAAM,IAAI8lD,EAAE,uBAAuBzlD,eAAe,EAAE,cAAcL,EAAE,GAAG,iDAAiD+B,EAAE,OAAO,EAAE,QAAQC,EAAE,EAAEA,EAAEhC,EAAE,GAAG,OAAO,EAAEgC,EAAE,CAAC,GAAGA,IAAI,GAAG,CAAC5B,EAAE,SAAS,IAAI6B,EAAEF,EAAE,MAAMC,GAAGU,EAAE1C,EAAE,GAAGgC,GAAG,GAAGU,GAAG,MAAMA,GAAG,GAAGT,IAAIS,EAAE,MAAM,IAAIojD,EAAE,GAAGzlD,gEAAgEL,EAAE,GAAG,MAAM,EAAEA,EAAE,GAAG,MAAM,4BAA4BA,EAAE,GAAG,MAAM,EAAEA,EAAE,GAAG,MAAM,eAAeK,4BAA4B0B,EAAE,MAAM,iCAAiCA,EAAE,MAAM,MAAM,EAAEA,EAAE,MAAM,MAAM,qBAAqBA,EAAE,SAAS,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,SAAS64D,GAAG96D,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE2mD,GAAGjnD,EAAE,IAAI,GAAG,EAAE,MAAM,EAAE,CAAC,EAAEM,EAAE,KAAK,EAAE,IAAIC,EAAE0mD,GAAG,EAAE,IAAI,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE,GAAG1mD,EAAE,KAAK,EAAED,EAAE,OAAO,EAAE,MAAM,IAAI0lD,EAAE,mFAAmF,KAAK,UAAUhmD,EAAE,IAAI,GAAG,EAAE,KAAK,CAAC,GAAG,EAAE,GAAGO,EAAE,OAAO,EAAE,MAAM,IAAIylD,EAAE,oFAAoF,KAAK,UAAU,EAAE,IAAI,GAAG,EAAE,KAAK,CAAC,GAAG,EAAE,GAAG1lD,EAAE,OAAO,GAAGC,EAAE,OAAO,GAAG,CAACumB,EAAE,YAAYxmB,EAAEC,CAAC,EAAE,MAAM,IAAIylD,EAAE,iFAAiF1lD,EAAE,0BAA0BC,EAAE,sBAAsB,CAAC,CAAC,SAASw6D,GAAG/6D,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC21D,GAAGW,GAAGH,EAAE,EAAE,QAAQl2D,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAEO,EAAE,CAAC,IAAI,EAAEP,EAAEO,GAAG,EAAE,EAAEA,GAAG0B,EAAE/B,EAAEK,GAAG,GAAG,GAAG,KAAK,CAAC,GAAG,IAAIk2D,IAAI,EAAE,MAAM,EAAE,MAAM,OAAO,KAAK,EAAE,MAAM,IAAIzQ,EAAE,2CAA2C,EAAE,+JAA+J,EAAE,GAAG1lD,EAAE,QAAQ,CAAC,IAAI,GAAG,CAAC,IAAI4B,EAAE,EAAE,MAAM,MAAM,CAAC,EAAEC,EAAEF,EAAE,MAAM,CAAC,EAAE,QAAQW,EAAE,EAAEA,EAAEV,EAAE,OAAO,EAAEU,EAAE,CAAC,IAAIC,EAAEX,EAAEU,GAAGE,EAAEX,EAAES,GAAG,GAAGE,GAAG,MAAMD,IAAIC,EAAE,MAAM,IAAIkjD,EAAE,8BAA8B,EAAE,2CAA2C/jD,2FAA2F,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS+4D,GAAGh7D,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,GAAG,CAAC,IAAI,EAAE,GAAG,MAAM,QAAQP,CAAC,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,MAAM,IAAIgmD,EAAE,6BAA6BzlD,sHAAsH,EAAE,qCAAqCP,EAAE,oBAAoB,EAAE,EAAEA,CAAC,KAAK,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,qBAAqB,EAAE,UAAUzlD,oEAAoE,KAAK,UAAUP,EAAE,KAAK,IAAI,EAAE,EAAE,CAACA,CAAC,CAAC,CAAC,GAAGE,GAAG,KAAK,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,CAAC,GAAGA,EAAE,IAAI,KAAK,SAAS,IAAI+B,EAAE,EAAE,GAAG,GAAGA,EAAE,MAAM,SAAS/B,EAAE,GAAG,OAAO,MAAM,IAAI8lD,EAAE,uBAAuBzlD,eAAe,EAAE,cAAcL,EAAE,GAAG,iDAAiD,KAAK,UAAU+B,EAAE,KAAK,GAAG,EAAE,QAAQC,EAAE,EAAEA,EAAEhC,EAAE,GAAG,OAAO,EAAEgC,EAAE,CAAC,GAAGA,IAAI,GAAG,CAAC5B,EAAE,SAAS,IAAI6B,EAAEF,EAAE,MAAMC,GAAGU,EAAE1C,EAAE,GAAGgC,GAAG,GAAGU,GAAG,MAAMA,IAAIT,EAAE,MAAM,IAAI6jD,EAAE,uBAAuBzlD,eAAe,EAAE,oBAAoB,KAAK,UAAUL,EAAE,EAAE,8BAA8B,KAAK,UAAU+B,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,SAASg5D,GAAGj7D,EAAE,EAAE,CAAC,GAAGA,GAAG,MAAM,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,EAAE,IAAIM,GAAG,CAAC,CAAC,EAAE,IAAIJ,EAAE,GAAG,OAAOF,GAAG,UAAU,OAAOA,GAAG,WAAWE,EAAE,CAACF,CAAC,UAAU,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,SAASE,EAAEF,MAAO,OAAM,IAAI,UAAU,kGAAkGA,GAAG,EAAE,GAAG,MAAM,QAAQE,CAAC,EAAE,OAAO,EAAE,IAAII,GAAGJ,CAAC,EAAE,CAAC,IAAII,EAAE,CAAC,EAAE,QAAQC,KAAK,EAAE,CAAC,IAAI,EAAEL,EAAE,eAAeK,CAAC,EAAEL,EAAEK,GAAG,CAAC,EAAE,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,GAAGD,EAAE,KAAK,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,CAAC,IAAI46D,GAAG,eAAeC,GAAG,cAAc/B,EAAE,CAAC,YAAYt4D,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,WAAW,EAAE,CAAC,QAAQA,EAAEZ,EAAE,EAAE,QAAQ,IAAI,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI8lD,EAAE,8KAA8K,EAAE0S,GAAG,KAAK53D,EAAEZ,EAAE,CAAC,CAAC,CAAC,QAAQY,EAAE,CAAC,GAAGA,EAAE,MAAM,OAAOA,EAAE,KAAK,CAAC,GAAG,KAAK,KAAKA,EAAE,KAAK,OAAOA,EAAE,WAAW,SAAS,KAAK,WAAWy3D,GAAGz3D,EAAE,SAAS,EAAE,KAAK,iBAAiB,OAAO,CAAC,GAAG,EAAEA,EAAE,qBAAqB84C,IAAI,MAAM,IAAIoM,EAAE,6DAA6D,EAAE,KAAK,WAAWllD,EAAE,UAAU,KAAK,iBAAiB,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,GAAG,CAAC,MAAM,QAAQY,EAAE,IAAI,GAAG,OAAOA,EAAE,MAAM,UAAU,OAAOA,EAAE,MAAM,WAAW,CAACA,EAAE,KAAKA,EAAE,KAAK,QAAQkB,KAAKlB,EAAE,KAAK,GAAG,KAAK,YAAY,QAAQkB,CAAC,IAAI,GAAG,MAAM,IAAIgkD,EAAE,sCAAsChkD,yCAAyC,KAAK,aAAa,EAAE,QAAQA,KAAK,KAAK,YAAYlB,EAAE,KAAKkB,IAAI,MAAM,QAAQ,KAAK,WAAWA,6HAA6HA,mBAAmB,EAAE9B,EAAE,KAAK+2D,GAAGn2D,EAAE,KAAKkB,EAAE,CAAC,CAAC,SAAS,MAAM,QAAQlB,EAAE,IAAI,EAAE,CAAC,GAAGA,EAAE,KAAK,SAAS,KAAK,QAAQ,OAAO,MAAM,IAAIklD,EAAE,2FAA2F,KAAK,QAAQ,yCAAyCllD,EAAE,OAAO,EAAEZ,EAAEY,EAAE,KAAK,IAAImB,GAAGg1D,GAAGh1D,CAAC,CAAC,CAAC,KAAK,CAAC,IAAID,EAAEi1D,GAAGn2D,EAAE,IAAI,EAAE,KAAK,QAAQ,QAAQmB,GAAG,CAAC/B,EAAE,KAAK8B,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,cAAc9B,EAAE,KAAK,gBAAgB,CAAC,EAAE,KAAK,iBAAiB,CAAC,EAAE,KAAK,YAAY,CAAC,EAAE,QAAQ8B,EAAE,EAAEA,EAAE,KAAK,QAAQ,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE,KAAK,qBAAqBD,GAAGE,EAAE,KAAK,YAAYF,GAAG,KAAK,gBAAgB,KAAKE,CAAC,EAAE,KAAK,iBAAiB,KAAKD,CAAC,EAAE,KAAK,YAAY,KAAK,KAAK,cAAcD,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,QAAQlB,EAAE,QAAQ,KAAK,aAAa,CAAC,MAAM,EAAE,KAAK,eAAe,CAAC,EAAE2nD,GAAG,OAAO,IAAI,CAAC,QAAQzmD,EAAE,EAAEA,EAAE,KAAK,QAAQ,OAAO,EAAEA,EAAE,CAAC,GAAG,EAAE,QAAQA,CAAC,IAAI,GAAG,SAAS,IAAIC,EAAE,KAAK,cAAcD,GAAG,KAAK,QAAQ,OAAO,IAAI,KAAK,eAAe,KAAK,CAACC,EAAED,CAAC,CAAC,EAAE,KAAK,aAAa,KAAK,KAAK,YAAYA,GAAG,OAAO,EAAE,CAAC,CAAC,EAAE,IAAIzB,EAAE06D,GAAGn6D,EAAE,QAAQ,KAAK,WAAW,EAAEiB,EAAE,CAACC,EAAEC,EAAEC,IAAI,CAAC,KAAK,YAAY,OAAO,IAAID,EAAE,KAAK,YAAYD,GAAG,IAAIC,GAAG,KAAK,aAAa,KAAKA,CAAC,EAAE,KAAK,eAAe,KAAK,CAACC,EAAEF,CAAC,CAAC,CAAC,EAAEymD,GAAG,SAAS,IAAI,CAAC,QAAQzmD,EAAE,EAAEA,EAAE,KAAK,QAAQ,OAAO,EAAEA,EAAE,CAAC,GAAG,EAAE,QAAQA,CAAC,IAAI,GAAG,SAAS,IAAIC,EAAE1B,EAAEyB,IAAIG,GAAG,CAAC,IAAIS,EAAE,GAAGC,EAAEC,EAAEC,EAAE,QAAQC,KAAKb,EAAE,CAAC,GAAG,OAAOa,GAAG,UAAU,CAAC,WAAW,MAAM,eAAe,IAAI,EAAE,QAAQA,CAAC,IAAI,GAAG,CAAC,IAAImB,EAAE,KAAK,qBAAqBnC,GAAGmC,EAAEA,EAAE,OAAO,KAAK,GAAG,KAAK,cAAcnC,KAAK40D,GAAG,CAAC,WAAW,KAAK,EAAE,QAAQ5zD,CAAC,IAAI,GAAGF,EAAEo0D,GAAG,CAAC,eAAe,IAAI,EAAE,QAAQl0D,CAAC,IAAI,KAAKF,EAAE20D,IAAI,KAAK,cAAcz1D,KAAK00D,GAAG,CAAC,WAAW,KAAK,EAAE,QAAQ1zD,CAAC,IAAI,GAAGF,EAAE40D,GAAG,CAAC,eAAe,IAAI,EAAE,QAAQ10D,CAAC,IAAI,KAAKF,EAAEq1D,IAAI,CAAC,WAAW,KAAK,EAAE,QAAQn1D,CAAC,IAAI,GAAGF,EAAEq0D,GAAG,CAAC,eAAe,IAAI,EAAE,QAAQn0D,CAAC,IAAI,KAAKF,EAAEm1D,IAAI,IAAI1wD,EAAE,CAAC,WAAW,KAAK,EAAE,QAAQvE,CAAC,IAAI,GAAGuE,EAAE,MAAM,CAAC,eAAe,IAAI,EAAE,QAAQvE,CAAC,IAAI,KAAKuE,EAAE,MAAMxE,EAAED,EAAED,EAAED,EAAE2E,CAAC,MAAMxE,EAAEs1D,GAAGr1D,CAAC,EAAEH,EAAED,EAAE01D,GAAGt1D,CAAC,EAAE,IAAI,EAAEylD,GAAG5lD,EAAE,IAAI,CAAC,EAAEE,CAAC,CAAC,EAAEhB,EAAEC,EAAEa,EAAE,CAAC,CAAC,CAAC,GAAGZ,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,0BAA0B,KAAK,gBAAgB,CAAC,kCAAkC,CAAC,KAAK,2BAA2B,MAAM,KAAK,iBAAiB,SAAS,KAAK,0BAA0B,QAAQ,QAAQ,KAAK,mJAAmJ,CAAC,CAAC,SAASnB,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,IAAIK,EAAE,EAAE,WAAW,KAAK,GAAG,EAAE,UAAU25D,GAAG35D,CAAC,EAAE,IAAIwB,EAAE,GAAGC,EAAE,KAAK,sBAAsBlB,EAAEZ,EAAE6B,EAAExB,CAAC,EAAE,GAAG,CAAC,IAAI0B,EAAED,EAAE,GAAG,OAAOA,EAAE,EAAE,EAAE,KAAK,iBAAiB,EAAE,IAAIE,EAAE,KAAK,aAAa,EAAE,KAAK,SAASA,EAAED,EAAE1B,EAAE,EAAE,QAAQ,EAAE,KAAK,EAAE,OAAOgmD,GAAG,CAAC,CAAC,QAAC,CAAQiU,GAAGx4D,EAAE,GAAGlB,CAAC,EAAE05D,GAAGx4D,EAAE,GAAG9B,CAAC,CAAC,CAAC,CAAC,MAAM,gBAAgBY,EAAEZ,EAAE,CAAC,OAAO,KAAK,iBAAiB,EAAE+5D,GAAG,KAAKn5D,EAAEZ,CAAC,CAAC,CAAC,gBAAgBY,EAAEZ,EAAE,EAAEK,EAAE,QAAQ,CAAC,IAAIwB,EAAE,GAAG,GAAG,MAAM,GAAGA,EAAE,KAAK7B,GAAG,KAAK,MAAM,IAAI8lD,EAAE,MAAMzlD,iEAAiEL,GAAG,UAAUY,GAAG,KAAK,MAAM,QAAQA,CAAC,EAAEiB,EAAEjB,EAAE,GAAG,MAAM,GAAGiB,EAAEjB,EAAE,MAAM,OAAQ,OAAM,IAAIklD,EAAE,yDAAyDzlD,uBAAuB,EAAE,OAAOwB,CAAC,CAAC,QAAQjB,EAAEZ,EAAE,CAAC,GAAG,MAAM,QAAQA,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI8lD,EAAE,oDAAoD,EAAE,IAAI,EAAE,MAAM,QAAQ9lD,CAAC,EAAEK,EAAE,EAAEL,EAAE,CAACA,CAAC,EAAE6B,EAAE,KAAK,wBAAwBxB,CAAC,EAAEyB,EAAE,IAAI4rD,GAAG,GAAG9sD,aAAa+oB,KAAK/oB,EAAE,CAACA,CAAC,GAAG,MAAM,QAAQA,CAAC,EAAE,CAAC,GAAGA,EAAE,SAAS,KAAK,OAAO,OAAO,MAAM,IAAIklD,EAAE,kCAAkCllD,EAAE,8DAA8D,KAAK,OAAO,UAAU,EAAE,QAAQoB,EAAE,EAAEA,EAAE,KAAK,OAAO,OAAO,EAAEA,EAAEF,EAAE,IAAI,KAAK,OAAOE,GAAGpB,EAAEoB,EAAE,CAAC,KAAM,SAAQA,KAAK,KAAK,OAAO,CAAC,IAAI,EAAEpB,EAAEoB,EAAE,MAAM,GAAG,GAAG,KAAK,MAAM,IAAI8jD,EAAE,8CAA8C9jD,EAAE,MAAM,EAAEF,EAAE,IAAIE,EAAE,CAAC,CAAC,CAAC,IAAID,EAAE+rD,GAAGjsD,EAAEC,CAAC,EAAE,OAAO,EAAEC,EAAEA,EAAE,EAAE,CAAC,wBAAwBnB,EAAE,CAAC,IAAIZ,EAAEkmD,GAAG,KAAKtlD,EAAE,MAAM,EAAE,EAAEA,EAAE,OAAO,QAAQP,KAAK,KAAK,OAAO,CAAC,IAAIwB,EAAE,MAAM,QAAQxB,EAAE,MAAM,EAAEA,EAAE,OAAO,CAACA,EAAE,MAAM,EAAEyB,EAAED,EAAE,IAAIE,GAAGA,EAAE,IAAI,EAAE,QAAQA,EAAE,EAAEA,EAAEnB,EAAE,OAAO,EAAEmB,EAAE,CAAC,IAAIC,EAAEF,EAAE,QAAQlB,EAAEmB,EAAE,EAAE,GAAGC,IAAI,KAAKhC,EAAE+B,GAAGF,EAAEG,GAAG,KAAK,IAAI,EAAE,KAAK,CAAC,GAAG,IAAI,EAAE,KAAK,CAAC,GAAG,EAAE,EAAE,CAAC,IAAI3B,EAAE,CAAC,EAAE,MAAML,EAAE,QAAQ,CAAC6B,EAAEC,IAAI,CAACD,GAAG,MAAMxB,EAAE,KAAKO,EAAEkB,EAAE,CAAC,CAAC,EAAE,IAAIgkD,EAAE,mDAAmD,KAAK,UAAUzlD,CAAC,GAAG,CAAC,CAAC,OAAOL,CAAC,CAAC,YAAYY,EAAEZ,EAAE,GAAG,EAAE,GAAG,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAIlzB,EAAE,KAAK,gBAAgBO,CAAC,EAAE,GAAG,EAAE,MAAM,IAAImlD,GAAG,+CAA+C,EAAE,IAAIlkD,EAAEs4D,GAAG95D,EAAEL,CAAC,EAAE8B,EAAE,KAAK,QAAQ,IAAIC,GAAG,CAAC,CAAC,EAAE,QAAQA,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAEwxB,EAAE,IAAI,CAAC,IAAItxB,EAAEJ,EAAEE,GAAG,GAAGW,EAAEb,EAAEE,GAAG,GAAGY,EAAEs3D,GAAGr5D,EAAEqB,EAAES,CAAC,EAAEE,EAAE,CAAC,EAAE,GAAG,MAAM,QAAQD,CAAC,EAAE,QAAQG,EAAE,EAAEA,EAAEH,EAAE,OAAO,EAAEG,EAAEF,EAAE,KAAK,CAAC,IAAI,KAAK,OAAOE,GAAG,MAAMH,EAAEG,EAAE,CAAC,OAAOF,EAAE,KAAK,CAAC,IAAI,KAAK,OAAO,GAAG,MAAMD,CAAC,CAAC,EAAE,IAAIE,EAAE,IAAI6qD,GAAG9qD,CAAC,EAAE,OAAOkrD,GAAG,KAAK,QAAQjrD,CAAC,CAAC,CAAC,EAAE,QAAQ,CAACZ,EAAES,IAAIZ,EAAEY,GAAG,KAAKT,CAAC,CAAC,EAAE,OAAOokD,GAAGvkD,EAAE,IAAIC,GAAGu6B,GAAGv6B,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQnB,EAAEZ,EAAE,CAAC,EAAE,CAAC,IAAI,EAAEu6D,GAAG35D,CAAC,EAAEk6D,GAAG,EAAE,KAAK,WAAW,KAAK,gBAAgB,EAAE,EAAE,GAAG,CAAC,IAAIz6D,EAAEL,EAAE,WAAW,KAAK,GAAGA,EAAE,UAAU,OAAOg6D,GAAG35D,CAAC,EAAE,KAAK,YAAY,EAAEA,CAAC,CAAC,QAAC,CAAQi6D,GAAG,EAAE15D,CAAC,CAAC,CAAC,CAAC,eAAeA,EAAE,CAACk6D,GAAGl6D,EAAE,KAAK,WAAW,KAAK,gBAAgB,EAAE,EAAE,IAAIZ,GAAG,MAAM,QAAQY,CAAC,EAAEA,EAAE,GAAGA,GAAG,MAAM,GAAG,OAAO,KAAK,YAAYA,EAAEZ,CAAC,CAAC,CAAC,sBAAsBY,EAAEZ,EAAE,EAAE,GAAGK,EAAE,CAAC,GAAG,KAAK,YAAY,KAAK,MAAM,IAAIwlD,GAAG,8FAA8F,EAAE,IAAIhkD,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,KAAK,iBAAiB,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE,KAAK,iBAAiBD,GAAG,KAAK,YAAYA,KAAK00D,GAAG30D,EAAE,KAAKE,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,EAAEF,EAAE,KAAKE,CAAC,CAAC,CAAC,GAAGnB,EAAE+5D,GAAG/5D,EAAE,KAAK,eAAe,KAAK,gBAAgB,GAAG,OAAO,EAAEZ,EAAE26D,GAAG36D,EAAE,KAAK,gBAAgB6B,EAAE,GAAG,QAAQ,EAAE+4D,GAAGh6D,EAAEZ,EAAE,IAAI,EAAE66D,GAAG76D,EAAE,KAAK,YAAY,KAAK,gBAAgB,EAAE,KAAK,UAAUK,GAAG,MAAMA,EAAE,GAAGO,EAAE,GAAG,MAAM,GAAGP,IAAI,EAAE,MAAM,IAAIylD,EAAE,mHAAmHzlD,aAAaO,EAAE,GAAG,MAAM,eAAe,EAAE,MAAM,CAACA,EAAEZ,CAAC,CAAC,CAAC,MAAM,oBAAoBY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,GAAGC,EAAE,CAAC,GAAG,CAACC,EAAEC,CAAC,EAAE,KAAK,sBAAsBpB,EAAEZ,EAAE6B,EAAEC,CAAC,EAAE,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,qCAAqC,EAAE,IAAI,EAAE,KAAK,GAAGzB,GAAG,KAAK,CAAC,IAAIqC,EAAE02D,GAAG/4D,EAAE,KAAK,WAAW,EAAE,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEqC,EAAE,OAAO,EAAE,EAAE,EAAE,KAAK,MAAM22D,GAAGr3D,EAAE,GAAG,KAAKU,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,CAACX,EAAEC,EAAE,CAAC,CAAC,CAAC,SAASpB,EAAEZ,EAAE,EAAEK,EAAE,EAAEwB,EAAE,CAAC,OAAO0xB,EAAE,IAAI,CAAC,IAAIzxB,EAAE,KAAK,gBAAgB9B,EAAE,EAAE6B,EAAE,OAAO,EAAEE,EAAE,CAAC,EAAE,GAAG1B,EAAE,EAAE,MAAM,IAAI0lD,GAAG,sCAAsC,EAAE,GAAGlkD,GAAG,KAAK,MAAM,IAAIkkD,GAAG,iDAAiD,EAAE,CAAC,IAAI/jD,EAAEm4D,GAAGr4D,EAAE,CAAC,EAAE,EAAEmuC,GAAGgZ,GAAG,EAAEnnD,CAAC,CAAC,EAAE,QAAQY,EAAE,EAAEA,EAAEV,EAAE,OAAO,EAAEU,EAAE,CAAC,IAAI,EAAEV,EAAEU,GAAG,GAAGE,EAAEZ,EAAEU,GAAG,GAAGG,EAAE8mD,GAAG,EAAE,EAAE/mD,EAAE,CAAC,EAAEE,EAAEo3D,GAAGl6D,EAAE6C,CAAC,EAAEmB,EAAEpD,EAAEkC,CAAC,EAAE,GAAGJ,IAAI,EAAE,QAAQuB,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAEC,EAAElC,EAAE,KAAK4gC,GAAG,CAAC,CAAC,EAAE,QAAQ1+B,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAEC,EAAE,CAAC,IAAIoD,EAAErD,EAAEC,GAAGlC,EAAEkC,GAAG20B,EAAE72B,EAAEkC,GAAGi1B,EAAEt2B,EAAE,EAAEyE,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ3E,EAAE,EAAEA,EAAEX,EAAE,OAAO,EAAEW,EAAEX,EAAEW,GAAGs2B,GAAGj3B,EAAEW,GAAGZ,CAAC,CAAC,CAAC,OAAOC,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,IAAInB,EAAE,KAAK,aAAaZ,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEY,EAAE,OAAO,EAAE,EAAE,CAAC,IAAIP,EAAEO,EAAE,GAAGiB,EAAExB,EAAE+lD,GAAGxlD,EAAEP,CAAC,EAAE,IAAIwB,GAAG,IAAIukD,GAAGxlD,EAAE,MAAM,EAAE,CAAC,EAAEP,CAAC,KAAKL,EAAE,KAAK6B,CAAC,CAAC,CAAC,OAAO7B,CAAC,CAAC,mBAAmB,CAAC,OAAOY,GAAG,CAAC,IAAIZ,EAAE,CAAC,EAAE,EAAEY,EAAE,MAAM,EAAE,KAAK,OAAO,MAAM,EAAEP,EAAEO,EAAE,MAAM,KAAK,OAAO,OAAO,KAAK,OAAO,OAAO,KAAK,QAAQ,MAAM,EAAEiB,EAAEjB,EAAE,MAAM,KAAK,OAAO,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,OAAO,KAAK,QAAQ,OAAO,CAAC,EAAEkB,EAAE,CAAC,EAAEC,EAAE,IAAI,CAAC,IAAIY,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,OAAO,OAAO,EAAE,EAAEA,EAAE,KAAK,CAAC,IAAI,KAAK,OAAO,GAAG,MAAM,EAAE,EAAE,CAAC,EAAE,IAAIC,EAAE,IAAI8qD,GAAG/qD,CAAC,EAAEE,EAAEirD,GAAG,KAAK,QAAQlrD,EAAE,CAAC,SAAS,EAAE,CAAC,EAAEE,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,cAAc,OAAO,EAAE,EAAE,CAAC,IAAImB,EAAE,KAAK,cAAc,GAAGoD,EAAEpD,EAAE5D,EAAE,GAAGwC,EAAE,EAAE,EAAEhB,EAAE,IAAI,OAAOwF,EAAEiyD,GAAGjyD,EAAExF,EAAE,EAAE,GAAG,IAAIgD,EAAEqjC,GAAG7gC,CAAC,EAAErH,EAAE,KAAK6E,CAAC,EAAE,IAAI,EAAE/B,EAAEuE,EAAEvE,EAAE81B,EAAE91B,EAAEuE,CAAC,CAAC,CAAC,QAAQ,EAAE,EAAE,EAAE,KAAK,eAAe,OAAO,EAAE,EAAE,CAAC,IAAIpD,EAAE,GAAG,KAAK,QAAQ,OAAO,GAAG,EAAE,KAAK,QAAQ,OAAOA,EAAEjE,EAAE,OAAO,CAAC,IAAIqH,EAAE,KAAK,eAAe,GAAG,GAAGxC,EAAE,KAAK,eAAe,GAAG,GAAGZ,EAAEikC,GAAG7gC,EAAEhH,EAAEwE,GAAGhC,EAAEgC,EAAE,CAAC,CAAC,CAAC4uB,GAAGxvB,CAAC,EAAEnC,EAAE,KAAKmC,CAAC,CAAC,CAAC,OAAOnB,EAAEolC,GAAGplC,CAAC,EAAE,KAAK,gBAAgB,EAAE,QAAQ,GAAG,CAACA,EAAE81B,EAAE91B,EAAE,CAAC,CAAC,CAAC,EAAEA,CAAC,EAAEd,EAAE,KAAK,0BAA0B,IAAIW,GAAGA,EAAE,KAAK,CAAC,EAAE,EAAE,GAAG,MAAM,CAAC,KAAK,WAAW,SAASZ,EAAE,EAAEC,CAAC,CAAC,EAAE,OAAOF,CAAC,CAAC,CAAC,CAAC,kBAAkB,CAAC,KAAK,aAAalB,GAAG2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAE,CAAC,EAAE,EAAEK,EAAEO,EAAE,MAAM,EAAE,KAAK,OAAO,MAAM,EAAEiB,EAAEjB,EAAE,MAAM,KAAK,OAAO,OAAO,KAAK,OAAO,OAAO,KAAK,QAAQ,MAAM,EAAEkB,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,OAAO,OAAO,EAAE,EAAEA,EAAE,KAAK,CAAC,IAAI,KAAK,OAAO,GAAG,MAAMzB,EAAE,EAAE,CAAC,EAAE,IAAI0B,EAAE,IAAI2rD,GAAG5rD,CAAC,EAAEE,EAAE8rD,GAAG,KAAK,QAAQ/rD,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,cAAc,OAAO,EAAE,EAAE,CAAC,IAAIW,EAAE,KAAK,cAAc,GAAG,EAAEwlC,GAAGxlC,EAAEb,EAAE,GAAGG,EAAE,EAAE,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,EAAE42B,EAAE,EAAE,CAAC,EAAE54B,EAAE,KAAK,CAAC,CAAC,CAAC,QAAQ,EAAE,EAAE,EAAE,KAAK,eAAe,OAAO,EAAE,EAAE,CAAC,IAAI0C,EAAE,KAAK,eAAe,GAAG,GAAG,EAAE,KAAK,eAAe,GAAG,GAAGE,EAAEslC,GAAGxlC,EAAEb,EAAE,GAAGG,EAAE,EAAE,CAAC,EAAEhC,EAAE,KAAK4C,CAAC,CAAC,CAAC,OAAO5C,CAAC,CAAC,CAAC,CAAC,MAAM,IAAIY,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,OAAOq6D,GAAG,KAAKz5D,EAAEZ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,OAAO25D,GAAG,KAAK/4D,EAAEZ,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,oBAAoBY,EAAEZ,CAAC,EAAEK,EAAE,EAAE,GAAGwB,EAAE,EAAE,GAAGE,EAAE,KAAK,kBAAkB,EAAE1B,EAAE,OAAOwB,CAAC,CAAC,EAAEG,EAAE,CAAC,EAAE,QAAQC,KAAKF,EAAE,CAAC,IAAIW,EAAE,MAAMT,EAAE,KAAK,EAAED,EAAE,KAAKU,EAAE,EAAE,CAAC,CAAC,OAAO8wB,GAAGzxB,CAAC,EAAEu4D,GAAG,EAAE,GAAG15D,CAAC,EAAE05D,GAAG,EAAE,GAAGt6D,CAAC,EAAEqmD,GAAGrkD,CAAC,CAAC,CAAC,gBAAgBpB,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,EAAEY,GAAG,MAAMA,EAAE,cAAcP,EAAE,EAAE,KAAK,iBAAiB,KAAK,QAAQwB,EAAE,KAAK,WAAW,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEzB,EAAE,OAAO,EAAEyB,EAAE,GAAG,CAACzB,EAAEyB,GAAG,WAAW9B,EAAE,KAAK,CAAC,KAAKK,EAAEyB,GAAG,aAAa,OAAOD,EAAEC,EAAE,CAAC,EAAE,OAAO9B,CAAC,CAAC,IAAI,aAAaY,EAAE,CAAC,KAAK,cAAcA,CAAC,CAAC,IAAI,cAAc,CAAC,OAAO,KAAK,aAAa,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,aAAaA,IAAI,KAAK,WAAWA,EAAE,KAAK,iBAAiB,GAAG,CAAC,SAAS,CAAC,IAAIA,EAAE,MAAM,QAAQ,EAAE,GAAGA,EAAE,uBAAuB,GAAG,KAAK,WAAW,MAAM,KAAK,iBAAiB,CAAC,IAAIZ,EAAEqzB,GAAG,EAAE,WAAW,KAAK,WAAW,QAAQ,EAAEzyB,EAAE,sBAAsBZ,EAAEqzB,GAAG,EAAE,UAAU,CAAC,OAAOzyB,CAAC,CAAC,oBAAoB,CAAC,IAAIA,EAAE,GAAG,OAAO,KAAK,MAAM,SAASA,EAAE2lD,GAAG,KAAK,IAAI,UAAU,MAAM,QAAQ,KAAK,IAAI,EAAE,CAAC,QAAQvmD,KAAK,KAAK,KAAK,GAAG,OAAOA,GAAG,SAAS,MAAM,IAAI,MAAM,oDAAoD,EAAEY,EAAE,KAAK,KAAK,IAAIZ,GAAGumD,GAAGvmD,CAAC,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAE,OAAO,KAAK,KAAK,IAAI,EAAEY,EAAE,CAAC,EAAE,IAAI,EAAE,KAAK,KAAK,QAAQP,KAAKL,EAAE,GAAG,OAAO,EAAEK,IAAI,SAASO,EAAEP,GAAGkmD,GAAG,EAAElmD,EAAE,MAAO,OAAM,IAAI,MAAM,oDAAoD,CAAC,CAAC,OAAOO,CAAC,CAAC,sBAAsB,CAAC,GAAG,OAAO,KAAK,SAAS,UAAU,OAAO,KAAK,SAAS,WAAW,MAAM,CAAC2lD,GAAG6R,GAAG,KAAK,OAAO,CAAC,CAAC,EAAE,GAAG,MAAM,QAAQ,KAAK,OAAO,EAAE,OAAO,KAAK,QAAQ,IAAIx3D,GAAG2lD,GAAG6R,GAAGx3D,CAAC,CAAC,CAAC,EAAE,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,QAAQY,EAAEZ,GAAGumD,GAAG6R,GAAG,KAAK,QAAQp4D,EAAE,CAAC,EAAE,OAAOY,CAAC,CAAC,CAAC,mBAAmB,CAAC,MAAM,CAAC,KAAK,KAAK,mBAAmB,EAAE,QAAQ,KAAK,qBAAqB,EAAE,iBAAiB,CAAC,WAAW,KAAK,UAAU,aAAa,EAAE,OAAO,KAAK,UAAU,UAAU,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,GAAGA,EAAE,kBAAkB,KAAK,MAAM,IAAI,MAAM,8CAA8C,EAAE,GAAGA,EAAE,cAAc,KAAK,MAAM,IAAI,MAAM,4CAA4C,EAAE,GAAGA,EAAE,oBAAoB,KAAK,MAAM,IAAI,MAAM,kDAAkD,EAAE,IAAIZ,EAAE+4D,GAAGn4D,EAAE,gBAAgB,EAAE,EAAEi1D,GAAG71D,CAAC,EAAEK,EAAE,GAAG,OAAOO,EAAE,MAAM,SAASP,EAAEmmD,GAAG5lD,EAAE,IAAI,UAAU,MAAM,QAAQA,EAAE,IAAI,EAAEP,EAAEO,EAAE,KAAK,IAAIkB,GAAG0kD,GAAG1kD,CAAC,CAAC,UAAUlB,EAAE,MAAM,KAAK,CAACP,EAAE,CAAC,EAAE,QAAQyB,KAAKlB,EAAE,KAAKP,EAAEyB,GAAG0kD,GAAG5lD,EAAE,KAAKkB,EAAE,CAAC,CAAC,IAAID,EAAE,GAAG,MAAM,QAAQjB,EAAE,OAAO,EAAEiB,EAAEjB,EAAE,QAAQ,IAAIkB,GAAG0kD,GAAG1kD,CAAC,CAAC,UAAUlB,EAAE,SAAS,KAAK,CAACiB,EAAE,CAAC,EAAE,QAAQC,KAAKlB,EAAE,QAAQiB,EAAEC,GAAG0kD,GAAG5lD,EAAE,QAAQkB,EAAE,CAAC,CAAC,KAAK,QAAQ,CAAC,KAAKzB,EAAE,QAAQwB,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,MAAM,KAAKjB,EAAEZ,EAAE,CAAC,GAAG,OAAOY,GAAG,SAAS,CAAC,IAAIqB,EAAE2uB,GAAG,gBAAgBhwB,CAAC,EAAE,GAAGqB,EAAE,SAAS,EAAE,MAAM,IAAI6jD,EAAE,0CAA0CllD,IAAI,EAAE,GAAGqB,EAAE,OAAO,EAAE,MAAM,IAAI6jD,EAAE,wBAAwB7jD,EAAE,kCAAkCrB,IAAI,EAAEA,EAAEqB,EAAE,EAAE,CAAC,GAAGrB,EAAE,MAAM,KAAK,MAAM,IAAIklD,EAAE,8GAA8G,EAAE,IAAI,EAAE,MAAMl1B,GAAG,cAAc,KAAK,gBAAgB5wB,CAAC,CAAC,EAAEK,EAAE,GAAGwB,EAAE,KAAKE,EAAE,CAAC,cAAc,KAAK,OAAOF,EAAExB,CAAC,EAAE,OAAO26D,GAAG,YAAY,8BAA8B/B,KAAK,YAAY,IAAI,EAAE,IAAIj5D,GAAG,KAAK,GAAGA,EAAE,mBAAmB,KAAK,WAAW,KAAK,CAAC+B,EAAE,eAAe,KAAK,kBAAkB,EAAE,IAAIE,EAAE,YAAY,CAAC,KAAKS,EAAE,MAAMC,CAAC,EAAE,MAAMiuB,GAAG,cAAc,MAAM,KAAK,UAAU,WAAW,EAAE3uB,CAAC,EAAE,EAAE,MAAM,KAAK,GAAGU,CAAC,EAAE,EAAE,KAAKiuB,GAAG,wBAAwB,CAAC,EAAE,KAAKluB,CAAC,CAAC,CAAC,CAAC,OAAO,KAAK,qBAAqB,OAAO41D,GAAG,KAAK,oBAAoB,KAAK,KAAK,EAAE,EAAEv2D,EAAE,oBAAoB,KAAK,qBAAqBA,EAAE,WAAW,EAAE,KAAKA,EAAE,YAAY,EAAE,MAAMnB,EAAE,KAAKmB,CAAC,CAAC,CAAC,uBAAuBnB,EAAE,CAAC03D,GAAG13D,EAAE,KAAK,IAAI,EAAE,KAAK,oBAAoBA,CAAC,CAAC,wBAAwB,CAAC,OAAO,KAAK,mBAAmB,CAAC,EAAEq6D,GAAG,UAAU,QAAQ1jC,GAAE,cAAc0jC,EAAE,EAAE,IAAIC,GAAG,cAAcD,EAAE,CAAC,EAAEC,GAAG,UAAU,aAAa3jC,GAAE,cAAc2jC,EAAE,EAAE,eAAeC,GAAGr7D,EAAE,EAAE,CAAC,kBAAkBA,IAAIA,EAAE,CAAC,cAAcA,CAAC,GAAGA,EAAEA,EAAE,IAAIE,EAAEF,EAAE,cAAcE,EAAE,cAAc,OAAOA,EAAEA,EAAE,cAAc,IAAII,EAAE24D,GAAG/4D,CAAC,EAAEK,EAAEw1D,GAAGz1D,EAAE,CAAC,EAAE,GAAGN,EAAE,iBAAiB,KAAK,CAAC,IAAI,EAAE,MAAM8wB,GAAG,YAAY9wB,EAAE,gBAAgBA,EAAE,WAAWO,EAAE,QAAQ,IAAI0B,GAAGA,EAAE,YAAY,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQA,KAAK1B,EAAE,QAAQ,EAAE0B,EAAE,cAAc,EAAEA,EAAE,cAAc1B,EAAE,YAAY,CAAC,EAAEmzB,GAAG,CAAC,CAAC,CAAC,OAAOnzB,CAAC,CAAC,eAAe+6D,GAAGt7D,EAAE,EAAE,CAAC,GAAG,GAAG,OAAO,EAAE,CAAC,GAAG,OAAOA,GAAG,SAAS,CAAC,IAAIE,EAAE4wB,GAAG,gBAAgB9wB,EAAE,CAAC,EAAE,GAAGE,EAAE,SAAS,EAAEA,EAAE,KAAK4wB,GAAG,mBAAmB9wB,EAAE,CAAC,CAAC,UAAUE,EAAE,OAAO,EAAE,MAAM,IAAI8lD,EAAE,wBAAwB9lD,EAAE,kCAAkCF,IAAI,EAAEA,EAAEE,EAAE,EAAE,CAAC,OAAOq7D,GAAGv7D,EAAE,OAAO,CAAC,CAAC,CAAC,eAAeu7D,GAAGv7D,EAAE,EAAEE,EAAE,CAAC,GAAGA,GAAG,OAAOA,EAAE,CAAC,GAAGF,EAAE,MAAM,KAAK,MAAM,IAAIgmD,EAAE,+GAA+G,EAAE,IAAI1lD,EAAE,MAAMN,EAAE,KAAK,EAAEO,EAAED,EAAE,cAAcC,EAAE,cAAc,OAAOA,EAAEA,EAAE,cAAc,IAAI,EAAEL,EAAE,QAAQ,KAAK,GAAGA,EAAE,OAAO,EAAEI,EAAE,YAAY,MAAMA,EAAE,aAAa,MAAM,EAAE2B,EAAE8zD,GAAGkD,GAAG14D,CAAC,EAAE,EAAE,CAAC,EAAE2B,EAAE5B,EAAE,eAAe,GAAG4B,GAAG,MAAMD,EAAE,mBAAmBC,CAAC,EAAE5B,EAAE,qBAAqB,MAAM2B,EAAE,uBAAuB3B,EAAE,mBAAmB,EAAEA,EAAE,YAAY,KAAK,CAAC,GAAGA,EAAE,aAAa,KAAK,MAAM,IAAI0lD,EAAE,gHAAgH,EAAE,GAAG,CAAC,aAAa7jD,EAAE,iBAAiBS,CAAC,EAAE44D,GAAGl7D,EAAE,WAAWA,EAAE,WAAW,EAAE2B,EAAE,YAAYE,EAAE,CAAC,EAAEF,EAAE,WAAW,MAAMW,EAAE,OAAO,GAAG,MAAMX,EAAE,UAAU,WAAWW,CAAC,EAAE8wB,GAAGvxB,CAAC,EAAEuxB,GAAG9wB,EAAE,IAAIC,GAAGA,EAAE,MAAM,CAAC,CAAC,CAAC,OAAOZ,CAAC,CAAC,SAASu5D,GAAGx7D,EAAE,EAAE,CAAC,IAAIE,EAAE4wB,GAAG,cAAc9wB,EAAE,CAAC,EAAEM,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ,GAAG,CAAC,EAAE,QAAQ,YAAYA,EAAE,KAAK,CAAC,KAAK,EAAE,KAAK,OAAOL,EAAE,EAAE,KAAK,CAAC,EAAEI,EAAE,EAAE,MAAMJ,EAAE,EAAE,KAAK,CAAC,EAAE,CAAC,aAAaI,EAAE,iBAAiBC,CAAC,CAAC,CAAC,IAAIk7D,GAAG,cAAcN,EAAE,CAAC,YAAYr6D,EAAE,CAAC,GAAG,MAAM,CAAC,OAAO,CAAC,EAAE,QAAQ,CAAC,CAAC,CAAC,EAAEA,EAAEA,GAAG,CAAC,EAAE,KAAK,UAAU,GAAG,KAAK,MAAM,GAAG,KAAK,KAAKA,EAAE,MAAM,KAAKA,EAAE,KAAK8mD,GAAG,aAAa,EAAE9mD,EAAE,QAAQ,KAAK,QAAQZ,KAAKY,EAAE,OAAO,KAAK,IAAIZ,CAAC,CAAC,CAAC,WAAWY,EAAE,CAAC,GAAGA,EAAE,aAAa,GAAG,cAAc,GAAG,MAAM,KAAKR,GAAGA,EAAE,CAAC,EAAE,MAAM,IAAI0lD,EAAE,kDAAkDllD,EAAE,0BAA0BA,EAAE,aAAa,GAAG,aAAa,GAAG,QAAQ,CAAC,CAAC,IAAIA,EAAE,CAAC,IAAIZ,EAAEY,aAAa26D,IAAI36D,aAAaq6D,GAAG,EAAE,GAAGj7D,EAAE,CAAC,GAAG,EAAEY,EAAE,EAAE,QAAQ,SAAS,EAAE,MAAM,IAAIklD,EAAE,uHAAuH,EAAE,GAAG,EAAE,OAAO,SAAS,EAAE,MAAM,IAAIA,EAAE,qHAAqH,CAAC,CAAC,GAAG,KAAK,QAAQ,SAAS,EAAE,CAAC,GAAGllD,EAAE,aAAa,SAAS,EAAE,CAAC,GAAGA,EAAE,iBAAiB,KAAK,MAAM,IAAIklD,EAAE,+FAA+F,EAAE,IAAIzlD,EAAEmtD,GAAG,CAAC,WAAW5sD,EAAE,gBAAgB,MAAMA,EAAE,MAAM,KAAKA,EAAE,KAAK,QAAQ,CAAC,EAAEA,EAAE,MAAMP,CAAC,CAAC,CAAC,GAAGL,EAAE,KAAK,QAAQ,EAAE,QAAQ,KAAK,OAAO,EAAE,WAAW,CAAC,GAAGY,EAAE,aAAa,SAAS,EAAE,MAAM,IAAIklD,EAAE,gHAAgHllD,EAAE,kBAAkBA,EAAE,aAAa,0CAA0C,EAAE,GAAGA,EAAE,aAAa,GAAG,cAAc,SAAS,EAAE,MAAM,IAAIklD,EAAE,uHAAuH,EAAE,KAAK,WAAWllD,CAAC,EAAE,KAAK,QAAQ,CAACA,EAAE,aAAa,GAAG,cAAc,EAAE,EAAE,KAAK,OAAO0sD,GAAG,KAAK,QAAQ,EAAE,CAAC,CAAC,KAAK,aAAa,CAAC,EAAE,IAAIL,GAAG,CAAC,cAAc,KAAK,cAAc,CAAC,EAAE,YAAY,CAAC,EAAE,cAAc,CAAC,EAAE,aAAa,KAAK,OAAO,cAAc,KAAK,QAAQ,WAAW/G,GAAG,KAAK,KAAK,OAAO,MAAM,EAAE,YAAY,CAAC,IAAI,EAAE,YAAY,KAAK,OAAO,IAAI7lD,GAAGA,EAAE,KAAK,EAAE,aAAa,KAAK,QAAQ,GAAG,KAAK,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAEO,EAAE,MAAM,KAAK,QAAQ,EAAE,EAAE,GAAG,MAAM,QAAQP,CAAC,EAAE,MAAM,IAAI,UAAU,uHAAuH,EAAE,KAAK,WAAWO,CAAC,EAAE,KAAK,QAAQ,CAACP,CAAC,EAAE,KAAK,aAAa,GAAG,cAAc,KAAK,QAAQ,KAAK,aAAa,GAAG,aAAa,CAAC,KAAK,QAAQ,GAAG,KAAK,CAAC,CAAC,KAAK,OAAO,KAAKO,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAK,CAAC,GAAG,KAAK,OAAO,SAAS,EAAE,MAAM,IAAI,UAAU,mCAAmC,EAAE,GAAG,KAAK,OAAO,IAAI,EAAE,KAAK,OAAO,SAAS,EAAE,KAAK,QAAQ,CAAC,EAAE,KAAK,aAAa,CAAC,EAAE,KAAK,cAAc,CAAC,MAAM,CAAC,IAAIA,EAAE,KAAK,OAAO,OAAO,EAAE,KAAK,OAAOA,GAAG,cAAc,CAAC,EAAE,KAAK,QAAQ,CAAC,KAAK,OAAOA,GAAG,MAAM,EAAE,KAAK,aAAa,GAAG,cAAc,KAAK,QAAQ,KAAK,aAAa,GAAG,aAAa,CAAC,KAAK,QAAQ,GAAG,KAAK,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,EAAE,KAAK,MAAM,KAAKY,EAAEZ,CAAC,CAAC,CAAC,MAAMY,EAAE,CAAC,GAAG2rD,GAAG3rD,CAAC,EAAE,KAAK,OAAO,SAAS,GAAG,KAAK,QAAQ,SAAS,EAAE,MAAM,IAAI,UAAU,0EAA0E,EAAE,KAAK,MAAM,IAAIq6D,GAAG,CAAC,OAAO,KAAK,OAAO,QAAQ,KAAK,QAAQ,GAAG,KAAK,KAAK,KAAK,QAAQ,CAAC,EAAE,KAAK,MAAM,UAAU,KAAK,UAAU,KAAK,gBAAgB,KAAK,MAAM,gBAAgB,KAAK,YAAY,KAAK,MAAM,YAAY,KAAK,uBAAuB,KAAK,MAAM,uBAAuB,KAAK,yBAAyB,KAAK,MAAM,yBAAyB,KAAK,aAAa,KAAK,MAAM,aAAa,KAAK,wBAAwB,KAAK,MAAM,wBAAwB,KAAK,0BAA0B,KAAK,MAAM,0BAA0B,KAAK,aAAa,KAAK,MAAM,aAAa,KAAK,eAAe,KAAK,MAAM,eAAe,KAAK,YAAY,KAAK,MAAM,YAAY,KAAK,WAAW,KAAK,MAAM,WAAW,KAAK,MAAM,EAAE,CAAC,aAAa,CAAC,OAAO,KAAK,OAAO,KAAK,MAAM,EAAE,MAAM,YAAY,CAAC,CAAC,QAAQr6D,EAAEZ,EAAE,EAAE,QAAQ,IAAI,CAAC,KAAK,OAAO,KAAK,MAAM,EAAE,MAAM,QAAQY,EAAEZ,EAAE,CAAC,CAAC,CAAC,WAAWY,EAAE,CAAC,KAAK,OAAO,MAAM,KAAK,MAAM,EAAE,KAAK,MAAM,WAAWA,CAAC,CAAC,CAAC,SAASA,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI6lD,GAAG,mDAAmD,EAAE,OAAO,KAAK,MAAM,SAASjlD,EAAEZ,EAAE,CAAC,CAAC,CAAC,MAAM,gBAAgBY,EAAEZ,EAAE,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI6lD,GAAG,mDAAmD,EAAE,OAAO,KAAK,MAAM,gBAAgBjlD,EAAEZ,CAAC,CAAC,CAAC,QAAQY,EAAEZ,EAAE,CAAC,EAAE,CAAC,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,EAAE,KAAK,MAAM,QAAQY,EAAEZ,CAAC,CAAC,CAAC,eAAeY,EAAE,CAAC,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,EAAE,KAAK,MAAM,eAAeA,CAAC,CAAC,CAAC,QAAQA,EAAE,CAAC,KAAK,MAAM,EAAE,KAAK,MAAM,QAAQA,CAAC,EAAE,KAAK,WAAW,KAAK,MAAM,UAAU,KAAK,iBAAiB,KAAK,MAAM,iBAAiB,KAAK,KAAK,KAAK,MAAM,KAAK,KAAK,QAAQ,KAAK,MAAM,QAAQ,KAAK,eAAe,KAAK,MAAM,eAAe,KAAK,aAAa,KAAK,MAAM,YAAY,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,OAAO,KAAK,OAAO,KAAK,MAAM,SAAS,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,MAAM,UAAUA,CAAC,CAAC,MAAM,IAAIA,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI6lD,GAAG,mDAAmD,EAAE,OAAO,KAAK,MAAM,IAAIjlD,EAAEZ,EAAE,CAAC,CAAC,CAAC,MAAM,WAAWY,EAAEZ,EAAE,CAAC,GAAG,CAAC,KAAK,MAAM,MAAM,IAAI6lD,GAAG,mDAAmD,EAAE,OAAO,KAAK,MAAM,WAAWjlD,EAAEZ,CAAC,CAAC,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAAC,OAAO,KAAK,MAAM,aAAaY,EAAEZ,CAAC,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,EAAE,CAAC,EAAEK,EAAE,GAAG,CAAC,IAAIwB,EAAEC,EAAE,CAAC,EAAE,GAAG9B,aAAa,MAAM,CAAC,GAAGA,EAAE,GAAG,WAAW,MAAMA,EAAE,GAAG,YAAY,QAAQ,MAAM,IAAI8lD,EAAE,gDAAgD,EAAEjkD,EAAE7B,CAAC,MAAM4mB,EAAE,OAAO5mB,EAAE,QAAQ,KAAK,IAAI,qHAAqH,EAAE6B,EAAE7B,EAAE,OAAO,OAAOA,EAAE,OAAO8B,EAAE9B,EAAE,IAAI+B,EAAE,IAAInB,EAAEkB,CAAC,EAAE,GAAG,EAAEC,aAAaw5D,IAAI,MAAM,IAAIxV,GAAG,yDAAyDhkD,GAAG,EAAE,QAAQC,KAAKH,EAAE,CAAC,IAAIa,EAAEmzD,GAAG7zD,EAAE,OAAO3B,CAAC,EAAEA,GAAGqC,EAAE,6BAA6B,EAAE,EAAEX,EAAE,IAAIW,CAAC,CAAC,CAAC,OAAOX,CAAC,CAAC,IAAI,aAAanB,EAAE,CAAC,GAAG,KAAK,OAAO,KAAK,MAAM,IAAIklD,EAAE,mFAAmF,EAAE,KAAK,MAAM,aAAallD,CAAC,CAAC,IAAI,cAAc,CAAC,GAAG,KAAK,OAAO,KAAK,MAAM,IAAIklD,EAAE,mFAAmF,EAAE,OAAO,KAAK,MAAM,YAAY,CAAC,WAAW,CAAC,IAAIllD,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,OAAO,CAAC,IAAI,EAAE,CAAC,EAAE,EAAE,UAAUA,EAAE,aAAa,EAAE,EAAE,OAAOA,EAAE,UAAU,EAAEY,EAAE,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,KAAK,KAAK,OAAOA,CAAC,CAAC,CAAC,EAAE26D,GAAG,UAAU,aAAahkC,GAAE,cAAcgkC,EAAE,EAAE,SAASC,GAAG17D,EAAE,CAAC,OAAO,IAAIm7D,GAAGn7D,CAAC,CAAC,CAAC,SAAS27D,GAAG37D,EAAE,CAAC,OAAO,IAAIy7D,GAAGz7D,CAAC,CAAC,CAAC,SAASkzD,GAAGlzD,EAAE,CAAC,OAAO0tD,GAAG1tD,CAAC,CAAC,CAAC,SAAS47D,GAAG57D,EAAE,EAAE,CAAC61D,GAAG,4BAA4B71D,EAAE,CAAC,CAAC,CAAC,IAAI67D,GAAG,cAAcpkC,GAAE,YAAY,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,CAAC,EAAEqkC,GAAG,cAAcD,EAAE,CAAC,MAAM/6D,EAAEZ,EAAE,EAAE,CAAC,OAAOsqD,GAAG1pD,EAAEZ,CAAC,CAAC,CAAC,EAAE47D,GAAG,UAAU,MAAMrkC,GAAE,cAAcqkC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO0sC,GAAG1sC,CAAC,CAAC,CAAC,EAAEi7D,GAAG,UAAU,OAAOtkC,GAAE,cAAcskC,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAOwrC,GAAGxrC,CAAC,CAAC,CAAC,EAAEk7D,GAAG,UAAU,OAAOvkC,GAAE,cAAcukC,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO2yB,EAAE,IAAIgV,GAAG,EAAE6D,GAAGxrC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEm7D,GAAG,UAAU,QAAQxkC,GAAE,cAAcwkC,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAOA,CAAC,CAAC,EAAEo7D,GAAG,UAAU,SAASzkC,GAAE,cAAcykC,EAAE,EAAE,IAAIC,GAAG,cAAcN,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO47B,GAAG57B,CAAC,CAAC,CAAC,EAAEq7D,GAAG,UAAU,UAAU1kC,GAAE,cAAc0kC,EAAE,EAAE,IAAIC,GAAG,cAAcP,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO6pD,GAAG7pD,CAAC,CAAC,CAAC,EAAEs7D,GAAG,UAAU,cAAc3kC,GAAE,cAAc2kC,EAAE,EAAE,IAAIC,GAAG,cAAcR,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAOwlC,GAAGxlC,CAAC,CAAC,CAAC,EAAEu7D,GAAG,UAAU,WAAW5kC,GAAE,cAAc4kC,EAAE,EAAE,IAAIC,GAAG,cAAcT,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO2pD,GAAG3pD,CAAC,CAAC,CAAC,EAAEw7D,GAAG,UAAU,WAAW7kC,GAAE,cAAc6kC,EAAE,EAAE,IAAIC,GAAG,cAAcV,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAOg8B,GAAGh8B,CAAC,CAAC,CAAC,EAAEy7D,GAAG,UAAU,OAAO9kC,GAAE,cAAc8kC,EAAE,EAAE,IAAIC,GAAG,cAAcX,EAAE,CAAC,MAAM/6D,EAAEZ,EAAE,GAAG,CAAC,OAAO0uC,GAAG9tC,EAAEZ,CAAC,CAAC,CAAC,EAAEs8D,GAAG,UAAU,UAAU/kC,GAAE,cAAc+kC,EAAE,EAAE,IAAIC,GAAG,cAAcZ,EAAE,CAAC,MAAM/6D,EAAEZ,EAAE,GAAG,CAAC,OAAO0mC,GAAG9lC,EAAEZ,CAAC,CAAC,CAAC,EAAEu8D,GAAG,UAAU,aAAahlC,GAAE,cAAcglC,EAAE,EAAE,IAAIC,GAAG,cAAcb,EAAE,CAAC,MAAM/6D,EAAEZ,EAAE,EAAE,CAAC,OAAOuzB,EAAE,IAAI2F,EAAEsD,GAAGtD,EAAEt4B,EAAEZ,CAAC,CAAC,EAAEY,CAAC,CAAC,CAAC,CAAC,EAAE47D,GAAG,UAAU,QAAQjlC,GAAE,cAAcilC,EAAE,EAAE,IAAIC,GAAG,cAAcd,EAAE,CAAC,MAAM/6D,EAAE,CAAC,OAAO2yB,EAAE,IAAI2F,EAAEt4B,EAAEg8B,GAAGwJ,GAAGxlC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE67D,GAAG,UAAU,OAAOllC,GAAE,cAAcklC,EAAE,EAAE,SAASC,GAAG58D,EAAE,CAAC,OAAOA,EAAE,aAAa,CAAC,CAAC,SAAS68D,GAAG78D,EAAE,EAAE,CAAC,EAAE,CAAC,OAAO8mD,GAAG9mD,EAAEy3B,GAAE,iBAAiB,OAAO,EAAE,aAAa,EAAE,YAAY,CAAC,CAAC,SAASqlC,GAAG98D,EAAE,CAAC,GAAGA,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC,EAAE,OAAO,EAAE,UAAU,SAAS,EAAE,OAAO,CAAC,EAAE68D,GAAG,CAAC,CAAC,CAAC,GAAG,OAAO78D,GAAG,SAAS,CAAC,IAAI,EAAE,CAAC,EAAE,OAAO,EAAE,UAAUA,EAAE,EAAE,OAAO,CAAC,EAAE68D,GAAG,CAAC,CAAC,KAAM,QAAO78D,aAAa67D,GAAG77D,EAAE68D,GAAG78D,CAAC,CAAC,CAAC,SAAS+8D,GAAG/8D,EAAE,CAAC,GAAGA,GAAG,MAAM,OAAOA,GAAG,SAAS,MAAM,IAAI,MAAM,yFAAyFA,GAAG,CAAC,CAAC,IAAIg9D,GAAG,cAAcvlC,GAAE,YAAY,CAAC,EAAEwlC,GAAG,cAAcD,EAAE,CAAC,YAAYl8D,EAAE,CAAC,MAAM,EAAEi8D,GAAGj8D,CAAC,EAAE,KAAK,GAAGA,GAAG,MAAMA,EAAE,IAAI,KAAK,IAAIA,EAAE,GAAG,KAAK,GAAGA,GAAG,MAAMA,EAAE,IAAI,KAAK,IAAIA,EAAE,GAAG,KAAK,MAAM,KAAK,KAAK,EAAE,KAAK,MAAM,KAAK,KAAK,CAAC,CAAC,MAAMA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEmoC,GAAG,CAAC,CAAC,CAAC,EAAE,OAAO,KAAK,QAAQnoC,EAAE44B,EAAE54B,EAAEijC,GAAG/J,EAAE,KAAK,GAAGE,GAAGx4B,CAAC,CAAC,CAAC,CAAC,GAAG,KAAK,QAAQZ,EAAE44B,EAAE54B,EAAEijC,GAAG/J,EAAE,KAAK,GAAGkxB,GAAGxpD,CAAC,CAAC,CAAC,CAAC,GAAGo7B,EAAEh8B,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,GAAG,KAAK,GAAG,GAAG,KAAK,EAAE,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAE,CAAC,GAAGZ,EAAE,GAAG,GAAGA,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE+8D,GAAG,UAAU,OAAOxlC,GAAE,cAAcwlC,EAAE,EAAE,SAASC,GAAGl9D,EAAE,CAAC,OAAO+8D,GAAG/8D,CAAC,EAAE,IAAIi9D,GAAG,CAAC,GAAGj9D,GAAG,KAAKA,EAAE,GAAG,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,SAASm9D,GAAGn9D,EAAE,CAAC,OAAO+8D,GAAG/8D,CAAC,EAAE,IAAIi9D,GAAG,CAAC,GAAGj9D,GAAG,KAAKA,EAAE,GAAG,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,IAAIo9D,GAAG,CAAC,KAAK,MAAM,EAAE,SAASC,GAAGr9D,EAAE,CAAC,OAAO4mD,GAAG5mD,CAAC,CAAC,CAAC,SAASs9D,GAAGt9D,EAAE,EAAE,CAAC,EAAE,CAAC,OAAO8mD,GAAG9mD,EAAEy3B,GAAE,iBAAiB,OAAO,EAAE,aAAa,EAAE,aAAa,CAAC,CAAC,SAAS8lC,GAAGv9D,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,KAAK,GAAG,OAAOA,GAAG,SAAS,CAAC,IAAIE,EAAE,CAAC,UAAUF,KAAKo9D,GAAGA,GAAGp9D,GAAGA,EAAE,OAAO,CAAC,CAAC,EAAE,OAAOs9D,GAAGp9D,CAAC,CAAC,KAAM,QAAOF,aAAag9D,GAAGh9D,EAAEs9D,GAAGt9D,CAAC,CAAC,CAAC,IAAIw9D,GAAG,cAAcnQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,gBAAgB,GAAGA,GAAG,OAAO,KAAK,SAASA,EAAE,SAAS,CAAC,KAAKA,EAAEZ,EAAE,CAACY,EAAE0rD,GAAG1rD,CAAC,EAAE,IAAI,EAAEwrC,GAAGxrC,CAAC,EAAE,OAAO,KAAK,UAAU,OAAO,EAAEw9B,GAAG,EAAE,EAAE,KAAK,QAAQ,GAAG,CAAC,CAAC,mBAAmBx9B,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,SAAS,KAAK,QAAQ,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE08D,GAAG,UAAU,OAAO/lC,GAAE,cAAc+lC,EAAE,EAAE,IAAIC,GAAG,cAAcpQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,cAAc,GAAGA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,MAAMA,EAAE,OAAO,KAAK,KAAK,cAAcA,EAAE,KAAK,CAAC,KAAKA,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAOokC,GAAG,EAAE,KAAK,KAAK,CAAC,CAAC,mBAAmBpkC,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,MAAM,KAAK,KAAK,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE28D,GAAG,UAAU,YAAYhmC,GAAE,cAAcgmC,EAAE,EAAE,IAAIC,GAAG,cAAcrQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,0BAA0B,QAAQA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,gBAAgB,GAAG,KAAK,iBAAiBurD,GAAGvrD,EAAE,kBAAkB,KAAK,yBAAyB,EAAE,KAAK,iBAAiBy8D,GAAGz8D,EAAE,gBAAgB,EAAE,KAAK,gBAAgBsuD,GAAGtuD,EAAE,eAAe,EAAEA,EAAE,YAAY,KAAK,KAAK,WAAW,aAAa,MAAM,QAAQA,EAAE,UAAU,EAAE,KAAK,WAAWA,EAAE,mBAAmB,OAAOA,EAAE,YAAY,SAAS,KAAK,WAAW,CAACA,EAAE,UAAU,MAAO,OAAM,IAAIklD,EAAE,sEAAsEllD,EAAE,YAAY,CAAC,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,CAAC,EAAE,GAAG,KAAK,YAAY,KAAK,QAAQP,KAAK,KAAK,WAAWL,EAAEK,EAAE,GAAG,EAAE,KAAK,MAAM,KAAK,UAAU,QAAQL,EAAE,UAAU,KAAK,iBAAiB,KAAK,iBAAiB,GAAG,KAAK,eAAe,EAAE,IAAI,EAAE,CAAC,EAAE,GAAG,KAAK,YAAY,KAAK,QAAQK,EAAE,EAAEA,EAAEO,EAAE,OAAO,EAAEP,EAAE,EAAEA,GAAGO,EAAEP,GAAG,KAAK,UAAU,CAAC,IAAIysD,GAAG,CAAC,KAAKlsD,EAAE,OAAO,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOY,EAAE0rD,GAAG1rD,CAAC,EAAE6pC,GAAG7pC,EAAE,KAAK,MAAM,KAAK,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,iBAAiBsrD,GAAG,KAAK,gBAAgB,EAAE,iBAAiBiR,GAAG,KAAK,gBAAgB,EAAE,gBAAgBnO,GAAG,KAAK,eAAe,EAAE,WAAW,KAAK,UAAU,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE48D,GAAG,UAAU,QAAQjmC,GAAE,cAAcimC,EAAE,EAAE,IAAIC,GAAG,cAActQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,cAAc,EAAEA,GAAG,OAAOA,EAAE,CAAC,GAAGA,EAAE,OAAO,MAAMA,EAAE,QAAQ,KAAK,cAAc,MAAM,IAAImlD,GAAG,4BAA4BnlD,EAAE,+CAA+C,EAAE,KAAK,MAAMA,EAAE,OAAO,KAAK,KAAK,cAAcA,EAAE,KAAK,CAAC,KAAKA,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAO8gC,GAAG,CAAC,CAAC,CAAC,mBAAmB9gC,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,MAAM,KAAK,KAAK,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE68D,GAAG,UAAU,MAAMlmC,GAAE,cAAckmC,EAAE,EAAE,IAAIC,GAAG,cAAcvQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,cAAc,EAAEA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,MAAMA,EAAE,OAAO,KAAK,KAAK,cAAcA,EAAE,KAAK,CAAC,KAAKA,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAOs4B,EAAE,EAAE3I,GAAE+T,GAAG,EAAE,KAAK,KAAK,EAAE,SAAS,CAAC,CAAC,CAAC,mBAAmB1jC,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,MAAM,KAAK,KAAK,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE88D,GAAG,UAAU,kBAAkBnmC,GAAE,cAAcmmC,EAAE,EAAE,IAAIC,GAAG,cAAcxQ,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,aAAa,EAAEA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,QAAQ,IAAI07D,GAAG,EAAE,MAAM,KAAK,KAAK17D,EAAE,MAAM,KAAK,KAAK,aAAaA,EAAE,IAAI,CAAC,KAAKA,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAO,KAAK,QAAQ,EAAE,KAAK,IAAI,CAAC,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,KAAK,KAAK,IAAI,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE+8D,GAAG,UAAU,UAAUpmC,GAAE,cAAcomC,EAAE,EAAE,SAASC,GAAG99D,EAAE,EAAEE,EAAE,CAAC,GAAG,OAAOF,GAAG,SAAS,OAAOomD,GAAGpmD,EAAE,CAAC,EAAE,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAIgmD,EAAE,OAAO9lD,6CAA6C,yBAAyBF,EAAE,kBAAkB,EAAE,QAAQM,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAEP,EAAEM,GAAG,GAAG,CAACyoD,GAAGxoD,CAAC,EAAE,MAAM,IAAIylD,EAAE,OAAO9lD,6CAA6C,yBAAyB,KAAK,UAAUF,CAAC,oCAAoCO,GAAG,CAAC,CAAC,OAAOP,CAAC,CAAC,SAAS+9D,GAAG/9D,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,GAAGP,GAAG,KAAK,OAAOA,EAAE,IAAI,EAAE,GAAG,EAAE,IAAIO,EAAE,GAAG,EAAE,OAAOL,IAAI,OAAO,EAAEF,EAAE,EAAEA,EAAE,EAAE,EAAE,KAAK,OAAO,EAAEM,EAAE,GAAGA,CAAC,CAAC,CAAC,SAAS09D,GAAGh+D,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGN,GAAG,KAAK,OAAO,KAAK,GAAGM,IAAI,QAAQN,EAAEA,EAAE,EAAEkpD,GAAG,CAAChpD,EAAE,EAAE,CAAC,CAAC,UAAUI,IAAI,OAAON,EAAEA,EAAE,MAAO,OAAM,IAAIgmD,EAAE,2BAA2B1lD,IAAI,EAAE,OAAON,CAAC,CAAC,SAASi+D,GAAGj+D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,KAAK00B,GAAG,CAAC,EAAE,IAAI,gBAAgBtzB,GAAG70B,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAEA,EAAE,CAAC,CAAC,SAASk+D,GAAGl+D,EAAE,EAAE,CAAC,OAAOyzB,EAAE,KAAK00B,GAAG,CAAC,EAAE,IAAI,gBAAgBtzB,GAAG70B,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,EAAEA,EAAE,CAAC,CAAC,SAASm+D,GAAGn+D,EAAE,EAAEE,EAAEI,EAAE,EAAEC,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC,OAAOkzB,EAAE,IAAI,CAAC,GAAG,GAAG,OAAO,EAAE61B,GAAG,GAAGnB,GAAG,CAAC,EAAEnoD,EAAE,MAAM,SAAS,EAAE,MAAM,IAAIgmD,EAAE,+DAA+DhmD,EAAE,MAAM,iBAAiB,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAIgmD,EAAE,iEAAiE,EAAE,MAAM,gBAAgB,EAAE,GAAG9lD,GAAG,MAAMA,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI8lD,EAAE,+DAA+D,EAAE,MAAM,gBAAgB,EAAE,GAAG,IAAI,kBAAkBhmD,EAAE60B,GAAG70B,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAGO,IAAI,SAAS,MAAM,IAAI0lD,GAAG,+EAA+E,EAAE,IAAIhkD,EAAEi9B,GAAGl/B,EAAE,EAAEM,EAAEC,IAAI,OAAO,OAAO,QAAQ,MAAM,CAAC,EAAE,OAAOL,GAAG,OAAO+B,EAAEsoD,GAAGtoD,EAAE/B,CAAC,GAAG+B,CAAC,CAAC,CAAC,CAAC,SAASm8D,GAAGp+D,EAAE,EAAEE,EAAEI,EAAE,CAAC,EAAE,CAAC,EAAEC,EAAE,QAAQ,EAAE,EAAE0B,EAAE,KAAK,CAAC,OAAOwxB,EAAE,IAAI,CAAC,GAAG,GAAG,OAAO,EAAE61B,GAAG,GAAGnB,GAAG,CAAC,EAAEnoD,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,6EAA6EhmD,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,8EAA8EhmD,EAAE,OAAO,EAAE,IAAIkC,EAAE+7D,GAAGj+D,EAAE,CAAC,EAAE,GAAGO,IAAI,SAAS,MAAM,IAAI0lD,GAAG,+EAA+E,EAAE,OAAO/jD,EAAEuwC,GAAG,OAAO,CAAC,EAAEvwC,EAAE,OAAO,EAAE,QAAQ5B,EAAE,IAAIC,IAAI,OAAO,OAAO,QAAQ,UAAU,EAAE,WAAW,OAAO,KAAKL,EAAE,WAAW+B,CAAC,CAAC,EAAE,IAAI,kBAAkBC,EAAE2yB,GAAG3yB,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAGA,CAAC,CAAC,CAAC,CAAC,SAASm8D,GAAGr+D,EAAE,EAAEE,EAAEI,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE,QAAQ,EAAE,EAAE,CAAC,OAAOkzB,EAAE,IAAI,CAAC,GAAG,GAAG,OAAO,EAAE61B,GAAG,GAAGnB,GAAG,CAAC,EAAEnoD,EAAE,OAAO,GAAGA,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,mEAAmEhmD,EAAE,OAAO,EAAE,GAAG,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,oEAAoEhmD,EAAE,OAAO,EAAE,IAAIiC,EAAEi8D,GAAGl+D,EAAE,CAAC,EAAE,GAAGO,IAAI,SAAS,MAAM,IAAI0lD,GAAG,+EAA+E,EAAE,OAAOhkD,EAAEu9B,GAAGv9B,EAAE,EAAE3B,EAAEC,IAAI,OAAO,OAAO,QAAQ,QAAQ,CAAC,EAAEL,GAAG,OAAO+B,EAAEsoD,GAAGtoD,EAAE/B,CAAC,GAAG,IAAI,kBAAkB+B,EAAE4yB,GAAG5yB,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,GAAGA,CAAC,CAAC,CAAC,CAAC,IAAIq8D,GAAG,cAAcjR,EAAE,CAAC,YAAYvsD,EAAEZ,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,KAAK,KAAK,KAAK,2BAA2B,eAAe,KAAK,yBAAyB,QAAQo+D,GAAG,WAAWp+D,CAAC,EAAE,KAAK,KAAKY,EAAEumD,GAAG,KAAK,KAAK,MAAM,EAAE,KAAK,OAAO,GAAG,KAAK,OAAO,GAAG,KAAK,OAAO,EAAE,MAAM,IAAIpB,GAAG,qDAAqD,KAAK,+BAA+B,EAAE,GAAG,KAAK,WAAW6X,GAAG59D,EAAE,WAAWY,EAAE,YAAY,EAAE,KAAK,QAAQg9D,GAAG59D,EAAE,SAAS,KAAK,EAAEA,EAAE,QAAQY,EAAE,SAAS,EAAE,KAAK,QAAQZ,EAAE,SAAS,KAAK,QAAQA,EAAE,QAAQmoD,GAAG,KAAK,OAAO,EAAE,KAAK,WAAWnoD,EAAE,YAAY,KAAK,eAAeA,EAAE,WAAWioD,GAAG,KAAK,UAAU,EAAE,KAAK,WAAW2U,GAAG58D,EAAE,UAAU,EAAE,KAAK,QAAQA,EAAE,SAAS,KAAK,GAAGA,EAAE,QAAQ,KAAK,gBAAgBmsD,GAAGnsD,EAAE,iBAAiB,KAAK,wBAAwB,EAAE,KAAK,eAAekvD,GAAGlvD,EAAE,cAAc,EAAE,KAAK,gBAAgBq9D,GAAGr9D,EAAE,eAAe,EAAE,KAAK,oBAAoBq9D,GAAGr9D,EAAE,mBAAmB,EAAE,KAAK,aAAa49D,GAAG59D,EAAE,cAAc,KAAK,EAAEA,EAAE,aAAaY,EAAE,cAAc,EAAE,KAAK,OAAO,GAAG,MAAM,QAAQ,KAAK,YAAY,GAAG,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIklD,EAAE,iGAAiG,KAAK,UAAU,KAAK,YAAY,GAAG,EAAE,GAAG,KAAK,OAAO,GAAG,GAAG,OAAO,KAAK,cAAc,SAAS,KAAK,aAAa,CAAC,KAAK,aAAa,KAAK,YAAY,UAAU,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIA,EAAE,0FAA0F,KAAK,UAAU,KAAK,YAAY,GAAG,UAAU,KAAK,OAAO,GAAG,GAAG,OAAO,KAAK,cAAc,SAAS,KAAK,aAAa,CAAC,KAAK,aAAa,KAAK,aAAa,KAAK,YAAY,UAAU,KAAK,aAAa,SAAS,EAAE,MAAM,IAAIA,EAAE,4FAA4F,KAAK,UAAU,KAAK,YAAY,GAAG,EAAE,CAAC,OAAO,WAAWllD,EAAE,CAAC,GAAGulD,GAAG,eAAevlD,EAAE,yCAAyC,EAAE,OAAOA,EAAE,YAAY,UAAU,CAACsmD,GAAGtmD,EAAE,WAAW,SAAS,EAAE,CAAC,EAAE,MAAM,IAAIklD,EAAE,oGAAoG,KAAK,UAAUllD,EAAE,UAAU,IAAI,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,WAAW,KAAK,WAAW,QAAQ,KAAK,QAAQ,QAAQ,KAAK,QAAQ,WAAW,KAAK,WAAW,aAAa,KAAK,aAAa,WAAW87D,GAAG,KAAK,UAAU,EAAE,QAAQ,KAAK,QAAQ,gBAAgBxQ,GAAG,KAAK,eAAe,EAAE,gBAAgBiR,GAAG,KAAK,eAAe,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,eAAenO,GAAG,KAAK,cAAc,CAAC,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEy9D,GAAG,cAAcD,EAAE,CAAC,YAAYx9D,EAAEZ,EAAE,CAAC,MAAMY,EAAEZ,CAAC,EAAE,KAAK,OAAO,KAAKq+D,GAAG,WAAWr+D,CAAC,EAAE,KAAK,QAAQA,EAAE,QAAQmnD,GAAG,KAAK,QAAQ,SAAS,EAAE,KAAK,kBAAkBgF,GAAGnsD,EAAE,mBAAmB,KAAK,0BAA0B,EAAE,KAAK,iBAAiBkvD,GAAGlvD,EAAE,gBAAgB,EAAE,KAAK,kBAAkBq9D,GAAGr9D,EAAE,iBAAiB,CAAC,CAAC,MAAMY,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgB,EAAEY,EAAE,OAAO,EAAE,GAAGA,EAAEZ,IAAI,KAAK,MAAM,IAAI8lD,EAAE,+DAA+DllD,EAAEZ,IAAI,EAAE,IAAI,EAAEY,EAAEZ,GAAGK,EAAE,KAAK,WAAW,OAAO,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,KAAK,OAAO,KAAK,UAAU,SAASA,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,UAAU,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,OAAO,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,GAAG,KAAK,UAAU,CAAC,CAAC,KAAK,KAAK,KAAK,EAAE,KAAK,CAAC,CAACL,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0rD,GAAG1rD,CAAC,EAAE,IAAI,EAAEP,EAAE,KAAK,MAAM,KAAK,KAAK,KAAK,KAAK,KAAK,EAAEwB,EAAEylD,GAAG,KAAK,WAAW,aAAa,CAAC,EAAE,GAAGzlD,GAAG,MAAM,KAAK,OAAO,EAAE,EAAEq8D,GAAGt9D,EAAE,KAAK,OAAO,KAAK,EAAEP,EAAE,KAAK,QAAQ,KAAK,QAAQ,KAAK,WAAW,KAAK,aAAawB,CAAC,MAAM,CAAC,GAAG,KAAK,OAAO,EAAE,EAAEo8D,GAAGr9D,EAAE,KAAK,OAAO,KAAK,EAAEP,EAAE,KAAK,QAAQ,GAAG,KAAK,QAAQ,KAAK,WAAW,KAAK,aAAa,EAAE,UAAU,KAAK,OAAO,EAAE,EAAE69D,GAAGt9D,EAAE,KAAK,OAAO,KAAK,EAAEP,EAAE,KAAK,QAAQ,KAAK,QAAQ,KAAK,WAAW,KAAK,YAAY,UAAU,KAAK,OAAO,EAAE,EAAE89D,GAAGv9D,EAAE,KAAK,OAAO,KAAK,EAAEP,EAAE,KAAK,QAAQ,KAAK,QAAQ,KAAK,WAAW,KAAK,YAAY,MAAO,OAAM,IAAI0lD,GAAG,uDAAuD,EAAE,KAAK,YAAY,OAAO,EAAE,KAAK,WAAW,MAAM,CAAC,EAAE,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,mBAAmBnlD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,CAAC,EAAE,EAAE,KAAK,aAAa,eAAeY,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,EAAEA,EAAE,MAAM,CAAC,EAAE,QAAQiB,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE+7D,GAAG,EAAEh8D,GAAG,KAAK,WAAWA,GAAG,KAAK,QAAQ,KAAK,QAAQA,GAAG,OAAO,KAAK,cAAc,SAAS,KAAK,aAAa,KAAK,aAAaA,EAAE,EAAE7B,EAAE,KAAK8B,CAAC,CAAC,CAAC,IAAIzB,EAAE,CAACO,EAAE,EAAE,EAAE,OAAO,KAAK,aAAa,gBAAgBP,EAAEA,EAAE,OAAOL,CAAC,EAAEK,EAAE,KAAK,KAAK,OAAO,IAAIA,EAAE,KAAK,KAAK,OAAO,EAAEA,EAAEA,EAAE,OAAOL,CAAC,GAAGK,CAAC,CAAC,WAAW,CAAC,IAAIO,EAAE,CAAC,QAAQ,KAAK,QAAQ,kBAAkBsrD,GAAG,KAAK,iBAAiB,EAAE,kBAAkBiR,GAAG,KAAK,iBAAiB,EAAE,iBAAiBnO,GAAG,KAAK,gBAAgB,CAAC,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,OAAO,WAAWA,EAAE,CAAC,GAAG,EAAE,YAAYA,IAAI,OAAOA,EAAE,SAAS,UAAUA,EAAE,QAAQ,EAAE,MAAM,IAAIklD,EAAE,0EAA0E,KAAK,UAAUllD,EAAE,OAAO,GAAG,CAAC,CAAC,EAAE09D,GAAG,cAAcD,EAAE,CAAC,YAAYz9D,EAAE,CAAC,MAAM,EAAEA,CAAC,EAAE09D,GAAG,WAAW19D,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,KAAKA,CAAC,CAAC,OAAO,WAAWA,EAAE,CAAC,GAAG,OAAOA,EAAE,YAAY,UAAU,CAACsmD,GAAGtmD,EAAE,WAAW,SAAS,EAAE,CAAC,EAAE,MAAM,IAAIklD,EAAE,8FAA8F,KAAK,UAAUllD,EAAE,UAAU,IAAI,CAAC,CAAC,EAAE09D,GAAG,UAAU,SAAS/mC,GAAE,cAAc+mC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAYz9D,EAAE,CAAC,MAAM,EAAEA,CAAC,EAAE29D,GAAG,WAAW39D,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,KAAKA,CAAC,CAAC,OAAO,WAAWA,EAAE,CAAC,GAAG,OAAOA,EAAE,YAAY,UAAU,EAAE,MAAM,QAAQA,EAAE,UAAU,IAAIA,EAAE,WAAW,SAAS,GAAGA,EAAE,WAAW,SAAS,IAAI,MAAM,IAAIklD,EAAE,2FAA2F,KAAK,UAAUllD,EAAE,UAAU,IAAI,CAAC,CAAC,EAAE29D,GAAG,UAAU,SAAShnC,GAAE,cAAcgnC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY19D,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,UAAU,QAAQ,KAAK,UAAU,QAAQ,MAAM,IAAIhH,EAAE,uGAAuG,KAAK,SAAS,CAAC,CAAC,MAAMllD,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,mDAAmD,KAAK,UAAUllD,CAAC,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgB,EAAEY,EAAE,OAAO,EAAE,GAAGA,EAAEZ,IAAI,KAAK,MAAM,IAAI8lD,EAAE,sEAAsE,EAAE,IAAI,EAAEllD,EAAEZ,GAAGK,EAAE,KAAK,WAAW,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,EAAE,KAAK,OAAO,KAAK,UAAU,SAASA,EAAE,UAAU,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,UAAU,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,OAAO,EAAE,UAAU,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,GAAG,KAAK,UAAU,CAAC,IAAIysD,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC9sD,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAIklD,EAAE,2FAA2F,EAAE,MAAM,QAAQ,EAAE,IAAIzlD,EAAE,EAAE,MAAMwB,EAAExB,EAAE,GAAGyB,EAAEC,EAAE,KAAK,aAAa,iBAAiBD,EAAE,EAAEC,EAAE,IAAID,EAAE,EAAEC,EAAE,GAAG,IAAIC,EAAE3B,EAAEyB,GAAG,EAAEzB,EAAE0B,GAAGW,EAAE,KAAK,WAAW,GAAG,EAAE,KAAK,WAAW,GAAGE,EAAE,KAAK,QAAQ,GAAGC,EAAE,KAAK,QAAQ,GAAGC,EAAEg7D,GAAG97D,EAAEY,EAAEF,EAAE,KAAK,OAAO,EAAEsB,EAAE85D,GAAG,EAAEj7D,EAAE,EAAE,KAAK,OAAO,EAAEoB,EAAE,CAACpC,EAAEiB,EAAEkB,EAAE,KAAK,OAAO,EAAE,KAAK,aAAa,iBAAiB,EAAE2wB,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,IAAIttB,EAAE+3B,GAAG,EAAE,KAAK,OAAO,KAAK,EAAEn7B,EAAE,KAAK,QAAQ,KAAK,OAAO,EAAE,OAAO,KAAK,aAAa,iBAAiBoD,EAAEstB,GAAGttB,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,KAAK,MAAM,OAAOA,EAAEgjD,GAAGhjD,EAAE,KAAK,KAAK,KAAK,EAAE,KAAK,UAAU,GAAG,KAAK,YAAY,OAAOA,EAAE,KAAK,WAAW,MAAMA,CAAC,GAAGA,CAAC,CAAC,CAAC,CAAC,mBAAmBzG,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,EAAE,EAAEP,EAAEwB,EAAE,KAAK,aAAa,iBAAiB,EAAE,EAAExB,EAAE,EAAEwB,EAAE,IAAI,EAAE,EAAExB,EAAE,EAAEwB,EAAE,GAAG,IAAIC,EAAE,KAAK,WAAW,GAAGC,EAAE,KAAK,WAAW,GAAGC,EAAE,KAAK,QAAQ,GAAG,EAAE,KAAK,QAAQ,GAAG,OAAOhC,EAAE,GAAG,KAAK,QAAQA,EAAEK,GAAGy9D,GAAG99D,EAAEK,GAAG2B,EAAEF,EAAE,KAAK,OAAO,EAAE9B,EAAE6B,GAAGi8D,GAAG99D,EAAE6B,GAAG,EAAEE,EAAE,KAAK,OAAO,EAAE/B,CAAC,CAAC,WAAW,CAAC,IAAIY,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,aAAaA,CAAC,CAAC,EAAE49D,GAAG,UAAU,kBAAkBjnC,GAAE,cAAcinC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY39D,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,UAAU,QAAQ,KAAK,UAAU,QAAQ,MAAM,IAAIhH,EAAE,uGAAuG,KAAK,SAAS,CAAC,CAAC,MAAMllD,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,mDAAmD,KAAK,UAAUllD,CAAC,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgB,EAAEY,EAAE,OAAO,EAAE,GAAGA,EAAEZ,IAAI,KAAK,MAAM,IAAI8lD,EAAE,sEAAsE,EAAE,IAAI,EAAEllD,EAAEZ,GAAGK,EAAE,KAAK,WAAW,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,EAAE,KAAK,OAAO,KAAK,UAAU,SAASA,EAAE,UAAU,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,UAAU,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,OAAO,EAAE,UAAU,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,GAAG,KAAK,UAAU,CAAC,IAAIysD,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC9sD,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAIklD,EAAE,2FAA2F,EAAE,MAAM,QAAQ,EAAE,IAAIzlD,EAAE,EAAE,MAAMwB,EAAExB,EAAE,GAAGyB,EAAEC,EAAEC,EAAE,KAAK,aAAa,iBAAiBA,EAAE,EAAEF,EAAE,EAAEC,EAAE,IAAIC,EAAE,EAAEF,EAAE,EAAEC,EAAE,GAAG,IAAI,EAAE1B,EAAE2B,GAAGU,EAAErC,EAAEyB,GAAG,EAAEzB,EAAE0B,GAAGa,EAAE,KAAK,WAAW,GAAGC,EAAE,KAAK,WAAW,GAAGC,EAAE,KAAK,WAAW,GAAGkB,EAAE,KAAK,QAAQ,GAAGC,EAAE,KAAK,QAAQ,GAAGoD,EAAE,KAAK,QAAQ,GAAGxC,EAAEi5D,GAAG,EAAE95D,EAAEpB,EAAE,KAAK,OAAO,EAAE6B,EAAEq5D,GAAGp7D,EAAEuB,EAAEpB,EAAE,KAAK,OAAO,EAAE6B,EAAEo5D,GAAG,EAAEz2D,EAAEvE,EAAE,KAAK,OAAO,EAAE6B,EAAE,CAAC9C,EAAEgD,EAAEJ,EAAEC,EAAE,KAAK,OAAO,EAAE,KAAK,aAAa,iBAAiB,EAAEiwB,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,IAAI/vB,EAAE86B,GAAG,EAAE,KAAK,OAAO,KAAK,EAAE/6B,EAAE,KAAK,QAAQ,KAAK,OAAO,EAAE,OAAO,KAAK,aAAa,iBAAiBC,EAAE+vB,GAAG/vB,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,KAAK,OAAO,OAAOA,EAAEylD,GAAGzlD,EAAE,KAAK,KAAK,KAAK,EAAE,KAAK,UAAU,GAAG,KAAK,aAAa,OAAOA,EAAE,KAAK,WAAW,MAAMA,CAAC,GAAGA,CAAC,CAAC,CAAC,CAAC,mBAAmBhE,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,EAAE,EAAEP,EAAEwB,EAAEC,EAAE,KAAK,aAAa,iBAAiB,EAAE,EAAEzB,EAAE,EAAEwB,EAAE,EAAEC,EAAE,IAAI,EAAE,EAAEzB,EAAE,EAAEwB,EAAE,EAAEC,EAAE,GAAG,IAAIC,EAAE,KAAK,WAAW,GAAGC,EAAE,KAAK,WAAW,GAAG,EAAE,KAAK,WAAW,GAAGU,EAAE,KAAK,QAAQ,GAAG,EAAE,KAAK,QAAQ,GAAGE,EAAE,KAAK,QAAQ,GAAG,OAAO5C,EAAE,GAAG,KAAK,QAAQA,EAAEK,GAAGy9D,GAAG99D,EAAEK,GAAGqC,EAAEX,EAAE,KAAK,OAAO,EAAE/B,EAAE6B,GAAGi8D,GAAG99D,EAAE6B,GAAG,EAAEG,EAAE,KAAK,OAAO,EAAEhC,EAAE8B,GAAGg8D,GAAG99D,EAAE8B,GAAGc,EAAE,EAAE,KAAK,OAAO,EAAE5C,CAAC,CAAC,WAAW,CAAC,IAAIY,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,aAAaA,CAAC,CAAC,EAAE69D,GAAG,UAAU,kBAAkBlnC,GAAE,cAAcknC,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,YAAYz9D,EAAEZ,EAAE,CAAC,GAAG,MAAMY,EAAEZ,CAAC,EAAE,KAAK,8BAA8B,gBAAgB,KAAK,8BAA8B,gBAAgB,KAAK,gBAAgB,KAAK,KAAK,gBAAgB,KAAKA,EAAE,SAAS,KAAK,MAAM,IAAI8lD,EAAE,qFAAqF,EAAE,GAAG9lD,EAAE,mBAAmB,MAAMA,EAAE,mBAAmB,MAAMA,EAAE,kBAAkB,KAAK,MAAM,IAAI8lD,EAAE,oPAAoP,EAAE,GAAG9lD,EAAE,SAAS,MAAMA,EAAE,UAAU,QAAQA,EAAE,UAAU,QAAQ,MAAM,IAAI8lD,EAAE,gBAAgB,KAAK,uEAAuE,KAAK,UAAU9lD,EAAE,OAAO,GAAG,EAAE,KAAK,gBAAgBA,EAAE,iBAAiB,KAAK,EAAEA,EAAE,gBAAgB,KAAK,qBAAqBmsD,GAAGnsD,EAAE,sBAAsB,KAAK,6BAA6B,EAAE,KAAK,qBAAqBq9D,GAAGr9D,EAAE,oBAAoB,EAAE,KAAK,oBAAoBkvD,GAAGlvD,EAAE,mBAAmB,EAAE,KAAK,qBAAqBmsD,GAAGnsD,EAAE,sBAAsB,KAAK,6BAA6B,EAAE,KAAK,qBAAqBq9D,GAAGr9D,EAAE,oBAAoB,EAAE,KAAK,oBAAoBkvD,GAAGlvD,EAAE,mBAAmB,CAAC,CAAC,MAAMY,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAEA,EAAE,OAAO,KAAK,KAAK,EAAE,MAAM,IAAIklD,EAAE,0BAA0B,KAAK,0BAA0B,KAAK,KAAK,gCAAgC,KAAK,UAAUllD,CAAC,GAAG,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgB,EAAEY,EAAE,OAAO,EAAE,GAAGA,EAAEZ,IAAI,MAAMY,EAAEZ,GAAG,EAAE,MAAM,IAAI8lD,EAAE,oEAAoE,KAAK,UAAUllD,EAAEZ,EAAE,GAAG,EAAE,IAAI,EAAEY,EAAEZ,GAAGK,EAAE,KAAK,WAAW,OAAO,CAAC,EAAE,KAAK,eAAe,CAAC,EAAEwB,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAE,KAAK,KAAK,EAAEA,EAAEF,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAK,EAAE,KAAK,gBAAgB,KAAK,OAAO,EAAE,IAAIC,EAAE,GAAG,KAAK,gBAAgB,KAAK,UAAU,mBAAmBzB,EAAE,UAAU,KAAK,qBAAqB,KAAK,qBAAqByB,EAAE,KAAK,mBAAmB,EAAE,KAAK,gBAAgB,KAAK,UAAU,mBAAmBD,EAAE,UAAU,KAAK,qBAAqB,KAAK,qBAAqBC,EAAE,KAAK,mBAAmB,EAAE,KAAK,QAAQ,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,OAAO,EAAE,UAAU,KAAK,gBAAgB,KAAK,gBAAgBA,EAAE,KAAK,cAAc,EAAE,KAAK,KAAK,KAAK,KAAK,UAAU,CAAC,IAAIgrD,GAAG,CAAC,KAAK,KAAK,KAAK,EAAE,KAAK,CAAC,CAAC9sD,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0rD,GAAG1rD,CAAC,EAAE,IAAI,EAAE,GAAG,KAAK,OAAO,EAAE,MAAM,IAAImlD,GAAG,kDAAkD,EAAE,OAAO,KAAK,OAAO,IAAI,KAAK,aAAa,kBAAkBnlD,EAAE+zB,GAAG/zB,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,EAAE4sC,GAAG5sC,EAAE,KAAK,gBAAgB,KAAK,EAAE,KAAK,gBAAgB,KAAK,EAAE,KAAK,QAAQ,KAAK,QAAQ,KAAK,aAAa,MAAM,GAAG,KAAK,UAAU,EAAEypD,GAAG,EAAE,KAAK,KAAK,KAAK,EAAE,KAAK,UAAU,GAAG,KAAK,YAAY,OAAO,EAAE,KAAK,WAAW,MAAM,CAAC,GAAG,KAAK,aAAa,kBAAkB,EAAE11B,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI/zB,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,KAAK,OAAOA,EAAE,kBAAkB,OAAOA,EAAE,kBAAkB,OAAOA,EAAE,iBAAiBA,EAAE,qBAAqBsrD,GAAG,KAAK,oBAAoB,EAAEtrD,EAAE,qBAAqBsrD,GAAG,KAAK,oBAAoB,EAAEtrD,EAAE,qBAAqBu8D,GAAG,KAAK,oBAAoB,EAAEv8D,EAAE,qBAAqBu8D,GAAG,KAAK,oBAAoB,EAAEv8D,EAAE,oBAAoBouD,GAAG,KAAK,mBAAmB,EAAEpuD,EAAE,oBAAoBouD,GAAG,KAAK,mBAAmB,EAAEpuD,CAAC,CAAC,EAAE89D,GAAG,UAAU,gBAAgB,IAAIC,GAAG,cAAcD,EAAE,CAAC,YAAY99D,EAAE,CAAC,MAAM,EAAEA,CAAC,CAAC,CAAC,EAAE+9D,GAAG,UAAU,kBAAkBpnC,GAAE,cAAconC,EAAE,EAAE,IAAIC,GAAG,cAAcP,EAAE,CAAC,YAAYz9D,EAAE,CAAC,MAAM,EAAEA,CAAC,EAAEg+D,GAAG,WAAWh+D,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAE,OAAO,OAAOA,EAAE,KAAK,OAAOA,EAAE,WAAWA,CAAC,CAAC,OAAO,WAAWA,EAAE,CAAC,GAAG,OAAOA,EAAE,YAAY,UAAU,CAACsmD,GAAGtmD,EAAE,WAAW,SAAS,EAAE,CAAC,EAAE,MAAM,IAAIklD,EAAE,yFAAyF,KAAK,UAAUllD,EAAE,UAAU,IAAI,CAAC,CAAC,EAAEg+D,GAAG,UAAU,SAASrnC,GAAE,cAAcqnC,EAAE,EAAE,IAAIC,GAAG,cAAc1R,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,OAAOA,EAAE,UAAU,SAAS,KAAK,SAAS,CAAC,CAACA,EAAE,SAASA,EAAE,QAAQ,EAAE,CAACA,EAAE,SAASA,EAAE,QAAQ,CAAC,EAAE,OAAOA,EAAE,SAAS,IAAI,SAAS,KAAK,SAAS,CAAC,CAACA,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,EAAE,CAACA,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,CAAC,EAAE,KAAK,SAASA,EAAE,SAAS,KAAK,WAAWA,EAAE,aAAa,OAAO,eAAeA,EAAE,WAAW,KAAK,UAAU,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,OAAO,KAAK,aAAa,gBAAgB,CAACA,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAGA,EAAE,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,EAAE,EAAE,CAACA,EAAE,GAAGA,EAAE,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAGA,EAAE,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAGA,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAE0rD,GAAG1rD,CAAC,EAAE,KAAK,aAAa,eAAe,CAAC,IAAI,EAAEipD,GAAGjpD,EAAE,KAAK,SAAS,GAAG,GAAGA,EAAE,MAAM,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAG,CAAC,EAAE,OAAOipD,GAAG,EAAE,KAAK,SAAS,GAAG,GAAGjpD,EAAE,MAAM,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAG,CAAC,CAAC,KAAK,CAAC,IAAI,EAAEipD,GAAGjpD,EAAE,KAAK,SAAS,GAAG,GAAGA,EAAE,MAAM,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAG,CAAC,EAAE,OAAOipD,GAAG,EAAE,KAAK,SAAS,GAAG,GAAGjpD,EAAE,MAAM,GAAG,KAAK,SAAS,GAAG,GAAG,KAAK,SAAS,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,SAAS,KAAK,SAAS,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEi+D,GAAG,UAAU,aAAatnC,GAAE,cAAcsnC,EAAE,EAAE,IAAIC,GAAG,cAAc3R,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,aAAa,CAAC,EAAE,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,KAAK,KAAKA,EAAE,MAAM,KAAK,KAAK,aAAaA,EAAE,KAAK,KAAK,WAAWA,EAAE,YAAY,KAAK,eAAeA,EAAE,WAAWqnD,GAAG,KAAK,UAAU,EAAE,KAAK,cAAcrnD,EAAE,eAAe,KAAK,UAAUA,EAAE,cAAcsnD,GAAG,KAAK,aAAa,CAAC,CAAC,mBAAmBtnD,EAAE,CAAC,GAAG,KAAK,aAAa,gBAAgB,CAAC,IAAIZ,EAAEY,EAAE,IAAI,KAAK,KAAK,KAAK,KAAK,GAAGA,EAAE,GAAG,EAAEA,EAAE,IAAI,KAAK,KAAK,KAAK,KAAK,GAAGA,EAAE,GAAG,MAAM,CAACA,EAAE,GAAGA,EAAE,GAAGZ,EAAE,CAAC,CAAC,KAAK,CAAC,IAAIA,EAAEY,EAAE,IAAI,KAAK,KAAK,KAAK,KAAK,GAAGA,EAAE,GAAG,EAAEA,EAAE,IAAI,KAAK,KAAK,KAAK,KAAK,GAAGA,EAAE,GAAG,MAAM,CAACA,EAAE,GAAGZ,EAAE,EAAEY,EAAE,EAAE,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAEP,EAAE,EAAE,MAAM,GAAG,KAAK,aAAa,gBAAgB,CAAC,EAAEs0B,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,IAAI9yB,EAAE,KAAK,KAAK,GAAGxB,EAAE,GAAGyB,EAAE,KAAK,KAAK,GAAGzB,EAAE,GAAG0B,EAAE,KAAK,gBAAgB,UAAUs3C,GAAG,sBAAsB,EAAE,CAACx3C,EAAEC,CAAC,CAAC,EAAEu3C,GAAG,eAAe,EAAE,CAACx3C,EAAEC,CAAC,CAAC,EAAE,OAAO6yB,GAAG5yB,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,IAAIF,EAAE,KAAK,KAAK,GAAGxB,EAAE,GAAGyB,EAAE,KAAK,KAAK,GAAGzB,EAAE,GAAG,OAAO,KAAK,gBAAgB,UAAUg5C,GAAG,sBAAsB,EAAE,CAACx3C,EAAEC,CAAC,CAAC,EAAEu3C,GAAG,eAAe,EAAE,CAACx3C,EAAEC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIlB,EAAE,CAAC,KAAK,KAAK,KAAK,WAAW,KAAK,WAAW,cAAc,KAAK,aAAa,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEk+D,GAAG,UAAU,eAAevnC,GAAE,cAAcunC,EAAE,EAAE,SAASC,GAAGj/D,EAAE,EAAEE,EAAE,CAAC,EAAE,CAAC,EAAEI,EAAE,QAAQC,EAAE,EAAE,CAAC,OAAOkzB,EAAE,IAAI,CAAClzB,GAAG,OAAOA,EAAE+oD,GAAG,GAAGnB,GAAG5nD,CAAC,EAAE,IAAI,EAAE09D,GAAGj+D,EAAEO,CAAC,EAAE,GAAGP,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,mEAAmEhmD,EAAE,QAAQ,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,yDAAyD,EAAE,QAAQ,EAAE,OAAO,EAAEtlB,GAAG,EAAE,EAAExgC,EAAEI,IAAI,OAAO,OAAO,QAAQ,OAAO,CAAC,EAAEC,IAAI,kBAAkB,EAAEs0B,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAIqqC,GAAG,cAAcZ,EAAE,CAAC,YAAYx9D,EAAE,CAAC,MAAM,EAAEA,CAAC,EAAE,KAAK,gBAAgB,KAAK,KAAK,gBAAgBA,EAAE,iBAAiB,KAAK,EAAEA,EAAE,gBAAgB,KAAK,qBAAqBurD,GAAGvrD,EAAE,sBAAsB,KAAK,0BAA0B,EAAE,KAAK,oBAAoBsuD,GAAGtuD,EAAE,mBAAmB,EAAE,KAAK,qBAAqBy8D,GAAGz8D,EAAE,oBAAoB,CAAC,CAAC,MAAMA,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAEA,EAAE,OAAO,EAAE,MAAM,IAAIklD,EAAE,uEAAuE,KAAK,UAAUllD,CAAC,IAAI,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgB,EAAE,EAAE,GAAGY,EAAEZ,IAAI,MAAMY,EAAEZ,GAAG,EAAE,MAAM,IAAI8lD,EAAE,yFAAyFllD,EAAEZ,MAAM,EAAE,IAAI,EAAEY,EAAEZ,GAAGK,EAAE,CAAC,KAAK,WAAW,GAAG,KAAK,WAAW,GAAG,EAAE,KAAK,eAAe,EAAE,KAAK,gBAAgB,KAAK,UAAU,mBAAmBA,EAAE,KAAK,KAAK,qBAAqB,KAAK,qBAAqB,GAAG,KAAK,mBAAmB,EAAE,KAAK,QAAQ,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,EAAE,KAAK,eAAe,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,EAAE,KAAK,KAAK,KAAK,KAAK,MAAM,EAAE,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0rD,GAAG1rD,CAAC,EAAE,IAAI,EAAEm+D,GAAGn+D,EAAE,KAAK,gBAAgB,KAAK,EAAE,KAAK,QAAQ,KAAK,QAAQ,KAAK,WAAW,IAAI,EAAE,OAAO,KAAK,UAAU,EAAEypD,GAAG,EAAE,KAAK,KAAK,KAAK,EAAE,KAAK,UAAU,GAAG,KAAK,YAAY,OAAO,EAAE,KAAK,WAAW,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,mBAAmBzpD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgBY,EAAE,GAAGA,EAAE,GAAG,EAAE,KAAK,aAAa,gBAAgBA,EAAE,GAAGA,EAAE,GAAGP,EAAE,KAAK,aAAa,gBAAgBO,EAAE,GAAG,KAAK,gBAAgBA,EAAE,GAAG,KAAK,gBAAgBiB,EAAEg8D,GAAG79D,EAAE,KAAK,WAAW,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE8B,EAAE+7D,GAAG,EAAE,KAAK,WAAW,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,OAAO,KAAK,aAAa,gBAAgB,CAACj9D,EAAE,GAAGP,EAAEwB,EAAEC,CAAC,EAAE,CAAClB,EAAE,GAAGiB,EAAEC,EAAEzB,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIO,EAAE,MAAM,UAAU,EAAE,OAAOA,EAAE,gBAAgB,KAAK,gBAAgBA,EAAE,qBAAqBsrD,GAAG,KAAK,oBAAoB,EAAEtrD,EAAE,qBAAqBu8D,GAAG,KAAK,oBAAoB,EAAEv8D,EAAE,oBAAoBouD,GAAG,KAAK,oBAAoB,EAAEpuD,CAAC,CAAC,EAAEo+D,GAAG,UAAU,kBAAkBznC,GAAE,cAAcynC,EAAE,EAAE,SAASC,GAAGn/D,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,MAAM,QAAQN,CAAC,EAAE,CAAC,GAAG,GAAG,MAAME,GAAG,KAAK,MAAM,IAAI8lD,EAAE,+EAA+E,EAAE1lD,GAAG,OAAOJ,EAAEF,EAAE,MAAMA,EAAE,OAAOM,EAAEN,EAAE,MAAM,EAAEA,EAAEA,EAAE,MAAM,EAAEA,EAAE,OAAOM,CAAC,GAAGN,EAAE,OAAO,IAAI,EAAEA,EAAE,MAAM,EAAEA,EAAE,MAAM,GAAGA,EAAEA,EAAE,EAAE,CAAC,SAASO,EAAE,EAAE,CAAC,OAAO,GAAG,MAAM,MAAM,QAAQ,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO,EAAEA,EAAE,CAAC,EAAEL,EAAEK,EAAEL,CAAC,EAAE,CAAC,OAAOF,EAAE,aAAa,EAAE,UAAUE,CAAC,CAAC,CAAC,SAASk/D,GAAGp/D,EAAE,EAAEE,EAAEI,EAAE,GAAGC,EAAE,EAAE,EAAE,GAAG0B,EAAE,GAAG,CAAC,OAAOwxB,EAAE,IAAI,CAAC,IAAIvxB,EAAE,EAAE,MAAM,OAAO,GAAGA,EAAE,EAAE,MAAM,IAAI8jD,EAAE,uCAAuC9jD,KAAK,EAAE,IAAIC,EAAE,CAAC,EAAE,CAAC,EAAE,OAAOgnD,GAAG,EAAEjnD,CAAC,CAAC,EAAE,GAAG,EAAE2yB,GAAG,EAAE1yB,CAAC,EAAE,GAAG,KAAK,MAAM,IAAI8jD,GAAG,gFAAgF,EAAE,GAAG,QAAQ,KAAK,mGAAmG,EAAE1lD,GAAG,OAAOA,EAAEkwB,GAAEA,GAAElwB,EAAE,MAAM,EAAE,SAAS,EAAEA,EAAE,OAAO2B,EAAE,IAAI3B,EAAEqjC,GAAGrjC,EAAE,EAAE,GAAGA,EAAEs0B,GAAGt0B,EAAE4B,CAAC,GAAG7B,IAAI,EAAEosC,GAAG,EAAE,CAAC,EAAEnsC,GAAG,OAAOA,EAAEmsC,GAAGnsC,EAAE,CAAC,IAAI,IAAIqC,EAAE,CAAC,EAAEC,EAAEC,EAAE5C,EAAE6C,EAAE,EAAE,MAAM,GAAGC,EAAEiuC,GAAG,CAAC,EAAE/sC,EAAE3D,GAAG,OAAO2D,EAAE+sC,GAAG1wC,CAAC,GAAG,QAAQgH,EAAE,EAAEA,EAAExE,EAAE,EAAEwE,EAAE,CAAC,IAAI,EAAEvE,EAAEuE,GAAG5C,EAAE8uB,EAAE,IAAIzzB,EAAE,EAAE8C,CAAC,CAAC,EAAE,GAAGvC,GAAG,KAAKsC,EAAE8B,EAAE,GAAG7B,EAAE6B,EAAE,OAAO,CAAC,IAAIC,EAAE6uB,EAAE,IAAI,CAAC,IAAI5uB,EAAEX,EAAEqD,GAAGzC,EAAE4hC,GAAG6C,GAAG1kC,CAAC,EAAEA,CAAC,EAAE2C,EAAEsxB,EAAEM,EAAEz0B,EAAE,GAAGE,CAAC,EAAEu0B,EAAEt2B,EAAE,GAAGgC,CAAC,CAAC,EAAEE,EAAElC,EAAE,IAAI,CAAC2E,EAAExC,IAAI6zB,EAAEM,EAAEz0B,EAAE,GAAGM,GAAGJ,CAAC,EAAEu0B,EAAE3xB,EAAE3C,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,OAAO0C,EAAE,UAAUxC,CAAC,CAAC,CAAC,EAAEnC,EAAE+B,EAAE,OAAO9B,EAAE8B,EAAE,SAAS,CAAC3C,GAAGW,EAAE,KAAKC,CAAC,CAAC,CAAC,IAAI,EAAE,OAAOZ,IAAI,EAAE2tC,GAAGhtC,EAAE,CAAC,GAAG,CAACC,EAAE,EAAEC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIwtD,GAAG,cAAcjD,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,IAAIZ,EAAE,GAAGY,EAAE,MAAM,KAAK,MAAM,IAAIklD,EAAE,sDAAsD,EAAE,GAAG,MAAM,QAAQllD,EAAE,IAAI,EAAEZ,EAAE,IAAIm/D,GAAG,CAAC,MAAMv+D,EAAE,IAAI,CAAC,EAAEZ,EAAEY,EAAE,KAAKZ,EAAE,WAAW,KAAK,MAAM,IAAI8lD,EAAE,mGAAmG,EAAE,KAAK,KAAK9lD,EAAE,KAAK,gBAAgBY,EAAE,iBAAiB,KAAK,GAAGA,EAAE,gBAAgB,KAAK,YAAYA,EAAE,aAAa,KAAK,GAAGA,EAAE,YAAY,KAAK,YAAYA,EAAE,aAAa,KAAK,GAAGA,EAAE,YAAY,KAAK,UAAUA,EAAE,UAAU,KAAK,GAAGA,EAAE,SAAS,KAAK,OAAOA,EAAE,QAAQ,KAAK,GAAGA,EAAE,OAAO,KAAK,gBAAgB,GAAG,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,UAAU,KAAK,KAAK,QAAQ,KAAK,KAAK,aAAa,KAAK,KAAK,WAAW,CAAC,CAAC,CAAC,WAAW,CAAC,GAAG,KAAK,SAAS,KAAK,CAAC,IAAIlsD,EAAE,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,KAAK,UAAU,OAAO,EAAE,OAAOqoD,GAAG,EAAEroD,CAAC,EAAE,IAAIZ,GAAG,IAAI,CAAC,KAAM,QAAO,KAAK,OAAO,CAAC,UAAUY,EAAE,CAAC,KAAK,QAAQA,CAAC,CAAC,mBAAmBA,EAAE,CAACwrD,GAAGxrD,CAAC,IAAIA,EAAEA,EAAE,IAAIA,EAAEA,EAAE,IAAIZ,EAAE,KAAK,KAAK,UAAU,MAAM,QAAQA,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAG,IAAI,EAAEA,EAAE,GAAGK,EAAE,GAAG,KAAK,gBAAgBA,EAAE,CAACO,EAAE,GAAGA,EAAE,GAAG,CAAC,EAAEP,EAAE,CAACO,EAAE,GAAG,CAAC,EAAE,KAAK,YAAY,CAAC,IAAIiB,EAAE,CAAC,EAAE,QAAQC,KAAK9B,EAAE6B,EAAE,KAAK,CAACjB,EAAE,GAAGkB,CAAC,CAAC,EAAE,MAAM,CAACzB,CAAC,EAAE,OAAOwB,CAAC,CAAC,KAAM,QAAOxB,CAAC,CAAC,YAAYO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,MAAM,QAAQvzB,CAAC,IAAIA,EAAEA,EAAE,IAAI,IAAI,EAAE,KAAK,gBAAgBA,EAAE,KAAK,GAAG,KAAK,YAAY,CAAC,IAAIK,EAAE,KAAK,OAAO,IAAIwB,GAAG,IAAI,EAAE,MAAM,CAAC,CAAC,EAAE,OAAOxB,CAAC,CAAC,KAAM,QAAO,CAAC,CAAC,CAAC,CAAC,IAAI,QAAQ,CAAC,GAAG,KAAK,SAAS,KAAK,CAAC,IAAIO,EAAE,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,KAAK,UAAU,OAAO,EAAEZ,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEY,EAAE,EAAE,EAAEZ,EAAE,KAAK,IAAI,EAAE,OAAOA,CAAC,KAAM,QAAO,KAAK,OAAO,CAAC,IAAI,OAAOY,EAAE,CAAC,KAAK,QAAQA,CAAC,CAAC,MAAMA,EAAE,CAAC,GAAG,KAAK,cAAc,KAAK,MAAM,IAAImlD,GAAG,kDAAkD,EAAEqG,GAAGxrD,CAAC,IAAIA,EAAEA,EAAE,IAAIA,EAAEA,EAAE,IAAIR,EAAE,KAAK,SAASQ,EAAE,GAAG,KAAKP,EAAEO,EAAE,MAAM,CAAC,EAAE,KAAK,UAAU,GAAG,IAAIksD,GAAG,CAAC,MAAM,CAAC1sD,EAAE,KAAK,GAAGC,CAAC,CAAC,CAAC,EAAE,IAAIwB,EAAE,CAACjB,EAAE,EAAE,EAAE,OAAOA,EAAE,MAAM,CAAC,CAAC,EAAE,KAAK,KAAK,MAAMiB,CAAC,EAAE,IAAIC,EAAE,GAAG,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAEA,EAAE,KAAK,KAAK,UAAUA,EAAE,CAAC,KAAK,KAAK,SAAS,EAAE,KAAK,WAAW,MAAM,GAAG,CAAC8kB,EAAE,YAAY,KAAK,UAAU,IAAI7kB,GAAGA,EAAE,MAAMA,EAAE,MAAM,OAAO,EAAE,EAAED,CAAC,EAAE,MAAM,IAAIgkD,EAAE,6FAA6F,KAAK,wCAAwC,KAAK,KAAK,WAAW,OAAO,KAAK,UAAUhkD,EAAE,IAAIC,GAAG,IAAI+qD,GAAG,CAAC,MAAM,CAAC,KAAK/qD,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,UAAU,KAAK,YAAY,CAAC,CAAC,YAAYnB,EAAEZ,EAAE,GAAG,CAACuzB,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,SAAS,MAAM,IAAIqyB,GAAG,iEAAiE,EAAE,IAAI,EAAE,KAAK,UAAU,GAAG,MAAM,GAAG,GAAG,GAAG,KAAK,MAAM,IAAIE,EAAE,uUAAuU,EAAE,GAAG,KAAK,SAAS,KAAK,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,QAAQ,KAAK,KAAK,UAAU,IAAIzlD,GAAG8nC,GAAG,CAAC,EAAE9nC,CAAC,CAAC,CAAC,EAAE,KAAK,QAAQ,CAAC8nC,GAAG,CAAC,EAAE,KAAK,KAAK,SAAS,CAAC,CAAC,UAAUvnC,GAAG,KAAK4yB,GAAG,KAAK,OAAO,EAAE,KAAK,YAAY,OAAOA,GAAG,KAAK,UAAU,EAAE,KAAK,WAAW,CAAC,GAAG,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,QAAQ,KAAK,KAAK,UAAU,IAAInzB,GAAG8nC,GAAG,CAAC,EAAE9nC,CAAC,CAAC,CAAC,EAAE,KAAK,QAAQ,GAAG8nC,GAAG,CAAC,EAAE,KAAK,KAAK,SAAS,CAAC,MAAM,CAAC,GAAG,MAAM,QAAQvnC,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAGA,EAAE,SAAS,KAAK,QAAQ,OAAO,MAAM,IAAIklD,EAAE,SAAS,KAAK,gBAAgB,KAAK,QAAQ,oCAAoCllD,EAAE,0CAA0CA,GAAG,EAAEZ,IAAI,GAAG,KAAK,WAAW,KAAK,KAAK,QAAQ,MAAM,CAAC,EAAEwzB,GAAG,KAAK,OAAO,EAAE,QAAQnzB,EAAE,EAAEA,EAAE,KAAK,QAAQ,OAAO,EAAEA,EAAE,CAAC,IAAIwB,EAAEjB,EAAEP,GAAGyB,EAAE,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,KAAK,UAAUzB,GAAG,KAAK,KAAK,UAAU0B,EAAE,CAAC,EAAED,CAAC,EAAE,GAAG,CAAC8kB,EAAE,YAAY/kB,EAAE,MAAME,CAAC,EAAE,MAAM,IAAI+jD,EAAE,SAASzlD,gCAAgC,KAAK,wBAAwB0B,qBAAqBF,EAAE,OAAO,EAAE,KAAK,QAAQxB,GAAGwB,CAAC,CAAC,CAAC,KAAK,QAAQ,KAAK,QAAQ,IAAIxB,GAAGozB,GAAGpzB,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAMO,EAAEZ,EAAE,CAAC,IAAI,EAAEA,GAAG,KAAK,KAAKA,EAAE,aAAaK,EAAEL,GAAG,KAAK,KAAKA,EAAE,UAAUA,GAAG,OAAOA,EAAE,CAAC,GAAG,IAAI6B,EAAEo9D,GAAGr+D,EAAE,EAAEP,EAAE,KAAK,YAAY,EAAEO,EAAEiB,EAAE,OAAO,EAAEA,EAAE,aAAaxB,EAAEwB,EAAE,UAAU,IAAIC,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC/B,EAAE,aAAa,EAAE8B,EAAEA,EAAE,OAAO,CAAC,EAAE,KAAK,UAAU,CAAC,EAAE,QAAQG,KAAK,EAAE,KAAK,UAAU,KAAK,IAAI6qD,GAAG,CAAC,MAAM7qD,EAAE,KAAK,CAAC,CAAC,EAAEF,EAAEA,EAAE,OAAO,KAAK,SAAS,CAAC,CAAC,GAAG1B,GAAG,OAAOL,EAAE,UAAUK,EAAEyB,EAAEA,EAAE,OAAOzB,CAAC,EAAE,KAAK,aAAaA,EAAE,QAAQyB,EAAE,aAAairD,GAAG,CAAC,IAAI9qD,EAAE,CAACrB,CAAC,EAAE,OAAOkB,CAAC,EAAEY,EAAE,KAAK,UAAU,OAAOX,CAAC,EAAEY,EAAE,KAAK,UAAU,KAAK,UAAUD,EAAE,IAAIE,EAAE,MAAM,MAAMX,EAAEjC,CAAC,EAAE,OAAO,KAAK,UAAU2C,EAAEC,CAAC,KAAM,QAAO,MAAM,MAAMhC,EAAEZ,CAAC,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAEvzB,GAAG,KAAK,KAAKA,EAAE,KAAKK,EAAEL,GAAG,KAAK,KAAKA,EAAE,SAAS6B,EAAE7B,GAAG,KAAK,KAAKA,EAAE,aAAaY,EAAE0rD,GAAG1rD,CAAC,EAAEiB,GAAG,OAAO,KAAK,SAASA,EAAE,KAAK,QAAQA,EAAE,KAAK,gBAAgBjB,CAAC,GAAG,IAAIkB,EAAE,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,KAAK,UAAU,OAAO,EAAE,GAAGD,EAAE,SAASC,EAAE,MAAM,IAAIgkD,EAAE,iBAAiBhkD,6BAA6BD,EAAE,0BAA0B,EAAE,KAAK,QAAQ,QAAQ,KAAK,kEAAkE,EAAE,IAAIE,EAAE,CAAC,SAAS1B,CAAC,EAAE4B,EAAEi9D,GAAG,CAACp8D,EAAE,IAAI,CAAC,IAAImB,EAAE,KAAK,KAAK,KAAK,CAACnB,CAAC,EAAE,OAAO,CAAC,EAAEf,CAAC,EAAE,MAAM,CAACkC,EAAE,GAAGA,EAAE,MAAM,CAAC,CAAC,CAAC,EAAErD,EAAEiB,EAAE,KAAK,YAAY,EAAE,KAAK,KAAK,OAAO,KAAK,eAAe,EAAEa,EAAET,EAAE,GAAGU,EAAEV,EAAE,GAAGW,EAAEX,EAAE,GAAG,KAAK,UAAU,KAAK,YAAYW,EAAEvC,CAAC,EAAE,IAAIwC,EAAE,KAAK,gBAAgBF,EAAED,EAAE,OAAO,KAAK,YAAY,CAACG,CAAC,EAAE,OAAOD,CAAC,EAAEC,CAAC,CAAC,CAAC,CAAC,gBAAgBjC,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEmoC,GAAGvnC,EAAE,KAAK,EAAE,OAAOZ,EAAEijC,GAAGjjC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEA,EAAEspD,GAAGtpD,CAAC,EAAE,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,KAAK,UAAU,IAAI,GAAG,EAAE,EAAEwpD,GAAGxpD,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEA,CAAC,EAAE,KAAK,KAAK,UAAU,EAAE,CAACwpD,GAAGxpD,EAAE,CAAC,EAAE,KAAK,KAAK,SAAS,CAAC,CAAC,EAAE,CAACA,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,CAAC,OAAO,KAAK,UAAU,KAAK,KAAK,iBAAiB,CAAC,CAAC,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,UAAU,KAAK,KAAK,oBAAoB,KAAK,KAAK,OAAO,CAAC,6BAA6BY,EAAE,CAAC,MAAM,6BAA6BA,CAAC,EAAE,KAAK,MAAM,MAAM,KAAK,KAAK,6BAA6BA,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,gBAAgB,KAAK,gBAAgB,YAAY,KAAK,YAAY,YAAY,KAAK,YAAY,SAAS,KAAK,SAAS,OAAO,KAAK,MAAM,EAAE,KAAK,cAAc,OAAOA,EAAE,aAAa,KAAK,cAAc,IAAI,EAAE,KAAK,KAAK,UAAU,EAAE,OAAO,KAAK,aAAa,IAAIowD,GAAG,YAAYpwD,EAAE,KAAK,CAAC,UAAU,KAAK,KAAK,aAAa,EAAE,OAAO,CAAC,GAAG,OAAO,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAE,CAAC,EAAEY,CAAC,EAAEZ,CAAC,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,IAAIK,EAAEL,EAAE,KAAK6B,EAAEg0D,GAAGx1D,EAAE,CAAC,EAAE,OAAO,IAAIO,EAAE,OAAO,OAAOZ,EAAE,CAAC,KAAK6B,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEuuD,GAAG,UAAU,MAAM74B,GAAE,cAAc64B,EAAE,EAAE,IAAIC,GAAG,cAAclD,EAAE,CAAC,EAAEiS,GAAG,cAAc/O,EAAE,CAAC,YAAYzvD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,mBAAmB,OAAO,KAAK,2BAA2B,eAAe,KAAK,8BAA8B,aAAa,KAAK,yBAAyB,QAAQ,KAAK,MAAMA,EAAE,MAAMumD,GAAG,KAAK,MAAM,OAAO,EAAE,KAAK,WAAWyV,GAAGh8D,EAAE,YAAY,KAAK,KAAK,mBAAmBA,EAAE,UAAU,EAAE,KAAK,QAAQA,EAAE,SAAS,KAAK,GAAGA,EAAE,QAAQ,KAAK,kBAAkBurD,GAAGvrD,EAAE,mBAAmB,KAAK,0BAA0B,EAAE,KAAK,qBAAqBurD,GAAGvrD,EAAE,sBAAsB,KAAK,6BAA6B,EAAE,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,KAAK,wBAAwB,EAAE,KAAK,kBAAkBy8D,GAAGz8D,EAAE,iBAAiB,EAAE,KAAK,qBAAqBy8D,GAAGz8D,EAAE,oBAAoB,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,iBAAiBsuD,GAAGtuD,EAAE,gBAAgB,EAAE,KAAK,oBAAoBsuD,GAAGtuD,EAAE,mBAAmB,EAAE,KAAK,eAAesuD,GAAGtuD,EAAE,cAAc,EAAE,KAAK,QAAQmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,SAAS,KAAK,EAAEA,EAAE,OAAO,CAAC,CAAC,CAAC,EAAE,KAAK,iBAAiBmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,kBAAkB,KAAK,EAAEA,EAAE,gBAAgB,CAAC,CAAC,CAAC,EAAE,KAAK,YAAYA,EAAE,YAAY,KAAK,UAAU,KAAK,MAAM,KAAK,YAAY,KAAK,KAAK,qBAAqB,IAAI,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,KAAK,OAAO,KAAK,UAAU,SAAS,CAACA,EAAEA,EAAE,OAAO,GAAG,KAAK,KAAK,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,KAAK,UAAU,mBAAmB,CAAC,KAAK,MAAM,KAAK,KAAK,EAAE,KAAK,KAAK,qBAAqB,KAAK,qBAAqB,GAAG,KAAK,mBAAmB,EAAE,KAAK,QAAQ,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,KAAK,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,EAAE,KAAK,KAAK,KAAK,KAAK,MAAM,EAAE,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAEA,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,8CAA8CllD,EAAE,SAAS,EAAE,IAAI,EAAEA,EAAE,GAAGA,EAAEA,EAAE,GAAG,IAAIP,EAAEL,EAAE,UAAU,KAAK,GAAGA,EAAE,SAAS,EAAE,KAAK,SAAS,KAAK,QAAQ,GAAG,KAAK,aAAa,OAAO,KAAK,YAAYq/D,GAAG,CAAC,KAAK,IAAIh2B,GAAGzoC,CAAC,EAAE,KAAK,KAAK,QAAQ,SAASP,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,GAAG,KAAK,sBAAsB,OAAO,KAAK,qBAAqBg/D,GAAG,CAAC,KAAK,IAAIh2B,GAAG,CAAC,EAAE,KAAK,KAAK,iBAAiB,SAAShpC,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,IAAIwB,EAAEC,EAAE,KAAK,YAAYC,EAAE,KAAK,qBAAqBD,GAAG,KAAKD,EAAEooD,GAAG/wB,EAAEt4B,EAAEkB,CAAC,EAAE,KAAK,OAAO,KAAK,CAAC,EAAED,EAAEooD,GAAGrpD,EAAE,KAAK,OAAO,KAAK,CAAC,EAAE,KAAK,MAAM,OAAOiB,EAAEwoD,GAAGxoD,EAAE,KAAK,KAAK,KAAK,CAAC,GAAGE,GAAG,OAAO,EAAEm3B,EAAE,EAAEn3B,CAAC,GAAG,IAAIC,EAAE42B,EAAE/2B,EAAEooD,GAAG,EAAE,KAAK,gBAAgB,KAAK,CAAC,CAAC,EAAE,OAAO,KAAK,YAAY,OAAOjoD,EAAE,KAAK,WAAW,MAAMA,CAAC,GAAG,CAACA,EAAEA,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIpB,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,MAAM,KAAK,MAAM,WAAW08D,GAAG,KAAK,UAAU,EAAE,QAAQ,KAAK,QAAQ,kBAAkBxQ,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,kBAAkBiR,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,iBAAiBnO,GAAG,KAAK,gBAAgB,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,eAAeA,GAAG,KAAK,cAAc,EAAE,QAAQ,KAAK,QAAQ,iBAAiB,KAAK,gBAAgB,EAAE,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEpuD,CAAC,EAAEZ,CAAC,CAAC,CAAC,EAAEo/D,GAAG,UAAU,gBAAgB7nC,GAAE,cAAc6nC,EAAE,EAAE,IAAIE,GAAG,cAAclP,EAAE,CAAC,YAAYxvD,EAAE,CAACA,EAAE,KAAK,IAAIw+D,GAAGx+D,CAAC,EAAE,MAAMA,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,KAAK,aAAa,OAAOC,GAAG,KAAK,KAAK,WAAW,EAAE,KAAK,KAAK,YAAY,MAAM,KAAK,KAAK,sBAAsB,OAAOA,GAAG,KAAK,KAAK,oBAAoB,EAAE,KAAK,KAAK,qBAAqB,MAAM,IAAI,EAAExzB,GAAG,KAAK,KAAKA,EAAE,KAAKK,EAAEL,GAAG,KAAK,KAAKA,EAAE,SAAS6B,EAAE7B,GAAG,KAAK,KAAKA,EAAE,aAAa,OAAO,MAAM,KAAKY,EAAE,CAAC,KAAK,EAAE,SAASP,EAAE,aAAawB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,WAAWjB,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,CAAC,CAAC,CAAC,EAAEs/D,GAAG,UAAU,YAAY/nC,GAAE,cAAc+nC,EAAE,EAAE,IAAIC,GAAG,cAAclP,EAAE,CAAC,YAAYzvD,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,mBAAmB,OAAO,KAAK,6BAA6B,cAAc,KAAK,2BAA2B,eAAe,KAAK,8BAA8B,aAAa,KAAK,yBAAyB,QAAQA,EAAE,WAAW,MAAM,IAAIklD,EAAE,6DAA6D,EAAE,KAAK,MAAMllD,EAAE,MAAMumD,GAAG,KAAK,MAAM,OAAO,EAAE,KAAK,WAAWyV,GAAGh8D,EAAE,aAAa,OAAO,KAAK,mBAAmBA,EAAE,UAAU,EAAE,KAAK,oBAAoBg8D,GAAGh8D,EAAE,sBAAsB,OAAO,KAAK,6BAA6BA,EAAE,mBAAmB,EAAE,KAAK,QAAQA,EAAE,SAAS,KAAK,GAAGA,EAAE,QAAQ,KAAK,kBAAkBurD,GAAGvrD,EAAE,mBAAmB,KAAK,0BAA0B,EAAE,KAAK,qBAAqBurD,GAAGvrD,EAAE,sBAAsB,KAAK,6BAA6B,EAAE,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,KAAK,wBAAwB,EAAE,KAAK,kBAAkBy8D,GAAGz8D,EAAE,iBAAiB,EAAE,KAAK,qBAAqBy8D,GAAGz8D,EAAE,oBAAoB,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,iBAAiBsuD,GAAGtuD,EAAE,gBAAgB,EAAE,KAAK,oBAAoBsuD,GAAGtuD,EAAE,mBAAmB,EAAE,KAAK,eAAesuD,GAAGtuD,EAAE,cAAc,EAAE,KAAK,QAAQmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,SAAS,KAAK,EAAEA,EAAE,OAAO,CAAC,CAAC,CAAC,EAAE,KAAK,iBAAiBmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,kBAAkB,KAAK,EAAEA,EAAE,gBAAgB,CAAC,CAAC,CAAC,EAAE,KAAK,YAAYA,EAAE,YAAY,KAAK,eAAeA,EAAE,eAAe,KAAK,UAAU,KAAK,MAAM,KAAK,YAAY,KAAK,KAAK,qBAAqB,IAAI,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAEA,EAAE,OAAO,GAAG,KAAK,OAAO,KAAK,UAAU,SAAS,CAACZ,EAAE,KAAK,MAAM,CAAC,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,KAAK,UAAU,mBAAmB,CAAC,KAAK,MAAM,KAAK,MAAM,CAAC,EAAE,KAAK,KAAK,qBAAqB,KAAK,qBAAqB,GAAG,KAAK,mBAAmB,EAAE,KAAK,QAAQ,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,MAAM,CAAC,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,EAAE,KAAK,KAAK,KAAK,KAAK,MAAM,EAAE,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAEA,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,uDAAuDllD,EAAE,SAAS,EAAE,IAAI,EAAEZ,EAAE,UAAU,KAAK,GAAGA,EAAE,SAASK,EAAEO,EAAE,GAAGA,EAAEA,EAAE,GAAG,EAAE,KAAK,SAAS,KAAK,QAAQ,GAAG,KAAK,aAAa,OAAO,KAAK,YAAYy+D,GAAG,CAAC,KAAK,IAAIh2B,GAAGzoC,CAAC,EAAE,KAAK,KAAK,QAAQ,SAAS,EAAE,MAAM,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,GAAG,KAAK,sBAAsB,OAAO,KAAK,qBAAqBy+D,GAAG,CAAC,KAAK,IAAIh2B,GAAGhpC,CAAC,EAAE,KAAK,KAAK,iBAAiB,SAAS,EAAE,MAAM,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,IAAIwB,EAAE,KAAK,YAAYC,EAAE,KAAK,qBAAqBC,EAAEC,EAAE,EAAE,EAAE,KAAK,SAAS,KAAK,QAAQ,IAAIpB,EAAEs4B,EAAEt4B,EAAEiB,EAAE,EAAE,GAAG,IAAIa,EAAEunD,GAAGrpD,EAAE,KAAK,OAAO,KAAK,CAAC,EAAE,KAAK,UAAU8B,EAAE2nD,GAAG3nD,EAAE,KAAK,KAAK,KAAK,CAAC,GAAG,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,IAAIrC,EAAE64B,EAAE74B,EAAEyB,EAAE,EAAE,GAAG,IAAI,EAAE,KAAK,gBAAgB,KAAK,EAAE,CAACc,EAAEC,CAAC,EAAEqsC,GAAG,EAAE,CAAC,EAAE,KAAK,MAAM,KAAK,KAAK,EAAE,EAAE,KAAK,CAAC,EAAEpsC,EAAEmnD,GAAG5pD,EAAEuC,CAAC,EAAE,CAACoB,EAAEC,EAAEoD,CAAC,EAAE6nC,GAAGxsC,EAAE,EAAEA,EAAE,KAAK,CAAC,EAAE,CAACmC,EAAEJ,CAAC,EAAEyqC,GAAGpsC,EAAE,EAAEA,EAAE,KAAK,CAAC,EAAEf,EAAE,KAAK,oBAAoB,MAAM62B,EAAE50B,EAAEa,CAAC,CAAC,EAAE7C,EAAE,KAAK,oBAAoB,MAAM42B,EAAE30B,EAAEQ,CAAC,CAAC,EAAE,IAAIC,EAAEulD,GAAG/wB,EAAEl3B,EAAE3B,CAAC,EAAEwC,CAAC,EAAE,EAAE,KAAK,WAAW,MAAM+1B,EAAEvxB,EAAE3C,CAAC,CAAC,EAAE,IAAIC,EAAEi0B,EAAEM,EAAEn3B,EAAE1B,CAAC,EAAE64B,EAAEN,EAAE,EAAErE,GAAGxyB,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,MAAM,CAAC4C,EAAEA,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI/D,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,MAAM,KAAK,MAAM,WAAW08D,GAAG,KAAK,UAAU,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,QAAQ,KAAK,QAAQ,kBAAkBxQ,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,kBAAkBiR,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,iBAAiBnO,GAAG,KAAK,gBAAgB,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,eAAeA,GAAG,KAAK,cAAc,EAAE,QAAQ,KAAK,QAAQ,iBAAiB,KAAK,iBAAiB,eAAe,KAAK,eAAe,WAAW,EAAE,EAAE,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEpuD,CAAC,EAAEZ,CAAC,CAAC,CAAC,EAAEu/D,GAAG,UAAU,UAAUhoC,GAAE,cAAcgoC,EAAE,EAAE,IAAIC,GAAG,cAAcpP,EAAE,CAAC,YAAYxvD,EAAE,CAACA,EAAE,iBAAiB,GAAG,QAAQ,KAAK,gHAAgH,EAAEA,EAAE,KAAK,IAAI2+D,GAAG3+D,CAAC,EAAE,MAAMA,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,KAAK,aAAa,OAAOC,GAAG,KAAK,KAAK,WAAW,EAAE,KAAK,KAAK,YAAY,MAAM,KAAK,KAAK,sBAAsB,OAAOA,GAAG,KAAK,KAAK,oBAAoB,EAAE,KAAK,KAAK,qBAAqB,MAAM,IAAI,EAAExzB,GAAG,KAAK,KAAKA,EAAE,KAAKK,EAAEL,GAAG,KAAK,KAAKA,EAAE,SAAS6B,EAAE7B,GAAG,KAAK,KAAKA,EAAE,aAAa,OAAO,MAAM,KAAKY,EAAE,CAAC,KAAK,EAAE,SAASP,EAAE,aAAawB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,WAAWjB,EAAEZ,EAAE,CAAC,OAAOA,EAAE,gBAAgB,IAAIA,EAAE,eAAe,GAAG,IAAIY,EAAEZ,CAAC,CAAC,CAAC,EAAEw/D,GAAG,UAAU,MAAMjoC,GAAE,cAAcioC,EAAE,EAAE,IAAIC,GAAG,cAAcpP,EAAE,CAAC,YAAYzvD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,mBAAmB,OAAO,KAAK,6BAA6B,cAAc,KAAK,2BAA2B,eAAe,KAAK,8BAA8B,aAAa,KAAK,yBAAyB,QAAQ,KAAK,MAAMA,EAAE,MAAMumD,GAAG,KAAK,MAAM,OAAO,EAAE,KAAK,WAAWyV,GAAGh8D,EAAE,aAAa,OAAO,KAAK,mBAAmBA,EAAE,UAAU,EAAE,KAAK,oBAAoBg8D,GAAGh8D,EAAE,sBAAsB,OAAO,KAAK,6BAA6BA,EAAE,mBAAmB,EAAE,KAAK,QAAQA,EAAE,SAAS,KAAK,GAAGA,EAAE,QAAQ,KAAK,kBAAkBurD,GAAGvrD,EAAE,mBAAmB,KAAK,0BAA0B,EAAE,KAAK,qBAAqBurD,GAAGvrD,EAAE,sBAAsB,KAAK,6BAA6B,EAAE,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,KAAK,wBAAwB,EAAE,KAAK,eAAeA,EAAE,eAAe,KAAK,kBAAkBy8D,GAAGz8D,EAAE,iBAAiB,EAAE,KAAK,qBAAqBy8D,GAAGz8D,EAAE,oBAAoB,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,iBAAiBsuD,GAAGtuD,EAAE,gBAAgB,EAAE,KAAK,oBAAoBsuD,GAAGtuD,EAAE,mBAAmB,EAAE,KAAK,eAAesuD,GAAGtuD,EAAE,cAAc,EAAE,KAAK,QAAQmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,SAAS,KAAK,EAAEA,EAAE,OAAO,CAAC,CAAC,CAAC,EAAE,KAAK,iBAAiBmoD,GAAG,CAAC,EAAEC,GAAG,CAAC,EAAEpoD,EAAE,kBAAkB,KAAK,EAAEA,EAAE,gBAAgB,CAAC,CAAC,CAAC,EAAE,KAAK,YAAYA,EAAE,YAAY,KAAK,eAAeA,EAAE,eAAe,KAAK,UAAU,CAAC,KAAK,MAAM,KAAK,KAAK,EAAE,KAAK,YAAY,KAAK,KAAK,qBAAqB,IAAI,CAAC,MAAMA,EAAE,CAAC,IAAIZ,EAAEY,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAI,EAAEA,EAAEA,EAAE,OAAO,GAAG,KAAK,OAAO,KAAK,UAAU,SAAS,CAAC,EAAE,KAAK,MAAM,CAAC,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,KAAK,UAAU,mBAAmB,CAAC,KAAK,MAAM,KAAK,MAAM,CAAC,EAAE,KAAK,KAAK,qBAAqB,KAAK,qBAAqB,GAAG,KAAK,mBAAmB,EAAE,IAAIP,EAAE,GAAG,KAAK,QAAQ,CAAC,GAAG,KAAK,eAAe,CAAC,IAAIwB,EAAE,KAAK,gBAAgBC,EAAE,KAAK,MAAMzB,EAAE,IAAIL,EAAE,cAAc+qD,EAAE,CAAC,MAAM/oD,EAAEC,EAAE,CAAC,IAAIS,EAAEb,EAAE,MAAM,CAACC,CAAC,CAAC,EAAEa,EAAE,IAAIsoD,GAAG,EAAE,MAAM,CAACnpD,CAAC,CAAC,EAAEc,EAAEf,EAAE,MAAM,CAACC,EAAE,CAAC,CAAC,EAAE,OAAOioD,GAAGA,GAAGrnD,EAAEC,CAAC,EAAEC,CAAC,CAAC,CAAC,EAAE5C,EAAE,UAAU,aAAaA,EAAE,MAAMK,EAAE,KAAK,gBAAgB,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,MAAM,CAAC,EAAE,KAAKA,EAAE,KAAK,gBAAgB,GAAG,KAAK,cAAc,CAAC,MAAM,KAAK,KAAK,KAAK,KAAK,MAAM,EAAE,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAEvzB,EAAE,UAAU,KAAK,GAAGA,EAAE,SAAS,GAAGY,EAAEA,EAAEA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,wDAAwDllD,EAAE,SAAS,EAAE,IAAIP,EAAEO,EAAE,GAAGiB,EAAEjB,EAAE,GAAGA,EAAEA,EAAE,GAAG,EAAE,KAAK,SAAS,KAAK,QAAQ,GAAG,KAAK,aAAa,OAAO,KAAK,YAAYy+D,GAAG,CAAC,KAAK,IAAIh2B,GAAGzoC,CAAC,EAAE,KAAK,KAAK,QAAQ,SAAS,EAAE,MAAM,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,GAAG,KAAK,sBAAsB,OAAO,KAAK,qBAAqBy+D,GAAG,CAAC,KAAK,IAAIh2B,GAAGhpC,CAAC,EAAE,KAAK,KAAK,iBAAiB,SAAS,EAAE,MAAM,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,IAAIyB,EAAE,KAAK,YAAYC,EAAE,KAAK,qBAAqBC,EAAE,EAAEU,EAAE,EAAE,EAAE,KAAK,SAAS,KAAK,QAAQ,IAAI9B,EAAEs4B,EAAEt4B,EAAEkB,EAAE,EAAE,GAAG,IAAIc,EAAEqnD,GAAGrpD,EAAE,KAAK,OAAO,KAAK,CAAC,EAAE,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,IAAIP,EAAE64B,EAAE74B,EAAE0B,EAAE,EAAE,GAAGa,EAAEg2B,EAAEh2B,EAAEqnD,GAAG5pD,EAAE,KAAK,gBAAgB,KAAK,CAAC,CAAC,EAAE,KAAK,UAAUuC,EAAEynD,GAAGznD,EAAE,KAAK,KAAK,KAAK,CAAC,GAAG,GAAG,CAACC,EAAEC,EAAEkB,EAAEC,CAAC,EAAEirC,GAAGtsC,EAAE,EAAEA,EAAE,KAAK,CAAC,EAAEZ,EAAE,KAAK,oBAAoB,MAAMa,CAAC,EAAE,EAAE,KAAK,oBAAoB,MAAMC,CAAC,EAAEJ,EAAEk2B,EAAEM,EAAE,EAAEr3B,CAAC,EAAEq3B,EAAEl3B,EAAE,KAAK,WAAW,MAAMgC,CAAC,CAAC,CAAC,EAAE,EAAE,KAAK,oBAAoB,MAAMC,CAAC,EAAE,IAAIoD,EAAE6xB,EAAE,EAAE,KAAK,WAAW,MAAMx2B,CAAC,CAAC,EAAE,MAAM,CAAC2E,EAAEA,EAAE3E,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI9B,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,MAAM,KAAK,MAAM,WAAW08D,GAAG,KAAK,UAAU,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,QAAQ,KAAK,QAAQ,kBAAkBxQ,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,eAAe,KAAK,eAAe,kBAAkBiR,GAAG,KAAK,iBAAiB,EAAE,qBAAqBA,GAAG,KAAK,oBAAoB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,iBAAiBnO,GAAG,KAAK,gBAAgB,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,eAAeA,GAAG,KAAK,cAAc,EAAE,QAAQ,KAAK,QAAQ,iBAAiB,KAAK,iBAAiB,eAAe,KAAK,cAAc,EAAE,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEpuD,CAAC,EAAEZ,CAAC,CAAC,CAAC,EAAEy/D,GAAG,UAAU,WAAWloC,GAAE,cAAckoC,EAAE,EAAE,IAAIC,GAAG,cAActP,EAAE,CAAC,YAAYxvD,EAAE,CAACA,EAAE,iBAAiB,GAAG,QAAQ,KAAK,gHAAgH,EAAEA,EAAE,KAAK,IAAI6+D,GAAG7+D,CAAC,EAAE,MAAMA,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,KAAK,aAAa,OAAOC,GAAG,KAAK,KAAK,WAAW,EAAE,KAAK,KAAK,YAAY,MAAM,KAAK,KAAK,sBAAsB,OAAOA,GAAG,KAAK,KAAK,oBAAoB,EAAE,KAAK,KAAK,qBAAqB,MAAM,IAAI,EAAExzB,GAAG,KAAK,KAAKA,EAAE,KAAKK,EAAEL,GAAG,KAAK,KAAKA,EAAE,SAAS6B,EAAE7B,GAAG,KAAK,KAAKA,EAAE,aAAa,OAAO,MAAM,KAAKY,EAAE,CAAC,KAAK,EAAE,SAASP,EAAE,aAAawB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,WAAWjB,EAAEZ,EAAE,CAAC,OAAOA,EAAE,gBAAgB,IAAIA,EAAE,eAAe,GAAG,IAAIY,EAAEZ,CAAC,CAAC,CAAC,EAAE0/D,GAAG,UAAU,OAAOnoC,GAAE,cAAcmoC,EAAE,EAAE,IAAIP,GAAG,cAAc9O,EAAE,CAAC,YAAYzvD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,MAAMA,EAAE,KAAK,CAAC,IAAI,WAAW,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,MAAM,MAAM,EAAE,QAAQ,EAAE,MAAM,QAAQA,EAAE,SAAS,EAAEY,EAAE,KAAK,GAAGZ,EAAE,SAAS,EAAEY,EAAE,KAAKZ,EAAE,SAAS,EAAE,OAAOY,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAEA,EAAE,IAAI,EAAEA,EAAE,MAAM,CAAC,EAAEP,EAAE,CAAC,EAAE,QAAQ0B,KAAK,KAAK,MAAM,MAAM,EAAE,QAAQ,EAAE,MAAM,QAAQA,EAAE,SAAS,EAAE1B,EAAE,KAAK,EAAE,OAAO,EAAE0B,EAAE,UAAU,MAAM,CAAC,EAAE1B,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,CAAC,EAAEA,EAAE,QAAQ,EAAE,IAAIwB,EAAE,CAAC,EAAEC,EAAE,QAAQC,EAAE,EAAEA,EAAE,KAAK,MAAM,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE,KAAK,MAAMD,GAAG,EAAE1B,EAAE0B,GAAGA,IAAI,EAAED,EAAE,CAAClB,EAAE,EAAE,EAAE,OAAO,CAAC,EAAEkB,EAAE,CAACA,EAAE,EAAE,EAAE,OAAO,CAAC,EAAEA,EAAEE,EAAE,KAAKF,EAAE9B,CAAC,EAAE6B,EAAE,KAAKC,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,QAAQC,KAAKF,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAE,KAAK,GAAGE,CAAC,EAAE,MAAM,CAACD,EAAE,EAAE,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,MAAMlB,EAAE,CAACwrD,GAAGxrD,CAAC,IAAIA,EAAEA,EAAE,IAAIA,EAAEA,EAAE,IAAIZ,EAAE,KAAK,MAAM,QAAQ,CAAC,EAAEK,IAAI,CAACkoD,GAAG,WAAWloD,IAAI,IAAI,CAAC,EAAE,MAAMO,CAAC,EAAE,MAAM,QAAQ,EAAE,SAAS,EAAEZ,EAAE,EAAE,UAAU,GAAGA,EAAE,EAAE,UAAUY,EAAE,CAACA,EAAE,GAAGZ,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,WAAW,CAAC,IAAIY,EAAE,MAAM,UAAU,EAAEZ,EAAE6B,IAAI,CAAC,UAAUA,EAAE,aAAa,EAAE,OAAOA,EAAE,UAAU,CAAC,GAAGxB,EAAE,CAAC,MAAM,KAAK,MAAM,IAAIL,CAAC,CAAC,EAAE,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEY,CAAC,EAAEP,CAAC,CAAC,CAAC,OAAO,WAAWO,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,IAAIK,EAAE,CAAC,EAAE,QAAQwB,KAAK7B,EAAE,MAAMK,EAAE,KAAKw1D,GAAGh0D,EAAE,CAAC,CAAC,EAAE,OAAO,IAAIjB,EAAE,CAAC,MAAMP,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,CAAC,GAAG,CAAC,KAAK,UAAU,MAAM,CAAC,EAAE,IAAIO,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,MAAMY,EAAE,KAAK,GAAGZ,EAAE,gBAAgB,EAAE,OAAOY,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,MAAMY,EAAE,KAAK,GAAGZ,EAAE,mBAAmB,EAAE,GAAG,CAAC,KAAK,UAAU,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQ,KAAK,KAAK,MAAMA,EAAE,KAAK,GAAG,EAAE,gBAAgB,EAAE,OAAOA,EAAE,OAAOY,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,YAAY,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,KAAK,KAAK,MAAMY,EAAE,KAAK,GAAGZ,EAAE,OAAO,EAAE,OAAO4sD,GAAGhsD,CAAC,CAAC,CAAC,WAAWA,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,QAAQ,KAAK,KAAK,MAAM,CAAC,IAAIK,EAAE,EAAE,QAAQ,OAAOwB,EAAEjB,EAAE,OAAOP,CAAC,EAAE,QAAQyB,EAAE,EAAEA,EAAE,EAAE,QAAQ,OAAO,EAAEA,EAAE9B,EAAE,KAAK,CAAC,EAAE,QAAQ8B,GAAGD,EAAEC,EAAE,CAAC,CAAC,CAAC+qD,GAAG7sD,CAAC,CAAC,CAAC,EAAEm/D,GAAG,UAAU,kBAAkB5nC,GAAE,cAAc4nC,EAAE,EAAE,SAASE,GAAGv/D,EAAE,CAAC,GAAG,CAAC,KAAK,EAAE,KAAKE,EAAE,SAASI,EAAE,GAAG,MAAMC,EAAE,EAAE,YAAY,CAAC,EAAEP,EAAE,EAAE,IAAI,GAAG,KAAK,EAAE,EAAE,EAAEE,CAAC,EAAEwqD,GAAG,EAAE,EAAExqD,CAAC,EAAE+B,EAAE,IAAI2oD,GAAG,EAAE,EAAEtqD,CAAC,EAAE,MAAM,CAACC,GAAGA,GAAG,EAAEozB,GAAG1xB,EAAE,EAAE,MAAM,CAAC,EAAE,MAAM1B,CAAC,EAAE,KAAK,MAAM,EAAE,IAAI0B,CAAC,EAAE,IAAI,GAAG0xB,GAAG,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,IAAIksC,GAAG,SAAS7/D,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,KAAKN,EAAE,OAAO,UAAU,eAAe,KAAKA,EAAEM,CAAC,GAAG,EAAE,QAAQA,CAAC,EAAE,IAAIJ,EAAEI,GAAGN,EAAEM,IAAI,GAAGN,GAAG,MAAM,OAAO,OAAO,uBAAuB,WAAW,QAAQO,EAAE,EAAED,EAAE,OAAO,sBAAsBN,CAAC,EAAEO,EAAED,EAAE,OAAOC,IAAI,EAAE,QAAQD,EAAEC,EAAE,EAAE,GAAG,OAAO,UAAU,qBAAqB,KAAKP,EAAEM,EAAEC,EAAE,IAAIL,EAAEI,EAAEC,IAAIP,EAAEM,EAAEC,KAAK,OAAOL,CAAC,EAAM4/D,GAAG,cAAcxP,EAAE,CAAC,YAAYxvD,EAAE,CAAC,GAAGA,EAAE,OAAO,MAAM,IAAImlD,GAAG,oDAAoD,EAAE,GAAG,MAAM,QAAQnlD,EAAE,IAAI,EAAE,MAAM,IAAImlD,GAAG,gEAAgE,EAAE,MAAMnlD,CAAC,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,KAAKlsD,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG,KAAK,KAAK,aAAa,OAAOC,GAAG,KAAK,KAAK,WAAW,EAAE,KAAK,KAAK,YAAY,MAAM,KAAK,KAAK,sBAAsB,OAAOA,GAAG,KAAK,KAAK,oBAAoB,EAAE,KAAK,KAAK,qBAAqB,MAAMxzB,GAAGA,EAAE,UAAU,MAAM,IAAI8lD,EAAE,2CAA2C,EAAE,IAAI,EAAE9lD,GAAG,KAAK,KAAKA,EAAE,KAAKK,EAAEL,GAAG,KAAK,KAAKA,EAAE,SAAS6B,EAAE7B,GAAG,KAAK,KAAKA,EAAE,aAAa,OAAO,MAAM,KAAKY,EAAE,CAAC,KAAK,EAAE,SAASP,EAAE,aAAawB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBjB,EAAE,CAAC,IAAIZ,EAAE,KAAK,yBAAyBY,CAAC,EAAE,OAAO,KAAK,kBAAkBZ,EAAE,CAACA,EAAE,GAAG,GAAGA,EAAE,MAAM,CAAC,CAAC,GAAG,KAAK,cAAcA,EAAE,CAACA,EAAE,GAAG,MAAM,CAAC,EAAE,KAAK,CAACY,EAAE,GAAG,GAAGZ,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,GAAGA,CAAC,CAAC,gBAAgBY,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,GAAG,CAAC,UAAUvzB,CAAC,EAAE,KAAK,KAAK,EAAEY,EAAE,MAAMP,EAAE,KAAK,yBAAyB,CAAC,EAAEwB,EAAE,CAACxB,EAAE,GAAG,GAAGA,EAAE,MAAM,CAAC,CAAC,EAAEyB,EAAEqmC,GAAGtmC,CAAC,EAAE,OAAO,MAAM,QAAQ7B,CAAC,EAAE,MAAMA,EAAE,MAAM,EAAE,KAAK8B,CAAC,EAAE,CAACA,CAAC,CAAC,CAAC,CAAC,CAAC,YAAYlB,EAAEZ,EAAE,GAAG,CAACuzB,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,SAAS,MAAM,IAAIqyB,GAAG,iEAAiE,EAAE,IAAI,EAAE,KAAK,UAAU,GAAG,MAAMvlD,EAAE,KAAK,yBAAyB,CAAC,EAAEwB,EAAE,CAACxB,EAAE,GAAG,GAAGA,EAAE,MAAM,CAAC,CAAC,EAAE,GAAG,EAAE,IAAI,KAAK,MAAM,IAAIylD,EAAE,uUAAuU,EAAE,GAAG,KAAK,UAAU,GAAG,KAAK,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,QAAQ,KAAK,KAAK,UAAU,IAAI,IAAI3d,GAAGtmC,CAAC,CAAC,EAAE,KAAK,QAAQ,CAACsmC,GAAGtmC,CAAC,CAAC,UAAUjB,GAAG,KAAK4yB,GAAG,KAAK,OAAO,EAAE,KAAK,YAAY,OAAOA,GAAG,KAAK,UAAU,EAAE,KAAK,WAAW,CAAC,GAAG,MAAM,QAAQ,KAAK,KAAK,SAAS,EAAE,KAAK,QAAQ,KAAK,KAAK,UAAU,IAAI,IAAI2U,GAAGtmC,CAAC,CAAC,EAAE,KAAK,QAAQ,GAAGsmC,GAAGtmC,CAAC,MAAM,CAAC,GAAG,MAAM,QAAQjB,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAGA,EAAE,SAAS,KAAK,QAAQ,OAAO,MAAM,IAAIklD,EAAE,SAAS,KAAK,gBAAgB,KAAK,QAAQ,oCAAoCllD,EAAE,0CAA0CA,GAAG,EAAEZ,EAAE,KAAK,WAAW,KAAK,KAAK,QAAQ,MAAM,CAAC,EAAEwzB,GAAG,KAAK,OAAO,EAAE,QAAQzxB,EAAE,EAAEA,EAAE,KAAK,QAAQ,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAEpB,EAAEmB,GAAGE,EAAEJ,EAAE,GAAG,CAAC+kB,EAAE,YAAY5kB,EAAE,MAAMC,CAAC,EAAE,MAAM,IAAI6jD,EAAE,SAAS/jD,gCAAgC,KAAK,wBAAwBE,qBAAqBD,EAAE,OAAO,EAAE,KAAK,QAAQD,GAAGC,CAAC,CAAC,CAAC,KAAK,QAAQ,KAAK,QAAQ,IAAID,GAAG0xB,GAAG1xB,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,yBAAyBnB,EAAE,CAAC,GAAG,CAAC,WAAWZ,EAAE,QAAQ,EAAE,WAAWK,EAAE,QAAQwB,EAAE,QAAQC,EAAE,aAAaC,CAAC,EAAE,KAAK,KAAKC,EAAEhC,IAAI,gBAAgB,EAAEY,EAAEoB,EAAE,EAAE,GAAGU,EAAE9B,EAAEoB,EAAE,EAAE,GAAG,EAAE67D,GAAG,EAAEx9D,EAAE,GAAGwB,EAAEC,EAAE,GAAGC,EAAE,EAAE,EAAEa,EAAEi7D,GAAGn7D,EAAErC,EAAE,GAAGwB,EAAEC,EAAE,GAAGC,EAAE,EAAE,EAAE,MAAM,CAAC,GAAGnB,EAAE,MAAM,EAAE,CAAC,EAAE,GAAGoB,EAAE,CAAC,EAAE,EAAEY,CAAC,EAAE,CAAC,EAAEA,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEg9D,GAAG,UAAU,YAAY,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAY7+D,EAAE,CAAC,GAAG,CAAC,QAAQZ,EAAE,WAAW,EAAE,QAAQK,EAAE,QAAQwB,EAAE,WAAWC,EAAE,aAAaC,CAAC,EAAEnB,EAAE,MAAM,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEA,CAAC,EAAE,CAAC,MAAMZ,CAAC,CAAC,CAAC,EAAE,KAAK,QAAQA,EAAEmnD,GAAG,KAAK,QAAQ,SAAS,EAAE,KAAK,WAAWyW,GAAG,EAAE,EAAE,YAAY,EAAE,KAAK,WAAW,QAAQ57D,GAAGmlD,GAAGnlD,EAAE,YAAY,CAAC,EAAE,KAAK,QAAQ47D,GAAGv9D,GAAG,EAAE,EAAE,SAAS,EAAE,KAAK,QAAQ,QAAQ2B,GAAGmlD,GAAGnlD,EAAE,SAAS,CAAC,EAAE,KAAK,QAAQH,GAAG,QAAQsmD,GAAG,KAAK,OAAO,EAAE,KAAK,WAAWrmD,GAAG,eAAemmD,GAAG,KAAK,UAAU,EAAE,KAAK,aAAa2V,GAAG77D,GAAG,EAAE,EAAE,cAAc,EAAE,KAAK,aAAa,QAAQC,GAAGmlD,GAAGnlD,EAAE,cAAc,CAAC,CAAC,CAAC,MAAMpB,EAAE,CAAC,IAAIZ,EAAEY,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAI,EAAE,KAAK,aAAa,gBAAgB,EAAEA,EAAE,OAAO,EAAE,GAAGA,EAAE,IAAI,KAAK,MAAM,IAAIklD,EAAE,+DAA+DllD,EAAE,IAAI,EAAE,IAAIP,EAAEO,EAAE,GAAGiB,EAAE,EAAEC,EAAE,KAAK,WAAW,OAAO,CAACzB,EAAE,KAAK,QAAQwB,CAAC,CAAC,EAAE,KAAK,OAAO,KAAK,UAAU,SAASC,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,IAAIC,EAAE,KAAK,WAAW,OAAO,CAAC,KAAK,QAAQ,KAAK,QAAQF,CAAC,CAAC,EAAE,GAAG,KAAK,gBAAgB,KAAK,UAAU,mBAAmBE,EAAE,KAAK,KAAK,qBAAqB,KAAK,qBAAqB,GAAG,KAAK,mBAAmB,EAAE,KAAK,QAAQ,CAAC,IAAIC,EAAE,GAAG,KAAK,eAAe,CAAC,IAAI,EAAE,KAAK,gBAAgBU,EAAE,KAAK,QAAQV,EAAE,IAAIhC,EAAE,cAAc+qD,EAAE,CAAC,MAAMnoD,EAAEC,EAAE,CAAC,IAAIC,EAAE,EAAE,MAAM,CAACJ,CAAC,CAAC,EAAE,EAAE0lC,GAAG,CAAC1lC,CAAC,CAAC,EAAEuB,EAAE,EAAE,MAAM,CAACvB,EAAE,CAAC,CAAC,EAAE,OAAOonD,GAAG,CAAChnD,EAAE,EAAEmB,CAAC,CAAC,CAAC,CAAC,EAAEjE,EAAE,UAAU,aAAaA,EAAE,MAAMgC,EAAE,KAAK,gBAAgB,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,QAAQH,CAAC,EAAE,KAAKG,EAAE,KAAK,gBAAgB,GAAG,KAAK,cAAc,CAAC,CAAC,KAAK,MAAM,EAAE,CAAC,KAAKpB,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,8DAA8DllD,EAAE,SAAS,EAAE,IAAI,EAAEZ,EAAE,UAAU,GAAGK,EAAEO,EAAE,GAAGiB,EAAEjB,EAAE,GAAGkB,EAAElB,EAAE,GAAGmB,EAAE,EAAE,EAAE,KAAK,SAAS,KAAK,QAAQ,GAAG,KAAK,aAAa,OAAO,KAAK,YAAYs9D,GAAG,CAAC,KAAK,IAAIh2B,GAAGhpC,CAAC,EAAE,KAAK,KAAK,QAAQ,SAAS,EAAE,MAAM0B,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,IAAIC,EAAE,KAAK,YAAY,EAAE,CAACkG,EAAGC,EAAGC,KAAK,CAACD,GAAI,CAACA,EAAGC,IAAIF,EAAGgxB,EAAE/wB,EAAGC,IAAIF,CAAE,EAAExF,EAAE,EAAErC,EAAE2B,EAAE,CAAC,EAAE,EAAE,EAAE3B,EAAE2B,EAAE,CAAC,EAAEY,EAAE,EAAEvC,EAAE2B,EAAE,CAAC,EAAEa,EAAE,EAAExC,EAAE2B,EAAE,CAAC,EAAE,EAAE,KAAK,kBAAkB,KAAK,iBAAiB,GAAG,KAAK,sBAAsB,OAAO,KAAK,qBAAqBq9D,GAAG,CAAC,KAAK,IAAIh2B,GAAGxnC,CAAC,EAAE,KAAK,KAAK,iBAAiB,SAAS,EAAE,MAAME,EAAE,YAAY,KAAK,WAAW,CAAC,GAAG,IAAIe,EAAE,KAAK,qBAAqBkB,EAAE,EAAEnC,EAAEiB,EAAE,CAAC,EAAEmB,EAAE,EAAEpC,EAAEiB,EAAE,CAAC,EAAEuE,EAAE,EAAExF,EAAEiB,EAAE,CAAC,EAAE+B,EAAE,EAAEhD,EAAEiB,EAAE,CAAC,EAAE2B,EAAE,EAAE,CAACC,EAAEC,EAAEC,EAAE0C,CAAC,EAAE4nC,GAAG,KAAK,OAAO,KAAK,EAAEntC,EAAE0C,CAAC,EAAE,CAACK,EAAEyC,EAAExC,EAAEC,CAAC,EAAE,KAAK,QAAQkqC,GAAG,KAAK,KAAK,KAAK,EAAEntC,CAAC,EAAE,CAAC,KAAK,KAAK,KAAK,IAAI,EAAEW,EAAE,KAAK,UAAUA,EAAEgC,EAAEI,EAAE,KAAK,OAAO,EAAE,EAAE,KAAK,UAAU,EAAEH,EAAE4C,EAAE,KAAK,OAAO,EAAE3E,EAAE,KAAK,UAAUA,EAAEgC,EAAEG,EAAE,KAAK,OAAO,EAAElC,EAAE,KAAK,UAAUA,EAAEyE,EAAEtC,EAAE,KAAK,OAAO,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAEC,CAAC,EAAE8pC,GAAG,KAAK,gBAAgB,KAAK,EAAEntC,EAAE0C,CAAC,EAAET,EAAE,KAAK,cAAcA,EAAEiB,CAAC,EAAEhB,EAAE,KAAK,cAAcA,EAAEiB,CAAC,EAAEmC,EAAE,KAAK,cAAcA,EAAElC,CAAC,EAAEN,EAAE,KAAK,cAAcA,EAAEO,CAAC,EAAE,IAAI2C,EAAE,KAAK,oBAAoB,MAAM6wB,EAAEl2B,EAAEsB,CAAC,CAAC,EAAEyD,EAAE,KAAK,oBAAoB,MAAMmxB,EAAE,EAAE30B,CAAC,CAAC,EAAE+D,EAAE4wB,EAAEM,EAAEzxB,EAAE3F,CAAC,EAAEo3B,EAAEnxB,EAAE,KAAK,WAAW,MAAM6wB,EAAEh2B,EAAEyE,CAAC,CAAC,CAAC,CAAC,EAAEY,EAAGixB,EAAE,KAAK,oBAAoB,MAAMN,EAAE/1B,EAAEgC,CAAC,CAAC,EAAE,KAAK,WAAW,MAAMmD,CAAC,CAAC,EAAE,MAAM,CAACC,EAAGA,EAAGD,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIpH,EAAE,MAAM,UAAU,EAAE,CAAC,MAAMZ,CAAC,EAAEY,EAAE,EAAE++D,GAAG/+D,EAAE,CAAC,OAAO,CAAC,EAAEP,EAAE,CAAC,QAAQ,KAAK,QAAQ,WAAW,KAAK,WAAW,QAAQ,KAAK,QAAQ,WAAW,KAAK,WAAW,aAAa,KAAK,aAAa,QAAQ,KAAK,OAAO,EAAE,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,EAAE,CAAC,EAAEA,CAAC,CAAC,CAAC,UAAUO,EAAEZ,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAEi9B,GAAGl+B,EAAEZ,EAAE,KAAK,QAAQK,GAAG,QAAQ,KAAK,aAAa,gBAAgB,OAAO,OAAO,KAAK,YAAY,EAAE,OAAO,EAAEgqD,GAAGxoD,EAAE,EAAE,KAAK,UAAU,EAAEA,CAAC,CAAC,cAAcjB,EAAEZ,EAAE,CAAC,OAAO8+B,GAAGl+B,EAAEZ,EAAE,EAAE,OAAO,KAAK,aAAa,gBAAgB,OAAO,MAAM,CAAC,CAAC,EAAE6/D,GAAG,UAAU,iBAAiBtoC,GAAE,cAAcsoC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAYh/D,EAAE,CAAC,IAAIZ,EAAE,IAAI6/D,GAAGj/D,CAAC,EAAE,MAAM,OAAO,OAAO,OAAO,OAAO,CAAC,EAAEA,CAAC,EAAE,CAAC,KAAKZ,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,WAAWY,EAAEZ,EAAE,CAAC,OAAO,IAAIY,EAAEZ,CAAC,CAAC,CAAC,EAAE8/D,GAAG,UAAU,aAAavoC,GAAE,cAAcuoC,EAAE,EAAE,IAAIC,GAAG,cAAc5S,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,KAAK,KAAK,IAAI,KAAK,IAAIA,EAAE,KAAK,CAAC,EAAE,CAAC,EAAE,KAAK,WAAWA,EAAE,WAAW,KAAK,KAAKA,EAAE,KAAK,KAAK,gBAAgB,EAAE,CAAC,cAAcA,EAAE,CAAC,GAAG,KAAK,YAAY,KAAK,OAAO,KAAK,WAAW,IAAIZ,EAAEY,EAAE,MAAM,EAAE,CAAC,EAAE,QAAQP,EAAE,EAAEA,EAAE,KAAK,WAAW,OAAO,EAAEA,EAAE,EAAE,KAAK,KAAK,WAAWA,IAAI,KAAKL,EAAEK,GAAG,KAAK,WAAWA,EAAE,EAAE,OAAO,CAAC,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,GAAG,EAAE,KAAK,MAAM,KAAK,KAAK,EAAE,CAAC,IAAIP,EAAEL,EAAE,UAAU,KAAK,GAAGA,EAAE,SAAS6B,EAAE,KAAK,cAAc,CAAC,EAAE,OAAO6oD,GAAG,IAAIF,GAAG,EAAE,KAAK,KAAK3oD,EAAE,KAAK,IAAI,EAAE,IAAI,EAAExB,CAAC,CAAC,CAAC,OAAOO,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,KAAK,KAAK,KAAK,WAAW,KAAK,WAAW,KAAK,KAAK,IAAI,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,SAAS,CAAC,OAAO,MAAM,QAAQ,CAAC,CAAC,EAAEm/D,GAAG,UAAU,UAAUxoC,GAAE,cAAcwoC,EAAE,EAAE,IAAIC,GAAG,cAAcD,EAAE,CAAC,YAAYn/D,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,IAAIZ,EAAEY,EAAE,MAAM,MAAM,CAACZ,EAAE,GAAG,EAAEA,EAAE,EAAE,CAAC,CAAC,EAAEggE,GAAG,UAAU,mBAAmBzoC,GAAE,cAAcyoC,EAAE,EAAE,IAAIC,GAAG,cAAc9S,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,WAAW,KAAK,KAAK,QAAQ,GAAG,KAAK,OAAO,KAAK,KAAK,KAAK,KAAK,KAAK,2BAA2B,eAAe,KAAK,yBAAyB,QAAQA,EAAE,iBAAiB,MAAMA,EAAE,YAAY,MAAMA,EAAE,UAAU,KAAK,CAAC,IAAIZ,EAAE,KAAKY,EAAE,WAAW,OAAOZ,EAAEY,EAAE,WAAW,KAAK,gBAAgB,CAACZ,EAAEY,EAAE,QAAQ,CAAC,CAAC,KAAK,MAAMA,EAAE,MAAMumD,GAAG,KAAK,MAAM,OAAO,EAAE,KAAK,WAAWyV,GAAGh8D,EAAE,UAAU,EAAEA,EAAE,SAAS,OAAO,KAAK,QAAQA,EAAE,SAAS,KAAK,kBAAkBurD,GAAGvrD,EAAE,mBAAmB,KAAK,0BAA0B,EAAE,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,KAAK,wBAAwB,EAAE,KAAK,iBAAiBsuD,GAAGtuD,EAAE,gBAAgB,EAAE,KAAK,eAAesuD,GAAGtuD,EAAE,cAAc,EAAE,KAAK,kBAAkBy8D,GAAGz8D,EAAE,iBAAiB,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,oBAAoBy8D,GAAGz8D,EAAE,mBAAmB,EAAE,KAAK,gBAAgB,GAAG,KAAK,UAAU,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAEA,EAAE,OAAO,GAAG,KAAK,QAAQ,OAAO,KAAK,OAAO,KAAK,UAAU,SAAS,CAACZ,EAAE,KAAK,KAAK,EAAE,KAAK,KAAK,kBAAkB,KAAK,kBAAkB,GAAG,KAAK,gBAAgB,EAAE,KAAK,UAAU,KAAK,KAAK,KAAK,UAAU,OAAO,CAAC,KAAK,KAAK,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,IAAI,KAAK,UAAU,CAAC,CAAC,QAAQ,EAAE,KAAK,CAAC,CAAC,IAAIA,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,mBAAmBY,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,EAAE,OAAOZ,EAAEA,EAAE,OAAO,GAAG,KAAK,MAAMA,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAEP,EAAEinD,GAAG,KAAK,WAAW,aAAa,CAAC,EAAEzlD,EAAE,OAAOxB,GAAG,KAAKwB,EAAEooD,GAAG,EAAE,KAAK,OAAO,KAAK,EAAE5pD,EAAE,KAAK,KAAK,KAAK,KAAK,KAAK,EAAE,IAAI,GAAGwB,EAAEooD,GAAG,EAAE,KAAK,OAAO,KAAK,CAAC,EAAE,KAAK,MAAM,OAAOpoD,EAAEwoD,GAAGxoD,EAAE,KAAK,KAAK,KAAK,CAAC,GAAG,KAAK,YAAY,OAAOA,EAAE,KAAK,WAAW,MAAMA,CAAC,IAAIA,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIjB,EAAE,CAAC,MAAM,KAAK,MAAM,WAAW87D,GAAG,KAAK,UAAU,EAAE,QAAQ,KAAK,QAAQ,kBAAkBxQ,GAAG,KAAK,iBAAiB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,kBAAkBiR,GAAG,KAAK,iBAAiB,EAAE,gBAAgBA,GAAG,KAAK,eAAe,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,iBAAiBnO,GAAG,KAAK,gBAAgB,EAAE,eAAeA,GAAG,KAAK,cAAc,CAAC,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEq/D,GAAG,UAAU,QAAQ1oC,GAAE,cAAc0oC,EAAE,EAAE,IAAIC,GAAG,cAAc/S,EAAE,CAAC,YAAYvsD,EAAE,CAACA,EAAEA,GAAG,CAAC,EAAE,MAAMA,CAAC,EAAE,KAAK,UAAU,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,KAAK,WAAWA,EAAE,UAAU,CAAC,mBAAmBA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,QAAQZ,KAAKY,EAAE,MAAM,CAAC,EAAE,GAAGZ,GAAG,KAAK,MAAM,IAAI8lD,EAAE,iEAAiEllD,EAAE,MAAM,CAAC,kHAAkH,EAAE,MAAM,CAACA,EAAE,GAAGkoD,GAAGloD,EAAE,CAAC,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,GAAG,KAAK,aAAa,iBAAiB,EAAE,KAAK,EAAE,CAAC,IAAIP,EAAE,CAAC,CAAC,EAAE,QAAQwB,EAAE,EAAEA,EAAE,EAAE,KAAK,EAAEA,EAAExB,EAAE,KAAKwB,CAAC,EAAExB,EAAE,KAAK,CAAC,EAAE,EAAEs0B,GAAG,EAAEt0B,CAAC,CAAC,CAAC,OAAOqpD,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI9oD,EAAE,CAAC,EAAE,KAAK,YAAY,OAAOA,EAAE,WAAW,KAAK,YAAY,IAAIZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEs/D,GAAG,UAAU,UAAU3oC,GAAE,cAAc2oC,EAAE,EAAE,IAAIC,GAAG,cAAchT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,WAAWg8D,GAAGh8D,EAAE,UAAU,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAO,KAAK,WAAW,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,WAAW87D,GAAG,KAAK,UAAU,CAAC,EAAE18D,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEu/D,GAAG,UAAU,aAAa5oC,GAAE,cAAc4oC,EAAE,EAAE,IAAIC,GAAG,cAAcjT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,EAAEA,EAAE,EAAE,KAAK,UAAU,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,MAAM,CAACA,EAAE,GAAG,KAAK,EAAEA,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,KAAK3yB,EAAE0rD,GAAG1rD,CAAC,EAAE2oD,GAAG3oD,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,EAAE,KAAK,CAAC,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEw/D,GAAG,UAAU,eAAe7oC,GAAE,cAAc6oC,EAAE,EAAE,IAAIC,GAAG,cAAclT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,YAAYA,EAAE,YAAY,QAAQZ,EAAE,EAAEA,EAAE,KAAK,YAAY,OAAO,EAAEA,EAAE,KAAK,UAAU,KAAK,YAAYA,EAAE,IAAI,KAAK,YAAYA,GAAG,KAAK,CAAC,UAAUY,EAAE,CAAC,OAAOA,EAAE,GAAGA,GAAG,IAAI,CAAC,oBAAoBA,EAAEZ,EAAE,CAAC,IAAI,EAAE,6CAA6CK,EAAEL,EAAE,MAAM,EAAE6B,EAAE,EAAEC,EAAE,KAAK,QAAQE,EAAE,EAAEA,EAAE3B,EAAE,OAAO,EAAE2B,EAAE,CAAC,IAAI,EAAE3B,EAAE2B,GAAG,GAAG,KAAK,UAAU,CAAC,EAAE,GAAGF,IAAI,KAAKA,EAAEE,MAAO,OAAM,IAAI8jD,EAAE,0CAA0C,OAAOjkD,GAAG,CAAC,CAAC,IAAIE,EAAE+mD,GAAGloD,CAAC,EAAE,GAAGkB,IAAI,KAAK,CAAC,GAAGD,IAAI,GAAGE,EAAEF,IAAI,EAAE,MAAM,IAAIikD,EAAE,CAAC,EAAEzlD,EAAEyB,GAAGC,EAAEF,CAAC,SAASE,IAAIF,EAAE,MAAM,IAAIikD,EAAE,CAAC,EAAE,OAAOzlD,CAAC,CAAC,mBAAmBO,EAAE,CAAC,IAAIZ,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAEY,EAAE,OAAO,EAAE,EAAE,GAAG,KAAK,UAAUA,EAAE,EAAE,EAAE,CAACZ,EAAE,GAAG,KAAK,CAAC,OAAOA,EAAEY,EAAE,MAAM,EAAE,CAAC,EAAE,OAAO,KAAK,WAAW,EAAEA,EAAE,MAAM,EAAE,CAAC,EAAE,OAAO,KAAK,oBAAoBA,EAAE,MAAM,CAAC,EAAE,KAAK,WAAW,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAEP,EAAE,EAAE,MAAMwB,EAAExB,EAAE,MAAM,EAAE,CAAC,EAAE,OAAO,KAAK,oBAAoBA,EAAE,MAAM,CAAC,EAAE,KAAK,WAAW,CAAC,EAAE,OAAO27B,EAAE,EAAEn6B,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIjB,EAAE,CAAC,YAAY,KAAK,WAAW,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEy/D,GAAG,UAAU,UAAU9oC,GAAE,cAAc8oC,EAAE,EAAE,IAAIC,GAAG,cAAcnT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAEA,EAAE,MAAM,KAAK,MAAM,IAAI,MAAM,iFAAiF,EAAE,GAAG,CAAC,MAAM,QAAQA,EAAE,IAAI,EAAE,MAAM,IAAI,MAAM,sEAAsEA,EAAE,eAAe,EAAE,IAAIZ,EAAEipD,GAAG,EAAEroD,EAAE,KAAK,OAAO,CAAC,EAAE,GAAG,CAACgmB,EAAE,YAAYhmB,EAAE,KAAK,MAAM,EAAE,KAAK,EAAEZ,CAAC,EAAE,MAAM,IAAI,MAAM,+BAA+B,KAAK,UAAUY,EAAE,IAAI,EAAE,4DAA4D,EAAE,KAAK,KAAKA,EAAE,KAAK,KAAK,mBAAmB,CAAC,CAAC,EAAE,OAAO,KAAK,IAAI,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,KAAK,KAAK,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,EAAE,OAAO,KAAK,KAAK,QAAQ,CAAC,EAAEP,IAAI,CAACL,EAAEK,EAAE,GAAGO,EAAE,EAAE,CAAC,EAAEZ,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAO20B,GAAG23B,GAAG1rD,CAAC,EAAE,KAAK,kBAAkB,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,KAAK,KAAK,IAAI,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE0/D,GAAG,UAAU,UAAU/oC,GAAE,cAAc+oC,EAAE,EAAE,IAAIC,GAAG,cAAcpT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,KAAK,CAAC,EAAEA,CAAC,EAAE,KAAK,gBAAgB,GAAGA,GAAG,KAAK,KAAK,UAAUA,EAAE,WAAW,KAAK,EAAEA,EAAE,UAAU,KAAK,UAAU,CAAC,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,UAAU,KAAK,SAAS,EAAE,OAAO,OAAO,OAAOA,EAAEY,CAAC,EAAEZ,CAAC,CAAC,YAAYY,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAEP,EAAE,GAAG,OAAOy5B,GAAGqP,GAAG,EAAE,KAAK,SAAS,EAAE9oC,CAAC,CAAC,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAEP,EAAE,GAAGwB,EAAE,GAAGC,EAAEg4B,GAAGqP,GAAG,EAAE,KAAK,SAAS,EAAE9oC,EAAEwB,CAAC,EAAE,OAAOq3B,EAAE,EAAE3I,GAAEzuB,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEy+D,GAAG,UAAU,UAAUhpC,GAAE,cAAcgpC,EAAE,EAAE,IAAIC,GAAG,cAAcrT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,WAAW,KAAK,KAAK,+BAA+B,gBAAgBA,EAAE,iBAAiB,MAAMA,EAAE,YAAY,KAAK,CAAC,IAAIZ,EAAE,KAAKY,EAAE,WAAW,OAAOZ,EAAEY,EAAE,WAAWA,EAAE,aAAa,KAAK,KAAK,gBAAgB,CAACZ,EAAE,IAAI,EAAE,KAAK,gBAAgB,CAACA,CAAC,EAAE,OAAOsmD,GAAG1lD,EAAE,WAAW,CAAC,CAAC,CAAC,KAAK,SAASA,EAAE,SAASumD,GAAG,KAAK,SAAS,UAAU,EAAE,KAAK,UAAUvmD,EAAE,UAAUumD,GAAG,KAAK,UAAU,WAAW,EAAE,KAAK,sBAAsBgF,GAAGvrD,EAAE,uBAAuB,KAAK,8BAA8B,EAAE,KAAK,sBAAsBy8D,GAAGz8D,EAAE,qBAAqB,EAAE,KAAK,oBAAoBy8D,GAAGz8D,EAAE,mBAAmB,EAAE,KAAK,qBAAqBsuD,GAAGtuD,EAAE,oBAAoB,EAAE,KAAK,SAASA,EAAE,SAAS,KAAK,gBAAgBA,EAAE,SAAS,KAAK,YAAYA,EAAE,WAAW,CAAC,MAAMA,EAAE,CAAC,KAAK,WAAW,KAAK,UAAU,aAAa,CAAC,KAAK,SAAS,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,GAAG,KAAK,oBAAoB,EAAE,KAAK,MAAM,EAAE,CAAC,6BAA6BA,EAAE,CAAC,CAAC,YAAYA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,KAAK,UAAU3yB,EAAE0rD,GAAG1rD,CAAC,EAAEuoC,GAAGvoC,EAAEsgC,GAAGtgC,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC,mBAAmBA,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAE,KAAK,aAAa,KAAK,MAAM,CAAC,GAAGA,EAAE,KAAK,SAAS,EAAE,IAAIZ,EAAEsmD,GAAG,KAAK,WAAW,EAAE,GAAGtmD,EAAE,SAASY,EAAE,OAAO,EAAE,MAAM,IAAIklD,EAAE,oBAAoB,KAAK,mDAAmDllD,GAAG,EAAE,CAAC,IAAI,EAAE,EAAE,QAAQP,EAAE,EAAEA,EAAEL,EAAE,OAAO,EAAEK,EAAE,CAAC,IAAIwB,EAAE7B,EAAEK,GAAGyB,EAAElB,EAAEP,EAAE,GAAG,GAAGwB,GAAG,MAAMC,GAAG,MAAMD,IAAIC,EAAE,MAAM,IAAIgkD,EAAE,oBAAoB,KAAK,mDAAmDllD,GAAG,EAAEiB,GAAG,OAAO7B,EAAE,GAAG8B,GAAG,GAAG,CAAC,CAAC,MAAM,CAAClB,EAAE,GAAG,GAAGZ,EAAE,KAAK,SAAS,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,EAAE,QAAQ,UAAU,EAAEyoD,GAAG,EAAE,OAAO,GAAG,IAAIhpD,EAAE8pD,GAAG,KAAK,WAAW,KAAK,EAAEnuB,EAAE,EAAE,CAAC,EAAE,IAAI,CAAC,CAAC,EAAE,OAAOA,EAAE37B,EAAEksD,GAAG,KAAK,mBAAmB,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI3rD,EAAE,CAAC,SAAS,KAAK,SAAS,UAAU,KAAK,UAAU,sBAAsBsrD,GAAG,KAAK,qBAAqB,EAAE,sBAAsBiR,GAAG,KAAK,qBAAqB,EAAE,oBAAoBA,GAAG,KAAK,mBAAmB,EAAE,qBAAqBnO,GAAG,KAAK,oBAAoB,EAAE,SAAS,KAAK,SAAS,YAAY,KAAK,WAAW,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE4/D,GAAG,UAAU,YAAYjpC,GAAE,cAAcipC,EAAE,EAAE,IAAIC,GAAG,cAActT,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,GAAG,CAAC,CAAC,EAAE,KAAK,gBAAgB,EAAE,CAAC,cAAcA,EAAE,CAAC,MAAM,IAAImlD,EAAE,CAAC,gCAAgCnlD,EAAEZ,EAAE,CAAC,GAAGY,GAAG,MAAMZ,GAAG,KAAK,OAAO,KAAK,GAAGY,EAAE,OAAOZ,EAAE,OAAO,OAAO,KAAK,gCAAgCA,EAAEY,CAAC,EAAE,GAAGZ,EAAE,SAAS,EAAE,OAAOY,EAAE,IAAI,EAAEA,EAAE,MAAM,EAAEA,EAAE,OAAOZ,EAAE,MAAM,EAAE,QAAQK,EAAE,EAAEA,EAAEL,EAAE,OAAO,EAAEK,EAAE,CAAC,IAAIwB,EAAEjB,EAAEA,EAAE,OAAOZ,EAAE,OAAOK,GAAGyB,EAAE9B,EAAEK,GAAG,GAAGwB,GAAG,MAAMC,GAAG,MAAMD,EAAE,GAAGC,EAAE,EAAE,EAAE,KAAK,IAAI,UAAUD,IAAI,EAAE,EAAE,KAAKC,CAAC,UAAUA,IAAI,EAAE,EAAE,KAAKD,CAAC,MAAM,CAAC,GAAGA,IAAIC,EAAE,MAAM,IAAIgkD,EAAE,wDAAwD,KAAK,UAAUllD,CAAC,EAAE,IAAI,KAAK,UAAUZ,CAAC,CAAC,EAAE,EAAE,KAAK6B,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,MAAMjB,EAAE,CAAC,GAAG,MAAM,QAAQA,CAAC,GAAG,CAAC,MAAM,QAAQA,EAAE,EAAE,IAAIA,EAAE,CAAC2rD,GAAG3rD,CAAC,CAAC,GAAGA,EAAEA,EAAEA,EAAE,OAAO,EAAE,MAAM,IAAIklD,EAAE,wEAAwEllD,EAAE,kBAAkB,EAAE,IAAIZ,EAAE,CAAC,EAAE,QAAQ6B,KAAKjB,EAAEiB,GAAG,MAAMA,EAAE,KAAK,MAAM7B,EAAE,KAAK6B,EAAE,EAAE,EAAE,GAAG7B,EAAE+mD,GAAG/mD,CAAC,EAAEA,EAAE,OAAO,EAAE,MAAM,IAAI8lD,EAAE,8EAA8E,KAAK,UAAUllD,CAAC,IAAI,EAAE,IAAI,EAAEA,EAAE,IAAI,KAAK,KAAKA,EAAE,GAAG,MAAM,CAAC,EAAE,QAAQiB,EAAE,EAAEA,EAAEjB,EAAE,OAAO,EAAEiB,EAAE,CAAC,IAAIC,EAAElB,EAAEiB,IAAI,KAAK,KAAKjB,EAAEiB,GAAG,MAAM,CAAC,EAAE,EAAE,KAAK,gCAAgC,EAAEC,CAAC,CAAC,CAAC,IAAIzB,EAAEO,EAAE,IAAIiB,GAAGA,EAAE,MAAM,EAAEjB,EAAE,QAAQ,IAAI,IAAI,IAAImmD,GAAG1mD,CAAC,EAAE,SAAS,EAAE,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,EAAE,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG3yB,EAAEA,EAAE,KAAK,gBAAgB,CAAC,IAAI,EAAE,CAAC,EAAEP,EAAEO,EAAE,IAAIiB,GAAGA,EAAE,IAAI,EAAE,GAAGxB,EAAE,QAAQ,IAAI,IAAI,GAAG,CAAC,IAAIwB,EAAEmnD,GAAG3oD,CAAC,EAAE,QAAQyB,KAAKlB,EAAE,CAAC,IAAImB,EAAED,EAAE,KAAK,QAAQE,EAAE,EAAEA,EAAEH,EAAEE,EAAE,EAAEC,EAAEF,EAAEwnD,GAAGxnD,EAAE,CAAC,EAAE,EAAE,KAAKA,CAAC,CAAC,CAAC,OAAO,KAAK,cAAc,CAAC,CAAC,KAAK,CAAC,IAAID,EAAE,GAAG,QAAQG,KAAKpB,EAAE,CAAC,IAAI,EAAEoB,EAAE,KAAK,GAAG,GAAG,KAAK,CAAC,IAAIU,EAAEV,EAAE,MAAM,EAAEU,EAAE,GAAGE,EAAEF,EAAE,MAAM,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,EAAEG,EAAEm5B,EAAEh6B,EAAE,CAAC,CAAC,EAAE,OAAO8mD,GAAGpmD,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,EAAEG,EAAE8xB,GAAG9xB,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEA,EAAEm5B,EAAEn5B,EAAED,CAAC,EAAE,EAAE,KAAKC,CAAC,EAAEhB,EAAE,EAAE,SAAS,EAAE,EAAE,CAAC,IAAIa,EAAEumD,GAAG,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,KAAKt0B,GAAG3yB,EAAEU,CAAC,CAAC,EAAEb,EAAE,EAAE,MAAM,EAAE,KAAKG,CAAC,CAAC,CAAC,IAAIF,EAAE,KAAK,cAAc,CAAC,EAAEC,EAAED,EAAE,KAAK,GAAGD,GAAG,GAAGE,GAAG,KAAK,CAAC,IAAIC,EAAEF,EAAE,MAAM,EAAEE,EAAE,OAAOU,EAAEV,EAAE,EAAE,GAAG,EAAE,CAACU,CAAC,EAAE,OAAOV,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,CAAC,EAAEF,EAAEk6B,EAAErH,GAAGqH,EAAEl6B,EAAE,CAAC,GAAGY,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,SAASX,EAAE,EAAE,CAAC,IAAIC,EAAE,CAACD,EAAE,CAAC,EAAE,OAAOknD,GAAG,EAAElnD,EAAE,CAAC,CAAC,EAAED,EAAE6yB,GAAG7yB,EAAEE,CAAC,CAAC,EAAE,OAAOF,CAAC,CAAC,KAAM,QAAO,KAAK,cAAclB,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAACA,EAAEA,EAAE,IAAIZ,EAAEY,EAAE,IAAI,KAAKZ,EAAE,KAAKA,EAAEY,EAAE,GAAG,MAAM,CAAC,EAAE,QAAQP,EAAE,EAAEA,EAAEO,EAAE,OAAO,EAAEP,EAAE,CAAC,IAAIwB,EAAEjB,EAAEP,IAAI,KAAK,KAAKO,EAAEP,GAAG,MAAM,CAAC,EAAEL,EAAE,KAAK,gCAAgCA,EAAE6B,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQxB,KAAKO,EAAEP,GAAG,MAAMA,EAAE,KAAK,MAAM,EAAE,KAAKA,EAAE,EAAE,EAAE,OAAO,EAAE0mD,GAAG,CAAC,EAAE,EAAE,SAAS,EAAE/mD,EAAE,EAAE,OAAOA,CAAC,EAAEA,EAAE,CAAC,IAAI,EAAE,OAAOA,CAAC,EAAEA,CAAC,CAAC,YAAYY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAGvzB,GAAG,KAAK,OAAO,KAAK,GAAG,CAAC,MAAM,QAAQA,CAAC,EAAE,MAAM,IAAI8lD,EAAE,2BAA2B,EAAE,GAAG,CAAC,MAAM,QAAQllD,CAAC,EAAE,MAAM,IAAIklD,EAAE,6BAA6B,EAAE,GAAG9lD,EAAE,SAASY,EAAE,OAAO,MAAM,IAAIklD,EAAE,mGAAmGllD,EAAE,aAAaZ,EAAE,SAAS,EAAE,GAAGA,EAAE,MAAMK,GAAGA,GAAG,IAAI,EAAE,OAAO,KAAKL,EAAEA,EAAE,IAAIK,GAAGA,GAAG,KAAKA,EAAEqjC,GAAGrjC,EAAE,CAAC,CAAC,EAAE,IAAI,EAAEL,EAAE,GAAG,QAAQK,EAAE,EAAEA,EAAEL,EAAE,OAAO,EAAE,EAAEK,EAAE,EAAEymC,GAAG,EAAE9mC,EAAEK,EAAE,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,EAAEqgE,GAAG,cAAcD,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEY,EAAE,GAAG,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEZ,EAAE44B,EAAE54B,EAAEY,EAAE,EAAE,EAAE,OAAOZ,CAAC,CAAC,CAAC,CAAC,EAAE0gE,GAAG,UAAU,MAAMnpC,GAAE,cAAcmpC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEY,EAAE,GAAG,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEZ,EAAEk5B,EAAEl5B,EAAEY,EAAE,EAAE,EAAE,OAAOZ,CAAC,CAAC,CAAC,CAAC,EAAE2gE,GAAG,UAAU,WAAWppC,GAAE,cAAcopC,EAAE,EAAE,IAAIC,GAAG,cAAcH,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEY,EAAE,GAAG,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEZ,EAAE44B,EAAE54B,EAAEY,EAAE,EAAE,EAAE,OAAOs4B,EAAE,EAAEt4B,EAAE,OAAOZ,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE4gE,GAAG,UAAU,UAAUrpC,GAAE,cAAcqpC,EAAE,EAAE,IAAIC,GAAG,cAAcJ,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEY,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEZ,EAAEgoC,GAAGhoC,EAAEY,EAAE,EAAE,EAAE,OAAOZ,CAAC,CAAC,CAAC,CAAC,EAAE6gE,GAAG,UAAU,UAAUtpC,GAAE,cAAcspC,EAAE,EAAE,IAAIC,GAAG,cAAcL,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAEY,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEZ,EAAEuoC,GAAGvoC,EAAEY,EAAE,EAAE,EAAE,OAAOZ,CAAC,CAAC,CAAC,CAAC,EAAE8gE,GAAG,UAAU,UAAUvpC,GAAE,cAAcupC,EAAE,EAAE,IAAIC,GAAG,cAAcN,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,aAAa,GAAGA,GAAG,OAAOA,EAAE,CAAC,GAAG,KAAK,KAAKA,EAAE,MAAM,KAAK,KAAK,aAAaA,EAAE,KAAK,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,EAAE,CAAC,MAAMA,EAAE,CAAC,GAAG,EAAE,MAAM,QAAQA,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,IAAIA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,uEAAuE,EAAEllD,EAAEA,EAAE,IAAIZ,EAAE,GAAG,QAAQK,KAAKO,EAAE,GAAGP,GAAG,KAAK,CAACL,EAAE,GAAG,KAAK,CAAC,GAAGA,EAAE,OAAO,IAAI,EAAE,CAAC,EAAE,QAAQK,EAAE,EAAEA,EAAEO,EAAE,OAAO,EAAEP,EAAE,CAAC,IAAIwB,EAAEjB,EAAEP,GAAG,MAAM,EAAEwB,EAAE,OAAO,KAAK,KAAK,CAAC,EAAE,IAAIC,EAAE,GAAG,QAAQC,KAAK,EAAE,GAAG6kB,EAAE,YAAY7kB,EAAEF,CAAC,EAAE,CAACC,EAAE,GAAG,KAAK,CAACA,GAAG,EAAE,KAAKD,CAAC,CAAC,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,IAAIikD,EAAE,4GAA4G,KAAK,UAAUllD,CAAC,CAAC,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO2yB,EAAE,IAAIu2B,GAAGlpD,EAAE,KAAK,IAAI,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,GAAG,EAAE,MAAM,QAAQA,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,GAAG,MAAM,IAAIklD,EAAE,6DAA6D,EAAE,IAAI9lD,EAAEY,EAAE,EAAEZ,EAAE,GAAG,MAAM,EAAEK,EAAE,KAAK,KAAK,EAAE,EAAE,OAAO,KAAK,KAAK,KAAK,KAAK,QAAQwB,KAAK7B,EAAE,MAAM,CAAC,EAAE,CAAC,GAAG,EAAEK,IAAI,MAAMwB,EAAExB,IAAI,KAAK,CAAC,EAAEA,GAAG,KAAK,KAAK,CAAC,EAAEA,IAAIwB,EAAExB,EAAE,CAAC,OAAO,CAAC,CAAC,YAAYO,EAAEZ,EAAE,CAAC,GAAGA,GAAG,KAAK,OAAO,KAAK,GAAG,CAAC,MAAM,QAAQA,CAAC,EAAE,MAAM,IAAI8lD,EAAE,2CAA2C,EAAE,GAAG,CAAC,MAAM,QAAQllD,CAAC,EAAE,MAAM,IAAIklD,EAAE,6CAA6C,EAAE,GAAG9lD,EAAE,SAASY,EAAE,OAAO,MAAM,IAAIklD,EAAE,mCAAmC9lD,EAAE,qCAAqCY,EAAE,SAAS,EAAE,OAAO2yB,EAAE,IAAI,CAAC,IAAI,EAAE,GAAG,GAAGvzB,EAAE,QAAQ8B,GAAG,CAAC,GAAGA,GAAG,KAAK,CAAC,EAAE,GAAG,MAAM,CAAC,CAAC,EAAE,EAAE,OAAO,KAAK,IAAIzB,EAAE,CAAC,EAAE,QAAQyB,EAAE,EAAEA,EAAElB,EAAE,OAAO,EAAEkB,EAAE9B,EAAE8B,IAAI,KAAKzB,EAAE,KAAKkwB,GAAE8Y,GAAGzoC,EAAEkB,EAAE,EAAE,MAAM,CAAC,EAAE9B,EAAE8B,GAAG,KAAKlB,EAAEkB,GAAG,KAAKzB,EAAE,KAAKqjC,GAAG1jC,EAAE8B,GAAG,EAAE,CAAC,EAAEzB,EAAE,KAAKL,EAAE8B,EAAE,EAAE,IAAID,EAAEy6B,GAAGj8B,EAAE,KAAK,IAAI,EAAE,OAAOu5B,GAAG/3B,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIjB,EAAE,CAAC,KAAK,KAAK,IAAI,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEmgE,GAAG,UAAU,cAAcxpC,GAAE,cAAcwpC,EAAE,EAAE,SAASC,GAAGlhE,EAAE,EAAE,CAAC,KAAKA,EAAE,GAAGA,GAAG,EAAE,OAAOA,CAAC,CAAC,SAASmhE,GAAGnhE,EAAE,EAAEE,EAAE,CAAC,GAAGF,EAAE,MAAM,OAAO,GAAG,EAAE,MAAM,OAAO,EAAE,MAAM,IAAIimD,GAAG,kEAAkE,EAAE,GAAGn/B,EAAE,OAAO9mB,EAAE,MAAM,QAAQ,EAAE,IAAI,uDAAuDA,EAAE,MAAM,QAAQ,EAAE8mB,EAAE,OAAO9mB,EAAE,MAAM,QAAQ,EAAE,IAAI,uDAAuD,EAAE,MAAM,QAAQ,EAAE,OAAOE,GAAG,WAAWA,EAAE,CAACA,EAAEA,CAAC,GAAGF,EAAE,QAAQ,aAAa,EAAE,QAAQ,YAAY,MAAM,IAAIimD,GAAG,6DAA6D,EAAE,IAAI3lD,EAAEN,EAAE,MAAM,OAAOO,EAAE,EAAE,MAAM,OAAOL,GAAG,OAAOA,EAAE,CAACI,EAAE,EAAEC,EAAE,CAAC,GAAG,IAAI,EAAEL,EAAE,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE,GAAGnzB,EAAEC,EAAE,CAAC,EAAED,EAAEC,EAAE,IAAI2B,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAED,EAAE,KAAK,CAAC,EAAE,EAAEg6B,EAAE,EAAE,EAAE,MAAM,OAAOh6B,CAAC,CAAC,CAAC,SAAS3B,EAAED,EAAE,CAAC,EAAEC,EAAED,EAAE,IAAI4B,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAED,EAAE,KAAK,CAAC,EAAElC,EAAEk8B,EAAEl8B,EAAEA,EAAE,MAAM,OAAOkC,CAAC,CAAC,CAAC,MAAM,EAAE,EAAE,IAAID,EAAE,GAAGjC,EAAE,MAAM,SAAS,GAAG,EAAE,MAAM,SAAS,EAAE,EAAE,KAAK,EAAE,GAAGiC,EAAEkhC,GAAG/J,EAAEp5B,EAAE,CAAC,EAAE,EAAE,EAAE,EAAEiC,EAAEkhC,GAAG/J,EAAEvE,GAAG70B,EAAE,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,EAAE,MAAM,CAAC,IAAIkC,EAAE,EAAE,KAAKlC,EAAE,MAAM,OAAO,EAAEmC,EAAE,EAAE,KAAK,EAAE,MAAM,OAAO,EAAEF,EAAE6wB,GAAG9yB,EAAE,EAAEkC,EAAEC,CAAC,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,IAAID,EAAE5B,EAAEC,EAAE2B,EAAE5B,EAAEC,EAAE,EAAE2B,EAAE5B,EAAE,EAAE,IAAI6B,EAAE,CAAC,EAAE,QAAQS,EAAEV,EAAEU,EAAEV,EAAE,EAAE,EAAEU,EAAET,EAAE,KAAKS,CAAC,EAAEX,EAAEytC,GAAGztC,EAAEE,CAAC,CAAC,CAAC,OAAOF,EAAE,MAAM,SAAS,IAAIA,EAAE2hC,GAAG3hC,EAAE,CAAC,GAAGA,CAAC,CAAC,CAAC,CAAC,IAAIm/D,GAAG,cAAcT,EAAE,CAAC,YAAY7/D,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,KAAKA,EAAE,KAAK,KAAK,UAAUA,EAAE,WAAW,KAAK,GAAGA,EAAE,UAAU,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,EAAE,CAAC,MAAMA,EAAE,CAACgmB,EAAE,OAAO,MAAM,QAAQhmB,CAAC,GAAGA,EAAE,SAAS,GAAG,MAAM,QAAQA,EAAE,EAAE,GAAG,MAAM,QAAQA,EAAE,EAAE,EAAE,IAAI,+DAA+D,EAAE,IAAIZ,EAAEY,EAAE,GAAG,EAAEA,EAAE,GAAG,GAAGZ,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI+lD,GAAG,8DAA8D,EAAE,IAAI1lD,EAAE,KAAK,cAAcL,EAAE,CAAC,EAAE,GAAGA,EAAEK,EAAE,MAAM,EAAEA,EAAE,IAAI,MAAM,IAAIylD,EAAE,8BAA8B9lD,EAAEK,EAAE,WAAW,EAAEA,EAAE,KAAK,CAAC,CAAC,cAAcO,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAIklD,EAAE,oEAAoEllD,EAAE,kBAAkB,EAAE,IAAIZ,EAAEY,EAAE,GAAG,EAAEA,EAAE,GAAGP,EAAE,OAAO,MAAM,QAAQ,KAAK,IAAI,EAAEA,EAAE,KAAK,KAAK,IAAI,CAACwB,EAAEC,IAAIk/D,GAAGn/D,EAAEjB,EAAEkB,GAAG,MAAM,MAAM,CAAC,EAAEzB,EAAE,CAAC2gE,GAAG,KAAK,KAAKhhE,EAAE,MAAM,MAAM,EAAEghE,GAAG,KAAK,KAAK,EAAE,MAAM,MAAM,CAAC,EAAE,KAAK,YAAYhhE,EAAE81D,GAAG91D,EAAEK,EAAE,EAAE,EAAE,EAAEy1D,GAAG,EAAEz1D,EAAE,EAAE,GAAG4gE,GAAGjhE,EAAE,EAAEK,CAAC,CAAC,CAAC,cAAcO,EAAEZ,EAAE,CAAC,IAAI,EAAE,OAAO,MAAM,QAAQ,KAAK,IAAI,EAAE,EAAE,KAAK,KAAK,EAAE,CAACghE,GAAG,KAAK,KAAKpgE,EAAE,MAAM,EAAEogE,GAAG,KAAK,KAAKhhE,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,mBAAmBY,EAAE,CAACgmB,EAAE,OAAO,MAAM,QAAQhmB,CAAC,GAAGA,EAAE,SAAS,GAAG,MAAM,QAAQA,EAAE,EAAE,GAAG,MAAM,QAAQA,EAAE,EAAE,EAAE,IAAI,+DAA+D,EAAE,IAAIZ,EAAEY,EAAE,GAAG,MAAM,EAAE,EAAEA,EAAE,GAAG,MAAM,EAAE,GAAGZ,EAAE,OAAO,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI+lD,GAAG,8DAA8D,EAAE,IAAI1lD,EAAE,KAAK,cAAcL,EAAE,CAAC,EAAEA,EAAE,OAAOK,EAAE,GAAG,CAAC,EAAE,EAAE,OAAOA,EAAE,GAAG,CAAC,EAAE,EAAE,OAAO,EAAE,CAAC,EAAE,IAAIwB,EAAE7B,EAAE,OAAO,CAAC,EAAE,OAAO6B,EAAE,SAAS,GAAGA,EAAE,KAAK,CAAC,EAAEA,CAAC,CAAC,YAAYjB,EAAEZ,EAAE,CAAC,OAAO,IAAI,CAAC,WAAW,CAAC,IAAIY,EAAE,CAAC,KAAK,KAAK,KAAK,UAAU,KAAK,SAAS,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEsgE,GAAG,UAAU,MAAM3pC,GAAE,cAAc2pC,EAAE,EAAE,IAAIC,GAAG,cAAchU,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,OAAOA,EAAE,MAAM,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,OAAO,KAAK,MAAM,EAAE,OAAO,OAAO,OAAOA,EAAEY,CAAC,EAAEZ,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAO8pD,GAAG,IAAI9xB,EAAEoxB,GAAG,EAAE,MAAM,EAAE,KAAK,MAAM,EAAE,CAAC,EAAE,IAAI,EAAEhqD,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEmhE,GAAG,UAAU,gBAAgB5pC,GAAE,cAAc4pC,EAAE,EAAE,IAAIC,GAAG,cAAcjU,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,KAAKA,EAAE,IAAI,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,KAAK,KAAK,IAAI,EAAE,OAAO,OAAO,OAAOA,EAAEY,CAAC,EAAEZ,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,IAAI,EAAEssD,GAAG1rD,CAAC,EAAE,OAAO,KAAK,KAAK,GAAG,KAAK,KAAK,EAAE8pD,GAAG,IAAI,CAAC,IAAI7oD,EAAE,KAAK,KAAK,KAAK,MAAM,EAAE,KAAK,KAAK,EAAE,OAAOq3B,EAAE,EAAE8wB,GAAG,EAAE,MAAM,EAAEnoD,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE7B,EAAE,UAAU,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEohE,GAAG,UAAU,kBAAkB7pC,GAAE,cAAc6pC,EAAE,EAAE,IAAIC,GAAG,cAAclU,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,KAAKA,EAAE,KAAK,KAAK,WAAWA,EAAE,UAAU,CAAC,eAAeA,EAAE,CAAC,OAAO,KAAK,YAAY0rD,GAAG1rD,CAAC,EAAE,KAAK,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,MAAM,UAAU,EAAEZ,EAAE,CAAC,KAAK,KAAK,IAAI,EAAE,OAAO,OAAO,OAAOA,EAAEY,CAAC,EAAEZ,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAG,KAAK,KAAK,GAAG,KAAK,KAAK,EAAE,CAAC,IAAI,EAAE,KAAK,eAAe3yB,CAAC,EAAE,OAAO8pD,GAAG,IAAI,CAAC,IAAI7oD,EAAEyqD,GAAG1rD,CAAC,EAAEkB,EAAE,mBAAmBC,EAAE,mBAAmBC,EAAE,CAACF,EAAEC,EAAEE,EAAEuiC,GAAGuH,GAAG,CAAC,EAAE,KAAK,IAAI,EAAE9pC,EAAEonD,GAAGpnD,EAAE,SAAS,EAAE,IAAIS,IAAI,EAAE,KAAK,OAAO,EAAE,KAAK,KAAKV,GAAG,KAAK,IAAIW,EAAE,CAACD,EAAEV,EAAE,KAAK,KAAKY,EAAEg2B,EAAEM,EAAEr3B,EAAEI,CAAC,EAAEi3B,EAAEN,EAAE32B,EAAE,EAAE,EAAED,CAAC,CAAC,EAAE,OAAO42B,EAAEM,EAAEt2B,EAAEF,CAAC,EAAEC,CAAC,CAAC,EAAE,IAAI2pD,GAAG1rD,CAAC,EAAEZ,EAAE,UAAU,EAAE,CAAC,CAAC,OAAOY,CAAC,CAAC,CAAC,CAAC,EAAEygE,GAAG,UAAU,eAAe9pC,GAAE,cAAc8pC,EAAE,EAAE,SAASC,GAAGxhE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,KAAK,CAAC,IAAI,EAAE,GAAGP,EAAE,OAAO,EAAE,EAAEu9B,GAAGv9B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,UAAUP,EAAE,OAAO,EAAE,EAAEy9B,GAAGz9B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,UAAUP,EAAE,OAAO,EAAE,EAAE29B,GAAG39B,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,MAAO,OAAM,IAAI0lD,GAAG,2DAA2DjmD,EAAE,UAAU,EAAE,OAAO,CAAC,CAAC,SAASyhE,GAAGzhE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,KAAK,CAAC,OAAOkzB,EAAE,IAAI,CAAC,IAAI,EAAEsV,GAAG/oC,EAAEM,CAAC,EAAE,EAAE,EAAE,KAAK2B,EAAE,EAAE,SAAS,MAAM,CAACu/D,GAAGxhE,EAAE,EAAEiC,EAAE/B,EAAE,EAAEK,CAAC,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,CAAC,SAASy/D,GAAG1hE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,KAAK,CAAC,OAAOkzB,EAAE,IAAI,CAAC,IAAI,EAAEsV,GAAG/oC,EAAEM,CAAC,EAAE,EAAE,EAAE,KAAK2B,EAAE,EAAE,SAASC,EAAE,CAAC,EAAE,QAAQc,KAAKmmD,GAAG,EAAEnpD,EAAE,IAAI,EAAEM,EAAE,QAAQ0C,CAAC,IAAI,GAAGd,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAKlC,EAAE,MAAMgD,EAAE,EAAE,IAAIb,EAAE+5B,EAAE,EAAEh6B,CAAC,EAAEU,EAAEs5B,EAAEj6B,EAAEC,CAAC,EAAEW,EAAE,GAAG,KAAK,KAAKq5B,EAAE,EAAEh6B,CAAC,EAAEY,EAAE5C,GAAG,KAAK,KAAKg8B,EAAEh8B,EAAEgC,CAAC,EAAE,MAAM,CAACs/D,GAAGxhE,EAAEmC,EAAES,EAAEE,EAAED,EAAEtC,CAAC,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS0/D,GAAG3hE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,KAAK,CAAC,OAAOumB,EAAE,YAAYxmB,EAAE,MAAM,EAAE,KAAK,EAAE6oD,GAAG,EAAEnpD,EAAE,KAAK,CAAC,CAAC,EAAEyhE,GAAGzhE,EAAE,EAAEE,EAAEI,EAAEC,CAAC,EAAEmhE,GAAG1hE,EAAE,EAAEE,EAAEI,EAAEC,CAAC,CAAC,CAAC,IAAIqhE,GAAG,cAAcvU,EAAE,CAAC,YAAYvsD,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,gBAAgB,GAAG,KAAK,KAAKA,EAAE,MAAM,KAAK,GAAGA,EAAE,KAAK,KAAK,SAASA,EAAE,UAAU,KAAK,IAAIA,EAAE,SAAS,KAAK,QAAQA,EAAE,SAAS,KAAK,KAAKA,EAAE,QAAQ,KAAK,OAAOA,EAAE,QAAQ,KAAK,GAAGA,EAAE,OAAO,KAAK,MAAMA,EAAE,OAAO,KAAK,GAAGA,EAAE,MAAM,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,OAAO,EAAE,KAAK,iBAAiBurD,GAAGvrD,EAAE,kBAAkB,MAAM,EAAE,KAAK,sBAAsBurD,GAAGvrD,EAAE,uBAAuB,OAAO,EAAE,KAAK,0BAA0BurD,GAAGvrD,EAAE,2BAA2B,MAAM,EAAE,KAAK,eAAesuD,GAAGtuD,EAAE,cAAc,EAAE,KAAK,gBAAgBsuD,GAAGtuD,EAAE,eAAe,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,iBAAiBy8D,GAAGz8D,EAAE,gBAAgB,CAAC,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,KAAK,MAAM,EAAE,KAAK,KAAK,KAAK,KAAKY,EAAE,OAAO,EAAEA,EAAEZ,GAAG,GAAG,GAAG,KAAK,MAAM,IAAI8lD,EAAE,QAAQ9lD,gGAAgG,KAAK,UAAUY,CAAC,IAAI,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAKlsD,EAAE,OAAO,KAAK,CAAC,CAACZ,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,IAAIK,EAAE,CAAC,CAAC,EAAE,KAAK,QAAQ,KAAK,MAAM,KAAK,UAAU,QAAQA,EAAE,KAAK,KAAK,iBAAiB,KAAK,iBAAiB,GAAG,KAAK,eAAe,GAAG,KAAK,SAAS,KAAK,KAAK,KAAK,UAAU,OAAOA,EAAE,KAAK,KAAK,gBAAgB,KAAK,gBAAgB,GAAG,KAAK,cAAc,GAAG,KAAK,WAAW,KAAK,UAAU,cAAcA,EAAE,KAAK,KAAK,sBAAsB,KAAK,EAAE,EAAE,KAAK,eAAe,KAAK,UAAU,kBAAkBA,EAAE,KAAK,KAAK,0BAA0B,KAAK,EAAE,EAAE,KAAK,MAAM,EAAE,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAEvzB,EAAE,UAAU,KAAK,GAAGA,EAAE,SAASK,EAAEisD,GAAG1rD,CAAC,EAAEiB,EAAExB,EAAE,MAAMyB,EAAED,EAAE,OAAOE,EAAEknD,GAAG,EAAEnnD,CAAC,EAAEE,EAAE,KAAK,MAAM,EAAE,KAAK,KAAK,KAAK,KAAKF,EAAEC,EAAE,OAAOC,EAAE,CAAC,EAAE,IAAI,EAAEkkD,GAAG,EAAEpkD,CAAC,EAAE,EAAEE,GAAGH,EAAEG,GAAG,IAAIU,EAAEX,EAAE,MAAM,EAAEW,EAAE,KAAK,EAAE,IAAI,EAAE,CAACkkB,EAAE,YAAYlkB,EAAEumD,GAAG,EAAEnnD,CAAC,EAAE,MAAM,EAAEA,EAAE,CAAC,CAAC,EAAEc,EAAE,IAAI,CAAC,GAAG,EAAE,CAAC,IAAIiC,EAAEm3B,EAAE,KAAK,WAAW,KAAK,EAAE,CAAC,EAAEv3B,EAAEu3B,EAAE,KAAK,eAAe,KAAK,EAAE,CAAC,EAAEt3B,EAAE,KAAK,OAAOs3B,EAAE,KAAK,KAAK,KAAK,EAAE,CAAC,EAAE,KAAKr3B,EAAE,KAAK,MAAMq3B,EAAE,KAAK,MAAM,KAAK,EAAE,CAAC,EAAE,KAAK,OAAOslC,GAAGjhE,EAAEwE,EAAEJ,EAAEC,EAAEC,EAAE,KAAK,OAAO,CAAC,KAAM,QAAO28D,GAAGjhE,EAAE,KAAK,WAAW,KAAK,EAAE,KAAK,eAAe,KAAK,EAAE,KAAK,MAAM,KAAK,KAAK,KAAK,KAAK,KAAK,EAAE,KAAK,OAAO,KAAK,KAAK,KAAK,MAAM,KAAK,EAAE,KAAK,OAAO,CAAC,EAAE,GAAG,CAAC,EAAE,OAAOuC,EAAE,EAAE,GAAG,CAACC,EAAEC,EAAEkB,CAAC,EAAEy9D,GAAGphE,EAAE,KAAK,MAAM,KAAK,EAAE,KAAK,KAAK,KAAK,EAAE0B,EAAE,KAAK,OAAO,EAAEkC,EAAE,CAACY,EAAEJ,EAAEC,IAAI,CAAC6uB,EAAE,IAAI,CAAC,IAAI5uB,EAAE,EAAED,EAAEE,EAAEC,EAAE,KAAK,EAAEyC,EAAE4xB,EAAEsN,GAAG5hC,EAAEH,CAAC,EAAEE,CAAC,EAAEE,EAAE,MAAM2hC,GAAG5hC,EAAE0C,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,OAAYrD,EAAE,KAAK,WAAWnB,EAAE,KAAK,QAAQ,EAAEmB,EAAE,KAAK,eAAeD,EAAE,KAAK,QAAQ,EAAMnB,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIjC,EAAE,CAAC,KAAK,KAAK,KAAK,SAAS,KAAK,SAAS,QAAQ,KAAK,QAAQ,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,gBAAgBsrD,GAAG,KAAK,eAAe,EAAE,iBAAiBA,GAAG,KAAK,gBAAgB,EAAE,sBAAsBA,GAAG,KAAK,qBAAqB,EAAE,0BAA0BA,GAAG,KAAK,yBAAyB,EAAE,gBAAgBiR,GAAG,KAAK,eAAe,EAAE,iBAAiBA,GAAG,KAAK,gBAAgB,EAAE,eAAenO,GAAG,KAAK,cAAc,EAAE,gBAAgBA,GAAG,KAAK,eAAe,CAAC,EAAEhvD,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE8gE,GAAG,UAAU,qBAAqBnqC,GAAE,cAAcmqC,EAAE,EAAE,IAAIC,GAAG,cAAcxU,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAGA,GAAG,OAAOA,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,KAAKA,EAAE,MAAM,KAAK,GAAGA,EAAE,KAAK,OAAO,KAAK,MAAM,UAAU,GAAG,CAAC,OAAO,UAAU,KAAK,IAAI,EAAE,MAAM,IAAI,MAAM,gDAAgD,KAAK,MAAM,UAAU,MAAM,QAAQ,KAAK,IAAI,GAAG,QAAQZ,KAAK,KAAK,KAAK,GAAG,CAAC,OAAO,UAAUA,CAAC,EAAE,MAAM,IAAI,MAAM,0DAA0D,KAAK,UAAU,KAAK,IAAI,GAAG,MAAO,OAAM,IAAI,MAAM,wEAAwE,KAAK,UAAU,KAAK,IAAI,GAAG,EAAE,KAAK,QAAQY,EAAE,SAAS,KAAK,KAAKA,EAAE,QAAQ,KAAK,OAAOA,EAAE,QAAQ,KAAK,GAAGA,EAAE,OAAO,KAAK,MAAMA,EAAE,OAAO,KAAK,GAAGA,EAAE,MAAM,KAAK,gBAAgBurD,GAAGvrD,EAAE,iBAAiB,OAAO,EAAE,KAAK,iBAAiBurD,GAAGvrD,EAAE,kBAAkB,MAAM,EAAE,KAAK,gBAAgBy8D,GAAGz8D,EAAE,eAAe,EAAE,KAAK,iBAAiBy8D,GAAGz8D,EAAE,gBAAgB,EAAE,KAAK,gBAAgB,EAAE,CAAC,MAAMA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,OAAO,OAAO,KAAK,MAAM,WAAW,KAAK,KAAK,CAAC,KAAK,IAAI,GAAG,QAAQiB,EAAE,EAAEA,EAAE,KAAK,KAAK,OAAO,EAAEA,EAAE,KAAK,KAAKA,GAAG,IAAI,KAAK,KAAKA,IAAI7B,GAAG,QAAQ6B,KAAK,KAAK,KAAK,GAAGA,EAAE,GAAGA,GAAG7B,EAAE,MAAM,IAAI,MAAM,iBAAiB6B,GAAG,EAAE,GAAG,KAAK,KAAK,SAASklD,GAAG,KAAK,IAAI,EAAE,OAAO,MAAM,IAAI,MAAM,4BAA4B,KAAK,MAAM,EAAE,IAAI,EAAE,KAAK,KAAK,IAAIllD,GAAGjB,EAAEiB,EAAE,EAAExB,EAAE,GAAG,KAAK,MAAM,KAAK,MAAM,KAAK,UAAU,QAAQ,EAAE,UAAU,KAAK,iBAAiB,KAAK,iBAAiBA,CAAC,EAAE,KAAK,MAAM,KAAK,KAAK,OAAO,KAAK,KAAK,KAAK,UAAU,OAAO,EAAE,UAAU,KAAK,gBAAgB,KAAK,gBAAgBA,CAAC,EAAE,KAAK,KAAK,KAAK,KAAK,MAAM,EAAE,CAAC,KAAKO,EAAEZ,EAAE,CAAC,IAAI,EAAEssD,GAAG1rD,CAAC,EAAEP,EAAE,EAAE,MAAMwB,EAAExB,EAAE,OAAO,OAAOkzB,EAAE,IAAI,CAAC,GAAG,CAAC,KAAKxxB,EAAE,SAASC,CAAC,EAAE6mC,GAAG,EAAE,KAAK,KAAK,EAAE,EAAE5mC,EAAEikD,GAAG,EAAErkD,CAAC,EAAE,QAAQ,KAAK,KAAK,KAAKI,EAAE,GAAG5B,EAAE,GAAG,IAAIqC,EAAE,GAAG,GAAG,MAAM,EAAE,MAAM,SAASb,EAAEm6B,EAAE,EAAE/5B,CAAC,EAAE,EAAEU,EAAE,KAAK,MAAMD,EAAE,KAAK,MAAM,KAAK,CAAC,EAAE,KAAKE,EAAE,KAAK,OAAOF,EAAE,KAAK,KAAK,KAAK,CAAC,EAAE,KAAKG,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEjB,EAAE,EAAE,EAAE,KAAK,KAAK,QAAQ,CAAC,IAAI,IAAIgB,EAAE,KAAKxC,EAAE,EAAE,EAAEyC,EAAE,KAAK,CAAC,IAAID,EAAE,KAAK,CAAC,EAAEC,EAAE,KAAKzC,EAAE,EAAE,GAAG,OAAO0B,EAAE+hC,GAAG/hC,EAAEc,CAAC,EAAEb,EAAE8hC,GAAG9hC,EAAEa,CAAC,EAAEF,GAAG,OAAOA,EAAEmhC,GAAGnhC,EAAEG,CAAC,GAAGF,GAAG,OAAOA,EAAEkhC,GAAGlhC,EAAEE,CAAC,GAAGw+D,GAAG,EAAEv/D,EAAEC,EAAEY,EAAED,EAAE,KAAK,OAAO,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI/B,EAAE,CAAC,KAAK,KAAK,KAAK,QAAQ,KAAK,QAAQ,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,gBAAgBsrD,GAAG,KAAK,eAAe,EAAE,iBAAiBA,GAAG,KAAK,gBAAgB,EAAE,gBAAgBiR,GAAG,KAAK,eAAe,EAAE,iBAAiBA,GAAG,KAAK,gBAAgB,CAAC,EAAEn9D,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE+gE,GAAG,UAAU,qBAAqBpqC,GAAE,cAAcoqC,EAAE,EAAE,SAASC,GAAG9hE,EAAE,EAAEE,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,GAAGzzB,EAAE,OAAO,EAAE,MAAM,IAAIgmD,EAAE,kEAAkEhmD,EAAE,gBAAgB,EAAE,GAAG,GAAG,OAAO,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,GAAG,EAAE,SAAS,GAAG,EAAE,GAAG,SAAS,GAAG,EAAE,GAAG,SAAS,EAAE,MAAM,IAAIgmD,EAAE,6GAA6G,EAAE,GAAG9lD,GAAG,OAAOA,EAAEopD,GAAG,GAAGppD,IAAI,gBAAgBA,IAAI,gBAAgB,MAAM,IAAI8lD,EAAE,wBAAwB9lD,kEAAkE,EAAE,IAAII,EAAE,OAAOJ,IAAI,gBAAgBI,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,GAAG,EAAE,EAAE,EAAEA,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,GAAG,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC,EAAEqpC,GAAG3pC,EAAEM,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIyhE,GAAG,cAAc1U,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAGA,GAAG,OAAOA,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,WAAWA,EAAE,YAAY,KAAKwoD,GAAG,EAAExoD,EAAE,WAAWA,EAAE,SAAS,KAAK,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,UAAU,OAAOA,EAAE,SAAS,SAAS,KAAK,QAAQ,CAAC,CAACA,EAAE,QAAQA,EAAE,OAAO,EAAE,CAACA,EAAE,QAAQA,EAAE,OAAO,CAAC,MAAM,CAAC,GAAGA,EAAE,QAAQA,EAAE,QAAQA,EAAE,QAAQ,SAAS,EAAE,MAAM,IAAIklD,EAAE,+EAA+EllD,EAAE,QAAQ,eAAe,EAAE,IAAIZ,EAAE,EAAE,GAAG,OAAOY,EAAE,QAAQ,IAAI,SAASZ,EAAE,CAACY,EAAE,QAAQ,GAAGA,EAAE,QAAQ,EAAE,EAAE,EAAE,CAACA,EAAE,QAAQ,GAAGA,EAAE,QAAQ,EAAE,MAAM,CAAC,GAAGA,EAAE,QAAQA,EAAE,QAAQA,EAAE,QAAQ,GAAG,SAAS,EAAE,MAAM,IAAIklD,EAAE,sFAAsFllD,EAAE,QAAQ,GAAG,eAAe,EAAE,GAAGZ,EAAEY,EAAE,QAAQ,GAAGA,EAAE,QAAQ,GAAG,SAAS,EAAE,MAAM,IAAIklD,EAAE,qFAAqFllD,EAAE,QAAQ,GAAG,eAAe,EAAE,EAAEA,EAAE,QAAQ,EAAE,CAAC,KAAK,QAAQ,CAACZ,EAAE,CAAC,CAAC,CAAC,KAAK,UAAU,CAAC,IAAI8sD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,EAAE,OAAO,KAAK,aAAa,iBAAiBY,EAAE,IAAI,MAAMA,EAAE,IAAI,EAAEZ,EAAEY,EAAE,GAAG,KAAK,QAAQ,GAAG,GAAG,KAAK,QAAQ,GAAG,GAAGZ,EAAE,KAAKY,EAAE,IAAI,MAAMA,EAAE,IAAI,EAAE,EAAEA,EAAE,GAAG,KAAK,QAAQ,GAAG,GAAG,KAAK,QAAQ,GAAG,GAAG,EAAE,KAAK,CAACA,EAAE,GAAGA,EAAE,GAAGZ,EAAE,CAAC,IAAIY,EAAE,IAAI,MAAMA,EAAE,IAAI,EAAEZ,EAAEY,EAAE,GAAG,KAAK,QAAQ,GAAG,GAAG,KAAK,QAAQ,GAAG,GAAGZ,EAAE,KAAKY,EAAE,IAAI,MAAMA,EAAE,IAAI,EAAE,EAAEA,EAAE,GAAG,KAAK,QAAQ,GAAG,GAAG,KAAK,QAAQ,GAAG,GAAG,EAAE,KAAK,CAACA,EAAE,GAAGZ,EAAE,EAAEY,EAAE,EAAE,EAAE,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAIquC,GAAGtV,GAAG1rD,CAAC,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,QAAQ,KAAK,QAAQ,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEihE,GAAG,UAAU,gBAAgBtqC,GAAE,cAAcsqC,EAAE,EAAE,SAASC,GAAGhiE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAOkzB,EAAE,IAAI,CAAC00B,GAAG5nD,CAAC,EAAE+nD,GAAG,CAAC,EAAED,GAAG/nD,CAAC,EAAEJ,GAAG,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAGI,GAAG,OAAOA,EAAE,SAASC,GAAG,OAAOA,EAAE+oD,GAAG,GAAG,GAAG,OAAO,EAAE,OAAOtpD,EAAEi+D,GAAGj+D,EAAEO,CAAC,EAAE,IAAI,EAAE0B,EAAE3B,IAAI,OAAO,OAAO,QAAQ,OAAO,IAAI,MAAM,EAAEsnC,GAAG5nC,EAAE,EAAEE,EAAE+B,CAAC,EAAE,EAAEm6B,GAAGp8B,EAAE,EAAEE,EAAE+B,CAAC,EAAE1B,IAAI,kBAAkB,EAAEs0B,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,SAASotC,GAAGjiE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,OAAOkzB,EAAE,IAAI,CAAC00B,GAAG5nD,CAAC,EAAE+nD,GAAG,CAAC,EAAED,GAAG/nD,CAAC,EAAEJ,GAAG,OAAOA,EAAE,CAAC,EAAE,EAAE,CAAC,GAAGI,GAAG,OAAOA,EAAE,SAASC,GAAG,OAAOA,EAAE+oD,GAAG,GAAG,GAAG,OAAO,EAAE,OAAOtpD,EAAEk+D,GAAGl+D,EAAEO,CAAC,EAAE,IAAI,EAAE0B,EAAE3B,IAAI,OAAO,OAAO,QAAQ,OAAO,IAAI,MAAM,EAAEwnC,GAAG9nC,EAAE,EAAEE,EAAE+B,CAAC,EAAE,EAAEq6B,GAAGt8B,EAAE,EAAEE,EAAE+B,CAAC,EAAE1B,IAAI,kBAAkB,EAAEs0B,GAAG,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAIqtC,GAAG,cAAc7U,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAGA,EAAE,UAAU,OAAOA,EAAE,SAAS,GAAG,MAAMA,CAAC,EAAE,OAAOA,EAAE,UAAU,SAAS,KAAK,SAAS,CAACA,EAAE,QAAQ,UAAU,MAAM,QAAQA,EAAE,QAAQ,GAAGA,EAAE,SAAS,SAAS,GAAG,OAAOA,EAAE,SAAS,IAAI,SAAS,KAAK,SAASA,EAAE,aAAc,OAAM,IAAIklD,EAAE,qGAAqG,KAAK,UAAUllD,EAAE,QAAQ,GAAG,EAAE,GAAGumD,GAAG,KAAK,SAAS,UAAU,EAAEvmD,EAAE,SAAS,KAAK,KAAK,QAAQ,KAAK,iBAAiB,OAAOA,EAAE,SAAS,SAAS,KAAK,QAAQ,CAACA,EAAE,OAAO,UAAU,MAAM,QAAQA,EAAE,OAAO,GAAGA,EAAE,QAAQ,SAAS,GAAG,OAAOA,EAAE,QAAQ,IAAI,SAAS,KAAK,QAAQA,EAAE,YAAa,OAAM,IAAIklD,EAAE,oGAAoG,KAAK,UAAUllD,EAAE,OAAO,GAAG,EAAEumD,GAAG,KAAK,QAAQ,SAAS,EAAE,KAAK,QAAQvmD,EAAE,SAAS,KAAK,QAAQA,EAAE,QAAQunD,GAAG,KAAK,OAAO,EAAE,KAAK,UAAU,CAAC,IAAI2E,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE69D,GAAGj9D,EAAE,GAAG,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,MAAM,CAACA,EAAE,GAAGZ,EAAEY,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,KAAK,eAAe3yB,EAAEZ,CAAC,EAAEY,EAAE0oD,GAAGgD,GAAG1rD,CAAC,EAAE,CAAC,EAAE,IAAI,EAAE,KAAK,gBAAgB0rD,GAAG1rD,CAAC,EAAE,CAAC,KAAK,SAAS,GAAG,CAAC,EAAE,CAAC,KAAK,QAAQ,GAAG,CAAC,EAAE,KAAK,QAAQ,cAAc,EAAE,OAAO4uC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAI5uC,EAAE,CAAC,SAAS,KAAK,SAAS,QAAQ,KAAK,QAAQ,QAAQ,KAAK,OAAO,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEqhE,GAAG,cAAcD,EAAE,CAAC,YAAYphE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAEyhE,GAAGlhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAEogE,GAAG,UAAU,eAAe1qC,GAAE,cAAc0qC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAYphE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAEyhE,GAAGlhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAEqgE,GAAG,UAAU,mBAAmB3qC,GAAE,cAAc2qC,EAAE,EAAE,IAAIC,GAAG,cAAchV,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAGA,EAAE,UAAU,OAAOA,EAAE,SAAS,CAAC,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,SAAS,MAAM,QAAQA,EAAE,QAAQ,EAAEA,EAAE,SAAS,CAACA,EAAE,SAASA,EAAE,QAAQ,EAAEA,EAAE,SAAS,KAAK,KAAK,QAAQ,KAAK,iBAAiB,MAAM,QAAQA,EAAE,OAAO,EAAE,CAAC,GAAGA,EAAE,QAAQ,SAAS,EAAE,MAAM,IAAIklD,EAAE,wHAAwHllD,EAAE,QAAQ,SAAS,EAAE,KAAK,QAAQA,EAAE,OAAO,MAAM,KAAK,QAAQ,CAACA,EAAE,QAAQA,EAAE,OAAO,EAAEumD,GAAG,KAAK,SAAS,UAAU,EAAEA,GAAG,KAAK,QAAQ,SAAS,EAAE,KAAK,QAAQvmD,EAAE,SAAS,KAAK,QAAQA,EAAE,QAAQ,KAAK,WAAWA,EAAE,YAAY,KAAK,eAAeA,EAAE,WAAWqnD,GAAG,KAAK,UAAU,EAAEE,GAAG,KAAK,OAAO,EAAE,KAAK,UAAU,CAAC,IAAI2E,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgBY,EAAE,GAAGA,EAAE,GAAG,EAAE,KAAK,aAAa,gBAAgBA,EAAE,GAAGA,EAAE,GAAG,OAAOZ,EAAE69D,GAAG79D,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,EAAE69D,GAAG,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,KAAK,aAAa,gBAAgB,CAACj9D,EAAE,GAAGA,EAAE,GAAGZ,EAAE,CAAC,EAAE,CAACY,EAAE,GAAGZ,EAAE,EAAEY,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,KAAK,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,KAAK,gBAAgBssD,GAAG1rD,CAAC,EAAE,KAAK,SAAS,KAAK,QAAQ,KAAK,QAAQ,KAAK,UAAU,EAAE,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,SAAS,KAAK,SAAS,QAAQ,KAAK,QAAQ,QAAQ,KAAK,QAAQ,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEwhE,GAAG,cAAcD,EAAE,CAAC,YAAYvhE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAEyhE,GAAGlhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAEugE,GAAG,UAAU,eAAe7qC,GAAE,cAAc6qC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAYvhE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAEyhE,GAAGlhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAEwgE,GAAG,UAAU,mBAAmB9qC,GAAE,cAAc8qC,EAAE,EAAE,IAAIC,GAAG,cAAcnV,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAGA,EAAE,UAAU,OAAOA,EAAE,SAAS,CAAC,EAAE,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,SAAS,MAAM,QAAQA,EAAE,QAAQ,EAAEA,EAAE,SAAS,CAACA,EAAE,SAASA,EAAE,SAASA,EAAE,QAAQ,EAAEA,EAAE,SAAS,KAAK,KAAK,QAAQ,KAAK,iBAAiB,MAAM,QAAQA,EAAE,OAAO,EAAE,CAAC,GAAGA,EAAE,QAAQ,SAAS,EAAE,MAAM,IAAIklD,EAAE,wHAAwHllD,EAAE,QAAQ,SAAS,EAAE,KAAK,QAAQA,EAAE,OAAO,MAAM,KAAK,QAAQ,CAACA,EAAE,QAAQA,EAAE,QAAQA,EAAE,OAAO,EAAEumD,GAAG,KAAK,SAAS,UAAU,EAAEA,GAAG,KAAK,QAAQ,SAAS,EAAE,KAAK,QAAQvmD,EAAE,SAAS,KAAK,QAAQA,EAAE,QAAQ,KAAK,WAAWA,EAAE,YAAY,KAAK,eAAeA,EAAE,WAAWqnD,GAAG,KAAK,UAAU,EAAEE,GAAG,KAAK,OAAO,EAAE,KAAK,UAAU,CAAC,IAAI2E,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,KAAK,aAAa,gBAAgBY,EAAE,GAAGA,EAAE,GAAG,EAAE,KAAK,aAAa,gBAAgBA,EAAE,GAAGA,EAAE,GAAGP,EAAE,KAAK,aAAa,gBAAgBO,EAAE,GAAGA,EAAE,GAAG,OAAOZ,EAAE69D,GAAG79D,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,EAAE69D,GAAG,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAEx9D,EAAEw9D,GAAGx9D,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQ,KAAK,QAAQ,EAAE,EAAE,KAAK,aAAa,gBAAgB,CAACO,EAAE,GAAGA,EAAE,GAAGZ,EAAE,EAAEK,CAAC,EAAE,CAACO,EAAE,GAAGZ,EAAE,EAAEK,EAAEO,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,KAAK,KAAK,eAAe3yB,EAAEZ,CAAC,EAAE,KAAK,gBAAgBssD,GAAG1rD,CAAC,EAAE,KAAK,SAAS,KAAK,QAAQ,KAAK,QAAQ,KAAK,UAAU,EAAE,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,SAAS,KAAK,SAAS,QAAQ,KAAK,QAAQ,QAAQ,KAAK,QAAQ,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAE2hE,GAAG,cAAcD,EAAE,CAAC,YAAY1hE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAE0hE,GAAGnhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAE0gE,GAAG,UAAU,eAAehrC,GAAE,cAAcgrC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY1hE,EAAE,CAAC,MAAMA,CAAC,CAAC,CAAC,gBAAgBA,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,OAAOomD,GAAGpmD,CAAC,EAAEsmD,GAAG9nD,CAAC,EAAE0hE,GAAGnhE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,KAAK,CAAC,CAAC,EAAE2gE,GAAG,UAAU,mBAAmBjrC,GAAE,cAAcirC,EAAE,EAAE,IAAIC,GAAG,cAActV,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,UAAU,CAAC,IAAIksD,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAAC,MAAM,CAACA,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,MAAM,IAAI+lD,EAAE,CAAC,EAAE2c,GAAG,cAAcD,EAAE,CAAC,YAAY7hE,EAAE,CAAC,MAAMA,GAAG,CAAC,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,OAAOsnC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEw6B,GAAG,UAAU,yBAAyBnrC,GAAE,cAAcmrC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,YAAY7hE,EAAE,CAAC,MAAMA,GAAG,CAAC,CAAC,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,OAAO0hC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEqgC,GAAG,UAAU,qBAAqBprC,GAAE,cAAcorC,EAAE,EAAE,IAAIC,GAAG,cAAczV,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,WAAWA,EAAE,YAAY,KAAK,eAAeA,EAAE,WAAWqnD,GAAG,KAAK,UAAU,EAAE,KAAK,UAAU,CAAC,IAAI6E,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBlsD,EAAE,CAAC,OAAOA,EAAEA,EAAE,KAAK,aAAa,eAAe,CAACA,EAAE,GAAGA,EAAE,EAAE,EAAE,CAACA,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,MAAM,IAAI+lD,EAAE,CAAC,WAAW,CAAC,IAAInlD,EAAE,CAAC,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,EAAEiiE,GAAG,cAAcD,EAAE,CAAC,KAAKhiE,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,OAAO,KAAK,aAAa,eAAesnC,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEA,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE26B,GAAG,UAAU,yBAAyBtrC,GAAE,cAAcsrC,EAAE,EAAE,IAAIC,GAAG,cAAcF,EAAE,CAAC,KAAKhiE,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE+4B,GAAG1rD,CAAC,EAAE,OAAO,KAAK,aAAa,eAAe0hC,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC,EAAEA,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAEwgC,GAAG,UAAU,qBAAqBvrC,GAAE,cAAcurC,EAAE,EAAE,IAAIC,GAAG,cAAc5V,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,MAAMA,EAAE,KAAK,CAAC,MAAMA,EAAE,CAAC,KAAK,MAAM,EAAE,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,OAAO,KAAK,KAAK,MAAM,UAAU,EAAE,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,OAAO,OAAO,KAAK,MAAM,UAAUA,EAAE,CAAC,IAAI,kBAAkB,CAAC,OAAO,KAAK,MAAM,gBAAgB,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,MAAM,mBAAmB,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,MAAM,QAAQ,CAAC,IAAI,QAAQ,CAAC,OAAO,KAAK,MAAM,MAAM,CAAC,YAAY,CAAC,OAAO,KAAK,MAAM,WAAW,CAAC,CAAC,WAAWA,EAAE,CAAC,KAAK,MAAM,WAAWA,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,MAAM,CAAC,UAAU,KAAK,MAAM,aAAa,EAAE,OAAO,KAAK,MAAM,UAAU,CAAC,CAAC,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,6BAA6BA,EAAE,CAAC,MAAM,6BAA6BA,CAAC,EAAE,KAAK,OAAO,MAAM,KAAK,MAAM,6BAA6BA,CAAC,CAAC,CAAC,OAAO,WAAWA,EAAEZ,EAAE,EAAE,CAAC,EAAE,CAAC,IAAIK,EAAEL,EAAE,MAAM6B,EAAEg0D,GAAGx1D,EAAE,CAAC,EAAE,OAAOL,EAAE,MAAM,IAAI8B,EAAE,CAAC,MAAMD,CAAC,EAAE,OAAO,OAAO,OAAOC,EAAE9B,CAAC,EAAE,IAAIY,EAAEkB,CAAC,CAAC,CAAC,EAAEkhE,GAAG,cAAcD,EAAE,CAAC,YAAYniE,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,gBAAgB,EAAE,CAAC,MAAMA,EAAE,CAAC,GAAGA,EAAE2rD,GAAG3rD,CAAC,EAAEA,EAAE,OAAO,EAAE,MAAM,IAAIklD,EAAE,gFAAgF,KAAK,UAAUllD,CAAC,GAAG,EAAE,KAAK,UAAU,CAAC,CAAC,MAAMA,CAAC,CAAC,EAAE,IAAIZ,EAAE,CAACY,EAAE,EAAE,EAAE,OAAOA,EAAE,MAAM,CAAC,CAAC,EAAE,KAAK,MAAM,QAAQ,KAAK,MAAM,MAAMZ,CAAC,EAAE,KAAK,MAAM,MAAM,IAAI,MAAM,MAAMY,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAE,CAACY,EAAE,EAAE,EAAE,OAAOA,EAAE,MAAM,CAAC,CAAC,EAAE,EAAE,KAAK,MAAM,mBAAmBZ,CAAC,EAAEK,EAAEO,EAAE,GAAG,MAAM,CAAC,EAAE,GAAGP,CAAC,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,KAAKO,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,KAAK3yB,EAAE0rD,GAAG1rD,CAAC,EAAEs+D,GAAG,CAACp9D,EAAE,IAAI,CAACwqD,GAAG,KAAK,MAAM,KAAKxqD,EAAE9B,CAAC,CAAC,EAAE,CAAC,CAAC,EAAEY,EAAE,CAAC,EAAE,GAAG,KAAK,KAAK,GAAG,EAAE,EAAE,GAAG,CAAC,CAAC,EAAEoiE,GAAG,UAAU,kBAAkBzrC,GAAE,cAAcyrC,EAAE,EAAE,SAASC,GAAGnjE,EAAE,CAACmnD,GAAGc,GAAG,yBAAyBjoD,CAAC,CAAC,CAAC,IAAIojE,GAAG,SAASC,GAAG,cAAcJ,EAAE,CAAC,YAAYniE,EAAE,CAAC,MAAMA,CAAC,EAAE,IAAIZ,EAAEY,EAAE,MAAM,UAAU,EAAE,EAAE,CAAC,EAAE,EAAE,UAAUA,EAAE,MAAM,aAAa,EAAE,EAAE,OAAOZ,EAAE,KAAK,aAAa61D,GAAG,CAAC,EAAE71D,EAAE,YAAYA,EAAE,cAAc,GAAG,IAAIK,EAAE,CAAC,EAAE,GAAGA,EAAE,UAAUO,EAAE,MAAM,aAAa,EAAEP,EAAE,OAAOL,EAAE,KAAK,cAAc61D,GAAGx1D,CAAC,EAAE,KAAK,aAAa,KAAK,WAAW,KAAK,aAAa,KAAK,KAAK,cAAc,KAAK,YAAY,KAAK,cAAc,KAAK,KAAK,UAAUO,EAAE,YAAY,OAAOsiE,GAAGtiE,EAAE,UAAUqiE,GAAG,KAAK,SAAS,EAAEriE,EAAE,QAAQ,MAAM,IAAImlD,GAAG,iEAAiE,EAAE,KAAK,UAAUnlD,EAAE,MAAM,SAAS,KAAK,gBAAgBA,EAAE,MAAM,gBAAgB,KAAK,YAAYA,EAAE,MAAM,YAAY,KAAK,gBAAgB,GAAG,KAAK,WAAW,GAAG,KAAK,UAAUA,EAAE,MAAM,UAAU,KAAK,aAAa,IAAI,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,KAAK,WAAWA,EAAE,KAAK,cAAc,OAAO,KAAK,aAAa,UAAUA,GAAG,KAAK,eAAe,OAAO,KAAK,cAAc,UAAUA,EAAE,CAAC,YAAY,CAAC,OAAO,KAAK,aAAa,WAAW,EAAE,OAAO,KAAK,cAAc,WAAW,CAAC,CAAC,CAAC,WAAWA,EAAE,CAAC,IAAIZ,EAAEY,EAAE,OAAO,EAAE,KAAK,MAAMZ,EAAE,CAAC,EAAE,KAAK,aAAa,WAAWY,EAAE,MAAM,EAAE,CAAC,CAAC,EAAE,KAAK,cAAc,WAAWA,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAE,CAAC,IAAIZ,EAAE,KAAK,aAAa,mBAAmBY,CAAC,EAAE,MAAM,QAAQZ,CAAC,GAAG,MAAM,QAAQA,EAAE,EAAE,IAAIA,EAAE,CAACA,CAAC,GAAGA,EAAEA,EAAE,IAAI,EAAEK,EAAEwB,EAAE,OAAO,KAAK,cAAcA,EAAE7B,EAAE,MAAM,CAAC,GAAG,EAAEA,EAAE,GAAG,EAAE,EAAE,KAAK,YAAY,UAAU,EAAE,EAAE,OAAO,IAAI,EAAEK,EAAE,CAAC,CAAC,GAAG,KAAK,WAAW,KAAKA,EAAE,CAAC,EAAE,EAAE,MAAM,CAAC,EAAEA,EAAE,CAAC,CAAC,EAAE,KAAK,YAAY,KAAK,WAAW,KAAKA,EAAE,OAAOwB,CAAC,EAAE,OAAOA,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,EAAE,OAAOA,CAAC,EAAE,OAAOA,EAAE,MAAM,CAAC,EAAEwkD,GAAGhmD,CAAC,CAAC,CAAC,MAAMO,EAAEZ,EAAE,CAAC,IAAI,EAAEA,GAAG,KAAK,KAAKA,EAAE,aAAaK,EAAEL,GAAG,KAAK,KAAKA,EAAE,UAAUA,GAAG,OAAOA,EAAE,CAAC,GAAG,IAAI6B,EAAEo9D,GAAGr+D,EAAE,EAAEP,EAAE,KAAK,YAAY,EAAE,GAAGO,EAAEiB,EAAE,OAAO,EAAEA,EAAE,aAAaxB,EAAEwB,EAAE,UAAU,MAAM,QAAQjB,CAAC,IAAI,EAAEA,EAAE,MAAM,CAAC,EAAEA,EAAEA,EAAE,KAAK,GAAG,MAAM,EAAE,SAAS,IAAIP,GAAG,KAAK,OAAO,MAAM,MAAMO,EAAEZ,CAAC,EAAE,IAAI8B,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC,IAAI,EAAE,EAAE,OAAO,GAAG,EAAE,EAAE,EAAE,MAAM,IAAI+jD,EAAE,+HAA+H,EAAE9lD,EAAE,aAAa,EAAE8B,EAAE,KAAK,GAAG,CAAC,EAAE,IAAIY,EAAE,EAAE,IAAI,GAAG,IAAIoqD,GAAG,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC,EAAE,KAAK,aAAa,UAAUpqD,EAAE,MAAM,EAAE,EAAE,CAAC,EAAE,KAAK,cAAc,UAAUA,EAAE,MAAM,EAAE,CAAC,EAAEX,EAAE,KAAK,GAAGW,CAAC,CAAC,CAAC,GAAGrC,GAAG,KAAK,MAAM,IAAI0lD,GAAG,uEAAuE,EAAE,IAAI/jD,EAAEF,EAAE,aAAairD,GAAG,QAAQ,KAAKjrD,EAAE,GAAG,aAAairD,KAAK/qD,EAAE,MAAM,IAAI8jD,EAAE,8GAA8G,EAAE,GAAG9jD,EAAE,CAAC,IAAI,EAAE,CAACpB,CAAC,EAAE,OAAOkB,CAAC,EAAEY,EAAE,KAAK,UAAU,OAAOX,CAAC,EAAE,EAAE,KAAK,UAAU,KAAK,UAAUW,EAAE,IAAIE,EAAE,MAAM,MAAM,EAAE5C,CAAC,EAAE,OAAO,KAAK,UAAU,EAAE4C,CAAC,KAAM,QAAO,MAAM,MAAMhC,EAAEZ,CAAC,CAAC,CAAC,KAAKY,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAEvzB,EAAE,aAAaK,EAAEwB,EAAE,GAAG,GAAG,KAAKxB,EAAE,KAAK,aAAa,KAAKO,EAAEZ,CAAC,EAAE6B,EAAE,KAAK,cAAc,KAAKjB,EAAEZ,CAAC,MAAM,CAAC,IAAIgC,EAAE,EAAE,MAAM,EAAE,EAAE,OAAO,CAAC,EAAE,EAAE,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE3B,EAAE,KAAK,aAAa,KAAKO,EAAE,OAAO,OAAOZ,EAAE,CAAC,aAAagC,CAAC,CAAC,CAAC,EAAEH,EAAE,KAAK,cAAc,KAAKjB,EAAE,OAAO,OAAOZ,EAAE,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI8B,EAAE,KAAK,cAAc,MAAM,QAAQzB,CAAC,IAAIyB,EAAEzB,EAAE,MAAM,CAAC,EAAE,OAAOwB,EAAE,MAAM,CAAC,CAAC,GAAGxB,EAAEA,EAAE,GAAGwB,EAAEA,EAAE,IAAI,KAAK,kBAAkBA,EAAE2qC,GAAG3qC,EAAE,CAAC,GAAG,IAAIE,EAAE,OAAO,KAAK,YAAY,SAASA,EAAE+nD,GAAG,CAACzpD,EAAEwB,CAAC,CAAC,EAAE,KAAK,YAAY,MAAME,EAAE62B,EAAEv4B,EAAEwB,CAAC,EAAE,KAAK,YAAY,MAAME,EAAEm3B,EAAE,GAAGN,EAAEv4B,EAAEwB,CAAC,CAAC,EAAE,KAAK,YAAY,MAAME,EAAEm3B,EAAE74B,EAAEwB,CAAC,EAAE,KAAK,WAAW,OAAOE,EAAE,CAAC1B,EAAEwB,CAAC,GAAG,KAAK,YAAY,KAAK,WAAW,KAAKE,EAAE,OAAOD,CAAC,EAAE,CAACC,CAAC,EAAE,OAAOD,CAAC,EAAEC,CAAC,CAAC,CAAC,CAAC,YAAYnB,EAAE,CAAC,KAAK,aAAa,YAAY,EAAE,KAAK,cAAc,YAAY,CAAC,CAAC,MAAMA,EAAE,CAAC2nD,GAAG,KAAK,aAAa,KAAK,IAAI,CAAC,KAAK,aAAa,MAAM3nD,CAAC,CAAC,CAAC,EAAE2nD,GAAG,KAAK,cAAc,KAAK,IAAI,CAAC,KAAK,cAAc,MAAM3nD,CAAC,CAAC,CAAC,EAAE,KAAK,MAAM,EAAE,CAAC,YAAYA,EAAEZ,EAAE,CAAC,MAAM,QAAQA,CAAC,IAAIA,EAAEA,EAAE,IAAI,IAAI,EAAE,GAAG,KAAK,gBAAgB,KAAK,WAAW,KAAK,EAAE,CAACA,EAAEA,CAAC,EAAE,EAAEA,EAAE,KAAK,WAAW,KAAK,EAAE,CAAC,KAAK,IAAI,EAAE,EAAE,KAAK,KAAK,YAAY,CAAC,IAAI6B,EAAE,KAAK,aAAa,OAAO,IAAIC,GAAG,IAAI,EAAE,OAAO,MAAM,QAAQ,CAAC,EAAE,EAAE,OAAOD,CAAC,EAAE,OAAOA,CAAC,EAAE,CAAC,CAAC,EAAE,OAAOA,CAAC,EAAE,OAAOA,CAAC,CAAC,KAAM,QAAO,CAAC,CAAC,IAAI,kBAAkB,CAAC,OAAO,KAAK,aAAa,iBAAiB,OAAO,KAAK,cAAc,gBAAgB,CAAC,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,aAAa,oBAAoB,OAAO,KAAK,cAAc,mBAAmB,CAAC,CAAC,6BAA6BjB,EAAE,CAAC,MAAM,6BAA6BA,CAAC,EAAE,KAAK,cAAc,MAAM,KAAK,aAAa,6BAA6BA,CAAC,EAAE,KAAK,eAAe,MAAM,KAAK,cAAc,6BAA6BA,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,UAAU,KAAK,SAAS,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,OAAO,WAAWA,EAAEZ,EAAE,CAAC,IAAI,EAAE61D,GAAG71D,EAAE,KAAK,EAAE,GAAG,OAAOA,EAAE,MAAMA,EAAE,cAAc,KAAK,MAAM,IAAI+lD,GAAG,0FAA0F,EAAE,IAAI1lD,EAAEL,EAAE,OAAOK,EAAE,MAAM,EAAE,IAAIO,EAAEP,CAAC,CAAC,CAAC,EAAE8iE,GAAG,UAAU,gBAAgB5rC,GAAE,cAAc4rC,EAAE,EAAE,IAAIC,GAAG,cAAcjW,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,MAAMA,EAAE,MAAMA,EAAE,OAAO,KAAK,OAAOA,EAAE,OAAO,KAAK,OAAO,CAAC,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,MAAM,KAAK,MAAM,OAAO,KAAK,MAAM,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,KAAK3yB,EAAE0rD,GAAG1rD,CAAC,EAAEA,EAAE,QAAQ,YAAYA,EAAEyoD,GAAGzoD,EAAE,SAAS,GAAGg4B,EAAEM,EAAEt4B,EAAE,KAAK,KAAK,EAAE,KAAK,MAAM,EAAE,CAAC,CAAC,EAAEwiE,GAAG,UAAU,YAAY7rC,GAAE,cAAc6rC,EAAE,EAAE,IAAIC,GAAG,CAAC,WAAW,SAAS,EAAEC,GAAG,IAAI,IAAID,EAAE,EAAEE,GAAG,cAAcpW,EAAE,CAAC,YAAYvsD,EAAE,CAAC,GAAG,MAAMA,CAAC,EAAE,KAAK,OAAOA,EAAE,OAAO,KAAK,MAAMA,EAAE,MAAMA,EAAE,cAAc,GAAG0iE,GAAG,IAAI1iE,EAAE,aAAa,EAAE,KAAK,cAAcA,EAAE,kBAAmB,OAAM,IAAIklD,EAAE,oCAAoCllD,EAAE,kCAAkC,OAAO,KAAK,cAAc,WAAW,KAAK,kBAAkB,QAAQA,EAAE,iBAAiB,CAAC,CAAC,mBAAmBA,EAAE,CAACA,EAAE2rD,GAAG3rD,CAAC,EAAE,IAAIZ,EAAEY,EAAE,GAAG,MAAM,CAAC,KAAK,OAAO,KAAK,MAAMZ,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIY,EAAE,CAAC,OAAO,KAAK,OAAO,MAAM,KAAK,MAAM,cAAc,KAAK,cAAc,kBAAkB,KAAK,iBAAiB,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC,IAAI,EAAE,CAAC,KAAK,OAAO,KAAK,KAAK,EAAE,GAAG,KAAK,gBAAgB,WAAW,OAAO8lB,GAAG,eAAez4C,EAAE,EAAE,CAAC,KAAK,iBAAiB,EAAE,GAAG,KAAK,gBAAgB,UAAU,OAAOy4C,GAAG,sBAAsBz4C,EAAE,EAAE,CAAC,KAAK,iBAAiB,EAAE,MAAM,IAAI,MAAM,oBAAoB,KAAK,0BAA0B,CAAC,GAAG0iE,EAAE,iBAAiB,CAAC,CAAC,CAAC,CAAC,EAAEC,GAAG,UAAU,WAAWhsC,GAAE,cAAcgsC,EAAE,EAAE,SAASC,GAAG1jE,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEisD,GAAGxsD,CAAC,EAAE,GAAGO,EAAE,QAAQ,UAAUA,EAAEgpD,GAAGhpD,EAAE,OAAO,GAAG,IAAI,MAAM,OAAOA,EAAE,IAAI,EAAEA,EAAE,MAAM,GAAGA,EAAE,OAAO,IAAIA,EAAEqjC,GAAGrjC,EAAE,EAAE,GAAG,IAAI,UAAUA,EAAE,MAAMA,EAAE,MAAM,OAAO,KAAK,IAAIA,EAAEqjC,GAAGrjC,EAAE,EAAE,GAAGA,EAAE,KAAK,EAAE,MAAM,IAAIylD,EAAE,4EAA4E,qBAAqB,uCAAuCzlD,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,WAAW,QAAQ,EAAE,SAAS,CAAC,EAAE0B,EAAE1B,EAAE2B,EAAE,GAAG,OAAO5B,GAAG,aAAa,IAAI,QAAQ4B,EAAEo+B,GAAGr+B,EAAE3B,EAAEJ,EAAE,CAAC,EAAEgC,EAAEo+B,GAAGr+B,EAAE,CAAC,EAAE/B,EAAE,CAAC,EAAE,IAAI,QAAQ,OAAOgC,EAAE,GAAG5B,EAAE,OAAO84B,EAAEl3B,EAAE5B,CAAC,EAAE,MAAM,IAAI0lD,EAAE,uDAAuD,CAAC,CAAC,IAAI2d,GAAG,cAActW,EAAE,CAAC,YAAYvsD,EAAE,CAAC,MAAMA,CAAC,EAAE,KAAK,UAAUA,EAAE,UAAUA,EAAE,WAAW,KAAK,WAAWA,EAAE,WAAW,KAAK,WAAW,UAAU,CAAC,WAAW,CAAC,IAAIA,EAAE,CAAC,UAAU,KAAK,UAAU,WAAW,KAAK,UAAU,EAAEZ,EAAE,MAAM,UAAU,EAAE,OAAO,OAAO,OAAOY,EAAEZ,CAAC,EAAEY,CAAC,CAAC,mBAAmBA,EAAE,CAAC,OAAOA,EAAE2rD,GAAG3rD,CAAC,EAAEA,GAAG,KAAK,CAAC,KAAK,SAAS,EAAE,KAAK,aAAa,UAAUA,EAAEA,EAAE,OAAO,KAAK,GAAGA,EAAE,KAAK,KAAK,SAAS,EAAEA,IAAIA,EAAEA,EAAE,OAAO,GAAG,KAAK,UAAUA,EAAE,CAAC,KAAKA,EAAEZ,EAAE,CAAC,OAAOuzB,EAAE,IAAI,CAAC3yB,EAAE0rD,GAAG1rD,CAAC,EAAEA,EAAE,QAAQ,UAAUA,EAAEyoD,GAAGzoD,EAAE,OAAO,GAAG,IAAI,EAAE,GAAG,OAAOZ,EAAE,cAAc,YAAY,CAAC,GAAG,KAAK,aAAa,QAAQ,MAAM,IAAI8lD,EAAE;AAAA,sCACn3jM9lD,EAAE,cAAc,EAAE,EAAEssD,GAAGtsD,EAAE,YAAY,CAAC,CAAC,IAAIK,EAAEiiC,GAAG1hC,CAAC,EAAEiB,EAAE2gC,GAAG5hC,CAAC,EAAEkB,EAAEwiC,GAAG,KAAK,UAAUjkC,CAAC,EAAE,WAAW,EAAE,IAAI,CAAC,EAAE0B,EAAEyiC,GAAG3iC,EAAE,CAAC,EAAE,WAAW,EAAE,IAAI,CAAC,EAAE,GAAG,EAAEC,GAAGC,GAAG,MAAM,IAAI+jD,EAAE,uEAAuE,KAAK,WAAW,EAAE,OAAO0d,GAAG5iE,EAAE,KAAK,WAAW,KAAK,UAAU,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE6iE,GAAG,UAAU,mBAAmBlsC,GAAE,cAAcksC,EAAE,EAAE,SAASxQ,GAAGnzD,EAAE,CAAC,OAAO,IAAIytD,GAAGztD,CAAC,CAAC,CAAC,SAASqyD,GAAGryD,EAAE,CAAC,OAAO,IAAI29D,GAAG39D,CAAC,CAAC,CAAC,SAASm0D,GAAGn0D,EAAE,CAAC,OAAO,IAAIw9D,GAAGx9D,CAAC,CAAC,CAAC,SAASqzD,GAAGrzD,EAAE,CAAC,OAAO,IAAIy9D,GAAGz9D,CAAC,CAAC,CAAC,SAASk0D,GAAGl0D,EAAE,CAAC,OAAO,IAAI09D,GAAG19D,CAAC,CAAC,CAAC,SAAS40D,GAAG50D,EAAE,CAAC,OAAO,IAAI69D,GAAG79D,CAAC,CAAC,CAAC,SAAS+0D,GAAG/0D,EAAE,CAAC,OAAO,IAAI49D,GAAG59D,CAAC,CAAC,CAAC,SAASyxD,GAAGzxD,EAAE,CAAC,OAAO,IAAI8+D,GAAG9+D,CAAC,CAAC,CAAC,SAAS0xD,GAAG1xD,EAAE,CAAC,OAAO,IAAIw+D,GAAGx+D,CAAC,CAAC,CAAC,SAAS2xD,GAAG3xD,EAAE,CAAC,OAAO,IAAI0+D,GAAG1+D,CAAC,CAAC,CAAC,SAAS4xD,GAAG5xD,EAAE,CAAC,OAAO,IAAIy+D,GAAGz+D,CAAC,CAAC,CAAC,SAAS6xD,GAAG7xD,EAAE,CAAC,OAAO,IAAI2+D,GAAG3+D,CAAC,CAAC,CAAC,SAASy0D,GAAGz0D,EAAE,CAAC,OAAO,IAAI6+D,GAAG7+D,CAAC,CAAC,CAAC,SAASgyD,GAAGhyD,EAAE,CAAC,OAAO,IAAI++D,GAAG/+D,CAAC,CAAC,CAAC,SAASi1D,GAAGj1D,EAAE,CAAC,OAAO,IAAIg/D,GAAGh/D,CAAC,CAAC,CAAC,SAASkyD,GAAGlyD,EAAE,CAAC,OAAO,IAAIk/D,GAAGl/D,CAAC,CAAC,CAAC,SAASwwD,GAAGxwD,EAAE,CAAC,OAAO,IAAIqgE,GAAGrgE,CAAC,CAAC,CAAC,SAASiyD,GAAGjyD,EAAE,CAAC,OAAO,IAAImgE,GAAGngE,CAAC,CAAC,CAAC,SAASoyD,GAAGpyD,EAAE,CAAC,OAAO,IAAIigE,GAAGjgE,CAAC,CAAC,CAAC,SAAS60D,GAAG70D,EAAE,CAAC,OAAO,IAAIkgE,GAAGlgE,CAAC,CAAC,CAAC,SAASuyD,GAAGvyD,EAAE,CAAC,OAAO,IAAIogE,GAAGpgE,CAAC,CAAC,CAAC,SAASo0D,GAAGp0D,EAAE,CAAC,OAAO,IAAIsgE,GAAGtgE,CAAC,CAAC,CAAC,SAASs0D,GAAGt0D,EAAE,CAAC,OAAO,IAAIugE,GAAGvgE,CAAC,CAAC,CAAC,SAASi0D,GAAGj0D,EAAE,CAAC,OAAO,IAAIwgE,GAAGxgE,CAAC,CAAC,CAAC,SAASsyD,GAAGtyD,EAAE,CAAC,OAAO,IAAI0gE,GAAG1gE,CAAC,CAAC,CAAC,SAASywD,GAAGzwD,EAAE,CAAC,OAAO,IAAI4gE,GAAG5gE,CAAC,CAAC,CAAC,SAAS2wD,GAAG3wD,EAAE,CAAC,OAAO,IAAI8gE,GAAG9gE,CAAC,CAAC,CAAC,SAASwxD,GAAGxxD,EAAE,CAAC,OAAO,IAAIihE,GAAGjhE,CAAC,CAAC,CAAC,SAAS8zD,GAAG9zD,EAAE,CAAC,OAAO,IAAI+gE,GAAG/gE,CAAC,CAAC,CAAC,SAAS+zD,GAAG/zD,EAAE,CAAC,OAAO,IAAIghE,GAAGhhE,CAAC,CAAC,CAAC,SAASg0D,GAAGh0D,EAAE,CAAC,OAAO,IAAI6gE,GAAG7gE,CAAC,CAAC,CAAC,SAASmyD,GAAGnyD,EAAE,CAAC,OAAO,IAAIohE,GAAGphE,CAAC,CAAC,CAAC,SAASqxD,GAAGrxD,EAAE,CAAC,OAAO,IAAI4hE,GAAG5hE,CAAC,CAAC,CAAC,SAASozD,GAAGpzD,EAAE,CAAC,OAAO,IAAI6hE,GAAG7hE,CAAC,CAAC,CAAC,SAASk1D,GAAGl1D,EAAE,CAAC,OAAO,IAAI+hE,GAAG/hE,CAAC,CAAC,CAAC,SAAS4wD,GAAG5wD,EAAE,CAAC,OAAO,IAAIoiE,GAAGpiE,CAAC,CAAC,CAAC,SAAS+wD,GAAG/wD,EAAE,CAAC,OAAO4wD,GAAG5wD,CAAC,CAAC,CAAC,SAASkxD,GAAGlxD,EAAE,CAAC,OAAO4wD,GAAG5wD,CAAC,CAAC,CAAC,SAAS6wD,GAAG7wD,EAAE,CAAC,OAAO,IAAIuiE,GAAGviE,CAAC,CAAC,CAAC,SAASgxD,GAAGhxD,EAAE,CAAC,OAAO6wD,GAAG7wD,CAAC,CAAC,CAAC,SAASmxD,GAAGnxD,EAAE,CAAC,OAAO6wD,GAAG7wD,CAAC,CAAC,CAAC,SAAS8wD,GAAG9wD,EAAE,CAAC,OAAO,IAAI0iE,GAAG1iE,CAAC,CAAC,CAAC,SAASixD,GAAGjxD,EAAE,CAAC,OAAO8wD,GAAG9wD,CAAC,CAAC,CAAC,SAASoxD,GAAGpxD,EAAE,CAAC,OAAO8wD,GAAG9wD,CAAC,CAAC,CAAC,SAAS0yD,GAAG1yD,EAAE,CAAC,OAAO,IAAI4iE,GAAG5iE,CAAC,CAAC,CAAC,SAAS2yD,GAAG3yD,EAAE,CAAC,OAAO,IAAI+iE,GAAG/iE,CAAC,CAAC,CAAC,SAAS8yD,GAAG9yD,EAAE,CAAC,OAAO,IAAI6iE,GAAG7iE,CAAC,CAAC,CAAC,SAAS+yD,GAAG/yD,EAAE,CAAC,OAAO,IAAIgjE,GAAGhjE,CAAC,CAAC,CAAC,SAAS2zD,GAAG3zD,EAAE,CAAC,OAAO,IAAImiE,GAAGniE,CAAC,CAAC,CAAC,SAAS4zD,GAAG5zD,EAAE,CAAC,OAAO,IAAIsiE,GAAGtiE,CAAC,CAAC,CAAC,SAAS6zD,GAAG7zD,EAAE,CAAC,OAAO,IAAIyiE,GAAGziE,CAAC,CAAC,CAAC,SAASgzD,GAAGhzD,EAAE,CAAC,OAAO,IAAI0/D,GAAG1/D,CAAC,CAAC,CAAC,SAASizD,GAAGjzD,EAAE,CAAC,OAAO,IAAIy/D,GAAGz/D,CAAC,CAAC,CAAC,SAASszD,GAAGtzD,EAAE,CAAC,OAAO,IAAI4/D,GAAG5/D,CAAC,CAAC,CAAC,SAASuzD,GAAGvzD,EAAE,CAAC,OAAO,IAAI2/D,GAAG3/D,CAAC,CAAC,CAAC,SAAS00D,GAAG10D,EAAE,CAAC,OAAO,IAAIw/D,GAAGx/D,CAAC,CAAC,CAAC,SAAS20D,GAAG30D,EAAE,CAAC,OAAO,IAAIs/D,GAAGt/D,CAAC,CAAC,CAAC,SAAS8xD,GAAG9xD,EAAE,CAAC,OAAO,IAAIggE,GAAGhgE,CAAC,CAAC,CAAC,SAAS+xD,GAAG/xD,EAAE,CAAC,OAAO,IAAI+/D,GAAG//D,CAAC,CAAC,CAAC,SAASw0D,GAAGx0D,EAAE,CAAC,OAAO,IAAIswD,GAAGtwD,CAAC,CAAC,CAAC,SAAS80D,GAAG90D,EAAE,CAAC,OAAO,IAAIq/D,GAAGr/D,CAAC,CAAC,CAAC,SAASsxD,GAAGtxD,EAAE,CAAC,OAAO,IAAIqjE,GAAGrjE,CAAC,CAAC,CAAC,SAASg1D,GAAGh1D,EAAE,CAAC,OAAO,IAAIkjE,GAAGljE,CAAC,CAAC,CAAC,IAAI4yD,GAAGE,GAAGD,GAAGE,GAAGU,GAAGE,GAAGD,GAAGE,GAAG,SAASnB,GAAGzyD,EAAE,CAAC,OAAO,IAAIqhE,GAAGrhE,CAAC,CAAC,CAAC,SAASwyD,GAAGxyD,EAAE,CAAC,OAAO,IAAIshE,GAAGthE,CAAC,CAAC,CAAC,SAAS0wD,GAAG1wD,EAAE,CAAC,OAAO,IAAIuhE,GAAGvhE,CAAC,CAAC,CAAC,SAASwzD,GAAGxzD,EAAE,CAAC,OAAO,IAAIygE,GAAGzgE,CAAC,CAAC,CAAC,SAASq0D,GAAGr0D,EAAE,CAAC,OAAO,IAAIsjE,GAAGtjE,CAAC,CAAC,CAAC,SAASu0D,GAAGv0D,EAAE,CAAC,OAAO,IAAIyjE,GAAGzjE,CAAC,CAAC,CAAC,SAASuxD,GAAGvxD,EAAE,CAAC,OAAO,IAAI2jE,GAAG3jE,CAAC,CAAC,CAAC,IAAI4jE,GAAG,CAAC,EAAExjE,GAAGwjE,GAAG,CAAC,KAAK,IAAIC,GAAG,IAAI,IAAIC,GAAG,eAAe,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,wBAAwB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,KAAK,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,4BAA4B,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,IAAI,IAAIC,GAAG,UAAU,IAAIC,GAAG,OAAO,IAAIC,GAAG,0BAA0B,IAAIC,EAAE,CAAC,EAAE,SAASZ,GAAG/jE,EAAE,EAAE,CAAC,OAAOk3D,GAAGl3D,EAAE,CAAC,CAAC,CAAC,SAASgkE,GAAGhkE,EAAE,EAAE,CAAC,OAAOy3D,GAAGz3D,EAAE,CAAC,CAAC,CAAC,SAAS2kE,GAAG3kE,EAAE,EAAE,CAAC,OAAO03D,GAAG13D,EAAE,CAAC,CAAC,CAAC,SAASikE,GAAGjkE,EAAE,EAAE,CAAC,OAAOm3D,GAAGn3D,EAAE,CAAC,CAAC,CAAC,SAASkkE,GAAGlkE,EAAE,EAAE,CAAC,OAAOi4D,GAAGj4D,EAAE,CAAC,CAAC,CAAC,SAASykE,GAAGzkE,EAAE,EAAE,CAAC,OAAOu3D,GAAGv3D,EAAE,CAAC,CAAC,CAAC,SAAS0kE,GAAG1kE,EAAE,EAAE,CAAC,OAAOw3D,GAAGx3D,EAAE,CAAC,CAAC,CAAC,SAASmkE,GAAGnkE,EAAE,EAAE,CAAC,OAAO+2D,GAAG/2D,EAAE,CAAC,CAAC,CAAC,SAASqkE,GAAGrkE,EAAE,EAAE,CAAC,OAAOk2D,GAAGl2D,EAAE,CAAC,CAAC,CAAC,SAASskE,GAAGtkE,EAAE,EAAE,CAAC,OAAOm2D,GAAGn2D,EAAE,CAAC,CAAC,CAAC,SAAS6jE,GAAG7jE,EAAE,EAAE,CAAC,OAAOm2D,GAAGn2D,EAAE,CAAC,CAAC,CAAC,SAASokE,GAAGpkE,EAAE,EAAE,CAAC,OAAOm2D,GAAGn2D,EAAE,CAAC,CAAC,CAAC,SAASukE,GAAGvkE,EAAE,EAAE,CAAC,OAAOi2D,GAAGj2D,EAAE,CAAC,CAAC,CAAC,SAAS8jE,GAAG9jE,EAAE,EAAE,CAAC,OAAOi2D,GAAGj2D,EAAE,CAAC,CAAC,CAAC,SAASwkE,GAAGxkE,EAAE,EAAE,CAAC,OAAOi2D,GAAGj2D,EAAE,CAAC,CAAC,CAAC,IAAI4kE,GAAG,CAAC,EAAExkE,GAAGwkE,GAAG,CAAC,cAAc,IAAIvJ,EAAE,CAAC,EAAE,IAAIwJ,GAAG,CAAC,EAAEzkE,GAAGykE,GAAG,CAAC,GAAG,IAAIC,GAAG,KAAK,IAAIC,GAAG,GAAG,IAAIC,EAAE,CAAC,EAAE,SAASD,GAAG/kE,EAAE,CAAC,OAAO,IAAIi9D,GAAGj9D,CAAC,CAAC,CAAC,SAAS8kE,GAAG9kE,EAAE,CAAC,OAAOk9D,GAAGl9D,CAAC,CAAC,CAAC,SAASglE,GAAGhlE,EAAE,CAAC,OAAOm9D,GAAGn9D,CAAC,CAAC,CAAC,IAAIilE,GAAG,cAAc1P,EAAE,CAAC,aAAa,CAAC,MAAM,GAAG,SAAS,EAAE,KAAK,MAAM,IAAI,CAAC,SAASz0D,EAAE,CAAC,GAAG,EAAEA,aAAaq6D,IAAI,MAAM,IAAI,MAAM,uDAAuD,EAAE,KAAK,MAAMr6D,CAAC,CAAC,EAAE,SAASokE,GAAGllE,EAAE,EAAE,CAAC,OAAOA,EAAE,CAAC,CAAC,SAASmlE,GAAGnlE,EAAE,EAAE,CAAC,OAAOA,EAAE,CAAC,CAAC,IAAIolE,GAAG,cAAcH,EAAE,CAAC,YAAYnkE,EAAE,CAAC,GAAG,MAAM,EAAEA,GAAG,OAAOA,EAAE,CAAC,GAAGA,EAAE,mBAAmB,MAAM,IAAImlD,GAAG,oEAAoE,EAAE,KAAK,QAAQnlD,EAAE,SAAS,WAAW,KAAK,SAAS,KAAK,IAAIA,EAAE,UAAU,CAAC,EAAE,KAAK,SAASA,EAAE,UAAU,EAAE,KAAK,QAAQA,EAAE,SAAS,EAAE,KAAK,KAAKA,EAAE,MAAM,OAAO,KAAK,SAASA,EAAE,SAAS,CAAC,OAAO,MAAM,KAAK,EAAE,QAAQ,KAAK,IAAI,IAAI,KAAK,QAAQ,KAAK,uBAAuB,KAAK,gDAAgD,EAAE,KAAK,KAAK,QAAQ,KAAK,OAAO,MAAM,KAAK,YAAYokE,GAAG,KAAK,OAAO,MAAM,KAAK,YAAYC,GAAG,KAAK,QAAQ,QAAQ,KAAK,IAAI,GAAG,KAAK,YAAYA,GAAG,KAAK,YAAYD,GAAG,KAAK,cAAcA,KAAK,KAAK,UAAU,GAAG,CAAC,MAAM,aAAapkE,EAAE,CAAC,KAAK,KAAK,EAAE,KAAK,aAAa,EAAE,KAAK,UAAU,KAAK,KAAK,KAAK,KAAK,SAAS,KAAK,KAAK,KAAK,cAAcokE,GAAG,EAAE,EAAE,GAAG,CAAC,CAAC,MAAM,WAAWpkE,EAAEZ,EAAE,CAAC,MAAMi1D,GAAGj1D,CAAC,EAAE,IAAI,EAAE,KAAK,gBAAgBA,CAAC,EAAE,GAAG,OAAO,KAAK,YAAY,EAAE,KAAK,SAAS,KAAK,IAAI,GAAG,KAAK,KAAK,EAAE,KAAK,KAAK,IAAI,KAAK,OAAO,KAAK,MAAM,KAAK,WAAW,KAAK,aAAaY,EAAE,KAAK,MAAM,aAAa,KAAK,CAAC,MAAM,WAAWA,EAAE,CAAC,KAAK,aAAa,GAAG,KAAK,SAAS,QAAQ,IAAI,SAAS,KAAK,+BAA+B,CAAC,CAAC,gBAAgBA,EAAE,CAACA,GAAG,OAAOA,EAAE,CAAC,GAAG,IAAIZ,EAAEY,EAAE,KAAK,SAAS,OAAOZ,GAAG,MAAM,QAAQ,KAAK,4BAA4B,KAAK,oDAAoD,OAAO,KAAKY,CAAC,GAAG,EAAEZ,CAAC,CAAC,EAAE,SAASmlE,GAAGrlE,EAAE,CAAC,OAAO,IAAIolE,GAAGplE,CAAC,CAAC,CAAC,IAAIslE,GAAG,CAAC,cAAcD,EAAE,EAAME,GAAG9qD,EAAE,EAAE8qD,GAAG,aAAa,4BAA4B,IAAI,GAAGvlE,GAAG,CAACA,GAAG,QAAQ,KAAK,+OAA+O,CAAC,CAAC,EAAE,IAAIwlE,IAAI,SAASxlE,EAAE,CAACA,EAAEA,EAAE,WAAW,GAAG,aAAaA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,UAAU,GAAG,YAAYA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,QAAQ,GAAG,UAAUA,EAAEA,EAAE,UAAU,GAAG,YAAYA,EAAEA,EAAE,aAAa,GAAG,eAAeA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,QAAQ,IAAI,UAAUA,EAAEA,EAAE,SAAS,IAAI,WAAWA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,YAAY,IAAI,cAAcA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,WAAW,IAAI,aAAaA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,cAAc,IAAI,gBAAgBA,EAAEA,EAAE,QAAQ,IAAI,UAAUA,EAAEA,EAAE,YAAY,IAAI,cAAcA,EAAEA,EAAE,WAAW,IAAI,aAAaA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,UAAU,IAAI,YAAYA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,YAAY,KAAK,cAAcA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,iBAAiB,KAAK,mBAAmBA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,YAAY,KAAK,cAAcA,EAAEA,EAAE,aAAa,KAAK,eAAeA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,gBAAgB,KAAK,kBAAkBA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,eAAe,KAAK,iBAAiBA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,kBAAkB,KAAK,oBAAoBA,EAAEA,EAAE,YAAY,KAAK,cAAcA,EAAEA,EAAE,gBAAgB,KAAK,kBAAkBA,EAAEA,EAAE,eAAe,KAAK,iBAAiBA,EAAEA,EAAE,cAAc,KAAK,gBAAgBA,EAAEA,EAAE,cAAc,KAAK,eAAe,GAAGwlE,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASzlE,EAAE,CAAC,IAAI,GAAG,SAASE,EAAE,CAACA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,GAAG,GAAG,KAAKA,EAAEA,EAAE,GAAG,GAAG,IAAI,GAAG,EAAEF,EAAE,0BAA0BA,EAAE,wBAAwB,CAAC,EAAE,CAAC,GAAGylE,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,CAAC,EAAE,SAASC,GAAG3lE,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,SAASF,EAAE,SAAS,SAAS,OAAO,CAAC,EAAE,MAAM,CAAC,EAAE,eAAe,CAAC,EAAE0lE,GAAG1lE,GAAGE,CAAC,CAAC,SAAS0lE,GAAG5lE,EAAE,CAAC,OAAO0lE,GAAG1lE,EAAE,CAAC,SAAS6lE,GAAG7lE,EAAE,CAAC,OAAO0lE,GAAG1lE,EAAE,CAAC,SAAS8lE,EAAE9lE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,EAAE,YAAYP,GAAG,GAAG,GAAG,EAAE,kBAAkB,OAAO,CAAC,IAAIiC,EAAE,EAAE,gBAAgBC,EAAE,EAAE,gBAAgB,EAAE,OAAO,EAAE,gBAAgB,OAAOD,EAAE,EAAE,EAAE,cAAc,GAAG,EAAE,OAAO,SAAS,OAAO8jE,GAAG,EAAE,WAAW,EAAE,iBAAiB7lE,EAAEI,EAAEC,CAAC,EAAE,GAAG,EAAE,OAAO,UAAU,OAAO,EAAE,WAAW,MAAM0B,EAAEC,CAAC,EAAE,IAAIY,GAAGijE,GAAGjjE,EAAE5C,EAAEI,EAAEC,CAAC,CAAC,EAAE,IAAI4B,EAAE4jE,GAAG,EAAE,WAAW,MAAM9jE,CAAC,EAAE,GAAG/B,EAAEI,EAAEC,CAAC,EAAEqC,EAAET,EAAE,SAAS,EAAE,OAAO,EAAE,OAAO,SAASS,EAAE,GAAGkkB,EAAE,cAAc3kB,EAAE,MAAMS,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,WAAW5C,GAAG,OAAO,GAAG,EAAE,KAAK,CAAC,SAAS+lE,GAAG/lE,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAEylE,GAAGhmE,CAAC,EAAE,GAAGM,GAAG,KAAK,CAAC,IAAI2B,EAAE3B,EAAE,yBAAyBC,CAAC,EAAE,GAAG0B,GAAG,KAAK,OAAOA,CAAC,CAAC,IAAI,EAAE/B,EAAE,kBAAkB,KAAK+B,GAAG,CAAC,CAAC,EAAEgkE,GAAG1lE,EAAE0B,CAAC,EAAE,EAAE,OAAO,IAAI,OAAO,EAAEgkE,GAAG1lE,EAAE,CAAC,GAAG,GAAG,MAAM,CAAC,SAAS2lE,GAAGlmE,EAAE,EAAEE,EAAE,CAAC,OAAO,EAAE+lE,GAAGjmE,EAAEE,EAAE,gBAAgB,EAAE,CAAC,SAASimE,GAAGnmE,EAAE,EAAE,CAAC,GAAG,CAACE,EAAEI,EAAEC,CAAC,EAAEylE,GAAGhmE,CAAC,EAAE,MAAM,CAACimE,GAAG/lE,EAAE,GAAG,EAAE,gBAAgB,EAAEI,EAAEC,CAAC,CAAC,CAAC,SAAS0lE,GAAGjmE,EAAE,EAAE,CAAC,OAAO,EAAE,GAAGA,KAAK,IAAIA,CAAC,CAAC,SAASgmE,GAAGhmE,EAAE,CAAC,IAAI,EAAEA,EAAE,MAAM,GAAG,EAAE,GAAG,EAAE,SAAS,EAAE,MAAM,CAACA,EAAE,EAAE,MAAM,EAAE,IAAIE,EAAE,EAAE,GAAGI,EAAE,EAAE,SAAS,EAAE,EAAE,GAAG,OAAOC,EAAE,OAAO,EAAE,EAAE,OAAO,EAAE,EAAE,MAAM,CAACL,EAAEK,EAAED,CAAC,CAAC,CAAC,SAAS8lE,GAAGpmE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEwlE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,GAAGI,IAAI,WAAW,CAACA,EAAEwlE,EAAE,mBAAmB9lE,EAAE,EAAEE,CAAC,EAAE,IAAIK,EAAE,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,IAAIA,EAAE,GAAG,GAAGD,EAAE,EAAE,GAAGC,EAAE,GAAG,GAAGD,EAAE,EAAE,EAAE,GAAG,OAAOC,CAAC,CAAC,OAAOD,CAAC,CAAC,SAAS+lE,GAAGrmE,EAAE,CAAC,OAAOA,EAAE,KAAKA,EAAE2wB,GAAG3wB,CAAC,CAAC,CAAC,IAAIsmE,GAAG,CAAC,EAAElmE,GAAGkmE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,UAAU,KAAK,SAAS,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEpmE,GAAGomE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,aAAa,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,aAAa,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEtmE,GAAGsmE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,kBAAkB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,UAAU,KAAK,SAAS,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,aAAa,KAAK,YAAY,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,OAAO,eAAe,KAAK,cAAc,KAAK,MAAM,EAAE,CAAC,OAAO,mBAAmB,KAAK,iBAAiB,KAAK,MAAM,EAAE,CAAC,OAAO,2BAA2B,KAAK,yBAAyB,KAAK,MAAM,EAAE,CAAC,OAAO,oBAAoB,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,qBAAqB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,uBAAuB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,wBAAwB,KAAK,sBAAsB,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,qBAAqB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,qBAAqB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,MAAM,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,KAAK,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,MAAM,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,iBAAiB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,MAAM,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,MAAM,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,uBAAuB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,OAAO,eAAe,KAAK,cAAc,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,qBAAqB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,EAAE,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,qBAAqB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAExmE,GAAGwmE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,UAAU,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,EAAE,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,OAAO,yBAAyB,KAAK,sBAAsB,KAAK,MAAM,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,KAAK,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,WAAW,KAAK,WAAW,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,gBAAgB,KAAK,gBAAgB,KAAK,MAAM,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,eAAe,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,WAAW,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,mBAAmB,KAAK,gBAAgB,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,WAAW,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,SAAS,aAAa,IAAI,EAAE,CAAC,OAAO,kBAAkB,KAAK,iBAAiB,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,WAAW,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,wBAAwB,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,6BAA6B,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,WAAW,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,WAAW,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,oBAAoB,KAAK,mBAAmB,KAAK,WAAW,aAAa,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,MAAM,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,cAAc,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,OAAO,QAAQ,KAAK,YAAY,KAAK,UAAU,EAAE,CAAC,OAAO,UAAU,KAAK,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE1mE,GAAG0mE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,OAAO,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,MAAM,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,SAAS,aAAa,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,uBAAuB,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,SAAS,aAAa,EAAE,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,SAAS,aAAa,EAAE,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,SAAS,aAAa,EAAE,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,eAAe,KAAK,eAAe,KAAK,OAAO,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE5mE,GAAG4mE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,cAAc,KAAK,YAAY,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,yBAAyB,KAAK,qBAAqB,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,YAAY,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE9mE,GAAG8mE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,aAAa,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,iBAAiB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEhnE,GAAGgnE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,yBAAyB,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,QAAQ,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,OAAO,EAAE,CAAC,SAAS,WAAW,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,IAAI,KAAK,SAAS,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,IAAI,KAAK,SAAS,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,OAAO,UAAU,KAAK,SAAS,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,QAAQ,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,eAAe,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,0BAA0B,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,EAAE,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAElnE,GAAGknE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,YAAY,SAAS,aAAa,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,QAAQ,EAAE,CAAC,OAAO,wBAAwB,KAAK,qBAAqB,KAAK,MAAM,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,OAAO,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,aAAa,OAAO,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,QAAQ,EAAE,CAAC,OAAO,wBAAwB,KAAK,qBAAqB,KAAK,MAAM,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,OAAO,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,sBAAsB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,kBAAkB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEpnE,GAAGonE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,iBAAiB,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,MAAM,EAAE,CAAC,OAAO,qBAAqB,KAAK,mBAAmB,KAAK,MAAM,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,wBAAwB,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,MAAM,EAAE,CAAC,OAAO,qBAAqB,KAAK,mBAAmB,KAAK,MAAM,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,SAAS,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,OAAO,sBAAsB,KAAK,qBAAqB,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,6BAA6B,SAAS,QAAQ,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,YAAY,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEtnE,GAAGsnE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,QAAQ,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,eAAe,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,YAAY,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,UAAU,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,YAAY,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAExnE,GAAGwnE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,eAAe,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,OAAO,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,WAAW,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,WAAW,aAAa,CAAC,CAAC,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,SAAS,aAAa,IAAI,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,kBAAkB,KAAK,iBAAiB,KAAK,SAAS,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,QAAQ,KAAK,aAAa,KAAK,OAAO,aAAa,EAAE,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,UAAU,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,WAAW,KAAK,WAAW,KAAK,QAAQ,EAAE,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE1nE,GAAG0nE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,gBAAgB,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,OAAO,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,iBAAiB,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,SAAS,aAAa,IAAI,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,SAAS,aAAa,IAAI,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,SAAS,aAAa,IAAI,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,eAAe,KAAK,SAAS,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,QAAQ,KAAK,QAAQ,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,gBAAgB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,mBAAmB,KAAK,kBAAkB,KAAK,OAAO,aAAa,GAAG,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE5nE,GAAG4nE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,WAAW,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,MAAM,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,YAAY,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,MAAM,EAAE,CAAC,OAAO,UAAU,KAAK,UAAU,KAAK,MAAM,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAE9nE,GAAG8nE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,GAAG,KAAK,UAAU,KAAK,SAAS,EAAE,CAAC,MAAM,GAAG,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,UAAU,KAAK,SAAS,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,IAAI,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,aAAa,KAAK,YAAY,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,mBAAmB,KAAK,kBAAkB,KAAK,OAAO,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,eAAe,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,MAAM,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,aAAa,KAAK,YAAY,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,WAAW,KAAK,UAAU,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,gBAAgB,KAAK,cAAc,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,gBAAgB,KAAK,eAAe,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,mBAAmB,KAAK,iBAAiB,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,UAAU,KAAK,SAAS,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,OAAO,MAAM,KAAK,MAAM,KAAK,SAAS,aAAa,EAAE,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,kBAAkB,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,SAAS,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,kBAAkB,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,aAAa,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,cAAc,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,mBAAmB,KAAK,kBAAkB,KAAK,OAAO,aAAa,GAAG,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEhoE,GAAGgoE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,sBAAsB,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,eAAe,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,IAAI,KAAK,QAAQ,KAAK,QAAQ,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,oBAAoB,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,mBAAmB,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEloE,GAAGkoE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,MAAM,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,OAAO,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,WAAW,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,SAAS,aAAa,EAAE,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEpoE,GAAGooE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,eAAe,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,YAAY,KAAK,YAAY,KAAK,QAAQ,EAAE,CAAC,OAAO,eAAe,KAAK,cAAc,KAAK,UAAU,EAAE,CAAC,OAAO,WAAW,KAAK,UAAU,KAAK,QAAQ,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,QAAQ,EAAE,CAAC,OAAO,YAAY,KAAK,WAAW,KAAK,QAAQ,EAAE,CAAC,OAAO,2BAA2B,KAAK,yBAAyB,KAAK,MAAM,CAAC,EAAE,QAAQ,CAAC,SAAS,eAAe,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,YAAY,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,aAAa,KAAK,YAAY,KAAK,MAAM,CAAC,EAAE,QAAQ,CAAC,UAAU,SAAS,OAAO,CAAC,EAAE,CAAC,SAAS,yBAAyB,SAAS,SAAS,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,QAAQ,CAAC,CAAC,CAAC,EAAMC,GAAG,CAAC,EAAEtoE,GAAGsoE,GAAG,CAAC,KAAK,IAAIC,EAAE,CAAC,EAAE,IAAIA,GAAG,CAAC,CAAC,SAAS,OAAO,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,SAAS,KAAK,QAAQ,aAAa,EAAE,EAAE,CAAC,OAAO,OAAO,KAAK,QAAQ,KAAK,OAAO,CAAC,CAAC,EAAE,CAAC,SAAS,aAAa,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,YAAY,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,KAAK,OAAO,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,MAAM,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,iBAAiB,KAAK,gBAAgB,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,QAAQ,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,gBAAgB,KAAK,SAAS,aAAa,CAAC,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,UAAU,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,OAAO,iBAAiB,eAAe,KAAK,OAAO,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,iBAAiB,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,WAAW,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,iBAAiB,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,aAAa,KAAK,UAAU,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,CAAC,EAAE,CAAC,SAAS,eAAe,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,OAAO,aAAa,KAAK,YAAY,KAAK,QAAQ,EAAE,CAAC,OAAO,cAAc,KAAK,aAAa,KAAK,QAAQ,CAAC,CAAC,EAAE,CAAC,SAAS,cAAc,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,IAAI,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,QAAQ,KAAK,UAAU,CAAC,EAAE,MAAM,CAAC,CAAC,EAAE,CAAC,SAAS,gBAAgB,SAAS,iBAAiB,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,KAAK,KAAK,QAAQ,EAAE,CAAC,MAAM,EAAE,KAAK,KAAK,KAAK,QAAQ,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC,EAAMC,GAAG,KAAK,CAAC,aAAa,CAAC,IAAI9nE,EAAE,CAACwlE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,EAAE,EAAExoE,EAAE,CAAC,EAAE,OAAO,GAAGY,EAAE,IAAI,GAAG,EAAE,IAAI,CAAC,EAAE,KAAK,UAAUZ,EAAE,OAAO,CAAC,EAAEK,KAAK,EAAEA,EAAE,UAAUA,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,WAAW,UAAU,CAAC,OAAO,KAAK,YAAY,KAAK,UAAU,IAAI,KAAK,CAAC,eAAeO,EAAEZ,EAAE,CAAC,EAAE,CAAC,IAAI,EAAEY,EAAE,KAAKP,EAAE,CAAC,EAAEwB,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,EAAE,OAAO,CAACiC,EAAEC,KAAKD,EAAEC,EAAE,MAAM,KAAK,QAAQA,CAAC,EAAEA,EAAE,GAAG,WAAW,aAAa,EAAE5D,EAAE,KAAK2D,EAAEC,EAAE,KAAK,EAAEA,EAAE,KAAK,QAAQpC,EAAE,KAAKmC,EAAEC,EAAE,KAAK,GAAGA,EAAE,OAAO,MAAMA,EAAE,MAAM,SAAS,IAAInC,EAAE,KAAKkC,EAAEC,EAAE,KAAK,EAAED,GAAG,CAAC,CAAC,EAAEhC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEU,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE1C,GAAG,OAAO0C,EAAE,KAAK,oBAAoB1C,EAAE,MAAM,EAAE,EAAE,KAAK,oBAAoBA,EAAE,OAAO,GAAG,IAAI4C,EAAE,OAAO,KAAKb,CAAC,EAAEa,EAAE,QAAQoB,GAAG,CAAC,IAAIC,EAAElC,EAAEiC,GAAGC,EAAE,WAAW,QAAQ,CAACoD,EAAExC,IAAI,CAAC,GAAG,CAACJ,EAAE,CAACC,CAAC,EAAEuhE,GAAG5+D,CAAC,EAAE1C,EAAE5C,EAAE0C,GAAG,GAAGE,EAAE,SAAS,KAAK,CAAC,IAAIC,EAAED,EAAE,QAAQ,QAAQD,CAAC,EAAE,GAAGE,IAAI,GAAG,CAAC,IAAI0C,EAAE,GAAG7C,KAAKG,IAAIX,EAAE,WAAWY,GAAGyC,CAAC,CAAC,CAACrD,EAAE,OAAO,KAAKU,CAAC,EAAEA,EAAE,SAAS,KAAKV,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,OAAO,KAAK,CAAC,EAAE,SAAS,EAAErB,EAAE,QAAQoB,GAAG,CAAC,IAAIC,EAAElC,EAAEiC,GAAGC,EAAE,SAAS,SAAS,GAAG,EAAE,KAAKA,CAAC,CAAC,CAAC,EAAE,OAAO,KAAK,CAAC,EAAE,QAAQD,GAAG,CAAC,GAAG,CAACC,CAAC,EAAEgiE,GAAGjiE,CAAC,EAAEqD,EAAEtF,EAAEkC,GAAGoD,GAAG,OAAOA,EAAE,aAAa,EAAErD,GAAG,EAAE,KAAKqD,CAAC,EAAE,CAAC,EAAE,OAAO,KAAK3E,CAAC,EAAE,OAAO,EAAE,OAAO,KAAKA,CAAC,EAAE,QAAQsB,GAAG,CAAC,GAAG,CAACC,CAAC,EAAEgiE,GAAGjiE,CAAC,EAAEqD,EAAEtF,EAAEkC,GAAGoD,IAAIA,EAAE,aAAa3E,EAAEsB,GAAGhC,EAAE,KAAKqF,CAAC,EAAE,CAAC,EAAErF,EAAE3B,EAAE,IAAIwC,EAAE,CAAC,EAAEjC,EAAE,SAAS,MAAMA,EAAE,QAAQ,UAAU,OAAOiC,EAAEjC,EAAE,QAAQ,SAAS,OAAO,CAACoD,EAAEC,KAAKD,EAAEC,EAAE,UAAU,MAAM,KAAK,YAAYA,CAAC,EAAED,GAAG,CAAC,CAAC,GAAG,IAAIlB,EAAE,CAAC,MAAMf,EAAE,OAAOC,EAAE,QAAQ,EAAE,QAAQH,EAAE,aAAaxB,EAAE,UAAUL,EAAE,UAAU6C,CAAC,EAAE,OAAOf,EAAE,OAAO,IAAIgB,EAAE,UAAUhB,GAAGgB,CAAC,CAAC,oBAAoBlC,EAAE,CAAC,OAAO,OAAO,KAAKA,GAAG,CAAC,CAAC,EAAE,OAAO,CAACZ,EAAE,KAAKA,EAAEY,EAAE,GAAG,MAAM,EAAEZ,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQY,EAAE,CAAC,IAAIZ,EAAE0lE,GAAG9kE,EAAE,EAAE,GAAG,KAAK,UAAUA,EAAE,KAAK,CAAC,EAAEA,EAAE,MAAM,OAAOA,EAAE,KAAK,CAAC,GAAG,IAAI,EAAE,CAAC,KAAKA,EAAE,KAAK,GAAGA,EAAE,GAAG,SAASZ,EAAE,SAAS,YAAYY,EAAE,OAAO,CAAC,GAAG,IAAIP,GAAGA,EAAE,WAAW,GAAG,EAAEA,EAAE,MAAM,CAAC,EAAEA,CAAC,EAAE,OAAO,CAAC,EAAE,SAAS,CAAC,EAAE,YAAY,CAAC,EAAE,WAAW,CAAC,EAAE,SAASO,EAAE,KAAK,QAAQZ,EAAE,OAAO,EAAE,OAAOA,EAAE,QAAQ,OAAO,EAAE,YAAYA,EAAE,OAAO,OAAO,CAACK,EAAEwB,KAAKxB,EAAEwB,EAAE,MAAM,CAAC,KAAKA,EAAE,KAAK,gBAAgBA,EAAE,MAAM,cAAcA,EAAE,GAAG,EAAExB,GAAG,CAAC,CAAC,GAAGL,EAAE,OAAO,OAAO,EAAE,WAAWA,EAAE,MAAM,OAAO,CAACK,EAAEwB,IAAI,CAAC,IAAIC,EAAED,EAAE,KAAKE,EAAE,OAAOF,EAAE,KAAK,CAAC,IAAI,SAASE,EAAE4mE,GAAG/nE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAE4mE,GAAG/nE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,WAAWE,EAAE6mE,GAAGhoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAE6mE,GAAGhoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,SAASE,EAAE8mE,GAAGjoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,cAAc,CAAC,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAE8mE,GAAGjoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,WAAWE,EAAE+mE,GAAGloE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAE+mE,GAAGloE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,OAAOE,EAAEgnE,GAAGnoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEgnE,GAAGnoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,SAASE,EAAEinE,GAAGpoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEinE,GAAGpoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,QAAQE,EAAEknE,GAAGroE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEknE,GAAGroE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,UAAUE,EAAEmnE,GAAGtoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEmnE,GAAGtoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,QAAQE,EAAEonE,GAAGvoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEonE,GAAGvoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,UAAUE,EAAEqnE,GAAGxoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEqnE,GAAGxoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,OAAOE,EAAEsnE,GAAGzoE,EAAE,KAAKiB,EAAE,OAAOA,EAAE,YAAY,EAAEE,IAAI,QAAQ,CAAC,CAACF,EAAE,mBAAmBE,EAAEsnE,GAAGzoE,EAAE,KAAKiB,EAAE,iBAAiBA,EAAE,YAAY,GAAG,MAAM,IAAI,SAAS,IAAI,UAAU,MAAM,QAAQ,MAAM,IAAI,MAAM,2BAA2BA,EAAE,gBAAgBjB,EAAE,IAAI,CAAC,CAAC,OAAOP,EAAEwB,EAAE,MAAM,CAAC,MAAME,EAAE,KAAKD,CAAC,EAAEzB,CAAC,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,YAAYO,EAAE,CAAC,IAAIZ,EAAEY,EAAE,QAAQ,EAAE,CAAC,EAAEP,EAAE,CAAC,EAAEwB,EAAE,CAAC,EAAE7B,GAAG,OAAO6B,EAAE7B,EAAE,OAAO,CAAC2C,EAAEC,KAAKD,EAAEC,EAAE,MAAM,KAAK,QAAQA,CAAC,EAAEA,EAAE,KAAK,SAASvC,EAAE,KAAKsC,EAAEC,EAAE,KAAK,EAAED,GAAG,CAAC,CAAC,GAAG,IAAIb,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAEnB,EAAE,UAAU,SAAS,QAAQ+B,GAAG,CAAC,GAAG,CAACC,CAAC,EAAEqjE,GAAGtjE,EAAE,IAAI,EAAEE,EAAE,CAAC,KAAKD,EAAE,GAAG,cAAc,OAAO,CAAC,EAAE,WAAW,CAAC,EAAE,SAAS,QAAQ,YAAY,CAAC,EAAE,WAAW,CAAC,MAAM,CAAC,MAAM0mE,GAAG3mE,EAAE,IAAI,EAAE,KAAK,OAAO,CAAC,EAAE,SAAS,CAAC,CAAC,EAAEE,EAAE,aAAaF,EAAE,KAAKb,EAAE,KAAKe,CAAC,EAAEhB,EAAEe,GAAGC,CAAC,CAAC,EAAE,OAAO,KAAKhB,CAAC,EAAE,QAAQc,GAAG,CAAC,IAAIC,EAAEf,EAAEc,GAAGC,EAAE,WAAW,QAAQ,CAACC,EAAEC,IAAI,CAAC,GAAG,CAAC,EAAE,CAACmB,CAAC,EAAEgiE,GAAGpjE,CAAC,EAAEwE,EAAExF,EAAE,GAAG,GAAGwF,EAAE,SAAS,KAAK,CAAC,IAAIxC,EAAEwC,EAAE,QAAQ,QAAQpD,CAAC,EAAE,GAAGY,IAAI,GAAG,CAAC,IAAIJ,EAAE,GAAG,KAAKI,IAAIjC,EAAE,WAAWE,GAAG2B,CAAC,CAAC,CAAC7B,EAAE,OAAO,KAAKyE,CAAC,EAAEA,EAAE,SAAS,KAAKzE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,IAAIX,EAAErB,EAAE,IAAIA,EAAE,UAAU,UAAU,QAAQ+B,GAAG,CAAC,GAAG,CAACC,EAAEC,CAAC,EAAEojE,GAAGhkE,EAAEU,EAAE,KAAK,EAAEG,EAAEjB,EAAEe,GAAGE,GAAG,OAAOA,EAAE,cAAcD,EAAEd,EAAE,KAAKe,CAAC,EAAE,CAAC,EAAE,IAAIJ,EAAE,KAAK,mBAAmB9B,CAAC,EAAE,MAAM,CAAC,MAAMiB,EAAE,OAAOC,EAAE,QAAQC,EAAE,QAAQ1B,EAAE,aAAa,EAAE,UAAUqC,CAAC,CAAC,CAAC,mBAAmB9B,EAAE,CAAC,MAAM,CAAC,WAAWA,EAAE,UAAU,KAAK,OAAOA,EAAE,UAAU,SAAS,OAAO,CAACZ,EAAE,KAAKA,EAAE,EAAE,MAAM,KAAK,mBAAmB,CAAC,EAAEA,GAAG,CAAC,CAAC,EAAE,QAAQY,EAAE,UAAU,UAAU,OAAO,CAACZ,EAAE,KAAKA,EAAE,EAAE,MAAM,KAAK,mBAAmB,EAAEY,EAAE,GAAG,EAAEZ,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBY,EAAEZ,EAAE,CAAC,IAAI,EAAEY,EAAE,KAAK,OAAOZ,GAAG,OAAO,EAAEA,EAAE,IAAI,CAAC,KAAK,EAAE,MAAMY,EAAE,IAAI,CAAC,CAAC,EAAE,SAAS2oE,GAAGzpE,EAAE,CAAC,IAAI,EAAEya,EAAE,EAAE,OAAO,GAAG,OAAO,EAAE,MAAM,YAAY,OAAO,EAAE,KAAKza,CAAC,EAAE,GAAG,OAAO,QAAQ,YAAY,OAAO,IAAI,OAAOA,EAAE,QAAQ,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,kFAAkF,CAAC,CAAC,SAAS0pE,GAAG1pE,EAAE,EAAE,CAAC,IAAIE,EAAE,MAAM,QAAQF,CAAC,EAAE,OAAO,aAAa,MAAM,KAAKA,CAAC,EAAEypE,GAAGzpE,CAAC,EAAE,OAAO,EAAEE,EAAEA,EAAE,YAAY,CAAC,CAAC,SAAS2oE,GAAG7oE,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEP,EAAE,GAAG,OAAOO,GAAG,KAAKmpE,GAAGnpE,EAAE,EAAED,CAAC,EAAEJ,CAAC,CAAC,SAAS+oE,GAAGjpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,EAAEA,EAAE,EAAEJ,CAAC,CAAC,SAAS6oE,GAAG/oE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,IAAI,CAAC,EAAEO,EAAED,EAAE,GAAG,KAAKA,EAAE,EAAEA,EAAE,GAAG,KAAKA,EAAE,EAAEJ,EAAE,OAAO,OAAOK,GAAG,SAASA,EAAE,SAASA,EAAE,EAAE,CAAC,CAAC,SAASipE,GAAGxpE,EAAE,CAAC,OAAO,OAAOA,GAAG,WAAWA,EAAEwlE,GAAGxlE,IAAIA,EAAE,CAAC,KAAKwlE,GAAG,SAAS,KAAKA,GAAG,QAAQ,MAAM,UAAU,KAAKA,GAAG,SAAS,KAAKA,GAAG,SAAS,KAAKA,GAAG,QAAQ,KAAKA,GAAG,SAAS,MAAM,QAAQ,KAAKA,GAAG,QAAQ,MAAM,OAAO,KAAKA,GAAG,UAAU,MAAM,UAAU,KAAKA,GAAG,UAAU,MAAM,SAAS,QAAQ,OAAO,IAAI,CAAC,CAAC,SAAS+D,GAAGvpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,KAAKA,EAAE,KAAK,KAAKJ,CAAC,CAAC,SAASmpE,GAAGrpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,KAAKkpE,GAAGlpE,EAAE,IAAI,EAAEJ,CAAC,CAAC,SAASopE,GAAGtpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,MAAMA,EAAE,KAAK,KAAKA,EAAE,KAAK,KAAK,IAAIC,GAAGipE,GAAGjpE,CAAC,CAAC,EAAEL,CAAC,CAAC,SAASypE,GAAG3pE,EAAE,CAAC,GAAG,CAACA,EAAE,YAAY,OAAOA,EAAE,KAAK,KAAKA,EAAE,IAAI,IAAI,GAAG,OAAO,EAAE,MAAM,SAAS,EAAE,KAAK,SAAS,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,SAASmpE,GAAGnpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,MAAMqpE,GAAGrpE,EAAE,KAAK,EAAEJ,CAAC,CAAC,SAAS8oE,GAAGhpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,IAAIA,EAAE,KAAK,GAAGA,EAAE,KAAK,EAAE,OAAOA,EAAE,KAAK,EAAEA,EAAE,KAAK,IAAI,CAAC,GAAG,IAAIC,GAAG,OAAOA,GAAG,SAASA,EAAE,SAASA,EAAE,EAAE,CAAC,EAAEL,CAAC,CAAC,SAAS4oE,GAAG9oE,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEP,EAAE,GAAG,OAAOO,GAAGA,EAAE,MAAMA,EAAE,KAAK,EAAEA,EAAE,KAAK,EAAE,IAAI,GAAGmpE,GAAG,EAAEppE,CAAC,CAAC,EAAEJ,CAAC,CAAC,SAASkpE,GAAGppE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,MAAMA,EAAE,KAAK,MAAMA,EAAE,KAAK,MAAM,IAAIC,GAAGopE,GAAGppE,CAAC,CAAC,EAAEL,CAAC,CAAC,SAASgpE,GAAGlpE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,OAAOM,GAAGA,EAAE,MAAMA,EAAE,KAAK,EAAEA,EAAE,KAAK,EAAEJ,CAAC,CAAC,IAAI0pE,GAAG,KAAK,CAAC,YAAY9oE,EAAEZ,EAAE,EAAE,CAAC,KAAK,KAAKY,EAAE,KAAK,UAAUZ,EAAE,KAAK,QAAQ,EAAE,KAAK,OAAO,CAAC,EAAE,KAAK,MAAM,CAAC,EAAE,KAAK,OAAOY,EAAE,WAAW,IAAIP,GAAG,KAAK,SAASA,CAAC,CAAC,EAAEO,EAAE,UAAU,OAAO,KAAK,MAAM,OAAO,KAAKA,EAAE,QAAQ,EAAE,OAAO,CAACP,EAAEwB,KAAKxB,EAAEwB,GAAG,KAAK,QAAQA,CAAC,EAAExB,GAAG,CAAC,CAAC,EAAE,CAAC,SAASO,EAAE,CAAC,OAAOilE,GAAGjlE,EAAE,KAAK,UAAU,KAAK,OAAO,CAAC,CAAC,QAAQA,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,KAAK,SAASY,GAAG,GAAG,EAAE,QAAQ,KAAK,OAAOilE,GAAGjlE,EAAE,KAAK,UAAU,KAAK,OAAO,EAAE,GAAG,EAAE,GAAG,MAAM,EAAE,GAAG,KAAK,OAAOioE,GAAG,KAAK,KAAK,SAASjoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,GAAG,KAAK,OAAO2oE,GAAG,KAAK,KAAK,SAAS/nE,EAAEZ,CAAC,EAAE,GAAG,EAAE,GAAG,KAAK,OAAO+oE,GAAG,KAAK,KAAK,SAASnoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,OAAO,KAAK,OAAOipE,GAAG,KAAK,KAAK,SAASroE,EAAEZ,CAAC,EAAE,GAAG,EAAE,MAAM,KAAK,OAAOmpE,GAAG,KAAK,KAAK,SAASvoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,MAAM,KAAK,CAAC,GAAG,EAAE,KAAK,GAAG,MAAM,EAAE,KAAK,GAAG,KAAK,OAAO8oE,GAAG,KAAK,KAAK,SAASloE,EAAEZ,CAAC,EAAE,GAAG,EAAE,KAAK,GAAG,KAAK,OAAO4oE,GAAG,KAAK,KAAK,SAAShoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,KAAK,OAAO,KAAK,OAAOkpE,GAAG,KAAK,KAAK,SAAStoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,KAAK,GAAG,KAAK,OAAOgpE,GAAG,KAAK,KAAK,SAASpoE,EAAEZ,CAAC,EAAE,GAAG,EAAE,KAAK,MAAM,KAAK,OAAOopE,GAAG,KAAK,KAAK,SAASxoE,EAAEZ,CAAC,CAAC,CAAC,OAAOA,CAAC,CAAC,EAAM2pE,GAAG,CAAC,EAAEzpE,GAAGypE,GAAG,CAAC,gBAAgB,IAAI39C,GAAG,IAAI,IAAIoN,GAAG,KAAK,IAAIE,GAAG,MAAM,IAAIE,GAAG,IAAI,IAAIZ,EAAE,KAAK,IAAIc,GAAG,IAAI,IAAIE,GAAG,IAAI,IAAIE,GAAG,OAAO,IAAIE,GAAG,OAAO,IAAIE,GAAG,KAAK,IAAIE,GAAG,MAAM,IAAIE,GAAG,KAAK,IAAIE,GAAG,MAAM,IAAIE,GAAG,MAAM,IAAIE,GAAG,QAAQ,IAAIsB,GAAG,UAAU,IAAIE,GAAG,cAAc,IAAIU,GAAG,UAAU,IAAIK,GAAG,YAAY,IAAIE,GAAG,YAAY,IAAIE,GAAG,YAAY,IAAIE,GAAG,eAAe,IAAIT,GAAG,SAAS,IAAIW,GAAG,iBAAiB,IAAI2T,GAAG,cAAc,IAAIzT,GAAG,YAAY,IAAIE,GAAG,OAAO,IAAI1N,GAAG,KAAK,IAAIE,GAAE,KAAK,IAAI0N,GAAG,YAAY,IAAIG,GAAG,MAAM,IAAI3N,GAAG,QAAQ,IAAItE,GAAG,OAAO,IAAImQ,GAAG,SAAS,IAAIgC,GAAG,SAAS,IAAIE,GAAG,SAAS,IAAIE,GAAG,SAAS,IAAIE,GAAG,OAAO,IAAII,GAAG,OAAO,IAAIF,GAAG,gBAAgB,IAAIM,GAAG,OAAO,IAAIE,GAAG,gBAAgB,IAAII,GAAG,IAAI,IAAIE,GAAG,KAAK,IAAIE,GAAG,aAAa,IAAIsS,GAAG,QAAQ,IAAIpS,GAAG,OAAO,IAAIE,GAAG,cAAc,IAAIE,GAAG,aAAa,IAAIE,GAAG,gBAAgB,IAAIE,GAAG,KAAK,IAAIE,GAAG,WAAW,IAAIE,GAAG,IAAI,IAAI5H,GAAG,SAAS,IAAIoI,GAAG,IAAI,IAAIE,GAAG,QAAQ,IAAI4Q,GAAG,OAAO,IAAI1Q,GAAG,IAAI,IAAIE,GAAG,oBAAoB,IAAIyQ,GAAG,MAAM,IAAIrR,GAAG,IAAI,IAAIc,GAAG,cAAc,IAAI0B,GAAG,IAAI,IAAIE,GAAG,WAAW,IAAIE,GAAG,MAAM,IAAIE,GAAG,IAAI,IAAII,GAAG,IAAI,IAAI4K,GAAG,KAAK,IAAI1Q,GAAG,MAAM,IAAIgG,GAAG,SAAS,IAAIpL,GAAG,MAAM,IAAIyZ,GAAG,OAAO,IAAInO,GAAG,SAAS,IAAI2N,GAAG,QAAQ,IAAIzN,GAAG,aAAa,IAAIE,GAAG,KAAK,IAAIsK,GAAG,KAAK,IAAIza,GAAG,MAAM,IAAIglB,GAAG,YAAY,IAAI/G,GAAG,MAAM,IAAItD,GAAG,SAAS,IAAItK,GAAG,MAAM,IAAIE,GAAG,MAAM,IAAIE,GAAG,UAAU,IAAIE,GAAG,KAAK,IAAIE,GAAG,UAAU,IAAIE,GAAG,OAAO,IAAIkU,GAAG,SAAS,IAAIjU,GAAG,2BAA2B,IAAIE,GAAG,IAAI,IAAIE,GAAG,MAAM,IAAIE,GAAG,WAAW,IAAIW,GAAG,WAAW,IAAII,GAAG,UAAU,IAAIE,GAAG,WAAW,IAAIE,GAAG,WAAW,IAAIE,GAAG,UAAU,IAAIE,GAAG,WAAW,IAAIE,GAAG,OAAO,IAAImS,GAAG,WAAW,IAAI/R,GAAG,OAAO,IAAI5U,GAAG,IAAI,IAAI0P,GAAG,QAAQ,IAAIoF,GAAG,UAAU,IAAIE,GAAG,kBAAkB,IAAIE,GAAG,QAAQ,IAAIE,GAAG,KAAK,IAAIE,GAAG,SAAS,IAAIG,GAAG,IAAI,IAAI7F,GAAG,QAAQ,IAAI+F,GAAG,UAAU,IAAIE,GAAG,IAAI,IAAIE,GAAG,QAAQ,IAAIE,GAAG,cAAc,IAAI2I,GAAG,IAAI,IAAItY,EAAE,aAAa,IAAI6P,GAAG,YAAY,IAAIE,GAAG,IAAI,IAAI1U,GAAG,KAAK,IAAI6O,GAAG,SAAS,IAAI+F,GAAG,OAAO,IAAIrW,GAAG,KAAK,IAAIsV,GAAG,SAAS,IAAIiB,GAAG,GAAG,IAAIpd,EAAE,aAAa,IAAIsd,GAAG,IAAI,IAAIE,GAAG,MAAM,IAAIE,GAAG,MAAM,IAAIE,GAAG,MAAM,IAAIE,GAAG,MAAM,IAAIE,GAAG,KAAK,IAAIM,GAAG,IAAI,IAAI7H,GAAG,MAAM,IAAI+H,GAAG,MAAM,IAAI/Z,GAAG,KAAK,IAAIia,GAAG,aAAa,IAAIE,GAAG,YAAY,IAAIE,GAAG,qBAAqB,IAAIE,GAAG,KAAK,IAAIE,GAAG,YAAY,IAAIM,GAAG,aAAa,IAAIE,GAAG,qBAAqB,IAAIE,GAAG,cAAc,IAAIE,GAAG,MAAM,IAAIC,GAAG,KAAK,IAAIvX,GAAG,WAAW,IAAIyX,GAAG,KAAK,IAAIE,GAAG,MAAM,IAAIE,GAAG,QAAQ,IAAItQ,EAAE,QAAQ,IAAIwQ,GAAG,UAAU,IAAIE,GAAG,UAAU,IAAIE,GAAG,UAAU,IAAIE,GAAG,UAAU,IAAIE,GAAG,KAAK,IAAIoC,GAAG,MAAM,IAAIlC,GAAG,MAAM,IAAIE,GAAG,OAAO,IAAIzK,GAAG,UAAU,IAAI+O,GAAG,aAAa,IAAInK,GAAG,KAAK,IAAI+F,GAAG,gBAAgB,IAAIE,GAAG,eAAe,IAAIE,GAAG,QAAQ,IAAIlR,GAAG,KAAK,IAAIoR,GAAG,OAAO,IAAIwL,GAAG,IAAI,IAAItL,GAAG,KAAK,IAAIE,GAAG,MAAM,IAAItR,GAAG,QAAQ,IAAIwR,GAAG,QAAQ,IAAIE,GAAG,QAAQ,IAAIE,GAAG,QAAQ,IAAIE,GAAG,QAAQ,IAAIE,GAAG,SAAS,IAAItI,GAAG,eAAe,IAAI+D,GAAG,OAAO,IAAIqP,GAAG,cAAc,IAAI3H,GAAG,SAAS,IAAIsH,GAAG,MAAM,IAAIjK,GAAG,KAAK,IAAIrM,GAAG,OAAO,IAAIE,GAAG,kBAAkB,IAAIuM,GAAG,QAAQ,IAAIE,GAAG,MAAM,IAAIE,GAAG,KAAK,IAAIE,GAAG,aAAa,IAAIE,GAAG,OAAO,IAAI2J,GAAG,IAAI,IAAIjT,GAAG,IAAI,IAAIvD,GAAG,IAAI,IAAI+M,GAAG,KAAK,IAAIpT,GAAG,OAAO,IAAIvQ,GAAG,SAAS,IAAI4jB,GAAG,SAAS,IAAIC,GAAG,SAAS,IAAI7a,GAAG,SAAS,IAAI8a,GAAG,SAAS,IAAIC,GAAG,SAAS,IAAIC,GAAG,KAAK,IAAIvM,GAAG,KAAK,IAAIyM,GAAG,UAAU,IAAI5b,GAAG,gBAAgB,IAAI8b,GAAG,OAAO,IAAIE,GAAG,mBAAmB,IAAIE,GAAG,QAAQ,IAAIE,GAAG,WAAW,IAAIC,GAAG,SAAS,IAAIC,GAAG,MAAM,IAAIjQ,GAAG,WAAW,IAAIoQ,GAAG,MAAM,IAAIjJ,GAAG,UAAU,IAAIjH,EAAE,CAAC,EAAE,IAAI0oC,GAAG,CAAC9pE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,UAAU,IAAI,QAAQ,IAAI,MAAM,MAAM,CAACM,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,MAAM,CAACI,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,MAAM,CAACI,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,oBAAoB,MAAM,CAACI,EAAE,kBAAkBwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAM+pE,GAAG,CAAC/pE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,MAAM,IAAI,aAAa,MAAM,CAACM,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,aAAa,MAAM,CAACI,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,MAAM,CAACI,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,SAAS,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,cAAc,MAAM,CAACI,EAAE,YAAYwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMylE,GAAG/lE,EAAE,WAAW,GAAG,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMylE,GAAG/lE,EAAE,WAAW,GAAG,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAE,SAASgqE,GAAGhqE,EAAE,EAAEE,EAAE,GAAG,CAAC,GAAG,EAAE,OAAOF,GAAG,UAAU,OAAO,GAAG,UAAU,CAAC8mB,EAAE,OAAO9mB,EAAE,SAAS,EAAE,OAAO,IAAIE,EAAE,WAAWF,SAAS,cAAc,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAI,CAAC,IAAIC,EAAEP,EAAEM,GAAG,EAAE,EAAEA,GAAGwmB,EAAE,OAAOvmB,EAAE,GAAG,EAAE,GAAGA,IAAI,EAAE,IAAIL,EAAE,WAAWF,SAAS,cAAc,CAAC,CAAC,CAAC,CAAC,SAASiqE,GAAGjqE,EAAE,CAAC,MAAM,EAAE,OAAOA,GAAG,UAAUA,EAAE,KAAK,GAAG,EAAE,CAAC,EAAE,CAAC,SAASkqE,GAAGlqE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE6pE,GAAGnqE,EAAEE,CAAC,EAAEK,EAAE,CAAC0pE,GAAG3pE,CAAC,EAAE,GAAGC,GAAG,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,qFAAqFD,GAAG,EAAE,GAAGC,GAAG,EAAE,QAAQ,GAAG,CAACD,EAAE6pE,GAAG,EAAE,MAAM7pE,CAAC,CAAC,CAAC,EAAE,CAAC2pE,GAAG3pE,CAAC,EAAE,MAAM,IAAI,MAAM,mCAAmCA,GAAG,EAAE,OAAOA,CAAC,CAAC,SAAS6pE,GAAGnqE,EAAE,EAAE,CAAC,GAAG,OAAOA,GAAG,SAAS,OAAO,EAAE,GAAG,OAAO,GAAG,SAAS,OAAOA,EAAE,GAAGA,EAAE,SAAS,EAAE,OAAO,MAAM,IAAI,MAAM,oCAAoCA,SAAS,GAAG,EAAE,IAAIE,EAAE,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAO,EAAEM,EAAE,CAAC,IAAIC,EAAEP,EAAEM,GAAG,EAAE,EAAEA,GAAG,GAAGC,GAAG,GAAG,GAAG,GAAGA,IAAI,EAAE,MAAM,IAAI,MAAM,oCAAoCP,SAAS,GAAG,EAAEE,EAAEI,GAAGC,GAAG,EAAEA,EAAE,CAAC,CAAC,OAAOL,CAAC,CAAC,IAAIkqE,GAAG,KAAK,CAAC,YAAYtpE,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAE,CAAC,KAAK,KAAKnB,EAAE,KAAK,MAAMZ,EAAE,KAAK,QAAQ,EAAE,KAAK,aAAaK,EAAE,KAAK,uBAAuBwB,EAAE,KAAK,YAAYC,EAAE,KAAK,eAAeC,EAAE,KAAK,QAAQ,CAAC,EAAE,KAAK,QAAQ,GAAG,KAAK,SAAS4gC,GAAG,CAAC,EAAElP,GAAG,KAAK,QAAQ,CAAC,CAAC,IAAI,IAAI,CAAC,OAAO,KAAK,SAAS,EAAE,CAAC,IAAI,QAAQ,CAAC,OAAO,KAAK,OAAO,CAAC,cAAc7yB,EAAE,CAAC,KAAK,QAAQ,QAAQZ,GAAG,EAAEY,GAAG,MAAM,CAACA,EAAE,IAAIZ,EAAE,OAAO,EAAE,IAAIA,EAAE,OAAO,QAAQ,CAAC,CAAC,EAAE,KAAK,QAAQ,CAAC,EAAE,KAAK,QAAQ,GAAG,KAAK,SAAS,QAAQ,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,QAAQ,MAAM,CAAC,KAAKY,EAAE,CAAC,GAAG,KAAK,QAAQ,MAAM,IAAI,MAAM,eAAe,KAAK,+BAA+B,EAAE,GAAGA,EAAE,GAAGA,GAAG,KAAK,KAAK,EAAE,MAAM,IAAI,MAAM,4BAA4BA,yBAAyB,KAAK,KAAK,GAAG,EAAE,IAAIZ,EAAE,KAAK,QAAQY,GAAG,GAAGZ,EAAE,QAAQ,MAAM,IAAI,MAAM,eAAe,KAAK,8BAA8BY,uGAAuG,EAAE,OAAO,KAAK,iBAAiBZ,EAAE,QAAQ,IAAIA,EAAE,KAAK,GAAGA,EAAE,MAAM,CAAC,SAASY,EAAE,CAAC,OAAOA,EAAE,IAAIZ,GAAG,KAAK,KAAKA,CAAC,CAAC,CAAC,CAAC,MAAMY,EAAEZ,EAAE,CAAC,GAAG,KAAK,QAAQ,MAAM,IAAI,MAAM,eAAe,KAAK,+BAA+B,EAAE,GAAGY,EAAE,GAAG,CAAC,KAAK,aAAaA,GAAG,KAAK,QAAQ,MAAM,IAAI,MAAM,2BAA2BA,+CAA+C,KAAK,SAAS,EAAE,IAAI,EAAE,KAAK,QAAQA,IAAI,CAAC,EAAE,GAAGZ,EAAE,QAAQ,KAAK,MAAM,MAAM,IAAI,MAAM,eAAe,KAAK,8CAA8CY;AAAA,uCACvwzEZ,EAAE,mCAAmC,KAAK,QAAQ,EAAE,GAAG,KAAK,KAAK,IAAI,IAAI,KAAK,cAAc,MAAM,KAAK,aAAa,SAAS,KAAK,KAAK,aAAaA,EAAE,OAAO8pE,GAAG,KAAK,aAAa9pE,EAAE,MAAM,eAAe,KAAK,8CAA8CY,IAAI,EAAE,EAAE,KAAK,MAAM,IAAI,MAAM,eAAe,KAAK,8CAA8CA,sCAAsC,EAAE,GAAG,EAAE,QAAQ,MAAM,IAAI,MAAM,eAAe,KAAK,8CAA8CA,yCAAyC,EAAE,EAAE,OAAOZ,EAAEyzB,GAAGzzB,CAAC,EAAE,EAAE,QAAQ,GAAG,KAAK,QAAQY,GAAG,CAAC,CAAC,UAAUA,EAAEZ,EAAE,CAAC,GAAGY,EAAE,SAASZ,EAAE,OAAO,MAAM,IAAI,MAAM,eAAe,KAAK,kEAAkEY,EAAE,2CAA2CZ,EAAE,SAAS,EAAEY,EAAE,QAAQ,CAAC,EAAEP,IAAI,KAAK,MAAM,EAAEL,EAAEK,EAAE,CAAC,CAAC,CAAC,OAAOO,EAAEZ,EAAE,CAAC,GAAG,CAAC,CAACA,GAAGA,IAAI,KAAK,MAAM,MAAM,IAAI,MAAM,wBAAwB,KAAK,oCAAoCA,GAAG,EAAE,GAAGY,EAAEA,EAAEA,EAAE,MAAM,EAAE,KAAK,KAAK,CAAC,MAAM,CAACA,EAAE,CAAC,EAAE,QAAQP,EAAE,EAAEA,EAAE,KAAK,KAAK,EAAEA,IAAIO,EAAE,KAAKP,CAAC,CAAC,CAAC,GAAGO,EAAE,SAAS,EAAE,OAAOyrB,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,OAAO,KAAK,YAAY,CAAC,EAAE,IAAI,EAAE,KAAK,SAASzrB,CAAC,EAAE,OAAOkpE,GAAG,KAAK,aAAa,EAAE,GAAG,MAAM,8BAA8B,EAAEp6B,GAAG,EAAE,CAAC,CAAC,CAAC,OAAO9uC,EAAE,CAAC,GAAG,CAAC,CAACA,GAAGA,IAAI,KAAK,MAAM,MAAM,IAAI,MAAM,wBAAwB,KAAK,oCAAoCA,GAAG,EAAE,GAAG,KAAK,KAAK,IAAI,EAAE,OAAOyrB,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,OAAO,KAAK,YAAY,CAAC,EAAE,IAAIrsB,EAAE,CAAC,EAAE,QAAQK,EAAE,EAAEA,EAAE,KAAK,KAAK,EAAEA,IAAIL,EAAE,KAAKK,CAAC,EAAE,IAAI,EAAE,KAAK,SAASL,CAAC,EAAE,OAAO8pE,GAAG,KAAK,aAAa,EAAE,GAAG,MAAM,mDAAmD,KAAK,wCAAwC,EAAE,GAAG,QAAQ,EAAExtC,GAAG,EAAE,CAAC,CAAC,CAAC,QAAQ17B,EAAEZ,EAAE,CAAC,GAAGA,EAAE,QAAQ,KAAK,MAAM,MAAM,IAAI,MAAM,wBAAwB,KAAK,8BAA8BA,EAAE,OAAO,EAAE,GAAGY,EAAE,SAASZ,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,sDAAsDY,EAAE,cAAcZ,EAAE,MAAM,IAAI,EAAE,IAAI,EAAE,KAAK,IAAI,GAAGY,CAAC,EAAE,GAAG,CAAC,KAAK,aAAa,GAAG,KAAK,QAAQ,MAAM,IAAI,MAAM,mCAAmC,UAAU,KAAK,UAAU,EAAE,KAAK,UAAUA,EAAEmwC,GAAG/wC,EAAE,CAAC,CAAC,CAAC,CAAC,MAAMY,EAAEZ,EAAE,CAAC,GAAGA,EAAE,QAAQ,KAAK,MAAM,MAAM,IAAI,MAAM,wBAAwB,KAAK,8BAA8BA,EAAE,OAAO,EAAE,IAAI,EAAE,EAAEK,EAAEO,EAAE,IAAIoB,IAAI,GAAGA,EAAE,EAAE,EAAE,GAAG,IAAIhC,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM;AAAA;AAAA,UAE/pE,6BAA6BA,EAAE,OAAO,EAAE,GAAG,CAAC,KAAK,aAAaY,EAAE,SAAS,KAAK,QAAQ,MAAM,IAAI,MAAM,2DAA2D,KAAK,eAAeA,EAAE,sEAAsE,EAAE,IAAIiB,EAAE,IAAI,EAAE,EAAE7B,EAAE,KAAK,EAAE8B,EAAE,CAAC,EAAEyxB,EAAE,IAAI,CAACvzB,EAAEg8B,EAAEh8B,EAAE,CAAC,EAAE,EAAE6B,CAAC,CAAC,EAAE,QAAQG,EAAE,EAAEA,EAAEpB,EAAE,OAAO,EAAEoB,EAAE,CAAC,IAAI,EAAEA,IAAI,EAAE,EAAE3B,EAAE2B,EAAE,GAAGU,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE9B,EAAEoB,GAAGH,CAAC,EAAEC,EAAEE,GAAGg6B,EAAEU,GAAG18B,EAAE0C,EAAE,CAAC,EAAE,KAAK,YAAY,CAAC,CAAC,OAAOZ,CAAC,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEpB,EAAE,OAAOoB,IAAID,EAAEC,GAAGA,EAAE,KAAK,UAAUD,EAAED,CAAC,CAAC,CAAC,EAAM,GAAG,KAAK,CAAC,YAAYlB,EAAEZ,EAAE,EAAEK,EAAE,GAAG,CAAC,KAAK,QAAQO,EAAE,KAAK,aAAaZ,EAAE,KAAK,aAAa,EAAEY,GAAG,MAAMA,EAAE,QAAQiB,GAAG,CAAC,GAAG,IAAIA,EAAE,MAAM,MAAM,IAAI,MAAM,mCAAmC,wBAAwBA,EAAE,OAAO,EAAEioE,GAAG9pE,EAAE6B,EAAE,MAAM,6BAA6B,EAAE4xB,GAAG5xB,CAAC,CAAC,CAAC,EAAE,KAAK,SAAS8gC,GAAG,CAAC,EAAE,KAAK,eAAetiC,EAAEozB,GAAG,KAAK,QAAQ,CAAC,CAAC,IAAI,IAAI,CAAC,OAAO,KAAK,SAAS,EAAE,CAAC,MAAM,CAAC,OAAO,IAAI,GAAG,CAAC,GAAG,KAAK,OAAO,EAAE,KAAK,aAAa,KAAK,YAAY,CAAC,CAAC,cAAc7yB,EAAE,CAAC,KAAK,QAAQ,QAAQZ,GAAG,EAAEY,GAAG,MAAM,CAACA,EAAE,IAAIZ,EAAE,EAAE,IAAIA,EAAE,QAAQ,CAAC,CAAC,EAAE,KAAK,QAAQ,OAAO,EAAE,KAAK,SAAS,QAAQ,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,QAAQ,MAAM,CAAC,MAAMY,EAAEZ,EAAE,EAAE,GAAG,CAAC,GAAGA,IAAI,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmCA,wBAAwB,KAAK,cAAc,EAAE,GAAG,IAAI,IAAI,KAAK,QAAQ,SAAS,EAAE,MAAM,IAAI,MAAM,kCAAkC,kCAAkC,KAAK,QAAQ,kBAAkB,EAAE8pE,GAAGlpE,EAAE,KAAK,aAAa,6BAA6B,EAAE,IAAIP,EAAE2pE,GAAG,KAAK,aAAa,KAAK,QAAQppE,CAAC,EAAE,OAAO2yB,EAAE,IAAI,CAAC,IAAI1xB,EAAE,KAAK,QAAQ,IAAIC,GAAGk6B,EAAEl6B,EAAEzB,CAAC,CAAC,EAAE,OAAOqvC,GAAG7tC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQjB,EAAEZ,EAAE,CAAC,GAAGA,IAAI,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmCA,wBAAwB,KAAK,cAAc,EAAE,GAAG,KAAK,KAAK,IAAI,EAAE,MAAM,IAAI,MAAM,mCAAmC,EAAE,IAAI,EAAEgqE,GAAG,KAAK,aAAa,KAAK,QAAQppE,CAAC,EAAEP,EAAE,KAAK,QAAQ,IAAI,EAAE,OAAOA,EAAE,KAAK,GAAGypE,GAAGzpE,EAAE,MAAMO,EAAE,6BAA6B,EAAEo7B,EAAE37B,EAAE,CAAC,CAAC,CAAC,SAASO,EAAE,CAAC,GAAGA,EAAE,QAAQ,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmCA,EAAE,4BAA4B,KAAK,cAAc,EAAE,GAAGkpE,GAAGlpE,EAAE,MAAM,KAAK,aAAa,6BAA6B,EAAE,KAAK,iBAAiB,KAAK,KAAK,EAAE,MAAM,IAAI,MAAM,0CAA0C,EAAE6yB,GAAG7yB,CAAC,EAAE,KAAK,QAAQ,KAAKA,CAAC,CAAC,CAAC,OAAOA,EAAE,CAAC,GAAGA,EAAE,EAAE,MAAM,IAAI,MAAM,0DAA0DA,GAAG,EAAE,GAAG,KAAK,iBAAiB,IAAIA,EAAE,KAAK,eAAe,MAAM,IAAI,MAAM,+BAA+BA,8BAA8B,KAAK,iBAAiB,EAAE,IAAIZ,EAAE,IAAI,GAAG,CAAC,EAAE,KAAK,aAAa,KAAK,aAAa,KAAK,cAAc,EAAEA,EAAE,QAAQ,OAAOY,EAAE,QAAQ,EAAE,EAAE,EAAE,KAAK,IAAI,KAAK,QAAQ,OAAOA,CAAC,EAAE,EAAE,EAAEZ,EAAE,QAAQ,GAAG,KAAK,QAAQ,GAAG,OAAOA,CAAC,CAAC,QAAQY,EAAEZ,EAAE,EAAE,CAAC,GAAG,IAAI,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmC,wBAAwB,KAAK,cAAc,EAAE,GAAGY,EAAE,GAAGA,EAAE,KAAK,QAAQ,OAAO,MAAM,IAAI,MAAM,4BAA4BA,oBAAoB,KAAK,QAAQ,kBAAkB,EAAE,GAAG,KAAK,QAAQA,IAAI,KAAK,MAAM,IAAI,MAAM,oBAAoBA,YAAY,EAAEkpE,GAAG,KAAK,QAAQlpE,GAAG,MAAMZ,EAAE,6BAA6B,EAAE,IAAIK,EAAE2pE,GAAG,KAAK,aAAa,KAAK,QAAQhqE,CAAC,EAAE,OAAOg8B,EAAE,KAAK,QAAQp7B,GAAGP,CAAC,CAAC,CAAC,QAAQO,EAAEZ,EAAE,CAAC,GAAGA,EAAE,QAAQ,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmCA,EAAE,4BAA4B,KAAK,cAAc,EAAE,GAAGY,EAAE,GAAG,KAAK,iBAAiB,IAAIA,GAAG,KAAK,eAAe,MAAM,IAAI,MAAM,yBAAyBA,wBAAwB,KAAK,0BAA0B,EAAEkpE,GAAG,KAAK,aAAa9pE,EAAE,MAAM,6BAA6B,EAAEyzB,GAAGzzB,CAAC,EAAE,KAAK,QAAQY,IAAI,OAAO,KAAK,QAAQA,GAAG,KAAK,IAAI,KAAK,QAAQA,GAAGZ,CAAC,CAAC,OAAOY,EAAEZ,EAAE,EAAE,CAAC,GAAGA,IAAI,KAAK,aAAa,MAAM,IAAI,MAAM,mCAAmCA,wBAAwB,KAAK,cAAc,EAAE8pE,GAAG,KAAK,aAAa,EAAE,6BAA6B,EAAElpE,EAAEA,EAAE,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,IAAIP,EAAE2pE,GAAG,KAAK,aAAa,KAAK,QAAQ,CAAC,EAAE,OAAOppE,EAAE,SAAS,EAAEyrB,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,OAAOhsB,CAAC,CAAC,EAAEkzB,EAAE,IAAI,CAAC,IAAI1xB,EAAEjB,EAAE,IAAIkB,GAAGk6B,EAAE,KAAK,QAAQl6B,GAAGzB,CAAC,CAAC,EAAE,OAAOqvC,GAAG7tC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOjB,EAAEZ,EAAE,CAAC,GAAG,CAAC,CAACY,GAAGA,IAAI,KAAK,aAAa,MAAM,IAAI,MAAM,uBAAuB,KAAK,2CAA2CA,GAAG,EAAEkpE,GAAG,KAAK,aAAa9pE,EAAE,6BAA6B,EAAE,IAAI,EAAEgqE,GAAG,KAAK,aAAa,KAAK,QAAQhqE,CAAC,EAAE,OAAO,KAAK,KAAK,IAAI,EAAEqsB,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,OAAO,CAAC,CAAC,EAAEkH,EAAE,IAAI,CAAC,IAAIlzB,EAAE,KAAK,QAAQ,IAAIwB,GAAGm6B,EAAEn6B,EAAE,CAAC,CAAC,EAAE,OAAOy6B,GAAGj8B,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,SAAS8pE,GAAGrqE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAM,GAAGA,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,oDAAoDA,EAAE,OAAO,EAAE,GAAGA,EAAE,QAAQE,EAAE,MAAM,IAAI,MAAM,mCAAmCF,EAAE,4BAA4BE,GAAG,EAAE,IAAIK,EAAEP,EAAE,MAAM,MAAM,CAAC,EAAEgqE,GAAGzpE,EAAE,EAAE,6BAA6B,EAAE,IAAI,EAAE0wC,GAAGjxC,CAAC,EAAE,OAAO,IAAI,GAAG,EAAE,EAAEM,CAAC,CAAC,CAAC,SAASgqE,GAAGtqE,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAO,IAAI,GAAG,CAAC,EAAEN,EAAE,EAAEM,CAAC,CAAC,CAAC,SAASiqE,GAAGvqE,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,EAAE,SAASN,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,sDAAsD,EAAE,cAAcA,EAAE,MAAM,IAAI,EAAE,IAAIO,EAAE,KAAK,IAAI,GAAG,CAAC,EAAE,GAAGD,GAAG,MAAMA,IAAI,IAAIC,GAAGD,EAAE,MAAM,IAAI,MAAM,mCAAmCC,UAAUD,IAAI,EAAE,IAAI,EAAE,IAAI,GAAG,CAAC,EAAEJ,EAAEF,EAAE,MAAMM,CAAC,EAAE,EAAE2wC,GAAGjxC,EAAE,CAAC,EAAE,OAAO,EAAE,QAAQ,CAACiC,EAAEC,IAAI,CAAC,EAAE,QAAQD,EAAE,EAAEC,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,SAASsoE,GAAGxqE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,EAAEC,EAAE,EAAE,IAAIqC,IAAItC,GAAGsC,EAAEtC,EAAE,EAAE,GAAGA,IAAIN,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM;AAAA;AAAA,UAEn6JM,6BAA6BN,EAAE,OAAO,EAAE,IAAI,EAAEA,EAAE,MAAM,MAAM,CAAC,EAAE,EAAEmqE,GAAG,EAAEjqE,CAAC,EAAE+B,EAAE3B,IAAI,EAAE,EAAEN,EAAE,KAAKM,EAAE4B,EAAEuxB,EAAE,IAAI,CAAC,IAAI7wB,EAAE,CAAC,EAAE5C,EAAEk8B,EAAEl8B,EAAE,CAAC,EAAEM,EAAE2B,CAAC,CAAC,EAAE,QAAQY,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAED,IAAI,EAAE,EAAEtC,EAAEsC,EAAE,GAAGE,EAAE,CAAC,EAAED,EAAE,CAAC,EAAEE,EAAE,CAAC,EAAE,EAAEH,GAAGZ,CAAC,EAAEW,EAAEC,GAAGq5B,EAAEU,GAAG58B,EAAE+C,EAAEC,CAAC,EAAE,CAAC,CAAC,CAAC,OAAOhD,EAAE,QAAQ,EAAE4C,CAAC,CAAC,EAAET,EAAE,IAAI,GAAG,CAAC,EAAEjC,EAAEF,EAAE,MAAM,EAAE,MAAM,EAAE,QAAQ4C,EAAE,EAAEA,EAAEV,EAAE,OAAOU,IAAIT,EAAE,QAAQS,EAAEV,EAAEU,EAAE,EAAE,OAAOT,CAAC,CAAC,IAAIsoE,GAAG,MAAMzqE,EAAE,EAAEE,IAAI,CAAC,OAAOF,EAAE,GAAG,CAAC,IAAI,KAAK,IAAI,cAAc,CAAC,IAAIM,EAAEwlE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,OAAO,MAAM,EAAE,KAAK,GAAG,GAAGA,EAAE,YAAYI,GAAG,qBAAqB,EAAEJ,EAAE,eAAeA,EAAE,aAAa,EAAEA,EAAE,YAAYK,GAAG,qBAAqB,EAAEL,EAAE,eAAeA,EAAE,aAAa,CAAC,CAAC,IAAI,QAAQ,IAAI,iBAAiB,CAAC,IAAII,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE,MAAMA,EAAE,YAAYK,GAAG,qBAAqB,EAAEL,EAAE,eAAeA,EAAE,aAAa,EAAE+B,EAAE,EAAE,IAAIW,GAAGA,EAAE,EAAE,EAAEV,EAAE,MAAM,EAAE,GAAG,KAAK,EAAE,EAAE,QAAQU,GAAG,CAAC,CAACA,EAAE,MAAMX,EAAE,QAAQW,EAAE,EAAE,IAAI,IAAIA,EAAE,QAAQ,CAAC,CAAC,EAAE,IAAIT,EAAE,EAAE,KAAKD,EAAE,IAAI,CAAC,IAAIU,EAAET,EAAEA,EAAE,MAAMjC,EAAE,YAAYI,GAAG,qBAAqB6B,EAAEjC,EAAE,eAAeA,EAAE,aAAa,EAAE,IAAI2C,EAAEV,EAAE,IAAIY,GAAGA,EAAE,EAAE,EAAEH,EAAE,QAAQG,GAAG,CAAC,CAACA,EAAE,MAAMd,EAAE,QAAQc,EAAE,EAAE,IAAI,IAAIF,EAAE,QAAQE,EAAE,EAAE,IAAI,IAAIA,EAAE,QAAQ,CAAC,CAAC,EAAE,IAAID,EAAE,MAAM5C,EAAE,YAAYK,GAAG,qBAAqB4B,EAAEjC,EAAE,eAAeA,EAAE,aAAa,EAAEgC,EAAE,MAAMY,EAAE,GAAG,KAAK,EAAEA,EAAE,QAAQC,GAAG,CAAC,CAACA,EAAE,MAAMd,EAAE,QAAQc,EAAE,EAAE,IAAI,IAAIF,EAAE,QAAQE,EAAE,EAAE,IAAI,IAAIA,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,OAAOZ,CAAC,CAAC,IAAI,WAAW,CAAC,IAAI7B,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACmmE,GAAG/lE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIA,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,OAAOK,EAAE,OAAOA,EAAE8lE,GAAG9lE,CAAC,IAAI,MAAMD,EAAE,KAAK,GAAG,GAAG,CAAC,OAAOC,CAAC,EAAE,CAACA,EAAE,MAAM,CAAC,CAAC,IAAI,QAAQ,CAAC,IAAID,EAAEN,EAAE,WAAW,KAAKO,GAAGwlE,GAAGxlE,EAAE,EAAEL,CAAC,IAAI,MAAM,EAAE,GAAGI,EAAE,CAAC,IAAIC,EAAEwlE,GAAGzlE,EAAE,EAAEJ,CAAC,EAAE,MAAM,CAACmmE,GAAG9lE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,IAAI,QAAQ,CAAC,IAAID,EAAEwlE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,OAAOA,EAAE,WAAWI,CAAC,EAAE,CAAC+lE,GAAG9lE,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,IAAID,EAAEwlE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,OAAOA,EAAE,UAAU,EAAE,CAACmmE,GAAG/lE,CAAC,CAAC,CAAC,CAAC,IAAI,gBAAgB,CAAC,IAAIA,EAAEwlE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,OAAOA,EAAE,cAAc,EAAE,CAACmmE,GAAG/lE,CAAC,CAAC,CAAC,CAAC,IAAI,gBAAgB,CAAC,IAAIA,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,yBAAyB9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE0C,EAAE,IAAIwnE,GAAGjoE,EAAE5B,EAAED,EAAE,EAAE4B,EAAE,EAAED,CAAC,EAAE,OAAO/B,EAAE,eAAe0C,CAAC,EAAE,CAACA,EAAE,SAASigC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAIviC,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAEA,EAAE,eAAeI,EAAE,EAAE,EAAE,OAAO,EAAE,MAAMC,EAAE,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,eAAeI,EAAE,EAAE,EAAE,KAAKC,CAAC,CAAC,CAAC,CAAC,IAAI,sBAAsB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,eAAeI,EAAE,EAAE,EAAE,OAAOC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,uBAAuB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAEA,EAAE,eAAeI,EAAE,EAAE,EAAE,OAAO,EAAE,QAAQC,EAAE,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,sBAAsB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEL,EAAE,eAAeI,EAAE,EAAE,EAAE,EAAEwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACK,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAEA,EAAE,eAAeI,EAAE,EAAE,EAAE,OAAO,EAAE,MAAM,EAAEC,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEL,EAAE,eAAeI,EAAE,EAAE,EAAE,MAAM,CAACuiC,GAAGtiC,EAAE,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAID,EAAEwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEL,EAAE,eAAeI,EAAE,EAAE,EAAE,OAAOC,EAAE,cAAc,EAAE,CAACA,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAEA,EAAE,cAAcI,EAAE,EAAE,EAAE,OAAO,EAAE,QAAQC,EAAE,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,cAAcI,EAAE,EAAE,EAAE,QAAQC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,sBAAsB,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEsoE,GAAGhqE,EAAED,EAAE,EAAE,CAAC,EAAE,OAAOJ,EAAE,cAAc+B,CAAC,EAAE,CAACA,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,IAAI,kBAAkB,CAAC,IAAI3B,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAEF,EAAE,KAAK,oBAAoB,EAAE,cAAc,EAAE,iBAAiB,IAAI,EAAE8lE,EAAE,EAAE9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEjC,EAAE,KAAK,oBAAoB,GAAG,EAAEkC,EAAEooE,GAAGhqE,EAAEC,EAAE,EAAE0B,CAAC,EAAE,OAAO/B,EAAE,cAAcgC,CAAC,EAAE,CAACA,EAAE,QAAQ,CAAC,CAAC,IAAI,mBAAmB,CAAC,IAAI5B,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,cAAcI,EAAE,EAAE,EAAE,OAAOC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,cAAcI,EAAE,EAAE,EAAE,MAAMC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,uBAAuB,CAAC,IAAID,EAAEwlE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAEmqE,GAAG/pE,EAAEC,EAAE,CAAC,EAAE,OAAOL,EAAE,cAAc,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,mBAAmB,IAAI,qBAAqB,CAAC,IAAII,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEL,EAAE,cAAcI,EAAE,EAAE,EAAE,EAAEwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACK,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,qBAAqB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAEA,EAAE,cAAcI,EAAE,EAAE,EAAE,OAAO,EAAE,SAASC,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACA,EAAE,cAAcI,EAAE,EAAE,EAAE,QAAQC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,CAAC,IAAID,EAAEwlE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAEsqE,GAAGlqE,EAAE,EAAEC,CAAC,EAAE,OAAOL,EAAE,cAAc,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC,IAAI,mBAAmB,CAAC,IAAII,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEL,EAAE,cAAcI,EAAE,EAAE,EAAE,MAAM,CAACuiC,GAAGtiC,EAAE,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,mBAAmB,CAAC,IAAID,EAAEwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE8B,EAAE9B,EAAE,cAAcI,EAAE,EAAE,EAAE,OAAOC,CAAC,EAAE,OAAOL,EAAE,cAAc8B,CAAC,EAAE,CAACA,EAAE,QAAQ,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAahC,EAAE,uBAAuB,CAAC,CAAC,EAAE,SAAS0qE,GAAG1qE,EAAE,EAAEE,EAAE,CAAC,GAAG,CAACI,EAAEC,CAAC,EAAEulE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,EAAEI,IAAI,UAAU,EAAE,CAAC,EAAE2B,EAAE1B,IAAI,QAAQ2B,EAAE5B,IAAI,iBAAiB6B,EAAE2jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,GAAG,EAAE,CAAC,GAAG+B,GAAGE,IAAI,EAAE,MAAM,IAAI,MAAM,uGAAuG,EAAE,GAAG,CAACF,GAAG,GAAGE,IAAI,EAAE,MAAM,IAAI,MAAM,kFAAkF,CAAC,CAAC,GAAGD,EAAE,MAAM,IAAI,MAAM,sEAAsE,EAAE,IAAIU,EAAEkjE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE2C,EAAEujE,GAAGpmE,EAAE,EAAEE,CAAC,EAAE4C,EAAEgjE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE6C,EAAE+iE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,CAAC8C,EAAEkB,CAAC,EAAE4hE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,IAAIgE,EAAElB,EAAEA,EAAE,QAAQ,IAAI,EAAE8iE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,OAAO0C,EAAE,IAAIC,EAAE,WAAWC,EAAE,UAAUC,EAAE,QAAQC,EAAE,SAASkB,EAAE,eAAe3D,EAAE,eAAe,CAAC,CAAC,CAAC,IAAIoqE,GAAG,CAAC3qE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,SAAS,CAAC,IAAIO,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE+B,EAAE6jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAI1B,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAEkmE,GAAGpmE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE+B,EAAE6jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC0B,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,eAAe,CAAC,GAAG,CAAC,OAAO1B,EAAE,IAAI,EAAE,WAAW,EAAE,UAAU0B,EAAE,QAAQC,EAAE,SAASC,EAAE,eAAeS,EAAE,eAAeC,CAAC,EAAE6nE,GAAG1qE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,OAAO,CAAC,EAAEwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,OAAO4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,QAAQ,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,UAAU,CAAC0B,EAAE,GAAGA,EAAE,EAAE,EAAE,KAAKC,EAAE,WAAWU,EAAE,uBAAuBT,EAAE,eAAeU,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,6BAA6B,CAAC,GAAG,CAAC,OAAOtC,EAAE,IAAI,EAAE,WAAW,EAAE,UAAU0B,EAAE,QAAQC,EAAE,SAASC,EAAE,eAAeS,EAAE,eAAeC,CAAC,EAAE6nE,GAAG1qE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,gBAAgB,CAAC,EAAEwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,OAAO4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,QAAQ,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,UAAU,CAAC0B,EAAE,GAAGA,EAAE,EAAE,EAAE,KAAKC,EAAE,WAAWU,EAAE,uBAAuBT,EAAE,eAAeU,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,sBAAsB,IAAI,kBAAkB,CAAC,IAAItC,EAAEulE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAEkmE,GAAGpmE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,gBAAgBwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,wBAAwB,IAAI,kBAAkB,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAEkmE,GAAGpmE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE,MAAM,CAACI,EAAE,gBAAgBwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAE0B,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAI1B,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE+B,EAAE6jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,CAACK,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC0B,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,IAAI1B,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,oBAAoB,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,sBAAsB9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,OAAOgC,EAAE,QAAQC,CAAC,EAAE7B,EAAE,kBAAkBwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,CAACK,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAE0B,CAAC,EAAE,MAAM,CAACC,EAAEC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,IAAI5B,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,EAAE,CAACK,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,CAAC,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,EAAE,CAACK,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,aAAa,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE1B,EAAE,GAAG2B,EAAE3B,EAAE,GAAG4B,EAAE,EAAE,GAAGS,EAAE,EAAE,GAAG,MAAM,CAACtC,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,CAAC+B,EAAEC,CAAC,EAAE,EAAE,CAACC,EAAES,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAa5C,EAAE,uBAAuB,CAAC,CAAC,EAAM4qE,GAAG,CAAC5qE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,OAAO,CAAC,IAAIO,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,KAAKC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,CAAC,IAAIA,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,SAASC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,cAAc,CAAC,IAAIA,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,YAAYC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOC,EAAE,EAAE,EAAE0B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,MAAM,CAAC5B,EAAE,KAAKwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,MAAM,CAACI,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,uBAAuB,MAAM,CAACI,EAAE,qBAAqBwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,gBAAgB,MAAM,CAACI,EAAE,cAAcwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,CAAC,IAAIK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAMC,EAAE,EAAE,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,CAAC,IAAIK,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,gBAAgBC,EAAE,EAAE,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE+B,CAAC,CAAC,CAAC,CAAC,IAAI,QAAQ,MAAM,CAAC3B,EAAE,MAAMwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAE,SAAS6qE,GAAG7qE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAEK,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,MAAMI,EAAE,OAAOC,EAAE,cAAc,EAAE,aAAa,EAAE,eAAe0B,EAAE,aAAaC,CAAC,CAAC,CAAC,IAAI4oE,GAAG,MAAM9qE,EAAE,EAAEE,EAAEI,EAAEC,EAAEspE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,sBAAsB,CAAC,GAAG,CAAC,MAAM,EAAE,OAAO,EAAE,cAAciC,EAAE,aAAaC,EAAE,eAAeC,EAAE,aAAaS,CAAC,EAAEioE,GAAG7qE,EAAE,EAAEE,CAAC,EAAE2C,EAAE,MAAMtC,EAAE,MAAM,gCAAgC,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,CAAC,EAAE,MAAM,CAACC,EAAE,gBAAgBA,EAAE,cAAc,CAAC,CAAC,IAAI,sBAAsB,CAAC,GAAG,CAAC,MAAM,EAAE,OAAO,EAAE,cAAcZ,EAAE,aAAaC,EAAE,eAAeC,CAAC,EAAE0oE,GAAG7qE,EAAE,EAAEE,CAAC,EAAE0C,EAAEkjE,EAAE,qBAAqB9lE,EAAE,EAAEE,CAAC,EAAE2C,EAAE,MAAMtC,EAAE,MAAM,6BAA6B,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,CAAC,EAAE,MAAM,CAACC,EAAE,gBAAgBA,EAAE,YAAY,CAAC,CAAC,IAAI,sBAAsB,IAAI,sBAAsB,CAAC,GAAG,CAAC,MAAM,EAAE,OAAO,EAAE,cAAcZ,EAAE,aAAaC,EAAE,eAAeC,CAAC,EAAE0oE,GAAG7qE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,MAAMK,EAAE,MAAM,uBAAuB,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,QAAQ,CAAC,IAAI,EAAE5B,EAAE,KAAKulE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC,MAAMK,EAAE,WAAW,CAAC,CAAC,EAAE,OAAO,EAAE,QAAQ,EAAE,CAAC,CAAC,IAAI,WAAW,OAAOA,EAAE,eAAeulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAM+qE,GAAG,CAAC/qE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,aAAa,CAAC,IAAIO,EAAEulE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,WAAWC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIA,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE3B,EAAE,KAAKC,EAAE,EAAE,CAAC,EAAE,MAAM,CAAC0B,EAAE,OAAOA,EAAE,OAAO,CAAC,CAAC,IAAI,aAAa,CAAC,IAAI1B,EAAEulE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,WAAWC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIA,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAEI,EAAE,OAAOC,CAAC,EAAE,MAAM,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC,IAAI,WAAW,CAAC,IAAIA,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAEI,EAAE,OAAOC,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAaP,EAAE,uBAAuB,CAAC,CAAC,EAAMgrE,GAAG,CAAChrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,QAAQ,OAAO,EAAEA,EAAE,MAAM,IAAI,yBAAyB,IAAIO,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC6lE,GAAG/lE,EAAE,KAAK,EAAEE,CAAC,GAAGK,CAAC,EAAE,IAAI,cAAc,MAAM,CAACwlE,GAAG/lE,EAAE,KAAK,EAAEE,CAAC,CAAC,EAAE,IAAI,WAAW,IAAI,eAAe,IAAI,0BAA0B,CAAC,IAAI0C,EAAEkjE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACmmE,GAAGzjE,CAAC,CAAC,CAAC,CAAC,IAAI,YAAY,OAAOkjE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,IAAI0C,GAAGyjE,GAAGzjE,CAAC,CAAC,EAAE,IAAI,WAAW,IAAI,EAAEkjE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACmmE,GAAG,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAAC/lE,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,OAAO,CAAC,EAAE,IAAI,SAAS,OAAO4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,IAAI0C,GAAGtC,EAAE,SAASsC,EAAE,KAAK,CAAC,EAAE,IAAI,OAAO,MAAM,CAACtC,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,OAAO,CAAC,CAAC,EAAE,IAAI,QAAQ,IAAI,EAAEwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,QAAQ,KAAK,gGAAgG,EAAE,QAAQ,IAAIgC,CAAC,EAAE,QAAQU,EAAE,EAAEA,EAAEX,EAAE,OAAOW,IAAI,QAAQ,IAAI,MAAM,UAAU,MAAM,KAAKX,EAAEW,GAAG,SAAS,CAAC,EAAE,MAAM,EAAET,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAanC,EAAE,uBAAuB,CAAC,CAAC,EAAMirE,GAAG,KAAK,CAAC,YAAYnqE,EAAEZ,EAAE,CAAC,KAAK,SAASY,EAAE,KAAK,WAAWZ,EAAE,KAAK,OAAO2iC,GAAG,CAAC,EAAE,KAAK,UAAU,IAAI,IAAIlP,GAAG,KAAK,MAAM,CAAC,CAAC,IAAI,IAAI,CAAC,OAAO,KAAK,OAAO,EAAE,CAAC,eAAe,CAAC,KAAK,UAAU,QAAQ7yB,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,UAAU,MAAM,EAAE,KAAK,OAAO,QAAQ,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,UAAU,IAAI,CAAC,YAAY,CAAC,OAAO+hC,GAAG,KAAK,KAAK,EAAE,OAAO,CAAC,CAAC,MAAM,OAAO/hC,EAAEZ,EAAE,CAAC,KAAK,uBAAuBY,EAAEZ,CAAC,EAAE,IAAI,EAAE,MAAMY,EAAE,KAAK,EAAE,OAAO,KAAK,UAAU,QAAQP,GAAGA,EAAE,QAAQ,CAAC,EAAE,KAAK,UAAU,MAAM,EAAEkzB,EAAE,IAAI,CAAC,IAAIlzB,EAAE0wC,GAAG/wC,CAAC,EAAE6B,EAAE,EAAE,OAAOC,EAAEzB,EAAE,OAAOumB,EAAE,OAAO/kB,IAAIC,EAAE,IAAI,kDAAkDD,8BAA8BC,aAAa,EAAE,QAAQC,EAAE,EAAEA,EAAEF,EAAEE,IAAI,CAAC,IAAIC,EAAE,EAAED,GAAG,EAAE1B,EAAE0B,GAAG0xB,GAAG,CAAC,EAAE,KAAK,UAAU,IAAIzxB,EAAE,CAAC,CAAC,CAAC,OAAO,KAAK,MAAM,CAAC,CAAC,CAAC,MAAM,KAAKpB,EAAEZ,EAAE,CAAC,KAAK,uBAAuBY,EAAEZ,CAAC,EAAE,IAAI,EAAE,MAAMY,EAAE,KAAK,EAAE,OAAO2yB,EAAE,IAAI,CAAC,IAAIlzB,EAAE,CAAC,EAAE,QAAQwB,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,CAAC,IAAIC,EAAE,EAAED,GAAGE,EAAE,KAAK,gBAAgBD,EAAE9B,CAAC,EAAEK,EAAE,KAAK0B,CAAC,CAAC,CAAC,OAAO2tC,GAAGrvC,CAAC,CAAC,CAAC,CAAC,CAAC,gBAAgBO,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,UAAU,IAAIY,CAAC,EAAE,OAAO,GAAG,KAAK,EAAEZ,CAAC,CAAC,uBAAuBY,EAAEZ,EAAE,CAAC,GAAGY,EAAE,QAAQ,KAAK,SAAS,MAAM,IAAI,MAAM,oBAAoB,KAAK,qBAAqBA,EAAE,OAAO,EAAE,GAAGZ,EAAE,QAAQ,KAAK,WAAW,MAAM,IAAI,MAAM,sBAAsB,KAAK,uBAAuBA,EAAE,OAAO,CAAC,CAAC,EAAMgrE,GAAG,MAAMlrE,EAAE,EAAEE,EAAEI,IAAI,CAAC,OAAON,EAAE,GAAG,CAAC,IAAI,YAAY,IAAI,cAAc,CAAC,IAAIO,EAAED,EAAE,yBAAyBN,EAAE,IAAI,EAAE,GAAGO,GAAG,KAAK,MAAM,CAACA,CAAC,EAAE,CAAC,IAAI,EAAEulE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE,IAAIgpE,GAAG,EAAE,CAAC,EAAE,OAAO3qE,EAAE,aAAaN,EAAE,KAAKiC,CAAC,EAAE,CAACA,EAAE,MAAM,CAAC,CAAC,CAAC,IAAI,oBAAoB,IAAI,sBAAsB,CAAC,IAAI1B,EAAEulE,EAAE,cAAc9lE,EAAE,EAAEE,EAAEI,CAAC,EAAE,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,MAAMI,EAAE,iBAAiBC,EAAE,EAAE,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,IAAI,oBAAoB,CAAC,IAAIA,EAAEulE,EAAE,cAAc9lE,EAAE,EAAEE,EAAEI,CAAC,EAAE,EAAEwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAAC,MAAMI,EAAE,iBAAiBC,EAAE,EAAE,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,kBAAkB,IAAI,oBAAoB,CAAC,IAAIA,EAAEulE,EAAE,cAAc9lE,EAAE,EAAEE,EAAEI,CAAC,EAAE,MAAM,CAACA,EAAE,iBAAiBC,EAAE,EAAE,EAAE,WAAW,CAAC,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAaP,EAAE,uBAAuB,CAAC,CAAC,EAAMmrE,GAAG,CAACnrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,iBAAiB,CAAC,IAAIO,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,mBAAmB9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,eAAeC,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,IAAI,wBAAwB,CAAC,IAAI1B,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,mBAAmB9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,sBAAsBC,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,IAAI,gBAAgB,CAAC,IAAI1B,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,qBAAqB9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,cAAcC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,6BAA6B,CAAC,IAAI5B,EAAEulE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,UAAUC,EAAE,EAAE2B,EAAE,YAAY,EAAEC,EAAE,YAAY,EAAEF,EAAE,CAAC,CAAC,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAajC,EAAE,uBAAuB,CAAC,CAAC,EAAMorE,GAAG,CAACprE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,QAAQ,MAAM,CAACM,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,WAAW,MAAM,CAACI,EAAE,SAASwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,eAAe,MAAM,CAACI,EAAE,aAAawlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,aAAa,MAAM,CAACI,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,aAAa,MAAM,CAACI,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,SAAS,IAAI,WAAW,MAAM,CAACI,EAAE,MAAMwlE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAMqrE,GAAG,CAACrrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,cAAc,IAAI,gBAAgB,IAAI,SAAS,MAAM,CAACM,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,SAAS,MAAM,CAACI,EAAE,OAAOwlE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,GAAG4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,eAAe,GAAG,CAACK,EAAE,CAAC,EAAEulE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,EAAEK,IAAI,UAAU0B,EAAE,IAAI,QAAQC,EAAE4jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAE,GAAG,EAAE,CAAC,GAAG+B,GAAGC,IAAI,EAAE,MAAM,IAAI,MAAM,oFAAoF,EAAE,GAAG,CAACD,GAAGC,IAAI,EAAE,MAAM,IAAI,MAAM,+DAA+D,CAAC,CAAC,GAAG,CAACU,EAAEC,CAAC,EAAEijE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAM,OAAO,CAAC,EAAEwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,WAAW4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,WAAW4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,KAAK0C,EAAE,WAAW,EAAE,uBAAuBC,EAAE,eAAeV,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAanC,EAAE,uBAAuB,CAAC,CAAC,EAAMsrE,GAAG,CAACtrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,gBAAgB,MAAM,CAACM,EAAE,cAAcwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,iBAAiB,IAAI,mBAAmB,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,mBAAmB,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,MAAM,MAAM,CAACI,EAAE,2BAA2BwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,UAAU,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,aAAa,MAAM,CAACI,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,gBAAgB,MAAM,CAACI,EAAE,cAAcwlE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAMurE,GAAG,CAACvrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,MAAM,CAAC,IAAIiC,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIA,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,IAAIA,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,IAAID,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIF,EAAE6jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAEC,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,IAAI5B,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,SAASC,EAAE,EAAE,CAAC,CAAC,EAAE,IAAI,gBAAgB,CAAC,IAAI0B,EAAE6jE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE0C,EAAEkjE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,cAAc2B,EAAEC,EAAEC,EAAES,CAAC,CAAC,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAa5C,EAAE,uBAAuB,CAAC,CAAC,EAAMwrE,GAAG,CAACxrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,WAAW,IAAI,SAAS,CAAC,IAAIO,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,OAAO,EAAE,EAAE,MAAM,EAAEK,CAAC,EAAE,CAACD,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,SAAS,CAAC,IAAIC,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAOC,EAAED,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,CAAC,IAAIC,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,OAAO,EAAEA,EAAE,KAAK2B,EAAE,OAAO,EAAE1B,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,IAAIA,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQ+B,EAAE,EAAEA,EAAE1B,EAAE,OAAO0B,IAAI1B,EAAE0B,IAAI,EAAE,KAAKA,CAAC,EAAE,IAAI,EAAE6jE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,IAAIC,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQ,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,QAAQ,CAAC,IAAIA,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,eAAe,CAAC,IAAIA,EAAEulE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,MAAM9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAEgC,EAAE4jE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAEiC,EAAE2jE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE0C,EAAEkjE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE2C,EAAEijE,EAAE,iBAAiB9lE,EAAE,EAAEE,CAAC,EAAE4C,EAAEgjE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,aAAawC,EAAEvC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,CAAC,CAAC,CAAC,CAAC,IAAI,OAAO,OAAO4wB,EAAE,IAAI,CAAC,IAAIlzB,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE,EAAE,GAAG,MAAM+B,EAAE3B,EAAE,QAAQ,EAAE,EAAE,EAAE,MAAM4B,EAAE,EAAE,IAAIC,GAAG,CAAC,IAAIS,EAAEkkB,EAAE,YAAY3kB,EAAE,MAAM,CAAC,EAAE,GAAG,CAACS,GAAG,CAACkkB,EAAE,YAAYxmB,EAAE,QAAQ6B,CAAC,EAAE,MAAMF,CAAC,EAAE,MAAM,IAAI,MAAM,wCAAwC,EAAE,OAAOW,EAAET,EAAE7B,EAAE,QAAQ6B,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC7B,EAAE,MAAM4B,EAAE3B,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,SAAS,CAAC,IAAIA,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,OAAOI,EAAE,QAAQ,EAAEC,CAAC,CAAC,CAAC,IAAI,OAAO,CAAC,IAAIA,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,CAAC,CAAC,CAAC,CAAC,IAAI,QAAQ,IAAI,SAAS,CAAC,IAAIA,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,kBAAkB9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,OAAOI,EAAE,MAAM,EAAE,EAAEC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,IAAIA,EAAEulE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,UAAUC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,CAAC,IAAIA,EAAEulE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,SAASC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,gBAAgB,CAAC,IAAIA,EAAEulE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE+B,EAAE6jE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,cAAcC,EAAE,EAAE,EAAE,EAAE,QAAQ0B,EAAE,MAAMA,EAAE3B,EAAE,KAAK2B,EAAE,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,QAAQ,MAAM,UAAU,aAAajC,EAAE,uBAAuB,CAAC,CAAC,EAAMyrE,GAAG,CAACzrE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,sBAAsB,CAAC,GAAG,CAAC,cAAcO,EAAE,aAAa,EAAE,kBAAkB,EAAE,gBAAgB0B,CAAC,EAAE3B,EAAE,OAAO,oBAAoBwlE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,SAAS9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,CAAC,EAAE,MAAM,CAACK,EAAE,EAAE,EAAE0B,CAAC,CAAC,CAAC,IAAI,gBAAgB,CAAC,GAAG,CAAC,cAAc1B,EAAE,YAAY,CAAC,EAAED,EAAE,OAAO,cAAcwlE,EAAE,eAAe9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,CAAC,EAAE,MAAM,CAACK,EAAE,CAAC,CAAC,CAAC,IAAI,oBAAoB,MAAM,CAACD,EAAE,OAAO,kBAAkBwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,mBAAmB,MAAM,CAACI,EAAE,OAAO,iBAAiBwlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAM0rE,GAAG,CAAC1rE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,MAAM,MAAM,CAACM,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,OAAO,MAAM,CAACI,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,MAAM,CAACI,EAAE,MAAMwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAM2rE,GAAG,CAAC3rE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,eAAe,CAAC,GAAG,CAAC,OAAOO,EAAE,aAAa,CAAC,EAAED,EAAE,OAAO,aAAawlE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,cAAc9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,yBAAyB9lE,EAAE,EAAEE,CAAC,CAAC,EAAE,MAAM,CAACK,EAAE,CAAC,CAAC,CAAC,IAAI,cAAc,CAAC,GAAG,CAAC,QAAQA,EAAE,OAAO,EAAE,MAAM,CAAC,EAAED,EAAE,OAAO,YAAYwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,CAAC,EAAE,MAAM,CAACK,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,yBAAyB,MAAM,CAACD,EAAE,OAAO,uBAAuBwlE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAM4rE,GAAG,CAAC5rE,EAAE,EAAEE,EAAEI,EAAEupE,KAAK,CAAC,OAAO7pE,EAAE,GAAG,CAAC,IAAI,OAAO,MAAM,CAACM,EAAE,KAAKwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,aAAa,CAAC,IAAIK,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,WAAWwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC,IAAIA,EAAEulE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,CAAC,CAAC,CAAC,CAAC,IAAI,UAAU,MAAM,CAACD,EAAE,QAAQwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,YAAY,MAAM,CAACI,EAAE,UAAUwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,OAAO9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,QAAQ,IAAI,MAAM,MAAM,CAACI,EAAE,IAAIwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,UAAU9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,gBAAgB9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,iBAAiB,CAAC,IAAIK,EAAEulE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,WAAW9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,eAAewlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,iBAAiB,CAAC,IAAIA,EAAEulE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,EAAE,MAAM,CAACI,EAAE,eAAewlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,eAAe,CAAC,IAAIA,EAAEulE,EAAE,YAAY9lE,EAAE,EAAEE,CAAC,EAAE,EAAE4lE,EAAE,aAAa9lE,EAAE,EAAEE,CAAC,EAAE,YAAY,EAAE,MAAM,CAACI,EAAE,aAAawlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAEK,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,cAAc,MAAM,CAACD,EAAE,YAAYwlE,EAAE,IAAI9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,QAAQ9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,IAAI,gBAAgB,MAAM,CAACI,EAAE,cAAcwlE,EAAE,KAAK9lE,EAAE,EAAEE,CAAC,EAAE4lE,EAAE,KAAK9lE,EAAE,EAAEE,CAAC,CAAC,CAAC,EAAE,QAAQ,MAAM,UAAU,aAAaF,EAAE,uBAAuB,CAAC,CAAC,EAAE,SAAS6rE,GAAG7rE,EAAE,EAAEE,EAAEI,EAAEC,EAAEkzB,EAAE,CAAC,IAAI,GAAG,CAAC,EAAExxB,EAAEC,IAAI,CAAC,OAAO,EAAE,SAAS,CAAC,IAAI,aAAa,OAAO3B,EAAE,IAAIupE,GAAG,EAAE7nE,EAAEC,CAAC,CAAC,EAAE,IAAI,aAAa,OAAO3B,EAAE,IAAIwpE,GAAG,EAAE9nE,EAAEC,CAAC,CAAC,EAAE,IAAI,UAAU,OAAOuoE,GAAG,EAAExoE,EAAEC,CAAC,EAAE,IAAI,cAAc,OAAO3B,EAAE,IAAIoqE,GAAG,EAAE1oE,EAAEC,CAAC,CAAC,EAAE,IAAI,WAAW,OAAO3B,EAAE,IAAIqqE,GAAG,EAAE3oE,EAAEC,CAAC,CAAC,EAAE,IAAI,UAAU,OAAO4oE,GAAG,EAAE7oE,EAAEC,CAAC,EAAE,IAAI,aAAa,OAAO3B,EAAE,IAAIwqE,GAAG,EAAE9oE,EAAEC,CAAC,CAAC,EAAE,IAAI,QAAQ,OAAO3B,EAAE,IAAI4qE,GAAG,EAAElpE,EAAEC,CAAC,CAAC,EAAE,IAAI,QAAQ,OAAO3B,EAAE,IAAIyqE,GAAG,EAAE/oE,EAAEC,CAAC,CAAC,EAAE,IAAI,UAAU,OAAO3B,EAAE,IAAI6qE,GAAG,EAAEnpE,EAAEC,CAAC,CAAC,EAAE,IAAI,WAAW,OAAO3B,EAAE,IAAI8qE,GAAG,EAAEppE,EAAEC,CAAC,CAAC,EAAE,IAAI,gBAAgB,OAAO3B,EAAE,IAAI+qE,GAAG,EAAErpE,EAAEC,CAAC,CAAC,EAAE,IAAI,YAAY,OAAO3B,EAAE,IAAIgrE,GAAG,EAAEtpE,EAAEC,CAAC,CAAC,EAAE,IAAI,aAAa,OAAO3B,EAAE,IAAIirE,GAAG,EAAEvpE,EAAEC,CAAC,CAAC,EAAE,IAAI,SAAS,OAAO3B,EAAE,IAAIkrE,GAAG,EAAExpE,EAAEC,CAAC,CAAC,EAAE,IAAI,WAAW,OAAO3B,EAAE,IAAImrE,GAAG,EAAEzpE,EAAEC,CAAC,CAAC,EAAE,IAAI,SAAS,OAAO3B,EAAE,IAAIorE,GAAG,EAAE1pE,EAAEC,CAAC,CAAC,EAAE,IAAI,iBAAiB,OAAO3B,EAAE,IAAIqrE,GAAG,EAAE3pE,EAAEC,CAAC,CAAC,EAAE,IAAI,aAAa,OAAOgpE,GAAG,EAAEjpE,EAAEC,EAAE5B,CAAC,EAAE,IAAI,SAAS,IAAI6B,EAAEyjE,GAAG,EAAE,EAAE,EAAE,GAAGzjE,GAAGA,EAAE,eAAe,OAAOA,EAAE,eAAe,IAAIynE,GAAG,EAAE3nE,EAAEC,CAAC,CAAC,EAAE,MAAM,UAAU,aAAa,EAAE,uBAAuB,EAAE,QAAQ,MAAM,UAAU,eAAe,EAAE,uIAAuI,CAAC,CAAC,GAAGlC,EAAE,EAAEE,CAAC,EAAE,OAAO4mB,EAAE,UAAU,CAAC,EAAE,EAAE,KAAK,GAAG,CAAC,EAAE,OAAO,CAAC,CAAC,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,IAAIglD,GAAG,KAAK,CAAC,YAAYhrE,EAAE,CAAC,EAAEZ,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEK,EAAE,CAAC,EAAE,CAAC,KAAK,UAAUO,EAAE,KAAK,eAAeZ,EAAE,KAAK,cAAc,EAAE,KAAK,YAAYK,EAAE,KAAK,YAAY,CAAC,GAAG,EAAE,UAAU,GAAG,YAAY,CAAC,EAAE,KAAK,SAAS,CAAC,KAAK,WAAW,EAAE,KAAK,OAAO,EAAE,KAAK,0BAA0B,CAAC,CAAC,SAASO,EAAEZ,EAAE,CAAC,MAAM,CAAC,GAAGY,EAAE,UAAUZ,EAAE,YAAY,CAAC,CAAC,CAAC,IAAI,eAAeY,EAAE,CAAC,KAAK,WAAWA,IAAI,KAAK,SAASA,EAAE,KAAK,0BAA0B,EAAE,CAAC,IAAI,gBAAgB,CAAC,OAAO,KAAK,QAAQ,CAAC,IAAI,kBAAkB,CAAC,OAAO,KAAK,mBAAmB,EAAE,CAAC,IAAI,mBAAmB,CAAC,OAAO,KAAK,kBAAkB,CAAC,2BAA2B,CAAC,IAAIA,EAAE,CAAC,EAAE,QAAQZ,EAAE,EAAEA,EAAE,KAAK,SAAS,OAAO,EAAEA,IAAI,CAAC,IAAI,EAAE,KAAK,SAAS,MAAM,EAAE,KAAK,SAAS,OAAOA,CAAC,EAAEY,EAAE,KAAK,KAAK,qBAAqB,CAAC,CAAC,CAAC,CAACA,EAAE,KAAK,EAAE,EAAE,KAAK,mBAAmBA,CAAC,CAAC,qBAAqBA,EAAE,CAAC,OAAOA,EAAEA,EAAE,IAAIZ,GAAGA,EAAE,KAAK,GAAGA,EAAE,cAAc,EAAE,GAAG,GAAGA,EAAE,aAAaA,EAAE,aAAa,EAAE,KAAK,GAAG,EAAE,EAAE,CAAC,WAAWY,EAAE,CAAC,KAAK,WAAW,KAAK,SAAS,KAAK,SAAS,KAAK,SAAS,MAAM,EAAE,KAAK,SAAS,KAAK,KAAK,SAAS,KAAK,OAAOA,CAAC,CAAC,EAAE,KAAK,mBAAmB,QAAQ,KAAK,qBAAqB,KAAK,QAAQ,CAAC,EAAE,CAAC,WAAW,CAAC,GAAG,KAAK,UAAU,KAAK,SAAS,OAAO,EAAE,KAAK,SAAS,KAAK,SAAS,MAAM,EAAE,KAAK,SAAS,OAAO,EAAE,EAAE,KAAK,kBAAkB,MAAM,MAAO,OAAM,IAAI,MAAM,yCAAyC,CAAC,CAAC,eAAe,CAAC,GAAG,KAAK,UAAU,KAAK,SAAS,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,SAAS,MAAM,EAAE,KAAK,SAAS,IAAIA,EAAE,OAAO,OAAO,CAAC,EAAE,KAAK,SAAS,KAAK,SAAS,OAAO,EAAE,EAAEA,EAAE,aAAa,EAAEA,EAAE,GAAG,KAAK,OAAO,KAAK,SAAS,OAAO,GAAG,EAAEA,CAAC,EAAE,KAAK,mBAAmB,OAAO,EAAE,EAAE,KAAK,qBAAqB,KAAK,QAAQ,CAAC,CAAC,KAAM,OAAM,IAAI,MAAM,uDAAuD,CAAC,CAAC,UAAUA,EAAE,CAAC,OAAO,KAAK,UAAUA,EAAE,CAAC,eAAeA,EAAE,CAAC,KAAK,eAAeA,EAAE,IAAIA,CAAC,CAAC,eAAeA,EAAE,CAAC,OAAO,KAAK,eAAeA,EAAE,CAAC,cAAcA,EAAE,CAAC,KAAK,cAAcA,EAAE,IAAIA,CAAC,CAAC,cAAcA,EAAE,CAAC,OAAO,KAAK,cAAcA,EAAE,CAAC,QAAQA,EAAE,CAAC,QAAQZ,KAAK,KAAK,eAAe,KAAK,eAAeA,GAAG,cAAcY,CAAC,EAAE,QAAQZ,KAAK,KAAK,cAAc,KAAK,cAAcA,GAAG,cAAcY,CAAC,CAAC,CAAC,EAAE,SAASirE,GAAG/rE,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,IAAI,IAAI,EAAE,CAAC,EAAE,EAAE,KAAK0B,EAAE,KAAKC,EAAE,IAAI,IAAIC,EAAE,OAAO,KAAKnC,CAAC,EAAE,IAAI8C,GAAGkjE,GAAGljE,CAAC,EAAE,EAAE,EAAEF,EAAE,CAAC,EAAEtC,GAAG,OAAOsC,EAAEtC,EAAE,IAAIwC,GAAGkjE,GAAGljE,EAAE,IAAI,EAAE,EAAE,GAAG,IAAID,EAAE,CAAC,GAAG,CAAC,EAAE,KAAKA,EAAE,OAAO,GAAG,CAAC,IAAIC,EAAED,EAAE,IAAI,EAAE,IAAImpE,GAAGlpE,CAAC,GAAGmpE,GAAGnpE,CAAC,GAAGopE,GAAGppE,CAAC,IAAI,GAAG,OAAO,EAAEA,EAAEb,EAAE,EAAE,SAAS,IAAIc,GAAGA,EAAE,IAAI,EAAE,OAAOA,GAAGxC,EAAE,IAAIwC,CAAC,CAAC,GAAGxC,EAAE,IAAIuC,EAAE,IAAI,EAAE5C,EAAE4C,EAAE,OAAO,MAAMX,EAAE,QAAQW,EAAE,IAAI,IAAI,IAAIF,EAAE,QAAQE,EAAE,IAAI,IAAI,GAAG,CAAC,GAAGA,EAAE,OAAO,SAAS,EAAE,CAAC,EAAE,KAAKA,EAAE,IAAI,EAAE,QAAQ,CAACA,EAAE,OAAO,QAAQC,GAAG,CAACb,EAAE,IAAIa,EAAE,IAAI,IAAIb,EAAE,IAAIa,EAAE,IAAI,EAAEF,EAAE,KAAKE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO/C,EAAE,QAAQ,EAAE,UAAUO,EAAE,cAAc,EAAE,YAAY,EAAE,WAAW0B,CAAC,CAAC,CAAC,SAASkqE,GAAGnsE,EAAE,EAAEE,EAAE,CAAC,GAAG,CAAC,UAAUI,EAAE,OAAOC,CAAC,EAAEL,EAAE,EAAE,CAAC,EAAE,EAAE,OAAO,KAAKK,CAAC,EAAE,IAAIqC,GAAGojE,GAAGpjE,CAAC,EAAE,EAAE,EAAE,IAAIA,GAAG5C,EAAE,MAAM4C,EAAE,EAAEX,EAAEjC,EAAE,UAAU,EAAE,QAAQ4C,GAAG,CAACtC,EAAE,IAAIsC,EAAE,IAAI,GAAG,EAAE,KAAKA,CAAC,CAAC,CAAC,EAAE5C,EAAE,QAAQ,QAAQ4C,GAAG,CAACtC,EAAE,IAAIsC,EAAE,IAAI,GAAG,EAAE,KAAKA,CAAC,CAAC,CAAC,EAAEX,GAAG,MAAMA,EAAE,QAAQW,GAAG,CAACtC,EAAE,IAAIsC,EAAE,IAAI,GAAG,EAAE,KAAKA,CAAC,CAAC,CAAC,EAAE,IAAIV,EAAE,IAAI,IAAIC,EAAE,CAAC,EAAE,KAAK,EAAE,OAAO,GAAG,CAAC,IAAIS,EAAE,EAAE,IAAI,EAAEV,EAAE,IAAIU,EAAE,IAAI,EAAE,EAAEA,EAAE,OAAOT,EAAE,KAAKS,CAAC,EAAEA,EAAE,SAAS,QAAQC,GAAG,CAAC,CAACX,EAAE,IAAIW,EAAE,IAAI,GAAGvC,EAAE,IAAIuC,EAAE,IAAI,GAAGA,EAAE,OAAO,MAAMC,GAAGZ,EAAE,IAAIY,EAAE,IAAI,CAAC,GAAG,EAAE,KAAKD,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOV,CAAC,CAAC,IAAIiqE,GAAG,CAAC,SAAS,QAAQ,QAAQ,OAAO,gBAAgB,cAAc,iBAAiB,KAAK,OAAO,EAAEC,GAAG,CAAC,sBAAsB,sBAAsB,sBAAsB,OAAO,EAAEC,GAAG,CAAC,YAAY,cAAc,oBAAoB,sBAAsB,kBAAkB,oBAAoB,kBAAkB,mBAAmB,EAAE,SAASN,GAAGhsE,EAAE,CAAC,OAAOosE,GAAG,QAAQpsE,EAAE,EAAE,GAAG,CAAC,CAAC,SAASisE,GAAGjsE,EAAE,CAAC,OAAOqsE,GAAG,QAAQrsE,EAAE,EAAE,GAAG,CAAC,CAAC,SAASksE,GAAGlsE,EAAE,CAAC,OAAOssE,GAAG,QAAQtsE,EAAE,EAAE,GAAG,CAAC,CAAC,IAAIusE,GAAG,KAAK,CAAC,YAAYzrE,EAAEZ,EAAE,CAAC,KAAK,MAAMY,EAAE,KAAK,OAAOZ,EAAE,KAAK,YAAY,IAAI,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,UAAU,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,qBAAqB,CAAC,EAAE,KAAK,oBAAoB,CAAC,EAAE,KAAK,mBAAmB,GAAG,KAAK,SAASY,EAAE,QAAQ,KAAK,QAAQA,EAAE,OAAO,KAAK,WAAWA,EAAE,UAAU,KAAK,WAAWA,EAAE,UAAU,KAAK,WAAWA,EAAE,UAAUA,EAAE,WAAW,MAAM,OAAO,KAAKA,EAAE,SAAS,EAAE,QAAQ,GAAG,CAAC,KAAK,qBAAqB,GAAG,IAAIyrE,GAAGzrE,EAAE,UAAU,GAAG,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,KAAK,UAAU,CAAC,IAAI,qBAAqB,CAAC,OAAO,KAAK,OAAO,KAAK,OAAO,oBAAoB,KAAK,oBAAoB,CAAC,IAAI,WAAW,CAAC,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,KAAK,UAAU,CAAC,IAAI,UAAUA,EAAE,CAAC,IAAIZ,EAAE,OAAO,KAAKY,CAAC,EAAE,IAAI,GAAGA,EAAE,GAAG,IAAIP,GAAGA,EAAE,EAAE,CAAC,EAAE,KAAK,WAAW,CAAC,EAAE,OAAO,GAAGL,CAAC,EAAE,KAAK,WAAWY,CAAC,CAAC,IAAI,gBAAgBA,EAAE,CAAC,KAAK,iBAAiBA,CAAC,CAAC,IAAI,QAAQ,CAAC,OAAO,KAAK,QAAQ,IAAIA,IAAI,CAAC,KAAKA,EAAE,KAAK,MAAMA,EAAE,WAAW,MAAMA,EAAE,WAAW,MAAM,MAAM,OAAO,MAAMA,EAAE,WAAW,MAAMA,EAAE,WAAW,MAAM,MAAM,MAAM,EAAE,CAAC,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,SAAS,IAAIA,IAAI,CAAC,KAAKA,EAAE,KAAK,MAAMA,EAAE,WAAW,MAAMA,EAAE,WAAW,MAAM,MAAM,OAAO,MAAMA,EAAE,WAAW,MAAMA,EAAE,WAAW,MAAM,MAAM,MAAM,EAAE,CAAC,CAAC,IAAI,YAAY,CAAC,OAAO,KAAK,QAAQ,IAAIA,GAAGA,EAAE,cAAcA,EAAE,IAAI,CAAC,CAAC,IAAI,aAAa,CAAC,OAAO,KAAK,SAAS,IAAIA,GAAG,CAAC,IAAIZ,EAAEY,EAAE,cAAcA,EAAE,KAAK,OAAOA,EAAE,cAAc,GAAGZ,KAAKY,EAAE,gBAAgBZ,CAAC,CAAC,CAAC,CAAC,IAAI,WAAW,CAAC,OAAO,OAAO,KAAK,KAAK,UAAU,EAAE,OAAO,CAACY,EAAEZ,KAAKY,EAAEZ,GAAG,KAAK,WAAWA,GAAG,UAAUY,GAAG,CAAC,CAAC,CAAC,CAAC,kBAAkBA,EAAEZ,EAAE,CAAC,IAAI,EAAEY,EAAE,IAAIiB,GAAGA,EAAE,IAAI,EAAE,KAAK,EAAExB,EAAEL,EAAE,IAAI6B,GAAGA,EAAE,IAAI,EAAE,KAAK,EAAE,OAAO,EAAE,KAAK,KAAK,SAAS,EAAE,KAAKxB,EAAE,KAAK,KAAK,SAAS,CAAC,CAAC,QAAQO,EAAEZ,EAAE,CAAC,IAAI,EAAE6rE,GAAGjrE,EAAEZ,EAAE,KAAK,UAAU,KAAK,UAAU,EAAE,CAAC,cAAcK,EAAE,YAAYwB,EAAE,WAAWC,CAAC,EAAE,EAAE,GAAGD,GAAG,KAAK,MAAM,IAAI,MAAM,qCAAqCA,EAAE,oCAAoCA,EAAE,8GAA8GC,IAAI,EAAE,GAAGzB,EAAE,OAAO,EAAE,CAAC,IAAI0B,EAAE/B,EAAE,IAAI,GAAG,EAAE,IAAI,EAAEgC,EAAE,OAAO,KAAKpB,CAAC,EAAE,MAAM,IAAI,MAAM,+BAA+BmB,gCAAgCC,sCAAsC3B,IAAI,CAAC,CAAC,OAAO4rE,GAAG,KAAK,MAAM,KAAK,UAAU,CAAC,CAAC,CAAC,QAAQrrE,EAAEZ,EAAE,CAACY,EAAE,KAAK,UAAUA,CAAC,EAAE,IAAI,EAAE,OAAO,KAAKA,CAAC,EAAE,KAAK,EAAE,KAAK,YAAYA,CAAC,EAAE,KAAK,uBAAuBA,CAAC,EAAEZ,EAAE,KAAK,WAAWA,CAAC,EAAE,KAAK,aAAaA,CAAC,EAAE,IAAIK,EAAE,EAAE,IAAI,GAAG,KAAK,MAAM,MAAMylE,GAAG,CAAC,EAAE,GAAG,EAAEjkE,EAAE7B,EAAE,IAAI,GAAG8lE,GAAG,CAAC,EAAE,EAAE,EAAEhkE,EAAED,EAAE,IAAI,GAAG,KAAK,MAAM,MAAM,EAAE,EAAE,KAAK,yBAAyB,EAAEC,EAAE,SAAS,IAAIA,EAAE,KAAK,UAAU,IAAIC,EAAE,KAAK,kBAAkB1B,EAAEyB,CAAC,EAAEE,EAAE,KAAK,YAAY,IAAID,CAAC,EAAEC,GAAG,OAAOA,EAAE,KAAK,QAAQpB,EAAEkB,CAAC,EAAE,KAAK,YAAY,IAAIC,EAAEC,CAAC,GAAG,IAAI,EAAE,CAAC,EAAEU,EAAE,CAAC,EAAE,OAAO6wB,EAAE,IAAI,CAAC,IAAI,EAAE,IAAIq4C,GAAG,KAAK,UAAU,EAAElpE,EAAE,KAAK,mBAAmB,EAAEE,EAAE,OAAO,OAAO,CAAC,EAAE,KAAK,SAAS,EAAE,OAAO,KAAKhC,CAAC,EAAE,QAAQoD,GAAG,CAAC,GAAG,CAACC,EAAEoD,CAAC,EAAEy+D,GAAG9hE,CAAC,EAAEa,EAAE,CAAC,EAAEA,EAAEwC,GAAGzG,EAAEoD,GAAGpB,EAAEqB,GAAGY,CAAC,CAAC,EAAE,IAAIhC,EAAE,KAAK,mBAAmBD,CAAC,EAAEE,EAAE,CAAC,EAAE,QAAQkB,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,IAAI,CAAC,IAAIC,EAAEjC,EAAEgC,GAAG,GAAG,CAACpB,EAAEqB,EAAE,MAAM,CAAC,IAAIoD,EAAEskE,GAAG1nE,EAAErB,EAAE,EAAE,KAAK,gBAAgB,EAAE,GAAGgkB,EAAE,UAAUvf,CAAC,EAAE,MAAM,IAAI,MAAM,4BAA4BpD,EAAE,kEAAkE,EAAErB,EAAEqB,EAAE,MAAMoD,EAAE,KAAK,uBAAuBpD,EAAE,KAAKA,EAAErB,EAAE,EAAEC,EAAEhB,EAAEiB,CAAC,CAAC,CAAC,CAAC,OAAO,KAAK,QAAQ,MAAM,EAAE,QAAQD,CAAC,EAAE7C,EAAE,IAAIgE,GAAG6hE,GAAG7hE,EAAEpB,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,mBAAmBhC,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,OAAO,MAAM,CAAC,EAAE,OAAO,KAAKY,CAAC,EAAE,IAAI,GAAGA,EAAE,EAAE,EAAE,IAAI,GAAG,EAAE,IAAIP,GAAGA,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO,IAAI,IAAIL,CAAC,CAAC,CAAC,uBAAuBY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAE,CAAC/B,EAAE,WAAW,WAAW8B,EAAE,QAAQlB,CAAC,IAAI,KAAK,EAAEA,GAAG,QAAQoB,GAAG,CAACA,GAAG,OAAOD,EAAEC,EAAE,KAAKD,EAAEC,EAAE,KAAK,GAAGhC,EAAE,SAAS,OAAO,CAAC,EAAEA,EAAE,OAAO,QAAQgC,GAAG,CAAC,GAAGA,EAAE,WAAW,UAAU,CAAC,IAAI,EAAEgkE,GAAGhkE,EAAE,KAAK,EAAE3B,CAAC,EAAE,GAAG,MAAM,EAAE,QAAQqC,GAAG,CAAC,GAAGA,GAAG,CAACA,EAAE,MAAM,CAACb,EAAE,IAAIa,EAAE,EAAE,EAAE,CAAC,IAAI,EAAEX,EAAEW,EAAE,IAAI,GAAG,IAAI,EAAE,CAAC,GAAG,CAAC,KAAK,mBAAmBA,EAAE,QAAQ,MAAM,CAAC,GAAG,CAACE,EAAEC,CAAC,EAAEojE,GAAGjmE,EAAE,KAAKK,CAAC,EAAE,KAAK,oBAAoBuC,GAAG,KAAK,oBAAoBA,GAAGC,GAAGH,GAAG,KAAK,oBAAoBE,GAAG,CAAC,EAAE,KAAK,oBAAoBA,GAAGC,GAAGH,EAAE,CAAC,OAAOX,EAAEW,EAAE,GAAG,MAAM,GAAG,MAAMX,EAAEW,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,MAAM,aAAa9B,EAAEZ,EAAE,CAAC,OAAO,KAAK,cAAcY,EAAEZ,CAAC,CAAC,CAAC,4BAA4B,CAAC,CAAC,KAAK,sBAAsB,OAAO,KAAK,KAAK,mBAAmB,EAAE,QAAQY,GAAG,KAAK,oBAAoBA,GAAG,QAAQZ,GAAGA,EAAE,QAAQ,CAAC,CAAC,EAAE,KAAK,kBAAkB,EAAE,CAAC,mBAAmB,CAAC,CAAC,KAAK,YAAY,OAAO,KAAK,KAAK,UAAU,EAAE,QAAQY,GAAG,CAAC,KAAK,WAAWA,GAAG,QAAQR,GAAG,CAACA,GAAG,CAACA,EAAE,MAAM,CAACA,EAAE,YAAY,CAAC,KAAK,QAAQ,IAAIA,EAAE,EAAE,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,wBAAwB,CAAC,OAAO,KAAK,UAAU,CAAC,0BAA0B,CAAC,QAAQQ,KAAK,KAAK,oBAAoB,KAAK,oBAAoBA,GAAG,QAAQZ,GAAGA,EAAE,QAAQ,CAAC,EAAE,OAAO,KAAK,oBAAoBY,EAAE,CAAC,MAAM,cAAcA,EAAEZ,EAAE,EAAE,GAAGK,EAAE,CAAC,EAAEwB,EAAE,CAAC,EAAE,CAAC,IAAIjB,EAAE,KAAK,UAAUA,CAAC,EAAE,KAAK,YAAYA,CAAC,EAAE,KAAK,uBAAuBA,CAAC,EAAEZ,EAAE,KAAK,WAAWA,CAAC,EAAE,KAAK,aAAaA,CAAC,GAAG,GAAG,CAAC,KAAK,mBAAmBua,EAAE,EAAE,QAAQ,2BAA2B,CAAC,OAAO7X,EAAN,CAAS,QAAQ,KAAKA,EAAE,OAAO,CAAC,CAAC,KAAK,yBAAyB,EAAE,IAAIZ,EAAE,IAAI8pE,GAAG,KAAK,UAAUvrE,EAAEwB,EAAE,KAAK,mBAAmB,EAAE,KAAK,WAAW,MAAM,KAAK,uBAAuBjB,EAAEkB,EAAE9B,EAAE,CAAC,EAAE,IAAI+B,EAAE/B,EAAE,IAAI0C,GAAGmjE,GAAGnjE,EAAE,KAAK,WAAWZ,CAAC,CAAC,EAAEE,EAAED,EAAE,IAAIW,GAAGA,EAAE,EAAE,EAAE,EAAE,OAAO,KAAK9B,CAAC,EAAE,IAAI8B,GAAG9B,EAAE8B,GAAG,EAAE,EAAE,OAAO,KAAK,QAAQ,IAAI,IAAI,CAAC,GAAGV,EAAE,GAAG,EAAE,GAAG,KAAK,SAAS,CAAC,EAAE,KAAK,oBAAoB,KAAK,kBAAkB,EAAE,KAAK,QAAQ,MAAMF,EAAE,QAAQ,KAAK,OAAO,EAAEC,CAAC,CAAC,MAAM,qBAAqBnB,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAEO,EAAE,OAAO,CAACiB,EAAEC,EAAEC,KAAKF,EAAE,KAAK,OAAOE,GAAG,MAAMD,EAAED,GAAG,CAAC,CAAC,EAAE,OAAO,KAAK,cAAcxB,EAAE,KAAK,YAAY,GAAGL,EAAE,CAAC,CAAC,CAAC,MAAM,uBAAuBY,EAAEZ,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAE,OAAO,KAAKjB,CAAC,EAAEkB,EAAED,EAAE,IAAI4C,GAAG,KAAK,MAAM,MAAMqhE,GAAGrhE,CAAC,EAAE,GAAG,EAAE1C,EAAE,EAAE,IAAI0C,GAAGqhE,GAAGrhE,CAAC,EAAE,EAAE,EAAEzC,EAAED,EAAE,IAAI0C,GAAG,KAAK,MAAM,MAAMA,EAAE,EAAEzC,EAAE,SAAS,IAAIA,EAAE,KAAK,UAAU,GAAG,CAAC,UAAU,EAAE,cAAcU,EAAE,YAAY,EAAE,WAAWE,CAAC,EAAEipE,GAAGjrE,EAAEoB,EAAE,KAAK,UAAU,KAAK,UAAU,EAAEa,EAAE,CAAC,GAAGf,EAAE,GAAG,KAAK,MAAM,QAAQ,GAAG,KAAK,YAAY,CAAC,CAAC,EAAE,IAAI2C,IAAI,CAAC,KAAKA,EAAE,SAASzE,EAAE,cAAc,EAAE,EAAE8C,EAAE,OAAO,OAAO,CAAC,EAAE,KAAK,SAAS,EAAE,OAAO,KAAKlC,CAAC,EAAE,QAAQ6D,GAAG,CAAC,GAAG,CAACC,EAAEC,CAAC,EAAEmhE,GAAGrhE,CAAC,EAAEG,EAAE,CAAC,EAAEA,EAAED,GAAG/D,EAAE6D,GAAG3B,EAAE4B,GAAGE,CAAC,CAAC,EAAE,IAAIZ,EAAE,CAAC,EAAEC,EAAE,KAAK,mBAAmBnB,CAAC,EAAEuE,EAAE,CAAC,EAAE,KAAKxE,EAAE,OAAO,GAAG,CAAC,IAAI4B,EAAE,KAAK,aAAa3C,EAAEe,EAAE7C,EAAE8C,EAAEuE,EAAEpD,EAAElC,EAAEiC,EAAE,CAAC,EAAE,MAAM,QAAQ,IAAIS,CAAC,CAAC,CAAC,GAAG,MAAM,CAACpE,GAAG,QAAQ,KAAK,iIAAiI,EAAE,IAAIwE,EAAE7C,EAAE,OAAOyC,GAAG,CAACqnE,GAAGrnE,CAAC,GAAG,CAACohE,GAAGphE,EAAE,KAAK3B,EAAE9C,CAAC,CAAC,EAAE,IAAIyE,GAAGA,EAAE,IAAI,EAAE,GAAGI,EAAE,OAAO,EAAE,CAAC,IAAIJ,EAAE,GAAG,MAAM,GAAG,OAAOA,EAAE,wFAAwF7B,MAAM,IAAI,MAAM,+BAA+BiC,gCAAgChD,iDAAiDa,OAAO+B,GAAG,CAAC,CAAC,OAAO3B,CAAC,CAAC,aAAalC,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAEC,EAAE,EAAE,CAAC,IAAIU,EAAE,CAAC,EAAE,KAAK1C,EAAE,OAAO,GAAG,CAAC,IAAI,EAAEA,EAAE,IAAI,EAAE,EAAE,eAAe,EAAE,SAAS,IAAI4C,EAAE,GAAG,GAAG,EAAE,KAAK,KAAK,SAASgjE,EAAE,aAAa,EAAE,KAAKvlE,EAAE,CAAC,IAAI,CAACuC,CAAC,EAAEqjE,GAAG,EAAE,KAAK,KAAK,CAAC,GAAG5lE,EAAE,EAAE,KAAK,OAAO,KAAK,CAAC,IAAIwC,EAAE8oE,GAAG,EAAE,KAAKtrE,EAAE,EAAE,KAAK,gBAAgB,EAAEuC,IAAI,CAACA,CAAC,EAAEqjE,GAAG,EAAE,KAAK,KAAK,CAAC,GAAG,IAAInjE,EAAE,EAAE,eAAe8jB,EAAE,UAAU/jB,CAAC,EAAEH,EAAE,KAAKG,EAAE,KAAKmB,IAAI3D,EAAEuC,GAAGoB,EAAE,EAAE,eAAelB,EAAE,KAAK,uBAAuBF,EAAE,EAAE,KAAKvC,EAAE,EAAEyB,EAAEC,EAAEC,CAAC,EAAE,KAAK,kBAAkB,EAAE,KAAKhC,EAAE,EAAEK,EAAEwB,EAAE,CAAC,EAAEmC,EAAE,CAAC,GAAG3D,EAAEuC,GAAGC,EAAE,KAAK,uBAAuBD,EAAE,EAAE,KAAKvC,EAAE,EAAEyB,EAAEC,EAAEC,CAAC,EAAE,KAAK,kBAAkB,EAAE,KAAKhC,EAAE,EAAEK,EAAEwB,EAAE,CAAC,EAAE,MAAM,KAAK,kBAAkB,EAAE,KAAK7B,EAAE,EAAEK,EAAEwB,EAAE,CAAC,CAAC,CAAC,OAAOa,CAAC,CAAC,kBAAkB9B,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAClB,EAAE,SAAS,QAAQmB,GAAG,CAAC,GAAG,CAACC,CAAC,EAAEikE,GAAGlkE,EAAE,KAAK,CAAC,EAAEF,EAAEG,IAAI,CAACF,EAAE,IAAIC,EAAE,IAAI,IAAIA,EAAE,KAAK,QAAQA,EAAE,WAAW,KAAK,GAAG,CAAC,CAAC8jE,GAAG,EAAExlE,EAAE,CAAC,CAAC,IAAIwB,EAAEG,GAAG,GAAGhC,EAAE,KAAK,CAAC,SAAS,EAAE,eAAe,KAAK+B,CAAC,CAAC,GAAGA,EAAE,WAAW,MAAM,GAAG,CAAC,CAAC8jE,GAAG,EAAExlE,EAAE,CAAC,CAAC,IAAIwB,EAAEG,GAAG,GAAGhC,EAAE,KAAK,CAAC,SAAS,EAAE,eAAe,KAAK+B,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,OAAO,KAAK,KAAK,SAAS,EAAE,QAAQnB,GAAG,KAAK,UAAUA,GAAG,QAAQZ,GAAGA,EAAE,QAAQ,CAAC,CAAC,CAAC,CAAC,uBAAuBY,EAAE,CAAC,OAAO,KAAKA,CAAC,EAAE,QAAQZ,GAAG,CAAC,IAAI,EAAEY,EAAEZ,GAAG,CAACK,CAAC,EAAEylE,GAAG9lE,CAAC,EAAE6B,EAAE,KAAK,MAAM,MAAMxB,GAAG,GAAGwB,EAAE,WAAW,OAAOA,EAAE,WAAW,MAAM,MAAM,CAAC,IAAIC,EAAED,EAAE,WAAW,MAAM,MAAME,EAAED,EAAE,SAAS,EAAE,MAAM,QAAQ,EAAE,MAAM,MAAM,CAACE,EAAE,IAAIF,EAAE,KAAK,IAAIA,EAAE,KAAKE,CAAC,EAAE4kB,EAAE,OAAO7kB,EAAE,IAAI,sBAAsBF,EAAE,mDAAmDC,gBAAgB,EAAE,QAAQ,CAAC,CAACD,EAAE,WAAW,OAAOA,EAAE,WAAW,MAAM,OAAO+kB,EAAE,OAAO,EAAE,QAAQ/kB,EAAE,WAAW,MAAM,MAAM,IAAI,sBAAsBA,EAAE,kDAAkDA,EAAE,WAAW,MAAM,kBAAkB,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,UAAUjB,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,QAAQ,KAAKY,EAAE,GAAG,KAAK,YAAY,MAAM,KAAK,WAAW,QAAQ,MAAM,KAAK,WAAW,OAAO,IAAI,KAAK,CAAC,IAAIP,EAAE,KAAK,WAAW,OAAO,GAAGL,EAAEK,EAAE,MAAMO,EAAE,EAAE,MAAMZ,EAAE,GAAGY,EAAE,GAAG,OAAOZ,CAAC,CAAC,YAAYY,EAAE,CAAC,IAAIZ,EAAE,OAAO,KAAKY,CAAC,EAAE,OAAO,GAAG,CAAC,GAAG,CAACP,CAAC,EAAEylE,GAAG,CAAC,EAAE,OAAO,KAAK,MAAM,MAAMzlE,IAAI,IAAI,CAAC,EAAE,GAAGL,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,uDAAuDA,+BAA+B,CAAC,CAAC,WAAWY,EAAE,CAAC,OAAOA,EAAE,IAAIZ,GAAG,KAAK,YAAY,MAAM,KAAK,WAAW,SAAS,MAAM,KAAK,WAAW,QAAQA,IAAI,KAAK,KAAK,WAAW,QAAQA,GAAG,KAAKA,EAAE,CAAC,CAAC,CAAC,CAAC,aAAaY,EAAE,CAACA,EAAE,QAAQZ,GAAG,CAAC,GAAG,CAAC,CAAC,EAAE8lE,GAAG9lE,CAAC,EAAE,GAAG,CAAC,KAAK,MAAM,MAAM,GAAG,MAAM,IAAI,MAAM,eAAeA,8BAA8B,CAAC,CAAC,CAAC,CAAC,EAAMssE,GAAG,KAAK,CAAC,YAAY1rE,EAAE,CAAC,EAAEZ,EAAE,CAAC,EAAE,CAAC,KAAK,sBAAsBY,EAAE,KAAK,aAAaZ,CAAC,CAAC,aAAaY,EAAEZ,EAAE,CAAC,KAAK,sBAAsBY,GAAGZ,EAAE,OAAO,KAAK,aAAaA,EAAE,IAAIA,CAAC,CAAC,yBAAyBY,EAAE,CAAC,OAAO,KAAK,sBAAsBA,EAAE,CAAC,iBAAiBA,EAAE,CAAC,OAAO,KAAK,aAAaA,EAAE,CAAC,SAAS,CAAC,QAAQA,KAAK,KAAK,aAAa,KAAK,aAAaA,GAAG,cAAc,EAAE,OAAO,KAAK,aAAaA,GAAG,QAAQA,KAAK,KAAK,sBAAsB,KAAK,sBAAsBA,GAAG,QAAQ,EAAE,OAAO,KAAK,sBAAsBA,EAAE,CAAC,EAAM2rE,GAAG,oBAAoBC,GAAG,aAAaC,GAAG,KAAK,CAAC,YAAY7rE,EAAEZ,EAAE,CAAC,EAAE,EAAE4wB,GAAG,CAAC,KAAK,SAAShwB,EAAE,KAAK,YAAYZ,EAAE,KAAK,QAAQ,MAAM,KAAK,GAAG,EAAEA,GAAG,OAAO,KAAK,YAAY,CAAC,GAAG,KAAK,gBAAgB,IAAIssE,EAAE,CAAC,IAAI,cAAc,CAAC,OAAO,KAAK,OAAO,CAAC,IAAI,YAAY,CAAC,OAAO,KAAK,SAAS,UAAU,CAAC,IAAI,aAAa,CAAC,OAAO,KAAK,SAAS,WAAW,CAAC,IAAI,QAAQ,CAAC,OAAO,KAAK,SAAS,MAAM,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,SAAS,OAAO,CAAC,IAAI,SAAS,CAAC,OAAO,KAAK,SAAS,SAAS,CAAC,IAAI,UAAU,CAAC,OAAO,KAAK,UAAU,mBAAmB,CAAC,IAAI,gBAAgB,CAAC,OAAO,KAAK,SAAS,CAAC,IAAI,2BAA2B,CAAC,OAAO,KAAK,oBAAoB,CAAC,eAAe,CAAC,IAAI1rE,EAAE,KAAK,SAAS,GAAGA,EAAE,MAAM,KAAK,KAAK,QAAQA,UAAU,KAAK,YAAY,aAAa,KAAK,KAAK,QAAQ,KAAK,GAAG,mBAAmBA,EAAE,KAAK,WAAW,MAAM,CAAC,IAAIZ,EAAE,KAAK,GAAG,gBAAgBY,EAAE,KAAK,WAAW,EAAE,GAAGZ,EAAE,SAAS,EAAEA,EAAE,KAAK,KAAK,GAAG,mBAAmBY,EAAE,KAAK,WAAW,CAAC,UAAUZ,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,wBAAwBA,EAAE,kCAAkC,CAACY,CAAC,IAAI,EAAE,KAAK,QAAQZ,EAAE,EAAE,CAAC,CAAC,MAAM,CAAC,GAAG,KAAK,cAAc,EAAE,KAAK,QAAQ,MAAM,KAAK,MAAM,IAAI,MAAM,+GAA+G,EAAE,IAAIY,EAAE,KAAK,QAAQ,KAAK,EAAE,OAAOgmB,EAAE,UAAUhmB,CAAC,EAAEA,EAAE,KAAKZ,GAAG,KAAK,SAASA,CAAC,CAAC,EAAE,KAAK,SAASY,CAAC,CAAC,CAAC,SAASA,EAAE,CAAC,KAAK,UAAUA,EAAE,IAAIZ,EAAE,KAAK,UAAU,cAAc,EAAE,KAAK,UAAU,UAAU,GAAG,KAAK,UAAU,qBAAqB,KAAK,CAAC,IAAI6B,EAAE,KAAK,UAAU,oBAAoBA,EAAE,WAAW,OAAO,EAAEA,EAAE,WAAWA,EAAE,sBAAsB,OAAO,KAAK,qBAAqBA,EAAE,qBAAqB,CAAC,KAAK,UAAU,EAAE,KAAK,QAAQ,GAAG7B,EAAE,SAAS,YAAYA,EAAE,SAAS,cAAc,IAAIK,EAAE,KAAK,GAAG,cAAc,KAAK,UAAU,WAAW,KAAK,UAAU,WAAW,EAAE,GAAG,KAAK,SAAS,IAAIgsE,GAAG3D,GAAG,SAAS,eAAe1oE,EAAE,KAAK,SAAS,CAAC,EAAE,KAAK,SAAS,UAAU,KAAK,6BAA6BK,CAAC,EAAE,KAAK,SAAS,gBAAgB,KAAK,gBAAgBO,EAAE,kBAAkB,MAAMA,EAAE,iBAAiB,MAAM,KAAK,CAAC,IAAIiB,EAAE6mE,GAAG,SAAS,eAAe9nE,EAAE,gBAAgB,EAAE,KAAK,YAAY,IAAIyrE,GAAGxqE,CAAC,EAAE,KAAK,YAAY,UAAU,KAAK,SAAS,UAAU,KAAK,YAAY,gBAAgB,KAAK,gBAAgB,KAAK,qBAAqBjB,EAAE,oBAAoB,CAAC,MAAM,EAAE,CAAC,MAAM,KAAKA,EAAEZ,EAAE,CAAC,GAAG,OAAOY,GAAG,SAAS,CAAC,IAAI,EAAE,KAAK,GAAG,gBAAgBA,CAAC,EAAE,GAAG,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,0CAA0CA,IAAI,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,wBAAwB,EAAE,kCAAkCA,IAAI,EAAEA,EAAE,EAAE,EAAE,CAAC,GAAGA,EAAE,MAAM,KAAK,MAAM,IAAI,MAAM,6GAA6G,EAAE,OAAOA,EAAE,KAAK,KAAK,SAAS,CAAC,CAAC,QAAQA,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,QAAQY,EAAE,KAAK,WAAW,EAAE,GAAG,KAAK,qBAAqB,CAAC,IAAIP,EAAE,aAAaspB,GAAG,CAAC,CAAC,EAAE,EAAE9nB,EAAE,CAAC,EAAE,OAAOxB,EAAE,QAAQ,CAACyB,EAAEC,IAAIF,EAAE,KAAK,qBAAqBE,IAAID,CAAC,EAAED,CAAC,CAAC,OAAO,CAAC,CAAC,gBAAgBjB,EAAE,CAAC,GAAG,EAAEA,aAAa+oB,KAAK,CAAC,MAAM,QAAQ/oB,CAAC,EAAE,CAAC,GAAG,KAAK,WAAW,MAAM,KAAK,UAAU,QAAQ,KAAK,QAAQP,KAAK,KAAK,UAAU,OAAO,CAAC,IAAIwB,EAAE,KAAK,UAAU,OAAOxB,GAAGwB,EAAE,YAAY,OAAOjB,EAAEP,GAAG,KAAK,0BAA0BwB,EAAE,YAAY,CAAC,OAAOjB,CAAC,CAACA,EAAE,MAAM,QAAQA,CAAC,EAAEA,EAAE,CAACA,CAAC,EAAE,IAAIZ,EAAE,OAAO,KAAK,KAAK,yBAAyB,EAAE,OAAO,GAAGY,EAAE,OAAOZ,IAAI,KAAK,WAAW,OAAO,MAAM,IAAI,MAAM,oDAAoD,KAAK,WAAW,OAAOA,gDAAgDY,EAAE,gCAAgC,EAAE,IAAI,EAAE,EAAE,OAAO,KAAK,WAAW,OAAO,CAACP,EAAEwB,IAAI,CAAC,IAAIC,EAAE,KAAK,UAAU,KAAK,UAAU,OAAOD,GAAG,KAAK,OAAOC,GAAG,MAAMA,EAAE,YAAY,KAAKzB,EAAEwB,GAAG,KAAK,0BAA0BC,EAAE,YAAYzB,EAAEwB,GAAGjB,EAAE,KAAKP,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,iBAAiBO,EAAE,CAAC,OAAOA,EAAEA,GAAG,KAAK,YAAY,MAAM,QAAQA,CAAC,EAAEA,EAAE,CAACA,CAAC,CAAC,CAAC,yBAAyB,CAAC,OAAO,KAAK,aAAa,KAAK,CAAC,EAAE,KAAK,sBAAsB,KAAK,KAAK,YAAY,QAAQ,CAAC,EAAE,CAAC,CAAC,EAAE,KAAK,YAAY,QAAQ,CAAC,EAAE,OAAO,KAAK,KAAK,qBAAqB,OAAO,CAAC,CAAC,CAAC,MAAM,8BAA8B,CAAC,OAAO,KAAK,aAAa,KAAK,CAAC,EAAE,KAAK,sBAAsB,KAAK,KAAK,YAAY,aAAa,CAAC,EAAE,CAAC,CAAC,EAAE,KAAK,YAAY,aAAa,CAAC,EAAE,OAAO,KAAK,KAAK,qBAAqB,OAAO,CAAC,CAAC,CAAC,6BAA6BA,EAAE,CAAC,GAAG,KAAK,0BAA0B,CAAC,EAAE,KAAK,qBAAqB,CAAC,IAAIZ,EAAE,OAAO,KAAK,KAAK,qBAAqB,OAAO,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,IAAI,CAAC,IAAIK,EAAEL,EAAE,GAAG6B,EAAE,KAAK,qBAAqB,QAAQxB,GAAG,KAAK,0BAA0BwB,EAAE,YAAYjB,EAAE,EAAE,CAAC,CAAC,CAAC,QAAQA,EAAEZ,EAAE,CAAC,KAAK,2BAA2B,MAAM,KAAK,6BAA6B,KAAK,wBAAwB,CAAC,EAAEY,EAAE,KAAK,gBAAgBA,CAAC,EAAEZ,EAAE,KAAK,iBAAiBA,CAAC,EAAE,IAAI,EAAE,KAAK,SAAS,QAAQY,EAAEZ,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,EAAE,CAAC,MAAM,aAAaY,EAAEZ,EAAE,CAAC,KAAK,2BAA2B,MAAM,KAAK,6BAA6B,MAAM,KAAK,6BAA6B,CAAC,EAAEY,EAAE,KAAK,gBAAgBA,CAAC,EAAEZ,EAAE,KAAK,iBAAiBA,CAAC,EAAE,IAAI,EAAE,MAAM,KAAK,SAAS,aAAaY,EAAEZ,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,EAAE,CAAC,wBAAwB,CAAC,OAAO,KAAK,SAAS,uBAAuB,CAAC,CAAC,4BAA4B,CAAC,KAAK,SAAS,2BAA2B,CAAC,CAAC,6BAA6BY,EAAE,CAAC,OAAO,OAAO,KAAKA,CAAC,EAAE,OAAO,CAACZ,EAAE,KAAKA,EAAE,GAAG,CAACY,EAAE,EAAE,EAAEZ,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,KAAK,SAAS,QAAQ,EAAE,KAAK,cAAc,KAAK,YAAY,QAAQ,EAAE,KAAK,2BAA2BwzB,GAAG,KAAK,yBAAyB,GAAG,KAAK,gBAAgB,QAAQ,CAAC,CAAC,EAAE,eAAek5C,GAAG5sE,EAAE,EAAE,CAAC,EAAEE,EAAE4wB,GAAG,CAAC,GAAG9wB,GAAG,KAAK,MAAM,IAAI,MAAM,wGAAwG,EAAE,GAAG,OAAO,EAAE,CAAC,GAAG,EAAE,WAAW,OAAOA,GAAG,WAAWA,EAAE6sE,GAAG7sE,CAAC,GAAG,IAAIM,EAAE,IAAIqsE,GAAG3sE,EAAE,EAAEE,CAAC,EAAE,OAAO,MAAMI,EAAE,KAAK,EAAEA,CAAC,CAAC,SAASwsE,GAAG9sE,EAAE,CAAC,GAAGA,GAAG,KAAK,MAAM,IAAI,MAAM,sHAAsH,EAAE,IAAI,EAAE,GAAGA,aAAa,MAAM,CAAC,GAAG,CAACM,EAAEC,CAAC,EAAEP,EAAE,GAAG,CAACM,EAAE,MAAM,IAAI,MAAM,kDAAkD,EAAE,GAAG,CAACC,GAAG,EAAEA,aAAa,aAAa,MAAM,IAAI,MAAM,mEAAmE,EAAE,GAAG,EAAE,kBAAkBD,GAAG,MAAM,IAAI,MAAM,uCAAuC,EAAE,GAAG,EAAE,oBAAoBA,GAAG,MAAM,IAAI,MAAM,yCAAyC,EAAE,IAAI,EAAEwwB,GAAG,eAAexwB,EAAE,eAAe,EAAE,EAAEwwB,GAAG,6BAA6BxwB,EAAE,EAAEC,CAAC,EAAE,EAAEuwB,GAAG,eAAe,CAAC,CAAC,SAAS,SAAS9wB,EAAE,EAAEA,UAAU,kBAAkBA,GAAG,gBAAgBA,GAAG,eAAeA,EAAE,EAAE8wB,GAAG,eAAe9wB,CAAC,MAAO,OAAM,IAAI,MAAM,sBAAsB,EAAE,IAAIE,EAAE,IAAIysE,GAAG,CAAC,EAAE,OAAOzsE,EAAE,KAAK,EAAEA,CAAC,CAAC,SAAS2sE,GAAG7sE,EAAE,CAAC,OAAOA,EAAE,SAAS,GAAG,IAAIA,EAAEA,EAAE,KAAK,GAAGA,IAAI0sE,KAAKD,IAAI,CAAC,IAAIM,GAAG,QAAYC,GAAG,CAAC,EAAE5sE,GAAG4sE,GAAG,CAAC,WAAW,IAAIC,GAAG,QAAQ,IAAIC,GAAG,eAAe,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,cAAc,IAAIC,GAAG,MAAM,IAAIC,GAAG,IAAI,IAAIC,GAAG,KAAK,IAAIC,GAAG,UAAU,IAAIC,GAAG,WAAW,IAAIC,GAAG,aAAa,IAAIC,GAAG,OAAO,IAAIC,GAAG,IAAI,IAAIC,EAAE,CAAC,EAAE,IAAIC,GAAGttE,GAAG+E,GAAG,CAAC,EAAMwoE,GAAGvtE,GAAG+E,GAAG,CAAC,EAAE,SAASyoE,GAAGhuE,EAAE,EAAE,CAAC,OAAOiuE,GAAGjuE,EAAE,CAAC,CAAC,CAAC,SAASiuE,GAAGjuE,EAAE,EAAEE,EAAE,IAAI,IAAII,EAAE,IAAI,IAAI,CAAC,GAAGN,GAAG,KAAK,OAAO,KAAK,GAAG,OAAO,MAAM,YAAYA,aAAa,KAAK,OAAOA,EAAE,MAAM,EAAE,GAAGM,EAAE,IAAIN,CAAC,EAAE,MAAM,IAAI,MAAM,wCAAwC,EAAE,GAAGE,EAAE,IAAIF,CAAC,EAAE,OAAOE,EAAE,IAAIF,CAAC,EAAE,IAAIO,EAAE,EAAEP,CAAC,EAAE,GAAGO,EAAE,SAASA,EAAE,QAAQ,KAAK,MAAM,IAAI,MAAM,mEAAmE,EAAE,GAAGA,EAAE,QAAQ,GAAG2tE,GAAGluE,CAAC,EAAE,CAAC,IAAI,EAAE,MAAM,QAAQA,CAAC,EAAE,CAAC,EAAE,CAAC,EAAEM,EAAE,IAAIN,CAAC,EAAE,QAAQ,KAAKA,EAAE,CAAC,IAAIiC,EAAEjC,EAAE,GAAGkC,EAAE+rE,GAAGhsE,EAAE,EAAE/B,EAAEI,CAAC,EAAE,EAAE,GAAG4B,CAAC,CAAC,OAAO5B,EAAE,OAAON,CAAC,EAAEA,EAAE,YAAY,EAAE,UAAUA,EAAE,WAAW,CAAC,KAAM,OAAM,IAAI,MAAM,yCAAyCA,GAAG,MAAO,QAAOE,EAAE,IAAIF,EAAEO,EAAE,KAAK,EAAEA,EAAE,KAAK,CAAC,SAAS4tE,GAAGnuE,EAAE,EAAEouE,GAAG,CAAC,OAAOC,GAAGruE,EAAE,CAAC,CAAC,CAAC,SAASquE,GAAGruE,EAAE,EAAEE,EAAE,IAAI,IAAI,CAAC,IAAII,EAAEN,EAAE,GAAG,GAAGE,EAAE,IAAII,CAAC,EAAE,MAAM,IAAI,MAAM,wCAAwC,EAAE,IAAIC,EAAE,EAAEP,CAAC,EAAE,GAAGO,EAAE,SAASA,EAAE,QAAQ,KAAK,MAAM,IAAI,MAAM,mEAAmE,EAAE,GAAGA,EAAE,QAAQ,GAAG2tE,GAAG5tE,CAAC,EAAE,CAAC,IAAI,EAAE,MAAM,QAAQA,CAAC,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE,IAAII,CAAC,EAAE,QAAQ,KAAKA,EAAE,CAAC,IAAI2B,EAAEjC,EAAE,IAAImC,GAAGA,EAAE,EAAE,EAAED,EAAEmsE,GAAGpsE,EAAE,EAAE/B,CAAC,EAAE,EAAE,GAAGgC,CAAC,CAAC,OAAOhC,EAAE,OAAOI,CAAC,EAAE,CAAC,KAAM,OAAM,IAAI,MAAM,yCAAyCA,GAAG,MAAO,QAAOC,EAAE,KAAK,CAAC,SAAS6tE,GAAGpuE,EAAE,CAAC,OAAOA,IAAI,KAAK,KAAKkuE,GAAGluE,EAAE,EAAE,EAAE,CAAC,MAAM,KAAK,QAAQ,EAAE,EAAE,CAAC,MAAMA,EAAE,QAAQ,EAAE,CAAC,CAAC,eAAesuE,GAAGtuE,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,IAAI+tE,GAAGjuE,EAAE,EAAEE,CAAC,EAAE,QAAQK,KAAK,MAAM,KAAKL,EAAE,KAAK,CAAC,EAAE,CAAC,IAAI6B,EAAE7B,EAAE,IAAIK,CAAC,EAAE,GAAGumB,EAAE,UAAU/kB,CAAC,EAAE,CAAC,IAAIC,EAAE,MAAMD,EAAE7B,EAAE,IAAIK,EAAEyB,CAAC,CAAC,CAAC,CAAC,OAAOisE,GAAGjuE,EAAE,EAAEE,CAAC,CAAC,CAAC,SAASguE,GAAGluE,EAAE,CAAC,IAAI,EAAE,GAAG,GAAGya,EAAE,EAAE,IAAI,YAAY,EAAE,EAAEza,aAAa,gBAAgB,CAAC,GAAG,CAAC,cAAcE,CAAC,EAAE+F,GAAG,EAAE,EAAEjG,aAAaE,CAAC,CAAC,OAAOF,GAAG,MAAM,CAAC,YAAY,OAAOA,CAAC,IAAI,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,UAAU,EAAEA,aAAa6pB,KAAK,EAAE7pB,aAAa,UAAU,CAAC,EAAE,CAAC,SAASuuE,GAAGvuE,EAAE,CAAC,OAAOA,GAAG,MAAMwuE,GAAGxuE,CAAC,GAAG,MAAM,QAAQA,CAAC,GAAG,OAAOA,GAAG,UAAUA,aAAa6pB,IAAI/C,EAAE,aAAa9mB,CAAC,CAAC,CAAC,SAASwuE,GAAGxuE,EAAE,CAAC,OAAOA,IAAI,MAAM,OAAOA,GAAG,UAAU,OAAOA,GAAG,UAAU,CAAC,SAASyuE,GAAGzuE,EAAE,CAAC,OAAOguE,GAAGhuE,EAAE0uE,EAAE,CAAC,CAAC,SAASA,GAAG1uE,EAAE,CAAC,OAAOA,aAAa6pB,GAAG,CAAC,MAAM7pB,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAEkuE,GAAGluE,CAAC,EAAE,CAAC,MAAM,KAAK,QAAQ,EAAE,EAAE,CAAC,MAAMA,EAAE,QAAQ,EAAE,CAAC,CAAC,IAAI2uE,GAAG,KAAK,CAAC,YAAY7tE,EAAE,CAAC,GAAG,KAAK,SAASA,EAAE,KAAK,MAAM,EAAE,KAAK,IAAI,EAAEA,GAAG,KAAK,MAAM,IAAI,WAAW,iDAAiD,EAAE,GAAGA,EAAE,EAAE,MAAM,IAAI,WAAW,2CAA2C,EAAE,KAAK,KAAK,IAAI,MAAMA,CAAC,EAAE,KAAK,gBAAgB,EAAEA,CAAC,CAAC,KAAKA,EAAE,CAAC,KAAKA,EAAE,GAAGA,GAAG,KAAK,gBAAgB,OAAOA,EAAE,KAAK,eAAe,CAAC,IAAIA,EAAE,CAAC,GAAGA,EAAE,EAAE,MAAM,IAAI,WAAW,qCAAqC,EAAE,OAAO,KAAK,KAAKA,EAAE,KAAK,SAAS,CAAC,IAAIA,EAAEZ,EAAE,CAAC,GAAGY,EAAE,EAAE,MAAM,IAAI,WAAW,qCAAqC,EAAE,KAAK,KAAKA,EAAE,KAAK,UAAUZ,CAAC,CAAC,QAAQ,CAAC,IAAIY,EAAE,KAAK,IAAI,KAAK,MAAM,OAAOA,EAAE,IAAIA,EAAE,KAAK,gBAAgBA,GAAGA,CAAC,CAAC,QAAQ,CAAC,OAAO,KAAK,OAAO,IAAI,KAAK,QAAQ,CAAC,SAAS,CAAC,OAAO,KAAK,OAAO,IAAI,CAAC,CAAC,KAAKA,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAM,IAAI,WAAW,sBAAsB,EAAE,KAAK,IAAI,KAAK,IAAIA,CAAC,EAAE,KAAK,IAAI,KAAK,KAAK,KAAK,IAAI,CAAC,CAAC,CAAC,QAAQA,EAAE,CAAC,QAAQZ,KAAKY,EAAE,KAAK,KAAKZ,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,KAAK,QAAQ,EAAE,MAAM,IAAI,WAAW,uBAAuB,EAAE,KAAK,IAAI,KAAK,KAAK,KAAK,IAAI,CAAC,EAAE,IAAIY,EAAE,KAAK,IAAI,KAAK,GAAG,EAAE,OAAO,KAAK,IAAI,KAAK,IAAI,MAAM,EAAEA,CAAC,CAAC,QAAQA,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAM,IAAI,WAAW,sBAAsB,EAAE,KAAK,MAAM,KAAK,KAAK,KAAK,MAAM,CAAC,EAAE,KAAK,IAAI,KAAK,MAAMA,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,KAAK,QAAQ,EAAE,MAAM,IAAI,WAAW,uBAAuB,EAAE,IAAIA,EAAE,KAAK,IAAI,KAAK,KAAK,EAAE,OAAO,KAAK,IAAI,KAAK,MAAM,MAAM,EAAE,KAAK,MAAM,KAAK,KAAK,KAAK,MAAM,CAAC,EAAEA,CAAC,CAAC,cAAcA,EAAE,CAAC,GAAG,KAAK,QAAQ,EAAE,MAAM,IAAI,WAAW,uBAAuB,EAAE,IAAIZ,EAAE,KAAK,KAAK,KAAK,MAAMY,CAAC,EAAE,EAAE,KAAK,IAAIZ,CAAC,EAAE,OAAO,KAAK,IAAIA,EAAE,KAAK,IAAI,CAAC,EAAE,CAAC,CAAC,EAAM0uE,GAAG,cAAcD,EAAE,CAAC,aAAa,CAAC,MAAMC,GAAG,gBAAgB,CAAC,CAAC,QAAQ,CAAC,MAAM,EAAE,CAAC,KAAK9tE,EAAE,CAAC,MAAM,OAAO,GAAG,KAAK,OAAO,EAAE,MAAM,KAAKA,CAAC,CAAC,CAAC,QAAQA,EAAE,CAAC,MAAM,OAAO,GAAG,KAAK,OAAO,EAAE,MAAM,QAAQA,CAAC,CAAC,CAAC,QAAQ,CAAC,IAAIA,EAAE,KAAK,SAAS,EAAEZ,EAAE,IAAI,MAAMY,CAAC,EAAE,EAAE,KAAK,OAAO,EAAE,QAAQP,EAAE,EAAEA,EAAE,EAAEA,IAAIL,EAAEK,GAAG,KAAK,IAAI,KAAK,KAAK,KAAK,MAAMA,CAAC,CAAC,EAAE,KAAK,KAAKL,EAAE,KAAK,SAASY,EAAE,KAAK,gBAAgB,EAAE,KAAK,SAAS,KAAK,MAAM,EAAE,KAAK,IAAI,CAAC,CAAC,EAAE8tE,GAAG,iBAAiB,GAAG,SAASC,GAAG7uE,EAAE,CAAC,OAAO,IAAI8uE,GAAG9uE,CAAC,CAAC,CAAC,SAAS+uE,GAAG/uE,EAAE,CAAC,OAAO,IAAIgvE,GAAGhvE,CAAC,CAAC,CAAC,SAASivE,GAAGjvE,EAAE,EAAE,CAAC,OAAO,IAAIkvE,GAAGlvE,EAAE,CAAC,CAAC,CAAC,SAASmvE,GAAGnvE,EAAE,EAAEovE,GAAG,KAAK,CAAC,OAAO,IAAIC,GAAGrvE,EAAE,CAAC,CAAC,CAAC,IAAIsvE,GAAG,KAAK,CAAC,MAAM,SAAS,CAAC,IAAIxuE,EAAE,CAAC,EAAEZ,EAAE,MAAM,KAAK,KAAK,EAAE,KAAK,CAACA,EAAE,MAAMY,EAAE,KAAKZ,EAAE,KAAK,EAAEA,EAAE,MAAM,KAAK,KAAK,EAAE,OAAOY,CAAC,CAAC,MAAM,gBAAgB,CAAC,IAAIA,EAAE,KAAK,SAAS,GAAG,EAAEZ,EAAE,CAAC,EAAE,EAAE,MAAMY,EAAE,KAAK,EAAE,KAAK,CAAC,EAAE,MAAMZ,EAAE,KAAK,EAAE,KAAK,EAAE,EAAE,MAAMY,EAAE,KAAK,EAAE,OAAOZ,CAAC,CAAC,MAAM,cAAc,CAAC,IAAIY,EAAE,MAAM,KAAK,KAAK,EAAE,KAAK,CAACA,EAAE,MAAMA,EAAE,MAAM,KAAK,KAAK,CAAC,CAAC,MAAM,aAAaA,EAAE,CAAC,IAAIZ,EAAE,MAAM,KAAK,KAAK,EAAE,EAAEY,EAAEZ,EAAE,KAAK,EAAE,KAAK,CAACA,EAAE,MAAM,GAAGA,EAAE,MAAM,KAAK,KAAK,EAAE,EAAEY,EAAEZ,EAAE,KAAK,CAAC,CAAC,aAAaY,EAAE,CAAC,OAAO,IAAIyuE,GAAG,KAAKzuE,CAAC,CAAC,CAAC,OAAOA,EAAE,CAAC,OAAO,IAAI0uE,GAAG,KAAK1uE,CAAC,CAAC,CAAC,IAAIA,EAAE,CAAC,OAAO,IAAI2uE,GAAG,KAAK3uE,CAAC,CAAC,CAAC,SAASA,EAAE,CAAC,OAAO,IAAI4uE,GAAG,KAAK5uE,CAAC,CAAC,CAAC,eAAeA,EAAE,CAAC,OAAO,IAAI4uE,GAAG,KAAK5uE,CAAC,EAAE,OAAO,CAAC,CAAC,QAAQA,EAAE,CAAC,OAAO,IAAI6uE,GAAG,KAAK7uE,CAAC,CAAC,CAAC,MAAM,aAAaA,EAAE,CAAC,OAAO,KAAK,IAAIA,CAAC,EAAE,aAAa,CAAC,CAAC,MAAM,cAAcA,EAAE,CAAC,OAAO,KAAK,eAAeA,CAAC,EAAE,aAAaZ,GAAGA,IAAI,EAAE,CAAC,CAAC,cAAcY,EAAEZ,EAAE,GAAG,CAAC,OAAO,IAAI0vE,GAAG,KAAK9uE,EAAEZ,CAAC,CAAC,CAAC,iBAAiBY,EAAEZ,EAAE,GAAG,EAAEkuE,GAAG,CAAC,OAAO,KAAK,cAActtE,EAAEZ,CAAC,EAAE,IAAI6B,GAAGosE,GAAGpsE,EAAE,CAAC,CAAC,CAAC,CAAC,YAAYjB,EAAEZ,EAAE,CAAC,OAAO,IAAIgvE,GAAGL,GAAG,CAAC,KAAK/tE,CAAC,CAAC,EAAEZ,CAAC,CAAC,CAAC,KAAKY,EAAE,CAAC,OAAOA,EAAE,GAAGA,GAAG,KAAK,KAAK,IAAI+uE,GAAG,KAAK/uE,CAAC,CAAC,CAAC,KAAKA,EAAE,CAAC,OAAOA,EAAE,GAAGA,GAAG,KAAK,KAAK,IAAIgvE,GAAG,KAAKhvE,CAAC,CAAC,CAAC,SAASA,EAAE,CAAC,OAAO,IAAIivE,GAAG,KAAKjvE,CAAC,CAAC,CAAC,QAAQA,EAAEZ,EAAE,CAAC,OAAO,IAAI8vE,GAAG,KAAKlvE,EAAEZ,CAAC,CAAC,CAAC,QAAQ,CAAC,OAAO,IAAI+vE,GAAG,IAAI,CAAC,CAAC,EAAEnB,GAAG,cAAcQ,EAAE,CAAC,YAAYxuE,EAAE,CAAC,MAAM,EAAE,KAAK,MAAMA,EAAE,KAAK,KAAK,CAAC,CAAC,SAAS,CAAC,MAAM,YAAY,KAAK,MAAM,cAAc,CAAC,MAAM,MAAM,CAAC,GAAG,KAAK,MAAM,KAAK,MAAM,OAAO,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,IAAIA,EAAE,KAAK,MAAM,KAAK,MAAM,OAAO,KAAK,OAAO,CAAC,MAAM2tE,GAAG3tE,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,EAAEkuE,GAAG,cAAcM,EAAE,CAAC,YAAYxuE,EAAE,CAAC,MAAM,EAAE,KAAK,OAAOA,CAAC,CAAC,SAAS,CAAC,MAAM,eAAe,CAAC,MAAM,MAAM,CAAC,GAAG,CAAC,OAAO,KAAK,OAAO,CAAC,OAAOA,EAAN,CAAS,MAAMA,EAAE,QAAQ,mDAAmDA,EAAE,UAAUA,CAAC,CAAC,CAAC,EAAEmvE,GAAG,cAAcX,EAAE,CAAC,YAAYxuE,EAAE,CAAC,MAAM,EAAE,KAAK,SAASA,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,aAAa,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,OAAO,KAAK,SAAS,KAAK,CAAC,CAAC,EAAEgvE,GAAG,cAAcR,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,SAASZ,EAAE,KAAK,MAAM,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,WAAW,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,KAAK,KAAK,QAAQ,KAAK,UAAU,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,OAAOA,EAAE4yB,GAAG5yB,EAAE,KAAK,CAAC,CAAC,OAAO,KAAK,SAAS,KAAK,CAAC,CAAC,EAAE+uE,GAAG,cAAcP,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,SAASZ,EAAE,KAAK,MAAM,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,WAAW,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,KAAK,SAAS,KAAK,CAAC,CAAC,EAAE0vE,GAAG,cAAcN,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,EAAE,GAAG,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,EAAE,KAAK,qBAAqB,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,oBAAoB,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,IAAIY,EAAE,CAAC,EAAE,KAAKA,EAAE,OAAO,KAAK,WAAW,CAAC,IAAIZ,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,OAAO,KAAK,sBAAsBY,EAAE,OAAO,EAAE,CAAC,MAAMA,EAAE,KAAK,EAAE,EAAE,CAAC,MAAM,KAAK,KAAK,EAAE,EAAEA,EAAE,KAAKZ,EAAE,KAAK,CAAC,CAAC,MAAM,CAAC,MAAMY,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE0uE,GAAG,cAAcF,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,aAAa,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,OAAO,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,MAAM,KAAK,UAAUA,EAAE,KAAK,EAAE,OAAOA,EAAE4yB,GAAG5yB,EAAE,KAAK,CAAC,CAAC,CAAC,EAAE2uE,GAAG,cAAcH,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,UAAU,CAAC,MAAM,MAAM,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,IAAIZ,EAAE8pB,GAAG,sBAAsBlpB,EAAE,KAAK,EAAE,EAAE,KAAK,UAAUA,EAAE,KAAK,EAAEP,EAAEypB,GAAG,sBAAsB,CAAC,EAAE,QAAQjoB,KAAK7B,EAAE8pB,GAAG,eAAejoB,EAAExB,CAAC,GAAGwB,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,EAAEwtE,GAAG,cAAcD,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,QAAQZ,EAAE,KAAK,MAAM,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,mBAAmB,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,OAAO,GAAG,CAAC,OAAO,MAAM,KAAK,SAAS,KAAK,CAAC,OAAOY,EAAN,CAAS,GAAG,CAAC,KAAK,QAAQA,CAAC,EAAE,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE4uE,GAAG,cAAcJ,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,eAAe,CAAC,MAAM,MAAM,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,IAAIZ,EAAE8pB,GAAG,sBAAsBlpB,EAAE,KAAK,EAAE,EAAE,MAAM,KAAK,UAAUA,EAAE,KAAK,EAAEP,EAAEypB,GAAG,sBAAsB,CAAC,EAAE,QAAQjoB,KAAK7B,EAAE8pB,GAAG,eAAejoB,EAAExB,CAAC,GAAGwB,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,EAAEmuE,GAAG,cAAcZ,EAAE,CAAC,aAAa,CAAC,MAAM,EAAE,KAAK,YAAY,IAAIV,GAAG,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,MAAM,YAAY,CAAC,KAAK,KAAK,YAAY,OAAO,IAAI,GAAG,GAAG,CAAC,MAAM,KAAK,KAAK,EAAE,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,MAAM,CAAC,MAAM,KAAK,YAAY,MAAM,EAAE,KAAK,EAAE,CAAC,CAAC,EAAEe,GAAG,cAAcO,EAAE,CAAC,YAAYpvE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,cAAc,CAAC,MAAM,MAAM,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,MAAM,GAAG,IAAIZ,EAAE8pB,GAAG,sBAAsBlpB,EAAE,KAAK,EAAE,EAAE,KAAK,UAAUA,EAAE,KAAK,EAAEP,EAAEypB,GAAG,sBAAsB,CAAC,EAAE,KAAK,YAAY,QAAQ,CAAC,EAAE,QAAQjoB,KAAK7B,EAAE8pB,GAAG,eAAejoB,EAAExB,CAAC,GAAGwB,EAAE,QAAQ,EAAE,MAAM,EAAE,CAAC,EAAEmtE,GAAG,cAAcI,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiBA,EAAE,KAAK,SAAS,KAAK,KAAK,SAAS,KAAK,KAAK,cAAcY,CAAC,CAAC,SAAS,CAAC,MAAM,wDAAwD,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,cAAc,KAAK,QAAQ,EAAE,KAAK,QAAQ,CAAC,MAAM,cAAcA,EAAE,CAAC,GAAG,MAAMA,EAAE,KAAK,UAAU,KAAK,CAAC,IAAI,EAAE,MAAM,KAAK,cAAc,KAAK,EAAE,GAAG,EAAE,KAAK,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,KAAK,SAAS,EAAE,MAAM,KAAK,kBAAkB,OAAO,KAAK,SAAS,KAAK,SAAS,aAAa,KAAK,gBAAgB,EAAE,CAAC,IAAIZ,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,OAAOA,EAAE,MAAM,KAAK,SAAS,KAAK,KAAK,cAAcY,CAAC,GAAGZ,CAAC,CAAC,EAAEkvE,IAAI,SAASpvE,EAAE,CAACA,EAAEA,EAAE,KAAK,GAAG,OAAOA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,QAAQ,GAAG,SAAS,GAAGovE,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,cAAcC,EAAE,CAAC,YAAYxuE,EAAEZ,EAAEkvE,GAAG,KAAK,CAAC,MAAM,EAAE,KAAK,UAAUtuE,EAAE,KAAK,aAAaZ,EAAE,KAAK,MAAM,EAAE,KAAK,eAAe,IAAI,CAAC,SAAS,CAAC,MAAM,kDAAkD,CAAC,MAAM,UAAUY,EAAE,CAAC,MAAMA,EAAE,IAAIZ,EAAE,EAAE,EAAE,EAAE,SAASK,EAAEyB,EAAE,CAAC,OAAOA,aAAastE,GAAG,CAAC,MAAMttE,EAAE,KAAK,EAAE,KAAKE,IAAIhC,IAAIgC,EAAE,MAAM,IAAIA,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAE,CAAC,MAAM,KAAK,QAAQ,EAAE,CAAC,CAAC,IAAIH,EAAE,MAAMusE,GAAG,KAAK,UAAU/tE,CAAC,EAAE,GAAGL,IAAI,EAAE,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,GAAG,EAAE,EAAE,OAAO,KAAK,aAAa,CAAC,KAAKkvE,GAAG,KAAK,MAAM,IAAI,MAAM,qEAAqE,KAAK,QAAQ,EAAE,KAAKA,GAAG,SAAS,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,KAAKA,GAAG,QAAQ,QAAQ,CAAC,OAAO,KAAK,QAAQ,CAAC,MAAMrtE,EAAE,KAAK,EAAE,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,eAAe,KAAK,UAAU,KAAK,cAAc,EAAE,KAAK,cAAc,CAAC,EAAEguE,GAAG,cAAcT,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,WAAWZ,EAAE,KAAK,OAAO,IAAIyuE,GAAGzuE,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,eAAe,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,OAAO,OAAO,GAAG,CAAC,IAAIY,EAAE,KAAK,SAAS,KAAK,EAAE,KAAK,OAAO,KAAKA,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO,KAAK,OAAO,EAAE,KAAK,OAAO,MAAM,CAAC,CAAC,EAAEkvE,GAAG,cAAcD,EAAE,CAAC,YAAYjvE,EAAEZ,EAAE,EAAE,CAAC,MAAMY,EAAEZ,CAAC,EAAE,KAAK,SAASY,EAAE,KAAK,WAAWZ,EAAE,KAAK,kBAAkB,GAAG,KAAK,OAAO6tE,GAAG,KAAK,GAAGjnD,EAAE,IAAI,EAAE,SAAS,CAAC,EAAE,KAAK,SAAS,QAAQ,QAAQ,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,SAAS,KAAK,IAAI,KAAK,WAAW,CAAC,EAAE,KAAK,QAAQ,CAAC,UAAUhmB,EAAE,CAAC,OAAO,KAAK,MAAM,KAAK,OAAO,EAAEA,CAAC,CAAC,CAAC,aAAa,CAAC,OAAO,KAAK,UAAU,KAAK,OAAO,OAAO,CAAC,CAAC,CAAC,MAAM,YAAY,CAAC,IAAI,KAAK,mBAAmB,KAAK,OAAO,EAAE,CAAC,KAAK,OAAO,QAAQ,GAAG,CAAC,IAAIA,EAAE,KAAK,YAAY,EAAEZ,EAAE,MAAM,KAAK,OAAO,cAAcY,CAAC,EAAE,GAAGZ,EAAE,KAAK,KAAK,kBAAkB,OAAQ,QAAO,KAAK,OAAO,EAAEA,CAAC,CAAC,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,CAAC,CAAC,EAAMgtE,GAAG,KAAK,CAAC,aAAa,CAAC,KAAK,KAAK,IAAI,CAAC,MAAMpsE,EAAEZ,EAAE,GAAG,CAAC,IAAI,EAAE,KAAK4mB,EAAE,OAAOhmB,EAAE,EAAE,IAAI;AAAA,QAClqvDA,GAAG,EAAE,IAAIP,EAAE,OAAO,KAAK,OAAO,EAAE,GAAG,KAAK,MAAM,KAAKA,EAAE,KAAK,KAAKL,EAAEK,EAAE,KAAK,KAAK,KAAK,KAAKO,CAAC,EAAEP,EAAE,KAAK,MAAM,KAAK,KAAKO,CAAC,EAAEqvE,GAAG,UAAU,MAAM,EAAE,SAAS,GAAG,iBAAiBrvE,EAAEZ,EAAEkwE,EAAE,EAAE7vE,CAAC,CAAC,CAAC,YAAYO,EAAE,CAAC,IAAIZ,EAAE,KAAK,EAAE,OAAO,KAAK,OAAO,EAAE,GAAGY,EAAE,OAAO,EAAE,EAAE,EAAE,EAAE,EAAE,KAAK,MAAM,MAAMA,EAAE,MAAM,KAAK,EAAE,KAAK,KAAKA,EAAE,KAAK,EAAE,KAAKqvE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,YAAY,MAAMY,EAAE,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,OAAOA,EAAE,CAAC,IAAIZ,EAAE,KAAK,EAAE,OAAO,KAAK,OAAO,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,KAAKiwE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,OAAOK,GAAGkzB,EAAE,IAAI3yB,EAAEP,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,aAAaO,EAAE,CAAC,OAAO,MAAM,KAAK,SAAS,GAAG,aAAaA,CAAC,CAAC,CAAC,IAAIA,EAAE,CAAC,IAAIZ,EAAE,KAAK,OAAOiwE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,IAAI,GAAGuzB,EAAE,IAAI3yB,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,SAASA,EAAE,CAAC,IAAIZ,EAAE,KAAK,OAAOiwE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,SAASY,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,SAASA,EAAE,CAAC,GAAGA,GAAG,KAAK,MAAM,IAAI,WAAW,2DAA2D,EAAE,IAAIZ,EAAE,KAAK,OAAOiwE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,SAASY,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,OAAOA,EAAE,CAAC,IAAIZ,EAAE,KAAK,EAAE,OAAO,KAAK,MAAM,MAAMY,EAAE,EAAE,EAAE,KAAK,KAAKA,EAAEA,IAAI,EAAE,EAAE,EAAE,KAAK,MAAM,OAAOA,IAAI,QAAQA,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE,KAAKqvE,GAAG,SAAS,CAAC,IAAI5vE,EAAEwuE,GAAG,UAAU,CAAC,MAAM,MAAM7uE,EAAE,SAAS,EAAE,KAAK,EAAE,EAAE,EAAE,OAAO+uE,GAAG1uE,EAAE,KAAKO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAKA,EAAE,CAAC,IAAIZ,EAAE,KAAK,EAAE,OAAO,KAAK,MAAM,MAAMY,GAAG,GAAG,KAAK,MAAMA,EAAE,EAAE,KAAK,KAAKA,EAAE,KAAK,MAAM,OAAO,KAAK,KAAKA,GAAGA,IAAI,QAAQA,EAAE,GAAG,EAAE,EAAE,EAAE,KAAKqvE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,KAAKY,CAAC,EAAE,CAAC,CAAC,CAAC,QAAQA,EAAEZ,EAAE,EAAE,GAAG,CAAC,GAAGY,GAAG,MAAMA,EAAE,EAAE,MAAM,KAAK,MAAM,KAAK,IAAI,WAAW,0DAA0D,EAAE,IAAI,WAAW,mNAAmN,KAAK,gBAAgB,EAAE,IAAIP,EAAE,KAAKwB,EAAE+rE,GAAG,KAAK5tE,GAAG4mB,EAAE,IAAI,EAAE,SAAS,CAAC,EAAE,OAAOqpD,GAAG,SAAS,CAAC,IAAInuE,EAAED,EAAE,MAAM,EAAE,OAAO,IAAIC,GAAGD,EAAE,MAAM,IAAI,MAAMxB,EAAE,SAAS,GAAG,QAAQO,EAAEkB,EAAE,SAAS,CAAC,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,KAAKlB,EAAE,CAAC,IAAIZ,EAAE,KAAK,EAAE,OAAO,KAAK,MAAM,MAAM,KAAK,KAAKY,EAAE,EAAEA,EAAE,KAAK,MAAM,MAAM,KAAK,MAAMA,EAAE,EAAE,KAAK,KAAK,EAAE,KAAKqvE,GAAG,UAAU,MAAMjwE,EAAE,SAAS,GAAG,KAAKY,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,SAAS,CAAC,GAAG,KAAK,OAAO,EAAE,EAAE,MAAM,IAAI,MAAM,gDAAgD,EAAE,OAAO,MAAM,KAAK,SAAS,GAAG,QAAQ,CAAC,CAAC,MAAM,gBAAgB,CAAC,GAAG,KAAK,OAAO,EAAE,EAAE,MAAM,IAAI,MAAM,gDAAgD,EAAE,OAAO,MAAM,KAAK,SAAS,GAAG,eAAe,CAAC,CAAC,EAAEosE,GAAG,gBAAgB,IAAI,SAASiD,GAAGnwE,EAAE,EAAE,KAAK,CAAC,OAAO,IAAI,cAAcktE,EAAE,CAAC,aAAa,CAAC,MAAM,GAAG,SAAS,EAAE,KAAK,KAAK,CAAC,CAAC,MAAM,UAAU,CAAC,OAAOltE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASstE,GAAGttE,EAAE,CAAC,OAAOmwE,GAAG,SAAStB,GAAG7uE,CAAC,EAAEA,EAAE,MAAM,CAAC,CAAC,SAAS6tE,GAAG7tE,EAAE,CAAC,GAAG,CAACkuE,GAAGluE,CAAC,EAAE,MAAM,IAAI,MAAM,mDAAmD,EAAE,IAAI,EAAE,GAAG,MAAM,QAAQA,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,EAAE,GAAG,KAAKF,EAAEE,GAAG,KAAK,KAAK,IAAI,EAAEF,EAAEE,GAAG,IAAI,UAAUF,aAAa,OAAO,QAAQE,KAAKF,EAAE,EAAE,GAAG,KAAKA,EAAEE,GAAG,KAAK,KAAK,IAAI,EAAEF,EAAEE,GAAG,IAAI,EAAE,OAAOiwE,GAAG,SAAS,CAAC,IAAIjwE,EAAE,MAAMouE,GAAGtuE,EAAEM,GAAG,CAAC,GAAGA,aAAa4sE,GAAG,MAAM,CAAC,MAAM5sE,EAAE,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG4tE,GAAG5tE,CAAC,EAAE,MAAM,CAAC,MAAM,KAAK,QAAQ,EAAE,EAAE,MAAM,IAAI,MAAM,2EAA2E,CAAC,CAAC,EAAE,OAAO6uE,GAAGjvE,EAAEkvE,GAAG,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC,SAASgB,GAAGpwE,EAAE,CAAC,GAAGA,IAAI,KAAK,OAAO,KAAK,IAAI,EAAEA,EAAE,GAAG,OAAOuuE,GAAG,CAAC,EAAE,CAAC,MAAM8B,GAAGrwE,CAAC,EAAE,QAAQ,EAAE,EAAE,CAAC,MAAM,KAAK,QAAQ,EAAE,CAAC,CAAC,SAASqwE,GAAGrwE,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,OAAOA,EAAE,aAAa6pB,GAAG+lB,GAAG5vC,CAAC,EAAEusB,GAAGvsB,CAAC,CAAC,CAAC,IAAIotE,GAAG,cAAcF,EAAE,CAAC,YAAYpsE,EAAE,CAAC,MAAM,EAAE,KAAK,MAAMA,CAAC,CAAC,MAAM,UAAU,CAAC,OAAO,MAAM,KAAK,MAAM,SAAS,GAAG,WAAW,EAAE,MAAM;AAAA,CACr1G,EAAE,IAAIP,IAAIA,EAAE,SAAS,IAAI,IAAIA,EAAEA,EAAE,MAAM,EAAE,EAAE,GAAGA,EAAE,CAAC,CAAC,EAAM+vE,GAAG,IAAIC,GAAG,OAAO,KAAK,EAAEC,GAAG,OAAO,OAAO,EAAEC,GAAG,OAAO,OAAO,EAAEC,GAAG,OAAO,iBAAiB,EAAEC,GAAG,OAAO,cAAc,EAAE1D,GAAG,cAAcC,EAAE,CAAC,YAAYpsE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,MAAMY,EAAE,KAAK,UAAU,GAAG,KAAK,gBAAgB,KAAK,KAAK,qBAAqB,GAAG,KAAK,cAAc,KAAK,KAAK,sBAAsB,GAAG,KAAK,UAAU,IAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK,IAAIssE,GAAGtsE,CAAC,EAAEZ,IAAIA,EAAE,CAAC,GAAG,KAAK,UAAUA,EAAE,YAAY,GAAG,KAAK,gBAAgBA,EAAE,YAAY,KAAK,cAAcA,EAAE,cAAc,KAAK,sBAAsBA,EAAE,sBAAsBA,EAAE,iBAAiB4mB,EAAE,OAAO5mB,EAAE,WAAW,KAAK,IAAI,gEAAgE,EAAE,KAAK,gBAAgB,GAAG,KAAK,UAAU,KAAK,KAAK,UAAUA,EAAE,UAAUA,EAAE,UAAU,GAAG,CAAC,MAAM,aAAa,CAAC,OAAO,KAAK,sBAAsB,MAAM,KAAK,eAAe,EAAE,KAAK,sBAAsB,OAAO,KAAK,KAAK,aAAa,EAAE,KAAK,eAAe,CAAC,MAAM,gBAAgB,CAAC,IAAIY,EAAE,MAAM,KAAK,oBAAoB,EAAE,GAAG,CAAC,KAAK,iBAAiB,CAACA,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,KAAK,iBAAiBA,GAAGgmB,EAAE,OAAOhmB,EAAE,SAAS,KAAK,gBAAgB,OAAO,IAAI,uCAAuC,KAAK,gBAAgB,OAAO,SAAS,EAAE,kEAAkEA,EAAE,OAAO,SAAS,EAAE,IAAI,EAAE,KAAK,kBAAkB,KAAK,gBAAgBA,GAAG,IAAIZ,EAAE,KAAK,gBAAgB,OAAO,CAACK,EAAEwB,KAAKxB,EAAEwB,GAAGxB,EAAEwB,GAAG,GAAG,EAAExB,GAAG,CAAC,CAAC,EAAE,EAAE,OAAO,KAAKL,CAAC,EAAE,OAAOK,GAAGL,EAAEK,GAAG,CAAC,EAAE,GAAGumB,EAAE,OAAO,EAAE,SAAS,EAAE,IAAI,iCAAiC,EAAE,SAAS,CAAC,EAAE,KAAK,eAAe,QAAQvmB,KAAK,OAAO,KAAK,KAAK,aAAa,EAAE,GAAG,KAAK,gBAAgB,QAAQA,CAAC,IAAI,GAAG,MAAM,IAAI,MAAM,YAAYA,EAAE,uEAAuE,KAAK,gBAAgB,SAAS,EAAE,IAAI,EAAE,KAAK,qBAAqB,EAAE,CAAC,MAAM,qBAAqB,CAAC,GAAG,KAAK,UAAU,CAAC,IAAI,EAAE,MAAM,MAAM,KAAK,KAAK,SAAS,GAAG,KAAK,EAAE,GAAG,EAAE,KAAK,MAAM,IAAI,MAAM,oCAAoC,EAAE,IAAID,EAAE,EAAE,MAAM,OAAO,KAAK,SAASA,EAAE,EAAE,CAAC,KAAM,QAAO,IAAI,CAAC,MAAM,UAAU,CAAC,KAAK,sBAAsB,MAAM,KAAK,eAAe,EAAE,IAAIQ,EAAE,MAAM,KAAK,KAAK,SAAS,EAAE,OAAO,KAAK,YAAYA,EAAEA,EAAE,KAAK,CAAC,GAAGA,EAAE,IAAIZ,GAAG,KAAK,gBAAgBA,CAAC,CAAC,CAAC,CAAC,gBAAgBY,EAAE,CAAC,IAAIZ,EAAE,KAAK,SAASY,CAAC,EAAE,EAAE,CAAC,EAAEP,EAAE,CAAC,EAAE,QAAQwB,EAAE,EAAEA,EAAE,KAAK,gBAAgB,OAAOA,IAAI,CAAC,IAAIC,EAAE,KAAK,gBAAgBD,GAAGE,EAAE,KAAK,cAAc,KAAK,cAAcD,GAAG,KAAK,GAAG,EAAE,KAAK,uBAAuB,CAACC,GAAG,CAAC,IAAIC,EAAEhC,EAAE6B,GAAG,EAAE,KAAK,GAAGG,IAAI,GAAG,GAAGD,GAAGA,EAAE,UAAU,OAAO,EAAEA,EAAE,YAAY,CAAC,GAAGA,IAAIA,EAAE,UAAUA,EAAE,SAAS,MAAM,IAAI,MAAM,mBAAmBD,4BAA4BlB,GAAG,EAAE,EAAE,MAAM,KAAK,CAAC,IAAI8B,EAAE,OAAOV,CAAC,EAAE,GAAG,MAAMU,CAAC,EAAEX,GAAGA,EAAE,QAAQ,OAAO,EAAE,KAAK,WAAWC,CAAC,EAAE,EAAEA,UAAU,CAACD,GAAG,CAACA,EAAE,MAAM,EAAEW,MAAO,QAAOX,EAAE,MAAM,CAAC,IAAI,UAAU,EAAEW,EAAE,MAAM,IAAI,QAAQ,EAAE,KAAK,MAAMA,CAAC,EAAE,MAAM,IAAI,OAAO,EAAE,KAAK,WAAWV,CAAC,EAAE,MAAM,QAAQ,EAAEU,CAAC,CAAC,CAACX,GAAGA,EAAE,QAAQ1B,EAAEyB,GAAG,EAAE,EAAEA,GAAG,CAAC,CAAC,CAAC,OAAO,OAAO,KAAKzB,CAAC,EAAE,SAAS,EAAE,EAAE,CAAC,GAAG,EAAE,GAAGA,CAAC,CAAC,CAAC,WAAWO,EAAE,CAAC,OAAOA,IAAI,KAAKA,EAAE,YAAY,IAAI,OAAO,EAAE,CAAC,CAAC,SAASA,EAAEZ,EAAE,GAAG,CAAC,IAAI,EAAE,CAAC,EAAEK,EAAE,EAAEwB,EAAEjB,EAAE,OAAOkB,EAAEuuE,GAAG,QAAQtuE,EAAE,EAAEA,EAAEF,EAAEE,IAAI,OAAOD,EAAE,CAAC,KAAKuuE,GAAG,OAAOzvE,EAAE,OAAOmB,CAAC,EAAE,CAAC,KAAKquE,GAAG/vE,EAAE0B,EAAE,EAAED,EAAEyuE,GAAG,MAAM,KAAK,KAAK,UAAU,GAAGlwE,EAAE0B,EAAE,EAAE,KAAK,YAAY,KAAK,KAAK,gBAAgB,MAAM,EAAE,KAAK,EAAE,EAAED,EAAEuuE,GAAG,MAAM,QAAQvuE,EAAEwuE,GAAGjwE,EAAE0B,EAAE,KAAK,CAAC,MAAM,KAAKuuE,GAAG,OAAO1vE,EAAE,OAAOmB,CAAC,EAAE,CAAC,KAAK,KAAK,UAAU,EAAE,KAAKnB,EAAE,UAAUP,EAAE0B,CAAC,CAAC,EAAED,EAAEuuE,GAAGhwE,EAAE0B,EAAE,EAAE,MAAM,QAAQ,CAAC,MAAM,KAAKwuE,GAAG,OAAO3vE,EAAE,OAAOmB,CAAC,EAAE,CAAC,KAAKquE,GAAGtuE,EAAE0uE,GAAG,MAAM,QAAQ,CAAC,MAAM,KAAKA,GAAG,OAAO5vE,EAAE,OAAOmB,CAAC,EAAE,CAAC,KAAK,KAAK,UAAU,EAAE,KAAKnB,EAAE,UAAUP,EAAE0B,EAAE,CAAC,CAAC,EAAED,EAAEuuE,GAAGhwE,EAAE0B,EAAE,EAAE,MAAM,KAAKquE,GAAGtuE,EAAEyuE,GAAG,MAAM,QAAQzuE,EAAE2uE,GAAG,KAAK,CAAC,MAAM,KAAKA,GAAG,OAAO7vE,EAAE,OAAOmB,CAAC,EAAE,CAAC,KAAKquE,GAAGtuE,EAAEyuE,GAAG,MAAM,QAAQ,CAAC,MAAM,QAAQ,CAAC,GAAGzuE,IAAI0uE,GAAG,EAAE,KAAK5vE,EAAE,UAAUP,EAAEwB,EAAE,CAAC,CAAC,EAAE,EAAE,KAAKjB,EAAE,UAAUP,CAAC,CAAC,EAAEL,GAAG,EAAE,SAAS,KAAK,gBAAgB,OAAO,MAAM,IAAI,MAAM,wCAAwC,KAAK,gBAAgB,qCAAqC,GAAG,EAAE,OAAO,CAAC,CAAC,EAAM0wE,GAAG,cAActB,EAAE,CAAC,YAAYxuE,EAAE,CAAC,MAAM,EAAE,KAAK,iBAAiBA,EAAE,KAAK,SAAS,GAAG,KAAK,QAAQA,EAAE,SAAS,KAAK,IAAIZ,EAAE,KAAK,KAAK,KAAK,OAAO,EAAE,GAAG,KAAK,QAAQ,GAAGA,EAAE,GAAGA,EAAE,IAAI,CAAC,OAAO,UAAUA,CAAC,EAAE,MAAM,IAAI,MAAM,gFAAgF,KAAK,SAAS,EAAE,GAAG,KAAK,UAAUY,EAAE,yBAAyB,GAAG,KAAK,aAAaA,EAAE,aAAa,KAAK,qBAAqBA,EAAE,sBAAsB,KAAK,QAAQ,KAAK,sBAAsBA,EAAE,sBAAsB,KAAK,sBAAsBA,EAAE,uBAAuB,EAAE,KAAK,mBAAmBA,EAAE,qBAAqB,GAAG,KAAK,gBAAgBA,EAAE,kBAAkB,GAAG,CAAC,KAAK,oBAAoB,CAAC,KAAK,gBAAgB,MAAM,IAAI,MAAM,sGAAsG,CAAC,CAAC,SAAS,CAAC,MAAM,YAAY,CAAC,aAAa,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC2Z,EAAE,EAAE,IAAI,YAAY,EAAE,MAAM,IAAI,MAAM,0DAA0D,EAAE,IAAIva,EAAE,IAAI0wE,GAAG9vE,CAAC,EAAE,OAAO,MAAMZ,EAAE,MAAM,EAAEA,CAAC,CAAC,MAAM,OAAO,CAAC,GAAG,CAAC,KAAK,OAAO,MAAM,UAAU,aAAa,aAAa,CAAC,MAAM,KAAK,uBAAuB,KAAK,GAAG,KAAK,sBAAsB,MAAM,EAAE,CAAC,CAAC,OAAO,EAAN,CAAS,MAAM,IAAI,MAAM,iDAAiD,EAAE,SAAS,CAAC,CAAC,GAAG,CAAC,KAAK,OAAO,MAAM,IAAI,MAAM,yCAAyC,EAAE,IAAIY,EAAE,OAAO,cAAc,OAAO,mBAAmB,GAAG,KAAK,aAAa,IAAIA,EAAE,CAAC,KAAK,aAAa,KAAK,aAAa,KAAK,aAAa,mBAAmB,KAAK,aAAa,aAAa,KAAK,aAAa,MAAM,IAAI,MAAM,wCAAwC,KAAK,yBAAyB,KAAK,aAAa,YAAY,EAAE,IAAIZ,EAAE,KAAK,aAAa,wBAAwB,KAAK,MAAM,EAAE,KAAK,SAAS,KAAK,aAAa,eAAe,EAAE,KAAK,SAAS,QAAQ,KAAK,QAAQ,EAAE,KAAK,SAAS,sBAAsB,KAAK,sBAAsBA,EAAE,QAAQ,KAAK,QAAQ,EAAE,KAAK,SAAS,IAAI,aAAa,KAAK,OAAO,EAAE,KAAK,SAAS,IAAI,aAAa,KAAK,OAAO,CAAC,CAAC,MAAM,MAAM,CAAC,GAAG,KAAK,SAAS,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,IAAIY,EAAEZ,EAAE,EAAE,MAAM,KAAK,aAAa,EAAE,GAAG,KAAK,mBAAmB,CAAC,IAAIK,EAAE,KAAK,aAAa,EAAE,aAAa,EAAEO,EAAE,KAAK,4BAA4BP,EAAE,CAAC,KAAK,UAAU,KAAK,qBAAqB,CAAC,CAAC,CAAC,CAAC,GAAG,KAAK,gBAAgB,CAAC,IAAIA,EAAE,KAAK,aAAa,EAAE,aAAa,EAAEL,EAAE,KAAK,4BAA4BK,EAAE,CAAC,KAAK,UAAU,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,YAAYO,EAAE,SAASZ,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,MAAM,SAAS,CAAC,OAAO,MAAM,KAAK,KAAK,GAAG,KAAK,CAAC,MAAM,cAAc,CAAC,IAAIY,EAAE,CAAC,EAAEZ,EAAE,CAAC,EAAE,EAAE,EAAE,OAAO,IAAI,QAAQK,GAAG,CAAC,IAAIwB,EAAE,YAAY,IAAI,CAAC,KAAK,qBAAqB,KAAK,SAAS,sBAAsB,KAAK,QAAQ,EAAE,KAAK,SAAS,KAAK,MAAMxB,EAAE,CAAC,cAAcO,EAAE,cAAcZ,CAAC,CAAC,EAAEY,EAAE,KAAK,KAAK,SAAS,MAAM,EAAE,KAAK,oBAAoB,CAAC,GAAG,KAAK,kBAAkB,KAAK,SAAS,uBAAuB,KAAK,QAAQ,EAAEZ,EAAE,KAAK,KAAK,SAAS,MAAM,CAAC,GAAG,EAAE,IAAI,KAAK,YAAY,cAAc6B,CAAC,EAAExB,EAAE,CAAC,cAAcO,EAAE,cAAcZ,CAAC,CAAC,EAAE,EAAE,KAAK,QAAQ,KAAK,aAAa,GAAG,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,WAAW,KAAK,SAAS,GAAG,KAAK,SAAS,WAAW,EAAE,KAAK,aAAa,MAAM,EAAE,KAAK,QAAQ,MAAM,KAAK,OAAO,UAAU,EAAE,OAAO,GAAG,KAAK,OAAO,UAAU,EAAE,GAAG,KAAK,EAAE,CAAC,SAAS,CAAC,MAAM,IAAI,MAAM,iDAAiD,CAAC,CAAC,eAAe,CAAC,OAAO,KAAK,YAAY,CAAC,aAAaY,EAAE,CAAC,IAAIZ,EAAEY,EAAE,GAAG,OAAO,EAAE,IAAI,aAAaA,EAAE,OAAOZ,CAAC,EAAE,OAAOY,EAAE,QAAQ,CAACP,EAAEwB,IAAI,EAAE,IAAIxB,EAAEwB,EAAE7B,CAAC,CAAC,EAAE,CAAC,CAAC,4BAA4BY,EAAEZ,EAAE,CAAC,IAAI,EAAE,IAAI,aAAa4mB,EAAE,cAAc5mB,CAAC,CAAC,EAAE,OAAO,EAAE,IAAIY,EAAE,EAAE,OAAOA,EAAE,MAAM,EAAEyrB,GAAG,EAAErsB,CAAC,CAAC,CAAC,EAAM2wE,GAAG,cAAcvB,EAAE,CAAC,YAAYxuE,EAAEZ,EAAE,CAAC,GAAG,MAAM,EAAE,KAAK,mBAAmBY,EAAE,KAAK,aAAaZ,EAAE,KAAK,SAAS,GAAG,KAAK,OAAO,GAAG,KAAK,aAAa,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK,SAAS,CAAC,KAAK,aAAa,aAAa,KAAK,aAAa,WAAW,EAAE,KAAK,WAAWiwC,GAAG,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,aAAa,WAAW,CAAC,IAAI,EAAE,KAAK,aAAa,YAAY,EAAE,KAAK,mBAAmB,MAAM5vC,EAAE,KAAK,aAAa,aAAa,EAAE,KAAK,mBAAmB,OAAOwB,GAAG,EAAE,GAAG,EAAEC,GAAG,EAAEzB,GAAG,EAAE0B,EAAEF,EAAE,EAAEG,EAAE3B,EAAEyB,EAAE,KAAK,QAAQouC,GAAG,CAACpuC,EAAED,EAAEG,EAAED,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,KAAK,QAAQmuC,GAAG,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,QAAQ,CAAC,aAAa,OAAOtvC,EAAEZ,EAAE,CAAC,EAAE,CAAC,GAAG,CAACua,EAAE,EAAE,IAAI,YAAY,EAAE,MAAM,IAAI,MAAM,0DAA0D,EAAE,GAAG,CAAC3Z,EAAE,CAAC,GAAGA,EAAE,SAAS,cAAc,OAAO,EAAE,CAACZ,EAAE,aAAa,CAACA,EAAE,aAAa,MAAM,IAAI,MAAM,wGAAwG,EAAEY,EAAE,MAAMZ,EAAE,YAAYY,EAAE,OAAOZ,EAAE,YAAY,CAAC,IAAI,EAAE,IAAI2wE,GAAG/vE,EAAEZ,CAAC,EAAE,OAAO,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC,MAAM,OAAO,CAAC,KAAK,aAAa,YAAY4mB,EAAE,OAAO,KAAK,aAAa,aAAa,QAAQ,KAAK,aAAa,aAAa,cAAc,IAAI,+BAA+B,KAAK,aAAa,oDAAoD,EAAE,GAAG,CAAC,KAAK,OAAO,MAAM,UAAU,aAAa,aAAa,CAAC,MAAM,CAAC,SAAS,KAAK,aAAa,SAAS,WAAW,KAAK,aAAa,WAAW,KAAK,aAAa,WAAW,OAAO,MAAM,KAAK,mBAAmB,MAAM,OAAO,KAAK,mBAAmB,MAAM,CAAC,CAAC,CAAC,OAAOhmB,EAAN,CAAS,MAAMA,EAAE,QAAQ,iDAAiDA,EAAE,UAAUA,CAAC,CAAC,GAAG,CAAC,KAAK,OAAO,MAAM,IAAI,MAAM,qCAAqC,EAAE,GAAG,CAAC,KAAK,mBAAmB,UAAU,KAAK,MAAM,OAAOA,EAAN,CAAS,QAAQ,IAAIA,CAAC,EAAE,KAAK,mBAAmB,IAAI,OAAO,IAAI,gBAAgB,KAAK,MAAM,CAAC,CAAC,OAAO,KAAK,mBAAmB,KAAK,EAAE,KAAK,SAAS,GAAG,IAAI,QAAQA,GAAG,CAAC,KAAK,mBAAmB,iBAAiB,IAAI,CAACA,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,MAAM,CAAC,GAAG,KAAK,SAAS,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,IAAIA,EAAE,GAAG,CAACA,EAAEq0B,GAAG,WAAW,KAAK,kBAAkB,CAAC,OAAOj1B,EAAN,CAAS,MAAM,IAAI,MAAM,4CAA4C,KAAK,UAAUA,CAAC,GAAG,CAAC,CAAC,GAAG,KAAK,OAAO,GAAG,CAAC,MAAM,CAAC,MAAM,KAAK,mBAAmBY,CAAC,EAAE,KAAK,EAAE,CAAC,OAAOZ,EAAN,CAAS,MAAM,IAAI,MAAM,oCAAoCA,EAAE,SAAS,CAAC,QAAC,CAAQY,EAAE,QAAQ,CAAC,KAAM,OAAM,CAAC,MAAMA,EAAE,KAAK,EAAE,CAAC,CAAC,cAAc,CAAC,MAAM,CAAC,EAAE,KAAK,aAAa,aAAa,KAAK,aAAa,eAAe,KAAK,mBAAmB,QAAQ,KAAK,aAAa,aAAa,KAAK,mBAAmB,SAAS,KAAK,aAAa,cAAc,CAAC,mBAAmBA,EAAE,CAAC,OAAO2yB,EAAE,IAAI,CAAC,IAAIvzB,EAAE0jC,GAAGnT,GAAE3vB,EAAE,SAAS,EAAE,CAAC,EAAE,EAAE,EAAEy4C,GAAG,cAAcr5C,EAAE,KAAK,QAAQ,KAAK,WAAW,KAAK,SAAS,UAAU,EAAE,IAAIK,EAAE,EAAE,MAAM,OAAO27B,EAAE,EAAE37B,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,SAAS,CAAC,OAAO,MAAM,KAAK,KAAK,GAAG,KAAK,CAAC,MAAM,CAAC,KAAK,OAAO,UAAU,EAAE,QAAQ,GAAG,EAAE,KAAK,CAAC,EAAE,GAAG,CAAC,KAAK,mBAAmB,UAAU,IAAI,OAAO,EAAN,CAAS,QAAQ,IAAI,CAAC,EAAE,KAAK,mBAAmB,IAAI,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC,SAAS,CAAC,MAAM,IAAI,MAAM,iDAAiD,CAAC,CAAC,EAAMuwE,GAAG,KAAK,CAAC,EAAMC,GAAG,cAAczB,EAAE,CAAC,MAAMxuE,EAAE,CAAC,OAAO,IAAIkwE,GAAG,KAAKlwE,CAAC,CAAC,CAAC,EAAEkwE,GAAG,cAAcD,EAAE,CAAC,YAAYjwE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,KAAK,IAAImwE,GAAGnwE,EAAEZ,CAAC,CAAC,CAAC,SAAS,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,KAAK,KAAK,CAAC,CAAC,EAAE+wE,GAAG,cAAcf,EAAE,CAAC,YAAYpvE,EAAEZ,EAAE,CAAC,MAAM,EAAE,KAAK,SAASY,EAAE,KAAK,UAAUZ,EAAE,KAAK,UAAU,EAAE,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,eAAe,KAAK,aAAa,CAAC,MAAM,MAAM,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAE,GAAGA,EAAE,KAAK,OAAO,KAAK,YAAY,GAAG,IAAI,KAAK,YAAY,KAAK,KAAK,SAAS,EAAE,KAAK,UAAU,GAAG,IAAI,IAAIZ,EAAEY,EAAE,MAAM,MAAM,KAAK,SAAS,EAAEZ,EAAE,GAAG,KAAK,UAAUA,EAAE,GAAG,QAAQ,KAAKA,EAAE,MAAM,EAAE,EAAE,EAAE,KAAK,YAAY,KAAK,CAAC,EAAE,OAAO,KAAK,UAAUA,EAAEA,EAAE,OAAO,GAAG,EAAE,CAAC,EAAMgxE,GAAG,cAAc5B,EAAE,CAAC,YAAY,CAAC,OAAO,IAAI6B,GAAG,IAAI,CAAC,CAAC,EAAEA,GAAG,cAAcJ,EAAE,CAAC,YAAYjwE,EAAE,CAAC,MAAM,EAAE,KAAK,SAASA,EAAE,KAAK,KAAK,IAAIswE,GAAGtwE,CAAC,CAAC,CAAC,SAAS,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,KAAK,KAAK,CAAC,CAAC,EAAEswE,GAAG,cAAclB,EAAE,CAAC,YAAYpvE,EAAE,CAAC,GAAG,MAAM,EAAE,KAAK,SAASA,EAAE2Z,EAAE,EAAE,IAAI,YAAY,EAAE,KAAK,QAAQ,IAAI,YAAY,OAAO,MAAM,CAAC,GAAG,CAAC,cAAcva,CAAC,EAAE+F,GAAG,EAAE,KAAK,QAAQ,IAAI/F,EAAE,MAAM,CAAC,CAAC,CAAC,SAAS,CAAC,MAAM,GAAG,KAAK,SAAS,QAAQ,WAAW,CAAC,MAAM,MAAM,CAAC,IAAIY,EAAE,MAAM,KAAK,SAAS,KAAK,EAAEZ,EAAE,GAAGY,EAAE,KAAK,MAAM,GAAGZ,EAAEY,EAAE,MAAM,IAAI,EAAE,OAAO2Z,EAAE,EAAE,IAAI,YAAY,EAAE,EAAE,KAAK,QAAQ,OAAOva,EAAE,CAAC,OAAO,EAAE,CAAC,EAAE,EAAE,KAAK,QAAQ,MAAM,OAAO,KAAKA,EAAE,MAAM,CAAC,EAAE,KAAK,YAAY,KAAK,CAAC,EAAE,EAAE,CAAC,EAAMmxE,GAAG,cAAcH,EAAE,CAAC,YAAYpwE,EAAEZ,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,KAAKY,EAAE,KAAK,QAAQZ,EAAE4mB,EAAE,OAAOhmB,aAAa,aAAa2Z,EAAE,EAAE,IAAI,YAAY,EAAE3Z,aAAa,MAAMA,aAAa,KAAK,IAAI,IAAI,sEAAsE,EAAE,KAAK,OAAOZ,EAAE,QAAQ,EAAE,KAAK,UAAUA,EAAE,WAAW,KAAK,IAAI,CAAC,SAAS,CAAC,MAAM,cAAc,KAAK,MAAM,CAAC,MAAM,MAAM,CAAC,OAAO,KAAK,SAAS,KAAK,gBAAgB,WAAW,KAAK,KAAK,WAAW,KAAK,KAAK,MAAM,CAAC,MAAM,KAAK,KAAK,EAAE,EAAE,CAAC,MAAM,MAAM,IAAI,QAAQ,CAAC,EAAEI,IAAI,CAAC,IAAIC,EAAE,KAAK,OAAO,KAAK,UAAU,GAAG,KAAK,gBAAgB,WAAW,EAAE,IAAI,WAAW,KAAK,KAAK,MAAM,KAAK,OAAOA,CAAC,CAAC,CAAC,MAAM,CAAC,IAAIwB,EAAE,IAAI,WAAWA,EAAE,OAAOE,GAAG,CAAC,IAAIC,EAAEH,EAAE,OAAO,GAAGG,aAAa,cAAcA,EAAE,IAAI,WAAWA,CAAC,GAAG,EAAEA,aAAa,YAAY,OAAO5B,EAAE,IAAI,UAAU,mCAAmC,CAAC,EAAE,EAAE4B,CAAC,CAAC,EAAEH,EAAE,QAAQE,GAAG3B,EAAE,IAAI,MAAM,SAAS,CAAC,EAAEyB,EAAE,QAAQE,GAAG3B,EAAE,IAAI,MAAM2B,EAAE,IAAI,CAAC,EAAE,IAAID,EAAE,KAAK,KAAK,MAAM,KAAK,OAAOzB,CAAC,EAAEwB,EAAE,kBAAkBC,CAAC,CAAC,CAAC,KAAK,OAAOzB,CAAC,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,EAAE,eAAe+wE,GAAGtxE,EAAE,EAAE,CAAC,EAAEE,EAAE,CAAC,IAAII,EAAEC,EAAE,OAAOP,GAAG,SAASM,EAAEN,GAAGM,EAAEN,EAAE,IAAIO,EAAEgxE,GAAGvxE,CAAC,GAAG,IAAI,EAAE,MAAME,GAAG4mB,EAAE,OAAOxmB,EAAEC,CAAC,EAAE,GAAG,EAAE,GAAG,CAAC,IAAI,EAAE,IAAI,WAAW,MAAM,EAAE,YAAY,CAAC,EAAE,OAAO,IAAI8wE,GAAG,EAAE,CAAC,CAAC,KAAM,OAAM,IAAI,MAAM,EAAE,UAAU,CAAC,CAAC,IAAIE,GAAGvxE,IAAI,CAAC,OAAOA,EAAE,OAAO,QAAQA,EAAE,QAAQ,KAAKA,EAAE,KAAK,KAAKA,EAAE,KAAK,YAAYA,EAAE,YAAY,MAAMA,EAAE,MAAM,SAASA,EAAE,SAAS,SAASA,EAAE,SAAS,UAAUA,EAAE,SAAS,GAAG,SAASwxE,GAAGxxE,EAAE,CAAC,OAAO,OAAOA,GAAG,UAAUA,EAAE,MAAM,EAAE,CAAC,IAAI,SAAS,CAAC,IAAImtE,GAAG,cAAc2D,EAAE,CAAC,YAAYhwE,EAAEZ,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,MAAMY,EAAE,KAAK,QAAQZ,CAAC,CAAC,MAAM,UAAU,CAAC,GAAGsxE,GAAG,KAAK,KAAK,GAAG/2D,EAAE,EAAE,IAAI,SAAS,EAAE,CAAC,IAAI3Z,EAAEoF,GAAG,EAAE,KAAK,MAAMpF,EAAE,aAAa,KAAK,MAAM,MAAM,CAAC,CAAC,CAAC,CAAC,OAAO,IAAIuwE,GAAG,KAAK,MAAM,KAAK,OAAO,CAAC,CAAC,EAAMhE,GAAG,cAAcyD,EAAE,CAAC,YAAYhwE,EAAEZ,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,KAAK,IAAIY,EAAE,KAAK,YAAYZ,CAAC,CAAC,MAAM,UAAU,CAAC,OAAOsxE,GAAG,KAAK,GAAG,EAAE,IAAIrE,GAAG,KAAK,IAAI,KAAK,WAAW,EAAE,SAAS,EAAEmE,GAAG,KAAK,IAAI,KAAK,WAAW,CAAC,CAAC,EAAE,SAAS/D,GAAGvtE,EAAE,EAAE,CAAC,EAAE,CAAC,OAAO,IAAIitE,GAAG,IAAII,GAAGrtE,CAAC,EAAE,CAAC,CAAC,CAAC,SAASwtE,GAAGxtE,EAAE,CAAC,IAAI,EAAE+uE,GAAG/uE,CAAC,EAAE,OAAOmwE,GAAG,SAAS,CAAC,CAAC,CAAC,SAAS1C,GAAGztE,EAAE,CAAC,OAAOmwE,GAAG,SAAS,CAAC,IAAI,EAAE,MAAMnwE,EAAE,EAAE,OAAO+uE,GAAG,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,eAAenB,GAAG5tE,EAAE,EAAE,CAAC,OAAO6wE,GAAG,OAAO7wE,EAAE,CAAC,CAAC,CAAC,eAAe0tE,GAAG1tE,EAAE,CAAC,OAAO4wE,GAAG,OAAO5wE,CAAC,CAAC,CAAC,IAAI2tE,GAAG,QAAQ,SAAS8D,GAAGzxE,EAAE,EAAE,CAAC,MAAM,QAAQA,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAGA,EAAE,QAAQE,GAAG,CAACA,GAAG,MAAM4mB,EAAE,OAAO5mB,EAAE,QAAQ,YAAY,IAAI,GAAG,0DAA0D,CAAC,CAAC,CAAC,CAAC,IAAIwxE,GAAGrzB,GAAG,UAAUszB,GAAG,cAAcv6D,EAAE,CAAC,aAAa,CAAC,MAAM,EAAE,KAAK,UAAU,GAAG,KAAK,SAAS,GAAG,KAAK,KAAK,IAAID,GAAG,KAAKmc,GAAG,CAAC,CAAC,CAAC,YAAY,CAAC,OAAOq+C,GAAG,YAAY,CAAC,MAAM7wE,EAAEZ,EAAE,EAAE,CAAC,KAAK,WAAW,KAAK,SAAS,GAAGua,EAAE,EAAE,IAAI,SAAS,GAAG+/B,EAAE,KAAK;AAAA;AAAA;AAAA,6BAGv6b,GAAG,IAAIj6C,EAAE,CAAC,GAAG,KAAK,WAAW,CAAC,EAAE,OAAO,KAAK,KAAK,IAAIA,EAAE,CAAC,OAAOO,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC,EAAEP,CAAC,CAAC,eAAeO,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,GAAGL,IAAI,UAAU,GAAG,MAAM,EAAE,OAAO,GAAG4mB,EAAE,SAAS,EAAE,EAAE,EAAE,CAAC,IAAI/kB,EAAE,EAAE,IAAIC,GAAG8kB,EAAE,aAAa9kB,CAAC,CAAC,EAAEzB,EAAE,KAAK,MAAMwB,EAAEjB,EAAEZ,CAAC,CAAC,MAAMK,EAAE,KAAK,MAAM,EAAEO,EAAEZ,CAAC,EAAE,MAAM,CAAC,OAAOK,EAAE,MAAMO,EAAE,MAAMZ,CAAC,CAAC,CAAC,SAASY,EAAE,CAAC,OAAO,KAAK,KAAK,IAAIA,CAAC,EAAE,KAAK,KAAK,IAAIA,CAAC,EAAE,SAAS,CAAC,CAAC,OAAOA,EAAE,CAAC,IAAIZ,EAAE,KAAK,KAAK,IAAIY,CAAC,EAAEZ,EAAE,UAAU,CAAC,OAAOY,EAAE,CAAC,GAAG,KAAK,KAAK,IAAIA,CAAC,EAAE,CAAC,IAAIZ,EAAE,KAAK,KAAK,IAAIY,CAAC,EAAEZ,EAAE,UAAU,CAAC,CAAC,KAAKY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,KAAK,IAAIjB,EAAE,CAAC,OAAOZ,EAAE,MAAMK,EAAE,SAASwB,CAAC,CAAC,CAAC,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK,WAAW,CAAC,CAAC,MAAM,KAAKjB,EAAE,CAAC,OAAO,KAAK,SAASA,CAAC,CAAC,CAAC,SAASA,EAAE,CAAC,GAAG,CAAC,MAAMZ,EAAE,mBAAmB,CAAC,EAAE,KAAK,KAAK,IAAIY,CAAC,EAAE,GAAGZ,IAAI,YAAY,CAAC,IAAIK,EAAE,KAAK,SAAS,EAAE,KAAK,MAAM,EAAEwB,EAAE,KAAK,SAAS,EAAE,KAAK,MAAM,EAAE,OAAOy4C,EAAE,uBAAuBj6C,EAAEwB,CAAC,CAAC,CAAC,OAAO,KAAK,KAAK,IAAIjB,CAAC,EAAE,MAAM,CAAC,WAAWA,EAAE,CAAC,IAAIZ,EAAE,KAAK,SAASY,EAAE,MAAM,EAAE,GAAGA,EAAE,QAAQ,SAAS,GAAG,CAAC,IAAI,EAAEZ,EAAE,IAAIK,GAAGumB,EAAE,aAAavmB,CAAC,CAAC,EAAE,OAAOgwB,GAAGzvB,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,OAAO,EAAN,CAAS,MAAM,IAAI,MAAM,kDAAkD,CAAC,CAAC,OAAOyvB,GAAGzvB,EAAE,MAAMA,EAAE,MAAMZ,CAAC,CAAC,CAAC,WAAWY,EAAEZ,EAAE,EAAE,CAAC,OAAOozB,GAAG,EAAE,yBAAyB,KAAK,eAAepzB,EAAE,EAAEY,CAAC,EAAE,IAAI,CAAC,CAAC,YAAYA,EAAEZ,EAAE,GAAG,CAAC,GAAG,KAAK,KAAK,IAAIY,CAAC,EAAE,CAAC,GAAG,KAAK,KAAK,IAAIA,CAAC,EAAE,WAAW,CAACZ,GAAG,KAAK,KAAK,IAAIY,CAAC,EAAE,SAAS,EAAE,MAAM,GAAG,GAAG,CAAC,mBAAmB,CAAC,EAAE,KAAK,KAAK,IAAIA,CAAC,EAAE,GAAG,OAAO,KAAK,YAAY,EAAE,KAAK,OAAO,EAAE,EAAE,KAAK,YAAY,EAAE,KAAK,OAAO,EAAE,GAAG,KAAK,KAAK,OAAOA,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,8BAA8BA,EAAE,CAAC,KAAK,YAAYA,EAAE,MAAM,CAAC,CAAC,MAAM,KAAKA,EAAE,CAAC,IAAIZ,EAAE4mB,EAAE,IAAI,EAAE,OAAOhmB,EAAE,EAAE,CAAC,SAASgmB,EAAE,IAAI,EAAE5mB,CAAC,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,WAAW,GAAG,QAAQ,CAAC,oHAAoH,CAAC,CAAC,CAAC,MAAMY,EAAE,CAAC2wE,GAAG,CAAC3wE,CAAC,EAAE,OAAO,EAAE,IAAIZ,EAAE,KAAK,SAASY,EAAE,MAAM,EAAE,OAAO4wE,GAAG5wE,EAAE,MAAMZ,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,gBAAgB,CAAC,MAAO,GAAE,CAAC,SAAS,CAAC,OAAO,MAAM,QAAQ,CAAC,CAAC,EAAEyxE,GAAG,WAAW,EAAE,IAAIC,GAAG,CAAC,EAAExxE,GAAGwxE,GAAG,CAAC,QAAQ,IAAIC,GAAG,aAAa,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,SAAS,IAAIC,GAAG,SAAS,IAAIC,GAAG,WAAW,IAAIC,GAAG,UAAU,IAAIC,GAAG,QAAQ,IAAIC,GAAG,UAAU,IAAIC,GAAG,UAAU,IAAIC,GAAG,aAAa,IAAIC,GAAG,aAAa,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,YAAY,IAAIC,GAAG,cAAc,IAAIC,GAAG,SAAS,IAAIC,GAAG,aAAa,IAAIC,GAAG,QAAQ,IAAIC,GAAG,QAAQ,IAAIC,GAAG,YAAY,IAAIC,GAAG,YAAY,IAAIC,GAAG,aAAa,IAAIC,GAAG,QAAQ,IAAIC,GAAG,aAAa,IAAIC,GAAG,SAAS,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,yBAAyB,IAAIC,GAAG,UAAU,IAAIC,GAAG,UAAU,IAAIC,GAAG,YAAY,IAAIC,GAAG,YAAY,IAAIC,GAAG,cAAc,IAAIC,GAAG,UAAU,IAAIC,GAAG,wBAAwB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,SAAS,IAAIC,GAAG,sBAAsB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,QAAQ,IAAIC,GAAG,SAAS,IAAIC,GAAG,SAAS,IAAIC,GAAG,cAAc,IAAIC,GAAG,WAAW,IAAIC,EAAE,CAAC,EAAE,SAASf,GAAG7zE,EAAE,CAAC,IAAI,EAAE,IAAI,aAAaA,EAAE,MAAM,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,EAAEA,GAAG,KAAK,IAAIF,EAAEE,EAAE,EAAE,OAAO,CAAC,CAAC,IAAI20E,GAAG70E,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,EAAEA,EAAE,OAAOE,EAAEF,EAAE,QAAQyxE,GAAG,EAAE,KAAK,EAAE,IAAInxE,EAAE,IAAI,aAAawmB,EAAE,cAAc,EAAE,KAAK,CAAC,EAAEvmB,EAAEL,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,OAAOI,EAAEuzE,GAAGtzE,CAAC,EAAEL,EAAE,WAAWI,EAAE,EAAE,MAAM,EAAE,KAAK,CAAC,EAAEw0E,GAAG,CAAC,WAAW55D,GAAG,YAAY,MAAM,WAAW25D,EAAE,EAAE,SAASE,GAAG/0E,EAAE,CAAC,MAAM,CAAC,EAAEE,EAAEI,EAAEC,EAAE,IAAI,CAAC,IAAI,EAAEi6C,EAAE,2BAA2B,EAAEt6C,CAAC,EAAE+B,EAAE,EAAE,OAAOC,EAAE4kB,EAAE,eAAe,CAAC,EAAE3kB,EAAE2kB,EAAE,cAAc,CAAC,EAAElkB,EAAEkkB,EAAE,uBAAuB,EAAE3kB,CAAC,EAAEU,EAAE,EAAE,OAAOC,EAAE5C,EAAE,OAAO6C,EAAE+jB,EAAE,eAAe,CAAC,EAAE9jB,EAAE8jB,EAAE,eAAe5mB,CAAC,EAAEgE,EAAEs2C,EAAE,iBAAiB,EAAE,CAAC,EAAE,EAAEA,EAAE,iBAAiBt6C,EAAE,CAAC,EAAE,GAAGgE,EAAE,OAAO,EAAE,SAAS,EAAE,QAAQqD,EAAE,EAAEA,EAAE3E,EAAE,OAAO,EAAE2E,EAAE3E,EAAE2E,GAAGvH,EAAEM,EAAEiH,EAAEjH,EAAE,QAAQC,EAAEgH,EAAEhH,EAAE,OAAO,MAAO,SAAQgH,EAAE,EAAEA,EAAE3E,EAAE,OAAO,EAAE2E,EAAE,CAAC,IAAI,EAAEuf,EAAE,WAAWvf,EAAEtF,EAAEC,CAAC,EAAEyC,EAAE,EAAE,MAAM,CAAC9B,CAAC,EAAEqB,EAAE,QAAQsD,GAAG7C,EAAE6C,GAAG,CAAC,EAAE,IAAI5C,EAAEkiB,EAAE,WAAWniB,EAAE9B,EAAEE,CAAC,EAAE8B,EAAE,EAAE,MAAM,CAAC/B,CAAC,EAAE,EAAE,QAAQ0E,GAAG3C,EAAE2C,GAAG,CAAC,EAAE,IAAI1C,EAAEgiB,EAAE,WAAWjiB,EAAE/B,EAAEE,CAAC,EAAEJ,EAAE2E,GAAGvH,EAAEM,EAAEsE,GAAGrE,EAAEuE,EAAE,CAAC,CAAC,MAAM,CAAClC,EAAE,CAAC,CAAC,CAAC,CAAC,SAASoyE,GAAGh1E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,KAAKC,CAAC,EAAE,EAAE,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,EAAEJ,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO0B,EAAE/B,EAAE,eAAeI,EAAE,MAAM,WAAW,EAAE4B,EAAEhC,EAAE,KAAK,IAAI+B,EAAE,MAAM,EAAE,OAAOC,EAAE,mBAAmB,CAAC,KAAKhC,EAAE,eAAeI,EAAE,MAAM,UAAU,CAAC,EAAE,KAAKJ,EAAE,eAAeK,EAAE,MAAM,UAAU,CAAC,CAAC,EAAE0B,CAAC,CAAC,IAAIgzE,GAAG,CAAC,WAAWr4D,GAAG,YAAY,MAAM,WAAWo4D,EAAE,EAAE,SAASE,GAAGl1E,EAAE,EAAEE,EAAE,UAAU,CAAC,GAAGA,IAAI,YAAY,CAAC,IAAIK,EAAE20E,GAAGl1E,EAAE,EAAE,SAAS,EAAE,EAAEk1E,GAAGl1E,EAAE,EAAE,SAAS,EAAE,OAAOg1E,GAAG,CAAC,OAAO,CAAC,KAAKz0E,EAAE,KAAK,CAAC,EAAE,QAAQP,CAAC,CAAC,CAAC,CAAC,IAAIM,EAAEwmB,EAAE,oBAAoBA,EAAE,cAAc,CAAC,EAAE5mB,CAAC,EAAE,OAAOF,EAAE,eAAe,EAAEE,EAAEI,CAAC,CAAC,CAAC,SAAS60E,GAAGn1E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,OAAOJ,EAAE,OAAOI,EAAE,MAAM,EAAE,CAAC,OAAOA,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,IAAI80E,GAAG,CAAC,WAAW91D,GAAG,YAAY,MAAM,WAAW61D,EAAE,EAAE,SAASE,GAAGr1E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,mBAAmB,KAAK,EAAEJ,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,OAAOL,EAAE,eAAeK,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,CAAC,IAAI+0E,GAAG,CAAC,WAAW7yD,GAAG,YAAY,MAAM,WAAW4yD,EAAE,EAAE,SAASrD,GAAGhyE,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAGA,IAAI,QAAQ,CAAC,IAAIC,EAAE,WAAW,KAAKP,CAAC,EAAE,MAAM,CAAC,EAAE,QAAQO,CAAC,CAAC,CAAC,GAAGD,IAAI,OAAO,CAAC,IAAIC,EAAEumB,EAAE,aAAa,CAAC,CAAC,EAAE5mB,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE60E,GAAG,CAAC9yE,EAAEC,IAAID,IAAIC,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,EAAElC,EAAEO,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,MAAM,IAAI,MAAM,iCAAiCL,QAAQI,GAAG,CAAC,CAAC,SAASi1E,GAAGv1E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,GAAG,IAAI,YAAY,CAAC,GAAGC,EAAE,QAAQ,YAAY,OAAO40E,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAI0C,EAAEsyE,GAAGh1E,EAAEK,EAAE,MAAMA,EAAE,KAAK,EAAEsC,EAAE0yE,GAAG,CAAC,OAAO,CAAC,EAAEh1E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAE4C,EAAEkyE,GAAG,CAAC,OAAO,CAAC,KAAKnyE,EAAE,KAAKD,CAAC,EAAE,QAAQ1C,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B0C,CAAC,EAAE1C,EAAE,8BAA8B2C,CAAC,EAAEC,CAAC,CAAC,GAAGvC,EAAE,QAAQ,YAAY,CAAC,IAAIqC,EAAEyyE,GAAG,CAAC,OAAO,CAAC,MAAM90E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE2C,EAAE0yE,GAAG,CAAC,OAAO,CAAC,EAAE3yE,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B0C,CAAC,EAAEC,CAAC,CAAC,GAAG,CAACikB,EAAE,gBAAgBvmB,EAAE,MAAM,CAAC,EAAE,CAAC,IAAIqC,EAAEuyE,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,MAAM,CAAC,OAAO0C,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAM,CAAC,CAAC,CAAC,IAAI,EAAE1C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAAC0B,EAAEC,EAAEC,CAAC,EAAE6vE,GAAG,EAAEzxE,EAAE,MAAMA,EAAE,MAAM,CAAC,EAAE,OAAOL,EAAE,eAAe+B,EAAEC,EAAEC,CAAC,CAAC,CAAC,IAAIqzE,GAAG,CAAC,WAAW/4D,GAAG,YAAY,MAAM,WAAW84D,EAAE,EAAE,SAASE,GAAGz1E,EAAE,EAAEE,EAAEI,EAAE,CAAC,OAAOJ,GAAG,KAAK,CAAC,CAAC,OAAOK,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE0B,CAAC,EAAE1B,EAAE2B,EAAE,EAAEuvE,GAAG,CAAC,EAAExvE,CAAC,EAAEjC,CAAC,EAAE,IAAImC,EAAED,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOU,EAAEV,EAAE,KAAK,IAAID,EAAE,MAAM,EAAE,OAAOY,EAAE,EAAE,QAAQ,SAAS23C,EAAE,uBAAuBr4C,CAAC,EAAEA,EAAEW,EAAE,EAAE,QAAQ,SAAS03C,EAAE,uBAAuB53C,CAAC,EAAEA,EAAEG,EAAEzC,GAAG,EAAE,MAAM,CAAC0C,EAAEkB,CAAC,EAAE,EAAE,EAAE,MAAMjC,EAAE,MAAMY,EAAEC,EAAEC,CAAC,EAAE,OAAOb,EAAE,eAAegC,EAAEnB,EAAEC,CAAC,CAAC,EAAE,CAAC,CAAC,OAAOzC,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,EAAE0B,CAAC,EAAE1B,EAAE2B,EAAE,EAAE,GAAG,EAAE,QAAQ,aAAaD,EAAE,QAAQ,YAAY,CAAC,IAAIE,EAAEozE,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQrzE,EAAE,MAAM,CAAC,MAAM,WAAW,CAAC,CAAC,EAAEU,EAAEV,EAAE,KAAK,IAAIC,EAAE,MAAM,EAAEU,EAAED,EAAE,mBAAmB,KAAKE,EAAEF,EAAE,mBAAmB,KAAKG,EAAEb,EAAE,KAAK,IAAIW,EAAE,MAAM,EAAE,OAAOG,EAAEd,EAAE,KAAK,IAAIY,EAAE,MAAM,EAAE,OAAOoB,EAAEqxE,GAAG,CAAC,OAAO,CAAC,EAAEtzE,CAAC,EAAE,QAAQC,EAAE,MAAM,CAAC,MAAM,WAAW,CAAC,CAAC,EAAE,EAAEA,EAAE,KAAK,IAAIgC,EAAE,MAAM,EAAEqD,EAAE,EAAE,mBAAmB,KAAK,EAAE,EAAE,mBAAmB,KAAK5C,EAAEzC,EAAE,KAAK,IAAIqF,EAAE,MAAM,EAAE,OAAO3C,EAAE1C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC2C,EAAEC,EAAE0C,CAAC,EAAEtH,EAAE,EAAE,MAAM+B,EAAE,MAAMc,EAAEC,EAAE2B,EAAEC,CAAC,EAAEI,EAAE9C,EAAE,eAAesF,EAAE,UAAU3C,CAAC,EAAE4C,EAAEvF,EAAE,eAAesF,EAAE,UAAU1C,CAAC,EAAEG,EAAE+vE,GAAG,CAAC,OAAO,CAAC,KAAKhwE,EAAE,KAAKyC,CAAC,EAAE,QAAQvF,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BC,CAAC,EAAED,EAAE,8BAA8BgC,CAAC,EAAEhC,EAAE,8BAA8B8C,CAAC,EAAE9C,EAAE,8BAA8BuF,CAAC,EAAExC,CAAC,KAAK,CAAC,IAAI9C,EAAED,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOU,EAAEV,EAAE,KAAK,IAAID,EAAE,MAAM,EAAE,OAAOY,EAAEvC,GAAG,EAAE,MAAM,CAACwC,EAAEC,CAAC,EAAE,EAAE,EAAE,MAAMd,EAAE,MAAME,EAAES,EAAEC,CAAC,EAAE,OAAOX,EAAE,eAAea,EAAEF,EAAEC,CAAC,CAAC,CAAC,CAAC,CAAC,SAAS4yE,GAAG11E,EAAE,CAAC,MAAM,CAAC,EAAEE,EAAEI,EAAEC,EAAE,EAAE,IAAI,CAAC,IAAI0B,EAAEu4C,EAAE,2BAA2B,EAAEt6C,CAAC,EAAEgC,EAAE4kB,EAAE,cAAc7kB,CAAC,EAAEE,EAAEF,EAAE,OAAOW,EAAEkkB,EAAE,eAAe7kB,CAAC,EAAEY,EAAEikB,EAAE,uBAAuB,UAAU5kB,CAAC,EAAEY,EAAEgkB,EAAE,uBAAuB,UAAU5kB,CAAC,EAAEa,EAAEy3C,EAAE,iBAAiB,EAAEv4C,CAAC,EAAEe,EAAEw3C,EAAE,iBAAiBt6C,EAAE+B,CAAC,EAAEiC,EAAEs2C,EAAE,uBAAuBl6C,EAAEC,CAAC,EAAE,EAAEi6C,EAAE,uBAAuB,EAAE,CAAC,EAAEjzC,EAAE,EAAE,OAAO,EAAEuf,EAAE,eAAe,CAAC,EAAEniB,EAAEzE,EAAE,OAAO0E,EAAEkiB,EAAE,eAAe5mB,CAAC,EAAE,GAAG6C,EAAE,OAAOC,EAAE,SAAS,EAAE,QAAQ6B,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,IAAI,CAAC,IAAIC,EAAED,EAAEX,EAAE,OAAOsD,EAAE3C,EAAE,EAAE,OAAOG,EAAEhF,EAAEkE,EAAEY,EAAE,GAAGZ,EAAEY,EAAE,EAAE,GAAG,EAAE0C,EAAE,GAAG,EAAEA,EAAE,EAAE,EAAE,EAAE3E,EAAEgC,GAAGG,EAAE,KAAKlC,EAAE+B,GAAGG,EAAE,IAAI,KAAM,SAAQH,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,IAAI,CAAC,IAAIC,EAAEgiB,EAAE,WAAWjiB,EAAE1C,EAAES,CAAC,EAAE4E,EAAE1C,EAAE,MAAM,CAACyC,CAAC,EAAExE,EAAE,QAAQoC,GAAGqC,EAAErC,GAAG,CAAC,EAAE,IAAIH,EAAE8hB,EAAE,WAAWtf,EAAED,EAAE,CAAC,EAAEE,EAAE3C,EAAE,MAAM,CAACH,CAAC,EAAE3B,EAAE,QAAQmC,GAAGsC,EAAEtC,GAAG,CAAC,EAAE,IAAIF,EAAE6hB,EAAE,WAAWrf,EAAE9C,EAAEC,CAAC,EAAEM,EAAElF,EAAEkE,EAAEc,EAAE,GAAGd,EAAEc,EAAE,EAAE,GAAG,EAAEC,EAAE,GAAG,EAAEA,EAAE,EAAE,EAAE,EAAEpC,EAAEgC,GAAGK,EAAE,KAAKpC,EAAE+B,GAAGK,EAAE,IAAI,CAAC,MAAM,CAACrC,EAAEC,EAAEb,CAAC,CAAC,CAAC,CAAC,IAAI4vE,GAAGkD,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,CAAC,EAAE21E,GAAGD,GAAG,CAAC11E,EAAE,EAAEE,EAAEI,KAAK,CAAC,KAAKN,EAAEE,EAAE,KAAK,EAAEI,CAAC,EAAE,EAAEs1E,GAAGH,GAAGp6D,GAAGw2D,GAAG8D,EAAE,EAAEE,GAAG,CAAC,WAAWx6D,GAAG,YAAY,MAAM,WAAWu6D,EAAE,EAAE,SAAS9D,GAAG9xE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEumB,EAAE,cAAcxmB,CAAC,EAAE,EAAEwmB,EAAE,oBAAoBvmB,EAAEL,CAAC,EAAE,QAAQ+B,EAAE,EAAEA,EAAEjC,EAAE,OAAOiC,IAAI,CAAC,IAAIC,EAAElC,EAAEiC,GAAG,GAAGC,EAAE,EAAE,MAAM,IAAI,MAAM,+BAA+B,EAAEA,GAAG3B,IAAI,EAAE,EAAE,EAAE2B,IAAI,EAAED,GAAG,EAAEC,IAAI,EAAE,CAAC,OAAO,CAAC,CAAC,SAAS6vE,GAAG/xE,EAAE,EAAEE,EAAEI,EAAE,GAAG,CAAC,IAAIC,EAAEP,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,GAAG,EAAEuwB,GAAG,CAAChwB,EAAEL,CAAC,EAAE,EAAE,KAAK,EAAE,QAAQ+B,EAAE,EAAEA,EAAE1B,EAAE0B,IAAI,QAAQC,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIC,EAAEnC,EAAE,IAAIiC,EAAEC,CAAC,EAAE,GAAGC,EAAE,EAAE,MAAM,IAAI,MAAM,+BAA+B,EAAEA,GAAGjC,IAAII,EAAE,EAAE,IAAI,EAAE2B,EAAEE,CAAC,EAAE,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,IAAIF,EAAEE,CAAC,EAAE,EAAE,IAAIF,EAAEC,CAAC,EAAED,EAAEE,CAAC,EAAE,EAAE,IAAI,EAAE,IAAIF,EAAEE,CAAC,EAAE,EAAEF,EAAEE,CAAC,EAAE,CAAC,OAAO,CAAC,CAAC,SAAS2zE,GAAG91E,EAAE,CAAC,MAAM,CAAC,EAAEE,EAAEI,IAAI,CAAC,IAAIC,EAAEumB,EAAE,uBAAuB5mB,EAAE,EAAE,MAAM,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,OAAO,EAAE,EAAEK,EAAE,GAAGP,EAAE,EAAE,GAAGM,CAAC,EAAE,OAAOC,CAAC,CAAC,CAAC,SAASw1E,GAAG/1E,EAAE,EAAEE,EAAE,CAAC,MAAM,CAAC,CAAC,OAAOI,EAAE,MAAMC,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,EAAED,EAAE,GAAGmxE,GAAG,EAAEzxE,CAAC,EAAE,EAAE,QAAQ,UAAUE,IAAI,SAAS,MAAM,IAAI,MAAM,sDAAsD,EAAE,IAAI+B,EAAE,EAAEC,EAAED,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOE,EAAE2kB,EAAE,cAAc,EAAE,KAAK,EAAElkB,EAAE1C,GAAG,EAAE,MAAM2C,EAAEikB,EAAE,kBAAkBlkB,EAAET,CAAC,EAAE,QAAQW,EAAE,EAAEA,EAAEX,EAAE,EAAEW,EAAED,EAAEC,GAAG,EAAEZ,EAAEY,GAAGvC,CAAC,EAAE,OAAO0B,EAAE,eAAe,EAAE,MAAMW,EAAEC,CAAC,CAAC,CAAC,CAAC,SAASmzE,GAAGh2E,EAAE,EAAEE,EAAE,CAAC,MAAM,CAAC,CAAC,OAAOI,EAAE,MAAMC,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,EAAED,EAAE,GAAGmxE,GAAG,EAAEzxE,CAAC,EAAE,EAAE,QAAQ,UAAUE,IAAI,SAAS,MAAM,IAAI,MAAM,sDAAsD,EAAE,IAAI+B,EAAE,EAAEC,EAAED,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOE,EAAEjC,GAAG,EAAE,MAAM0C,EAAE,EAAEV,EAAEC,EAAE5B,CAAC,EAAE,OAAO0B,EAAE,eAAe,EAAE,MAAME,EAAES,CAAC,CAAC,CAAC,CAAC,IAAIqvE,GAAG6D,GAAG91E,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEi2E,GAAGD,GAAGt5D,GAAGu1D,EAAE,EAAEiE,GAAG,CAAC,WAAWx5D,GAAG,YAAY,MAAM,WAAWu5D,EAAE,EAAE,SAAS/D,GAAGlyE,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEumB,EAAE,kBAAkB5mB,EAAE4mB,EAAE,cAAc,CAAC,CAAC,EAAE,GAAGxmB,GAAGJ,IAAI,SAAS,CAAC,IAAI,EAAE,EAAEF,EAAE,QAAQ,GAAG,CAAC,IAAIiC,EAAE6kB,EAAE,cAAc,EAAE,KAAK,EAAEvmB,EAAE,IAAI,EAAE,KAAK,CAAC,EAAE,GAAG0B,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,EAAE,EAAEjC,EAAE,QAAQ,GAAG,CAAC,IAAIiC,EAAE/B,IAAI,SAASs6C,EAAE,uBAAuB,EAAE,IAAI,EAAE,EAAE,KAAKt4C,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,MAAM,GAAG,EAAEA,EAAE,CAAC,IAAIS,EAAET,EAAE,EAAE,GAAG,EAAE,QAAQU,EAAE,EAAEA,EAAE,EAAE,MAAM,GAAG,EAAEA,EAAEtC,EAAEqC,EAAEC,GAAGZ,EAAEC,IAAI,CAAC,GAAG,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,OAAO3B,CAAC,CAAC,IAAI4xE,GAAG4C,GAAG,CAAC/0E,EAAE,IAAIA,IAAI,EAAE,EAAE,CAAC,EAAEm2E,GAAGV,GAAGj3D,GAAG2zD,GAAG,KAAK,MAAM,EAAEiE,GAAG,CAAC,WAAW53D,GAAG,YAAY,MAAM,WAAW23D,EAAE,EAAM/D,GAAG0D,GAAG91E,GAAG,KAAK,IAAIA,CAAC,CAAC,EAAEq2E,GAAGL,GAAGv3D,GAAG2zD,GAAG,SAAS,EAAEkE,GAAG,CAAC,WAAW73D,GAAG,YAAY,MAAM,WAAW43D,EAAE,EAAMhE,GAAGyD,GAAG91E,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAEu2E,GAAGP,GAAGr3D,GAAG0zD,EAAE,EAAEmE,GAAG,CAAC,WAAW73D,GAAG,YAAY,MAAM,WAAW43D,EAAE,EAAMjE,GAAGwD,GAAG91E,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAEy2E,GAAGT,GAAGj3D,GAAGuzD,EAAE,EAAEoE,GAAG,CAAC,WAAW33D,GAAG,YAAY,MAAM,WAAW03D,EAAE,EAAE,SAASlE,GAAGvyE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAE,CAAC,IAAIC,EAAEouB,GAAG,CAACjwB,EAAE,CAAC,EAAEJ,CAAC,EAAE,QAAQ0C,EAAE,EAAEA,EAAEtC,EAAEsC,IAAI,CAAC,IAAIC,EAAE,CAAC,EAAEC,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAExC,EAAEwC,IAAI,CAAC,IAAIC,EAAEhD,EAAE4C,EAAErC,EAAEwC,GAAGD,GAAGE,EAAE,EAAED,GAAGF,EAAE,KAAKG,CAAC,CAAC,CAAC,GAAGF,EAAE,GAAGA,GAAGZ,EAAE,EAAE,MAAM,IAAI,MAAM,oBAAoBW,yBAAyBZ,GAAG,EAAE,QAAQc,EAAE,EAAEA,EAAE,EAAEA,IAAIZ,EAAE,OAAOS,EAAE,EAAEG,GAAG,EAAE,IAAI,GAAG,EAAE,WAAWD,EAAE,EAAEC,CAAC,CAAC,CAAC,CAAC,OAAOZ,CAAC,CAAC,SAASqwE,GAAGxyE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEiwB,GAAGrwB,EAAEF,EAAE,KAAK,EAAE,QAAQO,EAAE,EAAEA,EAAED,EAAE,KAAK,EAAEC,EAAE,CAAC,IAAIyB,EAAE1B,EAAE,WAAWC,CAAC,EAAE,MAAM,EAAE0B,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAG,EAAE,EAAE,WAAW,CAACC,EAAEC,CAAC,CAAC,EAAEF,EAAE,GAAG,EAAE,OAAO,GAAG,IAAIY,EAAE5C,EAAE,WAAWgC,CAAC,EAAE,GAAGY,GAAGA,EAAE5C,EAAE,OAAO,SAASM,EAAE,OAAOC,GAAGP,EAAE,OAAO4C,GAAG,CAAC,OAAOtC,CAAC,CAAC,IAAIoyE,GAAGqC,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,EAAE,EAAE,CAAC,EAAE22E,GAAGlB,GAAGr2D,GAAGszD,GAAG,KAAK,MAAM,EAAEkE,GAAG,CAAC,WAAWx3D,GAAG,YAAY,MAAM,WAAWu3D,EAAE,EAAMlE,GAAGsC,GAAG,CAAC/0E,EAAE,IAAIA,GAAG,EAAE,EAAE,CAAC,EAAE62E,GAAGpB,GAAGp2D,GAAGozD,GAAG,KAAK,MAAM,EAAEqE,GAAG,CAAC,WAAWz3D,GAAG,YAAY,MAAM,WAAWw3D,EAAE,EAAMjE,GAAGmC,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,EAAE,EAAE,CAAC,EAAE+2E,GAAGtB,GAAG51D,GAAG+yD,GAAG,KAAK,MAAM,EAAEoE,GAAG,CAAC,WAAWn3D,GAAG,YAAY,MAAM,WAAWk3D,EAAE,EAAMpE,GAAGoC,GAAG,CAAC/0E,EAAE,IAAIA,GAAG,EAAE,EAAE,CAAC,EAAEi3E,GAAGxB,GAAG31D,GAAG6yD,GAAG,KAAK,MAAM,EAAEuE,GAAG,CAAC,WAAWp3D,GAAG,YAAY,MAAM,WAAWm3D,EAAE,EAAE,SAASpE,GAAG7yE,EAAE,EAAEE,EAAE,CAAC,IAAII,GAAG,EAAEN,IAAIE,EAAE,GAAGK,EAAEumB,EAAE,oBAAoB5mB,EAAE,SAAS,EAAEK,EAAE,GAAGP,EAAE,QAAQ,EAAE,EAAE,EAAEO,EAAE,OAAO,IAAIA,EAAE,GAAGA,EAAE,EAAE,GAAGD,EAAE,OAAOC,CAAC,CAAC,IAAIuyE,GAAGgD,GAAG91E,GAAG,KAAK,IAAIA,CAAC,CAAC,EAAEm3E,GAAGnB,GAAGh2D,GAAG8yD,EAAE,EAAEsE,GAAG,CAAC,WAAWp3D,GAAG,YAAY,MAAM,WAAWm3D,EAAE,EAAE,SAASpE,GAAG/yE,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEumB,EAAE,uBAAuBxmB,EAAEwmB,EAAE,cAAc5mB,CAAC,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEK,EAAE,OAAO,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE,EAAE0B,EAAEjC,EAAE,GAAG,QAAQkC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAEnC,EAAE,EAAEkC,IAAI,OAAO,MAAMC,CAAC,GAAGA,EAAEF,KAAKA,EAAEE,EAAE,CAAC5B,EAAE,GAAG0B,CAAC,CAAC,OAAO1B,CAAC,CAAC,IAAIyyE,GAAG+B,GAAG,CAAC/0E,EAAE,IAAI,KAAK,IAAIA,EAAE,CAAC,CAAC,EAAEq3E,GAAG5B,GAAG90D,GAAGqyD,EAAE,EAAEsE,GAAG,CAAC,WAAW32D,GAAG,YAAY,MAAM,WAAW02D,EAAE,EAAMpE,GAAG8B,GAAG,CAAC/0E,EAAE,IAAI,KAAK,IAAIA,EAAE,CAAC,CAAC,EAAEu3E,GAAG9B,GAAGt0D,GAAG8xD,EAAE,EAAEuE,GAAG,CAAC,WAAWr2D,GAAG,YAAY,MAAM,WAAWo2D,EAAE,EAAMrE,GAAG6B,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,CAAC,EAAEy3E,GAAG/B,GAAG,CAAC11E,EAAE,EAAEE,EAAEI,KAAK,CAAC,KAAKN,EAAEE,EAAE,EAAEI,EAAE,KAAKN,EAAEM,EAAE,EAAEJ,CAAC,EAAE,EAAEw3E,GAAGjC,GAAGl0D,GAAG2xD,GAAGuE,EAAE,EAAEE,GAAG,CAAC,WAAWp2D,GAAG,YAAY,MAAM,WAAWm2D,EAAE,EAAE,SAASvE,GAAGnzE,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEwmB,EAAE,kBAAkB,GAAG5mB,CAAC,EAAE,OAAOgzE,GAAG,CAAC,EAAE,EAAE5yE,EAAEN,EAAEE,CAAC,CAAC,CAAC,SAAS03E,GAAG53E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAEmxE,GAAGnxE,EAAE,KAAK,EAAE,IAAIC,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,CAAC,EAAE6yE,GAAG5yE,EAAED,EAAE,MAAMA,EAAE,KAAK,EAAE,OAAOJ,EAAE,eAAe,EAAEI,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIu3E,GAAG,CAAC,WAAWr2D,GAAG,YAAY,MAAM,WAAWo2D,EAAE,EAAMxE,GAAG2B,GAAG,CAAC/0E,EAAE,IAAIA,IAAI,EAAE,EAAE,CAAC,EAAE83E,GAAGrC,GAAGh0D,GAAG2xD,GAAG,KAAK,MAAM,EAAE2E,GAAG,CAAC,WAAWt2D,GAAG,YAAY,MAAM,WAAWq2D,EAAE,EAAE,SAASnD,GAAG30E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,EAAE,OAAO,EAAEumB,EAAE,cAAc,CAAC,EAAE7kB,EAAE6kB,EAAE,eAAe,CAAC,EAAE5kB,EAAE4kB,EAAE,eAAevmB,CAAC,EAAE4B,EAAE2kB,EAAE,uBAAuB5mB,EAAE4mB,EAAE,cAAcvmB,CAAC,CAAC,EAAE,QAAQqC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAEikB,EAAE,WAAWlkB,EAAE,EAAEX,CAAC,EAAEa,EAAE,IAAI,MAAMD,EAAE,MAAM,EAAE,QAAQG,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAIF,EAAEE,GAAGH,EAAEvC,EAAE0C,IAAI,IAAID,EAAE+jB,EAAE,WAAWhkB,EAAE,EAAEZ,CAAC,EAAEC,EAAEY,GAAG/C,EAAE4C,EAAE,CAAC,OAAOT,CAAC,CAAC,SAAS61E,GAAGh4E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAEL,EAAEuxE,GAAGlxE,EAAE,WAAW,EAAE,IAAI,EAAEA,EAAE,MAAM,OAAO0B,EAAE,IAAI,MAAM,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,IAAIA,EAAE,GAAG1B,EAAE,MAAM,EAAE,IAAI,IAAI2B,EAAE5B,EAAE,KAAK,IAAIC,EAAE,MAAM,EAAE,OAAO4B,EAAEwyE,GAAGzyE,EAAE3B,EAAE,MAAMA,EAAE,MAAM,EAAE0B,CAAC,EAAE,MAAM,CAAC,OAAO3B,EAAE,MAAM6B,EAAEF,EAAE1B,EAAE,KAAK,EAAE,MAAM0B,EAAE,MAAM1B,EAAE,KAAK,CAAC,CAAC,IAAI03E,GAAG,CAAC,WAAW5yD,GAAG,YAAY,MAAM,WAAW2yD,EAAE,EAAE,SAAS3E,GAAGrzE,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAEi6C,EAAE,0BAA0Bx6C,EAAEM,CAAC,EAAE,EAAEqqB,GAAG,EAAE,OAAO,EAAE1oB,EAAE6kB,EAAE,oBAAoBA,EAAE,cAAcvmB,CAAC,EAAE,CAAC,EAAE2B,EAAE4kB,EAAE,cAAc,CAAC,EAAE,QAAQ3kB,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,CAAC,IAAIS,EAAET,EAAED,EAAEW,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAEZ,EAAE,EAAEY,EAAED,GAAG3C,EAAE0C,EAAEE,GAAGb,EAAEE,GAAGU,CAAC,CAAC,MAAM,CAAC,QAAQZ,EAAE,SAAS1B,EAAE,SAAS,CAAC,CAAC,CAAC,SAAS23E,GAAGl4E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,MAAM,EAAE,IAAI0B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE4B,EAAEq4C,EAAE,mBAAmBt4C,EAAED,CAAC,EAAEW,EAAEV,EAAEW,EAAEtC,EAAEuC,EAAE,CAAC,EAAEX,GAAG,OAAOU,EAAEm1E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAEW,EAAE,KAAKD,CAAC,EAAED,EAAE43C,EAAE,iBAAiB53C,EAAE,OAAOX,CAAC,GAAG,IAAIc,EAAE7C,EAAE,KAAK,IAAI2C,EAAE,MAAM,EAAE,OAAO,CAAC,QAAQG,EAAE,SAASkB,EAAE,SAAS,CAAC,EAAEmvE,GAAGxwE,EAAE,MAAMA,EAAE,MAAME,EAAEH,CAAC,EAAE2E,EAAErD,EAAE,OAAO,IAAIqD,EAAEizC,EAAE,qBAAqBt2C,EAAEhC,CAAC,GAAGY,EAAE,QAAQ,GAAG5C,EAAE,8BAA8B,CAAC,CAAC,EAAEA,EAAE,eAAeqH,EAAE,EAAEvE,CAAC,CAAC,CAAC,IAAIm1E,GAAG,CAAC,WAAW/1D,GAAG,YAAY,MAAM,WAAW81D,EAAE,EAAE,SAASE,GAAGp4E,EAAE,EAAEE,EAAE,CAACF,EAAE,QAAQ,CAACM,EAAEC,IAAI,CAAC,GAAGD,EAAE,GAAGA,GAAGJ,EAAE,CAAC,IAAI,EAAE4mB,EAAE,WAAWvmB,EAAE,EAAE,OAAOumB,EAAE,eAAe,CAAC,CAAC,EAAE,KAAK,GAAG,EAAE,MAAM,IAAI,MAAM,WAAW,QAAQxmB,mBAAmBJ,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,SAASm4E,GAAGr4E,EAAE,EAAE,CAAC,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAEE,GAAGK,EAAEL,IAAIF,EAAE,OAAO,EAAE,EAAEA,EAAEE,EAAE,GAAG,OAAO,GAAGI,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,gCAAgC,EAAE,GAAGA,EAAE,GAAG,EAAE,MAAM,IAAI,MAAM,oCAAoC,EAAE,GAAGA,EAAEA,EAAE,OAAO,GAAGC,EAAE,MAAM,IAAI,MAAM,0CAA0C,EAAE,QAAQ,EAAE,EAAE,EAAED,EAAE,OAAO,EAAE,EAAE,GAAGA,EAAE,EAAE,GAAGA,EAAE,GAAG,MAAM,IAAI,MAAM,iDAAiD,CAAC,CAAC,CAAC,SAASg4E,GAAGt4E,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,OAAO,EAAEL,EAAE,OAAO+B,EAAE,IAAI,MAAM,CAAC,EAAE,KAAK,IAAI,EAAE,IAAI,IAAI,CAAC,CAAC,CAAC,EAAEo2E,GAAGn4E,EAAEI,CAAC,EAAE,IAAI4B,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAE,EAAEA,EAAE,CAACD,GAAG,EAAEC,GAAG,IAAIS,EAAE,EAAET,EAAE,GAAG,QAAQU,EAAE,EAAEA,EAAEX,EAAE,EAAE,EAAEW,EAAEZ,EAAEE,GAAG,KAAKU,EAAED,CAAC,CAAC,CAAC,QAAQT,EAAE,EAAEA,EAAEnC,EAAE,OAAO,EAAEmC,EAAE,CAAC,IAAIS,EAAE5C,EAAEmC,GAAGU,EAAE7C,EAAEmC,GAAG,EAAE,QAAQW,EAAE,EAAEA,EAAE5C,EAAE,OAAO,EAAE4C,EAAE,CAAC,IAAIC,EAAE7C,EAAE4C,GAAGE,EAAEF,EAAE,EAAE,OAAO,EAAE,GAAGE,GAAG,EAAE,CAAC,IAAIkB,EAAEjC,EAAEe,GAAG,EAAEkB,EAAEA,EAAE,OAAO,GAAGnB,EAAEH,GAAG,QAAQ2E,EAAE3E,EAAE2E,EAAE1E,EAAE,EAAE0E,EAAEtF,EAAEe,GAAG,KAAKD,EAAEwE,EAAE,GAAG,CAAC,CAAC,CAAC3E,EAAEG,EAAEH,GAAGC,EAAEE,EAAEF,EAAE,CAACA,IAAID,IAAIrC,EAAE,KAAK,CAACqC,EAAEC,CAAC,CAAC,EAAE,GAAGA,EAAED,EAAE,CAAC,MAAM,CAAC,UAAUX,EAAE,YAAY1B,EAAE,UAAU,CAAC,CAAC,CAAC,SAASg4E,GAAGv4E,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEF,EAAE,OAAO,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAEE,GAAG,OAAOK,EAAEumB,EAAE,kBAAkB,QAAQxmB,CAAC,EAAE,EAAE,KAAKC,CAAC,EAAEP,EAAEE,GAAG,QAAQ,CAAC,EAAE,IAAIK,EAAE,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,SAASi4E,GAAGx4E,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,MAAM,EAAE,CAAC,EAAE,KAAKE,EAAE,OAAO,GAAGA,EAAE,KAAK,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEN,EAAE,OAAOM,IAAIJ,EAAE,EAAE,IAAIF,EAAEM,GAAG,OAAOJ,CAAC,CAAC,SAASu4E,GAAGz4E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEi4E,GAAG,EAAE,CAAC,EAAE,GAAGv2E,EAAEu2E,GAAG,EAAE,CAAC,EAAE,GAAGt2E,EAAE,EAAE,QAAQC,KAAKjC,EAAE,QAAQ0C,EAAET,EAAE,GAAGS,EAAET,EAAE,GAAG,EAAES,EAAE,CAAC,QAAQC,EAAE,EAAEA,EAAEvC,EAAE,EAAEuC,EAAEtC,EAAE2B,EAAED,EAAEY,GAAG7C,EAAE4C,EAAE,EAAEC,GAAG,EAAEX,CAAC,CAAC,CAAC,SAASw2E,GAAG14E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,EAAE,GAAGA,EAAE,IAAI,EAAEumB,EAAE,kBAAkB5mB,EAAE4mB,EAAE,cAAc,CAAC,CAAC,EAAE7kB,EAAEjC,EAAE,OAAOkC,EAAED,IAAI,EAAE,EAAEA,EAAE,EAAE,GAAG,OAAOw2E,GAAGz4E,EAAE,EAAEM,EAAE4B,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,SAASoxE,GAAGtzE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAE,CAAC,GAAGjC,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,GAAG,EAAE,GAAG,SAAS,EAAE,MAAM,IAAI,MAAM,mCAAmC,EAAE,IAAIkC,EAAE,EAAE,GAAG,GAAG,EAAE,GAAGk2E,GAAG,EAAE,EAAEl2E,CAAC,EAAE5B,EAAE,SAAS,EAAE,MAAM,IAAI,MAAM,6BAA6B,EAAE,IAAI6B,EAAE7B,EAAE,GAAG,CAAC,UAAUsC,EAAE,YAAYC,EAAE,UAAUC,CAAC,EAAEw1E,GAAG,EAAE,EAAEt4E,EAAEmC,CAAC,EAAEY,EAAEw1E,GAAG31E,CAAC,EAAEI,EAAE01E,GAAGx4E,EAAEI,EAAEC,EAAEsC,EAAEC,CAAC,EAAE,MAAM,CAACC,EAAEC,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,IAAI21E,GAAG,WAAW,SAASpF,GAAGvzE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,mCAAmC,EAAE,GAAGA,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,mCAAmC,EAAE,GAAG,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,mCAAmC,EAAE,IAAI0B,EAAE,EAAE,SAAS,EAAEC,EAAE3B,EAAE,SAAS,EAAE4B,EAAE,EAAE,SAAS,EAAES,EAAE,CAAC,EAAEX,GAAGW,EAAE,KAAK,EAAE,EAAE,EAAEV,GAAGU,EAAE,KAAKrC,EAAE,EAAE,EAAE4B,GAAGS,EAAE,KAAK,EAAE,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAE,GAAGA,EAAE,KAAKA,EAAE,EAAE,GAAG,MAAM,IAAI,MAAM,qDAAqD,EAAE,IAAIC,EAAED,EAAE,SAAS,EAAE,EAAEA,EAAE,GAAGE,EAAEgkB,EAAE,kBAAkB,QAAQjkB,EAAE,CAAC,EAAEC,EAAE,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAED,EAAE,EAAE,EAAE,CAAC,IAAI0E,EAAEtF,EAAEjC,EAAE,GAAGA,EAAE,GAAG,EAAEkC,EAAE5B,EAAE,GAAGA,EAAE,GAAGqE,EAAExC,EAAE,EAAE,GAAG,EAAE,GAAG,GAAGwC,IAAI,EAAE,MAAM,IAAI,MAAM,qBAAqB,EAAE,IAAIC,EAAE,GAAGD,EAAE,GAAG,EAAE4C,GAAG5C,EAAE,GAAG,EAAE4C,EAAE3C,EAAE,UAAUA,EAAE,KAAK,KAAK,KAAK,KAAK,EAAE2C,GAAG5C,CAAC,CAAC,EAAEC,EAAE+zE,GAAG,MAAM,IAAI,MAAM,yCAAyCA,IAAI,EAAE71E,EAAE,EAAE,GAAGA,EAAE,GAAG8B,CAAC,CAAC,IAAI7B,EAAED,EAAED,GAAGG,EAAE8jB,EAAE,kBAAkB5mB,EAAE6C,CAAC,EAAEmB,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAErB,EAAE,EAAE,EAAE,CAAC,IAAI0E,EAAEzE,EAAE,EAAE,GAAGA,EAAE,GAAG,EAAEb,EAAEjC,EAAE,GAAGA,EAAE,GAAG2E,EAAExC,EAAE,EAAE,GAAG,EAAE,GAAG,QAAQyC,EAAE,EAAEA,EAAE2C,EAAE,EAAE3C,EAAE5B,EAAEkB,KAAK,EAAE,GAAGS,CAAC,CAAC,MAAM,CAAC7B,EAAEE,CAAC,CAAC,CAAC,IAAI41E,GAAGp+B,EAAE,iBAAiBq+B,GAAG,KAAK,CAAC,YAAY/3E,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAEC,EAAE,EAAEU,EAAE,CAAC,KAAK,MAAM9B,EAAE,KAAK,WAAWZ,EAAE,KAAK,OAAO,EAAE,KAAK,YAAYK,EAAE,KAAK,YAAYwB,EAAE,KAAK,aAAaC,EAAE,KAAK,kBAAkBC,EAAE,KAAK,mBAAmBC,EAAE,KAAK,yBAAyB,EAAE,KAAK,kBAAkBs4C,EAAE,2BAA2B53C,CAAC,EAAE,KAAK,WAAW43C,EAAE,cAAc,KAAK,iBAAiB,CAAC,CAAC,+BAA+B15C,EAAE,CAAC,OAAO,KAAK,kBAAkB,KAAK83E,GAAG,eAAe,KAAK,kBAAkB93E,EAAE,GAAG,KAAK,kBAAkBA,EAAE,CAAC,sBAAsBA,EAAE,CAAC,OAAO,KAAK,kBAAkB,KAAK83E,GAAG,eAAe,KAAK,mBAAmB93E,EAAE,GAAG,KAAK,mBAAmBA,EAAE,CAAC,YAAYA,EAAE,CAAC,IAAIZ,EAAE,KAAK,sBAAsBY,EAAE,CAAC,EAAE,OAAO,KAAK,+BAA+BA,EAAE,CAAC,EAAE,CAAC,KAAK83E,GAAG,aAAa,OAAOC,GAAG,sBAAsB34E,CAAC,EAAE,KAAK04E,GAAG,WAAW,OAAOC,GAAG,oBAAoB34E,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,gCAAgC04E,GAAG,KAAK,+BAA+B93E,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC,OAAO,oBAAoBA,EAAE,CAAC,IAAIZ,EAAEY,EAAE,OAAO,GAAGZ,IAAI,GAAGA,IAAI,EAAE,MAAO,GAAE,IAAI,EAAE,EAAE,QAAQK,EAAE,EAAEA,EAAEL,EAAE,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAEjB,EAAEP,EAAE,GAAGO,EAAEP,GAAGwB,EAAE,IAAI,EAAEA,EAAE,CAAC,OAAO,CAAC,CAAC,OAAO,sBAAsBjB,EAAE,CAAC,IAAIZ,EAAEY,EAAE,OAAO,GAAGZ,IAAI,EAAE,MAAO,GAAE,IAAI,EAAE,EAAEK,EAAEO,EAAE,GAAGiB,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE9B,EAAE,EAAE8B,EAAE,CAAC,IAAIC,EAAEnB,EAAEkB,GAAGC,IAAI1B,IAAIA,EAAE0B,EAAEF,EAAE,KAAK,IAAIC,EAAE,EAAED,CAAC,EAAE,EAAEC,EAAE,CAAC,OAAO,KAAK,IAAI9B,EAAE,EAAE6B,CAAC,CAAC,CAAC,sBAAsBjB,EAAEZ,EAAE,EAAE,GAAG,CAAC,GAAGA,EAAE,SAAS,EAAE,CAAC,GAAGY,EAAE,KAAK,GAAG,MAAM,CAAC,EAAE,MAAM,IAAI,MAAM,gFAAgF,CAAC,CAAC,OAAOg4E,GAAGh4E,EAAE,CAAC,CAAC,CAAC,oBAAoBA,EAAE,CAAC,IAAIZ,EAAE,KAAK,YAAY,EAAE,KAAK,kBAAkBs6C,EAAE,0BAA0B,EAAEt6C,CAAC,EAAE,IAAIK,EAAE,KAAK,sBAAsB,KAAK,MAAM,KAAK,UAAU,EAAEyB,EAAEw4C,EAAE,kCAAkC,KAAK,WAAWj6C,EAAEL,CAAC,EAAE8B,EAAE,GAAG,IAAIA,EAAE,GAAGlB,GAAG,QAAQmB,EAAE,EAAEA,GAAG,KAAK,WAAW,EAAEA,EAAED,EAAEC,GAAG,IAAID,EAAEC,GAAG,KAAK,YAAYA,CAAC,GAAG,OAAOD,CAAC,CAAC,gCAAgClB,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,KAAK,IAAIO,EAAE,CAAC,EAAEiB,EAAE,CAAC,EAAEC,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE1B,EAAE,EAAE0B,EAAED,GAAG9B,EAAE6B,EAAE,KAAKC,CAAC,EAAE,QAAQC,EAAE1B,EAAE0B,EAAEnB,EAAE,EAAEmB,EAAEF,EAAE,KAAK,EAAE,EAAE,OAAO+kB,EAAE,OAAO/kB,EAAE,SAASjB,EAAE,IAAI,yDAAyD,EAAEiB,CAAC,CAAC,6BAA6BjB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAEjB,EAAE,OAAOkB,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEF,EAAE,EAAE,EAAEE,EAAE,CAAC,IAAIC,EAAEpB,EAAEmB,EAAE,GAAGnB,EAAEmB,GAAG,EAAE,KAAK,IAAI1B,EAAE2B,CAAC,EAAEU,EAAE1C,EAAE+B,GAAGW,IAAI,KAAK,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAE,EAAE,EAAE,EAAEZ,EAAE,KAAKY,CAAC,EAAEA,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAEV,EAAE,EAAE,EAAE,EAAEF,EAAE,KAAK,EAAE,CAAC,CAAC,GAAGD,EAAE,GAAGC,EAAE,SAASlB,EAAEiB,EAAE,GAAG,MAAM,IAAI,MAAM,yBAAyB,EAAE,OAAOC,CAAC,CAAC,+BAA+BlB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAEjB,EAAE,OAAOkB,EAAE,CAAC,EAAE,GAAGD,IAAI,EAAE,MAAM,CAAC,EAAE,IAAIE,EAAE,EAAEC,EAAEpB,EAAE,GAAG,GAAGoB,GAAGhC,EAAE,OAAO,MAAM,IAAI,MAAM,yBAAyBgC,6BAA6BhC,EAAE,QAAQ,EAAE,IAAI,EAAEA,EAAEgC,GAAGF,EAAE,KAAK,CAAC,EAAE,QAAQY,EAAE,EAAEA,EAAEb,EAAE,EAAEa,EAAE,CAAC,IAAI,EAAE9B,EAAE8B,GAAG,GAAG,IAAIV,EAAE,GAAG,IAAI,EAAED,EAAEA,EAAE1B,EAAE,GAAG,EAAE,EAAE,QAAQ,CAAC,GAAG0B,EAAE,EAAEC,EAAE,EAAE,GAAGhC,EAAE,OAAO,MAAM,IAAI,MAAM,sBAAsB,4BAA4BA,EAAE,QAAQ,EAAE,EAAEA,EAAE,EAAE,CAAC8B,EAAE,KAAK,CAAC,CAAC,CAAC,GAAGA,EAAE,SAASlB,EAAE,OAAO,MAAM,IAAI,MAAM,kBAAkB,EAAE,OAAOkB,CAAC,CAAC,qBAAqBlB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,IAAIwB,EAAE,KAAK,sBAAsBjB,CAAC,EAAEkB,EAAE,KAAK,+BAA+BlB,CAAC,EAAE,OAAOkB,EAAE,CAAC,KAAK42E,GAAG,aAAa,OAAO,KAAK,+BAA+B72E,EAAE7B,EAAE,EAAEK,CAAC,EAAE,KAAKq4E,GAAG,WAAW,GAAG72E,EAAE,OAAO,EAAE7B,EAAE,OAAO,MAAM,IAAI,MAAM,mDAAmD6B,EAAE,OAAO,OAAO7B,EAAE,QAAQ,EAAE,OAAO,KAAK,6BAA6B6B,EAAE7B,EAAE,EAAEK,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,+BAA+Bq4E,GAAG52E,IAAI,CAAC,CAAC,CAAC,uBAAuB,CAAC,IAAIlB,EAAE,KAAK,mBAAmB,GAAG,GAAG,KAAK,kBAAkB,SAAS,EAAE,MAAM,IAAI,MAAM,+BAA+B,EAAE,IAAIZ,EAAE,KAAK,kBAAkB,GAAG,OAAOA,EAAE,CAAC,KAAK04E,GAAG,eAAe,OAAO93E,EAAE,GAAG,KAAK83E,GAAG,aAAa,MAAM,IAAI,MAAM,gDAAgD,EAAE,KAAKA,GAAG,WAAW,OAAO,KAAK,yBAAyB,GAAG,GAAG,EAAE,QAAQ,MAAM,IAAI,MAAM,sBAAsBA,GAAG14E,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,GAAG,KAAK,mBAAmB,GAAG,QAAQ,EAAE,MAAM,IAAI,MAAM,sEAAsE,EAAE,IAAI,EAAE,KAAK,sBAAsB,EAAEI,EAAE,KAAK,oBAAoB,CAAC,EAAEC,EAAE,IAAI,MAAM,KAAK,WAAW,CAAC,EAAEA,EAAEA,EAAE,OAAO,GAAG,EAAE,QAAQ2B,EAAE3B,EAAE,OAAO,EAAE2B,GAAG,EAAE,EAAEA,EAAE3B,EAAE2B,GAAG3B,EAAE2B,EAAE,GAAG5B,EAAE4B,EAAE,GAAG,IAAIH,EAAE+2E,GAAGx4E,EAAE,EAAE,EAAE0B,EAAE8kB,EAAE,kBAAkB,KAAK,YAAYA,EAAE,cAAc/kB,CAAC,CAAC,EAAE,GAAGxB,EAAE,GAAGD,EAAE,GAAG,EAAE,CAAC,IAAI4B,EAAE,KAAK,gCAAgC,EAAE3B,EAAE,GAAGD,EAAE,EAAE,EAAE,QAAQ6B,EAAE,EAAEA,GAAG,KAAK,WAAW,EAAEA,EAAED,EAAE,KAAK,qBAAqBC,EAAE,EAAED,EAAE3B,EAAE4B,GAAG7B,EAAE6B,EAAE,EAAE,KAAK,UAAU,KAAK,WAAWD,EAAEF,EAAED,CAAC,CAAC,CAAC,MAAM,CAACA,EAAEC,CAAC,CAAC,CAAC,UAAUlB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,GAAG,EAAE,SAAS,EAAE,OAAO,IAAIwB,EAAE,KAAK,OAAOC,EAAE,EAAEC,EAAE1B,EAAE,MAAM,EAAE0B,EAAEA,EAAE,MAAMnB,EAAE,CAAC,EAAE,IAAIoB,EAAE4kB,EAAE,cAAc7kB,CAAC,EAAE,EAAE/B,EAAE,OAAO0C,EAAE,KAAK,aAAa,GAAGA,EAAE,SAASV,GAAGU,EAAE,SAAS,EAAE,CAAC,IAAII,EAAE,KAAK,kBAAkBywB,EAAE,IAAI,CAAC,IAAIvvB,EAAEg4B,EAAEt5B,EAAEI,CAAC,EAAEJ,EAAEq7B,GAAG/5B,EAAEjC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAEa,EAAE,EAAEC,EAAE,EAAE,QAAQC,EAAE,EAAEA,GAAG,EAAE,EAAEA,EAAE,CAAC,IAAIkB,EAAElB,EAAE,EAAE9C,EAAE8C,GAAG,GAAG,GAAGkB,IAAInB,EAAE,CAAC,EAAEA,EAAE,QAAQ,CAAC,GAAGD,EAAEC,EAAE,CAAC,IAAIoB,EAAEpC,EAAE,SAAS,EAAEG,CAAC,EAAEqF,EAAEvF,EAAE,SAASc,EAAEZ,CAAC,EAAE6C,GAAGhC,EAAED,GAAGZ,EAAE62E,GAAGxxE,EAAEpD,EAAEY,CAAC,CAAC,CAAC,GAAG/B,GAAG,EAAE,CAAC,IAAImB,EAAE,EAAE,OAAOD,EAAE,KAAK,MAAMC,EAAEjC,CAAC,CAAC,CAAC,GAAGgC,EAAEnB,EAAE,GAAG,KAAK,aAAa,SAAS,EAAEf,EAAE,SAASe,EAAEb,EAAEgC,EAAEhC,CAAC,EAAE,KAAK,KAAK,aAAa,EAAE,EAAEa,EAAEmB,MAAO,MAAKA,EAAEnB,GAAG,CAAC,IAAIoB,EAAEnC,EAAE,MAAMe,EAAEb,CAAC,EAAE62E,GAAG50E,EAAEvB,EAAEV,CAAC,EAAE,EAAEa,CAAC,CAACmB,EAAE,GAAG,EAAElB,EAAE,EAAEF,EAAEC,IAAI,EAAEC,EAAEF,EAAEC,EAAEA,EAAED,EAAE,EAAE,CAAC,CAAC,EAAE,SAASi2E,GAAG/4E,EAAE,EAAEE,EAAE,CAAC,QAAQI,EAAE,EAAEA,EAAEJ,EAAEI,IAAIN,EAAEM,GAAG,EAAEA,EAAE,CAAC,SAASw4E,GAAG94E,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAE,QAAQI,KAAKN,EAAE,CAAC,GAAGM,EAAE,EAAE,CAAC,GAAG,CAAC,EAAE,MAAM,IAAI,MAAM,aAAaA,gBAAgB,EAAE,GAAGA,EAAE,GAAG,MAAM,IAAI,MAAM,aAAaA,iBAAiB,EAAEA,EAAE,EAAE,CAACJ,EAAE,KAAKI,CAAC,CAAC,CAAC,OAAOJ,CAAC,CAAC,SAASszE,GAAGxzE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAE,CAAC,OAAO,IAAI02E,GAAG74E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,EAAE,QAAQ,CAAC,CAAC,SAASsxE,GAAGzzE,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEP,IAAI,EAAE,EAAEA,EAAE,GAAGE,EAAE,EAAE,EAAE,EAAEF,GAAGE,EAAE,EAAE,GAAGK,GAAG,GAAG,EAAE,OAAOumB,EAAE,oBAAoB,EAAExmB,CAAC,EAAE,IAAI2B,EAAE,KAAK,IAAI,KAAK,MAAM,EAAEjC,GAAGE,CAAC,CAAC,EAAEgC,EAAE4kB,EAAE,oBAAoB7kB,EAAE3B,CAAC,EAAE,EAAEN,GAAGE,IAAI,IAAIA,EAAE,IAAIgC,EAAE,GAAGlC,EAAE,QAAQmC,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAID,EAAEC,GAAGD,EAAEC,EAAE,GAAGjC,EAAE,OAAOgC,CAAC,CAAC,IAAIwxE,GAAGoC,GAAG91E,GAAG,EAAE,KAAK,KAAKA,CAAC,CAAC,EAAEg5E,GAAGhD,GAAG5yD,GAAGswD,EAAE,EAAEuF,GAAG,CAAC,WAAW71D,GAAG,YAAY,MAAM,WAAW41D,EAAE,EAAE,SAASrF,GAAG3zE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAE,CAAC,IAAIS,EAAE,CAACtC,EAAEC,EAAEA,CAAC,EAAEsC,EAAE7C,EAAE,OAAO8C,EAAE,EAAE,OAAO,GAAGxC,IAAI,EAAE,OAAOiwB,GAAGrwB,EAAE,EAAE,KAAK,EAAE,IAAI6C,EAAEwtB,GAAG3tB,EAAE,EAAE,KAAK,EAAE,OAAOV,GAAG,UAAU,OAAOA,GAAG,SAASa,EAAE,OAAO,KAAKb,CAAC,EAAE,OAAOA,GAAG,WAAWa,EAAE,OAAO,KAAK,CAACb,CAAC,EAAE,QAAQc,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIkB,EAAE,CAAC,EAAE,EAAE,EAAE,QAAQqD,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAI,EAAE1E,EAAEG,EAAE,EAAEuE,GAAGrD,EAAE,KAAK,CAAC,EAAE,GAAG,EAAEjC,EAAEsF,EAAE,CAAC,GAAG,EAAE,GAAG,GAAGjH,EAAEC,EAAE,MAAM,IAAI,MAAM,oBAAoB2D,yBAAyBhE,GAAG,EAAE,QAAQqH,EAAE,EAAEA,EAAEhH,EAAEgH,IAAIpF,EAAEY,EAAE,OAAO,EAAExC,EAAEgH,IAAIzE,EAAEE,EAAEzC,EAAEgH,GAAGxE,EAAE,OAAO,EAAExC,EAAEgH,GAAG,EAAE,OAAO,EAAEzE,EAAE,GAAGA,EAAEE,EAAEzC,EAAEgH,EAAE,CAAC,OAAOxE,CAAC,CAAC,IAAI6wE,GAAGkC,GAAG91E,GAAG,GAAG,EAAE,KAAK,IAAI,CAACA,CAAC,EAAE,EAAEk5E,GAAGnD,GAAGlyD,GAAG7jB,GAAG,GAAG,EAAE,KAAK,IAAI,CAACA,CAAC,EAAE,EAAEm5E,GAAG,CAAC,WAAWt1D,GAAG,YAAY,MAAM,WAAWq1D,EAAE,EAAE,SAASpF,GAAG9zE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE61B,GAAG,iBAAiB91B,EAAE,EAAEJ,CAAC,EAAE,EAAE4mB,EAAE,cAAc5mB,CAAC,EAAE+B,EAAE6kB,EAAE,eAAexmB,CAAC,EAAE,GAAG,EAAE,CAAC,IAAIuC,EAAEuzB,GAAG,kBAAkB,EAAEn0B,CAAC,EAAE,OAAO1B,IAAI,SAASP,EAAE,MAAM6C,EAAEA,EAAE,CAAC,EAAE7C,EAAE,SAAS6C,EAAEA,EAAE,CAAC,CAAC,CAAC,IAAIX,EAAE3B,IAAI,SAASi6C,EAAE,uBAAuBx6C,CAAC,EAAEA,EAAEmC,EAAEouB,GAAGjwB,EAAEC,EAAE2B,CAAC,EAAEU,EAAE2tB,GAAGrwB,EAAEK,CAAC,EAAE,QAAQsC,EAAE,EAAEA,EAAED,EAAE,KAAK,EAAEC,EAAE,CAAC,IAAIC,EAAEF,EAAE,WAAWC,CAAC,EAAEE,EAAED,EAAE,IAAI,CAACE,EAAEkB,IAAIlB,EAAE,EAAEkB,EAAE,EAAEtB,EAAE,IAAIT,EAAE,IAAI,GAAGY,CAAC,EAAE,GAAGD,CAAC,CAAC,CAAC,OAAOvC,IAAI,SAASi6C,EAAE,uBAAuB53C,EAAE,MAAM,EAAEA,EAAE,MAAM,CAAC,SAASw2E,GAAGp5E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,OAAO,EAAE,GAAG,CAAC0B,EAAEC,CAAC,EAAEk0B,GAAG,iBAAiB71B,EAAE,EAAE,CAAC,EAAE61B,GAAG,kBAAkB71B,EAAE0B,EAAEC,CAAC,EAAE,IAAIC,EAAEjC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOqC,EAAEkxE,GAAG3xE,EAAEF,EAAEC,EAAE3B,EAAE,MAAMA,EAAE,KAAK,EAAE,OAAOL,EAAE,eAAegC,EAAE3B,EAAE,MAAMqC,CAAC,CAAC,CAAC,IAAIy2E,GAAG,CAAC,WAAW51D,GAAG,YAAY,MAAM,WAAW21D,EAAE,EAAE,SAASrF,GAAG/zE,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAE,EAAE,GAAGC,EAAE,EAAE,GAAGC,EAAE,IAAI,MAAMD,CAAC,EAAEU,EAAE,IAAI,MAAMX,CAAC,EAAEY,EAAE,EAAE,GAAG,GAAGX,IAAI,EAAE,CAAC,GAAGD,IAAI,EAAE,MAAM,IAAI,MAAMu4C,EAAE,gDAAgDv4C,CAAC,CAAC,EAAE,IAAI,EAAE6kB,EAAE,kBAAkB5mB,EAAE,CAAC,EAAEqH,EAAEuf,EAAE,kBAAkBvmB,EAAE,CAAC,EAAE,MAAM,CAAC,EAAE,CAAC,EAAEsC,CAAC,EAAE0E,EAAEpF,EAAES,CAAC,CAAC,CAAC,IAAIE,EAAE,GAAGC,EAAE,EAAEC,EAAE,IAAI,MAAMd,CAAC,EAAE,KAAK,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAED,EAAE,EAAE,EAAE,CAAC,IAAIsF,EAAEvH,EAAE,EAAE6C,GAAG,GAAG0E,EAAE,EAAE,MAAM,IAAI,MAAMizC,EAAE,gDAAgD,EAAEjzC,CAAC,CAAC,EAAE,GAAGA,GAAGrF,EAAE,MAAM,IAAI,MAAMs4C,EAAE,kDAAkD,EAAEjzC,EAAErF,CAAC,CAAC,EAAE,EAAEc,EAAEuE,GAAGzE,EAAEA,GAAGyE,GAAGxE,EAAEA,EAAEwE,CAAC,CAAC,IAAIrD,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAEhC,EAAE,EAAE,EAAE,CAAC,IAAIqF,EAAEvE,EAAE,KAAK,EAAEb,EAAE,GAAGoF,EAAErD,EAAEA,GAAG,CAACqD,EAAEvE,EAAE,GAAG,KAAK,IAAIA,EAAE,GAAG,CAAC,EAAE,EAAE,IAAIA,EAAE,IAAIA,EAAE,EAAE,GAAG,CAAC,GAAGkB,GAAGpB,EAAE,CAAC,IAAI,EAAE9C,EAAEuH,EAAEjH,EAAE,QAAQ,EAAE,EAAE,EAAE2B,EAAE,EAAE,EAAEW,EAAE,GAAG,EAAE,MAAM,CAAC,EAAE,CAACX,EAAEY,CAAC,EAAE0E,EAAEpF,EAAES,CAAC,CAAC,KAAK,CAAC,IAAI,EAAEI,EAAEd,EAAE,GAAGqF,EAAEuf,EAAE,kBAAkB5mB,EAAE,EAAE2C,CAAC,EAAE,EAAEikB,EAAE,kBAAkBvmB,EAAE,CAAC,EAAEoE,EAAE,IAAI,MAAMzC,CAAC,EAAE,KAAK,CAAC,EAAE,QAAQ0C,EAAE,EAAEA,EAAE3C,EAAE,EAAE2C,EAAE,CAAC,IAAIC,EAAE7E,EAAE4E,EAAE/B,GAAGiC,EAAEH,EAAEE,GAAG2C,GAAG3C,IAAI,EAAE,EAAE7B,EAAE6B,EAAE,IAAIC,EAAEH,EAAEE,KAAK,QAAQG,EAAE,EAAEA,EAAEnC,EAAE,EAAEmC,EAAEuC,EAAEC,EAAE3E,EAAEmC,GAAGhF,EAAE4E,EAAE/B,EAAEmC,GAAG,EAAEwC,GAAGlH,EAAEsE,GAAGhC,EAAEgC,GAAG4C,CAAC,CAAC,QAAQ5C,EAAE,EAAEA,EAAE1C,EAAE,EAAE0C,EAAE,GAAGD,EAAEC,KAAK,EAAE,CAAC,IAAIE,EAAEF,IAAI,EAAE,EAAE5B,EAAE4B,EAAE,GAAG2C,EAAEzC,EAAEjC,EAAE,GAAG+B,EAAE,QAAQ4C,EAAE,EAAEA,EAAE3E,EAAE,EAAE2E,EAAED,EAAEzC,EAAEjC,EAAE2E,GAAG,EAAE,EAAE1C,GAAG,CAAC,CAAC,MAAM,CAACyC,EAAE,CAAC,EAAE1E,CAAC,EAAE,EAAEV,EAAES,CAAC,CAAC,CAAC,CAAC,SAASoxE,GAAGh0E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEumB,EAAE,cAAcxmB,CAAC,EAAE,EAAE,EAAE,GAAG2B,EAAE1B,EAAE,OAAO2B,EAAE,CAAC,EAAEC,EAAE,EAAES,EAAE,GAAG,QAAQuB,EAAE,EAAEA,EAAElC,EAAE,EAAEkC,EAAE,CAAC,IAAIoD,EAAEhH,EAAE4D,GAAG,GAAGoD,IAAI,GAAG,CAAC,GAAG3E,IAAI,GAAG,MAAM,IAAI,MAAM43C,EAAE,yDAAyD53C,EAAEuB,CAAC,CAAC,EAAEvB,EAAEuB,EAAEjC,EAAE,KAAK,CAAC,CAAC,KAAK,CAAC,GAAGqF,EAAE,EAAE,MAAM,IAAI,MAAMizC,EAAE,8CAA8Cr2C,EAAEoD,CAAC,CAAC,EAAEpF,GAAGoF,EAAErF,EAAE,KAAKqF,CAAC,CAAC,CAAC,CAAC,GAAG3E,IAAI,GAAG,CAAC,GAAGT,GAAG,EAAE,MAAM,IAAI,MAAMq4C,EAAE,qDAAqD,CAAC,EAAE,IAAIr2C,EAAE,KAAK,MAAM,EAAEhC,CAAC,EAAE,GAAGA,EAAEgC,IAAI,EAAE,MAAM,IAAI,MAAMq2C,EAAE,gDAAgDl6C,EAAE4B,CAAC,CAAC,EAAEA,EAAEU,GAAGuB,CAAC,CAAC,GAAG2iB,EAAE,cAAc5kB,CAAC,IAAI,EAAE,MAAM,IAAI,MAAMs4C,EAAE,gDAAgDl6C,EAAE4B,CAAC,CAAC,EAAE,IAAIY,EAAExC,EAAE,OAAOyC,EAAE,CAAC,EAAE,GAAGD,EAAE,EAAE,CAACC,EAAED,EAAE,GAAG,EAAE,QAAQqB,EAAErB,EAAE,EAAEqB,GAAG,EAAE,EAAEA,EAAEpB,EAAEoB,GAAGpB,EAAEoB,EAAE,GAAG7D,EAAE6D,EAAE,EAAE,CAAC,IAAInB,EAAE,CAAC,EAAE,GAAGf,EAAE,EAAE,CAACe,EAAEf,EAAE,GAAG,EAAE,QAAQkC,EAAElC,EAAE,EAAEkC,GAAG,EAAE,EAAEA,EAAEnB,EAAEmB,GAAGnB,EAAEmB,EAAE,GAAGjC,EAAEiC,EAAE,EAAE,CAAC,IAAID,EAAE4iB,EAAE,kBAAkB5mB,EAAE,EAAE+B,CAAC,EAAE,QAAQkC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIoD,EAAE,EAAE,QAAQxC,EAAE,EAAEA,EAAEjC,EAAE,EAAEiC,EAAEwC,GAAGvH,EAAEmE,EAAErB,EAAEiC,GAAGhC,EAAEgC,GAAG,QAAQA,EAAE,EAAEA,EAAE9C,EAAE,EAAE8C,EAAEb,EAAEC,EAAElC,EAAE8C,GAAG,KAAK,MAAMwC,EAAEvE,EAAE+B,EAAE,EAAEwC,GAAGvE,EAAE+B,EAAE,CAAC,MAAM,CAACb,EAAE,CAAC,EAAEjC,CAAC,EAAEC,CAAC,CAAC,CAAC,SAAS+xE,GAAGj0E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,GAAG,EAAE,EAAE,CAAC,IAAI0B,EAAE3B,EAAE,OAAO4B,EAAE,CAAC,EAAE,GAAGlC,EAAE,OAAO,EAAE,EAAE,EAAEmC,EAAED,EAAE,GAAG,EAAED,EAAE,EAAE1B,EAAE0B,EAAE,GAAG,EAAE,EAAE,GAAG,EAAE,EAAE,MAAM,IAAI,MAAMu4C,EAAE,wDAAwD,CAAC,EAAE,IAAI13C,EAAE,EAAE,MAAM,EAAEA,EAAE,GAAG,EAAE,IAAIC,EAAED,EAAE,OAAO,CAAC6B,EAAEC,IAAID,EAAEC,EAAE,CAAC,EAAE5B,EAAE8jB,EAAE,kBAAkB5mB,EAAE6C,CAAC,EAAE,GAAGd,IAAI,EAAE,OAAO,EAAE,GAAGe,EAAE,KAAK,CAAC,EAAE,CAACA,EAAEF,CAAC,EAAE,GAAG,GAAG,EAAE,MAAM,IAAI,MAAM03C,EAAE,wDAAwD,CAAC,EAAE,IAAIt2C,EAAE,EAAEC,EAAE,EAAEoD,EAAE,EAAExC,EAAExE,EAAE2D,GAAG,OAAO,CAAC,IAAIS,EAAE,EAAE,GAAGR,EAAElC,EAAE,CAAC,GAAG0C,EAAEpE,EAAE4D,GAAGY,IAAIJ,EAAE,CAAC,EAAER,EAAE,QAAQ,CAAC,GAAGY,GAAGJ,EAAE,MAAM,IAAI,MAAM61C,EAAE,6DAA6D,CAAC,CAAC,CAAC,GAAGz1C,EAAE,GAAGA,GAAG,EAAE,MAAM,IAAI,MAAMy1C,EAAE,yDAAyDz1C,EAAE,CAAC,CAAC,EAAEA,EAAEwC,GAAGvE,EAAE,KAAK,EAAEuE,EAAEpF,EAAE4C,EAAE5C,CAAC,EAAE,QAAQyC,EAAEV,EAAEU,EAAET,EAAE,EAAES,EAAE,CAAC,IAAIC,EAAEvE,EAAEsE,GAAG,GAAGC,EAAE,GAAGA,GAAG3C,EAAE,GAAG,MAAM,IAAI,MAAMs4C,EAAE,uDAAuD51C,EAAEtE,EAAEsE,GAAG1C,EAAE,EAAE,CAAC,EAAE,QAAQ4C,EAAE,EAAEA,EAAE3C,EAAE2C,IAAI9B,EAAE+B,EAAE5C,EAAE2C,IAAI9E,EAAE6E,EAAE1C,EAAE2C,EAAE,CAAC,GAAG,EAAE,QAAQF,EAAE,EAAEA,EAAEzC,EAAEyC,IAAI5B,EAAE+B,EAAE5C,EAAEyC,IAAIT,EAAED,EAAE,GAAGA,EAAEC,EAAE,EAAEA,EAAEoD,EAAExC,EAAE,EAAEA,EAAEJ,EAAER,EAAElC,EAAE,KAAK,CAAC,OAAOsF,EAAE,GAAGvE,EAAE,KAAK,EAAEuE,EAAEpF,EAAE,EAAEA,CAAC,EAAE,CAACa,EAAEF,CAAC,CAAC,CAAC,IAAIoxE,GAAG4B,GAAG91E,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEs5E,GAAGvD,GAAGhyD,GAAG/jB,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEu5E,GAAG,CAAC,WAAWx1D,GAAG,YAAY,MAAM,WAAWu1D,EAAE,EAAMnF,GAAGY,GAAG,CAAC/0E,EAAE,IAAI,CAAC,IAAIE,EAAEF,EAAE,EAAE,OAAOE,EAAEA,CAAC,CAAC,EAAEs5E,GAAG/D,GAAGhxD,GAAG0vD,EAAE,EAAEsF,GAAG,CAAC,WAAWh1D,GAAG,YAAY,MAAM,WAAW+0D,EAAE,EAAE,SAASpF,GAAGp0E,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEgwB,GAAGvwB,EAAE,EAAE,KAAK,EAAE,QAAQ,EAAE,EAAE,EAAEO,EAAE,KAAK,IAAI,CAAC,IAAI,EAAEA,EAAE,WAAW,CAAC,EAAE0B,EAAE,IAAI,MAAM,EAAE,MAAM,EAAE,QAAQC,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAID,EAAEC,GAAG,EAAEA,GAAGhC,EAAEgC,GAAG5B,EAAE4B,GAAG3B,EAAE,IAAI,EAAE,IAAI,GAAG0B,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,OAAO1B,CAAC,CAAC,IAAIm5E,GAAG,KAAK,CAAC,YAAY54E,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,KAAK,UAAU8kB,EAAE,aAAahmB,CAAC,EAAE,KAAK,YAAYZ,EAAE,KAAK,QAAQ4mB,EAAE,aAAa,CAAC,EAAE,KAAK,SAASA,EAAE,aAAavmB,CAAC,EAAE,KAAK,SAASwB,EAAE,KAAK,cAAcC,CAAC,CAAC,YAAYlB,EAAE,CAAC,OAAO,KAAK,IAAI,KAAK,SAAS,EAAEA,EAAE,EAAE,KAAK,SAASA,EAAE,CAAC,CAAC,CAAC,aAAaA,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,YAAYA,CAAC,EAAE,OAAO,KAAK,IAAI,EAAEY,EAAE,EAAE,EAAEZ,EAAE,CAAC,CAAC,CAAC,aAAaY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,QAAQC,EAAE,EAAEA,EAAEF,EAAE,EAAEE,EAAE,CAAC,IAAIC,EAAE,KAAK,YAAYF,CAAC,EAAE,EAAE,KAAK,IAAI,EAAEE,EAAED,CAAC,EAAEW,EAAE,KAAK,IAAI,EAAEV,GAAGH,GAAGE,EAAE,GAAG,EAAE,EAAED,GAAG,EAAEY,GAAGE,EAAE5C,GAAG,EAAE,EAAE,EAAE+B,EAAEC,GAAGa,EAAE,EAAEA,GAAG,EAAE,KAAK,QAAQ,OAAO,QAAQgC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAEhC,GAAGjC,EAAEgC,EAAEiC,GAAG,OAAOhC,GAAGH,EAAE,KAAK,SAAS,OAAOG,IAAI,EAAEH,EAAE,EAAE,GAAG,KAAK,UAAU,OAAO,EAAErC,EAAE0B,GAAG,IAAI,WAAWc,CAAC,EAAE,IAAI,EAAE,EAAExC,EAAE0B,GAAGkC,EAAE,EAAEoD,EAAExC,GAAGA,EAAE,QAAQJ,GAAG,EAAER,KAAKQ,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAEwC,EAAE,KAAK,OAAO,EAAEA,EAAE,KAAK,SAAS,EAAE,QAAQxC,EAAE,EAAEA,EAAE,EAAE,EAAE,EAAEA,EAAEwC,EAAEzG,EAAEgC,EAAEiC,EAAE,EAAEwC,EAAE,KAAK,SAAS,EAAE,GAAG,EAAE,EAAE,CAACA,EAAEzG,EAAEgC,EAAE,EAAE,EAAE,EAAE,QAAQiC,EAAE,EAAEA,EAAEnC,EAAE,EAAEmC,EAAEwC,EAAE,KAAK,SAAS,EAAEA,EAAE,KAAK,QAAQ,CAAC,KAAK,CAAC,QAAQxC,EAAE,EAAEA,EAAEnC,EAAE,EAAE,EAAEmC,EAAEwC,EAAE,KAAK,QAAQ,EAAEA,EAAE,KAAK,SAAS,EAAEA,EAAE,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,QAAQzG,EAAEZ,EAAE,CAAC,IAAI,EAAEY,EAAE,OAAOP,EAAEL,EAAE,OAAO,GAAGK,EAAE,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,GAAG,GAAGgC,IAAI,EAAE,MAAM,IAAI,MAAM,oCAAoCA,GAAG,EAAE,QAAQ,EAAE,EAAE,EAAE3B,EAAE,EAAE,EAAE,CAAC,IAAIqC,EAAE1C,EAAE,IAAIgC,EAAE,GAAGU,EAAEA,GAAG1C,EAAE,IAAI,EAAE,CAAC0C,EAAE,MAAM,IAAI,MAAM,uBAAuB1C,EAAE,mBAAmBgC,MAAM,IAAI,EAAEA,EAAEhC,EAAE,EAAE,CAAC,GAAGgC,IAAI,EAAE,MAAM,IAAI,MAAM,gDAAgD,UAAUA,GAAG,CAAC,CAAC,IAAIH,EAAExB,EAAE,EAAEyB,EAAE8kB,EAAE,kBAAkB,QAAQvmB,CAAC,EAAE,GAAG,IAAI,GAAGA,IAAI,EAAE,CAAC,IAAI2B,EAAE,IAAI,MAAM,CAAC,EAAE,QAAQ,EAAE,EAAE,GAAGH,EAAE,EAAE,EAAEC,EAAE,GAAG,EAAE,MAAM,CAACE,EAAEF,CAAC,CAAC,CAACA,EAAE,GAAG,EAAE,QAAQE,EAAE,EAAEA,GAAGH,EAAE,EAAEG,EAAE,CAAC,IAAI,EAAEhC,EAAEgC,GAAGhC,EAAEgC,EAAE,GAAGU,EAAE,EAAE,KAAK,YAAY,QAAQ,GAAG,CAACA,GAAG,KAAK,aAAa,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,eAAe,EAAE,GAAGA,IAAI,IAAIA,EAAE,GAAGZ,EAAEE,GAAGF,EAAEE,EAAE,GAAGU,CAAC,CAAC,IAAIX,EAAE,IAAI,MAAMD,EAAED,EAAE,EAAE,QAAQG,EAAE,EAAEA,EAAEH,EAAE,EAAEG,EAAE,CAAC,IAAI,EAAEhC,EAAEgC,GAAGU,EAAEZ,EAAEE,GAAG,GAAG,KAAK,YAAY,QAAQ,GAAG,CAAC,IAAIY,EAAE5C,EAAEgC,EAAE,GAAGhC,EAAEgC,GAAGa,EAAE,KAAK,aAAaD,EAAE,CAAC,EAAE,KAAK,aAAahC,EAAE,EAAEmB,EAAEW,EAAEG,EAAE,CAAC,EAAEH,GAAGG,CAAC,CAAC,EAAE,KAAK,eAAeH,IAAIZ,EAAEE,GAAG,CAAC,IAAI,EAAEhC,EAAEgC,EAAE,GAAGhC,EAAEgC,GAAG,GAAG,IAAI,EAAE,SAAS,IAAIY,EAAE,EAAE,EAAE,KAAK,SAASC,EAAE,EAAE,KAAK,aAAajC,EAAE,EAAEmB,EAAEW,EAAEG,EAAED,CAAC,CAAC,CAAC,CAAC,MAAM,CAACb,EAAED,CAAC,CAAC,CAAC,EAAE,SAASqyE,GAAGr0E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAE,CAAC,OAAO,IAAIy3E,GAAGx5E,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,CAAC,EAAE,QAAQjC,EAAE,CAAC,CAAC,CAAC,SAAS25E,GAAG35E,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACN,EAAE,OAAO,OAAO,GAAG,EAAE,SAAS,EAAE,CAAC,QAAQ,EAAE,EAAE,EAAEA,EAAE,OAAO,EAAE,EAAEM,EAAE,KAAKN,EAAE,SAAS,EAAE,EAAE,CAAC,CAAC,EAAE,MAAM,CAAC,GAAG,EAAE,SAAS,EAAE,CAAC,IAAI,EAAE,EAAE,GAAG,EAAEA,EAAE,QAAQ,CAAC,EAAE,KAAK,IAAI,IAAI,CAAC,IAAIiC,EAAEjC,EAAE,SAAS,EAAE,CAAC,GAAG,CAACE,GAAG+B,EAAE,SAAS,IAAI3B,EAAE,KAAK2B,CAAC,EAAEjC,EAAEA,EAAE,SAAS,EAAE,CAAC,EAAE,EAAEA,EAAE,QAAQ,CAAC,CAAC,EAAE,CAACE,GAAGF,EAAE,SAAS,IAAIM,EAAE,KAAKN,CAAC,EAAE,MAAM,CAAC,IAAIO,EAAE,EAAE,QAAQ,EAAE,EAAE,EAAEP,EAAE,OAAO,EAAE,IAAI,GAAG,IAAIA,EAAE,QAAQ,EAAE,QAAQA,EAAE,EAAE,IAAI,GAAG,CAAC,IAAI,EAAEA,EAAE,SAASO,EAAE,CAAC,GAAG,CAACL,GAAG,EAAE,SAAS,IAAII,EAAE,KAAK,CAAC,EAAEC,EAAE,EAAE,CAAC,CAAC,CAAC,SAAS+zE,GAAGt0E,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,OAAOO,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE0B,EAAE,IAAI,MAAM3B,CAAC,EAAE,QAAQwC,EAAE,EAAEA,EAAExC,EAAE,EAAEwC,EAAE,CAAC,IAAIC,EAAExC,EAAE,OAAOo5E,GAAG35E,EAAE8C,GAAG,EAAE5C,EAAEK,CAAC,EAAE,IAAIyC,EAAEzC,EAAE,OAAOwC,EAAEd,EAAEa,GAAGE,EAAE,GAAGA,EAAE,EAAE,KAAK,IAAI,EAAEA,CAAC,CAAC,CAAC,IAAId,EAAE4kB,EAAE,kBAAkB,QAAQ,EAAE,CAAC,EAAE3kB,EAAE,IAAI,MAAM,CAAC,EAAES,EAAE,CAACtC,EAAE,CAAC,EAAEuC,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAExC,EAAE,EAAEwC,EAAE,QAAQC,EAAE,EAAEA,EAAEd,EAAEa,GAAG,EAAEC,EAAEb,EAAEW,EAAE,GAAGC,EAAEZ,EAAEW,EAAE,EAAE,GAAGE,EAAEZ,EAAEU,GAAGtC,EAAEsC,GAAG,EAAEA,EAAE,MAAM,CAACX,EAAEC,EAAES,CAAC,CAAC,CAAC,SAAS2xE,GAAGv0E,EAAE,EAAE,CAAC,IAAIE,EAAE4mB,EAAE,kBAAkB,QAAQ9mB,EAAE,MAAM,EAAE,QAAQM,EAAE,EAAEA,EAAEN,EAAE,OAAO,EAAEM,EAAEJ,EAAEI,GAAGwmB,EAAE,cAAc9mB,EAAEM,EAAE,EAAE,OAAO,CAAC,EAAE,mBAAmB,EAAE,OAAOJ,CAAC,CAAC,IAAIs0E,GAAGO,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,CAAC,EAAE45E,GAAGlE,GAAG,CAAC11E,EAAE,EAAEE,EAAEI,KAAK,CAAC,KAAKN,EAAEE,EAAE,KAAK,EAAEI,CAAC,EAAE,EAAEu5E,GAAGpE,GAAG1wD,GAAGyvD,GAAGoF,EAAE,EAAEE,GAAG,CAAC,WAAW/0D,GAAG,YAAY,MAAM,WAAW80D,EAAE,EAAE,SAASpF,GAAGz0E,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,MAAMF,EAAE,IAAI,EAAE,QAAQO,EAAE,EAAEA,EAAEL,EAAE,OAAOK,IAAIL,EAAEK,GAAGP,EAAE,MAAMO,GAAG,EAAEA,GAAG,IAAID,EAAEiwB,GAAGrwB,EAAEF,EAAE,KAAK,EAAE,QAAQO,EAAE,EAAEA,EAAED,EAAE,OAAO,OAAO,EAAEC,EAAE,CAAC,IAAI,EAAED,EAAE,WAAWC,CAAC,EAAE,EAAE,IAAI,MAAMP,EAAE,IAAI,EAAE,QAAQkC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,GAAG,EAAEA,GAAGlC,EAAE,MAAMkC,GAAG,IAAID,EAAEjC,EAAE,WAAW,CAAC,EAAEM,EAAE,OAAOC,GAAGP,EAAE,OAAOiC,EAAE,CAAC,OAAO3B,CAAC,CAAC,IAAIy5E,GAAG,CAAC/5E,EAAE,IAAI,CAAC,IAAIE,EAAE,EAAE,MAAMF,EAAE,MAAM,OAAOE,IAAI,EAAEF,EAAE,MAAM,EAAE,MAAME,CAAC,EAAE,SAAS85E,GAAGh6E,EAAE,EAAEE,EAAE,EAAEI,EAAEN,EAAE,OAAO,EAAE,CAAC,KAAKM,EAAEJ,GAAG,CAAC,GAAGI,EAAEJ,EAAE,IAAI,CAAC,IAAI+B,EAAE3B,EAAEJ,EAAE,EAAEgC,EAAE,EAAEhC,EAAE,EAAEiC,EAAE,KAAK,IAAIF,CAAC,EAAEW,EAAE,GAAG,KAAK,IAAI,EAAET,EAAE,CAAC,EAAEU,EAAE,GAAG,KAAK,KAAKV,EAAES,GAAGX,EAAEW,GAAGX,CAAC,EAAE,KAAK,KAAKC,EAAED,EAAE,CAAC,EAAEa,EAAE,KAAK,IAAI5C,EAAE,KAAK,MAAM,EAAEgC,EAAEU,EAAEX,EAAEY,CAAC,CAAC,EAAEE,EAAE,KAAK,IAAIzC,EAAE,KAAK,MAAM,GAAG2B,EAAEC,GAAGU,EAAEX,EAAEY,CAAC,CAAC,EAAEm3E,GAAGh6E,EAAE,EAAE8C,EAAEC,CAAC,CAAC,CAAC,IAAIxC,EAAEP,EAAE,GAAG,EAAEE,EAAE,EAAEI,EAAE,IAAIwmB,EAAE,KAAK9mB,EAAEE,EAAE,CAAC,EAAE65E,GAAG/5E,EAAEM,GAAGC,CAAC,EAAE,GAAGumB,EAAE,KAAK9mB,EAAEE,EAAEI,CAAC,EAAE,EAAE,GAAG,CAAC,IAAIwmB,EAAE,KAAK9mB,EAAE,EAAE,CAAC,EAAE,IAAI,IAAI+5E,GAAG/5E,EAAE,GAAGO,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,KAAKw5E,GAAG/5E,EAAE,GAAGO,CAAC,EAAE,GAAG,EAAE,EAAE,CAAC,CAACw5E,GAAG/5E,EAAEE,GAAGK,CAAC,IAAI,EAAEumB,EAAE,KAAK9mB,EAAEE,EAAE,CAAC,GAAG,EAAE,EAAE,EAAE4mB,EAAE,KAAK9mB,EAAE,EAAEM,CAAC,GAAG,GAAG,IAAIJ,EAAE,EAAE,GAAG,GAAG,IAAII,EAAE,EAAE,EAAE,CAAC,CAAC,SAASo0E,GAAG10E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,EAAE,EAAE,OAAO,GAAG,CAAC,EAAE0B,CAAC,EAAE,CAACjC,EAAE,OAAO,EAAE,CAAC,EAAEkC,EAAE4kB,EAAE,uBAAuB5mB,EAAE,EAAEI,CAAC,EAAE6B,EAAE2kB,EAAE,uBAAuB,QAAQ,EAAExmB,CAAC,EAAE,QAAQuC,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIC,EAAED,EAAEZ,EAAEc,EAAE/C,EAAE,SAAS8C,EAAEA,EAAEb,CAAC,EAAEe,EAAE,IAAI,MAAMD,EAAE,MAAM,EAAEA,EAAE,QAAQ,CAAC,EAAE4B,IAAI3B,EAAE2B,GAAG,CAAC,MAAM,EAAE,MAAMA,CAAC,CAAC,EAAErE,EAAE0C,EAAE,SAASg3E,GAAGh3E,EAAE1C,CAAC,EAAE0C,EAAEA,EAAE,MAAM,EAAE1C,CAAC,GAAGC,GAAGyC,EAAE,KAAK+2E,EAAE,EAAE,IAAI71E,EAAErB,EAAEvC,EAAE,EAAE4B,EAAE,SAASgC,EAAEA,EAAE5D,CAAC,EAAEiH,EAAEpF,EAAE,SAAS+B,EAAEA,EAAE5D,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,IAAI,EAAE,GAAG0C,EAAE,GAAG,MAAMuE,EAAE,GAAGvE,EAAE,GAAG,KAAK,CAAC,IAAIJ,EAAE,EAAE,MAAM,EAAE,OAAOA,EAAEA,EAAE,OAAO,GAAGtC,EAAE,CAACiwB,GAAG3tB,EAAE1C,EAAEgC,CAAC,EAAEquB,GAAG3tB,EAAE,QAAQT,CAAC,CAAC,CAAC,CAAC,SAASyyE,GAAG50E,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEumB,EAAE,eAAe,EAAE5mB,CAAC,EAAE,GAAG,EAAE,CAAC,EAAEA,EAAE,GAAG,CAAC,EAAE,QAAQ8C,EAAE,EAAEA,EAAEzC,EAAEyC,IAAI,EAAE,IAAI9C,EAAE8C,GAAG,EAAE,GAAG9C,EAAEK,GAAG,QAAQyC,EAAEzC,EAAE,EAAEyC,EAAE9C,EAAE,OAAO8C,IAAI,EAAE,IAAI9C,EAAE8C,GAAG,IAAI,EAAE,CAAC,EAAEf,EAAE,IAAI,WAAW/B,EAAEK,EAAE,EAAE2B,EAAE,IAAIonB,GAAG,EAAEhpB,EAAEN,CAAC,EAAEmC,EAAE,CAAC,EAAES,EAAE,EAAE,KAAK,GAAG,EAAE,KAAK,EAAE,QAAQI,EAAE,EAAEA,EAAE9C,EAAEK,GAAGyC,IAAI,CAAC,IAAIkB,EAAE,GAAGtB,EAAEsB,EAAElE,EAAEgD,GAAG,SAAS,MAAM,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQuE,EAAE,EAAEA,EAAE,EAAE,GAAGA,IAAI,QAAQ,EAAE,EAAE,EAAE,EAAE,GAAG,IAAI,EAAE,KAAKrF,EAAE,IAAIqF,EAAEvE,EAAE,CAAC,CAAC,EAAEkB,EAAE,EAAE,KAAK,GAAG,CAAC,CAAC,GAAG,EAAEA,KAAK,OAAOjC,EAAEe,GAAG,EAAEkB,OAAO,CAAC,IAAI,EAAE,OAAO,KAAK,CAAC,EAAE,OAAO,EAAEA,GAAG,EAAEjC,EAAEe,GAAG,EAAEb,EAAE,KAAKa,CAAC,CAAC,CAAC,CAAC,IAAIH,EAAE,EAAE,MAAM,EAAEA,EAAE,GAAG,OAAO,KAAK,CAAC,EAAE,OAAO,IAAIC,EAAE,IAAIwmB,GAAGzmB,EAAEvC,CAAC,EAAE6B,EAAE,QAAQ,CAACa,EAAEkB,IAAI,CAAC,QAAQ,EAAE,EAAE,EAAE,EAAE,GAAG,IAAI,QAAQqD,EAAE,EAAEA,EAAE,EAAE,GAAGA,IAAIzE,EAAE,IAAIZ,EAAE,IAAI,EAAEc,EAAEuE,CAAC,EAAE,EAAErD,EAAEqD,CAAC,CAAC,CAAC,EAAE,IAAIxE,EAAE7C,EAAE,MAAM,EAAE,OAAO6C,EAAExC,GAAGsC,EAAE,GAAG,CAAC,aAAaC,EAAE,OAAO,YAAYC,EAAE,QAAQd,CAAC,CAAC,CAACkyB,GAAG,MAAM,IAAI,IAAIw9C,GAAG,CAAC,EAAE,IAAIsI,GAAGlE,GAAG13D,GAAGre,GAAGA,GAAG,EAAEA,EAAE,KAAK,IAAIA,CAAC,EAAE,CAAC,EAAEk6E,GAAG,CAAC,WAAW77D,GAAG,YAAY,MAAM,WAAW47D,EAAE,EAAE,SAASE,GAAGn6E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAEmxE,GAAG,CAAClxE,CAAC,EAAE,WAAW,EAAE,IAAI,EAAEumB,EAAE,cAAcvmB,EAAE,KAAK,EAAE0B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAE4kB,EAAE,uBAAuB,UAAU,CAAC,EAAE,QAAQ3kB,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAID,EAAEC,GAAGF,EAAEE,GAAG,EAAE,EAAEF,EAAEE,GAAGF,EAAEE,GAAG,OAAOjC,EAAE,eAAeK,EAAE,MAAM,UAAU2B,CAAC,CAAC,CAAC,IAAIk4E,GAAG,CAAC,WAAWx6D,GAAG,YAAY,MAAM,WAAWu6D,EAAE,EAAME,GAAGtF,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,EAAE,EAAEA,EAAEA,CAAC,EAAE,SAASs6E,GAAGt6E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,EAAE,MAAMC,CAAC,EAAE,EAAEkxE,GAAG,CAACnxE,EAAEC,CAAC,EAAE,OAAO,EAAE,IAAI,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,EAAEJ,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAAC0B,EAAEC,CAAC,EAAEm4E,GAAG/5E,EAAE,MAAMC,EAAE,MAAM,EAAE,EAAE,SAAS,EAAE,OAAOL,EAAE,eAAegC,EAAE,UAAUD,CAAC,CAAC,CAAC,IAAIs4E,GAAG,CAAC,WAAWp4D,GAAG,YAAY,MAAM,WAAWm4D,EAAE,EAAME,GAAGzE,GAAGpzD,GAAG3iB,GAAG,KAAK,IAAI,EAAEA,CAAC,CAAC,EAAEy6E,GAAG,CAAC,WAAW93D,GAAG,YAAY,MAAM,WAAW63D,EAAE,EAAME,GAAG3E,GAAG9yD,GAAGjjB,GAAG,KAAK,IAAI,KAAK,IAAI,EAAEA,CAAC,EAAE,CAAC,CAAC,EAAE26E,GAAG,CAAC,WAAW13D,GAAG,YAAY,MAAM,WAAWy3D,EAAE,EAAE,SAASE,GAAG56E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,GAAGL,IAAI,SAAS,OAAOi1E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn1E,CAAC,CAAC,EAAE,GAAGE,IAAI,OAAO,OAAOs6E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQx6E,CAAC,CAAC,EAAE,GAAGE,IAAI,MAAM,OAAO+5E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQj6E,CAAC,CAAC,EAAE,GAAGE,IAAI,QAAQ,OAAOw6E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ16E,CAAC,CAAC,EAAE,GAAGE,IAAI,QAAQ,OAAOo6E,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,MAAMh6E,CAAC,EAAE,QAAQN,CAAC,CAAC,EAAE,GAAGE,IAAI,YAAY,OAAOi6E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn6E,EAAE,MAAM,CAAC,MAAMO,CAAC,CAAC,CAAC,EAAE,GAAGL,IAAI,UAAU,OAAOg5E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQl5E,CAAC,CAAC,EAAE,MAAM,IAAI,MAAM,cAAcE,iDAAiD,CAAC,CAAC,SAAS26E,GAAG76E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,EAAEwmB,EAAE,cAAcvmB,EAAE,KAAK,EAAE0B,EAAE6kB,EAAE,uBAAuB,EAAE,CAAC,EAAE5kB,EAAE4kB,EAAE,cAAc7kB,CAAC,EAAE6kB,EAAE,OAAO,IAAI5kB,EAAE,IAAI,kBAAkBD,UAAUC,iCAAiC3B,EAAE,cAAc,gFAAgF,EAAEL,EAAE,OAAOK,EAAE,MAAM,EAAE,IAAI4B,EAAEjC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,GAAG4B,EAAE,oBAAoB,KAAK,CAAC,IAAIS,EAAET,EAAE,mBAAmB,KAAKU,EAAEV,EAAE,mBAAmB,KAAKS,EAAE,MAAMX,EAAEY,EAAE,MAAMZ,CAAC,CAAC,MAAM,CAAC,OAAO1B,EAAE,OAAO,MAAM0B,EAAE,MAAM1B,EAAE,KAAK,CAAC,CAAC,IAAIu6E,GAAG,CAAC,WAAWl4D,GAAG,YAAY,MAAM,WAAWi4D,EAAE,EAAE,SAASE,GAAG/6E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,WAAW0B,CAAC,EAAE3B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,QAAQ,EAAE,IAAI2B,EAAE3B,EAAE,MAAM,OAAO4B,EAAE,EAAE,MAAM,OAAOS,EAAE,EAAErC,EAAE,MAAM2B,EAAE,GAAG3B,EAAE,MAAM2B,EAAE,GAAGW,EAAEZ,EAAE,EAAE,MAAME,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGW,EAAE,EAAEvC,EAAE,MAAM2B,EAAE,GAAG3B,EAAE,MAAM2B,EAAE,GAAGa,EAAEd,EAAE,EAAE,MAAME,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGa,EAAEzC,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE2D,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE4iB,EAAE,cAAc9jB,CAAC,EAAEuE,EAAEuf,EAAE,cAAc5iB,CAAC,EAAES,EAAEowB,GAAG,2BAA2Bx0B,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,CAAC,EAAE,OAAO,CAACuC,EAAEC,CAAC,CAAC,EAAE+jB,EAAE,OAAOlkB,IAAIC,EAAE,IAAI,kCAAkCD,WAAWC,6BAA6BtC,EAAE,aAAa,EAAE,wBAAwB,oBAAoB0B,eAAe,EAAE,IAAI2C,EAAE,EAAE,CAAC,EAAEhC,EAAEE,CAAC,EAAE,CAAC,EAAEA,EAAEF,CAAC,EAAEiC,EAAE5C,EAAE,CAACsF,EAAExE,EAAEF,CAAC,EAAE,CAAC0E,EAAE1E,EAAEE,CAAC,EAAE+B,EAAE+1E,GAAG,CAAC,OAAO,CAAC,EAAEt6E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM0E,CAAC,CAAC,CAAC,EAAE4C,EAAEqzE,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAM2E,CAAC,CAAC,CAAC,EAAEG,EAAE,EAAEF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG2C,EAAE,EAAE3C,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGG,EAAEhD,EAAEuF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGtC,EAAE,KAAK,IAAI,EAAEqC,CAAC,EAAEpC,EAAEjF,EAAE,KAAK,IAAI4E,EAAE,MAAM,EAAE,OAAOM,EAAElF,EAAE,KAAK,IAAIsH,EAAE,MAAM,EAAE,OAAOnC,EAAEyhB,EAAE,eAAehiB,EAAE,KAAK,EAAEQ,EAAEwhB,EAAE,eAAetf,EAAE,KAAK,EAAE,CAACS,EAAEN,EAAEO,CAAC,EAAE,EAAE,CAAC7C,EAAE,GAAG,EAAEA,EAAE,EAAE,EAAE,CAACA,EAAE,GAAGA,EAAE,GAAG,CAAC,EAAE,CAAC8C,EAAGC,EAAGC,CAAE,EAAEpG,EAAE,CAAC,EAAEqD,EAAE,GAAGA,EAAE,EAAE,EAAE,CAACA,EAAE,GAAG,EAAEA,EAAE,EAAE,EAAEgD,GAAGb,EAAExC,EAAEsD,GAAGgoB,GAAG,CAACrrB,EAAEuC,EAAExC,CAAC,EAAEH,EAAE,KAAK,EAAE6B,GAAG4B,GAAG,OAAOE,GAAGvI,EAAE,UAAU,QAAQwI,GAAG,EAAEA,GAAGxD,EAAEwD,KAAK,QAAQC,GAAG,EAAEA,GAAGlB,EAAEkB,IAAIF,GAAG,QAAQG,GAAG,EAAEA,GAAG3D,EAAE2D,IAAIH,GAAG,QAAQI,GAAG,EAAEA,GAAG7D,EAAE6D,IAAIJ,GAAG,CAAC,IAAIK,GAAG,KAAK,IAAIH,GAAGF,GAAGhB,CAAC,EAAE0B,GAAG,KAAK,IAAIP,GAAGH,GAAGxD,CAAC,EAAEmE,GAAG,KAAK,IAAIP,GAAGJ,GAAGzD,CAAC,EAAE,QAAQqE,GAAGV,GAAGU,GAAGP,GAAGO,KAAK,QAAQzC,GAAGgC,GAAGhC,GAAGuC,GAAGvC,KAAK,CAAC,IAAIE,GAAG,EAAE,QAAQC,GAAG8B,GAAG9B,GAAGqC,GAAGrC,KAAK,CAAC,IAAIC,GAAG,KAAK,IAAI0B,GAAG,EAAE,CAAC,EAAET,EAAEqB,GAAG,KAAK,IAAIZ,GAAGnB,EAAE,CAAC,EAAEc,EAAGpB,GAAG9B,EAAE6B,GAAGqC,GAAG1B,EAAEZ,GAAGmB,GAAGhB,GAAG9B,EAAE2B,GAAGoB,EAAGvB,GAAGwB,EAAGkB,IAAIxC,IAAIG,GAAGC,EAAE,CAACP,GAAG+B,GAAGJ,IAAIe,GAAGpE,EAAE2B,MAAME,EAAE,CAAC,CAAC,OAAO5G,EAAE,8BAA8B4E,CAAC,EAAE5E,EAAE,8BAA8BsH,CAAC,EAAEtH,EAAE,eAAeyE,EAAE4D,GAAG,MAAMA,GAAG,MAAM,CAAC,CAAC,IAAIyyE,GAAG,CAAC,WAAW5+D,GAAG,YAAY,MAAM,WAAW2+D,EAAE,EAAE,SAASE,GAAGj7E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,WAAWC,EAAE,WAAWC,EAAE,WAAWS,EAAE,eAAeC,CAAC,EAAEvC,EAAEwC,EAAEC,EAAEC,EAAEkB,EAAE,CAAC,EAAEpB,EAAEi4E,GAAG,CAAC,OAAO,CAAC,EAAEx6E,EAAE,EAAE,CAAC,EAAE,MAAM,CAAC,WAAW2B,EAAE,WAAWC,CAAC,EAAE,QAAQjC,CAAC,CAAC,EAAE,IAAI6C,EAAE6yE,GAAG,CAAC,OAAO,CAAC,EAAE9yE,EAAE,EAAE,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAEgE,EAAE,KAAKpB,CAAC,EAAEA,EAAEC,GAAGH,IAAII,EAAE43E,GAAG16E,EAAE4C,EAAEF,EAAEX,EAAEY,CAAC,EAAEqB,EAAE,KAAKpB,CAAC,EAAEA,EAAEE,GAAG,QAAQuE,KAAKrD,EAAEhE,EAAE,8BAA8BqH,CAAC,EAAE,OAAOzE,CAAC,CAAC,IAAIo4E,GAAG,CAAC,WAAWp1D,GAAG,YAAY,MAAM,WAAWm1D,EAAE,EAAME,GAAGpF,GAAG56D,GAAGnb,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEo7E,GAAG,CAAC,WAAWjgE,GAAG,YAAY,MAAM,WAAWggE,EAAE,EAAME,GAAGtF,GAAG36D,GAAGpb,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAEs7E,GAAG,CAAC,WAAWlgE,GAAG,YAAY,MAAM,WAAWigE,EAAE,EAAE,SAASE,GAAGv7E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAE,EAAEmxE,GAAG,EAAE,MAAM,EAAE,IAAIlxE,EAAED,EAAE,IAAI2B,GAAG/B,EAAE,KAAK,IAAI+B,EAAE,MAAM,EAAE,MAAM,EAAE,EAAEsuB,GAAGjwB,EAAE,GAAG,MAAMA,EAAE,GAAG,KAAK,EAAE,EAAE,EAAE,OAAO,QAAQ2B,EAAE,EAAEA,EAAE3B,EAAE,OAAO2B,IAAI,CAAC,IAAIC,EAAE3B,EAAE0B,GAAG,QAAQE,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,IAAID,EAAEC,EAAE,CAAC,OAAOjC,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAIs7E,GAAG,CAAC,WAAWlgE,GAAG,YAAY,MAAM,WAAWigE,EAAE,EAAE,SAASE,GAAGz7E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI0B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE2B,EAAED,EAAEE,EAAEq4C,EAAE,mBAAmBt4C,EAAE3B,EAAE,MAAM,MAAM,EAAEqC,EAAErC,EAAE4B,GAAG,OAAOS,EAAEo1E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAED,EAAEs4C,EAAE,iBAAiBt4C,EAAE,OAAO3B,EAAE,MAAM,MAAM,GAAGi6C,EAAE,2BAA2B,MAAMt4C,EAAEU,EAAE,MAAM,MAAM,EAAE,GAAG,CAACC,EAAEC,CAAC,EAAE03C,EAAE,0BAA0B53C,EAAE,MAAMV,CAAC,EAAEa,EAAE+jB,EAAE,cAAchkB,CAAC,EAAEE,EAAE8jB,EAAE,oBAAoBA,EAAE,cAAcjkB,CAAC,EAAED,EAAE,KAAK,EAAEsB,EAAEhE,EAAE,KAAK,IAAI0C,EAAE,MAAM,EAAE,OAAO,QAAQ2E,EAAE,EAAEA,EAAEvE,EAAE,OAAO,EAAEuE,EAAE,CAAC,IAAI,EAAEA,EAAExE,EAAE4B,EAAET,EAAE,GAAG,QAAQU,EAAE,EAAEA,EAAE7B,EAAE,EAAE6B,EAAE,CAAC,IAAIC,EAAEX,EAAE,EAAEU,GAAGD,EAAEA,GAAGE,CAAC,CAAC7B,EAAEuE,GAAG5C,CAAC,CAACxC,GAAG,MAAMjC,EAAE,8BAA8B0C,CAAC,EAAE,IAAI,EAAE1C,EAAE,eAAe2C,EAAED,EAAE,MAAMI,CAAC,EAAE,GAAG,EAAE,CAAC,IAAIuE,EAAEizC,EAAE,qBAAqB33C,EAAEZ,CAAC,EAAE,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAMqH,CAAC,CAAC,CAAC,EAAE,OAAOrH,EAAE,8BAA8B,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,IAAIw7E,GAAG,CAAC,WAAWngE,GAAG,YAAY,MAAM,WAAWkgE,EAAE,EAAE,SAASE,GAAG37E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI0B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE2B,EAAED,EAAEE,EAAEq4C,EAAE,mBAAmBt4C,EAAE3B,EAAE,MAAM,MAAM,EAAEqC,EAAErC,EAAE4B,GAAG,OAAOS,EAAEo1E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAED,EAAEs4C,EAAE,iBAAiBt4C,EAAE,OAAO3B,EAAE,MAAM,MAAM,GAAGi6C,EAAE,2BAA2B,MAAMt4C,EAAEU,EAAE,MAAM,MAAM,EAAE,GAAG,CAACC,EAAEC,CAAC,EAAE03C,EAAE,0BAA0B53C,EAAE,MAAMV,CAAC,EAAEa,EAAE+jB,EAAE,cAAchkB,CAAC,EAAEE,EAAE8jB,EAAE,oBAAoBA,EAAE,cAAcjkB,CAAC,EAAED,EAAE,KAAK,EAAEsB,EAAEhE,EAAE,KAAK,IAAI0C,EAAE,MAAM,EAAE,OAAO,QAAQ2E,EAAE,EAAEA,EAAEvE,EAAE,OAAO,EAAEuE,EAAE,CAAC,IAAI,EAAEA,EAAExE,EAAE4B,EAAET,EAAE,GAAG,QAAQU,EAAE,EAAEA,EAAE7B,EAAE,EAAE6B,EAAE,CAAC,IAAIC,EAAEX,EAAE,EAAEU,GAAGD,EAAEA,GAAGE,CAAC,CAAC7B,EAAEuE,GAAG5C,CAAC,CAACxC,GAAG,MAAMjC,EAAE,8BAA8B0C,CAAC,EAAE,IAAI,EAAE1C,EAAE,eAAe2C,EAAED,EAAE,MAAMI,CAAC,EAAE,GAAG,EAAE,CAAC,IAAIuE,EAAEizC,EAAE,qBAAqB33C,EAAEZ,CAAC,EAAE,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAMqH,CAAC,CAAC,CAAC,EAAE,OAAOrH,EAAE,8BAA8B,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI07E,GAAG,CAAC,WAAWpgE,GAAG,YAAY,MAAM,WAAWmgE,EAAE,EAAE,SAASE,GAAG77E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,QAAQ,EAAE,IAAI,EAAEumB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE0B,EAAEu4C,EAAE,mBAAmB,EAAEj6C,EAAE,MAAM,MAAM,EAAE2B,EAAE3B,EAAE4B,EAAE,CAAC,EAAEF,GAAG,OAAOC,EAAE81E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK+B,CAAC,CAAC,CAAC,EAAEE,EAAE,KAAKD,CAAC,EAAE,EAAEs4C,EAAE,iBAAiB,EAAE,OAAOt4C,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC,EAAE,EAAE,EAAEs4C,EAAE,2BAA2B,SAAS,EAAEt4C,EAAE,MAAM,MAAM,EAAE,GAAG,CAACU,EAAEC,CAAC,EAAE23C,EAAE,0BAA0Bt4C,EAAE,MAAM,CAAC,EAAEY,EAAEgkB,EAAE,cAAclkB,CAAC,EAAEG,EAAE+jB,EAAE,oBAAoBhkB,EAAE,OAAO,EAAEE,EAAE8jB,EAAE,cAAcjkB,CAAC,EAAEqB,EAAEhE,EAAE,KAAK,IAAIgC,EAAE,MAAM,EAAE,OAAO,QAAQ,EAAE,EAAE,EAAEa,EAAE,OAAO,EAAE,EAAE,CAAC,IAAIwE,EAAE,EAAEvE,EAAE,EAAEkB,EAAEqD,GAAG5C,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE5B,EAAE,EAAE4B,EAAE,CAAC,IAAIC,EAAEX,EAAEqD,EAAE3C,GAAGC,EAAE,IAAI,EAAEA,EAAEF,EAAEC,EAAE,CAAC7B,EAAE,GAAG4B,CAAC,CAAC,OAAOxC,EAAE,QAAQ,GAAGjC,EAAE,8BAA8B,CAAC,CAAC,EAAEA,EAAE,eAAe0C,EAAE,QAAQG,CAAC,CAAC,CAAC,IAAI+4E,GAAG,CAAC,WAAWrgE,GAAG,YAAY,MAAM,WAAWogE,EAAE,EAAE,SAASE,GAAG/7E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,QAAQ,EAAE,IAAI,EAAEumB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE0B,EAAEu4C,EAAE,mBAAmB,EAAEj6C,EAAE,MAAM,MAAM,EAAE2B,EAAE3B,EAAE4B,EAAE,CAAC,EAAEF,GAAG,OAAOC,EAAE81E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK+B,CAAC,CAAC,CAAC,EAAEE,EAAE,KAAKD,CAAC,EAAE,EAAEs4C,EAAE,iBAAiB,EAAE,OAAOt4C,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC,EAAE,EAAE,EAAEs4C,EAAE,2BAA2B,SAAS,EAAEt4C,EAAE,MAAM,MAAM,EAAE,GAAG,CAACU,EAAEC,CAAC,EAAE23C,EAAE,0BAA0Bt4C,EAAE,MAAM,CAAC,EAAEY,EAAEgkB,EAAE,cAAclkB,CAAC,EAAEG,EAAE+jB,EAAE,oBAAoBhkB,EAAE,OAAO,EAAEE,EAAE8jB,EAAE,cAAcjkB,CAAC,EAAEqB,EAAEhE,EAAE,KAAK,IAAIgC,EAAE,MAAM,EAAE,OAAO,QAAQ,EAAE,EAAE,EAAEa,EAAE,OAAO,EAAE,EAAE,CAAC,IAAIwE,EAAE,EAAEvE,EAAE,EAAEkB,EAAEqD,GAAG5C,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE5B,EAAE,EAAE4B,EAAE,CAAC,IAAIC,EAAEX,EAAEqD,EAAE3C,GAAGC,EAAE,IAAI,EAAEA,EAAEF,EAAEC,EAAE,CAAC7B,EAAE,GAAG4B,CAAC,CAAC,OAAOxC,EAAE,QAAQ,GAAGjC,EAAE,8BAA8B,CAAC,CAAC,EAAEA,EAAE,eAAe0C,EAAE,QAAQG,CAAC,CAAC,CAAC,IAAIi5E,GAAG,CAAC,WAAWtgE,GAAG,YAAY,MAAM,WAAWqgE,EAAE,EAAME,GAAGlG,GAAGp6D,GAAG3b,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEk8E,GAAG,CAAC,WAAWvgE,GAAG,YAAY,MAAM,WAAWsgE,EAAE,EAAME,GAAGpG,GAAGn6D,GAAG5b,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAEo8E,GAAG,CAAC,WAAWxgE,GAAG,YAAY,MAAM,WAAWugE,EAAE,EAAME,GAAGtG,GAAGl6D,GAAG7b,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEs8E,GAAG,CAAC,WAAWzgE,GAAG,YAAY,MAAM,WAAWwgE,EAAE,EAAME,GAAGxH,GAAG,CAAC/0E,EAAE,IAAI,KAAK,MAAMA,EAAE,CAAC,CAAC,EAAEw8E,GAAG/G,GAAG15D,GAAGwgE,EAAE,EAAEE,GAAG,CAAC,WAAW1gE,GAAG,YAAY,MAAM,WAAWygE,EAAE,EAAME,GAAG3G,GAAGj6D,GAAG9b,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAE28E,GAAG,CAAC,WAAW7gE,GAAG,YAAY,MAAM,WAAW4gE,EAAE,EAAE,SAASE,GAAG58E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEA,EAAE,aAAa0B,EAAE1B,EAAE,YAAY2B,EAAE3B,EAAE,eAAe4B,EAAE5B,EAAE,cAAcqC,EAAErC,EAAE,sBAAsBsC,EAAEtC,EAAE,qBAAqBuC,EAAEvC,EAAE,QAAQ,IAAIwC,EAAExC,EAAE,QAAQ,KAAKyC,EAAE,IAAI,MAAM,OAAO,kBAAkB,OAAO,kBAAkBkB,EAAEqsB,GAAGhwB,EAAE,SAASL,CAAC,EAAE,EAAEgE,EAAE,OAAOqD,EAAEhH,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAG,EAAEA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGoE,EAAEpE,EAAE,SAAS,GAAG,QAAQqE,EAAE,EAAEA,EAAErE,EAAE,UAAU,EAAEqE,EAAE,CAAC,IAAIC,EAAED,EAAE2C,EAAEzC,EAAEF,EAAEtE,EAAE,GAAG,QAAQkH,EAAE,EAAEA,EAAEjH,EAAE,WAAW,EAAEiH,EAAE,QAAQxC,EAAE,EAAEA,EAAEzE,EAAE,UAAU,EAAEyE,EAAE,CAAC,IAAIyC,EAAEzC,EAAE,EAAElC,EAAEmC,EAAE,KAAK,IAAI,EAAEwC,CAAC,EAAEvC,EAAE,KAAK,IAAI3E,EAAE,SAASqC,EAAE6E,CAAC,EAAEtC,EAAEN,EAAEG,EAAE,EAAE,QAAQI,EAAE,EAAEA,EAAE7E,EAAE,SAAS,EAAE6E,EAAE,CAAC,IAAIC,EAAED,EAAEnD,EAAEc,EAAEuC,EAAE,KAAK,IAAI,EAAED,CAAC,EAAE4C,EAAE,KAAK,IAAI1H,EAAE,QAAQsC,EAAEwC,CAAC,EAAEsC,EAAE3E,EAAEkF,EAAE,EAAEC,EAAG,EAAE,QAAQE,GAAGpD,EAAEoD,GAAGnD,EAAEmD,IAAInG,EAAE,CAAC,IAAIoG,GAAGxD,EAAEuD,GAAG/H,EAAE,GAAG,QAAQiI,GAAGjD,EAAEiD,GAAGN,EAAEM,IAAIpG,EAAE,CAAC,IAAIwE,GAAG2B,GAAGC,GAAGjI,EAAE,GAAGmI,GAAGzI,EAAE2G,GAAGa,GAAG,IAAI,OAAOiB,GAAGd,EAAEA,EAAEc,GAAG,IAAI,QAAQP,GAAGO,GAAGN,IAAK,CAAC,GAAG,MAAMR,CAAC,EAAE,KAAK,CAAC,IAAIS,EAAGjD,EAAEC,EAAET,EAAE6C,EAAE,EAAEY,GAAI,IAAI,MAAMF,EAAEC,EAAGR,CAAC,CAAC,CAAC,CAAC,OAAOzD,CAAC,CAAC,SAAS24E,GAAG78E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,GAAG,EAAE,GAAG,CAAC,IAAI,EAAEgwB,GAAGjwB,EAAE,SAAS,OAAO,EAAE2B,EAAE3B,EAAE,aAAa4B,EAAE5B,EAAE,YAAY6B,EAAE7B,EAAE,eAAesC,EAAEtC,EAAE,cAAcuC,EAAEvC,EAAE,sBAAsBwC,EAAExC,EAAE,qBAAqByC,EAAEzC,EAAE,QAAQ,IAAI0C,EAAE1C,EAAE,QAAQ,KAAK4D,EAAEqsB,GAAG,EAAErwB,EAAEF,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAEM,EAAE,UAAU,EAAE,EAAE,QAAQiH,EAAE,EAAEA,EAAEjH,EAAE,WAAW,EAAEiH,EAAE,QAAQ,EAAE,EAAE,EAAEjH,EAAE,UAAU,EAAE,EAAE,CAAC,IAAIqE,EAAE,EAAE1C,EAAEc,EAAE6B,EAAED,EAAE,KAAKC,EAAE,GAAGA,GAAGzC,EAAE,IAAI0C,EAAE,KAAK,IAAIvE,EAAE,SAASuC,EAAE8B,CAAC,EAAE,QAAQG,EAAE,EAAEA,EAAExE,EAAE,SAAS,EAAEwE,EAAE,CAAC,IAAI0C,EAAE1C,EAAE5C,EAAEc,EAAEgC,EAAEwC,EAAE,KAAKxC,EAAE,GAAGA,GAAGpC,EAAE,IAAI6E,EAAE,KAAK,IAAInH,EAAE,QAAQwC,EAAE0E,CAAC,EAAEvC,EAAE,OAAO,kBAAkBC,EAAE,GAAG,QAAQC,EAAEP,EAAEO,EAAEN,EAAEM,GAAGhD,EAAE,CAAC,IAAIiD,EAAED,EAAER,EAAE,QAAQU,EAAEL,EAAEK,EAAEoC,EAAEpC,GAAGzC,EAAE,CAAC,IAAI0C,EAAED,EAAEmC,EAAES,EAAE/D,EAAE,IAAI,EAAEiB,EAAEE,EAAEkC,CAAC,EAAEU,EAAEhD,IAAIA,EAAEgD,EAAE1H,EAAE2E,EAAE,IAAI,EAAE5E,EAAE,SAAS6E,GAAG7E,EAAE,QAAQ+E,GAAG/E,EAAE,WAAWiH,GAAGpC,EAAE7E,EAAE,QAAQ+E,GAAG/E,EAAE,WAAWiH,EAAErC,EAAEE,EAAEtC,EAAEwC,EAAE,CAAC,CAAC,EAAE,IAAIJ,EAAE,EAAE,EAAEJ,EAAEyC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,SAASu1E,GAAG98E,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEA,EAAE,YAAY0B,EAAE1B,EAAE,aAAa2B,EAAE3B,EAAE,YAAY4B,EAAE5B,EAAE,cAAcqC,EAAErC,EAAE,eAAesC,EAAEtC,EAAE,cAAcuC,EAAEvC,EAAE,qBAAqBwC,EAAExC,EAAE,sBAAsByC,EAAEzC,EAAE,qBAAqB2D,EAAE3D,EAAE,QAAQ,MAAM,EAAEA,EAAE,QAAQ,IAAIgH,EAAEhH,EAAE,QAAQ,KAAK,EAAE,IAAI,MAAM,OAAO,kBAAkB,OAAO,kBAAkBoE,EAAE4rB,GAAGhwB,EAAE,SAASL,CAAC,EAAE0E,EAAED,EAAE,OAAOE,EAAEtE,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGuE,EAAEvE,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGiH,EAAEjH,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGyE,EAAEzE,EAAE,SAAS,GAAG,QAAQkH,EAAE,EAAEA,EAAElH,EAAE,UAAU,EAAEkH,EAAE,CAAC,IAAIxC,EAAEwC,EAAE5C,EAAEK,EAAEuC,EAAEnH,EAAE,GAAG,QAAQ6E,EAAE,EAAEA,EAAE5E,EAAE,WAAW,EAAE4E,EAAE,QAAQC,EAAE,EAAEA,EAAE7E,EAAE,SAAS,EAAE6E,EAAE,CAAC,IAAIC,EAAED,EAAE,EAAElB,EAAEoB,EAAED,EAAE,KAAKC,EAAE,GAAGA,GAAGnD,EAAE,IAAI8F,EAAE,KAAK,IAAI1H,EAAE,QAAQuC,EAAEuC,CAAC,EAAEsC,EAAE1C,EAAEG,EAAEN,EAAE,QAAQoD,EAAE,EAAEA,EAAE3H,EAAE,UAAU,EAAE2H,EAAE,CAAC,IAAIC,EAAGD,EAAEjG,EAAE,EAAEmG,EAAGD,EAAG,KAAKC,EAAG,GAAGA,GAAIxF,EAAE,IAAIyF,GAAG,KAAK,IAAI9H,EAAE,SAASwC,EAAEoF,CAAE,EAAEG,GAAGX,EAAEO,EAAEV,EAAE,QAAQe,GAAG,EAAEA,GAAGhI,EAAE,SAAS,EAAEgI,GAAG,CAAC,IAAI5B,GAAG4B,GAAGrG,EAAEqF,EAAEkB,GAAG9B,GAAG,KAAK8B,GAAG,GAAGA,IAAI5F,EAAE,IAAI6F,GAAG,KAAK,IAAInI,EAAE,QAAQyC,EAAE2D,EAAE,EAAEgC,GAAGL,GAAGC,GAAGvD,EAAE4D,GAAG,EAAEC,GAAG,EAAEC,GAAG,EAAE,QAAQM,GAAG9D,EAAE8D,GAAGnB,EAAEmB,IAAIjH,EAAE,CAAC,IAAIkH,GAAGnE,EAAEkE,GAAG9I,EAAE,GAAG,QAAQsG,GAAGwB,EAAGxB,GAAGyB,GAAGzB,IAAIhE,EAAE,CAAC,IAAIkE,GAAGuC,GAAGzC,GAAGtG,EAAE,GAAG,QAAQyG,GAAG0B,GAAG1B,GAAG2B,GAAG3B,IAAIlE,EAAE,CAAC,IAAImE,GAAGF,GAAGC,GAAGzG,EAAE,GAAGgJ,GAAGtJ,EAAEgH,GAAG7B,GAAG,GAAG,IAAI,OAAOmE,GAAGV,GAAGA,GAAGU,GAAG,IAAI,QAAQT,IAAIS,GAAGR,MAAM,MAAMF,EAAE,EAAE,KAAK,CAAC,GAAG,MAAMA,EAAE,EAAE,KAAK,CAAC,GAAG,MAAMA,EAAE,EAAE,KAAK,CAAC,IAAIO,GAAGR,GAAGxD,EAAEP,EAAEuE,IAAI,IAAI,MAAMN,GAAGC,GAAGF,EAAE,CAAC,CAAC,CAAC,CAAC,OAAOjE,CAAC,CAAC,SAASo4E,GAAG/8E,EAAE,EAAE,CAAC,IAAIE,EAAEqwB,GAAG,EAAE,SAAS,OAAO,EAAEjwB,EAAE,EAAE,YAAYC,EAAE,EAAE,aAAa,EAAE,EAAE,YAAY,EAAE,EAAE,cAAc0B,EAAE,EAAE,eAAeC,EAAE,EAAE,cAAcC,EAAE,EAAE,qBAAqBS,EAAE,EAAE,sBAAsBC,EAAE,EAAE,qBAAqBC,EAAE,EAAE,QAAQ,MAAMC,EAAE,EAAE,QAAQ,IAAIC,EAAE,EAAE,QAAQ,KAAK,QAAQkB,EAAE,EAAEA,EAAE,EAAE,UAAU,EAAEA,EAAE,QAAQ,EAAE,EAAE,EAAE,EAAE,WAAW,EAAE,EAAE,QAAQqD,EAAE,EAAEA,EAAE,EAAE,SAAS,EAAEA,EAAE,CAAC,IAAI,EAAEA,EAAEjH,EAAEwC,EAAE6B,EAAE,EAAE,KAAKA,EAAE,GAAGA,GAAG,EAAE,IAAIC,EAAE,KAAK,IAAI,EAAE,QAAQzC,EAAE,CAAC,EAAE,QAAQ0C,EAAE,EAAEA,EAAE,EAAE,UAAU,EAAEA,EAAE,CAAC,IAAIC,EAAED,EAAEtE,EAAEwC,EAAEyE,EAAE1C,EAAE,KAAK0C,EAAE,GAAGA,GAAGvF,EAAE,IAAI+C,EAAE,KAAK,IAAI,EAAE,SAASpC,EAAEkC,CAAC,EAAE,QAAQ2C,EAAE,EAAEA,EAAE,EAAE,SAAS,EAAEA,EAAE,CAAC,IAAIxC,EAAEwC,EAAE,EAAEzE,EAAEkC,EAAED,EAAE,KAAKC,EAAE,GAAGA,GAAGhD,EAAE,IAAIiD,EAAE,KAAK,IAAI,EAAE,QAAQtC,EAAEoC,CAAC,EAAEG,EAAE,OAAO,kBAAkBC,EAAE,GAAG,QAAQC,EAAEX,EAAEW,EAAEV,EAAEU,GAAG,EAAE,CAAC,IAAI2C,EAAE3C,EAAE,EAAE,QAAQqC,EAAEH,EAAEG,EAAE3C,EAAE2C,GAAG1F,EAAE,CAAC,IAAIiG,EAAEP,EAAE7C,EAAE,QAAQqD,EAAGjD,EAAEiD,EAAGhD,EAAEgD,GAAIjG,EAAE,CAAC,IAAIkG,EAAGD,EAAGlD,EAAEoD,GAAGrI,EAAE,IAAIkE,EAAEoB,EAAEqC,EAAEQ,EAAG,CAAC,EAAEE,IAAIjD,IAAIA,EAAEiD,GAAGhD,EAAE4C,EAAErF,EAAEC,EAAEqF,EAAEtF,EAAEwF,EAAG,CAAC,CAAC,CAAClI,EAAE,IAAImF,EAAEnB,EAAEqD,EAAE1C,EAAE4C,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOvH,CAAC,CAAC,SAAS88E,GAAGh9E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAEkxE,GAAGlxE,EAAE,SAAS,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,CAAC,EAAE5B,EAAE6B,EAAE,EAAE2kB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAEr4C,CAAC,EAAE,IAAI,wEAAwE,oBAAoBA,IAAI,EAAE,IAAIS,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE4B,EAAEF,EAAEC,CAAC,EAAEW,EAAE,GAAGD,EAAE,cAAc,GAAGA,EAAE,eAAe,GAAGkkB,EAAE,YAAYlkB,EAAE,QAAQA,EAAE,QAAQ,EAAEC,EAAEsyE,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,MAAM,CAAC,IAAI4C,EAAE5C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOwC,EAAE+jB,EAAE,eAAevmB,EAAE,KAAK,EAAEyC,EAAE45E,GAAG95E,EAAEvC,EAAE,MAAMA,EAAE,MAAMwC,EAAEH,EAAE,KAAK,EAAEC,EAAE3C,EAAE,eAAe0C,EAAE,SAASrC,EAAE,MAAMyC,EAAE,MAAM,CAAC,CAAC,OAAOH,CAAC,CAAC,IAAIo6E,GAAG,CAAC,WAAWjhE,GAAG,YAAY,MAAM,WAAWghE,EAAE,EAAE,SAASE,GAAGl9E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,EAAE,WAAWC,CAAC,EAAE7B,EAAEmxE,GAAGlxE,EAAE,WAAW,EAAE,IAAIqC,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,EAAEU,EAAE3C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuC,EAAEg6E,GAAGj6E,EAAEtC,EAAE,MAAMA,EAAE,MAAMumB,EAAE,eAAevmB,EAAE,KAAK,EAAEqC,EAAE,KAAK,EAAE,OAAO1C,EAAE,eAAe4C,EAAE,MAAM,UAAUA,EAAE,MAAM,CAAC,CAAC,IAAIq6E,GAAG,CAAC,WAAWjhE,GAAG,YAAY,MAAM,WAAWghE,EAAE,EAAE,SAASE,GAAGp9E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,gBAAgBC,CAAC,EAAE7B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,eAAe,EAAE,IAAIqC,EAAE43C,EAAE,kBAAkB,EAAE,MAAM,EAAEv4C,EAAE,EAAEC,EAAEC,CAAC,EAAEU,EAAED,EAAE,YAAYE,EAAEF,EAAE,aAAaG,EAAEH,EAAE,YAAYI,EAAEJ,EAAE,YAAYsB,EAAEtB,EAAE,aAAa,EAAEA,EAAE,YAAY2E,EAAE3E,EAAE,cAAc,EAAEA,EAAE,eAAe+B,EAAE/B,EAAE,cAAcgC,EAAEhC,EAAE,qBAAqBiC,EAAEjC,EAAE,sBAAsBkC,EAAElC,EAAE,qBAAqB4E,EAAE5C,EAAE,EAAEhC,EAAE,QAAQ,MAAMoC,EAAEF,EAAE,EAAElC,EAAE,QAAQ,KAAK6E,EAAE5C,EAAE,EAAEjC,EAAE,QAAQ,IAAIqC,EAAEsrB,GAAG,EAAE,MAAM,SAAS,EAAErrB,EAAE,GAAGlC,EAAEkB,EAAE,GAAGiB,EAAEjF,EAAE,WAAWK,CAAC,EAAE,QAAQ6E,EAAE,EAAEA,EAAExC,EAAE,UAAU,EAAEwC,EAAE,QAAQC,EAAE,EAAEA,EAAEzC,EAAE,WAAW,EAAEyC,EAAE,QAAQC,EAAE,EAAEA,EAAE1C,EAAE,QAAQ,EAAE0C,EAAE,QAAQ2C,EAAE,EAAEA,EAAErF,EAAE,SAAS,EAAEqF,EAAE,QAAQN,EAAE,EAAEA,EAAE/E,EAAE,QAAQ,EAAE+E,EAAE,CAAC,IAAIO,EAAE5C,EAAEkC,EAAEW,EAAGF,EAAER,EAAEW,EAAGT,EAAE3C,EAAEqD,GAAG,EAAE,QAAQC,GAAG,EAAEA,GAAG1D,EAAE0D,IAAIf,EAAE,CAAC,IAAIgB,IAAIL,EAAEI,IAAIzF,EAAE,GAAG,EAAE0F,GAAG,GAAGA,IAAI3F,EAAE,UAAU,KAAK,MAAM2F,EAAE,IAAIA,IAAI,QAAQ5B,GAAG,EAAEA,GAAG9B,EAAE8B,IAAI,EAAE,CAAC,IAAI8B,IAAIN,EAAGxB,IAAI7D,EAAE,GAAG,EAAE2F,GAAG,GAAGA,IAAI7F,EAAE,WAAW,KAAK,MAAM6F,EAAE,IAAIA,IAAI,QAAQC,GAAG,EAAEA,GAAG5D,EAAE4D,IAAI/D,EAAE,CAAC,IAAIgE,IAAIP,EAAGM,IAAI3F,EAAK4F,GAAG,GAAGA,IAAI/F,EAAE,UAAU,KAAK,MAAM+F,EAAE,IAAIA,KAAYN,IAAIlD,EAAE,IAAIC,EAAEmD,GAAGE,GAAGE,GAAGtD,CAAC,EAAC,CAAC,CAAC,CAACJ,EAAE,IAAIoD,GAAGnD,EAAEE,EAAEE,EAAE2C,EAAEN,EAAEtC,CAAC,CAAC,CAAC,OAAOnF,EAAE,eAAe+E,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIo4E,GAAG,CAAC,WAAWlhE,GAAG,YAAY,MAAM,WAAWihE,EAAE,EAAE,SAASE,GAAGt9E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,EAAE,EAAEkxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,aAAa,EAAE,GAAG,CAAC,WAAW0B,EAAE,QAAQC,EAAE,IAAIC,CAAC,EAAE7B,EAAEsC,EAAE43C,EAAE,kBAAkB,EAAE,MAAMv4C,EAAEC,EAAE,EAAEC,CAAC,EAAEU,EAAED,EAAE,aAAaE,EAAEF,EAAE,YAAYG,EAAEH,EAAE,aAAaI,EAAEJ,EAAE,YAAYsB,EAAEtB,EAAE,eAAe,EAAEA,EAAE,cAAc2E,EAAE3E,EAAE,sBAAsB,EAAEA,EAAE,qBAAqB+B,EAAE,EAAE,EAAE/B,EAAE,QAAQ,KAAKgC,EAAE2C,EAAE,EAAE3E,EAAE,QAAQ,IAAIiC,EAAE0rB,GAAG,EAAE,MAAM,SAAS,EAAEzrB,EAAE,GAAG/B,EAAEC,GAAGwE,EAAEtH,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOyE,EAAEurB,GAAGhwB,EAAE,MAAM,UAAUiH,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE7E,EAAE,UAAU,EAAE6E,EAAE,QAAQxC,EAAE,EAAEA,EAAErC,EAAE,WAAW,EAAEqC,EAAE,QAAQC,EAAE,EAAEA,EAAEtC,EAAE,SAAS,EAAEsC,EAAE,QAAQC,EAAE,EAAEA,EAAEvC,EAAE,QAAQ,EAAEuC,EAAE,CAAC,IAAIC,EAAEF,EAAEN,EAAES,EAAEF,EAAER,EAAEW,EAAE,EAAE,QAAQ2C,EAAE,EAAEA,EAAEV,EAAEU,GAAG/D,EAAE,CAAC,IAAIyD,GAAGvC,EAAE6C,GAAGpF,EAAE,GAAG,EAAE8E,EAAE,GAAGA,GAAG/E,EAAE,WAAW,KAAK,MAAM+E,CAAC,IAAIA,GAAG,QAAQO,EAAE,EAAEA,EAAE,EAAEA,GAAG,EAAE,CAAC,IAAIC,GAAI9C,EAAE6C,GAAGpF,EAAKqF,EAAG,GAAGA,GAAIvF,EAAE,UAAU,KAAK,MAAMuF,CAAE,IAAIA,IAAY7C,GAAGN,EAAE,IAAIyC,EAAEE,EAAEQ,EAAGlD,CAAC,EAAC,CAAC,CAACJ,EAAE,IAAIS,EAAER,EAAE2C,EAAEvC,EAAEC,EAAEF,CAAC,CAAC,CAAC,OAAO/E,EAAE,eAAe2E,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI04E,GAAG,CAAC,WAAWthE,GAAG,YAAY,MAAM,WAAWqhE,EAAE,EAAE,SAASE,GAAGx9E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK0B,EAAE,SAASC,CAAC,EAAE,EAAE4kB,EAAE,OAAO7kB,EAAE,MAAM,SAASC,EAAE,MAAM,OAAO,IAAI,8EAA8E,EAAE4kB,EAAE,OAAO,GAAG,MAAM7kB,EAAE,MAAM,SAAS,EAAE,MAAM,OAAO,IAAI,4EAA4E,EAAE6kB,EAAE,OAAO,GAAG,MAAM7kB,EAAE,MAAM,SAAS,EAAE,MAAM,OAAO,IAAI,2EAA2E,EAAEwvE,GAAG,CAAClxE,EAAE0B,EAAEC,EAAE,EAAE,CAAC,EAAE,WAAW,EAAE,GAAG,CAAC,gBAAgBC,CAAC,EAAE7B,EAAE6B,GAAG,OAAOA,EAAE,MAAM,IAAIS,EAAE1C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOsC,EAAE3C,EAAE,KAAK,IAAI+B,EAAE,MAAM,EAAE,OAAOa,EAAE5C,EAAE,KAAK,IAAIgC,EAAE,MAAM,EAAE,OAAOa,EAAE,EAAE7C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,IAAI,aAAa,CAAC,CAAC,CAAC,EAAE8C,EAAE,EAAE9C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,IAAI,aAAa,CAAC,CAAC,CAAC,EAAEgE,EAAE,IAAI,aAAatB,EAAE,MAAM,EAAE,EAAEI,EAAE,OAAOuE,EAAExE,EAAE,OAAO,EAAED,EAAE,OAAO6B,EAAE9B,EAAE,OAAO+B,EAAE,EAAEC,EAAE,EAAEC,EAAE,EAAE0C,EAAE,EAAE,QAAQxC,EAAE,EAAEA,EAAEpC,EAAE,OAAO,EAAEoC,EAAEd,EAAEc,GAAGhC,EAAE4B,MAAMhC,EAAEoC,GAAGnC,EAAEgC,MAAM9B,EAAE+B,KAAK,KAAK,KAAKhC,EAAE0E,KAAKrF,CAAC,EAAEyC,GAAG,IAAIA,EAAE,GAAGC,GAAGF,IAAIE,EAAE,GAAGC,GAAGyC,IAAIzC,EAAE,GAAG0C,GAAG,IAAIA,EAAE,GAAG,OAAOtH,EAAE,eAAeK,EAAE,MAAMA,EAAE,MAAM2D,CAAC,CAAC,CAAC,IAAIu5E,GAAG,CAAC,WAAWx+D,GAAG,YAAY,MAAM,WAAWu+D,EAAE,EAAE,SAASE,GAAG19E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,MAAM,CAAC,EAAED,EAAEmxE,GAAG,CAAClxE,CAAC,EAAE,gBAAgB,EAAE,IAAI0B,EAAE,EAAE,OAAO,CAACsF,EAAE,IAAIA,EAAE,CAAC,EAAErF,EAAEs4C,EAAE,YAAYj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEE,EAAEq4C,EAAE,YAAYt4C,EAAE,OAAO,EAAE,MAAM,EAAEU,EAAE43C,EAAE,oBAAoBj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEY,EAAE23C,EAAE,oBAAoB,EAAE,EAAE,MAAM,EAAE13C,EAAE03C,EAAE,aAAa53C,EAAE,EAAE,EAAE,MAAM,EAAEG,EAAE83E,GAAG,CAAC,OAAO,CAAC,EAAEt6E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEc,EAAEg1E,GAAG,CAAC,OAAO,CAAC,EAAEj1E,CAAC,EAAE,QAAQ7C,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAE+B,EAAE22E,GAAG,CAAC,OAAO,CAAC,EAAE73E,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAE,EAAEw2E,GAAG,CAAC,OAAO,CAAC,EAAEl1E,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM2C,EAAE,KAAKC,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAE9C,EAAE,8BAA8BgE,CAAC,EAAE,CAAC,CAAC,IAAIy5E,GAAG,CAAC,WAAWthE,GAAG,YAAY,MAAM,WAAWqhE,EAAE,EAAE,SAASE,GAAG59E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE2B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOiC,EAAE2vE,GAAG7vE,EAAEC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,OAAOhC,EAAE,eAAe,CAAC,CAAC,EAAE,EAAE,MAAMiC,CAAC,CAAC,CAAC,IAAI07E,GAAG,CAAC,WAAWvhE,GAAG,YAAY,MAAM,WAAWshE,EAAE,EAAE,SAASE,GAAG99E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,GAAGM,EAAE,GAAGC,CAAC,EAAE,EAAE,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,EAAEJ,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO0B,EAAEu4C,EAAE,2BAA2B,MAAM,KAAK,CAAC,EAAE,MAAM,KAAK,CAAC,CAAC,EAAE,OAAOt6C,EAAE,eAAe,CAAC+B,EAAE,MAAM,EAAE,QAAQ,WAAW,KAAKA,CAAC,CAAC,CAAC,CAAC,IAAI87E,GAAG,CAAC,WAAWvhE,GAAG,YAAY,MAAM,WAAWshE,EAAE,EAAME,GAAGjI,GAAGp5D,GAAG,CAAC3c,EAAE,IAAI,CAAC,IAAIE,EAAE,EAAE,OAAOF,EAAEE,EAAE,aAAaA,EAAE,aAAaF,EAAEE,EAAE,aAAaA,EAAE,aAAaF,CAAC,CAAC,EAAEi+E,GAAG,CAAC,WAAWthE,GAAG,YAAY,MAAM,WAAWqhE,EAAE,EAAME,GAAGl+E,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,EAAEA,EAAE,OAAOE,EAAEF,EAAE,QAAQM,EAAE,IAAI,aAAawmB,EAAE,cAAc,EAAE,KAAK,CAAC,EAAEvmB,EAAEL,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,EAAEK,EAAE,mBAAmB,KAAK,EAAEA,EAAE,mBAAmB,KAAK0B,EAAE/B,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOgC,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,QAAQiC,EAAE,EAAEA,EAAEF,EAAE,OAAOE,IAAI,CAAC,IAAIS,EAAEX,EAAEE,GAAGU,EAAEX,EAAEC,GAAG7B,EAAE6B,GAAG,KAAK,MAAMS,EAAEC,CAAC,CAAC,CAAC,OAAO3C,EAAE,WAAWI,EAAE,EAAE,MAAM,SAAS,CAAC,EAAE69E,GAAG,CAAC,WAAWthE,GAAG,YAAY,MAAM,WAAWqhE,EAAE,EAAE,SAASE,GAAGp+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,mBAAmB,KAAK,EAAEJ,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,OAAOL,EAAE,eAAeK,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,CAAC,IAAI89E,GAAG,CAAC,WAAW7+D,GAAG,YAAY,MAAM,WAAW4+D,EAAE,EAAE,SAASE,GAAGt+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAevmB,EAAE,EAAE,GAAG,KAAK,EAAE,GAAG,EAAE,EAAE,IAAI2D,GAAGA,EAAE,KAAK,EAAEs2C,EAAE,uBAAuB,EAAE,CAAC,EAAE,IAAIv4C,EAAEu4C,EAAE,gBAAgB,EAAE,IAAIt2C,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,GAAG4iB,EAAE,cAAc7kB,CAAC,IAAI,EAAE,OAAO/B,EAAE,eAAe+B,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,EAAE,IAAIC,EAAE,EAAE,OAAOgC,GAAG4iB,EAAE,cAAc5iB,EAAE,KAAK,EAAE,CAAC,EAAE,GAAGhC,EAAE,SAAS,EAAE,OAAOizE,GAAG,CAAC,OAAO,CAAC,EAAEjzE,EAAE,EAAE,EAAE,QAAQhC,CAAC,CAAC,EAAE,GAAGgC,EAAE,GAAG,QAAQ,YAAY,CAAC,IAAIgC,EAAEhC,EAAE,IAAI0C,GAAGywE,GAAG,CAAC,OAAO,CAAC,MAAMzwE,CAAC,EAAE,QAAQ1E,CAAC,CAAC,CAAC,EAAE,EAAEgC,EAAE,IAAI0C,GAAGw5E,GAAG,CAAC,OAAO,CAAC,MAAMx5E,CAAC,EAAE,QAAQ1E,CAAC,CAAC,CAAC,EAAEqH,EAAE+2E,GAAG,CAAC,OAAOp6E,EAAE,QAAQhE,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,EAAEo+E,GAAG,CAAC,OAAO,EAAE,QAAQp+E,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAEyE,EAAEqwE,GAAG,CAAC,OAAO,CAAC,KAAKztE,EAAE,KAAK,CAAC,EAAE,QAAQrH,CAAC,CAAC,EAAE,OAAOgE,EAAE,QAAQU,GAAG1E,EAAE,8BAA8B0E,CAAC,CAAC,EAAE,EAAE,QAAQA,GAAG1E,EAAE,8BAA8B0E,CAAC,CAAC,EAAE1E,EAAE,8BAA8BqH,CAAC,EAAErH,EAAE,8BAA8B,CAAC,EAAEyE,CAAC,CAAC,IAAIxC,EAAED,EAAE,IAAIgC,GAAG,CAAC,IAAI,EAAE4iB,EAAE,cAAc5iB,EAAE,MAAM,MAAM,CAAC,CAAC,EAAE,OAAO22E,GAAG,CAAC,OAAO,CAAC,EAAE32E,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE0C,EAAET,EAAE,IAAI+B,IAAI,CAAC,KAAKhE,EAAE,KAAK,IAAIgE,EAAE,MAAM,EAAE,OAAO,MAAMA,EAAE,KAAK,EAAE,EAAEjC,EAAEu4C,EAAE,gBAAgBr4C,EAAE,IAAI+B,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,IAAIrB,EAAEV,EAAE,GAAG,MAAM,KAAK,EAAEW,EAAEovE,GAAGtvE,EAAEX,EAAE,EAAE,GAAG,MAAMY,CAAC,EAAEE,EAAEy3C,EAAE,gBAAgBt4C,EAAE,IAAIgC,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAElB,EAAE9C,EAAE,eAAe6C,EAAE,EAAE,GAAG,MAAMD,CAAC,EAAE,OAAOX,EAAE,QAAQ+B,GAAGhE,EAAE,8BAA8BgE,CAAC,CAAC,EAAElB,CAAC,CAAC,IAAIu7E,GAAG,CAAC,WAAWzhE,GAAG,YAAY,MAAM,WAAWwhE,EAAE,EAAE,SAASE,GAAGx+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,EAAE,UAAUC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,QAAQ,EAAE,IAAIsC,EAAE23C,EAAE,wBAAwBt4C,CAAC,EAAEY,EAAE03C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAE4B,EAAEF,EAAEW,EAAE,GAAGC,CAAC,EAAEE,EAAED,EAAE,aAAaE,EAAEF,EAAE,YAAYoB,EAAEpB,EAAE,eAAe,EAAEA,EAAE,cAAcyE,EAAEzE,EAAE,QAAQ,KAAK,EAAEA,EAAE,QAAQ,IAAI6B,EAAE7B,EAAE,aAAa,eAAe8B,EAAE,IAAI0kB,GAAGxmB,EAAE,SAASvC,EAAE,KAAK,EAAEsE,EAAEiiB,EAAE,eAAevmB,EAAE,KAAK,EAAEuE,EAAEgiB,EAAE,eAAe,EAAE,KAAK,EAAEtf,EAAE3C,EAAE,GAAGG,EAAEL,EAAEE,EAAE,GAAGA,EAAE,GAAG4C,EAAE9C,EAAEE,EAAE,GAAG,EAAEI,EAAEN,EAAE,EAAEE,EAAE,GAAGK,EAAEN,EAAE,QAAQ,GAAGO,EAAER,EAAEC,EAAE,QAAQ,GAAGA,EAAE,QAAQ,GAAGQ,EAAET,EAAEC,EAAE,QAAQ,GAAG,EAAES,EAAEV,EAAE,EAAEC,EAAE,QAAQ,GAAGU,EAAEpF,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO0H,EAAE/H,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOyH,EAAE/C,EAAE,OAAO,QAAQsD,EAAE,EAAEA,EAAEpF,EAAE,UAAU,EAAEoF,EAAE,CAAC,IAAIC,EAAGD,EAAEV,EAAEY,EAAGF,EAAEhD,EAAE,QAAQmD,GAAG,EAAEA,GAAGvF,EAAE,UAAU,EAAEuF,GAAG,CAAC,IAAIC,GAAGF,EAAGC,GAAGlD,EAAEoD,GAAGF,GAAGvF,EAAE,aAAa,EAAE,QAAQ6D,GAAG,EAAEA,GAAG5D,EAAE,EAAE4D,GAAG,CAAC,IAAI8B,GAAGF,GAAG5B,GAAGzC,EAAE,GAAGuE,GAAG,GAAGA,IAAI3F,EAAE,SAAS,SAAS,IAAI4F,GAAG/B,GAAG7B,EAAE,GAAG6D,GAAGR,EAAGM,GAAGzD,EAAE,QAAQ4D,GAAG,EAAEA,GAAG9F,EAAE,SAAS,EAAE8F,GAAG,CAAC,IAAIC,GAAGP,GAAGM,GAAGxD,EAAE0D,GAAGF,GAAG9F,EAAE,YAAYyE,EAAE,QAAQ4B,GAAG,EAAEA,GAAGnG,EAAE,EAAEmG,GAAG,CAAC,IAAIC,GAAGN,GAAGK,GAAG,EAAE,GAAGC,GAAG,GAAGA,IAAItG,EAAE,QAAQ,SAAS,IAAIuG,GAAGX,GAAGS,GAAGrE,EAAE,GAAG8B,GAAG+B,GAAGS,GAAG3B,EAAEX,GAAGuC,GAAG,QAAQtC,GAAG,EAAEA,GAAGjE,EAAE,WAAW,EAAEiE,GAAG,CAAC,IAAIC,GAAG1B,EAAEsB,GAAGG,GAAG9B,GAAG,QAAQqE,GAAG,EAAEA,GAAGxG,EAAE,YAAY,EAAEwG,GAAG3B,EAAEkB,GAAGS,GAAGjE,IAAI2B,GAAGiB,EAAEnB,GAAGwC,IAAIxC,IAAIhE,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO5C,EAAE,eAAe0E,EAAE,MAAMA,EAAE,MAAM+C,CAAC,CAAC,CAAC,IAAI82E,GAAG,CAAC,WAAW1hE,GAAG,YAAY,MAAM,WAAWyhE,EAAE,EAAE,SAASE,GAAG1+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,EAAE,gBAAgBC,EAAE,YAAYS,CAAC,EAAEtC,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,sBAAsB,EAAE,IAAIsC,EAAE23C,EAAE,wBAAwBt4C,CAAC,EAAEY,EAAE03C,EAAE,kBAAkBj6C,EAAE,MAAMqC,EAAE,EAAE,EAAEX,EAAEE,EAAE,GAAGU,CAAC,EAAE,CAAC,aAAaE,EAAE,YAAYC,EAAE,aAAakB,EAAE,YAAY,CAAC,EAAEpB,EAAEyE,EAAEzE,EAAE,aAAa,eAAe,EAAE,IAAIwmB,GAAGxmB,EAAE,YAAY,SAAS,EAAE6B,EAAE7B,EAAE,QAAQ,KAAK8B,EAAE9B,EAAE,QAAQ,IAAI+B,EAAE3E,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuE,EAAE5E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOsH,EAAE,IAAI8hB,GAAG/oB,EAAE,MAAMA,EAAE,MAAMsE,CAAC,EAAEG,EAAE,IAAIskB,GAAG,EAAE,MAAM,EAAE,MAAMxkB,CAAC,EAAE,QAAQ2C,EAAE,EAAEA,EAAEvD,EAAE,EAAEuD,EAAE,CAAC,IAAIxC,EAAE,KAAK,IAAI,EAAE,KAAK,MAAML,EAAE6C,GAAG1E,CAAC,CAAC,EAAEmC,EAAE,KAAK,IAAIpC,EAAE,WAAWA,EAAE,SAAS8B,EAAE6C,GAAG1E,CAAC,EAAE,QAAQoC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAE,KAAK,IAAI,EAAE,KAAK,MAAMT,EAAEQ,GAAGnC,CAAC,CAAC,EAAEqC,EAAE,KAAK,IAAIvC,EAAE,UAAUA,EAAE,QAAQ6B,EAAEQ,GAAGnC,CAAC,EAAE,QAAQsC,EAAE,EAAEA,EAAExC,EAAE,WAAW,EAAEwC,EAAE,QAAQ2C,EAAE,EAAEA,EAAEnF,EAAE,YAAY,EAAEmF,EAAE,CAAC,IAAIN,EAAE,EAAE,QAAQO,EAAE,EAAEA,EAAEpF,EAAE,UAAU,EAAEoF,EAAE,QAAQC,EAAGlD,EAAEkD,EAAGjD,EAAE,EAAEiD,EAAG,CAAC,IAAIC,EAAGX,EAAEU,EAAGpF,EAAE6B,EAAE,QAAQyD,GAAGjD,EAAEiD,GAAGhD,EAAE,EAAEgD,GAAG,CAAC,IAAIC,GAAGnD,EAAEkD,GAAGrF,EAAE2B,EAAE4C,EAAEI,GAAGH,EAAE,IAAIU,EAAEE,EAAGE,GAAGhD,CAAC,EAAEN,EAAE,IAAIkD,EAAEC,EAAGE,GAAGJ,CAAC,EAAEN,GAAGH,EAAE,IAAIU,EAAE5C,EAAE8C,EAAGE,EAAE,EAAEtD,EAAE,IAAIkD,EAAED,EAAEE,EAAGE,EAAE,CAAC,CAAC,CAAC,EAAE,IAAIV,EAAEF,EAAEtC,EAAEG,EAAE2C,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO/H,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAIy+E,GAAG,CAAC,WAAW3hE,GAAG,YAAY,MAAM,WAAW0hE,EAAE,EAAE,SAASE,GAAG5+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,WAAWC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,qBAAqB,EAAE,IAAIsC,EAAEikB,EAAE,eAAe,EAAE,KAAK,EAAEhkB,EAAEgkB,EAAE,eAAevmB,EAAE,KAAK,EAAEwC,EAAEy3C,EAAE,wBAAwBr4C,CAAC,EAAEa,EAAEw3C,EAAE,kBAAkB,EAAE,EAAE,MAAMv4C,EAAE,EAAEC,EAAEU,EAAE,GAAGG,CAAC,EAAEmB,EAAE,IAAIolB,GAAGtmB,EAAE,QAAQ,SAAS,EAAE,EAAEkB,EAAE,OAAOqD,EAAErH,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,EAAEL,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAACyE,EAAEC,EAAEC,CAAC,EAAEhC,EAAE,CAAC,UAAUiC,EAAE,aAAa0C,EAAE,YAAYxC,EAAE,WAAWyC,EAAE,SAASxC,EAAE,QAAQC,EAAE,YAAYC,EAAE,UAAUC,EAAE,SAASC,EAAE,aAAaC,EAAE,YAAY2C,CAAC,EAAEjF,EAAED,EAAEC,EAAE,WAAW,IAAI2E,EAAEH,EAAE,EAAExE,EAAE,QAAQ,IAAIkF,EAAElD,EAAE,EAAEhC,EAAE,QAAQ,KAAKmF,EAAGpF,IAAI,eAAeqF,EAAGlE,EAAE,QAAQ,GAAGmE,GAAGF,EAAGjE,EAAE,QAAQ,GAAGA,EAAE,QAAQ,GAAGoE,GAAGH,EAAGjE,EAAE,QAAQ,GAAG,EAAEqE,GAAGJ,EAAG,EAAEjE,EAAE,QAAQ,GAAGyC,GAAG7D,EAAE,GAAG2F,GAAGN,EAAGrF,EAAE,GAAGA,EAAE,GAAG4F,GAAGP,EAAGrF,EAAE,GAAG,EAAE6F,GAAGR,EAAG,EAAErF,EAAE,GAAG,QAAQ8F,GAAG,EAAEA,GAAG9D,EAAE,EAAE8D,GAAG,QAAQC,GAAG,EAAEA,GAAGpB,EAAE,EAAEoB,GAAG,QAAQC,GAAG,EAAEA,GAAG7D,EAAE,EAAE6D,GAAG,CAAC,IAAIK,GAAGL,GAAGnB,EAAEyB,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKD,GAAG7D,CAAC,CAAC,EAAE+D,GAAG,KAAK,IAAIjE,GAAGoC,EAAE2B,IAAI7D,CAAC,EAAE,QAAQsB,GAAG,EAAEA,GAAG1B,EAAE,EAAE0B,GAAG,CAAC,IAAIE,GAAGF,GAAGsB,EAAEnB,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKD,GAAGmB,CAAC,CAAC,EAAEjB,GAAG,KAAK,IAAI3B,GAAGL,EAAE8B,IAAImB,CAAC,EAAEqB,GAAG,EAAE,QAAQpC,GAAGkC,GAAGlC,GAAGmC,GAAG,EAAEnC,GAAG,CAAC,IAAIC,GAAGD,GAAG5B,EAAE6D,GAAG,QAAQ/B,GAAGL,GAAGK,GAAGJ,GAAG,EAAEI,GAAG,CAAC,IAAIP,GAAGO,GAAGa,EAAEnB,GAAGyC,GAAG5C,GAAGiC,GAAGH,GAAGvB,GAAGwB,GAAGtB,GAAGoC,GAAG7E,GAAG6C,EAAE,EAAEL,IAAIvC,GAAGI,EAAE,EAAE6B,IAAIhC,EAAEgE,GAAG,QAAQY,GAAG,EAAEA,GAAGtE,EAAE,EAAEsE,GAAG,CAAC,IAAIC,GAAGnC,EAAEgC,GAAGZ,GAAGc,IAAIE,GAAG,EAAEH,GAAGC,IAAIH,IAAII,GAAGC,EAAE,CAAC,CAAC,CAAC,IAAI1C,GAAGmB,EAAGQ,GAAGP,GAAGS,GAAGR,GAAG1B,GAAG2B,GAAGM,GAAG,EAAE5B,IAAIqC,EAAE,CAAC,CAAC,OAAOpJ,EAAE,eAAegE,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI26E,GAAG,CAAC,WAAW5hE,GAAG,YAAY,MAAM,WAAW2hE,EAAE,EAAE,SAASE,GAAG9+E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,UAAUC,CAAC,EAAE5B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,QAAQ,EAAE,IAAI4B,EAAEq4C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAE2B,EAAED,CAAC,EAAE,CAAC,YAAYW,EAAE,aAAaC,EAAE,YAAYC,EAAE,cAAcC,EAAE,eAAeC,EAAE,cAAckB,EAAE,QAAQ,CAAC,EAAE/B,EAAEoF,EAAE,EAAE,MAAM,EAAE,EAAE,KAAK5C,EAAE,EAAE,IAAIC,EAAE,IAAI0kB,GAAGnnB,EAAE,SAAS5B,EAAE,KAAK,EAAEsE,EAAE3E,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuE,EAAE5E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOsH,EAAE5C,EAAE,OAAOI,EAAE8hB,EAAE,eAAevmB,EAAE,KAAK,EAAEkH,EAAEqf,EAAE,eAAe,EAAE,KAAK,EAAE,QAAQ7hB,EAAE,EAAEA,EAAE9C,EAAE,UAAU,EAAE8C,EAAE,CAAC,IAAIC,EAAED,EAAED,EAAE,GAAGG,EAAEF,EAAEL,EAAE,QAAQ,GAAG,QAAQQ,EAAE,EAAEA,EAAEjD,EAAE,SAAS,EAAEiD,EAAE,CAAC,IAAIC,EAAEF,EAAEC,EAAER,EAAE,QAAQ,GAAGU,EAAEF,EAAEjD,EAAE,YAAYoF,EAAE,QAAQU,EAAE,EAAEA,EAAErF,EAAE,EAAEqF,EAAE,CAAC,IAAIN,EAAErC,EAAE2C,EAAElF,EAAE,GAAG4E,EAAE,GAAGA,GAAGxF,EAAE,QAAQ,SAAS,IAAI+F,EAAED,EAAER,EAAE,GAAGU,EAAGjD,EAAEyC,EAAE3C,EAAE,GAAG,QAAQoD,EAAG,EAAEA,EAAGjG,EAAE,UAAU,EAAEiG,EAAG,CAAC,IAAIC,GAAGhD,EAAE+C,EAAGxD,EAAE,QAAQ,GAAG0D,GAAGF,EAAGjG,EAAE,aAAawC,EAAE,QAAQ4D,GAAG,EAAEA,GAAG1F,EAAE,EAAE0F,GAAG,CAAC,IAAI5B,GAAG2B,GAAGC,GAAGvF,EAAE,GAAG2D,GAAG,GAAGA,IAAIxE,EAAE,SAAS,SAAS,IAAIsG,GAAGP,EAAEK,GAAGd,EAAE,GAAGiB,GAAGP,EAAGxB,GAAG3B,EAAE,GAAG,QAAQ2D,GAAG,EAAEA,GAAGxG,EAAE,SAAS,EAAEwG,GAAG,CAAC,IAAIC,GAAGP,GAAGM,GAAGxG,EAAE,YAAY0G,GAAGF,GAAGxG,EAAE,YAAY,EAAE,QAAQ2G,GAAG,EAAEA,GAAGhG,EAAE,EAAEgG,GAAG,CAAC,IAAIK,GAAGN,GAAGC,GAAG5E,EAAE,GAAGiF,GAAG,GAAGA,IAAIhH,EAAE,QAAQ,SAAS,IAAIiH,GAAGX,GAAGK,GAAGrB,EAAE,GAAG4B,GAAGX,GAAGS,GAAGhH,EAAE,WAAWyE,GAAGwC,GAAG,QAAQtC,GAAG,EAAEA,GAAG3E,EAAE,WAAW,EAAE2E,GAAG,CAAC,IAAIC,GAAGlC,EAAEwE,GAAGvC,IAAI,QAAQE,GAAG,EAAEA,GAAG7E,EAAE,YAAY,EAAE6E,GAAGQ,EAAEoB,GAAG5B,KAAKD,GAAGjC,EAAE8B,GAAGI,IAAIJ,IAAIzE,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOjC,EAAE,eAAe0E,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIm6E,GAAG,CAAC,WAAW7hE,GAAG,YAAY,MAAM,WAAW4hE,EAAE,EAAE,SAASE,GAAGh/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,YAAYC,CAAC,EAAE5B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,wBAAwB,EAAE,IAAI4B,EAAE2kB,EAAE,eAAevmB,EAAE,KAAK,EAAEqC,EAAEkkB,EAAE,eAAe,EAAE,KAAK,EAAEjkB,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAM2B,EAAE,EAAE,EAAED,CAAC,EAAEa,EAAED,EAAE,YAAYE,EAAEF,EAAE,aAAaG,EAAEH,EAAE,YAAYqB,EAAErB,EAAE,YAAY,EAAEA,EAAE,aAAa0E,EAAE1E,EAAE,YAAY,EAAE,IAAIymB,GAAGzmB,EAAE,YAAY,SAAS,EAAE8B,EAAE,EAAE,OAAO,CAACC,EAAEC,EAAEC,EAAE0C,CAAC,EAAE,EAAE,QAAQxC,EAAE9E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAACuH,EAAExC,EAAEC,EAAEC,CAAC,EAAEvC,EAAEwC,EAAElF,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAAC8E,EAAEC,EAAE2C,EAAEN,CAAC,EAAExF,EAAE+F,EAAErF,EAAE,QAAQ,MAAMsF,EAAGtF,EAAE,QAAQ,KAAKuF,EAAGvF,EAAE,QAAQ,IAAI,QAAQwF,GAAG,EAAEA,GAAGnE,EAAE,EAAEmE,GAAG,CAAC,IAAIC,GAAG,KAAK,IAAI,EAAE,KAAK,MAAMJ,EAAEG,IAAIvF,CAAC,CAAC,EAAEyF,GAAG,KAAK,IAAI1F,EAAE,UAAUA,EAAE,QAAQqF,EAAEG,IAAIvF,CAAC,EAAE6D,GAAG0B,GAAGzD,EAAE,QAAQ6D,GAAG,EAAEA,GAAG,EAAE,EAAEA,GAAG,CAAC,IAAIC,GAAG,KAAK,IAAI,EAAE,KAAK,MAAMN,EAAGK,IAAI1F,CAAC,CAAC,EAAE4F,GAAG,KAAK,IAAI9F,EAAE,WAAWA,EAAE,SAASuF,EAAGK,IAAI1F,CAAC,EAAE6F,GAAGH,GAAG5D,EAAE8B,GAAG,QAAQkC,GAAG,EAAEA,GAAGtB,EAAE,EAAEsB,GAAG,CAAC,IAAIC,GAAG,KAAK,IAAI,EAAE,KAAK,MAAMX,EAAGU,IAAI7F,CAAC,CAAC,EAAEmG,GAAG,KAAK,IAAItG,EAAE,UAAUA,EAAE,QAAQsF,EAAGU,IAAI7F,CAAC,EAAEoG,GAAGP,GAAG/D,EAAE8D,GAAG,QAAQS,GAAG,EAAEA,GAAGxG,EAAE,WAAW,EAAEwG,GAAG,CAAC,IAAIzC,GAAGyC,GAAG7B,EAAE4B,GAAG,QAAQtC,GAAG,EAAEA,GAAGjE,EAAE,YAAY,EAAEiE,GAAG,CAAC,IAAIC,GAAG,EAAE,QAAQC,GAAG,EAAEA,GAAGnE,EAAE,UAAU,EAAEmE,GAAG,CAAC,IAAIsC,GAAGtC,GAAG3B,EAAE4B,GAAGD,GAAGS,EAAE,QAAQP,GAAGoB,GAAGpB,GAAGqB,GAAG,EAAErB,GAAG,CAAC,IAAIE,IAAIiB,GAAGnB,GAAGpE,EAAEoF,GAAG5C,EAAEgE,GAAGzC,GAAGK,GAAGjC,EAAEgC,GAAG,QAAQsC,GAAGb,GAAGa,GAAGZ,GAAG,EAAEY,GAAG,CAAC,IAAIE,IAAIhB,GAAGc,GAAGxG,EAAEqF,GAAIH,EAAEb,GAAGsC,GAAGH,GAAGrE,EAAE2B,GAAG,QAAQ8C,GAAGb,GAAGa,GAAGR,GAAG,EAAEQ,GAAG,CAAC,IAAI3B,IAAIa,GAAGc,GAAG3G,EAAEmF,GAAIR,EAAE8B,GAAGI,GAAGF,GAAGxE,EAAEuE,GAAG3C,IAAI3B,EAAE4C,GAAGqB,IAAIrE,EAAE6E,GAAG/C,GAAG,CAAC,CAAC,CAAC,CAACnC,EAAEiC,GAAGE,IAAIC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO7G,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAI++E,GAAG,CAAC,WAAW9hE,GAAG,YAAY,MAAM,WAAW6hE,EAAE,EAAE,SAASE,GAAGl/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,IAAI,EAAE,QAAQ0B,EAAE,WAAWC,CAAC,EAAE5B,EAAEmxE,GAAG,CAAClxE,CAAC,EAAE,uBAAuB,EAAE,IAAI4B,EAAE2kB,EAAE,eAAevmB,EAAE,KAAK,EAAEqC,EAAEkkB,EAAE,eAAe,EAAE,KAAK,EAAEjkB,EAAE23C,EAAE,kBAAkBt4C,EAAE,EAAE,MAAMD,EAAE,EAAE,CAAC,EAAEa,EAAE,IAAIwmB,GAAGzmB,EAAE,QAAQ,SAAS,EAAEE,EAAED,EAAE,OAAO,CAACE,EAAEkB,EAAE,EAAEqD,CAAC,EAAEzE,EAAE,QAAQ,EAAE5C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAACoE,EAAEC,EAAEC,EAAEC,CAAC,EAAE3C,EAAEqF,EAAEtH,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC8E,EAAEyC,EAAExC,EAAEC,CAAC,EAAEtC,EAAE,CAAC,UAAUuC,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAYC,EAAE,WAAW2C,EAAE,QAAQN,EAAE,SAASO,EAAE,QAAQC,EAAG,YAAYC,EAAG,SAASC,GAAG,UAAUC,GAAG,SAASC,GAAG,YAAY5B,GAAG,aAAa8B,GAAG,YAAYC,EAAE,EAAE7F,EAAE8F,GAAGvD,EAAE,EAAEvC,EAAE,QAAQ,MAAM+F,GAAGvD,EAAE,EAAExC,EAAE,QAAQ,IAAIgG,GAAGvD,EAAE,EAAEzC,EAAE,QAAQ,KAAK,QAAQiG,GAAG,EAAEA,GAAG3D,EAAE,EAAE2D,GAAG,QAAQK,GAAG,EAAEA,GAAGlB,EAAE,EAAEkB,GAAG,QAAQC,GAAG,EAAEA,GAAGzB,EAAE,EAAEyB,GAAG,CAAC,IAAIC,GAAGD,GAAGT,GAAG/B,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKyC,GAAG1C,EAAE,CAAC,EAAEG,GAAG,KAAK,IAAIuB,IAAIjD,EAAEiE,IAAI1C,EAAE,EAAE,QAAQI,GAAG,EAAEA,GAAGmB,EAAE,EAAEnB,GAAG,CAAC,IAAIC,GAAGD,GAAG6B,GAAGU,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKtC,GAAGyB,EAAE,CAAC,EAAExB,GAAG,KAAK,IAAIqB,IAAIjD,EAAE2B,IAAIyB,EAAE,EAAE,QAAQvB,GAAG,EAAEA,GAAGiB,EAAG,EAAEjB,GAAG,CAAC,IAAIC,GAAGD,GAAG2B,GAAGzB,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKD,GAAGuB,EAAE,CAAC,EAAE7B,GAAG,KAAK,IAAI0B,IAAIjD,EAAE6B,IAAIuB,EAAE,EAAEa,GAAG,EAAE,QAAQC,GAAG5C,GAAG4C,GAAG1C,GAAG,EAAE0C,GAAG,CAAC,IAAIC,GAAGD,GAAG7C,GAAG0C,GAAG,QAAQK,GAAGJ,GAAGI,GAAGzC,GAAG,EAAEyC,GAAG,CAAC,IAAIC,GAAGD,GAAGjB,GAAGzB,GAAG,QAAQ4C,GAAGxC,GAAGwC,GAAG/C,GAAG,EAAE+C,GAAG,CAAC,IAAI,GAAGA,GAAGlB,GAAGvB,GAAG0C,GAAGlF,EAAEmE,GAAGlE,EAAE4E,GAAG3E,EAAE6E,GAAG5E,EAAE8E,GAAGI,GAAGhF,GAAGI,EAAE,EAAEqE,IAAIhC,GAAGpC,EAAE,EAAEsE,IAAI1E,GAAGK,EAAE,EAAE,IAAIJ,EAAEiE,GAAG,QAAQc,GAAG,EAAEA,GAAG7B,EAAG,EAAE6B,GAAG,CAAC,IAAIH,GAAG,EAAED,GAAGI,IAAIE,GAAG3C,EAAEwC,GAAGC,IAAIV,IAAIO,GAAGK,EAAE,CAAC,CAAC,CAAC,CAACpH,EAAEC,EAAE8F,GAAG5E,EAAEkF,GAAG,EAAErC,GAAGQ,EAAEL,GAAGiC,IAAII,EAAE,CAAC,CAAC,CAAC,OAAOrJ,EAAE,eAAe4C,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIq8E,GAAG,CAAC,WAAW/hE,GAAG,YAAY,MAAM,WAAW8hE,EAAE,EAAME,GAAGrJ,GAAG14D,GAAGrd,GAAG,KAAK,IAAIA,CAAC,CAAC,EAAEq/E,GAAG,CAAC,WAAWhiE,GAAG,YAAY,MAAM,WAAW+hE,EAAE,EAAME,GAAGvJ,GAAGz4D,GAAGtd,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEu/E,GAAG,CAAC,WAAWjiE,GAAG,YAAY,MAAM,WAAWgiE,EAAE,EAAE,SAASE,GAAGx/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,SAAS0B,EAAE,OAAOC,EAAE,mBAAmBC,CAAC,EAAE7B,EAAE,CAACsC,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAMyC,EAAE,EAAE,MAAM,GAAG,CAACkB,EAAE,CAAC,EAAEjC,EAAEsF,EAAEgpB,GAAG,CAACvtB,EAAEkB,EAAE,EAAEnB,CAAC,EAAE,SAAS,EAAE,EAAE7C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOyE,EAAEzE,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0E,EAAE1E,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOsE,EAAEiiB,EAAE,eAAevmB,EAAE,KAAK,EAAEuE,EAAEgiB,EAAE,eAAevf,EAAE,KAAK,EAAE,QAAQC,EAAE,EAAEA,EAAExE,EAAEwE,IAAI,CAAC,IAAIxC,EAAEwC,EAAE,EAAEC,EAAE,EAAEzC,GAAGC,EAAE,EAAED,EAAE,GAAGE,EAAE,EAAEF,EAAE,GAAGG,EAAE,EAAEH,EAAE,GAAGI,EAAET,EAAE6C,GAAG,GAAGpC,GAAGxC,EAAE,SAAS,IAAIyC,EAAEnB,EAAE,GAAGgB,EAAEuC,IAAI5E,EAAE,IAAIqB,EAAE,GAAG,EAAEoB,EAAE,EAAE,GAAGH,EAAEF,IAAInC,EAAE,IAAI,EAAE,GAAG,EAAE,QAAQmF,EAAE,EAAEA,EAAE/D,EAAE+D,IAAI,CAAC,IAAIN,EAAEzD,EAAE,EAAEuD,GAAG5E,EAAE,GAAGoF,EAAE5C,EAAE,IAAIoC,EAAEvC,IAAIrC,EAAE,GAAG,GAAG8E,EAAE,GAAGA,EAAE9E,EAAE,EAAE,CAAC,QAAQqF,EAAE,EAAEA,EAAE,EAAEA,IAAI,QAAQC,EAAG,EAAEA,EAAGpF,EAAEoF,IAAK,CAAC,IAAIC,EAAGD,EAAGD,EAAEpD,EAAE,GAAGmD,EAAEnD,EAAE,GAAG0C,EAAE1C,EAAE,GAAGyC,EAAE,OAAOa,GAAIjG,CAAC,CAAC,QAAQ,CAAC,GAAGD,IAAI,WAAW,CAAC,IAAIgG,EAAE,KAAK,MAAMP,CAAC,EAAEQ,EAAG,KAAK,KAAKR,CAAC,EAAES,EAAGT,EAAEO,EAAE,QAAQG,GAAG,EAAEA,GAAG,EAAEA,KAAK,CAAC,IAAIC,GAAG,EAAE,EAAErD,GAAGnC,EAAE,GAAGuF,GAAG/C,EAAE,IAAIL,EAAEE,IAAIrC,EAAE,GAAG,GAAGwF,GAAG,GAAGA,GAAGxF,EAAE,EAAE,CAAC,QAAQ4F,GAAG,EAAEA,GAAG3F,EAAE2F,KAAK,CAAC,IAAIC,GAAGD,GAAGL,GAAGvD,EAAE,GAAGmD,EAAEnD,EAAE,GAAG0C,EAAE1C,EAAE,GAAGyC,EAAE,OAAOoB,IAAIxG,CAAC,CAAC,QAAQ,CAAC,IAAIoG,GAAG,KAAK,MAAMD,EAAE,EAAE3B,GAAG,KAAK,KAAK2B,EAAE,EAAEG,GAAGH,GAAGC,GAAG,QAAQG,GAAG,EAAEA,GAAG3F,EAAE2F,KAAK,CAAC,IAAIC,GAAGD,GAAGH,GAAG1D,EAAE,GAAGqD,EAAErD,EAAE,GAAGO,EAAEP,EAAE,GAAG+D,GAAGhE,EAAE+D,IAAIA,GAAGD,GAAG/B,GAAG9B,EAAE,GAAGqD,EAAErD,EAAE,GAAGO,EAAEP,EAAE,GAAG,IAAIgE,GAAGjE,EAAE+D,IAAIA,GAAGD,GAAGH,GAAG1D,EAAE,GAAGsD,EAAGtD,EAAE,GAAGO,EAAEP,EAAE,GAAG,IAAIiE,GAAGlE,EAAE+D,IAAIA,GAAGD,GAAG/B,GAAG9B,EAAE,GAAGsD,EAAGtD,EAAE,GAAGO,EAAEP,EAAE,GAAG,IAAIsE,GAAGvE,EAAE+D,IAAIS,GAAGR,IAAIC,GAAGD,IAAIH,GAAGY,GAAGP,IAAIK,GAAGL,IAAIL,GAAGE,GAAGD,GAAGL,GAAGvD,EAAE,GAAGmD,EAAEnD,EAAE,GAAG0C,EAAE1C,EAAE,GAAGyC,EAAE,OAAOoB,IAAIS,IAAIC,GAAGD,IAAIhB,CAAE,CAAC,CAAC,KAAM,SAAQF,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAG,EAAE,EAAElD,GAAGnC,EAAE,GAAGoF,EAAE5C,EAAE,IAAIL,EAAEE,IAAIrC,EAAE,GAAG,GAAGqF,EAAG,GAAGA,EAAGrF,EAAE,EAAE,CAAC,QAAQwF,GAAG,EAAEA,GAAGvF,EAAEuF,KAAK,CAAC,IAAIC,GAAGD,GAAGJ,EAAEpD,EAAE,GAAGmD,EAAEnD,EAAE,GAAG0C,EAAE1C,EAAE,GAAGyC,EAAE,OAAOgB,IAAIpG,CAAC,CAAC,QAAQ,CAAC,IAAIiG,EAAG,KAAK,MAAMD,CAAE,EAAEE,GAAG,KAAK,MAAMV,CAAC,EAAE,QAAQW,GAAG,EAAEA,GAAGvF,EAAEuF,KAAK,CAAC,IAAIC,GAAGD,GAAGF,EAAGvD,EAAE,GAAGwD,GAAGxD,EAAE,GAAGO,EAAEP,EAAE,GAAG8B,GAAG2B,GAAGJ,EAAEpD,EAAE,GAAGmD,EAAEnD,EAAE,GAAG0C,EAAE1C,EAAE,GAAGyC,EAAE,OAAOZ,IAAI/B,EAAE2D,GAAG,CAAC,CAAC,CAAC,CAAC,OAAOrI,EAAE,eAAeqH,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIk4E,GAAG,CAAC,WAAWhiE,GAAG,YAAY,MAAM,WAAW+hE,EAAE,EAAE,SAASE,GAAG1/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAEmxE,GAAGlxE,EAAE,SAAS,EAAE,IAAI2B,EAAEs4C,EAAE,mBAAmB,CAAC,CAAC,EAAEj6C,EAAE,MAAM,MAAM,EAAE4B,EAAE5B,EAAE2B,GAAG,OAAOC,EAAE61E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKgC,CAAC,CAAC,CAAC,GAAG,IAAIU,EAAE43C,EAAE,iBAAiB,EAAEj6C,EAAE,MAAM,MAAM,EAAE,GAAG,GAAGqC,IAAIT,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,qDAAqDA,EAAE,MAAM,OAAO,kBAAkBS,GAAG,EAAE,IAAIC,EAAE8nB,GAAGxoB,EAAE,MAAM,OAAO,EAAEW,EAAEgkB,EAAE,mBAAmBA,EAAE,cAAc3kB,EAAE,KAAK,EAAEU,CAAC,EAAEE,EAAE7C,EAAE,KAAK,IAAIiC,EAAE,MAAM,EAAE,OAAOa,EAAEb,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG+B,EAAEjC,EAAE,CAACsF,EAAE,IAAIA,EAAEvE,EAAE,EAAE,EAAE,CAACuE,EAAE,IAAIA,EAAE,EAAE,QAAQA,EAAE,EAAEA,EAAExE,EAAE,OAAOwE,GAAGvE,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,IAAI,CAAC,IAAI2B,EAAET,EAAEqD,EAAE,CAAC,EAAE,GAAG,IAAI,EAAEzE,EAAE6B,GAAG,EAAE,EAAE5B,EAAE4B,OAAO,CAAC,IAAIC,EAAEV,EAAEqD,EAAE,EAAE,CAAC,EAAEzE,EAAE6B,GAAG,EAAE5B,EAAE6B,GAAG9B,EAAE8B,GAAG7B,EAAE4B,GAAG7B,EAAE8B,EAAE,CAAC,CAAC,IAAI,EAAE1E,EAAE,eAAeiC,EAAE,MAAMU,EAAEC,CAAC,EAAE,GAAGZ,GAAG,KAAK,CAAC,IAAIqF,EAAEizC,EAAE,uBAAuBt4C,CAAC,EAAE,EAAE81E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ93E,EAAE,MAAM,CAAC,KAAKqH,CAAC,CAAC,CAAC,EAAE,OAAOrH,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8BiC,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,IAAIw9E,GAAG,CAAC,WAAWpiE,GAAG,YAAY,MAAM,WAAWmiE,EAAE,EAAE,SAASE,GAAG5/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAEmxE,GAAGlxE,EAAE,QAAQ,EAAE,IAAI2B,EAAEs4C,EAAE,mBAAmB,CAAC,CAAC,EAAEj6C,EAAE,MAAM,MAAM,EAAE4B,EAAE5B,EAAE2B,GAAG,OAAOC,EAAE61E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKgC,CAAC,CAAC,CAAC,GAAG,IAAIU,EAAE43C,EAAE,iBAAiB,EAAEj6C,EAAE,MAAM,MAAM,EAAE,GAAG,GAAGqC,IAAIT,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,oDAAoDA,EAAE,MAAM,OAAO,kBAAkBS,GAAG,EAAE,IAAIC,EAAE8nB,GAAGxoB,EAAE,MAAM,OAAO,EAAEW,EAAEgkB,EAAE,oBAAoBA,EAAE,cAAc3kB,EAAE,KAAK,EAAEU,CAAC,EAAEE,EAAE7C,EAAE,KAAK,IAAIiC,EAAE,MAAM,EAAE,OAAOa,EAAEb,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG+B,EAAEjC,EAAE,CAACsF,EAAE,IAAIA,EAAEvE,EAAE,EAAE,EAAE,CAACuE,EAAE,IAAIA,EAAE,EAAE,QAAQA,EAAE,EAAEA,EAAExE,EAAE,OAAOwE,GAAGvE,EAAE,QAAQ,EAAE,EAAE,EAAEA,EAAE,IAAI,CAAC,IAAI2B,EAAET,EAAEqD,EAAE,CAAC,EAAE,GAAG,IAAI,EAAEzE,EAAE6B,GAAG,EAAE,EAAE5B,EAAE4B,OAAO,CAAC,IAAIC,EAAEV,EAAEqD,EAAE,EAAE,CAAC,EAAEzE,EAAE6B,GAAG,EAAE5B,EAAE6B,GAAG9B,EAAE8B,GAAG7B,EAAE4B,GAAG7B,EAAE8B,EAAE,CAAC,CAAC,IAAI,EAAE1E,EAAE,eAAeiC,EAAE,MAAMU,EAAEC,CAAC,EAAE,GAAGZ,GAAG,KAAK,CAAC,IAAIqF,EAAEizC,EAAE,uBAAuBt4C,CAAC,EAAE,EAAE81E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ93E,EAAE,MAAM,CAAC,KAAKqH,CAAC,CAAC,CAAC,EAAE,OAAOrH,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8BiC,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI09E,GAAG,CAAC,WAAWriE,GAAG,YAAY,MAAM,WAAWoiE,EAAE,EAAE,SAASE,GAAG9/E,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,aAAa0B,CAAC,EAAE3B,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO4B,EAAEjC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0C,EAAEkvE,GAAG5vE,EAAEC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,OAAOjC,EAAE,eAAe,CAAC,CAAC,EAAE,EAAE,MAAM0C,CAAC,CAAC,SAASrC,EAAE,MAAM,SAAS,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,WAAWK,CAAC,EAAE4B,EAAEjC,EAAE,WAAW,CAAC,EAAE0C,EAAEmvE,GAAG7vE,EAAEC,EAAE,EAAEF,CAAC,EAAE,OAAO/B,EAAE,eAAe0C,EAAE,MAAM,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,MAAM,IAAI,MAAM,qEAAqErC,EAAE,MAAM,SAAS,CAAC,CAAC,IAAIw/E,GAAG,CAAC,WAAWriE,GAAG,YAAY,MAAM,WAAWoiE,EAAE,EAAE,SAASE,GAAGhgF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,UAAU,EAAE,WAAW,CAAC,EAAED,EAAEwmB,EAAE,OAAO,IAAI,OAAO,IAAI,+DAA+D,GAAG,EAAE,IAAI7kB,EAAE1B,EAAE,MAAM,GAAG2B,EAAE3B,EAAE,MAAM,GAAG4B,EAAE5B,EAAE,MAAM,GAAGqC,EAAErC,EAAE,MAAM,GAAGsC,EAAEX,EAAE,EAAEY,EAAEX,EAAE,EAAEY,EAAEH,GAAG,EAAE,GAAGI,EAAE9C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2D,EAAE,IAAI,aAAajC,EAAEY,EAAEC,EAAEC,CAAC,EAAE,EAAE,EAAE,QAAQwE,EAAE,EAAEA,EAAEtF,EAAE,EAAEsF,EAAE,QAAQ,EAAE,EAAE,EAAE1E,EAAE,EAAE,EAAE,CAAC,IAAI8B,EAAE,KAAK,MAAM,EAAE,CAAC,EAAEC,EAAE,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE/B,EAAE,EAAE+B,EAAE,CAAC,IAAIC,EAAE,KAAK,MAAMD,EAAE,CAAC,EAAE2C,EAAE3C,EAAE,EAAEG,GAAGJ,EAAE,EAAE4C,GAAGzE,EAAE,QAAQ0E,EAAE,EAAEA,EAAE1E,EAAE,EAAE0E,EAAE,CAAC,IAAIvC,EAAEuC,EAAEzC,EAAEpC,GAAGkC,EAAE3C,GAAGwC,EAAEzC,EAAEqF,IAAIrD,EAAE,KAAKlB,EAAEkC,EAAE,CAAC,CAAC,CAAC,OAAOhF,EAAE,eAAe,CAAC+B,EAAEY,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAM2D,CAAC,CAAC,CAAC,IAAI+7E,GAAG,CAAC,WAAWtiE,GAAG,YAAY,MAAM,WAAWqiE,EAAE,EAAE,SAASE,GAAGlgF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,UAAUC,EAAE,gBAAgBC,CAAC,EAAE7B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,uBAAuB,EAAE,IAAIqC,EAAEkkB,EAAE,eAAevmB,EAAE,KAAK,EAAEsC,EAAEikB,EAAE,eAAe,EAAE,KAAK,EAAEhkB,EAAEZ,EAAEY,GAAG,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAGgkB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAE13C,CAAC,EAAE,IAAI,gFAAgF,oBAAoBA,IAAI,EAAE,IAAIC,EAAEy3C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAEuC,EAAEb,EAAEE,EAAE,EAAE,EAAE,CAAC,aAAaa,EAAE,YAAYkB,EAAE,eAAe,EAAE,cAAcqD,EAAE,QAAQ,CAAC,EAAExE,EAAE4B,EAAE,EAAE,KAAKC,EAAE,EAAE,IAAIC,EAAE9B,EAAE,YAAYA,EAAE,WAAW+B,EAAE,IAAIwkB,GAAGvmB,EAAE,SAASxC,EAAE,KAAK,EAAEiH,EAAEtH,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOyE,EAAE9E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOuH,EAAE3C,EAAE,OAAO,QAAQG,EAAE,EAAEA,EAAElC,EAAE,UAAU,EAAEkC,EAAE,CAAC,IAAIC,EAAED,EAAErC,EAAE,GAAGuC,EAAEF,EAAEH,EAAE,QAAQ,GAAG,QAAQM,EAAE,EAAEA,EAAErC,EAAE,UAAU,EAAEqC,EAAE,CAAC,IAAIC,EAAEF,EAAEC,EAAEN,EAAE,QAAQ,GAAGQ,EAAEF,EAAErC,EAAE,aAAa6B,EAAE,QAAQqD,EAAE,EAAEA,EAAEjF,EAAE,EAAEiF,EAAE,CAAC,IAAIN,EAAErC,EAAE2C,EAAE,EAAE,GAAGN,EAAE,GAAGA,GAAG5E,EAAE,SAAS,SAAS,IAAImF,EAAED,EAAEpF,EAAE,GAAGsF,EAAGjD,EAAEyC,EAAE/E,EAAE,GAAG,QAAQwF,EAAG,EAAEA,EAAGrF,EAAE,SAAS,EAAEqF,EAAG,CAAC,IAAIC,GAAGhD,EAAE+C,EAAGtD,EAAE,QAAQ,GAAGwD,GAAGF,EAAGrF,EAAE,YAAY4B,EAAE,QAAQ4D,GAAG,EAAEA,GAAGrE,EAAE,EAAEqE,GAAG,CAAC,IAAI5B,GAAG2B,GAAGC,GAAGhB,EAAE,GAAGZ,GAAG,GAAGA,IAAI5D,EAAE,QAAQ,SAAS,IAAI0F,GAAGP,EAAEK,GAAG1F,EAAE,GAAG6F,GAAGP,EAAGxB,GAAG5D,EAAE,WAAW4F,GAAGN,GAAGO,GAAGH,GAAG,QAAQI,GAAG,EAAEA,GAAG9F,EAAE,WAAW,EAAE8F,GAAG,CAAC,IAAIC,GAAGtB,EAAEkB,GAAGG,IAAI,QAAQM,GAAG,EAAEA,GAAGtE,EAAE,EAAEsE,GAAG1B,EAAEkB,GAAGQ,KAAKL,GAAG9D,EAAE4D,GAAGO,IAAIR,IAAI9D,EAAE+D,IAAI/D,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO3E,EAAE,eAAe4E,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIq7E,GAAG,CAAC,WAAWviE,GAAG,YAAY,MAAM,WAAWsiE,EAAE,EAAE,SAASE,GAAGpgF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAU0B,EAAE,IAAIC,EAAE,gBAAgBC,EAAE,YAAYS,CAAC,EAAEtC,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,qCAAqC,EAAE,IAAIsC,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAMqC,EAAE,EAAEX,EAAEC,EAAEC,EAAE,EAAE,EAAE,CAAC,aAAaW,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAYkB,CAAC,EAAErB,EAAE,EAAE,IAAIymB,GAAGzmB,EAAE,YAAY,SAAS,EAAE0E,EAAE1E,EAAE,QAAQ,KAAK,EAAEA,EAAE,QAAQ,IAAI8B,EAAE9B,EAAE,YAAYA,EAAE,WAAW+B,EAAE1E,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOsE,EAAE,IAAIykB,GAAG/oB,EAAE,MAAMA,EAAE,MAAMqE,CAAC,EAAEE,EAAE5E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOsH,EAAE,IAAI8hB,GAAG,EAAE,MAAM,EAAE,MAAMxkB,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEhC,EAAE,EAAEgC,EAAE,CAAC,IAAIyC,EAAE,KAAK,IAAI,EAAE,KAAK,MAAM,EAAEzC,GAAGlC,CAAC,CAAC,EAAEmC,EAAE,KAAK,IAAIpC,EAAE,WAAWA,EAAE,SAAS,EAAEmC,GAAGlC,CAAC,EAAE,QAAQoC,EAAE,EAAEA,EAAEhB,EAAE,EAAEgB,EAAE,CAAC,IAAIC,EAAE,KAAK,IAAI,EAAE,KAAK,MAAMoC,EAAErC,GAAGnC,CAAC,CAAC,EAAEqC,EAAE,KAAK,IAAIvC,EAAE,UAAUA,EAAE,QAAQ0E,EAAErC,GAAGnC,CAAC,EAAE,QAAQsC,EAAE,EAAEA,EAAExC,EAAE,YAAY,EAAEwC,EAAE,CAAC,IAAIC,EAAE,KAAK,MAAMD,EAAEV,CAAC,EAAEsD,EAAE5C,EAAEV,EAAEgD,EAAE,EAAE,QAAQO,EAAE,EAAEA,EAAErF,EAAE,UAAU,EAAEqF,EAAE,QAAQC,EAAGV,EAAEU,EAAGlD,EAAE,EAAEkD,EAAG,CAAC,IAAIC,EAAGpD,EAAEmD,EAAGrF,EAAE,EAAE,QAAQuF,GAAGlD,EAAEkD,GAAGjD,EAAE,EAAEiD,GAAG,CAAC,IAAIC,GAAGpD,EAAEmD,GAAGtF,EAAEwE,EAAEI,GAAG9C,EAAE,IAAIqD,EAAEE,EAAGE,GAAGhD,CAAC,EAAEkC,EAAE,IAAIU,EAAEC,EAAGE,GAAGhD,CAAC,CAAC,CAAC,CAAC,EAAE,IAAIsC,EAAE3C,EAAEE,EAAEI,EAAE2C,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO/H,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAImgF,GAAG,CAAC,WAAWxiE,GAAG,YAAY,MAAM,WAAWuiE,EAAE,EAAE,SAASE,GAAGtgF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAU0B,EAAE,IAAIC,EAAE,gBAAgBC,EAAE,WAAWS,CAAC,EAAEtC,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,oCAAoC,EAAE,IAAIsC,EAAEikB,EAAE,eAAevmB,EAAE,KAAK,EAAEuC,EAAEgkB,EAAE,eAAe,EAAE,KAAK,EAAE/jB,EAAEy3C,EAAE,kBAAkB53C,EAAE,EAAE,MAAM,EAAEX,EAAEC,EAAEC,EAAE,EAAE,EAAEa,EAAE,IAAIsmB,GAAGvmB,EAAE,QAAQ,SAAS,EAAEmB,EAAElB,EAAE,OAAO,CAAC,EAAEuE,EAAE,CAAC,EAAEvE,EAAE,QAAQ2B,EAAEzE,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAACqE,EAAEC,EAAEC,CAAC,EAAEjC,EAAE2E,EAAEtH,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC8E,EAAEyC,EAAExC,CAAC,EAAEnC,EAAE,CAAC,UAAUoC,EAAE,aAAaC,EAAE,YAAYC,EAAE,WAAWC,EAAE,SAASC,EAAE,QAAQ2C,EAAE,YAAYN,EAAE,UAAUO,EAAE,SAASC,EAAG,aAAaC,EAAG,YAAYC,EAAE,EAAEtF,EAAEuF,GAAGnD,EAAE,EAAEpC,EAAE,QAAQ,IAAIwF,GAAGnD,EAAE,EAAErC,EAAE,QAAQ,KAAK4D,GAAGgB,EAAEtC,EAAE,QAAQoD,GAAG,EAAEA,GAAGvD,EAAE,EAAEuD,GAAG,QAAQC,GAAG,EAAEA,GAAGrD,EAAE,EAAEqD,GAAG,QAAQC,GAAG,EAAEA,GAAGrD,EAAE,EAAEqD,GAAG,CAAC,IAAIC,GAAGD,GAAGL,GAAGO,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKD,GAAGR,CAAE,CAAC,EAAEU,GAAG,KAAK,IAAIZ,GAAG/C,EAAEyD,IAAIR,CAAE,EAAE,QAAQe,GAAG,EAAEA,GAAGlB,EAAE,EAAEkB,GAAG,CAAC,IAAIC,GAAGD,GAAGZ,GAAGc,GAAG,KAAK,IAAI,EAAE,KAAK,KAAKD,GAAGf,EAAE,CAAC,EAAEzB,GAAG,KAAK,IAAIuB,GAAI/C,EAAEgE,IAAIf,EAAE,EAAEvB,GAAG,EAAE,QAAQC,GAAG8B,GAAG9B,GAAG+B,GAAG,EAAE/B,GAAG,CAAC,IAAIC,GAAGD,GAAGqB,EAAGQ,GAAG,QAAQU,GAAGD,GAAGC,GAAG1C,GAAG,EAAE0C,GAAG,CAAC,IAAIrC,GAAGqC,GAAGjB,GAAGe,GAAGlC,GAAGtC,EAAE6D,GAAG5D,EAAEkC,GAAGjC,EAAEwE,GAAGnC,GAAGnC,GAAGG,EAAE,EAAE6B,IAAIS,GAAGrC,EAAE,EAAE6B,IAAIhC,EAAEyD,GAAG,QAAQtB,GAAG,EAAEA,GAAGT,GAAG,EAAES,GAAG,CAAC,IAAIP,GAAG6B,GAAG/B,GAAGS,GAAGmC,GAAG5E,EAAEuC,GAAGL,IAAI2C,GAAGhC,EAAEL,GAAGC,IAAIN,IAAIyC,GAAGC,EAAE,CAAC,CAAC,CAACtF,EAAE,EAAEuE,GAAGlB,EAAEoB,GAAG,EAAEQ,GAAGT,IAAI5B,EAAE,CAAC,CAAC,OAAO5G,EAAE,eAAe8C,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIu9E,GAAG,CAAC,WAAWziE,GAAG,YAAY,MAAM,WAAWwiE,EAAE,EAAE,SAASE,GAAGxgF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAEC,EAAEumB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAEJ,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,EAAEiwB,GAAG,CAAChwB,EAAEA,CAAC,EAAED,EAAE,KAAK,EAAE2B,EAAE,EAAE,OAAO,QAAQE,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAIF,EAAEE,EAAE5B,EAAE4B,GAAG,EAAEA,GAAG,IAAID,EAAE,CAAC,GAAG5B,EAAE,MAAM,GAAGA,EAAE,KAAK,EAAE,OAAOJ,EAAE,eAAegC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAIu+E,GAAG,CAAC,WAAW1iE,GAAG,YAAY,MAAM,WAAWyiE,EAAE,EAAME,GAAG,CAAC,WAAW1iE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAOhe,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,EAAE,OAAOC,CAAC,EAAEP,EAAE,CAAC,QAAQ,EAAE,IAAI,EAAE,UAAUiC,CAAC,EAAE/B,EAAEgC,EAAE,EAAEC,EAAED,EAAE,KAAK,IAAI5B,EAAE,MAAM,EAAE,OAAOsC,EAAEtC,EAAE,MAAM,OAAOuC,EAAEX,EAAE,KAAK,IAAI3B,EAAE,MAAM,EAAE,OAAOuC,EAAEvC,EAAE,MAAM,OAAO,CAAC,UAAUwC,EAAE,SAASC,EAAE,QAAQkB,EAAE,WAAW,EAAE,UAAUqD,EAAE,SAAS,EAAE,QAAQ5C,EAAE,aAAaC,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAY0C,EAAE,eAAexC,EAAE,cAAcyC,EAAE,SAASxC,CAAC,EAAEu1C,EAAE,sBAAsBl6C,EAAE,MAAMC,EAAE,MAAM,EAAE,EAAE,OAAO0B,CAAC,EAAEiD,EAAE4hB,EAAE,cAAc7hB,CAAC,EAAEE,EAAEF,EAAE,OAAOG,EAAE0hB,EAAE,kBAAkBxmB,EAAE,MAAM4E,CAAC,EAAE,QAAQI,EAAE,EAAEA,EAAEvC,EAAE,EAAEuC,EAAE,QAAQ2C,EAAE,EAAEA,EAAEV,EAAE,EAAEU,EAAE,CAAC,IAAIN,EAAEM,EAAErD,EAAED,EAAE,IAAI,QAAQuD,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAGD,EAAErD,EAAEF,EAAE,KAAK,QAAQyD,EAAG,EAAEA,EAAG,EAAE,EAAEA,EAAG,CAAC,IAAIC,EAAG,OAAO,iBAAiB,QAAQE,GAAG,EAAEA,GAAGzD,EAAE,EAAEyD,GAAG,CAAC,IAAI5B,GAAGgB,EAAEY,GAAGvD,EAAE,GAAG2B,IAAI,GAAGA,GAAG3D,EAAE,QAAQyF,GAAG,EAAEA,GAAGjB,EAAE,EAAEiB,GAAG,CAAC,IAAIC,GAAGP,EAAGM,GAAGhB,EAAE,GAAGiB,IAAI,GAAGA,GAAGxE,EAAE,CAAC,IAAIyE,GAAGme,EAAE,WAAW,CAACxhB,EAAEqB,GAAG+B,GAAGN,CAAE,EAAExF,EAAEkkB,EAAE,eAAexmB,EAAE,KAAK,CAAC,EAAEsI,GAAGke,EAAE,WAAW,CAACve,GAAGE,GAAGL,CAAE,EAAEtF,EAAEgkB,EAAE,eAAevmB,EAAE,KAAK,CAAC,EAAEsI,GAAG1G,EAAEwG,IAAI9F,EAAE+F,IAAIC,GAAGR,IAAKA,EAAGQ,GAAG,CAAC,CAAC,CAAC,IAAIP,GAAGwe,EAAE,WAAW,CAACxhB,EAAE2C,EAAEC,EAAEE,CAAE,EAAEjD,EAAE2hB,EAAE,eAAe7hB,CAAC,CAAC,EAAEG,EAAEkD,IAAID,CAAE,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOnG,EAAE,MAAM4kB,EAAE,aAAa1hB,EAAE9E,EAAE,KAAK,EAAE2E,EAAE3E,EAAE,KAAK,EAAE,MAAM2E,EAAE,MAAM3E,EAAE,KAAK,CAAC,CAAC,EAAMqgF,GAAG,CAAC,WAAWziE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAOle,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,EAAE,OAAOC,EAAE,GAAG,CAAC,EAAEP,EAAE,CAAC,QAAQ,EAAE,IAAIiC,EAAE,UAAUC,CAAC,EAAEhC,EAAEiC,EAAE,EAAES,EAAEkkB,EAAE,cAAcxmB,EAAE,MAAM6B,EAAE,KAAK,IAAI7B,EAAE,MAAM,EAAE,MAAM,EAAEuC,EAAEikB,EAAE,cAAcvmB,EAAE,MAAM4B,EAAE,KAAK,IAAI5B,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,UAAUuC,EAAE,SAASC,EAAE,QAAQC,EAAE,WAAWkB,EAAE,UAAU,EAAE,SAASqD,EAAE,QAAQ,EAAE,aAAa5C,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAYC,EAAE,eAAe0C,EAAE,cAAcxC,EAAE,SAASyC,CAAC,EAAE+yC,EAAE,sBAAsBl6C,EAAE,MAAMC,EAAE,MAAM,EAAE0B,EAAE,OAAOC,CAAC,EAAE4kB,EAAE,OAAO,EAAE,OAAOrf,EAAE,OAAO,IAAI,YAAYyW,4CAA4CzW,EAAE,mBAAmB,EAAE,MAAM,EAAE,IAAIxC,EAAE6hB,EAAE,cAAcrf,EAAEtF,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE+C,EAAE4hB,EAAE,0BAA0BvmB,EAAE,MAAMA,EAAE,KAAK,EAAE,QAAQ6E,EAAE,EAAEA,EAAEtC,EAAE,EAAEsC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAED,EAAEV,EAAE,EAAE,IAAI,QAAQsD,EAAE,EAAEA,EAAEV,EAAE,EAAEU,EAAE,CAAC,IAAIN,EAAEM,EAAErD,EAAE,EAAE,KAAK,QAAQsD,EAAE,EAAEA,EAAEhE,EAAE,EAAEgE,EAAE,CAAC,IAAIC,EAAG,OAAO,iBAAiBC,EAAG,EAAEC,EAAG,EAAE,QAAQC,GAAG,EAAEA,GAAGzD,EAAE,EAAEyD,GAAG,CAAC,IAAIC,GAAGjD,EAAEgD,GAAGd,EAAE,GAAGe,IAAI,GAAGA,GAAGxF,EAAE,QAAQ4D,GAAG,EAAEA,GAAG7B,EAAE,EAAE6B,GAAG,CAAC,IAAI8B,GAAGd,EAAEhB,GAAG3B,EAAE,GAAGyD,IAAI,GAAGA,GAAGzF,EAAE,CAAC,IAAI0F,GAAG9F,EAAEwC,GAAGmD,IAAIE,IAAIP,GAAGrF,EAAEyF,IAAI3B,IAAIuB,GAAGQ,GAAGP,IAAKA,EAAGO,GAAGN,EAAGE,GAAGD,EAAG1B,GAAG,CAAC,CAAC,CAACzB,EAAEkD,GAAIC,GAAIH,IAAIjD,EAAEG,GAAGC,GAAG4C,GAAGC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO/F,EAAE,MAAM2kB,EAAE,aAAa5hB,EAAE5E,EAAE,KAAK,EAAEC,EAAE,MAAMA,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,EAAMqgF,GAAG,CAAC,WAAW3iE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAOje,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,EAAE,OAAOC,EAAE,GAAG,CAAC,EAAEP,EAAE,CAAC,QAAQ,EAAE,IAAIiC,EAAE,UAAUC,CAAC,EAAEhC,EAAEiC,EAAE,EAAES,EAAEkkB,EAAE,cAAcxmB,EAAE,MAAM6B,EAAE,KAAK,IAAI7B,EAAE,MAAM,EAAE,MAAM,EAAEuC,EAAEikB,EAAE,cAAcvmB,EAAE,MAAM4B,EAAE,KAAK,IAAI5B,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,UAAUuC,EAAE,SAASC,EAAE,QAAQC,EAAE,WAAWkB,EAAE,UAAU,EAAE,SAASqD,EAAE,QAAQ,EAAE,aAAa5C,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAYC,EAAE,eAAe0C,EAAE,cAAcxC,EAAE,SAASyC,CAAC,EAAE+yC,EAAE,sBAAsBl6C,EAAE,MAAMC,EAAE,MAAM,EAAE0B,EAAE,OAAOC,CAAC,EAAE4kB,EAAE,OAAO,EAAE,OAAOrf,EAAE,OAAO,IAAI,YAAYwW,4CAA4CxW,EAAE,mBAAmB,EAAE,MAAM,EAAE,IAAIxC,EAAE6hB,EAAE,cAAcrf,EAAEtF,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE+C,EAAE4hB,EAAE,0BAA0BxmB,EAAE,MAAMA,EAAE,KAAK,EAAE,QAAQ8E,EAAE,EAAEA,EAAEtC,EAAE,EAAEsC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAED,EAAEV,EAAE,EAAE,IAAI,QAAQsD,EAAE,EAAEA,EAAEV,EAAE,EAAEU,EAAE,CAAC,IAAIN,EAAEM,EAAErD,EAAE,EAAE,KAAK,QAAQsD,EAAE,EAAEA,EAAEhE,EAAE,EAAEgE,EAAE,CAAC,IAAIC,EAAG,OAAO,iBAAiBC,EAAG9C,EAAE,EAAE,EAAEA,EAAE+C,EAAGV,EAAE,EAAE,EAAEA,EAAE,QAAQW,GAAG,EAAEA,GAAGzD,EAAE,EAAEyD,GAAG,CAAC,IAAIC,GAAGjD,EAAEgD,GAAGd,EAAE,GAAGe,IAAI,GAAGA,GAAGxF,EAAE,QAAQ4D,GAAG,EAAEA,GAAG7B,EAAE,EAAE6B,GAAG,CAAC,IAAI8B,GAAGd,EAAEhB,GAAG3B,EAAE,GAAGyD,IAAI,GAAGA,GAAGzF,EAAE,CAAC,IAAI0F,GAAG9F,EAAEwC,GAAGmD,IAAIE,IAAIP,GAAGrF,EAAEyF,IAAI3B,IAAIuB,GAAGQ,GAAGP,IAAKA,EAAGO,GAAGN,EAAGG,GAAGF,EAAGI,GAAG,CAAC,CAAC,CAACvD,EAAEE,GAAGgD,GAAIC,GAAIH,IAAIjD,EAAEG,GAAGC,GAAG4C,GAAGC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,OAAO/F,EAAE,MAAM2kB,EAAE,aAAa5hB,EAAE5E,EAAE,KAAK,EAAEA,EAAE,MAAMA,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,EAAE,SAASugF,GAAG7gF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI0B,EAAE1B,EAAE,QAAQ,OAAO0B,EAAEszE,GAAG,CAAC,OAAO,CAAC,EAAEh1E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,OAAO,CAAC,CAAC,EAAE+B,EAAEkzE,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAIgC,EAAED,EAAE,MAAM,OAAOE,EAAE2kB,EAAE,eAAe,EAAE7kB,EAAE,KAAK,EAAEW,EAAE43C,EAAE,mBAAmBr4C,EAAED,CAAC,EAAEW,EAAEV,EAAEW,EAAEb,EAAEW,GAAG,OAAOE,EAAEk1E,GAAG,CAAC,OAAO,CAAC,EAAE/1E,CAAC,EAAE,QAAQ/B,EAAE,MAAM,CAAC,KAAK0C,CAAC,CAAC,CAAC,EAAEC,EAAE23C,EAAE,iBAAiB33C,EAAE,OAAOX,CAAC,GAAGs4C,EAAE,2BAA2B,MAAM33C,EAAEC,EAAE,MAAM,MAAM,EAAE,GAAG,CAACC,EAAEC,CAAC,EAAEw3C,EAAE,0BAA0B13C,EAAE,MAAMD,CAAC,EAAEqB,EAAEs2C,EAAE,WAAW13C,EAAE,MAAM,OAAO,EAAE,EAAEoyE,GAAGh1E,EAAE6C,EAAEmB,CAAC,EAAEqD,EAAEuf,EAAE,cAAc9jB,CAAC,EAAE,EAAE9C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOyE,EAAEzE,EAAE,KAAK,IAAI4C,EAAE,MAAM,EAAE,OAAO,QAAQ8B,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAED,EAAE2C,EAAEzC,EAAE,EAAE,QAAQ0C,EAAE,EAAEA,EAAED,EAAE,EAAEC,EAAE1C,GAAGH,EAAEE,EAAE2C,GAAG,EAAE5C,GAAGE,CAAC,CAAC,GAAG,EAAE,CAAC,IAAIF,EAAE41C,EAAE,qBAAqB,EAAE,MAAMr4C,CAAC,EAAE0C,EAAE,EAAE,EAAEg2E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAM0E,CAAC,CAAC,CAAC,EAAE1E,EAAE,8BAA8B2E,CAAC,CAAC,CAAC,OAAO3E,EAAE,8BAA8B+B,CAAC,EAAEW,GAAG,MAAM1C,EAAE,8BAA8B4C,CAAC,EAAE,CAAC,CAAC,IAAIg+E,GAAG,CAAC,WAAW98D,GAAG,YAAY,MAAM,WAAW68D,EAAE,EAAE,SAASE,GAAG/gF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,SAASO,CAAC,EAAED,EAAE,EAAE,EAAE,CAAC,QAAQ,EAAE,WAAW2B,EAAE,OAAOC,CAAC,EAAEs4C,EAAE,qBAAqBj6C,EAAE,EAAE,MAAM,EAAEi6C,EAAE,oBAAoB,EAAE,OAAOt4C,EAAE,CAAC,EAAE,GAAG,CAAC,KAAKC,EAAE,MAAMS,CAAC,EAAE43C,EAAE,qBAAqBv4C,EAAEC,CAAC,EAAEW,EAAED,EAAE,OAAOE,EAAE,KAAKC,EAAE,EAAE,OAAOC,EAAE,CAAC,EAAE,QAAQkB,EAAE,EAAEA,EAAErB,EAAE,EAAEqB,EAAE,CAAC,QAAQ,KAAKtB,EAAEsB,GAAG,CAAC,GAAG,CAAC,mBAAmBqD,EAAE,WAAW,CAAC,EAAEizC,EAAE,qBAAqBz3C,EAAEb,EAAE,EAAE,EAAEyC,EAAE61C,EAAE,sBAAsBjzC,CAAC,EAAE5C,EAAE,EAAE,IAAIA,EAAEqzE,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,EAAE,EAAE,QAAQ93E,EAAE,MAAM,CAAC,KAAKqH,CAAC,CAAC,CAAC,EAAEvE,EAAE,KAAK2B,CAAC,GAAG,IAAIC,EAAED,EAAE,MAAM,MAAM,EAAE,QAAQE,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAED,EAAE,OAAO,EAAEC,GAAG,EAAE,CAAC,EAAEiiB,EAAE,YAAYniB,EAAE,MAAMC,CAAC,IAAID,EAAEk2E,GAAG,CAAC,OAAO,CAACl2E,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,MAAM0E,CAAC,CAAC,CAAC,EAAE5B,EAAE,KAAK2B,CAAC,GAAG7B,IAAI,KAAKA,EAAE6B,GAAG7B,EAAE40E,GAAG,CAAC,OAAO,CAAC,EAAE/yE,EAAE,EAAE7B,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE8C,EAAE,KAAKF,CAAC,EAAE,CAACoB,EAAErB,EAAE,IAAIV,EAAE+B,IAAI,IAAIpB,EAAE+9E,GAAG,CAAC,OAAO,CAAC,EAAE/9E,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,KAAKiC,EAAE+B,IAAI,EAAE,OAAOnB,GAAG,SAAS,EAAE,CAAC,CAAC,EAAEC,EAAE,KAAKF,CAAC,GAAGC,IAAI,CAAC,QAAQmB,KAAKlB,EAAEkB,IAAIpB,GAAG5C,EAAE,8BAA8BgE,CAAC,EAAE,OAAOpB,CAAC,CAAC,IAAIk+E,GAAG,CAAC,WAAW5iE,GAAG,YAAY,MAAM,WAAW2iE,EAAE,EAAE,SAASE,GAAGjhF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,GAAGM,EAAE,EAAEC,CAAC,EAAE,EAAEkxE,GAAG,CAACnxE,EAAEC,CAAC,EAAE,SAAS,EAAE,IAAI,EAAE,IAAI,aAAaumB,EAAE,cAAcvmB,EAAE,KAAK,CAAC,EAAE,EAAEL,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO0B,EAAE/B,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO,QAAQ4B,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE,EAAED,GAAGC,GAAG,EAAE,EAAED,GAAGD,EAAEC,GAAG,EAAEA,GAAGD,EAAEC,IAAIC,EAAE,EAAE,CAAC,OAAOjC,EAAE,eAAeK,EAAE,MAAM,UAAU,CAAC,CAAC,CAAC,IAAI2gF,GAAG,CAAC,WAAW5iE,GAAG,YAAY,MAAM,WAAW2iE,EAAE,EAAME,GAAG3mC,EAAE,MAAM4mC,GAAG5mC,EAAE,OAAO6mC,GAAG7mC,EAAE,OAAO8mC,GAAG9mC,EAAE,OAAO+mC,GAAG/mC,EAAE,OAAOgnC,GAAGhnC,EAAE,OAAOinC,GAAG1L,GAAGx3D,GAAGve,GAAG,CAAC,IAAI,EAAE,KAAK,KAAKA,CAAC,EAAEE,EAAE,KAAK,IAAIF,CAAC,EAAEM,EAAE,GAAG,EAAE6gF,GAAGjhF,GAAG,OAAO,GAAG,MAAMshF,GAAGlhF,EAAEihF,IAAIjhF,EAAEghF,IAAIhhF,EAAE+gF,IAAI/gF,EAAE8gF,IAAI9gF,EAAE,KAAK,IAAI,CAACJ,EAAEA,CAAC,EAAE,CAAC,EAAEwhF,GAAG,CAAC,WAAWnjE,GAAG,YAAY,MAAM,WAAWkjE,EAAE,EAAE,SAASE,GAAG3hF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,IAAI,CAAC,EAAED,EAAE,EAAEC,EAAE,MAAM,OAAO0B,EAAE1B,EAAE,MAAM,MAAM,EAAE2B,EAAE,EAAE,OAAO,EAAE,IAAI4kB,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,IAAI,iCAAiC,EAAE,EAAE,OAAO,IAAI,EAAE5kB,EAAE,EAAE,EAAE,GAAGD,EAAE,OAAOC,EAAE,EAAE,CAAC,EAAE24E,GAAG,CAAC,OAAO,CAAC,EAAEt6E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM+B,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI2/E,GAAG,CAAC,WAAWljE,GAAG,YAAY,MAAM,WAAWijE,EAAE,EAAME,GAAG9M,GAAG,CAAC/0E,EAAE,IAAIA,EAAE,CAAC,EAAE8hF,GAAGrM,GAAGt3D,GAAG0jE,EAAE,EAAEE,GAAG,CAAC,WAAW5jE,GAAG,YAAY,MAAM,WAAW2jE,EAAE,EAAE,SAASE,GAAGhiF,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAMO,EAAED,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAEJ,EAAE,KAAK,IAAIF,EAAE,MAAM,EAAEiC,EAAE,EAAE,mBAAmB,KAAKC,EAAE,EAAE,mBAAmB,KAAKC,EAAE,CAAC5B,EAAE,CAAC,EAAEqC,EAAEkkB,EAAE,cAAc3kB,CAAC,EAAEU,EAAEikB,EAAE,uBAAuB,UAAUlkB,CAAC,EAAEE,EAAEgkB,EAAE,uBAAuB,UAAUlkB,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAErC,EAAE,IAAI,CAAC,IAAIgH,EAAE6xE,GAAG,CAAC,OAAO,CAAC,EAAEn3E,CAAC,EAAE,QAAQ/B,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,EAAEk5E,GAAG,CAAC,OAAO,CAAC,EAAEl3E,CAAC,EAAE,QAAQhC,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEyE,EAAEqwE,GAAG,CAAC,OAAO,CAAC,KAAKztE,EAAE,KAAK,CAAC,EAAE,QAAQrH,CAAC,CAAC,EAAE,CAAC,KAAK0E,EAAE,KAAKC,CAAC,EAAEo9E,GAAGt9E,EAAE,EAAEzE,CAAC,EAAE4E,EAAE01C,EAAE,uBAAuB51C,EAAEC,CAAC,EAAE,QAAQ2C,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIxC,EAAEw1C,EAAE,oBAAoB11C,EAAE0C,CAAC,EAAE3E,EAAE,EAAE,EAAE2E,GAAGxC,EAAE,KAAKlC,EAAE,EAAE,EAAE0E,GAAGxC,EAAE,IAAI,CAAC9E,EAAE,8BAA8BqH,CAAC,EAAErH,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8ByE,CAAC,CAAC,CAAC,IAAI5B,EAAE7C,EAAE,eAAeiC,EAAE,UAAUU,CAAC,EAAEG,EAAE9C,EAAE,eAAeiC,EAAE,UAAUW,CAAC,EAAEoB,EAAE8wE,GAAG,CAAC,OAAO,CAAC,KAAKjyE,EAAE,KAAKC,CAAC,EAAE,QAAQ9C,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAEkB,CAAC,CAAC,SAAS+9E,GAAGjiF,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEwmB,EAAE,cAAc9mB,EAAE,KAAK,EAAEO,EAAEL,EAAE,KAAK,IAAIF,EAAE,MAAM,EAAE,EAAEE,EAAE,KAAK,IAAIK,EAAE,mBAAmB,KAAK,MAAM,EAAE,OAAO,EAAEL,EAAE,KAAK,IAAIK,EAAE,mBAAmB,KAAK,MAAM,EAAE,OAAO,GAAG2hF,GAAG5hF,CAAC,EAAE,CAAC,IAAI2B,EAAEkgF,GAAG,EAAE,EAAE7hF,EAAE,EAAEJ,CAAC,EAAEgC,EAAE,CAAClC,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,EAAE,GAAG,EAAE,CAAC,IAAImC,EAAEjC,EAAE,eAAegC,EAAE,UAAUD,EAAE,IAAI,EAAEW,EAAE1C,EAAE,eAAegC,EAAE,UAAUD,EAAE,IAAI,EAAEY,EAAE3C,EAAE,eAAe,CAAC,EAAE,UAAU4mB,EAAE,kBAAkBxmB,EAAE,SAAS,CAAC,EAAEwC,EAAEqyE,GAAG,CAAC,OAAO,CAAC,EAAEtyE,CAAC,EAAE,QAAQ3C,CAAC,CAAC,EAAE6C,EAAEg/E,GAAG,WAAW,CAAC,OAAO,CAAC,EAAE5/E,EAAE,EAAEU,CAAC,EAAE,QAAQ3C,CAAC,CAAC,EAAE8C,EAAE++E,GAAG,WAAW,CAAC,OAAO,CAAC,EAAEn/E,EAAE,EAAEE,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAEgE,EAAEhE,EAAE,KAAK,IAAI6C,EAAE,MAAM,EAAE,OAAO,EAAE7C,EAAE,KAAK,IAAI8C,EAAE,MAAM,EAAE,OAAO,OAAO9C,EAAE,8BAA8BiC,CAAC,EAAEjC,EAAE,8BAA8B0C,CAAC,EAAE1C,EAAE,8BAA8B2C,CAAC,EAAE3C,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAE,CAAC,KAAKkB,EAAE,KAAK,CAAC,CAAC,CAAC,OAAOjC,CAAC,KAAK,CAAC,IAAIA,EAAEu4C,EAAE,uBAAuB,EAAE,CAAC,EAAEt4C,EAAEkgF,GAAGngF,EAAE3B,EAAE,CAAC,EAAE,OAAOk6C,EAAE,uBAAuBt4C,CAAC,CAAC,CAAC,CAAC,SAASggF,GAAGliF,EAAE,CAAC,OAAOA,EAAEA,EAAE,KAAK,CAAC,CAAC,SAASmiF,GAAGniF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,GAAGL,IAAI,EAAE,MAAM,CAAC,KAAKF,EAAE,KAAK,CAAC,EAAE,IAAI,EAAEw6C,EAAE,uBAAuBx6C,EAAE,CAAC,EAAE,EAAEE,EAAE,EAAE+B,EAAEu4C,EAAE,qBAAqB,CAAC,EAAEt4C,EAAED,EAAE,KAAKE,EAAEF,EAAE,KAAKW,EAAE,CAACV,EAAE,MAAM,EAAEW,EAAEtC,EAAE,eAAeqC,EAAE,UAAUV,CAAC,EAAEY,EAAEvC,EAAE,eAAeqC,EAAE,UAAUT,CAAC,EAAEY,EAAEiyE,GAAG,CAAC,OAAO,CAAC,KAAKnyE,EAAE,KAAKC,CAAC,EAAE,QAAQvC,CAAC,CAAC,EAAEyC,EAAEw3C,EAAE,oBAAoB,CAAC,EAAEt2C,EAAElB,EAAE,KAAK,EAAEA,EAAE,KAAKuE,EAAE,CAACrD,EAAE,MAAM,EAAE,EAAE3D,EAAE,eAAegH,EAAE,UAAUrD,CAAC,EAAES,EAAEpE,EAAE,eAAegH,EAAE,UAAU,CAAC,EAAE3C,EAAEowE,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAKrwE,CAAC,EAAE,QAAQpE,CAAC,CAAC,EAAEsE,EAAEs9E,GAAGjgF,EAAEC,EAAE,EAAE7B,EAAEC,CAAC,EAAEuE,EAAED,EAAE,KAAK2C,EAAE3C,EAAE,KAAKG,EAAE,CAACF,EAAE,MAAM,EAAE2C,EAAElH,EAAE,eAAeyE,EAAE,UAAUF,CAAC,EAAEG,EAAE1E,EAAE,eAAeyE,EAAE,UAAUwC,CAAC,EAAEtC,EAAE8vE,GAAG,CAAC,OAAO,CAAC,KAAKvtE,EAAE,KAAKxC,CAAC,EAAE,QAAQ1E,CAAC,CAAC,EAAE4E,EAAEg9E,GAAGj+E,EAAE,EAAE,EAAE5D,EAAEC,CAAC,EAAE6E,EAAED,EAAE,KAAKE,EAAEF,EAAE,KAAKG,EAAE,CAACF,EAAE,MAAM,EAAE6C,EAAE1H,EAAE,eAAe+E,EAAE,UAAUF,CAAC,EAAEuC,EAAEpH,EAAE,eAAe+E,EAAE,UAAUD,CAAC,EAAE6C,EAAE8sE,GAAG,CAAC,OAAO,CAAC,KAAK/sE,EAAE,KAAKN,CAAC,EAAE,QAAQpH,CAAC,CAAC,EAAE4H,EAAGqyC,EAAE,UAAUt6C,EAAEI,CAAC,EAAE8H,EAAG,CAACD,EAAG,KAAK,MAAM,EAAEE,GAAG9H,EAAE,eAAe6H,EAAG,UAAUD,EAAG,IAAI,EAAEG,GAAG/H,EAAE,eAAe6H,EAAG,UAAUD,EAAG,IAAI,EAAEI,GAAGysE,GAAG,CAAC,OAAO,CAAC,KAAK3sE,GAAG,KAAKC,EAAE,EAAE,QAAQ/H,CAAC,CAAC,EAAEoG,GAAG+wE,GAAG,CAAC,OAAO,CAAC,EAAEnvE,GAAG,EAAEL,CAAC,EAAE,QAAQ3H,CAAC,CAAC,EAAEkI,GAAGmtE,GAAG,CAAC,OAAO,CAAC,EAAE1wE,EAAE,EAAEyB,EAAE,EAAE,QAAQpG,CAAC,CAAC,EAAEmI,GAAGmxE,GAAG,CAAC,OAAO,CAAC,EAAE30E,EAAE,EAAEyB,EAAE,EAAE,QAAQpG,CAAC,CAAC,EAAEoI,GAAG0sE,GAAG,CAAC,OAAO,CAAC,MAAM5sE,EAAE,EAAE,QAAQlI,CAAC,CAAC,EAAEqI,GAAGysE,GAAG,CAAC,OAAO,CAAC,MAAM3sE,EAAE,EAAE,QAAQnI,CAAC,CAAC,EAAEsI,GAAGu1E,GAAG,CAAC,OAAO,CAAC,MAAM31E,EAAE,EAAE,QAAQlI,CAAC,CAAC,EAAEuI,GAAGs1E,GAAG,CAAC,OAAO,CAAC,MAAM11E,EAAE,EAAE,QAAQnI,CAAC,CAAC,EAAE4I,GAAGm1E,GAAG,CAAC,OAAO,CAAC31E,GAAGC,EAAE,EAAE,QAAQrI,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE6I,GAAGk1E,GAAG,CAAC,OAAO,CAACz1E,GAAGC,EAAE,EAAE,QAAQvI,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE8I,GAAG9I,EAAE,KAAK,IAAI4I,GAAG,MAAM,EAAE,OAAOvC,GAAGrG,EAAE,KAAK,IAAI6I,GAAG,MAAM,EAAE,OAAO,OAAO7I,EAAE,8BAA8BsC,CAAC,EAAEtC,EAAE,8BAA8BuC,CAAC,EAAEvC,EAAE,8BAA8BwC,CAAC,EAAExC,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8BoE,CAAC,EAAEpE,EAAE,8BAA8BqE,CAAC,EAAErE,EAAE,8BAA8BkH,CAAC,EAAElH,EAAE,8BAA8B0E,CAAC,EAAE1E,EAAE,8BAA8B2E,CAAC,EAAE3E,EAAE,8BAA8B0H,CAAC,EAAE1H,EAAE,8BAA8BoH,CAAC,EAAEpH,EAAE,8BAA8B2H,CAAC,EAAE3H,EAAE,8BAA8B8H,EAAE,EAAE9H,EAAE,8BAA8B+H,EAAE,EAAE/H,EAAE,8BAA8BgI,EAAE,EAAEhI,EAAE,8BAA8BoG,EAAE,EAAEpG,EAAE,8BAA8BkI,EAAE,EAAElI,EAAE,8BAA8BmI,EAAE,EAAEnI,EAAE,8BAA8BoI,EAAE,EAAEpI,EAAE,8BAA8BsI,EAAE,EAAEtI,EAAE,8BAA8BqI,EAAE,EAAErI,EAAE,8BAA8BuI,EAAE,EAAEvI,EAAE,8BAA8B4I,EAAE,EAAE5I,EAAE,8BAA8B6I,EAAE,EAAE,CAAC,KAAKC,GAAG,KAAKzC,EAAE,CAAC,CAAC,SAASw7E,GAAGpiF,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,IAAI,aAAa,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAI,EAAE,EAAE,EAAE,EAAE,QAAQ0B,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIC,EAAEs4C,EAAE,SAASj6C,EAAE0B,EAAE,EAAE/B,CAAC,EAAEiC,EAAEq4C,EAAE,oBAAoBx6C,EAAEiC,CAAC,EAAE,GAAGE,EAAE,KAAKD,EAAE,KAAKC,EAAE,KAAKD,EAAE,KAAK,GAAGC,EAAE,KAAKD,EAAE,KAAKC,EAAE,KAAKD,EAAE,IAAI,CAAChC,IAAI,GAAG,EAAE,GAAG,GAAGs6C,EAAE,mBAAmBl6C,EAAE,EAAE,EAAEC,CAAC,CAAC,CAAC,OAAOD,CAAC,CAAC,SAAS+hF,GAAGriF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEumB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAEA,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG,EAAEC,EAAE,EAAE0B,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAEv6E,CAAC,EAAE,QAAQJ,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEgC,EAAE8/E,GAAG//E,EAAE,GAAG/B,CAAC,EAAEiC,EAAE04E,GAAG,CAAC,OAAO,CAAC,EAAE34E,CAAC,EAAE,QAAQhC,EAAE,MAAM,CAAC,MAAMI,EAAE,KAAK,CAAC,CAAC,EAAE,OAAOJ,EAAE,8BAA8B+B,CAAC,EAAE/B,EAAE,8BAA8BgC,CAAC,EAAEC,CAAC,CAAC,IAAImgF,GAAG,CAAC,WAAW1jE,GAAG,YAAY,MAAM,WAAWyjE,EAAE,EAAE,SAASE,GAAGviF,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,MAAMC,EAAE,MAAM,CAAC,EAAEL,EAAE,EAAE,GAAG4mB,EAAE,WAAWvmB,CAAC,EAAE0B,EAAE6kB,EAAE,kBAAkB,EAAEA,EAAE,cAAcxmB,CAAC,CAAC,EAAE,OAAOkiF,GAAGvgF,EAAE1B,EAAE,CAAC,EAAE,EAAE,eAAeD,EAAE,EAAE2B,CAAC,CAAC,CAAC,IAAIwgF,GAAG,CAAC,WAAW5jE,GAAG,YAAY,MAAM,WAAW0jE,EAAE,EAAE,SAASC,GAAGxiF,EAAE,EAAEE,EAAE,CAACF,EAAE,KAAK,CAAC,CAAC,CAAC,IAAI0iF,GAAG,CAAC,WAAW5jE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAO9e,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,MAAMI,CAAC,EAAEN,EAAEO,EAAEL,EAAE,EAAE4mB,EAAE,uBAAuBxmB,EAAE,MAAMwmB,EAAE,cAAcxmB,EAAE,KAAK,CAAC,EAAE,CAAC,EAAE2B,EAAEC,EAAEC,CAAC,EAAE7B,EAAE,MAAMsC,EAAErC,EAAE,KAAK,IAAID,EAAE,MAAM,EAAE,OAAO,QAAQwC,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIC,EAAED,EAAEZ,EAAED,EAAEE,EAAE,QAAQa,EAAE,EAAEA,EAAEf,EAAEe,IAAI,CAAC,IAAIkB,EAAElB,GAAGd,EAAEC,GAAG,QAAQgC,EAAE,EAAEA,EAAEjC,EAAEiC,IAAI,CAAC,IAAIoD,EAAEpD,EAAEhC,EAAE,QAAQ4C,EAAE,EAAEA,EAAE5C,EAAE4C,IAAI,CAAC,IAAIJ,EAAE,KAAK,MAAMzC,EAAEiC,EAAE,CAAC,EAAES,EAAE7B,EAAEmB,EAAEqD,EAAExC,EAAEF,EAAEjC,EAAEgC,GAAG,GAAGD,GAAG,GAAGA,EAAEzC,EAAE,CAAC,IAAI4C,EAAEH,EAAExC,EAAEqF,EAAEzE,EAAEmB,EAAEY,EAAEC,EAAEF,EAAEjC,EAAE4E,EAAE,CAAC,EAAE5C,GAAGC,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOtE,EAAE,MAAM,EAAED,EAAE,MAAMA,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,EAAMqiF,GAAG5N,GAAG,CAAC/0E,EAAE,IAAI,KAAK,MAAMA,EAAE,CAAC,CAAC,EAAE4iF,GAAGnN,GAAGz2D,GAAG2jE,GAAG,KAAK,OAAO,EAAEE,GAAG,CAAC,WAAW7jE,GAAG,YAAY,MAAM,WAAW4jE,EAAE,EAAE,SAASE,GAAG9iF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,WAAWS,EAAE,UAAUC,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAE1C,EAAE4D,EAAEs6E,GAAG,CAAC,OAAO,CAAC,EAAEj+E,EAAE,OAAO,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,QAAQgC,EAAE,IAAIC,EAAE,WAAWS,EAAE,UAAUC,EAAE,gBAAgBC,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,IAAI,EAAEoB,EAAE,GAAGtB,IAAI,QAAQ,EAAE,MAAM,SAAS,GAAG,EAAE,MAAM,KAAK,EAAE,CAAC,IAAI2E,EAAEszE,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,MAAM,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEgE,EAAE0xE,GAAG,CAAC,OAAO,CAAC,EAAE1xE,EAAE,EAAEqD,CAAC,EAAE,QAAQrH,CAAC,CAAC,EAAEA,EAAE,8BAA8BqH,CAAC,CAAC,MAAMrD,EAAE0xE,GAAG,CAAC,OAAO,CAAC,EAAE1xE,EAAE,EAAE,CAAC,EAAE,QAAQhE,CAAC,CAAC,EAAEA,EAAE,8BAA8B,CAAC,CAAC,CAAC,GAAG6C,EAAE,CAAC,IAAI,EAAEmB,EAAE,GAAGtB,IAAI,QAAQG,IAAI,SAASd,EAAE,MAAM,SAAS,GAAGA,EAAE,MAAM,KAAK,EAAE,CAAC,IAAIsF,EAAEszE,GAAG,CAAC,OAAO,CAAC,EAAE54E,CAAC,EAAE,QAAQ/B,EAAE,MAAM,CAAC,MAAM,CAAC+B,EAAE,MAAM,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEiC,EAAE02E,GAAG16E,EAAEgE,EAAEnB,EAAEwE,EAAEvE,CAAC,EAAE9C,EAAE,8BAA8BqH,CAAC,CAAC,MAAMrD,EAAE02E,GAAG16E,EAAEgE,EAAEnB,EAAEd,EAAEe,CAAC,EAAE9C,EAAE,8BAA8B,CAAC,CAAC,CAAC,OAAOgE,CAAC,CAAC,IAAI6+E,GAAG,CAAC,WAAWh9D,GAAG,YAAY,MAAM,WAAW+8D,EAAE,EAAE,SAASE,GAAGhjF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,WAAWS,EAAE,UAAUC,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAE1C,EAAE4D,EAAEg8E,GAAG,CAAC,OAAO,CAAC,EAAE3/E,EAAE,OAAO,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,QAAQgC,EAAE,IAAIC,EAAE,WAAWS,EAAE,UAAUC,EAAE,gBAAgBC,CAAC,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,IAAI,EAAEoB,EAAEA,EAAE0xE,GAAG,CAAC,OAAO,CAAC,EAAE1xE,EAAE,EAAE,CAAC,EAAE,QAAQhE,CAAC,CAAC,EAAEA,EAAE,8BAA8B,CAAC,CAAC,CAAC,GAAG6C,EAAE,CAAC,IAAI,EAAEmB,EAAEA,EAAE02E,GAAG16E,EAAEgE,EAAEnB,EAAEd,EAAEe,CAAC,EAAE9C,EAAE,8BAA8B,CAAC,CAAC,CAAC,OAAOgE,CAAC,CAAC,IAAI++E,GAAG,CAAC,WAAWj9D,GAAG,YAAY,MAAM,WAAWg9D,EAAE,EAAE,SAASE,GAAGljF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,OAAOM,EAAE,QAAQC,CAAC,EAAE,EAAE,EAAEumB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAEC,EAAE,MAAM0B,EAAE,EAAE,EAAE,OAAO,GAAG,CAACC,EAAEC,EAAES,EAAEC,CAAC,EAAE23C,EAAE,mBAAmBl6C,EAAEC,CAAC,EAAE,GAAG4B,IAAI,EAAE,OAAOjC,EAAE,eAAegC,EAAE5B,EAAE,MAAM,CAAC,CAAC,EAAE,IAAIwC,EAAE5C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOwC,EAAE7C,EAAE,WAAWI,CAAC,EAAE0C,EAAEuvE,GAAGzvE,EAAEC,EAAEzC,EAAE,MAAM6B,EAAEF,EAAEW,EAAEC,EAAEvC,EAAE,MAAM,CAAC,EAAE,OAAOJ,EAAE,eAAegC,EAAE5B,EAAE,MAAM0C,EAAE,MAAM,CAAC,CAAC,IAAImgF,GAAG,CAAC,WAAWhkE,GAAG,YAAY,MAAM,WAAW+jE,EAAE,EAAE,SAASE,GAAGpjF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU0B,CAAC,EAAE3B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,UAAU,EAAE,IAAI2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG4B,EAAEjC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0C,EAAErC,EAAE,MAAM2B,GAAG,QAAQ0C,EAAE,EAAEA,EAAEzC,EAAE,OAAO,EAAEyC,EAAE,CAAC,IAAIC,EAAE1C,EAAEyC,GAAGkiB,EAAE,OAAOjiB,GAAGjC,EAAE,GAAGiC,GAAG,EAAE,IAAI,6BAA6BA,mBAAmBjC,EAAE,IAAI,CAAC,CAAC,IAAIC,EAAEZ,EAAEA,GAAG,OAAOY,EAAE,GAAG,IAAIC,EAAEgkB,EAAE,cAAc,EAAE,KAAK,EAAE/jB,EAAEy3C,EAAE,aAAa,yBAAyBj6C,EAAE,EAAE2B,EAAEW,CAAC,EAAEG,EAAE63E,GAAG,CAAC,OAAO,CAAC,EAAEt6E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,CAAC6C,EAAE,UAAUA,EAAE,UAAUA,EAAE,QAAQA,EAAE,SAAS,CAAC,CAAC,CAAC,EAAEmB,EAAE22E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAM,CAAC6C,EAAE,UAAUD,EAAEC,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE,EAAE,CAACA,EAAE,UAAUA,EAAE,UAAUD,EAAEC,EAAE,UAAUA,EAAE,SAAS,EAAEwE,EAAErH,EAAE,WAAWgE,CAAC,EAAE,EAAEhE,EAAE,WAAW8C,CAAC,EAAE2B,EAAE6tE,GAAG,EAAEjrE,EAAE,CAAC,EAAE,OAAOrH,EAAE,8BAA8B8C,CAAC,EAAE9C,EAAE,8BAA8BgE,CAAC,EAAEhE,EAAE,eAAe6C,EAAE,YAAY4B,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI0+E,GAAG,CAAC,WAAWnkE,GAAG,YAAY,MAAM,WAAWkkE,EAAE,EAAE,SAASE,GAAGtjF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEumB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAEA,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG,EAAEC,EAAE,EAAE0B,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAEv6E,CAAC,EAAE,QAAQJ,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEgC,EAAE8/E,GAAG//E,EAAE,GAAG/B,CAAC,EAAEiC,EAAE04E,GAAG,CAAC,OAAO,CAAC,EAAE34E,CAAC,EAAE,QAAQhC,EAAE,MAAM,CAAC,MAAMI,EAAE,KAAK,CAAC,CAAC,EAAE,OAAOJ,EAAE,8BAA8B+B,CAAC,EAAE/B,EAAE,8BAA8BgC,CAAC,EAAEC,CAAC,CAAC,IAAIohF,GAAG,CAAC,WAAWhkE,GAAG,YAAY,MAAM,WAAW+jE,EAAE,EAAME,GAAGzN,GAAGt2D,GAAGzf,GAAG,OAAO,SAASA,CAAC,EAAE,EAAE,EAAE,MAAM,EAAEyjF,GAAG,CAAC,WAAWhkE,GAAG,YAAY,MAAM,WAAW+jE,EAAE,EAAME,GAAG3N,GAAGr2D,GAAG1f,GAAG,KAAK,IAAIA,CAAC,IAAI,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE2jF,GAAG,CAAC,WAAWjkE,GAAG,YAAY,MAAM,WAAWgkE,EAAE,EAAME,GAAG7N,GAAGp2D,GAAG3f,GAAG,OAAO,MAAMA,CAAC,EAAE,EAAE,EAAE,MAAM,EAAE6jF,GAAG,CAAC,WAAWlkE,GAAG,YAAY,MAAM,WAAWikE,EAAE,EAAE,SAASE,GAAG9jF,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,KAAKC,EAAE,IAAI,CAAC,EAAEL,EAAE,EAAE2yE,GAAGvyE,EAAEC,EAAE,CAAC,EAAE,OAAO,EAAE,eAAe,CAAC,EAAE,MAAM,EAAE,UAAU,CAAC,CAAC,CAAC,IAAIwjF,GAAG,CAAC,WAAWhkE,GAAG,YAAY,MAAM,WAAW+jE,EAAE,EAAME,GAAGjO,GAAG91D,GAAGjgB,GAAG,KAAK,MAAMA,CAAC,CAAC,EAAEikF,GAAG,CAAC,WAAWhkE,GAAG,YAAY,MAAM,WAAW+jE,EAAE,EAAME,GAAGnP,GAAG,CAAC/0E,EAAE,IAAIA,GAAG,CAAC,EAAEmkF,GAAG1O,GAAGv1D,GAAGgkE,GAAG,KAAK,MAAM,EAAEE,GAAG,CAAC,WAAWlkE,GAAG,YAAY,MAAM,WAAWikE,EAAE,EAAME,GAAGtO,GAAG51D,GAAGngB,GAAGA,EAAE,EAAE,EAAE,MAAM,EAAEskF,GAAG,CAAC,WAAWnkE,GAAG,YAAY,MAAM,WAAWkkE,EAAE,EAAME,GAAGxP,GAAG,CAAC/0E,EAAE,IAAIA,GAAG,CAAC,EAAEwkF,GAAG/O,GAAGr1D,GAAGmkE,GAAG,KAAK,MAAM,EAAEE,GAAG,CAAC,WAAWrkE,GAAG,YAAY,MAAM,WAAWokE,EAAE,EAAE,SAASE,GAAG1kF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,YAAY,EAAE,KAAK,EAAE,MAAM0B,EAAE,KAAKC,CAAC,EAAE5B,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI4B,EAAE5B,EAAE,MAAM,GAAGqC,EAAET,EAAE,EAAEU,EAAE3C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuC,EAAEgkB,EAAE,cAAcvmB,EAAE,KAAK,EAAEwC,EAAE,IAAI,aAAaD,CAAC,EAAE,SAASE,EAAEkB,EAAE,CAAC,IAAI,EAAEA,EAAE/B,EAAEoF,EAAErD,EAAE,EAAE,KAAK,IAAI,EAAE,EAAE,CAAC,EAAE,EAAEA,EAAE,EAAE,KAAK,IAAI,EAAE,EAAEtB,CAAC,EAAE+B,EAAE,EAAE,KAAK4C,GAAG,EAAEA,IAAI,CAAC,IAAI3C,EAAE/B,EAAE0E,GAAG5C,GAAGC,EAAEA,CAAC,CAAC,OAAOD,CAAC,CAAC,QAAQT,EAAE,EAAEA,EAAEpB,EAAEoB,IAAI,CAAC,IAAI,EAAElB,EAAEkB,CAAC,EAAEqD,EAAE1E,EAAEqB,GAAG,KAAK,IAAI,EAAEjC,EAAE,EAAE,CAACC,CAAC,EAAEa,EAAEmB,GAAGqD,CAAC,CAAC,OAAOrH,EAAE,eAAeK,EAAE,MAAMA,EAAE,MAAMwC,CAAC,CAAC,CAAC,IAAI4hF,GAAG,CAAC,WAAWnkE,GAAG,YAAY,MAAM,WAAWkkE,EAAE,EAAE,SAASE,GAAG5kF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,YAAY0B,EAAE,KAAKC,EAAE,MAAMC,EAAE,KAAKS,CAAC,EAAEtC,EAAEmxE,GAAG,EAAE,SAAS,EAAE,IAAI5uE,EAAEikB,EAAE,cAAc,EAAE,KAAK,EAAEhkB,EAAE,EAAE,MAAM,GAAGC,EAAE7C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO8C,EAAE9C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2D,EAAEhE,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,aAAa2C,CAAC,EAAE0E,EAAE1E,EAAE,QAAQ,EAAE,EAAE,EAAE0E,EAAE,IAAI,CAAC,IAAI5C,EAAE,EAAE7B,EAAE8B,EAAE,EAAED,EAAE,KAAK,IAAI,EAAEA,EAAE1C,CAAC,EAAE4C,EAAE,EAAEF,EAAE,KAAK,IAAI7B,EAAE6B,EAAE1C,EAAE,CAAC,EAAE6C,EAAE,EAAE,QAAQ0C,EAAE5C,EAAE4C,EAAE3C,EAAE2C,IAAI1C,GAAG,KAAK,IAAI9B,EAAEwE,GAAG,CAAC,EAAE1C,EAAE3C,EAAE2C,EAAE5C,EAAE,QAAQsF,EAAE5C,EAAE4C,EAAE3C,EAAE2C,IAAI,CAAC,IAAIxC,EAAE,GAAG7C,EAAES,EAAEI,EAAEwE,GAAGtD,EAAE,GAAGY,EAAE,IAAI0C,IAAIxC,GAAG,KAAK,IAAIF,EAAE,CAAClC,CAAC,GAAGoC,GAAGjC,EAAE,GAAG,EAAEyE,IAAIxC,CAAC,CAAC,CAAC,OAAO9E,EAAE,eAAe,EAAE,MAAMK,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIskF,GAAG,CAAC,WAAWpkE,GAAG,YAAY,MAAM,WAAWmkE,EAAE,EAAE,SAASE,GAAG9kF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,iBAAiB,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE/B,EAAEgC,EAAE3B,EAAE,MAAM4B,EAAED,EAAE,OAAOU,EAAEkkB,EAAE,eAAe,EAAE5kB,CAAC,EAAEW,EAAED,EAAEE,EAAE03C,EAAE,mBAAmB33C,EAAEV,CAAC,EAAEY,EAAEd,EAAE,KAAK,IAAI1B,EAAE,MAAM,EAAE,OAAO,GAAGuC,GAAG,KAAK,CAAC,IAAI8B,EAAE,IAAI,MAAMzC,CAAC,EAAE,QAAQ0C,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAID,EAAEC,GAAG3C,EAAEY,EAAE+B,IAAI9B,EAAE4xE,GAAG5xE,EAAEb,EAAE3B,EAAE,MAAMuC,EAAE8B,CAAC,EAAE/B,EAAE23C,EAAE,iBAAiB33C,EAAE,OAAOV,CAAC,EAAED,EAAE0C,CAAC,CAAC6sE,GAAGlxE,EAAE,KAAK,EAAEi6C,EAAE,2BAA2B,MAAM33C,EAAEV,CAAC,EAAE,GAAG,CAACa,EAAEkB,CAAC,EAAEs2C,EAAE,0BAA0Bt4C,EAAEW,CAAC,EAAE,EAAEikB,EAAE,cAAc5iB,CAAC,EAAEqD,EAAEwrE,GAAGhwE,EAAE,EAAEC,EAAEzC,EAAE,KAAK,EAAE,EAAE0B,EAAE,MAAMsF,EAAEvE,EAAEzC,EAAE,KAAK,EAAEoE,EAAE3B,EAAE,OAAO,IAAI2B,EAAE61C,EAAE,qBAAqBx3C,EAAEJ,CAAC,GAAG,CAAC,OAAO,EAAE,MAAM+B,EAAE,MAAMpE,EAAE,KAAK,CAAC,CAAC,IAAIwkF,GAAG,CAAC,WAAWrkE,GAAG,YAAY,MAAM,WAAWokE,EAAE,EAAE,SAASE,GAAGhlF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAEkxE,GAAGlxE,EAAE,SAAS,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,CAAC,EAAE5B,EAAE6B,EAAE,EAAE2kB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAEr4C,CAAC,EAAE,IAAI,wEAAwE,oBAAoBA,IAAI,EAAE,IAAIS,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE4B,EAAEF,EAAEC,CAAC,EAAEW,EAAE,GAAGD,EAAE,cAAc,GAAGA,EAAE,eAAe,GAAGkkB,EAAE,YAAYlkB,EAAE,QAAQA,EAAE,QAAQ,EAAEC,EAAEsyE,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,MAAM,CAAC,IAAI4C,EAAE5C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOwC,EAAE+jB,EAAE,eAAevmB,EAAE,KAAK,EAAEyC,EAAE45E,GAAG95E,EAAEvC,EAAE,MAAMA,EAAE,MAAMwC,EAAEH,EAAE,KAAK,EAAEC,EAAE3C,EAAE,eAAe0C,EAAE,SAASrC,EAAE,MAAMyC,EAAE,MAAM,CAAC,CAAC,OAAOH,CAAC,CAAC,IAAIoiF,GAAG,CAAC,WAAWrkE,GAAG,YAAY,MAAM,WAAWokE,EAAE,EAAE,SAASE,GAAGllF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,EAAE,WAAWC,CAAC,EAAE7B,EAAEmxE,GAAGlxE,EAAE,WAAW,EAAE,IAAIqC,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,EAAEU,EAAE3C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuC,EAAEg6E,GAAGj6E,EAAEtC,EAAE,MAAMA,EAAE,MAAMumB,EAAE,eAAevmB,EAAE,KAAK,EAAEqC,EAAE,KAAK,EAAE,OAAO1C,EAAE,eAAe4C,EAAE,MAAM,UAAUA,EAAE,MAAM,CAAC,CAAC,IAAIqiF,GAAG,CAAC,WAAWrkE,GAAG,YAAY,MAAM,WAAWokE,EAAE,EAAE,SAASE,GAAGplF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,gBAAgBC,CAAC,EAAE7B,EAAEmxE,GAAG,CAAClxE,EAAE,CAAC,EAAE,eAAe,EAAE,IAAIqC,EAAE43C,EAAE,kBAAkB,EAAE,MAAM,EAAEv4C,EAAE,EAAEC,EAAEC,CAAC,EAAEU,EAAE3C,EAAE,WAAW,CAAC,EAAE4C,EAAEi6E,GAAGl6E,EAAED,CAAC,EAAEG,EAAEH,EAAE,YAAYI,EAAEJ,EAAE,aAAasB,EAAEtB,EAAE,YAAY,EAAEA,EAAE,cAAc2E,EAAE3E,EAAE,eAAe,EAAEA,EAAE,cAAc+B,EAAE/B,EAAE,qBAAqBgC,EAAEhC,EAAE,sBAAsBiC,EAAEjC,EAAE,qBAAqBkC,EAAEH,EAAE,EAAE/B,EAAE,QAAQ,MAAM4E,EAAE3C,EAAE,EAAEjC,EAAE,QAAQ,KAAKoC,EAAEJ,EAAE,EAAEhC,EAAE,QAAQ,IAAI6E,EAAE8oB,GAAG,EAAE,MAAM,SAAS,EAAEtrB,EAAE/E,EAAE,WAAWK,CAAC,EAAE,QAAQ2E,EAAE,EAAEA,EAAEtC,EAAE,UAAU,EAAEsC,EAAE,QAAQC,EAAE,EAAEA,EAAEvC,EAAE,WAAW,EAAEuC,EAAE,QAAQC,EAAE,EAAEA,EAAExC,EAAE,QAAQ,EAAEwC,EAAE,QAAQC,EAAE,EAAEA,EAAEzC,EAAE,SAAS,EAAEyC,EAAE,QAAQC,EAAE,EAAEA,EAAE1C,EAAE,QAAQ,EAAE0C,EAAE,CAAC,IAAI2C,EAAE7C,EAAEN,EAAE6C,EAAEtC,EAAEL,EAAEkD,EAAE5C,EAAEkC,EAAEW,EAAG,EAAE,QAAQC,EAAG,EAAEA,EAAGzD,EAAEyD,GAAI,EAAE,CAAC,IAAIC,IAAIJ,EAAEG,GAAIrF,EAAE,GAAG,EAAEsF,GAAG,GAAGA,IAAIzF,EAAE,UAAU,KAAK,MAAMyF,EAAE,IAAIA,IAAI,QAAQC,GAAG,EAAEA,GAAG1D,EAAE0D,IAAIf,EAAE,CAAC,IAAIgB,IAAIZ,EAAEW,IAAItF,EAAE,GAAG,EAAEuF,GAAG,GAAGA,IAAI3F,EAAE,WAAW,KAAK,MAAM2F,EAAE,IAAIA,IAAI,QAAQ5B,GAAG,EAAEA,GAAG9B,EAAE8B,IAAI,EAAE,CAAC,IAAI8B,IAAIP,EAAEvB,IAAIzC,EAAE,GAAGuE,GAAG,GAAGA,IAAI7F,EAAE,UAAU,KAAK,MAAM6F,EAAE,IAAIA,GAAG,SAAS,IAAIC,GAAG/D,EAAEC,EAAEC,EAAE,EAAE/B,EAAE,IAAIoC,EAAEmD,GAAGE,GAAGE,GAAGtD,CAAC,EAAEwD,GAAGP,EAAGxD,EAAEC,EAAEyD,GAAGzD,EAAE8B,GAAGiC,GAAGF,KAAKC,GAAG,EAAE,EAAKC,KAAK,IAAWT,GAAIlD,EAAE,IAAIC,EAAEmD,GAAGE,GAAGE,GAAGtD,CAAC,EAAEyD,GAAE,CAAC,CAAC,CAACnB,EAAE,IAAIU,EAAGjD,EAAEE,EAAEC,EAAEC,EAAEH,CAAC,CAAC,CAAC,OAAOjF,EAAE,eAAeuH,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI49E,GAAG,CAAC,WAAWtkE,GAAG,YAAY,MAAM,WAAWqkE,EAAE,EAAE,SAASE,GAAGtlF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,EAAE0B,EAAE,EAAEwvE,GAAG,CAAC,EAAE,CAAC,EAAE,aAAa,EAAE,GAAG,CAAC,WAAWvvE,EAAE,QAAQC,EAAE,IAAIS,EAAE,gBAAgBC,CAAC,EAAEvC,EAAEwC,EAAE03C,EAAE,kBAAkBv4C,EAAE,MAAMC,EAAEC,EAAE,EAAES,EAAEC,CAAC,EAAEE,EAAE7C,EAAE,KAAK,IAAI+B,EAAE,MAAM,EAAE,OAAOe,EAAEutB,GAAGztB,EAAE,SAASb,EAAE,MAAM46E,GAAG95E,EAAEd,EAAE,MAAMA,EAAE,MAAMa,CAAC,EAAE,MAAM,EAAEoB,EAAEpB,EAAE,aAAa,EAAEA,EAAE,YAAYyE,EAAEzE,EAAE,eAAe,EAAEA,EAAE,cAAc6B,EAAE7B,EAAE,sBAAsB8B,EAAE9B,EAAE,qBAAqB+B,EAAED,EAAE,EAAE9B,EAAE,QAAQ,KAAKgC,EAAEH,EAAE,EAAE7B,EAAE,QAAQ,IAAI0E,EAAE+oB,GAAGtuB,EAAE,MAAM,SAAS,EAAE+C,EAAE9E,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOkH,EAAE8oB,GAAGhwB,EAAE,MAAM,UAAUyE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEnC,EAAE,UAAU,EAAEmC,EAAE,QAAQC,EAAE,EAAEA,EAAEpC,EAAE,WAAW,EAAEoC,EAAE,QAAQC,EAAE,EAAEA,EAAErC,EAAE,SAAS,EAAEqC,EAAE,QAAQC,EAAE,EAAEA,EAAEtC,EAAE,QAAQ,EAAEsC,EAAE,CAAC,IAAIC,EAAEF,EAAEL,EAAEQ,EAAEF,EAAEP,EAAEoD,EAAE,EAAE,QAAQN,EAAE,EAAEA,EAAEhD,EAAEgD,GAAGJ,EAAE,CAAC,IAAIW,GAAG7C,EAAEsC,GAAGzD,EAAE,GAAG,EAAEgE,EAAE,GAAGA,GAAGpF,EAAE,WAAW,KAAK,MAAMoF,CAAC,IAAIA,GAAG,QAAQC,EAAG,EAAEA,EAAGvD,EAAEuD,GAAI,EAAE,CAAC,IAAIC,GAAI9C,EAAE6C,GAAI,EAAE,GAAGC,EAAG,GAAGA,GAAItF,EAAE,UAAU,KAAK,MAAMsF,CAAE,IAAIA,EAAG,SAAS,IAAIC,GAAG1D,EAAEC,EAAE,EAAE5B,EAAE,IAAIiC,EAAEiD,EAAEE,EAAGlD,CAAC,EAAEoD,GAAGX,EAAE/C,EAAEuD,EAAGI,GAAGF,KAAKC,GAAG,EAAE,EAAKC,KAAK,IAAWN,GAAGR,EAAE,IAAIxC,EAAEiD,EAAEE,EAAGlD,CAAC,EAAEqD,GAAE,CAAC,CAACf,EAAE,IAAIS,EAAEhD,EAAEE,EAAEC,EAAEF,CAAC,CAAC,CAAC,OAAOhF,EAAE,eAAesH,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI+9E,GAAG,CAAC,WAAW1kE,GAAG,YAAY,MAAM,WAAWykE,EAAE,EAAE,SAASE,GAAGxlF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEumB,EAAE,eAAe,CAAC,EAAE,EAAE81D,GAAG58E,EAAE,EAAEE,EAAE,EAAEK,EAAE,KAAK,EAAE0B,EAAE46E,GAAG78E,EAAE,EAAEE,EAAEK,EAAE,GAAGD,CAAC,EAAE,MAAM,CAAC,EAAE,OAAO2B,EAAE,MAAM,CAAC,CAAC,IAAIwjF,GAAG,CAAC,WAAWzkE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAOhhB,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAE,CAAC,WAAWO,EAAE,QAAQ,EAAE,IAAI,EAAE,oBAAoB0B,CAAC,EAAE,EAAEC,EAAEhC,EAAEuxE,GAAGnxE,EAAE,mBAAmB,EAAE,IAAI6B,EAAED,EAAE,KAAK,IAAI5B,EAAE,MAAM,EAAE,OAAOsC,EAAE43C,EAAE,kBAAkBl6C,EAAE,MAAMC,EAAE,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAACsC,EAAEC,CAAC,EAAE0iF,GAAGrjF,EAAE7B,EAAE,MAAMA,EAAE,MAAM2B,EAAEW,CAAC,EAAEG,EAAEb,EAAE,MAAMW,EAAED,EAAE,SAAStC,EAAE,KAAK,EAAE0C,EAAEd,EAAE,MAAMY,EAAEF,EAAE,SAAStC,EAAE,KAAK,EAAE,MAAM,CAAC,CAAC,OAAOyC,EAAE,MAAMH,EAAE,SAAS,MAAMtC,EAAE,KAAK,EAAE,CAAC,OAAO0C,EAAE,MAAMJ,EAAE,SAAS,MAAM,OAAO,CAAC,CAAC,CAAC,EAAE,SAAS8iF,GAAG1lF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,EAAEi6C,EAAE,0BAA0Bj6C,EAAE,MAAM0B,CAAC,EAAE,GAAGW,EAAEkkB,EAAE,cAAc,CAAC,EAAE,EAAE,CAAC,EAAEhkB,EAAE5C,EAAE,eAAe,CAAC,EAAE,UAAU,IAAI,aAAa,CAAC0C,CAAC,CAAC,CAAC,EAAE,EAAE,KAAKE,CAAC,EAAE,IAAIC,EAAEwyE,GAAG,CAAC,OAAO,CAAC,EAAEh1E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAE,EAAE,KAAK6C,CAAC,EAAE,IAAIC,EAAE8+E,GAAG,CAAC,OAAO,CAAC,EAAE/+E,EAAE,EAAED,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE,EAAE,KAAK8C,CAAC,EAAE,IAAIkB,EAAE28E,GAAG,CAAC,OAAO,CAAC,EAAE79E,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,QAAQiE,GAAGjE,EAAE,8BAA8BiE,CAAC,CAAC,EAAED,CAAC,CAAC,IAAIyhF,GAAG,CAAC,WAAW1kE,GAAG,YAAY,MAAM,WAAWykE,EAAE,EAAE,SAASE,GAAG5lF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI0B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE2B,EAAED,EAAEE,EAAEq4C,EAAE,mBAAmBt4C,EAAE3B,EAAE,MAAM,MAAM,EAAEqC,EAAErC,EAAE4B,GAAG,OAAOS,EAAEo1E,GAAG,CAAC,OAAO,CAAC,EAAEz3E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAED,EAAEs4C,EAAE,iBAAiBt4C,EAAE,OAAO3B,EAAE,MAAM,MAAM,GAAGi6C,EAAE,2BAA2B,MAAMt4C,EAAEU,EAAE,MAAM,MAAM,EAAE,GAAG,CAACC,EAAEC,CAAC,EAAE03C,EAAE,0BAA0B53C,EAAE,MAAMV,CAAC,EAAEa,EAAE+jB,EAAE,cAAchkB,CAAC,EAAEE,EAAE8jB,EAAE,oBAAoBA,EAAE,cAAcjkB,CAAC,EAAED,EAAE,KAAK,EAAEsB,EAAEhE,EAAE,KAAK,IAAI0C,EAAE,MAAM,EAAE,OAAO,QAAQ2E,EAAE,EAAEA,EAAEvE,EAAE,OAAO,EAAEuE,EAAE,CAAC,IAAI,EAAEA,EAAExE,EAAE4B,EAAET,EAAE,GAAG,QAAQU,EAAE,EAAEA,EAAE7B,EAAE,EAAE6B,EAAE,CAAC,IAAIC,EAAEX,EAAE,EAAEU,IAAI,OAAO,MAAMC,CAAC,GAAGA,EAAEF,KAAKA,EAAEE,EAAE,CAAC7B,EAAEuE,GAAG5C,CAAC,CAACxC,GAAG,MAAMjC,EAAE,8BAA8B0C,CAAC,EAAE,IAAI,EAAE1C,EAAE,eAAe2C,EAAED,EAAE,MAAMI,CAAC,EAAE,GAAG,EAAE,CAAC,IAAIuE,EAAEizC,EAAE,qBAAqB33C,EAAEZ,CAAC,EAAE,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ36E,EAAE,MAAM,CAAC,MAAMqH,CAAC,CAAC,CAAC,EAAE,OAAOrH,EAAE,8BAA8B,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI2lF,GAAG,CAAC,WAAW3kE,GAAG,YAAY,MAAM,WAAW0kE,EAAE,EAAE,SAASE,GAAG9lF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,SAAS,EAAE,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,WAAW,EAAE,IAAI0B,EAAE,EAAE,IAAI,CAAC0C,EAAEC,IAAID,EAAE,GAAGpE,EAAE,MAAMqE,GAAGD,EAAE,EAAE,EAAEzC,EAAE,EAAE,IAAIyC,GAAGA,EAAE,EAAE,EAAExC,EAAE,EAAE,IAAI,CAACwC,EAAEC,IAAID,EAAE,GAAGpE,EAAE,MAAMqE,EAAE,EAAEhC,EAAE,IAAI,UAAU,EAAE,EAAEC,EAAE3C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOuC,EAAEvC,EAAE,MAAM,OAAOwC,EAAE+jB,EAAE,eAAevmB,EAAE,KAAK,EAAEyC,EAAE8jB,EAAE,cAAc7kB,CAAC,EAAEiC,EAAEjC,EAAE,OAAO,EAAE6kB,EAAE,eAAe7kB,CAAC,EAAEsF,EAAEuf,EAAE,uBAAuBvmB,EAAE,MAAMyC,CAAC,EAAE,QAAQ2B,EAAE,EAAEA,EAAE3B,EAAE2B,IAAI,CAAC,IAAIC,EAAEkiB,EAAE,WAAWniB,EAAET,EAAE,CAAC,EAAE,QAAQY,EAAE,EAAEA,EAAEZ,EAAEY,IAAIF,EAAEE,GAAG5C,EAAE4C,GAAGF,EAAEE,GAAG5C,EAAE4C,GAAG,EAAEF,EAAEE,GAAGlC,EAAEgC,EAAEE,IAAI3C,EAAE2C,KAAKF,EAAEE,IAAI3C,EAAE2C,GAAG,GAAG,EAAEF,EAAEE,GAAGlC,GAAGgC,EAAEA,EAAE,IAAI,CAACE,EAAE0C,IAAI1C,EAAE5C,EAAEsF,EAAE,EAAE,IAAI3C,EAAEiiB,EAAE,WAAWliB,EAAE9B,EAAEC,CAAC,EAAEwE,EAAE5C,GAAG9B,EAAEgC,EAAE,CAAC,MAAM,CAAC,OAAO3E,EAAE,MAAMqH,EAAEtF,EAAE1B,EAAE,KAAK,EAAE,MAAM0B,EAAE,MAAM1B,EAAE,KAAK,CAAC,CAAC,IAAIwlF,GAAG,CAAC,WAAW3kE,GAAG,YAAY,MAAM,WAAW0kE,EAAE,EAAME,GAAGjR,GAAG,CAAC/0E,EAAE,IAAI,CAAC,IAAIE,EAAEF,EAAE,EAAE,OAAOA,EAAE,GAAG,EAAE,GAAGA,GAAG,GAAG,GAAG,EAAEE,GAAGA,EAAE,GAAG,CAAC,CAAC,EAAE+lF,GAAGxQ,GAAGp0D,GAAG2kE,EAAE,EAAEE,GAAG,CAAC,WAAW7kE,GAAG,YAAY,MAAM,WAAW4kE,EAAE,EAAME,GAAG3lF,GAAG+E,GAAG,CAAC,EAAE,SAAS6gF,GAAGpmF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,IAAI,CAAC,EAAED,EAAE,EAAEC,EAAE,MAAM,OAAO0B,EAAE,EAAE,GAAGA,IAAI,KAAKA,EAAE,EAAE,GAAGA,IAAI,EAAE,EAAE,MAAM,MAAM,4EAA4E,iBAAiBA,GAAG,EAAE,IAAIC,EAAE4kB,EAAE,eAAe,CAAC7kB,CAAC,EAAE1B,EAAE,KAAK,EAAE4B,EAAE2iF,GAAG,CAAC,OAAO,CAAC,EAAEvkF,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,iBAAiBgC,EAAE,SAAS,EAAE,CAAC,CAAC,EAAEU,EAAE43C,EAAE,qBAAqBr4C,EAAE,MAAMD,CAAC,EAAEW,EAAEg4E,GAAG,CAAC,OAAO,CAAC,EAAE14E,CAAC,EAAE,QAAQjC,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAEE,EAAE+2E,GAAG,CAAC,OAAO,CAAC,EAAEt5E,EAAE,EAAEsC,CAAC,EAAE,QAAQ3C,CAAC,CAAC,EAAE6C,EAAEszE,GAAG,CAAC,OAAO,CAAC,EAAEvzE,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE8C,EAAE69E,GAAG,CAAC,OAAO,CAAC,EAAE99E,CAAC,EAAE,QAAQ7C,EAAE,MAAM,CAAC,KAAKgC,EAAE,SAAS,EAAE,CAAC,CAAC,EAAEgC,EAAE22E,GAAG,CAAC,OAAO,CAAC,EAAE73E,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAE,EAAEk/E,GAAG,CAAC,OAAO,CAAC,EAAE/+E,EAAE,EAAEmB,CAAC,EAAE,QAAQhE,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BiC,CAAC,EAAEjC,EAAE,8BAA8B2C,CAAC,EAAE3C,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAE9C,EAAE,8BAA8BgE,CAAC,EAAE,CAAC,CAAC,IAAImiF,GAAG,CAAC,WAAWliE,GAAG,YAAY,MAAM,WAAWiiE,EAAE,EAAE,SAASE,GAAGtmF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,KAAK,EAAE,WAAW0B,CAAC,EAAE3B,EAAEmxE,GAAGlxE,EAAE,aAAa,EAAE,IAAI2B,EAAED,EAAE1B,EAAE6lF,GAAG,CAAC,OAAO,CAAC,OAAO7lF,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC,EAAEiC,EAAED,EAAE,MAAM,GAAGU,EAAEV,EAAE,MAAM,GAAGW,EAAE3C,EAAE,KAAK,IAAIgC,EAAE,MAAM,EAAE,OAAOY,EAAE,CAACX,EAAE,CAAC,EAAEY,EAAE+jB,EAAE,oBAAoBA,EAAE,cAAchkB,CAAC,EAAE,OAAO,EAAE,QAAQE,EAAE,EAAEA,EAAEb,EAAE,EAAEa,EAAE,CAAC,IAAIkB,EAAElB,EAAEJ,EAAE,EAAE,IAAI,aAAaA,EAAE,CAAC,EAAE,EAAE,GAAGC,EAAEqB,GAAG,QAAQS,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,EAAEA,GAAG,EAAEA,EAAE,GAAG9B,EAAEqB,EAAES,GAAG,IAAI4C,EAAE4+E,GAAG,KAAK,EAAE,SAAS,CAAC,EAAE,EAAEnjF,EAAE,EAAE,QAAQ2B,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIC,EAAE2C,EAAE,EAAExE,EAAE,EAAE4B,GAAG,EAAE,OAAO,QAAQE,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,GAAGD,EAAE,EAAEC,GAAG,CAAC9B,EAAE,EAAE4B,GAAGE,EAAE,KAAK,CAAC,CAAC,CAAC,OAAO5C,GAAG/B,EAAE,8BAA8BgC,CAAC,EAAEhC,EAAE,eAAe4C,EAAE,QAAQC,CAAC,CAAC,CAAC,IAAIwjF,GAAG,CAAC,WAAWjlE,GAAG,YAAY,MAAM,WAAWglE,EAAE,EAAME,GAAGnoC,GAAG,wBAAwB,SAASooC,GAAGzmF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,CAAC,EAAE5B,EAAEmxE,GAAGlxE,EAAE,mBAAmB,EAAE,IAAI4B,EAAEjC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOqC,EAAE1C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,gBAAgB2C,CAAC,EAAE2jF,GAAGrkF,EAAES,EAAE,EAAEX,EAAEC,CAAC,EAAE,OAAOhC,EAAE,eAAe,CAAC2C,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAWA,CAAC,CAAC,CAAC,CAAC,IAAI6jF,GAAG,CAAC,WAAWhlE,GAAG,YAAY,MAAM,WAAW+kE,EAAE,EAAME,GAAGtoC,GAAG,wBAAwB,SAASuoC,GAAG5mF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,EAAE,mBAAmBC,CAAC,EAAE7B,EAAEmxE,GAAGlxE,EAAE,yBAAyB,EAAE,IAAIqC,EAAE1C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOsC,EAAE3C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,gBAAgB4C,EAAE,aAAaC,CAAC,EAAE4jF,GAAG/jF,EAAEC,EAAE,EAAEZ,EAAEC,EAAEC,CAAC,EAAE,MAAM,CAACjC,EAAE,eAAe,CAAC4C,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAWA,CAAC,CAAC,EAAE5C,EAAE,eAAe,CAAC,EAAE,QAAQ,IAAI,WAAW,CAAC6C,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI8jF,GAAG,CAAC,WAAWllE,GAAG,YAAY,MAAM,WAAWilE,EAAE,EAAME,GAAGzoC,GAAG,wBAAwB,SAAS0oC,GAAG/mF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,EAAE,aAAaC,CAAC,EAAE7B,EAAEmxE,GAAGlxE,EAAE,4BAA4B,EAAE,IAAIqC,EAAE1C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOsC,EAAE3C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO4C,EAAE,EAAEC,EAAEd,EAAEe,EAAEd,EAAEgC,EAAE/B,EAAE,CAAC,gBAAgB,EAAE,eAAeoF,CAAC,EAAEu/E,GAAGlkF,EAAEC,EAAEC,EAAEC,EAAEC,EAAEkB,CAAC,EAAE,MAAM,CAAChE,EAAE,eAAe,CAAC,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAW,CAAC,CAAC,EAAEA,EAAE,eAAe,CAACqH,EAAE,MAAM,EAAE,UAAU,IAAI,aAAaA,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIy/E,GAAG,CAAC,WAAWplE,GAAG,YAAY,MAAM,WAAWmlE,EAAE,EAAE,SAASE,GAAGjnF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ0B,EAAE,SAASC,CAAC,EAAE5B,EAAEmxE,GAAGlxE,EAAE,QAAQ,EAAE,IAAI4B,EAAE2kB,EAAE,cAAcvmB,EAAE,KAAK,EAAEqC,EAAE,IAAI,aAAaT,EAAE,CAAC,EAAES,EAAE,KAAKV,CAAC,EAAE,IAAIW,EAAE3C,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,QAAQuC,EAAE,EAAEA,EAAEX,EAAE,EAAEW,EAAED,EAAEC,IAAI,GAAGD,EAAEC,GAAG,IAAIF,EAAEE,EAAE,EAAED,EAAEC,IAAIb,GAAG,OAAO/B,EAAE,eAAe,CAAC,GAAGK,EAAE,MAAM,CAAC,EAAE,EAAEqC,CAAC,CAAC,CAAC,IAAIskF,GAAG,CAAC,WAAWplE,GAAG,YAAY,MAAM,WAAWmlE,EAAE,EAAE,SAASE,GAAGnnF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGA,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,+CAA+C,EAAE,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAIC,EAAE80E,GAAG,CAAC,OAAO,CAAC,MAAM/0E,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE,EAAEinF,GAAG,CAAC,OAAO,CAAC,EAAE5mF,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,EAAEk+E,GAAG,CAAC,OAAO,CAAC,MAAM99E,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE+B,EAAEklF,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQjnF,CAAC,CAAC,EAAEgC,EAAE8yE,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK/yE,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BK,CAAC,EAAEL,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B+B,CAAC,EAAEC,CAAC,KAAM,QAAOqgF,GAAG,CAAC,QAAQriF,EAAE,MAAM,CAAC,MAAMI,EAAE,MAAM,MAAM,EAAE,MAAMA,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI8mF,GAAG,CAAC,WAAW1hE,GAAG,YAAY,MAAM,WAAWyhE,EAAE,EAAE,SAASE,GAAGrnF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGA,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,8CAA8C,EAAE,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAIC,EAAE80E,GAAG,CAAC,OAAO,CAAC,MAAM/0E,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE,EAAEmnF,GAAG,CAAC,OAAO,CAAC,EAAE9mF,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,EAAEk+E,GAAG,CAAC,OAAO,CAAC,MAAM99E,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE+B,EAAEklF,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQjnF,CAAC,CAAC,EAAEgC,EAAE8yE,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK/yE,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BK,CAAC,EAAEL,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B+B,CAAC,EAAEC,CAAC,KAAM,QAAOqgF,GAAG,CAAC,QAAQriF,EAAE,MAAM,CAAC,MAAMI,EAAE,MAAM,MAAM,EAAE,MAAMA,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAIgnF,GAAG,CAAC,WAAWzlE,GAAG,YAAY,MAAM,WAAWwlE,EAAE,EAAE,SAASE,GAAGvnF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,GAAG,EAAE,SAAS,EAAE,OAAOqhF,GAAG,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,EAAE,QAAQzhF,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,GAAG,MAAM,EAAE,EAAE,GAAG,MAAM,EAAE,QAAQqC,GAAG,CAACkkB,EAAE,kBAAkB,EAAElkB,EAAE,MAAM,uDAAuD,EAAEkkB,EAAE,OAAO,IAAIlkB,EAAE,MAAM,IAAI,uDAAuD,CAAC,CAAC,EAAE,IAAIX,EAAE,CAAC,EAAEC,EAAE,EAAE,IAAIU,GAAG,CAAC,IAAIC,EAAE8+E,GAAG,CAAC,OAAO,CAAC,MAAM/+E,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,KAAKY,CAAC,EAAEA,CAAC,CAAC,EAAEV,EAAEm8E,GAAG,CAAC,OAAOp8E,EAAE,QAAQhC,EAAE,MAAM,CAAC,KAAKK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,QAAQW,GAAG1C,EAAE,8BAA8B0C,CAAC,CAAC,EAAET,CAAC,CAAC,IAAIqlF,GAAG,CAAC,WAAWzlE,GAAG,YAAY,MAAM,WAAWwlE,EAAE,EAAE,SAASE,GAAGznF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,SAAS,EAAE,cAAc,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,KAAK,EAAE,IAAI0B,EAAE,EAAE,IAAI,CAAC8C,EAAEJ,IAAII,EAAE,GAAGxE,EAAE,MAAMoE,GAAGI,EAAE,EAAE,EAAE7C,EAAE,EAAE,IAAI6C,GAAGA,EAAE,EAAE,EAAE5C,EAAEjC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOqC,EAAEkkB,EAAE,cAAcvmB,EAAE,KAAK,EAAEsC,EAAEtC,EAAE,MAAM,OAAOuC,EAAEgkB,EAAE,eAAevmB,EAAE,KAAK,EAAEwC,EAAE+jB,EAAE,cAAc7kB,CAAC,EAAEe,EAAEf,EAAE,OAAOiC,EAAE4iB,EAAE,eAAe7kB,CAAC,EAAE,EAAE6kB,EAAE,uBAAuBvmB,EAAE,MAAMwC,CAAC,EAAE,IAAI,GAAG,EAAE,KAAK,CAAC,EAAE,QAAQgC,EAAE,EAAEA,EAAEnC,EAAEmC,IAAI,CAAC,IAAIH,EAAEkiB,EAAE,WAAW/hB,EAAElC,EAAEC,CAAC,EAAE,IAAI,CAACgC,EAAE0C,IAAI1C,EAAE5C,EAAEsF,EAAE,EAAE3C,EAAEiiB,EAAE,WAAWliB,EAAE5B,EAAEkB,CAAC,EAAE,EAAEW,GAAG1C,EAAE4C,EAAE,CAAC,MAAM,CAAC,OAAO7E,EAAE,MAAM,EAAE+B,EAAE1B,EAAE,KAAK,EAAE,MAAM0B,EAAE,MAAM1B,EAAE,KAAK,CAAC,CAAC,IAAImnF,GAAG,CAAC,WAAW1lE,GAAG,YAAY,MAAM,WAAWylE,EAAE,EAAME,GAAG5S,GAAG,CAAC/0E,EAAE,IAAI,KAAK,IAAIA,EAAE,CAAC,CAAC,EAAE4nF,GAAGnS,GAAGvzD,GAAGylE,EAAE,EAAEE,GAAG,CAAC,WAAW3lE,GAAG,YAAY,MAAM,WAAW0lE,EAAE,EAAE,SAASE,GAAG9nF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,mBAAmBO,EAAE,kBAAkB,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,iBAAiB0B,CAAC,EAAE3B,EAAE4B,EAAE3B,EAAE,IAAIgH,GAAGrH,EAAE,KAAK,IAAIqH,EAAE,MAAM,EAAE,MAAM,EAAEpF,EAAE5B,EAAE,IAAIgH,GAAGA,EAAE,KAAK,EAAE3E,EAAE1C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO2C,EAAE3C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC4C,EAAEC,EAAEC,CAAC,EAAEswE,GAAGpxE,EAAEC,EAAES,EAAE,EAAE,MAAM,EAAE,MAAMC,EAAE,EAAE,MAAMZ,CAAC,EAAEiC,EAAEpB,EAAE,IAAIyE,GAAGrH,EAAE,eAAe,CAACqH,EAAE,MAAM,EAAE,QAAQA,CAAC,CAAC,EAAE,EAAErH,EAAE,eAAe8C,EAAE,EAAE,MAAMD,CAAC,EAAE,OAAOmB,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI6jF,GAAG,CAAC,WAAW1lE,GAAG,YAAY,MAAM,WAAWylE,EAAE,EAAE,SAASE,GAAGhoF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,OAAOM,EAAE,OAAOC,EAAE,OAAO,CAAC,EAAE,EAAE,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO2B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAACiC,EAAES,CAAC,EAAE2wE,GAAG,EAAEjzE,EAAE,MAAMA,EAAE,MAAM2B,EAAE1B,EAAE,MAAM2B,EAAE,EAAE,KAAK,EAAEW,EAAE3C,EAAE,eAAe,CAACiC,EAAE,MAAM,EAAE,QAAQA,CAAC,EAAEW,EAAE5C,EAAE,eAAe,CAAC0C,EAAE,MAAM,EAAEtC,EAAE,MAAMsC,CAAC,EAAE,MAAM,CAACC,EAAEC,CAAC,CAAC,CAAC,IAAImlF,GAAG,CAAC,WAAW3lE,GAAG,YAAY,MAAM,WAAW0lE,EAAE,EAAE,SAASE,GAAGloF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,EAAE,aAAa,EAAE,oBAAoB0B,CAAC,EAAE,EAAE,CAAC,kBAAkBC,CAAC,EAAE5B,EAAE6B,EAAEjC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAOqC,EAAE1C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO2C,EAAE3C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO4C,EAAEb,EAAE,IAAI,GAAG/B,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE6C,EAAEd,EAAE,IAAI,GAAG,EAAE,KAAK,EAAE,CAACe,EAAEkB,CAAC,EAAEsvE,GAAGrxE,EAAE5B,EAAE,MAAMqC,EAAE,EAAE,MAAM,EAAE,MAAMC,EAAE,EAAE,MAAMC,EAAEC,EAAEb,CAAC,EAAE,OAAOhC,EAAE,eAAe8C,EAAE,EAAE,MAAMkB,CAAC,CAAC,CAAC,IAAIikF,GAAG,CAAC,WAAW5lE,GAAG,YAAY,MAAM,WAAW2lE,EAAE,EAAE,SAASE,GAAGpoF,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,KAAKC,EAAE,MAAM,EAAE,KAAK,CAAC,EAAEL,EAAE+B,EAAEwxE,GAAGnzE,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAO,EAAE,eAAe,CAAC0B,EAAE,MAAM,EAAE,EAAEA,CAAC,CAAC,CAAC,IAAIomF,GAAG,CAAC,WAAW7lE,GAAG,YAAY,MAAM,WAAW4lE,EAAE,EAAME,GAAGvS,GAAGrzD,GAAG1iB,GAAG,EAAEA,CAAC,EAAEuoF,GAAG,CAAC,WAAW7lE,GAAG,YAAY,MAAM,WAAW4lE,EAAE,EAAE,SAASE,GAAGxoF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK0B,CAAC,EAAE3B,EAAEmxE,GAAGlxE,EAAE,gBAAgB,EAAE,IAAI2B,EAAE4kB,EAAE,eAAevmB,EAAE,KAAK,EAAE,CAAC4B,EAAES,CAAC,EAAEX,EAAE,CAACY,EAAEC,EAAEC,EAAEC,CAAC,EAAEzC,EAAE,MAAM2D,EAAEhE,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,aAAaumB,EAAE,cAAc,CAACjkB,EAAEV,EAAES,EAAEI,CAAC,CAAC,CAAC,EAAEuE,EAAE,CAAC,GAAGpF,EAAE,EAAEW,EAAE,EAAEA,EAAE,GAAGF,EAAE,EAAEG,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGZ,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGS,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAE+B,EAAE,EAAEC,EAAE2C,EAAE,GAAG,EAAE,GAAG1C,EAAE0C,EAAE,GAAG,EAAE,GAAG,QAAQzC,EAAE,EAAEA,EAAEjC,EAAEiC,IAAI,QAAQ0C,EAAE,EAAEA,EAAErF,EAAEqF,IAAI,CAAC,IAAIxC,EAAE,EAAEA,EAAEJ,GAAG4C,EAAE,IAAI,GAAGxC,EAAEJ,EAAE4C,EAAE,IAAIC,EAAE,KAAK,IAAI,EAAE,KAAK,MAAMzC,CAAC,CAAC,EAAEC,EAAED,EAAEyC,EAAEvC,EAAE,KAAK,IAAIpC,EAAE,EAAE,KAAK,KAAKkC,CAAC,CAAC,EAAEG,EAAEL,EAAE5C,EAAE,GAAGuF,EAAEvF,EAAE,GAAGkD,EAAEN,EAAE5C,EAAE,GAAGgD,EAAEhD,EAAE,GAAG,QAAQmD,EAAE,EAAEA,EAAEzC,EAAEyC,IAAI,CAAC,IAAIC,EAAE,EAAEA,EAAET,GAAGQ,EAAE,IAAI,GAAGC,EAAET,EAAEQ,EAAE,IAAI4C,EAAE,KAAK,IAAI,EAAE,KAAK,MAAM3C,CAAC,CAAC,EAAEqC,EAAErC,EAAE2C,EAAEC,EAAE,KAAK,IAAInF,EAAE,EAAE,KAAK,KAAKuC,CAAC,CAAC,EAAE6C,EAAGhD,EAAE8C,EAAE/F,EAAE,GAAGkG,EAAGhD,EAAE6C,EAAE/F,EAAE,GAAGmG,GAAGlD,EAAE+C,EAAEhG,EAAE,GAAGoG,GAAGlD,EAAE8C,EAAEhG,EAAE,GAAG,QAAQqG,GAAG,EAAEA,GAAGvF,EAAEuF,KAAK,CAAC,IAAI5B,GAAGzC,EAAEiE,EAAGI,IAAIE,GAAGvE,EAAEkE,EAAGG,IAAIG,GAAGxE,EAAEmE,GAAGE,IAAII,GAAGzE,EAAEoE,GAAGC,IAAIK,GAAGjC,IAAI+B,GAAG/B,IAAIgB,EAAEkB,GAAGJ,IAAIE,GAAGF,IAAId,EAAEmB,GAAGF,IAAIC,GAAGD,IAAI3D,EAAE,EAAEN,KAAKmE,EAAE,CAAC,CAAC,CAAC,OAAO5I,EAAE,eAAe,CAAC2C,EAAEV,EAAES,EAAEI,CAAC,EAAE,UAAU,CAAC,CAAC,CAAC,IAAIylF,GAAG,CAAC,WAAW1lE,GAAG,YAAY,MAAM,WAAWylE,EAAE,EAAE,SAASE,GAAG1oF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,aAAa,CAAC,EAAED,EAAEmxE,GAAG,CAAC,EAAElxE,CAAC,EAAE,oBAAoB,EAAE,IAAI0B,EAAE6kB,EAAE,eAAevmB,EAAE,KAAK,EAAE,CAAC2B,EAAEC,EAAES,EAAEC,CAAC,EAAEtC,EAAE,MAAM,CAAC,CAACuC,EAAEC,CAAC,EAAE,EAAE,MAAMC,EAAE,IAAI,aAAad,EAAEC,EAAES,EAAEC,CAAC,EAAEqB,EAAE,CAAC,GAAGpB,EAAE,EAAEX,EAAE,EAAEA,EAAE,GAAGY,EAAE,EAAEH,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGE,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGC,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAEwE,EAAErD,EAAE,GAAG,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAE,GAAGS,EAAEzE,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0E,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAE3C,EAAE2C,IAAI,CAAC,IAAIC,EAAED,EAAE5C,EAAE,GAAG,QAAQuF,EAAE,EAAEA,EAAE1E,EAAE0E,IAAI,CAAC,IAAIxC,EAAEwC,EAAED,EAAEE,EAAE,KAAK,MAAMzC,CAAC,EAAEC,EAAE,KAAK,IAAI,KAAK,KAAKD,CAAC,EAAE7C,EAAE,CAAC,EAAE+C,EAAEJ,EAAE2C,EAAExF,EAAE,GAAGkD,EAAEL,EAAEG,EAAEhD,EAAE,GAAGmD,EAAEJ,EAAEyC,EAAEpC,EAAE,EAAED,EAAE,QAAQE,EAAE,EAAEA,EAAEvC,EAAEuC,IAAI,CAAC,IAAI2C,EAAE3C,EAAE,EAAEqC,EAAE,KAAK,MAAMM,CAAC,EAAEC,EAAE,KAAK,IAAI,KAAK,KAAKD,CAAC,EAAErF,EAAE,CAAC,EAAEuF,EAAGF,EAAEN,EAAES,EAAG,EAAED,EAAGE,GAAGnD,EAAEyC,EAAE1F,EAAE,GAAGqG,GAAGpD,EAAEgD,EAAEjG,EAAE,GAAGsG,GAAGpD,EAAEwC,EAAE1F,EAAE,GAAG0E,GAAGxB,EAAE+C,EAAEjG,EAAE,GAAGwG,GAAGpD,EAAE+C,EAAGM,GAAGrD,EAAE8C,EAAGQ,GAAGvD,EAAEgD,EAAGQ,GAAGxD,EAAE+C,EAAG,QAAQU,GAAG,EAAEA,GAAGhG,EAAEgG,KAAK,CAAC,IAAIC,GAAGnE,EAAEC,KAAK5B,EAAEqF,GAAGQ,KAAKC,GAAGL,GAAGzF,EAAEsF,GAAGO,KAAKC,GAAGJ,GAAG1F,EAAEuF,GAAGM,KAAKC,GAAGH,GAAG3F,EAAE2D,GAAGkC,KAAKC,GAAGF,EAAE,CAAC,CAAC,CAAC,CAAC,OAAO1I,EAAE,eAAe,CAACgC,EAAEU,EAAET,EAAEU,CAAC,EAAE,UAAUG,CAAC,CAAC,CAAC,IAAI2lF,GAAG,CAAC,WAAW3lE,GAAG,YAAY,MAAM,WAAW0lE,EAAE,EAAE,SAASE,GAAG5oF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK0B,CAAC,EAAE3B,EAAEmxE,GAAGlxE,EAAE,uBAAuB,EAAE,IAAI2B,EAAE4kB,EAAE,eAAevmB,EAAE,KAAK,EAAE,CAAC4B,EAAES,CAAC,EAAEX,EAAE,CAACY,EAAEC,EAAEC,EAAEC,CAAC,EAAEzC,EAAE,MAAM2D,EAAEhE,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,aAAasC,EAAEV,EAAES,EAAEI,CAAC,EAAEuE,EAAE,CAAC,GAAGpF,EAAE,EAAEW,EAAE,EAAEA,EAAE,GAAGF,EAAE,EAAEG,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGZ,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGS,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAE+B,EAAE4C,EAAE,GAAG,EAAE,GAAG3C,EAAE2C,EAAE,GAAG,EAAE,GAAG1C,EAAE,EAAE,QAAQC,EAAE,EAAEA,EAAEjC,EAAEiC,IAAI,CAAC,IAAI0C,EAAE1C,EAAE5C,EAAE,GAAG,QAAQ8C,EAAE,EAAEA,EAAE7C,EAAE6C,IAAI,CAAC,IAAIyC,EAAE,EAAE9C,GAAGK,EAAE,IAAIL,EAAEK,EAAEC,EAAE,KAAK,IAAInC,EAAE,EAAE,EAAE,KAAK,MAAM2E,CAAC,EAAE,KAAK,MAAMA,CAAC,CAAC,EAAE,IAAIxC,EAAE,KAAK,IAAI,EAAEA,CAAC,GAAG,IAAIC,EAAEsC,EAAEvC,EAAE/C,EAAE,GAAG,QAAQiD,EAAE,EAAEA,EAAEvC,EAAEuC,IAAI,CAAC,IAAIC,EAAE,EAAER,GAAGO,EAAE,IAAIP,EAAEO,EAAEE,EAAE,KAAK,IAAItC,EAAE,EAAE,EAAE,KAAK,MAAMqC,CAAC,EAAE,KAAK,MAAMA,CAAC,CAAC,EAAE,IAAIC,EAAE,KAAK,IAAI,EAAEA,CAAC,GAAG,IAAIC,EAAEJ,EAAEG,EAAEnD,EAAE,GAAG,QAAQ+F,EAAE,EAAEA,EAAEjF,EAAEiF,IAAI,CAAC,IAAIN,EAAEzD,EAAEoB,EAAE2C,GAAG,EAAEpD,KAAK8C,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOzH,EAAE,eAAe,CAAC2C,EAAEV,EAAES,EAAEI,CAAC,EAAEzC,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIsoF,GAAG,CAAC,WAAWhmE,GAAG,YAAY,MAAM,WAAW+lE,EAAE,EAAE,SAASE,GAAG9oF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,aAAa,CAAC,EAAED,EAAEmxE,GAAG,CAAC,EAAElxE,CAAC,EAAE,2BAA2B,EAAE,IAAI0B,EAAE6kB,EAAE,eAAevmB,EAAE,KAAK,EAAE2B,EAAE4kB,EAAE,eAAe,EAAE,KAAK,EAAE,CAAC3kB,EAAES,EAAEC,EAAEC,CAAC,EAAEvC,EAAE,MAAM,CAAC,CAACwC,EAAEC,CAAC,EAAE,EAAE,MAAMkB,EAAE,IAAI,aAAa/B,EAAES,EAAEC,EAAEC,CAAC,EAAE,EAAE5C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOqH,EAAE,CAAC,GAAGxE,EAAE,EAAEH,EAAE,EAAEA,EAAE,GAAGI,EAAE,EAAEH,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGE,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGC,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAE2B,EAAE4C,EAAE,GAAG,EAAE,GAAG3C,EAAE2C,EAAE,GAAG,EAAE,GAAG1C,EAAE,EAAEF,EAAEG,EAAE,EAAEF,EAAE4C,EAAE,KAAK,KAAK3C,CAAC,EAAE,EAAE,EAAEG,EAAE,KAAK,KAAKF,CAAC,EAAE,EAAE,EAAE,QAAQ2C,EAAE,EAAEA,EAAEtF,EAAEsF,IAAI,CAAC,IAAIxC,EAAEwC,EAAExF,EAAE,GAAG,QAAQiD,EAAE,EAAEA,EAAEtC,EAAEsC,IAAI,CAAC,IAAIC,EAAEF,EAAEC,EAAEjD,EAAE,GAAGmD,EAAE,KAAK,MAAMF,EAAEL,CAAC,EAAEQ,EAAE,KAAK,MAAMD,EAAEoC,EAAE,CAAC,EAAE,QAAQlC,EAAE,EAAEA,EAAEzC,EAAEyC,IAAI,CAAC,IAAI2C,EAAE9C,EAAEG,EAAErD,EAAE,GAAG0F,EAAE,KAAK,MAAMrC,EAAER,CAAC,EAAEoD,EAAE,KAAK,MAAMP,EAAE3C,EAAE,CAAC,EAAE,QAAQmD,EAAG,EAAEA,EAAGrF,EAAEqF,IAAK,CAAC,IAAIC,EAAG,EAAE,QAAQC,GAAG,EAAEA,GAAGb,EAAEa,KAAK,CAAC,IAAIC,GAAGD,GAAGhD,EAAE,GAAGiD,GAAG,GAAGA,IAAIvF,EAAE,SAAS,IAAIwF,GAAGtD,EAAEqD,GAAGpG,EAAE,GAAGyE,GAAG2B,GAAG3D,EAAE8D,GAAG,KAAK,IAAI7F,EAAE,EAAE,EAAE,KAAK,MAAM+D,EAAE,EAAE,KAAK,MAAMA,EAAE,CAAC,EAAE,GAAGzB,IAAIuD,GAAG,QAAQC,GAAG,EAAEA,GAAG1D,EAAE0D,KAAK,CAAC,IAAIC,GAAGD,GAAGR,EAAE,GAAGS,GAAG,GAAGA,IAAI3F,EAAE,SAAS,IAAI4F,GAAGL,GAAGI,GAAGzG,EAAE,GAAG2G,GAAGF,GAAG/D,EAAEkE,GAAG,KAAK,IAAIjG,EAAE,EAAE,EAAE,KAAK,MAAMgG,EAAE,EAAE,KAAK,MAAMA,EAAE,CAAC,EAAEvD,IAAIwD,KAAKV,GAAI,EAAEQ,GAAGT,GAAI,CAAC,CAACjE,EAAE+D,EAAEE,GAAIC,CAAE,CAAC,CAAC,CAAC,CAAC,OAAOlI,EAAE,eAAeK,EAAE,MAAMA,EAAE,MAAM2D,CAAC,CAAC,CAAC,IAAI6kF,GAAG,CAAC,WAAWjmE,GAAG,YAAY,MAAM,WAAWgmE,EAAE,EAAE,SAASE,GAAGhpF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,SAAS,EAAE,IAAI,EAAEA,EAAE,MAAM,OAAO0B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG,IAAI,EAAE,OAAO40E,GAAG,CAAC,OAAO,CAAC,EAAE50E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAIgC,EAAE,IAAIonB,GAAG/oB,EAAE,MAAMA,EAAE,KAAK,EAAE4B,EAAEjC,EAAE,WAAWK,CAAC,EAAE,QAAQqC,EAAE,EAAEA,EAAEV,EAAE,KAAKU,IAAI,CAAC,IAAIC,EAAEX,EAAE,WAAWU,CAAC,EAAEE,EAAED,EAAE,MAAM,EAAEZ,EAAE,QAAQc,GAAGD,EAAEC,GAAGxC,EAAE,MAAMwC,GAAG,EAAED,EAAEC,EAAE,EAAEb,EAAE,IAAIC,EAAE,IAAI,GAAGW,CAAC,EAAE,GAAGD,CAAC,CAAC,CAAC,OAAO3C,EAAE,eAAegC,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI+mF,GAAG,CAAC,WAAW/lE,GAAG,YAAY,MAAM,WAAW8lE,EAAE,EAAME,GAAG,CAAC,WAAWrjE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAO7lB,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,EAAE,UAAU,EAAE,OAAO,CAAC,EAAE,EAAE0B,EAAE/B,EAAEgC,EAAE4kB,EAAE,uBAAuBxmB,EAAE,MAAMwmB,EAAE,cAAcxmB,EAAE,KAAK,CAAC,EAAE,CAAC6B,EAAES,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAM,CAACyC,EAAEC,CAAC,EAAEw3C,EAAE,eAAe,EAAE53C,EAAEC,CAAC,EAAEqB,EAAE,IAAI,EAAE,KAAK,IAAI3D,CAAC,EAAEgH,EAAE,KAAK,IAAIhH,CAAC,EAAE,EAAE0B,EAAE,KAAK,IAAI3B,EAAE,MAAM,EAAE,OAAO,QAAQsE,EAAE,EAAEA,EAAEzC,EAAEyC,IAAI,CAAC,IAAIC,EAAED,EAAE/B,EAAED,EAAEE,EAAE,QAAQgC,EAAE,EAAEA,EAAElC,EAAEkC,IAAI,CAAC,IAAI0C,EAAE1C,GAAGjC,EAAEC,GAAG,QAAQkC,EAAE,EAAEA,EAAEnC,EAAEmC,IAAI,CAAC,IAAIyC,EAAEzC,EAAElC,EAAE,QAAQmC,EAAE,EAAEA,EAAEnC,EAAEmC,IAAI,CAAC,IAAIC,EAAE,CAAC/C,EAAE2C,EAAEE,EAAEC,CAAC,EAAEE,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAGG,GAAGF,EAAEpC,GAAGwE,GAAGnC,EAAEpC,GAAG,EAAEsC,GAAGH,EAAEpC,GAAG,GAAGqC,EAAEpC,GAAGuE,EAAElC,EAAE,KAAK,MAAMA,EAAEtC,CAAC,EAAEuC,EAAE,KAAK,MAAMA,EAAEtC,CAAC,EAAE,IAAIiF,EAAE,EAAE,GAAG,OAAO,GAAG,WAAWhD,IAAI,EAAEgD,EAAE/D,EAAE+D,EAAE,EAAEhD,IAAII,GAAG,GAAGA,EAAExC,GAAGyC,GAAG,GAAGA,EAAE1C,EAAE,CAAC,IAAIsF,EAAE5C,GAAGzC,EAAEC,GAAGqF,EAAG9C,EAAEvC,EAAEsF,EAAGvD,EAAEqD,EAAEC,EAAGlD,EAAEgD,EAAE,EAAEG,EAAG,CAAC,IAAIT,EAAE9C,EAAE2C,EAAEC,EAAExC,EAAE/C,EAAEyF,GAAGM,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOhG,EAAE,MAAMC,EAAE5B,EAAE,MAAMA,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,EAAM6oF,GAAGpT,GAAG5yD,GAAGnjB,GAAG,CAAC,IAAI,EAAE,KAAK,MAAMA,CAAC,EAAE,OAAOA,EAAE,EAAE,GAAG,KAAK,MAAMA,CAAC,EAAEA,EAAE,EAAE,GAAG,KAAK,KAAKA,CAAC,EAAE,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC,EAAEopF,GAAG,CAAC,WAAWjmE,GAAG,YAAY,MAAM,WAAWgmE,EAAE,EAAE,SAASE,GAAGrpF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,CAAC,UAAU2B,EAAE,WAAWC,EAAE,UAAUC,EAAE,QAAQS,EAAE,WAAWC,CAAC,EAAE23C,EAAE,gBAAgB,EAAEj6C,EAAE,CAAC,EAAEuC,EAAE,GAAGC,EAAE7C,EAAE,WAAWK,CAAC,EAAEyC,EAAE9C,EAAE,WAAW,CAAC,EAAEgE,EAAEyvE,GAAG5wE,EAAEC,EAAE,EAAEH,EAAEV,EAAED,EAAED,EAAEW,EAAE,EAAEE,CAAC,EAAE,OAAO5C,EAAE,eAAe,EAAEgE,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIolF,GAAG,CAAC,WAAWjmE,GAAG,YAAY,MAAM,WAAWgmE,EAAE,EAAE,SAASE,GAAGvpF,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAEI,EAAEN,EAAE,OAAOO,EAAE,EAAE,KAAKL,EAAEI,GAAGC,EAAE,KAAK,OAAOL,EAAEI,GAAG,CAAC,EAAEN,EAAEO,GAAG,EAAEL,EAAEK,EAAE,EAAED,EAAEC,EAAE,OAAOD,CAAC,CAAC,SAASkpF,GAAGxpF,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAEI,EAAEN,EAAE,OAAOO,EAAE,EAAE,KAAKL,EAAEI,GAAGC,EAAE,KAAK,OAAOL,EAAEI,GAAG,CAAC,EAAEN,EAAEO,IAAI,EAAEL,EAAEK,EAAE,EAAED,EAAEC,EAAE,OAAOD,CAAC,CAAC,SAASmpF,GAAGzpF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAEumB,EAAE,kBAAkB,QAAQ5mB,EAAEK,CAAC,EAAE,QAAQ0B,EAAE,EAAEA,EAAE/B,EAAE,EAAE+B,EAAE,CAAC,IAAIC,EAAElC,EAAE,MAAMiC,EAAE3B,GAAG2B,EAAE,GAAG3B,CAAC,EAAE6B,EAAEF,EAAE1B,EAAE,QAAQqC,EAAE,EAAEA,EAAErC,EAAE,EAAEqC,EAAE,EAAET,EAAES,GAAG,IAAI,OAAO2mF,GAAGrnF,EAAE,EAAEU,EAAET,EAAE,EAAEqnF,GAAGtnF,EAAE,EAAEU,EAAET,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,SAASunF,GAAG1pF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,eAAeO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE2B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOiC,EAAEsnF,GAAGxnF,EAAEC,EAAE3B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,CAAC,EAAE,OAAOL,EAAE,eAAe,EAAE,MAAM,QAAQiC,CAAC,CAAC,CAAC,IAAIwnF,GAAG,CAAC,WAAWrmE,GAAG,YAAY,MAAM,WAAWomE,EAAE,EAAE,SAASE,GAAG5pF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,UAAUM,EAAE,EAAEC,EAAE,EAAE,CAAC,EAAE,EAAEkxE,GAAG,CAACnxE,EAAEC,EAAE,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAED,EAAE,MAAM,OAAO2B,EAAE/B,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO4B,EAAEhC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO4B,EAAEjC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0C,EAAE+nB,GAAGpqB,EAAE,MAAM,EAAE,KAAK,EAAEsC,EAAEikB,EAAE,oBAAoBA,EAAE,cAAcvmB,EAAE,KAAK,EAAEqC,CAAC,EAAEE,EAAE,EAAEC,EAAE,IAAI,GAAG,EAAE,GAAGxC,EAAE,MAAM,SAAS,EAAE,EAAEumB,EAAE,cAAcvmB,EAAE,MAAM,MAAM,CAAC,CAAC,EAAE,QAAQyC,EAAE,EAAEA,EAAEf,EAAE,OAAOe,IAAI,QAAQkB,EAAE,EAAEA,EAAEnB,EAAEmB,IAAIjC,EAAEe,KAAK,EAAEH,EAAEC,KAAKZ,EAAEc,GAAGH,EAAEC,KAAKX,EAAEa,GAAG,OAAO9C,EAAE,eAAeK,EAAE,MAAMqC,EAAEC,CAAC,CAAC,CAAC,IAAIgnF,GAAG,CAAC,WAAWtmE,GAAG,YAAY,MAAM,WAAWqmE,EAAE,EAAME,GAAGtvC,EAAE,gBAAgBuvC,GAAGvvC,EAAE,WAAWwvC,GAAGjU,GAAGvyD,GAAGxjB,GAAGA,GAAG,EAAE+pF,GAAG/pF,EAAE8pF,IAAI,KAAK,IAAI9pF,CAAC,EAAE,EAAE,EAAEiqF,GAAG,CAAC,WAAWzmE,GAAG,YAAY,MAAM,WAAWwmE,EAAE,EAAME,GAAGnU,GAAGnyD,GAAG5jB,GAAGA,EAAE,EAAE,GAAGA,EAAE,EAAE,EAAE,CAAC,EAAEmqF,GAAG,CAAC,WAAWvmE,GAAG,YAAY,MAAM,WAAWsmE,EAAE,EAAME,GAAGrU,GAAGryD,GAAG1jB,GAAG,KAAK,IAAIA,CAAC,CAAC,EAAEqqF,GAAG,CAAC,WAAW3mE,GAAG,YAAY,MAAM,WAAW0mE,EAAE,EAAME,GAAGvU,GAAGpyD,GAAG3jB,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEuqF,GAAG,CAAC,WAAW5mE,GAAG,YAAY,MAAM,WAAW2mE,EAAE,EAAME,GAAG,sBAAsBC,GAAG,KAAK,IAAID,EAAE,EAAE,EAAEE,GAAG3U,GAAGjyD,GAAG9jB,GAAG,CAAC,IAAI,EAAEA,EAAE,CAACyqF,GAAGvqF,EAAEF,EAAEyqF,GAAGnqF,EAAE,KAAK,IAAIN,CAAC,EAAEO,EAAE,OAAOL,EAAEK,EAAED,EAAE,EAAEC,EAAEP,EAAEO,EAAE,KAAK,IAAI,EAAED,CAAC,EAAEC,CAAC,CAAC,EAAEoqF,GAAG,CAAC,WAAW7mE,GAAG,YAAY,MAAM,WAAW4mE,EAAE,EAAE,SAASE,GAAG5qF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,SAAS,CAAC,EAAED,EAAEmxE,GAAG,CAAClxE,CAAC,EAAE,gBAAgB,EAAE,IAAI0B,EAAE6kB,EAAE,cAAc,CAAC,EAAE5kB,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,EAAEA,EAAE,KAAK,GAAG,CAAC,EAAE,QAAQ4C,EAAE,EAAE,EAAE,OAAOA,EAAEvE,EAAE,MAAM,OAAO,EAAEuE,EAAE5C,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,EAAE,IAAIC,EAAEulF,GAAG,WAAW,CAAC,OAAO,CAAC,EAAEnnF,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,SAASgC,EAAE,cAAc,CAAC,CAAC,CAAC,EAAEU,EAAE43C,EAAE,YAAYr4C,EAAE,MAAM,EAAEF,EAAE,EAAE,EAAEY,EAAE23C,EAAE,YAAY53C,EAAE,OAAO,EAAE,OAAO,EAAE,EAAEE,EAAE03C,EAAE,oBAAoBr4C,EAAE,MAAM,EAAEF,EAAE,EAAE,EAAE,EAAE44E,GAAG,CAAC,OAAO,CAAC,EAAE14E,CAAC,EAAE,QAAQjC,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAEmC,EAAEizE,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ93E,EAAE,MAAM,CAAC,KAAK2C,CAAC,CAAC,CAAC,EAAEgC,EAAEg2E,GAAG,CAAC,OAAO,CAAC,EAAE91E,CAAC,EAAE,QAAQ7E,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8BiC,CAAC,EAAEjC,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B6E,CAAC,EAAEF,CAAC,CAAC,IAAIgmF,GAAG,CAAC,WAAW5mE,GAAG,YAAY,MAAM,WAAW2mE,EAAE,EAAE,SAASE,GAAG9qF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,QAAQM,EAAE,OAAOC,EAAE,WAAW,EAAE,aAAa,CAAC,EAAE,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACj7uG,EAAE,OAAO,EAAE,GAAGD,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACjDA,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACjDA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACjD,EAAE,OAAO,EAAE,IAAI0B,EAAE/B,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO4B,EAAEhC,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO4B,EAAEjC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO0C,EAAE1C,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,GAAG,CAAC2C,EAAEC,EAAEC,EAAEC,EAAEkB,CAAC,EAAE6vE,GAAG9xE,EAAE3B,EAAE,MAAMA,EAAE,MAAM4B,EAAE3B,EAAE,MAAM4B,EAAES,CAAC,EAAE,MAAM,CAAC1C,EAAE,eAAe4C,EAAExC,EAAE,MAAMuC,CAAC,EAAE3C,EAAE,eAAe,CAAC4C,EAAE,EAAE,EAAEvC,EAAE,MAAMwC,CAAC,EAAE7C,EAAE,eAAe,CAAC8C,EAAE,MAAM,EAAE,OAAO,IAAI,WAAWA,EAAE,IAAI,GAAG,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE9C,EAAE,eAAe,CAACgE,EAAE,MAAM,EAAE5D,EAAE,MAAM,IAAI,WAAW4D,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI6mF,GAAG,CAAC,WAAW3mE,GAAG,YAAY,MAAM,WAAW0mE,EAAE,EAAE,SAASE,GAAGhrF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,aAAaM,EAAE,WAAWC,EAAE,SAAS,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACnjBA,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACjDA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,KAAKL,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,MAAM,EAAE0B,EAAE/B,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO4B,EAAE,MAAM,KAAKhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAACiC,EAAES,EAAEC,CAAC,EAAEmxE,GAAG/xE,EAAE3B,EAAE,MAAMA,EAAE,MAAM,EAAE4B,CAAC,EAAE,MAAM,CAAChC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,EAAEjC,EAAE,eAAe,CAAC2C,EAAE,MAAM,EAAE,EAAE,MAAM,IAAI,WAAWA,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIooF,GAAG,CAAC,WAAW5mE,GAAG,YAAY,MAAM,WAAW2mE,EAAE,EAAE,SAASE,GAAGlrF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,YAC9mBA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,YACjD,EAAE,OAAO,EAAE,GAAGA,EAAE,MAAM,KAAK,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,+CAA+C,EAAE,IAAI,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO2B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAACiC,EAAES,CAAC,EAAEqxE,GAAG,EAAE3zE,EAAE,MAAMA,EAAE,MAAM2B,EAAEC,EAAE,EAAE,EAAE,OAAOhC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,CAAC,CAAC,IAAIgpF,GAAG,CAAC,WAAW7mE,GAAG,YAAY,MAAM,WAAW4mE,EAAE,EAAE,SAASE,GAAGprF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,WACvhBA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,WACjD,EAAE,OAAO,EAAE,GAAGA,EAAE,MAAM,KAAK,EAAE,MAAM,GAAG,MAAM,IAAI,MAAM,+CAA+C,EAAE,IAAI,EAAEL,EAAE,KAAK,IAAII,EAAE,MAAM,EAAE,OAAO2B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO2B,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAACiC,EAAES,CAAC,EAAEqxE,GAAG,EAAE3zE,EAAE,MAAMA,EAAE,MAAM2B,EAAEC,CAAC,EAAE,OAAOhC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,CAAC,CAAC,IAAIkpF,GAAG,CAAC,WAAW9mE,GAAG,YAAY,MAAM,WAAW6mE,EAAE,EAAE,SAASE,GAAGtrF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,cAAcO,EAAE,aAAa,EAAE,aAAa,CAAC,EAAE,EAAE,CAAC,YAAY0B,CAAC,EAAE3B,EAAE,CAAC,UAAU4B,EAAE,WAAWC,EAAE,UAAUS,EAAE,QAAQC,EAAE,WAAWC,CAAC,EAAE03C,EAAE,gBAAgB,EAAEj6C,EAAE0B,CAAC,EAAEc,EAAE,GAAGC,EAAE9C,EAAE,WAAWK,CAAC,EAAE2D,EAAE,OAAO,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,IAAI,EAAEhE,EAAE,WAAW,CAAC,EAAEqH,EAAE,QAAQrH,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,EAAEgE,EAAEyvE,GAAG3wE,EAAE,EAAEf,EAAEa,EAAEF,EAAET,EAAED,EAAEW,EAAE0E,EAAExE,CAAC,EAAE,KAAK,CAAC,IAAI,UAAU,CAAC,IAAI,EAAE7C,EAAE,WAAW,CAAC,EAAEqH,EAAErH,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,GAAGgE,EAAEyvE,GAAG3wE,EAAE,EAAEf,EAAEa,EAAEF,EAAET,EAAED,EAAEW,EAAE0E,EAAExE,CAAC,EAAE,KAAK,CAAC,IAAI,QAAQ,CAAC,IAAI,EAAE7C,EAAE,WAAW,CAAC,EAAEqH,EAAErH,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,GAAGgE,EAAEyvE,GAAG3wE,EAAE,EAAEf,EAAEa,EAAEF,EAAET,EAAED,EAAEW,EAAE0E,EAAExE,CAAC,EAAE,KAAK,CAAC,IAAI,SAAS,CAAC,IAAI,EAAE7C,EAAE,WAAW,CAAC,EAAEqH,EAAEuf,EAAE,aAAa5mB,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,EAAEgE,EAAEyvE,GAAG3wE,EAAE,EAAEf,EAAEa,EAAEF,EAAET,EAAED,EAAEW,EAAE0E,EAAExE,CAAC,EAAE,KAAK,CAAC,QAAQ,MAAM,IAAI,MAAM,oBAAoB,EAAE,OAAO,CAAC,CAAC,OAAO7C,EAAE,eAAe+B,EAAEiC,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIqnF,GAAG,CAAC,WAAW/mE,GAAG,YAAY,MAAM,WAAW8mE,EAAE,EAAE,SAASE,GAAGxrF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,gBAAgB,EAAE,KAAK,CAAC,EAAED,EAAE2B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG2B,EAAEs4C,EAAE,iBAAiBj6C,EAAE,EAAE0B,CAAC,EAAEE,EAAE,IAAI,MAAM5B,EAAE,MAAM,MAAM,EAAE,KAAK,CAAC,EAAEqC,EAAErC,EAAE,MAAM,MAAM,EAAE,OAAO2B,EAAE,IAAIW,GAAG,CAAC,IAAIC,EAAE,CAAC,GAAGF,CAAC,EAAEE,EAAEb,GAAGY,EAAE,IAAIE,EAAEq2E,GAAG,CAAC,OAAO,CAAC,EAAE74E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAMiC,EAAE,KAAKW,CAAC,CAAC,CAAC,EAAE,OAAOX,EAAEF,IAAIY,EAAEE,CAAC,CAAC,CAAC,CAAC,IAAI0oF,GAAG,CAAC,WAAWvnE,GAAG,YAAY,MAAM,WAAWsnE,EAAE,EAAME,GAAG,CAAC,WAAWhnE,GAAG,YAAY,MAAM,WAAW,CAAC,CAAC,OAAO1kB,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEE,CAAC,EAAEF,EAAEM,EAAE,EAAEmxE,GAAGvxE,EAAE,QAAQ,EAAE,IAAIK,EAAED,EAAE,KAAK,IAAIJ,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,aAAaK,EAAE,MAAM,EAAE,QAAQ0B,EAAE,EAAEA,EAAE1B,EAAE,OAAO,EAAE0B,EAAE,CAAC,IAAIC,EAAE3B,EAAE0B,GAAG,EAAEA,GAAGC,EAAEA,CAAC,CAAC,MAAM,CAAC,OAAO5B,EAAE,MAAM,EAAEJ,EAAE,MAAMA,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,EAAMyrF,GAAG5V,GAAGpwD,GAAG,CAAC3lB,EAAE,IAAI,CAAC,IAAIE,EAAE,EAAE,OAAO,MAAMF,CAAC,EAAE,IAAIA,EAAE,EAAE,EAAEE,EAAE,KAAK,CAAC,EAAE0rF,GAAG,CAAC,WAAWjmE,GAAG,YAAY,MAAM,WAAWgmE,EAAE,EAAE,SAASE,GAAG7rF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,QAAQ0B,EAAE,UAAUC,EAAE,QAAQC,EAAE,aAAaS,EAAE,YAAYC,EAAE,eAAeC,CAAC,EAAExC,EAAEmxE,GAAGlxE,EAAE,cAAc,EAAE,GAAG,CAAC,iBAAiBwC,EAAE,WAAWC,EAAE,WAAWkB,EAAE,UAAU,EAAE,cAAcqD,EAAE,MAAM,EAAE,IAAI5C,EAAE,QAAQC,CAAC,EAAEwxB,GAAG,UAAU71B,EAAE,MAAM,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,CAAC,EAAE+B,EAAE,GAAGX,EAAEW,EAAEg2E,GAAG,CAAC,OAAO,CAAC,EAAEt6E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,UAAU,GAAGuE,EAAE,CAACuf,EAAE,OAAOvmB,EAAE,MAAM,QAAQ,EAAE,IAAI,yCAAyCA,EAAE,MAAM,QAAQ,EAAE,IAAIuE,EAAEsxB,GAAG,gBAAgB,EAAEzxB,EAAEC,CAAC,EAAE4C,EAAE4xE,GAAG,CAAC,OAAO,CAAC,EAAE74E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK4E,CAAC,CAAC,CAAC,EAAED,EAAEg2E,GAAG,CAAC,OAAO,CAAC,EAAErzE,CAAC,EAAE,QAAQtH,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,EAAE9C,EAAE,8BAA8BsH,CAAC,CAAC,KAAK,CAAC,IAAI1C,EAAE5E,EAAE,WAAWK,CAAC,EAAEiH,EAAE4sE,GAAGrxE,EAAE+B,EAAEF,EAAE,CAAC,EAAEC,EAAE3E,EAAE,eAAe8C,EAAEwE,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,OAAO3C,CAAC,CAAC,IAAIinF,GAAG,CAAC,WAAWnnE,GAAG,YAAY,MAAM,WAAWknE,EAAE,EAAE,SAASE,GAAG/rF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,UAAUO,EAAE,YAAY,EAAE,QAAQ,EAAE,SAAS0B,EAAE,SAASC,EAAE,uBAAuBC,CAAC,EAAE7B,EAAE,CAAC,KAAKsC,EAAE,WAAWC,CAAC,EAAE,EAAEC,EAAE5C,EAAE,KAAK,IAAI0C,EAAE,MAAM,EAAE,OAAOG,EAAE7C,EAAE,KAAK,IAAI2C,EAAE,MAAM,EAAE,OAAO,CAACG,EAAEkB,CAAC,EAAEmwE,GAAGvxE,EAAEC,EAAExC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,EAAE,MAAM,CAACjC,EAAE,eAAe,CAAC8C,EAAE,MAAM,EAAE,SAASA,CAAC,EAAE9C,EAAE,eAAe2C,EAAE,MAAM,QAAQqB,CAAC,CAAC,CAAC,CAAC,IAAI8nF,GAAG,CAAC,WAAWpnE,GAAG,YAAY,MAAM,WAAWmnE,EAAE,EAAE,SAASE,GAAGjsF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,UAAUO,CAAC,EAAED,EAAE,CAAC,MAAM,EAAE,UAAU,CAAC,EAAE,EAAE,GAAG,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,kCAAkC,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,0CAA0C,EAAE,OAAO,EAAE,IAAI2B,EAAE/B,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAOgC,EAAEhC,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,GAAG,CAACiC,EAAES,EAAEC,CAAC,EAAEyxE,GAAGryE,EAAEC,EAAE3B,CAAC,EAAEuC,EAAEF,EAAE,OAAO,MAAM,CAAC1C,EAAE,eAAe,CAAC4C,EAAE,CAAC,EAAE,QAAQX,CAAC,EAAEjC,EAAE,eAAe,CAAC4C,CAAC,EAAE,SAASF,CAAC,EAAE1C,EAAE,eAAe,CAAC,CAAC,EAAE,QAAQ,IAAI,WAAW2C,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIqpF,GAAG,CAAC,WAAWrnE,GAAG,YAAY,MAAM,WAAWonE,EAAE,EAAE,SAASE,GAAGnsF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,WAAWO,CAAC,EAAED,EAAE,CAAC,MAAM,CAAC,EAAE,EAAE,GAAG,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,kCAAkC,EAAE,GAAGC,GAAG,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,IAAI,EAAEL,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO+B,EAAEsyE,GAAG,EAAEh0E,CAAC,EAAE,OAAOL,EAAE,eAAe,EAAE,MAAM,QAAQ+B,CAAC,CAAC,CAAC,IAAImqF,GAAG,CAAC,WAAWtnE,GAAG,YAAY,MAAM,WAAWqnE,EAAE,EAAME,GAAGtW,GAAG/wD,GAAGhlB,GAAG,KAAK,IAAIA,CAAC,CAAC,EAAEssF,GAAG,CAAC,WAAWtnE,GAAG,YAAY,MAAM,WAAWqnE,EAAE,EAAME,GAAGxW,GAAG9wD,GAAGjlB,GAAG,KAAK,KAAKA,CAAC,CAAC,EAAEwsF,GAAG,CAAC,WAAWvnE,GAAG,YAAY,MAAM,WAAWsnE,EAAE,EAAE,SAASE,GAAGzsF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,MAAM,EAAE,IAAI,EAAEk0E,GAAGv0E,EAAE,WAAWK,CAAC,EAAE,CAAC,EAAE,OAAOL,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAIwsF,GAAG,CAAC,WAAWxnE,GAAG,YAAY,MAAM,WAAWunE,EAAE,EAAE,SAASE,GAAG3sF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,OAAO,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,MAAM,EAAE,IAAI0B,EAAE/B,EAAE,KAAK,IAAIK,EAAE,MAAM,EAAE,OAAO,CAAC2B,EAAEC,CAAC,EAAEuyE,GAAGzyE,EAAE1B,EAAE,MAAMA,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,CAACL,EAAE,eAAegC,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,EAAEhC,EAAE,eAAeiC,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIyqF,GAAG,CAAC,WAAWznE,GAAG,YAAY,MAAM,WAAWwnE,EAAE,EAAE,SAASE,GAAG7sF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,WAAW,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,SAAS0B,EAAE,UAAUC,EAAE,YAAYC,CAAC,EAAEjC,EAAE,CAAC0C,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAM,CAACyC,EAAEkB,CAAC,EAAE/B,GAAG,KAAKA,EAAE,CAACU,EAAEC,CAAC,EAAE,EAAE,CAACF,EAAEI,EAAEkB,EAAEnB,CAAC,EAAEwE,EAAEuf,EAAE,eAAevmB,EAAE,KAAK,EAAE,EAAEgH,EAAE,GAAG5C,EAAE4C,EAAE,GAAG3C,EAAE2C,EAAE,GAAG1C,EAAEiiB,EAAE,eAAe,CAAC,EAAEhiB,EAAED,EAAE,GAAG2C,EAAE3C,EAAE,GAAGG,EAAEH,EAAE,GAAG4C,EAAEqf,EAAE,uBAAuBvmB,EAAE,MAAMumB,EAAE,cAAc,CAAC,CAAC,EAAErf,EAAE,KAAKvF,CAAC,EAAE,IAAI+C,EAAE3E,EAAE,KAAK,IAAIC,EAAE,MAAM,EAAE,OAAO2E,EAAE5E,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,QAAQ8E,EAAE,EAAEA,EAAExC,EAAE,EAAEwC,EAAE,CAAC,IAAIC,EAAE,EAAE,MAAM,KAAK,EAAEH,EAAEA,EAAE,SAASE,EAAE,EAAEA,EAAE,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAEtC,EAAE,EAAEsC,EAAE,QAAQ2C,EAAE,EAAEA,EAAE/D,EAAE,EAAE+D,EAAE,QAAQN,EAAE,EAAEA,EAAE5E,EAAE,EAAE4E,EAAE,CAAC,IAAIO,EAAEC,EAAG9C,EAAE,GAAG4C,EAAE5C,EAAE,GAAGC,EAAE,EAAE,GAAG6C,IAAK,EAAE,SAAS,IAAIC,GAAI/C,EAAE,GAAG4C,EAAE5C,EAAE,GAAGC,EAAED,EAAE,IAAI8C,EAAGE,GAAIhD,EAAE,GAAG4C,EAAE5C,EAAE,GAAGC,EAAED,EAAE,IAAI8C,EAAGG,GAAGwkF,GAAG1kF,EAAGtF,EAAEb,CAAC,EAAEsG,GAAGukF,GAAGzkF,EAAGxF,EAAEZ,CAAC,EAAE,OAAO,EAAE,CAAC,IAAI,UAAUiG,EAAE6kF,GAAG9nF,EAAEpC,EAAEC,EAAE,EAAE6B,EAAEC,EAAEQ,EAAEmD,GAAGD,GAAGX,EAAEzF,CAAC,EAAE,MAAM,IAAI,WAAWgG,EAAE8kF,GAAG/nF,EAAEpC,EAAEC,EAAE,EAAE6B,EAAEC,EAAEQ,EAAEmD,GAAGD,GAAGX,EAAEzF,CAAC,EAAE,MAAM,QAAQ,MAAM,IAAI,MAAM,+DAA+D,GAAG,CAAC,CAAC,IAAIyE,GAAGvB,EAAEN,EAAEQ,EAAEkC,EAAES,EAAEjD,EAAE2C,EAAEF,EAAEd,IAAIuB,CAAC,CAAC,OAAO5H,EAAE,eAAe,EAAEC,EAAE,MAAMkH,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOnH,EAAE,MAAMmH,EAAE,EAAElH,EAAE,KAAK,EAAE,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,IAAI0sF,GAAG,CAAC,WAAW7nE,GAAG,YAAY,MAAM,WAAWynE,EAAE,EAAE,SAASC,GAAG9sF,EAAE,EAAEE,EAAE,CAAC,OAAOA,EAAE,CAAC,IAAI,UAAU,OAAOgtF,GAAGltF,EAAE,CAAC,EAAE,IAAI,OAAO,OAAOmtF,GAAGntF,EAAE,CAAC,EAAE,IAAI,UAAU,OAAOotF,GAAGptF,EAAE,CAAC,EAAE,IAAI,WAAW,QAAQ,OAAOqtF,GAAGrtF,EAAE,CAAC,CAAC,CAAC,CAAC,SAASktF,GAAGltF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAGE,EAAE,EAAE,GAAG,GAAG,EAAEA,EAAE,MAAM,CAAC,IAAII,EAAE,EAAE,EAAEJ,EAAEI,IAAIJ,EAAEI,EAAE,KAAK,MAAM,CAACJ,EAAEI,CAAC,EAAEJ,GAAGA,EAAEA,EAAE,CAAC,EAAEA,EAAEI,EAAE,CAACJ,EAAE,CAAC,SAASA,EAAE,EAAE,EAAE,GAAG,GAAG,EAAEA,EAAE,MAAM,CAAC,IAAII,EAAE,EAAE,EAAEJ,GAAGI,EAAE,KAAK,MAAMJ,EAAEI,CAAC,EAAEJ,GAAG,IAAIA,EAAEI,EAAEJ,EAAE,EAAE,CAAC,OAAO4mB,EAAE,MAAM,EAAE5mB,EAAE,EAAE,CAAC,CAAC,CAAC,SAASitF,GAAGntF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAGE,EAAE,EAAE,GAAG,GAAG,EAAEA,EAAE,MAAM,CAAC,IAAII,EAAE,EAAE,EAAEJ,GAAG,GAAG,KAAK,MAAM,CAACA,EAAEI,CAAC,EAAE,EAAE,SAASJ,EAAE,EAAE,EAAE,GAAG,GAAG,EAAEA,EAAE,MAAM,CAAC,IAAII,EAAE,EAAE,EAAEJ,GAAG,EAAE,KAAK,MAAMA,EAAEI,CAAC,CAAC,CAAC,OAAOwmB,EAAE,MAAM,EAAE5mB,EAAE,EAAE,CAAC,CAAC,CAAC,SAASmtF,GAAGrtF,EAAE,EAAE,CAAC,OAAOA,CAAC,CAAC,SAASotF,GAAGptF,EAAE,EAAE,CAAC,OAAO8mB,EAAE,MAAM,EAAE9mB,EAAE,EAAE,CAAC,CAAC,CAAC,SAASstF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAE,CAAC,IAAIC,EAAE,EAAEvC,EAAE2B,EAAE1B,EAAE2B,EAAE,EAAEC,EAAE,MAAO,IAAGF,GAAGA,EAAE,GAAG,GAAGC,GAAGA,EAAEhC,EAAEF,EAAE6C,GAAGD,CAAC,CAAC,SAASmqF,GAAG/sF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAE,CAAC,IAAIC,EAAE,KAAK,MAAMZ,CAAC,EAAEa,EAAE,KAAK,MAAMZ,CAAC,EAAE,OAAOorF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAEsC,EAAEC,EAAEX,EAAES,CAAC,CAAC,CAAC,SAASoqF,GAAGhtF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAE,CAAC,IAAIC,EAAE,KAAK,MAAMZ,CAAC,EAAEa,EAAE,KAAK,MAAMZ,CAAC,EAAEa,EAAEF,EAAE,EAAEG,EAAEF,EAAE,EAAEoB,GAAGlB,EAAEd,GAAGorF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAEsC,EAAEC,EAAEX,EAAES,CAAC,GAAGV,EAAEY,GAAGwqF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAEsC,EAAEG,EAAEb,EAAES,CAAC,EAAE,GAAGI,EAAEd,GAAGorF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAEwC,EAAED,EAAEX,EAAES,CAAC,GAAGV,EAAEY,GAAGwqF,GAAGttF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAEwC,EAAEC,EAAEb,EAAES,CAAC,EAAE,OAAOG,EAAEd,GAAGiC,GAAGjC,EAAEY,GAAG,CAAC,CAAC,SAAS0qF,GAAIvtF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAEL,EAAE,CAAC,EAAE,CAAC,EAAE,EAAEuxE,GAAG,EAAE,QAAQ,EAAE,IAAI,EAAEnxE,EAAE,KAAK,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,aAAa2B,EAAE,YAAYC,EAAE,QAAQC,CAAC,EAAEyyE,GAAG,EAAEr0E,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAACD,EAAE,eAAe4B,EAAE,EAAE,MAAMD,CAAC,EAAE3B,EAAE,eAAe,CAAC6B,EAAE,MAAM,EAAE,QAAQA,CAAC,CAAC,CAAC,CAAC,IAAIqrF,GAAG,CAAC,WAAWloE,GAAG,YAAY,MAAM,WAAWioE,EAAG,EAAE,SAASE,GAAIztF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAE,IAAI,GAAGC,EAAE,MAAM,QAAQ,IAAI,EAAEA,EAAE,MAAM,OAAO0B,EAAE1B,EAAE,MAAM,GAAG2B,EAAE,IAAI,MAAM,EAAE,CAAC,EAAEC,EAAE,EAAE,QAAQY,EAAE,EAAEA,EAAE,EAAEA,IAAIA,IAAI,IAAIb,EAAEC,KAAK5B,EAAE,MAAMwC,IAAI,IAAIH,EAAE,IAAI,MAAM,CAAC,EAAE,KAAK,CAAC,EAAEC,EAAEtC,EAAE,MAAM,MAAM,EAAEsC,EAAE,GAAG,EAAE,IAAIC,EAAE,IAAI,MAAMb,CAAC,EAAE,QAAQc,EAAE,EAAEA,EAAED,EAAE,OAAOC,IAAI,CAACH,EAAE,GAAGG,EAAE,IAAIC,EAAEo2E,GAAG,CAAC,OAAO,CAAC,EAAE74E,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM0C,EAAE,KAAKC,CAAC,CAAC,CAAC,EAAEC,EAAEC,GAAG83E,GAAG,CAAC,OAAO,CAAC,EAAE73E,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEhC,EAAE,8BAA8B8C,CAAC,CAAC,CAAC,OAAOF,CAAC,CAAC,IAAI4qF,GAAG,CAAC,WAAWnoE,GAAG,YAAY,MAAM,WAAWkoE,EAAG,EAAE,SAASE,GAAI3tF,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,WAAW,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,EAAED,EAAEmxE,GAAGlxE,EAAE,oBAAoB,EAAE,IAAI0B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE,EAAE,MAAM,OAAOC,EAAE,CAAC,EAAES,EAAE,CAAC,EAAEC,EAAEZ,EAAEC,EAAEY,EAAE,EAAE,QAAQE,EAAE,EAAEA,EAAEH,EAAE,EAAEG,EAAE,CAAC,IAAIkB,EAAEy9E,GAAG,CAAC,OAAO,CAAC,MAAM7+E,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,IAAI8C,EAAE,CAAC,CAAC,CAAC,EAAEF,EAAEoB,EAAEtB,EAAE,KAAKsB,CAAC,CAAC,CAAC,QAAQlB,EAAE,EAAEA,EAAE,EAAE,EAAEA,EAAE,CAAC,IAAIkB,EAAE4iB,EAAE,kBAAkB9jB,EAAE,OAAO,EAAE,EAAE9C,EAAE,eAAe,CAAC,EAAE,QAAQgE,CAAC,EAAEqD,EAAE4uE,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,EAAErzE,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE,EAAEq1E,GAAG,CAAC,OAAO,CAAC,EAAAhuE,CAAC,EAAE,QAAQrH,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAEyE,EAAE+yE,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,EAAEn3E,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE0E,EAAEi8E,GAAG,CAAC,OAAO,CAACl8E,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,KAAK,EAAE,SAAS,EAAE,CAAC,CAAC,EAAEiC,EAAE,KAAKyC,CAAC,EAAEhC,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAK2E,CAAC,EAAE3E,EAAE,KAAK,CAAC,EAAEA,EAAE,KAAK+B,CAAC,EAAE/B,EAAE,KAAKgC,CAAC,CAAC,CAAC,IAAI7B,EAAEwkF,GAAG,CAAC,OAAOplF,EAAE,QAAQjC,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,OAAO0C,EAAE,QAAQI,GAAG9C,EAAE,8BAA8B8C,CAAC,CAAC,EAAED,CAAC,CAAC,IAAI6qF,GAAG,CAAC,WAAWpoE,GAAG,YAAY,MAAM,WAAWmoE,EAAG,EAAME,GAAI,CAAC3S,GAAGpG,GAAGsG,GAAGE,GAAGzF,GAAG2F,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGG,GAAGE,GAAGM,GAAGE,GAAGE,GAAGE,GAAGvC,GAAGyC,GAAGE,GAAGE,GAAGE,GAAGvI,GAAGU,GAAG+H,GAAGhJ,GAAGkJ,GAAGI,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGC,GAAGC,GAAGC,GAAGI,GAAG9G,GAAGgH,GAAG9K,GAAGsL,GAAGpL,GAAGsL,GAAGpL,GAAG8L,GAAGG,GAAGC,GAAGhM,GAAGmM,GAAGE,GAAGE,GAAGE,GAAGE,GAAGzM,GAAGE,GAAG1B,GAAGmO,GAAGlF,GAAGoF,GAAGE,GAAGE,GAAGzJ,GAAGpD,GAAGE,GAAG6M,GAAG3M,GAAG6M,GAAGG,GAAGE,GAAGG,GAAGE,GAAGE,GAAGE,GAAGzN,GAAG2N,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGrO,GAAGuO,GAAGG,GAAGK,GAAG5O,GAAGE,GAAG6O,GAAGG,GAAGG,GAAGjP,GAAGmP,GAAGI,GAAGE,GAAGE,GAAGG,GAAGtN,GAAGpC,GAAG4P,GAAGE,GAAGE,GAAGE,GAAG/S,GAAGyM,GAAGwG,GAAG9N,GAAGE,GAAGG,GAAG2N,GAAGE,GAAGE,GAAGE,GAAGE,GAAGC,GAAGE,GAAGnQ,GAAGqQ,GAAGK,GAAGE,GAAGI,GAAG9Q,GAAGgR,GAAGE,GAAGE,GAAGlR,GAAGgN,GAAGsE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGE,GAAGlS,GAAGmS,GAAGjS,GAAGmS,GAAGE,GAAGE,GAAGE,GAAGE,GAAGtS,GAAGgH,GAAGwL,GAAGE,GAAGE,GAAGE,GAAGK,GAAGhV,GAAGuV,GAAGE,GAAGE,GAAGxG,EAAE,EAAE,QAAQpnF,KAAK6tF,GAAIpnE,GAAGzmB,CAAC,EAAE,IAAI8tF,GAAG,CAAC,EAAE1tF,GAAG0tF,GAAG,CAAC,iBAAiB,IAAIC,GAAG,wBAAwB,IAAIC,GAAI,8BAA8B,IAAIC,GAAG,mCAAmC,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,mCAAmC,IAAIC,GAAG,aAAa,IAAIC,GAAG,iBAAiB,IAAIC,GAAG,qBAAqB,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,cAAc,IAAIC,GAAG,wBAAwB,IAAIC,GAAG,yBAAyB,IAAIC,GAAG,cAAc,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,YAAY,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,eAAe,IAAIC,GAAI,0BAA0B,IAAIC,GAAG,iCAAiC,IAAIC,GAAG,YAAY,IAAIC,GAAG,aAAa,IAAIC,GAAG,gCAAgC,IAAIC,GAAG,kCAAkC,IAAIC,GAAG,qBAAqB,IAAIC,GAAG,uBAAuB,IAAIC,GAAG,aAAa,IAAIC,GAAG,mCAAmC,IAAIC,GAAG,8BAA8B,IAAIC,GAAG,cAAc,IAAIC,GAAG,oBAAoB,IAAIC,GAAG,sBAAsB,IAAIC,GAAG,YAAY,IAAIC,GAAG,0BAA0B,IAAIC,GAAG,oBAAoB,IAAIC,GAAI,yBAAyB,IAAIC,GAAI,kCAAkC,IAAIC,GAAG,kBAAkB,IAAIC,GAAI,oBAAoB,IAAIC,GAAG,gBAAgB,IAAIC,GAAG,oBAAoB,IAAIC,EAAE,CAAC,EAAE,IAAIC,GAAG,CAAC,EAAEC,GAAG,CAAC,MAAM,GAAG,UAAU,GAAG,mBAAmB,GAAG,sBAAsB,GAAG,MAAM,GAAG,QAAQ,GAAG,6BAA6B,EAAE,EAAE,SAASC,GAAG5wF,EAAE,EAAE,CAAC0wF,GAAG1wF,GAAG,CAAC,CAAC,SAAS6wF,GAAG7wF,EAAE,EAAE,CAAC,GAAG,EAAEA,KAAK0wF,KAAK,GAAG,KAAK,CAAC,IAAIpwF,EAAEwwF,GAAI9wF,EAAE,CAAC,EAAE,GAAGM,IAAI,KAAKowF,GAAG1wF,GAAGM,MAAO,QAAO,QAAQ,IAAI,0CAA0CN,CAAC,EAAE,IAAI,CAAC,IAAIE,EAAEwwF,GAAG1wF,GAAG,OAAOE,GAAG,MAAMA,EAAE,cAAc,GAAG,OAAOwwF,GAAG1wF,GAAG6wF,GAAG7wF,CAAC,IAAIE,EAAE,QAAQA,EAAE,UAAU,EAAEA,EAAE,QAAQA,EAAE,YAAY,EAAEA,EAAE,QAAQA,EAAE,KAAK,EAAEA,EAAE,QAAQA,EAAE,MAAM,EAAEA,EAAE,QAAQA,EAAE,mBAAmB,EAAEA,EAAE,QAAQA,EAAE,eAAe,EAAEA,EAAE,OAAOA,EAAE,YAAY,EAAEA,EAAE,OAAOA,EAAE,SAAS,EAAEA,EAAE,SAASA,EAAE,IAAI,EAAEwwF,GAAG1wF,GAAG,CAAC,SAAS+wF,GAAI/wF,EAAE,CAAC,GAAG,OAAO,iBAAiB,aAAaA,IAAI,EAAE,OAAO,IAAI,gBAAgB,IAAI,GAAG,EAAE,GAAG,OAAO,UAAU,YAAY,OAAO,SAAS,cAAc,QAAQ,EAAE,MAAM,IAAI,MAAM,wCAAwC,CAAC,CAAC,SAAS8wF,GAAI9wF,EAAE,EAAE,CAAC,GAAGA,IAAI,GAAGA,IAAI,EAAE,MAAM,IAAI,MAAM,wDAAwD,EAAE,IAAIE,EAAE,GAAG,KAAK6wF,GAAI/wF,CAAC,EAAE,EAAE,OAAOE,EAAE,iBAAiB,mBAAmBI,GAAG,CAACA,EAAE,eAAe,EAAE,OAAOowF,GAAG1wF,EAAE,EAAE,EAAE,EAAEya,EAAE,EAAE,QAAQ,wBAAwB,IAAIk2E,GAAG,6BAA6B,IAAI3wF,IAAI,EAAEE,EAAE,WAAW,QAAQywF,EAAE,GAAGzwF,EAAE,WAAW,qBAAqBywF,EAAE,EAAEzwF,EAAE,WAAW,SAASywF,EAAE,CAAC,CAAC,IAAIK,IAAI,SAAShxF,EAAE,CAACA,EAAEA,EAAE,MAAM,GAAG,QAAQA,EAAEA,EAAE,aAAa,GAAG,cAAc,GAAGgxF,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASjxF,EAAE,CAACA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,SAAS,GAAG,UAAU,GAAGixF,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASlxF,EAAE,CAACA,EAAEA,EAAE,iBAAiB,GAAG,mBAAmBA,EAAEA,EAAE,iBAAiB,GAAG,mBAAmBA,EAAEA,EAAE,yBAAyB,GAAG,2BAA2BA,EAAEA,EAAE,mBAAmB,GAAG,qBAAqBA,EAAEA,EAAE,mBAAmB,GAAG,oBAAoB,GAAGkxF,KAAKA,GAAG,CAAC,EAAE,EAAE,SAASC,GAAGnxF,EAAE,EAAE,CAAC,MAAM,CAAC,EAAEA,CAAC,CAAC,CAAC,SAASoxF,GAAGpxF,EAAE,EAAE,CAAC,OAAOA,EAAE,CAAC,CAAC,SAASqxF,GAAGrxF,EAAE,CAAC,IAAI,EAAE8mB,EAAE,cAAc9mB,CAAC,EAAEE,EAAE,KAAK,KAAK,EAAE,CAAC,EAAE,OAAO4mB,EAAE,oBAAoB5mB,CAAC,CAAC,CAAC,SAASoxF,GAAGtxF,EAAE,EAAE,CAAC,MAAM,CAAC,KAAK,IAAI,EAAE,KAAK,KAAK,EAAE,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE,KAAK,KAAKA,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,SAASuxF,GAAGvxF,EAAE,EAAE,CAAC,GAAG,CAACE,EAAEI,CAAC,EAAEgxF,GAAGtxF,EAAE,CAAC,EAAE,OAAOE,EAAEI,EAAE,CAAC,CAAC,SAASkxF,GAAGxxF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAEM,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,EAAE,OAAO2X,EAAE,EAAE,UAAU,eAAe,IAAI,GAAGna,EAAEJ,EAAE,KAAKK,EAAEL,EAAE,KAAK,EAAEA,EAAE,QAAQ,EAAEA,EAAE,QAAQ+B,EAAE/B,EAAE,IAAIiC,EAAE,EAAES,EAAE,EAAEC,EAAE3C,EAAE,WAAW4C,EAAE5C,EAAE,MAAMgC,EAAEhC,EAAE,QAAQI,EAAEN,EAAE,KAAKO,EAAEP,EAAE,KAAK,EAAEA,EAAE,KAAK,EAAEE,EAAE,KAAK+B,EAAEjC,EAAE,KAAKmC,EAAE,EAAES,EAAE,EAAEC,EAAE,GAAG,KAAK,EAAE,eAAe,KAAKC,EAAE9C,EAAE,MAAMkC,EAAElC,EAAE,MAAM,CAAC,oBAAoBM,EAAE,wBAAwBC,EAAE,8BAA8B,EAAE,0BAA0B,EAAE,mBAAmB0B,EAAE,sBAAsBC,EAAE,0BAA0BC,EAAE,mBAAmBS,EAAE,qBAAqBC,EAAE,iBAAiBC,CAAC,CAAC,CAAC,SAASurF,GAAGruF,EAAE,EAAE,CAAC,IAAIE,EAAE,EAAE,EAAE,OAAOua,EAAE,EAAE,QAAQ,OAAO,GAAGg3E,GAAIzxF,CAAC,EAAEE,CAAC,CAAC,SAASuxF,GAAIzxF,EAAE,CAAC,IAAI,EAAEA,EAAE,SAAS,EAAE,GAAG,IAAIA,EAAE,SAAS,MAAM,IAAI,MAAM,gBAAgByvF,GAAGzvF,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI0xF,GAAI,QAAQC,GAAI,MAAM,SAASrD,GAAGtuF,EAAE,CAAC,MAAM,CAAC,EAAEya,EAAE,EAAE,QAAQ,8BAA8B,GAAGza,IAAI,GAAG0xF,GAAI,KAAK,IAAI1xF,CAAC,GAAG,KAAK,IAAIA,CAAC,EAAE2xF,GAAI,CAAC,SAASlC,GAAGzvF,EAAE,EAAE,CAAC,OAAO,EAAE,CAAC,KAAKA,EAAE,SAAS,MAAM,WAAW,KAAKA,EAAE,aAAa,MAAM,eAAe,KAAKA,EAAE,cAAc,MAAM,gBAAgB,KAAKA,EAAE,kBAAkB,MAAM,oBAAoB,KAAKA,EAAE,8BAA8B,MAAM,gCAAgC,KAAKA,EAAE,cAAc,MAAM,gBAAgB,KAAKA,EAAE,mBAAmB,MAAM,qBAAqB,QAAQ,MAAM,sBAAsB,GAAG,CAAC,CAAC,SAAS+uF,GAAG/uF,EAAE,EAAE,CAAC,OAAO4xF,GAAG5xF,EAAE,IAAIA,EAAE,aAAa,CAAC,EAAE,cAAc,EAAE,kCAAkC,CAAC,CAAC,SAAS6uF,GAAG7uF,EAAE,EAAE,CAAC,IAAIE,EAAE0xF,GAAG5xF,EAAE,IAAIA,EAAE,aAAaA,EAAE,aAAa,EAAE,sCAAsC,EAAE,GAAGquF,GAAGruF,EAAE,IAAIA,EAAE,aAAaE,EAAE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,cAAcE,CAAC,CAAC,EAAEF,EAAE,mBAAmBE,EAAEF,EAAE,cAAc,IAAI,GAAG,MAAM,QAAQ,IAAIA,EAAE,iBAAiBE,CAAC,CAAC,EAAE,IAAI,MAAM,kCAAkC,EAAE,OAAOA,CAAC,CAAC,SAASquF,GAAGvuF,EAAE,EAAE,CAAC,IAAIE,EAAE0xF,GAAG5xF,EAAE,IAAIA,EAAE,aAAaA,EAAE,eAAe,EAAE,wCAAwC,EAAE,GAAGquF,GAAGruF,EAAE,IAAIA,EAAE,aAAaE,EAAE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,cAAcE,CAAC,CAAC,EAAEua,EAAE,EAAE,IAAI,qBAAqB,EAAE,OAAOva,EAAE,GAAGF,EAAE,mBAAmBE,EAAEF,EAAE,cAAc,IAAI,GAAG,MAAMkwF,GAAG,EAAElwF,EAAE,iBAAiBE,CAAC,CAAC,EAAE,IAAI,MAAM,oCAAoC,EAAE,OAAOA,CAAC,CAAC,IAAI2xF,GAAI,2BAA2B,SAAS3B,GAAGlwF,EAAE,EAAE,CAAC,IAAIE,EAAE2xF,GAAI,KAAK,CAAC,EAAE,GAAG3xF,GAAG,KAAK,CAAC,QAAQ,IAAI,wCAAwC,GAAG,EAAE,QAAQ,IAAIF,CAAC,EAAE,MAAM,CAAC,IAAIM,EAAE,CAACJ,EAAE,GAAGK,EAAEP,EAAE,MAAM;AAAA,CAC9ic,EAAE,EAAEO,EAAE,OAAO,SAAS,EAAE,OAAO,EAAE,EAAEA,EAAE,IAAI,CAACsC,EAAEC,IAAIgkB,EAAE,UAAUhkB,EAAE,GAAG,SAAS,EAAE,CAAC,EAAED,CAAC,EAAEZ,EAAE,EAAE,QAAQY,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAIZ,EAAE,KAAK,IAAI,EAAEY,GAAG,OAAOZ,CAAC,EAAE,IAAIC,EAAE,EAAE,MAAM,EAAE5B,EAAE,CAAC,EAAE6B,EAAE,EAAE,MAAM7B,EAAE,EAAEA,CAAC,EAAEsC,EAAE,EAAE,MAAMtC,CAAC,EAAE,QAAQ,IAAI4B,EAAE,KAAK;AAAA,CACjN,CAAC,EAAE,QAAQ,IAAI,EAAE,MAAM;AAAA,CACvB,EAAE,EAAE,EAAE,QAAQ,IAAI,MAAM4kB,EAAE,SAAS3kB,EAAE,GAAGF,CAAC,IAAI,+DAA+D,EAAE,QAAQ,IAAIW,EAAE,KAAK;AAAA,CACjI,CAAC,CAAC,CAAC,SAAS6rF,GAAGzuF,EAAE,CAAC,OAAO4xF,GAAG5xF,EAAE,IAAIA,EAAE,cAAc,EAAE,gCAAgC,CAAC,CAAC,SAASiwF,GAAGjwF,EAAE,EAAE,CAAC,GAAGquF,GAAGruF,EAAE,IAAIA,EAAE,YAAY,CAAC,CAAC,EAAE,CAACya,EAAE,EAAE,IAAI,qBAAqB,GAAGza,EAAE,oBAAoB,EAAEA,EAAE,WAAW,IAAI,GAAG,MAAM,QAAQ,IAAIA,EAAE,kBAAkB,CAAC,CAAC,EAAE,IAAI,MAAM,6CAA6C,CAAC,CAAC,SAASwwF,GAAGxwF,EAAE,EAAE,CAAC,GAAGquF,GAAGruF,EAAE,IAAIA,EAAE,gBAAgB,CAAC,CAAC,EAAEA,EAAE,oBAAoB,EAAEA,EAAE,eAAe,IAAI,GAAG,MAAM,QAAQ,IAAIA,EAAE,kBAAkB,CAAC,CAAC,EAAE,IAAI,MAAM,mCAAmC,CAAC,CAAC,SAAS2uF,GAAG3uF,EAAE,EAAE,CAAC,IAAIE,EAAE0xF,GAAG5xF,EAAE,IAAIA,EAAE,aAAa,EAAE,8BAA8B,EAAE,OAAOquF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,aAAaE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,aAAa,EAAEA,EAAE,WAAW,CAAC,EAAEE,CAAC,CAAC,SAASwuF,GAAG1uF,EAAE,EAAE,CAAC,IAAIE,EAAE0xF,GAAG5xF,EAAE,IAAIA,EAAE,aAAa,EAAE,8BAA8B,EAAE,OAAOquF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,qBAAqBE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,qBAAqB,EAAEA,EAAE,WAAW,CAAC,EAAEE,CAAC,CAAC,SAASgvF,IAAK,CAAC,OAAOz0E,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE,EAAE,CAAC,CAAC,SAASm0E,GAAG5uF,EAAE,CAAC,OAAO4xF,GAAG5xF,EAAE,IAAIA,EAAE,cAAc,EAAE,gCAAgC,CAAC,CAAC,SAASywF,GAAGzwF,EAAE,EAAE,CAAC,IAAIE,EAAEua,EAAE,EAAE,UAAU,wBAAwB,EAAE,GAAGza,GAAG,GAAG,GAAG,EAAE,CAAC,IAAIM,EAAE,IAAIN,KAAK,KAAK,MAAM,IAAI,MAAM,0BAA0BM,EAAE,cAAc,CAAC,CAAC,GAAGN,EAAEE,GAAG,EAAEA,EAAE,CAAC,IAAII,EAAE,IAAIN,KAAK,KAAKO,EAAE,IAAIL,KAAKA,KAAK,MAAM,IAAI,MAAM,0BAA0BI,EAAE,qDAAqDC,EAAE,GAAG,CAAC,CAAC,CAAC,SAASiuF,GAAGxuF,EAAE,CAAC,OAAO4xF,GAAG5xF,EAAE,IAAIA,EAAE,kBAAkB,EAAE,oCAAoC,CAAC,CAAC,SAASouF,GAAGpuF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAEjC,EAAE,kBAAkB,EAAEE,CAAC,EAAE,OAAO+B,IAAI,GAAG,IAAIosF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,aAAaM,CAAC,CAAC,EAAE+tF,GAAGruF,EAAE,IAAIA,EAAE,oBAAoBiC,EAAE1B,EAAEP,EAAE,MAAM,GAAG,EAAE,CAAC,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,wBAAwBiC,CAAC,CAAC,EAAE,GAAG,CAAC,SAASksF,GAAGnuF,EAAE,EAAEE,EAAE,CAAC4xF,GAAG9xF,EAAEE,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,cAAcA,EAAE,SAASE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,SAASswF,GAAItwF,EAAE,EAAE,CAAC8xF,GAAG9xF,EAAE,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,cAAcA,EAAE,SAAS,CAAC,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,IAAI,CAAC,CAAC,CAAC,SAASovF,GAAGpvF,EAAE,EAAEE,EAAE,CAAC,OAAO0xF,GAAG5xF,EAAE,IAAIA,EAAE,mBAAmB,EAAEE,CAAC,EAAE,YAAYA,EAAE,2BAA2B,CAAC,CAAC,SAASivF,GAAGnvF,EAAE,EAAEE,EAAE,CAAC,OAAOF,EAAE,mBAAmB,EAAEE,CAAC,CAAC,CAAC,SAASguF,GAAGluF,EAAE,EAAEE,EAAEI,EAAE,CAAC+tF,GAAGruF,EAAE,IAAImuF,GAAGnuF,EAAE,EAAEM,CAAC,CAAC,EAAE+tF,GAAGruF,EAAE,IAAIA,EAAE,UAAUE,EAAEI,CAAC,CAAC,CAAC,CAAC,SAAS0tF,GAAIhuF,EAAE,CAACquF,GAAGruF,EAAE,IAAIA,EAAE,gBAAgBA,EAAE,YAAY,IAAI,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,SAAS,EAAE,EAAEA,EAAE,OAAO,MAAMA,EAAE,OAAO,MAAM,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,QAAQ,EAAE,EAAEA,EAAE,OAAO,MAAMA,EAAE,OAAO,MAAM,CAAC,CAAC,CAAC,SAASiuF,GAAGjuF,EAAE,EAAEE,EAAE,CAACmuF,GAAGruF,EAAE,IAAIA,EAAE,gBAAgBA,EAAE,YAAYE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,qBAAqBA,EAAE,YAAYA,EAAE,kBAAkBA,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC,CAAC,SAASqwF,GAAGrwF,EAAE,EAAE,CAACquF,GAAGruF,EAAE,IAAIA,EAAE,gBAAgBA,EAAE,YAAY,CAAC,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,qBAAqBA,EAAE,YAAYA,EAAE,kBAAkBA,EAAE,WAAW,KAAK,CAAC,CAAC,CAAC,CAAC,SAASuwF,GAAGvwF,EAAE,CAAC,IAAI,EAAEA,EAAE,uBAAuBA,EAAE,WAAW,EAAE,GAAG,IAAIA,EAAE,qBAAqB,MAAM,IAAI,MAAM,8BAA8BgvF,GAAGhvF,EAAE,CAAC,CAAC,CAAC,CAAC,SAASgvF,GAAGhvF,EAAE,EAAE,CAAC,OAAO,EAAE,CAAC,KAAKA,EAAE,kCAAkC,MAAM,oCAAoC,KAAKA,EAAE,0CAA0C,MAAM,4CAA4C,KAAKA,EAAE,kCAAkC,MAAM,oCAAoC,KAAKA,EAAE,wBAAwB,MAAM,0BAA0B,QAAQ,MAAM,iBAAiB,GAAG,CAAC,CAAC,SAAS4xF,GAAG5xF,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE+tF,GAAGruF,EAAE,IAAI,EAAE,CAAC,EAAE,GAAGM,GAAG,KAAK,MAAM,IAAI,MAAMJ,CAAC,EAAE,OAAOI,CAAC,CAAC,SAASwxF,GAAG9xF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,iCAAiC,EAAEM,EAAE,EAAEN,EAAE,SAAS,GAAGM,EAAEN,EAAE,UAAUM,EAAEJ,EAAE,CAAC,IAAIK,EAAE,2BAA2BL,KAAK,MAAM,IAAI,MAAM,0BAA0BK,IAAI,CAAC,CAAC,CAAC,SAASuuF,GAAG9uF,EAAE,EAAE,EAAE,CAAC,OAAO8mB,EAAE,cAAc9mB,EAAE,MAAM,EAAEA,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,SAASqvF,GAAGrvF,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,MAAM,MAAM,sDAAsD,EAAE,MAAM,CAACA,EAAE,OAAO,EAAEA,EAAEA,EAAE,OAAO,GAAG,EAAEA,EAAEA,EAAE,OAAO,EAAE,CAAC,CAAC,SAASsvF,GAAGtvF,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,OAAOA,EAAE,SAAS,GAAGA,EAAE,SAAS,GAAGA,EAAE,KAAK,IAAI,EAAE,CAAC8uF,GAAG9uF,CAAC,EAAE,GAAGqvF,GAAGrvF,CAAC,CAAC,GAAG,CAAC,CAAC,SAASuvF,GAAGvvF,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAEua,EAAE,EAAE,UAAU,wBAAwB,EAAEna,EAAEma,EAAE,EAAE,UAAU,mCAAmC,EAAEna,IAAI,EAAE,GAAGma,EAAE,EAAE,QAAQ,0CAA0C,IAAIna,EAAEJ,EAAE,GAAG,IAAIA,EAAEA,EAAE,EAAEI,EAAEA,EAAE,EAAEN,EAAEA,EAAE,IAAI,CAACiC,EAAEC,IAAIA,GAAGlC,EAAE,OAAO,EAAE8mB,EAAE,kBAAkB9mB,EAAEkC,EAAE,EAAElC,EAAEkC,EAAE,EAAElC,EAAE,SAAS,IAAIA,EAAE,CAAC,EAAEA,EAAE,EAAE,IAAIA,EAAE,SAAS,IAAIA,EAAE8mB,EAAE,aAAa9mB,CAAC,EAAE,UAAU,IAAIO,EAAEumB,EAAE,cAAc9mB,CAAC,EAAE,EAAE,KAAKA,EAAE,QAAQ,GAAGO,GAAGL,EAAE,EAAE,CAAC,EAAEK,CAAC,EAAEP,EAAE,SAAS,GAAGA,EAAE,IAAIE,GAAGF,EAAE,IAAIE,EAAE,EAAEF,EAAEA,EAAE,SAAS,GAAGA,EAAE,GAAGA,EAAE,IAAIE,GAAGF,EAAE,IAAIE,EAAE,EAAE,CAACF,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAEA,EAAE,SAAS,GAAGA,EAAE,IAAIE,GAAGF,EAAE,GAAGA,EAAE,IAAIE,EAAE,EAAE,CAACF,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAEA,EAAE,SAAS,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,IAAIE,GAAGF,EAAE,IAAIE,EAAE,EAAE,CAACF,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAEA,EAAE,SAAS,GAAGA,EAAE,IAAIE,GAAGF,EAAE,GAAGA,EAAE,GAAGA,EAAE,IAAIE,IAAI,EAAE,CAACF,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,GAAG,IAAI,EAAE,GAAG,MAAM,KAAK,IAAI,GAAG,CAAC,EAAEM,GAAG,KAAK,IAAI,GAAG,CAAC,IAAI,EAAE,EAAE,IAAI,KAAK,IAAI,GAAG,CAAC,EAAE,EAAE,GAAG,GAAG,MAAM,EAAE,GAAG,EAAE,CAAC,IAAI2B,EAAE6sF,GAAG9uF,CAAC,EAAEkC,EAAE,EAAEC,EAAE,EAAEnC,EAAE,SAAS,CAACkC,EAAEC,CAAC,EAAEktF,GAAGrvF,CAAC,GAAGO,EAAE0B,GAAGC,EAAE,IAAIC,EAAE,GAAG,EAAE2kB,EAAE,oBAAoBvmB,CAAC,EAAE,IAAIqC,GAAGA,EAAE,CAAC,CAAC,MAAM,EAAEkkB,EAAE,oBAAoBvmB,CAAC,EAAE,OAAO,CAAC,CAAC,SAASwxF,GAAG/xF,EAAE,CAAC,OAAOA,EAAE,IAAI,CAAC,CAAC,SAAS8vF,GAAG9vF,EAAE,EAAE,CAAC,GAAGA,EAAEA,EAAE,MAAM,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE,EAAE8mB,EAAE,YAAY9mB,EAAE,CAAC,GAAG,CAACA,EAAE,QAAQ,CAAC,EAAE,QAAQA,EAAE,KAAK,GAAGA,EAAE,KAAK,GAAG,EAAE,KAAK,GAAG,EAAE,KAAK,EAAE,MAAM,GAAG,GAAGA,EAAE,SAAS,EAAE,OAAO,CAAC,IAAIE,EAAEF,EAAE,MAAM,EAAE,EAAE,GAAGM,EAAE,EAAE,MAAM,EAAE,EAAE,GAAG,GAAGJ,IAAII,GAAGyxF,GAAG7xF,CAAC,GAAG6xF,GAAGzxF,CAAC,IAAIN,EAAE,KAAK,GAAG,EAAE,KAAK,GAAG,MAAM,EAAE,CAAC,OAAOA,EAAE,KAAK,EAAE,IAAI+xF,GAAG/xF,EAAE,EAAE,GAAG+xF,GAAG,EAAE,EAAE,CAAC,CAAC,IAAIC,GAAGC,GAAG,SAASvC,GAAG1vF,EAAE,CAAC,GAAGgyF,IAAI,KAAK,CAAC,IAAI,EAAEnB,GAAG7wF,CAAC,EAAEgyF,GAAG,EAAE,aAAa,EAAE,gBAAgB,CAAC,CAAC,OAAOA,EAAE,CAAC,SAAS7B,IAAK,CAAC6B,GAAG,IAAI,CAAC,SAAS5B,IAAK,CAAC6B,GAAG,IAAI,CAAC,SAAShD,GAAGjvF,EAAE,CAAC,GAAGiyF,IAAI,KAAK,CAAC,IAAI,EAAEpB,GAAG7wF,CAAC,EAAEiyF,GAAG,EAAE,aAAa,EAAE,uBAAuB,CAAC,CAAC,OAAO,KAAK,IAAI,GAAGA,EAAE,CAAC,CAAC,SAASzC,GAAGxvF,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAO,GAAE,IAAI,EAAEE,EAAE2wF,GAAG7wF,CAAC,EAAE,OAAO2vF,GAAGzvF,EAAE,iCAAiC,GAAGF,IAAI,EAAE,EAAE,EAAE2vF,GAAGzvF,EAAE,0BAA0B,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,SAASyvF,GAAG3vF,EAAE,EAAE,CAAC,OAAOA,EAAE,aAAa,CAAC,GAAG,IAAI,CAAC,SAASgwF,GAAGhwF,EAAE,CAAC,GAAG,CAAC,GAAG6wF,GAAG7wF,CAAC,GAAG,KAAK,MAAM,EAAE,OAAO,EAAN,CAAS,OAAO,QAAQ,IAAI,qCAAqC,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,CAAC,SAAS4vF,GAAG5vF,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,GAAG,IAAI,EAAE6wF,GAAG7wF,CAAC,EAAE,GAAGA,IAAI,GAAG,GAAG,CAAC2vF,GAAG,EAAE,mBAAmB,EAAE,MAAM,WAAW,CAACA,GAAG,EAAE,wBAAwB,EAAE,MAAM,GAAG,OAAOuC,GAAG,CAAC,CAAC,CAAC,SAASrC,GAAG7vF,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,GAAG,IAAI,EAAE6wF,GAAG7wF,CAAC,EAAE,GAAGA,IAAI,GAAG,GAAG,CAAC2vF,GAAG,EAAE,mBAAmB,GAAG,CAACA,GAAG,EAAE,0BAA0B,EAAE,MAAM,OAAO,CAAC,GAAGA,GAAG,EAAE,wBAAwB,EAAE,OAAOuC,GAAG,CAAC,EAAE,IAAI,EAAE,8BAA8B,GAAGvC,GAAG,EAAE,CAAC,EAAE,CAAC,IAAIpvF,EAAE,EAAE,aAAa,CAAC,EAAE,OAAO4xF,GAAI,EAAE5xF,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,OAAO2xF,GAAG,CAAC,CAAC,CAAC,SAASA,GAAGlyF,EAAE,CAAC,IAAI,EAAEwxF,GAAGxxF,CAAC,EAAEE,EAAEF,EAAE,cAAc,EAAEA,EAAE,YAAYA,EAAE,WAAWE,CAAC,EAAE,IAAII,EAAE,EAAEC,EAAE,EAAEP,EAAE,WAAWA,EAAE,WAAW,EAAE,EAAE,oBAAoBM,EAAEC,EAAE,EAAE,EAAE,mBAAmB,EAAE,iBAAiB,IAAI,EAAE,IAAI,EAAEP,EAAE,kBAAkB,EAAEA,EAAE,gBAAgBA,EAAE,YAAY,CAAC,EAAEA,EAAE,qBAAqBA,EAAE,YAAYA,EAAE,kBAAkBA,EAAE,WAAWE,EAAE,CAAC,EAAE,IAAI,EAAEF,EAAE,uBAAuBA,EAAE,WAAW,IAAIA,EAAE,qBAAqB,OAAOA,EAAE,YAAYA,EAAE,WAAW,IAAI,EAAEA,EAAE,gBAAgBA,EAAE,YAAY,IAAI,EAAEA,EAAE,cAAcE,CAAC,EAAEF,EAAE,kBAAkB,CAAC,EAAE,CAAC,CAAC,SAASmyF,GAAInyF,EAAE,EAAE,CAAC,IAAIE,EAAEsxF,GAAGxxF,EAAE,CAAC,EAAEM,EAAEN,EAAE,cAAc,EAAEA,EAAE,YAAYA,EAAE,WAAWM,CAAC,EAAE,IAAIC,EAAE,EAAE,EAAE,EAAEP,EAAE,WAAWA,EAAE,WAAW,EAAEE,EAAE,wBAAwBK,EAAE,EAAE,EAAEL,EAAE,mBAAmBA,EAAE,qBAAqB,IAAI,EAAE,IAAI,EAAEF,EAAE,kBAAkB,EAAEA,EAAE,gBAAgBA,EAAE,YAAY,CAAC,EAAEA,EAAE,qBAAqBA,EAAE,YAAYA,EAAE,kBAAkBA,EAAE,WAAWM,EAAE,CAAC,EAAE,IAAI2B,EAAEjC,EAAE,uBAAuBA,EAAE,WAAW,IAAIA,EAAE,qBAAqB,OAAOA,EAAE,YAAYA,EAAE,WAAW,IAAI,EAAEA,EAAE,gBAAgBA,EAAE,YAAY,IAAI,EAAEA,EAAE,cAAcM,CAAC,EAAEN,EAAE,kBAAkB,CAAC,EAAEiC,CAAC,CAAC,SAAS8tF,GAAG/vF,EAAE,CAAC,OAAOA,IAAI,EAAE,GAAG6wF,GAAG7wF,CAAC,EAAE,WAAW,IAAI,CAAC,SAAS+tF,GAAG/tF,EAAE,EAAE,CAAC,MAAM,QAAQA,CAAC,IAAIA,EAAE,CAACA,CAAC,GAAGA,EAAE,QAAQE,GAAG,CAACA,GAAG,MAAM4mB,EAAE,OAAO5mB,EAAE,QAAQ,YAAY,IAAI,GAAG,4DAA4D,CAAC,CAAC,CAAC,CAAC,IAAIkyF,GAAG33E,EAAE,EAAE23E,GAAG,aAAa,YAAY,IAAIA,GAAG,UAAU,eAAe,EAAE,CAAC,EAAEA,GAAG,aAAa,gBAAgB,IAAIpC,GAAG,CAAC,EAAE,EAAEA,GAAG,CAAC,EAAE,EAAE,CAAC,EAAEoC,GAAG,aAAa,iCAAiC,IAAI,EAAE,EAAEA,GAAG,aAAa,yBAAyB,IAAIA,GAAG,IAAI,eAAe,IAAI,CAAC,EAAEA,GAAG,aAAa,oBAAoB,IAAI,EAAE,EAAEA,GAAG,aAAa,2BAA2B,IAAI,EAAE,EAAEA,GAAG,aAAa,aAAa,IAAIA,GAAG,QAAQ,WAAW,CAAC,EAAEA,GAAG,aAAa,2BAA2B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,kBAAkB,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,2BAA2B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,+BAA+B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,8BAA8B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,8BAA8B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,8BAA8B,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,oBAAoB,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,sBAAsB,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,oBAAoB,IAAIA,GAAG,QAAQ,YAAY,CAAC,EAAEA,GAAG,aAAa,yBAAyB,IAAI1C,GAAG0C,GAAG,UAAU,eAAe,CAAC,CAAC,EAAEA,GAAG,aAAa,+BAA+B,IAAInD,GAAGmD,GAAG,UAAU,eAAe,CAAC,CAAC,EAAEA,GAAG,aAAa,+CAA+C,IAAI,CAAC,IAAIpyF,EAAEoyF,GAAG,UAAU,eAAe,EAAE,OAAOpyF,IAAI,EAAE,EAAEwvF,GAAGxvF,CAAC,CAAC,CAAC,EAAEoyF,GAAG,aAAa,gDAAgD,IAAIA,GAAG,UAAU,8CAA8C,EAAE,GAAG,CAAC9mE,GAAG,SAAS,CAAC,EAAE8mE,GAAG,aAAa,+BAA+B,IAAIxC,GAAGwC,GAAG,UAAU,eAAe,CAAC,CAAC,EAAEA,GAAG,aAAa,+BAA+B,IAAIA,GAAG,QAAQ,0BAA0B,EAAE,GAAGA,GAAG,QAAQ,8BAA8B,CAAC,EAAEA,GAAG,aAAa,+BAA+B,IAAIvC,GAAGuC,GAAG,UAAU,eAAe,CAAC,CAAC,EAAEA,GAAG,aAAa,0BAA0B,IAAIrC,GAAGqC,GAAG,UAAU,eAAe,CAAC,CAAC,EAAEA,GAAG,aAAa,4BAA4B,IAAIA,GAAG,QAAQ,8BAA8B,EAAE,EAAE,CAAC,EAAEA,GAAG,aAAa,iCAAiC,IAAI,GAAGpyF,GAAG,CAAC,GAAGA,EAAE,GAAGA,IAAI,GAAG,MAAM,IAAI,MAAM,8FAA8FA,IAAI,CAAC,CAAC,EAAEoyF,GAAG,aAAa,wBAAwB,IAAI9mE,GAAG,SAAS,EAAE,EAAE,GAAGtrB,GAAG,CAAC,GAAGA,EAAE,GAAGA,IAAI,GAAG,MAAM,IAAI,MAAM,2FAA2FA,IAAI,CAAC,CAAC,EAAEoyF,GAAG,aAAa,6BAA6B,IAAI,GAAG,EAAEA,GAAG,aAAa,4BAA4B,IAAI,EAAE,EAAEA,GAAG,aAAa,2CAA2C,IAAI,GAAG,EAAEA,GAAG,aAAa,+BAA+B,IAAI,GAAG,EAAEA,GAAG,aAAa,iBAAiB,IAAI,EAAE,EAAEA,GAAG,aAAa,yBAAyB,IAAIA,GAAG,QAAQ,SAAS,CAAC,EAAEA,GAAG,aAAa,oCAAoC,IAAI,EAAE,CAAC,EAAEA,GAAG,aAAa,2CAA2C,IAAI,EAAE,EAAEA,GAAG,aAAa,sBAAsB,IAAI,EAAE,EAAE,SAASC,IAAI,CAAC,IAAIryF,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,EAAE,OAAOsY,EAAE,EAAE,UAAU,eAAe,IAAI,GAAGza,EAAE,kBAAkB,EAAE,KAAKE,EAAE,MAAMI,EAAE,KAAKC,EAAE,UAAU,EAAE,cAAc,EAAE,wBAAwB0B,EAAEwY,EAAE,EAAE,QAAQ,qBAAqB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAYnoU,GAAGvY,EAAE,GAAGC,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QASRnC,EAAE,GAAG,EAAE,YAAYE,EAAE,UAAUI,EAAE,UAAUC,EAAE,YAAY,EAAE,eAAe,EAAE,GAAG0B,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAQnFC,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MASFC,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,OAQD,CAAC,QAAQnC,EAAE,UAAU,EAAE,UAAUE,EAAE,UAAUI,EAAE,UAAUC,EAAE,OAAO,EAAE,aAAa,EAAE,iBAAiB0B,EAAE,iBAAiBC,EAAE,YAAYC,CAAC,CAAC,CAAC,SAASmwF,GAAGtyF,EAAE,EAAEE,EAAE,QAAQ,CAAC,IAAII,EAAEwmB,EAAE,eAAe,CAAC,EAAE,OAAOxmB,EAAE,IAAI,CAACC,EAAE,IAAI,CAAC,IAAI,EAAE,OAAOP,EAAE,QAAQE,OAAOK,IAAI0B,EAAE,IAAI3B,EAAE,OAAO,EAAE,OAAON,EAAE,EAAE,QAAQE,OAAOF,EAAE,QAAQO,IAAI,YAAYP,EAAE,QAAQO,IAAI,MAAM,GAAG,MAAM0B,IAAI,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,SAASswF,GAAGvyF,EAAE,EAAEE,EAAE,QAAQ,CAAC,IAAII,EAAEwmB,EAAE,eAAe,CAAC,EAAE,OAAOxmB,EAAE,IAAI,CAACC,EAAE,IAAI,CAAC,IAAI,EAAE,OAAOP,EAAE,QAAQE,uBAAuB,KAAK+B,EAAE,IAAI3B,EAAE,OAAO,EAAE,OAAON,EAAE,EAAE,QAAQE,OAAOF,EAAE,wBAAwB,KAAK,YAAYA,EAAE,wBAAwB,KAAK,MAAM,GAAG,MAAMiC,IAAI,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,SAASuwF,GAAIxyF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAOM,EAAEN,EAAE,IAAI,GAAG,GAAG,KAAK,IAAI,EAAEO,EAAE,IAAI,MAAML,EAAE,CAAC,EAAEK,EAAEL,EAAE,GAAGI,EAAEJ,EAAE,GAAG,QAAQ,EAAEA,EAAE,EAAE,GAAG,EAAE,EAAE,EAAEK,EAAE,GAAG,IAAIA,EAAE,EAAE,QAAQD,EAAE,EAAE,MAAM,OAAOC,CAAC,CAAC,SAASkyF,GAAGzyF,EAAE,EAAEE,EAAE,QAAQ,CAAC,IAAII,EAAEN,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC,EAAEO,EAAEiyF,GAAIlyF,EAAE,CAAC,EAAE,OAAOC,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC,IAAI0B,EAAE,OAAOjC,EAAE,QAAQE,OAAOK,EAAE,KAAK2B,EAAE,IAAI3B,EAAE,OAAO,EAAE,OAAOP,EAAE,EAAE,QAAQE,OAAOF,EAAE,QAAQO,EAAE,KAAK,YAAYP,EAAE,QAAQO,EAAE,KAAK,MAAM,GAAG0B,MAAMC,IAAI,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,SAASwwF,GAAG1yF,EAAE,CAAC,IAAI,EAAE8mB,EAAE,eAAe9mB,CAAC,EAAE,IAAIE,GAAGA,EAAE,SAAS,CAAC,EAAE,MAAM;AAAA;AAAA,wBAEhjC,EAAE,mBAAmB,EAAE;AAAA;AAAA,CAE9C,CAAC,SAASyyF,IAAI,CAAC,MAAM;AAAA;AAAA;AAAA;AAAA,CAIrB,CAAC,IAAIC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAuCJ,CAAC,iBAAiBC,EAAE,EAAEr4C,EAAE,SAASs4C,GAAG9yF,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAE,GAAGN,EAAE,QAAQ+C,GAAG,CAAC,IAAIC,EAAE8jB,EAAE,cAAc/jB,EAAE,UAAU,YAAY,EAAE,GAAGA,EAAE,UAAU,UAAUzC,EAAE,KAAK,iBAAiByC,EAAE,OAAOC,EAAE,EAAE,IAAIA,KAAK,KAAK,GAAG1C,EAAE,KAAK,qBAAqByC,EAAE,OAAO,EAAEzC,EAAE,KAAK,qBAAqByC,EAAE,OAAO,GAAG7C,EAAE,oBAAoB,CAAC,GAAG,CAAC,aAAagE,CAAC,EAAE6uF,GAAG7yF,EAAE,aAAa6C,EAAE,UAAU,aAAaA,EAAE,UAAU,QAAQ,EAAE,OAAOmB,EAAE,OAAO,CAAC,IAAK,GAAE5D,EAAE,KAAK,eAAeyC,EAAE,YAAY,EAAE,MAAM,IAAK,GAAEzC,EAAE,KAAK,iBAAiByC,EAAE,YAAY,EAAE,MAAM,IAAK,GAAEzC,EAAE,KAAK,iBAAiByC,EAAE,YAAY,EAAE,MAAM,IAAK,GAAEzC,EAAE,KAAK,iBAAiByC,EAAE,YAAY,EAAE,MAAM,QAAQ,KAAK,CAACzC,EAAE,KAAK,iBAAiByC,EAAE,eAAe,CAAC,CAAC,CAAC,EAAE7C,EAAE,oBAAoB,CAAC,OAAO,EAAE,aAAa,OAAO,CAAC,IAAK,GAAEI,EAAE,KAAK,uBAAuB,EAAE,MAAM,IAAK,GAAEA,EAAE,KAAK,yBAAyB,EAAEA,EAAE,KAAK,8BAA8B,EAAE,MAAM,IAAK,GAAEA,EAAE,KAAK,yBAAyB,EAAEA,EAAE,KAAK,gCAAgC,EAAE,MAAM,IAAK,GAAEA,EAAE,KAAK,yBAAyB,EAAEA,EAAE,KAAK,gCAAgC,EAAE,MAAM,QAAQ,KAAK,CAACA,EAAE,KAAK,4BAA4B,CAAC,CAACJ,EAAE,gBAAgBA,EAAE,eAAe,QAAQ6C,GAAG,CAACzC,EAAE,KAAK,WAAWyC,EAAE,QAAQA,EAAE,OAAOA,EAAE,WAAW,IAAIA,EAAE,cAAc,KAAK,CAAC,CAAC,EAAE,IAAIxC,EAAED,EAAE,KAAK;AAAA,CACxsC,EAAE,EAAEN,EAAE,IAAI+C,GAAGiwF,GAAIjwF,EAAE,EAAE7C,EAAE,aAAaA,EAAE,mBAAmB,CAAC,EAAE,KAAK;AAAA,CACjE,EAAE,EAAE,EAAE,SAAS+B,EAAEowF,GAAG,EAAEnwF,EAAE+wF,GAAIhxF,CAAC,EAAEE,EAAES,EAAEC,EAAEqwF,GAAIjxF,CAAC,EAAE,OAAO,EAAE,UAAUE,EAAEgxF,GAAI,EAAE,aAAa,EAAEjzF,EAAE,mBAAmB,EAAE0C,EAAEwwF,GAAInxF,CAAC,IAAIE,EAAEkxF,GAAI,EAAE,aAAa,EAAEnzF,EAAE,mBAAmB,EAAE0C,EAAE0wF,GAAIrxF,CAAC,GAAG/B,EAAE,eAAe2C,GAAG0wF,IAAK,CAAC1wF,EAAEX,EAAEU,EAAErC,EAAE4B,EAAE,EAAEjC,EAAE,QAAQ,EAAE,KAAK;AAAA,CACvO,CAAC,CAAC,SAASszF,GAAGxzF,EAAE,EAAE,GAAG,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAa,OAAOE,EAAE,OAAO,CAAC,IAAK,GAAE,OAAOuzF,GAAIzzF,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO0zF,GAAI1zF,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO2zF,GAAI3zF,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO4zF,GAAI5zF,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO6zF,GAAI7zF,EAAE,CAAC,EAAE,IAAK,GAAE,OAAO8zF,GAAI9zF,CAAC,EAAE,IAAK,GAAE,OAAO+zF,GAAI/zF,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,GAAGE,EAAE,8CAA8C,CAAC,CAAC,CAAC,SAAS8zF,GAAGh0F,EAAE,EAAE,CAAC,OAAOA,EAAE,UAAU,aAAa,OAAO,CAAC,IAAK,GAAE,OAAOi0F,GAAIj0F,CAAC,EAAE,IAAK,GAAE,OAAOk0F,GAAIl0F,EAAE,CAAC,EAAE,IAAK,GAAE,OAAOm0F,GAAIn0F,EAAE,CAAC,EAAE,IAAK,GAAE,OAAOo0F,GAAIp0F,EAAE,CAAC,EAAE,QAAQ,OAAOq0F,GAAIr0F,EAAE,CAAC,CAAC,CAAC,CAAC,SAASgzF,GAAIhzF,EAAE,EAAEE,EAAE,GAAGI,EAAE,CAAC,IAAIC,EAAE,GAAGL,EAAEK,GAAGyzF,GAAGh0F,EAAEM,CAAC,EAAEC,GAAGizF,GAAGxzF,EAAEM,CAAC,EAAE,IAAI,EAAEN,EAAE,UAAU,aAAa,EAAE,EAAE,aAAa,OAAO,EAAE,QAAQ,EAAE,SAASE,EAAEK,GAAG+zF,GAAIt0F,EAAE,CAAC,EAAEO,GAAGg0F,GAAIv0F,EAAE,CAAC,GAAGO,CAAC,CAAC,SAAS4yF,GAAInzF,EAAE,EAAEE,EAAE,CAAC,OAAOF,EAAE,OAAO,CAAC,IAAK,GAAE,OAAOw0F,GAAG,EAAE,IAAK,GAAE,OAAOC,GAAIz0F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOw0F,GAAI10F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAOy0F,GAAI30F,EAAE,EAAEE,CAAC,EAAE,QAAQ,OAAO00F,GAAI50F,EAAE,EAAEE,CAAC,CAAC,CAAC,CAAC,SAASmzF,GAAIrzF,EAAE,EAAEE,EAAE,CAAC,OAAOF,EAAE,OAAO,CAAC,IAAK,GAAE,OAAOw0F,GAAG,EAAE,IAAK,GAAE,OAAOK,GAAI70F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAO40F,GAAI90F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAO60F,GAAI/0F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAO80F,GAAIh1F,EAAE,EAAEE,CAAC,EAAE,IAAK,GAAE,OAAO+0F,GAAIj1F,EAAE,CAAC,EAAE,IAAK,GAAE,OAAOk1F,GAAIl1F,EAAE,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,GAAGA,EAAE,+CAA+C,CAAC,CAAC,CAAC,SAASizF,GAAIjzF,EAAE,CAAC,MAAM;AAAA;AAAA,eAE/jCA,EAAE;AAAA;AAAA,GAEd,CAAC,SAASszF,GAAItzF,EAAE,CAAC,MAAM;AAAA;AAAA,QAElBA,EAAE;AAAA;AAAA,GAEP,CAAC,SAASozF,GAAIpzF,EAAE,CAAC,MAAM;AAAA;AAAA,QAElBA,EAAE;AAAA;AAAA,GAEP,CAAC,SAASkzF,GAAIlzF,EAAE,CAAC,MAAM,GAAGA,EAAE;AAAA;AAAA;AAAA;AAAA,MAIzBA,EAAE;AAAA,MACFA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAuBFA,EAAE;AAAA,MACFA,EAAE;AAAA,MACFA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAyBFm1F;AAAA,MACAC;AAAA,MACAC;AAAA,GACH,CAAC,IAAIF,GAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYVC,GAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQJC,GAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASJ9B,GAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWJ,SAASiB,IAAI,CAAC,MAAM;AAAA;AAAA;AAAA;AAAA,GAInB,CAAC,SAASC,GAAIz0F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAE,OAAOA,EAAE,KAAK,EAAEJ,EAAE;AAAA;AAAA;AAAA;AAAA,MAIhF;AAAA;AAAA,sCAEgCI,EAAE;AAAA;AAAA,MAElCA,EAAE,KAAK,EAAEJ,EAAE;AAAA;AAAA;AAAA;AAAA,MAIX;AAAA;AAAA,sCAEgCI,EAAE;AAAA;AAAA,MAElCJ,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAOJ;AAAA;AAAA;AAAA,oCAGgCI,EAAE,OAAOA,EAAE;AAAA,iCACdA,EAAE;AAAA;AAAA,GAEhC,CAAC,SAASu0F,GAAI70F,EAAE,EAAEE,EAAE,CAAC,OAAO,EAAE,KAAK,EAAEA,EAAE;AAAA;AAAA;AAAA;AAAA,MAIpC;AAAA;AAAA,kCAE4B,EAAE;AAAA;AAAA,MAE9B,EAAE,KAAK,EAAEA,EAAE;AAAA;AAAA;AAAA;AAAA,MAIX;AAAA;AAAA,kCAE4B,EAAE;AAAA;AAAA,MAE9BA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAMJ;AAAA;AAAA;AAAA,oCAGgC,EAAE,OAAO,EAAE;AAAA,4BACnB,EAAE;AAAA;AAAA,GAE3B,CAAC,SAASy0F,GAAI30F,EAAE,EAAEE,EAAE,CAAC,GAAGA,EAAE,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAiB/B,IAAII,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAEC,EAAE,KAAK,KAAKP,EAAE,GAAG,CAAC,EAAE,EAAEO,EAAE,KAAK,KAAKP,EAAE,GAAG,CAAC,EAAE,MAAM;AAAA;AAAA;AAAA,oCAG5DM,EAAE,OAAOA,EAAE;AAAA,iCACdA,EAAE;AAAA;AAAA,wBAEX;AAAA,qBACH;AAAA;AAAA,6BAEQC;AAAA,4BACDA;AAAA;AAAA;AAAA;AAAA,GAIzB,CAAC,SAASw0F,GAAI/0F,EAAE,EAAEE,EAAE,CAAC,GAAGA,EAAE,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA,MAK7BqyF,GAAG,CAAC,IAAI,IAAI,GAAG,EAAEvyF,CAAC;AAAA;AAAA;AAAA,EAGtB,IAAIM,EAAEgyF,GAAG,CAAC,IAAI,IAAI,GAAG,EAAEtyF,CAAC,EAAE,MAAM;AAAA;AAAA;AAAA,oCAGE,EAAE,OAAO,EAAE;AAAA,iCACd,EAAE;AAAA,QAC3BM;AAAA;AAAA;AAAA,GAGL,CAAC,SAASs0F,GAAI50F,EAAE,EAAEE,EAAE,CAAC,GAAGA,EAAE,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAsB/B,IAAII,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAEC,EAAE,KAAK,KAAKP,EAAEA,EAAE,OAAO,GAAG,CAAC,EAAE,EAAEO,EAAE,KAAK,KAAKP,EAAEA,EAAE,OAAO,GAAG,CAAC,EAAE,EAAE,EAAEiC,EAAE,GAAGC,EAAE,UAAU,QAAQC,EAAE,EAAEA,EAAEnC,EAAE,OAAO,EAAEmC,IAAI,GAAGnC,EAAEA,EAAE,OAAOmC,EAAE,GAAGF,EAAE;AAAA,aACtKE,eAAe;AAAA,kBACVA,OAAO;AAAA,MACnBF,EAAEC,EAAE,IAAIC,MAAMD,EAAE,MAAM;AAAA,UAClBlC,EAAE;AAAA;AAAA,oCAEwBM,EAAE,OAAOA,EAAE;AAAA,iCACdA,EAAE;AAAA;AAAA,QAE3B2B;AAAA;AAAA,wBAEgB;AAAA,qBACH;AAAA;AAAA,6BAEQ1B;AAAA,4BACDA;AAAA;AAAA,mBAETP,EAAE,UAAUkC;AAAA;AAAA,GAE5B,CAAC,SAAS8yF,GAAIh1F,EAAE,EAAEE,EAAE,CAAC,GAAGA,EAAE,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA,QAK3BqyF,GAAG,CAAC,IAAI,IAAI,IAAI,IAAI,EAAEvyF,CAAC;AAAA;AAAA;AAAA,IAG3B,IAAIM,EAAEgyF,GAAG,CAAC,IAAI,IAAI,IAAI,IAAI,EAAEtyF,CAAC,EAAE,MAAM;AAAA;AAAA;AAAA,eAG1B,EAAE,OAAO,EAAE;AAAA,iCACO,EAAE;AAAA,QAC3BM;AAAA;AAAA;AAAA,GAGL,CAAC,SAAS20F,GAAIj1F,EAAE,EAAE,CAAC,IAAIE,EAAEoyF,GAAG,CAAC,IAAI,IAAI,IAAI,KAAK,IAAI,EAAEtyF,CAAC,EAAE,MAAM;AAAA;AAAA,kDAEd,EAAE;AAAA,+BACrB,EAAE;AAAA;AAAA,iCAEA,EAAE;AAAA;AAAA,QAE3BE;AAAA;AAAA;AAAA;AAAA;AAAA,GAKL,CAAC,SAASg1F,GAAIl1F,EAAE,EAAE,CAAC,IAAIE,EAAEoyF,GAAG,CAAC,IAAI,IAAI,IAAI,KAAK,KAAK,IAAI,EAAEtyF,CAAC,EAAE,MAAM;AAAA;AAAA;AAAA,eAGtD,EAAE,OAAO,EAAE;AAAA,iCACO,EAAE;AAAA;AAAA,QAE3BE;AAAA;AAAA;AAAA;AAAA;AAAA,GAKL,CAAC,SAASw0F,GAAI10F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAE,GAAGwmB,EAAE,YAAY9mB,EAAE,CAAC,EAAE,OAAOE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,MAK7F;AAAA;AAAA,8CAEwCI,EAAE,OAAOA,EAAE;AAAA;AAAA,MAEnD,IAAIC,EAAE,KAAK,KAAKP,EAAE,GAAG,CAAC,EAAE,OAAOE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAanC;AAAA;AAAA;AAAA,oCAGgCI,EAAE,OAAOA,EAAE;AAAA;AAAA,iCAEdA,EAAE;AAAA,6BACNC;AAAA,4BACDA;AAAA;AAAA;AAAA;AAAA,GAIzB,CAAC,SAASu0F,GAAI90F,EAAE,EAAEE,EAAE,CAAC,OAAO4mB,EAAE,YAAY9mB,EAAE,CAAC,EAAEE,EAAE;AAAA;AAAA;AAAA;AAAA,MAI9C;AAAA;AAAA,0CAEoC,EAAE,OAAO,EAAE;AAAA;AAAA,MAE/CF,EAAE,KAAK,EAAEE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAOX;AAAA;AAAA;AAAA,sCAGgC,EAAE,OAAO,EAAE;AAAA,mCACd,EAAE;AAAA;AAAA;AAAA,MAG/BF,EAAE,KAAK,EAAEE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAOX;AAAA;AAAA;AAAA,sCAGgC,EAAE,OAAO,EAAE;AAAA,mCACd,EAAE;AAAA;AAAA;AAAA,MAG/BA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IASJ;AAAA;AAAA;AAAA,oCAGgC,EAAE,OAAO,EAAE;AAAA,iCACd,EAAE;AAAA,wBACXF,EAAE;AAAA,4BACEA,EAAE;AAAA;AAAA;AAAA,GAG3B,CAAC,SAASs1F,GAAGt1F,EAAE,CAAC,MAAM,SAASA,GAAG,CAAC,SAASi0F,GAAIj0F,EAAE,CAAC,IAAI,EAAEA,EAAE,KAAKE,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,YAAY,EAAE,EAAE,MAAM,CAAC,EAAEI,EAAE+xF,GAAG,EAAE,MAAM;AAAA,WAClHnyF;AAAA,eACII,EAAE,aAAa;AAAA;AAAA,GAE3B,CAAC,SAASmzF,GAAIzzF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,GAAGF,EAAE,UAAU,UAAU,MAAM,SAASM,eAAeJ,MAAM,GAAG,CAACK,EAAE,CAAC,EAAEP,EAAE,UAAU,SAAS,GAAGO,IAAI,GAAG,IAAI,EAAE,MAAM;AAAA,cACnLD;AAAA,+BACiBJ;AAAA;AAAA,MAEzB,IAAI,EAAEo1F,GAAGp1F,CAAC,EAAE,GAAG,EAAE,MAAM;AAAA,YACjBI;AAAA,6BACiBJ,iBAAiBA,iBAAiB;AAAA,6BAClCA;AAAA;AAAA,IAEzB,GAAG,CAAC+B,EAAEC,CAAC,EAAElC,EAAE,UAAU,SAAS,MAAM;AAAA,YAC5BM;AAAA,6BACiB2B,MAAMC,MAAM;AAAA,6BACZhC;AAAA;AAAA,GAE1B,CAAC,SAASg0F,GAAIl0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEK,EAAEP,EAAE,UAAU,SAAS,EAAEqyF,GAAG,EAAE,GAAG,EAAE,MAAM;AAAA,WAC9G/xF;AAAA,gDACqCJ,oCAAoCA;AAAA;AAAA;AAAA,eAGrE,EAAE,aAAaA;AAAA;AAAA,IAE1B,IAAI,EAAE,CAAC,KAAK,KAAKK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAKA,EAAE,GAAG,CAAC,CAAC,EAAE,MAAM;AAAA,WAC3CD;AAAA;AAAA,UAED,EAAE,OAAO,EAAE;AAAA,eACN,EAAE,aAAaJ;AAAA;AAAA,GAE3B,CAAC,SAASwzF,GAAI1zF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,GAAGF,EAAE,UAAU,UAAU,MAAM;AAAA,cACjGM;AAAA,UACJi1F,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAIO,EAAEP,EAAE,UAAU,SAAS,EAAEO,EAAE,GAAG,EAAEA,EAAE,GAAG,GAAG,IAAI,GAAG,IAAI,EAAE,MAAM;AAAA,cACvDD;AAAA,+BACiBJ;AAAA;AAAA,MAEzB,IAAI+B,EAAEqzF,GAAGp1F,CAAC,EAAE,OAAO,IAAI,EAAE,EAAE;AAAA,cACnBI;AAAA,6CAC+B2B,qBAAqB/B;AAAA,+BACnCA;AAAA;AAAA,MAEzB;AAAA,cACQI;AAAA,6CAC+B2B,eAAe;AAAA,+BAC7B/B;AAAA;AAAA,MAEzB,IAAI,EAAE,EAAE;AAAA,cACAI;AAAA,wCAC0B2B,qBAAqB/B;AAAA,+BAC9BA;AAAA;AAAA,MAEzB;AAAA,cACQI;AAAA,wCAC0B2B,eAAe;AAAA,+BACxB/B;AAAA;AAAA,MAEzB,EAAE;AAAA,YACII;AAAA,6BACiBJ,iBAAiBA,yBAAyB+B;AAAA,6BAC1C/B;AAAA;AAAA,IAEzB;AAAA,YACQI;AAAA,6BACiB,MAAM,cAAc2B;AAAA,6BACpB/B;AAAA;AAAA,GAE1B,CAAC,SAASi0F,GAAIn0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAaM,EAAEN,EAAE,KAAKO,EAAE,MAAMD,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,EAAEN,EAAE,UAAU,SAAS,EAAE,EAAE,GAAGiC,EAAE,EAAE,GAAGC,EAAEmwF,GAAG,EAAE,GAAG,GAAG,MAAMvrE,EAAE,YAAY5mB,EAAE,CAAC,EAAE,OAAO,EAAE;AAAA,aAClLK;AAAA,qDACwCD,iBAAiBA;AAAA;AAAA,iBAErD4B,EAAE,aAAa5B;AAAA;AAAA,MAE1B;AAAA,aACOC;AAAA,qDACwC0B,QAAQ;AAAA;AAAA,iBAE5CC,EAAE,aAAa5B;AAAA;AAAA,MAE1B,GAAG,EAAE,MAAM;AAAA,WACNC;AAAA,gDACqCD,oCAAoCA;AAAA,0CAC1CA;AAAA;AAAA,eAE3B4B,EAAE,aAAa5B;AAAA;AAAA,IAE1B,IAAI6B,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAES,EAAE,KAAK,KAAK1C,EAAE,GAAG,CAAC,EAAE,MAAM;AAAA,WAC/DK;AAAA,iCACsBqC,MAAMT,EAAE,OAAOA,EAAE;AAAA,eACnCD,EAAE,aAAa5B;AAAA;AAAA,GAE3B,CAAC,SAASqzF,GAAI3zF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAaM,EAAEN,EAAE,KAAKO,EAAE,MAAMD,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,EAAEN,EAAE,UAAU,SAAS,GAAG,GAAG,MAAM8mB,EAAE,YAAY5mB,EAAE,CAAC,EAAE,CAAC,GAAG,EAAE,MAAM;AAAA,cAC/JK;AAAA,qDACuCD,iBAAiBA;AAAA,+BACvCA;AAAA;AAAA,MAEzB,IAAIwC,EAAE,EAAE,GAAGC,EAAE,EAAE,GAAG,MAAM;AAAA,YAClBxC;AAAA,mDACuCwC,QAAQD;AAAA,6BAC9BxC;AAAA;AAAA,GAE1B,CAAC,GAAG,CAAC,SAAS,EAAE,SAAS2B,CAAC,EAAE6kB,EAAE,aAAa5mB,CAAC,EAAEgC,EAAE,EAAE,GAAGA,EAAE,OAAOhC,EAAE,OAAO,CAAC,IAAI4C,EAAE0yF,GAAGx1F,EAAEkC,CAAC,EAAEa,EAAE,CAAC,MAAM,KAAK,EAAE,MAAM;AAAA,QACvGywF,GAAG1wF,EAAE,CAAC;AAAA,cACAvC;AAAA,iBACGA,KAAKk1F,GAAG1yF,EAAEd,CAAC;AAAA;AAAA,KAEvB,CAAC,GAAGjC,EAAE,UAAU,UAAU,MAAM;AAAA,cACvBO;AAAA,qDACuCL,EAAE;AAAA,UAC7Cq1F,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAImC,EAAE,EAAE,GAAGS,EAAE,EAAE,GAAGC,EAAEyyF,GAAGh1F,CAAC,EAAE,OAAOsC,IAAI,EAAE,EAAE;AAAA,cACjCrC;AAAA,2CAC6BsC,YAAYvC;AAAA,oDACHA;AAAA,+BACrBA;AAAA;AAAA,MAEzB;AAAA,YACMC;AAAA,yCAC6BsC,YAAY3C,EAAE;AAAA,4CACXiC;AAAA,6BACf7B;AAAA;AAAA,IAEzB6B,IAAI,EAAE,EAAE;AAAA,cACE5B;AAAA,2CAC6BsC,YAAYvC;AAAA,+CACRA;AAAA,+BAChBA;AAAA;AAAA,MAEzB;AAAA,YACMC;AAAA,yCAC6BsC,YAAY3C,EAAE;AAAA,uCAChB0C;AAAA,6BACVtC;AAAA;AAAA,IAEzB,EAAE;AAAA,cACQC;AAAA;AAAA,4BAEcD,qBAAqBuC;AAAA,+BAClBvC,iBAAiBA;AAAA,+BACjBA;AAAA;AAAA,MAEzB;AAAA,UACIC;AAAA;AAAA,wBAEcL,EAAE,cAAc2C;AAAA,2BACbV,MAAMS;AAAA,2BACNtC;AAAA;AAAA,CAE1B,CAAC,SAAS8zF,GAAIp0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAaM,EAAEN,EAAE,KAAKO,EAAE,MAAMD,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,EAAEN,EAAE,UAAU,SAAS,EAAE,CAAC,KAAK,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,KAAK,EAAE,GAAG,CAAC,CAAC,EAAE,GAAGE,EAAE,KAAK,EAAE,CAAC,IAAI4C,EAAE5C,EAAE,MAAM,CAAC,EAAE6C,EAAE,CAAC,EAAE,CAAC,EAAEC,EAAEwyF,GAAGx1F,EAAE8C,CAAC,EAAEoB,EAAE,CAAC,IAAI,MAAM,KAAK,EAAE,MAAM;AAAA,UACxO8vF,GAAGhxF,EAAE,CAAC;AAAA,eACDzC;AAAA,mBACIA,KAAKk1F,GAAGvxF,EAAEnB,CAAC;AAAA;AAAA,OAEvB,CAAC,IAAId,EAAEowF,GAAG,EAAE,GAAG,EAAE,MAAM;AAAA,WACnB9xF;AAAA,gDACqCD,oCAAoCA;AAAA,0CAC1CA;AAAA,0DACgBA;AAAA;AAAA;AAAA,eAG3C2B,EAAE,aAAa3B;AAAA;AAAA,IAE1B,IAAI4B,EAAE,EAAE,GAAGC,EAAE,EAAE,GAAGS,EAAE,KAAK,KAAK1C,EAAE,GAAG,CAAC,EAAE2C,EAAED,EAAE,KAAK,KAAK1C,EAAE,GAAG,CAAC,EAAE,MAAM;AAAA,WAC3DK;AAAA;AAAA,UAED2B,MAAMC,MAAMU,MAAMD;AAAA,eACbX,EAAE,aAAa3B;AAAA;AAAA,GAE3B,CAAC,SAASszF,GAAI5zF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAaM,EAAEN,EAAE,KAAKO,EAAE,MAAMD,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,EAAEJ,EAAE,GAAGA,EAAE,GAAG,EAAEA,EAAE,GAAG,CAAC,SAAS+B,EAAE,SAASC,CAAC,EAAE4kB,EAAE,aAAa5mB,CAAC,EAAEiC,EAAEF,EAAE,GAAGE,EAAE,OAAOjC,EAAE,OAAO,CAAC,IAAIgE,EAAEsxF,GAAGx1F,EAAEmC,CAAC,EAAE,EAAE,CAAC,MAAM,MAAM,OAAO,EAAE,MAAM;AAAA,UACpOqxF,GAAGtvF,EAAE,CAAC;AAAA,gBACA3D;AAAA,mBACGA,KAAKk1F,GAAG,EAAEvzF,CAAC;AAAA;AAAA,OAEvB,CAAC,GAAGlC,EAAE,UAAU,UAAU,MAAM;AAAA,cACzBO;AAAA;AAAA,iCAEmB,MAAM;AAAA,UAC7Bg1F,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAI4C,EAAE5C,EAAE,UAAU,SAAS6C,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAGG,EAAE/C,EAAE,UAAU,WAAW,GAAG8C,IAAI,GAAGC,GAAG,KAAK,OAAO,EAAE;AAAA,cACrFxC;AAAA,wBACUD;AAAA;AAAA;AAAA;AAAA,0BAIEA,iBAAiBA;AAAA,+BACZA;AAAA;AAAA,MAEzB;AAAA,gBACUC;AAAA;AAAA,oDAEoC;AAAA;AAAA,4BAExBuC,QAAQD;AAAA,iCACHvC;AAAA;AAAA,QAEzB,GAAGwC,IAAI,GAAGC,GAAG,KAAK,OAAO,EAAE;AAAA,cACrBxC;AAAA,gDACkCD;AAAA;AAAA,uDAEOA,iBAAiBA;AAAA,+BACzCA;AAAA;AAAA,MAEzB;AAAA,YACMC;AAAA,8CACkCL,EAAE;AAAA;AAAA,qDAEK4C,QAAQD;AAAA,6BAChCvC;AAAA;AAAA,IAEzB,IAAI0C,EAAEsyF,GAAGh1F,CAAC,EAAE,OAAO,EAAE;AAAA,YACbC;AAAA;AAAA,sBAEUD,eAAeA;AAAA,sBACfA;AAAA,4DACsC0C;AAAA,6BAC/B1C,iBAAiBA;AAAA,6BACjBA;AAAA;AAAA,MAEvB;AAAA,cACQC;AAAA;AAAA,4BAEc,aAAa,eAAeyC;AAAA,+BACzBH,MAAMC;AAAA,+BACNxC;AAAA;AAAA,GAE5B,CAAC,SAAS+zF,GAAIr0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEK,EAAE8xF,GAAG,EAAE,GAAG,EAAE,MAAM;AAAA,WACvF/xF;AAAA,0CAC+BJ;AAAA,0DACgBA;AAAA;AAAA,yBAEjCA;AAAA;AAAA,gDAEuBA,oCAAoCA;AAAA;AAAA;AAAA,mGAGeK,EAAE,aAAaL;AAAA;AAAA,IAE9G,IAAI,EAAEF,EAAE,UAAU,aAAa,EAAE,EAAE,OAAOiC,EAAEjC,EAAE,UAAU,SAASkC,EAAE,CAAC,KAAK,KAAKD,EAAE,GAAG,CAAC,EAAE,KAAK,KAAKA,EAAE,GAAG,CAAC,CAAC,EAAEE,EAAED,EAAE,GAAGU,EAAEV,EAAE,GAAGW,EAAE,KAAK,KAAK,EAAE,EAAE,GAAG,CAAC,EAAEC,EAAED,EAAE,KAAK,KAAK,EAAE,EAAE,GAAG,CAAC,EAAEE,EAAE,0BAA0BC,EAAE,OAAOF,mBAAmBD,gBAAgB,QAAQqB,EAAE,EAAEA,EAAE,EAAE,EAAEA,IAAInB,EAAE,QAAQmB,MAAMnB,EAAED,GAAG,EAAE,EAAEoB,EAAE,GAAGlB,EAAE,IAAIkB,OAAOpB,OAAOE,EAAE,MAAM;AAAA,WACtT1C,KAAKyC;AAAA,oBACIC;AAAA,2BACOJ;AAAA,kCACOA;AAAA,qDACmBA,MAAMT;AAAA,eAC5C5B,EAAE,aAAaL;AAAA;AAAA,GAE3B,CAAC,SAAS2zF,GAAI7zF,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,UAAU,aAAaM,EAAEN,EAAE,KAAKO,EAAE,MAAMD,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,EAAEJ,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAE+B,EAAE/B,EAAE,GAAG,EAAE,CAAC,SAASgC,EAAE,SAASC,CAAC,EAAE2kB,EAAE,aAAa5mB,CAAC,EAAE,GAAGgC,EAAE,OAAOhC,EAAE,OAAO,CAAC,IAAI,EAAEs1F,GAAGx1F,EAAEkC,CAAC,EAAEyC,EAAE,CAAC,MAAM,MAAM,QAAQ,QAAQ,EAAE,MAAM;AAAA,QACjP6uF,GAAG,EAAE,CAAC;AAAA,cACAjzF;AAAA,iBACGA,KAAKk1F,GAAG9wF,EAAExC,CAAC;AAAA;AAAA,KAEvB,CAAC,GAAGnC,EAAE,UAAU,UAAU,MAAM;AAAA,cACvBO;AAAA;AAAA,iCAEmB0B,MAAM,MAAM;AAAA,UACnCszF,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAI4C,EAAE5C,EAAE,UAAU,WAAW6C,EAAE7C,EAAE,UAAU,SAAS8C,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAGG,EAAE,iBAAiB1C,aAAa4D,EAAE,iBAAiB5D,uBAAuB,EAAE,iBAAiBA,uBAAuB,GAAGyC,IAAId,GAAGW,GAAG,KAAK,OAAO,EAAE;AAAA,cACzMrC;AAAA,UACJyC;AAAA,UACAkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAMgB5D,iBAAiBA;AAAA,+BACZA;AAAA;AAAA,MAEzB;AAAA,cACQC;AAAA;AAAA;AAAA;AAAA,uBAIS,MAAM;AAAA;AAAA,0BAEHwC,QAAQD;AAAA,+BACHxC;AAAA;AAAA,MAEzB,GAAGyC,IAAI,GAAGH,GAAG,KAAK,OAAO,EAAE;AAAA,cACnBrC;AAAA;AAAA,gCAEkBD,eAAeA,cAAcA;AAAA;AAAA;AAAA,yBAGpCA,iBAAiBA;AAAA,+BACXA;AAAA;AAAA,MAEzB;AAAA,cACQC;AAAA;AAAA,gCAEkBL,EAAE,GAAGA,EAAE,OAAOA,EAAE;AAAA;AAAA;AAAA,yBAGvB6C,QAAQD;AAAA,+BACFxC;AAAA;AAAA,MAEzB,IAAIiH,EAAE+tF,GAAGh1F,CAAC,EAAE,OAAO,EAAE;AAAA,YACfC;AAAA;AAAA,QAEJyC;AAAA,QACAkB;AAAA,QACA;AAAA;AAAA;AAAA,6BAGqB5D,iBAAiBA,yBAAyBiH;AAAA,6BAC1CjH;AAAA;AAAA,IAEzB;AAAA,YACQC;AAAA;AAAA,0BAEc0B,aAAa;AAAA,oBACnB;AAAA,6BACSa,MAAMC,cAAcwE;AAAA,6BACpBjH;AAAA;AAAA,GAE1B,CAAC,SAASwzF,GAAI9zF,EAAE,CAAC,IAAI,EAAEA,EAAE,UAAU,aAAaE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEK,EAAE,EAAE,GAAG,EAAE,EAAE,GAAGA,EAAE,EAAE,EAAE,GAAG,EAAE0B,EAAE,EAAE,GAAG,EAAE,CAAC,SAASC,EAAE,SAASC,CAAC,EAAE2kB,EAAE,aAAa,CAAC,EAAE,GAAG5kB,EAAE,OAAO,EAAE,OAAO,CAAC,IAAIgC,EAAEsxF,GAAGx1F,EAAEkC,CAAC,EAAE,EAAE,CAAC,MAAM,MAAM,QAAQ,SAAS,QAAQ,EAAE,MAAM;AAAA,QACjQsxF,GAAGtvF,CAAC;AAAA,cACE5D;AAAA,iBACGA,KAAKm1F,GAAG,EAAEtzF,CAAC;AAAA;AAAA,KAEvB,CAAC,GAAGnC,EAAE,UAAU,UAAU,MAAM;AAAA,cACvBM;AAAA;AAAA;AAAA,iBAGG2B,MAAM,MAAM,MAAM1B;AAAA;AAAA,UAEzBg1F,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAI4C,EAAE5C,EAAE,UAAU,WAAW6C,EAAE7C,EAAE,UAAU,SAAS8C,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAG,GAAGE,IAAId,GAAGW,GAAG,KAAK,MAAM;AAAA,cAClFtC;AAAA;AAAA;AAAA,gCAGkB,MAAM,MAAMC;AAAA;AAAA,0BAElBwC,QAAQD;AAAA,+BACH5C;AAAA;AAAA,MAEzB,GAAG6C,IAAIxC,GAAGqC,GAAG,KAAK,MAAM;AAAA,cAChBtC;AAAA;AAAA;AAAA,iBAGG,EAAE,GAAG,EAAE,GAAG,EAAE;AAAA,iBACZ,EAAE,GAAG,EAAE,OAAO,EAAE;AAAA;AAAA;AAAA,yBAGRyC,QAAQD;AAAA,+BACF5C;AAAA;AAAA,MAEzB,IAAI8C,EAAEsyF,GAAGp1F,CAAC,EAAE,MAAM;AAAA,YACZI;AAAA;AAAA,0BAEc2B,aAAa,eAAe;AAAA,qBACjC1B,gBAAgByC;AAAA,6BACRF,MAAMC;AAAA,6BACN7C;AAAA;AAAA,GAE1B,CAAC,SAAS6zF,GAAI/zF,EAAE,CAAC,IAAI,EAAEA,EAAE,UAAU,aAAaE,EAAEF,EAAE,KAAKM,EAAE,MAAMJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAE,CAAC,SAASK,EAAE,SAAS,CAAC,EAAEumB,EAAE,aAAa,CAAC,EAAE,GAAGvmB,EAAE,OAAO,EAAE,OAAO,CAAC,IAAI,EAAEi1F,GAAGx1F,EAAEO,CAAC,EAAEgH,EAAE,CAAC,MAAM,MAAM,QAAQ,SAAS,SAAS,QAAQ,EAAE,MAAM;AAAA,QACxOisF,GAAG,CAAC;AAAA,cACElzF;AAAA;AAAA,iBAEGA,KAAKm1F,GAAGluF,EAAE,CAAC;AAAA;AAAA,KAEvB,CAAC,IAAI,EAAE,EAAE,GAAGtF,EAAE,EAAE,GAAG,EAAEC,EAAE,EAAE,GAAGD,EAAEE,EAAE,EAAE,GAAGD,EAAEU,EAAE,EAAE,GAAGT,EAAE,GAAGnC,EAAE,UAAU,UAAU,MAAM;AAAA,cACtEM;AAAA;AAAA;AAAA;AAAA,iBAIGsC,MAAMT,MAAMD,MAAMD;AAAA;AAAA;AAAA,mBAGhB;AAAA,UACTszF,GAAGv1F,CAAC;AAAA;AAAA,MAER,IAAI6C,EAAE7C,EAAE,UAAU,WAAW8C,EAAE9C,EAAE,UAAU,SAAS+C,EAAED,EAAE,GAAGE,EAAEF,EAAE,GAAG,GAAGE,IAAIJ,GAAGC,GAAG,KAAK,MAAM;AAAA,cAClFvC;AAAA;AAAA;AAAA;AAAA,iBAIG6B,MAAMD,MAAMD,MAAM;AAAA;AAAA;AAAA,0BAGTe,QAAQD;AAAA,+BACH7C;AAAA;AAAA,MAEzB,GAAG8C,IAAI,GAAGH,GAAG,KAAK,MAAM;AAAA,cAChBvC;AAAA;AAAA;AAAA,iBAGG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE;AAAA,iBACjB,EAAE,GAAG,EAAE,GAAG,EAAE;AAAA,iBACZ,EAAE,GAAG,EAAE;AAAA,iBACP,EAAE;AAAA;AAAA;AAAA,yBAGM0C,QAAQD;AAAA,+BACF7C;AAAA;AAAA,MAEzB,IAAIgE,EAAEoxF,GAAGp1F,CAAC,EAAE,MAAM;AAAA,YACZI;AAAA;AAAA;AAAA,0BAGcsC,aAAaT,eAAeD;AAAA,qBACjCD,gBAAgB,gBAAgBiC;AAAA,6BACxBnB,MAAMC;AAAA,6BACN9C;AAAA;AAAA,GAE1B,CAAC,SAASq1F,GAAGv1F,EAAE,CAAC,IAAI,EAAEA,EAAE,KAAKE,EAAE4mB,EAAE,cAAc9mB,EAAE,UAAU,YAAY,EAAE,OAAOE,EAAE,EAAE,UAAU,KAAK;AAAA,0BAC5EA;AAAA;AAAA,iBAET;AAAA;AAAA;AAAA,GAGd,CAAC,SAASo0F,GAAIt0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAEJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEK,EAAE,MAAMD,EAAE,cAAc,EAAEN,EAAE,UAAU,aAAa,OAAO,EAAE,EAAE,aAAa,OAAOiC,EAAE4wF,GAAG7yF,EAAE,UAAU,aAAa,EAAE,YAAY,EAAEkC,EAAEwzF,GAAG,CAAC,EAAEvzF,EAAE,EAAE,EAAES,EAAEC,EAAE,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAE,IAAI,EAAED,EAAE,GAAG,EAAE,GAAGX,EAAE,QAAQ,EAAEW,EAAE,cAAcA,EAAEX,EAAE,IAAI8C,GAAG,UAAUlC,EAAEkC,EAAE5C,SAAS,EAAE,KAAK;AAAA,CAC1U,EAAE,IAAIW,EAAE,GAAG,EAAE,GAAG,EAAE,EAAEA,EAAE,SAASA,EAAE9C,EAAE,UAAU,aAAa,IAAI,CAAC+E,EAAEJ,IAAI,UAAU9B,EAAE8B,EAAExC,IAAI,EAAE,KAAK,IAAI,EAAE,IAAIY,EAAE,sBAAsBmB,EAAE4iB,EAAE,cAAc9mB,EAAE,UAAU,YAAY,IAAI,EAAEuH,EAAEuf,EAAE,cAAc,EAAE,YAAY,IAAI,EAAE,GAAG,IAAI,GAAG,CAAC5iB,GAAG,CAACqD,EAAExE,EAAE;AAAA;AAAA,cAE5NmB,GAAG,CAACqD,EAAE,IAAI,EAAExE,EAAE;AAAA;AAAA,QAEpBA,EAAE;AAAA;AAAA,gBAEMd,EAAE,OAAO,CAAC,IAAI8C,EAAE,EAAE,EAAEJ,EAAE,EAAE,EAAE1C,EAAE,QAAQ8C,CAAC,EAAE,IAAI9C,EAAE,QAAQ0C,CAAC,EAAE,GAAG5B,EAAE,8BAA8Bd,EAAE,QAAQ8C,CAAC,EAAE,GAAGhC,EAAE,2EAA2Ed,EAAE,QAAQ0C,CAAC,EAAE,KAAK5B,EAAE,+CAA+C,CAAC,MAAM;AAAA,WACvQxC;AAAA,QACH2B;AAAA,QACAU;AAAA,8BACsBtC,KAAKwC;AAAA,QAC3BC;AAAA;AAAA,GAEL,CAAC,SAASwxF,GAAIv0F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,KAAKM,EAAEJ,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEK,EAAE,MAAMD,EAAE,cAAc,EAAE,EAAE,SAAS,EAAEN,EAAE,UAAU,SAASiC,EAAEjC,EAAE,UAAU,aAAa,OAAOkC,EAAE,EAAE,aAAa,OAAO,GAAG,CAAClC,EAAE,UAAU,WAAWiC,IAAIC,GAAGlC,EAAE,UAAU,YAAY,MAAM8mB,EAAE,YAAY,EAAE,CAAC,EAAE,MAAM;AAAA,cAC3QvmB;AAAA,+BACiBL;AAAA;AAAA,MAEzB,IAAIiC,EAAEuzF,GAAGxzF,CAAC,EAAEU,EAAEiwF,GAAG7yF,EAAE,UAAU,aAAa,EAAE,YAAY,EAAE6C,EAAEX,EAAED,EAAEa,EAAEC,EAAE,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAEd,IAAI,EAAEa,EAAE,GAAGZ,EAAE,GAAGU,EAAE,QAAQ,EAAEE,EAAE,cAAcA,EAAEF,EAAE,IAAIsB,GAAG,UAAUnB,EAAEmB,EAAErB,SAAS,EAAE,KAAK;AAAA,CACxL,EAAE,IAAIG,EAAE,GAAG,OAAOd,EAAE,GAAGD,EAAE,EAAEe,EAAE,SAASA,EAAEhD,EAAE,UAAU,aAAa,IAAI,CAACkE,EAAE,IAAI,UAAUnB,EAAE,EAAEF,IAAI,EAAE,KAAK,IAAI,EAAE;AAAA,YAChGtC;AAAA,QACJ4B;AAAA,QACAW;AAAA,kBACUxC,KAAK0C;AAAA;AAAA,GAEpB,CAAC,SAAS0yF,GAAG11F,EAAE,CAAC,GAAGA,GAAG,EAAE,MAAM,MAAM,GAAGA,IAAI,EAAE,MAAM,QAAQ,GAAGA,IAAI,EAAE,MAAM,QAAQ,GAAGA,IAAI,EAAE,MAAM,QAAQ,GAAGA,IAAI,EAAE,MAAM,QAAQ,GAAGA,IAAI,EAAE,MAAM,QAAQ,MAAM,MAAM,gBAAgBA,wBAAwB,CAAC,CAAC,SAAS+yF,GAAG/yF,EAAE,EAAEE,EAAE,CAAC,GAAG,CAAC,SAASI,EAAE,SAASC,CAAC,EAAEumB,EAAE,aAAa,CAAC,EAAE,EAAE,EAAE,OAAO,EAAE9mB,GAAG,IAAI,GAAG,EAAE,KAAK,EAAEiC,EAAE,EAAE,EAAE,MAAM,CAAC,EAAE3B,EAAE4B,EAAE,CAAClC,GAAG,EAAE,GAAG,CAAC8mB,EAAE,YAAY,EAAE5mB,CAAC,GAAGI,EAAE,OAAO,GAAG,EAAE,MAAM,CAAC,gBAAgB4B,EAAE,aAAaA,EAAED,EAAE,EAAE,SAAS1B,CAAC,CAAC,CAAC,SAASi1F,GAAGx1F,EAAE,EAAE,CAAC,IAAIE,EAAE,KAAK,MAAM,KAAK,UAAUF,CAAC,CAAC,EAAE,OAAOE,EAAE,UAAU,aAAa,EAAEA,CAAC,CAAC,SAASu1F,GAAGz1F,EAAE,EAAE,CAAC,OAAO,EAAE,IAAIE,GAAGF,EAAEE,EAAE,EAAE,KAAK,IAAI,CAAC,CAAC,SAASy1F,GAAG31F,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEL,EAAE,IAAI,CAAC0C,EAAEC,IAAI,CAAC,IAAIC,EAAE,CAAC,aAAaF,EAAE,MAAM,SAASA,EAAE,UAAU,KAAKA,EAAE,QAAQ,SAAS,UAAUA,EAAE,UAAU,SAASA,EAAE,UAAU,GAAGA,EAAE,QAAQ,SAAS,WAAW,IAAI,EAAE,OAAOA,EAAE,SAAS,MAAMA,EAAE,QAAQ,OAAO,MAAMA,EAAE,QAAQ,MAAM,WAAW,IAAIE,EAAE,WAAWF,EAAE,QAAQ,MAAM,YAAY,CAAC,KAAK,EAAE,cAAcC,GAAG,UAAUC,CAAC,CAAC,CAAC,EAAE,EAAEvC,EAAE,IAAIqC,GAAGA,EAAE,SAAS,EAAE,EAAE,CAAC,aAAatC,EAAE,MAAM,SAASA,EAAE,QAAQ,SAAS,UAAU,GAAG,SAASA,EAAE,QAAQ,SAAS,WAAW,IAAI,EAAE2B,EAAE6wF,GAAGvyF,EAAE,EAAE,CAAC,EAAE2B,EAAEqsF,GAAGvuF,EAAE,GAAGiC,CAAC,EAAEE,EAAEnC,EAAE,cAAckC,CAAC,EAAE,OAAOuY,EAAE,EAAE,IAAI,qBAAqB,EAAE,CAAC,QAAQ,EAAE,eAAevY,EAAE,OAAOD,EAAE,aAAaE,EAAE,aAAa,EAAE,aAAa,EAAE,iBAAiB,KAAK,uBAAuB,KAAK,OAAO,KAAK,OAAO,KAAK,kBAAkB,KAAK,qBAAqB,KAAK,iBAAiB,KAAK,wBAAwB,KAAK,oBAAoB,IAAI,EAAE,OAAO,OAAO,CAAC,QAAQ,EAAE,eAAeD,EAAE,OAAOD,EAAE,aAAaE,EAAE,aAAa,EAAE,aAAa,CAAC,EAAEyzF,GAAG51F,EAAE,EAAEmC,CAAC,CAAC,CAAC,CAAC,SAASyzF,GAAG51F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAAC,EAAEC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE0B,EAAEC,EAAEC,EAAES,EAAE,KAAKC,EAAE,KAAKA,EAAE7C,EAAE,mBAAmBE,EAAE,MAAM,EAAE,EAAEua,EAAE,EAAE,UAAU,eAAe,IAAI,IAAI7X,EAAE5C,EAAE,mBAAmBE,EAAE,WAAW,EAAE,GAAG,IAAI4C,EAAE,GAAG,QAAQC,EAAE,EAAEA,EAAE,EAAE,cAAc,OAAOA,IAAI,CAAC,IAAIC,EAAE,EAAE,cAAcD,GAAGzC,EAAE0C,GAAGhD,EAAE,mBAAmBE,EAAE8C,EAAEF,CAAC,EAAExC,EAAE,SAAS0C,KAAKhD,EAAE,mBAAmBE,EAAE,SAAS8C,IAAIF,CAAC,EAAE,EAAE,sBAAsBvC,EAAE,GAAGyC,UAAUhD,EAAE,mBAAmBE,EAAE,GAAG8C,SAASF,CAAC,EAAE,EAAE,GAAGE,aAAahD,EAAE,mBAAmBE,EAAE,GAAG8C,YAAYF,CAAC,EAAE,CAAC,OAAO,EAAE,sBAAsBb,EAAEjC,EAAE,mBAAmBE,EAAE,WAAW4C,CAAC,EAAEX,EAAEnC,EAAE,mBAAmBE,EAAE,kBAAkB4C,CAAC,EAAEZ,EAAElC,EAAE,mBAAmBE,EAAE,cAAc4C,CAAC,GAAG,EAAE,gBAAgB,EAAE,eAAe,QAAQ,CAACC,EAAEC,IAAI,CAAC,EAAEA,GAAGhD,EAAE,mBAAmBE,EAAE6C,EAAE,KAAKD,CAAC,CAAC,CAAC,EAAE,CAAC,iBAAiBxC,EAAE,uBAAuB,EAAE,OAAOsC,EAAE,OAAOC,EAAE,kBAAkBtC,EAAE,qBAAqB,EAAE,iBAAiB0B,EAAE,wBAAwBE,EAAE,oBAAoBD,CAAC,CAAC,CAAC,SAAS2zF,GAAG71F,EAAE,EAAE,CAAC,GAAGA,EAAE,SAAS,EAAE,OAAO,MAAM,MAAM,4BAA4BA,EAAE,wCAAwC,EAAE,eAAe,EAAEA,EAAE,QAAQ,CAACE,EAAEI,IAAI,CAAC,IAAIC,EAAEL,EAAE,aAAa,EAAE,EAAEI,GAAG,EAAE,EAAE,MAAM,GAAG,CAACwmB,EAAE,YAAYvmB,EAAE,CAAC,EAAE,MAAM,MAAM,2EAA2EA,SAAS,cAAc,EAAE,GAAGL,EAAE,WAAW,EAAE,UAAU,OAAO,IAAI+B,EAAE/B,EAAE,SAASgC,EAAE,EAAE,UAAU,KAAK,EAAE,QAAQ,SAAS,GAAG,CAAC4kB,EAAE,YAAY7kB,EAAEC,CAAC,EAAE,MAAM,MAAM,kFAAkFD,SAASC,cAAc,CAAC,CAAC,CAAC,CAAC,SAAS4zF,GAAG91F,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,EAAE,QAAQ,sBAAsBs1F,GAAG,EAAE,aAAa31F,CAAC,EAAE21F,GAAG,CAAC,EAAE,YAAY,EAAE,CAACv1F,CAAC,CAAC,GAAG,IAAI,EAAEA,EAAE,QAAQ,QAAQ,EAAEA,EAAE,QAAQ,SAASA,EAAE,QAAQ,SAASN,EAAE,6BAA6B,EAAE,QAAQ,EAAE,GAAG,EAAE,EAAE,EAAEA,EAAE,uBAAuB,EAAE,QAAQ,EAAE,GAAG,EAAE,EAAE,EAAEA,EAAE,WAAW,EAAE,YAAY,EAAEya,EAAE,EAAE,UAAU,eAAe,IAAI,GAAG,EAAE,SAAS,MAAMza,EAAE,GAAG,UAAU,EAAE,OAAO,EAAE,CAAC,EAAE,EAAE,SAAS,MAAMA,EAAE,GAAG,UAAU,EAAE,OAAO,GAAG,EAAEE,EAAE,QAAQ,CAACgC,EAAEC,IAAI,CAAC,IAAIS,EAAE,EAAE,QAAQ,cAAcT,GAAGU,EAAE,EAAE,iBAAiBD,GAAGE,EAAE,EAAE,iBAAiB,SAASF,KAAKG,EAAE,EAAE,kBAAkB,GAAGH,UAAUI,EAAE,EAAE,qBAAqB,GAAGJ,aAAa,GAAGG,EAAE,CAAC,GAAG,CAAC,aAAamB,CAAC,EAAE6uF,GAAG,EAAE,QAAQ,aAAa7wF,EAAE,MAAMA,EAAE,QAAQ,QAAQ,EAAE,OAAOgC,EAAE,OAAO,CAAC,IAAK,GAAElE,EAAE,GAAG,WAAW+C,EAAE,IAAI,WAAWmB,CAAC,CAAC,EAAE,MAAM,IAAK,GAAElE,EAAE,GAAG,WAAW+C,EAAE,IAAI,WAAWmB,CAAC,CAAC,EAAE,MAAM,IAAK,GAAElE,EAAE,GAAG,WAAW+C,EAAE,IAAI,WAAWmB,CAAC,CAAC,EAAE,MAAM,IAAK,GAAElE,EAAE,GAAG,WAAW+C,EAAE,IAAI,WAAWmB,CAAC,CAAC,EAAE,MAAM,QAAQ,KAAK,CAAC,CAAC,GAAGlB,GAAGhD,EAAE,GAAG,UAAUgD,EAAEd,EAAE,QAAQ,SAAS,GAAGA,EAAE,QAAQ,SAAS,EAAE,EAAEW,GAAG,KAAK,CAAC,GAAGX,EAAE,UAAU,CAAC,GAAG4kB,EAAE,cAAc5kB,EAAE,KAAK,EAAE,EAAElC,EAAE,GAAG,UAAU6C,EAAEX,EAAE,cAAc,EAAE,MAAM,CAAC,IAAIgC,EAAEhC,EAAE,cAAcgC,aAAa,eAAeA,EAAE,IAAI,aAAaA,CAAC,GAAGlE,EAAE,GAAG,WAAW6C,EAAEqB,CAAC,CAAC,CAAC,MAAM,CAAChC,EAAE,QAAQ,OAAO,MAAMY,GAAG,MAAM9C,EAAE,GAAG,UAAU8C,EAAEZ,EAAE,QAAQ,MAAM,UAAU,EAAElC,EAAE,sBAAsBkC,EAAE,QAAQ,QAAQ,QAAQW,EAAEV,CAAC,CAAC,CAAC,CAAC,EAAE,IAAIF,EAAE,EAAE,iBAAiB,GAAGA,EAAE,OAAO3B,EAAE,MAAM,OAAO,CAAC,IAAK,GAAEN,EAAE,GAAG,WAAWiC,EAAE,IAAI,WAAW3B,EAAE,KAAK,CAAC,EAAE,MAAM,IAAK,GAAEN,EAAE,GAAG,WAAWiC,EAAE,IAAI,WAAW3B,EAAE,KAAK,CAAC,EAAE,MAAM,IAAK,GAAEN,EAAE,GAAG,WAAWiC,EAAE,IAAI,WAAW3B,EAAE,KAAK,CAAC,EAAE,MAAM,IAAK,GAAEN,EAAE,GAAG,WAAWiC,EAAE,IAAI,WAAW3B,EAAE,KAAK,CAAC,EAAE,MAAM,QAAQ,KAAK,CAAC,GAAG,EAAE,wBAAwB,CAAC,IAAI4B,EAAE4kB,EAAE,eAAexmB,EAAE,KAAK,EAAE,OAAOA,EAAE,MAAM,OAAO,CAAC,IAAK,GAAEN,EAAE,GAAG,WAAW,EAAE,wBAAwB,IAAI,WAAWkC,CAAC,CAAC,EAAE,MAAM,IAAK,GAAElC,EAAE,GAAG,WAAW,EAAE,wBAAwB,IAAI,WAAWkC,CAAC,CAAC,EAAE,MAAM,IAAK,GAAElC,EAAE,GAAG,WAAW,EAAE,wBAAwB,IAAI,WAAWkC,CAAC,CAAC,EAAE,MAAM,QAAQ,KAAK,CAAC,CAAC,EAAE,qBAAqBlC,EAAE,GAAG,UAAU,EAAE,oBAAoBM,EAAE,QAAQ,SAAS,GAAGA,EAAE,QAAQ,SAAS,EAAE,EAAE,EAAE,QAAQ,gBAAgBC,GAAG,EAAE,QAAQ,eAAe,QAAQ,CAAC2B,EAAEC,IAAI,CAAC,IAAIS,EAAE,EAAE,uBAAuBT,GAAGU,EAAEtC,EAAE4B,GAAG,GAAGD,EAAE,OAAO,QAAQlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,OAAOlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,OAAOlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,OAAOlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,MAAMlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,QAAQlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,QAAQlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,UAAUX,EAAE,OAAO,QAAQlC,EAAE,GAAG,WAAW4C,EAAEC,CAAC,MAAO,OAAM,MAAM,gBAAgBX,EAAE,4BAA4B,CAAC,CAAC,EAAElC,EAAE,eAAe,CAAC,CAAC,SAAS+1F,GAAG/1F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,GAAG,EAAE,OAAOJ,CAAC,EAAE,QAAQ,GAAG,CAAC,IAAI+B,EAAE,EAAE,SAAS,MAAM,EAAE,QAAQ,OAAO,MAAM,EAAE,QAAQ,MAAM,WAAW,EAAE,GAAGjC,EAAE,qBAAqB,CAAC,EAAE,UAAU,CAAC,IAAIkC,EAAE,EAAE,QAAQ,SAAS,CAAC,gBAAgBC,EAAE,aAAaS,EAAE,SAASC,CAAC,EAAEkwF,GAAG/yF,EAAE,aAAa,EAAE,MAAMkC,CAAC,EAAEY,EAAE,GAAGC,EAAE,GAAGC,EAAE,GAAG,GAAGJ,EAAE,SAAS,GAAG5C,EAAE,aAAa,CAAC,IAAI6E,EAAE,CAAC,KAAK,KAAK3C,EAAE,GAAG,CAAC,EAAE,KAAK,KAAKA,EAAE,GAAG,CAAC,CAAC,EAAEY,EAAE,GAAG+B,EAAE,GAAG,KAAKA,EAAE,GAAG,GAAG,SAASjC,EAAE,SAAS,GAAG,CAAC5C,EAAE,aAAa+C,EAAE,GAAGH,EAAE,GAAG,KAAKA,EAAE,GAAG,YAAYA,EAAE,OAAO,GAAG,CAAC5C,EAAE,aAAa,CAAC,IAAI6E,EAAEiiB,EAAE,eAAelkB,CAAC,EAAEI,EAAE,GAAG6B,EAAE,KAAK3C,EAAE,MAAM2C,EAAEA,EAAE,OAAO,KAAK3C,EAAE,IAAI,CAAC,IAAIgC,EAAE,EAAE,MAAM,OAAO,EAAEtB,EAAE,SAAS,GAAGkkB,EAAE,YAAY,EAAE,MAAM5kB,CAAC,EAAEqF,EAAEuf,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,EAAE0zB,EAAE,iBAAiB,EAAE,MAAMt6C,EAAE,KAAK,EAAEyE,EAAE,CAAC3E,EAAE,cAAckE,IAAIhE,EAAE,MAAM,QAAQ4mB,EAAE,YAAY5kB,EAAEhC,EAAE,QAAQ,QAAQ,EAAE0E,EAAE5E,EAAE,cAAc4C,EAAE,OAAO,EAAE,GAAG,GAAGV,EAAE,GAAG,KAAKA,EAAE,GAAG,IAAI5B,GAAG,GAAG4D,KAAKS,KAAKxC,EAAEU,EAAE,MAAMD,EAAE,UAAU2E,KAAK,KAAK,KAAKzE,KAAKC,KAAKC,KAAK4B,KAAK3C,GAAG,KAAK,CAAC,IAAIC,EAAE,EAAE,UAAU,UAAU,EAAE,QAAQ,SAAS5B,GAAG,GAAG,EAAE,SAAS4B,KAAKD,GAAG,CAAC,CAAC,EAAE,IAAI1B,EAAEP,EAAE,SAAS,EAAEA,EAAE,YAAY,KAAK,OAAO,GAAG,IAAIM,EAAE,IAAIC,EAAE,GAAGka,EAAE,EAAE,UAAU,eAAe,IAAI,CAAC,CAAC,SAASu7E,GAAGh2F,EAAE,CAAC,OAAOya,EAAE,EAAE,QAAQ,2BAA2B,GAAGza,GAAG,CAAC,CAAC,IAAIi2F,GAAG,KAAK,CAAC,YAAYn1F,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,iBAAiBkwF,GAAG,MAAM,KAAK,eAAe,CAAC,CAAC,KAAK,WAAW,KAAK,OAAO,CAAC,EAAE,IAAI9wF,EAAEmyF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA,UAE30N,KAAK,oBAAoBzD,GAAG,CAAC,IAAI,IAAI,GAAG,EAAEzxF,CAAC,EAAEwxF,GAAG,CAAC,IAAI,IAAI,GAAG,EAAExxF,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgB/DZ,EAAE;AAAA;AAAA,KAEP,CAAC,EAAMg2F,GAAG,KAAK,CAAC,YAAYp1F,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,iBAAiBkwF,GAAG,MAAM,KAAK,eAAe,CAAC,CAAC,KAAK,WAAW,KAAK,OAAO,CAAC,EAAE,IAAI9wF,EAAEmyF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA,UAElR,KAAK,oBAAoBzD,GAAG,CAAC,IAAI,IAAI,GAAG,EAAEzxF,CAAC,EAAEwxF,GAAG,CAAC,IAAI,IAAI,GAAG,EAAExxF,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgB/DZ,EAAE;AAAA;AAAA,KAEP,CAAC,EAAMi2F,GAAG,KAAK,CAAC,YAAYr1F,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAYmwF,GAAG,SAAS,IAAI/wF,EAAEmyF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,SAAS;AAAA,QAC9H8xF;AAAA;AAAA;AAAA;AAAA,UAIE1yF,EAAE;AAAA;AAAA,KAEP,CAAC,EAAMk2F,GAAG,KAAK,CAAC,YAAYt1F,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYmwF,GAAG,SAAS,IAAI/wF,EAAEmyF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,SAAS;AAAA,QACxK8xF;AAAA;AAAA;AAAA;AAAA;AAAA,UAKE1yF,EAAE;AAAA;AAAA,KAEP,CAAC,EAAMm2F,GAAI,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEC,GAAG,KAAK,CAAC,YAAYx1F,EAAEZ,EAAE,GAAG,EAAE,OAAO,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,WAAW,KAAK,OAAO,CAAC,EAAE,IAAIK,EAAE8xF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIj0F,EAAE,SAAS7B,IAAI6B,EAAE,8BAA8B,IAAIC,EAAE,GAAG,QAAQC,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,CAAC,IAAIC,EAAE,EAAED,GAAGD,GAAG;AAAA,yBACtTC;AAAA,8BACKo0F,GAAIn0F;AAAA,YACtB,CAAC,KAAK,SAAS;AAAA,QACnB,KAAK,oBAAoBywF,GAAG,EAAED,GAAG5xF,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uCAMH,EAAE;AAAA;AAAA,sCAEH,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAMdP,EAAE;AAAA,YAChByB;AAAA;AAAA,UAEFzB,EAAE,iBAAiBwB;AAAA;AAAA,KAExB,CAAC,EAAMw0F,GAAG,KAAK,CAAC,YAAYz1F,EAAEZ,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,WAAW,KAAK,OAAO,CAAC,EAAE,IAAI,EAAEmyF,GAAG,EAAE,KAAK,YAAYvxF,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIz1F,EAAE,GAAGwB,EAAE,SAAS7B,IAAI6B,EAAE,8BAA8B,QAAQC,EAAE,EAAEA,GAAG,EAAEA,IAAI,QAAQC,EAAE,EAAEA,GAAG,EAAEA,IAAI,CAAC,IAAIC,EAAEF,EAAE,EAAEC,EAAE1B,GAAG;AAAA;AAAA,gCAEtU0B,OAAO,KAAK,oBAAoB,cAAc,GAAGnB,EAAE;AAAA,8BACrDmB;AAAA,iCACGD,OAAO,KAAK,oBAAoB,cAAc,GAAGlB,EAAE;AAAA,gCACpDkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uBAUT,EAAE;AAAA;AAAA;AAAA,uBAGFE;AAAA;AAAA,uBAEAA;AAAA;AAAA,uBAEAA;AAAA;AAAA,uBAEAA;AAAA;AAAA;AAAA;AAAA,SAId,CAAC,KAAK,SAAS;AAAA,UACd,KAAK,oBAAoBywF,GAAG,EAAED,GAAG5xF,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAWhCP;AAAA;AAAA,YAEA,EAAE,YAAYwB;AAAA;AAAA,KAErB,CAAC,EAAMy0F,GAAG,CAAC,EAAEp2F,GAAGo2F,GAAG,CAAC,kCAAkC,IAAIC,GAAG,8BAA8B,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,iCAAiC,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,kBAAkB,IAAIC,GAAG,0BAA0B,IAAIC,GAAG,iCAAiC,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,mBAAmB,IAAIC,GAAG,gDAAgD,IAAIC,GAAG,gCAAgC,IAAIC,GAAG,sCAAsC,IAAIC,GAAG,+BAA+B,IAAIC,GAAG,yCAAyC,IAAIC,GAAG,+CAA+C,IAAIC,GAAG,yCAAyC,IAAIC,GAAG,wCAAwC,IAAIC,GAAG,+CAA+C,IAAIC,GAAG,2BAA2B,IAAIC,GAAG,yBAAyB,IAAIC,EAAE,CAAC,EAAE,SAASX,GAAGl3F,EAAE,CAAC,IAAI,EAAEqyF,GAAG,EAAEnyF,EAAE,GAAG,EAAE;AAAA;AAAA,MAEp3B,EAAE;AAAA,MACF,EAAE;AAAA,MACF,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,OAKD,OAAO2uF,GAAG7uF,EAAEE,CAAC,CAAC,CAAC,SAAS+2F,GAAGj3F,EAAE,CAAC,IAAI,EAAE,IAAI,aAAa,CAAC,GAAG,EAAE,EAAE,EAAE,EAAE,GAAG,GAAG,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO2uF,GAAG3uF,EAAE,CAAC,CAAC,CAAC,SAAS82F,GAAG92F,EAAE,CAAC,IAAI,EAAE,IAAI,YAAY,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO0uF,GAAG1uF,EAAE,CAAC,CAAC,CAAC,SAAS83F,GAAG93F,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAACkwF,GAAG,EAAEvwF,CAAC,EAAE,IAAI,EAAE0uF,GAAG5uF,CAAC,EAAEiC,EAAEjC,EAAE,WAAW,OAAOquF,GAAGruF,EAAE,IAAIA,EAAE,YAAYiC,EAAE,CAAC,CAAC,EAAEosF,GAAGruF,EAAE,IAAIA,EAAE,cAAciC,EAAEjC,EAAE,eAAeA,EAAE,aAAa,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,cAAciC,EAAEjC,EAAE,eAAeA,EAAE,aAAa,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,cAAciC,EAAEjC,EAAE,mBAAmBA,EAAE,OAAO,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,cAAciC,EAAEjC,EAAE,mBAAmBA,EAAE,OAAO,CAAC,EAAEya,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE4zE,GAAGruF,EAAE,IAAIA,EAAE,WAAWiC,EAAE,EAAE3B,EAAE,EAAEJ,EAAE,EAAEK,EAAE,EAAE,IAAI,CAAC,EAAE8tF,GAAGruF,EAAE,IAAIA,EAAE,aAAaiC,EAAE,EAAE3B,EAAE,EAAEJ,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,IAAI,CAAC,EAAE,CAAC,QAAQ,EAAE,SAAS,CAACE,EAAE,CAAC,CAAC,CAAC,CAAC,SAASu3F,GAAGz3F,EAAE,CAAC,OAAOA,EAAE,mBAAmB,CAAC,SAAS62F,GAAG72F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE4wF,GAAG,EAAEjxF,CAAC,EAAE,OAAO43F,GAAG93F,EAAEO,EAAE,EAAEk3F,GAAGn3F,CAAC,EAAEA,EAAE,mBAAmBN,EAAE,KAAK,CAAC,CAAC,SAASu3F,GAAGv3F,EAAE,CAAC,OAAOA,EAAE,uBAAuB,CAAC,SAAS22F,GAAG32F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE4wF,GAAG,EAAEjxF,CAAC,EAAE,OAAO43F,GAAG93F,EAAEO,EAAE,EAAEg3F,GAAGj3F,CAAC,EAAEA,EAAE,mBAAmBA,EAAE,oBAAoB,CAAC,CAAC,SAASq3F,GAAG33F,EAAE,CAAC,OAAOA,EAAE,qBAAqB,CAAC,SAASg3F,GAAGh3F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE4wF,GAAG,EAAEjxF,CAAC,EAAE,OAAO43F,GAAG93F,EAAEO,EAAE,EAAEo3F,GAAGr3F,CAAC,EAAEN,EAAE,KAAKA,EAAE,aAAa,CAAC,CAAC,SAAS03F,GAAG13F,EAAE,CAAC,OAAOA,EAAE,yBAAyB,CAAC,SAAS+2F,GAAG/2F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE+wF,GAAG,EAAEpxF,CAAC,EAAE,OAAO43F,GAAG93F,EAAEO,EAAE,EAAEm3F,GAAGp3F,CAAC,EAAEN,EAAE,KAAKA,EAAE,KAAK,CAAC,CAAC,SAASw3F,GAAGx3F,EAAE,CAAC,OAAOA,EAAE,6BAA6B,CAAC,SAAS42F,GAAG52F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE+wF,GAAG,EAAEpxF,CAAC,EAAE,OAAO43F,GAAG93F,EAAEO,EAAE,EAAEi3F,GAAGl3F,CAAC,EAAEN,EAAE,KAAKM,EAAE,oBAAoB,CAAC,CAAC,SAASm2F,GAAGz2F,EAAE,EAAEE,EAAE,CAAC,OAAOmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,aAAaE,CAAC,CAAC,EAAEkuF,GAAGpuF,EAAE,EAAE,eAAeE,EAAE,EAAE,GAAG,CAAC,GAAGkuF,GAAGpuF,EAAE,EAAE,KAAKE,EAAE,EAAE,GAAG,EAAE,CAAC,CAAC,SAAS03F,GAAG53F,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC8tF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,CAAC,CAAC,EAAE,IAAI,EAAEiC,EAAEC,EAAE3B,aAAa,YAAY,EAAE,IAAI,WAAWL,EAAEI,EAAE,CAAC,EAAE2B,EAAEjC,EAAE,cAAckC,EAAElC,EAAE,OAAO,EAAE,IAAI,aAAaE,EAAEI,EAAE,CAAC,EAAE2B,EAAEjC,EAAE,MAAMkC,EAAE,EAAE,2BAA2B,EAAE,IAAI3B,CAAC,EAAEka,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE4zE,GAAGruF,EAAE,IAAIA,EAAE,cAAcA,EAAE,WAAW,EAAE,EAAE,EAAEE,EAAEI,EAAEN,EAAE,KAAKiC,EAAE,CAAC,CAAC,EAAEosF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,WAAW,EAAEkC,EAAEhC,EAAEI,EAAE,EAAEN,EAAE,KAAKiC,EAAE,CAAC,CAAC,EAAEosF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,IAAI,CAAC,CAAC,CAAC,SAAS63F,GAAG73F,EAAE,EAAEE,EAAE,CAACmuF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,CAAC,CAAC,EAAEE,EAAE,gBAAgB,WAAWua,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE4zE,GAAGruF,EAAE,IAAIA,EAAE,cAAcA,EAAE,WAAW,EAAE,EAAE,EAAEE,EAAE,MAAMA,EAAE,OAAOF,EAAE,KAAKA,EAAE,cAAcE,EAAE,IAAI,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,WAAW,EAAEA,EAAE,KAAKE,EAAE,MAAMA,EAAE,OAAO,EAAEF,EAAE,KAAKA,EAAE,cAAcE,EAAE,IAAI,CAAC,EAAEua,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE4zE,GAAGruF,EAAE,IAAIA,EAAE,cAAcA,EAAE,WAAW,EAAE,EAAE,EAAEA,EAAE,KAAKA,EAAE,cAAcE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,WAAW,EAAEA,EAAE,KAAKA,EAAE,KAAKA,EAAE,cAAcE,CAAC,CAAC,EAAEmuF,GAAGruF,EAAE,IAAIA,EAAE,YAAYA,EAAE,WAAW,IAAI,CAAC,CAAC,CAAC,SAAS02F,GAAG12F,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEP,EAAE,aAAa,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,kBAAkBO,CAAC,CAAC,EAAE,IAAI0B,EAAE,EAAE,EAAE,EAAE/B,EAAE,OAAOmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,kBAAkBiC,EAAEjC,EAAE,WAAW,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,WAAW,EAAE,EAAEE,EAAE,EAAEF,EAAE,KAAKA,EAAE,MAAM,CAAC,CAAC,EAAEquF,GAAGruF,EAAE,IAAIA,EAAE,WAAWA,EAAE,kBAAkB,IAAI,CAAC,EAAEO,CAAC,CAAC,SAAS62F,GAAGp3F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAEO,EAAE,IAAI,aAAaL,CAAC,EAAE,OAAOI,EAAE,WAAWA,EAAE,kBAAkB,CAAC,EAAEA,EAAE,iBAAiBA,EAAE,kBAAkB,EAAEC,CAAC,EAAED,EAAE,WAAWA,EAAE,kBAAkB,IAAI,EAAEC,CAAC,CAAC,SAAS42F,GAAGn3F,EAAE,EAAEE,EAAEI,EAAE,CAAC,GAAG,CAACC,EAAE,CAAC,EAAE4wF,GAAG,EAAEjxF,CAAC,EAAE,EAAE,EAAE+B,EAAE,IAAI,WAAWmvF,GAAG,EAAElxF,EAAE,CAAC,CAAC,EAAE,OAAOmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAW,EAAE,EAAEO,EAAE,EAAED,EAAE,sBAAsBN,EAAE,cAAciC,CAAC,CAAC,EAAE,IAAI,aAAaA,EAAE,MAAM,CAAC,CAAC,SAASq1F,GAAGt3F,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE0B,EAAE,CAAC,IAAIC,EAAElC,EAAEmC,EAAE,IAAI,aAAaovF,GAAG,EAAE,CAAC,CAAC,EAAE,OAAOrvF,EAAE,WAAWA,EAAE,kBAAkB,CAAC,EAAEA,EAAE,iBAAiBA,EAAE,kBAAkB,EAAEC,CAAC,EAAED,EAAE,WAAWA,EAAE,kBAAkB,IAAI,EAAEC,CAAC,CAAC,SAASk1F,GAAGr3F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,IAAI,aAAa,EAAEJ,EAAE,CAAC,EAAE,OAAOmuF,GAAGruF,EAAE,IAAIA,EAAE,WAAW,EAAE,EAAEE,EAAE,EAAEF,EAAE,KAAKA,EAAE,MAAMM,CAAC,CAAC,EAAEA,CAAC,CAAC,IAAIy3F,GAAG,KAAK,CAAC,YAAYj3F,EAAE,CAAC,KAAK,cAAc,KAAK,KAAK,QAAQ,KAAK,KAAK,SAAS,GAAG,KAAK,oBAAoB,GAAG,KAAK,YAAY,CAAC,EAAE,IAAIZ,EAAEua,EAAE,EAAE,UAAU,eAAe,EAAE3Z,GAAG,MAAM,KAAK,GAAGA,EAAE8vF,GAAG1wF,EAAEY,CAAC,GAAG,KAAK,GAAG+vF,GAAG3wF,CAAC,EAAE,IAAI,EAAE,2BAA2BK,EAAE,8BAA8B,GAAG,KAAK,6BAA6B,KAAK,GAAG,aAAa,6BAA6B,EAAEka,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE,CAAC,IAAI1Y,EAAE,oBAAoBC,EAAE,yBAAyB,GAAG,KAAK,sBAAsB+sF,GAAG,KAAK,GAAGhtF,CAAC,EAAE4tF,GAAG,KAAK,GAAG3tF,CAAC,EAAE,KAAK,0BAA0B+sF,GAAG,KAAK,GAAG/sF,CAAC,UAAUyY,EAAE,EAAE,IAAI,0BAA0B,EAAE,MAAM,IAAI,MAAM,oHAAoH,EAAE,GAAG,KAAK,0BAA0B,KAAK,GAAG,aAAa,CAAC,EAAEk1E,GAAG,KAAK,GAAGpvF,CAAC,EAAE,KAAK,8BAA8BwuF,GAAG,KAAK,GAAGxuF,CAAC,UAAUka,EAAE,EAAE,IAAI,0BAA0B,EAAE,MAAM,IAAI,MAAM,6HAA6H,CAAC,SAAS,EAAE,yBAAyBk1E,GAAG,KAAK,GAAG,CAAC,EAAE,KAAK,0BAA0B,KAAK,GAAG,aAAa,CAAC,UAAUA,GAAG,KAAK,GAAGpvF,CAAC,EAAE,KAAK,8BAA8B,KAAK,GAAG,aAAaA,CAAC,MAAO,OAAM,IAAI,MAAM,qDAAqD,EAAE,KAAK,aAAa02F,GAAG,KAAK,EAAE,EAAE,KAAK,YAAYH,GAAG,KAAK,EAAE,EAAE,KAAK,YAAYtI,GAAG,KAAK,EAAE,EAAE,KAAK,cAAcgD,GAAG,KAAK,GAAG,KAAK,yBAAyB,CAAC,CAAC,IAAI,OAAO,CAAC,OAAO/2E,EAAE,EAAE,QAAQ,OAAO,CAAC,CAAC,SAAS,CAAC,GAAG,KAAK,SAAS,OAAO,KAAK,SAAS,MAAM,QAAQ,KAAK,sKAAsK,EAAE,KAAK,eAAe,MAAM,QAAQ,KAAK,oMAAoM,EAAE,IAAI3Z,EAAE,KAAK,GAAGutF,GAAGvtF,EAAE,IAAIA,EAAE,OAAO,CAAC,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,gBAAgBA,EAAE,YAAY,IAAI,CAAC,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,kBAAkB,KAAK,WAAW,CAAC,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,WAAWA,EAAE,aAAa,IAAI,CAAC,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,WAAWA,EAAE,qBAAqB,IAAI,CAAC,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,aAAa,KAAK,WAAW,CAAC,EAAE,KAAK,SAAS,EAAE,CAAC,2BAA2BA,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE22F,GAAG,KAAK,GAAG/1F,EAAEZ,EAAE,KAAK,aAAa,CAAC,CAAC,2BAA2BY,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEy2F,GAAG,KAAK,GAAG71F,EAAEZ,EAAE,KAAK,aAAa,CAAC,CAAC,iCAAiCY,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE82F,GAAG,KAAK,GAAGl2F,EAAEZ,EAAE,KAAK,aAAa,CAAC,CAAC,yBAAyBY,EAAEZ,EAAE,CAAC,KAAK,gBAAgB,EAAE23F,GAAG,KAAK,GAAG/2F,EAAEZ,CAAC,CAAC,CAAC,2BAA2BY,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,gBAAgB,EAAEq3F,GAAG,KAAK,GAAG92F,EAAEZ,EAAE,EAAEK,EAAE,KAAK,aAAa,CAAC,CAAC,iCAAiCO,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE02F,GAAG,KAAK,GAAG91F,EAAEZ,EAAE,KAAK,aAAa,CAAC,CAAC,0BAA0BY,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE62F,GAAG,KAAK,GAAGj2F,EAAEZ,EAAE,KAAK,aAAa,CAAC,CAAC,oBAAoBY,EAAE,CAAC,KAAK,gBAAgB,EAAE,KAAK,gBAAgBA,IAAIuvF,GAAG,KAAK,GAAG,KAAK,WAAW,EAAE,KAAK,cAAc,MAAMhC,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,cAAcvtF,CAAC,CAAC,CAAC,CAAC,gDAAgDA,EAAEZ,EAAE,EAAE,CAAC,OAAO,KAAK,qBAAqBY,EAAE,IAAIq2F,GAAG,KAAK,GAAGj3F,EAAE,EAAE,KAAK,aAAa,CAAC,CAAC,CAAC,+BAA+BY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,OAAOs1F,GAAG,KAAK,GAAGx2F,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,KAAK,aAAa,CAAC,CAAC,gCAAgClB,EAAEZ,EAAE,CAAC,OAAOk3F,GAAG,KAAK,GAAGt2F,EAAEZ,CAAC,CAAC,CAAC,wBAAwBY,EAAEZ,EAAE,EAAE,CAAC,KAAK,yBAAyBY,CAAC,EAAE,IAAIP,EAAEm2F,GAAG,KAAK,GAAGx2F,EAAE,EAAE,KAAK,aAAa,EAAE,OAAO,KAAK,2BAA2B,EAAEK,CAAC,CAAC,uBAAuB,CAAC,IAAIO,EAAE,KAAK,YAAY,KAAK,EAAE,EAAE,OAAO,KAAK,UAAUA,CAAC,CAAC,CAAC,YAAYA,EAAE,CAAC,IAAIZ,EAAE,EAAE,GAAGua,EAAE,EAAE,QAAQ,yBAAyB,EAAE,CAAC,IAAIla,EAAEO,EAAEiB,EAAExB,EAAE,UAAUA,EAAE,2BAA2B,CAAC,EAAEO,EAAE,MAAM,EAAE,EAAE,IAAI,CAAC,IAAIkB,EAAEzB,EAAE,eAAewB,EAAE,EAAE,CAAC,EAAE,OAAOC,IAAIzB,EAAE,kBAAkByB,IAAIzB,EAAE,mBAAmB,EAAEL,EAAE6B,CAAC,MAAM0Y,EAAE,EAAE,UAAU,8CAA8C,EAAE,GAAGva,EAAE,KAAK,WAAW,EAAE,KAAK,SAAS,EAAE,EAAE,IAAI,KAAK,iBAAiBA,EAAEua,EAAE,EAAE,UAAU,8CAA8C,CAAC,GAAG,EAAE,IAAI,GAAG,MAAM,CAAC,MAAMva,EAAE,cAAc,CAAC,CAAC,CAAC,gCAAgCY,EAAEZ,EAAE,EAAE,CAAC,OAAO,KAAK,qBAAqBY,EAAE,IAAIu2F,GAAG,KAAK,GAAGn3F,EAAE,CAAC,CAAC,CAAC,CAAC,cAAcY,EAAE,CAAC,KAAK,gBAAgB,EAAE,IAAIZ,EAAE,KAAK,GAAG,KAAK,cAAc,OAAO,KAAK,aAAag3F,GAAGh3F,CAAC,GAAG,IAAI,EAAEuuF,GAAGvuF,CAAC,EAAE,OAAOmuF,GAAGnuF,EAAE,IAAIA,EAAE,aAAa,EAAE,KAAK,YAAY,CAAC,EAAEmuF,GAAGnuF,EAAE,IAAIA,EAAE,aAAa,EAAEY,CAAC,CAAC,EAAEmvF,GAAG/vF,EAAE,CAAC,EAAE,KAAK,OAAOswF,GAAGtwF,EAAE,CAAC,EAAE,KAAK,sBAAsB,KAAK,WAAW,CAAC,EAAE,KAAK,oBAAoBu2F,GAAGv2F,EAAE,KAAK,QAAQ,KAAK,YAAY,GAAG,CAAC,CAAC,cAAcY,EAAE,CAAC,KAAK,gBAAgB,EAAEA,IAAI,KAAK,UAAU,KAAK,QAAQ,MAAMA,GAAG,MAAMutF,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,cAAcvtF,CAAC,CAAC,CAAC,CAAC,WAAWA,EAAE,CAAC,KAAK,gBAAgB,EAAE,KAAK,QAAQA,EAAE,KAAK,SAAS,MAAM,KAAK,OAAO0vF,GAAG,KAAK,GAAG,KAAK,OAAO,EAAEnC,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,WAAWvtF,CAAC,CAAC,CAAC,CAAC,mBAAmBA,EAAEZ,EAAE,EAAE,GAAG,CAAC,OAAO,KAAK,gBAAgB,EAAE,EAAEkvF,GAAG,KAAK,GAAGtuF,EAAEZ,CAAC,EAAEivF,GAAG,KAAK,GAAGruF,EAAEZ,CAAC,CAAC,CAAC,qBAAqBY,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAEmuF,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,kBAAkBvtF,EAAEZ,CAAC,CAAC,CAAC,CAAC,0BAA0BY,EAAEZ,EAAE,CAAC,OAAO,KAAK,gBAAgB,EAAE,KAAK,GAAG,mBAAmBY,EAAEZ,CAAC,CAAC,CAAC,sBAAsBY,EAAEZ,EAAE,EAAE,CAAC,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAEguF,GAAG,KAAK,GAAGptF,EAAEZ,EAAE,CAAC,CAAC,CAAC,uBAAuBY,EAAEZ,EAAE,EAAE,CAAC,KAAK,6BAA6BY,EAAE,EAAEZ,CAAC,CAAC,CAAC,6BAA6BY,EAAEZ,EAAE,EAAE,CAAC,KAAK,gBAAgB,EAAE,GAAG,CAACK,EAAEwB,CAAC,EAAEuvF,GAAGpxF,EAAE,CAAC,EAAE,KAAK,6BAA6BY,EAAEP,EAAEwB,CAAC,CAAC,CAAC,2BAA2BjB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,iCAAiC,EAAEO,EAAEP,EAAEL,CAAC,CAAC,CAAC,iCAAiCY,EAAEZ,EAAE,EAAEK,EAAE,CAAC,MAAM,IAAI,MAAM,mDAAmD,CAAC,CAAC,eAAe,CAAC,KAAK,SAAS,MAAMiwF,GAAG,KAAK,GAAG,KAAK,OAAO,EAAED,GAAG,KAAK,EAAE,CAAC,CAAC,gBAAgB,CAAC,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,IAAIzvF,EAAE,KAAK,GAAG,KAAK,OAAO,KAAK,cAAc,EAAEutF,GAAGvtF,EAAE,IAAIA,EAAE,aAAaA,EAAE,UAAU,EAAEA,EAAE,eAAe,CAAC,CAAC,CAAC,CAAC,gCAAgC,CAAC,KAAK,gBAAgB,EAAEutF,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,OAAO,CAAC,CAAC,CAAC,wBAAwB,CAAC,OAAO,KAAK,6BAA6B,OAAO,KAAK,4BAA4BU,GAAG,KAAK,GAAGt0E,EAAE,EAAE,UAAU,8CAA8C,IAAI,EAAE,kCAAkC,0BAA0B,GAAG,KAAK,2BAA2B,CAAC,8BAA8B,CAAC,OAAO,KAAK,uBAAuB,CAAC,CAAC,8BAA8B,CAAC,OAAO,KAAK,uBAAuB,CAAC,CAAC,YAAY,CAAC,GAAGA,EAAE,EAAE,UAAU,8CAA8C,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,GAAGla,EAAE,KAAK,6BAA6B,EAAEwB,EAAE,EAAE,YAAY,EAAE,OAAO,EAAE,WAAWxB,EAAE,iBAAiBwB,CAAC,EAAEA,CAAC,CAAC,IAAIjB,EAAE,KAAK,6BAA6B,EAAEZ,EAAEY,EAAE,eAAe,EAAE,OAAOA,EAAE,cAAcA,EAAE,iBAAiBZ,CAAC,EAAEA,CAAC,CAAC,UAAU,CAAC,GAAGua,EAAE,EAAE,UAAU,8CAA8C,IAAI,EAAE,CAAC,IAAIva,EAAE,KAAK,GAAG,EAAE,KAAK,6BAA6B,EAAEA,EAAE,SAAS,EAAE,gBAAgB,EAAE,MAAM,CAAC,IAAIY,EAAE,KAAK,6BAA6B,EAAEA,EAAE,YAAYA,EAAE,gBAAgB,CAAC,CAAC,MAAM,uBAAuBA,EAAE,CAAC,OAAO,MAAMgmB,EAAE,YAAY,IAAI,KAAK,UAAU,KAAK,iBAAiBhmB,EAAE2Z,EAAE,EAAE,UAAU,8CAA8C,CAAC,CAAC,EAAE,KAAK,aAAa3Z,EAAE2Z,EAAE,EAAE,UAAU,8CAA8C,CAAC,CAAC,CAAC,aAAa3Z,EAAEZ,EAAE,CAAC,GAAGA,IAAI,EAAE,OAAO,KAAK,GAAGA,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,GAAG,OAAO,EAAE,kBAAkBY,EAAE,EAAE,YAAY,EAAE,GAAG,KAAK,CAAC,IAAI,EAAE,KAAK,6BAA6B,EAAE,OAAO,EAAE,kBAAkBA,EAAE,EAAE,gBAAgB,EAAE,GAAG,CAAC,CAAC,iBAAiBA,EAAEZ,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,GAAG,GAAGA,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,GAAGK,EAAE,KAAK,6BAA6B,EAAEwB,EAAE,EAAE,kBAAkBjB,EAAE,EAAE,sBAAsB,EAAE,OAAO,KAAK,UAAU,OAAO,KAAK,SAAS,KAAK,GAAG,aAAaP,EAAE,gBAAgB,GAAGwB,GAAG,CAAC,KAAK,QAAQ,KAAK,CAAC,IAAI,EAAE,KAAK,6BAA6B,EAAExB,EAAE,EAAE,kBAAkBO,EAAE,EAAE,0BAA0B,EAAE,OAAO,KAAK,UAAU,OAAO,KAAK,SAAS,KAAK,GAAG,aAAa,EAAE,gBAAgB,GAAGP,GAAG,CAAC,KAAK,QAAQ,CAAC,CAAC,UAAUO,EAAE,CAAC,OAAO,IAAI,QAAQZ,GAAG,CAAC,KAAK,cAAc,IAAIY,EAAE,cAAc,EAAE,IAAIZ,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,CAAC,IAAIY,EAAEk3F,GAAI,KAAK,YAAY,IAAI93F,GAAGA,EAAE,QAAQ,CAAC,EAAE,QAAQA,EAAE,EAAEA,GAAGY,EAAE,EAAEZ,EAAE,CAAC,GAAG,CAAC,UAAU,CAAC,EAAE,KAAK,YAAYA,GAAG,EAAE,CAAC,CAAC,KAAK,YAAY,KAAK,YAAY,MAAMY,EAAE,CAAC,CAAC,CAAC,cAAcA,EAAEZ,EAAE,CAAC,GAAG,KAAK,YAAY,KAAK,CAAC,SAASY,EAAE,UAAUZ,CAAC,CAAC,EAAE,KAAK,YAAY,OAAO,EAAE,OAAO,IAAI,EAAE,qBAAqBua,EAAE,EAAE,WAAW,EAAEA,EAAE,EAAE,SAAS,iBAAiB,KAAKA,EAAE,EAAE,QAAQ,GAAGqM,EAAE,YAAY,KAAK,KAAK,UAAU,EAAE,KAAK,YAAY,SAAS,GAAG,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,yBAAyBhmB,EAAE,CAAC,KAAK,gBAAgB,EAAEmtF,GAAG,KAAK,GAAGntF,EAAE,KAAK,WAAW,EAAE,KAAK,OAAOyvF,GAAG,KAAK,EAAE,CAAC,CAAC,4BAA4B,CAAC,KAAK,eAAe,MAAMtC,GAAG,KAAK,GAAG,KAAK,cAAc,KAAK,WAAW,EAAE,KAAK,OAAOsC,GAAG,KAAK,EAAE,GAAGF,GAAG,KAAK,GAAG,KAAK,WAAW,CAAC,CAAC,qBAAqBvvF,EAAEZ,EAAE,CAAC,KAAK,yBAAyBY,CAAC,EAAE,IAAI,EAAEZ,EAAE,EAAE,OAAO,KAAK,2BAA2B,EAAE,CAAC,CAAC,6BAA6BY,EAAEZ,EAAE,EAAE,CAAC,KAAK,gBAAgB,EAAE,IAAIK,EAAE,KAAK,GAAG0tF,GAAG1tF,EAAEO,EAAE,KAAK,WAAW,EAAE,KAAK,OAAOyvF,GAAGhwF,CAAC,EAAE,KAAK,cAAcO,EAAEutF,GAAG9tF,EAAE,IAAIA,EAAE,SAAS,EAAE,EAAEL,EAAE,CAAC,CAAC,EAAEmuF,GAAG9tF,EAAE,IAAIA,EAAE,QAAQ,EAAE,EAAEL,EAAE,CAAC,CAAC,CAAC,CAAC,iCAAiCY,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,gBAAgB,EAAE8tF,GAAG,KAAK,GAAG,IAAI,KAAK,GAAG,QAAQvtF,EAAEZ,EAAE,EAAEK,CAAC,CAAC,CAAC,CAAC,iBAAiB,CAAC,GAAG,KAAK,SAAS,MAAM,IAAI,MAAM,yCAAyC,CAAC,CAAC,kBAAkB,CAAC,GAAG,KAAK,SAAS,KAAK,MAAM,IAAI,MAAM,kCAAkC,CAAC,CAAC,EAAE,SAASy3F,GAAIh4F,EAAE,CAAC,IAAI,EAAE,EAAE,KAAK,EAAEA,EAAE,QAAQA,EAAE,GAAG,EAAE,EAAE,EAAE,CAAC,OAAO,EAAE,CAAC,CAAC,GAAG,CAAC,QAAQi4F,GAAG,aAAaC,GAAG,mBAAmBC,GAAG,SAASC,GAAG,SAASC,GAAG,WAAWC,GAAG,UAAUC,GAAG,QAAQC,GAAG,UAAUC,GAAG,UAAUC,GAAG,aAAaC,GAAG,aAAaC,GAAG,YAAYC,GAAG,iBAAiBC,GAAG,SAASC,GAAG,cAAcC,GAAG,aAAaC,GAAG,QAAQC,GAAG,QAAQC,GAAG,YAAYC,GAAG,YAAYC,GAAG,aAAaC,GAAG,QAAQC,GAAG,aAAaC,GAAG,SAASC,GAAG,iBAAiBC,GAAG,gBAAgBC,GAAG,yBAAyBC,GAAG,UAAUC,GAAG,UAAUC,GAAG,YAAYC,GAAG,YAAYC,GAAG,cAAcC,GAAG,UAAUC,GAAG,wBAAwBC,GAAG,kBAAkBC,GAAG,2BAA2BC,GAAG,SAASC,GAAG,iBAAiBC,GAAG,iBAAiBC,GAAG,gBAAgBC,GAAG,2BAA2BC,GAAG,QAAQC,GAAG,SAASC,GAAG,SAASC,GAAG,cAAcC,GAAG,WAAWC,EAAE,EAAEnpB,GAAG,SAASopB,GAAGh7F,EAAE,EAAE,CAAC,MAAM,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAE,MAAM,EAAE,CAAC,EAAE,IAAIE,GAAG,GAAGF,KAAKE,GAAG,CAAC,CAAC,SAAS+6F,GAAGj7F,EAAE,EAAE,CAAC,OAAO,IAAI,EAAE,CAACA,CAAC,EAAEg7F,GAAGh7F,EAAE,CAAC,CAAC,CAAC,SAASk7F,GAAGl7F,EAAE,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,KAAK,IAAIE,EAAE,GAAG,QAAQI,EAAE,EAAEA,EAAEN,EAAEM,IAAIJ,GAAG,EAAEI,GAAGA,EAAEN,EAAE,IAAIE,GAAG,KAAK,OAAOA,CAAC,CAAC,IAAIi7F,GAAG,KAAK,CAAC,YAAYr6F,EAAE,CAAC,GAAG,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYA,EAAE,KAAK,KAAKA,EAAE,OAAO,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,OAAO,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA,YAI//a,CAAC,IAAI91F,EAAE+6F,GAAG,KAAK,KAAK,IAAI,EAAE,EAAEvF,GAAG,KAAK,IAAI,EAAEn1F,EAAE,KAAK,wBAAwBL,CAAC,EAAE6B,EAAE,KAAK,SAAS7B,CAAC,EAAE8B,EAAE,KAAK,UAAU9B,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,YAEjI;AAAA;AAAA,eAEGK;AAAA;AAAA;AAAA,cAGDwB;AAAA;AAAA,6BAEeC;AAAA;AAAA;AAAA,OAGtB,CAAC,CAAC,mBAAmBlB,EAAE,CAAC,IAAIZ,EAAE,CAAC,EAAE,QAAQ,EAAE,EAAE,GAAG,EAAE,IAAI,QAAQK,EAAE,EAAEA,GAAG,EAAEA,IAAI,CAAC,IAAIwB,EAAE,GAAG,IAAI,EAAE,IAAI,UAAUxB,IAAI,EAAE,IAAI,QAAQ,QAAQyB,EAAE,EAAEA,EAAE,KAAK,KAAKA,IAAID,EAAE,GAAGjB,EAAEA,EAAE,OAAO,EAAEkB,MAAMD,EAAE7B,EAAE,KAAK6B,CAAC,CAAC,CAAC,OAAO7B,CAAC,CAAC,wBAAwBY,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAM,QAAQ,KAAK,oBAAoB,WAAW,KAAK,YAAY,KAAK,IAAIZ,EAAE,GAAG,QAAQ,EAAE,KAAK,KAAK,EAAE,EAAE,KAAK,KAAK,IAAIA,GAAG,GAAGY,EAAE,SAAS,KAAK,oBAAoB,YAAY,KAAK,KAAK,YAAY,KAAK,EAAE,KAAK,KAAK,IAAIZ,GAAG,MAAM,OAAOA,CAAC,CAAC,SAASY,EAAE,CAAC,GAAG,KAAK,OAAO,EAAE,MAAM,GAAG,IAAIZ,EAAEY,EAAE,MAAM,EAAE,EAAE,EAAE,KAAK,oBAAoB,YAAY,KAAK,YAAY,KAAK,YAAY,KAAK,KAAK,GAAGP,EAAE,KAAK,oBAAoB,YAAY,KAAK,YAAY,KAAK,YAAY,KAAK,KAAK,GAAG,MAAM;AAAA,gBAC3rBL,EAAE;AAAA,gBACFA,EAAE;AAAA;AAAA;AAAA;AAAA,4BAIU;AAAA,4BACAK;AAAA,KACvB,CAAC,UAAUO,EAAE,CAAC,IAAIZ,EAAE,KAAK,mBAAmBY,CAAC,EAAE,OAAO,KAAK,OAAO,EAAE,wBAAwB,KAAK,oBAAoB,WAAW,KAAK,YAAY,gCAAgC,QAAQZ,EAAE;AAAA,gCAChKA,EAAE;AAAA,gCACFA,EAAE;AAAA,yCACOA,EAAE,KAAK,CAAC,EAAMk7F,GAAG,KAAK,CAAC,YAAYt6F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,aAAa,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYY,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAI,EAAE,GAAG,QAAQz1F,EAAE,EAAEA,EAAE,EAAEA,IAAI,CAAC,IAAIwB,EAAE,eAAexB,EAAE,IAAI,IAAIwB,GAAG,kBAAkBxB,EAAE,IAAIwB,GAAG,kBAAkB,GAAG;AAAA,UACzXA;AAAA,UACAxB,EAAE,EAAE,0CAA0C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mBAMrCA;AAAA;AAAA,UAETA,EAAE,EAAE,IAAI;AAAA,OACX,CAAC,KAAK,SAAS;AAAA,QACd86F,GAAIn7F,EAAE,KAAK,mBAAmB;AAAA,QAC9B,KAAK,oBAAoByyF,GAAG,EAAED,GAAG5xF,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,qBAQrB,KAAK,oBAAoB,cAAcA,EAAE;AAAA,qBACzC,KAAK,oBAAoB,cAAcA,EAAE;AAAA;AAAA,UAEpD;AAAA;AAAA;AAAA;AAAA,KAIL,CAAC,EAAE,SAASu6F,GAAIr7F,EAAE,EAAE,CAAC,MAAM;AAAA;AAAA,QAExB,EAAEyyF,GAAG,CAAC,IAAI,IAAI,GAAG,EAAE,YAAY,EAAEH,GAAG,CAAC,IAAI,IAAI,GAAG,EAAEtyF,CAAC;AAAA;AAAA;AAAA,GAGxD,CAAC,IAAIs7F,GAAG,KAAK,CAAC,YAAYx6F,EAAE,CAAC,KAAK,MAAMA,EAAE,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,EAAE,KAAK,mBAAmB,EAAE,KAAK,cAAc,EAAE,KAAK,aAAa,CAAC,EAAE,KAAK,WAAW,GAAG,KAAK,aAAa,CAAC,CAAC,CAAC,eAAeA,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAEg7F,GAAGr7F,EAAE,CAAC,EAAE6B,EAAEy5F,GAAG16F,EAAEP,EAAE,CAAC,EAAEwB,KAAK,KAAK,eAAe,KAAK,aAAaA,GAAG,CAAC,GAAGA,KAAK,KAAK,eAAe,KAAK,aAAaA,GAAG,CAAC,GAAG,IAAIC,EAAEy5F,GAAG36F,EAAEP,EAAE,KAAK,MAAM,GAAG,KAAK,MAAM,cAAc,CAAC,EAAE,GAAG,KAAK,aAAawB,GAAG,OAAO,EAAE,CAAC,KAAK,kBAAkB,KAAK,kBAAkB,KAAK,eAAeC,EAAE,KAAK,IAAI,EAAE,IAAIE,EAAE,KAAK,aAAaH,GAAG,MAAM,EAAE,OAAO,KAAK,aAAaA,GAAG,KAAKG,CAAC,EAAEA,CAAC,CAAC,IAAID,EAAE,OAAO1B,IAAI2wF,GAAG,mBAAmBjvF,EAAE,KAAK,MAAM,0BAA0BnB,EAAE,GAAGA,EAAE,EAAE,EAAEP,IAAI2wF,GAAG,mBAAmBjvF,EAAE,KAAK,MAAM,iCAAiCnB,EAAE,GAAGA,EAAE,EAAE,EAAEP,IAAI2wF,GAAG,iBAAiBjvF,EAAE,KAAK,MAAM,2BAA2BnB,EAAE,GAAGA,EAAE,EAAE,EAAEP,IAAI2wF,GAAG,iBAAiBjvF,EAAE,KAAK,MAAM,2BAA2BnB,EAAE,GAAGA,EAAE,EAAE,EAAEP,IAAI2wF,GAAG,2BAA2BjvF,EAAE,KAAK,MAAM,iCAAiCnB,EAAE,GAAGA,EAAE,EAAE,GAAG,KAAK,aAAaiB,GAAG,KAAKE,CAAC,EAAE,KAAK,kBAAkB,KAAK,oBAAoBD,EAAE,KAAK,IAAI,EAAEC,CAAC,CAAC,eAAenB,EAAEZ,EAAE,EAAEK,EAAE,CAAC,GAAG,KAAK,cAAc,KAAK,OAAO,IAAIwB,EAAEw5F,GAAG,EAAEh7F,CAAC,EAAEyB,EAAEw5F,GAAGt7F,EAAE6B,EAAExB,CAAC,EAAEyB,KAAK,KAAK,eAAe,KAAK,aAAaA,GAAG,CAAC,GAAG,IAAIC,EAAEw5F,GAAGv7F,EAAE6B,EAAE,KAAK,MAAM,GAAG,KAAK,MAAM,cAAcxB,CAAC,EAAE2B,EAAEuY,EAAE,EAAE,IAAI,gCAAgC,EAAEvY,IAAI,IAAI,KAAK,mBAAmBA,GAAG,KAAK,MAAM,oBAAoBpB,EAAE,OAAO,EAAE,KAAK,oBAAoBmB,IAAI,KAAK,aAAaD,GAAG,KAAKlB,CAAC,EAAE,KAAK,kBAAkB,KAAK,eAAemB,GAAG,KAAK,kBAAkB,IAAI,EAAE,KAAK,aAAaD,GAAGY,EAAE,EAAE,QAAQ9B,CAAC,EAAE,GAAG8B,EAAE,EAAE,MAAM,IAAI,MAAM,0EAA0E,EAAE,EAAE,OAAOA,EAAE,CAAC,EAAE,KAAK,IAAI,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,WAAW,OAAO,IAAI9B,EAAE,KAAK,gBAAgB,KAAK,gBAAgB,QAAQ,IAAI,YAAY,GAAG,KAAK,qBAAqB,KAAK,kBAAkB,IAAIA,IAAI,EAAE,IAAIZ,EAAE,KAAK,cAAc,KAAK,mBAAmB,QAAQ,IAAI,oBAAoB,KAAK,oBAAoB,EAAE,QAAQ,IAAI,iBAAiB,KAAK,kBAAkB,KAAK,MAAM,IAAIA,CAAC,KAAK,CAAC,CAAC,IAAI,mBAAmB,CAAC,OAAO,KAAK,kBAAkB,CAAC,IAAI,cAAc,CAAC,OAAO,KAAK,aAAa,CAAC,oBAAoB,CAAC,OAAO,KAAK,eAAe,CAAC,oBAAoB,CAAC,OAAO,KAAK,eAAe,CAAC,SAAS,CAAC,GAAG,KAAK,cAAc,KAAK,CAAC,QAAQY,KAAK,KAAK,aAAa,KAAK,aAAaA,GAAG,QAAQZ,GAAG,CAAC,KAAK,MAAM,oBAAoBA,EAAE,OAAO,CAAC,CAAC,EAAE,QAAQY,KAAK,KAAK,aAAa,KAAK,aAAaA,GAAG,QAAQZ,GAAG,CAAC,KAAK,MAAM,oBAAoBA,EAAE,OAAO,CAAC,CAAC,EAAE,KAAK,aAAa,KAAK,KAAK,aAAa,KAAK,KAAK,gBAAgB,EAAE,KAAK,gBAAgB,EAAE,KAAK,mBAAmB,EAAE,KAAK,cAAc,CAAC,CAAC,CAAC,EAAE,SAASw7F,GAAI17F,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,GAAG,IAAIE,EAAE,KAAK,MAAO,GAAE,GAAG,IAAIA,EAAE,KAAK,MAAO,GAA6B,GAAxB,IAAIA,EAAE,SAAqB,IAAIF,EAAE,KAAK,MAAO,IAAG,GAAG,IAAIE,EAAE,QAAQ,MAAO,GAAE,GAAG,IAAIA,EAAE,MAAM,MAAO,GAAE,MAAM,IAAI,MAAM,2BAA2B,GAAG,CAAC,CAAC,SAASu7F,GAAGz7F,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAEo7F,GAAI,EAAEr7F,CAAC,EAAE,EAAE,GAAGC,EAAE,CAAC,GAAG,CAAC2B,EAAEC,CAAC,EAAEmvF,GAAGtxF,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAEkC,EAAEC,CAAC,KAAK,CAAC,GAAG,CAACD,EAAEC,CAAC,EAAEgvF,GAAGnxF,EAAE,GAAGA,EAAE,EAAE,EAAE,EAAEkC,EAAEC,CAAC,CAAC,IAAIF,EAAEy5F,GAAIx7F,EAAE,CAAC,EAAE,OAAO,EAAE+B,CAAC,CAAC,SAAS05F,GAAI37F,EAAE,EAAE,CAAC,OAAOA,EAAE,CAAC,KAAKkxF,GAAG,mBAAmB,OAAOwG,GAAG,CAAC,EAAE,KAAKxG,GAAG,mBAAmB,OAAOsG,GAAG,CAAC,EAAE,KAAKtG,GAAG,iBAAiB,OAAOuG,GAAG,CAAC,EAAE,KAAKvG,GAAG,iBAAiB,OAAOqG,GAAG,CAAC,EAAE,KAAKrG,GAAG,yBAAyB,OAAOyG,GAAG,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,iCAAiC33F,GAAG,CAAC,CAAC,CAAC,SAAS47F,GAAI57F,EAAE,CAAC,OAAOya,EAAE,EAAE,QAAQ,8BAA8B,EAAEza,EAAEkxF,GAAG,mBAAmBA,GAAG,iBAAiBlxF,EAAEkxF,GAAG,mBAAmBA,GAAG,gBAAgB,CAAC,SAASqK,GAAGv7F,EAAE,EAAE,CAAC,GAAGA,IAAIixF,GAAG,OAAO,OAAOC,GAAG,mBAAmB,GAAGlxF,IAAIixF,GAAG,QAAQjxF,GAAG,KAAK,OAAO47F,GAAI,CAAC,EAAE,GAAG57F,IAAIixF,GAAG,UAAUjxF,IAAIixF,GAAG,OAAO,OAAOC,GAAG,yBAAyB,MAAM,IAAI,MAAM,gCAAgClxF,GAAG,CAAC,CAAC,SAASw7F,GAAGx7F,EAAE,EAAEE,EAAE,CAAC,MAAM,GAAGF,EAAE,MAAMA,EAAE,MAAM,KAAKE,GAAG,CAAC,IAAI27F,GAAG,KAAK,CAAC,YAAY/6F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAYY,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA,UAE9wH91F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KASL,CAAC,EAAE47F,GAAG,0BAA0BC,GAAG,YAAYC,GAAG,iBAAqBC,GAAG,0CAA0CC,GAAGJ,GAAG;AAAA;AAAA,EAE7HK,GAAGL,GAAG;AAAA;AAAA,EAENM,GAAG,YAAYC,GAAG,sCAA0CC,GAAG,YAAYC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS9EC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUHC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUHC,GAAG,sCAAsCC,GAAG,KAAK,CAAC,YAAY77F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYY,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA,UAEpN91F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KASL,CAAC,EAAM08F,GAAG,KAAK,CAAC,YAAY97F,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYA,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAI91F,EAAEY,EAAE,OAAO,EAAEm6F,GAAG,KAAK/6F,CAAC,EAAEK,EAAEm1F,GAAGx1F,CAAC,EAAE6B,EAAEm5F,GAAGh7F,EAAE,CAAC,EAAE8B,EAAE,EAAE,MAAM,EAAE,EAAEC,EAAE/B,GAAG,EAAE,KAAK,QAAQ8B,EAAE,KAAK,GAAG,KAAK,KAAK,SAAS;AAAA;AAAA,UAElRzB;AAAA,kCACwBwB;AAAA;AAAA,4CAEUE;AAAA;AAAA,KAEvC,CAAC,EAAM46F,GAAIx+C,GAAG,UAAUy+C,GAAI,KAAKC,GAAI,KAAKC,GAAG,CAAC,EAAE,SAASC,GAAIj9F,EAAE,CAAC,OAAOA,KAAKg9F,KAAKA,GAAGh9F,GAAG,CAAC,GAAGg9F,GAAGh9F,EAAE,CAAC,IAAIk9F,GAAIziF,EAAE,EAAE,UAAU,4BAA4B,EAAE0iF,GAAI,IAAI,SAASC,IAAK,CAAC,OAAO3iF,EAAE,EAAE,OAAO,QAAQ,KAAK,KAAKA,EAAE,EAAE,OAAO,OAAO,OAAOA,EAAE,EAAE,OAAO,OAAO,MAAM,OAAO,iBAAiB0iF,GAAI,KAAK,IAAI,CAAC,IAAIE,GAAG,cAAcjmF,EAAE,CAAC,YAAYtW,EAAE,CAAC,GAAG,MAAM,EAAE,KAAK,YAAY,IAAI,QAAQ,KAAK,gBAAgB,IAAI,QAAQ,KAAK,aAAa,IAAI,QAAQ,KAAK,cAAc,EAAE,KAAK,aAAa,EAAE,KAAK,eAAe,EAAE,KAAK,gBAAgB,EAAE,KAAK,kBAAkB,GAAG,KAAK,eAAe,EAAE,KAAK,SAAS,GAAG,CAAC2Z,EAAE,EAAE,QAAQ,WAAW,EAAE,MAAM,IAAI,MAAM,uCAAuC,EAAE,IAAIva,EAAE,GAAGY,GAAG,KAAK,CAAC,GAAGA,aAAai3F,GAAG73F,EAAEY,MAAM,CAAC,IAAI,EAAE+vF,GAAGp2E,EAAE,EAAE,UAAU,eAAe,EAAE3Z,CAAC,EAAEZ,EAAE,IAAI63F,GAAG,CAAC,CAAC,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,oBAAoB,EAAE,KAAK,CAAC,IAAI,EAAElH,GAAGp2E,EAAE,EAAE,UAAU,eAAe,CAAC,EAAEva,EAAE,IAAI63F,GAAG,CAAC,EAAE,KAAK,YAAYkF,GAAIxiF,EAAE,EAAE,UAAU,eAAe,CAAC,EAAE,KAAK,oBAAoB,EAAE,CAAC,KAAK,MAAMva,EAAE,KAAK,OAAO,KAAK,MAAM,GAAG,OAAO,KAAK,eAAe,IAAIo7F,GAAG,KAAK,KAAK,EAAE,KAAK,mBAAmB8B,GAAI,EAAE,KAAK,QAAQ,IAAIjmF,GAAG,KAAKmc,GAAG,CAAC,CAAC,CAAC,YAAY,CAAC,OAAO+pE,GAAG,YAAY,CAAC,YAAY,CAAC,OAAO,KAAK,QAAQ,WAAW,EAAE,KAAK,cAAc,CAAC,aAAav8F,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,IAAIC,EAAE,KAAK,eAAe/B,EAAE,CAAC,EAAEgC,EAAE,KAAK,QAAQ,IAAID,EAAE,MAAM,EAAEC,EAAE,SAAS,GAAGA,EAAE,QAAQ,CAAC,QAAQpB,EAAE,SAAS,CAACP,EAAEwB,CAAC,CAAC,EAAEG,EAAE,SAAS,CAAC3B,EAAEwB,CAAC,EAAE,IAAI,EAAEutF,GAAGpvF,CAAC,EAAE0C,EAAE,IAAI0zF,GAAG,EAAE,GAAGt0F,CAAC,EAAE,EAAE,KAAK,gBAAgBY,EAAE,CAACX,CAAC,EAAE,EAAE,CAAC,CAAC1B,EAAEwB,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,MAAM7B,EAAEgC,EAAE,QAAQ,KAAK,KAAK,8BAA8BD,CAAC,EAAE,EAAE,MAAM,CAAC,MAAMnB,EAAEZ,EAAE,EAAE,CAAC,IAAIua,EAAE,EAAE,QAAQ,gCAAgC,GAAGA,EAAE,EAAE,QAAQ,OAAO,IAAI,KAAK,uBAAuB3Z,CAAC,EAAE,IAAI,aAAaA,GAAG,KAAK,MAAM,IAAI,MAAM,uEAAuE,EAAE,IAAIP,EAAE,CAAC,GAAG,KAAK,WAAW,CAAC,EAAE,OAAO,KAAK,QAAQ,IAAIA,EAAE,CAAC,MAAML,EAAE,MAAM,EAAE,OAAOY,EAAE,MAAMmwF,GAAG,OAAO,SAAS,CAAC,CAAC,EAAE1wF,CAAC,CAAC,SAASO,EAAE,CAAC,OAAO,KAAK,QAAQ,IAAIA,CAAC,EAAE,KAAK,QAAQ,IAAIA,CAAC,EAAE,SAAS,CAAC,CAAC,OAAOA,EAAE,CAAC,IAAIZ,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAEZ,EAAE,UAAU,CAAC,OAAOY,EAAE,CAAC,GAAG,KAAK,QAAQ,IAAIA,CAAC,EAAE,CAAC,IAAIZ,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAEZ,EAAE,UAAU,CAAC,CAAC,KAAKY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,GAAG0Y,EAAE,EAAE,QAAQ,OAAO,GAAG,KAAK,uBAAuBva,CAAC,EAAEK,IAAI,YAAY,MAAM,IAAI,MAAM,uEAAuE,EAAE,KAAK,QAAQ,IAAIO,EAAE,CAAC,MAAM,EAAE,MAAMP,EAAE,OAAOL,EAAE,MAAM+wF,GAAG,OAAO,SAASlvF,CAAC,CAAC,CAAC,CAAC,8BAA8BjB,EAAE,CAAC,KAAK,YAAYA,EAAE,MAAM,CAAC,CAAC,SAASA,EAAE,CAAC,IAAIZ,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,OAAO,EAAE,MAAMP,EAAE,mBAAmBwB,EAAE,MAAMC,EAAE,MAAMC,EAAE,SAASC,CAAC,EAAEhC,EAAE,GAAG8B,GAAG,KAAK,CAAC,IAAIc,EAAEZ,EAAEY,EAAE,IAAI65F,GAAG16F,EAAEm6F,EAAE,EAAEt5F,EAAE,IAAI+4F,GAAG55F,EAAEm6F,EAAE,EAAE,IAAIr5F,EAAE,KAAK,gBAAgBD,EAAE,CAAC,CAAC,OAAOhC,EAAE,MAAMmB,EAAE,MAAM1B,CAAC,CAAC,EAAEA,CAAC,EAAEyC,EAAE,KAAK,SAASD,EAAE,MAAM,EAAE,OAAO,KAAK,8BAA8BA,CAAC,EAAEC,CAAC,CAAC,GAAG,GAAG,KAAK,OAAO,KAAK,qBAAqBlC,CAAC,EAAE,GAAGP,IAAI,SAAS,OAAO,EAAE,IAAI,EAAE,KAAK,cAAc,KAAKqC,EAAE,IAAIA,EAAEkkB,EAAE,IAAI,GAAG,IAAI,EAAE,GAAGvmB,IAAI,YAAY,CAAC,IAAIuC,EAAE,KAAK,SAASf,EAAE,KAAK,MAAM,EAAEgB,EAAE,KAAK,SAAShB,EAAE,KAAK,MAAM,EAAE,EAAEy4C,EAAE,uBAAuB13C,EAAEC,CAAC,CAAC,MAAM,EAAE,KAAK,qBAAqBjC,CAAC,EAAE,OAAO,IAAI,KAAK,gBAAgBgmB,EAAE,IAAI,EAAElkB,GAAG,KAAK,qBAAqB9B,EAAE,CAAC,CAAC,CAAC,MAAM,KAAKA,EAAE,CAAC,GAAG,KAAK,YAAY,IAAIA,CAAC,EAAE,CAAC,IAAIkC,EAAE,KAAK,YAAY,IAAIlC,CAAC,EAAE,OAAO,IAAI,QAAQoD,GAAGlB,EAAE,KAAKkB,CAAC,CAAC,CAAC,CAAC,IAAIhE,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,OAAO,EAAE,MAAMP,EAAE,MAAMwB,EAAE,MAAMC,EAAE,mBAAmBC,EAAE,SAASC,CAAC,EAAEhC,EAAE,GAAG6B,GAAG,KAAK,CAAC,IAAIiB,EAAEd,EAAEc,EAAE,IAAI25F,GAAGp8F,EAAE67F,EAAE,EAAEp5F,EAAE,IAAI64F,GAAGt7F,EAAE67F,EAAE,EAAE,IAAIl4F,EAAE,KAAK,gBAAgBlB,EAAE,CAAC,CAAC,OAAOlC,EAAE,MAAMP,EAAE,MAAMyB,CAAC,CAAC,EAAEA,CAAC,EAAEmC,EAAE,KAAK,KAAKD,EAAE,MAAM,EAAE,OAAO,KAAK,8BAA8BA,CAAC,EAAEC,CAAC,CAAC,GAAG,GAAG,KAAK,OAAO,KAAK,qBAAqBrD,CAAC,EAAE,GAAG2Z,EAAE,EAAE,QAAQ,OAAO,GAAG,CAACA,EAAE,EAAE,QAAQ,8BAA8B,GAAGA,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE,MAAM,IAAI,MAAM,8FAA8F,EAAE,IAAI,EAAE,KAAK7X,EAAE,GAAGZ,IAAI,aAAayY,EAAE,EAAE,IAAI,wBAAwB,EAAE,CAAC7X,EAAE,KAAK,OAAO9B,CAAC,EAAE,IAAIkC,EAAE,KAAK,QAAQ,IAAIJ,EAAE,MAAM,EAAE,EAAE,KAAK,MAAM,wBAAwBI,EAAE,QAAQ,QAAQ,GAAGquF,GAAG9wF,CAAC,CAAC,CAAC,CAAC,KAAK,YAAY,IAAIO,EAAE,CAAC,CAAC,EAAEkB,IAAI,aAAa,MAAM,KAAK,MAAM,sBAAsB,EAAE,IAAI,EAAE,GAAGA,IAAI,YAAY,CAAC,IAAIgB,EAAE,MAAM,QAAQ,IAAI,CAAC,KAAK,KAAKf,EAAE,KAAK,MAAM,EAAE,KAAK,KAAKA,EAAE,KAAK,MAAM,CAAC,CAAC,EAAEiC,EAAElB,EAAE,GAAGmB,EAAEnB,EAAE,GAAG,EAAEw3C,EAAE,uBAAuBt2C,EAAEC,CAAC,CAAC,SAAS,GAAG,KAAK,EAAE,KAAK,qBAAqBrD,CAAC,MAAM,CAAC,IAAIkC,EAAE8jB,EAAE,cAAcvmB,CAAC,EAAE,EAAE,KAAK,MAAM,gCAAgC,EAAEyC,CAAC,CAAC,CAAC,GAAGJ,GAAG,MAAM,KAAK,8BAA8BA,CAAC,EAAE,GAAG,KAAK,CAAC,IAAII,EAAE,KAAK,MAAM,GAAGqrF,GAAGrrF,EAAE,IAAIA,EAAE,aAAa,CAAC,CAAC,CAAC,CAAC,IAAIF,EAAE,KAAK,qBAAqBhC,EAAE,CAAC,EAAEiC,EAAE,KAAK,YAAY,IAAIjC,CAAC,EAAE,OAAO,KAAK,YAAY,OAAOA,CAAC,EAAEiC,EAAE,QAAQC,GAAGA,EAAEF,CAAC,CAAC,EAAE,KAAK,gBAAgB,IAAIhC,CAAC,IAAI,KAAK,gBAAgB,OAAOA,CAAC,EAAE,KAAK,YAAYA,CAAC,GAAGwyB,GAAG,EAAE,aAAaxyB,EAAE,IAAI,EAAE,KAAK,kBAAkBgC,CAAC,CAAC,UAAUhC,EAAEZ,EAAE,CAAC,EAAE,CAAC,IAAI,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,OAAOP,EAAE,MAAMwB,EAAE,MAAMC,EAAE,MAAMC,EAAE,SAASC,EAAE,QAAQ,CAAC,EAAE,EAAE,GAAGD,IAAI,YAAY,MAAM,IAAI,MAAM,uDAAuD,EAAE,GAAGD,GAAG,KAAK,CAAC,IAAIe,EAAEb,EAAEa,EAAE,IAAI45F,GAAG56F,EAAEq6F,EAAE,EAAEr5F,EAAE,IAAI84F,GAAG95F,EAAEq6F,EAAE,EAAE,IAAIp5F,EAAE,KAAK,gBAAgBD,EAAE,CAAC,CAAC,OAAOjC,EAAE,MAAMiB,EAAE,MAAME,CAAC,CAAC,EAAEA,CAAC,EAAEiC,EAAE,KAAK,UAAUlB,EAAE9C,CAAC,EAAE,OAAO,KAAK,8BAA8B8C,CAAC,EAAEkB,CAAC,CAAC,GAAG,GAAG,KAAK,MAAM3D,GAAG,KAAK,IAAI,MAAM,gCAAgC,EAAE,IAAI,MAAM,iCAAiC,EAAE,IAAIqC,EAAE,KAAK,OAAO9B,EAAEZ,EAAE,cAAc,EAAE,EAAEozB,GAAG,EAAE,yBAAyB1wB,CAAC,EAAEE,EAAE,KAAK,QAAQ,IAAIF,EAAE,MAAM,EAAE,OAAO,OAAO,OAAO,CAAC,UAAU,CAAC,EAAEE,EAAE,OAAO,CAAC,CAAC,WAAWhC,EAAE,CAAC,IAAIZ,EAAE,KAAK,SAASY,EAAE,MAAM,EAAE,GAAGA,EAAE,QAAQ,SAAS,GAAG,CAAC,IAAI,EAAEZ,EAAE,IAAIK,GAAGumB,EAAE,aAAavmB,CAAC,CAAC,EAAE,OAAOgwB,GAAGzvB,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,OAAO,EAAN,CAAS,MAAM,IAAI,MAAM,kDAAkD,CAAC,CAAC,OAAOyvB,GAAGzvB,EAAE,MAAMA,EAAE,MAAMZ,CAAC,CAAC,CAAC,uBAAuBY,EAAE,CAAC,GAAGA,GAAG,KAAK,QAAQZ,EAAE,EAAEA,EAAEY,EAAE,OAAOZ,IAAI,CAAC,IAAI,EAAEY,EAAEZ,GAAG,GAAG,CAACouF,GAAG,CAAC,EAAE,MAAM7zE,EAAE,EAAE,QAAQ,8BAA8B,EAAE,MAAM,aAAa,gJAAgJ,EAAE,MAAM,aAAa,yCAAyC,CAAC,CAAC,CAAC,qBAAqB3Z,EAAE,CAAC,GAAG,CAAC,MAAMZ,EAAE,MAAM,EAAE,SAASK,CAAC,EAAE,KAAK,QAAQ,IAAIO,CAAC,EAAEiB,EAAE+kB,EAAE,cAAc5mB,CAAC,EAAE,GAAGua,EAAE,EAAE,QAAQ,8BAA8B,EAAE,CAAC,IAAI3X,EAAE,KAAK,OAAOhC,CAAC,EAAEiC,EAAE,KAAK,QAAQ,IAAID,EAAE,MAAM,EAAEE,EAAE,KAAK,MAAM,gCAAgCD,EAAE,QAAQ,QAAQ,GAAGsuF,GAAGnxF,CAAC,CAAC,EAAE,SAAS,EAAE6B,CAAC,EAAE,OAAO,KAAK,8BAA8Be,CAAC,EAAEE,CAAC,CAAC,IAAIhB,EAAEyY,EAAE,EAAE,QAAQ,YAAY,GAAGla,IAAI,GAAG0B,EAAED,EAAEstF,GAAGpvF,CAAC,EAAEA,EAAEgC,EAAEF,EAAE,IAAIo0F,GAAGn0F,CAAC,EAAE,IAAIk0F,GAAGl0F,CAAC,EAAE,EAAE,KAAK,gBAAgBC,EAAE,CAAC,CAAC,MAAMD,EAAE,MAAM,EAAE,OAAOnB,CAAC,CAAC,EAAE,SAAS,EAAE8B,EAAE,KAAK,QAAQ,IAAI,EAAE,MAAM,EAAE,EAAE,KAAK,MAAM,gDAAgDA,EAAE,QAAQ,QAAQA,EAAE,SAAS,GAAGA,EAAE,SAAS,EAAE,EAAE,SAAS,EAAEb,CAAC,EAAE,OAAO,KAAK,8BAA8B,CAAC,EAAE,CAAC,CAAC,gBAAgB,CAAC,OAAO0Y,EAAE,EAAE,UAAU,+CAA+C,EAAE,CAAC,CAAC,KAAK3Z,EAAE,CAAC,IAAIZ,EAAE,KAAK,aAAa,EAAE,CAAC,EAAEK,EAAE,GAAG,KAAK,oBAAoB,MAAM,KAAK,mBAAmB,EAAEA,EAAE,IAAI,KAAK,aAAa,KAAK,CAAC,EAAE,KAAK,aAAa,EAAEO,EAAE,EAAE,IAAIiB,EAAE+kB,EAAE,QAAQ,KAAK,aAAa,IAAI5kB,GAAGA,EAAE,KAAK,CAAC,EAAE,OAAOA,GAAGA,GAAG,IAAI,EAAEF,EAAE8kB,EAAE,QAAQ,KAAK,aAAa,IAAI5kB,GAAGA,EAAE,IAAI,CAAC,EAAE,OAAOA,GAAGA,GAAG,IAAI,EAAE,KAAK,aAAahC,EAAEK,IAAI,KAAK,mBAAmB,MAAM,IAAI0B,EAAE,CAAC,aAAa,KAAK,aAAa,eAAe,KAAK,eAAe,SAAS,KAAK,OAAO,IAAI,EAAE,OAAO,SAAS,CAAC,GAAGwY,EAAE,EAAE,UAAU,+CAA+C,EAAE,EAAE,CAAC,IAAIvY,EAAE,MAAM,QAAQ,IAAIH,CAAC,EAAEE,EAAE,SAAS6kB,EAAE,IAAI5kB,CAAC,EAAED,EAAE,oBAAoB,IAAIC,EAAE,IAAI,CAAC,EAAEU,KAAK,CAAC,KAAKZ,EAAEY,GAAG,GAAG,CAAC,EAAE,EAAE,IAAI,GAAG,GAAG,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,IAAI,CAAC,MAAMX,EAAE,SAAS,CAAC,MAAM,2DAA2D,EAAE,OAAO,KAAK,aAAa,EAAE,KAAK,eAAe,EAAEA,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,WAAW,GAAG,cAAc,KAAK,cAAc,uBAAuB,KAAK,eAAe,kBAAkB,kBAAkB,KAAK,eAAe,YAAY,CAAC,CAAC,YAAY,CAAC,OAAOwY,EAAE,EAAE,UAAU,+CAA+C,EAAE,EAAE,KAAK,MAAM,WAAW,EAAE,CAAC,QAAQqM,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC,CAAC,SAAShmB,EAAE,CAAC,OAAO2Z,EAAE,EAAE,UAAU,+CAA+C,EAAE,GAAG,KAAK,MAAM,SAAS,EAAE3Z,IAAIA,EAAE,MAAMgmB,EAAE,IAAI,EAAEhmB,EAAE,CAAC,MAAM,aAAaA,EAAE,CAAC,GAAG2Z,EAAE,EAAE,UAAU,+CAA+C,EAAE,EAAE,OAAO,KAAK,MAAM,uBAAuB3Z,CAAC,EAAE,IAAIZ,EAAEY,EAAE,OAAOZ,EAAE,MAAMA,EAAE,OAAO,CAAC,YAAYY,EAAEZ,EAAE,GAAG,CAAC,GAAG,KAAK,gBAAgB,IAAIY,CAAC,EAAE,MAAM,GAAG,GAAG,CAAC,KAAK,QAAQ,IAAIA,CAAC,EAAE,MAAM,GAAG,GAAGZ,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,SAAS,EAAE,KAAK,QAAQ,IAAIA,CAAC,EAAE,WAAW,CAACZ,GAAG,KAAK,QAAQ,IAAIY,CAAC,EAAE,SAAS,EAAE,MAAM,GAAG,GAAG,KAAK,YAAY,IAAIA,CAAC,EAAE,OAAO,KAAK,gBAAgB,IAAIA,CAAC,EAAE,KAAK,iBAAiB,GAAG,KAAK,eAAeA,CAAC,EAAE,GAAG,CAAC,mBAAmB,CAAC,EAAE,KAAK,QAAQ,IAAIA,CAAC,EAAE,OAAO,GAAG,OAAO,KAAK,YAAY,EAAE,KAAK,OAAOZ,CAAC,EAAE,KAAK,YAAY,EAAE,KAAK,OAAOA,CAAC,GAAG,KAAK,QAAQ,OAAOY,CAAC,EAAE,EAAE,CAAC,eAAeA,EAAE,CAAC,GAAG,CAAC,QAAQZ,EAAE,MAAM,EAAE,SAASK,EAAE,MAAMwB,EAAE,SAASC,EAAE,MAAMC,CAAC,EAAE,KAAK,QAAQ,IAAInB,CAAC,EAAEoB,EAAED,GAAGA,EAAE,YAAYnB,EAAE,EAAE,KAAK,aAAa,IAAIoB,CAAC,EAAE,EAAE,EAAE,KAAK,aAAa,IAAIA,EAAE,EAAE,CAAC,GAAG,KAAK,aAAa,OAAOA,CAAC,EAAEhC,GAAG,OAAO,KAAK,eAAe,KAAK,aAAaK,EAAE,CAAC,EAAE,KAAK,eAAe,eAAeL,EAAEK,EAAEwB,EAAEC,CAAC,IAAI,IAAIY,EAAE,KAAK,QAAQ,IAAI9B,CAAC,EAAE8B,EAAE,QAAQ,KAAKA,EAAE,SAAS,KAAKA,EAAE,SAAS,GAAGA,EAAE,MAAM,IAAI,CAAC,WAAW9B,EAAE,CAAC,OAAO,KAAK,YAAYA,CAAC,EAAE,KAAK,QAAQ,IAAIA,CAAC,EAAE,QAAQ,OAAO,CAAC,YAAYA,EAAE,CAAC,OAAO,KAAK,QAAQ,IAAIA,CAAC,CAAC,CAAC,mBAAmBA,EAAEZ,EAAEg9F,GAAI,CAAC,OAAOziF,EAAE,EAAE,QAAQ,mBAAmB,GAAG3Z,EAAE,MAAM,GAAG,KAAK,QAAQ,IAAI,EAAE,MAAM,EAAE,SAAS,MAAMgmB,EAAE,cAAc,EAAE,KAAK,EAAE5mB,CAAC,CAAC,CAAC,iBAAiB,CAAC,OAAO,KAAK,KAAK,CAAC,MAAMY,EAAE,CAAC05C,EAAE,KAAK,uEAAuE,EAAE,IAAIt6C,EAAEY,EAAE,SAAS,EAAE,OAAO+7F,GAAI/7F,EAAE,MAAMZ,CAAC,CAAC,CAAC,cAAcY,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,IAAIo8F,GAAG77F,EAAE,MAAMZ,CAAC,EAAE6B,EAAE,KAAK,cAAcxB,EAAE,CAACO,CAAC,EAAE,CAAC,EAAE,OAAOwyB,GAAG,EAAE,yBAAyBvxB,CAAC,CAAC,CAAC,IAAIjB,EAAE,CAAC,GAAG,KAAK,mBAAmB,CAACA,CAAC,CAAC,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAIP,EAAE05F,GAAG,KAAK,QAAQ,IAAIn5F,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,KAAK,WAAWA,EAAE,MAAMA,EAAE,MAAMP,CAAC,CAAC,CAAC,GAAGka,EAAE,EAAE,QAAQ,6BAA6B,EAAE,OAAO,KAAK,cAAc3Z,EAAEk7F,GAAGl7F,EAAE,KAAK,EAAE,IAAIZ,EAAE,IAAI27F,GAAG/6F,EAAE,MAAMk7F,EAAE,EAAE,EAAE,KAAK,cAAc97F,EAAE,CAACY,CAAC,CAAC,EAAE,OAAOwyB,GAAG,EAAE,yBAAyB,CAAC,CAAC,CAAC,eAAexyB,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,GAAGL,IAAI,UAAU,GAAG,MAAM,EAAE,OAAO,GAAG4mB,EAAE,SAAS,EAAE,EAAE,EAAE,CAAC,IAAI/kB,EAAE,EAAE,IAAIC,GAAG8kB,EAAE,aAAa9kB,CAAC,CAAC,EAAEzB,EAAE,KAAK,MAAMwB,EAAEjB,EAAEZ,CAAC,CAAC,MAAMK,EAAE,KAAK,MAAM,EAAEO,EAAEZ,CAAC,EAAE,OAAO,KAAK,QAAQ,IAAIK,CAAC,EAAE,MAAM,KAAK,CAAC,OAAOA,EAAE,MAAMO,EAAE,MAAMZ,CAAC,CAAC,CAAC,WAAWY,EAAEZ,EAAE,EAAE,CAAC,OAAOozB,GAAG,EAAE,yBAAyB,KAAK,eAAexyB,EAAEZ,EAAE,CAAC,EAAE,IAAI,CAAC,CAAC,aAAaY,EAAE,CAAC,IAAIZ,EAAE,IAAI08F,GAAG97F,EAAE,KAAK,EAAE,OAAO,KAAK,gBAAgBZ,EAAE,CAACY,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,WAAWA,EAAE,CAAC,IAAIZ,EAAE,IAAIi7F,GAAGr6F,EAAE,KAAK,EAAE,EAAE,GAAG,OAAO,KAAK,gBAAgBZ,EAAE,CAACY,CAAC,EAAEA,EAAE,MAAM,KAAK,CAAC,CAAC,CAAC,cAAcA,EAAEZ,EAAE,CAAC,IAAI,EAAE,CAAC4uF,GAAGhuF,EAAE,KAAK,EAAE,GAAGuuF,GAAGvuF,EAAE,KAAK,CAAC,EAAEP,EAAE,CAAC,MAAMO,EAAE,MAAM,MAAM,EAAE,OAAOA,EAAE,MAAM,EAAEiB,EAAE,CAAC+sF,GAAG5uF,CAAC,EAAE,GAAGmvF,GAAGnvF,CAAC,CAAC,EAAE8B,EAAE,IAAIo5F,GAAGr5F,EAAE,CAAC,EAAEE,EAAE,GAAGC,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,gBAAgBF,EAAE,CAACzB,CAAC,EAAEO,EAAE,MAAMoB,EAAED,CAAC,EAAE,MAAM,CAAC,OAAO,EAAE,OAAO,MAAM/B,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,OAAOY,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,SAASP,EAAE,MAAMwB,EAAE,MAAMC,CAAC,EAAE,EAAE,GAAG9B,GAAG,KAAK,CAAC,IAAI4C,EAAEgkB,EAAE,cAAc/kB,CAAC,EAAEgB,EAAE7C,EAAE,GAAGA,EAAE,GAAG,EAAE4mB,EAAE,OAAOhkB,GAAGC,EAAE,IAAI,2GAA2G,CAAC,CAAC,IAAId,EAAEqtF,GAAGvtF,CAAC,EAAEG,EAAE3B,EAAE2B,EAAE,IAAIg0F,GAAGj0F,CAAC,EAAEC,EAAE,IAAI+zF,GAAGh0F,CAAC,EAAE,IAAI,EAAE,GAAGW,EAAE,CAAC1C,GAAG,KAAKA,EAAEmxF,GAAGpvF,CAAC,CAAC,EAAE,EAAE,KAAK,gBAAgBC,EAAE,CAAC,CAAC,MAAMD,EAAE,MAAMD,EAAE,OAAOlB,CAAC,CAAC,EAAEkB,EAAEY,EAAE,EAAE1C,CAAC,EAAE,MAAM,CAAC,MAAM8B,EAAE,MAAMD,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,gBAAgBjB,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,GAAGC,EAAE,CAAC,IAAIC,EAAE,KAAK,eAAenB,EAAE,YAAY,CAAC,EAAEoB,EAAE,KAAK,QAAQ,IAAID,EAAE,MAAM,EAAE,GAAGnB,EAAE,eAAeoB,EAAE,SAAS,IAAIpB,EAAE,mBAAmBkwF,GAAG,MAAM,CAAC,IAAIzpF,EAAEvF,GAAG,KAAKA,EAAEqvF,GAAGvwF,EAAE,WAAW,EAAEoB,EAAE,SAASqF,EAAE,IAAIxC,GAAGA,EAAE,CAAC,CAAC,CAAC,GAAGjE,EAAE,aAAa,OAAOoB,EAAE,MAAMpB,EAAE,aAAagmB,EAAE,cAAc7kB,EAAE,KAAK,IAAI,EAAE,OAAOC,EAAE,OAAO4kB,EAAE,uBAAuB7kB,EAAE,MAAM,CAAC,EAAEA,EAAE,IAAI,EAAE,CAAC,EAAEW,EAAE1C,EAAE,IAAIqH,GAAG,CAAC,GAAGA,EAAE,QAAQ,YAAY,MAAM,IAAI,MAAM,iIAAiI,EAAE,IAAIxC,EAAE,KAAK,QAAQ,IAAIwC,EAAE,MAAM,EAAE,GAAGxC,EAAE,SAAS,KAAK,CAAC,GAAG,CAACjE,EAAE,cAAcgmB,EAAE,cAAcvf,EAAE,KAAK,GAAGkT,EAAE,EAAE,UAAU,2BAA2B,EAAE,MAAM,CAAC,MAAMlT,EAAE,MAAM,QAAQ,KAAK,UAAU,GAAG,cAAcxC,EAAE,MAAM,EAAEjE,EAAE,eAAeiE,EAAE,SAAS,GAAGA,EAAE,MAAMwC,EAAE,MAAM,CAAC,GAAG,KAAK,YAAYA,EAAE,MAAM,EAAE,CAAC,CAACxC,EAAE,UAAU,CAAC,CAACjE,EAAE,aAAayG,EAAExC,EAAE,SAAS,KAAK,aAAawC,CAAC,EAAE,KAAK,WAAWA,CAAC,EAAE,EAAE,KAAKA,CAAC,EAAExC,EAAE,KAAK,QAAQ,IAAIwC,EAAE,MAAM,UAAUxC,EAAE,UAAU,CAAC+qF,GAAG/qF,EAAE,MAAMwC,EAAE,KAAK,EAAE,CAAC,IAAI5C,EAAE4C,EAAE3C,EAAE2C,EAAE,MAAMA,EAAE,MAAMxC,EAAE,MAAMwC,EAAE,KAAK,cAAcA,EAAE3C,CAAC,EAAE,EAAE,KAAK2C,CAAC,EAAExC,EAAE,KAAK,QAAQ,IAAIwC,EAAE,MAAM,EAAE5C,EAAE,MAAMC,CAAC,CAAC,MAAM,CAAC,MAAM2C,EAAE,MAAM,QAAQxC,EAAE,UAAU,EAAE,CAAC,CAAC,EAAE,KAAK,YAAY9C,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,MAAMA,EAAE,MAAM,QAAQC,EAAE,UAAU,EAAE,EAAEY,EAAEizF,GAAGj1F,EAAE8B,EAAE,CAAC,EAAEG,EAAE,KAAK,iBAAiBD,EAAE,IAAI6yF,GAAG,KAAK,MAAM70F,EAAE8B,EAAE,CAAC,CAAC,EAAEI,EAAE,KAAK,cAAc,KAAKkB,EAAElB,IAAIkB,EAAE,KAAK,WAAW,GAAGuW,EAAE,EAAE,IAAI,qBAAqB,GAAGq7E,GAAG,KAAK,MAAM/yF,EAAEH,EAAE,EAAErC,CAAC,EAAE,EAAE,QAAQgH,GAAG,KAAK,8BAA8BA,CAAC,CAAC,EAAEvE,IAAIkB,EAAE,KAAK,SAASA,CAAC,EAAE,KAAK,aAAa,KAAK,CAAC,KAAKpD,EAAE,YAAY,KAAK,MAAM,KAAK,aAAaoD,CAAC,CAAC,CAAC,GAAG,IAAIC,EAAEsW,EAAE,EAAE,IAAI,uBAAuB,EAAE,GAAGtW,EAAE,EAAE,CAAC,IAAIoD,EAAEuf,EAAE,IAAI,EAAEvf,EAAE,KAAK,gBAAgBpD,IAAI,KAAK,MAAM,GAAG,MAAM,EAAE,KAAK,gBAAgBoD,EAAE,CAAC,GAAG,CAACkT,EAAE,EAAE,QAAQ,qBAAqB,GAAGvY,EAAE,UAAUH,IAAI,GAAG,CAAC,IAAIwF,EAAE,KAAK,aAAatF,CAAC,EAAE,OAAO,KAAK,8BAA8BA,CAAC,EAAEsF,CAAC,CAAC,OAAOtF,CAAC,CAAC,cAAcnB,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,GAAG,CAAC,OAAO,EAAE,GAAG7B,EAAE,GAAG,MAAM,KAAK,gBAAgBY,EAAEZ,EAAE,EAAEK,EAAEwB,CAAC,CAAC,CAAC,iBAAiBjB,EAAEZ,EAAE,CAAC,OAAOY,KAAK,KAAK,cAAc,KAAK,YAAYA,GAAGZ,EAAE,GAAG,KAAK,YAAYY,EAAE,CAAC,mBAAmB,CAAC,OAAO,KAAK,cAAc,CAAC,SAAS,CAAC,KAAK,WAAW2Z,EAAE,EAAE,QAAQ,SAAS,GAAG,OAAO,KAAK,KAAK,WAAW,EAAE,QAAQ,GAAG,CAAC,KAAK,MAAM,cAAc,KAAK,YAAY,GAAG,YAAY,EAAE,OAAO,KAAK,YAAY,EAAE,CAAC,EAAE,KAAK,eAAe,QAAQ,EAAE,KAAK,QAAQ,MAAM,OAAO,mBAAmB,aAAa,KAAK,kBAAkB,kBAAkB,KAAK,OAAO,OAAO,EAAE,KAAK,OAAO,KAAK,KAAK,sBAAsB,KAAK,MAAM,QAAQ,KAAK,KAAK,MAAM,QAAQ,GAAG,KAAK,SAAS,GAAG,CAAC,gBAAgB,CAAC,OAAO,KAAK,qBAAqB,OAAO,KAAK,oBAAoBgZ,EAAE,IAAI,CAAC,GAAG,CAAChZ,EAAE,EAAE,IAAI,8BAA8B,EAAE,CAAC,IAAI3Z,EAAE2Z,EAAE,EAAE,QAAQ,OAAO,EAAEA,EAAE,EAAE,IAAI,QAAQ,EAAE,EAAE,IAAIva,EAAE,KAAK,IAAI2iC,GAAG,IAAI,CAAC,EAAE,SAAS,EAAE,GAAG,GAAGpoB,EAAE,EAAE,IAAI,QAAQ3Z,CAAC,EAAEZ,EAAE,EAAE,MAAO,GAAE,CAAC,MAAO,GAAE,CAAC,GAAG,KAAK,mBAAmB,CAAC,SAAS,CAAC,OAAO,KAAK,eAAe,IAAI,GAAG48F,GAAIC,EAAG,CAAC,YAAYj8F,EAAE,CAAC,IAAIZ,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,MAAM,EAAE,MAAMP,EAAE,OAAOwB,EAAE,QAAQC,EAAE,MAAMC,EAAE,SAASC,CAAC,EAAEhC,EAAE,GAAG8B,GAAG,KAAK,OAAO,IAAI,EAAE,KAAK,cAAc,KAAKY,EAAE,IAAIA,EAAEkkB,EAAE,IAAI,GAAG,IAAI,EAAE5mB,EAAE,SAAS,GAAG,GAAG,OAAO,EAAEqvF,GAAG,EAAErtF,CAAC,EAAEhC,EAAE,SAAS,GAAG6B,GAAG,KAAK,CAAC,IAAIe,EAAEwsF,GAAG,CAAC,EAAEvsF,EAAEC,EAAE,EAAE,GAAGkB,EAAE,EAAE,GAAGC,EAAEpC,aAAa,YAAYA,aAAa,mBAAmBG,GAAG,CAACiC,KAAK,CAACnB,EAAEkB,CAAC,EAAEotF,GAAG,EAAE,GAAG,EAAE,EAAE,GAAGpvF,EAAEa,EAAE,IAAIwzF,GAAGzzF,EAAEqB,CAAC,EAAEpB,EAAE,IAAIuzF,GAAGxzF,EAAEqB,CAAC,EAAE,IAAIoD,EAAEpD,EAAE,CAACD,EAAElB,CAAC,EAAE,EAAE+B,EAAE,KAAK,eAAewC,EAAEhH,CAAC,EAAEoE,EAAE,KAAK,QAAQ,IAAII,EAAE,MAAM,EAAEZ,EAAEQ,EAAE,MAAMssF,GAAG,OAAOtsF,EAAE,MAAMssF,GAAG,OAAOtsF,EAAE,SAAS4C,EAAE,KAAK,MAAM,2BAA2B,KAAK,WAAWxC,EAAE,MAAM,EAAE/B,EAAEkB,EAAEnC,CAAC,EAAE,IAAI6C,EAAE,CAAC,CAACV,EAAElB,CAAC,CAAC,EAAE6B,EAAE,GAAGC,EAAE,KAAK,gBAAgB/B,EAAE,CAACgC,CAAC,EAAExE,EAAEqE,EAAEC,CAAC,EAAE2C,EAAE,KAAK,QAAQ,IAAI1C,EAAE,MAAM,EAAE5E,EAAE,SAASsH,EAAE,SAAStH,EAAE,SAASsH,EAAE,SAAStH,EAAE,MAAMsH,EAAE,MAAMiT,EAAE,EAAE,IAAI,qBAAqB,EAAE,KAAK,YAAY3V,EAAE,MAAM,GAAG5E,EAAE,QAAQsH,EAAE,QAAQtH,EAAE,OAAO,KAAK,KAAK,QAAQ,OAAO4E,EAAE,MAAM,GAAG,KAAK,8BAA8BC,CAAC,EAAE,IAAI,KAAK,cAAc+hB,EAAE,IAAI,EAAElkB,EAAE,KAAK,CAAC,IAAIE,EAAE,KAAK,eAAe,EAAEb,EAAE1B,EAAE2B,CAAC,EAAEhC,EAAE,QAAQ4C,CAAC,CAAC,CAAC,qBAAqBhC,EAAEZ,EAAE,CAAC,IAAI,EAAE,KAAK,QAAQ,IAAIY,CAAC,EAAE,CAAC,MAAMP,CAAC,EAAE,EAAE,OAAO,KAAK,eAAeO,CAAC,EAAEZ,GAAG,OAAO,EAAE,OAAOo9F,GAAIp9F,EAAEK,CAAC,GAAG,EAAE,MAAM,CAAC,eAAeO,EAAEZ,EAAE,EAAEK,EAAE,CAAC,GAAG,KAAK,eAAe,KAAK,aAAaO,EAAE,CAAC,EAAE,CAAC,KAAK,mBAAmB,KAAK,cAAc,KAAK,mBAAmB,KAAK,KAAK,CAAC,IAAIiB,GAAG,KAAK,cAAc,KAAK,MAAM,QAAQ,CAAC,EAAE,KAAK,kBAAkB,GAAG,QAAQ,KAAK,6BAA6BA,wCAAwC,CAAC,CAAC,OAAO,KAAK,eAAe,eAAejB,EAAEZ,EAAEK,CAAC,CAAC,CAAC,aAAaO,EAAEZ,EAAE,CAAC,OAAOY,EAAE,GAAGA,EAAE,GAAGgmB,EAAE,gBAAgB5mB,CAAC,CAAC,CAAC,wBAAwB,CAAC,OAAO,CAAC,CAACY,CAAC,IAAI,OAAO,QAAQ,KAAK,WAAW,EAAE,KAAK,iBAAiBA,CAAC,CAAC,CAAC,MAAM,6BAA6B,CAAC,IAAIA,EAAE,CAAC,EAAE,GAAG,KAAK,MAAM,6BAA6B,CAAC,OAAO,CAAC,CAACZ,CAAC,IAAI,OAAO,QAAQ,KAAK,WAAW,EAAEY,EAAE,KAAK,KAAK,sBAAsBZ,CAAC,CAAC,EAAE,OAAO,QAAQ,IAAIY,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAACZ,CAAC,IAAI,OAAO,QAAQ,KAAK,WAAW,EAAE,CAAC,IAAI,EAAE,IAAI,QAAQK,GAAG,CAAC,GAAG,CAAC,KAAK,iBAAiBL,CAAC,EAAEK,EAAE,EAAE,CAAC,OAAOwB,EAAN,CAAS,MAAMA,CAAC,CAAC,CAAC,EAAEjB,EAAE,KAAK,CAAC,CAAC,CAAC,OAAO,QAAQ,IAAIA,CAAC,CAAC,CAAC,CAAC,MAAM,sBAAsBA,EAAE,CAAC,OAAO,KAAK,MAAM,GAAG,oBAAoBA,EAAE,aAAa,KAAK,MAAM,6BAA6B,qBAAqB,EAAE,KAAK,iBAAiBA,CAAC,GAAG,MAAMy5C,GAAG,EAAE,KAAK,sBAAsBz5C,CAAC,EAAE,CAAC,iBAAiBA,EAAE,CAAC,GAAG,KAAK,MAAM,GAAG,oBAAoBA,EAAE,aAAa,KAAK,MAAM,GAAG,WAAW,IAAI,GAAG,MAAM,QAAQ,IAAI,KAAK,MAAM,GAAG,kBAAkBA,EAAE,YAAY,CAAC,EAAE,KAAK,MAAM,GAAG,mBAAmBA,EAAE,eAAe,KAAK,MAAM,GAAG,cAAc,IAAI,IAAIovF,GAAGpvF,EAAE,OAAO,KAAK,MAAM,GAAG,iBAAiBA,EAAE,cAAc,CAAC,EAAE,IAAI,MAAM,oCAAoC,GAAG,IAAI,MAAM,6CAA6C,EAAE,MAAM,EAAE,CAAC,qBAAqB,CAAC,OAAO,CAAC,CAACA,CAAC,IAAI,OAAO,QAAQ,KAAK,WAAW,EAAE,CAAC,GAAG,CAAC,iBAAiBZ,EAAE,uBAAuB,EAAE,OAAOK,EAAE,OAAOwB,EAAE,kBAAkBC,EAAE,qBAAqBC,EAAE,iBAAiBC,EAAE,wBAAwB,EAAE,oBAAoBU,CAAC,EAAEgzF,GAAG,KAAK,MAAM90F,EAAE,QAAQA,EAAE,YAAY,EAAEA,EAAE,iBAAiBZ,EAAEY,EAAE,uBAAuB,EAAEA,EAAE,OAAOP,EAAEO,EAAE,OAAOiB,EAAEjB,EAAE,kBAAkBkB,EAAElB,EAAE,qBAAqBmB,EAAEnB,EAAE,iBAAiBoB,EAAEpB,EAAE,wBAAwB,EAAEA,EAAE,oBAAoB8B,CAAC,CAAC,CAAC,wBAAwB9B,EAAEZ,EAAE,EAAE,CAAC,GAAG,CAAC,QAAQK,EAAE,OAAOwB,EAAE,MAAMC,EAAE,SAASC,CAAC,EAAEnB,EAAEoB,EAAEoxB,GAAG,EAAE,QAAQ,GAAG,CAACpxB,EAAE,MAAM,GAAG,UAAU3B,CAAC,EAAE,MAAM,IAAI,MAAM,+RAA+R,EAAE,IAAI,EAAE2B,EAAE,aAAa3B,EAAEL,EAAE,EAAE6B,EAAEC,EAAEC,CAAC,EAAE,OAAOqxB,GAAG,EAAE,qBAAqB,EAAEpzB,EAAE,EAAEgC,CAAC,CAAC,CAAC,EAAEm7F,GAAG,WAAW,EAAE,SAASC,GAAIt9F,EAAE,EAAE,CAAC,GAAG,IAAI,WAAW,IAAI,YAAY,OAAOA,EAAE,GAAG,IAAI,SAAS,IAAI,OAAO,CAAC,IAAIE,EAAE,IAAI,QAAQ,IAAI,WAAWF,EAAE,MAAM,EAAE,IAAI,WAAWA,EAAE,MAAM,EAAE,QAAQM,EAAE,EAAEA,EAAEJ,EAAE,OAAO,EAAEI,EAAEJ,EAAEI,GAAG,KAAK,MAAMN,EAAEM,EAAE,EAAE,OAAOJ,CAAC,KAAM,OAAM,IAAI,MAAM,iBAAiB,GAAG,CAAC,CAAC,IAAIq9F,GAAG,QAAQ,SAASC,IAAI,CAAC/iF,EAAE,EAAE,IAAI,2BAA2B,EAAE,CAAC,CAAC6Q,GAAG,UAAU,GAAG6I,GAAG,QAAQ,IAAI,IAAIkpE,GAAG,CAAC,EAAE,IAAII,GAAI,CAAC,eAAeD,EAAE,EAAME,GAAG;AAAA;AAAA;AAAA,EAG1yiBC,GAAG,KAAK,CAAC,YAAY78F,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,YAAYs6C,EAAE,2BAA2Bt6C,EAAE,CAAC,EAAE,KAAK,oBAAoB81F,GAAG,KAAK,YAAY,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA,UAE3Kl1F;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAQL,CAAC,EAAM88F,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA,EAKTC,GAAG,KAAK,CAAC,YAAY/8F,EAAEZ,EAAE,EAAEK,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,qBAAqB,GAAG,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYi6C,EAAE,2BAA2Bt6C,EAAE,CAAC,EAAE,IAAI6B,EAAE,KAAK,YAAY,OAAO,KAAK,oBAAoBi0F,GAAGj0F,CAAC,EAAE,IAAIC,EAAE,GAAG,GAAGzB,EAAE,GAAGwB,IAAI,GAAG+kB,EAAE,cAAc,KAAK,WAAW,IAAI,EAAE9kB,EAAE;AAAA;AAAA;AAAA;AAAA,kBAIzSA,EAAE;AAAA,YACR0zF,GAAG3zF,CAAC;AAAA,UACNA,IAAI,EAAE,KAAK,oBAAoBC,GAAG;AAAA;AAAA;AAAA;AAAA,YAIhCA,GAAG;AAAA,yCAC0B,KAAK,YAAY;AAAA;AAAA;AAAA,gBAG1C,CAAC,IAAIE,EAAE+4F,GAAG,SAASl5F,CAAC,EAAE,KAAK,oBAAoBC,GAAG;AAAA;AAAA,iBAEjDE,EAAEH,EAAE,uBAAuBA;AAAA;AAAA,iBAE3BG,EAAEH,EAAE,uBAAuBA;AAAA;AAAA;AAAA;AAAA,YAIhCC,GAAG;AAAA;AAAA,iBAEEE,EAAEH,EAAE,cAAc,KAAK,YAAYA,EAAE;AAAA;AAAA,iBAErCG,EAAEH,EAAE,cAAc,KAAK,YAAYA,EAAE;AAAA;AAAA;AAAA;AAAA,WAI3C,CAAC,KAAK,SAAS;AAAA;AAAA,UAEhBjB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAQAkB;AAAA;AAAA;AAAA;AAAA,KAIL,CAAC,EAAE,SAAS87F,GAAG99F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,OAAOJ,EAAE,OAAOI,EAAE,MAAM,EAAE,CAAC,OAAOA,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,IAAIy9F,GAAG,CAAC,WAAWz+E,GAAG,YAAY,QAAQ,WAAWw+E,EAAE,EAAE,SAASE,GAAGh+F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,KAAKC,CAAC,EAAE,EAAE,EAAEL,EAAE,eAAeI,EAAE,MAAM,WAAW,EAAE,EAAEJ,EAAE,QAAQ,IAAI,EAAE,MAAM,EAAE+B,EAAE67F,GAAG,CAAC,OAAO,CAAC,EAAEx9F,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAEgC,EAAE47F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,OAAO,EAAE,mBAAmB,CAAC,KAAK+B,EAAE,KAAKC,CAAC,EAAE,CAAC,CAAC,IAAI+7F,GAAG,CAAC,WAAWrhF,GAAG,YAAY,QAAQ,WAAWohF,EAAE,EAAME,GAAG,+BAA+BC,GAAG;AAAA;AAAA;AAAA,EAGngB,SAASC,GAAIp+F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,EAAEJ,EAAE,eAAe,CAAC,EAAE,UAAU4mB,EAAE,kBAAkB,EAAE,SAAS,CAAC,EAAE7kB,EAAEwY,EAAE,EAAE,QAAQ,8BAA8B,EAAE,IAAIojF,GAAGM,GAAG59F,EAAE,MAAM,EAAE,KAAK,EAAE,IAAIo9F,GAAGO,GAAG39F,EAAE,MAAM,EAAE,KAAK,EAAE2B,EAAEhC,EAAE,gBAAgB+B,EAAE,CAAC1B,EAAE,CAAC,EAAE,SAAS,EAAE,OAAOL,EAAE,8BAA8B,CAAC,EAAEgC,CAAC,CAAC,IAAIm8F,GAAG,CAAC,WAAWz+E,GAAG,YAAY,QAAQ,WAAWw+E,EAAG,EAAME,GAAG,+BAA+BC,GAAG;AAAA;AAAA;AAAA,EAGla,SAASC,GAAIx+F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,EAAE,MAAMC,CAAC,EAAE,EAAE,EAAEka,EAAE,EAAE,QAAQ,8BAA8B,EAAE,IAAIojF,GAAGU,GAAGj+F,EAAE,MAAMC,EAAE,KAAK,EAAE,IAAIo9F,GAAGW,GAAGh+F,EAAE,MAAMC,EAAE,KAAK,EAAE,OAAOL,EAAE,gBAAgB,EAAE,CAACI,EAAEC,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIk+F,GAAG,CAAC,WAAWt8E,GAAG,YAAY,QAAQ,WAAWq8E,EAAG,EAAME,GAAG,0BAA0B,SAASC,GAAG,CAAC,UAAU3+F,EAAE,gBAAgB,EAAE,cAAcE,EAAE,MAAMI,CAAC,EAAE,CAAC,MAAM,CAAC,CAAC,OAAOC,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,EAAEA,EAAE0B,EAAE,EAAEC,EAAE5B,GAAG,EAAE,MAAM,GAAG2B,EAAE,mBAAmB,CAAC,CAAC,CAAC,GAAG/B,GAAG,KAAK,CAAC,IAAI2C,EAAEZ,EAAE,QAAQ,IAAI,EAAE,MAAM,EAAEa,EAAE5C,EAAE2C,EAAE,OAAOX,CAAC,EAAE,OAAOD,EAAE,eAAe,EAAE,MAAMC,EAAEY,CAAC,CAAC,CAAC,IAAIX,EAAEsY,EAAE,EAAE,QAAQ,6BAA6B,GAAG,GAAG,KAAK7X,EAAE,OAAOT,EAAES,EAAE,IAAI+5F,GAAG,EAAE,MAAM,CAAC,EAAE/5F,EAAE,IAAIi5F,GAAG,EAAE,MAAM77F,CAAC,EAAEiC,EAAE,gBAAgBW,EAAE,CAAC,CAAC,EAAEV,CAAC,CAAC,CAAC,CAAC,SAAS08F,GAAG,CAAC,UAAU5+F,EAAE,gBAAgB,EAAE,iBAAiBE,EAAE,GAAG,gBAAgBI,EAAE,GAAG,cAAcC,EAAE,MAAM,CAAC,EAAE,CAAC,MAAM,CAAC,CAAC,OAAO,EAAE,QAAQ0B,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEC,EAAE,EAAEC,CAAC,EAAE,EAAES,EAAEX,EAAE,GAAG3B,GAAG4B,EAAE,QAAQ,YAAY,CAAC,IAAIc,EAAEJ,EAAE,QAAQ,IAAIV,EAAE,MAAM,EAAEgC,EAAEtB,EAAE,QAAQ,IAAIT,EAAE,MAAM,EAAE,CAAC,EAAEoF,CAAC,EAAE,CAAC,CAACvE,EAAE,mBAAmB,KAAKkB,EAAE,mBAAmB,IAAI,EAAE,CAAClB,EAAE,mBAAmB,KAAKkB,EAAE,mBAAmB,IAAI,CAAC,EAAE,IAAIS,GAAG,CAAC,GAAG,CAACC,EAAEC,CAAC,EAAEF,EAAEG,EAAE,CAAC,OAAOF,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAM1C,EAAE,KAAK,EAAEsF,EAAE,CAAC,OAAO3C,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAM1C,EAAE,KAAK,EAAE6C,EAAE,IAAI24F,GAAG39F,EAAEkC,EAAE,MAAMC,EAAE,KAAK,EAAE,OAAOS,EAAE,gBAAgBoC,EAAE,CAACF,EAAE0C,CAAC,EAAEmjB,GAAG/lB,EAAE,MAAMC,EAAE,KAAK,CAAC,CAAC,CAAC,EAAE,EAAEm5F,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAKz2F,CAAC,EAAE,QAAQ3E,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B2E,CAAC,EAAE,CAAC,CAAC,IAAI1E,EAAE,GAAG8nB,GAAGzoB,EAAE,MAAMC,EAAE,KAAK,EAAE,IAAID,EAAE,QAAQ,UAAUC,EAAE,QAAQ,UAAUS,EAAE,mBAAmB,CAACV,EAAEC,CAAC,CAAC,IAAI5B,GAAG,KAAK,CAAC,IAAIyC,EAAEJ,EAAE,QAAQ,IAAIV,EAAE,MAAM,EAAE,OAAOgC,EAAEtB,EAAE,QAAQ,IAAIT,EAAE,MAAM,EAAE,OAAO,EAAED,EAAE,QAAQ,SAASs4C,EAAE,uBAAuBx3C,CAAC,EAAEA,EAAEuE,EAAErF,EAAE,QAAQ,SAASs4C,EAAE,uBAAuBt2C,CAAC,EAAEA,EAAE,CAAC,EAAES,CAAC,EAAEpE,EAAE2B,EAAE,MAAMC,EAAE,MAAM,EAAEoF,EAAE1E,CAAC,EAAE+B,EAAEhC,EAAE,eAAe+B,EAAE9B,CAAC,EAAEgC,EAAEjC,EAAE,QAAQ,IAAIgC,EAAE,MAAM,EAAE,OAAOC,EAAE,OAAO,EAAED,CAAC,CAAC,IAAI9B,EAAE2X,EAAE,EAAE,QAAQ,8BAA8B,GAAG,GAAG,KAAK1X,EAAE,OAAOD,EAAEC,EAAE,IAAI86F,GAAG,EAAE37F,EAAE,MAAMC,EAAE,MAAMjC,CAAC,EAAE6C,EAAE,IAAI46F,GAAG39F,EAAEkC,EAAE,MAAMC,EAAE,KAAK,EAAES,EAAE,gBAAgBG,EAAE,CAACb,EAAEC,CAAC,EAAEU,CAAC,CAAC,CAAC,CAAC,SAASg8F,GAAG7+F,EAAE,EAAE,GAAG,CAAC,GAAGA,IAAI,SAAS,OAAO,EAAEs8F,GAAGP,GAAG,GAAG/7F,IAAI,OAAO,OAAO,EAAEw8F,GAAGN,GAAG,GAAGl8F,IAAI,MAAM,OAAO,EAAEu8F,GAAGN,GAAG,GAAGj8F,IAAI,QAAQ,OAAO,EAAEy8F,GAAGN,GAAG,GAAGn8F,IAAI,QAAQ,OAAO,EAAEu+F,GAAGD,GAAG,GAAGt+F,IAAI,YAAY,OAAO,EAAEm+F,GAAGD,GAAG,GAAGl+F,IAAI,UAAU,OAAO,EAAE08F,GAAGL,GAAG,MAAM,IAAI,MAAM,cAAcr8F,mDAAmD,CAAC,CAAC,IAAI8+F,GAAG,KAAK,CAAC,YAAYh+F,EAAEZ,EAAE,EAAEK,EAAE,GAAGwB,EAAE,GAAGC,EAAE,GAAGC,EAAE,KAAKC,EAAE,GAAG,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,UAAU,SAAS,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAY,EAAE,KAAK,oBAAoB8zF,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIpzF,EAAErC,EAAEO,EAAE,GAAGA,EAAE,GAAG,EAAE,KAAK,KAAK8B,EAAE,CAAC,EAAEE,EAAEvC,EAAE,cAAc,cAAcwC,EAAEhB,EAAE,cAAc,cAAciB,EAAEzC,EAAE,CAAC,SAAS,QAAQ,EAAE,CAAC,SAAS,QAAQ,EAAE2D,EAAEnC,EAAE,CAAC,SAAS,QAAQ,EAAE,CAAC,SAAS,QAAQ,EAAEoC,EAAE,GAAGoD,EAAE,GAAGtF,IAAIC,EAAEiC,EAAE;AAAA;AAAA,YAE5kFlC;AAAA,WACD,EAAEkC,EAAE;AAAA;AAAA,YAEHlC;AAAA,WACDkC,EAAE;AAAA,YACDlC;AAAA,WACDsF,EAAE,gCAAgC,IAAIxC,EAAE/C,EAAE,kCAAkC,GAAGA,GAAG,KAAK,cAAc,KAAK,MAAM,EAAEE,GAAG,KAAK,cAAc,KAAK,wBAAwB,EAAE,GAAG,KAAK,cAAc,KAAK,gBAAgB,EAAE,IAAIyC,EAAE,OAAOC,EAAE,OAAO9D,EAAE,GAAGZ,EAAE,GAAGyE,EAAE,wBAAwB7D,EAAE,GAAG,OAAOZ,EAAE,GAAGY,EAAE,KAAK8D,EAAE,wBAAwB1E,EAAE,GAAG,QAAQ,KAAK,SAAS;AAAA,QAC9ViE;AAAA;AAAA,sCAE8B;AAAA;AAAA;AAAA;AAAA,8BAIR;AAAA,yBACLQ;AAAA,yBACAC;AAAA,wCACe9B;AAAA,wCACAC;AAAA;AAAA;AAAA;AAAA,uBAIjBC,EAAE,QAAQkB,EAAE;AAAA,uBACZlB,EAAE,QAAQkB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UASzBa;AAAA;AAAA,UAEAwC;AAAA;AAAA;AAAA;AAAA,KAIL,CAAC,EAAMw3F,GAAG,CAAC,KAAK,wCAAwC,KAAK,uCAAuC,EAAEC,GAAG,KAAK,CAAC,YAAYl+F,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,QAAQ,QAAQ,QAAQ,OAAO,EAAE,KAAK,YAAYs6C,EAAE,2BAA2Bt6C,EAAE,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA,UAGnPY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAUL,CAAC,EAAMm+F,GAAG,gBAAgB,SAASC,GAAGl/F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAACM,EAAI,EAAEC,CAAC,EAAE,EAAE,EAAEi6C,EAAE,WAAWl6C,EAAE,MAAMC,EAAE,KAAK,EAAE,GAAGD,EAAE,QAAQ,YAAY,CAAC,IAAI2B,EAAE/B,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE4B,EAAEhC,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAE4B,EAAE,IAAI68F,GAAGD,GAAG,KAAKz+F,EAAE,MAAMC,EAAE,KAAK,EAAEqC,EAAE,IAAIo8F,GAAGD,GAAG,KAAKz+F,EAAE,MAAMC,EAAE,KAAK,EAAEsC,EAAE,CAAC,CAAC,OAAOZ,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM3B,EAAE,KAAK,EAAE,CAAC,OAAO2B,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM3B,EAAE,KAAK,EAAE,CAAC,OAAO4B,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM3B,EAAE,KAAK,EAAE,CAAC,OAAO2B,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM3B,EAAE,KAAK,CAAC,EAAEuC,EAAE5C,EAAE,gBAAgBiC,EAAEU,EAAE,SAAS,EAAEE,EAAE7C,EAAE,gBAAgB0C,EAAEC,EAAE,SAAS,EAAEG,EAAEg7F,GAAG,CAAC,OAAO,CAAC,KAAKl7F,EAAE,KAAKC,CAAC,EAAE,QAAQ7C,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAEC,CAAC,CAAC,GAAG9C,EAAE,mBAAmB,CAACI,EAAEC,CAAC,CAAC,EAAE,CAAC,IAAI0B,EAAE/B,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE4B,EAAEhC,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAE,CAAC4B,EAAES,CAAC,EAAE02F,GAAGh5F,EAAE,MAAMC,EAAE,MAAM0B,EAAE,OAAOC,EAAE,OAAO,CAAC,EAAEW,EAAE3C,EAAE,eAAe0C,EAAE,CAAC,EAAEE,EAAE5C,EAAE,QAAQ,IAAI2C,EAAE,MAAM,EAAE,OAAOC,EAAE,OAAOX,EAAEU,CAAC,CAAC,IAAI,EAAE,OAAO4X,EAAE,EAAE,QAAQ,8BAA8B,EAAE,EAAE,IAAIojF,GAAGoB,GAAG3+F,EAAE,MAAMC,EAAE,KAAK,EAAE,EAAE,IAAIo9F,GAAGsB,GAAG3+F,EAAE,MAAMC,EAAE,KAAK,EAAEL,EAAE,gBAAgB,EAAE,CAACI,EAAEC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI4+F,GAAG,CAAC,WAAW59E,GAAG,YAAY,QAAQ,WAAW29E,EAAE,EAAE,SAASE,GAAGp/F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,CAACwuF,GAAG9uF,EAAE,KAAK,EAAE,GAAGqvF,GAAGrvF,EAAE,KAAK,CAAC,EAAEO,EAAE,CAAC,MAAMP,EAAE,MAAM,MAAMM,EAAE,OAAON,EAAE,MAAM,EAAE,EAAE,CAAC8uF,GAAG,CAAC,EAAE,GAAGO,GAAG,CAAC,CAAC,EAAE,EAAE,IAAI+L,GAAG,EAAE96F,CAAC,EAAE2B,EAAE,GAAGC,EAAE,CAAC5B,CAAC,EAAE6B,EAAEjC,EAAE,gBAAgB,EAAE,CAACK,CAAC,EAAEP,EAAE,MAAMkC,EAAED,CAAC,EAAE,MAAM,CAAC,OAAOE,EAAE,OAAO,MAAM,EAAE,MAAMA,EAAE,KAAK,CAAC,CAAC,SAASk9F,GAAGr/F,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,EAAEJ,EAAE+B,EAAE6kB,EAAE,cAAcvmB,EAAE,KAAK,EAAE2B,EAAE4kB,EAAE,uBAAuB,EAAE7kB,CAAC,EAAEE,EAAE2kB,EAAE,cAAc5kB,CAAC,EAAE4kB,EAAE,OAAO7kB,IAAIE,EAAE,IAAI,kBAAkBD,UAAUC,iCAAiC5B,EAAE,cAAc0B,gFAAgF,EAAE,IAAIW,EAAE,EAAE,QAAQ,IAAIrC,EAAE,MAAM,EAAE,OAAOqC,EAAE,UAAU,CAACktF,GAAGvvF,EAAE,MAAM2B,CAAC,GAAG,EAAEU,EAAE,UAAU,MAAMktF,GAAGltF,EAAE,MAAMV,CAAC,GAAGk9F,GAAG7+F,EAAE2B,EAAE,CAAC,GAAG,EAAE,OAAO3B,EAAE,MAAM,EAAE,CAAC,OAAOA,EAAE,OAAO,MAAM2B,EAAE,MAAM3B,EAAE,KAAK,EAAE,CAAC,IAAI++F,GAAG,CAAC,WAAW18E,GAAG,YAAY,QAAQ,WAAWy8E,EAAE,EAAME,GAAG,KAAK,CAAC,YAAYz+F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,GAAG,CAAC,WAAW,EAAE,UAAUK,EAAE,OAAOwB,EAAE,QAAQC,CAAC,EAAElB,EAAE,KAAK,YAAY,CAACP,EAAEyB,CAAC,EAAE,IAAIC,EAAE,KAAK,MAAM,EAAE,CAAC,EAAE,EAAEC,EAAE,EAAE,EAAE,EAAE,iCAAiC,GAAGhC,GAAG,KAAK,CAAC,IAAI,EAAE,EAAEA,EAAE,EAAE,4BAA4B4mB,EAAE,MAAM,CAAC,EAAE,EAAE,YAAY,CAAC,EAAE,WAAW,CAAC,IAAIlkB,EAAE,GAAGb,EAAE,EAAE,IAAIa,EAAE;AAAA,oCACtvEb;AAAA;AAAA;AAAA,SAG3B,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA,UAIba;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAQwB;AAAA;AAAA;AAAA;AAAA,8BAIJX;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YASlB;AAAA;AAAA;AAAA,iCAGqBA;AAAA,cACnBC,IAAI;AAAA;AAAA;AAAA,YAGN;AAAA,qBACSA,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA,YAKb;AAAA,qBACSA,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAMb;AAAA;AAAA;AAAA;AAAA,KAIP,CAAC,EAAMs9F,GAAG,KAAK,CAAC,YAAY1+F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,GAAG,CAAC,WAAW,EAAE,UAAUK,EAAE,OAAOwB,EAAE,QAAQC,CAAC,EAAElB,EAAE,KAAK,YAAY,CAACP,EAAEyB,CAAC,EAAE,IAAIC,EAAE,MAAMC,EAAE,GAAGhC,IAAI,OAAO+B,EAAE,MAAM/B,IAAI,OAAO+B,EAAE,cAAcC,EAAE,OAAOhC,IAAI,QAAQ+B,EAAE,eAAeC,EAAE,OAAO,IAAI,EAAE,GAAGhC,KAAKA,KAAKA,sEAAsEA,IAAI,MAAM,EAAE,WAAWA,IAAI,OAAO,EAAE,YAAYA,IAAI,MAAM,EAAE,WAAWA,IAAI,QAAQ,EAAE,YAAY,IAAI0C,EAAE,KAAK,MAAM,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAEE,EAAE;AAAA,YACrc5C,IAAI;AAAA;AAAA,mBAEGA,IAAI;AAAA;AAAA;AAAA;AAAA,wBAICgC;AAAA,cACVhC,IAAI,YAAYA,IAAI;AAAA,0BACRgC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAOpBa,EAAE,OAAO7C,IAAI,OAAO+B,EAAE,MAAMa,EAAE;AAAA;AAAA;AAAA;AAAA,QAI5BC,EAAE,SAAS7C,IAAI,QAAQ+B,EAAE,MAAMa,EAAE;AAAA;AAAA;AAAA;AAAA,QAIjCC,EAAE,SAAS,IAAIC,EAAE,GAAGjB,EAAE,EAAE,IAAIiB,EAAE;AAAA,oCACFjB;AAAA;AAAA;AAAA,SAG3B,KAAK,SAAS;AAAA,0CACmBE;AAAA;AAAA;AAAA;AAAA,UAIhCe;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAQwB;AAAA;AAAA,kCAEAf;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMJW;AAAA;AAAA,YAElBG,cAAcA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOdD;AAAA;AAAA;AAAA,iCAGqBF;AAAA,cACnB,IAAI;AAAA,YACNG,cAAcA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOdD;AAAA,qBACS,IAAI;AAAA,YACbC,cAAcA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOdD;AAAA,qBACS,IAAI;AAAA,YACbC,cAAcA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOdD;AAAA;AAAA,oBAEQ;AAAA;AAAA,KAEf,CAAC,EAAE,SAAS28F,GAAIz/F,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,KAAK,EAAE,SAAS,GAAG,EAAE,EAAE,OAAO,GAAG,UAAU,GAAG,CAAC,IAAIE,EAAE,EAAE,OAAO,EAAE,EAAE,OAAO,GAAG,QAAQF,EAAE,GAAGM,EAAEk6C,EAAE,yBAAyBt6C,CAAC,EAAE,EAAE,KAAK,CAAC,OAAOA,EAAE,WAAWI,EAAE,QAAQ,KAAK,KAAKJ,EAAEI,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,SAASo/F,GAAG1/F,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEk/F,GAAIz/F,EAAE,KAAK,EAAE,EAAEA,EAAE,QAAQ,EAAE,EAAE,EAAEO,EAAE,OAAO,IAAI,CAAC,GAAG,CAAC,OAAO0B,EAAE,WAAWC,EAAE,QAAQC,CAAC,EAAE5B,EAAE,GAAGqC,EAAEC,EAAE3C,IAAI,OAAO0C,EAAE,IAAI,EAAE,IAAI28F,GAAG,CAAC,WAAWr9F,EAAE,OAAOD,EAAE,UAAUjC,EAAE,MAAM,GAAG,QAAQmC,CAAC,EAAEF,CAAC,EAAE,IAAIs9F,GAAG,CAAC,WAAWr9F,EAAE,OAAOD,EAAE,UAAUjC,EAAE,MAAM,GAAG,QAAQmC,CAAC,CAAC,EAAES,EAAE,IAAI48F,GAAG,CAAC,WAAWt9F,EAAE,OAAOD,EAAE,UAAUjC,EAAE,MAAM,GAAG,QAAQmC,CAAC,EAAEjC,CAAC,EAAE2C,EAAE,EAAE,EAAEvC,EAAE,gBAAgBsC,EAAE,CAAC,CAAC,EAAE,CAAC,EAAEC,EAAE,SAAS7C,EAAE,QAAQM,EAAE,8BAA8BuC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI88F,GAAG,KAAK,CAAC,YAAY7+F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,IAAI,EAAE,IAAI,MAAMY,EAAE,MAAM,EAAE,QAAQkB,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,GAAGlB,EAAEZ,EAAE8B,IAAI,KAAK,YAAY,EAAE,KAAK,KAAK,EAAE,OAAO,IAAIzB,EAAEm1F,GAAG,KAAK,IAAI,EAAE3zF,EAAE69F,GAAI1/F,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,QAE10BK;AAAA,uBACewB;AAAA;AAAA,KAElB,CAAC,EAAE,SAAS69F,GAAI5/F,EAAE,CAAC,IAAI,EAAEA,EAAE,OAAO,GAAG,EAAE,EAAE,MAAM,MAAM,sBAAsB,wBAAwB,EAAE,IAAIE,EAAE,CAAC,UAAU,UAAU,UAAU,UAAU,UAAU,SAAS,EAAEI,EAAE,IAAI,MAAM,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,IAAID,EAAEN,EAAEO,IAAIL,EAAEK,GAAG,OAAOD,EAAE,KAAK,CAAC,CAAC,IAAIu/F,GAAG,KAAK,CAAC,YAAY/+F,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,IAAI,EAAE,IAAI,MAAMY,EAAE,MAAM,EAAE,QAAQ8B,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,GAAG9B,EAAEZ,EAAE0C,IAAI,GAAG,KAAK,YAAY,EAAE,KAAK,KAAK,EAAE,OAAO,KAAK,KAAK,EAAE,MAAM,MAAM,6BAA6B,KAAK,4BAA4B,EAAE,IAAIrC,EAAEm1F,GAAG,KAAK,IAAI,EAAE3zF,EAAEi5F,GAAG,KAAK,KAAK,IAAI,EAAEh5F,EAAE,IAAI,MAAM,KAAK,IAAI,EAAE,QAAQY,EAAE,EAAEA,EAAE1C,EAAE,OAAO0C,IAAIZ,EAAE9B,EAAE0C,IAAIb,EAAEa,GAAG,IAAIX,EAAE,QAAQD,EAAE,MAAM,EAAE,EAAE,KAAK,KAAKE,EAAE,KAAKH,EAAE,KAAK,KAAK,QAAQ,EAAE,KAAK,KAAK,KAAK,EAAE,mBAAmBC,EAAE,KAAK,OAAOC,KAAK,KAAK,SAAS;AAAA;AAAA,QAE/vB1B;AAAA;AAAA,oBAEY;AAAA,WACT2B;AAAA,sBACW;AAAA;AAAA,UAEZH,EAAE,KAAK,KAAK;AAAA,aACTA,EAAE,KAAK,KAAK,QAAQ,EAAE,KAAK,KAAK;AAAA,sBACvB;AAAA,aACTG;AAAA,wBACW;AAAA;AAAA;AAAA;AAAA;AAAA,KAKnB,CAAC,EAAE,SAAS49F,GAAG9/F,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEma,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAIolF,GAAG7/F,EAAE,MAAM,CAAC,EAAE,IAAI2/F,GAAG3/F,EAAE,MAAM,CAAC,EAAE,OAAOE,EAAE,gBAAgBI,EAAE,CAACN,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,SAAS+/F,GAAG//F,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,EAAE,EAAEP,EAAE,MAAM,OAAO,EAAE8mB,EAAE,eAAevmB,EAAEP,EAAE,KAAK,EAAEiC,EAAE,EAAEC,EAAEs4C,EAAE,mBAAmBv4C,EAAE,CAAC,EAAEE,EAAED,GAAG,KAAKU,EAAE5C,EAAEmC,IAAIS,EAAEk9F,GAAG9/F,EAAEkC,EAAE5B,CAAC,EAAE2B,EAAEu4C,EAAE,iBAAiBv4C,EAAE,OAAO,CAAC,GAAGu4C,EAAE,2BAA2B,MAAMv4C,EAAE,CAAC,EAAE,GAAG,CAACY,EAAEC,CAAC,EAAE03C,EAAE,0BAA0B53C,EAAE,MAAMX,CAAC,EAAEc,EAAEF,EAAE3C,IAAI6C,EAAEy3C,EAAE,qBAAqB33C,EAAE,CAAC,GAAG,IAAIG,EAAE8jB,EAAE,cAAchkB,CAAC,EAAEqB,EAAE2iB,EAAE,cAAc9mB,EAAE,KAAK,EAAEgD,EAAEuE,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAEz8F,CAAC,EAAE,MAAM,CAAC,MAAM,CAACuB,EAAEnB,CAAC,CAAC,EAAE,QAAQ1C,CAAC,CAAC,EAAEyE,EAAE6lB,GAAG5qB,EAAE,KAAK,EAAE2E,EAAE+6F,GAAGn4F,EAAExC,EAAE,MAAMzE,CAAC,EAAEsE,EAAEy6F,GAAG,CAAC,OAAO,CAAC,EAAE16F,CAAC,EAAE,MAAM,CAAC,MAAM5B,CAAC,EAAE,QAAQzC,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BiH,CAAC,EAAEjH,EAAE,8BAA8BqE,CAAC,EAAExC,GAAG7B,EAAE,8BAA8BsC,CAAC,EAAEgC,CAAC,CAAC,SAASo7F,GAAGhgG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE,OAAOy/F,GAAGx/F,EAAE,EAAE,EAAEL,CAAC,CAAC,CAAC,IAAI+/F,GAAG,CAAC,WAAWj8E,GAAG,YAAY,QAAQ,WAAWg8E,EAAE,EAAE,SAASE,GAAGlgG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAEJ,EAAE+B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE,IAAI,MAAMD,CAAC,EAAE,QAAQW,EAAE,EAAEA,EAAEV,EAAE,OAAOU,IAAIV,EAAEU,GAAGrC,EAAE,MAAM,EAAEqC,IAAI,IAAIT,EAAE,GAAG,EAAE,mBAAmB,CAAC5B,CAAC,CAAC,EAAE,CAAC,IAAI,EAAE,EAAE,QAAQ,IAAIA,EAAE,MAAM,EAAE,OAAOuC,EAAEg4F,GAAG,EAAEv6F,EAAE,MAAMA,EAAE,MAAM,EAAE2B,CAAC,EAAEC,EAAE,EAAE,eAAeD,EAAE3B,EAAE,KAAK,EAAE,IAAIwC,EAAE,EAAE,QAAQ,IAAIZ,EAAE,MAAM,EAAEY,EAAE,OAAOD,CAAC,MAAMX,EAAE29F,GAAGv/F,EAAE,EAAE,CAAC,EAAE,OAAO4B,CAAC,CAAC,IAAIg+F,GAAG,CAAC,WAAW96E,GAAG,YAAY,QAAQ,WAAW66E,EAAE,EAAME,GAAG,IAAI,SAASC,GAAG,CAAC,EAAErgG,EAAE,EAAE,EAAE,WAAWE,EAAE,WAAWI,EAAE,QAAQC,EAAE,KAAK,EAAE,KAAK,uBAAuB,EAAE,KAAK,eAAe0B,EAAE,EAAE,WAAWC,EAAE,IAAI,EAAE,CAAC,IAAIC,EAAEnC,EAAE,MAAM,OAAO4C,EAAE,EAAE,MAAM,OAAOC,EAAE3C,EAAEF,EAAE,MAAMmC,EAAE,GAAGnC,EAAE,MAAMmC,EAAE,GAAGW,EAAExC,EAAE,EAAE,MAAMsC,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGG,EAAE7C,EAAEF,EAAE,MAAMmC,EAAE,GAAGnC,EAAE,MAAMmC,EAAE,GAAGa,EAAE1C,EAAE,EAAE,MAAMsC,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGsB,EAAElE,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,EAAEuH,EAAEuf,EAAE,cAAc5iB,CAAC,EAAE,EAAE4iB,EAAE,cAAc,CAAC,EAAEliB,EAAEmwB,GAAG,2BAA2B/0B,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,CAAC,EAAE,OAAO,CAAC+C,EAAEC,CAAC,CAAC,EAAE8jB,EAAE,OAAOjkB,IAAIC,EAAE,IAAI,kCAAkCD,WAAWC,6BAA6B9C,EAAE,aAAa,EAAE,wBAAwBE,oBAAoBI,eAAe,EAAE,IAAIuE,EAAE3E,EAAE,CAACqH,EAAE1E,EAAEE,CAAC,EAAE,CAACwE,EAAExE,EAAEF,CAAC,EAAEiC,EAAExE,EAAE,CAAC,EAAE0C,EAAEF,CAAC,EAAE,CAAC,EAAEA,EAAEE,CAAC,EAAEwE,EAAE63F,GAAG,CAAC,OAAO,CAAC,EAAEr/F,CAAC,EAAE,QAAQO,EAAE,MAAM,CAAC,MAAMsE,CAAC,CAAC,CAAC,EAAEG,EAAEq6F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ9+F,EAAE,MAAM,CAAC,MAAMuE,CAAC,CAAC,CAAC,EAAE2C,EAAE,CAACD,EAAExC,CAAC,EAAEC,EAAE,KAAK,IAAIsC,EAAE,CAAC,EAAErC,EAAEhF,EAAEsH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGrC,EAAE,GAAG,KAAKC,EAAE,GAAG,KAAKC,EAAEnD,IAAI,YAAYoD,EAAEpD,GAAG,KAAK28F,GAAG38F,EAAE,EAAE,EAAE,KAAK+F,EAAE9C,GAAGC,GAAGC,GAAGC,GAAG,KAAKqC,EAAE,IAAI5E,IAAI,GAAGC,IAAI,IAAIkC,EAAEk7F,IAAIn4F,IAAI,GAAG,CAAC,IAAIE,EAAGX,EAAEY,EAAGpD,EAAE9E,IAAIiI,EAAG+3F,GAAG,CAAC,OAAO,CAAC,EAAE14F,CAAC,EAAE,QAAQjH,EAAE,MAAM,CAAC,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEkH,EAAE,KAAKU,CAAE,GAAG7H,IAAI8H,EAAG83F,GAAG,CAAC,OAAO,CAAC,EAAEl7F,CAAC,EAAE,QAAQzE,EAAE,MAAM,CAAC,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEkH,EAAE,KAAKW,CAAE,GAAG,IAAIC,EAAGrF,IAAI,EAAEsF,GAAGtF,IAAI,EAAEuF,GAAGJ,EAAGE,IAAKE,GAAG82F,GAAG,CAAC,OAAO,CAAC,EAAEl3F,CAAE,EAAE,QAAQ5H,EAAE,MAAM,CAAC,MAAM,CAAC0E,EAAEC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEuC,EAAE,KAAKc,EAAE,GAAG,IAAI5B,GAAG3D,IAAI,EAAE,EAAE,EAAEyF,GAAGL,EAAGE,KAAKG,GAAG42F,GAAG,CAAC,OAAO,CAAC,EAAEj3F,CAAE,EAAE,QAAQ7H,EAAE,MAAM,CAAC,MAAM,CAAC0E,EAAE,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAEuC,EAAE,KAAKgB,EAAE,GAAG,IAAIC,GAAGw2F,GAAG,CAAC,OAAO,CAAC,EAAE32F,GAAG,EAAEE,EAAE,EAAE,QAAQlI,CAAC,CAAC,EAAEoH,EAAEq4F,GAAG,CAAC,OAAO,CAAC,EAAEt3F,EAAE,EAAE,QAAQnI,EAAE,MAAM,CAAC,KAAKoG,GAAG,SAAS,EAAE,CAAC,CAAC,EAAEc,EAAE,KAAKiB,EAAE,CAAC,KAAK,CAAC,IAAIP,EAAGwiB,GAAG3qB,EAAE,MAAM,EAAE,KAAK,EAAEoI,EAAG,IAAI02F,GAAGj6F,EAAEC,EAAE,CAACG,EAAElC,EAAEC,CAAC,EAAE9C,EAAEI,EAAE6E,EAAEG,EAAEF,EAAEC,CAAC,EAAEgD,EAAG,CAACb,EAAExC,CAAC,EAAE,GAAG,GAAG,MAAMqD,EAAG,KAAK,CAAC,EAAEjD,GAAGiD,EAAG,KAAK,CAAC,EAAEhD,EAAE,CAAC,IAAIiD,GAAG/H,EAAE,eAAe,CAAC,EAAE,UAAUumB,EAAE,kBAAkB7kB,EAAE,SAAS,CAAC,EAAEoG,EAAG,KAAKC,EAAE,EAAEb,EAAE,KAAKa,EAAE,CAAC,CAACX,EAAEpH,EAAE,gBAAgB6H,EAAGC,EAAGF,CAAE,CAAC,CAAC,IAAID,EAAEm3F,GAAG,CAAC,OAAO,CAAC,EAAE13F,CAAC,EAAE,QAAQpH,EAAE,MAAM,CAAC,MAAMqE,CAAC,CAAC,CAAC,EAAE6C,EAAE,KAAKE,CAAC,EAAE,QAAQQ,KAAMV,EAAElH,EAAE,8BAA8B4H,CAAE,EAAE,OAAOD,CAAC,CAAC,SAASo4F,GAAItgG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,WAAWC,EAAE,WAAWC,EAAE,WAAWS,EAAE,eAAeC,CAAC,EAAEvC,EAAE,OAAO+/F,GAAG,CAAC,EAAE9/F,EAAE,EAAE,EAAE,WAAW2B,EAAE,WAAWC,EAAE,QAAQjC,EAAE,KAAK,EAAE,uBAAuB+B,EAAE,eAAeY,EAAE,WAAWD,CAAC,CAAC,CAAC,CAAC,IAAI29F,GAAG,CAAC,WAAWz6E,GAAG,YAAY,QAAQ,WAAWw6E,EAAG,EAAME,GAAG,iBAAiB,SAASC,GAAIzgG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGJ,EAAE,mBAAmB,CAACI,CAAC,CAAC,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAI,EAAEJ,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE,EAAE25F,GAAG,EAAE,MAAM,EAAE,OAAO/5F,EAAE,eAAeI,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIC,EAAE,OAAOka,EAAE,EAAE,QAAQ,6BAA6B,EAAEla,EAAE,IAAIo8F,GAAGr8F,EAAE,MAAMkgG,EAAE,EAAEjgG,EAAE,IAAIs7F,GAAGv7F,EAAE,MAAMkgG,EAAE,EAAEtgG,EAAE,gBAAgBK,EAAE,CAACD,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAIogG,GAAG,CAAC,WAAWxlF,GAAG,YAAY,QAAQ,WAAWulF,EAAG,EAAME,GAAI7E,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA,EAKzvH8E,GAAIjC,GAAG,CAAC,UAAUgC,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAW1lF,GAAG,YAAY,QAAQ,WAAWylF,EAAG,EAAME,GAAIhF,GAAG;AAAA;AAAA,oCAEvDiF,GAAIpC,GAAG,CAAC,UAAUmC,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAW5lF,GAAG,YAAY,QAAQ,WAAW2lF,EAAG,EAAME,GAAG,gBAAgBC,GAAItC,GAAG,CAAC,UAAUqC,GAAG,gBAAgBA,GAAG,gBAAgB,GAAG,cAAchJ,EAAE,CAAC,EAAEkJ,GAAG,CAAC,WAAW9lF,GAAG,YAAY,QAAQ,WAAW6lF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYtgG,EAAEZ,EAAE,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYY,EAAE,KAAK,cAAcZ,EAAE,IAAI,CAAC6B,EAAEC,IAAI,IAAIA,GAAG,EAAE,IAAI,EAAE,CAAC,EAAE,KAAK,cAAc,QAAQD,GAAG,CAAC,EAAE,KAAK,UAAUA,UAAUA,iBAAiB,CAAC,CAAC,EAAE,IAAIxB,EAAE,KAAK,cAAc,IAAIwB,GAAG,IAAIA,GAAG,EAAE,KAAK,KAAK,EAAE,KAAK,SAAS;AAAA;AAAA,UAE5gB,EAAE,KAAK;AAAA,SACR;AAAA;AAAA,yBAEgBxB;AAAA;AAAA;AAAA,KAGpB,CAAC,EAAM8gG,GAAG,KAAK,CAAC,YAAYvgG,EAAEZ,EAAE,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYY,EAAE,KAAK,cAAcZ,EAAE,IAAI,CAAC6B,EAAEC,IAAI,IAAIA,GAAG,EAAE,IAAI,EAAE,CAAC,EAAE,KAAK,cAAc,QAAQD,GAAG,CAAC,EAAE,KAAK,SAASA,UAAUA,iBAAiB,CAAC,CAAC,EAAE,IAAIxB,EAAE,KAAK,cAAc,IAAIwB,GAAG,IAAIA,GAAG,EAAE,KAAK,KAAK,EAAE,KAAK,SAAS;AAAA;AAAA,UAEhT,EAAE,KAAK;AAAA,SACR;AAAA;AAAA,wBAEexB;AAAA;AAAA;AAAA,KAGnB,CAAC,EAAE,SAAS+gG,GAAGthG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAE,EAAE,GAAGA,EAAE,SAAS,EAAE,OAAOw9F,GAAG,CAAC,OAAO,CAAC,EAAEx9F,EAAE,EAAE,EAAE,QAAQJ,CAAC,CAAC,EAAE,GAAGI,EAAE,OAAOma,EAAE,EAAE,IAAI,8BAA8B,EAAE,CAAC,IAAIvY,EAAE,KAAK,MAAM5B,EAAE,OAAO,CAAC,EAAE,EAAEghG,GAAG,CAAC,OAAOhhG,EAAE,MAAM,EAAE4B,CAAC,EAAE,QAAQhC,CAAC,CAAC,EAAE0C,EAAE0+F,GAAG,CAAC,OAAOhhG,EAAE,MAAM4B,CAAC,EAAE,QAAQhC,CAAC,CAAC,EAAE,OAAOohG,GAAG,CAAC,OAAO,CAAC,EAAE1+F,CAAC,EAAE,QAAQ1C,CAAC,CAAC,CAAC,CAAC,IAAIK,EAAED,EAAE,IAAI4B,GAAGA,EAAE,KAAK,EAAE,OAAO,CAACA,EAAE,IAAIyoB,GAAGzoB,EAAE,CAAC,CAAC,EAAE,EAAE5B,EAAE,IAAI4B,GAAGA,EAAE,KAAK,EAAED,EAAEwY,EAAE,EAAE,QAAQ,YAAY,EAAE,IAAI4mF,GAAG/gG,EAAE,GAAG,MAAM,CAAC,EAAE,IAAI8gG,GAAG9gG,EAAE,GAAG,MAAM,CAAC,EAAE,OAAOJ,EAAE,gBAAgB+B,EAAE3B,EAAEC,CAAC,CAAC,CAAC,IAAIghG,GAAG,CAAC,WAAWjmF,GAAG,YAAY,QAAQ,WAAWgmF,EAAE,EAAE,SAASE,GAAIxhG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE4B,EAAED,EAAEU,EAAE43C,EAAE,mBAAmBr4C,EAAEF,CAAC,EAAEY,EAAEtC,EAAEqC,GAAG,OAAOC,EAAEq9F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK0C,CAAC,CAAC,CAAC,EAAET,EAAEq4C,EAAE,iBAAiBr4C,EAAE,OAAOF,CAAC,GAAGu4C,EAAE,2BAA2B,MAAMr4C,EAAEF,CAAC,EAAE,GAAG,CAACa,EAAEC,CAAC,EAAEy3C,EAAE,0BAA0B33C,EAAE,MAAMV,CAAC,EAAEa,EAAE8jB,EAAE,cAAc/jB,CAAC,EAAEmB,EAAEm7F,GAAG,CAAC,OAAO,CAAC,EAAEx8F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG8C,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE08F,GAAGx7F,EAAEA,EAAE,MAAM,MAAMhE,CAAC,EAAEqH,EAAE,GAAG,EAAE,CAAC,IAAI,EAAEizC,EAAE,qBAAqB13C,EAAEZ,CAAC,EAAEqF,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,MAAMqH,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8BgE,CAAC,EAAEhE,EAAE,8BAA8B,CAAC,EAAE0C,GAAG,MAAM1C,EAAE,8BAA8B2C,CAAC,EAAE0E,CAAC,CAAC,IAAIk6F,GAAG,CAAC,WAAWlmF,GAAG,YAAY,QAAQ,WAAWimF,EAAG,EAAE,SAASE,GAAI1hG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE4B,EAAED,EAAEU,EAAE43C,EAAE,mBAAmBr4C,EAAEF,CAAC,EAAEY,EAAEtC,EAAEqC,GAAG,OAAOC,EAAEq9F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK0C,CAAC,CAAC,CAAC,EAAET,EAAEq4C,EAAE,iBAAiBr4C,EAAE,OAAOF,CAAC,GAAGu4C,EAAE,2BAA2B,MAAMr4C,EAAEF,CAAC,EAAE,GAAG,CAACa,EAAEC,CAAC,EAAEy3C,EAAE,0BAA0B33C,EAAE,MAAMV,CAAC,EAAEa,EAAE8jB,EAAE,cAAc/jB,CAAC,EAAEmB,EAAEm7F,GAAG,CAAC,OAAO,CAAC,EAAEx8F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG8C,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE08F,GAAGx7F,EAAEA,EAAE,MAAM,MAAMhE,CAAC,EAAEqH,EAAE,GAAG,EAAE,CAAC,IAAI,EAAEizC,EAAE,qBAAqB13C,EAAEZ,CAAC,EAAEqF,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,MAAMqH,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8BgE,CAAC,EAAEhE,EAAE,8BAA8B,CAAC,EAAE0C,GAAG,MAAM1C,EAAE,8BAA8B2C,CAAC,EAAE0E,CAAC,CAAC,IAAIo6F,GAAG,CAAC,WAAWnmF,GAAG,YAAY,QAAQ,WAAWkmF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY9gG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,GAAG,CAAC,WAAWK,EAAE,UAAUwB,EAAE,QAAQC,CAAC,EAAElB,EAAE,GAAG,KAAK,cAAc,KAAK,cAAc,EAAE,KAAK,YAAY,CAACiB,EAAEC,CAAC,EAAE,IAAIC,EAAE/B,IAAI,MAAM,IAAI,IAAIgC,EAAE,EAAE,gBAAgB,+CAA+C,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,kCAK1uE3B;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKJA;AAAA,wBACN2B;AAAA;AAAA,0BAEED;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAOrB,CAAC,EAAM4/F,GAAG,KAAK,CAAC,YAAY/gG,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAGumB,EAAE,OAAOhmB,EAAE,OAAO,EAAE,IAAI,aAAa,EAAE,OAAO,CAAC,EAAE,YAAY,EAAE,EAAE,MAAM,CAAC,2CAA2C,EAAE,IAAIiB,EAAEjB,EAAEA,EAAE,OAAO,GAAGkB,EAAE,KAAK,KAAKD,EAAE7B,CAAC,EAAE,KAAK,YAAYY,EAAE,MAAM,EAAE,EAAE,EAAEkB,EAAE,GAAG,KAAK,YAAY,KAAKA,CAAC,EAAEzB,GAAG,KAAK,cAAc,KAAK,cAAc,EAAE,IAAI0B,EAAE,KAAK,YAAYC,EAAED,EAAE,OAAO,EAAEyzF,GAAGxzF,CAAC,EAAEU,EAAEq4F,GAAG,SAAS/4F,CAAC,EAAE,EAAEY,EAAE,GAAGd,IAAI,EAAE,CAACc,EAAEZ,EAAE,EAAE,IAAI8C,EAAE0wF,GAAG5yF,CAAC,EAAE,EAAE;AAAA,UAClckC,kBAAkBA,KAAKpC,EAAE,KAAK;AAAA,YAC5BA,EAAEV,EAAE;AAAA,UACN8C,kBAAkBA,KAAKpC,EAAE,KAAK;AAAA,YAC5BA,EAAEV,EAAE;AAAA,UACN8C,kBAAkBA,KAAKpC,EAAE,KAAK;AAAA,YAC5BA,EAAEV,EAAE;AAAA,UACN8C,kBAAkBA,KAAKpC,EAAE,KAAK;AAAA,YAC5BA,EAAEV,EAAE,KAAK,MAAMY,EAAEZ,EAAE,EAAE;AAAA,UACvB;AAAA,YACEU,EAAEV,EAAE;AAAA,UACN;AAAA,YACEU,EAAEV,EAAE;AAAA,UACN;AAAA,YACEU,EAAEV,EAAE;AAAA,UACN;AAAA,YACEU,EAAEV,EAAE,MAAM,IAAIa,EAAE,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAE,MAAM,EAAED,CAAC,EAAEE,EAAE,IAAID,EAAED,EAAE,GAAGoB,EAAEnB,EAAE,IAAIiC,GAAG,OAAOA,CAAC,EAAEb,EAAE82F,GAAG,aAAan4F,EAAE,CAAC,EAAE,OAAO,SAAS,EAAEyE,EAAE0zF,GAAG,aAAan4F,EAAE,CAAC,EAAE,OAAO,SAAS,EAAEiC,EAAEk2F,GAAG,aAAan4F,EAAE,CAAC,EAAE,OAAO,SAAS,EAAE6B,EAAEs2F,GAAG,aAAan4F,EAAE,CAAC,EAAE,OAAO,SAAS,EAAE8B,EAAE,IAAI,MAAM,cAAc,WAAWC,EAAEtE,EAAE,GAAG;AAAA,sDAC7P4D,EAAE,KAAK;AAAA,sDACPoD,EAAE,KAAK;AAAA,sDACPxC,EAAE,KAAK;AAAA,sDACPJ,EAAE,KAAK,QAAQG,EAAE;AAAA,0BAC7CX,EAAE,KAAK;AAAA,uCACMoD,EAAE,KAAK;AAAA,uCACPxC,EAAE,KAAK;AAAA,qDACOJ,EAAE,KAAK,WAAW6C,EAAEjH,EAAE,GAAG;AAAA,qCACzC2D,EAAE,KAAK;AAAA,4CACAnB,EAAE,KAAK;AAAA,iDACFA,EAAE,MAAM,EAAE,EAAE,KAAK;AAAA,SACzD,KAAK,SAAS;AAAA,0BACGmB,EAAE,KAAK;AAAA,iCACAnB,EAAE,KAAK;AAAA,sCACFA,EAAE,MAAM,EAAE,EAAE,KAAK;AAAA;AAAA,QAE/CyE;AAAA;AAAA,UAEE;AAAA,4BACkB5E,EAAEV,EAAE,QAAQD,EAAEC,EAAE,GAAG;AAAA,4BACnBU,EAAEV,EAAE,QAAQD,EAAEC,EAAE,GAAG;AAAA,UACrC;AAAA,yCAC+Bc,gBAAgBA;AAAA,sBACnCA,gBAAgBA,QAAQ9C;AAAA;AAAA;AAAA,2BAGnB4E;AAAA;AAAA,8BAEG5E;AAAA;AAAA,YAElB2E;AAAA,6BACiBC;AAAA;AAAA;AAAA,mBAGVF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAWd,CAAC,EAAE,SAASk9F,GAAG9hG,EAAE,EAAEE,EAAEI,EAAE,KAAK,CAAC,IAAIC,EAAE,EAAE,MAAM,GAAG,EAAE,EAAE,MAAM,GAAGD,GAAG,OAAOC,EAAED,EAAE,MAAM,GAAG,EAAEA,EAAE,MAAM,IAAI,IAAI,EAAEk6C,EAAE,yBAAyB,CAAC,EAAEv4C,EAAE,CAAC,WAAW,EAAE,OAAO,EAAE,UAAU1B,EAAE,QAAQ,KAAK,KAAK,EAAE,CAAC,CAAC,EAAE2B,EAAE,IAAI0/F,GAAG3/F,EAAE/B,EAAEI,GAAG,IAAI,EAAE6B,EAAE,CAAC,CAAC,EAAE7B,GAAG,MAAM6B,EAAE,KAAK7B,CAAC,EAAE,IAAIsC,EAAE5C,EAAE,gBAAgBkC,EAAEC,EAAE,OAAO,EAAE,GAAGS,EAAE,MAAM,KAAK,EAAE,OAAOA,EAAE,IAAIC,EAAEi/F,GAAG9hG,EAAE,EAAEE,EAAE0C,CAAC,EAAE,OAAO5C,EAAE,8BAA8B4C,CAAC,EAAEC,CAAC,CAAC,SAASk/F,GAAG/hG,EAAE,EAAEE,EAAEI,EAAE,KAAK,CAAC,IAAIC,EAAED,GAAG,KAAKA,EAAE,MAAM,EAAE,MAAM,EAAEC,EAAEA,EAAE,OAAO,GAAG,EAAEi6C,EAAE,yBAAyB,CAAC,EAAEv4C,EAAE,IAAI4/F,GAAGthG,EAAE,EAAEL,EAAEI,GAAG,IAAI,EAAE4B,EAAE5B,GAAG,KAAK,CAAC,CAAC,EAAE,CAAC,EAAEA,CAAC,EAAE6B,EAAEnC,EAAE,gBAAgBiC,EAAEC,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,OAAO,CAAC,IAAIS,EAAEm/F,GAAG/hG,EAAE,EAAEE,EAAEiC,CAAC,EAAE,OAAOnC,EAAE,8BAA8BmC,CAAC,EAAES,CAAC,CAAC,OAAOT,CAAC,CAAC,SAAS6/F,GAAGhiG,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,CAACL,CAAC,EAAE,GAAGs6C,EAAE,2BAA2B,MAAMl6C,EAAE,OAAO,CAAC,EAAE,YAAY,EAAEA,EAAE,MAAM,CAAC,EAAEC,EAAE,EAAE,MAAM,MAAM,EAAE,CAACka,EAAE,EAAE,QAAQ,mBAAmB,GAAG,EAAE,MAAM,QAAQ,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,EAAEza,EAAE,QAAQ,IAAI,EAAE,MAAM,EAAEiC,EAAE,IAAI,MAAM,EAAE,SAASC,EAAE,EAAED,IAAIC,EAAElC,EAAE,aAAa,CAAC,EAAE,EAAE,KAAKkC,CAAC,GAAG,GAAG,CAACC,EAAES,CAAC,EAAE43C,EAAE,0BAA0Bt4C,EAAE,MAAM3B,CAAC,EAAEsC,EAAEikB,EAAE,cAAclkB,CAAC,EAAEE,EAAEu8F,GAAG,CAAC,OAAO,CAAC,EAAEn9F,CAAC,EAAE,QAAQlC,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG6C,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,KAAKC,CAAC,EAAE,IAAIC,EAAE++F,GAAG9hG,EAAE8C,EAAExC,CAAC,EAAE,EAAE,KAAKyC,CAAC,EAAE,IAAIC,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAEt8F,CAAC,EAAE,QAAQ/C,EAAE,MAAM,CAAC,MAAMmC,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,QAAQ+B,GAAGlE,EAAE,8BAA8BkE,CAAC,CAAC,EAAElB,CAAC,CAAC,OAAO++F,GAAG/hG,EAAE,EAAEM,CAAC,CAAC,CAAC,SAAS2hG,GAAIjiG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE0B,EAAEu4C,EAAE,mBAAmB,EAAEj6C,EAAE,MAAM,MAAM,EAAE2B,EAAE3B,EAAE4B,EAAE,CAAC,EAAEF,GAAG,OAAOC,EAAEg+F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK+B,CAAC,CAAC,CAAC,EAAEE,EAAE,KAAKD,CAAC,EAAE,EAAEs4C,EAAE,iBAAiB,EAAE,OAAOt4C,EAAE,MAAM,MAAM,GAAGs4C,EAAE,2BAA2B,SAAS,CAAC,EAAE,EAAE,EAAEt4C,EAAE,MAAM,MAAM,EAAE,IAAIU,EAAEo/F,GAAG9hG,EAAEgC,EAAE,EAAE,GAAG,KAAK,EAAE,OAAOC,EAAE,QAAQU,GAAG3C,EAAE,8BAA8B2C,CAAC,CAAC,EAAED,CAAC,CAAC,IAAIs/F,GAAG,CAAC,WAAWzmF,GAAG,YAAY,QAAQ,WAAWwmF,EAAG,EAAE,SAASE,GAAIniG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE0B,EAAEu4C,EAAE,mBAAmB,EAAEj6C,EAAE,MAAM,MAAM,EAAE2B,EAAE3B,EAAE4B,EAAE,CAAC,EAAEF,GAAG,OAAOC,EAAEg+F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK+B,CAAC,CAAC,CAAC,EAAEE,EAAE,KAAKD,CAAC,EAAE,EAAEs4C,EAAE,iBAAiB,EAAE,OAAOt4C,EAAE,MAAM,MAAM,GAAGs4C,EAAE,2BAA2B,SAAS,CAAC,EAAE,EAAE,EAAEt4C,EAAE,MAAM,MAAM,EAAE,IAAIU,EAAEo/F,GAAG9hG,EAAEgC,EAAE,EAAE,GAAG,KAAK,EAAE,OAAOC,EAAE,QAAQU,GAAG3C,EAAE,8BAA8B2C,CAAC,CAAC,EAAED,CAAC,CAAC,IAAIw/F,GAAG,CAAC,WAAW1mF,GAAG,YAAY,QAAQ,WAAWymF,EAAG,EAAME,GAAIvG,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA,EAK1mEwG,GAAI3D,GAAG,CAAC,UAAU0D,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAW5mF,GAAG,YAAY,QAAQ,WAAW2mF,EAAG,EAAME,GAAI1G,GAAG,qCAAqC2G,GAAI9D,GAAG,CAAC,UAAU6D,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAW9mF,GAAG,YAAY,QAAQ,WAAW6mF,EAAG,EAAME,GAAI7G,GAAG;AAAA;AAAA,EAEvN8G,GAAIjE,GAAG,CAAC,UAAUgE,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAWhnF,GAAG,YAAY,QAAQ,WAAW+mF,EAAG,EAAME,GAAIpF,GAAG;AAAA;AAAA,EAEzFqF,GAAI;AAAA;AAAA;AAAA;AAAA;AAAA,IAKFnF,GAAG;AAAA;AAAA,EAELoF,GAAIpE,GAAG,CAAC,UAAUkE,GAAI,gBAAgBC,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAWlnF,GAAG,YAAY,QAAQ,WAAWinF,EAAG,EAAME,GAAIpH,GAAG;AAAA;AAAA,6CAElEqH,GAAIxE,GAAG,CAAC,UAAUuE,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAWtnF,GAAG,YAAY,QAAQ,WAAWqnF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYviG,EAAEZ,EAAE,EAAEK,EAAE,GAAGwB,EAAE,GAAG,CAAC,GAAG,KAAK,cAAc,CAAC,GAAG,EAAE7B,IAAI,OAAO,EAAE,MAAM,IAAI,MAAM,4CAA4C,EAAE,IAAI8B,EAAElB,EAAE,YAAYmB,EAAEnB,EAAE,aAAaoB,EAAEpB,EAAE,YAAY,EAAEA,EAAE,eAAe8B,EAAE9B,EAAE,cAAc,EAAEA,EAAE,sBAAsBgC,EAAEhC,EAAE,qBAAqBiC,EAAEjC,EAAE,QAAQ,IAAIkC,EAAElC,EAAE,QAAQ,KAAK,KAAK,YAAYA,EAAE,SAAS,IAAIoD,EAAEhE,IAAI,MAAMiE,EAAE,cAAcrD,EAAE,oBAAoBA,EAAE,mBAAmBA,EAAE,iBAAiByG,EAAE,SAASzG,EAAE,mBAAmBA,EAAE,iBAAiBiE,EAAE,MAAM,GAAGb,IAAIa,EAAE,gBAAgB,EAAE,CAAC,IAAIC,EAAE,KAAK,KAAK,SAAS;AAAA,sCAC9nB/C,MAAMC;AAAA,mCACTa,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBP;AAAA,sBACZ;AAAA;AAAA;AAAA,kCAGYlC,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAgC;AAAA,wBACZF;AAAA;AAAA;AAAA,oCAGY9B,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAUZkE;AAAA;AAAA;AAAA,mCAGSzE,EAAEwB,EAAEoC,EAAEoD,EAAE,QAAQzE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAM3C,MAAM,CAAC,IAAI6B,EAAE,MAAMC,EAAE,GAAG1E,KAAKA,KAAKA,sEAAsEA,IAAI,QAAQ0E,EAAE,oBAAoB,IAAIC,EAAE,KAAK,MAAM7C,EAAE,CAAC,EAAE,EAAE8C,EAAE9C,EAAE,EAAEwF,EAAE;AAAA,YACtKtD;AAAA;AAAA;AAAA,wBAGYS;AAAA;AAAA,MAElB,KAAK,SAAS;AAAA,oCACgB1C,MAAMC;AAAA,iCACTa,MAAMC;AAAA,0CACG+B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMZjE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBEiE;AAAA;AAAA;AAAA;AAAA,gCAIF;AAAA,oBACZ;AAAA;AAAA;AAAA,gCAGYjE,EAAE;AAAA;AAAA;AAAA;AAAA,kCAIA+D;AAAA,uCACKjC;AAAA;AAAA;AAAA;AAAA,yCAIEA;AAAA,6CACIA;AAAA,6CACAA;AAAA;AAAA;AAAA,cAG/B4E;AAAA;AAAA;AAAA,gCAGkB3C;AAAA,gBAChBC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,cAQN0C;AAAA,uBACS1C,IAAI;AAAA;AAAA;AAAA,yCAGclC;AAAA;AAAA;AAAA;AAAA;AAAA,cAK3B4E;AAAA,uBACS1C,IAAI;AAAA;AAAA;AAAA,yCAGclC;AAAA,6CACIA;AAAA;AAAA;AAAA;AAAA,cAI/B4E;AAAA;AAAA;AAAA,oBAGM5C;AAAA;AAAA,KAEf,CAAC,EAAE0+F,GAAG,KAAK,CAAC,YAAYxiG,EAAEZ,EAAE,EAAEK,EAAE,GAAGwB,EAAE,GAAG,CAAC,GAAG,KAAK,cAAc,CAAC,GAAG,EAAE7B,IAAI,OAAO,EAAE,MAAM,IAAI,MAAM,4CAA4C,EAAE,IAAI8B,EAAElB,EAAE,YAAYmB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,aAAa,EAAEA,EAAE,YAAY8B,EAAE9B,EAAE,cAAc,EAAEA,EAAE,eAAegC,EAAEhC,EAAE,cAAciC,EAAEjC,EAAE,qBAAqBkC,EAAElC,EAAE,sBAAsBoD,EAAEpD,EAAE,qBAAqBqD,EAAErD,EAAE,QAAQ,MAAMyG,EAAEzG,EAAE,QAAQ,IAAIiE,EAAEjE,EAAE,QAAQ,KAAK,KAAK,YAAYA,EAAE,SAAS,IAAI6D,EAAEzE,IAAI,MAAM0E,EAAE,MAAM,GAAGD,IAAIC,EAAE,gBAAgB,EAAE,CAAC,IAAIK,EAAE,KAAK,KAAK,SAAS;AAAA;AAAA,oBAErehD,MAAMC,MAAM;AAAA,mCACGiC,MAAMoD,MAAMxC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBbhC;AAAA,sBACZH;AAAA;AAAA;AAAA,kCAGY9B,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAkC;AAAA,wBACZ;AAAA;AAAA;AAAA,oCAGYlC,EAAE;AAAA;AAAA;AAAA;AAAA,sCAIAoD;AAAA,0BACZpB;AAAA;AAAA;AAAA,sCAGYhC,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,4BAUZmE;AAAA;AAAA;AAAA,qCAGS1E,EAAEwB,EAAE,cAAcjB,EAAE,mBAAmBA,EAAE,oBAAoBA,EAAE,mBAAmBA,EAAE,kBAAkB,UAAUA,EAAE,oBAAoBA,EAAE,mBAAmBA,EAAE,kBAAkB,QAAQkC,OAAOkB;AAAA,6BACtMA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAOrB,MAAM,CAAC,IAAIW,EAAE,MAAMC,EAAE,GAAG5E,KAAKA,KAAKA,sEAAsEA,IAAI,QAAQ4E,EAAE,oBAAoB,IAAI0C,EAAE,KAAK,MAAMxF,EAAE,CAAC,EAAE,EAAEgD,EAAEhD,EAAE,EAAEyF,EAAE;AAAA,YACtK9C;AAAA;AAAA;AAAA,wBAGYE;AAAA;AAAA,MAElB,KAAK,SAAS;AAAA;AAAA,gBAEJ5C,MAAMC,MAAM;AAAA,iCACKiC,MAAMoD,MAAMxC;AAAA,0CACHH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMZ9D,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAmBE8D;AAAA;AAAA;AAAA;AAAA,gCAIF7B;AAAA,oBACZH;AAAA;AAAA;AAAA,gCAGY9B,EAAE;AAAA;AAAA;AAAA;AAAA,kCAIAkC;AAAA,oBACd;AAAA;AAAA;AAAA,kCAGclC,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIA0G;AAAA,yCACK1E;AAAA;AAAA;AAAA;AAAA,+CAIMA;AAAA,mDACIA;AAAA,mDACAA;AAAA;AAAA;AAAA,gBAGnC2E;AAAA;AAAA;AAAA,kCAGkBD;AAAA,kBAChBxC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gBAQNyC;AAAA,yBACSzC,IAAI;AAAA;AAAA;AAAA,+CAGkBlC;AAAA;AAAA;AAAA;AAAA;AAAA,gBAK/B2E;AAAA,yBACSzC,IAAI;AAAA;AAAA;AAAA,+CAGkBlC;AAAA,mDACIA;AAAA;AAAA;AAAA;AAAA,gBAInC2E;AAAA;AAAA;AAAA,sBAGM3C;AAAA;AAAA;AAAA,KAGjB,CAAC,EAAE,SAASy+F,GAAIvjG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAEwtF,GAAGxtF,EAAE,SAAS,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,CAAC,EAAE5B,EAAE6B,EAAE,EAAE2kB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAEr4C,CAAC,EAAE,IAAI,wEAAwE,oBAAoBA,IAAI,EAAE,IAAIS,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE4B,EAAEF,EAAEC,CAAC,EAAE,GAAGU,EAAE,cAAc,GAAGA,EAAE,eAAe,GAAGkkB,EAAE,YAAYlkB,EAAE,QAAQA,EAAE,QAAQ,EAAE,OAAOk7F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAI2C,EAAE,IAAIwgG,GAAGzgG,EAAE,MAAM,EAAE,EAAE,OAAO1C,EAAE,gBAAgB2C,EAAE,CAACtC,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIijG,GAAG,CAAC,WAAWxnF,GAAG,YAAY,QAAQ,WAAWunF,EAAG,EAAE,SAASE,GAAIzjG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,EAAE,WAAWC,CAAC,EAAE7B,EAAEsC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAEqC,EAAEX,EAAEC,EAAEC,CAAC,EAAEW,EAAE,IAAIwgG,GAAGzgG,EAAE,MAAM,EAAE,EAAE,OAAO3C,EAAE,gBAAgB4C,EAAE,CAACvC,CAAC,EAAE,SAAS,CAAC,CAAC,IAAImjG,GAAG,CAAC,WAAWxnF,GAAG,YAAY,QAAQ,WAAWunF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY7iG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,aAAaiB,EAAEjB,EAAE,YAAYkB,EAAElB,EAAE,eAAemB,EAAEnB,EAAE,cAAcoB,EAAEpB,EAAE,sBAAsB,EAAEA,EAAE,qBAAqB8B,EAAEV,EAAE,EAAEpB,EAAE,QAAQ,IAAI,EAAE,EAAE,EAAEA,EAAE,QAAQ,KAAKgC,EAAE,GAAG5C,EAAE,GAAG,KAAK,SAAS;AAAA,iCACpnC0C,MAAM;AAAA,0CACGE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAcVZ;AAAA,oBACZF;AAAA,gDAC4BzB;AAAA;AAAA,oCAEZO,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKJ;AAAA,mBACfmB;AAAA,kDAC+BF;AAAA;AAAA,sCAEZjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAanC,CAAC,EAAE8iG,GAAG,KAAK,CAAC,YAAY9iG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,YAAY,EAAEA,EAAE,aAAaP,EAAEO,EAAE,YAAYiB,EAAEjB,EAAE,YAAYkB,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,cAAc,EAAEA,EAAE,eAAe8B,EAAE9B,EAAE,cAAc,EAAEA,EAAE,qBAAqBgC,EAAEhC,EAAE,sBAAsBiC,EAAEjC,EAAE,qBAAqBkC,EAAE,EAAE,EAAElC,EAAE,QAAQ,MAAMoD,EAAEpB,EAAE,EAAEhC,EAAE,QAAQ,IAAIqD,EAAEpB,EAAE,EAAEjC,EAAE,QAAQ,KAAKyG,EAAE,GAAGrH,EAAE,EAAEK,GAAG,KAAK,SAAS;AAAA,iCACtXyC,MAAMkB,MAAMC;AAAA,0CACHoD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAiBV;AAAA,oBACZrF;AAAA,gDAC4BH;AAAA;AAAA,oCAEZjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKJgC;AAAA,sBACZ;AAAA,kDAC4Bd;AAAA;AAAA,sCAEZlB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAMJiC;AAAA,wBACZH;AAAA,oDAC4BX;AAAA;AAAA,wCAEZnB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAcrC,CAAC,EAAE,SAAS+iG,GAAI7jG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,WAAW0B,EAAE,QAAQC,EAAE,IAAIC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEuC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE03C,EAAE,kBAAkB,EAAE,MAAMv4C,EAAEC,EAAEW,EAAEV,EAAES,CAAC,EAAEG,EAAE,IAAI6gG,GAAG9gG,CAAC,EAAE,OAAO5C,EAAE,gBAAgB6C,EAAE,CAACxC,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,IAAIujG,GAAG,CAAC,WAAW3nF,GAAG,YAAY,QAAQ,WAAW0nF,EAAG,EAAE,SAASE,GAAI/jG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,EAAE,EAAEwtF,GAAG,CAACxtF,EAAE,CAAC,EAAE,aAAa,EAAE,GAAG,CAAC,WAAW0B,EAAE,QAAQC,EAAE,IAAIC,CAAC,EAAE7B,EAAEsC,EAAE43C,EAAE,kBAAkB,EAAE,MAAMv4C,EAAEC,EAAE,EAAEC,CAAC,EAAEU,EAAE,IAAI8gG,GAAG/gG,CAAC,EAAE,OAAO1C,EAAE,gBAAgB2C,EAAE,CAACtC,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,IAAIyjG,GAAG,CAAC,WAAW/nF,GAAG,YAAY,QAAQ,WAAW8nF,EAAG,EAAE,SAASE,GAAIjkG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,WAAW0B,CAAC,EAAE3B,EAAE,OAAO+/F,GAAG,CAAC,EAAE9/F,EAAE,EAAE,EAAE,WAAW,EAAE,WAAW0B,EAAE,QAAQ/B,CAAC,CAAC,CAAC,CAAC,IAAIgkG,GAAG,CAAC,WAAW9nF,GAAG,YAAY,QAAQ,WAAW6nF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYrjG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,cAAc,CAAC,IAAI,OAAO,UAAU,EAAEw4C,EAAE,2BAA2B15C,EAAEZ,CAAC,EAAEs6C,EAAE,2BAA2B15C,EAAE,CAAC,EAAE,IAAImB,EAAE,MAAM1B,GAAG,OAAOi6C,EAAE,2BAA2B15C,EAAEP,CAAC,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE0B,EAAE,0BAA0B,IAAIC,EAAE,MAAMH,GAAG,OAAOy4C,EAAE,2BAA2B15C,EAAEiB,CAAC,EAAE,KAAK,cAAc,KAAK,OAAO,EAAEG,EAAE,yBAAyB,KAAK,YAAYpB,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKrqCmB;AAAA,wBACDC;AAAA,2DACmCF;AAAA;AAAA;AAAA,KAGtD,CAAC,EAAMoiG,GAAG,KAAK,CAAC,YAAYtjG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,cAAc,CAAC,IAAI,OAAO,UAAU,EAAEw4C,EAAE,2BAA2B15C,EAAEZ,CAAC,EAAEs6C,EAAE,2BAA2B15C,EAAE,CAAC,EAAE,IAAImB,EAAE,YAAY1B,GAAG,OAAOi6C,EAAE,2BAA2B15C,EAAEP,CAAC,EAAE,KAAK,cAAc,KAAK,QAAQ,EAAE0B,EAAE,0BAA0B,IAAIC,EAAE,YAAYH,GAAG,OAAOy4C,EAAE,2BAA2B15C,EAAEiB,CAAC,EAAE,KAAK,cAAc,KAAK,OAAO,EAAEG,EAAE,yBAAyB,KAAK,YAAYpB,EAAE,KAAK,SAAS;AAAA;AAAA,wBAEtcmB;AAAA,uBACDC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yDAMkCF;AAAA;AAAA;AAAA;AAAA,KAIpD,CAAC,EAAMqiG,GAAI,CAAC,CAAC,OAAOrkG,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,EAAE,KAAKC,EAAE,SAAS,EAAE,OAAO,EAAE,MAAM0B,CAAC,EAAEjC,EAAE8mB,EAAE,OAAOvmB,EAAE,MAAM,SAAS,EAAE,MAAM,OAAO,IAAI,8EAA8E,EAAEumB,EAAE,OAAO,GAAG,MAAMvmB,EAAE,MAAM,SAAS,EAAE,MAAM,OAAO,IAAI,4EAA4E,EAAEumB,EAAE,OAAO7kB,GAAG,MAAM1B,EAAE,MAAM,SAAS0B,EAAE,MAAM,OAAO,IAAI,2EAA2E,EAAE,GAAG,CAAC,gBAAgBC,CAAC,EAAEhC,EAAEgC,GAAG,OAAOA,EAAE,MAAM,IAAIC,EAAE,CAAC7B,EAAEC,EAAE,CAAC,EAAEqC,EAAE,KAAK,GAAG,OAAOA,EAAE,EAAE,MAAMT,EAAE,KAAK,CAAC,GAAG,IAAIU,EAAE,KAAKZ,GAAG,OAAOY,EAAEZ,EAAE,MAAME,EAAE,KAAKF,CAAC,GAAG,IAAIa,EAAE2X,EAAE,EAAE,QAAQ,0BAA0B,EAAE,IAAI2pF,GAAG9jG,EAAE,MAAMC,EAAE,MAAM,EAAE,MAAMqC,EAAEC,EAAEX,CAAC,EAAE,IAAIiiG,GAAG7jG,EAAE,MAAMC,EAAE,MAAM,EAAE,MAAMqC,EAAEC,EAAEX,CAAC,EAAE,OAAO,EAAE,gBAAgBY,EAAEX,EAAEA,EAAE,GAAG,KAAK,CAAC,EAAEmiG,GAAG,CAAC,WAAWrlF,GAAG,YAAY,QAAQ,WAAWolF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYzjG,EAAE,CAAC,KAAK,cAAc,CAAC,QAAQ,EAAE,KAAK,YAAYA,EAAE,KAAK,KAAKA,EAAE,OAAO,IAAIZ,EAAEw1F,GAAG,KAAK,IAAI,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,WAAW,KAAK,KAAK,KAAK,KAAK,CAAC,EAAE,IAAI,EAAE8O,GAAI,KAAK,IAAI,EAAEjkG,EAAEwB,EAAEjB,EAAE,IAAI,CAACkB,EAAEC,IAAI,aAAawiG,GAAGxiG,cAAcA,eAAewiG,GAAGxiG,KAAK,EAAE1B,EAAE;AAAA,UACplCL;AAAA,UACAA;AAAA,UACA6B,EAAE,KAAK;AAAA,CAChB;AAAA,QACO,KAAK,SAAS;AAAA;AAAA,UAEZxB;AAAA,8BACoB;AAAA;AAAA,KAEzB,CAAC,EAAEkkG,GAAG,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAE,SAASD,GAAIxkG,EAAE,CAAC,GAAGA,IAAI,EAAE,MAAM,YAAY,GAAGA,GAAG,EAAE,OAAOykG,GAAG,MAAM,EAAEzkG,CAAC,EAAE,IAAI,GAAG,aAAa,CAAC,EAAE,KAAK,GAAG,EAAE,MAAM,MAAM,oBAAoBA,wBAAwB,CAAC,CAAC,IAAI0kG,GAAG,KAAK,CAAC,YAAY5jG,EAAE,CAAC,KAAK,cAAc,CAAC,QAAQ,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYA,EAAE,KAAK,KAAKA,EAAE,OAAO,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,WAAW,KAAK,KAAK,KAAK,KAAK,CAAC,EAAE,IAAIZ,EAAEw1F,GAAG,KAAK,IAAI,EAAE,EAAEuF,GAAG,SAAS,KAAK,IAAI,EAAE16F,EAAE06F,GAAG,YAAY,KAAK,IAAI,EAAEl5F,EAAE,KAAK,OAAO,EAAE,YAAY,QAAQxB,EAAE,MAAM,EAAE,EAAE,KAAK,KAAKyB,EAAE,wBAAwBzB,EAAE,KAAK,OAAOwB,KAAKE,EAAE;AAAA,mBACtjBD;AAAA,cACL,EAAE,KAAK,KAAK,QAAQlB,EAAE,KAAK,KAAK;AAAA,YAClCP,EAAE,KAAK,KAAK;AAAA,qBACHyB;AAAA,YACTzB,EAAE,KAAK,KAAK;AAAA;AAAA,MAElB2B,EAAE,KAAK,OAAO,EAAE,GAAG;AAAA,UACf,EAAE,KAAK,KAAK;AAAA,cACR,EAAE,KAAK,KAAK,QAAQpB,EAAE,KAAK,KAAK;AAAA,YAClCP,EAAE,KAAK,KAAK;AAAA,qBACHyB;AAAA,gBACL,EAAE,KAAK,KAAK,QAAQlB,EAAE,KAAK,KAAK;AAAA,cAClCP,EAAE,KAAK,KAAK;AAAA,uBACHyB;AAAA;AAAA;AAAA,MAGjB,EAAE,KAAK,MAAM,EAAE;AAAA,cACP9B,KAAKY,EAAE,IAAI,CAAC8B,EAAE,IAAI,SAAS,IAAI,EAAE,KAAK,MAAM9B,EAAE,IAAI,CAAC8B,EAAE,IAAI,GAAGrC,EAAE,QAAQ,EAAE,cAAc,KAAK,EAAE,KAAK;AAAA,CAC/G,EAAE,KAAK,SAAS;AAAA;AAAA,UAEPL;AAAA,UACAA;AAAA,UACA;AAAA;AAAA,UAEA+B;AAAA,UACAC;AAAA;AAAA;AAAA,KAGL,CAAC,EAAE,SAASyiG,GAAI3kG,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAED,EAAE,QAAQ,IAAIN,EAAE,MAAM,EAAE,EAAEM,EAAE,eAAeJ,EAAEF,EAAE,KAAK,EAAE,EAAEM,EAAE,QAAQ,IAAI,EAAE,MAAM,EAAE,OAAO,OAAO,EAAEC,CAAC,EAAE,EAAE,SAAS,EAAE,EAAE,MAAML,EAAE,EAAE,MAAMF,EAAE,MAAM,IAAIiC,EAAEm0B,GAAG,kBAAkB,EAAEtP,EAAE,eAAe9mB,EAAE,KAAK,CAAC,EAAEO,EAAE,QAAQ0B,GAAG1B,EAAE,MAAM,YAAY,EAAE,MAAM,CAAC,WAAW0B,EAAE,WAAW1B,EAAE,OAAOA,EAAE,MAAM,YAAYP,EAAE,MAAM,EAAE,IAAIkC,EAAE5B,EAAE,aAAa,IAAI,EAAE,MAAM,UAAU,GAAG,EAAE,OAAOA,EAAE,aAAa,IAAI,EAAE,MAAM,WAAW4B,EAAE,CAAC,EAAE,CAAC,CAAC,SAAS0iG,GAAG5kG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,KAAK,CAAC,EAAED,EAAE,CAAC2B,EAAEC,CAAC,EAAEk0B,GAAG,iBAAiB71B,EAAE,EAAE,CAAC,EAAE,GAAG61B,GAAG,kBAAkB71B,EAAE0B,EAAEC,CAAC,EAAE4kB,EAAE,cAAc5kB,CAAC,IAAI,EAAE,OAAOhC,EAAE,eAAegC,EAAE3B,EAAE,MAAM,CAAC,CAAC,EAAE,GAAGL,EAAE,mBAAmB,CAACK,CAAC,CAAC,GAAGA,EAAE,QAAQ,SAAS,CAAC,IAAIsC,EAAE3C,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAEuC,EAAEo3F,GAAGr3F,EAAE,OAAOZ,EAAEC,EAAE3B,EAAE,MAAMA,EAAE,KAAK,EAAE,OAAOL,EAAE,eAAegC,EAAE3B,EAAE,MAAMuC,CAAC,CAAC,CAAC,GAAG,CAAC,SAASX,CAAC,EAAEjC,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAEqC,EAAEwzB,GAAG,iBAAiB71B,EAAE,MAAM0B,EAAEC,CAAC,EAAE,GAAGC,GAAG,CAACS,EAAE,CAAC,IAAIC,EAAE4X,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAIiqF,GAAGxiG,CAAC,EAAE,IAAIqiG,GAAGriG,CAAC,EAAEY,EAAE,CAACb,CAAC,EAAE,OAAO/B,EAAE,gBAAgB2C,EAAE,CAACtC,CAAC,EAAEA,EAAE,MAAMuC,CAAC,CAAC,CAAC,OAAO5C,EAAE,YAAYK,EAAE,MAAM,EAAEokG,GAAIpkG,EAAE0B,EAAEC,EAAEhC,CAAC,CAAC,CAAC,IAAI2kG,GAAG,CAAC,WAAWphF,GAAG,YAAY,QAAQ,WAAWmhF,EAAE,EAAME,GAAI9kG,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,MAAM,CAAC,EAAED,EAAEwmB,EAAE,OAAOvmB,EAAE,MAAM,QAAQ,EAAE,IAAI,sEAAsE,EAAE,IAAI0B,EAAE,EAAE,OAAO,CAAC,EAAE0C,IAAI,EAAEA,CAAC,EAAEzC,EAAEs4C,EAAE,YAAYj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEE,EAAEq4C,EAAE,YAAYt4C,EAAE,OAAO,EAAE,MAAM,EAAEU,EAAE43C,EAAE,oBAAoBj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEY,EAAE23C,EAAE,oBAAoB,EAAE,EAAE,MAAM,EAAE13C,EAAE03C,EAAE,aAAa53C,EAAE,EAAE,EAAE,MAAM,EAAEG,EAAE,CAAC,EAAEC,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEgC,EAAEg8F,GAAG,CAAC,OAAO,CAAC,EAAEl9F,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAE,EAAEk9F,GAAG,CAAC,OAAO,CAAC,EAAEn7F,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAE2E,EAAEq9F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ1kG,EAAE,MAAM,CAAC,MAAM2C,EAAE,KAAKC,CAAC,CAAC,CAAC,EAAE,OAAOC,EAAE,KAAKC,CAAC,EAAED,EAAE,KAAKmB,CAAC,EAAEnB,EAAE,KAAK,CAAC,EAAEA,EAAE,QAAQ,GAAG7C,EAAE,8BAA8B,CAAC,CAAC,EAAEqH,CAAC,EAAEw9F,GAAG,CAAC,WAAW1oF,GAAG,YAAY,QAAQ,WAAWyoF,EAAG,EAAE,SAASE,GAAIhlG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE2B,EAAE/B,EAAE,SAASK,EAAE,MAAM,EAAE2B,EAAEhC,EAAE,SAAS,EAAE,MAAM,EAAEiC,EAAE+1F,GAAGj2F,EAAEC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,OAAOhC,EAAE,eAAe,CAAC,CAAC,EAAE,EAAE,MAAMiC,CAAC,CAAC,CAAC,IAAI8iG,GAAG,CAAC,WAAW3oF,GAAG,YAAY,QAAQ,WAAW0oF,EAAG,EAAE,SAASE,GAAIllG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,GAAGM,EAAE,GAAGC,CAAC,EAAE,EAAE,EAAEL,EAAE,SAASI,EAAE,MAAM,EAAE,EAAEJ,EAAE,SAASK,EAAE,MAAM,EAAE0B,EAAEu4C,EAAE,2BAA2B,MAAM,KAAK,CAAC,EAAE,MAAM,KAAK,CAAC,CAAC,EAAE,OAAOt6C,EAAE,eAAe,CAAC+B,EAAE,MAAM,EAAE,QAAQ,WAAW,KAAKA,CAAC,CAAC,CAAC,CAAC,IAAIkjG,GAAG,CAAC,WAAW3oF,GAAG,YAAY,QAAQ,WAAW0oF,EAAG,EAAME,GAAI,wBAAwBC,GAAGzG,GAAG,CAAC,UAAUwG,GAAI,cAAc5L,GAAG,MAAM,MAAM,CAAC,EAAE8L,GAAG,CAAC,WAAW7jF,GAAG,YAAY,QAAQ,WAAW4jF,EAAE,EAAE,SAASE,GAAGvlG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEL,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE,OAAOw9F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,EAAE,mBAAmB,IAAI,EAAE,QAAQL,CAAC,CAAC,CAAC,CAAC,IAAIslG,GAAG,CAAC,WAAW/iF,GAAG,YAAY,QAAQ,WAAW8iF,EAAE,EAAME,GAAI,wBAAwB,SAASC,GAAG1lG,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI27F,GAAG77F,EAAE,MAAMylG,EAAG,EAAEnlG,EAAE,EAAE,gBAAgBJ,EAAE,CAACF,CAAC,EAAE,OAAO,EAAE,MAAM,CAAC,OAAOM,EAAE,OAAO,MAAMA,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,SAASqlG,GAAG3lG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,GAAG,IAAI,YAAY,CAAC,GAAGC,EAAE,QAAQ,YAAY,OAAOu9F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAI,EAAEmoC,GAAG9nC,EAAE,KAAK,EAAE0B,EAAE0jG,GAAG,CAAC,OAAO,CAAC,EAAEplG,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAEgC,EAAE87F,GAAG,CAAC,OAAO,CAAC,KAAK/7F,EAAE,KAAK,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAO,EAAE,QAAQ,EAAEA,EAAE,8BAA8B+B,CAAC,EAAEC,CAAC,CAAC,GAAG3B,EAAE,QAAQ,YAAY,CAAC,IAAI,EAAEglG,GAAG,CAAC,OAAO,CAAC,MAAMhlG,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE+B,EAAE0jG,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQzlG,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B,CAAC,EAAE+B,CAAC,CAAC,GAAG,CAAC6kB,EAAE,gBAAgBvmB,EAAE,MAAM,CAAC,EAAE,CAAC,IAAI,EAAEu9F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,MAAM,CAAC,OAAO,EAAE,OAAO,MAAM,EAAE,MAAM,MAAM,CAAC,CAAC,CAAC,GAAGA,EAAE,mBAAmB,CAACK,CAAC,CAAC,EAAE,CAAC,IAAI,EAAEL,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAE,OAAO,CAAC0B,EAAEC,EAAEC,CAAC,EAAEi2F,GAAG,EAAE73F,EAAE,MAAMA,EAAE,MAAM,CAAC,EAAE,OAAOL,EAAE,eAAe+B,EAAEC,EAAEC,CAAC,CAAC,CAAC,GAAG,IAAI,QAAQ,OAAOujG,GAAGnlG,EAAEL,CAAC,EAAE,GAAG,IAAI,OAAO,CAAC,IAAI,EAAEA,EAAE,eAAe,CAAC,EAAE,OAAO4mB,EAAE,uBAAuB,OAAO,CAAC,CAAC,EAAE5kB,EAAEmjG,GAAG,CAAC,OAAO,CAAC,EAAE9kG,EAAE,EAAE,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B,CAAC,EAAEgC,CAAC,CAAC,MAAM,IAAI,MAAM,iCAAiC3B,EAAE,YAAY,GAAG,CAAC,CAAC,IAAIqlG,GAAG,CAAC,WAAWnpF,GAAG,YAAY,QAAQ,WAAWkpF,EAAE,EAAME,GAAG,kBAAkBC,GAAInH,GAAG,CAAC,UAAUkH,GAAG,gBAAgBA,GAAG,cAAcxN,EAAE,CAAC,EAAE0N,GAAG,CAAC,WAAWrpF,GAAG,YAAY,QAAQ,WAAWopF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYllG,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,SAAS,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAWjlI,CAAC,EAAMmlG,GAAG,KAAK,CAAC,YAAYnlG,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,SAAS,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAWnN,CAAC,EAAE,SAASolG,GAAIlmG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,aAAa,CAAC,EAAED,EAAE2B,EAAEwY,EAAE,EAAE,QAAQ,iBAAiB,EAAExY,EAAE,IAAIgkG,GAAG1lG,EAAE,KAAK,EAAE0B,EAAE,IAAI+jG,GAAGzlG,EAAE,KAAK,EAAE,IAAI2B,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,OAAOhC,EAAE,gBAAgB+B,EAAE,CAAC1B,CAAC,EAAEA,EAAE,MAAM2B,CAAC,CAAC,CAAC,IAAIikG,GAAG,CAAC,WAAWxpF,GAAG,YAAY,QAAQ,WAAWupF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYtlG,EAAE,CAAC,KAAK,cAAc,CAAC,OAAO,MAAM,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAaxX,CAAC,EAAE,SAASulG,GAAGrmG,EAAE,EAAE,CAAC,MAAM,CAAC,OAAO,EAAE,OAAO,MAAM,EAAE,MAAM,MAAMA,EAAE,KAAK,CAAC,CAAC,SAASsmG,GAAItmG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAEC,EAAEL,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE,EAAE,IAAI8lG,GAAG9lG,EAAE,KAAK,EAAE,EAAE,CAAC+lG,GAAG/lG,EAAEC,EAAE,mBAAmB,IAAI,EAAE8lG,GAAG/lG,EAAEC,EAAE,mBAAmB,IAAI,CAAC,EAAE,OAAOL,EAAE,gBAAgB,EAAE,EAAE,EAAE,GAAG,KAAK,CAAC,CAAC,IAAIqmG,GAAG,CAAC,WAAW1pF,GAAG,YAAY,QAAQ,WAAWypF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY1lG,EAAE,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,YAAY05C,EAAE,gBAAgB15C,EAAE,CAAC,EAAE,KAAK,cAAcA,EAAE,IAAI,CAACkB,EAAEC,IAAI,IAAIA,GAAG,EAAE,IAAI/B,EAAE,IAAI,MAAMY,EAAE,OAAO,CAAC,EAAEZ,EAAE,GAAGY,EAAE,GAAG,GAAG,QAAQkB,EAAE,EAAEA,EAAE9B,EAAE,OAAO8B,IAAI9B,EAAE8B,GAAG9B,EAAE8B,EAAE,GAAGlB,EAAEkB,GAAG,GAAG,IAAI,EAAE,CAAC,YAAY9B,EAAE,+BAA+B,EAAE,QAAQ8B,EAAE,EAAEA,EAAE9B,EAAE,OAAO8B,IAAI,CAAC,IAAIC,EAAE/B,EAAE8B,EAAE,GAAG,EAAE,KAAK,iBAAiB9B,EAAE8B,qBAAqBA,YAAYC,MAAM,CAAC,CAAC,IAAI1B,EAAEL,EAAE,OAAO6B,EAAE7B,EAAEA,EAAE,OAAO,GAAG,EAAE,KAAK,sBAAsBK,YAAYwB,MAAM,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAMjyB,EAAE,KAAK;AAAA,SACR;AAAA;AAAA,KAEJ,CAAC,EAAM0kG,GAAG,KAAK,CAAC,YAAY3lG,EAAEZ,EAAE,CAAC,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYs6C,EAAE,gBAAgB15C,EAAEZ,CAAC,EAAE,IAAI,EAAE,KAAK,YAAYK,EAAE,EAAE,OAAOwB,EAAE2zF,GAAGn1F,CAAC,EAAEyB,EAAEi5F,GAAG,SAAS16F,CAAC,EAAE0B,EAAE,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,EAAE,MAAM,EAAE1B,CAAC,EAAE,KAAK,cAAcO,EAAE,IAAI,CAACoD,EAAEC,IAAI,IAAIA,GAAG,EAAE,IAAIjC,EAAE,IAAI,MAAMpB,EAAE,OAAO,CAAC,EAAEoB,EAAE,GAAGpB,EAAE,GAAGZ,GAAG,QAAQgE,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,IAAIhC,EAAEgC,GAAGhC,EAAEgC,EAAE,GAAGpD,EAAEoD,GAAGhE,GAAG,IAAI,EAAE+B,EAAE/B,GAAG0C,EAAEX,EAAE,MAAM,EAAE,EAAE,EAAEA,EAAE,KAAK,EAAEa,EAAE,OAAO,OAAOZ,EAAE;AAAA;AAAA,oBAEjZ,YAAYU,EAAE,KAAK;AAAA,WAC5B,QAAQsB,EAAE,EAAEA,EAAEhC,EAAE,OAAOgC,IAAI,CAAC,IAAIC,EAAEjC,EAAEgC,EAAE,GAAGpB,GAAG;AAAA,cACzC,OAAOZ,EAAEgC,UAAU,QAAQhC,EAAEgC,EAAE;AAAA;AAAA,kBAE3BA,KAAKwiG,GAAGzkG,EAAE,EAAEkC,CAAC;AAAA,mBACZuiG,GAAG9jG,EAAE,EAAEuB,CAAC;AAAA,UACjB,CAAC,IAAIpB,EAAEb,EAAE,OAAOc,EAAEd,EAAEA,EAAE,OAAO,GAAGY,GAAG;AAAA;AAAA,gBAE7BC,KAAK2jG,GAAGzkG,EAAE,EAAEe,CAAC;AAAA,iBACZ0jG,GAAG9jG,EAAE,EAAEI,CAAC,OAAO,KAAK,SAAS;AAAA,uBACvBf,EAAE,IAAIiC,GAAG,OAAOA,CAAC;AAAA,UAC9BpB;AAAA;AAAA;AAAA;AAAA,UAIAf;AAAA,sCAC4BC;AAAA;AAAA,UAE5BA,EAAEzB,EAAE,QAAQyB,EAAEzB,EAAE;AAAA,cACZyB,EAAEzB,EAAE,QAAQ,EAAEA,EAAE;AAAA,gCACEyB;AAAA;AAAA;AAAA,UAGtBA,EAAEzB,EAAE,QAAQyB,EAAEzB,EAAE;AAAA,cACZyB,EAAEzB,EAAE,QAAQ,EAAEA,EAAE;AAAA,gCACEyB;AAAA;AAAA;AAAA,UAGtBA,EAAEzB,EAAE,QAAQyB,EAAEzB,EAAE;AAAA,cACZyB,EAAEzB,EAAE,QAAQ,EAAEA,EAAE;AAAA,cAChByB,EAAEzB,EAAE,QAAQ,EAAEA,EAAE;AAAA,gCACEyB;AAAA;AAAA;AAAA;AAAA,KAI3B,CAAC,EAAE,SAAS0kG,GAAG1mG,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,QAAQ,CAAC,EAAE,OAAOA,EAAE,IAAI,CAAC+B,EAAEC,IAAIA,IAAI1B,EAAE,GAAGyB,OAAO7B,IAAI6B,CAAC,EAAE,KAAK,CAAC,CAAC,SAAS4kG,GAAG3mG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEL,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE,OAAOw9F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,EAAE,mBAAmB,IAAI,EAAE,QAAQL,CAAC,CAAC,CAAC,CAAC,IAAI0mG,GAAG,CAAC,WAAWpnF,GAAG,YAAY,QAAQ,WAAWmnF,EAAE,EAAE,SAASE,GAAG7mG,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,GAAG,MAAM,GAAGM,IAAI,YAAY,CAAC,IAAIuC,EAAE7C,EAAE,IAAI,GAAGulG,GAAG,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,QAAQrlG,CAAC,CAAC,CAAC,EAAE4C,EAAE9C,EAAE,IAAI,GAAG2mG,GAAG,CAAC,OAAO,CAAC,MAAM,CAAC,EAAE,QAAQzmG,CAAC,CAAC,CAAC,EAAE6C,EAAE8jG,GAAGhkG,EAAE,EAAE3C,CAAC,EAAE8C,EAAE6jG,GAAG/jG,EAAE,EAAE5C,CAAC,EAAEgE,EAAE85F,GAAG,CAAC,OAAO,CAAC,KAAKj7F,EAAE,KAAKC,CAAC,EAAE,QAAQ9C,CAAC,CAAC,EAAE,OAAO2C,EAAE,QAAQ,GAAG3C,EAAE,8BAA8B,CAAC,CAAC,EAAE4C,EAAE,QAAQ,GAAG5C,EAAE,8BAA8B,CAAC,CAAC,EAAEA,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAEkB,CAAC,CAAC,IAAI3D,EAAEL,EAAE,mBAAmBF,CAAC,EAAE,GAAGM,IAAI,WAAWC,EAAE,IAAIA,EAAE,CAAC,IAAIsC,EAAE7C,EAAE,IAAI,GAAG,CAAC,IAAI2E,EAAEmiB,EAAE,cAAc,EAAE,MAAM,MAAM,CAAC,CAAC,EAAE,OAAOu4E,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAAC,GAAGyE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE7B,EAAED,EAAE,IAAI,IAAI,CAAC,KAAK3C,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,EAAE6C,EAAEy3C,EAAE,gBAAgB33C,EAAE,IAAI,GAAG,EAAE,KAAK,EAAE,CAAC,EAAEG,EAAEH,EAAE,GAAG,MAAM,KAAK,EAAEqB,EAAEo0F,GAAGx1F,EAAEC,EAAEzC,EAAE0C,CAAC,EAAE,EAAEw3C,EAAE,gBAAgBx6C,EAAE,IAAI,GAAG,EAAE,KAAK,EAAE,CAAC,EAAEuH,EAAErH,EAAE,eAAe,EAAEI,EAAE4D,CAAC,EAAE,OAAOrB,EAAE,QAAQ,GAAG3C,EAAE,8BAA8B,CAAC,CAAC,EAAEqH,CAAC,CAAC,IAAI,EAAEkT,EAAE,EAAE,UAAU,8BAA8B,EAAE,GAAGza,EAAE,OAAO,EAAE,CAAC,IAAI6C,EAAE,CAAC,EAAE,QAAQE,EAAE,EAAEA,EAAE/C,EAAE,OAAO+C,GAAG,EAAE,CAAC,IAAIC,EAAEhD,EAAE,MAAM+C,EAAEA,EAAE,CAAC,EAAEF,EAAE,KAAKgkG,GAAG7jG,EAAE,EAAE9C,CAAC,CAAC,CAAC,CAAC,IAAI4C,EAAE+jG,GAAGhkG,EAAE,EAAE3C,CAAC,EAAE,QAAQ6C,KAAKF,EAAE3C,EAAE,8BAA8B6C,CAAC,EAAE,OAAOD,CAAC,CAAC,GAAG2X,EAAE,EAAE,QAAQ,6BAA6B,GAAGza,EAAE,GAAG,MAAM,OAAO,EAAE,CAAC,IAAI6C,EAAE,IAAI4jG,GAAGzmG,EAAE,IAAI8C,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,OAAO5C,EAAE,gBAAgB2C,EAAE7C,EAAEM,CAAC,CAAC,CAAC,GAAG,CAAC,UAAU,EAAE,SAAS2B,CAAC,EAAE6kG,GAAI9mG,EAAE,EAAEE,CAAC,EAAEgC,EAAE,IAAIskG,GAAG,EAAE,IAAI3jG,GAAGA,EAAE,KAAK,CAAC,EAAEV,EAAEjC,EAAE,gBAAgBgC,EAAE,EAAE5B,CAAC,EAAE,EAAE,QAAQuC,GAAG3C,EAAE,8BAA8B2C,CAAC,CAAC,EAAE,IAAID,EAAEy8F,GAAG,CAAC,OAAO,CAAC,EAAEl9F,CAAC,EAAE,MAAM,CAAC,MAAMF,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BiC,CAAC,EAAES,CAAC,CAAC,SAASkkG,GAAI9mG,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEk6C,EAAE,gBAAgBx6C,EAAE,IAAI+B,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,MAAM,CAAC,UAAU/B,EAAE,IAAI+B,GAAGs9F,GAAG,CAAC,OAAO,CAAC,EAAEt9F,CAAC,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG+kB,EAAE,cAAc/kB,EAAE,MAAM,MAAM,CAAC,CAAC,CAAC,CAAC,EAAE,QAAQ7B,CAAC,CAAC,CAAC,EAAE,SAASI,CAAC,CAAC,CAAC,SAASymG,GAAG/mG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAevmB,EAAE,EAAE,GAAG,KAAK,EAAE,GAAG,EAAE,EAAE,IAAI4B,GAAGA,EAAE,KAAK,EAAEq4C,EAAE,uBAAuB,EAAE,CAAC,EAAE,IAAIv4C,EAAEu4C,EAAE,gBAAgB,EAAE,IAAIr4C,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,GAAG2kB,EAAE,cAAc7kB,CAAC,IAAI,EAAE,OAAO/B,EAAE,eAAe+B,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,EAAE,IAAIC,EAAE,EAAE,OAAOC,GAAG2kB,EAAE,cAAc3kB,EAAE,KAAK,EAAE,CAAC,EAAE,OAAOD,EAAE,SAAS,EAAE47F,GAAG,CAAC,OAAO,CAAC,EAAE57F,EAAE,EAAE,EAAE,QAAQhC,CAAC,CAAC,EAAE2mG,GAAG3kG,EAAE,EAAEhC,CAAC,CAAC,CAAC,IAAI8mG,GAAG,CAAC,WAAWlqF,GAAG,YAAY,QAAQ,WAAWiqF,EAAE,EAAME,GAAG,KAAK,CAAC,YAAYnmG,EAAEZ,EAAE,GAAG,EAAE,KAAKK,EAAE,GAAGwB,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,YAAYjB,EAAE,SAAS,IAAIkB,EAAElB,EAAE,QAAQ,IAAImB,EAAEnB,EAAE,QAAQ,KAAKoB,EAAEpB,EAAE,aAAa,EAAEA,EAAE,YAAY8B,EAAE9B,EAAE,eAAe,EAAEA,EAAE,cAAcgC,EAAEhC,EAAE,aAAaiC,EAAEjC,EAAE,YAAYkC,EAAE,KAAK,MAAMlC,EAAE,WAAW,CAAC,EAAE,EAAEoD,EAAEpD,EAAE,WAAW,EAAEqD,EAAErD,EAAE,aAAa,eAAeyG,EAAEpD,EAAE,EAAE,EAAEY,EAAEZ,EAAE,EAAE,EAAEQ,EAAER,EAAE,EAAE,EAAES,EAAE,GAAGC,EAAE,GAAG,IAAItE,EAAEqE,EAAE;AAAA;AAAA,YAE7mF;AAAA,WACD7C,EAAE6C,EAAE;AAAA;AAAA,YAEH;AAAA,WACDA,EAAE;AAAA;AAAA,cAEC;AAAA;AAAA,UAEJC,EAAE,gCAAgC,IAAIC,EAAE5E,EAAE,kCAAkC,GAAGA,GAAG,KAAK,cAAc,KAAK,MAAM,EAAEK,GAAG,KAAK,cAAc,KAAK,wBAAwB,EAAEwB,GAAG,KAAK,cAAc,KAAK,gBAAgB,EAAE,KAAK,SAAS;AAAA,QACpO6C;AAAA;AAAA,oCAE4B1C,MAAM;AAAA,iCACTF,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKb0C;AAAA;AAAA;AAAA,2BAGC4C,cAAcxC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAOTjC;AAAA,qCACKF;AAAA;AAAA,gCAEL9B,EAAE;AAAA;AAAA;AAAA;AAAA,kCAIAiC;AAAA,uCACK;AAAA;AAAA,kCAELjC,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAkC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oBAQhBmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kBAmBFD,IAAI;AAAA;AAAA,oBAEFC;AAAA;AAAA,0CAEsBnB;AAAA,mCACPA;AAAA;AAAA;AAAA,kCAGDA;AAAA,mCACCA;AAAA;AAAA;AAAA,yBAGVkB,IAAI;AAAA;AAAA,+BAEElB;AAAA,+BACAA;AAAA;AAAA;AAAA,oBAGXmB;AAAA;AAAA,wCAEoBnB;AAAA,wCACAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKRA;AAAA,gCACAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKPkB,IAAI;AAAA;AAAA,+BAEElB;AAAA,+BACAA;AAAA,+BACAA;AAAA;AAAA;AAAA,oBAGXmB;AAAA;AAAA,wCAEoBnB;AAAA,wCACAA;AAAA,wCACAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKRA;AAAA,gCACAA;AAAA,gCACAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAUtB8B;AAAA,UACAD;AAAA;AAAA;AAAA,KAGL,CAAC,EAAEqiG,GAAG,KAAK,CAAC,YAAYpmG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,YAAYA,EAAE,SAAS,IAAIZ,EAAEY,EAAE,QAAQ,MAAM,EAAEA,EAAE,QAAQ,IAAIP,EAAEO,EAAE,QAAQ,KAAKiB,EAAEjB,EAAE,YAAYkB,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,cAAc,EAAEA,EAAE,eAAe8B,EAAE9B,EAAE,cAAc,EAAEA,EAAE,YAAYgC,EAAEhC,EAAE,aAAaiC,EAAEjC,EAAE,YAAYkC,EAAE,KAAK,MAAMlC,EAAE,WAAW,CAAC,EAAE,EAAEoD,EAAEpD,EAAE,WAAW,EAAE,KAAK,SAAS;AAAA,oCACnUiB,MAAMC,MAAMC;AAAA,iCACf/B,MAAM,MAAMK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAgBb;AAAA,qCACK2B;AAAA;AAAA,gCAELpB,EAAE;AAAA;AAAA;AAAA;AAAA,kCAIAgC;AAAA,uCACK;AAAA;AAAA,kCAELhC,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAiC;AAAA,yCACKH;AAAA;AAAA,oCAEL9B,EAAE;AAAA;AAAA;AAAA;AAAA,sCAIAkC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oBAiBlBkB,IAAI;AAAA;AAAA,4CAEoBlB;AAAA,qCACPA;AAAA,2BACVkB,IAAI;AAAA;AAAA,4CAEalB;AAAA,4CACAA;AAAA;AAAA;AAAA,qCAGPA;AAAA,qCACAA;AAAA;AAAA;AAAA,2BAGVkB,IAAI;AAAA;AAAA,4CAEalB;AAAA,4CACAA;AAAA,4CACAA;AAAA;AAAA;AAAA,qCAGPA;AAAA,qCACAA;AAAA,qCACAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAShC,CAAC,EAAMmkG,GAAG,KAAK,CAAC,YAAYrmG,EAAEZ,EAAE,GAAG,EAAE,KAAKK,EAAE,GAAGwB,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,OAAO,KAAK,OAAO,EAAE,CAAC,KAAK,UAAU,KAAK,OAAO,EAAE,CAAC,KAAK,YAAY,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYjB,EAAE,SAAS,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIh0F,EAAElB,EAAE,QAAQ,KAAKmB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,cAAc,EAAEA,EAAE,aAAa8B,EAAE9B,EAAE,YAAY,EAAE8B,EAAEE,EAAE;AAAA;AAAA,gDAE5Y,QAAQqB,EAAE,EAAEA,EAAEvB,EAAEuB,IAAIrB,GAAG;AAAA,yBAC9CqB,EAAE;AAAA,wBACHA,EAAE;AAAA,yBACDA,EAAE,EAAE;AAAA,wBACLA,EAAE,EAAE;AAAA,oBACRA,KAAKrB,GAAG;AAAA,2BACD;AAAA,8BACGhC,EAAE;AAAA,SACvB,QAAQqD,EAAE,EAAEA,EAAEvB,EAAEuB,IAAIrB,GAAG;AAAA,oBACZqB,EAAE;AAAA,oBACFA,EAAE;AAAA,oBACFA,EAAE,EAAE;AAAA,oBACJA,EAAE,EAAE;AAAA,eACTA,iBAAiBrB,GAAG;AAAA;AAAA;AAAA,SAG1B,QAAQqB,EAAE,EAAEA,GAAG,EAAE,GAAG,EAAEA,IAAI,CAAC,IAAIoD,EAAEpD,EAAE,EAAE,GAAGrB,GAAG;AAAA,6BACvByE,EAAErF;AAAA,aAClBD,IAAI,GAAG,GAAGsF,EAAE3E,IAAIZ,EAAE,IAAI,GAAGc,GAAG;AAAA;AAAA,uEAE8ByE;AAAA,4BAC3CA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKEA;AAAA;AAAA,4BAEFA;AAAA;AAAA,iBAEXrF,IAAI,GAAGqF,EAAE,EAAEzE,GAAG;AAAA,qBACVyE,mBAAmBA,EAAE,gBAAgBA;AAAA,mBACvCzE,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAYGyE,gCAAgCA;AAAA;AAAA,yBAEhCA,6BAA6BA;AAAA;AAAA,sBAEhCzE,GAAG;AAAA,2DACkCyE;AAAA,4BAC/BA;AAAA;AAAA,8BAEEA;AAAA;AAAA,4BAEFA;AAAA;AAAA;AAAA,qBAGPA,cAAcA;AAAA,mBAChBA,EAAE,EAAE3E,GAAG,CAAC,IAAImC,EAAE/C,EAAE,IAAI,EAAE8kB,EAAE,kBAAkB5kB,CAAC,EAAEA,EAAEA,EAAE,IAAI,GAAGF,EAAE,IAAI,GAAGE,EAAE,IAAI,GAAGF,EAAE,IAAI,GAAGc,GAAG;AAAA,wDACjDiC;AAAA;AAAA,yEAEiBwC,EAAE;AAAA,8BAC7CA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKAA,EAAE;AAAA;AAAA,8BAEJA,EAAE;AAAA;AAAA,qBAEXrF,EAAE,EAAEY,GAAG;AAAA;AAAA;AAAA;AAAA,0BAIFyE,EAAE,gCAAgCA,EAAE;AAAA;AAAA,0BAEpCA,EAAE,6BAA6BA,EAAE;AAAA;AAAA,uBAEpCzE,GAAG;AAAA,yBACDyE,EAAE,mBAAmBA,gBAAgBA,EAAE;AAAA,wBACxCxC,IAAI,EAAEjC,GAAG;AAAA,yBACRyE,EAAE,cAAcA;AAAA,uBAClBzE,GAAG;AAAA,uCACaiC;AAAA;AAAA,2EAEoCwC,EAAE;AAAA,gCAC7CA,EAAE;AAAA;AAAA,kCAEAA,EAAE;AAAA;AAAA,gCAEJA,EAAE;AAAA;AAAA;AAAA,yBAGTA,EAAE,cAAcA,EAAE;AAAA,sBACrB,OAAOA,EAAE3E,IAAIZ,EAAE,IAAI,GAAGc,GAAG;AAAA;AAAA,sEAEuByE;AAAA,4BAC1CA;AAAA;AAAA;AAAA;AAAA,8BAIEA;AAAA;AAAA,4BAEFA;AAAA;AAAA;AAAA,kEAGsCA,EAAE;AAAA,4BACxCA,EAAE;AAAA;AAAA;AAAA;AAAA,8BAIAA,EAAE;AAAA;AAAA,4BAEJA,EAAE;AAAA;AAAA;AAAA,qBAGTA,mBAAmBA,gBAAgBA,EAAE;AAAA,iBACzCA,EAAE,EAAE3E,IAAIE,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uBAMLyE,EAAE,mBAAmBA,EAAE;AAAA,sBACxBzE,GAAG;AAAA,0DACiCyE;AAAA,4BAC9BA;AAAA;AAAA,8BAEEA;AAAA;AAAA,4BAEFA;AAAA;AAAA;AAAA;AAAA,sEAI0CA,EAAE;AAAA,4BAC5CA,EAAE;AAAA;AAAA,8BAEAA,EAAE;AAAA;AAAA,4BAEJA,EAAE;AAAA;AAAA;AAAA,qBAGTA;AAAA,4BACOA,gBAAgBA,EAAE;AAAA,iBAC7BA,EAAE,EAAE3E,IAAIE,GAAG;AAAA,uBACLyE,EAAE,mBAAmBA,gBAAgBA,EAAE;AAAA,sBACxCA,EAAE3E,IAAIE,GAAG;AAAA,gCACCyE;AAAA,4BACJA;AAAA,2BACDzG,EAAE;AAAA,8BACCyG;AAAA;AAAA,aAEjBA,EAAE,EAAE3E,IAAIE,GAAG;AAAA,kCACUyE,EAAE;AAAA,8BACNA,EAAE;AAAA,6BACHzG,EAAE;AAAA,gCACCyG,EAAE;AAAA;AAAA,gBAElB,CAACzE,GAAG;AAAA;AAAA,KAEfA,GAAG;AAAA;AAAA,KAEHA,GAAG;AAAA;AAAA,KAEH,IAAIC,EAAE,GAAGC,EAAE,GAAG,IAAIzC,EAAEwC,EAAE;AAAA;AAAA,aAEd;AAAA,YACDhB,EAAEgB,EAAE;AAAA;AAAA,aAEH;AAAA,YACDA,EAAE;AAAA,aACD;AAAA,YACDC,EAAE,gCAAgC,IAAIkB,EAAEhE,EAAE,kCAAkC,GAAGA,GAAG,KAAK,cAAc,KAAK,MAAM,EAAEK,GAAG,KAAK,cAAc,KAAK,wBAAwB,EAAEwB,GAAG,KAAK,cAAc,KAAK,gBAAgB,EAAE,KAAK,SAAS;AAAA,SACrOgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,WAaED;AAAA;AAAA;AAAA,WAGAoB;AAAA,WACAlB;AAAA;AAAA;AAAA,MAGL,CAAC,EAAMokG,GAAG,KAAK,CAAC,YAAYtmG,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,aAAa,KAAK,OAAO,EAAE,CAAC,KAAK,MAAM,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,EAAE,CAAC,KAAK,WAAW,KAAK,OAAO,EAAE,CAAC,KAAK,aAAa,KAAK,KAAK,EAAE,CAAC,KAAK,mBAAmB,KAAK,KAAK,EAAE,CAAC,KAAK,WAAW,KAAK,KAAK,CAAC,EAAE,KAAK,YAAYY,EAAE,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,GAAG,CAAC,WAAW,CAAC,EAAE91F,EAAEK,EAAE8xF,GAAG,EAAEtwF,EAAE,IAAI,eAAeC,EAAED,EAAE,EAAE,EAAEE,EAAEF,EAAE,EAAE,EAAEG,EAAE,KAAK,oBAAoB,sDAAsD,mBAAmBpB,EAAE,eAAeA,EAAE,QAAQ,EAAE,GAAG,QAAQ8B,EAAE,EAAEA,GAAG,EAAEA,IAAI,QAAQ,EAAE,EAAE,GAAG,EAAE,IAAI,GAAG;AAAA,gCACjnB;AAAA,yBACPA;AAAA;AAAA,YAEbV;AAAA;AAAA;AAAA;AAAA,iCAIqBF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAQEC;AAAA;AAAA;AAAA;AAAA,sBAIbF;AAAA;AAAA,2BAEKa,EAAE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,2BAKJA,EAAE,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAOrB,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UASd;AAAA;AAAA,UAEArC,EAAE;AAAA;AAAA,KAEP,CAAC,EAAE,SAAS8mG,GAAGrnG,EAAE,EAAE,CAAC,IAAIE,EAAEF,EAAE,OAAO,OAAOE,GAAG,EAAE,EAAE,CAAC,GAAGF,EAAE,MAAM,EAAE,EAAE,EAAEA,EAAEE,EAAE,GAAGF,EAAEE,EAAE,GAAGF,EAAEE,EAAE,EAAE,EAAE,CAAC,GAAGF,EAAE,MAAM,EAAE,EAAE,EAAEA,EAAEE,EAAE,GAAGF,EAAEE,EAAE,GAAGF,EAAEE,EAAE,EAAE,EAAE,CAAC,GAAGA,IAAI,GAAGF,EAAE,GAAG,EAAE,CAACA,EAAE,GAAG,CAAC,EAAE,IAAI,CAAC,SAASsnG,GAAG,CAAC,EAAEtnG,EAAE,OAAO,EAAE,SAASE,EAAE,QAAQI,EAAE,KAAKC,EAAE,KAAK,uBAAuB,EAAE,KAAK,eAAe,EAAE,EAAE,WAAW0B,EAAE,IAAI,EAAE,CAAC,IAAIC,EAAElC,EAAE,MAAMmC,EAAE7B,EAAE,QAAQ,IAAIN,EAAE,MAAM,EAAE4C,EAAE1C,EAAE,WAAW2C,EAAEX,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGY,EAAE5C,EAAE,YAAY6C,EAAE7C,EAAE,aAAa,eAAe8C,EAAE,GAAGkB,EAAE,GAAG,EAAEqD,EAAE,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC,IAAI3C,EAAEyiG,GAAG,EAAE,MAAMtkG,CAAC,EAAE6B,GAAG,OAAO,EAAEy6F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAMsE,CAAC,CAAC,CAAC,EAAE2C,EAAE,KAAK,CAAC,EAAE,CAAC,GAAGhH,GAAG,KAAK,CAAC,IAAIqE,EAAEyiG,GAAG9mG,EAAE,MAAMwC,CAAC,EAAE6B,GAAG,OAAOrE,EAAE8+F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQD,EAAE,MAAM,CAAC,MAAMsE,CAAC,CAAC,CAAC,EAAE2C,EAAE,KAAKhH,CAAC,EAAE,CAAC,GAAG,GAAGsC,IAAI,GAAGC,IAAI,IAAIF,EAAEw9F,KAAKj+F,EAAE,UAAUY,GAAGZ,EAAE,SAAS,MAAMD,EAAE,GAAG,IAAI,GAAG4kB,EAAE,YAAY3kB,EAAE,MAAM,MAAM,EAAE,EAAED,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC,IAAI0C,EAAE1C,EAAE,GAAGA,EAAE,IAAIA,EAAE,GAAG,GAAG2C,EAAE,CAAC,OAAO7E,EAAE,OAAO,MAAM,CAAC,EAAE4E,EAAE1E,EAAE,UAAU,EAAE,MAAMF,EAAE,KAAK,EAAE8E,EAAE3C,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,MAAM,EAAEA,EAAE,MAAMA,EAAE,MAAM,OAAO,KAAK2kB,EAAE,OAAOgpE,GAAG3tF,EAAE,MAAM0C,EAAE,KAAK,EAAE,IAAI,kBAAkB1C,EAAE,YAAY0C,EAAE,kBAAkB,EAAE,IAAI2C,EAAE63F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAM,CAAC,EAAEJ,EAAE,WAAWA,EAAE,WAAW,CAAC,CAAC,CAAC,EAAEqH,EAAE,KAAKC,CAAC,EAAE,IAAIxC,EAAEq7F,GAAG,CAAC,EAAEx7F,EAAE,EAAE2C,EAAE,QAAQlH,EAAE,WAAW0C,EAAE,WAAWkB,EAAE,KAAK3D,EAAE,WAAW0B,EAAE,uBAAuB,EAAE,eAAe,CAAC,CAAC,EAAEwF,EAAEnH,EAAE,QAAQ,IAAI0E,EAAE,MAAM,EAAE8hB,EAAE,OAAOrf,EAAE,SAAS,IAAI,6CAA6C,EAAEtF,EAAE,MAAM2C,EAAE2C,EAAE,MAAMvH,EAAE,SAAS,EAAE49F,GAAG,CAAC,OAAO,CAAC,EAAE94F,CAAC,EAAE,QAAQ1E,CAAC,CAAC,EAAE,EAAE,MAAMJ,EAAE,SAASqH,EAAE,KAAKvC,CAAC,CAAC,KAAK,CAAC,IAAIJ,EAAE1E,EAAE,UAAUA,EAAE,SAAS2E,EAAEw6F,GAAG,CAAC,OAAO,CAAC,EAAEr/F,CAAC,EAAE,QAAQM,EAAE,MAAM,CAAC,MAAMyC,EAAE,CAAC7C,EAAE,UAAU0E,EAAE1E,EAAE,UAAU,EAAE,CAACA,EAAE,UAAUA,EAAE,WAAW0E,CAAC,CAAC,CAAC,CAAC,EAAEE,EAAEu6F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAM,CAAC,EAAEJ,EAAE,WAAWA,EAAE,WAAW,CAAC,CAAC,CAAC,EAAEsH,EAAE64F,GAAG,CAAC,EAAEt9F,EAAE8B,EAAEC,EAAE,EAAE/B,EAAE+B,EAAED,EAAE,WAAW,CAAC9B,EAAE,WAAWmB,EAAE,QAAQ5D,EAAE,KAAKC,EAAE,WAAW0B,EAAE,uBAAuB,EAAE,eAAe,CAAC,CAAC,EAAE,EAAEo9F,GAAG,CAAC,OAAO,CAAC,EAAE73F,CAAC,EAAE,QAAQlH,EAAE,MAAM,CAAC,MAAMJ,EAAE,QAAQ,CAAC,CAAC,EAAEqH,EAAE,KAAK1C,CAAC,EAAE0C,EAAE,KAAKzC,CAAC,EAAEyC,EAAE,KAAKC,CAAC,CAAC,CAAC,QAAQ5C,KAAK2C,EAAEjH,EAAE,8BAA8BsE,CAAC,EAAE,OAAO,CAAC,CAAC,SAAS2iG,GAAG,CAAC,EAAEvnG,EAAE,OAAO,EAAE,SAASE,EAAE,QAAQI,EAAE,KAAKC,EAAE,KAAK,uBAAuB,EAAE,KAAK,eAAe,EAAE,EAAE,WAAW0B,EAAE,IAAI,EAAE,CAAC,GAAG,CAAC,YAAYC,EAAE,aAAaC,EAAE,WAAWS,EAAE,SAASC,EAAE,UAAUC,EAAE,WAAWC,CAAC,EAAE7C,EAAE8C,EAAED,IAAI,eAAemB,EAAEhC,EAAEC,EAAES,EAAE,EAAEE,EAAED,EAAE0E,EAAE,CAACrH,EAAE,UAAUgE,EAAE,CAAC,EAAE,EAAE,GAAGS,EAAE,GAAGC,EAAE,CAAC,EAAE,GAAG,GAAG,KAAK,CAAC,IAAIsD,EAAEm/F,GAAG,EAAE,MAAMrkG,CAAC,EAAEkF,GAAG,OAAO,EAAEm3F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAM4H,CAAC,CAAC,CAAC,EAAEtD,EAAE,KAAK,CAAC,EAAE,CAAC,GAAGrE,GAAG,KAAK,CAAC,IAAI2H,EAAEm/F,GAAG9mG,EAAE,MAAMyC,CAAC,EAAEkF,GAAG,OAAO3H,EAAE8+F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQD,EAAE,MAAM,CAAC,MAAM4H,CAAC,CAAC,CAAC,EAAEtD,EAAE,KAAKrE,CAAC,EAAE,CAAC,IAAIsE,EAAEw6F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE4D,EAAE4iB,EAAE,cAAc,EAAE,KAAK,EAAE5iB,CAAC,CAAC,CAAC,CAAC,EAAEU,EAAE,KAAKC,CAAC,EAAE,IAAIC,EAAE,IAAIsiG,GAAG7/F,EAAErH,CAAC,EAAEsH,EAAE,CAACxH,EAAE,MAAM,CAACE,EAAE,QAAQ,IAAIA,EAAE,QAAQ,IAAI,EAAE,CAACA,EAAE,aAAaA,EAAE,WAAW,EAAE,CAACA,EAAE,eAAeA,EAAE,aAAa,EAAE,CAACA,EAAE,UAAU,EAAE,CAACA,EAAE,YAAYA,EAAE,UAAU,EAAE,CAACA,EAAE,QAAQ,CAAC,EAAE8E,EAAE1E,EAAE,gBAAgBwE,EAAE,CAAC9E,CAAC,EAAE,UAAUwH,CAAC,EAAEC,EAAE43F,GAAG,CAAC,OAAO,CAAC,EAAEr6F,CAAC,EAAE,QAAQ1E,EAAE,MAAM,CAAC,MAAMiH,CAAC,CAAC,CAAC,EAAE3C,EAAE,KAAKI,CAAC,EAAEJ,EAAE,KAAK6C,CAAC,EAAE,IAAIxC,EAAE1E,GAAG,KAAK2E,EAAE,GAAG,KAAKC,EAAElD,IAAI,YAAYmD,EAAEnD,EAAE48F,GAAG58F,EAAE,EAAE,EAAE,KAAKoD,EAAE,IAAIy5F,GAAG97F,EAAEyE,EAAE,MAAM5C,EAAE,MAAM7B,EAAE6B,EAAE,MAAM4C,EAAE,MAAMzE,EAAE,CAAC9C,EAAE,UAAU,EAAEA,EAAE,WAAW,EAAE,CAACA,EAAE,UAAUA,EAAE,YAAY,CAAC,EAAE,EAAEyE,EAAEM,EAAEG,EAAEF,EAAEC,CAAC,EAAEG,EAAEtC,EAAE,CAACyE,EAAE5C,CAAC,EAAE,CAACA,EAAE4C,CAAC,EAAE,GAAGlH,GAAG+E,EAAE,KAAK/E,CAAC,EAAE2E,GAAGI,EAAE,KAAK,CAAC,EAAEH,EAAE,CAAC,IAAI+C,EAAE5H,EAAE,eAAe,CAAC,EAAE,UAAUwmB,EAAE,kBAAkB,EAAE,SAAS,CAAC,EAAExhB,EAAE,KAAK4C,CAAC,EAAEtD,EAAE,KAAKsD,CAAC,CAAC,CAAC,IAAID,EAAE3H,EAAE,gBAAgB+E,EAAEC,EAAE,SAAS,EAAEqC,EAAE03F,GAAG,CAAC,OAAO,CAAC,EAAEp3F,CAAC,EAAE,QAAQ3H,EAAE,MAAM,CAAC,MAAMJ,EAAE,QAAQ,CAAC,CAAC,EAAE0E,EAAE,KAAKqD,CAAC,EAAE,QAAQC,KAAKtD,EAAEtE,EAAE,8BAA8B4H,CAAC,EAAE,OAAOP,CAAC,CAAC,SAAS6/F,GAAIxnG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,EAAE,UAAUC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEuC,EAAE23C,EAAE,wBAAwBt4C,CAAC,EAAEY,EAAE03C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAE4B,EAAEF,EAAEW,EAAE,GAAGC,CAAC,EAAEE,EAAE,GAAGD,EAAE,eAAe,GAAGA,EAAE,cAAc,GAAGA,EAAE,iBAAiB,GAAGA,EAAE,gBAAgB,GAAGA,EAAE,eAAe,GAAGA,EAAE,cAAc,IAAIA,EAAE,QAAQ,OAAO,QAAQA,EAAE,QAAQ,OAAO,SAASC,EAAEukG,GAAG,CAAC,EAAE/mG,EAAE,OAAO,EAAE,SAASuC,EAAE,QAAQ5C,CAAC,CAAC,UAAU4C,EAAE,aAAa,GAAGD,IAAI,gBAAgB4X,EAAE,EAAE,QAAQ,gBAAgB,EAAE,CAAC,IAAIvW,EAAE,IAAIijG,GAAGrkG,CAAC,EAAE,EAAE,CAAC,CAACA,EAAE,QAAQ,IAAIA,EAAE,QAAQ,IAAI,EAAE,CAACA,EAAE,aAAaA,EAAE,WAAW,EAAE,CAACA,EAAE,eAAeA,EAAE,aAAa,EAAE,CAACA,EAAE,SAASA,EAAE,OAAO,CAAC,EAAEC,EAAE7C,EAAE,gBAAgBgE,EAAE,CAAC3D,EAAE,CAAC,EAAE,UAAU,CAAC,CAAC,SAASka,EAAE,EAAE,QAAQ,mBAAmB,EAAE1X,EAAEwkG,GAAG,CAAC,EAAEhnG,EAAE,OAAO,EAAE,SAASuC,EAAE,QAAQ5C,CAAC,CAAC,MAAM,CAAC,IAAIgE,EAAE,IAAI+iG,GAAGnkG,CAAC,EAAEC,EAAE7C,EAAE,gBAAgBgE,EAAE,CAAC3D,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIyC,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAEt8F,CAAC,EAAE,QAAQ7C,EAAE,MAAM,CAAC,MAAM4C,EAAE,QAAQ,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8B6C,CAAC,EAAEC,CAAC,CAAC,IAAIykG,GAAG,CAAC,WAAW1qF,GAAG,YAAY,QAAQ,WAAWyqF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY5mG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,IAAI,EAAE,KAAK,YAAYA,EAAE,YAAY,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,QAAQ,IAAIiB,EAAEjB,EAAE,QAAQ,KAAKkB,EAAElB,EAAE,aAAa,eAAe,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAYzxIA,EAAE;AAAA,kCACEA,EAAE;AAAA,iCACHZ,OAAOK;AAAA;AAAA,kCAENO,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAA,EAAE;AAAA,mCACH,OAAOiB;AAAA;AAAA,oCAENjB,EAAE;AAAA;AAAA;AAAA;AAAA,oBAIlBkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAef,CAAC,EAAE2lG,GAAG,KAAK,CAAC,YAAY7mG,EAAE,CAAC,KAAK,cAAc,CAAC,KAAK,GAAG,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,aAAaiB,EAAEjB,EAAE,YAAYkB,EAAElB,EAAE,aAAa,eAAemB,EAAE/B,EAAE,EAAEY,EAAE,QAAQ,IAAIoB,EAAE,EAAE,EAAEpB,EAAE,QAAQ,KAAK,EAAEkB,EAAE,EAAE,EAAEY,EAAEZ,EAAE,EAAE,EAAE,EAAEA,EAAE,EAAE,EAAE,KAAK,SAAS;AAAA,iCAC7OC,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKb;AAAA;AAAA,wCAEc,cAAcU;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAOtB1C;AAAA,gDACgBK;AAAA;AAAA,oCAEZO,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKbZ;AAAA;AAAA,kCAES;AAAA,kDACgB6B;AAAA;AAAA,sCAEZjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMb;AAAA;AAAA,oCAESA,EAAE;AAAA;AAAA,oBAElBkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAef,CAAC,EAAE4lG,GAAG,KAAK,CAAC,YAAY9mG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,IAAI,EAAE,KAAK,YAAYA,EAAE,YAAY,IAAIZ,EAAEY,EAAE,YAAY,EAAEA,EAAE,aAAaP,EAAEO,EAAE,YAAYiB,EAAEjB,EAAE,QAAQ,MAAMkB,EAAElB,EAAE,QAAQ,IAAImB,EAAEnB,EAAE,QAAQ,KAAK,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAWrLA,EAAE;AAAA,kCACEA,EAAE;AAAA,iCACHZ,OAAO6B;AAAA;AAAA,kCAENjB,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAA,EAAE;AAAA,mCACH,OAAOkB;AAAA;AAAA,oCAENlB,EAAE;AAAA;AAAA;AAAA;AAAA,sCAIAA,EAAE;AAAA,qCACHP,OAAO0B;AAAA;AAAA,sCAENnB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAanC,CAAC,EAAE+mG,GAAG,KAAK,CAAC,YAAY/mG,EAAE,CAAC,KAAK,cAAc,CAAC,KAAK,GAAG,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,YAAY,EAAEA,EAAE,aAAaP,EAAEO,EAAE,YAAYiB,EAAEjB,EAAE,YAAYkB,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEhC,EAAE,EAAEY,EAAE,QAAQ,MAAM,EAAE,EAAE,EAAEA,EAAE,QAAQ,IAAI8B,EAAErC,EAAE,EAAEO,EAAE,QAAQ,KAAK,KAAK,SAAS;AAAA,iCAC3OoB,MAAM,MAAMU;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAcb1C;AAAA,gDACgB6B;AAAA;AAAA,oCAEZjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKbZ;AAAA;AAAA,kCAES;AAAA,kDACgB8B;AAAA;AAAA,sCAEZlB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMb;AAAA;AAAA,oCAESP;AAAA,oDACgB0B;AAAA;AAAA,wCAEZnB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,6BAMbP;AAAA;AAAA,sCAESO,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAUnC,CAAC,EAAE,SAASgnG,GAAI9nG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,EAAE,gBAAgBC,EAAE,YAAYS,CAAC,EAAEtC,EAAEuC,EAAE23C,EAAE,wBAAwBt4C,CAAC,EAAEY,EAAE03C,EAAE,kBAAkBj6C,EAAE,MAAMqC,EAAE,EAAE,EAAEX,EAAEE,EAAE,GAAGU,CAAC,EAAEE,EAAE,IAAI2kG,GAAG5kG,CAAC,EAAE,OAAO5C,EAAE,gBAAgB6C,EAAE,CAACxC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIwnG,GAAG,CAAC,WAAW/qF,GAAG,YAAY,QAAQ,WAAW8qF,EAAG,EAAE,SAASE,GAAIhoG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,WAAWC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEuC,EAAE23C,EAAE,wBAAwBr4C,CAAC,EAAEW,EAAE03C,EAAE,kBAAkB,EAAE,EAAE,MAAMv4C,EAAE,EAAEC,EAAEU,EAAE,GAAGC,CAAC,EAAEE,EAAE,IAAI4kG,GAAG7kG,CAAC,EAAE,OAAO5C,EAAE,gBAAgB6C,EAAE,CAACxC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI0nG,GAAG,CAAC,WAAWhrF,GAAG,YAAY,QAAQ,WAAW+qF,EAAG,EAAE,SAASE,GAAIloG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,UAAUC,CAAC,EAAE5B,EAAE6B,EAAEq4C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAE2B,EAAED,CAAC,EAAEW,EAAE,IAAIskG,GAAG/kG,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI4nG,GAAG,CAAC,WAAWjrF,GAAG,YAAY,QAAQ,WAAWgrF,EAAG,EAAE,SAASE,GAAIpoG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,YAAYC,CAAC,EAAE5B,EAAE6B,EAAEq4C,EAAE,kBAAkBj6C,EAAE,MAAM2B,EAAE,EAAE,EAAED,CAAC,EAAEW,EAAE,IAAIglG,GAAGzlG,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI8nG,GAAG,CAAC,WAAWlrF,GAAG,YAAY,QAAQ,WAAWirF,EAAG,EAAE,SAASE,GAAItoG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,IAAI,EAAE,QAAQ0B,EAAE,WAAWC,CAAC,EAAE5B,EAAE6B,EAAEq4C,EAAE,kBAAkBt4C,EAAE,EAAE,MAAMD,EAAE,EAAE,CAAC,EAAEW,EAAE,IAAIilG,GAAG1lG,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIgoG,GAAG,CAAC,WAAWnrF,GAAG,YAAY,QAAQ,WAAWkrF,EAAG,EAAME,GAAI9J,GAAG;AAAA;AAAA,EAE74C+J,GAAI9J,GAAG,CAAC,UAAU6J,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAWrrF,GAAG,YAAY,QAAQ,WAAWorF,EAAG,EAAME,GAAI;AAAA;AAAA;AAAA,EAGtFC,GAAIjK,GAAG,CAAC,UAAUgK,EAAG,CAAC,EAAEE,GAAG,CAAC,WAAWvrF,GAAG,YAAY,QAAQ,WAAWsrF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYhoG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,QAAQ,QAAQ,QAAQ,EAAE,KAAK,YAAY,CAAC,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAE,CAAC,EAAEpB,EAAE,CAAC8B,CAAC,EAAE1C,EAAE,CAAC,EAAE4C,CAAC,EAAE,EAAE,KAAK,YAAY,CAACF,EAAE,EAAEE,EAAE,CAAC,EAAE,IAAIC,EAAExC,IAAI,WAAW,EAAE,EAAE,CAACyC,EAAEkB,CAAC,EAAE,CAAC,GAAGjC,EAAE,MAAM,GAAGC,EAAE,KAAK,EAAE,CAACiC,EAAEoD,EAAExC,CAAC,EAAE,EAAE,EAAE,CAAC,IAAI9C,EAAE,IAAI,EAAE,KAAK,yBAAyB,MAAMe,6BAA6B,EAAE,CAAC,MAAM,MAAM,mBAAmBA,GAAG,EAAE,CAAC2B,EAAEC,EAAEC,CAAC,EAAE/B,EAAE,EAAE,CAAC,IAAIZ,EAAE,IAAIY,EAAE,KAAK,wBAAwB,MAAMoB,4BAA4B,EAAE,CAAC,MAAM,MAAM,mBAAmBA,GAAG,EAAE,KAAK,SAAS;AAAA,yCAC1gBC;AAAA,wCACDQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,iCAgBP3C;AAAA;AAAA;AAAA;AAAA,+BAIFuF;AAAA,8BACD3C;AAAA;AAAA,uBAEPG;AAAA,mCACY/B;AAAA,4BACPjB;AAAA;AAAA;AAAA,uBAGL8C;AAAA,mCACYX;AAAA,4BACPnC;AAAA;AAAA;AAAA;AAAA;AAAA,aAKfgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAwBR,CAAC,EAAMgmG,GAAI/oG,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,SAAS0B,EAAE,OAAOC,EAAE,mBAAmBC,CAAC,EAAE7B,EAAEsC,EAAE,IAAIkmG,GAAGvoG,EAAE,MAAM,EAAE,MAAM0B,EAAEC,EAAEC,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,EAAE,EAAE,CAAC,EAAE,SAAS,CAAC,EAAEyoG,GAAG,CAAC,WAAWvrF,GAAG,YAAY,QAAQ,WAAWsrF,EAAG,EAAME,IAAI,SAASjpG,EAAE,CAACA,EAAE,KAAK,IAAIA,EAAE,IAAI,GAAG,GAAGipG,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,KAAK,CAAC,YAAYpoG,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,GAAGO,EAAE,KAAK,YAAYZ,EAAE,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,KAAK,OAAO,CAAC,EAAE,IAAI6B,EAAE,KAAK,YAAY,OAAOC,EAAE,KAAK,KAAKinG,GAAG,KAAK,MAAM,MAAMhnG,EAAE,EAAED,EAAE,QAAQmnG,GAAGpnG,EAAE,SAAS,KAAK,EAAE,KAAKG,EAAE,KAAK,YAAY,KAAK,YAAY,OAAO,GAAG,EAAE,GAAGU,EAAE,GAAG,GAAG,EAAErC,EAAE,UAAU2B,EAAE,IAAI,WAAWU,EAAErC,EAAE,UAAU,YAAY,EAAEA,EAAE,gBAAgB2B,IAAI,cAAcU,EAAErC,EAAE,aAAa,cAAc,KAAK,SAAS;AAAA;AAAA,UAE1uBm1F,GAAG3zF,CAAC;AAAA,oBACMqnG,GAAGrnG,EAAE,SAAS,KAAK,EAAE;AAAA,sBACnBE;AAAA;AAAA,cAER;AAAA,sBACQW;AAAA,YACVwmG,GAAGrnG,EAAE,SAAS,KAAK,EAAE;AAAA,gBACjB,KAAK,YAAYonG,GAAGpnG,EAAE,SAAS,KAAK,EAAE;AAAA;AAAA;AAAA;AAAA,KAIjD,CAAC,EAAE,SAASonG,GAAGnpG,EAAE,EAAEE,EAAE,CAAC,GAAGF,IAAI,EAAE,MAAM,GAAG,IAAI,GAAGA,IAAI,EAAE,MAAM,GAAG,QAAQ,MAAM,GAAGA,IAAI,EAAE,MAAM,GAAG,QAAQ,QAAQ,MAAM,GAAGA,IAAI,EAAE,MAAM,GAAG,QAAQ,QAAQ,QAAQ,MAAM,MAAM,IAAI,MAAM,cAAcE,cAAcF,wBAAwB,CAAC,CAAC,SAASopG,GAAGppG,EAAE,EAAEE,EAAE,CAAC,GAAGF,IAAI,EAAE,MAAM,GAAG,IAAI,GAAGA,IAAI,EAAE,MAAM,GAAG,MAAM,GAAGA,IAAI,EAAE,MAAM,GAAG,MAAM,GAAGA,IAAI,EAAE,MAAM,GAAG,MAAM,MAAM,IAAI,MAAM,cAAcE,cAAcF,wBAAwB,CAAC,CAAC,SAASqpG,GAAGrpG,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE,MAAM,OAAO0B,EAAEu4C,EAAE,mBAAmB,CAACl6C,CAAC,EAAE,CAAC,EAAE4B,EAAE,EAAED,GAAG,OAAOC,EAAEg+F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQhgG,EAAE,MAAM,CAAC,KAAK+B,CAAC,CAAC,CAAC,GAAG,IAAIE,EAAEq4C,EAAE,iBAAiB,EAAE,CAAC,EAAE,GAAG,GAAGr4C,IAAI,EAAE,EAAE,MAAM,IAAI,MAAM,mDAAmD,EAAE,MAAM,OAAO,kBAAkB7B,GAAG,EAAE,IAAIsC,EAAEV,EAAE,MAAMC,GAAGU,EAAEi7F,GAAG,CAAC,OAAO,CAAC,EAAE57F,CAAC,EAAE,QAAQhC,CAAC,CAAC,EAAE,QAAQ4C,EAAE,EAAEA,GAAG,KAAK,KAAK,KAAK,KAAKF,CAAC,CAAC,EAAE,EAAEE,IAAI,CAAC,IAAIC,EAAE,IAAImmG,GAAGlpG,EAAEkC,EAAE,MAAM,GAAG,CAAC,EAAEc,EAAE,CAAC,CAACF,CAAC,CAAC,EAAEoB,EAAErB,EAAEA,EAAE3C,EAAE,gBAAgB6C,EAAE,CAACF,CAAC,EAAEA,EAAE,MAAMG,CAAC,EAAE9C,EAAE,8BAA8BgE,CAAC,CAAC,CAAC,GAAG3D,EAAE,CAAC,IAAIuC,EAAE,IAAIomG,GAAGlpG,EAAEkC,EAAE,MAAM3B,EAAE,CAAC,EAAEwC,EAAEF,EAAEA,EAAE3C,EAAE,gBAAgB4C,EAAE,CAACD,CAAC,EAAEA,EAAE,KAAK,EAAE3C,EAAE,8BAA8B6C,CAAC,CAAC,CAAC,GAAGd,GAAG,KAAK,CAAC,IAAIa,EAAE03C,EAAE,uBAAuBv4C,CAAC,EAAEc,EAAEm9F,GAAG,CAAC,OAAO,CAAC,EAAEr9F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,KAAK4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8B2C,CAAC,EAAE3C,EAAE,8BAA8BgC,CAAC,EAAEa,CAAC,CAAC,OAAOF,CAAC,CAAC,SAASymG,GAAItpG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAE,OAAO+oG,GAAGJ,GAAG,KAAK1oG,EAAEL,EAAE,EAAE,EAAE+B,CAAC,CAAC,CAAC,IAAIsnG,GAAG,CAAC,WAAWhsF,GAAG,YAAY,QAAQ,WAAW+rF,EAAG,EAAE,SAASE,GAAIxpG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAE,OAAO+oG,GAAGJ,GAAG,IAAI1oG,EAAEL,EAAE,EAAE,EAAE+B,CAAC,CAAC,CAAC,IAAIwnG,GAAG,CAAC,WAAWjsF,GAAG,YAAY,QAAQ,WAAWgsF,EAAG,EAAE,SAASE,GAAI1pG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,aAAa0B,CAAC,EAAE3B,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,SAASK,EAAE,MAAM,EAAE4B,EAAEjC,EAAE,SAAS,EAAE,MAAM,EAAE0C,EAAEs1F,GAAGh2F,EAAEC,EAAE,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,OAAOjC,EAAE,eAAe,CAAC,CAAC,EAAE,EAAE,MAAM0C,CAAC,CAAC,SAASrC,EAAE,MAAM,SAAS,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,WAAWK,CAAC,EAAE4B,EAAEjC,EAAE,WAAW,CAAC,EAAE0C,EAAEu1F,GAAGj2F,EAAEC,EAAE,EAAEF,CAAC,EAAE,OAAO/B,EAAE,eAAe0C,EAAE,MAAM,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,MAAM,IAAI,MAAM,qEAAqErC,EAAE,MAAM,SAAS,CAAC,CAAC,IAAIopG,GAAG,CAAC,WAAWjsF,GAAG,YAAY,QAAQ,WAAWgsF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY9oG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYY,EAAE,KAAK,UAAUZ,EAAE,KAAK,WAAW,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA,gBAIvqE,KAAK,qBAAqB;AAAA,gBAC1B,KAAK,oBAAoB;AAAA,gBACzB,KAAK,oBAAoB;AAAA;AAAA,uBAElBA;AAAA,+BACQA;AAAA,uBACRA;AAAA,+BACQA;AAAA,mCACIA;AAAA,UACzB,KAAK,mBAAmB;AAAA;AAAA;AAAA,uBAGX,KAAK,uBAAuB;AAAA;AAAA;AAAA,GAGhD,CAAC,sBAAsB,CAAC,OAAO,KAAK,aAAa,OAAO,YAAY,WAAW,CAAC,qBAAqB,CAAC,OAAO,KAAK,aAAa,OAAO,YAAY,WAAW,CAAC,qBAAqB,CAAC,OAAO,KAAK,aAAa,OAAO,YAAY,WAAW,CAAC,oBAAoB,CAAC,OAAO,KAAK,aAAa,OAAO,KAAK,YAAY,GAAG,KAAK,YAAY,EAAE,CAAC,wBAAwB,CAAC,OAAO,KAAK,aAAa,OAAO,4BAA4B,2BAA2B,CAAC,EAAE,SAAS2pG,GAAI7pG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,UAAU,EAAE,WAAW,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,GAAG2B,EAAE,IAAI,OAAO3B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG4B,EAAE,IAAI,OAAO5B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGqC,EAAE,IAAI,OAAOrC,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGsC,EAAEX,EAAE,EAAEY,EAAEX,EAAE,EAAEY,EAAEH,GAAG,EAAE,GAAGI,EAAE,IAAI,OAAO,CAACf,EAAEY,EAAEC,EAAEC,CAAC,EAAE,CAACd,EAAEc,EAAEF,EAAEC,CAAC,EAAEoB,EAAE,IAAI0lG,GAAG5mG,EAAE,EAAE,CAAC,EAAE,OAAO9C,EAAE,gBAAgBgE,EAAE,CAAC3D,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAIupG,GAAG,CAAC,WAAWnsF,GAAG,YAAY,QAAQ,WAAWksF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYjpG,EAAEZ,EAAE,GAAG,EAAE,KAAKK,EAAE,GAAGwB,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,OAAO,KAAK,OAAO,EAAE,CAAC,KAAK,UAAU,KAAK,OAAO,EAAE,CAAC,KAAK,YAAY,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYjB,EAAE,SAAS,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIh0F,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,YAAYA,EAAE,WAAW,EAAE,GAAG8B,EAAE,GAAG,IAAIrC,EAAE,EAAE;AAAA;AAAA,YAEzqC;AAAA,WACDwB,EAAE,EAAE;AAAA;AAAA,YAEH;AAAA,WACD,EAAE;AAAA;AAAA,cAEC;AAAA;AAAA,UAEJa,EAAE,gCAAgC,IAAI,EAAE1C,EAAE,kCAAkC,GAAGA,GAAG,KAAK,cAAc,KAAK,MAAM,EAAEK,GAAG,KAAK,cAAc,KAAK,wBAAwB,EAAEwB,GAAG,KAAK,cAAc,KAAK,gBAAgB,EAAE,KAAK,SAAS;AAAA,QACpO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOgBG;AAAA,4BACIA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCASIF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAOEC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAcxB;AAAA,UACAW;AAAA;AAAA;AAAA,KAGL,CAAC,EAAMonG,GAAG,KAAK,CAAC,YAAYlpG,EAAEZ,EAAE,GAAG,EAAE,KAAKK,EAAE,GAAGwB,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,OAAO,KAAK,OAAO,EAAE,CAAC,KAAK,UAAU,KAAK,OAAO,EAAE,CAAC,KAAK,YAAY,KAAK,OAAO,EAAE,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYjB,EAAE,SAAS,KAAK,oBAAoBk1F,GAAG,KAAK,YAAY,MAAM,EAAE,IAAIh0F,EAAElB,EAAE,YAAYA,EAAE,WAAWmB,EAAEnB,EAAE,QAAQ,KAAKoB,EAAEpB,EAAE,YAAY,EAAEA,EAAE,cAAc8B,EAAE9B,EAAE,aAAa,EAAEA,EAAE,YAAYgC,EAAE,EAAEC,EAAE;AAAA;AAAA,+CAE1a,QAAQwE,EAAE,EAAEA,EAAE,EAAEA,IAAIxE,GAAG;AAAA,wBAC9CwE,EAAE;AAAA,uBACHA,EAAE;AAAA,wBACDA,EAAE,EAAE;AAAA,uBACLA,EAAE,EAAE;AAAA,mBACRA,KAAKxE,GAAG;AAAA,0BACDH;AAAA,QAClB,QAAQ2E,EAAE,EAAEA,EAAE,EAAEA,IAAIxE,GAAG;AAAA,mBACZwE,EAAE;AAAA,mBACFA,EAAE;AAAA,mBACFA,EAAE,EAAE;AAAA,mBACJA,EAAE,EAAE;AAAA,cACTA,iBAAiBxE,GAAG;AAAA;AAAA;AAAA,QAG1B,QAAQwE,EAAE,EAAEA,GAAGzE,EAAE,GAAG,EAAEyE,IAAI,CAAC,IAAIxC,EAAEwC,EAAE,EAAE,GAAGxE,GAAG;AAAA,4BACvBgC,EAAE;AAAA,YAClB7C,IAAI,GAAG,GAAG6C,EAAE,IAAI9C,EAAE,IAAI,GAAGc,GAAG;AAAA;AAAA,sEAE8BgC;AAAA,2BAC3CA;AAAA;AAAA;AAAA;AAAA;AAAA,6BAKEA;AAAA;AAAA,2BAEFA;AAAA;AAAA,gBAEX,IAAI,GAAGA,EAAE,EAAEhC,GAAG;AAAA,oBACVgC,mBAAmBA,EAAE,gBAAgBA;AAAA,kBACvChC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAYGgC,gCAAgCA;AAAA;AAAA,wBAEhCA,6BAA6BA;AAAA;AAAA,qBAEhChC,GAAG;AAAA,0DACkCgC;AAAA,2BAC/BA;AAAA;AAAA,6BAEEA;AAAA;AAAA,2BAEFA;AAAA;AAAA;AAAA,oBAGPA,cAAcA;AAAA,kBAChBA,EAAE,EAAE,GAAG,CAAC,IAAIJ,EAAE1C,EAAE,IAAI,EAAE6kB,EAAE,kBAAkB,CAAC,EAAE,EAAE,EAAE,IAAI,GAAG7kB,EAAE,IAAI,GAAG,EAAE,IAAI,GAAGA,EAAE,IAAI,GAAGc,GAAG;AAAA,uDACjD4B;AAAA;AAAA,wEAEiBI,EAAE;AAAA,6BAC7CA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,+BAKAA,EAAE;AAAA;AAAA,6BAEJA,EAAE;AAAA;AAAA,oBAEX,EAAE,EAAEhC,GAAG;AAAA;AAAA;AAAA;AAAA,yBAIFgC,EAAE,gCAAgCA,EAAE;AAAA;AAAA,yBAEpCA,EAAE,6BAA6BA,EAAE;AAAA;AAAA,sBAEpChC,GAAG;AAAA,wBACDgC,EAAE,mBAAmBA,gBAAgBA,EAAE;AAAA,uBACxCJ,IAAI,EAAE5B,GAAG;AAAA,wBACRgC,EAAE,cAAcA;AAAA,sBAClBhC,GAAG;AAAA,sCACa4B;AAAA;AAAA,0EAEoCI,EAAE;AAAA,+BAC7CA,EAAE;AAAA;AAAA,iCAEAA,EAAE;AAAA;AAAA,+BAEJA,EAAE;AAAA;AAAA;AAAA,wBAGTA,EAAE,cAAcA,EAAE;AAAA,qBACrB,OAAOA,EAAE,IAAI9C,EAAE,IAAI,GAAGc,GAAG;AAAA;AAAA,qEAEuBgC;AAAA,2BAC1CA;AAAA;AAAA;AAAA;AAAA,6BAIEA;AAAA;AAAA,2BAEFA;AAAA;AAAA;AAAA,iEAGsCA,EAAE;AAAA,2BACxCA,EAAE;AAAA;AAAA;AAAA;AAAA,6BAIAA,EAAE;AAAA;AAAA,2BAEJA,EAAE;AAAA;AAAA;AAAA,oBAGTA,mBAAmBA,gBAAgBA,EAAE;AAAA,gBACzCA,EAAE,EAAE,IAAIhC,GAAG;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sBAMLgC,EAAE,mBAAmBA,EAAE;AAAA,qBACxBhC,GAAG;AAAA,yDACiCgC;AAAA,2BAC9BA;AAAA;AAAA,6BAEEA;AAAA;AAAA,2BAEFA;AAAA;AAAA;AAAA;AAAA,qEAI0CA,EAAE;AAAA,2BAC5CA,EAAE;AAAA;AAAA,6BAEAA,EAAE;AAAA;AAAA,2BAEJA,EAAE;AAAA;AAAA;AAAA,oBAGTA;AAAA,2BACOA,gBAAgBA,EAAE;AAAA,gBAC7BA,EAAE,EAAE,IAAIhC,GAAG;AAAA,sBACLgC,EAAE,mBAAmBA,gBAAgBA,EAAE;AAAA,qBACxCA,EAAE,IAAIhC,GAAG;AAAA,+BACCgC;AAAA,2BACJA;AAAA,YACfA,EAAE,EAAE,IAAIhC,GAAG;AAAA,iCACUgC,EAAE;AAAA,6BACNA,EAAE;AAAA,eAChB,CAAChC,GAAG;AAAA;AAAA,IAEfA,GAAG;AAAA;AAAA,MAED,IAAIC,EAAE,GAAGkB,EAAE,GAAG,IAAI3D,EAAEyC,EAAE;AAAA;AAAA,YAEhB;AAAA,WACDjB,EAAEiB,EAAE;AAAA;AAAA,YAEH;AAAA,WACDA,EAAE;AAAA,YACD;AAAA,WACDkB,EAAE,gCAAgC,IAAIC,EAAEjE,EAAE,kCAAkC,GAAGA,GAAG,KAAK,cAAc,KAAK,MAAM,EAAEK,GAAG,KAAK,cAAc,KAAK,wBAAwB,EAAEwB,GAAG,KAAK,cAAc,KAAK,gBAAgB,EAAE,KAAK,SAAS;AAAA,QACrOiB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOgBhB;AAAA,4BACIA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAOlBe;AAAA;AAAA;AAAA,UAGAoB;AAAA,UACAD;AAAA;AAAA;AAAA,KAGL,CAAC,EAAE,SAAS+lG,GAAIjqG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,UAAUC,EAAE,gBAAgBC,CAAC,EAAE7B,EAAEsC,EAAEV,EAAEU,GAAG,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAGkkB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAE53C,CAAC,EAAE,IAAI,gFAAgF,oBAAoBA,IAAI,EAAE,IAAIC,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM,EAAEqC,EAAEX,EAAEE,EAAE,EAAE,EAAEW,EAAE2X,EAAE,EAAE,QAAQ,0BAA0B,GAAG5X,EAAE,aAAa,GAAGA,EAAE,YAAYA,EAAE,aAAa,EAAEC,EAAE,IAAIknG,GAAGnnG,CAAC,EAAEC,EAAE,IAAIinG,GAAGlnG,CAAC,EAAE,IAAIE,EAAE,CAAC,CAACF,EAAE,QAAQ,IAAIA,EAAE,QAAQ,IAAI,EAAE,CAACA,EAAE,aAAaA,EAAE,WAAW,EAAE,CAACA,EAAE,eAAeA,EAAE,aAAa,EAAE,CAACA,EAAE,SAASA,EAAE,OAAO,CAAC,EAAE,OAAO3C,EAAE,gBAAgB4C,EAAE,CAACvC,EAAE,CAAC,EAAE,UAAUwC,CAAC,CAAC,CAAC,IAAImnG,GAAG,CAAC,WAAWtsF,GAAG,YAAY,QAAQ,WAAWqsF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAYrpG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,IAAI,EAAE,KAAK,YAAYA,EAAE,YAAY,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,QAAQ,IAAIiB,EAAEjB,EAAE,QAAQ,KAAKkB,EAAElB,EAAE,YAAYA,EAAE,WAAW,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOr3BkB;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKMlB,EAAE;AAAA,kCACEA,EAAE;AAAA,iCACHZ,OAAOK;AAAA;AAAA,kCAENO,EAAE;AAAA;AAAA;AAAA;AAAA,oCAIAA,EAAE;AAAA,mCACH,OAAOiB;AAAA;AAAA,oCAENjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAYjC,CAAC,EAAEspG,GAAG,KAAK,CAAC,YAAYtpG,EAAE,CAAC,KAAK,cAAc,CAAC,KAAK,GAAG,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,aAAaiB,EAAEjB,EAAE,YAAYkB,EAAE9B,EAAE,EAAEY,EAAE,QAAQ,IAAImB,EAAE,EAAE,EAAEnB,EAAE,QAAQ,KAAKoB,EAAEpB,EAAE,YAAYA,EAAE,WAAW,KAAK,SAAS;AAAA,iCAClNkB,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAYP/B;AAAA,gDACgBK;AAAA;AAAA,oCAEZO,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKbZ;AAAA;AAAA,kCAES;AAAA,kDACgB6B;AAAA;AAAA,sCAEZjB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMb;AAAA;AAAA;AAAA,oCAGSoB;AAAA,8BACNA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KASzB,CAAC,EAAE,SAASmoG,GAAIrqG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAU0B,EAAE,IAAIC,EAAE,gBAAgBC,EAAE,YAAYS,CAAC,EAAEtC,EAAEuC,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAMqC,EAAE,EAAEX,EAAEC,EAAEC,EAAE,EAAE,EAAEW,EAAE,IAAIqnG,GAAGtnG,CAAC,EAAE,OAAO3C,EAAE,gBAAgB4C,EAAE,CAACvC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI+pG,GAAG,CAAC,WAAWzsF,GAAG,YAAY,QAAQ,WAAWwsF,EAAG,EAAE,SAASE,GAAIvqG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAU0B,EAAE,IAAIC,EAAE,gBAAgBC,EAAE,WAAWS,CAAC,EAAEtC,EAAEuC,EAAE23C,EAAE,kBAAkB53C,EAAE,EAAE,MAAM,EAAEX,EAAEC,EAAEC,EAAE,EAAE,EAAEW,EAAE,IAAIsnG,GAAGvnG,CAAC,EAAE,OAAO3C,EAAE,gBAAgB4C,EAAE,CAACvC,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIiqG,GAAG,CAAC,WAAW1sF,GAAG,YAAY,QAAQ,WAAWysF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY3pG,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAACA,EAAEA,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAM3pB,CAAC,EAAE,SAAS4pG,GAAI1qG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAEC,EAAE,CAAC,GAAGD,EAAE,MAAM,GAAGA,EAAE,KAAK,EAAE,EAAEwmB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAE++F,GAAG,CAAC,OAAO,CAAC,EAAE/+F,CAAC,EAAE,QAAQJ,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE+B,EAAE,IAAIwoG,GAAG,CAAC,EAAEvoG,EAAEhC,EAAE,gBAAgB+B,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,EAAEE,EAAEk9F,GAAG,CAAC,OAAO,CAAC,EAAEn9F,CAAC,EAAE,QAAQhC,EAAE,MAAM,CAAC,MAAMK,CAAC,CAAC,CAAC,EAAE,OAAOL,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8BgC,CAAC,EAAEC,CAAC,CAAC,IAAIwoG,GAAG,CAAC,WAAW5sF,GAAG,YAAY,QAAQ,WAAW2sF,EAAG,EAAME,GAAG,KAAK,CAAC,YAAY9pG,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,GAAG,EAAE,KAAK,YAAYA,EAAE,SAAS,GAAG,CAAC,SAASZ,EAAE,QAAQ,EAAE,QAAQK,EAAE,aAAawB,EAAE,YAAYC,EAAE,aAAaC,EAAE,YAAYC,EAAE,eAAe,EAAE,cAAcU,CAAC,EAAE9B,EAAE,CAAC,IAAI,EAAE,KAAKgC,CAAC,EAAEvC,EAAE,KAAK,SAAS;AAAA,oCACxlBwB,MAAMC;AAAA,iCACT,MAAMc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAaTb;AAAA,iCACG;AAAA;AAAA,kCAEC/B;AAAA,kCACAgC;AAAA,qCACGU;AAAA;AAAA,sCAEC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAgBjC,CAAC,EAAE,SAASioG,GAAI7qG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,IAAI0B,EAAE,UAAUC,CAAC,EAAE5B,EAAE6B,EAAEq4C,EAAE,sBAAsBj6C,EAAE,MAAM,EAAE,MAAM,EAAE0B,EAAE,OAAOC,CAAC,EAAEU,EAAEC,EAAE,IAAI+nG,GAAGzoG,CAAC,EAAES,EAAE1C,EAAE,gBAAgB2C,EAAE,CAACtC,EAAE,CAAC,EAAE,SAAS,EAAE,IAAIuC,EAAEu8F,GAAG,CAAC,OAAO,CAAC,EAAEz8F,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,MAAMiC,EAAE,QAAQ,CAAC,CAAC,EAAE,OAAOjC,EAAE,8BAA8B0C,CAAC,EAAEE,CAAC,CAAC,IAAIgoG,GAAG,CAAC,WAAW9sF,GAAG,YAAY,QAAQ,WAAW6sF,EAAG,EAAE,SAASE,GAAI/qG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,SAASO,CAAC,EAAED,EAAE,EAAE,EAAE,CAAC,QAAQ,EAAE,WAAW2B,EAAE,OAAOC,CAAC,EAAEs4C,EAAE,qBAAqBj6C,EAAE,EAAE,MAAM,EAAEi6C,EAAE,oBAAoB,EAAE,OAAOt4C,EAAE,CAAC,EAAE,GAAG,CAAC,KAAKC,EAAE,MAAMS,CAAC,EAAE43C,EAAE,qBAAqBv4C,EAAEC,CAAC,EAAEW,EAAED,EAAE,OAAOE,EAAE,KAAKC,EAAE,EAAE,OAAOC,EAAE,CAAC,EAAE,QAAQkB,EAAE,EAAEA,EAAErB,EAAE,EAAEqB,EAAE,CAAC,QAAQ,KAAKtB,EAAEsB,GAAG,CAAC,GAAG,CAAC,mBAAmBqD,EAAE,WAAW,CAAC,EAAEizC,EAAE,qBAAqBz3C,EAAEb,EAAE,EAAE,EAAEyC,EAAE61C,EAAE,sBAAsBjzC,CAAC,EAAE5C,EAAE,EAAE,IAAIA,EAAEu7F,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,EAAE,EAAE,QAAQhgG,EAAE,MAAM,CAAC,KAAKqH,CAAC,CAAC,CAAC,EAAEvE,EAAE,KAAK2B,CAAC,GAAG,IAAIC,EAAED,EAAE,MAAM,MAAM,EAAE,QAAQE,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAED,EAAE,OAAO,EAAEC,GAAG,EAAE,CAAC,EAAEiiB,EAAE,YAAYniB,EAAE,MAAMC,CAAC,IAAID,EAAE06F,GAAG,CAAC,OAAO,CAAC16F,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,MAAM0E,CAAC,CAAC,CAAC,EAAE5B,EAAE,KAAK2B,CAAC,GAAG7B,IAAI,KAAKA,EAAE6B,GAAG7B,EAAEo8F,GAAG,CAAC,OAAO,CAAC,EAAEv6F,EAAE,EAAE7B,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE8C,EAAE,KAAKF,CAAC,EAAE,CAACoB,EAAErB,EAAE,IAAIV,EAAE+B,IAAI,IAAIpB,EAAEk9F,GAAG,CAAC,OAAO,CAAC,EAAEl9F,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,KAAKiC,EAAE+B,IAAI,EAAE,OAAOnB,GAAG,SAAS,EAAE,CAAC,CAAC,EAAEC,EAAE,KAAKF,CAAC,GAAGC,IAAI,CAAC,QAAQmB,KAAKlB,EAAEkB,IAAIpB,GAAG5C,EAAE,8BAA8BgE,CAAC,EAAE,OAAOpB,CAAC,CAAC,IAAIkoG,IAAG,CAAC,WAAW5sF,GAAG,YAAY,QAAQ,WAAW2sF,EAAG,EAAME,IAAI,0CAA0CC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS1zCC,IAAIxM,GAAG,CAAC,UAAUsM,IAAI,gBAAgBC,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW/sF,GAAG,YAAY,QAAQ,WAAW8sF,GAAG,EAAME,IAAI,yCAAyCC,IAAI;AAAA;AAAA;AAAA,EAGvJC,IAAIvrG,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,GAAGM,EAAE,EAAEC,CAAC,EAAE,EAAE,EAAEka,EAAE,EAAE,QAAQ,8BAA8B,EAAE,IAAIojF,GAAGyN,IAAIhrG,EAAE,MAAMC,EAAE,KAAK,EAAE,IAAIo9F,GAAG0N,IAAI/qG,EAAE,MAAMC,EAAE,KAAK,EAAE,OAAOL,EAAE,gBAAgB,EAAE,CAACI,EAAEC,CAAC,EAAED,EAAE,KAAK,CAAC,EAAEkrG,IAAG,CAAC,WAAWltF,GAAG,YAAY,QAAQ,WAAWitF,GAAG,EAAME,IAAI;AAAA;AAAA,EAE9PC,IAAI,wBAAwBC,IAAI/M,GAAG,CAAC,UAAU8M,IAAI,gBAAgBD,IAAI,MAAM,OAAO,cAAclT,EAAE,CAAC,EAAEqT,IAAG,CAAC,WAAWptF,GAAG,YAAY,QAAQ,WAAWmtF,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA;AAAA,cAIxJrxD,EAAE;AAAA,eACDA,EAAE;AAAA,eACFA,EAAE;AAAA,eACFA,EAAE;AAAA,eACFA,EAAE;AAAA,eACFA,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMfsxD,IAAInN,GAAG,CAAC,UAAUkN,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWxtF,GAAG,YAAY,QAAQ,WAAWutF,GAAG,EAAME,IAAItN,GAAG;AAAA;AAAA,EAEzFuN,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASJC,GAAGvN,GAAG,CAAC,UAAUqN,IAAI,gBAAgBC,IAAI,cAAczT,GAAG,MAAM,SAAS,CAAC,EAAE2T,IAAG,CAAC,WAAW1tF,GAAG,YAAY,QAAQ,WAAWytF,EAAE,EAAE,SAASE,GAAGpsG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,IAAIO,CAAC,EAAEL,EAAE,CAAC,MAAM,CAAC,EAAE,EAAE,EAAE,EAAE,MAAM,OAAO+B,EAAE,EAAE,MAAM,MAAM,EAAEC,EAAE3B,EAAE,OAAOA,EAAE,IAAIumB,EAAE,OAAO,EAAE,EAAE,IAAIvmB,EAAE,IAAI,iCAAiC,EAAE,EAAE,OAAO,IAAI,EAAE2B,EAAE,EAAE3B,EAAE,GAAG0B,EAAE,OAAOC,EAAE,EAAE,CAAC,EAAEm9F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ/+F,EAAE,MAAM,CAAC,MAAM2B,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIoqG,IAAG,CAAC,WAAW3tF,GAAG,YAAY,QAAQ,WAAW0tF,EAAE,EAAME,GAAG,uBAAuBC,IAAI5N,GAAG,CAAC,UAAU2N,GAAG,gBAAgBA,GAAG,cAAc7T,EAAE,CAAC,EAAE+T,IAAG,CAAC,WAAW7tF,GAAG,YAAY,QAAQ,WAAW4tF,GAAG,EAAME,GAAG,KAAK,CAAC,YAAY3rG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,OAAO,MAAM,EAAE,IAAIK,EAAEL,EAAE,GAAG,KAAK,YAAYA,EAAE,IAAI6B,EAAE,EAAE,SAAS,KAAK,KAAK,UAAU,KAAK,KAAKC,EAAE,EAAE,GAAGzB,MAAM,MAAM0B,EAAE,GAAGnB,IAAI,OAAOmB,EAAE,4CAA4CnB,IAAI,OAAOmB,EAAE,wCAAyC,OAAM,IAAI,MAAM,sDAAsDnB,IAAI,EAAE,KAAK,SAAS;AAAA,yCAC75BiB;AAAA;AAAA;AAAA,UAG/BE;AAAA;AAAA;AAAA;AAAA,kDAIwC1B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMpBA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yDAS2ByB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAUpD,CAAC,EAAE,SAAS0qG,GAAG1sG,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEJ,EAAE,QAAQ,IAAIF,EAAE,MAAM,EAAEO,EAAEumB,EAAE,cAAc9mB,EAAE,KAAK,EAAE,EAAEA,EAAE,MAAMA,EAAE,MAAM,OAAO,GAAG,EAAEO,EAAE,EAAE0B,EAAEo9F,GAAG,CAAC,OAAO,CAAC,EAAEr/F,CAAC,EAAE,QAAQE,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEgC,EAAED,EAAE,MAAME,EAAE,IAAIsqG,GAAG,OAAOvqG,EAAE,CAAC,EAAEU,EAAE,IAAI6pG,GAAG,OAAOvqG,EAAE,CAAC,EAAEW,EAAE,CAAC,CAAC,OAAOvC,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM4B,CAAC,EAAE,CAAC,OAAO5B,EAAE,mBAAmB,KAAK,OAAO,MAAMA,EAAE,mBAAmB,KAAK,MAAM,MAAM4B,CAAC,CAAC,EAAEY,EAAE5C,EAAE,gBAAgBiC,EAAEU,EAAE,SAAS,EAAEE,EAAE7C,EAAE,gBAAgB0C,EAAEC,EAAE,SAAS,EAAEG,EAAEg7F,GAAG,CAAC,OAAO,CAAC,KAAKl7F,EAAE,KAAKC,CAAC,EAAE,QAAQ7C,CAAC,CAAC,EAAEA,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAE,IAAImB,EAAEm7F,GAAG,CAAC,OAAO,CAAC,EAAEr8F,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,MAAMF,EAAE,KAAK,CAAC,CAAC,EAAE,OAAOE,EAAE,8BAA8B+B,CAAC,EAAE/B,EAAE,8BAA8B8C,CAAC,EAAEkB,CAAC,CAAC,SAASyoG,IAAI3sG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAE,OAAOosG,GAAGpsG,EAAE,GAAGJ,CAAC,CAAC,CAAC,IAAI0sG,IAAG,CAAC,WAAWhuF,GAAG,YAAY,QAAQ,WAAW+tF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY/rG,EAAEZ,EAAE,CAAC,KAAK,YAAY,CAAC,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,KAAK,OAAO,CAAC,EAAE,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAYY,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,KAKz+B,CAAC,EAAE,SAASgsG,GAAG9sG,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,MAAMC,CAAC,EAAEL,EAAE,CAAC,MAAM,CAAC,EAAEA,EAAE,GAAG,EAAE,GAAG4mB,EAAE,WAAWvmB,CAAC,EAAE,IAAI,SAAS,CAAC,IAAI,EAAEumB,EAAE,kBAAkB,EAAEA,EAAE,cAAcxmB,CAAC,CAAC,EAAE,OAAO,EAAE,KAAKC,CAAC,EAAE,EAAE,eAAeD,EAAE,EAAE,CAAC,CAAC,KAAK,CAAC,IAAI,EAAE,IAAIusG,IAAGvsG,EAAEC,CAAC,EAAE0B,EAAE,CAAC,CAAC1B,CAAC,CAAC,EAAE,OAAO,EAAE,gBAAgB,EAAE,CAAC,EAAE,EAAE0B,CAAC,CAAC,CAAC,CAAC,IAAI8qG,IAAG,CAAC,WAAWluF,GAAG,YAAY,QAAQ,WAAWiuF,EAAE,EAAME,IAAG,KAAK,CAAC,YAAYlsG,EAAE,CAAC,KAAK,cAAc,CAAC,OAAO,EAAE,KAAK,YAAY,CAAC,EAAE,IAAIZ,EAAEY,EAAE,GAAG,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,yBAK3aZ;AAAA;AAAA,uCAEcA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAOlC,CAAC,EAAM+sG,IAAG,CAAC,WAAWnuF,GAAG,YAAY,QAAQ,WAAW,CAAC,CAAC,OAAO9e,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,MAAME,CAAC,EAAEF,EAAEM,EAAE,EAAEC,EAAE,IAAIysG,IAAG9sG,EAAE,KAAK,EAAE,OAAOI,EAAE,gBAAgBC,EAAE,CAACL,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,EAAMgtG,GAAG,mBAAmBC,IAAIxO,GAAG,CAAC,UAAUuO,GAAG,gBAAgBA,GAAG,cAAcxU,EAAE,CAAC,EAAE0U,IAAG,CAAC,WAAWruF,GAAG,YAAY,QAAQ,WAAWouF,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUtTC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAqBJC,IAAI3O,GAAG,CAAC,UAAUyO,IAAI,gBAAgBC,IAAI,MAAM,OAAO,CAAC,EAAEE,IAAG,CAAC,WAAWxuF,GAAG,YAAY,QAAQ,WAAWuuF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY3sG,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,IAAIZ,EAAEmyF,GAAG,EAAE,CAAC,EAAE9xF,CAAC,EAAEO,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uDAMpKP,QAAQ;AAAA;AAAA,wBAEvCL,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAcrB,CAAC,EAAMwtG,IAAG,KAAK,CAAC,YAAY5sG,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,IAAIZ,EAAEmyF,GAAG,EAAE,CAAC,EAAE9xF,CAAC,EAAEO,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAe7HP,QAAQ;AAAA,4BACVL,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgBpBA,EAAE;AAAA;AAAA,KAEP,CAAC,EAAMytG,IAAG,CAAC,WAAW/nF,GAAG,YAAY,QAAQ,WAAWgoF,GAAG,EAAEC,GAAGC,GAAGrzF,EAAE,EAAE,QAAQ,uCAAuC,EAAE,SAASmzF,IAAI5tG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,EAAED,EAAE,EAAE,OAAO,kBAAkB,aAAaC,aAAa,iBAAiB0B,EAAE,OAAO,kBAAkB,aAAa1B,aAAa,iBAAiB,CAAC2B,EAAEC,CAAC,EAAE,EAAE,CAAC5B,EAAE,WAAWA,EAAE,WAAW,EAAE,CAACA,EAAE,MAAMA,EAAE,MAAM,EAAEqC,EAAE,CAACT,EAAED,CAAC,EAAEW,EAAE,CAACV,EAAED,EAAE,CAAC,EAAE,GAAGD,GAAG,EAAE,CAAC,IAAIiC,EAAEuW,EAAE,EAAE,QAAQ,uCAAuC,GAAGozF,IAAI,MAAM3pG,IAAI4pG,MAAMA,GAAG5pG,EAAE2pG,GAAG,SAAS,cAAc,QAAQ,EAAE,WAAW,KAAK,CAAC,mBAAmBC,EAAE,CAAC,GAAGD,GAAG,OAAO,MAAM3rG,EAAE2rG,GAAG,OAAO,OAAO1rG,EAAE0rG,GAAG,UAAUttG,EAAE,EAAE,EAAE2B,EAAEC,CAAC,EAAE5B,EAAEstG,GAAG,MAAM,CAAC,IAAI/qG,EAAE5C,EAAE,eAAe0C,EAAE,OAAO,EAAE1C,EAAE,QAAQ,IAAI4C,EAAE,MAAM,EAAE,MAAMmuF,GAAG,OAAO/wF,EAAE,MAAM,yBAAyBA,EAAE,WAAW4C,EAAE,MAAM,EAAEvC,CAAC,EAAE,IAAIwC,EAAE0X,EAAE,EAAE,QAAQ,YAAY,EAAE,IAAIizF,IAAG7qG,CAAC,EAAE,IAAI4qG,IAAG5qG,CAAC,EAAEG,EAAE9C,EAAE,gBAAgB6C,EAAE,CAACD,CAAC,EAAE,OAAO,EAAE,OAAO5C,EAAE,YAAY4C,EAAE,MAAM,EAAEE,CAAC,CAAC,SAAS+qG,IAAI/tG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,WAAWS,EAAE,UAAUC,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAE1C,EAAE4D,EAAEs2C,EAAE,wBAAwB53C,CAAC,EAAE,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEW,EAAEV,EAAEW,EAAE,GAAGoB,CAAC,EAAEqD,EAAE,EAAE,CAAC,EAAE5C,EAAE,GAAG,KAAKC,EAAE3C,GAAG,KAAK4C,EAAE9B,IAAI,YAAY+B,EAAE,IAAI,CAAC,IAAIE,EAAE,CAACzE,EAAE,CAAC,EAAEkH,EAAE,CAACxC,EAAEC,IAAI,CAAC,GAAGA,IAAI,QAAQD,EAAE,MAAM,SAAS,GAAGA,EAAE,MAAM,KAAK,EAAE,CAAC,IAAIE,EAAEk6F,GAAG,CAAC,OAAO,CAAC,EAAEp6F,CAAC,EAAE,QAAQ/E,EAAE,MAAM,CAAC,MAAM,CAAC+E,EAAE,MAAM,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAKE,CAAC,EAAEA,CAAC,CAAC,OAAOF,CAAC,EAAE,GAAGN,GAAGK,EAAE,KAAKyC,EAAE,EAAE7E,CAAC,CAAC,EAAEgC,GAAGI,EAAE,KAAKyC,EAAExF,EAAEW,CAAC,CAAC,EAAEiC,EAAE,CAAC,IAAII,EAAE/E,EAAE,eAAe,CAAC,EAAE,UAAU4mB,EAAE,kBAAkB9jB,EAAE,SAAS,CAAC,EAAEgC,EAAE,KAAKC,CAAC,EAAE,EAAE,KAAKA,CAAC,CAAC,CAAC,OAAOD,CAAC,EAAE,GAAG,EAAE,eAAe,GAAG,EAAE,cAAc,GAAG,EAAE,iBAAiB,GAAG,EAAE,gBAAgB,GAAG,EAAE,eAAe,GAAG,EAAE,cAAc,IAAI,EAAE,QAAQ,OAAO,QAAQ,EAAE,QAAQ,OAAO,SAASuC,EAAE+/F,GAAG,CAAC,EAAE/mG,EAAE,OAAO,EAAE,SAAS,EAAE,QAAQL,EAAE,KAAK,EAAE,WAAW6C,EAAE,uBAAuBd,EAAE,eAAee,CAAC,CAAC,UAAU,EAAE,aAAa,GAAGkB,IAAI,gBAAgBuW,EAAE,EAAE,QAAQ,gBAAgB,EAAE,CAAC,IAAIzV,EAAEjC,EAAE87F,GAAG97F,EAAE,EAAE,EAAE,KAAK0E,EAAE,IAAI0/F,GAAG,EAAExiG,EAAEK,EAAEJ,EAAEC,CAAC,EAAEI,EAAE,CAAC,CAAC,EAAE,QAAQ,IAAI,EAAE,QAAQ,IAAI,EAAE,CAAC,EAAE,aAAa,EAAE,WAAW,EAAE,CAAC,EAAE,eAAe,EAAE,aAAa,EAAE,CAAC,EAAE,SAAS,EAAE,OAAO,CAAC,EAAEC,EAAEJ,EAAE,EAAEyC,EAAErH,EAAE,gBAAgBuH,EAAEvC,EAAE,UAAUD,CAAC,CAAC,SAASwV,EAAE,EAAE,QAAQ,mBAAmB,EAAElT,EAAEggG,GAAG,CAAC,EAAEhnG,EAAE,OAAO,EAAE,SAAS,EAAE,QAAQL,EAAE,KAAK,EAAE,WAAW6C,EAAE,uBAAuBd,EAAE,eAAee,CAAC,CAAC,MAAM,CAAC,IAAIgC,EAAEjC,EAAE87F,GAAG97F,EAAE,EAAE,EAAE,KAAK0E,EAAE,IAAIw/F,GAAG,EAAEtiG,EAAEK,EAAEJ,EAAEC,CAAC,EAAEI,EAAEH,EAAE,EAAEyC,EAAErH,EAAE,gBAAgBuH,EAAExC,EAAE,SAAS,CAAC,CAAC,IAAIuC,EAAE63F,GAAG,CAAC,OAAO,CAAC,EAAA93F,CAAC,EAAE,QAAQrH,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC,EAAE,OAAO,EAAE,KAAKqH,CAAC,EAAE,EAAE,QAAQvC,GAAG9E,EAAE,8BAA8B8E,CAAC,CAAC,EAAEwC,CAAC,CAAC,IAAIwmG,IAAG,CAAC,WAAWjoF,GAAG,YAAY,QAAQ,WAAWgoF,GAAG,EAAE,SAASE,IAAIjuG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,UAAUS,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAEzC,EAAE0C,EAAE,CAAC,EAAEkB,EAAEtB,EAAEsB,GAAG,OAAOA,EAAE,CAAC,EAAE,CAAC,GAAG4iB,EAAE,OAAO0zB,EAAE,+BAA+Bt4C,EAAEgC,CAAC,EAAE,IAAI,gFAAgFhC,oBAAoBgC,IAAI,EAAE,IAAI,EAAEs2C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEgC,EAAE/B,EAAEU,EAAE,EAAE,EAAE0E,EAAEkT,EAAE,EAAE,QAAQ,0BAA0B,GAAG,EAAE,aAAa,GAAG,EAAE,YAAY,EAAE,aAAa,EAAE,EAAE3X,EAAE+7F,GAAG/7F,EAAEyE,CAAC,EAAE,KAAK5C,EAAE,CAACpE,EAAE,CAAC,EAAEqE,EAAE,GAAG,KAAKC,EAAE5C,GAAG,KAAK6C,EAAEhC,IAAI,YAAY,GAAG8B,GAAGD,EAAE,KAAK,CAAC,EAAEE,GAAGF,EAAE,KAAK1C,CAAC,EAAE6C,EAAE,CAAC,IAAIG,EAAE/E,EAAE,eAAe,CAAC,EAAE,UAAU4mB,EAAE,kBAAkB/jB,EAAE,SAAS,CAAC,EAAE4B,EAAE,KAAKM,CAAC,EAAEjC,EAAE,KAAKiC,CAAC,CAAC,CAAC,IAAIuC,EAAED,EAAEC,EAAE,IAAIwiG,GAAG,EAAEplG,EAAE,EAAEC,EAAEC,CAAC,EAAE0C,EAAE,IAAIuiG,GAAG,EAAEnlG,EAAE,EAAEC,EAAEC,CAAC,EAAE,IAAIE,EAAE,CAAC,CAAC,EAAE,QAAQ,IAAI,EAAE,QAAQ,IAAI,EAAE,CAAC,EAAE,aAAa,EAAE,WAAW,EAAE,CAAC,EAAE,eAAe,EAAE,aAAa,EAAE,CAAC,EAAE,SAAS,EAAE,OAAO,CAAC,EAAEyC,EAAEvH,EAAE,gBAAgBsH,EAAE7C,EAAE,UAAUK,CAAC,EAAE,OAAOhC,EAAE,QAAQiC,GAAG/E,EAAE,8BAA8B+E,CAAC,CAAC,EAAEwC,CAAC,CAAC,IAAIymG,IAAG,CAAC,WAAWloF,GAAG,YAAY,QAAQ,WAAWioF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYrtG,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,SAASO,EAAE,KAAK,QAAQZ,EAAE,KAAK,YAAYK,EAAE,KAAK,cAAc,CAAC,IAAI,SAAS,EAAE,KAAK,YAAY,EAAE,IAAIwB,EAAE2zF,GAAG,EAAE,MAAM,EAAE1zF,EAAE;AAAA,gBACznH,QAAQC,EAAE,EAAEA,EAAE,KAAK,SAASA,IAAID,GAAG;AAAA,gDACHC;AAAA;AAAA,sDAEM,KAAK,YAAYA;AAAA,oCACnC,KAAK,QAAQA,MAAM,KAAK,SAAS;AAAA;AAAA,YAEzDF;AAAA;AAAA;AAAA;AAAA,YAIAC;AAAA;AAAA;AAAA;AAAA,OAIL,CAAC,EAAE,SAASosG,IAAIpuG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,OAAOM,EAAE,QAAQC,CAAC,EAAE,EAAE,EAAEA,EAAE,MAAM,EAAE,EAAE,EAAE,OAAO,GAAG0B,EAAE6kB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,CAAC4B,EAAEC,EAAES,EAAEC,CAAC,EAAE23C,EAAE,mBAAmBl6C,EAAEC,CAAC,EAAEuC,EAAEu8F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,CAACiC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEY,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAE/+F,CAAC,EAAE,QAAQJ,EAAE,MAAM,CAAC,MAAM,CAAC4mB,EAAE,cAAcxmB,EAAE,KAAK,EAAEsC,EAAEA,CAAC,CAAC,CAAC,CAAC,EAAE,GAAG1C,EAAE,mBAAmB,CAACI,EAAEC,CAAC,CAAC,GAAGD,EAAE,QAAQ,SAAS,CAAC,IAAIiH,EAAErH,EAAE,SAASK,EAAE,MAAM,EAAE,EAAEL,EAAE,WAAWI,CAAC,EAAEqE,EAAEg0F,GAAGpxF,EAAE,EAAEjH,EAAE,MAAM6B,EAAE,EAAES,EAAEC,EAAEvC,EAAE,MAAM2B,CAAC,EAAE,OAAO/B,EAAE,eAAegC,EAAE5B,EAAE,MAAMqE,EAAE,MAAM,CAAC,CAAC,IAAI3B,EAAE,IAAImrG,IAAG,EAAEtrG,EAAE,CAACV,EAAES,CAAC,EAAEtC,EAAE,KAAK,EAAE4D,EAAEhE,EAAE,gBAAgB8C,EAAE,CAACD,EAAED,CAAC,EAAEC,EAAE,KAAK,EAAE,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAEn7F,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAE,OAAOhC,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8BgE,CAAC,EAAE,CAAC,CAAC,IAAImqG,IAAG,CAAC,WAAWlvF,GAAG,YAAY,QAAQ,WAAWivF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYxtG,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,SAAS,EAAE,KAAK,YAAYA,EAAE,KAAK,KAAKA,EAAE,OAAO,IAAI,EAAEw1F,GAAG,KAAK,IAAI,EAAEn1F,EAAEguG,IAAIztG,EAAE,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,UAE93B;AAAA;AAAA,oDAE0CA,EAAE;AAAA,oCAClBP;AAAA;AAAA,KAE/B,CAAC,EAAE,SAASguG,IAAIvuG,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,UAAU,UAAU,UAAU,SAAS,EAAEI,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,IAAIA,IAAI,EAAED,EAAE,KAAK,OAAO,EAAEA,EAAE,KAAK,GAAGJ,EAAEK,IAAI,EAAE,OAAOD,EAAE,KAAK,CAAC,CAAC,SAASkuG,GAAGxuG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU0B,CAAC,EAAE3B,EAAE4B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG,GAAGka,EAAE,EAAE,IAAI,OAAO,EAAE,CAAC,IAAI,EAAEva,EAAE,SAAS,EAAE,MAAM,EAAEyE,EAAEpE,EAAE,MAAM2B,GAAG,QAAQ0C,EAAE,EAAEA,EAAE,EAAE,OAAO,EAAEA,EAAE,CAAC,IAAIC,EAAE,EAAED,GAAGkiB,EAAE,OAAOjiB,GAAGF,EAAE,GAAGE,GAAG,EAAE,IAAI,6BAA6BA,mBAAmBF,EAAE,IAAI,CAAC,CAAC,CAAC,IAAIxC,EAAEq4C,EAAE,aAAa,yBAAyBj6C,EAAE,EAAE2B,EAAED,CAAC,EAAEW,EAAEkkB,EAAE,cAAc,EAAE,KAAK,EAAEjkB,EAAE,CAAC,EAAEC,EAAEu8F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,CAACiC,EAAE,UAAUA,EAAE,UAAUA,EAAE,QAAQA,EAAE,SAAS,CAAC,CAAC,CAAC,EAAEY,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAACiC,EAAE,UAAUS,EAAET,EAAE,SAAS,CAAC,CAAC,CAAC,EAAEU,EAAE,KAAKC,CAAC,EAAED,EAAE,KAAKE,CAAC,EAAE,IAAIC,EAAE,CAACb,EAAE,UAAUA,EAAE,UAAUS,EAAET,EAAE,UAAUA,EAAE,SAAS,EAAE,GAAGjC,EAAE,mBAAmB,CAACK,EAAE,CAAC,CAAC,GAAGA,EAAE,QAAQ,SAAS,CAAC,IAAI,EAAEL,EAAE,WAAW6C,CAAC,EAAE4B,EAAEzE,EAAE,WAAW4C,CAAC,EAAE8B,EAAEg0F,GAAGj0F,EAAE,EAAE3B,CAAC,EAAE,OAAOH,EAAE,QAAQgC,GAAG3E,EAAE,8BAA8B2E,CAAC,CAAC,EAAE3E,EAAE,eAAeiC,EAAE,YAAYyC,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAIV,EAAE,IAAIoqG,IAAGxrG,EAAE,MAAME,CAAC,EAAE,EAAE9C,EAAE,gBAAgBgE,EAAE,CAACpB,EAAEC,CAAC,EAAED,EAAE,KAAK,EAAED,EAAE,KAAK,CAAC,EAAE,IAAI0E,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAMiC,EAAE,WAAW,CAAC,CAAC,EAAE,OAAOU,EAAE,QAAQ,GAAG3C,EAAE,8BAA8B,CAAC,CAAC,EAAEqH,CAAC,CAAC,IAAIknG,IAAG,CAAC,WAAWvvF,GAAG,YAAY,QAAQ,WAAWsvF,EAAE,EAAME,IAAI,uBAAuBC,IAAI;AAAA;AAAA,EAE1wCC,IAAIhQ,GAAG,CAAC,UAAU8P,IAAI,gBAAgBC,IAAI,cAAc9V,GAAG,MAAM,MAAM,CAAC,EAAEgW,IAAG,CAAC,WAAWzvF,GAAG,YAAY,QAAQ,WAAWwvF,GAAG,EAAME,IAAI,wBAAwBC,IAAI;AAAA;AAAA,EAEpKC,IAAIpQ,GAAG,CAAC,UAAUkQ,IAAI,gBAAgBC,IAAI,MAAM,OAAO,cAAcjW,EAAE,CAAC,EAAEmW,IAAG,CAAC,WAAW5vF,GAAG,YAAY,QAAQ,WAAW2vF,GAAG,EAAE,SAASE,IAAIlvG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAE,OAAOosG,GAAGpsG,EAAE,GAAGJ,CAAC,CAAC,CAAC,IAAIivG,IAAG,CAAC,WAAW5vF,GAAG,YAAY,QAAQ,WAAW2vF,GAAG,EAAME,IAAI,wCAAwCC,IAAI1Q,GAAG,CAAC,UAAUyQ,IAAI,MAAM,MAAM,CAAC,EAAEE,IAAG,CAAC,WAAW7vF,GAAG,YAAY,QAAQ,WAAW4vF,GAAG,EAAME,IAAI,0BAA0BC,IAAI7Q,GAAG,CAAC,UAAU4Q,IAAI,MAAM,MAAM,CAAC,EAAEE,IAAG,CAAC,WAAW/vF,GAAG,YAAY,QAAQ,WAAW8vF,GAAG,EAAME,IAAI,0BAA0BC,IAAIhR,GAAG,CAAC,UAAU+Q,IAAI,MAAM,MAAM,CAAC,EAAEE,IAAG,CAAC,WAAWjwF,GAAG,YAAY,QAAQ,WAAWgwF,GAAG,EAAME,IAAI,uBAAuBC,IAAI;AAAA;AAAA,EAE1qBC,IAAInR,GAAG,CAAC,UAAUiR,IAAI,gBAAgBC,IAAI,cAAc/W,GAAG,MAAM,MAAM,CAAC,EAAEiX,IAAG,CAAC,WAAWnwF,GAAG,YAAY,QAAQ,WAAWkwF,GAAG,EAAME,IAAI,wBAAwBC,IAAI;AAAA;AAAA,EAEpKC,IAAIvR,GAAG,CAAC,UAAUqR,IAAI,gBAAgBC,IAAI,cAAclX,GAAG,MAAM,MAAM,CAAC,EAAEoX,IAAG,CAAC,WAAWtwF,GAAG,YAAY,QAAQ,WAAWqwF,GAAG,EAAE,SAASE,IAAIrwG,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,KAAKC,EAAE,IAAI,CAAC,EAAEL,EAAE,EAAE+4F,GAAG34F,EAAEC,EAAE,CAAC,EAAE,OAAO,EAAE,eAAe,CAAC,EAAE,MAAM,EAAE,UAAU,CAAC,CAAC,CAAC,IAAI+vG,IAAG,CAAC,WAAWvwF,GAAG,YAAY,QAAQ,WAAWswF,GAAG,EAAME,IAAI7R,GAAG;AAAA;AAAA,EAEnU8R,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQJC,IAAI9R,GAAG,CAAC,UAAU4R,IAAI,gBAAgBC,IAAI,cAActX,EAAE,CAAC,EAAEwX,IAAG,CAAC,WAAW1wF,GAAG,YAAY,QAAQ,WAAWywF,GAAG,EAAME,IAAIjS,GAAG;AAAA;AAAA,EAE9HkS,IAAIjS,GAAG,CAAC,UAAUgS,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW5wF,GAAG,YAAY,QAAQ,WAAW2wF,GAAG,EAAME,IAAI,sCAAsCC,IAAI;AAAA;AAAA;AAAA;AAAA,EAIhIC,IAAIpS,GAAG,CAAC,UAAUkS,IAAI,gBAAgBC,IAAI,MAAM,MAAM,CAAC,EAAEE,IAAG,CAAC,WAAW/wF,GAAG,YAAY,QAAQ,WAAW8wF,GAAG,EAAME,IAAI,6BAA6BC,IAAIxS,GAAG,CAAC,UAAUuS,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWjxF,GAAG,YAAY,QAAQ,WAAWgxF,GAAG,EAAME,IAAI,sCAAsCC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA,EAKpRC,IAAI3S,GAAG,CAAC,UAAUyS,IAAI,gBAAgBC,IAAI,MAAM,MAAM,CAAC,EAAEE,IAAG,CAAC,WAAWpxF,GAAG,YAAY,QAAQ,WAAWmxF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY3wG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,IAAIC,EAAE9B,EAAE+B,EAAEnB,EAAE,GAAG,EAAE,KAAK,YAAYA,EAAE,IAAIoB,EAAE,EAAE,SAAS,cAAc3B,WAAWwB,IAAI,GAAGG,EAAE,eAAe,KAAKH,IAAI,EAAEG,EAAE,QAAQ,KAAKA,EAAE,WAAW,eAAeH,OAAO,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAS3VC,WAAWA;AAAA;AAAA,oCAECC;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKVC;AAAA;AAAA;AAAA,KAGrB,CAAC,EAAMwvG,IAAG,KAAK,CAAC,YAAY5wG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,IAAIC,EAAE9B,EAAE+B,EAAEnB,EAAE,GAAG,EAAE,KAAK,YAAYA,EAAE,IAAIoB,EAAE,EAAE,SAAS,cAAc3B,WAAWwB,IAAI,GAAGG,EAAE,eAAe,KAAKH,IAAI,EAAEG,EAAE,QAAQ,KAAKA,EAAE,WAAW,eAAeH,OAAO,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAQpR,KAAK,YAAY;AAAA,gCACjB,KAAK,YAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,iCAehBC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAWRA,WAAWA;AAAA;AAAA;AAAA,6DAGyBC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAqBnBC;AAAA;AAAA;AAAA,KAGrC,CAAC,EAAMyvG,IAAI3xG,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,YAAY,EAAE,KAAK,EAAE,MAAM0B,EAAE,KAAKC,CAAC,EAAE5B,EAAE6B,EAAEsY,EAAE,EAAE,QAAQ,0BAA0B,EAAE,IAAIi3F,IAAGnxG,EAAE,MAAM,EAAE,EAAE0B,EAAEC,CAAC,EAAE,IAAIuvG,IAAGlxG,EAAE,MAAM,EAAE,EAAE0B,EAAEC,CAAC,EAAE,OAAOhC,EAAE,gBAAgBiC,EAAE,CAAC5B,CAAC,EAAEA,EAAE,KAAK,CAAC,EAAEqxG,IAAG,CAAC,WAAWpxF,GAAG,YAAY,QAAQ,WAAWmxF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY/wG,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,aAAa,cAAc,IAAI,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYjB,EAAE,KAAK,MAAMA,EAAE,GAAG,KAAK,YAAYZ,EAAE,KAAK,KAAK,EAAE,KAAK,MAAMK,EAAE,KAAK,KAAKwB,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAQvd,KAAK;AAAA,oDACiB7B;AAAA,yCACX,KAAK;AAAA,0BACpBA;AAAA;AAAA;AAAA,sCAGY,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAelBK,qBAAqB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yCAOLA;AAAA,0BACfwB;AAAA;AAAA;AAAA;AAAA,0CAIgBA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAcrC,CAAC,EAAM+vG,IAAI9xG,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,YAAY0B,EAAE,KAAKC,EAAE,MAAMC,EAAE,KAAKS,CAAC,EAAEtC,EAAEuC,EAAE,IAAIgvG,IAAGtxG,EAAE,MAAM0B,EAAEC,EAAEC,EAAES,CAAC,EAAE,OAAO1C,EAAE,gBAAgB2C,EAAE,CAACtC,EAAE,EAAE,CAAC,EAAEA,EAAE,KAAK,CAAC,EAAEwxG,IAAG,CAAC,WAAWtxF,GAAG,YAAY,QAAQ,WAAWqxF,GAAG,EAAE,SAASE,IAAGhyG,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEumB,EAAE,cAAc,CAAC,EAAE9kB,EAAE8kB,EAAE,cAAc9mB,EAAE,KAAK,EAAEO,EAAE0B,EAAEo9F,GAAG,CAAC,OAAO,CAAC,EAAEr/F,CAAC,EAAE,MAAM,CAAC,MAAM,CAACgC,EAAEzB,CAAC,CAAC,EAAE,QAAQD,CAAC,CAAC,EAAE4B,EAAEw9F,GAAGz9F,EAAEjC,EAAE,MAAM,MAAMM,CAAC,EAAE,EAAE++F,GAAG,CAAC,OAAO,CAAC,EAAEn9F,CAAC,EAAE,MAAM,CAAC,MAAMhC,CAAC,EAAE,QAAQI,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B2B,CAAC,EAAE3B,EAAE,8BAA8B4B,CAAC,EAAE,CAAC,CAAC,SAAS+vG,GAAGjyG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,iBAAiB,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE4B,EAAED,EAAEU,EAAE43C,EAAE,mBAAmBr4C,EAAEF,CAAC,EAAEY,EAAED,GAAG,KAAKE,EAAE5C,EAAE,mBAAmB,CAACK,CAAC,CAAC,EAAEwC,EAAExC,EAAE,GAAGsC,EAAE,CAAC,GAAGC,EAAE,CAAC,IAAI6B,EAAEzE,EAAE,QAAQ,IAAI6C,EAAE,MAAM,EAAE,OAAO6B,EAAE,IAAI,MAAM3C,CAAC,EAAE,QAAQuF,EAAE,EAAEA,EAAE5C,EAAE,OAAO4C,IAAI5C,EAAE4C,GAAGjH,EAAE,MAAMqC,EAAE4E,IAAI,IAAI3C,EAAEi2F,GAAGn2F,EAAEpE,EAAE,MAAMA,EAAE,MAAMqC,EAAEgC,CAAC,EAAE7B,EAAE7C,EAAE,eAAe0E,EAAErE,EAAE,KAAK,EAAE,IAAIuE,EAAE5E,EAAE,QAAQ,IAAI6C,EAAE,MAAM,EAAE+B,EAAE,OAAOD,CAAC,MAAM9B,EAAE+8F,GAAGv/F,EAAEqC,EAAE1C,CAAC,EAAEiC,EAAEq4C,EAAE,iBAAiBr4C,EAAE,OAAOF,CAAC,CAAC,CAACu4C,EAAE,2BAA2B,MAAMr4C,EAAEF,CAAC,EAAE,GAAG,CAACe,EAAEkB,CAAC,EAAEs2C,EAAE,0BAA0Bz3C,EAAE,MAAMZ,CAAC,EAAE,EAAEa,EAAE,IAAI,EAAEw3C,EAAE,qBAAqBx3C,EAAEd,CAAC,GAAG,IAAIqF,EAAE,GAAGzE,EAAE,CAAC,IAAI6B,EAAEzE,EAAE,QAAQ,IAAI6C,EAAE,MAAM,EAAE,OAAO6B,EAAEu0F,GAAGx0F,EAAEmiB,EAAE,cAAc5iB,CAAC,EAAE,EAAE3D,EAAE,KAAK,EAAEgH,EAAErH,EAAE,eAAe,EAAEK,EAAE,KAAK,EAAE,IAAIsE,EAAE3E,EAAE,QAAQ,IAAIqH,EAAE,MAAM,EAAE1C,EAAE,OAAOD,CAAC,MAAM2C,EAAEyqG,IAAGjvG,EAAEmB,EAAE,EAAEhE,CAAC,EAAE,OAAO2C,GAAG3C,EAAE,8BAA8B6C,CAAC,EAAEwE,CAAC,CAAC,IAAI2qG,IAAG,CAAC,WAAWxxF,GAAG,YAAY,QAAQ,WAAWuxF,EAAE,EAAME,IAAIzU,GAAG;AAAA;AAAA,EAE33C0U,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA,IAKFxU,GAAG;AAAA;AAAA,EAELyU,IAAIzT,GAAG,CAAC,UAAUuT,IAAI,gBAAgBC,IAAI,cAAchZ,EAAE,CAAC,EAAEkZ,IAAG,CAAC,WAAW3xF,GAAG,YAAY,QAAQ,WAAW0xF,GAAG,EAAE,SAASE,IAAIvyG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAEwtF,GAAGxtF,EAAE,SAAS,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,gBAAgBC,CAAC,EAAE5B,EAAE6B,EAAE,EAAE2kB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAEr4C,CAAC,EAAE,IAAI,wEAAwE,oBAAoBA,IAAI,EAAE,IAAIS,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAE4B,EAAEF,EAAEC,CAAC,EAAE,GAAGU,EAAE,cAAc,GAAGA,EAAE,eAAe,GAAGkkB,EAAE,YAAYlkB,EAAE,QAAQA,EAAE,QAAQ,EAAE,OAAOk7F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAI2C,EAAE,IAAIwgG,GAAGzgG,EAAE,MAAM,EAAE,EAAE,OAAO1C,EAAE,gBAAgB2C,EAAE,CAACtC,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAIiyG,IAAG,CAAC,WAAW5xF,GAAG,YAAY,QAAQ,WAAW2xF,GAAG,EAAE,SAASE,IAAIzyG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,QAAQ,EAAE,IAAI0B,EAAE,WAAWC,EAAE,gBAAgBC,CAAC,EAAE7B,EAAEsC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE23C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,EAAEqC,EAAEX,EAAEE,EAAED,CAAC,EAAEY,EAAE,IAAIwgG,GAAGzgG,EAAE,MAAM,EAAE,EAAE,OAAO3C,EAAE,gBAAgB4C,EAAE,CAACvC,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAImyG,IAAG,CAAC,WAAW5xF,GAAG,YAAY,QAAQ,WAAW2xF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY7xG,EAAE,CAAC,KAAK,cAAc,CAAC,KAAK,QAAQ,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,aAAa,EAAEA,EAAE,YAAYP,EAAEO,EAAE,eAAeiB,EAAEjB,EAAE,sBAAsBkB,EAAElB,EAAE,qBAAqBmB,EAAEF,EAAE,EAAEjB,EAAE,QAAQ,IAAIoB,EAAEF,EAAE,EAAElB,EAAE,QAAQ,KAAK,EAAEiB,EAAEC,EAAE,EAAE,KAAK,SAAS;AAAA,iCACjrCC,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAcPH;AAAA,kBACdxB;AAAA,gDAC8BL;AAAA;AAAA,oCAEZY,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKJkB;AAAA,kDACgB;AAAA;AAAA,sCAEZlB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAOR;AAAA;AAAA;AAAA;AAAA,qCAIKkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAQhC,CAAC,EAAE4wG,IAAG,KAAK,CAAC,YAAY9xG,EAAE,CAAC,KAAK,cAAc,CAAC,KAAK,QAAQ,EAAE,KAAK,YAAYA,EAAE,QAAQ,IAAIZ,EAAEY,EAAE,YAAY,EAAEA,EAAE,aAAaP,EAAEO,EAAE,YAAYiB,EAAEjB,EAAE,cAAckB,EAAElB,EAAE,eAAemB,EAAEnB,EAAE,cAAcoB,EAAEpB,EAAE,qBAAqB,EAAEA,EAAE,sBAAsB8B,EAAE9B,EAAE,qBAAqB,EAAEoB,EAAE,EAAEpB,EAAE,QAAQ,MAAMgC,EAAE,EAAE,EAAEhC,EAAE,QAAQ,IAAIiC,EAAEH,EAAE,EAAE9B,EAAE,QAAQ,KAAKkC,EAAEd,EAAE,EAAEU,EAAE,EAAE,KAAK,SAAS;AAAA,iCAC5U,MAAME,MAAMC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAiBbb;AAAA,mBACbH;AAAA,gDAC6B7B;AAAA;AAAA,oCAEZY,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKJ;AAAA,sBACZkB;AAAA,kDAC4B;AAAA;AAAA,sCAEZlB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAMJ8B;AAAA,wBACZX;AAAA,oDAC4B1B;AAAA;AAAA,wCAEZO,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAORkC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAMT,OAAOJ;AAAA,yBACPA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KASpB,CAAC,EAAE,SAASiwG,IAAI7yG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,WAAW0B,EAAE,QAAQC,EAAE,IAAIC,EAAE,gBAAgBS,CAAC,EAAEtC,EAAEuC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE03C,EAAE,kBAAkB,EAAE,MAAMv4C,EAAEC,EAAEW,EAAEV,EAAES,CAAC,EAAEG,EAAE,IAAIugG,GAAGxgG,EAAE,MAAM,EAAE,EAAEE,EAAE9C,EAAE,gBAAgB6C,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,EAAEmB,EAAE,IAAI0uG,IAAG9vG,CAAC,EAAE,EAAE5C,EAAE,gBAAgBgE,EAAE,CAAC3D,EAAEyC,CAAC,EAAE,EAAE,KAAK,EAAE,OAAO9C,EAAE,8BAA8B8C,CAAC,EAAE,CAAC,CAAC,IAAI8vG,IAAG,CAAC,WAAW/xF,GAAG,YAAY,QAAQ,WAAW8xF,GAAG,EAAE,SAASE,IAAI/yG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,EAAE0B,EAAE,EAAE8rF,GAAG,CAAC,EAAE,CAAC,EAAE,aAAa,EAAE,GAAG,CAAC,WAAW7rF,EAAE,QAAQC,EAAE,IAAIS,EAAE,gBAAgBC,CAAC,EAAEvC,EAAEwC,EAAE03C,EAAE,kBAAkBv4C,EAAE,MAAMC,EAAEC,EAAE,EAAES,EAAEC,CAAC,EAAEE,EAAE,GAAGC,EAAE,IAAIqgG,GAAGvgG,EAAE,MAAMC,CAAC,EAAEmB,EAAEhE,EAAE,gBAAgB8C,EAAE,CAACf,CAAC,EAAEA,EAAE,KAAK,EAAE,EAAE,IAAI0wG,IAAG7vG,CAAC,EAAEyE,EAAErH,EAAE,gBAAgB,EAAE,CAACK,EAAE2D,CAAC,EAAEjC,EAAE,KAAK,EAAE,OAAO/B,EAAE,8BAA8BgE,CAAC,EAAEqD,CAAC,CAAC,IAAIyrG,IAAG,CAAC,WAAWnyF,GAAG,YAAY,QAAQ,WAAWkyF,GAAG,EAAE,SAASE,IAAGjzG,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAE,IAAI8iG,GAAGnjG,EAAE,MAAM,EAAE,EAAE,EAAEI,EAAE,gBAAgBC,EAAE,CAACP,CAAC,EAAE,SAAS,EAAEO,EAAE,IAAI8iG,GAAGnjG,EAAE,MAAM,GAAG,GAAG,CAAC,EAAE,IAAI,EAAEI,EAAE,gBAAgBC,EAAE,CAACP,CAAC,EAAE,SAAS,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,IAAIkzG,IAAG,CAAC,WAAWlyF,GAAG,YAAY,QAAQ,WAAW,CAAC,CAAC,OAAOhhB,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAE,CAAC,WAAWO,EAAE,QAAQ,EAAE,IAAI,EAAE,oBAAoB0B,CAAC,EAAE,EAAEC,EAAEhC,EAAE4mB,EAAE,OAAOxmB,EAAE,MAAM,SAAS,EAAE,IAAI,uDAAuDA,EAAE,MAAM,SAAS,EAAE,IAAI6B,EAAE,CAAC,EAAE,CAAC,EAAE2kB,EAAE,OAAO0zB,EAAE,+BAA+B,EAAEr4C,CAAC,EAAE,IAAI,wEAAwE,oBAAoBA,IAAI,EAAE,IAAIS,EAAE43C,EAAE,kBAAkBl6C,EAAE,MAAMC,EAAE,EAAE4B,EAAE,CAAC,EAAE,CAACU,EAAEC,CAAC,EAAEmwG,IAAG3yG,EAAE2B,EAAEW,EAAEV,CAAC,EAAE,MAAM,CAACW,EAAEC,CAAC,CAAC,CAAC,EAAE,SAASqwG,IAAGnzG,EAAE,EAAEE,EAAEI,EAAE,CAAC,IAAIC,EAAEumB,EAAE,cAAc,CAAC,EAAE9kB,EAAE8kB,EAAE,cAAc9mB,EAAE,KAAK,EAAEO,EAAE0B,EAAEo9F,GAAG,CAAC,OAAO,CAAC,EAAEr/F,CAAC,EAAE,MAAM,CAAC,MAAM,CAACgC,EAAEzB,CAAC,CAAC,EAAE,QAAQD,CAAC,CAAC,EAAE4B,EAAEw9F,GAAGz9F,EAAE,UAAU,OAAO3B,CAAC,EAAE,EAAE++F,GAAG,CAAC,OAAO,CAAC,EAAEn9F,CAAC,EAAE,MAAM,CAAC,MAAMhC,CAAC,EAAE,QAAQI,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B2B,CAAC,EAAE3B,EAAE,8BAA8B4B,CAAC,EAAE,CAAC,CAAC,IAAIkxG,IAAG,CAAC,WAAWnyF,GAAG,YAAY,QAAQ,WAAW,CAAC,CAAC,OAAOjhB,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAE,CAAC,SAASO,EAAE,KAAK,CAAC,EAAE,EAAE,EAAEL,EAAE+B,EAAE3B,EAAE,MAAM,OAAO4B,EAAE4kB,EAAE,eAAe,EAAExmB,EAAE,KAAK,EAAE6B,EAAED,EAAEU,EAAE43C,EAAE,mBAAmBr4C,EAAEF,CAAC,EAAEY,EAAED,GAAG,KAAKE,EAAE,EAAE,mBAAmB,CAACxC,CAAC,CAAC,EAAEyC,EAAE,CAAC,EAAEC,EAAE1C,EAAE,GAAGuC,EAAE,CAAC,GAAGC,EAAE,CAAC,IAAI8B,EAAE,EAAE,QAAQ,IAAI5B,EAAE,MAAM,EAAE,OAAO6B,EAAE,IAAI,MAAM5C,CAAC,EAAE,QAAQ+C,EAAE,EAAEA,EAAEH,EAAE,OAAOG,IAAIH,EAAEG,GAAG1E,EAAE,MAAMsC,EAAEoC,IAAI,IAAIF,EAAEg2F,GAAGl2F,EAAEtE,EAAE,MAAMA,EAAE,MAAMsC,EAAEiC,CAAC,EAAE7B,EAAE,EAAE,eAAe6B,EAAEvE,EAAE,KAAK,EAAE,IAAIkH,EAAE,EAAE,QAAQ,IAAIxE,EAAE,MAAM,EAAEwE,EAAE,OAAO1C,CAAC,MAAM9B,EAAE88F,GAAGx/F,EAAEsC,EAAE,CAAC,EAAEG,EAAE,KAAKC,CAAC,EAAEb,EAAEq4C,EAAE,iBAAiBr4C,EAAE,OAAOF,CAAC,CAAC,CAACu4C,EAAE,2BAA2B,MAAMr4C,EAAEF,CAAC,EAAE,GAAG,CAACiC,EAAE,CAAC,EAAEs2C,EAAE,0BAA0Bx3C,EAAE,MAAMb,CAAC,EAAEoF,EAAErD,EAAE3D,IAAIgH,EAAEizC,EAAE,qBAAqBt2C,EAAEhC,CAAC,GAAG,IAAI,EAAEixG,IAAGnwG,EAAE,EAAEuE,EAAE,CAAC,EAAE,QAAQ5C,KAAK5B,EAAE,EAAE,8BAA8B4B,CAAC,EAAE,OAAO,CAAC,CAAC,EAAE,SAAS0uG,IAAIrzG,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE4B,EAAED,EAAEU,EAAE43C,EAAE,mBAAmBr4C,EAAEF,CAAC,EAAEY,EAAEtC,EAAEqC,GAAG,OAAOC,EAAEq9F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK0C,CAAC,CAAC,CAAC,EAAET,EAAEq4C,EAAE,iBAAiBr4C,EAAE,OAAO5B,EAAE,MAAM,MAAM,GAAGi6C,EAAE,2BAA2B,MAAMr4C,EAAEF,CAAC,EAAE,GAAG,CAACa,EAAEC,CAAC,EAAEy3C,EAAE,0BAA0B33C,EAAE,MAAMV,CAAC,EAAEa,EAAE8jB,EAAE,cAAc/jB,CAAC,EAAEmB,EAAEm7F,GAAG,CAAC,OAAO,CAAC,EAAEx8F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG8C,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE08F,GAAGx7F,EAAEA,EAAE,MAAM,MAAMhE,CAAC,EAAEqH,EAAE,GAAG,EAAE,CAAC,IAAI,EAAEizC,EAAE,qBAAqB13C,EAAEZ,CAAC,EAAEqF,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,MAAMqH,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,8BAA8BgE,CAAC,EAAEhE,EAAE,8BAA8B,CAAC,EAAE0C,GAAG,MAAM1C,EAAE,8BAA8B2C,CAAC,EAAE0E,CAAC,CAAC,IAAI+rG,IAAG,CAAC,WAAWpyF,GAAG,YAAY,QAAQ,WAAWmyF,GAAG,EAAME,IAAI7V,GAAG;AAAA;AAAA,EAE/qG8V,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA,IAKF5V,GAAG;AAAA;AAAA,EAEL6V,IAAI7U,GAAG,CAAC,UAAU2U,IAAI,gBAAgBC,IAAI,cAAcna,EAAE,CAAC,EAAEqa,IAAG,CAAC,WAAWvyF,GAAG,YAAY,QAAQ,WAAWsyF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY7yG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAYA,EAAE,IAAI,CAAC0C,EAAE,IAAIA,EAAE,GAAG9B,EAAE,GAAG8B,EAAE,EAAE,EAAE,IAAIrC,EAAEO,EAAE,OAAOiB,EAAE2zF,GAAGn1F,CAAC,EAAEyB,EAAE9B,EAAE,IAAI0C,GAAGA,EAAE,EAAE,EAAE,KAAK,GAAG,EAAEX,EAAE/B,EAAE,IAAI,CAAC0C,EAAE,IAAIA,EAAE,GAAG9B,EAAE,EAAE,EAAE,KAAK,GAAG,EAAEoB,EAAE,CAAC,YAAY,YAAY,YAAY,WAAW,EAAE,MAAM,EAAE3B,CAAC,EAAE,EAAE,IAAI,UAAU,EAAE,EAAE,GAAGA,IAAI,EAAE,CAAC,KAAK,SAAS;AAAA,sBACvYyB;AAAA,oBACFC;AAAA;AAAA;AAAA;AAAA;AAAA,wCAKoB;AAAA;AAAA,4CAEI;AAAA;AAAA;AAAA;AAAA,QAIpC,MAAM,CAAC,KAAK,SAAS;AAAA,QACrBF,aAAaA,KAAKC;AAAA,QAClBD,WAAWA,KAAKE;AAAA;AAAA;AAAA,UAGdF;AAAA,8BACoBxB;AAAA;AAAA,iDAEmB;AAAA;AAAA,qDAEI;AAAA;AAAA;AAAA,UAG3CwB;AAAA,yBACeG;AAAA;AAAA,KAEpB,CAAC,EAAM0xG,IAAG,KAAK,CAAC,YAAY9yG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAYA,EAAE,IAAI,CAAC8C,EAAEkB,IAAIlB,EAAE,GAAGlC,EAAEoD,GAAGlB,EAAE,EAAE,EAAE,IAAIzC,EAAEO,EAAE,OAAOiB,EAAE2zF,GAAGn1F,CAAC,EAAEyB,EAAE9B,EAAE,IAAI8C,GAAGA,EAAE,EAAE,EAAE,KAAK,GAAG,EAAEf,EAAE/B,EAAE,IAAI,CAAC8C,EAAEkB,IAAIlB,EAAE,GAAGlC,EAAEoD,EAAE,EAAE,KAAK,GAAG,EAAEhC,EAAE+4F,GAAG,KAAK16F,CAAC,EAAE,EAAE06F,GAAG,SAAS16F,CAAC,EAAEqC,EAAE,GAAGV,EAAE3B,EAAE,QAAQ,KAAK,YAAYA,EAAE,KAAK,EAAEA,IAAI,EAAE,SAAS,QAAQ,EAAE,MAAM,EAAE,EAAE,KAAK,KAAKuC,EAAE,IAAI,UAAU,EAAE,EAAEC,EAAE,GAAG,GAAGxC,IAAI,EAAE,CAAC,IAAIyC,EAAE;AAAA,UACpYjB;AAAA;AAAA,0CAEgCe;AAAA;AAAA,8CAEIA;AAAA;AAAA;AAAA,QAGtCC,EAAE;AAAA,UACAhB;AAAA,UACAiB;AAAA,sCAC4B,EAAE,KAAK,OAAO;AAAA,UAC1Cd,EAAE3B,EAAE;AAAA,aACDqC;AAAA,YACDI;AAAA,wCAC4B,EAAE,KAAK,OAAO;AAAA;AAAA,OAE/C,KAAK,CAAC,IAAIA,EAAE;AAAA,UACTjB;AAAA,UACAA,UAAUA;AAAA,UACVA,WAAWA;AAAA,UACXA;AAAA;AAAA,6CAEmCe;AAAA,kDACKA;AAAA;AAAA,QAE1CC,EAAE;AAAA,UACAhB;AAAA,UACAiB;AAAA,sCAC4B,EAAE,KAAK,OAAO;AAAA,UAC1Cd,EAAE3B,EAAE;AAAA,aACDqC;AAAA,YACDI;AAAA,wCAC4B,EAAE,KAAK,OAAO;AAAA;AAAA;AAAA,UAG5Cd,EAAE3B,EAAE;AAAA,aACD2B,EAAE3B,EAAE,QAAQ,KAAK,YAAYA,EAAE;AAAA,YAChCyC;AAAA,wCAC4B,EAAE,KAAK,OAAO;AAAA,YAC1Cd,EAAE3B,EAAE;AAAA,eACDqC;AAAA,cACDI;AAAA,0CAC4B,EAAE,KAAK,OAAO;AAAA;AAAA;AAAA,OAGjD,CAAC,KAAK,SAAS;AAAA,cACRjB,aAAaA,KAAKC;AAAA,cAClBD,WAAWA,KAAKE;AAAA;AAAA;AAAA,UAGpBF;AAAA;AAAA,UAEAgB;AAAA;AAAA;AAAA,KAGL,CAAC,EAAM8wG,IAAI,CAAC,CAAC,OAAO7zG,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAE,CAAC,SAASO,EAAE,KAAK,CAAC,EAAEL,EAAE,EAAEua,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAIm5F,IAAGtzG,EAAE,MAAMC,EAAE,CAAC,EAAE,IAAIozG,IAAGrzG,EAAE,MAAMC,EAAE,CAAC,EAAE,OAAO,EAAE,gBAAgB,EAAE,CAACD,CAAC,EAAEA,EAAE,KAAK,CAAC,EAAEwzG,IAAG,CAAC,WAAW1yF,GAAG,YAAY,QAAQ,WAAWyyF,GAAG,EAAME,IAAI;AAAA,qBACzPC,IAAI;AAAA;AAAA;AAAA,IAGrBpW,GAAG;AAAA;AAAA,EAELqW,IAAIrV,GAAG,CAAC,UAAUmV,IAAI,gBAAgBC,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW7yF,GAAG,YAAY,QAAQ,WAAW4yF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYrzG,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,OAAO,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,OAAO,KAAK,OAAO,CAAC,EAAE,KAAK,YAAY,CAACY,EAAE,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAQzNZ,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAUNA,EAAE;AAAA;AAAA,KAEvB,CAAC,EAAMk0G,IAAI;AAAA;AAAA;AAAA;AAAA,eAIDC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkBjBC,GAAG1V,GAAG,CAAC,UAAUwV,IAAI,gBAAgBC,IAAI,iBAAiB,EAAE,CAAC,EAAEE,IAAG,CAAC,WAAWp2F,GAAG,YAAY,QAAQ,WAAWm2F,EAAE,EAAME,GAAG,gBAAgBC,GAAG7V,GAAG,CAAC,UAAU4V,GAAG,gBAAgBA,GAAG,gBAAgB,GAAG,cAAc7Z,EAAE,CAAC,EAAE+Z,IAAG,CAAC,WAAW3vF,GAAG,YAAY,QAAQ,WAAW0vF,EAAE,EAAE,SAASE,GAAG30G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,IAAI,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAe,CAAC,CAAC,EAAEvmB,EAAE,KAAK,EAAE0B,EAAEgwG,GAAG,CAAC,OAAO,CAAC,EAAE1xG,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,iBAAiB,EAAE,SAAS,EAAE,CAAC,CAAC,EAAEgC,EAAEs4C,EAAE,qBAAqBv4C,EAAE,MAAM,CAAC,EAAEE,EAAEk9F,GAAG,CAAC,OAAO,CAAC,EAAEp9F,CAAC,EAAE,QAAQ/B,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEU,EAAE6xG,GAAG,CAAC,OAAO,CAAC,EAAEl0G,EAAE,EAAE4B,CAAC,EAAE,QAAQjC,CAAC,CAAC,EAAE2C,EAAEqpG,GAAG,CAAC,OAAO,CAAC,EAAEtpG,CAAC,EAAE,QAAQ1C,CAAC,CAAC,EAAE4C,EAAEk9F,GAAG,CAAC,OAAO,CAAC,EAAEn9F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,KAAK,EAAE,SAAS,EAAE,CAAC,CAAC,EAAE6C,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAEv8F,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEc,EAAEsxG,GAAG,CAAC,OAAO,CAAC,EAAEzxG,EAAE,EAAEE,CAAC,EAAE,QAAQ7C,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B+B,CAAC,EAAE/B,EAAE,8BAA8BiC,CAAC,EAAEjC,EAAE,8BAA8B0C,CAAC,EAAE1C,EAAE,8BAA8B2C,CAAC,EAAE3C,EAAE,8BAA8B4C,CAAC,EAAE5C,EAAE,8BAA8B6C,CAAC,EAAEC,CAAC,CAAC,IAAI4xG,IAAG,CAAC,WAAWzwF,GAAG,YAAY,QAAQ,WAAWwwF,EAAE,EAAE,SAASE,IAAI70G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,KAAK,EAAE,WAAW0B,CAAC,EAAE3B,EAAE4B,EAAED,EAAE1B,EAAEo0G,GAAG,CAAC,OAAO,CAAC,OAAOp0G,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,IAAIK,EAAE,MAAM,OAAO,CAAC,CAAC,CAAC,EAAE4B,EAAED,EAAE,MAAM,GAAGU,EAAEV,EAAE,MAAM,GAAGW,EAAE,IAAIsxG,IAAGhyG,EAAES,EAAE,CAAC,EAAEE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEC,EAAE7C,EAAE,gBAAgB2C,EAAE,CAACX,CAAC,EAAE,QAAQY,CAAC,EAAE,OAAOb,GAAG/B,EAAE,8BAA8BgC,CAAC,EAAEa,CAAC,CAAC,IAAI+xG,IAAG,CAAC,WAAWxzF,GAAG,YAAY,QAAQ,WAAWuzF,GAAG,EAAME,IAAIjZ,GAAG;AAAA;AAAA,EAEp2CkZ,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUJ,SAASC,IAAIj1G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGJ,EAAE,mBAAmB,CAACI,CAAC,CAAC,EAAE,CAAC,IAAI,EAAEJ,EAAE,QAAQ,IAAII,EAAE,MAAM,EAAE,CAAC,EAAE2B,CAAC,EAAEs3F,GAAG,EAAE,OAAOj5F,EAAE,MAAMA,EAAE,KAAK,EAAE,OAAOJ,EAAE,eAAe+B,EAAE3B,EAAE,MAAM,CAAC,CAAC,CAAC,IAAIC,EAAE,OAAOka,EAAE,EAAE,QAAQ,6BAA6B,EAAEla,EAAE,IAAIo8F,GAAGr8F,EAAE,MAAM00G,GAAG,EAAEz0G,EAAE,IAAIs7F,GAAGv7F,EAAE,MAAMy0G,GAAG,EAAE70G,EAAE,gBAAgBK,EAAE,CAACD,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAI40G,IAAG,CAAC,WAAW1zF,GAAG,YAAY,QAAQ,WAAWyzF,GAAG,EAAME,IAAI92D,GAAG,wBAAwB,SAAS+2D,IAAIp1G,EAAE,CAACw6C,EAAE,KAAK,+FAA+F,EAAE,GAAG,CAAC,OAAO,EAAE,QAAQt6C,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,CAAC,EAAE5B,EAAE6B,EAAEjC,EAAE,SAASK,EAAE,MAAM,EAAEqC,EAAE1C,EAAE,SAAS,EAAE,MAAM,EAAE,CAAC,gBAAgB2C,CAAC,EAAEsyG,IAAIhzG,EAAES,EAAE,EAAEX,EAAEC,CAAC,EAAE,OAAOhC,EAAE,eAAe,CAAC2C,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAWA,CAAC,CAAC,CAAC,CAAC,IAAIwyG,IAAG,CAAC,WAAW3zF,GAAG,YAAY,QAAQ,WAAW0zF,GAAG,EAAME,IAAIj3D,GAAG,wBAAwB,SAASk3D,IAAIv1G,EAAE,CAACw6C,EAAE,KAAK,+FAA+F,EAAE,GAAG,CAAC,OAAO,EAAE,QAAQt6C,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,EAAE,mBAAmBC,CAAC,EAAE7B,EAAEsC,EAAE1C,EAAE,SAASK,EAAE,MAAM,EAAEsC,EAAE3C,EAAE,SAAS,EAAE,MAAM,EAAE,CAAC,gBAAgB4C,EAAE,aAAaC,CAAC,EAAEuyG,IAAI1yG,EAAEC,EAAE,EAAEZ,EAAEC,EAAEC,CAAC,EAAE,MAAM,CAACjC,EAAE,eAAe,CAAC4C,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAWA,CAAC,CAAC,EAAE5C,EAAE,eAAe,CAAC,EAAE,QAAQ,IAAI,WAAW,CAAC6C,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIyyG,IAAG,CAAC,WAAW7zF,GAAG,YAAY,QAAQ,WAAW4zF,GAAG,EAAME,IAAIp3D,GAAG,wBAAwB,SAASq3D,IAAI11G,EAAE,CAACw6C,EAAE,KAAK,+FAA+F,EAAE,GAAG,CAAC,OAAO,EAAE,QAAQt6C,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,aAAa0B,EAAE,eAAeC,EAAE,aAAaC,CAAC,EAAE7B,EAAEsC,EAAE1C,EAAE,SAASK,EAAE,MAAM,EAAEsC,EAAE3C,EAAE,SAAS,EAAE,MAAM,EAAE4C,EAAE,EAAEC,EAAEd,EAAEe,EAAEd,EAAEgC,EAAE/B,EAAE,CAAC,gBAAgB,EAAE,eAAeoF,CAAC,EAAEkuG,IAAI7yG,EAAEC,EAAEC,EAAEC,EAAEC,EAAEkB,CAAC,EAAE,MAAM,CAAChE,EAAE,eAAe,CAAC,EAAE,MAAM,EAAE,QAAQ,IAAI,WAAW,CAAC,CAAC,EAAEA,EAAE,eAAe,CAACqH,EAAE,MAAM,EAAE,UAAU,IAAI,aAAaA,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIouG,IAAG,CAAC,WAAW/zF,GAAG,YAAY,QAAQ,WAAW8zF,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY90G,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,cAAc,CAAC,SAAS,EAAE,KAAK,YAAY,CAACO,EAAEZ,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA,8BAIl/DK,aAAa;AAAA;AAAA;AAAA,KAGtC,CAAC,EAAMs1G,IAAI71G,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ0B,EAAE,SAASC,CAAC,EAAE5B,EAAE6B,EAAE2kB,EAAE,cAAcvmB,EAAE,KAAK,EAAEqC,EAAE,IAAIgzG,IAAGzzG,EAAE,EAAEF,EAAEC,CAAC,EAAEW,EAAEw8F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,CAACiC,CAAC,CAAC,CAAC,CAAC,EAAEW,EAAE5C,EAAE,gBAAgB0C,EAAE,CAACC,CAAC,EAAE,CAAC,EAAE3C,EAAE,8BAA8B2C,CAAC,EAAE,IAAIE,EAAE,CAAC,GAAGxC,EAAE,MAAM,CAAC,EAAEyC,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAEv8F,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,MAAM6C,CAAC,CAAC,CAAC,EAAE,OAAO7C,EAAE,8BAA8B4C,CAAC,EAAEE,CAAC,EAAE8yG,IAAG,CAAC,WAAWh0F,GAAG,YAAY,QAAQ,WAAW+zF,GAAG,EAAE,SAASE,GAAG/1G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAIC,EAAEglG,GAAG,CAAC,OAAO,CAAC,MAAMjlG,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE,EAAE61G,GAAG,CAAC,OAAO,CAAC,EAAEx1G,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,EAAEymG,GAAG,CAAC,OAAO,CAAC,MAAMrmG,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE+B,EAAE8zG,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ71G,CAAC,CAAC,EAAEgC,EAAE87F,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK/7F,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BK,CAAC,EAAEL,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B+B,CAAC,EAAEC,CAAC,KAAM,QAAO4qG,GAAG,CAAC,MAAM,CAAC,MAAMxsG,EAAE,MAAM,MAAMA,EAAE,MAAM,MAAMA,EAAE,QAAQ,SAAS,GAAG,CAAC,EAAE,QAAQJ,CAAC,CAAC,CAAC,CAAC,IAAI81G,IAAG,CAAC,WAAWtwF,GAAG,YAAY,QAAQ,WAAWqwF,EAAE,EAAE,SAASE,GAAGj2G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,GAAGA,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,8CAA8C,EAAE,GAAGA,EAAE,QAAQ,YAAY,CAAC,IAAIC,EAAEglG,GAAG,CAAC,OAAO,CAAC,MAAMjlG,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE,EAAE+1G,GAAG,CAAC,OAAO,CAAC,EAAE11G,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,EAAEymG,GAAG,CAAC,OAAO,CAAC,MAAMrmG,CAAC,EAAE,QAAQJ,CAAC,CAAC,EAAE+B,EAAE8zG,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ71G,CAAC,CAAC,EAAEgC,EAAE87F,GAAG,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK/7F,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8BK,CAAC,EAAEL,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B,CAAC,EAAEA,EAAE,8BAA8B+B,CAAC,EAAEC,CAAC,KAAM,QAAO4qG,GAAG,CAAC,MAAM,CAAC,MAAMxsG,EAAE,MAAM,MAAMA,EAAE,MAAM,MAAM,CAAC,EAAE,QAAQJ,CAAC,CAAC,CAAC,CAAC,IAAIg2G,IAAG,CAAC,WAAWr0F,GAAG,YAAY,QAAQ,WAAWo0F,EAAE,EAAE,SAASE,IAAIn2G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,GAAG,EAAE,SAAS,EAAE,OAAO8rG,GAAG,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,EAAE,QAAQlsG,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,GAAG,MAAM,EAAE,EAAE,GAAG,MAAM,EAAE,QAAQqC,GAAG,CAACkkB,EAAE,kBAAkB,EAAElkB,EAAE,MAAM,uDAAuD,EAAEkkB,EAAE,OAAO,IAAIlkB,EAAE,MAAM,IAAI,uDAAuD,CAAC,CAAC,EAAE,IAAIX,EAAE,CAAC,EAAEC,EAAE,EAAE,IAAIU,GAAG,CAAC,IAAIC,EAAEupG,GAAG,CAAC,OAAO,CAAC,MAAMxpG,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,KAAKY,CAAC,EAAEA,CAAC,CAAC,EAAEV,EAAE4kG,GAAG,CAAC,OAAO7kG,EAAE,QAAQhC,EAAE,MAAM,CAAC,KAAKK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,QAAQW,GAAG1C,EAAE,8BAA8B0C,CAAC,CAAC,EAAET,CAAC,CAAC,IAAIi0G,IAAG,CAAC,WAAWr0F,GAAG,YAAY,QAAQ,WAAWo0F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYv1G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYA,EAAE,IAAI,CAAC,EAAE0C,IAAI,EAAE,GAAG9B,EAAE8B,GAAG,EAAE,EAAE,EAAE,IAAIrC,EAAEO,EAAE,OAAOiB,EAAE2zF,GAAGn1F,CAAC,EAAEyB,EAAE9B,EAAE,IAAI,GAAG,EAAE,EAAE,EAAE,KAAK,GAAG,EAAE+B,EAAE/B,EAAE,IAAI,CAAC,EAAE0C,IAAI,EAAE,GAAG9B,EAAE8B,EAAE,EAAE,KAAK,GAAG,EAAEV,EAAE,CAAC,YAAY,YAAY,YAAY,WAAW,EAAE,MAAM,EAAE3B,CAAC,EAAE,GAAGA,IAAI,EAAE,CAAC,KAAK,SAAS;AAAA,sBAC58EyB;AAAA,oBACFC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAUZ,MAAM,CAAC,KAAK,SAAS;AAAA,QACrBF,aAAaA,KAAKC;AAAA,QAClBD,WAAWA,KAAKE;AAAA;AAAA;AAAA,UAGdF;AAAA;AAAA;AAAA;AAAA,YAIEA;AAAA,2BACeG;AAAA;AAAA;AAAA,KAGtB,CAAC,EAAMo0G,IAAG,KAAK,CAAC,YAAYx1G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,eAAe,CAAC,CAAC,KAAK,QAAQ,KAAK,OAAO,CAAC,EAAE,KAAK,YAAYA,EAAE,IAAI,CAACgE,EAAEC,IAAID,EAAE,GAAGpD,EAAEqD,GAAGD,EAAE,EAAE,EAAE,IAAI3D,EAAEO,EAAE,OAAOiB,EAAE2zF,GAAGn1F,CAAC,EAAEyB,EAAE9B,EAAE,IAAIgE,GAAGA,EAAE,EAAE,EAAE,KAAK,GAAG,EAAEjC,EAAE/B,EAAE,IAAI,CAACgE,EAAEC,IAAID,EAAE,GAAGpD,EAAEqD,EAAE,EAAE,KAAK,GAAG,EAAEjC,EAAE+4F,GAAG,KAAK16F,CAAC,EAAE,EAAE06F,GAAG,SAAS16F,CAAC,EAAEqC,EAAE,GAAGV,EAAE3B,EAAE,QAAQ,KAAK,YAAYA,EAAE,KAAK,EAAEA,IAAI,EAAE,SAAS,QAAQ,EAAE,MAAM,EAAE,EAAE,KAAK,KAAKuC,EAAE,CAAC,GAAGf,oBAAoB,GAAGG,EAAE3B,EAAE;AAAA,YAC5aqC;AAAA,QACJrC,IAAI,EAAE,GAAG;AAAA;AAAA,SAER2B,EAAE3B,EAAE;AAAA,YACD2B,EAAE3B,EAAE,QAAQ,KAAK,YAAYA,EAAE,QAAQA,IAAI,EAAE,GAAG,KAAK2B,EAAE3B,EAAE;AAAA,cACvDqC,MAAM,EAAEG,EAAExC,IAAI,EAAE,0BAA0B,6DAA6DyC,EAAE,GAAG,QAAQkB,EAAE,EAAEC,EAAE5D,IAAI,EAAE,EAAE,EAAE2D,EAAEC,EAAED,IAAIlB,GAAG;AAAA,UACnJF,EAAEoB;AAAA,cACEnB;AAAA,mBACKmB;AAAA;AAAA,YAEPnC;AAAA,mBACOmC,wBAAwB,EAAE,KAAK,OAAO;AAAA;AAAA,QAEjDlB,GAAGzC,IAAI,EAAE,KAAK,KAAK,KAAK,SAAS;AAAA,cAC3BwB,aAAaA,KAAKC;AAAA,cAClBD,WAAWA,KAAKE;AAAA;AAAA;AAAA,UAGpBF;AAAA;AAAA,UAEAiB;AAAA;AAAA;AAAA,KAGL,CAAC,EAAMuzG,GAAGv2G,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,SAAS,EAAE,cAAc,CAAC,EAAED,EAAE,GAAGwmB,EAAE,cAAcvmB,EAAE,KAAK,IAAI,EAAE,CAAC,IAAI4B,EAAE,EAAE,IAAI,CAACS,EAAEC,IAAID,EAAE,GAAGrC,EAAE,MAAMsC,GAAGD,EAAE,EAAE,EAAE,OAAOkqG,GAAG,CAAC,QAAQ5sG,EAAE,MAAM,CAAC,MAAMiC,EAAE,MAAM,EAAE,MAAM5B,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI0B,EAAEwY,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAI67F,IAAG/1G,EAAE,MAAM,EAAE,CAAC,EAAE,IAAI81G,IAAG91G,EAAE,MAAM,EAAE,CAAC,EAAE2B,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,OAAOhC,EAAE,gBAAgB+B,EAAE,CAAC1B,CAAC,EAAEA,EAAE,MAAM2B,CAAC,CAAC,EAAEs0G,IAAG,CAAC,WAAWx0F,GAAG,YAAY,QAAQ,WAAWu0F,EAAE,EAAME,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAS1aC,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAgBF9Y,GAAG;AAAA;AAAA,EAEL+Y,IAAI/X,GAAG,CAAC,UAAU6X,IAAI,gBAAgBC,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW10F,GAAG,YAAY,QAAQ,WAAWy0F,GAAG,EAAE,SAASE,IAAI72G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE,CAAC,EAAEC,EAAE2kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAEqC,EAAET,EAAEU,EAAE23C,EAAE,mBAAmB53C,EAAEX,CAAC,EAAEa,EAAEvC,EAAEsC,GAAG,OAAOC,EAAEo9F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK2C,CAAC,CAAC,CAAC,EAAED,EAAE43C,EAAE,iBAAiB53C,EAAE,OAAOX,CAAC,EAAEC,EAAE,KAAKY,CAAC,GAAG03C,EAAE,2BAA2B,OAAO53C,EAAEX,CAAC,EAAE,IAAIc,EAAE,GAAG7C,EAAE,mBAAmB,CAAC4C,CAAC,CAAC,EAAE,CAAC,IAAIE,EAAE9C,EAAE,QAAQ,IAAI4C,EAAE,MAAM,EAAE,OAAO,CAAC,QAAQoB,EAAE,SAAS,EAAE,SAASqD,CAAC,EAAEkyF,GAAG32F,EAAE,MAAMA,EAAE,MAAME,EAAEJ,CAAC,EAAEG,EAAE7C,EAAE,eAAe,EAAEqH,EAAErD,CAAC,CAAC,KAAK,CAAC,GAAG,CAAClB,EAAEkB,CAAC,EAAEs2C,EAAE,0BAA0B13C,EAAE,MAAMF,CAAC,EAAE,EAAEkkB,EAAE,cAAc5iB,CAAC,EAAEqD,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAEv8F,CAAC,EAAE,QAAQ5C,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE0qB,GAAGrqB,EAAE,KAAK,EAAEoE,EAAE+6F,GAAGn4F,EAAE,EAAE,OAAOrH,CAAC,EAAE6C,EAAEs8F,GAAG,CAAC,OAAO,CAAC16F,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,EAAEd,EAAE,KAAKqF,CAAC,EAAErF,EAAE,KAAKyC,CAAC,CAAC,CAAC,GAAG,EAAE,CAACzC,EAAE,KAAKa,CAAC,EAAE,IAAIC,EAAEw3C,EAAE,qBAAqBz3C,EAAE,MAAMZ,CAAC,EAAEY,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAEt8F,CAAC,EAAE,QAAQ7C,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,CAAC,CAAC,OAAOd,EAAE,QAAQc,GAAG9C,EAAE,8BAA8B8C,CAAC,CAAC,EAAED,CAAC,CAAC,IAAI+zG,IAAG,CAAC,WAAW10F,GAAG,YAAY,QAAQ,WAAWy0F,GAAG,EAAE,SAASE,IAAI/2G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,mBAAmBO,EAAE,kBAAkB,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,iBAAiB0B,CAAC,EAAE3B,EAAE4B,EAAE3B,EAAE,IAAIgH,GAAGrH,EAAE,SAASqH,EAAE,MAAM,CAAC,EAAEpF,EAAE5B,EAAE,IAAIgH,GAAGA,EAAE,KAAK,EAAE3E,EAAE1C,EAAE,SAAS,EAAE,MAAM,EAAE2C,EAAE3C,EAAE,SAAS,EAAE,MAAM,EAAE,CAAC4C,EAAEC,EAAEC,CAAC,EAAE02F,GAAGx3F,EAAEC,EAAES,EAAE,EAAE,MAAM,EAAE,MAAMC,EAAE,EAAE,MAAMZ,CAAC,EAAEiC,EAAEpB,EAAE,IAAIyE,GAAGrH,EAAE,eAAe,CAACqH,EAAE,MAAM,EAAE,QAAQA,CAAC,CAAC,EAAE,EAAErH,EAAE,eAAe8C,EAAE,EAAE,MAAMD,CAAC,EAAE,OAAOmB,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI8yG,IAAG,CAAC,WAAW30F,GAAG,YAAY,QAAQ,WAAW00F,GAAG,EAAE,SAASE,IAAIj3G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,OAAOM,EAAE,OAAOC,EAAE,OAAO,CAAC,EAAE,EAAE,EAAEL,EAAE,SAASI,EAAE,MAAM,EAAE2B,EAAE/B,EAAE,SAASK,EAAE,MAAM,EAAE2B,EAAEhC,EAAE,SAAS,EAAE,MAAM,EAAE,CAACiC,EAAES,CAAC,EAAE+2F,GAAG,EAAEr5F,EAAE,MAAMA,EAAE,MAAM2B,EAAE1B,EAAE,MAAM2B,EAAE,EAAE,KAAK,EAAEW,EAAE3C,EAAE,eAAe,CAACiC,EAAE,MAAM,EAAE,QAAQA,CAAC,EAAEW,EAAE5C,EAAE,eAAe,CAAC0C,EAAE,MAAM,EAAEtC,EAAE,MAAMsC,CAAC,EAAE,MAAM,CAACC,EAAEC,CAAC,CAAC,CAAC,IAAIo0G,IAAG,CAAC,WAAW50F,GAAG,YAAY,QAAQ,WAAW20F,GAAG,EAAE,SAASE,IAAIn3G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,OAAO,EAAE,aAAa,EAAE,oBAAoB0B,CAAC,EAAE,EAAE,CAAC,kBAAkBC,CAAC,EAAE5B,EAAE6B,EAAEjC,EAAE,SAASK,EAAE,MAAM,EAAEqC,EAAE1C,EAAE,SAAS,EAAE,MAAM,EAAE2C,EAAE3C,EAAE,SAAS,EAAE,MAAM,EAAE4C,EAAEb,EAAE,IAAI,GAAG/B,EAAE,SAAS,EAAE,MAAM,CAAC,EAAE6C,EAAEd,EAAE,IAAI,GAAG,EAAE,KAAK,EAAE,CAACe,EAAEkB,CAAC,EAAE01F,GAAGz3F,EAAE5B,EAAE,MAAMqC,EAAE,EAAE,MAAM,EAAE,MAAMC,EAAE,EAAE,MAAMC,EAAEC,EAAEb,CAAC,EAAE,OAAOhC,EAAE,eAAe8C,EAAE,EAAE,MAAMkB,CAAC,CAAC,CAAC,IAAIkzG,IAAG,CAAC,WAAW70F,GAAG,YAAY,QAAQ,WAAW40F,GAAG,EAAME,GAAGr3G,GAAG,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,KAAKC,EAAE,KAAK,EAAE,MAAM,CAAC,EAAEL,EAAE+B,EAAE43F,GAAGv5F,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAO,EAAE,eAAe,CAAC0B,EAAE,MAAM,EAAE,EAAEA,CAAC,CAAC,EAAEq1G,IAAG,CAAC,WAAW90F,GAAG,YAAY,QAAQ,WAAW60F,EAAE,EAAME,IAAI,kBAAkBC,IAAI7Y,GAAG,CAAC,UAAU4Y,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW/0F,GAAG,YAAY,QAAQ,WAAW80F,GAAG,EAAME,IAAI5b,GAAG;AAAA;AAAA,EAEl7E6b,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUJC,IAAIjZ,GAAG,CAAC,UAAU+Y,IAAI,gBAAgBC,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWl1F,GAAG,YAAY,QAAQ,WAAWi1F,GAAG,EAAME,IAAIhc,GAAG;AAAA;AAAA,EAE7Gic,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUJC,IAAIrZ,GAAG,CAAC,UAAUmZ,IAAI,gBAAgBC,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWh1F,GAAG,YAAY,QAAQ,WAAW+0F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYp3G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAE,CAAC,EAAEpB,EAAE,KAAK,YAAY,CAACkB,EAAE9B,EAAE,EAAE,CAAC,EAAE,IAAI0C,EAAE,CAACrC,GAAGL,EAAE,EAAE+B,EAAE,EAAEA,EAAE1B,GAAG,EAAE,EAAE2B,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC3B,GAAGL,EAAE,EAAEA,EAAE,EAAEA,EAAEK,GAAG,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEuC,EAAEf,EAAEe,EAAE,wEAAwEA,EAAE,8CAA8C,KAAK,SAAS;AAAA;AAAA,YAE/ZF,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,uCACoBX,QAAQC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCASZY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAoB9B,CAAC,EAAMq1G,IAAG,KAAK,CAAC,YAAYr3G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAY,CAAC,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAE,CAAC,EAAEpB,EAAE,KAAK,YAAY,CAACkB,EAAE9B,EAAE,EAAE,CAAC,EAAE,IAAI0C,EAAE,CAACrC,GAAGL,EAAE,EAAE+B,EAAE,EAAEA,EAAE1B,GAAG,EAAE,EAAE2B,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC3B,GAAGL,EAAE,EAAEA,EAAE,EAAEA,EAAEK,GAAG,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEuC,EAAEf,EAAEe,EAAE,wEAAwEA,EAAE,8CAA8C,KAAK,SAAS;AAAA;AAAA,YAE7WF,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,uCACoBX,QAAQC;AAAA,uCACRA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAcJY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAQH,EAAE;AAAA,uCACK,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAgDpC,CAAC,EAAE,SAASs1G,IAAIp4G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK0B,CAAC,EAAE3B,EAAE,CAAC4B,EAAEC,CAAC,EAAEF,EAAEW,EAAE6X,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAI09F,IAAG53G,EAAE,MAAM2B,EAAEC,EAAE,EAAE,CAAC,EAAE,IAAI+1G,IAAG33G,EAAE,MAAM2B,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,CAAC,EAAE,SAAS,CAAC,CAAC,IAAI83G,IAAG,CAAC,WAAWt1F,GAAG,YAAY,QAAQ,WAAWq1F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYx3G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYA,EAAE,GAAG,CAAC,CAACK,EAAEwB,CAAC,EAAE7B,EAAE,CAAC,CAAC8B,EAAEC,CAAC,EAAEnB,EAAEoB,EAAE,CAAC,GAAGF,EAAE,EAAEzB,EAAE,EAAEA,EAAE,GAAG0B,EAAE,EAAEF,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGC,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGC,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAEW,EAAEV,EAAE,GAAG,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAE,GAAGY,EAAE,EAAEF,EAAEG,EAAE,EAAE,EAAEC,EAAE,KAAK,KAAKF,CAAC,EAAE,EAAE,EAAEoB,EAAE,KAAK,KAAKnB,CAAC,EAAE,EAAE,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAU7hBH;AAAA,yCACD;AAAA;AAAA,6CAEIE;AAAA,4CACDC;AAAA;AAAA,oCAERC;AAAA,mCACDkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAcDlC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAQEC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sDAMkB1B,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,qDAMHwB,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KA8BlD,CAAC,EAAE,SAASw2G,IAAIv4G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,aAAa,CAAC,EAAED,EAAE2B,EAAE,IAAIq2G,IAAG,EAAE,MAAM/3G,EAAE,MAAM,CAAC,EAAE,OAAOL,EAAE,gBAAgB+B,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,IAAIu2G,IAAG,CAAC,WAAWx1F,GAAG,YAAY,QAAQ,WAAWu1F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY33G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,CAAC,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAE,CAAC,EAAEpB,EAAE,KAAK,YAAY,CAACkB,EAAE9B,EAAE,EAAE,CAAC,EAAE,IAAI0C,EAAE,CAACrC,GAAGL,EAAE,EAAE+B,EAAE,EAAEA,EAAE1B,GAAG,EAAE,EAAE2B,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC3B,GAAGL,EAAE,EAAEA,EAAE,EAAEA,EAAEK,GAAG,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEuC,EAAEvC,EAAE,MAAM,MAAMwC,EAAEhB,EAAEgB,EAAE,4EAA4EA,EAAE,8CAA8C,KAAK,SAAS;AAAA;AAAA,YAE5iBH,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,uCACoBX,QAAQC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCASZa;AAAA;AAAA;AAAA;AAAA,8DAI2BD;AAAA;AAAA;AAAA;AAAA;AAAA,KAKzD,CAAC,EAAM41G,IAAG,KAAK,CAAC,YAAY53G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,KAAK,YAAY,CAAC,EAAE,GAAG,CAACC,EAAEC,EAAEC,EAAE,CAAC,EAAEpB,EAAE,KAAK,YAAY,CAACkB,EAAE9B,EAAE,EAAE,CAAC,EAAE,IAAI0C,EAAE,CAACrC,GAAGL,EAAE,EAAE+B,EAAE,EAAEA,EAAE1B,GAAG,EAAE,EAAE2B,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC3B,GAAGL,EAAE,EAAEA,EAAE,EAAEA,EAAEK,GAAG,EAAE,EAAE,EAAE,EAAE,CAAC,EAAEuC,EAAEvC,EAAE,MAAM,MAAMwC,EAAEhB,EAAEgB,EAAE,4EAA4EA,EAAE,8CAA8C,KAAK,SAAS;AAAA;AAAA,YAEjYH,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,YACPA,EAAE,GAAG,EAAE;AAAA,uCACoBX,QAAQC;AAAA,uCACRA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAcJa;AAAA;AAAA;AAAA;AAAA,8DAI2BD;AAAA;AAAA;AAAA,gCAG9B,EAAE;AAAA,uCACK,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAapC,CAAC,EAAE,SAAS61G,IAAI34G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK0B,CAAC,EAAE3B,EAAE,CAAC4B,EAAEC,CAAC,EAAEF,EAAEW,EAAE6X,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAIi+F,IAAGn4G,EAAE,MAAM2B,EAAEC,EAAE,EAAE,CAAC,EAAE,IAAIs2G,IAAGl4G,EAAE,MAAM2B,EAAEC,EAAE,EAAE,CAAC,EAAE,OAAOjC,EAAE,gBAAgB0C,EAAE,CAACrC,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAIq4G,IAAG,CAAC,WAAW/1F,GAAG,YAAY,QAAQ,WAAW81F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY/3G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,YAAYA,EAAE,GAAG,CAAC,CAACK,EAAEwB,CAAC,EAAE7B,EAAE,CAAC,CAAC8B,EAAEC,CAAC,EAAEnB,EAAEoB,EAAE,CAAC,GAAGF,EAAE,EAAEzB,EAAE,EAAEA,EAAE,GAAG0B,EAAE,EAAEF,EAAE,EAAEA,CAAC,EAAE,EAAE,CAAC,GAAGC,EAAE,EAAEA,EAAE,EAAEA,EAAE,GAAGC,EAAE,EAAEA,EAAE,EAAEA,CAAC,EAAEW,EAAEV,EAAE,GAAG,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAE,GAAGY,EAAE,EAAEF,EAAEG,EAAE,EAAE,EAAEC,EAAE,KAAK,KAAKF,CAAC,EAAE,EAAE,EAAEoB,EAAE,KAAK,KAAKnB,CAAC,EAAE,EAAE,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAU3hBH;AAAA,yCACD;AAAA;AAAA,6CAEIE;AAAA,4CACDC;AAAA;AAAA,oCAERC;AAAA,mCACDkB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAcDlC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAQEC;AAAA;AAAA;AAAA;AAAA;AAAA,sBAKdC,EAAE;AAAA,sCACc,EAAE;AAAA;AAAA;AAAA,wBAGhBA,EAAE;AAAA,wCACc,EAAE;AAAA;AAAA;AAAA,4BAGd3B;AAAA,kBACV;AAAA;AAAA;AAAA;AAAA,4BAIUwB;AAAA,kBACV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,KAYb,CAAC,EAAE,SAAS+2G,IAAI94G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,aAAa,CAAC,EAAED,EAAE2B,EAAE,IAAI42G,IAAG,EAAE,MAAMt4G,EAAE,MAAM,CAAC,EAAE,OAAOL,EAAE,gBAAgB+B,EAAE,CAAC,CAAC,EAAE,EAAE,KAAK,CAAC,CAAC,IAAI82G,IAAG,CAAC,WAAWj2F,GAAG,YAAY,QAAQ,WAAWg2F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYl4G,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,IAAI,EAAEY,EAAE,OAAO,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,kCAAkC,+BAA+B,EAAE,GAAG,KAAK,YAAYA,EAAE,IAAI,EAAE,CAAC,KAAK,SAAS;AAAA;AAAA;AAAA,2BAG7YA,EAAE;AAAA;AAAA,QAErB,MAAM,CAAC,IAAIP,EAAE0B,GAAG/B,EAAE,QAAQ+B,CAAC,IAAI,IAAInB,EAAEmB,KAAK,EAAE,GAAGnB,EAAEmB,eAAeA,SAAS,UAAUA,KAAKF,EAAEjB,EAAE,IAAI,CAACmB,EAAEC,IAAI3B,EAAE2B,CAAC,CAAC,EAAE,KAAK,GAAG,EAAEF,EAAE0zF,GAAG,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,UAE3I1zF;AAAA,yBACeD;AAAA;AAAA,KAEpB,CAAC,EAAMk3G,IAAG,KAAK,CAAC,YAAYn4G,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,aAAa,GAAG,KAAK,aAAa,GAAG,IAAI,EAAEY,EAAE,OAAO,GAAG,EAAE,EAAE,MAAM,IAAI,MAAM,kCAAkC,+BAA+B,EAAE,KAAK,YAAYA,EAAE,IAAIP,EAAE06F,GAAG,KAAK,CAAC,EAAEl5F,EAAE,GAAGxB,EAAE,EAAE,YAAY,KAAK,YAAY,EAAE,KAAKyB,EAAE,GAAGzB,EAAE,EAAE,YAAY,KAAK,YAAY,EAAE,KAAK0B,EAAEyzF,GAAG,CAAC,EAAE,IAAI,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA,uCAIrU50F,EAAE;AAAA,cAC3BA,EAAE;AAAA,eACDiB;AAAA,2CAC4BjB,EAAE;AAAA,kBAC3BA,EAAE;AAAA;AAAA;AAAA;AAAA,QAIZ,KAAK,SAAS;AAAA;AAAA,YAEVmB;AAAA;AAAA,uBAEWC,EAAE3B,EAAE,MAAM,CAAC;AAAA,eACnBwB;AAAA,yBACU,EAAExB,EAAE,MAAM,CAAC;AAAA;AAAA,eAErByB;AAAA,yBACUY,EAAErC,EAAE,MAAM,CAAC;AAAA,iBACnBwB;AAAA,2BACU,EAAExB,EAAE,MAAM,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA,MAKhC,SAAS2B,EAAEc,EAAE,CAAC,OAAOF,EAAEE,CAAC,CAAC,CAAC,SAAS,EAAEA,EAAE,CAAC,OAAOA,EAAE,EAAE,GAAG,IAAIA,EAAE,EAAE,GAAG,QAAQF,EAAEE,CAAC,CAAC,CAAC,SAASJ,EAAEI,EAAE,CAAC,OAAOA,EAAE,EAAE,GAAG,IAAIA,EAAE,EAAE,GAAG,QAAQF,EAAEE,CAAC,CAAC,CAAC,SAAS,EAAEA,EAAE,CAAC,OAAOA,EAAE,EAAE,GAAG,IAAIA,EAAE,EAAE,GAAG,QAAQA,EAAE,EAAE,GAAG,IAAIA,EAAE,EAAE,GAAG,QAAQF,EAAEE,CAAC,CAAC,CAAC,SAASF,EAAEE,EAAE,CAAC,IAAIkB,EAAEpD,EAAE,IAAI,CAACiE,EAAEJ,IAAI5B,EAAE4B,EAAE3B,CAAC,CAAC,EAAEmB,EAAED,EAAE,KAAK,GAAG,EAAEqD,EAAErD,EAAE,MAAM,EAAE,EAAE,KAAK,GAAG,EAAE,MAAM,mBAAmBC,YAAYoD,KAAK,CAAC,SAASxE,EAAEC,EAAEkB,EAAE,CAAC,OAAOhE,EAAE,QAAQ8C,CAAC,IAAI,IAAIlC,EAAEkC,KAAK,EAAE,GAAGlC,EAAEkC,QAAQkB,EAAElB,SAAS,GAAGkB,EAAElB,IAAI,CAAC,CAAC,EAAE,SAASk2G,IAAIl5G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAEC,EAAE,MAAM,OAAO0B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG,IAAI,EAAE,OAAOu9F,GAAG,CAAC,OAAO,CAAC,EAAEv9F,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAIgC,EAAEuY,EAAE,EAAE,QAAQ,6BAA6B,EAAE,IAAIw+F,IAAG14G,EAAE,MAAM0B,CAAC,EAAE,IAAI+2G,IAAGz4G,EAAE,MAAM0B,CAAC,EAAE,OAAO/B,EAAE,gBAAgBgC,EAAE,CAAC3B,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAI44G,IAAG,CAAC,WAAWj2F,GAAG,YAAY,QAAQ,WAAWg2F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYt4G,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,OAAO,EAAE,KAAK,YAAY,CAAC,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,SAAS,KAAK,MAAM,CAAC,EAAE,IAAI,EAAEY,EAAE,GAAGP,EAAEO,EAAE,GAAG,KAAK,YAAYA,EAAE,IAAIiB,EAAE,GAAG,OAAO7B,GAAG,SAAS6B,EAAE,uBAAuB7B,EAAE,QAAQ,CAAC,KAAK6B,EAAE;AAAA,2BACl9B7B,EAAE,KAAK,GAAG;AAAA,8CACS,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAWhD6B;AAAA,uCAC2BxB,gCAAgC;AAAA;AAAA;AAAA;AAAA;AAAA,KAKlE,CAAC,EAAM84G,IAAG,CAAC,WAAWxzF,GAAG,YAAY,QAAQ,WAAW,CAAC,CAAC,OAAO7lB,EAAE,MAAM,EAAE,QAAQE,CAAC,IAAI,CAAC,GAAG,CAAC,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,EAAE,UAAU,EAAE,OAAO,CAAC,EAAE,EAAE0B,EAAE/B,EAAEgC,EAAE,IAAIk3G,IAAG94G,EAAE,MAAM,CAAC,EAAE,CAAC6B,EAAES,CAAC,EAAE43C,EAAE,eAAe,EAAEl6C,EAAE,MAAM,GAAGA,EAAE,MAAM,EAAE,EAAEuC,EAAE,CAAC,CAACV,EAAES,EAAE,KAAK,IAAIrC,CAAC,EAAE,KAAK,IAAIA,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,gBAAgBC,EAAE,CAAC5B,CAAC,EAAEA,EAAE,MAAMuC,CAAC,CAAC,CAAC,EAAMy2G,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAe5SC,IAAI5a,GAAG,CAAC,UAAU2a,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWr2F,GAAG,YAAY,QAAQ,WAAWo2F,GAAG,EAAME,IAAI,yBAAyBC,IAAI/a,GAAG,CAAC,UAAU8a,IAAI,cAAc3f,EAAE,CAAC,EAAE6f,IAAG,CAAC,WAAWv2F,GAAG,YAAY,QAAQ,WAAWs2F,GAAG,EAAME,GAAG,KAAK,CAAC,YAAY94G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAEC,EAAE,GAAG,CAAC,KAAK,cAAc,CAAC,UAAU,UAAU,cAAc,EAAE,KAAK,YAAYD,EAAE,IAAIE,EAAEwzF,GAAG3zF,EAAE,MAAM,EAAE,EAAE2zF,GAAG1zF,EAAE,MAAM,EAAEY,EAAE,GAAG,IAAI,EAAEA,EAAE,IAAI,IAAI,IAAIA,EAAE,QAAQ,IAAI,EAAE,cAAcA,KAAKE,EAAE,GAAGvC,IAAI,EAAEuC,EAAE,IAAIvC,IAAI,IAAIuC,EAAE,gBAAgB,IAAIC,EAAE,cAAcD,KAAKE,EAAE9C,EAAE,EAAE,aAAa,UAAU,KAAK,SAAS;AAAA,UACzgBgC,eAAeA,KAAKH;AAAA;AAAA;AAAA,YAGlB;AAAA;AAAA;AAAA,gCAGoBjB;AAAA;AAAA,kCAEEZ;AAAA,kCACA;AAAA,0CACQ8C;AAAA;AAAA;AAAA,uBAGnBD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,OAMhB,CAAC,EAAE,SAAS82G,IAAI75G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,EAAED,EAAE,CAAC,UAAU2B,EAAE,WAAWC,EAAE,UAAUC,EAAE,QAAQS,EAAE,WAAWC,CAAC,EAAE23C,EAAE,gBAAgB,EAAEj6C,EAAE,CAAC,EAAEuC,EAAE,CAACD,EAAEV,EAAEA,CAAC,EAAE,GAAGU,IAAI,EAAE,OAAO3C,EAAE,eAAe,EAAEK,EAAE,KAAK,EAAE,IAAIwC,EAAEs8F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,CAACgC,EAAED,CAAC,CAAC,CAAC,CAAC,EAAEe,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAM,CAACgC,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAE+B,EAAEhE,EAAE,eAAe,CAAC,EAAE,UAAU,IAAI,aAAa,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,IAAI05G,GAAG13G,EAAED,EAAEc,EAAE,MAAM,OAAOC,EAAE,MAAM,OAAOJ,EAAEE,CAAC,EAAEyE,EAAErH,EAAE,gBAAgB,EAAE,CAAC8C,EAAED,EAAEmB,CAAC,EAAElB,EAAE,KAAK,EAAE,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAA93F,CAAC,EAAE,QAAQrH,EAAE,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,OAAOA,EAAE,8BAA8B6C,CAAC,EAAE7C,EAAE,8BAA8B8C,CAAC,EAAE9C,EAAE,8BAA8BqH,CAAC,EAAErH,EAAE,8BAA8BgE,CAAC,EAAE,CAAC,CAAC,IAAI41G,IAAG,CAAC,WAAWz2F,GAAG,YAAY,QAAQ,WAAWw2F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYj5G,EAAEZ,EAAE,EAAEK,EAAE,CAAC,KAAK,cAAc,CAAC,iBAAiB,QAAQ,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,YAAY,KAAK,KAAK,CAAC,EAAE,KAAK,YAAY,CAACO,EAAE,CAAC,EAAE,IAAIiB,EAAE,yBAAyBC,EAAE,uBAAuB,KAAK,KAAK,KAAK,KAAK9B,EAAE,CAAC,CAAC,sCAAsC+B,EAAEwY,EAAE,EAAE,UAAU,eAAe,IAAI,EAAE1Y,EAAEC,EAAEE,EAAE3B,IAAI,OAAO,IAAI,KAAK,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,WAKrjC0B;AAAA;AAAA,+CAEoCC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAkBzC,CAAC,EAAE,SAAS83G,IAAIh6G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,eAAeO,EAAE,OAAO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE2B,EAAE,IAAI83G,IAAGx5G,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG,EAAE,MAAM,GAAG,CAAC,EAAE2B,EAAE,CAAC,CAAC3B,EAAE,MAAM,EAAE,CAAC,EAAE,OAAOL,EAAE,gBAAgB+B,EAAE,CAAC1B,EAAE,CAAC,EAAE,QAAQ2B,CAAC,CAAC,CAAC,IAAI+3G,IAAG,CAAC,WAAW32F,GAAG,YAAY,QAAQ,WAAW02F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYp5G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,IAAI,GAAG,EAAE,KAAK,YAAYA,EAAE,IAAIK,EAAEwB,EAAE,GAAG,EAAE,EAAE,MAAM,MAAM,kBAAkB,wBAAwB,EAAE,GAAG,IAAI,EAAEA,EAAE,QAAQxB,EAAE,YAAY,CAAC,IAAI0B,EAAE,CAAC,UAAU,UAAU,UAAU,SAAS,EAAEC,EAAE,CAAC,EAAE,EAAE,CAAC,EAAE,QAAQU,EAAE,EAAEA,EAAE1C,EAAE,OAAO0C,IAAI,EAAE,KAAK,GAAGX,EAAEW,IAAI,EAAEA,EAAE9B,GAAGoB,EAAE,KAAK,GAAGD,EAAEW,IAAI,EAAErC,EAAE2B,EAAE,KAAK,EAAEH,EAAE,EAAE,KAAK,CAAC,CAAC,IAAIC,EAAE0zF,GAAG,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,UAEzmB1zF;AAAA,4BACkBzB;AAAA;AAAA,2BAEDwB;AAAA;AAAA,2BAEAA;AAAA;AAAA;AAAA,KAGtB,CAAC,EAAE,SAASo4G,IAAIn6G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,UAAUM,EAAE,EAAEC,EAAE,EAAE,CAAC,EAAE,EAAE,EAAE,IAAI25G,IAAG55G,EAAE,MAAM,OAAOC,EAAE,MAAMA,EAAE,MAAM,MAAM,EAAE,OAAOL,EAAE,gBAAgB,EAAE,CAACI,EAAEC,EAAE,CAAC,EAAEoqB,GAAGpqB,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC,IAAI65G,IAAG,CAAC,WAAW72F,GAAG,YAAY,QAAQ,WAAW42F,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA,uBAG7N7/D,EAAE;AAAA,kBACPA,EAAE;AAAA;AAAA,EAElB8/D,IAAI3b,GAAG,CAAC,UAAU0b,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW/2F,GAAG,YAAY,QAAQ,WAAW82F,GAAG,EAAME,IAAI9b,GAAG;AAAA;AAAA,EAEzF+b,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUJC,IAAI/b,GAAG,CAAC,UAAU6b,IAAI,gBAAgBC,IAAI,cAAczgB,EAAE,CAAC,EAAE2gB,IAAG,CAAC,WAAW92F,GAAG,YAAY,QAAQ,WAAW62F,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA,EAG3HC,IAAIlc,GAAG,CAAC,UAAUic,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWl3F,GAAG,YAAY,QAAQ,WAAWi3F,GAAG,EAAME,IAAIrc,GAAG;AAAA;AAAA,EAEzFsc,IAAIrc,GAAG,CAAC,UAAUoc,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWv3F,GAAG,YAAY,QAAQ,WAAWs3F,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA,EAGtFC,IAAIxc,GAAG,CAAC,UAAUuc,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWz3F,GAAG,YAAY,QAAQ,WAAWw3F,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBtFC,IAAI3c,GAAG,CAAC,UAAU0c,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWz3F,GAAG,YAAY,QAAQ,WAAWw3F,GAAG,EAAME,IAAIx7G,GAAG,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,SAAS,CAAC,EAAED,EAAEwmB,EAAE,OAAOvmB,EAAE,MAAM,QAAQ,EAAE,IAAI,sEAAsE,EAAE,IAAI0B,EAAE,EAAE,OAAO,CAACsF,EAAE,IAAIA,EAAE,CAAC,EAAErF,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,EAAEA,EAAE,KAAK,GAAG,CAAC,EAAE,QAAQqF,EAAE,EAAE,EAAE,OAAOA,EAAEhH,EAAE,MAAM,OAAO,EAAEgH,EAAErF,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,EAAE,IAAIC,EAAE,CAAC,EAAES,EAAE2zG,GAAG,CAAC,OAAO,CAAC,EAAEh2G,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,SAASgC,EAAE,cAAc,CAAC,CAAC,CAAC,EAAEW,EAAE23C,EAAE,YAAY53C,EAAE,MAAM,EAAEX,EAAE,EAAE,EAAEa,EAAE03C,EAAE,YAAY33C,EAAE,OAAO,EAAE,OAAO,EAAE,EAAEE,EAAEy3C,EAAE,oBAAoB53C,EAAE,MAAM,EAAEX,EAAE,EAAE,EAAEe,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAEz8F,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,MAAM2C,CAAC,CAAC,CAAC,EAAEqB,EAAEg8F,GAAG,CAAC,OAAO,CAAC,EAAEl9F,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,KAAK4C,CAAC,CAAC,CAAC,EAAE,EAAEu8F,GAAG,CAAC,OAAO,CAAC,EAAEn7F,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM6C,CAAC,CAAC,CAAC,EAAE,OAAOZ,EAAE,KAAKS,CAAC,EAAET,EAAE,KAAKa,CAAC,EAAEb,EAAE,KAAK+B,CAAC,EAAE/B,EAAE,QAAQoF,GAAGrH,EAAE,8BAA8BqH,CAAC,CAAC,EAAE,CAAC,EAAEk0G,IAAG,CAAC,WAAWx3F,GAAG,YAAY,QAAQ,WAAWu3F,GAAG,EAAE,SAASE,IAAI17G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,QAAQM,EAAE,OAAOC,EAAE,WAAW,EAAE,aAAa,CAAC,EAAE,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,WAC/7B,EAAE,OAAO,EAAE,GAAGD,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,WACjDA,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,WACjDA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UAClD,EAAE,OAAO,EAAE,IAAI0B,EAAE/B,EAAE,SAASI,EAAE,MAAM,EAAE4B,EAAEhC,EAAE,SAASK,EAAE,MAAM,EAAE4B,EAAEjC,EAAE,SAAS,EAAE,MAAM,EAAE0C,EAAE1C,EAAE,SAAS,EAAE,MAAM,EAAE,GAAG,CAAC2C,EAAEC,EAAEC,EAAEC,EAAEkB,CAAC,EAAEi2F,GAAGl4F,EAAE3B,EAAE,MAAMA,EAAE,MAAM4B,EAAE3B,EAAE,MAAM4B,EAAES,CAAC,EAAE,MAAM,CAAC1C,EAAE,eAAe4C,EAAExC,EAAE,MAAMuC,CAAC,EAAE3C,EAAE,eAAe,CAAC4C,EAAE,EAAE,EAAEvC,EAAE,MAAMwC,CAAC,EAAE7C,EAAE,eAAe,CAAC8C,EAAE,MAAM,EAAE,OAAO,IAAI,WAAWA,EAAE,IAAI,GAAG,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE9C,EAAE,eAAe,CAACgE,EAAE,MAAM,EAAE5D,EAAE,MAAM,IAAI,WAAW4D,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIy3G,IAAG,CAAC,WAAWv3F,GAAG,YAAY,QAAQ,WAAWs3F,GAAG,EAAE,SAASE,IAAI57G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,aAAaM,EAAE,WAAWC,EAAE,SAAS,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,uDAAuDA,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,qDAAqDA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,KAAKL,EAAE,SAASK,EAAE,MAAM,CAAC,EAAE0B,EAAE/B,EAAE,SAASI,EAAE,MAAM,EAAE4B,EAAE,MAAM,KAAKhC,EAAE,SAAS,EAAE,MAAM,CAAC,EAAE,CAACiC,EAAES,EAAEC,CAAC,EAAEu3F,GAAGn4F,EAAE3B,EAAE,MAAMA,EAAE,MAAM,EAAE4B,CAAC,EAAE,MAAM,CAAChC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,EAAEjC,EAAE,eAAe,CAAC2C,EAAE,MAAM,EAAE,EAAE,MAAM,IAAI,WAAWA,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIg5G,IAAG,CAAC,WAAWx3F,GAAG,YAAY,QAAQ,WAAWu3F,GAAG,EAAE,SAASE,IAAI97G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,gBACjxCA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,gBACjD,EAAE,OAAO,EAAE,IAAI,EAAEL,EAAE,SAASI,EAAE,MAAM,EAAE2B,EAAE/B,EAAE,SAASK,EAAE,MAAM,EAAE2B,EAAEhC,EAAE,SAAS,EAAE,MAAM,EAAE,CAACiC,EAAES,CAAC,EAAEy3F,GAAG,EAAE/5F,EAAE,MAAMA,EAAE,MAAM2B,EAAEC,EAAE,EAAE,EAAE,OAAOhC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,CAAC,CAAC,IAAI45G,IAAG,CAAC,WAAWz3F,GAAG,YAAY,QAAQ,WAAWw3F,GAAG,EAAE,SAASE,IAAIh8G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,KAAKM,EAAE,QAAQC,EAAE,WAAW,CAAC,EAAE,EAAE,GAAGD,EAAE,MAAM,OAAO,EAAE,MAAM,IAAI,MAAM,2DAA2D,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,eAC1aA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,eACjD,EAAE,OAAO,EAAE,IAAI,EAAEL,EAAE,SAASI,EAAE,MAAM,EAAE2B,EAAE/B,EAAE,SAASK,EAAE,MAAM,EAAE2B,EAAEhC,EAAE,SAAS,EAAE,MAAM,EAAE,CAACiC,EAAES,CAAC,EAAEy3F,GAAG,EAAE/5F,EAAE,MAAMA,EAAE,MAAM2B,EAAEC,CAAC,EAAE,OAAOhC,EAAE,eAAe0C,EAAEtC,EAAE,MAAM6B,CAAC,CAAC,CAAC,IAAI85G,IAAG,CAAC,WAAW13F,GAAG,YAAY,QAAQ,WAAWy3F,GAAG,EAAE,SAASE,IAAIl8G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,cAAcO,EAAE,aAAa,EAAE,aAAa,CAAC,EAAE,EAAE,CAAC,YAAY0B,CAAC,EAAE3B,EAAE,CAAC,UAAU4B,EAAE,WAAWC,EAAE,UAAUS,EAAE,QAAQC,EAAE,WAAWC,CAAC,EAAE03C,EAAE,gBAAgB,EAAEj6C,EAAE0B,CAAC,EAAEc,EAAE,GAAG,GAAG,EAAE,QAAQ,SAAS,CAAC,IAAIwE,EAAErH,EAAE,WAAWK,CAAC,EAAE,EAAEL,EAAE,WAAW,CAAC,EAAEyE,EAAEmiB,EAAE,aAAa5mB,EAAE,SAAS,EAAE,MAAM,EAAE,EAAE,EAAE0E,EAAEm1F,GAAGxyF,EAAE,EAAEtF,EAAEa,EAAEF,EAAET,EAAED,EAAEW,EAAE8B,EAAE5B,CAAC,EAAE,OAAO7C,EAAE,eAAe+B,EAAE2C,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,IAAI5B,EAAE,IAAI42G,GAAGz3G,EAAED,EAAE3B,EAAE,MAAM,OAAO,EAAE,MAAM,OAAOsC,EAAE,CAACC,EAAE,CAAC,EAAEC,CAAC,EAAEmB,EAAEhE,EAAE,gBAAgB8C,EAAE,CAAC,EAAEzC,EAAE,CAAC,EAAE,EAAE,KAAK,EAAE,EAAE8+F,GAAG,CAAC,OAAO,CAAC,EAAEn7F,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM+B,CAAC,CAAC,CAAC,EAAE,OAAO/B,EAAE,8BAA8BgE,CAAC,EAAE,CAAC,CAAC,IAAIi4G,IAAG,CAAC,WAAW33F,GAAG,YAAY,QAAQ,WAAW03F,GAAG,EAAE,SAASE,IAAIp8G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,gBAAgB,EAAE,KAAK,CAAC,EAAED,EAAE2B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG2B,EAAEs4C,EAAE,iBAAiBj6C,EAAE,EAAE0B,CAAC,EAAEE,EAAE5B,EAAE,MAAM,OAAOqC,EAAE,IAAI,MAAMT,CAAC,EAAE,KAAK,CAAC,EAAEU,EAAEtC,EAAE,MAAM,MAAM,EAAE,OAAO2B,EAAE,IAAIY,GAAG,CAAC,IAAIC,EAAE,CAAC,GAAGF,CAAC,EAAEE,EAAEd,GAAGa,EAAE,IAAIE,EAAE4hG,GAAG,CAAC,OAAO,CAAC,EAAErkG,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM0C,EAAE,KAAKG,CAAC,CAAC,CAAC,EAAE,OAAOH,EAAEX,IAAIa,EAAEE,CAAC,CAAC,CAAC,CAAC,IAAIq5G,IAAG,CAAC,WAAWn4F,GAAG,YAAY,QAAQ,WAAWk4F,GAAG,EAAME,GAAG,kBAAkBC,IAAI5d,GAAG,CAAC,UAAU2d,GAAG,gBAAgBA,GAAG,cAAchiB,EAAE,CAAC,EAAEkiB,IAAG,CAAC,WAAWz4F,GAAG,YAAY,QAAQ,WAAWw4F,GAAG,EAAME,IAAI,gBAAgBC,IAAI/d,GAAG,CAAC,UAAU8d,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAWj4F,GAAG,YAAY,QAAQ,WAAWg4F,GAAG,EAAME,GAAG,4BAA4BC,IAAIje,GAAG,CAAC,UAAUge,GAAG,gBAAgBA,EAAE,CAAC,EAAEE,IAAG,CAAC,WAAWr4F,GAAG,YAAY,QAAQ,WAAWo4F,GAAG,EAAE,SAASE,IAAI,CAAC,OAAO/8G,EAAE,MAAM,EAAE,QAAQE,CAAC,EAAE,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAEO,EAAEu7F,GAAG;AAAA,mCAC9lD,EAAE;AAAA,IACjC,EAAE,IAAID,GAAGv7F,EAAE,MAAMC,CAAC,EAAE,OAAOL,EAAE,gBAAgB,EAAE,CAACI,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAI08G,IAAG,CAAC,WAAWr3F,GAAG,YAAY,QAAQ,WAAWo3F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYn8G,EAAEZ,EAAE,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,KAAK,YAAY,EAAE,IAAIK,EAAE,EAAE,OAAOwB,EAAE2zF,GAAG,EAAE,MAAM,EAAE1zF,EAAE0zF,GAAG,EAAE,MAAM,EAAEzzF,EAAE,GAAG,GAAG1B,IAAI,EAAE0B,EAAE,+BAA+B,CAAC,IAAIC,EAAE,EAAED,EAAE,EAAE,IAAI,CAAC,EAAEW,KAAKV,IAAI,EAAE,SAAS,EAAE,oBAAoBU,cAAcA,KAAK,UAAUV,EAAE,gBAAgBU,cAAcA,KAAK,EAAE,KAAK,GAAG,CAAC,CAAC,KAAK,SAAS;AAAA,QACpbb,aAAaA,KAAKjB;AAAA,QAClBiB,eAAeA,KAAK7B;AAAA;AAAA;AAAA,UAGlB8B;AAAA,yBACeC;AAAA;AAAA,KAEpB,CAAC,EAAE,SAASi7G,IAAIl9G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,QAAQ0B,EAAE,UAAUC,EAAE,QAAQC,EAAE,aAAaS,EAAE,YAAYC,EAAE,eAAeC,CAAC,EAAExC,EAAE,CAAC,iBAAiByC,EAAE,WAAWC,EAAE,WAAWkB,EAAE,UAAU,EAAE,cAAcqD,EAAE,MAAM,EAAE,IAAI5C,EAAE,QAAQC,CAAC,EAAEwxB,GAAG,UAAU71B,EAAE,MAAM,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,CAAC,EAAE+B,EAAE,GAAGX,EAAEW,EAAEw6F,GAAG,CAAC,OAAO,CAAC,EAAE9+F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,UAAU,GAAGuE,EAAE,CAACuf,EAAE,OAAOvmB,EAAE,MAAM,QAAQ,EAAE,IAAI,yCAAyCA,EAAE,MAAM,QAAQ,EAAE,IAAIiH,EAAE4uB,GAAG,gBAAgB,EAAEzxB,EAAEC,CAAC,EAAEI,EAAE4/F,GAAG,CAAC,OAAO,CAAC,EAAErkG,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM,EAAE,KAAKsH,CAAC,CAAC,CAAC,EAAE3C,EAAEw6F,GAAG,CAAC,OAAO,CAAC,EAAEr6F,CAAC,EAAE,QAAQ9E,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,EAAE9C,EAAE,8BAA8B8E,CAAC,CAAC,SAAS9E,EAAE,mBAAmB,CAACK,CAAC,CAAC,EAAE,CAAC,IAAIyE,EAAE9E,EAAE,SAASK,EAAE,MAAM,EAAEkH,EAAE8oB,GAAGhwB,EAAE,MAAMA,EAAE,MAAMyE,CAAC,EAAEC,EAAEs1F,GAAGx3F,EAAE0E,EAAE7C,EAAE,CAAC,EAAEC,EAAE3E,EAAE,eAAe8C,EAAEzC,EAAE,MAAM0E,EAAE,MAAM,CAAC,KAAK,CAAC,IAAID,EAAE,IAAIi4G,IAAG,EAAEr4G,EAAE7B,CAAC,EAAE8B,EAAE3E,EAAE,gBAAgB8E,EAAE,CAACzE,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAIuE,EAAEu6F,GAAG,CAAC,OAAO,CAAC,EAAEx6F,CAAC,EAAE,QAAQ3E,EAAE,MAAM,CAAC,MAAM8C,CAAC,CAAC,CAAC,EAAE,OAAO9C,EAAE,8BAA8B2E,CAAC,EAAEC,CAAC,CAAC,IAAIq4G,IAAG,CAAC,WAAWx4F,GAAG,YAAY,QAAQ,WAAWu4F,GAAG,EAAE,SAASE,IAAIp9G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,UAAUO,EAAE,YAAY,EAAE,QAAQ,EAAE,SAAS0B,EAAE,SAASC,EAAE,uBAAuBC,CAAC,EAAE7B,EAAE,CAAC,KAAKsC,EAAE,WAAWC,CAAC,EAAE,EAAEC,EAAE5C,EAAE,SAAS0C,EAAE,MAAM,EAAEG,EAAE7C,EAAE,SAAS2C,EAAE,MAAM,EAAE,CAACG,EAAEkB,CAAC,EAAEs2F,GAAG13F,EAAEC,EAAExC,EAAE,EAAE,EAAE0B,EAAEC,EAAEC,CAAC,EAAE,MAAM,CAACjC,EAAE,eAAe,CAAC8C,EAAE,MAAM,EAAE,SAASA,CAAC,EAAE9C,EAAE,eAAe2C,EAAE,MAAM,QAAQqB,CAAC,CAAC,CAAC,CAAC,IAAIm5G,IAAG,CAAC,WAAWz4F,GAAG,YAAY,QAAQ,WAAWw4F,GAAG,EAAE,SAASE,IAAIt9G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,UAAUO,CAAC,EAAED,EAAE,CAAC,MAAM,EAAE,UAAU,CAAC,EAAE,EAAE,GAAG,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,kCAAkC,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,0CAA0C,EAAE,OAAO,EAAE,IAAI2B,EAAE/B,EAAE,SAAS,EAAE,MAAM,EAAEgC,EAAEhC,EAAE,SAAS,EAAE,MAAM,EAAE,GAAG,CAACiC,EAAES,EAAEC,CAAC,EAAE43F,GAAGx4F,EAAEC,EAAE3B,CAAC,EAAEuC,EAAEF,EAAE,OAAO,MAAM,CAAC1C,EAAE,eAAe,CAAC4C,EAAE,CAAC,EAAE,QAAQX,CAAC,EAAEjC,EAAE,eAAe,CAAC4C,CAAC,EAAE,SAASF,CAAC,EAAE1C,EAAE,eAAe,CAAC,CAAC,EAAE,QAAQ,IAAI,WAAW2C,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI06G,IAAG,CAAC,WAAW14F,GAAG,YAAY,QAAQ,WAAWy4F,GAAG,EAAE,SAASE,IAAIx9G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,WAAWO,CAAC,EAAED,EAAE,CAAC,MAAM,CAAC,EAAE,EAAE,GAAG,EAAE,QAAQ,SAAS,MAAM,IAAI,MAAM,kCAAkC,EAAE,GAAGC,GAAG,EAAE,MAAM,IAAI,MAAM,sCAAsC,EAAE,IAAI,EAAEL,EAAE,SAAS,EAAE,MAAM,EAAE+B,EAAEy4F,GAAG,EAAEn6F,CAAC,EAAE,OAAOL,EAAE,eAAe,EAAE,MAAM,QAAQ+B,CAAC,CAAC,CAAC,IAAIw7G,IAAG,CAAC,WAAW34F,GAAG,YAAY,QAAQ,WAAW04F,GAAG,EAAME,IAAI,iBAAiBC,IAAIhf,GAAG,CAAC,UAAU+e,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW54F,GAAG,YAAY,QAAQ,WAAW24F,GAAG,EAAME,IAAI;AAAA;AAAA;AAAA,EAGr3EC,IAAInf,GAAG,CAAC,UAAUkf,GAAG,CAAC,EAAEE,IAAG,CAAC,WAAW94F,GAAG,YAAY,QAAQ,WAAW64F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYl9G,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,GAAG,EAAE,IAAI,EAAE,IAAI,MAAMY,EAAE,MAAM,EAAE,QAAQkB,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,GAAGlB,EAAEkB,GAAG9B,EAAE8B,GAAG,KAAK,YAAY,EAAE,KAAK,KAAK,EAAE,OAAO,IAAIzB,EAAEm1F,GAAG,KAAK,IAAI,EAAE3zF,EAAEk8G,IAAIn9G,CAAC,EAAE,KAAK,SAAS;AAAA;AAAA,UAElRP;AAAA,yBACewB;AAAA;AAAA,KAEpB,CAAC,EAAE,SAASk8G,IAAIj+G,EAAE,CAAC,IAAI,EAAEA,EAAE,OAAO,GAAG,EAAE,EAAE,MAAM,MAAM,iBAAiB,wBAAwB,EAAE,GAAG,IAAI,EAAE,MAAM,eAAeA,EAAE,MAAM,IAAIE,EAAE,CAAC,UAAU,UAAU,UAAU,UAAU,SAAS,EAAEI,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAOO,IAAID,EAAE,KAAK,QAAQJ,EAAEK,OAAOP,EAAEO,KAAK,EAAE,OAAOD,EAAE,KAAK,CAAC,CAAC,SAAS49G,GAAGl+G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,GAAGC,EAAE,QAAQ,UAAUA,EAAE,MAAM,OAAO,EAAE,CAAC,IAAI2B,EAAEhC,EAAE,SAASK,EAAE,MAAM,EAAE,EAAEA,EAAE,QAAQ,SAAS2B,EAAE,IAAIY,GAAGgkB,EAAE,aAAahkB,CAAC,CAAC,EAAEZ,EAAEU,EAAE2tB,GAAGhwB,EAAE,MAAMA,EAAE,MAAM,CAAC,EAAE,EAAEq6F,GAAGh4F,EAAE,CAAC,EAAE,OAAO1C,EAAE,eAAe,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,IAAI,EAAE,IAAI89G,IAAGz9G,EAAE,MAAM,CAAC,EAAE,OAAOL,EAAE,gBAAgB,EAAE,CAACK,CAAC,EAAEA,EAAE,KAAK,CAAC,CAAC,IAAI49G,IAAG,CAAC,WAAWj5F,GAAG,YAAY,QAAQ,WAAWg5F,EAAE,EAAME,IAAG,KAAK,CAAC,YAAYt9G,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,SAAS,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,IAAI,KAAK,KAAK,EAAE,CAAC,KAAK,YAAY,KAAK,KAAK,EAAE,CAAC,KAAK,cAAc,KAAK,OAAO,EAAE,CAAC,KAAK,MAAM,KAAK,KAAK,EAAE,CAAC,KAAK,MAAM,KAAK,KAAK,CAAC,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAwCp5B,CAAC,EAAEu9G,IAAG,KAAK,CAAC,YAAYv9G,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,SAAS,EAAE,KAAK,eAAe,CAAC,CAAC,KAAK,IAAI,KAAK,KAAK,EAAE,CAAC,KAAK,YAAY,KAAK,KAAK,EAAE,CAAC,KAAK,IAAI,KAAK,KAAK,CAAC,EAAE,KAAK,YAAYA,EAAE,KAAK,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAkC/L,CAAC,EAAE,SAASw9G,GAAGt+G,EAAE,EAAE,CAAC,IAAI,MAAMA,EAAE,8BAA8B,CAAC,CAAC,CAAC,SAASu+G,GAAGv+G,EAAE,CAAC,IAAI,EAAE,EAAE,KAAK,EAAEA,GAAG,GAAG,EAAE,OAAO,CAAC,CAAC,SAASw+G,IAAIx+G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,EAAE,EAAE,OAAO,CAAC,EAAED,EAAE2B,EAAEwY,EAAE,EAAE,UAAU,0CAA0C,EAAEvY,EAAEuY,EAAE,EAAE,UAAU,8BAA8B,EAAEtY,EAAE5B,EAAE,MAAMqC,EAAET,EAAEA,EAAE,OAAO,GAAG,GAAGjC,EAAE,mBAAmB,CAACK,CAAC,CAAC,GAAGqC,EAAEX,GAAG,EAAEC,EAAE,CAAC,IAAI+C,EAAE/E,EAAE,SAASK,EAAE,MAAM,EAAE,CAAC2E,EAAEC,CAAC,EAAE01F,GAAG51F,EAAE9C,EAAE5B,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,CAACL,EAAE,eAAegF,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,EAAEhF,EAAE,eAAeiF,EAAE,MAAMA,EAAE,MAAMA,EAAE,MAAM,CAAC,CAAC,CAAC,GAAG,IAAI,EAAE,OAAOhD,EAAEA,EAAE,OAAO,GAAG,EAAE,CAACjC,EAAE,eAAeiC,EAAE5B,EAAE,MAAM,CAAC,CAAC,EAAEL,EAAE,eAAeiC,EAAE,QAAQ,CAAC,CAAC,CAAC,EAAE,GAAGS,IAAI,EAAE,MAAM,CAACrC,EAAEusG,GAAG,CAAC,MAAM,CAAC,MAAM3qG,EAAE,MAAM,QAAQ,MAAM,CAAC,EAAE,QAAQjC,CAAC,CAAC,CAAC,EAAE,IAAI2C,EAAE3C,EAAE,QAAQ,IAAIK,EAAE,MAAM,EAAEuC,EAAED,IAAI,MAAMA,EAAE,SAASE,EAAED,EAAE5C,EAAE,aAAaK,CAAC,EAAEA,EAAE2D,EAAE4iB,EAAE,cAAc3kB,CAAC,EAAES,EAAEuB,EAAEk7F,GAAG,CAAC,OAAO,CAAC,EAAEt8F,CAAC,EAAE,MAAM,CAAC,MAAM,CAACmB,EAAEtB,CAAC,CAAC,EAAE,QAAQ1C,CAAC,CAAC,EAAE4C,GAAGw7G,GAAGp+G,EAAE6C,CAAC,EAAE,IAAIwE,EAAEg3G,GAAG,CAAC,EAAEx5G,EAAEw5G,GAAG37G,CAAC,EAAE+B,EAAE,KAAKC,EAAE,IAAID,IAAI,KAAK,CAACR,EAAEA,CAAC,EAAE,CAACA,EAAEQ,CAAC,EAAEE,EAAE,CAACI,EAAEC,EAAEC,IAAI,CAAC,IAAIC,EAAER,EAAE,EAAES,EAAE,IAAI+4G,IAAGj5G,CAAC,EAAE8C,EAAE,CAAC,CAACrF,CAAC,EAAE,CAAC+B,IAAI,KAAK,EAAE,CAAC,EAAE,CAAC,OAAO,iBAAiB,EAAE,CAACM,CAAC,EAAE,CAACC,CAAC,CAAC,EAAEyC,EAAEhD,EAAEA,EAAEzE,EAAE,gBAAgBmF,EAAED,EAAE,QAAQ6C,CAAC,EAAEq2G,GAAGp+G,EAAEyH,CAAC,CAAC,EAAE,QAAQ1C,EAAE,EAAEA,EAAEsC,EAAEtC,GAAG,EAAE,CAAC,IAAIC,EAAED,EAAE,EAAE,QAAQE,EAAEF,EAAEE,GAAG,EAAEA,GAAG,EAAEN,EAAEK,EAAEC,EAAE,CAACjB,EAAEa,CAAC,CAAC,CAAC,CAAC,QAAQE,EAAEF,EAAEE,EAAEsC,EAAEtC,GAAG,EAAE,CAAC,IAAIC,EAAEN,EAAE,EAAEO,EAAE,IAAIk5G,IAAG,CAACn6G,EAAEe,EAAE,CAAC,CAAC,EAAEI,EAAE,CAAC,CAACzC,CAAC,EAAE,CAAC+B,IAAI,KAAK,EAAE,CAAC,EAAE,CAAC4C,CAAC,CAAC,EAAEjC,EAAEX,EAAEA,EAAEzE,EAAE,gBAAgBiF,EAAED,EAAE,QAAQG,CAAC,EAAEi5G,GAAGp+G,EAAEoF,CAAC,EAAE,IAAI2C,EAAEV,EAAE,EAAEI,EAAEM,EAAE,EAAE,QAAQC,EAAED,EAAEC,GAAG,EAAEA,GAAG,EAAErD,EAAE8C,EAAEO,EAAEvD,EAAE,KAAK,CAAC,CAAC,IAAIG,EAAEH,EAAEA,EAAEigG,GAAG,CAAC,OAAO,CAAC,EAAEjgG,CAAC,EAAE,QAAQzE,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,CAACgE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAEo6G,GAAGp+G,EAAE4E,CAAC,EAAE,IAAI0C,EAAEgnG,GAAG,CAAC,OAAO,CAAC,EAAErqG,EAAE,QAAQQ,CAAC,EAAE,QAAQzE,EAAE,MAAM,CAAC,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,EAAEo+G,GAAGp+G,EAAEiE,CAAC,EAAE,IAAIa,EAAE7C,EAAE,MAAM,EAAE,EAAE,EAAE6C,EAAE,KAAK,CAAC,EAAEF,EAAEH,EAAEA,EAAE06F,GAAG,CAAC,OAAO,CAAC,EAAE16F,CAAC,EAAE,MAAM,CAAC,MAAMK,CAAC,EAAE,QAAQ9E,CAAC,CAAC,EAAEo+G,GAAGp+G,EAAE4E,CAAC,EAAE,IAAI2C,EAAED,EAAE,OAAOA,EAAE63F,GAAG,CAAC,OAAO,CAAC,EAAE73F,CAAC,EAAE,MAAM,CAAC,MAAMxC,CAAC,EAAE,QAAQ9E,CAAC,CAAC,EAAEo+G,GAAGp+G,EAAEuH,CAAC,EAAE,CAACD,EAAE7C,CAAC,CAAC,CAAC,IAAI85G,IAAG,CAAC,WAAWt5F,GAAG,YAAY,QAAQ,WAAWq5F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAY59G,EAAEZ,EAAE,EAAEK,EAAEwB,EAAEC,EAAE,CAAC,KAAK,cAAc,CAAC,QAAQ,YAAY,EAAE,KAAK,YAAYA,EAAE,IAAIC,EAAE,IAAI,UAAU,EAAE,EAAEC,EAAE,OAAO3B,EAAE,CAAC,IAAI,WAAW2B,EAAE,EAAE,MAAM,IAAI,UAAUA,EAAE,EAAE,MAAM,IAAI,OAAOA,EAAE,EAAE,MAAM,IAAI,UAAUA,EAAE,EAAE,MAAM,QAAQA,EAAE,EAAE,KAAK,CAAC,KAAK,SAAS;AAAA;AAAA;AAAA,mBAG92DA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAwBQA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAiBAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,4CAUiBpB,gCAAgCZ;AAAA;AAAA;AAAA,sCAGtC6B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sCAwBAA;AAAA;AAAA;AAAA;AAAA,mDAIa7B;AAAA,mDACAY;AAAA;AAAA,sBAE7BmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,SAwBb,CAAC,EAAE,SAAS08G,IAAI3+G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,WAAW,CAAC,EAAE,EAAE,CAAC,cAAc,EAAE,SAAS0B,EAAE,UAAUC,EAAE,YAAYC,CAAC,EAAE7B,EAAE,CAACsC,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAM,CAACyC,EAAEkB,CAAC,EAAE/B,GAAG,KAAKA,EAAE,CAACU,EAAEC,CAAC,EAAE,EAAE,CAACF,EAAEI,EAAEkB,EAAEnB,CAAC,EAAEwE,EAAE,IAAIm3G,IAAG77G,EAAEC,EAAE,EAAEb,EAAEC,EAAE,CAAC,EAAE,OAAOhC,EAAE,gBAAgBqH,EAAE,CAAChH,EAAE,CAAC,EAAE,SAAS,CAAC,CAAC,IAAIq+G,IAAG,CAAC,WAAWx5F,GAAG,YAAY,QAAQ,WAAWu5F,GAAG,EAAE,SAASE,IAAI7+G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAEL,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE6tF,GAAG,EAAE,QAAQ,EAAE,QAAQ,KAAK,YAAY,4DAA4D,EAAE,IAAI,EAAEztF,EAAE,SAAS,EAAE,MAAM,EAAE,CAAC,aAAa2B,EAAE,YAAYC,EAAE,QAAQC,CAAC,EAAE44F,GAAG,EAAEx6F,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,CAACD,EAAE,eAAe4B,EAAE,EAAE,MAAMD,CAAC,EAAE3B,EAAE,eAAe,CAAC6B,EAAE,MAAM,EAAE,QAAQA,CAAC,CAAC,CAAC,CAAC,IAAI28G,IAAG,CAAC,WAAWx5F,GAAG,YAAY,QAAQ,WAAWu5F,GAAG,EAAE,SAASE,IAAI/+G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAE,IAAI,GAAGC,EAAE,MAAM,QAAQ,IAAI,EAAEA,EAAE0B,EAAE,EAAE,MAAM,OAAOC,EAAE3B,EAAE,MAAM,GAAG4B,EAAE,IAAI,MAAMF,EAAE,CAAC,EAAEW,EAAE,EAAE,QAAQsB,EAAE,EAAEA,EAAEjC,EAAEiC,IAAIA,IAAI,IAAI/B,EAAES,KAAK,EAAE,MAAMsB,IAAI,IAAIrB,EAAE,CAAC,EAAEC,EAAE,IAAI,MAAMb,CAAC,EAAE,KAAK,CAAC,EAAEc,EAAE,EAAE,MAAM,MAAM,EAAEA,EAAE,GAAG,EAAE,IAAIC,EAAE,IAAI,MAAMd,CAAC,EAAE,QAAQgC,EAAE,EAAEA,EAAElB,EAAE,OAAOkB,IAAI,CAACpB,EAAE,GAAGoB,EAAE,IAAI,EAAE0gG,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ1kG,EAAE,MAAM,CAAC,MAAM4C,EAAE,KAAKC,CAAC,CAAC,CAAC,EAAEwE,EAAE83F,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQn/F,EAAE,MAAM,CAAC,MAAMiC,CAAC,CAAC,CAAC,EAAEa,EAAEkB,GAAGqD,EAAE1E,EAAE,KAAK,CAAC,CAAC,CAAC,OAAOA,EAAE,QAAQqB,GAAGhE,EAAE,8BAA8BgE,CAAC,CAAC,EAAElB,CAAC,CAAC,IAAIg8G,IAAG,CAAC,WAAWz5F,GAAG,YAAY,QAAQ,WAAWw5F,GAAG,EAAME,IAAG,KAAK,CAAC,YAAYn+G,EAAEZ,EAAE,CAAC,KAAK,cAAc,CAAC,IAAI,YAAY,EAAE,IAAI,EAAEY,EAAE,WAAWP,EAAEO,EAAE,UAAUiB,EAAEjB,EAAE,OAAOkB,EAAElB,EAAE,YAAYmB,EAAED,EAAE,KAAK,KAAKD,EAAE,CAAC,EAAE,KAAK,YAAY,CAACxB,EAAE0B,CAAC,EAAE,IAAIC,EAAE,MAAM,EAAE,WAAWU,EAAE,KAAK,MAAM,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAEE,EAAE;AAAA;AAAA,MAEh9CC,EAAE,GAAGhB,EAAE,EAAE,IAAIgB,EAAE;AAAA,oCACehB;AAAA;AAAA;AAAA,SAG3B,IAAIiB,EAAE,GAAGjB,EAAE,EAAE,IAAIiB,EAAE;AAAA,oCACQjB;AAAA;AAAA;AAAA,SAG3B,KAAK,SAAS;AAAA,0CACmBG;AAAA;AAAA;AAAA,UAGhCa;AAAA;AAAA;AAAA;AAAA;AAAA,UAKAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YASEhB,eAAe;AAAA,wDAC6BA;AAAA;AAAA;AAAA;AAAA,8BAI1BY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAgBlBE;AAAA;AAAA;AAAA,iCAGqBF;AAAA,cACnB,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAiBNE;AAAA,qBACS,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAebA;AAAA,qBACS,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAebA;AAAA;AAAA,oBAEQ;AAAA;AAAA,KAEf,CAAC,EAAE,SAASo8G,IAAIl/G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,WAAW,CAAC,EAAE,EAAE,CAAC,YAAY,CAAC,EAAED,EAAE2B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE,CAAC,EAAEC,EAAE,EAAES,EAAE43C,EAAE,mBAAmB,CAACr4C,CAAC,EAAEF,CAAC,EAAEY,EAAEtC,EAAEqC,GAAG,OAAOC,EAAEq9F,GAAG,CAAC,OAAO,CAAC,EAAE3/F,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,KAAK0C,CAAC,CAAC,CAAC,EAAEV,EAAE,KAAKW,CAAC,EAAEV,EAAEq4C,EAAE,iBAAiB,EAAEv4C,CAAC,EAAE,IAAI,IAAIa,EAAE03C,EAAE,aAAa,gBAAgB33C,EAAE,MAAMV,EAAE,CAAC,EAAEY,EAAE+jB,EAAE,cAAc,CAACjkB,EAAE,MAAMV,EAAE,CAAC,EAAEa,EAAEq8F,GAAG,CAAC,OAAO,CAAC,EAAEx8F,CAAC,EAAE,QAAQ3C,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG6C,CAAC,CAAC,CAAC,CAAC,EAAEb,EAAE,KAAKc,CAAC,EAAE,IAAIkB,EAAE0mB,GAAGrqB,EAAE,KAAK,EAAE,EAAE,CAACqE,EAAEC,EAAEC,EAAE0C,EAAExC,IAAI,CAAC,IAAIyC,EAAE7C,EAAE,MAAM,GAAGK,EAAEL,EAAE,MAAM,GAAGM,EAAEs1C,EAAE,aAAa,8BAA8Bv1C,EAAED,CAAC,EAAEG,EAAE,CAAC,WAAWD,EAAE,OAAOD,EAAE,UAAUwC,EAAE,YAAYzC,CAAC,EAAEI,EAAE,IAAI65G,IAAG95G,EAAEN,CAAC,EAAEQ,EAAEnF,EAAE,cAAckF,EAAE,CAACR,EAAEE,CAAC,EAAE0C,CAAC,EAAE,GAAGtF,EAAE,KAAKmD,CAAC,EAAEA,EAAE,MAAM,KAAKL,EAAE,OAAOK,EAAE,IAAIC,EAAE+xG,GAAG,CAAC,QAAQn3G,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK8E,EAAE,KAAK,EAAE,MAAM,SAAS,CAAC,CAAC,EAAEiD,EAAEi2G,GAAG,CAAC,OAAO,CAAC,EAAE54G,CAAC,EAAE,QAAQpF,EAAE,MAAM,CAAC,KAAK,CAAC+E,EAAEC,CAAC,CAAC,CAAC,CAAC,EAAE,OAAOhD,EAAE,KAAKoD,CAAC,EAAEpD,EAAE,KAAK+F,CAAC,EAAE,EAAE5C,EAAER,EAAEoD,EAAET,EAAExC,CAAC,CAAC,EAAEuC,EAAE,EAAEvE,EAAE,qBAAqB,EAAEkB,EAAE,CAAC,EAAE,EAAEm7F,GAAG,CAAC,OAAO,CAAC,EAAA93F,CAAC,EAAE,QAAQrH,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE6B,EAAE,EAAE,GAAG/B,GAAG,KAAK,CAACV,EAAE,KAAK,CAAC,EAAE,IAAI0C,EAAE41C,EAAE,uBAAuB53C,CAAC,EAAE+B,EAAEu7F,GAAG,CAAC,OAAO,CAACv7F,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,KAAK0E,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO1C,EAAE,QAAQ0C,GAAG1E,EAAE,8BAA8B0E,CAAC,CAAC,EAAED,CAAC,CAAC,IAAIw6G,IAAG,CAAC,WAAW35F,GAAG,YAAY,QAAQ,WAAW05F,GAAG,EAAME,IAAI,CAAC7e,GAAGG,GAAGG,GAAGG,GAAGG,GAAGI,GAAGE,GAAGE,GAAGO,GAAGE,GAAGG,GAAGG,GAAGG,GAAGI,GAAGG,GAAGI,GAAGE,GAAGI,GAAGE,GAAGE,GAAGI,GAAGS,GAAGE,GAAGE,GAAGS,GAAGG,GAAGI,GAAGlI,GAAGsI,GAAGS,GAAGS,GAAGM,GAAGE,GAAGE,GAAGE,GAAGE,GAAGG,GAAGG,GAAGG,GAAGO,GAAGE,GAAGE,GAAGG,GAAGI,GAAGI,GAAGE,GAAGG,GAAGG,GAAGE,IAAGI,IAAGI,IAAGI,IAAGG,IAAGI,IAAGE,IAAGG,IAAGI,IAAGG,IAAGE,IAAGG,IAAGI,IAAGG,IAAGK,IAAGE,IAAGG,IAAGI,IAAGI,IAAGI,IAAGlR,GAAGoR,IAAGvI,GAAG0I,IAAGG,IAAGG,IAAGvR,GAAG2R,IAAGI,IAAGE,IAAGI,IAAGG,IAAGI,IAAGG,IAAGI,IAAGI,IAAGG,IAAGG,IAAGI,IAAGE,IAAGE,IAAGI,IAAGE,IAAGE,IAAGE,IAAGE,IAAGI,IAAGI,IAAGI,IAAGY,IAAG3V,GAAG+V,IAAGG,IAAGG,IAAGG,IAAGrQ,GAAGwQ,IAAGI,IAAGE,IAAGI,IAAGI,IAAGnY,GAAGqY,IAAGE,IAAGE,IAAGE,IAAGE,IAAG9R,GAAG+O,IAAGkD,IAAGI,IAAGI,IAAG3Y,GAAG+Y,IAAGG,IAAGI,IAAGG,IAAGI,IAAGE,IAAGG,IAAGG,IAAGG,IAAGG,IAAGG,IAAGG,IAAGI,IAAGG,IAAGG,IAAGG,IAAGvW,GAAG+P,IAAG2G,IAAGE,IAAGE,IAAGE,IAAGE,IAAGE,IAAGE,IAAGE,IAAGG,IAAGG,IAAGG,IAAGE,IAAGG,IAAGE,IAAGE,IAAGE,IAAG/I,IAAGzU,GAAG2d,IAAGG,IAAGI,IAAGM,IAAGG,IAAGze,GAAG2e,IAAGE,IAAGG,IAAGnJ,GAAE,EAAE,QAAQh2G,KAAKo/G,IAAI34F,GAAGzmB,CAAC,EAAE,IAAIq/G,IAAI,SAASr/G,EAAE,CAACA,EAAEA,EAAE,QAAQ,GAAG,UAAUA,EAAEA,EAAE,MAAM,GAAG,QAAQA,EAAEA,EAAE,KAAK,GAAG,OAAOA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,UAAU,GAAG,WAAW,GAAGq/G,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,IAAI,SAASt/G,EAAE,CAACA,EAAEA,EAAE,OAAO,GAAG,SAASA,EAAEA,EAAE,KAAK,GAAG,OAAOA,EAAEA,EAAE,MAAM,GAAG,QAAQA,EAAEA,EAAE,MAAM,GAAG,QAAQA,EAAEA,EAAE,UAAU,GAAG,YAAYA,EAAEA,EAAE,QAAQ,GAAG,UAAUA,EAAEA,EAAE,IAAI,GAAG,KAAK,GAAGs/G,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,SAASC,IAAIx/G,EAAE,CAACu/G,GAAGv/G,EAAE,KAAK,MAAM8lB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,QAAQ,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS25F,IAAIz/G,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,GAAG1B,EAAE,QAAQ,WAAW,EAAE,QAAQ,UAAU,MAAM,IAAI,MAAM,6DAA6D,EAAE,GAAG,CAAC,WAAW2B,EAAE,WAAWC,EAAE,WAAWS,EAAE,eAAeC,CAAC,EAAEvC,EAAEwC,EAAE5C,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAGwC,EAAE7C,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG8C,EAAE,EAAE,GAAG,GAAG,KAAK,CAAC,IAAIgC,EAAE9E,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG8E,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,uDAAuDA,EAAE,MAAM,SAAS,EAAEhC,EAAEgC,EAAE,EAAE,CAAC,IAAId,EAAEjC,GAAG,KAAK,EAAE/B,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,GAAG,EAAEq9G,GAAG18G,GAAG,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,GAAGA,qEAAqE,EAAE,IAAI2E,EAAErF,EAAE3B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG,EAAE4B,EAAE,EAAE,MAAM,GAAG,EAAE,MAAM,GAAGwC,EAAEowB,GAAG,2BAA2Bx0B,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,CAAC,EAAEqE,EAAE1E,EAAE,WAAW,CAAC,GAAGyE,EAAE4C,EAAE,CAAC,EAAEhH,EAAE,KAAK,EAAEsE,EAAE3E,EAAE,UAAU,IAAI0E,EAAE,MAAM,EAAE,GAAGE,EAAE,IAAI,WAAW,IAAI,WAAWvE,EAAE,KAAK,EAAE,MAAM,EAAEiH,EAAE,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO+3G,GAAGz8G,EAAEgC,EAAEvE,EAAE,MAAM,OAAOwC,EAAEyE,EAAE,EAAE,MAAM,OAAOtF,EAAEC,EAAE,EAAEa,EAAEkB,EAAErB,GAAG,EAAEgC,CAAC,EAAED,CAAC,CAAC,IAAI86G,IAAG,CAAC,WAAW55F,GAAG,YAAY,OAAO,UAAU05F,IAAI,WAAWC,GAAG,EAAE,SAASE,GAAG3/G,EAAE,EAAE,CAAC,IAAIE,EAAE,SAASI,EAAE,EAAE,CAACJ,EAAE,EAAE,KAAK,MAAMF,EAAE,KAAK,CAAC,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASO,EAAE,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAE0B,CAAC,CAAC,EAAE,EAAEC,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGE,EAAE,EAAE,WAAWF,EAAE,MAAM,GAAGA,EAAE,KAAK,EAAEW,EAAE,EAAE,UAAU,IAAIT,EAAE,MAAM,EAAE,GAAG,OAAO2kB,EAAE,cAAc3kB,EAAE,KAAK,IAAI,GAAGjC,EAAEgC,EAAEm9G,GAAGp9G,EAAE,OAAOW,CAAC,EAAET,CAAC,CAAC,MAAM,CAAC,WAAWnC,EAAE,YAAY,OAAO,UAAUM,EAAE,WAAWC,CAAC,CAAC,CAAC,IAAIq/G,IAAGD,GAAGzkG,EAAE,EAAE,SAAS2kG,GAAG7/G,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAE,SAASC,EAAE,EAAE,CAACD,EAAE,EAAE,KAAK,MAAMN,EAAE,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,QAAQ,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,GAAG,CAAC,QAAQiC,EAAE,OAAOC,CAAC,EAAE,EAAE,CAAC,EAAEC,EAAE,EAAES,CAAC,EAAEV,EAAEW,EAAEZ,EAAE,UAAU,IAAIE,EAAE,MAAM,EAAE,GAAGW,EAAEb,EAAE,UAAU,IAAIW,EAAE,MAAM,EAAE,GAAGG,EAAE7C,GAAG,KAAKA,EAAEiC,EAAE,MAAMa,EAAEw3C,EAAE,2BAA2Br4C,EAAE,MAAMS,EAAE,KAAK,EAAEsB,EAAEjC,EAAE,WAAWe,EAAED,CAAC,EAAE,GAAG+jB,EAAE,cAAc9jB,CAAC,IAAI,EAAE,OAAOkB,EAAE,IAAI,EAAE,IAAI,WAAW,IAAI,WAAW/B,EAAE,KAAK,EAAE,MAAM,EAAEoF,EAAE,IAAI,WAAW,IAAI,WAAW3E,EAAE,KAAK,EAAE,MAAM,EAAE,EAAEX,EAAE,UAAU,IAAIiC,EAAE,MAAM,EAAE,GAAG,OAAW5D,EAAEuC,EAAE,EAAEV,EAAE,MAAM,OAAOW,EAAEyE,EAAE3E,EAAE,MAAM,OAAOy8G,GAAGl9G,EAAE,OAAO,CAAC,EAAK+B,CAAC,CAAC,MAAM,CAAC,WAAWlE,EAAE,YAAY,OAAO,UAAUO,EAAE,WAAW,CAAC,CAAC,CAAC,IAAIu/G,IAAI,GAAGC,IAAGF,GAAGxkG,GAAGykG,GAAG,EAAME,GAAG,SAASC,IAAIjgH,EAAE,CAACggH,GAAGhgH,EAAE,KAAK,MAAMsb,GAAG,KAAK,CAAC,QAAQ,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS4kG,IAAIlgH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAEJ,EAAE,WAAW,EAAE,GAAG,MAAM,EAAE,GAAG,KAAK,EAAE,GAAG4mB,EAAE,cAAcxmB,EAAE,KAAK,IAAI,EAAE,OAAOA,EAAE,IAAIC,EAAE,EAAE,IAAI0B,GAAG/B,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,EAAE,EAAE,EAAE,IAAI,WAAW,IAAI,WAAW1B,CAAC,EAAE,MAAM,EAAE,EAAEL,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG,OAAO0/G,GAAG,EAAEz/G,EAAE,OAAO8+G,GAAG/+G,EAAE,OAAO,CAAC,EAAEA,CAAC,CAAC,IAAI6/G,IAAG,CAAC,WAAW7kG,GAAG,YAAY,OAAO,UAAU2kG,IAAI,WAAWC,GAAG,EAAE,SAASE,GAAGpgH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQE,CAAC,EAAEF,EAAE,GAAG,EAAE,QAAQ,SAAS,OAAOusB,GAAGrsB,EAAE,SAAS,EAAE,MAAM,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE,IAAII,EAAEJ,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAEK,EAAEL,EAAE,mBAAmB,CAAC,EAAE,OAAOA,EAAE,mBAAmBI,CAAC,EAAE,IAAIC,CAAC,EAAED,CAAC,CAAC,IAAI+/G,IAAG,CAAC,WAAW/gG,GAAG,YAAY,OAAO,WAAW8gG,EAAE,EAAME,GAAG,SAASC,IAAIvgH,EAAE,CAACsgH,GAAGtgH,EAAE,KAAK,MAAMqlB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,SAAS,QAAQ,QAAQ,CAAC,CAAC,CAAC,SAASm7F,GAAGxgH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAACO,EAAE,CAAC,EAAEkgH,IAAI,EAAE,EAAE,MAAMngH,EAAE,IAAI,EAAE,EAAE,GAAG,QAAQ0C,EAAE,EAAEA,EAAE,EAAE,OAAOA,IAAI,EAAEA,KAAKA,IAAI,EAAE,IAAI,IAAIf,EAAEy+G,IAAI,EAAE,EAAE,MAAMpgH,EAAE,IAAI,EAAE4B,EAAE,CAAC,OAAO,EAAE,EAAE,OAAO,MAAM3B,EAAE,MAAM,EAAE,EAAE,KAAK,EAAE,GAAG,EAAE,CAAC,IAAIyC,EAAEo9G,GAAG,CAAC,OAAO,EAAE,QAAQlgH,CAAC,CAAC,EAAE,OAAO8C,EAAE,MAAMf,EAAEe,CAAC,CAAC,IAAIb,EAAEjC,EAAE,WAAW+B,EAAEC,EAAE,KAAK,EAAEU,EAAE1C,EAAE,UAAU,IAAIgC,EAAE,MAAM,EAAE,GAAGW,EAAE3C,EAAE,UAAU,IAAIiC,EAAE,MAAM,EAAE,GAAGW,EAAE,IAAI,WAAW,IAAI,WAAW,CAAC,EAAE,MAAM,EAAEC,EAAE,IAAI,WAAW,IAAI,WAAWb,EAAE,KAAK,EAAE,MAAM,EAAE,OAAOo+G,GAAG19G,EAAEG,EAAEb,EAAE,MAAM,OAAOm9G,GAAGn9G,EAAE,OAAOW,EAAEC,EAAE,EAAE,MAAM,EAAEX,CAAC,CAAC,SAASu+G,IAAI1gH,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,MAAMF,EAAE,MAAM,EAAE,QAAQM,EAAE,EAAEA,EAAEJ,EAAE,OAAOI,IAAIJ,EAAEI,GAAGN,EAAE,EAAEM,IAAI,OAAOJ,CAAC,CAAC,SAASugH,IAAIzgH,EAAE,EAAE,CAAC,IAAIE,EAAE,CAAC,EAAEI,EAAE,CAAC,EAAE,QAAQC,EAAE,EAAEA,EAAEP,EAAE,OAAO,EAAEO,EAAEP,EAAEO,KAAK,GAAGL,EAAE,KAAKF,EAAEO,EAAE,EAAEP,EAAE,EAAEO,MAAM,GAAGD,EAAE,KAAK,EAAEC,EAAE,EAAE,QAAQA,EAAE,EAAEA,EAAED,EAAE,OAAO,EAAEC,EAAE,CAAC,IAAI,EAAE,GAAG,QAAQ,EAAE,EAAE,EAAED,EAAE,OAAO,EAAE,EAAEA,EAAE,IAAIC,IAAI,IAAI,IAAID,EAAE,GAAGA,EAAE,MAAM,EAAE,GAAGA,EAAE,GAAGC,CAAC,CAAC,MAAM,CAACL,EAAEI,CAAC,CAAC,CAAC,IAAIqgH,IAAG,CAAC,WAAWt7F,GAAG,YAAY,OAAO,WAAWm7F,GAAG,UAAUD,GAAG,EAAE,SAASK,GAAG5gH,EAAE,EAAEE,EAAE,CAAC,IAAII,EAAEN,EAAE,MAAMO,EAAEP,EAAE,MAAM,OAAO,EAAE8mB,EAAE,eAAe,EAAExmB,CAAC,EAAE,EAAE,EAAE2B,EAAEu4C,EAAE,mBAAmB,EAAEj6C,CAAC,EAAE2B,EAAE,KAAKC,EAAE,GAAG,GAAGF,GAAG,KAAK,CAAC,IAAIW,EAAE,IAAI,MAAMrC,CAAC,EAAE,QAAQwC,EAAE,EAAEA,EAAEH,EAAE,OAAOG,IAAIH,EAAEG,GAAGzC,EAAE2B,EAAEc,IAAI,EAAEy3C,EAAE,iBAAiB,EAAE,OAAOj6C,CAAC,EAAE2B,EAAEs+G,GAAG,CAAC,OAAO,CAAC,EAAExgH,CAAC,EAAE,MAAM,CAAC,KAAKiC,CAAC,EAAE,QAAQ/B,CAAC,CAAC,EAAE,IAAI2C,EAAE3C,EAAE,UAAU,IAAIF,EAAE,MAAM,EAAE,GAAGE,EAAE,UAAU,IAAIgC,EAAE,MAAM,EAAE,KAAKW,IAAIV,EAAE,GAAG,CAAC,MAAM,CAAC,WAAWD,EAAE,aAAa,EAAE,KAAK,EAAE,mBAAmBC,CAAC,CAAC,CAAC,IAAI0+G,GAAG,SAASC,IAAI9gH,EAAE,CAAC6gH,GAAG7gH,EAAE,KAAK,MAAMub,GAAG,KAAK,CAAC,wBAAwB,CAAC,CAAC,CAAC,SAASwlG,IAAI/gH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAEgC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,CAAC,WAAWU,EAAE,KAAK,EAAE,aAAaE,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAE,GAAGwC,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG,EAAEA,EAAEV,EAAEyC,CAAC,CAAC,IAAI3B,EAAE,EAAE,MAAM,OAAOw3C,EAAE,2BAA2B,MAAM,EAAEx3C,CAAC,EAAE,GAAG,CAACkB,EAAEC,CAAC,EAAEq2C,EAAE,0BAA0B,EAAE,MAAM,CAAC,EAAEjzC,EAAEuf,EAAE,cAAc3iB,CAAC,EAAEY,EAAE,EAAE,WAAWb,EAAE,EAAE,KAAK,EAAE,GAAG4iB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIniB,EAAE,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG87G,GAAG3+G,EAAEqF,EAAE5C,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqBz1C,EAAE,MAAMjC,CAAC,EAAEiC,EAAE,MAAMJ,CAAC,CAAC,OAAOI,CAAC,CAAC,IAAIi8G,IAAG,CAAC,WAAWzlG,GAAG,YAAY,OAAO,UAAUulG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIlhH,EAAE,CAACihH,GAAGjhH,EAAE,KAAK,MAAMwb,GAAG,KAAK,CAAC,wBAAwB,CAAC,CAAC,CAAC,SAAS2lG,IAAInhH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAEgC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,CAAC,WAAWU,EAAE,KAAK,EAAE,aAAaE,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAE,GAAGwC,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG,EAAEA,EAAEV,EAAEyC,CAAC,CAAC,IAAI3B,EAAE,EAAE,MAAM,OAAOw3C,EAAE,2BAA2B,MAAM,EAAEx3C,CAAC,EAAE,GAAG,CAACkB,EAAEC,CAAC,EAAEq2C,EAAE,0BAA0B,EAAE,MAAM,CAAC,EAAEjzC,EAAEuf,EAAE,cAAc3iB,CAAC,EAAEY,EAAE,EAAE,WAAWb,EAAE,EAAE,KAAK,EAAE,GAAG4iB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIniB,EAAE,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAGk8G,GAAG/+G,EAAEqF,EAAE5C,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqBz1C,EAAE,MAAMjC,CAAC,EAAEiC,EAAE,MAAMJ,CAAC,CAAC,OAAOI,CAAC,CAAC,IAAIq8G,IAAG,CAAC,WAAW5lG,GAAG,YAAY,OAAO,UAAU0lG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIthH,EAAE,CAACqhH,GAAGrhH,EAAE,KAAK,MAAMyb,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS8lG,IAAIvhH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+B,EAAE,EAAEC,EAAE,EAAE,CAAC,WAAWC,EAAE,KAAKS,EAAE,mBAAmBC,CAAC,EAAE+9G,GAAG,EAAErgH,EAAE,CAAC,EAAE,GAAGsC,EAAE,CAAC,IAAI0E,EAAE,EAAE,UAAU,IAAIpF,EAAE,MAAM,EAAE,GAAGoF,IAAI,IAAIrF,EAAEC,EAAEF,EAAEsF,EAAE,CAAC,IAAIzE,EAAEZ,EAAE,MAAM,MAAM,EAAE,EAAE,EAAEa,EAAE,EAAE,WAAWD,EAAE,OAAO,EAAEE,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGmB,EAAE4iB,EAAE,cAAc/jB,EAAE,KAAK,EAAE,EAAEb,EAAE,MAAMU,EAAE,IAAI,OAAOy+G,GAAGp/G,EAAEo9G,GAAGn9G,EAAE,OAAOgC,EAAE,EAAElB,CAAC,EAAEH,GAAG,EAAE,YAAYV,EAAE,MAAM,EAAEY,CAAC,CAAC,IAAIy+G,IAAG,CAAC,WAAW/lG,GAAG,YAAY,OAAO,WAAW8lG,IAAI,UAAUD,GAAG,EAAMG,GAAG,SAASC,IAAI1hH,EAAE,CAACyhH,GAAGzhH,EAAE,KAAK,MAAMgc,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS2lG,IAAI3hH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAEO,EAAE,EAAE,EAAE,EAAED,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,gBAAgBC,CAAC,EAAEjC,EAAE0C,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE0B,EAAE,EAAEC,EAAEC,CAAC,EAAEU,EAAED,EAAE,aAAaE,EAAEF,EAAE,YAAYG,EAAEH,EAAE,QAAQ,IAAII,EAAEJ,EAAE,QAAQ,MAAMsB,EAAEtB,EAAE,QAAQ,OAAO,EAAEA,EAAE,QAAQ,KAAK2E,EAAE3E,EAAE,aAAa,EAAEA,EAAE,YAAY+B,EAAE/B,EAAE,WAAW,GAAGA,EAAE,aAAa,eAAe,MAAM,IAAI,MAAM,6CAA6CA,EAAE,yCAAyC,EAAE,GAAGA,EAAE,gBAAgB,GAAGA,EAAE,iBAAiB,EAAE,MAAM,IAAI,MAAM,0EAA0EA,EAAE,mBAAmBA,EAAE,iBAAiB,EAAE,IAAIgC,EAAEtE,EAAE,WAAWsC,EAAE,SAAS,SAAS,EAAEiC,EAAEvE,EAAE,UAAU,IAAIsE,EAAE,MAAM,EAAE,GAAG,OAAO68G,GAAG,EAAElhH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGsC,EAAEC,EAAEC,EAAEC,EAAEkB,EAAE,EAAEqD,EAAE,EAAE5C,EAAEE,CAAC,EAAED,CAAC,CAAC,IAAIg9G,IAAG,CAAC,WAAW5lG,GAAG,YAAY,OAAO,UAAU0lG,IAAI,WAAWC,GAAG,EAAE,SAASE,GAAG7hH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,EAAEM,CAAC,EAAE,EAAE,CAAC,MAAMC,CAAC,EAAEL,EAAE,EAAE4mB,EAAE,cAAcxmB,EAAE,KAAK,EAAE,EAAEwmB,EAAE,uBAAuBvmB,EAAE,CAAC,EAAE,OAAOumB,EAAE,OAAO,IAAIA,EAAE,cAAc,CAAC,EAAE,IAAI,cAAc,iBAAiBxmB,EAAE,uEAAuE,EAAEN,EAAE,QAAQ,OAAOM,EAAE,MAAM,EAAE,CAAC,OAAOA,EAAE,OAAO,MAAM,EAAE,MAAMA,EAAE,KAAK,CAAC,CAAC,IAAIwhH,IAAG,CAAC,WAAWl/F,GAAG,YAAY,OAAO,WAAWi/F,EAAE,EAAME,GAAG,SAASC,IAAIhiH,EAAE,CAAC+hH,GAAG/hH,EAAE,KAAK,MAAMoc,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,QAAQ,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS6lG,IAAIjiH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,EAAE,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,WAAW0B,CAAC,EAAE3B,EAAE,GAAGC,EAAE,QAAQ,WAAW,EAAE,QAAQ,UAAU,MAAM,IAAI,MAAM,4DAA4D,EAAE,IAAI2B,EAAE3B,EAAE,MAAM,OAAO4B,EAAE,EAAE,MAAM,OAAOS,EAAE,EAAErC,EAAE,MAAM2B,EAAE,GAAG3B,EAAE,MAAM2B,EAAE,GAAGW,EAAEZ,EAAE,EAAE,MAAME,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGW,EAAE,EAAEvC,EAAE,MAAM2B,EAAE,GAAG3B,EAAE,MAAM2B,EAAE,GAAGa,EAAEd,EAAE,EAAE,MAAME,EAAE,GAAG,EAAE,MAAMA,EAAE,GAAGa,EAAEzC,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE2D,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE4iB,EAAE,cAAc9jB,CAAC,EAAEuE,EAAEuf,EAAE,cAAc5iB,CAAC,EAAES,EAAEowB,GAAG,2BAA2Bx0B,EAAE,MAAM,MAAM,EAAE,EAAE,EAAE,EAAE,MAAM,MAAM,EAAE,EAAE,CAAC,EAAE,OAAO,CAACuC,EAAEC,CAAC,CAAC,EAAE+jB,EAAE,OAAOlkB,IAAIC,EAAE,IAAI,kCAAkCD,WAAWC,6BAA6BtC,EAAE,aAAa,EAAE,wBAAwB,oBAAoB0B,eAAe,EAAE,IAAI2C,EAAE,EAAE,CAAC,EAAEhC,EAAEE,CAAC,EAAE,CAAC,EAAEA,EAAEF,CAAC,EAAEiC,EAAE5C,EAAE,CAACsF,EAAExE,EAAEF,CAAC,EAAE,CAAC0E,EAAE1E,EAAEE,CAAC,EAAE+B,EAAE+8G,GAAG,CAAC,OAAO,CAAC,EAAEthH,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAM0E,CAAC,CAAC,CAAC,EAAE4C,EAAEq6G,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQ3hH,EAAE,MAAM,CAAC,MAAM2E,CAAC,CAAC,CAAC,EAAEG,EAAE9E,EAAE,UAAU,IAAI4E,EAAE,MAAM,EAAE,GAAG2C,EAAEvH,EAAE,UAAU,IAAIsH,EAAE,MAAM,EAAE,GAAGvC,EAAE,EAAEH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGI,EAAEjD,EAAEuF,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGrC,EAAE,KAAK,IAAI,EAAEoC,CAAC,EAAEnC,EAAElF,EAAE,WAAW,CAACiF,EAAEF,EAAEC,CAAC,EAAEJ,EAAE,KAAK,EAAEO,EAAEnF,EAAE,UAAU,IAAIkF,EAAE,MAAM,EAAE,GAAGE,EAAE,IAAI,WAAW,IAAI,WAAWR,EAAE,KAAK,EAAE,MAAM,EAAEmD,EAAE,IAAI,WAAW,IAAI,WAAWT,EAAE,KAAK,EAAE,MAAM,EAAE,OAAOu6G,GAAG/8G,EAAEM,EAAER,EAAE,MAAM,OAAO2C,EAAEQ,EAAET,EAAE,MAAM,OAAO,EAAEvF,EAAEoD,CAAC,EAAEnF,EAAE,YAAY4E,EAAE,MAAM,EAAE5E,EAAE,YAAYsH,EAAE,MAAM,EAAEpC,EAAE,MAAMT,EAAES,CAAC,CAAC,IAAI88G,IAAG,CAAC,WAAW9lG,GAAG,YAAY,OAAO,UAAU4lG,IAAI,WAAWC,GAAG,EAAE,SAASE,GAAGniH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,MAAM,CAAC,MAAME,EAAE,KAAKI,CAAC,EAAE,QAAQC,CAAC,EAAEP,EAAE,CAAC,EAAE,CAAC,EAAEo2B,GAAG,iBAAiB,EAAEl2B,EAAEI,CAAC,EAAE2B,EAAEm0B,GAAG,iBAAiB,EAAE,MAAM,EAAE,CAAC,EAAEl0B,EAAE3B,EAAE,SAAS,EAAE,MAAM,EAAE4B,EAAE5B,EAAE,WAAW,EAAE,EAAE,KAAK,EAAEqC,EAAEkkB,EAAE,eAAe,EAAE,KAAK,EAAEjkB,EAAEtC,EAAE,UAAU,IAAI4B,EAAE,MAAM,EAAE,GAAGF,EAAE,CAAC,IAAIe,EAAEozB,GAAG,kBAAkB,EAAExzB,CAAC,EAAE,OAAO,EAAE,QAAQ,SAASC,EAAE,YAAYX,EAAE,MAAMc,EAAEA,EAAE8jB,EAAE,cAAc,CAAC,CAAC,EAAEvmB,EAAE,mBAAmB4B,CAAC,EAAE,IAAID,EAAE,SAASc,EAAEA,EAAE8jB,EAAE,cAAc,CAAC,CAAC,CAAC,EAAE3kB,CAAC,CAAC,GAAG,EAAE,QAAQ,SAAS,CAAC,IAAIa,EAAE8wE,GAAG5xE,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE,KAAK,EAAE,OAAOW,EAAE,YAAYG,EAAEb,CAAC,CAAC,IAAIW,EAAEvC,EAAE,mBAAmB4B,CAAC,EAAEY,EAAE,EAAE,MAAM,OAAO,GAAGA,IAAI,EAAEq/G,IAAIlgH,EAAEU,EAAE,GAAGE,EAAE,EAAE,CAAC,UAAUC,IAAI,EAAEs/G,IAAIngH,EAAEU,EAAE,GAAGA,EAAE,GAAGE,EAAE,EAAE,CAAC,UAAUC,IAAI,EAAEu/G,IAAIpgH,EAAEU,EAAE,GAAGA,EAAE,GAAGA,EAAE,GAAGE,EAAE,EAAE,CAAC,MAAM,CAAC,IAAIE,EAAE8wE,GAAG5xE,EAAE,EAAE,EAAE,EAAE,MAAM,EAAE,KAAK,EAAEY,EAAE,IAAIE,CAAC,CAAC,CAAC,OAAOb,CAAC,CAAC,SAASigH,IAAIpiH,EAAE,EAAEE,EAAEI,EAAEC,EAAE,CAAC,IAAI,EAAE,EAAE,EAAED,EAAE,GAAG2B,EAAE3B,EAAE,GAAG4B,EAAE,EAAE3B,EAAE,GAAG,QAAQ4B,EAAE,EAAEA,EAAED,EAAEC,IAAI,CAAC,IAAIS,EAAET,EAAE,EAAEF,EAAE/B,EAAE,IAAIF,EAAE,SAAS4C,EAAEA,EAAErC,EAAE,EAAE,EAAE,CAAC,EAAE,GAAGA,EAAE,EAAE,CAAC,CAAC,SAAS8hH,IAAIriH,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,CAAC,IAAI,EAAE,EAAE0B,EAAE1B,EAAE,GAAG2B,EAAE3B,EAAE,GAAG4B,EAAE5B,EAAE,GAAGqC,EAAEX,EAAE,EAAE,GAAGY,EAAEX,EAAE,EAAE,GAAG,QAAQY,EAAEb,EAAEa,EAAEF,EAAEE,IAAI,QAAQC,EAAEb,EAAEa,EAAEF,EAAEE,IAAI,CAAC,IAAIC,EAAEF,EAAE,EAAEC,EAAE7C,EAAEiC,EAAE7B,EAAE,IAAIN,EAAE,SAASgD,EAAEA,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,GAAG,EAAE,EAAE,CAAC,CAAC,SAASs/G,IAAItiH,EAAE,EAAEE,EAAEI,EAAEC,EAAE,EAAE,EAAE,CAAC,IAAI0B,EAAE,EAAEC,EAAE,EAAE,GAAGC,EAAE,EAAE,GAAGS,EAAE,EAAE,GAAGC,EAAEX,EAAE,EAAE,GAAGY,EAAEX,EAAE,EAAE,GAAGY,EAAEH,EAAE,EAAE,GAAGI,EAAE,EAAE,GAAG,QAAQkB,EAAEhC,EAAEgC,EAAErB,EAAEqB,IAAI,QAAQ,EAAE/B,EAAE,EAAEW,EAAE,IAAI,QAAQyE,EAAE3E,EAAE2E,EAAExE,EAAEwE,IAAI,CAAC,IAAI,EAAErD,EAAE,EAAE,EAAEhE,EAAEqH,EAAEjH,EAAE0C,EAAEzC,EAAE,IAAIP,EAAE,SAAS,EAAE,EAAE,EAAE,EAAE,EAAEiC,CAAC,EAAEA,GAAG,EAAE,EAAE,CAAC,CAAC,IAAIsgH,IAAG,CAAC,WAAW9+F,GAAG,YAAY,OAAO,WAAW0+F,EAAE,EAAE,SAASK,IAAIxiH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,MAAM,CAAC,EAAED,EAAE2B,EAAE,EAAE,OAAO,CAACsF,EAAE,IAAIA,EAAE,CAAC,EAAErF,EAAEs4C,EAAE,YAAYj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEE,EAAEq4C,EAAE,YAAYt4C,EAAE,OAAO,EAAE,MAAM,EAAEU,EAAE43C,EAAE,oBAAoBj6C,EAAE,MAAM,EAAE0B,CAAC,EAAEY,EAAE23C,EAAE,oBAAoB,EAAE,EAAE,MAAM,EAAE13C,EAAE03C,EAAE,aAAa53C,EAAE,EAAE,EAAE,MAAM,EAAEG,EAAE8+G,GAAG,CAAC,OAAO,CAAC,EAAEthH,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,MAAMgC,CAAC,CAAC,CAAC,EAAEc,EAAEw9G,GAAG,CAAC,OAAO,CAAC,EAAEz9G,CAAC,EAAE,QAAQ7C,EAAE,MAAM,CAAC,KAAKiC,CAAC,CAAC,CAAC,EAAE+B,EAAE29G,GAAG,CAAC,OAAO,CAAC,EAAE7+G,CAAC,EAAE,QAAQ9C,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAE,EAAEu/G,GAAG,CAAC,OAAO,CAAC,EAAEj+G,CAAC,EAAE,QAAQhE,EAAE,MAAM,CAAC,MAAM2C,EAAE,KAAKC,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,YAAY6C,EAAE,MAAM,EAAE7C,EAAE,YAAY8C,EAAE,MAAM,EAAE9C,EAAE,YAAY6C,EAAE,MAAM,EAAE,CAAC,CAAC,IAAI0/G,IAAG,CAAC,WAAWpmG,GAAG,YAAY,OAAO,WAAWmmG,GAAG,EAAE,SAASE,GAAG1iH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,MAAM,CAAC,MAAME,CAAC,EAAE,QAAQI,CAAC,EAAEN,EAAEO,EAAED,EAAE,WAAW,EAAE,MAAMJ,CAAC,EAAE,EAAEI,EAAE,mBAAmB,CAAC,EAAE,OAAOA,EAAE,mBAAmBC,CAAC,EAAE,IAAI,CAAC,EAAEA,CAAC,CAAC,IAAIoiH,IAAG,CAAC,WAAWlmG,GAAG,YAAY,OAAO,WAAWimG,EAAE,EAAME,IAAGjD,GAAGjjG,EAAE,EAAMmmG,GAAG,SAASC,IAAI9iH,EAAE,CAAC6iH,GAAG7iH,EAAE,KAAK,MAAM2c,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASomG,IAAI/iH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,aAAa,EAAE,aAAa,CAAC,EAAED,EAAE2B,EAAE/B,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG2B,EAAEhC,EAAE,WAAWK,EAAE,MAAMA,EAAE,KAAK,EAAE4B,EAAEjC,EAAE,UAAU,IAAIgC,EAAE,MAAM,EAAE,GAAG,OAAO2gH,GAAG5gH,EAAE,EAAE,EAAEE,CAAC,EAAED,CAAC,CAAC,IAAI8gH,IAAG,CAAC,WAAWrmG,GAAG,YAAY,OAAO,UAAUmmG,IAAI,WAAWC,GAAG,EAAE,SAASE,GAAGjjH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAEwmB,EAAE,eAAe9mB,EAAE,MAAM,KAAK,EAAE,GAAG,KAAK,EAAE,GAAGO,EAAE,EAAE,IAAIwC,GAAGA,EAAE,KAAK,EAAEy3C,EAAE,uBAAuBj6C,EAAED,CAAC,EAAE,IAAI,EAAEk6C,EAAE,gBAAgB,EAAE,IAAIz3C,GAAGA,EAAE,KAAK,EAAEzC,CAAC,EAAE,EAAE,EAAE,OAAOyC,GAAG+jB,EAAE,cAAc/jB,EAAE,KAAK,EAAE,CAAC,EAAE,GAAG,EAAE,SAAS,EAAE,OAAOq9G,GAAG,CAAC,OAAO,CAAC,EAAE,EAAE,EAAE,EAAE,QAAQlgH,CAAC,CAAC,EAAE,IAAI+B,EAAE/B,EAAE,WAAW,EAAE,EAAE,GAAG,KAAK,EAAE,GAAG4mB,EAAE,cAAc,CAAC,IAAI,EAAE,OAAO7kB,EAAE,GAAG,EAAE,GAAG,QAAQ,SAAS,CAAC,IAAIc,EAAE,EAAE,IAAI4B,GAAG,CAAC,IAAIC,EAAEkiB,EAAE,cAAcniB,EAAE,MAAM,MAAMrE,CAAC,CAAC,EAAE,OAAOuhH,GAAG,CAAC,OAAO,CAACl9G,CAAG,EAAE,QAAQzE,EAAE,MAAM,CAAC,MAAM,CAAC,GAAG0E,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE5B,EAAED,EAAE,IAAI4B,IAAI,CAAC,KAAKzE,EAAE,SAASyE,EAAE,MAAM,EAAE,MAAMA,EAAE,KAAK,EAAE,EAAE,EAAE61C,EAAE,gBAAgBz3C,EAAE,IAAI4B,GAAGA,EAAE,KAAK,EAAE,CAAC,EAAE,IAAIT,EAAEnB,EAAE,GAAG,MAAM,KAAK,EAAE,EAAEmvE,GAAGlvE,EAAE,EAAE,EAAE,GAAG,MAAMkB,CAAC,EAAEqD,EAAEizC,EAAE,gBAAgB,EAAE,IAAI71C,GAAGA,EAAE,KAAK,EAAErE,CAAC,EAAE2B,EAAE,MAAMsF,EAAE,IAAI,EAAErH,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,OAAO,EAAE,YAAYu4C,EAAE,uBAAuB,CAAC,EAAEz3C,EAAE,QAAQ4B,GAAGzE,EAAE,YAAYyE,EAAE,MAAM,CAAC,EAAE1C,CAAC,CAAC,IAAIC,EAAE4kB,EAAE,cAAc,EAAE,GAAG,MAAM,MAAM,EAAExmB,CAAC,CAAC,EAAE6B,EAAE,EAAES,EAAE,EAAE,IAAIG,GAAG,CAAC,IAAIC,EAAE8jB,EAAE,cAAc/jB,EAAE,MAAM,MAAMzC,CAAC,CAAC,EAAE,OAAO6B,GAAGa,EAAEA,CAAC,CAAC,EAAEH,EAAE,EAAE,IAAIE,GAAG7C,EAAE,mBAAmB6C,CAAC,CAAC,EAAED,EAAE5C,EAAE,mBAAmB+B,CAAC,EAAE,QAAQc,EAAE,EAAEA,EAAEb,EAAEa,IAAI,CAAC,IAAIC,EAAED,EAAEZ,EAAE,QAAQ+B,EAAE,EAAEA,EAAErB,EAAE,OAAOqB,IAAI,CAAC,IAAI,EAAEtB,EAAEsB,GAAGqD,EAAExE,EAAE,EAAE,EAAEF,EAAEqB,GAAG,SAASqD,EAAEA,EAAE,CAAC,EAAEzE,EAAE,IAAI,EAAEE,CAAC,EAAEA,GAAG,CAAC,CAAC,CAAC,OAAOf,CAAC,CAAC,IAAIihH,IAAG,CAAC,WAAWpmG,GAAG,YAAY,OAAO,WAAWmmG,EAAE,EAAME,GAAG,SAASC,IAAIpjH,EAAE,CAACmjH,GAAGnjH,EAAE,KAAK,MAAM+c,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASsmG,IAAIrjH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,EAAED,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG0B,EAAE3B,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,CAAC,QAAQ4B,EAAE,UAAUC,EAAE,IAAIS,EAAE,gBAAgBC,EAAE,WAAWC,CAAC,EAAE5C,EAAE6C,EAAEy3C,EAAE,wBAAwB13C,CAAC,EAAEE,EAAEw3C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEC,EAAES,EAAEC,EAAE,GAAGE,CAAC,EAAEmB,EAAElB,EAAE,aAAa,EAAEA,EAAE,YAAYuE,EAAEvE,EAAE,QAAQ,IAAI,EAAEA,EAAE,QAAQ,MAAM2B,EAAE3B,EAAE,QAAQ,OAAO4B,EAAE5B,EAAE,QAAQ,KAAK6B,EAAE7B,EAAE,eAAe8B,EAAE9B,EAAE,cAAcwE,EAAExE,EAAE,aAAagC,EAAEhC,EAAE,YAAYyE,EAAEzE,EAAE,WAAWiC,EAAEjC,EAAE,YAAYkC,EAAElC,EAAE,QAAQ,OAAO,OAAO,EAAE,EAAE,GAAGA,EAAE,aAAa,eAAe,MAAM,IAAI,MAAM,oDAAoDA,EAAE,yCAAyC,EAAE,IAAImC,EAAE7E,EAAE,WAAW0C,EAAE,SAAS,SAAS,EAAEoC,EAAE9E,EAAE,UAAU,IAAI6E,EAAE,MAAM,EAAE,GAAG,OAAOg+G,GAAG,EAAE5iH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG0B,EAAEiC,EAAE,EAAEqD,EAAE,EAAE5C,EAAEC,EAAEM,EAAEL,EAAEC,EAAE0C,EAAExC,EAAEyC,EAAExC,EAAEG,CAAC,EAAED,CAAC,CAAC,IAAIm+G,IAAG,CAAC,WAAWvmG,GAAG,YAAY,OAAO,UAAUqmG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIxjH,EAAE,CAACujH,GAAGvjH,EAAE,KAAK,MAAMid,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASwmG,IAAIzjH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,GAAGO,EAAE,OAAO,CAAC,EAAEL,EAAE,CAAC,QAAQ,EAAE,IAAI+B,EAAE,WAAWC,EAAE,gBAAgBC,EAAE,WAAWS,CAAC,EAAEtC,EAAEuC,EAAE,EAAEC,EAAE03C,EAAE,wBAAwBt4C,CAAC,EAAEa,EAAEy3C,EAAE,kBAAkB53C,EAAE,EAAE,MAAM,EAAEC,EAAEZ,EAAEE,EAAE,GAAGW,CAAC,EAAE,CAAC,UAAUE,EAAE,aAAakB,EAAE,YAAY,EAAE,WAAWqD,EAAE,SAAS,EAAE,QAAQ5C,EAAE,YAAYC,EAAE,UAAUC,EAAE,SAASC,EAAE,aAAa0C,EAAE,YAAYxC,CAAC,EAAEjC,EAAE0E,EAAEvD,EAAE,EAAEnB,EAAE,QAAQ,IAAIkC,EAAE,EAAE,EAAElC,EAAE,QAAQ,KAAKmC,EAAEnC,EAAE,aAAa,eAAeoC,EAAE2hB,EAAE,eAAe/jB,EAAE,OAAO,EAAEqC,EAAE0hB,EAAE,eAAevmB,EAAE,KAAK,EAAE,CAAC8E,EAAEC,EAAE2C,CAAC,EAAE6e,EAAE,eAAe,EAAE,KAAK,EAAEnf,EAAExC,EAAE,GAAG+C,EAAEhD,EAAEC,EAAE,GAAGA,EAAE,GAAGgD,EAAGjD,EAAEC,EAAE,GAAG,EAAEiD,EAAGlD,EAAE,EAAEC,EAAE,GAAGkD,GAAGjD,EAAE,GAAGkD,GAAGpD,EAAEE,EAAE,GAAGA,EAAE,GAAGmD,GAAGrD,EAAEE,EAAE,GAAG,EAAEuB,GAAGzB,EAAE,EAAEE,EAAE,GAAGqD,GAAG,EAAE,WAAW1F,EAAE,QAAQ,SAAS,EAAE2F,GAAG,EAAE,UAAU,IAAID,GAAG,MAAM,EAAE,GAAGE,GAAG,EAAE,UAAU,IAAIpI,EAAE,MAAM,EAAE,GAAGqI,GAAG,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,OAAO26G,GAAG56G,GAAGC,GAAG5F,EAAEkB,EAAE,EAAE,EAAES,EAAE4C,EAAE1C,EAAEC,EAAEF,EAAE4C,EAAExC,EAAEyC,EAAExC,EAAEI,EAAEC,EAAE2C,EAAEN,EAAEO,EAAEC,EAAGC,EAAGC,GAAGC,GAAGC,GAAG5B,GAAG+B,EAAE,EAAED,EAAE,CAAC,IAAIi7G,IAAG,CAAC,WAAWzmG,GAAG,YAAY,OAAO,UAAUumG,IAAI,WAAWC,GAAG,EAAME,IAAGhE,GAAGtiG,EAAE,EAAMumG,IAAGjE,GAAGriG,EAAE,EAAMumG,IAAI,SAAS7jH,EAAE,CAACA,EAAEA,EAAE,SAAS,GAAG,WAAWA,EAAEA,EAAE,QAAQ,GAAG,SAAS,GAAG6jH,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,SAASC,IAAI/jH,EAAE,CAAC8jH,GAAG9jH,EAAE,KAAK,MAAMyd,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,QAAQ,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASumG,IAAIhkH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,EAAE,mBAAmB,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,MAAM2B,EAAE,MAAMC,EAAE,OAAOC,CAAC,EAAEjC,EAAE0C,EAAEV,EAAE,MAAM,GAAG,CAACW,EAAEC,CAAC,EAAE,EAAEC,EAAE,CAACH,EAAEC,EAAEC,EAAEb,EAAE,MAAM,EAAE,EAAEe,EAAE,EAAE,UAAU,IAAIf,EAAE,MAAM,EAAEiC,EAAEjC,EAAE,QAAQ,YAAYiC,EAAEw+G,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAEzgH,CAAC,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAEe,EAAE,EAAE,UAAU,IAAIkB,EAAE,MAAM,GAAG,IAAI,EAAElB,EAAE,GAAGuE,EAAE,EAAE,UAAU,IAAIrF,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAGwC,EAAE,EAAE,WAAW5B,EAAE,SAAS,EAAE6B,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGE,EAAE,IAAI,WAAW,IAAI,WAAW5C,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO6hH,GAAG,EAAEv8G,EAAE,EAAE3E,EAAEiC,EAAEhC,EAAEC,EAAE+gH,GAAGtjH,GAAG,EAAEqE,CAAC,EAAEV,GAAG,MAAM,EAAE,YAAYA,EAAE,MAAM,EAAES,CAAC,CAAC,IAAIs/G,IAAG,CAAC,WAAWxmG,GAAG,YAAY,OAAO,UAAUsmG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAInkH,EAAE,CAACkkH,GAAGlkH,EAAE,KAAK,MAAMud,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS6mG,IAAIpkH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAE4B,EAAE3B,EAAE,MAAM,OAAOumB,EAAE,OAAOvmB,EAAE,QAAQ,WAAWA,EAAE,QAAQ,QAAQ,IAAI,4BAA4BA,EAAE,mCAAmC,EAAE,IAAI4B,EAAEq4C,EAAE,mBAAmB,CAAC,CAAC,EAAEt4C,CAAC,EAAEU,EAAErC,EAAE4B,IAAI,OAAOS,EAAE49G,GAAG,CAAC,OAAO,CAAC,EAAEjgH,CAAC,EAAE,MAAM,CAAC,KAAK4B,CAAC,EAAE,QAAQjC,CAAC,CAAC,GAAG,IAAI2C,EAAE23C,EAAE,iBAAiB,EAAEt4C,CAAC,EAAE,GAAGs4C,EAAE,2BAA2B,UAAU,CAAC33C,CAAC,EAAEX,CAAC,EAAE,IAAIY,EAAE5C,EAAE,WAAW0C,EAAE,MAAMA,EAAE,KAAK,EAAEG,EAAEH,EAAE,MAAMC,GAAGG,EAAE9C,EAAE,UAAU,IAAI0C,EAAE,MAAM,EAAE,GAAGsB,EAAEhE,EAAE,UAAU,IAAI4C,EAAE,MAAM,EAAE,GAAGohH,GAAGlhH,EAAE,EAAE,EAAE,EAAEf,EAAE,EAAE,EAAEc,EAAEmB,EAAEm7G,GAAG9+G,EAAE,MAAM,EAAE,IAAI,EAAEuC,EAAE,GAAGX,IAAI,KAAK,CAAC,IAAIoF,EAAEizC,EAAE,uBAAuBr4C,CAAC,EAAE,EAAEq+G,GAAG,CAAC,OAAO,CAAC,EAAE19G,CAAC,EAAE,MAAM,CAAC,KAAKyE,CAAC,EAAE,QAAQrH,CAAC,CAAC,EAAEA,EAAE,YAAY0C,EAAE,MAAM,EAAE1C,EAAE,YAAY4C,EAAE,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,IAAIuhH,IAAG,CAAC,WAAW9mG,GAAG,YAAY,OAAO,UAAU4mG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIvkH,EAAE,CAACskH,GAAGtkH,EAAE,KAAK,MAAMwd,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASgnG,IAAIxkH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,EAAE,UAAU,EAAE,QAAQ0B,CAAC,EAAE3B,EAAE4B,EAAE3B,EAAE,MAAM,OAAOumB,EAAE,OAAOvmB,EAAE,QAAQ,WAAWA,EAAE,QAAQ,QAAQ,IAAI,2BAA2BA,EAAE,mCAAmC,EAAE,IAAI4B,EAAEq4C,EAAE,mBAAmB,CAAC,CAAC,EAAEt4C,CAAC,EAAEU,EAAErC,EAAE4B,IAAI,OAAOS,EAAE49G,GAAG,CAAC,OAAO,CAAC,EAAEjgH,CAAC,EAAE,MAAM,CAAC,KAAK4B,CAAC,EAAE,QAAQjC,CAAC,CAAC,GAAG,IAAI2C,EAAE23C,EAAE,iBAAiB,EAAEt4C,CAAC,EAAE,GAAGs4C,EAAE,2BAA2B,SAAS,CAAC33C,CAAC,EAAEX,CAAC,EAAE,IAAIY,EAAE5C,EAAE,WAAW0C,EAAE,MAAMA,EAAE,KAAK,EAAEG,EAAEH,EAAE,MAAMC,GAAGG,EAAE9C,EAAE,UAAU,IAAI0C,EAAE,MAAM,EAAE,GAAGsB,EAAEhE,EAAE,UAAU,IAAI4C,EAAE,MAAM,EAAE,GAAGwhH,GAAGthH,EAAE,EAAE,EAAE,EAAEf,EAAE,EAAE,EAAEc,EAAEmB,EAAEm7G,GAAG9+G,EAAE,MAAM,EAAE,IAAI,EAAEuC,EAAE,GAAGX,IAAI,KAAK,CAAC,IAAIoF,EAAEizC,EAAE,uBAAuBr4C,CAAC,EAAE,EAAEq+G,GAAG,CAAC,OAAO,CAAC,EAAE19G,CAAC,EAAE,MAAM,CAAC,KAAKyE,CAAC,EAAE,QAAQrH,CAAC,CAAC,EAAEA,EAAE,YAAY0C,EAAE,MAAM,EAAE1C,EAAE,YAAY4C,EAAE,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI2hH,IAAG,CAAC,WAAWjnG,GAAG,YAAY,OAAO,UAAU+mG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI3kH,EAAE,CAAC0kH,GAAG1kH,EAAE,KAAK,MAAM2d,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,SAAS,QAAQ,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASinG,IAAI5kH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAEL,EAAE,CAAC,UAAU,EAAE,WAAW,CAAC,EAAEI,EAAE2B,EAAE1B,EAAE,MAAM,GAAG2B,EAAE,IAAI,OAAO3B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG4B,EAAE,IAAI,OAAO5B,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGqC,EAAE,IAAI,OAAOrC,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGsC,EAAEX,EAAE,EAAEY,EAAEX,EAAE,EAAEY,EAAEH,GAAG,EAAE,GAAGI,EAAE,IAAI,OAAO,CAACf,EAAEY,EAAEC,EAAEC,CAAC,EAAE,CAACd,EAAEc,EAAEF,EAAEC,CAAC,EAAEoB,EAAE,EAAE,WAAWlB,EAAE,SAAS,EAAEuE,EAAE,EAAE,UAAU,IAAIhH,EAAE,MAAM,EAAE,GAAGwE,EAAE,IAAI,WAAW,IAAI,WAAW+hB,EAAE,eAAevmB,EAAE,KAAK,CAAC,EAAE,MAAM,EAAEoE,EAAE,IAAI,WAAW,IAAI,WAAW3B,CAAC,EAAE,MAAM,EAAE4B,EAAE,IAAI,WAAW,IAAI,WAAWkiB,EAAE,eAAe9jB,CAAC,CAAC,EAAE,MAAM,EAAE6B,EAAE,EAAE,UAAU,IAAIX,EAAE,MAAM,EAAE,GAAG,OAAOwgH,GAAGn9G,EAAE,EAAE,IAAI,OAAO,EAAE,EAAExC,EAAExE,EAAE,MAAM,OAAO,EAAEoE,EAAEC,EAAE5B,EAAE,OAAO6B,CAAC,EAAEX,CAAC,CAAC,IAAI2gH,IAAG,CAAC,WAAWlnG,GAAG,YAAY,OAAO,UAAUgnG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI/kH,EAAE,CAAC8kH,GAAG9kH,EAAE,KAAK,MAAM4d,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASonG,IAAIhlH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAE,EAAE,EAAED,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG0B,EAAE3B,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,CAAC,QAAQ4B,EAAE,UAAUC,EAAE,IAAIS,EAAE,gBAAgBC,CAAC,EAAE3C,EAAE4C,EAAEX,GAAG,KAAK,CAAC,EAAE,CAAC,EAAEA,EAAEY,EAAEy3C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEY,EAAEF,EAAEC,EAAE,EAAE,EAAEG,EAAED,EAAE,aAAamB,EAAEnB,EAAE,YAAY,EAAEA,EAAE,QAAQ,IAAIwE,EAAExE,EAAE,QAAQ,MAAM,EAAEA,EAAE,QAAQ,OAAO4B,EAAE5B,EAAE,QAAQ,KAAK6B,EAAE7B,EAAE,eAAe8B,EAAE9B,EAAE,cAAc+B,EAAE/B,EAAE,aAAayE,EAAEzE,EAAE,YAAYiC,EAAEjC,EAAE,WAAW0E,EAAE1E,EAAE,YAAYkC,EAAElC,EAAE,QAAQ,OAAO,OAAO,EAAE,EAAE,GAAGA,EAAE,aAAa,eAAe,MAAM,IAAI,MAAM,mEAAmEA,EAAE,yCAAyC,EAAE,IAAImC,EAAE5E,EAAE,WAAWyC,EAAE,SAAS,SAAS,EAAEoC,EAAE7E,EAAE,UAAU,IAAI4E,EAAE,MAAM,EAAE,GAAG,OAAO4/G,GAAG,EAAEvkH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAG0B,EAAEe,EAAEkB,EAAE,EAAEqD,EAAE,EAAE5C,EAAEM,EAAEL,EAAEC,EAAEC,EAAE0C,EAAExC,EAAEyC,EAAEtC,CAAC,EAAED,CAAC,CAAC,IAAI+/G,IAAG,CAAC,WAAWrnG,GAAG,YAAY,OAAO,UAAUmnG,IAAI,WAAWC,GAAG,EAAME,IAAGvF,GAAGthG,EAAE,EAAM8mG,IAAI,GAAGC,IAAGvF,GAAGrhG,GAAG2mG,IAAI,MAAM,EAAME,IAAG1F,GAAGlhG,GAAG,SAAS,EAAE,SAAS6mG,GAAGtlH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,IAAI,CAAC,EAAEL,EAAE,EAAEK,EAAE,MAAM,OAAO0B,EAAE1B,EAAE,MAAM,MAAM,EAAE2B,EAAE,EAAE,OAAO,EAAE,IAAI4kB,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,IAAI,iCAAiC,EAAE,EAAE,OAAO,IAAI,EAAE5kB,EAAE,EAAE,EAAE,GAAGD,EAAE,OAAOC,EAAE,EAAE,CAAC,EAAE2/G,GAAG,CAAC,OAAO,CAAC,EAAEthH,CAAC,EAAE,QAAQD,EAAE,MAAM,CAAC,MAAM2B,CAAC,CAAC,CAAC,CAAC,CAAC,IAAIsjH,IAAG,CAAC,WAAW7mG,GAAG,YAAY,OAAO,WAAW4mG,EAAE,EAAE,SAASE,GAAGxlH,EAAE,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,EAAE,MAAME,EAAE,MAAMI,CAAC,EAAE,QAAQC,CAAC,EAAEP,EAAE,EAAEO,EAAE,WAAW,EAAED,CAAC,EAAE,OAAOC,EAAE,mBAAmB,CAAC,EAAE,KAAKL,CAAC,EAAE,CAAC,CAAC,IAAIulH,IAAG,CAAC,WAAW5mG,GAAG,YAAY,OAAO,WAAW2mG,EAAE,EAAME,GAAG,SAASC,IAAI3lH,EAAE,CAAC0lH,GAAG1lH,EAAE,KAAK,MAAM8e,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS8mG,IAAI5lH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,MAAMM,CAAC,EAAE,EAAEC,EAAEL,EAAE,WAAWI,EAAE,MAAMA,EAAE,KAAK,EAAE,EAAEJ,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG,EAAEJ,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG,CAAC0B,EAAEC,EAAEC,EAAES,CAAC,EAAEtC,EAAE,MAAM,OAAOolH,GAAG,EAAEzjH,EAAEC,EAAEC,EAAES,EAAE,CAAC,EAAErC,CAAC,CAAC,IAAIslH,IAAG,CAAC,WAAW/mG,GAAG,YAAY,OAAO,WAAW8mG,IAAI,UAAUD,GAAG,EAAMG,IAAGnG,GAAG5gG,EAAE,EAAMgnG,IAAI,GAAGC,IAAGnG,GAAG7gG,GAAG+mG,GAAG,EAAME,GAAG,SAASC,IAAIlmH,EAAE,CAACimH,GAAGjmH,EAAE,KAAK,MAAMif,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASknG,IAAInmH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,gBAAgBO,CAAC,EAAED,EAAE,CAAC,EAAE,EAAE,KAAK,EAAE,SAAS2B,EAAE,OAAOC,EAAE,MAAMC,CAAC,EAAEjC,EAAE0C,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAE,EAAE,UAAU,IAAIb,EAAE,MAAM,EAAE,GAAGc,EAAEb,GAAG,KAAK,EAAE,UAAU,IAAIA,EAAE,MAAM,EAAE,GAAG,EAAEc,EAAEb,GAAG,KAAK,EAAE,UAAU,IAAIA,EAAE,MAAM,EAAE,GAAG,EAAE+B,EAAE,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAE,GAAG4iB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,OAAO5iB,EAAE,IAAI,EAAE,EAAE,UAAU,IAAIA,EAAE,MAAM,EAAE,GAAG,OAAO+hH,GAAGrjH,EAAEC,EAAEC,EAAEC,EAAEC,EAAEzC,EAAE,CAAC,EAAE2D,CAAC,CAAC,IAAIkiH,IAAG,CAAC,WAAWnnG,GAAG,YAAY,OAAO,UAAUinG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAItmH,EAAE,CAACqmH,GAAGrmH,EAAE,KAAK,MAAM+lB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASwgG,IAAIvmH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,UAAUS,EAAE,WAAWC,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAE9C,EAAEgE,EAAEs2C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEU,EAAET,EAAEW,CAAC,EAAE,EAAEw8G,GAAGv8G,GAAG,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,GAAGA,qEAAqE,EAAE,IAAIwE,EAAEjH,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG,EAAED,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGqE,EAAET,EAAE,YAAYU,EAAE,EAAE,GAAG,GAAG,KAAK,CAAC,IAAI2D,GAAGjI,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGiI,GAAG,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sDAAsDA,GAAG,MAAM,SAAS,EAAE,GAAGA,GAAG,MAAM,KAAK5D,EAAE,MAAM,IAAI,MAAM,2BAA2B4D,GAAG,wDAAwD5D,IAAI,EAAEC,EAAE2D,GAAG,EAAE,CAAC,IAAI1D,EAAEX,EAAE,aAAaY,EAAEZ,EAAE,YAAYsD,EAAEtD,EAAE,QAAQ,IAAIc,EAAEd,EAAE,QAAQ,MAAMuD,EAAEvD,EAAE,QAAQ,OAAOe,EAAEf,EAAE,QAAQ,KAAKgB,EAAEhB,EAAE,eAAeiB,EAAEjB,EAAE,cAAckB,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,WAAW+D,EAAE/D,EAAE,QAAQ,OAAO,OAAO,EAAE,EAAEyD,EAAEzD,EAAE,UAAUgE,EAAEhE,EAAE,SAASiE,EAAGjE,EAAE,QAAQ,GAAGrB,IAAI,OAAO,MAAM,IAAI,MAAM,yDAAyDA,wBAAwB,EAAE,IAAIuF,EAAG9H,EAAE,WAAW4D,EAAE,SAAS,SAAS,EAAEmE,GAAG/H,EAAE,UAAU,IAAI8H,EAAG,MAAM,EAAE,GAAGE,GAAGrG,GAAG,KAAK,EAAE3B,EAAE,UAAU,IAAI2B,EAAE,MAAM,EAAE,GAAG,OAAOokH,GAAG9+G,EAAEI,EAAEO,EAAEC,EAAG,EAAEtD,EAAEC,EAAEF,EAAE4C,EAAExC,EAAEyC,EAAExC,EAAEgD,EAAE/C,EAAEC,EAAEC,EAAEC,EAAEC,EAAEX,EAAE,EAAE2D,GAAGtF,GAAG,EAAEqF,EAAE,EAAED,CAAE,CAAC,IAAIo+G,IAAG,CAAC,WAAWzgG,GAAG,YAAY,OAAO,UAAUugG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI1mH,EAAE,CAACymH,GAAGzmH,EAAE,KAAK,MAAMgmB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS2gG,IAAI3mH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB0B,CAAC,EAAE,EAAE,CAAC,QAAQC,EAAE,IAAIC,EAAE,UAAUS,EAAE,WAAWC,EAAE,gBAAgBC,EAAE,WAAWC,EAAE,eAAeC,CAAC,EAAE9C,EAAEgE,EAAEs2C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE,MAAM2B,EAAEU,EAAET,EAAEW,EAAE,EAAE,EAAE,EAAEw8G,GAAGv8G,GAAG,GAAG,GAAG,KAAK,MAAM,IAAI,MAAM,GAAGA,8EAA8E,EAAE,IAAIwE,EAAEjH,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG,EAAED,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGqE,EAAET,EAAE,YAAYU,EAAE,EAAE,GAAG,GAAG,KAAK,CAAC,IAAI2D,GAAGjI,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGiI,GAAG,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,+DAA+DA,GAAG,MAAM,SAAS,EAAE,GAAGA,GAAG,MAAM,KAAK5D,EAAE,MAAM,IAAI,MAAM,oCAAoC4D,GAAG,wDAAwD5D,IAAI,EAAEC,EAAE2D,GAAG,EAAE,CAAC,IAAI1D,EAAEX,EAAE,aAAaY,EAAEZ,EAAE,YAAYsD,EAAEtD,EAAE,QAAQ,IAAIc,EAAEd,EAAE,QAAQ,MAAMuD,EAAEvD,EAAE,QAAQ,OAAOe,EAAEf,EAAE,QAAQ,KAAKgB,EAAEhB,EAAE,eAAeiB,EAAEjB,EAAE,cAAckB,EAAElB,EAAE,aAAamB,EAAEnB,EAAE,YAAYoB,EAAEpB,EAAE,WAAW+D,EAAE/D,EAAE,QAAQ,OAAO,OAAO,EAAE,EAAEyD,EAAEzD,EAAE,UAAUgE,EAAEhE,EAAE,SAASiE,EAAGjE,EAAE,QAAQ,GAAGrB,IAAI,OAAO,MAAM,IAAI,MAAM,kEAAkEA,wBAAwB,EAAE,IAAIuF,EAAG9H,EAAE,WAAW4D,EAAE,SAAS,SAAS,EAAEmE,GAAG/H,EAAE,UAAU,IAAI8H,EAAG,MAAM,EAAE,GAAGE,GAAGrG,GAAG,KAAK,EAAE3B,EAAE,UAAU,IAAI2B,EAAE,MAAM,EAAE,GAAG,OAAOwkH,GAAGl/G,EAAEI,EAAEO,EAAEC,EAAG,EAAEtD,EAAEC,EAAEF,EAAE4C,EAAExC,EAAEyC,EAAExC,EAAEgD,EAAE/C,EAAEC,EAAEC,EAAEC,EAAEC,EAAEX,EAAE,EAAE2D,GAAGtF,GAAG,EAAEqF,EAAE,EAAED,CAAE,CAAC,IAAIw+G,IAAG,CAAC,WAAW5gG,GAAG,YAAY,OAAO,UAAU0gG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI9mH,EAAE,CAAC6mH,GAAG7mH,EAAE,KAAK,MAAMmf,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,QAAQ,CAAC,CAAC,CAAC,SAAS4nG,IAAI/mH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,CAAC,EAAEF,EAAE,CAAC,OAAOM,EAAE,QAAQC,CAAC,EAAEL,EAAE,CAAC,EAAE,EAAE+B,EAAEC,CAAC,EAAE4zB,GAAG,mBAAmBx1B,EAAEC,CAAC,EAAE4B,EAAE,EAAE,WAAW,EAAE7B,EAAE,KAAK,EAAE,GAAG,IAAI,EAAE,OAAO6B,EAAE,IAAIS,EAAErC,EAAE,MAAMsC,EAAED,EAAEA,EAAE,OAAO,GAAGG,EAAE,EAAE,UAAU,IAAIzC,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAG4D,EAAE,IAAI,WAAW,IAAI,WAAWjC,CAAC,EAAE,MAAM,EAAEqF,EAAE,EAAE,UAAU,IAAIpF,EAAE,MAAM,EAAE,GAAG,OAAO0kH,GAAG9jH,EAAEs8G,GAAG/+G,EAAE,OAAO,EAAE,EAAEuC,EAAEZ,EAAEkC,EAAEoD,CAAC,EAAEpF,CAAC,CAAC,IAAI6kH,IAAG,CAAC,WAAW7nG,GAAG,YAAY,OAAO,UAAU2nG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIlnH,EAAE,CAACinH,GAAGjnH,EAAE,KAAK,MAAM,SAAS,KAAK,CAAC,SAAS,SAAS,QAAQ,SAAS,SAAS,SAAS,QAAQ,QAAQ,CAAC,CAAC,CAAC,SAASmnH,IAAInnH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,QAAQ,CAAC,EAAEL,EAAE,CAAC,KAAK,EAAE,UAAU+B,CAAC,EAAE3B,EAAE4B,EAAE4kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG4B,EAAE,EAAE,SAAS,EAAE,MAAM,EAAES,EAAErC,EAAE,MAAM2B,GAAG,QAAQuF,EAAE,EAAEA,EAAEtF,EAAE,OAAO,EAAEsF,EAAE,CAAC,IAAIxC,EAAE9C,EAAEsF,GAAGqf,EAAE,OAAO7hB,GAAGrC,EAAE,GAAGqC,GAAG,EAAE,IAAI,6BAA6BA,mBAAmBrC,EAAE,IAAI,CAAC,CAAC,IAAIC,EAAE23C,EAAE,aAAa,yBAAyBj6C,EAAE,EAAE2B,EAAED,CAAC,EAAEa,EAAE++G,GAAG,CAAC,OAAO,CAAC,EAAEthH,CAAC,EAAE,MAAM,CAAC,MAAM,CAACsC,EAAE,UAAUA,EAAE,UAAUA,EAAE,QAAQA,EAAE,SAAS,CAAC,EAAE,QAAQ,CAAC,CAAC,EAAEE,EAAE+jB,EAAE,cAAc,EAAE,KAAK,EAAE9jB,EAAE6+G,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,MAAM,CAAC,MAAM,CAACh/G,EAAE,UAAUE,EAAEF,EAAE,SAAS,CAAC,EAAE,QAAQ,CAAC,CAAC,EAAEqB,EAAE,CAACrB,EAAE,UAAUA,EAAE,UAAUE,EAAEF,EAAE,UAAUA,EAAE,SAAS,EAAE,EAAE,EAAE,WAAWqB,EAAE3D,EAAE,KAAK,EAAE,GAAGumB,EAAE,cAAcvmB,EAAE,KAAK,IAAI,EAAE,OAAO,EAAE,IAAIgH,EAAEzE,EAAE,MAAM,OAAO,EAAE6B,EAAE,EAAE,UAAU,IAAI7B,EAAE,MAAM,EAAE,GAAG+B,EAAE,EAAE,UAAU,IAAI7B,EAAE,MAAM,EAAE,GAAG8B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG0C,EAAE,IAAI,WAAW,IAAI,WAAWsf,EAAE,eAAehkB,EAAE,KAAK,CAAC,EAAE,MAAM,EAAEkC,EAAE,IAAI,WAAW,IAAI,WAAW8hB,EAAE,eAAe5iB,CAAC,CAAC,EAAE,MAAM,EAAE,OAAO+iH,GAAGtiH,EAAE06G,GAAG9+G,EAAE,OAAOiH,EAAED,EAAE1C,EAAEhC,EAAE,UAAUmC,EAAEF,CAAC,EAAE,EAAE,YAAYhC,EAAE,MAAM,EAAE,EAAE,YAAYE,EAAE,MAAM,EAAE,EAAE,MAAMH,EAAE,YAAY,CAAC,CAAC,IAAIukH,IAAG,CAAC,WAAWloG,GAAG,YAAY,OAAO,UAAUgoG,IAAI,WAAWC,GAAG,EAAME,IAAI,GAAGC,IAAGzH,GAAGzgG,GAAGioG,IAAI,MAAM,EAAME,IAAI,GAAGC,IAAG3H,GAAGxgG,GAAGkoG,IAAI,MAAM,EAAME,GAAG,SAASC,IAAI1nH,EAAE,CAACynH,GAAGznH,EAAE,KAAK,MAAM4f,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS+nG,IAAI3nH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,MAAM,CAAC,MAAME,CAAC,EAAE,QAAQI,CAAC,EAAEN,EAAEO,EAAED,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,EAAEA,EAAE,WAAW,EAAE,MAAM,SAAS,EAAE,GAAGwmB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,CAAC,IAAI,EAAExmB,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGmnH,GAAGlnH,EAAE8+G,GAAG,EAAE,OAAOn/G,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI0nH,IAAG,CAAC,WAAWhoG,GAAG,YAAY,OAAO,UAAU8nG,IAAI,WAAWC,GAAG,EAAME,IAAI,GAAGC,IAAGjI,GAAGhgG,GAAGgoG,IAAI,MAAM,EAAME,IAAI,GAAGC,IAAGnI,GAAG//F,GAAGioG,IAAI,MAAM,EAAME,IAAGtI,GAAG3/F,EAAE,EAAMkoG,IAAI,GAAGC,IAAGtI,GAAG3/F,GAAGgoG,IAAI,MAAM,EAAME,IAAGzI,GAAGx/F,EAAE,EAAMkoG,IAAI,GAAGC,IAAGzI,GAAGz/F,GAAGioG,IAAI,MAAM,EAAME,IAAI,GAAGC,IAAG3I,GAAGx/F,GAAGkoG,IAAI,MAAM,EAAME,GAAG,SAASC,IAAI1oH,EAAE,CAACyoH,GAAGzoH,EAAE,KAAK,MAAM0gB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASioG,IAAI3oH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,iBAAiBO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAEgC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,CAAC,WAAWU,EAAE,KAAK,EAAE,aAAaE,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAE,GAAGwC,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG,EAAEA,EAAEV,EAAEyC,CAAC,CAAC,IAAI3B,EAAE,EAAE,MAAM,OAAOw3C,EAAE,2BAA2B,MAAM,EAAEx3C,CAAC,EAAE,GAAG,CAACkB,EAAEC,CAAC,EAAEq2C,EAAE,0BAA0B,EAAE,MAAM,CAAC,EAAEjzC,EAAEuf,EAAE,cAAc3iB,CAAC,EAAEY,EAAE,EAAE,WAAWb,EAAE,EAAE,KAAK,EAAE,GAAG4iB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIniB,EAAE,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG0jH,GAAGvmH,EAAEm9G,GAAG,EAAE,OAAO93G,EAAE5C,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqBz1C,EAAE,MAAMjC,CAAC,EAAEiC,EAAE,MAAMJ,CAAC,CAAC,OAAOI,CAAC,CAAC,IAAI6jH,IAAG,CAAC,WAAWloG,GAAG,YAAY,OAAO,UAAUgoG,IAAI,WAAWC,GAAG,EAAME,IAAI,GAAGC,IAAGjJ,GAAGl/F,GAAGkoG,GAAG,EAAME,GAAG,SAASC,IAAIhpH,EAAE,CAAC+oH,GAAG/oH,EAAE,KAAK,MAAM4gB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASqoG,IAAIjpH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAEO,EAAE,EAAE,EAAE,EAAED,EAAE,UAAU,IAAIC,EAAE,MAAM,EAAE,GAAGumB,EAAE,OAAOvmB,EAAE,QAAQ,UAAU,IAAI,0DAA0DA,EAAE,QAAQ,EAAE,GAAG,CAAC,WAAW,EAAE,QAAQ0B,EAAE,IAAIC,EAAE,gBAAgBC,CAAC,EAAEjC,EAAE0C,EAAE43C,EAAE,kBAAkBj6C,EAAE,MAAM,EAAE0B,EAAE,EAAEC,EAAEC,CAAC,EAAEU,EAAED,EAAE,aAAaE,EAAEF,EAAE,YAAYG,EAAEH,EAAE,QAAQ,IAAII,EAAEJ,EAAE,QAAQ,MAAMsB,EAAEtB,EAAE,QAAQ,OAAO,EAAEA,EAAE,QAAQ,KAAK2E,EAAE3E,EAAE,eAAe,EAAEA,EAAE,cAAc+B,EAAE/B,EAAE,aAAagC,EAAEhC,EAAE,YAAYiC,EAAEjC,EAAE,WAAWkC,EAAElC,EAAE,YAAY,GAAGA,EAAE,aAAa,eAAe,MAAM,IAAI,MAAM,6CAA6CA,EAAE,yCAAyC,EAAE,IAAI4E,EAAElH,EAAE,WAAWsC,EAAE,SAAS,SAAS,EAAEoC,EAAE1E,EAAE,UAAU,IAAIkH,EAAE,MAAM,EAAE,GAAG,OAAOuhH,GAAG,EAAExoH,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGA,EAAE,MAAM,GAAGsC,EAAEC,EAAEC,EAAEC,EAAEkB,EAAE,EAAEqD,EAAE,EAAE5C,EAAEC,EAAEC,EAAEC,EAAEE,CAAC,EAAEwC,CAAC,CAAC,IAAI0hH,IAAG,CAAC,WAAWtoG,GAAG,YAAY,OAAO,UAAUooG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIppH,EAAE,CAACmpH,GAAGnpH,EAAE,KAAK,MAAMihB,GAAG,KAAK,CAAC,wBAAwB,CAAC,CAAC,CAAC,SAASooG,IAAIrpH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE+B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAED,EAAEE,EAAE,EAAE,CAAC,WAAWS,EAAE,KAAKC,EAAE,aAAaC,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAEyC,EAAEH,EAAE,GAAGE,EAAE,CAAC,IAAI6B,EAAE,EAAE,UAAU,IAAIhC,EAAE,MAAM,EAAE,GAAGgC,IAAI3C,IAAIE,EAAES,EAAEV,EAAE0C,EAAE5B,EAAEw3C,EAAE,iBAAiBx3C,EAAE,OAAOb,EAAE,MAAM,MAAM,EAAE,CAACq4C,EAAE,2BAA2B,OAAOx3C,EAAEb,EAAE,MAAM,MAAM,EAAE,GAAG,CAAC+B,EAAE,CAAC,EAAEs2C,EAAE,0BAA0Br4C,EAAE,MAAMa,CAAC,EAAEuE,EAAEuf,EAAE,cAAc,CAAC,EAAE,EAAE3kB,EAAEA,EAAE,QAAQ,YAAY,EAAEugH,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAEvgH,CAAC,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAED,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,IAAI,IAAIyC,EAAE,EAAE,WAAWT,EAAE,SAAS,EAAE,GAAG4iB,EAAE,cAAc3kB,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIyC,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGwkH,GAAGjnH,EAAEqF,EAAE3C,CAAC,CAAC,CAAC,GAAG7B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAIgC,EAAE41C,EAAE,qBAAqB71C,EAAE,MAAM7B,CAAC,EAAE6B,EAAE,MAAMC,CAAC,CAAC,OAAOzC,EAAE,QAAQ,WAAW,EAAE,YAAY,EAAE,MAAM,EAAEwC,CAAC,CAAC,IAAI2kH,IAAG,CAAC,WAAWroG,GAAG,YAAY,OAAO,UAAUmoG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIxpH,EAAE,CAACupH,GAAGvpH,EAAE,KAAK,MAAMkhB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASuoG,IAAIzpH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE+B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAED,EAAEE,EAAE,EAAE,CAAC,WAAWS,EAAE,KAAKC,EAAE,aAAaC,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAE,GAAGwC,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG+B,IAAI1C,IAAIE,EAAES,EAAEV,EAAEyC,EAAE,CAAC,IAAI3B,EAAEb,EAAE,MAAM,OAAOq4C,EAAE,2BAA2B,MAAM33C,EAAEG,CAAC,EAAE,GAAG,CAACkB,EAAE,CAAC,EAAEs2C,EAAE,0BAA0Br4C,EAAE,MAAMU,CAAC,EAAE0E,EAAEuf,EAAE,cAAc,CAAC,EAAE,EAAE,EAAE,WAAW5iB,EAAE/B,EAAE,KAAK,EAAE,GAAG2kB,EAAE,cAAc3kB,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIwC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG4kH,GAAGrnH,EAAEm9G,GAAG,EAAE,OAAO93G,EAAE5C,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqB,EAAE,MAAM13C,CAAC,EAAE,EAAE,MAAM6B,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI+kH,IAAG,CAAC,WAAWxoG,GAAG,YAAY,OAAO,UAAUsoG,IAAI,WAAWC,GAAG,EAAME,IAAI,GAAGC,IAAG/J,GAAG1+F,GAAGwoG,GAAG,EAAME,IAAI,SAAS7pH,EAAE,CAACA,EAAEA,EAAE,QAAQ,GAAG,UAAUA,EAAEA,EAAE,UAAU,GAAG,WAAW,GAAG6pH,KAAKA,GAAG,CAAC,EAAE,EAAE,IAAIC,GAAG,SAASC,IAAI/pH,EAAE,CAAC8pH,GAAG9pH,EAAE,KAAK,MAAMohB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,QAAQ,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS4oG,IAAIhqH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQE,EAAE,MAAM,CAAC,SAASI,EAAE,KAAKC,CAAC,CAAC,EAAEP,EAAE,EAAEM,EAAE,IAAI,CAAC0C,EAAEkB,IAAIlB,EAAE,GAAG,EAAE,MAAMkB,GAAGlB,EAAE,EAAE,EAAE,EAAE9C,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+B,EAAE/B,EAAE,WAAW,EAAE,EAAE,KAAK,EAAEgC,EAAEhC,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,GAAGE,EAAE,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM,EAAES,EAAEtC,EAAE,IAAI0C,GAAGA,EAAE,EAAE,EAAEH,EAAEvC,EAAE,IAAI0C,GAAGA,EAAE,EAAE,EAAEF,EAAE,IAAI,WAAW,IAAI,WAAWF,CAAC,EAAE,MAAM,EAAEG,EAAE,IAAI,WAAW,IAAI,WAAWF,CAAC,EAAE,MAAM,EAAE,OAAOinH,GAAG,EAAE3nH,EAAE,EAAE,MAAM,OAAOk9G,GAAG,EAAE,OAAOv8G,EAAEC,EAAE8mH,GAAGtpH,GAAG2B,CAAC,EAAED,CAAC,CAAC,IAAIgoH,IAAG,CAAC,WAAW7oG,GAAG,YAAY,OAAO,WAAW4oG,IAAI,UAAUD,GAAG,EAAMG,IAAI,GAAGC,IAAGtK,GAAGt+F,GAAG2oG,GAAG,EAAME,IAAGzK,GAAGn+F,EAAE,EAAE,SAAS6oG,GAAGrqH,EAAE,EAAE,CAAC,IAAIE,EAAE,IAAI,WAAWF,EAAE,KAAK,OAAO,OAAO,EAAE,CAAC,EAAEM,EAAEJ,EAAE,GAAGK,EAAEL,EAAE,GAAG,EAAEA,EAAE,GAAG,EAAEA,EAAE,GAAG,OAAOF,EAAE,KAAK,MAAM,CAAC,EAAE,CAAC,iBAAiBM,EAAE,aAAaC,EAAE,gBAAgB,EAAE,cAAc,CAAC,CAAC,CAAC,IAAI+pH,GAAG,SAASC,IAAIvqH,EAAE,CAACsqH,GAAGtqH,EAAE,KAAK,MAAM0hB,GAAG,SAAS,CAAC,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS8oG,IAAIxqH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,aAAaO,EAAE,cAAc,EAAE,eAAe,CAAC,EAAED,EAAE,CAAC,MAAM2B,EAAE,OAAOC,CAAC,EAAEhC,EAAEiC,EAAE,EAAE,UAAU,IAAIF,EAAE,MAAM,EAAE,GAAGW,EAAE,EAAE,UAAU,IAAIV,EAAE,MAAM,EAAE,GAAGW,EAAEynH,GAAGnoH,EAAES,EAAE,EAAErC,EAAE,CAAC,EAAE,CAAC,iBAAiBuC,EAAE,aAAaC,EAAE,gBAAgBC,EAAE,cAAckB,CAAC,EAAEmmH,GAAG,EAAExnH,CAAC,EAAE,OAAO,EAAE,KAAK,MAAMG,CAAC,EAAE,EAAE,KAAK,MAAMkB,CAAC,EAAE,EAAE,WAAW,CAACnB,CAAC,EAAE,QAAQD,CAAC,CAAC,CAAC,IAAI2nH,IAAG,CAAC,WAAW/oG,GAAG,YAAY,OAAO,UAAU6oG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI3qH,EAAE,CAAC0qH,GAAG1qH,EAAE,KAAK,MAAM2hB,GAAG,SAAS,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,MAAM,CAAC,CAAC,CAAC,SAASipG,IAAI5qH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,aAAaO,EAAE,cAAc,EAAE,eAAe,EAAE,mBAAmB0B,CAAC,EAAE3B,EAAE,CAAC,MAAM4B,EAAE,OAAOC,CAAC,EAAEjC,EAAE0C,EAAE,EAAE,UAAU,IAAIV,EAAE,MAAM,EAAE,GAAGW,EAAE,EAAE,UAAU,IAAIV,EAAE,MAAM,EAAE,GAAGW,EAAE4nH,GAAG9nH,EAAEC,EAAE,EAAEtC,EAAE,EAAE0B,CAAC,EAAE,CAAC,iBAAiBc,EAAE,aAAaC,EAAE,gBAAgBkB,EAAE,cAAc,CAAC,EAAEmmH,GAAG,EAAEvnH,CAAC,EAAE,EAAE,KAAK,MAAMoB,CAAC,EAAE,IAAIqD,EAAE,EAAE,WAAW,CAACvE,CAAC,EAAE,QAAQD,CAAC,EAAE,EAAE,EAAE,WAAW,CAAC,EAAE,QAAQ,CAAC,EAAE,MAAM,CAACwE,EAAE,CAAC,CAAC,CAAC,IAAIsjH,IAAG,CAAC,WAAWlpG,GAAG,YAAY,OAAO,UAAUgpG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI/qH,EAAE,CAAC8qH,GAAG9qH,EAAE,KAAK,MAAM4hB,GAAG,SAAS,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASopG,IAAIhrH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,aAAaO,EAAE,cAAc,EAAE,eAAe,EAAE,aAAa0B,CAAC,EAAE3B,EAAE,CAAC,MAAM4B,EAAE,OAAOC,CAAC,EAAEjC,EAAE0C,EAAE,EAAE,UAAU,IAAIV,EAAE,MAAM,EAAE,GAAGW,EAAE,EAAE,UAAU,IAAIV,EAAE,MAAM,EAAE,GAAGW,EAAEgoH,GAAGloH,EAAEC,EAAE,EAAEtC,EAAE,EAAE0B,CAAC,EAAE,CAAC,iBAAiBc,EAAE,aAAaC,EAAE,gBAAgBkB,EAAE,cAAc,CAAC,EAAEmmH,GAAG,EAAEvnH,CAAC,EAAE,EAAE,KAAK,MAAM,CAAC,EAAE,IAAIyE,EAAE,EAAE,WAAW,CAACvE,CAAC,EAAE,QAAQD,CAAC,EAAE,EAAE,EAAE,WAAW,CAACC,CAAC,EAAE,UAAUkB,CAAC,EAAE,MAAM,CAACqD,EAAE,CAAC,CAAC,CAAC,IAAI0jH,IAAG,CAAC,WAAWrpG,GAAG,YAAY,OAAO,UAAUmpG,IAAI,WAAWC,GAAG,EAAME,IAAI,GAAGC,IAAGtL,GAAGp+F,GAAGypG,IAAI,MAAM,EAAME,GAAG,SAASC,IAAIrrH,EAAE,CAACorH,GAAGprH,EAAE,KAAK,MAAM8hB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASwpG,IAAItrH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ0B,EAAE,SAASC,CAAC,EAAE5B,EAAE6B,EAAEjC,EAAE,WAAW,CAAC,GAAGK,EAAE,MAAM,CAAC,EAAE,CAAC,EAAEqC,EAAE1C,EAAE,UAAU,IAAIiC,EAAE,MAAM,EAAE,GAAGW,EAAE5C,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG,OAAO6qH,GAAGtoH,EAAE,EAAEb,EAAEC,EAAEU,CAAC,EAAET,CAAC,CAAC,IAAIopH,IAAG,CAAC,WAAWzpG,GAAG,YAAY,OAAO,UAAUupG,IAAI,WAAWC,GAAG,EAAE,SAASE,IAAIxrH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAEJ,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAE,OAAOA,EAAE,mBAAmBI,CAAC,EAAE,KAAK,CAAC,EAAEA,CAAC,CAAC,IAAImrH,IAAG,CAAC,WAAW5pG,GAAG,YAAY,OAAO,WAAW2pG,GAAG,EAAE,SAASE,IAAI1rH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,CAAC,EAAED,EAAE,GAAG,EAAE,SAAS,EAAE,OAAOglH,GAAG,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,EAAE,QAAQplH,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,GAAG,MAAM,EAAE,EAAE,GAAG,MAAM,EAAE,QAAQqC,GAAG,CAACkkB,EAAE,kBAAkB,EAAElkB,EAAE,MAAM,uDAAuD,EAAEkkB,EAAE,OAAO,IAAIlkB,EAAE,MAAM,IAAI,uDAAuD,CAAC,CAAC,EAAE,IAAIX,EAAE,CAAC,EAAEC,EAAE,EAAE,IAAIU,GAAG,CAAC,IAAIC,EAAEyiH,GAAG,CAAC,OAAO,CAAC,MAAM1iH,CAAC,EAAE,QAAQ1C,EAAE,MAAM,CAAC,IAAIK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,KAAKY,CAAC,EAAEA,CAAC,CAAC,EAAEV,EAAE8gH,GAAG,CAAC,OAAO/gH,EAAE,QAAQhC,EAAE,MAAM,CAAC,KAAKK,CAAC,CAAC,CAAC,EAAE,OAAO0B,EAAE,QAAQW,GAAG1C,EAAE,YAAY0C,EAAE,MAAM,CAAC,EAAET,CAAC,CAAC,IAAIwpH,IAAG,CAAC,WAAW5pG,GAAG,YAAY,OAAO,WAAW2pG,GAAG,EAAME,GAAG,SAASC,IAAI7rH,EAAE,CAAC4rH,GAAG5rH,EAAE,KAAK,MAAMgiB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,QAAQ,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS8pG,IAAI9rH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQE,EAAE,MAAM,CAAC,SAASI,EAAE,cAAcC,CAAC,CAAC,EAAEP,EAAE,EAAEM,EAAE,IAAI,CAAC4D,EAAEC,IAAID,EAAE,GAAG,EAAE,MAAMC,GAAGD,EAAE,EAAE,EAAE,GAAG4iB,EAAE,cAAc,EAAE,KAAK,IAAI,EAAE,OAAO0+F,GAAG,CAAC,QAAQtlH,EAAE,MAAM,CAAC,MAAM,EAAE,MAAMK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,EAAE,IAAI,EAAEL,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+B,EAAE/B,EAAE,WAAW,EAAE,EAAE,KAAK,EAAE,EAAEA,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,GAAGW,EAAE,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM,EAAE,EAAEtC,EAAE,IAAI4D,GAAGA,EAAE,EAAE,EAAEpB,EAAExC,EAAE,IAAI4D,GAAGA,EAAE,EAAE,EAAEnB,EAAE,IAAI,WAAW,IAAI,WAAW,CAAC,EAAE,MAAM,EAAEC,EAAE,IAAI,WAAW,IAAI,WAAWF,CAAC,EAAE,MAAM,EAAE,OAAO8oH,GAAG,EAAEhpH,EAAE,EAAE,MAAM,OAAOy8G,GAAG,EAAE,OAAOt8G,EAAEC,EAAEzC,EAAE,CAAC,EAAE0B,CAAC,CAAC,IAAI8pH,GAAG,CAAC,WAAW/pG,GAAG,YAAY,OAAO,WAAW8pG,IAAI,UAAUD,GAAG,EAAMG,IAAI,GAAGC,IAAGpM,GAAG39F,GAAG8pG,GAAG,EAAME,GAAG,SAASC,IAAInsH,EAAE,CAACksH,GAAGlsH,EAAE,KAAK,MAAMmiB,GAAG,KAAK,CAAC,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASiqG,IAAIpsH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,EAAEM,EAAE,MAAMC,CAAC,EAAE,EAAE,EAAEL,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG,EAAEJ,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG0B,EAAE,EAAEC,EAAE5B,EAAE6B,EAAED,EAAEA,EAAE,QAAQ,YAAYC,EAAEugH,GAAG,CAAC,QAAQxiH,EAAE,OAAO,CAAC,EAAEI,CAAC,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAE2B,EAAE/B,EAAE,UAAU,IAAIiC,EAAE,MAAM,EAAE,IAAI,IAAIS,EAAE1C,EAAE,WAAWI,EAAE,MAAM,SAAS,EAAEuC,EAAE3C,EAAE,UAAU,IAAI0C,EAAE,MAAM,EAAE,GAAG,OAAOspH,GAAGjqH,EAAE,EAAEY,CAAC,EAAEX,EAAE,QAAQ,WAAWhC,EAAE,YAAYiC,EAAE,MAAM,EAAES,CAAC,CAAC,IAAIypH,IAAG,CAAC,WAAWlqG,GAAG,YAAY,OAAO,UAAUgqG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIvsH,EAAE,CAACssH,GAAGtsH,EAAE,KAAK,MAAMoiB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASoqG,IAAIxsH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE+B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAED,EAAEE,EAAE,EAAE,CAAC,WAAWS,EAAE,KAAKC,EAAE,aAAaC,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAEyC,EAAEH,EAAE,GAAGE,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG+B,IAAI1C,IAAIE,EAAES,EAAEV,EAAEyC,EAAE3B,EAAEw3C,EAAE,iBAAiBx3C,EAAE,OAAOb,EAAE,MAAM,MAAM,EAAE,CAACq4C,EAAE,2BAA2B,OAAOx3C,EAAEb,EAAE,MAAM,MAAM,EAAE,GAAG,CAAC+B,EAAE,CAAC,EAAEs2C,EAAE,0BAA0Br4C,EAAE,MAAMa,CAAC,EAAEuE,EAAEuf,EAAE,cAAc,CAAC,EAAE,EAAE,EAAE,WAAW5iB,EAAE/B,EAAE,KAAK,EAAE,GAAG2kB,EAAE,cAAc3kB,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIwC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG2nH,GAAGpqH,EAAEqF,EAAE83G,GAAG,EAAE,OAAO16G,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqB,EAAE,MAAM13C,CAAC,EAAE,EAAE,MAAM6B,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI8nH,IAAG,CAAC,WAAWrqG,GAAG,YAAY,OAAO,UAAUmqG,IAAI,WAAWC,GAAG,EAAME,IAAI1sH,GAAG,CAAC,GAAG,CAAC,QAAQ,EAAE,MAAME,CAAC,EAAEF,EAAE,CAAC,MAAMM,EAAE,KAAKC,EAAE,KAAK,EAAE,MAAM,CAAC,EAAEL,EAAE+B,EAAEwxE,GAAGnzE,EAAEC,EAAE,EAAE,CAAC,EAAE2B,EAAE,EAAE,WAAW,CAACD,EAAE,MAAM,EAAE,CAAC,EAAE,OAAO,EAAE,mBAAmBC,CAAC,EAAE,IAAID,CAAC,EAAEC,CAAC,EAAEyqH,IAAG,CAAC,WAAWnqG,GAAG,YAAY,OAAO,WAAWkqG,GAAG,EAAME,IAAI,GAAGC,IAAGhN,GAAG1hG,GAAGyuG,GAAG,EAAME,IAAGnN,GAAGh9F,EAAE,EAAMoqG,IAAGpN,GAAG18F,EAAE,EAAM+pG,GAAG,SAASC,IAAIjtH,EAAE,CAACgtH,GAAGhtH,EAAE,KAAK,MAAM+iB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASmqG,IAAIltH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAEL,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK+B,CAAC,EAAE3B,EAAE,CAAC4B,EAAEC,CAAC,EAAEF,EAAE,CAACW,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAMyC,EAAE,CAACJ,EAAEV,EAAEC,EAAEY,CAAC,EAAEmB,EAAE,EAAE,UAAU,IAAI3D,EAAE,MAAM,EAAE,EAAE2D,EAAE,QAAQ,YAAY,EAAEw+G,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAEniH,CAAC,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAE2D,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,GAAG,IAAIqD,EAAErD,EAAE,GAAG,EAAE,EAAE,WAAWlB,EAAE,SAAS,EAAE,GAAG8jB,EAAE,cAAcvmB,EAAE,KAAK,IAAI,EAAE,OAAO,EAAE,IAAIoE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,OAAOqoH,GAAGzlH,EAAE3E,EAAEC,EAAEC,EAAEC,EAAEb,EAAEC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAEwC,CAAC,EAAE,GAAG,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,CAAC,CAAC,IAAIwoH,IAAG,CAAC,WAAWpqG,GAAG,YAAY,OAAO,UAAUkqG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIrtH,EAAE,CAACotH,GAAGptH,EAAE,KAAK,MAAM6iB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASyqG,IAAIttH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,OAAOO,CAAC,EAAEL,EAAE,CAAC,aAAa,EAAE,iBAAiB,EAAE,KAAK+B,CAAC,EAAE3B,EAAE,CAAC4B,EAAEC,CAAC,EAAEF,EAAE,CAACW,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAMyC,EAAE,CAACJ,EAAEV,EAAEC,EAAEY,CAAC,EAAEmB,EAAE,EAAE,WAAWlB,EAAE,SAAS,EAAE,GAAG8jB,EAAE,cAAcvmB,EAAE,KAAK,IAAI,EAAE,OAAO2D,EAAE,IAAI,EAAE,EAAE,UAAU,IAAI3D,EAAE,MAAM,EAAEgH,EAAE,EAAE,QAAQ,YAAYA,EAAEm7G,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAEniH,CAAC,EAAE,MAAM,CAAC,MAAM,SAAS,CAAC,CAAC,EAAE,EAAE,EAAE,UAAU,IAAIgH,EAAE,MAAM,GAAG,IAAI,EAAE,EAAE,GAAG5C,EAAE,EAAE,UAAU,IAAIT,EAAE,MAAM,EAAE,GAAG,OAAOkpH,GAAG,EAAExqH,EAAEC,EAAEC,EAAEC,EAAEb,EAAEC,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAEwC,CAAC,EAAE4C,GAAG,MAAM,EAAE,YAAYA,EAAE,MAAM,EAAErD,CAAC,CAAC,IAAIqpH,IAAG,CAAC,WAAW1qG,GAAG,YAAY,OAAO,UAAUwqG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIztH,EAAE,CAACwtH,GAAGxtH,EAAE,KAAK,MAAMkjB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASwqG,IAAI1tH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAEwmB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAGA,EAAE,MAAM,SAAS,EAAE,OAAO6/G,GAAG,CAAC,OAAO,CAAC,EAAE7/G,CAAC,EAAE,QAAQL,CAAC,CAAC,EAAE,IAAI+B,EAAE/B,EAAE,WAAWK,EAAE,MAAMA,EAAE,KAAK,EAAE2B,EAAEhC,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG4B,EAAEjC,EAAE,UAAU,IAAI+B,EAAE,MAAM,EAAE,GAAGW,EAAE,IAAI,WAAW,IAAI,WAAW,CAAC,EAAE,MAAM,EAAEC,EAAE,IAAI,WAAW,IAAI,WAAWtC,EAAE,KAAK,EAAE,MAAM,EAAEitH,GAAGtrH,EAAEU,EAAE,EAAE,OAAOC,EAAEtC,EAAE,MAAM,OAAO4B,CAAC,EAAE,IAAIW,EAAE++G,GAAG,CAAC,OAAO,CAAC,EAAE5/G,CAAC,EAAE,MAAM,CAAC,MAAM1B,EAAE,KAAK,EAAE,QAAQL,CAAC,CAAC,EAAE,OAAOA,EAAE,YAAY+B,EAAE,MAAM,EAAEa,CAAC,CAAC,IAAI6qH,IAAG,CAAC,WAAWzqG,GAAG,YAAY,OAAO,WAAWwqG,IAAI,UAAUD,GAAG,EAAMG,GAAG,SAASC,IAAI7tH,EAAE,CAAC4tH,GAAG5tH,EAAE,KAAK,MAAM6lB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASioG,IAAI9tH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,QAAQ,EAAE,UAAU,EAAE,OAAO0B,CAAC,EAAE3B,EAAE4B,EAAEhC,EAAE,WAAWK,EAAE,MAAMA,EAAE,KAAK,EAAE4B,EAAEjC,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAGqC,EAAE1C,EAAE,UAAU,IAAIgC,EAAE,MAAM,EAAE,GAAG,CAACW,EAAEC,EAAEC,EAAEC,CAAC,EAAEzC,EAAE,MAAM,CAAC2D,EAAE,CAAC,EAAEs2C,EAAE,eAAev4C,EAAEa,EAAEC,CAAC,EAAEwE,EAAE,IAAI,EAAE,EAAE,IAAI5C,EAAE,OAAO,GAAG,SAAS,CAAC,EAAE,EAAE,EAAE4C,EAAE,EAAE,CAAC,EAAE,CAAC,GAAG,EAAE,CAAC,EAAE3C,EAAE,IAAI,WAAW,IAAI,WAAWD,CAAC,EAAE,MAAM,EAAE,OAAOipH,GAAGzrH,EAAEU,EAAEC,EAAEC,EAAEC,EAAE,EAAEkB,EAAE,EAAEU,EAAED,EAAE,OAAO/B,CAAC,EAAEV,CAAC,CAAC,IAAI6rH,IAAG,CAAC,WAAWloG,GAAG,YAAY,OAAO,WAAWioG,IAAI,UAAUD,GAAG,EAAMG,IAAGrO,GAAGx8F,EAAE,EAAM8qG,IAAGtO,GAAGv8F,EAAE,EAAM8qG,GAAG,SAASC,IAAInuH,EAAE,CAACkuH,GAAGluH,EAAE,KAAK,MAAMqjB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS+qG,IAAIpuH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,QAAQO,EAAE,QAAQ,CAAC,EAAEL,EAAE,CAAC,MAAM,CAAC,EAAEI,EAAE2B,EAAE,EAAE,WAAW,EAAE,EAAE,KAAK,EAAE,GAAG6kB,EAAE,cAAc,CAAC,IAAI,EAAE,OAAO7kB,EAAE,GAAG,CAAC,UAAUC,EAAE,WAAWC,EAAE,UAAUS,EAAE,QAAQC,EAAE,WAAWC,CAAC,EAAEkzB,GAAG,gBAAgB,EAAEz1B,EAAE,CAAC,EAAEyC,EAAE,EAAE,UAAU,IAAIzC,EAAE,MAAM,EAAE,GAAG4D,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGoD,EAAE,IAAI,WAAW,IAAI,WAAW1E,CAAC,EAAE,MAAM,EAAEkC,EAAE,EAAE,UAAU,IAAI9C,EAAE,MAAM,EAAE,GAAG,OAAOisH,GAAGlrH,EAAEmB,EAAEk7G,GAAG,EAAE,OAAOn9G,EAAEC,EAAES,EAAE2E,EAAEzE,EAAEiC,CAAC,EAAE9C,CAAC,CAAC,IAAIosH,IAAG,CAAC,WAAWhrG,GAAG,YAAY,OAAO,UAAU8qG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIvuH,EAAE,CAACsuH,GAAGtuH,EAAE,KAAK,MAAM,WAAW,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASwuH,IAAIxuH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,CAAC,EAAEF,EAAE,CAAC,UAAUM,EAAE,EAAEC,EAAE,EAAE,CAAC,EAAE,EAAE,EAAEL,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG2B,EAAE/B,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG2B,EAAEhC,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGiC,EAAEjC,EAAE,WAAWK,EAAE,MAAMA,EAAE,KAAK,EAAEqC,EAAE1C,EAAE,UAAU,IAAIiC,EAAE,MAAM,EAAE,GAAGU,EAAEvC,EAAE,MAAM,OAAOwC,EAAEvC,EAAE,MAAM,OAAOwC,EAAEF,IAAI,GAAGA,EAAE,GAAGC,IAAI,EAAE,EAAEgkB,EAAE,cAAcvmB,EAAE,MAAM,MAAM,CAAC,CAAC,EAAE,OAAO+tH,GAAG,EAAErsH,EAAEC,EAAEa,EAAEH,CAAC,EAAET,CAAC,CAAC,IAAIssH,IAAG,CAAC,WAAWlrG,GAAG,YAAY,OAAO,WAAWirG,IAAI,UAAUD,GAAG,EAAMG,GAAG,SAASC,IAAI3uH,EAAE,CAAC0uH,GAAG1uH,EAAE,KAAK,MAAM6jB,GAAG,KAAK,CAAC,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS+qG,IAAI5uH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,EAAEE,CAAC,CAAC,EAAEF,EAAEM,EAAE,EAAE,UAAU,IAAIJ,EAAE,MAAM,EAAE,GAAGK,EAAE,EAAE,WAAWL,EAAE,MAAMA,EAAE,KAAK,EAAE,EAAE,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG,OAAOumB,EAAE,cAAcvmB,EAAE,KAAK,IAAI,GAAGmuH,GAAGpuH,EAAE,CAAC,EAAEC,CAAC,CAAC,IAAIsuH,IAAG,CAAC,WAAW,UAAU,YAAY,OAAO,UAAUF,IAAI,WAAWC,GAAG,EAAME,IAAGnP,GAAGj8F,EAAE,EAAMqrG,GAAG,SAASC,IAAIhvH,EAAE,CAAC+uH,GAAG/uH,EAAE,KAAK,MAAMmkB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS8qG,IAAIjvH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAO,CAAC,OAAOE,CAAC,EAAE,MAAM,CAAC,IAAII,CAAC,CAAC,EAAEN,EAAEO,EAAE,EAAE,UAAU,IAAIL,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,WAAWA,EAAE,MAAMA,EAAE,KAAK,EAAE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+B,EAAE/B,EAAE,MAAMI,GAAG4B,EAAE4kB,EAAE,cAAc5mB,EAAE,KAAK,EAAE+B,EAAE,OAAO6kB,EAAE,cAAc,EAAE,KAAK,IAAI,GAAGioG,GAAGxuH,EAAE,EAAE0B,EAAEC,CAAC,EAAE,CAAC,CAAC,IAAIgtH,IAAG,CAAC,WAAW/qG,GAAG,YAAY,OAAO,UAAU6qG,IAAI,WAAWC,GAAG,EAAE,SAASE,IAAInvH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,WAAW,EAAE,SAAS,CAAC,EAAED,EAAE2B,EAAE6kB,EAAE,cAAc,CAAC,EAAE5kB,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,EAAEA,EAAE,KAAK,GAAG,CAAC,EAAE,QAAQ4C,EAAE,EAAE,EAAE,OAAOA,EAAEvE,EAAE,MAAM,OAAO,EAAEuE,EAAE5C,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC,EAAE,IAAIC,EAAE4pH,GAAG,WAAW,CAAC,OAAO,CAAC,EAAExrH,CAAC,EAAE,QAAQL,EAAE,MAAM,CAAC,SAASgC,EAAE,cAAc,CAAC,CAAC,CAAC,EAAEU,EAAE43C,EAAE,YAAYr4C,EAAE,MAAM,EAAEF,EAAE,EAAE,EAAEY,EAAE23C,EAAE,YAAY53C,EAAE,OAAO,EAAE,OAAO,EAAE,EAAEE,EAAE03C,EAAE,oBAAoBr4C,EAAE,MAAM,EAAEF,EAAE,EAAE,EAAE,EAAE4/G,GAAG,CAAC,OAAO,CAAC,EAAE1/G,CAAC,EAAE,QAAQjC,EAAE,MAAM,CAAC,MAAM0C,CAAC,CAAC,CAAC,EAAEmC,EAAEy7G,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQtgH,EAAE,MAAM,CAAC,KAAK2C,CAAC,CAAC,CAAC,EAAEgC,EAAEg9G,GAAG,CAAC,OAAO,CAAC,EAAE98G,CAAC,EAAE,QAAQ7E,EAAE,MAAM,CAAC,MAAM4C,CAAC,CAAC,CAAC,EAAE,OAAO5C,EAAE,YAAYiC,EAAE,MAAM,EAAEjC,EAAE,YAAY,EAAE,MAAM,EAAEA,EAAE,YAAY6E,EAAE,MAAM,EAAEF,CAAC,CAAC,IAAIuqH,IAAG,CAAC,WAAWnrG,GAAG,YAAY,OAAO,WAAWkrG,GAAG,EAAME,GAAG,SAASC,IAAItvH,EAAE,CAACqvH,GAAGrvH,EAAE,KAAK,MAAM,sBAAsB,SAAS,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASuvH,IAAIvvH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,CAAC,EAAEF,EAAE,CAAC,QAAQM,EAAE,OAAOC,EAAE,WAAW,EAAE,aAAa,CAAC,EAAEL,EAAE+B,EAAE3B,EAAE,MAAM,GAAG4B,EAAE5B,EAAE,MAAM,GAAG6B,EAAE,EAAE,SAAS,EAAE,MAAM,EAAE,GAAGS,EAAE,CAACX,EAAEE,EAAED,CAAC,EAAEW,EAAE,EAAE,UAAU,IAAIvC,EAAE,MAAM,EAAE,GAAGwC,EAAE,EAAE,UAAU,IAAIvC,EAAE,MAAM,EAAE,GAAGwC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAE,EAAE,WAAWJ,EAAEtC,EAAE,KAAK,EAAE4D,EAAE,EAAE,UAAU,IAAIlB,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,WAAWJ,EAAE,MAAM,EAAE,CAAC,EAAErC,EAAE,KAAK,EAAEgH,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG,EAAE,EAAE,WAAW,CAACpF,CAAC,EAAE,MAAM,EAAEwC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAE,EAAE,WAAW,CAAC3C,CAAC,EAAE3B,EAAE,KAAK,EAAEuE,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGE,EAAE,EAAE,WAAW,CAAC,CAAC,EAAE,OAAO,EAAE0C,EAAE,EAAE,UAAU,IAAI1C,EAAE,MAAM,EAAE,GAAGE,EAAEqqH,GAAGxsH,EAAEC,EAAEu8G,GAAG9+G,EAAE,OAAO0B,EAAEE,EAAED,EAAEa,EAAEmB,EAAEqD,EAAE5C,EAAEE,EAAE2C,CAAC,EAAEC,EAAE,EAAE,SAAS3C,EAAE,MAAM,EAAEG,EAAE,OAAOwC,EAAE,GAAG,CAAC,IAAK,GAAE,CAACxC,EAAEu1C,EAAE,gDAAgD/yC,EAAE,EAAE,EAAE,KAAK,CAAC,IAAK,GAAE,CAACxC,EAAEu1C,EAAE,gDAAgD/yC,EAAE,GAAGA,EAAE,EAAE,EAAE,KAAK,CAAC,IAAK,GAAExC,EAAEu1C,EAAE,kDAAkD/yC,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAE,MAAM,QAAQxC,EAAE,EAAE,CAAC,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAEG,EAAE,MAAM,EAAE,YAAYjC,EAAE,MAAM,EAAE,EAAE,YAAY,EAAE,MAAM,EAAE,EAAE,YAAY,EAAE,MAAM,EAAE,EAAE,YAAY4B,EAAE,MAAM,EAAE,IAAI,MAAMK,CAAC,EAAE,IAAIC,EAAElC,EAAEmC,EAAE,EAAE,OAAOH,IAAIpC,EAAE,KAAKsC,EAAEi9G,GAAG,CAAC,OAAO,CAAC,EAAEn/G,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,CAACgC,EAAE9C,CAAC,CAAC,EAAE,QAAQ,CAAC,CAAC,EAAEiD,EAAEg9G,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,KAAKn9G,CAAC,EAAE,QAAQ,CAAC,CAAC,EAAE,EAAE,YAAYhC,EAAE,MAAM,EAAE,EAAE,YAAY,EAAE,MAAM,GAAG,CAACkC,EAAEC,EAAE,EAAEP,CAAC,CAAC,CAAC,IAAI4qH,IAAG,CAAC,WAAWprG,GAAG,YAAY,OAAO,UAAUkrG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI1vH,EAAE,CAACyvH,GAAGzvH,EAAE,KAAK,MAAMqkB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASsrG,IAAI3vH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,CAAC,EAAEF,EAAE,CAAC,aAAaM,EAAE,WAAWC,EAAE,SAAS,CAAC,EAAEL,EAAE,GAAGI,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACty4CA,EAAE,OAAO,EAAE,GAAGC,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM;AAAA,UACjDA,EAAE,OAAO,EAAE,GAAG,EAAE,MAAM,SAAS,EAAE,MAAM,IAAI,MAAM,sDAAsD,EAAE,OAAO,EAAE,IAAI,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAG2B,EAAE,EAAE,UAAU,IAAI1B,EAAE,MAAM,EAAE,GAAG2B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAE7B,EAAE,MAAM,GAAGsC,EAAEkkB,EAAE,cAAc,EAAE,KAAK,EAAEjkB,EAAE,EAAE,WAAW,CAACV,EAAES,CAAC,EAAEtC,EAAE,KAAK,EAAEwC,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGE,EAAE,EAAE,WAAW,CAACH,CAAC,EAAE,EAAE,KAAK,EAAEI,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGmB,EAAE,EAAE,WAAW,CAAC,CAAC,EAAE,OAAO,EAAE,EAAE,EAAE,UAAU,IAAIA,EAAE,MAAM,EAAE,GAAGurH,GAAG,EAAExtH,EAAEC,EAAEC,EAAEW,EAAEE,EAAE,CAAC,EAAE,IAAIuE,EAAE,EAAE,SAASrD,EAAE,MAAM,EAAE,EAAE,OAAOqD,EAAE,GAAG,CAAC,IAAK,GAAE,CAAC,EAAEizC,EAAE,yDAAyDjzC,EAAE,GAAGA,EAAE,EAAE,EAAE,KAAK,CAAC,IAAK,GAAE,CAAC,EAAEizC,EAAE,8CAA8CjzC,EAAE,GAAGA,EAAE,EAAE,EAAE,KAAK,CAAC,IAAK,GAAE,EAAEizC,EAAE,qDAAqD,EAAE,MAAM,IAAK,GAAE,CAAC,IAAI71C,EAAE,MAAM,KAAK,EAAE,SAASpE,EAAE,MAAM,CAAC,EAAEqE,EAAE,MAAM,KAAK,EAAE,SAAS7B,EAAE,MAAM,CAAC,EAAE,EAAEy3C,EAAE,gDAAgD71C,EAAEC,CAAC,EAAE,KAAK,CAAC,IAAK,GAAE,CAAC,IAAID,EAAE,MAAM,KAAK,EAAE,SAASpE,EAAE,MAAM,CAAC,EAAEqE,EAAE,MAAM,KAAK,EAAE,SAAS7B,EAAE,MAAM,CAAC,EAAE,EAAEy3C,EAAE,gDAAgD71C,EAAEC,CAAC,EAAE,KAAK,CAAC,QAAQ,EAAE,EAAE,CAAC,GAAG,EAAE,YAAYV,EAAE,MAAM,EAAE,EAAE,MAAM,EAAE,YAAYrB,EAAE,MAAM,EAAE,EAAE,YAAYE,EAAE,MAAM,EAAE,IAAI,MAAM,CAAC,EAAE,MAAM,CAACF,EAAEE,CAAC,CAAC,CAAC,IAAI6sH,IAAG,CAAC,WAAWvrG,GAAG,YAAY,OAAO,UAAUqrG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,GAAG9vH,EAAE,CAAC6vH,GAAG7vH,EAAE,KAAK,MAAM,yBAAyB,KAAK,CAAC,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS+vH,GAAG/vH,EAAE,EAAE,CAAC,GAAG,CAAC,QAAQE,EAAE,OAAOI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,QAAQ,EAAE,WAAW,CAAC,EAAED,EAAE2B,EAAE,EAAE,MAAM,GAAGC,EAAEhC,EAAE,SAAS,EAAE,OAAO+B,EAAE,EAAEA,CAAC,EAAE,GAAGW,EAAEX,EAAE,EAAEC,EAAE,EAAE,EAAE,GAAGU,EAAE,EAAE,MAAM,IAAI,MAAM43C,EAAE,wDAAwD,CAAC,EAAE,IAAI,EAAEj6C,EAAE,MAAM,MAAM,EAAE,EAAE,GAAGqC,EAAE,IAAIE,EAAE5C,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAGwC,EAAE7C,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG8C,EAAE9C,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGgE,EAAEhE,EAAE,WAAW,EAAEK,EAAE,KAAK,EAAE4D,EAAEjE,EAAE,UAAU,IAAIgE,EAAE,MAAM,EAAE,GAAGqD,EAAErH,EAAE,WAAW,CAAC,CAAC,EAAE,OAAO,EAAE6E,EAAE7E,EAAE,UAAU,IAAIqH,EAAE,MAAM,EAAE,GAAGsoH,GAAG/sH,EAAEu8G,GAAG9+G,EAAE,OAAOA,EAAE,MAAM,GAAGwC,EAAEC,EAAEmB,EAAEY,EAAE,EAAE,CAAC,EAAE,IAAIJ,EAAEzE,EAAE,SAASqH,EAAE,MAAM,EAAE3C,EAAE,OAAOD,EAAE,GAAG,CAAC,IAAK,GAAE,CAACC,EAAE41C,EAAE,wDAAwD,EAAE,KAAK,CAAC,IAAK,GAAE,CAAC51C,EAAE41C,EAAE,6DAA6D,EAAE,KAAK,CAAC,IAAK,GAAE51C,EAAE41C,EAAE,yDAAyD71C,EAAE,GAAGA,EAAE,EAAE,EAAE,MAAM,IAAK,GAAEC,EAAE41C,EAAE,uDAAuD71C,EAAE,GAAGA,EAAE,GAAGA,EAAE,EAAE,EAAE,MAAM,QAAQC,EAAE,EAAE,CAAC,GAAG1E,EAAE,YAAYqH,EAAE,MAAM,EAAE3C,EAAE,MAAM1E,EAAE,YAAYgE,EAAE,MAAM,EAAE,IAAI,MAAMU,CAAC,EAAE,OAAOV,CAAC,CAAC,SAAS8rH,IAAIhwH,EAAE,CAAC,OAAO+vH,GAAG/vH,EAAE,EAAE,CAAC,CAAC,IAAIiwH,IAAG,CAAC,WAAW3rG,GAAG,YAAY,OAAO,UAAUwrG,GAAG,WAAWE,GAAG,EAAE,SAASE,IAAIlwH,EAAE,CAAC,OAAO+vH,GAAG/vH,EAAE,EAAE,CAAC,CAAC,IAAImwH,IAAG,CAAC,WAAW5rG,GAAG,YAAY,OAAO,UAAUurG,GAAG,WAAWI,GAAG,EAAE,SAASE,IAAIpwH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,MAAME,EAAE,QAAQI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,CAAC,gBAAgB,EAAE,KAAK,CAAC,EAAEL,EAAE+B,EAAE6kB,EAAE,eAAe,EAAEvmB,EAAE,KAAK,EAAE,GAAG2B,EAAEs4C,EAAE,iBAAiBj6C,EAAE,EAAE0B,CAAC,EAAEE,EAAE,IAAI,MAAM5B,EAAE,MAAM,MAAM,EAAE,KAAK,CAAC,EAAEqC,EAAErC,EAAE,MAAM,MAAM,EAAE,OAAO2B,EAAE,IAAIW,GAAG,CAAC,IAAIC,EAAE,CAAC,GAAGF,CAAC,EAAEE,EAAEb,GAAGY,EAAE,IAAIE,EAAEo/G,GAAG,CAAC,OAAO,CAAC,EAAE5hH,CAAC,EAAE,MAAM,CAAC,MAAM4B,EAAE,KAAKW,CAAC,EAAE,QAAQxC,CAAC,CAAC,EAAE,OAAO6B,EAAEF,IAAIY,EAAEE,CAAC,CAAC,CAAC,CAAC,IAAIstH,IAAG,CAAC,WAAWnsG,GAAG,YAAY,OAAO,WAAWksG,GAAG,EAAME,IAAG3Q,GAAG57F,EAAE,EAAMwsG,IAAG5Q,GAAGj7F,EAAE,EAAM8rG,IAAI,GAAGC,IAAG5Q,GAAGp7F,GAAG+rG,GAAG,EAAME,GAAG,SAASC,IAAI3wH,EAAE,CAAC0wH,GAAG1wH,EAAE,KAAK,MAAM2lB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASirG,IAAI5wH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+B,EAAE,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAEC,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAG,OAAOyuH,GAAG,EAAEnwH,EAAE8+G,GAAG,EAAE,OAAOn9G,CAAC,EAAED,CAAC,CAAC,IAAI4uH,IAAG,CAAC,WAAWlrG,GAAG,YAAY,OAAO,UAAUgrG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAI/wH,EAAE,CAAC8wH,GAAG9wH,EAAE,KAAK,MAAM2kB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,QAAQ,QAAQ,QAAQ,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASqsG,IAAIhxH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAEL,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,QAAQ+B,EAAE,UAAUC,EAAE,QAAQC,EAAE,aAAaS,EAAE,YAAYC,EAAE,eAAeC,CAAC,EAAExC,EAAE,CAAC,iBAAiByC,EAAE,WAAWC,EAAE,WAAWkB,EAAE,UAAU,EAAE,cAAcqD,EAAE,MAAM,EAAE,IAAI5C,EAAE,QAAQC,CAAC,EAAEwxB,GAAG,UAAU71B,EAAE,MAAM,EAAE,EAAE0B,EAAEC,EAAEC,EAAES,EAAEC,EAAEC,CAAC,EAAE+B,EAAE,GAAGX,EAAEW,EAAEg9G,GAAG,CAAC,OAAO,CAAC,EAAEthH,CAAC,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAMyC,CAAC,CAAC,CAAC,UAAU,GAAGuE,EAAE,CAACuf,EAAE,OAAOvmB,EAAE,MAAM,QAAQ,EAAE,IAAI,yCAAyCA,EAAE,MAAM,QAAQ,EAAE,IAAIuE,EAAEsxB,GAAG,gBAAgB,EAAEzxB,EAAEC,CAAC,EAAE4C,EAAE26G,GAAG,CAAC,OAAO,CAAC,EAAE5hH,CAAC,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,KAAKuE,CAAC,CAAC,CAAC,EAAED,EAAEg9G,GAAG,CAAC,OAAO,CAAC,EAAEr6G,CAAC,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAMxE,CAAC,CAAC,CAAC,EAAE,EAAE,YAAYwE,EAAE,MAAM,CAAC,KAAK,CAAC,IAAI1C,EAAE,EAAE,WAAW/B,EAAE,SAAS,EAAEyE,EAAE,EAAE,UAAU,IAAIjH,EAAE,MAAM,EAAE,GAAGyE,EAAE,IAAI,WAAW,IAAI,WAAW8hB,EAAE,eAAevmB,EAAE,KAAK,CAAC,EAAE,MAAM,EAAEkH,EAAE,IAAI,WAAW,IAAI,WAAW,CAAC,EAAE,MAAM,EAAExC,EAAE,IAAI,WAAW,IAAI,WAAWN,CAAC,EAAE,MAAM,EAAEO,EAAE,IAAI,WAAW,IAAI,WAAWN,CAAC,EAAE,MAAM,EAAEO,EAAE,IAAI,WAAW,IAAI,WAAWpC,CAAC,EAAE,MAAM,EAAEqC,EAAE,IAAI,WAAW,IAAI,WAAW0hB,EAAE,eAAe/jB,CAAC,CAAC,EAAE,MAAM,EAAEsC,EAAE,EAAE,UAAU,IAAIP,EAAE,MAAM,EAAE,GAAGgsH,GAAGtpH,EAAExC,EAAEzE,EAAE,MAAM,OAAOkH,EAAExC,EAAEC,EAAEC,EAAEC,EAAErC,EAAE,OAAOsC,CAAC,EAAER,EAAEg9G,GAAG,CAAC,OAAO,CAAC,EAAE/8G,CAAC,EAAE,QAAQ,EAAE,MAAM,CAAC,MAAM9B,CAAC,CAAC,CAAC,EAAE,EAAE,YAAY8B,EAAE,MAAM,CAAC,CAAC,OAAOD,CAAC,CAAC,IAAIosH,IAAG,CAAC,WAAWtsG,GAAG,YAAY,OAAO,UAAUosG,IAAI,WAAWC,GAAG,EAAE,SAASE,IAAIlxH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,WAAW,CAAC,EAAEL,EAAE,CAAC,UAAU,EAAE,YAAY+B,EAAE,QAAQC,EAAE,SAASC,EAAE,SAASS,EAAE,uBAAuBC,CAAC,EAAEvC,EAAEwC,EAAE,EAAE,SAASvC,EAAE,MAAM,EAAEwC,EAAE,EAAE,SAAS,EAAE,MAAM,EAAE,CAACC,EAAEkB,CAAC,EAAEmwE,GAAGvxE,EAAEC,EAAE,EAAEd,EAAEC,EAAEC,EAAES,EAAEC,CAAC,EAAE,EAAE,EAAE,WAAW,CAACG,EAAE,MAAM,EAAE,QAAQ,EAAEuE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAEA,EAAE,YAAYvE,EAAE,IAAI,EAAE,EAAE,WAAW,EAAE,MAAM,OAAO,EAAE,OAAO,EAAE,mBAAmB,CAAC,EAAE,IAAIkB,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,IAAIitH,IAAG,CAAC,WAAWvsG,GAAG,YAAY,OAAO,WAAWssG,GAAG,EAAE,SAASE,IAAIpxH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,UAAU,CAAC,EAAEL,EAAE,CAAC,UAAU,CAAC,EAAEI,EAAE2B,EAAE,EAAE,SAAS1B,EAAE,MAAM,EAAE2B,EAAE,EAAE,SAAS,EAAE,MAAM,EAAE,CAACC,EAAES,EAAEC,CAAC,EAAEyxE,GAAGryE,EAAEC,EAAE,GAAG,CAAC,EAAEY,EAAEF,EAAE,OAAOG,EAAE,EAAE,WAAW,CAACD,EAAE,CAAC,EAAE,OAAO,EAAE,EAAE,mBAAmBC,CAAC,EAAE,IAAIZ,CAAC,EAAE,IAAI+B,EAAE,EAAE,WAAW,CAACpB,CAAC,EAAE,QAAQ,EAAEqB,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAEC,EAAE,YAAYvB,EAAE,IAAI2E,EAAE,EAAE,WAAW,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,mBAAmBA,CAAC,EAAE,IAAI1E,CAAC,EAAE,CAACE,EAAEmB,EAAEqD,CAAC,CAAC,CAAC,IAAI8pH,IAAG,CAAC,WAAWxsG,GAAG,YAAY,OAAO,WAAWusG,GAAG,EAAE,SAASE,IAAItxH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAEL,EAAE,CAAC,WAAW,CAAC,EAAEI,EAAE,EAAE,EAAE,SAASC,EAAE,MAAM,EAAE0B,EAAEsyE,GAAG,EAAE,CAAC,EAAEryE,EAAE,EAAE,WAAW3B,EAAE,MAAM,OAAO,EAAE,OAAO,EAAE,mBAAmB2B,CAAC,EAAE,IAAID,CAAC,EAAEC,CAAC,CAAC,IAAIqvH,IAAG,CAAC,WAAWzsG,GAAG,YAAY,OAAO,WAAWwsG,GAAG,EAAME,IAAI,GAAGC,IAAG5R,GAAG96F,GAAGysG,GAAG,EAAME,GAAG,SAASC,IAAI3xH,EAAE,CAAC0xH,GAAG1xH,EAAE,KAAK,MAAMgkB,GAAG,KAAK,CAAC,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAAS4tG,IAAI5xH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,KAAKO,EAAE,SAAS,CAAC,EAAED,EAAE,CAAC,EAAE,CAAC,EAAEJ,EAAE+B,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGC,EAAED,EAAEE,EAAE,EAAE,CAAC,WAAWS,EAAE,KAAKC,EAAE,aAAaC,EAAE,mBAAmBC,CAAC,EAAE69G,GAAG,EAAErgH,EAAE,CAAC,EAAEyC,EAAEH,EAAE,GAAGE,EAAE,CAAC,IAAI4B,EAAE,EAAE,UAAU,IAAI/B,EAAE,MAAM,EAAE,GAAG+B,IAAI1C,IAAIE,EAAES,EAAEV,EAAEyC,EAAE3B,EAAEw3C,EAAE,iBAAiBx3C,EAAE,OAAOb,EAAE,MAAM,MAAM,EAAE,CAACq4C,EAAE,2BAA2B,MAAMx3C,EAAEb,EAAE,MAAM,MAAM,EAAE,GAAG,CAAC+B,EAAE,CAAC,EAAEs2C,EAAE,0BAA0Br4C,EAAE,MAAMa,CAAC,EAAEuE,EAAEuf,EAAE,cAAc,CAAC,EAAE,EAAE,EAAE,WAAW5iB,EAAE/B,EAAE,KAAK,EAAE,GAAG2kB,EAAE,cAAc3kB,EAAE,KAAK,IAAI,EAAE,CAAC,IAAIwC,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAG+sH,GAAGxvH,EAAEqF,EAAE83G,GAAG,EAAE,OAAO16G,CAAC,CAAC,CAAC,GAAG5B,GAAG,EAAE,YAAYH,EAAE,MAAM,EAAE,EAAE,CAAC,IAAI+B,EAAE61C,EAAE,qBAAqB,EAAE,MAAM13C,CAAC,EAAE,EAAE,MAAM6B,CAAC,CAAC,OAAO,CAAC,CAAC,IAAIktH,IAAG,CAAC,WAAW7tG,GAAG,YAAY,OAAO,UAAU2tG,IAAI,WAAWC,GAAG,EAAME,IAAGnS,GAAG36F,EAAE,EAAM+sG,IAAGpS,GAAG16F,EAAE,EAAM+sG,GAAG,SAASC,IAAIjyH,EAAE,CAACgyH,GAAGhyH,EAAE,KAAK,MAAMklB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASgtG,IAAIlyH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,EAAEO,CAAC,EAAE,EAAE,EAAEL,EAAE,UAAU,IAAIK,EAAE,MAAM,EAAE,GAAG,CAAC,KAAK,CAAC,EAAED,EAAE2B,EAAE,IAAI,MAAM1B,EAAE,MAAM,MAAM,EAAE,QAAQuC,EAAE,EAAEA,EAAEb,EAAE,OAAOa,IAAIb,EAAEa,GAAGvC,EAAE,MAAMuC,GAAG,EAAEA,GAAG,IAAIZ,EAAE,IAAI,WAAW,IAAI,WAAW3B,EAAE,KAAK,EAAE,MAAM,EAAE4B,EAAE,IAAI,WAAW,IAAI,WAAWF,CAAC,EAAE,MAAM,EAAEW,EAAE1C,EAAE,WAAW+B,EAAE1B,EAAE,KAAK,EAAEsC,EAAE3C,EAAE,UAAU,IAAI0C,EAAE,MAAM,EAAE,GAAG,OAAOovH,GAAG,EAAE9vH,EAAE3B,EAAE,MAAM,OAAO4B,EAAEF,EAAE,OAAOo9G,GAAGz8G,EAAE,OAAOC,CAAC,EAAED,CAAC,CAAC,IAAIuvH,IAAG,CAAC,WAAWjtG,GAAG,YAAY,OAAO,UAAU+sG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIryH,EAAE,CAACoyH,GAAGpyH,EAAE,KAAK,MAAMmlB,GAAG,KAAK,CAAC,SAAS,QAAQ,SAAS,SAAS,SAAS,OAAO,SAAS,QAAQ,CAAC,CAAC,CAAC,IAAImtG,IAAI,CAAC,CAAC,OAAOtyH,EAAE,QAAQ,EAAE,MAAME,CAAC,IAAI,CAAC,GAAG,CAAC,EAAEI,CAAC,EAAEN,EAAE,CAAC,EAAEO,EAAE,OAAO,CAAC,EAAEL,EAAE,EAAE,EAAE,UAAU,IAAII,EAAE,MAAM,EAAE,GAAG2B,EAAE,IAAI,WAAW,IAAI,WAAW3B,EAAE,KAAK,EAAE,MAAM,EAAE4B,EAAE5B,EAAE,MAAM,MAAM,EAAE4B,EAAEA,EAAE,OAAO,GAAG3B,EAAE,IAAI4B,EAAE,EAAE,WAAWD,EAAE5B,EAAE,KAAK,EAAEsC,EAAE,EAAE,UAAU,IAAIT,EAAE,MAAM,EAAE,GAAGU,EAAE,EAAE,WAAWX,EAAE,OAAO,EAAEY,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAG,OAAOuvH,GAAG,EAAEnwH,EAAE3B,EAAE,MAAM,OAAO++G,GAAG/+G,EAAE,OAAOC,EAAE,EAAEqC,EAAEE,CAAC,EAAE,CAACX,EAAEU,CAAC,CAAC,EAAE0vH,IAAG,CAAC,WAAWptG,GAAG,YAAY,OAAO,UAAUktG,IAAI,WAAWC,GAAG,EAAME,GAAG,SAASC,IAAIzyH,EAAE,CAACwyH,GAAGxyH,EAAE,KAAK,MAAMolB,GAAG,KAAK,CAAC,SAAS,SAAS,OAAO,SAAS,SAAS,SAAS,SAAS,SAAS,SAAS,QAAQ,SAAS,QAAQ,SAAS,SAAS,SAAS,SAAS,QAAQ,CAAC,CAAC,CAAC,SAASstG,IAAI1yH,EAAE,CAAC,GAAG,CAAC,QAAQ,EAAE,OAAOE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,EAAE,WAAW,CAAC,EAAEL,EAAE,CAAC,cAAc,EAAE,SAAS+B,EAAE,UAAUC,EAAE,YAAYC,CAAC,EAAE7B,EAAE,CAACsC,EAAEC,EAAEC,EAAEC,CAAC,EAAExC,EAAE,MAAM,CAACyC,EAAEkB,CAAC,EAAE/B,GAAG,KAAKA,EAAE,CAACU,EAAEC,CAAC,EAAE,EAAE,CAACF,EAAEI,EAAEkB,EAAEnB,CAAC,EAAEwE,EAAE,IAAI,WAAW,IAAI,WAAWuf,EAAE,eAAevmB,EAAE,KAAK,CAAC,EAAE,MAAM,EAAE,EAAE,IAAI,WAAW,IAAI,WAAWumB,EAAE,eAAe,CAAC,CAAC,EAAE,MAAM,EAAEniB,EAAE,EAAE,WAAW,EAAEpE,EAAE,KAAK,EAAEqE,EAAE,EAAE,UAAU,IAAID,EAAE,MAAM,EAAE,GAAGG,EAAE,EAAE,UAAU,IAAIvE,EAAE,MAAM,EAAE,GAAGyE,EAAE,EAAE,UAAU,IAAI,EAAE,MAAM,EAAE,GAAGyC,EAAE,IAAI,UAAU,EAAE,EAAExC,EAAE,OAAOhD,EAAE,CAAC,IAAI,WAAWgD,EAAE,EAAE,MAAM,IAAI,UAAUA,EAAE,EAAE,MAAM,IAAI,OAAOA,EAAE,EAAE,MAAM,IAAI,UAAUA,EAAE,EAAE,MAAM,QAAQA,EAAE,EAAE,KAAK,CAAC,OAAOutH,GAAG1tH,EAAEE,EAAE,EAAE,MAAM,GAAG,EAAEpC,EAAEI,EAAEkB,EAAEnB,EAAED,EAAED,EAAE0E,EAAEhH,EAAE,MAAM,OAAO,EAAE,EAAE,EAAE,OAAO,EAAEkH,EAAExC,EAAE/C,EAAE0C,CAAC,EAAED,CAAC,CAAC,IAAIguH,IAAG,CAAC,WAAWvtG,GAAG,YAAY,OAAO,UAAUqtG,IAAI,WAAWC,GAAG,EAAE,SAASE,IAAI5yH,EAAE,CAAC,GAAG,CAAC,OAAO,EAAE,QAAQE,EAAE,MAAMI,CAAC,EAAEN,EAAE,CAAC,MAAMO,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,EAAED,EAAE,EAAE,IAAI,GAAGC,EAAE,MAAM,QAAQ,IAAI,EAAEA,EAAE,MAAM,GAAG0B,EAAE1B,EAAE,MAAM,OAAO2B,EAAE,IAAI,MAAMD,EAAE,CAAC,EAAEE,EAAE,EAAE,QAAQY,EAAE,EAAEA,EAAEd,EAAEc,IAAIA,IAAI,IAAIb,EAAEC,KAAK5B,EAAE,MAAMwC,IAAI,IAAIH,EAAE,IAAI,MAAM,CAAC,EAAEC,EAAE,IAAI,MAAMZ,CAAC,EAAE,KAAK,CAAC,EAAEa,EAAEvC,EAAE,MAAM,MAAM,EAAEuC,EAAE,GAAG,EAAE,QAAQC,EAAE,EAAEA,EAAEH,EAAE,OAAOG,IAAIF,EAAE,GAAGE,EAAEH,EAAEG,GAAGo/G,GAAG,CAAC,OAAO,CAAC,EAAE5hH,CAAC,EAAE,MAAM,CAAC,MAAMsC,EAAE,KAAKC,CAAC,EAAE,QAAQ5C,CAAC,CAAC,EAAE,OAAO0C,EAAE,IAAI,CAAC,CAAC,OAAOG,EAAE,MAAMC,CAAC,KAAK,CAAC,OAAOD,EAAE,MAAMC,EAAE,MAAMd,CAAC,EAAE,CAAC,CAAC,IAAI2wH,IAAG,CAAC,WAAWttG,GAAG,YAAY,OAAO,WAAWqtG,GAAG,EAAE,SAASE,IAAI9yH,EAAE,CAAC,GAAG,CAAC,OAAO,CAAC,EAAE,CAAC,EAAE,QAAQE,CAAC,EAAEF,EAAEM,EAAEJ,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAE,OAAOA,EAAE,mBAAmBI,CAAC,EAAE,KAAK,CAAC,EAAEA,CAAC,CAAC,IAAIyyH,IAAG,CAAC,WAAWrtG,GAAG,YAAY,OAAO,WAAWotG,GAAG,EAAME,IAAI,CAACtT,IAAGE,IAAGG,IAAGI,IAAGa,IAAGI,IAAGI,IAAGI,IAAGM,IAAGO,IAAGE,IAAGC,IAAGI,IAAGE,IAAGI,IAAGI,IAAGC,IAAGC,IAAGK,IAAGI,IAAGI,IAAGI,IAAGI,IAAGC,IAAGE,IAAGC,IAAGE,IAAGE,IAAGI,IAAGC,IAAGE,IAAGI,IAAGI,IAAGI,IAAGI,IAAGI,IAAGE,IAAGE,IAAGnH,IAAGuH,IAAGE,IAAGE,IAAGC,IAAGE,IAAGC,IAAGE,IAAGE,IAAGI,IAAGE,IAAGI,IAAGI,IAAGI,IAAGE,IAAGK,IAAGE,IAAGC,IAAGK,IAAGI,IAAGI,IAAGE,IAAGI,IAAGE,IAAGE,IAAGI,GAAGE,IAAGI,IAAGI,IAAGE,IAAGE,IAAGC,IAAGC,IAAGjL,IAAGqL,IAAGI,IAAGI,IAAGI,IAAGC,IAAGC,IAAGI,IAAGI,IAAGI,IAAGC,IAAGvM,IAAG2M,IAAGE,IAAGI,IAAGI,IAAGK,IAAGE,IAAGE,IAAGC,IAAGC,IAAGE,IAAGI,IAAGI,IAAGE,IAAGE,IAAGE,IAAGE,IAAGI,IAAGC,IAAGC,IAAGI,IAAGI,IAAGI,IAAGhS,IAAGkS,IAAGE,GAAE,EAAE,QAAQ/yH,KAAKgzH,IAAIvsG,GAAGzmB,CAAC,EAAE,IAAIizH,GAAGx4G,EAAE,EAAEw4G,GAAG,aAAa,wBAAwB,SAAS,CAAC,GAAG,CAAC,OAAO,YAAY,SAAS,IAAI,WAAW,CAAC,EAAE,GAAG,IAAI,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,IAAI,GAAG,GAAG,EAAE,CAAC,CAAC,CAAC,OAAOjzH,EAAN,CAAS,MAAM,EAAE,CAAC,CAAC,EAAEizH,GAAG,aAAa,+BAA+B,SAAS,CAAC,GAAGA,GAAG,IAAI,SAAS,EAAE,MAAM,GAAG,GAAG,CAAC,OAAO,IAAI,eAAe,EAAE,MAAM,YAAY,IAAI,kBAAkB,CAAC,CAAC,EAAE,YAAY,SAAS,IAAI,WAAW,CAAC,EAAE,GAAG,IAAI,IAAI,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,GAAG,EAAE,EAAE,EAAE,GAAG,EAAE,IAAI,GAAG,EAAE,EAAE,GAAG,EAAE,CAAC,CAAC,CAAC,OAAOjzH,EAAN,CAAS,MAAM,EAAE,CAAC,CAAC,EAAE,IAAIkzH,GAAG1yH,GAAG+F,GAAG,CAAC,EAAE4sH,IAAG3yH,GAAGoW,GAAG,CAAC,EAAEw8G,GAAG5yH,GAAGuW,GAAG,CAAC,EAAMs8G,GAAGH,GAAG,SAASA,GAAGI,IAAIF,GAAG,SAASA,GAAGG,GAAG,cAAcn8G,EAAE,CAAC,YAAYtW,EAAE,CAAC,MAAM,EAAE,KAAK,KAAKA,EAAE,KAAK,iBAAiB,EAAE,KAAK,KAAK,KAAK,qBAAqB0yH,EAAE,EAAEC,GAAG,KAAK,KAAK,KAAK,gBAAgB,EAAE,KAAK,UAAU,IAAIt8G,GAAG,KAAKmc,GAAG,CAAC,CAAC,CAAC,MAAMxyB,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,CAAC,GAAG,KAAK,kBAAkB,EAAE,OAAO,KAAK,KAAKA,EAAEO,EAAEZ,EAAE,EAAE,CAAC,EAAEK,CAAC,CAAC,YAAY,CAAC,OAAO,KAAK,UAAU,WAAW,CAAC,CAAC,MAAM,KAAKO,EAAE,CAAC,IAAIZ,EAAE4mB,EAAE,IAAI,EAAE,OAAOhmB,EAAE,EAAE,CAAC,SAASgmB,EAAE,IAAI,EAAE5mB,CAAC,CAAC,CAAC,KAAKY,EAAEZ,EAAE,EAAEK,EAAEwB,EAAE,CAAC,IAAIC,EAAE,KAAK,mBAAmB,GAAGzB,IAAI,SAAS,CAAC,IAAIqC,EAAE1C,EAAE,KAAK,UAAU,IAAIY,EAAE,CAAC,GAAGkB,EAAE,YAAYY,EAAE,MAAM,EAAE,MAAMrC,EAAE,aAAa,KAAK,SAASwB,CAAC,CAAC,EAAE,MAAM,CAAC,IAAIE,EAAE6kB,EAAE,cAAc,CAAC,EAAE5kB,EAAED,EAAE6kB,EAAE,gBAAgBvmB,CAAC,EAAE,EAAE,KAAK,KAAK,QAAQ2B,CAAC,EAAE,KAAK,UAAU,IAAIpB,EAAE,CAAC,GAAGkB,EAAE,aAAa,EAAE,MAAM,EAAE,MAAMzB,EAAE,SAASwB,CAAC,CAAC,EAAE,KAAK,KAAK,KAAK,eAAeC,EAAEC,EAAE,CAAC,EAAE/B,GAAG,MAAM,KAAK,KAAK,OAAO,IAAI,IAAI,WAAWA,EAAE,OAAOA,EAAE,WAAWgC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,KAAKpB,EAAE,CAAC,OAAO,KAAK,SAASA,CAAC,CAAC,CAAC,SAASA,EAAEZ,EAAE,EAAE,CAAC,GAAG,CAAC,aAAaK,EAAE,MAAMwB,EAAE,MAAMC,EAAE,YAAYC,CAAC,EAAE,KAAK,UAAU,IAAInB,CAAC,EAAE,GAAGiB,IAAI,SAAS,OAAO7B,GAAG,MAAMA,IAAI,KAAK,GAAG,MAAM,GAAG+B,EAAE,QAAQA,EAAEA,EAAE,MAAM/B,EAAE,CAAC,EAAEA,EAAEA,GAAG,EAAE,EAAE,GAAG4mB,EAAE,cAAc9kB,CAAC,EAAE,IAAIE,EAAE4kB,EAAE,gBAAgB/kB,CAAC,EAAE,EAAE,KAAK,KAAK,OAAO,MAAMxB,EAAEL,EAAEgC,EAAE3B,EAAE,EAAE2B,CAAC,EAAE,OAAOwxH,IAAI,EAAE,OAAO3xH,CAAC,CAAC,CAAC,YAAYjB,EAAEZ,EAAE,GAAG,CAAC,GAAG,KAAK,UAAU,IAAIY,CAAC,EAAE,CAAC,IAAI,EAAE,KAAK,UAAU,IAAIA,CAAC,EAAE,GAAG,EAAE,WAAW,CAACZ,GAAG,EAAE,SAAS,EAAE,MAAM,GAAG,KAAK,KAAK,MAAM,EAAE,YAAY,EAAE,KAAK,KAAK,KAAK,YAAY,EAAE,EAAE,EAAE,KAAK,UAAU,OAAOY,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC,SAASA,EAAE,CAAC,OAAO,KAAK,UAAU,IAAIA,CAAC,EAAE,KAAK,UAAU,IAAIA,CAAC,EAAE,SAAS,CAAC,CAAC,OAAOA,EAAE,CAAC,IAAIZ,EAAE,KAAK,UAAU,IAAIY,CAAC,EAAEZ,GAAG,MAAMA,EAAE,UAAU,CAAC,gBAAgB,CAAC,MAAO,GAAE,CAAC,gBAAgBY,EAAE,CAAC,OAAO,KAAK,UAAU,IAAIA,CAAC,EAAE,YAAY,CAAC,SAAS,CAAC,KAAK,KAAK,KAAK,QAAQ,EAAE,YAAY,KAAK,MAAM,KAAK,KAAK,QAAQ,oBAAoB,EAAE,KAAK,KAAK,IAAI,CAAC,QAAQ,CAAC,MAAM,CAAC,WAAW,EAAE,CAAC,CAAC,WAAWA,EAAEZ,EAAE,EAAE,CAAC,IAAIK,EAAE,GAAG,GAAG,KAAKA,EAAE,KAAK,MAAM,KAAKO,EAAEZ,CAAC,MAAM,CAAC,IAAI6B,EAAE,KAAK,mBAAmBxB,EAAE,CAAC,GAAGwB,CAAC,EAAE,KAAK,UAAU,IAAIxB,EAAE,CAAC,GAAGwB,EAAE,aAAa,EAAE,MAAMjB,EAAE,MAAMZ,EAAE,SAAS,CAAC,CAAC,EAAE,IAAI8B,EAAE8kB,EAAE,cAAchmB,CAAC,EAAE,KAAK,KAAK,KAAK,eAAeiB,EAAEC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,OAAOzB,EAAE,MAAMO,EAAE,MAAMZ,CAAC,CAAC,CAAC,mBAAmB,CAAC,MAAMY,EAAE,MAAMZ,EAAE,OAAO,CAAC,EAAE,CAAC,IAAIK,EAAE,KAAK,KAAK,OAAO,OAAO,CAAC,aAAawB,CAAC,EAAE,KAAK,UAAU,IAAI,CAAC,EAAEC,EAAE8kB,EAAE,cAAchmB,CAAC,EAAE,OAAOZ,EAAE,CAAC,IAAI,UAAU,OAAO,IAAI,aAAaK,EAAEwB,EAAEC,CAAC,EAAE,IAAI,QAAQ,OAAO,IAAI,WAAWzB,EAAEwB,EAAEC,CAAC,EAAE,IAAI,OAAO,OAAO,IAAI,WAAWzB,EAAEwB,EAAEC,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,iBAAiB9B,GAAG,CAAC,CAAC,CAAC,EAAE,SAASyzH,IAAI3zH,EAAE,CAAC,MAAM,CAAC,EAAEE,KAAK4mB,EAAE,MAAM9mB,EAAE,CAAC,YAAY,aAAa,CAAC,EAAE,KAAKM,GAAG,CAACA,EAAE,IAAI,EAAE,IAAI,EAAE,uCAAuCN,IAAI,EAAEM,EAAE,YAAY,EAAE,KAAKC,GAAG,CAAC,YAAY,YAAYA,EAAE,CAAC,EAAE,KAAK,GAAG,CAACL,EAAE,EAAE,SAAS,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,SAAS0zH,GAAG5zH,EAAE,EAAEE,EAAE,CAAC,GAAG2zH,IAAI,KAAK,OAAOA,GAAG,IAAIvzH,EAAE,yBAAyB,OAAON,GAAG,EAAEM,EAAE,uCAAuCN,IAAIM,EAAE,+BAA+BwzH,IAAI,MAAMA,GAAGxzH,IAAI,KAAKwzH,GAAGxzH,GAAGJ,EAAEI,CAAC,CAAC,eAAeyzH,KAAI,CAAC,GAAG,CAAC/zH,EAAE,CAAC,EAAE,MAAM,QAAQ,IAAI,CAACya,EAAE,EAAE,SAAS,uBAAuB,EAAEA,EAAE,EAAE,SAAS,8BAA8B,CAAC,CAAC,EAAE,OAAO,IAAI,QAAQ,CAACva,EAAEI,IAAI,CAAC,IAAIC,EAAE,CAAC,EAAEA,EAAE,WAAW,CAAC0B,EAAEC,IAAI,CAAC,GAAGD,EAAE,SAAS,YAAY,EAAE,CAAC,IAAIE,EAAEgxH,IAAG,mBAAmB,QAAQ,MAAM,KAAK,EAAEvwH,EAAE,IAAI,KAAK,CAACT,CAAC,EAAE,CAAC,KAAK,wBAAwB,CAAC,EAAE,OAAO,IAAI,gBAAgBS,CAAC,CAAC,CAAC,OAAOX,EAAE,SAAS,OAAO,EAAE2xH,GAAG5zH,EAAE,EAAEg0H,IAAI,KAAKA,GAAG9xH,CAAC,EAAEA,EAAED,CAAC,EAAEgyH,KAAK1zH,EAAE,gBAAgBozH,IAAIC,GAAG5zH,EAAE,EAAEg0H,IAAI,KAAKA,GAAG,EAAE,CAAC,GAAG,IAAI,EAAE,GAAGzzH,EAAE,QAAQ,IAAI,CAAI,GAAG2zH,KAAUA,GAAG,GAAG5zH,EAAE,CAAC,QAAQ,iMAAiM,CAAC,EAAC,EAAE,IAAI,EAAE,GAAGN,GAAG6zH,IAAI,MAAMtzH,EAAE,oBAAoB,IAAI,KAAK,CAAC,uCAAuC8yH,GAAG,SAAS,CAAC,EAAE,CAAC,KAAK,iBAAiB,CAAC,EAAE,EAAEA,GAAG9yH,CAAC,GAAG,EAAE+yH,IAAI/yH,CAAC,EAAE,EAAE,KAAK0B,GAAG,CAAC,EAAE,GAAGiyH,GAAG,GAAG,IAAIhyH,EAAE,KAAKD,EAAE,KAAK,CAAC,KAAKA,EAAE,MAAM,OAAO,KAAK,CAAC,CAAC,EAAE,qBAAqBA,EAAE,MAAM,0BAA0B,KAAK,CAAC,QAAQ,CAAC,EAAE,gBAAgBA,EAAE,MAAM,oBAAoB,SAAS,CAAC,CAAC,EAAE,eAAeA,EAAE,MAAM,kBAAkB,KAAK,CAAC,SAAS,SAAS,QAAQ,CAAC,EAAE,YAAYA,EAAE,MAAM,eAAeC,EAAE,CAAC,QAAQ,CAAC,EAAE,QAAQD,EAAE,MAAM,UAAUC,EAAE,CAAC,CAAC,CAAC,EAAEhC,EAAE,CAAC,KAAK+B,CAAC,CAAC,CAAC,CAAC,EAAE,MAAM3B,CAAC,CAAC,CAAC,CAAC,CAAC,SAASozH,IAAI1zH,EAAE,EAAE,CAAC,OAAO,EAAE,CAAC,IAAI,UAAU,OAAO,IAAI,aAAaA,CAAC,EAAE,IAAI,QAAQ,OAAO,IAAI,WAAWA,CAAC,EAAE,IAAI,OAAO,OAAO,IAAI,WAAWA,CAAC,EAAE,QAAQ,MAAM,IAAI,MAAM,iBAAiB,GAAG,CAAC,CAAC,CAAC,IAAIm0H,IAAI,CAAC,yBAAyB,8BAA8B,sCAAsC,EAAEN,GAAG,KAAKG,GAAG,KAAKF,GAAG,CAAC,EAAEI,GAAG,GAAGD,GAAG,GAAG,SAASG,IAAIp0H,EAAE,EAAE,GAAG,CAAC,GAAGozB,GAAG,mGAAmG,EAAE8gG,GAAG,MAAM,IAAI,MAAM,gIAAgI,EAAEL,GAAG7zH,EAAEi0H,GAAG,CAAC,CAAC,SAASI,IAAIr0H,EAAE,EAAE,GAAG,CAAC,GAAGk0H,GAAG,MAAM,IAAI,MAAM,iIAAiI,EAAE,GAAG,OAAOl0H,GAAG,SAASg0H,GAAGh0H,MAAM,CAAC8zH,GAAG9zH,EAAE,IAAIE,EAAEi0H,IAAI,OAAO7zH,GAAGwzH,GAAGxzH,IAAI,IAAI,EAAE,GAAGJ,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,2DAA2DA,EAAE,KAAK,GAAG,gKAAgK,CAAC,CAAC+zH,GAAG,CAAC,CAAC,IAAIT,GAAG,GAAGC,GAAG,GAAG,SAASa,IAAIt0H,EAAE,CAACwzH,GAAGxzH,CAAC,CAAC,SAASu0H,KAAK,CAAC,GAAGd,KAAK,GAAG,MAAM,IAAI,MAAM,+BAA+B,EAAE,OAAOA,EAAE,CAAC,IAAIe,IAAI,QAAYC,IAAI,EAAEtgG,GAAG,OAAO,SAAS,CAAC,GAAG,CAAC,KAAKn0B,CAAC,EAAE,MAAM+zH,IAAG,EAAE,OAAO,IAAIR,GAAGvzH,CAAC,CAAC,EAAEy0H,GAAG,EAAE,IAAIC,GAAG,QAAQC,IAAI,QAAQC,IAAI,QAAQC,IAAI,QAAQC,IAAI,QAAQC,IAAI,CAAC,KAAKL,GAAG,YAAYA,GAAG,iBAAiBC,IAAI,mBAAmBC,IAAI,qBAAqBC,IAAI,oBAAoBC,GAAG,EChxJr3e,IAAAE,GAAA,GAAAC,GAAAD,GAAA,oBAAAE,GAAA,YAAAC,GAAA,mBAAAC,GAAA,sBAAAC,GAAA,6BAAAC,GAAA,kBAAAC,GAAA,yBAAAC,GAAA,gBAAAC,GAAA,mBAAAC,IAAA,wBAAAC,IAAA,sBAAAC,MCEO,SAASC,GACdC,EACAC,EACAC,EAAW,GACX,CASA,GARAF,EAAI,UAAU,EAEdC,EAAO,MAAM,CAAC,EAAE,QAAQ,CAAC,CAAE,EAAAE,EAAG,EAAAC,CAAE,EAAGC,IAAY,CAC7C,IAAMC,EAAOL,EAAOI,GACpBL,EAAI,OAAOM,EAAK,EAAGA,EAAK,CAAC,EACzBN,EAAI,OAAOG,EAAGC,CAAC,CACjB,CAAC,EAEGF,EAAU,CACZ,IAAMI,EAAOL,EAAOA,EAAO,OAAS,GAC9BM,EAAKN,EAAO,GAClB,GAAI,CAACK,GAAQ,CAACC,EACZ,OAGFP,EAAI,OAAOM,EAAK,EAAGA,EAAK,CAAC,EACzBN,EAAI,OAAOO,EAAG,EAAGA,EAAG,CAAC,CACvB,CAEAP,EAAI,OAAO,CACb,CC3BA,IAAAQ,GAAA,GAAAC,GAAAD,GAAA,+BAAAE,GAAA,mBAAAC,GAAA,iBAAAC,GAAA,WAAAC,GAAA,YAAAC,GAAA,aAAAC,GAAA,eAAAC,IAAA,eAAAC,GAAA,eAAAC,GAAA,eAAAC,GAAA,kBAAAC,GAAA,uBAAAC,GAAA,UAAAC,GAAA,UAAAC,KCOO,IAAMC,GAAN,KAAwC,CAK7C,YAAYC,EAAeC,EAAgB,CACzC,GAAI,CAACC,GAAcF,CAAK,GAAK,CAACE,GAAcD,CAAM,EAChD,MAAM,IAAI,MAAM,wFAAwF,KAAK,UAAU,CAAE,MAAAD,EAAO,OAAAC,CAAO,CAAC,GAAG,EAG7I,KAAK,OAASD,EACd,KAAK,QAAUC,CACjB,CAEA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,QAAiB,CAAE,OAAO,KAAK,OAAS,CAE5C,SAAsB,CAC3B,OAAO,IAAIF,GAAW,EAAI,KAAK,MAAO,EAAI,KAAK,MAAM,CACvD,CACF,EDvBO,SAASI,GAASC,EAAaC,EAAa,CACjD,OAAOD,aAAqBE,IAAUF,EAAO,MAAM,SAAWC,CAChE,CAEO,SAASE,IAAWH,EAAoC,CAC7D,OAAOD,GAASC,EAAQ,CAAC,CAC3B,CAEO,SAASI,GAAWJ,EAAoC,CAC7D,OAAOD,GAASC,EAAQ,CAAC,CAC3B,CAEO,SAASK,GAAWL,EAAoC,CAC7D,OAAOD,GAASC,EAAQ,CAAC,CAC3B,CAEO,SAASM,GAAWN,EAAoC,CAC7D,OAAOD,GAASC,EAAQ,CAAC,CAC3B,CAEO,SAASO,GAAQC,EAAa,CACnC,OAAOA,EAAM,IAAM,CACrB,CAEO,SAASC,GAAOD,EAAa,CAClC,OAAOA,EAAM,IAAM,CACrB,CAEO,SAASE,GAAMF,EAAaG,EAAO,EAAG,CAC3C,IAAMC,EAAI,IAAMD,EAChB,OAAO,KAAK,MAAMH,EAAMI,CAAC,EAAIA,CAC/B,CAEO,SAASC,GAAaC,EAAmB,CAC9C,OAAOA,GAAOA,EAAI,OAASA,EAAI,MACjC,CAEO,SAASC,GAA0B,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAgBC,EAAmB,CAC3F,IAAMC,EAAQD,EAAY,KAAK,IAAID,EAAQD,CAAK,EAChD,OAAO,IAAII,GAAW,KAAK,MAAMJ,EAAQG,CAAK,EAAG,KAAK,MAAMF,EAASE,CAAK,CAAC,CAC7E,CAEO,SAASE,GAAeC,EAAqB,CAClD,OAAOA,EAAI,OAAO,CAACC,EAAKC,IAAOD,EAAI,IAAIC,CAAE,EAAG,IAAIC,GAAM,EAAG,CAAC,CAAC,EACxD,IAAI,IAAIA,GAAMH,EAAI,OAAQA,EAAI,MAAM,CAAC,CAC1C,CAEO,SAASI,GAAMlB,EAAamB,EAAeC,EAAwB,CACxE,OAAO,MAAMpB,CAAG,EAAE,KAAK,CAAC,EAAE,IAAI,CAACqB,EAAGC,IAAMH,EAASG,EAAIF,CAAK,CAC5D,CAEO,SAASG,GAAcvB,EAAU,CACtC,MAAO,CAAC,CAACA,GAAQA,IAAQ,KAAcA,IAAQ,MAAc,CAAC,OAAO,MAAMA,CAAG,GAAKA,IAAQ,CAC7F,CAEO,SAASwB,GAAmBxB,EAAU,CAC3C,OAAOuB,GAAcvB,CAAG,GAAKA,GAAO,GAAKA,GAAO,CAClD,CEzDO,IAAMyB,GAAN,KAA8B,CAKnC,YAAYC,EAAWC,EAAW,CAChC,KAAK,GAAKD,EACV,KAAK,GAAKC,CACZ,CAEA,IAAI,GAAY,CAAE,OAAO,KAAK,EAAI,CAElC,IAAI,GAAY,CAAE,OAAO,KAAK,EAAI,CAE3B,IAAIC,EAAmB,CAC5B,OAAO,IAAIH,GAAM,KAAK,EAAIG,EAAG,EAAG,KAAK,EAAIA,EAAG,CAAC,CAC/C,CAEO,IAAIA,EAAmB,CAC5B,OAAO,IAAIH,GAAM,KAAK,EAAIG,EAAG,EAAG,KAAK,EAAIA,EAAG,CAAC,CAC/C,CAEO,IAAIA,EAAmB,CAC5B,OAAO,IAAIH,GAAM,KAAK,EAAIG,EAAG,EAAG,KAAK,EAAIA,EAAG,CAAC,CAC/C,CAEO,IAAIA,EAAmB,CAC5B,OAAO,IAAIH,GAAM,KAAK,EAAIG,EAAG,EAAG,KAAK,EAAIA,EAAG,CAAC,CAC/C,CAEO,KAAa,CAClB,OAAO,IAAIH,GAAM,KAAK,IAAI,KAAK,CAAC,EAAG,KAAK,IAAI,KAAK,CAAC,CAAC,CACrD,CAEO,WAAoB,CACzB,OAAO,KAAK,KAAM,KAAK,GAAK,EAAM,KAAK,GAAK,CAAE,CAChD,CAEO,OAAe,CACpB,OAAO,IAAIA,GAAM,KAAK,MAAM,KAAK,CAAC,EAAG,KAAK,MAAM,KAAK,CAAC,CAAC,CACzD,CACF,ECxCO,IAAMI,GAAN,KAAwD,CAC7D,OAAc,OAAOC,EAAoB,CACvC,MAAO,CAAC,CAACA,GAAQ,CAACA,EAAK,EAAGA,EAAK,EAAGA,EAAK,MAAOA,EAAK,MAAM,EAAE,MAAMC,EAAa,CAChF,CAEA,OAAc,iBAAiBC,EAAUC,EAAgBC,EAA0B,GAAO,CACxF,GAAI,CAACL,GAAI,OAAOG,CAAG,EACjB,MAAM,IAAI,MAAM,GAAGC,oBAAyB,KAAK,UAAUD,CAAG,wDAAwD,EAGxH,GAAI,CAACE,IAA4BF,EAAI,MAAQ,GAAKA,EAAI,OAAS,GAC7D,MAAM,IAAI,MAAM,GAAGC,cAAmBD,EAAI,sBAAsBA,EAAI,kCAAkC,CAE1G,CAUA,YAAYG,EAA4BD,EAA0B,GAAM,CACtE,IAAMF,EAAOG,GAAQ,CAAC,EAEhBC,EAAS,CAACJ,EAAI,KAAMA,EAAI,IAAKA,EAAI,MAAOA,EAAI,MAAM,EAAE,MAAMD,EAAa,EACvEM,EAAS,CAACL,EAAI,EAAGA,EAAI,EAAGA,EAAI,MAAOA,EAAI,MAAM,EAAE,MAAMD,EAAa,EAExE,GAAI,CAACM,GAAU,CAACD,EACd,MAAM,IAAI,MAAM,2EAA2E,KAAK,UAAUJ,CAAG,GAAG,EAGlH,GAAM,CAACM,EAAGC,EAAGC,EAAOC,CAAM,EAAIJ,EAC1B,CAACL,EAAI,EAAGA,EAAI,EAAGA,EAAI,MAAOA,EAAI,MAAM,EACpC,CAACA,EAAI,KAAMA,EAAI,IAAKA,EAAI,MAAQA,EAAI,KAAMA,EAAI,OAASA,EAAI,GAAG,EAElEH,GAAI,iBAAiB,CACnB,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CACf,EAAG,kBAAmBP,CAAuB,EAE7C,KAAK,GAAKI,EACV,KAAK,GAAKC,EACV,KAAK,OAASC,EACd,KAAK,QAAUC,CACjB,CAEA,IAAW,GAAY,CAAE,OAAO,KAAK,EAAI,CAEzC,IAAW,GAAY,CAAE,OAAO,KAAK,EAAI,CAEzC,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,QAAiB,CAAE,OAAO,KAAK,OAAS,CAEnD,IAAW,MAAe,CAAE,OAAO,KAAK,CAAG,CAE3C,IAAW,KAAc,CAAE,OAAO,KAAK,CAAG,CAE1C,IAAW,OAAgB,CAAE,OAAO,KAAK,EAAI,KAAK,KAAO,CAEzD,IAAW,QAAiB,CAAE,OAAO,KAAK,EAAI,KAAK,MAAQ,CAE3D,IAAW,MAAe,CAAE,OAAO,KAAK,MAAQ,KAAK,MAAQ,CAE7D,IAAW,SAAiB,CAAE,OAAO,IAAIC,GAAM,KAAK,KAAM,KAAK,GAAG,CAAG,CAErE,IAAW,UAAkB,CAAE,OAAO,IAAIA,GAAM,KAAK,MAAO,KAAK,GAAG,CAAG,CAEvE,IAAW,YAAoB,CAAE,OAAO,IAAIA,GAAM,KAAK,KAAM,KAAK,MAAM,CAAG,CAE3E,IAAW,aAAqB,CAAE,OAAO,IAAIA,GAAM,KAAK,MAAO,KAAK,MAAM,CAAG,CAEtE,OAAsB,CAC3B,GAAM,CAACJ,EAAGC,EAAGC,EAAOC,CAAM,EAAI,CAAC,KAAK,EAAG,KAAK,EAAG,KAAK,MAAO,KAAK,MAAM,EACnE,IAAKE,GAAQ,KAAK,MAAMA,CAAG,CAAC,EAC/B,OAAO,IAAId,GAAI,CACb,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CACf,CAAC,CACH,CAEO,OAAsB,CAC3B,GAAM,CAACH,EAAGC,EAAGC,EAAOC,CAAM,EAAI,CAAC,KAAK,EAAG,KAAK,EAAG,KAAK,MAAO,KAAK,MAAM,EACnE,IAAKE,GAAQ,KAAK,MAAMA,CAAG,CAAC,EAC/B,OAAO,IAAId,GAAI,CACb,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CACf,CAAC,CACH,CAEO,UAAyB,CAC9B,GAAI,CACF,EAAAH,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CACf,EAAI,KACEG,EAAO,KAAK,IAAIJ,EAAQC,CAAM,EACpC,OAAID,EAAQC,IACVH,GAAMM,EAAO,EACbJ,GAASI,GAEPH,EAASD,IACXD,GAAMK,EAAO,EACbH,GAAUG,GAGL,IAAIf,GAAI,CAAE,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,CAAC,CACxC,CAEO,QAAQI,EAAuC,CACpD,IAAMC,EAASC,GAAaF,CAAC,EAAKA,EAAkB,MAAQA,EACtDG,EAASD,GAAaF,CAAC,EAAKA,EAAkB,OAASA,EAC7D,OAAO,IAAIhB,GAAI,CACb,EAAG,KAAK,EAAIiB,EACZ,EAAG,KAAK,EAAIE,EACZ,MAAO,KAAK,MAAQF,EACpB,OAAQ,KAAK,OAASE,CACxB,CAAC,CACH,CAEO,IAAIC,EAAcC,EAA4B,CACnD,GAAM,CAACZ,EAAGC,EAAGC,EAAOC,CAAM,EAAI,CAC5B,KAAK,EAAKQ,EAAO,EACjB,KAAK,EAAKC,EAAO,EACjB,KAAK,MAAQD,EACb,KAAK,OAASC,CAChB,EACA,OAAO,IAAIrB,GAAI,CAAE,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,CAAC,CACxC,CAEO,mBAAmBU,EAAkBC,EAAiC,CAC3E,GAAM,CAAE,EAAAd,EAAG,EAAAC,EAAG,MAAAc,EAAO,OAAAC,CAAO,EAAI,KAC1BC,EAAW,KAAK,IAAIjB,EAAG,CAAC,EACxBkB,EAAW,KAAK,IAAIjB,EAAG,CAAC,EAExBkB,EAAWJ,EAAQE,EACnBG,EAAYJ,EAASE,EACrBG,EAAe,KAAK,IAAIF,EAAUN,EAAWI,CAAQ,EACrDK,EAAgB,KAAK,IAAIF,EAAWN,EAAYI,CAAQ,EAE9D,OAAQ,IAAI3B,GAAI,CAAE,EAAG0B,EAAU,EAAGC,EAAU,MAAOG,EAAc,OAAQC,CAAc,CAAC,EAAG,MAAM,CACnG,CAEO,MAAMC,EAAYC,EAA0B,CACjD,GAAM,CAAE,MAAAtB,EAAO,OAAAC,CAAO,EAAI,KACpBH,EAAI,KAAK,EAAIuB,EACbtB,EAAI,KAAK,EAAIuB,EAEnB,OAAO,IAAIjC,GAAI,CAAE,EAAAS,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,CAAC,CACxC,CAEO,aAAasB,EAAqBC,EAAoB,CAC3D,IAAMC,EAAI,KAAK,MAAQ,EACjBC,EAAI,KAAK,OAAS,EAElBC,EAAK,EACLC,EAAK,EACPC,EAAMJ,EACNK,EAAMJ,EAEN5B,EAAI,KAAK,KACTC,EAAI,KAAK,IACTgC,EAAK,KAAK,MACVC,EAAK,KAAK,OAEd,OAAID,EAAKP,IACPK,EAAM,CAACE,EAAKP,EAAaC,EACzBM,EAAKP,GAEHQ,EAAKT,IACPO,EAAM,CAACE,EAAKT,EAAcG,EAC1BM,EAAKT,GAEHzB,EAAI,IACNgC,EAAM,EAAIhC,EACVA,EAAI,GAEFC,EAAI,IACN+B,EAAM,EAAI/B,EACVA,EAAI,GAGC,CAAE,GAAA6B,EAAI,IAAAE,EAAK,GAAAH,EAAI,IAAAE,EAAK,EAAA9B,EAAG,GAAAiC,EAAI,EAAAlC,EAAG,GAAAiC,EAAI,EAAAN,EAAG,EAAAC,CAAE,CAChD,CAEO,UAAUO,EAAa,CAC5B,OAAO,IAAI5C,GAAI,CACb,KAAM,KAAK,KAAQ4C,EAAO,KAAO,KAAK,MACtC,IAAK,KAAK,IAAOA,EAAO,IAAM,KAAK,OACnC,MAAO,KAAK,MAASA,EAAO,MAAQ,KAAK,MACzC,OAAQ,KAAK,OAAUA,EAAO,OAAS,KAAK,MAC9C,CAAC,EAAE,SAAS,EAAE,MAAM,CACtB,CACF,EC3LO,IAAMC,GAAN,cAA0BC,EAA4B,CAC3D,YAAYC,EAAcC,EAAaC,EAAeC,EAAgBC,EAA0B,GAAO,CACrG,MAAM,CAAE,KAAAJ,EAAM,IAAAC,EAAK,MAAAC,EAAO,OAAAC,CAAO,EAAGC,CAAuB,CAC7D,CACF,ECTO,IAAMC,GAAN,KAAsB,CAW3B,YACEC,EACAC,EACAC,EACAC,EACAC,EACA,CACA,KAAK,WAAa,IAAIC,GAAWD,EAAU,MAAOA,EAAU,MAAM,EAClE,KAAK,OAASJ,EACd,KAAK,YAAcC,EACnB,KAAK,WAAaC,EAClB,KAAK,KAAO,IAAII,GAAIH,CAAW,EAAE,QAAQ,KAAK,UAAU,CAC1D,CAEA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,YAAqB,CAAE,OAAO,KAAK,WAAa,CAE3D,IAAW,WAAoB,CAAE,OAAO,KAAK,UAAY,CAEzD,IAAW,KAAW,CAAE,OAAO,KAAK,IAAM,CAE1C,IAAW,WAAwB,CAAE,OAAO,KAAK,UAAY,CAE7D,IAAW,YAAqB,CAAE,OAAO,KAAK,UAAU,KAAO,CAE/D,IAAW,aAAsB,CAAE,OAAO,KAAK,UAAU,MAAQ,CAEjE,IAAW,aAAmB,CAAE,OAAO,IAAIG,GAAI,KAAK,IAAI,EAAE,QAAQ,KAAK,UAAU,QAAQ,CAAC,CAAG,CAEtF,QAAQC,EAAeC,EAAiC,CAC7D,OAAO,IAAIT,GACT,KAAK,MACL,KAAK,WACL,KAAK,UACL,KAAK,YACL,CAAE,MAAAQ,EAAO,OAAAC,CAAO,CAClB,CACF,CACF,EC5CO,IAAMC,GAAN,cAA4BC,EAAyC,CAC1E,YACEC,EACAC,EACAC,EACA,CACA,MAAMF,EAAOA,EAAO,GAAIC,EAAaC,CAAS,CAChD,CAEgB,QAAQC,EAAeC,EAA+B,CACpE,GAAM,CAAE,MAAAJ,EAAO,YAAAC,EAAa,UAAAC,CAAU,EAAI,MAAM,QAAQC,EAAOC,CAAM,EACrE,OAAO,IAAIN,GAAcE,EAAOC,EAAaC,CAAS,CACxD,CACF,ECrBO,SAASG,GAAIC,EAAWC,EAAWC,EAAQ,GAAM,CACtD,IAAMC,EAAQ,KAAK,IAAI,EAAK,KAAK,IAAIH,EAAK,MAAOC,EAAK,KAAK,EAAI,KAAK,IAAID,EAAK,KAAMC,EAAK,IAAI,CAAC,EACvFG,EAAS,KAAK,IAAI,EAAK,KAAK,IAAIJ,EAAK,OAAQC,EAAK,MAAM,EAAI,KAAK,IAAID,EAAK,IAAKC,EAAK,GAAG,CAAC,EACxFI,EAAeF,EAAQC,EAE7B,OAAOF,EACHG,GAAgBL,EAAK,KAAOC,EAAK,KAAOI,GACxCA,EAAe,KAAK,IAAIL,EAAK,KAAMC,EAAK,IAAI,CAClD,CCRO,SAASK,GAAQC,EAA4B,CAClD,IAAMC,EAAKD,EAAI,IAAKE,GAAOA,EAAG,CAAC,EACzBC,EAAKH,EAAI,IAAKE,GAAOA,EAAG,CAAC,EACzBE,EAAOH,EAAG,OAAO,CAACI,EAAKC,IAAOA,EAAID,EAAMC,EAAID,EAAM,GAAQ,EAC1DE,EAAOJ,EAAG,OAAO,CAACE,EAAKG,IAAOA,EAAIH,EAAMG,EAAIH,EAAM,GAAQ,EAC1DI,EAAOR,EAAG,OAAO,CAACS,EAAKJ,IAAOI,EAAMJ,EAAIA,EAAII,EAAM,CAAC,EACnDC,EAAOR,EAAG,OAAO,CAACO,EAAKF,IAAOE,EAAMF,EAAIA,EAAIE,EAAM,CAAC,EAEzD,OAAO,IAAIE,GAAYR,EAAMG,EAAME,EAAME,CAAI,CAC/C,CCRO,SAASE,GACdC,EACAC,EACAC,EACAC,EAAQ,GACE,CACV,IAAIC,EAAuBH,EACxB,IAAI,CAACI,EAAOC,KAAc,CAAE,MAAAD,EAAO,SAAAC,CAAS,EAAE,EAC9C,KAAK,CAACC,EAAIC,IAAOD,EAAG,MAAQC,EAAG,KAAK,EACpC,IAAKC,GAAMA,EAAE,QAAQ,EAElBC,EAAiB,CAAC,EAExB,KAAON,EAAqB,OAAS,GAAG,CACtC,IAAMO,EAAOP,EAAqB,IAAI,EACtCM,EAAK,KAAKC,CAAI,EAEd,IAAMC,EAAUR,EAEVS,EAAoB,CAAC,EAC3B,QAASC,EAAI,EAAGA,EAAIF,EAAQ,OAAQE,IAAK,CACvC,IAAMC,EAAMH,EAAQE,GAEdE,EAAUhB,EAAMW,GAChBM,EAASjB,EAAMe,GAErBF,EAAQ,KAAKK,GAAIF,EAASC,EAAQd,CAAK,CAAC,CAC1C,CAEAC,EAAuBA,EAAqB,OAC1C,CAACe,EAAGC,IAAMP,EAAQO,IAAMlB,CAC1B,CACF,CAEA,OAAOQ,CACT,CCpCO,SAASW,GAAUC,EAAgBC,EAAgC,CACxE,OAAUC,EAAK,IAAM,CACnB,GAAM,CAACC,EAAGC,EAAGC,CAAC,EAAIJ,EACZK,EAAWC,GAAK,CAAC,GAAGP,EAAE,MAAM,MAAM,EAAG,CAAC,EAAG,CAAC,EAAGG,EAAG,SAAS,EACzDK,EAAWD,GAAK,CAAC,GAAGP,EAAE,MAAM,MAAM,EAAG,CAAC,EAAG,CAAC,EAAGI,EAAG,SAAS,EACzDK,EAAWF,GAAK,CAAC,GAAGP,EAAE,MAAM,MAAM,EAAG,CAAC,EAAG,CAAC,EAAGK,EAAG,SAAS,EACzDK,EAAaC,GAAO,CAACL,EAAOE,EAAOC,CAAK,EAAG,CAAC,EAElD,OAAUG,GAAIZ,EAAGU,CAAO,CAC1B,CAAC,CACH,CCFO,SAASG,GAAYC,EAAwBC,EAAgB,GAAoB,CACtF,OAAUC,EAAK,IAAM,CACnB,GAAM,CAACC,EAAQC,CAAK,EAAIJ,EAAU,MAAM,MAAM,CAAC,EAC/C,GAAIG,IAAWC,EAAO,OAAOJ,EAC7B,IAAMK,EAAU,KAAK,IAAIF,EAASC,CAAK,EACjCE,EAAgB,KAAK,MAAMD,GAAWJ,EAAgB,GAAM,EAAE,EAC9DM,EAAcJ,EAASC,EAAQ,EAAI,EACnCI,EAAuBC,GAA0C,CACrE,IAAMC,EAAqBV,EAAU,MAAM,MAAM,EACjD,OAAAU,EAAmBH,GAAeE,EACxBE,GAAKD,EAAoB,EAAG,SAAS,CACjD,EACME,EAAsBJ,EAAoBF,CAAa,EACvDO,EAAyBR,EAAWO,EAAoB,MAAML,GAE9DO,EAAiB,CADMb,GAAiBY,EAAyBL,EAAoBK,CAAsB,EAAI,KACvEb,EAAWY,CAAmB,EACzE,OAAQG,GAAM,CAAC,CAACA,CAAC,EACjB,IAAKA,GAASC,GAAKD,EAAkB,SAAS,CAAC,EAClD,OAAUE,GAAOH,EAAgBP,CAAW,CAC9C,CAAC,CACH,CC9BO,SAASW,IAAaC,EAAmB,CAC9C,IAAMC,EAAQD,EAAW,MAAM,EAC/B,QAASE,EAAID,EAAM,OAAS,EAAGC,EAAI,EAAGA,IAAK,CACzC,IAAMC,EAAI,KAAK,MAAM,KAAK,OAAO,GAAKD,EAAI,EAAE,EACtCE,EAAIH,EAAMC,GAChBD,EAAMC,GAAKD,EAAME,GACjBF,EAAME,GAAKC,CACb,CACA,OAAOH,CACT,CCFO,SAASI,GAAQC,EAAW,CACjC,MAAO,IAAK,EAAI,KAAK,IAAI,CAACA,CAAC,EAC7B,CAEO,SAASC,IAAeD,EAAW,CACxC,OAAO,KAAK,IAAIA,GAAK,EAAIA,EAAE,CAC7B,CCJO,IAAME,GAAN,cAAmBC,EAAqB,CAC7C,YAAYC,EAAWC,EAAWC,EAAeC,EAAgBC,EAA0B,GAAO,CAChG,MAAM,CAAE,EAAAJ,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,EAAGC,CAAuB,CACxD,CACF,ECHA,IAAMC,IAAO,GACPC,IAAO,IACPC,IAAW,IAOJC,GAAN,KAA8C,CAOnD,YACEC,EACAC,EACAC,EAAe,IAAIC,GAAM,EAAG,CAAC,EAC7B,CACA,GAAM,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAIJ,EAC1B,KAAK,SAAW,IAAIK,GAAWF,EAAOC,CAAM,EAC5C,KAAK,OAASH,EACd,KAAK,WAAaF,EAA8B,IAC7CO,GAAOA,EAAG,IAAI,IAAIJ,GAAMC,EAAOC,CAAM,CAAC,EAAE,IAAIH,CAAK,CACpD,CACF,CAEA,IAAW,OAAe,CAAE,OAAO,IAAIC,GAAM,KAAK,OAAO,EAAG,KAAK,OAAO,CAAC,CAAG,CAE5E,IAAW,YAAqB,CAAE,OAAO,KAAK,SAAS,KAAO,CAE9D,IAAW,aAAsB,CAAE,OAAO,KAAK,SAAS,MAAQ,CAEhE,IAAW,WAAqB,CAAE,OAAO,KAAK,UAAY,CAE1D,IAAW,mBAA6B,CACtC,OAAO,KAAK,WAAW,IACpBI,GAAOA,EAAG,IAAI,KAAK,MAAM,EAAE,IAAI,IAAIJ,GAAM,KAAK,WAAY,KAAK,WAAW,CAAC,CAC9E,CACF,CAEO,QAAiCC,EAAeC,EAAmB,CACxE,OAAO,IAAK,KAAK,YACf,KAAK,kBACL,CAAE,MAAAD,EAAO,OAAAC,CAAO,CAClB,CACF,CAEO,QAAiCG,EAAWC,EAAc,CAC/D,OAAO,IAAK,KAAK,YACf,KAAK,kBACL,KAAK,SACL,IAAIN,GAAMK,EAAGC,CAAC,CAChB,CACF,CAEO,aAAsCF,EAAc,CACzD,OAAO,KAAK,QAAQA,EAAG,EAAGA,EAAG,CAAC,CAChC,CAaO,MACLG,EACAC,EAAkE,CAAE,EAC/D,CACL,GAAID,EAAW,CACb,IAAME,EAAMF,aAAqBG,GAC7BH,EAAU,IAAI,MAAM,EACpB,IAAII,GAAIJ,CAAS,EAErB,OAAO,KAAK,QAAQE,EAAI,EAAGA,EAAI,CAAC,EAAE,MAAM,KAAMD,CAAO,CACvD,CAEA,GAAM,CAAE,iBAAAI,EAAkB,cAAAC,CAAc,EAAI,CAAE,iBAAkB,GAAO,cAAe,GAAK,GAAGL,CAAQ,EAEtG,OAAII,EACK,KAAK,UAAU,EAGjB,KAAK,aAAaC,CAAa,CACxC,CAEQ,WAAiB,CACvB,IAAMC,EAAU,KAAK,yBAAyB,EAExC,CAACC,EAAeC,EAAgBC,CAAW,EAAIH,EAC/CI,EAAed,GAAca,EAAY,IAAIb,CAAE,EAAE,UAAU,EAC3De,GAAkBD,EAAYH,CAAa,EAAIG,EAAYF,CAAc,GAAK,EAE9EI,EAAO,KAAK,MAAMD,EAAiBxB,GAAQ,EAE3C0B,EAAWC,GAAeR,CAAO,EAEjCT,EAAI,KAAK,MAAM,KAAK,IAAI,EAAGgB,EAAS,EAAK5B,IAAO2B,CAAK,CAAC,EACtDd,EAAI,KAAK,MAAM,KAAK,IAAI,EAAGe,EAAS,EAAK3B,IAAO0B,CAAK,CAAC,EAE5D,OAAO,IAAIG,GAAKlB,EAAGC,EAAG,KAAK,IAAIc,EAAM,KAAK,WAAaf,CAAC,EAAG,KAAK,IAAIe,EAAM,KAAK,YAAcd,CAAC,CAAC,CACjG,CAEQ,aAAakB,EAAsB,CACzC,IAAMf,EAAMgB,GAAQ,KAAK,SAAS,EAClC,OAAOhB,EAAI,IAAIA,EAAI,MAAQe,EAASf,EAAI,OAASe,CAAO,CAC1D,CAEU,0BAAoC,CAC5C,MAAM,IAAI,MAAM,wDAAwD,CAC1E,CACF,EC7HO,IAAME,GAAN,cAA6BC,EAAc,CAC7B,0BAAoC,CACrD,IAAMC,EAAM,KAAK,UACjB,MAAO,CACLA,EAAI,GACJA,EAAI,GACJC,GAAe,CAACD,EAAI,GAAIA,EAAI,EAAE,CAAC,CACjC,CACF,CACF,ECTO,IAAME,GAAN,cAA8BC,EAAc,CAC1C,eAAyB,CAC9B,OAAO,KAAK,UAAU,MAAM,EAAG,EAAE,CACnC,CAEO,gBAA0B,CAC/B,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEO,iBAA2B,CAChC,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEO,SAAmB,CACxB,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEO,YAAsB,CAC3B,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEO,aAAuB,CAC5B,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEO,UAAoB,CACzB,OAAO,KAAK,UAAU,MAAM,GAAI,EAAE,CACpC,CAEmB,0BAAoC,CACrD,MAAO,CACL,KAAK,WAAW,EAChB,KAAK,YAAY,EACjB,KAAK,SAAS,CAChB,EAAE,IAAIC,EAAc,CACtB,CACF,ECjCO,IAAMC,GAAN,KAAsC,CAI3C,YAAYC,EAAeC,EAAkB,CAC3C,KAAK,OAASD,EACd,KAAK,UAAYC,CACnB,CAEA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,UAAmB,CAAE,OAAO,KAAK,SAAW,CAEhD,SAASC,EAAe,GAAc,CAC3C,MAAO,GAAG,KAAK,QAAQA,EAAe,KAAKC,GAAM,KAAK,QAAQ,KAAO,IACvE,CACF,EClBO,IAAMC,GAAN,cAAyBC,EAAI,CAUlC,YAAYC,EAAiCC,EAAe,CAC1D,MAAMD,CAAG,EACT,KAAK,OAASC,CAChB,CAZA,OAAc,wBAAwBD,EAAUE,EAAgB,CAE9D,GADAH,GAAI,iBAAiBC,EAAKE,CAAM,EAC5B,CAACC,GAAcH,EAAI,KAAK,EAC1B,MAAM,IAAI,MAAM,GAAGE,gCAAqCF,EAAI,uBAAuB,CAEvF,CASA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CACnD,ECrBO,IAAMI,GAAN,KAA6B,CAKlC,YAAYC,EAAeC,EAA6B,CACtD,GAAM,OAAOD,GAAU,SACrB,MAAM,IAAI,MAAM,oEAAoE,EAGtF,GAAI,CAAC,MAAM,QAAQC,CAAW,GAAKA,EAAY,KAAMC,GAAS,EAAEA,aAAgB,aAAa,EAC3F,MAAM,IAAI,MAAM,0FAA0F,EAG5G,KAAK,OAASF,EACd,KAAK,aAAeC,CACtB,CAEA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,aAA8B,CAAE,OAAO,KAAK,YAAc,CAE9D,QAAc,CACnB,MAAO,CACL,MAAO,KAAK,MACZ,YAAa,KAAK,YAAY,IAAKE,GAAM,MAAM,KAAKA,CAAC,CAAC,CACxD,CACF,CAEA,OAAc,SAASC,EAAmC,CACxD,IAAMH,EAAcG,EAAK,YAAY,IAAKD,GAAW,IAAI,aAAaA,CAAC,CAAC,EACxE,OAAO,IAAIJ,GAAuBK,EAAK,MAAOH,CAAW,CAC3D,CACF,EC5BO,IAAMI,GAAN,cAA2BC,EAAW,CAgB3C,YAAYC,EAAiCC,EAAeC,EAAeC,EAAoB,CAC7F,MAAMH,EAAKC,CAAK,EAChB,KAAK,OAASC,EACd,KAAK,YAAcC,CACrB,CAnBA,OAAc,0BAA0BH,EAAUI,EAAgB,CAGhE,GAFAL,GAAW,wBAAwBC,EAAKI,CAAM,EAG5C,CAACC,GAAmBL,EAAI,KAAK,GAC1B,CAACK,GAAmBL,EAAI,UAAU,EAErC,MAAM,IAAI,MAAM,GAAGI,kCAAuCJ,EAAI,eAAeA,EAAI,2CAA2C,CAEhI,CAYA,IAAW,OAAgB,CAAE,OAAO,KAAK,MAAQ,CAEjD,IAAW,YAAqB,CAAE,OAAO,KAAK,WAAa,CAC7D,ECxBO,SAASM,GAAoBC,EAAwC,CAC1E,OAAOA,EAAI,qBAAqBC,EAClC,CAEO,SAASC,GAAiCC,EAAoBC,EAAsD,CAEzH,MAAO,CAAE,GAAGD,EAAW,GADL,CAAE,UAAAC,CAAU,CACM,CACtC,CCXO,SAASC,IAAgC,CAC9C,IAAMC,EAAQ,OAAO,MACrB,GAAI,CAACA,EAAO,MAAM,IAAI,MAAM,8DAA8D,EAM1F,MAAO,CACL,OAAQ,kBACR,yBACA,MAAO,iBACP,UACA,MAAO,iBACP,oBAAqB,IAAM,SAAS,cAAc,QAAQ,EAC1D,mBAAoB,IAAM,SAAS,cAAc,KAAK,EACtD,mBAAoB,IAAM,SAAS,cAAc,OAAO,EACxD,MAAAA,EACA,SAde,IAAM,CACrB,MAAM,IAAI,MAAM,6DAA6D,CAC/E,CAaA,CACF,CCtBO,SAASC,IAAoB,CAClC,OAAO,OAAO,QAAW,UACpB,OAAO,SAAY,aACnB,QAAQ,UAAY,MACpB,QAAQ,SAAS,MAAQ,IAChC,CCFO,SAASC,GAAiBC,EAAsB,CACrD,IAAIC,EAAiB,GACrB,GAAI,CAACD,GAAME,GAAS,EAClB,GAAI,CAEFF,EAAK,GAAQ,KACf,OAASG,EAAP,CACAF,EAAkBE,EAAY,SAAS,CACzC,CAMF,MAAO,CAAE,SAHQH,EACZI,GAAqB,IAAI,QAAQ,CAACC,EAASC,IAAW,CAAEN,EAAG,SAASI,EAAU,CAACD,EAAUI,IAAYJ,EAAMG,EAAOH,CAAG,EAAIE,EAAQE,CAAM,CAAE,CAAG,CAAC,EAC9I,IAAM,CAAE,MAAM,IAAI,MAAM,qEAAqEN,GAAgB,CAAG,CAClG,CACpB,CCdO,SAASO,IAA+B,CAE7C,IAAMC,EAAS,OAAO,QAAa,OAAO,kBACpCC,EAAQ,OAAO,OAAS,OAAO,iBAE/BC,EAAQ,OAAO,OAAY,OAAO,iBAElCC,EAAsB,IAAM,CAChC,GAAIH,EAAQ,OAAO,IAAIA,EACvB,MAAM,IAAI,MAAM,4EAA4E,CAC9F,EAEMI,EAAqB,IAAM,CAC/B,GAAIH,EAAO,OAAO,IAAIA,EACtB,MAAM,IAAI,MAAM,0EAA0E,CAC5F,EAEMI,EAAqB,IAAM,CAC/B,GAAIH,EAAO,OAAO,IAAIA,EACtB,MAAM,IAAI,MAAM,0EAA0E,CAC5F,EAEMI,EAAQ,OAAO,MAGfC,EAAaC,GAAiB,EAEpC,MAAO,CACL,OAAQR,GAAU,KAAM,CAAC,EACzB,yBAA0B,OAAO,0BAA4B,KAAM,CAAC,EACpE,MAAOC,GAAS,KAAM,CAAC,EACvB,UAAW,OAAO,WAAa,KAAM,CAAC,EACtC,MAAO,OAAO,kBAAoB,KAAM,CAAC,EACzC,oBAAAE,EACA,mBAAAC,EACA,mBAAAC,EACA,MAAAC,EACA,GAAGC,CACL,CACF,CC3CO,SAASE,IAAqB,CACnC,OAAO,OAAO,QAAW,UACpB,OAAO,UAAa,aACpB,OAAO,kBAAqB,aAC5B,OAAO,mBAAsB,aAC7B,OAAO,kBAAqB,aAC5B,OAAO,WAAc,aACrB,OAAO,0BAA6B,WAC3C,CCDA,IAAIC,GAEJ,SAASC,KAAsB,CAC7B,GAAI,CAACD,GACH,MAAM,IAAI,MAAM,uEAAuE,EAEzF,OAAOA,EACT,CAEA,SAASE,GAAOC,EAAkB,CAChCH,GAAcG,CAChB,CAEA,SAASC,IAAa,CAGpB,OAAIC,GAAU,EAAUH,GAAOI,GAAiB,CAAC,EAC7CC,GAAS,EAAUL,GAAOM,GAAgB,CAAC,EACxC,IACT,CAEA,SAASC,IAAYN,EAA2B,CAK9C,GAJKH,IACHI,GAAW,EAGT,CAACJ,GACH,MAAM,IAAI,MAAM,4EAA4E,EAG9F,GAAM,CAAE,OAAAU,EAASV,GAAY,OAAQ,MAAAW,EAAQX,GAAY,KAAM,EAAIG,EACnEH,GAAY,OAASU,EACrBV,GAAY,MAAQW,EACpBX,GAAY,oBAAsBG,EAAI,sBAAwB,IAAM,IAAIO,GACxEV,GAAY,mBAAqBG,EAAI,qBAAuB,IAAM,IAAIQ,GAEtEX,GAAY,UAAYG,EAAI,WAAaH,GAAY,UACrDA,GAAY,MAAQG,EAAI,OAASH,GAAY,MAC7CA,GAAY,MAAQG,EAAI,OAASH,GAAY,MAC7CA,GAAY,SAAWG,EAAI,UAAYH,GAAY,QACrD,CAEO,IAAMG,GAAM,CACjB,OAAAF,IACA,OAAAC,GACA,WAAAE,GACA,iBAAAE,GACA,iBAAAM,GACA,gBAAAJ,GACA,YAAAC,IACA,UAAAJ,GACA,SAAAE,EACF,EAEAH,GAAW,EC3DJ,SAASS,GAAaC,EAAmB,CAC9C,MAAI,CAACC,GAAI,SAAS,GAAK,OAAOD,GAAQ,SAC7B,SAAS,eAAeA,CAAG,EAE7BA,CACT,CCJO,SAASE,GAAoBC,EAA4F,CAC9H,GAAM,CAAE,OAAAC,EAAQ,yBAAAC,CAAyB,EAAIC,GAAI,OAAO,EAExD,GAAIH,aAAqBE,EACvB,OAAOF,EAGT,IAAMI,EAASC,GAAaL,CAAS,EAErC,GAAI,EAAEI,aAAkBH,GACtB,MAAM,IAAI,MAAM,gEAAgE,EAGlF,IAAMK,EAAMF,EAAO,WAAW,IAAI,EAClC,GAAI,CAACE,EACH,MAAM,IAAI,MAAM,8CAA8C,EAGhE,OAAOA,CACT,CChBO,IAAKC,QAEVA,EAAA,SAAW,WAEXA,EAAA,UAAY,YAEZA,EAAA,YAAc,cAEdA,EAAA,aAAe,eARLA,QAAA,IAoBCC,GAAN,KAA4D,CAajE,YAAYC,EAAiC,CAAC,EAAG,CAC/C,GAAM,CACJ,eAAAC,EAAgB,gBAAAC,EAAiB,UAAAC,EAAW,SAAAC,EAAU,UAAAC,EAAW,QAAAC,CACnE,EAAIN,EACJ,KAAK,eAAiBC,GAAkB,WACxC,KAAK,gBAAkBC,GAAmB,qBAC1C,KAAK,UAAYC,GAAa,yBAC9B,KAAK,SAAWC,GAAY,GAC5B,KAAK,UAAYC,GAAa,UAC9B,KAAK,QAAUC,GAAW,CAC5B,CACF,EAEaC,GAAN,KAAoB,CAOzB,YACEC,EACAC,EACAT,EAAiC,CAAC,EAClC,CAEA,KAAK,KAAO,OAAOQ,GAAS,SACxB,CAACA,CAAI,EACJA,aAAgBD,GAAgBC,EAAK,KAAOA,EACjD,KAAK,OAASC,EACd,KAAK,QAAU,IAAIV,GAAqBC,CAAO,CACjD,CAEA,aAAaU,EAAuC,CAClD,GAAM,CAAE,QAAAJ,CAAQ,EAAI,KAAK,QACzB,OAAO,KAAK,KAAK,IAAKK,GAAMD,EAAI,YAAYC,CAAC,EAAE,KAAK,EAAE,OAAO,CAACC,EAAIC,IAAQD,EAAKC,EAAKA,EAAKD,EAAK,CAAC,EAAK,EAAIN,CAC1G,CAEA,eAAwB,CACtB,GAAM,CAAE,SAAAF,EAAU,QAAAE,CAAQ,EAAI,KAAK,QACnC,OAAO,KAAK,KAAK,OAASF,EAAY,EAAIE,CAC5C,CAEA,aAAaI,EAA+BI,EAAkC,CAC5E,GAAM,CAAE,eAAAb,CAAe,EAAI,KAAK,QAC1Bc,EAAcd,IAAmB,gBAA+BA,IAAmB,YACnFe,EAAaf,IAAmB,eAA8BA,IAAmB,eAEjFgB,EAAiB,KAAK,aAAaP,CAAG,EACtCQ,EAAkB,KAAK,cAAc,EACrCC,EAAKJ,EAAc,KAAK,OAAO,EAAIE,EAAiB,KAAK,OAAO,EAChEG,EAAIJ,EAAa,KAAK,OAAO,EAAIE,EAAkB,KAAK,OAAO,EAGrE,GAAIJ,EAAY,CACd,GAAM,CAAE,MAAAO,EAAO,OAAAC,CAAO,EAAIR,EACpBS,EAAO,KAAK,IAAI,KAAK,IAAIJ,EAAGE,EAAQJ,CAAc,EAAG,CAAC,EACtDO,EAAO,KAAK,IAAI,KAAK,IAAIJ,EAAGE,EAASJ,CAAe,EAAG,CAAC,EAC9D,MAAO,CAAE,EAAGK,EAAM,EAAGC,CAAK,CAC5B,CACA,MAAO,CAAE,EAAAL,EAAG,EAAAC,CAAE,CAChB,CAEA,KAAKK,EAAkE,CACrE,IAAMC,EAASC,GAAaF,CAAS,EAC/Bf,EAAMkB,GAAoBF,CAAM,EAEhC,CACJ,gBAAAxB,EAAiB,UAAAC,EAAW,SAAAC,EAAU,UAAAC,EAAW,QAAAC,CACnD,EAAI,KAAK,QAETI,EAAI,KAAO,GAAGN,OAAcC,IAC5B,IAAMwB,EAAe,KAAK,aAAanB,CAAG,EACpCoB,EAAa,KAAK,cAAc,EAEtCpB,EAAI,UAAYR,EAChB,IAAM6B,EAAY,KAAK,aAAarB,EAAKgB,CAAM,EAC/ChB,EAAI,SAASqB,EAAU,EAAGA,EAAU,EAAGF,EAAcC,CAAU,EAE/DpB,EAAI,UAAYP,EAChB,KAAK,KAAK,QAAQ,CAAC6B,EAAUC,IAAM,CACjC,IAAMd,EAAIb,EAAUyB,EAAU,EACxBX,EAAId,EAAUyB,EAAU,GAAME,EAAI,GAAK7B,EAC7CM,EAAI,SAASsB,EAAUb,EAAGC,CAAC,CAC7B,CAAC,CACH,CACF,ECjHO,IAAMc,GAAN,KAAqB,CAS1B,YAAYC,EAA2B,CAAC,EAAG,CACzC,GAAM,CACJ,SAAAC,EAAU,UAAAC,EAAW,MAAAC,EAAO,iBAAAC,CAC9B,EAAIJ,EACJ,KAAK,SAAWC,GAAY,qBAC5B,KAAK,UAAYC,GAAa,EAC9B,KAAK,MAAQC,EAEb,IAAME,EAA0B,CAC9B,6BACA,gBAAiB,KAAK,QACxB,EACA,KAAK,iBAAmB,IAAIC,GAAqB,CAAE,GAAGD,EAAyB,GAAGD,CAAiB,CAAC,CACtG,CACF,EAEaG,GAAN,KAAc,CAKnB,YACEC,EACAR,EAA2B,CAAC,EAC5B,CACA,KAAK,IAAM,IAAIS,GAAID,CAAG,EACtB,KAAK,QAAU,IAAIT,GAAeC,CAAO,CAC3C,CAEA,KAAKU,EAAkE,CACrE,IAAMC,EAAMC,GAAoBF,CAAS,EAEnC,CAAE,SAAAT,EAAU,UAAAC,CAAU,EAAI,KAAK,QAE/B,CACJ,EAAAW,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CACf,EAAI,KAAK,IACTL,EAAI,YAAcV,EAClBU,EAAI,UAAYT,EAChBS,EAAI,WAAWE,EAAGC,EAAGC,EAAOC,CAAM,EAElC,GAAM,CAAE,MAAAb,CAAM,EAAI,KAAK,QACnBA,GACF,IAAIc,GAAc,CAACd,CAAK,EAAG,CAAE,EAAGU,EAAKX,EAAY,EAAI,EAAAY,CAAE,EAAG,KAAK,QAAQ,gBAAgB,EAAE,KAAKJ,CAAS,CAE3G,CACF,EC3DO,SAASQ,IACdC,EACAC,EACA,EACwB,MAAM,QAAQA,CAAU,EAAIA,EAAa,CAACA,CAAU,GAE5D,QAASC,GAAQ,CAE/B,IAAMC,EAAQD,aAAeE,GACzBF,EAAI,MACHG,GAAoBH,CAAG,EAAIA,EAAI,UAAU,MAAQ,OAGhDI,EAAMJ,aAAeE,GACvBF,EAAI,IACHG,GAAoBH,CAAG,EAAIA,EAAI,UAAU,IAAM,IAAIK,GAAIL,CAAG,EAEzDM,EAAQL,EAAQ,GAAGM,GAAMN,CAAK,IAAM,OAC1C,IAAIO,GAAQJ,EAAK,CAAE,MAAAE,CAAM,CAAC,EAAE,KAAKR,CAAS,CAC5C,CAAC,CACH,CC1BO,SAASW,GAAcC,EAAsD,CAClF,GAAM,CAAE,MAAAC,EAAO,MAAAC,CAAM,EAAIC,GAAI,OAAO,EAEpC,OAAQH,aAAiBC,GAASD,EAAM,UAClCA,aAAiBE,GAASF,EAAM,YAAc,CACtD,CCJO,SAASI,GAAiBC,EAAgE,CAE/F,OAAO,IAAI,QAAQ,CAACC,EAASC,IAAW,EAClCF,aAAiBG,GAAI,OAAO,EAAE,QAAUC,GAAcJ,CAAK,IAAGC,EAAQ,IAAI,EAE9E,SAASI,EAAQC,EAAU,CACrB,CAACA,EAAE,gBAEPA,EAAE,cAAc,oBAAoB,OAAQC,CAAM,EAClDD,EAAE,cAAc,oBAAoB,QAASD,CAAO,EACpDH,EAAOI,CAAC,EACV,CAEA,SAASC,EAAOD,EAAU,CACpB,CAACA,EAAE,gBACPA,EAAE,cAAc,oBAAoB,OAAQC,CAAM,EAClDD,EAAE,cAAc,oBAAoB,QAASD,CAAO,EACpDJ,EAAQK,CAAC,EACX,CAEAN,EAAM,iBAAiB,OAAQO,CAAM,EACrCP,EAAM,iBAAiB,QAASK,CAAO,CACzC,CAAC,CACH,CCxBO,SAASG,GAAcC,EAAsC,CAClE,OAAO,IAAI,QAAQ,CAACC,EAASC,IAAW,CAChCF,aAAe,MAAOE,EAAO,IAAI,MAAM,kDAAkD,CAAC,EAChG,IAAMC,EAAS,IAAI,WACnBA,EAAO,OAAS,IAAM,CAChB,OAAOA,EAAO,QAAW,UAAUD,EAAO,IAAI,MAAM,kEAAkE,CAAC,EAC3H,IAAME,EAAMC,GAAI,OAAO,EAAE,mBAAmB,EAC5CD,EAAI,OAAS,IAAMH,EAAQG,CAAG,EAC9BA,EAAI,QAAUF,EACdE,EAAI,IAAMD,EAAO,MACnB,EACAA,EAAO,QAAUD,EACjBC,EAAO,cAAcH,CAAG,CAC1B,CAAC,CACH,CCbO,SAASM,GAAmBC,EAA0F,CAC3H,GAAM,CAAE,MAAAC,EAAO,MAAAC,CAAM,EAAIC,GAAI,OAAO,EAEpC,OAAIH,aAAiBC,EACZ,IAAIG,GAAWJ,EAAM,aAAcA,EAAM,aAAa,EAE3DA,aAAiBE,EACZ,IAAIE,GAAWJ,EAAM,WAAYA,EAAM,WAAW,EAEpD,IAAII,GAAWJ,EAAM,MAAOA,EAAM,MAAM,CACjD,CCPO,SAASK,GAAa,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAmC,CAC9E,GAAM,CAAE,oBAAAC,CAAoB,EAAIC,GAAI,OAAO,EACrCC,EAASF,EAAoB,EACnC,OAAAE,EAAO,MAAQJ,EACfI,EAAO,OAASH,EACTG,CACT,CAEO,SAASC,GAAsBC,EAAwDC,EAAuC,CACnI,GAAM,CAAE,UAAAC,CAAU,EAAIL,GAAI,OAAO,EAEjC,GAAI,EAAEG,aAAiBE,IAAc,CAACC,GAAcH,CAAK,EACvD,MAAM,IAAI,MAAM,4DAA4D,EAG9E,GAAM,CAAE,MAAAN,EAAO,OAAAC,CAAO,EAAIM,GAAQG,GAAmBJ,CAAK,EACpDF,EAASL,GAAa,CAAE,MAAAC,EAAO,OAAAC,CAAO,CAAC,EAE7C,OAAIK,aAAiBE,EACnBG,GAAoBP,CAAM,EAAE,aAAaE,EAAO,EAAG,CAAC,EAEpDK,GAAoBP,CAAM,EAAE,UAAUE,EAAO,EAAG,EAAGN,EAAOC,CAAM,EAE3DG,CACT,CCzBA,eAAsBQ,GACpBC,EACAC,EAC4B,CAC5B,IAAMC,EAAeD,GAAUE,GAAI,OAAO,EAAE,oBAAoB,EAE1D,CAACC,EAAQC,EAAOC,CAAW,EAAIN,EAAU,MAAM,MAAMO,GAAWP,CAAS,EAAI,EAAI,CAAC,EAClFQ,EAAiBC,EAAK,IAAMT,EAAU,KAAKI,EAAQC,EAAOC,CAAW,EAAE,MAAM,CAAC,EACpF,aAASI,GAAW,SAASF,EAAaN,CAAY,EAEtDM,EAAY,QAAQ,EAEbN,CACT,CChBO,SAASS,GAAeC,EAAY,CACzC,GAAM,CAAE,MAAAC,EAAO,OAAAC,EAAQ,MAAAC,CAAM,EAAIC,GAAI,OAAO,EAE5C,OAAOJ,aAAiBC,GACnBD,aAAiBE,GACjBF,aAAiBG,CACxB,CCHO,SAASE,GAAcC,EAA6CC,EAAmBC,EAAc,GAAO,CACjH,GAAM,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAIC,GAAI,OAAO,EAErC,GAAI,EAAEL,aAAiBG,GAASH,aAAiBI,GAC/C,MAAM,IAAI,MAAM,0EAA0E,EAG5F,GAAIH,GAAa,EAAG,OAAOK,GAAa,CAAE,MAAO,EAAG,OAAQ,CAAE,CAAC,EAC/D,IAAMC,EAAOC,GAAmBR,CAAK,EAC/BS,EAAQR,EAAY,KAAK,IAAIM,EAAK,OAAQA,EAAK,KAAK,EACpDG,EAAQD,EAAQF,EAAK,MACrBI,EAASF,EAAQF,EAAK,OAEtBK,EAAeN,GAAa,CAAE,MAAOL,EAAW,OAAQA,CAAU,CAAC,EACnEY,EAAcb,aAAiBI,EAASJ,EAAQc,GAAsBd,CAAK,EAE3Ee,EAAS,KAAK,IAAIL,EAAQC,CAAM,EAAI,EACpCK,EAAKd,GAAeQ,EAAQC,EAASI,EAAS,EAC9CE,EAAKf,GAAeS,EAASD,EAAQK,EAAS,EACpD,OAAIF,EAAY,MAAQ,GAAKA,EAAY,OAAS,GAAGK,GAAoBN,CAAY,EAAE,UAAUC,EAAaG,EAAIC,EAAIP,EAAOC,CAAM,EAE5HC,CACT,CCjBO,IAAMO,GAAN,KAAe,CAapB,YAAYC,EAAkCC,EAAoB,GAAO,CAZzE,KAAQ,cAAkD,CAAC,EAE3D,KAAQ,UAAiC,CAAC,EAI1C,KAAQ,mBAAqB,GAE7B,KAAQ,iBAA+B,CAAC,EAExC,KAAQ,WAAa,EAGnB,GAAI,CAAC,MAAM,QAAQD,CAAM,EACvB,MAAM,IAAI,MAAM,4HAA4HA,GAAQ,EAGtJ,KAAK,mBAAqBC,EAC1B,KAAK,WAAaD,EAAO,OAEzBA,EAAO,QAAQ,CAACE,EAAOC,IAAQ,CAC7B,GAAIC,GAAWF,CAAK,EAAG,CACrB,KAAK,cAAcC,GAAOD,EAC1B,KAAK,iBAAiBC,GAAOD,EAAM,MACnC,MACF,CAEA,GAAIG,GAAWH,CAAK,EAAG,CACrB,IAAMI,EAAaJ,EAAc,MAAM,GACvC,GAAII,IAAc,EAChB,MAAM,IAAI,MAAM,yCAAyCA,4CAAoD,EAG/G,KAAK,cAAcH,GAAOD,EAC1B,KAAK,iBAAiBC,GAAQD,EAAc,MAAM,MAAM,CAAC,EACzD,MACF,CAGA,IAAMK,EAAUL,aAAyBM,GAAI,OAAO,EAAE,OAASN,EAAQO,GAAsBP,CAAK,EAClG,KAAK,UAAUC,GAAOI,EACtB,KAAK,iBAAiBJ,GAAO,CAACI,EAAO,OAAQA,EAAO,MAAO,CAAC,CAC9D,CAAC,CACH,CAEA,IAAW,cAAiD,CAC1D,OAAO,KAAK,aACd,CAEA,IAAW,UAAgC,CACzC,OAAO,KAAK,SACd,CAEA,IAAW,cAAwB,CACjC,OAAO,KAAK,UAAY,GAAK,KAAK,kBACpC,CAEA,IAAW,WAAoB,CAC7B,OAAO,KAAK,UACd,CAEA,IAAW,iBAA8B,CACvC,OAAO,KAAK,gBACd,CAEA,IAAW,WAAgC,CACzC,OAAO,KAAK,UACd,CAEA,IAAW,yBAAwC,CACjD,OAAOG,GAAM,KAAK,UAAW,EAAG,CAAC,EAAE,IACjC,CAACC,EAAGC,IAAa,KAAK,2BAA2BA,CAAQ,CAC3D,CACF,CAEO,SAASA,EAAiE,CAC/E,OAAO,KAAK,SAASA,IAAa,KAAK,aAAaA,EACtD,CAEO,mBAAmBA,EAA4B,CACpD,OAAO,KAAK,iBAAiBA,EAC/B,CAEO,eAAeA,EAA0B,CAC9C,OAAO,KAAK,iBAAiBA,GAAU,EACzC,CAEO,cAAcA,EAA0B,CAC7C,OAAO,KAAK,iBAAiBA,GAAU,EACzC,CAEO,2BAA2BA,EAA8B,CAC9D,GAAI,OAAO,KAAK,WAAc,SAC5B,MAAM,IAAI,MAAM,uFAAuF,EAGzG,IAAMC,EAAQ,KAAK,cAAcD,CAAQ,EACnCE,EAAS,KAAK,eAAeF,CAAQ,EAC3C,OAAOG,GAA0B,CAAE,MAAAF,EAAO,OAAAC,CAAO,EAAG,KAAK,SAAS,CACpE,CAWO,cAAcE,EAAmBC,EAAiB,GAAmB,CAC1E,YAAK,WAAaD,EAERE,EAAK,IAAM,CACnB,IAAMC,EAAeT,GAAM,KAAK,UAAW,EAAG,CAAC,EAAE,IAAKE,GAAa,CACjE,IAAMV,EAAQ,KAAK,SAASU,CAAQ,EAEpC,GAAIV,aAAoBkB,GAAQ,CAC9B,IAAIC,EAAYhB,GAAWH,CAAK,EAAIA,EAAWoB,GAAWpB,CAAK,EAC/D,OAAAmB,EAAYE,GAAYF,EAA0BJ,CAAc,GAE5DI,EAAU,MAAM,KAAOL,GAAaK,EAAU,MAAM,KAAOL,KAC7DK,EAAeG,GAAS,eAAeH,EAA0B,CAACL,EAAWA,CAAS,EAAG,GAAO,EAAK,GAGhGK,EAAU,KAAKL,EAAWA,EAAW,CAAC,CAC/C,CAEA,GAAId,aAAiBM,GAAI,OAAO,EAAE,OAChC,OAAUiB,GAAW,WAAWC,GAAcxB,EAAOc,EAAWC,CAAc,CAAC,EAGjF,MAAM,IAAI,MAAM,+BAA+BL,8FAAqGV,GAAO,CAC7J,CAAC,EAKD,OAHuByB,GAAMR,EAAa,IAAKS,GAASC,GAAKD,EAAG,SAAS,CAAC,CAAC,EAAE,KAAK,KAAK,UAAWZ,EAAWA,EAAW,CAAC,CAI3H,CAAC,CACH,CACF,EC1IA,eAAsBc,GAAWC,EAAsC,CACrE,GAAIA,aAAkBC,GAAU,OAAOD,EACvC,IAAME,EAAgB,MAAM,QAAQF,CAAM,EAAIA,EAAS,CAACA,CAAM,EAC9D,GAAI,CAACE,EAAc,OAAQ,MAAM,IAAI,MAAM,0CAA0C,EACrF,IAAMC,EAAcC,GAAiB,MAAM,QAAQJ,CAAM,EAAI,mBAAmBI,KAAS,GACnFC,EAAaH,EAAc,IAAII,EAAY,EACjD,OAAAD,EAAW,QAAQ,CAACE,EAAOC,IAAM,CAC/B,GAAI,CAACC,GAAeF,CAAK,GAAK,CAACG,GAAWH,CAAK,GAAK,CAACI,GAAWJ,CAAK,EACnE,MAAI,OAAOL,EAAcM,IAAO,SAAgB,IAAI,MAAM,eAAeL,EAAWK,CAAC,qEAAqEN,EAAcM,IAAI,EACtK,IAAI,MAAM,eAAeL,EAAWK,CAAC,8HAA8H,EAE3K,GAAIG,GAAWJ,CAAK,EAAG,CAErB,IAAMK,EAAYL,EAAM,MAAM,GAC9B,GAAIK,IAAc,EAAG,MAAM,IAAI,MAAM,eAAeT,EAAWK,CAAC,gCAAgCI,4CAAoD,CACtJ,CACF,CAAC,EAED,MAAM,QAAQ,IAAIP,EAAW,IAAKE,GAAUE,GAAeF,CAAK,GAAKM,GAAiBN,CAAK,CAAC,CAAC,EACtF,IAAIN,GAASI,EAAY,MAAM,QAAQL,CAAM,CAAC,CACvD,CClBA,eAAsBc,GAAaC,EAAkBC,EAAuE,CAC1H,GAAM,CAAE,OAAAC,CAAO,EAAIC,GAAI,OAAO,EAC1BC,EAASJ,EACb,GAAI,EAAEA,aAAiBE,GAAS,CAC9B,IAAMG,EAAW,MAAMC,GAAWN,CAAK,EACvC,GAAIK,EAAS,UAAY,EAAG,MAAM,IAAI,MAAM,4CAA4C,EACxF,IAAME,EAAiBF,EAAS,SAAS,CAAC,EAC1CD,EAASG,aAA0BL,EAASK,EAAiB,MAAMC,GAAoBD,CAAc,CACvG,CACA,IAAME,EAAMC,GAAoBN,CAAM,EAItC,OAHcH,EACX,IAAKU,GAASA,aAAeC,GAAgBD,EAAI,QAAQP,EAAO,MAAOA,EAAO,MAAM,EAAE,IAAI,MAAM,EAAIO,CAAI,EACxG,IAAKE,GAAQA,EAAI,mBAAmBT,EAAO,MAAOA,EAAO,MAAM,CAAC,EACtD,IAAI,CAAC,CAAE,EAAAU,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,IAAM,CAC5C,IAAMC,EAAUC,GAAa,CAAE,MAAAH,EAAO,OAAAC,CAAO,CAAC,EAC9C,OAAID,EAAQ,GAAKC,EAAS,GAAGP,GAAoBQ,CAAO,EAAE,aAAaT,EAAI,aAAaK,EAAGC,EAAGC,EAAOC,CAAM,EAAG,EAAG,CAAC,EAC3GC,CACT,CAAC,CACH,CClBA,eAAsBE,GAAmBC,EAAwCC,EAAiE,CAChJ,GAAI,CAACC,GAAWF,CAAW,GAAK,CAACG,GAAWH,CAAW,EACrD,MAAM,IAAI,MAAM,2DAA2D,EAG7E,GAAIG,GAAWH,CAAW,GAAKA,EAAY,MAAM,GAAK,EACpD,MAAM,IAAI,MAAM,kDAAkD,EAGpE,OAAUI,EAAK,IAAM,CACnB,GAAM,CAACC,EAAWC,EAAUC,CAAW,EAAIP,EAAY,MAAM,MAAMG,GAAWH,CAAW,EAAI,EAAI,CAAC,EAMlG,OALcC,EAAW,IAAKO,GAASA,aAAeC,GAAgBD,EAAI,QAAQF,EAAUD,CAAS,EAAE,IAAMG,CAAI,EAC9G,IAAKE,GAAQA,EAAI,mBAAmBJ,EAAUD,CAAS,CAAC,EAExD,OAAQK,GAAQA,EAAI,MAAQ,GAAKA,EAAI,OAAS,CAAC,EAC/C,IAAI,CAAC,CAAE,EAAAC,EAAG,EAAAC,EAAG,MAAAC,EAAO,OAAAC,CAAO,IAASC,GAAQf,EAAY,KAAKK,EAAWC,EAAUC,CAAW,EAAG,CAACK,EAAGD,EAAG,CAAC,EAAG,CAACG,EAAQD,EAAON,CAAW,CAAC,CAAC,CAE7I,CAAC,CACH,CChCA,eAAsBS,GACpBC,EAEAC,EACmB,CACnB,GAAM,CAAE,MAAAC,CAAM,EAAIC,GAAI,OAAO,EACvBC,EAAM,MAAMF,EAAMF,EAAKC,CAAI,EACjC,GAAI,EAAEG,EAAI,OAAS,KACjB,MAAM,IAAI,MAAM,qBAAqBA,EAAI,WAAWA,EAAI,yBAAyBA,EAAI,KAAK,EAE5F,OAAOA,CACT,CCVA,eAAsBC,IAAWC,EAAwC,CACvE,IAAMC,EAAM,MAAMC,GAAaF,CAAG,EAC5BG,EAAO,MAAOF,EAAK,KAAK,EAE9B,GAAI,CAACE,EAAK,KAAK,WAAW,QAAQ,EAChC,MAAM,IAAI,MAAM,wEAAwEA,EAAK,kBAAkBF,EAAI,KAAK,EAE1H,OAAOG,GAAcD,CAAI,CAC3B,CCTA,eAAsBE,GAAaC,EAAyB,CAC1D,OAAQ,MAAMC,GAAaD,CAAG,GAAG,KAAK,CACxC,CCFA,eAAsBE,IAAgBC,EAAoC,CACxE,OAAO,IAAI,aAAa,MAAO,MAAMC,GAAaD,CAAG,GAAG,YAAY,CAAC,CACvE,CCFO,SAASE,GAAcC,EAAsC,CAClE,OAAO,IAAI,QAAQ,CAACC,EAASC,IAAW,CAChCF,aAAe,MAAOE,EAAO,IAAI,MAAM,kDAAkD,CAAC,EAEhG,IAAMC,EAAQC,GAAI,OAAO,EAAE,mBAAmB,EAC9CD,EAAM,UAAY,IAAMF,EAAQE,CAAK,EACrCA,EAAM,QAAUD,EAChBC,EAAM,YAAc,GACpBA,EAAM,MAAQ,GACdA,EAAM,IAAM,IAAI,gBAAgBH,CAAG,EACnCG,EAAM,KAAK,CACb,CAAC,CACH,CCXA,eAAsBE,IAAWC,EAAwC,CACvE,IAAMC,EAAM,MAAMC,GAAaF,CAAG,EAC5BG,EAAO,MAAOF,EAAK,KAAK,EAE9B,GAAI,CAACE,EAAK,KAAK,WAAW,QAAQ,EAChC,MAAM,IAAI,MAAM,wEAAwEA,EAAK,kBAAkBF,EAAI,KAAK,EAE1H,OAAOG,GAAcD,CAAI,CAC3B,CCXO,SAASE,GAAaC,EAAyBC,EAA0B,CAC9E,IAAMC,EAA0B,GAAGD,0BAEnC,GAAI,CAACD,EACH,MAAO,CACL,aAAc,GACd,YAAaE,CACf,EAGF,GAAIF,IAAQ,IACV,MAAO,CACL,aAAc,IACd,YAAa,IAAIE,GACnB,EAGF,IAAMC,EAAWH,EAAI,WAAW,SAAS,EAAI,UAAYA,EAAI,WAAW,UAAU,EAAI,WAAa,GACnGA,EAAMA,EAAI,QAAQG,EAAU,EAAE,EAE9B,IAAMC,EAAQJ,EAAI,MAAM,GAAG,EAAE,OAAQK,GAAMA,CAAC,EAEtCC,EAAeN,EAAI,SAAS,OAAO,EACrCI,EAAMA,EAAM,OAAS,GACrBF,EAEAK,EAAeJ,GAAYH,EAAI,SAAS,OAAO,EAAII,EAAM,MAAM,EAAGA,EAAM,OAAS,CAAC,EAAIA,GAAO,KAAK,GAAG,EACzG,OAAAG,EAAeP,EAAI,WAAW,GAAG,EAAI,IAAIO,IAAiBA,EAEnD,CACL,aAAAA,EACA,YAAaA,IAAiB,IAAM,IAAID,IAAiB,GAAGC,KAAgBD,GAC9E,CACF,CC5BA,eAAsBE,GACpBC,EACAC,EAC4B,CAC5B,GAAM,CAAE,YAAAC,EAAa,aAAAC,CAAa,EAAIC,GAAaJ,EAAKC,CAAgB,EAElEI,EAAW,MAAMC,GAAuCJ,CAAW,EAEzE,OAAUK,GAAM,YAAYF,EAAUF,CAAY,CACpD,CCXO,SAASK,IAAgBC,EAAoBC,EAAwBC,EAAqB,GAAO,CACtG,GAAM,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAIF,EACtBG,GAAmBJ,CAAS,EAC5BA,EACJ,OAAAD,EAAM,MAAQG,EACdH,EAAM,OAASI,EACR,CAAE,MAAAD,EAAO,OAAAC,CAAO,CACzB,CCHO,IAAeE,GAAf,KAAyC,CAC9C,YAAYC,EAAc,CAI1B,KAAU,QAAkC,OAE5C,KAAU,eAAiC,CAAC,EAL1C,KAAK,MAAQA,CACf,CAQA,IAAW,QAAiC,CAAE,OAAO,KAAK,OAAS,CAEnE,IAAW,eAAgC,CAAE,OAAO,KAAK,cAAgB,CAEzE,IAAW,UAAoB,CAAE,MAAO,CAAC,CAAC,KAAK,MAAQ,CAEhD,iBAAiBC,EAA8B,CACpD,GAAM,CAAE,IAAAC,EAAK,QAAAC,CAAQ,EAAI,KAAK,qBAAqBF,CAAS,EAC5D,OAAOC,EAAIC,EACb,CAEO,sBAAsBF,EAAmBG,EAAmB,CACjE,GAAM,CAAE,IAAAF,EAAK,QAAAC,CAAQ,EAAI,KAAK,qBAAqBF,CAAS,EAC5DC,EAAIC,GAAS,QAAQ,EACrBD,EAAIC,GAAWC,CACjB,CAEO,cAAe,CACpB,OAAO,KAAK,eAAe,IAAI,CAAC,CAAE,UAAAH,CAAU,KAAO,CACjD,KAAMA,EACN,OAAQ,KAAK,iBAAiBA,CAAS,CACzC,EAAE,CACJ,CAEO,oBAAqB,CAC1B,OAAO,KAAK,aAAa,EAAE,OAAQI,GAAUA,EAAM,kBAAqBC,EAAQ,CAClF,CAEO,iBAAkB,CACvB,OAAO,KAAK,aAAa,EAAE,OAAQD,GAAU,EAAEA,EAAM,kBAAqBC,GAAS,CACrF,CAEO,UAAW,CAChB,KAAK,gBAAgB,EAAE,QAAQ,CAAC,CAAE,KAAAC,EAAM,OAAAH,CAAO,IAAM,CACnD,KAAK,sBAAsBG,EAAMH,EAAO,SAAS,CAAC,CACpD,CAAC,CACH,CAEO,QAAS,CACd,KAAK,mBAAmB,EAAE,QAAQ,CAAC,CAAE,KAAAG,EAAM,OAAQC,CAAS,IAAM,CAChE,IAAMJ,EAAYK,GAAOD,EAAS,SAAS,CAAC,EAC5CA,EAAS,QAAQ,EACjB,KAAK,sBAAsBD,EAAMH,CAAM,CACzC,CAAC,CACH,CAEO,QAAQM,EAAmB,GAAM,CACtC,KAAK,aAAa,EAAE,QAASL,GAAU,CACrC,GAAIK,GAAoBL,EAAM,OAAO,WACnC,MAAM,IAAI,MAAM,mDAAmDA,EAAM,MAAM,EAEjFA,EAAM,OAAO,QAAQ,CACvB,CAAC,EACD,KAAK,QAAU,MACjB,CAEO,iBAAgC,CACrC,OAAO,IAAI,aACT,KAAK,aAAa,EACf,IAAI,CAAC,CAAE,OAAAD,CAAO,IAAM,MAAM,KAAKA,EAAO,SAAS,CAAC,CAAa,EAC7D,OAAO,CAACO,EAAMC,IAAQD,EAAK,OAAOC,CAAG,CAAC,CAC3C,CACF,CAEA,MAAa,KAAKC,EAAgE,CAChF,GAAIA,aAAwB,aAAc,CACxC,KAAK,eAAeA,CAAY,EAChC,MACF,CACA,MAAM,KAAK,YAAYA,CAAY,CACrC,CAEA,MAAa,YAAYC,EAAyB,CAChD,GAAIA,GAAO,OAAOA,GAAQ,SACxB,MAAM,IAAI,MAAM,GAAG,KAAK,wCAAwC,EAElE,IAAMC,EAAY,MAAMC,GAAcF,EAAK,KAAK,oBAAoB,CAAC,EACrE,KAAK,kBAAkBC,CAAS,CAClC,CAEA,MAAa,aAAaE,EAA8B,CACtD,GAAIA,GAAY,OAAOA,GAAa,SAClC,MAAM,IAAI,MAAM,GAAG,KAAK,+CAA+C,EAEzE,GAAM,CAAE,SAAAC,CAAS,EAAIC,GAAI,OAAO,EAC1B,CAAE,YAAAC,EAAa,aAAAC,CAAa,EAAIC,GAAaL,EAAU,KAAK,oBAAoB,CAAC,EACjFM,EAAwBC,GAAwB,QAAQ,IAAIA,EAAU,IAAKC,GAAOP,EAASO,CAAE,EAAE,KAAMC,GAAQA,EAAI,MAAM,CAAC,CAAC,EACzHC,EAAiBC,GAAM,qBAAqBL,CAAoB,EAChEM,EAAW,KAAK,OAAO,MAAMX,EAASE,CAAW,GAAG,SAAS,CAAC,EAC9DL,EAAY,MAAMY,EAAYE,EAAUR,CAAY,EAC1D,KAAK,kBAAkBN,CAAS,CAClC,CAEO,kBAAkBA,EAA8B,CACrD,GAAM,CAAE,cAAAe,EAAe,OAAAC,CAAO,EAAI,KAAK,2BAA2BhB,CAAS,EAC3E,KAAK,eAAiBe,EACtB,KAAK,QAAUC,CACjB,CAEO,eAAeC,EAAuB,CAC3C,GAAM,CAAE,cAAAF,EAAe,OAAAC,CAAO,EAAI,KAAK,cAAcC,CAAO,EAC5D,KAAK,eAAiBF,EACtB,KAAK,QAAUC,CACjB,CAEQ,qBAAqB9B,EAAmB,CAC9C,GAAI,CAAC,KAAK,OACR,MAAM,IAAI,MAAM,mDAAmD,EAGrE,IAAMgC,EAAShC,EAAU,MAAM,GAAG,EAAE,OAAO,CAACiC,EAAoD/B,IAAY,CAE1G,GAAI,CAAC+B,EAAI,QAAQ,eAAe/B,CAAO,EACrC,MAAM,IAAI,MAAM,wDAAwDA,eAAqBF,GAAW,EAE1G,MAAO,CAAE,IAAKiC,EAAI,QAAS,QAAA/B,EAAS,QAAS+B,EAAI,QAAQ/B,EAAS,CACpE,EAAG,CAAE,QAAS,KAAK,MAAO,CAAC,EAErB,CAAE,IAAAD,EAAK,QAAAC,CAAQ,EAAI8B,EACzB,GAAI,CAAC/B,GAAO,CAACC,GAAW,EAAED,EAAIC,aAAuBgC,IACnD,MAAM,IAAI,MAAM,8DAA8DlC,GAAW,EAG3F,MAAO,CAAE,IAAAC,EAAK,QAAAC,CAAQ,CACxB,CASF,ECnJO,SAASiC,GACdC,EACAC,EACAC,EACa,CACb,OAAUC,EAAK,IAAM,CACnB,IAAIC,EAASC,GAAgBL,EAAGC,EAAO,iBAAkBA,EAAO,iBAAkBC,EAAQ,MAAM,EAChG,OAAAE,EAASE,EAAIF,EAAKH,EAAO,IAAI,EACtBG,CACT,CAAC,CACH,CCRO,SAASG,GACdC,EACAC,EACAC,EAAe,GACF,CACb,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAUC,GACdH,EACOI,EACAC,GAAOP,EAAIC,EAAiB,MAAqB,QAAS,CAAC,EAAG,CAAC,EAAG,MAAM,EAC3EA,EAAiB,MAAM,IACzB,EACEO,GAAuBR,EAAGC,EAAiB,MAA8B,CAAC,EAAG,CAAC,CAAC,CACrF,EACMQ,EAAOD,GAAuBJ,EAAMH,EAAiB,MAAO,CAAC,EAAG,CAAC,CAAC,EAElES,EAASL,GAAQC,EAAIF,EAAMK,CAAI,CAAC,EAChCE,EAAOH,GAAuBE,EAAKT,EAAiB,MAAO,CAAC,EAAG,CAAC,CAAC,EAEvE,OAAUI,GAAQC,EAAIF,EAASE,EAAIG,EAAME,CAAI,CAAC,CAAC,CACjD,CAAC,CACH,CAEO,SAASC,GACdZ,EACAC,EACAC,EAAe,GACfW,EAAc,GACD,CACb,OAAUV,EAAK,IAAM,CACnB,IAAMC,EAAUC,GACdH,EACOI,EACAC,GAAOP,EAAIC,EAAiB,MAAqB,QAASY,EAAc,CAAC,EAAG,CAAC,EAAI,CAAC,EAAG,CAAC,EAAG,MAAM,EAClGZ,EAAiB,MAAM,IACzB,EACEO,GAAuBR,EAAGC,EAAiB,MAA8BY,EAAc,CAAC,EAAG,CAAC,EAAI,CAAC,EAAG,CAAC,CAAC,CAC5G,EACMJ,EAAOD,GAAuBJ,EAAMH,EAAiB,MAAO,CAAC,EAAG,CAAC,CAAC,EAElES,EAASL,GAAQC,EAAIF,EAAMK,CAAI,CAAC,EAChCE,EAAOH,GAAuBE,EAAKT,EAAiB,MAAO,CAAC,EAAG,CAAC,CAAC,EAEjEa,EAAST,GAAQC,EAAIF,EAASE,EAAIG,EAAME,CAAI,CAAC,CAAC,EAC9CI,EAAOP,GAAuBM,EAAKb,EAAiB,MAAO,CAAC,EAAG,CAAC,CAAC,EAEvE,OAAUI,GAAQC,EAAIF,EAASE,EAAIG,EAASH,EAAIK,EAAMI,CAAI,CAAC,CAAC,CAAC,CAC/D,CAAC,CACH,CClDO,SAASC,GACdC,EACAC,EACAC,EAA4B,OAC5BC,EAAW,GACE,CACb,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAASC,EACVC,GAAOP,EAAGC,EAAO,QAAS,CAAC,EAAG,CAAC,EAAGC,CAAO,EAC5CD,EAAO,IACT,EAEA,OAAOE,EAAcK,GAAKH,CAAG,EAAIA,CACnC,CAAC,CACH,CChBO,SAASI,GAA2BC,EAAgBC,EAA+B,CACxF,OAAO,KAAKD,CAAS,EAAE,QAASE,GAAS,CAClCD,EAAc,KAAME,GAAOA,EAAG,eAAiBD,CAAI,GACtDF,EAAUE,GAAM,QAAQ,CAE5B,CAAC,CACH,CCJO,SAASE,GACdC,EACAC,EACA,CACA,MAAO,CACLC,EACAC,EACAC,EACAC,IACe,CACf,IAAMC,EAAaC,GACjBP,EAAeE,EAAaC,EAAcC,EAAaA,CAAU,EACjE,CAACA,EAAYA,EAAYF,EAAYC,CAAW,CAClD,EACMK,EAAUC,GAAST,EAAeG,CAAW,CAAC,EAEpD,OAAAF,EAAc,KACZ,CAAE,UAAW,GAAGI,WAAuB,EACvC,CAAE,UAAW,GAAGA,QAAoB,CACtC,EAEO,CAAE,QAAAC,EAAS,KAAAE,CAAK,CACzB,CACF,CCvBO,SAASE,GACdC,EACAC,EACA,CACA,MAAO,CACLC,EACAC,EACAC,IACa,CACb,IAAMC,EAAgBC,GAASN,EAAeE,EAAaC,CAAW,EAAG,CAACD,EAAYC,CAAW,CAAC,EAC5FI,EAAaC,GAASR,EAAeG,CAAW,CAAC,EAEvD,OAAAF,EAAc,KACZ,CAAE,UAAW,GAAGG,WAAuB,EACvC,CAAE,UAAW,GAAGA,QAAoB,CACtC,EAEO,CACL,QAASC,EACT,KAAME,CACR,CACF,CACF,CCNO,IAAME,GAAN,KAA0B,CAE/B,YAESC,EAEAC,EAEAC,EAEP,CANO,sBAAAF,EAEA,sBAAAC,EAEA,UAAAC,CAEN,CACL,EC3BO,SAASC,GACdC,EACAC,EACA,CACA,MAAO,CAACC,EAAoBC,EAAqBC,IAA8C,CAC7F,IAAMC,EAAsBC,GAASN,EAAe,EAAQE,CAAU,EAAG,CAAC,EAAG,EAAGA,EAAY,CAAC,CAAC,EACxFK,EAAsBD,GAASN,EAAeE,EAAaC,CAAW,EAAG,CAAC,EAAG,EAAGD,EAAYC,CAAW,CAAC,EACxGK,EAAUC,GAAST,EAAeG,CAAW,CAAC,EAEpD,OAAAF,EAAc,KACZ,CAAE,UAAW,GAAGG,oBAAgC,EAChD,CAAE,UAAW,GAAGA,oBAAgC,EAChD,CAAE,UAAW,GAAGA,QAAoB,CACtC,EAEO,IAAIM,GACTL,EACAE,EACAC,CACF,CACF,CACF,CAEO,SAASG,GAEdC,EACA,CACA,OAAQC,GAAwC,CAC9C,IAAMR,EAAmBO,EAAgC,GAAGC,qBAA2B,CAAC,EAClFN,EAAmBK,EAAgC,GAAGC,qBAA2B,CAAC,EAClFL,EAAOI,EAAgC,GAAGC,SAAe,CAAC,EAEhE,OAAO,IAAIH,GACTL,EACAE,EACAC,CACF,CACF,CACF,CCvCO,SAASM,GAA0BC,EAAgBC,EAA+B,CACvF,MAAO,CAACC,EAAsBC,EAAmBC,IAAwB,CACvE,IAAMC,EAASL,EAAUE,GAEzB,GAAI,CAACI,GAASD,EAAQF,CAAS,EAC7B,MAAM,IAAI,MAAM,sBAAsBD,oBAA+BC,oBAA4BE,GAAQ,EAG3G,OAAAJ,EAAc,KACZ,CAAE,aAAAC,EAAc,UAAWE,GAAcF,CAAa,CACxD,EAEOG,CACT,CACF,CCjBO,SAASE,GAAsBC,EAAuB,CAC3D,IAAIC,EAAmBD,EAEvB,SAASE,EAAeC,EAAkC,CACxD,IAAMC,EAAMH,EAAiB,MAAM,EAAGE,CAAU,EAChD,OAAAF,EAAmBA,EAAiB,MAAME,CAAU,EAC7CC,CACT,CAEA,SAASC,GAAoC,CAC3C,OAAOJ,CACT,CAEA,MAAO,CACL,eAAAC,EACA,oBAAAG,CACF,CACF,CCdO,SAASC,GAAkBC,EAAwCC,EAA+B,CACvG,IAAMC,EAAoBC,GAAyBH,EAAgBC,CAAa,EAC1EG,EAA6BC,GAAkCL,EAAgBC,CAAa,EAElG,SAASK,EAAyBC,EAAoBC,EAAqBC,EAAsBC,EAAe,GAA0B,CACxI,IAAMC,EAAQD,EACVR,EAAkBK,EAAYC,EAAa,EAAG,GAAGC,SAAoB,EACrEL,EAA2BG,EAAYC,EAAa,GAAGC,SAAoB,EACzEG,EAAQR,EAA2BI,EAAaA,EAAa,GAAGC,SAAoB,EACpFI,EAAQT,EAA2BI,EAAaA,EAAa,GAAGC,SAAoB,EAE1F,MAAO,CAAE,MAAAE,EAAO,MAAAC,EAAO,MAAAC,CAAM,CAC/B,CAEA,SAASC,EAAyBP,EAAoBC,EAAqBC,EAAsBC,EAAe,GAA0B,CACxI,GAAM,CAAE,MAAAC,EAAO,MAAAC,EAAO,MAAAC,CAAM,EAAIP,EAAyBC,EAAYC,EAAaC,EAAcC,CAAY,EACtGK,EAAQX,EAA2BI,EAAaA,EAAa,GAAGC,SAAoB,EAE1F,MAAO,CACL,MAAAE,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAE,CACvB,CACF,CAEA,MAAO,CACL,yBAAAT,EACA,yBAAAQ,CACF,CACF,CC1BO,SAASE,GAAcC,EAA8F,CAC1H,IAAMC,EAAgC,CAAC,EAEjC,CACJ,eAAAC,EACA,oBAAAC,CACF,EAAIC,GAAsBJ,CAAO,EAE3B,CACJ,yBAAAK,CACF,EAAIC,GAAkBJ,EAAgBD,CAAa,EAE7CM,EAASF,EAAyB,EAAG,GAAI,SAAU,EAAI,EACvDG,EAASH,EAAyB,GAAI,GAAI,QAAQ,EAClDI,EAASJ,EAAyB,GAAI,IAAK,QAAQ,EACnDK,EAASL,EAAyB,IAAK,IAAK,QAAQ,EAE1D,GAAIF,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,cAAAF,EACA,OAAQ,CACN,OAAAM,EAAQ,OAAAC,EAAQ,OAAAC,EAAQ,OAAAC,CAC1B,CACF,CACF,CC1BO,SAASC,GAAsBC,EAAuE,CAC3G,OAAQC,GAA+B,CACrC,IAAMC,EAAUF,EAAgC,GAAGC,YAAkB,CAAC,EAChEE,EAAOH,EAAgC,GAAGC,SAAe,CAAC,EAEhE,MAAO,CAAE,QAAAC,EAAS,KAAAC,CAAK,CACzB,CACF,CCRO,SAASC,GAAkBC,EAAgBC,EAA+B,CAC/E,IAAMC,EAAqBC,GAA0BH,EAAWC,CAAa,EAEvEG,EAAoBC,GAAsBH,CAAkB,EAC5DI,EAA6BC,GAA+BL,CAAkB,EAEpF,SAASM,EAAyBC,EAAgBC,EAAe,GAA0B,CACzF,IAAMC,EAAQD,EACVN,EAAkB,GAAGK,SAAc,EACnCH,EAA2B,GAAGG,SAAc,EAC1CG,EAAQN,EAA2B,GAAGG,SAAc,EACpDI,EAAQP,EAA2B,GAAGG,SAAc,EAE1D,MAAO,CAAE,MAAAE,EAAO,MAAAC,EAAO,MAAAC,CAAM,CAC/B,CAEA,SAASC,EAAyBL,EAAgBC,EAAe,GAA0B,CACzF,IAAMC,EAAQD,EACVN,EAAkB,GAAGK,SAAc,EACnCH,EAA2B,GAAGG,SAAc,EAC1CG,EAAQN,EAA2B,GAAGG,SAAc,EACpDI,EAAQP,EAA2B,GAAGG,SAAc,EACpDM,EAAQT,EAA2B,GAAGG,SAAc,EAE1D,MAAO,CACL,MAAAE,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAE,CACvB,CACF,CAEA,MAAO,CACL,yBAAAP,EACA,yBAAAM,CACF,CACF,CC/BO,SAASE,GACdC,EACuE,CACvE,IAAMC,EAAgC,CAAC,EAEjC,CACJ,yBAAAC,CACF,EAAIC,GAAkBH,EAAWC,CAAa,EAExCG,EAAS,CACb,OAAQF,EAAyB,SAAU,EAAI,EAC/C,OAAQA,EAAyB,QAAQ,EACzC,OAAQA,EAAyB,QAAQ,EACzC,OAAQA,EAAyB,QAAQ,CAC3C,EAEA,OAAAG,GAA2BL,EAAWC,CAAa,EAE5C,CAAE,OAAAG,EAAQ,cAAAH,CAAc,CACjC,CCfO,IAAMK,GAAN,cAAmCC,EAAuG,CAC/I,aAAc,CACZ,MAAM,sBAAsB,CAC9B,CAEO,aAAaC,EAA8B,CAChD,GAAM,CAAE,OAAAC,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,oDAAoD,EAGtE,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAiBC,GAAKJ,EAAM,cAAc,IAAK,EAAI,EAAG,SAAS,EAE/DK,EAAaC,GAAUH,EADb,CAAC,QAAS,QAAS,OAAO,CACO,EAAE,IAAI,GAAG,EAEtDI,EAAMC,GAAYH,EAAYJ,EAAO,OAAQ,EAAI,EACrD,OAAAM,EAAMC,GAAYD,EAAKN,EAAO,MAAM,EACpCM,EAAMC,GAAYD,EAAKN,EAAO,MAAM,EACpCM,EAAMC,GAAYD,EAAKN,EAAO,MAAM,EACpCM,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,OAAO,EAEtCA,CACT,CAAC,CACH,CAEA,MAAa,QAAQP,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAMU,GAAWV,CAAK,CAAC,CAClD,CAEU,qBAA8B,CACtC,MAAO,8BACT,CAEU,2BAA2BW,EAA8B,CACjE,OAAOC,GAA2BD,CAAS,CAC7C,CAEU,cAAcE,EAAuB,CAC7C,OAAOC,GAAcD,CAAO,CAC9B,CACF,EChDO,SAASE,GACdC,EACAC,EACa,CACb,OAAUC,EAAK,IAASC,EACnBC,GAAOJ,EAAGC,EAAO,OAAO,EAC3BA,EAAO,IACT,CAAC,CACH,CCTO,SAASI,GAAcC,EAAuBC,EAAoBC,EAA2E,CAClJ,IAAMC,EAAgC,CAAC,EAEjC,CACJ,eAAAC,EACA,oBAAAC,CACF,EAAIC,GAAsBN,CAAO,EAI3BO,EAFkBC,GAAuBJ,EAAgBD,CAAa,EAEjDF,EAAYC,EAAa,IAAI,EAExD,GAAIG,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,cAAAF,EACA,OAAQ,CAAE,GAAAI,CAAG,CACf,CACF,CClBO,SAASE,GACdC,EACsD,CACtD,IAAMC,EAAgC,CAAC,EAEjCC,EAAqBC,GAA0BH,EAAWC,CAAa,EAE7E,SAASG,EAAgBC,EAA0B,CACjD,IAAMC,EAAUJ,EAAmB,GAAGG,YAAkB,CAAC,EACnDE,EAAOL,EAAmB,GAAGG,SAAe,CAAC,EACnD,MAAO,CAAE,QAAAC,EAAS,KAAAC,CAAK,CACzB,CAEA,IAAMC,EAAS,CACb,GAAIJ,EAAgB,IAAI,CAC1B,EAEA,OAAAK,GAA2BT,EAAWC,CAAa,EAE5C,CAAE,OAAAO,EAAQ,cAAAP,CAAc,CACjC,CCvBO,SAASS,GAAmBC,EAA8B,CAC/D,IAAMC,EAAyC,CAAC,EAC1CC,EAAmC,CAAC,EAE1C,cAAO,KAAKF,CAAS,EAAE,QAASG,GAAQ,CACtC,IAAMC,EAAMD,EAAI,WAAW,IAAI,EAAID,EAAgBD,EACnDG,EAAID,GAAOH,EAAUG,EACvB,CAAC,EAEM,CAAE,oBAAAF,EAAqB,cAAAC,CAAc,CAC9C,CCDO,IAAeG,GAAf,cAGGC,EAAyB,CAGjC,YAAYC,EAAeC,EAA+D,CACxF,MAAMD,CAAK,EACX,KAAK,sBAAwBC,CAC/B,CAEA,IAAW,sBAAgE,CACzE,OAAO,KAAK,qBACd,CAQO,OAAOC,EAA4C,CACxD,GAAM,CAAE,OAAAC,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,GAAG,KAAK,qCAAqC,EAG/D,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAqBH,aAAiBI,GACxC,KAAK,qBAAqB,aAAaJ,CAAK,EAC5CA,EACJ,OAAOK,GAAoBF,EAAmB,KAAKA,EAAmB,MAAM,GAAI,EAAE,EAAGF,EAAO,EAAE,CAChG,CAAC,CACH,CAEgB,QAAQK,EAAmB,GAAM,CAC/C,KAAK,qBAAqB,QAAQA,CAAgB,EAClD,MAAM,QAAQA,CAAgB,CAChC,CAEO,qBAAqBC,EAAuB,CACjD,GAAM,CAAE,OAAAN,EAAQ,cAAAO,CAAc,EAAI,KAAK,wBAAwBD,CAAO,EACtE,KAAK,QAAUN,EACf,KAAK,eAAiBO,CACxB,CAEO,wBAAwBD,EAAuB,CACpD,OAAOE,GAAcF,EAAS,KAAK,wBAAwB,EAAG,KAAK,yBAAyB,CAAC,CAC/F,CAEU,2BAA2BG,EAA8B,CACjE,GAAM,CAAE,oBAAAC,EAAqB,cAAAC,CAAc,EAAIC,GAAmBH,CAAS,EAE3E,YAAK,qBAAqB,kBAAkBC,CAAmB,EAExDG,GAA2BF,CAAa,CACjD,CAEU,cAAcL,EAAuB,CAC7C,IAAMQ,EAAM,KAAK,wBAAwB,EACnCC,EAAO,KAAK,yBAAyB,EACrCC,EAAwBD,EAAOD,EAAOC,EAEtCE,EAA0BX,EAAQ,MAAM,EAAGA,EAAQ,OAASU,CAAoB,EAChFE,EAAoBZ,EAAQ,MAAMA,EAAQ,OAASU,CAAoB,EAE7E,YAAK,qBAAqB,eAAeC,CAAuB,EACzD,KAAK,wBAAwBC,CAAiB,CACvD,CACF,ECjFO,IAAMC,GAAyB,CAAC,UAAW,QAAS,MAAO,QAAS,UAAW,YAAa,WAAW,EAEjGC,GAAN,KAAsB,CAS3B,YAAYC,EAAwC,CARpD,KAAO,QAAU,EACjB,KAAO,MAAQ,EACf,KAAO,IAAM,EACb,KAAO,MAAQ,EACf,KAAO,QAAU,EACjB,KAAO,UAAY,EACnB,KAAO,UAAY,EAGjB,GAAIA,EAAc,SAAW,EAC3B,MAAM,IAAI,MAAM,8EAA8EA,EAAc,QAAQ,EAGtHF,GAAuB,QAAQ,CAACG,EAAYC,IAAQ,CAClD,KAAKD,GAAcD,EAAcE,EACnC,CAAC,CACH,CAEA,eAAgB,CACd,OAAOJ,GACJ,IAAKG,IAAgB,CAAE,WAAAA,EAAY,YAAa,KAAKA,EAAsB,EAAE,EAC7E,KAAK,CAACE,EAAIC,IAAOA,EAAG,YAAcD,EAAG,WAAW,CACrD,CACF,EClBO,IAAME,GAAN,cAAgCC,EAA0C,CAC/E,YAAYC,EAA6C,IAAIC,GAAwB,CACnF,MAAM,oBAAqBD,CAAoB,CACjD,CAEO,aAAaE,EAA4C,CAC9D,OAAUC,EAAK,IAASC,GAAQ,KAAK,OAAOF,CAAK,CAAC,CAAC,CACrD,CAEA,MAAa,QAAQA,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAMG,GAAWH,CAAK,CAAC,CAClD,CAEA,MAAa,mBAAmBA,EAAkB,CAChD,IAAMI,EAAW,MAAMD,GAAWH,CAAK,EACjCK,EAAM,MAAM,KAAK,aAAaD,CAAQ,EACtCE,EAAsB,MAAM,QAAQ,IAAOC,GAAQF,CAAG,EAAE,IAAI,MAAOG,GAAM,CAC7E,IAAMC,EAAOD,EAAE,SAAS,EACxB,OAAAA,EAAE,QAAQ,EACHC,CACT,CAAC,CAAC,EACFJ,EAAI,QAAQ,EAEZ,IAAMK,EAAqBJ,EACxB,IAAKK,GAAiB,IAAIC,GAAgBD,CAA4B,CAAC,EAE1E,OAAOP,EAAS,aACZM,EACAA,EAAmB,EACzB,CAEU,qBAA8B,CACtC,MAAO,uBACT,CAEU,yBAAkC,CAC1C,MAAO,IACT,CAEU,0BAAmC,CAC3C,MAAO,EACT,CACF,EC9CO,SAASG,GAAsBC,EAA0C,CAC9E,OAAOA,EAAI,uBAAuBC,EACpC,CAEO,SAASC,GAAmCC,EAAoBC,EAA4D,CAEjI,MAAO,CAAE,GAAGD,EAAW,GADL,CAAE,YAAAC,CAAY,CACI,CACtC,CCFO,SAASC,IAAoBC,EAAuCC,EAA6EC,EAAgB,GAAKC,EAA0B,EACxK,MAAM,QAAQF,CAAe,EAAIA,EAAkB,CAACA,CAAe,GAE3E,QAASG,GAAM,CAElC,IAAMC,EAAOD,aAAaE,GACtBF,EACCG,GAAsBH,CAAC,EAAIA,EAAE,YAAc,OAChD,GAAI,CAACC,EACH,MAAM,IAAI,MAAM,iHAAiH,EAInI,IAAMG,EADSH,EAAK,cAAc,EACF,OAAQI,GAAcA,EAAU,YAAcP,CAAa,EAErFQ,EAASC,GAAoBP,CAAC,EAChCA,EAAE,UAAU,IAAI,WACfD,GAAmB,IAAIS,GAAM,EAAG,CAAC,EAEhB,IAAIC,GACxBL,EAAiB,IAAKC,GAAc,GAAGA,EAAU,eAAeK,GAAML,EAAU,WAAW,IAAI,EAC/FC,CACF,EACc,KAAKV,CAAS,CAC9B,CAAC,CACH,CCfO,SAASe,GACdC,EACgE,CAChE,OACEC,GAAoBD,CAAG,GAEpBA,EAAI,qBAAwBE,IAE5BF,EAAI,8BAAiCE,IAErCF,EAAI,uBAA0BG,EAErC,CAEA,SAASC,IAAmBC,EAAM,CAGhC,IAAMC,EAAWC,GAAaA,EAAU,IAAO,KAAK,GAC9CC,EAA6B,CAACC,EAAGC,IAAM,KAAK,MAAMD,EAAE,GAAKC,EAAE,KAAO,GAAKD,EAAE,GAAKC,EAAE,KAAO,CAAC,EAExFC,EAAQ,CACZ,KAA0B,OAC1B,MAA2B,OAC3B,IAAyB,MAC3B,EAEMC,EAAU,CAACC,EAAWC,EAAUC,IAAe,CAEnD,IAAMC,EAAiB,KAAK,MAAMH,EAAU,GAAKC,EAAS,EAAE,EAEtDG,EAAkB,KAAK,MAAMH,EAAS,GAAKC,EAAW,EAAE,EAE9D,OAAOC,EAAiBC,CAC1B,EAEMC,EAAW,CAACC,EAAOC,IAAU,CAIjC,IAAMC,EAAa,KAAK,MAAMD,EAAM,GAAKD,EAAM,GAAIC,EAAM,GAAKD,EAAM,EAAE,EAChEG,EAAWF,EAAM,GAAKD,EAAM,GAC5BI,EAAiB,KAAK,KAAKD,EAAWD,CAAU,EAChDG,EAAiBlB,EAAQiB,CAAc,EACvCE,EAAiB,KAAK,MAAM,GAAKD,CAAc,EAG/CE,EAAgBN,EAAM,GAAKD,EAAM,GAAK,EAAI,GAAK,EAErD,OADeM,EAAiBC,CAElC,EAEMC,EAAY,CAACd,EAAWC,EAAUC,IAAe,CAKrD,IAAMa,EAAOpB,EAA2BK,EAAWE,CAAU,EAEvDc,EAAa,CACjB,IAAKhB,EAAU,GAAKE,EAAW,IAAM,EACrC,IAAKF,EAAU,GAAKE,EAAW,IAAM,CACvC,EACMe,EAAkBtB,EAA2BM,EAAUe,CAAU,EACjEN,EAAiB,KAAK,KAAKO,EAAkBF,CAAI,EACjDJ,EAAiB,KAAK,MAAMlB,EAAQiB,CAAc,CAAC,EAGnDQ,EAAYF,EAAW,GAAKf,EAAS,GAAK,EAAI,GAAK,EAEzD,OADeU,EAAiBO,CAElC,EAEA,GAAI,CAAC1B,GAAQ,CAACA,EAAK,YAAcA,EAAK,WAAW,SAAW,GAAI,OAAOM,EACvE,IAAMqB,EAAK3B,EAAK,WAChB,OAAAM,EAAM,KAAOO,EAASc,EAAG,IAAKA,EAAG,GAAG,EACpCrB,EAAM,MAAQgB,EAAUK,EAAG,IAAKA,EAAG,IAAKA,EAAG,EAAE,EAC7CrB,EAAM,IAAMC,EAAQoB,EAAG,IAAKA,EAAG,IAAKA,EAAG,EAAE,EAClCrB,CACT,CAEO,SAASsB,GACdC,EACAC,EAC4C,CAC5C,GAAM,CAAE,IAAKC,CAAM,EAAIF,EAAU,UAC3BG,EAAYF,EAAmB,QAAwBC,EAAM,EAAGA,EAAM,CAAC,EACvEE,EAAOD,EAAU,MAAM,EACvB,CAAE,UAAAE,CAAU,EAAIL,EAAU,UAC1BM,EAAc,IAAIrC,GACtB+B,EAAU,UAAU,MACpBI,EAAK,QAAQC,EAAU,QAAQ,CAAC,EAChCA,CACF,EACM5B,EAAQP,IAAmB+B,CAAkB,EAEnD,MAAO,CAAE,GAAGD,EAAW,GADL,CAAE,UAAAG,EAAW,mBAAAF,EAAoB,YAAAK,EAAa,MAAA7B,CAAM,CAClC,CACtC,CCjGO,IAAM8B,GAAN,KAA+B,CAapC,YAAYC,EAAqC,CAAC,EAAG,CACnD,GAAM,CACJ,UAAAC,EAAY,GAAM,WAAAC,EAAa,GAAM,UAAAC,EAAW,UAAAC,EAAW,UAAAC,EAAW,WAAAC,CACxE,EAAIN,EACJ,KAAK,UAAYC,EACjB,KAAK,WAAaC,EAClB,KAAK,UAAYC,GAAa,EAC9B,KAAK,UAAYE,GAAa,EAC9B,KAAK,UAAYD,GAAa,uBAC9B,KAAK,WAAaE,GAAc,sBAClC,CACF,EAEaC,GAAN,KAAwB,CAK7B,YACEC,EACAR,EAAqC,CAAC,EACtC,CACA,KAAK,cAAgBQ,EACrB,KAAK,QAAU,IAAIT,GAAyBC,CAAO,CACrD,CAEA,KAAKS,EAAkE,CACrE,IAAMC,EAAMC,GAAoBF,CAAS,EAEnC,CACJ,UAAAR,EAAW,WAAAC,EAAY,UAAAC,EAAW,UAAAC,EAAW,UAAAC,EAAW,WAAAC,CAC1D,EAAI,KAAK,QAcT,GAZIL,GAAa,KAAK,yBAAyBW,KAC7CF,EAAI,YAAcN,EAClBM,EAAI,UAAYP,EAChBU,GAAYH,EAAK,KAAK,cAAc,cAAc,CAAC,EACnDG,GAAYH,EAAK,KAAK,cAAc,eAAe,CAAC,EACpDG,GAAYH,EAAK,KAAK,cAAc,gBAAgB,CAAC,EACrDG,GAAYH,EAAK,KAAK,cAAc,QAAQ,CAAC,EAC7CG,GAAYH,EAAK,KAAK,cAAc,WAAW,EAAG,EAAI,EACtDG,GAAYH,EAAK,KAAK,cAAc,YAAY,EAAG,EAAI,EACvDG,GAAYH,EAAK,KAAK,cAAc,SAAS,EAAG,EAAI,GAGlDR,EAAY,CACdQ,EAAI,YAAcJ,EAClBI,EAAI,UAAYJ,EAEhB,IAAMQ,EAAaC,GAAe,CAChCL,EAAI,UAAU,EACdA,EAAI,IAAIK,EAAG,EAAGA,EAAG,EAAGV,EAAW,EAAG,EAAI,KAAK,EAAE,EAC7CK,EAAI,KAAK,CACX,EACA,KAAK,cAAc,UAAU,QAAQI,CAAS,CAChD,CACF,CACF,EAIO,SAASE,IACdP,EACAD,EACA,EAC2B,MAAM,QAAQA,CAAa,EAAIA,EAAgB,CAACA,CAAa,GACrE,QAASS,GAAM,CAEhC,IAAMC,EAAYD,aAAaE,GAC3BF,EACCG,GAAoBH,CAAC,EAAIA,EAAE,UAAY,OAC5C,GAAI,CAACC,EACH,MAAM,IAAI,MAAM,8HAA8H,EAGhJ,IAAIX,GAAkBW,CAAS,EAAE,KAAKT,CAAS,CACjD,CAAC,CACH,gBCvGA,SAASY,IAAkBC,EAAwCC,EAA+B,CAChG,IAAMC,EAAoBC,GAAyBH,EAAgBC,CAAa,EAC1EG,EAA6BC,GAAkCL,EAAgBC,CAAa,EAElG,SAASK,EAA4BC,EAAoBC,EAAqBC,EAA4C,CACxH,IAAMC,EAAkBN,EAA2BG,EAAYC,EAAa,GAAGC,mBAA8B,EACvGE,EAAkBP,EAA2BI,EAAaA,EAAa,GAAGC,mBAA8B,EACxGG,EAAiBV,EAAkBK,EAAYC,EAAa,EAAG,GAAGC,kBAA6B,EAErG,MAAO,CAAE,gBAAAC,EAAiB,gBAAAC,EAAiB,eAAAC,CAAe,CAC5D,CAEA,SAASC,EAAuBC,EAAkBL,EAAuC,CACvF,IAAMC,EAAkBN,EAA2BU,EAAUA,EAAU,GAAGL,mBAA8B,EAClGE,EAAkBP,EAA2BU,EAAUA,EAAU,GAAGL,mBAA8B,EAClGM,EAAkBX,EAA2BU,EAAUA,EAAU,GAAGL,mBAA8B,EAExG,MAAO,CAAE,gBAAAC,EAAiB,gBAAAC,EAAiB,gBAAAI,CAAgB,CAC7D,CAEA,MAAO,CACL,kBAAAb,EACA,2BAAAE,EACA,4BAAAE,EACA,uBAAAO,CACF,CACF,CAEO,SAASG,GAAcC,EAAuBC,EAAsF,CACzI,IAAMjB,EAAgC,CAAC,EAEjC,CACJ,eAAAD,EACA,oBAAAmB,CACF,EAAIC,GAAsBH,CAAO,EAE3B,CACJ,kBAAAf,EACA,2BAAAE,EACA,4BAAAE,EACA,uBAAAO,CACF,EAAId,IAAkBC,EAAgBC,CAAa,EAE7CoB,EAAqBnB,EAAkB,EAAG,GAAI,EAAG,oBAAoB,EACrEoB,EAA+BhB,EAA4B,GAAI,GAAI,8BAA8B,EACjGiB,EAA+BjB,EAA4B,GAAI,IAAK,8BAA8B,EAElGkB,EAAa,CACjB,QAASH,EACT,kBAAmBC,EACnB,kBAAmBC,CACrB,EAEME,EAAc,CAAC,EACrBC,GAAMR,EAAe,EAAG,CAAC,EAAE,QAASS,GAAQ,CAC1CF,EAAY,cAAcE,KAASd,EAAuB,IAAK,0BAA0Bc,GAAK,CAChG,CAAC,EAED,IAAMC,EAA4BtB,EAA4B,IAAK,IAAK,2BAA2B,EAC7FuB,EAA2BzB,EAA2B,IAAK,IAAK,0BAA0B,EAE1F0B,EAAY,CAChB,gBAAiBF,EACjB,eAAgBC,CAClB,EAEA,GAAIV,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,cAAAlB,EACA,OAAQ,CAAE,WAAAuB,EAAY,YAAAC,EAAa,UAAAK,CAAU,CAC/C,CACF,CCxEA,SAASC,IAAkBC,EAAgBC,EAA+B,CACxE,IAAMC,EAAqBC,GAA0BH,EAAWC,CAAa,EAEvEG,EAAoBC,GAAsBH,CAAkB,EAC5DI,EAA6BC,GAA+BL,CAAkB,EAEpF,SAASM,EAA4BC,EAA4C,CAC/E,IAAMC,EAAkBJ,EAA2B,GAAGG,mBAA8B,EAC9EE,EAAkBL,EAA2B,GAAGG,mBAA8B,EAC9EG,EAAiBR,EAAkB,GAAGK,kBAA6B,EAEzE,MAAO,CAAE,gBAAAC,EAAiB,gBAAAC,EAAiB,eAAAC,CAAe,CAC5D,CAEA,SAASC,EAAuBJ,EAAuC,CACrE,IAAMC,EAAkBJ,EAA2B,GAAGG,mBAA8B,EAC9EE,EAAkBL,EAA2B,GAAGG,mBAA8B,EAC9EK,EAAkBR,EAA2B,GAAGG,mBAA8B,EAEpF,MAAO,CAAE,gBAAAC,EAAiB,gBAAAC,EAAiB,gBAAAG,CAAgB,CAC7D,CAEA,MAAO,CACL,kBAAAV,EACA,2BAAAE,EACA,4BAAAE,EACA,uBAAAK,CACF,CACF,CAEO,SAASE,GACdf,EACAgB,EAC+D,CAC/D,IAAMf,EAAgC,CAAC,EAEjC,CACJ,kBAAAG,EACA,2BAAAE,EACA,4BAAAE,EACA,uBAAAK,CACF,EAAId,IAAkBC,EAAWC,CAAa,EAExCgB,EAAqBb,EAAkB,oBAAoB,EAC3Dc,EAA+BV,EAA4B,8BAA8B,EACzFW,EAA+BX,EAA4B,8BAA8B,EAEzFY,EAAa,CACjB,QAASH,EACT,kBAAmBC,EACnB,kBAAmBC,CACrB,EAEME,EAAc,CAAC,EACrBC,GAAMN,EAAe,EAAG,CAAC,EAAE,QAASO,GAAQ,CAC1CF,EAAY,cAAcE,KAASV,EAAuB,0BAA0BU,GAAK,CAC3F,CAAC,EAED,IAAMC,EAA4BhB,EAA4B,2BAA2B,EACnFiB,EAA2BnB,EAA2B,0BAA0B,EAEhFoB,EAAY,CAChB,gBAAiBF,EACjB,eAAgBC,CAClB,EAEA,OAAAE,GAA2B3B,EAAWC,CAAa,EAE5C,CAAE,OAAQ,CAAE,WAAAmB,EAAY,YAAAC,EAAa,UAAAK,CAAU,EAAG,cAAAzB,CAAc,CACzE,CCjEA,SAAS2B,GAAKC,EAAgBC,EAAoBC,EAAuC,CACvF,OAAUC,EAAOC,GAAOJ,EAAGC,EAAO,QAASC,EAAQ,MAAM,EAAGD,EAAO,IAAI,CACzE,CAEA,SAASI,GAAeL,EAAgBC,EAA8BK,EAAkB,GAAmB,CACzG,IAAIC,EAAMD,EAAqBE,GAAKR,CAAC,EAAIA,EACzC,OAAAO,EAAME,GAAuBF,EAAKN,EAAO,gBAAiB,CAAC,EAAG,CAAC,CAAC,EAChEM,EAAME,GAA0BD,GAAKD,CAAG,EAAGN,EAAO,gBAAiB,CAAC,EAAG,CAAC,CAAC,EACzEM,EAASG,GAAQH,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAASJ,EAAII,EAAKR,GAAKC,EAAGC,EAAO,eAAgB,CAAC,EAAG,CAAC,CAAC,CAAC,EACjDM,CACT,CAEA,SAASI,IAAUX,EAAgBC,EAAsC,CACvE,IAAIM,EAAME,GAA0BD,GAAKR,CAAC,EAAGC,EAAO,gBAAiB,CAAC,EAAG,CAAC,CAAC,EAC3E,OAAAM,EAAME,GAA0BD,GAAKD,CAAG,EAAGN,EAAO,gBAAiB,CAAC,EAAG,CAAC,CAAC,EACzEM,EAAME,GAA0BD,GAAKD,CAAG,EAAGN,EAAO,gBAAiB,CAAC,EAAG,CAAC,CAAC,EACzEM,EAASJ,EAAII,EAAKP,CAAC,EACZO,CACT,CAEO,IAAMK,GAAN,cAA2BC,EAAkC,CAGlE,YAAYC,EAAuB,CACjC,MAAM,cAAc,EACpB,KAAK,eAAiBA,CACxB,CAEO,aAAaC,EAA8B,CAChD,GAAM,CAAE,OAAAd,CAAO,EAAI,KACnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,4CAA4C,EAE9D,OAAUe,EAAK,IAAM,CACnB,IAAMC,EAAiBC,GAAKH,EAAM,cAAc,IAAK,EAAI,EAAG,SAAS,EAE/DI,EAAaC,GAAUH,EADb,CAAC,QAAS,QAAS,OAAO,CACO,EAAE,IAAI,GAAG,EACtDV,EAASC,GAAKT,GAAKoB,EAAYlB,EAAO,WAAW,QAAS,CAAC,EAAG,CAAC,CAAC,CAAC,EACrE,OAAAM,EAAMF,GAAeE,EAAKN,EAAO,WAAW,kBAAmB,EAAK,EACpEM,EAAMF,GAAeE,EAAKN,EAAO,WAAW,iBAAiB,EAC7DoB,GAAM,KAAK,eAAgB,EAAG,CAAC,EAAE,QAASC,GAAQ,CAChDf,EAAMI,IAAUJ,EAAKN,EAAO,YAAY,cAAcqB,IAAM,CAC9D,CAAC,EACDf,EAAMF,GAAeE,EAAKN,EAAO,UAAU,eAAe,EAC1DM,EAASC,GAAKC,GAAuBF,EAAKN,EAAO,UAAU,eAAgB,CAAC,EAAG,CAAC,CAAC,CAAC,EAC3EM,CACT,CAAC,CACH,CAEA,MAAa,QAAQQ,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAMQ,GAAWR,CAAK,CAAC,CAClD,CAEU,qBAA8B,CACtC,MAAO,qBACT,CAEU,2BAA2BS,EAA8B,CACjE,OAAOC,GAA2BD,EAAW,KAAK,cAAc,CAClE,CAEU,cAAcE,EAAuB,CAC7C,OAAOC,GAAcD,EAAS,KAAK,cAAc,CACnD,CACF,ECzEO,SAASE,GAAcC,EAA6E,CACzG,IAAMC,EAAgC,CAAC,EAEjC,CACJ,eAAAC,EACA,oBAAAC,CACF,EAAIC,GAAsBJ,CAAO,EAE3BK,EAAkBC,GAAuBJ,EAAgBD,CAAa,EAEtEM,EAAMF,EAAgB,IAAK,EAAG,QAAQ,EACtCG,EAASH,EAAgB,IAAK,EAAG,WAAW,EAElD,GAAIF,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,cAAAF,EACA,OAAQ,CAAE,GAAI,CAAE,IAAAM,EAAK,OAAAC,CAAO,CAAE,CAChC,CACF,CCnBO,SAASC,GACdC,EACsD,CACtD,IAAMC,EAAgC,CAAC,EAEjCC,EAAqBC,GAA0BH,EAAWC,CAAa,EAE7E,SAASG,EAAgBC,EAA0B,CACjD,IAAMC,EAAUJ,EAAmB,GAAGG,YAAkB,CAAC,EACnDE,EAAOL,EAAmB,GAAGG,SAAe,CAAC,EACnD,MAAO,CAAE,QAAAC,EAAS,KAAAC,CAAK,CACzB,CAEA,IAAMC,EAAS,CACb,GAAI,CACF,IAAKJ,EAAgB,QAAQ,EAC7B,OAAQA,EAAgB,WAAW,CACrC,CACF,EAEA,OAAAK,GAA2BT,EAAWC,CAAa,EAE5C,CAAE,OAAAO,EAAQ,cAAAP,CAAc,CACjC,CCvBO,IAAKS,QAEVA,EAAA,OAAS,SAETA,EAAA,KAAO,OAJGA,QAAA,ICKL,IAAMC,GAAN,cAA2BC,EAAyB,CAGzD,YAAYC,EAAqC,IAAIC,GAAa,CAAC,EAAG,CACpE,MAAM,cAAc,EACpB,KAAK,sBAAwBD,CAC/B,CAEA,IAAW,sBAAqC,CAC9C,OAAO,KAAK,qBACd,CAEO,OAAOE,EAA0C,CACtD,GAAM,CAAE,OAAAC,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,GAAG,KAAK,qCAAqC,EAG/D,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAqBH,aAAiBI,GACxC,KAAK,qBAAqB,aAAaJ,CAAK,EAC5CA,EAEEK,EAAYC,GAAQH,EAAoB,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,OAAO,EAAE,KAAKA,EAAmB,MAAM,GAAI,EAAE,EACrGI,EAAMC,GAAoBH,EAAQJ,EAAO,GAAG,GAAG,EAAE,KAAK,EACtDQ,EAASD,GAAoBH,EAAQJ,EAAO,GAAG,MAAM,EAC3D,MAAO,CAAE,IAAAM,EAAK,OAAAE,CAAO,CACvB,CAAC,CACH,CAEO,aAAaT,EAA0C,CAC5D,OAAUE,EAAK,IAAM,CACnB,GAAM,CAAE,IAAAK,EAAK,OAAAE,CAAO,EAAI,KAAK,OAAOT,CAAK,EACzC,MAAO,CAAE,IAAAO,EAAK,OAAWG,GAAQD,CAAM,CAAE,CAC3C,CAAC,CACH,CAEA,MAAa,QAAQT,EAAsC,CACzD,OAAO,KAAK,aAAa,MAAMW,GAAWX,CAAK,CAAC,CAClD,CAEA,MAAa,oBAAoBA,EAA8E,CAC7G,IAAMY,EAAW,MAAMD,GAAWX,CAAK,EACjCa,EAAM,MAAM,KAAK,aAAaD,CAAQ,EAEtCE,EAAUC,GAAQF,EAAI,GAAG,EACzBG,EAAaD,GAAQF,EAAI,MAAM,EAC/BI,EAAsBH,EAAK,IAAI,CAACI,EAAWC,KAAO,CACtD,UAAAD,EACA,aAAcF,EAAQG,EACxB,EAAE,EAEIC,EAAqB,MAAM,QAAQ,IACvCH,EAAoB,IAAI,MAAO,CAAE,UAAAC,EAAW,aAAAG,CAAa,IAAM,CAC7D,IAAMd,EAAOW,EAAU,SAAS,EAAG,GAC7BI,EAAYD,EAAa,SAAS,EAAG,GACrCE,EAASD,EAAW,GACpBb,EAASc,kBACTC,EAAoBD,EAASD,EAAY,EAAIA,EAEnD,OAAAJ,EAAU,QAAQ,EAClBG,EAAa,QAAQ,EACd,CAAE,IAAAd,EAAK,OAAAE,EAAQ,kBAAAe,CAAkB,CAC1C,CAAC,CACH,EACA,OAAAX,EAAI,IAAI,QAAQ,EAChBA,EAAI,OAAO,QAAQ,EAEZD,EAAS,aAAeQ,EAAiDA,EAAmB,EACrG,CAEU,qBAA8B,CACtC,MAAO,kBACT,CAEgB,QAAQK,EAAmB,GAAM,CAC/C,KAAK,qBAAqB,QAAQA,CAAgB,EAClD,MAAM,QAAQA,CAAgB,CAChC,CAEO,qBAAqBC,EAAuB,CACjD,GAAM,CAAE,OAAAzB,EAAQ,cAAA0B,CAAc,EAAI,KAAK,wBAAwBD,CAAO,EACtE,KAAK,QAAUzB,EACf,KAAK,eAAiB0B,CACxB,CAEO,wBAAwBD,EAAuB,CACpD,OAAOE,GAAcF,CAAO,CAC9B,CAEU,2BAA2BG,EAA8B,CACjE,GAAM,CAAE,oBAAAC,EAAqB,cAAAC,CAAc,EAAIC,GAAmBH,CAAS,EAE3E,YAAK,qBAAqB,kBAAkBC,CAAmB,EAExDG,GAA2BF,CAAa,CACjD,CAEU,cAAcL,EAAuB,CAG7C,IAAMQ,EAA0BR,EAAQ,MAAM,EAAGA,EAAQ,OAAS,IAAoB,EAChFS,EAAoBT,EAAQ,MAAMA,EAAQ,OAAS,IAAoB,EAE7E,YAAK,qBAAqB,eAAeQ,CAAuB,EACzD,KAAK,wBAAwBC,CAAiB,CACvD,CACF,EC7GO,IAAeC,GAAf,cAGGC,EAAgC,CACjC,YAAYC,EAAqBC,EAAmBC,EAAgD,CACzG,IAAMC,EAAkBD,EAAmB,IAAI,CAAC,CAAE,MAAAE,EAAO,OAAAC,CAAO,IAAM,CACpE,IAAMC,EAAQL,EAAY,KAAK,IAAII,EAAQD,CAAK,EAChD,MAAO,CACL,MAAOA,EAAQE,EACf,OAAQD,EAASC,CACnB,CACF,CAAC,EAEKC,EAAYJ,EAAgB,OAElC,OAAUK,EAAK,IAAM,CACnB,IAAMC,EAA0B,CAACC,EAAeC,IAAqBC,GAAM,CAAIC,GAAK,CAAC,EAAE,EAAGH,EAAO,SAAS,EAAMG,GAAK,CAAC,EAAE,EAAGF,EAAO,SAAS,CAAC,EAAG,CAAC,EAAE,KAAK,EAAG,GAAG,EAAE,KAAK,EAG9JG,EAAa,CAACC,EAAkBC,IAAoD,CACxF,GAAM,CAAE,MAAAZ,EAAO,OAAAC,CAAO,EAAIF,EAAgBY,GAC1C,OAAOC,EAAKZ,EAAOC,CAAM,EAAI,KAAK,IAAID,EAAQC,CAAM,EAAI,EAAI,CAC9D,EAEMY,EAAeF,GAAqBD,EAAWC,EAAU,CAACG,EAAG,IAAMA,EAAI,CAAC,EACxEC,EAAeJ,GAAqBD,EAAWC,EAAU,CAACG,EAAG,IAAM,EAAIA,CAAC,EAa9E,OAXwBlB,EACrB,IAAOa,GAAK,CAACN,EAAW,GAAG,EAAGN,EAAW,SAAS,CAAC,EACnD,IAAOW,GAAM,MAAM,KAAK,MAAML,CAAS,EAAG,CAACa,EAAGL,IAAaN,EAC1DQ,EAAYF,CAAQ,EACpBI,EAAYJ,CAAQ,CACtB,CAAC,CAAC,CAAC,EACF,IAAOH,GAAM,MAAM,KAAK,MAAML,CAAS,EAAG,CAACa,EAAGL,IAAaN,EAC1DN,EAAgBY,GAAU,MAC1BZ,EAAgBY,GAAU,MAC5B,CAAC,CAAC,CAAC,CAGP,CAAC,CACH,CAEO,aAAaM,EAA8B,CAChD,OAAUb,EAAK,IAAM,CACnB,IAAMc,EAAM,KAAK,OAAOD,CAAK,EAC7B,OAAO,KAAK,YACVC,EACAD,EAAM,UACNA,EAAM,gBAAgB,IAAI,CAAC,CAAChB,EAAQD,CAAK,KAAO,CAAE,OAAAC,EAAQ,MAAAD,CAAM,EAAE,CACpE,CACF,CAAC,CACH,CAEA,MAAa,QAAQiB,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAME,GAAWF,CAAK,CAAC,CAClD,CAEA,MAAa,gBAAgBA,EAAgE,CAC3F,IAAMG,EAAW,MAAMD,GAAWF,CAAK,EACjCI,EAAqBjB,EACzB,IAASkB,GAAQ,KAAK,aAAaF,CAAQ,CAAC,CAC9C,EAEMG,EAAoB,MAAM,QAAQ,IAAIF,EAAgB,IAC1D,MAAOG,EAAgBb,IAAa,CAClC,IAAMc,EAAiB,MAAM,KAAKD,EAAe,SAAS,CAAC,EACrDE,EAAUD,EAAe,OAAO,CAACT,EAAGW,IAAMC,GAAOD,CAAC,CAAC,EACnDE,EAAUJ,EAAe,OAAO,CAACT,EAAGW,IAAM,CAACC,GAAOD,CAAC,CAAC,EAE1D,OAAO,IAAIG,GACT,MAAM,EAAE,EAAE,KAAK,CAAC,EAAE,IAAI,CAACd,EAAGW,IAAM,IAAII,GAAML,EAAQC,GAAcE,EAAQF,EAAY,CAAC,EACrF,CACE,OAAQP,EAAS,eAAeT,CAAQ,EACxC,MAAOS,EAAS,cAAcT,CAAQ,CACxC,CACF,CACF,CACF,CAAC,EAED,OAAAU,EAAgB,QAASW,GAAMA,EAAE,QAAQ,CAAC,EAEnCZ,EAAS,aAAeG,EAAyCA,EAAkB,EAC5F,CAEU,0BAAmC,CAC3C,MAAO,IACT,CACF,EC5FO,IAAMU,GAAN,cAAgCC,EAAkD,CACvF,YAAYC,EAA6C,IAAIC,GAAwB,CACnF,MAAM,oBAAqBD,CAAoB,CACjD,CAEU,qBAA8B,CACtC,MAAO,wBACT,CAEU,yBAAkC,CAC1C,MAAO,IACT,CACF,ECVO,SAASE,GACdC,EAC2E,CAC3E,IAAMC,EAAgC,CAAC,EAEjC,CACJ,yBAAAC,CACF,EAAIC,GAAkBH,EAAWC,CAAa,EAExCG,EAAS,CACb,OAAQF,EAAyB,SAAU,EAAI,EAC/C,OAAQA,EAAyB,QAAQ,EACzC,OAAQA,EAAyB,QAAQ,CAC3C,EAEA,OAAAG,GAA2BL,EAAWC,CAAa,EAE5C,CAAE,OAAAG,EAAQ,cAAAH,CAAc,CACjC,CCpBO,SAASK,GAAkBC,EAAkG,CAClI,IAAMC,EAAgC,CAAC,EAEjC,CACJ,eAAAC,EACA,oBAAAC,CACF,EAAIC,GAAsBJ,CAAO,EAE3B,CACJ,yBAAAK,CACF,EAAIC,GAAkBJ,EAAgBD,CAAa,EAE7CM,EAASF,EAAyB,EAAG,GAAI,SAAU,EAAI,EACvDG,EAASH,EAAyB,GAAI,GAAI,QAAQ,EAClDI,EAASJ,EAAyB,GAAI,IAAK,QAAQ,EAEzD,GAAIF,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,cAAAF,EACA,OAAQ,CAAE,OAAAM,EAAQ,OAAAC,EAAQ,OAAAC,CAAO,CACnC,CACF,CClBO,IAAMC,GAAN,cAAuCC,EAA+G,CAC3J,aAAc,CACZ,MAAM,0BAA0B,CAClC,CAEO,aAAaC,EAA8B,CAChD,GAAM,CAAE,OAAAC,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,wDAAwD,EAG1E,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAiBC,GAAKJ,EAAM,cAAc,IAAK,EAAI,EAAG,SAAS,EAE/DK,EAAaC,GAAUH,EADb,CAAC,QAAS,QAAS,OAAO,CACO,EAAE,IAAI,GAAG,EAEtDI,EAAMC,GAAYH,EAAYJ,EAAO,OAAQ,EAAI,EACrD,OAAAM,EAAMC,GAAYD,EAAKN,EAAO,MAAM,EACpCM,EAAMC,GAAYD,EAAKN,EAAO,MAAM,EACpCM,EAASE,GAAQF,EAAK,CAAC,GAAI,EAAE,EAAG,CAAC,EAAG,CAAC,EAAG,OAAO,EAExCA,CACT,CAAC,CACH,CAEA,MAAa,QAAQP,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAMU,GAAWV,CAAK,CAAC,CAClD,CAEU,qBAA8B,CACtC,MAAO,mCACT,CAEU,2BAA2BW,EAA8B,CACjE,OAAOC,GAA+BD,CAAS,CACjD,CAEU,cAAcE,EAAuB,CAC7C,OAAOC,GAAkBD,CAAO,CAClC,CACF,EC/CO,IAAME,GAAN,cAAoCC,EAAsD,CAC/F,YAAYC,EAAiD,IAAIC,GAA4B,CAC3F,MAAM,wBAAyBD,CAAoB,CACrD,CAEU,qBAA8B,CACtC,MAAO,6BACT,CAEU,yBAAkC,CAC1C,MAAO,IACT,CACF,ECZO,IAAME,GAAN,cAA8BC,EAAkB,CAAC,ECAjD,SAASC,GAAMC,EAAgBC,EAAuC,CAC3E,OAAUC,EAAOC,EAAIH,EAAGC,EAAO,OAAO,EAAGA,EAAO,MAAM,CACxD,CCDA,SAASG,GACPC,EACAC,EACAC,EACAC,EACAC,EAA4B,OACf,CACb,GAAM,CAAE,QAAAC,EAAS,KAAAC,CAAK,EAAIL,EAAO,KAE7BM,EAASC,GAAOR,EAAGK,EAASH,EAASE,CAAO,EAChD,OAAAG,EAASE,EAAIF,EAAKD,CAAI,EACtBC,EAAMG,GAAMH,EAAKN,EAAO,KAAK,EACtBE,EAAcQ,GAAKJ,CAAG,EAAIA,CACnC,CAEO,SAASK,GAAKZ,EAAgBC,EAAyB,CAC5D,OAAOF,GAAUC,EAAGC,EAAQ,CAAC,EAAG,CAAC,EAAG,EAAI,CAC1C,CAEO,SAASY,GAAWb,EAAgBC,EAAyB,CAClE,OAAOF,GAAUC,EAAGC,EAAQ,CAAC,EAAG,CAAC,EAAG,EAAK,CAC3C,CAEO,SAASa,GAASd,EAAgBC,EAAyB,CAChE,OAAOF,GAAUC,EAAGC,EAAQ,CAAC,EAAG,CAAC,EAAG,GAAM,OAAO,CACnD,CCxBA,SAASc,IAAkBC,EAAwCC,EAA+B,CAChG,SAASC,EAAoBC,EAAyBC,EAAoBC,EAAiC,CACzG,IAAMC,EAAUN,EAAeG,CAAe,EACxCI,EAAQD,EAAQ,QAAUF,EAAaC,EAAaA,GAE1D,GAAIG,GAAQD,CAAK,EACf,MAAM,IAAI,MAAM,+BAA+BA,sBAA0BD,EAAQ,uBAAuBF,kBAA2BC,GAAY,EAGjJ,OAAUI,EACR,IAASC,GACJC,GAASL,EAAS,CAACF,EAAYG,EAAOF,EAAYA,CAAU,CAAC,EAChE,CAAC,EAAG,EAAG,EAAG,CAAC,CACb,CACF,CACF,CAEA,SAASO,EACPT,EACAC,EACAC,EACAQ,EACY,CACZ,IAAMC,EAAUZ,EAAoBC,EAAiBC,EAAYC,CAAU,EACrEU,EAAUC,GAAShB,EAAeI,CAAU,CAAC,EAEnD,OAAAH,EAAc,KACZ,CAAE,UAAW,GAAGY,WAAuB,EACvC,CAAE,UAAW,GAAGA,QAAoB,CACtC,EAEO,CAAE,QAAAC,EAAS,KAAAC,CAAK,CACzB,CAEA,SAASE,EAAwBC,EAAoBL,EAAwC,CAC3F,IAAMP,EAAaU,GAAShB,EAAekB,CAAU,CAAC,EAChDC,EAAYH,GAAShB,EAAekB,CAAU,CAAC,EAErD,OAAAjB,EAAc,KACZ,CAAE,UAAW,GAAGY,WAAuB,EACvC,CAAE,UAAW,GAAGA,UAAsB,CACxC,EAEO,CACL,QAAAP,EACA,OAAAa,CACF,CACF,CAEA,SAASC,EACPjB,EACAC,EACAC,EACAQ,EACiB,CACjB,IAAMQ,EAAOT,EAAkBT,EAAiBC,EAAYC,EAAY,GAAGQ,QAAmB,EACxFS,EAAQL,EAAwBb,EAAY,GAAGS,SAAoB,EAEzE,MAAO,CAAE,KAAAQ,EAAM,MAAAC,CAAM,CACvB,CAEA,SAASC,EACPpB,EACAC,EACAC,EACAQ,EACAW,EAAS,GACY,CACrB,IAAMC,EAAQL,GAAwBI,EAAS,GAAM,GAAKrB,EAAiBC,EAAYC,EAAY,GAAGQ,SAAoB,EACpHa,EAAQN,EAAuBjB,EAAiBC,EAAYC,EAAY,GAAGQ,SAAoB,EAErG,MAAO,CAAE,MAAAY,EAAO,MAAAC,CAAM,CACxB,CAEA,MAAO,CACL,uBAAAN,EACA,2BAAAG,CACF,CACF,CAEO,SAASI,GAAcrB,EAA6E,CACzG,GAAM,CACJ,eAAAN,EACA,oBAAA4B,CACF,EAAIC,GAAsBvB,CAAO,EAE3BL,EAAgC,CAAC,EAEjC,CACJ,uBAAAmB,EACA,2BAAAG,CACF,EAAIxB,IAAkBC,EAAgBC,CAAa,EAE7C6B,EAAcV,EAAuB,KAAM,GAAI,EAAG,aAAa,EAC/DW,EAAWR,EAA2B,KAAM,GAAI,EAAG,UAAU,EAC7DS,EAAWT,EAA2B,KAAM,GAAI,EAAG,UAAU,EAC7DU,EAAWV,EAA2B,KAAM,GAAI,EAAG,UAAU,EAE7DW,EAAcX,EAA2B,MAAO,GAAI,EAAG,cAAe,EAAI,EAC1EY,EAAWZ,EAA2B,MAAO,GAAI,EAAG,UAAU,EAC9Da,EAAWb,EAA2B,MAAO,GAAI,EAAG,UAAU,EAC9Dc,EAAWd,EAA2B,MAAO,GAAI,EAAG,UAAU,EAE9De,EAAef,EAA2B,OAAQ,IAAK,EAAG,eAAgB,EAAI,EAC9EgB,EAAYhB,EAA2B,OAAQ,IAAK,EAAG,WAAW,EAClEiB,EAAYjB,EAA2B,OAAQ,IAAK,EAAG,WAAW,EAElEkB,EAAelB,EAA2B,OAAQ,IAAK,EAAG,eAAgB,EAAI,EAC9EmB,EAAYnB,EAA2B,OAAQ,IAAK,EAAG,WAAW,EAClEoB,EAAYpB,EAA2B,OAAQ,IAAK,EAAG,WAAW,EAClEqB,EAAmBrB,EAA2B,OAAQ,IAAK,EAAG,kBAAkB,EAEhFsB,EAAQpC,EACZ,IAASC,GAAaoC,GAAS9C,EAAe,IAAM,GAAG,EAAG,CAAC,IAAK,GAAG,CAAC,EAAG,CAAC,EAAG,CAAC,CAAC,CAC/E,EAGA,GAFAC,EAAc,KAAK,CAAE,UAAW,IAAK,CAAC,EAElC2B,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAsBlF,MAAO,CAAE,OAnBM,CACb,YAAAE,EACA,SAAAC,EACA,SAAAC,EACA,SAAAC,EACA,YAAAC,EACA,SAAAC,EACA,SAAAC,EACA,SAAAC,EACA,aAAAC,EACA,UAAAC,EACA,UAAAC,EACA,aAAAC,EACA,UAAAC,EACA,UAAAC,EACA,iBAAAC,EACA,GAAAC,CACF,EAEiB,cAAA5C,CAAc,CACjC,CC7IA,SAAS8C,IAAkBC,EAAgBC,EAA+B,CACxE,IAAMC,EAAqBC,GAA0BH,EAAWC,CAAa,EAE7E,SAASG,EAAwBC,EAAkC,CACjE,IAAMC,EAAUJ,EAAmB,GAAGG,kBAAwB,CAAC,EACzDE,EAASL,EAAmB,GAAGG,iBAAuB,CAAC,EAE7D,MAAO,CAAE,QAAAC,EAAS,OAAAC,CAAO,CAC3B,CAEA,SAASC,EAAuBH,EAAiC,CAC/D,IAAMI,EAAUP,EAAmB,GAAGG,iBAAuB,CAAC,EACxDK,EAAOR,EAAmB,GAAGG,cAAoB,CAAC,EAClDM,EAAQP,EAAwBC,CAAM,EAE5C,MAAO,CAAE,KAAM,CAAE,QAAAI,EAAS,KAAAC,CAAK,EAAG,MAAAC,CAAM,CAC1C,CAEA,SAASC,EAA2BP,EAAqC,CACvE,MAAO,CACL,MAAOG,EAAuB,GAAGH,SAAc,EAC/C,MAAOG,EAAuB,GAAGH,SAAc,CACjD,CACF,CAEA,MAAO,CACL,uBAAAG,EACA,2BAAAI,CACF,CACF,CAEO,SAASC,GACdb,EACsD,CACtD,IAAMC,EAAgC,CAAC,EAEjC,CACJ,uBAAAO,EACA,2BAAAI,CACF,EAAIb,IAAkBC,EAAWC,CAAa,EAExCa,EAAcN,EAAuB,aAAa,EAClDO,EAAWH,EAA2B,UAAU,EAChDI,EAAWJ,EAA2B,UAAU,EAChDK,EAAWL,EAA2B,UAAU,EAEhDM,EAAcN,EAA2B,aAAa,EACtDO,EAAWP,EAA2B,UAAU,EAChDQ,EAAWR,EAA2B,UAAU,EAChDS,EAAWT,EAA2B,UAAU,EAEhDU,EAAeV,EAA2B,cAAc,EACxDW,EAAYX,EAA2B,WAAW,EAClDY,EAAYZ,EAA2B,WAAW,EAElDa,EAAeb,EAA2B,cAAc,EACxDc,EAAYd,EAA2B,WAAW,EAClDe,EAAYf,EAA2B,WAAW,EAClDgB,EAAmBhB,EAA2B,kBAAkB,EAEhE,CAAE,GAAAiB,CAAG,EAAI7B,EAGf,GAFAC,EAAc,KAAK,CAAE,aAAc,KAAM,UAAW,IAAK,CAAC,EAEtD,CAAC6B,GAAWD,CAAE,EAChB,MAAM,IAAI,MAAM,yDAAyDA,GAAI,EAG/E,IAAME,EAAS,CACb,YAAAjB,EACA,SAAAC,EACA,SAAAC,EACA,SAAAC,EACA,YAAAC,EACA,SAAAC,EACA,SAAAC,EACA,SAAAC,EACA,aAAAC,EACA,UAAAC,EACA,UAAAC,EACA,aAAAC,EACA,UAAAC,EACA,UAAAC,EACA,iBAAAC,EACA,GAAAC,CACF,EAEA,OAAAG,GAA2BhC,EAAWC,CAAa,EAE5C,CAAE,OAAA8B,EAAQ,cAAA9B,CAAc,CACjC,CC1FO,SAASgC,GAASC,EAAgBC,EAA0C,CACjF,IAAIC,EAAMC,GAAKH,EAAGC,EAAO,KAAK,EAC9B,OAAAC,EAAME,GAAWF,EAAKD,EAAO,KAAK,EAClCC,EAASG,EAAIH,EAAKF,CAAC,EACnBE,EAASI,GAAKJ,CAAG,EACVA,CACT,CAEO,SAASK,GAAaP,EAAgBC,EAA0C,CACrF,IAAIC,EAAMM,GAASR,EAAGC,EAAO,KAAK,EAClCC,EAAME,GAAWF,EAAKD,EAAO,KAAK,EAElC,IAAIQ,EAAYC,GAAQV,EAAG,EAAG,EAAG,OAAO,EAClCW,EAAWC,GAAkBH,EAAO,KAAK,EACzCI,EAAQJ,EAAO,MAAM,KAAOP,EAAI,MAAM,GAG5C,GAFsBO,EAAO,MAAM,KAAOP,EAAI,MAAM,IAAMO,EAAO,MAAM,KAAOP,EAAI,MAAM,GAErE,CACjB,IAAMY,EAAY,CAAC,GAAGZ,EAAI,KAAK,EAC/BY,EAAU,GAAK,EACf,IAAMC,EAAYH,GAAkBE,CAAS,EAC7CZ,EAASc,GAAO,CAACd,EAAKa,CAAM,EAAG,CAAC,EAEhC,IAAME,EAAY,CAAC,GAAGf,EAAI,KAAK,EAC/Be,EAAU,GAAK,EACf,IAAMC,EAAYN,GAAkBK,CAAS,EAC7Cf,EAASc,GAAO,CAACd,EAAKgB,CAAM,EAAG,CAAC,CAClC,CAEA,OAAAT,EAASI,EAAWG,GAAO,CAACP,EAAQE,CAAK,EAAG,CAAC,EAAIF,EACjDP,EAASG,EAAII,EAAQP,CAAG,EAExBA,EAASI,GAAKJ,CAAG,EACVA,CACT,CC5BO,IAAMiB,GAAN,cAAiCC,EAAyB,CAC/D,aAAc,CACZ,MAAM,oBAAoB,CAC5B,CAEO,aAAaC,EAA8B,CAChD,GAAM,CAAE,OAAAC,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,kDAAkD,EAGpE,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAiBC,GAAKJ,EAAM,cAAc,IAAK,EAAI,EAAG,SAAS,EAG/DK,EAAaC,GAAUH,EADb,CAAC,QAAS,QAAS,OAAO,CACO,EAAE,IAAI,GAAG,EAEtDI,EAAMC,GAASH,EAAYJ,EAAO,WAAW,EACjDM,EAASE,GAAQF,EAAK,EAAG,EAAG,OAAO,EAEnCA,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EACnCM,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EACnCM,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EAEnCM,EAAMI,GAAaJ,EAAKN,EAAO,WAAW,EAC1CM,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EACnCM,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EACnCM,EAAMG,GAASH,EAAKN,EAAO,QAAQ,EAEnCM,EAAMI,GAAaJ,EAAKN,EAAO,YAAY,EAC3CM,EAAMG,GAASH,EAAKN,EAAO,SAAS,EACpCM,EAAMG,GAASH,EAAKN,EAAO,SAAS,EAEpCM,EAAMI,GAAaJ,EAAKN,EAAO,YAAY,EAC3CM,EAAMG,GAASH,EAAKN,EAAO,SAAS,EACpCM,EAAMG,GAASH,EAAKN,EAAO,SAAS,EACpCM,EAAMI,GAAaJ,EAAKN,EAAO,gBAAgB,EAE/C,IAAMW,EAAYL,EAAI,KAAK,CAAC,EAAG,CAAC,CAAC,EAGjC,OAF0BM,GAAOD,EAAWX,EAAO,EAAE,CAGvD,CAAC,CACH,CAEA,MAAa,QAAQD,EAAwC,CAC3D,OAAO,KAAK,aAAa,MAAMc,GAAWd,CAAK,CAAC,CAClD,CAEA,MAAa,sBAAsBA,EAAwD,CA7D7F,IAAAe,EA+DI,IAAIA,EAAAf,GAAA,YAAAA,EAAO,QAAP,MAAAe,EAAc,KAAMC,GAAQA,GAAO,GAAI,OAAO,IAAI,aAAa,GAAG,EACtE,IAAMC,EAAW,MAAMH,GAAWd,CAAK,EACjCkB,EAA2BhB,EAAK,IAASiB,GAAQ,KAAK,aAAaF,CAAQ,CAAC,CAAC,EAC7EG,EAA0B,MAAM,QAAQ,IAAIF,EAAsB,IAAKG,GAAMA,EAAE,KAAK,CAAC,CAAC,EAC5F,OAAAH,EAAsB,QAASG,GAAMA,EAAE,QAAQ,CAAC,EACzCJ,EAAS,aAAeG,EAA0BA,EAAwB,EACnF,CAEU,qBAA8B,CACtC,MAAO,wBACT,CAEU,2BAA2BE,EAA8B,CACjE,OAAOC,GAA2BD,CAAS,CAC7C,CAEU,cAAcE,EAAuB,CAC7C,OAAOC,GAAcD,CAAO,CAC9B,CACF,EC9EO,SAASE,IAAyBC,EAAuB,CAC9D,IAAMC,EAAM,IAAIC,GAChB,OAAAD,EAAI,eAAeD,CAAO,EACnBC,CACT,CCJO,SAASE,GAGdC,EACAC,EAC6B,CAE7B,MAAO,CAAE,GAAGD,EAAW,GADL,CAAE,WAAAC,CAAW,CACK,CACtC,CCRO,SAASC,IAAUC,EAA8B,CACtD,OAAO,OAAOA,EAAI,KAAQ,QAC5B,CAEO,SAASC,GAGdC,EACAC,EACkB,CAElB,MAAO,CAAE,GAAGD,EAAW,GADL,CAAE,IAAAC,CAAI,CACY,CACtC,CCRO,SAASC,IAAaC,EAAiC,CAC5D,OAAQA,EAAI,iBAA0BA,EAAI,oBACrCC,GAAmBD,EAAI,iBAAiB,CAC/C,CAEO,SAASE,GAGdC,EACAC,EACAC,EACqB,CAErB,MAAO,CAAE,GAAGF,EAAW,GADL,CAAE,OAAAC,EAAQ,kBAAAC,CAAkB,CACV,CACtC,CCjBA,SAASC,IAAkBC,EAAwCC,EAA+B,CAChG,SAASC,EAA2BC,EAAqBC,EAAuD,CAC9G,IAAMC,EAAaC,GAASN,EAAe,EAAQG,CAAW,EAAG,CAAC,EAAG,EAAGA,EAAa,CAAC,CAAC,EACjFI,EAAsBC,GAASR,EAAeG,CAAW,CAAC,EAC1DM,EAAuBD,GAASR,EAAeG,CAAW,CAAC,EAC3DO,EAAqBF,GAASR,EAAeG,CAAW,CAAC,EACzDQ,EAAyBH,GAASR,EAAeG,CAAW,CAAC,EAEnE,OAAAF,EAAc,KACZ,CAAE,UAAW,GAAGG,WAAuB,EACvC,CAAE,UAAW,GAAGA,oBAAgC,EAChD,CAAE,UAAW,GAAGA,qBAAiC,EACjD,CAAE,UAAW,GAAGA,mBAA+B,EAC/C,CAAE,UAAW,GAAGA,uBAAmC,CACrD,EAEO,CACL,QAAAC,EACA,iBAAAE,EACA,kBAAAE,EACA,gBAAAC,EACA,oBAAAC,CACF,CACF,CAEA,SAASC,EACPC,EACAC,EACAC,EACAX,EACAY,EACY,CACZ,IAAMX,EAAaC,GACjBN,EAAea,EAAaC,EAAcC,EAAaA,CAAU,EACjE,CAACA,EAAYA,EAAYF,EAAYC,CAAW,CAClD,EACMG,EAAUT,GAASR,EAAec,CAAW,CAAC,EAEpD,OAAAb,EAAc,KACZ,CAAE,UAAW,GAAGG,WAAuB,EACvC,CAAE,UAAW,GAAGA,KAAgBY,EAAkB,oBAAsB,QAAS,CACnF,EAEO,CAAE,QAAAX,EAAS,KAAAY,CAAK,CACzB,CAEA,SAASC,EACPL,EACAC,EACAC,EACAX,EACqB,CACrB,GAAM,CACJ,QAAAC,EACA,KAAAY,CACF,EAAIL,EAAkBC,EAAYC,EAAaC,EAAYX,EAAc,EAAI,EAE7E,MAAO,CACL,QAAAC,EACA,kBAAmBY,CACrB,CACF,CAEA,SAASE,EACPN,EACAC,EACAV,EAC4B,CAC5B,IAAMgB,EAAiBlB,EAA2BW,EAAY,GAAGT,kBAA6B,EACxFiB,EAAiBH,EAA2BL,EAAYC,EAAa,EAAG,GAAGV,kBAA6B,EAE9G,MAAO,CAAE,eAAAgB,EAAgB,eAAAC,CAAe,CAC1C,CAEA,SAASC,GAA+C,CACtD,IAAMC,EAASL,EAA2B,EAAG,GAAI,EAAG,oBAAoB,EAClEM,EAASL,EAAsB,GAAI,GAAI,oBAAoB,EAC3DM,EAASN,EAAsB,GAAI,IAAK,oBAAoB,EAC5DO,EAASP,EAAsB,IAAK,IAAK,oBAAoB,EAC7DQ,EAASR,EAAsB,IAAK,IAAK,oBAAoB,EAC7DS,EAAST,EAAsB,IAAK,IAAK,oBAAoB,EAC7DU,EAASV,EAAsB,IAAK,IAAK,oBAAoB,EAC7DW,EAASX,EAAsB,IAAK,IAAK,oBAAoB,EAC7DY,EAASZ,EAAsB,IAAK,IAAK,oBAAoB,EAC7Da,EAASb,EAAsB,IAAK,IAAK,oBAAoB,EAC7Dc,EAAUd,EAAsB,IAAK,IAAK,qBAAqB,EAC/De,EAAUf,EAAsB,IAAK,IAAK,qBAAqB,EAC/DgB,EAAUhB,EAAsB,IAAK,KAAM,qBAAqB,EAChEiB,EAAUjB,EAAsB,KAAM,KAAM,qBAAqB,EACvE,MAAO,CACL,OAAAI,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,QAAAC,EACA,QAAAC,EACA,QAAAC,EACA,QAAAC,CACF,CACF,CAEA,SAASC,GAAsD,CAC7D,IAAMd,EAASL,EAA2B,KAAM,IAAK,EAAG,yBAAyB,EAC3EM,EAASN,EAA2B,IAAK,IAAK,EAAG,yBAAyB,EAC1EO,EAASP,EAA2B,IAAK,IAAK,EAAG,yBAAyB,EAC1EQ,EAASR,EAA2B,IAAK,IAAK,EAAG,yBAAyB,EAC1ES,EAAST,EAA2B,IAAK,IAAK,EAAG,yBAAyB,EAC1EU,EAASV,EAA2B,IAAK,IAAK,EAAG,yBAAyB,EAC1EW,EAASX,EAA2B,IAAK,GAAI,EAAG,yBAAyB,EACzEY,EAASZ,EAA2B,GAAI,IAAK,EAAG,yBAAyB,EACzEoB,EAA2B1B,EAAkB,IAAK,GAAI,EAAG,yDAAyD,EAClH2B,EAAoB3B,EAAkB,IAAK,EAAG,EAAG,kDAAkD,EACnG4B,EAA2B5B,EAAkB,KAAM,GAAI,EAAG,yDAAyD,EACnH6B,EAAoB7B,EAAkB,KAAM,GAAI,EAAG,kDAAkD,EACrG8B,EAA2B9B,EAAkB,IAAK,GAAI,EAAG,yDAAyD,EAClH+B,EAAoB/B,EAAkB,IAAK,GAAI,EAAG,kDAAkD,EACpGgC,EAA2BhC,EAAkB,IAAK,GAAI,EAAG,yDAAyD,EAClHiC,EAAoBjC,EAAkB,IAAK,GAAI,EAAG,kDAAkD,EACpGkC,EAA2BlC,EAAkB,IAAK,GAAI,EAAG,yDAAyD,EAClHmC,EAAoBnC,EAAkB,IAAK,GAAI,EAAG,kDAAkD,EACpGoC,EAA2BpC,EAAkB,IAAK,GAAI,EAAG,yDAAyD,EAClHqC,EAAoBrC,EAAkB,IAAK,GAAI,EAAG,kDAAkD,EA0B1G,MAAO,CACL,OAAAW,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,OAAAC,EACA,gBAjCsB,CACtB,uBAAwBQ,EACxB,gBAAiBC,CACnB,EA+BE,gBA9BsB,CACtB,uBAAwBC,EACxB,gBAAiBC,CACnB,EA4BE,gBA3BsB,CACtB,uBAAwBC,EACxB,gBAAiBC,CACnB,EAyBE,gBAxBsB,CACtB,uBAAwBC,EACxB,gBAAiBC,CACnB,EAsBE,gBArBsB,CACtB,uBAAwBC,EACxB,gBAAiBC,CACnB,EAmBE,gBAlBsB,CACtB,uBAAwBC,EACxB,gBAAiBC,CACnB,CAgBA,CACF,CAEA,MAAO,CACL,yBAAA3B,EACA,6BAAAe,CACF,CACF,CAEO,SAASa,GAAcC,EAA6E,CACzG,IAAMlD,EAAgC,CAAC,EACjC,CACJ,eAAAD,EACA,oBAAAoD,CACF,EAAIC,GAAsBF,CAAO,EAC3B,CACJ,yBAAA7B,EACA,6BAAAe,CACF,EAAItC,IAAkBC,EAAgBC,CAAa,EAC7CqD,EAAchC,EAAyB,EACvCiC,EAAmBlB,EAA6B,EAKhDmB,EAAe,CACnB,UALmBC,GACnBzD,EAAe,KAAO,CAAC,EACvB,CAAC,EAAG,KAAM,CAAC,CACb,CAGA,EAEA,GADAC,EAAc,KAAK,CAAE,UAAW,wBAAyB,CAAC,EACtDmD,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAGlF,MAAO,CACL,OAAQ,CACN,YAAAE,EACA,iBAAAC,EACA,aAAAC,CACF,EACA,cAAAvD,CACF,CACF,CChNA,SAASyD,IAAkBC,EAAgBC,EAA+B,CACxE,IAAMC,EAAqBC,GAA0BH,EAAWC,CAAa,EAE7E,SAASG,EAA2BC,EAAgBC,EAAaC,EAA2C,CAC1G,IAAMC,EAAUN,EAAmB,GAAGG,YAAiBC,sBAAyB,EAAG,GAAGC,WAAsB,EACtGE,EAAoBP,EAAmB,GAAGG,YAAiBC,oCAAuC,EAAG,GAAGC,qBAAgC,EAC9I,MAAO,CAAE,QAAAC,EAAS,kBAAAC,CAAkB,CACtC,CAEA,SAASC,EAAsBJ,EAAyC,CACtE,IAAMC,EAAe,oBAAoBD,IACnCK,EAAsB,sBAAsBL,cAC5CM,EAA4B,GAAGL,mBAC/BM,EAA4B,GAAGN,mBAE/BC,EAAUN,EAAmB,GAAGS,sBAAyC,EAAG,GAAGC,WAAmC,EAClHE,EAAmBZ,EAAmB,GAAGS,oBAAuC,EAAG,GAAGC,oBAA4C,EAClIH,EAAoBP,EAAmB,GAAGS,mBAAsC,EAAG,GAAGC,qBAA6C,EACnIG,EAAkBb,EAAmB,GAAGS,0BAA6C,EAAG,GAAGC,mBAA2C,EACtII,EAAsBd,EAAmB,GAAGS,8BAAiD,EAAG,GAAGC,uBAA+C,EAExJ,MAAO,CACL,eAAgB,CACd,QAAAJ,EACA,iBAAAM,EACA,kBAAAL,EACA,gBAAAM,EACA,oBAAAC,CACF,EACA,eAAgBZ,EAA2B,cAAeE,EAAKO,CAAyB,CAC1F,CACF,CAEA,SAASI,GAA+C,CACtD,MAAO,CACL,OAAQb,EAA2B,cAAe,EAAG,oBAAoB,EACzE,OAAQM,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,OAAQA,EAAsB,CAAC,EAC/B,QAASA,EAAsB,EAAE,EACjC,QAASA,EAAsB,EAAE,EACjC,QAASA,EAAsB,EAAE,EACjC,QAASA,EAAsB,EAAE,CACnC,CACF,CAEA,SAASQ,EAAkBb,EAAgBE,EAAkC,CAC3E,IAAMC,EAAUN,EAAmB,GAAGG,YAAkB,EAAG,GAAGE,WAAsB,EAC9EY,EAAOjB,EAAmB,GAAGG,WAAiB,EAAG,GAAGE,QAAmB,EAC7E,MAAO,CAAE,QAAAC,EAAS,KAAAW,CAAK,CACzB,CAEA,SAASC,EAA0Bd,EAAkC,CACnE,IAAMe,EAAyBH,EAC7B,2BAA2BZ,yBAC3B,kCAAkCA,0BACpC,EACMgB,EAAkBJ,EACtB,2BAA2BZ,mBAC3B,kCAAkCA,mBACpC,EACA,MAAO,CAAE,uBAAAe,EAAwB,gBAAAC,CAAgB,CACnD,CAEA,SAASC,GAAsD,CAC7D,MAAO,CACL,OAAQnB,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,OAAQA,EAA2B,aAAc,EAAG,yBAAyB,EAC7E,gBAAiBgB,EAA0B,CAAC,EAC5C,gBAAiBA,EAA0B,CAAC,EAC5C,gBAAiBA,EAA0B,CAAC,EAC5C,gBAAiBA,EAA0B,CAAC,EAC5C,gBAAiBA,EAA0B,CAAC,EAC5C,gBAAiBA,EAA0B,CAAC,CAC9C,CACF,CAEA,MAAO,CACL,yBAAAH,EACA,6BAAAM,CACF,CACF,CAEO,SAASC,GACdxB,EACsD,CACtD,IAAMC,EAAgC,CAAC,EACjC,CACJ,yBAAAgB,EACA,6BAAAM,CACF,EAAIxB,IAAkBC,EAAWC,CAAa,EACxCwB,EAAYzB,EAAU,oBAE5B,GADAC,EAAc,KAAK,CAAE,aAAc,mBAAoB,UAAW,wBAAyB,CAAC,EACxF,CAACyB,GAAWD,CAAS,EACvB,MAAM,IAAI,MAAM,yEAAyEA,GAAW,EAGtG,IAAME,EAAS,CACb,YAAaV,EAAyB,EACtC,iBAAkBM,EAA6B,EAC/C,aAAc,CACZ,UAAAE,CACF,CACF,EAEA,OAAAG,GAA2B5B,EAAWC,CAAa,EAC5C,CAAE,OAAA0B,EAAQ,cAAA1B,CAAc,CACjC,CCzHO,SAAS4B,GAAmBC,EAAgBC,EAA6BC,EAA2B,CACzG,OAAUC,EAAK,IAAM,CACnB,IAAIC,EAASC,GAAOL,EAAGC,EAAO,QAASC,EAAS,MAAM,EACtD,OAAAE,EAASE,EAAIF,EAAKH,EAAO,iBAAiB,EAChCM,GAAYH,EAAK,EAAG,CAAC,CACjC,CAAC,CACH,CCLA,IAAMI,IAAU,qBAEhB,SAASC,IAAmBC,EAAgBC,EAAyCC,EAA2B,CAC9G,OAAUC,EAAK,IAAM,CACnB,IAAIC,EAASC,GAAgBL,EAAGC,EAAO,QAASC,EAAS,MAAM,EAC/D,OAAAE,EAASE,GACPF,EACAH,EAAO,gBACPA,EAAO,oBACPA,EAAO,kBACPA,EAAO,iBACPH,GACF,EACUS,GAAYH,EAAK,EAAG,CAAC,CACjC,CAAC,CACH,CAEA,SAASI,IAAsBC,EAAoC,CACjE,MAAO,CAAC,EAAG,EAAG,EAAG,EAAE,EAAE,KAAMC,GAAQA,IAAQD,CAAQ,EAAI,CAAC,EAAG,CAAC,EAAI,CAAC,EAAG,CAAC,CACvE,CAEO,SAASE,GAAYX,EAAgBC,EAA4B,CACtE,OAAUE,EAAK,IAAM,CACnB,IAAIS,EACAR,EAAMS,GAAmBb,EAAGC,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EA0BrD,GAxBuB,CACrBA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,OACPA,EAAO,QACPA,EAAO,QACPA,EAAO,QACPA,EAAO,OACT,EAEe,QAAQ,CAACa,EAAO,IAAM,CACnC,IAAML,EAAW,EAAI,EACfM,EAAuBP,IAAsBC,CAAQ,EAC3DL,EAAML,IAAmBK,EAAKU,EAAM,eAAgBC,CAAoB,EACxEX,EAAMS,GAAmBT,EAAKU,EAAM,eAAgB,CAAC,EAAG,CAAC,CAAC,EACtDL,IAAa,KAAIG,EAASR,EAChC,CAAC,EAEGQ,IAAW,KACb,MAAM,IAAI,MAAM,+CAA+C,EAGjE,MAAO,CACL,IAAAR,EACA,OAAQQ,CACV,CACF,CAAC,CACH,CC9DA,SAASI,IAAIC,EAAoBC,EAAWC,EAAW,CACrD,IAAMC,EAAYH,EAAM,UAAU,EAC5BI,EAAQ,KAAK,IAAID,EAAUF,GAAG,GAAIE,EAAUF,GAAG,EAAE,EACjDI,EAAQ,KAAK,IAAIF,EAAUF,GAAG,GAAIE,EAAUF,GAAG,EAAE,EACjDK,EAAQ,KAAK,IAAIH,EAAUF,GAAG,GAAIE,EAAUF,GAAG,EAAE,EACjDM,EAAQ,KAAK,IAAIJ,EAAUF,GAAG,GAAIE,EAAUF,GAAG,EAAE,EACjDO,EAAQ,KAAK,IAAIL,EAAUD,GAAG,GAAIC,EAAUD,GAAG,EAAE,EACjDO,EAAQ,KAAK,IAAIN,EAAUD,GAAG,GAAIC,EAAUD,GAAG,EAAE,EACjDQ,EAAQ,KAAK,IAAIP,EAAUD,GAAG,GAAIC,EAAUD,GAAG,EAAE,EACjDS,EAAQ,KAAK,IAAIR,EAAUD,GAAG,GAAIC,EAAUD,GAAG,EAAE,EACjDU,GAASN,EAAQF,IAAUG,EAAQF,GACnCQ,GAASH,EAAQF,IAAUG,EAAQF,GACzC,GAAIG,GAAS,GAAKC,GAAS,EAAG,MAAO,GACrC,IAAMC,EAAmB,KAAK,IAAIV,EAAOI,CAAK,EACxCO,EAAmB,KAAK,IAAIV,EAAOI,CAAK,EACxCO,EAAmB,KAAK,IAAIV,EAAOI,CAAK,EACxCO,EAAmB,KAAK,IAAIV,EAAOI,CAAK,EACxCO,EAAmB,KAAK,IAAIF,EAAmBF,EAAkB,CAAG,EAAI,KAAK,IAAIG,EAAmBF,EAAkB,CAAG,EAC/H,OAAOG,GAAoBN,EAAQC,EAAQK,EAC7C,CAEO,SAASC,GACdnB,EACAoB,EACAC,EACAC,EACAC,EACU,CACV,IAAMC,EAAWxB,EAAM,MAAM,GACvByB,EAAa,KAAK,IAAIJ,EAAeG,CAAQ,EAE7CE,EAAaN,EAChB,IAAI,CAACO,EAAOC,KAAc,CAAE,MAAAD,EAAO,SAAAC,CAAS,EAAE,EAC9C,OAAQC,GAAMA,EAAE,MAAQN,CAAc,EACtC,KAAK,CAACO,EAAIC,IAAOA,EAAG,MAAQD,EAAG,KAAK,EAEjCE,EAAgBC,GAAeA,GAAKX,EAAe,EAAI,EACvDY,EAAqB,CAAC,EAE5B,OAAAR,EAAW,QAASG,GAAM,CACxB,GAAIK,EAAS,QAAUT,EAAY,OACnC,IAAMU,EAAgBN,EAAE,MACxB,QAAS3B,EAAIgC,EAAS,OAAS,EAAGhC,GAAK,EAAG,EAAEA,EAAG,CAC7C,IAAMkC,EAAMrC,IAAIC,EAAO6B,EAAE,SAAUK,EAAShC,EAAE,EAC9C,GAAIkC,IAAQ,IACZP,EAAE,OAASG,EAAaI,CAAG,EACvBP,EAAE,OAASN,GAAgB,KACjC,CACIY,IAAkBN,EAAE,OACtBK,EAAS,KAAKL,EAAE,QAAQ,CAE5B,CAAC,EACMK,CACT,CCnDA,SAASG,IAAkCC,EAAgB,CACzD,IAAMC,EAASC,GAAWC,GAAUH,EAAG,CAAC,EAAG,CAAC,CAAC,CAAC,EAExCI,EAAQ,CACTC,GAAIJ,EAAI,GAAIA,EAAI,EAAE,EAClBI,GAAIJ,EAAI,GAAIA,EAAI,EAAE,CACvB,EACMK,EAAU,CACXC,EAAIN,EAAI,GAAOO,GAAIJ,EAAM,GAAI,CAAC,CAAC,EAC/BG,EAAIN,EAAI,GAAOO,GAAIJ,EAAM,GAAI,CAAC,CAAC,CACpC,EACA,MAAO,CAAE,MAAAA,EAAO,QAAAE,CAAQ,CAC1B,CAEA,SAASG,IAAiBC,EAAiBC,EAAiB,CAC1D,GAAM,CAAE,MAAAP,EAAO,QAAAE,CAAQ,EAAIP,IAAkCW,CAAE,EAEzDT,EAASC,GAAWC,GAAUQ,EAAI,CAAC,EAAG,CAAC,CAAC,CAAC,EACzCC,EAAcJ,GAAOK,EAAOC,GAAON,GAAIP,EAAI,GAAI,CAAC,CAAC,EAAGG,EAAM,EAAE,EAAG,CAAC,EAChEW,EAAcR,EAAOM,EAAOL,GAAIP,EAAI,GAAI,EAAE,EAAGG,EAAM,EAAE,EAAGE,EAAQ,EAAE,EAClEU,EAAcR,GAAOK,EAAOC,GAAON,GAAIP,EAAI,GAAI,CAAC,CAAC,EAAGG,EAAM,EAAE,EAAG,CAAC,EAChEa,EAAcV,EAAOM,EAAOL,GAAIP,EAAI,GAAI,EAAE,EAAGG,EAAM,EAAE,EAAGE,EAAQ,EAAE,EAExE,OAAUH,GACLe,GAAM,CACJb,GAAIU,EAAUH,CAAQ,EACtBP,GAAIY,EAAUD,CAAQ,EACtBT,EAAIQ,EAAUH,CAAQ,EACtBL,EAAIU,EAAUD,CAAQ,CAC3B,CAAC,EACD,CAAC,EAAG,CAAC,CACP,CACF,CAEO,SAASG,GAAYC,EAA6BC,EAA+BC,EAA2B,CACjH,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAYJ,EAAe,MAAM,GAEnCK,EAAQhB,IACPiB,EAAWC,GAAKL,EAAO,UAAW,CAACE,EAAW,EAAG,CAAC,CAAC,EAAG,CAAC,GAAI,CAAC,CAAC,EAC7DE,EAAQN,EAAgB,CAAC,GAAI,CAAC,CAAC,CACpC,EACAK,EAAWC,EAAQD,EAAO,CAACD,EAAYC,EAAM,MAAM,GAAKD,EAAY,CAAC,CAAC,EAEtE,IAAMI,EAAsBC,GAAWC,GAAMT,EAAkB,CAAC,EAAG,EAAG,CAAC,EAAG,CAAC,GAAI,GAAI,EAAE,CAAC,CAAC,EACnFU,EAAYD,GAAMF,EAAkB,CAAC,EAAG,EAAG,CAAC,EAAG,CAAC,GAAI,GAAI,CAAC,CAAC,EAE9DG,EAAYL,EAAQK,EAAQ,CAACP,EAAWO,EAAO,MAAM,EAAY,CAAC,EAElE,IAAMC,EAAkB9B,GAAQuB,CAAK,EAC/BQ,EAAmB/B,GAAQ6B,CAAM,EAEvC,MAAO,CAAE,MAAOC,EAAc,OAAQC,CAAc,CACtD,CAAC,CACH,CCrDO,SAASC,GACdC,EACAC,EACA,CACA,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAYH,EAAE,MAAM,GACpBI,EAA2BC,EAC/BC,GAAUN,EAAGC,EAAO,sBAAsB,EAC1C,CAACE,EAAW,GAAI,EAAG,CAAC,CACtB,EACMI,EAAqBF,EACzBC,GAAUN,EAAGC,EAAO,eAAe,EACnC,CAACE,EAAW,GAAI,CAAC,CACnB,EACA,MAAO,CAAE,sBAAAC,EAAuB,gBAAAG,CAAgB,CAClD,CAAC,CACH,CCfO,SAASC,GACdC,EACAC,EACAC,EACA,CACA,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAQC,GAAmBL,EAAGE,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACnDI,EAAQD,GAAmBD,EAAOF,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDK,EAAQF,GAAmBC,EAAOJ,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDM,EAAQH,GAAmBE,EAAOL,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDO,EAAQJ,GAAmBG,EAAON,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDQ,EAAQL,GAAmBI,EAAOP,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDS,EAAQN,GAAmBK,EAAOR,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EACvDU,EAAQP,GAAmBM,EAAOT,EAAO,OAAQ,CAAC,EAAG,CAAC,CAAC,EAEvDW,EAAiBC,GAAmBb,EAAQC,EAAO,eAAe,EAClEa,EAAiBD,GAAmBd,EAAGE,EAAO,eAAe,EAC7Dc,EAAiBF,GAAmBR,EAAOJ,EAAO,eAAe,EACjEe,EAAiBH,GAAmBN,EAAON,EAAO,eAAe,EACjEgB,EAAiBJ,GAAmBJ,EAAOR,EAAO,eAAe,EACjEiB,EAAiBL,GAAmBF,EAAOV,EAAO,eAAe,EAEjEkB,EAAoBC,GAAO,CAC/BR,EAAe,sBACfE,EAAe,sBACfC,EAAe,sBACfC,EAAe,sBACfC,EAAe,sBACfC,EAAe,qBACjB,EAAG,CAAC,EAEEG,EAAsBD,GAAO,CACjCR,EAAe,gBACfE,EAAe,gBACfC,EAAe,gBACfC,EAAe,gBACfC,EAAe,gBACfC,EAAe,eACjB,EAAG,CAAC,EAEJ,MAAO,CACL,eAAAC,EACA,iBAAAE,CACF,CACF,CAAC,CACH,CC9CO,IAAMC,GAAN,KAA4B,CAOjC,YAAY,CAAE,cAAAC,EAAe,WAAAC,CAAW,EAA4B,CAAC,EAAG,CANxE,KAAU,MAAQ,wBAUhB,GAHA,KAAK,eAAiBD,GAAiB,GACvC,KAAK,YAAcC,GAAc,IAE7B,OAAO,KAAK,gBAAmB,UAAY,KAAK,gBAAkB,GAAK,KAAK,gBAAkB,EAChG,MAAM,IAAI,MAAM,GAAG,KAAK,+DAA+D,EAGzF,GAAI,OAAO,KAAK,aAAgB,SAC9B,MAAM,IAAI,MAAM,GAAG,KAAK,4CAA4C,CAExE,CAEA,IAAI,eAAwB,CAAE,OAAO,KAAK,cAAgB,CAE1D,IAAI,YAAqB,CAAE,OAAO,KAAK,WAAa,CACtD,ECbO,IAAMC,GAAN,cAA6BC,EAAyB,CAC3D,aAAc,CACZ,MAAM,gBAAgB,CACxB,CAEO,aAAaC,EAAiB,CACnC,GAAM,CAAE,OAAAC,CAAO,EAAI,KACnB,GAAI,CAACA,EAAQ,MAAM,IAAI,MAAM,8CAA8C,EAC3E,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAAiBC,GAAKJ,EAAM,cAAc,IAAK,EAAK,EAAG,SAAS,EAChEK,EAAOC,GAAOC,GAAIJ,EAAa,KAAK,EAAG,CAAC,EACxCK,EAAWC,GAAYJ,EAAGJ,EAAO,WAAW,EAC5C,CAAE,eAAAS,EAAgB,iBAAAC,CAAiB,EAAIC,GAAgBJ,EAAS,IAAKA,EAAS,OAAQP,EAAO,gBAAgB,EACnH,OAAOY,GAAYH,EAAgBC,EAAkBV,EAAO,YAAY,CAC1E,CAAC,CACH,CAEA,MAAa,QAAQD,EAAkB,CACrC,OAAO,KAAK,aAAa,MAAMc,GAAWd,CAAK,CAAC,CAClD,CAEA,MAAa,YAAYA,EAAkBe,EAAkC,CAAC,EAA6B,CACzG,GAAM,CAAE,WAAAC,EAAY,cAAAC,CAAc,EAAI,IAAIC,GAAsBH,CAAO,EACjEI,EAAW,MAAML,GAAWd,CAAK,EACjC,CAAE,MAAOoB,EAAQ,OAAQC,CAAQ,EAAI,KAAK,aAAaF,CAAQ,EAC/DG,EAAQF,EAAO,GACfG,EAASF,EAAQ,GACvB,QAASG,EAAI,EAAGA,EAAIJ,EAAO,OAAQI,IACjCJ,EAAOI,GAAG,QAAQ,EAClBH,EAAQG,GAAG,QAAQ,EAErB,IAAMC,EAAa,MAAM,KAAKF,EAAO,SAAS,CAAC,EAEzCG,EAAUC,GAAkBL,EAAOG,EAAwBT,EAD5C,GACsEC,CAAa,EAClGW,EAAeT,EAAS,2BAA2B,CAAC,EACpDU,EAAYV,EAAS,UACrBW,EAAOD,EAAYD,EAAa,MAChCG,EAAOF,EAAYD,EAAa,OAChCI,EAAYV,EAAM,UAAU,EAC5BW,EAAUP,EACb,IAAKQ,GAAQ,CACZ,GAAM,CAACC,EAAKC,CAAM,EAAI,CACpB,KAAK,IAAI,EAAGJ,EAAUE,GAAK,EAAE,EAC7B,KAAK,IAAI,EAAKF,EAAUE,GAAK,EAAE,CACjC,EAAE,IAAKG,GAAQA,EAAMN,CAAI,EACnB,CAACO,EAAMC,CAAK,EAAI,CACpB,KAAK,IAAI,EAAGP,EAAUE,GAAK,EAAE,EAC7B,KAAK,IAAI,EAAKF,EAAUE,GAAK,EAAE,CACjC,EAAE,IAAKG,GAAQA,EAAMP,CAAI,EACzB,OAAO,IAAIU,GACTf,EAAWS,GACX,IAAIO,GAAKH,EAAMH,EAAKI,EAAQD,EAAMF,EAASD,CAAG,EAC9C,CAAE,OAAQhB,EAAS,eAAe,CAAC,EAAG,MAAOA,EAAS,cAAc,CAAC,CAAE,CACzE,CACF,CAAC,EACH,OAAAG,EAAM,QAAQ,EACdC,EAAO,QAAQ,EACRU,CACT,CAEU,qBAA8B,CACtC,MAAO,uBACT,CAEU,2BAA2BS,EAA8B,CACjE,OAAOC,GAA2BD,CAAS,CAC7C,CAEU,cAAcE,EAAuB,CAC7C,OAAOC,GAAcD,CAAO,CAC9B,CACF,ECjFO,SAASE,IAAqBC,EAAuB,CAC1D,IAAMC,EAAM,IAAIC,GAChB,OAAAD,EAAI,eAAeD,CAAO,EACnBC,CACT,CAEO,SAASE,IAAuBH,EAAuB,CAC5D,OAAOD,IAAqBC,CAAO,CACrC,CAGO,IAAMI,GAAN,cAA+BF,EAAe,CAAC,ECd/C,IAAMG,GAAgB,GAEhBC,GAAc,CACzB,IAAIC,GAAM,QAAU,OAAQ,EAC5B,IAAIA,GAAM,QAAS,OAAO,EAC1B,IAAIA,GAAM,QAAS,OAAO,EAC1B,IAAIA,GAAM,OAAQ,OAAO,EACzB,IAAIA,GAAM,QAAS,OAAO,CAC5B,EAEaC,GAAwB,CACnC,IAAID,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,CAC9B,EAEaE,GAA+C,CAAC,QAAS,QAAS,MAAM,EAExEC,GAAqB,oBACrBC,GAAoC,mCCVjD,IAAMC,GAAYC,GAAa,OAAOA,GAAQ,SAEvC,SAASC,GAAeC,EAAa,CAC1C,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,mBAAmBA,GAAQ,EAG7C,GAAI,OAAOA,EAAO,oBAAuB,UACvC,MAAM,IAAI,MAAM,wDAAwDA,EAAO,oBAAoB,EAGrG,GAAI,CAACH,GAASG,EAAO,YAAY,GAAKA,EAAO,aAAe,GAAKA,EAAO,aAAe,EACrF,MAAM,IAAI,MAAM,gEAAgEA,EAAO,cAAc,EAGvG,GACE,CAAC,MAAM,QAAQA,EAAO,OAAO,GAC1B,CAACA,EAAO,QAAQ,QAChB,CAACA,EAAO,QAAQ,MAAOC,GAAW,OAAOA,GAAM,QAAQ,EAE1D,MAAM,IAAI,MAAM,kEAAkE,KAAK,UAAUD,EAAO,OAAO,GAAG,EAGpH,GACE,CAAC,MAAM,QAAQA,EAAO,OAAO,GAC1B,CAACA,EAAO,QAAQ,QAChB,CAACA,EAAO,QAAQ,IAAKE,GAAWA,GAAK,CAAC,CAAC,EAAE,MAAOA,GAAWL,GAASK,EAAE,CAAC,GAAKL,GAASK,EAAE,CAAC,CAAC,EAE5F,MAAM,IAAI,MAAM,wEAAwE,KAAK,UAAUF,EAAO,OAAO,GAAG,EAG1H,GAAIA,EAAO,UACT,CAAC,MAAM,QAAQA,EAAO,OAAO,GAC1BA,EAAO,QAAQ,SAAW,GAC1B,CAACA,EAAO,QAAQ,MAAMH,EAAQ,GAEjC,MAAM,IAAI,MAAM,8EAA8E,KAAK,UAAUG,EAAO,OAAO,GAAG,CAElI,CCjDO,SAASG,GAAMC,EAA6B,CACjD,OAAUC,EAAK,IAAM,CACnB,IAAMC,EAASC,EAAIH,EAAMI,GAAO,kBAAmB,CAAC,EACpD,OAAUC,EAAOC,GAAQC,GAAIP,EAAGE,CAAG,CAAC,EAAGA,CAAG,CAC5C,CAAC,CACH,CCFO,SAASM,GAAkBC,EAAgBC,EAAwC,CACxF,OAAUC,EAAK,IAAM,CACnB,IAAIC,EAASC,GAAIJ,EAAG,CAAC,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,CAAC,CAAC,EACpD,OAAAG,EAASE,GAAOF,EAAKF,EAAO,KAAK,QAAS,CAAC,EAAG,CAAC,EAAG,OAAO,EACzDE,EAASG,GAAIH,EAAKF,EAAO,GAAG,GAAG,EAC/BE,EAASI,EAAIJ,EAAKF,EAAO,GAAG,OAAO,EACnCE,EAASK,EAAIL,EAAKF,EAAO,KAAK,IAAI,EAC3BQ,GAAMN,CAAG,CAClB,CAAC,CACH,CCTO,SAASO,GAAuBC,EAAgBC,EAA0C,CAC/F,OAAUC,EAAK,IAAM,CACnB,IAAIC,EAASC,GAAIJ,EAAG,CAAC,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,CAAC,CAAC,EACpD,OAAAG,EAASE,GAAgBF,EAAKF,EAAO,iBAAkBA,EAAO,iBAAkB,CAAC,EAAG,CAAC,EAAG,OAAO,EAC/FE,EAASG,EAAIH,EAAKF,EAAO,IAAI,EACtBM,GAAMJ,CAAG,CAClB,CAAC,CACH,CCHA,SAASK,IAAkBC,EAAwCC,EAA+B,CAChG,IAAMC,EAAoBC,GAAyBH,EAAgBC,CAAa,EAEhF,SAASG,EAAuBC,EAAcC,EAAiC,CAC7E,IAAMC,EAASC,GAASR,EAAeK,CAAI,CAAC,EACtCI,EAAaD,GAASR,EAAeK,CAAI,CAAC,EAEhD,OAAAJ,EAAc,KACZ,CAAE,UAAW,GAAGK,OAAmB,EACnC,CAAE,UAAW,GAAGA,WAAuB,CACzC,EACO,CAAE,IAAAC,EAAK,QAAAE,CAAQ,CACxB,CAEA,SAASC,EAA+BC,EAAoBC,EAAqBN,EAAyC,CACxH,IAAMO,EAAOX,EAAkBS,EAAYC,EAAa,EAAG,GAAGN,QAAmB,EAC3EQ,EAAKV,EAAuBQ,EAAa,GAAGN,MAAiB,EACnE,MAAO,CAAE,KAAAO,EAAM,GAAAC,CAAG,CACpB,CACA,IAAMC,EAA6BC,GAAkChB,EAAgBC,CAAa,EAElG,MAAO,CACL,kBAAAC,EACA,+BAAAQ,EACA,2BAAAK,CACF,CACF,CAEO,SAASE,GACdC,EACAC,EACAC,EACAC,EACgE,CAChE,GAAM,CACJ,eAAArB,EACA,oBAAAsB,CACF,EAAIC,GAAsBL,CAAO,EAE3BjB,EAAgC,CAAC,EACjC,CACJ,kBAAAC,EACA,+BAAAQ,EACA,2BAAAK,CACF,EAAIhB,IAAkBC,EAAgBC,CAAa,EAC/CuB,EAEJ,GAAIL,EAAO,mBAAoB,CAC7B,GAAM,CAACM,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,CAAE,EAAIZ,EACvCa,EAAQf,EAAO,mBACjBjB,EAAkBuB,EAAIC,EAAI,EAAG,OAAO,EACpCX,EAA2BU,EAAIC,EAAI,OAAO,EACxCS,EAAQpB,EAA2BW,EAAIC,EAAI,OAAO,EAClDS,EAAQrB,EAA2BY,EAAIC,EAAI,OAAO,EAClDS,EAAQtB,EAA2Ba,EAAIC,EAAI,OAAO,EAClDS,EAAQvB,EAA2Bc,EAAIC,EAAI,OAAO,EAClDS,EAAQxB,EAA2Be,EAAIC,EAAI,OAAO,EAClDS,EAAQR,EAAKjB,EAA2BgB,EAAIC,EAAI,OAAO,EAAI,OAC3DS,EAAQR,EAAKlB,EAA2BiB,EAAIC,EAAI,OAAO,EAAI,OAC3DS,EAAQxC,EAAkB+B,GAAMD,GAAMD,EAAI,EAAIX,EAAiB,EAAG,OAAO,EAC/EI,EAAS,CACP,MAAAU,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,CAC1D,CACF,KAAO,CACL,GAAM,CAACjB,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,EAAIC,CAAE,EAAIZ,EACvCa,EAAQxB,EAA+Be,EAAIC,EAAI,OAAO,EACtDS,EAAQzB,EAA+BgB,EAAIC,EAAI,OAAO,EACtDS,EAAQ1B,EAA+BiB,EAAIC,EAAI,OAAO,EACtDS,EAAQ3B,EAA+BkB,EAAIC,EAAI,OAAO,EACtDS,EAAQ5B,EAA+BmB,EAAIC,EAAI,OAAO,EACtDS,EAAQ7B,EAA+BoB,EAAIC,EAAI,OAAO,EACtDS,EAAQ9B,EAA+BqB,EAAIC,EAAI,OAAO,EACtDS,EAAQ/B,EAA+BsB,EAAIC,EAAI,OAAO,EACtDS,EAAQxC,EAAkB+B,EAAI,EAAIb,EAAiB,EAAG,OAAO,EACnEI,EAAS,CACP,MAAAU,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,EAAO,MAAAC,CAC1D,CACF,CACA,GAAIpB,EAAoB,EAAE,SAAW,EACnC,MAAM,IAAI,MAAM,kCAAkCA,EAAoB,EAAE,QAAQ,EAElF,MAAO,CAAE,OAAAE,EAAQ,cAAAvB,CAAc,CACjC,CCjFA,SAAS0C,IAAkBC,EAAgBC,EAA+B,CACxE,IAAMC,EAAqBC,GAA0BH,EAAWC,CAAa,EAE7E,SAASG,EAAuBC,EAA2B,CACzD,IAAMC,EAAMJ,EAAmB,GAAGG,QAAc,CAAC,EAC3CE,EAAUL,EAAmB,GAAGG,YAAkB,CAAC,EACzD,MAAO,CAAE,IAAAC,EAAK,QAAAC,CAAQ,CACxB,CAEA,SAASC,EAAkBH,EAA4B,CACrD,IAAMI,EAAUP,EAAmB,GAAGG,YAAkB,CAAC,EACnDK,EAAOR,EAAmB,GAAGG,SAAe,CAAC,EACnD,MAAO,CAAE,QAAAI,EAAS,KAAAC,CAAK,CACzB,CAEA,SAASC,EAA+BN,EAAmC,CACzE,IAAMO,EAAOJ,EAAkB,GAAGH,QAAa,EACzCQ,EAAKT,EAAuB,GAAGC,MAAW,EAChD,MAAO,CAAE,KAAAO,EAAM,GAAAC,CAAG,CACpB,CAEA,IAAMC,EAA6BC,GAA+Bb,CAAkB,EACpF,MAAO,CACL,kBAAAM,EACA,+BAAAG,EACA,2BAAAG,CACF,CACF,CAEO,SAASE,GACdhB,EACAiB,EACgE,CAChE,IAAMhB,EAAgC,CAAC,EAEjC,CACJ,kBAAAO,EACA,+BAAAG,EACA,2BAAAG,CACF,EAAIf,IAAkBC,EAAWC,CAAa,EAE1CiB,EAEJ,GAAID,EAAO,mBAAoB,CAE7B,IAAME,EAAcF,EAAO,aAAeA,EAAO,YAAY,QAAU,EACvEC,EAAS,CACP,MAAOD,EAAO,mBAAqBT,EAAkB,OAAO,EAAIM,EAA2B,OAAO,EAClG,MAAOA,EAA2B,OAAO,EACzC,MAAOA,EAA2B,OAAO,EACzC,MAAOA,EAA2B,OAAO,EACzC,MAAOA,EAA2B,OAAO,EACzC,MAAOA,EAA2B,OAAO,EACzC,MAAOK,EAAa,EAAIL,EAA2B,OAAO,EAAI,OAC9D,MAAOK,EAAa,EAAIL,EAA2B,OAAO,EAAI,OAC9D,MAAON,EAAkB,OAAO,CAClC,CACF,MACEU,EAAS,CACP,MAAOP,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOA,EAA+B,OAAO,EAC7C,MAAOH,EAAkB,OAAO,CAClC,EAGF,OAAAY,GAA2BpB,EAAWC,CAAa,EAC5C,CAAE,OAAAiB,EAAQ,cAAAjB,CAAc,CACjC,CC9EO,IAAMoB,GAAN,KAAwB,CAO7B,YAAY,CAAE,UAAAC,EAAW,eAAAC,CAAe,EAAwB,CAAC,EAAG,CANpE,KAAU,MAAQ,oBAUhB,GAHA,KAAK,WAAaD,GAAa,IAC/B,KAAK,gBAAkBC,GAAkB,GAErC,OAAO,KAAK,YAAe,UAAY,KAAK,WAAa,KAAO,EAClE,MAAM,IAAI,MAAM,GAAG,KAAK,2DAA2D,EAGrF,GAAI,OAAO,KAAK,iBAAoB,UAAY,KAAK,iBAAmB,GAAK,KAAK,iBAAmB,EACnG,MAAM,IAAI,MAAM,GAAG,KAAK,gEAAgE,CAE5F,CAEA,IAAI,WAAoB,CAAE,OAAO,KAAK,UAAY,CAElD,IAAI,gBAAyB,CAAE,OAAO,KAAK,eAAiB,CAC9D,ECLO,IAAMC,GAAN,cAA6BC,EAAmC,CAKrE,YAAYC,EAA0B,CACpC,MAAM,YAAY,EAClBC,GAAeD,CAAM,EACrB,KAAK,QAAUA,CACjB,CAEA,IAAW,QAA2B,CACpC,OAAO,KAAK,OACd,CAEA,IAAW,iBAA2B,CACpC,OAAO,KAAK,OAAO,iBAAmB,KAAK,OAAO,QAAQ,OAAS,CACrE,CAEA,IAAW,iBAA0B,CACnC,MAAO,IAAK,KAAK,gBAAkB,KAAK,OAAO,QAAQ,OAAS,EAClE,CAEO,cAAcE,EAAgBC,EAAiD,CACpF,IAAIC,EAAMC,GAAkBH,EAAGC,EAAO,KAAK,EAC3C,OAAAC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EACzCC,EAAMC,GAAkBD,EAAKD,EAAO,KAAK,EAClCI,GAAUH,EAAKD,EAAO,MAAO,QAAS,EAAK,CACpD,CAEO,aAAaD,EAAgBC,EAAsC,CACxE,IAAIC,EAAM,KAAK,OAAO,mBAClBI,GAAMD,GAAUL,EAAGC,EAAO,MAAqB,QAAS,EAAK,CAAC,EAC9DM,GAAuBP,EAAGC,EAAO,KAA4B,EACjE,OAAAC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMK,GAAuBL,EAAKD,EAAO,KAAK,EAC9CC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMK,GAAuBL,EAAKD,EAAO,KAAK,EAC9CC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMK,GAAuBL,EAAKD,EAAO,KAAK,EAC9CC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMK,GAAuBL,EAAKD,EAAO,KAAK,EAC9CC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMK,GAAuBL,EAAKD,EAAO,KAAK,EAC9CC,EAASE,GAAQF,EAAK,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,CAAC,EAAG,MAAM,EAC5CA,EAAMD,EAAO,MAAQM,GAAuBL,EAAKD,EAAO,KAAK,EAAIC,EACjEA,EAAMD,EAAO,MAAQM,GAAuBL,EAAKD,EAAO,KAAK,EAAIC,EAC1DG,GAAUH,EAAKD,EAAO,MAAO,QAAS,EAAK,CACpD,CAEO,aAAaO,EAAiBC,EAAgC,CACnE,GAAM,CAAE,OAAAR,CAAO,EAAI,KAEnB,GAAI,CAACA,EACH,MAAM,IAAI,MAAM,0CAA0C,EAG5D,OAAUS,EAAK,IAAM,CACnB,IAAIC,EAAiBC,GAAKJ,EAAM,cAAcC,EAAW,EAAK,EAAG,SAAS,EAC1E,OAAAE,EAAc,KAAK,OAAO,QACtBE,GAAUF,EAAa,KAAK,OAAO,OAAO,EAC1CA,EACJA,EAAcA,EAAY,IAAI,GAAG,EAC1B,KAAK,OAAO,mBACf,KAAK,aAAaA,EAAaV,CAAyB,EACxD,KAAK,cAAcU,EAAaV,CAAoC,CAC1E,CAAC,CACH,CAEA,MAAa,QAAQO,EAAkBC,EAAyC,CAC9E,OAAO,KAAK,aAAa,MAAMK,GAAWN,CAAK,EAAGC,CAAS,CAC7D,CAEA,MAAa,OAAOD,EAAkBO,EAAoC,CAAC,EAA+B,CACxG,GAAM,CAAE,UAAAN,EAAW,eAAAO,CAAe,EAAI,IAAIC,GAAkBF,CAAa,EACnEG,EAAW,MAAMJ,GAAWN,CAAK,EACjCN,EAAM,MAAM,KAAK,aAAagB,EAAUT,CAAS,EACjDU,EAAUT,EAAK,IAASU,GAAQlB,CAAG,EAAE,GAAG,WAAW,CAAC,EACpDmB,EAAkB,CACtB,MAAOH,EAAS,cAAc,CAAC,EAC/B,OAAQA,EAAS,eAAe,CAAC,CACnC,EAEMI,EAAU,MAAM,KAAK,aAAaH,EAAMD,EAAS,2BAA2B,CAAC,EAAGF,CAAc,EACpGd,EAAI,QAAQ,EACZiB,EAAK,QAAQ,EAEb,IAAMI,EAAQD,EAAQ,IAAKE,GAAQA,EAAI,GAAG,EACpCC,EAASH,EAAQ,IAAKE,GAAQA,EAAI,KAAK,EACvCE,EAAcJ,EAAQ,IAAKE,GAAQA,EAAI,UAAU,EACjDG,EAAaL,EAAQ,IAAKE,GAAQ,KAAK,OAAO,QAAQA,EAAI,MAAM,EAgBtE,OAdgBI,GACdL,EAAM,IAAKM,GAAQA,EAAI,QAAQpB,CAAS,CAAC,EACzCgB,EACA,KAAK,OAAO,aACZ,EACF,EAE2B,IAAKK,GAAQ,IAAIC,GAC1CN,EAAOK,GACPJ,EAAYI,GACZH,EAAWG,GACXP,EAAMO,GACNT,CACF,CAAC,CAEH,CAEU,qBAA8B,CACtC,MAAO,EACT,CAEU,2BAA2BW,EAA8B,CACjE,OAAOC,GAA2BD,EAAW,KAAK,MAAM,CAC1D,CAEU,cAAcE,EAAuB,CAC7C,IAAMC,EAAc,KAAK,OAAO,aAAevC,GAAe,qBAExDwC,EAAaD,EAAcA,EAAY,OAAS,OACtD,GAAIC,IAAe,GAAKA,IAAe,GAAKA,IAAe,EACzD,MAAM,IAAI,MAAM,oEAAoEA,yBAAkC,EAExH,OAAOC,GAAcH,EAAS,KAAK,OAAQ,KAAK,gBAAiBC,CAAW,CAC9E,CAEA,MAAgB,aACdG,EACAC,EACAvB,EACA,CACA,GAAM,CAAE,MAAAwB,EAAO,OAAAC,CAAO,EAAIF,EACpB9B,EAAY,KAAK,IAAI+B,EAAOC,CAAM,EAClCC,EAAoBjC,EAAY+B,EAChCG,EAAoBlC,EAAYgC,EAEhCG,EAAWN,EAAa,MAAM,GAC9BO,EAAW,KAAK,OAAO,QAAQ,OAE/B,CAACC,EAAaC,EAAcC,CAAiB,EAAOtC,EAAK,IAAM,CACnE,IAAMuC,EAAWX,EAAa,QAAQ,CAACM,EAAUA,EAAUC,EAAU,KAAK,eAAe,CAAC,EAEpFtB,EAAQ0B,EAAS,MAAM,CAAC,EAAG,EAAG,EAAG,CAAC,EAAG,CAACL,EAAUA,EAAUC,EAAU,CAAC,CAAC,EACtEpB,EAASwB,EAAS,MAAM,CAAC,EAAG,EAAG,EAAG,CAAC,EAAG,CAACL,EAAUA,EAAUC,EAAU,CAAC,CAAC,EACvEnB,EAAc,KAAK,gBAClBwB,GAAQD,EAAS,MAAM,CAAC,EAAG,EAAG,EAAG,CAAC,EAAG,CAACL,EAAUA,EAAUC,EAAU,KAAK,OAAO,QAAQ,MAAM,CAAC,EAAG,CAAC,EACnGM,GAAO,CAAC,EACf,MAAO,CAAC5B,EAAOE,EAAQC,CAAW,CACpC,CAAC,EAEKJ,EAAU,CAAC,EACX8B,EAAa,MAAML,EAAa,MAAM,EACtCM,EAAY,MAAMP,EAAY,MAAM,EAC1C,QAASQ,EAAM,EAAGA,EAAMV,EAAUU,IAChC,QAASC,EAAM,EAAGA,EAAMX,EAAUW,IAChC,QAASC,EAAS,EAAGA,EAASX,EAAUW,IAAU,CAChD,IAAMC,EAAQC,GAAQN,EAAWE,GAAKC,GAAKC,GAAQ,EAAE,EACrD,GAAI,CAACxC,GAAkByC,EAAQzC,EAAgB,CAC7C,IAAM2C,GAAQJ,EAAMG,GAAQL,EAAUC,GAAKC,GAAKC,GAAQ,EAAE,GAAKZ,EAAYF,EACrEkB,GAAQN,EAAMI,GAAQL,EAAUC,GAAKC,GAAKC,GAAQ,EAAE,GAAKZ,EAAYD,EACrEkB,EAAe,KAAK,IAAIR,EAAUC,GAAKC,GAAKC,GAAQ,EAAE,EAAI,KAAK,OAAO,QAAQA,GAAQ,EAAKZ,EAAYF,EACvGoB,EAAgB,KAAK,IAAIT,EAAUC,GAAKC,GAAKC,GAAQ,EAAE,EAAI,KAAK,OAAO,QAAQA,GAAQ,EAAKZ,EAAYD,EACxG3C,EAAK2D,EAAOE,EAAa,EACzBE,EAAKH,EAAOE,EAAc,EAC1BE,EAAM,CAAE,IAAAV,EAAK,IAAAC,EAAK,OAAAC,CAAO,EACzB,CAAE,WAAAS,EAAY,MAAAC,CAAM,EAAI,KAAK,gBAC/B,MAAM,KAAK,sBAAsBlB,EAAkCgB,CAAG,EACtE,CAAE,WAAY,EAAG,MAAO,CAAE,EAC9B1C,EAAQ,KAAK,CACX,IAAK,IAAI6C,GAAYnE,EAAG+D,EAAG/D,EAAI6D,EAAYE,EAAID,CAAW,EAC1D,MAAAL,EACA,WAAYA,EAAQQ,EACpB,MAAAC,EACA,GAAGF,CACL,CAAC,CACH,CACF,CAIJ,OAAAlB,EAAY,QAAQ,EACpBC,EAAa,QAAQ,EACrBC,EAAkB,QAAQ,EACnB1B,CACT,CAEA,MAAc,sBAAsB8C,EAA4BJ,EAAmD,CACjH,GAAM,CAAE,IAAAV,EAAK,IAAAC,EAAK,OAAAC,CAAO,EAAIQ,EACvBK,EAAc,MAAMD,EAAc,MAAM,EAC9C,OAAO,MAAM,KAAK,OAAO,QAAQ,MAAM,EAAE,KAAK,CAAC,EAC5C,IAAI,CAACE,EAAGC,IAAMF,EAAYf,GAAKC,GAAKC,GAAQe,EAAE,EAC9C,IAAI,CAACN,EAAYC,KAAW,CAC3B,WAAAD,EACA,MAAAC,CACF,EAAE,EACD,OAAO,CAACM,EAAKC,IAAUD,EAAI,WAAaC,EAAK,WAAaD,EAAMC,CAAK,CAC1E,CACF,EAjNaC,GAAN9E,GAAM8E,GACG,qBAAuB,CAAC,EAAG,GAAI,GAAI,GAAI,IAAK,IAAK,IAAK,KAAM,IAAI,ECPzE,IAAMC,GAAN,cAAyBC,EAAe,CAC7C,YAAYC,EAAqB,GAAM,CACrC,IAAMC,EAAS,CACb,mBAAAD,EACA,aAAcE,GACd,QAAS,CAAC,MAAM,EAChB,GAAIF,EACA,CACA,QAASG,GACT,QAASC,EACX,EACE,CACA,QAASC,GACT,gBAAiB,EACnB,CACJ,EAEA,MAAMJ,CAAM,CACd,CAEA,IAAW,oBAA8B,CACvC,OAAO,KAAK,OAAO,kBACrB,CAEA,IAAW,SAAmB,CAC5B,OAAO,KAAK,OAAO,OACrB,CAEA,MAAa,YAAYK,EAAkBC,EAA6D,CAEtG,OADyB,MAAM,KAAK,OAAOD,EAAOC,CAAa,GACvC,IAAKC,GAAQ,IAAIC,GAAcD,EAAI,MAAOA,EAAI,YAAa,CAAE,MAAOA,EAAI,WAAY,OAAQA,EAAI,WAAY,CAAC,CAAC,CACxI,CAEmB,qBAA8B,CAC/C,OAAO,KAAK,mBAAqBE,GAAoCC,EACvE,CAEmB,2BAA2BC,EAA8F,CAC1I,OAAO,MAAM,2BAA2BA,CAAS,CACnD,CACF,EClDO,SAASC,IAAiBC,EAAuBC,EAAqB,GAAM,CACjF,IAAMC,EAAM,IAAIC,GAAWF,CAAkB,EAC7C,OAAAC,EAAI,eAAeF,CAAO,EACnBE,CACT,CCPO,IAAME,GAAN,cAAsCC,EAAkB,CAAxD,kCACL,KAAmB,MAAQ,0BAC7B,ECNO,IAAMC,GAAN,KAAwB,CAE7B,MAAa,KAAKC,EAA2D,CAC3E,OAAOA,EAAY,MAAM,KAAK,IAAI,CAAC,CACrC,CAEA,MAAa,KAAkB,CAC7B,MAAM,IAAI,MAAM,yCAAyC,CAC3D,CACF,ECFA,eAAsBC,GACpBC,EACAC,EAEAC,EACAC,EAEAC,EAAwF,CAAC,CAAE,YAAAC,CAAY,IAAMA,EAC7G,CACA,IAAMC,EAAYN,EAAc,IAAKO,GAAkBC,GAAoBD,CAAY,EACnFH,EAAoBG,CAAY,EAChCA,EAAa,SAAU,EACrBE,EAAgDN,IACpDF,aAAoBS,GAChB,MAAMC,GAAmBV,EAAOK,CAAS,EACzC,MAAMM,GAAaX,EAAOK,CAAS,GAEnCO,EAAU,MAAMX,EAAeO,CAAK,EAC1C,OAAAA,EAAM,QAASK,GAAMA,aAAgBJ,IAAUI,EAAE,QAAQ,CAAC,EACnDD,CACT,CAEA,eAAsBE,GACpBR,EACAN,EAEAe,EACAb,EAEAC,EACA,CACA,OAAOL,GACL,CAACQ,CAAY,EACbN,EACA,MAAOQ,GAAUO,EAAcP,EAAM,EAAE,EACvCN,EACAC,CACF,CACF,CC3CO,IAAMa,GAAgB,GAEhBC,GAAc,CACzB,IAAIC,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,EAC5B,IAAIA,GAAM,SAAU,QAAQ,CAC9B,EAEaC,GAAqC,CAAC,QAAS,QAAS,MAAM,ECFpE,IAAMC,GAAN,cAA+BC,EAAe,CACnD,aAAc,CACZ,IAAMC,EAAS,CACb,mBAAoB,GACpB,aAAcC,GACd,QAAS,CAAC,MAAM,EAChB,QAASC,GACT,QAASC,GACT,mBAAoB,GACpB,YAAa,CAAC,EAAG,GAAI,GAAI,GAAI,IAAK,IAAK,GAAG,CAC5C,EAEA,MAAMH,CAAM,CACd,CAEA,IAAW,SAAmB,CAC5B,OAAO,KAAK,OAAO,OACrB,CAEA,MAAa,YAAYI,EAAkBC,EAA6D,CAEtG,OADyB,MAAM,KAAK,OAAOD,EAAOC,CAAa,GACvC,IAAKC,GAAQ,IAAIC,GAAcD,EAAI,MAAOA,EAAI,YAAa,CAAE,MAAOA,EAAI,WAAY,OAAQA,EAAI,WAAY,CAAC,CAAC,CACxI,CAEmB,qBAA8B,CAC/C,MAAO,0BACT,CAEmB,2BAA2BE,EAA8F,CAC1I,OAAO,MAAM,2BAA2BA,CAAS,CACnD,CACF,ECzBO,IAAMC,GAAO,CAClB,eAAgB,IAAIC,GACpB,iBAAkB,IAAIC,GACtB,WAAY,IAAIC,GAChB,kBAAmB,IAAIC,GACvB,sBAAuB,IAAIC,GAC3B,mBAAoB,IAAIC,GACxB,kBAAmB,IAAIC,GACvB,aAAc,IAAIC,EACpB,EASaC,IAAiB,CAACC,EAAkBC,IAA6DX,GAAK,eAAe,YAAYU,EAAOC,CAAO,EAS/IC,IAAmB,CAACF,EAAkBC,IAA+DX,GAAK,iBAAiB,YAAYU,EAAOC,CAAO,EASrJE,IAAa,CAACH,EAAkBC,IAA0DX,GAAK,WAAW,YAAYU,EAAOC,CAAO,EASpIG,IAAuBJ,GAAmEV,GAAK,kBAAkB,gBAAgBU,CAAK,EAWtIK,IAA2BL,GAAmEV,GAAK,sBAAsB,gBAAgBU,CAAK,EAY9IM,IAAyBN,GAA6DV,GAAK,mBAAmB,sBAAsBU,CAAK,EASzIO,IAA4BP,GAAmEV,GAAK,kBAAkB,mBAAmBU,CAAK,EAS9IQ,IAAuBR,GAAiFV,GAAK,aAAa,oBAAoBU,CAAK,EAEnJS,IAA2BC,GAAgBpB,GAAK,eAAe,KAAKoB,CAAG,EACvEC,IAA6BD,GAAgBpB,GAAK,iBAAiB,KAAKoB,CAAG,EAC3EE,IAAuBF,GAAgBpB,GAAK,WAAW,KAAKoB,CAAG,EAC/DG,IAAyBH,GAAgBpB,GAAK,kBAAkB,KAAKoB,CAAG,EACxEI,IAA6BJ,GAAgBpB,GAAK,sBAAsB,KAAKoB,CAAG,EAChFK,IAA4BL,GAAgBpB,GAAK,mBAAmB,KAAKoB,CAAG,EAC5EM,IAA2BN,GAAgBpB,GAAK,kBAAkB,KAAKoB,CAAG,EAC1EO,IAAsBP,GAAgBpB,GAAK,aAAa,KAAKoB,CAAG,EAGhEQ,IAAyBT,IACzBU,IAAcpB,IACdqB,IAAkBhB,ICtGxB,IAAMiB,GAAN,cAAqEC,EAAwB,CAClG,YAEYC,EAEAC,EAEAC,EACV,CACA,MAAM,EANI,gBAAAF,EAEA,WAAAC,EAEA,oBAAAC,CAGZ,CACF,EAEaC,GAAN,cAAmFL,EAA0E,CAClK,MAAsB,KAA+C,CACnE,IAAMM,EAAgB,MAAM,KAAK,WAE3BC,EAAwB,MAAMC,GAClCF,EACA,KAAK,MACL,MAAOG,GAAU,QAAQ,IACvBA,EAAM,IAAKC,GAASC,GAAK,kBAAkB,mBAAmBD,CAAI,CAA6B,CACjG,EACA,KAAK,cACP,EAEA,OAAOJ,EAAc,IACnB,CAACM,EAAcC,IAAMC,GAAmCF,EAAcL,EAAsBM,EAAE,CAChG,CACF,CAEA,kBAAmB,CACjB,OAAO,IAAIE,GAA2B,KAAM,KAAK,KAAK,CACxD,CACF,EAEaC,GAAN,cAAsFhB,EAA8F,CACzL,MAAsB,KAAyD,CAC7E,IAAMY,EAAe,MAAM,KAAK,WAChC,GAAI,CAACA,EACH,OAGF,IAAMK,EAAkB,MAAMC,GAC5BN,EACA,KAAK,MACJF,GAASC,GAAK,kBAAkB,mBAAmBD,CAAI,EACxD,KAAK,cACP,EAEA,OAAOI,GAA0BF,EAAcK,CAAe,CAChE,CAEA,kBAAmB,CACjB,OAAO,IAAIE,GAA8B,KAAM,KAAK,KAAK,CAC3D,CACF,EAEaC,GAAN,cAAuHf,EAAuC,CAC1J,kBAAmB,CAC1B,OAAO,IAAIgB,GAA4C,KAAM,KAAK,KAAK,CACzE,CAEA,qBAAsB,CACpB,OAAO,IAAIC,GAA8B,KAAM,KAAK,KAAK,CAC3D,CACF,EAEaC,GAAN,cAA0HP,EAA0C,CAChK,kBAAmB,CAC1B,OAAO,IAAIQ,GAA+C,KAAM,KAAK,KAAK,CAC5E,CAEA,oBAAqB,CACnB,OAAO,IAAIC,GAAgC,KAAM,KAAK,KAAK,CAC7D,CACF,EC3EO,IAAMC,GAAN,cAAkEC,EAAwB,CAC/F,YAEYC,EAEAC,EAEAC,EACV,CACA,MAAM,EANI,gBAAAF,EAEA,WAAAC,EAEA,oBAAAC,CAGZ,CACF,EAEaC,GAAN,cAAgFL,EAAuE,CAC5J,MAAsB,KAA+C,CACnE,IAAMM,EAAgB,MAAM,KAAK,WAC3BC,EAAqB,MAAMC,GAC/BF,EACA,KAAK,MACL,MAAOG,GAAU,QAAQ,IAAIA,EAAM,IAAKC,GAASC,GAAK,aAAa,oBAAoBD,CAAI,CAAoC,CAAC,EAChI,KAAK,cACP,EACA,OAAOJ,EAAc,IAAI,CAACM,EAAcC,IAAM,CAC5C,GAAM,CAAE,IAAAC,EAAK,OAAAC,EAAQ,kBAAAC,CAAkB,EAAIT,EAAmBM,GAC9D,OAAOI,GAAcC,GAAiBN,EAAcG,EAAQC,CAAiB,EAAGF,CAAG,CACrF,CAAC,CACH,CAEA,qBAAsB,CACpB,OAAO,IAAIK,GAA8B,KAAM,KAAK,KAAK,CAC3D,CACF,EAEaC,GAAN,cAAmFpB,EAA2F,CACnL,MAAsB,KAAyD,CAC7E,IAAMY,EAAe,MAAM,KAAK,WAChC,GAAI,CAACA,EAAc,OACnB,GAAM,CAAE,IAAAE,EAAK,OAAAC,EAAQ,kBAAAC,CAAkB,EAAI,MAAMK,GAC/CT,EACA,KAAK,MACJF,GAASC,GAAK,aAAa,oBAAoBD,CAAI,EACpD,KAAK,cACP,EACA,OAAOO,GAAcC,GAAiBN,EAAcG,EAAQC,CAAiB,EAAGF,CAAG,CACrF,CAEA,qBAAsB,CACpB,OAAO,IAAIQ,GAAiC,KAAM,KAAK,KAAK,CAC9D,CACF,EAEaC,GAAN,cAAoHlB,EAAoC,CACpJ,qBAAsB,CAC7B,OAAO,IAAImB,GAA+C,KAAM,KAAK,KAAK,CAC5E,CAEA,qBAAsB,CACpB,OAAO,IAAIC,GAA8B,KAAM,KAAK,KAAK,CAC3D,CACF,EAEaC,GAAN,cAAuHN,EAAuC,CAC1J,qBAAsB,CAC7B,OAAO,IAAIO,GAAkD,KAAM,KAAK,KAAK,CAC/E,CAEA,oBAAqB,CACnB,OAAO,IAAIC,GAAgC,KAAM,KAAK,KAAK,CAC7D,CACF,ECzEO,IAAMC,GAAN,cAAqEC,EAAwB,CAClG,YAEYC,EAEAC,EACV,CACA,MAAM,EAJI,gBAAAD,EAEA,WAAAC,CAGZ,CACF,EAEaC,GAAN,cAAsGJ,EAAyE,CACpL,MAAsB,KAA8C,CAClE,IAAMK,EAAgB,MAAM,KAAK,WAQjC,OAPoB,MAAMC,GACxBD,EACA,KAAK,MACJE,GAAU,QAAQ,IAAIA,EAAM,IAAKC,GAASC,GAAK,mBAAmB,sBAAsBD,CAAI,CAA0B,CAAC,EACxH,KACCE,GAAiBA,EAAa,UAAU,MAAM,KAAM,CAAE,iBAAkB,EAAK,CAAC,CACjF,GACmB,IAAI,CAACC,EAAYC,IAAMC,GAAkCR,EAAcO,GAAID,CAAU,CAAC,CAC3G,CAEA,qBAAsB,CACpB,OAAO,IAAIG,GAA+C,KAAM,KAAK,KAAK,CAC5E,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GAA4C,KAAM,KAAK,KAAK,CACzE,CACF,EAEaC,GAAN,cAAwGhB,EAA6F,CAC1M,MAAsB,KAAwD,CAC5E,IAAMU,EAAe,MAAM,KAAK,WAChC,GAAI,CAACA,EAAc,OACnB,IAAMC,EAAa,MAAMM,GACvBP,EACA,KAAK,MACJF,GAASC,GAAK,mBAAmB,sBAAsBD,CAAI,EAC5D,KAECE,GAAiBA,EAAa,UAAU,MAAM,KAAM,CAAE,iBAAkB,EAAK,CAAC,CACjF,EACA,OAAOG,GAAyBH,EAAcC,CAAU,CAC1D,CAEA,qBAAsB,CACpB,OAAO,IAAIO,GAAkD,KAAM,KAAK,KAAK,CAC/E,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GAA+C,KAAM,KAAK,KAAK,CAC5E,CACF,ECnDO,IAAMC,GAAN,cAAkEC,EAAwB,CAC/F,YAEYC,EAEAC,EAEAC,EACV,CACA,MAAM,EANI,gBAAAF,EAEA,WAAAC,EAEA,wBAAAC,CAGZ,CAEA,IAAc,aAAyD,CACrE,OAAO,KAAK,mBACRC,GAAK,sBACLA,GAAK,iBACX,CACF,EAEaC,GAAN,cAAgFN,EAAqE,CAC1J,MAAsB,KAA6C,CACjE,IAAMO,EAAgB,MAAM,KAAK,WAC3BC,EAAaD,EAAc,IAAKE,GAAQA,EAAI,SAAS,EACrDC,EAAgD,KAAK,iBAAoBC,GAC3E,MAAMC,GAAmB,KAAK,MAAOJ,CAAU,EAC/C,MAAMK,GAAa,KAAK,MAAOL,CAAU,EACvCM,EAAsB,MAAM,QAAQ,IAAIJ,EAAM,IAAKK,GAAS,KAAK,YAAY,gBAAgBA,CAAI,CAAC,CAAC,EACzG,OAAAL,EAAM,QAASM,GAAMA,aAAgBL,IAAUK,EAAE,QAAQ,CAAC,EAC3CT,EACZ,OAAO,CAACU,EAAeC,IAAMJ,EAAoBI,EAAE,EACnD,IAAI,CAACC,EAAcD,IAAME,GAAiCD,EAAcL,EAAoBI,EAAE,CAAC,CAEpG,CAEA,qBAAsB,CACpB,OAAO,IAAIG,GAA+C,KAAM,KAAK,KAAK,CAC5E,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GAA4C,KAAM,KAAK,KAAK,CACzE,CAEA,qBAAsB,CACpB,OAAO,IAAIC,GAA8B,KAAM,KAAK,KAAK,CAC3D,CACF,EAEaC,GAAN,cAAmFxB,EAAyF,CACjL,MAAsB,KAAuD,CAC3E,IAAMmB,EAAe,MAAM,KAAK,WAChC,GAAI,CAACA,EACH,OAEF,GAAM,CAAE,UAAAM,CAAU,EAAIN,EAChBT,EAAgD,KAAK,iBAAoBC,GAC3E,MAAMC,GAAmB,KAAK,MAAO,CAACa,CAAS,CAAC,EAChD,MAAMZ,GAAa,KAAK,MAAO,CAACY,CAAS,CAAC,EACxCC,EAAY,MAAM,KAAK,YAAY,gBAAgBhB,EAAM,EAAE,EACjE,OAAAA,EAAM,QAASM,GAAMA,aAAgBL,IAAUK,EAAE,QAAQ,CAAC,EACnDI,GAAiCD,EAAcO,CAAS,CACjE,CAEA,qBAAsB,CACpB,OAAO,IAAIC,GAAkD,KAAM,KAAK,KAAK,CAC/E,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GAA+C,KAAM,KAAK,KAAK,CAC5E,CAEA,oBAAqB,CACnB,OAAO,IAAIC,GAAgC,KAAM,KAAK,KAAK,CAC7D,CACF,EC1EO,IAAMC,GAAN,cAA2CC,EAAwB,CAExE,YAAsBC,EAA4BC,EAAgC,IAAIC,GAAyB,CAC7G,MAAM,EADc,WAAAF,EAA4B,aAAAC,CAElD,CACF,EAEaE,GAAN,cAAiCL,EAAqC,CAC3E,MAAsB,KAAgC,CACpD,GAAM,CAAE,MAAAE,EAAO,QAAAC,CAAQ,EAAI,KACvBG,EACJ,GAAIH,aAAmBI,GAAyBD,EAASE,GAAK,iBAAiB,YAAYN,EAAOC,CAAO,UAChGA,aAAmBC,GAAuBE,EAASE,GAAK,eAAe,YAAYN,EAAOC,CAAO,UACjGA,aAAmBM,GAAmBH,EAASE,GAAK,WAAW,YAAYN,EAAOC,CAAO,MAC7F,OAAM,IAAI,MAAM,sHAAsH,EAC3I,OAAOG,CACT,CAEQ,gCAAmE,CACzE,OAAO,IAAI,QAAiC,CAACI,EAASC,IAAW,CAC/D,KAAK,IAAI,EACN,KAAMC,GAAeF,EAAQE,EAAW,IAAKC,GAAcC,GAAwB,CAAC,EAAGD,CAAS,CAAC,CAAC,CAAC,EACnG,MAAOE,GAAQJ,EAAOI,CAAG,CAAC,CAC/B,CAAC,CACH,CAEA,kBAAkBC,EAAqB,GAAO,CAC5C,OAAO,IAAIC,GACT,KAAK,+BAA+B,EACpC,KAAK,MACLD,CACF,CACF,CAEA,qBAAsB,CACpB,OAAO,IAAIE,GACT,KAAK,+BAA+B,EACpC,KAAK,KACP,CACF,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GACT,KAAK,+BAA+B,EACpC,KAAK,KACP,CACF,CACF,EAEaC,GAAN,cAAmCpB,EAA+C,CACvF,MAAsB,KAA0C,CAC9D,IAAMqB,EAAiB,MAAM,IAAIhB,GAAmB,KAAK,MAAO,KAAK,OAAO,EACxEiB,EAAgCD,EAAe,GACnD,OAAAA,EAAe,QAASE,GAAkB,CACpCA,EAAc,MAAQD,EAA8B,QAAOA,EAAgCC,EACjG,CAAC,EACMD,CACT,CAEQ,+BAA4E,CAElF,OAAO,IAAI,QAA2C,MAAOZ,GAAY,CACvE,IAAMG,EAAY,MAAM,KAAK,IAAI,EACjCH,EAAQG,EAAYC,GAA4B,CAAC,EAAGD,CAAS,EAAI,MAAS,CAC5E,CAAC,CACH,CAEA,kBAAkBG,EAAqB,GAAO,CAC5C,OAAO,IAAIQ,GACT,KAAK,8BAA8B,EACnC,KAAK,MACLR,CACF,CACF,CAEA,qBAAsB,CACpB,OAAO,IAAIS,GACT,KAAK,8BAA8B,EACnC,KAAK,KACP,CACF,CAEA,kBAAmB,CACjB,OAAO,IAAIC,GACT,KAAK,8BAA8B,EACnC,KAAK,KACP,CACF,CACF,ECjGO,SAASC,IAAiBC,EAAkBC,EAAgC,IAAIC,GAA+C,CACpI,OAAO,IAAIC,GAAqBH,EAAOC,CAAO,CAChD,CAEO,SAASG,GAAeJ,EAAkBC,EAAgC,IAAIC,GAA6C,CAChI,OAAO,IAAIG,GAAmBL,EAAOC,CAAO,CAC9C,CCLA,eAAsBK,IAAuBC,EAAkBC,EAAiG,CAC9J,OAAOC,GAAeF,EAAO,IAAIG,GAAsBF,EAAgB,CAAE,cAAAA,CAAc,EAAI,CAAC,CAAC,CAAC,EAC3F,kBAAkB,EAClB,oBAAoB,CACzB,CAEA,eAAsBG,IAAmBJ,EAAkBK,EAAoC,CAAC,EAA4E,CAC1K,OAAOH,GAAeF,EAAO,IAAIM,GAAkBD,CAAa,CAAC,EAC9D,kBAAkB,EAClB,oBAAoB,CACzB,CAEO,IAAME,IAAWR,IClBjB,SAASS,GAAkBC,EAA+BC,EAA+B,CAC9F,GAAID,EAAK,SAAWC,EAAK,OAAQ,MAAM,IAAI,MAAM,gDAAgD,EACjG,IAAMC,EAAQ,MAAM,KAAKF,CAAI,EACvBG,EAAQ,MAAM,KAAKF,CAAI,EAC7B,OAAO,KAAK,KACVC,EACG,IAAI,CAACE,EAAKC,IAAMD,EAAMD,EAAME,EAAE,EAC9B,OAAO,CAACC,EAAKC,IAASD,EAAOC,EAAOA,EAAO,CAAC,CACjD,CACF,CCJO,IAAMC,GAAN,KAAkB,CAIvB,YAAYC,EAAkJC,EAAoB,GAAK,CACrL,KAAK,mBAAqBA,EAC1B,IAAMC,EAAa,MAAM,QAAQF,CAAM,EAAIA,EAAS,CAACA,CAAM,EAC3D,GAAI,CAACE,EAAW,OAAQ,MAAM,IAAI,MAAM,yDAAyD,EACjG,IAAIC,EAAQ,EACNC,EAAoB,IAAM,UAAUD,MAC1C,KAAK,oBAAsBD,EAAW,IAAKG,GAAS,CAClD,GAAIA,aAAgBC,GAAwB,OAAOD,EACnD,GAAIA,aAAgB,aAAc,OAAO,IAAIC,GAAuBF,EAAkB,EAAG,CAACC,CAAI,CAAC,EAC/F,GAAIA,EAAK,YAAcA,EAAK,sBAAsB,aAAc,OAAO,IAAIC,GAAuBF,EAAkB,EAAG,CAACC,EAAK,UAAU,CAAC,EACxI,MAAM,IAAI,MAAM,qMAAqM,CACvN,CAAC,CACH,CAEA,IAAW,oBAA+C,CAAE,OAAO,KAAK,mBAAqB,CAE7F,IAAW,mBAA4B,CAAE,OAAO,KAAK,kBAAoB,CAElE,oBAAoBE,EAA+BC,EAAqC,CAC7F,OAAOA,EACJ,IAAKC,GAAMC,GAAkBD,EAAGF,CAAe,CAAC,EAChD,OAAO,CAACI,EAAIC,IAAOD,EAAKC,EAAI,CAAC,GAAKJ,EAAY,QAAU,EAC7D,CAEO,gBAAgBD,EAA0C,CAC/D,OAAO,KAAK,mBACT,IAAI,CAAC,CAAE,YAAAC,EAAa,MAAAK,CAAM,IAAM,IAAIC,GAAUD,EAAO,KAAK,oBAAoBN,EAAiBC,CAAW,CAAC,CAAC,EAC5G,OAAO,CAACO,EAAMC,IAAUD,EAAK,SAAWC,EAAK,SAAWD,EAAOC,CAAK,CACzE,CAEO,cAAcT,EAA0C,CAC7D,IAAMU,EAAY,KAAK,gBAAgBV,CAAe,EACtD,OAAQU,EAAU,SAAW,KAAK,mBAAsBA,EAAY,IAAIH,GAAU,UAAWG,EAAU,QAAQ,CACjH,CAEO,QAAc,CACnB,MAAO,CACL,kBAAmB,KAAK,mBACxB,mBAAoB,KAAK,oBAAoB,IAAKC,GAAOA,EAAG,OAAO,CAAC,CACtE,CACF,CAEA,OAAc,SAASC,EAAwB,CAC7C,IAAMC,EAAqBD,EAAK,mBAAmB,IAAKD,GAAYZ,GAAuB,SAASY,CAAE,CAAC,EACvG,OAAO,IAAInB,GAAYqB,EAAoBD,EAAK,iBAAiB,CACnE,CACF,EClDO,SAASE,IAAuBC,EAAuB,CAC5D,IAAMC,EAAM,IAAIC,GAChB,OAAAD,EAAI,eAAeD,CAAO,EACnBC,CACT,CCHO,SAASE,IAAiBC,EAAYC,EAA4B,CACvE,GAAM,CAAE,MAAAC,EAAO,OAAAC,CAAO,EAAI,IAAIC,GAAWH,EAAW,MAAOA,EAAW,MAAM,EAE5E,GAAIC,GAAS,GAAKC,GAAU,EAC1B,MAAM,IAAI,MAAM,uCAAuC,KAAK,UAAU,CAAE,MAAAD,EAAO,OAAAC,CAAO,CAAC,GAAG,EAG5F,GAAI,MAAM,QAAQH,CAAO,EAEvB,OAAQA,EAAuB,IAAKK,GAAQN,IAAcM,EAAK,CAAE,MAAAH,EAAO,OAAAC,CAAO,CAAgB,CAAC,EAGlG,GAAIG,GAAoBN,CAAO,EAAG,CAChC,IAAMO,EAAmBP,EAAQ,UAAU,QAAQE,EAAOC,CAAM,EAC1DK,EAAmBR,EAAQ,mBAAmB,QAAQO,EAAiB,IAAI,MAAOA,EAAiB,IAAI,MAAM,EACnH,OAAOE,GAAwBC,GAAwBV,EAASO,CAAgB,EAAGC,CAAgB,CACrG,CAEA,OAAIG,GAAoBX,CAAO,EACtBU,GAAwBV,EAASA,EAAQ,UAAU,QAAQE,EAAOC,CAAM,CAAC,EAG9EH,aAAmBY,IAAiBZ,aAAmBa,GACjDb,EAAgB,QAAQE,EAAOC,CAAM,EAGxCH,CACT,CCTO,IAAMc,IAAcA",
-  "names": ["tfjs_esm_exports", "__export", "ii", "oa", "sa", "cu", "pu", "mu", "fu", "Zn", "Go", "ia", "aa", "Wo", "kl", "la", "ua", "ca", "ma", "pa", "Uo", "El", "lp", "ap", "cg", "Ho", "ai", "up", "cp", "p1", "Sb", "Py", "lo", "qo", "uo", "pp", "_l", "li", "Ko", "mp", "jo", "Al", "fp", "dp", "Xo", "Yo", "da", "fa", "Zo", "My", "ra", "hp", "ha", "Jo", "gp", "xp", "yp", "$l", "Xd", "jd", "i0", "vb", "bp", "ts", "wp", "qd", "xa", "ga", "es", "ui", "ya", "Cp", "Dl", "ba", "rs", "ns", "Yd", "os", "Ii", "Si", "Bc", "wa", "ci", "Ph", "Ca", "ss", "Ly", "Ip", "co", "Sp", "ye", "Ia", "Sa", "va", "zo", "Rl", "Np", "Ch", "Bn", "is", "Na", "Ta", "vp", "as", "ka", "f1", "Ea", "_a", "Aa", "m1", "xlt", "_u", "ls", "cs", "Fl", "kp", "Tp", "Ep", "us", "ps", "ms", "fs", "ds", "$a", "du", "_p", "hs", "pi", "Ra", "Fa", "Oa", "Da", "k0", "gs", "mi", "Wr", "Ws", "fi", "xs", "ylt", "ys", "bs", "ws", "hu", "Tn", "Ap", "$p", "Dp", "Ol", "x0", "Rp", "Qo", "Pa", "Xe", "Cs", "vs", "di", "Ss", "Op", "Is", "Fp", "Ns", "qa", "Ts", "ks", "Bi", "La", "Pp", "hi", "Ma", "qi", "_s", "Ba", "Es", "za", "gi", "Ds", "Va", "xi", "Pl", "Ga", "Ll", "Ml", "Lp", "yi", "As", "zl", "Rs", "po", "Wa", "Bl", "Vl", "Gl", "Fs", "$s", "Jr", "Os", "Ps", "Ft", "pe", "Jn", "Ua", "Ha", "Qn", "Mp", "bi", "Wl", "blt", "Ka", "wi", "Ci", "Ee", "ax", "lx", "X", "LE", "Zp", "qu", "Ai", "ux", "cx", "px", "mx", "fx", "dx", "Yl", "gx", "gE", "v", "BE", "Di", "xx", "yx", "bx", "Zl", "wx", "n6", "GE", "Ri", "Vr", "nx", "wt", "VZ", "J", "Cx", "Cr", "sn", "wn", "ne", "Ix", "Sx", "vx", "Nx", "K$", "Qp", "In", "em", "Tx", "Ex", "Tlt", "Jl", "rm", "hh", "Xu", "nm", "un", "AR", "ch", "W0", "_x", "Fi", "HZ", "Kl", "WE", "Ax", "gpt", "vt", "xpt", "pt", "$x", "Dx", "lv", "UE", "Oi", "hpt", "dpt", "uv", "Pn", "z", "$r", "Rx", "Fx", "er", "rr", "Ox", "Yu", "au", "xo", "Spt", "vpt", "Pi", "Yp", "hM", "uu", "Li", "m6", "ox", "Cpt", "u0", "Jd", "zg", "olt", "ik", "bK", "wK", "Re", "ln", "tl", "Xl", "Gs", "h6", "j$", "Pv", "_r", "xm", "Px", "Lx", "Mx", "De", "Ur", "ED", "Ql", "om", "Ln", "pv", "KE", "M7", "z7", "hD", "zx", "Sr", "tu", "Gx", "sm", "im", "Rr", "eu", "am", "Wx", "hX", "jE", "Lt", "yE", "Ir", "ru", "Hx", "XE", "Sn", "ve", "ah", "YE", "_D", "Ja", "Mi", "qx", "Kx", "q8", "AD", "Zu", "s6", "D", "ZE", "JE", "Ht", "gh", "Qa", "Bs", "Ei", "cr", "yr", "T", "QE", "cn", "t_", "e_", "r_", "n_", "jx", "an", "ou", "Jg", "Xx", "ypt", "o_", "s_", "i_", "a_", "v_", "tc", "N_", "zi", "su", "wpt", "Za", "ty", "Xp", "j8", "h1", "Lu", "UZ", "$D", "Fr", "lm", "Ipt", "R", "pr", "T_", "k_", "E_", "__", "lu", "um", "cm", "mt", "a6", "lh", "mh", "pm", "mm", "K8", "Q", "tH", "Npt", "nlt", "elt", "rlt", "wT", "A_", "Yr", "ey", "dX", "fm", "dm", "Rt", "hm", "dh", "gm", "ec", "Le", "iu", "zs", "nu", "gX", "c6", "fX", "mr", "Se", "Mt", "ym", "Mn", "nr", "bo", "ry", "xX", "ct", "ft", "Wu", "ny", "$i", "ur", "Me", "Vs", "rx", "$_", "D_", "R_", "go", "OE", "B", "Dr", "bpt", "oy", "ic", "Ot", "bm", "sy", "Nlt", "vlt", "wm", "vr", "sr", "F_", "y", "CK", "IK", "iy", "Bx", "plt", "cR", "PE", "Um", "slt", "dM", "Zke", "dd", "_e", "ly", "Ne", "It", "DU", "QS", "RU", "FU", "OU", "PU", "Pg", "r", "__require", "e", "gr", "Wt", "LU", "n", "o", "Tl", "T1", "Elt", "N1", "Ue", "mo", "t", "Fn", "g1", "x1", "zu", "He", "fo", "Mu", "ho", "C1", "v1", "b1", "Rn", "S1", "Vp", "Bg", "m0", "s", "i", "a", "u", "l", "Ls", "y1", "s4", "w1", "Bp", "I1", "p0", "gt", "c", "p", "m", "f", "d", "iE", "aE", "u_", "l_", "rv", "p_", "c_", "nv", "f_", "m_", "ov", "h_", "d_", "sv", "x_", "g_", "iv", "h", "g", "b_", "y_", "av", "w_", "I_", "C_", "Yx", "w", "C", "N", "_", "b", "$", "P", "V", "G", "W", "q", "H", "fh", "fNt", "S_", "lj", "uj", "cj", "pj", "mj", "fj", "Ju", "gN", "Zb", "ig", "BW", "VW", "GW", "WW", "fI", "Ok", "Fk", "it", "jt", "qe", "ke", "fe", "Ae", "_n", "or", "Hn", "Lr", "L", "U", "x", "A", "F", "Ru", "Y", "ut", "xt", "Dt", "ie", "Mo", "j", "Z", "et", "rt", "ot", "at", "nt", "Yc", "dt", "ht", "bt", "Et", "At", "Vt", "Gt", "Xt", "hr", "Xn", "Zt", "ce", "he", "We", "Mr", "zr", "qn", "Kn", "Xr", "Zi", "Il", "Rd", "Md", "Tr", "Ji", "Gk", "Fd", "Sl", "Xc", "Qi", "Wk", "Uk", "gI", "pg", "Od", "en", "mg", "xI", "yI", "vg", "EI", "Kt", "ea", "bI", "Hk", "wI", "CI", "YS", "II", "Pd", "fg", "dg", "Ld", "vl", "qk", "gg", "hg", "SI", "Ag", "zd", "vI", "ZS", "$g", "Kk", "jk", "NI", "e1", "Zc", "TI", "kI", "t1", "Xk", "_I", "AI", "Zk", "$I", "xg", "yg", "DI", "RI", "FI", "Qk", "OI", "PI", "LI", "Fu", "MI", "zI", "bg", "BI", "Jc", "VI", "GI", "Dg", "Jk", "Bd", "WI", "ZI", "UI", "HI", "qI", "wg", "Cg", "KI", "jI", "Ig", "Sg", "XI", "An", "rp", "o1", "n1", "JS", "$U", "YI", "Yk", "JI", "QI", "tS", "eS", "rS", "nS", "oS", "sS", "iS", "aS", "lS", "uS", "cS", "pS", "mS", "fS", "dS", "hS", "gS", "xS", "yS", "bS", "wS", "CS", "IS", "SS", "vS", "NS", "TS", "kS", "ES", "_S", "AS", "$S", "DS", "RS", "FS", "OS", "PS", "LS", "MS", "zS", "BS", "VS", "GS", "WS", "US", "HS", "Ng", "Tg", "Vd", "qS", "KS", "Qc", "kg", "tp", "ep", "jS", "K", "lt", "_t", "re", "Ke", "je", "ee", "te", "dr", "jn", "ta", "Eg", "Gd", "XS", "rn", "Nl", "_g", "QW", "tU", "eU", "rU", "nU", "oU", "sU", "iU", "aU", "lU", "uU", "cU", "pU", "mU", "fU", "dU", "hU", "gU", "xU", "yU", "bU", "wU", "CU", "IU", "SU", "vU", "NU", "TU", "kU", "EU", "_U", "Rg", "r1", "Fg", "Og", "AU", "HW", "G5e", "UW", "qW", "dI", "Lk", "Pk", "ra", "zo", "$n", "s1", "Lg", "MU", "np", "zU", "BU", "VU", "GU", "E", "$e", "Dn", "Yn", "Bo", "xr", "Jt", "WU", "na", "UU", "HU", "qU", "Pu", "KU", "jU", "lr", "t0", "e0", "r0", "n0", "o0", "XU", "Mg", "s0", "Vo", "i1", "a1", "op", "oi", "sp", "si", "l1", "Ou", "Wd", "ip", "YU", "Ud", "ZU", "JU", "Hd", "u1", "qd", "t4", "z", "r4", "e4", "i0", "c1", "a0", "l0", "n4", "Kd", "ii", "oa", "sa", "Zn", "Go", "ia", "aa", "Wo", "kl", "la", "ua", "ca", "pa", "ma", "Uo", "ap", "El", "lp", "Ho", "ai", "up", "p1", "cp", "lo", "qo", "uo", "pp", "_l", "li", "Ko", "mp", "jo", "Al", "fp", "dp", "Xo", "Yo", "fa", "Zo", "da", "hp", "ha", "Jo", "gp", "xp", "yp", "$l", "jd", "Xd", "Qo", "bp", "ts", "wp", "ga", "xa", "es", "ui", "ya", "Cp", "Dl", "ba", "rs", "ns", "os", "ci", "wa", "Ca", "ss", "co", "Ip", "Sp", "Ia", "Sa", "va", "is", "Na", "Ta", "vp", "as", "ka", "Ea", "_a", "Aa", "m1", "f1", "xlt", "Rl", "Np", "ls", "us", "cs", "Tp", "Fl", "kp", "Ep", "ps", "ms", "fs", "ds", "$a", "_p", "hs", "pi", "Da", "Ra", "Fa", "Oa", "mi", "gs", "fi", "xs", "ylt", "ys", "bs", "ws", "Ap", "$p", "Dp", "Ol", "Rp", "Pa", "Cs", "di", "Is", "Fp", "Ss", "Op", "vs", "Ns", "Ts", "ks", "La", "Pp", "hi", "Ma", "gi", "Es", "za", "Ba", "_s", "Va", "As", "$s", "xi", "yi", "Ds", "Pl", "Ga", "Ll", "Ml", "Lp", "Rs", "zl", "Wa", "Bl", "Vl", "Gl", "Fs", "Os", "Ps", "Jn", "Ua", "Ha", "Qn", "Mp", "bi", "Wl", "blt", "wi", "po", "Yd", "qa", "Ci", "Ii", "Si", "vi", "o4", "zp", "Zd", "Jd", "c0", "u0", "zg", "Lu", "h1", "vlt", "Nlt", "Tlt", "y", "p4", "Wp", "Hl", "f4", "c4", "Qd", "Gu", "Gp", "d0", "Vu", "E1", "Bu", "nn", "f0", "_1", "Pe", "k1", "kr", "Ul", "i4", "Vg", "a4", "l4", "u4", "m4", "Gg", "h0", "d4", "A1", "$1", "D1", "th", "g0", "R1", "h4", "Wg", "rh", "eh", "F1", "pe", "Ms", "Up", "g4", "O1", "P1", "L1", "Ft", "O", "Ka", "go", "I0", "nh", "y4", "Ut", "x0", "y0", "b0", "w0", "C0", "x4", "sr", "Wu", "M1", "b4", "S0", "Ug", "ql", "k", "w4", "C4", "v0", "Kl", "T0", "v4", "S4", "I4", "N0", "On", "Br", "B1", "z1", "I", "ja", "k0", "T", "N4", "wn", "on", "ur", "oh", "Hg", "G1", "T4", "qg", "A4", "E0", "V1", "W1", "U1", "Hp", "_0", "Kg", "A0", "qp", "Ni", "jg", "k4", "E4", "_4", "Ce", "H1", "q1", "K1", "j1", "$0", "D0", "Uu", "jl", "X1", "R0", "Ti", "Y1", "$4", "D4", "Xg", "Xa", "Kp", "Z1", "R4", "F4", "O4", "P4", "J1", "Q1", "L4", "M4", "ki", "tE", "z4", "Yg", "jp", "Er", "Zg", "eE", "rE", "nE", "oE", "sE", "F0", "B4", "O0", "P0", "wt", "V4", "J", "G4", "sn", "Jg", "W4", "_r", "uE", "mE", "fE", "G0", "tx", "Qg", "cE", "B0", "dE", "hE", "U4", "H4", "q4", "lE", "Ya", "L0", "K4", "j4", "M0", "z0", "X4", "Y4", "sh", "Z4", "pE", "ih", "ex", "V0", "yE", "xE", "J4", "Lt", "Q4", "Ei", "dpt", "hpt", "gpt", "W0", "xpt", "Pn", "ah", "ypt", "B", "vt", "De", "bpt", "tH", "wpt", "Cpt", "Ipt", "Spt", "vpt", "Xp", "gE", "Npt", "eH", "Xl", "rH", "Ht", "nH", "Za", "oH", "Ot", "sH", "Vr", "Pt", "bE", "ge", "nx", "mH", "cH", "pH", "rx", "Hu", "wE", "iH", "aH", "lH", "uH", "ox", "CE", "lh", "IE", "sx", "U0", "Le", "dH", "bH", "gH", "xH", "yH", "hH", "q0", "wH", "AE", "kE", "$E", "EE", "_E", "vE", "H0", "fH", "NE", "TE", "CH", "SE", "Q", "uh", "_i", "Cn", "OE", "DE", "EH", "FE", "kH", "SH", "NH", "RE", "vH", "TH", "_H", "ix", "IH", "K0", "j0", "PE", "AH", "X", "$H", "Yp", "DH", "pt", "RH", "D", "FH", "Ee", "OH", "ax", "PH", "lx", "LH", "LE", "MH", "Zp", "zH", "qu", "BH", "Ai", "VH", "ux", "GH", "cx", "WH", "px", "UH", "mx", "HH", "fx", "qH", "dx", "KH", "zE", "ju", "Y0", "hx", "jH", "X0", "ME", "Jp", "ZH", "JH", "XH", "Z0", "Ku", "YH", "to", "Ar", "Ie", "QH", "R", "tq", "Yl", "eq", "gx", "rq", "ne", "nq", "Yr", "oq", "Rt", "sq", "$i", "iq", "BE", "aq", "Zl", "VE", "lq", "Di", "uq", "xx", "cq", "yx", "pq", "bx", "mq", "wx", "fq", "GE", "dq", "Ri", "hq", "Cx", "xo", "gq", "Cr", "xq", "Ix", "yq", "Sx", "bq", "vx", "wq", "Nx", "Cq", "In", "Iq", "Qp", "Sq", "tm", "vq", "em", "Nq", "Tx", "Tq", "kx", "kq", "Ex", "Eq", "Jl", "_q", "rm", "Aq", "Xu", "$q", "nm", "Dq", "ch", "Rq", "_x", "Fq", "Fi", "Oq", "WE", "Pq", "Ax", "Lq", "$r", "Mq", "_e", "zq", "It", "Bq", "$x", "Vq", "Dx", "Gq", "UE", "Wq", "Oi", "Uq", "Rx", "J0", "HE", "Q0", "yo", "Hq", "tv", "ph", "qq", "Kq", "Ir", "jq", "Ja", "Xq", "an", "mt", "Yq", "Se", "Zq", "Mt", "Jq", "ft", "Qq", "qE", "Qa", "tK", "Fx", "eK", "er", "rK", "rr", "nK", "Ox", "oK", "Dr", "sK", "Yu", "iK", "Pi", "aK", "Li", "lK", "Re", "uK", "ln", "cK", "Px", "pK", "Lx", "mK", "Mx", "fK", "Ql", "dK", "om", "hK", "Ln", "KE", "gK", "zx", "xK", "Sr", "yK", "tu", "bK", "Vx", "wK", "CK", "IK", "Bx", "un", "SK", "zs", "vK", "Gx", "NK", "ct", "TK", "sm", "kK", "im", "EK", "Rr", "_K", "eu", "AK", "am", "$K", "Wx", "Ux", "DK", "mh", "jE", "RK", "ru", "FK", "Hx", "OK", "XE", "PK", "Sn", "LK", "ve", "Ne", "cr", "YE", "MK", "Mi", "zK", "qx", "BK", "Kx", "VK", "Zu", "GK", "ZE", "WK", "JE", "UK", "Bs", "HK", "yr", "qK", "QE", "KK", "cn", "jK", "t_", "XK", "e_", "YK", "r_", "ZK", "n_", "JK", "nu", "QK", "ej", "tj", "jx", "rj", "ou", "nj", "Xx", "oj", "o_", "sj", "s_", "ij", "i_", "aj", "a_", "Qx", "Qu", "Zx", "Jx", "dj", "v_", "hj", "tc", "gj", "N_", "xj", "zi", "su", "yj", "ty", "bj", "Fr", "wj", "lm", "Cj", "pr", "Ij", "T_", "Sj", "k_", "vj", "E_", "Nj", "__", "Tj", "um", "kj", "cm", "Ej", "pm", "_j", "mm", "Aj", "A_", "$j", "ey", "Dj", "fm", "Rj", "dm", "Fj", "hm", "Oj", "dh", "Pj", "gm", "Lj", "ec", "Mj", "iu", "zj", "au", "Bj", "tl", "Vj", "xm", "Gj", "mr", "Wj", "lu", "Uj", "ym", "Hj", "Mn", "qj", "nr", "Kj", "bo", "jj", "ry", "Xj", "ny", "Me", "Vs", "$_", "D_", "R_", "Yj", "oy", "Zj", "bm", "Jj", "sy", "Qj", "wm", "t6", "vr", "F_", "iy", "ay", "e6", "ly", "r6", "n6", "o6", "s6", "i6", "a6", "O_", "u6", "c6", "p6", "m6", "P_", "f6", "lv", "uv", "hh", "d6", "h6", "uu", "L_", "M_", "z_", "g6", "Cm", "rc", "nc", "oc", "sc", "x6", "y6", "uy", "b6", "cy", "w6", "C6", "I6", "B_", "S6", "py", "v6", "my", "N6", "V_", "T6", "G_", "k6", "W_", "E6", "U_", "_6", "H_", "wo", "A6", "q_", "K_", "$6", "R6", "D6", "fy", "cv", "dy", "hy", "j_", "F6", "O6", "P6", "X_", "L6", "Y_", "M6", "Z_", "z6", "J_", "B6", "Q_", "V6", "gy", "G6", "xy", "W6", "U6", "tA", "H6", "eA", "q6", "rA", "K6", "nA", "j6", "oA", "sA", "Xe", "X6", "Gr", "Y6", "iA", "Z6", "aA", "J6", "lA", "Q6", "uA", "tX", "cA", "eX", "pA", "rX", "nX", "mA", "oX", "sX", "fA", "iX", "dA", "aX", "hA", "lX", "gA", "uX", "xA", "cX", "yA", "pX", "bA", "mX", "wA", "fX", "dX", "Gs", "pv", "hX", "gX", "xX", "Wr", "cu", "pu", "mu", "fu", "Bi", "du", "hu", "Ws", "ic", "yX", "gh", "v", "RX", "FX", "OX", "PX", "LX", "DX", "yy", "Vi", "fv", "mv", "bX", "WX", "XX", "CX", "BX", "VX", "NX", "wX", "KX", "HX", "UX", "g5", "h5", "GX", "YX", "jX", "TX", "EX", "SX", "kX", "_X", "IX", "AX", "$X", "t5", "e5", "r5", "s5", "a5", "i5", "n5", "o5", "p5", "l5", "u5", "c5", "ZX", "MX", "QX", "hv", "zX", "vX", "dv", "qX", "CA", "IA", "JX", "d5", "f5", "m5", "Ur", "by", "SA", "vA", "NA", "TA", "kA", "EA", "_A", "AA", "$A", "DA", "RA", "x5", "FA", "OA", "y5", "PA", "LA", "MA", "zA", "BA", "VA", "GA", "WA", "UA", "HA", "qA", "KA", "b5", "jA", "XA", "YA", "ZA", "JA", "QA", "t2", "e2", "r2", "n2", "o2", "s2", "i2", "a2", "l2", "p2", "u2", "c2", "m2", "f2", "d2", "h2", "g2", "x2", "y2", "b2", "w2", "w5", "C2", "I2", "wy", "gv", "S2", "C5", "v2", "N2", "I5", "T2", "k2", "E2", "_2", "A2", "$2", "D2", "R2", "F2", "O2", "P2", "L2", "xv", "M2", "z2", "S5", "v5", "B2", "V2", "G2", "W2", "U2", "H2", "q2", "K2", "j2", "X2", "Y2", "Z2", "J2", "Q2", "t$", "e$", "r$", "n$", "o$", "s$", "yv", "bv", "i$", "a$", "l$", "u$", "c$", "p$", "m$", "f$", "d$", "h$", "g$", "x$", "N5", "y$", "T5", "vn", "Hr", "M", "St", "Im", "xh", "Io", "ro", "Cv", "Nr", "xe", "So", "el", "Co", "Sm", "wv", "Gi", "k5", "yh", "vo", "b$", "Wi", "Cy", "Ze", "w$", "C$", "Iy", "E5", "vy", "Sy", "gu", "I$", "S$", "v$", "N$", "T$", "vm", "Fe", "E$", "pn", "Iv", "bh", "k$", "Hs", "_5", "Ny", "_$", "Ty", "A5", "A$", "No", "ac", "qs", "Zr", "Sv", "ir", "mn", "no", "nl", "$$", "Ey", "D$", "R$", "rl", "vv", "wh", "Nm", "Tv", "Tm", "To", "Nv", "_y", "lc", "fn", "F$", "O$", "Ay", "P$", "xu", "L$", "M$", "$5", "D5", "dn", "km", "yu", "Em", "_m", "Am", "$m", "Dm", "R5", "qr", "uc", "cc", "pc", "mc", "fc", "dc", "Rm", "z$", "B$", "Te", "de", "$y", "Fm", "Nt", "Bt", "Om", "G$", "Ch", "O5", "Ih", "Pm", "ye", "Jr", "P5", "ol", "L5", "$t", "M5", "z5", "kv", "Ks", "Dy", "B5", "ko", "Ry", "Fy", "U$", "hc", "V5", "W5", "W$", "G5", "U5", "K$", "H5", "j5", "K5", "q5", "Ev", "gc", "Lm", "Mm", "zm", "Bm", "H$", "ze", "q$", "Be", "j$", "Z5", "o8", "n8", "s8", "i8", "e8", "a8", "l8", "Y5", "u8", "Q5", "J5", "t8", "r8", "X5", "ED", "Tn", "ll", "NY", "FY", "xZ", "OY", "Gv", "Wv", "Uv", "UY", "qY", "jY", "HY", "KY", "XY", "VY", "uZ", "CZ", "PY", "gY", "xY", "yY", "bY", "wY", "sZ", "iZ", "IY", "TY", "vY", "BY", "kY", "cY", "RY", "_Y", "gZ", "hZ", "YY", "ZY", "pZ", "mZ", "vD", "ND", "QY", "tZ", "Pv", "uY", "GY", "mY", "eZ", "rZ", "yZ", "fZ", "dZ", "TD", "kD", "JY", "LY", "MY", "zY", "DY", "fY", "pY", "AY", "bZ", "$Y", "wZ", "aZ", "CY", "nZ", "oZ", "dY", "EY", "lZ", "hY", "cZ", "SY", "WY", "Ui", "Oy", "X$", "c8", "sl", "Py", "_v", "Ly", "My", "zy", "hn", "By", "gn", "Sh", "Hi", "Vm", "bu", "p8", "m8", "f8", "d8", "h8", "xc", "Gm", "g8", "Wm", "x8", "y8", "Nh", "vh", "Vy", "Th", "kh", "Z$", "b8", "w8", "Av", "J$", "Wy", "Uy", "C8", "I8", "S8", "v8", "N8", "T8", "Eh", "k8", "$v", "Gy", "Q$", "_h", "eD", "Rv", "Dv", "rD", "A8", "Hy", "$8", "D8", "_8", "nD", "yc", "qy", "Um", "zn", "R8", "Ky", "jy", "oD", "F8", "aD", "sD", "O8", "lD", "iD", "P8", "L8", "uD", "Xy", "Hm", "Yy", "Zy", "M8", "cD", "Eo", "Fv", "z8", "Ov", "pD", "mD", "B8", "V8", "fD", "G8", "W8", "Bn", "Jy", "dD", "hD", "U8", "H8", "qi", "q8", "K8", "j8", "Qr", "Qy", "tb", "eb", "rb", "nb", "ob", "sb", "ib", "ab", "lb", "qm", "ub", "cb", "pb", "js", "Lv", "Xs", "Mv", "mb", "wu", "yD", "bD", "gD", "me", "xD", "be", "Km", "jm", "Xm", "Ym", "Zm", "Jm", "Cu", "Nn", "Ys", "Ah", "zv", "Y8", "wD", "Z8", "bc", "Iu", "il", "al", "Qm", "tf", "fb", "ef", "Su", "rf", "nf", "J8", "of", "Bv", "Vv", "Ic", "wc", "cl", "sf", "Cc", "af", "ul", "lf", "Q8", "db", "Sc", "uf", "vc", "cf", "pf", "mf", "ff", "df", "hf", "gf", "xf", "yf", "pl", "bf", "wf", "Cf", "If", "Sf", "vf", "$h", "tY", "Nf", "Tf", "kf", "Ef", "Dh", "eY", "rY", "nY", "_f", "Af", "oY", "$f", "wb", "CD", "hb", "Df", "Rf", "gb", "Ff", "Of", "xb", "Pf", "Lf", "yb", "Mf", "zf", "bb", "Bf", "Vf", "Cb", "Gf", "sY", "iY", "Wf", "Uf", "aY", "ID", "Hf", "SD", "qf", "_D", "DZ", "OZ", "IZ", "SZ", "NZ", "TZ", "_Z", "RZ", "AZ", "$Z", "FZ", "PZ", "kZ", "EZ", "vZ", "AD", "$D", "MZ", "LZ", "zZ", "Sb", "Ib", "DD", "vb", "BZ", "VZ", "GZ", "oo", "RD", "Hv", "UZ", "Nb", "HZ", "S", "br", "xn", "Tb", "FD", "_o", "Rh", "Zs", "qv", "qZ", "Kv", "KZ", "jv", "jZ", "Xv", "XZ", "Yv", "YZ", "Zv", "ZZ", "Jv", "JZ", "Qv", "QZ", "tN", "t7", "eN", "e7", "rN", "r7", "nN", "n7", "oN", "o7", "sN", "s7", "iN", "i7", "aN", "a7", "lN", "l7", "uN", "u7", "cN", "c7", "Fh", "kb", "Fb", "_b", "Rb", "Eb", "Pb", "Db", "Ob", "Ab", "$b", "OD", "pN", "p7", "PD", "LD", "Lb", "ae", "MD", "zD", "Vn", "BD", "Kf", "Mb", "zb", "VD", "GD", "WD", "UD", "HD", "qD", "KD", "jD", "mN", "XD", "YD", "ZD", "Bb", "JD", "QD", "tR", "eR", "rR", "nR", "oR", "sR", "iR", "aR", "lR", "fN", "Oh", "dN", "hN", "F7", "O7", "uR", "$7", "D7", "R7", "Nc", "Vb", "P7", "L7", "Ph", "M7", "B7", "z7", "cR", "AR", "Yf", "Js", "ed", "Xf", "rd", "wR", "NR", "TR", "kR", "_R", "PN", "ER", "CR", "bR", "gR", "pR", "Gb", "vu", "mR", "xN", "fR", "Wb", "dR", "V7", "hR", "G7", "jf", "Tc", "AN", "yN", "Lh", "bN", "xR", "Hb", "yR", "fl", "EN", "Je", "TN", "vN", "NN", "Ub", "kN", "SN", "IN", "CN", "qb", "_N", "wN", "kc", "kn", "W7", "U7", "Kb", "Mh", "IR", "jb", "$N", "SR", "Zf", "Jf", "Qf", "zh", "DN", "RN", "Xb", "FN", "ON", "td", "vR", "H7", "Yb", "tt", "q7", "Nu", "mw", "zN", "sd", "Jb", "MN", "BN", "Ec", "VN", "WN", "HN", "qN", "Qb", "tw", "jN", "KN", "YN", "XN", "ew", "ZN", "rw", "JN", "QN", "Bh", "tT", "eT", "rT", "nw", "ow", "sw", "Ac", "nT", "dl", "sF", "LN", "$c", "iw", "aw", "ld", "lF", "sT", "lw", "Dc", "Rc", "Fc", "aT", "uw", "cw", "id", "pw", "K7", "$R", "Qt", "wr", "DR", "nd", "Kr", "RR", "Ao", "FR", "$o", "OR", "oe", "od", "j7", "Ki", "PR", "yn", "kt", "Do", "X7", "LR", "GN", "MR", "UN", "zR", "Y7", "BR", "Z7", "VR", "J7", "GR", "Q7", "WR", "tJ", "UR", "eJ", "HR", "rJ", "qR", "nJ", "KR", "oJ", "jR", "sJ", "_c", "XR", "iJ", "YR", "aJ", "ZR", "Ve", "JR", "lJ", "QR", "uJ", "cJ", "pJ", "mJ", "tF", "fJ", "dJ", "eF", "Ro", "ad", "nF", "rF", "hJ", "oF", "oT", "iF", "Fo", "aF", "gJ", "uF", "xJ", "cF", "iT", "yJ", "bJ", "Vh", "pF", "Gh", "mF", "lT", "fF", "uT", "dF", "CJ", "cT", "hF", "pT", "gF", "mT", "xF", "Oc", "Yt", "yF", "fT", "bF", "IJ", "wF", "SJ", "CF", "vJ", "IF", "NJ", "SF", "TJ", "vF", "kJ", "NF", "EJ", "TF", "_J", "kF", "AJ", "EF", "$J", "_F", "DJ", "AF", "RJ", "FJ", "$F", "OJ", "DF", "ud", "fw", "dw", "RF", "PJ", "FF", "LJ", "OF", "MJ", "PF", "zJ", "LF", "BJ", "MF", "VJ", "zF", "GJ", "BF", "WJ", "VF", "UJ", "GF", "HJ", "WF", "ji", "UF", "Tu", "HF", "dT", "qF", "qJ", "KF", "KJ", "jF", "jJ", "XF", "XJ", "YF", "YJ", "ZF", "ZJ", "JF", "JJ", "QF", "QJ", "tO", "tQ", "eO", "eQ", "rO", "rQ", "nO", "nQ", "oO", "hT", "sO", "oQ", "iO", "sQ", "aO", "iQ", "lO", "uO", "cO", "pO", "hl", "mO", "aQ", "fO", "lQ", "dO", "uQ", "cQ", "pQ", "mQ", "fQ", "dQ", "hQ", "hO", "cd", "gO", "gQ", "Wh", "Uh", "hw", "xQ", "yQ", "gT", "bQ", "wQ", "xO", "Hh", "CQ", "yO", "bO", "IQ", "SQ", "wO", "vQ", "CO", "NQ", "IO", "TQ", "SO", "kQ", "vO", "EQ", "NO", "_Q", "TO", "AQ", "kO", "$Q", "EO", "DQ", "_O", "RQ", "AO", "FQ", "OQ", "$O", "PQ", "DO", "LQ", "MQ", "RO", "zQ", "FO", "BQ", "OO", "xT", "PO", "VQ", "LO", "GQ", "MO", "WQ", "zO", "UQ", "BO", "VO", "GO", "HQ", "WO", "qQ", "UO", "KQ", "HO", "jQ", "XQ", "qO", "jO", "yT", "KO", "YQ", "XO", "ZQ", "JQ", "YO", "QQ", "t9", "ZO", "e9", "r9", "JO", "n9", "QO", "qh", "tP", "eP", "rP", "bT", "nP", "o9", "gw", "s9", "i9", "oP", "a9", "sP", "l9", "iP", "u9", "aP", "c9", "lP", "p9", "uP", "m9", "cP", "f9", "pP", "d9", "mP", "h9", "fP", "g9", "dP", "hP", "x9", "gP", "y9", "xP", "b9", "w9", "yP", "C9", "bP", "I9", "wP", "S9", "v9", "N9", "CP", "T9", "IP", "k9", "SP", "E9", "vP", "_9", "NP", "A9", "TP", "$9", "kP", "D9", "EP", "R9", "_P", "F9", "AP", "O9", "$P", "P9", "DP", "L9", "RP", "FP", "M9", "OP", "z9", "PP", "B9", "LP", "V9", "MP", "G9", "zP", "W9", "BP", "U9", "VP", "H9", "GP", "q9", "WP", "K9", "UP", "J9", "Q9", "HP", "j9", "X9", "Z9", "Y9", "Kh", "ttt", "qP", "ett", "KP", "rtt", "jP", "ntt", "dd", "Qs", "mtt", "Zh", "OT", "JP", "Iw", "yt", "ST", "NT", "DT", "TT", "_T", "ET", "AT", "vT", "xl", "pd", "QP", "MT", "ctt", "FT", "RT", "yl", "fd", "PT", "zT", "ZP", "LT", "Wn", "BT", "VT", "Eu", "GT", "vw", "kT", "Cw", "ftt", "dtt", "Sw", "ptt", "md", "Yh", "$T", "Pc", "xw", "wT", "Gn", "stt", "ott", "ku", "jr", "Pr", "Lc", "XP", "jh", "Xi", "YP", "Xh", "itt", "att", "ltt", "gl", "utt", "tL", "yw", "bw", "ww", "IT", "htt", "Tt", "Ge", "ti", "Mc", "gtt", "eL", "hd", "gd", "Nw", "rL", "nL", "Tw", "xtt", "wtt", "Stt", "ytt", "Itt", "btt", "Ctt", "ktt", "yd", "ztt", "Vtt", "Wtt", "Htt", "Ktt", "jtt", "Xtt", "oL", "Mtt", "Btt", "Gtt", "Utt", "qtt", "Ytt", "Ztt", "sL", "Ett", "Ptt", "Att", "Dtt", "_tt", "Ltt", "$tt", "Rtt", "Ftt", "Ott", "vtt", "Ntt", "Ttt", "zc", "bd", "wd", "Cd", "zt", "aL", "WT", "iL", "lL", "uL", "we", "kw", "Ew", "_w", "Aw", "tet", "Jh", "$w", "ik", "JT", "ek", "jT", "ZT", "KT", "qT", "YT", "XT", "HT", "UT", "nk", "rk", "sk", "ok", "Rw", "Pw", "Dw", "Ow", "Fw", "QT", "tk", "Qh", "Bc", "eet", "cL", "Lw", "pL", "mL", "fL", "dL", "hL", "gL", "xL", "yL", "bL", "wL", "CL", "IL", "SL", "vL", "NL", "TL", "kL", "EL", "_L", "AL", "$L", "DL", "RL", "FL", "OL", "PL", "LL", "ML", "zL", "BL", "Mw", "VL", "GL", "WL", "zw", "UL", "HL", "qL", "KL", "jL", "XL", "YL", "ZL", "Vc", "JL", "ak", "Qe", "QL", "Bw", "Id", "ret", "Vw", "eM", "rM", "tM", "net", "oet", "set", "tn", "fr", "nM", "lk", "oM", "sM", "iM", "Gc", "aM", "uM", "cM", "pM", "mM", "fM", "so", "Gw", "aet", "uet", "cet", "Ww", "pet", "met", "fet", "det", "_u", "het", "dM", "hM", "Zke", "Sd", "io", "Yi", "Oo", "tr", "gM", "En", "xM", "uk", "ck", "get", "yM", "pk", "mk", "xet", "bM", "Po", "Ct", "le", "bl", "vd", "fk", "tg", "wM", "eg", "CM", "IM", "st", "SM", "rg", "Uw", "bet", "Un", "Hw", "wet", "qw", "Au", "vM", "Wc", "NM", "Oe", "TM", "dk", "Uc", "Cet", "kM", "EM", "Iet", "_M", "vet", "Net", "AM", "Tet", "ket", "$M", "DM", "Eet", "RM", "Kw", "jw", "Xw", "FM", "_et", "OM", "Aet", "PM", "Yw", "Zw", "LM", "MM", "Jw", "$et", "zM", "Det", "BM", "Ret", "Fet", "VM", "Oet", "Pet", "GM", "Let", "Met", "WM", "zet", "Bet", "Vet", "UM", "Get", "Wet", "HM", "ei", "$u", "Uet", "qM", "Het", "KM", "Qw", "tC", "qet", "jM", "Ket", "XM", "jet", "YM", "eC", "rC", "Xet", "ZM", "nC", "Yet", "hk", "oC", "Zet", "ri", "JM", "Jet", "QM", "Qet", "tz", "trt", "ez", "ert", "gk", "rz", "wl", "nz", "rrt", "oz", "xk", "sz", "iz", "nrt", "az", "sC", "iC", "ort", "lz", "aC", "uz", "srt", "cz", "lC", "cC", "uC", "Hc", "pz", "Nd", "irt", "yk", "mz", "Td", "pC", "kd", "mC", "fC", "dC", "hC", "art", "fz", "gC", "xC", "yC", "bC", "lrt", "dz", "urt", "hz", "crt", "gz", "prt", "xz", "mrt", "yz", "frt", "drt", "bz", "hrt", "grt", "wz", "wC", "xrt", "Cz", "qc", "ng", "Iz", "Sz", "CC", "yrt", "vz", "brt", "Nz", "wrt", "Tz", "IC", "Crt", "kz", "Ed", "_d", "Irt", "Ez", "SC", "vC", "Srt", "_z", "vrt", "Az", "NC", "Nrt", "$z", "TC", "Trt", "Dz", "krt", "Rz", "Ert", "_rt", "Art", "Fz", "$rt", "Drt", "Rrt", "Oz", "Frt", "Ort", "Prt", "Pz", "Lrt", "Mrt", "Lz", "zrt", "Brt", "bk", "Mz", "kC", "zz", "Bz", "Vrt", "Vz", "og", "EC", "Grt", "Gz", "_C", "Cl", "Wz", "AC", "Uz", "Hz", "Wrt", "qz", "Urt", "Hrt", "qrt", "Kz", "$C", "DC", "jz", "Krt", "Ad", "wk", "jrt", "Xz", "Xrt", "Yz", "RC", "Yrt", "Zz", "FC", "Zrt", "Ck", "Jz", "Jrt", "Qrt", "tnt", "Qz", "ent", "rnt", "nnt", "t3", "ont", "e3", "snt", "int", "r3", "ant", "lnt", "n3", "unt", "cnt", "o3", "pnt", "mnt", "fnt", "s3", "dnt", "hnt", "gnt", "i3", "xnt", "a3", "ynt", "bnt", "wnt", "l3", "Cnt", "Int", "u3", "Snt", "vnt", "Nnt", "c3", "Tnt", "knt", "p3", "Ent", "_nt", "Ant", "m3", "OC", "PC", "$nt", "f3", "LC", "Dnt", "d3", "h3", "Ik", "g3", "Rnt", "Fnt", "Ont", "x3", "Pnt", "y3", "Lnt", "b3", "MC", "zC", "Mnt", "w3", "znt", "C3", "I3", "S3", "v3", "N3", "Bnt", "T3", "Vnt", "Gnt", "Wnt", "k3", "BC", "VC", "Unt", "E3", "Hnt", "qnt", "Knt", "_3", "GC", "jnt", "Xnt", "Sk", "A3", "$3", "vk", "D3", "Nk", "R3", "Ynt", "F3", "Znt", "Jnt", "Qnt", "O3", "tot", "eot", "P3", "rot", "not", "L3", "oot", "sot", "M3", "WC", "iot", "z3", "sg", "B3", "V3", "G3", "aot", "W3", "UC", "HC", "Tk", "U3", "lot", "uot", "cot", "H3", "pot", "q3", "mot", "K3", "fot", "j3", "dot", "X3", "kk", "Y3", "hot", "got", "Z3", "xot", "yot", "bot", "J3", "wot", "Cot", "Iot", "Q3", "qC", "KC", "Sot", "tB", "jC", "vot", "eB", "XC", "YC", "Not", "rB", "ZC", "Tot", "nB", "JC", "QC", "kot", "oB", "tI", "sB", "Eot", "_ot", "iB", "Aot", "$ot", "aB", "$d", "Dot", "lB", "eI", "Rot", "uB", "rI", "Fot", "cB", "Oot", "Pot", "pB", "Lot", "Mot", "zot", "mB", "Bot", "Vot", "fB", "Got", "Wot", "dB", "Uot", "Hot", "hB", "qot", "Kot", "gB", "jot", "xB", "Xot", "yB", "Yot", "bB", "Zot", "wB", "Jot", "CB", "Qot", "IB", "tst", "SB", "vB", "est", "NB", "rst", "nst", "TB", "kB", "ost", "EB", "sst", "_B", "nI", "ist", "AB", "ast", "$B", "lst", "DB", "ust", "RB", "cst", "pst", "FB", "mst", "fst", "OB", "oI", "dst", "Ek", "PB", "sI", "iI", "Kc", "LB", "hst", "MB", "aI", "gst", "zB", "xst", "BB", "yst", "VB", "lI", "bst", "GB", "wst", "qt", "Du", "WB", "Cst", "Ist", "UB", "se", "HB", "ue", "Sst", "qB", "KB", "vst", "Nst", "jB", "jc", "XB", "YB", "Tst", "ao", "Est", "kst", "ZB", "bn", "JB", "_st", "Ast", "QB", "tV", "$st", "Dst", "eV", "rV", "Rst", "Fst", "nV", "oV", "Ost", "Pst", "sV", "ar", "iV", "aV", "Lst", "Mst", "lV", "Lo", "zst", "Bst", "Vst", "uV", "Gst", "cV", "ni", "pV", "mV", "fV", "Wst", "Ust", "dV", "_k", "hV", "gV", "Hst", "qst", "xV", "yV", "Kst", "jst", "bV", "wV", "CV", "Ak", "IV", "Xst", "Yst", "SV", "vV", "Zst", "Jst", "NV", "TV", "Qst", "tit", "kV", "EV", "eit", "rit", "_V", "AV", "nit", "oit", "$V", "DV", "sit", "RV", "FV", "uI", "OV", "$k", "PV", "LV", "iit", "ait", "MV", "zV", "lit", "BV", "VV", "uit", "cit", "GV", "WV", "pit", "mit", "UV", "HV", "fit", "dit", "qV", "KV", "hit", "git", "jV", "XV", "xit", "yit", "YV", "bit", "ZV", "wit", "JV", "QV", "Cit", "Iit", "tG", "Sit", "eG", "vit", "rG", "nG", "Nit", "oG", "sG", "Tit", "iG", "kit", "aG", "lG", "Eit", "_it", "uG", "Ait", "cG", "pG", "$it", "Dit", "mG", "fG", "Rit", "Fit", "dG", "hG", "Oit", "Pit", "gG", "Lit", "xG", "Dk", "yG", "Mit", "zit", "bG", "Bit", "wG", "CG", "Dd", "IG", "Vit", "Git", "SG", "vG", "Wit", "Uit", "NG", "TG", "Hit", "qit", "kG", "Kit", "EG", "_G", "jit", "Xit", "AG", "Yit", "$G", "Zit", "DG", "RG", "Jit", "Qit", "cI", "tat", "FG", "OG", "eat", "rat", "PG", "LG", "nat", "oat", "MG", "sat", "zG", "iat", "BG", "VG", "GG", "WG", "aat", "lat", "UG", "HG", "uat", "cat", "qG", "KG", "pat", "mat", "jG", "XG", "fat", "dat", "YG", "ZG", "JG", "QG", "hat", "gat", "tW", "eW", "xat", "yat", "rW", "nW", "bat", "wat", "oW", "sW", "iW", "Cat", "Iat", "aW", "Sat", "lW", "uW", "vat", "Nat", "cW", "pW", "Tat", "kat", "mW", "fW", "pI", "mI", "Eat", "dW", "_at", "hW", "Aat", "gW", "xW", "yW", "$at", "bW", "wW", "Dat", "Rat", "CW", "IW", "Fat", "Oat", "SW", "Pat", "vW", "Lat", "NW", "Mat", "TW", "zat", "kW", "EW", "Bat", "Vat", "_W", "AW", "$W", "DW", "Gat", "Wat", "RW", "FW", "Uat", "Hat", "OW", "PW", "qat", "Kat", "LW", "jat", "MW", "Xat", "zW", "Yat", "Rk", "zk", "XW", "Bk", "KW", "Zat", "cg", "ZW", "Mk", "Qat", "Jat", "jW", "hI", "lg", "YW", "ag", "Vk", "ug", "tlt", "elt", "rlt", "nlt", "olt", "slt", "ilt", "JW", "alt", "llt", "ult", "clt", "plt", "draw_exports", "__export", "AnchorPosition", "DrawBox", "DrawBoxOptions", "DrawFaceLandmarks", "DrawFaceLandmarksOptions", "DrawTextField", "DrawTextFieldOptions", "drawContour", "drawDetections", "drawFaceExpressions", "drawFaceLandmarks", "drawContour", "ctx", "points", "isClosed", "x", "y", "prevIdx", "from", "to", "utils_exports", "__export", "computeReshapedDimensions", "getCenterPoint", "isDimensions", "isEven", "isFloat", "isTensor", "isTensor1D", "isTensor2D", "isTensor3D", "isTensor4D", "isValidNumber", "isValidProbablitiy", "range", "round", "Dimensions", "width", "height", "isValidNumber", "isTensor", "tensor", "dim", "Ft", "isTensor1D", "isTensor2D", "isTensor3D", "isTensor4D", "isFloat", "num", "isEven", "round", "prec", "f", "isDimensions", "obj", "computeReshapedDimensions", "width", "height", "inputSize", "scale", "Dimensions", "getCenterPoint", "pts", "sum", "pt", "Point", "range", "start", "step", "_", "i", "isValidNumber", "isValidProbablitiy", "Point", "x", "y", "pt", "Box", "rect", "isValidNumber", "box", "callee", "allowNegativeDimensions", "_box", "isBbox", "isRect", "x", "y", "width", "height", "Point", "val", "diff", "s", "scaleX", "isDimensions", "scaleY", "padX", "padY", "imgWidth", "imgHeight", "right", "bottom", "clippedX", "clippedY", "newWidth", "newHeight", "clippedWidth", "clippedHeight", "sx", "sy", "imageHeight", "imageWidth", "w", "h", "dx", "dy", "edx", "edy", "ex", "ey", "region", "BoundingBox", "Box", "left", "top", "right", "bottom", "allowNegativeDimensions", "ObjectDetection", "score", "classScore", "className", "relativeBox", "imageDims", "Dimensions", "Box", "width", "height", "FaceDetection", "ObjectDetection", "score", "relativeBox", "imageDims", "width", "height", "iou", "box1", "box2", "isIOU", "width", "height", "interSection", "minBbox", "pts", "xs", "pt", "ys", "minX", "min", "x", "minY", "y", "maxX", "max", "maxY", "BoundingBox", "nonMaxSuppression", "boxes", "scores", "iouThreshold", "isIOU", "indicesSortedByScore", "score", "boxIndex", "c1", "c2", "c", "pick", "curr", "indices", "outputs", "i", "idx", "currBox", "idxBox", "iou", "_", "j", "normalize", "x", "meanRgb", "B", "r", "g", "b", "avg_r", "xo", "avg_g", "avg_b", "avg_rgb", "ne", "ct", "padToSquare", "imgTensor", "isCenterImage", "B", "height", "width", "dimDiff", "paddingAmount", "paddingAxis", "createPaddingTensor", "paddingAmountLocal", "paddingTensorShape", "xo", "paddingTensorAppend", "remainingPaddingAmount", "tensorsToStack", "t", "J", "ne", "shuffleArray", "inputArray", "array", "i", "j", "x", "sigmoid", "x", "inverseSigmoid", "Rect", "Box", "x", "y", "width", "height", "allowNegativeDimensions", "relX", "relY", "relScale", "FaceLandmarks", "relativeFaceLandmarkPositions", "imgDims", "shift", "Point", "width", "height", "Dimensions", "pt", "x", "y", "detection", "options", "box", "FaceDetection", "Box", "useDlibAlignment", "minBoxPadding", "centers", "leftEyeCenter", "rightEyeCenter", "mouthCenter", "distToMouth", "eyeToMouthDist", "size", "refPoint", "getCenterPoint", "Rect", "padding", "minBbox", "FaceLandmarks5", "FaceLandmarks", "pts", "getCenterPoint", "FaceLandmarks68", "FaceLandmarks", "getCenterPoint", "FaceMatch", "label", "distance", "withDistance", "round", "LabeledBox", "Box", "box", "label", "callee", "isValidNumber", "LabeledFaceDescriptors", "label", "descriptors", "desc", "d", "json", "PredictedBox", "LabeledBox", "box", "label", "score", "classScore", "callee", "isValidProbablitiy", "isWithFaceDetection", "obj", "FaceDetection", "extendWithFaceDetection", "sourceObj", "detection", "createBrowserEnv", "fetch", "isNodejs", "createFileSystem", "fs", "requireFsError", "isNodejs", "err", "filePath", "resolve", "reject", "buffer", "createNodejsEnv", "Canvas", "Image", "Video", "createCanvasElement", "createImageElement", "createVideoElement", "fetch", "fileSystem", "createFileSystem", "isBrowser", "environment", "getEnv", "setEnv", "env", "initialize", "isBrowser", "createBrowserEnv", "isNodejs", "createNodejsEnv", "monkeyPatch", "Canvas", "Image", "createFileSystem", "resolveInput", "arg", "env", "getContext2dOrThrow", "canvasArg", "Canvas", "CanvasRenderingContext2D", "env", "canvas", "resolveInput", "ctx", "AnchorPosition", "DrawTextFieldOptions", "options", "anchorPosition", "backgroundColor", "fontColor", "fontSize", "fontStyle", "padding", "DrawTextField", "text", "anchor", "ctx", "l", "w0", "w1", "canvasDims", "isShiftLeft", "isShiftTop", "textFieldWidth", "textFieldHeight", "x", "y", "width", "height", "newX", "newY", "canvasArg", "canvas", "resolveInput", "getContext2dOrThrow", "maxTextWidth", "textHeight", "upperLeft", "textLine", "i", "DrawBoxOptions", "options", "boxColor", "lineWidth", "label", "drawLabelOptions", "defaultDrawLabelOptions", "DrawTextFieldOptions", "DrawBox", "box", "Box", "canvasArg", "ctx", "getContext2dOrThrow", "x", "y", "width", "height", "DrawTextField", "drawDetections", "canvasArg", "detections", "det", "score", "FaceDetection", "isWithFaceDetection", "box", "Box", "label", "round", "DrawBox", "isMediaLoaded", "media", "Image", "Video", "env", "awaitMediaLoaded", "media", "resolve", "reject", "env", "isMediaLoaded", "onError", "e", "onLoad", "bufferToImage", "buf", "resolve", "reject", "reader", "img", "env", "getMediaDimensions", "input", "Image", "Video", "env", "Dimensions", "createCanvas", "width", "height", "createCanvasElement", "env", "canvas", "createCanvasFromMedia", "media", "dims", "ImageData", "isMediaLoaded", "getMediaDimensions", "getContext2dOrThrow", "imageTensorToCanvas", "imgTensor", "canvas", "targetCanvas", "env", "height", "width", "numChannels", "isTensor4D", "imgTensor3D", "B", "nx", "isMediaElement", "input", "Image", "Canvas", "Video", "env", "imageToSquare", "input", "inputSize", "centerImage", "Image", "Canvas", "env", "createCanvas", "dims", "getMediaDimensions", "scale", "width", "height", "targetCanvas", "inputCanvas", "createCanvasFromMedia", "offset", "dx", "dy", "getContext2dOrThrow", "NetInput", "inputs", "treatAsBatchInput", "input", "idx", "isTensor3D", "isTensor4D", "batchSize", "canvas", "env", "createCanvasFromMedia", "range", "_", "batchIdx", "width", "height", "computeReshapedDimensions", "inputSize", "isCenterInputs", "B", "inputTensors", "Ft", "imgTensor", "rr", "padToSquare", "Gs", "nx", "imageToSquare", "nr", "t", "J", "toNetInput", "inputs", "NetInput", "inputArgArray", "getIdxHint", "idx", "inputArray", "resolveInput", "input", "i", "isMediaElement", "isTensor3D", "isTensor4D", "batchSize", "awaitMediaLoaded", "extractFaces", "input", "detections", "Canvas", "env", "canvas", "netInput", "toNetInput", "tensorOrCanvas", "imageTensorToCanvas", "ctx", "getContext2dOrThrow", "det", "FaceDetection", "box", "x", "y", "width", "height", "faceImg", "createCanvas", "extractFaceTensors", "imageTensor", "detections", "isTensor3D", "isTensor4D", "B", "imgHeight", "imgWidth", "numChannels", "det", "FaceDetection", "box", "x", "y", "width", "height", "gm", "fetchOrThrow", "url", "init", "fetch", "env", "res", "fetchImage", "uri", "res", "fetchOrThrow", "blob", "bufferToImage", "fetchJson", "uri", "fetchOrThrow", "fetchNetWeights", "uri", "fetchOrThrow", "bufferToVideo", "buf", "resolve", "reject", "video", "env", "fetchVideo", "uri", "res", "fetchOrThrow", "blob", "bufferToVideo", "getModelUris", "uri", "defaultModelName", "defaultManifestFilename", "protocol", "parts", "s", "manifestFile", "modelBaseUri", "loadWeightMap", "uri", "defaultModelName", "manifestUri", "modelBaseUri", "getModelUris", "manifest", "fetchJson", "_r", "matchDimensions", "input", "reference", "useMediaDimensions", "width", "height", "getMediaDimensions", "NeuralNetwork", "name", "paramPath", "obj", "objProp", "tensor", "param", "Ka", "path", "variable", "ur", "throwOnRedispose", "flat", "arr", "weightsOrUrl", "uri", "weightMap", "loadWeightMap", "filePath", "readFile", "env", "manifestUri", "modelBaseUri", "getModelUris", "fetchWeightsFromDisk", "filePaths", "fp", "buf", "loadWeights", "_r", "manifest", "paramMappings", "params", "weights", "result", "res", "Ft", "depthwiseSeparableConv", "x", "params", "stride", "B", "out", "mm", "X", "denseBlock3", "x", "denseBlockParams", "isFirstLayer", "B", "out1", "Fr", "X", "In", "depthwiseSeparableConv", "out2", "in3", "out3", "denseBlock4", "isScaleDown", "in4", "out4", "convLayer", "x", "params", "padding", "withRelu", "B", "out", "X", "In", "Fr", "disposeUnusedWeightTensors", "weightMap", "paramMappings", "path", "pm", "extractConvParamsFactory", "extractWeights", "paramMappings", "channelsIn", "channelsOut", "filterSize", "mappedPrefix", "filters", "$_", "bias", "Me", "extractFCParamsFactory", "extractWeights", "paramMappings", "channelsIn", "channelsOut", "mappedPrefix", "fc_weights", "Vs", "fc_bias", "Me", "SeparableConvParams", "depthwise_filter", "pointwise_filter", "bias", "extractSeparableConvParamsFactory", "extractWeights", "paramMappings", "channelsIn", "channelsOut", "mappedPrefix", "depthwise_filter", "$_", "pointwise_filter", "bias", "Me", "SeparableConvParams", "loadSeparableConvParamsFactory", "extractWeightEntry", "prefix", "extractWeightEntryFactory", "weightMap", "paramMappings", "originalPath", "paramRank", "mappedPath", "tensor", "isTensor", "extractWeightsFactory", "weights", "remainingWeights", "extractWeights", "numWeights", "ret", "getRemainingWeights", "extractorsFactory", "extractWeights", "paramMappings", "extractConvParams", "extractConvParamsFactory", "extractSeparableConvParams", "extractSeparableConvParamsFactory", "extractDenseBlock3Params", "channelsIn", "channelsOut", "mappedPrefix", "isFirstLayer", "conv0", "conv1", "conv2", "extractDenseBlock4Params", "conv3", "extractParams", "weights", "paramMappings", "extractWeights", "getRemainingWeights", "extractWeightsFactory", "extractDenseBlock4Params", "extractorsFactory", "dense0", "dense1", "dense2", "dense3", "loadConvParamsFactory", "extractWeightEntry", "prefix", "filters", "bias", "loadParamsFactory", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractConvParams", "loadConvParamsFactory", "extractSeparableConvParams", "loadSeparableConvParamsFactory", "extractDenseBlock3Params", "prefix", "isFirstLayer", "conv0", "conv1", "conv2", "extractDenseBlock4Params", "conv3", "extractParamsFromWeightMap", "weightMap", "paramMappings", "extractDenseBlock4Params", "loadParamsFactory", "params", "disposeUnusedWeightTensors", "FaceFeatureExtractor", "NeuralNetwork", "input", "params", "B", "batchTensor", "J", "normalized", "normalize", "out", "denseBlock4", "Yl", "toNetInput", "weightMap", "extractParamsFromWeightMap", "weights", "extractParams", "fullyConnectedLayer", "x", "params", "B", "X", "Lt", "extractParams", "weights", "channelsIn", "channelsOut", "paramMappings", "extractWeights", "getRemainingWeights", "extractWeightsFactory", "fc", "extractFCParamsFactory", "extractParamsFromWeightMap", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractFcParams", "prefix", "weights", "bias", "params", "disposeUnusedWeightTensors", "seperateWeightMaps", "weightMap", "featureExtractorMap", "classifierMap", "key", "map", "FaceProcessor", "NeuralNetwork", "_name", "faceFeatureExtractor", "input", "params", "B", "bottleneckFeatures", "NetInput", "fullyConnectedLayer", "throwOnRedispose", "weights", "paramMappings", "extractParams", "weightMap", "featureExtractorMap", "classifierMap", "seperateWeightMaps", "extractParamsFromWeightMap", "cIn", "cOut", "classifierWeightSize", "featureExtractorWeights", "classifierWeights", "FACE_EXPRESSION_LABELS", "FaceExpressions", "probabilities", "expression", "idx", "e0", "e1", "FaceExpressionNet", "FaceProcessor", "faceFeatureExtractor", "FaceFeatureExtractor", "input", "B", "iu", "toNetInput", "netInput", "out", "probabilitesByBatch", "vr", "t", "data", "predictionsByBatch", "probabilites", "FaceExpressions", "isWithFaceExpressions", "obj", "FaceExpressions", "extendWithFaceExpressions", "sourceObj", "expressions", "drawFaceExpressions", "canvasArg", "faceExpressions", "minConfidence", "textFieldAnchor", "e", "expr", "FaceExpressions", "isWithFaceExpressions", "resultsToDisplay", "exprLocal", "anchor", "isWithFaceDetection", "Point", "DrawTextField", "round", "isWithFaceLandmarks", "obj", "isWithFaceDetection", "FaceLandmarks", "FaceDetection", "calculateFaceAngle", "mesh", "degrees", "radians", "calcLengthBetweenTwoPoints", "a", "b", "angle", "calcYaw", "leftPoint", "midPoint", "rightPoint", "leftToMidpoint", "rightToMidpoint", "calcRoll", "lever", "pivot", "hypotenuse", "opposite", "angleInRadians", "angleInDegrees", "normalizeAngle", "tiltDirection", "calcPitch", "base", "baseCoords", "midToBaseLength", "direction", "pt", "extendWithFaceLandmarks", "sourceObj", "unshiftedLandmarks", "shift", "landmarks", "rect", "imageDims", "alignedRect", "DrawFaceLandmarksOptions", "options", "drawLines", "drawPoints", "lineWidth", "lineColor", "pointSize", "pointColor", "DrawFaceLandmarks", "faceLandmarks", "canvasArg", "ctx", "getContext2dOrThrow", "FaceLandmarks68", "drawContour", "drawPoint", "pt", "drawFaceLandmarks", "f", "landmarks", "FaceLandmarks", "isWithFaceLandmarks", "extractorsFactory", "extractWeights", "paramMappings", "extractConvParams", "extractConvParamsFactory", "extractSeparableConvParams", "extractSeparableConvParamsFactory", "extractReductionBlockParams", "channelsIn", "channelsOut", "mappedPrefix", "separable_conv0", "separable_conv1", "expansion_conv", "extractMainBlockParams", "channels", "separable_conv2", "extractParams", "weights", "numMainBlocks", "getRemainingWeights", "extractWeightsFactory", "entry_flow_conv_in", "entry_flow_reduction_block_0", "entry_flow_reduction_block_1", "entry_flow", "middle_flow", "range", "idx", "exit_flow_reduction_block", "exit_flow_separable_conv", "exit_flow", "loadParamsFactory", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractConvParams", "loadConvParamsFactory", "extractSeparableConvParams", "loadSeparableConvParamsFactory", "extractReductionBlockParams", "mappedPrefix", "separable_conv0", "separable_conv1", "expansion_conv", "extractMainBlockParams", "separable_conv2", "extractParamsFromWeightMap", "numMainBlocks", "entry_flow_conv_in", "entry_flow_reduction_block_0", "entry_flow_reduction_block_1", "entry_flow", "middle_flow", "range", "idx", "exit_flow_reduction_block", "exit_flow_separable_conv", "exit_flow", "disposeUnusedWeightTensors", "conv", "x", "params", "stride", "X", "In", "reductionBlock", "isActivateInput", "out", "Fr", "depthwiseSeparableConv", "ru", "mainBlock", "TinyXception", "NeuralNetwork", "numMainBlocks", "input", "B", "batchTensor", "J", "normalized", "normalize", "range", "idx", "toNetInput", "weightMap", "extractParamsFromWeightMap", "weights", "extractParams", "extractParams", "weights", "paramMappings", "extractWeights", "getRemainingWeights", "extractWeightsFactory", "extractFCParams", "extractFCParamsFactory", "age", "gender", "extractParamsFromWeightMap", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractFcParams", "prefix", "weights", "bias", "params", "disposeUnusedWeightTensors", "Gender", "AgeGenderNet", "NeuralNetwork", "faceFeatureExtractor", "TinyXception", "input", "params", "B", "bottleneckFeatures", "NetInput", "pooled", "Yl", "age", "fullyConnectedLayer", "gender", "iu", "toNetInput", "netInput", "out", "ages", "vr", "genders", "ageAndGenderTensors", "ageTensor", "i", "predictionsByBatch", "genderTensor", "probMale", "isMale", "genderProbability", "throwOnRedispose", "weights", "paramMappings", "extractParams", "weightMap", "featureExtractorMap", "classifierMap", "seperateWeightMaps", "extractParamsFromWeightMap", "featureExtractorWeights", "classifierWeights", "FaceLandmark68NetBase", "FaceProcessor", "output", "inputSize", "originalDimensions", "inputDimensions", "width", "height", "scale", "batchSize", "B", "createInterleavedTensor", "fillX", "fillY", "nr", "xo", "getPadding", "batchIdx", "cond", "getPaddingX", "w", "getPaddingY", "_", "input", "out", "toNetInput", "netInput", "landmarkTensors", "vr", "landmarksForBatch", "landmarkTensor", "landmarksArray", "xCoords", "i", "isEven", "yCoords", "FaceLandmarks68", "Point", "t", "FaceLandmark68Net", "FaceLandmark68NetBase", "faceFeatureExtractor", "FaceFeatureExtractor", "extractParamsFromWeightMapTiny", "weightMap", "paramMappings", "extractDenseBlock3Params", "loadParamsFactory", "params", "disposeUnusedWeightTensors", "extractParamsTiny", "weights", "paramMappings", "extractWeights", "getRemainingWeights", "extractWeightsFactory", "extractDenseBlock3Params", "extractorsFactory", "dense0", "dense1", "dense2", "TinyFaceFeatureExtractor", "NeuralNetwork", "input", "params", "B", "batchTensor", "J", "normalized", "normalize", "out", "denseBlock3", "Yl", "toNetInput", "weightMap", "extractParamsFromWeightMapTiny", "weights", "extractParamsTiny", "FaceLandmark68TinyNet", "FaceLandmark68NetBase", "faceFeatureExtractor", "TinyFaceFeatureExtractor", "FaceLandmarkNet", "FaceLandmark68Net", "scale", "x", "params", "X", "D", "convLayer", "x", "params", "strides", "withRelu", "padding", "filters", "bias", "out", "In", "X", "scale", "Fr", "conv", "convNoRelu", "convDown", "extractorsFactory", "extractWeights", "paramMappings", "extractFilterValues", "numFilterValues", "numFilters", "filterSize", "weights", "depth", "isFloat", "B", "Ot", "$_", "extractConvParams", "mappedPrefix", "filters", "bias", "Me", "extractScaleLayerParams", "numWeights", "biases", "extractConvLayerParams", "conv", "scale", "extractResidualLayerParams", "isDown", "conv1", "conv2", "extractParams", "getRemainingWeights", "extractWeightsFactory", "conv32_down", "conv32_1", "conv32_2", "conv32_3", "conv64_down", "conv64_1", "conv64_2", "conv64_3", "conv128_down", "conv128_1", "conv128_2", "conv256_down", "conv256_1", "conv256_2", "conv256_down_out", "fc", "Vs", "extractorsFactory", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractScaleLayerParams", "prefix", "weights", "biases", "extractConvLayerParams", "filters", "bias", "scale", "extractResidualLayerParams", "extractParamsFromWeightMap", "conv32_down", "conv32_1", "conv32_2", "conv32_3", "conv64_down", "conv64_1", "conv64_2", "conv64_3", "conv128_down", "conv128_1", "conv128_2", "conv256_down", "conv256_1", "conv256_2", "conv256_down_out", "fc", "isTensor2D", "params", "disposeUnusedWeightTensors", "residual", "x", "params", "out", "conv", "convNoRelu", "X", "Fr", "residualDown", "convDown", "pooled", "Yl", "zeros", "Ne", "isPad", "padShapeX", "zerosW", "ne", "padShapeY", "zerosH", "FaceRecognitionNet", "NeuralNetwork", "input", "params", "B", "batchTensor", "J", "normalized", "normalize", "out", "convDown", "ru", "residual", "residualDown", "globalAvg", "Lt", "toNetInput", "_a", "dim", "netInput", "faceDescriptorTensors", "vr", "faceDescriptorsForBatch", "t", "weightMap", "extractParamsFromWeightMap", "weights", "extractParams", "createFaceRecognitionNet", "weights", "net", "FaceRecognitionNet", "extendWithFaceDescriptor", "sourceObj", "descriptor", "isWithAge", "obj", "extendWithAge", "sourceObj", "age", "isWithGender", "obj", "isValidProbablitiy", "extendWithGender", "sourceObj", "gender", "genderProbability", "extractorsFactory", "extractWeights", "paramMappings", "extractDepthwiseConvParams", "numChannels", "mappedPrefix", "filters", "$_", "batch_norm_scale", "Me", "batch_norm_offset", "batch_norm_mean", "batch_norm_variance", "extractConvParams", "channelsIn", "channelsOut", "filterSize", "isPointwiseConv", "bias", "extractPointwiseConvParams", "extractConvPairParams", "depthwise_conv", "pointwise_conv", "extractMobilenetV1Params", "conv_0", "conv_1", "conv_2", "conv_3", "conv_4", "conv_5", "conv_6", "conv_7", "conv_8", "conv_9", "conv_10", "conv_11", "conv_12", "conv_13", "extractPredictionLayerParams", "box_encoding_0_predictor", "class_predictor_0", "box_encoding_1_predictor", "class_predictor_1", "box_encoding_2_predictor", "class_predictor_2", "box_encoding_3_predictor", "class_predictor_3", "box_encoding_4_predictor", "class_predictor_4", "box_encoding_5_predictor", "class_predictor_5", "extractParams", "weights", "getRemainingWeights", "extractWeightsFactory", "mobilenetv1", "prediction_layer", "output_layer", "rx", "extractorsFactory", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractPointwiseConvParams", "prefix", "idx", "mappedPrefix", "filters", "batch_norm_offset", "extractConvPairParams", "prefixDepthwiseConv", "mappedPrefixDepthwiseConv", "mappedPrefixPointwiseConv", "batch_norm_scale", "batch_norm_mean", "batch_norm_variance", "extractMobilenetV1Params", "extractConvParams", "bias", "extractBoxPredictorParams", "box_encoding_predictor", "class_predictor", "extractPredictionLayerParams", "extractParamsFromWeightMap", "extra_dim", "isTensor3D", "params", "disposeUnusedWeightTensors", "pointwiseConvLayer", "x", "params", "strides", "B", "out", "In", "X", "Cr", "epsilon", "depthwiseConvLayer", "x", "params", "strides", "B", "out", "Fi", "Di", "Cr", "getStridesForLayerIdx", "layerIdx", "idx", "mobileNetV1", "conv11", "pointwiseConvLayer", "param", "depthwiseConvStrides", "IOU", "boxes", "i", "j", "boxesData", "yminI", "xminI", "ymaxI", "xmaxI", "yminJ", "xminJ", "ymaxJ", "xmaxJ", "areaI", "areaJ", "intersectionYmin", "intersectionXmin", "intersectionYmax", "intersectionXmax", "intersectionArea", "nonMaxSuppression", "scores", "maxOutputSize", "iouThreshold", "scoreThreshold", "numBoxes", "outputSize", "candidates", "score", "boxIndex", "c", "c1", "c2", "suppressFunc", "x", "selected", "originalScore", "iou", "getCenterCoordinatesAndSizesLayer", "x", "vec", "vr", "Ot", "sizes", "ct", "centers", "X", "pt", "decodeBoxesLayer", "x0", "x1", "div0_out", "D", "er", "add0_out", "div1_out", "add1_out", "nr", "outputLayer", "boxPredictions", "classPredictions", "params", "B", "batchSize", "boxes", "R", "Dr", "scoresAndClasses", "Yr", "Rt", "scores", "boxesByBatch", "scoresByBatch", "boxPredictionLayer", "x", "params", "B", "batchSize", "boxPredictionEncoding", "R", "convLayer", "classPrediction", "predictionLayer", "x", "conv11", "params", "B", "conv0", "pointwiseConvLayer", "conv1", "conv2", "conv3", "conv4", "conv5", "conv6", "conv7", "boxPrediction0", "boxPredictionLayer", "boxPrediction1", "boxPrediction2", "boxPrediction3", "boxPrediction4", "boxPrediction5", "boxPredictions", "ne", "classPredictions", "SsdMobilenetv1Options", "minConfidence", "maxResults", "SsdMobilenetv1", "NeuralNetwork", "input", "params", "B", "batchTensor", "J", "x", "ct", "pt", "features", "mobileNetV1", "boxPredictions", "classPredictions", "predictionLayer", "outputLayer", "toNetInput", "options", "maxResults", "minConfidence", "SsdMobilenetv1Options", "netInput", "_boxes", "_scores", "boxes", "scores", "i", "scoresData", "indices", "nonMaxSuppression", "reshapedDims", "inputSize", "padX", "padY", "boxesData", "results", "idx", "top", "bottom", "val", "left", "right", "FaceDetection", "Rect", "weightMap", "extractParamsFromWeightMap", "weights", "extractParams", "createSsdMobilenetv1", "weights", "net", "SsdMobilenetv1", "createFaceDetectionNet", "FaceDetectionNet", "IOU_THRESHOLD", "BOX_ANCHORS", "Point", "BOX_ANCHORS_SEPARABLE", "MEAN_RGB_SEPARABLE", "DEFAULT_MODEL_NAME", "DEFAULT_MODEL_NAME_SEPARABLE_CONV", "isNumber", "arg", "validateConfig", "config", "c", "a", "leaky", "x", "B", "min", "D", "mt", "X", "Fr", "ct", "convWithBatchNorm", "x", "params", "B", "out", "cn", "In", "ct", "D", "X", "leaky", "depthwiseSeparableConv", "x", "params", "B", "out", "cn", "mm", "X", "leaky", "extractorsFactory", "extractWeights", "paramMappings", "extractConvParams", "extractConvParamsFactory", "extractBatchNormParams", "size", "mappedPrefix", "sub", "Me", "truediv", "extractConvWithBatchNormParams", "channelsIn", "channelsOut", "conv", "bn", "extractSeparableConvParams", "extractSeparableConvParamsFactory", "extractParams", "weights", "config", "boxEncodingSize", "filterSizes", "getRemainingWeights", "extractWeightsFactory", "params", "s0", "s1", "s2", "s3", "s4", "s5", "s6", "s7", "s8", "conv0", "conv1", "conv2", "conv3", "conv4", "conv5", "conv6", "conv7", "conv8", "extractorsFactory", "weightMap", "paramMappings", "extractWeightEntry", "extractWeightEntryFactory", "extractBatchNormParams", "prefix", "sub", "truediv", "extractConvParams", "filters", "bias", "extractConvWithBatchNormParams", "conv", "bn", "extractSeparableConvParams", "loadSeparableConvParamsFactory", "extractParamsFromWeightMap", "config", "params", "numFilters", "disposeUnusedWeightTensors", "TinyYolov2Options", "inputSize", "scoreThreshold", "_TinyYolov2Base", "NeuralNetwork", "config", "validateConfig", "x", "params", "out", "convWithBatchNorm", "ru", "convLayer", "leaky", "depthwiseSeparableConv", "input", "inputSize", "B", "batchTensor", "J", "normalize", "toNetInput", "forwardParams", "scoreThreshold", "TinyYolov2Options", "netInput", "out0", "vr", "inputDimensions", "results", "boxes", "res", "scores", "classScores", "classNames", "nonMaxSuppression", "box", "idx", "ObjectDetection", "weightMap", "extractParamsFromWeightMap", "weights", "filterSizes", "numFilters", "extractParams", "outputTensor", "inputBlobDimensions", "width", "height", "correctionFactorX", "correctionFactorY", "numCells", "numBoxes", "boxesTensor", "scoresTensor", "classScoresTensor", "reshaped", "iu", "mt", "scoresData", "boxesData", "row", "col", "anchor", "score", "sigmoid", "ctX", "ctY", "widthLocal", "heightLocal", "y", "pos", "classScore", "label", "BoundingBox", "classesTensor", "classesData", "_", "i", "max", "curr", "TinyYolov2Base", "TinyYolov2", "TinyYolov2Base", "withSeparableConvs", "config", "IOU_THRESHOLD", "BOX_ANCHORS_SEPARABLE", "MEAN_RGB_SEPARABLE", "BOX_ANCHORS", "input", "forwardParams", "det", "FaceDetection", "DEFAULT_MODEL_NAME_SEPARABLE_CONV", "DEFAULT_MODEL_NAME", "weightMap", "createTinyYolov2", "weights", "withSeparableConvs", "net", "TinyYolov2", "TinyFaceDetectorOptions", "TinyYolov2Options", "ComposableTask", "onfulfilled", "extractAllFacesAndComputeResults", "parentResults", "input", "computeResults", "extractedFaces", "getRectForAlignment", "alignedRect", "faceBoxes", "parentResult", "isWithFaceLandmarks", "faces", "Ft", "extractFaceTensors", "extractFaces", "results", "f", "extractSingleFaceAndComputeResult", "computeResult", "IOU_THRESHOLD", "BOX_ANCHORS", "Point", "MEAN_RGB", "TinyFaceDetector", "TinyYolov2Base", "config", "IOU_THRESHOLD", "BOX_ANCHORS", "MEAN_RGB", "input", "forwardParams", "det", "FaceDetection", "weightMap", "nets", "SsdMobilenetv1", "TinyFaceDetector", "TinyYolov2", "FaceLandmark68Net", "FaceLandmark68TinyNet", "FaceRecognitionNet", "FaceExpressionNet", "AgeGenderNet", "ssdMobilenetv1", "input", "options", "tinyFaceDetector", "tinyYolov2", "detectFaceLandmarks", "detectFaceLandmarksTiny", "computeFaceDescriptor", "recognizeFaceExpressions", "predictAgeAndGender", "loadSsdMobilenetv1Model", "url", "loadTinyFaceDetectorModel", "loadTinyYolov2Model", "loadFaceLandmarkModel", "loadFaceLandmarkTinyModel", "loadFaceRecognitionModel", "loadFaceExpressionModel", "loadAgeGenderModel", "loadFaceDetectionModel", "locateFaces", "detectLandmarks", "PredictFaceExpressionsTaskBase", "ComposableTask", "parentTask", "input", "extractedFaces", "PredictAllFaceExpressionsTask", "parentResults", "faceExpressionsByFace", "extractAllFacesAndComputeResults", "faces", "face", "nets", "parentResult", "i", "extendWithFaceExpressions", "PredictAllAgeAndGenderTask", "PredictSingleFaceExpressionsTask", "faceExpressions", "extractSingleFaceAndComputeResult", "PredictSingleAgeAndGenderTask", "PredictAllFaceExpressionsWithFaceAlignmentTask", "PredictAllAgeAndGenderWithFaceAlignmentTask", "ComputeAllFaceDescriptorsTask", "PredictSingleFaceExpressionsWithFaceAlignmentTask", "PredictSingleAgeAndGenderWithFaceAlignmentTask", "ComputeSingleFaceDescriptorTask", "PredictAgeAndGenderTaskBase", "ComposableTask", "parentTask", "input", "extractedFaces", "PredictAllAgeAndGenderTask", "parentResults", "ageAndGenderByFace", "extractAllFacesAndComputeResults", "faces", "face", "nets", "parentResult", "i", "age", "gender", "genderProbability", "extendWithAge", "extendWithGender", "PredictAllFaceExpressionsTask", "PredictSingleAgeAndGenderTask", "extractSingleFaceAndComputeResult", "PredictSingleFaceExpressionsTask", "PredictAllAgeAndGenderWithFaceAlignmentTask", "PredictAllFaceExpressionsWithFaceAlignmentTask", "ComputeAllFaceDescriptorsTask", "PredictSingleAgeAndGenderWithFaceAlignmentTask", "PredictSingleFaceExpressionsWithFaceAlignmentTask", "ComputeSingleFaceDescriptorTask", "ComputeFaceDescriptorsTaskBase", "ComposableTask", "parentTask", "input", "ComputeAllFaceDescriptorsTask", "parentResults", "extractAllFacesAndComputeResults", "faces", "face", "nets", "parentResult", "descriptor", "i", "extendWithFaceDescriptor", "PredictAllFaceExpressionsWithFaceAlignmentTask", "PredictAllAgeAndGenderWithFaceAlignmentTask", "ComputeSingleFaceDescriptorTask", "extractSingleFaceAndComputeResult", "PredictSingleFaceExpressionsWithFaceAlignmentTask", "PredictSingleAgeAndGenderWithFaceAlignmentTask", "DetectFaceLandmarksTaskBase", "ComposableTask", "parentTask", "input", "useTinyLandmarkNet", "nets", "DetectAllFaceLandmarksTask", "parentResults", "detections", "res", "faces", "Ft", "extractFaceTensors", "extractFaces", "faceLandmarksByFace", "face", "f", "_parentResult", "i", "parentResult", "extendWithFaceLandmarks", "PredictAllFaceExpressionsWithFaceAlignmentTask", "PredictAllAgeAndGenderWithFaceAlignmentTask", "ComputeAllFaceDescriptorsTask", "DetectSingleFaceLandmarksTask", "detection", "landmarks", "PredictSingleFaceExpressionsWithFaceAlignmentTask", "PredictSingleAgeAndGenderWithFaceAlignmentTask", "ComputeSingleFaceDescriptorTask", "DetectFacesTaskBase", "ComposableTask", "input", "options", "SsdMobilenetv1Options", "DetectAllFacesTask", "result", "TinyFaceDetectorOptions", "nets", "TinyYolov2Options", "resolve", "reject", "detections", "detection", "extendWithFaceDetection", "err", "useTinyLandmarkNet", "DetectAllFaceLandmarksTask", "PredictAllFaceExpressionsTask", "PredictAllAgeAndGenderTask", "DetectSingleFaceTask", "faceDetections", "faceDetectionWithHighestScore", "faceDetection", "DetectSingleFaceLandmarksTask", "PredictSingleFaceExpressionsTask", "PredictSingleAgeAndGenderTask", "detectSingleFace", "input", "options", "SsdMobilenetv1Options", "DetectSingleFaceTask", "detectAllFaces", "DetectAllFacesTask", "allFacesSsdMobilenetv1", "input", "minConfidence", "detectAllFaces", "SsdMobilenetv1Options", "allFacesTinyYolov2", "forwardParams", "TinyYolov2Options", "allFaces", "euclideanDistance", "arr1", "arr2", "desc1", "desc2", "val", "i", "res", "diff", "FaceMatcher", "inputs", "distanceThreshold", "inputArray", "count", "createUniqueLabel", "desc", "LabeledFaceDescriptors", "queryDescriptor", "descriptors", "d", "euclideanDistance", "d1", "d2", "label", "FaceMatch", "best", "curr", "bestMatch", "ld", "json", "labeledDescriptors", "createTinyFaceDetector", "weights", "net", "TinyFaceDetector", "resizeResults", "results", "dimensions", "width", "height", "Dimensions", "obj", "isWithFaceLandmarks", "resizedDetection", "resizedLandmarks", "extendWithFaceLandmarks", "extendWithFaceDetection", "isWithFaceDetection", "FaceLandmarks", "FaceDetection", "version"]
+  "sourcesContent": ["/*\n  Face-API\n  homepage: <https://github.com/vladmandic/face-api>\n  author: <https://github.com/vladmandic>'\n*/\n\nvar __create = Object.create;\nvar __defProp = Object.defineProperty;\nvar __getOwnPropDesc = Object.getOwnPropertyDescriptor;\nvar __getOwnPropNames = Object.getOwnPropertyNames;\nvar __getProtoOf = Object.getPrototypeOf;\nvar __hasOwnProp = Object.prototype.hasOwnProperty;\nvar __require = /* @__PURE__ */ ((x) => typeof require !== \"undefined\" ? require : typeof Proxy !== \"undefined\" ? new Proxy(x, {\n  get: (a, b) => (typeof require !== \"undefined\" ? require : a)[b]\n}) : x)(function(x) {\n  if (typeof require !== \"undefined\")\n    return require.apply(this, arguments);\n  throw new Error('Dynamic require of \"' + x + '\" is not supported');\n});\nvar __commonJS = (cb, mod4) => function __require2() {\n  return mod4 || (0, cb[__getOwnPropNames(cb)[0]])((mod4 = { exports: {} }).exports, mod4), mod4.exports;\n};\nvar __export = (target, all5) => {\n  for (var name in all5)\n    __defProp(target, name, { get: all5[name], enumerable: true });\n};\nvar __copyProps = (to, from, except, desc) => {\n  if (from && typeof from === \"object\" || typeof from === \"function\") {\n    for (let key of __getOwnPropNames(from))\n      if (!__hasOwnProp.call(to, key) && key !== except)\n        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });\n  }\n  return to;\n};\nvar __toESM = (mod4, isNodeMode, target) => (target = mod4 != null ? __create(__getProtoOf(mod4)) : {}, __copyProps(\n  isNodeMode || !mod4 || !mod4.__esModule ? __defProp(target, \"default\", { value: mod4, enumerable: true }) : target,\n  mod4\n));\n\n// node_modules/long/src/long.js\nvar require_long = __commonJS({\n  \"node_modules/long/src/long.js\"(exports, module) {\n    module.exports = Long2;\n    var wasm = null;\n    try {\n      wasm = new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([\n        0,\n        97,\n        115,\n        109,\n        1,\n        0,\n        0,\n        0,\n        1,\n        13,\n        2,\n        96,\n        0,\n        1,\n        127,\n        96,\n        4,\n        127,\n        127,\n        127,\n        127,\n        1,\n        127,\n        3,\n        7,\n        6,\n        0,\n        1,\n        1,\n        1,\n        1,\n        1,\n        6,\n        6,\n        1,\n        127,\n        1,\n        65,\n        0,\n        11,\n        7,\n        50,\n        6,\n        3,\n        109,\n        117,\n        108,\n        0,\n        1,\n        5,\n        100,\n        105,\n        118,\n        95,\n        115,\n        0,\n        2,\n        5,\n        100,\n        105,\n        118,\n        95,\n        117,\n        0,\n        3,\n        5,\n        114,\n        101,\n        109,\n        95,\n        115,\n        0,\n        4,\n        5,\n        114,\n        101,\n        109,\n        95,\n        117,\n        0,\n        5,\n        8,\n        103,\n        101,\n        116,\n        95,\n        104,\n        105,\n        103,\n        104,\n        0,\n        0,\n        10,\n        191,\n        1,\n        6,\n        4,\n        0,\n        35,\n        0,\n        11,\n        36,\n        1,\n        1,\n        126,\n        32,\n        0,\n        173,\n        32,\n        1,\n        173,\n        66,\n        32,\n        134,\n        132,\n        32,\n        2,\n        173,\n        32,\n        3,\n        173,\n        66,\n        32,\n        134,\n        132,\n        126,\n        34,\n        4,\n        66,\n        32,\n        135,\n        167,\n        36,\n        0,\n        32,\n        4,\n        167,\n        11,\n        36,\n        1,\n        1,\n        126,\n        32,\n        0,\n        173,\n        32,\n        1,\n        173,\n        66,\n        32,\n        134,\n        132,\n        32,\n        2,\n        173,\n        32,\n        3,\n        173,\n        66,\n        32,\n        134,\n        132,\n        127,\n        34,\n        4,\n        66,\n        32,\n        135,\n        167,\n        36,\n        0,\n        32,\n        4,\n        167,\n        11,\n        36,\n        1,\n        1,\n        126,\n        32,\n        0,\n        173,\n        32,\n        1,\n        173,\n        66,\n        32,\n        134,\n        132,\n        32,\n        2,\n        173,\n        32,\n        3,\n        173,\n        66,\n        32,\n        134,\n        132,\n        128,\n        34,\n        4,\n        66,\n        32,\n        135,\n        167,\n        36,\n        0,\n        32,\n        4,\n        167,\n        11,\n        36,\n        1,\n        1,\n        126,\n        32,\n        0,\n        173,\n        32,\n        1,\n        173,\n        66,\n        32,\n        134,\n        132,\n        32,\n        2,\n        173,\n        32,\n        3,\n        173,\n        66,\n        32,\n        134,\n        132,\n        129,\n        34,\n        4,\n        66,\n        32,\n        135,\n        167,\n        36,\n        0,\n        32,\n        4,\n        167,\n        11,\n        36,\n        1,\n        1,\n        126,\n        32,\n        0,\n        173,\n        32,\n        1,\n        173,\n        66,\n        32,\n        134,\n        132,\n        32,\n        2,\n        173,\n        32,\n        3,\n        173,\n        66,\n        32,\n        134,\n        132,\n        130,\n        34,\n        4,\n        66,\n        32,\n        135,\n        167,\n        36,\n        0,\n        32,\n        4,\n        167,\n        11\n      ])), {}).exports;\n    } catch (e) {\n    }\n    function Long2(low, high, unsigned) {\n      this.low = low | 0;\n      this.high = high | 0;\n      this.unsigned = !!unsigned;\n    }\n    Long2.prototype.__isLong__;\n    Object.defineProperty(Long2.prototype, \"__isLong__\", { value: true });\n    function isLong(obj) {\n      return (obj && obj[\"__isLong__\"]) === true;\n    }\n    Long2.isLong = isLong;\n    var INT_CACHE = {};\n    var UINT_CACHE = {};\n    function fromInt(value, unsigned) {\n      var obj, cachedObj, cache;\n      if (unsigned) {\n        value >>>= 0;\n        if (cache = 0 <= value && value < 256) {\n          cachedObj = UINT_CACHE[value];\n          if (cachedObj)\n            return cachedObj;\n        }\n        obj = fromBits(value, (value | 0) < 0 ? -1 : 0, true);\n        if (cache)\n          UINT_CACHE[value] = obj;\n        return obj;\n      } else {\n        value |= 0;\n        if (cache = -128 <= value && value < 128) {\n          cachedObj = INT_CACHE[value];\n          if (cachedObj)\n            return cachedObj;\n        }\n        obj = fromBits(value, value < 0 ? -1 : 0, false);\n        if (cache)\n          INT_CACHE[value] = obj;\n        return obj;\n      }\n    }\n    Long2.fromInt = fromInt;\n    function fromNumber(value, unsigned) {\n      if (isNaN(value))\n        return unsigned ? UZERO : ZERO;\n      if (unsigned) {\n        if (value < 0)\n          return UZERO;\n        if (value >= TWO_PWR_64_DBL)\n          return MAX_UNSIGNED_VALUE;\n      } else {\n        if (value <= -TWO_PWR_63_DBL)\n          return MIN_VALUE;\n        if (value + 1 >= TWO_PWR_63_DBL)\n          return MAX_VALUE;\n      }\n      if (value < 0)\n        return fromNumber(-value, unsigned).neg();\n      return fromBits(value % TWO_PWR_32_DBL | 0, value / TWO_PWR_32_DBL | 0, unsigned);\n    }\n    Long2.fromNumber = fromNumber;\n    function fromBits(lowBits, highBits, unsigned) {\n      return new Long2(lowBits, highBits, unsigned);\n    }\n    Long2.fromBits = fromBits;\n    var pow_dbl = Math.pow;\n    function fromString(str, unsigned, radix) {\n      if (str.length === 0)\n        throw Error(\"empty string\");\n      if (str === \"NaN\" || str === \"Infinity\" || str === \"+Infinity\" || str === \"-Infinity\")\n        return ZERO;\n      if (typeof unsigned === \"number\") {\n        radix = unsigned, unsigned = false;\n      } else {\n        unsigned = !!unsigned;\n      }\n      radix = radix || 10;\n      if (radix < 2 || 36 < radix)\n        throw RangeError(\"radix\");\n      var p2;\n      if ((p2 = str.indexOf(\"-\")) > 0)\n        throw Error(\"interior hyphen\");\n      else if (p2 === 0) {\n        return fromString(str.substring(1), unsigned, radix).neg();\n      }\n      var radixToPower = fromNumber(pow_dbl(radix, 8));\n      var result = ZERO;\n      for (var i = 0; i < str.length; i += 8) {\n        var size = Math.min(8, str.length - i), value = parseInt(str.substring(i, i + size), radix);\n        if (size < 8) {\n          var power = fromNumber(pow_dbl(radix, size));\n          result = result.mul(power).add(fromNumber(value));\n        } else {\n          result = result.mul(radixToPower);\n          result = result.add(fromNumber(value));\n        }\n      }\n      result.unsigned = unsigned;\n      return result;\n    }\n    Long2.fromString = fromString;\n    function fromValue(val, unsigned) {\n      if (typeof val === \"number\")\n        return fromNumber(val, unsigned);\n      if (typeof val === \"string\")\n        return fromString(val, unsigned);\n      return fromBits(val.low, val.high, typeof unsigned === \"boolean\" ? unsigned : val.unsigned);\n    }\n    Long2.fromValue = fromValue;\n    var TWO_PWR_16_DBL = 1 << 16;\n    var TWO_PWR_24_DBL = 1 << 24;\n    var TWO_PWR_32_DBL = TWO_PWR_16_DBL * TWO_PWR_16_DBL;\n    var TWO_PWR_64_DBL = TWO_PWR_32_DBL * TWO_PWR_32_DBL;\n    var TWO_PWR_63_DBL = TWO_PWR_64_DBL / 2;\n    var TWO_PWR_24 = fromInt(TWO_PWR_24_DBL);\n    var ZERO = fromInt(0);\n    Long2.ZERO = ZERO;\n    var UZERO = fromInt(0, true);\n    Long2.UZERO = UZERO;\n    var ONE = fromInt(1);\n    Long2.ONE = ONE;\n    var UONE = fromInt(1, true);\n    Long2.UONE = UONE;\n    var NEG_ONE = fromInt(-1);\n    Long2.NEG_ONE = NEG_ONE;\n    var MAX_VALUE = fromBits(4294967295 | 0, 2147483647 | 0, false);\n    Long2.MAX_VALUE = MAX_VALUE;\n    var MAX_UNSIGNED_VALUE = fromBits(4294967295 | 0, 4294967295 | 0, true);\n    Long2.MAX_UNSIGNED_VALUE = MAX_UNSIGNED_VALUE;\n    var MIN_VALUE = fromBits(0, 2147483648 | 0, false);\n    Long2.MIN_VALUE = MIN_VALUE;\n    var LongPrototype = Long2.prototype;\n    LongPrototype.toInt = function toInt() {\n      return this.unsigned ? this.low >>> 0 : this.low;\n    };\n    LongPrototype.toNumber = function toNumber() {\n      if (this.unsigned)\n        return (this.high >>> 0) * TWO_PWR_32_DBL + (this.low >>> 0);\n      return this.high * TWO_PWR_32_DBL + (this.low >>> 0);\n    };\n    LongPrototype.toString = function toString(radix) {\n      radix = radix || 10;\n      if (radix < 2 || 36 < radix)\n        throw RangeError(\"radix\");\n      if (this.isZero())\n        return \"0\";\n      if (this.isNegative()) {\n        if (this.eq(MIN_VALUE)) {\n          var radixLong = fromNumber(radix), div3 = this.div(radixLong), rem1 = div3.mul(radixLong).sub(this);\n          return div3.toString(radix) + rem1.toInt().toString(radix);\n        } else\n          return \"-\" + this.neg().toString(radix);\n      }\n      var radixToPower = fromNumber(pow_dbl(radix, 6), this.unsigned), rem = this;\n      var result = \"\";\n      while (true) {\n        var remDiv = rem.div(radixToPower), intval = rem.sub(remDiv.mul(radixToPower)).toInt() >>> 0, digits = intval.toString(radix);\n        rem = remDiv;\n        if (rem.isZero())\n          return digits + result;\n        else {\n          while (digits.length < 6)\n            digits = \"0\" + digits;\n          result = \"\" + digits + result;\n        }\n      }\n    };\n    LongPrototype.getHighBits = function getHighBits() {\n      return this.high;\n    };\n    LongPrototype.getHighBitsUnsigned = function getHighBitsUnsigned() {\n      return this.high >>> 0;\n    };\n    LongPrototype.getLowBits = function getLowBits() {\n      return this.low;\n    };\n    LongPrototype.getLowBitsUnsigned = function getLowBitsUnsigned() {\n      return this.low >>> 0;\n    };\n    LongPrototype.getNumBitsAbs = function getNumBitsAbs() {\n      if (this.isNegative())\n        return this.eq(MIN_VALUE) ? 64 : this.neg().getNumBitsAbs();\n      var val = this.high != 0 ? this.high : this.low;\n      for (var bit = 31; bit > 0; bit--)\n        if ((val & 1 << bit) != 0)\n          break;\n      return this.high != 0 ? bit + 33 : bit + 1;\n    };\n    LongPrototype.isZero = function isZero() {\n      return this.high === 0 && this.low === 0;\n    };\n    LongPrototype.eqz = LongPrototype.isZero;\n    LongPrototype.isNegative = function isNegative() {\n      return !this.unsigned && this.high < 0;\n    };\n    LongPrototype.isPositive = function isPositive() {\n      return this.unsigned || this.high >= 0;\n    };\n    LongPrototype.isOdd = function isOdd() {\n      return (this.low & 1) === 1;\n    };\n    LongPrototype.isEven = function isEven2() {\n      return (this.low & 1) === 0;\n    };\n    LongPrototype.equals = function equals(other) {\n      if (!isLong(other))\n        other = fromValue(other);\n      if (this.unsigned !== other.unsigned && this.high >>> 31 === 1 && other.high >>> 31 === 1)\n        return false;\n      return this.high === other.high && this.low === other.low;\n    };\n    LongPrototype.eq = LongPrototype.equals;\n    LongPrototype.notEquals = function notEquals(other) {\n      return !this.eq(other);\n    };\n    LongPrototype.neq = LongPrototype.notEquals;\n    LongPrototype.ne = LongPrototype.notEquals;\n    LongPrototype.lessThan = function lessThan(other) {\n      return this.comp(other) < 0;\n    };\n    LongPrototype.lt = LongPrototype.lessThan;\n    LongPrototype.lessThanOrEqual = function lessThanOrEqual(other) {\n      return this.comp(other) <= 0;\n    };\n    LongPrototype.lte = LongPrototype.lessThanOrEqual;\n    LongPrototype.le = LongPrototype.lessThanOrEqual;\n    LongPrototype.greaterThan = function greaterThan(other) {\n      return this.comp(other) > 0;\n    };\n    LongPrototype.gt = LongPrototype.greaterThan;\n    LongPrototype.greaterThanOrEqual = function greaterThanOrEqual(other) {\n      return this.comp(other) >= 0;\n    };\n    LongPrototype.gte = LongPrototype.greaterThanOrEqual;\n    LongPrototype.ge = LongPrototype.greaterThanOrEqual;\n    LongPrototype.compare = function compare(other) {\n      if (!isLong(other))\n        other = fromValue(other);\n      if (this.eq(other))\n        return 0;\n      var thisNeg = this.isNegative(), otherNeg = other.isNegative();\n      if (thisNeg && !otherNeg)\n        return -1;\n      if (!thisNeg && otherNeg)\n        return 1;\n      if (!this.unsigned)\n        return this.sub(other).isNegative() ? -1 : 1;\n      return other.high >>> 0 > this.high >>> 0 || other.high === this.high && other.low >>> 0 > this.low >>> 0 ? -1 : 1;\n    };\n    LongPrototype.comp = LongPrototype.compare;\n    LongPrototype.negate = function negate() {\n      if (!this.unsigned && this.eq(MIN_VALUE))\n        return MIN_VALUE;\n      return this.not().add(ONE);\n    };\n    LongPrototype.neg = LongPrototype.negate;\n    LongPrototype.add = function add5(addend) {\n      if (!isLong(addend))\n        addend = fromValue(addend);\n      var a48 = this.high >>> 16;\n      var a32 = this.high & 65535;\n      var a16 = this.low >>> 16;\n      var a00 = this.low & 65535;\n      var b48 = addend.high >>> 16;\n      var b32 = addend.high & 65535;\n      var b16 = addend.low >>> 16;\n      var b00 = addend.low & 65535;\n      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;\n      c00 += a00 + b00;\n      c16 += c00 >>> 16;\n      c00 &= 65535;\n      c16 += a16 + b16;\n      c32 += c16 >>> 16;\n      c16 &= 65535;\n      c32 += a32 + b32;\n      c48 += c32 >>> 16;\n      c32 &= 65535;\n      c48 += a48 + b48;\n      c48 &= 65535;\n      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);\n    };\n    LongPrototype.subtract = function subtract(subtrahend) {\n      if (!isLong(subtrahend))\n        subtrahend = fromValue(subtrahend);\n      return this.add(subtrahend.neg());\n    };\n    LongPrototype.sub = LongPrototype.subtract;\n    LongPrototype.multiply = function multiply4(multiplier) {\n      if (this.isZero())\n        return ZERO;\n      if (!isLong(multiplier))\n        multiplier = fromValue(multiplier);\n      if (wasm) {\n        var low = wasm.mul(\n          this.low,\n          this.high,\n          multiplier.low,\n          multiplier.high\n        );\n        return fromBits(low, wasm.get_high(), this.unsigned);\n      }\n      if (multiplier.isZero())\n        return ZERO;\n      if (this.eq(MIN_VALUE))\n        return multiplier.isOdd() ? MIN_VALUE : ZERO;\n      if (multiplier.eq(MIN_VALUE))\n        return this.isOdd() ? MIN_VALUE : ZERO;\n      if (this.isNegative()) {\n        if (multiplier.isNegative())\n          return this.neg().mul(multiplier.neg());\n        else\n          return this.neg().mul(multiplier).neg();\n      } else if (multiplier.isNegative())\n        return this.mul(multiplier.neg()).neg();\n      if (this.lt(TWO_PWR_24) && multiplier.lt(TWO_PWR_24))\n        return fromNumber(this.toNumber() * multiplier.toNumber(), this.unsigned);\n      var a48 = this.high >>> 16;\n      var a32 = this.high & 65535;\n      var a16 = this.low >>> 16;\n      var a00 = this.low & 65535;\n      var b48 = multiplier.high >>> 16;\n      var b32 = multiplier.high & 65535;\n      var b16 = multiplier.low >>> 16;\n      var b00 = multiplier.low & 65535;\n      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;\n      c00 += a00 * b00;\n      c16 += c00 >>> 16;\n      c00 &= 65535;\n      c16 += a16 * b00;\n      c32 += c16 >>> 16;\n      c16 &= 65535;\n      c16 += a00 * b16;\n      c32 += c16 >>> 16;\n      c16 &= 65535;\n      c32 += a32 * b00;\n      c48 += c32 >>> 16;\n      c32 &= 65535;\n      c32 += a16 * b16;\n      c48 += c32 >>> 16;\n      c32 &= 65535;\n      c32 += a00 * b32;\n      c48 += c32 >>> 16;\n      c32 &= 65535;\n      c48 += a48 * b00 + a32 * b16 + a16 * b32 + a00 * b48;\n      c48 &= 65535;\n      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);\n    };\n    LongPrototype.mul = LongPrototype.multiply;\n    LongPrototype.divide = function divide(divisor) {\n      if (!isLong(divisor))\n        divisor = fromValue(divisor);\n      if (divisor.isZero())\n        throw Error(\"division by zero\");\n      if (wasm) {\n        if (!this.unsigned && this.high === -2147483648 && divisor.low === -1 && divisor.high === -1) {\n          return this;\n        }\n        var low = (this.unsigned ? wasm.div_u : wasm.div_s)(\n          this.low,\n          this.high,\n          divisor.low,\n          divisor.high\n        );\n        return fromBits(low, wasm.get_high(), this.unsigned);\n      }\n      if (this.isZero())\n        return this.unsigned ? UZERO : ZERO;\n      var approx, rem, res;\n      if (!this.unsigned) {\n        if (this.eq(MIN_VALUE)) {\n          if (divisor.eq(ONE) || divisor.eq(NEG_ONE))\n            return MIN_VALUE;\n          else if (divisor.eq(MIN_VALUE))\n            return ONE;\n          else {\n            var halfThis = this.shr(1);\n            approx = halfThis.div(divisor).shl(1);\n            if (approx.eq(ZERO)) {\n              return divisor.isNegative() ? ONE : NEG_ONE;\n            } else {\n              rem = this.sub(divisor.mul(approx));\n              res = approx.add(rem.div(divisor));\n              return res;\n            }\n          }\n        } else if (divisor.eq(MIN_VALUE))\n          return this.unsigned ? UZERO : ZERO;\n        if (this.isNegative()) {\n          if (divisor.isNegative())\n            return this.neg().div(divisor.neg());\n          return this.neg().div(divisor).neg();\n        } else if (divisor.isNegative())\n          return this.div(divisor.neg()).neg();\n        res = ZERO;\n      } else {\n        if (!divisor.unsigned)\n          divisor = divisor.toUnsigned();\n        if (divisor.gt(this))\n          return UZERO;\n        if (divisor.gt(this.shru(1)))\n          return UONE;\n        res = UZERO;\n      }\n      rem = this;\n      while (rem.gte(divisor)) {\n        approx = Math.max(1, Math.floor(rem.toNumber() / divisor.toNumber()));\n        var log22 = Math.ceil(Math.log(approx) / Math.LN2), delta = log22 <= 48 ? 1 : pow_dbl(2, log22 - 48), approxRes = fromNumber(approx), approxRem = approxRes.mul(divisor);\n        while (approxRem.isNegative() || approxRem.gt(rem)) {\n          approx -= delta;\n          approxRes = fromNumber(approx, this.unsigned);\n          approxRem = approxRes.mul(divisor);\n        }\n        if (approxRes.isZero())\n          approxRes = ONE;\n        res = res.add(approxRes);\n        rem = rem.sub(approxRem);\n      }\n      return res;\n    };\n    LongPrototype.div = LongPrototype.divide;\n    LongPrototype.modulo = function modulo(divisor) {\n      if (!isLong(divisor))\n        divisor = fromValue(divisor);\n      if (wasm) {\n        var low = (this.unsigned ? wasm.rem_u : wasm.rem_s)(\n          this.low,\n          this.high,\n          divisor.low,\n          divisor.high\n        );\n        return fromBits(low, wasm.get_high(), this.unsigned);\n      }\n      return this.sub(this.div(divisor).mul(divisor));\n    };\n    LongPrototype.mod = LongPrototype.modulo;\n    LongPrototype.rem = LongPrototype.modulo;\n    LongPrototype.not = function not() {\n      return fromBits(~this.low, ~this.high, this.unsigned);\n    };\n    LongPrototype.and = function and(other) {\n      if (!isLong(other))\n        other = fromValue(other);\n      return fromBits(this.low & other.low, this.high & other.high, this.unsigned);\n    };\n    LongPrototype.or = function or(other) {\n      if (!isLong(other))\n        other = fromValue(other);\n      return fromBits(this.low | other.low, this.high | other.high, this.unsigned);\n    };\n    LongPrototype.xor = function xor(other) {\n      if (!isLong(other))\n        other = fromValue(other);\n      return fromBits(this.low ^ other.low, this.high ^ other.high, this.unsigned);\n    };\n    LongPrototype.shiftLeft = function shiftLeft(numBits) {\n      if (isLong(numBits))\n        numBits = numBits.toInt();\n      if ((numBits &= 63) === 0)\n        return this;\n      else if (numBits < 32)\n        return fromBits(this.low << numBits, this.high << numBits | this.low >>> 32 - numBits, this.unsigned);\n      else\n        return fromBits(0, this.low << numBits - 32, this.unsigned);\n    };\n    LongPrototype.shl = LongPrototype.shiftLeft;\n    LongPrototype.shiftRight = function shiftRight(numBits) {\n      if (isLong(numBits))\n        numBits = numBits.toInt();\n      if ((numBits &= 63) === 0)\n        return this;\n      else if (numBits < 32)\n        return fromBits(this.low >>> numBits | this.high << 32 - numBits, this.high >> numBits, this.unsigned);\n      else\n        return fromBits(this.high >> numBits - 32, this.high >= 0 ? 0 : -1, this.unsigned);\n    };\n    LongPrototype.shr = LongPrototype.shiftRight;\n    LongPrototype.shiftRightUnsigned = function shiftRightUnsigned(numBits) {\n      if (isLong(numBits))\n        numBits = numBits.toInt();\n      numBits &= 63;\n      if (numBits === 0)\n        return this;\n      else {\n        var high = this.high;\n        if (numBits < 32) {\n          var low = this.low;\n          return fromBits(low >>> numBits | high << 32 - numBits, high >>> numBits, this.unsigned);\n        } else if (numBits === 32)\n          return fromBits(high, 0, this.unsigned);\n        else\n          return fromBits(high >>> numBits - 32, 0, this.unsigned);\n      }\n    };\n    LongPrototype.shru = LongPrototype.shiftRightUnsigned;\n    LongPrototype.shr_u = LongPrototype.shiftRightUnsigned;\n    LongPrototype.toSigned = function toSigned() {\n      if (!this.unsigned)\n        return this;\n      return fromBits(this.low, this.high, false);\n    };\n    LongPrototype.toUnsigned = function toUnsigned() {\n      if (this.unsigned)\n        return this;\n      return fromBits(this.low, this.high, true);\n    };\n    LongPrototype.toBytes = function toBytes(le) {\n      return le ? this.toBytesLE() : this.toBytesBE();\n    };\n    LongPrototype.toBytesLE = function toBytesLE() {\n      var hi = this.high, lo = this.low;\n      return [\n        lo & 255,\n        lo >>> 8 & 255,\n        lo >>> 16 & 255,\n        lo >>> 24,\n        hi & 255,\n        hi >>> 8 & 255,\n        hi >>> 16 & 255,\n        hi >>> 24\n      ];\n    };\n    LongPrototype.toBytesBE = function toBytesBE() {\n      var hi = this.high, lo = this.low;\n      return [\n        hi >>> 24,\n        hi >>> 16 & 255,\n        hi >>> 8 & 255,\n        hi & 255,\n        lo >>> 24,\n        lo >>> 16 & 255,\n        lo >>> 8 & 255,\n        lo & 255\n      ];\n    };\n    Long2.fromBytes = function fromBytes(bytes, unsigned, le) {\n      return le ? Long2.fromBytesLE(bytes, unsigned) : Long2.fromBytesBE(bytes, unsigned);\n    };\n    Long2.fromBytesLE = function fromBytesLE(bytes, unsigned) {\n      return new Long2(\n        bytes[0] | bytes[1] << 8 | bytes[2] << 16 | bytes[3] << 24,\n        bytes[4] | bytes[5] << 8 | bytes[6] << 16 | bytes[7] << 24,\n        unsigned\n      );\n    };\n    Long2.fromBytesBE = function fromBytesBE(bytes, unsigned) {\n      return new Long2(\n        bytes[4] << 24 | bytes[5] << 16 | bytes[6] << 8 | bytes[7],\n        bytes[0] << 24 | bytes[1] << 16 | bytes[2] << 8 | bytes[3],\n        unsigned\n      );\n    };\n  }\n});\n\n// (disabled):node_modules/node-fetch/browser.js\nvar require_browser = __commonJS({\n  \"(disabled):node_modules/node-fetch/browser.js\"() {\n  }\n});\n\n// (disabled):util\nvar require_util = __commonJS({\n  \"(disabled):util\"() {\n  }\n});\n\n// node_modules/seedrandom/lib/alea.js\nvar require_alea = __commonJS({\n  \"node_modules/seedrandom/lib/alea.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function Alea(seed) {\n        var me = this, mash = Mash();\n        me.next = function() {\n          var t = 2091639 * me.s0 + me.c * 23283064365386963e-26;\n          me.s0 = me.s1;\n          me.s1 = me.s2;\n          return me.s2 = t - (me.c = t | 0);\n        };\n        me.c = 1;\n        me.s0 = mash(\" \");\n        me.s1 = mash(\" \");\n        me.s2 = mash(\" \");\n        me.s0 -= mash(seed);\n        if (me.s0 < 0) {\n          me.s0 += 1;\n        }\n        me.s1 -= mash(seed);\n        if (me.s1 < 0) {\n          me.s1 += 1;\n        }\n        me.s2 -= mash(seed);\n        if (me.s2 < 0) {\n          me.s2 += 1;\n        }\n        mash = null;\n      }\n      function copy(f, t) {\n        t.c = f.c;\n        t.s0 = f.s0;\n        t.s1 = f.s1;\n        t.s2 = f.s2;\n        return t;\n      }\n      function impl(seed, opts) {\n        var xg = new Alea(seed), state = opts && opts.state, prng = xg.next;\n        prng.int32 = function() {\n          return xg.next() * 4294967296 | 0;\n        };\n        prng.double = function() {\n          return prng() + (prng() * 2097152 | 0) * 11102230246251565e-32;\n        };\n        prng.quick = prng;\n        if (state) {\n          if (typeof state == \"object\")\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      function Mash() {\n        var n = 4022871197;\n        var mash = function(data) {\n          data = String(data);\n          for (var i = 0; i < data.length; i++) {\n            n += data.charCodeAt(i);\n            var h = 0.02519603282416938 * n;\n            n = h >>> 0;\n            h -= n;\n            h *= n;\n            n = h >>> 0;\n            h -= n;\n            n += h * 4294967296;\n          }\n          return (n >>> 0) * 23283064365386963e-26;\n        };\n        return mash;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.alea = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// node_modules/seedrandom/lib/xor128.js\nvar require_xor128 = __commonJS({\n  \"node_modules/seedrandom/lib/xor128.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function XorGen(seed) {\n        var me = this, strseed = \"\";\n        me.x = 0;\n        me.y = 0;\n        me.z = 0;\n        me.w = 0;\n        me.next = function() {\n          var t = me.x ^ me.x << 11;\n          me.x = me.y;\n          me.y = me.z;\n          me.z = me.w;\n          return me.w ^= me.w >>> 19 ^ t ^ t >>> 8;\n        };\n        if (seed === (seed | 0)) {\n          me.x = seed;\n        } else {\n          strseed += seed;\n        }\n        for (var k = 0; k < strseed.length + 64; k++) {\n          me.x ^= strseed.charCodeAt(k) | 0;\n          me.next();\n        }\n      }\n      function copy(f, t) {\n        t.x = f.x;\n        t.y = f.y;\n        t.z = f.z;\n        t.w = f.w;\n        return t;\n      }\n      function impl(seed, opts) {\n        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {\n          return (xg.next() >>> 0) / 4294967296;\n        };\n        prng.double = function() {\n          do {\n            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);\n          } while (result === 0);\n          return result;\n        };\n        prng.int32 = xg.next;\n        prng.quick = prng;\n        if (state) {\n          if (typeof state == \"object\")\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.xor128 = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// node_modules/seedrandom/lib/xorwow.js\nvar require_xorwow = __commonJS({\n  \"node_modules/seedrandom/lib/xorwow.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function XorGen(seed) {\n        var me = this, strseed = \"\";\n        me.next = function() {\n          var t = me.x ^ me.x >>> 2;\n          me.x = me.y;\n          me.y = me.z;\n          me.z = me.w;\n          me.w = me.v;\n          return (me.d = me.d + 362437 | 0) + (me.v = me.v ^ me.v << 4 ^ (t ^ t << 1)) | 0;\n        };\n        me.x = 0;\n        me.y = 0;\n        me.z = 0;\n        me.w = 0;\n        me.v = 0;\n        if (seed === (seed | 0)) {\n          me.x = seed;\n        } else {\n          strseed += seed;\n        }\n        for (var k = 0; k < strseed.length + 64; k++) {\n          me.x ^= strseed.charCodeAt(k) | 0;\n          if (k == strseed.length) {\n            me.d = me.x << 10 ^ me.x >>> 4;\n          }\n          me.next();\n        }\n      }\n      function copy(f, t) {\n        t.x = f.x;\n        t.y = f.y;\n        t.z = f.z;\n        t.w = f.w;\n        t.v = f.v;\n        t.d = f.d;\n        return t;\n      }\n      function impl(seed, opts) {\n        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {\n          return (xg.next() >>> 0) / 4294967296;\n        };\n        prng.double = function() {\n          do {\n            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);\n          } while (result === 0);\n          return result;\n        };\n        prng.int32 = xg.next;\n        prng.quick = prng;\n        if (state) {\n          if (typeof state == \"object\")\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.xorwow = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// node_modules/seedrandom/lib/xorshift7.js\nvar require_xorshift7 = __commonJS({\n  \"node_modules/seedrandom/lib/xorshift7.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function XorGen(seed) {\n        var me = this;\n        me.next = function() {\n          var X = me.x, i = me.i, t, v, w;\n          t = X[i];\n          t ^= t >>> 7;\n          v = t ^ t << 24;\n          t = X[i + 1 & 7];\n          v ^= t ^ t >>> 10;\n          t = X[i + 3 & 7];\n          v ^= t ^ t >>> 3;\n          t = X[i + 4 & 7];\n          v ^= t ^ t << 7;\n          t = X[i + 7 & 7];\n          t = t ^ t << 13;\n          v ^= t ^ t << 9;\n          X[i] = v;\n          me.i = i + 1 & 7;\n          return v;\n        };\n        function init2(me2, seed2) {\n          var j, w, X = [];\n          if (seed2 === (seed2 | 0)) {\n            w = X[0] = seed2;\n          } else {\n            seed2 = \"\" + seed2;\n            for (j = 0; j < seed2.length; ++j) {\n              X[j & 7] = X[j & 7] << 15 ^ seed2.charCodeAt(j) + X[j + 1 & 7] << 13;\n            }\n          }\n          while (X.length < 8)\n            X.push(0);\n          for (j = 0; j < 8 && X[j] === 0; ++j)\n            ;\n          if (j == 8)\n            w = X[7] = -1;\n          else\n            w = X[j];\n          me2.x = X;\n          me2.i = 0;\n          for (j = 256; j > 0; --j) {\n            me2.next();\n          }\n        }\n        init2(me, seed);\n      }\n      function copy(f, t) {\n        t.x = f.x.slice();\n        t.i = f.i;\n        return t;\n      }\n      function impl(seed, opts) {\n        if (seed == null)\n          seed = +new Date();\n        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {\n          return (xg.next() >>> 0) / 4294967296;\n        };\n        prng.double = function() {\n          do {\n            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);\n          } while (result === 0);\n          return result;\n        };\n        prng.int32 = xg.next;\n        prng.quick = prng;\n        if (state) {\n          if (state.x)\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.xorshift7 = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// node_modules/seedrandom/lib/xor4096.js\nvar require_xor4096 = __commonJS({\n  \"node_modules/seedrandom/lib/xor4096.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function XorGen(seed) {\n        var me = this;\n        me.next = function() {\n          var w = me.w, X = me.X, i = me.i, t, v;\n          me.w = w = w + 1640531527 | 0;\n          v = X[i + 34 & 127];\n          t = X[i = i + 1 & 127];\n          v ^= v << 13;\n          t ^= t << 17;\n          v ^= v >>> 15;\n          t ^= t >>> 12;\n          v = X[i] = v ^ t;\n          me.i = i;\n          return v + (w ^ w >>> 16) | 0;\n        };\n        function init2(me2, seed2) {\n          var t, v, i, j, w, X = [], limit = 128;\n          if (seed2 === (seed2 | 0)) {\n            v = seed2;\n            seed2 = null;\n          } else {\n            seed2 = seed2 + \"\\0\";\n            v = 0;\n            limit = Math.max(limit, seed2.length);\n          }\n          for (i = 0, j = -32; j < limit; ++j) {\n            if (seed2)\n              v ^= seed2.charCodeAt((j + 32) % seed2.length);\n            if (j === 0)\n              w = v;\n            v ^= v << 10;\n            v ^= v >>> 15;\n            v ^= v << 4;\n            v ^= v >>> 13;\n            if (j >= 0) {\n              w = w + 1640531527 | 0;\n              t = X[j & 127] ^= v + w;\n              i = 0 == t ? i + 1 : 0;\n            }\n          }\n          if (i >= 128) {\n            X[(seed2 && seed2.length || 0) & 127] = -1;\n          }\n          i = 127;\n          for (j = 4 * 128; j > 0; --j) {\n            v = X[i + 34 & 127];\n            t = X[i = i + 1 & 127];\n            v ^= v << 13;\n            t ^= t << 17;\n            v ^= v >>> 15;\n            t ^= t >>> 12;\n            X[i] = v ^ t;\n          }\n          me2.w = w;\n          me2.X = X;\n          me2.i = i;\n        }\n        init2(me, seed);\n      }\n      function copy(f, t) {\n        t.i = f.i;\n        t.w = f.w;\n        t.X = f.X.slice();\n        return t;\n      }\n      ;\n      function impl(seed, opts) {\n        if (seed == null)\n          seed = +new Date();\n        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {\n          return (xg.next() >>> 0) / 4294967296;\n        };\n        prng.double = function() {\n          do {\n            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);\n          } while (result === 0);\n          return result;\n        };\n        prng.int32 = xg.next;\n        prng.quick = prng;\n        if (state) {\n          if (state.X)\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.xor4096 = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// node_modules/seedrandom/lib/tychei.js\nvar require_tychei = __commonJS({\n  \"node_modules/seedrandom/lib/tychei.js\"(exports, module) {\n    (function(global2, module2, define2) {\n      function XorGen(seed) {\n        var me = this, strseed = \"\";\n        me.next = function() {\n          var b = me.b, c = me.c, d = me.d, a = me.a;\n          b = b << 25 ^ b >>> 7 ^ c;\n          c = c - d | 0;\n          d = d << 24 ^ d >>> 8 ^ a;\n          a = a - b | 0;\n          me.b = b = b << 20 ^ b >>> 12 ^ c;\n          me.c = c = c - d | 0;\n          me.d = d << 16 ^ c >>> 16 ^ a;\n          return me.a = a - b | 0;\n        };\n        me.a = 0;\n        me.b = 0;\n        me.c = 2654435769 | 0;\n        me.d = 1367130551;\n        if (seed === Math.floor(seed)) {\n          me.a = seed / 4294967296 | 0;\n          me.b = seed | 0;\n        } else {\n          strseed += seed;\n        }\n        for (var k = 0; k < strseed.length + 20; k++) {\n          me.b ^= strseed.charCodeAt(k) | 0;\n          me.next();\n        }\n      }\n      function copy(f, t) {\n        t.a = f.a;\n        t.b = f.b;\n        t.c = f.c;\n        t.d = f.d;\n        return t;\n      }\n      ;\n      function impl(seed, opts) {\n        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {\n          return (xg.next() >>> 0) / 4294967296;\n        };\n        prng.double = function() {\n          do {\n            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);\n          } while (result === 0);\n          return result;\n        };\n        prng.int32 = xg.next;\n        prng.quick = prng;\n        if (state) {\n          if (typeof state == \"object\")\n            copy(state, xg);\n          prng.state = function() {\n            return copy(xg, {});\n          };\n        }\n        return prng;\n      }\n      if (module2 && module2.exports) {\n        module2.exports = impl;\n      } else if (define2 && define2.amd) {\n        define2(function() {\n          return impl;\n        });\n      } else {\n        this.tychei = impl;\n      }\n    })(\n      exports,\n      typeof module == \"object\" && module,\n      typeof define == \"function\" && define\n    );\n  }\n});\n\n// (disabled):crypto\nvar require_crypto = __commonJS({\n  \"(disabled):crypto\"() {\n  }\n});\n\n// node_modules/seedrandom/seedrandom.js\nvar require_seedrandom = __commonJS({\n  \"node_modules/seedrandom/seedrandom.js\"(exports, module) {\n    (function(global2, pool3, math) {\n      var width = 256, chunks = 6, digits = 52, rngname = \"random\", startdenom = math.pow(width, chunks), significance = math.pow(2, digits), overflow = significance * 2, mask = width - 1, nodecrypto;\n      function seedrandom5(seed, options, callback) {\n        var key = [];\n        options = options == true ? { entropy: true } : options || {};\n        var shortseed = mixkey(flatten4(\n          options.entropy ? [seed, tostring(pool3)] : seed == null ? autoseed() : seed,\n          3\n        ), key);\n        var arc4 = new ARC4(key);\n        var prng = function() {\n          var n = arc4.g(chunks), d = startdenom, x = 0;\n          while (n < significance) {\n            n = (n + x) * width;\n            d *= width;\n            x = arc4.g(1);\n          }\n          while (n >= overflow) {\n            n /= 2;\n            d /= 2;\n            x >>>= 1;\n          }\n          return (n + x) / d;\n        };\n        prng.int32 = function() {\n          return arc4.g(4) | 0;\n        };\n        prng.quick = function() {\n          return arc4.g(4) / 4294967296;\n        };\n        prng.double = prng;\n        mixkey(tostring(arc4.S), pool3);\n        return (options.pass || callback || function(prng2, seed2, is_math_call, state) {\n          if (state) {\n            if (state.S) {\n              copy(state, arc4);\n            }\n            prng2.state = function() {\n              return copy(arc4, {});\n            };\n          }\n          if (is_math_call) {\n            math[rngname] = prng2;\n            return seed2;\n          } else\n            return prng2;\n        })(\n          prng,\n          shortseed,\n          \"global\" in options ? options.global : this == math,\n          options.state\n        );\n      }\n      function ARC4(key) {\n        var t, keylen = key.length, me = this, i = 0, j = me.i = me.j = 0, s = me.S = [];\n        if (!keylen) {\n          key = [keylen++];\n        }\n        while (i < width) {\n          s[i] = i++;\n        }\n        for (i = 0; i < width; i++) {\n          s[i] = s[j = mask & j + key[i % keylen] + (t = s[i])];\n          s[j] = t;\n        }\n        (me.g = function(count2) {\n          var t2, r = 0, i2 = me.i, j2 = me.j, s2 = me.S;\n          while (count2--) {\n            t2 = s2[i2 = mask & i2 + 1];\n            r = r * width + s2[mask & (s2[i2] = s2[j2 = mask & j2 + t2]) + (s2[j2] = t2)];\n          }\n          me.i = i2;\n          me.j = j2;\n          return r;\n        })(width);\n      }\n      function copy(f, t) {\n        t.i = f.i;\n        t.j = f.j;\n        t.S = f.S.slice();\n        return t;\n      }\n      ;\n      function flatten4(obj, depth) {\n        var result = [], typ = typeof obj, prop;\n        if (depth && typ == \"object\") {\n          for (prop in obj) {\n            try {\n              result.push(flatten4(obj[prop], depth - 1));\n            } catch (e) {\n            }\n          }\n        }\n        return result.length ? result : typ == \"string\" ? obj : obj + \"\\0\";\n      }\n      function mixkey(seed, key) {\n        var stringseed = seed + \"\", smear, j = 0;\n        while (j < stringseed.length) {\n          key[mask & j] = mask & (smear ^= key[mask & j] * 19) + stringseed.charCodeAt(j++);\n        }\n        return tostring(key);\n      }\n      function autoseed() {\n        try {\n          var out;\n          if (nodecrypto && (out = nodecrypto.randomBytes)) {\n            out = out(width);\n          } else {\n            out = new Uint8Array(width);\n            (global2.crypto || global2.msCrypto).getRandomValues(out);\n          }\n          return tostring(out);\n        } catch (e) {\n          var browser = global2.navigator, plugins = browser && browser.plugins;\n          return [+new Date(), global2, plugins, global2.screen, tostring(pool3)];\n        }\n      }\n      function tostring(a) {\n        return String.fromCharCode.apply(0, a);\n      }\n      mixkey(math.random(), pool3);\n      if (typeof module == \"object\" && module.exports) {\n        module.exports = seedrandom5;\n        try {\n          nodecrypto = require_crypto();\n        } catch (ex) {\n        }\n      } else if (typeof define == \"function\" && define.amd) {\n        define(function() {\n          return seedrandom5;\n        });\n      } else {\n        math[\"seed\" + rngname] = seedrandom5;\n      }\n    })(\n      typeof self !== \"undefined\" ? self : exports,\n      [],\n      Math\n    );\n  }\n});\n\n// node_modules/seedrandom/index.js\nvar require_seedrandom2 = __commonJS({\n  \"node_modules/seedrandom/index.js\"(exports, module) {\n    var alea5 = require_alea();\n    var xor128 = require_xor128();\n    var xorwow = require_xorwow();\n    var xorshift7 = require_xorshift7();\n    var xor4096 = require_xor4096();\n    var tychei = require_tychei();\n    var sr = require_seedrandom();\n    sr.alea = alea5;\n    sr.xor128 = xor128;\n    sr.xorwow = xorwow;\n    sr.xorshift7 = xorshift7;\n    sr.xor4096 = xor4096;\n    sr.tychei = tychei;\n    module.exports = sr;\n  }\n});\n\n// (disabled):node_modules/string_decoder/lib/string_decoder.js\nvar require_string_decoder = __commonJS({\n  \"(disabled):node_modules/string_decoder/lib/string_decoder.js\"() {\n  }\n});\n\n// (disabled):fs\nvar require_fs = __commonJS({\n  \"(disabled):fs\"() {\n  }\n});\n\n// (disabled):path\nvar require_path = __commonJS({\n  \"(disabled):path\"() {\n  }\n});\n\n// (disabled):worker_threads\nvar require_worker_threads = __commonJS({\n  \"(disabled):worker_threads\"() {\n  }\n});\n\n// (disabled):perf_hooks\nvar require_perf_hooks = __commonJS({\n  \"(disabled):perf_hooks\"() {\n  }\n});\n\n// (disabled):os\nvar require_os = __commonJS({\n  \"(disabled):os\"() {\n  }\n});\n\n// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js\nvar require_tfjs_backend_wasm_threaded_simd = __commonJS({\n  \"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js\"(exports, module) {\n    var WasmBackendModuleThreadedSimd2 = (() => {\n      var _scriptDir = typeof document !== \"undefined\" && document.currentScript ? document.currentScript.src : void 0;\n      if (typeof __filename !== \"undefined\")\n        _scriptDir = _scriptDir || __filename;\n      return function(WasmBackendModuleThreadedSimd3) {\n        WasmBackendModuleThreadedSimd3 = WasmBackendModuleThreadedSimd3 || {};\n        function GROWABLE_HEAP_I8() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAP8;\n        }\n        function GROWABLE_HEAP_U8() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAPU8;\n        }\n        function GROWABLE_HEAP_I16() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAP16;\n        }\n        function GROWABLE_HEAP_I32() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAP32;\n        }\n        function GROWABLE_HEAP_U32() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAPU32;\n        }\n        function GROWABLE_HEAP_F32() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAPF32;\n        }\n        function GROWABLE_HEAP_F64() {\n          if (wasmMemory.buffer != buffer2) {\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n          }\n          return HEAPF64;\n        }\n        var Module = typeof WasmBackendModuleThreadedSimd3 != \"undefined\" ? WasmBackendModuleThreadedSimd3 : {};\n        var readyPromiseResolve, readyPromiseReject;\n        Module[\"ready\"] = new Promise(function(resolve, reject) {\n          readyPromiseResolve = resolve;\n          readyPromiseReject = reject;\n        });\n        var beforeListeners;\n        if (typeof process !== \"undefined\" && process.listeners) {\n          beforeListeners = { uncaughtException: process.listeners(\"uncaughtException\"), unhandledRejection: process.listeners(\"unhandledRejection\") };\n        }\n        var moduleOverrides = Object.assign({}, Module);\n        var arguments_ = [];\n        var thisProgram = \"./this.program\";\n        var quit_ = (status, toThrow) => {\n          throw toThrow;\n        };\n        var ENVIRONMENT_IS_WEB = typeof window == \"object\";\n        var ENVIRONMENT_IS_WORKER = typeof importScripts == \"function\";\n        var ENVIRONMENT_IS_NODE = typeof process == \"object\" && typeof process.versions == \"object\" && typeof process.versions.node == \"string\";\n        var ENVIRONMENT_IS_PTHREAD = Module[\"ENVIRONMENT_IS_PTHREAD\"] || false;\n        var scriptDirectory = \"\";\n        function locateFile(path) {\n          if (Module[\"locateFile\"]) {\n            return Module[\"locateFile\"](path, scriptDirectory);\n          }\n          return scriptDirectory + path;\n        }\n        var read_, readAsync, readBinary, setWindowTitle;\n        function logExceptionOnExit(e) {\n          if (e instanceof ExitStatus)\n            return;\n          let toLog = e;\n          err(\"exiting due to exception: \" + toLog);\n        }\n        if (ENVIRONMENT_IS_NODE) {\n          if (ENVIRONMENT_IS_WORKER) {\n            scriptDirectory = require_path().dirname(scriptDirectory) + \"/\";\n          } else {\n            scriptDirectory = __dirname + \"/\";\n          }\n          var fs, nodePath;\n          if (typeof __require === \"function\") {\n            fs = require_fs();\n            nodePath = require_path();\n          }\n          read_ = (filename, binary) => {\n            filename = nodePath[\"normalize\"](filename);\n            return fs.readFileSync(filename, binary ? void 0 : \"utf8\");\n          };\n          readBinary = (filename) => {\n            var ret = read_(filename, true);\n            if (!ret.buffer) {\n              ret = new Uint8Array(ret);\n            }\n            return ret;\n          };\n          readAsync = (filename, onload, onerror) => {\n            filename = nodePath[\"normalize\"](filename);\n            fs.readFile(filename, function(err2, data) {\n              if (err2)\n                onerror(err2);\n              else\n                onload(data.buffer);\n            });\n          };\n          if (process[\"argv\"].length > 1) {\n            thisProgram = process[\"argv\"][1].replace(/\\\\/g, \"/\");\n          }\n          arguments_ = process[\"argv\"].slice(2);\n          process[\"on\"](\"uncaughtException\", function(ex) {\n            if (!(ex instanceof ExitStatus)) {\n              throw ex;\n            }\n          });\n          process[\"on\"](\"unhandledRejection\", function(reason) {\n            throw reason;\n          });\n          quit_ = (status, toThrow) => {\n            if (keepRuntimeAlive()) {\n              process[\"exitCode\"] = status;\n              throw toThrow;\n            }\n            logExceptionOnExit(toThrow);\n            process[\"exit\"](status);\n          };\n          Module[\"inspect\"] = function() {\n            return \"[Emscripten Module object]\";\n          };\n          let nodeWorkerThreads;\n          try {\n            nodeWorkerThreads = require_worker_threads();\n          } catch (e) {\n            console.error('The \"worker_threads\" module is not supported in this node.js build - perhaps a newer version is needed?');\n            throw e;\n          }\n          global.Worker = nodeWorkerThreads.Worker;\n        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {\n          if (ENVIRONMENT_IS_WORKER) {\n            scriptDirectory = self.location.href;\n          } else if (typeof document != \"undefined\" && document.currentScript) {\n            scriptDirectory = document.currentScript.src;\n          }\n          if (typeof _scriptDir !== \"undefined\" && _scriptDir) {\n            scriptDirectory = _scriptDir;\n          }\n          if (scriptDirectory.indexOf(\"blob:\") !== 0) {\n            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, \"\").lastIndexOf(\"/\") + 1);\n          } else {\n            scriptDirectory = \"\";\n          }\n          if (!ENVIRONMENT_IS_NODE) {\n            read_ = (url) => {\n              var xhr = new XMLHttpRequest();\n              xhr.open(\"GET\", url, false);\n              xhr.send(null);\n              return xhr.responseText;\n            };\n            if (ENVIRONMENT_IS_WORKER) {\n              readBinary = (url) => {\n                var xhr = new XMLHttpRequest();\n                xhr.open(\"GET\", url, false);\n                xhr.responseType = \"arraybuffer\";\n                xhr.send(null);\n                return new Uint8Array(xhr.response);\n              };\n            }\n            readAsync = (url, onload, onerror) => {\n              var xhr = new XMLHttpRequest();\n              xhr.open(\"GET\", url, true);\n              xhr.responseType = \"arraybuffer\";\n              xhr.onload = () => {\n                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {\n                  onload(xhr.response);\n                  return;\n                }\n                onerror();\n              };\n              xhr.onerror = onerror;\n              xhr.send(null);\n            };\n          }\n          setWindowTitle = (title) => document.title = title;\n        } else {\n        }\n        if (ENVIRONMENT_IS_NODE) {\n          if (typeof performance == \"undefined\") {\n            global.performance = require_perf_hooks().performance;\n          }\n        }\n        var defaultPrint = console.log.bind(console);\n        var defaultPrintErr = console.warn.bind(console);\n        if (ENVIRONMENT_IS_NODE) {\n          defaultPrint = (str) => fs.writeSync(1, str + \"\\n\");\n          defaultPrintErr = (str) => fs.writeSync(2, str + \"\\n\");\n        }\n        var out = Module[\"print\"] || defaultPrint;\n        var err = Module[\"printErr\"] || defaultPrintErr;\n        Object.assign(Module, moduleOverrides);\n        moduleOverrides = null;\n        if (Module[\"arguments\"])\n          arguments_ = Module[\"arguments\"];\n        if (Module[\"thisProgram\"])\n          thisProgram = Module[\"thisProgram\"];\n        if (Module[\"quit\"])\n          quit_ = Module[\"quit\"];\n        var POINTER_SIZE = 4;\n        var Atomics_load = Atomics.load;\n        var Atomics_store = Atomics.store;\n        var Atomics_compareExchange = Atomics.compareExchange;\n        var wasmBinary;\n        if (Module[\"wasmBinary\"])\n          wasmBinary = Module[\"wasmBinary\"];\n        var noExitRuntime = Module[\"noExitRuntime\"] || true;\n        if (typeof WebAssembly != \"object\") {\n          abort(\"no native wasm support detected\");\n        }\n        var wasmMemory;\n        var wasmModule;\n        var ABORT = false;\n        var EXITSTATUS;\n        function assert3(condition, text) {\n          if (!condition) {\n            abort(text);\n          }\n        }\n        var UTF8Decoder = typeof TextDecoder != \"undefined\" ? new TextDecoder(\"utf8\") : void 0;\n        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {\n          var endIdx = idx + maxBytesToRead;\n          var endPtr = idx;\n          while (heapOrArray[endPtr] && !(endPtr >= endIdx))\n            ++endPtr;\n          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {\n            return UTF8Decoder.decode(heapOrArray.buffer instanceof SharedArrayBuffer ? heapOrArray.slice(idx, endPtr) : heapOrArray.subarray(idx, endPtr));\n          }\n          var str = \"\";\n          while (idx < endPtr) {\n            var u0 = heapOrArray[idx++];\n            if (!(u0 & 128)) {\n              str += String.fromCharCode(u0);\n              continue;\n            }\n            var u1 = heapOrArray[idx++] & 63;\n            if ((u0 & 224) == 192) {\n              str += String.fromCharCode((u0 & 31) << 6 | u1);\n              continue;\n            }\n            var u2 = heapOrArray[idx++] & 63;\n            if ((u0 & 240) == 224) {\n              u0 = (u0 & 15) << 12 | u1 << 6 | u2;\n            } else {\n              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;\n            }\n            if (u0 < 65536) {\n              str += String.fromCharCode(u0);\n            } else {\n              var ch = u0 - 65536;\n              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);\n            }\n          }\n          return str;\n        }\n        function UTF8ToString(ptr, maxBytesToRead) {\n          return ptr ? UTF8ArrayToString(GROWABLE_HEAP_U8(), ptr, maxBytesToRead) : \"\";\n        }\n        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {\n          if (!(maxBytesToWrite > 0))\n            return 0;\n          var startIdx = outIdx;\n          var endIdx = outIdx + maxBytesToWrite - 1;\n          for (var i = 0; i < str.length; ++i) {\n            var u = str.charCodeAt(i);\n            if (u >= 55296 && u <= 57343) {\n              var u1 = str.charCodeAt(++i);\n              u = 65536 + ((u & 1023) << 10) | u1 & 1023;\n            }\n            if (u <= 127) {\n              if (outIdx >= endIdx)\n                break;\n              heap[outIdx++] = u;\n            } else if (u <= 2047) {\n              if (outIdx + 1 >= endIdx)\n                break;\n              heap[outIdx++] = 192 | u >> 6;\n              heap[outIdx++] = 128 | u & 63;\n            } else if (u <= 65535) {\n              if (outIdx + 2 >= endIdx)\n                break;\n              heap[outIdx++] = 224 | u >> 12;\n              heap[outIdx++] = 128 | u >> 6 & 63;\n              heap[outIdx++] = 128 | u & 63;\n            } else {\n              if (outIdx + 3 >= endIdx)\n                break;\n              heap[outIdx++] = 240 | u >> 18;\n              heap[outIdx++] = 128 | u >> 12 & 63;\n              heap[outIdx++] = 128 | u >> 6 & 63;\n              heap[outIdx++] = 128 | u & 63;\n            }\n          }\n          heap[outIdx] = 0;\n          return outIdx - startIdx;\n        }\n        function stringToUTF8(str, outPtr, maxBytesToWrite) {\n          return stringToUTF8Array(str, GROWABLE_HEAP_U8(), outPtr, maxBytesToWrite);\n        }\n        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;\n        if (ENVIRONMENT_IS_PTHREAD) {\n          buffer2 = Module[\"buffer\"];\n        }\n        function updateGlobalBufferAndViews(buf) {\n          buffer2 = buf;\n          Module[\"HEAP8\"] = HEAP8 = new Int8Array(buf);\n          Module[\"HEAP16\"] = HEAP16 = new Int16Array(buf);\n          Module[\"HEAP32\"] = HEAP32 = new Int32Array(buf);\n          Module[\"HEAPU8\"] = HEAPU8 = new Uint8Array(buf);\n          Module[\"HEAPU16\"] = HEAPU16 = new Uint16Array(buf);\n          Module[\"HEAPU32\"] = HEAPU32 = new Uint32Array(buf);\n          Module[\"HEAPF32\"] = HEAPF32 = new Float32Array(buf);\n          Module[\"HEAPF64\"] = HEAPF64 = new Float64Array(buf);\n        }\n        var INITIAL_MEMORY = Module[\"INITIAL_MEMORY\"] || 16777216;\n        if (ENVIRONMENT_IS_PTHREAD) {\n          wasmMemory = Module[\"wasmMemory\"];\n          buffer2 = Module[\"buffer\"];\n        } else {\n          if (Module[\"wasmMemory\"]) {\n            wasmMemory = Module[\"wasmMemory\"];\n          } else {\n            wasmMemory = new WebAssembly.Memory({ \"initial\": INITIAL_MEMORY / 65536, \"maximum\": 2147483648 / 65536, \"shared\": true });\n            if (!(wasmMemory.buffer instanceof SharedArrayBuffer)) {\n              err(\"requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag\");\n              if (ENVIRONMENT_IS_NODE) {\n                console.log(\"(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)\");\n              }\n              throw Error(\"bad memory\");\n            }\n          }\n        }\n        if (wasmMemory) {\n          buffer2 = wasmMemory.buffer;\n        }\n        INITIAL_MEMORY = buffer2.byteLength;\n        updateGlobalBufferAndViews(buffer2);\n        var wasmTable;\n        var __ATPRERUN__ = [];\n        var __ATINIT__ = [];\n        var __ATPOSTRUN__ = [];\n        var runtimeInitialized = false;\n        function keepRuntimeAlive() {\n          return noExitRuntime;\n        }\n        function preRun() {\n          if (Module[\"preRun\"]) {\n            if (typeof Module[\"preRun\"] == \"function\")\n              Module[\"preRun\"] = [Module[\"preRun\"]];\n            while (Module[\"preRun\"].length) {\n              addOnPreRun(Module[\"preRun\"].shift());\n            }\n          }\n          callRuntimeCallbacks(__ATPRERUN__);\n        }\n        function initRuntime() {\n          runtimeInitialized = true;\n          if (ENVIRONMENT_IS_PTHREAD)\n            return;\n          callRuntimeCallbacks(__ATINIT__);\n        }\n        function postRun() {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return;\n          if (Module[\"postRun\"]) {\n            if (typeof Module[\"postRun\"] == \"function\")\n              Module[\"postRun\"] = [Module[\"postRun\"]];\n            while (Module[\"postRun\"].length) {\n              addOnPostRun(Module[\"postRun\"].shift());\n            }\n          }\n          callRuntimeCallbacks(__ATPOSTRUN__);\n        }\n        function addOnPreRun(cb) {\n          __ATPRERUN__.unshift(cb);\n        }\n        function addOnInit(cb) {\n          __ATINIT__.unshift(cb);\n        }\n        function addOnPostRun(cb) {\n          __ATPOSTRUN__.unshift(cb);\n        }\n        var runDependencies = 0;\n        var runDependencyWatcher = null;\n        var dependenciesFulfilled = null;\n        function addRunDependency(id) {\n          runDependencies++;\n          if (Module[\"monitorRunDependencies\"]) {\n            Module[\"monitorRunDependencies\"](runDependencies);\n          }\n        }\n        function removeRunDependency(id) {\n          runDependencies--;\n          if (Module[\"monitorRunDependencies\"]) {\n            Module[\"monitorRunDependencies\"](runDependencies);\n          }\n          if (runDependencies == 0) {\n            if (runDependencyWatcher !== null) {\n              clearInterval(runDependencyWatcher);\n              runDependencyWatcher = null;\n            }\n            if (dependenciesFulfilled) {\n              var callback = dependenciesFulfilled;\n              dependenciesFulfilled = null;\n              callback();\n            }\n          }\n        }\n        function abort(what) {\n          if (ENVIRONMENT_IS_PTHREAD) {\n            postMessage({ \"cmd\": \"onAbort\", \"arg\": what });\n          } else {\n            if (Module[\"onAbort\"]) {\n              Module[\"onAbort\"](what);\n            }\n          }\n          what = \"Aborted(\" + what + \")\";\n          err(what);\n          ABORT = true;\n          EXITSTATUS = 1;\n          what += \". Build with -sASSERTIONS for more info.\";\n          var e = new WebAssembly.RuntimeError(what);\n          readyPromiseReject(e);\n          throw e;\n        }\n        var dataURIPrefix = \"data:application/octet-stream;base64,\";\n        function isDataURI(filename) {\n          return filename.startsWith(dataURIPrefix);\n        }\n        function isFileURI(filename) {\n          return filename.startsWith(\"file://\");\n        }\n        var wasmBinaryFile;\n        wasmBinaryFile = \"tfjs-backend-wasm-threaded-simd.wasm\";\n        if (!isDataURI(wasmBinaryFile)) {\n          wasmBinaryFile = locateFile(wasmBinaryFile);\n        }\n        function getBinary(file) {\n          try {\n            if (file == wasmBinaryFile && wasmBinary) {\n              return new Uint8Array(wasmBinary);\n            }\n            if (readBinary) {\n              return readBinary(file);\n            }\n            throw \"both async and sync fetching of the wasm failed\";\n          } catch (err2) {\n            abort(err2);\n          }\n        }\n        function getBinaryPromise() {\n          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {\n            if (typeof fetch == \"function\" && !isFileURI(wasmBinaryFile)) {\n              return fetch(wasmBinaryFile, { credentials: \"same-origin\" }).then(function(response) {\n                if (!response[\"ok\"]) {\n                  throw \"failed to load wasm binary file at '\" + wasmBinaryFile + \"'\";\n                }\n                return response[\"arrayBuffer\"]();\n              }).catch(function() {\n                return getBinary(wasmBinaryFile);\n              });\n            } else {\n              if (readAsync) {\n                return new Promise(function(resolve, reject) {\n                  readAsync(wasmBinaryFile, function(response) {\n                    resolve(new Uint8Array(response));\n                  }, reject);\n                });\n              }\n            }\n          }\n          return Promise.resolve().then(function() {\n            return getBinary(wasmBinaryFile);\n          });\n        }\n        function createWasm() {\n          var info = { \"env\": asmLibraryArg, \"wasi_snapshot_preview1\": asmLibraryArg };\n          function receiveInstance(instance, module2) {\n            var exports3 = instance.exports;\n            Module[\"asm\"] = exports3;\n            registerTLSInit(Module[\"asm\"][\"_emscripten_tls_init\"]);\n            wasmTable = Module[\"asm\"][\"__indirect_function_table\"];\n            addOnInit(Module[\"asm\"][\"__wasm_call_ctors\"]);\n            wasmModule = module2;\n            if (!ENVIRONMENT_IS_PTHREAD) {\n              var numWorkersToLoad = PThread.unusedWorkers.length;\n              PThread.unusedWorkers.forEach(function(w) {\n                PThread.loadWasmModuleToWorker(w, function() {\n                  if (!--numWorkersToLoad)\n                    removeRunDependency(\"wasm-instantiate\");\n                });\n              });\n            }\n          }\n          if (!ENVIRONMENT_IS_PTHREAD) {\n            addRunDependency(\"wasm-instantiate\");\n          }\n          function receiveInstantiationResult(result) {\n            receiveInstance(result[\"instance\"], result[\"module\"]);\n          }\n          function instantiateArrayBuffer(receiver) {\n            return getBinaryPromise().then(function(binary) {\n              return WebAssembly.instantiate(binary, info);\n            }).then(function(instance) {\n              return instance;\n            }).then(receiver, function(reason) {\n              err(\"failed to asynchronously prepare wasm: \" + reason);\n              abort(reason);\n            });\n          }\n          function instantiateAsync() {\n            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == \"function\" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == \"function\") {\n              return fetch(wasmBinaryFile, { credentials: \"same-origin\" }).then(function(response) {\n                var result = WebAssembly.instantiateStreaming(response, info);\n                return result.then(receiveInstantiationResult, function(reason) {\n                  err(\"wasm streaming compile failed: \" + reason);\n                  err(\"falling back to ArrayBuffer instantiation\");\n                  return instantiateArrayBuffer(receiveInstantiationResult);\n                });\n              });\n            } else {\n              return instantiateArrayBuffer(receiveInstantiationResult);\n            }\n          }\n          if (Module[\"instantiateWasm\"]) {\n            try {\n              var exports2 = Module[\"instantiateWasm\"](info, receiveInstance);\n              return exports2;\n            } catch (e) {\n              err(\"Module.instantiateWasm callback failed with error: \" + e);\n              readyPromiseReject(e);\n            }\n          }\n          instantiateAsync().catch(readyPromiseReject);\n          return {};\n        }\n        var tempDouble;\n        var tempI64;\n        var ASM_CONSTS = {};\n        function ExitStatus(status) {\n          this.name = \"ExitStatus\";\n          this.message = \"Program terminated with exit(\" + status + \")\";\n          this.status = status;\n        }\n        function killThread(pthread_ptr) {\n          var worker = PThread.pthreads[pthread_ptr];\n          delete PThread.pthreads[pthread_ptr];\n          worker.terminate();\n          __emscripten_thread_free_data(pthread_ptr);\n          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);\n          worker.pthread_ptr = 0;\n        }\n        function cancelThread(pthread_ptr) {\n          var worker = PThread.pthreads[pthread_ptr];\n          worker.postMessage({ \"cmd\": \"cancel\" });\n        }\n        function cleanupThread(pthread_ptr) {\n          var worker = PThread.pthreads[pthread_ptr];\n          assert3(worker);\n          PThread.returnWorkerToPool(worker);\n        }\n        function spawnThread(threadParams) {\n          var worker = PThread.getNewWorker();\n          if (!worker) {\n            return 6;\n          }\n          PThread.runningWorkers.push(worker);\n          PThread.pthreads[threadParams.pthread_ptr] = worker;\n          worker.pthread_ptr = threadParams.pthread_ptr;\n          var msg = { \"cmd\": \"run\", \"start_routine\": threadParams.startRoutine, \"arg\": threadParams.arg, \"pthread_ptr\": threadParams.pthread_ptr };\n          worker.runPthread = () => {\n            msg.time = performance.now();\n            worker.postMessage(msg, threadParams.transferList);\n          };\n          if (worker.loaded) {\n            worker.runPthread();\n            delete worker.runPthread;\n          }\n          return 0;\n        }\n        var SYSCALLS = { varargs: void 0, get: function() {\n          SYSCALLS.varargs += 4;\n          var ret = GROWABLE_HEAP_I32()[SYSCALLS.varargs - 4 >> 2];\n          return ret;\n        }, getStr: function(ptr) {\n          var ret = UTF8ToString(ptr);\n          return ret;\n        } };\n        function _proc_exit(code) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(1, 1, code);\n          EXITSTATUS = code;\n          if (!keepRuntimeAlive()) {\n            PThread.terminateAllThreads();\n            if (Module[\"onExit\"])\n              Module[\"onExit\"](code);\n            ABORT = true;\n          }\n          quit_(code, new ExitStatus(code));\n        }\n        function exitJS(status, implicit) {\n          EXITSTATUS = status;\n          if (!implicit) {\n            if (ENVIRONMENT_IS_PTHREAD) {\n              exitOnMainThread(status);\n              throw \"unwind\";\n            } else {\n            }\n          }\n          _proc_exit(status);\n        }\n        var _exit = exitJS;\n        function handleException(e) {\n          if (e instanceof ExitStatus || e == \"unwind\") {\n            return EXITSTATUS;\n          }\n          quit_(1, e);\n        }\n        var PThread = { unusedWorkers: [], runningWorkers: [], tlsInitFunctions: [], pthreads: {}, init: function() {\n          if (ENVIRONMENT_IS_PTHREAD) {\n            PThread.initWorker();\n          } else {\n            PThread.initMainThread();\n          }\n        }, initMainThread: function() {\n          var pthreadPoolSize = 8;\n          while (pthreadPoolSize--) {\n            PThread.allocateUnusedWorker();\n          }\n        }, initWorker: function() {\n          noExitRuntime = false;\n        }, setExitStatus: function(status) {\n          EXITSTATUS = status;\n        }, terminateAllThreads: function() {\n          for (var worker of Object.values(PThread.pthreads)) {\n            PThread.returnWorkerToPool(worker);\n          }\n          for (var worker of PThread.unusedWorkers) {\n            worker.terminate();\n          }\n          PThread.unusedWorkers = [];\n        }, returnWorkerToPool: function(worker) {\n          var pthread_ptr = worker.pthread_ptr;\n          delete PThread.pthreads[pthread_ptr];\n          PThread.unusedWorkers.push(worker);\n          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);\n          worker.pthread_ptr = 0;\n          __emscripten_thread_free_data(pthread_ptr);\n        }, receiveObjectTransfer: function(data) {\n        }, threadInitTLS: function() {\n          PThread.tlsInitFunctions.forEach((f) => f());\n        }, loadWasmModuleToWorker: function(worker, onFinishedLoading) {\n          worker.onmessage = (e) => {\n            var d = e[\"data\"];\n            var cmd = d[\"cmd\"];\n            if (worker.pthread_ptr)\n              PThread.currentProxiedOperationCallerThread = worker.pthread_ptr;\n            if (d[\"targetThread\"] && d[\"targetThread\"] != _pthread_self()) {\n              var targetWorker = PThread.pthreads[d.targetThread];\n              if (targetWorker) {\n                targetWorker.postMessage(d, d[\"transferList\"]);\n              } else {\n                err('Internal error! Worker sent a message \"' + cmd + '\" to target pthread ' + d[\"targetThread\"] + \", but that thread no longer exists!\");\n              }\n              PThread.currentProxiedOperationCallerThread = void 0;\n              return;\n            }\n            if (cmd === \"processProxyingQueue\") {\n              executeNotifiedProxyingQueue(d[\"queue\"]);\n            } else if (cmd === \"spawnThread\") {\n              spawnThread(d);\n            } else if (cmd === \"cleanupThread\") {\n              cleanupThread(d[\"thread\"]);\n            } else if (cmd === \"killThread\") {\n              killThread(d[\"thread\"]);\n            } else if (cmd === \"cancelThread\") {\n              cancelThread(d[\"thread\"]);\n            } else if (cmd === \"loaded\") {\n              worker.loaded = true;\n              if (onFinishedLoading)\n                onFinishedLoading(worker);\n              if (worker.runPthread) {\n                worker.runPthread();\n                delete worker.runPthread;\n              }\n            } else if (cmd === \"print\") {\n              out(\"Thread \" + d[\"threadId\"] + \": \" + d[\"text\"]);\n            } else if (cmd === \"printErr\") {\n              err(\"Thread \" + d[\"threadId\"] + \": \" + d[\"text\"]);\n            } else if (cmd === \"alert\") {\n              alert(\"Thread \" + d[\"threadId\"] + \": \" + d[\"text\"]);\n            } else if (d.target === \"setimmediate\") {\n              worker.postMessage(d);\n            } else if (cmd === \"onAbort\") {\n              if (Module[\"onAbort\"]) {\n                Module[\"onAbort\"](d[\"arg\"]);\n              }\n            } else if (cmd) {\n              err(\"worker sent an unknown command \" + cmd);\n            }\n            PThread.currentProxiedOperationCallerThread = void 0;\n          };\n          worker.onerror = (e) => {\n            var message = \"worker sent an error!\";\n            err(message + \" \" + e.filename + \":\" + e.lineno + \": \" + e.message);\n            throw e;\n          };\n          if (ENVIRONMENT_IS_NODE) {\n            worker.on(\"message\", function(data) {\n              worker.onmessage({ data });\n            });\n            worker.on(\"error\", function(e) {\n              worker.onerror(e);\n            });\n            worker.on(\"detachedExit\", function() {\n            });\n          }\n          worker.postMessage({ \"cmd\": \"load\", \"urlOrBlob\": Module[\"mainScriptUrlOrBlob\"] || _scriptDir, \"wasmMemory\": wasmMemory, \"wasmModule\": wasmModule });\n        }, allocateUnusedWorker: function() {\n          var pthreadMainJs = locateFile(\"tfjs-backend-wasm-threaded-simd.worker.js\");\n          PThread.unusedWorkers.push(new Worker(pthreadMainJs));\n        }, getNewWorker: function() {\n          if (PThread.unusedWorkers.length == 0) {\n            PThread.allocateUnusedWorker();\n            PThread.loadWasmModuleToWorker(PThread.unusedWorkers[0]);\n          }\n          return PThread.unusedWorkers.pop();\n        } };\n        Module[\"PThread\"] = PThread;\n        function callRuntimeCallbacks(callbacks2) {\n          while (callbacks2.length > 0) {\n            callbacks2.shift()(Module);\n          }\n        }\n        function withStackSave(f) {\n          var stack2 = stackSave();\n          var ret = f();\n          stackRestore(stack2);\n          return ret;\n        }\n        function demangle(func2) {\n          return func2;\n        }\n        function demangleAll(text) {\n          var regex = /\\b_Z[\\w\\d_]+/g;\n          return text.replace(regex, function(x) {\n            var y = demangle(x);\n            return x === y ? x : y + \" [\" + x + \"]\";\n          });\n        }\n        function establishStackSpace() {\n          var pthread_ptr = _pthread_self();\n          var stackTop = GROWABLE_HEAP_I32()[pthread_ptr + 44 >> 2];\n          var stackSize = GROWABLE_HEAP_I32()[pthread_ptr + 48 >> 2];\n          var stackMax = stackTop - stackSize;\n          _emscripten_stack_set_limits(stackTop, stackMax);\n          stackRestore(stackTop);\n        }\n        Module[\"establishStackSpace\"] = establishStackSpace;\n        function exitOnMainThread(returnCode) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(2, 0, returnCode);\n          try {\n            _exit(returnCode);\n          } catch (e) {\n            handleException(e);\n          }\n        }\n        var wasmTableMirror = [];\n        function getWasmTableEntry(funcPtr) {\n          var func2 = wasmTableMirror[funcPtr];\n          if (!func2) {\n            if (funcPtr >= wasmTableMirror.length)\n              wasmTableMirror.length = funcPtr + 1;\n            wasmTableMirror[funcPtr] = func2 = wasmTable.get(funcPtr);\n          }\n          return func2;\n        }\n        function invokeEntryPoint(ptr, arg) {\n          var result = getWasmTableEntry(ptr)(arg);\n          if (keepRuntimeAlive()) {\n            PThread.setExitStatus(result);\n          } else {\n            __emscripten_thread_exit(result);\n          }\n        }\n        Module[\"invokeEntryPoint\"] = invokeEntryPoint;\n        function jsStackTrace() {\n          var error = new Error();\n          if (!error.stack) {\n            try {\n              throw new Error();\n            } catch (e) {\n              error = e;\n            }\n            if (!error.stack) {\n              return \"(no stack trace available)\";\n            }\n          }\n          return error.stack.toString();\n        }\n        function registerTLSInit(tlsInitFunc) {\n          PThread.tlsInitFunctions.push(tlsInitFunc);\n        }\n        function writeArrayToMemory(array2, buffer3) {\n          GROWABLE_HEAP_I8().set(array2, buffer3);\n        }\n        function ___emscripten_init_main_thread_js(tb) {\n          __emscripten_thread_init(tb, !ENVIRONMENT_IS_WORKER, 1, !ENVIRONMENT_IS_WEB);\n          PThread.threadInitTLS();\n        }\n        function ___emscripten_thread_cleanup(thread) {\n          if (!ENVIRONMENT_IS_PTHREAD)\n            cleanupThread(thread);\n          else\n            postMessage({ \"cmd\": \"cleanupThread\", \"thread\": thread });\n        }\n        function pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(3, 1, pthread_ptr, attr, startRoutine, arg);\n          return ___pthread_create_js(pthread_ptr, attr, startRoutine, arg);\n        }\n        function ___pthread_create_js(pthread_ptr, attr, startRoutine, arg) {\n          if (typeof SharedArrayBuffer == \"undefined\") {\n            err(\"Current environment does not support SharedArrayBuffer, pthreads are not available!\");\n            return 6;\n          }\n          var transferList = [];\n          var error = 0;\n          if (ENVIRONMENT_IS_PTHREAD && (transferList.length === 0 || error)) {\n            return pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg);\n          }\n          if (error)\n            return error;\n          var threadParams = { startRoutine, pthread_ptr, arg, transferList };\n          if (ENVIRONMENT_IS_PTHREAD) {\n            threadParams.cmd = \"spawnThread\";\n            postMessage(threadParams, transferList);\n            return 0;\n          }\n          return spawnThread(threadParams);\n        }\n        function __emscripten_default_pthread_stack_size() {\n          return 2097152;\n        }\n        var nowIsMonotonic = true;\n        function __emscripten_get_now_is_monotonic() {\n          return nowIsMonotonic;\n        }\n        function executeNotifiedProxyingQueue(queue) {\n          Atomics.store(GROWABLE_HEAP_I32(), queue >> 2, 1);\n          if (_pthread_self()) {\n            __emscripten_proxy_execute_task_queue(queue);\n          }\n          Atomics.compareExchange(GROWABLE_HEAP_I32(), queue >> 2, 1, 0);\n        }\n        Module[\"executeNotifiedProxyingQueue\"] = executeNotifiedProxyingQueue;\n        function __emscripten_notify_task_queue(targetThreadId, currThreadId, mainThreadId, queue) {\n          if (targetThreadId == currThreadId) {\n            setTimeout(() => executeNotifiedProxyingQueue(queue));\n          } else if (ENVIRONMENT_IS_PTHREAD) {\n            postMessage({ \"targetThread\": targetThreadId, \"cmd\": \"processProxyingQueue\", \"queue\": queue });\n          } else {\n            var worker = PThread.pthreads[targetThreadId];\n            if (!worker) {\n              return;\n            }\n            worker.postMessage({ \"cmd\": \"processProxyingQueue\", \"queue\": queue });\n          }\n          return 1;\n        }\n        function __emscripten_set_offscreencanvas_size(target, width, height) {\n          return -1;\n        }\n        function _abort() {\n          abort(\"\");\n        }\n        function warnOnce(text) {\n          if (!warnOnce.shown)\n            warnOnce.shown = {};\n          if (!warnOnce.shown[text]) {\n            warnOnce.shown[text] = 1;\n            if (ENVIRONMENT_IS_NODE)\n              text = \"warning: \" + text;\n            err(text);\n          }\n        }\n        function _emscripten_check_blocking_allowed() {\n          if (ENVIRONMENT_IS_NODE)\n            return;\n          if (ENVIRONMENT_IS_WORKER)\n            return;\n          warnOnce(\"Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread\");\n        }\n        function _emscripten_date_now() {\n          return Date.now();\n        }\n        function getHeapMax() {\n          return 2147483648;\n        }\n        function _emscripten_get_heap_max() {\n          return getHeapMax();\n        }\n        var _emscripten_get_now;\n        if (ENVIRONMENT_IS_NODE) {\n          _emscripten_get_now = () => {\n            var t = process[\"hrtime\"]();\n            return t[0] * 1e3 + t[1] / 1e6;\n          };\n        } else if (ENVIRONMENT_IS_PTHREAD) {\n          _emscripten_get_now = () => performance.now() - Module[\"__performance_now_clock_drift\"];\n        } else\n          _emscripten_get_now = () => performance.now();\n        function _emscripten_memcpy_big(dest, src, num) {\n          GROWABLE_HEAP_U8().copyWithin(dest, src, src + num);\n        }\n        function _emscripten_num_logical_cores() {\n          if (ENVIRONMENT_IS_NODE)\n            return require_os().cpus().length;\n          return navigator[\"hardwareConcurrency\"];\n        }\n        function _emscripten_proxy_to_main_thread_js(index, sync) {\n          var numCallArgs = arguments.length - 2;\n          var outerArgs = arguments;\n          return withStackSave(() => {\n            var serializedNumCallArgs = numCallArgs;\n            var args = stackAlloc(serializedNumCallArgs * 8);\n            var b = args >> 3;\n            for (var i = 0; i < numCallArgs; i++) {\n              var arg = outerArgs[2 + i];\n              GROWABLE_HEAP_F64()[b + i] = arg;\n            }\n            return _emscripten_run_in_main_runtime_thread_js(index, serializedNumCallArgs, args, sync);\n          });\n        }\n        var _emscripten_receive_on_main_thread_js_callArgs = [];\n        function _emscripten_receive_on_main_thread_js(index, numCallArgs, args) {\n          _emscripten_receive_on_main_thread_js_callArgs.length = numCallArgs;\n          var b = args >> 3;\n          for (var i = 0; i < numCallArgs; i++) {\n            _emscripten_receive_on_main_thread_js_callArgs[i] = GROWABLE_HEAP_F64()[b + i];\n          }\n          var isEmAsmConst = index < 0;\n          var func2 = !isEmAsmConst ? proxiedFunctionTable[index] : ASM_CONSTS[-index - 1];\n          return func2.apply(null, _emscripten_receive_on_main_thread_js_callArgs);\n        }\n        function emscripten_realloc_buffer(size) {\n          try {\n            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n            return 1;\n          } catch (e) {\n          }\n        }\n        function _emscripten_resize_heap(requestedSize) {\n          var oldSize = GROWABLE_HEAP_U8().length;\n          requestedSize = requestedSize >>> 0;\n          if (requestedSize <= oldSize) {\n            return false;\n          }\n          var maxHeapSize = getHeapMax();\n          if (requestedSize > maxHeapSize) {\n            return false;\n          }\n          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;\n          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {\n            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);\n            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);\n            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));\n            var replacement = emscripten_realloc_buffer(newSize);\n            if (replacement) {\n              return true;\n            }\n          }\n          return false;\n        }\n        function _emscripten_unwind_to_js_event_loop() {\n          throw \"unwind\";\n        }\n        function _fd_close(fd) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(4, 1, fd);\n          return 52;\n        }\n        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(5, 1, fd, offset_low, offset_high, whence, newOffset);\n          return 70;\n        }\n        var printCharBuffers = [null, [], []];\n        function printChar(stream, curr) {\n          var buffer3 = printCharBuffers[stream];\n          if (curr === 0 || curr === 10) {\n            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));\n            buffer3.length = 0;\n          } else {\n            buffer3.push(curr);\n          }\n        }\n        function _fd_write(fd, iov, iovcnt, pnum) {\n          if (ENVIRONMENT_IS_PTHREAD)\n            return _emscripten_proxy_to_main_thread_js(6, 1, fd, iov, iovcnt, pnum);\n          var num = 0;\n          for (var i = 0; i < iovcnt; i++) {\n            var ptr = GROWABLE_HEAP_U32()[iov >> 2];\n            var len = GROWABLE_HEAP_U32()[iov + 4 >> 2];\n            iov += 8;\n            for (var j = 0; j < len; j++) {\n              printChar(fd, GROWABLE_HEAP_U8()[ptr + j]);\n            }\n            num += len;\n          }\n          GROWABLE_HEAP_U32()[pnum >> 2] = num;\n          return 0;\n        }\n        function getCFunc(ident) {\n          var func2 = Module[\"_\" + ident];\n          return func2;\n        }\n        function ccall(ident, returnType, argTypes, args, opts) {\n          var toC = { \"string\": (str) => {\n            var ret2 = 0;\n            if (str !== null && str !== void 0 && str !== 0) {\n              var len = (str.length << 2) + 1;\n              ret2 = stackAlloc(len);\n              stringToUTF8(str, ret2, len);\n            }\n            return ret2;\n          }, \"array\": (arr) => {\n            var ret2 = stackAlloc(arr.length);\n            writeArrayToMemory(arr, ret2);\n            return ret2;\n          } };\n          function convertReturnValue(ret2) {\n            if (returnType === \"string\") {\n              return UTF8ToString(ret2);\n            }\n            if (returnType === \"boolean\")\n              return Boolean(ret2);\n            return ret2;\n          }\n          var func2 = getCFunc(ident);\n          var cArgs = [];\n          var stack2 = 0;\n          if (args) {\n            for (var i = 0; i < args.length; i++) {\n              var converter = toC[argTypes[i]];\n              if (converter) {\n                if (stack2 === 0)\n                  stack2 = stackSave();\n                cArgs[i] = converter(args[i]);\n              } else {\n                cArgs[i] = args[i];\n              }\n            }\n          }\n          var ret = func2.apply(null, cArgs);\n          function onDone(ret2) {\n            if (stack2 !== 0)\n              stackRestore(stack2);\n            return convertReturnValue(ret2);\n          }\n          ret = onDone(ret);\n          return ret;\n        }\n        function cwrap(ident, returnType, argTypes, opts) {\n          argTypes = argTypes || [];\n          var numericArgs = argTypes.every((type) => type === \"number\" || type === \"boolean\");\n          var numericRet = returnType !== \"string\";\n          if (numericRet && numericArgs && !opts) {\n            return getCFunc(ident);\n          }\n          return function() {\n            return ccall(ident, returnType, argTypes, arguments, opts);\n          };\n        }\n        PThread.init();\n        var proxiedFunctionTable = [null, _proc_exit, exitOnMainThread, pthreadCreateProxied, _fd_close, _fd_seek, _fd_write];\n        var asmLibraryArg = { \"__emscripten_init_main_thread_js\": ___emscripten_init_main_thread_js, \"__emscripten_thread_cleanup\": ___emscripten_thread_cleanup, \"__pthread_create_js\": ___pthread_create_js, \"_emscripten_default_pthread_stack_size\": __emscripten_default_pthread_stack_size, \"_emscripten_get_now_is_monotonic\": __emscripten_get_now_is_monotonic, \"_emscripten_notify_task_queue\": __emscripten_notify_task_queue, \"_emscripten_set_offscreencanvas_size\": __emscripten_set_offscreencanvas_size, \"abort\": _abort, \"emscripten_check_blocking_allowed\": _emscripten_check_blocking_allowed, \"emscripten_date_now\": _emscripten_date_now, \"emscripten_get_heap_max\": _emscripten_get_heap_max, \"emscripten_get_now\": _emscripten_get_now, \"emscripten_memcpy_big\": _emscripten_memcpy_big, \"emscripten_num_logical_cores\": _emscripten_num_logical_cores, \"emscripten_receive_on_main_thread_js\": _emscripten_receive_on_main_thread_js, \"emscripten_resize_heap\": _emscripten_resize_heap, \"emscripten_unwind_to_js_event_loop\": _emscripten_unwind_to_js_event_loop, \"exit\": _exit, \"fd_close\": _fd_close, \"fd_seek\": _fd_seek, \"fd_write\": _fd_write, \"memory\": wasmMemory || Module[\"wasmMemory\"] };\n        var asm = createWasm();\n        var ___wasm_call_ctors = Module[\"___wasm_call_ctors\"] = function() {\n          return (___wasm_call_ctors = Module[\"___wasm_call_ctors\"] = Module[\"asm\"][\"__wasm_call_ctors\"]).apply(null, arguments);\n        };\n        var _init = Module[\"_init\"] = function() {\n          return (_init = Module[\"_init\"] = Module[\"asm\"][\"init\"]).apply(null, arguments);\n        };\n        var _init_with_threads_count = Module[\"_init_with_threads_count\"] = function() {\n          return (_init_with_threads_count = Module[\"_init_with_threads_count\"] = Module[\"asm\"][\"init_with_threads_count\"]).apply(null, arguments);\n        };\n        var _get_threads_count = Module[\"_get_threads_count\"] = function() {\n          return (_get_threads_count = Module[\"_get_threads_count\"] = Module[\"asm\"][\"get_threads_count\"]).apply(null, arguments);\n        };\n        var _register_tensor = Module[\"_register_tensor\"] = function() {\n          return (_register_tensor = Module[\"_register_tensor\"] = Module[\"asm\"][\"register_tensor\"]).apply(null, arguments);\n        };\n        var _dispose_data = Module[\"_dispose_data\"] = function() {\n          return (_dispose_data = Module[\"_dispose_data\"] = Module[\"asm\"][\"dispose_data\"]).apply(null, arguments);\n        };\n        var _dispose = Module[\"_dispose\"] = function() {\n          return (_dispose = Module[\"_dispose\"] = Module[\"asm\"][\"dispose\"]).apply(null, arguments);\n        };\n        var _Abs = Module[\"_Abs\"] = function() {\n          return (_Abs = Module[\"_Abs\"] = Module[\"asm\"][\"Abs\"]).apply(null, arguments);\n        };\n        var _Add = Module[\"_Add\"] = function() {\n          return (_Add = Module[\"_Add\"] = Module[\"asm\"][\"Add\"]).apply(null, arguments);\n        };\n        var _AddN = Module[\"_AddN\"] = function() {\n          return (_AddN = Module[\"_AddN\"] = Module[\"asm\"][\"AddN\"]).apply(null, arguments);\n        };\n        var _All = Module[\"_All\"] = function() {\n          return (_All = Module[\"_All\"] = Module[\"asm\"][\"All\"]).apply(null, arguments);\n        };\n        var _Any = Module[\"_Any\"] = function() {\n          return (_Any = Module[\"_Any\"] = Module[\"asm\"][\"Any\"]).apply(null, arguments);\n        };\n        var _ArgMax = Module[\"_ArgMax\"] = function() {\n          return (_ArgMax = Module[\"_ArgMax\"] = Module[\"asm\"][\"ArgMax\"]).apply(null, arguments);\n        };\n        var _AvgPool = Module[\"_AvgPool\"] = function() {\n          return (_AvgPool = Module[\"_AvgPool\"] = Module[\"asm\"][\"AvgPool\"]).apply(null, arguments);\n        };\n        var _BatchMatMul = Module[\"_BatchMatMul\"] = function() {\n          return (_BatchMatMul = Module[\"_BatchMatMul\"] = Module[\"asm\"][\"BatchMatMul\"]).apply(null, arguments);\n        };\n        var _Ceil = Module[\"_Ceil\"] = function() {\n          return (_Ceil = Module[\"_Ceil\"] = Module[\"asm\"][\"Ceil\"]).apply(null, arguments);\n        };\n        var _ClipByValue = Module[\"_ClipByValue\"] = function() {\n          return (_ClipByValue = Module[\"_ClipByValue\"] = Module[\"asm\"][\"ClipByValue\"]).apply(null, arguments);\n        };\n        var _Conv2D = Module[\"_Conv2D\"] = function() {\n          return (_Conv2D = Module[\"_Conv2D\"] = Module[\"asm\"][\"Conv2D\"]).apply(null, arguments);\n        };\n        var _Conv2DBackpropInput = Module[\"_Conv2DBackpropInput\"] = function() {\n          return (_Conv2DBackpropInput = Module[\"_Conv2DBackpropInput\"] = Module[\"asm\"][\"Conv2DBackpropInput\"]).apply(null, arguments);\n        };\n        var _Cos = Module[\"_Cos\"] = function() {\n          return (_Cos = Module[\"_Cos\"] = Module[\"asm\"][\"Cos\"]).apply(null, arguments);\n        };\n        var _Cosh = Module[\"_Cosh\"] = function() {\n          return (_Cosh = Module[\"_Cosh\"] = Module[\"asm\"][\"Cosh\"]).apply(null, arguments);\n        };\n        var _CropAndResize = Module[\"_CropAndResize\"] = function() {\n          return (_CropAndResize = Module[\"_CropAndResize\"] = Module[\"asm\"][\"CropAndResize\"]).apply(null, arguments);\n        };\n        var _Cumprod = Module[\"_Cumprod\"] = function() {\n          return (_Cumprod = Module[\"_Cumprod\"] = Module[\"asm\"][\"Cumprod\"]).apply(null, arguments);\n        };\n        var _Cumsum = Module[\"_Cumsum\"] = function() {\n          return (_Cumsum = Module[\"_Cumsum\"] = Module[\"asm\"][\"Cumsum\"]).apply(null, arguments);\n        };\n        var _DepthToSpace = Module[\"_DepthToSpace\"] = function() {\n          return (_DepthToSpace = Module[\"_DepthToSpace\"] = Module[\"asm\"][\"DepthToSpace\"]).apply(null, arguments);\n        };\n        var _DepthwiseConv2dNative = Module[\"_DepthwiseConv2dNative\"] = function() {\n          return (_DepthwiseConv2dNative = Module[\"_DepthwiseConv2dNative\"] = Module[\"asm\"][\"DepthwiseConv2dNative\"]).apply(null, arguments);\n        };\n        var _Elu = Module[\"_Elu\"] = function() {\n          return (_Elu = Module[\"_Elu\"] = Module[\"asm\"][\"Elu\"]).apply(null, arguments);\n        };\n        var _Equal = Module[\"_Equal\"] = function() {\n          return (_Equal = Module[\"_Equal\"] = Module[\"asm\"][\"Equal\"]).apply(null, arguments);\n        };\n        var _Exp = Module[\"_Exp\"] = function() {\n          return (_Exp = Module[\"_Exp\"] = Module[\"asm\"][\"Exp\"]).apply(null, arguments);\n        };\n        var _FlipLeftRight = Module[\"_FlipLeftRight\"] = function() {\n          return (_FlipLeftRight = Module[\"_FlipLeftRight\"] = Module[\"asm\"][\"FlipLeftRight\"]).apply(null, arguments);\n        };\n        var _Floor = Module[\"_Floor\"] = function() {\n          return (_Floor = Module[\"_Floor\"] = Module[\"asm\"][\"Floor\"]).apply(null, arguments);\n        };\n        var _FloorDiv = Module[\"_FloorDiv\"] = function() {\n          return (_FloorDiv = Module[\"_FloorDiv\"] = Module[\"asm\"][\"FloorDiv\"]).apply(null, arguments);\n        };\n        var _FusedBatchNorm = Module[\"_FusedBatchNorm\"] = function() {\n          return (_FusedBatchNorm = Module[\"_FusedBatchNorm\"] = Module[\"asm\"][\"FusedBatchNorm\"]).apply(null, arguments);\n        };\n        var _FusedConv2D = Module[\"_FusedConv2D\"] = function() {\n          return (_FusedConv2D = Module[\"_FusedConv2D\"] = Module[\"asm\"][\"FusedConv2D\"]).apply(null, arguments);\n        };\n        var _FusedDepthwiseConv2D = Module[\"_FusedDepthwiseConv2D\"] = function() {\n          return (_FusedDepthwiseConv2D = Module[\"_FusedDepthwiseConv2D\"] = Module[\"asm\"][\"FusedDepthwiseConv2D\"]).apply(null, arguments);\n        };\n        var _Gather = Module[\"_Gather\"] = function() {\n          return (_Gather = Module[\"_Gather\"] = Module[\"asm\"][\"Gather\"]).apply(null, arguments);\n        };\n        var _GatherNd = Module[\"_GatherNd\"] = function() {\n          return (_GatherNd = Module[\"_GatherNd\"] = Module[\"asm\"][\"GatherNd\"]).apply(null, arguments);\n        };\n        var _Greater = Module[\"_Greater\"] = function() {\n          return (_Greater = Module[\"_Greater\"] = Module[\"asm\"][\"Greater\"]).apply(null, arguments);\n        };\n        var _GreaterEqual = Module[\"_GreaterEqual\"] = function() {\n          return (_GreaterEqual = Module[\"_GreaterEqual\"] = Module[\"asm\"][\"GreaterEqual\"]).apply(null, arguments);\n        };\n        var _LeakyRelu = Module[\"_LeakyRelu\"] = function() {\n          return (_LeakyRelu = Module[\"_LeakyRelu\"] = Module[\"asm\"][\"LeakyRelu\"]).apply(null, arguments);\n        };\n        var _Less = Module[\"_Less\"] = function() {\n          return (_Less = Module[\"_Less\"] = Module[\"asm\"][\"Less\"]).apply(null, arguments);\n        };\n        var _LessEqual = Module[\"_LessEqual\"] = function() {\n          return (_LessEqual = Module[\"_LessEqual\"] = Module[\"asm\"][\"LessEqual\"]).apply(null, arguments);\n        };\n        var _Log = Module[\"_Log\"] = function() {\n          return (_Log = Module[\"_Log\"] = Module[\"asm\"][\"Log\"]).apply(null, arguments);\n        };\n        var _LogicalAnd = Module[\"_LogicalAnd\"] = function() {\n          return (_LogicalAnd = Module[\"_LogicalAnd\"] = Module[\"asm\"][\"LogicalAnd\"]).apply(null, arguments);\n        };\n        var _LogicalNot = Module[\"_LogicalNot\"] = function() {\n          return (_LogicalNot = Module[\"_LogicalNot\"] = Module[\"asm\"][\"LogicalNot\"]).apply(null, arguments);\n        };\n        var _LogicalOr = Module[\"_LogicalOr\"] = function() {\n          return (_LogicalOr = Module[\"_LogicalOr\"] = Module[\"asm\"][\"LogicalOr\"]).apply(null, arguments);\n        };\n        var _LogicalXor = Module[\"_LogicalXor\"] = function() {\n          return (_LogicalXor = Module[\"_LogicalXor\"] = Module[\"asm\"][\"LogicalXor\"]).apply(null, arguments);\n        };\n        var _Max = Module[\"_Max\"] = function() {\n          return (_Max = Module[\"_Max\"] = Module[\"asm\"][\"Max\"]).apply(null, arguments);\n        };\n        var _MaxPool = Module[\"_MaxPool\"] = function() {\n          return (_MaxPool = Module[\"_MaxPool\"] = Module[\"asm\"][\"MaxPool\"]).apply(null, arguments);\n        };\n        var _Maximum = Module[\"_Maximum\"] = function() {\n          return (_Maximum = Module[\"_Maximum\"] = Module[\"asm\"][\"Maximum\"]).apply(null, arguments);\n        };\n        var _Mean = Module[\"_Mean\"] = function() {\n          return (_Mean = Module[\"_Mean\"] = Module[\"asm\"][\"Mean\"]).apply(null, arguments);\n        };\n        var _Min = Module[\"_Min\"] = function() {\n          return (_Min = Module[\"_Min\"] = Module[\"asm\"][\"Min\"]).apply(null, arguments);\n        };\n        var _Minimum = Module[\"_Minimum\"] = function() {\n          return (_Minimum = Module[\"_Minimum\"] = Module[\"asm\"][\"Minimum\"]).apply(null, arguments);\n        };\n        var _MirrorPad = Module[\"_MirrorPad\"] = function() {\n          return (_MirrorPad = Module[\"_MirrorPad\"] = Module[\"asm\"][\"MirrorPad\"]).apply(null, arguments);\n        };\n        var _Multiply = Module[\"_Multiply\"] = function() {\n          return (_Multiply = Module[\"_Multiply\"] = Module[\"asm\"][\"Multiply\"]).apply(null, arguments);\n        };\n        var _Neg = Module[\"_Neg\"] = function() {\n          return (_Neg = Module[\"_Neg\"] = Module[\"asm\"][\"Neg\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV3 = Module[\"_NonMaxSuppressionV3\"] = function() {\n          return (_NonMaxSuppressionV3 = Module[\"_NonMaxSuppressionV3\"] = Module[\"asm\"][\"NonMaxSuppressionV3\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV4 = Module[\"_NonMaxSuppressionV4\"] = function() {\n          return (_NonMaxSuppressionV4 = Module[\"_NonMaxSuppressionV4\"] = Module[\"asm\"][\"NonMaxSuppressionV4\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV5 = Module[\"_NonMaxSuppressionV5\"] = function() {\n          return (_NonMaxSuppressionV5 = Module[\"_NonMaxSuppressionV5\"] = Module[\"asm\"][\"NonMaxSuppressionV5\"]).apply(null, arguments);\n        };\n        var _NotEqual = Module[\"_NotEqual\"] = function() {\n          return (_NotEqual = Module[\"_NotEqual\"] = Module[\"asm\"][\"NotEqual\"]).apply(null, arguments);\n        };\n        var _OneHot = Module[\"_OneHot\"] = function() {\n          return (_OneHot = Module[\"_OneHot\"] = Module[\"asm\"][\"OneHot\"]).apply(null, arguments);\n        };\n        var _PadV2 = Module[\"_PadV2\"] = function() {\n          return (_PadV2 = Module[\"_PadV2\"] = Module[\"asm\"][\"PadV2\"]).apply(null, arguments);\n        };\n        var _Pow = Module[\"_Pow\"] = function() {\n          return (_Pow = Module[\"_Pow\"] = Module[\"asm\"][\"Pow\"]).apply(null, arguments);\n        };\n        var _Prelu = Module[\"_Prelu\"] = function() {\n          return (_Prelu = Module[\"_Prelu\"] = Module[\"asm\"][\"Prelu\"]).apply(null, arguments);\n        };\n        var _Prod = Module[\"_Prod\"] = function() {\n          return (_Prod = Module[\"_Prod\"] = Module[\"asm\"][\"Prod\"]).apply(null, arguments);\n        };\n        var _RealDiv = Module[\"_RealDiv\"] = function() {\n          return (_RealDiv = Module[\"_RealDiv\"] = Module[\"asm\"][\"RealDiv\"]).apply(null, arguments);\n        };\n        var _Relu = Module[\"_Relu\"] = function() {\n          return (_Relu = Module[\"_Relu\"] = Module[\"asm\"][\"Relu\"]).apply(null, arguments);\n        };\n        var _Relu6 = Module[\"_Relu6\"] = function() {\n          return (_Relu6 = Module[\"_Relu6\"] = Module[\"asm\"][\"Relu6\"]).apply(null, arguments);\n        };\n        var _ResizeBilinear = Module[\"_ResizeBilinear\"] = function() {\n          return (_ResizeBilinear = Module[\"_ResizeBilinear\"] = Module[\"asm\"][\"ResizeBilinear\"]).apply(null, arguments);\n        };\n        var _ResizeNearestNeighbor = Module[\"_ResizeNearestNeighbor\"] = function() {\n          return (_ResizeNearestNeighbor = Module[\"_ResizeNearestNeighbor\"] = Module[\"asm\"][\"ResizeNearestNeighbor\"]).apply(null, arguments);\n        };\n        var _Reverse = Module[\"_Reverse\"] = function() {\n          return (_Reverse = Module[\"_Reverse\"] = Module[\"asm\"][\"Reverse\"]).apply(null, arguments);\n        };\n        var _RotateWithOffset = Module[\"_RotateWithOffset\"] = function() {\n          return (_RotateWithOffset = Module[\"_RotateWithOffset\"] = Module[\"asm\"][\"RotateWithOffset\"]).apply(null, arguments);\n        };\n        var _Round = Module[\"_Round\"] = function() {\n          return (_Round = Module[\"_Round\"] = Module[\"asm\"][\"Round\"]).apply(null, arguments);\n        };\n        var _Rsqrt = Module[\"_Rsqrt\"] = function() {\n          return (_Rsqrt = Module[\"_Rsqrt\"] = Module[\"asm\"][\"Rsqrt\"]).apply(null, arguments);\n        };\n        var _ScatterNd = Module[\"_ScatterNd\"] = function() {\n          return (_ScatterNd = Module[\"_ScatterNd\"] = Module[\"asm\"][\"ScatterNd\"]).apply(null, arguments);\n        };\n        var _SelectV2 = Module[\"_SelectV2\"] = function() {\n          return (_SelectV2 = Module[\"_SelectV2\"] = Module[\"asm\"][\"SelectV2\"]).apply(null, arguments);\n        };\n        var _Sigmoid = Module[\"_Sigmoid\"] = function() {\n          return (_Sigmoid = Module[\"_Sigmoid\"] = Module[\"asm\"][\"Sigmoid\"]).apply(null, arguments);\n        };\n        var _Sin = Module[\"_Sin\"] = function() {\n          return (_Sin = Module[\"_Sin\"] = Module[\"asm\"][\"Sin\"]).apply(null, arguments);\n        };\n        var _Softmax = Module[\"_Softmax\"] = function() {\n          return (_Softmax = Module[\"_Softmax\"] = Module[\"asm\"][\"Softmax\"]).apply(null, arguments);\n        };\n        var _SparseFillEmptyRows = Module[\"_SparseFillEmptyRows\"] = function() {\n          return (_SparseFillEmptyRows = Module[\"_SparseFillEmptyRows\"] = Module[\"asm\"][\"SparseFillEmptyRows\"]).apply(null, arguments);\n        };\n        var _SparseReshape = Module[\"_SparseReshape\"] = function() {\n          return (_SparseReshape = Module[\"_SparseReshape\"] = Module[\"asm\"][\"SparseReshape\"]).apply(null, arguments);\n        };\n        var _SparseSegmentReduction = Module[\"_SparseSegmentReduction\"] = function() {\n          return (_SparseSegmentReduction = Module[\"_SparseSegmentReduction\"] = Module[\"asm\"][\"SparseSegmentReduction\"]).apply(null, arguments);\n        };\n        var _Sqrt = Module[\"_Sqrt\"] = function() {\n          return (_Sqrt = Module[\"_Sqrt\"] = Module[\"asm\"][\"Sqrt\"]).apply(null, arguments);\n        };\n        var _Square = Module[\"_Square\"] = function() {\n          return (_Square = Module[\"_Square\"] = Module[\"asm\"][\"Square\"]).apply(null, arguments);\n        };\n        var _SquaredDifference = Module[\"_SquaredDifference\"] = function() {\n          return (_SquaredDifference = Module[\"_SquaredDifference\"] = Module[\"asm\"][\"SquaredDifference\"]).apply(null, arguments);\n        };\n        var _Step = Module[\"_Step\"] = function() {\n          return (_Step = Module[\"_Step\"] = Module[\"asm\"][\"Step\"]).apply(null, arguments);\n        };\n        var _StridedSlice = Module[\"_StridedSlice\"] = function() {\n          return (_StridedSlice = Module[\"_StridedSlice\"] = Module[\"asm\"][\"StridedSlice\"]).apply(null, arguments);\n        };\n        var _Sub = Module[\"_Sub\"] = function() {\n          return (_Sub = Module[\"_Sub\"] = Module[\"asm\"][\"Sub\"]).apply(null, arguments);\n        };\n        var _Sum = Module[\"_Sum\"] = function() {\n          return (_Sum = Module[\"_Sum\"] = Module[\"asm\"][\"Sum\"]).apply(null, arguments);\n        };\n        var _Tan = Module[\"_Tan\"] = function() {\n          return (_Tan = Module[\"_Tan\"] = Module[\"asm\"][\"Tan\"]).apply(null, arguments);\n        };\n        var _Tanh = Module[\"_Tanh\"] = function() {\n          return (_Tanh = Module[\"_Tanh\"] = Module[\"asm\"][\"Tanh\"]).apply(null, arguments);\n        };\n        var _Tile = Module[\"_Tile\"] = function() {\n          return (_Tile = Module[\"_Tile\"] = Module[\"asm\"][\"Tile\"]).apply(null, arguments);\n        };\n        var _TopK = Module[\"_TopK\"] = function() {\n          return (_TopK = Module[\"_TopK\"] = Module[\"asm\"][\"TopK\"]).apply(null, arguments);\n        };\n        var _Transform = Module[\"_Transform\"] = function() {\n          return (_Transform = Module[\"_Transform\"] = Module[\"asm\"][\"Transform\"]).apply(null, arguments);\n        };\n        var _Transpose = Module[\"_Transpose\"] = function() {\n          return (_Transpose = Module[\"_Transpose\"] = Module[\"asm\"][\"Transpose\"]).apply(null, arguments);\n        };\n        var __FusedMatMul = Module[\"__FusedMatMul\"] = function() {\n          return (__FusedMatMul = Module[\"__FusedMatMul\"] = Module[\"asm\"][\"_FusedMatMul\"]).apply(null, arguments);\n        };\n        var _malloc = Module[\"_malloc\"] = function() {\n          return (_malloc = Module[\"_malloc\"] = Module[\"asm\"][\"malloc\"]).apply(null, arguments);\n        };\n        var _free = Module[\"_free\"] = function() {\n          return (_free = Module[\"_free\"] = Module[\"asm\"][\"free\"]).apply(null, arguments);\n        };\n        var __emscripten_tls_init = Module[\"__emscripten_tls_init\"] = function() {\n          return (__emscripten_tls_init = Module[\"__emscripten_tls_init\"] = Module[\"asm\"][\"_emscripten_tls_init\"]).apply(null, arguments);\n        };\n        var _pthread_self = Module[\"_pthread_self\"] = function() {\n          return (_pthread_self = Module[\"_pthread_self\"] = Module[\"asm\"][\"pthread_self\"]).apply(null, arguments);\n        };\n        var ___errno_location = Module[\"___errno_location\"] = function() {\n          return (___errno_location = Module[\"___errno_location\"] = Module[\"asm\"][\"__errno_location\"]).apply(null, arguments);\n        };\n        var __emscripten_thread_init = Module[\"__emscripten_thread_init\"] = function() {\n          return (__emscripten_thread_init = Module[\"__emscripten_thread_init\"] = Module[\"asm\"][\"_emscripten_thread_init\"]).apply(null, arguments);\n        };\n        var __emscripten_thread_crashed = Module[\"__emscripten_thread_crashed\"] = function() {\n          return (__emscripten_thread_crashed = Module[\"__emscripten_thread_crashed\"] = Module[\"asm\"][\"_emscripten_thread_crashed\"]).apply(null, arguments);\n        };\n        var _emscripten_main_thread_process_queued_calls = Module[\"_emscripten_main_thread_process_queued_calls\"] = function() {\n          return (_emscripten_main_thread_process_queued_calls = Module[\"_emscripten_main_thread_process_queued_calls\"] = Module[\"asm\"][\"emscripten_main_thread_process_queued_calls\"]).apply(null, arguments);\n        };\n        var _emscripten_main_browser_thread_id = Module[\"_emscripten_main_browser_thread_id\"] = function() {\n          return (_emscripten_main_browser_thread_id = Module[\"_emscripten_main_browser_thread_id\"] = Module[\"asm\"][\"emscripten_main_browser_thread_id\"]).apply(null, arguments);\n        };\n        var _emscripten_run_in_main_runtime_thread_js = Module[\"_emscripten_run_in_main_runtime_thread_js\"] = function() {\n          return (_emscripten_run_in_main_runtime_thread_js = Module[\"_emscripten_run_in_main_runtime_thread_js\"] = Module[\"asm\"][\"emscripten_run_in_main_runtime_thread_js\"]).apply(null, arguments);\n        };\n        var _emscripten_dispatch_to_thread_ = Module[\"_emscripten_dispatch_to_thread_\"] = function() {\n          return (_emscripten_dispatch_to_thread_ = Module[\"_emscripten_dispatch_to_thread_\"] = Module[\"asm\"][\"emscripten_dispatch_to_thread_\"]).apply(null, arguments);\n        };\n        var __emscripten_proxy_execute_task_queue = Module[\"__emscripten_proxy_execute_task_queue\"] = function() {\n          return (__emscripten_proxy_execute_task_queue = Module[\"__emscripten_proxy_execute_task_queue\"] = Module[\"asm\"][\"_emscripten_proxy_execute_task_queue\"]).apply(null, arguments);\n        };\n        var __emscripten_thread_free_data = Module[\"__emscripten_thread_free_data\"] = function() {\n          return (__emscripten_thread_free_data = Module[\"__emscripten_thread_free_data\"] = Module[\"asm\"][\"_emscripten_thread_free_data\"]).apply(null, arguments);\n        };\n        var __emscripten_thread_exit = Module[\"__emscripten_thread_exit\"] = function() {\n          return (__emscripten_thread_exit = Module[\"__emscripten_thread_exit\"] = Module[\"asm\"][\"_emscripten_thread_exit\"]).apply(null, arguments);\n        };\n        var _emscripten_stack_set_limits = Module[\"_emscripten_stack_set_limits\"] = function() {\n          return (_emscripten_stack_set_limits = Module[\"_emscripten_stack_set_limits\"] = Module[\"asm\"][\"emscripten_stack_set_limits\"]).apply(null, arguments);\n        };\n        var stackSave = Module[\"stackSave\"] = function() {\n          return (stackSave = Module[\"stackSave\"] = Module[\"asm\"][\"stackSave\"]).apply(null, arguments);\n        };\n        var stackRestore = Module[\"stackRestore\"] = function() {\n          return (stackRestore = Module[\"stackRestore\"] = Module[\"asm\"][\"stackRestore\"]).apply(null, arguments);\n        };\n        var stackAlloc = Module[\"stackAlloc\"] = function() {\n          return (stackAlloc = Module[\"stackAlloc\"] = Module[\"asm\"][\"stackAlloc\"]).apply(null, arguments);\n        };\n        var dynCall_iijjiiii = Module[\"dynCall_iijjiiii\"] = function() {\n          return (dynCall_iijjiiii = Module[\"dynCall_iijjiiii\"] = Module[\"asm\"][\"dynCall_iijjiiii\"]).apply(null, arguments);\n        };\n        var dynCall_jiji = Module[\"dynCall_jiji\"] = function() {\n          return (dynCall_jiji = Module[\"dynCall_jiji\"] = Module[\"asm\"][\"dynCall_jiji\"]).apply(null, arguments);\n        };\n        Module[\"keepRuntimeAlive\"] = keepRuntimeAlive;\n        Module[\"wasmMemory\"] = wasmMemory;\n        Module[\"cwrap\"] = cwrap;\n        Module[\"ExitStatus\"] = ExitStatus;\n        Module[\"PThread\"] = PThread;\n        var calledRun;\n        dependenciesFulfilled = function runCaller() {\n          if (!calledRun)\n            run();\n          if (!calledRun)\n            dependenciesFulfilled = runCaller;\n        };\n        function run(args) {\n          args = args || arguments_;\n          if (runDependencies > 0) {\n            return;\n          }\n          if (ENVIRONMENT_IS_PTHREAD) {\n            readyPromiseResolve(Module);\n            initRuntime();\n            postMessage({ \"cmd\": \"loaded\" });\n            return;\n          }\n          preRun();\n          if (runDependencies > 0) {\n            return;\n          }\n          function doRun() {\n            if (calledRun)\n              return;\n            calledRun = true;\n            Module[\"calledRun\"] = true;\n            if (ABORT)\n              return;\n            initRuntime();\n            readyPromiseResolve(Module);\n            if (Module[\"onRuntimeInitialized\"])\n              Module[\"onRuntimeInitialized\"]();\n            postRun();\n          }\n          if (Module[\"setStatus\"]) {\n            Module[\"setStatus\"](\"Running...\");\n            setTimeout(function() {\n              setTimeout(function() {\n                Module[\"setStatus\"](\"\");\n              }, 1);\n              doRun();\n            }, 1);\n          } else {\n            doRun();\n          }\n        }\n        if (Module[\"preInit\"]) {\n          if (typeof Module[\"preInit\"] == \"function\")\n            Module[\"preInit\"] = [Module[\"preInit\"]];\n          while (Module[\"preInit\"].length > 0) {\n            Module[\"preInit\"].pop()();\n          }\n        }\n        run();\n        var listenersAdded;\n        if (beforeListeners) {\n          listenersAdded = { uncaughtException: process.listeners(\"uncaughtException\").filter(function(listener) {\n            return !beforeListeners.uncaughtException.indexOf(listener) > -1;\n          }), unhandledRejection: process.listeners(\"unhandledRejection\").filter(function(listener) {\n            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;\n          }) };\n        }\n        var actualModule;\n        if (typeof WasmBackendModule !== \"undefined\") {\n          actualModule = WasmBackendModule;\n        } else if (typeof WasmBackendModuleThreadedSimd3 !== \"undefined\") {\n          actualModule = WasmBackendModuleThreadedSimd3;\n        } else {\n          throw new Error(\"Could not find wasm module in post.js\");\n        }\n        if (listenersAdded) {\n          var tmpDispose = actualModule[\"_dispose\"];\n          actualModule[\"_dispose\"] = function() {\n            tmpDispose();\n            listenersAdded.uncaughtException.forEach(function(listener) {\n              process.removeListener(\"uncaughtException\", listener);\n            });\n            listenersAdded.unhandledRejection.forEach(function(listener) {\n              process.removeListener(\"unhandledRejection\", listener);\n            });\n          };\n        }\n        return WasmBackendModuleThreadedSimd3.ready;\n      };\n    })();\n    if (typeof exports === \"object\" && typeof module === \"object\")\n      module.exports = WasmBackendModuleThreadedSimd2;\n    else if (typeof define === \"function\" && define[\"amd\"])\n      define([], function() {\n        return WasmBackendModuleThreadedSimd2;\n      });\n    else if (typeof exports === \"object\")\n      exports[\"WasmBackendModuleThreadedSimd\"] = WasmBackendModuleThreadedSimd2;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js\nvar require_tfjs_backend_wasm_threaded_simd_worker = __commonJS({\n  \"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js\"(exports, module) {\n    module.exports.wasmWorkerContents = `\"use strict\";var Module={};var ENVIRONMENT_IS_NODE=typeof process==\"object\"&&typeof process.versions==\"object\"&&typeof process.versions.node==\"string\";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require(\"worker_threads\");var parentPort=nodeWorkerThreads.parentPort;parentPort.on(\"message\",data=>onmessage({data:data}));var fs=require(\"fs\");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,\"utf8\"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(\" \");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+\"\n\");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(\" \");postMessage({cmd:\"alert\",text:text,threadId:Module[\"_pthread_self\"]()})}var err=threadPrintErr;self.alert=threadAlert;Module[\"instantiateWasm\"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module[\"wasmModule\"],info);receiveInstance(instance);Module[\"wasmModule\"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd===\"load\"){Module[\"wasmModule\"]=e.data.wasmModule;Module[\"wasmMemory\"]=e.data.wasmMemory;Module[\"buffer\"]=Module[\"wasmMemory\"].buffer;Module[\"ENVIRONMENT_IS_PTHREAD\"]=true;if(typeof e.data.urlOrBlob==\"string\"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd===\"run\"){Module[\"__performance_now_clock_drift\"]=performance.now()-e.data.time;Module[\"__emscripten_thread_init\"](e.data.pthread_ptr,0,0,1);Module[\"establishStackSpace\"]();Module[\"PThread\"].receiveObjectTransfer(e.data);Module[\"PThread\"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module[\"executeNotifiedProxyingQueue\"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module[\"invokeEntryPoint\"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!=\"unwind\"){if(ex instanceof Module[\"ExitStatus\"]){if(Module[\"keepRuntimeAlive\"]()){}else{Module[\"__emscripten_thread_exit\"](ex.status)}}else{throw ex}}}}else if(e.data.cmd===\"cancel\"){if(Module[\"_pthread_self\"]()){Module[\"__emscripten_thread_exit\"](-1)}}else if(e.data.target===\"setimmediate\"){}else if(e.data.cmd===\"processProxyingQueue\"){if(initializedJS){Module[\"executeNotifiedProxyingQueue\"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err(\"worker.js received unknown command \"+e.data.cmd);err(e.data)}}catch(ex){if(Module[\"__emscripten_thread_crashed\"]){Module[\"__emscripten_thread_crashed\"]()}throw ex}};`;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js\nvar require_tfjs_backend_wasm = __commonJS({\n  \"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js\"(exports, module) {\n    var WasmBackendModule2 = (() => {\n      var _scriptDir = typeof document !== \"undefined\" && document.currentScript ? document.currentScript.src : void 0;\n      if (typeof __filename !== \"undefined\")\n        _scriptDir = _scriptDir || __filename;\n      return function(WasmBackendModule3) {\n        WasmBackendModule3 = WasmBackendModule3 || {};\n        var Module = typeof WasmBackendModule3 != \"undefined\" ? WasmBackendModule3 : {};\n        var readyPromiseResolve, readyPromiseReject;\n        Module[\"ready\"] = new Promise(function(resolve, reject) {\n          readyPromiseResolve = resolve;\n          readyPromiseReject = reject;\n        });\n        var beforeListeners;\n        if (typeof process !== \"undefined\" && process.listeners) {\n          beforeListeners = { uncaughtException: process.listeners(\"uncaughtException\"), unhandledRejection: process.listeners(\"unhandledRejection\") };\n        }\n        var moduleOverrides = Object.assign({}, Module);\n        var arguments_ = [];\n        var thisProgram = \"./this.program\";\n        var quit_ = (status, toThrow) => {\n          throw toThrow;\n        };\n        var ENVIRONMENT_IS_WEB = typeof window == \"object\";\n        var ENVIRONMENT_IS_WORKER = typeof importScripts == \"function\";\n        var ENVIRONMENT_IS_NODE = typeof process == \"object\" && typeof process.versions == \"object\" && typeof process.versions.node == \"string\";\n        var scriptDirectory = \"\";\n        function locateFile(path) {\n          if (Module[\"locateFile\"]) {\n            return Module[\"locateFile\"](path, scriptDirectory);\n          }\n          return scriptDirectory + path;\n        }\n        var read_, readAsync, readBinary, setWindowTitle;\n        function logExceptionOnExit(e) {\n          if (e instanceof ExitStatus)\n            return;\n          let toLog = e;\n          err(\"exiting due to exception: \" + toLog);\n        }\n        if (ENVIRONMENT_IS_NODE) {\n          if (ENVIRONMENT_IS_WORKER) {\n            scriptDirectory = require_path().dirname(scriptDirectory) + \"/\";\n          } else {\n            scriptDirectory = __dirname + \"/\";\n          }\n          var fs, nodePath;\n          if (typeof __require === \"function\") {\n            fs = require_fs();\n            nodePath = require_path();\n          }\n          read_ = (filename, binary) => {\n            filename = nodePath[\"normalize\"](filename);\n            return fs.readFileSync(filename, binary ? void 0 : \"utf8\");\n          };\n          readBinary = (filename) => {\n            var ret = read_(filename, true);\n            if (!ret.buffer) {\n              ret = new Uint8Array(ret);\n            }\n            return ret;\n          };\n          readAsync = (filename, onload, onerror) => {\n            filename = nodePath[\"normalize\"](filename);\n            fs.readFile(filename, function(err2, data) {\n              if (err2)\n                onerror(err2);\n              else\n                onload(data.buffer);\n            });\n          };\n          if (process[\"argv\"].length > 1) {\n            thisProgram = process[\"argv\"][1].replace(/\\\\/g, \"/\");\n          }\n          arguments_ = process[\"argv\"].slice(2);\n          process[\"on\"](\"uncaughtException\", function(ex) {\n            if (!(ex instanceof ExitStatus)) {\n              throw ex;\n            }\n          });\n          process[\"on\"](\"unhandledRejection\", function(reason) {\n            throw reason;\n          });\n          quit_ = (status, toThrow) => {\n            if (keepRuntimeAlive()) {\n              process[\"exitCode\"] = status;\n              throw toThrow;\n            }\n            logExceptionOnExit(toThrow);\n            process[\"exit\"](status);\n          };\n          Module[\"inspect\"] = function() {\n            return \"[Emscripten Module object]\";\n          };\n        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {\n          if (ENVIRONMENT_IS_WORKER) {\n            scriptDirectory = self.location.href;\n          } else if (typeof document != \"undefined\" && document.currentScript) {\n            scriptDirectory = document.currentScript.src;\n          }\n          if (_scriptDir) {\n            scriptDirectory = _scriptDir;\n          }\n          if (scriptDirectory.indexOf(\"blob:\") !== 0) {\n            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, \"\").lastIndexOf(\"/\") + 1);\n          } else {\n            scriptDirectory = \"\";\n          }\n          {\n            read_ = (url) => {\n              var xhr = new XMLHttpRequest();\n              xhr.open(\"GET\", url, false);\n              xhr.send(null);\n              return xhr.responseText;\n            };\n            if (ENVIRONMENT_IS_WORKER) {\n              readBinary = (url) => {\n                var xhr = new XMLHttpRequest();\n                xhr.open(\"GET\", url, false);\n                xhr.responseType = \"arraybuffer\";\n                xhr.send(null);\n                return new Uint8Array(xhr.response);\n              };\n            }\n            readAsync = (url, onload, onerror) => {\n              var xhr = new XMLHttpRequest();\n              xhr.open(\"GET\", url, true);\n              xhr.responseType = \"arraybuffer\";\n              xhr.onload = () => {\n                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {\n                  onload(xhr.response);\n                  return;\n                }\n                onerror();\n              };\n              xhr.onerror = onerror;\n              xhr.send(null);\n            };\n          }\n          setWindowTitle = (title) => document.title = title;\n        } else {\n        }\n        var out = Module[\"print\"] || console.log.bind(console);\n        var err = Module[\"printErr\"] || console.warn.bind(console);\n        Object.assign(Module, moduleOverrides);\n        moduleOverrides = null;\n        if (Module[\"arguments\"])\n          arguments_ = Module[\"arguments\"];\n        if (Module[\"thisProgram\"])\n          thisProgram = Module[\"thisProgram\"];\n        if (Module[\"quit\"])\n          quit_ = Module[\"quit\"];\n        var POINTER_SIZE = 4;\n        var wasmBinary;\n        if (Module[\"wasmBinary\"])\n          wasmBinary = Module[\"wasmBinary\"];\n        var noExitRuntime = Module[\"noExitRuntime\"] || true;\n        if (typeof WebAssembly != \"object\") {\n          abort(\"no native wasm support detected\");\n        }\n        var wasmMemory;\n        var ABORT = false;\n        var EXITSTATUS;\n        function assert3(condition, text) {\n          if (!condition) {\n            abort(text);\n          }\n        }\n        var UTF8Decoder = typeof TextDecoder != \"undefined\" ? new TextDecoder(\"utf8\") : void 0;\n        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {\n          var endIdx = idx + maxBytesToRead;\n          var endPtr = idx;\n          while (heapOrArray[endPtr] && !(endPtr >= endIdx))\n            ++endPtr;\n          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {\n            return UTF8Decoder.decode(heapOrArray.subarray(idx, endPtr));\n          }\n          var str = \"\";\n          while (idx < endPtr) {\n            var u0 = heapOrArray[idx++];\n            if (!(u0 & 128)) {\n              str += String.fromCharCode(u0);\n              continue;\n            }\n            var u1 = heapOrArray[idx++] & 63;\n            if ((u0 & 224) == 192) {\n              str += String.fromCharCode((u0 & 31) << 6 | u1);\n              continue;\n            }\n            var u2 = heapOrArray[idx++] & 63;\n            if ((u0 & 240) == 224) {\n              u0 = (u0 & 15) << 12 | u1 << 6 | u2;\n            } else {\n              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;\n            }\n            if (u0 < 65536) {\n              str += String.fromCharCode(u0);\n            } else {\n              var ch = u0 - 65536;\n              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);\n            }\n          }\n          return str;\n        }\n        function UTF8ToString(ptr, maxBytesToRead) {\n          return ptr ? UTF8ArrayToString(HEAPU8, ptr, maxBytesToRead) : \"\";\n        }\n        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {\n          if (!(maxBytesToWrite > 0))\n            return 0;\n          var startIdx = outIdx;\n          var endIdx = outIdx + maxBytesToWrite - 1;\n          for (var i = 0; i < str.length; ++i) {\n            var u = str.charCodeAt(i);\n            if (u >= 55296 && u <= 57343) {\n              var u1 = str.charCodeAt(++i);\n              u = 65536 + ((u & 1023) << 10) | u1 & 1023;\n            }\n            if (u <= 127) {\n              if (outIdx >= endIdx)\n                break;\n              heap[outIdx++] = u;\n            } else if (u <= 2047) {\n              if (outIdx + 1 >= endIdx)\n                break;\n              heap[outIdx++] = 192 | u >> 6;\n              heap[outIdx++] = 128 | u & 63;\n            } else if (u <= 65535) {\n              if (outIdx + 2 >= endIdx)\n                break;\n              heap[outIdx++] = 224 | u >> 12;\n              heap[outIdx++] = 128 | u >> 6 & 63;\n              heap[outIdx++] = 128 | u & 63;\n            } else {\n              if (outIdx + 3 >= endIdx)\n                break;\n              heap[outIdx++] = 240 | u >> 18;\n              heap[outIdx++] = 128 | u >> 12 & 63;\n              heap[outIdx++] = 128 | u >> 6 & 63;\n              heap[outIdx++] = 128 | u & 63;\n            }\n          }\n          heap[outIdx] = 0;\n          return outIdx - startIdx;\n        }\n        function stringToUTF8(str, outPtr, maxBytesToWrite) {\n          return stringToUTF8Array(str, HEAPU8, outPtr, maxBytesToWrite);\n        }\n        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;\n        function updateGlobalBufferAndViews(buf) {\n          buffer2 = buf;\n          Module[\"HEAP8\"] = HEAP8 = new Int8Array(buf);\n          Module[\"HEAP16\"] = HEAP16 = new Int16Array(buf);\n          Module[\"HEAP32\"] = HEAP32 = new Int32Array(buf);\n          Module[\"HEAPU8\"] = HEAPU8 = new Uint8Array(buf);\n          Module[\"HEAPU16\"] = HEAPU16 = new Uint16Array(buf);\n          Module[\"HEAPU32\"] = HEAPU32 = new Uint32Array(buf);\n          Module[\"HEAPF32\"] = HEAPF32 = new Float32Array(buf);\n          Module[\"HEAPF64\"] = HEAPF64 = new Float64Array(buf);\n        }\n        var INITIAL_MEMORY = Module[\"INITIAL_MEMORY\"] || 16777216;\n        var wasmTable;\n        var __ATPRERUN__ = [];\n        var __ATINIT__ = [];\n        var __ATPOSTRUN__ = [];\n        var runtimeInitialized = false;\n        function keepRuntimeAlive() {\n          return noExitRuntime;\n        }\n        function preRun() {\n          if (Module[\"preRun\"]) {\n            if (typeof Module[\"preRun\"] == \"function\")\n              Module[\"preRun\"] = [Module[\"preRun\"]];\n            while (Module[\"preRun\"].length) {\n              addOnPreRun(Module[\"preRun\"].shift());\n            }\n          }\n          callRuntimeCallbacks(__ATPRERUN__);\n        }\n        function initRuntime() {\n          runtimeInitialized = true;\n          callRuntimeCallbacks(__ATINIT__);\n        }\n        function postRun() {\n          if (Module[\"postRun\"]) {\n            if (typeof Module[\"postRun\"] == \"function\")\n              Module[\"postRun\"] = [Module[\"postRun\"]];\n            while (Module[\"postRun\"].length) {\n              addOnPostRun(Module[\"postRun\"].shift());\n            }\n          }\n          callRuntimeCallbacks(__ATPOSTRUN__);\n        }\n        function addOnPreRun(cb) {\n          __ATPRERUN__.unshift(cb);\n        }\n        function addOnInit(cb) {\n          __ATINIT__.unshift(cb);\n        }\n        function addOnPostRun(cb) {\n          __ATPOSTRUN__.unshift(cb);\n        }\n        var runDependencies = 0;\n        var runDependencyWatcher = null;\n        var dependenciesFulfilled = null;\n        function addRunDependency(id) {\n          runDependencies++;\n          if (Module[\"monitorRunDependencies\"]) {\n            Module[\"monitorRunDependencies\"](runDependencies);\n          }\n        }\n        function removeRunDependency(id) {\n          runDependencies--;\n          if (Module[\"monitorRunDependencies\"]) {\n            Module[\"monitorRunDependencies\"](runDependencies);\n          }\n          if (runDependencies == 0) {\n            if (runDependencyWatcher !== null) {\n              clearInterval(runDependencyWatcher);\n              runDependencyWatcher = null;\n            }\n            if (dependenciesFulfilled) {\n              var callback = dependenciesFulfilled;\n              dependenciesFulfilled = null;\n              callback();\n            }\n          }\n        }\n        function abort(what) {\n          {\n            if (Module[\"onAbort\"]) {\n              Module[\"onAbort\"](what);\n            }\n          }\n          what = \"Aborted(\" + what + \")\";\n          err(what);\n          ABORT = true;\n          EXITSTATUS = 1;\n          what += \". Build with -sASSERTIONS for more info.\";\n          var e = new WebAssembly.RuntimeError(what);\n          readyPromiseReject(e);\n          throw e;\n        }\n        var dataURIPrefix = \"data:application/octet-stream;base64,\";\n        function isDataURI(filename) {\n          return filename.startsWith(dataURIPrefix);\n        }\n        function isFileURI(filename) {\n          return filename.startsWith(\"file://\");\n        }\n        var wasmBinaryFile;\n        wasmBinaryFile = \"tfjs-backend-wasm.wasm\";\n        if (!isDataURI(wasmBinaryFile)) {\n          wasmBinaryFile = locateFile(wasmBinaryFile);\n        }\n        function getBinary(file) {\n          try {\n            if (file == wasmBinaryFile && wasmBinary) {\n              return new Uint8Array(wasmBinary);\n            }\n            if (readBinary) {\n              return readBinary(file);\n            }\n            throw \"both async and sync fetching of the wasm failed\";\n          } catch (err2) {\n            abort(err2);\n          }\n        }\n        function getBinaryPromise() {\n          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {\n            if (typeof fetch == \"function\" && !isFileURI(wasmBinaryFile)) {\n              return fetch(wasmBinaryFile, { credentials: \"same-origin\" }).then(function(response) {\n                if (!response[\"ok\"]) {\n                  throw \"failed to load wasm binary file at '\" + wasmBinaryFile + \"'\";\n                }\n                return response[\"arrayBuffer\"]();\n              }).catch(function() {\n                return getBinary(wasmBinaryFile);\n              });\n            } else {\n              if (readAsync) {\n                return new Promise(function(resolve, reject) {\n                  readAsync(wasmBinaryFile, function(response) {\n                    resolve(new Uint8Array(response));\n                  }, reject);\n                });\n              }\n            }\n          }\n          return Promise.resolve().then(function() {\n            return getBinary(wasmBinaryFile);\n          });\n        }\n        function createWasm() {\n          var info = { \"env\": asmLibraryArg, \"wasi_snapshot_preview1\": asmLibraryArg };\n          function receiveInstance(instance, module2) {\n            var exports3 = instance.exports;\n            Module[\"asm\"] = exports3;\n            wasmMemory = Module[\"asm\"][\"memory\"];\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n            wasmTable = Module[\"asm\"][\"__indirect_function_table\"];\n            addOnInit(Module[\"asm\"][\"__wasm_call_ctors\"]);\n            removeRunDependency(\"wasm-instantiate\");\n          }\n          addRunDependency(\"wasm-instantiate\");\n          function receiveInstantiationResult(result) {\n            receiveInstance(result[\"instance\"]);\n          }\n          function instantiateArrayBuffer(receiver) {\n            return getBinaryPromise().then(function(binary) {\n              return WebAssembly.instantiate(binary, info);\n            }).then(function(instance) {\n              return instance;\n            }).then(receiver, function(reason) {\n              err(\"failed to asynchronously prepare wasm: \" + reason);\n              abort(reason);\n            });\n          }\n          function instantiateAsync() {\n            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == \"function\" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == \"function\") {\n              return fetch(wasmBinaryFile, { credentials: \"same-origin\" }).then(function(response) {\n                var result = WebAssembly.instantiateStreaming(response, info);\n                return result.then(receiveInstantiationResult, function(reason) {\n                  err(\"wasm streaming compile failed: \" + reason);\n                  err(\"falling back to ArrayBuffer instantiation\");\n                  return instantiateArrayBuffer(receiveInstantiationResult);\n                });\n              });\n            } else {\n              return instantiateArrayBuffer(receiveInstantiationResult);\n            }\n          }\n          if (Module[\"instantiateWasm\"]) {\n            try {\n              var exports2 = Module[\"instantiateWasm\"](info, receiveInstance);\n              return exports2;\n            } catch (e) {\n              err(\"Module.instantiateWasm callback failed with error: \" + e);\n              readyPromiseReject(e);\n            }\n          }\n          instantiateAsync().catch(readyPromiseReject);\n          return {};\n        }\n        var tempDouble;\n        var tempI64;\n        function ExitStatus(status) {\n          this.name = \"ExitStatus\";\n          this.message = \"Program terminated with exit(\" + status + \")\";\n          this.status = status;\n        }\n        function callRuntimeCallbacks(callbacks2) {\n          while (callbacks2.length > 0) {\n            callbacks2.shift()(Module);\n          }\n        }\n        function demangle(func2) {\n          return func2;\n        }\n        function demangleAll(text) {\n          var regex = /\\b_Z[\\w\\d_]+/g;\n          return text.replace(regex, function(x) {\n            var y = demangle(x);\n            return x === y ? x : y + \" [\" + x + \"]\";\n          });\n        }\n        function jsStackTrace() {\n          var error = new Error();\n          if (!error.stack) {\n            try {\n              throw new Error();\n            } catch (e) {\n              error = e;\n            }\n            if (!error.stack) {\n              return \"(no stack trace available)\";\n            }\n          }\n          return error.stack.toString();\n        }\n        function writeArrayToMemory(array2, buffer3) {\n          HEAP8.set(array2, buffer3);\n        }\n        function _abort() {\n          abort(\"\");\n        }\n        function getHeapMax() {\n          return 2147483648;\n        }\n        function _emscripten_get_heap_max() {\n          return getHeapMax();\n        }\n        function _emscripten_memcpy_big(dest, src, num) {\n          HEAPU8.copyWithin(dest, src, src + num);\n        }\n        function emscripten_realloc_buffer(size) {\n          try {\n            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);\n            updateGlobalBufferAndViews(wasmMemory.buffer);\n            return 1;\n          } catch (e) {\n          }\n        }\n        function _emscripten_resize_heap(requestedSize) {\n          var oldSize = HEAPU8.length;\n          requestedSize = requestedSize >>> 0;\n          var maxHeapSize = getHeapMax();\n          if (requestedSize > maxHeapSize) {\n            return false;\n          }\n          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;\n          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {\n            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);\n            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);\n            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));\n            var replacement = emscripten_realloc_buffer(newSize);\n            if (replacement) {\n              return true;\n            }\n          }\n          return false;\n        }\n        var SYSCALLS = { varargs: void 0, get: function() {\n          SYSCALLS.varargs += 4;\n          var ret = HEAP32[SYSCALLS.varargs - 4 >> 2];\n          return ret;\n        }, getStr: function(ptr) {\n          var ret = UTF8ToString(ptr);\n          return ret;\n        } };\n        function _fd_close(fd) {\n          return 52;\n        }\n        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {\n          return 70;\n        }\n        var printCharBuffers = [null, [], []];\n        function printChar(stream, curr) {\n          var buffer3 = printCharBuffers[stream];\n          if (curr === 0 || curr === 10) {\n            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));\n            buffer3.length = 0;\n          } else {\n            buffer3.push(curr);\n          }\n        }\n        function _fd_write(fd, iov, iovcnt, pnum) {\n          var num = 0;\n          for (var i = 0; i < iovcnt; i++) {\n            var ptr = HEAPU32[iov >> 2];\n            var len = HEAPU32[iov + 4 >> 2];\n            iov += 8;\n            for (var j = 0; j < len; j++) {\n              printChar(fd, HEAPU8[ptr + j]);\n            }\n            num += len;\n          }\n          HEAPU32[pnum >> 2] = num;\n          return 0;\n        }\n        function getCFunc(ident) {\n          var func2 = Module[\"_\" + ident];\n          return func2;\n        }\n        function ccall(ident, returnType, argTypes, args, opts) {\n          var toC = { \"string\": (str) => {\n            var ret2 = 0;\n            if (str !== null && str !== void 0 && str !== 0) {\n              var len = (str.length << 2) + 1;\n              ret2 = stackAlloc(len);\n              stringToUTF8(str, ret2, len);\n            }\n            return ret2;\n          }, \"array\": (arr) => {\n            var ret2 = stackAlloc(arr.length);\n            writeArrayToMemory(arr, ret2);\n            return ret2;\n          } };\n          function convertReturnValue(ret2) {\n            if (returnType === \"string\") {\n              return UTF8ToString(ret2);\n            }\n            if (returnType === \"boolean\")\n              return Boolean(ret2);\n            return ret2;\n          }\n          var func2 = getCFunc(ident);\n          var cArgs = [];\n          var stack2 = 0;\n          if (args) {\n            for (var i = 0; i < args.length; i++) {\n              var converter = toC[argTypes[i]];\n              if (converter) {\n                if (stack2 === 0)\n                  stack2 = stackSave();\n                cArgs[i] = converter(args[i]);\n              } else {\n                cArgs[i] = args[i];\n              }\n            }\n          }\n          var ret = func2.apply(null, cArgs);\n          function onDone(ret2) {\n            if (stack2 !== 0)\n              stackRestore(stack2);\n            return convertReturnValue(ret2);\n          }\n          ret = onDone(ret);\n          return ret;\n        }\n        function cwrap(ident, returnType, argTypes, opts) {\n          argTypes = argTypes || [];\n          var numericArgs = argTypes.every((type) => type === \"number\" || type === \"boolean\");\n          var numericRet = returnType !== \"string\";\n          if (numericRet && numericArgs && !opts) {\n            return getCFunc(ident);\n          }\n          return function() {\n            return ccall(ident, returnType, argTypes, arguments, opts);\n          };\n        }\n        var asmLibraryArg = { \"abort\": _abort, \"emscripten_get_heap_max\": _emscripten_get_heap_max, \"emscripten_memcpy_big\": _emscripten_memcpy_big, \"emscripten_resize_heap\": _emscripten_resize_heap, \"fd_close\": _fd_close, \"fd_seek\": _fd_seek, \"fd_write\": _fd_write };\n        var asm = createWasm();\n        var ___wasm_call_ctors = Module[\"___wasm_call_ctors\"] = function() {\n          return (___wasm_call_ctors = Module[\"___wasm_call_ctors\"] = Module[\"asm\"][\"__wasm_call_ctors\"]).apply(null, arguments);\n        };\n        var _init = Module[\"_init\"] = function() {\n          return (_init = Module[\"_init\"] = Module[\"asm\"][\"init\"]).apply(null, arguments);\n        };\n        var _init_with_threads_count = Module[\"_init_with_threads_count\"] = function() {\n          return (_init_with_threads_count = Module[\"_init_with_threads_count\"] = Module[\"asm\"][\"init_with_threads_count\"]).apply(null, arguments);\n        };\n        var _get_threads_count = Module[\"_get_threads_count\"] = function() {\n          return (_get_threads_count = Module[\"_get_threads_count\"] = Module[\"asm\"][\"get_threads_count\"]).apply(null, arguments);\n        };\n        var _register_tensor = Module[\"_register_tensor\"] = function() {\n          return (_register_tensor = Module[\"_register_tensor\"] = Module[\"asm\"][\"register_tensor\"]).apply(null, arguments);\n        };\n        var _dispose_data = Module[\"_dispose_data\"] = function() {\n          return (_dispose_data = Module[\"_dispose_data\"] = Module[\"asm\"][\"dispose_data\"]).apply(null, arguments);\n        };\n        var _dispose = Module[\"_dispose\"] = function() {\n          return (_dispose = Module[\"_dispose\"] = Module[\"asm\"][\"dispose\"]).apply(null, arguments);\n        };\n        var _Abs = Module[\"_Abs\"] = function() {\n          return (_Abs = Module[\"_Abs\"] = Module[\"asm\"][\"Abs\"]).apply(null, arguments);\n        };\n        var _Add = Module[\"_Add\"] = function() {\n          return (_Add = Module[\"_Add\"] = Module[\"asm\"][\"Add\"]).apply(null, arguments);\n        };\n        var _AddN = Module[\"_AddN\"] = function() {\n          return (_AddN = Module[\"_AddN\"] = Module[\"asm\"][\"AddN\"]).apply(null, arguments);\n        };\n        var _All = Module[\"_All\"] = function() {\n          return (_All = Module[\"_All\"] = Module[\"asm\"][\"All\"]).apply(null, arguments);\n        };\n        var _Any = Module[\"_Any\"] = function() {\n          return (_Any = Module[\"_Any\"] = Module[\"asm\"][\"Any\"]).apply(null, arguments);\n        };\n        var _ArgMax = Module[\"_ArgMax\"] = function() {\n          return (_ArgMax = Module[\"_ArgMax\"] = Module[\"asm\"][\"ArgMax\"]).apply(null, arguments);\n        };\n        var _AvgPool = Module[\"_AvgPool\"] = function() {\n          return (_AvgPool = Module[\"_AvgPool\"] = Module[\"asm\"][\"AvgPool\"]).apply(null, arguments);\n        };\n        var _BatchMatMul = Module[\"_BatchMatMul\"] = function() {\n          return (_BatchMatMul = Module[\"_BatchMatMul\"] = Module[\"asm\"][\"BatchMatMul\"]).apply(null, arguments);\n        };\n        var _Ceil = Module[\"_Ceil\"] = function() {\n          return (_Ceil = Module[\"_Ceil\"] = Module[\"asm\"][\"Ceil\"]).apply(null, arguments);\n        };\n        var _ClipByValue = Module[\"_ClipByValue\"] = function() {\n          return (_ClipByValue = Module[\"_ClipByValue\"] = Module[\"asm\"][\"ClipByValue\"]).apply(null, arguments);\n        };\n        var _Conv2D = Module[\"_Conv2D\"] = function() {\n          return (_Conv2D = Module[\"_Conv2D\"] = Module[\"asm\"][\"Conv2D\"]).apply(null, arguments);\n        };\n        var _Conv2DBackpropInput = Module[\"_Conv2DBackpropInput\"] = function() {\n          return (_Conv2DBackpropInput = Module[\"_Conv2DBackpropInput\"] = Module[\"asm\"][\"Conv2DBackpropInput\"]).apply(null, arguments);\n        };\n        var _Cos = Module[\"_Cos\"] = function() {\n          return (_Cos = Module[\"_Cos\"] = Module[\"asm\"][\"Cos\"]).apply(null, arguments);\n        };\n        var _Cosh = Module[\"_Cosh\"] = function() {\n          return (_Cosh = Module[\"_Cosh\"] = Module[\"asm\"][\"Cosh\"]).apply(null, arguments);\n        };\n        var _CropAndResize = Module[\"_CropAndResize\"] = function() {\n          return (_CropAndResize = Module[\"_CropAndResize\"] = Module[\"asm\"][\"CropAndResize\"]).apply(null, arguments);\n        };\n        var _Cumprod = Module[\"_Cumprod\"] = function() {\n          return (_Cumprod = Module[\"_Cumprod\"] = Module[\"asm\"][\"Cumprod\"]).apply(null, arguments);\n        };\n        var _Cumsum = Module[\"_Cumsum\"] = function() {\n          return (_Cumsum = Module[\"_Cumsum\"] = Module[\"asm\"][\"Cumsum\"]).apply(null, arguments);\n        };\n        var _DepthToSpace = Module[\"_DepthToSpace\"] = function() {\n          return (_DepthToSpace = Module[\"_DepthToSpace\"] = Module[\"asm\"][\"DepthToSpace\"]).apply(null, arguments);\n        };\n        var _DepthwiseConv2dNative = Module[\"_DepthwiseConv2dNative\"] = function() {\n          return (_DepthwiseConv2dNative = Module[\"_DepthwiseConv2dNative\"] = Module[\"asm\"][\"DepthwiseConv2dNative\"]).apply(null, arguments);\n        };\n        var _Elu = Module[\"_Elu\"] = function() {\n          return (_Elu = Module[\"_Elu\"] = Module[\"asm\"][\"Elu\"]).apply(null, arguments);\n        };\n        var _Equal = Module[\"_Equal\"] = function() {\n          return (_Equal = Module[\"_Equal\"] = Module[\"asm\"][\"Equal\"]).apply(null, arguments);\n        };\n        var _Exp = Module[\"_Exp\"] = function() {\n          return (_Exp = Module[\"_Exp\"] = Module[\"asm\"][\"Exp\"]).apply(null, arguments);\n        };\n        var _FlipLeftRight = Module[\"_FlipLeftRight\"] = function() {\n          return (_FlipLeftRight = Module[\"_FlipLeftRight\"] = Module[\"asm\"][\"FlipLeftRight\"]).apply(null, arguments);\n        };\n        var _Floor = Module[\"_Floor\"] = function() {\n          return (_Floor = Module[\"_Floor\"] = Module[\"asm\"][\"Floor\"]).apply(null, arguments);\n        };\n        var _FloorDiv = Module[\"_FloorDiv\"] = function() {\n          return (_FloorDiv = Module[\"_FloorDiv\"] = Module[\"asm\"][\"FloorDiv\"]).apply(null, arguments);\n        };\n        var _FusedBatchNorm = Module[\"_FusedBatchNorm\"] = function() {\n          return (_FusedBatchNorm = Module[\"_FusedBatchNorm\"] = Module[\"asm\"][\"FusedBatchNorm\"]).apply(null, arguments);\n        };\n        var _FusedConv2D = Module[\"_FusedConv2D\"] = function() {\n          return (_FusedConv2D = Module[\"_FusedConv2D\"] = Module[\"asm\"][\"FusedConv2D\"]).apply(null, arguments);\n        };\n        var _FusedDepthwiseConv2D = Module[\"_FusedDepthwiseConv2D\"] = function() {\n          return (_FusedDepthwiseConv2D = Module[\"_FusedDepthwiseConv2D\"] = Module[\"asm\"][\"FusedDepthwiseConv2D\"]).apply(null, arguments);\n        };\n        var _Gather = Module[\"_Gather\"] = function() {\n          return (_Gather = Module[\"_Gather\"] = Module[\"asm\"][\"Gather\"]).apply(null, arguments);\n        };\n        var _GatherNd = Module[\"_GatherNd\"] = function() {\n          return (_GatherNd = Module[\"_GatherNd\"] = Module[\"asm\"][\"GatherNd\"]).apply(null, arguments);\n        };\n        var _Greater = Module[\"_Greater\"] = function() {\n          return (_Greater = Module[\"_Greater\"] = Module[\"asm\"][\"Greater\"]).apply(null, arguments);\n        };\n        var _GreaterEqual = Module[\"_GreaterEqual\"] = function() {\n          return (_GreaterEqual = Module[\"_GreaterEqual\"] = Module[\"asm\"][\"GreaterEqual\"]).apply(null, arguments);\n        };\n        var _LeakyRelu = Module[\"_LeakyRelu\"] = function() {\n          return (_LeakyRelu = Module[\"_LeakyRelu\"] = Module[\"asm\"][\"LeakyRelu\"]).apply(null, arguments);\n        };\n        var _Less = Module[\"_Less\"] = function() {\n          return (_Less = Module[\"_Less\"] = Module[\"asm\"][\"Less\"]).apply(null, arguments);\n        };\n        var _LessEqual = Module[\"_LessEqual\"] = function() {\n          return (_LessEqual = Module[\"_LessEqual\"] = Module[\"asm\"][\"LessEqual\"]).apply(null, arguments);\n        };\n        var _Log = Module[\"_Log\"] = function() {\n          return (_Log = Module[\"_Log\"] = Module[\"asm\"][\"Log\"]).apply(null, arguments);\n        };\n        var _LogicalAnd = Module[\"_LogicalAnd\"] = function() {\n          return (_LogicalAnd = Module[\"_LogicalAnd\"] = Module[\"asm\"][\"LogicalAnd\"]).apply(null, arguments);\n        };\n        var _LogicalNot = Module[\"_LogicalNot\"] = function() {\n          return (_LogicalNot = Module[\"_LogicalNot\"] = Module[\"asm\"][\"LogicalNot\"]).apply(null, arguments);\n        };\n        var _LogicalOr = Module[\"_LogicalOr\"] = function() {\n          return (_LogicalOr = Module[\"_LogicalOr\"] = Module[\"asm\"][\"LogicalOr\"]).apply(null, arguments);\n        };\n        var _LogicalXor = Module[\"_LogicalXor\"] = function() {\n          return (_LogicalXor = Module[\"_LogicalXor\"] = Module[\"asm\"][\"LogicalXor\"]).apply(null, arguments);\n        };\n        var _Max = Module[\"_Max\"] = function() {\n          return (_Max = Module[\"_Max\"] = Module[\"asm\"][\"Max\"]).apply(null, arguments);\n        };\n        var _MaxPool = Module[\"_MaxPool\"] = function() {\n          return (_MaxPool = Module[\"_MaxPool\"] = Module[\"asm\"][\"MaxPool\"]).apply(null, arguments);\n        };\n        var _Maximum = Module[\"_Maximum\"] = function() {\n          return (_Maximum = Module[\"_Maximum\"] = Module[\"asm\"][\"Maximum\"]).apply(null, arguments);\n        };\n        var _Mean = Module[\"_Mean\"] = function() {\n          return (_Mean = Module[\"_Mean\"] = Module[\"asm\"][\"Mean\"]).apply(null, arguments);\n        };\n        var _Min = Module[\"_Min\"] = function() {\n          return (_Min = Module[\"_Min\"] = Module[\"asm\"][\"Min\"]).apply(null, arguments);\n        };\n        var _Minimum = Module[\"_Minimum\"] = function() {\n          return (_Minimum = Module[\"_Minimum\"] = Module[\"asm\"][\"Minimum\"]).apply(null, arguments);\n        };\n        var _MirrorPad = Module[\"_MirrorPad\"] = function() {\n          return (_MirrorPad = Module[\"_MirrorPad\"] = Module[\"asm\"][\"MirrorPad\"]).apply(null, arguments);\n        };\n        var _Multiply = Module[\"_Multiply\"] = function() {\n          return (_Multiply = Module[\"_Multiply\"] = Module[\"asm\"][\"Multiply\"]).apply(null, arguments);\n        };\n        var _Neg = Module[\"_Neg\"] = function() {\n          return (_Neg = Module[\"_Neg\"] = Module[\"asm\"][\"Neg\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV3 = Module[\"_NonMaxSuppressionV3\"] = function() {\n          return (_NonMaxSuppressionV3 = Module[\"_NonMaxSuppressionV3\"] = Module[\"asm\"][\"NonMaxSuppressionV3\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV4 = Module[\"_NonMaxSuppressionV4\"] = function() {\n          return (_NonMaxSuppressionV4 = Module[\"_NonMaxSuppressionV4\"] = Module[\"asm\"][\"NonMaxSuppressionV4\"]).apply(null, arguments);\n        };\n        var _NonMaxSuppressionV5 = Module[\"_NonMaxSuppressionV5\"] = function() {\n          return (_NonMaxSuppressionV5 = Module[\"_NonMaxSuppressionV5\"] = Module[\"asm\"][\"NonMaxSuppressionV5\"]).apply(null, arguments);\n        };\n        var _NotEqual = Module[\"_NotEqual\"] = function() {\n          return (_NotEqual = Module[\"_NotEqual\"] = Module[\"asm\"][\"NotEqual\"]).apply(null, arguments);\n        };\n        var _OneHot = Module[\"_OneHot\"] = function() {\n          return (_OneHot = Module[\"_OneHot\"] = Module[\"asm\"][\"OneHot\"]).apply(null, arguments);\n        };\n        var _PadV2 = Module[\"_PadV2\"] = function() {\n          return (_PadV2 = Module[\"_PadV2\"] = Module[\"asm\"][\"PadV2\"]).apply(null, arguments);\n        };\n        var _Pow = Module[\"_Pow\"] = function() {\n          return (_Pow = Module[\"_Pow\"] = Module[\"asm\"][\"Pow\"]).apply(null, arguments);\n        };\n        var _Prelu = Module[\"_Prelu\"] = function() {\n          return (_Prelu = Module[\"_Prelu\"] = Module[\"asm\"][\"Prelu\"]).apply(null, arguments);\n        };\n        var _Prod = Module[\"_Prod\"] = function() {\n          return (_Prod = Module[\"_Prod\"] = Module[\"asm\"][\"Prod\"]).apply(null, arguments);\n        };\n        var _RealDiv = Module[\"_RealDiv\"] = function() {\n          return (_RealDiv = Module[\"_RealDiv\"] = Module[\"asm\"][\"RealDiv\"]).apply(null, arguments);\n        };\n        var _Relu = Module[\"_Relu\"] = function() {\n          return (_Relu = Module[\"_Relu\"] = Module[\"asm\"][\"Relu\"]).apply(null, arguments);\n        };\n        var _Relu6 = Module[\"_Relu6\"] = function() {\n          return (_Relu6 = Module[\"_Relu6\"] = Module[\"asm\"][\"Relu6\"]).apply(null, arguments);\n        };\n        var _ResizeBilinear = Module[\"_ResizeBilinear\"] = function() {\n          return (_ResizeBilinear = Module[\"_ResizeBilinear\"] = Module[\"asm\"][\"ResizeBilinear\"]).apply(null, arguments);\n        };\n        var _ResizeNearestNeighbor = Module[\"_ResizeNearestNeighbor\"] = function() {\n          return (_ResizeNearestNeighbor = Module[\"_ResizeNearestNeighbor\"] = Module[\"asm\"][\"ResizeNearestNeighbor\"]).apply(null, arguments);\n        };\n        var _Reverse = Module[\"_Reverse\"] = function() {\n          return (_Reverse = Module[\"_Reverse\"] = Module[\"asm\"][\"Reverse\"]).apply(null, arguments);\n        };\n        var _RotateWithOffset = Module[\"_RotateWithOffset\"] = function() {\n          return (_RotateWithOffset = Module[\"_RotateWithOffset\"] = Module[\"asm\"][\"RotateWithOffset\"]).apply(null, arguments);\n        };\n        var _Round = Module[\"_Round\"] = function() {\n          return (_Round = Module[\"_Round\"] = Module[\"asm\"][\"Round\"]).apply(null, arguments);\n        };\n        var _Rsqrt = Module[\"_Rsqrt\"] = function() {\n          return (_Rsqrt = Module[\"_Rsqrt\"] = Module[\"asm\"][\"Rsqrt\"]).apply(null, arguments);\n        };\n        var _ScatterNd = Module[\"_ScatterNd\"] = function() {\n          return (_ScatterNd = Module[\"_ScatterNd\"] = Module[\"asm\"][\"ScatterNd\"]).apply(null, arguments);\n        };\n        var _SelectV2 = Module[\"_SelectV2\"] = function() {\n          return (_SelectV2 = Module[\"_SelectV2\"] = Module[\"asm\"][\"SelectV2\"]).apply(null, arguments);\n        };\n        var _Sigmoid = Module[\"_Sigmoid\"] = function() {\n          return (_Sigmoid = Module[\"_Sigmoid\"] = Module[\"asm\"][\"Sigmoid\"]).apply(null, arguments);\n        };\n        var _Sin = Module[\"_Sin\"] = function() {\n          return (_Sin = Module[\"_Sin\"] = Module[\"asm\"][\"Sin\"]).apply(null, arguments);\n        };\n        var _Softmax = Module[\"_Softmax\"] = function() {\n          return (_Softmax = Module[\"_Softmax\"] = Module[\"asm\"][\"Softmax\"]).apply(null, arguments);\n        };\n        var _SparseFillEmptyRows = Module[\"_SparseFillEmptyRows\"] = function() {\n          return (_SparseFillEmptyRows = Module[\"_SparseFillEmptyRows\"] = Module[\"asm\"][\"SparseFillEmptyRows\"]).apply(null, arguments);\n        };\n        var _SparseReshape = Module[\"_SparseReshape\"] = function() {\n          return (_SparseReshape = Module[\"_SparseReshape\"] = Module[\"asm\"][\"SparseReshape\"]).apply(null, arguments);\n        };\n        var _SparseSegmentReduction = Module[\"_SparseSegmentReduction\"] = function() {\n          return (_SparseSegmentReduction = Module[\"_SparseSegmentReduction\"] = Module[\"asm\"][\"SparseSegmentReduction\"]).apply(null, arguments);\n        };\n        var _Sqrt = Module[\"_Sqrt\"] = function() {\n          return (_Sqrt = Module[\"_Sqrt\"] = Module[\"asm\"][\"Sqrt\"]).apply(null, arguments);\n        };\n        var _Square = Module[\"_Square\"] = function() {\n          return (_Square = Module[\"_Square\"] = Module[\"asm\"][\"Square\"]).apply(null, arguments);\n        };\n        var _SquaredDifference = Module[\"_SquaredDifference\"] = function() {\n          return (_SquaredDifference = Module[\"_SquaredDifference\"] = Module[\"asm\"][\"SquaredDifference\"]).apply(null, arguments);\n        };\n        var _Step = Module[\"_Step\"] = function() {\n          return (_Step = Module[\"_Step\"] = Module[\"asm\"][\"Step\"]).apply(null, arguments);\n        };\n        var _StridedSlice = Module[\"_StridedSlice\"] = function() {\n          return (_StridedSlice = Module[\"_StridedSlice\"] = Module[\"asm\"][\"StridedSlice\"]).apply(null, arguments);\n        };\n        var _Sub = Module[\"_Sub\"] = function() {\n          return (_Sub = Module[\"_Sub\"] = Module[\"asm\"][\"Sub\"]).apply(null, arguments);\n        };\n        var _Sum = Module[\"_Sum\"] = function() {\n          return (_Sum = Module[\"_Sum\"] = Module[\"asm\"][\"Sum\"]).apply(null, arguments);\n        };\n        var _Tan = Module[\"_Tan\"] = function() {\n          return (_Tan = Module[\"_Tan\"] = Module[\"asm\"][\"Tan\"]).apply(null, arguments);\n        };\n        var _Tanh = Module[\"_Tanh\"] = function() {\n          return (_Tanh = Module[\"_Tanh\"] = Module[\"asm\"][\"Tanh\"]).apply(null, arguments);\n        };\n        var _Tile = Module[\"_Tile\"] = function() {\n          return (_Tile = Module[\"_Tile\"] = Module[\"asm\"][\"Tile\"]).apply(null, arguments);\n        };\n        var _TopK = Module[\"_TopK\"] = function() {\n          return (_TopK = Module[\"_TopK\"] = Module[\"asm\"][\"TopK\"]).apply(null, arguments);\n        };\n        var _Transform = Module[\"_Transform\"] = function() {\n          return (_Transform = Module[\"_Transform\"] = Module[\"asm\"][\"Transform\"]).apply(null, arguments);\n        };\n        var _Transpose = Module[\"_Transpose\"] = function() {\n          return (_Transpose = Module[\"_Transpose\"] = Module[\"asm\"][\"Transpose\"]).apply(null, arguments);\n        };\n        var __FusedMatMul = Module[\"__FusedMatMul\"] = function() {\n          return (__FusedMatMul = Module[\"__FusedMatMul\"] = Module[\"asm\"][\"_FusedMatMul\"]).apply(null, arguments);\n        };\n        var _malloc = Module[\"_malloc\"] = function() {\n          return (_malloc = Module[\"_malloc\"] = Module[\"asm\"][\"malloc\"]).apply(null, arguments);\n        };\n        var _free = Module[\"_free\"] = function() {\n          return (_free = Module[\"_free\"] = Module[\"asm\"][\"free\"]).apply(null, arguments);\n        };\n        var ___errno_location = Module[\"___errno_location\"] = function() {\n          return (___errno_location = Module[\"___errno_location\"] = Module[\"asm\"][\"__errno_location\"]).apply(null, arguments);\n        };\n        var stackSave = Module[\"stackSave\"] = function() {\n          return (stackSave = Module[\"stackSave\"] = Module[\"asm\"][\"stackSave\"]).apply(null, arguments);\n        };\n        var stackRestore = Module[\"stackRestore\"] = function() {\n          return (stackRestore = Module[\"stackRestore\"] = Module[\"asm\"][\"stackRestore\"]).apply(null, arguments);\n        };\n        var stackAlloc = Module[\"stackAlloc\"] = function() {\n          return (stackAlloc = Module[\"stackAlloc\"] = Module[\"asm\"][\"stackAlloc\"]).apply(null, arguments);\n        };\n        var dynCall_iijjiiii = Module[\"dynCall_iijjiiii\"] = function() {\n          return (dynCall_iijjiiii = Module[\"dynCall_iijjiiii\"] = Module[\"asm\"][\"dynCall_iijjiiii\"]).apply(null, arguments);\n        };\n        var dynCall_jiji = Module[\"dynCall_jiji\"] = function() {\n          return (dynCall_jiji = Module[\"dynCall_jiji\"] = Module[\"asm\"][\"dynCall_jiji\"]).apply(null, arguments);\n        };\n        Module[\"cwrap\"] = cwrap;\n        var calledRun;\n        dependenciesFulfilled = function runCaller() {\n          if (!calledRun)\n            run();\n          if (!calledRun)\n            dependenciesFulfilled = runCaller;\n        };\n        function run(args) {\n          args = args || arguments_;\n          if (runDependencies > 0) {\n            return;\n          }\n          preRun();\n          if (runDependencies > 0) {\n            return;\n          }\n          function doRun() {\n            if (calledRun)\n              return;\n            calledRun = true;\n            Module[\"calledRun\"] = true;\n            if (ABORT)\n              return;\n            initRuntime();\n            readyPromiseResolve(Module);\n            if (Module[\"onRuntimeInitialized\"])\n              Module[\"onRuntimeInitialized\"]();\n            postRun();\n          }\n          if (Module[\"setStatus\"]) {\n            Module[\"setStatus\"](\"Running...\");\n            setTimeout(function() {\n              setTimeout(function() {\n                Module[\"setStatus\"](\"\");\n              }, 1);\n              doRun();\n            }, 1);\n          } else {\n            doRun();\n          }\n        }\n        if (Module[\"preInit\"]) {\n          if (typeof Module[\"preInit\"] == \"function\")\n            Module[\"preInit\"] = [Module[\"preInit\"]];\n          while (Module[\"preInit\"].length > 0) {\n            Module[\"preInit\"].pop()();\n          }\n        }\n        run();\n        var listenersAdded;\n        if (beforeListeners) {\n          listenersAdded = { uncaughtException: process.listeners(\"uncaughtException\").filter(function(listener) {\n            return !beforeListeners.uncaughtException.indexOf(listener) > -1;\n          }), unhandledRejection: process.listeners(\"unhandledRejection\").filter(function(listener) {\n            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;\n          }) };\n        }\n        var actualModule;\n        if (typeof WasmBackendModule3 !== \"undefined\") {\n          actualModule = WasmBackendModule3;\n        } else if (typeof WasmBackendModuleThreadedSimd !== \"undefined\") {\n          actualModule = WasmBackendModuleThreadedSimd;\n        } else {\n          throw new Error(\"Could not find wasm module in post.js\");\n        }\n        if (listenersAdded) {\n          var tmpDispose = actualModule[\"_dispose\"];\n          actualModule[\"_dispose\"] = function() {\n            tmpDispose();\n            listenersAdded.uncaughtException.forEach(function(listener) {\n              process.removeListener(\"uncaughtException\", listener);\n            });\n            listenersAdded.unhandledRejection.forEach(function(listener) {\n              process.removeListener(\"unhandledRejection\", listener);\n            });\n          };\n        }\n        return WasmBackendModule3.ready;\n      };\n    })();\n    if (typeof exports === \"object\" && typeof module === \"object\")\n      module.exports = WasmBackendModule2;\n    else if (typeof define === \"function\" && define[\"amd\"])\n      define([], function() {\n        return WasmBackendModule2;\n      });\n    else if (typeof exports === \"object\")\n      exports[\"WasmBackendModule\"] = WasmBackendModule2;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/backend.js\nvar EPSILON_FLOAT32 = 1e-7;\nvar EPSILON_FLOAT16 = 1e-4;\nvar DataStorage = class {\n  constructor(backend2, dataMover) {\n    this.backend = backend2;\n    this.dataMover = dataMover;\n    this.data = /* @__PURE__ */ new WeakMap();\n    this.dataIdsCount = 0;\n  }\n  get(dataId) {\n    if (!this.data.has(dataId)) {\n      this.dataMover.moveData(this.backend, dataId);\n    }\n    return this.data.get(dataId);\n  }\n  set(dataId, value) {\n    this.dataIdsCount++;\n    this.data.set(dataId, value);\n  }\n  has(dataId) {\n    return this.data.has(dataId);\n  }\n  delete(dataId) {\n    this.dataIdsCount--;\n    return this.data.delete(dataId);\n  }\n  numDataIds() {\n    return this.dataIdsCount;\n  }\n};\nvar KernelBackend = class {\n  refCount(dataId) {\n    return notYetImplemented(\"refCount\");\n  }\n  incRef(dataId) {\n    return notYetImplemented(\"incRef\");\n  }\n  timerAvailable() {\n    return true;\n  }\n  time(f) {\n    return notYetImplemented(\"time\");\n  }\n  read(dataId) {\n    return notYetImplemented(\"read\");\n  }\n  readSync(dataId) {\n    return notYetImplemented(\"readSync\");\n  }\n  readToGPU(dataId, options) {\n    return notYetImplemented(\"readToGPU\");\n  }\n  numDataIds() {\n    return notYetImplemented(\"numDataIds\");\n  }\n  disposeData(dataId, force) {\n    return notYetImplemented(\"disposeData\");\n  }\n  write(values, shape, dtype) {\n    return notYetImplemented(\"write\");\n  }\n  move(dataId, values, shape, dtype, refCount) {\n    return notYetImplemented(\"move\");\n  }\n  createTensorFromTexture(values, shape, dtype) {\n    return notYetImplemented(\"createTensorFromTexture\");\n  }\n  memory() {\n    return notYetImplemented(\"memory\");\n  }\n  floatPrecision() {\n    return notYetImplemented(\"floatPrecision\");\n  }\n  epsilon() {\n    return this.floatPrecision() === 32 ? EPSILON_FLOAT32 : EPSILON_FLOAT16;\n  }\n  dispose() {\n    return notYetImplemented(\"dispose\");\n  }\n};\nfunction notYetImplemented(kernelName) {\n  throw new Error(`'${kernelName}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/util_base.js\nfunction shuffle(array2) {\n  let counter = array2.length;\n  let index = 0;\n  while (counter > 0) {\n    index = Math.random() * counter | 0;\n    counter--;\n    swap(array2, counter, index);\n  }\n}\nfunction shuffleCombo(array2, array22) {\n  if (array2.length !== array22.length) {\n    throw new Error(`Array sizes must match to be shuffled together First array length was ${array2.length}Second array length was ${array22.length}`);\n  }\n  let counter = array2.length;\n  let index = 0;\n  while (counter > 0) {\n    index = Math.random() * counter | 0;\n    counter--;\n    swap(array2, counter, index);\n    swap(array22, counter, index);\n  }\n}\nfunction clamp(min6, x, max6) {\n  return Math.max(min6, Math.min(x, max6));\n}\nfunction nearestLargerEven(val) {\n  return val % 2 === 0 ? val : val + 1;\n}\nfunction swap(object, left, right) {\n  const temp = object[left];\n  object[left] = object[right];\n  object[right] = temp;\n}\nfunction sum(arr) {\n  let sum6 = 0;\n  for (let i = 0; i < arr.length; i++) {\n    sum6 += arr[i];\n  }\n  return sum6;\n}\nfunction randUniform(a, b) {\n  const r = Math.random();\n  return b * r + (1 - r) * a;\n}\nfunction distSquared(a, b) {\n  let result = 0;\n  for (let i = 0; i < a.length; i++) {\n    const diff = Number(a[i]) - Number(b[i]);\n    result += diff * diff;\n  }\n  return result;\n}\nfunction assert(expr, msg) {\n  if (!expr) {\n    throw new Error(typeof msg === \"string\" ? msg : msg());\n  }\n}\nfunction assertShapesMatch(shapeA, shapeB, errorMessagePrefix = \"\") {\n  assert(arraysEqual(shapeA, shapeB), () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);\n}\nfunction assertNonNull(a) {\n  assert(a != null, () => `The input to the tensor constructor must be a non-null value.`);\n}\nfunction flatten(arr, result = [], skipTypedArray = false) {\n  if (result == null) {\n    result = [];\n  }\n  if (Array.isArray(arr) || isTypedArray(arr) && !skipTypedArray) {\n    for (let i = 0; i < arr.length; ++i) {\n      flatten(arr[i], result, skipTypedArray);\n    }\n  } else {\n    result.push(arr);\n  }\n  return result;\n}\nfunction sizeFromShape(shape) {\n  if (shape.length === 0) {\n    return 1;\n  }\n  let size = shape[0];\n  for (let i = 1; i < shape.length; i++) {\n    size *= shape[i];\n  }\n  return size;\n}\nfunction isScalarShape(shape) {\n  return shape.length === 0;\n}\nfunction arraysEqual(n1, n2) {\n  if (n1 === n2) {\n    return true;\n  }\n  if (n1 == null || n2 == null) {\n    return false;\n  }\n  if (n1.length !== n2.length) {\n    return false;\n  }\n  for (let i = 0; i < n1.length; i++) {\n    if (n1[i] !== n2[i]) {\n      return false;\n    }\n  }\n  return true;\n}\nfunction isInt(a) {\n  return a % 1 === 0;\n}\nfunction tanh(x) {\n  if (Math.tanh != null) {\n    return Math.tanh(x);\n  }\n  if (x === Infinity) {\n    return 1;\n  } else if (x === -Infinity) {\n    return -1;\n  } else {\n    const e2x = Math.exp(2 * x);\n    return (e2x - 1) / (e2x + 1);\n  }\n}\nfunction sizeToSquarishShape(size) {\n  const width = Math.ceil(Math.sqrt(size));\n  return [width, Math.ceil(size / width)];\n}\nfunction createShuffledIndices(n) {\n  const shuffledIndices = new Uint32Array(n);\n  for (let i = 0; i < n; ++i) {\n    shuffledIndices[i] = i;\n  }\n  shuffle(shuffledIndices);\n  return shuffledIndices;\n}\nfunction rightPad(a, size) {\n  if (size <= a.length) {\n    return a;\n  }\n  return a + \" \".repeat(size - a.length);\n}\nfunction repeatedTry(checkFn, delayFn = (counter) => 0, maxCounter, scheduleFn) {\n  return new Promise((resolve, reject) => {\n    let tryCount = 0;\n    const tryFn = () => {\n      if (checkFn()) {\n        resolve();\n        return;\n      }\n      tryCount++;\n      const nextBackoff = delayFn(tryCount);\n      if (maxCounter != null && tryCount >= maxCounter) {\n        reject();\n        return;\n      }\n      if (scheduleFn != null) {\n        scheduleFn(tryFn, nextBackoff);\n      } else {\n        setTimeout(tryFn, nextBackoff);\n      }\n    };\n    tryFn();\n  });\n}\nfunction inferFromImplicitShape(shape, size) {\n  let shapeProd = 1;\n  let implicitIdx = -1;\n  for (let i = 0; i < shape.length; ++i) {\n    if (shape[i] >= 0) {\n      shapeProd *= shape[i];\n    } else if (shape[i] === -1) {\n      if (implicitIdx !== -1) {\n        throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${implicitIdx} and dim ${i}`);\n      }\n      implicitIdx = i;\n    } else if (shape[i] < 0) {\n      throw Error(`Shapes can not be < 0. Found ${shape[i]} at dim ${i}`);\n    }\n  }\n  if (implicitIdx === -1) {\n    if (size > 0 && size !== shapeProd) {\n      throw Error(`Size(${size}) must match the product of shape ${shape}`);\n    }\n    return shape;\n  }\n  if (shapeProd === 0) {\n    throw Error(`Cannot infer the missing size in [${shape}] when there are 0 elements`);\n  }\n  if (size % shapeProd !== 0) {\n    throw Error(`The implicit shape can't be a fractional number. Got ${size} / ${shapeProd}`);\n  }\n  const newShape = shape.slice();\n  newShape[implicitIdx] = size / shapeProd;\n  return newShape;\n}\nfunction parseAxisParam(axis, shape) {\n  const rank = shape.length;\n  axis = axis == null ? shape.map((s, i) => i) : [].concat(axis);\n  assert(axis.every((ax) => ax >= -rank && ax < rank), () => `All values in axis param must be in range [-${rank}, ${rank}) but got axis ${axis}`);\n  assert(axis.every((ax) => isInt(ax)), () => `All values in axis param must be integers but got axis ${axis}`);\n  return axis.map((a) => a < 0 ? rank + a : a);\n}\nfunction squeezeShape(shape, axis) {\n  const newShape = [];\n  const keptDims = [];\n  const isEmptyArray = axis != null && Array.isArray(axis) && axis.length === 0;\n  const axes = axis == null || isEmptyArray ? null : parseAxisParam(axis, shape).sort();\n  let j = 0;\n  for (let i = 0; i < shape.length; ++i) {\n    if (axes != null) {\n      if (axes[j] === i && shape[i] !== 1) {\n        throw new Error(`Can't squeeze axis ${i} since its dim '${shape[i]}' is not 1`);\n      }\n      if ((axes[j] == null || axes[j] > i) && shape[i] === 1) {\n        newShape.push(shape[i]);\n        keptDims.push(i);\n      }\n      if (axes[j] <= i) {\n        j++;\n      }\n    }\n    if (shape[i] !== 1) {\n      newShape.push(shape[i]);\n      keptDims.push(i);\n    }\n  }\n  return { newShape, keptDims };\n}\nfunction getTypedArrayFromDType(dtype, size) {\n  let values = null;\n  if (dtype == null || dtype === \"float32\") {\n    values = new Float32Array(size);\n  } else if (dtype === \"int32\") {\n    values = new Int32Array(size);\n  } else if (dtype === \"bool\") {\n    values = new Uint8Array(size);\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n  return values;\n}\nfunction getArrayFromDType(dtype, size) {\n  let values = null;\n  if (dtype == null || dtype === \"float32\") {\n    values = new Float32Array(size);\n  } else if (dtype === \"int32\") {\n    values = new Int32Array(size);\n  } else if (dtype === \"bool\") {\n    values = new Uint8Array(size);\n  } else if (dtype === \"string\") {\n    values = new Array(size);\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n  return values;\n}\nfunction checkConversionForErrors(vals, dtype) {\n  for (let i = 0; i < vals.length; i++) {\n    const num = vals[i];\n    if (isNaN(num) || !isFinite(num)) {\n      throw Error(`A tensor of type ${dtype} being uploaded contains ${num}.`);\n    }\n  }\n}\nfunction isValidDtype(dtype) {\n  return dtype === \"bool\" || dtype === \"complex64\" || dtype === \"float32\" || dtype === \"int32\" || dtype === \"string\";\n}\nfunction hasEncodingLoss(oldType, newType) {\n  if (newType === \"complex64\") {\n    return false;\n  }\n  if (newType === \"float32\" && oldType !== \"complex64\") {\n    return false;\n  }\n  if (newType === \"int32\" && oldType !== \"float32\" && oldType !== \"complex64\") {\n    return false;\n  }\n  if (newType === \"bool\" && oldType === \"bool\") {\n    return false;\n  }\n  return true;\n}\nfunction isTypedArray(a) {\n  return a instanceof Float32Array || a instanceof Int32Array || a instanceof Uint8Array || a instanceof Uint8ClampedArray;\n}\nfunction bytesPerElement(dtype) {\n  if (dtype === \"float32\" || dtype === \"int32\") {\n    return 4;\n  } else if (dtype === \"complex64\") {\n    return 8;\n  } else if (dtype === \"bool\") {\n    return 1;\n  } else {\n    throw new Error(`Unknown dtype ${dtype}`);\n  }\n}\nfunction bytesFromStringArray(arr) {\n  if (arr == null) {\n    return 0;\n  }\n  let bytes = 0;\n  arr.forEach((x) => bytes += x.length);\n  return bytes;\n}\nfunction isString(value) {\n  return typeof value === \"string\" || value instanceof String;\n}\nfunction isBoolean(value) {\n  return typeof value === \"boolean\";\n}\nfunction isNumber(value) {\n  return typeof value === \"number\";\n}\nfunction inferDtype(values) {\n  if (Array.isArray(values)) {\n    return inferDtype(values[0]);\n  }\n  if (values instanceof Float32Array) {\n    return \"float32\";\n  } else if (values instanceof Int32Array || values instanceof Uint8Array || values instanceof Uint8ClampedArray) {\n    return \"int32\";\n  } else if (isNumber(values)) {\n    return \"float32\";\n  } else if (isString(values)) {\n    return \"string\";\n  } else if (isBoolean(values)) {\n    return \"bool\";\n  }\n  return \"float32\";\n}\nfunction isFunction(f) {\n  return !!(f && f.constructor && f.call && f.apply);\n}\nfunction nearestDivisor(size, start) {\n  for (let i = start; i < size; ++i) {\n    if (size % i === 0) {\n      return i;\n    }\n  }\n  return size;\n}\nfunction computeStrides(shape) {\n  const rank = shape.length;\n  if (rank < 2) {\n    return [];\n  }\n  const strides = new Array(rank - 1);\n  strides[rank - 2] = shape[rank - 1];\n  for (let i = rank - 3; i >= 0; --i) {\n    strides[i] = strides[i + 1] * shape[i + 1];\n  }\n  return strides;\n}\nfunction createNestedArray(offset, shape, a, isComplex = false) {\n  const ret = new Array();\n  if (shape.length === 1) {\n    const d = shape[0] * (isComplex ? 2 : 1);\n    for (let i = 0; i < d; i++) {\n      ret[i] = a[offset + i];\n    }\n  } else {\n    const d = shape[0];\n    const rest = shape.slice(1);\n    const len = rest.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);\n    for (let i = 0; i < d; i++) {\n      ret[i] = createNestedArray(offset + i * len, rest, a, isComplex);\n    }\n  }\n  return ret;\n}\nfunction toNestedArray(shape, a, isComplex = false) {\n  if (shape.length === 0) {\n    return a[0];\n  }\n  const size = shape.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);\n  if (size === 0) {\n    return [];\n  }\n  if (size !== a.length) {\n    throw new Error(`[${shape}] does not match the input size ${a.length}${isComplex ? \" for a complex tensor\" : \"\"}.`);\n  }\n  return createNestedArray(0, shape, a, isComplex);\n}\nfunction makeOnesTypedArray(size, dtype) {\n  const array2 = makeZerosTypedArray(size, dtype);\n  for (let i = 0; i < array2.length; i++) {\n    array2[i] = 1;\n  }\n  return array2;\n}\nfunction makeZerosTypedArray(size, dtype) {\n  if (dtype == null || dtype === \"float32\" || dtype === \"complex64\") {\n    return new Float32Array(size);\n  } else if (dtype === \"int32\") {\n    return new Int32Array(size);\n  } else if (dtype === \"bool\") {\n    return new Uint8Array(size);\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n}\nfunction makeZerosNestedTypedArray(shape, dtype) {\n  const size = shape.reduce((prev, curr) => prev * curr, 1);\n  if (dtype == null || dtype === \"float32\") {\n    return toNestedArray(shape, new Float32Array(size));\n  } else if (dtype === \"int32\") {\n    return toNestedArray(shape, new Int32Array(size));\n  } else if (dtype === \"bool\") {\n    return toNestedArray(shape, new Uint8Array(size));\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n}\nfunction assertNonNegativeIntegerDimensions(shape) {\n  shape.forEach((dimSize) => {\n    assert(Number.isInteger(dimSize) && dimSize >= 0, () => `Tensor must have a shape comprised of positive integers but got shape [${shape}].`);\n  });\n}\nfunction locToIndex(locs, rank, strides) {\n  if (rank === 0) {\n    return 0;\n  } else if (rank === 1) {\n    return locs[0];\n  }\n  let index = locs[locs.length - 1];\n  for (let i = 0; i < locs.length - 1; ++i) {\n    index += strides[i] * locs[i];\n  }\n  return index;\n}\nfunction indexToLoc(index, rank, strides) {\n  if (rank === 0) {\n    return [];\n  } else if (rank === 1) {\n    return [index];\n  }\n  const locs = new Array(rank);\n  for (let i = 0; i < locs.length - 1; ++i) {\n    locs[i] = Math.floor(index / strides[i]);\n    index -= locs[i] * strides[i];\n  }\n  locs[locs.length - 1] = index;\n  return locs;\n}\nfunction isPromise(object) {\n  return object && object.then && typeof object.then === \"function\";\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/environment.js\nvar TENSORFLOWJS_FLAGS_PREFIX = \"tfjsflags\";\nvar Environment = class {\n  constructor(global2) {\n    this.global = global2;\n    this.flags = {};\n    this.flagRegistry = {};\n    this.urlFlags = {};\n    this.getQueryParams = getQueryParams;\n    this.populateURLFlags();\n  }\n  setPlatform(platformName, platform) {\n    if (this.platform != null) {\n      if (!(env().getBool(\"IS_TEST\") || env().getBool(\"PROD\"))) {\n        console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${platformName}.`);\n      }\n    }\n    this.platformName = platformName;\n    this.platform = platform;\n  }\n  registerFlag(flagName, evaluationFn, setHook) {\n    this.flagRegistry[flagName] = { evaluationFn, setHook };\n    if (this.urlFlags[flagName] != null) {\n      const flagValue = this.urlFlags[flagName];\n      if (!(env().getBool(\"IS_TEST\") || env().getBool(\"PROD\"))) {\n        console.warn(`Setting feature override from URL ${flagName}: ${flagValue}.`);\n      }\n      this.set(flagName, flagValue);\n    }\n  }\n  async getAsync(flagName) {\n    if (flagName in this.flags) {\n      return this.flags[flagName];\n    }\n    this.flags[flagName] = await this.evaluateFlag(flagName);\n    return this.flags[flagName];\n  }\n  get(flagName) {\n    if (flagName in this.flags) {\n      return this.flags[flagName];\n    }\n    const flagValue = this.evaluateFlag(flagName);\n    if (isPromise(flagValue)) {\n      throw new Error(`Flag ${flagName} cannot be synchronously evaluated. Please use getAsync() instead.`);\n    }\n    this.flags[flagName] = flagValue;\n    return this.flags[flagName];\n  }\n  getNumber(flagName) {\n    return this.get(flagName);\n  }\n  getBool(flagName) {\n    return this.get(flagName);\n  }\n  getFlags() {\n    return this.flags;\n  }\n  get features() {\n    return this.flags;\n  }\n  set(flagName, value) {\n    if (this.flagRegistry[flagName] == null) {\n      throw new Error(`Cannot set flag ${flagName} as it has not been registered.`);\n    }\n    this.flags[flagName] = value;\n    if (this.flagRegistry[flagName].setHook != null) {\n      this.flagRegistry[flagName].setHook(value);\n    }\n  }\n  evaluateFlag(flagName) {\n    if (this.flagRegistry[flagName] == null) {\n      throw new Error(`Cannot evaluate flag '${flagName}': no evaluation function found.`);\n    }\n    return this.flagRegistry[flagName].evaluationFn();\n  }\n  setFlags(flags) {\n    this.flags = Object.assign({}, flags);\n  }\n  reset() {\n    this.flags = {};\n    this.urlFlags = {};\n    this.populateURLFlags();\n  }\n  populateURLFlags() {\n    if (typeof this.global === \"undefined\" || typeof this.global.location === \"undefined\" || typeof this.global.location.search === \"undefined\") {\n      return;\n    }\n    const urlParams = this.getQueryParams(this.global.location.search);\n    if (TENSORFLOWJS_FLAGS_PREFIX in urlParams) {\n      const keyValues = urlParams[TENSORFLOWJS_FLAGS_PREFIX].split(\",\");\n      keyValues.forEach((keyValue) => {\n        const [key, value] = keyValue.split(\":\");\n        this.urlFlags[key] = parseValue(key, value);\n      });\n    }\n  }\n};\nfunction getQueryParams(queryString) {\n  const params = {};\n  queryString.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g, (s, ...t) => {\n    decodeParam(params, t[0], t[1]);\n    return t.join(\"=\");\n  });\n  return params;\n}\nfunction decodeParam(params, name, value) {\n  params[decodeURIComponent(name)] = decodeURIComponent(value || \"\");\n}\nfunction parseValue(flagName, value) {\n  value = value.toLowerCase();\n  if (value === \"true\" || value === \"false\") {\n    return value === \"true\";\n  } else if (`${+value}` === value) {\n    return +value;\n  }\n  throw new Error(`Could not parse value flag value ${value} for flag ${flagName}.`);\n}\nfunction env() {\n  return ENV;\n}\nvar ENV = null;\nfunction setEnvironmentGlobal(environment) {\n  ENV = environment;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/global_util.js\nvar globalNameSpace;\nfunction getGlobalNamespace() {\n  if (globalNameSpace == null) {\n    let ns;\n    if (typeof window !== \"undefined\") {\n      ns = window;\n    } else if (typeof global !== \"undefined\") {\n      ns = global;\n    } else if (typeof process !== \"undefined\") {\n      ns = process;\n    } else if (typeof self !== \"undefined\") {\n      ns = self;\n    } else {\n      throw new Error(\"Could not find a global object\");\n    }\n    globalNameSpace = ns;\n  }\n  return globalNameSpace;\n}\nfunction getGlobalMap() {\n  const ns = getGlobalNamespace();\n  if (ns._tfGlobals == null) {\n    ns._tfGlobals = /* @__PURE__ */ new Map();\n  }\n  return ns._tfGlobals;\n}\nfunction getGlobal(key, init2) {\n  const globalMap = getGlobalMap();\n  if (globalMap.has(key)) {\n    return globalMap.get(key);\n  } else {\n    const singleton = init2();\n    globalMap.set(key, singleton);\n    return globalMap.get(key);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/kernel_names.js\nvar Abs = \"Abs\";\nvar Acos = \"Acos\";\nvar Acosh = \"Acosh\";\nvar Add = \"Add\";\nvar AddN = \"AddN\";\nvar All = \"All\";\nvar Any = \"Any\";\nvar ArgMax = \"ArgMax\";\nvar ArgMin = \"ArgMin\";\nvar Asin = \"Asin\";\nvar Asinh = \"Asinh\";\nvar Atan = \"Atan\";\nvar Atanh = \"Atanh\";\nvar Atan2 = \"Atan2\";\nvar AvgPool = \"AvgPool\";\nvar AvgPoolGrad = \"AvgPoolGrad\";\nvar AvgPool3D = \"AvgPool3D\";\nvar AvgPool3DGrad = \"AvgPool3DGrad\";\nvar BatchMatMul = \"BatchMatMul\";\nvar BatchToSpaceND = \"BatchToSpaceND\";\nvar Bincount = \"Bincount\";\nvar BroadcastTo = \"BroadcastTo\";\nvar BroadcastArgs = \"BroadcastArgs\";\nvar Cast = \"Cast\";\nvar Ceil = \"Ceil\";\nvar ClipByValue = \"ClipByValue\";\nvar Complex = \"Complex\";\nvar ComplexAbs = \"ComplexAbs\";\nvar Concat = \"Concat\";\nvar Conv2D = \"Conv2D\";\nvar Conv2DBackpropFilter = \"Conv2DBackpropFilter\";\nvar Conv2DBackpropInput = \"Conv2DBackpropInput\";\nvar Conv3D = \"Conv3D\";\nvar Conv3DBackpropFilterV2 = \"Conv3DBackpropFilterV2\";\nvar Conv3DBackpropInputV2 = \"Conv3DBackpropInputV2\";\nvar Cos = \"Cos\";\nvar Cosh = \"Cosh\";\nvar Cumprod = \"Cumprod\";\nvar Cumsum = \"Cumsum\";\nvar CropAndResize = \"CropAndResize\";\nvar DenseBincount = \"DenseBincount\";\nvar DepthToSpace = \"DepthToSpace\";\nvar DepthwiseConv2dNative = \"DepthwiseConv2dNative\";\nvar DepthwiseConv2dNativeBackpropFilter = \"DepthwiseConv2dNativeBackpropFilter\";\nvar DepthwiseConv2dNativeBackpropInput = \"DepthwiseConv2dNativeBackpropInput\";\nvar Diag = \"Diag\";\nvar Dilation2D = \"Dilation2D\";\nvar Dilation2DBackpropInput = \"Dilation2DBackpropInput\";\nvar Dilation2DBackpropFilter = \"Dilation2DBackpropFilter\";\nvar RealDiv = \"RealDiv\";\nvar Einsum = \"Einsum\";\nvar Elu = \"Elu\";\nvar EluGrad = \"EluGrad\";\nvar Erf = \"Erf\";\nvar Equal = \"Equal\";\nvar Exp = \"Exp\";\nvar ExpandDims = \"ExpandDims\";\nvar Expm1 = \"Expm1\";\nvar FFT = \"FFT\";\nvar Fill = \"Fill\";\nvar FlipLeftRight = \"FlipLeftRight\";\nvar Floor = \"Floor\";\nvar FloorDiv = \"FloorDiv\";\nvar FusedBatchNorm = \"FusedBatchNorm\";\nvar GatherV2 = \"GatherV2\";\nvar GatherNd = \"GatherNd\";\nvar Greater = \"Greater\";\nvar GreaterEqual = \"GreaterEqual\";\nvar Identity = \"Identity\";\nvar IFFT = \"IFFT\";\nvar Imag = \"Imag\";\nvar IsFinite = \"IsFinite\";\nvar IsInf = \"IsInf\";\nvar IsNan = \"IsNan\";\nvar LeakyRelu = \"LeakyRelu\";\nvar Less = \"Less\";\nvar LessEqual = \"LessEqual\";\nvar LinSpace = \"LinSpace\";\nvar Log = \"Log\";\nvar Log1p = \"Log1p\";\nvar LogicalAnd = \"LogicalAnd\";\nvar LogicalNot = \"LogicalNot\";\nvar LogicalOr = \"LogicalOr\";\nvar LogicalXor = \"LogicalXor\";\nvar LogSoftmax = \"LogSoftmax\";\nvar LowerBound = \"LowerBound\";\nvar LRN = \"LRN\";\nvar LRNGrad = \"LRNGrad\";\nvar Max = \"Max\";\nvar Maximum = \"Maximum\";\nvar MaxPool = \"MaxPool\";\nvar MaxPoolGrad = \"MaxPoolGrad\";\nvar MaxPool3D = \"MaxPool3D\";\nvar MaxPool3DGrad = \"MaxPool3DGrad\";\nvar MaxPoolWithArgmax = \"MaxPoolWithArgmax\";\nvar Mean = \"Mean\";\nvar Min = \"Min\";\nvar Minimum = \"Minimum\";\nvar MirrorPad = \"MirrorPad\";\nvar Mod = \"Mod\";\nvar Multinomial = \"Multinomial\";\nvar Multiply = \"Multiply\";\nvar Neg = \"Neg\";\nvar NotEqual = \"NotEqual\";\nvar NonMaxSuppressionV3 = \"NonMaxSuppressionV3\";\nvar NonMaxSuppressionV4 = \"NonMaxSuppressionV4\";\nvar NonMaxSuppressionV5 = \"NonMaxSuppressionV5\";\nvar OnesLike = \"OnesLike\";\nvar OneHot = \"OneHot\";\nvar Pack = \"Pack\";\nvar PadV2 = \"PadV2\";\nvar Pool = \"Pool\";\nvar Pow = \"Pow\";\nvar Prelu = \"Prelu\";\nvar Prod = \"Prod\";\nvar RaggedGather = \"RaggedGather\";\nvar RaggedRange = \"RaggedRange\";\nvar RaggedTensorToTensor = \"RaggedTensorToTensor\";\nvar Range = \"Range\";\nvar Real = \"Real\";\nvar Reciprocal = \"Reciprocal\";\nvar Relu = \"Relu\";\nvar Reshape = \"Reshape\";\nvar ResizeNearestNeighbor = \"ResizeNearestNeighbor\";\nvar ResizeNearestNeighborGrad = \"ResizeNearestNeighborGrad\";\nvar ResizeBilinear = \"ResizeBilinear\";\nvar ResizeBilinearGrad = \"ResizeBilinearGrad\";\nvar Relu6 = \"Relu6\";\nvar Reverse = \"Reverse\";\nvar Round = \"Round\";\nvar Rsqrt = \"Rsqrt\";\nvar ScatterNd = \"ScatterNd\";\nvar SearchSorted = \"SearchSorted\";\nvar Select = \"Select\";\nvar Selu = \"Selu\";\nvar Slice = \"Slice\";\nvar Sin = \"Sin\";\nvar Sinh = \"Sinh\";\nvar Sign = \"Sign\";\nvar Sigmoid = \"Sigmoid\";\nvar Softplus = \"Softplus\";\nvar Sqrt = \"Sqrt\";\nvar Sum = \"Sum\";\nvar SpaceToBatchND = \"SpaceToBatchND\";\nvar SplitV = \"SplitV\";\nvar Softmax = \"Softmax\";\nvar SparseFillEmptyRows = \"SparseFillEmptyRows\";\nvar SparseReshape = \"SparseReshape\";\nvar SparseSegmentMean = \"SparseSegmentMean\";\nvar SparseSegmentSum = \"SparseSegmentSum\";\nvar SparseToDense = \"SparseToDense\";\nvar SquaredDifference = \"SquaredDifference\";\nvar Square = \"Square\";\nvar StridedSlice = \"StridedSlice\";\nvar StringNGrams = \"StringNGrams\";\nvar StringSplit = \"StringSplit\";\nvar StringToHashBucketFast = \"StringToHashBucketFast\";\nvar Sub = \"Sub\";\nvar Tan = \"Tan\";\nvar Tanh = \"Tanh\";\nvar Tile = \"Tile\";\nvar TopK = \"TopK\";\nvar Transform = \"Transform\";\nvar Transpose = \"Transpose\";\nvar Unique = \"Unique\";\nvar Unpack = \"Unpack\";\nvar UnsortedSegmentSum = \"UnsortedSegmentSum\";\nvar UpperBound = \"UpperBound\";\nvar ZerosLike = \"ZerosLike\";\nvar Step = \"Step\";\nvar FromPixels = \"FromPixels\";\nvar RotateWithOffset = \"RotateWithOffset\";\nvar _FusedMatMul = \"_FusedMatMul\";\nvar FusedConv2D = \"FusedConv2D\";\nvar FusedDepthwiseConv2D = \"FusedDepthwiseConv2D\";\n\n// node_modules/@tensorflow/tfjs-core/dist/log.js\nfunction warn(...msg) {\n  if (!(env().getBool(\"IS_TEST\") || env().getBool(\"PROD\"))) {\n    console.warn(...msg);\n  }\n}\nfunction log(...msg) {\n  if (!(env().getBool(\"IS_TEST\") || env().getBool(\"PROD\"))) {\n    console.log(...msg);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/kernel_registry.js\nvar kernelRegistry = getGlobal(\"kernelRegistry\", () => /* @__PURE__ */ new Map());\nvar gradRegistry = getGlobal(\"gradRegistry\", () => /* @__PURE__ */ new Map());\nfunction getKernel(kernelName, backendName) {\n  const key = makeKey(kernelName, backendName);\n  return kernelRegistry.get(key);\n}\nfunction getGradient(kernelName) {\n  return gradRegistry.get(kernelName);\n}\nfunction getKernelsForBackend(backendName) {\n  const it = kernelRegistry.entries();\n  const result = [];\n  while (true) {\n    const { done, value } = it.next();\n    if (done) {\n      break;\n    }\n    const [key, config] = value;\n    const [backend2] = key.split(\"_\");\n    if (backend2 === backendName) {\n      result.push(config);\n    }\n  }\n  return result;\n}\nfunction registerKernel(config) {\n  const { kernelName, backendName } = config;\n  const key = makeKey(kernelName, backendName);\n  if (kernelRegistry.has(key)) {\n    warn(`The kernel '${kernelName}' for backend '${backendName}' is already registered`);\n  }\n  kernelRegistry.set(key, config);\n}\nfunction registerGradient(config) {\n  const { kernelName } = config;\n  if (gradRegistry.has(kernelName)) {\n    if (env().getBool(\"DEBUG\")) {\n      warn(`Overriding the gradient for '${kernelName}'`);\n    }\n  }\n  gradRegistry.set(kernelName, config);\n}\nfunction unregisterKernel(kernelName, backendName) {\n  const key = makeKey(kernelName, backendName);\n  if (!kernelRegistry.has(key)) {\n    throw new Error(`The kernel '${kernelName}' for backend '${backendName}' is not registered`);\n  }\n  kernelRegistry.delete(key);\n}\nfunction unregisterGradient(kernelName) {\n  if (!gradRegistry.has(kernelName)) {\n    throw new Error(`The gradient '${kernelName}' for backend is not registered`);\n  }\n  gradRegistry.delete(kernelName);\n}\nfunction copyRegisteredKernels(registeredBackendName, newBackendName) {\n  const kernels = getKernelsForBackend(registeredBackendName);\n  kernels.forEach((kernelConfig) => {\n    const newKernelConfig = Object.assign({}, kernelConfig, { backendName: newBackendName });\n    registerKernel(newKernelConfig);\n  });\n}\nfunction makeKey(kernelName, backendName) {\n  return `${backendName}_${kernelName}`;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/util.js\nvar util_exports = {};\n__export(util_exports, {\n  arraysEqual: () => arraysEqual,\n  assert: () => assert,\n  assertNonNegativeIntegerDimensions: () => assertNonNegativeIntegerDimensions,\n  assertNonNull: () => assertNonNull,\n  assertShapesMatch: () => assertShapesMatch,\n  bytesFromStringArray: () => bytesFromStringArray,\n  bytesPerElement: () => bytesPerElement,\n  checkConversionForErrors: () => checkConversionForErrors,\n  clamp: () => clamp,\n  computeStrides: () => computeStrides,\n  createScalarValue: () => createScalarValue,\n  createShuffledIndices: () => createShuffledIndices,\n  decodeString: () => decodeString,\n  distSquared: () => distSquared,\n  encodeString: () => encodeString,\n  fetch: () => fetch3,\n  fingerPrint64: () => fingerPrint64,\n  flatten: () => flatten,\n  getArrayFromDType: () => getArrayFromDType,\n  getTypedArrayFromDType: () => getTypedArrayFromDType,\n  hasEncodingLoss: () => hasEncodingLoss,\n  hexToLong: () => hexToLong,\n  indexToLoc: () => indexToLoc,\n  inferDtype: () => inferDtype,\n  inferFromImplicitShape: () => inferFromImplicitShape,\n  isBoolean: () => isBoolean,\n  isFunction: () => isFunction,\n  isInt: () => isInt,\n  isNumber: () => isNumber,\n  isPromise: () => isPromise,\n  isScalarShape: () => isScalarShape,\n  isString: () => isString,\n  isTypedArray: () => isTypedArray,\n  isValidDtype: () => isValidDtype,\n  locToIndex: () => locToIndex,\n  makeOnesTypedArray: () => makeOnesTypedArray,\n  makeZerosNestedTypedArray: () => makeZerosNestedTypedArray,\n  makeZerosTypedArray: () => makeZerosTypedArray,\n  nearestDivisor: () => nearestDivisor,\n  nearestLargerEven: () => nearestLargerEven,\n  now: () => now,\n  parseAxisParam: () => parseAxisParam,\n  randUniform: () => randUniform,\n  repeatedTry: () => repeatedTry,\n  rightPad: () => rightPad,\n  shuffle: () => shuffle,\n  shuffleCombo: () => shuffleCombo,\n  sizeFromShape: () => sizeFromShape,\n  sizeToSquarishShape: () => sizeToSquarishShape,\n  squeezeShape: () => squeezeShape,\n  sum: () => sum,\n  swap: () => swap,\n  tanh: () => tanh,\n  toNestedArray: () => toNestedArray,\n  toTypedArray: () => toTypedArray\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/hash_util.js\nvar LongExports = __toESM(require_long());\nvar Long = LongExports.default || LongExports;\nfunction hexToLong(hex) {\n  return Long.fromString(hex, true, 16);\n}\nvar k0 = hexToLong(\"c3a5c85c97cb3127\");\nvar k1 = hexToLong(\"b492b66fbe98f273\");\nvar k2 = hexToLong(\"9ae16a3b2f90404f\");\nfunction shiftMix(val) {\n  return val.xor(val.shru(47));\n}\nfunction fetch2(s, offset, numBytes) {\n  const bytes = s.slice(offset, offset + numBytes);\n  return Long.fromBytes(Array.from(bytes), true, true);\n}\nfunction fetch64(s, offset) {\n  return fetch2(s, offset, 8);\n}\nfunction fetch32(s, offset) {\n  return fetch2(s, offset, 4);\n}\nfunction rotate64(val, shift) {\n  return shift === 0 ? val : val.shru(shift).or(val.shl(64 - shift));\n}\nfunction hashLen16(u, v, mul2 = hexToLong(\"9ddfea08eb382d69\")) {\n  let a = u.xor(v).mul(mul2);\n  a = a.xor(a.shru(47));\n  let b = v.xor(a).mul(mul2);\n  b = b.xor(b.shru(47));\n  b = b.mul(mul2);\n  return b;\n}\nfunction weakHashLen32WithSeeds(w, x, y, z, a, b) {\n  a = a.add(w);\n  b = rotate64(b.add(a).add(z), 21);\n  const c = a;\n  a = a.add(x);\n  a = a.add(y);\n  b = b.add(rotate64(a, 44));\n  return [a.add(z), b.add(c)];\n}\nfunction weakHashLen32WithSeedsStr(s, offset, a, b) {\n  return weakHashLen32WithSeeds(fetch64(s, offset), fetch64(s, offset + 8), fetch64(s, offset + 16), fetch64(s, offset + 24), a, b);\n}\nfunction hashLen0to16(s, len = s.length) {\n  if (len >= 8) {\n    const mul2 = k2.add(len * 2);\n    const a = fetch64(s, 0).add(k2);\n    const b = fetch64(s, len - 8);\n    const c = rotate64(b, 37).mul(mul2).add(a);\n    const d = rotate64(a, 25).add(b).mul(mul2);\n    return hashLen16(c, d, mul2);\n  }\n  if (len >= 4) {\n    const mul2 = k2.add(len * 2);\n    const a = fetch32(s, 0);\n    return hashLen16(a.shl(3).add(len), fetch32(s, len - 4), mul2);\n  }\n  if (len > 0) {\n    const a = s[0];\n    const b = s[len >> 1];\n    const c = s[len - 1];\n    const y = a + (b << 8);\n    const z = len + (c << 2);\n    return shiftMix(k2.mul(y).xor(k0.mul(z))).mul(k2);\n  }\n  return k2;\n}\nfunction hashLen17to32(s, len = s.length) {\n  const mul2 = k2.add(len * 2);\n  const a = fetch64(s, 0).mul(k1);\n  const b = fetch64(s, 8);\n  const c = fetch64(s, len - 8).mul(mul2);\n  const d = fetch64(s, len - 16).mul(k2);\n  return hashLen16(rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d), a.add(rotate64(b.add(k2), 18)).add(c), mul2);\n}\nfunction hashLen33to64(s, len = s.length) {\n  const mul2 = k2.add(len * 2);\n  const a = fetch64(s, 0).mul(k2);\n  const b = fetch64(s, 8);\n  const c = fetch64(s, len - 8).mul(mul2);\n  const d = fetch64(s, len - 16).mul(k2);\n  const y = rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d);\n  const z = hashLen16(y, a.add(rotate64(b.add(k2), 18)).add(c), mul2);\n  const e = fetch64(s, 16).mul(mul2);\n  const f = fetch64(s, 24);\n  const g = y.add(fetch64(s, len - 32)).mul(mul2);\n  const h = z.add(fetch64(s, len - 24)).mul(mul2);\n  return hashLen16(rotate64(e.add(f), 43).add(rotate64(g, 30)).add(h), e.add(rotate64(f.add(a), 18)).add(g), mul2);\n}\nfunction fingerPrint64(s, len = s.length) {\n  const seed = Long.fromNumber(81, true);\n  if (len <= 32) {\n    if (len <= 16) {\n      return hashLen0to16(s, len);\n    } else {\n      return hashLen17to32(s, len);\n    }\n  } else if (len <= 64) {\n    return hashLen33to64(s, len);\n  }\n  let x = seed;\n  let y = seed.mul(k1).add(113);\n  let z = shiftMix(y.mul(k2).add(113)).mul(k2);\n  let v = [Long.UZERO, Long.UZERO];\n  let w = [Long.UZERO, Long.UZERO];\n  x = x.mul(k2).add(fetch64(s, 0));\n  let offset = 0;\n  const end = (len - 1 >> 6) * 64;\n  const last64 = end + (len - 1 & 63) - 63;\n  do {\n    x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(k1);\n    y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(k1);\n    x = x.xor(w[1]);\n    y = y.add(v[0]).add(fetch64(s, offset + 40));\n    z = rotate64(z.add(w[0]), 33).mul(k1);\n    v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(k1), x.add(w[0]));\n    w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));\n    [z, x] = [x, z];\n    offset += 64;\n  } while (offset !== end);\n  const mul2 = k1.add(z.and(255).shl(1));\n  offset = last64;\n  w[0] = w[0].add(len - 1 & 63);\n  v[0] = v[0].add(w[0]);\n  w[0] = w[0].add(v[0]);\n  x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(mul2);\n  y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(mul2);\n  x = x.xor(w[1].mul(9));\n  y = y.add(v[0].mul(9).add(fetch64(s, offset + 40)));\n  z = rotate64(z.add(w[0]), 33).mul(mul2);\n  v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(mul2), x.add(w[0]));\n  w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));\n  [z, x] = [x, z];\n  return hashLen16(hashLen16(v[0], w[0], mul2).add(shiftMix(y).mul(k0)).add(z), hashLen16(v[1], w[1], mul2).add(x), mul2);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/util.js\nfunction createScalarValue(value, dtype) {\n  if (dtype === \"string\") {\n    return encodeString(value);\n  }\n  return toTypedArray([value], dtype);\n}\nfunction noConversionNeeded(a, dtype) {\n  return a instanceof Float32Array && dtype === \"float32\" || a instanceof Int32Array && dtype === \"int32\" || a instanceof Uint8Array && dtype === \"bool\";\n}\nfunction toTypedArray(a, dtype) {\n  if (dtype === \"string\") {\n    throw new Error(\"Cannot convert a string[] to a TypedArray\");\n  }\n  if (Array.isArray(a)) {\n    a = flatten(a);\n  }\n  if (env().getBool(\"DEBUG\")) {\n    checkConversionForErrors(a, dtype);\n  }\n  if (noConversionNeeded(a, dtype)) {\n    return a;\n  }\n  if (dtype == null || dtype === \"float32\" || dtype === \"complex64\") {\n    return new Float32Array(a);\n  } else if (dtype === \"int32\") {\n    return new Int32Array(a);\n  } else if (dtype === \"bool\") {\n    const bool = new Uint8Array(a.length);\n    for (let i = 0; i < bool.length; ++i) {\n      if (Math.round(a[i]) !== 0) {\n        bool[i] = 1;\n      }\n    }\n    return bool;\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n}\nfunction now() {\n  return env().platform.now();\n}\nfunction fetch3(path, requestInits) {\n  return env().platform.fetch(path, requestInits);\n}\nfunction encodeString(s, encoding = \"utf-8\") {\n  encoding = encoding || \"utf-8\";\n  return env().platform.encode(s, encoding);\n}\nfunction decodeString(bytes, encoding = \"utf-8\") {\n  encoding = encoding || \"utf-8\";\n  return env().platform.decode(bytes, encoding);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/profiler.js\nvar Profiler = class {\n  constructor(backendTimer, logger) {\n    this.backendTimer = backendTimer;\n    this.logger = logger;\n    if (logger == null) {\n      this.logger = new Logger();\n    }\n  }\n  profileKernel(kernelName, inputs, f) {\n    let outputs;\n    const holdResultWrapperFn = () => {\n      outputs = f();\n    };\n    let timer;\n    const start = now();\n    if (this.backendTimer.timerAvailable()) {\n      timer = this.backendTimer.time(holdResultWrapperFn);\n    } else {\n      holdResultWrapperFn();\n      for (const output of outputs) {\n        output.dataSync();\n      }\n      timer = Promise.resolve({ kernelMs: now() - start });\n    }\n    if (env().getBool(\"CHECK_COMPUTATION_FOR_ERRORS\")) {\n      for (let i = 0; i < outputs.length; i++) {\n        const output = outputs[i];\n        output.data().then((tensorVals) => {\n          checkComputationForErrors(tensorVals, output.dtype, kernelName);\n        });\n      }\n    }\n    const kernelProfile = {\n      kernelName,\n      outputs,\n      inputs,\n      timeMs: timer.then((timing) => timing.kernelMs),\n      extraInfo: timer.then((timing) => timing.getExtraProfileInfo != null ? timing.getExtraProfileInfo() : \"\")\n    };\n    return kernelProfile;\n  }\n  logKernelProfile(kernelProfile) {\n    const { kernelName, outputs, timeMs, inputs, extraInfo } = kernelProfile;\n    outputs.forEach((result) => {\n      Promise.all([result.data(), timeMs, extraInfo]).then((valueContainer) => {\n        this.logger.logKernelProfile(kernelName, result, valueContainer[0], valueContainer[1], inputs, valueContainer[2]);\n      });\n    });\n  }\n};\nfunction checkComputationForErrors(vals, dtype, kernelName) {\n  if (dtype !== \"float32\") {\n    return false;\n  }\n  for (let i = 0; i < vals.length; i++) {\n    const num = vals[i];\n    if (isNaN(num) || !isFinite(num)) {\n      console.warn(`Found ${num} in the result of '${kernelName}'`);\n      return true;\n    }\n  }\n  return false;\n}\nvar Logger = class {\n  logKernelProfile(name, result, vals, timeMs, inputs, extraInfo) {\n    const time2 = typeof timeMs === \"number\" ? rightPad(`${timeMs}ms`, 9) : timeMs[\"error\"];\n    const paddedName = rightPad(name, 25);\n    const rank = result.rank;\n    const size = result.size;\n    const shape = rightPad(result.shape.toString(), 14);\n    let inputShapesDescription = \"\";\n    for (const name2 in inputs) {\n      const input2 = inputs[name2];\n      if (input2 != null) {\n        const inputShape = input2.shape || result.shape;\n        const inputRank = inputShape.length;\n        inputShapesDescription += `${name2}: ${inputRank}D ${inputRank > 0 ? inputShape : \"\"} `;\n      }\n    }\n    console.log(`%c${paddedName}\t%c${time2}\t%c${rank}D ${shape}\t%c${size}\t%c${inputShapesDescription}\t%c${extraInfo}`, \"font-weight:bold\", \"color:red\", \"color:blue\", \"color: orange\", \"color: green\", \"color: steelblue\");\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/tape.js\nfunction getFilteredNodesXToY(tape, xs, y) {\n  const tensorsFromX = {};\n  const nodesFromX = {};\n  for (let i = 0; i < xs.length; i++) {\n    tensorsFromX[xs[i].id] = true;\n  }\n  for (let i = 0; i < tape.length; i++) {\n    const node = tape[i];\n    const nodeInputs = node.inputs;\n    for (const inputName in nodeInputs) {\n      const input2 = nodeInputs[inputName];\n      let anyInputFromX = false;\n      for (let j = 0; j < xs.length; j++) {\n        if (tensorsFromX[input2.id]) {\n          node.outputs.forEach((output) => tensorsFromX[output.id] = true);\n          anyInputFromX = true;\n          nodesFromX[node.id] = true;\n          break;\n        }\n      }\n      if (anyInputFromX) {\n        break;\n      }\n    }\n  }\n  const tensorsLeadToY = {};\n  tensorsLeadToY[y.id] = true;\n  const nodesToY = {};\n  for (let i = tape.length - 1; i >= 0; i--) {\n    const node = tape[i];\n    const nodeInputs = node.inputs;\n    for (let j = 0; j < node.outputs.length; j++) {\n      if (tensorsLeadToY[node.outputs[j].id]) {\n        for (const inputName in nodeInputs) {\n          tensorsLeadToY[nodeInputs[inputName].id] = true;\n          nodesToY[node.id] = true;\n        }\n        break;\n      }\n    }\n  }\n  const filteredTape = [];\n  for (let i = 0; i < tape.length; i++) {\n    const node = tape[i];\n    if (nodesFromX[node.id] && nodesToY[node.id]) {\n      const prunedInputs = {};\n      for (const inputName in node.inputs) {\n        const nodeInput = node.inputs[inputName];\n        if (tensorsFromX[nodeInput.id]) {\n          prunedInputs[inputName] = nodeInput;\n        }\n      }\n      const prunedNode = Object.assign({}, node);\n      prunedNode.inputs = prunedInputs;\n      prunedNode.outputs = node.outputs;\n      filteredTape.push(prunedNode);\n    }\n  }\n  return filteredTape;\n}\nfunction backpropagateGradients(tensorAccumulatedGradientMap, filteredTape, tidy2, add5) {\n  for (let i = filteredTape.length - 1; i >= 0; i--) {\n    const node = filteredTape[i];\n    const dys = [];\n    node.outputs.forEach((o) => {\n      const gradTensor = tensorAccumulatedGradientMap[o.id];\n      if (gradTensor != null) {\n        dys.push(gradTensor);\n      } else {\n        dys.push(null);\n      }\n    });\n    if (node.gradient == null) {\n      throw new Error(`Cannot compute gradient: gradient function not found for ${node.kernelName}.`);\n    }\n    const inputGradients = node.gradient(dys);\n    for (const inputName in node.inputs) {\n      if (!(inputName in inputGradients)) {\n        throw new Error(`Cannot backprop through input ${inputName}. Available gradients found: ${Object.keys(inputGradients)}.`);\n      }\n      const dx = tidy2(() => inputGradients[inputName]());\n      if (dx.dtype !== \"float32\") {\n        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input ${inputName} must have 'float32' dtype, but has '${dx.dtype}'`);\n      }\n      const x = node.inputs[inputName];\n      if (!arraysEqual(dx.shape, x.shape)) {\n        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input '${inputName}' has shape '${dx.shape}', which does not match the shape of the input '${x.shape}'`);\n      }\n      if (tensorAccumulatedGradientMap[x.id] == null) {\n        tensorAccumulatedGradientMap[x.id] = dx;\n      } else {\n        const curGradient = tensorAccumulatedGradientMap[x.id];\n        tensorAccumulatedGradientMap[x.id] = add5(curGradient, dx);\n        curGradient.dispose();\n      }\n    }\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/tensor_format.js\nvar FORMAT_LIMIT_NUM_VALS = 20;\nvar FORMAT_NUM_FIRST_LAST_VALS = 3;\nvar FORMAT_NUM_SIG_DIGITS = 7;\nfunction tensorToString(vals, shape, dtype, verbose) {\n  const strides = computeStrides(shape);\n  const padPerCol = computeMaxSizePerColumn(vals, shape, dtype, strides);\n  const rank = shape.length;\n  const valsLines = subTensorToString(vals, shape, dtype, strides, padPerCol);\n  const lines = [\"Tensor\"];\n  if (verbose) {\n    lines.push(`  dtype: ${dtype}`);\n    lines.push(`  rank: ${rank}`);\n    lines.push(`  shape: [${shape}]`);\n    lines.push(`  values:`);\n  }\n  lines.push(valsLines.map((l) => \"    \" + l).join(\"\\n\"));\n  return lines.join(\"\\n\");\n}\nfunction computeMaxSizePerColumn(vals, shape, dtype, strides) {\n  const n = sizeFromShape(shape);\n  const numCols = strides[strides.length - 1];\n  const padPerCol = new Array(numCols).fill(0);\n  const rank = shape.length;\n  const valuesOrTuples = dtype === \"complex64\" ? createComplexTuples(vals) : vals;\n  if (rank > 1) {\n    for (let row = 0; row < n / numCols; row++) {\n      const offset = row * numCols;\n      for (let j = 0; j < numCols; j++) {\n        padPerCol[j] = Math.max(padPerCol[j], valToString(valuesOrTuples[offset + j], 0, dtype).length);\n      }\n    }\n  }\n  return padPerCol;\n}\nfunction valToString(val, pad3, dtype) {\n  let valStr;\n  if (Array.isArray(val)) {\n    valStr = `${parseFloat(val[0].toFixed(FORMAT_NUM_SIG_DIGITS))} + ${parseFloat(val[1].toFixed(FORMAT_NUM_SIG_DIGITS))}j`;\n  } else if (isString(val)) {\n    valStr = `'${val}'`;\n  } else if (dtype === \"bool\") {\n    valStr = boolNumToString(val);\n  } else {\n    valStr = parseFloat(val.toFixed(FORMAT_NUM_SIG_DIGITS)).toString();\n  }\n  return rightPad(valStr, pad3);\n}\nfunction boolNumToString(v) {\n  return v === 0 ? \"false\" : \"true\";\n}\nfunction subTensorToString(vals, shape, dtype, strides, padPerCol, isLast = true) {\n  const storagePerElement = dtype === \"complex64\" ? 2 : 1;\n  const size = shape[0];\n  const rank = shape.length;\n  if (rank === 0) {\n    if (dtype === \"complex64\") {\n      const complexTuple = createComplexTuples(vals);\n      return [valToString(complexTuple[0], 0, dtype)];\n    }\n    if (dtype === \"bool\") {\n      return [boolNumToString(vals[0])];\n    }\n    return [vals[0].toString()];\n  }\n  if (rank === 1) {\n    if (size > FORMAT_LIMIT_NUM_VALS) {\n      const firstValsSize = FORMAT_NUM_FIRST_LAST_VALS * storagePerElement;\n      let firstVals = Array.from(vals.slice(0, firstValsSize));\n      let lastVals = Array.from(vals.slice((size - FORMAT_NUM_FIRST_LAST_VALS) * storagePerElement, size * storagePerElement));\n      if (dtype === \"complex64\") {\n        firstVals = createComplexTuples(firstVals);\n        lastVals = createComplexTuples(lastVals);\n      }\n      return [\n        \"[\" + firstVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(\", \") + \", ..., \" + lastVals.map((x, i) => valToString(x, padPerCol[size - FORMAT_NUM_FIRST_LAST_VALS + i], dtype)).join(\", \") + \"]\"\n      ];\n    }\n    const displayVals = dtype === \"complex64\" ? createComplexTuples(vals) : Array.from(vals);\n    return [\n      \"[\" + displayVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(\", \") + \"]\"\n    ];\n  }\n  const subshape = shape.slice(1);\n  const substrides = strides.slice(1);\n  const stride = strides[0] * storagePerElement;\n  const lines = [];\n  if (size > FORMAT_LIMIT_NUM_VALS) {\n    for (let i = 0; i < FORMAT_NUM_FIRST_LAST_VALS; i++) {\n      const start = i * stride;\n      const end = start + stride;\n      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, false));\n    }\n    lines.push(\"...\");\n    for (let i = size - FORMAT_NUM_FIRST_LAST_VALS; i < size; i++) {\n      const start = i * stride;\n      const end = start + stride;\n      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));\n    }\n  } else {\n    for (let i = 0; i < size; i++) {\n      const start = i * stride;\n      const end = start + stride;\n      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));\n    }\n  }\n  const sep = rank === 2 ? \",\" : \"\";\n  lines[0] = \"[\" + lines[0] + sep;\n  for (let i = 1; i < lines.length - 1; i++) {\n    lines[i] = \" \" + lines[i] + sep;\n  }\n  let newLineSep = \",\\n\";\n  for (let i = 2; i < rank; i++) {\n    newLineSep += \"\\n\";\n  }\n  lines[lines.length - 1] = \" \" + lines[lines.length - 1] + \"]\" + (isLast ? \"\" : newLineSep);\n  return lines;\n}\nfunction createComplexTuples(vals) {\n  const complexTuples = [];\n  for (let i = 0; i < vals.length; i += 2) {\n    complexTuples.push([vals[i], vals[i + 1]]);\n  }\n  return complexTuples;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/tensor.js\nvar TensorBuffer = class {\n  constructor(shape, dtype, values) {\n    this.dtype = dtype;\n    this.shape = shape.slice();\n    this.size = sizeFromShape(shape);\n    if (values != null) {\n      const n = values.length;\n      assert(n === this.size, () => `Length of values '${n}' does not match the size inferred by the shape '${this.size}'.`);\n    }\n    if (dtype === \"complex64\") {\n      throw new Error(`complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).`);\n    }\n    this.values = values || getArrayFromDType(dtype, this.size);\n    this.strides = computeStrides(shape);\n  }\n  set(value, ...locs) {\n    if (locs.length === 0) {\n      locs = [0];\n    }\n    assert(locs.length === this.rank, () => `The number of provided coordinates (${locs.length}) must match the rank (${this.rank})`);\n    const index = this.locToIndex(locs);\n    this.values[index] = value;\n  }\n  get(...locs) {\n    if (locs.length === 0) {\n      locs = [0];\n    }\n    let i = 0;\n    for (const loc of locs) {\n      if (loc < 0 || loc >= this.shape[i]) {\n        const msg = `Requested out of range element at ${locs}.   Buffer shape=${this.shape}`;\n        throw new Error(msg);\n      }\n      i++;\n    }\n    let index = locs[locs.length - 1];\n    for (let i2 = 0; i2 < locs.length - 1; ++i2) {\n      index += this.strides[i2] * locs[i2];\n    }\n    return this.values[index];\n  }\n  locToIndex(locs) {\n    if (this.rank === 0) {\n      return 0;\n    } else if (this.rank === 1) {\n      return locs[0];\n    }\n    let index = locs[locs.length - 1];\n    for (let i = 0; i < locs.length - 1; ++i) {\n      index += this.strides[i] * locs[i];\n    }\n    return index;\n  }\n  indexToLoc(index) {\n    if (this.rank === 0) {\n      return [];\n    } else if (this.rank === 1) {\n      return [index];\n    }\n    const locs = new Array(this.shape.length);\n    for (let i = 0; i < locs.length - 1; ++i) {\n      locs[i] = Math.floor(index / this.strides[i]);\n      index -= locs[i] * this.strides[i];\n    }\n    locs[locs.length - 1] = index;\n    return locs;\n  }\n  get rank() {\n    return this.shape.length;\n  }\n  toTensor() {\n    return trackerFn().makeTensor(this.values, this.shape, this.dtype);\n  }\n};\nvar trackerFn = null;\nvar opHandler = null;\nvar deprecationWarningFn = null;\nfunction setTensorTracker(fn) {\n  trackerFn = fn;\n}\nfunction setOpHandler(handler) {\n  opHandler = handler;\n}\nfunction setDeprecationWarningFn(fn) {\n  deprecationWarningFn = fn;\n}\nvar Tensor = class {\n  constructor(shape, dtype, dataId, id) {\n    this.kept = false;\n    this.isDisposedInternal = false;\n    this.shape = shape.slice();\n    this.dtype = dtype || \"float32\";\n    this.size = sizeFromShape(shape);\n    this.strides = computeStrides(shape);\n    this.dataId = dataId;\n    this.id = id;\n    this.rankType = this.rank < 5 ? this.rank.toString() : \"higher\";\n  }\n  get rank() {\n    return this.shape.length;\n  }\n  async buffer() {\n    const vals = await this.data();\n    return opHandler.buffer(this.shape, this.dtype, vals);\n  }\n  bufferSync() {\n    return opHandler.buffer(this.shape, this.dtype, this.dataSync());\n  }\n  async array() {\n    const vals = await this.data();\n    return toNestedArray(this.shape, vals, this.dtype === \"complex64\");\n  }\n  arraySync() {\n    return toNestedArray(this.shape, this.dataSync(), this.dtype === \"complex64\");\n  }\n  async data() {\n    this.throwIfDisposed();\n    const data = trackerFn().read(this.dataId);\n    if (this.dtype === \"string\") {\n      const bytes = await data;\n      try {\n        return bytes.map((b) => decodeString(b));\n      } catch (_a) {\n        throw new Error(\"Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().\");\n      }\n    }\n    return data;\n  }\n  dataToGPU(options) {\n    this.throwIfDisposed();\n    return trackerFn().readToGPU(this.dataId, options);\n  }\n  dataSync() {\n    this.throwIfDisposed();\n    const data = trackerFn().readSync(this.dataId);\n    if (this.dtype === \"string\") {\n      try {\n        return data.map((b) => decodeString(b));\n      } catch (_a) {\n        throw new Error(\"Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().\");\n      }\n    }\n    return data;\n  }\n  async bytes() {\n    this.throwIfDisposed();\n    const data = await trackerFn().read(this.dataId);\n    if (this.dtype === \"string\") {\n      return data;\n    } else {\n      return new Uint8Array(data.buffer);\n    }\n  }\n  dispose() {\n    if (this.isDisposed) {\n      return;\n    }\n    trackerFn().disposeTensor(this);\n    this.isDisposedInternal = true;\n  }\n  get isDisposed() {\n    return this.isDisposedInternal;\n  }\n  throwIfDisposed() {\n    if (this.isDisposed) {\n      throw new Error(`Tensor is disposed.`);\n    }\n  }\n  print(verbose = false) {\n    return opHandler.print(this, verbose);\n  }\n  clone() {\n    this.throwIfDisposed();\n    return opHandler.clone(this);\n  }\n  toString(verbose = false) {\n    const vals = this.dataSync();\n    return tensorToString(vals, this.shape, this.dtype, verbose);\n  }\n  cast(dtype) {\n    this.throwIfDisposed();\n    return opHandler.cast(this, dtype);\n  }\n  variable(trainable = true, name, dtype) {\n    this.throwIfDisposed();\n    return trackerFn().makeVariable(this, trainable, name, dtype);\n  }\n};\nObject.defineProperty(Tensor, Symbol.hasInstance, {\n  value: (instance) => {\n    return !!instance && instance.data != null && instance.dataSync != null && instance.throwIfDisposed != null;\n  }\n});\nfunction getGlobalTensorClass() {\n  return getGlobal(\"Tensor\", () => {\n    return Tensor;\n  });\n}\ngetGlobalTensorClass();\nvar Variable = class extends Tensor {\n  constructor(initialValue, trainable, name, tensorId) {\n    super(initialValue.shape, initialValue.dtype, initialValue.dataId, tensorId);\n    this.trainable = trainable;\n    this.name = name;\n  }\n  assign(newValue) {\n    if (newValue.dtype !== this.dtype) {\n      throw new Error(`dtype of the new value (${newValue.dtype}) and previous value (${this.dtype}) must match`);\n    }\n    if (!arraysEqual(newValue.shape, this.shape)) {\n      throw new Error(`shape of the new value (${newValue.shape}) and previous value (${this.shape}) must match`);\n    }\n    trackerFn().disposeTensor(this);\n    this.dataId = newValue.dataId;\n    trackerFn().incRef(this, null);\n  }\n  dispose() {\n    trackerFn().disposeVariable(this);\n    this.isDisposedInternal = true;\n  }\n};\nObject.defineProperty(Variable, Symbol.hasInstance, {\n  value: (instance) => {\n    return instance instanceof Tensor && instance.assign != null && instance.assign instanceof Function;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/tensor_util.js\nvar tensor_util_exports = {};\n__export(tensor_util_exports, {\n  assertTypesMatch: () => assertTypesMatch,\n  getTensorsInContainer: () => getTensorsInContainer,\n  isTensorInList: () => isTensorInList,\n  makeTypesMatch: () => makeTypesMatch\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/types.js\nvar Rank;\n(function(Rank2) {\n  Rank2[\"R0\"] = \"R0\";\n  Rank2[\"R1\"] = \"R1\";\n  Rank2[\"R2\"] = \"R2\";\n  Rank2[\"R3\"] = \"R3\";\n  Rank2[\"R4\"] = \"R4\";\n  Rank2[\"R5\"] = \"R5\";\n  Rank2[\"R6\"] = \"R6\";\n})(Rank || (Rank = {}));\nvar UpcastInt32AndMap;\n(function(UpcastInt32AndMap2) {\n  UpcastInt32AndMap2[\"float32\"] = \"float32\";\n  UpcastInt32AndMap2[\"int32\"] = \"int32\";\n  UpcastInt32AndMap2[\"bool\"] = \"int32\";\n  UpcastInt32AndMap2[\"complex64\"] = \"complex64\";\n})(UpcastInt32AndMap || (UpcastInt32AndMap = {}));\nvar UpcastBoolAndMap;\n(function(UpcastBoolAndMap2) {\n  UpcastBoolAndMap2[\"float32\"] = \"float32\";\n  UpcastBoolAndMap2[\"int32\"] = \"int32\";\n  UpcastBoolAndMap2[\"bool\"] = \"bool\";\n  UpcastBoolAndMap2[\"complex64\"] = \"complex64\";\n})(UpcastBoolAndMap || (UpcastBoolAndMap = {}));\nvar UpcastFloat32AndMap;\n(function(UpcastFloat32AndMap2) {\n  UpcastFloat32AndMap2[\"float32\"] = \"float32\";\n  UpcastFloat32AndMap2[\"int32\"] = \"float32\";\n  UpcastFloat32AndMap2[\"bool\"] = \"float32\";\n  UpcastFloat32AndMap2[\"complex64\"] = \"complex64\";\n})(UpcastFloat32AndMap || (UpcastFloat32AndMap = {}));\nvar UpcastComplex64AndMap;\n(function(UpcastComplex64AndMap2) {\n  UpcastComplex64AndMap2[\"float32\"] = \"complex64\";\n  UpcastComplex64AndMap2[\"int32\"] = \"complex64\";\n  UpcastComplex64AndMap2[\"bool\"] = \"complex64\";\n  UpcastComplex64AndMap2[\"complex64\"] = \"complex64\";\n})(UpcastComplex64AndMap || (UpcastComplex64AndMap = {}));\nvar upcastTypeMap = {\n  \"float32\": UpcastFloat32AndMap,\n  \"int32\": UpcastInt32AndMap,\n  \"bool\": UpcastBoolAndMap,\n  \"complex64\": UpcastComplex64AndMap\n};\nfunction upcastType(typeA, typeB) {\n  if (typeA === \"string\" || typeB === \"string\") {\n    if (typeA === \"string\" && typeB === \"string\") {\n      return \"string\";\n    }\n    throw new Error(`Can not upcast ${typeA} with ${typeB}`);\n  }\n  return upcastTypeMap[typeA][typeB];\n}\nfunction sumOutType(type) {\n  return upcastType(type, \"int32\");\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/tensor_util.js\nfunction makeTypesMatch(a, b) {\n  if (a.dtype === b.dtype) {\n    return [a, b];\n  }\n  const dtype = upcastType(a.dtype, b.dtype);\n  return [a.cast(dtype), b.cast(dtype)];\n}\nfunction assertTypesMatch(a, b) {\n  assert(a.dtype === b.dtype, () => `The dtypes of the first(${a.dtype}) and second(${b.dtype}) input must match`);\n}\nfunction isTensorInList(tensor2, tensorList) {\n  return tensorList.some((x) => x.id === tensor2.id);\n}\nfunction getTensorsInContainer(result) {\n  const list = [];\n  const seen = /* @__PURE__ */ new Set();\n  walkTensorContainer(result, list, seen);\n  return list;\n}\nfunction walkTensorContainer(container, list, seen) {\n  if (container == null) {\n    return;\n  }\n  if (container instanceof Tensor) {\n    list.push(container);\n    return;\n  }\n  if (!isIterable(container)) {\n    return;\n  }\n  const iterable = container;\n  for (const k in iterable) {\n    const val = iterable[k];\n    if (!seen.has(val)) {\n      seen.add(val);\n      walkTensorContainer(val, list, seen);\n    }\n  }\n}\nfunction isIterable(obj) {\n  return Array.isArray(obj) || typeof obj === \"object\";\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/engine.js\nfunction isRegisteredKernelInvocation(kernelInvocation) {\n  return kernelInvocation.kernelName != null;\n}\nvar EngineState = class {\n  constructor() {\n    this.registeredVariables = {};\n    this.nextTapeNodeId = 0;\n    this.numBytes = 0;\n    this.numTensors = 0;\n    this.numStringTensors = 0;\n    this.numDataBuffers = 0;\n    this.gradientDepth = 0;\n    this.kernelDepth = 0;\n    this.scopeStack = [];\n    this.numDataMovesStack = [];\n    this.nextScopeId = 0;\n    this.tensorInfo = /* @__PURE__ */ new WeakMap();\n    this.profiling = false;\n    this.activeProfile = {\n      newBytes: 0,\n      newTensors: 0,\n      peakBytes: 0,\n      kernels: [],\n      result: null,\n      get kernelNames() {\n        return Array.from(new Set(this.kernels.map((k) => k.name)));\n      }\n    };\n  }\n  dispose() {\n    for (const variableName in this.registeredVariables) {\n      this.registeredVariables[variableName].dispose();\n    }\n  }\n};\nvar Engine = class {\n  constructor(ENV7) {\n    this.ENV = ENV7;\n    this.registry = {};\n    this.registryFactory = {};\n    this.pendingBackendInitId = 0;\n    this.state = new EngineState();\n  }\n  async ready() {\n    if (this.pendingBackendInit != null) {\n      return this.pendingBackendInit.then(() => {\n      });\n    }\n    if (this.backendInstance != null) {\n      return;\n    }\n    const sortedBackends = this.getSortedBackends();\n    for (let i = 0; i < sortedBackends.length; i++) {\n      const backendName = sortedBackends[i];\n      const success = await this.initializeBackend(backendName).success;\n      if (success) {\n        await this.setBackend(backendName);\n        return;\n      }\n    }\n    throw new Error(`Could not initialize any backends, all backend initializations failed.`);\n  }\n  get backend() {\n    if (this.pendingBackendInit != null) {\n      throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);\n    }\n    if (this.backendInstance == null) {\n      const { name, asyncInit } = this.initializeBackendsAndReturnBest();\n      if (asyncInit) {\n        throw new Error(`The highest priority backend '${name}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);\n      }\n      this.setBackend(name);\n    }\n    return this.backendInstance;\n  }\n  backendNames() {\n    return Object.keys(this.registryFactory);\n  }\n  findBackend(backendName) {\n    if (!(backendName in this.registry)) {\n      if (backendName in this.registryFactory) {\n        const { asyncInit } = this.initializeBackend(backendName);\n        if (asyncInit) {\n          return null;\n        }\n      } else {\n        return null;\n      }\n    }\n    return this.registry[backendName];\n  }\n  findBackendFactory(backendName) {\n    if (!(backendName in this.registryFactory)) {\n      return null;\n    }\n    return this.registryFactory[backendName].factory;\n  }\n  registerBackend(backendName, factory, priority = 1) {\n    if (backendName in this.registryFactory) {\n      warn(`${backendName} backend was already registered. Reusing existing backend factory.`);\n      return false;\n    }\n    this.registryFactory[backendName] = { factory, priority };\n    return true;\n  }\n  async setBackend(backendName) {\n    if (this.registryFactory[backendName] == null) {\n      throw new Error(`Backend name '${backendName}' not found in registry`);\n    }\n    this.backendName = backendName;\n    if (this.registry[backendName] == null) {\n      this.backendInstance = null;\n      const { success, asyncInit } = this.initializeBackend(backendName);\n      const result = asyncInit ? await success : success;\n      if (!result) {\n        return false;\n      }\n    }\n    this.backendInstance = this.registry[backendName];\n    this.setupRegisteredKernels();\n    this.profiler = new Profiler(this.backendInstance);\n    return true;\n  }\n  setupRegisteredKernels() {\n    const kernels = getKernelsForBackend(this.backendName);\n    kernels.forEach((kernel) => {\n      if (kernel.setupFunc != null) {\n        kernel.setupFunc(this.backendInstance);\n      }\n    });\n  }\n  disposeRegisteredKernels(backendName) {\n    const kernels = getKernelsForBackend(backendName);\n    kernels.forEach((kernel) => {\n      if (kernel.disposeFunc != null) {\n        kernel.disposeFunc(this.registry[backendName]);\n      }\n    });\n  }\n  initializeBackend(backendName) {\n    const registryFactoryEntry = this.registryFactory[backendName];\n    if (registryFactoryEntry == null) {\n      throw new Error(`Cannot initialize backend ${backendName}, no registration found.`);\n    }\n    try {\n      const backend2 = registryFactoryEntry.factory();\n      if (backend2 && !(backend2 instanceof KernelBackend) && typeof backend2.then === \"function\") {\n        const promiseId = ++this.pendingBackendInitId;\n        const success = backend2.then((backendInstance) => {\n          if (promiseId < this.pendingBackendInitId) {\n            return false;\n          }\n          this.registry[backendName] = backendInstance;\n          this.pendingBackendInit = null;\n          return true;\n        }).catch((err) => {\n          if (promiseId < this.pendingBackendInitId) {\n            return false;\n          }\n          this.pendingBackendInit = null;\n          warn(`Initialization of backend ${backendName} failed`);\n          warn(err.stack || err.message);\n          return false;\n        });\n        this.pendingBackendInit = success;\n        return { success, asyncInit: true };\n      } else {\n        this.registry[backendName] = backend2;\n        return { success: true, asyncInit: false };\n      }\n    } catch (err) {\n      warn(`Initialization of backend ${backendName} failed`);\n      warn(err.stack || err.message);\n      return { success: false, asyncInit: false };\n    }\n  }\n  removeBackend(backendName) {\n    if (!(backendName in this.registryFactory)) {\n      throw new Error(`${backendName} backend not found in registry`);\n    }\n    if (this.backendName === backendName && this.pendingBackendInit != null) {\n      this.pendingBackendInitId++;\n    }\n    if (backendName in this.registry) {\n      this.disposeRegisteredKernels(backendName);\n      this.registry[backendName].dispose();\n      delete this.registry[backendName];\n    }\n    delete this.registryFactory[backendName];\n    if (this.backendName === backendName) {\n      this.pendingBackendInit = null;\n      this.backendName = null;\n      this.backendInstance = null;\n    }\n  }\n  getSortedBackends() {\n    if (Object.keys(this.registryFactory).length === 0) {\n      throw new Error(\"No backend found in registry.\");\n    }\n    return Object.keys(this.registryFactory).sort((a, b) => {\n      return this.registryFactory[b].priority - this.registryFactory[a].priority;\n    });\n  }\n  initializeBackendsAndReturnBest() {\n    const sortedBackends = this.getSortedBackends();\n    for (let i = 0; i < sortedBackends.length; i++) {\n      const backendName = sortedBackends[i];\n      const { success, asyncInit } = this.initializeBackend(backendName);\n      if (asyncInit || success) {\n        return { name: backendName, asyncInit };\n      }\n    }\n    throw new Error(`Could not initialize any backends, all backend initializations failed.`);\n  }\n  moveData(backend2, dataId) {\n    const info = this.state.tensorInfo.get(dataId);\n    const srcBackend = info.backend;\n    const values = this.readSync(dataId);\n    const refCount = srcBackend.refCount(dataId);\n    srcBackend.disposeData(dataId, true);\n    info.backend = backend2;\n    backend2.move(dataId, values, info.shape, info.dtype, refCount);\n    if (this.shouldCheckForMemLeaks()) {\n      this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1]++;\n    }\n  }\n  tidy(nameOrFn, fn) {\n    let name = null;\n    if (fn == null) {\n      if (typeof nameOrFn !== \"function\") {\n        throw new Error(\"Please provide a function to tidy()\");\n      }\n      fn = nameOrFn;\n    } else {\n      if (typeof nameOrFn !== \"string\" && !(nameOrFn instanceof String)) {\n        throw new Error(\"When calling with two arguments, the first argument to tidy() must be a string\");\n      }\n      if (typeof fn !== \"function\") {\n        throw new Error(\"When calling with two arguments, the 2nd argument to tidy() must be a function\");\n      }\n      name = nameOrFn;\n    }\n    let result;\n    return this.scopedRun(() => this.startScope(name), () => this.endScope(result), () => {\n      result = fn();\n      if (result instanceof Promise) {\n        console.error(\"Cannot return a Promise inside of tidy.\");\n      }\n      return result;\n    });\n  }\n  scopedRun(start, end, f) {\n    start();\n    try {\n      const res = f();\n      end();\n      return res;\n    } catch (ex) {\n      end();\n      throw ex;\n    }\n  }\n  nextTensorId() {\n    return Engine.nextTensorId++;\n  }\n  nextVariableId() {\n    return Engine.nextVariableId++;\n  }\n  clone(x) {\n    const y = ENGINE.runKernel(Identity, { x });\n    const inputs = { x };\n    const grad2 = (dy) => ({\n      x: () => {\n        const dtype = \"float32\";\n        const gradInputs = { x: dy };\n        const attrs = { dtype };\n        return ENGINE.runKernel(\n          Cast,\n          gradInputs,\n          attrs\n        );\n      }\n    });\n    const saved = [];\n    this.addTapeNode(this.state.activeScope.name, inputs, [y], grad2, saved, {});\n    return y;\n  }\n  runKernel(kernelName, inputs, attrs) {\n    if (this.backendName == null) {\n      this.backend;\n    }\n    const hasKernel = getKernel(kernelName, this.backendName) != null;\n    if (!hasKernel) {\n      throw new Error(`Kernel '${kernelName}' not registered for backend '${this.backendName}'`);\n    }\n    return this.runKernelFunc({ kernelName, inputs, attrs });\n  }\n  shouldCheckForMemLeaks() {\n    return this.ENV.getBool(\"IS_TEST\");\n  }\n  checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos) {\n    const numDataIdsAfter = this.backend.numDataIds();\n    let numOutputDataIds = 0;\n    outInfos.forEach((info) => {\n      numOutputDataIds += info.dtype === \"complex64\" ? 3 : 1;\n    });\n    const numMoves = this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1];\n    const dataIdsLeaked = numDataIdsAfter - numDataIdsBefore - numOutputDataIds - numMoves;\n    if (dataIdsLeaked > 0) {\n      throw new Error(`Backend '${this.backendName}' has an internal memory leak (${dataIdsLeaked} data ids) after running '${kernelName}'`);\n    }\n  }\n  runKernelFunc(kernelParams) {\n    let outputs;\n    let saved = [];\n    const isTapeOn = this.isTapeOn();\n    const startingBytecount = this.state.numBytes;\n    const startingNumTensors = this.state.numTensors;\n    if (this.shouldCheckForMemLeaks()) {\n      this.state.numDataMovesStack.push(0);\n    }\n    let kernelFunc3;\n    if (this.backendName == null) {\n      this.backend;\n    }\n    let out;\n    const kernelOrScopeName = isRegisteredKernelInvocation(kernelParams) ? kernelParams.kernelName : this.state.activeScope != null ? this.state.activeScope.name : \"\";\n    if (isRegisteredKernelInvocation(kernelParams)) {\n      const { kernelName, inputs: inputs2, attrs: attrs2 } = kernelParams;\n      if (this.backendName == null) {\n        this.backend;\n      }\n      const kernel = getKernel(kernelName, this.backendName);\n      assert(kernel != null, () => `Cannot find registered kernel '${kernelName}' for backend '${this.backendName}'`);\n      kernelFunc3 = () => {\n        const numDataIdsBefore = this.backend.numDataIds();\n        out = kernel.kernelFunc({ inputs: inputs2, attrs: attrs2, backend: this.backend });\n        const outInfos = Array.isArray(out) ? out : [out];\n        if (this.shouldCheckForMemLeaks()) {\n          this.checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos);\n        }\n        const outTensors = outInfos.map((outInfo) => {\n          if (outInfo.rank != null) {\n            return outInfo;\n          }\n          return this.makeTensorFromTensorInfo(outInfo);\n        });\n        if (isTapeOn) {\n          const tensorsToSave = this.getTensorsForGradient(kernelName, inputs2, outTensors);\n          saved = this.saveTensorsForBackwardMode(tensorsToSave);\n        }\n        return outTensors;\n      };\n    } else {\n      const { forwardFunc } = kernelParams;\n      const saveFunc = (tensors) => {\n        if (!isTapeOn) {\n          return;\n        }\n        saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));\n      };\n      kernelFunc3 = () => {\n        const numDataIdsBefore = this.backend.numDataIds();\n        out = this.tidy(() => forwardFunc(this.backend, saveFunc));\n        const outs = Array.isArray(out) ? out : [out];\n        if (this.shouldCheckForMemLeaks()) {\n          this.checkKernelForMemLeak(kernelOrScopeName, numDataIdsBefore, outs);\n        }\n        return outs;\n      };\n    }\n    const { inputs, attrs } = kernelParams;\n    const backwardsFunc = isRegisteredKernelInvocation(kernelParams) ? null : kernelParams.backwardsFunc;\n    let kernelProfile;\n    this.scopedRun(\n      () => this.state.kernelDepth++,\n      () => this.state.kernelDepth--,\n      () => {\n        if (!this.ENV.getBool(\"DEBUG\") && !this.state.profiling) {\n          outputs = kernelFunc3();\n        } else {\n          kernelProfile = this.profiler.profileKernel(kernelOrScopeName, inputs, () => kernelFunc3());\n          if (this.ENV.getBool(\"DEBUG\")) {\n            this.profiler.logKernelProfile(kernelProfile);\n          }\n          outputs = kernelProfile.outputs;\n        }\n      }\n    );\n    if (isTapeOn) {\n      this.addTapeNode(kernelOrScopeName, inputs, outputs, backwardsFunc, saved, attrs);\n    }\n    if (this.state.profiling) {\n      this.state.activeProfile.kernels.push({\n        name: kernelOrScopeName,\n        bytesAdded: this.state.numBytes - startingBytecount,\n        totalBytesSnapshot: this.state.numBytes,\n        tensorsAdded: this.state.numTensors - startingNumTensors,\n        totalTensorsSnapshot: this.state.numTensors,\n        inputShapes: Object.keys(inputs).map((key) => inputs[key] != null ? inputs[key].shape : null),\n        outputShapes: outputs.map((item) => item.shape),\n        kernelTimeMs: kernelProfile.timeMs,\n        extraInfo: kernelProfile.extraInfo\n      });\n    }\n    return Array.isArray(out) ? outputs : outputs[0];\n  }\n  saveTensorsForBackwardMode(tensors) {\n    const saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));\n    return saved;\n  }\n  getTensorsForGradient(kernelName, inputs, outputs) {\n    const gradConfig = getGradient(kernelName);\n    if (gradConfig != null) {\n      const inputsToSave = gradConfig.inputsToSave || [];\n      const outputsToSave = gradConfig.outputsToSave || [];\n      let inputTensorsToSave;\n      if (gradConfig.saveAllInputs) {\n        assert(Array.isArray(inputs), () => \"saveAllInputs is true, expected inputs to be an array.\");\n        inputTensorsToSave = Object.keys(inputs).map((key) => inputs[key]);\n      } else {\n        inputTensorsToSave = inputsToSave.map((inputName) => inputs[inputName]);\n      }\n      const outputTensorsToSave = outputs.filter((_, i) => outputsToSave[i]);\n      return inputTensorsToSave.concat(outputTensorsToSave);\n    }\n    return [];\n  }\n  makeTensor(values, shape, dtype, backend2) {\n    if (values == null) {\n      throw new Error(\"Values passed to engine.makeTensor() are null\");\n    }\n    dtype = dtype || \"float32\";\n    backend2 = backend2 || this.backend;\n    let backendVals = values;\n    if (dtype === \"string\" && isString(values[0])) {\n      backendVals = values.map((d) => encodeString(d));\n    }\n    const dataId = backend2.write(backendVals, shape, dtype);\n    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());\n    this.trackTensor(t, backend2);\n    if (dtype === \"string\") {\n      const info = this.state.tensorInfo.get(dataId);\n      const newBytes = bytesFromStringArray(backendVals);\n      this.state.numBytes += newBytes - info.bytes;\n      info.bytes = newBytes;\n    }\n    return t;\n  }\n  makeTensorFromDataId(dataId, shape, dtype, backend2) {\n    dtype = dtype || \"float32\";\n    const tensorInfo = { dataId, shape, dtype };\n    return this.makeTensorFromTensorInfo(tensorInfo, backend2);\n  }\n  makeTensorFromTensorInfo(tensorInfo, backend2) {\n    const { dataId, shape, dtype } = tensorInfo;\n    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());\n    this.trackTensor(t, backend2);\n    return t;\n  }\n  makeVariable(initialValue, trainable = true, name, dtype) {\n    name = name || this.nextVariableId().toString();\n    if (dtype != null && dtype !== initialValue.dtype) {\n      initialValue = initialValue.cast(dtype);\n    }\n    const v = new Variable(initialValue, trainable, name, this.nextTensorId());\n    if (this.state.registeredVariables[v.name] != null) {\n      throw new Error(`Variable with name ${v.name} was already registered`);\n    }\n    this.state.registeredVariables[v.name] = v;\n    this.incRef(v, this.backend);\n    return v;\n  }\n  trackTensor(a, backend2) {\n    this.state.numTensors++;\n    if (a.dtype === \"string\") {\n      this.state.numStringTensors++;\n    }\n    let bytes = 0;\n    if (a.dtype !== \"complex64\" && a.dtype !== \"string\") {\n      bytes = a.size * bytesPerElement(a.dtype);\n    }\n    this.state.numBytes += bytes;\n    if (!this.state.tensorInfo.has(a.dataId)) {\n      this.state.numDataBuffers++;\n      this.state.tensorInfo.set(a.dataId, {\n        backend: backend2 || this.backend,\n        dtype: a.dtype,\n        shape: a.shape,\n        bytes\n      });\n    }\n    if (!(a instanceof Variable)) {\n      this.track(a);\n    }\n  }\n  incRef(a, backend2) {\n    this.trackTensor(a, backend2);\n    this.backend.incRef(a.dataId);\n  }\n  removeDataId(dataId, backend2) {\n    if (this.state.tensorInfo.has(dataId) && this.state.tensorInfo.get(dataId).backend === backend2) {\n      this.state.tensorInfo.delete(dataId);\n      this.state.numDataBuffers--;\n    }\n  }\n  disposeTensor(a) {\n    if (!this.state.tensorInfo.has(a.dataId)) {\n      return;\n    }\n    const info = this.state.tensorInfo.get(a.dataId);\n    this.state.numTensors--;\n    if (a.dtype === \"string\") {\n      this.state.numStringTensors--;\n      this.state.numBytes -= info.bytes;\n    }\n    if (a.dtype !== \"complex64\" && a.dtype !== \"string\") {\n      const bytes = a.size * bytesPerElement(a.dtype);\n      this.state.numBytes -= bytes;\n    }\n    if (info.backend.disposeData(a.dataId)) {\n      this.removeDataId(a.dataId, info.backend);\n    }\n  }\n  disposeVariables() {\n    for (const varName in this.state.registeredVariables) {\n      const v = this.state.registeredVariables[varName];\n      this.disposeVariable(v);\n    }\n  }\n  disposeVariable(v) {\n    this.disposeTensor(v);\n    if (this.state.registeredVariables[v.name] != null) {\n      delete this.state.registeredVariables[v.name];\n    }\n  }\n  memory() {\n    const info = this.backend.memory();\n    info.numTensors = this.state.numTensors;\n    info.numDataBuffers = this.state.numDataBuffers;\n    info.numBytes = this.state.numBytes;\n    if (this.state.numStringTensors > 0) {\n      info.unreliable = true;\n      if (info.reasons == null) {\n        info.reasons = [];\n      }\n      info.reasons.push(\"Memory usage by string tensors is approximate (2 bytes per character)\");\n    }\n    return info;\n  }\n  async profile(query) {\n    this.state.profiling = true;\n    const startBytes = this.state.numBytes;\n    const startNumTensors = this.state.numTensors;\n    this.state.activeProfile.kernels = [];\n    this.state.activeProfile.result = await query();\n    this.state.profiling = false;\n    this.state.activeProfile.peakBytes = Math.max(...this.state.activeProfile.kernels.map((d) => d.totalBytesSnapshot));\n    this.state.activeProfile.newBytes = this.state.numBytes - startBytes;\n    this.state.activeProfile.newTensors = this.state.numTensors - startNumTensors;\n    for (const kernel of this.state.activeProfile.kernels) {\n      kernel.kernelTimeMs = await kernel.kernelTimeMs;\n      kernel.extraInfo = await kernel.extraInfo;\n    }\n    return this.state.activeProfile;\n  }\n  isTapeOn() {\n    return this.state.gradientDepth > 0 && this.state.kernelDepth === 0;\n  }\n  addTapeNode(kernelName, inputs, outputs, gradientsFunc, saved, attrs) {\n    const tapeNode = { id: this.state.nextTapeNodeId++, kernelName, inputs, outputs, saved };\n    const gradConfig = getGradient(kernelName);\n    if (gradConfig != null) {\n      gradientsFunc = gradConfig.gradFunc;\n    }\n    if (gradientsFunc != null) {\n      tapeNode.gradient = (dys) => {\n        dys = dys.map((dy, i) => {\n          if (dy == null) {\n            const output = outputs[i];\n            const vals = makeZerosTypedArray(output.size, output.dtype);\n            return this.makeTensor(vals, output.shape, output.dtype);\n          }\n          return dy;\n        });\n        return gradientsFunc(dys.length > 1 ? dys : dys[0], saved, attrs);\n      };\n    }\n    this.state.activeTape.push(tapeNode);\n  }\n  keep(result) {\n    result.kept = true;\n    return result;\n  }\n  startTape() {\n    if (this.state.gradientDepth === 0) {\n      this.state.activeTape = [];\n    }\n    this.state.gradientDepth++;\n  }\n  endTape() {\n    this.state.gradientDepth--;\n  }\n  startScope(name) {\n    const scopeInfo = {\n      track: [],\n      name: \"unnamed scope\",\n      id: this.state.nextScopeId++\n    };\n    if (name) {\n      scopeInfo.name = name;\n    }\n    this.state.scopeStack.push(scopeInfo);\n    this.state.activeScope = scopeInfo;\n  }\n  endScope(result) {\n    const tensorsToTrackInParent = getTensorsInContainer(result);\n    const tensorsToTrackInParentSet = new Set(tensorsToTrackInParent.map((t) => t.id));\n    for (let i = 0; i < this.state.activeScope.track.length; i++) {\n      const tensor2 = this.state.activeScope.track[i];\n      if (!tensor2.kept && !tensorsToTrackInParentSet.has(tensor2.id)) {\n        tensor2.dispose();\n      }\n    }\n    const oldScope = this.state.scopeStack.pop();\n    this.state.activeScope = this.state.scopeStack.length === 0 ? null : this.state.scopeStack[this.state.scopeStack.length - 1];\n    tensorsToTrackInParent.forEach((tensor2) => {\n      if (!tensor2.kept && tensor2.scopeId === oldScope.id) {\n        this.track(tensor2);\n      }\n    });\n  }\n  gradients(f, xs, dy, allowNoGradients = false) {\n    assert(xs.length > 0, () => \"gradients() received an empty list of xs.\");\n    if (dy != null && dy.dtype !== \"float32\") {\n      throw new Error(`dy must have 'float32' dtype, but has '${dy.dtype}'`);\n    }\n    const y = this.scopedRun(() => this.startTape(), () => this.endTape(), () => this.tidy(\"forward\", f));\n    assert(y instanceof Tensor, () => \"The result y returned by f() must be a tensor.\");\n    const filteredTape = getFilteredNodesXToY(this.state.activeTape, xs, y);\n    if (!allowNoGradients && filteredTape.length === 0 && xs.length > 0) {\n      throw new Error(\"Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.\");\n    }\n    return this.tidy(\"backward\", () => {\n      const accumulatedGradientMap = {};\n      accumulatedGradientMap[y.id] = dy == null ? ones(y.shape) : dy;\n      backpropagateGradients(\n        accumulatedGradientMap,\n        filteredTape,\n        (f2) => this.tidy(f2),\n        add\n      );\n      const grads2 = xs.map((x) => accumulatedGradientMap[x.id]);\n      if (this.state.gradientDepth === 0) {\n        this.state.activeTape.forEach((node) => {\n          for (const tensor2 of node.saved) {\n            tensor2.dispose();\n          }\n        });\n        this.state.activeTape = null;\n      }\n      return { value: y, grads: grads2 };\n    });\n  }\n  customGrad(f) {\n    assert(isFunction(f), () => \"The f passed in customGrad(f) must be a function.\");\n    return (...inputs) => {\n      assert(inputs.every((t) => t instanceof Tensor), () => \"The args passed in customGrad(f)(x1, x2,...) must all be tensors\");\n      let res;\n      const inputMap = {};\n      inputs.forEach((input2, i) => {\n        inputMap[i] = input2;\n      });\n      const forwardFunc = (_, save) => {\n        res = f(...[...inputs, save]);\n        assert(res.value instanceof Tensor, () => \"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor\");\n        assert(isFunction(res.gradFunc), () => \"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function.\");\n        return res.value;\n      };\n      const backwardsFunc = (dy, saved) => {\n        const gradRes = res.gradFunc(dy, saved);\n        const grads2 = Array.isArray(gradRes) ? gradRes : [gradRes];\n        assert(grads2.length === inputs.length, () => \"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...).\");\n        assert(grads2.every((t) => t instanceof Tensor), () => \"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.\");\n        const gradMap = {};\n        grads2.forEach((grad2, i) => {\n          gradMap[i] = () => grad2;\n        });\n        return gradMap;\n      };\n      return this.runKernelFunc({\n        forwardFunc,\n        backwardsFunc,\n        inputs: inputMap\n      });\n    };\n  }\n  readSync(dataId) {\n    const info = this.state.tensorInfo.get(dataId);\n    return info.backend.readSync(dataId);\n  }\n  read(dataId) {\n    const info = this.state.tensorInfo.get(dataId);\n    return info.backend.read(dataId);\n  }\n  readToGPU(dataId, options) {\n    const info = this.state.tensorInfo.get(dataId);\n    return info.backend.readToGPU(dataId, options);\n  }\n  async time(query) {\n    const start = now();\n    const timingInfo = await this.backend.time(query);\n    timingInfo.wallMs = now() - start;\n    return timingInfo;\n  }\n  track(result) {\n    if (this.state.activeScope != null) {\n      result.scopeId = this.state.activeScope.id;\n      this.state.activeScope.track.push(result);\n    }\n    return result;\n  }\n  get registeredVariables() {\n    return this.state.registeredVariables;\n  }\n  reset() {\n    this.pendingBackendInitId++;\n    this.state.dispose();\n    this.ENV.reset();\n    this.state = new EngineState();\n    for (const backendName in this.registry) {\n      this.disposeRegisteredKernels(backendName);\n      this.registry[backendName].dispose();\n      delete this.registry[backendName];\n    }\n    this.backendName = null;\n    this.backendInstance = null;\n    this.pendingBackendInit = null;\n  }\n};\nEngine.nextTensorId = 0;\nEngine.nextVariableId = 0;\nfunction ones(shape) {\n  const values = makeOnesTypedArray(sizeFromShape(shape), \"float32\");\n  return ENGINE.makeTensor(values, shape, \"float32\");\n}\nfunction getOrMakeEngine() {\n  const ns = getGlobalNamespace();\n  if (ns._tfengine == null) {\n    const environment = new Environment(ns);\n    ns._tfengine = new Engine(environment);\n  }\n  setEnvironmentGlobal(ns._tfengine.ENV);\n  setTensorTracker(() => ns._tfengine);\n  return ns._tfengine;\n}\nvar ENGINE = getOrMakeEngine();\nfunction add(a, b) {\n  const inputs = { a, b };\n  return ENGINE.runKernel(Add, inputs);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/device_util.js\nvar device_util_exports = {};\n__export(device_util_exports, {\n  isBrowser: () => isBrowser,\n  isMobile: () => isMobile,\n  mockIsMobile: () => mockIsMobile\n});\nfunction _isNavigatorDefined() {\n  return typeof navigator !== \"undefined\" && navigator != null;\n}\nvar isMobileMockValue;\nfunction mockIsMobile(value) {\n  isMobileMockValue = value;\n}\nfunction isMobile(nav) {\n  if (isMobileMockValue !== void 0) {\n    return isMobileMockValue;\n  }\n  if (nav || _isNavigatorDefined()) {\n    if (!nav) {\n      nav = navigator;\n    }\n    if (nav.product === \"ReactNative\") {\n      return true;\n    }\n    const a = nav.userAgent || nav.vendor || (typeof window !== \"undefined\" ? window.opera : \"\");\n    if (!a) {\n      const navAny = nav;\n      return navAny.userAgentData && navAny.userAgentData.mobile;\n    }\n    return /(android|bb\\d+|meego).+mobile|avantgo|bada\\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(a) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\\-(n|u)|c55\\/|capi|ccwa|cdm\\-|cell|chtm|cldc|cmd\\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\\-s|devi|dica|dmob|do(c|p)o|ds(12|\\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\\-|_)|g1 u|g560|gene|gf\\-5|g\\-mo|go(\\.w|od)|gr(ad|un)|haie|hcit|hd\\-(m|p|t)|hei\\-|hi(pt|ta)|hp( i|ip)|hs\\-c|ht(c(\\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\\-(20|go|ma)|i230|iac( |\\-|\\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\\/)|klon|kpt |kwc\\-|kyo(c|k)|le(no|xi)|lg( g|\\/(k|l|u)|50|54|\\-[a-w])|libw|lynx|m1\\-w|m3ga|m50\\/|ma(te|ui|xo)|mc(01|21|ca)|m\\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\\-2|po(ck|rt|se)|prox|psio|pt\\-g|qa\\-a|qc(07|12|21|32|60|\\-[2-7]|i\\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\\-|oo|p\\-)|sdk\\/|se(c(\\-|0|1)|47|mc|nd|ri)|sgh\\-|shar|sie(\\-|m)|sk\\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\\-|v\\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\\-|tdg\\-|tel(i|m)|tim\\-|t\\-mo|to(pl|sh)|ts(70|m\\-|m3|m5)|tx\\-9|up(\\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\\-|your|zeto|zte\\-/i.test(a.substr(0, 4));\n  }\n  return false;\n}\nfunction isBrowser() {\n  return typeof window !== \"undefined\" && window.document != null || typeof WorkerGlobalScope !== \"undefined\";\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/flags.js\nvar ENV2 = env();\nENV2.registerFlag(\"DEBUG\", () => false, (debugValue) => {\n  if (debugValue) {\n    console.warn(\"Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.\");\n  }\n});\nENV2.registerFlag(\"IS_BROWSER\", () => isBrowser());\nENV2.registerFlag(\"IS_NODE\", () => typeof process !== \"undefined\" && typeof process.versions !== \"undefined\" && typeof process.versions.node !== \"undefined\");\nENV2.registerFlag(\"IS_CHROME\", () => typeof navigator !== \"undefined\" && navigator != null && navigator.userAgent != null && /Chrome/.test(navigator.userAgent) && /Google Inc/.test(navigator.vendor));\nENV2.registerFlag(\"PROD\", () => false);\nENV2.registerFlag(\"TENSORLIKE_CHECK_SHAPE_CONSISTENCY\", () => ENV2.getBool(\"DEBUG\"));\nENV2.registerFlag(\"DEPRECATION_WARNINGS_ENABLED\", () => true);\nENV2.registerFlag(\"IS_TEST\", () => false);\nENV2.registerFlag(\"CHECK_COMPUTATION_FOR_ERRORS\", () => true);\nENV2.registerFlag(\"WRAP_TO_IMAGEBITMAP\", () => false);\nENV2.registerFlag(\"ENGINE_COMPILE_ONLY\", () => false);\nENV2.registerFlag(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\", () => false);\nENV2.registerFlag(\"USE_SETTIMEOUTCUSTOM\", () => false);\n\n// node_modules/@tensorflow/tfjs-core/dist/tensor_util_env.js\nfunction inferShape(val, dtype) {\n  let firstElem = val;\n  if (isTypedArray(val)) {\n    return dtype === \"string\" ? [] : [val.length];\n  }\n  if (typeof val === \"object\" && \"texture\" in val) {\n    const usedChannels = val.channels || \"RGBA\";\n    return [val.height, val.width * usedChannels.length];\n  }\n  if (!Array.isArray(val)) {\n    return [];\n  }\n  const shape = [];\n  while (Array.isArray(firstElem) || isTypedArray(firstElem) && dtype !== \"string\") {\n    shape.push(firstElem.length);\n    firstElem = firstElem[0];\n  }\n  if (Array.isArray(val) && env().getBool(\"TENSORLIKE_CHECK_SHAPE_CONSISTENCY\")) {\n    deepAssertShapeConsistency(val, shape, []);\n  }\n  return shape;\n}\nfunction deepAssertShapeConsistency(val, shape, indices) {\n  indices = indices || [];\n  if (!Array.isArray(val) && !isTypedArray(val)) {\n    assert(shape.length === 0, () => `Element arr[${indices.join(\"][\")}] is a primitive, but should be an array/TypedArray of ${shape[0]} elements`);\n    return;\n  }\n  assert(shape.length > 0, () => `Element arr[${indices.join(\"][\")}] should be a primitive, but is an array of ${val.length} elements`);\n  assert(val.length === shape[0], () => `Element arr[${indices.join(\"][\")}] should have ${shape[0]} elements, but has ${val.length} elements`);\n  const subShape = shape.slice(1);\n  for (let i = 0; i < val.length; ++i) {\n    deepAssertShapeConsistency(val[i], subShape, indices.concat(i));\n  }\n}\nfunction assertDtype(expectedDtype, actualDType, argName, functionName) {\n  if (expectedDtype === \"string_or_numeric\") {\n    return;\n  }\n  if (expectedDtype == null) {\n    throw new Error(`Expected dtype cannot be null.`);\n  }\n  if (expectedDtype !== \"numeric\" && expectedDtype !== actualDType || expectedDtype === \"numeric\" && actualDType === \"string\") {\n    throw new Error(`Argument '${argName}' passed to '${functionName}' must be ${expectedDtype} tensor, but got ${actualDType} tensor`);\n  }\n}\nfunction convertToTensor(x, argName, functionName, parseAsDtype = \"numeric\") {\n  if (x instanceof Tensor) {\n    assertDtype(parseAsDtype, x.dtype, argName, functionName);\n    return x;\n  }\n  let inferredDtype = inferDtype(x);\n  if (inferredDtype !== \"string\" && [\"bool\", \"int32\", \"float32\"].indexOf(parseAsDtype) >= 0) {\n    inferredDtype = parseAsDtype;\n  }\n  assertDtype(parseAsDtype, inferredDtype, argName, functionName);\n  if (x == null || !isTypedArray(x) && !Array.isArray(x) && typeof x !== \"number\" && typeof x !== \"boolean\" && typeof x !== \"string\") {\n    const type = x == null ? \"null\" : x.constructor.name;\n    throw new Error(`Argument '${argName}' passed to '${functionName}' must be a Tensor or TensorLike, but got '${type}'`);\n  }\n  const inferredShape = inferShape(x, inferredDtype);\n  if (!isTypedArray(x) && !Array.isArray(x)) {\n    x = [x];\n  }\n  const skipTypedArray = true;\n  const values = inferredDtype !== \"string\" ? toTypedArray(x, inferredDtype) : flatten(x, [], skipTypedArray);\n  return ENGINE.makeTensor(values, inferredShape, inferredDtype);\n}\nfunction convertToTensorArray(arg, argName, functionName, parseAsDtype = \"numeric\") {\n  if (!Array.isArray(arg)) {\n    throw new Error(`Argument ${argName} passed to ${functionName} must be a \\`Tensor[]\\` or \\`TensorLike[]\\``);\n  }\n  const tensors = arg;\n  return tensors.map((t, i) => convertToTensor(t, `${argName}[${i}]`, functionName, parseAsDtype));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/operation.js\nvar OP_SCOPE_SUFFIX = \"__op\";\nfunction op(f) {\n  const keys = Object.keys(f);\n  if (keys.length !== 1) {\n    throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${keys.length} keys.`);\n  }\n  let opName = keys[0];\n  const fn = f[opName];\n  if (opName.endsWith(\"_\")) {\n    opName = opName.substring(0, opName.length - 1);\n  }\n  opName = opName + OP_SCOPE_SUFFIX;\n  const f2 = (...args) => {\n    ENGINE.startScope(opName);\n    try {\n      const result = fn(...args);\n      if (isPromise(result)) {\n        console.error(\"Cannot return a Promise inside of tidy.\");\n      }\n      ENGINE.endScope(result);\n      return result;\n    } catch (ex) {\n      ENGINE.endScope(null);\n      throw ex;\n    }\n  };\n  Object.defineProperty(f2, \"name\", { value: opName, configurable: true });\n  return f2;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/complex.js\nfunction complex_(real4, imag4) {\n  const $real = convertToTensor(real4, \"real\", \"complex\");\n  const $imag = convertToTensor(imag4, \"imag\", \"complex\");\n  assertShapesMatch($real.shape, $imag.shape, `real and imag shapes, ${$real.shape} and ${$imag.shape}, must match in call to tf.complex().`);\n  const inputs = { real: $real, imag: $imag };\n  return ENGINE.runKernel(Complex, inputs);\n}\nvar complex = op({ complex_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor_ops_util.js\nfunction makeTensor(values, shape, inferredShape, dtype) {\n  if (dtype == null) {\n    dtype = inferDtype(values);\n  }\n  if (dtype === \"complex64\") {\n    throw new Error(`Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).`);\n  }\n  if (typeof values === \"object\" && \"texture\" in values) {\n    if (dtype !== \"float32\" && dtype !== \"int32\") {\n      throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${dtype}.`);\n    }\n    values.channels = values.channels || \"RGBA\";\n    return ENGINE.backend.createTensorFromTexture(values, shape || inferredShape, dtype);\n  }\n  if (!isTypedArray(values) && !Array.isArray(values) && typeof values !== \"number\" && typeof values !== \"boolean\" && typeof values !== \"string\") {\n    throw new Error(\"values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray\");\n  }\n  if (shape != null) {\n    assertNonNegativeIntegerDimensions(shape);\n    const providedSize = sizeFromShape(shape);\n    const inferredSize = sizeFromShape(inferredShape);\n    assert(providedSize === inferredSize, () => `Based on the provided shape, [${shape}], the tensor should have ${providedSize} values but has ${inferredSize}`);\n    for (let i = 0; i < inferredShape.length; ++i) {\n      const inferred = inferredShape[i];\n      const flatDimsDontMatch = i === inferredShape.length - 1 ? inferred !== sizeFromShape(shape.slice(i)) : true;\n      assert(inferredShape[i] === shape[i] || !flatDimsDontMatch, () => `Error creating a new Tensor. Inferred shape (${inferredShape}) does not match the provided shape (${shape}). `);\n    }\n  }\n  if (!isTypedArray(values) && !Array.isArray(values)) {\n    values = [values];\n  }\n  shape = shape || inferredShape;\n  values = dtype !== \"string\" ? toTypedArray(values, dtype) : flatten(values, [], true);\n  return ENGINE.makeTensor(values, shape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor.js\nfunction tensor(values, shape, dtype) {\n  const inferredShape = inferShape(values, dtype);\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/types.js\nvar DTYPE_VALUE_SIZE_MAP = {\n  \"float32\": 4,\n  \"float16\": 2,\n  \"int32\": 4,\n  \"uint16\": 2,\n  \"uint8\": 1,\n  \"bool\": 1,\n  \"complex64\": 8\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/io/io_utils.js\nvar NUM_BYTES_STRING_LENGTH = 4;\nasync function encodeWeights(tensors, group) {\n  const specs = [];\n  const dataPromises = [];\n  const names = Array.isArray(tensors) ? tensors.map((tensor2) => tensor2.name) : Object.keys(tensors);\n  for (let i = 0; i < names.length; ++i) {\n    const name = names[i];\n    const t = Array.isArray(tensors) ? tensors[i].tensor : tensors[name];\n    if (t.dtype !== \"float32\" && t.dtype !== \"int32\" && t.dtype !== \"bool\" && t.dtype !== \"string\" && t.dtype !== \"complex64\") {\n      throw new Error(`Unsupported dtype in weight '${name}': ${t.dtype}`);\n    }\n    const spec = { name, shape: t.shape, dtype: t.dtype };\n    if (t.dtype === \"string\") {\n      const utf8bytes = new Promise(async (resolve) => {\n        const vals = await t.bytes();\n        const totalNumBytes = vals.reduce((p2, c) => p2 + c.length, 0) + NUM_BYTES_STRING_LENGTH * vals.length;\n        const bytes = new Uint8Array(totalNumBytes);\n        let offset = 0;\n        for (let i2 = 0; i2 < vals.length; i2++) {\n          const val = vals[i2];\n          const bytesOfLength = new Uint8Array(new Uint32Array([val.length]).buffer);\n          bytes.set(bytesOfLength, offset);\n          offset += NUM_BYTES_STRING_LENGTH;\n          bytes.set(val, offset);\n          offset += val.length;\n        }\n        resolve(bytes);\n      });\n      dataPromises.push(utf8bytes);\n    } else {\n      dataPromises.push(t.data());\n    }\n    if (group != null) {\n      spec.group = group;\n    }\n    specs.push(spec);\n  }\n  const tensorValues = await Promise.all(dataPromises);\n  return { data: concatenateTypedArrays(tensorValues), specs };\n}\nfunction decodeWeights(buffer2, specs) {\n  const out = {};\n  let float16Decode;\n  let offset = 0;\n  for (const spec of specs) {\n    const name = spec.name;\n    const dtype = spec.dtype;\n    const shape = spec.shape;\n    const size = sizeFromShape(shape);\n    let values;\n    if (\"quantization\" in spec) {\n      const quantization = spec.quantization;\n      if (quantization.dtype === \"uint8\" || quantization.dtype === \"uint16\") {\n        if (!(\"min\" in quantization && \"scale\" in quantization)) {\n          throw new Error(`Weight ${spec.name} with quantization ${quantization.dtype} doesn't have corresponding metadata min and scale.`);\n        }\n      } else if (quantization.dtype === \"float16\") {\n        if (dtype !== \"float32\") {\n          throw new Error(`Weight ${spec.name} is quantized with ${quantization.dtype} which only supports weights of type float32 not ${dtype}.`);\n        }\n      } else {\n        throw new Error(`Weight ${spec.name} has unknown quantization dtype ${quantization.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);\n      }\n      const quantizationSizeFactor = DTYPE_VALUE_SIZE_MAP[quantization.dtype];\n      const byteBuffer = buffer2.slice(offset, offset + size * quantizationSizeFactor);\n      const quantizedArray = quantization.dtype === \"uint8\" ? new Uint8Array(byteBuffer) : new Uint16Array(byteBuffer);\n      if (dtype === \"float32\") {\n        if (quantization.dtype === \"uint8\" || quantization.dtype === \"uint16\") {\n          values = new Float32Array(quantizedArray.length);\n          for (let i = 0; i < quantizedArray.length; i++) {\n            const v = quantizedArray[i];\n            values[i] = v * quantization.scale + quantization.min;\n          }\n        } else if (quantization.dtype === \"float16\") {\n          if (float16Decode === void 0) {\n            float16Decode = getFloat16Decoder();\n          }\n          values = float16Decode(quantizedArray);\n        } else {\n          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type float32.`);\n        }\n      } else if (dtype === \"int32\") {\n        if (quantization.dtype !== \"uint8\" && quantization.dtype !== \"uint16\") {\n          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type int32.`);\n        }\n        values = new Int32Array(quantizedArray.length);\n        for (let i = 0; i < quantizedArray.length; i++) {\n          const v = quantizedArray[i];\n          values[i] = Math.round(v * quantization.scale + quantization.min);\n        }\n      } else {\n        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);\n      }\n      offset += size * quantizationSizeFactor;\n    } else if (dtype === \"string\") {\n      const size2 = sizeFromShape(spec.shape);\n      values = [];\n      for (let i = 0; i < size2; i++) {\n        const byteLength = new Uint32Array(buffer2.slice(offset, offset + NUM_BYTES_STRING_LENGTH))[0];\n        offset += NUM_BYTES_STRING_LENGTH;\n        const bytes = new Uint8Array(buffer2.slice(offset, offset + byteLength));\n        values.push(bytes);\n        offset += byteLength;\n      }\n    } else {\n      const dtypeFactor = DTYPE_VALUE_SIZE_MAP[dtype];\n      const byteBuffer = buffer2.slice(offset, offset + size * dtypeFactor);\n      if (dtype === \"float32\") {\n        values = new Float32Array(byteBuffer);\n      } else if (dtype === \"int32\") {\n        values = new Int32Array(byteBuffer);\n      } else if (dtype === \"bool\") {\n        values = new Uint8Array(byteBuffer);\n      } else if (dtype === \"complex64\") {\n        values = new Float32Array(byteBuffer);\n        const real4 = new Float32Array(values.length / 2);\n        const image2 = new Float32Array(values.length / 2);\n        for (let i = 0; i < real4.length; i++) {\n          real4[i] = values[i * 2];\n          image2[i] = values[i * 2 + 1];\n        }\n        const realTensor = tensor(real4, shape, \"float32\");\n        const imageTensor = tensor(image2, shape, \"float32\");\n        out[name] = complex(realTensor, imageTensor);\n        realTensor.dispose();\n        imageTensor.dispose();\n      } else {\n        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);\n      }\n      offset += size * dtypeFactor;\n    }\n    if (dtype !== \"complex64\") {\n      out[name] = tensor(values, shape, dtype);\n    }\n  }\n  return out;\n}\nfunction concatenateTypedArrays(xs) {\n  if (xs === null) {\n    throw new Error(`Invalid input value: ${JSON.stringify(xs)}`);\n  }\n  let totalByteLength = 0;\n  const normalizedXs = [];\n  xs.forEach((x) => {\n    totalByteLength += x.byteLength;\n    normalizedXs.push(x.byteLength === x.buffer.byteLength ? x : new x.constructor(x));\n    if (!(x instanceof Float32Array || x instanceof Int32Array || x instanceof Uint8Array)) {\n      throw new Error(`Unsupported TypedArray subtype: ${x.constructor.name}`);\n    }\n  });\n  const y = new Uint8Array(totalByteLength);\n  let offset = 0;\n  normalizedXs.forEach((x) => {\n    y.set(new Uint8Array(x.buffer), offset);\n    offset += x.byteLength;\n  });\n  return y.buffer;\n}\nvar useNodeBuffer = typeof Buffer !== \"undefined\" && (typeof Blob === \"undefined\" || typeof atob === \"undefined\" || typeof btoa === \"undefined\");\nfunction stringByteLength(str) {\n  if (useNodeBuffer) {\n    return Buffer.byteLength(str);\n  }\n  return new Blob([str]).size;\n}\nfunction arrayBufferToBase64String(buffer2) {\n  if (useNodeBuffer) {\n    return Buffer.from(buffer2).toString(\"base64\");\n  }\n  const buf = new Uint8Array(buffer2);\n  let s = \"\";\n  for (let i = 0, l = buf.length; i < l; i++) {\n    s += String.fromCharCode(buf[i]);\n  }\n  return btoa(s);\n}\nfunction base64StringToArrayBuffer(str) {\n  if (useNodeBuffer) {\n    const buf = Buffer.from(str, \"base64\");\n    return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);\n  }\n  const s = atob(str);\n  const buffer2 = new Uint8Array(s.length);\n  for (let i = 0; i < s.length; ++i) {\n    buffer2.set([s.charCodeAt(i)], i);\n  }\n  return buffer2.buffer;\n}\nfunction concatenateArrayBuffers(buffers) {\n  if (buffers.length === 1) {\n    return buffers[0];\n  }\n  let totalByteLength = 0;\n  buffers.forEach((buffer2) => {\n    totalByteLength += buffer2.byteLength;\n  });\n  const temp = new Uint8Array(totalByteLength);\n  let offset = 0;\n  buffers.forEach((buffer2) => {\n    temp.set(new Uint8Array(buffer2), offset);\n    offset += buffer2.byteLength;\n  });\n  return temp.buffer;\n}\nfunction basename(path) {\n  const SEPARATOR = \"/\";\n  path = path.trim();\n  while (path.endsWith(SEPARATOR)) {\n    path = path.slice(0, path.length - 1);\n  }\n  const items = path.split(SEPARATOR);\n  return items[items.length - 1];\n}\nfunction getModelJSONForModelArtifacts(artifacts, manifest) {\n  const result = {\n    modelTopology: artifacts.modelTopology,\n    format: artifacts.format,\n    generatedBy: artifacts.generatedBy,\n    convertedBy: artifacts.convertedBy,\n    weightsManifest: manifest\n  };\n  if (artifacts.signature != null) {\n    result.signature = artifacts.signature;\n  }\n  if (artifacts.userDefinedMetadata != null) {\n    result.userDefinedMetadata = artifacts.userDefinedMetadata;\n  }\n  if (artifacts.modelInitializer != null) {\n    result.modelInitializer = artifacts.modelInitializer;\n  }\n  if (artifacts.initializerSignature != null) {\n    result.initializerSignature = artifacts.initializerSignature;\n  }\n  if (artifacts.trainingConfig != null) {\n    result.trainingConfig = artifacts.trainingConfig;\n  }\n  return result;\n}\nfunction getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData) {\n  const modelArtifacts = {\n    modelTopology: modelJSON.modelTopology,\n    format: modelJSON.format,\n    generatedBy: modelJSON.generatedBy,\n    convertedBy: modelJSON.convertedBy\n  };\n  if (modelJSON.trainingConfig != null) {\n    modelArtifacts.trainingConfig = modelJSON.trainingConfig;\n  }\n  if (modelJSON.weightsManifest != null) {\n    if (!weightSpecs) {\n      throw new Error(\"modelJSON has weightsManifest but weightSpecs is null\");\n    }\n    if (!weightData) {\n      throw new Error(\"modelJSON has weightsManifest but weightData is null\");\n    }\n    modelArtifacts.weightSpecs = weightSpecs;\n    modelArtifacts.weightData = weightData;\n  }\n  if (modelJSON.signature != null) {\n    modelArtifacts.signature = modelJSON.signature;\n  }\n  if (modelJSON.userDefinedMetadata != null) {\n    modelArtifacts.userDefinedMetadata = modelJSON.userDefinedMetadata;\n  }\n  if (modelJSON.modelInitializer != null) {\n    modelArtifacts.modelInitializer = modelJSON.modelInitializer;\n  }\n  if (modelJSON.initializerSignature != null) {\n    modelArtifacts.initializerSignature = modelJSON.initializerSignature;\n  }\n  return modelArtifacts;\n}\nasync function getModelArtifactsForJSON(modelJSON, loadWeights2) {\n  let weightSpecs;\n  let weightData;\n  if (modelJSON.weightsManifest != null) {\n    [weightSpecs, weightData] = await loadWeights2(modelJSON.weightsManifest);\n  }\n  return getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData);\n}\nfunction getModelArtifactsInfoForJSON(modelArtifacts) {\n  if (modelArtifacts.modelTopology instanceof ArrayBuffer) {\n    throw new Error(\"Expected JSON model topology, received ArrayBuffer.\");\n  }\n  return {\n    dateSaved: new Date(),\n    modelTopologyType: \"JSON\",\n    modelTopologyBytes: modelArtifacts.modelTopology == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.modelTopology)),\n    weightSpecsBytes: modelArtifacts.weightSpecs == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.weightSpecs)),\n    weightDataBytes: modelArtifacts.weightData == null ? 0 : modelArtifacts.weightData.byteLength\n  };\n}\nfunction getWeightSpecs(weightsManifest) {\n  const weightSpecs = [];\n  for (const entry of weightsManifest) {\n    weightSpecs.push(...entry.weights);\n  }\n  return weightSpecs;\n}\nfunction computeFloat16MantisaTable() {\n  const convertMantissa = (i) => {\n    let m = i << 13;\n    let e = 0;\n    while ((m & 8388608) === 0) {\n      e -= 8388608;\n      m <<= 1;\n    }\n    m &= ~8388608;\n    e += 947912704;\n    return m | e;\n  };\n  const mantisaTable = new Uint32Array(2048);\n  mantisaTable[0] = 0;\n  for (let i = 1; i < 1024; i++) {\n    mantisaTable[i] = convertMantissa(i);\n  }\n  for (let i = 1024; i < 2048; i++) {\n    mantisaTable[i] = 939524096 + (i - 1024 << 13);\n  }\n  return mantisaTable;\n}\nfunction computeFloat16ExponentTable() {\n  const exponentTable = new Uint32Array(64);\n  exponentTable[0] = 0;\n  exponentTable[31] = 1199570944;\n  exponentTable[32] = 2147483648;\n  exponentTable[63] = 3347054592;\n  for (let i = 1; i < 31; i++) {\n    exponentTable[i] = i << 23;\n  }\n  for (let i = 33; i < 63; i++) {\n    exponentTable[i] = 2147483648 + (i - 32 << 23);\n  }\n  return exponentTable;\n}\nfunction computeFloat16OffsetTable() {\n  const offsetTable = new Uint32Array(64);\n  for (let i = 0; i < 64; i++) {\n    offsetTable[i] = 1024;\n  }\n  offsetTable[0] = offsetTable[32] = 0;\n  return offsetTable;\n}\nfunction getFloat16Decoder() {\n  const mantisaTable = computeFloat16MantisaTable();\n  const exponentTable = computeFloat16ExponentTable();\n  const offsetTable = computeFloat16OffsetTable();\n  return (quantizedArray) => {\n    const buffer2 = new ArrayBuffer(4 * quantizedArray.length);\n    const bufferUint32View = new Uint32Array(buffer2);\n    for (let index = 0; index < quantizedArray.length; index++) {\n      const float16Bits = quantizedArray[index];\n      const float32Bits = mantisaTable[offsetTable[float16Bits >> 10] + (float16Bits & 1023)] + exponentTable[float16Bits >> 10];\n      bufferUint32View[index] = float32Bits;\n    }\n    return new Float32Array(buffer2);\n  };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/router_registry.js\nvar IORouterRegistry = class {\n  constructor() {\n    this.saveRouters = [];\n    this.loadRouters = [];\n  }\n  static getInstance() {\n    if (IORouterRegistry.instance == null) {\n      IORouterRegistry.instance = new IORouterRegistry();\n    }\n    return IORouterRegistry.instance;\n  }\n  static registerSaveRouter(saveRouter) {\n    IORouterRegistry.getInstance().saveRouters.push(saveRouter);\n  }\n  static registerLoadRouter(loadRouter) {\n    IORouterRegistry.getInstance().loadRouters.push(loadRouter);\n  }\n  static getSaveHandlers(url) {\n    return IORouterRegistry.getHandlers(url, \"save\");\n  }\n  static getLoadHandlers(url, loadOptions) {\n    return IORouterRegistry.getHandlers(url, \"load\", loadOptions);\n  }\n  static getHandlers(url, handlerType, loadOptions) {\n    const validHandlers = [];\n    const routers = handlerType === \"load\" ? IORouterRegistry.getInstance().loadRouters : IORouterRegistry.getInstance().saveRouters;\n    routers.forEach((router) => {\n      const handler = router(url, loadOptions);\n      if (handler !== null) {\n        validHandlers.push(handler);\n      }\n    });\n    return validHandlers;\n  }\n};\nvar registerSaveRouter = (loudRouter) => IORouterRegistry.registerSaveRouter(loudRouter);\nvar registerLoadRouter = (loudRouter) => IORouterRegistry.registerLoadRouter(loudRouter);\nvar getSaveHandlers = (url) => IORouterRegistry.getSaveHandlers(url);\nvar getLoadHandlers = (url, loadOptions) => IORouterRegistry.getLoadHandlers(url, loadOptions);\n\n// node_modules/@tensorflow/tfjs-core/dist/io/indexed_db.js\nvar DATABASE_NAME = \"tensorflowjs\";\nvar DATABASE_VERSION = 1;\nvar MODEL_STORE_NAME = \"models_store\";\nvar INFO_STORE_NAME = \"model_info_store\";\nfunction getIndexedDBFactory() {\n  if (!env().getBool(\"IS_BROWSER\")) {\n    throw new Error(\"Failed to obtain IndexedDB factory because the current environmentis not a web browser.\");\n  }\n  const theWindow = typeof window === \"undefined\" ? self : window;\n  const factory = theWindow.indexedDB || theWindow.mozIndexedDB || theWindow.webkitIndexedDB || theWindow.msIndexedDB || theWindow.shimIndexedDB;\n  if (factory == null) {\n    throw new Error(\"The current browser does not appear to support IndexedDB.\");\n  }\n  return factory;\n}\nfunction setUpDatabase(openRequest) {\n  const db = openRequest.result;\n  db.createObjectStore(MODEL_STORE_NAME, { keyPath: \"modelPath\" });\n  db.createObjectStore(INFO_STORE_NAME, { keyPath: \"modelPath\" });\n}\nvar BrowserIndexedDB = class {\n  constructor(modelPath) {\n    this.indexedDB = getIndexedDBFactory();\n    if (modelPath == null || !modelPath) {\n      throw new Error(\"For IndexedDB, modelPath must not be null, undefined or empty.\");\n    }\n    this.modelPath = modelPath;\n  }\n  async save(modelArtifacts) {\n    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {\n      throw new Error(\"BrowserLocalStorage.save() does not support saving model topology in binary formats yet.\");\n    }\n    return this.databaseAction(this.modelPath, modelArtifacts);\n  }\n  async load() {\n    return this.databaseAction(this.modelPath);\n  }\n  databaseAction(modelPath, modelArtifacts) {\n    return new Promise((resolve, reject) => {\n      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);\n      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);\n      openRequest.onsuccess = () => {\n        const db = openRequest.result;\n        if (modelArtifacts == null) {\n          const modelTx = db.transaction(MODEL_STORE_NAME, \"readonly\");\n          const modelStore = modelTx.objectStore(MODEL_STORE_NAME);\n          const getRequest = modelStore.get(this.modelPath);\n          getRequest.onsuccess = () => {\n            if (getRequest.result == null) {\n              db.close();\n              return reject(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));\n            } else {\n              resolve(getRequest.result.modelArtifacts);\n            }\n          };\n          getRequest.onerror = (error) => {\n            db.close();\n            return reject(getRequest.error);\n          };\n          modelTx.oncomplete = () => db.close();\n        } else {\n          const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);\n          const infoTx = db.transaction(INFO_STORE_NAME, \"readwrite\");\n          let infoStore = infoTx.objectStore(INFO_STORE_NAME);\n          const putInfoRequest = infoStore.put({ modelPath: this.modelPath, modelArtifactsInfo });\n          let modelTx;\n          putInfoRequest.onsuccess = () => {\n            modelTx = db.transaction(MODEL_STORE_NAME, \"readwrite\");\n            const modelStore = modelTx.objectStore(MODEL_STORE_NAME);\n            const putModelRequest = modelStore.put({\n              modelPath: this.modelPath,\n              modelArtifacts,\n              modelArtifactsInfo\n            });\n            putModelRequest.onsuccess = () => resolve({ modelArtifactsInfo });\n            putModelRequest.onerror = (error) => {\n              infoStore = infoTx.objectStore(INFO_STORE_NAME);\n              const deleteInfoRequest = infoStore.delete(this.modelPath);\n              deleteInfoRequest.onsuccess = () => {\n                db.close();\n                return reject(putModelRequest.error);\n              };\n              deleteInfoRequest.onerror = (error2) => {\n                db.close();\n                return reject(putModelRequest.error);\n              };\n            };\n          };\n          putInfoRequest.onerror = (error) => {\n            db.close();\n            return reject(putInfoRequest.error);\n          };\n          infoTx.oncomplete = () => {\n            if (modelTx == null) {\n              db.close();\n            } else {\n              modelTx.oncomplete = () => db.close();\n            }\n          };\n        }\n      };\n      openRequest.onerror = (error) => reject(openRequest.error);\n    });\n  }\n};\nBrowserIndexedDB.URL_SCHEME = \"indexeddb://\";\nvar indexedDBRouter = (url) => {\n  if (!env().getBool(\"IS_BROWSER\")) {\n    return null;\n  } else {\n    if (!Array.isArray(url) && url.startsWith(BrowserIndexedDB.URL_SCHEME)) {\n      return browserIndexedDB(url.slice(BrowserIndexedDB.URL_SCHEME.length));\n    } else {\n      return null;\n    }\n  }\n};\nIORouterRegistry.registerSaveRouter(indexedDBRouter);\nIORouterRegistry.registerLoadRouter(indexedDBRouter);\nfunction browserIndexedDB(modelPath) {\n  return new BrowserIndexedDB(modelPath);\n}\nfunction maybeStripScheme(key) {\n  return key.startsWith(BrowserIndexedDB.URL_SCHEME) ? key.slice(BrowserIndexedDB.URL_SCHEME.length) : key;\n}\nvar BrowserIndexedDBManager = class {\n  constructor() {\n    this.indexedDB = getIndexedDBFactory();\n  }\n  async listModels() {\n    return new Promise((resolve, reject) => {\n      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);\n      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);\n      openRequest.onsuccess = () => {\n        const db = openRequest.result;\n        const tx = db.transaction(INFO_STORE_NAME, \"readonly\");\n        const store = tx.objectStore(INFO_STORE_NAME);\n        const getAllInfoRequest = store.getAll();\n        getAllInfoRequest.onsuccess = () => {\n          const out = {};\n          for (const item of getAllInfoRequest.result) {\n            out[item.modelPath] = item.modelArtifactsInfo;\n          }\n          resolve(out);\n        };\n        getAllInfoRequest.onerror = (error) => {\n          db.close();\n          return reject(getAllInfoRequest.error);\n        };\n        tx.oncomplete = () => db.close();\n      };\n      openRequest.onerror = (error) => reject(openRequest.error);\n    });\n  }\n  async removeModel(path) {\n    path = maybeStripScheme(path);\n    return new Promise((resolve, reject) => {\n      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);\n      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);\n      openRequest.onsuccess = () => {\n        const db = openRequest.result;\n        const infoTx = db.transaction(INFO_STORE_NAME, \"readwrite\");\n        const infoStore = infoTx.objectStore(INFO_STORE_NAME);\n        const getInfoRequest = infoStore.get(path);\n        let modelTx;\n        getInfoRequest.onsuccess = () => {\n          if (getInfoRequest.result == null) {\n            db.close();\n            return reject(new Error(`Cannot find model with path '${path}' in IndexedDB.`));\n          } else {\n            const deleteInfoRequest = infoStore.delete(path);\n            const deleteModelData = () => {\n              modelTx = db.transaction(MODEL_STORE_NAME, \"readwrite\");\n              const modelStore = modelTx.objectStore(MODEL_STORE_NAME);\n              const deleteModelRequest = modelStore.delete(path);\n              deleteModelRequest.onsuccess = () => resolve(getInfoRequest.result.modelArtifactsInfo);\n              deleteModelRequest.onerror = (error) => reject(getInfoRequest.error);\n            };\n            deleteInfoRequest.onsuccess = deleteModelData;\n            deleteInfoRequest.onerror = (error) => {\n              deleteModelData();\n              db.close();\n              return reject(getInfoRequest.error);\n            };\n          }\n        };\n        getInfoRequest.onerror = (error) => {\n          db.close();\n          return reject(getInfoRequest.error);\n        };\n        infoTx.oncomplete = () => {\n          if (modelTx == null) {\n            db.close();\n          } else {\n            modelTx.oncomplete = () => db.close();\n          }\n        };\n      };\n      openRequest.onerror = (error) => reject(openRequest.error);\n    });\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/io/local_storage.js\nvar PATH_SEPARATOR = \"/\";\nvar PATH_PREFIX = \"tensorflowjs_models\";\nvar INFO_SUFFIX = \"info\";\nvar MODEL_TOPOLOGY_SUFFIX = \"model_topology\";\nvar WEIGHT_SPECS_SUFFIX = \"weight_specs\";\nvar WEIGHT_DATA_SUFFIX = \"weight_data\";\nvar MODEL_METADATA_SUFFIX = \"model_metadata\";\nfunction getModelKeys(path) {\n  return {\n    info: [PATH_PREFIX, path, INFO_SUFFIX].join(PATH_SEPARATOR),\n    topology: [PATH_PREFIX, path, MODEL_TOPOLOGY_SUFFIX].join(PATH_SEPARATOR),\n    weightSpecs: [PATH_PREFIX, path, WEIGHT_SPECS_SUFFIX].join(PATH_SEPARATOR),\n    weightData: [PATH_PREFIX, path, WEIGHT_DATA_SUFFIX].join(PATH_SEPARATOR),\n    modelMetadata: [PATH_PREFIX, path, MODEL_METADATA_SUFFIX].join(PATH_SEPARATOR)\n  };\n}\nfunction removeItems(keys) {\n  for (const key of Object.values(keys)) {\n    window.localStorage.removeItem(key);\n  }\n}\nfunction getModelPathFromKey(key) {\n  const items = key.split(PATH_SEPARATOR);\n  if (items.length < 3) {\n    throw new Error(`Invalid key format: ${key}`);\n  }\n  return items.slice(1, items.length - 1).join(PATH_SEPARATOR);\n}\nfunction maybeStripScheme2(key) {\n  return key.startsWith(BrowserLocalStorage.URL_SCHEME) ? key.slice(BrowserLocalStorage.URL_SCHEME.length) : key;\n}\nvar BrowserLocalStorage = class {\n  constructor(modelPath) {\n    if (!env().getBool(\"IS_BROWSER\") || typeof window === \"undefined\" || typeof window.localStorage === \"undefined\") {\n      throw new Error(\"The current environment does not support local storage.\");\n    }\n    this.LS = window.localStorage;\n    if (modelPath == null || !modelPath) {\n      throw new Error(\"For local storage, modelPath must not be null, undefined or empty.\");\n    }\n    this.modelPath = modelPath;\n    this.keys = getModelKeys(this.modelPath);\n  }\n  async save(modelArtifacts) {\n    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {\n      throw new Error(\"BrowserLocalStorage.save() does not support saving model topology in binary formats yet.\");\n    } else {\n      const topology = JSON.stringify(modelArtifacts.modelTopology);\n      const weightSpecs = JSON.stringify(modelArtifacts.weightSpecs);\n      const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);\n      try {\n        this.LS.setItem(this.keys.info, JSON.stringify(modelArtifactsInfo));\n        this.LS.setItem(this.keys.topology, topology);\n        this.LS.setItem(this.keys.weightSpecs, weightSpecs);\n        this.LS.setItem(this.keys.weightData, arrayBufferToBase64String(modelArtifacts.weightData));\n        const metadata = {\n          format: modelArtifacts.format,\n          generatedBy: modelArtifacts.generatedBy,\n          convertedBy: modelArtifacts.convertedBy,\n          signature: modelArtifacts.signature != null ? modelArtifacts.signature : void 0,\n          userDefinedMetadata: modelArtifacts.userDefinedMetadata != null ? modelArtifacts.userDefinedMetadata : void 0,\n          modelInitializer: modelArtifacts.modelInitializer != null ? modelArtifacts.modelInitializer : void 0,\n          initializerSignature: modelArtifacts.initializerSignature != null ? modelArtifacts.initializerSignature : void 0,\n          trainingConfig: modelArtifacts.trainingConfig != null ? modelArtifacts.trainingConfig : void 0\n        };\n        this.LS.setItem(this.keys.modelMetadata, JSON.stringify(metadata));\n        return { modelArtifactsInfo };\n      } catch (err) {\n        removeItems(this.keys);\n        throw new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${modelArtifactsInfo.modelTopologyBytes}, weightSpecsBytes=${modelArtifactsInfo.weightSpecsBytes}, weightDataBytes=${modelArtifactsInfo.weightDataBytes}.`);\n      }\n    }\n  }\n  async load() {\n    const info = JSON.parse(this.LS.getItem(this.keys.info));\n    if (info == null) {\n      throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);\n    }\n    if (info.modelTopologyType !== \"JSON\") {\n      throw new Error(\"BrowserLocalStorage does not support loading non-JSON model topology yet.\");\n    }\n    const out = {};\n    const topology = JSON.parse(this.LS.getItem(this.keys.topology));\n    if (topology == null) {\n      throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);\n    }\n    out.modelTopology = topology;\n    const weightSpecs = JSON.parse(this.LS.getItem(this.keys.weightSpecs));\n    if (weightSpecs == null) {\n      throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);\n    }\n    out.weightSpecs = weightSpecs;\n    const metadataString = this.LS.getItem(this.keys.modelMetadata);\n    if (metadataString != null) {\n      const metadata = JSON.parse(metadataString);\n      out.format = metadata.format;\n      out.generatedBy = metadata.generatedBy;\n      out.convertedBy = metadata.convertedBy;\n      if (metadata.signature != null) {\n        out.signature = metadata.signature;\n      }\n      if (metadata.userDefinedMetadata != null) {\n        out.userDefinedMetadata = metadata.userDefinedMetadata;\n      }\n      if (metadata.modelInitializer != null) {\n        out.modelInitializer = metadata.modelInitializer;\n      }\n      if (metadata.initializerSignature != null) {\n        out.initializerSignature = metadata.initializerSignature;\n      }\n      if (metadata.trainingConfig != null) {\n        out.trainingConfig = metadata.trainingConfig;\n      }\n    }\n    const weightDataBase64 = this.LS.getItem(this.keys.weightData);\n    if (weightDataBase64 == null) {\n      throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);\n    }\n    out.weightData = base64StringToArrayBuffer(weightDataBase64);\n    return out;\n  }\n};\nBrowserLocalStorage.URL_SCHEME = \"localstorage://\";\nvar localStorageRouter = (url) => {\n  if (!env().getBool(\"IS_BROWSER\")) {\n    return null;\n  } else {\n    if (!Array.isArray(url) && url.startsWith(BrowserLocalStorage.URL_SCHEME)) {\n      return browserLocalStorage(url.slice(BrowserLocalStorage.URL_SCHEME.length));\n    } else {\n      return null;\n    }\n  }\n};\nIORouterRegistry.registerSaveRouter(localStorageRouter);\nIORouterRegistry.registerLoadRouter(localStorageRouter);\nfunction browserLocalStorage(modelPath) {\n  return new BrowserLocalStorage(modelPath);\n}\nvar BrowserLocalStorageManager = class {\n  constructor() {\n    assert(env().getBool(\"IS_BROWSER\"), () => \"Current environment is not a web browser\");\n    assert(typeof window === \"undefined\" || typeof window.localStorage !== \"undefined\", () => \"Current browser does not appear to support localStorage\");\n    this.LS = window.localStorage;\n  }\n  async listModels() {\n    const out = {};\n    const prefix = PATH_PREFIX + PATH_SEPARATOR;\n    const suffix = PATH_SEPARATOR + INFO_SUFFIX;\n    for (let i = 0; i < this.LS.length; ++i) {\n      const key = this.LS.key(i);\n      if (key.startsWith(prefix) && key.endsWith(suffix)) {\n        const modelPath = getModelPathFromKey(key);\n        out[modelPath] = JSON.parse(this.LS.getItem(key));\n      }\n    }\n    return out;\n  }\n  async removeModel(path) {\n    path = maybeStripScheme2(path);\n    const keys = getModelKeys(path);\n    if (this.LS.getItem(keys.info) == null) {\n      throw new Error(`Cannot find model at path '${path}'`);\n    }\n    const info = JSON.parse(this.LS.getItem(keys.info));\n    removeItems(keys);\n    return info;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/io/model_management.js\nvar URL_SCHEME_SUFFIX = \"://\";\nvar ModelStoreManagerRegistry = class {\n  constructor() {\n    this.managers = {};\n  }\n  static getInstance() {\n    if (ModelStoreManagerRegistry.instance == null) {\n      ModelStoreManagerRegistry.instance = new ModelStoreManagerRegistry();\n    }\n    return ModelStoreManagerRegistry.instance;\n  }\n  static registerManager(scheme, manager) {\n    assert(scheme != null, () => \"scheme must not be undefined or null.\");\n    if (scheme.endsWith(URL_SCHEME_SUFFIX)) {\n      scheme = scheme.slice(0, scheme.indexOf(URL_SCHEME_SUFFIX));\n    }\n    assert(scheme.length > 0, () => \"scheme must not be an empty string.\");\n    const registry = ModelStoreManagerRegistry.getInstance();\n    assert(registry.managers[scheme] == null, () => `A model store manager is already registered for scheme '${scheme}'.`);\n    registry.managers[scheme] = manager;\n  }\n  static getManager(scheme) {\n    const manager = ModelStoreManagerRegistry.getInstance().managers[scheme];\n    if (manager == null) {\n      throw new Error(`Cannot find model manager for scheme '${scheme}'`);\n    }\n    return manager;\n  }\n  static getSchemes() {\n    return Object.keys(ModelStoreManagerRegistry.getInstance().managers);\n  }\n};\nfunction parseURL(url) {\n  if (url.indexOf(URL_SCHEME_SUFFIX) === -1) {\n    throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${ModelStoreManagerRegistry.getSchemes().join(\",\")}`);\n  }\n  return {\n    scheme: url.split(URL_SCHEME_SUFFIX)[0],\n    path: url.split(URL_SCHEME_SUFFIX)[1]\n  };\n}\nasync function cloneModelInternal(sourceURL, destURL, deleteSource = false) {\n  assert(sourceURL !== destURL, () => `Old path and new path are the same: '${sourceURL}'`);\n  const loadHandlers = IORouterRegistry.getLoadHandlers(sourceURL);\n  assert(loadHandlers.length > 0, () => `Copying failed because no load handler is found for source URL ${sourceURL}.`);\n  assert(loadHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) load handlers for source URL ${sourceURL}.`);\n  const loadHandler = loadHandlers[0];\n  const saveHandlers = IORouterRegistry.getSaveHandlers(destURL);\n  assert(saveHandlers.length > 0, () => `Copying failed because no save handler is found for destination URL ${destURL}.`);\n  assert(saveHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) save handlers for destination URL ${destURL}.`);\n  const saveHandler = saveHandlers[0];\n  const sourceScheme = parseURL(sourceURL).scheme;\n  const sourcePath = parseURL(sourceURL).path;\n  const sameMedium = sourceScheme === parseURL(sourceURL).scheme;\n  const modelArtifacts = await loadHandler.load();\n  if (deleteSource && sameMedium) {\n    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);\n  }\n  const saveResult = await saveHandler.save(modelArtifacts);\n  if (deleteSource && !sameMedium) {\n    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);\n  }\n  return saveResult.modelArtifactsInfo;\n}\nasync function listModels() {\n  const schemes = ModelStoreManagerRegistry.getSchemes();\n  const out = {};\n  for (const scheme of schemes) {\n    const schemeOut = await ModelStoreManagerRegistry.getManager(scheme).listModels();\n    for (const path in schemeOut) {\n      const url = scheme + URL_SCHEME_SUFFIX + path;\n      out[url] = schemeOut[path];\n    }\n  }\n  return out;\n}\nasync function removeModel(url) {\n  const schemeAndPath = parseURL(url);\n  const manager = ModelStoreManagerRegistry.getManager(schemeAndPath.scheme);\n  return manager.removeModel(schemeAndPath.path);\n}\nasync function copyModel(sourceURL, destURL) {\n  const deleteSource = false;\n  return cloneModelInternal(sourceURL, destURL, deleteSource);\n}\nasync function moveModel(sourceURL, destURL) {\n  const deleteSource = true;\n  return cloneModelInternal(sourceURL, destURL, deleteSource);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/platforms/platform_browser.js\nvar PlatformBrowser = class {\n  constructor() {\n    this.messageName = \"setTimeoutCustom\";\n    this.functionRefs = [];\n    this.handledMessageCount = 0;\n    this.hasEventListener = false;\n  }\n  fetch(path, init2) {\n    return fetch(path, init2);\n  }\n  now() {\n    return performance.now();\n  }\n  encode(text, encoding) {\n    if (encoding !== \"utf-8\" && encoding !== \"utf8\") {\n      throw new Error(`Browser's encoder only supports utf-8, but got ${encoding}`);\n    }\n    if (this.textEncoder == null) {\n      this.textEncoder = new TextEncoder();\n    }\n    return this.textEncoder.encode(text);\n  }\n  decode(bytes, encoding) {\n    return new TextDecoder(encoding).decode(bytes);\n  }\n  setTimeoutCustom(functionRef, delay) {\n    if (typeof window === \"undefined\" || !env().getBool(\"USE_SETTIMEOUTCUSTOM\")) {\n      setTimeout(functionRef, delay);\n      return;\n    }\n    this.functionRefs.push(functionRef);\n    setTimeout(() => {\n      window.postMessage({ name: this.messageName, index: this.functionRefs.length - 1 }, \"*\");\n    }, delay);\n    if (!this.hasEventListener) {\n      this.hasEventListener = true;\n      window.addEventListener(\"message\", (event) => {\n        if (event.source === window && event.data.name === this.messageName) {\n          event.stopPropagation();\n          const functionRef2 = this.functionRefs[event.data.index];\n          functionRef2();\n          this.handledMessageCount++;\n          if (this.handledMessageCount === this.functionRefs.length) {\n            this.functionRefs = [];\n            this.handledMessageCount = 0;\n          }\n        }\n      }, true);\n    }\n  }\n};\nif (env().get(\"IS_BROWSER\")) {\n  env().setPlatform(\"browser\", new PlatformBrowser());\n  try {\n    ModelStoreManagerRegistry.registerManager(BrowserLocalStorage.URL_SCHEME, new BrowserLocalStorageManager());\n  } catch (err) {\n  }\n  try {\n    ModelStoreManagerRegistry.registerManager(BrowserIndexedDB.URL_SCHEME, new BrowserIndexedDBManager());\n  } catch (err) {\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/platforms/platform_node.js\nvar getNodeFetch = {\n  importFetch: () => require_browser()\n};\nvar systemFetch;\nvar PlatformNode = class {\n  constructor() {\n    this.util = require_util();\n    this.textEncoder = new this.util.TextEncoder();\n  }\n  fetch(path, requestInits) {\n    if (env().global.fetch != null) {\n      return env().global.fetch(path, requestInits);\n    }\n    if (systemFetch == null) {\n      systemFetch = getNodeFetch.importFetch();\n    }\n    return systemFetch(path, requestInits);\n  }\n  now() {\n    const time2 = process.hrtime();\n    return time2[0] * 1e3 + time2[1] / 1e6;\n  }\n  encode(text, encoding) {\n    if (encoding !== \"utf-8\" && encoding !== \"utf8\") {\n      throw new Error(`Node built-in encoder only supports utf-8, but got ${encoding}`);\n    }\n    return this.textEncoder.encode(text);\n  }\n  decode(bytes, encoding) {\n    if (bytes.length === 0) {\n      return \"\";\n    }\n    return new this.util.TextDecoder(encoding).decode(bytes);\n  }\n};\nif (env().get(\"IS_NODE\") && !env().get(\"IS_BROWSER\")) {\n  env().setPlatform(\"node\", new PlatformNode());\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/buffer.js\nfunction buffer(shape, dtype = \"float32\", values) {\n  dtype = dtype || \"float32\";\n  assertNonNegativeIntegerDimensions(shape);\n  return new TensorBuffer(shape, dtype, values);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/cast.js\nfunction cast_(x, dtype) {\n  const $x = convertToTensor(x, \"x\", \"cast\");\n  if (!isValidDtype(dtype)) {\n    throw new Error(`Failed to cast to unknown dtype ${dtype}`);\n  }\n  if (dtype === \"string\" && $x.dtype !== \"string\" || dtype !== \"string\" && $x.dtype === \"string\") {\n    throw new Error(\"Only strings can be casted to strings\");\n  }\n  const inputs = { x: $x };\n  const attrs = { dtype };\n  return ENGINE.runKernel(Cast, inputs, attrs);\n}\nvar cast = op({ cast_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/clone.js\nfunction clone_(x) {\n  const $x = convertToTensor(x, \"x\", \"clone\", \"string_or_numeric\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Identity, inputs);\n}\nvar clone = op({ clone_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/print.js\nfunction print(x, verbose = false) {\n  console.log(x.toString(verbose));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/base_side_effects.js\ngetOrMakeEngine();\nvar opHandler2 = {\n  buffer,\n  cast,\n  clone,\n  print\n};\nsetOpHandler(opHandler2);\n\n// node_modules/@tensorflow/tfjs-core/dist/io/io.js\nvar io_exports = {};\n__export(io_exports, {\n  browserFiles: () => browserFiles,\n  browserHTTPRequest: () => browserHTTPRequest,\n  concatenateArrayBuffers: () => concatenateArrayBuffers,\n  copyModel: () => copyModel,\n  decodeWeights: () => decodeWeights,\n  encodeWeights: () => encodeWeights,\n  fromMemory: () => fromMemory,\n  fromMemorySync: () => fromMemorySync,\n  getLoadHandlers: () => getLoadHandlers,\n  getModelArtifactsForJSON: () => getModelArtifactsForJSON,\n  getModelArtifactsForJSONSync: () => getModelArtifactsForJSONSync,\n  getModelArtifactsInfoForJSON: () => getModelArtifactsInfoForJSON,\n  getSaveHandlers: () => getSaveHandlers,\n  getWeightSpecs: () => getWeightSpecs,\n  http: () => http,\n  isHTTPScheme: () => isHTTPScheme,\n  listModels: () => listModels,\n  loadWeights: () => loadWeights,\n  moveModel: () => moveModel,\n  registerLoadRouter: () => registerLoadRouter,\n  registerSaveRouter: () => registerSaveRouter,\n  removeModel: () => removeModel,\n  weightsLoaderFactory: () => weightsLoaderFactory,\n  withSaveHandler: () => withSaveHandler,\n  withSaveHandlerSync: () => withSaveHandlerSync\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/io/browser_files.js\nvar DEFAULT_FILE_NAME_PREFIX = \"model\";\nvar DEFAULT_JSON_EXTENSION_NAME = \".json\";\nvar DEFAULT_WEIGHT_DATA_EXTENSION_NAME = \".weights.bin\";\nfunction defer(f) {\n  return new Promise((resolve) => setTimeout(resolve)).then(f);\n}\nvar BrowserDownloads = class {\n  constructor(fileNamePrefix) {\n    if (!env().getBool(\"IS_BROWSER\")) {\n      throw new Error(\"browserDownloads() cannot proceed because the current environment is not a browser.\");\n    }\n    if (fileNamePrefix.startsWith(BrowserDownloads.URL_SCHEME)) {\n      fileNamePrefix = fileNamePrefix.slice(BrowserDownloads.URL_SCHEME.length);\n    }\n    if (fileNamePrefix == null || fileNamePrefix.length === 0) {\n      fileNamePrefix = DEFAULT_FILE_NAME_PREFIX;\n    }\n    this.modelJsonFileName = fileNamePrefix + DEFAULT_JSON_EXTENSION_NAME;\n    this.weightDataFileName = fileNamePrefix + DEFAULT_WEIGHT_DATA_EXTENSION_NAME;\n  }\n  async save(modelArtifacts) {\n    if (typeof document === \"undefined\") {\n      throw new Error(\"Browser downloads are not supported in this environment since `document` is not present\");\n    }\n    const weightsURL = window.URL.createObjectURL(new Blob([modelArtifacts.weightData], { type: \"application/octet-stream\" }));\n    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {\n      throw new Error(\"BrowserDownloads.save() does not support saving model topology in binary formats yet.\");\n    } else {\n      const weightsManifest = [{\n        paths: [\"./\" + this.weightDataFileName],\n        weights: modelArtifacts.weightSpecs\n      }];\n      const modelJSON = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);\n      const modelJsonURL = window.URL.createObjectURL(new Blob([JSON.stringify(modelJSON)], { type: \"application/json\" }));\n      const jsonAnchor = this.modelJsonAnchor == null ? document.createElement(\"a\") : this.modelJsonAnchor;\n      jsonAnchor.download = this.modelJsonFileName;\n      jsonAnchor.href = modelJsonURL;\n      await defer(() => jsonAnchor.dispatchEvent(new MouseEvent(\"click\")));\n      if (modelArtifacts.weightData != null) {\n        const weightDataAnchor = this.weightDataAnchor == null ? document.createElement(\"a\") : this.weightDataAnchor;\n        weightDataAnchor.download = this.weightDataFileName;\n        weightDataAnchor.href = weightsURL;\n        await defer(() => weightDataAnchor.dispatchEvent(new MouseEvent(\"click\")));\n      }\n      return { modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts) };\n    }\n  }\n};\nBrowserDownloads.URL_SCHEME = \"downloads://\";\nvar BrowserFiles = class {\n  constructor(files) {\n    if (files == null || files.length < 1) {\n      throw new Error(`When calling browserFiles, at least 1 file is required, but received ${files}`);\n    }\n    this.jsonFile = files[0];\n    this.weightsFiles = files.slice(1);\n  }\n  async load() {\n    return new Promise((resolve, reject) => {\n      const jsonReader = new FileReader();\n      jsonReader.onload = (event) => {\n        const modelJSON = JSON.parse(event.target.result);\n        const modelTopology = modelJSON.modelTopology;\n        if (modelTopology == null) {\n          reject(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));\n          return;\n        }\n        const weightsManifest = modelJSON.weightsManifest;\n        if (weightsManifest == null) {\n          reject(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));\n          return;\n        }\n        if (this.weightsFiles.length === 0) {\n          resolve({ modelTopology });\n          return;\n        }\n        const modelArtifactsPromise = getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));\n        resolve(modelArtifactsPromise);\n      };\n      jsonReader.onerror = (error) => reject(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`);\n      jsonReader.readAsText(this.jsonFile);\n    });\n  }\n  loadWeights(weightsManifest) {\n    const weightSpecs = [];\n    const paths = [];\n    for (const entry of weightsManifest) {\n      weightSpecs.push(...entry.weights);\n      paths.push(...entry.paths);\n    }\n    const pathToFile = this.checkManifestAndWeightFiles(weightsManifest);\n    const promises = paths.map((path) => this.loadWeightsFile(path, pathToFile[path]));\n    return Promise.all(promises).then((buffers) => [weightSpecs, concatenateArrayBuffers(buffers)]);\n  }\n  loadWeightsFile(path, file) {\n    return new Promise((resolve, reject) => {\n      const weightFileReader = new FileReader();\n      weightFileReader.onload = (event) => {\n        const weightData = event.target.result;\n        resolve(weightData);\n      };\n      weightFileReader.onerror = (error) => reject(`Failed to weights data from file of path '${path}'.`);\n      weightFileReader.readAsArrayBuffer(file);\n    });\n  }\n  checkManifestAndWeightFiles(manifest) {\n    const basenames = [];\n    const fileNames = this.weightsFiles.map((file) => basename(file.name));\n    const pathToFile = {};\n    for (const group of manifest) {\n      group.paths.forEach((path) => {\n        const pathBasename = basename(path);\n        if (basenames.indexOf(pathBasename) !== -1) {\n          throw new Error(`Duplicate file basename found in weights manifest: '${pathBasename}'`);\n        }\n        basenames.push(pathBasename);\n        if (fileNames.indexOf(pathBasename) === -1) {\n          throw new Error(`Weight file with basename '${pathBasename}' is not provided.`);\n        } else {\n          pathToFile[path] = this.weightsFiles[fileNames.indexOf(pathBasename)];\n        }\n      });\n    }\n    if (basenames.length !== this.weightsFiles.length) {\n      throw new Error(`Mismatch in the number of files in weights manifest (${basenames.length}) and the number of weight files provided (${this.weightsFiles.length}).`);\n    }\n    return pathToFile;\n  }\n};\nvar browserDownloadsRouter = (url) => {\n  if (!env().getBool(\"IS_BROWSER\")) {\n    return null;\n  } else {\n    if (!Array.isArray(url) && url.startsWith(BrowserDownloads.URL_SCHEME)) {\n      return browserDownloads(url.slice(BrowserDownloads.URL_SCHEME.length));\n    } else {\n      return null;\n    }\n  }\n};\nIORouterRegistry.registerSaveRouter(browserDownloadsRouter);\nfunction browserDownloads(fileNamePrefix = \"model\") {\n  return new BrowserDownloads(fileNamePrefix);\n}\nfunction browserFiles(files) {\n  return new BrowserFiles(files);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/progress.js\nfunction monitorPromisesProgress(promises, onProgress, startFraction, endFraction) {\n  checkPromises(promises);\n  startFraction = startFraction == null ? 0 : startFraction;\n  endFraction = endFraction == null ? 1 : endFraction;\n  checkFraction(startFraction, endFraction);\n  let resolvedPromise = 0;\n  const registerMonitor = (promise) => {\n    promise.then((value) => {\n      const fraction = startFraction + ++resolvedPromise / promises.length * (endFraction - startFraction);\n      onProgress(fraction);\n      return value;\n    });\n    return promise;\n  };\n  function checkPromises(promises2) {\n    assert(promises2 != null && Array.isArray(promises2) && promises2.length > 0, () => \"promises must be a none empty array\");\n  }\n  function checkFraction(startFraction2, endFraction2) {\n    assert(startFraction2 >= 0 && startFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got startFraction ${startFraction2}`);\n    assert(endFraction2 >= 0 && endFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got endFraction ${endFraction2}`);\n    assert(endFraction2 >= startFraction2, () => `startFraction must be no more than endFraction, but got startFraction ${startFraction2} and endFraction ${endFraction2}`);\n  }\n  return Promise.all(promises.map(registerMonitor));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/weights_loader.js\nasync function loadWeightsAsArrayBuffer(fetchURLs, loadOptions) {\n  if (loadOptions == null) {\n    loadOptions = {};\n  }\n  const fetchFunc = loadOptions.fetchFunc == null ? env().platform.fetch : loadOptions.fetchFunc;\n  const requests = fetchURLs.map((fetchURL) => fetchFunc(fetchURL, loadOptions.requestInit, { isBinary: true }));\n  const fetchStartFraction = 0;\n  const fetchEndFraction = 0.5;\n  const responses = loadOptions.onProgress == null ? await Promise.all(requests) : await monitorPromisesProgress(requests, loadOptions.onProgress, fetchStartFraction, fetchEndFraction);\n  const bufferPromises = responses.map((response) => response.arrayBuffer());\n  const bufferStartFraction = 0.5;\n  const bufferEndFraction = 1;\n  const buffers = loadOptions.onProgress == null ? await Promise.all(bufferPromises) : await monitorPromisesProgress(bufferPromises, loadOptions.onProgress, bufferStartFraction, bufferEndFraction);\n  return buffers;\n}\nasync function loadWeights(manifest, filePathPrefix = \"\", weightNames, requestInit) {\n  const fetchWeights = (fetchUrls) => loadWeightsAsArrayBuffer(fetchUrls, { requestInit });\n  const loadWeights2 = weightsLoaderFactory(fetchWeights);\n  return loadWeights2(manifest, filePathPrefix, weightNames);\n}\nfunction weightsLoaderFactory(fetchWeightsFunction) {\n  return async (manifest, filePathPrefix = \"\", weightNames) => {\n    const groupIndicesToFetchMap = manifest.map(() => false);\n    const groupWeightsToFetch = {};\n    const weightsFound = weightNames != null ? weightNames.map(() => false) : [];\n    const allManifestWeightNames = [];\n    manifest.forEach((manifestGroupConfig, groupIndex) => {\n      let groupOffset = 0;\n      manifestGroupConfig.weights.forEach((weightsEntry) => {\n        const rawDtype = \"quantization\" in weightsEntry ? weightsEntry.quantization.dtype : weightsEntry.dtype;\n        const weightsBytes = DTYPE_VALUE_SIZE_MAP[rawDtype] * sizeFromShape(weightsEntry.shape);\n        const enqueueWeightsForFetchingFn = () => {\n          groupIndicesToFetchMap[groupIndex] = true;\n          if (groupWeightsToFetch[groupIndex] == null) {\n            groupWeightsToFetch[groupIndex] = [];\n          }\n          groupWeightsToFetch[groupIndex].push({\n            manifestEntry: weightsEntry,\n            groupOffset,\n            sizeBytes: weightsBytes\n          });\n        };\n        if (weightNames != null) {\n          weightNames.forEach((weightName, weightIndex) => {\n            if (weightName === weightsEntry.name) {\n              enqueueWeightsForFetchingFn();\n              weightsFound[weightIndex] = true;\n            }\n          });\n        } else {\n          enqueueWeightsForFetchingFn();\n        }\n        allManifestWeightNames.push(weightsEntry.name);\n        groupOffset += weightsBytes;\n      });\n    });\n    if (!weightsFound.every((found) => found)) {\n      const weightsNotFound = weightNames.filter((_, i) => !weightsFound[i]);\n      throw new Error(`Could not find weights in manifest with names: ${weightsNotFound.join(\", \")}. \nManifest JSON has weights with names: ${allManifestWeightNames.join(\", \")}.`);\n    }\n    const groupIndicesToFetch = groupIndicesToFetchMap.reduce((accumulator, shouldFetch, i) => {\n      if (shouldFetch) {\n        accumulator.push(i);\n      }\n      return accumulator;\n    }, []);\n    const fetchUrls = [];\n    groupIndicesToFetch.forEach((i) => {\n      manifest[i].paths.forEach((filepath) => {\n        const fetchUrl = filePathPrefix + (!filePathPrefix.endsWith(\"/\") ? \"/\" : \"\") + filepath;\n        fetchUrls.push(fetchUrl);\n      });\n    });\n    const buffers = await fetchWeightsFunction(fetchUrls);\n    const weightsTensorMap = {};\n    let bufferIndexOffset = 0;\n    groupIndicesToFetch.forEach((i) => {\n      const numBuffers = manifest[i].paths.length;\n      let groupBytes = 0;\n      for (let i2 = 0; i2 < numBuffers; i2++) {\n        groupBytes += buffers[bufferIndexOffset + i2].byteLength;\n      }\n      const groupBuffer = new ArrayBuffer(groupBytes);\n      const groupByteBuffer = new Uint8Array(groupBuffer);\n      let groupBufferOffset = 0;\n      for (let i2 = 0; i2 < numBuffers; i2++) {\n        const buffer2 = new Uint8Array(buffers[bufferIndexOffset + i2]);\n        groupByteBuffer.set(buffer2, groupBufferOffset);\n        groupBufferOffset += buffer2.byteLength;\n      }\n      const weightsEntries = groupWeightsToFetch[i];\n      weightsEntries.forEach((weightsEntry) => {\n        const byteBuffer = groupBuffer.slice(weightsEntry.groupOffset, weightsEntry.groupOffset + weightsEntry.sizeBytes);\n        const nameToTensorMap = decodeWeights(byteBuffer, [weightsEntry.manifestEntry]);\n        for (const name in nameToTensorMap) {\n          weightsTensorMap[name] = nameToTensorMap[name];\n        }\n      });\n      bufferIndexOffset += numBuffers;\n    });\n    return weightsTensorMap;\n  };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/http.js\nvar OCTET_STREAM_MIME_TYPE = \"application/octet-stream\";\nvar JSON_TYPE = \"application/json\";\nvar HTTPRequest = class {\n  constructor(path, loadOptions) {\n    this.DEFAULT_METHOD = \"POST\";\n    if (loadOptions == null) {\n      loadOptions = {};\n    }\n    this.weightPathPrefix = loadOptions.weightPathPrefix;\n    this.onProgress = loadOptions.onProgress;\n    this.weightUrlConverter = loadOptions.weightUrlConverter;\n    if (loadOptions.fetchFunc != null) {\n      assert(typeof loadOptions.fetchFunc === \"function\", () => \"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)\");\n      this.fetch = loadOptions.fetchFunc;\n    } else {\n      this.fetch = env().platform.fetch;\n    }\n    assert(path != null && path.length > 0, () => \"URL path for http must not be null, undefined or empty.\");\n    if (Array.isArray(path)) {\n      assert(path.length === 2, () => `URL paths for http must have a length of 2, (actual length is ${path.length}).`);\n    }\n    this.path = path;\n    if (loadOptions.requestInit != null && loadOptions.requestInit.body != null) {\n      throw new Error(\"requestInit is expected to have no pre-existing body, but has one.\");\n    }\n    this.requestInit = loadOptions.requestInit || {};\n  }\n  async save(modelArtifacts) {\n    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {\n      throw new Error(\"BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.\");\n    }\n    const init2 = Object.assign({ method: this.DEFAULT_METHOD }, this.requestInit);\n    init2.body = new FormData();\n    const weightsManifest = [{\n      paths: [\"./model.weights.bin\"],\n      weights: modelArtifacts.weightSpecs\n    }];\n    const modelTopologyAndWeightManifest = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);\n    init2.body.append(\"model.json\", new Blob([JSON.stringify(modelTopologyAndWeightManifest)], { type: JSON_TYPE }), \"model.json\");\n    if (modelArtifacts.weightData != null) {\n      init2.body.append(\"model.weights.bin\", new Blob([modelArtifacts.weightData], { type: OCTET_STREAM_MIME_TYPE }), \"model.weights.bin\");\n    }\n    const response = await this.fetch(this.path, init2);\n    if (response.ok) {\n      return {\n        modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts),\n        responses: [response]\n      };\n    } else {\n      throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${response.status}.`);\n    }\n  }\n  async load() {\n    const modelConfigRequest = await this.fetch(this.path, this.requestInit);\n    if (!modelConfigRequest.ok) {\n      throw new Error(`Request to ${this.path} failed with status code ${modelConfigRequest.status}. Please verify this URL points to the model JSON of the model to load.`);\n    }\n    let modelJSON;\n    try {\n      modelJSON = await modelConfigRequest.json();\n    } catch (e) {\n      let message = `Failed to parse model JSON of response from ${this.path}.`;\n      if (this.path.endsWith(\".pb\")) {\n        message += \" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.\";\n      } else {\n        message += \" Please make sure the server is serving valid JSON for this request.\";\n      }\n      throw new Error(message);\n    }\n    const modelTopology = modelJSON.modelTopology;\n    const weightsManifest = modelJSON.weightsManifest;\n    if (modelTopology == null && weightsManifest == null) {\n      throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);\n    }\n    return getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));\n  }\n  async loadWeights(weightsManifest) {\n    const weightPath = Array.isArray(this.path) ? this.path[1] : this.path;\n    const [prefix, suffix] = parseUrl(weightPath);\n    const pathPrefix = this.weightPathPrefix || prefix;\n    const weightSpecs = getWeightSpecs(weightsManifest);\n    const fetchURLs = [];\n    const urlPromises = [];\n    for (const weightsGroup of weightsManifest) {\n      for (const path of weightsGroup.paths) {\n        if (this.weightUrlConverter != null) {\n          urlPromises.push(this.weightUrlConverter(path));\n        } else {\n          fetchURLs.push(pathPrefix + path + suffix);\n        }\n      }\n    }\n    if (this.weightUrlConverter) {\n      fetchURLs.push(...await Promise.all(urlPromises));\n    }\n    const buffers = await loadWeightsAsArrayBuffer(fetchURLs, {\n      requestInit: this.requestInit,\n      fetchFunc: this.fetch,\n      onProgress: this.onProgress\n    });\n    return [weightSpecs, concatenateArrayBuffers(buffers)];\n  }\n};\nHTTPRequest.URL_SCHEME_REGEX = /^https?:\\/\\//;\nfunction parseUrl(url) {\n  const lastSlash = url.lastIndexOf(\"/\");\n  const lastSearchParam = url.lastIndexOf(\"?\");\n  const prefix = url.substring(0, lastSlash);\n  const suffix = lastSearchParam > lastSlash ? url.substring(lastSearchParam) : \"\";\n  return [prefix + \"/\", suffix];\n}\nfunction isHTTPScheme(url) {\n  return url.match(HTTPRequest.URL_SCHEME_REGEX) != null;\n}\nvar httpRouter = (url, loadOptions) => {\n  if (typeof fetch === \"undefined\" && (loadOptions == null || loadOptions.fetchFunc == null)) {\n    return null;\n  } else {\n    let isHTTP = true;\n    if (Array.isArray(url)) {\n      isHTTP = url.every((urlItem) => isHTTPScheme(urlItem));\n    } else {\n      isHTTP = isHTTPScheme(url);\n    }\n    if (isHTTP) {\n      return http(url, loadOptions);\n    }\n  }\n  return null;\n};\nIORouterRegistry.registerSaveRouter(httpRouter);\nIORouterRegistry.registerLoadRouter(httpRouter);\nfunction http(path, loadOptions) {\n  return new HTTPRequest(path, loadOptions);\n}\nfunction browserHTTPRequest(path, loadOptions) {\n  return http(path, loadOptions);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/io/passthrough.js\nvar PassthroughLoader = class {\n  constructor(modelArtifacts) {\n    this.modelArtifacts = modelArtifacts;\n  }\n  load() {\n    return this.modelArtifacts;\n  }\n};\nvar PassthroughSaver = class {\n  constructor(saveHandler) {\n    this.saveHandler = saveHandler;\n  }\n  save(modelArtifacts) {\n    return this.saveHandler(modelArtifacts);\n  }\n};\nvar PassthroughAsync = class {\n  constructor(handler) {\n    if (handler.load) {\n      this.load = () => Promise.resolve(handler.load());\n    }\n    if (handler.save) {\n      this.save = (modelArtifacts) => Promise.resolve(handler.save(modelArtifacts));\n    }\n  }\n};\nfunction fromMemory(modelArtifacts, weightSpecs, weightData, trainingConfig) {\n  const args = arguments;\n  return new PassthroughAsync(fromMemorySync(...args));\n}\nfunction fromMemorySync(modelArtifacts, weightSpecs, weightData, trainingConfig) {\n  if (arguments.length === 1) {\n    const isModelArtifacts = modelArtifacts.modelTopology != null || modelArtifacts.weightSpecs != null;\n    if (isModelArtifacts) {\n      return new PassthroughLoader(modelArtifacts);\n    } else {\n      console.warn(\"Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.\");\n      return new PassthroughLoader({ modelTopology: modelArtifacts });\n    }\n  } else {\n    console.warn(\"Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.\");\n    return new PassthroughLoader({\n      modelTopology: modelArtifacts,\n      weightSpecs,\n      weightData,\n      trainingConfig\n    });\n  }\n}\nfunction withSaveHandler(saveHandler) {\n  return new PassthroughSaver(saveHandler);\n}\nfunction withSaveHandlerSync(saveHandler) {\n  return new PassthroughSaver(saveHandler);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/math.js\nvar math_exports = {};\n__export(math_exports, {\n  confusionMatrix: () => confusionMatrix\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/mat_mul.js\nfunction matMul_(a, b, transposeA = false, transposeB = false) {\n  let $a = convertToTensor(a, \"a\", \"matMul\");\n  let $b = convertToTensor(b, \"b\", \"matMul\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  const attrs = { transposeA, transposeB };\n  return ENGINE.runKernel(BatchMatMul, inputs, attrs);\n}\nvar matMul = op({ matMul_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/one_hot.js\nfunction oneHot_(indices, depth, onValue = 1, offValue = 0, dtype = \"int32\") {\n  if (depth < 2) {\n    throw new Error(`Error in oneHot: depth must be >=2, but it is ${depth}`);\n  }\n  const $indices = convertToTensor(indices, \"indices\", \"oneHot\", \"int32\");\n  const inputs = { indices: $indices };\n  const attrs = { dtype, depth, onValue, offValue };\n  return ENGINE.runKernel(OneHot, inputs, attrs);\n}\nvar oneHot = op({ oneHot_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/globals.js\nfunction enableProdMode() {\n  env().set(\"PROD\", true);\n}\nfunction enableDebugMode() {\n  env().set(\"DEBUG\", true);\n}\nfunction disableDeprecationWarnings() {\n  env().set(\"DEPRECATION_WARNINGS_ENABLED\", false);\n  console.warn(`TensorFlow.js deprecation warnings have been disabled.`);\n}\nfunction deprecationWarn(msg) {\n  if (env().getBool(\"DEPRECATION_WARNINGS_ENABLED\")) {\n    console.warn(msg + \" You can disable deprecation warnings with tf.disableDeprecationWarnings().\");\n  }\n}\nsetDeprecationWarningFn(deprecationWarn);\nfunction disposeVariables() {\n  ENGINE.disposeVariables();\n}\nfunction engine() {\n  return ENGINE;\n}\nfunction memory() {\n  return ENGINE.memory();\n}\nfunction profile(f) {\n  return ENGINE.profile(f);\n}\nfunction tidy(nameOrFn, fn) {\n  return ENGINE.tidy(nameOrFn, fn);\n}\nfunction dispose(container) {\n  const tensors = getTensorsInContainer(container);\n  tensors.forEach((tensor2) => tensor2.dispose());\n}\nfunction keep(result) {\n  return ENGINE.keep(result);\n}\nfunction time(f) {\n  return ENGINE.time(f);\n}\nfunction setBackend(backendName) {\n  return ENGINE.setBackend(backendName);\n}\nfunction ready() {\n  return ENGINE.ready();\n}\nfunction getBackend() {\n  return ENGINE.backendName;\n}\nfunction removeBackend(name) {\n  ENGINE.removeBackend(name);\n}\nfunction findBackend(name) {\n  return ENGINE.findBackend(name);\n}\nfunction findBackendFactory(name) {\n  return ENGINE.findBackendFactory(name);\n}\nfunction registerBackend(name, factory, priority = 1) {\n  return ENGINE.registerBackend(name, factory, priority);\n}\nfunction backend() {\n  return ENGINE.backend;\n}\nfunction setPlatform(platformName, platform) {\n  env().setPlatform(platformName, platform);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/imag.js\nfunction imag_(input2) {\n  const $input = convertToTensor(input2, \"input\", \"imag\");\n  const inputs = { input: $input };\n  return ENGINE.runKernel(Imag, inputs);\n}\nvar imag = op({ imag_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/neg.js\nfunction neg_(x) {\n  const $x = convertToTensor(x, \"x\", \"neg\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Neg, inputs);\n}\nvar neg = op({ neg_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/real.js\nfunction real_(input2) {\n  const $input = convertToTensor(input2, \"input\", \"real\");\n  const inputs = { input: $input };\n  return ENGINE.runKernel(Real, inputs);\n}\nvar real = op({ real_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/transpose.js\nfunction transpose_(x, perm, conjugate) {\n  const $x = convertToTensor(x, \"x\", \"transpose\");\n  if (perm == null) {\n    perm = $x.shape.map((s, i) => i).reverse();\n  }\n  assert($x.rank === perm.length, () => `Error in transpose: rank of input ${$x.rank} must match length of perm ${perm}.`);\n  perm.forEach((axis) => {\n    assert(axis >= 0 && axis < $x.rank, () => `All entries in 'perm' must be between 0 and ${$x.rank - 1} but got ${perm}`);\n  });\n  if ($x.rank <= 1) {\n    return $x.clone();\n  }\n  const inputs = { x: $x };\n  const attrs = { perm };\n  if ($x.dtype === \"complex64\") {\n    return tidy(() => {\n      let $real = real($x);\n      let $imag = imag($x);\n      $real = ENGINE.runKernel(Transpose, { x: $real }, attrs);\n      $imag = ENGINE.runKernel(Transpose, { x: $imag }, attrs);\n      if (conjugate) {\n        $imag = neg($imag);\n      }\n      return complex($real, $imag);\n    });\n  }\n  return ENGINE.runKernel(Transpose, inputs, attrs);\n}\nvar transpose = op({ transpose_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/confusion_matrix.js\nfunction confusionMatrix_(labels, predictions, numClasses) {\n  const $labels = convertToTensor(labels, \"labels\", \"confusionMatrix\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"confusionMatrix\");\n  assert(numClasses == null || numClasses > 0 && Number.isInteger(numClasses), () => `If provided, numClasses must be a positive integer, but got ${numClasses}`);\n  assert($labels.rank === 1, () => `Expected the rank of labels to be 1, but got ${$labels.rank}`);\n  assert($predictions.rank === 1, () => `Expected the rank of predictions to be 1, but got ${$predictions.rank}`);\n  assert($labels.shape[0] === $predictions.shape[0], () => `Mismatch in the number of examples: ${$labels.shape[0]} vs. ${$predictions.shape[0]}. Labels and predictions should have the same number of elements.`);\n  assert(numClasses > 0 && Number.isInteger(numClasses), () => `numClasses is required to be a positive integer, but got ${numClasses}`);\n  const oneHotLabels = oneHot(cast($labels, \"int32\"), numClasses);\n  const oneHotPredictions = oneHot(cast($predictions, \"int32\"), numClasses);\n  const oneHotLabelsT = transpose(oneHotLabels);\n  const product = matMul(oneHotLabelsT, oneHotPredictions);\n  return cast(product, \"int32\");\n}\nvar confusionMatrix = op({ confusionMatrix_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_util.js\nvar broadcast_util_exports = {};\n__export(broadcast_util_exports, {\n  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,\n  getBroadcastDims: () => getBroadcastDims,\n  getReductionAxes: () => getReductionAxes\n});\nfunction getBroadcastDims(inShape, outShape) {\n  const inRank = inShape.length;\n  const dims = [];\n  for (let i = 0; i < inRank; i++) {\n    const dim = inRank - 1 - i;\n    const a = inShape[dim] || 1;\n    const b = outShape[outShape.length - 1 - i] || 1;\n    if (b > 1 && a === 1) {\n      dims.unshift(dim);\n    }\n  }\n  return dims;\n}\nfunction getReductionAxes(inShape, outShape) {\n  const result = [];\n  for (let i = 0; i < outShape.length; i++) {\n    const inDim = inShape[inShape.length - i - 1];\n    const outAxis = outShape.length - i - 1;\n    const outDim = outShape[outAxis];\n    if (inDim == null || inDim === 1 && outDim > 1) {\n      result.unshift(outAxis);\n    }\n  }\n  return result;\n}\nfunction assertAndGetBroadcastShape(shapeA, shapeB) {\n  const result = [];\n  const l = Math.max(shapeA.length, shapeB.length);\n  for (let i = 0; i < l; i++) {\n    let a = shapeA[shapeA.length - i - 1];\n    if (a == null) {\n      a = 1;\n    }\n    let b = shapeB[shapeB.length - i - 1];\n    if (b == null) {\n      b = 1;\n    }\n    if (a === 1) {\n      result.unshift(b);\n    } else if (b === 1) {\n      result.unshift(a);\n    } else if (a !== b) {\n      const errMsg = `Operands could not be broadcast together with shapes ${shapeA} and ${shapeB}.`;\n      throw Error(errMsg);\n    } else {\n      result.unshift(a);\n    }\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/browser.js\nvar browser_exports = {};\n__export(browser_exports, {\n  fromPixels: () => fromPixels,\n  fromPixelsAsync: () => fromPixelsAsync,\n  toPixels: () => toPixels\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor3d.js\nfunction tensor3d(values, shape, dtype) {\n  assertNonNull(values);\n  if (shape != null && shape.length !== 3) {\n    throw new Error(\"tensor3d() requires shape to have three numbers\");\n  }\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 3 && inferredShape.length !== 1) {\n    throw new Error(\"tensor3d() requires values to be number[][][] or flat/TypedArray\");\n  }\n  if (inferredShape.length === 1 && shape == null) {\n    throw new Error(\"tensor3d() requires shape to be provided when `values` are a flat array\");\n  }\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/browser.js\nvar fromPixels2DContext;\nfunction fromPixels_(pixels, numChannels = 3) {\n  if (numChannels > 4) {\n    throw new Error(\"Cannot construct Tensor with more than 4 channels from pixels.\");\n  }\n  if (pixels == null) {\n    throw new Error(\"pixels passed to tf.browser.fromPixels() can not be null\");\n  }\n  let isPixelData2 = false;\n  let isImageData = false;\n  let isVideo = false;\n  let isImage = false;\n  let isCanvasLike = false;\n  let isImageBitmap = false;\n  if (pixels.data instanceof Uint8Array) {\n    isPixelData2 = true;\n  } else if (typeof ImageData !== \"undefined\" && pixels instanceof ImageData) {\n    isImageData = true;\n  } else if (typeof HTMLVideoElement !== \"undefined\" && pixels instanceof HTMLVideoElement) {\n    isVideo = true;\n  } else if (typeof HTMLImageElement !== \"undefined\" && pixels instanceof HTMLImageElement) {\n    isImage = true;\n  } else if (pixels.getContext != null) {\n    isCanvasLike = true;\n  } else if (typeof ImageBitmap !== \"undefined\" && pixels instanceof ImageBitmap) {\n    isImageBitmap = true;\n  } else {\n    throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${pixels.constructor.name}`);\n  }\n  const kernel = getKernel(FromPixels, ENGINE.backendName);\n  if (kernel != null) {\n    const inputs = { pixels };\n    const attrs = { numChannels };\n    return ENGINE.runKernel(FromPixels, inputs, attrs);\n  }\n  const [width, height] = isVideo ? [\n    pixels.videoWidth,\n    pixels.videoHeight\n  ] : [pixels.width, pixels.height];\n  let vals;\n  if (isCanvasLike) {\n    vals = pixels.getContext(\"2d\").getImageData(0, 0, width, height).data;\n  } else if (isImageData || isPixelData2) {\n    vals = pixels.data;\n  } else if (isImage || isVideo || isImageBitmap) {\n    if (fromPixels2DContext == null) {\n      if (typeof document === \"undefined\") {\n        if (typeof OffscreenCanvas !== \"undefined\" && typeof OffscreenCanvasRenderingContext2D !== \"undefined\") {\n          fromPixels2DContext = new OffscreenCanvas(1, 1).getContext(\"2d\");\n        } else {\n          throw new Error(\"Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.\");\n        }\n      } else {\n        fromPixels2DContext = document.createElement(\"canvas\").getContext(\"2d\", { willReadFrequently: true });\n      }\n    }\n    fromPixels2DContext.canvas.width = width;\n    fromPixels2DContext.canvas.height = height;\n    fromPixels2DContext.drawImage(pixels, 0, 0, width, height);\n    vals = fromPixels2DContext.getImageData(0, 0, width, height).data;\n  }\n  let values;\n  if (numChannels === 4) {\n    values = new Int32Array(vals);\n  } else {\n    const numPixels = width * height;\n    values = new Int32Array(numPixels * numChannels);\n    for (let i = 0; i < numPixels; i++) {\n      for (let channel = 0; channel < numChannels; ++channel) {\n        values[i * numChannels + channel] = vals[i * 4 + channel];\n      }\n    }\n  }\n  const outShape = [height, width, numChannels];\n  return tensor3d(values, outShape, \"int32\");\n}\nfunction isPixelData(pixels) {\n  return pixels != null && pixels.data instanceof Uint8Array;\n}\nfunction isImageBitmapFullySupported() {\n  return typeof window !== \"undefined\" && typeof ImageBitmap !== \"undefined\" && window.hasOwnProperty(\"createImageBitmap\");\n}\nfunction isNonEmptyPixels(pixels) {\n  return pixels != null && pixels.width !== 0 && pixels.height !== 0;\n}\nfunction canWrapPixelsToImageBitmap(pixels) {\n  return isImageBitmapFullySupported() && !(pixels instanceof ImageBitmap) && isNonEmptyPixels(pixels) && !isPixelData(pixels);\n}\nasync function fromPixelsAsync(pixels, numChannels = 3) {\n  let inputs = null;\n  if (env().getBool(\"WRAP_TO_IMAGEBITMAP\") && canWrapPixelsToImageBitmap(pixels)) {\n    let imageBitmap;\n    try {\n      imageBitmap = await createImageBitmap(pixels, { premultiplyAlpha: \"none\" });\n    } catch (e) {\n      imageBitmap = null;\n    }\n    if (imageBitmap != null && imageBitmap.width === pixels.width && imageBitmap.height === pixels.height) {\n      inputs = imageBitmap;\n    } else {\n      inputs = pixels;\n    }\n  } else {\n    inputs = pixels;\n  }\n  return fromPixels_(inputs, numChannels);\n}\nasync function toPixels(img, canvas) {\n  let $img = convertToTensor(img, \"img\", \"toPixels\");\n  if (!(img instanceof Tensor)) {\n    const originalImgTensor = $img;\n    $img = cast(originalImgTensor, \"int32\");\n    originalImgTensor.dispose();\n  }\n  if ($img.rank !== 2 && $img.rank !== 3) {\n    throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${$img.rank}.`);\n  }\n  const [height, width] = $img.shape.slice(0, 2);\n  const depth = $img.rank === 2 ? 1 : $img.shape[2];\n  if (depth > 4 || depth === 2) {\n    throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${depth}`);\n  }\n  if ($img.dtype !== \"float32\" && $img.dtype !== \"int32\") {\n    throw new Error(`Unsupported type for toPixels: ${$img.dtype}. Please use float32 or int32 tensors.`);\n  }\n  const data = await $img.data();\n  const multiplier = $img.dtype === \"float32\" ? 255 : 1;\n  const bytes = new Uint8ClampedArray(width * height * 4);\n  for (let i = 0; i < height * width; ++i) {\n    const rgba = [0, 0, 0, 255];\n    for (let d = 0; d < depth; d++) {\n      const value = data[i * depth + d];\n      if ($img.dtype === \"float32\") {\n        if (value < 0 || value > 1) {\n          throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${value}.`);\n        }\n      } else if ($img.dtype === \"int32\") {\n        if (value < 0 || value > 255) {\n          throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${value}.`);\n        }\n      }\n      if (depth === 1) {\n        rgba[0] = value * multiplier;\n        rgba[1] = value * multiplier;\n        rgba[2] = value * multiplier;\n      } else {\n        rgba[d] = value * multiplier;\n      }\n    }\n    const j = i * 4;\n    bytes[j + 0] = Math.round(rgba[0]);\n    bytes[j + 1] = Math.round(rgba[1]);\n    bytes[j + 2] = Math.round(rgba[2]);\n    bytes[j + 3] = Math.round(rgba[3]);\n  }\n  if (canvas != null) {\n    canvas.width = width;\n    canvas.height = height;\n    const ctx = canvas.getContext(\"2d\");\n    const imageData = new ImageData(bytes, width, height);\n    ctx.putImageData(imageData, 0, 0);\n  }\n  if ($img !== img) {\n    $img.dispose();\n  }\n  return bytes;\n}\nvar fromPixels = op({ fromPixels_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/gather_nd_util.js\nvar gather_nd_util_exports = {};\n__export(gather_nd_util_exports, {\n  prepareAndValidate: () => prepareAndValidate\n});\nfunction prepareAndValidate(tensor2, indices) {\n  const tensorRank = tensor2.shape.length;\n  const indicesRank = indices.shape.length;\n  if (tensorRank < 1) {\n    throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${tensorRank}.`);\n  }\n  if (indicesRank < 1) {\n    throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${indicesRank}.`);\n  }\n  if (indices.dtype !== \"int32\") {\n    throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${indices.dtype}.`);\n  }\n  if (indices.shape[indicesRank - 1] > tensorRank) {\n    throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${indices.shape[indicesRank - 1]} vs. ${tensorRank}`);\n  }\n  if (sizeFromShape(tensor2.shape) === 0) {\n    throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${tensor2.shape}.`);\n  }\n  const indicesShape = indices.shape;\n  const sliceRank = indicesShape[indicesShape.length - 1];\n  let nResult = 1;\n  for (let i = 0; i < indicesShape.length - 1; ++i) {\n    nResult *= indicesShape[i];\n  }\n  const inputShape = tensor2.shape;\n  const resultShape = indicesShape.slice();\n  resultShape.pop();\n  let sliceSize = 1;\n  for (let i = sliceRank; i < tensorRank; ++i) {\n    sliceSize *= inputShape[i];\n    resultShape.push(inputShape[i]);\n  }\n  const strides = [\n    ...computeStrides(tensor2.shape).map((stride) => stride / sliceSize),\n    1\n  ].slice(0, sliceRank);\n  return [resultShape, nResult, sliceSize, strides];\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/scatter_nd_util.js\nvar scatter_nd_util_exports = {};\n__export(scatter_nd_util_exports, {\n  calculateShapes: () => calculateShapes,\n  validateInput: () => validateInput,\n  validateUpdateShape: () => validateUpdateShape\n});\nfunction validateUpdateShape(shape, indices, updates) {\n  const sliceDim = indices.rank > 1 ? indices.shape[indices.rank - 1] : 1;\n  const batchDim = indices.rank > 1 ? indices.rank - 1 : 1;\n  const shapeError = `Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${updates.shape}, indices.shape: ${indices.shape}, shape: ${shape}, sliceDim: ${sliceDim}, and batchDim: ${batchDim}.`;\n  if (updates.rank < batchDim) {\n    throw new Error(shapeError + ` update.rank < ${batchDim}. `);\n  }\n  if (shape.length < sliceDim + (updates.rank - batchDim)) {\n    throw new Error(shapeError + ` Output shape length < ${sliceDim + (updates.rank - batchDim)}`);\n  }\n  if (updates.rank !== batchDim + shape.length - sliceDim) {\n    throw new Error(shapeError + ` update.rank != ${batchDim + shape.length - sliceDim}`);\n  }\n  for (let d = 0; d < batchDim; ++d) {\n    if (updates.shape[d] !== indices.shape[d]) {\n      throw new Error(shapeError + ` updates.shape[${d}] (${updates.shape[d]}) != indices.shape[${d}] (${indices.shape[d]}).`);\n    }\n  }\n  for (let d = 0; d < updates.rank - batchDim; ++d) {\n    if (updates.shape[d + batchDim] !== shape[d + sliceDim]) {\n      throw new Error(shapeError + ` updates.shape[${d + batchDim}] (${updates.shape[d + batchDim]}) != shape[${d + batchDim}] (${shape[d + batchDim]})`);\n    }\n  }\n}\nfunction validateInput(updates, indices, shape) {\n  if (indices.rank < 1) {\n    throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${indices.rank}.`);\n  }\n  if (updates.rank < 1) {\n    throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${updates.rank}.`);\n  }\n  if (indices.dtype !== \"int32\") {\n    throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${indices.dtype}`);\n  }\n  if (shape.length < 1) {\n    throw new Error(`Output rank must be greater or equal to 1, but got shape: ${shape}`);\n  }\n  if (shape.length === 0) {\n    if (indices.size === 0) {\n      throw new Error(`Indices specified for empty output. indices shape: ${indices.shape}`);\n    }\n    if (updates.size === 0) {\n      throw new Error(`Updates specified for empty output. updates shape: ${updates.shape}`);\n    }\n  }\n  validateUpdateShape(shape, indices, updates);\n}\nfunction calculateShapes(updates, indices, shape) {\n  const indicesRank = indices.shape.length;\n  const sliceRank = indicesRank > 1 ? indices.shape[indicesRank - 1] : 1;\n  const totalNd = shape.length;\n  let sliceSize = 1;\n  for (let i = sliceRank; i < totalNd; ++i) {\n    sliceSize *= shape[i];\n  }\n  const safeSliceDim = sliceRank < 1 ? 1 : sliceRank;\n  const numUpdates = sizeFromShape(indices.shape) / safeSliceDim;\n  const strides = [...computeStrides(shape.slice(0, sliceRank)), 1];\n  const outputSize = sizeFromShape(shape);\n  return { sliceRank, numUpdates, sliceSize, strides, outputSize };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice_util.js\nvar slice_util_exports = {};\n__export(slice_util_exports, {\n  assertParamsValid: () => assertParamsValid,\n  computeFlatOffset: () => computeFlatOffset,\n  computeOutShape: () => computeOutShape,\n  getNormalizedAxes: () => getNormalizedAxes,\n  isSliceContinous: () => isSliceContinous,\n  maskToAxes: () => maskToAxes,\n  parseSliceParams: () => parseSliceParams,\n  sliceInfo: () => sliceInfo,\n  startForAxis: () => startForAxis,\n  startIndicesWithElidedDims: () => startIndicesWithElidedDims,\n  stopForAxis: () => stopForAxis,\n  stopIndicesWithElidedDims: () => stopIndicesWithElidedDims,\n  stridesForAxis: () => stridesForAxis,\n  stridesWithElidedDims: () => stridesWithElidedDims\n});\nvar NEW_AXIS = -2;\nvar SHRINK_AXIS = -1;\nfunction assertParamsValid(input2, begin, size) {\n  const inputRank = input2.shape.length;\n  assert(inputRank === begin.length, () => `Error in slice${inputRank}D: Length of begin ${begin} must match the rank of the array (${inputRank}).`);\n  assert(inputRank === size.length, () => `Error in slice${inputRank}D: Length of size ${size} must match the rank of the array (${inputRank}).`);\n  for (let i = 0; i < inputRank; ++i) {\n    assert(begin[i] + size[i] <= input2.shape[i], () => `Error in slice${inputRank}D: begin[${i}] + size[${i}] (${begin[i] + size[i]}) would overflow input.shape[${i}] (${input2.shape[i]})`);\n  }\n}\nfunction maskToAxes(mask) {\n  const axes = [];\n  let axis = 0;\n  while (mask > 0) {\n    if (mask & 1) {\n      axes.push(axis);\n    }\n    mask /= 2;\n    axis++;\n  }\n  return axes;\n}\nfunction computeOutShape(begin, end, strides) {\n  const size = [];\n  for (let axis = 0; axis < begin.length; axis++) {\n    size[axis] = Math.ceil((end[axis] - begin[axis]) / strides[axis]);\n  }\n  return size;\n}\nfunction stridesWithElidedDims(strides, ellipsisInsertionIndex, numElidedAxes, inputShape) {\n  const newStrides = [...strides];\n  for (let i = newStrides.length; i < inputShape.length; i++) {\n    newStrides.push(1);\n  }\n  for (let i = 0; i < numElidedAxes; i++) {\n    if (i === 0) {\n      newStrides[ellipsisInsertionIndex] = 1;\n    } else {\n      newStrides.splice(ellipsisInsertionIndex, 0, 1);\n      newStrides.pop();\n    }\n  }\n  return newStrides;\n}\nfunction unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, normalizedAxis) {\n  if (normalizedAxis <= ellipsisInsertionIndex) {\n    return normalizedAxis;\n  }\n  return normalizedAxis - (numElidedAxes - 1);\n}\nfunction getElidedAxes(numElidedAxes, ellipsisInsertionIndex) {\n  const elidedAxes = [];\n  for (let i = 0; i < numElidedAxes; i++) {\n    elidedAxes.push(ellipsisInsertionIndex + i);\n  }\n  return elidedAxes;\n}\nfunction getNormalizedAxes(inputShape, ellipsisAxes, numInterpolatedAxes, begin, end, strides, beginMask, endMask, ellipsisMask) {\n  const inputRank = inputShape.length;\n  let normalizedBegin = new Array(inputRank), normalizedEnd = new Array(inputRank), normalizedStrides = new Array(inputRank);\n  if (ellipsisAxes.length && numInterpolatedAxes > 0) {\n    const fullIndex = ellipsisAxes[0];\n    const numElidedAxes = numInterpolatedAxes + 1;\n    normalizedBegin = startIndicesWithElidedDims(beginMask, fullIndex, numElidedAxes, begin, inputShape);\n    normalizedEnd = stopIndicesWithElidedDims(endMask, fullIndex, numElidedAxes, end, inputShape);\n    normalizedStrides = stridesWithElidedDims(strides, fullIndex, numElidedAxes, inputShape);\n  } else {\n    for (let axis = 0; axis < inputRank; axis++) {\n      normalizedBegin[axis] = startForAxis(beginMask, begin, strides, inputShape, axis, ellipsisMask);\n      normalizedEnd[axis] = stopForAxis(endMask, end, strides, inputShape, axis, ellipsisMask);\n      normalizedStrides[axis] = stridesForAxis(strides, axis, ellipsisMask);\n    }\n  }\n  return {\n    begin: normalizedBegin,\n    end: normalizedEnd,\n    strides: normalizedStrides\n  };\n}\nfunction startIndicesWithElidedDims(beginMask, ellipsisInsertionIndex, numElidedAxes, originalBegin, inputShape) {\n  const newIndices = [...inputShape];\n  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);\n  for (let axis = 0; axis < newIndices.length; axis++) {\n    if (elidedAxes.indexOf(axis) > -1) {\n      newIndices[axis] = 0;\n    } else {\n      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);\n      let originalValue = originalBegin[originalAxis];\n      if (beginMask & 1 << originalAxis) {\n        originalValue = 0;\n      }\n      newIndices[axis] = originalValue;\n    }\n  }\n  return newIndices;\n}\nfunction stopIndicesWithElidedDims(endMask, ellipsisInsertionIndex, numElidedAxes, originalEnd, inputShape) {\n  const newIndices = [...inputShape];\n  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);\n  for (let axis = 0; axis < newIndices.length; axis++) {\n    if (elidedAxes.indexOf(axis) > -1) {\n      newIndices[axis] = Number.MAX_SAFE_INTEGER;\n    } else {\n      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);\n      let originalValue = originalEnd[originalAxis];\n      if (endMask & 1 << originalAxis) {\n        originalValue = Number.MAX_SAFE_INTEGER;\n      }\n      newIndices[axis] = originalValue;\n    }\n  }\n  for (let i = 0; i < newIndices.length; i++) {\n    const axisSize = inputShape[i];\n    if (newIndices[i] < 0) {\n      newIndices[i] += axisSize;\n    }\n    newIndices[i] = clamp(0, newIndices[i], inputShape[i]);\n  }\n  return newIndices;\n}\nfunction stridesForAxis(strides, axis, ellipsisMask) {\n  let stride = strides[axis];\n  if (ellipsisMask & 1 << axis || stride == null) {\n    stride = 1;\n  }\n  return stride;\n}\nfunction startForAxis(beginMask, startIndices, strides, inputShape, axis, ellipsisMask) {\n  let start = startIndices[axis];\n  const stride = strides[axis] || 1;\n  if (beginMask & 1 << axis || ellipsisMask & 1 << axis || start == null) {\n    if (stride > 0) {\n      start = Number.MIN_SAFE_INTEGER;\n    } else {\n      start = Number.MAX_SAFE_INTEGER;\n    }\n  }\n  const axisSize = inputShape[axis];\n  if (start < 0) {\n    start += axisSize;\n  }\n  start = clamp(0, start, axisSize - 1);\n  return start;\n}\nfunction stopForAxis(endMask, stopIndices, strides, inputShape, axis, ellipsisMask) {\n  let stop = stopIndices[axis];\n  const stride = strides[axis] || 1;\n  if (endMask & 1 << axis || ellipsisMask & 1 << axis || stop == null) {\n    if (stride > 0) {\n      stop = Number.MAX_SAFE_INTEGER;\n    } else {\n      stop = Number.MIN_SAFE_INTEGER;\n    }\n  }\n  const axisSize = inputShape[axis];\n  if (stop < 0) {\n    stop += axisSize;\n  }\n  if (stride > 0) {\n    stop = clamp(0, stop, axisSize);\n  } else {\n    stop = clamp(-1, stop, axisSize - 1);\n  }\n  return stop;\n}\nfunction isSliceContinous(shape, begin, size) {\n  let firstNonOneAxis = size.length;\n  for (let i = 0; i < size.length; i++) {\n    if (size[i] > 1) {\n      firstNonOneAxis = i;\n      break;\n    }\n  }\n  for (let i = firstNonOneAxis + 1; i < size.length; i++) {\n    if (begin[i] > 0 || size[i] !== shape[i]) {\n      return false;\n    }\n  }\n  return true;\n}\nfunction computeFlatOffset(begin, strides) {\n  let flatOffset = begin.length > 0 ? begin[begin.length - 1] : 1;\n  for (let i = 0; i < begin.length - 1; i++) {\n    flatOffset += begin[i] * strides[i];\n  }\n  return flatOffset;\n}\nfunction parseSliceParams(x, begin, size) {\n  let begin_;\n  const xRank = x.shape.length;\n  if (typeof begin === \"number\") {\n    begin_ = [begin, ...new Array(xRank - 1).fill(0)];\n  } else if (begin.length < xRank) {\n    begin_ = begin.concat(new Array(xRank - begin.length).fill(0));\n  } else {\n    begin_ = begin.slice();\n  }\n  begin_.forEach((d) => {\n    assert(d !== -1, () => \"slice() does not support negative begin indexing.\");\n  });\n  let size_;\n  if (size == null) {\n    size_ = new Array(xRank).fill(-1);\n  } else if (typeof size === \"number\") {\n    size_ = [size, ...new Array(xRank - 1).fill(-1)];\n  } else if (size.length < xRank) {\n    size_ = size.concat(new Array(xRank - size.length).fill(-1));\n  } else {\n    size_ = size;\n  }\n  size_ = size_.map((d, i) => {\n    if (d >= 0) {\n      return d;\n    } else {\n      assert(d === -1, () => `Negative size values should be exactly -1 but got ${d} for the slice() size at index ${i}.`);\n      return x.shape[i] - begin_[i];\n    }\n  });\n  return [begin_, size_];\n}\nfunction sliceInfo(xShape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {\n  let stridesNonNull;\n  if (strides == null) {\n    stridesNonNull = new Array(begin.length);\n    stridesNonNull.fill(1);\n  } else {\n    stridesNonNull = strides;\n  }\n  if (ellipsisMask != null && (ellipsisMask & ellipsisMask - 1) !== 0) {\n    throw new Error(\"Multiple ellipses in slice is not allowed.\");\n  }\n  let ellipsisSeen = false;\n  const sparseSpec = {\n    dims: stridesNonNull.length,\n    numAddAxisAfterEllipsis: 0,\n    begin: begin.slice(),\n    end: end.slice(),\n    strides: stridesNonNull.slice(),\n    beginMask,\n    endMask,\n    ellipsisMask,\n    newAxisMask,\n    shrinkAxisMask\n  };\n  for (let i = 0; i < sparseSpec.dims; i++) {\n    if (ellipsisSeen && (1 << i & newAxisMask) !== 0) {\n      sparseSpec.numAddAxisAfterEllipsis++;\n    }\n    if (1 << i & ellipsisMask) {\n      ellipsisSeen = true;\n    }\n  }\n  if (!ellipsisSeen) {\n    sparseSpec.ellipsisMask |= 1 << sparseSpec.dims;\n    sparseSpec.dims++;\n  }\n  const denseSpec = {\n    dims: xShape.length,\n    beginMask: 0,\n    endMask: 0,\n    beginValid: false,\n    endValid: false\n  };\n  buildDenseSpec(sparseSpec, denseSpec);\n  let isIdentity = true;\n  let sliceDim0 = true;\n  let isSimpleSlice = true;\n  const processingShape = [];\n  const finalShape = [];\n  for (let i = 0; i < xShape.length; ++i) {\n    if (denseSpec.strides[i] === 0) {\n      throw Error(`strides[${i}] must be non-zero`);\n    }\n    const shrinkI = !!(denseSpec.shrinkAxisMask & 1 << i);\n    const dimI = xShape[i];\n    if (dimI === -1) {\n      processingShape.push(shrinkI ? 1 : -1);\n      continue;\n    }\n    const masks = [denseSpec.beginMask & 1 << i, denseSpec.endMask & 1 << i];\n    const validRange = [\n      denseSpec.strides[i] > 0 ? 0 : -1,\n      denseSpec.strides[i] > 0 ? dimI : dimI - 1\n    ];\n    if (shrinkI && denseSpec.strides[i] <= 0) {\n      throw Error(\"only stride 1 allowed on non-range indexing.\");\n    }\n    isSimpleSlice = isSimpleSlice && denseSpec.strides[i] === 1;\n    const beginAndEndMasked = !!(denseSpec.beginMask & 1 << i && denseSpec.endMask & 1 << i);\n    if (denseSpec.beginValid && denseSpec.endValid) {\n      if (shrinkI) {\n        const xFwd = denseSpec.begin[i] < 0 ? dimI + denseSpec.begin[i] : denseSpec.begin[i];\n        denseSpec.begin[i] = xFwd;\n        denseSpec.end[i] = denseSpec.begin[i] + 1;\n        if (xFwd < 0 || xFwd >= dimI) {\n          throw Error(`slice index ${denseSpec.begin[i]} of dimension ${i} out of bounds.`);\n        }\n      } else {\n        denseSpec.begin[i] = canonical(denseSpec.begin[i], 0, denseSpec.strides[i], dimI, masks, validRange);\n        denseSpec.end[i] = canonical(denseSpec.end[i], 1, denseSpec.strides[i], dimI, masks, validRange);\n      }\n      const takeAllInDimension = denseSpec.strides[i] === 1 && denseSpec.begin[i] === 0 && denseSpec.end[i] === dimI;\n      isIdentity = isIdentity && takeAllInDimension;\n      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || takeAllInDimension);\n    } else {\n      isIdentity = isIdentity && (denseSpec.strides[i] === 1 && beginAndEndMasked);\n      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || beginAndEndMasked);\n    }\n    let intervalLength;\n    let knownInterval = false;\n    if (denseSpec.beginValid && denseSpec.endValid) {\n      intervalLength = denseSpec.end[i] - denseSpec.begin[i];\n      knownInterval = true;\n    } else if (shrinkI) {\n      intervalLength = 1;\n      knownInterval = true;\n    } else if (beginAndEndMasked) {\n      if (dimI >= 0) {\n        if (denseSpec.strides[i] < 0) {\n          intervalLength = -dimI;\n        } else {\n          intervalLength = dimI;\n        }\n        knownInterval = true;\n      }\n    }\n    if (knownInterval) {\n      let sizeI;\n      if (intervalLength === 0 || intervalLength < 0 !== denseSpec.strides[i] < 0) {\n        sizeI = 0;\n      } else {\n        sizeI = Math.trunc(intervalLength / denseSpec.strides[i]) + (intervalLength % denseSpec.strides[i] !== 0 ? 1 : 0);\n      }\n      processingShape.push(sizeI);\n    } else {\n      processingShape.push(-1);\n    }\n  }\n  for (let denseDim = 0; denseDim < denseSpec.finalShapeGatherIndices.length; ++denseDim) {\n    const gatherIndex = denseSpec.finalShapeGatherIndices[denseDim];\n    if (gatherIndex >= 0) {\n      finalShape.push(processingShape[gatherIndex]);\n    } else if (gatherIndex === NEW_AXIS) {\n      finalShape.push(1);\n    }\n  }\n  const finalShapeSparse = finalShape.filter((dim, i) => denseSpec.finalShapeGatherIndices[i] !== NEW_AXIS);\n  return {\n    finalShapeSparse,\n    finalShape,\n    isIdentity,\n    sliceDim0,\n    isSimpleSlice,\n    begin: denseSpec.begin,\n    end: denseSpec.end,\n    strides: denseSpec.strides\n  };\n}\nfunction buildDenseSpec(sparse2, dense2) {\n  dense2.beginMask = 0;\n  dense2.endMask = 0;\n  dense2.shrinkAxisMask = 0;\n  let fullIndex = 0;\n  dense2.beginValid = sparse2.begin != null;\n  dense2.endValid = sparse2.end != null;\n  dense2.begin = new Array(dense2.dims);\n  dense2.end = new Array(dense2.dims);\n  dense2.strides = new Array(dense2.dims);\n  dense2.finalShapeGatherIndices = [];\n  dense2.finalShapeGatherIndicesSparse = [];\n  dense2.inputShapeGatherIndicesSparse = new Array(dense2.dims);\n  for (let i = 0; i < sparse2.dims; i++) {\n    if (1 << i & sparse2.ellipsisMask) {\n      const nextIndex = Math.min(dense2.dims - (sparse2.dims - i) + 1 + sparse2.numAddAxisAfterEllipsis, dense2.dims);\n      for (; fullIndex < nextIndex; fullIndex++) {\n        dense2.begin[fullIndex] = 0;\n        dense2.end[fullIndex] = 0;\n        dense2.strides[fullIndex] = 1;\n        dense2.beginMask |= 1 << fullIndex;\n        dense2.endMask |= 1 << fullIndex;\n        dense2.finalShapeGatherIndices.push(fullIndex);\n        dense2.finalShapeGatherIndicesSparse.push(-1);\n        dense2.inputShapeGatherIndicesSparse[fullIndex] = i;\n      }\n    } else if (1 << i & sparse2.newAxisMask) {\n      dense2.finalShapeGatherIndices.push(NEW_AXIS);\n      dense2.finalShapeGatherIndicesSparse.push(-1);\n    } else {\n      if (fullIndex === dense2.begin.length) {\n        throw Error(`Index out of range using input dim ${fullIndex}; input has only ${dense2.dims} dims, ${dense2.begin.length}.`);\n      }\n      if (sparse2.begin != null) {\n        dense2.begin[fullIndex] = sparse2.begin[i];\n      }\n      if (sparse2.end != null) {\n        dense2.end[fullIndex] = sparse2.end[i];\n      }\n      dense2.strides[fullIndex] = sparse2.strides[i];\n      if (sparse2.beginMask & 1 << i) {\n        dense2.beginMask |= 1 << fullIndex;\n      }\n      if (sparse2.endMask & 1 << i) {\n        dense2.endMask |= 1 << fullIndex;\n      }\n      if (sparse2.shrinkAxisMask & 1 << i) {\n        dense2.finalShapeGatherIndices.push(SHRINK_AXIS);\n        dense2.finalShapeGatherIndicesSparse.push(-1);\n        dense2.shrinkAxisMask |= 1 << fullIndex;\n      } else {\n        dense2.finalShapeGatherIndices.push(fullIndex);\n        dense2.finalShapeGatherIndicesSparse.push(i);\n      }\n      dense2.inputShapeGatherIndicesSparse[fullIndex] = i;\n      fullIndex++;\n    }\n  }\n}\nfunction canonical(x, c, strideI, dimI, masks, validRange) {\n  if (masks[c]) {\n    return strideI > 0 ? validRange[c] : validRange[c + 1 & 1];\n  } else {\n    const xFwd = x < 0 ? dimI + x : x;\n    return xFwd < validRange[0] ? validRange[0] : xFwd > validRange[1] ? validRange[1] : xFwd;\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/serialization.js\nvar serialization_exports = {};\n__export(serialization_exports, {\n  Serializable: () => Serializable,\n  SerializationMap: () => SerializationMap,\n  registerClass: () => registerClass\n});\nvar Serializable = class {\n  getClassName() {\n    return this.constructor.className;\n  }\n  static fromConfig(cls, config) {\n    return new cls(config);\n  }\n};\nvar SerializationMap = class {\n  constructor() {\n    this.classNameMap = {};\n  }\n  static getMap() {\n    if (SerializationMap.instance == null) {\n      SerializationMap.instance = new SerializationMap();\n    }\n    return SerializationMap.instance;\n  }\n  static register(cls) {\n    SerializationMap.getMap().classNameMap[cls.className] = [cls, cls.fromConfig];\n  }\n};\nfunction registerClass(cls) {\n  assert(cls.className != null, () => `Class being registered does not have the static className property defined.`);\n  assert(typeof cls.className === \"string\", () => `className is required to be a string, but got type ` + typeof cls.className);\n  assert(cls.className.length > 0, () => `Class being registered has an empty-string as its className, which is disallowed.`);\n  SerializationMap.register(cls);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/test_util.js\nvar test_util_exports = {};\n__export(test_util_exports, {\n  TEST_EPSILON_FLOAT16: () => TEST_EPSILON_FLOAT16,\n  createVideoElement: () => createVideoElement,\n  encodeStrings: () => encodeStrings,\n  expectArrayBuffersEqual: () => expectArrayBuffersEqual,\n  expectArraysClose: () => expectArraysClose,\n  expectArraysEqual: () => expectArraysEqual,\n  expectNumbersClose: () => expectNumbersClose,\n  expectPromiseToFail: () => expectPromiseToFail,\n  expectValuesInRange: () => expectValuesInRange,\n  play: () => play,\n  testEpsilon: () => testEpsilon\n});\nvar TEST_EPSILON_FLOAT32 = 1e-3;\nvar TEST_EPSILON_FLOAT16 = 0.1;\nfunction expectArraysClose(actual, expected, epsilon3) {\n  if (epsilon3 == null) {\n    epsilon3 = testEpsilon();\n  }\n  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, epsilon3));\n}\nfunction testEpsilon() {\n  return ENGINE.backend.floatPrecision() === 32 ? TEST_EPSILON_FLOAT32 : TEST_EPSILON_FLOAT16;\n}\nfunction expectArraysPredicate(actual, expected, predicate) {\n  let checkClassType = true;\n  if (isTypedArray(actual) || isTypedArray(expected)) {\n    checkClassType = false;\n  }\n  if (isTypedArray(actual) && isTypedArray(expected)) {\n    checkClassType = true;\n  }\n  if (checkClassType) {\n    const aType = actual.constructor.name;\n    const bType = expected.constructor.name;\n    if (aType !== bType) {\n      throw new Error(`Arrays are of different type. Actual: ${aType}. Expected: ${bType}`);\n    }\n  }\n  if (Array.isArray(actual) && Array.isArray(expected)) {\n    const actualShape = inferShape(actual);\n    const expectedShape = inferShape(expected);\n    if (!arraysEqual(actualShape, expectedShape)) {\n      throw new Error(`Arrays have different shapes. Actual: [${actualShape}]. Expected: [${expectedShape}]`);\n    }\n  }\n  const actualFlat = isTypedArray(actual) ? actual : flatten(actual);\n  const expectedFlat = isTypedArray(expected) ? expected : flatten(expected);\n  if (actualFlat.length !== expectedFlat.length) {\n    throw new Error(`Arrays have different lengths actual: ${actualFlat.length} vs expected: ${expectedFlat.length}.\nActual:   ${actualFlat}.\nExpected: ${expectedFlat}.`);\n  }\n  for (let i = 0; i < expectedFlat.length; ++i) {\n    const a = actualFlat[i];\n    const e = expectedFlat[i];\n    if (!predicate(a, e)) {\n      throw new Error(`Arrays differ: actual[${i}] = ${a}, expected[${i}] = ${e}.\nActual:   ${actualFlat}.\nExpected: ${expectedFlat}.`);\n    }\n  }\n  if (typeof expect !== \"undefined\") {\n    expect().nothing();\n  }\n}\nfunction expectPromiseToFail(fn, done) {\n  fn().then(() => done.fail(), () => done());\n  if (typeof expect !== \"undefined\") {\n    expect().nothing();\n  }\n}\nfunction expectArraysEqual(actual, expected) {\n  const exp4 = typeof expected === \"string\" || typeof expected === \"number\" || typeof expected === \"boolean\" ? [expected] : expected;\n  if (isString(actual) || isString(actual[0]) || isString(expected) || isString(expected[0])) {\n    return expectArraysPredicate(actual, exp4, (a, b) => a == b);\n  }\n  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, 0));\n}\nfunction expectNumbersClose(a, e, epsilon3) {\n  if (epsilon3 == null) {\n    epsilon3 = testEpsilon();\n  }\n  if (!areClose(a, e, epsilon3)) {\n    throw new Error(`Numbers differ: actual === ${a}, expected === ${e}`);\n  }\n  if (typeof expect !== \"undefined\") {\n    expect().nothing();\n  }\n}\nfunction areClose(a, e, epsilon3) {\n  if (!isFinite(a) && !isFinite(e)) {\n    return true;\n  }\n  if (isNaN(a) || isNaN(e) || Math.abs(a - e) > epsilon3) {\n    return false;\n  }\n  return true;\n}\nfunction expectValuesInRange(actual, low, high) {\n  for (let i = 0; i < actual.length; i++) {\n    if (actual[i] < low || actual[i] > high) {\n      throw new Error(`Value out of range:${actual[i]} low: ${low}, high: ${high}`);\n    }\n  }\n}\nfunction expectArrayBuffersEqual(actual, expected) {\n  const actualArray = new Float32Array(actual);\n  const expectedArray = new Float32Array(expected);\n  if (actualArray.length !== expectedArray.length) {\n    throw new Error(`Expected ArrayBuffer to be of length ${expectedArray.length}, but it was ${actualArray.length}`);\n  }\n  for (let i = 0; i < expectedArray.length; i++) {\n    if (actualArray[i] !== expectedArray[i]) {\n      throw new Error(`Expected ArrayBuffer value at ${i} to be ${expectedArray[i]} but got ${actualArray[i]} instead`);\n    }\n  }\n}\nfunction encodeStrings(a) {\n  for (let i = 0; i < a.length; i++) {\n    const val = a[i];\n    if (Array.isArray(val)) {\n      encodeStrings(val);\n    } else {\n      a[i] = encodeString(val);\n    }\n  }\n  return a;\n}\nfunction createVideoElement(source) {\n  const video = document.createElement(\"video\");\n  if (\"playsInline\" in video) {\n    video.playsInline = true;\n  }\n  video.muted = true;\n  video.loop = true;\n  video.style.position = \"fixed\";\n  video.style.left = \"0px\";\n  video.style.top = \"0px\";\n  video.preload = \"auto\";\n  video.appendChild(source);\n  return new Promise((resolve) => {\n    video.addEventListener(\"loadeddata\", (_) => resolve(video));\n    video.load();\n  });\n}\nasync function play(video) {\n  await video.play();\n  if (\"requestVideoFrameCallback\" in video) {\n    await new Promise((resolve) => {\n      video.requestVideoFrameCallback(resolve);\n    });\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/version.js\nvar version = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/add.js\nfunction add_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"add\");\n  let $b = convertToTensor(b, \"b\", \"add\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Add, inputs);\n}\nvar add2 = op({ add_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/floorDiv.js\nfunction floorDiv_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"floorDiv\");\n  let $b = convertToTensor(b, \"b\", \"floorDiv\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(FloorDiv, inputs);\n}\nvar floorDiv = op({ floorDiv_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/div.js\nfunction div_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"div\");\n  let $b = convertToTensor(b, \"b\", \"div\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  if ($a.dtype === \"int32\" && $b.dtype === \"int32\") {\n    return floorDiv($a, $b);\n  }\n  const inputs = { a: $a, b: $b };\n  const attrs = {};\n  return ENGINE.runKernel(RealDiv, inputs, attrs);\n}\nvar div = op({ div_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/mul.js\nfunction mul_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"mul\");\n  let $b = convertToTensor(b, \"b\", \"mul\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Multiply, inputs);\n}\nvar mul = op({ mul_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/abs.js\nfunction abs_(x) {\n  const $x = convertToTensor(x, \"x\", \"abs\");\n  if ($x.dtype === \"complex64\") {\n    const inputs = { x: $x };\n    return ENGINE.runKernel(ComplexAbs, inputs);\n  } else {\n    const inputs = { x: $x };\n    return ENGINE.runKernel(Abs, inputs);\n  }\n}\nvar abs = op({ abs_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/acos.js\nfunction acos_(x) {\n  const $x = convertToTensor(x, \"x\", \"acos\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Acos, inputs);\n}\nvar acos = op({ acos_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/acosh.js\nfunction acosh_(x) {\n  const $x = convertToTensor(x, \"x\", \"acosh\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Acosh, inputs);\n}\nvar acosh = op({ acosh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/add_n.js\nfunction addN_(tensors) {\n  assert(Array.isArray(tensors), () => \"The argument passed to tf.addN() must be a list of tensors\");\n  assert(tensors.length >= 1, () => `Must pass at least one tensor to tf.addN(), but got ${tensors.length}`);\n  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, \"addN\"));\n  const firstTensor = $tensors[0];\n  $tensors.forEach((t) => {\n    if (t.dtype !== firstTensor.dtype) {\n      throw new Error(\"All tensors passed to tf.addN() must have the same dtype\");\n    }\n  });\n  $tensors.forEach((t) => {\n    if (!arraysEqual(t.shape, firstTensor.shape)) {\n      throw new Error(\"All tensors passed to tf.addN() must have the same shape\");\n    }\n  });\n  const inputs = $tensors;\n  return ENGINE.runKernel(AddN, inputs);\n}\nvar addN = op({ addN_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/all.js\nfunction all_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"all\", \"bool\");\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(All, inputs, attrs);\n}\nvar all = op({ all_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/any.js\nfunction any_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"any\", \"bool\");\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(Any, inputs, attrs);\n}\nvar any = op({ any_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/arg_max.js\nfunction argMax_(x, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"argMax\");\n  const inputs = { x: $x };\n  const attrs = { axis };\n  return ENGINE.runKernel(ArgMax, inputs, attrs);\n}\nvar argMax = op({ argMax_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/arg_min.js\nfunction argMin_(x, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"argMin\");\n  const inputs = { x: $x };\n  const attrs = { axis };\n  return ENGINE.runKernel(ArgMin, inputs, attrs);\n}\nvar argMin = op({ argMin_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/asin.js\nfunction asin_(x) {\n  const $x = convertToTensor(x, \"x\", \"asin\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Asin, inputs);\n}\nvar asin = op({ asin_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/asinh.js\nfunction asinh_(x) {\n  const $x = convertToTensor(x, \"x\", \"asinh\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Asinh, inputs);\n}\nvar asinh = op({ asinh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/atan.js\nfunction atan_(x) {\n  const $x = convertToTensor(x, \"x\", \"atan\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Atan, inputs);\n}\nvar atan = op({ atan_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/atan2.js\nfunction atan2_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"atan2\");\n  let $b = convertToTensor(b, \"b\", \"atan2\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Atan2, inputs);\n}\nvar atan2 = op({ atan2_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/atanh.js\nfunction atanh_(x) {\n  const $x = convertToTensor(x, \"x\", \"atanh\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Atanh, inputs);\n}\nvar atanh = op({ atanh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv_util.js\nfunction computeDilation2DInfo(inputShape, filterShape, strides, pad3, dataFormat = \"NHWC\", dilations) {\n  const inputChannels = inputShape[3];\n  const $filterShape = [...filterShape, inputChannels];\n  const $dataFormat = convertConv2DDataFormat(dataFormat);\n  return computeConv2DInfo(inputShape, $filterShape, strides, dilations, pad3, null, null, $dataFormat);\n}\nfunction computePool2DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = \"channelsLast\") {\n  const [filterHeight, filterWidth] = parseTupleParam(filterSize);\n  let filterShape;\n  if (dataFormat === \"channelsLast\") {\n    filterShape = [filterHeight, filterWidth, inShape[3], inShape[3]];\n  } else if (dataFormat === \"channelsFirst\") {\n    filterShape = [filterHeight, filterWidth, inShape[1], inShape[1]];\n  } else {\n    throw new Error(`Unknown dataFormat ${dataFormat}`);\n  }\n  return computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, false, dataFormat);\n}\nfunction computePool3DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = \"NDHWC\") {\n  const [filterDepth, filterHeight, filterWidth] = parse3TupleParam(filterSize);\n  let filterShape;\n  let $dataFormat;\n  if (dataFormat === \"NDHWC\") {\n    $dataFormat = \"channelsLast\";\n    filterShape = [filterDepth, filterHeight, filterWidth, inShape[4], inShape[4]];\n  } else if (dataFormat === \"NCDHW\") {\n    $dataFormat = \"channelsFirst\";\n    filterShape = [filterDepth, filterHeight, filterWidth, inShape[1], inShape[1]];\n  } else {\n    throw new Error(`Unknown dataFormat ${dataFormat}`);\n  }\n  return computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, false, $dataFormat, roundingMode);\n}\nfunction computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, depthwise = false, dataFormat = \"channelsLast\") {\n  let [batchSize, inHeight, inWidth, inChannels] = [-1, -1, -1, -1];\n  if (dataFormat === \"channelsLast\") {\n    [batchSize, inHeight, inWidth, inChannels] = inShape;\n  } else if (dataFormat === \"channelsFirst\") {\n    [batchSize, inChannels, inHeight, inWidth] = inShape;\n  } else {\n    throw new Error(`Unknown dataFormat ${dataFormat}`);\n  }\n  const [filterHeight, filterWidth, , filterChannels] = filterShape;\n  const [strideHeight, strideWidth] = parseTupleParam(strides);\n  const [dilationHeight, dilationWidth] = parseTupleParam(dilations);\n  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);\n  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);\n  const { padInfo, outHeight, outWidth } = getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, effectiveFilterHeight, effectiveFilterWidth, roundingMode, dataFormat);\n  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;\n  let outShape;\n  if (dataFormat === \"channelsFirst\") {\n    outShape = [batchSize, outChannels, outHeight, outWidth];\n  } else if (dataFormat === \"channelsLast\") {\n    outShape = [batchSize, outHeight, outWidth, outChannels];\n  }\n  return {\n    batchSize,\n    dataFormat,\n    inHeight,\n    inWidth,\n    inChannels,\n    outHeight,\n    outWidth,\n    outChannels,\n    padInfo,\n    strideHeight,\n    strideWidth,\n    filterHeight,\n    filterWidth,\n    effectiveFilterHeight,\n    effectiveFilterWidth,\n    dilationHeight,\n    dilationWidth,\n    inShape,\n    outShape,\n    filterShape\n  };\n}\nfunction computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, depthwise = false, dataFormat = \"channelsLast\", roundingMode) {\n  let [batchSize, inDepth, inHeight, inWidth, inChannels] = [-1, -1, -1, -1, -1];\n  if (dataFormat === \"channelsLast\") {\n    [batchSize, inDepth, inHeight, inWidth, inChannels] = inShape;\n  } else if (dataFormat === \"channelsFirst\") {\n    [batchSize, inChannels, inDepth, inHeight, inWidth] = inShape;\n  } else {\n    throw new Error(`Unknown dataFormat ${dataFormat}`);\n  }\n  const [filterDepth, filterHeight, filterWidth, , filterChannels] = filterShape;\n  const [strideDepth, strideHeight, strideWidth] = parse3TupleParam(strides);\n  const [dilationDepth, dilationHeight, dilationWidth] = parse3TupleParam(dilations);\n  const effectiveFilterDepth = getEffectiveFilterSize(filterDepth, dilationDepth);\n  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);\n  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);\n  const { padInfo, outDepth, outHeight, outWidth } = get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, effectiveFilterDepth, effectiveFilterHeight, effectiveFilterWidth, roundingMode);\n  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;\n  let outShape;\n  if (dataFormat === \"channelsFirst\") {\n    outShape = [batchSize, outChannels, outDepth, outHeight, outWidth];\n  } else if (dataFormat === \"channelsLast\") {\n    outShape = [batchSize, outDepth, outHeight, outWidth, outChannels];\n  }\n  return {\n    batchSize,\n    dataFormat,\n    inDepth,\n    inHeight,\n    inWidth,\n    inChannels,\n    outDepth,\n    outHeight,\n    outWidth,\n    outChannels,\n    padInfo,\n    strideDepth,\n    strideHeight,\n    strideWidth,\n    filterDepth,\n    filterHeight,\n    filterWidth,\n    effectiveFilterDepth,\n    effectiveFilterHeight,\n    effectiveFilterWidth,\n    dilationDepth,\n    dilationHeight,\n    dilationWidth,\n    inShape,\n    outShape,\n    filterShape\n  };\n}\nfunction computeOutputShape2D(inShape, fieldSize, stride, zeroPad, roundingMode) {\n  if (zeroPad == null) {\n    zeroPad = computeDefaultPad(inShape, fieldSize, stride);\n  }\n  const inputRows = inShape[0];\n  const inputCols = inShape[1];\n  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);\n  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);\n  return [outputRows, outputCols];\n}\nfunction computeOutputShape4D(inShape, fieldSize, outChannels, stride, zeroPad, roundingMode) {\n  if (zeroPad == null) {\n    zeroPad = computeDefaultPad(inShape, fieldSize, stride);\n  }\n  const inputDepth = inShape[0];\n  const inputRows = inShape[1];\n  const inputCols = inShape[2];\n  const outputDepths = round((inputDepth - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);\n  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);\n  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);\n  return [outputDepths, outputRows, outputCols, outChannels];\n}\nfunction computeDefaultPad(inputShape, fieldSize, stride, dilation = 1) {\n  const effectiveFieldSize = getEffectiveFilterSize(fieldSize, dilation);\n  return Math.floor((inputShape[0] * (stride - 1) - stride + effectiveFieldSize) / 2);\n}\nfunction parseTupleParam(param) {\n  if (typeof param === \"number\") {\n    return [param, param, param];\n  }\n  if (param.length === 2) {\n    return [param[0], param[1], 1];\n  }\n  return param;\n}\nfunction parse3TupleParam(param) {\n  return typeof param === \"number\" ? [param, param, param] : param;\n}\nfunction getEffectiveFilterSize(filterSize, dilation) {\n  if (dilation <= 1) {\n    return filterSize;\n  }\n  return filterSize + (filterSize - 1) * (dilation - 1);\n}\nfunction getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, filterHeight, filterWidth, roundingMode, dataFormat) {\n  let padInfo;\n  let outHeight;\n  let outWidth;\n  if (typeof pad3 === \"number\") {\n    const padType = pad3 === 0 ? \"VALID\" : \"NUMBER\";\n    padInfo = { top: pad3, bottom: pad3, left: pad3, right: pad3, type: padType };\n    const outShape = computeOutputShape2D([inHeight, inWidth], filterHeight, strideHeight, pad3, roundingMode);\n    outHeight = outShape[0];\n    outWidth = outShape[1];\n  } else if (pad3 === \"same\") {\n    outHeight = Math.ceil(inHeight / strideHeight);\n    outWidth = Math.ceil(inWidth / strideWidth);\n    const padAlongHeight = Math.max(0, (outHeight - 1) * strideHeight + filterHeight - inHeight);\n    const padAlongWidth = Math.max(0, (outWidth - 1) * strideWidth + filterWidth - inWidth);\n    const top = Math.floor(padAlongHeight / 2);\n    const bottom = padAlongHeight - top;\n    const left = Math.floor(padAlongWidth / 2);\n    const right = padAlongWidth - left;\n    padInfo = { top, bottom, left, right, type: \"SAME\" };\n  } else if (pad3 === \"valid\") {\n    padInfo = { top: 0, bottom: 0, left: 0, right: 0, type: \"VALID\" };\n    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);\n    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);\n  } else if (typeof pad3 === \"object\") {\n    const top = dataFormat === \"channelsLast\" ? pad3[1][0] : pad3[2][0];\n    const bottom = dataFormat === \"channelsLast\" ? pad3[1][1] : pad3[2][1];\n    const left = dataFormat === \"channelsLast\" ? pad3[2][0] : pad3[3][0];\n    const right = dataFormat === \"channelsLast\" ? pad3[2][1] : pad3[3][1];\n    const padType = top === 0 && bottom === 0 && left === 0 && right === 0 ? \"VALID\" : \"EXPLICIT\";\n    padInfo = { top, bottom, left, right, type: padType };\n    outHeight = round((inHeight - filterHeight + top + bottom) / strideHeight + 1, roundingMode);\n    outWidth = round((inWidth - filterWidth + left + right) / strideWidth + 1, roundingMode);\n  } else {\n    throw Error(`Unknown padding parameter: ${pad3}`);\n  }\n  return { padInfo, outHeight, outWidth };\n}\nfunction get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, filterDepth, filterHeight, filterWidth, roundingMode) {\n  let padInfo;\n  let outDepth;\n  let outHeight;\n  let outWidth;\n  if (typeof pad3 === \"number\") {\n    const padType = pad3 === 0 ? \"VALID\" : \"NUMBER\";\n    padInfo = {\n      top: pad3,\n      bottom: pad3,\n      left: pad3,\n      right: pad3,\n      front: pad3,\n      back: pad3,\n      type: padType\n    };\n    const outShape = computeOutputShape4D([inDepth, inHeight, inWidth, 1], filterDepth, 1, strideDepth, pad3, roundingMode);\n    outDepth = outShape[0];\n    outHeight = outShape[1];\n    outWidth = outShape[2];\n  } else if (pad3 === \"same\") {\n    outDepth = Math.ceil(inDepth / strideDepth);\n    outHeight = Math.ceil(inHeight / strideHeight);\n    outWidth = Math.ceil(inWidth / strideWidth);\n    const padAlongDepth = (outDepth - 1) * strideDepth + filterDepth - inDepth;\n    const padAlongHeight = (outHeight - 1) * strideHeight + filterHeight - inHeight;\n    const padAlongWidth = (outWidth - 1) * strideWidth + filterWidth - inWidth;\n    const front = Math.floor(padAlongDepth / 2);\n    const back = padAlongDepth - front;\n    const top = Math.floor(padAlongHeight / 2);\n    const bottom = padAlongHeight - top;\n    const left = Math.floor(padAlongWidth / 2);\n    const right = padAlongWidth - left;\n    padInfo = { top, bottom, left, right, front, back, type: \"SAME\" };\n  } else if (pad3 === \"valid\") {\n    padInfo = {\n      top: 0,\n      bottom: 0,\n      left: 0,\n      right: 0,\n      front: 0,\n      back: 0,\n      type: \"VALID\"\n    };\n    outDepth = Math.ceil((inDepth - filterDepth + 1) / strideDepth);\n    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);\n    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);\n  } else {\n    throw Error(`Unknown padding parameter: ${pad3}`);\n  }\n  return { padInfo, outDepth, outHeight, outWidth };\n}\nfunction round(value, roundingMode) {\n  if (!roundingMode) {\n    return Math.trunc(value);\n  }\n  switch (roundingMode) {\n    case \"round\":\n      return Math.round(value);\n    case \"ceil\":\n      return Math.ceil(value);\n    case \"floor\":\n      return Math.floor(value);\n    default:\n      throw new Error(`Unknown roundingMode ${roundingMode}`);\n  }\n}\nfunction tupleValuesAreOne(param) {\n  const [dimA, dimB, dimC] = parseTupleParam(param);\n  return dimA === 1 && dimB === 1 && dimC === 1;\n}\nfunction eitherStridesOrDilationsAreOne(strides, dilations) {\n  return tupleValuesAreOne(strides) || tupleValuesAreOne(dilations);\n}\nfunction convertConv2DDataFormat(dataFormat) {\n  if (dataFormat === \"NHWC\") {\n    return \"channelsLast\";\n  } else if (dataFormat === \"NCHW\") {\n    return \"channelsFirst\";\n  } else {\n    throw new Error(`Unknown dataFormat ${dataFormat}`);\n  }\n}\nfunction checkPadOnDimRoundingMode(opDesc, pad3, dimRoundingMode) {\n  if (dimRoundingMode != null) {\n    if (typeof pad3 === \"string\") {\n      throw Error(`Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);\n    } else if (typeof pad3 === \"number\") {\n      assert(isInt(pad3), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);\n    } else if (typeof pad3 === \"object\") {\n      pad3.forEach((p2) => {\n        p2.forEach((v) => {\n          assert(isInt(v), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${v}.`);\n        });\n      });\n    } else {\n      throw Error(`Error in ${opDesc}: Unknown padding parameter: ${pad3}`);\n    }\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reshape.js\nfunction reshape_(x, shape) {\n  const $x = convertToTensor(x, \"x\", \"reshape\", \"string_or_numeric\");\n  const inputs = { x: $x };\n  const attrs = { shape };\n  return ENGINE.runKernel(Reshape, inputs, attrs);\n}\nvar reshape = op({ reshape_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool.js\nfunction avgPool_(x, filterSize, strides, pad3, dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"avgPool\", \"float32\");\n  const dilations = 1;\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in avgPool: x must be rank 4 but got rank ${x4D.rank}.`);\n  checkPadOnDimRoundingMode(\"avgPool\", pad3, dimRoundingMode);\n  const inputs = { x: x4D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };\n  let res = ENGINE.runKernel(AvgPool, inputs, attrs);\n  res = cast(res, $x.dtype);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar avgPool = op({ avgPool_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_3d.js\nfunction avgPool3d_(x, filterSize, strides, pad3, dimRoundingMode, dataFormat = \"NDHWC\") {\n  const $x = convertToTensor(x, \"x\", \"avgPool3d\", \"float32\");\n  let x5D = $x;\n  let reshapedTo5D = false;\n  if ($x.rank === 4) {\n    reshapedTo5D = true;\n    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);\n  }\n  assert(x5D.rank === 5, () => `Error in avgPool3d: x must be rank 5 but got rank ${x5D.rank}.`);\n  assert(dataFormat === \"NDHWC\", () => `Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);\n  checkPadOnDimRoundingMode(\"avgPool3d\", pad3, dimRoundingMode);\n  const inputs = { x: x5D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };\n  let res = ENGINE.runKernel(AvgPool3D, inputs, attrs);\n  res = cast(res, x5D.dtype);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar avgPool3d = op({ avgPool3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat.js\nfunction concat_(tensors, axis = 0) {\n  assert(tensors.length >= 1, () => \"Pass at least one tensor to concat\");\n  const $tensors = convertToTensorArray(tensors, \"tensors\", \"concat\", \"string_or_numeric\");\n  if ($tensors[0].dtype === \"complex64\") {\n    $tensors.forEach((tensor2) => {\n      if (tensor2.dtype !== \"complex64\") {\n        throw new Error(`Cannot concatenate complex64 tensors with a tensor\n          with dtype ${tensor2.dtype}. `);\n      }\n    });\n  }\n  if ($tensors.length === 1) {\n    return clone($tensors[0]);\n  }\n  const inputs = $tensors;\n  const attr = { axis };\n  return ENGINE.runKernel(Concat, inputs, attr);\n}\nvar concat = op({ concat_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sigmoid.js\nfunction sigmoid_(x) {\n  const $x = convertToTensor(x, \"x\", \"sigmoid\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Sigmoid, inputs);\n}\nvar sigmoid = op({ sigmoid_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice.js\nfunction slice_(x, begin, size) {\n  const $x = convertToTensor(x, \"x\", \"slice\", \"string_or_numeric\");\n  if ($x.rank === 0) {\n    throw new Error(\"Slicing scalar is not possible\");\n  }\n  const inputs = { x: $x };\n  const attrs = { begin, size };\n  return ENGINE.runKernel(Slice, inputs, attrs);\n}\nvar slice = op({ slice_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tanh.js\nfunction tanh_(x) {\n  const $x = convertToTensor(x, \"x\", \"tanh\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Tanh, inputs);\n}\nvar tanh2 = op({ tanh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/basic_lstm_cell.js\nfunction basicLSTMCell_(forgetBias, lstmKernel, lstmBias, data, c, h) {\n  const $forgetBias = convertToTensor(forgetBias, \"forgetBias\", \"basicLSTMCell\");\n  const $lstmKernel = convertToTensor(lstmKernel, \"lstmKernel\", \"basicLSTMCell\");\n  const $lstmBias = convertToTensor(lstmBias, \"lstmBias\", \"basicLSTMCell\");\n  const $data = convertToTensor(data, \"data\", \"basicLSTMCell\");\n  const $c = convertToTensor(c, \"c\", \"basicLSTMCell\");\n  const $h = convertToTensor(h, \"h\", \"basicLSTMCell\");\n  const combined = concat([$data, $h], 1);\n  const weighted = matMul(combined, $lstmKernel);\n  const res = add2(weighted, $lstmBias);\n  const batchSize = res.shape[0];\n  const sliceCols = res.shape[1] / 4;\n  const sliceSize = [batchSize, sliceCols];\n  const i = slice(res, [0, 0], sliceSize);\n  const j = slice(res, [0, sliceCols], sliceSize);\n  const f = slice(res, [0, sliceCols * 2], sliceSize);\n  const o = slice(res, [0, sliceCols * 3], sliceSize);\n  const newC = add2(mul(sigmoid(i), tanh2(j)), mul($c, sigmoid(add2($forgetBias, f))));\n  const newH = mul(tanh2(newC), sigmoid(o));\n  return [newC, newH];\n}\nvar basicLSTMCell = op({ basicLSTMCell_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batch_to_space_nd.js\nfunction batchToSpaceND_(x, blockShape, crops) {\n  const $x = convertToTensor(x, \"x\", \"batchToSpaceND\");\n  const prod5 = blockShape.reduce((a, b) => a * b);\n  assert($x.rank >= 1 + blockShape.length, () => `input rank is ${$x.rank} but should be > than blockShape.length ${blockShape.length}`);\n  assert(crops.length === blockShape.length, () => `crops.length is ${crops.length} but should be equal to blockShape.length  ${blockShape.length}`);\n  assert($x.shape[0] % prod5 === 0, () => `input tensor batch is ${$x.shape[0]} but is not divisible by the product of the elements of blockShape ${blockShape.join(\" * \")} === ${prod5}`);\n  const inputs = { x: $x };\n  const attrs = { blockShape, crops };\n  return ENGINE.runKernel(BatchToSpaceND, inputs, attrs);\n}\nvar batchToSpaceND = op({ batchToSpaceND_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm_util.js\nfunction xAs4D(x) {\n  let x4D;\n  if (x.rank === 0 || x.rank === 1) {\n    x4D = reshape(x, [1, 1, 1, x.size]);\n  } else if (x.rank === 2) {\n    x4D = reshape(x, [1, 1, x.shape[0], x.shape[1]]);\n  } else if (x.rank === 3) {\n    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);\n  } else {\n    x4D = x;\n  }\n  return x4D;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm.js\nfunction batchNorm_(x, mean4, variance, offset, scale2, varianceEpsilon) {\n  if (varianceEpsilon == null) {\n    varianceEpsilon = 1e-3;\n  }\n  const $x = convertToTensor(x, \"x\", \"batchNorm\");\n  const $mean = convertToTensor(mean4, \"mean\", \"batchNorm\");\n  const $variance = convertToTensor(variance, \"variance\", \"batchNorm\");\n  let $scale;\n  if (scale2 != null) {\n    $scale = convertToTensor(scale2, \"scale\", \"batchNorm\");\n  }\n  let $offset;\n  if (offset != null) {\n    $offset = convertToTensor(offset, \"offset\", \"batchNorm\");\n  }\n  assert($mean.rank === $variance.rank, () => \"Batch normalization gradient requires mean and variance to have equal ranks.\");\n  assert($offset == null || $mean.rank === $offset.rank, () => \"Batch normalization gradient requires mean and offset to have equal ranks.\");\n  assert($scale == null || $mean.rank === $scale.rank, () => \"Batch normalization gradient requires mean and scale to have equal ranks.\");\n  const x4D = xAs4D($x);\n  const inputs = {\n    x: x4D,\n    scale: $scale,\n    offset: $offset,\n    mean: $mean,\n    variance: $variance\n  };\n  const attrs = { varianceEpsilon };\n  const res = ENGINE.runKernel(FusedBatchNorm, inputs, attrs);\n  return reshape(res, $x.shape);\n}\nvar batchNorm = op({ batchNorm_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm2d.js\nfunction batchNorm2d_(x, mean4, variance, offset, scale2, varianceEpsilon) {\n  const $x = convertToTensor(x, \"x\", \"batchNorm\");\n  const $mean = convertToTensor(mean4, \"mean\", \"batchNorm\");\n  const $variance = convertToTensor(variance, \"variance\", \"batchNorm\");\n  let $scale;\n  if (scale2 != null) {\n    $scale = convertToTensor(scale2, \"scale\", \"batchNorm\");\n  }\n  let $offset;\n  if (offset != null) {\n    $offset = convertToTensor(offset, \"offset\", \"batchNorm\");\n  }\n  assert($x.rank === 2, () => `Error in batchNorm2D: x must be rank 2 but got rank ${$x.rank}.`);\n  assert($mean.rank === 2 || $mean.rank === 1, () => `Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${$mean.rank}.`);\n  assert($variance.rank === 2 || $variance.rank === 1, () => `Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${$variance.rank}.`);\n  if ($scale != null) {\n    assert($scale.rank === 2 || $scale.rank === 1, () => `Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${$scale.rank}.`);\n  }\n  if ($offset != null) {\n    assert($offset.rank === 2 || $offset.rank === 1, () => `Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${$offset.rank}.`);\n  }\n  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);\n}\nvar batchNorm2d = op({ batchNorm2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm3d.js\nfunction batchNorm3d_(x, mean4, variance, offset, scale2, varianceEpsilon) {\n  const $x = convertToTensor(x, \"x\", \"batchNorm\");\n  const $mean = convertToTensor(mean4, \"mean\", \"batchNorm\");\n  const $variance = convertToTensor(variance, \"variance\", \"batchNorm\");\n  let $scale;\n  if (scale2 != null) {\n    $scale = convertToTensor(scale2, \"scale\", \"batchNorm\");\n  }\n  let $offset;\n  if (offset != null) {\n    $offset = convertToTensor(offset, \"offset\", \"batchNorm\");\n  }\n  assert($x.rank === 3, () => `Error in batchNorm3D: x must be rank 3 but got rank ${$x.rank}.`);\n  assert($mean.rank === 3 || $mean.rank === 1, () => `Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${$mean.rank}.`);\n  assert($variance.rank === 3 || $variance.rank === 1, () => `Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${$variance.rank}.`);\n  if ($scale != null) {\n    assert($scale.rank === 3 || $scale.rank === 1, () => `Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${$scale.rank}.`);\n  }\n  if ($offset != null) {\n    assert($offset.rank === 3 || $offset.rank === 1, () => `Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${$offset.rank}.`);\n  }\n  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);\n}\nvar batchNorm3d = op({ batchNorm3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm4d.js\nfunction batchNorm4d_(x, mean4, variance, offset, scale2, varianceEpsilon) {\n  const $x = convertToTensor(x, \"x\", \"batchNorm\");\n  const $mean = convertToTensor(mean4, \"mean\", \"batchNorm\");\n  const $variance = convertToTensor(variance, \"variance\", \"batchNorm\");\n  let $scale;\n  if (scale2 != null) {\n    $scale = convertToTensor(scale2, \"scale\", \"batchNorm\");\n  }\n  let $offset;\n  if (offset != null) {\n    $offset = convertToTensor(offset, \"offset\", \"batchNorm\");\n  }\n  assert($x.rank === 4, () => `Error in batchNorm4D: x must be rank 4 but got rank ${$x.rank}.`);\n  assert($mean.rank === 4 || $mean.rank === 1, () => `Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${$mean.rank}.`);\n  assert($variance.rank === 4 || $variance.rank === 1, () => `Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${$variance.rank}.`);\n  if ($scale != null) {\n    assert($scale.rank === 4 || $scale.rank === 1, () => `Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${$scale.rank}.`);\n  }\n  if ($offset != null) {\n    assert($offset.rank === 4 || $offset.rank === 1, () => `Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${$offset.rank}.`);\n  }\n  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);\n}\nvar batchNorm4d = op({ batchNorm4d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/bincount.js\nfunction bincount_(x, weights, size) {\n  const $x = convertToTensor(x, \"x\", \"bincount\");\n  const $weights = convertToTensor(weights, \"weights\", \"bincount\");\n  assert($x.dtype === \"int32\", () => `Error in bincount: input dtype must be int32, but got ${$x.dtype}`);\n  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);\n  assert($weights.size === $x.size || $weights.size === 0, () => `Error in bincount: weights must have the same size as input or0-length, but got input shape: ${$x.shape}, weights shape: ${$weights.shape}.`);\n  const inputs = { x: $x, weights: $weights };\n  const attrs = { size };\n  return ENGINE.runKernel(Bincount, inputs, attrs);\n}\nvar bincount = op({ bincount_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_args.js\nfunction broadcastArgs_(s0, s1) {\n  const shape1Input = convertToTensor(s0, \"s0\", \"broadcastArgs\", \"int32\");\n  const shape2Input = convertToTensor(s1, \"s1\", \"broadcastArgs\", \"int32\");\n  if (shape1Input.rank !== 1) {\n    throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${shape1Input.rank}`);\n  }\n  if (shape2Input.rank !== 1) {\n    throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${shape2Input.rank}`);\n  }\n  const inputs = { s0: shape1Input, s1: shape2Input };\n  return ENGINE.runKernel(BroadcastArgs, inputs);\n}\nvar broadcastArgs = op({ broadcastArgs_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_to.js\nfunction broadcastTo_(x, shape) {\n  let input2 = convertToTensor(x, \"broadcastTo\", \"x\");\n  const xShape = input2.shape;\n  if (shape.some((d) => !(d > 0) || d % 1 !== 0)) {\n    throw new Error(`broadcastTo(): Invalid broadcast shape [${shape}].`);\n  }\n  if (shape.length < input2.rank) {\n    throw new Error(`broadcastTo(): shape.length=${shape.length} < input.rank=${input2.rank}.`);\n  }\n  if (shape.length > input2.rank) {\n    const newShape = input2.shape.slice();\n    while (newShape.length < shape.length) {\n      newShape.unshift(1);\n    }\n    input2 = reshape(input2, newShape);\n  }\n  const inputShape = input2.shape;\n  const reps = Array.from(shape);\n  for (let i = shape.length - 1; i >= 0; i--) {\n    if (inputShape[i] === shape[i]) {\n      reps[i] = 1;\n    } else if (input2.shape[i] !== 1) {\n      throw new Error(`broadcastTo(): [${xShape}] cannot be broadcast to [${shape}].`);\n    }\n  }\n  const axes = reps.map((n, i) => n > 1 ? i : -1).filter((i) => i >= 0);\n  if (axes.length === 0) {\n    return clone(input2);\n  }\n  const inputs = { x: input2 };\n  const attrs = { reps };\n  return ENGINE.runKernel(Tile, inputs, attrs);\n}\nvar broadcastTo = op({ broadcastTo_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ceil.js\nfunction ceil_(x) {\n  const $x = convertToTensor(x, \"x\", \"ceil\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Ceil, inputs);\n}\nvar ceil = op({ ceil_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fill.js\nfunction fill(shape, value, dtype) {\n  const attrs = { shape, value, dtype };\n  return ENGINE.runKernel(Fill, {}, attrs);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/clip_by_value.js\nfunction clipByValue_(x, clipValueMin, clipValueMax) {\n  const $x = convertToTensor(x, \"x\", \"clipByValue\");\n  assert(clipValueMin <= clipValueMax, () => `Error in clip: min (${clipValueMin}) must be less than or equal to max (${clipValueMax}).`);\n  if (clipValueMin === clipValueMax) {\n    return fill($x.shape, clipValueMin, $x.dtype);\n  }\n  const inputs = { x: $x };\n  const attrs = { clipValueMin, clipValueMax };\n  return ENGINE.runKernel(ClipByValue, inputs, attrs);\n}\nvar clipByValue = op({ clipByValue_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat_1d.js\nfunction concat1d_(tensors) {\n  return concat(tensors, 0);\n}\nvar concat1d = op({ concat1d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat_2d.js\nfunction concat2d_(tensors, axis) {\n  return concat(tensors, axis);\n}\nvar concat2d = op({ concat2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat_3d.js\nfunction concat3d_(tensors, axis) {\n  return concat(tensors, axis);\n}\nvar concat3d = op({ concat3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat_4d.js\nfunction concat4d_(tensors, axis) {\n  return concat(tensors, axis);\n}\nvar concat4d = op({ concat4d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d.js\nfunction conv2d_(x, filter, strides, pad3, dataFormat = \"NHWC\", dilations = [1, 1], dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"conv2d\", \"float32\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv2d\", \"float32\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in conv2d: input must be rank 4, but got rank ${x4D.rank}.`);\n  assert($filter.rank === 4, () => `Error in conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);\n  checkPadOnDimRoundingMode(\"conv2d\", pad3, dimRoundingMode);\n  const inDepth = dataFormat === \"NHWC\" ? x4D.shape[3] : x4D.shape[1];\n  assert(inDepth === $filter.shape[2], () => `Error in conv2d: depth of input (${inDepth}) must match input depth for filter ${$filter.shape[2]}.`);\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const inputs = { x: x4D, filter: $filter };\n  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };\n  const res = ENGINE.runKernel(Conv2D, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar conv2d = op({ conv2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv1d.js\nfunction conv1d_(x, filter, stride, pad3, dataFormat = \"NWC\", dilation = 1, dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"conv1d\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv1d\");\n  let x3D = $x;\n  let reshapedTo3D = false;\n  if ($x.rank === 2) {\n    reshapedTo3D = true;\n    x3D = reshape($x, [1, $x.shape[0], $x.shape[1]]);\n  }\n  assert(x3D.rank === 3, () => `Error in conv1d: input must be rank 3, but got rank ${x3D.rank}.`);\n  assert($filter.rank === 3, () => `Error in conv1d: filter must be rank 3, but got rank ${$filter.rank}.`);\n  checkPadOnDimRoundingMode(\"conv1d\", pad3, dimRoundingMode);\n  assert(x3D.shape[2] === $filter.shape[1], () => `Error in conv1d: depth of input (${x3D.shape[2]}) must match input depth for filter ${$filter.shape[1]}.`);\n  assert(eitherStridesOrDilationsAreOne(stride, dilation), () => `Error in conv1D: Either stride or dilation must be 1. Got stride ${stride} and dilation '${dilation}'`);\n  assert(dataFormat === \"NWC\", () => `Error in conv1d: got dataFormat of ${dataFormat} but only NWC is currently supported.`);\n  const filter4D = reshape($filter, [1, $filter.shape[0], $filter.shape[1], $filter.shape[2]]);\n  const input4D = reshape(x3D, [x3D.shape[0], 1, x3D.shape[1], x3D.shape[2]]);\n  const strides = [1, stride];\n  const dilations = [1, dilation];\n  const conv2dDataFormat = \"NHWC\";\n  const res = conv2d(input4D, filter4D, strides, pad3, conv2dDataFormat, dilations, dimRoundingMode);\n  if (reshapedTo3D) {\n    return reshape(res, [res.shape[2], res.shape[3]]);\n  }\n  return reshape(res, [res.shape[0], res.shape[2], res.shape[3]]);\n}\nvar conv1d = op({ conv1d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_backprop_input.js\nfunction conv2DBackpropInput_(xShape, dy, filter, strides, pad3, dataFormat = \"NHWC\", dimRoundingMode) {\n  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);\n  let xShape4D = xShape;\n  let dy4D = dy;\n  let reshapedTo4D = false;\n  if (dy.rank === 3) {\n    reshapedTo4D = true;\n    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);\n    xShape4D = [1, xShape[0], xShape[1], xShape[2]];\n  }\n  assert(xShape4D.length === 4, () => `Error in conv2dDerInput: inShape must be length 4, but got length ${xShape4D.length}.`);\n  assert(dy4D.rank === 4, () => `Error in conv2dDerInput: dy must be rank 4, but got rank ${dy4D.rank}`);\n  assert(filter.rank === 4, () => `Error in conv2dDerInput: filter must be rank 4, but got rank ${filter.rank}`);\n  const inDepth = dataFormat === \"NHWC\" ? xShape4D[3] : xShape4D[1];\n  const outDepth = dataFormat === \"NHWC\" ? dy4D.shape[3] : dy4D.shape[1];\n  assert(inDepth === filter.shape[2], () => `Error in conv2dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[2]}.`);\n  assert(outDepth === filter.shape[3], () => `Error in conv2dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[3]}.`);\n  checkPadOnDimRoundingMode(\"conv2dDerInput\", pad3, dimRoundingMode);\n  const inputs = { dy: dy4D, filter };\n  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape: xShape4D };\n  const res = ENGINE.runKernel(Conv2DBackpropInput, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar conv2DBackpropInput = op({ conv2DBackpropInput_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_transpose.js\nfunction conv2dTranspose_(x, filter, outputShape, strides, pad3, dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"conv2dTranspose\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv2dTranspose\");\n  return conv2DBackpropInput(outputShape, $x, $filter, strides, pad3, \"NHWC\", dimRoundingMode);\n}\nvar conv2dTranspose = op({ conv2dTranspose_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d.js\nfunction conv3d_(x, filter, strides, pad3, dataFormat = \"NDHWC\", dilations = [1, 1, 1]) {\n  const $x = convertToTensor(x, \"x\", \"conv3d\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv3d\");\n  let x5D = $x;\n  let reshapedTo5D = false;\n  if ($x.rank === 4) {\n    reshapedTo5D = true;\n    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);\n  }\n  assert(x5D.rank === 5, () => `Error in conv3d: input must be rank 5, but got rank ${x5D.rank}.`);\n  assert($filter.rank === 5, () => `Error in conv3d: filter must be rank 5, but got rank ${$filter.rank}.`);\n  assert(x5D.shape[4] === $filter.shape[3], () => `Error in conv3d: depth of input (${x5D.shape[4]}) must match input depth for filter ${$filter.shape[3]}.`);\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv3D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  assert(dataFormat === \"NDHWC\", () => `Error in conv3d: got dataFormat of ${dataFormat} but only NDHWC is currently supported.`);\n  const inputs = { x: x5D, filter: $filter };\n  const attrs = { strides, pad: pad3, dataFormat, dilations };\n  const res = ENGINE.runKernel(Conv3D, inputs, attrs);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar conv3d = op({ conv3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_backprop_input.js\nfunction conv3DBackpropInput_(xShape, dy, filter, strides, pad3) {\n  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);\n  let xShape5D = xShape;\n  let dy5D = dy;\n  let reshapedTo5D = false;\n  if (dy.rank === 4) {\n    reshapedTo5D = true;\n    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);\n    xShape5D = [1, xShape[0], xShape[1], xShape[2], xShape[3]];\n  }\n  const inDepth = xShape5D[4];\n  const outDepth = dy5D.shape[4];\n  assert(xShape5D.length === 5, () => `Error in conv3dDerInput: inShape must be length 5, but got length ${xShape5D.length}.`);\n  assert(dy5D.rank === 5, () => `Error in conv3dDerInput: dy must be rank 5, but got rank ${dy5D.rank}`);\n  assert(filter.rank === 5, () => `Error in conv3dDerInput: filter must be rank 5, but got rank ${filter.rank}`);\n  assert(inDepth === filter.shape[3], () => `Error in conv3dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[3]}.`);\n  assert(outDepth === filter.shape[4], () => `Error in conv3dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[4]}.`);\n  const inputs = { dy: dy5D, filter };\n  const attrs = { pad: pad3, strides, inputShape: xShape5D };\n  const res = ENGINE.runKernel(Conv3DBackpropInputV2, inputs, attrs);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar conv3DBackpropInput = op({ conv3DBackpropInput_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_transpose.js\nfunction conv3dTranspose_(x, filter, outputShape, strides, pad3) {\n  const $x = convertToTensor(x, \"x\", \"conv3dTranspose\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv3dTranspose\");\n  return conv3DBackpropInput(outputShape, $x, $filter, strides, pad3);\n}\nvar conv3dTranspose = op({ conv3dTranspose_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/cos.js\nfunction cos_(x) {\n  const $x = convertToTensor(x, \"x\", \"cos\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Cos, inputs);\n}\nvar cos = op({ cos_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/cosh.js\nfunction cosh_(x) {\n  const $x = convertToTensor(x, \"x\", \"cosh\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Cosh, inputs);\n}\nvar cosh = op({ cosh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/cumprod.js\nfunction cumprod_(x, axis = 0, exclusive = false, reverse5 = false) {\n  const $x = convertToTensor(x, \"x\", \"cumprod\");\n  const inputs = { x: $x };\n  const attrs = { axis, exclusive, reverse: reverse5 };\n  return ENGINE.runKernel(Cumprod, inputs, attrs);\n}\nvar cumprod = op({ cumprod_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/cumsum.js\nfunction cumsum_(x, axis = 0, exclusive = false, reverse5 = false) {\n  const $x = convertToTensor(x, \"x\", \"cumsum\");\n  const inputs = { x: $x };\n  const attrs = { axis, exclusive, reverse: reverse5 };\n  return ENGINE.runKernel(Cumsum, inputs, attrs);\n}\nvar cumsum = op({ cumsum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/dense_bincount.js\nfunction denseBincount_(x, weights, size, binaryOutput = false) {\n  const $x = convertToTensor(x, \"x\", \"denseBincount\");\n  const $weights = convertToTensor(weights, \"weights\", \"denseBincount\");\n  assert($x.dtype === \"int32\", () => `Error in denseBincount: input dtype must be int32, but got ${$x.dtype}`);\n  assert($x.rank <= 2, () => `Error in denseBincount: input must be at most rank 2, but got rank ${$x.rank}.`);\n  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);\n  assert($weights.size === $x.size || $weights.size === 0, () => `Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${$x.shape}, weights shape: ${$weights.shape}.`);\n  const inputs = { x: $x, weights: $weights };\n  const attrs = { size, binaryOutput };\n  return ENGINE.runKernel(DenseBincount, inputs, attrs);\n}\nvar denseBincount = op({ denseBincount_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/depth_to_space.js\nfunction depthToSpace_(x, blockSize, dataFormat = \"NHWC\") {\n  const $x = convertToTensor(x, \"x\", \"depthToSpace\", \"float32\");\n  const inputHeight = dataFormat === \"NHWC\" ? $x.shape[1] : $x.shape[2];\n  const inputWidth = dataFormat === \"NHWC\" ? $x.shape[2] : $x.shape[3];\n  const inputDepth = dataFormat === \"NHWC\" ? $x.shape[3] : $x.shape[1];\n  assert(blockSize > 1, () => `blockSize should be > 1 for depthToSpace, but was: ${blockSize}`);\n  assert(inputHeight * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying\n    ${inputHeight} and ${blockSize}  for depthToSpace with input shape\n    ${$x.shape}`);\n  assert(inputWidth * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying\n    ${inputWidth} and ${blockSize} for depthToSpace with input shape\n        ${$x.shape}`);\n  assert(inputDepth % (blockSize * blockSize) === 0, () => `Dimension size must be evenly divisible by ${blockSize * blockSize} but is ${inputDepth} for depthToSpace with input shape ${$x.shape}`);\n  const inputs = { x: $x };\n  const attrs = { blockSize, dataFormat };\n  return ENGINE.runKernel(DepthToSpace, inputs, attrs);\n}\nvar depthToSpace = op({ depthToSpace_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d.js\nfunction depthwiseConv2d_(x, filter, strides, pad3, dataFormat = \"NHWC\", dilations = [1, 1], dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"depthwiseConv2d\", \"float32\");\n  const $filter = convertToTensor(filter, \"filter\", \"depthwiseConv2d\", \"float32\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);\n  assert($filter.rank === 4, () => `Error in depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);\n  const inChannels = dataFormat === \"NHWC\" ? x4D.shape[3] : x4D.shape[1];\n  assert(inChannels === $filter.shape[2], () => `Error in depthwiseConv2d: number of input channels (${inChannels}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);\n  checkPadOnDimRoundingMode(\"depthwiseConv2d\", pad3, dimRoundingMode);\n  const inputs = { x: x4D, filter: $filter };\n  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };\n  const res = ENGINE.runKernel(DepthwiseConv2dNative, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar depthwiseConv2d = op({ depthwiseConv2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/diag.js\nfunction diag_(x) {\n  const $x = convertToTensor(x, \"x\", \"diag\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Diag, inputs);\n}\nvar diag = op({ diag_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/dilation2d.js\nfunction dilation2d_(x, filter, strides, pad3, dilations = [1, 1], dataFormat = \"NHWC\") {\n  const $x = convertToTensor(x, \"x\", \"dilation2d\");\n  const $filter = convertToTensor(filter, \"filter\", \"dilation2d\");\n  assert($x.rank === 3 || $x.rank === 4, () => `Error in dilation2d: input must be rank 3 or 4, but got rank ${$x.rank}.`);\n  assert($filter.rank === 3, () => `Error in dilation2d: filter must be rank 3, but got rank ${$filter.rank}.`);\n  assert(dataFormat === \"NHWC\", () => `Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${dataFormat}`);\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n    reshapedTo4D = true;\n  }\n  const inputs = { x: x4D, filter: $filter };\n  const attrs = { strides, pad: pad3, dilations };\n  const res = ENGINE.runKernel(Dilation2D, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar dilation2d = op({ dilation2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/equal.js\nfunction equal_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"equal\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"equal\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Equal, inputs);\n}\nvar equal = op({ equal_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/where.js\nfunction where_(condition, a, b) {\n  const $a = convertToTensor(a, \"a\", \"where\");\n  const $b = convertToTensor(b, \"b\", \"where\");\n  const $condition = convertToTensor(condition, \"condition\", \"where\", \"bool\");\n  const broadcastShape = assertAndGetBroadcastShape(assertAndGetBroadcastShape($condition.shape, $a.shape), $b.shape);\n  const $broadcastedCondition = broadcastTo($condition, broadcastShape);\n  const $broadcastedA = broadcastTo($a, broadcastShape);\n  const $broadcastedB = broadcastTo($b, broadcastShape);\n  const inputs = {\n    condition: $broadcastedCondition,\n    t: $broadcastedA,\n    e: $broadcastedB\n  };\n  return ENGINE.runKernel(Select, inputs);\n}\nvar where = op({ where_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/zeros_like.js\nfunction zerosLike_(x) {\n  const $x = convertToTensor(x, \"x\", \"zerosLike\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(ZerosLike, inputs);\n}\nvar zerosLike = op({ zerosLike_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/div_no_nan.js\nfunction divNoNan_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"div\");\n  let $b = convertToTensor(b, \"b\", \"div\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const divResult = div($a, $b);\n  const zeros4 = zerosLike(divResult);\n  const bEqualsZero = equal($b, zeros4);\n  return where(bEqualsZero, zeros4, divResult);\n}\nvar divNoNan = op({ divNoNan_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/dot.js\nfunction dot_(t1, t2) {\n  const $t1 = convertToTensor(t1, \"t1\", \"dot\");\n  const $t2 = convertToTensor(t2, \"t2\", \"dot\");\n  assert(($t1.rank === 1 || $t1.rank === 2) && ($t2.rank === 1 || $t2.rank === 2), () => `Error in dot: inputs must all be rank 1 or 2, but got ranks ${$t1.rank} and ${$t2.rank}.`);\n  const t1Inner = $t1.rank === 1 ? $t1.size : $t1.shape[1];\n  const t2Inner = $t2.rank === 1 ? $t2.size : $t2.shape[0];\n  assert(t1Inner === t2Inner, () => `Error in dot: inner dimensions of inputs must match, but got ${t1Inner} and ${t2Inner}.`);\n  if ($t1.rank === 1 && $t2.rank === 1) {\n    const t12D = reshape($t1, [1, -1]);\n    const t22D = reshape($t2, [-1, 1]);\n    const t1t2 = matMul(t12D, t22D);\n    return reshape(t1t2, []);\n  } else if ($t1.rank === 1 && $t2.rank === 2) {\n    const t12D = reshape($t1, [1, -1]);\n    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);\n    const t1t2 = matMul(t12D, t22D);\n    return reshape(t1t2, [t1t2.size]);\n  } else if ($t1.rank === 2 && $t2.rank === 1) {\n    const t22D = reshape($t2, [-1, 1]);\n    const t1t2 = matMul($t1, t22D);\n    return reshape(t1t2, [t1t2.size]);\n  } else {\n    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);\n    const t1t2 = matMul($t1, t22D);\n    return t1t2;\n  }\n}\nvar dot = op({ dot_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/einsum.js\nfunction einsum_(equation, ...tensors) {\n  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, \"einsum\"));\n  const attrs = { equation };\n  return ENGINE.runKernel(Einsum, $tensors, attrs);\n}\nvar einsum = op({ einsum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/elu.js\nfunction elu_(x) {\n  const $x = convertToTensor(x, \"x\", \"elu\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Elu, inputs);\n}\nvar elu = op({ elu_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/erf.js\nfunction erf_(x) {\n  let $x = convertToTensor(x, \"x\", \"erf\");\n  assert($x.dtype === \"int32\" || $x.dtype === \"float32\", () => \"Input dtype must be `int32` or `float32`.\");\n  if ($x.dtype === \"int32\") {\n    $x = cast($x, \"float32\");\n  }\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Erf, inputs);\n}\nvar erf = op({ erf_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/axis_util.js\nfunction axesAreInnerMostDims(axes, rank) {\n  for (let i = 0; i < axes.length; ++i) {\n    if (axes[axes.length - i - 1] !== rank - 1 - i) {\n      return false;\n    }\n  }\n  return true;\n}\nfunction combineLocations(outputLoc, reduceLoc, axes) {\n  const rank = outputLoc.length + reduceLoc.length;\n  const loc = [];\n  let outIdx = 0;\n  let reduceIdx = 0;\n  for (let dim = 0; dim < rank; dim++) {\n    if (axes.indexOf(dim) === -1) {\n      loc.push(outputLoc[outIdx++]);\n    } else {\n      loc.push(reduceLoc[reduceIdx++]);\n    }\n  }\n  return loc;\n}\nfunction computeOutAndReduceShapes(aShape, axes) {\n  const outShape = [];\n  const rank = aShape.length;\n  for (let dim = 0; dim < rank; dim++) {\n    if (axes.indexOf(dim) === -1) {\n      outShape.push(aShape[dim]);\n    }\n  }\n  const reduceShape = axes.map((dim) => aShape[dim]);\n  return [outShape, reduceShape];\n}\nfunction expandShapeToKeepDim(shape, axes) {\n  const reduceSubShape = axes.map((x) => 1);\n  return combineLocations(shape, reduceSubShape, axes);\n}\nfunction assertAxesAreInnerMostDims(msg, axes, rank) {\n  assert(axesAreInnerMostDims(axes, rank), () => `${msg} supports only inner-most axes for now. Got axes ${axes} and rank-${rank} input.`);\n}\nfunction getAxesPermutation(axes, rank) {\n  if (axesAreInnerMostDims(axes, rank)) {\n    return null;\n  }\n  const result = [];\n  for (let i = 0; i < rank; ++i) {\n    if (axes.indexOf(i) === -1) {\n      result.push(i);\n    }\n  }\n  axes.forEach((axis) => result.push(axis));\n  return result;\n}\nfunction getUndoAxesPermutation(axes) {\n  return axes.map((axis, i) => [i, axis]).sort((a, b) => a[1] - b[1]).map((x) => x[0]);\n}\nfunction getInnerMostAxes(numAxes, rank) {\n  const res = [];\n  for (let i = rank - numAxes; i < rank; ++i) {\n    res.push(i);\n  }\n  return res;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max.js\nfunction max_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"max\");\n  const inputs = { x: $x };\n  const attrs = { reductionIndices: axis, keepDims };\n  return ENGINE.runKernel(Max, inputs, attrs);\n}\nvar max = op({ max_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/min.js\nfunction min_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"min\");\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(Min, inputs, attrs);\n}\nvar min = op({ min_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pow.js\nfunction pow_(base, exp4) {\n  let $base = convertToTensor(base, \"base\", \"pow\");\n  let $exp = convertToTensor(exp4, \"exp\", \"pow\");\n  [$base, $exp] = makeTypesMatch($base, $exp);\n  const inputs = { a: $base, b: $exp };\n  return ENGINE.runKernel(Pow, inputs);\n}\nvar pow = op({ pow_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/scalar.js\nfunction scalar(value, dtype) {\n  if ((isTypedArray(value) && dtype !== \"string\" || Array.isArray(value)) && dtype !== \"complex64\") {\n    throw new Error(\"Error creating a new Scalar: value must be a primitive (number|boolean|string)\");\n  }\n  if (dtype === \"string\" && isTypedArray(value) && !(value instanceof Uint8Array)) {\n    throw new Error(\"When making a scalar from encoded string, the value must be `Uint8Array`.\");\n  }\n  const shape = [];\n  const inferredShape = [];\n  return makeTensor(value, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sqrt.js\nfunction sqrt_(x) {\n  const $x = convertToTensor(x, \"x\", \"sqrt\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Sqrt, inputs);\n}\nvar sqrt = op({ sqrt_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/square.js\nfunction square_(x) {\n  const $x = convertToTensor(x, \"x\", \"square\");\n  const attrs = {};\n  return ENGINE.runKernel(\"Square\", { x: $x }, attrs);\n}\nvar square = op({ square_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sum.js\nfunction sum_(x, axis = null, keepDims = false) {\n  let $x = convertToTensor(x, \"x\", \"sum\");\n  if ($x.dtype === \"bool\") {\n    $x = cast($x, \"int32\");\n  }\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(Sum, inputs, attrs);\n}\nvar sum2 = op({ sum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/norm.js\nfunction norm_(x, ord = \"euclidean\", axis = null, keepDims = false) {\n  x = convertToTensor(x, \"x\", \"norm\");\n  const norm2 = normImpl(x, ord, axis);\n  let keepDimsShape = norm2.shape;\n  if (keepDims) {\n    const axes = parseAxisParam(axis, x.shape);\n    keepDimsShape = expandShapeToKeepDim(norm2.shape, axes);\n  }\n  return reshape(norm2, keepDimsShape);\n}\nfunction normImpl(x, p2, axis = null) {\n  if (x.rank === 0) {\n    return abs(x);\n  }\n  if (x.rank !== 1 && axis === null) {\n    return normImpl(reshape(x, [-1]), p2, axis);\n  }\n  if (x.rank === 1 || typeof axis === \"number\" || Array.isArray(axis) && axis.length === 1) {\n    if (p2 === 1) {\n      return sum2(abs(x), axis);\n    }\n    if (p2 === Infinity) {\n      return max(abs(x), axis);\n    }\n    if (p2 === -Infinity) {\n      return min(abs(x), axis);\n    }\n    if (p2 === \"euclidean\" || p2 === 2) {\n      return sqrt(sum2(pow(abs(x), scalar(2, \"int32\")), axis));\n    }\n    throw new Error(`Error in norm: invalid ord value: ${p2}`);\n  }\n  if (Array.isArray(axis) && axis.length === 2) {\n    if (p2 === 1) {\n      return max(sum2(abs(x), axis[0]), axis[1] - 1);\n    }\n    if (p2 === Infinity) {\n      return max(sum2(abs(x), axis[1]), axis[0]);\n    }\n    if (p2 === -Infinity) {\n      return min(sum2(abs(x), axis[1]), axis[0]);\n    }\n    if (p2 === \"fro\" || p2 === \"euclidean\") {\n      return sqrt(sum2(square(x), axis));\n    }\n    throw new Error(`Error in norm: invalid ord value: ${p2}`);\n  }\n  throw new Error(`Error in norm: invalid axis: ${axis}`);\n}\nvar norm = op({ norm_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/euclidean_norm.js\nfunction euclideanNorm_(x, axis = null, keepDims = false) {\n  return norm(x, \"euclidean\", axis, keepDims);\n}\nvar euclideanNorm = op({ euclideanNorm_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/exp.js\nfunction exp_(x) {\n  const $x = convertToTensor(x, \"x\", \"exp\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Exp, inputs);\n}\nvar exp = op({ exp_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/expand_dims.js\nfunction expandDims_(x, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"expandDims\", \"string_or_numeric\");\n  assert(axis <= $x.rank, () => \"Axis must be <= rank of the tensor\");\n  const inputs = { input: $x };\n  const attrs = { dim: axis };\n  return ENGINE.runKernel(ExpandDims, inputs, attrs);\n}\nvar expandDims = op({ expandDims_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/expm1.js\nfunction expm1_(x) {\n  const $x = convertToTensor(x, \"x\", \"expm1\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Expm1, inputs);\n}\nvar expm1 = op({ expm1_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tile.js\nfunction tile_(x, reps) {\n  const $x = convertToTensor(x, \"x\", \"tile\", \"string_or_numeric\");\n  assert($x.rank === reps.length, () => `Error in transpose: rank of input ${$x.rank} must match length of reps ${reps}.`);\n  const inputs = { x: $x };\n  const attrs = { reps };\n  return ENGINE.runKernel(Tile, inputs, attrs);\n}\nvar tile = op({ tile_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/eye.js\nfunction eye_(numRows, numColumns, batchShape, dtype = \"float32\") {\n  if (numColumns == null) {\n    numColumns = numRows;\n  }\n  const buff = buffer([numRows, numColumns], dtype);\n  const n = numRows <= numColumns ? numRows : numColumns;\n  for (let i = 0; i < n; ++i) {\n    buff.set(1, i, i);\n  }\n  const out = reshape(buff.toTensor(), [numRows, numColumns]);\n  if (batchShape == null) {\n    return out;\n  } else {\n    if (batchShape.length === 1) {\n      return tile(expandDims(out, 0), [batchShape[0], 1, 1]);\n    } else if (batchShape.length === 2) {\n      return tile(expandDims(expandDims(out, 0), 0), [batchShape[0], batchShape[1], 1, 1]);\n    } else if (batchShape.length === 3) {\n      return tile(expandDims(expandDims(expandDims(out, 0), 0), 0), [\n        batchShape[0],\n        batchShape[1],\n        batchShape[2],\n        1,\n        1\n      ]);\n    } else {\n      throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${batchShape.length}D.`);\n    }\n  }\n}\nvar eye = op({ eye_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/floor.js\nfunction floor_(x) {\n  const $x = convertToTensor(x, \"x\", \"floor\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Floor, inputs);\n}\nvar floor = op({ floor_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/gather.js\nfunction gather_(x, indices, axis = 0, batchDims = 0) {\n  const $x = convertToTensor(x, \"x\", \"gather\");\n  const $indices = convertToTensor(indices, \"indices\", \"gather\", \"int32\");\n  const inputs = { x: $x, indices: $indices };\n  const attrs = { axis, batchDims };\n  return ENGINE.runKernel(GatherV2, inputs, attrs);\n}\nvar gather = op({ gather_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/greater.js\nfunction greater_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"greater\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"greater\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Greater, inputs);\n}\nvar greater = op({ greater_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/greater_equal.js\nfunction greaterEqual_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"greaterEqual\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"greaterEqual\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(GreaterEqual, inputs);\n}\nvar greaterEqual = op({ greaterEqual_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/is_finite.js\nfunction isFinite_(x) {\n  const $x = convertToTensor(x, \"x\", \"isFinite\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(IsFinite, inputs);\n}\nvar isFinite2 = op({ isFinite_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/is_inf.js\nfunction isInf_(x) {\n  const $x = convertToTensor(x, \"x\", \"isInf\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(IsInf, inputs);\n}\nvar isInf = op({ isInf_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/is_nan.js\nfunction isNaN_(x) {\n  const $x = convertToTensor(x, \"x\", \"isNaN\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(IsNan, inputs);\n}\nvar isNaN2 = op({ isNaN_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/leaky_relu.js\nfunction leakyRelu_(x, alpha = 0.2) {\n  const $x = convertToTensor(x, \"x\", \"leakyRelu\");\n  const inputs = { x: $x };\n  const attrs = { alpha };\n  return ENGINE.runKernel(LeakyRelu, inputs, attrs);\n}\nvar leakyRelu = op({ leakyRelu_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/less.js\nfunction less_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"less\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"less\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Less, inputs);\n}\nvar less = op({ less_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/less_equal.js\nfunction lessEqual_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"lessEqual\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"lessEqual\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(LessEqual, inputs);\n}\nvar lessEqual = op({ lessEqual_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/linspace.js\nfunction linspace(start, stop, num) {\n  if (num <= 0) {\n    throw new Error(\"The number of values should be positive.\");\n  }\n  const attrs = { start, stop, num };\n  return ENGINE.runKernel(LinSpace, {}, attrs);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/local_response_normalization.js\nfunction localResponseNormalization_(x, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {\n  const $x = convertToTensor(x, \"x\", \"localResponseNormalization\");\n  assert($x.rank === 4 || $x.rank === 3, () => `Error in localResponseNormalization: x must be rank 3 or 4 but got\n               rank ${$x.rank}.`);\n  assert(isInt(depthRadius), () => `Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${depthRadius}.`);\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  const inputs = { x: x4D };\n  const attrs = { depthRadius, bias, alpha, beta };\n  const res = ENGINE.runKernel(LRN, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  } else {\n    return res;\n  }\n}\nvar localResponseNormalization = op({ localResponseNormalization_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/log.js\nfunction log_(x) {\n  const $x = convertToTensor(x, \"x\", \"log\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Log, inputs);\n}\nvar log2 = op({ log_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/log1p.js\nfunction log1p_(x) {\n  const $x = convertToTensor(x, \"x\", \"log1p\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Log1p, inputs);\n}\nvar log1p = op({ log1p_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients.js\nfunction grad(f) {\n  assert(isFunction(f), () => \"The f passed in grad(f) must be a function\");\n  return (x, dy) => {\n    const $x = convertToTensor(x, \"x\", \"tf.grad\", \"string_or_numeric\");\n    const $dy = dy != null ? convertToTensor(dy, \"dy\", \"tf.grad\") : null;\n    return ENGINE.tidy(() => {\n      const { value, grads: grads2 } = ENGINE.gradients(() => f($x), [$x], $dy);\n      if ($dy != null) {\n        assertShapesMatch(value.shape, $dy.shape, \"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)\");\n      }\n      checkGrads(grads2);\n      return grads2[0];\n    });\n  };\n}\nfunction grads(f) {\n  assert(isFunction(f), () => \"The f passed in grads(f) must be a function\");\n  return (args, dy) => {\n    assert(Array.isArray(args), () => \"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s\");\n    const $args = convertToTensorArray(args, \"args\", \"tf.grads\", \"string_or_numeric\");\n    const $dy = dy != null ? convertToTensor(dy, \"dy\", \"tf.grads\") : null;\n    return ENGINE.tidy(() => {\n      const { value, grads: grads2 } = ENGINE.gradients(() => f(...$args), $args, $dy);\n      if ($dy != null) {\n        assertShapesMatch(value.shape, $dy.shape, \"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])\");\n      }\n      checkGrads(grads2);\n      return grads2;\n    });\n  };\n}\nfunction valueAndGrad(f) {\n  assert(isFunction(f), () => \"The f passed in valueAndGrad(f) must be a function\");\n  return (x, dy) => {\n    assert(x instanceof Tensor, () => \"The x passed in valueAndGrad(f)(x) must be a tensor\");\n    assert(dy == null || dy instanceof Tensor, () => \"The dy passed in valueAndGrad(f)(x, dy) must be a tensor\");\n    const { grads: grads2, value } = ENGINE.gradients(() => f(x), [x], dy);\n    checkGrads(grads2);\n    return { grad: grads2[0], value };\n  };\n}\nfunction valueAndGrads(f) {\n  assert(isFunction(f), () => \"The f passed in valueAndGrads(f) must be a function\");\n  return (args, dy) => {\n    assert(Array.isArray(args) && args.every((arg) => arg instanceof Tensor), () => \"The args passed in valueAndGrads(f)(args) must be array of tensors\");\n    assert(dy == null || dy instanceof Tensor, () => \"The dy passed in valueAndGrads(f)(args, dy) must be a tensor\");\n    const res = ENGINE.gradients(() => f(...args), args, dy);\n    if (dy != null) {\n      assertShapesMatch(res.value.shape, dy.shape, \"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])\");\n    }\n    checkGrads(res.grads);\n    return res;\n  };\n}\nfunction variableGrads(f, varList) {\n  assert(isFunction(f), () => \"The f passed in variableGrads(f) must be a function\");\n  assert(varList == null || Array.isArray(varList) && varList.every((v) => v instanceof Variable), () => \"The varList passed in variableGrads(f, varList) must be an array of variables\");\n  const specifiedVarList = varList != null;\n  if (!specifiedVarList) {\n    varList = [];\n    for (const varName in ENGINE.registeredVariables) {\n      varList.push(ENGINE.registeredVariables[varName]);\n    }\n  }\n  const specifiedNonTrainable = specifiedVarList ? varList.filter((variable2) => !variable2.trainable) : null;\n  const originalVarCount = varList.length;\n  varList = varList.filter((variable2) => variable2.trainable);\n  assert(varList.length > 0, () => `variableGrads() expects at least one of the input variables to be trainable, but none of the ${originalVarCount} variables is trainable.`);\n  const allowNoGradients = true;\n  const { value, grads: grads2 } = ENGINE.gradients(f, varList, null, allowNoGradients);\n  assert(grads2.some((g) => g != null), () => \"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize().\");\n  assert(value.rank === 0, () => `The f passed in variableGrads(f) must return a scalar, but it returned a rank-${value.rank} tensor`);\n  const namedGrads = {};\n  varList.forEach((v, i) => {\n    if (grads2[i] != null) {\n      namedGrads[v.name] = grads2[i];\n    }\n  });\n  if (specifiedNonTrainable != null) {\n    specifiedNonTrainable.forEach((v) => namedGrads[v.name] = null);\n  }\n  return { value, grads: namedGrads };\n}\nfunction customGrad(f) {\n  return ENGINE.customGrad(f);\n}\nfunction checkGrads(grads2) {\n  const numNullGradients = grads2.filter((g) => g == null).length;\n  if (numNullGradients > 0) {\n    throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that\n    the f you passed encloses all operations that lead from x to y.`);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/softplus.js\nfunction softplus_(x) {\n  const $x = convertToTensor(x, \"x\", \"softplus\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Softplus, inputs);\n}\nvar softplus = op({ softplus_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/log_sigmoid.js\nfunction logSigmoid_(x) {\n  const $x = convertToTensor(x, \"x\", \"logSigmoid\");\n  const customOp = customGrad((x2) => {\n    const value = neg(softplus(neg(x2)));\n    const gradFunc = (dy) => {\n      const derX = mul(dy, sigmoid(neg(x2)));\n      return derX;\n    };\n    return { value, gradFunc };\n  });\n  return customOp($x);\n}\nvar logSigmoid = op({ logSigmoid_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sub.js\nfunction sub_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"sub\");\n  let $b = convertToTensor(b, \"b\", \"sub\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Sub, inputs);\n}\nvar sub = op({ sub_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/log_softmax.js\nfunction logSoftmax_(logits, axis = -1) {\n  const $logits = convertToTensor(logits, \"logits\", \"logSoftmax\");\n  if (axis === -1) {\n    axis = $logits.rank - 1;\n  }\n  if (axis !== $logits.rank - 1) {\n    throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and axis was ${axis}`);\n  }\n  const customOp = customGrad((logits2, save) => {\n    const keepDims = true;\n    const xMax = max(logits2, axis, true);\n    const shifted = sub(logits2, xMax);\n    const value = sub(cast(shifted, \"float32\"), log2(sum2(exp(shifted), axis, keepDims)));\n    save([value]);\n    const gradFunc = (dy, saved) => {\n      const [value2] = saved;\n      const keepDims2 = true;\n      const softmax6 = exp(value2);\n      return sub(dy, mul(sum2(dy, axis, keepDims2), softmax6));\n    };\n    return { value, gradFunc };\n  });\n  return customOp($logits);\n}\nvar logSoftmax = op({ logSoftmax_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/log_sum_exp.js\nfunction logSumExp_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"logSumExp\");\n  const axes = parseAxisParam(axis, $x.shape);\n  const xMax = max($x, axes, true);\n  const a = sub($x, xMax);\n  const b = exp(a);\n  const c = sum2(b, axes);\n  const d = log2(c);\n  const res = add2(reshape(xMax, d.shape), d);\n  if (keepDims) {\n    const newShape = expandShapeToKeepDim(res.shape, axes);\n    return reshape(res, newShape);\n  }\n  return res;\n}\nvar logSumExp = op({ logSumExp_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/logical_and.js\nfunction logicalAnd_(a, b) {\n  const $a = convertToTensor(a, \"a\", \"logicalAnd\", \"bool\");\n  const $b = convertToTensor(b, \"b\", \"logicalAnd\", \"bool\");\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(LogicalAnd, inputs);\n}\nvar logicalAnd = op({ logicalAnd_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/logical_not.js\nfunction logicalNot_(x) {\n  const $x = convertToTensor(x, \"x\", \"logicalNot\", \"bool\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(LogicalNot, inputs);\n}\nvar logicalNot = op({ logicalNot_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/logical_or.js\nfunction logicalOr_(a, b) {\n  const $a = convertToTensor(a, \"a\", \"logicalOr\", \"bool\");\n  const $b = convertToTensor(b, \"b\", \"logicalOr\", \"bool\");\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(LogicalOr, inputs);\n}\nvar logicalOr = op({ logicalOr_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/logical_xor.js\nfunction logicalXor_(a, b) {\n  const $a = convertToTensor(a, \"a\", \"logicalXor\", \"bool\");\n  const $b = convertToTensor(b, \"b\", \"logicalXor\", \"bool\");\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  return logicalAnd(logicalOr(a, b), logicalNot(logicalAnd(a, b)));\n}\nvar logicalXor = op({ logicalXor_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/search_sorted.js\nvar INT32_MAX = 2147483648;\nfunction searchSorted_(sortedSequence, values, side = \"left\") {\n  const $sortedSequence = convertToTensor(sortedSequence, \"sortedSequence\", \"searchSorted\");\n  const $values = convertToTensor(values, \"values\", \"searchSorted\");\n  const sequenceSize = $sortedSequence.shape[$sortedSequence.shape.length - 1];\n  const valuesSize = $values.shape[$values.shape.length - 1];\n  const $sortedSequence2D = reshape($sortedSequence, [-1, sequenceSize]);\n  const $values2D = reshape($values, [-1, valuesSize]);\n  if ($sortedSequence2D.rank < 2) {\n    throw new Error(`Sorted input argument must be at least 2-dimensional`);\n  }\n  if ($sortedSequence2D.shape[0] !== $values2D.shape[0]) {\n    throw new Error(`Leading dimension of 'sortedSequence' and 'values' must match.`);\n  }\n  if (sizeFromShape($values2D.shape) >= INT32_MAX) {\n    throw new Error(`values tensor size must less than ${INT32_MAX}`);\n  }\n  if ($sortedSequence2D.shape[1] >= INT32_MAX) {\n    throw new Error(`trailing dim_size must less than ${INT32_MAX} for int32 output type, was ${$sortedSequence2D.shape[1]}`);\n  }\n  const inputs = {\n    sortedSequence: $sortedSequence2D,\n    values: $values2D\n  };\n  const attrs = { side };\n  return ENGINE.runKernel(SearchSorted, inputs, attrs);\n}\nvar searchSorted = op({ searchSorted_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/lower_bound.js\nfunction lowerBound(sortedSequence, values) {\n  return searchSorted(sortedSequence, values, \"left\");\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool.js\nfunction maxPool_(x, filterSize, strides, pad3, dimRoundingMode) {\n  const $x = convertToTensor(x, \"x\", \"maxPool\");\n  const dilations = 1;\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x4D.rank}.`);\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  checkPadOnDimRoundingMode(\"maxPool\", pad3, dimRoundingMode);\n  const inputs = { x: x4D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };\n  const res = ENGINE.runKernel(MaxPool, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar maxPool = op({ maxPool_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_3d.js\nfunction maxPool3d_(x, filterSize = [1, 1, 1], strides, pad3, dimRoundingMode, dataFormat = \"NDHWC\") {\n  const $x = convertToTensor(x, \"x\", \"maxPool3d\");\n  let x5D = $x;\n  let reshapedTo5D = false;\n  if ($x.rank === 4) {\n    reshapedTo5D = true;\n    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);\n  }\n  assert(x5D.rank === 5, () => `Error in maxPool3d: x must be rank 5 but got rank ${x5D.rank}.`);\n  assert(dataFormat === \"NDHWC\", () => `Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);\n  checkPadOnDimRoundingMode(\"maxPool3d\", pad3, dimRoundingMode);\n  const inputs = { x: x5D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };\n  const res = ENGINE.runKernel(MaxPool3D, inputs, attrs);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar maxPool3d = op({ maxPool3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_with_argmax.js\nfunction maxPoolWithArgmax_(x, filterSize, strides, pad3, includeBatchInIndex = false) {\n  const $x = convertToTensor(x, \"x\", \"maxPoolWithArgmax\");\n  const inputs = { x: $x };\n  const attrs = { filterSize, strides, pad: pad3, includeBatchInIndex };\n  const result = ENGINE.runKernel(MaxPoolWithArgmax, inputs, attrs);\n  return { result: result[0], indexes: result[1] };\n}\nvar maxPoolWithArgmax = op({ maxPoolWithArgmax_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/maximum.js\nfunction maximum_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"maximum\");\n  let $b = convertToTensor(b, \"b\", \"maximum\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  if ($a.dtype === \"bool\") {\n    $a = cast($a, \"int32\");\n    $b = cast($b, \"int32\");\n  }\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Maximum, inputs);\n}\nvar maximum = op({ maximum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/mean.js\nfunction mean_(x, axis = null, keepDims = false) {\n  const $x = convertToTensor(x, \"x\", \"mean\");\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(Mean, inputs, attrs);\n}\nvar mean = op({ mean_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/zeros.js\nfunction zeros(shape, dtype = \"float32\") {\n  if (dtype === \"complex64\") {\n    const real4 = zeros(shape, \"float32\");\n    const imag4 = zeros(shape, \"float32\");\n    return complex(real4, imag4);\n  }\n  const values = makeZerosTypedArray(sizeFromShape(shape), dtype);\n  return ENGINE.makeTensor(values, shape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ones.js\nfunction ones2(shape, dtype = \"float32\") {\n  if (dtype === \"complex64\") {\n    const real4 = ones2(shape, \"float32\");\n    const imag4 = zeros(shape, \"float32\");\n    return complex(real4, imag4);\n  }\n  const values = makeOnesTypedArray(sizeFromShape(shape), dtype);\n  return ENGINE.makeTensor(values, shape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/meshgrid.js\nfunction meshgrid(x, y, { indexing = \"xy\" } = {}) {\n  if (indexing !== \"xy\" && indexing !== \"ij\") {\n    throw new TypeError(`${indexing} is not a valid third argument to meshgrid`);\n  }\n  if (x === void 0) {\n    return [];\n  }\n  let $x = convertToTensor(x, \"x\", \"meshgrid\", x instanceof Tensor ? x.dtype : \"float32\");\n  if (y === void 0) {\n    return [$x];\n  }\n  let $y = convertToTensor(y, \"y\", \"meshgrid\", y instanceof Tensor ? y.dtype : \"float32\");\n  const w = sizeFromShape($x.shape);\n  const h = sizeFromShape($y.shape);\n  if (indexing === \"xy\") {\n    $x = reshape($x, [1, -1]);\n    $y = reshape($y, [-1, 1]);\n    return [\n      matMul(ones2([h, 1], $x.dtype), $x),\n      matMul($y, ones2([1, w], $y.dtype))\n    ];\n  }\n  $x = reshape($x, [-1, 1]);\n  $y = reshape($y, [1, -1]);\n  return [\n    matMul($x, ones2([1, h], $x.dtype)),\n    matMul(ones2([w, 1], $y.dtype), $y)\n  ];\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/minimum.js\nfunction minimum_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"minimum\");\n  let $b = convertToTensor(b, \"b\", \"minimum\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  if ($a.dtype === \"bool\") {\n    $a = cast($a, \"int32\");\n    $b = cast($b, \"int32\");\n  }\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Minimum, inputs);\n}\nvar minimum = op({ minimum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/mirror_pad.js\nfunction mirrorPad_(x, paddings, mode) {\n  assert(mode === \"reflect\" || mode === \"symmetric\", () => `Invalid mode. Mode must be either reflect or symmetric. Got ${mode}.`);\n  const $x = convertToTensor(x, \"x\", \"mirrorPad\");\n  if ($x.rank === 0) {\n    throw new Error(\"mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad\");\n  }\n  assert(paddings.length === $x.rank, () => `Padding doesn't match input. Must be ${$x.rank}. Got ${paddings.length}.`);\n  const shapeOffset = mode === \"reflect\" ? 1 : 0;\n  for (let i = 0; i < $x.rank; i++) {\n    assert(paddings[i].length === 2, () => `Invalid number of paddings. Must be length of 2 each.`);\n    assert(paddings[i][0] >= 0 && paddings[i][0] <= $x.shape[i] - shapeOffset && paddings[i][1] >= 0 && paddings[i][1] <= $x.shape[i] - shapeOffset, () => `Padding in dimension ${i} cannot be greater than or equal to ${$x.shape[i] - shapeOffset} or less than 0 for input of shape ${$x.shape}`);\n  }\n  const attrs = { paddings, mode };\n  const inputs = { x: $x };\n  return ENGINE.runKernel(MirrorPad, inputs, attrs);\n}\nvar mirrorPad = op({ mirrorPad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/mod.js\nfunction mod_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"mod\");\n  let $b = convertToTensor(b, \"b\", \"mod\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(Mod, inputs);\n}\nvar mod = op({ mod_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/moments.js\nfunction moments_(x, axis = null, keepDims = false) {\n  x = convertToTensor(x, \"x\", \"moments\");\n  const axes = parseAxisParam(axis, x.shape);\n  const xMean = mean(x, axes, keepDims);\n  let keepDimsShape = xMean.shape;\n  if (!keepDims) {\n    keepDimsShape = expandShapeToKeepDim(xMean.shape, axes);\n  }\n  const devSquared = square(sub(cast(x, \"float32\"), reshape(xMean, keepDimsShape)));\n  const variance = mean(devSquared, axes, keepDims);\n  return { mean: xMean, variance };\n}\nvar moments = op({ moments_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/multi_rnn_cell.js\nfunction multiRNNCell_(lstmCells, data, c, h) {\n  const $data = convertToTensor(data, \"data\", \"multiRNNCell\");\n  const $c = convertToTensorArray(c, \"c\", \"multiRNNCell\");\n  const $h = convertToTensorArray(h, \"h\", \"multiRNNCell\");\n  let input2 = $data;\n  const newStates = [];\n  for (let i = 0; i < lstmCells.length; i++) {\n    const output = lstmCells[i](input2, $c[i], $h[i]);\n    newStates.push(output[0]);\n    newStates.push(output[1]);\n    input2 = output[1];\n  }\n  const newC = [];\n  const newH = [];\n  for (let i = 0; i < newStates.length; i += 2) {\n    newC.push(newStates[i]);\n    newH.push(newStates[i + 1]);\n  }\n  return [newC, newH];\n}\nvar multiRNNCell = op({ multiRNNCell_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/multinomial.js\nfunction multinomial_(logits, numSamples, seed, normalized = false) {\n  const $logits = convertToTensor(logits, \"logits\", \"multinomial\");\n  const numOutcomes = $logits.size;\n  const origRank = $logits.rank;\n  if (numOutcomes < 2) {\n    throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${numOutcomes}.`);\n  }\n  if (origRank > 2) {\n    throw new Error(`Rank of probabilities must be 1 or 2, but is ${origRank}`);\n  }\n  seed = seed || Math.random();\n  const logits2D = origRank === 1 ? reshape($logits, [1, -1]) : $logits;\n  const inputs = { logits: logits2D };\n  const attrs = { numSamples, seed, normalized };\n  const res = ENGINE.runKernel(Multinomial, inputs, attrs);\n  return origRank === 1 ? reshape(res, [res.size]) : res;\n}\nvar multinomial = op({ multinomial_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/not_equal.js\nfunction notEqual_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"notEqual\", \"string_or_numeric\");\n  let $b = convertToTensor(b, \"b\", \"notEqual\", \"string_or_numeric\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  return ENGINE.runKernel(NotEqual, inputs);\n}\nvar notEqual = op({ notEqual_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ones_like.js\nfunction onesLike_(x) {\n  const $x = convertToTensor(x, \"x\", \"onesLike\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(OnesLike, inputs);\n}\nvar onesLike = op({ onesLike_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/outer_product.js\nfunction outerProduct_(v1, v2) {\n  const $v1 = convertToTensor(v1, \"v1\", \"outerProduct\");\n  const $v2 = convertToTensor(v2, \"v2\", \"outerProduct\");\n  assert($v1.rank === 1 && $v2.rank === 1, () => `Error in outerProduct: inputs must be rank 1, but got ranks ${$v1.rank} and ${$v2.rank}.`);\n  const v12D = reshape($v1, [-1, 1]);\n  const v22D = reshape($v2, [1, -1]);\n  return matMul(v12D, v22D);\n}\nvar outerProduct = op({ outerProduct_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pad.js\nfunction pad_(x, paddings, constantValue = 0) {\n  const $x = convertToTensor(x, \"x\", \"pad\");\n  if ($x.rank === 0) {\n    throw new Error(\"pad(scalar) is not defined. Pass non-scalar to pad\");\n  }\n  const attrs = { paddings, constantValue };\n  const inputs = { x: $x };\n  return ENGINE.runKernel(PadV2, inputs, attrs);\n}\nvar pad = op({ pad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pad1d.js\nfunction pad1d_(x, paddings, constantValue = 0) {\n  assert(paddings.length === 2, () => \"Invalid number of paddings. Must be length of 2.\");\n  return pad(x, [paddings], constantValue);\n}\nvar pad1d = op({ pad1d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pad2d.js\nfunction pad2d_(x, paddings, constantValue = 0) {\n  assert(paddings.length === 2 && paddings[0].length === 2 && paddings[1].length === 2, () => \"Invalid number of paddings. Must be length of 2 each.\");\n  return pad(x, paddings, constantValue);\n}\nvar pad2d = op({ pad2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pad3d.js\nfunction pad3d_(x, paddings, constantValue = 0) {\n  assert(paddings.length === 3 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2, () => \"Invalid number of paddings. Must be length of 2 each.\");\n  return pad(x, paddings, constantValue);\n}\nvar pad3d = op({ pad3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pad4d.js\nfunction pad4d_(x, paddings, constantValue = 0) {\n  assert(paddings.length === 4 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2 && paddings[3].length === 2, () => \"Invalid number of paddings. Must be length of 2 each.\");\n  return pad(x, paddings, constantValue);\n}\nvar pad4d = op({ pad4d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/space_to_batch_nd.js\nfunction spaceToBatchND_(x, blockShape, paddings) {\n  const $x = convertToTensor(x, \"x\", \"spaceToBatchND\");\n  assert($x.rank >= 1 + blockShape.length, () => `input rank ${$x.rank} should be > than [blockShape] ${blockShape.length}`);\n  assert(paddings.length === blockShape.length, () => `paddings.shape[0] ${paddings.length} must be equal to [blockShape] ${blockShape.length}`);\n  assert($x.shape.reduce((a, b, i) => {\n    if (i > 0 && i <= blockShape.length) {\n      return a && (b + paddings[i - 1][0] + paddings[i - 1][1]) % blockShape[i - 1] === 0;\n    }\n    return a;\n  }, true), () => `input spatial dimensions ${$x.shape.slice(1)} with paddings ${paddings.toString()} must be divisible by blockShapes ${blockShape.toString()}`);\n  const inputs = { x: $x };\n  const attrs = { blockShape, paddings };\n  return ENGINE.runKernel(SpaceToBatchND, inputs, attrs);\n}\nvar spaceToBatchND = op({ spaceToBatchND_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/pool.js\nfunction pool_(input2, windowShape, poolingType, pad3, dilations, strides, dimRoundingMode) {\n  if (dilations == null) {\n    dilations = [1, 1];\n  }\n  if (strides == null) {\n    strides = 1;\n  }\n  if (pad3 === 0) {\n    pad3 = \"valid\";\n  }\n  const $x = convertToTensor(input2, \"x\", \"maxPool\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in pool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = computePool2DInfo(x4D.shape, windowShape, strides, dilations, pad3);\n  const dilation = [convInfo.dilationHeight, convInfo.dilationWidth];\n  let basePadding;\n  if (pad3 === \"same\") {\n    basePadding = withSpaceToBatchBasePaddings([convInfo.filterHeight, convInfo.filterWidth], dilation);\n  } else {\n    basePadding = [[0, 0], [0, 0]];\n  }\n  const isDilationOne = dilation[0] === 1 && dilation[1] === 1;\n  const [adjustedPadding, adjustedCrops] = requiredSpaceToBatchPaddings([convInfo.inHeight, convInfo.inWidth], dilation, basePadding);\n  const convertedPad = isDilationOne ? pad3 : \"valid\";\n  const convertedX = isDilationOne ? x4D : spaceToBatchND(x4D, dilation, adjustedPadding);\n  const forwardOp = poolingType === \"avg\" ? () => avgPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode) : () => maxPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode);\n  const y = forwardOp();\n  const res = isDilationOne ? y : batchToSpaceND(y, dilation, adjustedCrops);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nfunction requiredSpaceToBatchPaddings(inputShape, blockShape, basePadding) {\n  const padStart = basePadding.map((b) => b[0]);\n  const origPadEnd = basePadding.map((b) => b[1]);\n  const fullInputShape = inputShape.concat(padStart, origPadEnd);\n  const padEndExtra = blockShape.map((b, i) => (b - fullInputShape[i] % b) % b);\n  const padEnd = origPadEnd.map((s, i) => s + padEndExtra[i]);\n  const paddings = blockShape.map((_, i) => [padStart[i], padEnd[i]]);\n  const crops = blockShape.map((_, i) => [0, padEndExtra[i]]);\n  return [paddings, crops];\n}\nfunction withSpaceToBatchBasePaddings(filterShape, dilation) {\n  const dilatedFilterShape = filterShape.map((s, i) => {\n    return s + (s - 1) * (dilation[i] - 1);\n  });\n  const padExtraShape = dilatedFilterShape.map((s) => s - 1);\n  const padExtraStart = padExtraShape.map((s) => Math.floor(s / 2));\n  const padExtraEnd = padExtraShape.map((s, i) => s - padExtraStart[i]);\n  return padExtraShape.map((_, i) => {\n    return [padExtraStart[i], padExtraEnd[i]];\n  });\n}\nvar pool = op({ pool_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/prelu.js\nfunction prelu_(x, alpha) {\n  const $x = convertToTensor(x, \"x\", \"prelu\");\n  const $alpha = convertToTensor(alpha, \"alpha\", \"prelu\");\n  const inputs = { x: $x, alpha: $alpha };\n  return ENGINE.runKernel(Prelu, inputs);\n}\nvar prelu = op({ prelu_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/prod.js\nfunction prod_(x, axis = null, keepDims = false) {\n  let $x = convertToTensor(x, \"x\", \"prod\");\n  if ($x.dtype === \"bool\") {\n    $x = cast($x, \"int32\");\n  }\n  const inputs = { x: $x };\n  const attrs = { axis, keepDims };\n  return ENGINE.runKernel(Prod, inputs, attrs);\n}\nvar prod = op({ prod_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_gather.js\nfunction raggedGather_(paramsNestedSplits, paramsDenseValues, indices, outputRaggedRank) {\n  const $paramsNestedSplits = paramsNestedSplits.map((t, i) => convertToTensor(t, `tensors${i}`, \"raggedGather\", \"int32\"));\n  const $paramsDenseValues = convertToTensor(paramsDenseValues, \"paramsDenseValues\", \"raggedGather\");\n  const $indices = convertToTensor(indices, \"indices\", \"raggedGather\", \"int32\");\n  const inputs = {\n    paramsNestedSplits: $paramsNestedSplits,\n    paramsDenseValues: $paramsDenseValues,\n    indices: $indices\n  };\n  const attrs = { outputRaggedRank };\n  const result = ENGINE.runKernel(RaggedGather, inputs, attrs);\n  return {\n    outputNestedSplits: result.slice(0, result.length - 1),\n    outputDenseValues: result[result.length - 1]\n  };\n}\nvar raggedGather = op({ raggedGather_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_range.js\nfunction raggedRange_(starts, limits, deltas) {\n  const $starts = convertToTensor(starts, \"starts\", \"raggedRange\");\n  const $limits = convertToTensor(limits, \"limits\", \"raggedRange\", $starts.dtype);\n  const $deltas = convertToTensor(deltas, \"deltas\", \"raggedRange\", $starts.dtype);\n  const inputs = {\n    starts: $starts,\n    limits: $limits,\n    deltas: $deltas\n  };\n  const result = ENGINE.runKernel(RaggedRange, inputs);\n  return {\n    rtNestedSplits: result[0],\n    rtDenseValues: result[1]\n  };\n}\nvar raggedRange = op({ raggedRange_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_tensor_to_tensor.js\nfunction raggedTensorToTensor_(shape, values, defaultValue, rowPartitionTensors, rowPartitionTypes) {\n  const $shape = convertToTensor(shape, \"shape\", \"raggedTensorToTensor\", \"int32\");\n  const $values = convertToTensor(values, \"values\", \"raggedTensorToTensor\");\n  const $defaultValue = convertToTensor(defaultValue, \"defaultValue\", \"raggedTensorToTensor\", $values.dtype);\n  const $rowPartitionTensors = rowPartitionTensors.map((t, i) => convertToTensor(t, `tensors${i}`, \"raggedTensorToTensor\", \"int32\"));\n  const inputs = {\n    shape: $shape,\n    values: $values,\n    defaultValue: $defaultValue,\n    rowPartitionTensors: $rowPartitionTensors\n  };\n  const attrs = { rowPartitionTypes };\n  return ENGINE.runKernel(RaggedTensorToTensor, inputs, attrs);\n}\nvar raggedTensorToTensor = op({ raggedTensorToTensor_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/rand.js\nfunction rand_(shape, randFunction, dtype) {\n  const size = sizeFromShape(shape);\n  let values = null;\n  if (dtype == null || dtype === \"float32\") {\n    values = new Float32Array(size);\n  } else if (dtype === \"int32\") {\n    values = new Int32Array(size);\n  } else if (dtype === \"bool\") {\n    values = new Uint8Array(size);\n  } else {\n    throw new Error(`Unknown data type ${dtype}`);\n  }\n  for (let i = 0; i < size; i++) {\n    values[i] = randFunction();\n  }\n  return ENGINE.makeTensor(values, shape, dtype);\n}\nvar rand = op({ rand_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/rand_util.js\nvar seedrandom = __toESM(require_seedrandom2());\nvar MPRandGauss = class {\n  constructor(mean4, stdDeviation, dtype, truncated, seed) {\n    this.mean = mean4;\n    this.stdDev = stdDeviation;\n    this.dtype = dtype;\n    this.nextVal = NaN;\n    this.truncated = truncated;\n    if (this.truncated) {\n      this.upper = this.mean + this.stdDev * 2;\n      this.lower = this.mean - this.stdDev * 2;\n    }\n    const seedValue = seed ? seed : Math.random();\n    this.random = seedrandom.alea(seedValue.toString());\n  }\n  nextValue() {\n    if (!isNaN(this.nextVal)) {\n      const value = this.nextVal;\n      this.nextVal = NaN;\n      return value;\n    }\n    let resultX, resultY;\n    let isValid = false;\n    while (!isValid) {\n      let v1, v2, s;\n      do {\n        v1 = 2 * this.random() - 1;\n        v2 = 2 * this.random() - 1;\n        s = v1 * v1 + v2 * v2;\n      } while (s >= 1 || s === 0);\n      const mul2 = Math.sqrt(-2 * Math.log(s) / s);\n      resultX = this.mean + this.stdDev * v1 * mul2;\n      resultY = this.mean + this.stdDev * v2 * mul2;\n      if (!this.truncated || this.isValidTruncated(resultX)) {\n        isValid = true;\n      }\n    }\n    if (!this.truncated || this.isValidTruncated(resultY)) {\n      this.nextVal = this.convertValue(resultY);\n    }\n    return this.convertValue(resultX);\n  }\n  convertValue(value) {\n    if (this.dtype == null || this.dtype === \"float32\") {\n      return value;\n    }\n    return Math.round(value);\n  }\n  isValidTruncated(value) {\n    return value <= this.upper && value >= this.lower;\n  }\n};\nvar RandGamma = class {\n  constructor(alpha, beta, dtype, seed) {\n    this.alpha = alpha;\n    this.beta = 1 / beta;\n    this.dtype = dtype;\n    const seedValue = seed ? seed : Math.random();\n    this.randu = seedrandom.alea(seedValue.toString());\n    this.randn = new MPRandGauss(0, 1, dtype, false, this.randu());\n    if (alpha < 1) {\n      this.d = alpha + 2 / 3;\n    } else {\n      this.d = alpha - 1 / 3;\n    }\n    this.c = 1 / Math.sqrt(9 * this.d);\n  }\n  nextValue() {\n    let x2, v0, v1, x, u, v;\n    while (true) {\n      do {\n        x = this.randn.nextValue();\n        v = 1 + this.c * x;\n      } while (v <= 0);\n      v *= v * v;\n      x2 = x * x;\n      v0 = 1 - 0.331 * x2 * x2;\n      v1 = 0.5 * x2 + this.d * (1 - v + Math.log(v));\n      u = this.randu();\n      if (u < v0 || Math.log(u) < v1) {\n        break;\n      }\n    }\n    v = 1 / this.beta * this.d * v;\n    if (this.alpha < 1) {\n      v *= Math.pow(this.randu(), 1 / this.alpha);\n    }\n    return this.convertValue(v);\n  }\n  convertValue(value) {\n    if (this.dtype === \"float32\") {\n      return value;\n    }\n    return Math.round(value);\n  }\n};\nvar UniformRandom = class {\n  constructor(min6 = 0, max6 = 1, dtype, seed) {\n    this.canReturnFloat = () => this.dtype == null || this.dtype === \"float32\";\n    this.min = min6;\n    this.range = max6 - min6;\n    this.dtype = dtype;\n    if (seed == null) {\n      seed = Math.random();\n    }\n    if (typeof seed === \"number\") {\n      seed = seed.toString();\n    }\n    if (!this.canReturnFloat() && this.range <= 1) {\n      throw new Error(`The difference between ${min6} - ${max6} <= 1 and dtype is not float`);\n    }\n    this.random = seedrandom.alea(seed);\n  }\n  convertValue(value) {\n    if (this.canReturnFloat()) {\n      return value;\n    }\n    return Math.round(value);\n  }\n  nextValue() {\n    return this.convertValue(this.min + this.range * this.random());\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/random_gamma.js\nfunction randomGamma_(shape, alpha, beta = 1, dtype = \"float32\", seed) {\n  if (beta == null) {\n    beta = 1;\n  }\n  if (dtype == null) {\n    dtype = \"float32\";\n  }\n  if (dtype !== \"float32\" && dtype !== \"int32\") {\n    throw new Error(`Unsupported data type ${dtype}`);\n  }\n  const rgamma = new RandGamma(alpha, beta, dtype, seed);\n  const res = buffer(shape, dtype);\n  for (let i = 0; i < res.values.length; i++) {\n    res.values[i] = rgamma.nextValue();\n  }\n  return res.toTensor();\n}\nvar randomGamma = op({ randomGamma_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/random_normal.js\nfunction randomNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {\n  if (dtype != null && dtype === \"bool\") {\n    throw new Error(`Unsupported data type ${dtype}`);\n  }\n  const randGauss = new MPRandGauss(mean4, stdDev, dtype, false, seed);\n  const res = buffer(shape, dtype);\n  for (let i = 0; i < res.values.length; i++) {\n    res.values[i] = randGauss.nextValue();\n  }\n  return res.toTensor();\n}\nvar randomNormal = op({ randomNormal_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/random_standard_normal.js\nfunction randomStandardNormal_(shape, dtype, seed) {\n  if (dtype != null && dtype === \"bool\") {\n    throw new Error(`Unsupported data type ${dtype}`);\n  }\n  return randomNormal(shape, 0, 1, dtype, seed);\n}\nvar randomStandardNormal = op({ randomStandardNormal_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/random_uniform.js\nfunction randomUniform_(shape, minval = 0, maxval = 1, dtype = \"float32\", seed) {\n  const res = buffer(shape, dtype);\n  const random = new UniformRandom(minval, maxval, null, seed);\n  for (let i = 0; i < res.values.length; i++) {\n    res.values[i] = random.nextValue();\n  }\n  return res.toTensor();\n}\nvar randomUniform = op({ randomUniform_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/range.js\nfunction range(start, stop, step5 = 1, dtype = \"float32\") {\n  if (step5 === 0) {\n    throw new Error(\"Cannot have a step of zero\");\n  }\n  const attrs = { start, stop, step: step5, dtype };\n  return ENGINE.runKernel(Range, {}, attrs);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reciprocal.js\nfunction reciprocal_(x) {\n  const $x = convertToTensor(x, \"x\", \"reciprocal\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Reciprocal, inputs);\n}\nvar reciprocal = op({ reciprocal_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/relu.js\nfunction relu_(x) {\n  const $x = convertToTensor(x, \"x\", \"relu\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Relu, inputs);\n}\nvar relu = op({ relu_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/relu6.js\nfunction relu6_(x) {\n  const $x = convertToTensor(x, \"x\", \"relu6\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Relu6, inputs);\n}\nvar relu6 = op({ relu6_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reverse.js\nfunction reverse_(x, axis) {\n  const $x = convertToTensor(x, \"x\", \"reverse\");\n  const inputs = { x: $x };\n  const attrs = { dims: axis };\n  return ENGINE.runKernel(Reverse, inputs, attrs);\n}\nvar reverse = op({ reverse_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_1d.js\nfunction reverse1d_(x) {\n  const $x = convertToTensor(x, \"x\", \"reverse\");\n  assert($x.rank === 1, () => `Error in reverse1D: x must be rank 1 but got rank ${$x.rank}.`);\n  return reverse($x, 0);\n}\nvar reverse1d = op({ reverse1d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_2d.js\nfunction reverse2d_(x, axis) {\n  const $x = convertToTensor(x, \"x\", \"reverse\");\n  assert($x.rank === 2, () => `Error in reverse2D: x must be rank 2 but got rank ${$x.rank}.`);\n  return reverse($x, axis);\n}\nvar reverse2d = op({ reverse2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_3d.js\nfunction reverse3d_(x, axis) {\n  const $x = convertToTensor(x, \"x\", \"reverse\");\n  assert($x.rank === 3, () => `Error in reverse3D: x must be rank 3 but got rank ${$x.rank}.`);\n  return reverse($x, axis);\n}\nvar reverse3d = op({ reverse3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_4d.js\nfunction reverse4d_(x, axis) {\n  const $x = convertToTensor(x, \"x\", \"reverse\");\n  assert($x.rank === 4, () => `Error in reverse4D: x must be rank 4 but got rank ${$x.rank}.`);\n  return reverse($x, axis);\n}\nvar reverse4d = op({ reverse4d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/round.js\nfunction round_(x) {\n  const $x = convertToTensor(x, \"x\", \"round\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Round, inputs);\n}\nvar round2 = op({ round_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/rsqrt.js\nfunction rsqrt_(x) {\n  const $x = convertToTensor(x, \"x\", \"rsqrt\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Rsqrt, inputs);\n}\nvar rsqrt = op({ rsqrt_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/selu.js\nfunction selu_(x) {\n  const $x = convertToTensor(x, \"x\", \"selu\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Selu, inputs);\n}\nvar selu = op({ selu_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/separable_conv2d.js\nfunction separableConv2d_(x, depthwiseFilter, pointwiseFilter, strides, pad3, dilation = [1, 1], dataFormat = \"NHWC\") {\n  const $x = convertToTensor(x, \"x\", \"separableConv2d\");\n  const $depthwiseFilter = convertToTensor(depthwiseFilter, \"depthwiseFilter\", \"separableConv2d\");\n  const $pointwiseFilter = convertToTensor(pointwiseFilter, \"pointwiseFilter\", \"separableConv2d\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  if (dataFormat === \"NCHW\") {\n    throw new Error(\"separableConv2d currently does not support dataFormat NCHW; only NHWC is supported\");\n  }\n  assert(x4D.rank === 4, () => `Error in separableConv2d: input must be rank 4, but got rank ${x4D.rank}.`);\n  assert($depthwiseFilter.rank === 4, () => `Error in separableConv2d: depthwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);\n  assert($pointwiseFilter.rank === 4, () => `Error in separableConv2d: pointwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);\n  assert($pointwiseFilter.shape[0] === 1, () => `Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${$pointwiseFilter.shape[0]}.`);\n  assert($pointwiseFilter.shape[1] === 1, () => `Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${$pointwiseFilter.shape[1]}.`);\n  const inChannels = $depthwiseFilter.shape[2];\n  const channelMultiplier = $depthwiseFilter.shape[3];\n  assert($pointwiseFilter.shape[2] === inChannels * channelMultiplier, () => `Error in separableConv2d: the third dimension of pointwise filter must be ${inChannels * channelMultiplier}, but got ${$pointwiseFilter.shape[2]}.`);\n  const depthwise = depthwiseConv2d(x4D, $depthwiseFilter, strides, pad3, dataFormat, dilation);\n  const pointwiseStride = 1;\n  const res = conv2d(depthwise, $pointwiseFilter, pointwiseStride, \"valid\", dataFormat);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar separableConv2d = op({ separableConv2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/setdiff1d_async.js\nasync function setdiff1dAsync_(x, y) {\n  const $x = convertToTensor(x, \"x\", \"setdiff1d\");\n  const $y = convertToTensor(y, \"y\", \"setdiff1d\");\n  assert($x.dtype === $y.dtype, () => `x and y should have the same dtype, but got x (${$x.dtype}) and y (${$y.dtype}).`);\n  assert($x.rank === 1, () => `x should be 1D tensor, but got x (${$x.shape}).`);\n  assert($y.rank === 1, () => `y should be 1D tensor, but got y (${$y.shape}).`);\n  const xVals = await $x.data();\n  const yVals = await $y.data();\n  const ySet = new Set(yVals);\n  let outputSize = 0;\n  for (let i = 0; i < xVals.length; i++) {\n    if (!ySet.has(xVals[i])) {\n      outputSize++;\n    }\n  }\n  const buffer2 = new TensorBuffer([outputSize], $x.dtype);\n  const indices = new TensorBuffer([outputSize], \"int32\");\n  for (let i = 0, p2 = 0; i < xVals.length; i++) {\n    if (!ySet.has(xVals[i])) {\n      buffer2.values[p2] = xVals[i];\n      indices.values[p2] = i;\n      p2++;\n    }\n  }\n  return [buffer2.toTensor(), indices.toTensor()];\n}\nvar setdiff1dAsync = setdiff1dAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sign.js\nfunction sign_(x) {\n  const $x = convertToTensor(x, \"x\", \"sign\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Sign, inputs);\n}\nvar sign = op({ sign_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sin.js\nfunction sin_(x) {\n  const $x = convertToTensor(x, \"x\", \"sin\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Sin, inputs);\n}\nvar sin = op({ sin_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sinh.js\nfunction sinh_(x) {\n  const $x = convertToTensor(x, \"x\", \"sinh\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Sinh, inputs);\n}\nvar sinh = op({ sinh_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice1d.js\nfunction slice1d_(x, begin, size) {\n  const $x = convertToTensor(x, \"x\", \"slice1d\");\n  assert($x.rank === 1, () => `slice1d expects a rank-1 tensor, but got a rank-${$x.rank} tensor`);\n  return slice($x, [begin], [size]);\n}\nvar slice1d = op({ slice1d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice2d.js\nfunction slice2d_(x, begin, size) {\n  const $x = convertToTensor(x, \"x\", \"slice2d\");\n  assert($x.rank === 2, () => `slice2d expects a rank-2 tensor, but got a rank-${$x.rank} tensor`);\n  return slice($x, begin, size);\n}\nvar slice2d = op({ slice2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice3d.js\nfunction slice3d_(x, begin, size) {\n  const $x = convertToTensor(x, \"x\", \"slice3d\");\n  assert($x.rank === 3, () => `slice3d expects a rank-3 tensor, but got a rank-${$x.rank} tensor`);\n  return slice($x, begin, size);\n}\nvar slice3d = op({ slice3d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/slice4d.js\nfunction slice4d_(x, begin, size) {\n  const $x = convertToTensor(x, \"x\", \"slice4d\");\n  assert($x.rank === 4, () => `slice4d expects a rank-4 tensor, but got a rank-${$x.rank} tensor`);\n  return slice($x, begin, size);\n}\nvar slice4d = op({ slice4d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/softmax.js\nfunction softmax_(logits, dim = -1) {\n  const $logits = convertToTensor(logits, \"logits\", \"softmax\", \"float32\");\n  if (dim === -1) {\n    dim = $logits.rank - 1;\n  }\n  if (dim !== $logits.rank - 1) {\n    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and dim was ${dim}`);\n  }\n  const inputs = { logits: $logits };\n  const attrs = { dim };\n  return ENGINE.runKernel(Softmax, inputs, attrs);\n}\nvar softmax = op({ softmax_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/fft.js\nfunction fft_(input2) {\n  assert(input2.dtype === \"complex64\", () => `The dtype for tf.spectral.fft() must be complex64 but got ${input2.dtype}.`);\n  const inputs = { input: input2 };\n  return ENGINE.runKernel(FFT, inputs);\n}\nvar fft = op({ fft_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/ifft.js\nfunction ifft_(input2) {\n  assert(input2.dtype === \"complex64\", () => `The dtype for tf.spectral.ifft() must be complex64 but got ${input2.dtype}.`);\n  const inputs = { input: input2 };\n  return ENGINE.runKernel(IFFT, inputs);\n}\nvar ifft = op({ ifft_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/irfft.js\nfunction irfft_(input2) {\n  const innerDimensionSize = input2.shape[input2.shape.length - 1];\n  const batch = input2.size / innerDimensionSize;\n  let ret;\n  if (innerDimensionSize <= 2) {\n    const complexInput = reshape(input2, [batch, innerDimensionSize]);\n    ret = ifft(complexInput);\n  } else {\n    const outputShape = [batch, 2 * (innerDimensionSize - 1)];\n    const realInput = reshape(real(input2), [batch, innerDimensionSize]);\n    const imagInput = reshape(imag(input2), [batch, innerDimensionSize]);\n    const realConjugate = reverse(slice(realInput, [0, 1], [batch, innerDimensionSize - 2]), 1);\n    const imagConjugate = mul(reverse(slice(imagInput, [0, 1], [batch, innerDimensionSize - 2]), 1), scalar(-1));\n    const r = concat([realInput, realConjugate], 1);\n    const i = concat([imagInput, imagConjugate], 1);\n    const complexInput = reshape(complex(r, i), [outputShape[0], outputShape[1]]);\n    ret = ifft(complexInput);\n  }\n  ret = real(ret);\n  if (input2.rank === 3 && input2.shape[0] !== 0) {\n    const temp = ret;\n    const batch2 = input2.shape[0];\n    ret = reshape(ret, [batch2, ret.shape[0] / batch2, ret.shape[1]]);\n    temp.dispose();\n  }\n  return ret;\n}\nvar irfft = op({ irfft_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/split.js\nfunction split_(x, numOrSizeSplits, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"split\");\n  const inputs = { x: $x };\n  const attr = { numOrSizeSplits, axis };\n  return ENGINE.runKernel(SplitV, inputs, attr);\n}\nvar split = op({ split_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/rfft.js\nfunction rfft_(input2, fftLength) {\n  assert(input2.dtype === \"float32\", () => `The dtype for rfft() must be real value but got ${input2.dtype}`);\n  let innerDimensionSize = input2.shape[input2.shape.length - 1];\n  const batch = input2.size / innerDimensionSize;\n  let adjustedInput;\n  if (fftLength != null && fftLength < innerDimensionSize) {\n    const begin = input2.shape.map((v) => 0);\n    const size = input2.shape.map((v) => v);\n    size[input2.shape.length - 1] = fftLength;\n    adjustedInput = slice(input2, begin, size);\n    innerDimensionSize = fftLength;\n  } else if (fftLength != null && fftLength > innerDimensionSize) {\n    const zerosShape = input2.shape.map((v) => v);\n    zerosShape[input2.shape.length - 1] = fftLength - innerDimensionSize;\n    adjustedInput = concat([input2, zeros(zerosShape)], input2.shape.length - 1);\n    innerDimensionSize = fftLength;\n  } else {\n    adjustedInput = input2;\n  }\n  const zerosInput = zerosLike(adjustedInput);\n  const complexInput = reshape(complex(adjustedInput, zerosInput), [batch, innerDimensionSize]);\n  const ret = fft(complexInput);\n  const half = Math.floor(innerDimensionSize / 2) + 1;\n  const realValues = real(ret);\n  const imagValues = imag(ret);\n  const realComplexConjugate = split(realValues, [half, innerDimensionSize - half], realValues.shape.length - 1);\n  const imagComplexConjugate = split(imagValues, [half, innerDimensionSize - half], imagValues.shape.length - 1);\n  const outputShape = adjustedInput.shape.slice();\n  outputShape[adjustedInput.shape.length - 1] = half;\n  return reshape(complex(realComplexConjugate[0], imagComplexConjugate[0]), outputShape);\n}\nvar rfft = op({ rfft_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/squared_difference.js\nfunction squaredDifference_(a, b) {\n  let $a = convertToTensor(a, \"a\", \"squaredDifference\");\n  let $b = convertToTensor(b, \"b\", \"squaredDifference\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  assertAndGetBroadcastShape($a.shape, $b.shape);\n  const inputs = { a: $a, b: $b };\n  const attrs = {};\n  return ENGINE.runKernel(SquaredDifference, inputs, attrs);\n}\nvar squaredDifference = op({ squaredDifference_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/squeeze.js\nfunction squeeze_(x, axis) {\n  const $x = convertToTensor(x, \"x\", \"squeeze\", \"string_or_numeric\");\n  return reshape($x, squeezeShape($x.shape, axis).newShape);\n}\nvar squeeze = op({ squeeze_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/stack.js\nfunction stack_(tensors, axis = 0) {\n  const $tensors = convertToTensorArray(tensors, \"tensors\", \"stack\", \"string_or_numeric\");\n  assert($tensors.length >= 1, () => \"Pass at least one tensor to tf.stack\");\n  if ($tensors.length > 0) {\n    assert(axis <= $tensors[0].rank, () => \"Axis must be <= rank of the tensor\");\n  }\n  const inputs = $tensors;\n  const attrs = { axis };\n  return ENGINE.runKernel(Pack, inputs, attrs);\n}\nvar stack = op({ stack_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/step.js\nfunction step_(x, alpha = 0) {\n  const $x = convertToTensor(x, \"x\", \"step\");\n  const inputs = { x: $x };\n  const attrs = { alpha };\n  return ENGINE.runKernel(Step, inputs, attrs);\n}\nvar step = op({ step_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/strided_slice.js\nfunction stridedSlice_(x, begin, end, strides, beginMask = 0, endMask = 0, ellipsisMask = 0, newAxisMask = 0, shrinkAxisMask = 0) {\n  const $x = convertToTensor(x, \"x\", \"stridedSlice\", \"string_or_numeric\");\n  const inputs = { x: $x };\n  const attrs = {\n    begin,\n    end,\n    strides,\n    beginMask,\n    endMask,\n    ellipsisMask,\n    newAxisMask,\n    shrinkAxisMask\n  };\n  return ENGINE.runKernel(StridedSlice, inputs, attrs);\n}\nvar stridedSlice = op({ stridedSlice_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tan.js\nfunction tan_(x) {\n  const $x = convertToTensor(x, \"x\", \"tan\", \"float32\");\n  const inputs = { x: $x };\n  return ENGINE.runKernel(Tan, inputs);\n}\nvar tan = op({ tan_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor1d.js\nfunction tensor1d(values, dtype) {\n  assertNonNull(values);\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 1) {\n    throw new Error(\"tensor1d() requires values to be a flat/TypedArray\");\n  }\n  const shape = null;\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor2d.js\nfunction tensor2d(values, shape, dtype) {\n  assertNonNull(values);\n  if (shape != null && shape.length !== 2) {\n    throw new Error(\"tensor2d() requires shape to have two numbers\");\n  }\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 2 && inferredShape.length !== 1) {\n    throw new Error(\"tensor2d() requires values to be number[][] or flat/TypedArray\");\n  }\n  if (inferredShape.length === 1 && shape == null) {\n    throw new Error(\"tensor2d() requires shape to be provided when `values` are a flat/TypedArray\");\n  }\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor4d.js\nfunction tensor4d(values, shape, dtype) {\n  assertNonNull(values);\n  if (shape != null && shape.length !== 4) {\n    throw new Error(\"tensor4d() requires shape to have four numbers\");\n  }\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 4 && inferredShape.length !== 1) {\n    throw new Error(\"tensor4d() requires values to be number[][][][] or flat/TypedArray\");\n  }\n  if (inferredShape.length === 1 && shape == null) {\n    throw new Error(\"tensor4d() requires shape to be provided when `values` are a flat array\");\n  }\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor5d.js\nfunction tensor5d(values, shape, dtype) {\n  assertNonNull(values);\n  if (shape != null && shape.length !== 5) {\n    throw new Error(\"tensor5d() requires shape to have five numbers\");\n  }\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 5 && inferredShape.length !== 1) {\n    throw new Error(\"tensor5d() requires values to be number[][][][][] or flat/TypedArray\");\n  }\n  if (inferredShape.length === 1 && shape == null) {\n    throw new Error(\"tensor5d() requires shape to be provided when `values` are a flat array\");\n  }\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/tensor6d.js\nfunction tensor6d(values, shape, dtype) {\n  assertNonNull(values);\n  if (shape != null && shape.length !== 6) {\n    throw new Error(\"tensor6d() requires shape to have six numbers\");\n  }\n  const inferredShape = inferShape(values, dtype);\n  if (inferredShape.length !== 6 && inferredShape.length !== 1) {\n    throw new Error(\"tensor6d() requires values to be number[][][][][][] or flat/TypedArray\");\n  }\n  if (inferredShape.length === 1 && shape == null) {\n    throw new Error(\"tensor6d() requires shape to be provided when `values` are a flat array\");\n  }\n  shape = shape || inferredShape;\n  return makeTensor(values, shape, inferredShape, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/topk.js\nfunction topk_(x, k = 1, sorted = true) {\n  const $x = convertToTensor(x, \"x\", \"topk\");\n  if ($x.rank === 0) {\n    throw new Error(\"topk() expects the input to be of rank 1 or higher\");\n  }\n  const lastDim = $x.shape[$x.shape.length - 1];\n  if (k < 0) {\n    throw new Error(`'k' passed to topk() must be >= 0 but got ${k}`);\n  }\n  if (k > lastDim) {\n    throw new Error(`'k' passed to topk() must be <= the last dimension (${lastDim}) but got ${k}`);\n  }\n  const inputs = { x: $x };\n  const attrs = { k, sorted };\n  const [values, indices] = ENGINE.runKernel(TopK, inputs, attrs);\n  return { values, indices };\n}\nvar topk = op({ topk_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/truncated_normal.js\nfunction truncatedNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {\n  if (dtype != null && dtype === \"bool\") {\n    throw new Error(`Unsupported data type $ { dtype }`);\n  }\n  const randGauss = new MPRandGauss(mean4, stdDev, dtype, true, seed);\n  const res = buffer(shape, dtype);\n  for (let i = 0; i < res.values.length; i++) {\n    res.values[i] = randGauss.nextValue();\n  }\n  return res.toTensor();\n}\nvar truncatedNormal = op({ truncatedNormal_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/unique.js\nfunction unique_(x, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"unique\", \"string_or_numeric\");\n  assert($x.rank > 0, () => \"The input tensor must be at least 1D\");\n  const inputs = { x: $x };\n  const attrs = { axis };\n  const [values, indices] = ENGINE.runKernel(Unique, inputs, attrs);\n  return { values, indices };\n}\nvar unique = op({ unique_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/unsorted_segment_sum.js\nfunction unsortedSegmentSum_(x, segmentIds, numSegments) {\n  const $x = convertToTensor(x, \"x\", \"unsortedSegmentSum\");\n  const $segmentIds = convertToTensor(segmentIds, \"segmentIds\", \"unsortedSegmentSum\", \"int32\");\n  assert(isInt(numSegments), () => \"numSegments must be of dtype int\");\n  const inputs = { x: $x, segmentIds: $segmentIds };\n  const attrs = { numSegments };\n  return ENGINE.runKernel(UnsortedSegmentSum, inputs, attrs);\n}\nvar unsortedSegmentSum = op({ unsortedSegmentSum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/unstack.js\nfunction unstack_(x, axis = 0) {\n  const $x = convertToTensor(x, \"x\", \"unstack\", \"string_or_numeric\");\n  assert(axis >= -$x.shape.length && axis < $x.shape.length, () => `Axis = ${axis} is not in [-${$x.shape.length}, ${$x.shape.length})`);\n  const inputs = { value: $x };\n  const attrs = { axis };\n  return ENGINE.runKernel(Unpack, inputs, attrs);\n}\nvar unstack = op({ unstack_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/upper_bound.js\nfunction upperBound(sortedSequence, values) {\n  return searchSorted(sortedSequence, values, \"right\");\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/variable.js\nfunction variable(initialValue, trainable = true, name, dtype) {\n  return ENGINE.makeVariable(initialValue, trainable, name, dtype);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/where_impl.js\nfunction whereImpl(condShape, condVals) {\n  const indices = [];\n  for (let i = 0; i < condVals.length; i++) {\n    if (condVals[i]) {\n      indices.push(i);\n    }\n  }\n  const inBuffer = buffer(condShape, \"int32\");\n  const out = buffer([indices.length, condShape.length], \"int32\");\n  for (let i = 0; i < indices.length; i++) {\n    const loc = inBuffer.indexToLoc(indices[i]);\n    const offset = i * condShape.length;\n    out.values.set(loc, offset);\n  }\n  return out.toTensor();\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/where_async.js\nasync function whereAsync_(condition) {\n  const $condition = convertToTensor(condition, \"condition\", \"whereAsync\", \"bool\");\n  const vals = await $condition.data();\n  const res = whereImpl($condition.shape, vals);\n  if (condition !== $condition) {\n    $condition.dispose();\n  }\n  return res;\n}\nvar whereAsync = whereAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/boolean_mask.js\nasync function booleanMaskAsync_(tensor2, mask, axis) {\n  const $tensor = convertToTensor(tensor2, \"tensor\", \"boolMask\");\n  const $mask = convertToTensor(mask, \"mask\", \"boolMask\", \"bool\");\n  const axisFrom = axis == null ? 0 : axis;\n  const maskDim = $mask.rank;\n  const tensorShape = $tensor.shape;\n  assert(maskDim > 0, () => \"mask cannot be scalar\");\n  assertShapesMatch(tensorShape.slice(axisFrom, axisFrom + maskDim), $mask.shape, `mask's shape must match the first K dimensions of tensor's shape,`);\n  let leadingSize = 1;\n  for (let i = axisFrom; i < axisFrom + maskDim; i++) {\n    leadingSize *= tensorShape[i];\n  }\n  const targetTensorShape = tensorShape.slice(0, axisFrom).concat([leadingSize], tensorShape.slice(axisFrom + maskDim));\n  const reshapedTensor = reshape($tensor, targetTensorShape);\n  const reshapedMask = reshape($mask, [-1]);\n  const positivePositions = await whereAsync(reshapedMask);\n  const indices = squeeze(positivePositions, [1]);\n  const res = gather(reshapedTensor, indices, axisFrom);\n  if (tensor2 !== $tensor) {\n    $tensor.dispose();\n  }\n  if (mask !== $mask) {\n    $mask.dispose();\n  }\n  indices.dispose();\n  reshapedTensor.dispose();\n  reshapedMask.dispose();\n  positivePositions.dispose();\n  return res;\n}\nvar booleanMaskAsync = booleanMaskAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/moving_average.js\nfunction movingAverage_(v, x, decay, step5, zeroDebias = true) {\n  const $v = convertToTensor(v, \"v\", \"movingAverage\");\n  const $x = convertToTensor(x, \"x\", \"movingAverage\");\n  const $decay = convertToTensor(decay, \"decay\", \"movingAverage\");\n  assertTypesMatch($v, $x);\n  assert(arraysEqual($v.shape, $x.shape), () => \"Shape mismatch in v and x\");\n  const one = scalar(1);\n  const oneMinusDecay = sub(one, $decay);\n  let update = mul(sub($x, $v), oneMinusDecay);\n  if (zeroDebias) {\n    assert(step5 != null, () => \"When using zeroDebias: true, step is required.\");\n    const $step = convertToTensor(step5, \"step\", \"movingAverage\");\n    update = div(update, sub(one, pow($decay, $step)));\n  }\n  return add2($v, update);\n}\nvar movingAverage = op({ movingAverage_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/scatter_nd.js\nfunction scatterND_(indices, updates, shape) {\n  const $indices = convertToTensor(indices, \"indices\", \"scatterND\", \"int32\");\n  const $updates = convertToTensor(updates, \"updates\", \"scatterND\");\n  validateInput($updates, $indices, shape);\n  const inputs = { indices: $indices, updates: $updates };\n  const attrs = { shape };\n  return ENGINE.runKernel(ScatterNd, inputs, attrs);\n}\nvar scatterND = op({ scatterND_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse_to_dense_util.js\nfunction validateInput2(sparseIndices, sparseValues, outputShape, defaultValues) {\n  if (sparseIndices.dtype !== \"int32\") {\n    throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${sparseIndices.dtype}.`);\n  }\n  if (sparseIndices.rank > 2) {\n    throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${sparseIndices.shape}.`);\n  }\n  const numElems = sparseIndices.rank > 0 ? sparseIndices.shape[0] : 1;\n  const numDims = sparseIndices.rank > 1 ? sparseIndices.shape[1] : 1;\n  if (outputShape.length !== numDims) {\n    throw new Error(`outputShape has incorrect number of elements:, ${outputShape.length}, should be: ${numDims}.`);\n  }\n  const numValues = sparseValues.size;\n  if (!(sparseValues.rank === 0 || sparseValues.rank === 1 && numValues === numElems)) {\n    throw new Error(`sparseValues has incorrect shape ${sparseValues.shape}, should be [] or [${numElems}]`);\n  }\n  if (sparseValues.dtype !== defaultValues.dtype) {\n    throw new Error(\"sparseValues.dtype must match defaultValues.dtype\");\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse_to_dense.js\nfunction sparseToDense_(sparseIndices, sparseValues, outputShape, defaultValue = 0) {\n  const $sparseIndices = convertToTensor(sparseIndices, \"sparseIndices\", \"sparseToDense\", \"int32\");\n  const $sparseValues = convertToTensor(sparseValues, \"sparseValues\", \"sparseToDense\", \"string_or_numeric\");\n  const $defaultValue = convertToTensor(defaultValue, \"defaultValue\", \"sparseToDense\", $sparseValues.dtype);\n  validateInput2($sparseIndices, $sparseValues, outputShape, $defaultValue);\n  const inputs = {\n    sparseIndices: $sparseIndices,\n    sparseValues: $sparseValues,\n    defaultValue: $defaultValue\n  };\n  const attrs = { outputShape };\n  return ENGINE.runKernel(SparseToDense, inputs, attrs);\n}\nvar sparseToDense = op({ sparseToDense_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/gather_nd.js\nfunction gatherND_(x, indices) {\n  const $indices = convertToTensor(indices, \"indices\", \"gatherND\", \"int32\");\n  const $x = convertToTensor(x, \"x\", \"gatherND\", \"string_or_numeric\");\n  const inputs = { params: $x, indices: $indices };\n  return ENGINE.runKernel(GatherNd, inputs);\n}\nvar gatherND = op({ gatherND_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/dropout_util.js\nfunction getNoiseShape(x, noiseShape) {\n  if (noiseShape == null) {\n    return x.shape.slice();\n  }\n  if (arraysEqual(x.shape, noiseShape)) {\n    return noiseShape;\n  }\n  if (x.shape.length === noiseShape.length) {\n    const newDimension = [];\n    for (let i = 0; i < x.shape.length; i++) {\n      if (noiseShape[i] == null && x.shape[i] != null) {\n        newDimension.push(x.shape[i]);\n      } else {\n        newDimension.push(noiseShape[i]);\n      }\n    }\n    return newDimension;\n  }\n  return noiseShape;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/dropout.js\nfunction dropout_(x, rate, noiseShape, seed) {\n  const $x = convertToTensor(x, \"x\", \"dropout\");\n  assert($x.dtype === \"float32\", () => `x has to be a floating point tensor since it's going to be scaled, but got a ${$x.dtype} tensor instead.`);\n  assert(rate >= 0 && rate < 1, () => `rate must be a float in the range [0, 1), but got ${rate}.`);\n  if (rate === 0) {\n    return x instanceof Tensor ? $x.clone() : $x;\n  }\n  const $noiseShape = getNoiseShape($x, noiseShape);\n  const keepProb = 1 - rate;\n  const multiplier = div(floor(add2(randomUniform($noiseShape, 0, 1, \"float32\", seed), keepProb)), keepProb);\n  return mul($x, multiplier);\n}\nvar dropout = op({ dropout_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/signal_ops_util.js\nfunction enclosingPowerOfTwo(value) {\n  return Math.floor(Math.pow(2, Math.ceil(Math.log(value) / Math.log(2))));\n}\nfunction cosineWindow(windowLength, a, b) {\n  const even = 1 - windowLength % 2;\n  const newValues = new Float32Array(windowLength);\n  for (let i = 0; i < windowLength; ++i) {\n    const cosArg = 2 * Math.PI * i / (windowLength + even - 1);\n    newValues[i] = a - b * Math.cos(cosArg);\n  }\n  return tensor1d(newValues, \"float32\");\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/in_top_k.js\nasync function inTopKAsync_(predictions, targets, k = 1) {\n  const $predictions = convertToTensor(predictions, \"predictions\", \"inTopK\");\n  const $targets = convertToTensor(targets, \"targets\", \"inTopK\");\n  assert($predictions.rank > 1, () => `inTopK() expects the predictions to be of rank 2 or higher, but got ${$predictions.rank}`);\n  assert($predictions.rank - 1 === $targets.rank, () => `predictions rank should be 1 larger than targets rank, but got predictions rank ${$predictions.rank} and targets rank ${$targets.rank}`);\n  assertShapesMatch($predictions.shape.slice(0, $predictions.shape.length - 1), $targets.shape, `predictions's shape should be align with the targets' shape, except the last dimension.`);\n  const lastDim = $predictions.shape[$predictions.shape.length - 1];\n  assert(k > 0 && k <= lastDim, () => `'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${lastDim}), but got ${k}`);\n  const predictionsVals = await $predictions.data();\n  const targetsVals = await $targets.data();\n  const [batch, size] = [predictionsVals.length / lastDim, lastDim];\n  const precision3 = getTypedArrayFromDType(\"bool\", batch);\n  for (let b = 0; b < batch; b++) {\n    const offset = b * size;\n    const vals = predictionsVals.subarray(offset, offset + size);\n    const valAndInd = [];\n    for (let i = 0; i < vals.length; i++) {\n      valAndInd.push({ value: vals[i], index: i });\n    }\n    valAndInd.sort((a, b2) => b2.value - a.value);\n    precision3[b] = 0;\n    for (let i = 0; i < k; i++) {\n      if (valAndInd[i].index === targetsVals[b]) {\n        precision3[b] = 1;\n        break;\n      }\n    }\n  }\n  if (predictions !== $predictions) {\n    $predictions.dispose();\n  }\n  if (targets !== $targets) {\n    $targets.dispose();\n  }\n  return tensor(precision3, $targets.shape, \"bool\");\n}\nvar inTopKAsync = inTopKAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fused_ops.js\nvar fused_ops_exports = {};\n__export(fused_ops_exports, {\n  conv2d: () => conv2d2,\n  depthwiseConv2d: () => depthwiseConv2d2,\n  matMul: () => matMul2\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_backprop_filter.js\nfunction conv2DBackpropFilter_(x, dy, filterShape, strides, pad3, dataFormat = \"NHWC\", dimRoundingMode) {\n  let x4D = x;\n  if (x.rank === 3) {\n    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);\n  }\n  let dy4D = dy;\n  if (dy4D.rank === 3) {\n    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in conv2dDerFilter: input must be rank 4, but got shape ${x4D.shape}.`);\n  assert(dy4D.rank === 4, () => `Error in conv2dDerFilter: dy must be rank 4, but got shape ${dy4D.shape}.`);\n  assert(filterShape.length === 4, () => `Error in conv2dDerFilter: filterShape must be length 4, but got ${filterShape}.`);\n  const inDepth = dataFormat === \"NHWC\" ? x4D.shape[3] : x4D.shape[1];\n  const outDepth = dataFormat === \"NHWC\" ? dy4D.shape[3] : dy4D.shape[1];\n  assert(inDepth === filterShape[2], () => `Error in conv2dDerFilter: depth of input ${inDepth}) must match input depth in filter (${filterShape[2]}.`);\n  assert(outDepth === filterShape[3], () => `Error in conv2dDerFilter: depth of dy (${outDepth}) must match output depth for filter (${filterShape[3]}).`);\n  checkPadOnDimRoundingMode(\"conv2dDerFilter\", pad3, dimRoundingMode);\n  const inputs = { x: x4D, dy: dy4D };\n  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape };\n  return ENGINE.runKernel(Conv2DBackpropFilter, inputs, attrs);\n}\nvar conv2DBackpropFilter = op({ conv2DBackpropFilter_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fused_util.js\nfunction getFusedDyActivation(dy, y, activation2) {\n  if (activation2 == null || activation2 === \"linear\") {\n    return dy;\n  }\n  if (activation2 === \"relu\") {\n    return mul(dy, step(y));\n  }\n  throw new Error(`Cannot compute gradient for fused activation ${activation2}.`);\n}\nfunction getFusedBiasGradient(bias, dyActivation) {\n  let res = dyActivation;\n  const reduceAxes = getReductionAxes(bias.shape, dyActivation.shape);\n  if (reduceAxes.length > 0) {\n    res = sum2(res, reduceAxes);\n  }\n  return reshape(res, bias.shape);\n}\nfunction applyActivation(x, activation2, preluActivationWeights, leakyreluAlpha) {\n  if (activation2 === \"linear\") {\n    return x;\n  } else if (activation2 === \"relu\") {\n    return relu(x);\n  } else if (activation2 === \"elu\") {\n    return elu(x);\n  } else if (activation2 === \"relu6\") {\n    return relu6(x);\n  } else if (activation2 === \"prelu\") {\n    return prelu(x, preluActivationWeights);\n  } else if (activation2 === \"leakyrelu\") {\n    return leakyRelu(x, leakyreluAlpha);\n  } else if (activation2 === \"sigmoid\") {\n    return sigmoid(x);\n  }\n  throw new Error(`Unknown fused activation ${activation2}.`);\n}\nvar shouldFuse = (gradientDepth, activation2) => {\n  const gradientMode = gradientDepth > 0;\n  return !gradientMode || activation2 === \"linear\";\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fused/conv2d.js\nfunction fusedConv2d_({ x, filter, strides, pad: pad3, dataFormat = \"NHWC\", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = \"linear\", preluActivationWeights, leakyreluAlpha }) {\n  activation2 = activation2 || \"linear\";\n  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {\n    assert(dataFormat === \"NHWC\", () => `Error in fused conv2d: got dataFormat of ${dataFormat} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);\n    let result = conv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);\n    if (bias != null) {\n      result = add2(result, bias);\n    }\n    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);\n  }\n  const $x = convertToTensor(x, \"x\", \"conv2d\", \"float32\");\n  const $filter = convertToTensor(filter, \"filter\", \"conv2d\", \"float32\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in fused conv2d: input must be rank 4, but got rank ${x4D.rank}.`);\n  assert($filter.rank === 4, () => `Error in fused conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);\n  checkPadOnDimRoundingMode(\"fused conv2d\", pad3, dimRoundingMode);\n  const inputChannels = dataFormat === \"NHWC\" ? x4D.shape[3] : x4D.shape[1];\n  assert($filter.shape[2] === inputChannels, () => `Error in conv2d: depth of input (${inputChannels}) must match input depth for filter ${$filter.shape[2]}.`);\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode);\n  let $bias;\n  if (bias != null) {\n    $bias = convertToTensor(bias, \"bias\", \"fused conv2d\");\n    [$bias] = makeTypesMatch($bias, $x);\n    if (dataFormat === \"NHWC\") {\n      assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);\n    } else {\n      assert($bias.shape.length <= 1, () => `Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${$bias.shape.length}.`);\n      assert($bias.shape.length === 0 || $bias.shape[0] === convInfo.outChannels || $bias.shape[0] === 1, () => `Error in fused conv2d: bias shape (${$bias.shape}) is not compatible with the number of output channels (${convInfo.outChannels})`);\n    }\n  }\n  let $preluActivationWeights;\n  if (preluActivationWeights != null) {\n    const alphaShape = preluActivationWeights.shape;\n    assert(alphaShape.length <= 1 || alphaShape.length === 3, () => `Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${alphaShape.length}.`);\n    if (alphaShape.length === 1) {\n      assert(alphaShape[0] === 1 || alphaShape[0] === convInfo.outChannels, () => `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the number of output channels (${convInfo.outChannels}).`);\n    } else if (alphaShape.length === 3) {\n      try {\n        assertAndGetBroadcastShape(alphaShape, convInfo.outShape);\n      } catch (e) {\n        const errMsg = `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the output shape of the conv2d (${convInfo.outShape}).`;\n        throw Error(errMsg);\n      }\n    }\n    $preluActivationWeights = convertToTensor(preluActivationWeights, \"prelu weights\", \"fused conv2d\");\n  }\n  const grad2 = (dy, saved) => {\n    assert(dataFormat === \"NHWC\", () => `Error in gradient of fused conv2D: got dataFormat of ${dataFormat} but only NHWC is currently supported.`);\n    const [$filter2, x4D2, y, $bias2] = saved;\n    const dyActivation = getFusedDyActivation(dy, y, activation2);\n    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);\n    const xDer = conv2DBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3);\n    const filterDer = conv2DBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3);\n    const der = [xDer, filterDer];\n    if ($bias2 != null) {\n      const biasDer = getFusedBiasGradient($bias2, dyActivation);\n      der.push(biasDer);\n    }\n    return der;\n  };\n  const inputs = {\n    x: x4D,\n    filter: $filter,\n    bias: $bias,\n    preluActivationWeights: $preluActivationWeights\n  };\n  const attrs = {\n    strides,\n    pad: pad3,\n    dataFormat,\n    dilations,\n    dimRoundingMode,\n    activation: activation2,\n    leakyreluAlpha\n  };\n  if (bias == null) {\n    const customOp = customGrad((x4D2, filter2, save) => {\n      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);\n      save([filter2, x4D2, res]);\n      if (reshapedTo4D) {\n        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n      }\n      return { value: res, gradFunc: grad2 };\n    });\n    return customOp(x4D, $filter);\n  } else {\n    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {\n      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);\n      save([filter2, x4D2, res, bias2]);\n      if (reshapedTo4D) {\n        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n      }\n      return { value: res, gradFunc: grad2 };\n    });\n    return customOpWithBias(x4D, $filter, $bias);\n  }\n}\nvar conv2d2 = op({ fusedConv2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d_native_backprop_filter.js\nfunction depthwiseConv2dNativeBackpropFilter_(x, dy, filterShape, strides, pad3, dilations = [1, 1], dimRoundingMode) {\n  let x4D = x;\n  if (x.rank === 3) {\n    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);\n  }\n  let dy4D = dy;\n  if (dy4D.rank === 3) {\n    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);\n  }\n  const inputs = { x: x4D, dy: dy4D };\n  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, filterShape };\n  return ENGINE.runKernel(DepthwiseConv2dNativeBackpropFilter, inputs, attrs);\n}\nvar depthwiseConv2dNativeBackpropFilter = op({ depthwiseConv2dNativeBackpropFilter_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d_native_backprop_input.js\nfunction depthwiseConv2dNativeBackpropInput_(xShape, dy, filter, strides, pad3, dilations = [1, 1], dimRoundingMode) {\n  let dy4D = dy;\n  let reshapedTo4D = false;\n  if (dy.rank === 3) {\n    reshapedTo4D = true;\n    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);\n  }\n  const inputs = { dy: dy4D, filter };\n  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, inputShape: xShape };\n  const res = ENGINE.runKernel(DepthwiseConv2dNativeBackpropInput, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar depthwiseConv2dNativeBackpropInput = op({ depthwiseConv2dNativeBackpropInput_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fused/depthwise_conv2d.js\nfunction fusedDepthwiseConv2d_({ x, filter, strides, pad: pad3, dataFormat = \"NHWC\", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = \"linear\", preluActivationWeights, leakyreluAlpha }) {\n  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {\n    let result = depthwiseConv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);\n    if (bias != null) {\n      result = add2(result, bias);\n    }\n    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);\n  }\n  const $x = convertToTensor(x, \"x\", \"depthwiseConv2d\", \"float32\");\n  const $filter = convertToTensor(filter, \"filter\", \"depthwiseConv2d\", \"float32\");\n  let x4D = $x;\n  let reshapedTo4D = false;\n  if ($x.rank === 3) {\n    reshapedTo4D = true;\n    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);\n  }\n  assert(x4D.rank === 4, () => `Error in fused depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);\n  assert($filter.rank === 4, () => `Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);\n  assert(x4D.shape[3] === $filter.shape[2], () => `Error in fused depthwiseConv2d: number of input channels (${x4D.shape[3]}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);\n  if (dilations == null) {\n    dilations = [1, 1];\n  }\n  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  checkPadOnDimRoundingMode(\"fused depthwiseConv2d\", pad3, dimRoundingMode);\n  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode, true);\n  let $bias;\n  if (bias != null) {\n    $bias = convertToTensor(bias, \"bias\", \"fused conv2d\");\n    [$bias] = makeTypesMatch($bias, $x);\n    assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);\n  }\n  let $preluActivationWeights;\n  if (preluActivationWeights != null) {\n    $preluActivationWeights = convertToTensor(preluActivationWeights, \"prelu weights\", \"fused depthwiseConv2d\");\n  }\n  const grad2 = (dy, saved) => {\n    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${dilations}'`);\n    const [$filter2, x4D2, y, bias2] = saved;\n    const dyActivation = getFusedDyActivation(dy, y, activation2);\n    const xDer = depthwiseConv2dNativeBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3, dilations, dimRoundingMode);\n    const filterDer = depthwiseConv2dNativeBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3, dilations, dimRoundingMode);\n    if (bias2 != null) {\n      const biasDer = getFusedBiasGradient($bias, dyActivation);\n      return [xDer, filterDer, biasDer];\n    }\n    return [xDer, filterDer];\n  };\n  const inputs = {\n    x: x4D,\n    filter: $filter,\n    bias: $bias,\n    preluActivationWeights: $preluActivationWeights\n  };\n  const attrs = {\n    strides,\n    pad: pad3,\n    dataFormat,\n    dilations,\n    dimRoundingMode,\n    activation: activation2,\n    leakyreluAlpha\n  };\n  if (bias == null) {\n    const customOp = customGrad((x4D2, filter2, save) => {\n      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);\n      save([filter2, x4D2, res]);\n      if (reshapedTo4D) {\n        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n      }\n      return { value: res, gradFunc: grad2 };\n    });\n    return customOp(x4D, $filter);\n  } else {\n    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {\n      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);\n      save([filter2, x4D2, res, bias2]);\n      if (reshapedTo4D) {\n        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n      }\n      return { value: res, gradFunc: grad2 };\n    });\n    return customOpWithBias(x4D, $filter, $bias);\n  }\n}\nvar depthwiseConv2d2 = op({ fusedDepthwiseConv2d_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/fused/mat_mul.js\nfunction fusedMatMul_({ a, b, transposeA = false, transposeB = false, bias, activation: activation2 = \"linear\", preluActivationWeights, leakyreluAlpha = 0.2 }) {\n  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {\n    let result = matMul(a, b, transposeA, transposeB);\n    if (bias != null) {\n      result = add2(result, bias);\n    }\n    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);\n  }\n  let $a = convertToTensor(a, \"a\", \"fused matMul\");\n  let $b = convertToTensor(b, \"b\", \"fused matMul\");\n  [$a, $b] = makeTypesMatch($a, $b);\n  const innerShapeA = transposeA ? $a.shape[$a.rank - 2] : $a.shape[$a.rank - 1];\n  const innerShapeB = transposeB ? $b.shape[$b.rank - 1] : $b.shape[$b.rank - 2];\n  const outerShapeA = transposeA ? $a.shape[$a.rank - 1] : $a.shape[$a.rank - 2];\n  const outerShapeB = transposeB ? $b.shape[$b.rank - 2] : $b.shape[$b.rank - 1];\n  const outerDimsA = $a.shape.slice(0, -2);\n  const outerDimsB = $b.shape.slice(0, -2);\n  const batchDimA = sizeFromShape(outerDimsA);\n  const batchDimB = sizeFromShape(outerDimsB);\n  assert(innerShapeA === innerShapeB, () => `Error in fused matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${$a.shape} and ${$b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);\n  const outShapeOuterDims = assertAndGetBroadcastShape($a.shape.slice(0, -2), $b.shape.slice(0, -2));\n  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);\n  const a3D = transposeA ? reshape($a, [batchDimA, innerShapeA, outerShapeA]) : reshape($a, [batchDimA, outerShapeA, innerShapeA]);\n  const b3D = transposeB ? reshape($b, [batchDimB, outerShapeB, innerShapeB]) : reshape($b, [batchDimB, innerShapeB, outerShapeB]);\n  let $bias;\n  if (bias != null) {\n    $bias = convertToTensor(bias, \"bias\", \"fused matMul\");\n    [$bias] = makeTypesMatch($bias, $a);\n    assertAndGetBroadcastShape(outShape, $bias.shape);\n  }\n  let $preluActivationWeights;\n  if (preluActivationWeights != null) {\n    $preluActivationWeights = convertToTensor(preluActivationWeights, \"prelu weights\", \"fused matMul\");\n  }\n  const grad2 = (dy, saved) => {\n    const [a3D2, b3D2, y, $bias2] = saved;\n    const dyActivation = getFusedDyActivation(reshape(dy, y.shape), y, activation2);\n    let aDer;\n    let bDer;\n    if (!transposeA && !transposeB) {\n      aDer = matMul(dyActivation, b3D2, false, true);\n      bDer = matMul(a3D2, dyActivation, true, false);\n    } else if (!transposeA && transposeB) {\n      aDer = matMul(dyActivation, b3D2, false, false);\n      bDer = matMul(dyActivation, a3D2, true, false);\n    } else if (transposeA && !transposeB) {\n      aDer = matMul(b3D2, dyActivation, false, true);\n      bDer = matMul(a3D2, dyActivation, false, false);\n    } else {\n      aDer = matMul(b3D2, dyActivation, true, true);\n      bDer = matMul(dyActivation, a3D2, true, true);\n    }\n    if (bias != null) {\n      const biasDer = getFusedBiasGradient($bias2, dyActivation);\n      return [aDer, bDer, biasDer];\n    } else {\n      return [aDer, bDer];\n    }\n  };\n  const inputs = {\n    a: a3D,\n    b: b3D,\n    bias: $bias,\n    preluActivationWeights: $preluActivationWeights\n  };\n  const attrs = { transposeA, transposeB, activation: activation2, leakyreluAlpha };\n  if (bias == null) {\n    const customOp = customGrad((a3D2, b3D2, save) => {\n      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);\n      save([a3D2, b3D2, res]);\n      return { value: reshape(res, outShape), gradFunc: grad2 };\n    });\n    return customOp(a3D, b3D);\n  } else {\n    const customOpWithBias = customGrad((a3D2, b3D2, $bias2, save) => {\n      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);\n      save([a3D2, b3D2, res, $bias2]);\n      return { value: reshape(res, outShape), gradFunc: grad2 };\n    });\n    return customOpWithBias(a3D, b3D, $bias);\n  }\n}\nvar matMul2 = op({ fusedMatMul_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/signal/hamming_window.js\nfunction hammingWindow_(windowLength) {\n  return cosineWindow(windowLength, 0.54, 0.46);\n}\nvar hammingWindow = op({ hammingWindow_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/signal/hann_window.js\nfunction hannWindow_(windowLength) {\n  return cosineWindow(windowLength, 0.5, 0.5);\n}\nvar hannWindow = op({ hannWindow_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/signal/frame.js\nfunction frame_(signal2, frameLength, frameStep, padEnd = false, padValue = 0) {\n  let start = 0;\n  const output = [];\n  while (start + frameLength <= signal2.size) {\n    output.push(slice(signal2, start, frameLength));\n    start += frameStep;\n  }\n  if (padEnd) {\n    while (start < signal2.size) {\n      const padLen = start + frameLength - signal2.size;\n      const pad3 = concat([\n        slice(signal2, start, frameLength - padLen),\n        fill([padLen], padValue)\n      ]);\n      output.push(pad3);\n      start += frameStep;\n    }\n  }\n  if (output.length === 0) {\n    return tensor2d([], [0, frameLength]);\n  }\n  return reshape(concat(output), [output.length, frameLength]);\n}\nvar frame = op({ frame_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/signal/stft.js\nfunction stft_(signal2, frameLength, frameStep, fftLength, windowFn = hannWindow) {\n  if (fftLength == null) {\n    fftLength = enclosingPowerOfTwo(frameLength);\n  }\n  const framedSignal = frame(signal2, frameLength, frameStep);\n  const windowedSignal = mul(framedSignal, windowFn(frameLength));\n  return rfft(windowedSignal, fftLength);\n}\nvar stft = op({ stft_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/crop_and_resize.js\nfunction cropAndResize_(image2, boxes, boxInd, cropSize, method = \"bilinear\", extrapolationValue = 0) {\n  const $image = convertToTensor(image2, \"image\", \"cropAndResize\");\n  const $boxes = convertToTensor(boxes, \"boxes\", \"cropAndResize\", \"float32\");\n  const $boxInd = convertToTensor(boxInd, \"boxInd\", \"cropAndResize\", \"int32\");\n  const numBoxes = $boxes.shape[0];\n  assert($image.rank === 4, () => `Error in cropAndResize: image must be rank 4,but got rank ${$image.rank}.`);\n  assert($boxes.rank === 2 && $boxes.shape[1] === 4, () => `Error in cropAndResize: boxes must be have size [${numBoxes},4] but had shape ${$boxes.shape}.`);\n  assert($boxInd.rank === 1 && $boxInd.shape[0] === numBoxes, () => `Error in cropAndResize: boxInd must be have size [${numBoxes}] but had shape ${$boxes.shape}.`);\n  assert(cropSize.length === 2, () => `Error in cropAndResize: cropSize must be of length 2, but got length ${cropSize.length}.`);\n  assert(cropSize[0] >= 1 && cropSize[1] >= 1, () => `cropSize must be atleast [1,1], but was ${cropSize}`);\n  assert(method === \"bilinear\" || method === \"nearest\", () => `method must be bilinear or nearest, but was ${method}`);\n  const inputs = { image: $image, boxes: $boxes, boxInd: $boxInd };\n  const attrs = { method, extrapolationValue, cropSize };\n  const res = ENGINE.runKernel(CropAndResize, inputs, attrs);\n  return res;\n}\nvar cropAndResize = op({ cropAndResize_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/flip_left_right.js\nfunction flipLeftRight_(image2) {\n  const $image = convertToTensor(image2, \"image\", \"flipLeftRight\", \"float32\");\n  assert($image.rank === 4, () => `Error in flipLeftRight: image must be rank 4,but got rank ${$image.rank}.`);\n  const inputs = { image: $image };\n  const res = ENGINE.runKernel(FlipLeftRight, inputs, {});\n  return res;\n}\nvar flipLeftRight = op({ flipLeftRight_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/grayscale_to_rgb.js\nfunction grayscaleToRGB_(image2) {\n  const $image = convertToTensor(image2, \"image\", \"grayscaleToRGB\");\n  const lastDimsIdx = $image.rank - 1;\n  const lastDims = $image.shape[lastDimsIdx];\n  assert($image.rank >= 2, () => `Error in grayscaleToRGB: images must be at least rank 2, but got rank ${$image.rank}.`);\n  assert(lastDims === 1, () => `Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${lastDims}.`);\n  const reps = new Array($image.rank);\n  reps.fill(1, 0, lastDimsIdx);\n  reps[lastDimsIdx] = 3;\n  return tile($image, reps);\n}\nvar grayscaleToRGB = op({ grayscaleToRGB_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/rotate_with_offset.js\nfunction rotateWithOffset_(image2, radians, fillValue = 0, center = 0.5) {\n  const $image = convertToTensor(image2, \"image\", \"rotateWithOffset\", \"float32\");\n  assert($image.rank === 4, () => `Error in rotateWithOffset: image must be rank 4,but got rank ${$image.rank}.`);\n  const inputs = { image: $image };\n  const attrs = { radians, fillValue, center };\n  const res = ENGINE.runKernel(RotateWithOffset, inputs, attrs);\n  return res;\n}\nvar rotateWithOffset = op({ rotateWithOffset_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/nonmax_util.js\nfunction nonMaxSuppSanityCheck(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {\n  if (iouThreshold == null) {\n    iouThreshold = 0.5;\n  }\n  if (scoreThreshold == null) {\n    scoreThreshold = Number.NEGATIVE_INFINITY;\n  }\n  if (softNmsSigma == null) {\n    softNmsSigma = 0;\n  }\n  const numBoxes = boxes.shape[0];\n  maxOutputSize = Math.min(maxOutputSize, numBoxes);\n  assert(0 <= iouThreshold && iouThreshold <= 1, () => `iouThreshold must be in [0, 1], but was '${iouThreshold}'`);\n  assert(boxes.rank === 2, () => `boxes must be a 2D tensor, but was of rank '${boxes.rank}'`);\n  assert(boxes.shape[1] === 4, () => `boxes must have 4 columns, but 2nd dimension was ${boxes.shape[1]}`);\n  assert(scores.rank === 1, () => \"scores must be a 1D tensor\");\n  assert(scores.shape[0] === numBoxes, () => `scores has incompatible shape with boxes. Expected ${numBoxes}, but was ${scores.shape[0]}`);\n  assert(0 <= softNmsSigma && softNmsSigma <= 1, () => `softNmsSigma must be in [0, 1], but was '${softNmsSigma}'`);\n  return { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression.js\nfunction nonMaxSuppression_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppression\", \"float32\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppression\", \"float32\");\n  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);\n  maxOutputSize = inputs.maxOutputSize;\n  iouThreshold = inputs.iouThreshold;\n  scoreThreshold = inputs.scoreThreshold;\n  const attrs = { maxOutputSize, iouThreshold, scoreThreshold };\n  return ENGINE.runKernel(NonMaxSuppressionV3, { boxes: $boxes, scores: $scores }, attrs);\n}\nvar nonMaxSuppression = op({ nonMaxSuppression_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/non_max_suppression_util.js\nfunction binaryInsert(arr, element, comparator) {\n  const index = binarySearch(arr, element, comparator);\n  const insertionPoint = index < 0 ? -(index + 1) : index;\n  arr.splice(insertionPoint, 0, element);\n}\nfunction binarySearch(arr, target, comparator) {\n  return binarySearch_(arr, target, comparator || defaultComparator);\n}\nfunction defaultComparator(a, b) {\n  return a > b ? 1 : a < b ? -1 : 0;\n}\nfunction binarySearch_(arr, target, comparator) {\n  let left = 0;\n  let right = arr.length;\n  let middle = 0;\n  let found = false;\n  while (left < right) {\n    middle = left + (right - left >>> 1);\n    const compareResult = comparator(target, arr[middle]);\n    if (compareResult > 0) {\n      left = middle + 1;\n    } else {\n      right = middle;\n      found = !compareResult;\n    }\n  }\n  return found ? left : -left - 1;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/non_max_suppression_impl.js\nfunction nonMaxSuppressionV3Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {\n  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, 0);\n}\nfunction nonMaxSuppressionV4Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize) {\n  return nonMaxSuppressionImpl_(\n    boxes,\n    scores,\n    maxOutputSize,\n    iouThreshold,\n    scoreThreshold,\n    0,\n    false,\n    padToMaxOutputSize,\n    true\n  );\n}\nfunction nonMaxSuppressionV5Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {\n  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, true);\n}\nfunction nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, returnScoresTensor = false, padToMaxOutputSize = false, returnValidOutputs = false) {\n  const candidates = [];\n  for (let i = 0; i < scores.length; i++) {\n    if (scores[i] > scoreThreshold) {\n      candidates.push({ score: scores[i], boxIndex: i, suppressBeginIndex: 0 });\n    }\n  }\n  candidates.sort(ascendingComparator);\n  const scale2 = softNmsSigma > 0 ? -0.5 / softNmsSigma : 0;\n  const selectedIndices = [];\n  const selectedScores = [];\n  while (selectedIndices.length < maxOutputSize && candidates.length > 0) {\n    const candidate = candidates.pop();\n    const { score: originalScore, boxIndex, suppressBeginIndex } = candidate;\n    if (originalScore < scoreThreshold) {\n      break;\n    }\n    let ignoreCandidate = false;\n    for (let j = selectedIndices.length - 1; j >= suppressBeginIndex; --j) {\n      const iou = intersectionOverUnion(boxes, boxIndex, selectedIndices[j]);\n      if (iou >= iouThreshold) {\n        ignoreCandidate = true;\n        break;\n      }\n      candidate.score = candidate.score * suppressWeight(iouThreshold, scale2, iou);\n      if (candidate.score <= scoreThreshold) {\n        break;\n      }\n    }\n    candidate.suppressBeginIndex = selectedIndices.length;\n    if (!ignoreCandidate) {\n      if (candidate.score === originalScore) {\n        selectedIndices.push(boxIndex);\n        selectedScores.push(candidate.score);\n      } else if (candidate.score > scoreThreshold) {\n        binaryInsert(candidates, candidate, ascendingComparator);\n      }\n    }\n  }\n  const validOutputs = selectedIndices.length;\n  const elemsToPad = maxOutputSize - validOutputs;\n  if (padToMaxOutputSize && elemsToPad > 0) {\n    selectedIndices.push(...new Array(elemsToPad).fill(0));\n    selectedScores.push(...new Array(elemsToPad).fill(0));\n  }\n  const result = { selectedIndices };\n  if (returnScoresTensor) {\n    result[\"selectedScores\"] = selectedScores;\n  }\n  if (returnValidOutputs) {\n    result[\"validOutputs\"] = validOutputs;\n  }\n  return result;\n}\nfunction intersectionOverUnion(boxes, i, j) {\n  const iCoord = boxes.subarray(i * 4, i * 4 + 4);\n  const jCoord = boxes.subarray(j * 4, j * 4 + 4);\n  const yminI = Math.min(iCoord[0], iCoord[2]);\n  const xminI = Math.min(iCoord[1], iCoord[3]);\n  const ymaxI = Math.max(iCoord[0], iCoord[2]);\n  const xmaxI = Math.max(iCoord[1], iCoord[3]);\n  const yminJ = Math.min(jCoord[0], jCoord[2]);\n  const xminJ = Math.min(jCoord[1], jCoord[3]);\n  const ymaxJ = Math.max(jCoord[0], jCoord[2]);\n  const xmaxJ = Math.max(jCoord[1], jCoord[3]);\n  const areaI = (ymaxI - yminI) * (xmaxI - xminI);\n  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);\n  if (areaI <= 0 || areaJ <= 0) {\n    return 0;\n  }\n  const intersectionYmin = Math.max(yminI, yminJ);\n  const intersectionXmin = Math.max(xminI, xminJ);\n  const intersectionYmax = Math.min(ymaxI, ymaxJ);\n  const intersectionXmax = Math.min(xmaxI, xmaxJ);\n  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);\n  return intersectionArea / (areaI + areaJ - intersectionArea);\n}\nfunction suppressWeight(iouThreshold, scale2, iou) {\n  const weight = Math.exp(scale2 * iou * iou);\n  return iou <= iouThreshold ? weight : 0;\n}\nfunction ascendingComparator(c1, c2) {\n  return c1.score - c2.score || c1.score === c2.score && c2.boxIndex - c1.boxIndex;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_async.js\nasync function nonMaxSuppressionAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppressionAsync\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppressionAsync\");\n  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);\n  maxOutputSize = inputs.maxOutputSize;\n  iouThreshold = inputs.iouThreshold;\n  scoreThreshold = inputs.scoreThreshold;\n  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);\n  const boxesVals = boxesAndScores[0];\n  const scoresVals = boxesAndScores[1];\n  const { selectedIndices } = nonMaxSuppressionV3Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);\n  if ($boxes !== boxes) {\n    $boxes.dispose();\n  }\n  if ($scores !== scores) {\n    $scores.dispose();\n  }\n  return tensor1d(selectedIndices, \"int32\");\n}\nvar nonMaxSuppressionAsync = nonMaxSuppressionAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_with_score.js\nfunction nonMaxSuppressionWithScore_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppression\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppression\");\n  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);\n  maxOutputSize = params.maxOutputSize;\n  iouThreshold = params.iouThreshold;\n  scoreThreshold = params.scoreThreshold;\n  softNmsSigma = params.softNmsSigma;\n  const inputs = { boxes: $boxes, scores: $scores };\n  const attrs = { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };\n  const result = ENGINE.runKernel(NonMaxSuppressionV5, inputs, attrs);\n  return { selectedIndices: result[0], selectedScores: result[1] };\n}\nvar nonMaxSuppressionWithScore = op({ nonMaxSuppressionWithScore_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_with_score_async.js\nasync function nonMaxSuppressionWithScoreAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppressionAsync\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppressionAsync\");\n  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);\n  maxOutputSize = params.maxOutputSize;\n  iouThreshold = params.iouThreshold;\n  scoreThreshold = params.scoreThreshold;\n  softNmsSigma = params.softNmsSigma;\n  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);\n  const boxesVals = boxesAndScores[0];\n  const scoresVals = boxesAndScores[1];\n  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);\n  if ($boxes !== boxes) {\n    $boxes.dispose();\n  }\n  if ($scores !== scores) {\n    $scores.dispose();\n  }\n  return {\n    selectedIndices: tensor1d(selectedIndices, \"int32\"),\n    selectedScores: tensor1d(selectedScores)\n  };\n}\nvar nonMaxSuppressionWithScoreAsync = nonMaxSuppressionWithScoreAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_padded.js\nfunction nonMaxSuppressionPadded_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppression\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppression\");\n  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);\n  const $maxOutputSize = params.maxOutputSize;\n  const $iouThreshold = params.iouThreshold;\n  const $scoreThreshold = params.scoreThreshold;\n  const inputs = { boxes: $boxes, scores: $scores };\n  const attrs = {\n    maxOutputSize: $maxOutputSize,\n    iouThreshold: $iouThreshold,\n    scoreThreshold: $scoreThreshold,\n    padToMaxOutputSize\n  };\n  const result = ENGINE.runKernel(NonMaxSuppressionV4, inputs, attrs);\n  return { selectedIndices: result[0], validOutputs: result[1] };\n}\nvar nonMaxSuppressionPadded = op({ nonMaxSuppressionPadded_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_padded_async.js\nasync function nonMaxSuppressionPaddedAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {\n  const $boxes = convertToTensor(boxes, \"boxes\", \"nonMaxSuppressionAsync\");\n  const $scores = convertToTensor(scores, \"scores\", \"nonMaxSuppressionAsync\");\n  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);\n  const $maxOutputSize = params.maxOutputSize;\n  const $iouThreshold = params.iouThreshold;\n  const $scoreThreshold = params.scoreThreshold;\n  const [boxesVals, scoresVals] = await Promise.all([$boxes.data(), $scores.data()]);\n  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl(boxesVals, scoresVals, $maxOutputSize, $iouThreshold, $scoreThreshold, padToMaxOutputSize);\n  if ($boxes !== boxes) {\n    $boxes.dispose();\n  }\n  if ($scores !== scores) {\n    $scores.dispose();\n  }\n  return {\n    selectedIndices: tensor1d(selectedIndices, \"int32\"),\n    validOutputs: scalar(validOutputs, \"int32\")\n  };\n}\nvar nonMaxSuppressionPaddedAsync = nonMaxSuppressionPaddedAsync_;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/resize_bilinear.js\nfunction resizeBilinear_(images, size, alignCorners = false, halfPixelCenters = false) {\n  const $images = convertToTensor(images, \"images\", \"resizeBilinear\");\n  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeBilinear: x must be rank 3 or 4, but got rank ${$images.rank}.`);\n  assert(size.length === 2, () => `Error in resizeBilinear: new shape must 2D, but got shape ${size}.`);\n  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.`);\n  let batchImages = $images;\n  let reshapedTo4D = false;\n  if ($images.rank === 3) {\n    reshapedTo4D = true;\n    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);\n  }\n  const [] = size;\n  const inputs = { images: batchImages };\n  const attrs = { alignCorners, halfPixelCenters, size };\n  const res = ENGINE.runKernel(ResizeBilinear, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar resizeBilinear = op({ resizeBilinear_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/resize_nearest_neighbor.js\nfunction resizeNearestNeighbor_(images, size, alignCorners = false, halfPixelCenters = false) {\n  const $images = convertToTensor(images, \"images\", \"resizeNearestNeighbor\");\n  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${$images.rank}.`);\n  assert(size.length === 2, () => `Error in resizeNearestNeighbor: new shape must 2D, but got shape ${size}.`);\n  assert($images.dtype === \"float32\" || $images.dtype === \"int32\", () => \"`images` must have `int32` or `float32` as dtype\");\n  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.`);\n  let batchImages = $images;\n  let reshapedTo4D = false;\n  if ($images.rank === 3) {\n    reshapedTo4D = true;\n    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);\n  }\n  const [] = size;\n  const inputs = { images: batchImages };\n  const attrs = { alignCorners, halfPixelCenters, size };\n  const res = ENGINE.runKernel(ResizeNearestNeighbor, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar resizeNearestNeighbor = op({ resizeNearestNeighbor_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/threshold.js\nfunction threshold_(image2, method = \"binary\", inverted = false, threshValue = 0.5) {\n  const $image = convertToTensor(image2, \"image\", \"threshold\");\n  const RED_INTENCITY_COEF = 0.2989;\n  const GREEN_INTENCITY_COEF = 0.587;\n  const BLUE_INTENCITY_COEF = 0.114;\n  const totalPixelsInImage = $image.shape[0] * $image.shape[1];\n  let $threshold = mul(tensor1d([threshValue]), 255);\n  let r, g, b, grayscale;\n  assert($image.rank === 3, () => `Error in threshold: image must be rank 3,but got rank ${$image.rank}.`);\n  assert($image.shape[2] === 3 || $image.shape[2] === 1, () => `Error in threshold: image color channel must be equal to 3 or 1but got ${$image.shape[2]}.`);\n  assert($image.dtype === \"int32\" || $image.dtype === \"float32\", () => `Error in dtype: image dtype must be int32 or float32,but got dtype ${$image.dtype}.`);\n  assert(method === \"otsu\" || method === \"binary\", () => `Method must be binary or otsu, but was ${method}`);\n  if ($image.shape[2] === 3) {\n    [r, g, b] = split($image, [1, 1, 1], -1);\n    const $r = mul(r, RED_INTENCITY_COEF);\n    const $g = mul(g, GREEN_INTENCITY_COEF);\n    const $b = mul(b, BLUE_INTENCITY_COEF);\n    grayscale = add2(add2($r, $g), $b);\n  } else {\n    grayscale = image2;\n  }\n  if (method === \"otsu\") {\n    const $histogram = bincount(cast(round2(grayscale), \"int32\"), tensor([]), 256);\n    $threshold = otsu($histogram, totalPixelsInImage);\n  }\n  const invCondition = inverted ? lessEqual(grayscale, $threshold) : greater(grayscale, $threshold);\n  const result = cast(mul(invCondition, 255), \"int32\");\n  return result;\n}\nfunction otsu(histogram, total) {\n  let bestThresh = tensor1d([-1]);\n  let bestInBetVar = tensor1d([0]);\n  let cInBetVar = tensor1d([0]);\n  let classFirst, classSecond, meanFirst, meanSec, weightForeground, weightBack;\n  for (let index = 0; index < histogram.size - 1; index++) {\n    classFirst = slice(histogram, 0, index + 1);\n    classSecond = slice(histogram, index + 1);\n    weightForeground = div(sum2(classFirst), total);\n    weightBack = div(sum2(classSecond), total);\n    const meanFirstDivA = sum2(mul(classFirst, range(0, classFirst.size)));\n    meanFirst = div(meanFirstDivA, sum2(classFirst));\n    const meanSecFill = fill(classSecond.shape, classFirst.size);\n    const meanSecAdd = add2(range(0, classSecond.size), meanSecFill);\n    const meanSecMul = mul(classSecond, meanSecAdd);\n    meanSec = div(sum2(meanSecMul), sum2(classSecond));\n    const cInBetVarSubA = sub(meanFirst, meanSec);\n    const cInBetVarSubB = sub(meanFirst, meanSec);\n    const cInBetVarMul = mul(weightForeground, weightBack);\n    cInBetVar = mul(mul(cInBetVarMul, cInBetVarSubA), cInBetVarSubB);\n    const condition = greater(cInBetVar, bestInBetVar);\n    bestInBetVar = where(condition, cInBetVar, bestInBetVar);\n    bestThresh = where(condition, tensor1d([index]), bestThresh);\n  }\n  return bestThresh;\n}\nvar threshold = op({ threshold_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/image/transform.js\nfunction transform_(image2, transforms, interpolation = \"nearest\", fillMode = \"constant\", fillValue = 0, outputShape) {\n  const $image = convertToTensor(image2, \"image\", \"transform\", \"float32\");\n  const $transforms = convertToTensor(transforms, \"transforms\", \"transform\", \"float32\");\n  assert($image.rank === 4, () => `Error in transform: image must be rank 4,but got rank ${$image.rank}.`);\n  assert($transforms.rank === 2 && ($transforms.shape[0] === $image.shape[0] || $transforms.shape[0] === 1) && $transforms.shape[1] === 8, () => `Error in transform: Input transform should be batch x 8 or 1 x 8`);\n  assert(outputShape == null || outputShape.length === 2, () => `Error in transform: outputShape must be [height, width] or null, but got ${outputShape}.`);\n  const inputs = { image: $image, transforms: $transforms };\n  const attrs = { interpolation, fillMode, fillValue, outputShape };\n  return ENGINE.runKernel(Transform, inputs, attrs);\n}\nvar transform = op({ transform_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/band_part.js\nfunction bandPart_(a, numLower, numUpper) {\n  assert(numLower % 1 === 0, () => `bandPart(): numLower must be an integer, got ${numLower}.`);\n  assert(numUpper % 1 === 0, () => `bandPart(): numUpper must be an integer, got ${numUpper}.`);\n  const $a = convertToTensor(a, \"a\", \"bandPart\");\n  assert($a.rank >= 2, () => `bandPart(): Rank must be at least 2, got ${$a.rank}.`);\n  const shape = $a.shape;\n  const [M, N] = $a.shape.slice(-2);\n  if (!(numLower <= M)) {\n    throw new Error(`bandPart(): numLower (${numLower}) must not be greater than the number of rows (${M}).`);\n  }\n  if (!(numUpper <= N)) {\n    throw new Error(`bandPart(): numUpper (${numUpper}) must not be greater than the number of columns (${N}).`);\n  }\n  if (numLower < 0) {\n    numLower = M;\n  }\n  if (numUpper < 0) {\n    numUpper = N;\n  }\n  const i = reshape(range(0, M, 1, \"int32\"), [-1, 1]);\n  const j = range(0, N, 1, \"int32\");\n  const ij = sub(i, j);\n  const inBand = logicalAnd(lessEqual(ij, scalar(+numLower, \"int32\")), greaterEqual(ij, scalar(-numUpper, \"int32\")));\n  const zero = zeros([M, N], $a.dtype);\n  return reshape(stack(unstack(reshape($a, [-1, M, N])).map((mat) => where(inBand, mat, zero))), shape);\n}\nvar bandPart = op({ bandPart_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/gram_schmidt.js\nfunction gramSchmidt_(xs) {\n  let inputIsTensor2D;\n  if (Array.isArray(xs)) {\n    inputIsTensor2D = false;\n    assert(xs != null && xs.length > 0, () => \"Gram-Schmidt process: input must not be null, undefined, or empty\");\n    const dim = xs[0].shape[0];\n    for (let i = 1; i < xs.length; ++i) {\n      assert(xs[i].shape[0] === dim, () => `Gram-Schmidt: Non-unique lengths found in the input vectors: (${xs[i].shape[0]} vs. ${dim})`);\n    }\n  } else {\n    inputIsTensor2D = true;\n    xs = split(xs, xs.shape[0], 0).map((x) => squeeze(x, [0]));\n  }\n  assert(xs.length <= xs[0].shape[0], () => `Gram-Schmidt: Number of vectors (${xs.length}) exceeds number of dimensions (${xs[0].shape[0]}).`);\n  const ys = [];\n  const xs1d = xs;\n  for (let i = 0; i < xs.length; ++i) {\n    ys.push(ENGINE.tidy(() => {\n      let x = xs1d[i];\n      if (i > 0) {\n        for (let j = 0; j < i; ++j) {\n          const proj = mul(sum2(mul(ys[j], x)), ys[j]);\n          x = sub(x, proj);\n        }\n      }\n      return div(x, norm(x, \"euclidean\"));\n    }));\n  }\n  if (inputIsTensor2D) {\n    return stack(ys, 0);\n  } else {\n    return ys;\n  }\n}\nvar gramSchmidt = op({ gramSchmidt_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/qr.js\nfunction qr_(x, fullMatrices = false) {\n  assert(x.rank >= 2, () => `qr() requires input tensor to have a rank >= 2, but got rank ${x.rank}`);\n  if (x.rank === 2) {\n    return qr2d(x, fullMatrices);\n  } else {\n    const outerDimsProd = x.shape.slice(0, x.shape.length - 2).reduce((value, prev) => value * prev);\n    const x2ds = unstack(reshape(x, [\n      outerDimsProd,\n      x.shape[x.shape.length - 2],\n      x.shape[x.shape.length - 1]\n    ]), 0);\n    const q2ds = [];\n    const r2ds = [];\n    x2ds.forEach((x2d) => {\n      const [q2d, r2d] = qr2d(x2d, fullMatrices);\n      q2ds.push(q2d);\n      r2ds.push(r2d);\n    });\n    const q = reshape(stack(q2ds, 0), x.shape);\n    const r = reshape(stack(r2ds, 0), x.shape);\n    return [q, r];\n  }\n}\nfunction qr2d(x, fullMatrices = false) {\n  return ENGINE.tidy(() => {\n    assert(x.shape.length === 2, () => `qr2d() requires a 2D Tensor, but got a ${x.shape.length}D Tensor.`);\n    const m = x.shape[0];\n    const n = x.shape[1];\n    let q = eye(m);\n    let r = clone(x);\n    const one2D = tensor2d([[1]], [1, 1]);\n    let w = clone(one2D);\n    const iters = m >= n ? n : m;\n    for (let j = 0; j < iters; ++j) {\n      const rTemp = r;\n      const wTemp = w;\n      const qTemp = q;\n      [w, r, q] = ENGINE.tidy(() => {\n        const rjEnd1 = slice(r, [j, j], [m - j, 1]);\n        const normX = norm(rjEnd1);\n        const rjj = slice(r, [j, j], [1, 1]);\n        const s = where(greater(rjj, 0), tensor2d([[-1]]), tensor2d([[1]]));\n        const u1 = sub(rjj, mul(s, normX));\n        const wPre = div(rjEnd1, u1);\n        if (wPre.shape[0] === 1) {\n          w = clone(one2D);\n        } else {\n          w = concat([\n            one2D,\n            slice(wPre, [1, 0], [wPre.shape[0] - 1, wPre.shape[1]])\n          ], 0);\n        }\n        const tau = neg(div(matMul(s, u1), normX));\n        const rjEndAll = slice(r, [j, 0], [m - j, n]);\n        const tauTimesW = mul(tau, w);\n        const wT = transpose(w);\n        if (j === 0) {\n          r = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));\n        } else {\n          const rTimesTau = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));\n          r = concat([slice(r, [0, 0], [j, n]), rTimesTau], 0);\n        }\n        const tawTimesWT = transpose(tauTimesW);\n        const qAllJEnd = slice(q, [0, j], [m, q.shape[1] - j]);\n        if (j === 0) {\n          q = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));\n        } else {\n          const qTimesTau = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));\n          q = concat([slice(q, [0, 0], [m, j]), qTimesTau], 1);\n        }\n        return [w, r, q];\n      });\n      dispose([rTemp, wTemp, qTemp]);\n    }\n    if (!fullMatrices && m > n) {\n      q = slice(q, [0, 0], [m, n]);\n      r = slice(r, [0, 0], [n, n]);\n    }\n    return [q, r];\n  });\n}\nvar qr = op({ qr_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/loss_ops_utils.js\nvar Reduction;\n(function(Reduction2) {\n  Reduction2[Reduction2[\"NONE\"] = 0] = \"NONE\";\n  Reduction2[Reduction2[\"MEAN\"] = 1] = \"MEAN\";\n  Reduction2[Reduction2[\"SUM\"] = 2] = \"SUM\";\n  Reduction2[Reduction2[\"SUM_BY_NONZERO_WEIGHTS\"] = 3] = \"SUM_BY_NONZERO_WEIGHTS\";\n})(Reduction || (Reduction = {}));\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/compute_weighted_loss.js\nfunction computeWeightedLoss_(losses2, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $losses = convertToTensor(losses2, \"losses\", \"computeWeightedLoss\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"computeWeightedLoss\");\n  }\n  const weightedLoss = $weights == null ? $losses : mul($losses, $weights);\n  if (reduction === Reduction.NONE) {\n    return weightedLoss;\n  }\n  if (reduction === Reduction.SUM) {\n    return sum2(weightedLoss);\n  }\n  if (reduction === Reduction.MEAN) {\n    if ($weights == null) {\n      return mean(weightedLoss);\n    } else {\n      const broadcastFactor = $losses.size / $weights.size;\n      const result = div(sum2(weightedLoss), sum2($weights));\n      return broadcastFactor > 1 ? div(result, scalar(broadcastFactor)) : result;\n    }\n  }\n  if (reduction === Reduction.SUM_BY_NONZERO_WEIGHTS) {\n    if ($weights == null) {\n      return div(sum2(weightedLoss), scalar($losses.size));\n    } else {\n      const broadcastedWeights = mul($weights, ones2($losses.shape));\n      const numNonZeros = cast(sum2(notEqual(broadcastedWeights, scalar(0))), \"float32\");\n      return div(sum2(weightedLoss), numNonZeros);\n    }\n  }\n  throw Error(`Unknown reduction: ${reduction}`);\n}\nvar computeWeightedLoss = op({ computeWeightedLoss_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/absolute_difference.js\nfunction absoluteDifference_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $labels = convertToTensor(labels, \"labels\", \"absoluteDifference\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"absoluteDifference\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"absoluteDifference\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in absoluteDifference: \");\n  const losses2 = abs(sub($labels, $predictions));\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar absoluteDifference = op({ absoluteDifference_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/cosine_distance.js\nfunction cosineDistance_(labels, predictions, axis, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $labels = convertToTensor(labels, \"labels\", \"cosineDistance\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"cosineDistance\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"cosineDistance\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in cosineDistance: \");\n  const one = scalar(1);\n  const losses2 = sub(one, sum2(mul($labels, $predictions), axis, true));\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar cosineDistance = op({ cosineDistance_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/hinge_loss.js\nfunction hingeLoss_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  let $labels = convertToTensor(labels, \"labels\", \"hingeLoss\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"hingeLoss\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"hingeLoss\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in hingeLoss: \");\n  const one = scalar(1);\n  $labels = sub(mul(scalar(2), $labels), one);\n  const losses2 = relu(sub(one, mul($labels, $predictions)));\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar hingeLoss = op({ hingeLoss_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/huber_loss.js\nfunction huberLoss_(labels, predictions, weights, delta = 1, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $labels = convertToTensor(labels, \"labels\", \"huberLoss\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"huberLoss\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"huberLoss\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in huberLoss: \");\n  const deltaScalar = scalar(delta);\n  const error = abs(sub($predictions, $labels));\n  const quadratic = minimum(error, deltaScalar);\n  const linear = sub(error, quadratic);\n  const losses2 = add2(mul(scalar(0.5), square(quadratic)), mul(deltaScalar, linear));\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar huberLoss = op({ huberLoss_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/log_loss.js\nfunction logLoss_(labels, predictions, weights, epsilon3 = 1e-7, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $labels = convertToTensor(labels, \"labels\", \"logLoss\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"logLoss\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"logLoss\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in logLoss: \");\n  const one = scalar(1);\n  const epsilonScalar = scalar(epsilon3);\n  const l13 = neg(mul($labels, log2(add2($predictions, epsilonScalar))));\n  const l23 = mul(sub(one, $labels), log2(add2(sub(one, $predictions), epsilonScalar)));\n  const losses2 = sub(l13, l23);\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar logLoss = op({ logLoss_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/mean_squared_error.js\nfunction meanSquaredError_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  const $labels = convertToTensor(labels, \"labels\", \"meanSquaredError\");\n  const $predictions = convertToTensor(predictions, \"predictions\", \"meanSquaredError\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"meanSquaredError\");\n  }\n  assertShapesMatch($labels.shape, $predictions.shape, \"Error in meanSquaredError: \");\n  const losses2 = squaredDifference($labels, $predictions);\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar meanSquaredError = op({ meanSquaredError_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/sigmoid_cross_entropy.js\nfunction sigmoidCrossEntropyWithLogits_(labels, logits) {\n  const $labels = convertToTensor(labels, \"labels\", \"sigmoidCrossEntropyWithLogits\");\n  const $logits = convertToTensor(logits, \"logits\", \"sigmoidCrossEntropyWithLogits\");\n  assertShapesMatch($labels.shape, $logits.shape, \"Error in sigmoidCrossEntropyWithLogits: \");\n  const maxOutput = relu($logits);\n  const outputXTarget = mul($logits, $labels);\n  const sigmoidOutput = log1p(exp(neg(abs($logits))));\n  return add2(sub(maxOutput, outputXTarget), sigmoidOutput);\n}\nfunction sigmoidCrossEntropy_(multiClassLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  let $multiClassLabels = convertToTensor(multiClassLabels, \"multiClassLabels\", \"sigmoidCrossEntropy\");\n  const $logits = convertToTensor(logits, \"logits\", \"sigmoidCrossEntropy\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"sigmoidCrossEntropy\");\n  }\n  assertShapesMatch($multiClassLabels.shape, $logits.shape, \"Error in sigmoidCrossEntropy: \");\n  if (labelSmoothing > 0) {\n    const labelSmoothingScalar = scalar(labelSmoothing);\n    const one = scalar(1);\n    const half = scalar(0.5);\n    $multiClassLabels = add2(mul($multiClassLabels, sub(one, labelSmoothingScalar)), mul(half, labelSmoothingScalar));\n  }\n  const losses2 = sigmoidCrossEntropyWithLogits_($multiClassLabels, $logits);\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar sigmoidCrossEntropy = op({ sigmoidCrossEntropy_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/losses/softmax_cross_entropy.js\nfunction softmaxCrossEntropyWithLogits_(labels, logits, dim = -1) {\n  if (dim === -1) {\n    dim = logits.rank - 1;\n  }\n  if (dim !== logits.rank - 1) {\n    throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${logits.rank} and dim was ${dim}`);\n  }\n  const customOp = customGrad((labels2, logits2, save) => {\n    const keepDims = true;\n    const lse = logSumExp(logits2, [dim], keepDims);\n    const logResult = sub(cast(logits2, \"float32\"), lse);\n    save([labels2, logResult]);\n    const costVector = neg(mul(logResult, labels2));\n    const value = sum2(costVector, [dim]);\n    const gradFunc = (dy, saved) => {\n      const [labels3, logResult2] = saved;\n      const dyShape = expandShapeToKeepDim(dy.shape, [dim]);\n      return [\n        mul(reshape(dy, dyShape), sub(cast(labels3, \"float32\"), exp(logResult2))),\n        mul(reshape(dy, dyShape), sub(exp(logResult2), cast(labels3, \"float32\")))\n      ];\n    };\n    return { value, gradFunc };\n  });\n  return customOp(labels, logits);\n}\nfunction softmaxCrossEntropy_(onehotLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {\n  let $onehotLabels = convertToTensor(onehotLabels, \"onehotLabels\", \"softmaxCrossEntropy\");\n  const $logits = convertToTensor(logits, \"logits\", \"softmaxCrossEntropy\");\n  let $weights = null;\n  if (weights != null) {\n    $weights = convertToTensor(weights, \"weights\", \"softmaxCrossEntropy\");\n  }\n  assertShapesMatch($onehotLabels.shape, $logits.shape, \"Error in softmaxCrossEntropy: \");\n  if (labelSmoothing > 0) {\n    const labelSmoothingScalar = scalar(labelSmoothing);\n    const one = scalar(1);\n    const numClasses = scalar($onehotLabels.shape[1]);\n    $onehotLabels = add2(mul($onehotLabels, sub(one, labelSmoothingScalar)), div(labelSmoothingScalar, numClasses));\n  }\n  const losses2 = softmaxCrossEntropyWithLogits_($onehotLabels, $logits);\n  return computeWeightedLoss(losses2, $weights, reduction);\n}\nvar softmaxCrossEntropy = op({ softmaxCrossEntropy_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_fill_empty_rows.js\nfunction sparseFillEmptyRows_(indices, values, denseShape, defaultValue) {\n  const $indices = convertToTensor(indices, \"indices\", \"sparseFillEmptyRows\", \"int32\");\n  const $values = convertToTensor(values, \"values\", \"sparseFillEmptyRows\");\n  const $denseShape = convertToTensor(denseShape, \"denseShape\", \"sparseFillEmptyRows\", \"int32\");\n  const $defaultValue = convertToTensor(defaultValue, \"defaultValue\", \"sparseFillEmptyRows\", $values.dtype);\n  if ($indices.rank !== 2) {\n    throw new Error(`Indices should be Tensor2D but received shape\n        ${$indices.shape}`);\n  }\n  if ($values.rank !== 1) {\n    throw new Error(`Values should be Tensor1D but received shape ${$values.shape}`);\n  }\n  if ($denseShape.rank !== 1) {\n    throw new Error(`Dense shape should be Tensor1D but received shape ${$denseShape.shape}`);\n  }\n  if ($defaultValue.rank !== 0) {\n    throw new Error(`Default value should be a scalar but received shape ${$defaultValue.shape}`);\n  }\n  const inputs = {\n    indices: $indices,\n    values: $values,\n    denseShape: $denseShape,\n    defaultValue: $defaultValue\n  };\n  const result = ENGINE.runKernel(SparseFillEmptyRows, inputs);\n  return {\n    outputIndices: result[0],\n    outputValues: result[1],\n    emptyRowIndicator: result[2],\n    reverseIndexMap: result[3]\n  };\n}\nvar sparseFillEmptyRows = op({ sparseFillEmptyRows_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_reshape.js\nfunction sparseReshape_(inputIndices, inputShape, newShape) {\n  const $inputIndices = convertToTensor(inputIndices, \"inputIndices\", \"sparseReshape\", \"int32\");\n  const $inputShape = convertToTensor(inputShape, \"inputShape\", \"sparseReshape\", \"int32\");\n  const $newShape = convertToTensor(newShape, \"newShape\", \"sparseReshape\", \"int32\");\n  if ($inputIndices.rank !== 2) {\n    throw new Error(`Input indices should be Tensor2D but received shape\n        ${$inputIndices.shape}`);\n  }\n  if ($inputShape.rank !== 1) {\n    throw new Error(`Input shape should be Tensor1D but received shape ${$inputShape.shape}`);\n  }\n  if ($newShape.rank !== 1) {\n    throw new Error(`New shape should be Tensor1D but received shape ${$newShape.shape}`);\n  }\n  const inputs = {\n    inputIndices: $inputIndices,\n    inputShape: $inputShape,\n    newShape: $newShape\n  };\n  const result = ENGINE.runKernel(SparseReshape, inputs);\n  return { outputIndices: result[0], outputShape: result[1] };\n}\nvar sparseReshape = op({ sparseReshape_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_mean.js\nfunction sparseSegmentMean_(data, indices, segmentIds) {\n  const $data = convertToTensor(data, \"data\", \"sparseSegmentMean\");\n  const $indices = convertToTensor(indices, \"indices\", \"sparseSegmentMean\", \"int32\");\n  const $segmentIds = convertToTensor(segmentIds, \"segmentIds\", \"sparseSegmentMean\", \"int32\");\n  if ($data.rank < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if ($indices.rank !== 1) {\n    throw new Error(`Indices should be Tensor1D but received shape\n          ${$indices.shape}`);\n  }\n  if ($segmentIds.rank !== 1) {\n    throw new Error(`Segment ids should be Tensor1D but received shape\n          ${$segmentIds.shape}`);\n  }\n  const inputs = {\n    data: $data,\n    indices: $indices,\n    segmentIds: $segmentIds\n  };\n  return ENGINE.runKernel(SparseSegmentMean, inputs);\n}\nvar sparseSegmentMean = op({ sparseSegmentMean_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_sum.js\nfunction sparseSegmentSum_(data, indices, segmentIds) {\n  const $data = convertToTensor(data, \"data\", \"sparseSegmentSum\");\n  const $indices = convertToTensor(indices, \"indices\", \"sparseSegmentSum\", \"int32\");\n  const $segmentIds = convertToTensor(segmentIds, \"segmentIds\", \"sparseSegmentSum\", \"int32\");\n  if ($data.rank < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if ($indices.rank !== 1) {\n    throw new Error(`Indices should be Tensor1D but received shape\n         ${$indices.shape}`);\n  }\n  if ($segmentIds.rank !== 1) {\n    throw new Error(`Segment ids should be Tensor1D but received shape\n         ${$segmentIds.shape}`);\n  }\n  const inputs = {\n    data: $data,\n    indices: $indices,\n    segmentIds: $segmentIds\n  };\n  return ENGINE.runKernel(SparseSegmentSum, inputs);\n}\nvar sparseSegmentSum = op({ sparseSegmentSum_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_n_grams.js\nfunction stringNGrams_(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {\n  const $data = convertToTensor(data, \"data\", \"stringNGrams\", \"string\");\n  if ($data.dtype !== \"string\") {\n    throw new Error(\"Data must be of datatype string\");\n  }\n  if ($data.shape.length !== 1) {\n    throw new Error(`Data must be a vector, saw: ${$data.shape}`);\n  }\n  const $dataSplits = convertToTensor(dataSplits, \"dataSplits\", \"stringNGrams\");\n  if ($dataSplits.dtype !== \"int32\") {\n    throw new Error(\"Data splits must be of datatype int32\");\n  }\n  const attrs = {\n    separator,\n    nGramWidths,\n    leftPad,\n    rightPad: rightPad2,\n    padWidth,\n    preserveShortSequences\n  };\n  const inputs = { data: $data, dataSplits: $dataSplits };\n  const result = ENGINE.runKernel(StringNGrams, inputs, attrs);\n  return { nGrams: result[0], nGramsSplits: result[1] };\n}\nvar stringNGrams = op({ stringNGrams_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_split.js\nfunction stringSplit_(input2, delimiter, skipEmpty = true) {\n  const $input = convertToTensor(input2, \"input\", \"stringSplit\", \"string\");\n  const $delimiter = convertToTensor(delimiter, \"delimiter\", \"stringSplit\", \"string\");\n  if ($input.rank !== 1) {\n    throw new Error(`Input should be Tensor1D but received shape ${$input.shape}`);\n  }\n  if ($delimiter.rank !== 0) {\n    throw new Error(`Delimiter should be a scalar but received shape ${$delimiter.shape}`);\n  }\n  const attrs = { skipEmpty };\n  const inputs = { input: $input, delimiter: $delimiter };\n  const result = ENGINE.runKernel(StringSplit, inputs, attrs);\n  return { indices: result[0], values: result[1], shape: result[2] };\n}\nvar stringSplit = op({ stringSplit_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_to_hash_bucket_fast.js\nfunction stringToHashBucketFast_(input2, numBuckets) {\n  const $input = convertToTensor(input2, \"input\", \"stringToHashBucketFast\", \"string\");\n  const attrs = { numBuckets };\n  if (numBuckets <= 0) {\n    throw new Error(`Number of buckets must be at least 1`);\n  }\n  const inputs = { input: $input };\n  return ENGINE.runKernel(StringToHashBucketFast, inputs, attrs);\n}\nvar stringToHashBucketFast = op({ stringToHashBucketFast_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ops.js\nvar spectral = {\n  fft,\n  ifft,\n  rfft,\n  irfft\n};\nvar signal = {\n  hammingWindow,\n  hannWindow,\n  frame,\n  stft\n};\nvar image = {\n  flipLeftRight,\n  grayscaleToRGB,\n  resizeNearestNeighbor,\n  resizeBilinear,\n  rotateWithOffset,\n  cropAndResize,\n  nonMaxSuppression,\n  nonMaxSuppressionAsync,\n  nonMaxSuppressionWithScore,\n  nonMaxSuppressionWithScoreAsync,\n  nonMaxSuppressionPadded,\n  nonMaxSuppressionPaddedAsync,\n  threshold,\n  transform\n};\nvar linalg = {\n  bandPart,\n  gramSchmidt,\n  qr\n};\nvar losses = {\n  absoluteDifference,\n  computeWeightedLoss,\n  cosineDistance,\n  hingeLoss,\n  huberLoss,\n  logLoss,\n  meanSquaredError,\n  sigmoidCrossEntropy,\n  softmaxCrossEntropy\n};\nvar sparse = {\n  sparseFillEmptyRows,\n  sparseReshape,\n  sparseSegmentMean,\n  sparseSegmentSum\n};\nvar string = {\n  stringNGrams,\n  stringSplit,\n  stringToHashBucketFast\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/optimizer.js\nvar Optimizer = class extends Serializable {\n  minimize(f, returnCost = false, varList) {\n    const { value, grads: grads2 } = this.computeGradients(f, varList);\n    if (varList != null) {\n      const gradArray = varList.map((v) => ({ name: v.name, tensor: grads2[v.name] }));\n      this.applyGradients(gradArray);\n    } else {\n      this.applyGradients(grads2);\n    }\n    dispose(grads2);\n    if (returnCost) {\n      return value;\n    } else {\n      value.dispose();\n      return null;\n    }\n  }\n  get iterations() {\n    if (this.iterations_ == null) {\n      this.iterations_ = 0;\n    }\n    return this.iterations_;\n  }\n  incrementIterations() {\n    this.iterations_ = this.iterations + 1;\n  }\n  computeGradients(f, varList) {\n    return variableGrads(f, varList);\n  }\n  dispose() {\n    if (this.iterations_ != null) {\n      dispose(this.iterations_);\n    }\n  }\n  async saveIterations() {\n    if (this.iterations_ == null) {\n      this.iterations_ = 0;\n    }\n    return {\n      name: \"iter\",\n      tensor: scalar(this.iterations_, \"int32\")\n    };\n  }\n  async getWeights() {\n    throw new Error(\"getWeights() is not implemented for this optimizer yet.\");\n  }\n  async setWeights(weightValues) {\n    throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`);\n  }\n  async extractIterations(weightValues) {\n    this.iterations_ = (await weightValues[0].tensor.data())[0];\n    return weightValues.slice(1);\n  }\n};\nObject.defineProperty(Optimizer, Symbol.hasInstance, {\n  value: (instance) => {\n    return instance.minimize != null && instance.computeGradients != null && instance.applyGradients != null;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/adadelta_optimizer.js\nvar AdadeltaOptimizer = class extends Optimizer {\n  constructor(learningRate, rho, epsilon3 = null) {\n    super();\n    this.learningRate = learningRate;\n    this.rho = rho;\n    this.epsilon = epsilon3;\n    this.accumulatedGrads = [];\n    this.accumulatedUpdates = [];\n    if (epsilon3 == null) {\n      this.epsilon = ENGINE.backend.epsilon();\n    }\n  }\n  applyGradients(variableGradients) {\n    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);\n    variableNames.forEach((name, i) => {\n      const value = ENGINE.registeredVariables[name];\n      const trainable = false;\n      if (this.accumulatedGrads[i] == null) {\n        this.accumulatedGrads[i] = {\n          originalName: `${name}/accum_grad`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      if (this.accumulatedUpdates[i] == null) {\n        this.accumulatedUpdates[i] = {\n          originalName: `${name}/accum_var`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n      if (gradient == null) {\n        return;\n      }\n      const accumulatedGrad = this.accumulatedGrads[i].variable;\n      const accumulatedUpdate = this.accumulatedUpdates[i].variable;\n      tidy(() => {\n        const newAccumulatedGrad = add2(mul(accumulatedGrad, this.rho), mul(square(gradient), 1 - this.rho));\n        const updates = mul(div(sqrt(add2(accumulatedUpdate, this.epsilon)), sqrt(add2(accumulatedGrad, this.epsilon))), gradient);\n        const newAccumulatedUpdate = add2(mul(accumulatedUpdate, this.rho), mul(square(updates), 1 - this.rho));\n        accumulatedGrad.assign(newAccumulatedGrad);\n        accumulatedUpdate.assign(newAccumulatedUpdate);\n        const newValue = add2(mul(updates, -this.learningRate), value);\n        value.assign(newValue);\n      });\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    if (this.accumulatedUpdates != null) {\n      dispose(this.accumulatedGrads.map((v) => v.variable));\n      dispose(this.accumulatedUpdates.map((v) => v.variable));\n    }\n  }\n  async getWeights() {\n    const variables = [...this.accumulatedGrads, ...this.accumulatedUpdates];\n    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    const variableCount = weightValues.length / 2;\n    const trainable = false;\n    this.accumulatedGrads = weightValues.slice(0, variableCount).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n    this.accumulatedUpdates = weightValues.slice(variableCount, variableCount * 2).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"rho\": this.rho,\n      \"epsilon\": this.epsilon\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"rho\"], config[\"epsilon\"]);\n  }\n};\nAdadeltaOptimizer.className = \"Adadelta\";\nregisterClass(AdadeltaOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/adagrad_optimizer.js\nvar AdagradOptimizer = class extends Optimizer {\n  constructor(learningRate, initialAccumulatorValue = 0.1) {\n    super();\n    this.learningRate = learningRate;\n    this.initialAccumulatorValue = initialAccumulatorValue;\n    this.accumulatedGrads = [];\n  }\n  applyGradients(variableGradients) {\n    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);\n    variableNames.forEach((name, i) => {\n      const value = ENGINE.registeredVariables[name];\n      if (this.accumulatedGrads[i] == null) {\n        const trainable = false;\n        this.accumulatedGrads[i] = {\n          originalName: `${name}/accumulator`,\n          variable: tidy(() => fill(value.shape, this.initialAccumulatorValue).variable(trainable))\n        };\n      }\n      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n      if (gradient == null) {\n        return;\n      }\n      const accumulatedGrad = this.accumulatedGrads[i].variable;\n      tidy(() => {\n        const newAccumulatedGrad = add2(accumulatedGrad, square(gradient));\n        accumulatedGrad.assign(newAccumulatedGrad);\n        const newValue = add2(mul(div(gradient, sqrt(add2(newAccumulatedGrad, ENGINE.backend.epsilon()))), -this.learningRate), value);\n        value.assign(newValue);\n      });\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    if (this.accumulatedGrads != null) {\n      dispose(this.accumulatedGrads.map((v) => v.variable));\n    }\n  }\n  async getWeights() {\n    return [await this.saveIterations()].concat(this.accumulatedGrads.map((v) => ({ name: v.originalName, tensor: v.variable })));\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    const trainable = false;\n    this.accumulatedGrads = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"initialAccumulatorValue\": this.initialAccumulatorValue\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"initialAccumulatorValue\"]);\n  }\n};\nAdagradOptimizer.className = \"Adagrad\";\nregisterClass(AdagradOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/adam_optimizer.js\nvar AdamOptimizer = class extends Optimizer {\n  constructor(learningRate, beta1, beta2, epsilon3 = null) {\n    super();\n    this.learningRate = learningRate;\n    this.beta1 = beta1;\n    this.beta2 = beta2;\n    this.epsilon = epsilon3;\n    this.accumulatedFirstMoment = [];\n    this.accumulatedSecondMoment = [];\n    tidy(() => {\n      this.accBeta1 = scalar(beta1).variable();\n      this.accBeta2 = scalar(beta2).variable();\n    });\n    if (epsilon3 == null) {\n      this.epsilon = ENGINE.backend.epsilon();\n    }\n  }\n  applyGradients(variableGradients) {\n    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);\n    tidy(() => {\n      const oneMinusAccBeta1 = sub(1, this.accBeta1);\n      const oneMinusAccBeta2 = sub(1, this.accBeta2);\n      varNames.forEach((name, i) => {\n        const value = ENGINE.registeredVariables[name];\n        const trainable = false;\n        if (this.accumulatedFirstMoment[i] == null) {\n          this.accumulatedFirstMoment[i] = {\n            originalName: `${name}/m`,\n            variable: tidy(() => zerosLike(value).variable(trainable))\n          };\n        }\n        if (this.accumulatedSecondMoment[i] == null) {\n          this.accumulatedSecondMoment[i] = {\n            originalName: `${name}/v`,\n            variable: tidy(() => zerosLike(value).variable(trainable))\n          };\n        }\n        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n        if (gradient == null) {\n          return;\n        }\n        const firstMoment = this.accumulatedFirstMoment[i].variable;\n        const secondMoment = this.accumulatedSecondMoment[i].variable;\n        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));\n        const newSecondMoment = add2(mul(secondMoment, this.beta2), mul(square(gradient), 1 - this.beta2));\n        const biasCorrectedFirstMoment = div(newFirstMoment, oneMinusAccBeta1);\n        const biasCorrectedSecondMoment = div(newSecondMoment, oneMinusAccBeta2);\n        firstMoment.assign(newFirstMoment);\n        secondMoment.assign(newSecondMoment);\n        const newValue = add2(mul(div(biasCorrectedFirstMoment, add2(sqrt(biasCorrectedSecondMoment), this.epsilon)), -this.learningRate), value);\n        value.assign(newValue);\n      });\n      this.accBeta1.assign(mul(this.accBeta1, this.beta1));\n      this.accBeta2.assign(mul(this.accBeta2, this.beta2));\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    this.accBeta1.dispose();\n    this.accBeta2.dispose();\n    if (this.accumulatedFirstMoment != null) {\n      dispose(this.accumulatedFirstMoment.map((v) => v.variable));\n    }\n    if (this.accumulatedSecondMoment != null) {\n      dispose(this.accumulatedSecondMoment.map((v) => v.variable));\n    }\n  }\n  async getWeights() {\n    const variables = [...this.accumulatedFirstMoment, ...this.accumulatedSecondMoment];\n    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    tidy(() => {\n      this.accBeta1.assign(pow(this.beta1, this.iterations_ + 1));\n      this.accBeta2.assign(pow(this.beta2, this.iterations_ + 1));\n    });\n    const variableCount = weightValues.length / 2;\n    const trainable = false;\n    this.accumulatedFirstMoment = weightValues.slice(0, variableCount).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n    this.accumulatedSecondMoment = weightValues.slice(variableCount, variableCount * 2).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"beta1\": this.beta1,\n      \"beta2\": this.beta2,\n      \"epsilon\": this.epsilon\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"beta1\"], config[\"beta2\"], config[\"epsilon\"]);\n  }\n};\nAdamOptimizer.className = \"Adam\";\nregisterClass(AdamOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/adamax_optimizer.js\nvar AdamaxOptimizer = class extends Optimizer {\n  constructor(learningRate, beta1, beta2, epsilon3 = null, decay = 0) {\n    super();\n    this.learningRate = learningRate;\n    this.beta1 = beta1;\n    this.beta2 = beta2;\n    this.epsilon = epsilon3;\n    this.decay = decay;\n    this.accumulatedFirstMoment = [];\n    this.accumulatedWeightedInfNorm = [];\n    tidy(() => {\n      this.iteration = scalar(0).variable();\n      this.accBeta1 = scalar(beta1).variable();\n    });\n    if (epsilon3 == null) {\n      this.epsilon = ENGINE.backend.epsilon();\n    }\n  }\n  applyGradients(variableGradients) {\n    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);\n    tidy(() => {\n      const oneMinusAccBeta1 = sub(1, this.accBeta1);\n      const lr = div(-this.learningRate, add2(mul(this.iteration, this.decay), 1));\n      variableNames.forEach((name, i) => {\n        const value = ENGINE.registeredVariables[name];\n        const trainable = false;\n        if (this.accumulatedFirstMoment[i] == null) {\n          this.accumulatedFirstMoment[i] = {\n            originalName: `${name}/m`,\n            variable: zerosLike(value).variable(trainable)\n          };\n        }\n        if (this.accumulatedWeightedInfNorm[i] == null) {\n          this.accumulatedWeightedInfNorm[i] = {\n            originalName: `${name}/v`,\n            variable: zerosLike(value).variable(trainable)\n          };\n        }\n        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n        if (gradient == null) {\n          return;\n        }\n        const firstMoment = this.accumulatedFirstMoment[i].variable;\n        const weightedInfNorm = this.accumulatedWeightedInfNorm[i].variable;\n        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));\n        const ut0 = mul(weightedInfNorm, this.beta2);\n        const ut1 = abs(gradient);\n        const newWeightedInfNorm = maximum(ut0, ut1);\n        firstMoment.assign(newFirstMoment);\n        weightedInfNorm.assign(newWeightedInfNorm);\n        const newValue = add2(mul(div(lr, oneMinusAccBeta1), div(newFirstMoment, add2(newWeightedInfNorm, this.epsilon))), value);\n        value.assign(newValue);\n      });\n      this.iteration.assign(add2(this.iteration, 1));\n      this.accBeta1.assign(mul(this.accBeta1, this.beta1));\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    this.accBeta1.dispose();\n    this.iteration.dispose();\n    if (this.accumulatedFirstMoment != null) {\n      dispose(this.accumulatedFirstMoment.map((v) => v.variable));\n    }\n    if (this.accumulatedWeightedInfNorm != null) {\n      dispose(this.accumulatedWeightedInfNorm.map((v) => v.variable));\n    }\n  }\n  async getWeights() {\n    throw new Error(\"getWeights() is not implemented for Adamax yet.\");\n  }\n  async setWeights(weightValues) {\n    throw new Error(\"setWeights() is not implemented for Adamax yet.\");\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"beta1\": this.beta1,\n      \"beta2\": this.beta2,\n      \"epsilon\": this.epsilon,\n      \"decay\": this.decay\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"beta1\"], config[\"beta2\"], config[\"epsilon\"], config[\"decay\"]);\n  }\n};\nAdamaxOptimizer.className = \"Adamax\";\nregisterClass(AdamaxOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/sgd_optimizer.js\nvar SGDOptimizer = class extends Optimizer {\n  constructor(learningRate) {\n    super();\n    this.learningRate = learningRate;\n    this.setLearningRate(learningRate);\n  }\n  applyGradients(variableGradients) {\n    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);\n    varNames.forEach((name, i) => {\n      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n      if (gradient == null) {\n        return;\n      }\n      const value = ENGINE.registeredVariables[name];\n      tidy(() => {\n        const newValue = add2(mul(this.c, gradient), value);\n        value.assign(newValue);\n      });\n    });\n    this.incrementIterations();\n  }\n  setLearningRate(learningRate) {\n    this.learningRate = learningRate;\n    if (this.c != null) {\n      this.c.dispose();\n    }\n    this.c = keep(scalar(-learningRate));\n  }\n  dispose() {\n    this.c.dispose();\n  }\n  async getWeights() {\n    return [await this.saveIterations()];\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    if (weightValues.length !== 0) {\n      throw new Error(\"SGD optimizer does not have settable weights.\");\n    }\n  }\n  getConfig() {\n    return { \"learningRate\": this.learningRate };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"]);\n  }\n};\nSGDOptimizer.className = \"SGD\";\nregisterClass(SGDOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/momentum_optimizer.js\nvar MomentumOptimizer = class extends SGDOptimizer {\n  constructor(learningRate, momentum, useNesterov = false) {\n    super(learningRate);\n    this.learningRate = learningRate;\n    this.momentum = momentum;\n    this.useNesterov = useNesterov;\n    this.accumulations = [];\n    this.m = scalar(this.momentum);\n  }\n  applyGradients(variableGradients) {\n    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);\n    variableNames.forEach((name, i) => {\n      const value = ENGINE.registeredVariables[name];\n      if (this.accumulations[i] == null) {\n        const trainable = false;\n        this.accumulations[i] = {\n          originalName: `${name}/momentum`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      const accumulation = this.accumulations[i].variable;\n      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n      if (gradient == null) {\n        return;\n      }\n      tidy(() => {\n        let newValue;\n        const newAccumulation = add2(mul(this.m, accumulation), gradient);\n        if (this.useNesterov) {\n          newValue = add2(mul(this.c, add2(gradient, mul(newAccumulation, this.m))), value);\n        } else {\n          newValue = add2(mul(this.c, newAccumulation), value);\n        }\n        accumulation.assign(newAccumulation);\n        value.assign(newValue);\n      });\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    this.m.dispose();\n    if (this.accumulations != null) {\n      dispose(this.accumulations.map((v) => v.variable));\n    }\n  }\n  setMomentum(momentum) {\n    this.momentum = momentum;\n  }\n  async getWeights() {\n    return [await this.saveIterations()].concat(this.accumulations.map((v) => ({ name: v.originalName, tensor: v.variable })));\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    const trainable = false;\n    this.accumulations = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"momentum\": this.momentum,\n      \"useNesterov\": this.useNesterov\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"momentum\"], config[\"useNesterov\"]);\n  }\n};\nMomentumOptimizer.className = \"Momentum\";\nregisterClass(MomentumOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/rmsprop_optimizer.js\nvar RMSPropOptimizer = class extends Optimizer {\n  constructor(learningRate, decay = 0.9, momentum = 0, epsilon3 = null, centered = false) {\n    super();\n    this.learningRate = learningRate;\n    this.decay = decay;\n    this.momentum = momentum;\n    this.epsilon = epsilon3;\n    this.accumulatedMeanSquares = [];\n    this.accumulatedMoments = [];\n    this.accumulatedMeanGrads = [];\n    this.centered = centered;\n    if (epsilon3 == null) {\n      this.epsilon = ENGINE.backend.epsilon();\n    }\n    if (learningRate == null) {\n      throw new Error(`learningRate for RMSPropOptimizer must be defined.`);\n    }\n  }\n  applyGradients(variableGradients) {\n    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);\n    variableNames.forEach((name, i) => {\n      const value = ENGINE.registeredVariables[name];\n      const trainable = false;\n      if (this.accumulatedMeanSquares[i] == null) {\n        this.accumulatedMeanSquares[i] = {\n          originalName: `${name}/rms`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      if (this.accumulatedMoments[i] == null) {\n        this.accumulatedMoments[i] = {\n          originalName: `${name}/momentum`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      if (this.accumulatedMeanGrads[i] == null && this.centered) {\n        this.accumulatedMeanGrads[i] = {\n          originalName: `${name}/mg`,\n          variable: tidy(() => zerosLike(value).variable(trainable))\n        };\n      }\n      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];\n      if (gradient == null) {\n        return;\n      }\n      const accumulatedMeanSquare = this.accumulatedMeanSquares[i].variable;\n      const accumulatedMoments = this.accumulatedMoments[i].variable;\n      tidy(() => {\n        const newAccumulatedMeanSquare = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));\n        if (this.centered) {\n          const accumulatedMeanGrad = this.accumulatedMeanGrads[i].variable;\n          const newAccumulatedMeanGrad = add2(mul(accumulatedMeanGrad, this.decay), mul(gradient, 1 - this.decay));\n          const gradContribution = div(mul(gradient, this.learningRate), sqrt(sub(newAccumulatedMeanSquare, add2(square(newAccumulatedMeanGrad), this.epsilon))));\n          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), gradContribution);\n          accumulatedMeanSquare.assign(newAccumulatedMeanSquare);\n          accumulatedMeanGrad.assign(newAccumulatedMeanGrad);\n          accumulatedMoments.assign(newAccumulatedMoments);\n          const newValue = sub(value, newAccumulatedMoments);\n          value.assign(newValue);\n        } else {\n          const newAccumulatedMeanSquare2 = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));\n          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), div(mul(gradient, this.learningRate), sqrt(add2(newAccumulatedMeanSquare2, this.epsilon))));\n          accumulatedMeanSquare.assign(newAccumulatedMeanSquare2);\n          accumulatedMoments.assign(newAccumulatedMoments);\n          const newValue = sub(value, newAccumulatedMoments);\n          value.assign(newValue);\n        }\n      });\n    });\n    this.incrementIterations();\n  }\n  dispose() {\n    if (this.accumulatedMeanSquares != null) {\n      dispose(this.accumulatedMeanSquares.map((v) => v.variable));\n    }\n    if (this.accumulatedMeanGrads != null && this.centered) {\n      dispose(this.accumulatedMeanGrads.map((v) => v.variable));\n    }\n    if (this.accumulatedMoments != null) {\n      dispose(this.accumulatedMoments.map((v) => v.variable));\n    }\n  }\n  async getWeights() {\n    const variables = [...this.accumulatedMeanSquares, ...this.accumulatedMoments];\n    if (this.centered) {\n      variables.push(...this.accumulatedMeanGrads);\n    }\n    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));\n  }\n  async setWeights(weightValues) {\n    weightValues = await this.extractIterations(weightValues);\n    const variableCount = this.centered ? weightValues.length / 3 : weightValues.length / 2;\n    const trainable = false;\n    this.accumulatedMeanSquares = weightValues.slice(0, variableCount).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n    this.accumulatedMoments = weightValues.slice(variableCount, variableCount * 2).map((v) => ({\n      originalName: v.name,\n      variable: v.tensor.variable(trainable)\n    }));\n    if (this.centered) {\n      this.accumulatedMeanGrads = weightValues.slice(variableCount * 2, variableCount * 3).map((v) => ({\n        originalName: v.name,\n        variable: v.tensor.variable(trainable)\n      }));\n    }\n  }\n  getConfig() {\n    return {\n      \"learningRate\": this.learningRate,\n      \"decay\": this.decay,\n      \"momentum\": this.momentum,\n      \"epsilon\": this.epsilon,\n      \"centered\": this.centered\n    };\n  }\n  static fromConfig(cls, config) {\n    return new cls(config[\"learningRate\"], config[\"decay\"], config[\"momentum\"], config[\"epsilon\"], config[\"centered\"]);\n  }\n};\nRMSPropOptimizer.className = \"RMSProp\";\nregisterClass(RMSPropOptimizer);\n\n// node_modules/@tensorflow/tfjs-core/dist/optimizers/optimizer_constructors.js\nvar OptimizerConstructors = class {\n  static sgd(learningRate) {\n    return new SGDOptimizer(learningRate);\n  }\n  static momentum(learningRate, momentum, useNesterov = false) {\n    return new MomentumOptimizer(learningRate, momentum, useNesterov);\n  }\n  static rmsprop(learningRate, decay = 0.9, momentum = 0, epsilon3 = null, centered = false) {\n    return new RMSPropOptimizer(learningRate, decay, momentum, epsilon3, centered);\n  }\n  static adam(learningRate = 1e-3, beta1 = 0.9, beta2 = 0.999, epsilon3 = null) {\n    return new AdamOptimizer(learningRate, beta1, beta2, epsilon3);\n  }\n  static adadelta(learningRate = 1e-3, rho = 0.95, epsilon3 = null) {\n    return new AdadeltaOptimizer(learningRate, rho, epsilon3);\n  }\n  static adamax(learningRate = 2e-3, beta1 = 0.9, beta2 = 0.999, epsilon3 = null, decay = 0) {\n    return new AdamaxOptimizer(learningRate, beta1, beta2, epsilon3, decay);\n  }\n  static adagrad(learningRate, initialAccumulatorValue = 0.1) {\n    return new AdagradOptimizer(learningRate, initialAccumulatorValue);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/train.js\nvar train = {\n  sgd: OptimizerConstructors.sgd,\n  momentum: OptimizerConstructors.momentum,\n  adadelta: OptimizerConstructors.adadelta,\n  adagrad: OptimizerConstructors.adagrad,\n  rmsprop: OptimizerConstructors.rmsprop,\n  adamax: OptimizerConstructors.adamax,\n  adam: OptimizerConstructors.adam\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/browser_util.js\nvar delayCallback = (() => {\n  if (typeof requestAnimationFrame !== \"undefined\") {\n    return requestAnimationFrame;\n  } else if (typeof setImmediate !== \"undefined\") {\n    return setImmediate;\n  }\n  return (f) => f();\n})();\nfunction nextFrame() {\n  return new Promise((resolve) => delayCallback(() => resolve()));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/backend_util.js\nvar backend_util_exports = {};\n__export(backend_util_exports, {\n  ERF_A1: () => ERF_A1,\n  ERF_A2: () => ERF_A2,\n  ERF_A3: () => ERF_A3,\n  ERF_A4: () => ERF_A4,\n  ERF_A5: () => ERF_A5,\n  ERF_P: () => ERF_P,\n  PARALLELIZE_THRESHOLD: () => PARALLELIZE_THRESHOLD,\n  RowPartitionType: () => RowPartitionType,\n  SELU_SCALE: () => SELU_SCALE,\n  SELU_SCALEALPHA: () => SELU_SCALEALPHA,\n  applyActivation: () => applyActivation,\n  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,\n  assertAxesAreInnerMostDims: () => assertAxesAreInnerMostDims,\n  assertParamsConsistent: () => assertParamsConsistent,\n  assignToTypedArray: () => assignToTypedArray,\n  axesAreInnerMostDims: () => axesAreInnerMostDims,\n  calculateShapes: () => calculateShapes,\n  checkEinsumDimSizes: () => checkEinsumDimSizes,\n  checkPadOnDimRoundingMode: () => checkPadOnDimRoundingMode,\n  combineLocations: () => combineLocations,\n  combineRaggedTensorToTensorShapes: () => combineRaggedTensorToTensorShapes,\n  complexWithEvenIndex: () => complexWithEvenIndex,\n  complexWithOddIndex: () => complexWithOddIndex,\n  computeConv2DInfo: () => computeConv2DInfo,\n  computeConv3DInfo: () => computeConv3DInfo,\n  computeDefaultPad: () => computeDefaultPad,\n  computeDilation2DInfo: () => computeDilation2DInfo,\n  computeOptimalWindowSize: () => computeOptimalWindowSize,\n  computeOutAndReduceShapes: () => computeOutAndReduceShapes,\n  computeOutShape: () => computeOutShape2,\n  computePool2DInfo: () => computePool2DInfo,\n  computePool3DInfo: () => computePool3DInfo,\n  convertConv2DDataFormat: () => convertConv2DDataFormat,\n  decodeEinsumEquation: () => decodeEinsumEquation,\n  eitherStridesOrDilationsAreOne: () => eitherStridesOrDilationsAreOne,\n  expandShapeToKeepDim: () => expandShapeToKeepDim,\n  exponent: () => exponent,\n  exponents: () => exponents,\n  fromStringArrayToUint8: () => fromStringArrayToUint8,\n  fromUint8ToStringArray: () => fromUint8ToStringArray,\n  getAxesPermutation: () => getAxesPermutation,\n  getBroadcastDims: () => getBroadcastDims,\n  getComplexWithIndex: () => getComplexWithIndex,\n  getEinsumComputePath: () => getEinsumComputePath,\n  getEinsumPermutation: () => getEinsumPermutation,\n  getFusedBiasGradient: () => getFusedBiasGradient,\n  getFusedDyActivation: () => getFusedDyActivation,\n  getImageCenter: () => getImageCenter,\n  getInnerMostAxes: () => getInnerMostAxes,\n  getPermuted: () => getPermuted,\n  getRaggedRank: () => getRaggedRank,\n  getReductionAxes: () => getReductionAxes,\n  getReshaped: () => getReshaped,\n  getReshapedPermuted: () => getReshapedPermuted,\n  getRowPartitionTypesHelper: () => getRowPartitionTypesHelper,\n  getSliceBeginCoords: () => getSliceBeginCoords,\n  getSliceSize: () => getSliceSize,\n  getSparseFillEmptyRowsIndicesDenseShapeMismatch: () => getSparseFillEmptyRowsIndicesDenseShapeMismatch,\n  getSparseFillEmptyRowsNegativeIndexErrorMessage: () => getSparseFillEmptyRowsNegativeIndexErrorMessage,\n  getSparseFillEmptyRowsOutOfRangeIndexErrorMessage: () => getSparseFillEmptyRowsOutOfRangeIndexErrorMessage,\n  getSparseReshapeEmptyTensorZeroOutputDimErrorMessage: () => getSparseReshapeEmptyTensorZeroOutputDimErrorMessage,\n  getSparseReshapeInputOutputMismatchErrorMessage: () => getSparseReshapeInputOutputMismatchErrorMessage,\n  getSparseReshapeInputOutputMultipleErrorMessage: () => getSparseReshapeInputOutputMultipleErrorMessage,\n  getSparseReshapeMultipleNegativeOneOutputDimErrorMessage: () => getSparseReshapeMultipleNegativeOneOutputDimErrorMessage,\n  getSparseReshapeNegativeOutputDimErrorMessage: () => getSparseReshapeNegativeOutputDimErrorMessage,\n  getSparseSegmentReductionIndicesOutOfRangeErrorMessage: () => getSparseSegmentReductionIndicesOutOfRangeErrorMessage,\n  getSparseSegmentReductionNegativeSegmentIdsErrorMessage: () => getSparseSegmentReductionNegativeSegmentIdsErrorMessage,\n  getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage: () => getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage,\n  getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage: () => getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage,\n  getUndoAxesPermutation: () => getUndoAxesPermutation,\n  isIdentityPermutation: () => isIdentityPermutation,\n  log: () => log,\n  mergeRealAndImagArrays: () => mergeRealAndImagArrays,\n  prepareAndValidate: () => prepareAndValidate,\n  prepareSplitSize: () => prepareSplitSize,\n  segment_util: () => segment_util_exports,\n  shouldFuse: () => shouldFuse,\n  slice_util: () => slice_util_exports,\n  splitRealAndImagArrays: () => splitRealAndImagArrays,\n  tupleValuesAreOne: () => tupleValuesAreOne,\n  upcastType: () => upcastType,\n  validateDefaultValueShape: () => validateDefaultValueShape,\n  validateInput: () => validateInput,\n  validateUpdateShape: () => validateUpdateShape,\n  warn: () => warn\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/concat_util.js\nfunction assertParamsConsistent(shapes, axis) {\n  const rank = shapes[0].length;\n  shapes.forEach((shape, i) => {\n    assert(shape.length === rank, () => `Error in concat${rank}D: rank of tensors[${i}] must be the same as the rank of the rest (${rank})`);\n  });\n  assert(axis >= 0 && axis < rank, () => `Error in concat${rank}D: axis must be between 0 and ${rank - 1}.`);\n  const firstShape = shapes[0];\n  shapes.forEach((shape, i) => {\n    for (let r = 0; r < rank; r++) {\n      assert(r === axis || shape[r] === firstShape[r], () => `Error in concat${rank}D: Shape of tensors[${i}] (${shape}) does not match the shape of the rest (${firstShape}) along the non-concatenated axis ${i}.`);\n    }\n  });\n}\nfunction computeOutShape2(shapes, axis) {\n  const outputShape = shapes[0].slice();\n  for (let i = 1; i < shapes.length; i++) {\n    outputShape[axis] += shapes[i][axis];\n  }\n  return outputShape;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_to_dense_util.js\nvar RowPartitionType;\n(function(RowPartitionType3) {\n  RowPartitionType3[RowPartitionType3[\"FIRST_DIM_SIZE\"] = 0] = \"FIRST_DIM_SIZE\";\n  RowPartitionType3[RowPartitionType3[\"VALUE_ROWIDS\"] = 1] = \"VALUE_ROWIDS\";\n  RowPartitionType3[RowPartitionType3[\"ROW_LENGTHS\"] = 2] = \"ROW_LENGTHS\";\n  RowPartitionType3[RowPartitionType3[\"ROW_SPLITS\"] = 3] = \"ROW_SPLITS\";\n  RowPartitionType3[RowPartitionType3[\"ROW_LIMITS\"] = 4] = \"ROW_LIMITS\";\n  RowPartitionType3[RowPartitionType3[\"ROW_STARTS\"] = 5] = \"ROW_STARTS\";\n})(RowPartitionType || (RowPartitionType = {}));\nfunction combineRaggedTensorToTensorShapes(raggedRank, shape, valueShape) {\n  let outputShape = new Array();\n  if (valueShape == null && shape == null) {\n    return outputShape;\n  }\n  if (shape == null) {\n    while (outputShape.length < raggedRank + valueShape.length) {\n      outputShape.push(-1);\n    }\n  } else {\n    outputShape = shape.slice();\n  }\n  if (valueShape == null) {\n    return outputShape;\n  }\n  if (raggedRank + valueShape.length !== outputShape.length) {\n    throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.rank = ${raggedRank + valueShape.length}, but shape.rank = ${outputShape.length}`);\n  }\n  for (let i = 1; i < valueShape.length; ++i) {\n    const valueDim = valueShape[i];\n    const outputShapeDimIndex = outputShape[outputShape.length - valueShape.length + i];\n    const outputShapeDim = outputShape[outputShapeDimIndex];\n    if (valueDim >= 0) {\n      if (outputShapeDim >= 0) {\n        if (outputShapeDim !== valueDim) {\n          throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.shape[${i + raggedRank}] = ${valueDim} but shape[${i + raggedRank}] = ${outputShapeDim}`);\n        }\n      } else {\n        outputShape[outputShapeDimIndex] = valueDim;\n      }\n    }\n  }\n  return outputShape;\n}\nfunction getRowPartitionTypesHelper(rowPartitionTypeStrings) {\n  const stringToType = {\n    \"FIRST_DIM_SIZE\": RowPartitionType.FIRST_DIM_SIZE,\n    \"VALUE_ROWIDS\": RowPartitionType.VALUE_ROWIDS,\n    \"ROW_LENGTHS\": RowPartitionType.ROW_LENGTHS,\n    \"ROW_SPLITS\": RowPartitionType.ROW_SPLITS,\n    \"ROW_LIMITS\": RowPartitionType.ROW_LIMITS,\n    \"ROW_STARTS\": RowPartitionType.ROW_STARTS\n  };\n  const result = [];\n  for (const typeStr of rowPartitionTypeStrings) {\n    if (typeStr in stringToType) {\n      result.push(stringToType[typeStr]);\n    } else {\n      break;\n    }\n  }\n  return result;\n}\nfunction getRaggedRank(rowPartitionTypes) {\n  if (rowPartitionTypes.length === 0) {\n    return 0;\n  }\n  if (rowPartitionTypes[0] === RowPartitionType.FIRST_DIM_SIZE) {\n    return rowPartitionTypes.length - 1;\n  }\n  return rowPartitionTypes.length;\n}\nfunction validateDefaultValueShape(defaultValueShape, valueShape) {\n  if (defaultValueShape == null || valueShape == null) {\n    return;\n  }\n  const defaultNDims = defaultValueShape.length;\n  const valuesNDims = valueShape.length;\n  if (defaultNDims >= valuesNDims) {\n    throw new Error(`defaultValue.shape=${defaultValueShape} and ragged tensor flatValues.shape=${valueShape}, are incompatible: defaultValue.rank = ${defaultNDims} must be less than ragged tensor input flatValues.rank = ${valuesNDims})`);\n  }\n  for (let i = 0; i < Math.min(defaultNDims, valuesNDims - 1); ++i) {\n    const defaultDim = defaultValueShape[i];\n    const valueDim = valueShape[i + 1];\n    if (defaultDim >= 0 && valueDim >= 0 && defaultDim !== 1 && defaultDim !== valueDim) {\n      throw new Error(`defaultValue.shape=${defaultValueShape}, and ragged tensor input flatValues.shape=${valueShape} are incompatible: defaultValue.shape[${i - defaultValueShape.length}] = ${defaultDim} but ragged tensor input.flatValues.shape[${i - defaultValueShape.length}] = ${valueDim}`);\n    }\n  }\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/reduce_util.js\nvar PARALLELIZE_THRESHOLD = 30;\nfunction computeOptimalWindowSize(inSize) {\n  if (inSize <= PARALLELIZE_THRESHOLD) {\n    return inSize;\n  }\n  return nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/rotate_util.js\nfunction getImageCenter(center, imageHeight, imageWidth) {\n  const centerX = imageWidth * (typeof center === \"number\" ? center : center[0]);\n  const centerY = imageHeight * (typeof center === \"number\" ? center : center[1]);\n  return [centerX, centerY];\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/array_ops_util.js\nfunction getReshaped(inputShape, blockShape, prod5, batchToSpace = true) {\n  let reshaped = [];\n  if (batchToSpace) {\n    reshaped = reshaped.concat(blockShape.slice(0));\n    reshaped.push(inputShape[0] / prod5);\n    reshaped = reshaped.concat(inputShape.slice(1));\n  } else {\n    reshaped = reshaped.concat(inputShape[0]);\n    const spatialLength = blockShape.length;\n    for (let i = 0; i < spatialLength; ++i) {\n      reshaped = reshaped.concat([inputShape[i + 1] / blockShape[i], blockShape[i]]);\n    }\n    reshaped = reshaped.concat(inputShape.slice(spatialLength + 1));\n  }\n  return reshaped;\n}\nfunction getPermuted(reshapedRank, blockShapeRank, batchToSpace = true) {\n  const permuted = [];\n  if (batchToSpace) {\n    permuted.push(blockShapeRank);\n    for (let i = blockShapeRank + 1; i < reshapedRank; ++i) {\n      if (i <= 2 * blockShapeRank) {\n        permuted.push(i);\n        permuted.push(i - (blockShapeRank + 1));\n      } else {\n        permuted.push(i);\n      }\n    }\n  } else {\n    const permutedBeforeBatch = [];\n    const permutedAfterBatch = [];\n    for (let i = 1; i < reshapedRank; ++i) {\n      if (i >= blockShapeRank * 2 + 1 || i % 2 === 1) {\n        permutedAfterBatch.push(i);\n      } else {\n        permutedBeforeBatch.push(i);\n      }\n    }\n    permuted.push(...permutedBeforeBatch);\n    permuted.push(0);\n    permuted.push(...permutedAfterBatch);\n  }\n  return permuted;\n}\nfunction getReshapedPermuted(inputShape, blockShape, prod5, batchToSpace = true) {\n  const reshapedPermuted = [];\n  if (batchToSpace) {\n    reshapedPermuted.push(inputShape[0] / prod5);\n  } else {\n    reshapedPermuted.push(inputShape[0] * prod5);\n  }\n  for (let i = 1; i < inputShape.length; ++i) {\n    if (i <= blockShape.length) {\n      if (batchToSpace) {\n        reshapedPermuted.push(blockShape[i - 1] * inputShape[i]);\n      } else {\n        reshapedPermuted.push(inputShape[i] / blockShape[i - 1]);\n      }\n    } else {\n      reshapedPermuted.push(inputShape[i]);\n    }\n  }\n  return reshapedPermuted;\n}\nfunction getSliceBeginCoords(crops, blockShape) {\n  const sliceBeginCoords = [0];\n  for (let i = 0; i < blockShape; ++i) {\n    sliceBeginCoords.push(crops[i][0]);\n  }\n  return sliceBeginCoords;\n}\nfunction getSliceSize(uncroppedShape, crops, blockShape) {\n  const sliceSize = uncroppedShape.slice(0, 1);\n  for (let i = 0; i < blockShape; ++i) {\n    sliceSize.push(uncroppedShape[i + 1] - crops[i][0] - crops[i][1]);\n  }\n  return sliceSize;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/selu_util.js\nvar SELU_SCALEALPHA = 1.7580993408473768;\nvar SELU_SCALE = 1.0507009873554805;\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/erf_util.js\nvar ERF_P = 0.3275911;\nvar ERF_A1 = 0.254829592;\nvar ERF_A2 = -0.284496736;\nvar ERF_A3 = 1.421413741;\nvar ERF_A4 = -1.453152027;\nvar ERF_A5 = 1.061405429;\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/complex_util.js\nfunction mergeRealAndImagArrays(real4, imag4) {\n  if (real4.length !== imag4.length) {\n    throw new Error(`Cannot merge real and imag arrays of different lengths. real:${real4.length}, imag: ${imag4.length}.`);\n  }\n  const result = new Float32Array(real4.length * 2);\n  for (let i = 0; i < result.length; i += 2) {\n    result[i] = real4[i / 2];\n    result[i + 1] = imag4[i / 2];\n  }\n  return result;\n}\nfunction splitRealAndImagArrays(complex4) {\n  const real4 = new Float32Array(complex4.length / 2);\n  const imag4 = new Float32Array(complex4.length / 2);\n  for (let i = 0; i < complex4.length; i += 2) {\n    real4[i / 2] = complex4[i];\n    imag4[i / 2] = complex4[i + 1];\n  }\n  return { real: real4, imag: imag4 };\n}\nfunction complexWithEvenIndex(complex4) {\n  const len = Math.ceil(complex4.length / 4);\n  const real4 = new Float32Array(len);\n  const imag4 = new Float32Array(len);\n  for (let i = 0; i < complex4.length; i += 4) {\n    real4[Math.floor(i / 4)] = complex4[i];\n    imag4[Math.floor(i / 4)] = complex4[i + 1];\n  }\n  return { real: real4, imag: imag4 };\n}\nfunction complexWithOddIndex(complex4) {\n  const len = Math.floor(complex4.length / 4);\n  const real4 = new Float32Array(len);\n  const imag4 = new Float32Array(len);\n  for (let i = 2; i < complex4.length; i += 4) {\n    real4[Math.floor(i / 4)] = complex4[i];\n    imag4[Math.floor(i / 4)] = complex4[i + 1];\n  }\n  return { real: real4, imag: imag4 };\n}\nfunction getComplexWithIndex(complex4, index) {\n  const real4 = complex4[index * 2];\n  const imag4 = complex4[index * 2 + 1];\n  return { real: real4, imag: imag4 };\n}\nfunction assignToTypedArray(data, real4, imag4, index) {\n  data[index * 2] = real4;\n  data[index * 2 + 1] = imag4;\n}\nfunction exponents(n, inverse) {\n  const real4 = new Float32Array(n / 2);\n  const imag4 = new Float32Array(n / 2);\n  for (let i = 0; i < Math.ceil(n / 2); i++) {\n    const x = (inverse ? 2 : -2) * Math.PI * (i / n);\n    real4[i] = Math.cos(x);\n    imag4[i] = Math.sin(x);\n  }\n  return { real: real4, imag: imag4 };\n}\nfunction exponent(k, n, inverse) {\n  const x = (inverse ? 2 : -2) * Math.PI * (k / n);\n  const real4 = Math.cos(x);\n  const imag4 = Math.sin(x);\n  return { real: real4, imag: imag4 };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/einsum_util.js\nvar ARROW = \"->\";\nvar ARROW_REGEX = /->/g;\nvar COMMA = \",\";\nvar ELLIPSIS = \"...\";\nfunction decodeEinsumEquation(equation, numTensors) {\n  equation = equation.replace(/\\s/g, \"\");\n  const numArrows = (equation.length - equation.replace(ARROW_REGEX, \"\").length) / ARROW.length;\n  if (numArrows < 1) {\n    throw new Error(\"Equations without an arrow are not supported.\");\n  } else if (numArrows > 1) {\n    throw new Error(`Equation must contain exactly one arrow (\"${ARROW}\").`);\n  }\n  const [inputString, outputString] = equation.split(ARROW);\n  assert(inputString.indexOf(ELLIPSIS) === -1, () => `The ellipsis notation (\"${ELLIPSIS}\") is not supported yet.`);\n  const inputTerms = inputString.split(COMMA);\n  const numInputs = inputTerms.length;\n  if (numTensors !== numInputs) {\n    throw new Error(`Expected ${numInputs} input tensors, received ${numTensors}`);\n  }\n  if (numInputs > 2) {\n    throw new Error(\"Support for more than 2 input tensors is not implemented yet.\");\n  }\n  const allDims = [];\n  for (let i = 0; i < outputString.length; ++i) {\n    const dimName = outputString[i];\n    if (!inputTerms.some((inputTerm) => inputTerm.indexOf(dimName) !== -1)) {\n      throw new Error(`Output subscripts contain the label ${dimName} not present in the input subscripts.`);\n    }\n    if (allDims.indexOf(dimName) === -1) {\n      allDims.push(dimName);\n    }\n  }\n  for (let i = 0; i < inputString.length; ++i) {\n    const dimName = inputString[i];\n    if (allDims.indexOf(dimName) === -1 && dimName !== COMMA) {\n      allDims.push(dimName);\n    }\n  }\n  const idDims = new Array(inputTerms.length);\n  for (let i = 0; i < numInputs; ++i) {\n    if (new Set(inputTerms[i].split(\"\")).size !== inputTerms[i].length) {\n      throw new Error(`Found duplicate axes in input component ${inputTerms[i]}. Support for duplicate axes in input is not implemented yet.`);\n    }\n    idDims[i] = [];\n    for (let j = 0; j < inputTerms[i].length; ++j) {\n      idDims[i].push(allDims.indexOf(inputTerms[i][j]));\n    }\n  }\n  const numDims = allDims.length;\n  const numOutDims = outputString.length;\n  const summedDims = [];\n  for (let i = numOutDims; i < numDims; ++i) {\n    summedDims.push(i);\n  }\n  return { allDims, summedDims, idDims };\n}\nfunction getEinsumPermutation(nDims, idDims) {\n  let permutationIndices = new Array(nDims);\n  permutationIndices.fill(-1);\n  for (let i = 0; i < idDims.length; ++i) {\n    permutationIndices[idDims[i]] = i;\n  }\n  const expandDims6 = [];\n  for (let i = 0; i < nDims; ++i) {\n    if (permutationIndices[i] === -1) {\n      expandDims6.push(i);\n    }\n  }\n  permutationIndices = permutationIndices.filter((d) => d !== -1);\n  return { permutationIndices, expandDims: expandDims6 };\n}\nfunction checkEinsumDimSizes(nDims, idDims, tensors) {\n  const dimSizes = new Array(nDims);\n  for (let i = 0; i < tensors.length; ++i) {\n    const shape = tensors[i].shape;\n    for (let j = 0; j < idDims[i].length; ++j) {\n      if (dimSizes[idDims[i][j]] === void 0) {\n        dimSizes[idDims[i][j]] = shape[j];\n      } else {\n        assert(dimSizes[idDims[i][j]] === shape[j], () => `Expected dimension ${dimSizes[idDims[i][j]]} at axis ${j} of input shaped ${JSON.stringify(shape)}, but got dimension ${shape[j]}`);\n      }\n    }\n  }\n}\nfunction getEinsumComputePath(summedDims, idDims) {\n  const path = summedDims;\n  const steps = [];\n  let nSteps = 0;\n  if (summedDims.length === 0) {\n    path.push(-1);\n  }\n  nSteps = summedDims.length + 1;\n  for (let i = 0; i < nSteps; ++i) {\n    steps.push([]);\n  }\n  const computedTermIndices = [];\n  for (let i = 0; i < path.length; ++i) {\n    const summedDim = path[i];\n    const termIndices = findTermsWithDim(idDims, summedDim);\n    for (const termIndex of termIndices) {\n      if (computedTermIndices.indexOf(termIndex) === -1) {\n        steps[i].push(termIndex);\n        computedTermIndices.push(termIndex);\n      }\n    }\n  }\n  return { path, steps };\n}\nfunction isIdentityPermutation(perm) {\n  return perm.every((dim, index) => dim === index);\n}\nfunction findTermsWithDim(idDims, dim) {\n  const termIndices = [];\n  for (let i = 0; i < idDims.length; ++i) {\n    if (idDims[i].length === 0 || idDims[i].indexOf(dim) !== -1 || dim === -1) {\n      termIndices.push(i);\n    }\n  }\n  return termIndices;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/split_util.js\nfunction prepareSplitSize(x, numOrSizeSplits, axis = 0) {\n  let splitSizes = [];\n  if (typeof numOrSizeSplits === \"number\") {\n    assert(x.shape[axis] % numOrSizeSplits === 0, () => \"Number of splits must evenly divide the axis.\");\n    splitSizes = new Array(numOrSizeSplits).fill(x.shape[axis] / numOrSizeSplits);\n  } else {\n    const numOfNegs = numOrSizeSplits.reduce((count2, value) => {\n      if (value === -1) {\n        count2 += 1;\n      }\n      return count2;\n    }, 0);\n    assert(numOfNegs <= 1, () => \"There should be only one negative value in split array.\");\n    const negIndex = numOrSizeSplits.indexOf(-1);\n    if (negIndex !== -1) {\n      const total = numOrSizeSplits.reduce((a, b) => b > 0 ? a + b : a);\n      numOrSizeSplits[negIndex] = x.shape[axis] - total;\n    }\n    assert(x.shape[axis] === numOrSizeSplits.reduce((a, b) => a + b), () => \"The sum of sizes must match the size of the axis dimension.\");\n    splitSizes = numOrSizeSplits;\n  }\n  return splitSizes;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_fill_empty_rows_util.js\nfunction getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesLength) {\n  return `Received SparseTensor with denseShape[0] = 0 but\n  indices.shape[0] = ${indicesLength}`;\n}\nfunction getSparseFillEmptyRowsNegativeIndexErrorMessage(index, value) {\n  return `indices(${index}, 0) is invalid: ${value} < 0`;\n}\nfunction getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(index, value, limit) {\n  return `indices(${index}, 0) is invalid: ${value} >= ${limit}`;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_reshape_util.js\nfunction getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(dim1, dim2) {\n  return `only one output dimension may be -1, not both ${dim1} and ${dim2}`;\n}\nfunction getSparseReshapeNegativeOutputDimErrorMessage(dim, value) {\n  return `size ${dim} must be non-negative, not ${value}`;\n}\nfunction getSparseReshapeEmptyTensorZeroOutputDimErrorMessage() {\n  return \"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero\";\n}\nfunction getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape) {\n  const inputSize = sizeFromShape(inputShape);\n  const outputSize = sizeFromShape(outputShape);\n  return `Input to reshape is a SparseTensor with ${inputSize}\n  dense values, but the requested shape requires a multiple of ${outputSize}. inputShape=${inputShape} outputShape= ${outputShape}`;\n}\nfunction getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape) {\n  const inputSize = sizeFromShape(inputShape);\n  const outputSize = sizeFromShape(outputShape);\n  return `Input to reshape is a tensor with ${inputSize} dense values, but the requested shape has ${outputSize}. inputShape=${inputShape} outputShape=${outputShape}`;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_reduction_util.js\nfunction getSparseSegmentReductionNegativeSegmentIdsErrorMessage() {\n  return `segment ids must be >= 0`;\n}\nfunction getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage() {\n  return `segment ids are not increasing`;\n}\nfunction getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(segmentId, outputRows) {\n  return `Segment id ${segmentId} out of range [0, ${outputRows}), possibly because segmentIds input is not sorted.`;\n}\nfunction getSparseSegmentReductionIndicesOutOfRangeErrorMessage(index, indexValue, inputRows) {\n  return `Bad: indices[${index}] == ${indexValue} out of range [0, ${inputRows})`;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/segment_util.js\nvar segment_util_exports = {};\n__export(segment_util_exports, {\n  collectGatherOpShapeInfo: () => collectGatherOpShapeInfo,\n  computeOutShape: () => computeOutShape3,\n  segOpComputeOptimalWindowSize: () => segOpComputeOptimalWindowSize\n});\nfunction segOpComputeOptimalWindowSize(inSize, numSegments) {\n  let done = false;\n  let res;\n  if (inSize <= PARALLELIZE_THRESHOLD) {\n    res = inSize;\n    done = true;\n  } else {\n    res = nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));\n  }\n  while (!done) {\n    if (res > numSegments || res === inSize) {\n      done = true;\n    } else {\n      res = nearestDivisor(inSize, res + 1);\n    }\n  }\n  return res;\n}\nfunction computeOutShape3(aShape, axis, numSegments) {\n  const outShape = [];\n  const rank = aShape.length;\n  for (let dim = 0; dim < rank; dim++) {\n    if (dim !== axis) {\n      outShape.push(aShape[dim]);\n    } else {\n      outShape.push(numSegments);\n    }\n  }\n  return outShape;\n}\nfunction collectGatherOpShapeInfo(x, indices, axis, batchDims) {\n  const indicesRank = indices.shape.length;\n  const xRank = x.shape.length;\n  if (batchDims !== 0) {\n    if (batchDims < -indicesRank || batchDims > indicesRank) {\n      throw new Error(`Expect batchDims in the range of [-${indicesRank}, ${indicesRank}], but got ${batchDims}`);\n    }\n  }\n  if (batchDims < 0) {\n    batchDims += indicesRank;\n  }\n  if (batchDims > xRank) {\n    throw new Error(`batchDims (${batchDims}) must be less than rank(x) (\n    ${xRank}).`);\n  }\n  if (axis < batchDims) {\n    throw new Error(`batchDims (${batchDims}) must be less than or equal to axis (${axis}).`);\n  }\n  for (let i = 0; i < batchDims; ++i) {\n    if (x.shape[i] !== indices.shape[i]) {\n      throw new Error(`x.shape[${i}]: ${x.shape[i]} should be equal to indices.shape[${i}]: ${indices.shape[i]}.`);\n    }\n  }\n  const dimSize = x.shape[axis];\n  const outputShape = [];\n  let batchSize = 1;\n  let outerSize = 1;\n  let sliceSize = 1;\n  for (let i = 0; i < batchDims; ++i) {\n    outputShape.push(x.shape[i]);\n    batchSize *= x.shape[i];\n  }\n  for (let i = batchDims; i < axis; i++) {\n    outputShape.push(x.shape[i]);\n    outerSize *= x.shape[i];\n  }\n  for (let i = batchDims; i < indicesRank; i++) {\n    outputShape.push(indices.shape[i]);\n  }\n  for (let i = axis + 1; i < xRank; i++) {\n    outputShape.push(x.shape[i]);\n    sliceSize *= x.shape[i];\n  }\n  return { batchSize, sliceSize, outerSize, dimSize, outputShape };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/backend_util.js\nfunction fromUint8ToStringArray(vals) {\n  try {\n    return vals.map((val) => decodeString(val));\n  } catch (err) {\n    throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${err}`);\n  }\n}\nfunction fromStringArrayToUint8(strings) {\n  return strings.map((s) => encodeString(s));\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/backends/kernel_impls.js\nvar kernel_impls_exports = {};\n__export(kernel_impls_exports, {\n  nonMaxSuppressionV3Impl: () => nonMaxSuppressionV3Impl,\n  nonMaxSuppressionV4Impl: () => nonMaxSuppressionV4Impl,\n  nonMaxSuppressionV5Impl: () => nonMaxSuppressionV5Impl,\n  whereImpl: () => whereImpl\n});\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Abs_grad.js\nvar absGradConfig = {\n  kernelName: Abs,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(dy, step(cast(x, \"float32\"), -1)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Acos_grad.js\nvar acosGradConfig = {\n  kernelName: Acos,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return {\n      x: () => {\n        const a = square(cast(x, \"float32\"));\n        const b = sqrt(sub(scalar(1), a));\n        return neg(div(dy, b));\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Acosh_grad.js\nvar acoshGradConfig = {\n  kernelName: Acosh,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return {\n      x: () => {\n        const a = sqrt(sub(square(cast(x, \"float32\")), 1));\n        return div(dy, a);\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Add_grad.js\nvar addGradConfig = {\n  kernelName: Add,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      let res = dy;\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, a.shape);\n    };\n    const derB = () => {\n      let res = dy;\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, b.shape);\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/AddN_grad.js\nvar addNGradConfig = {\n  kernelName: AddN,\n  saveAllInputs: true,\n  gradFunc: (dy, saved) => {\n    const ders = {};\n    saved.forEach((_, i) => {\n      ders[i] = () => dy.clone();\n    });\n    return ders;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ArgMax_grad.js\nvar argMaxGradConfig = {\n  kernelName: ArgMax,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => zerosLike(x) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ArgMin_grad.js\nvar argMinGradConfig = {\n  kernelName: ArgMin,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => zerosLike(x) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Asin_grad.js\nvar asinGradConfig = {\n  kernelName: Asin,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, sqrt(sub(scalar(1), square(cast(x, \"float32\"))))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Asinh_grad.js\nvar asinhGradConfig = {\n  kernelName: Asinh,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return {\n      x: () => {\n        const a = sqrt(add2(scalar(1), square(cast(x, \"float32\"))));\n        return div(dy, a);\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Atan2_grad.js\nvar atan2GradConfig = {\n  kernelName: Atan2,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      const d = add2(square(a), square(b));\n      let res = mul(dy, div(b, d));\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, a.shape);\n    };\n    const derB = () => {\n      const d = add2(square(a), square(b));\n      let res = neg(mul(dy, div(a, d)));\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, b.shape);\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Atan_grad.js\nvar atanGradConfig = {\n  kernelName: Atan,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, add2(square(cast(x, \"float32\")), 1)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Atanh_grad.js\nvar atanhGradConfig = {\n  kernelName: Atanh,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, sub(scalar(1), square(cast(x, \"float32\")))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_3d_grad.js\nfunction avgPool3dGrad_(dy, input2, filterSize, strides, pad3, dimRoundingMode) {\n  const $dy = convertToTensor(dy, \"dy\", \"avgPool3dGrad\");\n  const $input = convertToTensor(input2, \"input\", \"avgPool3dGrad\");\n  let dy5D = $dy;\n  let input5D = $input;\n  let reshapedTo5D = false;\n  if ($input.rank === 4) {\n    reshapedTo5D = true;\n    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);\n    input5D = reshape($input, [\n      1,\n      $input.shape[0],\n      $input.shape[1],\n      $input.shape[2],\n      $input.shape[3]\n    ]);\n  }\n  assert(dy5D.rank === 5, () => `Error in avgPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);\n  assert(input5D.rank === 5, () => `Error in avgPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);\n  checkPadOnDimRoundingMode(\"avgPool3dGrad\", pad3, dimRoundingMode);\n  const inputs = { dy: dy5D, input: input5D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };\n  const res = ENGINE.runKernel(AvgPool3DGrad, inputs, attrs);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar avgPool3dGrad = op({ avgPool3dGrad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/AvgPool3D_grad.js\nvar avgPool3DGradConfig = {\n  kernelName: AvgPool3D,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n    return {\n      x: () => avgPool3dGrad(dy, x, filterSize, strides, pad3, dimRoundingMode)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_grad.js\nfunction avgPoolGrad_(dy, input2, filterSize, strides, pad3) {\n  const $dy = convertToTensor(dy, \"dy\", \"avgPoolGrad\");\n  const $input = convertToTensor(input2, \"input\", \"avgPoolGrad\");\n  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);\n  let input4D = $input;\n  let dy4D = $dy;\n  let reshapedTo4D = false;\n  if ($input.rank === 3) {\n    reshapedTo4D = true;\n    input4D = reshape($input, [1, $input.shape[0], $input.shape[1], $input.shape[2]]);\n    dy4D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2]]);\n  }\n  assert(dy4D.rank === 4, () => `Error in avgPoolGrad: dy must be rank 4 but got rank ${dy4D.rank}.`);\n  assert(input4D.rank === 4, () => `Error in avgPoolGrad: input must be rank 4 but got rank ${input4D.rank}.`);\n  const inputs = { dy: dy4D, input: input4D };\n  const attrs = { filterSize, strides, pad: pad3 };\n  const res = ENGINE.runKernel(AvgPoolGrad, inputs, attrs);\n  if (reshapedTo4D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);\n  }\n  return res;\n}\nvar avgPoolGrad = op({ avgPoolGrad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/AvgPool_grad.js\nvar avgPoolGradConfig = {\n  kernelName: AvgPool,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { filterSize, strides, pad: pad3 } = attrs;\n    return { x: () => avgPoolGrad(dy, x, filterSize, strides, pad3) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/BatchMatMul_grad.js\nvar batchMatMulGradConfig = {\n  kernelName: BatchMatMul,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [a, b] = saved;\n    const { transposeA, transposeB } = attrs;\n    if (!transposeA && !transposeB) {\n      return {\n        a: () => matMul(dy, b, false, true),\n        b: () => matMul(a, dy, true, false)\n      };\n    } else if (!transposeA && transposeB) {\n      return {\n        a: () => matMul(dy, b, false, false),\n        b: () => matMul(dy, a, true, false)\n      };\n    } else if (transposeA && !transposeB) {\n      return {\n        a: () => matMul(b, dy, false, true),\n        b: () => matMul(a, dy, false, false)\n      };\n    } else {\n      return {\n        a: () => matMul(b, dy, true, true),\n        b: () => matMul(dy, a, true, true)\n      };\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/BatchToSpaceND_grad.js\nvar batchToSpaceNDGradConfig = {\n  kernelName: BatchToSpaceND,\n  gradFunc: (dy, saved, attrs) => {\n    const { blockShape, crops } = attrs;\n    return { x: () => spaceToBatchND(dy, blockShape, crops) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/BroadcastTo_grad.js\nvar broadcastToGradConfig = {\n  kernelName: BroadcastTo,\n  gradFunc: (dy, saved, attrs) => {\n    const broadCastToAttrs = attrs;\n    const inputShape = broadCastToAttrs.inputShape;\n    const outputShape = broadCastToAttrs.shape;\n    const reps = Array.from(outputShape);\n    for (let i = inputShape.length - 1; i >= 0; i--) {\n      if (inputShape[i] === outputShape[i]) {\n        reps[i] = 1;\n      } else if (inputShape[i] !== 1) {\n        throw new Error(`broadcastTo(): [${inputShape}] cannot be broadcast to [${outputShape}].`);\n      }\n    }\n    const axes = [];\n    for (let i = 0; i < reps.length; i++) {\n      if (reps[i] > 1) {\n        axes.push(i);\n      }\n    }\n    return { x: () => sum2(dy, axes, true) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Cast_grad.js\nvar castGradConfig = {\n  kernelName: Cast,\n  gradFunc: (dy) => {\n    return { x: () => dy.clone() };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Ceil_grad.js\nvar ceilGradConfig = {\n  kernelName: Ceil,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ClipByValue_grad.js\nvar clipByValueGradConfig = {\n  kernelName: ClipByValue,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { clipValueMin, clipValueMax } = attrs;\n    return {\n      x: () => where(logicalAnd(greaterEqual(x, clipValueMin), lessEqual(x, clipValueMax)), dy, zerosLike(dy))\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ComplexAbs_grad.js\nvar complexAbsGradConfig = {\n  kernelName: ComplexAbs,\n  inputsToSave: [\"x\"],\n  gradFunc: absGradConfig.gradFunc\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Concat_grad.js\nvar concatGradConfig = {\n  kernelName: Concat,\n  saveAllInputs: true,\n  gradFunc: (dy, saved, attrs) => {\n    const shapes = saved.map((t) => t.shape);\n    const { axis } = attrs;\n    const $axis = parseAxisParam(axis, saved[0].shape)[0];\n    const sizeSplits = shapes.map((s) => s[$axis]);\n    const derTensors = split(dy, sizeSplits, $axis);\n    return derTensors.map((t) => () => t);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv2D_grad.js\nvar conv2DGradConfig = {\n  kernelName: Conv2D,\n  inputsToSave: [\"x\", \"filter\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x4D, $filter] = saved;\n    const { dilations, strides, pad: pad3, dataFormat } = attrs;\n    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);\n    return {\n      x: () => conv2DBackpropInput(x4D.shape, dy, $filter, strides, pad3, dataFormat),\n      filter: () => conv2DBackpropFilter(x4D, dy, $filter.shape, strides, pad3, dataFormat)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv2DBackpropInput_grad.js\nvar conv2DBackpropInputGradConfig = {\n  kernelName: Conv2DBackpropInput,\n  inputsToSave: [\"dy\", \"filter\"],\n  gradFunc: (ddx, saved, attrs) => {\n    const [dy, filter] = saved;\n    const { strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;\n    return {\n      dy: () => conv2d(ddx, filter, strides, pad3, dataFormat, 1, dimRoundingMode),\n      filter: () => conv2DBackpropFilter(ddx, dy, filter.shape, strides, pad3, dataFormat, dimRoundingMode)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_backprop_filter.js\nfunction conv3DBackpropFilter_(x, dy, filterShape, strides, pad3) {\n  let x5D = x;\n  if (x.rank === 4) {\n    x5D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2], x.shape[3]]);\n  }\n  let dy5D = dy;\n  if (dy5D.rank === 4) {\n    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);\n  }\n  assert(x5D.rank === 5, () => `Error in conv3dDerFilter: input must be rank 5, but got shape ${x5D.shape}.`);\n  assert(dy5D.rank === 5, () => `Error in conv3dDerFilter: dy must be rank 5, but got shape ${dy5D.shape}.`);\n  assert(filterShape.length === 5, () => `Error in conv3dDerFilter: filterShape must be length 5, but got ${filterShape}.`);\n  assert(x5D.shape[4] === filterShape[3], () => `Error in conv3dDerFilter: depth of input ${x5D.shape[4]}) must match input depth in filter (${filterShape[3]}.`);\n  assert(dy5D.shape[4] === filterShape[4], () => `Error in conv3dDerFilter: depth of dy (${dy5D.shape[4]}) must match output depth for filter (${filterShape[4]}).`);\n  const inputs = { x: x5D, dy: dy5D };\n  const attrs = { strides, pad: pad3, filterShape };\n  return ENGINE.runKernel(Conv3DBackpropFilterV2, inputs, attrs);\n}\nvar conv3DBackpropFilter = op({ conv3DBackpropFilter_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv3D_grad.js\nvar conv3DGradConfig = {\n  kernelName: Conv3D,\n  inputsToSave: [\"x\", \"filter\"],\n  gradFunc: (dy, saved, attrs) => {\n    const { dilations, strides, pad: pad3 } = attrs;\n    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);\n    const [x5D, $filter] = saved;\n    return {\n      x: () => conv3DBackpropInput(x5D.shape, dy, $filter, strides, pad3),\n      filter: () => conv3DBackpropFilter(x5D, dy, $filter.shape, strides, pad3)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Cos_grad.js\nvar cosGradConfig = {\n  kernelName: Cos,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(neg(sin(cast(x, \"float32\"))), dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Cosh_grad.js\nvar coshGradConfig = {\n  kernelName: Cosh,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(sinh(cast(x, \"float32\")), dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Cumsum_grad.js\nvar cumsumGradConfig = {\n  kernelName: Cumsum,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { axis, exclusive, reverse: reverse5 } = attrs;\n    return {\n      x: () => {\n        const permutation = getAxesPermutation([axis], x.rank);\n        let out = cumsum(dy, axis, exclusive, !reverse5);\n        if (permutation != null) {\n          out = transpose(out, permutation);\n        }\n        return out;\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/DepthwiseConv2dNative_grad.js\nvar depthwiseConv2dNativeGradConfig = {\n  kernelName: DepthwiseConv2dNative,\n  inputsToSave: [\"x\", \"filter\"],\n  gradFunc: (dy, saved, attrs) => {\n    const { dilations, strides, pad: pad3, dimRoundingMode } = attrs;\n    const $dilations = dilations == null ? [1, 1] : dilations;\n    assert(tupleValuesAreOne($dilations), () => `Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${$dilations}'`);\n    const [x, filter] = saved;\n    assert(x.rank === 4, () => `Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${x.rank}.`);\n    assert(filter.rank === 4, () => `Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${filter.rank}.`);\n    assert(x.shape[3] === filter.shape[2], () => `Error in gradient of depthwiseConv2d: number of input channels (${x.shape[3]}) must match the inChannels dimension in filter ${filter.shape[2]}.`);\n    assert(eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${strides} and dilations '${$dilations}'.`);\n    checkPadOnDimRoundingMode(\"depthwiseConv2d\", pad3, dimRoundingMode);\n    return {\n      x: () => depthwiseConv2dNativeBackpropInput(x.shape, dy, filter, strides, pad3, $dilations, dimRoundingMode),\n      filter: () => depthwiseConv2dNativeBackpropFilter(x, dy, filter.shape, strides, pad3, $dilations, dimRoundingMode)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Dilation2D_grad.js\nvar dilation2dGradConfig = {\n  kernelName: Dilation2D,\n  inputsToSave: [\"x\", \"filter\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x, filter] = saved;\n    const inputInputs = { x, filter, dy };\n    const filterInputs = { x, filter, dy };\n    return {\n      x: () => ENGINE.runKernel(Dilation2DBackpropInput, inputInputs, attrs),\n      filter: () => ENGINE.runKernel(Dilation2DBackpropFilter, filterInputs, attrs)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Elu_grad.js\nvar eluGradConfig = {\n  kernelName: Elu,\n  outputsToSave: [true],\n  gradFunc: (dy, saved) => {\n    const [y] = saved;\n    const inputs = { dy, y };\n    return { x: () => ENGINE.runKernel(EluGrad, inputs) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Erf_grad.js\nvar erfGradConfig = {\n  kernelName: Erf,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    const a = mul(exp(neg(square(x))), 2 / Math.sqrt(Math.PI));\n    return { x: () => mul(dy, a) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Exp_grad.js\nvar expGradConfig = {\n  kernelName: Exp,\n  outputsToSave: [true],\n  gradFunc: (dy, saved) => {\n    const [y] = saved;\n    return { x: () => mul(dy, y) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ExpandDims_grad.js\nvar expandDimsGradConfig = {\n  kernelName: ExpandDims,\n  inputsToSave: [\"input\"],\n  gradFunc: (dy, saved) => {\n    const [input2] = saved;\n    return { input: () => reshape(dy, input2.shape) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Expm1_grad.js\nvar expm1GradConfig = {\n  kernelName: Expm1,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(dy, exp(x)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Floor_grad.js\nvar floorGradConfig = {\n  kernelName: Floor,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/FloorDiv_grad.js\nvar floorDivGradConfig = {\n  kernelName: FloorDiv,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      const res = div(dy, cast(b, \"float32\"));\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(res, reduceAxes), a.shape);\n      }\n      return res;\n    };\n    const derB = () => {\n      let res = mul(dy, cast(a, \"float32\"));\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = reshape(sum2(res, reduceAxes), b.shape);\n      }\n      const tmp = square(b);\n      return neg(div(res, cast(tmp, \"float32\")));\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/FusedBatchNorm_grad.js\nvar fusedBatchNormGradConfig = {\n  kernelName: FusedBatchNorm,\n  inputsToSave: [\"x\", \"mean\", \"variance\", \"scale\"],\n  gradFunc: (dy, saved, attrs) => {\n    const { varianceEpsilon } = attrs;\n    const [x, mean4, variance, scale2] = saved;\n    const scaleValue = scale2 == null ? scalar(1) : scale2;\n    const reductionAxes = getReductionAxes(mean4.shape, x.shape);\n    const tileShape = [];\n    if (mean4.rank === 1) {\n      for (let i = 0; i < x.shape.length - 1; ++i) {\n        tileShape.push(x.shape[i]);\n      }\n      tileShape.push(1);\n    }\n    const xMinusMean = sub(x, mean4);\n    const dyTimesScaleValue = mul(dy, scaleValue);\n    const oneOverSqrtVariance = rsqrt(add2(variance, scalar(varianceEpsilon)));\n    const minusHalfRCube = mul(mul(mul(oneOverSqrtVariance, oneOverSqrtVariance), oneOverSqrtVariance), scalar(-0.5));\n    const derX = () => {\n      if (mean4.rank === 1) {\n        return reshape(mul(mul(dy, tile(reshape(oneOverSqrtVariance, [1, 1, 1, mean4.shape[0]]), tileShape)), scaleValue), x.shape);\n      } else {\n        return reshape(mul(mul(dy, oneOverSqrtVariance), scaleValue), x.shape);\n      }\n    };\n    const derMean = () => {\n      let meanDer = mul(mul(oneOverSqrtVariance, scalar(-1)), dyTimesScaleValue);\n      if (mean4.rank === 1) {\n        meanDer = sum2(meanDer, reductionAxes);\n      }\n      return reshape(meanDer, mean4.shape);\n    };\n    const derVariance = () => {\n      let varianceDer = mul(mul(minusHalfRCube, xMinusMean), dyTimesScaleValue);\n      if (mean4.rank === 1) {\n        varianceDer = sum2(varianceDer, reductionAxes);\n      }\n      return reshape(varianceDer, mean4.shape);\n    };\n    const derScale = () => {\n      const xMinusMean2TimesRsqrt = mul(xMinusMean, oneOverSqrtVariance);\n      let scaleDer = mul(dy, xMinusMean2TimesRsqrt);\n      if (mean4.rank === 1) {\n        scaleDer = sum2(scaleDer, reductionAxes);\n      }\n      return reshape(scaleDer, mean4.shape);\n    };\n    const derOffset = () => {\n      let offsetDer = dy;\n      if (mean4.rank === 1) {\n        offsetDer = sum2(offsetDer, reductionAxes);\n      }\n      return reshape(offsetDer, mean4.shape);\n    };\n    return {\n      x: derX,\n      mean: derMean,\n      variance: derVariance,\n      scale: derScale,\n      offset: derOffset\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/GatherV2_grad.js\nvar gatherGradConfig = {\n  kernelName: GatherV2,\n  inputsToSave: [\"x\", \"indices\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x, indices] = saved;\n    const { axis } = attrs;\n    const parsedAxis = parseAxisParam(axis, x.shape)[0];\n    const derX = () => {\n      const paramsShape = x.shape;\n      const indicesSize = indices.size;\n      const outerShape = paramsShape.slice(0, parsedAxis);\n      const outerDims = outerShape.length;\n      const innerShape = paramsShape.slice(axis, paramsShape.length).slice(1);\n      const innerDims = innerShape.length;\n      const outerAxesIndices = arrayRange(0, outerDims);\n      const innerAxesIndices = arrayRange(outerDims + 1, outerDims + 1 + innerDims);\n      const valuesShape = arrayConcat([outerShape, [indicesSize], innerShape]);\n      const values = reshape(dy, valuesShape);\n      const reshapedIndices = reshape(indices, [indicesSize]);\n      const transposeDims = arrayConcat([[outerDims], outerAxesIndices, innerAxesIndices]);\n      const valuesTranspose = transpose(values, transposeDims);\n      let paramsGrad = unsortedSegmentSum(valuesTranspose, reshapedIndices, x.shape[parsedAxis]);\n      const invertTransposeDims = getUndoAxesPermutation(transposeDims);\n      paramsGrad = transpose(paramsGrad, invertTransposeDims);\n      return paramsGrad;\n    };\n    return { x: derX, indices: () => indices };\n  }\n};\nfunction arrayRange(start, stop) {\n  const result = [];\n  for (let i = start; i < stop; ++i) {\n    result.push(i);\n  }\n  return result;\n}\nfunction arrayConcat(arrays) {\n  const result = [];\n  for (let i = 0; i < arrays.length; ++i) {\n    for (let j = 0; j < arrays[i].length; ++j) {\n      result.push(arrays[i][j]);\n    }\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/GreaterEqual_grad.js\nvar greaterEqualGradConfig = {\n  kernelName: GreaterEqual,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    return { a: () => zerosLike(a), b: () => zerosLike(b) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Identity_grad.js\nvar identityGradConfig = {\n  kernelName: Identity,\n  gradFunc: (dy) => {\n    return { x: () => cast(dy, \"float32\") };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/IsFinite_grad.js\nvar isFiniteGradConfig = {\n  kernelName: IsFinite,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/IsInf_grad.js\nvar isInfGradConfig = {\n  kernelName: IsInf,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/IsNan_grad.js\nvar isNanGradConfig = {\n  kernelName: IsNan,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/LeakyRelu_grad.js\nvar leakyReluGradConfig = {\n  kernelName: LeakyRelu,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { alpha } = attrs;\n    const mask = greater(x, 0);\n    return { x: () => where(mask, dy, mul(dy, alpha)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Log1p_grad.js\nvar log1pGradConfig = {\n  kernelName: Log1p,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, add2(x, 1)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Log_grad.js\nvar logGradConfig = {\n  kernelName: Log,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, cast(x, \"float32\")) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/LogSoftmax_grad.js\nvar logSoftmaxGradConfig = {\n  kernelName: LogSoftmax,\n  inputsToSave: [],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const [value] = saved;\n    const { axis } = attrs;\n    return {\n      logits: () => {\n        const keepDims = true;\n        const softmax6 = exp(value);\n        return sub(dy, mul(sum2(dy, axis, keepDims), softmax6));\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/local_response_normalization_backprop.js\nfunction localResponseNormalizationBackprop_(x, y, dy, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {\n  const inputs = { x, y, dy };\n  const attrs = { depthRadius, bias, alpha, beta };\n  return ENGINE.runKernel(LRNGrad, inputs, attrs);\n}\nvar localResponseNormalizationBackprop = op({ localResponseNormalizationBackprop_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/LRN_grad.js\nvar lrnGradConfig = {\n  kernelName: LRN,\n  inputsToSave: [\"x\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const [x, y] = saved;\n    const { depthRadius, bias, alpha, beta } = attrs;\n    return {\n      x: () => localResponseNormalizationBackprop(x, y, dy, depthRadius, bias, alpha, beta)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/min_max_grad_util.js\nfunction gradForMinAndMax(dy, y, xOrig, origAxes) {\n  if (y.rank < xOrig.rank) {\n    y = reshape(y, expandShapeToKeepDim(y.shape, origAxes));\n  }\n  if (dy.rank < xOrig.rank) {\n    dy = reshape(dy, expandShapeToKeepDim(dy.shape, origAxes));\n  }\n  return {\n    x: () => {\n      const dx = mul(dy, cast(equal(xOrig, y), dy.dtype));\n      return dx;\n    }\n  };\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Max_grad.js\nvar maxGradConfig = {\n  kernelName: Max,\n  inputsToSave: [\"x\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const maxAttrs = attrs;\n    const { reductionIndices } = maxAttrs;\n    const x = saved[0];\n    const y = saved[1];\n    const origAxes = parseAxisParam(reductionIndices, x.shape);\n    const maxGrad = gradForMinAndMax(dy, y, x, origAxes);\n    return {\n      x: () => {\n        return maxGrad[\"x\"]();\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Maximum_grad.js\nvar maximumGradConfig = {\n  kernelName: Maximum,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const derA = () => mul(dy, cast(greaterEqual(a, b), \"float32\"));\n    const derB = () => mul(dy, cast(less(a, b), \"float32\"));\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_3d_grad.js\nfunction maxPool3dGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {\n  const $dy = convertToTensor(dy, \"dy\", \"maxPool3dGrad\");\n  const $input = convertToTensor(input2, \"input\", \"maxPool3dGrad\");\n  const $output = convertToTensor(output, \"output\", \"maxPool3dGrad\");\n  let dy5D = $dy;\n  let input5D = $input;\n  let output5D = $output;\n  let reshapedTo5D = false;\n  if ($input.rank === 4) {\n    reshapedTo5D = true;\n    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);\n    input5D = reshape($input, [\n      1,\n      $input.shape[0],\n      $input.shape[1],\n      $input.shape[2],\n      $input.shape[3]\n    ]);\n    output5D = reshape($output, [\n      1,\n      $output.shape[0],\n      $output.shape[1],\n      $output.shape[2],\n      $output.shape[3]\n    ]);\n  }\n  assert(dy5D.rank === 5, () => `Error in maxPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);\n  assert(input5D.rank === 5, () => `Error in maxPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);\n  assert(output5D.rank === 5, () => `Error in maxPool3dGrad: output must be rank 5 but got rank ${output5D.rank}.`);\n  checkPadOnDimRoundingMode(\"maxPool3dGrad\", pad3, dimRoundingMode);\n  const inputs = { dy: dy5D, input: input5D, output: output5D };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };\n  const res = ENGINE.runKernel(MaxPool3DGrad, inputs, attrs);\n  if (reshapedTo5D) {\n    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);\n  }\n  return res;\n}\nvar maxPool3dGrad = op({ maxPool3dGrad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/MaxPool3D_grad.js\nvar maxPool3DGradConfig = {\n  kernelName: MaxPool3D,\n  inputsToSave: [\"x\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const [x, y] = saved;\n    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n    return {\n      x: () => maxPool3dGrad(dy, x, y, filterSize, strides, pad3, dimRoundingMode)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_grad.js\nfunction maxPoolGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {\n  const $dy = convertToTensor(dy, \"dy\", \"maxPoolGrad\");\n  const $input = convertToTensor(input2, \"input\", \"maxPoolGrad\");\n  const $output = convertToTensor(output, \"output\", \"maxPoolGrad\");\n  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);\n  assert($dy.rank === 4, () => `Error in maxPoolGrad: dy must be rank 4 but got rank ${$dy.rank}.`);\n  assert($input.rank === 4, () => `Error in maxPoolGrad: input must be rank 4 but got rank ${$input.rank}.`);\n  checkPadOnDimRoundingMode(\"maxPoolGrad\", pad3, dimRoundingMode);\n  const inputs = { dy: $dy, input: $input, output: $output };\n  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };\n  return ENGINE.runKernel(MaxPoolGrad, inputs, attrs);\n}\nvar maxPoolGrad = op({ maxPoolGrad_ });\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/MaxPool_grad.js\nvar maxPoolGradConfig = {\n  kernelName: MaxPool,\n  inputsToSave: [\"x\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const [x, y] = saved;\n    const { filterSize, strides, pad: pad3 } = attrs;\n    return {\n      x: () => maxPoolGrad(dy, x, y, filterSize, strides, pad3)\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Mean_grad.js\nvar meanGradConfig = {\n  kernelName: Mean,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { axis } = attrs;\n    const axes = parseAxisParam(axis, x.shape);\n    const shapes = computeOutAndReduceShapes(x.shape, axes);\n    const reduceShape = shapes[1];\n    const reduceSize = sizeFromShape(reduceShape);\n    const derX = () => {\n      const expandedDyShape = x.shape.slice();\n      axes.forEach((axis2) => {\n        expandedDyShape[axis2] = 1;\n      });\n      const expandedDy = reshape(dy, expandedDyShape);\n      const res = div(mul(expandedDy, ones2(x.shape, \"float32\")), reduceSize);\n      return res;\n    };\n    return { x: derX };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Min_grad.js\nvar minGradConfig = {\n  kernelName: Min,\n  inputsToSave: [\"x\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const minAttrs = attrs;\n    const { axis } = minAttrs;\n    const [x, y] = saved;\n    const origAxes = parseAxisParam(axis, x.shape);\n    const minGrad = gradForMinAndMax(dy, y, x, origAxes);\n    return {\n      x: () => {\n        return minGrad[\"x\"]();\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Minimum_grad.js\nvar minimumGradConfig = {\n  kernelName: Minimum,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const derA = () => mul(dy, cast(lessEqual(a, b), \"float32\"));\n    const derB = () => mul(dy, cast(greater(a, b), \"float32\"));\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/MirrorPad_grad.js\nvar mirrorPadGradConfig = {\n  kernelName: MirrorPad,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const x = saved[0];\n    const { paddings } = attrs;\n    const begin = paddings.map((p2) => p2[0]);\n    return { x: () => slice(dy, begin, x.shape) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Mod_grad.js\nvar modGradConfig = {\n  kernelName: Mod,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(dy, reduceAxes), a.shape);\n      }\n      return dy;\n    };\n    const derB = () => {\n      const res = mul(dy, neg(floor(div(a, b))));\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(res, reduceAxes), b.shape);\n      }\n      return res;\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Multiply_grad.js\nvar multiplyGradConfig = {\n  kernelName: Multiply,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      const res = mul(dy, cast(b, \"float32\"));\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(res, reduceAxes), a.shape);\n      }\n      return res;\n    };\n    const derB = () => {\n      const res = mul(dy, cast(a, \"float32\"));\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(res, reduceAxes), b.shape);\n      }\n      return res;\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Neg_grad.js\nvar negGradConfig = {\n  kernelName: Neg,\n  gradFunc: (dy) => {\n    return { x: () => neg(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/OneHot_grad.js\nvar oneHotGradConfig = {\n  kernelName: OneHot,\n  inputsToSave: [\"indices\"],\n  gradFunc: (dy, saved) => {\n    const indices = saved[0];\n    return { indices: () => zeros(indices.shape, \"float32\") };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/OnesLike_grad.js\nvar onesLikeGradConfig = {\n  kernelName: OnesLike,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Pack_grad.js\nvar packGradConfig = {\n  kernelName: Pack,\n  saveAllInputs: true,\n  gradFunc: (dy, saved, attrs) => {\n    const { axis } = attrs;\n    const derTensors = unstack(dy, axis);\n    return derTensors.map((t) => () => t);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/PadV2_grad.js\nvar padV2GradConfig = {\n  kernelName: PadV2,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const x = saved[0];\n    const { paddings } = attrs;\n    const begin = paddings.map((p2) => p2[0]);\n    return { x: () => slice(dy, begin, x.shape) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Pow_grad.js\nvar powGradConfig = {\n  kernelName: Pow,\n  inputsToSave: [\"a\", \"b\"],\n  outputsToSave: [true],\n  gradFunc: (dy, saved) => {\n    const [a, b, y] = saved;\n    const base = a;\n    const exp4 = b;\n    const outShape = assertAndGetBroadcastShape(base.shape, exp4.shape);\n    const derBase = () => {\n      const expFloat = cast(exp4, \"float32\");\n      let res = mul(dy, mul(expFloat, pow(base, sub(expFloat, scalar(1)))));\n      const reduceAxes = getReductionAxes(base.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, base.shape);\n    };\n    const derExp = () => {\n      const condition = greater(base, 0);\n      const logBase = where(condition, log2(base), zerosLike(base));\n      let res = mul(dy, mul(y, logBase));\n      const reduceAxes = getReductionAxes(exp4.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, exp4.shape);\n    };\n    return { a: derBase, b: derExp };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Prelu_grad.js\nvar preluGradConfig = {\n  kernelName: Prelu,\n  inputsToSave: [\"x\", \"alpha\"],\n  gradFunc: (dy, saved) => {\n    const [x, alpha] = saved;\n    const mask = greater(x, 0);\n    return {\n      x: () => where(mask, dy, mul(dy, alpha)),\n      alpha: () => {\n        let res = where(mask, zerosLike(dy), mul(dy, x));\n        const reduceAxes = getReductionAxes(alpha.shape, dy.shape);\n        if (reduceAxes.length > 0) {\n          res = sum2(res, reduceAxes);\n        }\n        return reshape(res, alpha.shape);\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Prod_grad.js\nfunction prodGradFn_(x, dy, axis) {\n  const expandedYShape = x.shape.slice();\n  expandedYShape[axis] = 1;\n  const expandedDy = reshape(dy, expandedYShape);\n  const xCumProd = cumprod(x, axis, true, false);\n  const xCumRevProd = cumprod(x, axis, true, true);\n  const dx = mul(xCumProd, xCumRevProd);\n  return mul(expandedDy, dx);\n}\nfunction prodsGradFn_(x, dy, axis) {\n  const xRank = x.shape.length;\n  const finalProdAxis = xRank - axis.length;\n  const xPermutation = backend_util_exports.getAxesPermutation(axis, xRank);\n  let permutedX = x;\n  if (xPermutation != null) {\n    permutedX = transpose(x, xPermutation);\n  }\n  const newShape = permutedX.shape.slice();\n  const removedShape = newShape.splice(xRank - axis.length, axis.length);\n  const endPartShape = removedShape.reduce((p2, c) => p2 * c, 1);\n  newShape.push(endPartShape);\n  const reshapedPermutedX = permutedX.reshape(newShape);\n  let prodGrad = prodGradFn_(reshapedPermutedX, dy, finalProdAxis);\n  prodGrad = prodGrad.reshape(permutedX.shape);\n  if (xPermutation != null) {\n    const undoPermutation = backend_util_exports.getUndoAxesPermutation(xPermutation);\n    prodGrad = transpose(prodGrad, undoPermutation);\n  }\n  return prodGrad;\n}\nvar prodGradConfig = {\n  kernelName: Prod,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { axis } = attrs;\n    let axisArr = [];\n    if (axis === void 0 || axis === null) {\n      axisArr = x.shape.map((_, i) => i);\n    } else if (typeof axis === \"number\") {\n      axisArr = [axis];\n    } else {\n      axisArr = axis;\n    }\n    return { x: () => prodsGradFn_(x, dy, axisArr) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/RealDiv_grad.js\nvar divGradConfig = {\n  kernelName: RealDiv,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      const res = div(dy, cast(b, \"float32\"));\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        return reshape(sum2(res, reduceAxes), a.shape);\n      }\n      return res;\n    };\n    const derB = () => {\n      let res = mul(dy, cast(a, \"float32\"));\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = reshape(sum2(res, reduceAxes), b.shape);\n      }\n      const tmp = square(b);\n      return neg(div(res, cast(tmp, \"float32\")));\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Reciprocal_grad.js\nvar reciprocalGradConfig = {\n  kernelName: Reciprocal,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, neg(square(x))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Relu6_grad.js\nvar relu6GradConfig = {\n  kernelName: Relu6,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    const mask = mul(lessEqual(x, 6), step(x));\n    return { x: () => mul(dy, cast(mask, \"float32\")) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Relu_grad.js\nvar reluGradConfig = {\n  kernelName: Relu,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(dy, cast(step(x), \"float32\")) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Reshape_grad.js\nvar reshapeGradConfig = {\n  kernelName: Reshape,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => reshape(dy, x.shape) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ResizeBilinear_grad.js\nvar resizeBilinearGradConfig = {\n  kernelName: ResizeBilinear,\n  inputsToSave: [\"images\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [images] = saved;\n    const inputs = { dy, images };\n    const imagesDer = () => ENGINE.runKernel(ResizeBilinearGrad, inputs, attrs);\n    return { images: imagesDer };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ResizeNearestNeighbor_grad.js\nvar resizeNearestNeighborGradConfig = {\n  kernelName: ResizeNearestNeighbor,\n  inputsToSave: [\"images\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [images] = saved;\n    const inputs = { dy, images };\n    const imagesDer = () => ENGINE.runKernel(ResizeNearestNeighborGrad, inputs, attrs);\n    return { images: imagesDer };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Reverse_grad.js\nvar reverseGradConfig = {\n  kernelName: Reverse,\n  gradFunc: (dy, saved, attrs) => {\n    const { dims } = attrs;\n    const axes = parseAxisParam(dims, dy.shape);\n    return { x: () => reverse(dy, axes) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Round_grad.js\nvar roundGradConfig = {\n  kernelName: Round,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Rsqrt_grad.js\nvar rsqrtGradConfig = {\n  kernelName: Rsqrt,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => neg(div(dy, mul(pow(x, 1.5), 2))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Select_grad.js\nvar selectGradConfig = {\n  kernelName: Select,\n  inputsToSave: [\"condition\"],\n  gradFunc: (dy, saved) => {\n    const [condition] = saved;\n    return {\n      condition: () => cast(zerosLike(condition), \"float32\"),\n      t: () => mul(dy, cast(condition, dy.dtype)),\n      e: () => mul(dy, cast(logicalNot(condition), dy.dtype))\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Selu_grad.js\nvar seluGradConfig = {\n  kernelName: Selu,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return {\n      x: () => {\n        const mask = greater(x, scalar(0));\n        const scaleAlpha2 = scalar(SELU_SCALEALPHA);\n        const scale2 = scalar(SELU_SCALE);\n        const greaterThanZeroDer = mul(dy, scale2);\n        const lessEqualZeroDer = mul(mul(dy, scaleAlpha2), exp(cast(x, \"float32\")));\n        return where(mask, greaterThanZeroDer, lessEqualZeroDer);\n      }\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sigmoid_grad.js\nvar sigmoidGradConfig = {\n  kernelName: Sigmoid,\n  outputsToSave: [true],\n  gradFunc: (dy, saved) => {\n    const [y] = saved;\n    return { x: () => mul(dy, mul(y, sub(scalar(1), y))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sign_grad.js\nvar signGradConfig = {\n  kernelName: Sign,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sin_grad.js\nvar sinGradConfig = {\n  kernelName: Sin,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(cos(cast(x, \"float32\")), dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sinh_grad.js\nvar sinhGradConfig = {\n  kernelName: Sinh,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(cosh(cast(x, \"float32\")), dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Slice_grad.js\nvar sliceGradConfig = {\n  kernelName: Slice,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { begin, size } = attrs;\n    const inputShape = x.shape;\n    const [begin_, size_] = parseSliceParams(x, begin, size);\n    const paddings = [];\n    for (let i = 0; i < dy.rank; i++) {\n      paddings.push([begin_[i], inputShape[i] - begin_[i] - size_[i]]);\n    }\n    return { x: () => pad(dy, paddings) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Softmax_grad.js\nvar softmaxGradConfig = {\n  kernelName: Softmax,\n  outputsToSave: [true],\n  gradFunc: (dy, saved, attrs) => {\n    const [y] = saved;\n    const { dim } = attrs;\n    const keepDims = true;\n    const dyTimesY = mul(dy, y);\n    return {\n      logits: () => sub(dyTimesY, mul(sum2(dyTimesY, [dim], keepDims), y))\n    };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Softplus_grad.js\nvar softplusGradConfig = {\n  kernelName: Softplus,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(dy, sigmoid(x)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/SpaceToBatchND_grad.js\nvar spaceToBatchNDGradConfig = {\n  kernelName: SpaceToBatchND,\n  gradFunc: (dy, saved, attrs) => {\n    const { blockShape, paddings } = attrs;\n    return { x: () => batchToSpaceND(dy, blockShape, paddings) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/SplitV_grad.js\nvar splitVGradConfig = {\n  kernelName: SplitV,\n  gradFunc: (dy, saved, attrs) => {\n    const { axis } = attrs;\n    return { x: () => concat(dy, axis) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sqrt_grad.js\nvar sqrtGradConfig = {\n  kernelName: Sqrt,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, mul(sqrt(cast(x, \"float32\")), 2)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Square_grad.js\nvar squareGradConfig = {\n  kernelName: Square,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => mul(dy, mul(cast(x, \"float32\"), 2)) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/SquaredDifference_grad.js\nvar squaredDifferenceGradConfig = {\n  kernelName: SquaredDifference,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const two = scalar(2);\n    const derA = () => mul(dy, mul(two, sub(a, b)));\n    const derB = () => mul(dy, mul(two, sub(b, a)));\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Step_grad.js\nvar stepGradConfig = {\n  kernelName: Step,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sub_grad.js\nvar subGradConfig = {\n  kernelName: Sub,\n  inputsToSave: [\"a\", \"b\"],\n  gradFunc: (dy, saved) => {\n    const [a, b] = saved;\n    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);\n    const derA = () => {\n      let res = dy;\n      const reduceAxes = getReductionAxes(a.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(res, a.shape);\n    };\n    const derB = () => {\n      let res = dy;\n      const reduceAxes = getReductionAxes(b.shape, outShape);\n      if (reduceAxes.length > 0) {\n        res = sum2(res, reduceAxes);\n      }\n      return reshape(neg(res), b.shape);\n    };\n    return { a: derA, b: derB };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Sum_grad.js\nvar sumGradConfig = {\n  kernelName: Sum,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const expandedDyShape = x.shape.slice();\n    const { axis } = attrs;\n    const axes = parseAxisParam(axis, x.shape);\n    axes.forEach((axis2) => {\n      expandedDyShape[axis2] = 1;\n    });\n    const expandedDy = reshape(dy, expandedDyShape);\n    const derX = mul(expandedDy, ones2(x.shape, \"float32\"));\n    return { x: () => derX };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Tan_grad.js\nvar tanGradConfig = {\n  kernelName: Tan,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved) => {\n    const [x] = saved;\n    return { x: () => div(dy, square(cos(x))) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Tanh_grad.js\nvar tanhGradConfig = {\n  kernelName: Tanh,\n  outputsToSave: [true],\n  gradFunc: (dy, saved) => {\n    const [y] = saved;\n    return { x: () => mul(sub(scalar(1), square(y)), dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Tile_grad.js\nvar tileGradConfig = {\n  kernelName: Tile,\n  inputsToSave: [\"x\"],\n  gradFunc: (dy, saved, attrs) => {\n    const [x] = saved;\n    const { reps } = attrs;\n    const derX = () => {\n      let xGrad = zerosLike(x);\n      if (x.rank === 1) {\n        for (let i = 0; i < reps[0]; ++i) {\n          xGrad = add2(xGrad, slice(dy, [i * x.shape[0]], [x.shape[0]]));\n        }\n      } else if (x.rank === 2) {\n        for (let i = 0; i < reps[0]; ++i) {\n          for (let j = 0; j < reps[1]; ++j) {\n            xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1]], [\n              x.shape[0],\n              x.shape[1]\n            ]));\n          }\n        }\n      } else if (x.rank === 3) {\n        for (let i = 0; i < reps[0]; ++i) {\n          for (let j = 0; j < reps[1]; ++j) {\n            for (let k = 0; k < reps[2]; ++k) {\n              xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1], k * x.shape[2]], [x.shape[0], x.shape[1], x.shape[2]]));\n            }\n          }\n        }\n      } else if (x.rank === 4) {\n        for (let i = 0; i < reps[0]; ++i) {\n          for (let j = 0; j < reps[1]; ++j) {\n            for (let k = 0; k < reps[2]; ++k) {\n              for (let l = 0; l < reps[3]; ++l) {\n                xGrad = add2(xGrad, slice(dy, [\n                  i * x.shape[0],\n                  j * x.shape[1],\n                  k * x.shape[2],\n                  l * x.shape[3]\n                ], [x.shape[0], x.shape[1], x.shape[2], x.shape[3]]));\n              }\n            }\n          }\n        }\n      } else {\n        throw new Error(`Gradient for tile operation is not implemented for rank-${x.rank} tensors yet.`);\n      }\n      return xGrad;\n    };\n    return { x: derX };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Transpose_grad.js\nvar transposeGradConfig = {\n  kernelName: Transpose,\n  gradFunc: (dy, saved, attrs) => {\n    const transposeAttrs = attrs;\n    const { perm } = transposeAttrs;\n    const undoPerm = getUndoAxesPermutation(perm);\n    return { x: () => transpose(dy, undoPerm) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/Unpack_grad.js\nvar unpackGradConfig = {\n  kernelName: Unpack,\n  gradFunc: (dy, saved, attrs) => {\n    const unpackAttrs = attrs;\n    const { axis } = unpackAttrs;\n    return { value: () => stack(dy, axis) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/UnsortedSegmentSum_grad.js\nvar unsortedSegmentSumGradConfig = {\n  kernelName: UnsortedSegmentSum,\n  inputsToSave: [\"segmentIds\"],\n  gradFunc: (dy, saved) => {\n    const [segmentIds] = saved;\n    const derX = () => {\n      return gatherDropNegatives(dy, segmentIds);\n    };\n    return { x: derX };\n  }\n};\nfunction gatherDropNegatives(x, indices) {\n  const zeroClippedIndices = maximum(indices, zerosLike(indices));\n  const gathered = gather(x, zeroClippedIndices);\n  let isPositive = greaterEqual(indices, scalar(0, \"int32\"));\n  const numIters = gathered.rank - isPositive.rank;\n  for (let i = 0; i < numIters; ++i) {\n    isPositive = expandDims(isPositive, i + 1);\n  }\n  isPositive = logicalAnd(isPositive, ones2(gathered.shape, \"bool\"));\n  const zeroSlice = zerosLike(gathered);\n  return where(isPositive, gathered, zeroSlice);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/gradients/ZerosLike_grad.js\nvar zerosLikeGradConfig = {\n  kernelName: ZerosLike,\n  gradFunc: (dy) => {\n    return { x: () => zerosLike(dy) };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/register_all_gradients.js\nvar gradConfigs = [\n  absGradConfig,\n  acosGradConfig,\n  acoshGradConfig,\n  addGradConfig,\n  addNGradConfig,\n  argMaxGradConfig,\n  argMinGradConfig,\n  asinGradConfig,\n  asinhGradConfig,\n  atan2GradConfig,\n  atanGradConfig,\n  atanhGradConfig,\n  avgPool3DGradConfig,\n  avgPoolGradConfig,\n  batchMatMulGradConfig,\n  batchToSpaceNDGradConfig,\n  broadcastToGradConfig,\n  castGradConfig,\n  ceilGradConfig,\n  clipByValueGradConfig,\n  complexAbsGradConfig,\n  concatGradConfig,\n  conv2DBackpropInputGradConfig,\n  conv2DGradConfig,\n  conv3DGradConfig,\n  cosGradConfig,\n  coshGradConfig,\n  cumsumGradConfig,\n  depthwiseConv2dNativeGradConfig,\n  dilation2dGradConfig,\n  divGradConfig,\n  eluGradConfig,\n  erfGradConfig,\n  expGradConfig,\n  expandDimsGradConfig,\n  expm1GradConfig,\n  floorDivGradConfig,\n  floorGradConfig,\n  fusedBatchNormGradConfig,\n  gatherGradConfig,\n  greaterEqualGradConfig,\n  identityGradConfig,\n  isFiniteGradConfig,\n  isInfGradConfig,\n  isNanGradConfig,\n  leakyReluGradConfig,\n  log1pGradConfig,\n  logGradConfig,\n  logSoftmaxGradConfig,\n  lrnGradConfig,\n  maxGradConfig,\n  maxGradConfig,\n  maximumGradConfig,\n  maxPool3DGradConfig,\n  maxPoolGradConfig,\n  meanGradConfig,\n  minGradConfig,\n  minimumGradConfig,\n  mirrorPadGradConfig,\n  modGradConfig,\n  multiplyGradConfig,\n  negGradConfig,\n  oneHotGradConfig,\n  onesLikeGradConfig,\n  packGradConfig,\n  padV2GradConfig,\n  padV2GradConfig,\n  powGradConfig,\n  preluGradConfig,\n  prodGradConfig,\n  reciprocalGradConfig,\n  relu6GradConfig,\n  reluGradConfig,\n  reshapeGradConfig,\n  resizeBilinearGradConfig,\n  resizeNearestNeighborGradConfig,\n  reverseGradConfig,\n  roundGradConfig,\n  rsqrtGradConfig,\n  selectGradConfig,\n  seluGradConfig,\n  sigmoidGradConfig,\n  signGradConfig,\n  sinGradConfig,\n  sinhGradConfig,\n  sliceGradConfig,\n  softmaxGradConfig,\n  softplusGradConfig,\n  spaceToBatchNDGradConfig,\n  spaceToBatchNDGradConfig,\n  splitVGradConfig,\n  splitVGradConfig,\n  sqrtGradConfig,\n  squaredDifferenceGradConfig,\n  squareGradConfig,\n  stepGradConfig,\n  subGradConfig,\n  sumGradConfig,\n  tanGradConfig,\n  tanhGradConfig,\n  tileGradConfig,\n  transposeGradConfig,\n  unpackGradConfig,\n  unsortedSegmentSumGradConfig,\n  zerosLikeGradConfig\n];\nfor (const gradientConfig of gradConfigs) {\n  registerGradient(gradientConfig);\n}\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/abs.js\ngetGlobalTensorClass().prototype.abs = function() {\n  this.throwIfDisposed();\n  return abs(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/acos.js\ngetGlobalTensorClass().prototype.acos = function() {\n  this.throwIfDisposed();\n  return acos(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/acosh.js\ngetGlobalTensorClass().prototype.acosh = function() {\n  this.throwIfDisposed();\n  return acosh(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/add.js\ngetGlobalTensorClass().prototype.add = function(b) {\n  this.throwIfDisposed();\n  return add2(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/all.js\ngetGlobalTensorClass().prototype.all = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return all(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/any.js\ngetGlobalTensorClass().prototype.any = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return any(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/arg_max.js\ngetGlobalTensorClass().prototype.argMax = function(axis) {\n  this.throwIfDisposed();\n  return argMax(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/arg_min.js\ngetGlobalTensorClass().prototype.argMin = function(axis) {\n  this.throwIfDisposed();\n  return argMin(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as_scalar.js\ngetGlobalTensorClass().prototype.asScalar = function() {\n  this.throwIfDisposed();\n  assert(this.size === 1, () => \"The array must have only 1 element.\");\n  return reshape(this, []);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as_type.js\ngetGlobalTensorClass().prototype.asType = function(dtype) {\n  this.throwIfDisposed();\n  return cast(this, dtype);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as1d.js\ngetGlobalTensorClass().prototype.as1D = function() {\n  this.throwIfDisposed();\n  return reshape(this, [this.size]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as2d.js\ngetGlobalTensorClass().prototype.as2D = function(rows, columns) {\n  this.throwIfDisposed();\n  return reshape(this, [rows, columns]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as3d.js\ngetGlobalTensorClass().prototype.as3D = function(rows, columns, depth) {\n  this.throwIfDisposed();\n  return reshape(this, [rows, columns, depth]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as4d.js\ngetGlobalTensorClass().prototype.as4D = function(rows, columns, depth, depth2) {\n  this.throwIfDisposed();\n  return reshape(this, [rows, columns, depth, depth2]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as5d.js\ngetGlobalTensorClass().prototype.as5D = function(rows, columns, depth, depth2, depth3) {\n  this.throwIfDisposed();\n  return reshape(this, [rows, columns, depth, depth2, depth3]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/asin.js\ngetGlobalTensorClass().prototype.asin = function() {\n  this.throwIfDisposed();\n  return asin(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/asinh.js\ngetGlobalTensorClass().prototype.asinh = function() {\n  this.throwIfDisposed();\n  return asinh(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atan.js\ngetGlobalTensorClass().prototype.atan = function() {\n  this.throwIfDisposed();\n  return atan(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atan2.js\ngetGlobalTensorClass().prototype.atan2 = function(b) {\n  this.throwIfDisposed();\n  return atan2(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atanh.js\ngetGlobalTensorClass().prototype.atanh = function() {\n  this.throwIfDisposed();\n  return atanh(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/avg_pool.js\ngetGlobalTensorClass().prototype.avgPool = function(filterSize, strides, pad3, dimRoundingMode) {\n  this.throwIfDisposed();\n  return avgPool(this, filterSize, strides, pad3, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/batch_to_space_nd.js\ngetGlobalTensorClass().prototype.batchToSpaceND = function(blockShape, crops) {\n  this.throwIfDisposed();\n  return batchToSpaceND(this, blockShape, crops);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/batchnorm.js\ngetGlobalTensorClass().prototype.batchNorm = function(mean4, variance, offset, scale2, varianceEpsilon) {\n  this.throwIfDisposed();\n  return batchNorm(this, mean4, variance, offset, scale2, varianceEpsilon);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/broadcast_to.js\ngetGlobalTensorClass().prototype.broadcastTo = function(shape) {\n  this.throwIfDisposed();\n  return broadcastTo(this, shape);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cast.js\ngetGlobalTensorClass().prototype.cast = function(dtype) {\n  this.throwIfDisposed();\n  return cast(this, dtype);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ceil.js\ngetGlobalTensorClass().prototype.ceil = function() {\n  this.throwIfDisposed();\n  return ceil(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/clip_by_value.js\ngetGlobalTensorClass().prototype.clipByValue = function(min6, max6) {\n  this.throwIfDisposed();\n  return clipByValue(this, min6, max6);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/concat.js\ngetGlobalTensorClass().prototype.concat = function(x, axis) {\n  this.throwIfDisposed();\n  if (x instanceof Tensor) {\n    x = [x];\n  }\n  return concat([this, ...x], axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv1d.js\ngetGlobalTensorClass().prototype.conv1d = function(filter, stride, pad3, dataFormat, dilation, dimRoundingMode) {\n  this.throwIfDisposed();\n  return conv1d(this, filter, stride, pad3, dataFormat, dilation, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv2d_transpose.js\ngetGlobalTensorClass().prototype.conv2dTranspose = function(filter, outputShape, strides, pad3, dimRoundingMode) {\n  this.throwIfDisposed();\n  return conv2dTranspose(this, filter, outputShape, strides, pad3, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv2d.js\ngetGlobalTensorClass().prototype.conv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {\n  this.throwIfDisposed();\n  return conv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cos.js\ngetGlobalTensorClass().prototype.cos = function() {\n  this.throwIfDisposed();\n  return cos(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cosh.js\ngetGlobalTensorClass().prototype.cosh = function() {\n  this.throwIfDisposed();\n  return cosh(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cumprod.js\ngetGlobalTensorClass().prototype.cumprod = function(axis, exclusive, reverse5) {\n  this.throwIfDisposed();\n  return cumprod(this, axis, exclusive, reverse5);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cumsum.js\ngetGlobalTensorClass().prototype.cumsum = function(axis, exclusive, reverse5) {\n  this.throwIfDisposed();\n  return cumsum(this, axis, exclusive, reverse5);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/depth_to_space.js\ngetGlobalTensorClass().prototype.depthToSpace = function(blockSize, dataFormat) {\n  this.throwIfDisposed();\n  return depthToSpace(this, blockSize, dataFormat);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/depthwise_conv2d.js\ngetGlobalTensorClass().prototype.depthwiseConv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {\n  this.throwIfDisposed();\n  return depthwiseConv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/dilation2d.js\ngetGlobalTensorClass().prototype.dilation2d = function(filter, strides, pad3, dilations, dataFormat) {\n  this.throwIfDisposed();\n  return dilation2d(this, filter, strides, pad3, dilations, dataFormat);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/div_no_nan.js\ngetGlobalTensorClass().prototype.divNoNan = function(b) {\n  this.throwIfDisposed();\n  return divNoNan(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/div.js\ngetGlobalTensorClass().prototype.div = function(b) {\n  this.throwIfDisposed();\n  return div(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/dot.js\ngetGlobalTensorClass().prototype.dot = function(b) {\n  this.throwIfDisposed();\n  return dot(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/elu.js\ngetGlobalTensorClass().prototype.elu = function() {\n  this.throwIfDisposed();\n  return elu(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/equal.js\ngetGlobalTensorClass().prototype.equal = function(b) {\n  this.throwIfDisposed();\n  return equal(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/erf.js\ngetGlobalTensorClass().prototype.erf = function() {\n  this.throwIfDisposed();\n  return erf(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/euclidean_norm.js\ngetGlobalTensorClass().prototype.euclideanNorm = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return euclideanNorm(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/exp.js\ngetGlobalTensorClass().prototype.exp = function() {\n  this.throwIfDisposed();\n  return exp(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/expand_dims.js\ngetGlobalTensorClass().prototype.expandDims = function(axis) {\n  this.throwIfDisposed();\n  return expandDims(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/expm1.js\ngetGlobalTensorClass().prototype.expm1 = function() {\n  this.throwIfDisposed();\n  return expm1(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/fft.js\ngetGlobalTensorClass().prototype.fft = function() {\n  this.throwIfDisposed();\n  return fft(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/flatten.js\ngetGlobalTensorClass().prototype.flatten = function() {\n  this.throwIfDisposed();\n  return reshape(this, [this.size]);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/floor.js\ngetGlobalTensorClass().prototype.floor = function() {\n  this.throwIfDisposed();\n  return floor(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/floorDiv.js\ngetGlobalTensorClass().prototype.floorDiv = function(b) {\n  this.throwIfDisposed();\n  return floorDiv(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/gather.js\ngetGlobalTensorClass().prototype.gather = function(indices, axis) {\n  this.throwIfDisposed();\n  return gather(this, indices, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/greater_equal.js\ngetGlobalTensorClass().prototype.greaterEqual = function(b) {\n  this.throwIfDisposed();\n  return greaterEqual(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/greater.js\ngetGlobalTensorClass().prototype.greater = function(b) {\n  this.throwIfDisposed();\n  return greater(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ifft.js\ngetGlobalTensorClass().prototype.ifft = function() {\n  this.throwIfDisposed();\n  return ifft(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/irfft.js\ngetGlobalTensorClass().prototype.irfft = function() {\n  this.throwIfDisposed();\n  return irfft(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_finite.js\ngetGlobalTensorClass().prototype.isFinite = function() {\n  this.throwIfDisposed();\n  return isFinite2(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_inf.js\ngetGlobalTensorClass().prototype.isInf = function() {\n  this.throwIfDisposed();\n  return isInf(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_nan.js\ngetGlobalTensorClass().prototype.isNaN = function() {\n  this.throwIfDisposed();\n  return isNaN2(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/leaky_relu.js\ngetGlobalTensorClass().prototype.leakyRelu = function(alpha) {\n  this.throwIfDisposed();\n  return leakyRelu(this, alpha);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/less_equal.js\ngetGlobalTensorClass().prototype.lessEqual = function(b) {\n  this.throwIfDisposed();\n  return lessEqual(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/less.js\ngetGlobalTensorClass().prototype.less = function(b) {\n  this.throwIfDisposed();\n  return less(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/local_response_normalization.js\ngetGlobalTensorClass().prototype.localResponseNormalization = function(depthRadius, bias, alpha, beta) {\n  this.throwIfDisposed();\n  return localResponseNormalization(this, depthRadius, bias, alpha, beta);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_sigmoid.js\ngetGlobalTensorClass().prototype.logSigmoid = function() {\n  this.throwIfDisposed();\n  return logSigmoid(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_softmax.js\ngetGlobalTensorClass().prototype.logSoftmax = function(axis) {\n  this.throwIfDisposed();\n  return logSoftmax(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_sum_exp.js\ngetGlobalTensorClass().prototype.logSumExp = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return logSumExp(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log.js\ngetGlobalTensorClass().prototype.log = function() {\n  this.throwIfDisposed();\n  return log2(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log1p.js\ngetGlobalTensorClass().prototype.log1p = function() {\n  this.throwIfDisposed();\n  return log1p(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_and.js\ngetGlobalTensorClass().prototype.logicalAnd = function(b) {\n  this.throwIfDisposed();\n  return logicalAnd(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_not.js\ngetGlobalTensorClass().prototype.logicalNot = function() {\n  this.throwIfDisposed();\n  return logicalNot(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_or.js\ngetGlobalTensorClass().prototype.logicalOr = function(b) {\n  this.throwIfDisposed();\n  return logicalOr(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_xor.js\ngetGlobalTensorClass().prototype.logicalXor = function(b) {\n  this.throwIfDisposed();\n  return logicalXor(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mat_mul.js\ngetGlobalTensorClass().prototype.matMul = function(b, transposeA, transposeB) {\n  this.throwIfDisposed();\n  return matMul(this, b, transposeA, transposeB);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/max_pool.js\ngetGlobalTensorClass().prototype.maxPool = function(filterSize, strides, pad3, dimRoundingMode) {\n  this.throwIfDisposed();\n  return maxPool(this, filterSize, strides, pad3, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/max.js\ngetGlobalTensorClass().prototype.max = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return max(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/maximum.js\ngetGlobalTensorClass().prototype.maximum = function(b) {\n  this.throwIfDisposed();\n  return maximum(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mean.js\ngetGlobalTensorClass().prototype.mean = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return mean(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/min.js\ngetGlobalTensorClass().prototype.min = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return min(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/minimum.js\ngetGlobalTensorClass().prototype.minimum = function(b) {\n  this.throwIfDisposed();\n  return minimum(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mirror_pad.js\ngetGlobalTensorClass().prototype.mirrorPad = function(paddings, mode) {\n  this.throwIfDisposed();\n  return mirrorPad(this, paddings, mode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mod.js\ngetGlobalTensorClass().prototype.mod = function(b) {\n  this.throwIfDisposed();\n  return mod(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mul.js\ngetGlobalTensorClass().prototype.mul = function(b) {\n  this.throwIfDisposed();\n  return mul(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/neg.js\ngetGlobalTensorClass().prototype.neg = function() {\n  this.throwIfDisposed();\n  return neg(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/norm.js\ngetGlobalTensorClass().prototype.norm = function(ord, axis, keepDims) {\n  this.throwIfDisposed();\n  return norm(this, ord, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/not_equal.js\ngetGlobalTensorClass().prototype.notEqual = function(b) {\n  this.throwIfDisposed();\n  return notEqual(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/one_hot.js\ngetGlobalTensorClass().prototype.oneHot = function(depth, onValue = 1, offValue = 0) {\n  this.throwIfDisposed();\n  return oneHot(this, depth, onValue, offValue);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ones_like.js\ngetGlobalTensorClass().prototype.onesLike = function() {\n  this.throwIfDisposed();\n  return onesLike(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pad.js\ngetGlobalTensorClass().prototype.pad = function(paddings, constantValue) {\n  this.throwIfDisposed();\n  return pad(this, paddings, constantValue);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pool.js\ngetGlobalTensorClass().prototype.pool = function(windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode) {\n  this.throwIfDisposed();\n  return pool(this, windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pow.js\ngetGlobalTensorClass().prototype.pow = function(exp4) {\n  this.throwIfDisposed();\n  return pow(this, exp4);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/prelu.js\ngetGlobalTensorClass().prototype.prelu = function(alpha) {\n  this.throwIfDisposed();\n  return prelu(this, alpha);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/prod.js\ngetGlobalTensorClass().prototype.prod = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return prod(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reciprocal.js\ngetGlobalTensorClass().prototype.reciprocal = function() {\n  this.throwIfDisposed();\n  return reciprocal(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/relu.js\ngetGlobalTensorClass().prototype.relu = function() {\n  this.throwIfDisposed();\n  return relu(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/relu6.js\ngetGlobalTensorClass().prototype.relu6 = function() {\n  this.throwIfDisposed();\n  return relu6(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reshape_as.js\ngetGlobalTensorClass().prototype.reshapeAs = function(x) {\n  this.throwIfDisposed();\n  return reshape(this, x.shape);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reshape.js\ngetGlobalTensorClass().prototype.reshape = function(shape) {\n  this.throwIfDisposed();\n  return reshape(this, shape);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/resize_bilinear.js\ngetGlobalTensorClass().prototype.resizeBilinear = function(newShape2D, alignCorners, halfPixelCenters) {\n  this.throwIfDisposed();\n  return resizeBilinear(this, newShape2D, alignCorners, halfPixelCenters);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/resize_nearest_neighbor.js\ngetGlobalTensorClass().prototype.resizeNearestNeighbor = function(newShape2D, alignCorners, halfFloatCenters) {\n  this.throwIfDisposed();\n  return resizeNearestNeighbor(this, newShape2D, alignCorners, halfFloatCenters);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reverse.js\ngetGlobalTensorClass().prototype.reverse = function(axis) {\n  this.throwIfDisposed();\n  return reverse(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/rfft.js\ngetGlobalTensorClass().prototype.rfft = function() {\n  this.throwIfDisposed();\n  return rfft(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/round.js\ngetGlobalTensorClass().prototype.round = function() {\n  this.throwIfDisposed();\n  return round2(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/rsqrt.js\ngetGlobalTensorClass().prototype.rsqrt = function() {\n  this.throwIfDisposed();\n  return rsqrt(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/selu.js\ngetGlobalTensorClass().prototype.selu = function() {\n  this.throwIfDisposed();\n  return selu(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/separable_conv2d.js\ngetGlobalTensorClass().prototype.separableConv2d = function(depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat) {\n  this.throwIfDisposed();\n  return separableConv2d(this, depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sigmoid.js\ngetGlobalTensorClass().prototype.sigmoid = function() {\n  this.throwIfDisposed();\n  return sigmoid(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sign.js\ngetGlobalTensorClass().prototype.sign = function() {\n  this.throwIfDisposed();\n  return sign(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sin.js\ngetGlobalTensorClass().prototype.sin = function() {\n  this.throwIfDisposed();\n  return sin(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sinh.js\ngetGlobalTensorClass().prototype.sinh = function() {\n  this.throwIfDisposed();\n  return sinh(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/slice.js\ngetGlobalTensorClass().prototype.slice = function(begin, size) {\n  this.throwIfDisposed();\n  return slice(this, begin, size);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/softmax.js\ngetGlobalTensorClass().prototype.softmax = function(dim) {\n  this.throwIfDisposed();\n  return softmax(this, dim);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/softplus.js\ngetGlobalTensorClass().prototype.softplus = function() {\n  this.throwIfDisposed();\n  return softplus(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/space_to_batch_nd.js\ngetGlobalTensorClass().prototype.spaceToBatchND = function(blockShape, paddings) {\n  this.throwIfDisposed();\n  return spaceToBatchND(this, blockShape, paddings);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/split.js\ngetGlobalTensorClass().prototype.split = function(numOrSizeSplits, axis) {\n  this.throwIfDisposed();\n  return split(this, numOrSizeSplits, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sqrt.js\ngetGlobalTensorClass().prototype.sqrt = function() {\n  this.throwIfDisposed();\n  return sqrt(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/square.js\ngetGlobalTensorClass().prototype.square = function() {\n  this.throwIfDisposed();\n  return square(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/squared_difference.js\ngetGlobalTensorClass().prototype.squaredDifference = function(b) {\n  this.throwIfDisposed();\n  return squaredDifference(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/squeeze.js\ngetGlobalTensorClass().prototype.squeeze = function(axis) {\n  this.throwIfDisposed();\n  return squeeze(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/stack.js\ngetGlobalTensorClass().prototype.stack = function(x, axis) {\n  this.throwIfDisposed();\n  const tensorsToBeStacked = x instanceof Tensor ? [this, x] : [this, ...x];\n  return stack(tensorsToBeStacked, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/step.js\ngetGlobalTensorClass().prototype.step = function(alpha) {\n  this.throwIfDisposed();\n  return step(this, alpha);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/strided_slice.js\ngetGlobalTensorClass().prototype.stridedSlice = function(begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {\n  this.throwIfDisposed();\n  return stridedSlice(this, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sub.js\ngetGlobalTensorClass().prototype.sub = function(b) {\n  this.throwIfDisposed();\n  return sub(this, b);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sum.js\ngetGlobalTensorClass().prototype.sum = function(axis, keepDims) {\n  this.throwIfDisposed();\n  return sum2(this, axis, keepDims);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tan.js\ngetGlobalTensorClass().prototype.tan = function() {\n  this.throwIfDisposed();\n  return tan(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tanh.js\ngetGlobalTensorClass().prototype.tanh = function() {\n  this.throwIfDisposed();\n  return tanh2(this);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tile.js\ngetGlobalTensorClass().prototype.tile = function(reps) {\n  this.throwIfDisposed();\n  return tile(this, reps);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_bool.js\ngetGlobalTensorClass().prototype.toBool = function() {\n  this.throwIfDisposed();\n  return cast(this, \"bool\");\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_float.js\ngetGlobalTensorClass().prototype.toFloat = function() {\n  this.throwIfDisposed();\n  return cast(this, \"float32\");\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_int.js\ngetGlobalTensorClass().prototype.toInt = function() {\n  this.throwIfDisposed();\n  return cast(this, \"int32\");\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/topk.js\ngetGlobalTensorClass().prototype.topk = function(k, sorted) {\n  this.throwIfDisposed();\n  return topk(this, k, sorted);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/transpose.js\ngetGlobalTensorClass().prototype.transpose = function(perm) {\n  this.throwIfDisposed();\n  return transpose(this, perm);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unique.js\ngetGlobalTensorClass().prototype.unique = function(axis) {\n  this.throwIfDisposed();\n  return unique(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unsorted_segment_sum.js\ngetGlobalTensorClass().prototype.unsortedSegmentSum = function(segmentIds, numSegments) {\n  this.throwIfDisposed();\n  return unsortedSegmentSum(this, segmentIds, numSegments);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unstack.js\ngetGlobalTensorClass().prototype.unstack = function(axis) {\n  this.throwIfDisposed();\n  return unstack(this, axis);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/where.js\ngetGlobalTensorClass().prototype.where = function(condition, x) {\n  this.throwIfDisposed();\n  return where(condition, this, x);\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/zeros_like.js\ngetGlobalTensorClass().prototype.zerosLike = function() {\n  this.throwIfDisposed();\n  return zerosLike(this);\n};\n\n// node_modules/@tensorflow/tfjs-layers/dist/errors.js\nvar AttributeError = class extends Error {\n  constructor(message) {\n    super(message);\n    Object.setPrototypeOf(this, AttributeError.prototype);\n  }\n};\nvar RuntimeError = class extends Error {\n  constructor(message) {\n    super(message);\n    Object.setPrototypeOf(this, RuntimeError.prototype);\n  }\n};\nvar ValueError = class extends Error {\n  constructor(message) {\n    super(message);\n    Object.setPrototypeOf(this, ValueError.prototype);\n  }\n};\nvar NotImplementedError = class extends Error {\n  constructor(message) {\n    super(message);\n    Object.setPrototypeOf(this, NotImplementedError.prototype);\n  }\n};\nvar AssertionError = class extends Error {\n  constructor(message) {\n    super(message);\n    Object.setPrototypeOf(this, AssertionError.prototype);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/executor_utils.js\nvar LruCache = class {\n  constructor(maxEntries) {\n    this.maxEntries = maxEntries || 100;\n    this.cache = /* @__PURE__ */ new Map();\n  }\n  get(key) {\n    let entry;\n    if (this.cache.has(key)) {\n      entry = this.cache.get(key);\n      this.cache.delete(key);\n      this.cache.set(key, entry);\n    }\n    return entry;\n  }\n  put(key, value) {\n    if (this.cache.has(key)) {\n      this.cache.delete(key);\n    } else if (this.cache.size >= this.maxEntries) {\n      const keyToDelete = this.cache.keys().next().value;\n      this.cache.delete(keyToDelete);\n    }\n    this.cache.set(key, value);\n  }\n  getMaxEntries() {\n    return this.maxEntries;\n  }\n  setMaxEntries(maxEntries) {\n    if (maxEntries < 0) {\n      throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${maxEntries}.`);\n    }\n    if (this.maxEntries > maxEntries) {\n      for (let i = 0; i < this.maxEntries - maxEntries; i++) {\n        const keyToDelete = this.cache.keys().next().value;\n        this.cache.delete(keyToDelete);\n      }\n    }\n    this.maxEntries = maxEntries;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/generic_utils.js\nfunction pyListRepeat(value, numValues) {\n  if (Array.isArray(value)) {\n    let newArray = [];\n    for (let i = 0; i < numValues; i++) {\n      newArray = newArray.concat(value);\n    }\n    return newArray;\n  } else {\n    const newArray = new Array(numValues);\n    newArray.fill(value);\n    return newArray;\n  }\n}\nfunction assert2(val, message) {\n  if (!val) {\n    throw new AssertionError(message);\n  }\n}\nfunction count(array2, refernce) {\n  let counter = 0;\n  for (const item of array2) {\n    if (item === refernce) {\n      counter++;\n    }\n  }\n  return counter;\n}\nfunction singletonOrArray(xs) {\n  if (xs.length === 1) {\n    return xs[0];\n  }\n  return xs;\n}\nfunction toList(x) {\n  if (Array.isArray(x)) {\n    return x;\n  }\n  return [x];\n}\nfunction toSnakeCase(name) {\n  const intermediate = name.replace(/(.)([A-Z][a-z0-9]+)/g, \"$1_$2\");\n  const insecure = intermediate.replace(/([a-z])([A-Z])/g, \"$1_$2\").toLowerCase();\n  if (insecure[0] !== \"_\") {\n    return insecure;\n  }\n  return \"private\" + insecure;\n}\nfunction toCamelCase(identifier) {\n  if (identifier.length <= 1) {\n    return identifier;\n  }\n  if (identifier.indexOf(\"_\") === -1) {\n    return identifier;\n  }\n  return identifier.replace(/[_]+(\\w|$)/g, (m, p1) => p1.toUpperCase());\n}\nvar _GLOBAL_CUSTOM_OBJECTS = {};\nfunction serializeKerasObject(instance) {\n  if (instance === null || instance === void 0) {\n    return null;\n  }\n  const dict = {};\n  dict[\"className\"] = instance.getClassName();\n  dict[\"config\"] = instance.getConfig();\n  return dict;\n}\nfunction convertNDArrayScalarsInConfig(config) {\n  if (config == null || typeof config !== \"object\") {\n    return;\n  } else if (Array.isArray(config)) {\n    config.forEach((configItem) => convertNDArrayScalarsInConfig(configItem));\n  } else {\n    const fields = Object.keys(config);\n    for (const field of fields) {\n      const value = config[field];\n      if (value != null && typeof value === \"object\") {\n        if (!Array.isArray(value) && value[\"type\"] === \"ndarray\" && typeof value[\"value\"] === \"number\") {\n          config[field] = value[\"value\"];\n        } else {\n          convertNDArrayScalarsInConfig(value);\n        }\n      }\n    }\n  }\n}\nfunction deserializeKerasObject(identifier, moduleObjects = {}, customObjects = {}, printableModuleName = \"object\", fastWeightInit = false) {\n  if (typeof identifier === \"string\") {\n    const functionName = identifier;\n    let fn;\n    if (functionName in customObjects) {\n      fn = customObjects[functionName];\n    } else if (functionName in _GLOBAL_CUSTOM_OBJECTS) {\n      fn = _GLOBAL_CUSTOM_OBJECTS[functionName];\n    } else {\n      fn = moduleObjects[functionName];\n      if (fn == null) {\n        throw new ValueError(`Unknown ${printableModuleName}: ${identifier}. This may be due to one of the following reasons:\n1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);\n      }\n    }\n    return fn;\n  } else {\n    const config = identifier;\n    if (config[\"className\"] == null || config[\"config\"] == null) {\n      throw new ValueError(`${printableModuleName}: Improper config format: ${JSON.stringify(config)}.\n'className' and 'config' must set.`);\n    }\n    const className = config[\"className\"];\n    let cls, fromConfig;\n    if (className in customObjects) {\n      [cls, fromConfig] = customObjects[className];\n    } else if (className in _GLOBAL_CUSTOM_OBJECTS) {\n      [cls, fromConfig] = _GLOBAL_CUSTOM_OBJECTS[\"className\"];\n    } else if (className in moduleObjects) {\n      [cls, fromConfig] = moduleObjects[className];\n    }\n    if (cls == null) {\n      throw new ValueError(`Unknown ${printableModuleName}: ${className}. This may be due to one of the following reasons:\n1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);\n    }\n    if (fromConfig != null) {\n      const customObjectsCombined = {};\n      for (const key of Object.keys(_GLOBAL_CUSTOM_OBJECTS)) {\n        customObjectsCombined[key] = _GLOBAL_CUSTOM_OBJECTS[key];\n      }\n      for (const key of Object.keys(customObjects)) {\n        customObjectsCombined[key] = customObjects[key];\n      }\n      const nestedConfig = config[\"config\"];\n      nestedConfig[\"customObjects\"] = customObjectsCombined;\n      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);\n      for (const key of Object.keys(customObjects)) {\n        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];\n      }\n      convertNDArrayScalarsInConfig(config[\"config\"]);\n      const returnObj = fromConfig(cls, config[\"config\"], customObjects, fastWeightInit);\n      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);\n      return returnObj;\n    } else {\n      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);\n      for (const key of Object.keys(customObjects)) {\n        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];\n      }\n      const returnObj = new cls(config[\"config\"]);\n      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);\n      return returnObj;\n    }\n  }\n}\nfunction numberCompare(a, b) {\n  return a < b ? -1 : a > b ? 1 : 0;\n}\nfunction reverseNumberCompare(a, b) {\n  return -1 * numberCompare(a, b);\n}\nfunction unique2(xs) {\n  if (xs == null) {\n    return xs;\n  }\n  const out = [];\n  for (const x of xs) {\n    if (out.indexOf(x) === -1) {\n      out.push(x);\n    }\n  }\n  return out;\n}\nfunction isObjectEmpty(obj) {\n  if (obj == null) {\n    throw new ValueError(`Invalid value in obj: ${JSON.stringify(obj)}`);\n  }\n  for (const key in obj) {\n    if (obj.hasOwnProperty(key)) {\n      return false;\n    }\n  }\n  return true;\n}\nfunction checkStringTypeUnionValue(values, label, value) {\n  if (value == null) {\n    return;\n  }\n  if (values.indexOf(value) < 0) {\n    throw new ValueError(`${value} is not a valid ${label}.  Valid values are ${values} or null/undefined.`);\n  }\n}\nfunction checkArrayTypeAndLength(x, expectedType, minLength = 0, maxLength = Infinity) {\n  assert2(minLength >= 0);\n  assert2(maxLength >= minLength);\n  return Array.isArray(x) && x.length >= minLength && x.length <= maxLength && x.every((e) => typeof e === expectedType);\n}\nfunction assertPositiveInteger(value, name) {\n  if (Array.isArray(value)) {\n    util_exports.assert(value.length > 0, () => `${name} is unexpectedly an empty array.`);\n    value.forEach((v, i) => assertPositiveInteger(v, `element ${i + 1} of ${name}`));\n  } else {\n    util_exports.assert(Number.isInteger(value) && value > 0, () => `Expected ${name} to be a positive integer, but got ${formatAsFriendlyString(value)}.`);\n  }\n}\nfunction formatAsFriendlyString(value) {\n  if (value === null) {\n    return \"null\";\n  } else if (Array.isArray(value)) {\n    return \"[\" + value.map((v) => formatAsFriendlyString(v)).join(\",\") + \"]\";\n  } else if (typeof value === \"string\") {\n    return `\"${value}\"`;\n  } else {\n    return `${value}`;\n  }\n}\nfunction debounce(f, waitMs, nowFunc) {\n  let lastTime = nowFunc != null ? nowFunc() : util_exports.now();\n  let lastResult;\n  const f2 = (...args) => {\n    const now2 = nowFunc != null ? nowFunc() : util_exports.now();\n    if (now2 - lastTime < waitMs) {\n      return lastResult;\n    }\n    lastTime = now2;\n    lastResult = f(...args);\n    return lastResult;\n  };\n  return f2;\n}\nfunction mapActivationToFusedKernel(activationName) {\n  if (activationName === \"relu\") {\n    return \"relu\";\n  }\n  if (activationName === \"linear\") {\n    return \"linear\";\n  }\n  if (activationName === \"elu\") {\n    return \"elu\";\n  }\n  return null;\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/backend/state.js\nvar _nextUniqueTensorId = 0;\nfunction getNextUniqueTensorId() {\n  return _nextUniqueTensorId++;\n}\nvar _uidPrefixes = {};\nfunction getUid(prefix = \"\") {\n  if (!(prefix in _uidPrefixes)) {\n    _uidPrefixes[prefix] = 0;\n  }\n  _uidPrefixes[prefix] += 1;\n  return prefix + _uidPrefixes[prefix].toString();\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/keras_format/common.js\nvar VALID_DATA_FORMAT_VALUES = [\"channelsFirst\", \"channelsLast\"];\nvar VALID_INTERPOLATION_FORMAT_VALUES = [\"nearest\", \"bilinear\"];\nvar VALID_PADDING_MODE_VALUES = [\"valid\", \"same\", \"causal\"];\nvar VALID_POOL_MODE_VALUES = [\"max\", \"avg\"];\nvar VALID_BIDIRECTIONAL_MERGE_MODES = [\"sum\", \"mul\", \"concat\", \"ave\"];\n\n// node_modules/@tensorflow/tfjs-layers/dist/common.js\nvar nameMap = /* @__PURE__ */ new Map();\nfunction checkDataFormat(value) {\n  checkStringTypeUnionValue(VALID_DATA_FORMAT_VALUES, \"DataFormat\", value);\n}\nfunction checkInterpolationFormat(value) {\n  checkStringTypeUnionValue(VALID_INTERPOLATION_FORMAT_VALUES, \"InterpolationFormat\", value);\n}\nfunction checkPaddingMode(value) {\n  checkStringTypeUnionValue(VALID_PADDING_MODE_VALUES, \"PaddingMode\", value);\n}\nfunction checkPoolMode(value) {\n  checkStringTypeUnionValue(VALID_POOL_MODE_VALUES, \"PoolMode\", value);\n}\nvar _nameScopeStack = [];\nvar _nameScopeDivider = \"/\";\nfunction nameScope(name, fn) {\n  _nameScopeStack.push(name);\n  try {\n    const val = fn();\n    _nameScopeStack.pop();\n    return val;\n  } catch (e) {\n    _nameScopeStack.pop();\n    throw e;\n  }\n}\nfunction currentNameScopePrefix() {\n  if (_nameScopeStack.length === 0) {\n    return \"\";\n  } else {\n    return _nameScopeStack.join(_nameScopeDivider) + _nameScopeDivider;\n  }\n}\nfunction getScopedTensorName(tensorName) {\n  if (!isValidTensorName(tensorName)) {\n    throw new Error(\"Not a valid tensor name: '\" + tensorName + \"'\");\n  }\n  return currentNameScopePrefix() + tensorName;\n}\nfunction getUniqueTensorName(scopedName) {\n  if (!isValidTensorName(scopedName)) {\n    throw new Error(\"Not a valid tensor name: '\" + scopedName + \"'\");\n  }\n  if (!nameMap.has(scopedName)) {\n    nameMap.set(scopedName, 0);\n  }\n  const index = nameMap.get(scopedName);\n  nameMap.set(scopedName, nameMap.get(scopedName) + 1);\n  if (index > 0) {\n    const result = `${scopedName}_${index}`;\n    nameMap.set(result, 1);\n    return result;\n  } else {\n    return scopedName;\n  }\n}\nvar tensorNameRegex = new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\\._\\/]*$/);\nfunction isValidTensorName(name) {\n  return !!name.match(tensorNameRegex);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/math_utils.js\nfunction isInteger(x) {\n  return x === parseInt(x.toString(), 10);\n}\nfunction arrayProd(array2, begin, end) {\n  if (begin == null) {\n    begin = 0;\n  }\n  if (end == null) {\n    end = array2.length;\n  }\n  let prod5 = 1;\n  for (let i = begin; i < end; ++i) {\n    prod5 *= array2[i];\n  }\n  return prod5;\n}\nfunction min2(array2) {\n  if (array2.length === 0) {\n    return Number.NaN;\n  }\n  let min6 = Number.POSITIVE_INFINITY;\n  for (let i = 0; i < array2.length; i++) {\n    const value = array2[i];\n    if (value < min6) {\n      min6 = value;\n    }\n  }\n  return min6;\n}\nfunction max2(array2) {\n  if (array2.length === 0) {\n    return Number.NaN;\n  }\n  let max6 = Number.NEGATIVE_INFINITY;\n  for (let i = 0; i < array2.length; i++) {\n    const value = array2[i];\n    if (value > max6) {\n      max6 = value;\n    }\n  }\n  return max6;\n}\nfunction range2(begin, end) {\n  if (end < begin) {\n    throw new ValueError(`end (${end}) < begin (${begin}) is forbidden.`);\n  }\n  const out = [];\n  for (let i = begin; i < end; ++i) {\n    out.push(i);\n  }\n  return out;\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/backend/common.js\nvar _epsilon;\nfunction epsilon() {\n  if (_epsilon == null) {\n    _epsilon = backend().epsilon();\n  }\n  return _epsilon;\n}\nfunction imageDataFormat() {\n  return \"channelsLast\";\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/backend/tfjs_backend.js\nfunction cast2(x, dtype) {\n  return cast(x, dtype);\n}\nfunction expandDims2(x, axis = -1) {\n  const outShape = x.shape.slice();\n  if (axis < 0) {\n    axis = outShape.length + axis + 1;\n  }\n  outShape.splice(axis, 0, 1);\n  return reshape(x, outShape);\n}\nfunction repeat(x, n) {\n  return tidy(() => {\n    if (x.shape.length !== 2) {\n      throw new ValueError(`repeat() expects a rank-2 tensor, but received a rank-${x.shape.length} tensor.`);\n    }\n    const y = expandDims2(x, 1);\n    return tile2(y, [1, n, 1]);\n  });\n}\nfunction flatten2(x) {\n  const newShape = [arrayProd(x.shape)];\n  return reshape(x, newShape);\n}\nfunction batchFlatten(x) {\n  if (x.rank <= 1) {\n    throw new ValueError(`batchFlatten requires a minimum rank of 2. Got rank: ${x.rank}.`);\n  }\n  const newShape = [x.shape[0], arrayProd(x.shape, 1)];\n  return reshape(x, newShape);\n}\nfunction sliceAlongFirstAxis(array2, start, size) {\n  return tidy(() => {\n    switch (array2.rank) {\n      case 1:\n        return slice1d(array2, start, size);\n      case 2:\n        return slice2d(array2, [start, 0], [size, array2.shape[1]]);\n      case 3:\n        return slice3d(array2, [start, 0, 0], [size, array2.shape[1], array2.shape[2]]);\n      case 4:\n        return slice4d(array2, [start, 0, 0, 0], [size, array2.shape[1], array2.shape[2], array2.shape[3]]);\n      case 5:\n        return slice(array2, [start, 0, 0, 0, 0], [\n          size,\n          array2.shape[1],\n          array2.shape[2],\n          array2.shape[3],\n          array2.shape[4]\n        ]);\n      case 6:\n        return slice(array2, [start, 0, 0, 0, 0, 0], [\n          size,\n          array2.shape[1],\n          array2.shape[2],\n          array2.shape[3],\n          array2.shape[4],\n          array2.shape[5]\n        ]);\n      default:\n        throw new ValueError(`sliceAlongFirstAxis() received an unsupported tensor rank: ${array2.rank}`);\n    }\n  });\n}\nfunction sliceAlongLastAxis(array2, start, size) {\n  return tidy(() => {\n    switch (array2.rank) {\n      case 1:\n        return slice1d(array2, start, size);\n      case 2:\n        return slice2d(array2, [0, start], [array2.shape[0], size]);\n      case 3:\n        return slice3d(array2, [0, 0, start], [array2.shape[0], array2.shape[1], size]);\n      case 4:\n        return slice4d(array2, [0, 0, 0, start], [array2.shape[0], array2.shape[1], array2.shape[2], size]);\n      default:\n        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);\n    }\n  });\n}\nfunction sliceAlongAxis(array2, start, size, axis) {\n  return tidy(() => {\n    switch (array2.rank) {\n      case 1:\n        return slice1d(array2, start, size);\n      case 2:\n        switch (axis) {\n          case 1:\n            return sliceAlongFirstAxis(array2, start, size);\n          case 2:\n            return sliceAlongLastAxis(array2, start, size);\n          default:\n            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);\n        }\n      case 3:\n        switch (axis) {\n          case 1:\n            return sliceAlongFirstAxis(array2, start, size);\n          case 2:\n            return slice3d(array2, [0, start, 0], [array2.shape[0], size, array2.shape[2]]);\n          case 3:\n            return sliceAlongLastAxis(array2, start, size);\n          default:\n            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);\n        }\n      case 4:\n        switch (axis) {\n          case 1:\n            return sliceAlongFirstAxis(array2, start, size);\n          case 2:\n            return slice4d(array2, [0, start, 0, 0], [array2.shape[0], size, array2.shape[2], array2.shape[3]]);\n          case 3:\n            return slice4d(array2, [0, 0, start, 0], [array2.shape[0], array2.shape[1], size, array2.shape[3]]);\n          case 4:\n            return sliceAlongLastAxis(array2, start, size);\n          default:\n            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);\n        }\n      default:\n        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);\n    }\n  });\n}\nfunction concatenate(tensors, axis = -1) {\n  let rank;\n  if (axis < 0) {\n    rank = tensors[0].rank;\n    if (rank !== 0) {\n      axis = rank;\n    } else {\n      axis = 0;\n    }\n  }\n  if (axis === tensors[0].rank) {\n    axis = -1;\n  }\n  return concat(tensors, axis);\n}\nfunction concatAlongFirstAxis(a, b) {\n  switch (a.rank) {\n    case 1:\n      return concat1d([a, b]);\n    case 2:\n      return concat2d([a, b], 0);\n    case 3:\n      return concat3d([a, b], 0);\n    case 4:\n      return concat4d([a, b], 0);\n    default:\n      throw new ValueError(`concatAlongFirstAxis() received an unsupported tensor rank: ${a.rank}`);\n  }\n}\nfunction tile2(x, n) {\n  if (!Array.isArray(n)) {\n    n = [n];\n  }\n  if (x.rank !== n.length) {\n    throw new ValueError(`The length of input n (${n.length}) does not match the number of dimensions in input x (${x.rank})`);\n  }\n  return tile(x, n);\n}\nfunction randomNormal2(shape, mean4 = 0, stddev = 1, dtype, seed) {\n  return randomNormal(shape, mean4, stddev, dtype, seed);\n}\nfunction dot2(a, b, activation2, bias) {\n  if (a.rank < 2 || b.rank < 2) {\n    throw new NotImplementedError(`dot requires both inputs to be rank >= 2 but got x shape = ${a.shape} and y shape = ${b.shape}`);\n  }\n  if (b.rank >= 3) {\n    const xLastDim = a.shape.slice(-1)[0];\n    const ySecondLastDim = b.shape.slice(-2)[0];\n    if (xLastDim !== ySecondLastDim) {\n      throw new NotImplementedError(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${a.shape} and  y shape = ${b.shape}`);\n    }\n  }\n  if (a.rank === 2 && b.rank === 2) {\n    const transposeA = false;\n    const transposeB = false;\n    return fused_ops_exports.matMul({\n      a,\n      b,\n      transposeA,\n      transposeB,\n      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,\n      activation: activation2\n    });\n  } else {\n    const aFirstDims = a.shape.slice();\n    const aLastDim = aFirstDims.pop();\n    a = reshape(a, [-1, aLastDim]);\n    const bShape = b.shape.slice();\n    const bLastDim = bShape.pop();\n    const ySecondLastDim = bShape.pop();\n    const yOtherDims = [...bShape, bLastDim];\n    const perm = Array.from({ length: b.rank }, (_, i) => {\n      if (i === 0) {\n        return b.rank - 2;\n      } else if (i <= b.rank - 2) {\n        return i - 1;\n      }\n      return i;\n    });\n    b = reshape(transpose(b, perm), [ySecondLastDim, -1]);\n    const outputShape = [...aFirstDims, ...yOtherDims];\n    const transposeA = false;\n    const transposeB = false;\n    return reshape(fused_ops_exports.matMul({\n      a,\n      b,\n      transposeA,\n      transposeB,\n      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,\n      activation: activation2\n    }), outputShape);\n  }\n}\nfunction gather2(reference, indices, axis) {\n  return tidy(() => {\n    if (Array.isArray(indices)) {\n      indices = tensor1d(indices, \"int32\");\n    } else {\n      indices = cast(indices, \"int32\");\n    }\n    return gather(reference, indices, axis);\n  });\n}\nfunction square2(x) {\n  return mul(x, x);\n}\nfunction reshapeBias(xRank, bias, dataFormat) {\n  const biasShape = bias.shape;\n  if (bias.rank !== 1 && bias.rank !== xRank) {\n    throw new ValueError(`Unexpected bias dimensions: ${bias.rank}; expected it to be 1 or ${xRank}`);\n  }\n  if (xRank === 5) {\n    if (dataFormat === \"channelsFirst\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, biasShape[0], 1, 1, 1]);\n      } else {\n        return reshape(bias, [1, biasShape[3], biasShape[0], biasShape[1], biasShape[2]]);\n      }\n    } else if (dataFormat === \"channelsLast\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, 1, 1, 1, biasShape[0]]);\n      } else {\n        return reshape(bias, [1].concat(biasShape));\n      }\n    }\n  } else if (xRank === 4) {\n    if (dataFormat === \"channelsFirst\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, biasShape[0], 1, 1]);\n      } else {\n        return reshape(bias, [1, biasShape[2], biasShape[0], biasShape[1]]);\n      }\n    } else if (dataFormat === \"channelsLast\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, 1, 1, biasShape[0]]);\n      } else {\n        return reshape(bias, [1].concat(biasShape));\n      }\n    }\n  } else if (xRank === 3) {\n    if (dataFormat === \"channelsFirst\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, biasShape[0], 1]);\n      } else {\n        return reshape(bias, [1, biasShape[1], biasShape[0]]);\n      }\n    } else if (dataFormat === \"channelsLast\") {\n      if (biasShape.length === 1) {\n        return reshape(bias, [1, 1, biasShape[0]]);\n      } else {\n        return reshape(bias, [1].concat(biasShape));\n      }\n    }\n  } else if (xRank < 3) {\n    return bias;\n  }\n  throw new ValueError(`Unsupported input rank by biasAdd: ${bias.rank}`);\n}\nfunction biasAdd(x, bias, dataFormat) {\n  return tidy(() => {\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    checkDataFormat(dataFormat);\n    return add2(x, reshapeBias(x.rank, bias, dataFormat));\n  });\n}\nfunction elu2(x, alpha = 1) {\n  if (alpha !== 1) {\n    throw new NotImplementedError(`Support for alpha values other than 1 (${alpha}) is not implemented yet.`);\n  }\n  return elu(x);\n}\nfunction softsign(x) {\n  return tidy(() => div(x, add2(abs(x), 1)));\n}\nfunction dropout2(x, level, noiseShape, seed) {\n  return tidy(() => dropout(x, level, noiseShape, seed));\n}\nfunction hardSigmoid(x) {\n  return tidy(() => {\n    const y = add2(0.5, mul(0.2, x));\n    return clipByValue(y, 0, 1);\n  });\n}\nfunction inTrainPhase(x, alt, training = false) {\n  return training ? x() : alt();\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/keras_format/initializer_config.js\nvar VALID_FAN_MODE_VALUES = [\"fanIn\", \"fanOut\", \"fanAvg\"];\nvar VALID_DISTRIBUTION_VALUES = [\"normal\", \"uniform\", \"truncatedNormal\"];\n\n// node_modules/@tensorflow/tfjs-layers/dist/initializers.js\nfunction checkFanMode(value) {\n  checkStringTypeUnionValue(VALID_FAN_MODE_VALUES, \"FanMode\", value);\n}\nfunction checkDistribution(value) {\n  checkStringTypeUnionValue(VALID_DISTRIBUTION_VALUES, \"Distribution\", value);\n}\nvar Initializer = class extends serialization_exports.Serializable {\n  fromConfigUsesCustomObjects() {\n    return false;\n  }\n  getConfig() {\n    return {};\n  }\n};\nvar Zeros = class extends Initializer {\n  apply(shape, dtype) {\n    return zeros(shape, dtype);\n  }\n};\nZeros.className = \"Zeros\";\nserialization_exports.registerClass(Zeros);\nvar Ones = class extends Initializer {\n  apply(shape, dtype) {\n    return ones2(shape, dtype);\n  }\n};\nOnes.className = \"Ones\";\nserialization_exports.registerClass(Ones);\nvar Constant = class extends Initializer {\n  constructor(args) {\n    super();\n    if (typeof args !== \"object\") {\n      throw new ValueError(`Expected argument of type ConstantConfig but got ${args}`);\n    }\n    if (args.value === void 0) {\n      throw new ValueError(`config must have value set but got ${args}`);\n    }\n    this.value = args.value;\n  }\n  apply(shape, dtype) {\n    return tidy(() => mul(scalar(this.value), ones2(shape, dtype)));\n  }\n  getConfig() {\n    return {\n      value: this.value\n    };\n  }\n};\nConstant.className = \"Constant\";\nserialization_exports.registerClass(Constant);\nvar RandomUniform = class extends Initializer {\n  constructor(args) {\n    super();\n    this.DEFAULT_MINVAL = -0.05;\n    this.DEFAULT_MAXVAL = 0.05;\n    this.minval = args.minval || this.DEFAULT_MINVAL;\n    this.maxval = args.maxval || this.DEFAULT_MAXVAL;\n    this.seed = args.seed;\n  }\n  apply(shape, dtype) {\n    return randomUniform(shape, this.minval, this.maxval, dtype);\n  }\n  getConfig() {\n    return { minval: this.minval, maxval: this.maxval, seed: this.seed };\n  }\n};\nRandomUniform.className = \"RandomUniform\";\nserialization_exports.registerClass(RandomUniform);\nvar RandomNormal = class extends Initializer {\n  constructor(args) {\n    super();\n    this.DEFAULT_MEAN = 0;\n    this.DEFAULT_STDDEV = 0.05;\n    this.mean = args.mean || this.DEFAULT_MEAN;\n    this.stddev = args.stddev || this.DEFAULT_STDDEV;\n    this.seed = args.seed;\n  }\n  apply(shape, dtype) {\n    dtype = dtype || \"float32\";\n    if (dtype !== \"float32\" && dtype !== \"int32\") {\n      throw new NotImplementedError(`randomNormal does not support dType ${dtype}.`);\n    }\n    return randomNormal2(shape, this.mean, this.stddev, dtype, this.seed);\n  }\n  getConfig() {\n    return { mean: this.mean, stddev: this.stddev, seed: this.seed };\n  }\n};\nRandomNormal.className = \"RandomNormal\";\nserialization_exports.registerClass(RandomNormal);\nvar TruncatedNormal = class extends Initializer {\n  constructor(args) {\n    super();\n    this.DEFAULT_MEAN = 0;\n    this.DEFAULT_STDDEV = 0.05;\n    this.mean = args.mean || this.DEFAULT_MEAN;\n    this.stddev = args.stddev || this.DEFAULT_STDDEV;\n    this.seed = args.seed;\n  }\n  apply(shape, dtype) {\n    dtype = dtype || \"float32\";\n    if (dtype !== \"float32\" && dtype !== \"int32\") {\n      throw new NotImplementedError(`truncatedNormal does not support dType ${dtype}.`);\n    }\n    return truncatedNormal(shape, this.mean, this.stddev, dtype, this.seed);\n  }\n  getConfig() {\n    return { mean: this.mean, stddev: this.stddev, seed: this.seed };\n  }\n};\nTruncatedNormal.className = \"TruncatedNormal\";\nserialization_exports.registerClass(TruncatedNormal);\nvar Identity2 = class extends Initializer {\n  constructor(args) {\n    super();\n    this.gain = args.gain != null ? args.gain : 1;\n  }\n  apply(shape, dtype) {\n    return tidy(() => {\n      if (shape.length !== 2 || shape[0] !== shape[1]) {\n        throw new ValueError(\"Identity matrix initializer can only be used for 2D square matrices.\");\n      } else {\n        return mul(this.gain, eye(shape[0]));\n      }\n    });\n  }\n  getConfig() {\n    return { gain: this.gain };\n  }\n};\nIdentity2.className = \"Identity\";\nserialization_exports.registerClass(Identity2);\nfunction computeFans(shape, dataFormat = \"channelsLast\") {\n  let fanIn;\n  let fanOut;\n  checkDataFormat(dataFormat);\n  if (shape.length === 2) {\n    fanIn = shape[0];\n    fanOut = shape[1];\n  } else if ([3, 4, 5].indexOf(shape.length) !== -1) {\n    if (dataFormat === \"channelsFirst\") {\n      const receptiveFieldSize = arrayProd(shape, 2);\n      fanIn = shape[1] * receptiveFieldSize;\n      fanOut = shape[0] * receptiveFieldSize;\n    } else if (dataFormat === \"channelsLast\") {\n      const receptiveFieldSize = arrayProd(shape, 0, shape.length - 2);\n      fanIn = shape[shape.length - 2] * receptiveFieldSize;\n      fanOut = shape[shape.length - 1] * receptiveFieldSize;\n    }\n  } else {\n    const shapeProd = arrayProd(shape);\n    fanIn = Math.sqrt(shapeProd);\n    fanOut = Math.sqrt(shapeProd);\n  }\n  return [fanIn, fanOut];\n}\nvar VarianceScaling = class extends Initializer {\n  constructor(args) {\n    super();\n    if (args.scale < 0) {\n      throw new ValueError(`scale must be a positive float. Got: ${args.scale}`);\n    }\n    this.scale = args.scale == null ? 1 : args.scale;\n    this.mode = args.mode == null ? \"fanIn\" : args.mode;\n    checkFanMode(this.mode);\n    this.distribution = args.distribution == null ? \"normal\" : args.distribution;\n    checkDistribution(this.distribution);\n    this.seed = args.seed;\n  }\n  apply(shape, dtype) {\n    const fans = computeFans(shape);\n    const fanIn = fans[0];\n    const fanOut = fans[1];\n    let scale2 = this.scale;\n    if (this.mode === \"fanIn\") {\n      scale2 /= Math.max(1, fanIn);\n    } else if (this.mode === \"fanOut\") {\n      scale2 /= Math.max(1, fanOut);\n    } else {\n      scale2 /= Math.max(1, (fanIn + fanOut) / 2);\n    }\n    if (this.distribution === \"normal\") {\n      const stddev = Math.sqrt(scale2);\n      dtype = dtype || \"float32\";\n      if (dtype !== \"float32\" && dtype !== \"int32\") {\n        throw new NotImplementedError(`${this.getClassName()} does not support dType ${dtype}.`);\n      }\n      return truncatedNormal(shape, 0, stddev, dtype, this.seed);\n    } else {\n      const limit = Math.sqrt(3 * scale2);\n      return randomUniform(shape, -limit, limit, dtype);\n    }\n  }\n  getConfig() {\n    return {\n      scale: this.scale,\n      mode: this.mode,\n      distribution: this.distribution,\n      seed: this.seed\n    };\n  }\n};\nVarianceScaling.className = \"VarianceScaling\";\nserialization_exports.registerClass(VarianceScaling);\nvar GlorotUniform = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 1,\n      mode: \"fanAvg\",\n      distribution: \"uniform\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nGlorotUniform.className = \"GlorotUniform\";\nserialization_exports.registerClass(GlorotUniform);\nvar GlorotNormal = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 1,\n      mode: \"fanAvg\",\n      distribution: \"normal\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nGlorotNormal.className = \"GlorotNormal\";\nserialization_exports.registerClass(GlorotNormal);\nvar HeNormal = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 2,\n      mode: \"fanIn\",\n      distribution: \"normal\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nHeNormal.className = \"HeNormal\";\nserialization_exports.registerClass(HeNormal);\nvar HeUniform = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 2,\n      mode: \"fanIn\",\n      distribution: \"uniform\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nHeUniform.className = \"HeUniform\";\nserialization_exports.registerClass(HeUniform);\nvar LeCunNormal = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 1,\n      mode: \"fanIn\",\n      distribution: \"normal\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nLeCunNormal.className = \"LeCunNormal\";\nserialization_exports.registerClass(LeCunNormal);\nvar LeCunUniform = class extends VarianceScaling {\n  constructor(args) {\n    super({\n      scale: 1,\n      mode: \"fanIn\",\n      distribution: \"uniform\",\n      seed: args == null ? null : args.seed\n    });\n  }\n  getClassName() {\n    return VarianceScaling.className;\n  }\n};\nLeCunUniform.className = \"LeCunNormal\";\nserialization_exports.registerClass(LeCunUniform);\nvar Orthogonal = class extends Initializer {\n  constructor(args) {\n    super();\n    this.DEFAULT_GAIN = 1;\n    this.gain = args.gain == null ? this.DEFAULT_GAIN : args.gain;\n    this.seed = args.seed;\n    if (this.seed != null) {\n      throw new NotImplementedError(\"Random seed is not implemented for Orthogonal Initializer yet.\");\n    }\n  }\n  apply(shape, dtype) {\n    return tidy(() => {\n      if (shape.length < 2) {\n        throw new NotImplementedError(\"Shape must be at least 2D.\");\n      }\n      if (shape[0] * shape[1] > 2e3) {\n        console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${shape[0] * shape[1]}) elements: Slowness may result.`);\n      }\n      const normalizedShape = shape[0] > shape[1] ? [shape[1], shape[0]] : shape;\n      const a = randomNormal2(normalizedShape, 0, 1, \"float32\");\n      let q = linalg.gramSchmidt(a);\n      if (shape[0] > shape[1]) {\n        q = transpose(q);\n      }\n      return mul(this.gain, q);\n    });\n  }\n  getConfig() {\n    return {\n      gain: this.gain,\n      seed: this.seed\n    };\n  }\n};\nOrthogonal.className = \"Orthogonal\";\nserialization_exports.registerClass(Orthogonal);\nvar INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {\n  \"constant\": \"Constant\",\n  \"glorotNormal\": \"GlorotNormal\",\n  \"glorotUniform\": \"GlorotUniform\",\n  \"heNormal\": \"HeNormal\",\n  \"heUniform\": \"HeUniform\",\n  \"identity\": \"Identity\",\n  \"leCunNormal\": \"LeCunNormal\",\n  \"leCunUniform\": \"LeCunUniform\",\n  \"ones\": \"Ones\",\n  \"orthogonal\": \"Orthogonal\",\n  \"randomNormal\": \"RandomNormal\",\n  \"randomUniform\": \"RandomUniform\",\n  \"truncatedNormal\": \"TruncatedNormal\",\n  \"varianceScaling\": \"VarianceScaling\",\n  \"zeros\": \"Zeros\"\n};\nfunction deserializeInitializer(config, customObjects = {}) {\n  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, \"initializer\");\n}\nfunction serializeInitializer(initializer) {\n  return serializeKerasObject(initializer);\n}\nfunction getInitializer(identifier) {\n  if (typeof identifier === \"string\") {\n    const className = identifier in INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;\n    if (className === \"GlorotNormal\") {\n      return new GlorotNormal();\n    } else if (className === \"GlorotUniform\") {\n      return new GlorotUniform();\n    } else if (className === \"HeNormal\") {\n      return new HeNormal();\n    } else if (className === \"HeUniform\") {\n      return new HeUniform();\n    } else if (className === \"LeCunNormal\") {\n      return new LeCunNormal();\n    } else if (className === \"LeCunUniform\") {\n      return new LeCunUniform();\n    } else {\n      const config = {};\n      config[\"className\"] = className;\n      config[\"config\"] = {};\n      return deserializeInitializer(config);\n    }\n  } else if (identifier instanceof Initializer) {\n    return identifier;\n  } else {\n    return deserializeInitializer(identifier);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/types_utils.js\nfunction isArrayOfShapes(x) {\n  return Array.isArray(x) && Array.isArray(x[0]);\n}\nfunction normalizeShapeList(x) {\n  if (x.length === 0) {\n    return [];\n  }\n  if (!Array.isArray(x[0])) {\n    return [x];\n  }\n  return x;\n}\nfunction getExactlyOneTensor(xs) {\n  let x;\n  if (Array.isArray(xs)) {\n    if (xs.length !== 1) {\n      throw new ValueError(`Expected Tensor length to be 1; got ${xs.length}`);\n    }\n    x = xs[0];\n  } else {\n    x = xs;\n  }\n  return x;\n}\nfunction getExactlyOneShape(shapes) {\n  if (Array.isArray(shapes) && Array.isArray(shapes[0])) {\n    if (shapes.length === 1) {\n      shapes = shapes;\n      return shapes[0];\n    } else {\n      throw new ValueError(`Expected exactly 1 Shape; got ${shapes.length}`);\n    }\n  } else {\n    return shapes;\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/variable_utils.js\nfunction countParamsInWeights(weights) {\n  let count2 = 0;\n  for (const weight of weights) {\n    if (weight.shape.length === 0) {\n      count2 += 1;\n    } else {\n      count2 += weight.shape.reduce((a, b) => a * b);\n    }\n  }\n  return count2;\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/variables.js\nvar DEFAULT_VARIABLE_NAME_PREFIX = \"Variable\";\nvar LayerVariable = class {\n  constructor(val, dtype = \"float32\", name = DEFAULT_VARIABLE_NAME_PREFIX, trainable = true, constraint = null) {\n    this.dtype = dtype == null ? \"float32\" : dtype;\n    this.shape = val.shape;\n    this.id = getNextUniqueTensorId();\n    name = name == null ? DEFAULT_VARIABLE_NAME_PREFIX : name;\n    this.originalName = getScopedTensorName(name);\n    this.name = getUniqueTensorName(this.originalName);\n    this.trainable_ = trainable;\n    this.constraint = constraint;\n    this.val = variable(val, this.trainable_, this.name, this.dtype);\n  }\n  read() {\n    this.assertNotDisposed();\n    return this.val;\n  }\n  write(newVal) {\n    this.assertNotDisposed();\n    checkShapesMatch(this.val, newVal);\n    if (this.val.id !== newVal.id) {\n      this.val.assign(newVal);\n      if (this.constraint != null) {\n        this.val.assign(this.constraint.apply(this.val));\n      }\n    }\n    return this;\n  }\n  dispose() {\n    this.assertNotDisposed();\n    this.val.dispose();\n  }\n  assertNotDisposed() {\n    if (this.val.isDisposed) {\n      throw new Error(`LayersVariable ${this.name} is already disposed.`);\n    }\n  }\n  get trainable() {\n    return this.trainable_;\n  }\n  set trainable(trainable) {\n    this.trainable_ = trainable;\n    this.val.trainable = trainable;\n  }\n};\nfunction checkShapesMatch(x, y) {\n  if (x.shape.toString() !== y.shape.toString()) {\n    throw new Error(\"Shape mismatch: \" + JSON.stringify(x.shape) + \" vs. \" + JSON.stringify(y.shape));\n  }\n}\nfunction batchGetValue(xs) {\n  return xs.map((x) => x.read());\n}\nfunction batchSetValue(variablesAndValues) {\n  variablesAndValues.forEach((variableAndValue) => {\n    const variable2 = variableAndValue[0];\n    variable2.write(variableAndValue[1]);\n  });\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/topology.js\nvar InputSpec = class {\n  constructor(args) {\n    this.dtype = args.dtype;\n    this.shape = args.shape;\n    if (args.shape != null) {\n      this.ndim = args.shape.length;\n    } else {\n      this.ndim = args.ndim;\n    }\n    this.maxNDim = args.maxNDim;\n    this.minNDim = args.minNDim;\n    this.axes = args.axes || {};\n  }\n};\nvar SymbolicTensor = class {\n  constructor(dtype, shape, sourceLayer, inputs, callArgs, name, outputTensorIndex) {\n    this.dtype = dtype;\n    this.shape = shape;\n    this.sourceLayer = sourceLayer;\n    this.inputs = inputs;\n    this.callArgs = callArgs;\n    this.outputTensorIndex = outputTensorIndex;\n    this.id = getNextUniqueTensorId();\n    if (name != null) {\n      this.originalName = getScopedTensorName(name);\n      this.name = getUniqueTensorName(this.originalName);\n    }\n    this.rank = shape.length;\n  }\n};\nvar _nextNodeID = 0;\nvar Node = class {\n  constructor(args, callArgs) {\n    this.callArgs = callArgs;\n    this.id = _nextNodeID++;\n    this.outboundLayer = args.outboundLayer;\n    this.inboundLayers = args.inboundLayers;\n    this.nodeIndices = args.nodeIndices;\n    this.tensorIndices = args.tensorIndices;\n    this.inputTensors = args.inputTensors;\n    this.outputTensors = args.outputTensors;\n    this.inputMasks = args.inputMasks;\n    this.outputMasks = args.outputMasks;\n    this.inputShapes = args.inputShapes;\n    this.outputShapes = args.outputShapes;\n    for (const layer of args.inboundLayers) {\n      if (layer != null) {\n        layer.outboundNodes.push(this);\n      }\n    }\n    args.outboundLayer.inboundNodes.push(this);\n  }\n  getConfig() {\n    const inboundNames = [];\n    for (const layer of this.inboundLayers) {\n      if (layer != null) {\n        inboundNames.push(layer.name);\n      } else {\n        inboundNames.push(null);\n      }\n    }\n    return {\n      outboundLayer: this.outboundLayer ? this.outboundLayer.name : null,\n      inboundLayers: inboundNames,\n      nodeIndices: this.nodeIndices,\n      tensorIndices: this.tensorIndices\n    };\n  }\n};\nvar _nextLayerID = 0;\nvar Layer = class extends serialization_exports.Serializable {\n  constructor(args = {}) {\n    super();\n    this._callHook = null;\n    this._addedWeightNames = [];\n    this._stateful = false;\n    this.id = _nextLayerID++;\n    this.activityRegularizer = null;\n    this.inputSpec = null;\n    this.supportsMasking = false;\n    this._trainableWeights = [];\n    this._nonTrainableWeights = [];\n    this._losses = [];\n    this._updates = [];\n    this._built = false;\n    this.inboundNodes = [];\n    this.outboundNodes = [];\n    let name = args.name;\n    if (!name) {\n      const prefix = this.getClassName();\n      name = toSnakeCase(prefix) + \"_\" + getUid(prefix);\n    }\n    this.name = name;\n    this.trainable_ = args.trainable == null ? true : args.trainable;\n    if (args.inputShape != null || args.batchInputShape != null) {\n      let batchInputShape;\n      if (args.batchInputShape != null) {\n        batchInputShape = args.batchInputShape;\n      } else if (args.inputShape != null) {\n        let batchSize = null;\n        if (args.batchSize != null) {\n          batchSize = args.batchSize;\n        }\n        batchInputShape = [batchSize].concat(args.inputShape);\n      }\n      this.batchInputShape = batchInputShape;\n      let dtype = args.dtype;\n      if (dtype == null) {\n        dtype = args.inputDType;\n      }\n      if (dtype == null) {\n        dtype = \"float32\";\n      }\n      this.dtype = dtype;\n    }\n    if (args.weights != null) {\n      this.initialWeights = args.weights;\n    } else {\n      this.initialWeights = null;\n    }\n    this._refCount = null;\n    this.fastWeightInitDuringBuild = false;\n  }\n  static nodeKey(layer, nodeIndex) {\n    return layer.name + \"_ib-\" + nodeIndex.toString();\n  }\n  getNodeAtIndex(nodeIndex, attrName) {\n    if (this.inboundNodes.length === 0) {\n      throw new RuntimeError(`The layer has never been called and thus has no defined ${attrName}.`);\n    }\n    if (this.inboundNodes.length <= nodeIndex) {\n      throw new ValueError(`Asked to get ${attrName} at node ${nodeIndex}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);\n    }\n    return this.inboundNodes[nodeIndex];\n  }\n  getInputAt(nodeIndex) {\n    return singletonOrArray(this.getNodeAtIndex(nodeIndex, \"input\").inputTensors);\n  }\n  getOutputAt(nodeIndex) {\n    return singletonOrArray(this.getNodeAtIndex(nodeIndex, \"output\").outputTensors);\n  }\n  get input() {\n    if (this.inboundNodes.length > 1) {\n      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of \"layer input\" is ill-defined. Use \\`getInputAt(nodeIndex)\\` instead.`);\n    } else if (this.inboundNodes.length === 0) {\n      throw new AttributeError(`Layer ${this.name} is not connected, no input to return.`);\n    }\n    return singletonOrArray(this.getNodeAtIndex(0, \"input\").inputTensors);\n  }\n  get output() {\n    if (this.inboundNodes.length === 0) {\n      throw new AttributeError(`Layer ${this.name} has no inbound nodes.`);\n    }\n    if (this.inboundNodes.length > 1) {\n      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of \"layer output\" is ill-defined. Use \\`getOutputAt(nodeIndex)\\` instead.`);\n    }\n    return singletonOrArray(this.getNodeAtIndex(0, \"output\").outputTensors);\n  }\n  get losses() {\n    return this._losses;\n  }\n  calculateLosses() {\n    return this.losses.map((lossFn) => lossFn());\n  }\n  get updates() {\n    return this._updates;\n  }\n  get built() {\n    return this._built;\n  }\n  set built(built) {\n    this._built = built;\n  }\n  get trainable() {\n    return this.trainable_;\n  }\n  set trainable(trainable) {\n    this._trainableWeights.forEach((w) => w.trainable = trainable);\n    this.trainable_ = trainable;\n  }\n  get trainableWeights() {\n    if (this.trainable_) {\n      return this._trainableWeights.filter((w) => w.trainable);\n    } else {\n      return [];\n    }\n  }\n  set trainableWeights(weights) {\n    this._trainableWeights = weights;\n  }\n  get nonTrainableWeights() {\n    if (this.trainable) {\n      return this._trainableWeights.filter((w) => !w.trainable).concat(this._nonTrainableWeights);\n    } else {\n      return this._trainableWeights.concat(this._nonTrainableWeights);\n    }\n  }\n  set nonTrainableWeights(weights) {\n    this._nonTrainableWeights = weights;\n  }\n  get weights() {\n    return this.trainableWeights.concat(this.nonTrainableWeights);\n  }\n  get stateful() {\n    return this._stateful;\n  }\n  resetStates() {\n    if (!this.stateful) {\n      throw new Error(\"Cannot call the resetStates() method of a non-stateful Layer object.\");\n    }\n  }\n  assertInputCompatibility(inputs) {\n    inputs = toList(inputs);\n    if (this.inputSpec == null || this.inputSpec.length === 0) {\n      return;\n    }\n    const inputSpec = toList(this.inputSpec);\n    if (inputs.length !== inputSpec.length) {\n      throw new ValueError(`Layer ${this.name} expects ${inputSpec.length} inputs, but it received ${inputs.length} input tensors. Input received: ${inputs}`);\n    }\n    for (let inputIndex = 0; inputIndex < inputs.length; inputIndex++) {\n      const x = inputs[inputIndex];\n      const spec = inputSpec[inputIndex];\n      if (spec == null) {\n        continue;\n      }\n      const ndim = x.rank;\n      if (spec.ndim != null) {\n        if (ndim !== spec.ndim) {\n          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected ndim=${spec.ndim}, found ndim=${ndim}`);\n        }\n      }\n      if (spec.maxNDim != null) {\n        if (ndim > spec.maxNDim) {\n          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected max_ndim=${spec.maxNDim}, found ndim=${ndim}`);\n        }\n      }\n      if (spec.minNDim != null) {\n        if (ndim < spec.minNDim) {\n          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected min_ndim=${spec.minNDim}, found ndim=${ndim}.`);\n        }\n      }\n      if (spec.dtype != null) {\n        if (x.dtype !== spec.dtype) {\n          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name} : expected dtype=${spec.dtype}, found dtype=${x.dtype}.`);\n        }\n      }\n      if (spec.axes) {\n        const xShape = x.shape;\n        for (const key in spec.axes) {\n          const axis = Number(key);\n          const value = spec.axes[key];\n          const xShapeAtAxis = axis >= 0 ? xShape[axis] : xShape[xShape.length + axis];\n          if (value != null && [value, null].indexOf(xShapeAtAxis) === -1) {\n            throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected axis ${axis} of input shape to have value ${value} but got shape ${xShape}.`);\n          }\n        }\n      }\n      if (spec.shape != null) {\n        for (let i = 0; i < spec.shape.length; ++i) {\n          const specDim = spec.shape[i];\n          const dim = x.shape[i];\n          if (specDim != null && dim != null) {\n            if (specDim !== dim) {\n              throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected shape=${spec.shape}, found shape=${x.shape}.`);\n            }\n          }\n        }\n      }\n    }\n  }\n  call(inputs, kwargs) {\n    return inputs;\n  }\n  invokeCallHook(inputs, kwargs) {\n    if (this._callHook != null) {\n      this._callHook(inputs, kwargs);\n    }\n  }\n  setCallHook(callHook) {\n    this._callHook = callHook;\n  }\n  clearCallHook() {\n    this._callHook = null;\n  }\n  apply(inputs, kwargs) {\n    kwargs = kwargs || {};\n    this.assertNotDisposed();\n    const inputsList = toList(inputs);\n    let allAreSymbolic = true;\n    for (const input2 of inputsList) {\n      if (!(input2 instanceof SymbolicTensor)) {\n        allAreSymbolic = false;\n        break;\n      }\n    }\n    let noneAreSymbolic = true;\n    for (const input2 of inputsList) {\n      if (input2 instanceof SymbolicTensor) {\n        noneAreSymbolic = false;\n        break;\n      }\n    }\n    if (allAreSymbolic === noneAreSymbolic) {\n      throw new ValueError(\"Arguments to apply() must be all SymbolicTensors or all Tensors\");\n    }\n    return nameScope(this.name, () => {\n      if (!this.built) {\n        this.assertInputCompatibility(inputs);\n        const inputShapes = [];\n        for (const xElem of toList(inputs)) {\n          inputShapes.push(xElem.shape);\n        }\n        this.build(singletonOrArray(inputShapes));\n        this.built = true;\n        if (this.initialWeights) {\n          this.setWeights(this.initialWeights);\n        }\n        if (this._refCount === null && noneAreSymbolic) {\n          this._refCount = 1;\n        }\n      }\n      this.assertInputCompatibility(inputs);\n      if (noneAreSymbolic) {\n        let output = this.call(inputs, kwargs);\n        const outputList = toList(output);\n        const outputListCopy = [];\n        for (let x of outputList) {\n          if (inputsList.indexOf(x) !== -1) {\n            x = x.clone();\n          }\n          outputListCopy.push(x);\n        }\n        output = singletonOrArray(outputListCopy);\n        if (this.activityRegularizer != null) {\n          throw new NotImplementedError(\"Layer invocation in the presence of activity regularizer(s) is not supported yet.\");\n        }\n        return output;\n      } else {\n        const inputShape = collectInputShape(inputs);\n        const outputShape = this.computeOutputShape(inputShape);\n        let output;\n        const outputDType = guessOutputDType(inputs);\n        this.warnOnIncompatibleInputShape(Array.isArray(inputs) ? inputShape[0] : inputShape);\n        if (outputShape != null && outputShape.length > 0 && Array.isArray(outputShape[0])) {\n          output = outputShape.map((shape, index) => new SymbolicTensor(outputDType, shape, this, toList(inputs), kwargs, this.name, index));\n        } else {\n          output = new SymbolicTensor(outputDType, outputShape, this, toList(inputs), kwargs, this.name);\n        }\n        this.addInboundNode(inputs, output, null, null, inputShape, outputShape, kwargs);\n        this._refCount++;\n        if (this.activityRegularizer != null) {\n          throw new NotImplementedError(\"Layer invocation in the presence of activity regularizer(s) is not supported yet.\");\n        }\n        return output;\n      }\n    });\n  }\n  warnOnIncompatibleInputShape(inputShape) {\n    if (this.batchInputShape == null) {\n      return;\n    } else if (inputShape.length !== this.batchInputShape.length) {\n      console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(inputShape)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);\n    } else {\n      let dimMismatch = false;\n      this.batchInputShape.forEach((dimension, i) => {\n        if (dimension != null && inputShape[i] != null && inputShape[i] !== dimension) {\n          dimMismatch = true;\n        }\n      });\n      if (dimMismatch) {\n        console.warn(`The shape of the input tensor (${JSON.stringify(inputShape)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`);\n      }\n    }\n  }\n  get outputShape() {\n    if (this.inboundNodes == null || this.inboundNodes.length === 0) {\n      throw new AttributeError(`The layer ${this.name} has never been called and thus has no defined output shape.`);\n    }\n    const allOutputShapes = [];\n    for (const node of this.inboundNodes) {\n      const shapeString = JSON.stringify(node.outputShapes);\n      if (allOutputShapes.indexOf(shapeString) === -1) {\n        allOutputShapes.push(shapeString);\n      }\n    }\n    if (allOutputShapes.length === 1) {\n      const outputShapes = this.inboundNodes[0].outputShapes;\n      if (Array.isArray(outputShapes) && Array.isArray(outputShapes[0]) && outputShapes.length === 1) {\n        return outputShapes[0];\n      } else {\n        return outputShapes;\n      }\n    } else {\n      throw new AttributeError(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of \"output shape\" is ill-defined for the layer.`);\n    }\n  }\n  countParams() {\n    if (!this.built) {\n      throw new RuntimeError(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);\n    }\n    return countParamsInWeights(this.weights);\n  }\n  build(inputShape) {\n    this.built = true;\n  }\n  getWeights(trainableOnly = false) {\n    return batchGetValue(trainableOnly ? this.trainableWeights : this.weights);\n  }\n  setWeights(weights) {\n    tidy(() => {\n      const params = this.weights;\n      if (params.length !== weights.length) {\n        throw new ValueError(`You called setWeights(weights) on layer \"${this.name}\" with a weight list of length ${weights.length}, but the layer was expecting ${params.length} weights. Provided weights: ${weights}...`);\n      }\n      if (params.length === 0) {\n        return;\n      }\n      const weightValueTuples = [];\n      const paramValues = batchGetValue(params);\n      for (let i = 0; i < paramValues.length; ++i) {\n        const pv = paramValues[i];\n        const p2 = params[i];\n        const w = weights[i];\n        if (!util_exports.arraysEqual(pv.shape, w.shape)) {\n          throw new ValueError(`Layer weight shape ${pv.shape} not compatible with provided weight shape ${w.shape}`);\n        }\n        weightValueTuples.push([p2, w]);\n      }\n      batchSetValue(weightValueTuples);\n    });\n  }\n  addWeight(name, shape, dtype, initializer, regularizer, trainable, constraint, getInitializerFunc) {\n    if (this._addedWeightNames.indexOf(name) !== -1) {\n      throw new ValueError(`Duplicate weight name ${name} for layer ${this.name}`);\n    }\n    this._addedWeightNames.push(name);\n    if (dtype == null) {\n      dtype = \"float32\";\n    }\n    if (this.fastWeightInitDuringBuild) {\n      initializer = getInitializerFunc != null ? getInitializerFunc() : getInitializer(\"zeros\");\n    }\n    const initValue = initializer.apply(shape, dtype);\n    const weight = new LayerVariable(initValue, dtype, name, trainable, constraint);\n    initValue.dispose();\n    if (regularizer != null) {\n      this.addLoss(() => regularizer.apply(weight.read()));\n    }\n    if (trainable == null) {\n      trainable = true;\n    }\n    if (trainable) {\n      this._trainableWeights.push(weight);\n    } else {\n      this._nonTrainableWeights.push(weight);\n    }\n    return weight;\n  }\n  setFastWeightInitDuringBuild(value) {\n    this.fastWeightInitDuringBuild = value;\n  }\n  addLoss(losses2) {\n    if (losses2 == null || Array.isArray(losses2) && losses2.length === 0) {\n      return;\n    }\n    losses2 = toList(losses2);\n    if (this._losses !== void 0 && this._losses !== null) {\n      this.losses.push(...losses2);\n    }\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  computeMask(inputs, mask) {\n    if (!this.supportsMasking) {\n      if (mask != null) {\n        if (Array.isArray(mask)) {\n          mask.forEach((maskElement) => {\n            if (maskElement != null) {\n              throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);\n            }\n          });\n        } else {\n          throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);\n        }\n      }\n      return null;\n    }\n    return mask;\n  }\n  addInboundNode(inputTensors, outputTensors, inputMasks, outputMasks, inputShapes, outputShapes, kwargs = null) {\n    const inputTensorList = toList(inputTensors);\n    outputTensors = toList(outputTensors);\n    inputMasks = toList(inputMasks);\n    outputMasks = toList(outputMasks);\n    inputShapes = normalizeShapeList(inputShapes);\n    outputShapes = normalizeShapeList(outputShapes);\n    const inboundLayers = [];\n    const nodeIndices = [];\n    const tensorIndices = [];\n    for (const x of inputTensorList) {\n      inboundLayers.push(x.sourceLayer);\n      nodeIndices.push(x.nodeIndex);\n      tensorIndices.push(x.tensorIndex);\n    }\n    new Node({\n      outboundLayer: this,\n      inboundLayers,\n      nodeIndices,\n      tensorIndices,\n      inputTensors: inputTensorList,\n      outputTensors,\n      inputMasks,\n      outputMasks,\n      inputShapes,\n      outputShapes\n    }, kwargs);\n    for (let i = 0; i < outputTensors.length; i++) {\n      outputTensors[i].sourceLayer = this;\n      outputTensors[i].nodeIndex = this.inboundNodes.length - 1;\n      outputTensors[i].tensorIndex = i;\n    }\n  }\n  getConfig() {\n    const config = { name: this.name, trainable: this.trainable };\n    if (this.batchInputShape != null) {\n      config[\"batchInputShape\"] = this.batchInputShape;\n    }\n    if (this.dtype != null) {\n      config[\"dtype\"] = this.dtype;\n    }\n    return config;\n  }\n  disposeWeights() {\n    this.weights.forEach((weight) => weight.dispose());\n    return this.weights.length;\n  }\n  assertNotDisposed() {\n    if (this._refCount === 0) {\n      throw new Error(`Layer '${this.name}' is already disposed.`);\n    }\n  }\n  dispose() {\n    if (!this.built) {\n      throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);\n    }\n    if (this._refCount === null) {\n      throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);\n    }\n    this.assertNotDisposed();\n    let numDisposedVariables = 0;\n    if (--this._refCount === 0) {\n      numDisposedVariables = this.disposeWeights();\n    }\n    return { refCountAfterDispose: this._refCount, numDisposedVariables };\n  }\n};\nfunction collectInputShape(inputTensors) {\n  inputTensors = toList(inputTensors);\n  const shapes = [];\n  for (const x of inputTensors) {\n    shapes.push(x.shape);\n  }\n  return singletonOrArray(shapes);\n}\nfunction guessOutputDType(inputTensors) {\n  return \"float32\";\n}\nfunction getSourceInputs(tensor2, layer, nodeIndex) {\n  if (layer == null || nodeIndex != null && nodeIndex > 0) {\n    layer = tensor2.sourceLayer;\n    nodeIndex = tensor2.nodeIndex;\n  }\n  if (layer.inboundNodes.length === 0) {\n    return [tensor2];\n  } else {\n    const node = layer.inboundNodes[nodeIndex];\n    if (node.inboundLayers.length === 0) {\n      return node.inputTensors;\n    } else {\n      const sourceTensors = [];\n      for (let i = 0; i < node.inboundLayers.length; i++) {\n        const x = node.inputTensors[i];\n        const layer2 = node.inboundLayers[i];\n        const nodeIndex2 = node.nodeIndices[i];\n        const previousSources = getSourceInputs(x, layer2, nodeIndex2);\n        for (const x2 of previousSources) {\n          if (sourceTensors.indexOf(x2) === -1) {\n            sourceTensors.push(x2);\n          }\n        }\n      }\n      return sourceTensors;\n    }\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/input_layer.js\nvar InputLayer = class extends Layer {\n  constructor(args) {\n    super({\n      dtype: args.dtype,\n      name: args.name != null ? args.name : getUid(\"input\").toString()\n    });\n    if (args.batchSize == null) {\n      args.batchSize = null;\n    }\n    if (args.sparse == null) {\n      args.sparse = false;\n    }\n    this.trainable = false;\n    this.built = true;\n    this.sparse = args.sparse;\n    if (args.inputShape != null && args.batchInputShape != null) {\n      throw new ValueError(\"Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.\");\n    }\n    let batchInputShape = args.batchInputShape;\n    if (batchInputShape == null) {\n      if (args.inputShape == null) {\n        throw new ValueError(\"An InputLayer should be passed either a `batchInputShape` or an `inputShape`.\");\n      } else {\n        batchInputShape = [args.batchSize].concat(args.inputShape);\n      }\n    } else {\n      if (args.batchSize != null) {\n        throw new ValueError(\"Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.\");\n      }\n    }\n    const dtype = args.dtype || \"float32\";\n    this.batchInputShape = batchInputShape;\n    this.dtype = dtype;\n    this.inputSpec = [{ shape: batchInputShape }];\n    const inputTensor = new SymbolicTensor(this.dtype, this.batchInputShape, this, [], {}, this.name);\n    inputTensor.nodeIndex = 0;\n    inputTensor.tensorIndex = 0;\n    new Node({\n      outboundLayer: this,\n      inboundLayers: [],\n      nodeIndices: [],\n      tensorIndices: [],\n      inputTensors: [inputTensor],\n      outputTensors: [inputTensor],\n      inputMasks: [null],\n      outputMasks: [null],\n      inputShapes: [batchInputShape],\n      outputShapes: [batchInputShape]\n    });\n  }\n  apply(inputs, kwargs) {\n    throw new ValueError(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`);\n  }\n  dispose() {\n    return { refCountAfterDispose: this._refCount, numDisposedVariables: 0 };\n  }\n  getConfig() {\n    return {\n      batchInputShape: this.batchInputShape,\n      dtype: this.dtype,\n      sparse: this.sparse,\n      name: this.name\n    };\n  }\n};\nInputLayer.className = \"InputLayer\";\nserialization_exports.registerClass(InputLayer);\nfunction Input(config) {\n  if (config.batchShape == null && config.shape == null) {\n    throw new Error(\"Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.\");\n  }\n  if (config.batchShape != null && config.shape != null) {\n    throw new ValueError(\"Please provide either a `shape` or `batchShape` argument to Input, but not both.\");\n  }\n  let batchShape = config.batchShape;\n  if (config.shape != null && batchShape == null) {\n    batchShape = [null].concat(config.shape);\n  }\n  let dtype = config.dtype;\n  if (dtype == null) {\n    dtype = \"float32\";\n  }\n  const inputLayer2 = new InputLayer({\n    batchInputShape: batchShape,\n    name: config.name,\n    dtype,\n    sparse: config.sparse\n  });\n  const outputs = inputLayer2.inboundNodes[0].outputTensors;\n  return outputs[0];\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/executor.js\nfunction assertFeedCompatibility(key, val) {\n  if (key.dtype == null || key.dtype === val.dtype) {\n    return val;\n  }\n  try {\n    return cast(val, key.dtype);\n  } catch (err) {\n    throw new ValueError(`The dtype of the feed (${val.dtype}) can not be cast to the dtype of the key '${key.name}' (${key.dtype}).`);\n  }\n}\nvar FeedDict = class {\n  constructor(feeds) {\n    this.id2Value = {};\n    this.id2Mask = {};\n    this.name2Id = {};\n    if (feeds instanceof FeedDict) {\n      for (const id in feeds.id2Value) {\n        this.id2Value[id] = feeds.id2Value[id];\n        if (id in feeds.id2Mask) {\n          this.id2Mask[id] = feeds.id2Mask[id];\n        }\n      }\n    } else {\n      if (feeds == null) {\n        return;\n      }\n      for (const feed of feeds) {\n        this.add(feed.key, feed.value);\n      }\n    }\n  }\n  add(key, value, mask) {\n    if (this.id2Value[key.id] == null) {\n      this.id2Value[key.id] = assertFeedCompatibility(key, value);\n      this.name2Id[key.name] = key.id;\n      if (mask != null) {\n        this.id2Mask[key.id] = mask;\n      }\n    } else {\n      throw new ValueError(`Duplicate key: name=${key.name}, id=${key.id}`);\n    }\n    return this;\n  }\n  addFeed(feed) {\n    this.add(feed.key, feed.value);\n  }\n  hasKey(key) {\n    return this.id2Value[key.id] != null;\n  }\n  names() {\n    return Object.keys(this.name2Id);\n  }\n  getValue(key) {\n    if (key instanceof SymbolicTensor) {\n      if (this.id2Value[key.id] == null) {\n        throw new ValueError(`Nonexistent key: ${key.name}`);\n      } else {\n        return this.id2Value[key.id];\n      }\n    } else {\n      const id = this.name2Id[key];\n      if (id == null) {\n        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);\n      }\n      return this.id2Value[id];\n    }\n  }\n  getMask(key) {\n    if (key instanceof SymbolicTensor) {\n      if (this.id2Value[key.id] == null) {\n        throw new ValueError(`Nonexistent key: ${key.name}`);\n      } else {\n        return this.id2Mask[key.id];\n      }\n    } else {\n      const id = this.name2Id[key];\n      if (id == null) {\n        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);\n      }\n      return this.id2Mask[id];\n    }\n  }\n  disposeMasks() {\n    if (this.id2Mask != null) {\n      dispose(this.id2Mask);\n    }\n  }\n};\nvar cachedSorted = new LruCache();\nvar cachedRecipientCounts = new LruCache();\nfunction updateCacheMaxEntries(maxEntries) {\n  if (cachedSorted != null) {\n    cachedSorted.setMaxEntries(maxEntries);\n  }\n  if (cachedRecipientCounts != null) {\n    cachedRecipientCounts.setMaxEntries(maxEntries);\n  }\n}\nfunction execute(fetches, feedDict, kwargs, probe) {\n  const training = kwargs == null ? false : kwargs[\"training\"];\n  const arrayFetches = Array.isArray(fetches);\n  const fetchArray = arrayFetches ? fetches : [fetches];\n  const outputNames = fetchArray.map((t) => t.name);\n  const finalOutputs = [];\n  const feedNames = feedDict.names();\n  for (const outputName of outputNames) {\n    if (feedNames.indexOf(outputName) !== -1) {\n      finalOutputs.push(feedDict.getValue(outputName));\n    } else {\n      finalOutputs.push(null);\n    }\n  }\n  if (probe != null) {\n    probe.maxNumTensors = -Infinity;\n    probe.minNumTensors = Infinity;\n  }\n  const fetchAndFeedKey = outputNames.join(\",\") + \"|\" + feedDict.names().sort().join(\",\");\n  let sorted = cachedSorted.get(fetchAndFeedKey);\n  let recipientCounts;\n  if (sorted == null) {\n    const out = getTopologicalSortAndRecipientCounts(fetchArray, feedDict);\n    sorted = out.sorted;\n    recipientCounts = out.recipientCounts;\n    cachedSorted.put(fetchAndFeedKey, sorted);\n    cachedRecipientCounts.put(fetchAndFeedKey, recipientCounts);\n  }\n  recipientCounts = {};\n  if (!training) {\n    Object.assign(recipientCounts, cachedRecipientCounts.get(fetchAndFeedKey));\n  }\n  const internalFeedDict = new FeedDict(feedDict);\n  for (let i = 0; i < sorted.length; ++i) {\n    if (probe != null) {\n      const numTensors = memory().numTensors;\n      if (numTensors > probe.maxNumTensors) {\n        probe.maxNumTensors = numTensors;\n      }\n      if (numTensors < probe.minNumTensors) {\n        probe.minNumTensors = numTensors;\n      }\n    }\n    const symbolic = sorted[i];\n    const srcLayer = symbolic.sourceLayer;\n    if (srcLayer instanceof InputLayer) {\n      continue;\n    }\n    const inputValues = [];\n    const inputMasks = [];\n    const tensorsToDispose = [];\n    let maskExists = false;\n    for (const input2 of symbolic.inputs) {\n      const value = internalFeedDict.getValue(input2);\n      const mask = internalFeedDict.getMask(input2);\n      inputValues.push(value);\n      inputMasks.push(mask);\n      if (mask != null) {\n        maskExists = true;\n      }\n      if (!training) {\n        recipientCounts[input2.name]--;\n        if (recipientCounts[input2.name] === 0 && !feedDict.hasKey(input2) && outputNames.indexOf(input2.name) === -1 && !value.isDisposed && input2.sourceLayer.stateful !== true) {\n          tensorsToDispose.push(value);\n        }\n      }\n    }\n    if (maskExists) {\n      kwargs = kwargs || {};\n      kwargs[\"mask\"] = inputMasks[0];\n    }\n    const outputTensors = toList(srcLayer.apply(inputValues, kwargs));\n    let outputMask = null;\n    if (srcLayer.supportsMasking) {\n      outputMask = srcLayer.computeMask(inputValues, inputMasks);\n    }\n    const layerOutputs = getNodeOutputs(symbolic);\n    const outputSymbolicTensors = Array.isArray(layerOutputs) ? layerOutputs : [layerOutputs];\n    for (let i2 = 0; i2 < outputSymbolicTensors.length; ++i2) {\n      if (!internalFeedDict.hasKey(outputSymbolicTensors[i2])) {\n        internalFeedDict.add(outputSymbolicTensors[i2], outputTensors[i2], Array.isArray(outputMask) ? outputMask[0] : outputMask);\n      }\n      const index = outputNames.indexOf(outputSymbolicTensors[i2].name);\n      if (index !== -1) {\n        finalOutputs[index] = outputTensors[i2];\n      }\n    }\n    if (!training) {\n      dispose(tensorsToDispose);\n    }\n  }\n  internalFeedDict.disposeMasks();\n  return arrayFetches ? finalOutputs : finalOutputs[0];\n}\nfunction getTopologicalSortAndRecipientCounts(fetches, feedDict) {\n  util_exports.assert(fetches != null && fetches.length > 0, () => `Expected at least one fetch, got none`);\n  let finalSorted = [];\n  let finalRecipientMap = {};\n  if (fetches.length === 1) {\n    const out = getTopologicalSortAndRecipientCountsForOneFetch(fetches[0], feedDict);\n    finalSorted = out.sorted;\n    finalRecipientMap = out.recipientMap;\n  } else {\n    const visited = /* @__PURE__ */ new Set();\n    for (const fetch4 of fetches) {\n      const { sorted, recipientMap } = getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict);\n      for (const symbolicTensor of sorted) {\n        if (!visited.has(symbolicTensor.name)) {\n          finalSorted.push(symbolicTensor);\n          visited.add(symbolicTensor.name);\n        }\n      }\n      for (const name in recipientMap) {\n        if (finalRecipientMap[name] == null) {\n          finalRecipientMap[name] = /* @__PURE__ */ new Set();\n        }\n        recipientMap[name].forEach((recipient) => finalRecipientMap[name].add(recipient));\n      }\n    }\n  }\n  return {\n    sorted: finalSorted,\n    recipientCounts: recipientMap2Counts(finalRecipientMap)\n  };\n}\nfunction recipientMap2Counts(recipientMap) {\n  const recipientCounts = {};\n  for (const name in recipientMap) {\n    recipientCounts[name] = recipientMap[name].size;\n  }\n  return recipientCounts;\n}\nfunction getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict) {\n  const visited = /* @__PURE__ */ new Set();\n  const sorted = [];\n  const recipientMap = {};\n  for (const key of feedDict.names()) {\n    visited.add(key);\n  }\n  const stack2 = [];\n  const marks = [];\n  stack2.push(fetch4);\n  while (stack2.length > 0) {\n    const top = stack2[stack2.length - 1];\n    if (visited.has(top.name)) {\n      stack2.pop();\n      continue;\n    }\n    const topIsMarked = marks[marks.length - 1] === stack2.length - 1;\n    if (top.inputs.length === 0 || topIsMarked) {\n      stack2.pop();\n      sorted.push(top);\n      visited.add(top.name);\n      if (topIsMarked) {\n        marks.pop();\n      }\n    } else {\n      marks.push(stack2.length - 1);\n      for (const input2 of top.inputs) {\n        if (recipientMap[input2.name] == null) {\n          recipientMap[input2.name] = /* @__PURE__ */ new Set();\n        }\n        recipientMap[input2.name].add(top.name);\n        if (visited.has(input2.name)) {\n          continue;\n        }\n        stack2.push(input2);\n      }\n    }\n  }\n  return { sorted, recipientMap };\n}\nfunction getNodeOutputs(fetch4) {\n  let layerOutputs;\n  if (fetch4.sourceLayer.inboundNodes.length === 1) {\n    layerOutputs = fetch4.sourceLayer.output;\n  } else {\n    let nodeIndex = null;\n    for (let i = 0; i < fetch4.sourceLayer.inboundNodes.length; ++i) {\n      for (const outputTensor of fetch4.sourceLayer.inboundNodes[i].outputTensors) {\n        if (outputTensor.id === fetch4.id) {\n          nodeIndex = i;\n          break;\n        }\n      }\n    }\n    layerOutputs = fetch4.sourceLayer.getOutputAt(nodeIndex);\n  }\n  return layerOutputs;\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/flags_layers.js\nvar ENV3 = env();\nENV3.registerFlag(\"TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES\", () => 100, updateCacheMaxEntries);\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_constraints.js\nvar exports_constraints_exports = {};\n__export(exports_constraints_exports, {\n  maxNorm: () => maxNorm,\n  minMaxNorm: () => minMaxNorm,\n  nonNeg: () => nonNeg,\n  unitNorm: () => unitNorm\n});\n\n// node_modules/@tensorflow/tfjs-layers/dist/constraints.js\nfunction calcL2Norms(w, axis) {\n  return tidy(() => sqrt(sum2(mul(w, w), axis, true)));\n}\nvar Constraint = class extends serialization_exports.Serializable {\n  getConfig() {\n    return {};\n  }\n};\nvar MaxNorm = class extends Constraint {\n  constructor(args) {\n    super();\n    this.defaultMaxValue = 2;\n    this.defaultAxis = 0;\n    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;\n    this.axis = args.axis != null ? args.axis : this.defaultAxis;\n  }\n  apply(w) {\n    return tidy(() => {\n      const norms = calcL2Norms(w, this.axis);\n      const desired = clipByValue(norms, 0, this.maxValue);\n      return mul(w, div(desired, add2(epsilon(), norms)));\n    });\n  }\n  getConfig() {\n    return { maxValue: this.maxValue, axis: this.axis };\n  }\n};\nMaxNorm.className = \"MaxNorm\";\nserialization_exports.registerClass(MaxNorm);\nvar UnitNorm = class extends Constraint {\n  constructor(args) {\n    super();\n    this.defaultAxis = 0;\n    this.axis = args.axis != null ? args.axis : this.defaultAxis;\n  }\n  apply(w) {\n    return tidy(() => div(w, add2(epsilon(), calcL2Norms(w, this.axis))));\n  }\n  getConfig() {\n    return { axis: this.axis };\n  }\n};\nUnitNorm.className = \"UnitNorm\";\nserialization_exports.registerClass(UnitNorm);\nvar NonNeg = class extends Constraint {\n  apply(w) {\n    return relu(w);\n  }\n};\nNonNeg.className = \"NonNeg\";\nserialization_exports.registerClass(NonNeg);\nvar MinMaxNorm = class extends Constraint {\n  constructor(args) {\n    super();\n    this.defaultMinValue = 0;\n    this.defaultMaxValue = 1;\n    this.defaultRate = 1;\n    this.defaultAxis = 0;\n    this.minValue = args.minValue != null ? args.minValue : this.defaultMinValue;\n    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;\n    this.rate = args.rate != null ? args.rate : this.defaultRate;\n    this.axis = args.axis != null ? args.axis : this.defaultAxis;\n  }\n  apply(w) {\n    return tidy(() => {\n      const norms = calcL2Norms(w, this.axis);\n      const desired = add2(mul(this.rate, clipByValue(norms, this.minValue, this.maxValue)), mul(1 - this.rate, norms));\n      return mul(w, div(desired, add2(epsilon(), norms)));\n    });\n  }\n  getConfig() {\n    return {\n      minValue: this.minValue,\n      maxValue: this.maxValue,\n      rate: this.rate,\n      axis: this.axis\n    };\n  }\n};\nMinMaxNorm.className = \"MinMaxNorm\";\nserialization_exports.registerClass(MinMaxNorm);\nvar CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP = {\n  \"maxNorm\": \"MaxNorm\",\n  \"minMaxNorm\": \"MinMaxNorm\",\n  \"nonNeg\": \"NonNeg\",\n  \"unitNorm\": \"UnitNorm\"\n};\nfunction serializeConstraint(constraint) {\n  return serializeKerasObject(constraint);\n}\nfunction deserializeConstraint(config, customObjects = {}) {\n  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, \"constraint\");\n}\nfunction getConstraint(identifier) {\n  if (identifier == null) {\n    return null;\n  }\n  if (typeof identifier === \"string\") {\n    const className = identifier in CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP ? CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;\n    const config = { className, config: {} };\n    return deserializeConstraint(config);\n  } else if (identifier instanceof Constraint) {\n    return identifier;\n  } else {\n    return deserializeConstraint(identifier);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_constraints.js\nfunction maxNorm(args) {\n  return new MaxNorm(args);\n}\nfunction unitNorm(args) {\n  return new UnitNorm(args);\n}\nfunction nonNeg() {\n  return new NonNeg();\n}\nfunction minMaxNorm(config) {\n  return new MinMaxNorm(config);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_initializers.js\nvar exports_initializers_exports = {};\n__export(exports_initializers_exports, {\n  constant: () => constant,\n  glorotNormal: () => glorotNormal,\n  glorotUniform: () => glorotUniform,\n  heNormal: () => heNormal,\n  heUniform: () => heUniform,\n  identity: () => identity,\n  leCunNormal: () => leCunNormal,\n  leCunUniform: () => leCunUniform,\n  ones: () => ones3,\n  orthogonal: () => orthogonal,\n  randomNormal: () => randomNormal3,\n  randomUniform: () => randomUniform2,\n  truncatedNormal: () => truncatedNormal2,\n  varianceScaling: () => varianceScaling,\n  zeros: () => zeros2\n});\nfunction zeros2() {\n  return new Zeros();\n}\nfunction ones3() {\n  return new Ones();\n}\nfunction constant(args) {\n  return new Constant(args);\n}\nfunction randomUniform2(args) {\n  return new RandomUniform(args);\n}\nfunction randomNormal3(args) {\n  return new RandomNormal(args);\n}\nfunction truncatedNormal2(args) {\n  return new TruncatedNormal(args);\n}\nfunction identity(args) {\n  return new Identity2(args);\n}\nfunction varianceScaling(config) {\n  return new VarianceScaling(config);\n}\nfunction glorotUniform(args) {\n  return new GlorotUniform(args);\n}\nfunction glorotNormal(args) {\n  return new GlorotNormal(args);\n}\nfunction heNormal(args) {\n  return new HeNormal(args);\n}\nfunction heUniform(args) {\n  return new HeUniform(args);\n}\nfunction leCunNormal(args) {\n  return new LeCunNormal(args);\n}\nfunction leCunUniform(args) {\n  return new LeCunUniform(args);\n}\nfunction orthogonal(args) {\n  return new Orthogonal(args);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_layers.js\nvar exports_layers_exports = {};\n__export(exports_layers_exports, {\n  Layer: () => Layer,\n  RNN: () => RNN,\n  RNNCell: () => RNNCell,\n  activation: () => activation,\n  add: () => add3,\n  alphaDropout: () => alphaDropout,\n  average: () => average,\n  averagePooling1d: () => averagePooling1d,\n  averagePooling2d: () => averagePooling2d,\n  averagePooling3d: () => averagePooling3d,\n  avgPool1d: () => avgPool1d,\n  avgPool2d: () => avgPool2d,\n  avgPool3d: () => avgPool3d2,\n  avgPooling1d: () => avgPooling1d,\n  avgPooling2d: () => avgPooling2d,\n  avgPooling3d: () => avgPooling3d,\n  batchNormalization: () => batchNormalization2,\n  bidirectional: () => bidirectional,\n  categoryEncoding: () => categoryEncoding,\n  concatenate: () => concatenate2,\n  conv1d: () => conv1d2,\n  conv2d: () => conv2d3,\n  conv2dTranspose: () => conv2dTranspose2,\n  conv3d: () => conv3d2,\n  conv3dTranspose: () => conv3dTranspose2,\n  convLstm2d: () => convLstm2d,\n  convLstm2dCell: () => convLstm2dCell,\n  cropping2D: () => cropping2D,\n  dense: () => dense,\n  depthwiseConv2d: () => depthwiseConv2d4,\n  dot: () => dot3,\n  dropout: () => dropout3,\n  elu: () => elu3,\n  embedding: () => embedding,\n  flatten: () => flatten3,\n  gaussianDropout: () => gaussianDropout,\n  gaussianNoise: () => gaussianNoise,\n  globalAveragePooling1d: () => globalAveragePooling1d,\n  globalAveragePooling2d: () => globalAveragePooling2d,\n  globalMaxPool1d: () => globalMaxPool1d,\n  globalMaxPool2d: () => globalMaxPool2d,\n  globalMaxPooling1d: () => globalMaxPooling1d,\n  globalMaxPooling2d: () => globalMaxPooling2d,\n  gru: () => gru,\n  gruCell: () => gruCell,\n  input: () => input,\n  inputLayer: () => inputLayer,\n  layerNormalization: () => layerNormalization,\n  leakyReLU: () => leakyReLU,\n  lstm: () => lstm,\n  lstmCell: () => lstmCell,\n  masking: () => masking,\n  maxPool1d: () => maxPool1d,\n  maxPool2d: () => maxPool2d,\n  maxPooling1d: () => maxPooling1d,\n  maxPooling2d: () => maxPooling2d,\n  maxPooling3d: () => maxPooling3d,\n  maximum: () => maximum2,\n  minimum: () => minimum2,\n  multiply: () => multiply,\n  permute: () => permute,\n  prelu: () => prelu2,\n  reLU: () => reLU,\n  repeatVector: () => repeatVector,\n  rescaling: () => rescaling,\n  reshape: () => reshape2,\n  resizing: () => resizing,\n  rnn: () => rnn2,\n  separableConv2d: () => separableConv2d2,\n  simpleRNN: () => simpleRNN,\n  simpleRNNCell: () => simpleRNNCell,\n  softmax: () => softmax2,\n  spatialDropout1d: () => spatialDropout1d,\n  stackedRNNCells: () => stackedRNNCells,\n  thresholdedReLU: () => thresholdedReLU,\n  timeDistributed: () => timeDistributed,\n  upSampling2d: () => upSampling2d,\n  zeroPadding2d: () => zeroPadding2d\n});\n\n// node_modules/@tensorflow/tfjs-layers/dist/logs.js\nasync function resolveScalarsInLogs(logs) {\n  if (logs == null) {\n    return;\n  }\n  const promises = [];\n  const keys = [];\n  const scalarsToDispose = [];\n  for (const key in logs) {\n    const value = logs[key];\n    if (typeof value !== \"number\") {\n      const valueScalar = value;\n      promises.push(valueScalar.data());\n      keys.push(key);\n      scalarsToDispose.push(valueScalar);\n    }\n  }\n  if (promises.length > 0) {\n    const values = await Promise.all(promises);\n    for (let i = 0; i < values.length; ++i) {\n      logs[keys[i]] = values[i][0];\n    }\n    dispose(scalarsToDispose);\n  }\n}\nfunction disposeTensorsInLogs(logs) {\n  if (logs == null) {\n    return;\n  }\n  for (const key in logs) {\n    const value = logs[key];\n    if (typeof value !== \"number\") {\n      value.dispose();\n    }\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/base_callbacks.js\nvar ModelLoggingVerbosity;\n(function(ModelLoggingVerbosity2) {\n  ModelLoggingVerbosity2[ModelLoggingVerbosity2[\"SILENT\"] = 0] = \"SILENT\";\n  ModelLoggingVerbosity2[ModelLoggingVerbosity2[\"VERBOSE\"] = 1] = \"VERBOSE\";\n})(ModelLoggingVerbosity || (ModelLoggingVerbosity = {}));\nvar DEFAULT_YIELD_EVERY_MS = 125;\nvar BaseCallback = class {\n  constructor() {\n    this.validationData = null;\n  }\n  setParams(params) {\n    this.params = params;\n  }\n  async onEpochBegin(epoch, logs) {\n  }\n  async onEpochEnd(epoch, logs) {\n  }\n  async onBatchBegin(batch, logs) {\n  }\n  async onBatchEnd(batch, logs) {\n  }\n  async onTrainBegin(logs) {\n  }\n  async onTrainEnd(logs) {\n  }\n  setModel(model2) {\n  }\n};\nvar CallbackList = class {\n  constructor(callbacks2, queueLength = 10) {\n    if (callbacks2 == null) {\n      callbacks2 = [];\n    }\n    this.callbacks = callbacks2;\n    this.queueLength = queueLength;\n  }\n  append(callback) {\n    this.callbacks.push(callback);\n  }\n  setParams(params) {\n    for (const callback of this.callbacks) {\n      callback.setParams(params);\n    }\n  }\n  setModel(model2) {\n    for (const callback of this.callbacks) {\n      callback.setModel(model2);\n    }\n  }\n  async onEpochBegin(epoch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onEpochBegin(epoch, logs);\n    }\n  }\n  async onEpochEnd(epoch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onEpochEnd(epoch, logs);\n    }\n  }\n  async onBatchBegin(batch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onBatchBegin(batch, logs);\n    }\n  }\n  async onBatchEnd(batch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onBatchEnd(batch, logs);\n    }\n  }\n  async onTrainBegin(logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onTrainBegin(logs);\n    }\n  }\n  async onTrainEnd(logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    for (const callback of this.callbacks) {\n      await callback.onTrainEnd(logs);\n    }\n  }\n};\nvar BaseLogger = class extends BaseCallback {\n  constructor() {\n    super();\n  }\n  async onEpochBegin(epoch) {\n    this.seen = 0;\n    this.totals = {};\n  }\n  async onBatchEnd(batch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    const batchSize = logs[\"size\"] == null ? 0 : logs[\"size\"];\n    this.seen += batchSize;\n    for (const key in logs) {\n      const value = logs[key];\n      if (typeof value === \"number\") {\n        if (!this.totals.hasOwnProperty(key)) {\n          this.totals[key] = 0;\n        }\n        this.totals[key] = this.totals[key] + value * batchSize;\n      } else {\n        let oldTotalsToDispose;\n        if (key in this.totals) {\n          oldTotalsToDispose = this.totals[key];\n        } else {\n          this.totals[key] = 0;\n        }\n        const total = tidy(() => add2(this.totals[key], mul(value, batchSize)));\n        this.totals[key] = total;\n        if (oldTotalsToDispose != null) {\n          oldTotalsToDispose.dispose();\n        }\n      }\n    }\n  }\n  async onEpochEnd(epoch, logs) {\n    if (logs != null) {\n      for (const key of this.params[\"metrics\"]) {\n        if (this.totals[key] == null) {\n          continue;\n        }\n        if (typeof this.totals[key] === \"number\") {\n          logs[key] = this.totals[key] / this.seen;\n        } else {\n          tidy(() => {\n            const log5 = mul(div(1, this.seen), this.totals[key]);\n            logs[key] = log5;\n            this.totals[key].dispose();\n            keep(logs[key]);\n          });\n        }\n      }\n    }\n  }\n};\nvar History = class extends BaseCallback {\n  async onTrainBegin(logs) {\n    this.epoch = [];\n    this.history = {};\n  }\n  async onEpochEnd(epoch, logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    this.epoch.push(epoch);\n    for (const key in logs) {\n      if (this.history[key] == null) {\n        this.history[key] = [];\n      }\n      this.history[key].push(logs[key]);\n    }\n  }\n  async syncData() {\n    const promises = [];\n    const keys = [];\n    const indices = [];\n    for (const key in this.history) {\n      const valueArray = this.history[key];\n      for (let i = 0; i < valueArray.length; ++i) {\n        if (typeof valueArray[i] !== \"number\") {\n          const valueScalar = valueArray[i];\n          promises.push(valueScalar.data());\n          keys.push(key);\n          indices.push(i);\n        }\n      }\n    }\n    const values = await Promise.all(promises);\n    for (let n = 0; n < values.length; ++n) {\n      const tensorToDispose = this.history[keys[n]][indices[n]];\n      tensorToDispose.dispose();\n      this.history[keys[n]][indices[n]] = values[n][0];\n    }\n  }\n};\nvar CustomCallback = class extends BaseCallback {\n  constructor(args, yieldEvery) {\n    super();\n    this.currentEpoch = 0;\n    this.nowFunc = args.nowFunc;\n    this.nextFrameFunc = args.nextFrameFunc || nextFrame;\n    this.yieldEvery = yieldEvery || \"auto\";\n    if (this.yieldEvery === \"auto\") {\n      this.yieldEvery = DEFAULT_YIELD_EVERY_MS;\n    }\n    if (this.yieldEvery === \"never\" && args.onYield != null) {\n      throw new Error(\"yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback\");\n    }\n    if (util_exports.isNumber(this.yieldEvery)) {\n      this.maybeWait = debounce(this.maybeWait.bind(this), this.yieldEvery, this.nowFunc);\n    }\n    this.trainBegin = args.onTrainBegin;\n    this.trainEnd = args.onTrainEnd;\n    this.epochBegin = args.onEpochBegin;\n    this.epochEnd = args.onEpochEnd;\n    this.batchBegin = args.onBatchBegin;\n    this.batchEnd = args.onBatchEnd;\n    this.yield = args.onYield;\n  }\n  async maybeWait(epoch, batch, logs) {\n    const ps = [];\n    if (this.yield != null) {\n      await resolveScalarsInLogs(logs);\n      ps.push(this.yield(epoch, batch, logs));\n    }\n    ps.push(this.nextFrameFunc());\n    await Promise.all(ps);\n  }\n  async onEpochBegin(epoch, logs) {\n    this.currentEpoch = epoch;\n    if (this.epochBegin != null) {\n      await resolveScalarsInLogs(logs);\n      await this.epochBegin(epoch, logs);\n    }\n  }\n  async onEpochEnd(epoch, logs) {\n    const ps = [];\n    if (this.epochEnd != null) {\n      await resolveScalarsInLogs(logs);\n      ps.push(this.epochEnd(epoch, logs));\n    }\n    if (this.yieldEvery === \"epoch\") {\n      ps.push(this.nextFrameFunc());\n    }\n    await Promise.all(ps);\n  }\n  async onBatchBegin(batch, logs) {\n    if (this.batchBegin != null) {\n      await resolveScalarsInLogs(logs);\n      await this.batchBegin(batch, logs);\n    }\n  }\n  async onBatchEnd(batch, logs) {\n    const ps = [];\n    if (this.batchEnd != null) {\n      await resolveScalarsInLogs(logs);\n      ps.push(this.batchEnd(batch, logs));\n    }\n    if (this.yieldEvery === \"batch\") {\n      ps.push(this.nextFrameFunc());\n    } else if (util_exports.isNumber(this.yieldEvery)) {\n      ps.push(this.maybeWait(this.currentEpoch, batch, logs));\n    }\n    await Promise.all(ps);\n  }\n  async onTrainBegin(logs) {\n    if (this.trainBegin != null) {\n      await resolveScalarsInLogs(logs);\n      await this.trainBegin(logs);\n    }\n  }\n  async onTrainEnd(logs) {\n    if (this.trainEnd != null) {\n      await resolveScalarsInLogs(logs);\n      await this.trainEnd(logs);\n    }\n  }\n};\nfunction standardizeCallbacks(callbacks2, yieldEvery) {\n  if (callbacks2 == null) {\n    callbacks2 = {};\n  }\n  if (callbacks2 instanceof BaseCallback) {\n    return [callbacks2];\n  }\n  if (Array.isArray(callbacks2) && callbacks2[0] instanceof BaseCallback) {\n    return callbacks2;\n  }\n  const callbackConfigs = toList(callbacks2);\n  return callbackConfigs.map((callbackConfig) => new CustomCallback(callbackConfig, yieldEvery));\n}\nvar CallbackConstructorRegistry = class {\n  constructor() {\n  }\n  static registerCallbackConstructor(verbosityLevel, callbackConstructor) {\n    util_exports.assert(verbosityLevel >= 0 && Number.isInteger(verbosityLevel), () => `Verbosity level is expected to be an integer >= 0, but got ${verbosityLevel}`);\n    CallbackConstructorRegistry.checkForDuplicate(callbackConstructor);\n    if (CallbackConstructorRegistry.constructors[verbosityLevel] == null) {\n      CallbackConstructorRegistry.constructors[verbosityLevel] = [];\n    }\n    CallbackConstructorRegistry.constructors[verbosityLevel].push(callbackConstructor);\n  }\n  static checkForDuplicate(callbackConstructor) {\n    for (const levelName in CallbackConstructorRegistry.constructors) {\n      const constructors = CallbackConstructorRegistry.constructors[+levelName];\n      constructors.forEach((ctor) => {\n        if (ctor === callbackConstructor) {\n          throw new ValueError(\"Duplicate callback constructor.\");\n        }\n      });\n    }\n  }\n  static clear() {\n    CallbackConstructorRegistry.constructors = {};\n  }\n  static createCallbacks(verbosityLevel) {\n    const constructors = [];\n    for (const levelName in CallbackConstructorRegistry.constructors) {\n      const level = +levelName;\n      if (verbosityLevel >= level) {\n        constructors.push(...CallbackConstructorRegistry.constructors[level]);\n      }\n    }\n    return constructors.map((ctor) => new ctor());\n  }\n};\nCallbackConstructorRegistry.constructors = {};\nfunction configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics) {\n  const history = new History();\n  const actualCallbacks = [\n    new BaseLogger(),\n    ...CallbackConstructorRegistry.createCallbacks(verbose)\n  ];\n  if (callbacks2 != null) {\n    actualCallbacks.push(...callbacks2);\n  }\n  actualCallbacks.push(history);\n  const callbackList = new CallbackList(actualCallbacks);\n  callbackList.setParams({\n    epochs,\n    initialEpoch,\n    samples: numTrainSamples,\n    steps: stepsPerEpoch,\n    batchSize,\n    verbose,\n    doValidation,\n    metrics: callbackMetrics\n  });\n  return { callbackList, history };\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/serialization.js\nfunction deserialize(config, customObjects = {}, fastWeightInit = false) {\n  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, \"layer\", fastWeightInit);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/losses.js\nfunction l2Normalize(x, axis) {\n  return tidy(() => {\n    if (x.dtype !== \"float32\") {\n      x = cast(x, \"float32\");\n    }\n    const squareSum = sum2(square2(x), axis, true);\n    const epsilonTensor = fill(squareSum.shape, epsilon());\n    const norm2 = sqrt(maximum(squareSum, epsilonTensor));\n    return div(x, norm2);\n  });\n}\nfunction meanSquaredError2(yTrue, yPred) {\n  return tidy(() => mean(square2(sub(yPred, yTrue)), -1));\n}\nfunction meanAbsoluteError(yTrue, yPred) {\n  return tidy(() => mean(abs(sub(yPred, yTrue)), -1));\n}\nfunction meanAbsolutePercentageError(yTrue, yPred) {\n  return tidy(() => {\n    const diff = sub(yTrue, yPred);\n    const clippedTrue = clipByValue(abs(yTrue), epsilon(), Number.MAX_VALUE);\n    const absResult = abs(div(diff, clippedTrue));\n    return mul(100, mean(absResult, -1));\n  });\n}\nfunction meanSquaredLogarithmicError(yTrue, yPred) {\n  return tidy(() => {\n    const clippedPred = clipByValue(yPred, epsilon(), Number.MAX_VALUE);\n    const firstLog = log2(add2(1, clippedPred));\n    const clippedTrue = clipByValue(yTrue, epsilon(), Number.MAX_VALUE);\n    const secondLog = log2(add2(1, clippedTrue));\n    return mean(square2(sub(firstLog, secondLog)), -1);\n  });\n}\nfunction squaredHinge(yTrue, yPred) {\n  return tidy(() => {\n    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));\n    return mean(square2(maxResult), -1);\n  });\n}\nfunction hinge(yTrue, yPred) {\n  return tidy(() => {\n    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));\n    return mean(maxResult, -1);\n  });\n}\nfunction categoricalHinge(yTrue, yPred) {\n  return tidy(() => {\n    const pos = sum2(mul(yTrue, yPred), -1);\n    const neg4 = max(mul(sub(1, yTrue), yPred), -1);\n    return maximum(0, add2(1, sub(neg4, pos)));\n  });\n}\nfunction logcosh(yTrue, yPred) {\n  return tidy(() => {\n    const log22 = Math.log(2);\n    const predictionDiff = sub(yPred, yTrue);\n    const logcoshResult = sub(add2(predictionDiff, softplus(mul(-2, predictionDiff))), log22);\n    return mean(logcoshResult, -1);\n  });\n}\nfunction categoricalCrossentropy(target, output, fromLogits = false) {\n  return tidy(() => {\n    if (fromLogits) {\n      output = softmax(output);\n    } else {\n      const outputSum = sum2(output, output.shape.length - 1, true);\n      output = div(output, outputSum);\n    }\n    output = clipByValue(output, epsilon(), 1 - epsilon());\n    return neg(sum2(mul(cast(target, \"float32\"), log2(output)), output.shape.length - 1));\n  });\n}\nfunction sparseCategoricalCrossentropy(target, output, fromLogits = false) {\n  return tidy(() => {\n    const flatTarget = cast(floor(flatten2(target)), \"int32\");\n    output = clipByValue(output, epsilon(), 1 - epsilon());\n    const outputShape = output.shape;\n    const oneHotTarget = reshape(oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);\n    return categoricalCrossentropy(oneHotTarget, output, fromLogits);\n  });\n}\nfunction sigmoidCrossEntropyWithLogits(labels, logits) {\n  if (!util_exports.arraysEqual(labels.shape, logits.shape)) {\n    throw new ValueError(`logits and labels must have the same shape, but got shapes ${JSON.stringify(labels.shape)} and ${JSON.stringify(logits.shape)}`);\n  }\n  return tidy(() => {\n    const reluLogits = relu(logits);\n    const negAbsLogits = neg(abs(logits));\n    return add2(sub(reluLogits, mul(logits, labels)), log1p(exp(negAbsLogits)));\n  });\n}\nfunction binaryCrossentropy(yTrue, yPred) {\n  return tidy(() => {\n    let y;\n    y = clipByValue(yPred, epsilon(), 1 - epsilon());\n    y = log2(div(y, sub(1, y)));\n    return mean(sigmoidCrossEntropyWithLogits(yTrue, y), -1);\n  });\n}\nfunction kullbackLeiblerDivergence(yTrue, yPred) {\n  return tidy(() => {\n    const clippedTrue = clipByValue(yTrue, epsilon(), 1);\n    const clippedPred = clipByValue(yPred, epsilon(), 1);\n    return sum2(mul(yTrue, log2(div(clippedTrue, clippedPred))), -1);\n  });\n}\nfunction poisson(yTrue, yPred) {\n  return tidy(() => {\n    const logPred = log2(add2(epsilon(), yPred));\n    return mean(sub(yPred, mul(yTrue, logPred)), -1);\n  });\n}\nfunction cosineProximity(yTrue, yPred) {\n  return tidy(() => {\n    const trueNormalized = l2Normalize(yTrue, -1);\n    const predNormalized = l2Normalize(yPred, -1);\n    const trueXPred = mul(trueNormalized, predNormalized);\n    return neg(sum2(trueXPred, -1));\n  });\n}\nvar lossesMap = {\n  meanSquaredError: meanSquaredError2,\n  meanAbsoluteError,\n  meanAbsolutePercentageError,\n  meanSquaredLogarithmicError,\n  squaredHinge,\n  hinge,\n  categoricalHinge,\n  logcosh,\n  categoricalCrossentropy,\n  sparseCategoricalCrossentropy,\n  binaryCrossentropy,\n  kullbackLeiblerDivergence,\n  poisson,\n  cosineProximity\n};\nfunction get(identifierOrFn) {\n  if (typeof identifierOrFn === \"string\") {\n    if (identifierOrFn in lossesMap) {\n      return lossesMap[identifierOrFn];\n    }\n    let errMsg = `Unknown loss ${identifierOrFn}`;\n    if (identifierOrFn.toLowerCase().includes(\"softmaxcrossentropy\")) {\n      errMsg = `Unknown loss ${identifierOrFn}. Use \"categoricalCrossentropy\" as the string name for tf.losses.softmaxCrossEntropy`;\n    }\n    throw new ValueError(errMsg);\n  } else {\n    return identifierOrFn;\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/metrics.js\nfunction binaryAccuracy(yTrue, yPred) {\n  return tidy(() => {\n    const threshold3 = mul(0.5, onesLike(yPred));\n    const yPredThresholded = cast2(greater(yPred, threshold3), yTrue.dtype);\n    return mean(equal(yTrue, yPredThresholded), -1);\n  });\n}\nfunction categoricalAccuracy(yTrue, yPred) {\n  return tidy(() => cast2(equal(argMax(yTrue, -1), argMax(yPred, -1)), \"float32\"));\n}\nfunction truePositives(yTrue, yPred) {\n  return tidy(() => {\n    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 1))), \"float32\");\n  });\n}\nfunction falseNegatives(yTrue, yPred) {\n  return tidy(() => {\n    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 0))), \"float32\");\n  });\n}\nfunction falsePositives(yTrue, yPred) {\n  return tidy(() => {\n    return cast(sum2(logicalAnd(equal(yTrue, 0), equal(yPred, 1))), \"float32\");\n  });\n}\nfunction precision(yTrue, yPred) {\n  return tidy(() => {\n    const tp = truePositives(yTrue, yPred);\n    const fp = falsePositives(yTrue, yPred);\n    const denominator = add2(tp, fp);\n    return cast(where(greater(denominator, 0), div(tp, denominator), 0), \"float32\");\n  });\n}\nfunction recall(yTrue, yPred) {\n  return tidy(() => {\n    const tp = truePositives(yTrue, yPred);\n    const fn = falseNegatives(yTrue, yPred);\n    const denominator = add2(tp, fn);\n    return cast(where(greater(denominator, 0), div(tp, denominator), 0), \"float32\");\n  });\n}\nfunction binaryCrossentropy2(yTrue, yPred) {\n  return binaryCrossentropy(yTrue, yPred);\n}\nfunction sparseCategoricalAccuracy(yTrue, yPred) {\n  if (yTrue.rank === yPred.rank) {\n    yTrue = squeeze(yTrue, [yTrue.rank - 1]);\n  }\n  yPred = argMax(yPred, -1);\n  if (yPred.dtype !== yTrue.dtype) {\n    yPred = cast(yPred, yTrue.dtype);\n  }\n  return cast(equal(yTrue, yPred), \"float32\");\n}\nvar mse = meanSquaredError2;\nvar MSE = meanSquaredError2;\nvar mae = meanAbsoluteError;\nvar MAE = meanAbsoluteError;\nvar mape = meanAbsolutePercentageError;\nvar MAPE = meanAbsolutePercentageError;\nvar categoricalCrossentropy2 = categoricalCrossentropy;\nvar cosine = cosineProximity;\nvar sparseCategoricalCrossentropy2 = sparseCategoricalCrossentropy;\nvar metricsMap = {\n  binaryAccuracy,\n  categoricalAccuracy,\n  precision,\n  categoricalCrossentropy: categoricalCrossentropy2,\n  sparseCategoricalCrossentropy: sparseCategoricalCrossentropy2,\n  mse,\n  MSE,\n  mae,\n  MAE,\n  mape,\n  MAPE,\n  cosine\n};\nfunction get2(identifier) {\n  if (typeof identifier === \"string\" && identifier in metricsMap) {\n    return metricsMap[identifier];\n  } else if (typeof identifier !== \"string\" && identifier != null) {\n    return identifier;\n  } else {\n    throw new ValueError(`Unknown metric ${identifier}`);\n  }\n}\nfunction getLossOrMetricName(fn) {\n  assert2(fn !== null, `Unknown LossOrMetricFn ${fn}`);\n  if (typeof fn === \"string\") {\n    return fn;\n  } else {\n    let fnName;\n    for (const key of Object.keys(lossesMap)) {\n      if (lossesMap[key] === fn) {\n        fnName = key;\n        break;\n      }\n    }\n    if (fnName !== void 0) {\n      return fnName;\n    }\n    for (const key of Object.keys(metricsMap)) {\n      if (metricsMap[key] === fn) {\n        fnName = key;\n        break;\n      }\n    }\n    if (fnName !== void 0) {\n      return fnName;\n    }\n    return fn.name;\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/optimizers.js\nfunction getOptimizer(identifier) {\n  const optimizerMap = {\n    \"Adagrad\": () => train.adagrad(0.01),\n    \"Adadelta\": () => train.adadelta(1, 0.95, epsilon()),\n    \"Adam\": () => train.adam(1e-3, 0.9, 0.999, epsilon()),\n    \"Adamax\": () => train.adamax(2e-3, 0.9, 0.999, epsilon(), 0),\n    \"RMSProp\": () => train.rmsprop(1e-3, 0.9, 0, epsilon()),\n    \"SGD\": () => train.sgd(0.01)\n  };\n  optimizerMap[\"adagrad\"] = optimizerMap[\"Adagrad\"];\n  optimizerMap[\"adadelta\"] = optimizerMap[\"Adadelta\"];\n  optimizerMap[\"adam\"] = optimizerMap[\"Adam\"];\n  optimizerMap[\"adamax\"] = optimizerMap[\"Adamax\"];\n  optimizerMap[\"rmsprop\"] = optimizerMap[\"RMSProp\"];\n  optimizerMap[\"sgd\"] = optimizerMap[\"SGD\"];\n  if (identifier in optimizerMap) {\n    return optimizerMap[identifier]();\n  }\n  throw new ValueError(`Unknown Optimizer ${identifier}`);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/user_defined_metadata.js\nvar MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH = 1 * 1024 * 1024;\nfunction checkUserDefinedMetadata(userDefinedMetadata, modelName, checkSize = false) {\n  if (userDefinedMetadata == null || typeof userDefinedMetadata !== \"object\" || Object.getPrototypeOf(userDefinedMetadata) !== Object.prototype || !plainObjectCheck(userDefinedMetadata)) {\n    throw new Error(\"User-defined metadata is expected to be a JSON object, but is not.\");\n  }\n  if (checkSize) {\n    const out = JSON.stringify(userDefinedMetadata);\n    if (out.length > MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH) {\n      console.warn(`User-defined metadata of model \"${modelName}\" is too large in size (length=${out.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH}.`);\n    }\n  }\n}\nfunction plainObjectCheck(x) {\n  if (x === null) {\n    return true;\n  } else if (typeof x === \"object\") {\n    if (Object.getPrototypeOf(x) === Object.prototype) {\n      const keys = Object.keys(x);\n      for (const key of keys) {\n        if (typeof key !== \"string\") {\n          return false;\n        }\n        if (!plainObjectCheck(x[key])) {\n          return false;\n        }\n      }\n      return true;\n    } else {\n      if (Array.isArray(x)) {\n        for (const item of x) {\n          if (!plainObjectCheck(item)) {\n            return false;\n          }\n        }\n        return true;\n      } else {\n        return false;\n      }\n    }\n  } else {\n    const xType = typeof x;\n    return xType === \"string\" || xType === \"number\" || xType === \"boolean\";\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/layer_utils.js\nfunction printSummary(model2, lineLength, positions, printFn = console.log) {\n  const sequentialLike = isModelSequentialLike(model2);\n  const toDisplay = [\"Layer (type)\", \"Input Shape\", \"Output shape\", \"Param #\"];\n  if (sequentialLike) {\n    lineLength = lineLength || 90;\n    positions = positions || [0.32, 0.61, 0.89, 1];\n  } else {\n    lineLength = lineLength || 115;\n    positions = positions || [0.24, 0.48, 0.7, 0.8, 1];\n  }\n  if (positions[positions.length - 1] <= 1) {\n    positions = positions.map((p2) => Math.floor(lineLength * p2));\n  }\n  let relevantNodes;\n  if (!sequentialLike) {\n    toDisplay.push(\"Receives inputs\");\n    relevantNodes = [];\n    for (const depth in model2.nodesByDepth) {\n      relevantNodes.push(...model2.nodesByDepth[depth]);\n    }\n  }\n  printFn(\"_\".repeat(lineLength));\n  printRow(toDisplay, positions, printFn);\n  printFn(\"=\".repeat(lineLength));\n  const layers = model2.layers;\n  for (let i = 0; i < layers.length; ++i) {\n    if (sequentialLike) {\n      printLayerSummary(layers[i], positions, printFn);\n    } else {\n      printLayerSummaryWithConnections(layers[i], positions, relevantNodes, printFn);\n    }\n    printFn((i === layers.length - 1 ? \"=\" : \"_\").repeat(lineLength));\n  }\n  model2.checkTrainableWeightsConsistency();\n  const trainableCount = countTrainableParams(model2);\n  const nonTrainableCount = countParamsInWeights(model2.nonTrainableWeights);\n  printFn(`Total params: ${trainableCount + nonTrainableCount}`);\n  printFn(`Trainable params: ${trainableCount}`);\n  printFn(`Non-trainable params: ${nonTrainableCount}`);\n  printFn(\"_\".repeat(lineLength));\n}\nfunction countTrainableParams(model2) {\n  let trainableCount;\n  if (model2.collectedTrainableWeights != null) {\n    trainableCount = countParamsInWeights(model2.collectedTrainableWeights);\n  } else {\n    trainableCount = countParamsInWeights(model2.trainableWeights);\n  }\n  return trainableCount;\n}\nfunction isModelSequentialLike(model2) {\n  let sequentialLike = true;\n  const nodesByDepth = [];\n  const nodes = [];\n  for (const depth in model2.nodesByDepth) {\n    nodesByDepth.push(model2.nodesByDepth[depth]);\n  }\n  for (const depthNodes of nodesByDepth) {\n    if (depthNodes.length > 1 || depthNodes.length === 1 && depthNodes[0].inboundLayers.length > 1) {\n      sequentialLike = false;\n      break;\n    }\n    nodes.push(...depthNodes);\n  }\n  if (sequentialLike) {\n    for (const layer of model2.layers) {\n      let flag = false;\n      for (const node of layer.inboundNodes) {\n        if (nodes.indexOf(node) !== -1) {\n          if (flag) {\n            sequentialLike = false;\n            break;\n          } else {\n            flag = true;\n          }\n        }\n      }\n      if (!sequentialLike) {\n        break;\n      }\n    }\n  }\n  return sequentialLike;\n}\nfunction printRow(fields, positions, printFn = console.log) {\n  let line = \"\";\n  for (let i = 0; i < fields.length; ++i) {\n    if (i > 0) {\n      line = line.slice(0, line.length - 1) + \" \";\n    }\n    line += fields[i];\n    line = line.slice(0, positions[i]);\n    line += \" \".repeat(positions[i] - line.length);\n  }\n  printFn(line);\n}\nfunction printLayerSummary(layer, positions, printFn) {\n  let outputShape;\n  let inputShape;\n  try {\n    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(\",\");\n  } catch (err) {\n    inputShape = \"multiple\";\n  }\n  try {\n    outputShape = JSON.stringify(layer.outputShape);\n  } catch (err) {\n    outputShape = \"multiple\";\n  }\n  const name = layer.name;\n  const className = layer.getClassName();\n  const fields = [\n    `${name} (${className})`,\n    inputShape,\n    outputShape,\n    layer.countParams().toString()\n  ];\n  printRow(fields, positions, printFn);\n}\nfunction printLayerSummaryWithConnections(layer, positions, relevantNodes, printFn) {\n  let outputShape;\n  let inputShape;\n  try {\n    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(\",\");\n  } catch (err) {\n    inputShape = \"multiple\";\n  }\n  try {\n    outputShape = JSON.stringify(layer.outputShape);\n  } catch (err) {\n    outputShape = \"multiple\";\n  }\n  const connections = [];\n  for (const node of layer.inboundNodes) {\n    if (relevantNodes != null && relevantNodes.length > 0 && relevantNodes.indexOf(node) === -1) {\n      continue;\n    }\n    for (let i = 0; i < node.inboundLayers.length; ++i) {\n      const inboundLayer = node.inboundLayers[i].name;\n      const inboundLayerIndex = node.nodeIndices[i];\n      const inboundTensorIndex = node.tensorIndices[i];\n      connections.push(`${inboundLayer}[${inboundLayerIndex}][${inboundTensorIndex}]`);\n    }\n  }\n  const name = layer.name;\n  const className = layer.getClassName();\n  const firstConnection = connections.length === 0 ? \"\" : connections[0];\n  const fields = [\n    `${name} (${className})`,\n    inputShape,\n    outputShape,\n    layer.countParams().toString(),\n    firstConnection\n  ];\n  printRow(fields, positions, printFn);\n  for (let i = 1; i < connections.length; ++i) {\n    printRow([\"\", \"\", \"\", \"\", connections[i]], positions, printFn);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/serialization_utils.js\nfunction isArrayItemInputOrOutputName(key, index, value) {\n  return (key === \"inboundNodes\" || key === \"outputLayers\" || key === \"inputLayers\") && index === 0 && typeof value === \"string\";\n}\nfunction convertPythonicToTs(pythonicConfig, key) {\n  if (pythonicConfig === null) {\n    return null;\n  } else if (typeof pythonicConfig === \"string\") {\n    return toCamelCase(pythonicConfig);\n  } else if (typeof pythonicConfig === \"number\" || typeof pythonicConfig === \"boolean\") {\n    return pythonicConfig;\n  } else if (pythonicConfig instanceof Array) {\n    const tsArray = [];\n    const arrayLength = pythonicConfig.length;\n    for (let i = 0; i < arrayLength; ++i) {\n      const item = pythonicConfig[i];\n      if (isArrayItemInputOrOutputName(key, i, item)) {\n        tsArray.push(item);\n      } else {\n        tsArray.push(convertPythonicToTs(item, key));\n      }\n    }\n    return tsArray;\n  } else {\n    const tsDict = {};\n    for (const pythonicKey of Object.keys(pythonicConfig)) {\n      const pythonicValue = pythonicConfig[pythonicKey];\n      if (pythonicKey === \"name\" && typeof pythonicValue === \"string\") {\n        tsDict[pythonicKey] = pythonicValue;\n      } else {\n        const tsKey = toCamelCase(pythonicKey);\n        tsDict[tsKey] = convertPythonicToTs(pythonicValue, tsKey);\n      }\n    }\n    return tsDict;\n  }\n}\nfunction convertTsToPythonic(tsConfig, key) {\n  if (tsConfig === null || tsConfig === void 0) {\n    return null;\n  } else if (typeof tsConfig === \"string\") {\n    return toSnakeCase(tsConfig);\n  } else if (typeof tsConfig === \"number\" || typeof tsConfig === \"boolean\") {\n    return tsConfig;\n  } else if (tsConfig instanceof Array) {\n    const pyArray = [];\n    const arrayLength = tsConfig.length;\n    for (let i = 0; i < arrayLength; ++i) {\n      const item = tsConfig[i];\n      if (isArrayItemInputOrOutputName(key, i, item)) {\n        pyArray.push(item);\n      } else {\n        pyArray.push(convertTsToPythonic(item, key));\n      }\n    }\n    return pyArray;\n  } else {\n    const pyDict = {};\n    for (const tsKey of Object.keys(tsConfig)) {\n      const tsValue = tsConfig[tsKey];\n      const pyKey = toSnakeCase(tsKey);\n      if ((tsKey === \"name\" || tsKey === \"className\") && typeof tsValue === \"string\") {\n        pyDict[pyKey] = tsValue;\n      } else {\n        pyDict[pyKey] = convertTsToPythonic(tsValue, tsKey);\n      }\n    }\n    return pyDict;\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/version.js\nvar version2 = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/container.js\nvar Container = class extends Layer {\n  constructor(args) {\n    super({});\n    this.containerNodes = /* @__PURE__ */ new Set();\n    this.name = args.name;\n    if (this.name == null) {\n      const prefix = this.getClassName().toLowerCase();\n      this.name = getUid(prefix);\n    }\n    this.supportsMasking = false;\n    this.trainable_ = true;\n    if (Array.isArray(args.inputs)) {\n      this.inputs = args.inputs.slice();\n    } else {\n      this.inputs = [args.inputs];\n    }\n    if (Array.isArray(args.outputs)) {\n      this.outputs = args.outputs.slice();\n    } else {\n      this.outputs = [args.outputs];\n    }\n    if (unique2(this.inputs).length !== this.inputs.length) {\n      throw new ValueError(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map((x) => x.name)}`);\n    }\n    if (unique2(this.outputs).length !== this.outputs.length) {\n      console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map((x) => x.name)}`);\n    }\n    this.inputLayers = [];\n    this.inputLayersNodeIndices = [];\n    this.inputLayersTensorIndices = [];\n    this.outputLayers = [];\n    this.outputLayersNodeIndices = [];\n    this.outputLayersTensorIndices = [];\n    this.layers = [];\n    this.internalContainerRefs = [];\n    for (const x of this.outputs) {\n      const layer = x.sourceLayer;\n      const nodeIndex = x.nodeIndex;\n      const tensorIndex = x.tensorIndex;\n      this.outputLayers.push(layer);\n      this.outputLayersNodeIndices.push(nodeIndex);\n      this.outputLayersTensorIndices.push(tensorIndex);\n    }\n    for (const x of this.inputs) {\n      const layer = x.sourceLayer;\n      const nodeIndex = x.nodeIndex;\n      const tensorIndex = x.tensorIndex;\n      assert2(nodeIndex === 0, \"input layer has >1 nodes\");\n      assert2(tensorIndex === 0, \"input layer has >1 tensors\");\n      this.inputLayers.push(layer);\n      this.inputLayersNodeIndices.push(nodeIndex);\n      this.inputLayersTensorIndices.push(tensorIndex);\n    }\n    this.inputNames = [];\n    this.outputNames = [];\n    this.feedInputShapes = [];\n    this.feedInputNames = [];\n    this.feedOutputNames = [];\n    for (let i = 0; i < this.inputLayers.length; i++) {\n      const layer = this.inputLayers[i];\n      if (!(layer instanceof InputLayer)) {\n        throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${args.inputs}. Input ${i} (0-based) originates from layer type ${layer.getClassName()}.`);\n      }\n      this.inputNames.push(layer.name);\n      this.feedInputShapes.push(layer.batchInputShape);\n      this.feedInputNames.push(layer.name);\n    }\n    for (const layer of this.outputLayers) {\n      this.outputNames.push(layer.name);\n    }\n    this.internalInputShapes = this.inputs.map((x) => x.shape);\n    this.internalOutputShapes = this.outputs.map((x) => x.shape);\n    const nodesDepths = {};\n    const nodeIDToNode = {};\n    const layersDepths = {};\n    const layerIDToLayer = {};\n    const layerIndices = {};\n    const nodesInDecreasingDepth = [];\n    const buildMapOfGraph = (tensor2, finishedNodes2, nodesInProgress2, layer, nodeIndex, tensorIndex) => {\n      if (layer == null || nodeIndex == null || tensorIndex == null) {\n        layer = tensor2.sourceLayer;\n        nodeIndex = tensor2.nodeIndex;\n        tensorIndex = tensor2.tensorIndex;\n      }\n      const node = layer.inboundNodes[nodeIndex];\n      if (nodesInProgress2.indexOf(node) !== -1) {\n        throw new RuntimeError(`The tensor ${tensor2.name} at layer \"${layer.name}\" is part of a cycle.`);\n      }\n      if (finishedNodes2.indexOf(node) !== -1) {\n        return;\n      }\n      this.containerNodes.add(Container.nodeKey(layer, nodeIndex));\n      if (!(layer.id in layerIndices)) {\n        layerIndices[layer.id] = Object.keys(layerIndices).length;\n      }\n      if (nodesInProgress2.indexOf(node) === -1) {\n        nodesInProgress2.push(node);\n      }\n      const numInboundLayers = node.inboundLayers.length;\n      for (let i = 0; i < numInboundLayers; i++) {\n        const x = node.inputTensors[i];\n        const layer2 = node.inboundLayers[i];\n        const nodeIndex2 = node.nodeIndices[i];\n        const tensorIndex2 = node.tensorIndices[i];\n        buildMapOfGraph(x, finishedNodes2, nodesInProgress2, layer2, nodeIndex2, tensorIndex2);\n      }\n      finishedNodes2.push(node);\n      while (nodesInProgress2.indexOf(node) >= 0) {\n        nodesInProgress2.splice(nodesInProgress2.indexOf(node), 1);\n      }\n      nodesInDecreasingDepth.push(node);\n    };\n    const finishedNodes = [];\n    const nodesInProgress = [];\n    for (const x of this.outputs) {\n      buildMapOfGraph(x, finishedNodes, nodesInProgress);\n    }\n    const reversedNodesInDecreasingDepth = nodesInDecreasingDepth.slice().reverse();\n    for (const node of reversedNodesInDecreasingDepth) {\n      nodeIDToNode[node.id] = node;\n      if (!(node.id in nodesDepths)) {\n        nodesDepths[node.id] = 0;\n      }\n      let depth = nodesDepths[node.id];\n      const previousDepth = layersDepths[node.outboundLayer.id] == null ? 0 : layersDepths[node.outboundLayer.id];\n      depth = Math.max(depth, previousDepth);\n      layersDepths[node.outboundLayer.id] = depth;\n      layerIDToLayer[node.outboundLayer.id] = node.outboundLayer;\n      nodesDepths[node.id] = depth;\n      for (let i = 0; i < node.inboundLayers.length; i++) {\n        const inboundLayer = node.inboundLayers[i];\n        const nodeIndex = node.nodeIndices[i];\n        const inboundNode = inboundLayer.inboundNodes[nodeIndex];\n        const previousDepth2 = nodesDepths[inboundNode.id] == null ? 0 : nodesDepths[inboundNode.id];\n        nodesDepths[inboundNode.id] = Math.max(depth + 1, previousDepth2);\n        nodeIDToNode[inboundNode.id] = inboundNode;\n      }\n    }\n    const nodesByDepth = {};\n    for (const nodeID in nodesDepths) {\n      const depth = nodesDepths[nodeID];\n      if (!(depth in nodesByDepth)) {\n        nodesByDepth[depth] = [];\n      }\n      nodesByDepth[depth].push(nodeIDToNode[nodeID]);\n    }\n    const layersByDepth = {};\n    for (const layerID in layersDepths) {\n      const depth = layersDepths[layerID];\n      if (!(depth in layersByDepth)) {\n        layersByDepth[depth] = [];\n      }\n      layersByDepth[depth].push(layerIDToLayer[layerID]);\n    }\n    let depthKeys = Object.keys(layersByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);\n    this.layers = [];\n    for (const depth of depthKeys) {\n      const layersForDepth = layersByDepth[depth];\n      layersForDepth.sort((a, b) => {\n        const aIndex = layerIndices[a.id];\n        const bIndex = layerIndices[b.id];\n        if (aIndex < bIndex) {\n          return -1;\n        }\n        if (aIndex > bIndex) {\n          return 1;\n        }\n        return 0;\n      });\n      for (const layer of layersForDepth) {\n        if (layer instanceof Container) {\n          this.internalContainerRefs.push(layer);\n        }\n        this.layers.push(layer);\n      }\n    }\n    this.layersByDepth = layersByDepth;\n    depthKeys = Object.keys(nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);\n    const computableTensors = this.inputs.slice();\n    const layersWithCompleteInput = [];\n    for (const depth of depthKeys) {\n      for (const node of nodesByDepth[depth]) {\n        const layer = node.outboundLayer;\n        if (layer != null) {\n          for (const x of node.inputTensors) {\n            if (computableTensors.indexOf(x) === -1) {\n              throw new RuntimeError(`Graph disconnected: cannot obtain value for tensor ${x} at layer \"${layer.name}\". The following previous layers were accessed without issue: ${layersWithCompleteInput}`);\n            }\n          }\n          for (const x of node.outputTensors) {\n            computableTensors.push(x);\n          }\n          layersWithCompleteInput.push(layer.name);\n        }\n      }\n    }\n    this.nodesByDepth = nodesByDepth;\n    const allNames = this.layers.map((x) => x.name);\n    for (const name of allNames) {\n      const numOccurrences = allNames.filter((x) => x === name).length;\n      if (numOccurrences !== 1) {\n        throw new RuntimeError(`The name \"${name}\" is used ${numOccurrences} times in the model. All layer names should be unique. Layer names: ` + JSON.stringify(allNames));\n      }\n    }\n    this.outboundNodes = [];\n    this.inboundNodes = [];\n    new Node({\n      outboundLayer: this,\n      inboundLayers: [],\n      nodeIndices: [],\n      tensorIndices: [],\n      inputTensors: this.inputs,\n      outputTensors: this.outputs,\n      inputMasks: this.inputs.map((x) => null),\n      outputMasks: this.outputs.map((x) => null),\n      inputShapes: this.inputs.map((x) => x.shape),\n      outputShapes: this.outputs.map((x) => x.shape)\n    });\n    this.built = true;\n    this._refCount = 1;\n  }\n  assertNotDisposed() {\n    if (this._refCount === 0) {\n      throw new Error(`Container '${this.name}' is already disposed.`);\n    }\n  }\n  dispose() {\n    this.assertNotDisposed();\n    const result = { refCountAfterDispose: null, numDisposedVariables: 0 };\n    if (--this._refCount === 0) {\n      for (const layer of this.layers) {\n        result.numDisposedVariables += layer.dispose().numDisposedVariables;\n      }\n      for (const container of this.internalContainerRefs) {\n        result.numDisposedVariables += container.dispose().numDisposedVariables;\n      }\n    }\n    result.refCountAfterDispose = this._refCount;\n    return result;\n  }\n  get trainable() {\n    return this.trainable_;\n  }\n  set trainable(trainable) {\n    this.layers.forEach((layer) => {\n      layer._trainableWeights.forEach((w) => w.trainable = trainable);\n    });\n    this.trainable_ = trainable;\n  }\n  get trainableWeights() {\n    if (this._trainableWeights.length > 0) {\n      throw new ValueError(\"Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.\");\n    }\n    if (!this.trainable) {\n      return [];\n    }\n    let weights = [];\n    for (const layer of this.layers) {\n      weights = weights.concat(layer.trainableWeights);\n    }\n    return weights;\n  }\n  get nonTrainableWeights() {\n    const weights = [];\n    for (const layer of this.layers) {\n      weights.push(...layer.nonTrainableWeights);\n    }\n    if (!this.trainable) {\n      const trainableWeights = [];\n      for (const layer of this.layers) {\n        trainableWeights.push(...layer.trainableWeights);\n      }\n      return trainableWeights.concat(weights);\n    }\n    return weights;\n  }\n  get weights() {\n    return this.trainableWeights.concat(this.nonTrainableWeights);\n  }\n  loadWeights(weights, strict = true) {\n    const nameToWeight = {};\n    let totalWeightsCount = 0;\n    for (const layer of this.layers) {\n      for (const weight of layer.weights) {\n        if (nameToWeight[weight.originalName] != null) {\n          throw new ValueError(`Duplicate weight name: ${weight.originalName}`);\n        }\n        nameToWeight[weight.originalName] = weight;\n        totalWeightsCount++;\n      }\n    }\n    const weightValueTuples = [];\n    for (const name in weights) {\n      let validatedName = name;\n      if (nameToWeight[name] == null) {\n        const tokens = name.split(\"/\");\n        const shortenNameArray = tokens.slice(0, -2).concat([tokens[tokens.length - 1]]);\n        validatedName = shortenNameArray.join(\"/\");\n      }\n      if (nameToWeight[validatedName] != null) {\n        weightValueTuples.push([nameToWeight[validatedName], weights[name]]);\n      } else if (strict) {\n        throw new ValueError(`Provided weight data has no target variable: ${name}`);\n      }\n      delete nameToWeight[validatedName];\n    }\n    if (strict) {\n      const unsetNames = [];\n      for (const name in nameToWeight) {\n        unsetNames.push(name);\n      }\n      if (unsetNames.length > 0) {\n        throw new ValueError(`${unsetNames.length} of ${totalWeightsCount} weights are not set: ${unsetNames}`);\n      }\n    }\n    batchSetValue(weightValueTuples);\n  }\n  updatedConfig() {\n    const theConfig = this.getConfig();\n    const modelConfig = {};\n    modelConfig[\"className\"] = this.getClassName();\n    modelConfig[\"config\"] = theConfig;\n    modelConfig[\"kerasVersion\"] = `tfjs-layers ${version2}`;\n    modelConfig[\"backend\"] = \"TensorFlow.js\";\n    return modelConfig;\n  }\n  toJSON(unused, returnString = true) {\n    const modelConfig = convertTsToPythonic(this.updatedConfig());\n    return returnString ? JSON.stringify(modelConfig) : modelConfig;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = toList(inputs);\n      const feedDict = new FeedDict();\n      for (let i = 0; i < this.inputs.length; ++i) {\n        feedDict.add(this.inputs[i], inputs[i]);\n      }\n      return execute(this.outputs, feedDict, kwargs);\n    });\n  }\n  computeMask(inputs, mask) {\n    return tidy(() => {\n      inputs = toList(inputs);\n      let masks;\n      if (mask == null) {\n        masks = pyListRepeat(null, inputs.length);\n      } else {\n        masks = toList(mask);\n      }\n      return this.runInternalGraph(inputs, masks)[1];\n    });\n  }\n  computeOutputShape(inputShape) {\n    const inputShapes = normalizeShapeList(inputShape);\n    if (inputShapes.length !== this.inputLayers.length) {\n      throw new ValueError(`Invalid inputShape argument ${inputShape}: model has ${this.inputLayers.length} tensor inputs.`);\n    }\n    const layersToOutputShapes = {};\n    for (let i = 0; i < inputShapes.length; i++) {\n      const layer = this.inputLayers[i];\n      const inputShape2 = inputShapes[i];\n      const shapeKey = layer.name + \"_0_0\";\n      layersToOutputShapes[shapeKey] = inputShape2;\n    }\n    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);\n    if (depthKeys.length > 1) {\n      for (const depth of depthKeys) {\n        const nodes = this.nodesByDepth[depth];\n        for (const node of nodes) {\n          const layer = node.outboundLayer;\n          if (this.inputLayers.map((x) => x.id).indexOf(layer.id) !== -1) {\n            continue;\n          }\n          const inputShapes2 = [];\n          for (let j = 0; j < node.inboundLayers.length; j++) {\n            const inboundLayer = node.inboundLayers[j];\n            const nodeIndex2 = node.nodeIndices[j];\n            const tensorIndex = node.tensorIndices[j];\n            const shapeKey = `${inboundLayer.name}_${nodeIndex2}_${tensorIndex}`;\n            const inputShape2 = layersToOutputShapes[shapeKey];\n            inputShapes2.push(inputShape2);\n          }\n          const outputShape = layer.computeOutputShape(singletonOrArray(inputShapes2));\n          const outputShapes2 = normalizeShapeList(outputShape);\n          const nodeIndex = layer.inboundNodes.indexOf(node);\n          for (let j = 0; j < outputShapes2.length; j++) {\n            const shapeKey = `${layer.name}_${nodeIndex}_${j}`;\n            layersToOutputShapes[shapeKey] = outputShapes2[j];\n          }\n        }\n      }\n    }\n    const outputShapes = [];\n    const outputShapeKeys = [];\n    for (let i = 0; i < this.outputLayers.length; i++) {\n      const layer = this.outputLayers[i];\n      const nodeIndex = this.outputLayersNodeIndices[i];\n      const tensorIndex = this.outputLayersTensorIndices[i];\n      const shapeKey = `${layer.name}_${nodeIndex}_${tensorIndex}`;\n      outputShapeKeys.push(shapeKey);\n    }\n    for (let i = 0; i < outputShapeKeys.length; i++) {\n      const key = outputShapeKeys[i];\n      assert2(key in layersToOutputShapes);\n      outputShapes.push(layersToOutputShapes[key]);\n    }\n    return singletonOrArray(outputShapes);\n  }\n  runInternalGraph(inputs, masks) {\n    if (masks == null) {\n      masks = pyListRepeat(null, inputs.length);\n    }\n    const tensorMap = {};\n    for (let i = 0; i < this.inputs.length; ++i) {\n      const x = this.inputs[i];\n      const y = inputs[i];\n      const mask = masks[i];\n      tensorMap[x.id] = [y, mask];\n    }\n    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);\n    for (const depth of depthKeys) {\n      const nodes = this.nodesByDepth[depth];\n      for (const node of nodes) {\n        const layer = node.outboundLayer;\n        const referenceInputTensors = node.inputTensors;\n        const referenceOutputTensors = node.outputTensors;\n        const computedData = new Array();\n        for (const x of referenceInputTensors) {\n          if (x.id in tensorMap) {\n            computedData.push(tensorMap[x.id]);\n          }\n        }\n        if (computedData.length === referenceInputTensors.length) {\n          let kwargs = {};\n          let computedTensors;\n          let computedMasks;\n          let outputTensors2;\n          let outputMasks2;\n          if (node.callArgs != null) {\n            kwargs = node.callArgs;\n          }\n          if (computedData.length === 1) {\n            const [computedTensor, computedMask] = computedData[0];\n            if (kwargs[\"mask\"] == null) {\n              kwargs[\"mask\"] = computedMask;\n            }\n            outputTensors2 = toList(layer.call(computedTensor, kwargs));\n            outputMasks2 = toList(layer.computeMask(computedTensor, computedMask));\n            computedTensors = [computedTensor];\n            computedMasks = [computedMask];\n          } else {\n            computedTensors = computedData.map((x) => x[0]);\n            computedMasks = computedData.map((x) => x[1]);\n            if (kwargs[\"mask\"] == null) {\n              kwargs[\"mask\"] = computedMasks;\n            }\n            outputTensors2 = toList(layer.call(computedTensors, kwargs));\n            outputMasks2 = toList(layer.computeMask(computedTensors, computedMasks));\n          }\n          if (layer.activityRegularizer) {\n            throw new NotImplementedError(\"LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.\");\n          }\n          for (let i = 0; i < referenceOutputTensors.length; ++i) {\n            const x = referenceOutputTensors[i];\n            const y = outputTensors2[i];\n            const mask = outputMasks2[i];\n            tensorMap[x.id] = [y, mask];\n          }\n        }\n      }\n    }\n    const outputTensors = [];\n    const outputMasks = [];\n    const outputShapes = [];\n    for (const x of this.outputs) {\n      assert2(x.id in tensorMap, `Could not compute output ${x.name} : ${x.id}`);\n      const [tensor2, mask] = tensorMap[x.id];\n      outputShapes.push(tensor2.shape);\n      outputTensors.push(tensor2);\n      outputMasks.push(mask);\n    }\n    return [outputTensors, outputMasks, outputShapes];\n  }\n  buildNodeConversionMap(layers) {\n    const nodeConversionMap = {};\n    let keptNodes;\n    for (const layer of this.layers) {\n      keptNodes = layer instanceof Container ? 1 : 0;\n      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {\n        const nodeKey = Container.nodeKey(layer, originalNodeIndex);\n        if (this.containerNodes.has(nodeKey)) {\n          nodeConversionMap[nodeKey] = keptNodes;\n          keptNodes += 1;\n        }\n      }\n    }\n    return nodeConversionMap;\n  }\n  getLayer(name, index) {\n    if (index != null) {\n      if (this.layers.length <= index) {\n        throw new ValueError(`Was asked to retrieve layer at index ${index}, but model only has ${this.layers.length} layer(s).`);\n      } else {\n        return this.layers[index];\n      }\n    } else {\n      if (name == null) {\n        throw new ValueError(\"Provide either a layer name or layer index\");\n      }\n    }\n    for (const layer of this.layers) {\n      if (layer.name === name) {\n        return layer;\n      }\n    }\n    throw new ValueError(`No such layer: ${name}`);\n  }\n  calculateLosses() {\n    return tidy(() => {\n      const losses2 = [];\n      for (const layer of this.layers) {\n        for (let nodeIndex = 0; nodeIndex < layer.inboundNodes.length; ++nodeIndex) {\n          const nodeKey = Container.nodeKey(layer, nodeIndex);\n          if (this.containerNodes.has(nodeKey)) {\n            losses2.push(...layer.calculateLosses());\n          }\n        }\n      }\n      return losses2;\n    });\n  }\n  getConfig() {\n    const config = { name: this.name };\n    const nodeConversionMap = this.buildNodeConversionMap(this.layers);\n    const layerConfigs = [];\n    for (const layer of this.layers) {\n      const layerClassName = layer.getClassName();\n      const layerConfig = layer.getConfig();\n      const filteredInboundNodes = [];\n      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {\n        const node = layer.inboundNodes[originalNodeIndex];\n        const nodeKey = Container.nodeKey(layer, originalNodeIndex);\n        let kwargs = {};\n        if (this.containerNodes.has(nodeKey)) {\n          if (node.callArgs) {\n            try {\n              JSON.stringify(node.callArgs);\n              kwargs = node.callArgs;\n            } catch (err) {\n              console.warn(`Layer ${layer.name} was passed non-serializable keyword arguments: ${node.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`);\n              kwargs = {};\n            }\n          }\n          if (node.inboundLayers.length > 0) {\n            const nodeData = [];\n            for (let i = 0; i < node.inboundLayers.length; i++) {\n              const inboundLayer = node.inboundLayers[i];\n              const nodeIndex = node.nodeIndices[i];\n              const tensorIndex = node.tensorIndices[i];\n              const nodeKey2 = Container.nodeKey(inboundLayer, nodeIndex);\n              let newNodeIndex = nodeConversionMap[nodeKey2];\n              if (newNodeIndex == null) {\n                newNodeIndex = 0;\n              }\n              nodeData.push([inboundLayer.name, newNodeIndex, tensorIndex, kwargs]);\n            }\n            filteredInboundNodes.push(nodeData);\n          }\n        }\n      }\n      const dict = {};\n      dict[\"name\"] = layer.name;\n      dict[\"className\"] = layerClassName;\n      dict[\"config\"] = layerConfig;\n      dict[\"inboundNodes\"] = filteredInboundNodes;\n      layerConfigs.push(dict);\n    }\n    config[\"layers\"] = layerConfigs;\n    const modelInputs = [];\n    for (let i = 0; i < this.inputLayers.length; i++) {\n      const layer = this.inputLayers[i];\n      const nodeIndex = this.inputLayersNodeIndices[i];\n      const nodeKey = Container.nodeKey(layer, nodeIndex);\n      if (!this.containerNodes.has(nodeKey)) {\n        continue;\n      }\n      let newNodeIndex = nodeConversionMap[nodeKey];\n      if (newNodeIndex === null || newNodeIndex === void 0) {\n        newNodeIndex = 0;\n      }\n      const tensorIndex = this.inputLayersTensorIndices[i];\n      modelInputs.push([layer.name, newNodeIndex, tensorIndex]);\n    }\n    config[\"inputLayers\"] = modelInputs;\n    const modelOutputs = [];\n    for (let i = 0; i < this.outputLayers.length; i++) {\n      const layer = this.outputLayers[i];\n      const nodeIndex = this.outputLayersNodeIndices[i];\n      const nodeKey = Container.nodeKey(layer, nodeIndex);\n      if (!this.containerNodes.has(nodeKey)) {\n        continue;\n      }\n      let newNodeIndex = nodeConversionMap[nodeKey];\n      if (newNodeIndex === null || newNodeIndex === void 0) {\n        newNodeIndex = 0;\n      }\n      const tensorIndex = this.outputLayersTensorIndices[i];\n      modelOutputs.push([layer.name, newNodeIndex, tensorIndex]);\n    }\n    config[\"outputLayers\"] = modelOutputs;\n    return config;\n  }\n  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {\n    const createdLayers = {};\n    const unprocessedNodes = {};\n    function addUnprocessedNode(layer, nodeData) {\n      if (!(layer.name in unprocessedNodes)) {\n        unprocessedNodes[layer.name] = [nodeData];\n      } else {\n        unprocessedNodes[layer.name].push(nodeData);\n      }\n    }\n    function processNode(layer, nodeData) {\n      const inputTensors2 = [];\n      let kwargs;\n      for (const inputData of nodeData) {\n        const inboundLayerName = inputData[0];\n        const inboundNodeIndex = inputData[1];\n        const inboundTensorIndex = inputData[2];\n        kwargs = inputData[3] == null ? {} : inputData[3];\n        if (!(inboundLayerName in createdLayers)) {\n          addUnprocessedNode(layer, nodeData);\n          return;\n        }\n        const inboundLayer = createdLayers[inboundLayerName];\n        if (inboundLayer.inboundNodes.length <= inboundNodeIndex) {\n          addUnprocessedNode(layer, nodeData);\n          return;\n        }\n        const inboundNode = inboundLayer.inboundNodes[inboundNodeIndex];\n        inputTensors2.push(inboundNode.outputTensors[inboundTensorIndex]);\n      }\n      if (inputTensors2.length > 0) {\n        layer.apply(singletonOrArray(inputTensors2), kwargs);\n      }\n    }\n    function processLayer(layerData) {\n      const layerName = layerData[\"name\"];\n      const layer = deserialize(layerData, config[\"customObjects\"] != null ? config[\"customObjects\"] : {});\n      layer.setFastWeightInitDuringBuild(fastWeightInit);\n      createdLayers[layerName] = layer;\n      const inboundNodesData = layerData[\"inboundNodes\"];\n      inboundNodesData.forEach((nodeData) => {\n        if (!(nodeData instanceof Array)) {\n          throw new ValueError(`Corrupted configuration, expected array for nodeData: ${nodeData}`);\n        }\n        addUnprocessedNode(layer, nodeData);\n      });\n    }\n    const name = config[\"name\"];\n    const layersFromConfig = config[\"layers\"];\n    for (const layerData of layersFromConfig) {\n      processLayer(layerData);\n    }\n    while (!isObjectEmpty(unprocessedNodes)) {\n      for (const layerData of layersFromConfig) {\n        const layer = createdLayers[layerData[\"name\"]];\n        if (layer.name in unprocessedNodes) {\n          const currentUnprocessedNodesForLayer = unprocessedNodes[layer.name];\n          delete unprocessedNodes[layer.name];\n          for (const nodeData of currentUnprocessedNodesForLayer) {\n            processNode(layer, nodeData);\n          }\n        }\n      }\n    }\n    const inputTensors = [];\n    const outputTensors = [];\n    const inputLayersFromConfig = config[\"inputLayers\"];\n    for (const layerData of inputLayersFromConfig) {\n      const layerName = layerData[0];\n      const nodeIndex = layerData[1];\n      const tensorIndex = layerData[2];\n      assert2(layerName in createdLayers);\n      const layer = createdLayers[layerName];\n      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;\n      inputTensors.push(layerOutputTensors[tensorIndex]);\n    }\n    const outputLayersFromConfig = config[\"outputLayers\"];\n    for (const layerData of outputLayersFromConfig) {\n      const layerName = layerData[0];\n      const nodeIndex = layerData[1];\n      const tensorIndex = layerData[2];\n      assert2(layerName in createdLayers);\n      const layer = createdLayers[layerName];\n      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;\n      outputTensors.push(layerOutputTensors[tensorIndex]);\n    }\n    return new cls({ inputs: inputTensors, outputs: outputTensors, name });\n  }\n  get stateful() {\n    if (this._stateful) {\n      throw new ValueError(\"Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.\");\n    }\n    for (const layer of this.layers) {\n      if (layer.stateful) {\n        return true;\n      }\n    }\n    return false;\n  }\n  resetStates() {\n    tidy(() => {\n      this.layers.forEach((layer) => {\n        if (layer.stateful) {\n          layer.resetStates();\n        }\n      });\n    });\n  }\n};\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/training_utils.js\nfunction standardizeSampleOrClassWeights(xWeight, outputNames, weightType) {\n  const numOutputs = outputNames.length;\n  if (xWeight == null || Array.isArray(xWeight) && xWeight.length === 0) {\n    return outputNames.map((name) => null);\n  }\n  if (numOutputs === 1) {\n    if (Array.isArray(xWeight) && xWeight.length === 1) {\n      return xWeight;\n    } else if (typeof xWeight === \"object\" && outputNames[0] in xWeight) {\n      return [xWeight[outputNames[0]]];\n    } else {\n      return [xWeight];\n    }\n  }\n  if (Array.isArray(xWeight)) {\n    if (xWeight.length !== numOutputs) {\n      throw new Error(`Provided ${weightType} is an array of ${xWeight.length} element(s), but the model has ${numOutputs} outputs. Make sure a set of weights is provided for each model output.`);\n    }\n    return xWeight;\n  } else if (typeof xWeight === \"object\" && Object.keys(xWeight).length > 0 && typeof xWeight[Object.keys(xWeight)[0]] === \"object\") {\n    const output = [];\n    outputNames.forEach((outputName) => {\n      if (outputName in xWeight) {\n        output.push(xWeight[outputName]);\n      } else {\n        output.push(null);\n      }\n    });\n    return output;\n  } else {\n    throw new Error(`The model has multiple (${numOutputs}) outputs, so ${weightType} must be either an array with ${numOutputs} elements or an object with ${outputNames} keys. Provided ${weightType} not understood: ${JSON.stringify(xWeight)}`);\n  }\n}\nfunction standardizeClassWeights(classWeight, outputNames) {\n  return standardizeSampleOrClassWeights(classWeight, outputNames, \"classWeight\");\n}\nasync function standardizeWeights(y, sampleWeight, classWeight, sampleWeightMode) {\n  if (sampleWeight != null || sampleWeightMode != null) {\n    throw new Error(\"Support sampleWeight is not implemented yet\");\n  }\n  if (classWeight != null) {\n    const yClasses = tidy(() => {\n      if (y.shape.length === 1) {\n        return clone(y);\n      } else if (y.shape.length === 2) {\n        if (y.shape[1] > 1) {\n          const axis = 1;\n          return argMax(y, axis);\n        } else if (y.shape[1] === 1) {\n          return reshape(y, [y.shape[0]]);\n        } else {\n          throw new Error(`Encountered unexpected last-dimension size (${y.shape[1]}) during handling of class weights. The size is expected to be >= 1.`);\n        }\n      } else {\n        throw new Error(`Unexpected rank of target (y) tensor (${y.rank}) during handling of class weights. The rank is expected to be 1 or 2.`);\n      }\n    });\n    const yClassIndices = Array.from(await yClasses.data());\n    dispose(yClasses);\n    const classSampleWeight = [];\n    yClassIndices.forEach((classIndex) => {\n      if (classWeight[classIndex] == null) {\n        throw new Error(`classWeight must contain all classes in the training data. The class ${classIndex} exists in the data but not in classWeight`);\n      } else {\n        classSampleWeight.push(classWeight[classIndex]);\n      }\n    });\n    return tensor1d(classSampleWeight, \"float32\");\n  } else {\n    return null;\n  }\n}\nfunction computeWeightedLoss2(losses2, sampleWeights) {\n  return mul(losses2, sampleWeights);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/training_dataset.js\nvar DEFAULT_VALIDATION_BATCH_SIZE = 32;\nfunction standardizeDataIteratorOutput(model2, iteratorOut) {\n  let xs;\n  let ys;\n  const iteratorOutObj = iteratorOut;\n  xs = iteratorOutObj[\"xs\"];\n  ys = iteratorOutObj[\"ys\"];\n  util_exports.assert(xs != null && ys != null, () => `A Dataset iterator for fitDataset() is expected to generate objects of the form \\`{xs: xVal, ys: yVal}\\`, where the two values may be \\`tf.Tensor\\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${iteratorOut}`);\n  const flattenedXs = flattenTensorOrArrayOrMap(\"input\", model2.inputNames, xs);\n  const flattenedYs = flattenTensorOrArrayOrMap(\"output\", model2.outputNames, ys);\n  const batchSize = flattenedXs[0].shape[0];\n  util_exports.assert(flattenedXs.length === model2.inputs.length, () => `LayersModel has ${model2.inputs.length} inputs, but the dataset provides ${flattenedXs.length} inputs.  (Expected input keys: ${JSON.stringify(model2.inputNames)})`);\n  util_exports.assert(flattenedYs.length === model2.outputs.length, () => `LayersModel has ${model2.outputs.length} outputs, but the dataset provides ${flattenedYs.length} outputs.  (Expected output keys: ${JSON.stringify(model2.outputNames)})`);\n  for (let xIndex = 0; xIndex < flattenedXs.length; xIndex++) {\n    util_exports.assert(flattenedXs[xIndex].shape[0] === batchSize, () => `Batch size mismatch: input ${model2.inputNames[xIndex]} has ${flattenedXs[xIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);\n  }\n  for (let yIndex = 0; yIndex < flattenedYs.length; yIndex++) {\n    util_exports.assert(flattenedYs[yIndex].shape[0] === batchSize, () => `Batch size mismatch: output ${model2.outputNames[yIndex]} has ${flattenedYs[yIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);\n  }\n  return { xs: flattenedXs, ys: flattenedYs };\n}\nfunction flattenTensorOrArrayOrMap(inputOrOutput, names, values) {\n  if (values instanceof Tensor) {\n    return [values];\n  } else if (Array.isArray(values)) {\n    util_exports.assert(values.length === names.length, () => `Received an array of ${values.length} Tensors, but expected ${names.length} to match the ${inputOrOutput} keys ${names}.`);\n    return values;\n  } else {\n    const result = [];\n    for (const name of names) {\n      if (values[name] == null) {\n        throw new ValueError(`The feature data generated by the dataset lacks the required ${inputOrOutput} key '${name}'.`);\n      }\n      result.push(values[name]);\n    }\n    return result;\n  }\n}\nfunction standardizeTensorValidationData(data) {\n  if (data.length === 3) {\n    throw new NotImplementedError(\"Validation with sample weights is not implemented yet.\");\n  }\n  return { xs: data[0], ys: data[1] };\n}\nasync function fitDataset(model2, dataset, args) {\n  const hasBatchesPerEpoch = args.batchesPerEpoch != null;\n  util_exports.assert(model2.optimizer != null, () => \"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig).\");\n  util_exports.assert(args != null, () => `For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call.`);\n  util_exports.assert(args.epochs != null && args.epochs > 0 && Number.isInteger(args.epochs), () => `For fitDataset(), config.epochs is expected to be a positive integer, but got ${args.epochs}`);\n  util_exports.assert(!hasBatchesPerEpoch || args.batchesPerEpoch > 0 && Number.isInteger(args.batchesPerEpoch), () => `For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${args.batchesPerEpoch}`);\n  util_exports.assert(\n    args[\"validationSplit\"] == null,\n    () => \"`validationSplit` is not supported by `fitDataset()`. Use validationData instead.\"\n  );\n  if (model2.isTraining) {\n    throw new Error(\"Cannot start training because another fit() call is ongoing.\");\n  }\n  model2.isTraining = true;\n  try {\n    const doValidation = args.validationData != null;\n    let valXs;\n    let valYs;\n    if (doValidation) {\n      if (isDatasetObject(args.validationData)) {\n        util_exports.assert(args.validationBatches == null || args.validationBatches > 0 && Number.isInteger(args.validationBatches), () => `For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${args.validationBatches}`);\n      } else {\n        const validationData = standardizeTensorValidationData(args.validationData);\n        valXs = validationData.xs;\n        valYs = validationData.ys;\n      }\n    }\n    const trainFunction = model2.makeTrainFunction();\n    const outLabels = model2.getDedupedMetricsNames();\n    let callbackMetrics;\n    if (doValidation) {\n      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => \"val_\" + n));\n    } else {\n      callbackMetrics = outLabels.slice();\n    }\n    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);\n    const verbose = args.verbose == null ? 1 : args.verbose;\n    const { callbackList, history } = configureCallbacks(\n      callbacks2,\n      verbose,\n      args.epochs,\n      null,\n      null,\n      getStepsPerEpoch(dataset, args),\n      null,\n      doValidation,\n      callbackMetrics\n    );\n    callbackList.setModel(model2);\n    model2.history = history;\n    await callbackList.onTrainBegin();\n    model2.stopTraining_ = false;\n    let epoch = args.initialEpoch == null ? 0 : args.initialEpoch;\n    let dataIterator = await dataset.iterator();\n    while (epoch < args.epochs) {\n      const epochLogs = {};\n      await callbackList.onEpochBegin(epoch);\n      let stepsDone = 0;\n      let batchIndex = 0;\n      if (!hasBatchesPerEpoch) {\n        dataIterator = await dataset.iterator();\n      }\n      while (hasBatchesPerEpoch ? stepsDone < args.batchesPerEpoch : true) {\n        const iteratorOut = await dataIterator.next();\n        if (hasBatchesPerEpoch && iteratorOut.done) {\n          console.warn(`You provided \\`batchesPerEpoch\\` as ${args.batchesPerEpoch}, but your dataset iterator ran out of data after ${stepsDone} batches; interrupting training. Make sure that your dataset can generate at least \\`batchesPerEpoch * epochs\\` batches (in this case, ${args.batchesPerEpoch * args.epochs} batches). You may need to use the repeat() function when building your dataset.`);\n          break;\n        }\n        if (iteratorOut.value != null) {\n          const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);\n          const batchLogs = {};\n          batchLogs[\"batch\"] = batchIndex;\n          batchLogs[\"size\"] = xs[0].shape[0];\n          await callbackList.onBatchBegin(batchIndex, batchLogs);\n          const sampleWeights = [];\n          if (args.classWeight != null) {\n            const standardClassWeights = standardizeClassWeights(args.classWeight, model2.outputNames);\n            for (let i = 0; i < standardClassWeights.length; ++i) {\n              sampleWeights.push(await standardizeWeights(ys[i], null, standardClassWeights[i]));\n            }\n          }\n          const ins = xs.concat(ys).concat(sampleWeights);\n          const outs = trainFunction(ins);\n          dispose(ins);\n          for (let i = 0; i < outLabels.length; ++i) {\n            const label = outLabels[i];\n            const out = outs[i];\n            batchLogs[label] = out;\n            keep(out);\n          }\n          await callbackList.onBatchEnd(batchIndex, batchLogs);\n          disposeTensorsInLogs(batchLogs);\n          batchIndex++;\n          stepsDone++;\n        }\n        if (hasBatchesPerEpoch ? stepsDone >= args.batchesPerEpoch : iteratorOut.done) {\n          if (doValidation) {\n            let valOuts;\n            if (isDatasetObject(args.validationData)) {\n              valOuts = toList(await model2.evaluateDataset(args.validationData, { batches: args.validationBatches }));\n            } else {\n              valOuts = toList(model2.evaluate(valXs, valYs, {\n                batchSize: args.validationBatchSize == null ? DEFAULT_VALIDATION_BATCH_SIZE : args.validationBatchSize,\n                verbose: 0\n              }));\n            }\n            for (let i = 0; i < model2.metricsNames.length; ++i) {\n              epochLogs[`val_${model2.metricsNames[i]}`] = valOuts[i];\n            }\n          }\n          break;\n        }\n        if (model2.stopTraining_) {\n          break;\n        }\n      }\n      await callbackList.onEpochEnd(epoch, epochLogs);\n      epoch++;\n      if (model2.stopTraining_) {\n        break;\n      }\n    }\n    await callbackList.onTrainEnd();\n    await model2.history.syncData();\n    return model2.history;\n  } finally {\n    model2.isTraining = false;\n  }\n}\nfunction getStepsPerEpoch(dataset, args) {\n  let stepsPerEpoch = null;\n  if (args.batchesPerEpoch != null) {\n    stepsPerEpoch = args.batchesPerEpoch;\n  } else if (Number.isFinite(dataset.size)) {\n    stepsPerEpoch = dataset.size;\n  }\n  return stepsPerEpoch;\n}\nfunction isDatasetObject(dataset) {\n  return typeof dataset.iterator === \"function\";\n}\nfunction isLazyIteratorObject(iterator) {\n  return typeof iterator.next === \"function\";\n}\nasync function evaluateDataset(model2, dataset, args) {\n  args = args || {};\n  const hasBatches = args.batches != null;\n  const f = model2.testFunction;\n  let outs = [];\n  if (args.verbose > 0) {\n    throw new NotImplementedError(\"Verbose mode is not implemented yet.\");\n  }\n  util_exports.assert(!hasBatches || args.batches > 0 && Number.isInteger(args.batches), () => `Test loop expects \\`batches\\` to be a positive integer, but received ${JSON.stringify(args.batches)}`);\n  const dataIterator = isLazyIteratorObject(dataset) ? dataset : await dataset.iterator();\n  let numExamples = 0;\n  let batch = 0;\n  while (hasBatches ? batch < args.batches : true) {\n    const iteratorOut = await dataIterator.next();\n    outs = tidy(() => {\n      if (iteratorOut.value) {\n        const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);\n        const xsAndYs = xs.concat(ys);\n        const batchOuts = tidy(() => f(xsAndYs));\n        dispose(xsAndYs);\n        if (batch === 0) {\n          for (let i = 0; i < batchOuts.length; ++i) {\n            outs.push(scalar(0));\n          }\n        }\n        const batchSize = xsAndYs[0].shape[0];\n        for (let i = 0; i < batchOuts.length; ++i) {\n          const batchOut = batchOuts[i];\n          const oldScalar = outs[i];\n          outs[i] = tidy(() => add2(outs[i], mul(batchSize, batchOut)));\n          if (batch > 0) {\n            dispose(oldScalar);\n          }\n        }\n        dispose(batchOuts);\n        numExamples += batchSize;\n        ++batch;\n      }\n      return outs;\n    });\n    if (iteratorOut.done) {\n      if (hasBatches) {\n        console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \\`batches\\` batches (in this case, ${args.batches} batches). You may need to use the repeat() function when building your dataset.`);\n      }\n      break;\n    }\n  }\n  for (let i = 0; i < outs.length; ++i) {\n    const oldScalar = outs[i];\n    outs[i] = div(outs[i], numExamples);\n    dispose(oldScalar);\n  }\n  return singletonOrArray(outs);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/training_tensors.js\nfunction checkBatchSize(batchSize) {\n  util_exports.assert(batchSize > 0 && Number.isInteger(batchSize), () => `batchSize is required to be a positive integer, but got ${batchSize}`);\n}\nfunction sliceArrays(arrays, start, stop) {\n  if (arrays == null) {\n    return [null];\n  } else if (Array.isArray(arrays)) {\n    return arrays.map((array2) => sliceAlongFirstAxis(array2, start, stop - start));\n  } else {\n    return sliceAlongFirstAxis(arrays, start, stop - start);\n  }\n}\nfunction sliceArraysByIndices(arrays, indices) {\n  return tidy(() => {\n    if (arrays == null) {\n      return null;\n    } else if (Array.isArray(arrays)) {\n      return arrays.map((array2) => sliceArraysByIndices(array2, indices));\n    } else {\n      return gather2(arrays, indices.dtype === \"int32\" ? indices : cast(indices, \"int32\"));\n    }\n  });\n}\nfunction makeBatches(size, batchSize) {\n  const output = [];\n  let batchStart = 0;\n  let batchEnd = null;\n  while (batchStart < size) {\n    batchEnd = batchStart + batchSize;\n    if (batchEnd >= size) {\n      batchEnd = size;\n    }\n    output.push([batchStart, batchEnd]);\n    batchStart = batchEnd;\n  }\n  return output;\n}\nasync function fitLoop(model2, f, ins, outLabels, batchSize, epochs, verbose, callbacks2, valF, valIns, shuffle2, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps) {\n  if (batchSize == null) {\n    batchSize = 32;\n  }\n  if (epochs == null) {\n    epochs = 1;\n  }\n  if (shuffle2 == null) {\n    shuffle2 = true;\n  }\n  if (initialEpoch == null) {\n    initialEpoch = 0;\n  }\n  let doValidation = false;\n  if (valF != null && valIns != null) {\n    doValidation = true;\n  }\n  if (validationSteps != null) {\n    doValidation = true;\n    if (stepsPerEpoch == null) {\n      throw new ValueError(\"Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.\");\n    }\n  }\n  const numTrainSamples = model2.checkNumSamples(ins, batchSize, stepsPerEpoch, \"steps_per_epoch\");\n  let indexArray;\n  if (numTrainSamples != null) {\n    indexArray = range2(0, numTrainSamples);\n  }\n  if (verbose == null) {\n    verbose = 1;\n  }\n  const { callbackList, history } = configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics);\n  callbackList.setModel(model2);\n  model2.history = history;\n  await callbackList.onTrainBegin();\n  model2.stopTraining_ = false;\n  for (let epoch = initialEpoch; epoch < epochs; ++epoch) {\n    await callbackList.onEpochBegin(epoch);\n    const epochLogs = {};\n    if (stepsPerEpoch != null) {\n      throw new NotImplementedError(\"stepsPerEpoch mode is not implemented yet.\");\n    } else {\n      if (shuffle2 === \"batch\") {\n        throw new NotImplementedError(\"batch shuffling is not implemneted yet\");\n      } else if (shuffle2) {\n        util_exports.shuffle(indexArray);\n      }\n      const epochIndexArray1D = tensor1d(indexArray);\n      const batches = makeBatches(numTrainSamples, batchSize);\n      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {\n        const batchLogs = {};\n        await callbackList.onBatchBegin(batchIndex, batchLogs);\n        tidy(() => {\n          const batchStart = batches[batchIndex][0];\n          const batchEnd = batches[batchIndex][1];\n          const batchIds = sliceAlongFirstAxis(epochIndexArray1D, batchStart, batchEnd - batchStart);\n          batchLogs[\"batch\"] = batchIndex;\n          batchLogs[\"size\"] = batchEnd - batchStart;\n          const insBatch = sliceArraysByIndices(ins, batchIds);\n          const outs = f(insBatch);\n          for (let i = 0; i < outLabels.length; ++i) {\n            const label = outLabels[i];\n            const out = outs[i];\n            batchLogs[label] = out;\n            keep(out);\n          }\n          if (batchIndex === batches.length - 1) {\n            if (doValidation) {\n              const valOuts = model2.testLoop(valF, valIns, batchSize);\n              for (let i = 0; i < outLabels.length; ++i) {\n                const label = outLabels[i];\n                const out = valOuts[i];\n                keep(out);\n                epochLogs[\"val_\" + label] = out;\n              }\n            }\n          }\n        });\n        await callbackList.onBatchEnd(batchIndex, batchLogs);\n        disposeTensorsInLogs(batchLogs);\n        if (model2.stopTraining_) {\n          break;\n        }\n      }\n      epochIndexArray1D.dispose();\n    }\n    await callbackList.onEpochEnd(epoch, epochLogs);\n    if (model2.stopTraining_) {\n      break;\n    }\n  }\n  await callbackList.onTrainEnd();\n  await model2.history.syncData();\n  return model2.history;\n}\nasync function fitTensors(model2, x, y, args = {}) {\n  if (model2.isTraining) {\n    throw new Error(\"Cannot start training because another fit() call is ongoing.\");\n  }\n  model2.isTraining = true;\n  let inputs;\n  let targets;\n  let originalInputs;\n  let originalTargets;\n  let inputValX;\n  let inputValY;\n  let valX;\n  let valY;\n  let sampleWeights;\n  try {\n    const batchSize = args.batchSize == null ? 32 : args.batchSize;\n    checkBatchSize(batchSize);\n    const checkBatchAxis = false;\n    const standardizedOuts = await model2.standardizeUserData(x, y, args.sampleWeight, args.classWeight, checkBatchAxis, batchSize);\n    inputs = standardizedOuts[0];\n    targets = standardizedOuts[1];\n    sampleWeights = standardizedOuts[2];\n    let doValidation = false;\n    let valIns;\n    if (args.validationData != null && args.validationData.length > 0) {\n      doValidation = true;\n      if (args.validationData.length === 2) {\n        inputValX = args.validationData[0];\n        inputValY = args.validationData[1];\n      } else if (args.validationData.length === 3) {\n        throw new NotImplementedError(\"validationData including sample weights is not supported yet.\");\n      } else {\n        throw new ValueError(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${args.validationData} is invalid.`);\n      }\n      const checkBatchAxis2 = true;\n      const valStandardized = await model2.standardizeUserData(inputValX, inputValY, null, null, checkBatchAxis2, batchSize);\n      valX = valStandardized[0];\n      valY = valStandardized[1];\n      valIns = valX.concat(valY);\n    } else if (args.validationSplit != null && args.validationSplit > 0 && args.validationSplit < 1) {\n      doValidation = true;\n      const splitAt = Math.floor(inputs[0].shape[0] * (1 - args.validationSplit));\n      const originalBatchSize = inputs[0].shape[0];\n      valX = sliceArrays(inputs, splitAt, originalBatchSize);\n      originalInputs = inputs;\n      inputs = sliceArrays(inputs, 0, splitAt);\n      valY = sliceArrays(targets, splitAt, originalBatchSize);\n      originalTargets = targets;\n      targets = sliceArrays(targets, 0, splitAt);\n      valIns = valX.concat(valY);\n    } else if (args.validationSteps != null) {\n      doValidation = true;\n    }\n    const ins = inputs.concat(targets).concat(sampleWeights);\n    model2.checkTrainableWeightsConsistency();\n    const trainFunction = model2.makeTrainFunction();\n    const outLabels = model2.getDedupedMetricsNames();\n    let valFunction;\n    let callbackMetrics;\n    if (doValidation) {\n      model2.makeTestFunction();\n      valFunction = model2.testFunction;\n      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => \"val_\" + n));\n    } else {\n      valFunction = null;\n      valIns = [];\n      callbackMetrics = outLabels.slice();\n    }\n    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);\n    const out = await fitLoop(model2, trainFunction, ins, outLabels, batchSize, args.epochs, args.verbose, callbacks2, valFunction, valIns, args.shuffle, callbackMetrics, args.initialEpoch, null, null);\n    return out;\n  } finally {\n    model2.isTraining = false;\n    disposeNewTensors(inputs, x);\n    disposeNewTensors(targets, y);\n    disposeNewTensors(originalInputs, x);\n    disposeNewTensors(originalTargets, y);\n    disposeNewTensors(valX, inputValX);\n    disposeNewTensors(valY, inputValY);\n    if (sampleWeights != null) {\n      dispose(sampleWeights);\n    }\n  }\n}\nfunction ensureTensorsRank2OrHigher(tensors) {\n  const outs = [];\n  if (tensors instanceof Tensor) {\n    tensors = [tensors];\n  }\n  for (let i = 0; i < tensors.length; ++i) {\n    const tensor2 = tensors[i];\n    if (tensor2.rank === 1) {\n      outs.push(expandDims2(tensor2, 1));\n    } else if (tensor2.rank === 0) {\n      throw new Error(\"Expected tensor to be at least 1D, but received a 0D tensor (scalar).\");\n    } else {\n      outs.push(tensor2);\n    }\n  }\n  return outs;\n}\nfunction disposeNewTensors(tensors, refTensors) {\n  if (tensors == null) {\n    return;\n  }\n  const oldTensorIds = [];\n  if (refTensors instanceof Tensor) {\n    oldTensorIds.push(refTensors.id);\n  } else if (Array.isArray(refTensors)) {\n    refTensors.forEach((t) => oldTensorIds.push(t.id));\n  } else if (refTensors != null) {\n    for (const name in refTensors) {\n      const oldTensor = refTensors[name];\n      oldTensorIds.push(oldTensor.id);\n    }\n  }\n  const tensorsToDispose = [];\n  if (tensors instanceof Tensor) {\n    if (oldTensorIds.indexOf(tensors.id) === -1) {\n      tensorsToDispose.push(tensors);\n    }\n  } else if (Array.isArray(tensors)) {\n    tensors.forEach((t) => {\n      if (oldTensorIds.indexOf(t.id) === -1) {\n        tensorsToDispose.push(t);\n      }\n    });\n  } else if (tensors != null) {\n    for (const name in tensors) {\n      const tensor2 = tensors[name];\n      if (oldTensorIds.indexOf(tensor2.id) === -1) {\n        tensorsToDispose.push(tensor2);\n      }\n    }\n  }\n  tensorsToDispose.forEach((t) => {\n    if (!t.isDisposed) {\n      t.dispose();\n    }\n  });\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/engine/training.js\nfunction isDataTensor(x) {\n  return x instanceof Tensor;\n}\nfunction isDataArray(x) {\n  return Array.isArray(x);\n}\nfunction isDataDict(x) {\n  return !isDataTensor(x) && !isDataArray(x);\n}\nfunction standardizeInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = \"\") {\n  if (names == null || names.length === 0) {\n    if (data != null) {\n      let gotUnexpectedData = false;\n      if (isDataArray(data) && data.length > 0) {\n        gotUnexpectedData = true;\n      } else if (isDataDict(data)) {\n        for (const key in data) {\n          if (data.hasOwnProperty(key)) {\n            gotUnexpectedData = true;\n            break;\n          }\n        }\n      } else {\n        gotUnexpectedData = true;\n      }\n      if (gotUnexpectedData) {\n        throw new ValueError(`Error when checking model ${exceptionPrefix} expected no data, but got ${data}`);\n      }\n    }\n    return [];\n  }\n  if (data == null) {\n    return names.map((name) => null);\n  }\n  let arrays;\n  if (isDataDict(data)) {\n    data = data;\n    arrays = [];\n    for (const name of names) {\n      if (data[name] == null) {\n        throw new ValueError(`No data provided for \"${name}\". Need data for each key in: ${names}`);\n      }\n      arrays.push(data[name]);\n    }\n  } else if (isDataArray(data)) {\n    data = data;\n    if (data.length !== names.length) {\n      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${names.length} Tensor(s), but instead got the following list of Tensor(s): ${data}`);\n    }\n    arrays = data;\n  } else {\n    data = data;\n    if (names.length > 1) {\n      throw new ValueError(`The model ${exceptionPrefix} expects ${names.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${data.shape}`);\n    }\n    arrays = [data];\n  }\n  arrays = ensureTensorsRank2OrHigher(arrays);\n  if (shapes != null) {\n    for (let i = 0; i < names.length; ++i) {\n      if (shapes[i] == null) {\n        continue;\n      }\n      const array2 = arrays[i];\n      if (array2.shape.length !== shapes[i].length) {\n        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s). but got array with shape ${array2.shape}`);\n      }\n      for (let j = 0; j < shapes[i].length; ++j) {\n        if (j === 0 && !checkBatchAxis) {\n          continue;\n        }\n        const dim = array2.shape[j];\n        const refDim = shapes[i][j];\n        if (refDim != null && refDim >= 0 && dim !== refDim) {\n          throw new ValueError(`${exceptionPrefix} expected a batch of elements where each example has shape [${shapes[i].slice(1, shapes[i].length)}] (i.e.,tensor shape [*,${shapes[i].slice(1, shapes[i].length)}]) but the ${exceptionPrefix} received an input with ${array2.shape[0]} examples, each with shape [${array2.shape.slice(1, array2.shape.length)}] (tensor shape [${array2.shape}])`);\n        }\n      }\n    }\n  }\n  return arrays;\n}\nfunction checkArrayLengths(inputs, targets, weights) {\n  const setX = unique2(inputs.map((input2) => input2.shape[0]));\n  setX.sort();\n  const setY = unique2(targets.map((target) => target.shape[0]));\n  setY.sort();\n  if (setX.length > 1) {\n    throw new ValueError(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(inputs.map((input2) => input2.shape))}`);\n  }\n  if (setY.length > 1) {\n    throw new ValueError(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(targets.map((target) => target.shape))}`);\n  }\n  if (setX.length > 0 && setY.length > 0 && !util_exports.arraysEqual(setX, setY)) {\n    throw new ValueError(`Input Tensors should have the same number of samples as target Tensors. Found ${setX[0]} input sample(s) and ${setY[0]} target sample(s).`);\n  }\n}\nfunction checkLossAndTargetCompatibility(targets, lossFns, outputShapes) {\n  const keyLosses = [\n    meanSquaredError2,\n    binaryCrossentropy,\n    categoricalCrossentropy\n  ];\n  for (let i = 0; i < targets.length; ++i) {\n    const y = targets[i];\n    const loss = lossFns[i];\n    const shape = outputShapes[i];\n    if (loss == null) {\n      continue;\n    }\n    if (loss === categoricalCrossentropy) {\n      if (y.shape[y.shape.length - 1] === 1) {\n        throw new ValueError(`You are passing a target array of shape ${y.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);\n      }\n    }\n    if (keyLosses.indexOf(loss) !== -1) {\n      const slicedYShape = y.shape.slice(1);\n      const slicedShape = shape.slice(1);\n      for (let j = 0; j < slicedYShape.length; ++j) {\n        const targetDim = slicedYShape[j];\n        const outDim = slicedShape[j];\n        if (outDim != null && targetDim !== outDim) {\n          throw new ValueError(`A target Tensor with shape ${y.shape} was passed for an output of shape ${shape}, while using a loss function that expects targets to have the same shape as the output.`);\n        }\n      }\n    }\n  }\n}\nfunction checkInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = \"\") {\n  let arrays;\n  if (Array.isArray(data)) {\n    if (data.length !== names.length) {\n      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${names.length} Tensor(s), but instead got ${data.length} Tensors(s).`);\n    }\n    arrays = data;\n  } else {\n    if (names.length > 1) {\n      throw new ValueError(`The model expects ${names.length} ${exceptionPrefix} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(data.shape)}.`);\n    }\n    arrays = [data];\n  }\n  if (shapes != null) {\n    for (let i = 0; i < names.length; ++i) {\n      if (shapes[i] == null) {\n        continue;\n      }\n      const array2 = arrays[i];\n      if (array2.shape.length !== shapes[i].length) {\n        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s), but got array with shape ${JSON.stringify(array2.shape)}`);\n      }\n      for (let j = 0; j < shapes[i].length; ++j) {\n        if (j === 0 && !checkBatchAxis) {\n          continue;\n        }\n        const dim = array2.shape[j];\n        const refDim = shapes[i][j];\n        if (refDim != null) {\n          if (refDim !== dim) {\n            throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have shape ${JSON.stringify(shapes[i])} but got array with shape ${JSON.stringify(array2.shape)}.`);\n          }\n        }\n      }\n    }\n  }\n}\nfunction collectMetrics(metrics, outputNames) {\n  if (metrics == null || Array.isArray(metrics) && metrics.length === 0) {\n    return outputNames.map((name) => []);\n  }\n  let wrappedMetrics;\n  if (typeof metrics === \"string\" || typeof metrics === \"function\") {\n    wrappedMetrics = [metrics];\n  } else if (Array.isArray(metrics) || typeof metrics === \"object\") {\n    wrappedMetrics = metrics;\n  } else {\n    throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${metrics}`);\n  }\n  if (Array.isArray(wrappedMetrics)) {\n    return outputNames.map((name) => wrappedMetrics);\n  } else {\n    const nestedMetrics = [];\n    for (const name of outputNames) {\n      let outputMetrics = wrappedMetrics.hasOwnProperty(name) ? wrappedMetrics[name] : [];\n      if (!Array.isArray(outputMetrics)) {\n        outputMetrics = [outputMetrics];\n      }\n      nestedMetrics.push(outputMetrics);\n    }\n    return nestedMetrics;\n  }\n}\nvar LAYERS_MODEL_FORMAT_NAME = \"layers-model\";\nvar LayersModel = class extends Container {\n  constructor(args) {\n    super(args);\n    this.isTraining = false;\n  }\n  summary(lineLength, positions, printFn = console.log) {\n    if (!this.built) {\n      throw new ValueError(`This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).`);\n    }\n    printSummary(this, lineLength, positions, printFn);\n  }\n  compile(args) {\n    if (args.loss == null) {\n      args.loss = [];\n    }\n    this.loss = args.loss;\n    if (typeof args.optimizer === \"string\") {\n      this.optimizer_ = getOptimizer(args.optimizer);\n      this.isOptimizerOwned = true;\n    } else {\n      if (!(args.optimizer instanceof Optimizer)) {\n        throw new ValueError(`User-defined optimizer must be an instance of tf.Optimizer.`);\n      }\n      this.optimizer_ = args.optimizer;\n      this.isOptimizerOwned = false;\n    }\n    let lossFunctions = [];\n    if (!Array.isArray(args.loss) && typeof args.loss !== \"string\" && typeof args.loss !== \"function\") {\n      args.loss = args.loss;\n      for (const name in args.loss) {\n        if (this.outputNames.indexOf(name) === -1) {\n          throw new ValueError(`Unknown entry in loss dictionary: \"${name}\". Only expected the following keys: ${this.outputNames}`);\n        }\n      }\n      for (const name of this.outputNames) {\n        if (args.loss[name] == null) {\n          console.warn(`Output \"${name}\" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${name} during training`);\n        }\n        lossFunctions.push(get(args.loss[name]));\n      }\n    } else if (Array.isArray(args.loss)) {\n      if (args.loss.length !== this.outputs.length) {\n        throw new ValueError(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${args.loss}.`);\n      }\n      const theLosses = args.loss;\n      lossFunctions = theLosses.map((l) => get(l));\n    } else {\n      const lossFunction = get(args.loss);\n      this.outputs.forEach((_) => {\n        lossFunctions.push(lossFunction);\n      });\n    }\n    this.lossFunctions = lossFunctions;\n    this.feedOutputNames = [];\n    this.feedOutputShapes = [];\n    this.feedLossFns = [];\n    for (let i = 0; i < this.outputs.length; ++i) {\n      const shape = this.internalOutputShapes[i];\n      const name = this.outputNames[i];\n      this.feedOutputNames.push(name);\n      this.feedOutputShapes.push(shape);\n      this.feedLossFns.push(this.lossFunctions[i]);\n    }\n    const skipTargetIndices = [];\n    this.metrics = args.metrics;\n    this.metricsNames = [\"loss\"];\n    this.metricsTensors = [];\n    nameScope(\"loss\", () => {\n      for (let i = 0; i < this.outputs.length; ++i) {\n        if (skipTargetIndices.indexOf(i) !== -1) {\n          continue;\n        }\n        const weightedLoss = this.lossFunctions[i];\n        if (this.outputs.length > 1) {\n          this.metricsTensors.push([weightedLoss, i]);\n          this.metricsNames.push(this.outputNames[i] + \"_loss\");\n        }\n      }\n    });\n    const nestedMetrics = collectMetrics(args.metrics, this.outputNames);\n    const appendMetric = (outputIndex, metricName, metricTensor) => {\n      if (this.outputNames.length > 1) {\n        metricName = this.outputNames[outputIndex] + \"_\" + metricName;\n      }\n      this.metricsNames.push(metricName);\n      this.metricsTensors.push([metricTensor, outputIndex]);\n    };\n    nameScope(\"metric\", () => {\n      for (let i = 0; i < this.outputs.length; ++i) {\n        if (skipTargetIndices.indexOf(i) !== -1) {\n          continue;\n        }\n        const outputMetrics = nestedMetrics[i];\n        const handleMetrics = (metrics) => {\n          const metricNamePrefix = \"\";\n          let metricName;\n          let accFn;\n          let weightedMetricFn;\n          for (const metric of metrics) {\n            if (typeof metric === \"string\" && [\"accuracy\", \"acc\", \"crossentropy\", \"ce\"].indexOf(metric) !== -1) {\n              const outputShape = this.internalOutputShapes[i];\n              if (outputShape[outputShape.length - 1] === 1 || this.lossFunctions[i] === binaryCrossentropy) {\n                if ([\"accuracy\", \"acc\"].indexOf(metric) !== -1) {\n                  accFn = binaryAccuracy;\n                } else if ([\"crossentropy\", \"ce\"].indexOf(metric) !== -1) {\n                  accFn = binaryCrossentropy2;\n                }\n              } else if (this.lossFunctions[i] === sparseCategoricalCrossentropy) {\n                if ([\"accuracy\", \"acc\"].indexOf(metric) !== -1) {\n                  accFn = sparseCategoricalAccuracy;\n                } else if ([\"crossentropy\", \"ce\"].indexOf(metric) !== -1) {\n                  accFn = sparseCategoricalCrossentropy2;\n                }\n              } else {\n                if ([\"accuracy\", \"acc\"].indexOf(metric) !== -1) {\n                  accFn = categoricalAccuracy;\n                } else if ([\"crossentropy\", \"ce\"].indexOf(metric) !== -1) {\n                  accFn = categoricalCrossentropy2;\n                }\n              }\n              let suffix;\n              if ([\"accuracy\", \"acc\"].indexOf(metric) !== -1) {\n                suffix = \"acc\";\n              } else if ([\"crossentropy\", \"ce\"].indexOf(metric) !== -1) {\n                suffix = \"ce\";\n              }\n              weightedMetricFn = accFn;\n              metricName = metricNamePrefix + suffix;\n            } else {\n              const metricFn = get2(metric);\n              weightedMetricFn = metricFn;\n              metricName = metricNamePrefix + getLossOrMetricName(metric);\n            }\n            let metricResult;\n            nameScope(metricName, () => {\n              metricResult = weightedMetricFn;\n            });\n            appendMetric(i, metricName, metricResult);\n          }\n        };\n        handleMetrics(outputMetrics);\n      }\n    });\n    this.collectedTrainableWeights = this.trainableWeights;\n  }\n  checkTrainableWeightsConsistency() {\n    if (this.collectedTrainableWeights == null) {\n      return;\n    }\n    if (this.trainableWeights.length !== this.collectedTrainableWeights.length) {\n      console.warn(\"Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?\");\n    }\n  }\n  evaluate(x, y, args = {}) {\n    const batchSize = args.batchSize == null ? 32 : args.batchSize;\n    checkBatchSize(batchSize);\n    const checkBatchAxis = true;\n    const standardizedOuts = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);\n    try {\n      const ins = standardizedOuts[0].concat(standardizedOuts[1]);\n      this.makeTestFunction();\n      const f = this.testFunction;\n      const testOuts = this.testLoop(f, ins, batchSize, args.verbose, args.steps);\n      return singletonOrArray(testOuts);\n    } finally {\n      disposeNewTensors(standardizedOuts[0], x);\n      disposeNewTensors(standardizedOuts[1], y);\n    }\n  }\n  async evaluateDataset(dataset, args) {\n    this.makeTestFunction();\n    return evaluateDataset(this, dataset, args);\n  }\n  checkNumSamples(ins, batchSize, steps, stepsName = \"steps\") {\n    let numSamples;\n    if (steps != null) {\n      numSamples = null;\n      if (batchSize != null) {\n        throw new ValueError(`If ${stepsName} is set, batchSize must be null or undefined.Got batchSize = ${batchSize}`);\n      }\n    } else if (ins != null) {\n      if (Array.isArray(ins)) {\n        numSamples = ins[0].shape[0];\n      } else {\n        numSamples = ins.shape[0];\n      }\n    } else {\n      throw new ValueError(`Either the input data should have a defined shape, or ${stepsName} shoud be specified.`);\n    }\n    return numSamples;\n  }\n  execute(inputs, outputs) {\n    if (Array.isArray(outputs) && outputs.length === 0) {\n      throw new ValueError(\"`outputs` is an empty Array, which is not allowed.\");\n    }\n    const outputsIsArray = Array.isArray(outputs);\n    const outputNames = outputsIsArray ? outputs : [outputs];\n    const outputSymbolicTensors = this.retrieveSymbolicTensors(outputNames);\n    const feedDict = new FeedDict();\n    if (inputs instanceof Tensor) {\n      inputs = [inputs];\n    }\n    if (Array.isArray(inputs)) {\n      if (inputs.length !== this.inputs.length) {\n        throw new ValueError(`The number of inputs provided (${inputs.length}) does not match the number of inputs of this model (${this.inputs.length}).`);\n      }\n      for (let i = 0; i < this.inputs.length; ++i) {\n        feedDict.add(this.inputs[i], inputs[i]);\n      }\n    } else {\n      for (const input2 of this.inputs) {\n        const tensorValue = inputs[input2.name];\n        if (tensorValue == null) {\n          throw new ValueError(`No value is provided for the model's input ${input2.name}`);\n        }\n        feedDict.add(input2, tensorValue);\n      }\n    }\n    const executeOutputs = execute(outputSymbolicTensors, feedDict);\n    return outputsIsArray ? executeOutputs : executeOutputs[0];\n  }\n  retrieveSymbolicTensors(symbolicTensorNames) {\n    const outputSymbolicTensors = pyListRepeat(null, symbolicTensorNames.length);\n    let outputsRemaining = symbolicTensorNames.length;\n    for (const layer of this.layers) {\n      const layerOutputs = Array.isArray(layer.output) ? layer.output : [layer.output];\n      const layerOutputNames = layerOutputs.map((output) => output.name);\n      for (let i = 0; i < symbolicTensorNames.length; ++i) {\n        const index = layerOutputNames.indexOf(symbolicTensorNames[i]);\n        if (index !== -1) {\n          outputSymbolicTensors[i] = layerOutputs[index];\n          outputsRemaining--;\n        }\n        if (outputsRemaining === 0) {\n          break;\n        }\n      }\n      if (outputsRemaining === 0) {\n        break;\n      }\n    }\n    if (outputsRemaining > 0) {\n      const remainingNames = [];\n      outputSymbolicTensors.forEach((tensor2, i) => {\n        if (tensor2 == null) {\n          remainingNames.push(symbolicTensorNames[i]);\n        }\n      });\n      throw new ValueError(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(remainingNames)}`);\n    }\n    return outputSymbolicTensors;\n  }\n  predictLoop(ins, batchSize = 32, verbose = false) {\n    return tidy(() => {\n      const numSamples = this.checkNumSamples(ins);\n      if (verbose) {\n        throw new NotImplementedError(\"Verbose predictLoop() is not implemented yet.\");\n      }\n      const batches = makeBatches(numSamples, batchSize);\n      const outsBatches = this.outputs.map((output) => []);\n      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {\n        const batchOuts = tidy(() => {\n          const batchStart = batches[batchIndex][0];\n          const batchEnd = batches[batchIndex][1];\n          const insBatch = sliceArrays(ins, batchStart, batchEnd);\n          const feeds = [];\n          if (Array.isArray(insBatch)) {\n            for (let i = 0; i < insBatch.length; ++i) {\n              feeds.push({ key: this.inputs[i], value: insBatch[i] });\n            }\n          } else {\n            feeds.push({ key: this.inputs[0], value: insBatch });\n          }\n          const feedDict = new FeedDict(feeds);\n          return execute(this.outputs, feedDict);\n        });\n        batchOuts.forEach((batchOut, i) => outsBatches[i].push(batchOut));\n      }\n      return singletonOrArray(outsBatches.map((batches2) => concat(batches2, 0)));\n    });\n  }\n  predict(x, args = {}) {\n    const xsRank2OrHigher = ensureTensorsRank2OrHigher(x);\n    checkInputData(xsRank2OrHigher, this.inputNames, this.feedInputShapes, false);\n    try {\n      const batchSize = args.batchSize == null ? 32 : args.batchSize;\n      checkBatchSize(batchSize);\n      return this.predictLoop(xsRank2OrHigher, batchSize);\n    } finally {\n      disposeNewTensors(xsRank2OrHigher, x);\n    }\n  }\n  predictOnBatch(x) {\n    checkInputData(x, this.inputNames, this.feedInputShapes, true);\n    const batchSize = (Array.isArray(x) ? x[0] : x).shape[0];\n    return this.predictLoop(x, batchSize);\n  }\n  standardizeUserDataXY(x, y, checkBatchAxis = true, batchSize) {\n    if (this.optimizer_ == null) {\n      throw new RuntimeError(\"You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).\");\n    }\n    const outputShapes = [];\n    for (let i = 0; i < this.feedOutputShapes.length; ++i) {\n      const outputShape = this.feedOutputShapes[i];\n      const lossFn = this.feedLossFns[i];\n      if (lossFn === sparseCategoricalCrossentropy) {\n        outputShapes.push(outputShape.slice(0, outputShape.length - 1).concat([1]));\n      } else {\n        outputShapes.push(outputShape);\n      }\n    }\n    x = standardizeInputData(x, this.feedInputNames, this.feedInputShapes, false, \"input\");\n    y = standardizeInputData(y, this.feedOutputNames, outputShapes, false, \"target\");\n    checkArrayLengths(x, y, null);\n    checkLossAndTargetCompatibility(y, this.feedLossFns, this.feedOutputShapes);\n    if (this.stateful && batchSize != null && batchSize > 0) {\n      if (x[0].shape[0] % batchSize !== 0) {\n        throw new ValueError(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${batchSize}. Found: ${x[0].shape[0]} sample(s).`);\n      }\n    }\n    return [x, y];\n  }\n  async standardizeUserData(x, y, sampleWeight, classWeight, checkBatchAxis = true, batchSize) {\n    const [standardXs, standardYs] = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);\n    if (sampleWeight != null) {\n      throw new Error(\"sample weight is not supported yet.\");\n    }\n    let standardSampleWeights = null;\n    if (classWeight != null) {\n      const classWeights = standardizeClassWeights(classWeight, this.outputNames);\n      standardSampleWeights = [];\n      for (let i = 0; i < classWeights.length; ++i) {\n        standardSampleWeights.push(await standardizeWeights(standardYs[i], null, classWeights[i]));\n      }\n    }\n    return [standardXs, standardYs, standardSampleWeights];\n  }\n  testLoop(f, ins, batchSize, verbose = 0, steps) {\n    return tidy(() => {\n      const numSamples = this.checkNumSamples(ins, batchSize, steps, \"steps\");\n      const outs = [];\n      if (verbose > 0) {\n        throw new NotImplementedError(\"Verbose mode is not implemented yet.\");\n      }\n      if (steps != null) {\n        throw new NotImplementedError(\"steps mode in testLoop() is not implemented yet\");\n      } else {\n        const batches = makeBatches(numSamples, batchSize);\n        const indexArray = tensor1d(range2(0, numSamples));\n        for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {\n          const batchStart = batches[batchIndex][0];\n          const batchEnd = batches[batchIndex][1];\n          const batchIds = sliceAlongFirstAxis(indexArray, batchStart, batchEnd - batchStart);\n          const insBatch = sliceArraysByIndices(ins, batchIds);\n          const batchOuts = f(insBatch);\n          if (batchIndex === 0) {\n            for (let i = 0; i < batchOuts.length; ++i) {\n              outs.push(scalar(0));\n            }\n          }\n          for (let i = 0; i < batchOuts.length; ++i) {\n            const batchOut = batchOuts[i];\n            outs[i] = add2(outs[i], mul(batchEnd - batchStart, batchOut));\n          }\n        }\n        for (let i = 0; i < outs.length; ++i) {\n          outs[i] = div(outs[i], numSamples);\n        }\n      }\n      return outs;\n    });\n  }\n  getDedupedMetricsNames() {\n    const outLabels = this.metricsNames;\n    const dedupedOutLabels = [];\n    for (let i = 0; i < outLabels.length; ++i) {\n      const label = outLabels[i];\n      let newLabel = label;\n      if (count(outLabels, label) > 1) {\n        const dupIndex = count(outLabels.slice(0, i), label);\n        newLabel += `_${dupIndex}`;\n      }\n      dedupedOutLabels.push(newLabel);\n    }\n    return dedupedOutLabels;\n  }\n  makeTrainFunction() {\n    return (data) => {\n      const lossValues = [];\n      const inputs = data.slice(0, this.inputs.length);\n      const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);\n      const sampleWeights = data.slice(this.inputs.length + this.outputs.length, this.inputs.length + this.outputs.length * 2);\n      const metricsValues = [];\n      const totalLossFunction = () => {\n        const feeds = [];\n        for (let i = 0; i < this.inputs.length; ++i) {\n          feeds.push({ key: this.inputs[i], value: inputs[i] });\n        }\n        const feedDict = new FeedDict(feeds);\n        const outputs = execute(this.outputs, feedDict, { \"training\": true });\n        let totalLoss;\n        for (let i = 0; i < this.lossFunctions.length; ++i) {\n          const lossFunction = this.lossFunctions[i];\n          let loss = lossFunction(targets[i], outputs[i]);\n          if (sampleWeights[i] != null) {\n            loss = computeWeightedLoss2(loss, sampleWeights[i]);\n          }\n          const meanLoss = mean(loss);\n          lossValues.push(meanLoss);\n          if (i === 0) {\n            totalLoss = loss;\n          } else {\n            totalLoss = add2(totalLoss, loss);\n          }\n        }\n        for (let i = 0; i < this.metricsTensors.length; ++i) {\n          let weightedMetric;\n          if (this.outputs.length > 1 && i < this.outputs.length) {\n            weightedMetric = lossValues[i];\n          } else {\n            const metric = this.metricsTensors[i][0];\n            const outputIndex = this.metricsTensors[i][1];\n            weightedMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));\n          }\n          keep(weightedMetric);\n          metricsValues.push(weightedMetric);\n        }\n        totalLoss = mean(totalLoss);\n        this.calculateLosses().forEach((regularizerLoss) => {\n          totalLoss = add2(totalLoss, regularizerLoss);\n        });\n        return totalLoss;\n      };\n      const variables = this.collectedTrainableWeights.map((param) => param.read());\n      const returnCost = true;\n      const totalLossValue = this.optimizer_.minimize(totalLossFunction, returnCost, variables);\n      return [totalLossValue].concat(metricsValues);\n    };\n  }\n  makeTestFunction() {\n    this.testFunction = (data) => {\n      return tidy(() => {\n        const valOutputs = [];\n        let totalLoss;\n        const inputs = data.slice(0, this.inputs.length);\n        const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);\n        const feeds = [];\n        for (let i = 0; i < this.inputs.length; ++i) {\n          feeds.push({ key: this.inputs[i], value: inputs[i] });\n        }\n        const feedDict = new FeedDict(feeds);\n        const outputs = execute(this.outputs, feedDict);\n        for (let i = 0; i < this.lossFunctions.length; ++i) {\n          const lossFunction = this.lossFunctions[i];\n          const loss = mean(lossFunction(targets[i], outputs[i]));\n          if (i === 0) {\n            totalLoss = loss;\n          } else {\n            totalLoss = add2(totalLoss, loss);\n          }\n          valOutputs.push(totalLoss);\n        }\n        for (let i = 0; i < this.metricsTensors.length; ++i) {\n          const metric = this.metricsTensors[i][0];\n          const outputIndex = this.metricsTensors[i][1];\n          const meanMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));\n          valOutputs.push(meanMetric);\n        }\n        return valOutputs;\n      });\n    };\n  }\n  async fit(x, y, args = {}) {\n    return fitTensors(this, x, y, args);\n  }\n  async fitDataset(dataset, args) {\n    return fitDataset(this, dataset, args);\n  }\n  async trainOnBatch(x, y) {\n    const standardizeOut = await this.standardizeUserData(x, y);\n    const inputs = standardizeOut[0];\n    const targets = standardizeOut[1];\n    const trainFunction = this.makeTrainFunction();\n    const losses2 = trainFunction(inputs.concat(targets));\n    const lossValues = [];\n    for (const loss of losses2) {\n      const v = await loss.data();\n      lossValues.push(v[0]);\n    }\n    dispose(losses2);\n    disposeNewTensors(standardizeOut[0], x);\n    disposeNewTensors(standardizeOut[1], y);\n    return singletonOrArray(lossValues);\n  }\n  getNamedWeights(config) {\n    const namedWeights = [];\n    const trainableOnly = config != null && config.trainableOnly;\n    const weights = trainableOnly ? this.trainableWeights : this.weights;\n    const weightValues = this.getWeights(trainableOnly);\n    for (let i = 0; i < weights.length; ++i) {\n      if (trainableOnly && !weights[i].trainable) {\n        continue;\n      }\n      namedWeights.push({ name: weights[i].originalName, tensor: weightValues[i] });\n    }\n    return namedWeights;\n  }\n  set stopTraining(stop) {\n    this.stopTraining_ = stop;\n  }\n  get stopTraining() {\n    return this.stopTraining_;\n  }\n  get optimizer() {\n    return this.optimizer_;\n  }\n  set optimizer(optimizer) {\n    if (this.optimizer_ !== optimizer) {\n      this.optimizer_ = optimizer;\n      this.isOptimizerOwned = false;\n    }\n  }\n  dispose() {\n    const result = super.dispose();\n    if (result.refCountAfterDispose === 0 && this.optimizer != null && this.isOptimizerOwned) {\n      const numTensorsBeforeOptmizerDisposal = memory().numTensors;\n      this.optimizer_.dispose();\n      result.numDisposedVariables += numTensorsBeforeOptmizerDisposal - memory().numTensors;\n    }\n    return result;\n  }\n  getLossIdentifiers() {\n    let lossNames;\n    if (typeof this.loss === \"string\") {\n      lossNames = toSnakeCase(this.loss);\n    } else if (Array.isArray(this.loss)) {\n      for (const loss of this.loss) {\n        if (typeof loss !== \"string\") {\n          throw new Error(\"Serialization of non-string loss is not supported.\");\n        }\n      }\n      lossNames = this.loss.map((name) => toSnakeCase(name));\n    } else {\n      const outputNames = Object.keys(this.loss);\n      lossNames = {};\n      const losses2 = this.loss;\n      for (const outputName of outputNames) {\n        if (typeof losses2[outputName] === \"string\") {\n          lossNames[outputName] = toSnakeCase(losses2[outputName]);\n        } else {\n          throw new Error(\"Serialization of non-string loss is not supported.\");\n        }\n      }\n    }\n    return lossNames;\n  }\n  getMetricIdentifiers() {\n    if (typeof this.metrics === \"string\" || typeof this.metrics === \"function\") {\n      return [toSnakeCase(getLossOrMetricName(this.metrics))];\n    } else if (Array.isArray(this.metrics)) {\n      return this.metrics.map((metric) => toSnakeCase(getLossOrMetricName(metric)));\n    } else {\n      const metricsIdentifiers = {};\n      for (const key in this.metrics) {\n        metricsIdentifiers[key] = toSnakeCase(getLossOrMetricName(this.metrics[key]));\n      }\n      return metricsIdentifiers;\n    }\n  }\n  getTrainingConfig() {\n    return {\n      loss: this.getLossIdentifiers(),\n      metrics: this.getMetricIdentifiers(),\n      optimizer_config: {\n        class_name: this.optimizer.getClassName(),\n        config: this.optimizer.getConfig()\n      }\n    };\n  }\n  loadTrainingConfig(trainingConfig) {\n    if (trainingConfig.weighted_metrics != null) {\n      throw new Error(\"Loading weight_metrics is not supported yet.\");\n    }\n    if (trainingConfig.loss_weights != null) {\n      throw new Error(\"Loading loss_weights is not supported yet.\");\n    }\n    if (trainingConfig.sample_weight_mode != null) {\n      throw new Error(\"Loading sample_weight_mode is not supported yet.\");\n    }\n    const tsConfig = convertPythonicToTs(trainingConfig.optimizer_config);\n    const optimizer = deserialize(tsConfig);\n    let loss;\n    if (typeof trainingConfig.loss === \"string\") {\n      loss = toCamelCase(trainingConfig.loss);\n    } else if (Array.isArray(trainingConfig.loss)) {\n      loss = trainingConfig.loss.map((lossEntry) => toCamelCase(lossEntry));\n    } else if (trainingConfig.loss != null) {\n      loss = {};\n      for (const key in trainingConfig.loss) {\n        loss[key] = toCamelCase(trainingConfig.loss[key]);\n      }\n    }\n    let metrics;\n    if (Array.isArray(trainingConfig.metrics)) {\n      metrics = trainingConfig.metrics.map((metric) => toCamelCase(metric));\n    } else if (trainingConfig.metrics != null) {\n      metrics = {};\n      for (const key in trainingConfig.metrics) {\n        metrics[key] = toCamelCase(trainingConfig.metrics[key]);\n      }\n    }\n    this.compile({ loss, metrics, optimizer });\n  }\n  async save(handlerOrURL, config) {\n    if (typeof handlerOrURL === \"string\") {\n      const handlers = io_exports.getSaveHandlers(handlerOrURL);\n      if (handlers.length === 0) {\n        throw new ValueError(`Cannot find any save handlers for URL '${handlerOrURL}'`);\n      } else if (handlers.length > 1) {\n        throw new ValueError(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);\n      }\n      handlerOrURL = handlers[0];\n    }\n    if (handlerOrURL.save == null) {\n      throw new ValueError(\"LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.\");\n    }\n    const weightDataAndSpecs = await io_exports.encodeWeights(this.getNamedWeights(config));\n    const returnString = false;\n    const unusedArg = null;\n    const modelConfig = this.toJSON(unusedArg, returnString);\n    const modelArtifacts = {\n      modelTopology: modelConfig,\n      format: LAYERS_MODEL_FORMAT_NAME,\n      generatedBy: `TensorFlow.js tfjs-layers v${version2}`,\n      convertedBy: null\n    };\n    const includeOptimizer = config == null ? false : config.includeOptimizer;\n    if (includeOptimizer && this.optimizer != null) {\n      modelArtifacts.trainingConfig = this.getTrainingConfig();\n      const weightType = \"optimizer\";\n      const { data: optimizerWeightData, specs: optimizerWeightSpecs } = await io_exports.encodeWeights(await this.optimizer.getWeights(), weightType);\n      weightDataAndSpecs.specs.push(...optimizerWeightSpecs);\n      weightDataAndSpecs.data = io_exports.concatenateArrayBuffers([weightDataAndSpecs.data, optimizerWeightData]);\n    }\n    if (this.userDefinedMetadata != null) {\n      const checkSize = true;\n      checkUserDefinedMetadata(this.userDefinedMetadata, this.name, checkSize);\n      modelArtifacts.userDefinedMetadata = this.userDefinedMetadata;\n    }\n    modelArtifacts.weightData = weightDataAndSpecs.data;\n    modelArtifacts.weightSpecs = weightDataAndSpecs.specs;\n    return handlerOrURL.save(modelArtifacts);\n  }\n  setUserDefinedMetadata(userDefinedMetadata) {\n    checkUserDefinedMetadata(userDefinedMetadata, this.name);\n    this.userDefinedMetadata = userDefinedMetadata;\n  }\n  getUserDefinedMetadata() {\n    return this.userDefinedMetadata;\n  }\n};\nLayersModel.className = \"Model\";\nserialization_exports.registerClass(LayersModel);\nvar Functional = class extends LayersModel {\n};\nFunctional.className = \"Functional\";\nserialization_exports.registerClass(Functional);\n\n// node_modules/@tensorflow/tfjs-layers/dist/models.js\nasync function modelFromJSON(modelAndWeightsConfig, customObjects) {\n  if (!(\"modelTopology\" in modelAndWeightsConfig)) {\n    modelAndWeightsConfig = { modelTopology: modelAndWeightsConfig };\n  }\n  modelAndWeightsConfig = modelAndWeightsConfig;\n  let modelTopology = modelAndWeightsConfig.modelTopology;\n  if (modelTopology[\"model_config\"] != null) {\n    modelTopology = modelTopology[\"model_config\"];\n  }\n  const tsConfig = convertPythonicToTs(modelTopology);\n  const model2 = deserialize(tsConfig, customObjects);\n  if (modelAndWeightsConfig.weightsManifest != null) {\n    const weightValues = await io_exports.loadWeights(modelAndWeightsConfig.weightsManifest, modelAndWeightsConfig.pathPrefix, model2.weights.map((weight) => weight.originalName));\n    const uniqueWeightValues = {};\n    for (const weight of model2.weights) {\n      uniqueWeightValues[weight.originalName] = weightValues[weight.originalName];\n    }\n    model2.loadWeights(uniqueWeightValues);\n    dispose(weightValues);\n  }\n  return model2;\n}\nasync function loadLayersModel(pathOrIOHandler, options) {\n  if (options == null) {\n    options = {};\n  }\n  if (typeof pathOrIOHandler === \"string\") {\n    const handlers = io_exports.getLoadHandlers(pathOrIOHandler, options);\n    if (handlers.length === 0) {\n      handlers.push(io_exports.browserHTTPRequest(pathOrIOHandler, options));\n    } else if (handlers.length > 1) {\n      throw new ValueError(`Found more than one (${handlers.length}) load handlers for URL '${pathOrIOHandler}'`);\n    }\n    pathOrIOHandler = handlers[0];\n  }\n  return loadLayersModelFromIOHandler(pathOrIOHandler, void 0, options);\n}\nasync function loadLayersModelFromIOHandler(handler, customObjects, options) {\n  if (options == null) {\n    options = {};\n  }\n  if (handler.load == null) {\n    throw new ValueError(\"Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.\");\n  }\n  const artifacts = await handler.load();\n  let modelTopology = artifacts.modelTopology;\n  if (modelTopology[\"model_config\"] != null) {\n    modelTopology = modelTopology[\"model_config\"];\n  }\n  const strict = options.strict == null ? true : options.strict;\n  const fastWeightInit = artifacts.weightData != null && artifacts.weightSpecs != null && strict;\n  const model2 = deserialize(convertPythonicToTs(modelTopology), customObjects, fastWeightInit);\n  const trainingConfig = artifacts.trainingConfig;\n  if (trainingConfig != null) {\n    model2.loadTrainingConfig(trainingConfig);\n  }\n  if (artifacts.userDefinedMetadata != null) {\n    model2.setUserDefinedMetadata(artifacts.userDefinedMetadata);\n  }\n  if (artifacts.weightData != null) {\n    if (artifacts.weightSpecs == null) {\n      throw new ValueError(\"LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.\");\n    }\n    const { modelWeights, optimizerWeights } = decodeModelAndOptimizerWeights(artifacts.weightData, artifacts.weightSpecs);\n    model2.loadWeights(modelWeights, strict);\n    if (model2.optimizer != null && optimizerWeights.length > 0) {\n      await model2.optimizer.setWeights(optimizerWeights);\n    }\n    dispose(modelWeights);\n    dispose(optimizerWeights.map((w) => w.tensor));\n  }\n  return model2;\n}\nfunction decodeModelAndOptimizerWeights(buffer2, specs) {\n  const name2Tensor = io_exports.decodeWeights(buffer2, specs);\n  const modelWeights = {};\n  const optimizerWeights = [];\n  specs.forEach((spec) => {\n    if (spec.group === \"optimizer\") {\n      optimizerWeights.push({ name: spec.name, tensor: name2Tensor[spec.name] });\n    } else {\n      modelWeights[spec.name] = name2Tensor[spec.name];\n    }\n  });\n  return { modelWeights, optimizerWeights };\n}\nvar Sequential = class extends LayersModel {\n  constructor(args) {\n    super({ inputs: [], outputs: [] });\n    args = args || {};\n    this.trainable = true;\n    this.built = false;\n    this.name = args.name != null ? args.name : getUid(\"sequential_\");\n    if (args.layers != null) {\n      for (const layer of args.layers) {\n        this.add(layer);\n      }\n    }\n  }\n  checkShape(layer) {\n    const shape = layer.inboundNodes[0].outputTensors[0].shape;\n    if (shape.some((x) => x < 0)) {\n      throw new ValueError(`Negative dimension size caused by adding layer ${layer.name} with input shape [${layer.inboundNodes[0].inputTensors[0].shape}]`);\n    }\n  }\n  add(layer) {\n    const isLayerModelInstance = layer instanceof Sequential || layer instanceof LayersModel;\n    let modelLayer;\n    if (isLayerModelInstance) {\n      modelLayer = layer;\n      if (modelLayer.outputs.length !== 1) {\n        throw new ValueError(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");\n      }\n      if (modelLayer.inputs.length !== 1) {\n        throw new ValueError(\"All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.\");\n      }\n    }\n    if (this.outputs.length === 0) {\n      if (layer.inboundNodes.length === 0) {\n        if (layer.batchInputShape == null) {\n          throw new ValueError(\"The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.\");\n        }\n        const x = Input({\n          batchShape: layer.batchInputShape,\n          dtype: layer.dtype,\n          name: layer.name + \"_input\"\n        });\n        layer.apply(x);\n      }\n      if (isLayerModelInstance) {\n        this.outputs = modelLayer.outputs;\n        this.inputs = modelLayer.inputs;\n      } else {\n        if (layer.inboundNodes.length !== 1) {\n          throw new ValueError(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${layer.name} which has ${layer.inboundNodes.length} pre-existing inbound connections.`);\n        }\n        if (layer.inboundNodes[0].outputTensors.length !== 1) {\n          throw new ValueError(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");\n        }\n        this.checkShape(layer);\n        this.outputs = [layer.inboundNodes[0].outputTensors[0]];\n        this.inputs = getSourceInputs(this.outputs[0]);\n      }\n      this.inboundNodes = [];\n      new Node({\n        outboundLayer: this,\n        inboundLayers: [],\n        nodeIndices: [],\n        tensorIndices: [],\n        inputTensors: this.inputs,\n        outputTensors: this.outputs,\n        inputMasks: pyListRepeat(null, this.inputs.length),\n        outputMasks: [null],\n        inputShapes: this.inputs.map((x) => x.shape),\n        outputShapes: this.outputs[0].shape\n      });\n    } else {\n      const outputTensor = layer.apply(this.outputs[0]);\n      if (Array.isArray(outputTensor)) {\n        throw new TypeError(\"All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.\");\n      }\n      this.checkShape(layer);\n      this.outputs = [outputTensor];\n      this.inboundNodes[0].outputTensors = this.outputs;\n      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];\n    }\n    this.layers.push(layer);\n    this.built = false;\n  }\n  pop() {\n    if (this.layers.length === 0) {\n      throw new TypeError(\"There are no layers in the model.\");\n    }\n    this.layers.pop();\n    if (this.layers.length === 0) {\n      this.outputs = [];\n      this.inboundNodes = [];\n      this.outboundNodes = [];\n    } else {\n      const lastLayerIndex = this.layers.length - 1;\n      this.layers[lastLayerIndex].outboundNodes = [];\n      this.outputs = [this.layers[lastLayerIndex].output];\n      this.inboundNodes[0].outputTensors = this.outputs;\n      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];\n    }\n  }\n  call(inputs, kwargs) {\n    if (this.model == null) {\n      this.build();\n    }\n    return this.model.call(inputs, kwargs);\n  }\n  build(inputShape) {\n    getExactlyOneShape(inputShape);\n    if (this.inputs.length === 0 || this.outputs.length === 0) {\n      throw new TypeError(\"Sequential model cannot be built: model is empty. Add some layers first.\");\n    }\n    this.model = new LayersModel({\n      inputs: this.inputs,\n      outputs: this.outputs[0],\n      name: this.name + \"_model\"\n    });\n    this.model.trainable = this.trainable;\n    this.supportsMasking = this.model.supportsMasking;\n    this.inputLayers = this.model.inputLayers;\n    this.inputLayersNodeIndices = this.model.inputLayersNodeIndices;\n    this.inputLayersTensorIndices = this.model.inputLayersTensorIndices;\n    this.outputLayers = this.model.outputLayers;\n    this.outputLayersNodeIndices = this.model.outputLayersNodeIndices;\n    this.outputLayersTensorIndices = this.model.outputLayersTensorIndices;\n    this.nodesByDepth = this.model.nodesByDepth;\n    this.containerNodes = this.model.containerNodes;\n    this.outputNames = this.model.outputNames;\n    this.inputNames = this.model.inputNames;\n    this.built = true;\n  }\n  countParams() {\n    if (!this.built) {\n      this.build();\n    }\n    return super.countParams();\n  }\n  summary(lineLength, positions, printFn = console.log) {\n    if (!this.built) {\n      this.build();\n    }\n    super.summary(lineLength, positions, printFn);\n  }\n  setWeights(weights) {\n    if (this.model == null) {\n      this.build();\n    }\n    this.model.setWeights(weights);\n  }\n  evaluate(x, y, args = {}) {\n    if (!this.built) {\n      throw new RuntimeError(\"The model needs to be compiled before being used.\");\n    }\n    return this.model.evaluate(x, y, args);\n  }\n  async evaluateDataset(dataset, args) {\n    if (!this.built) {\n      throw new RuntimeError(\"The model needs to be compiled before being used.\");\n    }\n    return this.model.evaluateDataset(dataset, args);\n  }\n  predict(x, args = {}) {\n    if (this.model == null) {\n      this.build();\n    }\n    return this.model.predict(x, args);\n  }\n  predictOnBatch(x) {\n    if (this.model == null) {\n      this.build();\n    }\n    return this.model.predictOnBatch(x);\n  }\n  compile(args) {\n    this.build();\n    this.model.compile(args);\n    this.optimizer_ = this.model.optimizer;\n    this.isOptimizerOwned = this.model.isOptimizerOwned;\n    this.loss = this.model.loss;\n    this.metrics = this.model.metrics;\n    this.metricsTensors = this.model.metricsTensors;\n    this.metricsNames = this.model.metricsNames;\n  }\n  get optimizer() {\n    return this.model == null ? void 0 : this.model.optimizer;\n  }\n  set optimizer(optimizer) {\n    this.model.optimizer = optimizer;\n  }\n  async fit(x, y, args = {}) {\n    if (!this.built) {\n      throw new RuntimeError(\"The model needs to be compiled before being used.\");\n    }\n    return this.model.fit(x, y, args);\n  }\n  async fitDataset(dataset, args) {\n    if (!this.built) {\n      throw new RuntimeError(\"The model needs to be compiled before being used.\");\n    }\n    return this.model.fitDataset(dataset, args);\n  }\n  async trainOnBatch(x, y) {\n    return this.model.trainOnBatch(x, y);\n  }\n  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {\n    let configArray;\n    let extraModelConfig = {};\n    if (config instanceof Array) {\n      if (!(config[0].className != null) || config[0][\"className\"] === \"Merge\") {\n        throw new ValueError(\"Legacy serialization format not supported yet.\");\n      }\n      configArray = config;\n    } else {\n      util_exports.assert(config[\"layers\"] != null, () => `When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field.`);\n      configArray = config[\"layers\"];\n      delete config[\"layers\"];\n      extraModelConfig = config;\n    }\n    const model2 = new cls(extraModelConfig);\n    if (!(model2 instanceof Sequential)) {\n      throw new NotImplementedError(`Sequential.fromConfig called on non-Sequential input: ${model2}`);\n    }\n    for (const conf of configArray) {\n      const customObjects2 = void 0;\n      const layer = deserialize(conf, customObjects2, fastWeightInit);\n      if (fastWeightInit) {\n        layer.setFastWeightInitDuringBuild(true);\n      }\n      model2.add(layer);\n    }\n    return model2;\n  }\n  set stopTraining(stop) {\n    if (this.model == null) {\n      throw new ValueError(\"Cannot set the stopTraining property of a sequential model before it is compiled.\");\n    }\n    this.model.stopTraining = stop;\n  }\n  get stopTraining() {\n    if (this.model == null) {\n      throw new ValueError(\"Cannot get the stopTraining property of a sequential model before it is compiled.\");\n    }\n    return this.model.stopTraining;\n  }\n  getConfig() {\n    const layers = [];\n    for (const layer of this.layers) {\n      const dict = {};\n      dict[\"className\"] = layer.getClassName();\n      dict[\"config\"] = layer.getConfig();\n      layers.push(dict);\n    }\n    return { name: this.name, layers };\n  }\n};\nSequential.className = \"Sequential\";\nserialization_exports.registerClass(Sequential);\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports.js\nfunction model(args) {\n  return new LayersModel(args);\n}\nfunction sequential(config) {\n  return new Sequential(config);\n}\nfunction input(config) {\n  return Input(config);\n}\nfunction registerCallbackConstructor(verbosityLevel, callbackConstructor) {\n  CallbackConstructorRegistry.registerCallbackConstructor(verbosityLevel, callbackConstructor);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/activations.js\nvar Activation = class extends serialization_exports.Serializable {\n  getConfig() {\n    return {};\n  }\n};\nvar Elu2 = class extends Activation {\n  apply(x, alpha = 1) {\n    return elu2(x, alpha);\n  }\n};\nElu2.className = \"elu\";\nserialization_exports.registerClass(Elu2);\nvar Selu2 = class extends Activation {\n  apply(x) {\n    return selu(x);\n  }\n};\nSelu2.className = \"selu\";\nserialization_exports.registerClass(Selu2);\nvar Relu2 = class extends Activation {\n  apply(x) {\n    return relu(x);\n  }\n};\nRelu2.className = \"relu\";\nserialization_exports.registerClass(Relu2);\nvar Relu62 = class extends Activation {\n  apply(x) {\n    return tidy(() => minimum(6, relu(x)));\n  }\n};\nRelu62.className = \"relu6\";\nserialization_exports.registerClass(Relu62);\nvar Linear = class extends Activation {\n  apply(x) {\n    return x;\n  }\n};\nLinear.className = \"linear\";\nserialization_exports.registerClass(Linear);\nvar Sigmoid2 = class extends Activation {\n  apply(x) {\n    return sigmoid(x);\n  }\n};\nSigmoid2.className = \"sigmoid\";\nserialization_exports.registerClass(Sigmoid2);\nvar HardSigmoid = class extends Activation {\n  apply(x) {\n    return hardSigmoid(x);\n  }\n};\nHardSigmoid.className = \"hardSigmoid\";\nserialization_exports.registerClass(HardSigmoid);\nvar Softplus2 = class extends Activation {\n  apply(x) {\n    return softplus(x);\n  }\n};\nSoftplus2.className = \"softplus\";\nserialization_exports.registerClass(Softplus2);\nvar Softsign = class extends Activation {\n  apply(x) {\n    return softsign(x);\n  }\n};\nSoftsign.className = \"softsign\";\nserialization_exports.registerClass(Softsign);\nvar Tanh2 = class extends Activation {\n  apply(x) {\n    return tanh2(x);\n  }\n};\nTanh2.className = \"tanh\";\nserialization_exports.registerClass(Tanh2);\nvar Softmax2 = class extends Activation {\n  apply(x, axis = -1) {\n    return softmax(x, axis);\n  }\n};\nSoftmax2.className = \"softmax\";\nserialization_exports.registerClass(Softmax2);\nvar LogSoftmax2 = class extends Activation {\n  apply(x, axis = -1) {\n    return logSoftmax(x, axis);\n  }\n};\nLogSoftmax2.className = \"logSoftmax\";\nserialization_exports.registerClass(LogSoftmax2);\nvar Swish = class extends Activation {\n  apply(x, alpha = 1) {\n    return tidy(() => mul(sigmoid(mul(x, alpha)), x));\n  }\n};\nSwish.className = \"swish\";\nserialization_exports.registerClass(Swish);\nvar Mish = class extends Activation {\n  apply(x) {\n    return tidy(() => mul(x, tanh2(softplus(x))));\n  }\n};\nMish.className = \"mish\";\nserialization_exports.registerClass(Mish);\nfunction serializeActivation(activation2) {\n  return activation2.getClassName();\n}\nfunction deserializeActivation(config, customObjects = {}) {\n  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, \"activation\");\n}\nfunction getActivation(identifier) {\n  if (identifier == null) {\n    const config = {};\n    config[\"className\"] = \"linear\";\n    config[\"config\"] = {};\n    return deserializeActivation(config);\n  }\n  if (typeof identifier === \"string\") {\n    const config = {};\n    config[\"className\"] = identifier;\n    config[\"config\"] = {};\n    return deserializeActivation(config);\n  } else if (identifier instanceof Activation) {\n    return identifier;\n  } else {\n    return deserializeActivation(identifier);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/regularizers.js\nfunction assertObjectArgs(args) {\n  if (args != null && typeof args !== \"object\") {\n    throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${args}`);\n  }\n}\nvar Regularizer = class extends serialization_exports.Serializable {\n};\nvar L1L2 = class extends Regularizer {\n  constructor(args) {\n    super();\n    assertObjectArgs(args);\n    this.l1 = args == null || args.l1 == null ? 0.01 : args.l1;\n    this.l2 = args == null || args.l2 == null ? 0.01 : args.l2;\n    this.hasL1 = this.l1 !== 0;\n    this.hasL2 = this.l2 !== 0;\n  }\n  apply(x) {\n    return tidy(() => {\n      let regularization = zeros([1]);\n      if (this.hasL1) {\n        regularization = add2(regularization, sum2(mul(this.l1, abs(x))));\n      }\n      if (this.hasL2) {\n        regularization = add2(regularization, sum2(mul(this.l2, square2(x))));\n      }\n      return reshape(regularization, []);\n    });\n  }\n  getConfig() {\n    return { \"l1\": this.l1, \"l2\": this.l2 };\n  }\n  static fromConfig(cls, config) {\n    return new cls({ l1: config[\"l1\"], l2: config[\"l2\"] });\n  }\n};\nL1L2.className = \"L1L2\";\nserialization_exports.registerClass(L1L2);\nfunction l1(args) {\n  assertObjectArgs(args);\n  return new L1L2({ l1: args != null ? args.l1 : null, l2: 0 });\n}\nfunction l2(args) {\n  assertObjectArgs(args);\n  return new L1L2({ l2: args != null ? args.l2 : null, l1: 0 });\n}\nvar REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {\n  \"l1l2\": \"L1L2\"\n};\nfunction serializeRegularizer(constraint) {\n  return serializeKerasObject(constraint);\n}\nfunction deserializeRegularizer(config, customObjects = {}) {\n  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, \"regularizer\");\n}\nfunction getRegularizer(identifier) {\n  if (identifier == null) {\n    return null;\n  }\n  if (typeof identifier === \"string\") {\n    const className = identifier in REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;\n    const config = { className, config: {} };\n    return deserializeRegularizer(config);\n  } else if (identifier instanceof Regularizer) {\n    return identifier;\n  } else {\n    return deserializeRegularizer(identifier);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/advanced_activations.js\nvar ReLU = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.supportsMasking = true;\n    if (args != null) {\n      this.maxValue = args.maxValue;\n    }\n  }\n  call(inputs, kwargs) {\n    inputs = getExactlyOneTensor(inputs);\n    let output = relu(inputs);\n    if (this.maxValue != null) {\n      output = clipByValue(output, 0, this.maxValue);\n    }\n    return output;\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const config = { maxValue: this.maxValue };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nReLU.className = \"ReLU\";\nserialization_exports.registerClass(ReLU);\nvar LeakyReLU = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.DEFAULT_ALPHA = 0.3;\n    if (args == null) {\n      args = {};\n    }\n    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;\n  }\n  call(inputs, kwargs) {\n    const x = getExactlyOneTensor(inputs);\n    return leakyRelu(x, this.alpha);\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const config = { alpha: this.alpha };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nLeakyReLU.className = \"LeakyReLU\";\nserialization_exports.registerClass(LeakyReLU);\nvar PReLU = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.DEFAULT_ALPHA_INITIALIZER = \"zeros\";\n    if (args == null) {\n      args = {};\n    }\n    this.supportsMasking = true;\n    this.alphaInitializer = getInitializer(args.alphaInitializer || this.DEFAULT_ALPHA_INITIALIZER);\n    this.alphaRegularizer = getRegularizer(args.alphaRegularizer);\n    this.alphaConstraint = getConstraint(args.alphaConstraint);\n    if (args.sharedAxes == null) {\n      this.sharedAxes = null;\n    } else if (Array.isArray(args.sharedAxes)) {\n      this.sharedAxes = args.sharedAxes;\n    } else if (typeof args.sharedAxes === \"number\") {\n      this.sharedAxes = [args.sharedAxes];\n    } else {\n      throw new ValueError(`Expected sharedAxes to be a number or an array of numbers, but got ${args.sharedAxes}`);\n    }\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const paramShape = inputShape.slice(1);\n    if (this.sharedAxes != null) {\n      for (const i of this.sharedAxes) {\n        paramShape[i - 1] = 1;\n      }\n    }\n    this.alpha = this.addWeight(\"alpha\", paramShape, \"float32\", this.alphaInitializer, this.alphaRegularizer, true, this.alphaConstraint);\n    const axes = {};\n    if (this.sharedAxes != null) {\n      for (let i = 1; i < inputShape.length; ++i) {\n        axes[i] = inputShape[i];\n      }\n    }\n    this.inputSpec = [new InputSpec({\n      ndim: inputShape.length,\n      axes\n    })];\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    inputs = getExactlyOneTensor(inputs);\n    return prelu(inputs, this.alpha.read());\n  }\n  getConfig() {\n    const config = {\n      alphaInitializer: serializeInitializer(this.alphaInitializer),\n      alphaRegularizer: serializeRegularizer(this.alphaRegularizer),\n      alphaConstraint: serializeConstraint(this.alphaConstraint),\n      sharedAxes: this.sharedAxes\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nPReLU.className = \"PReLU\";\nserialization_exports.registerClass(PReLU);\nvar ELU = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.DEFAULT_ALPHA = 1;\n    if (args == null) {\n      args = {};\n    }\n    if (args.alpha != null && args.alpha !== this.DEFAULT_ALPHA) {\n      throw new NotImplementedError(`Non-default alpha value (${args.alpha}) is not supported by the ELU layer yet.`);\n    }\n    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;\n  }\n  call(inputs, kwargs) {\n    const x = getExactlyOneTensor(inputs);\n    return elu(x);\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const config = { alpha: this.alpha };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nELU.className = \"ELU\";\nserialization_exports.registerClass(ELU);\nvar ThresholdedReLU = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.DEFAULT_THETA = 1;\n    if (args == null) {\n      args = {};\n    }\n    this.theta = args.theta == null ? this.DEFAULT_THETA : args.theta;\n  }\n  call(inputs, kwargs) {\n    const x = getExactlyOneTensor(inputs);\n    return mul(x, cast(greater(x, this.theta), \"float32\"));\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const config = { theta: this.theta };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nThresholdedReLU.className = \"ThresholdedReLU\";\nserialization_exports.registerClass(ThresholdedReLU);\nvar Softmax3 = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.DEFAULT_AXIS = 1;\n    if (args == null) {\n      args = {};\n    }\n    this.softmax = new Softmax2().apply;\n    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;\n  }\n  call(inputs, kwargs) {\n    const x = getExactlyOneTensor(inputs);\n    return this.softmax(x, this.axis);\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const config = { axis: this.axis };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nSoftmax3.className = \"Softmax\";\nserialization_exports.registerClass(Softmax3);\n\n// node_modules/@tensorflow/tfjs-layers/dist/utils/conv_utils.js\nfunction normalizeArray(value, n, name) {\n  if (typeof value === \"number\") {\n    return pyListRepeat(value, n);\n  } else {\n    if (value.length !== n) {\n      throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${value.length} elements.`);\n    }\n    for (let i = 0; i < n; ++i) {\n      const singleValue = value[i];\n      if (!isInteger(singleValue)) {\n        throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${JSON.stringify(value)} including a non-integer number ${singleValue}`);\n      }\n    }\n    return value;\n  }\n}\nfunction convOutputLength(inputLength, filterSize, padding, stride, dilation = 1) {\n  if (inputLength == null) {\n    return inputLength;\n  }\n  const dilatedFilterSize = filterSize + (filterSize - 1) * (dilation - 1);\n  let outputLength;\n  if (padding === \"same\") {\n    outputLength = inputLength;\n  } else {\n    outputLength = inputLength - dilatedFilterSize + 1;\n  }\n  return Math.floor((outputLength + stride - 1) / stride);\n}\nfunction deconvLength(dimSize, strideSize, kernelSize, padding) {\n  if (dimSize == null) {\n    return null;\n  }\n  if (padding === \"valid\") {\n    dimSize = dimSize * strideSize + max2([kernelSize - strideSize, 0]);\n  } else if (padding === \"same\") {\n    dimSize = dimSize * strideSize;\n  } else {\n    throw new ValueError(`Unsupport padding mode: ${padding}.`);\n  }\n  return dimSize;\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional.js\nfunction preprocessConv2DInput(x, dataFormat) {\n  return tidy(() => {\n    checkDataFormat(dataFormat);\n    if (dataFormat === \"channelsFirst\") {\n      return transpose(x, [0, 2, 3, 1]);\n    } else {\n      return x;\n    }\n  });\n}\nfunction preprocessConv3DInput(x, dataFormat) {\n  return tidy(() => {\n    checkDataFormat(dataFormat);\n    if (dataFormat === \"channelsFirst\") {\n      return transpose(x, [0, 2, 3, 4, 1]);\n    } else {\n      return x;\n    }\n  });\n}\nfunction conv1dWithBias(x, kernel, bias, strides = 1, padding = \"valid\", dataFormat, dilationRate = 1) {\n  return tidy(() => {\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    checkDataFormat(dataFormat);\n    if (x.shape.length !== 3) {\n      throw new ValueError(`The input of a conv1dWithBias operation should be 3, but is ${x.shape.length} instead.`);\n    }\n    if (kernel.shape.length !== 3) {\n      throw new ValueError(`The kernel for a conv1dWithBias operation should be 3, but is ${kernel.shape.length} instead`);\n    }\n    if (bias != null && bias.shape.length !== 1) {\n      throw new ValueError(`The bias for a conv1dWithBias operation should be 1, but is ${kernel.shape.length} instead`);\n    }\n    if (dataFormat === \"channelsFirst\") {\n      x = transpose(x, [0, 2, 1]);\n    }\n    if (padding === \"causal\") {\n      throw new NotImplementedError(\"The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.\");\n    }\n    let y = conv1d(x, kernel, strides, padding === \"same\" ? \"same\" : \"valid\", \"NWC\", dilationRate);\n    if (bias != null) {\n      y = biasAdd(y, bias);\n    }\n    return y;\n  });\n}\nfunction conv2dWithBiasActivation(x, kernel, bias, strides = [1, 1], padding = \"valid\", dataFormat, dilationRate, activation2 = null) {\n  return tidy(() => {\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    checkDataFormat(dataFormat);\n    if (x.rank !== 3 && x.rank !== 4) {\n      throw new ValueError(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${x.rank}.`);\n    }\n    if (kernel.rank !== 3 && kernel.rank !== 4) {\n      throw new ValueError(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${x.rank}.`);\n    }\n    let y = preprocessConv2DInput(x, dataFormat);\n    if (padding === \"causal\") {\n      throw new NotImplementedError(\"The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.\");\n    }\n    y = fused_ops_exports.conv2d({\n      x: y,\n      filter: kernel,\n      strides,\n      pad: padding === \"same\" ? \"same\" : \"valid\",\n      dilations: dilationRate,\n      dataFormat: \"NHWC\",\n      bias,\n      activation: activation2\n    });\n    if (dataFormat === \"channelsFirst\") {\n      y = transpose(y, [0, 3, 1, 2]);\n    }\n    return y;\n  });\n}\nfunction conv3dWithBias(x, kernel, bias, strides = [1, 1, 1], padding = \"valid\", dataFormat, dilationRate) {\n  return tidy(() => {\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    checkDataFormat(dataFormat);\n    if (x.rank !== 4 && x.rank !== 5) {\n      throw new ValueError(`conv3dWithBias expects input to be of rank 4 or 5, but received ${x.rank}.`);\n    }\n    if (kernel.rank !== 4 && kernel.rank !== 5) {\n      throw new ValueError(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${x.rank}.`);\n    }\n    let y = preprocessConv3DInput(x, dataFormat);\n    if (padding === \"causal\") {\n      throw new NotImplementedError(\"The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.\");\n    }\n    y = conv3d(y, kernel, strides, padding === \"same\" ? \"same\" : \"valid\", \"NDHWC\", dilationRate);\n    if (bias != null) {\n      y = biasAdd(y, bias);\n    }\n    if (dataFormat === \"channelsFirst\") {\n      y = transpose(y, [0, 4, 1, 2, 3]);\n    }\n    return y;\n  });\n}\nvar BaseConv = class extends Layer {\n  constructor(rank, args) {\n    super(args);\n    this.bias = null;\n    this.DEFAULT_KERNEL_INITIALIZER = \"glorotNormal\";\n    this.DEFAULT_BIAS_INITIALIZER = \"zeros\";\n    BaseConv.verifyArgs(args);\n    this.rank = rank;\n    assertPositiveInteger(this.rank, \"rank\");\n    if (this.rank !== 1 && this.rank !== 2 && this.rank !== 3) {\n      throw new NotImplementedError(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);\n    }\n    this.kernelSize = normalizeArray(args.kernelSize, rank, \"kernelSize\");\n    this.strides = normalizeArray(args.strides == null ? 1 : args.strides, rank, \"strides\");\n    this.padding = args.padding == null ? \"valid\" : args.padding;\n    checkPaddingMode(this.padding);\n    this.dataFormat = args.dataFormat == null ? \"channelsLast\" : args.dataFormat;\n    checkDataFormat(this.dataFormat);\n    this.activation = getActivation(args.activation);\n    this.useBias = args.useBias == null ? true : args.useBias;\n    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);\n    this.biasConstraint = getConstraint(args.biasConstraint);\n    this.biasRegularizer = getRegularizer(args.biasRegularizer);\n    this.activityRegularizer = getRegularizer(args.activityRegularizer);\n    this.dilationRate = normalizeArray(args.dilationRate == null ? 1 : args.dilationRate, rank, \"dilationRate\");\n    if (this.rank === 1 && (Array.isArray(this.dilationRate) && this.dilationRate.length !== 1)) {\n      throw new ValueError(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);\n    } else if (this.rank === 2) {\n      if (typeof this.dilationRate === \"number\") {\n        this.dilationRate = [this.dilationRate, this.dilationRate];\n      } else if (this.dilationRate.length !== 2) {\n        throw new ValueError(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`);\n      }\n    } else if (this.rank === 3) {\n      if (typeof this.dilationRate === \"number\") {\n        this.dilationRate = [this.dilationRate, this.dilationRate, this.dilationRate];\n      } else if (this.dilationRate.length !== 3) {\n        throw new ValueError(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`);\n      }\n    }\n  }\n  static verifyArgs(args) {\n    assert2(\"kernelSize\" in args, `required key 'kernelSize' not in config`);\n    if (typeof args.kernelSize !== \"number\" && !checkArrayTypeAndLength(args.kernelSize, \"number\", 1, 3)) {\n      throw new ValueError(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(args.kernelSize)}.`);\n    }\n  }\n  getConfig() {\n    const config = {\n      kernelSize: this.kernelSize,\n      strides: this.strides,\n      padding: this.padding,\n      dataFormat: this.dataFormat,\n      dilationRate: this.dilationRate,\n      activation: serializeActivation(this.activation),\n      useBias: this.useBias,\n      biasInitializer: serializeInitializer(this.biasInitializer),\n      biasRegularizer: serializeRegularizer(this.biasRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      biasConstraint: serializeConstraint(this.biasConstraint)\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nvar Conv = class extends BaseConv {\n  constructor(rank, args) {\n    super(rank, args);\n    this.kernel = null;\n    Conv.verifyArgs(args);\n    this.filters = args.filters;\n    assertPositiveInteger(this.filters, \"filters\");\n    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.kernelConstraint = getConstraint(args.kernelConstraint);\n    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : inputShape.length - 1;\n    if (inputShape[channelAxis] == null) {\n      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);\n    }\n    const inputDim = inputShape[channelAxis];\n    const kernelShape = this.kernelSize.concat([inputDim, this.filters]);\n    this.kernel = this.addWeight(\"kernel\", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.filters], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    }\n    this.inputSpec = [{ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } }];\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      let outputs;\n      const biasValue = this.bias == null ? null : this.bias.read();\n      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());\n      if (fusedActivationName != null && this.rank === 2) {\n        outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate, fusedActivationName);\n      } else {\n        if (this.rank === 1) {\n          outputs = conv1dWithBias(inputs, this.kernel.read(), biasValue, this.strides[0], this.padding, this.dataFormat, this.dilationRate[0]);\n        } else if (this.rank === 2) {\n          outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);\n        } else if (this.rank === 3) {\n          outputs = conv3dWithBias(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);\n        } else {\n          throw new NotImplementedError(\"convolutions greater than 3D are not implemented yet.\");\n        }\n        if (this.activation != null) {\n          outputs = this.activation.apply(outputs);\n        }\n      }\n      return outputs;\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const newSpace = [];\n    const space = this.dataFormat === \"channelsLast\" ? inputShape.slice(1, inputShape.length - 1) : inputShape.slice(2);\n    for (let i = 0; i < space.length; ++i) {\n      const newDim = convOutputLength(space[i], this.kernelSize[i], this.padding, this.strides[i], typeof this.dilationRate === \"number\" ? this.dilationRate : this.dilationRate[i]);\n      newSpace.push(newDim);\n    }\n    let outputShape = [inputShape[0]];\n    if (this.dataFormat === \"channelsLast\") {\n      outputShape = outputShape.concat(newSpace);\n      outputShape.push(this.filters);\n    } else {\n      outputShape.push(this.filters);\n      outputShape = outputShape.concat(newSpace);\n    }\n    return outputShape;\n  }\n  getConfig() {\n    const config = {\n      filters: this.filters,\n      kernelInitializer: serializeInitializer(this.kernelInitializer),\n      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),\n      kernelConstraint: serializeConstraint(this.kernelConstraint)\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  static verifyArgs(args) {\n    if (!(\"filters\" in args) || typeof args.filters !== \"number\" || args.filters < 1) {\n      throw new ValueError(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(args.filters)}`);\n    }\n  }\n};\nvar Conv2D2 = class extends Conv {\n  constructor(args) {\n    super(2, args);\n    Conv2D2.verifyArgs(args);\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"rank\"];\n    return config;\n  }\n  static verifyArgs(args) {\n    if (typeof args.kernelSize !== \"number\" && !checkArrayTypeAndLength(args.kernelSize, \"number\", 1, 2)) {\n      throw new ValueError(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(args.kernelSize)}.`);\n    }\n  }\n};\nConv2D2.className = \"Conv2D\";\nserialization_exports.registerClass(Conv2D2);\nvar Conv3D2 = class extends Conv {\n  constructor(args) {\n    super(3, args);\n    Conv3D2.verifyArgs(args);\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"rank\"];\n    return config;\n  }\n  static verifyArgs(args) {\n    if (typeof args.kernelSize !== \"number\") {\n      if (!(Array.isArray(args.kernelSize) && (args.kernelSize.length === 1 || args.kernelSize.length === 3))) {\n        throw new ValueError(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(args.kernelSize)}.`);\n      }\n    }\n  }\n};\nConv3D2.className = \"Conv3D\";\nserialization_exports.registerClass(Conv3D2);\nvar Conv2DTranspose = class extends Conv2D2 {\n  constructor(args) {\n    super(args);\n    this.inputSpec = [new InputSpec({ ndim: 4 })];\n    if (this.padding !== \"same\" && this.padding !== \"valid\") {\n      throw new ValueError(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);\n    }\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape.length !== 4) {\n      throw new ValueError(\"Input should have rank 4; Received input shape: \" + JSON.stringify(inputShape));\n    }\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : inputShape.length - 1;\n    if (inputShape[channelAxis] == null) {\n      throw new ValueError(\"The channel dimension of the inputs should be defined. Found `None`.\");\n    }\n    const inputDim = inputShape[channelAxis];\n    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);\n    this.kernel = this.addWeight(\"kernel\", kernelShape, \"float32\", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.filters], \"float32\", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    }\n    this.inputSpec = [new InputSpec({ ndim: 4, axes: { [channelAxis]: inputDim } })];\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      let input2 = getExactlyOneTensor(inputs);\n      if (input2.shape.length !== 4) {\n        throw new ValueError(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);\n      }\n      const inputShape = input2.shape;\n      const batchSize = inputShape[0];\n      let hAxis;\n      let wAxis;\n      if (this.dataFormat === \"channelsFirst\") {\n        hAxis = 2;\n        wAxis = 3;\n      } else {\n        hAxis = 1;\n        wAxis = 2;\n      }\n      const height = inputShape[hAxis];\n      const width = inputShape[wAxis];\n      const kernelH = this.kernelSize[0];\n      const kernelW = this.kernelSize[1];\n      const strideH = this.strides[0];\n      const strideW = this.strides[1];\n      const outHeight = deconvLength(height, strideH, kernelH, this.padding);\n      const outWidth = deconvLength(width, strideW, kernelW, this.padding);\n      const outputShape = [batchSize, outHeight, outWidth, this.filters];\n      if (this.dataFormat !== \"channelsLast\") {\n        input2 = transpose(input2, [0, 2, 3, 1]);\n      }\n      let outputs = conv2dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);\n      if (this.dataFormat !== \"channelsLast\") {\n        outputs = transpose(outputs, [0, 3, 1, 2]);\n      }\n      if (this.bias != null) {\n        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);\n      }\n      if (this.activation != null) {\n        outputs = this.activation.apply(outputs);\n      }\n      return outputs;\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const outputShape = inputShape.slice();\n    let channelAxis;\n    let heightAxis;\n    let widthAxis;\n    if (this.dataFormat === \"channelsFirst\") {\n      channelAxis = 1;\n      heightAxis = 2;\n      widthAxis = 3;\n    } else {\n      channelAxis = 3;\n      heightAxis = 1;\n      widthAxis = 2;\n    }\n    const kernelH = this.kernelSize[0];\n    const kernelW = this.kernelSize[1];\n    const strideH = this.strides[0];\n    const strideW = this.strides[1];\n    outputShape[channelAxis] = this.filters;\n    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);\n    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);\n    return outputShape;\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"dilationRate\"];\n    return config;\n  }\n};\nConv2DTranspose.className = \"Conv2DTranspose\";\nserialization_exports.registerClass(Conv2DTranspose);\nvar Conv3DTranspose = class extends Conv3D2 {\n  constructor(args) {\n    super(args);\n    this.inputSpec = [new InputSpec({ ndim: 5 })];\n    if (this.padding !== \"same\" && this.padding !== \"valid\") {\n      throw new ValueError(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);\n    }\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape.length !== 5) {\n      throw new ValueError(\"Input should have rank 5; Received input shape: \" + JSON.stringify(inputShape));\n    }\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : inputShape.length - 1;\n    if (inputShape[channelAxis] == null) {\n      throw new ValueError(\"The channel dimension of the inputs should be defined. Found `None`.\");\n    }\n    const inputDim = inputShape[channelAxis];\n    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);\n    this.kernel = this.addWeight(\"kernel\", kernelShape, \"float32\", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.filters], \"float32\", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    }\n    this.inputSpec = [new InputSpec({ ndim: 5, axes: { [channelAxis]: inputDim } })];\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      let input2 = getExactlyOneTensor(inputs);\n      if (input2.shape.length !== 5) {\n        throw new ValueError(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);\n      }\n      const inputShape = input2.shape;\n      const batchSize = inputShape[0];\n      let hAxis;\n      let wAxis;\n      let dAxis;\n      if (this.dataFormat === \"channelsFirst\") {\n        dAxis = 2;\n        hAxis = 3;\n        wAxis = 4;\n      } else {\n        dAxis = 1;\n        hAxis = 2;\n        wAxis = 3;\n      }\n      const depth = inputShape[dAxis];\n      const height = inputShape[hAxis];\n      const width = inputShape[wAxis];\n      const kernelD = this.kernelSize[0];\n      const kernelH = this.kernelSize[1];\n      const kernelW = this.kernelSize[2];\n      const strideD = this.strides[0];\n      const strideH = this.strides[1];\n      const strideW = this.strides[2];\n      const outDepth = deconvLength(depth, strideD, kernelD, this.padding);\n      const outHeight = deconvLength(height, strideH, kernelH, this.padding);\n      const outWidth = deconvLength(width, strideW, kernelW, this.padding);\n      const outputShape = [batchSize, outDepth, outHeight, outWidth, this.filters];\n      if (this.dataFormat !== \"channelsLast\") {\n        input2 = transpose(input2, [0, 2, 3, 4, 1]);\n      }\n      let outputs = conv3dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);\n      if (this.dataFormat !== \"channelsLast\") {\n        outputs = transpose(outputs, [0, 4, 1, 2, 3]);\n      }\n      if (this.bias !== null) {\n        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);\n      }\n      if (this.activation !== null) {\n        outputs = this.activation.apply(outputs);\n      }\n      return outputs;\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const outputShape = inputShape.slice();\n    let channelAxis;\n    let depthAxis;\n    let heightAxis;\n    let widthAxis;\n    if (this.dataFormat === \"channelsFirst\") {\n      channelAxis = 1;\n      depthAxis = 2;\n      heightAxis = 3;\n      widthAxis = 4;\n    } else {\n      channelAxis = 4;\n      depthAxis = 1;\n      heightAxis = 2;\n      widthAxis = 3;\n    }\n    const kernelD = this.kernelSize[0];\n    const kernelH = this.kernelSize[1];\n    const kernelW = this.kernelSize[2];\n    const strideD = this.strides[0];\n    const strideH = this.strides[1];\n    const strideW = this.strides[2];\n    outputShape[channelAxis] = this.filters;\n    outputShape[depthAxis] = deconvLength(outputShape[depthAxis], strideD, kernelD, this.padding);\n    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);\n    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);\n    return outputShape;\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"dilationRate\"];\n    return config;\n  }\n};\nConv3DTranspose.className = \"Conv3DTranspose\";\nserialization_exports.registerClass(Conv3DTranspose);\nvar SeparableConv = class extends Conv {\n  constructor(rank, config) {\n    super(rank, config);\n    this.DEFAULT_DEPTHWISE_INITIALIZER = \"glorotUniform\";\n    this.DEFAULT_POINTWISE_INITIALIZER = \"glorotUniform\";\n    this.depthwiseKernel = null;\n    this.pointwiseKernel = null;\n    if (config.filters == null) {\n      throw new ValueError(\"The `filters` configuration field is required by SeparableConv, but is unspecified.\");\n    }\n    if (config.kernelInitializer != null || config.kernelRegularizer != null || config.kernelConstraint != null) {\n      throw new ValueError(\"Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.\");\n    }\n    if (config.padding != null && config.padding !== \"same\" && config.padding !== \"valid\") {\n      throw new ValueError(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(config.padding)}`);\n    }\n    this.depthMultiplier = config.depthMultiplier == null ? 1 : config.depthMultiplier;\n    this.depthwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_DEPTHWISE_INITIALIZER);\n    this.depthwiseRegularizer = getRegularizer(config.depthwiseRegularizer);\n    this.depthwiseConstraint = getConstraint(config.depthwiseConstraint);\n    this.pointwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_POINTWISE_INITIALIZER);\n    this.pointwiseRegularizer = getRegularizer(config.pointwiseRegularizer);\n    this.pointwiseConstraint = getConstraint(config.pointwiseConstraint);\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape.length < this.rank + 2) {\n      throw new ValueError(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank + 2}, but received input shape: ${JSON.stringify(inputShape)}`);\n    }\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : inputShape.length - 1;\n    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {\n      throw new ValueError(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(inputShape[channelAxis])}`);\n    }\n    const inputDim = inputShape[channelAxis];\n    const depthwiseKernelShape = this.kernelSize.concat([inputDim, this.depthMultiplier]);\n    const pointwiseKernelShape = [];\n    for (let i = 0; i < this.rank; ++i) {\n      pointwiseKernelShape.push(1);\n    }\n    pointwiseKernelShape.push(inputDim * this.depthMultiplier, this.filters);\n    const trainable = true;\n    this.depthwiseKernel = this.addWeight(\"depthwise_kernel\", depthwiseKernelShape, \"float32\", this.depthwiseInitializer, this.depthwiseRegularizer, trainable, this.depthwiseConstraint);\n    this.pointwiseKernel = this.addWeight(\"pointwise_kernel\", pointwiseKernelShape, \"float32\", this.pointwiseInitializer, this.pointwiseRegularizer, trainable, this.pointwiseConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.filters], \"float32\", this.biasInitializer, this.biasRegularizer, trainable, this.biasConstraint);\n    } else {\n      this.bias = null;\n    }\n    this.inputSpec = [new InputSpec({ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } })];\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      let output;\n      if (this.rank === 1) {\n        throw new NotImplementedError(\"1D separable convolution is not implemented yet.\");\n      } else if (this.rank === 2) {\n        if (this.dataFormat === \"channelsFirst\") {\n          inputs = transpose(inputs, [0, 2, 3, 1]);\n        }\n        output = separableConv2d(inputs, this.depthwiseKernel.read(), this.pointwiseKernel.read(), this.strides, this.padding, this.dilationRate, \"NHWC\");\n      }\n      if (this.useBias) {\n        output = biasAdd(output, this.bias.read(), this.dataFormat);\n      }\n      if (this.activation != null) {\n        output = this.activation.apply(output);\n      }\n      if (this.dataFormat === \"channelsFirst\") {\n        output = transpose(output, [0, 3, 1, 2]);\n      }\n      return output;\n    });\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"rank\"];\n    delete config[\"kernelInitializer\"];\n    delete config[\"kernelRegularizer\"];\n    delete config[\"kernelConstraint\"];\n    config[\"depthwiseInitializer\"] = serializeInitializer(this.depthwiseInitializer);\n    config[\"pointwiseInitializer\"] = serializeInitializer(this.pointwiseInitializer);\n    config[\"depthwiseRegularizer\"] = serializeRegularizer(this.depthwiseRegularizer);\n    config[\"pointwiseRegularizer\"] = serializeRegularizer(this.pointwiseRegularizer);\n    config[\"depthwiseConstraint\"] = serializeConstraint(this.depthwiseConstraint);\n    config[\"pointwiseConstraint\"] = serializeConstraint(this.pointwiseConstraint);\n    return config;\n  }\n};\nSeparableConv.className = \"SeparableConv\";\nvar SeparableConv2D = class extends SeparableConv {\n  constructor(args) {\n    super(2, args);\n  }\n};\nSeparableConv2D.className = \"SeparableConv2D\";\nserialization_exports.registerClass(SeparableConv2D);\nvar Conv1D = class extends Conv {\n  constructor(args) {\n    super(1, args);\n    Conv1D.verifyArgs(args);\n    this.inputSpec = [{ ndim: 3 }];\n  }\n  getConfig() {\n    const config = super.getConfig();\n    delete config[\"rank\"];\n    delete config[\"dataFormat\"];\n    return config;\n  }\n  static verifyArgs(args) {\n    if (typeof args.kernelSize !== \"number\" && !checkArrayTypeAndLength(args.kernelSize, \"number\", 1, 1)) {\n      throw new ValueError(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(args.kernelSize)}.`);\n    }\n  }\n};\nConv1D.className = \"Conv1D\";\nserialization_exports.registerClass(Conv1D);\nvar Cropping2D = class extends Layer {\n  constructor(args) {\n    super(args);\n    if (typeof args.cropping === \"number\") {\n      this.cropping = [[args.cropping, args.cropping], [args.cropping, args.cropping]];\n    } else if (typeof args.cropping[0] === \"number\") {\n      this.cropping = [\n        [args.cropping[0], args.cropping[0]],\n        [args.cropping[1], args.cropping[1]]\n      ];\n    } else {\n      this.cropping = args.cropping;\n    }\n    this.dataFormat = args.dataFormat === void 0 ? \"channelsLast\" : args.dataFormat;\n    this.inputSpec = [{ ndim: 4 }];\n  }\n  computeOutputShape(inputShape) {\n    if (this.dataFormat === \"channelsFirst\") {\n      return [\n        inputShape[0],\n        inputShape[1],\n        inputShape[2] - this.cropping[0][0] - this.cropping[0][1],\n        inputShape[3] - this.cropping[1][0] - this.cropping[1][1]\n      ];\n    } else {\n      return [\n        inputShape[0],\n        inputShape[1] - this.cropping[0][0] - this.cropping[0][1],\n        inputShape[2] - this.cropping[1][0] - this.cropping[1][1],\n        inputShape[3]\n      ];\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      if (this.dataFormat === \"channelsLast\") {\n        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[1] - this.cropping[0][0] - this.cropping[0][1], 2);\n        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[2] - this.cropping[1][1] - this.cropping[1][0], 3);\n      } else {\n        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[2] - this.cropping[0][0] - this.cropping[0][1], 3);\n        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[3] - this.cropping[1][1] - this.cropping[1][0], 4);\n      }\n    });\n  }\n  getConfig() {\n    const config = { cropping: this.cropping, dataFormat: this.dataFormat };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nCropping2D.className = \"Cropping2D\";\nserialization_exports.registerClass(Cropping2D);\nvar UpSampling2D = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.DEFAULT_SIZE = [2, 2];\n    this.inputSpec = [{ ndim: 4 }];\n    this.size = args.size == null ? this.DEFAULT_SIZE : args.size;\n    this.dataFormat = args.dataFormat == null ? \"channelsLast\" : args.dataFormat;\n    checkDataFormat(this.dataFormat);\n    this.interpolation = args.interpolation == null ? \"nearest\" : args.interpolation;\n    checkInterpolationFormat(this.interpolation);\n  }\n  computeOutputShape(inputShape) {\n    if (this.dataFormat === \"channelsFirst\") {\n      const height = inputShape[2] == null ? null : this.size[0] * inputShape[2];\n      const width = inputShape[3] == null ? null : this.size[1] * inputShape[3];\n      return [inputShape[0], inputShape[1], height, width];\n    } else {\n      const height = inputShape[1] == null ? null : this.size[0] * inputShape[1];\n      const width = inputShape[2] == null ? null : this.size[1] * inputShape[2];\n      return [inputShape[0], height, width, inputShape[3]];\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      let input2 = getExactlyOneTensor(inputs);\n      const inputShape = input2.shape;\n      if (this.dataFormat === \"channelsFirst\") {\n        input2 = transpose(input2, [0, 2, 3, 1]);\n        const height = this.size[0] * inputShape[2];\n        const width = this.size[1] * inputShape[3];\n        const resized = this.interpolation === \"nearest\" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);\n        return transpose(resized, [0, 3, 1, 2]);\n      } else {\n        const height = this.size[0] * inputShape[1];\n        const width = this.size[1] * inputShape[2];\n        return this.interpolation === \"nearest\" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);\n      }\n    });\n  }\n  getConfig() {\n    const config = {\n      size: this.size,\n      dataFormat: this.dataFormat,\n      interpolation: this.interpolation\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nUpSampling2D.className = \"UpSampling2D\";\nserialization_exports.registerClass(UpSampling2D);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional_depthwise.js\nfunction depthwiseConv2d3(x, depthwiseKernel, strides = [1, 1], padding = \"valid\", dataFormat, dilationRate) {\n  return tidy(() => {\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    checkDataFormat(dataFormat);\n    let y = preprocessConv2DInput(x, dataFormat);\n    if (x.rank !== 4) {\n      throw new ValueError(`Input for depthwiseConv2d is required to be 4-D, but is instead ${x.rank}-D`);\n    }\n    if (depthwiseKernel.rank !== 4) {\n      throw new ValueError(`depthwiseKernel is required to be 4-D, but is instead ${depthwiseKernel.rank}-D`);\n    }\n    y = depthwiseConv2d(y, depthwiseKernel, strides, padding === \"same\" ? \"same\" : \"valid\", \"NHWC\", dilationRate);\n    if (dataFormat === \"channelsFirst\") {\n      y = transpose(y, [0, 3, 1, 2]);\n    }\n    return y;\n  });\n}\nvar DepthwiseConv2D = class extends BaseConv {\n  constructor(args) {\n    super(2, args);\n    this.depthwiseKernel = null;\n    this.depthMultiplier = args.depthMultiplier == null ? 1 : args.depthMultiplier;\n    this.depthwiseInitializer = getInitializer(args.depthwiseInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.depthwiseConstraint = getConstraint(args.depthwiseConstraint);\n    this.depthwiseRegularizer = getRegularizer(args.depthwiseRegularizer);\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape.length < 4) {\n      throw new ValueError(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(inputShape)}.`);\n    }\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : 3;\n    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {\n      throw new ValueError(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${inputShape[channelAxis]}).`);\n    }\n    const inputDim = inputShape[channelAxis];\n    const depthwiseKernelShape = [\n      this.kernelSize[0],\n      this.kernelSize[1],\n      inputDim,\n      this.depthMultiplier\n    ];\n    this.depthwiseKernel = this.addWeight(\"depthwise_kernel\", depthwiseKernelShape, null, this.depthwiseInitializer, this.depthwiseRegularizer, true, this.depthwiseConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [inputDim * this.depthMultiplier], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    } else {\n      this.bias = null;\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      let outputs = depthwiseConv2d3(inputs, this.depthwiseKernel.read(), this.strides, this.padding, this.dataFormat, null);\n      if (this.useBias) {\n        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);\n      }\n      if (this.activation != null) {\n        outputs = this.activation.apply(outputs);\n      }\n      return outputs;\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const rows = this.dataFormat === \"channelsFirst\" ? inputShape[2] : inputShape[1];\n    const cols = this.dataFormat === \"channelsFirst\" ? inputShape[3] : inputShape[2];\n    const outFilters = this.dataFormat === \"channelsFirst\" ? inputShape[1] * this.depthMultiplier : inputShape[3] * this.depthMultiplier;\n    const outRows = convOutputLength(rows, this.kernelSize[0], this.padding, this.strides[0]);\n    const outCols = convOutputLength(cols, this.kernelSize[1], this.padding, this.strides[1]);\n    if (this.dataFormat === \"channelsFirst\") {\n      return [inputShape[0], outFilters, outRows, outCols];\n    } else {\n      return [inputShape[0], outRows, outCols, outFilters];\n    }\n  }\n  getConfig() {\n    const config = super.getConfig();\n    config[\"depthMultiplier\"] = this.depthMultiplier;\n    config[\"depthwiseInitializer\"] = serializeInitializer(this.depthwiseInitializer);\n    config[\"depthwiseRegularizer\"] = serializeRegularizer(this.depthwiseRegularizer);\n    config[\"depthwiseConstraint\"] = serializeConstraint(this.depthwiseRegularizer);\n    return config;\n  }\n};\nDepthwiseConv2D.className = \"DepthwiseConv2D\";\nserialization_exports.registerClass(DepthwiseConv2D);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/recurrent.js\nfunction standardizeArgs(inputs, initialState, constants, numConstants) {\n  if (Array.isArray(inputs)) {\n    if (initialState != null || constants != null) {\n      throw new ValueError(\"When inputs is an array, neither initialState or constants should be provided\");\n    }\n    if (numConstants != null) {\n      constants = inputs.slice(inputs.length - numConstants, inputs.length);\n      inputs = inputs.slice(0, inputs.length - numConstants);\n    }\n    if (inputs.length > 1) {\n      initialState = inputs.slice(1, inputs.length);\n    }\n    inputs = inputs[0];\n  }\n  function toListOrNull(x) {\n    if (x == null || Array.isArray(x)) {\n      return x;\n    } else {\n      return [x];\n    }\n  }\n  initialState = toListOrNull(initialState);\n  constants = toListOrNull(constants);\n  return { inputs, initialState, constants };\n}\nfunction rnn(stepFunction, inputs, initialStates, goBackwards = false, mask, constants, unroll = false, needPerStepOutputs = false) {\n  return tidy(() => {\n    const ndim = inputs.shape.length;\n    if (ndim < 3) {\n      throw new ValueError(`Input should be at least 3D, but is ${ndim}D.`);\n    }\n    const axes = [1, 0].concat(range2(2, ndim));\n    inputs = transpose(inputs, axes);\n    if (constants != null) {\n      throw new NotImplementedError(\"The rnn() functoin of the deeplearn.js backend does not support constants yet.\");\n    }\n    if (unroll) {\n      console.warn(\"Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend.\");\n    }\n    if (mask != null) {\n      mask = cast(cast(mask, \"bool\"), \"float32\");\n      if (mask.rank === ndim - 1) {\n        mask = expandDims(mask, -1);\n      }\n      mask = transpose(mask, axes);\n    }\n    if (goBackwards) {\n      inputs = reverse(inputs, 0);\n      if (mask != null) {\n        mask = reverse(mask, 0);\n      }\n    }\n    const perStepOutputs = [];\n    let lastOutput;\n    let states = initialStates;\n    const timeSteps = inputs.shape[0];\n    const perStepInputs = unstack(inputs);\n    let perStepMasks;\n    if (mask != null) {\n      perStepMasks = unstack(mask);\n    }\n    for (let t = 0; t < timeSteps; ++t) {\n      const currentInput = perStepInputs[t];\n      const stepOutputs = tidy(() => stepFunction(currentInput, states));\n      if (mask == null) {\n        lastOutput = stepOutputs[0];\n        states = stepOutputs[1];\n      } else {\n        const maskedOutputs = tidy(() => {\n          const stepMask = perStepMasks[t];\n          const negStepMask = sub(onesLike(stepMask), stepMask);\n          const output = add2(mul(stepOutputs[0], stepMask), mul(states[0], negStepMask));\n          const newStates = states.map((state, i) => {\n            return add2(mul(stepOutputs[1][i], stepMask), mul(state, negStepMask));\n          });\n          return { output, newStates };\n        });\n        lastOutput = maskedOutputs.output;\n        states = maskedOutputs.newStates;\n      }\n      if (needPerStepOutputs) {\n        perStepOutputs.push(lastOutput);\n      }\n    }\n    let outputs;\n    if (needPerStepOutputs) {\n      const axis = 1;\n      outputs = stack(perStepOutputs, axis);\n    }\n    return [lastOutput, outputs, states];\n  });\n}\nvar RNN = class extends Layer {\n  constructor(args) {\n    super(args);\n    let cell;\n    if (args.cell == null) {\n      throw new ValueError(\"cell property is missing for the constructor of RNN.\");\n    } else if (Array.isArray(args.cell)) {\n      cell = new StackedRNNCells({ cells: args.cell });\n    } else {\n      cell = args.cell;\n    }\n    if (cell.stateSize == null) {\n      throw new ValueError(\"The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).\");\n    }\n    this.cell = cell;\n    this.returnSequences = args.returnSequences == null ? false : args.returnSequences;\n    this.returnState = args.returnState == null ? false : args.returnState;\n    this.goBackwards = args.goBackwards == null ? false : args.goBackwards;\n    this._stateful = args.stateful == null ? false : args.stateful;\n    this.unroll = args.unroll == null ? false : args.unroll;\n    this.supportsMasking = true;\n    this.inputSpec = [new InputSpec({ ndim: 3 })];\n    this.stateSpec = null;\n    this.states_ = null;\n    this.numConstants = null;\n    this.keptStates = [];\n  }\n  getStates() {\n    if (this.states_ == null) {\n      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;\n      return range2(0, numStates).map((x) => null);\n    } else {\n      return this.states_;\n    }\n  }\n  setStates(states) {\n    this.states_ = states;\n  }\n  computeOutputShape(inputShape) {\n    if (isArrayOfShapes(inputShape)) {\n      inputShape = inputShape[0];\n    }\n    inputShape = inputShape;\n    let stateSize = this.cell.stateSize;\n    if (!Array.isArray(stateSize)) {\n      stateSize = [stateSize];\n    }\n    const outputDim = stateSize[0];\n    let outputShape;\n    if (this.returnSequences) {\n      outputShape = [inputShape[0], inputShape[1], outputDim];\n    } else {\n      outputShape = [inputShape[0], outputDim];\n    }\n    if (this.returnState) {\n      const stateShape = [];\n      for (const dim of stateSize) {\n        stateShape.push([inputShape[0], dim]);\n      }\n      return [outputShape].concat(stateShape);\n    } else {\n      return outputShape;\n    }\n  }\n  computeMask(inputs, mask) {\n    return tidy(() => {\n      if (Array.isArray(mask)) {\n        mask = mask[0];\n      }\n      const outputMask = this.returnSequences ? mask : null;\n      if (this.returnState) {\n        const stateMask = this.states.map((s) => null);\n        return [outputMask].concat(stateMask);\n      } else {\n        return outputMask;\n      }\n    });\n  }\n  get states() {\n    if (this.states_ == null) {\n      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;\n      const output = [];\n      for (let i = 0; i < numStates; ++i) {\n        output.push(null);\n      }\n      return output;\n    } else {\n      return this.states_;\n    }\n  }\n  set states(s) {\n    this.states_ = s;\n  }\n  build(inputShape) {\n    const constantShape = null;\n    if (this.numConstants != null) {\n      throw new NotImplementedError(\"Constants support is not implemented in RNN yet.\");\n    }\n    if (isArrayOfShapes(inputShape)) {\n      inputShape = inputShape[0];\n    }\n    inputShape = inputShape;\n    const batchSize = this.stateful ? inputShape[0] : null;\n    const inputDim = inputShape.slice(2);\n    this.inputSpec[0] = new InputSpec({ shape: [batchSize, null, ...inputDim] });\n    const stepInputShape = [inputShape[0]].concat(inputShape.slice(2));\n    if (constantShape != null) {\n      throw new NotImplementedError(\"Constants support is not implemented in RNN yet.\");\n    } else {\n      this.cell.build(stepInputShape);\n    }\n    let stateSize;\n    if (Array.isArray(this.cell.stateSize)) {\n      stateSize = this.cell.stateSize;\n    } else {\n      stateSize = [this.cell.stateSize];\n    }\n    if (this.stateSpec != null) {\n      if (!util_exports.arraysEqual(this.stateSpec.map((spec) => spec.shape[spec.shape.length - 1]), stateSize)) {\n        throw new ValueError(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`);\n      }\n    } else {\n      this.stateSpec = stateSize.map((dim) => new InputSpec({ shape: [null, dim] }));\n    }\n    if (this.stateful) {\n      this.resetStates();\n    }\n  }\n  resetStates(states, training = false) {\n    tidy(() => {\n      if (!this.stateful) {\n        throw new AttributeError(\"Cannot call resetStates() on an RNN Layer that is not stateful.\");\n      }\n      const batchSize = this.inputSpec[0].shape[0];\n      if (batchSize == null) {\n        throw new ValueError(\"If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \\n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.\");\n      }\n      if (this.states_ == null) {\n        if (Array.isArray(this.cell.stateSize)) {\n          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));\n        } else {\n          this.states_ = [zeros([batchSize, this.cell.stateSize])];\n        }\n      } else if (states == null) {\n        dispose(this.states_);\n        if (this.keptStates != null) {\n          dispose(this.keptStates);\n          this.keptStates = [];\n        }\n        if (Array.isArray(this.cell.stateSize)) {\n          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));\n        } else {\n          this.states_[0] = zeros([batchSize, this.cell.stateSize]);\n        }\n      } else {\n        if (!Array.isArray(states)) {\n          states = [states];\n        }\n        if (states.length !== this.states_.length) {\n          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);\n        }\n        if (training === true) {\n          this.keptStates.push(this.states_.slice());\n        } else {\n          dispose(this.states_);\n        }\n        for (let index = 0; index < this.states_.length; ++index) {\n          const value = states[index];\n          const dim = Array.isArray(this.cell.stateSize) ? this.cell.stateSize[index] : this.cell.stateSize;\n          const expectedShape = [batchSize, dim];\n          if (!util_exports.arraysEqual(value.shape, expectedShape)) {\n            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);\n          }\n          this.states_[index] = value;\n        }\n      }\n      this.states_ = this.states_.map((state) => keep(state.clone()));\n    });\n  }\n  apply(inputs, kwargs) {\n    let initialState = kwargs == null ? null : kwargs[\"initialState\"];\n    let constants = kwargs == null ? null : kwargs[\"constants\"];\n    if (kwargs == null) {\n      kwargs = {};\n    }\n    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);\n    inputs = standardized.inputs;\n    initialState = standardized.initialState;\n    constants = standardized.constants;\n    let additionalInputs = [];\n    let additionalSpecs = [];\n    if (initialState != null) {\n      kwargs[\"initialState\"] = initialState;\n      additionalInputs = additionalInputs.concat(initialState);\n      this.stateSpec = [];\n      for (const state of initialState) {\n        this.stateSpec.push(new InputSpec({ shape: state.shape }));\n      }\n      additionalSpecs = additionalSpecs.concat(this.stateSpec);\n    }\n    if (constants != null) {\n      kwargs[\"constants\"] = constants;\n      additionalInputs = additionalInputs.concat(constants);\n      this.numConstants = constants.length;\n    }\n    const isTensor = additionalInputs[0] instanceof SymbolicTensor;\n    if (isTensor) {\n      const fullInput = [inputs].concat(additionalInputs);\n      const fullInputSpec = this.inputSpec.concat(additionalSpecs);\n      const originalInputSpec = this.inputSpec;\n      this.inputSpec = fullInputSpec;\n      const output = super.apply(fullInput, kwargs);\n      this.inputSpec = originalInputSpec;\n      return output;\n    } else {\n      return super.apply(inputs, kwargs);\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const mask = kwargs == null ? null : kwargs[\"mask\"];\n      const training = kwargs == null ? null : kwargs[\"training\"];\n      let initialState = kwargs == null ? null : kwargs[\"initialState\"];\n      inputs = getExactlyOneTensor(inputs);\n      if (initialState == null) {\n        if (this.stateful) {\n          initialState = this.states_;\n        } else {\n          initialState = this.getInitialState(inputs);\n        }\n      }\n      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;\n      if (initialState.length !== numStates) {\n        throw new ValueError(`RNN Layer has ${numStates} state(s) but was passed ${initialState.length} initial state(s).`);\n      }\n      if (this.unroll) {\n        console.warn(\"Ignoring unroll = true for RNN layer, due to imperative backend.\");\n      }\n      const cellCallKwargs = { training };\n      const step5 = (inputs2, states2) => {\n        const outputs2 = this.cell.call([inputs2].concat(states2), cellCallKwargs);\n        return [outputs2[0], outputs2.slice(1)];\n      };\n      const rnnOutputs = rnn(step5, inputs, initialState, this.goBackwards, mask, null, this.unroll, this.returnSequences);\n      const lastOutput = rnnOutputs[0];\n      const outputs = rnnOutputs[1];\n      const states = rnnOutputs[2];\n      if (this.stateful) {\n        this.resetStates(states, training);\n      }\n      const output = this.returnSequences ? outputs : lastOutput;\n      if (this.returnState) {\n        return [output].concat(states);\n      } else {\n        return output;\n      }\n    });\n  }\n  getInitialState(inputs) {\n    return tidy(() => {\n      let initialState = zeros(inputs.shape);\n      initialState = sum2(initialState, [1, 2]);\n      initialState = expandDims2(initialState);\n      if (Array.isArray(this.cell.stateSize)) {\n        return this.cell.stateSize.map((dim) => dim > 1 ? tile2(initialState, [1, dim]) : initialState);\n      } else {\n        return this.cell.stateSize > 1 ? [tile2(initialState, [1, this.cell.stateSize])] : [initialState];\n      }\n    });\n  }\n  get trainableWeights() {\n    if (!this.trainable) {\n      return [];\n    }\n    return this.cell.trainableWeights;\n  }\n  get nonTrainableWeights() {\n    if (!this.trainable) {\n      return this.cell.weights;\n    }\n    return this.cell.nonTrainableWeights;\n  }\n  setFastWeightInitDuringBuild(value) {\n    super.setFastWeightInitDuringBuild(value);\n    if (this.cell != null) {\n      this.cell.setFastWeightInitDuringBuild(value);\n    }\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = {\n      returnSequences: this.returnSequences,\n      returnState: this.returnState,\n      goBackwards: this.goBackwards,\n      stateful: this.stateful,\n      unroll: this.unroll\n    };\n    if (this.numConstants != null) {\n      config[\"numConstants\"] = this.numConstants;\n    }\n    const cellConfig = this.cell.getConfig();\n    if (this.getClassName() === RNN.className) {\n      config[\"cell\"] = {\n        \"className\": this.cell.getClassName(),\n        \"config\": cellConfig\n      };\n    }\n    return Object.assign(Object.assign(Object.assign({}, cellConfig), baseConfig), config);\n  }\n  static fromConfig(cls, config, customObjects = {}) {\n    const cellConfig = config[\"cell\"];\n    const cell = deserialize(cellConfig, customObjects);\n    return new cls(Object.assign(config, { cell }));\n  }\n};\nRNN.className = \"RNN\";\nserialization_exports.registerClass(RNN);\nvar RNNCell = class extends Layer {\n};\nvar SimpleRNNCell = class extends RNNCell {\n  constructor(args) {\n    super(args);\n    this.DEFAULT_ACTIVATION = \"tanh\";\n    this.DEFAULT_KERNEL_INITIALIZER = \"glorotNormal\";\n    this.DEFAULT_RECURRENT_INITIALIZER = \"orthogonal\";\n    this.DEFAULT_BIAS_INITIALIZER = \"zeros\";\n    this.units = args.units;\n    assertPositiveInteger(this.units, `units`);\n    this.activation = getActivation(args.activation == null ? this.DEFAULT_ACTIVATION : args.activation);\n    this.useBias = args.useBias == null ? true : args.useBias;\n    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);\n    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);\n    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);\n    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);\n    this.biasRegularizer = getRegularizer(args.biasRegularizer);\n    this.kernelConstraint = getConstraint(args.kernelConstraint);\n    this.recurrentConstraint = getConstraint(args.recurrentConstraint);\n    this.biasConstraint = getConstraint(args.biasConstraint);\n    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);\n    this.recurrentDropout = min2([\n      1,\n      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])\n    ]);\n    this.dropoutFunc = args.dropoutFunc;\n    this.stateSize = this.units;\n    this.dropoutMask = null;\n    this.recurrentDropoutMask = null;\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    this.kernel = this.addWeight(\"kernel\", [inputShape[inputShape.length - 1], this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    this.recurrentKernel = this.addWeight(\"recurrent_kernel\", [this.units, this.units], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    } else {\n      this.bias = null;\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = inputs;\n      if (inputs.length !== 2) {\n        throw new ValueError(`SimpleRNNCell expects 2 input Tensors, got ${inputs.length}.`);\n      }\n      let prevOutput = inputs[1];\n      inputs = inputs[0];\n      const training = kwargs[\"training\"] == null ? false : kwargs[\"training\"];\n      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {\n        this.dropoutMask = generateDropoutMask({\n          ones: () => onesLike(inputs),\n          rate: this.dropout,\n          training,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {\n        this.recurrentDropoutMask = generateDropoutMask({\n          ones: () => onesLike(prevOutput),\n          rate: this.recurrentDropout,\n          training,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      let h;\n      const dpMask = this.dropoutMask;\n      const recDpMask = this.recurrentDropoutMask;\n      if (dpMask != null) {\n        h = dot2(mul(inputs, dpMask), this.kernel.read());\n      } else {\n        h = dot2(inputs, this.kernel.read());\n      }\n      if (this.bias != null) {\n        h = biasAdd(h, this.bias.read());\n      }\n      if (recDpMask != null) {\n        prevOutput = mul(prevOutput, recDpMask);\n      }\n      let output = add2(h, dot2(prevOutput, this.recurrentKernel.read()));\n      if (this.activation != null) {\n        output = this.activation.apply(output);\n      }\n      return [output, output];\n    });\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = {\n      units: this.units,\n      activation: serializeActivation(this.activation),\n      useBias: this.useBias,\n      kernelInitializer: serializeInitializer(this.kernelInitializer),\n      recurrentInitializer: serializeInitializer(this.recurrentInitializer),\n      biasInitializer: serializeInitializer(this.biasInitializer),\n      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),\n      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),\n      biasRegularizer: serializeRegularizer(this.biasRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      kernelConstraint: serializeConstraint(this.kernelConstraint),\n      recurrentConstraint: serializeConstraint(this.recurrentConstraint),\n      biasConstraint: serializeConstraint(this.biasConstraint),\n      dropout: this.dropout,\n      recurrentDropout: this.recurrentDropout\n    };\n    return Object.assign(Object.assign({}, baseConfig), config);\n  }\n};\nSimpleRNNCell.className = \"SimpleRNNCell\";\nserialization_exports.registerClass(SimpleRNNCell);\nvar SimpleRNN = class extends RNN {\n  constructor(args) {\n    args.cell = new SimpleRNNCell(args);\n    super(args);\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (this.cell.dropoutMask != null) {\n        dispose(this.cell.dropoutMask);\n        this.cell.dropoutMask = null;\n      }\n      if (this.cell.recurrentDropoutMask != null) {\n        dispose(this.cell.recurrentDropoutMask);\n        this.cell.recurrentDropoutMask = null;\n      }\n      const mask = kwargs == null ? null : kwargs[\"mask\"];\n      const training = kwargs == null ? null : kwargs[\"training\"];\n      const initialState = kwargs == null ? null : kwargs[\"initialState\"];\n      return super.call(inputs, { mask, training, initialState });\n    });\n  }\n  static fromConfig(cls, config) {\n    return new cls(config);\n  }\n};\nSimpleRNN.className = \"SimpleRNN\";\nserialization_exports.registerClass(SimpleRNN);\nvar GRUCell = class extends RNNCell {\n  constructor(args) {\n    super(args);\n    this.DEFAULT_ACTIVATION = \"tanh\";\n    this.DEFAULT_RECURRENT_ACTIVATION = \"hardSigmoid\";\n    this.DEFAULT_KERNEL_INITIALIZER = \"glorotNormal\";\n    this.DEFAULT_RECURRENT_INITIALIZER = \"orthogonal\";\n    this.DEFAULT_BIAS_INITIALIZER = \"zeros\";\n    if (args.resetAfter) {\n      throw new ValueError(`GRUCell does not support reset_after parameter set to true.`);\n    }\n    this.units = args.units;\n    assertPositiveInteger(this.units, \"units\");\n    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);\n    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);\n    this.useBias = args.useBias == null ? true : args.useBias;\n    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);\n    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);\n    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);\n    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);\n    this.biasRegularizer = getRegularizer(args.biasRegularizer);\n    this.kernelConstraint = getConstraint(args.kernelConstraint);\n    this.recurrentConstraint = getConstraint(args.recurrentConstraint);\n    this.biasConstraint = getConstraint(args.biasConstraint);\n    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);\n    this.recurrentDropout = min2([\n      1,\n      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])\n    ]);\n    this.dropoutFunc = args.dropoutFunc;\n    this.implementation = args.implementation;\n    this.stateSize = this.units;\n    this.dropoutMask = null;\n    this.recurrentDropoutMask = null;\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const inputDim = inputShape[inputShape.length - 1];\n    this.kernel = this.addWeight(\"kernel\", [inputDim, this.units * 3], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    this.recurrentKernel = this.addWeight(\"recurrent_kernel\", [this.units, this.units * 3], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);\n    if (this.useBias) {\n      this.bias = this.addWeight(\"bias\", [this.units * 3], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    } else {\n      this.bias = null;\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = inputs;\n      if (inputs.length !== 2) {\n        throw new ValueError(`GRUCell expects 2 input Tensors (inputs, h, c), got ${inputs.length}.`);\n      }\n      const training = kwargs[\"training\"] == null ? false : kwargs[\"training\"];\n      let hTMinus1 = inputs[1];\n      inputs = inputs[0];\n      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {\n        this.dropoutMask = generateDropoutMask({\n          ones: () => onesLike(inputs),\n          rate: this.dropout,\n          training,\n          count: 3,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {\n        this.recurrentDropoutMask = generateDropoutMask({\n          ones: () => onesLike(hTMinus1),\n          rate: this.recurrentDropout,\n          training,\n          count: 3,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      const dpMask = this.dropoutMask;\n      const recDpMask = this.recurrentDropoutMask;\n      let z;\n      let r;\n      let hh;\n      if (0 < this.dropout && this.dropout < 1) {\n        inputs = mul(inputs, dpMask[0]);\n      }\n      let matrixX = dot2(inputs, this.kernel.read());\n      if (this.useBias) {\n        matrixX = biasAdd(matrixX, this.bias.read());\n      }\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {\n        hTMinus1 = mul(hTMinus1, recDpMask[0]);\n      }\n      const recurrentKernelValue = this.recurrentKernel.read();\n      const [rk1, rk2] = split(recurrentKernelValue, [2 * this.units, this.units], recurrentKernelValue.rank - 1);\n      const matrixInner = dot2(hTMinus1, rk1);\n      const [xZ, xR, xH] = split(matrixX, 3, matrixX.rank - 1);\n      const [recurrentZ, recurrentR] = split(matrixInner, 2, matrixInner.rank - 1);\n      z = this.recurrentActivation.apply(add2(xZ, recurrentZ));\n      r = this.recurrentActivation.apply(add2(xR, recurrentR));\n      const recurrentH = dot2(mul(r, hTMinus1), rk2);\n      hh = this.activation.apply(add2(xH, recurrentH));\n      const h = add2(mul(z, hTMinus1), mul(add2(1, neg(z)), hh));\n      return [h, h];\n    });\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = {\n      units: this.units,\n      activation: serializeActivation(this.activation),\n      recurrentActivation: serializeActivation(this.recurrentActivation),\n      useBias: this.useBias,\n      kernelInitializer: serializeInitializer(this.kernelInitializer),\n      recurrentInitializer: serializeInitializer(this.recurrentInitializer),\n      biasInitializer: serializeInitializer(this.biasInitializer),\n      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),\n      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),\n      biasRegularizer: serializeRegularizer(this.biasRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      kernelConstraint: serializeConstraint(this.kernelConstraint),\n      recurrentConstraint: serializeConstraint(this.recurrentConstraint),\n      biasConstraint: serializeConstraint(this.biasConstraint),\n      dropout: this.dropout,\n      recurrentDropout: this.recurrentDropout,\n      implementation: this.implementation,\n      resetAfter: false\n    };\n    return Object.assign(Object.assign({}, baseConfig), config);\n  }\n};\nGRUCell.className = \"GRUCell\";\nserialization_exports.registerClass(GRUCell);\nvar GRU = class extends RNN {\n  constructor(args) {\n    if (args.implementation === 0) {\n      console.warn(\"`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.\");\n    }\n    args.cell = new GRUCell(args);\n    super(args);\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (this.cell.dropoutMask != null) {\n        dispose(this.cell.dropoutMask);\n        this.cell.dropoutMask = null;\n      }\n      if (this.cell.recurrentDropoutMask != null) {\n        dispose(this.cell.recurrentDropoutMask);\n        this.cell.recurrentDropoutMask = null;\n      }\n      const mask = kwargs == null ? null : kwargs[\"mask\"];\n      const training = kwargs == null ? null : kwargs[\"training\"];\n      const initialState = kwargs == null ? null : kwargs[\"initialState\"];\n      return super.call(inputs, { mask, training, initialState });\n    });\n  }\n  static fromConfig(cls, config) {\n    if (config[\"implmentation\"] === 0) {\n      config[\"implementation\"] = 1;\n    }\n    return new cls(config);\n  }\n};\nGRU.className = \"GRU\";\nserialization_exports.registerClass(GRU);\nvar LSTMCell = class extends RNNCell {\n  constructor(args) {\n    super(args);\n    this.DEFAULT_ACTIVATION = \"tanh\";\n    this.DEFAULT_RECURRENT_ACTIVATION = \"hardSigmoid\";\n    this.DEFAULT_KERNEL_INITIALIZER = \"glorotNormal\";\n    this.DEFAULT_RECURRENT_INITIALIZER = \"orthogonal\";\n    this.DEFAULT_BIAS_INITIALIZER = \"zeros\";\n    this.units = args.units;\n    assertPositiveInteger(this.units, \"units\");\n    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);\n    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);\n    this.useBias = args.useBias == null ? true : args.useBias;\n    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);\n    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);\n    this.unitForgetBias = args.unitForgetBias;\n    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);\n    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);\n    this.biasRegularizer = getRegularizer(args.biasRegularizer);\n    this.kernelConstraint = getConstraint(args.kernelConstraint);\n    this.recurrentConstraint = getConstraint(args.recurrentConstraint);\n    this.biasConstraint = getConstraint(args.biasConstraint);\n    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);\n    this.recurrentDropout = min2([\n      1,\n      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])\n    ]);\n    this.dropoutFunc = args.dropoutFunc;\n    this.implementation = args.implementation;\n    this.stateSize = [this.units, this.units];\n    this.dropoutMask = null;\n    this.recurrentDropoutMask = null;\n  }\n  build(inputShape) {\n    var _a;\n    inputShape = getExactlyOneShape(inputShape);\n    const inputDim = inputShape[inputShape.length - 1];\n    this.kernel = this.addWeight(\"kernel\", [inputDim, this.units * 4], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    this.recurrentKernel = this.addWeight(\"recurrent_kernel\", [this.units, this.units * 4], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);\n    let biasInitializer;\n    if (this.useBias) {\n      if (this.unitForgetBias) {\n        const capturedBiasInit = this.biasInitializer;\n        const capturedUnits = this.units;\n        biasInitializer = new (_a = class CustomInit extends Initializer {\n          apply(shape, dtype) {\n            const bI = capturedBiasInit.apply([capturedUnits]);\n            const bF = new Ones().apply([capturedUnits]);\n            const bCAndH = capturedBiasInit.apply([capturedUnits * 2]);\n            return concatAlongFirstAxis(concatAlongFirstAxis(bI, bF), bCAndH);\n          }\n        }, _a.className = \"CustomInit\", _a)();\n      } else {\n        biasInitializer = this.biasInitializer;\n      }\n      this.bias = this.addWeight(\"bias\", [this.units * 4], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    } else {\n      this.bias = null;\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const training = kwargs[\"training\"] == null ? false : kwargs[\"training\"];\n      inputs = inputs;\n      if (inputs.length !== 3) {\n        throw new ValueError(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);\n      }\n      let hTMinus1 = inputs[1];\n      const cTMinus1 = inputs[2];\n      inputs = inputs[0];\n      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {\n        this.dropoutMask = generateDropoutMask({\n          ones: () => onesLike(inputs),\n          rate: this.dropout,\n          training,\n          count: 4,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {\n        this.recurrentDropoutMask = generateDropoutMask({\n          ones: () => onesLike(hTMinus1),\n          rate: this.recurrentDropout,\n          training,\n          count: 4,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      const dpMask = this.dropoutMask;\n      const recDpMask = this.recurrentDropoutMask;\n      let i;\n      let f;\n      let c;\n      let o;\n      if (0 < this.dropout && this.dropout < 1) {\n        inputs = mul(inputs, dpMask[0]);\n      }\n      let z = dot2(inputs, this.kernel.read());\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {\n        hTMinus1 = mul(hTMinus1, recDpMask[0]);\n      }\n      z = add2(z, dot2(hTMinus1, this.recurrentKernel.read()));\n      if (this.useBias) {\n        z = biasAdd(z, this.bias.read());\n      }\n      const [z0, z1, z2, z3] = split(z, 4, z.rank - 1);\n      i = this.recurrentActivation.apply(z0);\n      f = this.recurrentActivation.apply(z1);\n      c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(z2)));\n      o = this.recurrentActivation.apply(z3);\n      const h = mul(o, this.activation.apply(c));\n      return [h, h, c];\n    });\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = {\n      units: this.units,\n      activation: serializeActivation(this.activation),\n      recurrentActivation: serializeActivation(this.recurrentActivation),\n      useBias: this.useBias,\n      kernelInitializer: serializeInitializer(this.kernelInitializer),\n      recurrentInitializer: serializeInitializer(this.recurrentInitializer),\n      biasInitializer: serializeInitializer(this.biasInitializer),\n      unitForgetBias: this.unitForgetBias,\n      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),\n      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),\n      biasRegularizer: serializeRegularizer(this.biasRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      kernelConstraint: serializeConstraint(this.kernelConstraint),\n      recurrentConstraint: serializeConstraint(this.recurrentConstraint),\n      biasConstraint: serializeConstraint(this.biasConstraint),\n      dropout: this.dropout,\n      recurrentDropout: this.recurrentDropout,\n      implementation: this.implementation\n    };\n    return Object.assign(Object.assign({}, baseConfig), config);\n  }\n};\nLSTMCell.className = \"LSTMCell\";\nserialization_exports.registerClass(LSTMCell);\nvar LSTM = class extends RNN {\n  constructor(args) {\n    if (args.implementation === 0) {\n      console.warn(\"`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.\");\n    }\n    args.cell = new LSTMCell(args);\n    super(args);\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (this.cell.dropoutMask != null) {\n        dispose(this.cell.dropoutMask);\n        this.cell.dropoutMask = null;\n      }\n      if (this.cell.recurrentDropoutMask != null) {\n        dispose(this.cell.recurrentDropoutMask);\n        this.cell.recurrentDropoutMask = null;\n      }\n      const mask = kwargs == null ? null : kwargs[\"mask\"];\n      const training = kwargs == null ? null : kwargs[\"training\"];\n      const initialState = kwargs == null ? null : kwargs[\"initialState\"];\n      return super.call(inputs, { mask, training, initialState });\n    });\n  }\n  static fromConfig(cls, config) {\n    if (config[\"implmentation\"] === 0) {\n      config[\"implementation\"] = 1;\n    }\n    return new cls(config);\n  }\n};\nLSTM.className = \"LSTM\";\nserialization_exports.registerClass(LSTM);\nvar StackedRNNCells = class extends RNNCell {\n  constructor(args) {\n    super(args);\n    this.cells = args.cells;\n  }\n  get stateSize() {\n    const stateSize = [];\n    for (const cell of this.cells.slice().reverse()) {\n      if (Array.isArray(cell.stateSize)) {\n        stateSize.push(...cell.stateSize);\n      } else {\n        stateSize.push(cell.stateSize);\n      }\n    }\n    return stateSize;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = inputs;\n      let states = inputs.slice(1);\n      const nestedStates = [];\n      for (const cell of this.cells.slice().reverse()) {\n        if (Array.isArray(cell.stateSize)) {\n          nestedStates.push(states.splice(0, cell.stateSize.length));\n        } else {\n          nestedStates.push(states.splice(0, 1));\n        }\n      }\n      nestedStates.reverse();\n      const newNestedStates = [];\n      let callInputs;\n      for (let i = 0; i < this.cells.length; ++i) {\n        const cell = this.cells[i];\n        states = nestedStates[i];\n        if (i === 0) {\n          callInputs = [inputs[0]].concat(states);\n        } else {\n          callInputs = [callInputs[0]].concat(states);\n        }\n        callInputs = cell.call(callInputs, kwargs);\n        newNestedStates.push(callInputs.slice(1));\n      }\n      states = [];\n      for (const cellStates of newNestedStates.slice().reverse()) {\n        states.push(...cellStates);\n      }\n      return [callInputs[0]].concat(states);\n    });\n  }\n  build(inputShape) {\n    if (isArrayOfShapes(inputShape)) {\n      inputShape = inputShape[0];\n    }\n    inputShape = inputShape;\n    let outputDim;\n    this.cells.forEach((cell, i) => {\n      nameScope(`RNNCell_${i}`, () => {\n        cell.build(inputShape);\n        if (Array.isArray(cell.stateSize)) {\n          outputDim = cell.stateSize[0];\n        } else {\n          outputDim = cell.stateSize;\n        }\n        inputShape = [inputShape[0], outputDim];\n      });\n    });\n    this.built = true;\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const getCellConfig = (cell) => {\n      return {\n        \"className\": cell.getClassName(),\n        \"config\": cell.getConfig()\n      };\n    };\n    const cellConfigs = this.cells.map(getCellConfig);\n    const config = { \"cells\": cellConfigs };\n    return Object.assign(Object.assign({}, baseConfig), config);\n  }\n  static fromConfig(cls, config, customObjects = {}) {\n    const cells = [];\n    for (const cellConfig of config[\"cells\"]) {\n      cells.push(deserialize(cellConfig, customObjects));\n    }\n    return new cls({ cells });\n  }\n  get trainableWeights() {\n    if (!this.trainable) {\n      return [];\n    }\n    const weights = [];\n    for (const cell of this.cells) {\n      weights.push(...cell.trainableWeights);\n    }\n    return weights;\n  }\n  get nonTrainableWeights() {\n    const weights = [];\n    for (const cell of this.cells) {\n      weights.push(...cell.nonTrainableWeights);\n    }\n    if (!this.trainable) {\n      const trainableWeights = [];\n      for (const cell of this.cells) {\n        trainableWeights.push(...cell.trainableWeights);\n      }\n      return trainableWeights.concat(weights);\n    }\n    return weights;\n  }\n  getWeights() {\n    const weights = [];\n    for (const cell of this.cells) {\n      weights.push(...cell.weights);\n    }\n    return batchGetValue(weights);\n  }\n  setWeights(weights) {\n    const tuples = [];\n    for (const cell of this.cells) {\n      const numParams = cell.weights.length;\n      const inputWeights = weights.splice(numParams);\n      for (let i = 0; i < cell.weights.length; ++i) {\n        tuples.push([cell.weights[i], inputWeights[i]]);\n      }\n    }\n    batchSetValue(tuples);\n  }\n};\nStackedRNNCells.className = \"StackedRNNCells\";\nserialization_exports.registerClass(StackedRNNCells);\nfunction generateDropoutMask(args) {\n  const { ones: ones4, rate, training = false, count: count2 = 1, dropoutFunc } = args;\n  const droppedInputs = () => dropoutFunc != null ? dropoutFunc(ones4(), rate) : dropout2(ones4(), rate);\n  const createMask = () => inTrainPhase(droppedInputs, ones4, training);\n  if (!count2 || count2 <= 1) {\n    return keep(createMask().clone());\n  }\n  const masks = Array(count2).fill(void 0).map(createMask);\n  return masks.map((m) => keep(m.clone()));\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional_recurrent.js\nvar __rest = function(s, e) {\n  var t = {};\n  for (var p2 in s)\n    if (Object.prototype.hasOwnProperty.call(s, p2) && e.indexOf(p2) < 0)\n      t[p2] = s[p2];\n  if (s != null && typeof Object.getOwnPropertySymbols === \"function\")\n    for (var i = 0, p2 = Object.getOwnPropertySymbols(s); i < p2.length; i++) {\n      if (e.indexOf(p2[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p2[i]))\n        t[p2[i]] = s[p2[i]];\n    }\n  return t;\n};\nvar ConvRNN2D = class extends RNN {\n  constructor(args) {\n    if (args.unroll) {\n      throw new NotImplementedError(\"Unrolling is not possible with convolutional RNNs.\");\n    }\n    if (Array.isArray(args.cell)) {\n      throw new NotImplementedError(\"It is not possible at the moment to stack convolutional cells.\");\n    }\n    super(args);\n    this.inputSpec = [new InputSpec({ ndim: 5 })];\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (this.cell.dropoutMask != null) {\n        dispose(this.cell.dropoutMask);\n        this.cell.dropoutMask = null;\n      }\n      if (this.cell.recurrentDropoutMask != null) {\n        dispose(this.cell.recurrentDropoutMask);\n        this.cell.recurrentDropoutMask = null;\n      }\n      if (kwargs && kwargs[\"constants\"]) {\n        throw new ValueError(\"ConvRNN2D cell does not support constants\");\n      }\n      const mask = kwargs == null ? null : kwargs[\"mask\"];\n      const training = kwargs == null ? null : kwargs[\"training\"];\n      const initialState = kwargs == null ? null : kwargs[\"initialState\"];\n      return super.call(inputs, { mask, training, initialState });\n    });\n  }\n  computeOutputShape(inputShape) {\n    let outShape = this.computeSingleOutputShape(inputShape);\n    if (!this.returnSequences) {\n      outShape = [outShape[0], ...outShape.slice(2)];\n    }\n    if (this.returnState) {\n      outShape = [outShape, ...Array(2).fill([inputShape[0], ...outShape.slice(-3)])];\n    }\n    return outShape;\n  }\n  getInitialState(inputs) {\n    return tidy(() => {\n      const { stateSize } = this.cell;\n      const inputShape = inputs.shape;\n      const outputShape = this.computeSingleOutputShape(inputShape);\n      const stateShape = [outputShape[0], ...outputShape.slice(2)];\n      const initialState = zeros(stateShape);\n      if (Array.isArray(stateSize)) {\n        return Array(stateSize.length).fill(initialState);\n      }\n      return [initialState];\n    });\n  }\n  resetStates(states, training = false) {\n    tidy(() => {\n      if (!this.stateful) {\n        throw new AttributeError(\"Cannot call resetStates() on an RNN Layer that is not stateful.\");\n      }\n      const inputShape = this.inputSpec[0].shape;\n      const outputShape = this.computeSingleOutputShape(inputShape);\n      const stateShape = [outputShape[0], ...outputShape.slice(2)];\n      const batchSize = inputShape[0];\n      if (batchSize == null) {\n        throw new ValueError(\"If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \\n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.\");\n      }\n      if (this.getStates() == null) {\n        if (Array.isArray(this.cell.stateSize)) {\n          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));\n        } else {\n          this.states_ = [zeros(stateShape)];\n        }\n      } else if (states == null) {\n        dispose(this.states_);\n        if (this.keptStates != null) {\n          dispose(this.keptStates);\n          this.keptStates = [];\n        }\n        if (Array.isArray(this.cell.stateSize)) {\n          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));\n        } else {\n          this.states_[0] = zeros(stateShape);\n        }\n      } else {\n        if (!Array.isArray(states)) {\n          states = [states];\n        }\n        if (states.length !== this.states_.length) {\n          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);\n        }\n        if (training) {\n          this.keptStates.push(this.states_.slice());\n        } else {\n          dispose(this.states_);\n        }\n        for (let index = 0; index < this.states_.length; ++index) {\n          const value = states[index];\n          const expectedShape = stateShape;\n          if (!util_exports.arraysEqual(value.shape, expectedShape)) {\n            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);\n          }\n          this.states_[index] = value;\n        }\n      }\n      this.states_ = this.states_.map((state) => keep(state.clone()));\n    });\n  }\n  computeSingleOutputShape(inputShape) {\n    const { dataFormat, filters, kernelSize, padding, strides, dilationRate } = this.cell;\n    const isChannelsFirst = dataFormat === \"channelsFirst\";\n    const h = inputShape[isChannelsFirst ? 3 : 2];\n    const w = inputShape[isChannelsFirst ? 4 : 3];\n    const hOut = convOutputLength(h, kernelSize[0], padding, strides[0], dilationRate[0]);\n    const wOut = convOutputLength(w, kernelSize[1], padding, strides[1], dilationRate[1]);\n    const outShape = [\n      ...inputShape.slice(0, 2),\n      ...isChannelsFirst ? [filters, hOut, wOut] : [hOut, wOut, filters]\n    ];\n    return outShape;\n  }\n};\nConvRNN2D.className = \"ConvRNN2D\";\nvar ConvLSTM2DCell = class extends LSTMCell {\n  constructor(args) {\n    const { filters, kernelSize, strides, padding, dataFormat, dilationRate } = args;\n    super(Object.assign(Object.assign({}, args), { units: filters }));\n    this.filters = filters;\n    assertPositiveInteger(this.filters, \"filters\");\n    this.kernelSize = normalizeArray(kernelSize, 2, \"kernelSize\");\n    this.kernelSize.forEach((size) => assertPositiveInteger(size, \"kernelSize\"));\n    this.strides = normalizeArray(strides || 1, 2, \"strides\");\n    this.strides.forEach((stride) => assertPositiveInteger(stride, \"strides\"));\n    this.padding = padding || \"valid\";\n    checkPaddingMode(this.padding);\n    this.dataFormat = dataFormat || \"channelsLast\";\n    checkDataFormat(this.dataFormat);\n    this.dilationRate = normalizeArray(dilationRate || 1, 2, \"dilationRate\");\n    this.dilationRate.forEach((rate) => assertPositiveInteger(rate, \"dilationRate\"));\n  }\n  build(inputShape) {\n    var _a;\n    inputShape = getExactlyOneShape(inputShape);\n    const channelAxis = this.dataFormat === \"channelsFirst\" ? 1 : inputShape.length - 1;\n    if (inputShape[channelAxis] == null) {\n      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);\n    }\n    const inputDim = inputShape[channelAxis];\n    const numOfKernels = 4;\n    const kernelShape = this.kernelSize.concat([inputDim, this.filters * numOfKernels]);\n    this.kernel = this.addWeight(\"kernel\", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n    const recurrentKernelShape = this.kernelSize.concat([this.filters, this.filters * numOfKernels]);\n    this.recurrentKernel = this.addWeight(\"recurrent_kernel\", recurrentKernelShape, null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);\n    if (this.useBias) {\n      let biasInitializer;\n      if (this.unitForgetBias) {\n        const init2 = this.biasInitializer;\n        const filters = this.filters;\n        biasInitializer = new (_a = class CustomInit extends Initializer {\n          apply(shape, dtype) {\n            const biasI = init2.apply([filters]);\n            const biasF = ones2([filters]);\n            const biasCAndO = init2.apply([filters * 2]);\n            return concatenate([biasI, biasF, biasCAndO]);\n          }\n        }, _a.className = \"CustomInit\", _a)();\n      } else {\n        biasInitializer = this.biasInitializer;\n      }\n      this.bias = this.addWeight(\"bias\", [this.filters * numOfKernels], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (inputs.length !== 3) {\n        throw new ValueError(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);\n      }\n      const training = kwargs[\"training\"] || false;\n      const x = inputs[0];\n      const hTMinus1 = inputs[1];\n      const cTMinus1 = inputs[2];\n      const numOfKernels = 4;\n      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {\n        this.dropoutMask = generateDropoutMask({\n          ones: () => onesLike(x),\n          rate: this.dropout,\n          training,\n          count: numOfKernels,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      const dropoutMask = this.dropoutMask;\n      const applyDropout = (x2, mask, index) => {\n        if (!mask || !mask[index]) {\n          return x2;\n        }\n        return mul(mask[index], x2);\n      };\n      let xI = applyDropout(x, dropoutMask, 0);\n      let xF = applyDropout(x, dropoutMask, 1);\n      let xC = applyDropout(x, dropoutMask, 2);\n      let xO = applyDropout(x, dropoutMask, 3);\n      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {\n        this.recurrentDropoutMask = generateDropoutMask({\n          ones: () => onesLike(hTMinus1),\n          rate: this.recurrentDropout,\n          training,\n          count: numOfKernels,\n          dropoutFunc: this.dropoutFunc\n        });\n      }\n      const recDropoutMask = this.recurrentDropoutMask;\n      let hI = applyDropout(hTMinus1, recDropoutMask, 0);\n      let hF = applyDropout(hTMinus1, recDropoutMask, 1);\n      let hC = applyDropout(hTMinus1, recDropoutMask, 2);\n      let hO = applyDropout(hTMinus1, recDropoutMask, 3);\n      const kernelChannelAxis = 3;\n      const [kernelI, kernelF, kernelC, kernelO] = split(this.kernel.read(), numOfKernels, kernelChannelAxis);\n      const [biasI, biasF, biasC, biasO] = this.useBias ? split(this.bias.read(), numOfKernels) : [null, null, null, null];\n      xI = this.inputConv(xI, kernelI, biasI, this.padding);\n      xF = this.inputConv(xF, kernelF, biasF, this.padding);\n      xC = this.inputConv(xC, kernelC, biasC, this.padding);\n      xO = this.inputConv(xO, kernelO, biasO, this.padding);\n      const [recKernelI, recKernelF, recKernelC, recKernelO] = split(this.recurrentKernel.read(), numOfKernels, kernelChannelAxis);\n      hI = this.recurrentConv(hI, recKernelI);\n      hF = this.recurrentConv(hF, recKernelF);\n      hC = this.recurrentConv(hC, recKernelC);\n      hO = this.recurrentConv(hO, recKernelO);\n      const i = this.recurrentActivation.apply(add2(xI, hI));\n      const f = this.recurrentActivation.apply(add2(xF, hF));\n      const c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(add2(xC, hC))));\n      const h = mul(this.recurrentActivation.apply(add2(xO, hO)), this.activation.apply(c));\n      return [h, h, c];\n    });\n  }\n  getConfig() {\n    const _a = super.getConfig(), { \"units\": _ } = _a, baseConfig = __rest(_a, [\"units\"]);\n    const config = {\n      filters: this.filters,\n      kernelSize: this.kernelSize,\n      padding: this.padding,\n      dataFormat: this.dataFormat,\n      dilationRate: this.dilationRate,\n      strides: this.strides\n    };\n    return Object.assign(Object.assign({}, baseConfig), config);\n  }\n  inputConv(x, w, b, padding) {\n    const out = conv2d(x, w, this.strides, padding || \"valid\", this.dataFormat === \"channelsFirst\" ? \"NCHW\" : \"NHWC\", this.dilationRate);\n    if (b) {\n      return biasAdd(out, b, this.dataFormat);\n    }\n    return out;\n  }\n  recurrentConv(x, w) {\n    const strides = 1;\n    return conv2d(x, w, strides, \"same\", this.dataFormat === \"channelsFirst\" ? \"NCHW\" : \"NHWC\");\n  }\n};\nConvLSTM2DCell.className = \"ConvLSTM2DCell\";\nserialization_exports.registerClass(ConvLSTM2DCell);\nvar ConvLSTM2D = class extends ConvRNN2D {\n  constructor(args) {\n    const cell = new ConvLSTM2DCell(args);\n    super(Object.assign(Object.assign({}, args), { cell }));\n  }\n  static fromConfig(cls, config) {\n    return new cls(config);\n  }\n};\nConvLSTM2D.className = \"ConvLSTM2D\";\nserialization_exports.registerClass(ConvLSTM2D);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/core.js\nvar Dropout = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.rate = Math.max(Math.min(args.rate, 1), 0);\n    this.noiseShape = args.noiseShape;\n    this.seed = args.seed;\n    this.supportsMasking = true;\n  }\n  getNoiseShape(input2) {\n    if (this.noiseShape == null) {\n      return this.noiseShape;\n    }\n    const inputShape = input2.shape;\n    const noiseShape = [];\n    for (let i = 0; i < this.noiseShape.length; ++i) {\n      noiseShape.push(this.noiseShape[i] == null ? inputShape[i] : this.noiseShape[i]);\n    }\n    return noiseShape;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      if (0 < this.rate && this.rate < 1) {\n        const training = kwargs[\"training\"] == null ? false : kwargs[\"training\"];\n        const noiseShape = this.getNoiseShape(input2);\n        const output = inTrainPhase(() => dropout2(input2, this.rate, noiseShape, this.seed), () => input2, training);\n        return output;\n      }\n      return inputs;\n    });\n  }\n  getConfig() {\n    const config = {\n      rate: this.rate,\n      noiseShape: this.noiseShape,\n      seed: this.seed\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  dispose() {\n    return super.dispose();\n  }\n};\nDropout.className = \"Dropout\";\nserialization_exports.registerClass(Dropout);\nvar SpatialDropout1D = class extends Dropout {\n  constructor(args) {\n    super(args);\n    this.inputSpec = [{ ndim: 3 }];\n  }\n  getNoiseShape(input2) {\n    const inputShape = input2.shape;\n    return [inputShape[0], 1, inputShape[2]];\n  }\n};\nSpatialDropout1D.className = \"SpatialDropout1D\";\nserialization_exports.registerClass(SpatialDropout1D);\nvar Dense = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.activation = null;\n    this.useBias = true;\n    this.kernel = null;\n    this.bias = null;\n    this.DEFAULT_KERNEL_INITIALIZER = \"glorotNormal\";\n    this.DEFAULT_BIAS_INITIALIZER = \"zeros\";\n    if (args.batchInputShape == null && args.inputShape == null && args.inputDim != null) {\n      let batchSize = null;\n      if (args.batchSize != null) {\n        batchSize = args.batchSize;\n      }\n      this.batchInputShape = [batchSize, args.inputDim];\n    }\n    this.units = args.units;\n    assertPositiveInteger(this.units, \"units\");\n    this.activation = getActivation(args.activation);\n    if (args.useBias != null) {\n      this.useBias = args.useBias;\n    }\n    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);\n    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);\n    this.kernelConstraint = getConstraint(args.kernelConstraint);\n    this.biasConstraint = getConstraint(args.biasConstraint);\n    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);\n    this.biasRegularizer = getRegularizer(args.biasRegularizer);\n    this.activityRegularizer = getRegularizer(args.activityRegularizer);\n    this.supportsMasking = true;\n    this.inputSpec = [{ minNDim: 2 }];\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const inputLastDim = inputShape[inputShape.length - 1];\n    if (this.kernel == null) {\n      this.kernel = this.addWeight(\"kernel\", [inputLastDim, this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);\n      if (this.useBias) {\n        this.bias = this.addWeight(\"bias\", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);\n      }\n    }\n    this.inputSpec = [{ minNDim: 2, axes: { [-1]: inputLastDim } }];\n    this.built = true;\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const outputShape = inputShape.slice();\n    outputShape[outputShape.length - 1] = this.units;\n    return outputShape;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());\n      let output;\n      if (fusedActivationName != null) {\n        output = dot2(input2, this.kernel.read(), fusedActivationName, this.bias ? this.bias.read() : null);\n      } else {\n        output = dot2(input2, this.kernel.read());\n        if (this.bias != null) {\n          output = biasAdd(output, this.bias.read());\n        }\n        if (this.activation != null) {\n          output = this.activation.apply(output);\n        }\n      }\n      return output;\n    });\n  }\n  getConfig() {\n    const config = {\n      units: this.units,\n      activation: serializeActivation(this.activation),\n      useBias: this.useBias,\n      kernelInitializer: serializeInitializer(this.kernelInitializer),\n      biasInitializer: serializeInitializer(this.biasInitializer),\n      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),\n      biasRegularizer: serializeRegularizer(this.biasRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      kernelConstraint: serializeConstraint(this.kernelConstraint),\n      biasConstraint: serializeConstraint(this.biasConstraint)\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nDense.className = \"Dense\";\nserialization_exports.registerClass(Dense);\nvar Flatten = class extends Layer {\n  constructor(args) {\n    args = args || {};\n    super(args);\n    this.inputSpec = [{ minNDim: 3 }];\n    this.dataFormat = args.dataFormat;\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    for (const dim of inputShape.slice(1)) {\n      if (dim == null) {\n        throw new ValueError(`The shape of the input to \"Flatten\" is not fully defined (got ${inputShape.slice(1)}). Make sure to pass a complete \"input_shape\" or \"batch_input_shape\" argument to the first layer in your model.`);\n      }\n    }\n    return [inputShape[0], arrayProd(inputShape, 1)];\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      let input2 = getExactlyOneTensor(inputs);\n      if (this.dataFormat === \"channelsFirst\" && input2.rank > 1) {\n        const permutation = [0];\n        for (let i = 2; i < input2.rank; ++i) {\n          permutation.push(i);\n        }\n        permutation.push(1);\n        input2 = transpose(input2, permutation);\n      }\n      return batchFlatten(input2);\n    });\n  }\n  getConfig() {\n    const config = {};\n    if (this.dataFormat != null) {\n      config[\"dataFormat\"] = this.dataFormat;\n    }\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nFlatten.className = \"Flatten\";\nserialization_exports.registerClass(Flatten);\nvar Activation2 = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.supportsMasking = true;\n    this.activation = getActivation(args.activation);\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      return this.activation.apply(input2);\n    });\n  }\n  getConfig() {\n    const config = { activation: serializeActivation(this.activation) };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nActivation2.className = \"Activation\";\nserialization_exports.registerClass(Activation2);\nvar RepeatVector = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.n = args.n;\n    this.inputSpec = [{ ndim: 2 }];\n  }\n  computeOutputShape(inputShape) {\n    return [inputShape[0], this.n, inputShape[1]];\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      return repeat(inputs, this.n);\n    });\n  }\n  getConfig() {\n    const config = {\n      n: this.n\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nRepeatVector.className = \"RepeatVector\";\nserialization_exports.registerClass(RepeatVector);\nvar Reshape2 = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.targetShape = args.targetShape;\n    for (let i = 0; i < this.targetShape.length; ++i) {\n      if (this.isUnknown(this.targetShape[i])) {\n        this.targetShape[i] = null;\n      }\n    }\n  }\n  isUnknown(dim) {\n    return dim < 0 || dim == null;\n  }\n  fixUnknownDimension(inputShape, outputShape) {\n    const errorMsg = \"Total size of new array must be unchanged.\";\n    const finalShape = outputShape.slice();\n    let known = 1;\n    let unknown = null;\n    for (let i = 0; i < finalShape.length; ++i) {\n      const dim = finalShape[i];\n      if (this.isUnknown(dim)) {\n        if (unknown === null) {\n          unknown = i;\n        } else {\n          throw new ValueError(\"Can only specifiy one unknown dimension.\");\n        }\n      } else {\n        known *= dim;\n      }\n    }\n    const originalSize = arrayProd(inputShape);\n    if (unknown !== null) {\n      if (known === 0 || originalSize % known !== 0) {\n        throw new ValueError(errorMsg);\n      }\n      finalShape[unknown] = originalSize / known;\n    } else if (originalSize !== known) {\n      throw new ValueError(errorMsg);\n    }\n    return finalShape;\n  }\n  computeOutputShape(inputShape) {\n    let anyUnknownDims = false;\n    for (let i = 0; i < inputShape.length; ++i) {\n      if (this.isUnknown(inputShape[i])) {\n        anyUnknownDims = true;\n        break;\n      }\n    }\n    if (anyUnknownDims) {\n      return inputShape.slice(0, 1).concat(this.targetShape);\n    } else {\n      return inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      const inputShape = input2.shape;\n      const outputShape = inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));\n      return reshape(input2, outputShape);\n    });\n  }\n  getConfig() {\n    const config = {\n      targetShape: this.targetShape\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nReshape2.className = \"Reshape\";\nserialization_exports.registerClass(Reshape2);\nvar Permute = class extends Layer {\n  constructor(args) {\n    super(args);\n    if (args.dims == null) {\n      throw new Error(\"Required configuration field `dims` is missing during Permute constructor call.\");\n    }\n    if (!Array.isArray(args.dims)) {\n      throw new Error(`Permute constructor requires \\`dims\\` to be an Array, but received ${args.dims} instead.`);\n    }\n    const expectedSortedIndices = range2(1, args.dims.length + 1);\n    if (!util_exports.arraysEqual(args.dims.slice().sort(), expectedSortedIndices)) {\n      throw new Error(\"Invalid permutation `dims`: \" + JSON.stringify(args.dims) + \" `dims` must contain consecutive integers starting from 1.\");\n    }\n    this.dims = args.dims;\n    this.dimsIncludingBatch = [0].concat(this.dims);\n    this.inputSpec = [new InputSpec({ ndim: this.dims.length + 1 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const outputShape = inputShape.slice();\n    this.dims.forEach((dim, i) => {\n      outputShape[i + 1] = inputShape[dim];\n    });\n    return outputShape;\n  }\n  call(inputs, kwargs) {\n    return transpose(getExactlyOneTensor(inputs), this.dimsIncludingBatch);\n  }\n  getConfig() {\n    const config = {\n      dims: this.dims\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nPermute.className = \"Permute\";\nserialization_exports.registerClass(Permute);\nvar Masking = class extends Layer {\n  constructor(args) {\n    super(args == null ? {} : args);\n    this.supportsMasking = true;\n    if (args != null) {\n      this.maskValue = args.maskValue == null ? 0 : args.maskValue;\n    } else {\n      this.maskValue = 0;\n    }\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = { maskValue: this.maskValue };\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  computeMask(inputs, mask) {\n    const input2 = getExactlyOneTensor(inputs);\n    const axis = -1;\n    return any(notEqual(input2, this.maskValue), axis);\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      const axis = -1;\n      const keepDims = true;\n      const booleanMask = any(notEqual(input2, this.maskValue), axis, keepDims);\n      const output = mul(input2, cast(booleanMask, input2.dtype));\n      return output;\n    });\n  }\n};\nMasking.className = \"Masking\";\nserialization_exports.registerClass(Masking);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/embeddings.js\nvar Embedding = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.embeddings = null;\n    this.DEFAULT_EMBEDDINGS_INITIALIZER = \"randomUniform\";\n    if (args.batchInputShape == null && args.inputShape == null) {\n      let batchSize = null;\n      if (args.batchSize != null) {\n        batchSize = args.batchSize;\n      }\n      if (args.inputLength == null) {\n        this.batchInputShape = [batchSize, null];\n      } else {\n        this.batchInputShape = [batchSize].concat(toList(args.inputLength));\n      }\n    }\n    this.inputDim = args.inputDim;\n    assertPositiveInteger(this.inputDim, \"inputDim\");\n    this.outputDim = args.outputDim;\n    assertPositiveInteger(this.outputDim, \"outputDim\");\n    this.embeddingsInitializer = getInitializer(args.embeddingsInitializer || this.DEFAULT_EMBEDDINGS_INITIALIZER);\n    this.embeddingsRegularizer = getRegularizer(args.embeddingsRegularizer);\n    this.activityRegularizer = getRegularizer(args.activityRegularizer);\n    this.embeddingsConstraint = getConstraint(args.embeddingsConstraint);\n    this.maskZero = args.maskZero;\n    this.supportsMasking = args.maskZero;\n    this.inputLength = args.inputLength;\n  }\n  build(inputShape) {\n    this.embeddings = this.addWeight(\"embeddings\", [this.inputDim, this.outputDim], this.dtype, this.embeddingsInitializer, this.embeddingsRegularizer, true, this.embeddingsConstraint);\n    this.built = true;\n  }\n  warnOnIncompatibleInputShape(inputShape) {\n  }\n  computeMask(inputs, mask) {\n    return tidy(() => {\n      if (!this.maskZero) {\n        return null;\n      } else {\n        inputs = getExactlyOneTensor(inputs);\n        return notEqual(inputs, zerosLike(inputs));\n      }\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (this.inputLength == null) {\n      return [...inputShape, this.outputDim];\n    }\n    const inLens = toList(this.inputLength);\n    if (inLens.length !== inputShape.length - 1) {\n      throw new ValueError(`\"inputLength\" is ${this.inputLength}, but received input shape has shape ${inputShape}`);\n    } else {\n      let i = 0;\n      for (let k = 0; k < inLens.length; ++k) {\n        const s1 = inLens[k];\n        const s2 = inputShape[k + 1];\n        if (s1 != null && s2 != null && s1 !== s2) {\n          throw new ValueError(`\"inputLength\" is ${this.inputLength}, but received input shape has shape ${inputShape}`);\n        } else if (s1 == null) {\n          inLens[i] = s2;\n        }\n        i++;\n      }\n    }\n    return [inputShape[0], ...inLens, this.outputDim];\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      let input2 = getExactlyOneTensor(inputs);\n      if (input2.dtype !== \"int32\") {\n        input2 = cast2(input2, \"int32\");\n      }\n      const output = gather2(this.embeddings.read(), reshape(input2, [input2.size]));\n      return reshape(output, getExactlyOneShape(this.computeOutputShape(input2.shape)));\n    });\n  }\n  getConfig() {\n    const config = {\n      inputDim: this.inputDim,\n      outputDim: this.outputDim,\n      embeddingsInitializer: serializeInitializer(this.embeddingsInitializer),\n      embeddingsRegularizer: serializeRegularizer(this.embeddingsRegularizer),\n      activityRegularizer: serializeRegularizer(this.activityRegularizer),\n      embeddingsConstraint: serializeConstraint(this.embeddingsConstraint),\n      maskZero: this.maskZero,\n      inputLength: this.inputLength\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nEmbedding.className = \"Embedding\";\nserialization_exports.registerClass(Embedding);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/merge.js\nvar Merge = class extends Layer {\n  constructor(args) {\n    super(args || {});\n    this.supportsMasking = true;\n  }\n  mergeFunction(inputs) {\n    throw new NotImplementedError();\n  }\n  computeElementwiseOpOutputShape(shape1, shape2) {\n    if (shape1 == null || shape2 == null) {\n      return null;\n    } else if (shape1.length < shape2.length) {\n      return this.computeElementwiseOpOutputShape(shape2, shape1);\n    } else if (shape2.length === 0) {\n      return shape1;\n    }\n    const outputShape = shape1.slice(0, shape1.length - shape2.length);\n    for (let k = 0; k < shape2.length; ++k) {\n      const i = shape1[shape1.length - shape2.length + k];\n      const j = shape2[k];\n      if (i == null || j == null || i < 0 || j < 0) {\n        outputShape.push(null);\n      } else if (i === 1) {\n        outputShape.push(j);\n      } else if (j === 1) {\n        outputShape.push(i);\n      } else {\n        if (i !== j) {\n          throw new ValueError(\"Operands could not be broadcast together with shapes \" + JSON.stringify(shape1) + \" \" + JSON.stringify(shape2));\n        }\n        outputShape.push(i);\n      }\n    }\n    return outputShape;\n  }\n  build(inputShape) {\n    if (Array.isArray(inputShape) && !Array.isArray(inputShape[0])) {\n      inputShape = [getExactlyOneShape(inputShape)];\n    }\n    inputShape = inputShape;\n    if (inputShape.length < 2) {\n      throw new ValueError(`A merge layer should be called on an Array of at least 2 inputs. Got ${inputShape.length} input(s).`);\n    }\n    let batchSizes = [];\n    for (const shape of inputShape) {\n      if (shape != null && shape[0] !== null) {\n        batchSizes.push(shape[0]);\n      }\n    }\n    batchSizes = unique2(batchSizes);\n    if (batchSizes.length > 1) {\n      throw new ValueError(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(inputShape)}.`);\n    }\n    let outputShape = inputShape[0] == null ? null : inputShape[0].slice(1);\n    for (let i = 1; i < inputShape.length; ++i) {\n      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);\n      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);\n    }\n    const allRanks = inputShape.map((shape) => shape.length);\n    if (inputShape.indexOf(null) === -1 && unique2(allRanks).length === 1) {\n      this.reshapeRequired = false;\n    } else {\n      this.reshapeRequired = true;\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = inputs;\n      if (this.reshapeRequired) {\n        const reshapedInputs = [];\n        const inputDims = inputs.map((input2) => input2.rank);\n        if (inputDims.indexOf(null) === -1) {\n          const maxNDim = max2(inputDims);\n          for (let x of inputs) {\n            const xNDim = x.rank;\n            for (let k = 0; k < maxNDim - xNDim; ++k) {\n              x = expandDims2(x, 1);\n            }\n            reshapedInputs.push(x);\n          }\n          return this.mergeFunction(reshapedInputs);\n        } else {\n          let transposed = false;\n          for (const x of inputs) {\n            const xNDim = x.rank;\n            if (xNDim == null) {\n              const xShape = x.shape;\n              const batchSize = xShape[0];\n              const newShape = xShape.slice(1).concat([batchSize]);\n              let xTransposed = reshape(x, [batchSize].concat(arrayProd(xShape.slice(1))));\n              xTransposed = transpose(xTransposed, [1, 0]);\n              xTransposed = reshape(xTransposed, newShape);\n              reshapedInputs.push(xTransposed);\n              transposed = true;\n            } else if (xNDim > 1) {\n              const dims = range2(1, xNDim).concat([0]);\n              reshapedInputs.push(transpose(x, dims));\n              transposed = true;\n            } else {\n              reshapedInputs.push(x);\n            }\n          }\n          let y = this.mergeFunction(reshapedInputs);\n          const yNDim = y.rank;\n          if (transposed) {\n            if (yNDim == null) {\n              const yShape = y.shape;\n              const yNDim2 = yShape.length;\n              const batchSize = yShape[yNDim2 - 1];\n              const newShape = [batchSize].concat(yShape.slice(0, yShape.length - 1));\n              y = reshape(transpose(reshape(y, [-1, batchSize]), [1, 0]), newShape);\n            } else if (yNDim > 1) {\n              const dims = [yNDim - 1].concat(range2(0, yNDim - 1));\n              y = transpose(y, dims);\n            }\n          }\n          return y;\n        }\n      } else {\n        return this.mergeFunction(inputs);\n      }\n    });\n  }\n  computeOutputShape(inputShape) {\n    inputShape = inputShape;\n    let outputShape;\n    if (inputShape[0] == null) {\n      outputShape = null;\n    } else {\n      outputShape = inputShape[0].slice(1);\n    }\n    for (let i = 1; i < inputShape.length; ++i) {\n      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);\n      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);\n    }\n    let batchSizes = [];\n    for (const shape of inputShape) {\n      if (shape != null && shape[0] !== null) {\n        batchSizes.push(shape[0]);\n      }\n    }\n    batchSizes = unique2(batchSizes);\n    if (batchSizes.length === 1) {\n      outputShape = batchSizes.concat(outputShape);\n    } else {\n      outputShape = [null].concat(outputShape);\n    }\n    return outputShape;\n  }\n  computeMask(inputs, mask) {\n    return tidy(() => {\n      if (mask == null) {\n        return null;\n      }\n      if (!Array.isArray(mask)) {\n        throw new ValueError(\"`mask` should be an Array\");\n      }\n      if (!Array.isArray(inputs)) {\n        throw new ValueError(\"`inputs` should be an Array\");\n      }\n      if (mask.length !== inputs.length) {\n        throw new ValueError(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${inputs.length} vs ${mask.length})`);\n      }\n      if (mask.every((m) => m == null)) {\n        return null;\n      }\n      mask = mask.map((m) => m == null ? m : expandDims(m, 0));\n      let output = mask[0];\n      for (let i = 1; i < mask.length - 1; ++i) {\n        output = logicalAnd(output, mask[i]);\n      }\n      return output;\n    });\n  }\n};\nvar Add2 = class extends Merge {\n  constructor(args) {\n    super(args);\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      let output = inputs[0].clone();\n      for (let i = 1; i < inputs.length; ++i) {\n        output = add2(output, inputs[i]);\n      }\n      return output;\n    });\n  }\n};\nAdd2.className = \"Add\";\nserialization_exports.registerClass(Add2);\nvar Multiply2 = class extends Merge {\n  constructor(args) {\n    super(args);\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      let output = inputs[0].clone();\n      for (let i = 1; i < inputs.length; ++i) {\n        output = mul(output, inputs[i]);\n      }\n      return output;\n    });\n  }\n};\nMultiply2.className = \"Multiply\";\nserialization_exports.registerClass(Multiply2);\nvar Average = class extends Merge {\n  constructor(args) {\n    super(args);\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      let output = inputs[0].clone();\n      for (let i = 1; i < inputs.length; ++i) {\n        output = add2(output, inputs[i]);\n      }\n      return mul(1 / inputs.length, output);\n    });\n  }\n};\nAverage.className = \"Average\";\nserialization_exports.registerClass(Average);\nvar Maximum2 = class extends Merge {\n  constructor(args) {\n    super(args);\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      let output = inputs[0];\n      for (let i = 1; i < inputs.length; ++i) {\n        output = maximum(output, inputs[i]);\n      }\n      return output;\n    });\n  }\n};\nMaximum2.className = \"Maximum\";\nserialization_exports.registerClass(Maximum2);\nvar Minimum2 = class extends Merge {\n  constructor(args) {\n    super(args);\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      let output = inputs[0];\n      for (let i = 1; i < inputs.length; ++i) {\n        output = minimum(output, inputs[i]);\n      }\n      return output;\n    });\n  }\n};\nMinimum2.className = \"Minimum\";\nserialization_exports.registerClass(Minimum2);\nvar Concatenate = class extends Merge {\n  constructor(args) {\n    super(args);\n    this.DEFAULT_AXIS = -1;\n    if (args == null) {\n      args = {};\n    }\n    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;\n    this.supportsMasking = true;\n    this.reshapeRequired = false;\n  }\n  build(inputShape) {\n    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0])) || inputShape.length === 1) {\n      throw new ValueError(\"A `Concatenate` layer should be called on a list of at least 2 inputs\");\n    }\n    inputShape = inputShape;\n    let allNoneShape = true;\n    for (const shape of inputShape) {\n      if (shape != null) {\n        allNoneShape = false;\n        break;\n      }\n    }\n    if (allNoneShape) {\n      return;\n    }\n    const shapeSet = [];\n    for (let i = 0; i < inputShape.length; ++i) {\n      const shapeWithoutConcatAxis = inputShape[i].slice();\n      shapeWithoutConcatAxis.splice(this.axis, 1);\n      let exists = false;\n      for (const shape of shapeSet) {\n        if (util_exports.arraysEqual(shape, shapeWithoutConcatAxis)) {\n          exists = true;\n          break;\n        }\n      }\n      if (!exists) {\n        shapeSet.push(shapeWithoutConcatAxis);\n      }\n    }\n    if (shapeSet.length > 1) {\n      throw new ValueError(\"A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: \" + JSON.stringify(inputShape));\n    }\n  }\n  mergeFunction(inputs) {\n    return tidy(() => {\n      return concatenate(inputs, this.axis);\n    });\n  }\n  computeOutputShape(inputShape) {\n    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0]))) {\n      throw new ValueError(\"A `Concatenate` layer should be called on a list of inputs.\");\n    }\n    const inputShapes = inputShape;\n    const outputShape = inputShapes[0].slice();\n    const axis = this.axis < 0 ? outputShape.length + this.axis : this.axis;\n    for (const shape of inputShapes.slice(1)) {\n      if (outputShape[axis] == null || shape[axis] == null) {\n        outputShape[axis] = null;\n        break;\n      }\n      outputShape[axis] += shape[axis];\n    }\n    return outputShape;\n  }\n  computeMask(inputs, mask) {\n    if (mask == null) {\n      return null;\n    }\n    if (!Array.isArray(mask)) {\n      throw new ValueError(\"`mask` should be an array for Concatenate\");\n    }\n    if (!Array.isArray(inputs)) {\n      throw new ValueError(\"`inputs` should be an array for Concatenate\");\n    }\n    if (mask.length !== inputs.length) {\n      throw new ValueError(`Mismatch in the length of mask (${mask.length}) and the legnth of inputs (${inputs.length})`);\n    }\n    return tidy(() => {\n      let allNullMasks = true;\n      mask.forEach((m) => {\n        if (m != null) {\n          allNullMasks = false;\n          return;\n        }\n      });\n      if (allNullMasks) {\n        return null;\n      }\n      const outputMasks = [];\n      for (let i = 0; i < inputs.length; ++i) {\n        if (mask[i] == null) {\n          outputMasks.push(cast(onesLike(inputs[i]), \"bool\"));\n        } else if (mask[i].rank < inputs[i].rank) {\n          outputMasks.push(expandDims(mask[i], -1));\n        } else {\n          outputMasks.push(mask[i]);\n        }\n      }\n      const concatenatedMasks = concat(outputMasks, this.axis);\n      return all(concatenatedMasks, -1, false);\n    });\n  }\n  getConfig() {\n    const config = {\n      \"axis\": this.axis\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nConcatenate.className = \"Concatenate\";\nserialization_exports.registerClass(Concatenate);\nfunction interpretAxis(axis, dim) {\n  while (axis < 0) {\n    axis += dim;\n  }\n  return axis;\n}\nfunction batchDot(x, y, axes) {\n  if (x.shape.length > 3 || y.shape.length > 3) {\n    throw new NotImplementedError(\"batchDot is not implemented for tensors of 4D or higher rank yet\");\n  }\n  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of x to be >= 2, but got ${x.shape.length}`);\n  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of y to be >= 2, but got ${y.shape.length}`);\n  if (typeof axes === \"number\") {\n    axes = [axes, axes];\n  }\n  if (x.dtype === \"complex64\" || y.dtype === \"complex64\") {\n    throw new NotImplementedError(\"batchDot is not implemented for complex64-type Tensors yet.\");\n  }\n  const xNDim = x.shape.length;\n  const yNDim = y.shape.length;\n  if (axes == null) {\n    axes = [xNDim - 1, yNDim - 2];\n  }\n  const axesArray = axes;\n  return tidy(() => {\n    let diff;\n    if (xNDim > yNDim) {\n      diff = xNDim - yNDim;\n      const diffShape = [];\n      for (let i = 0; i < diff; ++i) {\n        diffShape.push(1);\n      }\n      y = reshape(y, y.shape.concat(diffShape));\n    } else if (yNDim > xNDim) {\n      diff = yNDim - xNDim;\n      const diffShape = [];\n      for (let i = 0; i < diff; ++i) {\n        diffShape.push(1);\n      }\n      x = reshape(x, x.shape.concat(diffShape));\n    } else {\n      diff = 0;\n    }\n    let out;\n    if (x.shape.length === 2 && y.shape.length === 2) {\n      if (axesArray[0] === axesArray[1]) {\n        out = sum2(mul(x, y), axesArray[0]);\n      } else {\n        out = sum2(mul(transpose(x, [1, 0]), y), axesArray[1]);\n      }\n    } else {\n      const adjX = axesArray[0] !== x.shape.length - 1;\n      const adjY = axesArray[1] === y.shape.length - 1;\n      out = matMul(x, y, adjX, adjY);\n    }\n    if (diff > 0) {\n      let idx;\n      if (xNDim > yNDim) {\n        idx = xNDim + yNDim - 3;\n      } else {\n        idx = xNDim - 1;\n      }\n      const squeezeAxes = [];\n      for (let i = idx; i < idx + diff; ++i) {\n        squeezeAxes.push(i);\n      }\n      out = squeeze(out, squeezeAxes);\n    }\n    if (out.shape.length === 1) {\n      out = expandDims(out, 1);\n    }\n    return out;\n  });\n}\nvar Dot = class extends Merge {\n  constructor(args) {\n    super(args);\n    this.axes = args.axes;\n    this.normalize = args.normalize == null ? false : args.normalize;\n    this.supportsMasking = true;\n    this.reshapeRequired = false;\n  }\n  build(inputShape) {\n    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => \"A `Dot` layer should be called on a list of exactly 2 inputs.\");\n    const shape1 = inputShape[0];\n    const shape2 = inputShape[1];\n    if (shape1.length > 3 || shape2.length > 3) {\n      throw new NotImplementedError(\"Dot layer does not support tensors of 4D or higher rank yet.\");\n    }\n    const axes = this.interpretAxes(shape1, shape2);\n    if (shape1[axes[0]] !== shape2[axes[1]]) {\n      throw new ValueError(`Dimension incompatibility: ${shape1[axes[0]]} !== ${shape2[axes[1]]}`);\n    }\n  }\n  mergeFunction(inputs) {\n    if (inputs.length !== 2) {\n      throw new ValueError(`A \\`Dot\\` layer must be called on exactly 2 inputs, but received ${inputs.length} input(s).`);\n    }\n    let x1 = inputs[0];\n    let x2 = inputs[1];\n    let axes;\n    if (!Array.isArray(this.axes)) {\n      axes = [\n        interpretAxis(this.axes, x1.shape.length),\n        interpretAxis(this.axes, x2.shape.length)\n      ];\n    } else {\n      axes = this.axes.map((axis, i) => interpretAxis(axis, inputs[i].shape.length));\n    }\n    if (this.normalize) {\n      x1 = l2Normalize(x1, axes[0]);\n      x2 = l2Normalize(x2, axes[1]);\n    }\n    return batchDot(x1, x2, axes);\n  }\n  interpretAxes(shape1, shape2) {\n    let axes;\n    if (!Array.isArray(this.axes)) {\n      axes = [\n        interpretAxis(this.axes, shape1.length),\n        interpretAxis(this.axes, shape2.length)\n      ];\n    } else {\n      axes = this.axes;\n    }\n    return axes;\n  }\n  computeOutputShape(inputShape) {\n    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => \"A `Dot` layer should be called on a list of exactly 2 inputs.\");\n    const shape1 = inputShape[0].slice();\n    const shape2 = inputShape[1].slice();\n    if (shape1.length > 3 || shape2.length > 3) {\n      throw new NotImplementedError(\"Dot layer does not support tensors of 4D or higher rank yet.\");\n    }\n    const axes = this.interpretAxes(shape1, shape2);\n    shape1.splice(axes[0], 1);\n    shape2.splice(axes[1], 1);\n    shape2.splice(0, 1);\n    const outputShape = shape1.concat(shape2);\n    if (outputShape.length === 1) {\n      outputShape.push(1);\n    }\n    return outputShape;\n  }\n  computeMask(inputs, mask) {\n    return null;\n  }\n  getConfig() {\n    const config = {\n      \"axes\": this.axes,\n      \"normalize\": this.normalize\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nDot.className = \"Dot\";\nserialization_exports.registerClass(Dot);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/noise.js\nvar GaussianNoise = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.supportsMasking = true;\n    this.stddev = args.stddev;\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = { stddev: this.stddev };\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      const noised = () => add2(randomNormal2(input2.shape, 0, this.stddev), input2);\n      const output = inTrainPhase(noised, () => input2, kwargs[\"training\"] || false);\n      return output;\n    });\n  }\n};\nGaussianNoise.className = \"GaussianNoise\";\nserialization_exports.registerClass(GaussianNoise);\nvar GaussianDropout = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.supportsMasking = true;\n    this.rate = args.rate;\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = { rate: this.rate };\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      const input2 = getExactlyOneTensor(inputs);\n      if (this.rate > 0 && this.rate < 1) {\n        const noised = () => {\n          const stddev = Math.sqrt(this.rate / (1 - this.rate));\n          return mul(input2, randomNormal2(input2.shape, 1, stddev));\n        };\n        return inTrainPhase(noised, () => input2, kwargs[\"training\"] || false);\n      }\n      return input2;\n    });\n  }\n};\nGaussianDropout.className = \"GaussianDropout\";\nserialization_exports.registerClass(GaussianDropout);\nvar AlphaDropout = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.supportsMasking = true;\n    this.rate = args.rate;\n    this.noiseShape = args.noiseShape;\n  }\n  _getNoiseShape(inputs) {\n    return this.noiseShape || getExactlyOneTensor(inputs).shape;\n  }\n  computeOutputShape(inputShape) {\n    return inputShape;\n  }\n  getConfig() {\n    const baseConfig = super.getConfig();\n    const config = { rate: this.rate };\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      if (this.rate < 1 && this.rate > 0) {\n        const noiseShape = this._getNoiseShape(inputs);\n        const droppedInputs = () => {\n          const input2 = getExactlyOneTensor(inputs);\n          const alpha = 1.6732632423543772;\n          const scale2 = 1.0507009873554805;\n          const alphaP = -alpha * scale2;\n          let keptIdx = greaterEqual(randomUniform(noiseShape), this.rate);\n          keptIdx = cast2(keptIdx, \"float32\");\n          const a = ((1 - this.rate) * (1 + this.rate * alphaP ** 2)) ** -0.5;\n          const b = -a * alphaP * this.rate;\n          const x = add2(mul(input2, keptIdx), mul(add2(keptIdx, -1), alphaP));\n          return add2(mul(x, a), b);\n        };\n        return inTrainPhase(droppedInputs, () => getExactlyOneTensor(inputs), kwargs[\"training\"] || false);\n      }\n      return inputs;\n    });\n  }\n};\nAlphaDropout.className = \"AlphaDropout\";\nserialization_exports.registerClass(AlphaDropout);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/normalization.js\nfunction batchNormalization(x, mean4, variance, beta, gamma, epsilon3 = 1e-3) {\n  let out;\n  if (x.rank === 2) {\n    out = batchNorm2d(x, mean4, variance, beta, gamma, epsilon3);\n  } else if (x.rank === 3) {\n    out = batchNorm3d(x, mean4, variance, beta, gamma, epsilon3);\n  } else if (x.rank === 4) {\n    out = batchNorm4d(x, mean4, variance, beta, gamma, epsilon3);\n  } else {\n    throw new NotImplementedError(`batchNormalization is not implemented for array of rank ${x.rank} yet`);\n  }\n  return out;\n}\nfunction regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {\n  return tidy(() => {\n    const meanAndVariance = moments(x, reductionAxes);\n    const mean4 = meanAndVariance.mean;\n    const variance = meanAndVariance.variance;\n    const normed = batchNormalization(x, mean4, variance, beta, gamma, epsilon3);\n    return [normed, mean4, variance];\n  });\n}\nfunction broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {\n  return tidy(() => {\n    const meanAndVariance = moments(x, reductionAxes);\n    const mean4 = meanAndVariance.mean;\n    const variance = meanAndVariance.variance;\n    const targetShape = [];\n    for (const axis of range2(0, x.rank)) {\n      if (reductionAxes.indexOf(axis) !== -1) {\n        targetShape.push(1);\n      } else {\n        targetShape.push(x.shape[axis]);\n      }\n    }\n    const broadcastMean = reshape(mean4, targetShape);\n    const broadcastVariance = reshape(variance, targetShape);\n    const broadcastGamma = gamma == null ? null : reshape(gamma, targetShape);\n    const broadcastBeta = beta == null ? null : reshape(beta, targetShape);\n    const normed = batchNormalization(x, broadcastMean, broadcastVariance, broadcastBeta, broadcastGamma, epsilon3);\n    return [normed, mean4, variance];\n  });\n}\nfunction normalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {\n  if (util_exports.arraysEqual(reductionAxes.slice().sort(), range2(0, x.rank - 1))) {\n    return regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3);\n  } else {\n    return broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3);\n  }\n}\nvar BatchNormalization = class extends Layer {\n  constructor(args) {\n    if (args == null) {\n      args = {};\n    }\n    super(args);\n    this.supportsMasking = true;\n    this.axis = args.axis == null ? -1 : args.axis;\n    this.momentum = args.momentum == null ? 0.99 : args.momentum;\n    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;\n    this.center = args.center == null ? true : args.center;\n    this.scale = args.scale == null ? true : args.scale;\n    this.betaInitializer = getInitializer(args.betaInitializer || \"zeros\");\n    this.gammaInitializer = getInitializer(args.gammaInitializer || \"ones\");\n    this.movingMeanInitializer = getInitializer(args.movingMeanInitializer || \"zeros\");\n    this.movingVarianceInitializer = getInitializer(args.movingVarianceInitializer || \"ones\");\n    this.betaConstraint = getConstraint(args.betaConstraint);\n    this.gammaConstraint = getConstraint(args.gammaConstraint);\n    this.betaRegularizer = getRegularizer(args.betaRegularizer);\n    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const axis = this.axis >= 0 ? this.axis : this.axis + inputShape.length;\n    const dim = inputShape[axis];\n    if (dim == null) {\n      throw new ValueError(`Axis ${axis} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(inputShape)}.`);\n    }\n    this.inputSpec = [new InputSpec({ ndim: inputShape.length, axes: { [axis]: dim } })];\n    const shape = [dim];\n    if (this.scale) {\n      this.gamma = this.addWeight(\"gamma\", shape, null, this.gammaInitializer, this.gammaRegularizer, true, this.gammaConstraint);\n    }\n    if (this.center) {\n      this.beta = this.addWeight(\"beta\", shape, null, this.betaInitializer, this.betaRegularizer, true, this.betaConstraint);\n    }\n    this.movingMean = this.addWeight(\"moving_mean\", shape, null, this.movingMeanInitializer, null, false);\n    this.movingVariance = this.addWeight(\"moving_variance\", shape, null, this.movingVarianceInitializer, null, false);\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const training = kwargs[\"training\"] == null ? false : kwargs[\"training\"];\n      const input2 = getExactlyOneTensor(inputs);\n      const inputShape = input2.shape;\n      const ndim = inputShape.length;\n      const reductionAxes = range2(0, ndim);\n      const axis = this.axis >= 0 ? this.axis : this.axis + ndim;\n      reductionAxes.splice(axis, 1);\n      const broadcastShape = pyListRepeat(1, ndim);\n      broadcastShape[axis] = inputShape[axis];\n      const sortedReductionAxes = reductionAxes.slice();\n      sortedReductionAxes.sort();\n      const needsBroadcasting = !util_exports.arraysEqual(sortedReductionAxes, range2(0, ndim).slice(0, ndim - 1));\n      const normalizeInference = () => {\n        if (needsBroadcasting) {\n          const broadcastMovingMean = reshape(this.movingMean.read(), broadcastShape);\n          const broadcastMovingVariance = reshape(this.movingVariance.read(), broadcastShape);\n          const broadcastBeta = this.center ? reshape(this.beta.read(), broadcastShape) : null;\n          const broadcastGamma = this.scale ? reshape(this.gamma.read(), broadcastShape) : null;\n          return batchNormalization(input2, broadcastMovingMean, broadcastMovingVariance, broadcastBeta, broadcastGamma, this.epsilon);\n        } else {\n          return batchNormalization(input2, this.movingMean.read(), this.movingVariance.read(), this.beta == null ? null : this.beta.read(), this.gamma == null ? null : this.gamma.read(), this.epsilon);\n        }\n      };\n      if (!training) {\n        return normalizeInference();\n      }\n      const [normedTraining, mean4, variance] = normalizeBatchInTraining(input2, this.gamma.read(), this.beta.read(), reductionAxes, this.epsilon);\n      const doMovingAverage = (variable2, value, momentum) => {\n        tidy(() => {\n          const decay = 1 - momentum;\n          const origValue = variable2.read();\n          const updateDelta = mul(sub(origValue, value), decay);\n          variable2.write(sub(origValue, updateDelta));\n        });\n      };\n      const updateMovingMeanAndVariance = () => {\n        doMovingAverage(this.movingMean, mean4, this.momentum);\n        doMovingAverage(this.movingVariance, variance, this.momentum);\n      };\n      updateMovingMeanAndVariance();\n      return normedTraining;\n    });\n  }\n  getConfig() {\n    const config = {\n      axis: this.axis,\n      momentum: this.momentum,\n      epsilon: this.epsilon,\n      center: this.center,\n      scale: this.scale,\n      betaInitializer: serializeInitializer(this.betaInitializer),\n      gammaInitializer: serializeInitializer(this.gammaInitializer),\n      movingMeanInitializer: serializeInitializer(this.movingMeanInitializer),\n      movingVarianceInitializer: serializeInitializer(this.movingVarianceInitializer),\n      betaRegularizer: serializeRegularizer(this.betaRegularizer),\n      gammaRegularizer: serializeRegularizer(this.gammaRegularizer),\n      betaConstraint: serializeConstraint(this.betaConstraint),\n      gammaConstraint: serializeConstraint(this.gammaConstraint)\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nBatchNormalization.className = \"BatchNormalization\";\nserialization_exports.registerClass(BatchNormalization);\nvar LayerNormalization = class extends Layer {\n  constructor(args) {\n    if (args == null) {\n      args = {};\n    }\n    super(args);\n    this.axis = args.axis == null ? -1 : args.axis;\n    if (typeof this.axis === \"number\") {\n      if (!Number.isInteger(this.axis)) {\n        throw new Error(`Expected axis to be an integer, but received ${this.axis}`);\n      }\n    } else if (Array.isArray(this.axis)) {\n      for (const axis of this.axis) {\n        if (!Number.isInteger(axis)) {\n          throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`);\n        }\n      }\n    } else {\n      throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);\n    }\n    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;\n    this.center = args.center == null ? true : args.center;\n    this.scale = args.scale == null ? true : args.scale;\n    this.betaInitializer = getInitializer(args.betaInitializer || \"zeros\");\n    this.gammaInitializer = getInitializer(args.gammaInitializer || \"ones\");\n    this.betaRegularizer = getRegularizer(args.betaRegularizer);\n    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);\n    this.supportsMasking = true;\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const nDims = inputShape.length;\n    if (typeof this.axis === \"number\") {\n      this.axis = [this.axis];\n    }\n    for (let i = 0; i < this.axis.length; ++i) {\n      if (this.axis[i] < 0) {\n        this.axis[i] += nDims;\n      }\n    }\n    for (const axis of this.axis) {\n      if (axis < 0 || axis >= nDims) {\n        throw new Error(`Invalid axis: ${axis}`);\n      }\n    }\n    if (this.axis.length !== unique2(this.axis).length) {\n      throw new Error(`Found duplicate axes in: ${this.axis}`);\n    }\n    const paramShape = this.axis.map((axis) => inputShape[axis]);\n    const trainable = true;\n    if (this.scale) {\n      this.gamma = this.addWeight(\"gamma\", paramShape, \"float32\", this.gammaInitializer, this.gammaRegularizer, trainable);\n    } else {\n      this.gamma = null;\n    }\n    if (this.center) {\n      this.beta = this.addWeight(\"beta\", paramShape, \"float32\", this.betaInitializer, this.betaRegularizer, trainable);\n    } else {\n      this.beta = null;\n    }\n    this.built = true;\n  }\n  call(inputs, kwargs) {\n    const input2 = getExactlyOneTensor(inputs);\n    const inputShape = input2.shape;\n    const nDims = inputShape.length;\n    return tidy(() => {\n      const keepDims = true;\n      let { mean: mean4, variance } = moments(input2, this.axis, keepDims);\n      const broadcastShape = pyListRepeat(1, nDims);\n      for (const dim of this.axis) {\n        broadcastShape[dim] = inputShape[dim];\n      }\n      const broadcast = (v) => {\n        if (v != null && v.shape.length !== nDims) {\n          return reshape(v, broadcastShape);\n        } else {\n          return v;\n        }\n      };\n      let scale2 = this.scale ? broadcast(this.gamma.read()) : null;\n      let offset = this.center ? broadcast(this.beta.read()) : null;\n      const momentsTiling = [];\n      const scaleOffsetTiling = [];\n      for (let i = 0; i < nDims; ++i) {\n        if (this.axis.indexOf(i) !== -1) {\n          momentsTiling.push(inputShape[i]);\n          scaleOffsetTiling.push(1);\n        } else {\n          momentsTiling.push(1);\n          scaleOffsetTiling.push(inputShape[i]);\n        }\n      }\n      mean4 = tile(mean4, momentsTiling);\n      variance = tile(variance, momentsTiling);\n      if (scale2 != null) {\n        scale2 = tile(scale2, scaleOffsetTiling);\n      }\n      if (offset != null) {\n        offset = tile(offset, scaleOffsetTiling);\n      }\n      return batchNormalization(input2, mean4, variance, offset, scale2, this.epsilon);\n    });\n  }\n  getConfig() {\n    const config = {\n      axis: this.axis,\n      epsilon: this.epsilon,\n      center: this.center,\n      scale: this.scale,\n      betaInitializer: serializeInitializer(this.betaInitializer),\n      gammaInitializer: serializeInitializer(this.gammaInitializer),\n      betaRegularizer: serializeRegularizer(this.betaRegularizer),\n      gammaRegularizer: serializeRegularizer(this.gammaRegularizer)\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nLayerNormalization.className = \"LayerNormalization\";\nserialization_exports.registerClass(LayerNormalization);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/padding.js\nfunction spatial2dPadding(x, padding, dataFormat) {\n  return tidy(() => {\n    if (x.rank !== 4) {\n      throw new ValueError(`temporalPadding expects input tensor to be 4-D, but received a ${x.rank}-D tensor.`);\n    }\n    if (padding == null) {\n      padding = [[1, 1], [1, 1]];\n    }\n    if (padding.length !== 2 || padding[0].length !== 2 || padding[1].length !== 2) {\n      throw new ValueError(\"spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.\");\n    }\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    if (dataFormat !== \"channelsLast\" && dataFormat !== \"channelsFirst\") {\n      throw new ValueError(`Unknown data format: ${dataFormat}. Supported data formats are 'channelsLast' and 'channelsFirst.`);\n    }\n    let pattern;\n    if (dataFormat === \"channelsFirst\") {\n      pattern = [[0, 0], [0, 0], padding[0], padding[1]];\n    } else {\n      pattern = [[0, 0], padding[0], padding[1], [0, 0]];\n    }\n    return pad(x, pattern);\n  });\n}\nvar ZeroPadding2D = class extends Layer {\n  constructor(args) {\n    if (args == null) {\n      args = {};\n    }\n    super(args);\n    this.dataFormat = args.dataFormat == null ? imageDataFormat() : args.dataFormat;\n    if (args.padding == null) {\n      this.padding = [[1, 1], [1, 1]];\n    } else if (typeof args.padding === \"number\") {\n      this.padding = [[args.padding, args.padding], [args.padding, args.padding]];\n    } else {\n      args.padding = args.padding;\n      if (args.padding.length !== 2) {\n        throw new ValueError(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${args.padding.length} array.`);\n      }\n      let heightPadding;\n      let widthPadding;\n      if (typeof args.padding[0] === \"number\") {\n        heightPadding = [args.padding[0], args.padding[0]];\n        widthPadding = [args.padding[1], args.padding[1]];\n      } else {\n        args.padding = args.padding;\n        if (args.padding[0].length !== 2) {\n          throw new ValueError(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${args.padding[0].length} array.`);\n        }\n        heightPadding = args.padding[0];\n        if (args.padding[1].length !== 2) {\n          throw new ValueError(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${args.padding[1].length} array.`);\n        }\n        widthPadding = args.padding[1];\n      }\n      this.padding = [heightPadding, widthPadding];\n    }\n    this.inputSpec = [new InputSpec({ ndim: 4 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    let rows;\n    let cols;\n    if (this.dataFormat === \"channelsFirst\") {\n      if (inputShape[2] != null && inputShape[2] >= 0) {\n        rows = inputShape[2] + this.padding[0][0] + this.padding[0][1];\n      } else {\n        rows = null;\n      }\n      if (inputShape[3] != null && inputShape[3] >= 0) {\n        cols = inputShape[3] + this.padding[1][0] + this.padding[1][1];\n      } else {\n        cols = null;\n      }\n      return [inputShape[0], inputShape[1], rows, cols];\n    } else {\n      if (inputShape[1] != null && inputShape[1] >= 0) {\n        rows = inputShape[1] + this.padding[0][0] + this.padding[0][1];\n      } else {\n        rows = null;\n      }\n      if (inputShape[2] != null && inputShape[2] >= 0) {\n        cols = inputShape[2] + this.padding[1][0] + this.padding[1][1];\n      } else {\n        cols = null;\n      }\n      return [inputShape[0], rows, cols, inputShape[3]];\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => spatial2dPadding(getExactlyOneTensor(inputs), this.padding, this.dataFormat));\n  }\n  getConfig() {\n    const config = {\n      padding: this.padding,\n      dataFormat: this.dataFormat\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nZeroPadding2D.className = \"ZeroPadding2D\";\nserialization_exports.registerClass(ZeroPadding2D);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/pooling.js\nfunction pool2d(x, poolSize, strides, padding, dataFormat, poolMode) {\n  return tidy(() => {\n    checkDataFormat(dataFormat);\n    checkPoolMode(poolMode);\n    checkPaddingMode(padding);\n    if (strides == null) {\n      strides = [1, 1];\n    }\n    if (padding == null) {\n      padding = \"valid\";\n    }\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    if (poolMode == null) {\n      poolMode = \"max\";\n    }\n    x = preprocessConv2DInput(x, dataFormat);\n    let y;\n    const paddingString = padding === \"same\" ? \"same\" : \"valid\";\n    if (poolMode === \"max\") {\n      y = maxPool(x, poolSize, strides, paddingString);\n    } else {\n      y = avgPool(\n        x,\n        poolSize,\n        strides,\n        paddingString\n      );\n    }\n    if (dataFormat === \"channelsFirst\") {\n      y = transpose(y, [0, 3, 1, 2]);\n    }\n    return y;\n  });\n}\nfunction pool3d(x, poolSize, strides, padding, dataFormat, poolMode) {\n  return tidy(() => {\n    checkDataFormat(dataFormat);\n    checkPoolMode(poolMode);\n    checkPaddingMode(padding);\n    if (strides == null) {\n      strides = [1, 1, 1];\n    }\n    if (padding == null) {\n      padding = \"valid\";\n    }\n    if (dataFormat == null) {\n      dataFormat = imageDataFormat();\n    }\n    if (poolMode == null) {\n      poolMode = \"max\";\n    }\n    x = preprocessConv3DInput(x, dataFormat);\n    let y;\n    const paddingString = padding === \"same\" ? \"same\" : \"valid\";\n    if (poolMode === \"max\") {\n      y = maxPool3d(x, poolSize, strides, paddingString);\n    } else {\n      y = avgPool3d(x, poolSize, strides, paddingString);\n    }\n    if (dataFormat === \"channelsFirst\") {\n      y = transpose(y, [0, 4, 1, 2, 3]);\n    }\n    return y;\n  });\n}\nvar Pooling1D = class extends Layer {\n  constructor(args) {\n    if (args.poolSize == null) {\n      args.poolSize = 2;\n    }\n    super(args);\n    if (typeof args.poolSize === \"number\") {\n      this.poolSize = [args.poolSize];\n    } else if (Array.isArray(args.poolSize) && args.poolSize.length === 1 && typeof args.poolSize[0] === \"number\") {\n      this.poolSize = args.poolSize;\n    } else {\n      throw new ValueError(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.poolSize)}`);\n    }\n    assertPositiveInteger(this.poolSize, \"poolSize\");\n    if (args.strides == null) {\n      this.strides = this.poolSize;\n    } else {\n      if (typeof args.strides === \"number\") {\n        this.strides = [args.strides];\n      } else if (Array.isArray(args.strides) && args.strides.length === 1 && typeof args.strides[0] === \"number\") {\n        this.strides = args.strides;\n      } else {\n        throw new ValueError(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.strides)}`);\n      }\n    }\n    assertPositiveInteger(this.strides, \"strides\");\n    this.padding = args.padding == null ? \"valid\" : args.padding;\n    checkPaddingMode(this.padding);\n    this.inputSpec = [new InputSpec({ ndim: 3 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const length = convOutputLength(inputShape[1], this.poolSize[0], this.padding, this.strides[0]);\n    return [inputShape[0], length, inputShape[2]];\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      inputs = expandDims2(getExactlyOneTensor(inputs), 2);\n      const output = this.poolingFunction(getExactlyOneTensor(inputs), [this.poolSize[0], 1], [this.strides[0], 1], this.padding, \"channelsLast\");\n      return squeeze(output, [2]);\n    });\n  }\n  getConfig() {\n    const config = {\n      poolSize: this.poolSize,\n      padding: this.padding,\n      strides: this.strides\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nvar MaxPooling1D = class extends Pooling1D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool2d(inputs, poolSize, strides, padding, dataFormat, \"max\");\n  }\n};\nMaxPooling1D.className = \"MaxPooling1D\";\nserialization_exports.registerClass(MaxPooling1D);\nvar AveragePooling1D = class extends Pooling1D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool2d(inputs, poolSize, strides, padding, dataFormat, \"avg\");\n  }\n};\nAveragePooling1D.className = \"AveragePooling1D\";\nserialization_exports.registerClass(AveragePooling1D);\nvar Pooling2D = class extends Layer {\n  constructor(args) {\n    if (args.poolSize == null) {\n      args.poolSize = [2, 2];\n    }\n    super(args);\n    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize];\n    if (args.strides == null) {\n      this.strides = this.poolSize;\n    } else if (Array.isArray(args.strides)) {\n      if (args.strides.length !== 2) {\n        throw new ValueError(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${args.strides.length}.`);\n      }\n      this.strides = args.strides;\n    } else {\n      this.strides = [args.strides, args.strides];\n    }\n    assertPositiveInteger(this.poolSize, \"poolSize\");\n    assertPositiveInteger(this.strides, \"strides\");\n    this.padding = args.padding == null ? \"valid\" : args.padding;\n    this.dataFormat = args.dataFormat == null ? \"channelsLast\" : args.dataFormat;\n    checkDataFormat(this.dataFormat);\n    checkPaddingMode(this.padding);\n    this.inputSpec = [new InputSpec({ ndim: 4 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    let rows = this.dataFormat === \"channelsFirst\" ? inputShape[2] : inputShape[1];\n    let cols = this.dataFormat === \"channelsFirst\" ? inputShape[3] : inputShape[2];\n    rows = convOutputLength(rows, this.poolSize[0], this.padding, this.strides[0]);\n    cols = convOutputLength(cols, this.poolSize[1], this.padding, this.strides[1]);\n    if (this.dataFormat === \"channelsFirst\") {\n      return [inputShape[0], inputShape[1], rows, cols];\n    } else {\n      return [inputShape[0], rows, cols, inputShape[3]];\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);\n    });\n  }\n  getConfig() {\n    const config = {\n      poolSize: this.poolSize,\n      padding: this.padding,\n      strides: this.strides,\n      dataFormat: this.dataFormat\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nvar MaxPooling2D = class extends Pooling2D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool2d(inputs, poolSize, strides, padding, dataFormat, \"max\");\n  }\n};\nMaxPooling2D.className = \"MaxPooling2D\";\nserialization_exports.registerClass(MaxPooling2D);\nvar AveragePooling2D = class extends Pooling2D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool2d(inputs, poolSize, strides, padding, dataFormat, \"avg\");\n  }\n};\nAveragePooling2D.className = \"AveragePooling2D\";\nserialization_exports.registerClass(AveragePooling2D);\nvar Pooling3D = class extends Layer {\n  constructor(args) {\n    if (args.poolSize == null) {\n      args.poolSize = [2, 2, 2];\n    }\n    super(args);\n    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize, args.poolSize];\n    if (args.strides == null) {\n      this.strides = this.poolSize;\n    } else if (Array.isArray(args.strides)) {\n      if (args.strides.length !== 3) {\n        throw new ValueError(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${args.strides.length}.`);\n      }\n      this.strides = args.strides;\n    } else {\n      this.strides = [args.strides, args.strides, args.strides];\n    }\n    assertPositiveInteger(this.poolSize, \"poolSize\");\n    assertPositiveInteger(this.strides, \"strides\");\n    this.padding = args.padding == null ? \"valid\" : args.padding;\n    this.dataFormat = args.dataFormat == null ? \"channelsLast\" : args.dataFormat;\n    checkDataFormat(this.dataFormat);\n    checkPaddingMode(this.padding);\n    this.inputSpec = [new InputSpec({ ndim: 5 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    let depths = this.dataFormat === \"channelsFirst\" ? inputShape[2] : inputShape[1];\n    let rows = this.dataFormat === \"channelsFirst\" ? inputShape[3] : inputShape[2];\n    let cols = this.dataFormat === \"channelsFirst\" ? inputShape[4] : inputShape[3];\n    depths = convOutputLength(depths, this.poolSize[0], this.padding, this.strides[0]);\n    rows = convOutputLength(rows, this.poolSize[1], this.padding, this.strides[1]);\n    cols = convOutputLength(cols, this.poolSize[2], this.padding, this.strides[2]);\n    if (this.dataFormat === \"channelsFirst\") {\n      return [inputShape[0], inputShape[1], depths, rows, cols];\n    } else {\n      return [inputShape[0], depths, rows, cols, inputShape[4]];\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      this.invokeCallHook(inputs, kwargs);\n      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);\n    });\n  }\n  getConfig() {\n    const config = {\n      poolSize: this.poolSize,\n      padding: this.padding,\n      strides: this.strides,\n      dataFormat: this.dataFormat\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nvar MaxPooling3D = class extends Pooling3D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool3d(inputs, poolSize, strides, padding, dataFormat, \"max\");\n  }\n};\nMaxPooling3D.className = \"MaxPooling3D\";\nserialization_exports.registerClass(MaxPooling3D);\nvar AveragePooling3D = class extends Pooling3D {\n  constructor(args) {\n    super(args);\n  }\n  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {\n    checkDataFormat(dataFormat);\n    checkPaddingMode(padding);\n    return pool3d(inputs, poolSize, strides, padding, dataFormat, \"avg\");\n  }\n};\nAveragePooling3D.className = \"AveragePooling3D\";\nserialization_exports.registerClass(AveragePooling3D);\nvar GlobalPooling1D = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.inputSpec = [new InputSpec({ ndim: 3 })];\n  }\n  computeOutputShape(inputShape) {\n    return [inputShape[0], inputShape[2]];\n  }\n  call(inputs, kwargs) {\n    throw new NotImplementedError();\n  }\n};\nvar GlobalAveragePooling1D = class extends GlobalPooling1D {\n  constructor(args) {\n    super(args || {});\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const input2 = getExactlyOneTensor(inputs);\n      return mean(input2, 1);\n    });\n  }\n};\nGlobalAveragePooling1D.className = \"GlobalAveragePooling1D\";\nserialization_exports.registerClass(GlobalAveragePooling1D);\nvar GlobalMaxPooling1D = class extends GlobalPooling1D {\n  constructor(args) {\n    super(args || {});\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const input2 = getExactlyOneTensor(inputs);\n      return max(input2, 1);\n    });\n  }\n};\nGlobalMaxPooling1D.className = \"GlobalMaxPooling1D\";\nserialization_exports.registerClass(GlobalMaxPooling1D);\nvar GlobalPooling2D = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.dataFormat = args.dataFormat == null ? \"channelsLast\" : args.dataFormat;\n    checkDataFormat(this.dataFormat);\n    this.inputSpec = [new InputSpec({ ndim: 4 })];\n  }\n  computeOutputShape(inputShape) {\n    inputShape = inputShape;\n    if (this.dataFormat === \"channelsLast\") {\n      return [inputShape[0], inputShape[3]];\n    } else {\n      return [inputShape[0], inputShape[1]];\n    }\n  }\n  call(inputs, kwargs) {\n    throw new NotImplementedError();\n  }\n  getConfig() {\n    const config = { dataFormat: this.dataFormat };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n};\nvar GlobalAveragePooling2D = class extends GlobalPooling2D {\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const input2 = getExactlyOneTensor(inputs);\n      if (this.dataFormat === \"channelsLast\") {\n        return mean(input2, [1, 2]);\n      } else {\n        return mean(input2, [2, 3]);\n      }\n    });\n  }\n};\nGlobalAveragePooling2D.className = \"GlobalAveragePooling2D\";\nserialization_exports.registerClass(GlobalAveragePooling2D);\nvar GlobalMaxPooling2D = class extends GlobalPooling2D {\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const input2 = getExactlyOneTensor(inputs);\n      if (this.dataFormat === \"channelsLast\") {\n        return max(input2, [1, 2]);\n      } else {\n        return max(input2, [2, 3]);\n      }\n    });\n  }\n};\nGlobalMaxPooling2D.className = \"GlobalMaxPooling2D\";\nserialization_exports.registerClass(GlobalMaxPooling2D);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/wrappers.js\nvar Wrapper = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.layer = args.layer;\n  }\n  build(inputShape) {\n    this.built = true;\n  }\n  get trainable() {\n    if (this.layer != null) {\n      return this.layer.trainable;\n    } else {\n      return false;\n    }\n  }\n  set trainable(value) {\n    if (this.layer != null) {\n      this.layer.trainable = value;\n    }\n  }\n  get trainableWeights() {\n    return this.layer.trainableWeights;\n  }\n  get nonTrainableWeights() {\n    return this.layer.nonTrainableWeights;\n  }\n  get updates() {\n    return this.layer._updates;\n  }\n  get losses() {\n    return this.layer.losses;\n  }\n  getWeights() {\n    return this.layer.getWeights();\n  }\n  setWeights(weights) {\n    this.layer.setWeights(weights);\n  }\n  getConfig() {\n    const config = {\n      \"layer\": {\n        \"className\": this.layer.getClassName(),\n        \"config\": this.layer.getConfig()\n      }\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  setFastWeightInitDuringBuild(value) {\n    super.setFastWeightInitDuringBuild(value);\n    if (this.layer != null) {\n      this.layer.setFastWeightInitDuringBuild(value);\n    }\n  }\n  static fromConfig(cls, config, customObjects = {}) {\n    const layerConfig = config[\"layer\"];\n    const layer = deserialize(layerConfig, customObjects);\n    delete config[\"layer\"];\n    const newConfig = { layer };\n    Object.assign(newConfig, config);\n    return new cls(newConfig);\n  }\n};\nvar TimeDistributed = class extends Wrapper {\n  constructor(args) {\n    super(args);\n    this.supportsMasking = true;\n  }\n  build(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape.length < 3) {\n      throw new ValueError(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(inputShape)}`);\n    }\n    this.inputSpec = [{ shape: inputShape }];\n    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));\n    if (!this.layer.built) {\n      this.layer.build(childInputShape);\n      this.layer.built = true;\n    }\n    super.build(inputShape);\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));\n    const childOutputShape = this.layer.computeOutputShape(childInputShape);\n    const timesteps = inputShape[1];\n    return [childOutputShape[0], timesteps].concat(childOutputShape.slice(1));\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      const step5 = (inputs2, states) => {\n        const output = getExactlyOneTensor(this.layer.call(inputs2, kwargs));\n        return [output, []];\n      };\n      const rnnOutputs = rnn(step5, inputs, [], false, null, null, false, true);\n      const y = rnnOutputs[1];\n      return y;\n    });\n  }\n};\nTimeDistributed.className = \"TimeDistributed\";\nserialization_exports.registerClass(TimeDistributed);\nfunction checkBidirectionalMergeMode(value) {\n  checkStringTypeUnionValue(VALID_BIDIRECTIONAL_MERGE_MODES, \"BidirectionalMergeMode\", value);\n}\nvar DEFAULT_BIDIRECTIONAL_MERGE_MODE = \"concat\";\nvar Bidirectional = class extends Wrapper {\n  constructor(args) {\n    super(args);\n    const layerConfig = args.layer.getConfig();\n    const forwDict = {};\n    forwDict[\"className\"] = args.layer.getClassName();\n    forwDict[\"config\"] = layerConfig;\n    this.forwardLayer = deserialize(forwDict);\n    layerConfig[\"goBackwards\"] = layerConfig[\"goBackwards\"] === true ? false : true;\n    const backDict = {};\n    backDict[\"className\"] = args.layer.getClassName();\n    backDict[\"config\"] = layerConfig;\n    this.backwardLayer = deserialize(backDict);\n    this.forwardLayer.name = \"forward_\" + this.forwardLayer.name;\n    this.backwardLayer.name = \"backward_\" + this.backwardLayer.name;\n    this.mergeMode = args.mergeMode === void 0 ? DEFAULT_BIDIRECTIONAL_MERGE_MODE : args.mergeMode;\n    checkBidirectionalMergeMode(this.mergeMode);\n    if (args.weights) {\n      throw new NotImplementedError(\"weights support is not implemented for Bidirectional layer yet.\");\n    }\n    this._stateful = args.layer.stateful;\n    this.returnSequences = args.layer.returnSequences;\n    this.returnState = args.layer.returnState;\n    this.supportsMasking = true;\n    this._trainable = true;\n    this.inputSpec = args.layer.inputSpec;\n    this.numConstants = null;\n  }\n  get trainable() {\n    return this._trainable;\n  }\n  set trainable(value) {\n    this._trainable = value;\n    if (this.forwardLayer != null) {\n      this.forwardLayer.trainable = value;\n    }\n    if (this.backwardLayer != null) {\n      this.backwardLayer.trainable = value;\n    }\n  }\n  getWeights() {\n    return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights());\n  }\n  setWeights(weights) {\n    const numWeights = weights.length;\n    const numeightsOver2 = Math.floor(numWeights / 2);\n    this.forwardLayer.setWeights(weights.slice(0, numeightsOver2));\n    this.backwardLayer.setWeights(weights.slice(numeightsOver2));\n  }\n  computeOutputShape(inputShape) {\n    let layerShapes = this.forwardLayer.computeOutputShape(inputShape);\n    if (!(Array.isArray(layerShapes) && Array.isArray(layerShapes[0]))) {\n      layerShapes = [layerShapes];\n    }\n    layerShapes = layerShapes;\n    let outputShape;\n    let outputShapes;\n    let stateShape;\n    if (this.returnState) {\n      stateShape = layerShapes.slice(1);\n      outputShape = layerShapes[0];\n    } else {\n      outputShape = layerShapes[0];\n    }\n    outputShape = outputShape;\n    if (this.mergeMode === \"concat\") {\n      outputShape[outputShape.length - 1] *= 2;\n      outputShapes = [outputShape];\n    } else if (this.mergeMode == null) {\n      outputShapes = [outputShape, outputShape.slice()];\n    } else {\n      outputShapes = [outputShape];\n    }\n    if (this.returnState) {\n      if (this.mergeMode == null) {\n        return outputShapes.concat(stateShape).concat(stateShape.slice());\n      }\n      return [outputShape].concat(stateShape).concat(stateShape.slice());\n    }\n    return singletonOrArray(outputShapes);\n  }\n  apply(inputs, kwargs) {\n    let initialState = kwargs == null ? null : kwargs[\"initialState\"];\n    let constants = kwargs == null ? null : kwargs[\"constants\"];\n    if (kwargs == null) {\n      kwargs = {};\n    }\n    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);\n    inputs = standardized.inputs;\n    initialState = standardized.initialState;\n    constants = standardized.constants;\n    if (Array.isArray(inputs)) {\n      initialState = inputs.slice(1);\n      inputs = inputs[0];\n    }\n    if ((initialState == null || initialState.length === 0) && constants == null) {\n      return super.apply(inputs, kwargs);\n    }\n    const additionalInputs = [];\n    const additionalSpecs = [];\n    if (initialState != null) {\n      const numStates = initialState.length;\n      if (numStates % 2 > 0) {\n        throw new ValueError(\"When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.\");\n      }\n      kwargs[\"initialState\"] = initialState;\n      additionalInputs.push(...initialState);\n      const stateSpecs = initialState.map((state) => new InputSpec({ shape: state.shape }));\n      this.forwardLayer.stateSpec = stateSpecs.slice(0, numStates / 2);\n      this.backwardLayer.stateSpec = stateSpecs.slice(numStates / 2);\n      additionalSpecs.push(...stateSpecs);\n    }\n    if (constants != null) {\n      throw new NotImplementedError(\"Support for constants in Bidirectional layers is not implemented yet.\");\n    }\n    const isSymbolicTensor = additionalInputs[0] instanceof SymbolicTensor;\n    for (const tensor2 of additionalInputs) {\n      if (tensor2 instanceof SymbolicTensor !== isSymbolicTensor) {\n        throw new ValueError(\"The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors\");\n      }\n    }\n    if (isSymbolicTensor) {\n      const fullInput = [inputs].concat(additionalInputs);\n      const fullInputSpec = this.inputSpec.concat(additionalSpecs);\n      const originalInputSpec = this.inputSpec;\n      this.inputSpec = fullInputSpec;\n      const output = super.apply(fullInput, kwargs);\n      this.inputSpec = originalInputSpec;\n      return output;\n    } else {\n      return super.apply(inputs, kwargs);\n    }\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const initialState = kwargs[\"initialState\"];\n      let y;\n      let yRev;\n      if (initialState == null) {\n        y = this.forwardLayer.call(inputs, kwargs);\n        yRev = this.backwardLayer.call(inputs, kwargs);\n      } else {\n        const forwardState = initialState.slice(0, initialState.length / 2);\n        const backwardState = initialState.slice(initialState.length / 2);\n        y = this.forwardLayer.call(inputs, Object.assign(kwargs, { initialState: forwardState }));\n        yRev = this.backwardLayer.call(inputs, Object.assign(kwargs, { initialState: backwardState }));\n      }\n      let states;\n      if (this.returnState) {\n        if (Array.isArray(y)) {\n          states = y.slice(1).concat(yRev.slice(1));\n        } else {\n        }\n        y = y[0];\n        yRev = yRev[0];\n      }\n      if (this.returnSequences) {\n        yRev = reverse(yRev, 1);\n      }\n      let output;\n      if (this.mergeMode === \"concat\") {\n        output = concatenate([y, yRev]);\n      } else if (this.mergeMode === \"sum\") {\n        output = add2(y, yRev);\n      } else if (this.mergeMode === \"ave\") {\n        output = mul(0.5, add2(y, yRev));\n      } else if (this.mergeMode === \"mul\") {\n        output = mul(y, yRev);\n      } else if (this.mergeMode == null) {\n        output = [y, yRev];\n      }\n      if (this.returnState) {\n        if (this.mergeMode == null) {\n          return output.concat(states);\n        }\n        return [output].concat(states);\n      }\n      return output;\n    });\n  }\n  resetStates(states) {\n    this.forwardLayer.resetStates();\n    this.backwardLayer.resetStates();\n  }\n  build(inputShape) {\n    nameScope(this.forwardLayer.name, () => {\n      this.forwardLayer.build(inputShape);\n    });\n    nameScope(this.backwardLayer.name, () => {\n      this.backwardLayer.build(inputShape);\n    });\n    this.built = true;\n  }\n  computeMask(inputs, mask) {\n    if (Array.isArray(mask)) {\n      mask = mask[0];\n    }\n    let outputMask;\n    if (this.returnSequences) {\n      if (this.mergeMode == null) {\n        outputMask = [mask, mask];\n      } else {\n        outputMask = mask;\n      }\n    } else {\n      if (this.mergeMode == null) {\n        outputMask = [null, null];\n      } else {\n        outputMask = null;\n      }\n    }\n    if (this.returnState) {\n      const states = this.forwardLayer.states;\n      const stateMask = states.map((state) => null);\n      if (Array.isArray(outputMask)) {\n        return outputMask.concat(stateMask).concat(stateMask);\n      } else {\n        return [outputMask].concat(stateMask).concat(stateMask);\n      }\n    } else {\n      return outputMask;\n    }\n  }\n  get trainableWeights() {\n    return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights);\n  }\n  get nonTrainableWeights() {\n    return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights);\n  }\n  setFastWeightInitDuringBuild(value) {\n    super.setFastWeightInitDuringBuild(value);\n    if (this.forwardLayer != null) {\n      this.forwardLayer.setFastWeightInitDuringBuild(value);\n    }\n    if (this.backwardLayer != null) {\n      this.backwardLayer.setFastWeightInitDuringBuild(value);\n    }\n  }\n  getConfig() {\n    const config = {\n      \"mergeMode\": this.mergeMode\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  static fromConfig(cls, config) {\n    const rnnLayer = deserialize(config[\"layer\"]);\n    delete config[\"layer\"];\n    if (config[\"numConstants\"] != null) {\n      throw new NotImplementedError(`Deserialization of a Bidirectional layer with numConstants present is not supported yet.`);\n    }\n    const newConfig = config;\n    newConfig[\"layer\"] = rnnLayer;\n    return new cls(newConfig);\n  }\n};\nBidirectional.className = \"Bidirectional\";\nserialization_exports.registerClass(Bidirectional);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/image_preprocessing.js\nvar Rescaling = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.scale = args.scale;\n    if (args.offset) {\n      this.offset = args.offset;\n    } else {\n      this.offset = 0;\n    }\n  }\n  getConfig() {\n    const config = {\n      \"scale\": this.scale,\n      \"offset\": this.offset\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      if (inputs.dtype !== \"float32\") {\n        inputs = cast2(inputs, \"float32\");\n      }\n      return add2(mul(inputs, this.scale), this.offset);\n    });\n  }\n};\nRescaling.className = \"Rescaling\";\nserialization_exports.registerClass(Rescaling);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/image_resizing.js\nvar INTERPOLATION_KEYS = [\"bilinear\", \"nearest\"];\nvar INTERPOLATION_METHODS = new Set(INTERPOLATION_KEYS);\nvar Resizing = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.height = args.height;\n    this.width = args.width;\n    if (args.interpolation) {\n      if (INTERPOLATION_METHODS.has(args.interpolation)) {\n        this.interpolation = args.interpolation;\n      } else {\n        throw new ValueError(`Invalid interpolation parameter: ${args.interpolation} is not implemented`);\n      }\n    } else {\n      this.interpolation = \"bilinear\";\n    }\n    this.cropToAspectRatio = Boolean(args.cropToAspectRatio);\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    const numChannels = inputShape[2];\n    return [this.height, this.width, numChannels];\n  }\n  getConfig() {\n    const config = {\n      \"height\": this.height,\n      \"width\": this.width,\n      \"interpolation\": this.interpolation,\n      \"cropToAspectRatio\": this.cropToAspectRatio\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      const size = [this.height, this.width];\n      if (this.interpolation === \"bilinear\") {\n        return image.resizeBilinear(inputs, size, !this.cropToAspectRatio);\n      } else if (this.interpolation === \"nearest\") {\n        return image.resizeNearestNeighbor(inputs, size, !this.cropToAspectRatio);\n      } else {\n        throw new Error(`Interpolation is ${this.interpolation} but only ${[...INTERPOLATION_METHODS]} are supported`);\n      }\n    });\n  }\n};\nResizing.className = \"Resizing\";\nserialization_exports.registerClass(Resizing);\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/preprocessing_utils.js\nfunction encodeCategoricalInputs(inputs, outputMode, depth, weights) {\n  let input2 = getExactlyOneTensor(inputs);\n  if (input2.dtype !== \"int32\") {\n    input2 = cast2(input2, \"int32\");\n  }\n  if (outputMode === \"int\") {\n    return input2;\n  }\n  const originalShape = input2.shape;\n  if (input2.rank === 0) {\n    input2 = expandDims(input2, -1);\n  }\n  if (outputMode === \"oneHot\") {\n    if (input2.shape[input2.shape.length - 1] !== 1) {\n      input2 = expandDims(input2, -1);\n    }\n  }\n  if (input2.rank > 2) {\n    throw new ValueError(`When outputMode is not int, maximum output rank is 2 Received outputMode ${outputMode} and input shape ${originalShape} which would result in output rank ${input2.rank}.`);\n  }\n  const binaryOutput = [\"multiHot\", \"oneHot\"].includes(outputMode);\n  const denseBincountInput = input2;\n  let binCounts;\n  if (typeof weights !== \"undefined\" && outputMode === \"count\") {\n    binCounts = denseBincount(denseBincountInput, weights, depth, binaryOutput);\n  } else {\n    binCounts = denseBincount(denseBincountInput, [], depth, binaryOutput);\n  }\n  if (outputMode !== \"tfIdf\") {\n    return binCounts;\n  }\n  if (weights) {\n    return mul(binCounts, weights);\n  } else {\n    throw new ValueError(`When outputMode is 'tfIdf', weights must be provided.`);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/category_encoding.js\nvar CategoryEncoding = class extends Layer {\n  constructor(args) {\n    super(args);\n    this.numTokens = args.numTokens;\n    if (args.outputMode) {\n      this.outputMode = args.outputMode;\n    } else {\n      this.outputMode = \"multiHot\";\n    }\n  }\n  getConfig() {\n    const config = {\n      \"numTokens\": this.numTokens,\n      \"outputMode\": this.outputMode\n    };\n    const baseConfig = super.getConfig();\n    Object.assign(config, baseConfig);\n    return config;\n  }\n  computeOutputShape(inputShape) {\n    inputShape = getExactlyOneShape(inputShape);\n    if (inputShape == null) {\n      return [this.numTokens];\n    }\n    if (this.outputMode === \"oneHot\" && inputShape[inputShape.length - 1] !== 1) {\n      inputShape.push(this.numTokens);\n      return inputShape;\n    }\n    inputShape[inputShape.length - 1] = this.numTokens;\n    return inputShape;\n  }\n  call(inputs, kwargs) {\n    return tidy(() => {\n      inputs = getExactlyOneTensor(inputs);\n      if (inputs.dtype !== \"int32\") {\n        inputs = cast2(inputs, \"int32\");\n      }\n      let countWeights;\n      if (typeof kwargs[\"countWeights\"] !== \"undefined\") {\n        if (this.outputMode !== \"count\") {\n          throw new ValueError(`countWeights is not used when outputMode !== count.\n              Received countWeights=${kwargs[\"countWeights\"]}`);\n        }\n        countWeights = getExactlyOneTensor(kwargs[\"countWeights\"]);\n      }\n      const maxValue = max(inputs);\n      const minValue = min(inputs);\n      const greaterEqualMax = greater(this.numTokens, maxValue).bufferSync().get(0);\n      const greaterMin = greaterEqual(minValue, 0).bufferSync().get(0);\n      if (!(greaterEqualMax && greaterMin)) {\n        throw new ValueError(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);\n      }\n      return encodeCategoricalInputs(inputs, this.outputMode, this.numTokens, countWeights);\n    });\n  }\n};\nCategoryEncoding.className = \"CategoryEncoding\";\nserialization_exports.registerClass(CategoryEncoding);\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_layers.js\nfunction inputLayer(args) {\n  return new InputLayer(args);\n}\nfunction elu3(args) {\n  return new ELU(args);\n}\nfunction reLU(args) {\n  return new ReLU(args);\n}\nfunction leakyReLU(args) {\n  return new LeakyReLU(args);\n}\nfunction prelu2(args) {\n  return new PReLU(args);\n}\nfunction softmax2(args) {\n  return new Softmax3(args);\n}\nfunction thresholdedReLU(args) {\n  return new ThresholdedReLU(args);\n}\nfunction conv1d2(args) {\n  return new Conv1D(args);\n}\nfunction conv2d3(args) {\n  return new Conv2D2(args);\n}\nfunction conv2dTranspose2(args) {\n  return new Conv2DTranspose(args);\n}\nfunction conv3d2(args) {\n  return new Conv3D2(args);\n}\nfunction conv3dTranspose2(args) {\n  return new Conv3DTranspose(args);\n}\nfunction separableConv2d2(args) {\n  return new SeparableConv2D(args);\n}\nfunction cropping2D(args) {\n  return new Cropping2D(args);\n}\nfunction upSampling2d(args) {\n  return new UpSampling2D(args);\n}\nfunction depthwiseConv2d4(args) {\n  return new DepthwiseConv2D(args);\n}\nfunction activation(args) {\n  return new Activation2(args);\n}\nfunction dense(args) {\n  return new Dense(args);\n}\nfunction dropout3(args) {\n  return new Dropout(args);\n}\nfunction spatialDropout1d(args) {\n  return new SpatialDropout1D(args);\n}\nfunction flatten3(args) {\n  return new Flatten(args);\n}\nfunction repeatVector(args) {\n  return new RepeatVector(args);\n}\nfunction reshape2(args) {\n  return new Reshape2(args);\n}\nfunction permute(args) {\n  return new Permute(args);\n}\nfunction embedding(args) {\n  return new Embedding(args);\n}\nfunction add3(args) {\n  return new Add2(args);\n}\nfunction average(args) {\n  return new Average(args);\n}\nfunction concatenate2(args) {\n  return new Concatenate(args);\n}\nfunction maximum2(args) {\n  return new Maximum2(args);\n}\nfunction minimum2(args) {\n  return new Minimum2(args);\n}\nfunction multiply(args) {\n  return new Multiply2(args);\n}\nfunction dot3(args) {\n  return new Dot(args);\n}\nfunction batchNormalization2(args) {\n  return new BatchNormalization(args);\n}\nfunction layerNormalization(args) {\n  return new LayerNormalization(args);\n}\nfunction zeroPadding2d(args) {\n  return new ZeroPadding2D(args);\n}\nfunction averagePooling1d(args) {\n  return new AveragePooling1D(args);\n}\nfunction avgPool1d(args) {\n  return averagePooling1d(args);\n}\nfunction avgPooling1d(args) {\n  return averagePooling1d(args);\n}\nfunction averagePooling2d(args) {\n  return new AveragePooling2D(args);\n}\nfunction avgPool2d(args) {\n  return averagePooling2d(args);\n}\nfunction avgPooling2d(args) {\n  return averagePooling2d(args);\n}\nfunction averagePooling3d(args) {\n  return new AveragePooling3D(args);\n}\nfunction avgPool3d2(args) {\n  return averagePooling3d(args);\n}\nfunction avgPooling3d(args) {\n  return averagePooling3d(args);\n}\nfunction globalAveragePooling1d(args) {\n  return new GlobalAveragePooling1D(args);\n}\nfunction globalAveragePooling2d(args) {\n  return new GlobalAveragePooling2D(args);\n}\nfunction globalMaxPooling1d(args) {\n  return new GlobalMaxPooling1D(args);\n}\nfunction globalMaxPooling2d(args) {\n  return new GlobalMaxPooling2D(args);\n}\nfunction maxPooling1d(args) {\n  return new MaxPooling1D(args);\n}\nfunction maxPooling2d(args) {\n  return new MaxPooling2D(args);\n}\nfunction maxPooling3d(args) {\n  return new MaxPooling3D(args);\n}\nfunction gru(args) {\n  return new GRU(args);\n}\nfunction gruCell(args) {\n  return new GRUCell(args);\n}\nfunction lstm(args) {\n  return new LSTM(args);\n}\nfunction lstmCell(args) {\n  return new LSTMCell(args);\n}\nfunction simpleRNN(args) {\n  return new SimpleRNN(args);\n}\nfunction simpleRNNCell(args) {\n  return new SimpleRNNCell(args);\n}\nfunction convLstm2d(args) {\n  return new ConvLSTM2D(args);\n}\nfunction convLstm2dCell(args) {\n  return new ConvLSTM2DCell(args);\n}\nfunction rnn2(args) {\n  return new RNN(args);\n}\nfunction stackedRNNCells(args) {\n  return new StackedRNNCells(args);\n}\nfunction bidirectional(args) {\n  return new Bidirectional(args);\n}\nfunction timeDistributed(args) {\n  return new TimeDistributed(args);\n}\nvar globalMaxPool1d = globalMaxPooling1d;\nvar globalMaxPool2d = globalMaxPooling2d;\nvar maxPool1d = maxPooling1d;\nvar maxPool2d = maxPooling2d;\nfunction gaussianNoise(args) {\n  return new GaussianNoise(args);\n}\nfunction gaussianDropout(args) {\n  return new GaussianDropout(args);\n}\nfunction alphaDropout(args) {\n  return new AlphaDropout(args);\n}\nfunction masking(args) {\n  return new Masking(args);\n}\nfunction rescaling(args) {\n  return new Rescaling(args);\n}\nfunction resizing(args) {\n  return new Resizing(args);\n}\nfunction categoryEncoding(args) {\n  return new CategoryEncoding(args);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_metrics.js\nvar exports_metrics_exports = {};\n__export(exports_metrics_exports, {\n  MAPE: () => MAPE2,\n  MSE: () => MSE2,\n  binaryAccuracy: () => binaryAccuracy2,\n  binaryCrossentropy: () => binaryCrossentropy3,\n  categoricalAccuracy: () => categoricalAccuracy2,\n  categoricalCrossentropy: () => categoricalCrossentropy3,\n  cosineProximity: () => cosineProximity2,\n  mape: () => mape2,\n  meanAbsoluteError: () => meanAbsoluteError2,\n  meanAbsolutePercentageError: () => meanAbsolutePercentageError2,\n  meanSquaredError: () => meanSquaredError3,\n  mse: () => mse2,\n  precision: () => precision2,\n  recall: () => recall2,\n  sparseCategoricalAccuracy: () => sparseCategoricalAccuracy2\n});\nfunction binaryAccuracy2(yTrue, yPred) {\n  return binaryAccuracy(yTrue, yPred);\n}\nfunction binaryCrossentropy3(yTrue, yPred) {\n  return binaryCrossentropy2(yTrue, yPred);\n}\nfunction sparseCategoricalAccuracy2(yTrue, yPred) {\n  return sparseCategoricalAccuracy(yTrue, yPred);\n}\nfunction categoricalAccuracy2(yTrue, yPred) {\n  return categoricalAccuracy(yTrue, yPred);\n}\nfunction categoricalCrossentropy3(yTrue, yPred) {\n  return categoricalCrossentropy2(yTrue, yPred);\n}\nfunction precision2(yTrue, yPred) {\n  return precision(yTrue, yPred);\n}\nfunction recall2(yTrue, yPred) {\n  return recall(yTrue, yPred);\n}\nfunction cosineProximity2(yTrue, yPred) {\n  return cosineProximity(yTrue, yPred);\n}\nfunction meanAbsoluteError2(yTrue, yPred) {\n  return meanAbsoluteError(yTrue, yPred);\n}\nfunction meanAbsolutePercentageError2(yTrue, yPred) {\n  return meanAbsolutePercentageError(yTrue, yPred);\n}\nfunction MAPE2(yTrue, yPred) {\n  return meanAbsolutePercentageError(yTrue, yPred);\n}\nfunction mape2(yTrue, yPred) {\n  return meanAbsolutePercentageError(yTrue, yPred);\n}\nfunction meanSquaredError3(yTrue, yPred) {\n  return meanSquaredError2(yTrue, yPred);\n}\nfunction MSE2(yTrue, yPred) {\n  return meanSquaredError2(yTrue, yPred);\n}\nfunction mse2(yTrue, yPred) {\n  return meanSquaredError2(yTrue, yPred);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_models.js\nvar exports_models_exports = {};\n__export(exports_models_exports, {\n  modelFromJSON: () => modelFromJSON\n});\n\n// node_modules/@tensorflow/tfjs-layers/dist/exports_regularizers.js\nvar exports_regularizers_exports = {};\n__export(exports_regularizers_exports, {\n  l1: () => l12,\n  l1l2: () => l1l2,\n  l2: () => l22\n});\nfunction l1l2(config) {\n  return new L1L2(config);\n}\nfunction l12(config) {\n  return l1(config);\n}\nfunction l22(config) {\n  return l2(config);\n}\n\n// node_modules/@tensorflow/tfjs-layers/dist/callbacks.js\nvar Callback = class extends BaseCallback {\n  constructor() {\n    super(...arguments);\n    this.model = null;\n  }\n  setModel(model2) {\n    if (!(model2 instanceof LayersModel)) {\n      throw new Error(\"model must be a LayersModel, not some other Container\");\n    }\n    this.model = model2;\n  }\n};\nfunction less2(currVal, prevVal) {\n  return currVal < prevVal;\n}\nfunction greater2(currVal, prevVal) {\n  return currVal > prevVal;\n}\nvar EarlyStopping = class extends Callback {\n  constructor(args) {\n    super();\n    if (args == null) {\n      args = {};\n    }\n    if (args.restoreBestWeights) {\n      throw new NotImplementedError(\"restoreBestWeights = True is not implemented in EarlyStopping yet.\");\n    }\n    this.monitor = args.monitor || \"val_loss\";\n    this.minDelta = Math.abs(args.minDelta || 0);\n    this.patience = args.patience || 0;\n    this.verbose = args.verbose || 0;\n    this.mode = args.mode || \"auto\";\n    this.baseline = args.baseline;\n    if ([\"auto\", \"min\", \"max\"].indexOf(this.mode) === -1) {\n      console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`);\n      this.mode = \"auto\";\n    }\n    if (this.mode === \"min\") {\n      this.monitorFunc = less2;\n    } else if (this.mode === \"max\") {\n      this.monitorFunc = greater2;\n    } else {\n      if (this.monitor.indexOf(\"acc\") !== -1) {\n        this.monitorFunc = greater2;\n      } else {\n        this.monitorFunc = less2;\n      }\n    }\n    if (this.monitorFunc === less2) {\n      this.minDelta *= -1;\n    }\n  }\n  async onTrainBegin(logs) {\n    this.wait = 0;\n    this.stoppedEpoch = 0;\n    if (this.baseline != null) {\n      this.best = this.baseline;\n    } else {\n      this.best = this.monitorFunc === less2 ? Infinity : -Infinity;\n    }\n  }\n  async onEpochEnd(epoch, logs) {\n    await resolveScalarsInLogs(logs);\n    const current = this.getMonitorValue(logs);\n    if (current == null) {\n      return;\n    }\n    if (this.monitorFunc(current - this.minDelta, this.best)) {\n      this.best = current;\n      this.wait = 0;\n    } else {\n      this.wait++;\n      if (this.wait >= this.patience) {\n        this.stoppedEpoch = epoch;\n        this.model.stopTraining = true;\n      }\n    }\n  }\n  async onTrainEnd(logs) {\n    if (this.stoppedEpoch > 0 && this.verbose) {\n      console.log(`Epoch ${this.stoppedEpoch}: early stopping.`);\n    }\n  }\n  getMonitorValue(logs) {\n    if (logs == null) {\n      logs = {};\n    }\n    const monitorValue = logs[this.monitor];\n    if (monitorValue == null) {\n      console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(logs)}`);\n    }\n    return monitorValue;\n  }\n};\nfunction earlyStopping(args) {\n  return new EarlyStopping(args);\n}\nvar callbacks = { earlyStopping };\n\n// node_modules/@tensorflow/tfjs-converter/dist/flags.js\nvar ENV4 = env();\nENV4.registerFlag(\"KEEP_INTERMEDIATE_TENSORS\", () => false, (debugValue) => {\n  if (debugValue) {\n    console.warn(\"Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.\");\n  }\n});\n\n// node_modules/@tensorflow/tfjs-converter/dist/data/compiled_api.js\nvar DataType;\n(function(DataType2) {\n  DataType2[DataType2[\"DT_INVALID\"] = 0] = \"DT_INVALID\";\n  DataType2[DataType2[\"DT_FLOAT\"] = 1] = \"DT_FLOAT\";\n  DataType2[DataType2[\"DT_DOUBLE\"] = 2] = \"DT_DOUBLE\";\n  DataType2[DataType2[\"DT_INT32\"] = 3] = \"DT_INT32\";\n  DataType2[DataType2[\"DT_UINT8\"] = 4] = \"DT_UINT8\";\n  DataType2[DataType2[\"DT_INT16\"] = 5] = \"DT_INT16\";\n  DataType2[DataType2[\"DT_INT8\"] = 6] = \"DT_INT8\";\n  DataType2[DataType2[\"DT_STRING\"] = 7] = \"DT_STRING\";\n  DataType2[DataType2[\"DT_COMPLEX64\"] = 8] = \"DT_COMPLEX64\";\n  DataType2[DataType2[\"DT_INT64\"] = 9] = \"DT_INT64\";\n  DataType2[DataType2[\"DT_BOOL\"] = 10] = \"DT_BOOL\";\n  DataType2[DataType2[\"DT_QINT8\"] = 11] = \"DT_QINT8\";\n  DataType2[DataType2[\"DT_QUINT8\"] = 12] = \"DT_QUINT8\";\n  DataType2[DataType2[\"DT_QINT32\"] = 13] = \"DT_QINT32\";\n  DataType2[DataType2[\"DT_BFLOAT16\"] = 14] = \"DT_BFLOAT16\";\n  DataType2[DataType2[\"DT_QINT16\"] = 15] = \"DT_QINT16\";\n  DataType2[DataType2[\"DT_QUINT16\"] = 16] = \"DT_QUINT16\";\n  DataType2[DataType2[\"DT_UINT16\"] = 17] = \"DT_UINT16\";\n  DataType2[DataType2[\"DT_COMPLEX128\"] = 18] = \"DT_COMPLEX128\";\n  DataType2[DataType2[\"DT_HALF\"] = 19] = \"DT_HALF\";\n  DataType2[DataType2[\"DT_RESOURCE\"] = 20] = \"DT_RESOURCE\";\n  DataType2[DataType2[\"DT_VARIANT\"] = 21] = \"DT_VARIANT\";\n  DataType2[DataType2[\"DT_UINT32\"] = 22] = \"DT_UINT32\";\n  DataType2[DataType2[\"DT_UINT64\"] = 23] = \"DT_UINT64\";\n  DataType2[DataType2[\"DT_FLOAT_REF\"] = 101] = \"DT_FLOAT_REF\";\n  DataType2[DataType2[\"DT_DOUBLE_REF\"] = 102] = \"DT_DOUBLE_REF\";\n  DataType2[DataType2[\"DT_INT32_REF\"] = 103] = \"DT_INT32_REF\";\n  DataType2[DataType2[\"DT_UINT8_REF\"] = 104] = \"DT_UINT8_REF\";\n  DataType2[DataType2[\"DT_INT16_REF\"] = 105] = \"DT_INT16_REF\";\n  DataType2[DataType2[\"DT_INT8_REF\"] = 106] = \"DT_INT8_REF\";\n  DataType2[DataType2[\"DT_STRING_REF\"] = 107] = \"DT_STRING_REF\";\n  DataType2[DataType2[\"DT_COMPLEX64_REF\"] = 108] = \"DT_COMPLEX64_REF\";\n  DataType2[DataType2[\"DT_INT64_REF\"] = 109] = \"DT_INT64_REF\";\n  DataType2[DataType2[\"DT_BOOL_REF\"] = 110] = \"DT_BOOL_REF\";\n  DataType2[DataType2[\"DT_QINT8_REF\"] = 111] = \"DT_QINT8_REF\";\n  DataType2[DataType2[\"DT_QUINT8_REF\"] = 112] = \"DT_QUINT8_REF\";\n  DataType2[DataType2[\"DT_QINT32_REF\"] = 113] = \"DT_QINT32_REF\";\n  DataType2[DataType2[\"DT_BFLOAT16_REF\"] = 114] = \"DT_BFLOAT16_REF\";\n  DataType2[DataType2[\"DT_QINT16_REF\"] = 115] = \"DT_QINT16_REF\";\n  DataType2[DataType2[\"DT_QUINT16_REF\"] = 116] = \"DT_QUINT16_REF\";\n  DataType2[DataType2[\"DT_UINT16_REF\"] = 117] = \"DT_UINT16_REF\";\n  DataType2[DataType2[\"DT_COMPLEX128_REF\"] = 118] = \"DT_COMPLEX128_REF\";\n  DataType2[DataType2[\"DT_HALF_REF\"] = 119] = \"DT_HALF_REF\";\n  DataType2[DataType2[\"DT_RESOURCE_REF\"] = 120] = \"DT_RESOURCE_REF\";\n  DataType2[DataType2[\"DT_VARIANT_REF\"] = 121] = \"DT_VARIANT_REF\";\n  DataType2[DataType2[\"DT_UINT32_REF\"] = 122] = \"DT_UINT32_REF\";\n  DataType2[DataType2[\"DT_UINT64_REF\"] = 123] = \"DT_UINT64_REF\";\n})(DataType || (DataType = {}));\nvar SaverDef;\n(function(SaverDef2) {\n  let CheckpointFormatVersion;\n  (function(CheckpointFormatVersion2) {\n    CheckpointFormatVersion2[CheckpointFormatVersion2[\"LEGACY\"] = 0] = \"LEGACY\";\n    CheckpointFormatVersion2[CheckpointFormatVersion2[\"V1\"] = 1] = \"V1\";\n    CheckpointFormatVersion2[CheckpointFormatVersion2[\"V2\"] = 2] = \"V2\";\n  })(CheckpointFormatVersion = SaverDef2.CheckpointFormatVersion || (SaverDef2.CheckpointFormatVersion = {}));\n})(SaverDef || (SaverDef = {}));\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/custom_op/register.js\nvar CUSTOM_OPS = {};\nfunction registerOp(name, opFunc) {\n  const opMapper = {\n    tfOpName: name,\n    category: \"custom\",\n    inputs: [],\n    attrs: [],\n    customExecutor: opFunc\n  };\n  CUSTOM_OPS[name] = opMapper;\n}\nfunction getRegisteredOp(name) {\n  return CUSTOM_OPS[name];\n}\nfunction deregisterOp(name) {\n  delete CUSTOM_OPS[name];\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/utils.js\nfunction getParamValue(paramName, node, tensorMap, context, resourceManager) {\n  const inputParam = node.inputParams[paramName];\n  if (inputParam && inputParam.inputIndexStart !== void 0) {\n    const start = inputParam.inputIndexStart;\n    const end = inputParam.inputIndexEnd === 0 ? void 0 : inputParam.inputIndexEnd === void 0 ? start + 1 : inputParam.inputIndexEnd;\n    if (inputParam.type === \"tensor\") {\n      return getTensor(node.inputNames[inputParam.inputIndexStart], tensorMap, context, resourceManager);\n    }\n    if (inputParam.type === \"tensors\") {\n      const inputs = node.inputNames.slice(start, end);\n      return inputs.map((name) => getTensor(name, tensorMap, context, resourceManager));\n    }\n    const tensor2 = getTensor(node.inputNames.slice(start)[0], tensorMap, context, resourceManager);\n    const data = tensor2.dataSync();\n    return inputParam.type === \"number\" ? data[0] : util_exports.toNestedArray(tensor2.shape, data);\n  }\n  const attrParam = node.attrParams[paramName];\n  return attrParam && attrParam.value;\n}\nfunction getTensor(name, tensorsMap, context, resourceManager) {\n  const [nodeName, index] = parseNodeName(name);\n  if (resourceManager != null) {\n    const tensor2 = resourceManager.getHashTableHandleByName(nodeName);\n    if (tensor2 != null) {\n      return tensor2;\n    }\n  }\n  const contextId = context.currentContextIds.find((contextId2) => {\n    return !!tensorsMap[getNodeNameWithContextId(nodeName, contextId2)];\n  });\n  return contextId !== void 0 ? tensorsMap[getNodeNameWithContextId(nodeName, contextId)][index] : void 0;\n}\nfunction getTensorsForCurrentContenxt(name, tensorsMap, context) {\n  return tensorsMap[getNodeNameWithContextId(name, context.currentContextId)];\n}\nfunction getNodeNameAndIndex(inputName, context) {\n  const [nodeName, index, outputName] = parseNodeName(inputName);\n  return [\n    getNodeNameWithContextId(nodeName, context && context.currentContextId),\n    index,\n    outputName\n  ];\n}\nfunction getNodeNameWithContextId(name, contextId) {\n  return !!contextId ? `${name}-${contextId}` : name;\n}\nfunction parseNodeName(name) {\n  const parts = name.split(\":\");\n  if (parts.length === 1) {\n    return [name, 0, void 0];\n  }\n  const nodeName = parts[0];\n  const outputName = parts.length === 3 ? parts[1] : void 0;\n  const index = Number(parts[parts.length - 1]);\n  return [nodeName, index, outputName];\n}\nfunction getPadding(node, tensorMap, context) {\n  let pad3 = getParamValue(\"pad\", node, tensorMap, context);\n  if (pad3 === \"explicit\") {\n    pad3 = getParamValue(\"explicitPaddings\", node, tensorMap, context);\n    const explicitPadding = [[0, 0], [0, 0], [0, 0], [0, 0]];\n    for (let i = 0; i < 4; i++) {\n      explicitPadding[i][0] = pad3[i * 2];\n      explicitPadding[i][1] = pad3[i * 2 + 1];\n    }\n    return explicitPadding;\n  }\n  return pad3;\n}\nfunction cloneTensor(tensor2) {\n  return tensor2.kept ? tensor2 : clone(tensor2);\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/arithmetic.js\nvar arithmetic_exports = {};\n__export(arithmetic_exports, {\n  json: () => json\n});\nvar json = [\n  {\n    \"tfOpName\": \"Add\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"AddV2\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"AddN\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"BiasAdd\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sub\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"RealDiv\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Div\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"DivNoNan\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FloorDiv\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Mul\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Maximum\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Minimum\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Pow\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SquaredDifference\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Mod\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FloorMod\",\n    \"category\": \"arithmetic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/basic_math.js\nvar basic_math_exports = {};\n__export(basic_math_exports, {\n  json: () => json2\n});\nvar json2 = [\n  {\n    \"tfOpName\": \"Abs\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Acos\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Asin\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Atan\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Atan2\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"y\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Ceil\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ClipByValue\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"clipValueMin\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"clipValueMax\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Complex\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"real\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"imag\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ComplexAbs\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Cos\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Cosh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Elu\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Exp\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Floor\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Log\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Imag\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"outputType\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Neg\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Real\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"outputType\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Prelu\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"alpha\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Relu\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Relu6\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Selu\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sigmoid\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sin\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sinh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sqrt\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Rsqrt\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Square\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Tan\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Tanh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sign\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Round\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Expm1\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Log1p\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Reciprocal\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Softplus\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Asinh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Acosh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Atanh\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Erf\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Prod\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axes\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LeakyRelu\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"alpha\",\n        \"name\": \"alpha\",\n        \"type\": \"number\",\n        \"defaultValue\": 0.2\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"IsNan\",\n    \"category\": \"basic_math\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/control.js\nvar control_exports = {};\n__export(control_exports, {\n  json: () => json3\n});\nvar json3 = [\n  {\n    \"tfOpName\": \"EmptyTensorList\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"maxNumElements\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LoopCond\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"pred\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Switch\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"data\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"pred\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Merge\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Enter\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"frame_name\",\n        \"name\": \"frameName\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"is_constant\",\n        \"name\": \"isConstant\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Exit\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NextIteration\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"size\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"element_shape\",\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"tfName\": \"dynamic_size\",\n        \"name\": \"dynamicSize\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"clear_after_read\",\n        \"name\": \"clearAfterRead\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"identical_element_shapes\",\n        \"name\": \"identicalElementShapes\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"tensor_array_name\",\n        \"name\": \"name\",\n        \"type\": \"string\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayWriteV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"index\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayReadV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"index\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayGatherV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"element_shape\",\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayScatterV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayConcatV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"element_shape_except0\",\n        \"name\": \"elementShapeExcept0\",\n        \"type\": \"shape\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArraySplitV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"lengths\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArraySizeV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"flowIn\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorArrayCloseV3\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorArrayId\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"StatelessIf\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"cond\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"then_branch\",\n        \"name\": \"thenBranch\",\n        \"type\": \"func\"\n      },\n      {\n        \"tfName\": \"else_branch\",\n        \"name\": \"elseBranch\",\n        \"type\": \"func\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"If\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"cond\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"then_branch\",\n        \"name\": \"thenBranch\",\n        \"type\": \"func\"\n      },\n      {\n        \"tfName\": \"else_branch\",\n        \"name\": \"elseBranch\",\n        \"type\": \"func\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"StatelessWhile\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"cond\",\n        \"name\": \"cond\",\n        \"type\": \"func\"\n      },\n      {\n        \"tfName\": \"body\",\n        \"name\": \"body\",\n        \"type\": \"func\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"While\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"cond\",\n        \"name\": \"cond\",\n        \"type\": \"func\"\n      },\n      {\n        \"tfName\": \"body\",\n        \"name\": \"body\",\n        \"type\": \"func\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListScatter\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListScatterV2\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"numElements\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListGather\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListGetItem\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"index\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListSetItem\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"index\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListReserve\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"numElements\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListFromTensor\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListStack\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"num_elements\",\n        \"name\": \"numElements\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListSplit\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"lengths\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListConcat\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_shape\",\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListConcatV2\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_shape\",\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListPopBack\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"elementShape\",\n        \"type\": \"shape\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListPushBack\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"element_dtype\",\n        \"name\": \"elementDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListLength\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TensorListResize\",\n    \"category\": \"control\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensorListId\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"size\",\n        \"type\": \"number\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/convolution.js\nvar convolution_exports = {};\n__export(convolution_exports, {\n  json: () => json4\n});\nvar json4 = [\n  {\n    \"tfOpName\": \"AvgPool\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"ksize\",\n        \"name\": \"kernelSize\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"MaxPool\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"ksize\",\n        \"name\": \"kernelSize\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": [],\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"MaxPoolWithArgmax\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"ksize\",\n        \"name\": \"kernelSize\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"include_batch_in_index\",\n        \"name\": \"includeBatchInIndex\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"AvgPool3D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"ksize\",\n        \"name\": \"kernelSize\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"MaxPool3D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"ksize\",\n        \"name\": \"kernelSize\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Conv1D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"stride\",\n        \"name\": \"stride\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NWC\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"dilation\",\n        \"name\": \"dilation\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Conv2D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"useCudnnOnGpu\",\n        \"name\": \"useCudnnOnGpu\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"_FusedConv2D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"num_args\",\n        \"name\": \"numArgs\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"use_cudnn_on_gpu\",\n        \"name\": \"useCudnnOnGpu\",\n        \"type\": \"bool\",\n        \"defaultValue\": true\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\",\n        \"defaultValue\": [\n          1,\n          1,\n          1,\n          1\n        ]\n      },\n      {\n        \"tfName\": \"fused_ops\",\n        \"name\": \"fusedOps\",\n        \"type\": \"string[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"epsilon\",\n        \"name\": \"epsilon\",\n        \"type\": \"number\",\n        \"defaultValue\": 1e-4\n      },\n      {\n        \"tfName\": \"leakyrelu_alpha\",\n        \"name\": \"leakyreluAlpha\",\n        \"type\": \"number\",\n        \"defaultValue\": 0.2\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Conv2DBackpropInput\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 2,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 0,\n        \"name\": \"outputShape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"DepthwiseConv2d\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"input\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"DepthwiseConv2dNative\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"input\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FusedDepthwiseConv2dNative\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"num_args\",\n        \"name\": \"numArgs\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\",\n        \"defaultValue\": [\n          1,\n          1,\n          1,\n          1\n        ]\n      },\n      {\n        \"tfName\": \"fused_ops\",\n        \"name\": \"fusedOps\",\n        \"type\": \"string[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"explicit_paddings\",\n        \"name\": \"explicitPaddings\",\n        \"type\": \"number[]\",\n        \"defaultValue\": []\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Conv3D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"defaultValue\": \"NHWC\"\n      },\n      {\n        \"tfName\": \"dilations\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Dilation2D\",\n    \"category\": \"convolution\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"filter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"strides\",\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"rates\",\n        \"name\": \"dilations\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"padding\",\n        \"name\": \"pad\",\n        \"type\": \"string\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/creation.js\nvar creation_exports = {};\n__export(creation_exports, {\n  json: () => json5\n});\nvar json5 = [\n  {\n    \"tfOpName\": \"Fill\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"value\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LinSpace\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"start\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"stop\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"num\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"OneHot\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"depth\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"onValue\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      },\n      {\n        \"start\": 3,\n        \"name\": \"offValue\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"axis\",\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Ones\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"OnesLike\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"RandomStandardNormal\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"seed\",\n        \"name\": \"seed\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"seed2\",\n        \"name\": \"seed2\",\n        \"type\": \"number\",\n        \"defaultValue\": 0,\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"T\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"RandomUniform\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"minval\",\n        \"name\": \"minval\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"maxval\",\n        \"name\": \"maxval\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      },\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"seed\",\n        \"name\": \"seed\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"seed2\",\n        \"name\": \"seed2\",\n        \"type\": \"number\",\n        \"defaultValue\": 0,\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"T\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Range\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"start\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"stop\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"step\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"Tidx\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TruncatedNormal\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"means\",\n        \"name\": \"mean\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"stddev\",\n        \"name\": \"stdDev\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      },\n      {\n        \"tfName\": \"seed\",\n        \"name\": \"seed\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"seed2\",\n        \"name\": \"seed2\",\n        \"type\": \"number\",\n        \"defaultValue\": 0,\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"T\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Zeros\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ZerosLike\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Multinomial\",\n    \"category\": \"creation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"logits\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"numSamples\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"seed\",\n        \"name\": \"seed\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"seed2\",\n        \"name\": \"seed2\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"output_dtype\",\n        \"name\": \"output_dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/dynamic.js\nvar dynamic_exports = {};\n__export(dynamic_exports, {\n  json: () => json6\n});\nvar json6 = [\n  {\n    \"tfOpName\": \"NonMaxSuppressionV2\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"boxes\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scores\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"maxOutputSize\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"iouThreshold\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NonMaxSuppressionV3\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"boxes\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scores\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"maxOutputSize\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"iouThreshold\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"scoreThreshold\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NonMaxSuppressionV4\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"boxes\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scores\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"maxOutputSize\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"iouThreshold\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"scoreThreshold\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"T_threshold\",\n        \"name\": \"threshold\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"pad_to_max_output_size\",\n        \"name\": \"padToMaxOutputSize\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NonMaxSuppressionV5\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"boxes\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scores\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"maxOutputSize\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"iouThreshold\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"scoreThreshold\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 5,\n        \"name\": \"softNmsSigma\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Where\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"condition\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ListDiff\",\n    \"category\": \"dynamic\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"y\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/evaluation.js\nvar evaluation_exports = {};\n__export(evaluation_exports, {\n  json: () => json7\n});\nvar json7 = [\n  {\n    \"tfOpName\": \"LowerBound\",\n    \"category\": \"evaluation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"sortedSequence\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"TopKV2\",\n    \"category\": \"evaluation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"k\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"sorted\",\n        \"name\": \"sorted\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"UpperBound\",\n    \"category\": \"evaluation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"sortedSequence\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Unique\",\n    \"category\": \"evaluation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"UniqueV2\",\n    \"category\": \"evaluation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/graph.js\nvar graph_exports = {};\n__export(graph_exports, {\n  json: () => json8\n});\nvar json8 = [\n  {\n    \"tfOpName\": \"PlaceholderWithDefault\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"default\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"shape\",\n        \"name\": \"shape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Placeholder\",\n    \"category\": \"graph\",\n    \"attrs\": [\n      {\n        \"tfName\": \"shape\",\n        \"name\": \"shape\",\n        \"type\": \"shape\"\n      },\n      {\n        \"tfName\": \"dtype\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Const\",\n    \"category\": \"graph\"\n  },\n  {\n    \"tfOpName\": \"Identity\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"IdentityN\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensors\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Snapshot\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Rank\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Size\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Shape\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ShapeN\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensors\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Print\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"data\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"message\",\n        \"name\": \"message\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"first_n\",\n        \"name\": \"firstN\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"summarize\",\n        \"name\": \"summarize\",\n        \"type\": \"number\",\n        \"defaultValue\": 3\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NoOp\",\n    \"category\": \"graph\",\n    \"inputs\": []\n  },\n  {\n    \"tfOpName\": \"StopGradient\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FakeQuantWithMinMaxVars\",\n    \"category\": \"graph\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"min\",\n        \"name\": \"min\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"max\",\n        \"name\": \"max\",\n        \"type\": \"number\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/hash_table.js\nvar hash_table_exports = {};\n__export(hash_table_exports, {\n  json: () => json9\n});\nvar json9 = [\n  {\n    \"tfOpName\": \"HashTable\",\n    \"category\": \"hash_table\",\n    \"inputs\": [],\n    \"attrs\": [\n      {\n        \"tfName\": \"shared_name\",\n        \"name\": \"sharedName\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"use_node_name_sharing\",\n        \"name\": \"useNodeNameSharing\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"key_dtype\",\n        \"name\": \"keyDType\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"value_dtype\",\n        \"name\": \"valueDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"HashTableV2\",\n    \"category\": \"hash_table\",\n    \"inputs\": [],\n    \"attrs\": [\n      {\n        \"tfName\": \"shared_name\",\n        \"name\": \"sharedName\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"use_node_name_sharing\",\n        \"name\": \"useNodeNameSharing\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"key_dtype\",\n        \"name\": \"keyDType\",\n        \"type\": \"dtype\"\n      },\n      {\n        \"tfName\": \"value_dtype\",\n        \"name\": \"valueDType\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableImport\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"keys\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"Tin\",\n        \"name\": \"tIn\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"tOut\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableImportV2\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"keys\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"Tin\",\n        \"name\": \"tIn\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"tOut\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableFind\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"keys\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"defaultValue\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"Tin\",\n        \"name\": \"tIn\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"tOut\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableFindV2\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"keys\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"defaultValue\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"Tin\",\n        \"name\": \"tIn\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"Tout\",\n        \"name\": \"tOut\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableSize\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LookupTableSizeV2\",\n    \"category\": \"hash_table\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tableHandle\",\n        \"type\": \"tensor\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/image.js\nvar image_exports = {};\n__export(image_exports, {\n  json: () => json10\n});\nvar json10 = [\n  {\n    \"tfOpName\": \"ResizeBilinear\",\n    \"category\": \"image\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"images\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"size\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"align_corners\",\n        \"name\": \"alignCorners\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"half_pixel_centers\",\n        \"name\": \"halfPixelCenters\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ResizeNearestNeighbor\",\n    \"category\": \"image\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"images\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"size\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"align_corners\",\n        \"name\": \"alignCorners\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"half_pixel_centers\",\n        \"name\": \"halfPixelCenters\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"CropAndResize\",\n    \"category\": \"image\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"image\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"boxes\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"boxInd\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"cropSize\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"method\",\n        \"name\": \"method\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"extrapolation_value\",\n        \"name\": \"extrapolationValue\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ImageProjectiveTransformV3\",\n    \"category\": \"image\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"images\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"transforms\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"outputShape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"fillValue\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"interpolation\",\n        \"name\": \"interpolation\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"fill_mode\",\n        \"name\": \"fillMode\",\n        \"type\": \"string\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/logical.js\nvar logical_exports = {};\n__export(logical_exports, {\n  json: () => json11\n});\nvar json11 = [\n  {\n    \"tfOpName\": \"Equal\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"NotEqual\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Greater\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"GreaterEqual\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Less\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LessEqual\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LogicalAnd\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LogicalNot\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LogicalOr\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Select\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"condition\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SelectV2\",\n    \"category\": \"logical\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"condition\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/matrices.js\nvar matrices_exports = {};\n__export(matrices_exports, {\n  json: () => json12\n});\nvar json12 = [\n  {\n    \"tfOpName\": \"_FusedMatMul\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"end\": 0,\n        \"name\": \"args\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"num_args\",\n        \"name\": \"numArgs\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"fused_ops\",\n        \"name\": \"fusedOps\",\n        \"type\": \"string[]\",\n        \"defaultValue\": []\n      },\n      {\n        \"tfName\": \"epsilon\",\n        \"name\": \"epsilon\",\n        \"type\": \"number\",\n        \"defaultValue\": 1e-4\n      },\n      {\n        \"tfName\": \"transpose_a\",\n        \"name\": \"transposeA\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"transpose_b\",\n        \"name\": \"transposeB\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"leakyrelu_alpha\",\n        \"name\": \"leakyreluAlpha\",\n        \"type\": \"number\",\n        \"defaultValue\": 0.2\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"MatMul\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"transpose_a\",\n        \"name\": \"transposeA\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"transpose_b\",\n        \"name\": \"transposeB\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"BatchMatMul\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"adj_x\",\n        \"name\": \"transposeA\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"adj_y\",\n        \"name\": \"transposeB\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"BatchMatMulV2\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"a\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"b\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"adj_x\",\n        \"name\": \"transposeA\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"adj_y\",\n        \"name\": \"transposeB\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Transpose\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"perm\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Einsum\",\n    \"category\": \"matrices\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"equation\",\n        \"name\": \"equation\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"N\",\n        \"name\": \"n\",\n        \"type\": \"number\",\n        \"defaultValue\": 2\n      },\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/normalization.js\nvar normalization_exports = {};\n__export(normalization_exports, {\n  json: () => json13\n});\nvar json13 = [\n  {\n    \"tfOpName\": \"EuclideanNorm\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\",\n        \"defaultValue\": false\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FusedBatchNorm\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scale\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"offset\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"mean\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"variance\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"epsilon\",\n        \"name\": \"epsilon\",\n        \"type\": \"number\",\n        \"defaultValue\": 1e-3\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FusedBatchNormV2\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scale\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"offset\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"mean\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"variance\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"epsilon\",\n        \"name\": \"epsilon\",\n        \"type\": \"number\",\n        \"defaultValue\": 1e-3\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"FusedBatchNormV3\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"scale\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"offset\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"mean\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 4,\n        \"name\": \"variance\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"epsilon\",\n        \"name\": \"epsilon\",\n        \"type\": \"number\",\n        \"defaultValue\": 1e-3\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LRN\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"depth_radius\",\n        \"name\": \"radius\",\n        \"type\": \"number\",\n        \"defaultValue\": 5\n      },\n      {\n        \"tfName\": \"bias\",\n        \"name\": \"bias\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      },\n      {\n        \"tfName\": \"alpha\",\n        \"name\": \"alpha\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      },\n      {\n        \"tfName\": \"beta\",\n        \"name\": \"beta\",\n        \"type\": \"number\",\n        \"defaultValue\": 0.5\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Softmax\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"LogSoftmax\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SparseToDense\",\n    \"category\": \"normalization\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"sparseIndices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"outputShape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"sparseValues\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"defaultValue\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"validate_indices\",\n        \"name\": \"validateIndices\",\n        \"type\": \"bool\",\n        \"defaultValue\": true,\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/reduction.js\nvar reduction_exports = {};\n__export(reduction_exports, {\n  json: () => json14\n});\nvar json14 = [\n  {\n    \"tfOpName\": \"Bincount\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"size\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"weights\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"DenseBincount\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"size\",\n        \"type\": \"number\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"weights\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"binary_output\",\n        \"name\": \"binaryOutput\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Max\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Mean\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Min\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Sum\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"All\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Any\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ArgMax\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ArgMin\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Prod\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"keep_dims\",\n        \"name\": \"keepDims\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Cumprod\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"exclusive\",\n        \"name\": \"exclusive\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"reverse\",\n        \"name\": \"reverse\",\n        \"type\": \"bool\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Cumsum\",\n    \"category\": \"reduction\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"exclusive\",\n        \"name\": \"exclusive\",\n        \"type\": \"bool\"\n      },\n      {\n        \"tfName\": \"reverse\",\n        \"name\": \"reverse\",\n        \"type\": \"bool\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/slice_join.js\nvar slice_join_exports = {};\n__export(slice_join_exports, {\n  json: () => json15\n});\nvar json15 = [\n  {\n    \"tfOpName\": \"ConcatV2\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": -1,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      },\n      {\n        \"start\": -1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"N\",\n        \"name\": \"n\",\n        \"type\": \"number\",\n        \"defaultValue\": 2\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Concat\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 1,\n        \"end\": 0,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      },\n      {\n        \"start\": 0,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"N\",\n        \"name\": \"n\",\n        \"type\": \"number\",\n        \"defaultValue\": 2\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"GatherV2\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"batch_dims\",\n        \"name\": \"batchDims\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Gather\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"validate_indices\",\n        \"name\": \"validateIndices\",\n        \"type\": \"bool\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Reverse\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"dims\",\n        \"type\": \"bool[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ReverseV2\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Slice\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"begin\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"size\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"StridedSlice\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"begin\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"end\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"strides\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"begin_mask\",\n        \"name\": \"beginMask\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"end_mask\",\n        \"name\": \"endMask\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"new_axis_mask\",\n        \"name\": \"newAxisMask\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"ellipsis_mask\",\n        \"name\": \"ellipsisMask\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"shrink_axis_mask\",\n        \"name\": \"shrinkAxisMask\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Pack\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"end\": 0,\n        \"name\": \"tensors\",\n        \"type\": \"tensors\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"axis\",\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Unpack\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"tensor\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"axis\",\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"tfName\": \"num\",\n        \"name\": \"num\",\n        \"type\": \"number\",\n        \"defaultValue\": 0,\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Tile\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"reps\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Split\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      },\n      {\n        \"start\": 1,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"num_split\",\n        \"name\": \"numOrSizeSplits\",\n        \"type\": \"number\",\n        \"defaultValue\": 1\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SplitV\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"numOrSizeSplits\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"axis\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ScatterNd\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"GatherNd\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SparseToDense\",\n    \"category\": \"slice_join\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"sparseIndices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"outputShape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"sparseValues\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"defaultValue\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"validate_indices\",\n        \"name\": \"validateIndices\",\n        \"type\": \"bool\",\n        \"defaultValue\": false,\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/sparse.js\nvar sparse_exports = {};\n__export(sparse_exports, {\n  json: () => json16\n});\nvar json16 = [\n  {\n    \"tfOpName\": \"SparseFillEmptyRows\",\n    \"category\": \"sparse\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"values\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"denseShape\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 3,\n        \"name\": \"defaultValue\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SparseReshape\",\n    \"category\": \"sparse\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"inputIndices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"inputShape\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"newShape\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"T\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SparseSegmentMean\",\n    \"category\": \"sparse\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"data\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"segmentIds\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SparseSegmentSum\",\n    \"category\": \"sparse\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"data\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"indices\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"segmentIds\",\n        \"type\": \"tensor\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/spectral.js\nvar spectral_exports = {};\n__export(spectral_exports, {\n  json: () => json17\n});\nvar json17 = [\n  {\n    \"tfOpName\": \"FFT\",\n    \"category\": \"spectral\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"IFFT\",\n    \"category\": \"spectral\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"RFFT\",\n    \"category\": \"spectral\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"fft_length\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"IRFFT\",\n    \"category\": \"spectral\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"fft_length\",\n        \"type\": \"number\",\n        \"notSupported\": true\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/string.js\nvar string_exports = {};\n__export(string_exports, {\n  json: () => json18\n});\nvar json18 = [\n  {\n    \"tfOpName\": \"StringNGrams\",\n    \"category\": \"string\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"data\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"dataSplits\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"separator\",\n        \"name\": \"separator\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"ngram_widths\",\n        \"name\": \"nGramWidths\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"tfName\": \"left_pad\",\n        \"name\": \"leftPad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"right_pad\",\n        \"name\": \"rightPad\",\n        \"type\": \"string\"\n      },\n      {\n        \"tfName\": \"pad_width\",\n        \"name\": \"padWidth\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"preserve_short_sequences\",\n        \"name\": \"preserveShortSequences\",\n        \"type\": \"bool\"\n      }\n    ],\n    \"outputs\": [\n      \"ngrams\",\n      \"ngrams_splits\"\n    ]\n  },\n  {\n    \"tfOpName\": \"StringSplit\",\n    \"category\": \"string\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"input\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"delimiter\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"skip_empty\",\n        \"name\": \"skipEmpty\",\n        \"type\": \"bool\"\n      }\n    ],\n    \"outputs\": [\n      \"indices\",\n      \"values\",\n      \"shape\"\n    ]\n  },\n  {\n    \"tfOpName\": \"StringToHashBucketFast\",\n    \"category\": \"string\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"input\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"num_buckets\",\n        \"name\": \"numBuckets\",\n        \"type\": \"number\"\n      }\n    ]\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/transformation.js\nvar transformation_exports = {};\n__export(transformation_exports, {\n  json: () => json19\n});\nvar json19 = [\n  {\n    \"tfOpName\": \"Cast\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"SrcT\",\n        \"name\": \"sdtype\",\n        \"type\": \"dtype\",\n        \"notSupported\": true\n      },\n      {\n        \"tfName\": \"DstT\",\n        \"name\": \"dtype\",\n        \"type\": \"dtype\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"ExpandDims\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"axis\",\n        \"type\": \"number\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"MirrorPad\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"padding\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"mode\",\n        \"name\": \"mode\",\n        \"type\": \"string\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Pad\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"padding\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"constant_value\",\n        \"name\": \"constantValue\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"PadV2\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"padding\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"constantValue\",\n        \"type\": \"number\",\n        \"defaultValue\": 0\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Reshape\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"Squeeze\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"axis\",\n        \"tfDeprecatedName\": \"squeeze_dims\",\n        \"name\": \"axis\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"SpaceToBatchND\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"blockShape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"paddings\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"BatchToSpaceND\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"blockShape\",\n        \"type\": \"number[]\"\n      },\n      {\n        \"start\": 2,\n        \"name\": \"crops\",\n        \"type\": \"number[]\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"DepthToSpace\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": [\n      {\n        \"tfName\": \"block_size\",\n        \"name\": \"blockSize\",\n        \"type\": \"number\"\n      },\n      {\n        \"tfName\": \"data_format\",\n        \"name\": \"dataFormat\",\n        \"type\": \"string\"\n      }\n    ]\n  },\n  {\n    \"tfOpName\": \"BroadcastTo\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"x\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"shape\",\n        \"type\": \"number[]\"\n      }\n    ],\n    \"attrs\": []\n  },\n  {\n    \"tfOpName\": \"BroadcastArgs\",\n    \"category\": \"transformation\",\n    \"inputs\": [\n      {\n        \"start\": 0,\n        \"name\": \"s0\",\n        \"type\": \"tensor\"\n      },\n      {\n        \"start\": 1,\n        \"name\": \"s1\",\n        \"type\": \"tensor\"\n      }\n    ],\n    \"attrs\": []\n  }\n];\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/operation_mapper.js\nvar OperationMapper = class {\n  constructor() {\n    const ops = [\n      arithmetic_exports,\n      basic_math_exports,\n      control_exports,\n      convolution_exports,\n      creation_exports,\n      dynamic_exports,\n      evaluation_exports,\n      graph_exports,\n      hash_table_exports,\n      image_exports,\n      logical_exports,\n      matrices_exports,\n      normalization_exports,\n      reduction_exports,\n      slice_join_exports,\n      sparse_exports,\n      spectral_exports,\n      string_exports,\n      transformation_exports\n    ];\n    const mappersJson = [].concat(...ops.map((op2) => op2.json));\n    this.opMappers = mappersJson.reduce((map, mapper) => {\n      map[mapper.tfOpName] = mapper;\n      return map;\n    }, {});\n  }\n  static get Instance() {\n    return this._instance || (this._instance = new this());\n  }\n  transformGraph(graph, signature = {}) {\n    const tfNodes = graph.node;\n    const placeholders = [];\n    const weights = [];\n    const initNodes = [];\n    const nodes = tfNodes.reduce((map, node) => {\n      map[node.name] = this.mapNode(node);\n      if (node.op.startsWith(\"Placeholder\")) {\n        placeholders.push(map[node.name]);\n      } else if (node.op === \"Const\") {\n        weights.push(map[node.name]);\n      } else if (node.input == null || node.input.length === 0) {\n        initNodes.push(map[node.name]);\n      }\n      return map;\n    }, {});\n    let inputs = [];\n    const outputs = [];\n    let inputNodeNameToKey = {};\n    let outputNodeNameToKey = {};\n    if (signature != null) {\n      inputNodeNameToKey = this.mapSignatureEntries(signature.inputs);\n      outputNodeNameToKey = this.mapSignatureEntries(signature.outputs);\n    }\n    const allNodes = Object.keys(nodes);\n    allNodes.forEach((key) => {\n      const node = nodes[key];\n      node.inputNames.forEach((name, index) => {\n        const [nodeName, , outputName] = getNodeNameAndIndex(name);\n        const inputNode = nodes[nodeName];\n        if (inputNode.outputs != null) {\n          const outputIndex = inputNode.outputs.indexOf(outputName);\n          if (outputIndex !== -1) {\n            const inputName = `${nodeName}:${outputIndex}`;\n            node.inputNames[index] = inputName;\n          }\n        }\n        node.inputs.push(inputNode);\n        inputNode.children.push(node);\n      });\n    });\n    if (Object.keys(outputNodeNameToKey).length === 0) {\n      allNodes.forEach((key) => {\n        const node = nodes[key];\n        if (node.children.length === 0) {\n          outputs.push(node);\n        }\n      });\n    } else {\n      Object.keys(outputNodeNameToKey).forEach((name) => {\n        const [nodeName] = getNodeNameAndIndex(name);\n        const node = nodes[nodeName];\n        if (node != null) {\n          node.signatureKey = outputNodeNameToKey[name];\n          outputs.push(node);\n        }\n      });\n    }\n    if (Object.keys(inputNodeNameToKey).length > 0) {\n      Object.keys(inputNodeNameToKey).forEach((name) => {\n        const [nodeName] = getNodeNameAndIndex(name);\n        const node = nodes[nodeName];\n        if (node) {\n          node.signatureKey = inputNodeNameToKey[name];\n          inputs.push(node);\n        }\n      });\n    } else {\n      inputs = placeholders;\n    }\n    let functions = {};\n    if (graph.library != null && graph.library.function != null) {\n      functions = graph.library.function.reduce((functions2, func2) => {\n        functions2[func2.signature.name] = this.mapFunction(func2);\n        return functions2;\n      }, {});\n    }\n    const result = { nodes, inputs, outputs, weights, placeholders, signature, functions };\n    if (initNodes.length > 0) {\n      result.initNodes = initNodes;\n    }\n    return result;\n  }\n  mapSignatureEntries(entries) {\n    return Object.keys(entries || {}).reduce((prev, curr) => {\n      prev[entries[curr].name] = curr;\n      return prev;\n    }, {});\n  }\n  mapNode(node) {\n    const mapper = getRegisteredOp(node.op) || this.opMappers[node.op] || {};\n    if (node.attr == null) {\n      node.attr = {};\n    }\n    const newNode = {\n      name: node.name,\n      op: node.op,\n      category: mapper.category,\n      inputNames: (node.input || []).map((input2) => input2.startsWith(\"^\") ? input2.slice(1) : input2),\n      inputs: [],\n      children: [],\n      inputParams: {},\n      attrParams: {},\n      rawAttrs: node.attr,\n      outputs: mapper.outputs\n    };\n    if (mapper.inputs != null) {\n      newNode.inputParams = mapper.inputs.reduce((map, param) => {\n        map[param.name] = {\n          type: param.type,\n          inputIndexStart: param.start,\n          inputIndexEnd: param.end\n        };\n        return map;\n      }, {});\n    }\n    if (mapper.attrs != null) {\n      newNode.attrParams = mapper.attrs.reduce((map, param) => {\n        const type = param.type;\n        let value = void 0;\n        switch (param.type) {\n          case \"string\":\n            value = getStringParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getStringParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"string[]\":\n            value = getStringArrayParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getStringArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"number\":\n            value = getNumberParam(node.attr, param.tfName, param.defaultValue || 0);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getNumberParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"number[]\":\n            value = getNumericArrayParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getNumericArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"bool\":\n            value = getBoolParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getBoolParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"bool[]\":\n            value = getBoolArrayParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getBoolArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"shape\":\n            value = getTensorShapeParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getTensorShapeParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"shape[]\":\n            value = getTensorShapeArrayParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getTensorShapeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"dtype\":\n            value = getDtypeParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getDtypeParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"dtype[]\":\n            value = getDtypeArrayParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getDtypeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"func\":\n            value = getFuncParam(node.attr, param.tfName, param.defaultValue);\n            if (value === void 0 && !!param.tfDeprecatedName) {\n              value = getFuncParam(node.attr, param.tfDeprecatedName, param.defaultValue);\n            }\n            break;\n          case \"tensor\":\n          case \"tensors\":\n            break;\n          default:\n            throw new Error(`Unsupported param type: ${param.type} for op: ${node.op}`);\n        }\n        map[param.name] = { value, type };\n        return map;\n      }, {});\n    }\n    return newNode;\n  }\n  mapFunction(functionDef) {\n    const tfNodes = functionDef.nodeDef;\n    const placeholders = [];\n    const weights = [];\n    let nodes = {};\n    if (tfNodes != null) {\n      nodes = tfNodes.reduce((map, node) => {\n        map[node.name] = this.mapNode(node);\n        if (node.op === \"Const\") {\n          weights.push(map[node.name]);\n        }\n        return map;\n      }, {});\n    }\n    const inputs = [];\n    const outputs = [];\n    functionDef.signature.inputArg.forEach((arg) => {\n      const [nodeName] = getNodeNameAndIndex(arg.name);\n      const node = {\n        name: nodeName,\n        op: \"Placeholder\",\n        inputs: [],\n        inputNames: [],\n        category: \"graph\",\n        inputParams: {},\n        attrParams: { dtype: { value: parseDtypeParam(arg.type), type: \"dtype\" } },\n        children: []\n      };\n      node.signatureKey = arg.name;\n      inputs.push(node);\n      nodes[nodeName] = node;\n    });\n    const allNodes = Object.keys(nodes);\n    allNodes.forEach((key) => {\n      const node = nodes[key];\n      node.inputNames.forEach((name, index) => {\n        const [nodeName, , outputName] = getNodeNameAndIndex(name);\n        const inputNode = nodes[nodeName];\n        if (inputNode.outputs != null) {\n          const outputIndex = inputNode.outputs.indexOf(outputName);\n          if (outputIndex !== -1) {\n            const inputName = `${nodeName}:${outputIndex}`;\n            node.inputNames[index] = inputName;\n          }\n        }\n        node.inputs.push(inputNode);\n        inputNode.children.push(node);\n      });\n    });\n    const returnNodeMap = functionDef.ret;\n    functionDef.signature.outputArg.forEach((output) => {\n      const [nodeName, index] = getNodeNameAndIndex(returnNodeMap[output.name]);\n      const node = nodes[nodeName];\n      if (node != null) {\n        node.defaultOutput = index;\n        outputs.push(node);\n      }\n    });\n    const signature = this.mapArgsToSignature(functionDef);\n    return { nodes, inputs, outputs, weights, placeholders, signature };\n  }\n  mapArgsToSignature(functionDef) {\n    return {\n      methodName: functionDef.signature.name,\n      inputs: functionDef.signature.inputArg.reduce((map, arg) => {\n        map[arg.name] = this.mapArgToTensorInfo(arg);\n        return map;\n      }, {}),\n      outputs: functionDef.signature.outputArg.reduce((map, arg) => {\n        map[arg.name] = this.mapArgToTensorInfo(arg, functionDef.ret);\n        return map;\n      }, {})\n    };\n  }\n  mapArgToTensorInfo(arg, nameMap2) {\n    let name = arg.name;\n    if (nameMap2 != null) {\n      name = nameMap2[name];\n    }\n    return { name, dtype: arg.type };\n  }\n};\nfunction decodeBase64(text) {\n  const global2 = env().global;\n  if (typeof global2.atob !== \"undefined\") {\n    return global2.atob(text);\n  } else if (typeof Buffer !== \"undefined\") {\n    return new Buffer(text, \"base64\").toString();\n  } else {\n    throw new Error(\"Unable to decode base64 in this environment. Missing built-in atob() or Buffer()\");\n  }\n}\nfunction parseStringParam(s, keepCase) {\n  const value = Array.isArray(s) ? String.fromCharCode.apply(null, s) : decodeBase64(s);\n  return keepCase ? value : value.toLowerCase();\n}\nfunction getStringParam(attrs, name, def, keepCase = false) {\n  const param = attrs[name];\n  if (param != null) {\n    return parseStringParam(param.s, keepCase);\n  }\n  return def;\n}\nfunction getBoolParam(attrs, name, def) {\n  const param = attrs[name];\n  return param ? param.b : def;\n}\nfunction getNumberParam(attrs, name, def) {\n  const param = attrs[name] || {};\n  const value = param[\"i\"] != null ? param[\"i\"] : param[\"f\"] != null ? param[\"f\"] : def;\n  return typeof value === \"number\" ? value : parseInt(value, 10);\n}\nfunction parseDtypeParam(value) {\n  if (typeof value === \"string\") {\n    value = DataType[value];\n  }\n  switch (value) {\n    case DataType.DT_FLOAT:\n    case DataType.DT_HALF:\n      return \"float32\";\n    case DataType.DT_INT32:\n    case DataType.DT_INT64:\n    case DataType.DT_INT8:\n    case DataType.DT_UINT8:\n      return \"int32\";\n    case DataType.DT_BOOL:\n      return \"bool\";\n    case DataType.DT_DOUBLE:\n      return \"float32\";\n    case DataType.DT_STRING:\n      return \"string\";\n    default:\n      return null;\n  }\n}\nfunction getFuncParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.func) {\n    return param.func.name;\n  }\n  return def;\n}\nfunction getDtypeParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.type) {\n    return parseDtypeParam(param.type);\n  }\n  return def;\n}\nfunction getDtypeArrayParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.list && param.list.type) {\n    return param.list.type.map((v) => parseDtypeParam(v));\n  }\n  return def;\n}\nfunction parseTensorShapeParam(shape) {\n  if (shape.unknownRank) {\n    return void 0;\n  }\n  if (shape.dim != null) {\n    return shape.dim.map((dim) => typeof dim.size === \"number\" ? dim.size : parseInt(dim.size, 10));\n  }\n  return [];\n}\nfunction getTensorShapeParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.shape) {\n    return parseTensorShapeParam(param.shape);\n  }\n  return def;\n}\nfunction getNumericArrayParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param) {\n    return ((param.list.f && param.list.f.length ? param.list.f : param.list.i) || []).map((v) => typeof v === \"number\" ? v : parseInt(v, 10));\n  }\n  return def;\n}\nfunction getStringArrayParam(attrs, name, def, keepCase = false) {\n  const param = attrs[name];\n  if (param && param.list && param.list.s) {\n    return param.list.s.map((v) => {\n      return parseStringParam(v, keepCase);\n    });\n  }\n  return def;\n}\nfunction getTensorShapeArrayParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.list && param.list.shape) {\n    return param.list.shape.map((v) => {\n      return parseTensorShapeParam(v);\n    });\n  }\n  return def;\n}\nfunction getBoolArrayParam(attrs, name, def) {\n  const param = attrs[name];\n  if (param && param.list && param.list.b) {\n    return param.list.b;\n  }\n  return def;\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/custom_op/node_value_impl.js\nvar NodeValueImpl = class {\n  constructor(node, tensorMap, context) {\n    this.node = node;\n    this.tensorMap = tensorMap;\n    this.context = context;\n    this.inputs = [];\n    this.attrs = {};\n    this.inputs = node.inputNames.map((name) => this.getInput(name));\n    if (node.rawAttrs != null) {\n      this.attrs = Object.keys(node.rawAttrs).reduce((attrs, key) => {\n        attrs[key] = this.getAttr(key);\n        return attrs;\n      }, {});\n    }\n  }\n  getInput(name) {\n    return getTensor(name, this.tensorMap, this.context);\n  }\n  getAttr(name, defaultValue) {\n    const value = this.node.rawAttrs[name];\n    if (value.tensor != null) {\n      return getTensor(name, this.tensorMap, this.context);\n    }\n    if (value.i != null || value.f != null) {\n      return getNumberParam(this.node.rawAttrs, name, defaultValue);\n    }\n    if (value.s != null) {\n      return getStringParam(this.node.rawAttrs, name, defaultValue);\n    }\n    if (value.b != null) {\n      return getBoolParam(this.node.rawAttrs, name, defaultValue);\n    }\n    if (value.shape != null) {\n      return getTensorShapeParam(this.node.rawAttrs, name, defaultValue);\n    }\n    if (value.type != null) {\n      return getDtypeParam(this.node.rawAttrs, name, defaultValue);\n    }\n    if (value.list != null) {\n      if (value.list.i != null || value.list.f != null) {\n        return getNumericArrayParam(this.node.rawAttrs, name, defaultValue);\n      }\n      if (value.list.s != null) {\n        return getStringArrayParam(this.node.rawAttrs, name, defaultValue);\n      }\n      if (value.list.shape != null) {\n        return getTensorShapeArrayParam(this.node.rawAttrs, name, defaultValue);\n      }\n      if (value.list.b != null) {\n        return getBoolArrayParam(this.node.rawAttrs, name, defaultValue);\n      }\n      if (value.list.type != null) {\n        return getDtypeArrayParam(this.node.rawAttrs, name, defaultValue);\n      }\n    }\n    return defaultValue;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-core/dist/ops/ops_for_converter.js\nvar ops_for_converter_exports = {};\n__export(ops_for_converter_exports, {\n  OP_SCOPE_SUFFIX: () => OP_SCOPE_SUFFIX,\n  abs: () => abs,\n  acos: () => acos,\n  acosh: () => acosh,\n  add: () => add2,\n  addN: () => addN,\n  all: () => all,\n  any: () => any,\n  argMax: () => argMax,\n  argMin: () => argMin,\n  asin: () => asin,\n  asinh: () => asinh,\n  atan: () => atan,\n  atan2: () => atan2,\n  atanh: () => atanh,\n  avgPool: () => avgPool,\n  avgPool3d: () => avgPool3d,\n  basicLSTMCell: () => basicLSTMCell,\n  batchNorm: () => batchNorm,\n  batchNorm2d: () => batchNorm2d,\n  batchNorm3d: () => batchNorm3d,\n  batchNorm4d: () => batchNorm4d,\n  batchToSpaceND: () => batchToSpaceND,\n  bincount: () => bincount,\n  booleanMaskAsync: () => booleanMaskAsync,\n  broadcastArgs: () => broadcastArgs,\n  broadcastTo: () => broadcastTo,\n  buffer: () => buffer,\n  cast: () => cast,\n  ceil: () => ceil,\n  clipByValue: () => clipByValue,\n  clone: () => clone,\n  complex: () => complex,\n  concat: () => concat,\n  concat1d: () => concat1d,\n  concat2d: () => concat2d,\n  concat3d: () => concat3d,\n  concat4d: () => concat4d,\n  conv1d: () => conv1d,\n  conv2d: () => conv2d,\n  conv2dTranspose: () => conv2dTranspose,\n  conv3d: () => conv3d,\n  conv3dTranspose: () => conv3dTranspose,\n  cos: () => cos,\n  cosh: () => cosh,\n  cosineWindow: () => cosineWindow,\n  cumprod: () => cumprod,\n  cumsum: () => cumsum,\n  denseBincount: () => denseBincount,\n  depthToSpace: () => depthToSpace,\n  depthwiseConv2d: () => depthwiseConv2d,\n  diag: () => diag,\n  dilation2d: () => dilation2d,\n  div: () => div,\n  divNoNan: () => divNoNan,\n  dot: () => dot,\n  dropout: () => dropout,\n  einsum: () => einsum,\n  elu: () => elu,\n  enclosingPowerOfTwo: () => enclosingPowerOfTwo,\n  equal: () => equal,\n  erf: () => erf,\n  euclideanNorm: () => euclideanNorm,\n  exp: () => exp,\n  expandDims: () => expandDims,\n  expm1: () => expm1,\n  eye: () => eye,\n  fft: () => fft,\n  fill: () => fill,\n  floor: () => floor,\n  floorDiv: () => floorDiv,\n  fused: () => fused_ops_exports,\n  gather: () => gather,\n  gatherND: () => gatherND,\n  greater: () => greater,\n  greaterEqual: () => greaterEqual,\n  ifft: () => ifft,\n  imag: () => imag,\n  image: () => image,\n  inTopKAsync: () => inTopKAsync,\n  irfft: () => irfft,\n  isFinite: () => isFinite2,\n  isInf: () => isInf,\n  isNaN: () => isNaN2,\n  leakyRelu: () => leakyRelu,\n  less: () => less,\n  lessEqual: () => lessEqual,\n  linalg: () => linalg,\n  linspace: () => linspace,\n  localResponseNormalization: () => localResponseNormalization,\n  log: () => log2,\n  log1p: () => log1p,\n  logSigmoid: () => logSigmoid,\n  logSoftmax: () => logSoftmax,\n  logSumExp: () => logSumExp,\n  logicalAnd: () => logicalAnd,\n  logicalNot: () => logicalNot,\n  logicalOr: () => logicalOr,\n  logicalXor: () => logicalXor,\n  losses: () => losses,\n  lowerBound: () => lowerBound,\n  matMul: () => matMul,\n  max: () => max,\n  maxPool: () => maxPool,\n  maxPool3d: () => maxPool3d,\n  maxPoolWithArgmax: () => maxPoolWithArgmax,\n  maximum: () => maximum,\n  mean: () => mean,\n  meshgrid: () => meshgrid,\n  min: () => min,\n  minimum: () => minimum,\n  mirrorPad: () => mirrorPad,\n  mod: () => mod,\n  moments: () => moments,\n  movingAverage: () => movingAverage,\n  mul: () => mul,\n  multiRNNCell: () => multiRNNCell,\n  multinomial: () => multinomial,\n  neg: () => neg,\n  norm: () => norm,\n  notEqual: () => notEqual,\n  oneHot: () => oneHot,\n  ones: () => ones2,\n  onesLike: () => onesLike,\n  op: () => op,\n  outerProduct: () => outerProduct,\n  pad: () => pad,\n  pad1d: () => pad1d,\n  pad2d: () => pad2d,\n  pad3d: () => pad3d,\n  pad4d: () => pad4d,\n  pool: () => pool,\n  pow: () => pow,\n  prelu: () => prelu,\n  print: () => print,\n  prod: () => prod,\n  raggedGather: () => raggedGather,\n  raggedRange: () => raggedRange,\n  raggedTensorToTensor: () => raggedTensorToTensor,\n  rand: () => rand,\n  randomGamma: () => randomGamma,\n  randomNormal: () => randomNormal,\n  randomStandardNormal: () => randomStandardNormal,\n  randomUniform: () => randomUniform,\n  range: () => range,\n  real: () => real,\n  reciprocal: () => reciprocal,\n  relu: () => relu,\n  relu6: () => relu6,\n  reshape: () => reshape,\n  reverse: () => reverse,\n  reverse1d: () => reverse1d,\n  reverse2d: () => reverse2d,\n  reverse3d: () => reverse3d,\n  reverse4d: () => reverse4d,\n  rfft: () => rfft,\n  round: () => round2,\n  rsqrt: () => rsqrt,\n  scalar: () => scalar,\n  scatterND: () => scatterND,\n  searchSorted: () => searchSorted,\n  selu: () => selu,\n  separableConv2d: () => separableConv2d,\n  setdiff1dAsync: () => setdiff1dAsync,\n  sigmoid: () => sigmoid,\n  sign: () => sign,\n  signal: () => signal,\n  sin: () => sin,\n  sinh: () => sinh,\n  slice: () => slice,\n  slice1d: () => slice1d,\n  slice2d: () => slice2d,\n  slice3d: () => slice3d,\n  slice4d: () => slice4d,\n  softmax: () => softmax,\n  softplus: () => softplus,\n  spaceToBatchND: () => spaceToBatchND,\n  sparse: () => sparse,\n  sparseToDense: () => sparseToDense,\n  spectral: () => spectral,\n  split: () => split,\n  sqrt: () => sqrt,\n  square: () => square,\n  squaredDifference: () => squaredDifference,\n  squeeze: () => squeeze,\n  stack: () => stack,\n  step: () => step,\n  stridedSlice: () => stridedSlice,\n  string: () => string,\n  sub: () => sub,\n  sum: () => sum2,\n  tan: () => tan,\n  tanh: () => tanh2,\n  tensor: () => tensor,\n  tensor1d: () => tensor1d,\n  tensor2d: () => tensor2d,\n  tensor3d: () => tensor3d,\n  tensor4d: () => tensor4d,\n  tensor5d: () => tensor5d,\n  tensor6d: () => tensor6d,\n  tile: () => tile,\n  topk: () => topk,\n  transpose: () => transpose,\n  truncatedNormal: () => truncatedNormal,\n  unique: () => unique,\n  unsortedSegmentSum: () => unsortedSegmentSum,\n  unstack: () => unstack,\n  upperBound: () => upperBound,\n  variable: () => variable,\n  where: () => where,\n  whereAsync: () => whereAsync,\n  zeros: () => zeros,\n  zerosLike: () => zerosLike\n});\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/arithmetic_executor.js\nvar executeOp = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"BiasAdd\":\n    case \"AddV2\":\n    case \"Add\": {\n      return [ops.add(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"AddN\": {\n      return [ops.addN(getParamValue(\"tensors\", node, tensorMap, context))];\n    }\n    case \"FloorMod\":\n    case \"Mod\":\n      return [ops.mod(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    case \"Mul\":\n      return [ops.mul(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    case \"RealDiv\":\n    case \"Div\": {\n      return [ops.div(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"DivNoNan\": {\n      return [ops.divNoNan(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"FloorDiv\": {\n      return [ops.floorDiv(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Sub\": {\n      return [ops.sub(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Minimum\": {\n      return [ops.minimum(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Maximum\": {\n      return [ops.maximum(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Pow\": {\n      return [ops.pow(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"SquaredDifference\": {\n      return [ops.squaredDifference(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/basic_math_executor.js\nvar executeOp2 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Abs\":\n    case \"ComplexAbs\":\n      return [ops.abs(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Acos\":\n      return [ops.acos(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Acosh\":\n      return [ops.acosh(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Asin\":\n      return [ops.asin(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Asinh\":\n      return [ops.asinh(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Atan\":\n      return [ops.atan(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Atan2\":\n      return [ops.atan2(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"y\", node, tensorMap, context))];\n    case \"Atanh\":\n      return [ops.atanh(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Ceil\":\n      return [ops.ceil(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Complex\":\n      return [ops.complex(getParamValue(\"real\", node, tensorMap, context), getParamValue(\"imag\", node, tensorMap, context))];\n    case \"Cos\":\n      return [ops.cos(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Cosh\":\n      return [ops.cosh(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Elu\":\n      return [ops.elu(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Erf\":\n      return [ops.erf(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Exp\":\n      return [ops.exp(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Expm1\": {\n      return [ops.expm1(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Floor\":\n      return [ops.floor(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Log\":\n      return [ops.log(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Log1p\": {\n      return [ops.log1p(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Imag\":\n      return [ops.imag(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Neg\":\n      return [ops.neg(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Reciprocal\": {\n      return [ops.reciprocal(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Real\":\n      return [ops.real(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Relu\":\n      return [ops.relu(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Round\": {\n      return [ops.round(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Selu\":\n      return [ops.selu(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Sigmoid\":\n      return [ops.sigmoid(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Sin\":\n      return [ops.sin(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Sign\": {\n      return [ops.sign(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Sinh\": {\n      return [ops.sinh(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Softplus\": {\n      return [ops.softplus(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Sqrt\": {\n      return [ops.sqrt(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Square\": {\n      return [ops.square(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Tanh\": {\n      return [ops.tanh(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"Tan\":\n      return [ops.tan(getParamValue(\"x\", node, tensorMap, context))];\n    case \"ClipByValue\":\n      return [ops.clipByValue(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"clipValueMin\", node, tensorMap, context), getParamValue(\"clipValueMax\", node, tensorMap, context))];\n    case \"Relu6\":\n      return [ops.relu6(getParamValue(\"x\", node, tensorMap, context))];\n    case \"Rsqrt\":\n      return [ops.rsqrt(getTensor(node.inputNames[0], tensorMap, context))];\n    case \"Prod\":\n      return [ops.prod(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"axes\", node, tensorMap, context))];\n    case \"LeakyRelu\":\n      return [ops.leakyRelu(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"alpha\", node, tensorMap, context))];\n    case \"Prelu\":\n      return [ops.prelu(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"alpha\", node, tensorMap, context))];\n    case \"IsNan\":\n      return [ops.isNaN(getTensor(node.inputNames[0], tensorMap, context))];\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_utils.js\nfunction assertShapesMatchAllowUndefinedSize(shapeA, shapeB, errorMessagePrefix = \"\") {\n  if (typeof shapeA === \"number\" || typeof shapeB === \"number\") {\n    return;\n  }\n  util_exports.assert(shapeA.length === shapeB.length, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);\n  for (let i = 0; i < shapeA.length; i++) {\n    const dim0 = shapeA[i];\n    const dim1 = shapeB[i];\n    util_exports.assert(dim0 < 0 || dim1 < 0 || dim0 === dim1, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);\n  }\n}\nfunction fullDefinedShape(elementShape) {\n  if (typeof elementShape === \"number\" || elementShape.some((dim) => dim < 0)) {\n    return false;\n  }\n  return true;\n}\nfunction inferElementShape(listElementShape, tensors, elementShape) {\n  let partialShape = mergeElementShape(listElementShape, elementShape);\n  const notfullDefinedShape = !fullDefinedShape(partialShape);\n  if (notfullDefinedShape && tensors.length === 0) {\n    throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${partialShape}`);\n  }\n  if (notfullDefinedShape) {\n    tensors.forEach((tensor2) => {\n      partialShape = mergeElementShape(tensor2.shape, partialShape);\n    });\n  }\n  if (!fullDefinedShape(partialShape)) {\n    throw new Error(`Non-fully-defined elementShape: ${partialShape}`);\n  }\n  return partialShape;\n}\nfunction mergeElementShape(elementShapeA, elementShapeB) {\n  if (typeof elementShapeA === \"number\") {\n    return elementShapeB;\n  }\n  if (typeof elementShapeB === \"number\") {\n    return elementShapeA;\n  }\n  if (elementShapeA.length !== elementShapeB.length) {\n    throw new Error(`Incompatible ranks during merge: ${elementShapeA} vs. ${elementShapeB}`);\n  }\n  const result = [];\n  for (let i = 0; i < elementShapeA.length; ++i) {\n    const dim0 = elementShapeA[i];\n    const dim1 = elementShapeB[i];\n    if (dim0 >= 0 && dim1 >= 0 && dim0 !== dim1) {\n      throw new Error(`Incompatible shape during merge: ${elementShapeA} vs. ${elementShapeB}`);\n    }\n    result[i] = dim0 >= 0 ? dim0 : dim1;\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_array.js\nvar TensorArray = class {\n  constructor(name, dtype, maxSize, elementShape, identicalElementShapes, dynamicSize, clearAfterRead) {\n    this.name = name;\n    this.dtype = dtype;\n    this.maxSize = maxSize;\n    this.elementShape = elementShape;\n    this.identicalElementShapes = identicalElementShapes;\n    this.dynamicSize = dynamicSize;\n    this.clearAfterRead = clearAfterRead;\n    this.tensors = [];\n    this.closed_ = false;\n    this.idTensor = scalar(0);\n    keep(this.idTensor);\n  }\n  get id() {\n    return this.idTensor.id;\n  }\n  get closed() {\n    return this.closed_;\n  }\n  clearAndClose(keepIds) {\n    this.tensors.forEach((tensor2) => {\n      if (keepIds == null || !keepIds.has(tensor2.tensor.id)) {\n        tensor2.tensor.dispose();\n      }\n    });\n    this.tensors = [];\n    this.closed_ = true;\n    this.idTensor.dispose();\n  }\n  size() {\n    return this.tensors.length;\n  }\n  read(index) {\n    if (this.closed_) {\n      throw new Error(`TensorArray ${this.name} has already been closed.`);\n    }\n    if (index < 0 || index >= this.size()) {\n      throw new Error(`Tried to read from index ${index}, but array size is: ${this.size()}`);\n    }\n    const tensorWithState = this.tensors[index];\n    if (tensorWithState.cleared) {\n      throw new Error(`TensorArray ${this.name}: Could not read index ${index} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);\n    }\n    if (this.clearAfterRead) {\n      tensorWithState.cleared = true;\n    }\n    tensorWithState.read = true;\n    return tensorWithState.tensor;\n  }\n  readMany(indices) {\n    return indices.map((index) => this.read(index));\n  }\n  write(index, tensor2) {\n    if (this.closed_) {\n      throw new Error(`TensorArray ${this.name} has already been closed.`);\n    }\n    if (index < 0 || !this.dynamicSize && index >= this.maxSize) {\n      throw new Error(`Tried to write to index ${index}, but array is not resizeable and size is: ${this.maxSize}`);\n    }\n    const t = this.tensors[index] || {};\n    if (tensor2.dtype !== this.dtype) {\n      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index},\n          because the value dtype is ${tensor2.dtype}, but TensorArray dtype is ${this.dtype}.`);\n    }\n    if (this.size() === 0 && (this.elementShape == null || this.elementShape.length === 0)) {\n      this.elementShape = tensor2.shape;\n    }\n    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, `TensorArray ${this.name}: Could not write to TensorArray index ${index}.`);\n    if (t.read) {\n      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been read.`);\n    }\n    if (t.written) {\n      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been written.`);\n    }\n    t.tensor = tensor2;\n    keep(tensor2);\n    t.written = true;\n    this.tensors[index] = t;\n  }\n  writeMany(indices, tensors) {\n    if (indices.length !== tensors.length) {\n      throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${indices.length} is not the same as tensors size: ${tensors.length}.`);\n    }\n    indices.forEach((i, index) => this.write(i, tensors[index]));\n  }\n  gather(indices, dtype) {\n    if (!!dtype && dtype !== this.dtype) {\n      throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${dtype}`);\n    }\n    if (!indices) {\n      indices = [];\n      for (let i = 0; i < this.size(); i++) {\n        indices.push(i);\n      }\n    } else {\n      indices = indices.slice(0, this.size());\n    }\n    if (indices.length === 0) {\n      return tensor([], [0].concat(this.elementShape));\n    }\n    const tensors = this.readMany(indices);\n    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, \"TensorArray shape mismatch: \");\n    return stack(tensors, 0);\n  }\n  concat(dtype) {\n    if (!!dtype && dtype !== this.dtype) {\n      throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${dtype}`);\n    }\n    if (this.size() === 0) {\n      return tensor([], [0].concat(this.elementShape));\n    }\n    const indices = [];\n    for (let i = 0; i < this.size(); i++) {\n      indices.push(i);\n    }\n    const tensors = this.readMany(indices);\n    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, `TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${tensors[0].shape})`);\n    return concat(tensors, 0);\n  }\n  scatter(indices, tensor2) {\n    if (tensor2.dtype !== this.dtype) {\n      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);\n    }\n    if (indices.length !== tensor2.shape[0]) {\n      throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);\n    }\n    const maxIndex = Math.max(...indices);\n    if (!this.dynamicSize && maxIndex >= this.maxSize) {\n      throw new Error(`Max index must be < array size (${maxIndex}  vs. ${this.maxSize})`);\n    }\n    this.writeMany(indices, unstack(tensor2, 0));\n  }\n  split(length, tensor2) {\n    if (tensor2.dtype !== this.dtype) {\n      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);\n    }\n    let totalLength = 0;\n    const cumulativeLengths = length.map((len) => {\n      totalLength += len;\n      return totalLength;\n    });\n    if (totalLength !== tensor2.shape[0]) {\n      throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);\n    }\n    if (!this.dynamicSize && length.length !== this.maxSize) {\n      throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${length.length}), and the TensorArray is not marked as dynamically resizeable`);\n    }\n    const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;\n    const tensors = [];\n    tidy(() => {\n      tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);\n      for (let i = 0; i < length.length; ++i) {\n        const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];\n        const indices2 = [0, previousLength, 0];\n        const sizes = [1, length[i], elementPerRow];\n        tensors[i] = reshape(slice(tensor2, indices2, sizes), this.elementShape);\n      }\n      return tensors;\n    });\n    const indices = [];\n    for (let i = 0; i < length.length; i++) {\n      indices[i] = i;\n    }\n    this.writeMany(indices, tensors);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_list.js\nvar TensorList = class {\n  constructor(tensors, elementShape, elementDtype, maxNumElements = -1) {\n    this.tensors = tensors;\n    this.elementShape = elementShape;\n    this.elementDtype = elementDtype;\n    if (tensors != null) {\n      tensors.forEach((tensor2) => {\n        if (elementDtype !== tensor2.dtype) {\n          throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${tensor2.dtype}`);\n        }\n        assertShapesMatchAllowUndefinedSize(elementShape, tensor2.shape, \"TensorList shape mismatch: \");\n        keep(tensor2);\n      });\n    }\n    this.idTensor = scalar(0);\n    this.maxNumElements = maxNumElements;\n    keep(this.idTensor);\n  }\n  get id() {\n    return this.idTensor.id;\n  }\n  copy() {\n    return new TensorList([...this.tensors], this.elementShape, this.elementDtype);\n  }\n  clearAndClose(keepIds) {\n    this.tensors.forEach((tensor2) => {\n      if (keepIds == null || !keepIds.has(tensor2.id)) {\n        tensor2.dispose();\n      }\n    });\n    this.tensors.length = 0;\n    this.idTensor.dispose();\n  }\n  size() {\n    return this.tensors.length;\n  }\n  stack(elementShape, elementDtype, numElements = -1) {\n    if (elementDtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);\n    }\n    if (numElements !== -1 && this.tensors.length !== numElements) {\n      throw new Error(`Operation expected a list with ${numElements} elements but got a list with ${this.tensors.length} elements.`);\n    }\n    assertShapesMatchAllowUndefinedSize(elementShape, this.elementShape, \"TensorList shape mismatch: \");\n    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);\n    return tidy(() => {\n      const reshapedTensors = this.tensors.map((tensor2) => reshape(tensor2, outputElementShape));\n      return stack(reshapedTensors, 0);\n    });\n  }\n  popBack(elementShape, elementDtype) {\n    if (elementDtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);\n    }\n    if (this.size() === 0) {\n      throw new Error(\"Trying to pop from an empty list.\");\n    }\n    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);\n    const tensor2 = this.tensors.pop();\n    tensor2.kept = false;\n    assertShapesMatchAllowUndefinedSize(tensor2.shape, elementShape, \"TensorList shape mismatch: \");\n    return reshape(tensor2, outputElementShape);\n  }\n  pushBack(tensor2) {\n    if (tensor2.dtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);\n    }\n    assertShapesMatchAllowUndefinedSize(tensor2.shape, this.elementShape, \"TensorList shape mismatch: \");\n    if (this.maxNumElements === this.size()) {\n      throw new Error(`Trying to push element into a full list.`);\n    }\n    keep(tensor2);\n    this.tensors.push(tensor2);\n  }\n  resize(size) {\n    if (size < 0) {\n      throw new Error(`TensorListResize expects size to be non-negative. Got: ${size}`);\n    }\n    if (this.maxNumElements !== -1 && size > this.maxNumElements) {\n      throw new Error(`TensorListResize input size ${size} is greater maxNumElement ${this.maxNumElements}.`);\n    }\n    const destTensorList = new TensorList([], this.elementShape, this.elementDtype, this.maxNumElements);\n    destTensorList.tensors.length = size;\n    for (let i = 0; i < Math.min(this.tensors.length, size); ++i) {\n      destTensorList.tensors[i] = this.tensors[i];\n    }\n    return destTensorList;\n  }\n  getItem(elementIndex, elementShape, elementDtype) {\n    if (elementDtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);\n    }\n    if (elementIndex < 0 || elementIndex > this.tensors.length) {\n      throw new Error(`Trying to access element ${elementIndex} in a list with ${this.tensors.length} elements.`);\n    }\n    if (this.tensors[elementIndex] == null) {\n      throw new Error(`element at index ${elementIndex} is null.`);\n    }\n    assertShapesMatchAllowUndefinedSize(this.tensors[elementIndex].shape, elementShape, \"TensorList shape mismatch: \");\n    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);\n    return reshape(this.tensors[elementIndex], outputElementShape);\n  }\n  setItem(elementIndex, tensor2) {\n    if (tensor2.dtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);\n    }\n    if (elementIndex < 0 || this.maxNumElements !== -1 && elementIndex >= this.maxNumElements) {\n      throw new Error(`Trying to set element ${elementIndex} in a list with max ${this.maxNumElements} elements.`);\n    }\n    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, \"TensorList shape mismatch: \");\n    keep(tensor2);\n    if (this.tensors[elementIndex] != null) {\n      this.tensors[elementIndex].kept = false;\n    }\n    this.tensors[elementIndex] = tensor2;\n  }\n  gather(indices, elementDtype, elementShape) {\n    if (elementDtype !== this.elementDtype) {\n      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);\n    }\n    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, \"TensorList shape mismatch: \");\n    indices = indices.slice(0, this.size());\n    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);\n    if (indices.length === 0) {\n      return tensor([], [0].concat(outputElementShape));\n    }\n    return tidy(() => {\n      const tensors = indices.map((i) => reshape(this.tensors[i], outputElementShape));\n      return stack(tensors, 0);\n    });\n  }\n  concat(elementDtype, elementShape) {\n    if (!!elementDtype && elementDtype !== this.elementDtype) {\n      throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${elementDtype}`);\n    }\n    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, \"TensorList shape mismatch: \");\n    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);\n    if (this.size() === 0) {\n      return tensor([], [0].concat(outputElementShape));\n    }\n    return tidy(() => {\n      const tensors = this.tensors.map((t) => reshape(t, outputElementShape));\n      return concat(tensors, 0);\n    });\n  }\n};\nfunction fromTensor(tensor2, elementShape, elementDtype) {\n  const dtype = tensor2.dtype;\n  if (tensor2.shape.length < 1) {\n    throw new Error(`Tensor must be at least a vector, but saw shape: ${tensor2.shape}`);\n  }\n  if (tensor2.dtype !== elementDtype) {\n    throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${elementDtype}`);\n  }\n  const tensorElementShape = tensor2.shape.slice(1);\n  assertShapesMatchAllowUndefinedSize(tensorElementShape, elementShape, \"TensorList shape mismatch: \");\n  const tensorList = unstack(tensor2);\n  return new TensorList(tensorList, elementShape, dtype);\n}\nfunction reserve(elementShape, elementDtype, numElements, maxNumElements) {\n  return new TensorList([], elementShape, elementDtype, maxNumElements);\n}\nfunction scatter(tensor2, indices, elementShape, numElements) {\n  if (indices.length !== tensor2.shape[0]) {\n    throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);\n  }\n  const maxIndex = Math.max(...indices);\n  if (numElements != null && numElements !== -1 && maxIndex >= numElements) {\n    throw new Error(`Max index must be < array size (${maxIndex}  vs. ${numElements})`);\n  }\n  const list = new TensorList([], elementShape, tensor2.dtype, numElements);\n  const tensors = unstack(tensor2, 0);\n  indices.forEach((value, index) => {\n    list.setItem(value, tensors[index]);\n  });\n  return list;\n}\nfunction split2(tensor2, length, elementShape) {\n  let totalLength = 0;\n  const cumulativeLengths = length.map((len) => {\n    totalLength += len;\n    return totalLength;\n  });\n  if (totalLength !== tensor2.shape[0]) {\n    throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);\n  }\n  const shapeWithoutFirstDim = tensor2.shape.slice(1);\n  const outputElementShape = mergeElementShape(shapeWithoutFirstDim, elementShape);\n  const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;\n  const tensors = tidy(() => {\n    const tensors2 = [];\n    tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);\n    for (let i = 0; i < length.length; ++i) {\n      const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];\n      const indices = [0, previousLength, 0];\n      const sizes = [1, length[i], elementPerRow];\n      tensors2[i] = reshape(slice(tensor2, indices, sizes), outputElementShape);\n    }\n    tensor2.dispose();\n    return tensors2;\n  });\n  const list = new TensorList([], elementShape, tensor2.dtype, length.length);\n  for (let i = 0; i < tensors.length; i++) {\n    list.setItem(i, tensors[i]);\n  }\n  return list;\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/control_executor.js\nvar executeOp3 = async (node, tensorMap, context) => {\n  switch (node.op) {\n    case \"If\":\n    case \"StatelessIf\": {\n      const thenFunc = getParamValue(\"thenBranch\", node, tensorMap, context);\n      const elseFunc = getParamValue(\"elseBranch\", node, tensorMap, context);\n      const cond = getParamValue(\"cond\", node, tensorMap, context);\n      const args = getParamValue(\"args\", node, tensorMap, context);\n      const condValue = await cond.data();\n      if (condValue[0]) {\n        return context.functionMap[thenFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);\n      } else {\n        return context.functionMap[elseFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);\n      }\n    }\n    case \"While\":\n    case \"StatelessWhile\": {\n      const bodyFunc = getParamValue(\"body\", node, tensorMap, context);\n      const condFunc = getParamValue(\"cond\", node, tensorMap, context);\n      const args = getParamValue(\"args\", node, tensorMap, context);\n      const condResult = await context.functionMap[condFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);\n      const argIds = args.map((tensor2) => tensor2.id);\n      let condValue = await condResult[0].data();\n      condResult.forEach((tensor2) => {\n        if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1) {\n          tensor2.dispose();\n        }\n      });\n      let result = args;\n      while (condValue[0]) {\n        const origResult = result;\n        result = await context.functionMap[bodyFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);\n        const resultIds = result.map((tensor2) => tensor2.id);\n        origResult.forEach((tensor2) => {\n          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {\n            tensor2.dispose();\n          }\n        });\n        const condResult2 = await context.functionMap[condFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);\n        condValue = await condResult2[0].data();\n        condResult2.forEach((tensor2) => {\n          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {\n            tensor2.dispose();\n          }\n        });\n      }\n      return result;\n    }\n    case \"LoopCond\": {\n      const pred = getParamValue(\"pred\", node, tensorMap, context);\n      return [cloneTensor(pred)];\n    }\n    case \"Switch\": {\n      const pred = getParamValue(\"pred\", node, tensorMap, context);\n      let data = getParamValue(\"data\", node, tensorMap, context);\n      if (!data.kept) {\n        data = cloneTensor(data);\n      }\n      return (await pred.data())[0] ? [void 0, data] : [data, void 0];\n    }\n    case \"Merge\": {\n      const inputName = node.inputNames.find((name) => getTensor(name, tensorMap, context) !== void 0);\n      if (inputName) {\n        const data = getTensor(inputName, tensorMap, context);\n        return [cloneTensor(data)];\n      }\n      return void 0;\n    }\n    case \"Enter\": {\n      const frameId = getParamValue(\"frameName\", node, tensorMap, context);\n      const data = getParamValue(\"tensor\", node, tensorMap, context);\n      context.enterFrame(frameId);\n      return [cloneTensor(data)];\n    }\n    case \"Exit\": {\n      const data = getParamValue(\"tensor\", node, tensorMap, context);\n      context.exitFrame();\n      return [cloneTensor(data)];\n    }\n    case \"NextIteration\": {\n      const data = getParamValue(\"tensor\", node, tensorMap, context);\n      context.nextIteration();\n      return [cloneTensor(data)];\n    }\n    case \"TensorArrayV3\": {\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      const dtype = getParamValue(\"dtype\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const dynamicSize = getParamValue(\"dynamicSize\", node, tensorMap, context);\n      const clearAfterRead = getParamValue(\"clearAfterRead\", node, tensorMap, context);\n      const identicalElementShapes = getParamValue(\"identicalElementShapes\", node, tensorMap, context);\n      const name = getParamValue(\"name\", node, tensorMap, context);\n      const tensorArray = new TensorArray(name, dtype, size, elementShape, identicalElementShapes, dynamicSize, clearAfterRead);\n      context.addTensorArray(tensorArray);\n      return [tensorArray.idTensor, scalar(1)];\n    }\n    case \"TensorArrayWriteV3\": {\n      const id = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const index = getParamValue(\"index\", node, tensorMap, context);\n      const writeTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const writeTensorArray = context.getTensorArray(id.id);\n      writeTensorArray.write(index, writeTensor);\n      return [writeTensorArray.idTensor];\n    }\n    case \"TensorArrayReadV3\": {\n      const readId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const readIndex = getParamValue(\"index\", node, tensorMap, context);\n      const readTensorArray = context.getTensorArray(readId.id);\n      return [readTensorArray.read(readIndex)];\n    }\n    case \"TensorArrayGatherV3\": {\n      const gatherId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const gatherIndices = getParamValue(\"indices\", node, tensorMap, context);\n      const gatherDtype = getParamValue(\"dtype\", node, tensorMap, context);\n      const gatherTensorArray = context.getTensorArray(gatherId.id);\n      return [gatherTensorArray.gather(gatherIndices, gatherDtype)];\n    }\n    case \"TensorArrayScatterV3\": {\n      const scatterId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const scatterIndices = getParamValue(\"indices\", node, tensorMap, context);\n      const scatterTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const scatterTensorArray = context.getTensorArray(scatterId.id);\n      scatterTensorArray.scatter(scatterIndices, scatterTensor);\n      return [scatterTensorArray.idTensor];\n    }\n    case \"TensorArrayConcatV3\": {\n      const concatId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const concatTensorArray = context.getTensorArray(concatId.id);\n      const concatDtype = getParamValue(\"dtype\", node, tensorMap, context);\n      return [concatTensorArray.concat(concatDtype)];\n    }\n    case \"TensorArraySplitV3\": {\n      const splitId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const splitTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const lengths = getParamValue(\"lengths\", node, tensorMap, context);\n      const splitTensorArray = context.getTensorArray(splitId.id);\n      splitTensorArray.split(lengths, splitTensor);\n      return [splitTensorArray.idTensor];\n    }\n    case \"TensorArraySizeV3\": {\n      const sizeId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const sizeTensorArray = context.getTensorArray(sizeId.id);\n      return [scalar(sizeTensorArray.size(), \"int32\")];\n    }\n    case \"TensorArrayCloseV3\": {\n      const closeId = getParamValue(\"tensorArrayId\", node, tensorMap, context);\n      const closeTensorArray = context.getTensorArray(closeId.id);\n      closeTensorArray.clearAndClose();\n      return [closeTensorArray.idTensor];\n    }\n    case \"TensorListSetItem\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const index = getParamValue(\"index\", node, tensorMap, context);\n      const writeTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      tensorList.setItem(index, writeTensor);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListGetItem\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const readIndex = getParamValue(\"index\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDType = getParamValue(\"elementDType\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      return [tensorList.getItem(readIndex, elementShape, elementDType)];\n    }\n    case \"TensorListScatterV2\":\n    case \"TensorListScatter\": {\n      const scatterIndices = getParamValue(\"indices\", node, tensorMap, context);\n      const scatterTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const numElements = getParamValue(\"numElements\", node, tensorMap, context);\n      const tensorList = scatter(scatterTensor, scatterIndices, elementShape, numElements);\n      context.addTensorList(tensorList);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListReserve\":\n    case \"EmptyTensorList\": {\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDtype = getParamValue(\"elementDType\", node, tensorMap, context);\n      let numElementsParam;\n      if (node.op === \"TensorListReserve\") {\n        numElementsParam = \"numElements\";\n      } else {\n        numElementsParam = \"maxNumElements\";\n      }\n      const numElements = getParamValue(numElementsParam, node, tensorMap, context);\n      const maxNumElements = node.op === \"TensorListReserve\" ? -1 : numElements;\n      const tensorList = reserve(elementShape, elementDtype, numElements, maxNumElements);\n      context.addTensorList(tensorList);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListGather\": {\n      const gatherId = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const gatherIndices = getParamValue(\"indices\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDtype = getParamValue(\"elementDType\", node, tensorMap, context);\n      const tensorList = context.getTensorList(gatherId.id);\n      return [tensorList.gather(gatherIndices, elementDtype, elementShape)];\n    }\n    case \"TensorListStack\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDtype = getParamValue(\"elementDType\", node, tensorMap, context);\n      const numElements = getParamValue(\"numElements\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      return [tensorList.stack(elementShape, elementDtype, numElements)];\n    }\n    case \"TensorListFromTensor\": {\n      const tensor2 = getParamValue(\"tensor\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDtype = getParamValue(\"elementDType\", node, tensorMap, context);\n      const tensorList = fromTensor(tensor2, elementShape, elementDtype);\n      context.addTensorList(tensorList);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListConcat\":\n    case \"TensorListConcatV2\": {\n      const concatId = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const tensorList = context.getTensorList(concatId.id);\n      const concatDtype = getParamValue(\"dtype\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      return [tensorList.concat(concatDtype, elementShape)];\n    }\n    case \"TensorListPushBack\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const writeTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      tensorList.pushBack(writeTensor);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListPopBack\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const elementDType = getParamValue(\"elementDType\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      return [tensorList.popBack(elementShape, elementDType)];\n    }\n    case \"TensorListSplit\": {\n      const splitTensor = getParamValue(\"tensor\", node, tensorMap, context);\n      const elementShape = getParamValue(\"elementShape\", node, tensorMap, context);\n      const lengths = getParamValue(\"lengths\", node, tensorMap, context);\n      const tensorList = split2(splitTensor, lengths, elementShape);\n      context.addTensorList(tensorList);\n      return [tensorList.idTensor];\n    }\n    case \"TensorListLength\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const tensorList = context.getTensorList(idTensor.id);\n      return [scalar(tensorList.size(), \"int32\")];\n    }\n    case \"TensorListResize\": {\n      const idTensor = getParamValue(\"tensorListId\", node, tensorMap, context);\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      const srcTensorList = context.getTensorList(idTensor.id);\n      const destTensorList = srcTensorList.resize(size);\n      context.addTensorList(destTensorList);\n      return [destTensorList.idTensor];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/convolution_executor.js\nfunction fusedConvAndDepthWiseParams(node, tensorMap, context) {\n  const [extraOp, activationFunc] = getParamValue(\"fusedOps\", node, tensorMap, context);\n  const isBiasAdd = extraOp === \"biasadd\";\n  const noBiasAdd = !isBiasAdd;\n  const isPrelu = activationFunc === \"prelu\";\n  const isBatchNorm = extraOp === \"fusedbatchnorm\";\n  const numArgs = getParamValue(\"numArgs\", node, tensorMap, context);\n  if (isBiasAdd) {\n    if (isPrelu && numArgs !== 2) {\n      throw new Error(\"FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.\");\n    }\n    if (!isPrelu && isBiasAdd && numArgs !== 1) {\n      throw new Error(\"FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.\");\n    }\n  }\n  if (isBatchNorm) {\n    throw new Error(\"FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported\");\n  }\n  const stride = getParamValue(\"strides\", node, tensorMap, context);\n  const pad3 = getPadding(node, tensorMap, context);\n  const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n  const dilations = getParamValue(\"dilations\", node, tensorMap, context);\n  let [biasArg, preluArg] = getParamValue(\"args\", node, tensorMap, context);\n  if (noBiasAdd) {\n    preluArg = biasArg;\n    biasArg = void 0;\n  }\n  const leakyreluAlpha = getParamValue(\"leakyreluAlpha\", node, tensorMap, context);\n  return {\n    stride,\n    pad: pad3,\n    dataFormat,\n    dilations,\n    biasArg,\n    preluArg,\n    activationFunc,\n    leakyreluAlpha\n  };\n}\nvar executeOp4 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Conv1D\": {\n      const stride = getParamValue(\"stride\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n      const dilation = getParamValue(\"dilation\", node, tensorMap, context);\n      return [ops.conv1d(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), stride, pad3, dataFormat, dilation)];\n    }\n    case \"Conv2D\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getPadding(node, tensorMap, context);\n      const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n      const dilations = getParamValue(\"dilations\", node, tensorMap, context);\n      return [ops.conv2d(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];\n    }\n    case \"_FusedConv2D\": {\n      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);\n      return [ops.fused.conv2d({\n        x: getParamValue(\"x\", node, tensorMap, context),\n        filter: getParamValue(\"filter\", node, tensorMap, context),\n        strides: [stride[1], stride[2]],\n        pad: pad3,\n        dataFormat,\n        dilations: [dilations[1], dilations[2]],\n        bias: biasArg,\n        activation: activationFunc,\n        preluActivationWeights: preluArg,\n        leakyreluAlpha\n      })];\n    }\n    case \"FusedDepthwiseConv2dNative\": {\n      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);\n      return [ops.fused.depthwiseConv2d({\n        x: getParamValue(\"x\", node, tensorMap, context),\n        filter: getParamValue(\"filter\", node, tensorMap, context),\n        strides: [stride[1], stride[2]],\n        pad: pad3,\n        dataFormat,\n        dilations: [dilations[1], dilations[2]],\n        bias: biasArg,\n        activation: activationFunc,\n        preluActivationWeights: preluArg,\n        leakyreluAlpha\n      })];\n    }\n    case \"Conv2DBackpropInput\":\n    case \"Conv2dTranspose\": {\n      const shape = getParamValue(\"outputShape\", node, tensorMap, context);\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getPadding(node, tensorMap, context);\n      return [ops.conv2dTranspose(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), shape, [stride[1], stride[2]], pad3)];\n    }\n    case \"DepthwiseConv2dNative\":\n    case \"DepthwiseConv2d\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getPadding(node, tensorMap, context);\n      const dilations = getParamValue(\"dilations\", node, tensorMap, context);\n      const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n      return [ops.depthwiseConv2d(getParamValue(\"input\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];\n    }\n    case \"Conv3D\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n      const dilations = getParamValue(\"dilations\", node, tensorMap, context);\n      return [ops.conv3d(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), [stride[1], stride[2], stride[3]], pad3, dataFormat, [dilations[1], dilations[2], dilations[3]])];\n    }\n    case \"AvgPool\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const kernelSize = getParamValue(\"kernelSize\", node, tensorMap, context);\n      return [ops.avgPool(getParamValue(\"x\", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];\n    }\n    case \"MaxPool\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const kernelSize = getParamValue(\"kernelSize\", node, tensorMap, context);\n      return [ops.maxPool(getParamValue(\"x\", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];\n    }\n    case \"MaxPoolWithArgmax\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const kernelSize = getParamValue(\"kernelSize\", node, tensorMap, context);\n      const includeBatchInIndex = getParamValue(\"includeBatchInIndex\", node, tensorMap, context);\n      const { result, indexes } = ops.maxPoolWithArgmax(getParamValue(\"x\", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3, includeBatchInIndex);\n      return [result, indexes];\n    }\n    case \"AvgPool3D\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const kernelSize = getParamValue(\"kernelSize\", node, tensorMap, context);\n      return [ops.avgPool3d(getParamValue(\"x\", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];\n    }\n    case \"MaxPool3D\": {\n      const stride = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const kernelSize = getParamValue(\"kernelSize\", node, tensorMap, context);\n      return [ops.maxPool3d(getParamValue(\"x\", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];\n    }\n    case \"Dilation2D\": {\n      const strides = getParamValue(\"strides\", node, tensorMap, context);\n      const pad3 = getParamValue(\"pad\", node, tensorMap, context);\n      const dilations = getParamValue(\"dilations\", node, tensorMap, context);\n      const strideHeight = strides[1];\n      const strideWidth = strides[2];\n      const dilationHeight = dilations[1];\n      const dilationWidth = dilations[2];\n      return [ops.dilation2d(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"filter\", node, tensorMap, context), [strideHeight, strideWidth], pad3, [dilationHeight, dilationWidth], \"NHWC\")];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/creation_executor.js\nvar executeOp5 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Fill\": {\n      const shape = getParamValue(\"shape\", node, tensorMap, context);\n      const dtype = getParamValue(\"dtype\", node, tensorMap, context);\n      const value = getParamValue(\"value\", node, tensorMap, context);\n      return [ops.fill(shape, value, dtype)];\n    }\n    case \"LinSpace\": {\n      const start = getParamValue(\"start\", node, tensorMap, context);\n      const stop = getParamValue(\"stop\", node, tensorMap, context);\n      const num = getParamValue(\"num\", node, tensorMap, context);\n      return [ops.linspace(start, stop, num)];\n    }\n    case \"Multinomial\": {\n      const logits = getParamValue(\"logits\", node, tensorMap, context);\n      const numSamples = getParamValue(\"numSamples\", node, tensorMap, context);\n      const seed = getParamValue(\"seed\", node, tensorMap, context);\n      return [ops.multinomial(logits, numSamples, seed)];\n    }\n    case \"OneHot\": {\n      const indices = getParamValue(\"indices\", node, tensorMap, context);\n      const depth = getParamValue(\"depth\", node, tensorMap, context);\n      const onValue = getParamValue(\"onValue\", node, tensorMap, context);\n      const offValue = getParamValue(\"offValue\", node, tensorMap, context);\n      const dtype = getParamValue(\"dtype\", node, tensorMap, context);\n      return [ops.oneHot(indices, depth, onValue, offValue, dtype)];\n    }\n    case \"Ones\": {\n      return [ops.ones(getParamValue(\"shape\", node, tensorMap, context), getParamValue(\"dtype\", node, tensorMap, context))];\n    }\n    case \"OnesLike\": {\n      return [ops.onesLike(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"RandomStandardNormal\": {\n      return [ops.randomStandardNormal(getParamValue(\"shape\", node, tensorMap, context), getParamValue(\"dtype\", node, tensorMap, context), getParamValue(\"seed\", node, tensorMap, context))];\n    }\n    case \"RandomUniform\": {\n      return [ops.randomUniform(\n        getParamValue(\"shape\", node, tensorMap, context),\n        getParamValue(\"minval\", node, tensorMap, context),\n        getParamValue(\"maxval\", node, tensorMap, context),\n        getParamValue(\"dtype\", node, tensorMap, context)\n      )];\n    }\n    case \"Range\": {\n      const start = getParamValue(\"start\", node, tensorMap, context);\n      const stop = getParamValue(\"stop\", node, tensorMap, context);\n      const step5 = getParamValue(\"step\", node, tensorMap, context);\n      return [ops.range(start, stop, step5, getParamValue(\"dtype\", node, tensorMap, context))];\n    }\n    case \"TruncatedNormal\": {\n      const shape = getParamValue(\"shape\", node, tensorMap, context);\n      const mean4 = getParamValue(\"mean\", node, tensorMap, context);\n      const stdDev = getParamValue(\"stdDev\", node, tensorMap, context);\n      const seed = getParamValue(\"seed\", node, tensorMap, context);\n      return [ops.truncatedNormal(shape, mean4, stdDev, getParamValue(\"dtype\", node, tensorMap, context), seed)];\n    }\n    case \"Zeros\": {\n      return [ops.zeros(getParamValue(\"shape\", node, tensorMap, context), getParamValue(\"dtype\", node, tensorMap, context))];\n    }\n    case \"ZerosLike\": {\n      return [ops.zerosLike(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/dynamic_executor.js\nfunction nmsParams(node, tensorMap, context) {\n  const boxes = getParamValue(\"boxes\", node, tensorMap, context);\n  const scores = getParamValue(\"scores\", node, tensorMap, context);\n  const maxOutputSize = getParamValue(\"maxOutputSize\", node, tensorMap, context);\n  const iouThreshold = getParamValue(\"iouThreshold\", node, tensorMap, context);\n  const scoreThreshold = getParamValue(\"scoreThreshold\", node, tensorMap, context);\n  const softNmsSigma = getParamValue(\"softNmsSigma\", node, tensorMap, context);\n  return {\n    boxes,\n    scores,\n    maxOutputSize,\n    iouThreshold,\n    scoreThreshold,\n    softNmsSigma\n  };\n}\nvar executeOp6 = async (node, tensorMap, context, resourceManager, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"NonMaxSuppressionV5\": {\n      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = nmsParams(node, tensorMap, context);\n      const result = await ops.image.nonMaxSuppressionWithScoreAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);\n      return [result.selectedIndices, result.selectedScores];\n    }\n    case \"NonMaxSuppressionV4\": {\n      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);\n      const padToMaxOutputSize = getParamValue(\"padToMaxOutputSize\", node, tensorMap, context);\n      const result = await ops.image.nonMaxSuppressionPaddedAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);\n      return [result.selectedIndices, result.validOutputs];\n    }\n    case \"NonMaxSuppressionV3\":\n    case \"NonMaxSuppressionV2\": {\n      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);\n      return [await ops.image.nonMaxSuppressionAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold)];\n    }\n    case \"Where\": {\n      const condition = ops.cast(getParamValue(\"condition\", node, tensorMap, context), \"bool\");\n      const result = [await ops.whereAsync(condition)];\n      condition.dispose();\n      return result;\n    }\n    case \"ListDiff\": {\n      return ops.setdiff1dAsync(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"y\", node, tensorMap, context));\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/evaluation_executor.js\nvar executeOp7 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"LowerBound\": {\n      const sortedSequence = getParamValue(\"sortedSequence\", node, tensorMap, context);\n      const values = getParamValue(\"values\", node, tensorMap, context);\n      return [ops.lowerBound(sortedSequence, values)];\n    }\n    case \"TopKV2\": {\n      const x = getParamValue(\"x\", node, tensorMap, context);\n      const k = getParamValue(\"k\", node, tensorMap, context);\n      const sorted = getParamValue(\"sorted\", node, tensorMap, context);\n      const result = ops.topk(x, k, sorted);\n      return [result.values, result.indices];\n    }\n    case \"UpperBound\": {\n      const sortedSequence = getParamValue(\"sortedSequence\", node, tensorMap, context);\n      const values = getParamValue(\"values\", node, tensorMap, context);\n      return [ops.upperBound(sortedSequence, values)];\n    }\n    case \"Unique\": {\n      const x = getParamValue(\"x\", node, tensorMap, context);\n      const result = ops.unique(x);\n      return [result.values, result.indices];\n    }\n    case \"UniqueV2\": {\n      const x = getParamValue(\"x\", node, tensorMap, context);\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const result = ops.unique(x, axis);\n      return [result.values, result.indices];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/graph_executor.js\nvar executeOp8 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Const\": {\n      return tensorMap[node.name];\n    }\n    case \"PlaceholderWithDefault\":\n      const def = getParamValue(\"default\", node, tensorMap, context);\n      return [getTensor(node.name, tensorMap, context) || def];\n    case \"Placeholder\":\n      return [getTensor(node.name, tensorMap, context)];\n    case \"Identity\":\n    case \"StopGradient\":\n    case \"FakeQuantWithMinMaxVars\": {\n      const data2 = getParamValue(\"x\", node, tensorMap, context);\n      return [cloneTensor(data2)];\n    }\n    case \"IdentityN\":\n      return getParamValue(\"x\", node, tensorMap, context).map((t) => cloneTensor(t));\n    case \"Snapshot\":\n      const snapshot = getParamValue(\"x\", node, tensorMap, context);\n      return [cloneTensor(snapshot)];\n    case \"Shape\":\n      return [ops.tensor1d(getParamValue(\"x\", node, tensorMap, context).shape, \"int32\")];\n    case \"ShapeN\":\n      return getParamValue(\"x\", node, tensorMap, context).map((t) => ops.tensor1d(t.shape));\n    case \"Size\":\n      return [ops.scalar(getParamValue(\"x\", node, tensorMap, context).size, \"int32\")];\n    case \"Rank\":\n      return [ops.scalar(getParamValue(\"x\", node, tensorMap, context).rank, \"int32\")];\n    case \"NoOp\":\n      return [ops.scalar(1)];\n    case \"Print\":\n      const input2 = getParamValue(\"x\", node, tensorMap, context);\n      const data = getParamValue(\"data\", node, tensorMap, context);\n      const message = getParamValue(\"message\", node, tensorMap, context);\n      const summarize = getParamValue(\"summarize\", node, tensorMap, context);\n      console.warn(\"The graph has a tf.print() operation,usually used for debugging, which slows down performance.\");\n      console.log(message);\n      for (let i = 0; i < data.length; i++) {\n        console.log(Array.prototype.slice.call(data[i].dataSync()).slice(0, summarize));\n      }\n      return [input2];\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/hash_table.js\nvar HashTable = class {\n  constructor(keyDType, valueDType) {\n    this.keyDType = keyDType;\n    this.valueDType = valueDType;\n    this.handle = scalar(0);\n    this.tensorMap = /* @__PURE__ */ new Map();\n    keep(this.handle);\n  }\n  get id() {\n    return this.handle.id;\n  }\n  clearAndClose() {\n    this.tensorMap.forEach((value) => value.dispose());\n    this.tensorMap.clear();\n    this.handle.dispose();\n  }\n  size() {\n    return this.tensorMap.size;\n  }\n  tensorSize() {\n    return scalar(this.size(), \"int32\");\n  }\n  async import(keys, values) {\n    this.checkKeyAndValueTensor(keys, values);\n    const $keys = await keys.data();\n    this.tensorMap.forEach((value) => value.dispose());\n    this.tensorMap.clear();\n    return tidy(() => {\n      const $values = unstack(values);\n      const keysLength = $keys.length;\n      const valuesLength = $values.length;\n      util_exports.assert(keysLength === valuesLength, () => `The number of elements doesn't match, keys has ${keysLength} elements, the values has ${valuesLength} elements.`);\n      for (let i = 0; i < keysLength; i++) {\n        const key = $keys[i];\n        const value = $values[i];\n        keep(value);\n        this.tensorMap.set(key, value);\n      }\n      return this.handle;\n    });\n  }\n  async find(keys, defaultValue) {\n    this.checkKeyAndValueTensor(keys, defaultValue);\n    const $keys = await keys.data();\n    return tidy(() => {\n      const result = [];\n      for (let i = 0; i < $keys.length; i++) {\n        const key = $keys[i];\n        const value = this.findWithDefault(key, defaultValue);\n        result.push(value);\n      }\n      return stack(result);\n    });\n  }\n  findWithDefault(key, defaultValue) {\n    const result = this.tensorMap.get(key);\n    return result != null ? result : defaultValue;\n  }\n  checkKeyAndValueTensor(key, value) {\n    if (key.dtype !== this.keyDType) {\n      throw new Error(`Expect key dtype ${this.keyDType}, but got ${key.dtype}`);\n    }\n    if (value.dtype !== this.valueDType) {\n      throw new Error(`Expect value dtype ${this.valueDType}, but got ${value.dtype}`);\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/hash_table_executor.js\nvar executeOp9 = async (node, tensorMap, context, resourceManager) => {\n  switch (node.op) {\n    case \"HashTable\":\n    case \"HashTableV2\": {\n      const existingTableHandle = resourceManager.getHashTableHandleByName(node.name);\n      if (existingTableHandle != null) {\n        return [existingTableHandle];\n      } else {\n        const keyDType = getParamValue(\"keyDType\", node, tensorMap, context);\n        const valueDType = getParamValue(\"valueDType\", node, tensorMap, context);\n        const hashTable = new HashTable(keyDType, valueDType);\n        resourceManager.addHashTable(node.name, hashTable);\n        return [hashTable.handle];\n      }\n    }\n    case \"LookupTableImport\":\n    case \"LookupTableImportV2\": {\n      const handle = getParamValue(\"tableHandle\", node, tensorMap, context, resourceManager);\n      const keys = getParamValue(\"keys\", node, tensorMap, context);\n      const values = getParamValue(\"values\", node, tensorMap, context);\n      const hashTable = resourceManager.getHashTableById(handle.id);\n      return [await hashTable.import(keys, values)];\n    }\n    case \"LookupTableFind\":\n    case \"LookupTableFindV2\": {\n      const handle = getParamValue(\"tableHandle\", node, tensorMap, context, resourceManager);\n      const keys = getParamValue(\"keys\", node, tensorMap, context);\n      const defaultValue = getParamValue(\"defaultValue\", node, tensorMap, context);\n      const hashTable = resourceManager.getHashTableById(handle.id);\n      return [await hashTable.find(keys, defaultValue)];\n    }\n    case \"LookupTableSize\":\n    case \"LookupTableSizeV2\": {\n      const handle = getParamValue(\"tableHandle\", node, tensorMap, context, resourceManager);\n      const hashTable = resourceManager.getHashTableById(handle.id);\n      return [hashTable.tensorSize()];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/image_executor.js\nvar executeOp10 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"ResizeBilinear\": {\n      const images = getParamValue(\"images\", node, tensorMap, context);\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      const alignCorners = getParamValue(\"alignCorners\", node, tensorMap, context);\n      const halfPixelCenters = getParamValue(\"halfPixelCenters\", node, tensorMap, context);\n      return [ops.image.resizeBilinear(images, [size[0], size[1]], alignCorners, halfPixelCenters)];\n    }\n    case \"ResizeNearestNeighbor\": {\n      const images = getParamValue(\"images\", node, tensorMap, context);\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      const alignCorners = getParamValue(\"alignCorners\", node, tensorMap, context);\n      const halfPixelCenters = getParamValue(\"halfPixelCenters\", node, tensorMap, context);\n      return [ops.image.resizeNearestNeighbor(images, [size[0], size[1]], alignCorners, halfPixelCenters)];\n    }\n    case \"CropAndResize\": {\n      const image2 = getParamValue(\"image\", node, tensorMap, context);\n      const boxes = getParamValue(\"boxes\", node, tensorMap, context);\n      const boxInd = getParamValue(\"boxInd\", node, tensorMap, context);\n      const cropSize = getParamValue(\"cropSize\", node, tensorMap, context);\n      const method = getParamValue(\"method\", node, tensorMap, context);\n      const extrapolationValue = getParamValue(\"extrapolationValue\", node, tensorMap, context);\n      return [ops.image.cropAndResize(image2, boxes, boxInd, cropSize, method, extrapolationValue)];\n    }\n    case \"ImageProjectiveTransformV3\": {\n      const images = getParamValue(\"images\", node, tensorMap, context);\n      const transforms = getParamValue(\"transforms\", node, tensorMap, context);\n      const outputShape = getParamValue(\"outputShape\", node, tensorMap, context);\n      const fillValue = getParamValue(\"fillValue\", node, tensorMap, context);\n      const interpolation = getParamValue(\"interpolation\", node, tensorMap, context);\n      const fillMode = getParamValue(\"fillMode\", node, tensorMap, context);\n      return [ops.image.transform(images, transforms, interpolation.toLowerCase(), fillMode.toLowerCase(), fillValue, outputShape)];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/logical_executor.js\nvar executeOp11 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Equal\": {\n      return [ops.equal(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"NotEqual\": {\n      return [ops.notEqual(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Greater\": {\n      return [ops.greater(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"GreaterEqual\": {\n      return [ops.greaterEqual(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Less\": {\n      return [ops.less(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"LessEqual\": {\n      return [ops.lessEqual(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"LogicalAnd\": {\n      return [ops.logicalAnd(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"LogicalNot\": {\n      return [ops.logicalNot(getParamValue(\"a\", node, tensorMap, context))];\n    }\n    case \"LogicalOr\": {\n      return [ops.logicalOr(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    case \"Select\":\n    case \"SelectV2\": {\n      return [ops.where(getParamValue(\"condition\", node, tensorMap, context), getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/matrices_executor.js\nvar executeOp12 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"BatchMatMul\":\n    case \"BatchMatMulV2\":\n    case \"MatMul\":\n      return [ops.matMul(getParamValue(\"a\", node, tensorMap, context), getParamValue(\"b\", node, tensorMap, context), getParamValue(\"transposeA\", node, tensorMap, context), getParamValue(\"transposeB\", node, tensorMap, context))];\n    case \"Einsum\":\n      return [ops.einsum(getParamValue(\"equation\", node, tensorMap, context), ...getParamValue(\"tensors\", node, tensorMap, context))];\n    case \"Transpose\":\n      return [ops.transpose(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"perm\", node, tensorMap, context))];\n    case \"_FusedMatMul\":\n      const [extraOp, activationFunc] = getParamValue(\"fusedOps\", node, tensorMap, context);\n      const isBiasAdd = extraOp === \"biasadd\";\n      const isPrelu = activationFunc === \"prelu\";\n      const numArgs = getParamValue(\"numArgs\", node, tensorMap, context);\n      const leakyreluAlpha = getParamValue(\"leakyreluAlpha\", node, tensorMap, context);\n      if (isBiasAdd) {\n        if (isPrelu && numArgs !== 2) {\n          throw new Error(\"Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.\");\n        }\n        if (!isPrelu && numArgs !== 1) {\n          throw new Error(\"Fused MatMul with BiasAdd must have one extra argument: bias.\");\n        }\n      }\n      const [biasArg, preluArg] = getParamValue(\"args\", node, tensorMap, context);\n      return [ops.fused.matMul({\n        a: getParamValue(\"a\", node, tensorMap, context),\n        b: getParamValue(\"b\", node, tensorMap, context),\n        transposeA: getParamValue(\"transposeA\", node, tensorMap, context),\n        transposeB: getParamValue(\"transposeB\", node, tensorMap, context),\n        bias: biasArg,\n        activation: activationFunc,\n        preluActivationWeights: preluArg,\n        leakyreluAlpha\n      })];\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/normalization_executor.js\nvar executeOp13 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"EuclideanNorm\":\n      return [ops.euclideanNorm(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"axis\", node, tensorMap, context), getParamValue(\"keepDims\", node, tensorMap, context))];\n    case \"FusedBatchNorm\":\n    case \"FusedBatchNormV2\": {\n      return [ops.batchNorm(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"mean\", node, tensorMap, context), getParamValue(\"variance\", node, tensorMap, context), getParamValue(\"offset\", node, tensorMap, context), getParamValue(\"scale\", node, tensorMap, context), getParamValue(\"epsilon\", node, tensorMap, context))];\n    }\n    case \"FusedBatchNormV3\": {\n      return [ops.batchNorm(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"mean\", node, tensorMap, context), getParamValue(\"variance\", node, tensorMap, context), getParamValue(\"offset\", node, tensorMap, context), getParamValue(\"scale\", node, tensorMap, context), getParamValue(\"epsilon\", node, tensorMap, context))];\n    }\n    case \"LRN\": {\n      return [ops.localResponseNormalization(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"radius\", node, tensorMap, context), getParamValue(\"bias\", node, tensorMap, context), getParamValue(\"alpha\", node, tensorMap, context), getParamValue(\"beta\", node, tensorMap, context))];\n    }\n    case \"Softmax\": {\n      return [ops.softmax(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"LogSoftmax\": {\n      return [ops.logSoftmax(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"SparseToDense\": {\n      return [ops.sparseToDense(getParamValue(\"sparseIndices\", node, tensorMap, context), getParamValue(\"outputShape\", node, tensorMap, context), getParamValue(\"sparseValues\", node, tensorMap, context), getParamValue(\"defaultValue\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/reduction_executor.js\nvar executeOp14 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Max\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.max(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"Mean\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.mean(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"Min\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.min(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"Sum\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.sum(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"All\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.all(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"Any\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.any(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"ArgMax\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      return [ops.argMax(getParamValue(\"x\", node, tensorMap, context), axis)];\n    }\n    case \"ArgMin\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      return [ops.argMin(getParamValue(\"x\", node, tensorMap, context), axis)];\n    }\n    case \"Prod\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const keepDims = getParamValue(\"keepDims\", node, tensorMap, context);\n      return [ops.prod(getParamValue(\"x\", node, tensorMap, context), axis, keepDims)];\n    }\n    case \"Cumprod\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const exclusive = getParamValue(\"exclusive\", node, tensorMap, context);\n      const reverse5 = getParamValue(\"reverse\", node, tensorMap, context);\n      return [ops.cumprod(getParamValue(\"x\", node, tensorMap, context), axis, exclusive, reverse5)];\n    }\n    case \"Cumsum\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const exclusive = getParamValue(\"exclusive\", node, tensorMap, context);\n      const reverse5 = getParamValue(\"reverse\", node, tensorMap, context);\n      return [ops.cumsum(getParamValue(\"x\", node, tensorMap, context), axis, exclusive, reverse5)];\n    }\n    case \"Bincount\":\n      const x = getParamValue(\"x\", node, tensorMap, context);\n      const weights = getParamValue(\"weights\", node, tensorMap, context);\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      return [ops.bincount(x, weights, size)];\n    case \"DenseBincount\": {\n      const x2 = getParamValue(\"x\", node, tensorMap, context);\n      const weights2 = getParamValue(\"weights\", node, tensorMap, context);\n      const size2 = getParamValue(\"size\", node, tensorMap, context);\n      const binaryOutput = getParamValue(\"binaryOutput\", node, tensorMap, context);\n      return [ops.denseBincount(x2, weights2, size2, binaryOutput)];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/slice_join_executor.js\nvar executeOp15 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"ConcatV2\":\n    case \"Concat\": {\n      const n = getParamValue(\"n\", node, tensorMap, context);\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      let inputs = getParamValue(\"tensors\", node, tensorMap, context);\n      inputs = inputs.slice(0, n);\n      return [ops.concat(inputs, axis)];\n    }\n    case \"Gather\": {\n      const input2 = getParamValue(\"x\", node, tensorMap, context);\n      const indices = getParamValue(\"indices\", node, tensorMap, context);\n      return [ops.gather(input2, ops.cast(indices, \"int32\"), 0)];\n    }\n    case \"GatherV2\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const batchDims = getParamValue(\"batchDims\", node, tensorMap, context);\n      const input2 = getParamValue(\"x\", node, tensorMap, context);\n      const indices = getParamValue(\"indices\", node, tensorMap, context);\n      return [ops.gather(input2, ops.cast(indices, \"int32\"), axis, batchDims)];\n    }\n    case \"Reverse\": {\n      const dims = getParamValue(\"dims\", node, tensorMap, context);\n      const axis = [];\n      for (let i = 0; i < dims.length; i++) {\n        if (dims[i]) {\n          axis.push(i);\n        }\n      }\n      const input2 = getParamValue(\"x\", node, tensorMap, context);\n      return [ops.reverse(input2, axis)];\n    }\n    case \"ReverseV2\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const input2 = getParamValue(\"x\", node, tensorMap, context);\n      return [ops.reverse(input2, axis)];\n    }\n    case \"Slice\": {\n      const begin = getParamValue(\"begin\", node, tensorMap, context);\n      const size = getParamValue(\"size\", node, tensorMap, context);\n      return [ops.slice(getParamValue(\"x\", node, tensorMap, context), begin, size)];\n    }\n    case \"StridedSlice\": {\n      const begin = getParamValue(\"begin\", node, tensorMap, context);\n      const end = getParamValue(\"end\", node, tensorMap, context);\n      const strides = getParamValue(\"strides\", node, tensorMap, context);\n      const beginMask = getParamValue(\"beginMask\", node, tensorMap, context);\n      const endMask = getParamValue(\"endMask\", node, tensorMap, context);\n      const ellipsisMask = getParamValue(\"ellipsisMask\", node, tensorMap, context);\n      const newAxisMask = getParamValue(\"newAxisMask\", node, tensorMap, context);\n      const shrinkAxisMask = getParamValue(\"shrinkAxisMask\", node, tensorMap, context);\n      const tensor2 = getParamValue(\"x\", node, tensorMap, context);\n      return [ops.stridedSlice(tensor2, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask)];\n    }\n    case \"Pack\": {\n      return tidy(() => {\n        const axis = getParamValue(\"axis\", node, tensorMap, context);\n        const tensors = getParamValue(\"tensors\", node, tensorMap, context);\n        const shape = tensors[0].shape;\n        const squeezedShape = ops.squeeze(tensors[0]).shape;\n        const mapped = tensors.map((tensor2) => {\n          const sameShape = util_exports.arraysEqual(tensor2.shape, shape);\n          if (!sameShape && !util_exports.arraysEqual(ops.squeeze(tensor2).shape, squeezedShape)) {\n            throw new Error(\"the input tensors shape does not match\");\n          }\n          return sameShape ? tensor2 : ops.reshape(tensor2, shape);\n        });\n        return [ops.stack(mapped, axis)];\n      });\n    }\n    case \"Unpack\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const tensor2 = getParamValue(\"tensor\", node, tensorMap, context);\n      return ops.unstack(tensor2, axis);\n    }\n    case \"Tile\": {\n      const reps = getParamValue(\"reps\", node, tensorMap, context);\n      return [ops.tile(getParamValue(\"x\", node, tensorMap, context), reps)];\n    }\n    case \"Split\":\n    case \"SplitV\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      const numOrSizeSplits = getParamValue(\"numOrSizeSplits\", node, tensorMap, context);\n      const tensor2 = getParamValue(\"x\", node, tensorMap, context);\n      return ops.split(tensor2, numOrSizeSplits, axis);\n    }\n    case \"ScatterNd\": {\n      const indices = getParamValue(\"indices\", node, tensorMap, context);\n      const values = getParamValue(\"values\", node, tensorMap, context);\n      const shape = getParamValue(\"shape\", node, tensorMap, context);\n      return [ops.scatterND(indices, values, shape)];\n    }\n    case \"GatherNd\": {\n      const x = getParamValue(\"x\", node, tensorMap, context);\n      const indices = getParamValue(\"indices\", node, tensorMap, context);\n      return [ops.gatherND(x, indices)];\n    }\n    case \"SparseToDense\": {\n      const indices = getParamValue(\"sparseIndices\", node, tensorMap, context);\n      const shape = getParamValue(\"outputShape\", node, tensorMap, context);\n      const sparseValues = getParamValue(\"sparseValues\", node, tensorMap, context);\n      const defaultValue = getParamValue(\"defaultValue\", node, tensorMap, context);\n      return [ops.sparseToDense(indices, sparseValues, shape, sparseValues.dtype === defaultValue.dtype ? defaultValue : ops.cast(defaultValue, sparseValues.dtype))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/sparse_executor.js\nvar executeOp16 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"SparseFillEmptyRows\": {\n      const { outputIndices, outputValues, emptyRowIndicator, reverseIndexMap } = ops.sparse.sparseFillEmptyRows(getParamValue(\"indices\", node, tensorMap, context), getParamValue(\"values\", node, tensorMap, context), getParamValue(\"denseShape\", node, tensorMap, context), getParamValue(\"defaultValue\", node, tensorMap, context));\n      return [\n        outputIndices,\n        outputValues,\n        emptyRowIndicator,\n        reverseIndexMap\n      ];\n    }\n    case \"SparseReshape\": {\n      const { outputIndices, outputShape } = ops.sparse.sparseReshape(getParamValue(\"inputIndices\", node, tensorMap, context), getParamValue(\"inputShape\", node, tensorMap, context), getParamValue(\"newShape\", node, tensorMap, context));\n      return [outputIndices, outputShape];\n    }\n    case \"SparseSegmentMean\": {\n      const outputData = ops.sparse.sparseSegmentMean(getParamValue(\"data\", node, tensorMap, context), getParamValue(\"indices\", node, tensorMap, context), getParamValue(\"segmentIds\", node, tensorMap, context));\n      return [outputData];\n    }\n    case \"SparseSegmentSum\": {\n      const outputData = ops.sparse.sparseSegmentSum(getParamValue(\"data\", node, tensorMap, context), getParamValue(\"indices\", node, tensorMap, context), getParamValue(\"segmentIds\", node, tensorMap, context));\n      return [outputData];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/spectral_executor.js\nvar executeOp17 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"FFT\": {\n      return [ops.fft(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"IFFT\": {\n      return [ops.ifft(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"RFFT\": {\n      return [ops.rfft(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    case \"IRFFT\": {\n      return [ops.irfft(getParamValue(\"x\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/string_executor.js\nvar executeOp18 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"StringNGrams\": {\n      const { nGrams, nGramsSplits } = ops.string.stringNGrams(getParamValue(\"data\", node, tensorMap, context), getParamValue(\"dataSplits\", node, tensorMap, context), getParamValue(\"separator\", node, tensorMap, context), getParamValue(\"nGramWidths\", node, tensorMap, context), getParamValue(\"leftPad\", node, tensorMap, context), getParamValue(\"rightPad\", node, tensorMap, context), getParamValue(\"padWidth\", node, tensorMap, context), getParamValue(\"preserveShortSequences\", node, tensorMap, context));\n      return [nGrams, nGramsSplits];\n    }\n    case \"StringSplit\": {\n      const { indices, values, shape } = ops.string.stringSplit(getParamValue(\"input\", node, tensorMap, context), getParamValue(\"delimiter\", node, tensorMap, context), getParamValue(\"skipEmpty\", node, tensorMap, context));\n      return [indices, values, shape];\n    }\n    case \"StringToHashBucketFast\": {\n      const output = ops.string.stringToHashBucketFast(getParamValue(\"input\", node, tensorMap, context), getParamValue(\"numBuckets\", node, tensorMap, context));\n      return [output];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/transformation_executor.js\nvar executeOp19 = (node, tensorMap, context, ops = ops_for_converter_exports) => {\n  switch (node.op) {\n    case \"Cast\": {\n      return [ops.cast(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"dtype\", node, tensorMap, context))];\n    }\n    case \"ExpandDims\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      return [ops.expandDims(getParamValue(\"x\", node, tensorMap, context), axis)];\n    }\n    case \"Squeeze\": {\n      const axis = getParamValue(\"axis\", node, tensorMap, context);\n      return [ops.squeeze(getParamValue(\"x\", node, tensorMap, context), axis)];\n    }\n    case \"Reshape\": {\n      return [ops.reshape(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"shape\", node, tensorMap, context))];\n    }\n    case \"MirrorPad\": {\n      return [ops.mirrorPad(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"padding\", node, tensorMap, context), getParamValue(\"mode\", node, tensorMap, context))];\n    }\n    case \"PadV2\":\n    case \"Pad\": {\n      return [ops.pad(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"padding\", node, tensorMap, context), getParamValue(\"constantValue\", node, tensorMap, context))];\n    }\n    case \"SpaceToBatchND\": {\n      const blockShape = getParamValue(\"blockShape\", node, tensorMap, context);\n      const paddings = getParamValue(\"paddings\", node, tensorMap, context);\n      return [ops.spaceToBatchND(getParamValue(\"x\", node, tensorMap, context), blockShape, paddings)];\n    }\n    case \"BatchToSpaceND\": {\n      const blockShape = getParamValue(\"blockShape\", node, tensorMap, context);\n      const crops = getParamValue(\"crops\", node, tensorMap, context);\n      return [ops.batchToSpaceND(getParamValue(\"x\", node, tensorMap, context), blockShape, crops)];\n    }\n    case \"DepthToSpace\": {\n      const blockSize = getParamValue(\"blockSize\", node, tensorMap, context);\n      const dataFormat = getParamValue(\"dataFormat\", node, tensorMap, context).toUpperCase();\n      return [ops.depthToSpace(getParamValue(\"x\", node, tensorMap, context), blockSize, dataFormat)];\n    }\n    case \"BroadcastTo\": {\n      return [ops.broadcastTo(getParamValue(\"x\", node, tensorMap, context), getParamValue(\"shape\", node, tensorMap, context))];\n    }\n    case \"BroadcastArgs\": {\n      return [ops.broadcastArgs(getParamValue(\"s0\", node, tensorMap, context), getParamValue(\"s1\", node, tensorMap, context))];\n    }\n    default:\n      throw TypeError(`Node type ${node.op} is not implemented`);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/operations/operation_executor.js\nfunction executeOp20(node, tensorMap, context, resourceManager, tidy2 = tidy) {\n  const value = ((node2, tensorMap2, context2) => {\n    switch (node2.category) {\n      case \"arithmetic\":\n        return tidy2(() => executeOp(node2, tensorMap2, context2));\n      case \"basic_math\":\n        return tidy2(() => executeOp2(node2, tensorMap2, context2));\n      case \"control\":\n        return executeOp3(node2, tensorMap2, context2);\n      case \"convolution\":\n        return tidy2(() => executeOp4(node2, tensorMap2, context2));\n      case \"creation\":\n        return tidy2(() => executeOp5(node2, tensorMap2, context2));\n      case \"dynamic\":\n        return executeOp6(node2, tensorMap2, context2);\n      case \"evaluation\":\n        return tidy2(() => executeOp7(node2, tensorMap2, context2));\n      case \"image\":\n        return tidy2(() => executeOp10(node2, tensorMap2, context2));\n      case \"graph\":\n        return tidy2(() => executeOp8(node2, tensorMap2, context2));\n      case \"logical\":\n        return tidy2(() => executeOp11(node2, tensorMap2, context2));\n      case \"matrices\":\n        return tidy2(() => executeOp12(node2, tensorMap2, context2));\n      case \"normalization\":\n        return tidy2(() => executeOp13(node2, tensorMap2, context2));\n      case \"reduction\":\n        return tidy2(() => executeOp14(node2, tensorMap2, context2));\n      case \"slice_join\":\n        return tidy2(() => executeOp15(node2, tensorMap2, context2));\n      case \"sparse\":\n        return tidy2(() => executeOp16(node2, tensorMap2, context2));\n      case \"spectral\":\n        return tidy2(() => executeOp17(node2, tensorMap2, context2));\n      case \"string\":\n        return tidy2(() => executeOp18(node2, tensorMap2, context2));\n      case \"transformation\":\n        return tidy2(() => executeOp19(node2, tensorMap2, context2));\n      case \"hash_table\":\n        return executeOp9(node2, tensorMap2, context2, resourceManager);\n      case \"custom\":\n        const opMapper = getRegisteredOp(node2.op);\n        if (opMapper && opMapper.customExecutor) {\n          return opMapper.customExecutor(new NodeValueImpl(node2, tensorMap2, context2));\n        } else {\n          throw TypeError(`Custom op ${node2.op} is not registered.`);\n        }\n      default:\n        throw TypeError(`Unknown op '${node2.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`);\n    }\n  })(node, tensorMap, context);\n  if (util_exports.isPromise(value)) {\n    return value.then((data) => [].concat(data));\n  }\n  return [].concat(value);\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/execution_context.js\nvar ExecutionContext = class {\n  constructor(weightMap = {}, tensorArrayMap = {}, tensorListMap = {}, functionMap = {}) {\n    this.weightMap = weightMap;\n    this.tensorArrayMap = tensorArrayMap;\n    this.tensorListMap = tensorListMap;\n    this.functionMap = functionMap;\n    this.rootContext = { id: 0, frameName: \"\", iterationId: 0 };\n    this.contexts = [this.rootContext];\n    this.lastId = 0;\n    this.generateCurrentContextIds();\n  }\n  newFrame(id, frameName) {\n    return { id, frameName, iterationId: 0 };\n  }\n  set currentContext(contexts2) {\n    if (this.contexts !== contexts2) {\n      this.contexts = contexts2;\n      this.generateCurrentContextIds();\n    }\n  }\n  get currentContext() {\n    return this.contexts;\n  }\n  get currentContextId() {\n    return this._currentContextIds[0];\n  }\n  get currentContextIds() {\n    return this._currentContextIds;\n  }\n  generateCurrentContextIds() {\n    const names = [];\n    for (let i = 0; i < this.contexts.length - 1; i++) {\n      const contexts2 = this.contexts.slice(0, this.contexts.length - i);\n      names.push(this.contextIdforContexts(contexts2));\n    }\n    names.push(\"\");\n    this._currentContextIds = names;\n  }\n  contextIdforContexts(contexts2) {\n    return contexts2 ? contexts2.map((context) => context.id === 0 && context.iterationId === 0 ? \"\" : `${context.frameName}-${context.iterationId}`).join(\"/\") : \"\";\n  }\n  enterFrame(frameId) {\n    if (this.contexts) {\n      this.lastId++;\n      this.contexts = this.contexts.slice();\n      this.contexts.push(this.newFrame(this.lastId, frameId));\n      this._currentContextIds.unshift(this.contextIdforContexts(this.contexts));\n    }\n  }\n  exitFrame() {\n    if (this.contexts && this.contexts.length > 1) {\n      this.contexts = this.contexts.slice();\n      this.contexts.splice(-1);\n      this.currentContextIds.shift();\n    } else {\n      throw new Error(\"Cannot exit frame, the context is empty\");\n    }\n  }\n  nextIteration() {\n    if (this.contexts && this.contexts.length > 0) {\n      this.contexts = this.contexts.slice();\n      this.lastId++;\n      const context = Object.assign({}, this.contexts[this.contexts.length - 1]);\n      context.iterationId += 1;\n      context.id = this.lastId;\n      this.contexts.splice(-1, 1, context);\n      this._currentContextIds.splice(0, 1, this.contextIdforContexts(this.contexts));\n    } else {\n      throw new Error(\"Cannot increase frame iteration, the context is empty\");\n    }\n  }\n  getWeight(name) {\n    return this.weightMap[name];\n  }\n  addTensorArray(tensorArray) {\n    this.tensorArrayMap[tensorArray.id] = tensorArray;\n  }\n  getTensorArray(id) {\n    return this.tensorArrayMap[id];\n  }\n  addTensorList(tensorList) {\n    this.tensorListMap[tensorList.id] = tensorList;\n  }\n  getTensorList(id) {\n    return this.tensorListMap[id];\n  }\n  dispose(keepIds) {\n    for (const key in this.tensorArrayMap) {\n      this.tensorArrayMap[key].clearAndClose(keepIds);\n    }\n    for (const key in this.tensorListMap) {\n      this.tensorListMap[key].clearAndClose(keepIds);\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/model_analysis.js\nfunction getExecutionSubgraph(inputs, outputs, weightMap, initNodes) {\n  const usedNodes = /* @__PURE__ */ new Set();\n  const missingInputs = [];\n  let dynamicNode = null;\n  let syncInputs = null;\n  const seen = /* @__PURE__ */ new Set();\n  const inputNodeNames = Object.keys(inputs).map((name) => parseNodeName(name)[0]);\n  let initNodeNames = [];\n  if (initNodes != null) {\n    initNodeNames = initNodes.map((node) => parseNodeName(node.name)[0]);\n  }\n  const frontier = [...outputs];\n  while (frontier.length > 0) {\n    const node = frontier.pop();\n    if (isControlFlow(node) || isDynamicShape(node) || isHashTable(node)) {\n      if (dynamicNode == null) {\n        dynamicNode = node;\n        syncInputs = dynamicNode.children.map((child) => child.name).filter((name) => usedNodes.has(name));\n      }\n    }\n    usedNodes.add(node.name);\n    if (weightMap[node.name] != null) {\n      continue;\n    }\n    if (inputNodeNames.indexOf(node.name) !== -1) {\n      continue;\n    }\n    if (initNodeNames.indexOf(node.name) !== -1) {\n      continue;\n    }\n    if (node.inputs.length === 0) {\n      missingInputs.push(node.name);\n      continue;\n    }\n    node.inputs.forEach((input2) => {\n      if (seen.has(input2.name)) {\n        return;\n      }\n      seen.add(input2.name);\n      frontier.push(input2);\n    });\n  }\n  return { inputs, outputs, usedNodes, missingInputs, dynamicNode, syncInputs };\n}\nfunction getNodesInTopologicalOrder(graph, weightMap, executionInfo) {\n  const { usedNodes, inputs } = executionInfo;\n  const frontier = [];\n  const inputNodes = Object.keys(inputs).map((name) => parseNodeName(name)[0]).map((name) => graph.nodes[name]);\n  const initNodes = graph.initNodes;\n  inputNodes.forEach((input2) => {\n    if (usedNodes.has(input2.name)) {\n      frontier.push(input2);\n    }\n  });\n  graph.weights.forEach((weight) => {\n    if (usedNodes.has(weight.name)) {\n      frontier.push(weight);\n    }\n  });\n  if (initNodes != null) {\n    initNodes.forEach((node) => {\n      if (usedNodes.has(node.name)) {\n        frontier.push(node);\n      }\n    });\n  }\n  const seen = /* @__PURE__ */ new Set();\n  const orderedNodes = [];\n  while (frontier.length > 0) {\n    const node = frontier.pop();\n    seen.add(node.name);\n    if (!weightMap[node.name]) {\n      orderedNodes.push(node);\n    }\n    node.children.forEach((child) => {\n      if (!seen.has(child.name) && usedNodes.has(child.name) && child.inputs.every((input2) => seen.has(input2.name))) {\n        frontier.push(child);\n      }\n    });\n  }\n  return orderedNodes;\n}\nvar CONTROL_FLOW_OPS = [\n  \"Switch\",\n  \"Merge\",\n  \"Enter\",\n  \"Exit\",\n  \"NextIteration\",\n  \"StatelessIf\",\n  \"StatelessWhile\",\n  \"if\",\n  \"While\"\n];\nvar DYNAMIC_SHAPE_OPS = [\n  \"NonMaxSuppressionV2\",\n  \"NonMaxSuppressionV3\",\n  \"NonMaxSuppressionV5\",\n  \"Where\"\n];\nvar HASH_TABLE_OPS = [\n  \"HashTable\",\n  \"HashTableV2\",\n  \"LookupTableImport\",\n  \"LookupTableImportV2\",\n  \"LookupTableFind\",\n  \"LookupTableFindV2\",\n  \"LookupTableSize\",\n  \"LookupTableSizeV2\"\n];\nfunction isControlFlow(node) {\n  return CONTROL_FLOW_OPS.indexOf(node.op) >= 0;\n}\nfunction isDynamicShape(node) {\n  return DYNAMIC_SHAPE_OPS.indexOf(node.op) >= 0;\n}\nfunction isHashTable(node) {\n  return HASH_TABLE_OPS.indexOf(node.op) >= 0;\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/graph_executor.js\nvar GraphExecutor = class {\n  constructor(graph, parent) {\n    this.graph = graph;\n    this.parent = parent;\n    this.compiledMap = /* @__PURE__ */ new Map();\n    this._weightMap = {};\n    this.SEPERATOR = \",\";\n    this._functions = {};\n    this._functionExecutorMap = {};\n    this.intermediateTensors = {};\n    this.keepTensorForDebug = false;\n    this._outputs = graph.outputs;\n    this._inputs = graph.inputs;\n    this._initNodes = graph.initNodes;\n    this._signature = graph.signature;\n    this._functions = graph.functions;\n    if (graph.functions != null) {\n      Object.keys(graph.functions).forEach((name) => {\n        this._functionExecutorMap[name] = new GraphExecutor(graph.functions[name], this);\n      });\n    }\n  }\n  get weightIds() {\n    return this.parent ? this.parent.weightIds : this._weightIds;\n  }\n  get functionExecutorMap() {\n    return this.parent ? this.parent.functionExecutorMap : this._functionExecutorMap;\n  }\n  get weightMap() {\n    return this.parent ? this.parent.weightMap : this._weightMap;\n  }\n  set weightMap(weightMap) {\n    const weightIds = Object.keys(weightMap).map((key) => weightMap[key].map((tensor2) => tensor2.id));\n    this._weightIds = [].concat(...weightIds);\n    this._weightMap = weightMap;\n  }\n  set resourceManager(resourceManager) {\n    this._resourceManager = resourceManager;\n  }\n  get inputs() {\n    return this._inputs.map((node) => {\n      return {\n        name: node.name,\n        shape: node.attrParams[\"shape\"] ? node.attrParams[\"shape\"].value : void 0,\n        dtype: node.attrParams[\"dtype\"] ? node.attrParams[\"dtype\"].value : void 0\n      };\n    });\n  }\n  get outputs() {\n    return this._outputs.map((node) => {\n      return {\n        name: node.name,\n        shape: node.attrParams[\"shape\"] ? node.attrParams[\"shape\"].value : void 0,\n        dtype: node.attrParams[\"dtype\"] ? node.attrParams[\"dtype\"].value : void 0\n      };\n    });\n  }\n  get inputNodes() {\n    return this._inputs.map((node) => node.signatureKey || node.name);\n  }\n  get outputNodes() {\n    return this._outputs.map((node) => {\n      const name = node.signatureKey || node.name;\n      return node.defaultOutput ? `${name}:${node.defaultOutput}` : name;\n    });\n  }\n  get functions() {\n    return Object.keys(this._functions).reduce((map, key) => {\n      map[key] = this._functions[key].signature;\n      return map;\n    }, {});\n  }\n  getCompilationKey(inputs, outputs) {\n    const sortedInputs = inputs.map((node) => node.name).sort();\n    const sortedOutputs = outputs.map((node) => node.name).sort();\n    return sortedInputs.join(this.SEPERATOR) + \"--\" + sortedOutputs.join(this.SEPERATOR);\n  }\n  compile(inputs, outputs) {\n    const executionInfo = getExecutionSubgraph(inputs, outputs, this.weightMap, this._initNodes);\n    const { missingInputs, dynamicNode, syncInputs } = executionInfo;\n    if (dynamicNode != null) {\n      throw new Error(`This execution contains the node '${dynamicNode.name}', which has the dynamic op '${dynamicNode.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${syncInputs}]`);\n    }\n    if (missingInputs.length > 0) {\n      const outNames = outputs.map((n) => n.name);\n      const inNames = Object.keys(inputs);\n      throw new Error(`Cannot compute the outputs [${outNames}] from the provided inputs [${inNames}]. Missing the following inputs: [${missingInputs}]`);\n    }\n    return getNodesInTopologicalOrder(this.graph, this.weightMap, executionInfo);\n  }\n  execute(inputs, outputs) {\n    inputs = this.mapInputs(inputs);\n    const names = Object.keys(inputs).sort();\n    this.checkInputs(inputs);\n    this.checkInputShapeAndType(inputs);\n    outputs = this.mapOutputs(outputs);\n    this.checkOutputs(outputs);\n    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);\n    const outputNodeNames = outputs.map((name) => parseNodeName(name)[0]);\n    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);\n    this.resetIntermediateTensors();\n    if (outputNodes.length === 0) {\n      outputNodes = this._outputs;\n    }\n    const compilationKey = this.getCompilationKey(inputNodes, outputNodes);\n    let orderedNodes = this.compiledMap.get(compilationKey);\n    if (orderedNodes == null) {\n      orderedNodes = this.compile(inputs, outputNodes);\n      this.compiledMap.set(compilationKey, orderedNodes);\n    }\n    const tensorArrayMap = {};\n    const tensorListMap = {};\n    return tidy(() => {\n      const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);\n      const tensorsMap = Object.assign({}, this.weightMap);\n      Object.keys(inputs).forEach((name) => {\n        const [nodeName, index] = parseNodeName(name);\n        const tensors = [];\n        tensors[index] = inputs[name];\n        tensorsMap[nodeName] = tensors;\n      });\n      const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);\n      const intermediateTensorConsumerCount = {};\n      for (let i = 0; i < orderedNodes.length; i++) {\n        const node = orderedNodes[i];\n        if (!tensorsMap[node.name]) {\n          const tensors = executeOp20(node, tensorsMap, context, this._resourceManager);\n          if (util_exports.isPromise(tensors)) {\n            throw new Error(`The execution of the op '${node.op}' returned a promise. Please use model.executeAsync() instead.`);\n          }\n          tensorsMap[node.name] = tensors;\n          this.checkTensorForDisposal(node.name, node, tensorsMap, context, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount);\n        }\n      }\n      if (this.parent == null) {\n        context.dispose(tensorsToKeep);\n      }\n      return outputs.map((name) => getTensor(name, tensorsMap, context));\n    });\n  }\n  getFrozenTensorIds(tensorMap) {\n    const ids = [].concat.apply([], Object.keys(tensorMap).map((key) => tensorMap[key]).map((tensors) => tensors.map((tensor2) => tensor2.id)));\n    return new Set(ids);\n  }\n  checkTensorForDisposal(nodeName, node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount) {\n    if (node.category === \"control\" || outputNames.indexOf(nodeName) !== -1) {\n      return;\n    }\n    tensorMap[nodeName].forEach((tensor2) => {\n      if (tensor2 != null) {\n        intermediateTensorConsumerCount[tensor2.id] = (intermediateTensorConsumerCount[tensor2.id] || 0) + node.children.length;\n      }\n    });\n    node.inputs.forEach((input2) => {\n      if (input2.category !== \"control\") {\n        const tensors = getTensorsForCurrentContenxt(input2.name, tensorMap, context);\n        if (tensors != null) {\n          tensors.forEach((tensor2) => {\n            if (tensor2 && !tensor2.kept && !tensorsToKeep.has(tensor2.id)) {\n              const count2 = intermediateTensorConsumerCount[tensor2.id];\n              if (count2 === 1) {\n                if (!this.keepTensorForDebug) {\n                  tensor2.dispose();\n                } else {\n                  const [nodeName2, index] = getNodeNameAndIndex(node.name, context);\n                  if (this.intermediateTensors[nodeName2]) {\n                    this.intermediateTensors[nodeName2][index] = tensor2;\n                  } else {\n                    this.intermediateTensors[nodeName2] = [];\n                    this.intermediateTensors[nodeName2][index] = tensor2;\n                  }\n                }\n                delete intermediateTensorConsumerCount[tensor2.id];\n              } else if (count2 != null) {\n                intermediateTensorConsumerCount[tensor2.id]--;\n              }\n            }\n          });\n        }\n      }\n    });\n  }\n  async executeAsync(inputs, outputs) {\n    return this._executeAsync(inputs, outputs);\n  }\n  disposeIntermediateTensors() {\n    if (!this.intermediateTensors) {\n      return;\n    }\n    Object.keys(this.intermediateTensors).forEach((key) => this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose()));\n    this.disposeTensorsMap();\n  }\n  disposeTensorsMap() {\n    if (!this.tensorsMap) {\n      return;\n    }\n    Object.keys(this.tensorsMap).forEach((key) => {\n      const tensorArray = this.tensorsMap[key];\n      tensorArray.forEach((tensor2) => {\n        if (tensor2 && !tensor2.kept && !tensor2.isDisposed && !this.keepIds.has(tensor2.id)) {\n          tensor2.dispose();\n        }\n      });\n    });\n  }\n  getIntermediateTensors() {\n    return this.tensorsMap;\n  }\n  resetIntermediateTensors() {\n    for (const key in this.intermediateTensors) {\n      this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose());\n      delete this.intermediateTensors[key];\n    }\n  }\n  async _executeAsync(inputs, outputs, isFunctionExecution = false, tensorArrayMap = {}, tensorListMap = {}) {\n    if (!isFunctionExecution) {\n      inputs = this.mapInputs(inputs);\n      this.checkInputs(inputs);\n      this.checkInputShapeAndType(inputs);\n      outputs = this.mapOutputs(outputs);\n      this.checkOutputs(outputs);\n    }\n    try {\n      this.keepTensorForDebug = env().getBool(\"KEEP_INTERMEDIATE_TENSORS\");\n    } catch (e) {\n      console.warn(e.message);\n    }\n    this.resetIntermediateTensors();\n    const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);\n    this.tensorsMap = await this.executeWithControlFlow(inputs, context, outputs, isFunctionExecution);\n    const results = outputs.map((name) => getTensor(name, this.tensorsMap, context));\n    const outputIds = results.map((t) => t.id);\n    const inputIds = Object.keys(inputs).map((name) => inputs[name].id);\n    this.keepIds = /* @__PURE__ */ new Set([...outputIds, ...inputIds, ...this.weightIds]);\n    if (!this.keepTensorForDebug) {\n      this.disposeTensorsMap();\n    }\n    if (this.parent == null) {\n      context.dispose(this.keepIds);\n    }\n    return results;\n  }\n  async executeFunctionAsync(inputs, tensorArrayMap, tensorListMap) {\n    const mappedInputs = inputs.reduce((map, tensor2, index) => {\n      map[this.inputs[index].name] = tensor2;\n      return map;\n    }, {});\n    return this._executeAsync(mappedInputs, this.outputNodes, true, tensorArrayMap, tensorListMap);\n  }\n  async executeWithControlFlow(inputs, context, outputNames, isFunctionExecution) {\n    const names = Object.keys(inputs);\n    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);\n    const outputNodeNames = outputNames.map((name) => parseNodeName(name)[0]);\n    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);\n    if (outputNodes.length === 0) {\n      outputNodes = this._outputs;\n    }\n    const { usedNodes, missingInputs, dynamicNode, syncInputs } = getExecutionSubgraph(inputs, outputNodes, this.weightMap, this._initNodes);\n    const stack2 = [\n      ...inputNodes,\n      ...this.graph.weights,\n      ...this._initNodes || []\n    ].map((node) => {\n      return { node, contexts: context.currentContext };\n    });\n    const tensorsMap = Object.assign({}, this.weightMap);\n    Object.keys(inputs).forEach((name) => {\n      const [nodeName, index] = parseNodeName(name);\n      const tensors = [];\n      tensors[index] = inputs[name];\n      tensorsMap[nodeName] = tensors;\n    });\n    const intermediateTensorConsumerCount = {};\n    const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);\n    const added = {};\n    while (stack2.length > 0) {\n      const promises = this.processStack(inputNodes, stack2, context, tensorsMap, added, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount, usedNodes);\n      await Promise.all(promises);\n    }\n    if (dynamicNode == null && !isFunctionExecution) {\n      console.warn(`This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.`);\n    }\n    const missingOutputs = outputNodes.filter((node) => !isControlFlow(node) && !getTensor(node.name, tensorsMap, context)).map((node) => node.name);\n    if (missingOutputs.length > 0) {\n      let alternativeMsg = \"\";\n      if (dynamicNode != null) {\n        alternativeMsg = `Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${syncInputs}]`;\n      }\n      throw new Error(`Cannot compute the outputs [${missingOutputs}] from the provided inputs [${names}]. Consider providing the following inputs: [${missingInputs}]. ${alternativeMsg}`);\n    }\n    return tensorsMap;\n  }\n  processStack(inputNodes, stack2, context, tensorMap, added, tensorsToKeep, outputNames, intermediateTensorConsumerCount, usedNodes) {\n    const promises = [];\n    while (stack2.length > 0) {\n      const item = stack2.pop();\n      context.currentContext = item.contexts;\n      let nodeName = \"\";\n      if (item.node.op === \"Enter\" && getParamValue(\"isConstant\", item.node, tensorMap, context)) {\n        [nodeName] = getNodeNameAndIndex(item.node.name, context);\n      }\n      if (tensorMap[item.node.name] == null) {\n        const tensors = executeOp20(item.node, tensorMap, context, this._resourceManager);\n        if (!nodeName) {\n          [nodeName] = getNodeNameAndIndex(item.node.name, context);\n        }\n        const currentContext = context.currentContext;\n        if (util_exports.isPromise(tensors)) {\n          promises.push(tensors.then((t) => {\n            tensorMap[nodeName] = t;\n            context.currentContext = currentContext;\n            this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);\n            this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);\n            return t;\n          }));\n        } else {\n          tensorMap[nodeName] = tensors;\n          this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);\n          this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);\n        }\n      } else {\n        this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);\n      }\n    }\n    return promises;\n  }\n  processChildNodes(node, stack2, context, tensorMap, added, usedNodes) {\n    node.children.forEach((childNode) => {\n      const [nodeName] = getNodeNameAndIndex(childNode.name, context);\n      if (added[nodeName] || !usedNodes.has(childNode.name)) {\n        return;\n      }\n      if (childNode.op === \"Merge\") {\n        if (childNode.inputNames.some((name) => {\n          return !!getTensor(name, tensorMap, context);\n        })) {\n          added[nodeName] = true;\n          stack2.push({ contexts: context.currentContext, node: childNode });\n        }\n      } else if (childNode.inputNames.every((name) => {\n        return !!getTensor(name, tensorMap, context);\n      })) {\n        added[nodeName] = true;\n        stack2.push({ contexts: context.currentContext, node: childNode });\n      }\n    });\n  }\n  dispose() {\n    Object.keys(this.weightMap).forEach((key) => this.weightMap[key].forEach((tensor2) => tensor2.dispose()));\n  }\n  checkInputShapeAndType(inputs) {\n    Object.keys(inputs).forEach((name) => {\n      const input2 = inputs[name];\n      const [nodeName] = parseNodeName(name);\n      const node = this.graph.nodes[nodeName];\n      if (node.attrParams[\"shape\"] && node.attrParams[\"shape\"].value) {\n        const shape = node.attrParams[\"shape\"].value;\n        const match = shape.length === input2.shape.length && input2.shape.every((dim, index) => shape[index] === -1 || shape[index] === dim);\n        util_exports.assert(match, () => `The shape of dict['${node.name}'] provided in model.execute(dict) must be [${shape}], but was [${input2.shape}]`);\n      }\n      if (node.attrParams[\"dtype\"] && node.attrParams[\"dtype\"].value) {\n        util_exports.assert(input2.dtype === node.attrParams[\"dtype\"].value, () => `The dtype of dict['${node.name}'] provided in model.execute(dict) must be ${node.attrParams[\"dtype\"].value}, but was ${input2.dtype}`);\n      }\n    });\n  }\n  mapInputs(inputs) {\n    const result = {};\n    for (const inputName in inputs) {\n      if (this._signature != null && this._signature.inputs != null && this._signature.inputs[inputName] != null) {\n        const tensor2 = this._signature.inputs[inputName];\n        result[tensor2.name] = inputs[inputName];\n      } else {\n        result[inputName] = inputs[inputName];\n      }\n    }\n    return result;\n  }\n  checkInputs(inputs) {\n    const notInGraph = Object.keys(inputs).filter((name) => {\n      const [nodeName] = parseNodeName(name);\n      return this.graph.nodes[nodeName] == null;\n    });\n    if (notInGraph.length > 0) {\n      throw new Error(`The dict provided in model.execute(dict) has keys: [${notInGraph}] that are not part of graph`);\n    }\n  }\n  mapOutputs(outputs) {\n    return outputs.map((name) => {\n      if (this._signature != null && this._signature.outputs != null && this._signature.outputs[name] != null) {\n        const tensor2 = this._signature.outputs[name];\n        return tensor2.name;\n      }\n      return name;\n    }, {});\n  }\n  checkOutputs(outputs) {\n    outputs.forEach((name) => {\n      const [normalizedName] = parseNodeName(name);\n      if (!this.graph.nodes[normalizedName]) {\n        throw new Error(`The output '${name}' is not found in the graph`);\n      }\n    });\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/resource_manager.js\nvar ResourceManager = class {\n  constructor(hashTableNameToHandle = {}, hashTableMap = {}) {\n    this.hashTableNameToHandle = hashTableNameToHandle;\n    this.hashTableMap = hashTableMap;\n  }\n  addHashTable(name, hashTable) {\n    this.hashTableNameToHandle[name] = hashTable.handle;\n    this.hashTableMap[hashTable.id] = hashTable;\n  }\n  getHashTableHandleByName(name) {\n    return this.hashTableNameToHandle[name];\n  }\n  getHashTableById(id) {\n    return this.hashTableMap[id];\n  }\n  dispose() {\n    for (const key in this.hashTableMap) {\n      this.hashTableMap[key].clearAndClose();\n      delete this.hashTableMap[key];\n    }\n    for (const name in this.hashTableNameToHandle) {\n      this.hashTableNameToHandle[name].dispose();\n      delete this.hashTableNameToHandle[name];\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-converter/dist/executor/graph_model.js\nvar TFHUB_SEARCH_PARAM = \"?tfjs-format=file\";\nvar DEFAULT_MODEL_NAME = \"model.json\";\nvar GraphModel = class {\n  constructor(modelUrl, loadOptions = {}, tfio = io_exports) {\n    this.modelUrl = modelUrl;\n    this.loadOptions = loadOptions;\n    this.version = \"n/a\";\n    this.io = tfio;\n    if (loadOptions == null) {\n      this.loadOptions = {};\n    }\n    this.resourceManager = new ResourceManager();\n  }\n  get modelVersion() {\n    return this.version;\n  }\n  get inputNodes() {\n    return this.executor.inputNodes;\n  }\n  get outputNodes() {\n    return this.executor.outputNodes;\n  }\n  get inputs() {\n    return this.executor.inputs;\n  }\n  get outputs() {\n    return this.executor.outputs;\n  }\n  get weights() {\n    return this.executor.weightMap;\n  }\n  get metadata() {\n    return this.artifacts.userDefinedMetadata;\n  }\n  get modelSignature() {\n    return this.signature;\n  }\n  get modelStructuredOutputKeys() {\n    return this.structuredOutputKeys;\n  }\n  findIOHandler() {\n    const path = this.modelUrl;\n    if (path.load != null) {\n      this.handler = path;\n    } else if (this.loadOptions.requestInit != null) {\n      this.handler = this.io.browserHTTPRequest(path, this.loadOptions);\n    } else {\n      const handlers = this.io.getLoadHandlers(path, this.loadOptions);\n      if (handlers.length === 0) {\n        handlers.push(this.io.browserHTTPRequest(path, this.loadOptions));\n      } else if (handlers.length > 1) {\n        throw new Error(`Found more than one (${handlers.length}) load handlers for URL '${[path]}'`);\n      }\n      this.handler = handlers[0];\n    }\n  }\n  load() {\n    this.findIOHandler();\n    if (this.handler.load == null) {\n      throw new Error(\"Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.\");\n    }\n    const loadResult = this.handler.load();\n    if (util_exports.isPromise(loadResult)) {\n      return loadResult.then((artifacts) => this.loadSync(artifacts));\n    }\n    return this.loadSync(loadResult);\n  }\n  loadSync(artifacts) {\n    this.artifacts = artifacts;\n    const graph = this.artifacts.modelTopology;\n    let signature = this.artifacts.signature;\n    if (this.artifacts.userDefinedMetadata != null) {\n      const metadata = this.artifacts.userDefinedMetadata;\n      if (metadata.signature != null) {\n        signature = metadata.signature;\n      }\n      if (metadata.structuredOutputKeys != null) {\n        this.structuredOutputKeys = metadata.structuredOutputKeys;\n      }\n    }\n    this.signature = signature;\n    this.version = `${graph.versions.producer}.${graph.versions.minConsumer}`;\n    const weightMap = this.io.decodeWeights(this.artifacts.weightData, this.artifacts.weightSpecs);\n    this.executor = new GraphExecutor(OperationMapper.Instance.transformGraph(graph, this.signature));\n    this.executor.weightMap = this.convertTensorMapToTensorsMap(weightMap);\n    this.executor.resourceManager = this.resourceManager;\n    if (artifacts.modelInitializer != null && artifacts.modelInitializer.node != null) {\n      const initializer = OperationMapper.Instance.transformGraph(artifacts.modelInitializer);\n      this.initializer = new GraphExecutor(initializer);\n      this.initializer.weightMap = this.executor.weightMap;\n      this.initializer.resourceManager = this.resourceManager;\n      this.initializerSignature = artifacts.initializerSignature;\n    }\n    return true;\n  }\n  async save(handlerOrURL, config) {\n    if (typeof handlerOrURL === \"string\") {\n      const handlers = this.io.getSaveHandlers(handlerOrURL);\n      if (handlers.length === 0) {\n        throw new Error(`Cannot find any save handlers for URL '${handlerOrURL}'`);\n      } else if (handlers.length > 1) {\n        throw new Error(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);\n      }\n      handlerOrURL = handlers[0];\n    }\n    if (handlerOrURL.save == null) {\n      throw new Error(\"GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.\");\n    }\n    return handlerOrURL.save(this.artifacts);\n  }\n  predict(inputs, config) {\n    const outputTensors = this.execute(inputs, this.outputNodes);\n    if (this.structuredOutputKeys) {\n      const outputTensorsArray = outputTensors instanceof Tensor ? [outputTensors] : outputTensors;\n      const outputTensorMap = {};\n      outputTensorsArray.forEach((outputTensor, i) => outputTensorMap[this.structuredOutputKeys[i]] = outputTensor);\n      return outputTensorMap;\n    }\n    return outputTensors;\n  }\n  normalizeInputs(inputs) {\n    if (!(inputs instanceof Tensor) && !Array.isArray(inputs)) {\n      if (this.signature != null && this.signature.inputs != null) {\n        for (const input2 in this.signature.inputs) {\n          const tensor2 = this.signature.inputs[input2];\n          if (tensor2.resourceId != null) {\n            inputs[input2] = this.resourceIdToCapturedInput[tensor2.resourceId];\n          }\n        }\n      }\n      return inputs;\n    }\n    inputs = Array.isArray(inputs) ? inputs : [inputs];\n    const numCapturedInputs = Object.keys(this.resourceIdToCapturedInput).length;\n    if (inputs.length + numCapturedInputs !== this.inputNodes.length) {\n      throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length - numCapturedInputs} non-resource placeholders, while there are ${inputs.length} input tensors provided.`);\n    }\n    let inputIndex = 0;\n    return this.inputNodes.reduce((map, inputName) => {\n      const signature = this.signature ? this.signature.inputs[inputName] : null;\n      if (signature != null && signature.resourceId != null) {\n        map[inputName] = this.resourceIdToCapturedInput[signature.resourceId];\n      } else {\n        map[inputName] = inputs[inputIndex++];\n      }\n      return map;\n    }, {});\n  }\n  normalizeOutputs(outputs) {\n    outputs = outputs || this.outputNodes;\n    return !Array.isArray(outputs) ? [outputs] : outputs;\n  }\n  executeInitializerGraph() {\n    if (this.initializer == null) {\n      return [];\n    }\n    if (this.initializerSignature == null) {\n      return this.initializer.execute({}, []);\n    } else {\n      return this.initializer.execute({}, Object.keys(this.initializerSignature.outputs));\n    }\n  }\n  async executeInitializerGraphAsync() {\n    if (this.initializer == null) {\n      return [];\n    }\n    if (this.initializerSignature == null) {\n      return this.initializer.executeAsync({}, []);\n    } else {\n      return this.initializer.executeAsync({}, Object.keys(this.initializerSignature.outputs));\n    }\n  }\n  setResourceIdToCapturedInput(outputs) {\n    this.resourceIdToCapturedInput = {};\n    if (this.initializerSignature) {\n      const outputNames = Object.keys(this.initializerSignature.outputs);\n      for (let i = 0; i < outputNames.length; i++) {\n        const outputName = outputNames[i];\n        const tensorInfo = this.initializerSignature.outputs[outputName];\n        this.resourceIdToCapturedInput[tensorInfo.resourceId] = outputs[i];\n      }\n    }\n  }\n  execute(inputs, outputs) {\n    if (this.resourceIdToCapturedInput == null) {\n      this.setResourceIdToCapturedInput(this.executeInitializerGraph());\n    }\n    inputs = this.normalizeInputs(inputs);\n    outputs = this.normalizeOutputs(outputs);\n    const result = this.executor.execute(inputs, outputs);\n    return result.length > 1 ? result : result[0];\n  }\n  async executeAsync(inputs, outputs) {\n    if (this.resourceIdToCapturedInput == null) {\n      this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync());\n    }\n    inputs = this.normalizeInputs(inputs);\n    outputs = this.normalizeOutputs(outputs);\n    const result = await this.executor.executeAsync(inputs, outputs);\n    return result.length > 1 ? result : result[0];\n  }\n  getIntermediateTensors() {\n    return this.executor.getIntermediateTensors();\n  }\n  disposeIntermediateTensors() {\n    this.executor.disposeIntermediateTensors();\n  }\n  convertTensorMapToTensorsMap(map) {\n    return Object.keys(map).reduce((newMap, key) => {\n      newMap[key] = [map[key]];\n      return newMap;\n    }, {});\n  }\n  dispose() {\n    this.executor.dispose();\n    if (this.initializer) {\n      this.initializer.dispose();\n      if (this.resourceIdToCapturedInput) {\n        dispose(this.resourceIdToCapturedInput);\n      }\n    }\n    this.resourceManager.dispose();\n  }\n};\nasync function loadGraphModel(modelUrl, options = {}, tfio = io_exports) {\n  if (modelUrl == null) {\n    throw new Error(\"modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model\");\n  }\n  if (options == null) {\n    options = {};\n  }\n  if (options.fromTFHub && typeof modelUrl === \"string\") {\n    modelUrl = getTFHubUrl(modelUrl);\n  }\n  const model2 = new GraphModel(modelUrl, options, tfio);\n  await model2.load();\n  return model2;\n}\nfunction loadGraphModelSync(modelSource) {\n  if (modelSource == null) {\n    throw new Error(\"modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model\");\n  }\n  let ioHandler;\n  if (modelSource instanceof Array) {\n    const [modelJSON, weights] = modelSource;\n    if (!modelJSON) {\n      throw new Error(\"modelJSON must be the first element of the array\");\n    }\n    if (!weights || !(weights instanceof ArrayBuffer)) {\n      throw new Error(\"An ArrayBuffer of weights must be the second element of the array\");\n    }\n    if (!(\"modelTopology\" in modelJSON)) {\n      throw new Error(\"Model JSON is missing 'modelTopology'\");\n    }\n    if (!(\"weightsManifest\" in modelJSON)) {\n      throw new Error(\"Model JSON is missing 'weightsManifest'\");\n    }\n    const weightSpecs = io_exports.getWeightSpecs(modelJSON.weightsManifest);\n    const modelArtifacts = io_exports.getModelArtifactsForJSONSync(modelJSON, weightSpecs, weights);\n    ioHandler = io_exports.fromMemorySync(modelArtifacts);\n  } else if (\"load\" in modelSource) {\n    ioHandler = modelSource;\n  } else if (\"modelTopology\" in modelSource && \"weightSpecs\" in modelSource && \"weightData\" in modelSource) {\n    ioHandler = io_exports.fromMemorySync(modelSource);\n  } else {\n    throw new Error(\"Unknown model format\");\n  }\n  const model2 = new GraphModel(ioHandler);\n  model2.load();\n  return model2;\n}\nfunction getTFHubUrl(modelUrl) {\n  if (!modelUrl.endsWith(\"/\")) {\n    modelUrl = modelUrl + \"/\";\n  }\n  return `${modelUrl}${DEFAULT_MODEL_NAME}${TFHUB_SEARCH_PARAM}`;\n}\n\n// node_modules/@tensorflow/tfjs-converter/dist/version.js\nvar version3 = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-data/dist/index.js\nvar dist_exports2 = {};\n__export(dist_exports2, {\n  CSVDataset: () => CSVDataset,\n  Dataset: () => Dataset,\n  FileDataSource: () => FileDataSource,\n  TextLineDataset: () => TextLineDataset,\n  URLDataSource: () => URLDataSource,\n  array: () => array,\n  csv: () => csv,\n  func: () => func,\n  generator: () => generator,\n  microphone: () => microphone,\n  version_data: () => version4,\n  webcam: () => webcam,\n  zip: () => zip\n});\n\n// node_modules/@tensorflow/tfjs-data/dist/dataset.js\nvar seedrandom3 = __toESM(require_seedrandom2());\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/lazy_iterator.js\nvar seedrandom2 = __toESM(require_seedrandom2());\n\n// node_modules/@tensorflow/tfjs-data/dist/util/deep_map.js\nfunction deepMap(input2, mapFn) {\n  return deepMapInternal(input2, mapFn);\n}\nfunction deepMapInternal(input2, mapFn, seen = /* @__PURE__ */ new Map(), containedIn = /* @__PURE__ */ new Set()) {\n  if (input2 == null) {\n    return null;\n  }\n  if (typeof Blob === \"function\" && input2 instanceof Blob) {\n    return input2.slice();\n  }\n  if (containedIn.has(input2)) {\n    throw new Error(\"Circular references are not supported.\");\n  }\n  if (seen.has(input2)) {\n    return seen.get(input2);\n  }\n  const result = mapFn(input2);\n  if (result.recurse && result.value !== null) {\n    throw new Error(\"A deep map function may not return both a value and recurse=true.\");\n  }\n  if (!result.recurse) {\n    seen.set(input2, result.value);\n    return result.value;\n  } else if (isIterable2(input2)) {\n    const mappedIterable = Array.isArray(input2) ? [] : {};\n    containedIn.add(input2);\n    for (const k in input2) {\n      const child = input2[k];\n      const childResult = deepMapInternal(child, mapFn, seen, containedIn);\n      mappedIterable[k] = childResult;\n    }\n    containedIn.delete(input2);\n    if (input2.__proto__) {\n      mappedIterable.__proto__ = input2.__proto__;\n    }\n    return mappedIterable;\n  } else {\n    throw new Error(`Can't recurse into non-iterable type: ${input2}`);\n  }\n}\nfunction deepZip(inputs, zipFn = zipToList) {\n  return deepZipInternal(inputs, zipFn);\n}\nfunction deepZipInternal(inputs, zipFn, containedIn = /* @__PURE__ */ new Set()) {\n  const input2 = inputs[0];\n  if (containedIn.has(input2)) {\n    throw new Error(\"Circular references are not supported.\");\n  }\n  const result = zipFn(inputs);\n  if (result.recurse && result.value !== null) {\n    throw new Error(\"A deep zip function may not return both a value and recurse=true.\");\n  }\n  if (!result.recurse) {\n    return result.value;\n  } else if (isIterable2(input2)) {\n    const mappedIterable = Array.isArray(input2) ? [] : {};\n    containedIn.add(input2);\n    for (const k in input2) {\n      const children = inputs.map((x) => x[k]);\n      const childResult = deepZipInternal(children, zipFn, containedIn);\n      mappedIterable[k] = childResult;\n    }\n    containedIn.delete(input2);\n    return mappedIterable;\n  } else {\n    throw new Error(`Can't recurse into non-iterable type: ${input2}`);\n  }\n}\nfunction zipToList(x) {\n  if (x === null) {\n    return null;\n  }\n  if (isIterable2(x[0])) {\n    return { value: null, recurse: true };\n  } else {\n    return { value: x, recurse: false };\n  }\n}\nasync function deepMapAndAwaitAll(input2, mapFn) {\n  const seen = /* @__PURE__ */ new Map();\n  deepMapInternal(input2, mapFn, seen);\n  for (const key of Array.from(seen.keys())) {\n    const value = seen.get(key);\n    if (util_exports.isPromise(value)) {\n      const mappedValue = await value;\n      seen.set(key, mappedValue);\n    }\n  }\n  const result = deepMapInternal(input2, mapFn, seen);\n  return result;\n}\nfunction isIterable2(obj) {\n  let isTextDecoder = false;\n  if (env().get(\"IS_BROWSER\")) {\n    isTextDecoder = obj instanceof TextDecoder;\n  } else {\n    const { StringDecoder } = require_string_decoder();\n    isTextDecoder = obj instanceof StringDecoder;\n  }\n  return obj != null && !ArrayBuffer.isView(obj) && (Array.isArray(obj) || typeof obj === \"object\" && !(obj instanceof Tensor) && !(obj instanceof Promise) && !isTextDecoder);\n}\nfunction canTensorify(obj) {\n  return obj == null || isPrimitive(obj) || Array.isArray(obj) || typeof obj === \"object\" && obj instanceof Tensor || util_exports.isTypedArray(obj);\n}\nfunction isPrimitive(value) {\n  return value === null || typeof value !== \"object\" && typeof value !== \"function\";\n}\n\n// node_modules/@tensorflow/tfjs-data/dist/util/deep_clone.js\nfunction deepClone(container) {\n  return deepMap(container, cloneIfTensor);\n}\nfunction cloneIfTensor(item) {\n  if (item instanceof Tensor) {\n    return { value: item.clone(), recurse: false };\n  } else if (isIterable2(item)) {\n    return { value: null, recurse: true };\n  } else {\n    return { value: item, recurse: false };\n  }\n}\n\n// node_modules/@tensorflow/tfjs-data/dist/util/ring_buffer.js\nvar RingBuffer = class {\n  constructor(capacity) {\n    this.capacity = capacity;\n    this.begin = 0;\n    this.end = 0;\n    if (capacity == null) {\n      throw new RangeError(\"Can't create a ring buffer of unknown capacity.\");\n    }\n    if (capacity < 1) {\n      throw new RangeError(\"Can't create ring buffer of capacity < 1.\");\n    }\n    this.data = new Array(capacity);\n    this.doubledCapacity = 2 * capacity;\n  }\n  wrap(index) {\n    while (index < 0) {\n      index += this.doubledCapacity;\n    }\n    return index % this.doubledCapacity;\n  }\n  get(index) {\n    if (index < 0) {\n      throw new RangeError(\"Can't get item at a negative index.\");\n    }\n    return this.data[index % this.capacity];\n  }\n  set(index, value) {\n    if (index < 0) {\n      throw new RangeError(\"Can't set item at a negative index.\");\n    }\n    this.data[index % this.capacity] = value;\n  }\n  length() {\n    let length = this.end - this.begin;\n    if (length < 0) {\n      length = this.doubledCapacity + length;\n    }\n    return length;\n  }\n  isFull() {\n    return this.length() === this.capacity;\n  }\n  isEmpty() {\n    return this.length() === 0;\n  }\n  push(value) {\n    if (this.isFull()) {\n      throw new RangeError(\"Ring buffer is full.\");\n    }\n    this.set(this.end, value);\n    this.end = this.wrap(this.end + 1);\n  }\n  pushAll(values) {\n    for (const value of values) {\n      this.push(value);\n    }\n  }\n  pop() {\n    if (this.isEmpty()) {\n      throw new RangeError(\"Ring buffer is empty.\");\n    }\n    this.end = this.wrap(this.end - 1);\n    const result = this.get(this.end);\n    this.set(this.end, void 0);\n    return result;\n  }\n  unshift(value) {\n    if (this.isFull()) {\n      throw new RangeError(\"Ring buffer is full.\");\n    }\n    this.begin = this.wrap(this.begin - 1);\n    this.set(this.begin, value);\n  }\n  shift() {\n    if (this.isEmpty()) {\n      throw new RangeError(\"Ring buffer is empty.\");\n    }\n    const result = this.get(this.begin);\n    this.set(this.begin, void 0);\n    this.begin = this.wrap(this.begin + 1);\n    return result;\n  }\n  shuffleExcise(relativeIndex) {\n    if (this.isEmpty()) {\n      throw new RangeError(\"Ring buffer is empty.\");\n    }\n    const index = this.wrap(this.begin + relativeIndex);\n    const result = this.get(index);\n    this.set(index, this.pop());\n    return result;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/util/growing_ring_buffer.js\nvar GrowingRingBuffer = class extends RingBuffer {\n  constructor() {\n    super(GrowingRingBuffer.INITIAL_CAPACITY);\n  }\n  isFull() {\n    return false;\n  }\n  push(value) {\n    if (super.isFull()) {\n      this.expand();\n    }\n    super.push(value);\n  }\n  unshift(value) {\n    if (super.isFull()) {\n      this.expand();\n    }\n    super.unshift(value);\n  }\n  expand() {\n    const newCapacity = this.capacity * 2;\n    const newData = new Array(newCapacity);\n    const len = this.length();\n    for (let i = 0; i < len; i++) {\n      newData[i] = this.get(this.wrap(this.begin + i));\n    }\n    this.data = newData;\n    this.capacity = newCapacity;\n    this.doubledCapacity = 2 * this.capacity;\n    this.begin = 0;\n    this.end = len;\n  }\n};\nGrowingRingBuffer.INITIAL_CAPACITY = 32;\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/lazy_iterator.js\nfunction iteratorFromItems(items) {\n  return new ArrayIterator(items);\n}\nfunction iteratorFromFunction(func2) {\n  return new FunctionCallIterator(func2);\n}\nfunction iteratorFromConcatenated(baseIterators, baseErrorHandler) {\n  return new ChainedIterator(baseIterators, baseErrorHandler);\n}\nfunction iteratorFromZipped(iterators, mismatchMode = ZipMismatchMode.FAIL) {\n  return new ZipIterator(iterators, mismatchMode);\n}\nvar LazyIterator = class {\n  async toArray() {\n    const result = [];\n    let x = await this.next();\n    while (!x.done) {\n      result.push(x.value);\n      x = await this.next();\n    }\n    return result;\n  }\n  async toArrayForTest() {\n    const stream = this.prefetch(100);\n    const result = [];\n    let x = await stream.next();\n    while (!x.done) {\n      result.push(x.value);\n      x = await stream.next();\n    }\n    return result;\n  }\n  async resolveFully() {\n    let x = await this.next();\n    while (!x.done) {\n      x = await this.next();\n    }\n  }\n  async resolveWhile(predicate) {\n    let x = await this.next();\n    let shouldContinue = predicate(x.value);\n    while (!x.done && shouldContinue) {\n      x = await this.next();\n      shouldContinue = predicate(x.value);\n    }\n  }\n  handleErrors(handler) {\n    return new ErrorHandlingLazyIterator(this, handler);\n  }\n  filter(predicate) {\n    return new FilterIterator(this, predicate);\n  }\n  map(transform5) {\n    return new MapIterator(this, transform5);\n  }\n  mapAsync(transform5) {\n    return new AsyncMapIterator(this, transform5);\n  }\n  serialMapAsync(transform5) {\n    return new AsyncMapIterator(this, transform5).serial();\n  }\n  flatmap(transform5) {\n    return new FlatmapIterator(this, transform5);\n  }\n  async forEachAsync(f) {\n    return this.map(f).resolveFully();\n  }\n  async serialForEach(f) {\n    return this.serialMapAsync(f).resolveWhile((x) => x === true);\n  }\n  rowMajorBatch(batchSize, smallLastBatch = true) {\n    return new RowMajorBatchIterator(this, batchSize, smallLastBatch);\n  }\n  columnMajorBatch(batchSize, smallLastBatch = true, zipFn = zipToList) {\n    const rowBatches = this.rowMajorBatch(batchSize, smallLastBatch);\n    return rowBatches.map((x) => deepZip(x, zipFn));\n  }\n  concatenate(iterator, baseErrorHandler) {\n    return new ChainedIterator(iteratorFromItems([this, iterator]), baseErrorHandler);\n  }\n  take(count2) {\n    if (count2 < 0 || count2 == null) {\n      return this;\n    }\n    return new TakeIterator(this, count2);\n  }\n  skip(count2) {\n    if (count2 < 0 || count2 == null) {\n      return this;\n    }\n    return new SkipIterator(this, count2);\n  }\n  prefetch(bufferSize) {\n    return new PrefetchIterator(this, bufferSize);\n  }\n  shuffle(windowSize, seed) {\n    return new ShuffleIterator(this, windowSize, seed);\n  }\n  serial() {\n    return new SerialIterator(this);\n  }\n};\nvar ArrayIterator = class extends LazyIterator {\n  constructor(items) {\n    super();\n    this.items = items;\n    this.trav = 0;\n  }\n  summary() {\n    return `Array of ${this.items.length} items`;\n  }\n  async next() {\n    if (this.trav >= this.items.length) {\n      return { value: null, done: true };\n    }\n    const item = this.items[this.trav];\n    this.trav++;\n    return { value: deepClone(item), done: false };\n  }\n};\nvar FunctionCallIterator = class extends LazyIterator {\n  constructor(nextFn) {\n    super();\n    this.nextFn = nextFn;\n  }\n  summary() {\n    return `Function call`;\n  }\n  async next() {\n    try {\n      return this.nextFn();\n    } catch (e) {\n      e.message = `Error thrown while iterating through a dataset: ${e.message}`;\n      throw e;\n    }\n  }\n};\nvar SerialIterator = class extends LazyIterator {\n  constructor(upstream) {\n    super();\n    this.upstream = upstream;\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Serial`;\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    return this.upstream.next();\n  }\n};\nvar SkipIterator = class extends LazyIterator {\n  constructor(upstream, maxCount) {\n    super();\n    this.upstream = upstream;\n    this.maxCount = maxCount;\n    this.count = 0;\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Skip`;\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    while (this.count++ < this.maxCount) {\n      const skipped = await this.upstream.next();\n      if (skipped.done) {\n        return skipped;\n      }\n      dispose(skipped.value);\n    }\n    return this.upstream.next();\n  }\n};\nvar TakeIterator = class extends LazyIterator {\n  constructor(upstream, maxCount) {\n    super();\n    this.upstream = upstream;\n    this.maxCount = maxCount;\n    this.count = 0;\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Take`;\n  }\n  async next() {\n    if (this.count++ >= this.maxCount) {\n      return { value: null, done: true };\n    }\n    return this.upstream.next();\n  }\n};\nvar RowMajorBatchIterator = class extends LazyIterator {\n  constructor(upstream, batchSize, enableSmallLastBatch = true) {\n    super();\n    this.upstream = upstream;\n    this.batchSize = batchSize;\n    this.enableSmallLastBatch = enableSmallLastBatch;\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  summary() {\n    return `${this.upstream.summary()} -> RowMajorBatch`;\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    const batch = [];\n    while (batch.length < this.batchSize) {\n      const item = await this.upstream.next();\n      if (item.done) {\n        if (this.enableSmallLastBatch && batch.length > 0) {\n          return { value: batch, done: false };\n        }\n        return { value: null, done: true };\n      }\n      batch.push(item.value);\n    }\n    return { value: batch, done: false };\n  }\n};\nvar FilterIterator = class extends LazyIterator {\n  constructor(upstream, predicate) {\n    super();\n    this.upstream = upstream;\n    this.predicate = predicate;\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Filter`;\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    while (true) {\n      const item = await this.upstream.next();\n      if (item.done || this.predicate(item.value)) {\n        return item;\n      }\n      dispose(item.value);\n    }\n  }\n};\nvar MapIterator = class extends LazyIterator {\n  constructor(upstream, transform5) {\n    super();\n    this.upstream = upstream;\n    this.transform = transform5;\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Map`;\n  }\n  async next() {\n    const item = await this.upstream.next();\n    if (item.done) {\n      return { value: null, done: true };\n    }\n    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);\n    const mapped = this.transform(item.value);\n    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);\n    for (const t of inputTensors) {\n      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {\n        t.dispose();\n      }\n    }\n    return { value: mapped, done: false };\n  }\n};\nvar ErrorHandlingLazyIterator = class extends LazyIterator {\n  constructor(upstream, handler) {\n    super();\n    this.upstream = upstream;\n    this.handler = handler;\n    this.count = 0;\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  summary() {\n    return `${this.upstream.summary()} -> handleErrors`;\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    while (true) {\n      try {\n        return await this.upstream.next();\n      } catch (e) {\n        if (!this.handler(e)) {\n          return { value: null, done: true };\n        }\n      }\n    }\n  }\n};\nvar AsyncMapIterator = class extends LazyIterator {\n  constructor(upstream, transform5) {\n    super();\n    this.upstream = upstream;\n    this.transform = transform5;\n  }\n  summary() {\n    return `${this.upstream.summary()} -> AsyncMap`;\n  }\n  async next() {\n    const item = await this.upstream.next();\n    if (item.done) {\n      return { value: null, done: true };\n    }\n    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);\n    const mapped = await this.transform(item.value);\n    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);\n    for (const t of inputTensors) {\n      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {\n        t.dispose();\n      }\n    }\n    return { value: mapped, done: false };\n  }\n};\nvar OneToManyIterator = class extends LazyIterator {\n  constructor() {\n    super();\n    this.outputQueue = new GrowingRingBuffer();\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  async serialNext() {\n    while (this.outputQueue.length() === 0) {\n      if (!await this.pump()) {\n        return { value: null, done: true };\n      }\n    }\n    return { value: this.outputQueue.shift(), done: false };\n  }\n};\nvar FlatmapIterator = class extends OneToManyIterator {\n  constructor(upstream, transform5) {\n    super();\n    this.upstream = upstream;\n    this.transform = transform5;\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Flatmap`;\n  }\n  async pump() {\n    const item = await this.upstream.next();\n    if (item.done) {\n      return false;\n    }\n    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);\n    const mappedArray = this.transform(item.value);\n    const outputTensors = tensor_util_exports.getTensorsInContainer(mappedArray);\n    this.outputQueue.pushAll(mappedArray);\n    for (const t of inputTensors) {\n      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {\n        t.dispose();\n      }\n    }\n    return true;\n  }\n};\nvar ChainedIterator = class extends LazyIterator {\n  constructor(iterators, baseErrorHandler) {\n    super();\n    this.baseErrorHandler = baseErrorHandler;\n    this.lastRead = null;\n    this.iterator = null;\n    this.moreIterators = iterators;\n  }\n  summary() {\n    const upstreamSummaries = \"TODO: fill in upstream of chained summaries\";\n    return `${upstreamSummaries} -> Chained`;\n  }\n  async next() {\n    this.lastRead = this.readFromChain(this.lastRead);\n    return this.lastRead;\n  }\n  async readFromChain(lastRead) {\n    await lastRead;\n    if (this.iterator == null) {\n      const iteratorResult = await this.moreIterators.next();\n      if (iteratorResult.done) {\n        return { value: null, done: true };\n      }\n      this.iterator = iteratorResult.value;\n      if (this.baseErrorHandler != null) {\n        this.iterator = this.iterator.handleErrors(this.baseErrorHandler);\n      }\n    }\n    const itemResult = await this.iterator.next();\n    if (itemResult.done) {\n      this.iterator = null;\n      return this.readFromChain(lastRead);\n    }\n    return itemResult;\n  }\n};\nvar ZipMismatchMode;\n(function(ZipMismatchMode2) {\n  ZipMismatchMode2[ZipMismatchMode2[\"FAIL\"] = 0] = \"FAIL\";\n  ZipMismatchMode2[ZipMismatchMode2[\"SHORTEST\"] = 1] = \"SHORTEST\";\n  ZipMismatchMode2[ZipMismatchMode2[\"LONGEST\"] = 2] = \"LONGEST\";\n})(ZipMismatchMode || (ZipMismatchMode = {}));\nvar ZipIterator = class extends LazyIterator {\n  constructor(iterators, mismatchMode = ZipMismatchMode.FAIL) {\n    super();\n    this.iterators = iterators;\n    this.mismatchMode = mismatchMode;\n    this.count = 0;\n    this.currentPromise = null;\n  }\n  summary() {\n    const upstreamSummaries = \"TODO: fill in upstream of zip summaries\";\n    return `{${upstreamSummaries}} -> Zip`;\n  }\n  async nextState(afterState) {\n    await afterState;\n    let numIterators = 0;\n    let iteratorsDone = 0;\n    function getNext(container) {\n      if (container instanceof LazyIterator) {\n        const result = container.next();\n        return {\n          value: result.then((x) => {\n            numIterators++;\n            if (x.done) {\n              iteratorsDone++;\n            }\n            return x.value;\n          }),\n          recurse: false\n        };\n      } else {\n        return { value: null, recurse: true };\n      }\n    }\n    const mapped = await deepMapAndAwaitAll(this.iterators, getNext);\n    if (numIterators === iteratorsDone) {\n      return { value: null, done: true };\n    }\n    if (iteratorsDone > 0) {\n      switch (this.mismatchMode) {\n        case ZipMismatchMode.FAIL:\n          throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);\n        case ZipMismatchMode.SHORTEST:\n          return { value: null, done: true };\n        case ZipMismatchMode.LONGEST:\n        default:\n      }\n    }\n    this.count++;\n    return { value: mapped, done: false };\n  }\n  async next() {\n    this.currentPromise = this.nextState(this.currentPromise);\n    return this.currentPromise;\n  }\n};\nvar PrefetchIterator = class extends LazyIterator {\n  constructor(upstream, bufferSize) {\n    super();\n    this.upstream = upstream;\n    this.bufferSize = bufferSize;\n    this.buffer = new RingBuffer(bufferSize);\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Prefetch`;\n  }\n  refill() {\n    while (!this.buffer.isFull()) {\n      const v = this.upstream.next();\n      this.buffer.push(v);\n    }\n  }\n  next() {\n    this.refill();\n    return this.buffer.shift();\n  }\n};\nvar ShuffleIterator = class extends PrefetchIterator {\n  constructor(upstream, windowSize, seed) {\n    super(upstream, windowSize);\n    this.upstream = upstream;\n    this.windowSize = windowSize;\n    this.upstreamExhausted = false;\n    this.random = seedrandom2.alea(seed || util_exports.now().toString());\n    this.lastRead = Promise.resolve({ value: null, done: false });\n  }\n  async next() {\n    this.lastRead = this.lastRead.then(() => this.serialNext());\n    return this.lastRead;\n  }\n  randomInt(max6) {\n    return Math.floor(this.random() * max6);\n  }\n  chooseIndex() {\n    return this.randomInt(this.buffer.length());\n  }\n  async serialNext() {\n    if (!this.upstreamExhausted) {\n      this.refill();\n    }\n    while (!this.buffer.isEmpty()) {\n      const chosenIndex = this.chooseIndex();\n      const result = await this.buffer.shuffleExcise(chosenIndex);\n      if (result.done) {\n        this.upstreamExhausted = true;\n      } else {\n        this.refill();\n        return result;\n      }\n    }\n    return { value: null, done: true };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/dataset.js\nvar Dataset = class {\n  constructor() {\n    this.size = null;\n  }\n  batch(batchSize, smallLastBatch = true) {\n    const base = this;\n    util_exports.assert(batchSize > 0, () => `batchSize needs to be positive, but it is\n      ${batchSize}`);\n    let size;\n    if (this.size === Infinity || this.size == null) {\n      size = this.size;\n    } else if (smallLastBatch) {\n      size = Math.ceil(this.size / batchSize);\n    } else {\n      size = Math.floor(this.size / batchSize);\n    }\n    return datasetFromIteratorFn(async () => {\n      return (await base.iterator()).columnMajorBatch(batchSize, smallLastBatch, deepBatchConcat);\n    }, size);\n  }\n  concatenate(dataset) {\n    const base = this;\n    let size;\n    if (this.size === Infinity || dataset.size === Infinity) {\n      size = Infinity;\n    } else if (this.size != null && dataset.size != null) {\n      size = this.size + dataset.size;\n    } else {\n      size = null;\n    }\n    return datasetFromIteratorFn(async () => (await base.iterator()).concatenate(await dataset.iterator()), size);\n  }\n  filter(predicate) {\n    const base = this;\n    let size;\n    if (this.size === Infinity) {\n      size = Infinity;\n    } else {\n      size = null;\n    }\n    return datasetFromIteratorFn(async () => {\n      return (await base.iterator()).filter((x) => tidy(() => predicate(x)));\n    }, size);\n  }\n  async forEachAsync(f) {\n    return (await this.iterator()).forEachAsync(f);\n  }\n  map(transform5) {\n    const base = this;\n    return datasetFromIteratorFn(async () => {\n      return (await base.iterator()).map((x) => tidy(() => transform5(x)));\n    }, this.size);\n  }\n  mapAsync(transform5) {\n    const base = this;\n    return datasetFromIteratorFn(async () => {\n      return (await base.iterator()).mapAsync(transform5);\n    }, this.size);\n  }\n  prefetch(bufferSize) {\n    if (bufferSize == null) {\n      throw new RangeError(\"`Dataset.prefetch()` requires bufferSize to be specified.\");\n    }\n    const base = this;\n    return datasetFromIteratorFn(async () => (await base.iterator()).prefetch(bufferSize), this.size);\n  }\n  repeat(count2) {\n    const base = this;\n    let size;\n    if (this.size != null && count2 > 0) {\n      size = this.size * count2;\n    } else if (count2 === 0) {\n      size = 0;\n    } else if (this.size != null && (count2 === void 0 || count2 < 0)) {\n      size = Infinity;\n    } else {\n      size = null;\n    }\n    return datasetFromIteratorFn(async () => {\n      const iteratorIterator = iteratorFromFunction(async () => ({ value: await base.iterator(), done: false }));\n      return iteratorFromConcatenated(iteratorIterator.take(count2));\n    }, size);\n  }\n  skip(count2) {\n    const base = this;\n    let size;\n    if (this.size != null && count2 >= 0 && this.size >= count2) {\n      size = this.size - count2;\n    } else if (this.size != null && (this.size < count2 || count2 === void 0 || count2 < 0)) {\n      size = 0;\n    } else {\n      size = null;\n    }\n    return datasetFromIteratorFn(async () => (await base.iterator()).skip(count2), size);\n  }\n  shuffle(bufferSize, seed, reshuffleEachIteration = true) {\n    if (bufferSize == null || bufferSize < 0) {\n      if (this.size == null) {\n        throw new RangeError(\"`Dataset.shuffle()` requires bufferSize to be specified.\");\n      } else {\n        throw new RangeError(`\\`Dataset.shuffle()\\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \\`tf.Tensor\\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);\n      }\n    }\n    const base = this;\n    const random = seedrandom3.alea(seed || util_exports.now().toString());\n    return datasetFromIteratorFn(async () => {\n      let seed2 = random.int32();\n      if (reshuffleEachIteration) {\n        seed2 += random.int32();\n      }\n      return (await base.iterator()).shuffle(bufferSize, seed2.toString());\n    }, this.size);\n  }\n  take(count2) {\n    const base = this;\n    let size;\n    if (this.size != null && this.size > count2) {\n      size = count2;\n    } else if (this.size != null && this.size <= count2) {\n      size = this.size;\n    } else {\n      size = null;\n    }\n    return datasetFromIteratorFn(async () => (await base.iterator()).take(count2), size);\n  }\n  async toArray() {\n    if (this.size === Infinity) {\n      throw new Error(\"Can not convert infinite data stream to array.\");\n    }\n    return (await this.iterator()).toArray();\n  }\n  async toArrayForTest() {\n    if (this.size === Infinity) {\n      throw new Error(\"Can not convert infinite data stream to array.\");\n    }\n    return (await this.iterator()).toArrayForTest();\n  }\n};\nDataset.MAX_BUFFER_SIZE = 1e4;\nfunction datasetFromIteratorFn(iteratorFn, size = null) {\n  return new class extends Dataset {\n    constructor() {\n      super(...arguments);\n      this.size = size;\n    }\n    async iterator() {\n      return iteratorFn();\n    }\n  }();\n}\nfunction array(items) {\n  return datasetFromIteratorFn(async () => iteratorFromItems(items), items.length);\n}\nfunction zip(datasets) {\n  if (!isIterable2(datasets)) {\n    throw new Error(\"The argument to zip() must be an object or array.\");\n  }\n  let size;\n  if (Array.isArray(datasets)) {\n    for (let i = 0; i < datasets.length; i++) {\n      size = size == null ? datasets[i].size : Math.min(size, datasets[i].size);\n    }\n  } else if (datasets instanceof Object) {\n    for (const ds in datasets) {\n      size = size == null ? datasets[ds].size : Math.min(size, datasets[ds].size);\n    }\n  }\n  return datasetFromIteratorFn(async () => {\n    const streams = await deepMapAndAwaitAll(datasets, (d) => {\n      if (d instanceof Dataset) {\n        return { value: d.iterator(), recurse: false };\n      } else if (isIterable2(d)) {\n        return { value: null, recurse: true };\n      } else {\n        throw new Error(\"Leaves of the structure passed to zip() must be Datasets, not primitives.\");\n      }\n    });\n    return iteratorFromZipped(streams, ZipMismatchMode.SHORTEST);\n  }, size);\n}\nfunction deepBatchConcat(rows) {\n  if (rows === null) {\n    return null;\n  }\n  const exampleRow = rows[0];\n  if (canTensorify(exampleRow)) {\n    const value = batchConcat(rows);\n    return { value, recurse: false };\n  }\n  return { value: null, recurse: true };\n}\nfunction batchConcat(arrays) {\n  if (arrays.length === 0) {\n    throw new Error(\"Can't make a batch of zero elements.\");\n  }\n  if (arrays[0] instanceof Tensor) {\n    return stack(arrays);\n  } else {\n    return tensor(arrays);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-data/dist/datasets/text_line_dataset.js\nvar TextLineDataset = class extends Dataset {\n  constructor(input2) {\n    super();\n    this.input = input2;\n  }\n  async iterator() {\n    const inputIterator = await this.input.iterator();\n    const utf8Iterator = inputIterator.decodeUTF8();\n    const lineIterator = utf8Iterator.split(\"\\n\").map((line) => {\n      if (line.endsWith(\"\\r\")) {\n        line = line.slice(0, -1);\n      }\n      return line;\n    });\n    return lineIterator;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/datasets/csv_dataset.js\nvar CODE_QUOTE = '\"';\nvar STATE_OUT = Symbol(\"out\");\nvar STATE_FIELD = Symbol(\"field\");\nvar STATE_QUOTE = Symbol(\"quote\");\nvar STATE_QUOTE_AFTER_QUOTE = Symbol(\"quoteafterquote\");\nvar STATE_WITHIN_QUOTE_IN_QUOTE = Symbol(\"quoteinquote\");\nvar CSVDataset = class extends Dataset {\n  constructor(input2, csvConfig) {\n    super();\n    this.input = input2;\n    this.hasHeader = true;\n    this.fullColumnNames = null;\n    this.columnNamesValidated = false;\n    this.columnConfigs = null;\n    this.configuredColumnsOnly = false;\n    this.delimiter = \",\";\n    this.delimWhitespace = false;\n    this.base = new TextLineDataset(input2);\n    if (!csvConfig) {\n      csvConfig = {};\n    }\n    this.hasHeader = csvConfig.hasHeader === false ? false : true;\n    this.fullColumnNames = csvConfig.columnNames;\n    this.columnConfigs = csvConfig.columnConfigs;\n    this.configuredColumnsOnly = csvConfig.configuredColumnsOnly;\n    if (csvConfig.delimWhitespace) {\n      util_exports.assert(csvConfig.delimiter == null, () => \"Delimiter should not be provided when delimWhitespace is true.\");\n      this.delimWhitespace = true;\n      this.delimiter = \" \";\n    } else {\n      this.delimiter = csvConfig.delimiter ? csvConfig.delimiter : \",\";\n    }\n  }\n  async columnNames() {\n    if (!this.columnNamesValidated) {\n      await this.setColumnNames();\n    }\n    return this.configuredColumnsOnly ? Object.keys(this.columnConfigs) : this.fullColumnNames;\n  }\n  async setColumnNames() {\n    const columnNamesFromFile = await this.maybeReadHeaderLine();\n    if (!this.fullColumnNames && !columnNamesFromFile) {\n      throw new Error(\"Column names must be provided if there is no header line.\");\n    } else if (this.fullColumnNames && columnNamesFromFile) {\n      util_exports.assert(columnNamesFromFile.length === this.fullColumnNames.length, () => \"The length of provided columnNames (\" + this.fullColumnNames.length.toString() + \") does not match the length of the header line read from file (\" + columnNamesFromFile.length.toString() + \").\");\n    }\n    if (!this.fullColumnNames) {\n      this.fullColumnNames = columnNamesFromFile;\n    }\n    const counts = this.fullColumnNames.reduce((countAcc, name) => {\n      countAcc[name] = countAcc[name] + 1 || 1;\n      return countAcc;\n    }, {});\n    const duplicateNames = Object.keys(counts).filter((name) => counts[name] > 1);\n    util_exports.assert(duplicateNames.length === 0, () => \"Duplicate column names found: \" + duplicateNames.toString());\n    if (this.columnConfigs) {\n      for (const key of Object.keys(this.columnConfigs)) {\n        const index = this.fullColumnNames.indexOf(key);\n        if (index === -1) {\n          throw new Error('The key \"' + key + '\" provided in columnConfigs does not match any of the column names (' + this.fullColumnNames.toString() + \").\");\n        }\n      }\n    }\n    this.columnNamesValidated = true;\n  }\n  async maybeReadHeaderLine() {\n    if (this.hasHeader) {\n      const iter = await this.base.iterator();\n      const firstElement = await iter.next();\n      if (firstElement.done) {\n        throw new Error(\"No data was found for CSV parsing.\");\n      }\n      const firstLine = firstElement.value;\n      const headers = this.parseRow(firstLine, false);\n      return headers;\n    } else {\n      return null;\n    }\n  }\n  async iterator() {\n    if (!this.columnNamesValidated) {\n      await this.setColumnNames();\n    }\n    let lines = await this.base.iterator();\n    if (this.hasHeader) {\n      lines = lines.skip(1);\n    }\n    return lines.map((x) => this.makeDataElement(x));\n  }\n  makeDataElement(line) {\n    const values = this.parseRow(line);\n    const features = {};\n    const labels = {};\n    for (let i = 0; i < this.fullColumnNames.length; i++) {\n      const key = this.fullColumnNames[i];\n      const config = this.columnConfigs ? this.columnConfigs[key] : null;\n      if (this.configuredColumnsOnly && !config) {\n        continue;\n      } else {\n        const value = values[i];\n        let parsedValue = null;\n        if (value === \"\") {\n          if (config && config.default !== void 0) {\n            parsedValue = config.default;\n          } else if (config && (config.required || config.isLabel)) {\n            throw new Error(`Required column ${key} is empty in this line: ${line}`);\n          } else {\n            parsedValue = void 0;\n          }\n        } else {\n          const valueAsNum = Number(value);\n          if (isNaN(valueAsNum)) {\n            if (config && config.dtype === \"bool\") {\n              parsedValue = this.getBoolean(value);\n            } else {\n              parsedValue = value;\n            }\n          } else if (!config || !config.dtype) {\n            parsedValue = valueAsNum;\n          } else {\n            switch (config.dtype) {\n              case \"float32\":\n                parsedValue = valueAsNum;\n                break;\n              case \"int32\":\n                parsedValue = Math.floor(valueAsNum);\n                break;\n              case \"bool\":\n                parsedValue = this.getBoolean(value);\n                break;\n              default:\n                parsedValue = valueAsNum;\n            }\n          }\n        }\n        config && config.isLabel ? labels[key] = parsedValue : features[key] = parsedValue;\n      }\n    }\n    if (Object.keys(labels).length === 0) {\n      return features;\n    } else {\n      return { xs: features, ys: labels };\n    }\n  }\n  getBoolean(value) {\n    if (value === \"1\" || value.toLowerCase() === \"true\") {\n      return 1;\n    } else {\n      return 0;\n    }\n  }\n  parseRow(line, validateElementCount = true) {\n    const result = [];\n    let readOffset = 0;\n    const readLength = line.length;\n    let currentState = STATE_OUT;\n    for (let i = 0; i < readLength; i++) {\n      switch (currentState) {\n        case STATE_OUT:\n          switch (line.charAt(i)) {\n            case CODE_QUOTE:\n              readOffset = i + 1;\n              currentState = STATE_QUOTE;\n              break;\n            case this.delimiter:\n              readOffset = i + 1;\n              if (this.delimiter === \" \" && this.delimWhitespace) {\n                break;\n              }\n              result.push(\"\");\n              currentState = STATE_OUT;\n              break;\n            default:\n              currentState = STATE_FIELD;\n              readOffset = i;\n              break;\n          }\n          break;\n        case STATE_FIELD:\n          switch (line.charAt(i)) {\n            case this.delimiter:\n              result.push(line.substring(readOffset, i));\n              currentState = STATE_OUT;\n              readOffset = i + 1;\n              break;\n            default:\n          }\n          break;\n        case STATE_QUOTE:\n          switch (line.charAt(i)) {\n            case CODE_QUOTE:\n              currentState = STATE_QUOTE_AFTER_QUOTE;\n              break;\n            default:\n          }\n          break;\n        case STATE_QUOTE_AFTER_QUOTE:\n          switch (line.charAt(i)) {\n            case this.delimiter:\n              result.push(line.substring(readOffset, i - 1));\n              currentState = STATE_OUT;\n              readOffset = i + 1;\n              break;\n            case CODE_QUOTE:\n              currentState = STATE_QUOTE;\n              break;\n            default:\n              currentState = STATE_WITHIN_QUOTE_IN_QUOTE;\n              break;\n          }\n          break;\n        case STATE_WITHIN_QUOTE_IN_QUOTE:\n          switch (line.charAt(i)) {\n            case CODE_QUOTE:\n              currentState = STATE_QUOTE;\n              break;\n            default:\n          }\n          break;\n        default:\n      }\n    }\n    if (currentState === STATE_QUOTE_AFTER_QUOTE) {\n      result.push(line.substring(readOffset, readLength - 1));\n    } else {\n      result.push(line.substring(readOffset));\n    }\n    if (validateElementCount && result.length !== this.fullColumnNames.length) {\n      throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${result}`);\n    }\n    return result;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/microphone_iterator.js\nvar MicrophoneIterator = class extends LazyIterator {\n  constructor(microphoneConfig) {\n    super();\n    this.microphoneConfig = microphoneConfig;\n    this.isClosed = false;\n    this.fftSize = microphoneConfig.fftSize || 1024;\n    const fftSizeLog2 = Math.log2(this.fftSize);\n    if (this.fftSize < 0 || fftSizeLog2 < 4 || fftSizeLog2 > 14 || !Number.isInteger(fftSizeLog2)) {\n      throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);\n    }\n    this.numFrames = microphoneConfig.numFramesPerSpectrogram || 43;\n    this.sampleRateHz = microphoneConfig.sampleRateHz;\n    this.columnTruncateLength = microphoneConfig.columnTruncateLength || this.fftSize;\n    this.audioTrackConstraints = microphoneConfig.audioTrackConstraints;\n    this.smoothingTimeConstant = microphoneConfig.smoothingTimeConstant || 0;\n    this.includeSpectrogram = microphoneConfig.includeSpectrogram === false ? false : true;\n    this.includeWaveform = microphoneConfig.includeWaveform === true ? true : false;\n    if (!this.includeSpectrogram && !this.includeWaveform) {\n      throw new Error(\"Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.\");\n    }\n  }\n  summary() {\n    return `microphone`;\n  }\n  static async create(microphoneConfig = {}) {\n    if (!env().get(\"IS_BROWSER\")) {\n      throw new Error(\"microphone API is only supported in browser environment.\");\n    }\n    const microphoneIterator = new MicrophoneIterator(microphoneConfig);\n    await microphoneIterator.start();\n    return microphoneIterator;\n  }\n  async start() {\n    try {\n      this.stream = await navigator.mediaDevices.getUserMedia({\n        audio: this.audioTrackConstraints == null ? true : this.audioTrackConstraints,\n        video: false\n      });\n    } catch (e) {\n      throw new Error(`Error thrown while initializing video stream: ${e.message}`);\n    }\n    if (!this.stream) {\n      throw new Error(\"Could not obtain audio from microphone.\");\n    }\n    const ctxConstructor = window.AudioContext || window.webkitAudioContext;\n    this.audioContext = new ctxConstructor();\n    if (!this.sampleRateHz) {\n      this.sampleRateHz = this.audioContext.sampleRate;\n    } else if (this.audioContext.sampleRate !== this.sampleRateHz) {\n      throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);\n    }\n    const streamSource = this.audioContext.createMediaStreamSource(this.stream);\n    this.analyser = this.audioContext.createAnalyser();\n    this.analyser.fftSize = this.fftSize * 2;\n    this.analyser.smoothingTimeConstant = this.smoothingTimeConstant;\n    streamSource.connect(this.analyser);\n    this.freqData = new Float32Array(this.fftSize);\n    this.timeData = new Float32Array(this.fftSize);\n    return;\n  }\n  async next() {\n    if (this.isClosed) {\n      return { value: null, done: true };\n    }\n    let spectrogramTensor;\n    let waveformTensor;\n    const audioDataQueue = await this.getAudioData();\n    if (this.includeSpectrogram) {\n      const freqData = this.flattenQueue(audioDataQueue.freqDataQueue);\n      spectrogramTensor = this.getTensorFromAudioDataArray(freqData, [this.numFrames, this.columnTruncateLength, 1]);\n    }\n    if (this.includeWaveform) {\n      const timeData = this.flattenQueue(audioDataQueue.timeDataQueue);\n      waveformTensor = this.getTensorFromAudioDataArray(timeData, [this.numFrames * this.fftSize, 1]);\n    }\n    return {\n      value: { \"spectrogram\": spectrogramTensor, \"waveform\": waveformTensor },\n      done: false\n    };\n  }\n  async capture() {\n    return (await this.next()).value;\n  }\n  async getAudioData() {\n    const freqDataQueue = [];\n    const timeDataQueue = [];\n    let currentFrames = 0;\n    return new Promise((resolve) => {\n      const intervalID = setInterval(() => {\n        if (this.includeSpectrogram) {\n          this.analyser.getFloatFrequencyData(this.freqData);\n          if (this.freqData[0] === -Infinity) {\n            resolve({ freqDataQueue, timeDataQueue });\n          }\n          freqDataQueue.push(this.freqData.slice(0, this.columnTruncateLength));\n        }\n        if (this.includeWaveform) {\n          this.analyser.getFloatTimeDomainData(this.timeData);\n          timeDataQueue.push(this.timeData.slice());\n        }\n        if (++currentFrames === this.numFrames) {\n          clearInterval(intervalID);\n          resolve({ freqDataQueue, timeDataQueue });\n        }\n      }, this.fftSize / this.sampleRateHz * 1e3);\n    });\n  }\n  stop() {\n    if (!this.isClosed) {\n      this.isClosed = true;\n      this.analyser.disconnect();\n      this.audioContext.close();\n      if (this.stream != null && this.stream.getTracks().length > 0) {\n        this.stream.getTracks()[0].stop();\n      }\n    }\n  }\n  toArray() {\n    throw new Error(\"Can not convert infinite audio stream to array.\");\n  }\n  getSampleRate() {\n    return this.sampleRateHz;\n  }\n  flattenQueue(queue) {\n    const frameSize = queue[0].length;\n    const freqData = new Float32Array(queue.length * frameSize);\n    queue.forEach((data, i) => freqData.set(data, i * frameSize));\n    return freqData;\n  }\n  getTensorFromAudioDataArray(freqData, shape) {\n    const vals = new Float32Array(util_exports.sizeFromShape(shape));\n    vals.set(freqData, vals.length - freqData.length);\n    return tensor(vals, shape);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/webcam_iterator.js\nvar WebcamIterator = class extends LazyIterator {\n  constructor(webcamVideoElement, webcamConfig) {\n    super();\n    this.webcamVideoElement = webcamVideoElement;\n    this.webcamConfig = webcamConfig;\n    this.isClosed = true;\n    this.resize = false;\n    if (this.needToResize()) {\n      this.resize = true;\n      this.cropSize = [this.webcamConfig.resizeHeight, this.webcamConfig.resizeWidth];\n      this.cropBoxInd = tensor1d([0], \"int32\");\n      if (this.webcamConfig.centerCrop) {\n        const widthCroppingRatio = this.webcamConfig.resizeWidth * 1 / this.webcamVideoElement.width;\n        const heightCroppingRatio = this.webcamConfig.resizeHeight * 1 / this.webcamVideoElement.height;\n        const widthCropStart = (1 - widthCroppingRatio) / 2;\n        const heightCropStart = (1 - heightCroppingRatio) / 2;\n        const widthCropEnd = widthCropStart + widthCroppingRatio;\n        const heightCropEnd = heightCroppingRatio + heightCropStart;\n        this.cropBox = tensor2d([heightCropStart, widthCropStart, heightCropEnd, widthCropEnd], [1, 4]);\n      } else {\n        this.cropBox = tensor2d([0, 0, 1, 1], [1, 4]);\n      }\n    }\n  }\n  summary() {\n    return `webcam`;\n  }\n  static async create(webcamVideoElement, webcamConfig = {}) {\n    if (!env().get(\"IS_BROWSER\")) {\n      throw new Error(\"tf.data.webcam is only supported in browser environment.\");\n    }\n    if (!webcamVideoElement) {\n      webcamVideoElement = document.createElement(\"video\");\n      if (!webcamConfig.resizeWidth || !webcamConfig.resizeHeight) {\n        throw new Error(\"Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.\");\n      }\n      webcamVideoElement.width = webcamConfig.resizeWidth;\n      webcamVideoElement.height = webcamConfig.resizeHeight;\n    }\n    const webcamIterator = new WebcamIterator(webcamVideoElement, webcamConfig);\n    await webcamIterator.start();\n    return webcamIterator;\n  }\n  async start() {\n    if (this.webcamConfig.facingMode) {\n      util_exports.assert(this.webcamConfig.facingMode === \"user\" || this.webcamConfig.facingMode === \"environment\", () => `Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);\n    }\n    try {\n      this.stream = await navigator.mediaDevices.getUserMedia({\n        video: {\n          deviceId: this.webcamConfig.deviceId,\n          facingMode: this.webcamConfig.facingMode ? this.webcamConfig.facingMode : \"user\",\n          width: this.webcamVideoElement.width,\n          height: this.webcamVideoElement.height\n        }\n      });\n    } catch (e) {\n      e.message = `Error thrown while initializing video stream: ${e.message}`;\n      throw e;\n    }\n    if (!this.stream) {\n      throw new Error(\"Could not obtain video from webcam.\");\n    }\n    try {\n      this.webcamVideoElement.srcObject = this.stream;\n    } catch (error) {\n      console.log(error);\n      this.webcamVideoElement.src = window.URL.createObjectURL(this.stream);\n    }\n    this.webcamVideoElement.play();\n    this.isClosed = false;\n    return new Promise((resolve) => {\n      this.webcamVideoElement.onloadedmetadata = () => {\n        resolve();\n      };\n    });\n  }\n  async next() {\n    if (this.isClosed) {\n      return { value: null, done: true };\n    }\n    let img;\n    try {\n      img = browser_exports.fromPixels(this.webcamVideoElement);\n    } catch (e) {\n      throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`);\n    }\n    if (this.resize) {\n      try {\n        return { value: this.cropAndResizeFrame(img), done: false };\n      } catch (e) {\n        throw new Error(`Error thrown cropping the video: ${e.message}`);\n      } finally {\n        img.dispose();\n      }\n    } else {\n      return { value: img, done: false };\n    }\n  }\n  needToResize() {\n    if (this.webcamConfig.resizeWidth && this.webcamConfig.resizeHeight && (this.webcamVideoElement.width !== this.webcamConfig.resizeWidth || this.webcamVideoElement.height !== this.webcamConfig.resizeHeight)) {\n      return true;\n    }\n    return false;\n  }\n  cropAndResizeFrame(img) {\n    return tidy(() => {\n      const expandedImage = expandDims(cast(img, \"float32\"), 0);\n      let resizedImage;\n      resizedImage = image.cropAndResize(expandedImage, this.cropBox, this.cropBoxInd, this.cropSize, \"bilinear\");\n      const shape = resizedImage.shape;\n      return reshape(resizedImage, shape.slice(1));\n    });\n  }\n  async capture() {\n    return (await this.next()).value;\n  }\n  stop() {\n    const tracks = this.stream.getTracks();\n    tracks.forEach((track) => track.stop());\n    try {\n      this.webcamVideoElement.srcObject = null;\n    } catch (error) {\n      console.log(error);\n      this.webcamVideoElement.src = null;\n    }\n    this.isClosed = true;\n  }\n  toArray() {\n    throw new Error(\"Can not convert infinite video stream to array.\");\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/datasource.js\nvar DataSource = class {\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/string_iterator.js\nvar StringIterator = class extends LazyIterator {\n  split(separator) {\n    return new SplitIterator(this, separator);\n  }\n};\nvar SplitIterator = class extends StringIterator {\n  constructor(upstream, separator) {\n    super();\n    this.upstream = upstream;\n    this.impl = new SplitIteratorImpl(upstream, separator);\n  }\n  summary() {\n    return this.impl.summary();\n  }\n  async next() {\n    return this.impl.next();\n  }\n};\nvar SplitIteratorImpl = class extends OneToManyIterator {\n  constructor(upstream, separator) {\n    super();\n    this.upstream = upstream;\n    this.separator = separator;\n    this.carryover = \"\";\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Split('${this.separator}')`;\n  }\n  async pump() {\n    const chunkResult = await this.upstream.next();\n    if (chunkResult.done) {\n      if (this.carryover === \"\") {\n        return false;\n      }\n      this.outputQueue.push(this.carryover);\n      this.carryover = \"\";\n      return true;\n    }\n    const lines = chunkResult.value.split(this.separator);\n    lines[0] = this.carryover + lines[0];\n    for (const line of lines.slice(0, -1)) {\n      this.outputQueue.push(line);\n    }\n    this.carryover = lines[lines.length - 1];\n    return true;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/byte_chunk_iterator.js\nvar ByteChunkIterator = class extends LazyIterator {\n  decodeUTF8() {\n    return new Utf8Iterator(this);\n  }\n};\nvar Utf8Iterator = class extends StringIterator {\n  constructor(upstream) {\n    super();\n    this.upstream = upstream;\n    this.impl = new Utf8IteratorImpl(upstream);\n  }\n  summary() {\n    return this.impl.summary();\n  }\n  async next() {\n    return this.impl.next();\n  }\n};\nvar Utf8IteratorImpl = class extends OneToManyIterator {\n  constructor(upstream) {\n    super();\n    this.upstream = upstream;\n    if (env().get(\"IS_BROWSER\")) {\n      this.decoder = new TextDecoder(\"utf-8\");\n    } else {\n      const { StringDecoder } = require_string_decoder();\n      this.decoder = new StringDecoder(\"utf8\");\n    }\n  }\n  summary() {\n    return `${this.upstream.summary()} -> Utf8`;\n  }\n  async pump() {\n    const chunkResult = await this.upstream.next();\n    let chunk;\n    if (chunkResult.done) {\n      return false;\n    } else {\n      chunk = chunkResult.value;\n    }\n    let text;\n    if (env().get(\"IS_BROWSER\")) {\n      text = this.decoder.decode(chunk, { stream: true });\n    } else {\n      text = this.decoder.write(Buffer.from(chunk.buffer));\n    }\n    this.outputQueue.push(text);\n    return true;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/file_chunk_iterator.js\nvar FileChunkIterator = class extends ByteChunkIterator {\n  constructor(file, options = {}) {\n    super();\n    this.file = file;\n    this.options = options;\n    util_exports.assert(file instanceof Uint8Array || (env().get(\"IS_BROWSER\") ? file instanceof File || file instanceof Blob : false), () => \"FileChunkIterator only supports File, Blob and Uint8Array right now.\");\n    this.offset = options.offset || 0;\n    this.chunkSize = options.chunkSize || 1024 * 1024;\n  }\n  summary() {\n    return `FileChunks ${this.file}`;\n  }\n  async next() {\n    if (this.offset >= (this.file instanceof Uint8Array ? this.file.byteLength : this.file.size)) {\n      return { value: null, done: true };\n    }\n    const chunk = new Promise((resolve, reject) => {\n      const end = this.offset + this.chunkSize;\n      if (this.file instanceof Uint8Array) {\n        resolve(new Uint8Array(this.file.slice(this.offset, end)));\n      } else {\n        const fileReader = new FileReader();\n        fileReader.onload = (event) => {\n          let data = fileReader.result;\n          if (data instanceof ArrayBuffer) {\n            data = new Uint8Array(data);\n          }\n          if (!(data instanceof Uint8Array)) {\n            return reject(new TypeError(\"FileReader returned unknown type.\"));\n          }\n          resolve(data);\n        };\n        fileReader.onabort = (event) => {\n          return reject(new Error(\"Aborted\"));\n        };\n        fileReader.onerror = (event) => {\n          return reject(new Error(event.type));\n        };\n        const slice5 = this.file.slice(this.offset, end);\n        fileReader.readAsArrayBuffer(slice5);\n      }\n      this.offset = end;\n    });\n    return { value: await chunk, done: false };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/iterators/url_chunk_iterator.js\nasync function urlChunkIterator(url, options = {}, fetchFunc) {\n  let urlString;\n  let requestInit;\n  if (typeof url === \"string\") {\n    urlString = url;\n  } else {\n    urlString = url.url;\n    requestInit = getRequestInitFromRequest(url);\n  }\n  const response = await (fetchFunc || util_exports.fetch)(urlString, requestInit);\n  if (response.ok) {\n    const uint8Array = new Uint8Array(await response.arrayBuffer());\n    return new FileChunkIterator(uint8Array, options);\n  } else {\n    throw new Error(response.statusText);\n  }\n}\nvar getRequestInitFromRequest = (request) => {\n  const init2 = {\n    method: request.method,\n    headers: request.headers,\n    body: request.body,\n    mode: request.mode,\n    credentials: request.credentials,\n    cache: request.cache,\n    redirect: request.redirect,\n    referrer: request.referrer,\n    integrity: request.integrity\n  };\n  return init2;\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/util/source_util.js\nfunction isLocalPath(source) {\n  return typeof source === \"string\" && source.slice(0, 7) === \"file://\";\n}\n\n// node_modules/@tensorflow/tfjs-data/dist/sources/file_data_source.js\nvar FileDataSource = class extends DataSource {\n  constructor(input2, options = {}) {\n    super();\n    this.input = input2;\n    this.options = options;\n  }\n  async iterator() {\n    if (isLocalPath(this.input) && env().get(\"IS_NODE\")) {\n      const fs = require_fs();\n      this.input = fs.readFileSync(this.input.slice(7));\n    }\n    return new FileChunkIterator(this.input, this.options);\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/sources/url_data_source.js\nvar URLDataSource = class extends DataSource {\n  constructor(url, fileOptions = {}) {\n    super();\n    this.url = url;\n    this.fileOptions = fileOptions;\n  }\n  async iterator() {\n    if (isLocalPath(this.url)) {\n      return new FileDataSource(this.url, this.fileOptions).iterator();\n    } else {\n      return urlChunkIterator(this.url, this.fileOptions);\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-data/dist/readers.js\nfunction csv(source, csvConfig = {}) {\n  return new CSVDataset(new URLDataSource(source), csvConfig);\n}\nfunction func(f) {\n  const iter = iteratorFromFunction(f);\n  return datasetFromIteratorFn(async () => iter);\n}\nfunction generator(generator2) {\n  return datasetFromIteratorFn(async () => {\n    const gen = await generator2();\n    return iteratorFromFunction(() => gen.next());\n  });\n}\nasync function webcam(webcamVideoElement, webcamConfig) {\n  return WebcamIterator.create(webcamVideoElement, webcamConfig);\n}\nasync function microphone(microphoneConfig) {\n  return MicrophoneIterator.create(microphoneConfig);\n}\n\n// node_modules/@tensorflow/tfjs-data/dist/version.js\nvar version4 = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/cpu_util.js\nfunction assertNotComplex(tensor2, opName) {\n  if (!Array.isArray(tensor2)) {\n    tensor2 = [tensor2];\n  }\n  tensor2.forEach((t) => {\n    if (t != null) {\n      util_exports.assert(t.dtype !== \"complex64\", () => `${opName} does not support complex64 tensors in the CPU backend.`);\n    }\n  });\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/backend_cpu.js\nvar whereImpl2 = kernel_impls_exports.whereImpl;\nvar MathBackendCPU = class extends KernelBackend {\n  constructor() {\n    super();\n    this.blockSize = 48;\n    this.firstUse = true;\n    this.data = new DataStorage(this, engine());\n  }\n  nextDataId() {\n    return MathBackendCPU.nextDataId++;\n  }\n  write(values, shape, dtype) {\n    if (this.firstUse) {\n      this.firstUse = false;\n      if (env().get(\"IS_NODE\")) {\n        backend_util_exports.warn(\"\\n============================\\nHi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. \\n============================\");\n      }\n    }\n    const dataId = { id: this.nextDataId() };\n    this.data.set(dataId, { values, dtype, refCount: 1 });\n    return dataId;\n  }\n  makeTensorInfo(shape, dtype, values) {\n    let outId;\n    if (dtype === \"string\" && values != null && values.length > 0 && util_exports.isString(values[0])) {\n      const encodedValues = values.map((d) => util_exports.encodeString(d));\n      outId = this.write(encodedValues, shape, dtype);\n    } else {\n      outId = this.write(values, shape, dtype);\n    }\n    return { dataId: outId, shape, dtype };\n  }\n  refCount(dataId) {\n    if (this.data.has(dataId)) {\n      const tensorData = this.data.get(dataId);\n      return tensorData.refCount;\n    }\n    return 0;\n  }\n  incRef(dataId) {\n    const tensorData = this.data.get(dataId);\n    tensorData.refCount++;\n  }\n  decRef(dataId) {\n    if (this.data.has(dataId)) {\n      const tensorData = this.data.get(dataId);\n      tensorData.refCount--;\n    }\n  }\n  move(dataId, values, shape, dtype, refCount) {\n    this.data.set(dataId, { values, dtype, refCount });\n  }\n  numDataIds() {\n    return this.data.numDataIds();\n  }\n  async read(dataId) {\n    return this.readSync(dataId);\n  }\n  readSync(dataId) {\n    const { dtype, complexTensorInfos } = this.data.get(dataId);\n    if (dtype === \"complex64\") {\n      const realValues = this.readSync(complexTensorInfos.real.dataId);\n      const imagValues = this.readSync(complexTensorInfos.imag.dataId);\n      return backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);\n    }\n    return this.data.get(dataId).values;\n  }\n  bufferSync(t) {\n    const data = this.readSync(t.dataId);\n    if (t.dtype === \"string\") {\n      try {\n        const strings = data.map((d) => util_exports.decodeString(d));\n        return buffer(t.shape, t.dtype, strings);\n      } catch (_a) {\n        throw new Error(\"Failed to decode encoded string bytes into utf-8\");\n      }\n    }\n    return buffer(t.shape, t.dtype, data);\n  }\n  makeOutput(values, shape, dtype) {\n    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);\n  }\n  disposeData(dataId, force = false) {\n    if (this.data.has(dataId)) {\n      this.data.get(dataId).refCount--;\n      if (!force && this.data.get(dataId).refCount > 0) {\n        return false;\n      }\n      const { complexTensorInfos } = this.data.get(dataId);\n      if (complexTensorInfos != null) {\n        this.disposeData(complexTensorInfos.real.dataId, true);\n        this.disposeData(complexTensorInfos.imag.dataId, true);\n      }\n      this.data.delete(dataId);\n    }\n    return true;\n  }\n  disposeIntermediateTensorInfo(tensorInfo) {\n    this.disposeData(tensorInfo.dataId);\n  }\n  async time(f) {\n    const start = util_exports.now();\n    f();\n    const kernelMs = util_exports.now() - start;\n    return { kernelMs };\n  }\n  memory() {\n    return {\n      unreliable: true,\n      reasons: [\"The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less.\"]\n    };\n  }\n  where(condition) {\n    assertNotComplex([condition], \"where\");\n    const condVals = this.readSync(condition.dataId);\n    return whereImpl2(condition.shape, condVals);\n  }\n  dispose() {\n  }\n  floatPrecision() {\n    return 32;\n  }\n  epsilon() {\n    return super.epsilon();\n  }\n};\nMathBackendCPU.nextDataId = 0;\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/shared.js\nvar shared_exports = {};\n__export(shared_exports, {\n  addImpl: () => addImpl,\n  bincountImpl: () => bincountImpl,\n  bincountReduceImpl: () => bincountReduceImpl,\n  castImpl: () => castImpl,\n  ceilImpl: () => ceilImpl,\n  concatImpl: () => concatImpl,\n  equalImpl: () => equalImpl,\n  expImpl: () => expImpl,\n  expm1Impl: () => expm1Impl,\n  floorImpl: () => floorImpl,\n  gatherNdImpl: () => gatherNdImpl,\n  gatherV2Impl: () => gatherV2Impl,\n  greaterEqualImpl: () => greaterEqualImpl,\n  greaterImpl: () => greaterImpl,\n  lessEqualImpl: () => lessEqualImpl,\n  lessImpl: () => lessImpl,\n  linSpaceImpl: () => linSpaceImpl,\n  logImpl: () => logImpl,\n  maxImpl: () => maxImpl,\n  maximumImpl: () => maximumImpl,\n  minimumImpl: () => minimumImpl,\n  multiplyImpl: () => multiplyImpl,\n  negImpl: () => negImpl,\n  notEqualImpl: () => notEqualImpl,\n  prodImpl: () => prodImpl,\n  raggedGatherImpl: () => raggedGatherImpl,\n  raggedRangeImpl: () => raggedRangeImpl,\n  raggedTensorToTensorImpl: () => raggedTensorToTensorImpl,\n  rangeImpl: () => rangeImpl,\n  rsqrtImpl: () => rsqrtImpl,\n  scatterImpl: () => scatterImpl,\n  sigmoidImpl: () => sigmoidImpl,\n  simpleAbsImpl: () => simpleAbsImpl,\n  sliceImpl: () => sliceImpl,\n  sparseFillEmptyRowsImpl: () => sparseFillEmptyRowsImpl,\n  sparseReshapeImpl: () => sparseReshapeImpl,\n  sparseSegmentReductionImpl: () => sparseSegmentReductionImpl,\n  sqrtImpl: () => sqrtImpl,\n  squaredDifferenceImpl: () => squaredDifferenceImpl,\n  stridedSliceImpl: () => stridedSliceImpl,\n  stringNGramsImpl: () => stringNGramsImpl,\n  stringSplitImpl: () => stringSplitImpl,\n  stringToHashBucketFastImpl: () => stringToHashBucketFastImpl,\n  subImpl: () => subImpl,\n  tileImpl: () => tileImpl,\n  topKImpl: () => topKImpl,\n  transposeImpl: () => transposeImpl,\n  uniqueImpl: () => uniqueImpl\n});\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Abs.js\nfunction simpleAbsImpl(vals) {\n  const resultValues = new Float32Array(vals.length);\n  for (let i = 0; i < vals.length; ++i) {\n    resultValues[i] = Math.abs(vals[i]);\n  }\n  return resultValues;\n}\nvar abs2 = (args) => {\n  const { x } = args.inputs;\n  const cpuBackend = args.backend;\n  assertNotComplex(x, \"abs\");\n  let resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));\n  const values = cpuBackend.data.get(x.dataId).values;\n  resultValues = simpleAbsImpl(values);\n  return cpuBackend.makeOutput(resultValues, x.shape, x.dtype);\n};\nvar absConfig = {\n  kernelName: Abs,\n  backendName: \"cpu\",\n  kernelFunc: abs2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/binary_impl.js\nfunction createSimpleBinaryKernelImpl(op2) {\n  return (aShape, bShape, aVals, bVals, dtype) => {\n    const newShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);\n    const resultRank = newShape.length;\n    const resultStrides = util_exports.computeStrides(newShape);\n    const resultSize = util_exports.sizeFromShape(newShape);\n    const result = util_exports.getTypedArrayFromDType(dtype, resultSize);\n    const aRank = aShape.length;\n    const bRank = bShape.length;\n    const aStrides = util_exports.computeStrides(aShape);\n    const bStrides = util_exports.computeStrides(bShape);\n    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, newShape);\n    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, newShape);\n    if (aBroadcastDims.length + bBroadcastDims.length === 0) {\n      for (let i = 0; i < result.length; ++i) {\n        result[i] = op2(aVals[i % aVals.length], bVals[i % bVals.length]);\n      }\n    } else {\n      for (let i = 0; i < result.length; ++i) {\n        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);\n        const aLoc = loc.slice(-aRank);\n        aBroadcastDims.forEach((d) => aLoc[d] = 0);\n        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);\n        const bLoc = loc.slice(-bRank);\n        bBroadcastDims.forEach((d) => bLoc[d] = 0);\n        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);\n        result[i] = op2(aVals[aIndex], bVals[bIndex]);\n      }\n    }\n    return [result, newShape];\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Complex.js\nfunction complex2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { real: real4, imag: imag4 } = inputs;\n  const realVals = backend2.data.get(real4.dataId).values;\n  const imagVals = backend2.data.get(imag4.dataId).values;\n  const complexInfo = backend2.makeTensorInfo(real4.shape, \"complex64\");\n  const complex4 = backend2.data.get(complexInfo.dataId);\n  complex4.complexTensorInfos = {\n    real: backend2.makeTensorInfo(real4.shape, \"float32\", realVals),\n    imag: backend2.makeTensorInfo(imag4.shape, \"float32\", imagVals)\n  };\n  return complexInfo;\n}\nvar complexConfig = {\n  kernelName: Complex,\n  backendName: \"cpu\",\n  kernelFunc: complex2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/zeros_impl.js\nfunction zeros3(backend2, shape, dtype = \"float32\") {\n  if (dtype === \"complex64\") {\n    const real4 = zeros3(backend2, shape, \"float32\");\n    const imag4 = zeros3(backend2, shape, \"float32\");\n    return complex2({ inputs: { real: real4, imag: imag4 }, backend: backend2 });\n  }\n  const values = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(shape), dtype);\n  return backend2.makeTensorInfo(shape, dtype, values);\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Identity.js\nfunction identity2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  backend2.incRef(x.dataId);\n  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };\n}\nvar identityConfig = {\n  kernelName: Identity,\n  backendName: \"cpu\",\n  kernelFunc: identity2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Real.js\nfunction real2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const real4 = backend2.data.get(input2.dataId).complexTensorInfos.real;\n  const realVal = backend2.data.get(real4.dataId).values;\n  return backend2.makeTensorInfo(real4.shape, real4.dtype, realVal);\n}\nvar realConfig = {\n  kernelName: Real,\n  backendName: \"cpu\",\n  kernelFunc: real2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cast.js\nfunction castImpl(values, shape, inputType, dtype) {\n  if (dtype === \"int32\") {\n    const resultValues = Int32Array.from(values);\n    return [shape, \"int32\", resultValues];\n  }\n  if (dtype === \"bool\") {\n    const zero = util_exports.toTypedArray([0], inputType);\n    const [resultData, resultShape] = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0)(shape, [], values, zero, \"bool\");\n    return [resultShape, \"bool\", resultData];\n  }\n  throw new Error(`Error in Cast: failed to cast ${inputType} to ${dtype}`);\n}\nfunction cast3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { dtype } = attrs;\n  if (dtype === \"complex64\") {\n    if (x.dtype === \"complex64\") {\n      return identity2({ inputs: { x }, backend: backend2 });\n    }\n    const zerosTensorInfo = zeros3(backend2, x.shape, x.dtype);\n    const floatX = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: \"float32\" } });\n    const result = complex2({ inputs: { real: floatX, imag: zerosTensorInfo }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);\n    backend2.disposeIntermediateTensorInfo(floatX);\n    return result;\n  }\n  if (x.dtype === \"complex64\") {\n    const realPart = real2({ inputs: { input: x }, backend: backend2 });\n    const result = cast3({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    return result;\n  }\n  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {\n    const result = identity2({ inputs: { x }, backend: backend2 });\n    return { dataId: result.dataId, shape: result.shape, dtype };\n  }\n  const values = backend2.data.get(x.dataId).values;\n  const [resultShape, resultType, resultData] = castImpl(values, x.shape, x.dtype, dtype);\n  return backend2.makeTensorInfo(resultShape, resultType, resultData);\n}\nvar castConfig = {\n  kernelName: Cast,\n  backendName: \"cpu\",\n  kernelFunc: cast3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/binary_utils.js\nfunction binaryKernelFunc(name, simpleImpl, complexImpl, dtype) {\n  if (complexImpl == null) {\n    return ({ inputs, backend: backend2 }) => {\n      const { a, b } = inputs;\n      const cpuBackend = backend2;\n      assertNotComplex([a, b], name);\n      const aVals = cpuBackend.data.get(a.dataId).values;\n      const bVals = cpuBackend.data.get(b.dataId).values;\n      const decodedAVals = a.dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;\n      const decodedBVals = a.dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;\n      const $dtype = dtype || a.dtype;\n      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);\n      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);\n    };\n  }\n  return ({ inputs, backend: backend2 }) => {\n    const { a, b } = inputs;\n    const cpuBackend = backend2;\n    if (a.dtype === \"complex64\" || b.dtype === \"complex64\") {\n      const $aComplex = cast3({ inputs: { x: a }, backend: cpuBackend, attrs: { dtype: \"complex64\" } });\n      const $aComplexVals = cpuBackend.data.get($aComplex.dataId);\n      const aReal = $aComplexVals.complexTensorInfos.real;\n      const aImag = $aComplexVals.complexTensorInfos.imag;\n      const aRealVals = cpuBackend.data.get(aReal.dataId).values;\n      const aImagVals = cpuBackend.data.get(aImag.dataId).values;\n      const $bComplex = cast3({ inputs: { x: b }, backend: cpuBackend, attrs: { dtype: \"complex64\" } });\n      const $bComplexVals = cpuBackend.data.get($bComplex.dataId);\n      const bReal = $bComplexVals.complexTensorInfos.real;\n      const bImag = $bComplexVals.complexTensorInfos.imag;\n      const bRealVals = cpuBackend.data.get(bReal.dataId).values;\n      const bImagVals = cpuBackend.data.get(bImag.dataId).values;\n      const [resultRealData, resultImagData, resultShape] = complexImpl(a.shape, b.shape, aRealVals, aImagVals, bRealVals, bImagVals);\n      const resultReal = cpuBackend.makeTensorInfo(resultShape, \"float32\", resultRealData);\n      const resultImag = cpuBackend.makeTensorInfo(resultShape, \"float32\", resultImagData);\n      const result = complex2({ inputs: { real: resultReal, imag: resultImag }, backend: cpuBackend });\n      cpuBackend.disposeIntermediateTensorInfo($aComplex);\n      cpuBackend.disposeIntermediateTensorInfo($bComplex);\n      cpuBackend.disposeIntermediateTensorInfo(resultReal);\n      cpuBackend.disposeIntermediateTensorInfo(resultImag);\n      return result;\n    } else {\n      const aVals = cpuBackend.data.get(a.dataId).values;\n      const bVals = cpuBackend.data.get(b.dataId).values;\n      const $dtype = dtype || a.dtype;\n      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, aVals, bVals, $dtype);\n      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);\n    }\n  };\n}\nfunction createComplexBinaryKernelImpl(op2) {\n  return (aShape, bShape, aRealVals, aImagVals, bRealVals, bImagVals) => {\n    const resultShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);\n    const resultSize = util_exports.sizeFromShape(resultShape);\n    const resultRank = resultShape.length;\n    const resultStrides = util_exports.computeStrides(resultShape);\n    const resultRealVals = util_exports.getTypedArrayFromDType(\"float32\", resultSize);\n    const resultImagVals = util_exports.getTypedArrayFromDType(\"float32\", resultSize);\n    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, resultShape);\n    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, resultShape);\n    const aVals = backend_util_exports.mergeRealAndImagArrays(aRealVals, aImagVals);\n    const bVals = backend_util_exports.mergeRealAndImagArrays(bRealVals, bImagVals);\n    const aRank = aShape.length;\n    const aStrides = util_exports.computeStrides(aShape);\n    const bRank = bShape.length;\n    const bStrides = util_exports.computeStrides(bShape);\n    if (aBroadcastDims.length + bBroadcastDims.length === 0) {\n      for (let i = 0; i < resultRealVals.length; i++) {\n        const aIdx = i % aVals.length;\n        const bIdx = i % bVals.length;\n        const result = op2(aVals[aIdx * 2], aVals[aIdx * 2 + 1], bVals[bIdx * 2], bVals[bIdx * 2 + 1]);\n        resultRealVals[i] = result.real;\n        resultImagVals[i] = result.imag;\n      }\n    } else {\n      for (let i = 0; i < resultRealVals.length; i++) {\n        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);\n        const aLoc = loc.slice(-aRank);\n        aBroadcastDims.forEach((d) => aLoc[d] = 0);\n        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);\n        const bLoc = loc.slice(-bRank);\n        bBroadcastDims.forEach((d) => bLoc[d] = 0);\n        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);\n        const opResult = op2(aVals[aIndex * 2], aVals[aIndex * 2 + 1], bVals[bIndex * 2], bVals[bIndex * 2 + 1]);\n        resultRealVals[i] = opResult.real;\n        resultImagVals[i] = opResult.imag;\n      }\n    }\n    return [resultRealVals, resultImagVals, resultShape];\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Add.js\nvar addImpl = createSimpleBinaryKernelImpl((a, b) => a + b);\nvar addComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {\n  return { real: aReal + bReal, imag: aImag + bImag };\n});\nvar add4 = binaryKernelFunc(Add, addImpl, addComplexImpl);\nvar addConfig = {\n  kernelName: Add,\n  backendName: \"cpu\",\n  kernelFunc: add4\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Bincount_impl.js\nfunction bincountImpl(xVals, weightsVals, weightsDtype, weightsShape, size) {\n  const weightsSize = util_exports.sizeFromShape(weightsShape);\n  const outVals = util_exports.makeZerosTypedArray(size, weightsDtype);\n  for (let i = 0; i < xVals.length; i++) {\n    const value = xVals[i];\n    if (value < 0) {\n      throw new Error(\"Input x must be non-negative!\");\n    }\n    if (value >= size) {\n      continue;\n    }\n    if (weightsSize > 0) {\n      outVals[value] += weightsVals[i];\n    } else {\n      outVals[value] += 1;\n    }\n  }\n  return outVals;\n}\nfunction bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput = false) {\n  const numRows = xBuf.shape[0];\n  const numCols = xBuf.shape[1];\n  const outBuf = buffer([numRows, size], weightsBuf.dtype);\n  for (let i = 0; i < numRows; i++) {\n    for (let j = 0; j < numCols; j++) {\n      const value = xBuf.get(i, j);\n      if (value < 0) {\n        throw new Error(\"Input x must be non-negative!\");\n      }\n      if (value >= size) {\n        continue;\n      }\n      if (binaryOutput) {\n        outBuf.set(1, i, value);\n      } else {\n        if (weightsBuf.size > 0) {\n          outBuf.set(outBuf.get(i, value) + weightsBuf.get(i, j), i, value);\n        } else {\n          outBuf.set(outBuf.get(i, value) + 1, i, value);\n        }\n      }\n    }\n  }\n  return outBuf;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/unary_impl.js\nfunction createSimpleUnaryImpl(op2) {\n  return (values, dtype, attrs) => {\n    const newValues = util_exports.getTypedArrayFromDType(dtype, values.length);\n    for (let i = 0; i < values.length; ++i) {\n      newValues[i] = op2(values[i], attrs);\n    }\n    return newValues;\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/unary_utils.js\nfunction unaryKernelFunc(name, op2, dtype) {\n  return ({ inputs, attrs, backend: backend2 }) => {\n    const { x } = inputs;\n    assertNotComplex(x, name);\n    if (x.dtype === \"string\" || dtype === \"string\") {\n      throw new Error(\"unaryKernelFunc does not support string input/output\");\n    }\n    const cpuBackend = backend2;\n    const values = cpuBackend.data.get(x.dataId).values;\n    const xSize = util_exports.sizeFromShape(x.shape);\n    const $dtype = dtype || x.dtype;\n    const newValues = util_exports.getArrayFromDType($dtype, xSize);\n    for (let i = 0; i < xSize; ++i) {\n      newValues[i] = op2(values[i], attrs);\n    }\n    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);\n  };\n}\nfunction unaryKernelFuncFromImpl(name, unaryImpl, dtype) {\n  return ({ inputs, attrs, backend: backend2 }) => {\n    const { x } = inputs;\n    assertNotComplex(x, name);\n    if (x.dtype === \"string\" || dtype === \"string\") {\n      throw new Error(\"unaryKernelFunc does not support string input/output\");\n    }\n    const cpuBackend = backend2;\n    const values = cpuBackend.data.get(x.dataId).values;\n    const $dtype = dtype || x.dtype;\n    const newValues = unaryImpl(values, $dtype, attrs);\n    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Ceil.js\nvar ceilImpl = createSimpleUnaryImpl((xi) => Math.ceil(xi));\nvar ceil2 = unaryKernelFuncFromImpl(Ceil, ceilImpl);\nvar ceilConfig = {\n  kernelName: Ceil,\n  backendName: \"cpu\",\n  kernelFunc: ceil2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Concat_impl.js\nfunction concatImpl(inputs, outShape, dtype, simplyConcat) {\n  const outVals = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(outShape));\n  if (simplyConcat && dtype !== \"string\") {\n    let offset = 0;\n    inputs.forEach((input2) => {\n      const size = util_exports.sizeFromShape(input2.shape);\n      outVals.set(input2.vals, offset);\n      offset += size;\n    });\n  } else {\n    let colOffset = 0;\n    inputs.forEach((input2) => {\n      const decodedData = dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(input2.vals) : input2.vals;\n      let tIdx = 0;\n      for (let row = 0; row < input2.shape[0]; ++row) {\n        const resIdx = row * outShape[1] + colOffset;\n        for (let col = 0; col < input2.shape[1]; ++col) {\n          outVals[resIdx + col] = decodedData[tIdx++];\n        }\n      }\n      colOffset += input2.shape[1];\n    });\n  }\n  return outVals;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Equal.js\nvar equalImpl = createSimpleBinaryKernelImpl((a, b) => a === b ? 1 : 0);\nvar equal2 = binaryKernelFunc(Equal, equalImpl, null, \"bool\");\nvar equalConfig = {\n  kernelName: Equal,\n  backendName: \"cpu\",\n  kernelFunc: equal2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Exp.js\nvar expImpl = createSimpleUnaryImpl((xi) => Math.exp(xi));\nvar exp2 = unaryKernelFuncFromImpl(Exp, expImpl, \"float32\");\nvar expConfig = {\n  kernelName: Exp,\n  backendName: \"cpu\",\n  kernelFunc: exp2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Expm1.js\nvar expm1Impl = createSimpleUnaryImpl((xi) => Math.expm1(xi));\nvar expm12 = unaryKernelFuncFromImpl(Expm1, expm1Impl);\nvar expm1Config = {\n  kernelName: Expm1,\n  backendName: \"cpu\",\n  kernelFunc: expm12\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Floor.js\nvar floorImpl = createSimpleUnaryImpl((xi) => Math.floor(xi));\nvar floor2 = unaryKernelFuncFromImpl(Floor, floorImpl);\nvar floorConfig = {\n  kernelName: Floor,\n  backendName: \"cpu\",\n  kernelFunc: floor2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherNd_Impl.js\nfunction gatherNdImpl(indicesData, paramsBuf, dtype, numSlices, sliceRank, sliceSize, strides, paramsShape, paramsSize) {\n  const outBuf = buffer([numSlices, sliceSize], dtype);\n  for (let i = 0; i < numSlices; i++) {\n    const index = [];\n    let flattenIndex = 0;\n    for (let j = 0; j < sliceRank; j++) {\n      const dim = indicesData[i * sliceRank + j];\n      flattenIndex += dim * strides[j];\n      index.push(dim);\n    }\n    if (flattenIndex < 0 || flattenIndex >= paramsSize / sliceSize) {\n      throw new Error(`Invalid indices: ${index} does not index into ${paramsShape}`);\n    }\n    for (let k = 0; k < sliceSize; k++) {\n      outBuf.values[i * sliceSize + k] = paramsBuf.get(...paramsBuf.indexToLoc(flattenIndex * sliceSize + k));\n    }\n  }\n  return outBuf;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherV2_impl.js\nfunction gatherV2Impl(xBuf, indicesBuf, flattenOutputShape) {\n  const outBuf = buffer(flattenOutputShape, xBuf.dtype);\n  for (let i = 0; i < outBuf.size; ++i) {\n    const newLoc = outBuf.indexToLoc(i);\n    const originalLoc = newLoc.slice();\n    const batchIdx = originalLoc[0];\n    const indicesIdx = originalLoc[2];\n    const indicesIndex = indicesBuf.locToIndex([batchIdx, indicesIdx]);\n    originalLoc[2] = indicesBuf.values[indicesIndex];\n    const originalIndex = xBuf.locToIndex(originalLoc);\n    if (0 <= originalIndex && originalIndex < xBuf.values.length) {\n      outBuf.values[i] = xBuf.values[originalIndex];\n    }\n  }\n  return outBuf;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Greater.js\nvar greaterImpl = createSimpleBinaryKernelImpl((a, b) => a > b ? 1 : 0);\nvar greater3 = binaryKernelFunc(Greater, greaterImpl, null, \"bool\");\nvar greaterConfig = {\n  kernelName: Greater,\n  backendName: \"cpu\",\n  kernelFunc: greater3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GreaterEqual.js\nvar greaterEqualImpl = createSimpleBinaryKernelImpl((a, b) => a >= b ? 1 : 0);\nvar greaterEqual2 = binaryKernelFunc(GreaterEqual, greaterEqualImpl, null, \"bool\");\nvar greaterEqualConfig = {\n  kernelName: GreaterEqual,\n  backendName: \"cpu\",\n  kernelFunc: greaterEqual2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Less.js\nvar lessImpl = createSimpleBinaryKernelImpl((a, b) => a < b ? 1 : 0);\nvar less3 = binaryKernelFunc(Less, lessImpl, null, \"bool\");\nvar lessConfig = {\n  kernelName: Less,\n  backendName: \"cpu\",\n  kernelFunc: less3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LessEqual.js\nvar lessEqualImpl = createSimpleBinaryKernelImpl((a, b) => a <= b ? 1 : 0);\nvar lessEqual2 = binaryKernelFunc(LessEqual, lessEqualImpl, null, \"bool\");\nvar lessEqualConfig = {\n  kernelName: LessEqual,\n  backendName: \"cpu\",\n  kernelFunc: lessEqual2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LinSpace_impl.js\nfunction linSpaceImpl(start, stop, num) {\n  const step5 = (stop - start) / (num - 1);\n  const values = util_exports.makeZerosTypedArray(num, \"float32\");\n  values[0] = start;\n  for (let i = 1; i < values.length; i++) {\n    values[i] = values[i - 1] + step5;\n  }\n  return values;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Log.js\nvar logImpl = createSimpleUnaryImpl((xi) => Math.log(xi));\nvar log3 = unaryKernelFuncFromImpl(Log, logImpl);\nvar logConfig = {\n  kernelName: Log,\n  backendName: \"cpu\",\n  kernelFunc: log3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Max_impl.js\nfunction maxImpl(aVals, reduceSize, outShape, dtype) {\n  const vals = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(outShape));\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let max6 = aVals[offset];\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      if (Number.isNaN(value) || value > max6) {\n        max6 = value;\n      }\n    }\n    vals[i] = max6;\n  }\n  return vals;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Maximum.js\nvar maximumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.max(aValue, bValue));\nvar maximum3 = binaryKernelFunc(Maximum, maximumImpl);\nvar maximumConfig = {\n  kernelName: Maximum,\n  backendName: \"cpu\",\n  kernelFunc: maximum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Minimum.js\nvar minimumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.min(aValue, bValue));\nvar minimum3 = binaryKernelFunc(Minimum, minimumImpl);\nvar minimumConfig = {\n  kernelName: Minimum,\n  backendName: \"cpu\",\n  kernelFunc: minimum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multiply.js\nvar multiplyImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue * bValue);\nvar multiplyComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {\n  return {\n    real: aReal * bReal - aImag * bImag,\n    imag: aReal * bImag + aImag * bReal\n  };\n});\nvar multiply2 = binaryKernelFunc(Multiply, multiplyImpl, multiplyComplexImpl);\nvar multiplyConfig = {\n  kernelName: Multiply,\n  backendName: \"cpu\",\n  kernelFunc: multiply2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Neg.js\nfunction negImpl(xVals, xShape, xDtype) {\n  const minusOne = util_exports.createScalarValue(-1, xDtype);\n  return multiplyImpl([], xShape, minusOne, xVals, xDtype);\n}\nfunction neg2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  assertNotComplex(x, \"neg\");\n  const xVals = backend2.data.get(x.dataId).values;\n  const [res, newShape] = negImpl(xVals, x.shape, x.dtype);\n  return backend2.makeTensorInfo(newShape, x.dtype, res);\n}\nvar negConfig = {\n  kernelName: Neg,\n  backendName: \"cpu\",\n  kernelFunc: neg2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NotEqual.js\nvar notEqualImpl = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0);\nvar notEqual2 = binaryKernelFunc(NotEqual, notEqualImpl, null, \"bool\");\nvar notEqualConfig = {\n  kernelName: NotEqual,\n  backendName: \"cpu\",\n  kernelFunc: notEqual2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transpose_impl.js\nfunction transposeImpl(xVals, xShape, dtype, perm, newShape) {\n  const xRank = xShape.length;\n  const xSize = util_exports.sizeFromShape(xShape);\n  const xStrides = util_exports.computeStrides(xShape);\n  const newStrides = util_exports.computeStrides(newShape);\n  const result = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(newShape));\n  for (let i = 0; i < xSize; ++i) {\n    const loc = util_exports.indexToLoc(i, xRank, xStrides);\n    const newLoc = new Array(loc.length);\n    for (let i2 = 0; i2 < newLoc.length; i2++) {\n      newLoc[i2] = loc[perm[i2]];\n    }\n    const newIndex = util_exports.locToIndex(newLoc, xRank, newStrides);\n    result[newIndex] = xVals[i];\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transpose.js\nfunction transpose2(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x } = inputs;\n  const { perm } = attrs;\n  assertNotComplex(x, \"transpose\");\n  const xRank = x.shape.length;\n  const newShape = new Array(xRank);\n  for (let i = 0; i < newShape.length; i++) {\n    newShape[i] = x.shape[perm[i]];\n  }\n  const values = backend2.data.get(x.dataId).values;\n  const result = transposeImpl(values, x.shape, x.dtype, perm, newShape);\n  const dataId = backend2.write(result, newShape, x.dtype);\n  return { dataId, shape: newShape, dtype: x.dtype };\n}\nvar transposeConfig = {\n  kernelName: Transpose,\n  backendName: \"cpu\",\n  kernelFunc: transpose2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Prod.js\nfunction prodImpl(xShape, xDtype, xVals, reductionAxes) {\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, reductionAxes);\n  const outDtype = upcastType(xDtype, \"int32\");\n  const outVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), outDtype);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  for (let i = 0; i < outVals.length; ++i) {\n    const offset = i * reduceSize;\n    let prod5 = 1;\n    for (let j = 0; j < reduceSize; ++j) {\n      prod5 *= xVals[offset + j];\n    }\n    outVals[i] = prod5;\n  }\n  return { outVals, outShape, outDtype };\n}\nfunction prod2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  assertNotComplex(x, \"prod\");\n  const xRank = x.shape.length;\n  const axes = util_exports.parseAxisParam(axis, x.shape);\n  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);\n  let reductionAxes = axes;\n  let permutedX = x;\n  const intermediateTensorInfos = [];\n  if (permutation != null) {\n    permutedX = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });\n    intermediateTensorInfos.push(permutedX);\n    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);\n  }\n  const xVals = backend2.data.get(permutedX.dataId).values;\n  const { outVals, outShape, outDtype } = prodImpl(permutedX.shape, permutedX.dtype, xVals, reductionAxes);\n  let resultShape = outShape;\n  if (keepDims) {\n    resultShape = backend_util_exports.expandShapeToKeepDim(outShape, axes);\n  }\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return backend2.makeTensorInfo(resultShape, outDtype, outVals);\n}\nvar prodConfig = {\n  kernelName: Prod,\n  backendName: \"cpu\",\n  kernelFunc: prod2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedGather_impl.js\nfunction validateIndices(indices, indicesShape, numParams) {\n  indices.forEach((index, i) => {\n    if (index < 0 || index >= numParams) {\n      const locString = util_exports.indexToLoc(i, indicesShape.length, util_exports.computeStrides(indicesShape)).join(\",\");\n      throw new Error(`indices[${locString}] = ${index} is not in [0, ${numParams})`);\n    }\n  });\n}\nfunction validateSplits(paramsNestedSplits, numParamsDenseValues) {\n  for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {\n    const splits = paramsNestedSplits[dim];\n    const lastSplit = dim === paramsNestedSplits.length - 1 ? numParamsDenseValues : paramsNestedSplits[dim + 1].length;\n    if (splits.length === 0) {\n      throw new Error(\"Ragged splits may not be empty\");\n    }\n    if (splits[0] < 0) {\n      throw new Error(\"Ragged splits must be non-negative\");\n    }\n    if (splits[splits.length - 1] > lastSplit) {\n      throw new Error(\"Ragged splits must not point past values\");\n    }\n    for (let i = 1; i < splits.length; ++i) {\n      if (splits[i - 1] > splits[i]) {\n        throw new Error(\"Ragged splits must be sorted in ascending order\");\n      }\n    }\n  }\n}\nfunction makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues) {\n  const valueSlices = [];\n  let numValues = 0;\n  const numSplits = indicesShape.length - 1 + paramsNestedSplits.length;\n  const outSplits = new Array(numSplits).fill(null).map(() => [0]);\n  validateSplits(paramsNestedSplits, numParamsDenseValues);\n  let nrows = 1;\n  for (let dim = 0; dim < indicesShape.length - 1; ++dim) {\n    nrows *= indicesShape[dim];\n    const rowLength = indicesShape[dim + 1];\n    for (let i = 1; i < nrows + 1; ++i) {\n      outSplits[dim].push(i * rowLength);\n    }\n  }\n  for (let i = 0; i < indices.length; ++i) {\n    let start = indices[i];\n    let limit = indices[i] + 1;\n    for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {\n      const splits = paramsNestedSplits[dim];\n      const outDim = dim + indicesShape.length - 1;\n      if (outDim >= 0) {\n        const outSplitsOutDim = outSplits[outDim];\n        const delta = outSplitsOutDim[outSplitsOutDim.length - 1] - splits[start];\n        for (let j = start; j < limit; ++j) {\n          outSplits[outDim].push(splits[j + 1] + delta);\n        }\n      }\n      start = splits[start];\n      limit = splits[limit];\n    }\n    if (limit !== start) {\n      valueSlices.push([start, limit]);\n      numValues += limit - start;\n    }\n  }\n  return { outSplits, valueSlices, numValues };\n}\nfunction getSplits(outSplits) {\n  const splitsOut = [];\n  for (let i = 0; i < outSplits.length; ++i) {\n    const numSplits = outSplits[i].length;\n    const splits = util_exports.getArrayFromDType(\"int32\", numSplits);\n    splitsOut.push(splits);\n    outSplits[i].forEach((value, j) => splits[j] = value);\n  }\n  return splitsOut;\n}\nfunction computeFlatOuterDims(orig, numOutDims) {\n  const outDims = orig.slice(0, numOutDims);\n  while (outDims.length < numOutDims) {\n    outDims.push(1);\n  }\n  for (let inDim = numOutDims; inDim < orig.length; inDim++) {\n    outDims[numOutDims - 1] *= orig[inDim];\n  }\n  return outDims;\n}\nfunction writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, values, valuesShape) {\n  const denseM = computeFlatOuterDims(paramsDenseValuesShape, 2)[1];\n  const valuesM = computeFlatOuterDims(valuesShape, 2)[1];\n  let outPos = 0;\n  for (const slice5 of valueSlices) {\n    for (let i = slice5[0]; i < slice5[1]; ++i) {\n      for (let j = 0; j < valueSize; ++j) {\n        values[outPos * valuesM + j] = paramsDenseValues[i * denseM + j];\n      }\n      ++outPos;\n    }\n  }\n}\nfunction getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues) {\n  const valuesShape = paramsDenseValuesShape.slice();\n  valuesShape[0] = numValues;\n  const valuesOut = util_exports.getArrayFromDType(paramsDenseValuesDType, util_exports.sizeFromShape(valuesShape));\n  const numElements = paramsDenseValues.length;\n  const valueSize = numElements === 0 ? 0 : numElements / paramsDenseValuesShape[0];\n  writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, valuesOut, valuesShape);\n  return [valuesOut, valuesShape];\n}\nfunction raggedGatherImpl(paramsNestedSplits, paramsNestedSplitsShapes, paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, indices, indicesShape, outputRaggedRank) {\n  if (paramsNestedSplits.length === 0) {\n    throw new Error(\"paramsNestedSplits must be non empty\");\n  }\n  if (paramsNestedSplitsShapes[0].length === 0) {\n    throw new Error(\"Split tensors must not be scalars\");\n  }\n  const numParams = paramsNestedSplitsShapes[0][0] - 1;\n  validateIndices(indices, indicesShape, numParams);\n  if (paramsDenseValuesShape.length === 0) {\n    throw new Error(\"params.rank must be nonzero\");\n  }\n  const numParamsDenseValues = paramsDenseValuesShape[0];\n  const { outSplits, valueSlices, numValues } = makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues);\n  const outputNestedSplits = getSplits(outSplits);\n  const outputDenseValues = getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues);\n  return [outputNestedSplits, outputDenseValues[0], outputDenseValues[1]];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedRange_impl.js\nvar INT32_MAX2 = 2147483647;\nfunction raggedRangeImpl(starts, startsShape, startsDType, limits, limitsShape, deltas, deltasShape) {\n  if (startsShape.length > 1) {\n    throw new Error(\"starts must be a scalar or vector\");\n  }\n  if (limitsShape.length > 1) {\n    throw new Error(\"limits must be a scalar or vector\");\n  }\n  if (deltasShape.length > 1) {\n    throw new Error(\"deltas must be a scalar or vector\");\n  }\n  const broadcastStarts = startsShape.length === 0;\n  const broadcastLimits = limitsShape.length === 0;\n  const broadcastDeltas = deltasShape.length === 0;\n  const inSizes = [];\n  if (!broadcastStarts) {\n    inSizes.push(startsShape[0]);\n  }\n  if (!broadcastLimits) {\n    inSizes.push(limitsShape[0]);\n  }\n  if (!broadcastDeltas) {\n    inSizes.push(deltasShape[0]);\n  }\n  for (let i = 1; i < inSizes.length; ++i) {\n    if (inSizes[i] !== inSizes[i - 1]) {\n      throw new Error(\"starts, limits, and deltas must have the same shape\");\n    }\n  }\n  const nRows = inSizes.length === 0 ? 1 : inSizes[0];\n  const rtNestedSplits = util_exports.getArrayFromDType(\"int32\", nRows + 1);\n  rtNestedSplits[0] = 0;\n  for (let row = 0; row < nRows; ++row) {\n    const start = broadcastStarts ? starts[0] : starts[row];\n    const limit = broadcastLimits ? limits[0] : limits[row];\n    const delta = broadcastDeltas ? deltas[0] : deltas[row];\n    if (delta === 0) {\n      throw new Error(\"Requires delta != 0\");\n    }\n    let size;\n    if (delta > 0 && limit < start || delta < 0 && limit > start) {\n      size = 0;\n    } else {\n      size = Math.ceil(Math.abs((limit - start) / delta));\n      if (size > INT32_MAX2) {\n        throw new Error(`Requires ((limit - start) / delta) <= ${INT32_MAX2}`);\n      }\n    }\n    rtNestedSplits[row + 1] = rtNestedSplits[row] + size;\n  }\n  const nVals = rtNestedSplits[nRows];\n  const rtDenseValues = util_exports.getArrayFromDType(startsDType, nVals);\n  let valueIndex = 0;\n  for (let row = 0; row < nRows; ++row) {\n    const rowSize = rtNestedSplits[row + 1] - rtNestedSplits[row];\n    let value = broadcastStarts ? starts[0] : starts[row];\n    const delta = broadcastDeltas ? deltas[0] : deltas[row];\n    for (let i = 0; i < rowSize; ++i) {\n      rtDenseValues[valueIndex++] = value;\n      value += delta;\n    }\n  }\n  return [rtNestedSplits, rtDenseValues];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedTensorToTensor_impl.js\nvar RowPartitionType2 = backend_util_exports.RowPartitionType;\nvar RaggedTensorToTensorOp = class {\n  constructor(shape, shapeShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypeStrings) {\n    this.shape = shape;\n    this.shapeShape = shapeShape;\n    this.values = values;\n    this.valuesShape = valuesShape;\n    this.valuesDType = valuesDType;\n    this.defaultValue = defaultValue;\n    this.defaultValueShape = defaultValueShape;\n    this.rowPartitionValues = rowPartitionValues;\n    this.rowPartitionValuesShapes = rowPartitionValuesShapes;\n    this.rowPartitionTypes = backend_util_exports.getRowPartitionTypesHelper(rowPartitionTypeStrings);\n    this.raggedRank = backend_util_exports.getRaggedRank(this.rowPartitionTypes);\n  }\n  getRowPartitionTypeByDimension(dimension) {\n    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {\n      return this.rowPartitionTypes[dimension + 1];\n    } else {\n      return this.rowPartitionTypes[dimension];\n    }\n  }\n  getRowPartitionTensor(dimension) {\n    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {\n      return this.rowPartitionValues[dimension + 1];\n    } else {\n      return this.rowPartitionValues[dimension];\n    }\n  }\n  getMaxWidth(dimension) {\n    const rowPartitionTensor = this.getRowPartitionTensor(dimension - 1);\n    switch (this.getRowPartitionTypeByDimension(dimension - 1)) {\n      case RowPartitionType2.VALUE_ROWIDS:\n        return RaggedTensorToTensorOp.getMaxWidthValueRowID(rowPartitionTensor);\n      case RowPartitionType2.ROW_SPLITS:\n        return RaggedTensorToTensorOp.getMaxWidthRowSplit(rowPartitionTensor);\n      default:\n        throw new Error(`Cannot handle partition type ${RowPartitionType2[this.getRowPartitionTypeByDimension(dimension - 1)]}`);\n    }\n  }\n  static getMaxWidthRowSplit(rowSplit) {\n    const tensorLength = rowSplit.length;\n    if (tensorLength === 0 || tensorLength === 1) {\n      return 0;\n    }\n    let maxWidth = 0;\n    for (let i = 0; i < tensorLength - 1; ++i) {\n      const currentWidth = rowSplit[i + 1] - rowSplit[i];\n      if (currentWidth > maxWidth) {\n        maxWidth = currentWidth;\n      }\n    }\n    return maxWidth;\n  }\n  static getMaxWidthValueRowID(valueRowIds) {\n    const indexLength = valueRowIds.length;\n    if (indexLength === 0) {\n      return 0;\n    }\n    let firstEqualIndex = 0;\n    let firstEqualIndexValue = valueRowIds[0];\n    let maxWidth = 0;\n    for (let i = 1; i < indexLength; ++i) {\n      const value = valueRowIds[i];\n      if (value !== firstEqualIndexValue) {\n        firstEqualIndexValue = value;\n        maxWidth = Math.max(i - firstEqualIndex, maxWidth);\n        firstEqualIndex = i;\n      }\n    }\n    return Math.max(indexLength - firstEqualIndex, maxWidth);\n  }\n  tensorShapeFromTensor(t, tShape, isPartial = true) {\n    if (tShape.length === 0) {\n      if (t[0] === -1) {\n        return [];\n      }\n      throw new Error(`The only valid scalar shape tensor is the fully unknown shape specified as -1.`);\n    }\n    return makeShape(t, isPartial);\n  }\n  calculateOutputSize(firstDim) {\n    const valueShape = this.valuesShape;\n    const defaultValueShape = this.defaultValueShape;\n    backend_util_exports.validateDefaultValueShape(defaultValueShape, valueShape);\n    const shape = this.tensorShapeFromTensor(this.shape, this.shapeShape);\n    const outputShape = backend_util_exports.combineRaggedTensorToTensorShapes(this.raggedRank, shape, valueShape);\n    const result = outputShape;\n    if (result[0] < 0) {\n      result[0] = firstDim;\n    }\n    for (let i = 1; i <= this.raggedRank; ++i) {\n      if (result[i] < 0) {\n        result[i] = this.getMaxWidth(i);\n      }\n    }\n    return result;\n  }\n  calculateFirstParentOutputIndex(firstDimension, outputIndexMultiplier, firstDimensionOutput) {\n    const minDimension = Math.min(firstDimension, firstDimensionOutput);\n    const result = [];\n    let currentOutputIndex = 0;\n    for (let i = 0; i < minDimension; ++i, currentOutputIndex += outputIndexMultiplier) {\n      result.push(currentOutputIndex);\n    }\n    for (let i = minDimension; i < firstDimension; ++i) {\n      result.push(-1);\n    }\n    util_exports.assert(result.length === firstDimension, () => \"Final length of result must be equal to firstDimension.\");\n    return result;\n  }\n  calculateOutputIndexRowSplit(rowSplit, parentOutputIndex, outputIndexMultiplier, outputSize) {\n    const rowSplitSize = rowSplit.length;\n    const result = [];\n    for (let i = 0; i < rowSplitSize - 1; ++i) {\n      const rowLength = rowSplit[i + 1] - rowSplit[i];\n      let realLength = Math.min(outputSize, rowLength);\n      let parentOutputIndexCurrent = parentOutputIndex[i];\n      if (parentOutputIndexCurrent === -1) {\n        realLength = 0;\n      }\n      for (let j = 0; j < realLength; ++j) {\n        result.push(parentOutputIndexCurrent);\n        parentOutputIndexCurrent += outputIndexMultiplier;\n      }\n      for (let j = 0; j < rowLength - realLength; ++j) {\n        result.push(-1);\n      }\n    }\n    if (rowSplitSize > 0 && result.length !== rowSplit[rowSplitSize - 1]) {\n      throw new Error(\"Invalid row split size.\");\n    }\n    return result;\n  }\n  calculateOutputIndexValueRowID(valueRowIds, parentOutputIndex, outputIndexMultiplier, outputSize) {\n    const indexSize = valueRowIds.length;\n    const result = [];\n    if (indexSize === 0) {\n      return [];\n    }\n    let currentOutputColumn = 0;\n    let currentValueRowId = valueRowIds[0];\n    if (currentValueRowId >= parentOutputIndex.length) {\n      throw new Error(`Got currentValueRowId=${currentValueRowId}, which is not less than ${parentOutputIndex.length}`);\n    }\n    let currentOutputIndex = parentOutputIndex[currentValueRowId];\n    result.push(currentOutputIndex);\n    for (let i = 1; i < indexSize; ++i) {\n      const nextValueRowId = valueRowIds[i];\n      if (nextValueRowId === currentValueRowId) {\n        if (currentOutputIndex >= 0) {\n          ++currentOutputColumn;\n          if (currentOutputColumn < outputSize) {\n            currentOutputIndex += outputIndexMultiplier;\n          } else {\n            currentOutputIndex = -1;\n          }\n        }\n      } else {\n        currentOutputColumn = 0;\n        currentValueRowId = nextValueRowId;\n        if (nextValueRowId >= parentOutputIndex.length) {\n          throw new Error(`Got nextValueRowId=${nextValueRowId} which is not less than ${parentOutputIndex.length}`);\n        }\n        currentOutputIndex = parentOutputIndex[nextValueRowId];\n      }\n      result.push(currentOutputIndex);\n    }\n    if (result.length !== valueRowIds.length) {\n      throw new Error(\"Invalid row ids.\");\n    }\n    return result;\n  }\n  calculateOutputIndex(dimension, parentOutputIndex, outputIndexMultiplier, outputSize) {\n    const rowPartitionTensor = this.getRowPartitionTensor(dimension);\n    const partitionType = this.getRowPartitionTypeByDimension(dimension);\n    switch (partitionType) {\n      case RowPartitionType2.VALUE_ROWIDS:\n        return this.calculateOutputIndexValueRowID(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);\n      case RowPartitionType2.ROW_SPLITS:\n        if (rowPartitionTensor.length - 1 > parentOutputIndex.length) {\n          throw new Error(`Row partition size is greater than output size: ${rowPartitionTensor.length - 1} > ${parentOutputIndex.length}`);\n        }\n        return this.calculateOutputIndexRowSplit(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);\n      default:\n        throw new Error(`Unsupported partition type: ${RowPartitionType2[partitionType]}`);\n    }\n  }\n  getFirstDimensionSize() {\n    const firstPartitionTensor = this.rowPartitionValues[0];\n    if (this.rowPartitionTypes.length === 0) {\n      throw new Error(\"No row_partition_types given.\");\n    }\n    const firstPartitionType = this.rowPartitionTypes[0];\n    switch (firstPartitionType) {\n      case RowPartitionType2.FIRST_DIM_SIZE:\n        return firstPartitionTensor[0];\n      case RowPartitionType2.VALUE_ROWIDS:\n        throw new Error(\"Cannot handle VALUE_ROWIDS in first dimension.\");\n      case RowPartitionType2.ROW_SPLITS:\n        return this.rowPartitionValuesShapes[0][0] - 1;\n      default:\n        throw new Error(`Cannot handle type ${RowPartitionType2[firstPartitionType]}`);\n    }\n  }\n  compute() {\n    const firstPartitionTensor = this.rowPartitionValues[0];\n    if (firstPartitionTensor.length <= 0) {\n      throw new Error(\"Invalid first partition input. Tensor requires at least one element.\");\n    }\n    const firstDimension = this.getFirstDimensionSize();\n    const outputSize = this.calculateOutputSize(firstDimension);\n    const multiplier = new Array(this.raggedRank + 1);\n    multiplier[multiplier.length - 1] = 1;\n    for (let i = multiplier.length - 2; i >= 0; --i) {\n      multiplier[i] = multiplier[i + 1] * outputSize[i + 1];\n    }\n    const outputShape = makeShape(outputSize, false);\n    const outputTensor = util_exports.getArrayFromDType(this.valuesDType, util_exports.sizeFromShape(outputShape));\n    const fullSize = multiplier[0] * outputSize[0];\n    if (fullSize > 0) {\n      let outputIndex = this.calculateFirstParentOutputIndex(firstDimension, multiplier[0], outputSize[0]);\n      for (let i = 1; i <= this.raggedRank; ++i) {\n        const newOutputIndex = this.calculateOutputIndex(i - 1, outputIndex, multiplier[i], outputSize[i]);\n        outputIndex = newOutputIndex;\n      }\n      this.setOutput(this.raggedRank, outputIndex, outputTensor, outputShape);\n    }\n    return [outputShape, outputTensor];\n  }\n  setOutput(raggedRank, outputIndex, outputTensor, outputShape) {\n    if (outputTensor.length === 0) {\n      return;\n    }\n    const valuesBase = this.values;\n    const outputBase = outputTensor;\n    let elementShape = outputShape.slice();\n    elementShape = elementShape.slice(raggedRank + 1);\n    const valueElementSize = util_exports.sizeFromShape(elementShape);\n    const outputIndexSize = outputIndex.length;\n    let defaultValue = this.defaultValue;\n    if (defaultValue.length !== valueElementSize && defaultValue.length !== 1) {\n      const srcShape = this.defaultValueShape;\n      tidy(() => {\n        const defaultValueTensor = reshape(defaultValue, srcShape);\n        const bCastDefault = broadcastTo(defaultValueTensor, elementShape);\n        defaultValue = bCastDefault.dataSync();\n      });\n    }\n    let srcStart = 0;\n    let dstStart = 0;\n    let dstEnd = 0;\n    for (let srcI = 0; srcI <= outputIndexSize; ++srcI) {\n      let dstI = srcI < outputIndexSize ? outputIndex[srcI] : -1;\n      if (dstI === dstEnd) {\n        ++dstEnd;\n        continue;\n      }\n      if (dstStart < dstEnd) {\n        const src = valuesBase.subarray(srcStart * valueElementSize);\n        const dst = outputBase.subarray(dstStart * valueElementSize);\n        const nVals = (dstEnd - dstStart) * valueElementSize;\n        copyArray(dst, src, nVals);\n      }\n      if (srcI >= outputIndexSize) {\n        const outputSize = outputTensor.length;\n        dstI = Math.floor(outputSize / valueElementSize);\n      }\n      if (dstI > dstEnd) {\n        if (this.defaultValue.length === 1) {\n          outputBase.subarray(dstEnd * valueElementSize, dstI * valueElementSize).fill(this.defaultValue[0]);\n          dstEnd = dstI;\n        } else {\n          while (dstI > dstEnd) {\n            const dst = outputBase.slice(dstEnd * valueElementSize);\n            copyArray(dst, defaultValue, valueElementSize);\n            ++dstEnd;\n          }\n        }\n      }\n      if (dstI < 0) {\n        srcStart = srcI + 1;\n        dstStart = dstEnd;\n      } else {\n        srcStart = srcI;\n        dstStart = dstEnd;\n        dstEnd = dstStart + 1;\n      }\n    }\n  }\n};\nfunction copyArray(dst, src, size) {\n  for (let i = 0; i < size; i++) {\n    dst[i] = src[i];\n  }\n}\nfunction makeShape(shape, isPartial) {\n  const out = [];\n  for (let dim of shape) {\n    if (dim < 0) {\n      if (!isPartial) {\n        throw new Error(`Dimension ${dim} must be >= 0`);\n      }\n      if (dim < -1) {\n        throw new Error(`Dimension ${dim} must be >= -1`);\n      }\n      dim = -1;\n    }\n    out.push(dim);\n  }\n  return out;\n}\nfunction raggedTensorToTensorImpl(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes) {\n  return new RaggedTensorToTensorOp(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes).compute();\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Range_impl.js\nfunction rangeImpl(start, stop, step5, dtype) {\n  const sameStartStop = start === stop;\n  const increasingRangeNegativeStep = start < stop && step5 < 0;\n  const decreasingRangePositiveStep = stop < start && step5 > 1;\n  if (sameStartStop || increasingRangeNegativeStep || decreasingRangePositiveStep) {\n    return util_exports.makeZerosTypedArray(0, dtype);\n  }\n  const numElements = Math.abs(Math.ceil((stop - start) / step5));\n  const values = util_exports.makeZerosTypedArray(numElements, dtype);\n  if (stop < start && step5 === 1) {\n    step5 = -1;\n  }\n  values[0] = start;\n  for (let i = 1; i < values.length; i++) {\n    values[i] = values[i - 1] + step5;\n  }\n  return values;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Rsqrt.js\nvar rsqrtImpl = createSimpleUnaryImpl((xi) => 1 / Math.sqrt(xi));\nvar rsqrt2 = unaryKernelFuncFromImpl(Rsqrt, rsqrtImpl);\nvar rsqrtConfig = {\n  kernelName: Rsqrt,\n  backendName: \"cpu\",\n  kernelFunc: rsqrt2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Scatter_impl.js\nfunction scatterImpl(indices, updates, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, defaultValue, sumDupeIndices) {\n  const flattenShape = [outputSize / sliceSize, sliceSize];\n  const indicesData = indices.values;\n  const updatesData = updates.values;\n  if (outputSize === 0) {\n    return buffer(shape, updates.dtype);\n  }\n  const outBuf = buffer(flattenShape, updates.dtype);\n  if (typeof defaultValue === \"string\") {\n    outBuf.values.fill(defaultValue);\n  } else if (typeof defaultValue === \"number\") {\n    outBuf.values.fill(defaultValue);\n  } else if (typeof defaultValue === \"boolean\") {\n    outBuf.values.fill(+defaultValue);\n  }\n  for (let i = 0; i < numUpdates; i++) {\n    const index = [];\n    let flattenIndex = 0;\n    for (let j = 0; j < sliceRank; j++) {\n      const dim = indicesData[i * sliceRank + j];\n      index.push(dim);\n      flattenIndex += dim * strides[j];\n    }\n    if (flattenIndex < 0 || flattenIndex >= outputSize / sliceSize) {\n      throw new Error(`Invalid indices: ${index} does not index into ${shape}`);\n    }\n    for (let k = 0; k < sliceSize; k++) {\n      if (sumDupeIndices) {\n        outBuf.values[flattenIndex * sliceSize + k] += updatesData[i * sliceSize + k];\n      } else {\n        outBuf.values[flattenIndex * sliceSize + k] = updates.rank === 0 ? updatesData[0] : updatesData[i * sliceSize + k];\n      }\n    }\n  }\n  return outBuf;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sigmoid.js\nvar sigmoidImpl = createSimpleUnaryImpl((xi) => 1 / (1 + Math.exp(-xi)));\nvar sigmoid2 = unaryKernelFunc(Sigmoid, (xi) => 1 / (1 + Math.exp(-xi)));\nvar sigmoidConfig = {\n  kernelName: Sigmoid,\n  backendName: \"cpu\",\n  kernelFunc: sigmoid2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Slice.js\nfunction sliceImpl(vals, begin, size, shape, dtype) {\n  const isContinous = slice_util_exports.isSliceContinous(shape, begin, size);\n  const length = util_exports.sizeFromShape(size);\n  const xStrides = util_exports.computeStrides(shape);\n  if (isContinous) {\n    const flatOffset = slice_util_exports.computeFlatOffset(begin, xStrides);\n    if (dtype === \"string\") {\n      return vals.slice(flatOffset, flatOffset + length);\n    }\n    return vals.subarray(flatOffset, flatOffset + length);\n  }\n  const decodedData = dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(vals) : vals;\n  const inBuf = buffer(shape, dtype, decodedData);\n  const outBuf = buffer(size, dtype);\n  for (let i = 0; i < outBuf.size; ++i) {\n    const outLoc = outBuf.indexToLoc(i);\n    const inLoc = outLoc.map((idx, j) => idx + begin[j]);\n    outBuf.set(inBuf.get(...inLoc), ...outLoc);\n  }\n  if (dtype === \"string\") {\n    return backend_util_exports.fromStringArrayToUint8(outBuf.values);\n  }\n  return outBuf.values;\n}\nfunction slice2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { begin, size } = attrs;\n  assertNotComplex(x, \"slice\");\n  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);\n  slice_util_exports.assertParamsValid(x, $begin, $size);\n  const vals = backend2.data.get(x.dataId).values;\n  const outVals = sliceImpl(vals, $begin, $size, x.shape, x.dtype);\n  return backend2.makeTensorInfo($size, x.dtype, outVals);\n}\nvar sliceConfig = {\n  kernelName: Slice,\n  backendName: \"cpu\",\n  kernelFunc: slice2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseFillEmptyRows_impl.js\nfunction sparseFillEmptyRowsImpl(indices, indicesShape, indicesDType, values, valuesDType, denseShape, defaultValue) {\n  const indicesCount = indicesShape[0];\n  const denseRows = denseShape[0];\n  const emptyRowIndicator = new Array(denseRows);\n  const reverseIndexMap = new Array(indicesCount);\n  const rank = indicesShape[1];\n  if (denseRows === 0) {\n    if (indicesCount !== 0) {\n      throw new Error(backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesCount));\n    }\n    const outputIndices = util_exports.getArrayFromDType(indicesDType, 0);\n    const outputValues = util_exports.getArrayFromDType(valuesDType, 0);\n    return [\n      outputIndices,\n      [0, rank],\n      outputValues,\n      emptyRowIndicator,\n      reverseIndexMap\n    ];\n  }\n  let rowsAreOrdered = true;\n  let lastIndicesRow = 0;\n  const csrOffset = new Array(denseRows).fill(0);\n  for (let i = 0; i < indicesCount; ++i) {\n    const row = indices[i * rank];\n    if (row < 0) {\n      throw new Error(backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(i, row));\n    }\n    if (row >= denseRows) {\n      throw new Error(backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(i, row, denseRows));\n    }\n    ++csrOffset[row];\n    rowsAreOrdered = rowsAreOrdered && row >= lastIndicesRow;\n    lastIndicesRow = row;\n  }\n  let allRowsFull = true;\n  for (let row = 0; row < denseRows; ++row) {\n    const rowEmpty = csrOffset[row] === 0;\n    emptyRowIndicator[row] = rowEmpty;\n    allRowsFull = allRowsFull && !rowEmpty;\n    csrOffset[row] = Math.max(csrOffset[row], 1);\n    if (row > 0) {\n      csrOffset[row] += csrOffset[row - 1];\n    }\n  }\n  if (allRowsFull && rowsAreOrdered) {\n    const outputIndices = indices;\n    const outputValues = values;\n    for (let i = 0; i < indicesCount; ++i) {\n      reverseIndexMap[i] = i;\n    }\n    return [\n      outputIndices,\n      [indicesCount, rank],\n      outputValues,\n      emptyRowIndicator,\n      reverseIndexMap\n    ];\n  } else {\n    const fullIndicesCount = csrOffset[denseRows - 1];\n    const outputIndices = util_exports.getArrayFromDType(indicesDType, fullIndicesCount * rank);\n    const outputValues = util_exports.getArrayFromDType(valuesDType, fullIndicesCount);\n    const filledCount = new Array(denseRows).fill(0);\n    for (let i = 0; i < indicesCount; ++i) {\n      const row = indices[i * rank];\n      const offset = filledCount[row];\n      const outputI = (row === 0 ? 0 : csrOffset[row - 1]) + offset;\n      filledCount[row]++;\n      for (let j = 0; j < rank; ++j) {\n        outputIndices[outputI * rank + j] = indices[i * rank + j];\n      }\n      outputValues[outputI] = values[i];\n      reverseIndexMap[i] = outputI;\n    }\n    for (let row = 0; row < denseRows; ++row) {\n      const rowCount = filledCount[row];\n      if (rowCount === 0) {\n        const startingIndex = row === 0 ? 0 : csrOffset[row - 1];\n        outputIndices[startingIndex * rank + 0] = row;\n        for (let col = 1; col < rank; ++col) {\n          outputIndices[startingIndex * rank + col] = 0;\n        }\n        outputValues[startingIndex] = defaultValue;\n      }\n    }\n    return [\n      outputIndices,\n      [fullIndicesCount, rank],\n      outputValues,\n      emptyRowIndicator,\n      reverseIndexMap\n    ];\n  }\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseReshape_impl.js\nfunction sparseReshapeImpl(inputIndices, inputIndicesShape, inputDType, inputShape, targetShape) {\n  const denseSize = util_exports.sizeFromShape(inputShape);\n  const nnz = inputIndicesShape[0];\n  const outputRank = targetShape.length;\n  const outputShape = [];\n  let product = 1;\n  let unknownIndex = -1;\n  for (let d = 0; d < outputRank; ++d) {\n    const size = targetShape[d];\n    if (size === -1) {\n      if (unknownIndex !== -1) {\n        throw new Error(backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(unknownIndex, d));\n      }\n      unknownIndex = d;\n      outputShape.push(1);\n    } else {\n      if (size < 0) {\n        throw new Error(backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(d, size));\n      }\n      product *= size;\n      outputShape.push(size);\n    }\n  }\n  if (unknownIndex !== -1) {\n    if (product <= 0) {\n      throw new Error(backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());\n    }\n    const missing = Math.trunc(denseSize / product);\n    if (product * missing !== denseSize) {\n      throw new Error(backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape));\n    }\n    outputShape[unknownIndex] = missing;\n  }\n  const outputSize = util_exports.sizeFromShape(outputShape);\n  if (outputSize !== denseSize) {\n    throw new Error(backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape));\n  }\n  const inputRank = inputShape.length;\n  const inputStrides = [];\n  if (inputRank > 0) {\n    inputStrides[inputRank - 1] = 1;\n    for (let d = inputRank - 2; d >= 0; --d) {\n      inputStrides[d] = inputStrides[d + 1] * inputShape[d + 1];\n    }\n  }\n  const outputStrides = [];\n  if (outputRank > 0) {\n    outputStrides[outputRank - 1] = 1;\n    for (let d = outputRank - 2; d >= 0; --d) {\n      outputStrides[d] = outputStrides[d + 1] * outputShape[d + 1];\n    }\n  }\n  const newIndices = util_exports.getArrayFromDType(inputDType, nnz * outputRank);\n  for (let i = 0; i < nnz; ++i) {\n    let id = 0;\n    for (let j = 0; j < inputRank; ++j) {\n      id += inputIndices[i * inputRank + j] * inputStrides[j];\n    }\n    for (let j = 0; j < outputRank; ++j) {\n      newIndices[i * outputRank + j] = Math.trunc(id / outputStrides[j]);\n      id %= outputStrides[j];\n    }\n  }\n  return [newIndices, [nnz, outputRank], outputShape];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentReduction_impl.js\nfunction sparseSegmentReductionImpl(input2, inputShape, inputDType, indices, segmentIds, isMean = false, defaultValue = 0) {\n  const numIndices = indices.length;\n  const inputFlat = [inputShape[0], input2.length / inputShape[0]];\n  const numCol = inputFlat[1];\n  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIds[numIndices - 1] + 1 : 0;\n  const outputRows = lastSegmentIdPlusOne;\n  if (outputRows < 0) {\n    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());\n  }\n  const outputShape = inputShape.slice();\n  outputShape[0] = outputRows;\n  const outputLength = outputShape.reduce((product, value) => product * value, 1);\n  const output = util_exports.getArrayFromDType(inputDType, outputLength);\n  if (numIndices === 0) {\n    if (outputRows > 0) {\n      output.fill(defaultValue);\n    }\n    return [output, outputShape];\n  }\n  if (outputRows <= 0) {\n    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());\n  }\n  let start = 0, end = 1;\n  let uninitializedIndex = 0;\n  let outIndex = segmentIds[start];\n  while (true) {\n    let nextIndex = 0;\n    if (end < numIndices) {\n      nextIndex = segmentIds[end];\n      if (outIndex === nextIndex) {\n        ++end;\n        continue;\n      }\n      if (outIndex >= nextIndex) {\n        throw new Error(backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage());\n      }\n    }\n    if (outIndex < 0 || outIndex >= outputRows) {\n      throw new Error(backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(outIndex, outputRows));\n    }\n    if (outIndex > uninitializedIndex) {\n      output.fill(defaultValue, uninitializedIndex * numCol, outIndex * numCol);\n    }\n    for (let i = start; i < end; ++i) {\n      const index = indices[i];\n      if (index < 0 || index >= inputFlat[0]) {\n        throw new Error(backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(i, indices[i], inputFlat[0]));\n      }\n      for (let j = 0; j < numCol; j++) {\n        output[outIndex * numCol + j] += input2[index * numCol + j];\n      }\n    }\n    if (isMean) {\n      for (let j = 0; j < numCol; j++) {\n        output[outIndex * numCol + j] /= end - start;\n      }\n    }\n    start = end;\n    ++end;\n    uninitializedIndex = outIndex + 1;\n    outIndex = nextIndex;\n    if (end > numIndices) {\n      break;\n    }\n  }\n  if (uninitializedIndex < outputRows) {\n    output.fill(defaultValue, uninitializedIndex * numCol, outputRows * numCol);\n  }\n  return [output, outputShape];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sqrt.js\nvar sqrtImpl = createSimpleUnaryImpl((xi) => Math.sqrt(xi));\nvar sqrt2 = unaryKernelFunc(Sqrt, (xi) => Math.sqrt(xi));\nvar sqrtConfig = {\n  kernelName: Sqrt,\n  backendName: \"cpu\",\n  kernelFunc: sqrt2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SquaredDifference.js\nvar squaredDifferenceImpl = createSimpleBinaryKernelImpl((a, b) => {\n  const diff = a - b;\n  return diff * diff;\n});\nvar squaredDifference2 = binaryKernelFunc(SquaredDifference, squaredDifferenceImpl);\nvar squaredDifferenceConfig = {\n  kernelName: SquaredDifference,\n  backendName: \"cpu\",\n  kernelFunc: squaredDifference2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StridedSlice_impl.js\nfunction stridedSliceImpl(outShape, xBuf, strides, begin) {\n  const outBuf = buffer(outShape, xBuf.dtype);\n  for (let i = 0; i < outBuf.size; i++) {\n    const loc = outBuf.indexToLoc(i);\n    const newLoc = new Array(loc.length);\n    for (let j = 0; j < newLoc.length; j++) {\n      newLoc[j] = loc[j] * strides[j] + begin[j];\n    }\n    outBuf.set(xBuf.get(...newLoc), ...loc);\n  }\n  return outBuf;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringNGrams_impl.js\nvar StringNGramsOp = class {\n  constructor(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {\n    this.separator = util_exports.encodeString(separator);\n    this.nGramWidths = nGramWidths;\n    this.leftPad = util_exports.encodeString(leftPad);\n    this.rightPad = util_exports.encodeString(rightPad2);\n    this.padWidth = padWidth;\n    this.preserveShort = preserveShortSequences;\n  }\n  getPadWidth(nGramWidth) {\n    return Math.min(this.padWidth < 0 ? nGramWidth - 1 : this.padWidth, nGramWidth - 1);\n  }\n  getNumNGrams(length, nGramWidth) {\n    const padWidth = this.getPadWidth(nGramWidth);\n    return Math.max(0, length + 2 * padWidth - nGramWidth + 1);\n  }\n  createNGrams(data, splitIndex, output, outputStartIndex, numNGrams, nGramWidth) {\n    for (let nGramIndex = 0; nGramIndex < numNGrams; ++nGramIndex) {\n      const padWidth = this.getPadWidth(nGramWidth);\n      const leftPadding = Math.max(0, padWidth - nGramIndex);\n      const rightPadding = Math.max(0, padWidth - (numNGrams - (nGramIndex + 1)));\n      const numTokens = nGramWidth - (leftPadding + rightPadding);\n      const dataStartIndex = splitIndex + (leftPadding > 0 ? 0 : nGramIndex - padWidth);\n      let nGramSize = 0;\n      nGramSize += leftPadding * this.leftPad.length;\n      for (let n = 0; n < numTokens; ++n) {\n        nGramSize += data[dataStartIndex + n].length;\n      }\n      nGramSize += rightPadding * this.rightPad.length;\n      const numSeparators = leftPadding + rightPadding + numTokens - 1;\n      nGramSize += numSeparators * this.separator.length;\n      output[outputStartIndex + nGramIndex] = new Uint8Array(nGramSize);\n      const nGram = output[outputStartIndex + nGramIndex];\n      let nextNGramIndex = 0;\n      const appendToNGram = (str) => str.forEach((value) => nGram[nextNGramIndex++] = value);\n      for (let n = 0; n < leftPadding; ++n) {\n        appendToNGram(this.leftPad);\n        appendToNGram(this.separator);\n      }\n      for (let n = 0; n < numTokens - 1; ++n) {\n        appendToNGram(data[dataStartIndex + n]);\n        appendToNGram(this.separator);\n      }\n      if (numTokens > 0) {\n        appendToNGram(data[dataStartIndex + numTokens - 1]);\n        for (let n = 0; n < rightPadding; ++n) {\n          appendToNGram(this.separator);\n          appendToNGram(this.rightPad);\n        }\n      } else {\n        for (let n = 0; n < rightPadding - 1; ++n) {\n          appendToNGram(this.rightPad);\n          appendToNGram(this.separator);\n        }\n        appendToNGram(this.rightPad);\n      }\n    }\n  }\n  compute(data, splits) {\n    const inputDataSize = data.length;\n    const splitsSize = splits.length;\n    if (splitsSize > 0) {\n      let prevSplit = splits[0];\n      if (prevSplit !== 0) {\n        throw new Error(`First split value must be 0, got ${prevSplit}`);\n      }\n      for (let i = 1; i < splitsSize; ++i) {\n        let validSplits = splits[i] >= prevSplit;\n        validSplits = validSplits && splits[i] <= inputDataSize;\n        if (!validSplits) {\n          throw new Error(`Invalid split value ${splits[i]}, must be in [${prevSplit}, ${inputDataSize}]`);\n        }\n        prevSplit = splits[i];\n      }\n      if (prevSplit !== inputDataSize) {\n        throw new Error(`Last split value must be data size. Expected ${inputDataSize}, got ${prevSplit}`);\n      }\n    }\n    const numBatchItems = splitsSize - 1;\n    const nGramsSplits = util_exports.getArrayFromDType(\"int32\", splitsSize);\n    if (inputDataSize === 0 || splitsSize === 0) {\n      const empty = new Array(inputDataSize);\n      for (let i = 0; i <= numBatchItems; ++i) {\n        nGramsSplits[i] = 0;\n      }\n      return [empty, nGramsSplits];\n    }\n    nGramsSplits[0] = 0;\n    for (let i = 1; i <= numBatchItems; ++i) {\n      const length = splits[i] - splits[i - 1];\n      let numNGrams = 0;\n      this.nGramWidths.forEach((nGramWidth) => {\n        numNGrams += this.getNumNGrams(length, nGramWidth);\n      });\n      if (this.preserveShort && length > 0 && numNGrams === 0) {\n        numNGrams = 1;\n      }\n      nGramsSplits[i] = nGramsSplits[i - 1] + numNGrams;\n    }\n    const nGrams = new Array(nGramsSplits[numBatchItems]);\n    for (let i = 0; i < numBatchItems; ++i) {\n      const splitIndex = splits[i];\n      let outputStartIdx = nGramsSplits[i];\n      this.nGramWidths.forEach((nGramWidth) => {\n        const length = splits[i + 1] - splits[i];\n        const numNGrams = this.getNumNGrams(length, nGramWidth);\n        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);\n        outputStartIdx += numNGrams;\n      });\n      if (this.preserveShort && outputStartIdx === nGramsSplits[i]) {\n        const dataLength = splits[i + 1] - splits[i];\n        if (dataLength === 0) {\n          continue;\n        }\n        const nGramWidth = dataLength + 2 * this.padWidth;\n        const numNGrams = 1;\n        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);\n      }\n    }\n    return [nGrams, nGramsSplits];\n  }\n};\nfunction stringNGramsImpl(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {\n  return new StringNGramsOp(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences).compute(data, dataSplits);\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringSplit_impl.js\nfunction split3(str, delimiters, skipEmpty, result) {\n  if (!str.length) {\n    return;\n  }\n  if (delimiters.length === 0) {\n    for (let i = 0; i < str.length; ++i) {\n      result.push(str.subarray(i, i + 1));\n    }\n    return;\n  }\n  if (delimiters.length === 1) {\n    const delimiter = delimiters[0];\n    let f = str.indexOf(delimiter);\n    while (f !== -1) {\n      const token = str.subarray(0, f);\n      if (!skipEmpty || token.length !== 0) {\n        result.push(token);\n      }\n      str = str.subarray(f + 1);\n      f = str.indexOf(delimiter);\n    }\n    if (!skipEmpty || str.length !== 0) {\n      result.push(str);\n    }\n    return;\n  }\n  let tokenStart = 0;\n  for (let i = 0; i < str.length + 1; i++) {\n    if (i === str.length || delimiters.indexOf(str[i]) !== -1) {\n      const token = str.subarray(tokenStart, i);\n      if (!skipEmpty || token.length !== 0) {\n        result.push(token);\n      }\n      tokenStart = i + 1;\n    }\n  }\n}\nfunction stringSplitImpl(input2, delimiter, skipEmpty) {\n  const batchSize = input2.length;\n  const tokens = [];\n  let outputSize = 0;\n  let maxNumEntries = 0;\n  const numIndices = new Array(batchSize);\n  for (let i = 0; i < batchSize; ++i) {\n    const prevTokensLength = tokens.length;\n    split3(input2[i], delimiter, skipEmpty, tokens);\n    const nEntries = tokens.length - prevTokensLength;\n    numIndices[i] = nEntries;\n    outputSize += nEntries;\n    maxNumEntries = Math.max(maxNumEntries, nEntries);\n  }\n  const indices = util_exports.getArrayFromDType(\"int32\", outputSize * 2);\n  const values = new Array(outputSize);\n  const shape = [batchSize, maxNumEntries];\n  let c = 0;\n  for (let i = 0; i < batchSize; ++i) {\n    for (let j = 0; j < numIndices[i]; ++j) {\n      indices[c * 2] = i;\n      indices[c * 2 + 1] = j;\n      values[c] = tokens[c];\n      ++c;\n    }\n  }\n  return [indices, values, shape];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringToHashBucketFast_impl.js\nfunction stringToHashBucketFastImpl(input2, numBuckets) {\n  const output = util_exports.getArrayFromDType(\"int32\", input2.length);\n  for (let i = 0; i < input2.length; ++i) {\n    output[i] = util_exports.fingerPrint64(input2[i]).modulo(numBuckets).getLowBitsUnsigned();\n  }\n  return output;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sub.js\nvar subImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue - bValue);\nvar subComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {\n  return { real: aReal - bReal, imag: aImag - bImag };\n});\nvar sub2 = binaryKernelFunc(Sub, subImpl, subComplexImpl);\nvar subConfig = {\n  kernelName: Sub,\n  backendName: \"cpu\",\n  kernelFunc: sub2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tile_impl.js\nfunction tileImpl(xBuf, reps) {\n  const newShape = new Array(xBuf.rank);\n  for (let i = 0; i < newShape.length; i++) {\n    newShape[i] = xBuf.shape[i] * reps[i];\n  }\n  const result = buffer(newShape, xBuf.dtype);\n  for (let i = 0; i < result.values.length; ++i) {\n    const newLoc = result.indexToLoc(i);\n    const originalLoc = new Array(xBuf.rank);\n    for (let j = 0; j < originalLoc.length; j++) {\n      originalLoc[j] = newLoc[j] % xBuf.shape[j];\n    }\n    const originalIndex = xBuf.locToIndex(originalLoc);\n    result.values[i] = xBuf.values[originalIndex];\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/TopK_impl.js\nvar comparePair = (a, b) => {\n  const valueDiff = b.value - a.value;\n  return valueDiff === 0 ? a.index - b.index : valueDiff;\n};\nfunction select(array2, k, left = 0, right = array2.length - 1) {\n  while (right > left) {\n    if (right - left > 600) {\n      const n = right - left + 1;\n      const i2 = k - left + 1;\n      const z = Math.log(n);\n      const s = 0.5 * Math.exp(2 * z / 3);\n      const sd = 0.5 * Math.sqrt(z * s * (n - s) / n) * Math.sign(i2 - n / 2);\n      const newLeft = Math.max(left, Math.floor(k - i2 * s / n + sd));\n      const newRight = Math.min(right, Math.floor(k + (n - i2) * s / n + sd));\n      select(array2, k, newLeft, newRight);\n    }\n    const t = array2[k];\n    let i = left;\n    let j = right;\n    util_exports.swap(array2, left, k);\n    if (comparePair(array2[right], t) > 0) {\n      util_exports.swap(array2, left, right);\n    }\n    while (i < j) {\n      util_exports.swap(array2, i, j);\n      i++;\n      j--;\n      while (comparePair(array2[i], t) < 0) {\n        i = i + 1;\n      }\n      while (comparePair(array2[j], t) > 0) {\n        j = j - 1;\n      }\n    }\n    if (comparePair(array2[left], t) === 0) {\n      util_exports.swap(array2, left, j);\n    } else {\n      j = j + 1;\n      util_exports.swap(array2, j, right);\n    }\n    if (j <= k) {\n      left = j + 1;\n    }\n    if (k <= j) {\n      right = j - 1;\n    }\n  }\n}\nfunction topKImpl(x, xShape, xDtype, k, sorted) {\n  const lastDim = xShape[xShape.length - 1];\n  const [batch, size] = [x.length / lastDim, lastDim];\n  const allTopKVals = util_exports.getTypedArrayFromDType(xDtype, batch * k);\n  const allTopKIndices = util_exports.getTypedArrayFromDType(\"int32\", batch * k);\n  for (let b = 0; b < batch; b++) {\n    const offset = b * size;\n    const vals = x.subarray(offset, offset + size);\n    let valAndInd = new Array(vals.length);\n    vals.forEach((value, index) => valAndInd[index] = { value, index });\n    if (k < valAndInd.length) {\n      select(valAndInd, k);\n      valAndInd = valAndInd.slice(0, k);\n    }\n    if (sorted) {\n      valAndInd.sort(comparePair);\n    }\n    const outOffset = b * k;\n    const topKVals = allTopKVals.subarray(outOffset, outOffset + k);\n    const topKIndices = allTopKIndices.subarray(outOffset, outOffset + k);\n    for (let i = 0; i < k; i++) {\n      topKVals[i] = valAndInd[i].value;\n      topKIndices[i] = valAndInd[i].index;\n    }\n  }\n  const outputShape = xShape.slice();\n  outputShape[outputShape.length - 1] = k;\n  return [\n    buffer(outputShape, xDtype, allTopKVals),\n    buffer(outputShape, \"int32\", allTopKIndices)\n  ];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unique_impl.js\nfunction uniqueImpl(values, axis, shape, dtype) {\n  const $axis = util_exports.parseAxisParam(axis, shape)[0];\n  const newShape = [1, shape[0], 1];\n  for (let i = 0; i < $axis; i++) {\n    newShape[0] *= shape[i];\n  }\n  newShape[1] = shape[$axis];\n  for (let i = $axis + 1; i < shape.length; i++) {\n    newShape[2] *= shape[i];\n  }\n  const uniqueElements = {};\n  const indices = new Int32Array(shape[$axis]);\n  const inputBuffer = new TensorBuffer(newShape, dtype, values);\n  const uniqueIndices = [];\n  const is1DTensor = newShape[0] === 1 && newShape[2] === 1;\n  for (let i = 0; i < shape[$axis]; i++) {\n    let element;\n    if (is1DTensor) {\n      element = values[i].toString();\n    } else {\n      const axisValues = [];\n      for (let m = 0; m < newShape[0]; m++) {\n        for (let n = 0; n < newShape[2]; n++) {\n          axisValues.push(inputBuffer.get(m, i, n));\n        }\n      }\n      element = axisValues.join(\",\");\n    }\n    if (uniqueElements[element] !== void 0) {\n      indices[i] = uniqueElements[element];\n    } else {\n      const uniqueIndex = Object.keys(uniqueElements).length;\n      uniqueElements[element] = uniqueIndex;\n      indices[i] = uniqueIndex;\n      uniqueIndices.push(i);\n    }\n  }\n  const outputTmpShape = newShape.slice();\n  outputTmpShape[1] = Object.keys(uniqueElements).length;\n  const outputBuffer = new TensorBuffer(outputTmpShape, dtype);\n  uniqueIndices.forEach((uniqueElementIndex, i) => {\n    for (let m = 0; m < newShape[0]; m++) {\n      for (let n = 0; n < newShape[2]; n++) {\n        outputBuffer.set(inputBuffer.get(m, uniqueElementIndex, n), m, i, n);\n      }\n    }\n  });\n  const outputShape = shape.slice();\n  outputShape[$axis] = outputTmpShape[1];\n  return {\n    outputValues: outputBuffer.values,\n    outputShape,\n    indices\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/base.js\nregisterBackend(\"cpu\", () => new MathBackendCPU(), 1);\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Elu.js\nvar elu4 = unaryKernelFunc(Elu, (xi) => xi >= 0 ? xi : Math.exp(xi) - 1);\nvar eluConfig = {\n  kernelName: Elu,\n  backendName: \"cpu\",\n  kernelFunc: elu4\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LeakyRelu.js\nfunction leakyRelu2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { alpha } = attrs;\n  assertNotComplex([x], \"leakyRelu\");\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const xVals = backend2.data.get(x.dataId).values;\n  const outVals = util_exports.getTypedArrayFromDType(\"float32\", xSize);\n  for (let i = 0; i < xVals.length; i++) {\n    outVals[i] = xVals[i] < 0 ? alpha * xVals[i] : xVals[i];\n  }\n  return backend2.makeTensorInfo(x.shape, \"float32\", outVals);\n}\nvar leakyReluConfig = {\n  kernelName: LeakyRelu,\n  backendName: \"cpu\",\n  kernelFunc: leakyRelu2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Prelu.js\nvar preluImpl = createSimpleBinaryKernelImpl((xValue, aValue) => xValue < 0 ? aValue * xValue : xValue);\nfunction prelu3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x, alpha } = inputs;\n  assertNotComplex([x, alpha], \"prelu\");\n  const aVals = backend2.data.get(x.dataId).values;\n  const bVals = backend2.data.get(alpha.dataId).values;\n  const [resultData, resultShape] = preluImpl(x.shape, alpha.shape, aVals, bVals, \"float32\");\n  return backend2.makeTensorInfo(resultShape, \"float32\", resultData);\n}\nvar preluConfig = {\n  kernelName: Prelu,\n  backendName: \"cpu\",\n  kernelFunc: prelu3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Relu.js\nvar relu2 = unaryKernelFunc(Relu, (xi) => Math.max(0, xi));\nvar reluConfig = {\n  kernelName: Relu,\n  backendName: \"cpu\",\n  kernelFunc: relu2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Relu6.js\nvar relu62 = unaryKernelFunc(Relu6, (xi) => Math.min(Math.max(0, xi), 6));\nvar relu6Config = {\n  kernelName: Relu6,\n  backendName: \"cpu\",\n  kernelFunc: relu62\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/fused_utils.js\nfunction applyActivation2(backend2, x, activation2, preluActivationWeights, leakyreluAlpha) {\n  if (activation2 === \"linear\") {\n    return identity2({ inputs: { x }, backend: backend2 });\n  } else if (activation2 === \"relu\") {\n    return relu2({ inputs: { x }, backend: backend2 });\n  } else if (activation2 === \"elu\") {\n    return elu4({ inputs: { x }, backend: backend2 });\n  } else if (activation2 === \"relu6\") {\n    return relu62({ inputs: { x }, backend: backend2 });\n  } else if (activation2 === \"prelu\") {\n    return prelu3({ inputs: { x, alpha: preluActivationWeights }, backend: backend2 });\n  } else if (activation2 === \"leakyrelu\") {\n    return leakyRelu2({ inputs: { x }, backend: backend2, attrs: { alpha: leakyreluAlpha } });\n  } else if (activation2 === \"sigmoid\") {\n    return sigmoid2({ inputs: { x }, backend: backend2 });\n  }\n  throw new Error(`Activation ${activation2} has not been implemented for the CPU backend.`);\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reshape.js\nfunction reshape3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { shape } = attrs;\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const $shape = util_exports.inferFromImplicitShape(shape, xSize);\n  const $xSize = util_exports.sizeFromShape($shape);\n  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);\n  backend2.incRef(x.dataId);\n  const xData = backend2.data.get(x.dataId);\n  if (xData.complexTensorInfos != null) {\n    const real4 = xData.complexTensorInfos.real;\n    const imag4 = xData.complexTensorInfos.imag;\n    real4.shape = $shape;\n    imag4.shape = $shape;\n  }\n  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };\n}\nvar reshapeConfig = {\n  kernelName: Reshape,\n  backendName: \"cpu\",\n  kernelFunc: reshape3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchMatMul.js\nfunction batchMatMul(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b } = inputs;\n  const { transposeA, transposeB } = attrs;\n  assertNotComplex([a, b], \"matMul\");\n  const aRank = a.shape.length;\n  const bRank = b.shape.length;\n  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];\n  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];\n  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];\n  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];\n  const outerDimsA = a.shape.slice(0, -2);\n  const outerDimsB = b.shape.slice(0, -2);\n  const batchDimA = util_exports.sizeFromShape(outerDimsA);\n  const batchDimB = util_exports.sizeFromShape(outerDimsB);\n  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));\n  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);\n  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);\n  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];\n  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];\n  const a3d = reshape3({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });\n  const b3d = reshape3({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });\n  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];\n  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];\n  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];\n  const batchDim = Math.max(batchDimA, batchDimB);\n  const a3dValues = backend2.data.get(a3d.dataId).values;\n  const b3dValues = backend2.data.get(b3d.dataId).values;\n  const a3dStrides = util_exports.computeStrides(a3d.shape);\n  const b3dStrides = util_exports.computeStrides(b3d.shape);\n  const [aBatch, aOuterStep, aInnerStep] = transposeA ? [a3dStrides[0], 1, a3dStrides[1]] : [a3dStrides[0], a3dStrides[1], 1];\n  const [bInnerStep, bOuterStep, bBatch] = transposeB ? [1, b3dStrides[1], b3dStrides[0]] : [b3dStrides[1], 1, b3dStrides[0]];\n  const size = leftDim * rightDim;\n  const result = buffer([batchDim, leftDim, rightDim], a3d.dtype);\n  const resVals = result.values;\n  const blockSize = backend2.blockSize;\n  for (let bi = 0; bi < batchDim; bi++) {\n    for (let i0 = 0; i0 < leftDim; i0 += blockSize) {\n      for (let j0 = 0; j0 < rightDim; j0 += blockSize) {\n        for (let k02 = 0; k02 < sharedDim; k02 += blockSize) {\n          const iBlock = Math.min(i0 + blockSize, leftDim);\n          const jBlock = Math.min(j0 + blockSize, rightDim);\n          const kBlock = Math.min(k02 + blockSize, sharedDim);\n          for (let i = i0; i < iBlock; i++) {\n            for (let j = j0; j < jBlock; j++) {\n              let sum6 = 0;\n              for (let k = k02; k < kBlock; k++) {\n                const batchOffsetA = Math.min(bi, batchDimA - 1) * aBatch;\n                const batchOffsetB = Math.min(bi, batchDimB - 1) * bBatch;\n                const aVal = a3dValues[batchOffsetA + i * aOuterStep + k * aInnerStep];\n                const bVal = b3dValues[k * bInnerStep + j * bOuterStep + batchOffsetB];\n                sum6 += aVal * bVal;\n              }\n              resVals[bi * size + (i * rightDim + j)] += sum6;\n            }\n          }\n        }\n      }\n    }\n  }\n  backend2.disposeIntermediateTensorInfo(a3d);\n  backend2.disposeIntermediateTensorInfo(b3d);\n  return backend2.makeTensorInfo(outShape, result.dtype, result.values);\n}\nvar batchMatMulConfig = {\n  kernelName: BatchMatMul,\n  backendName: \"cpu\",\n  kernelFunc: batchMatMul\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/_FusedMatMul.js\nfunction _fusedMatMul(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b, bias, preluActivationWeights } = inputs;\n  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;\n  let current;\n  let addRes;\n  let activationRes;\n  const intermediates = [];\n  const matMulRes = batchMatMul({ inputs: { a, b }, attrs: { transposeA, transposeB }, backend: backend2 });\n  current = matMulRes;\n  if (bias) {\n    addRes = add4({ inputs: { a: current, b: bias }, backend: backend2 });\n    intermediates.push(current);\n    current = addRes;\n  }\n  if (activation2) {\n    activationRes = applyActivation2(backend2, current, activation2, preluActivationWeights, leakyreluAlpha);\n    intermediates.push(current);\n    current = activationRes;\n  }\n  for (const i of intermediates) {\n    backend2.disposeIntermediateTensorInfo(i);\n  }\n  return current;\n}\nvar _fusedMatMulConfig = {\n  kernelName: _FusedMatMul,\n  backendName: \"cpu\",\n  kernelFunc: _fusedMatMul\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Acos.js\nvar acos2 = unaryKernelFunc(Acos, (xi) => Math.acos(xi));\nvar acosConfig = {\n  kernelName: Acos,\n  backendName: \"cpu\",\n  kernelFunc: acos2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Acosh.js\nvar acosh2 = unaryKernelFunc(Acosh, (xi) => Math.acosh(xi));\nvar acoshConfig = {\n  kernelName: Acosh,\n  backendName: \"cpu\",\n  kernelFunc: acosh2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AddN.js\nfunction addN2(args) {\n  const { inputs, backend: backend2 } = args;\n  const tensors = inputs;\n  assertNotComplex(inputs, \"addN\");\n  const vals = tensors.map((t) => backend2.data.get(t.dataId).values);\n  const outBuf = buffer(tensors[0].shape, tensors[0].dtype);\n  const outVals = outBuf.values;\n  for (let i = 0; i < tensors.length; i++) {\n    const currVals = vals[i];\n    for (let j = 0; j < outVals.length; j++) {\n      outVals[j] += currVals[j];\n    }\n  }\n  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);\n}\nvar addNConfig = {\n  kernelName: AddN,\n  backendName: \"cpu\",\n  kernelFunc: addN2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/All.js\nfunction all2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  assertNotComplex(x, \"all\");\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  if (permutedAxes != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"all\", axes, $x.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);\n  const aVals = backend2.data.get($x.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let all5 = aVals[offset];\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      all5 = all5 && value;\n    }\n    vals[i] = all5;\n  }\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo($x);\n  }\n  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);\n  if (keepDims) {\n    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });\n    backend2.disposeIntermediateTensorInfo(result);\n    return reshapedResult;\n  }\n  return result;\n}\nvar allConfig = {\n  kernelName: All,\n  backendName: \"cpu\",\n  kernelFunc: all2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Any.js\nfunction any2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  assertNotComplex(x, \"any\");\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  if (permutedAxes != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"any\", axes, $x.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);\n  const aVals = backend2.data.get($x.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let anyVal = aVals[offset];\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      anyVal = anyVal || value;\n    }\n    vals[i] = anyVal;\n  }\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo($x);\n  }\n  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);\n  if (keepDims) {\n    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });\n    backend2.disposeIntermediateTensorInfo(result);\n    return reshapedResult;\n  }\n  return result;\n}\nvar anyConfig = {\n  kernelName: Any,\n  backendName: \"cpu\",\n  kernelFunc: any2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ArgMax.js\nfunction argMax2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis } = attrs;\n  assertNotComplex(x, \"argMax\");\n  let axes = util_exports.parseAxisParam(axis, x.shape);\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  const intermediateTensorInfos = [];\n  if (permutedAxes != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    intermediateTensorInfos.push($x);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);\n  }\n  axes = [axes[0]];\n  backend_util_exports.assertAxesAreInnerMostDims(\"argMax\", axes, $x.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);\n  const outSize = util_exports.sizeFromShape(outShape);\n  const vals = util_exports.makeZerosTypedArray(outSize, \"int32\");\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const aVals = backend2.data.get($x.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let max6 = aVals[offset];\n    let maxIndex = 0;\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      if (value > max6) {\n        max6 = value;\n        maxIndex = j;\n      }\n    }\n    vals[i] = maxIndex;\n  }\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return backend2.makeTensorInfo(outShape, \"int32\", vals);\n}\nvar argMaxConfig = {\n  kernelName: ArgMax,\n  backendName: \"cpu\",\n  kernelFunc: argMax2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ArgMin.js\nfunction argMin2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis } = attrs;\n  assertNotComplex(x, \"argMin\");\n  let axes = util_exports.parseAxisParam(axis, x.shape);\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  const intermediateTensorInfos = [];\n  if (permutedAxes != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    intermediateTensorInfos.push($x);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);\n  }\n  axes = [axes[0]];\n  backend_util_exports.assertAxesAreInnerMostDims(\"argMin\", axes, $x.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);\n  const outSize = util_exports.sizeFromShape(outShape);\n  const vals = util_exports.makeZerosTypedArray(outSize, \"int32\");\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const aVals = backend2.data.get($x.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let min6 = aVals[offset];\n    let minIndex = 0;\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      if (value < min6) {\n        min6 = value;\n        minIndex = j;\n      }\n    }\n    vals[i] = minIndex;\n  }\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return backend2.makeTensorInfo(outShape, \"int32\", vals);\n}\nvar argMinConfig = {\n  kernelName: ArgMin,\n  backendName: \"cpu\",\n  kernelFunc: argMin2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Asin.js\nvar asin2 = unaryKernelFunc(Asin, (xi) => Math.asin(xi));\nvar asinConfig = {\n  kernelName: Asin,\n  backendName: \"cpu\",\n  kernelFunc: asin2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Asinh.js\nvar asinh2 = unaryKernelFunc(Asinh, (xi) => Math.asinh(xi));\nvar asinhConfig = {\n  kernelName: Asinh,\n  backendName: \"cpu\",\n  kernelFunc: asinh2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atan.js\nvar atan3 = unaryKernelFunc(Atan, (xi) => Math.atan(xi));\nvar atanConfig = {\n  kernelName: Atan,\n  backendName: \"cpu\",\n  kernelFunc: atan3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atan2.js\nvar atan2Impl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.atan2(aValue, bValue));\nvar atan22 = binaryKernelFunc(Atan2, atan2Impl);\nvar atan2Config = {\n  kernelName: Atan2,\n  backendName: \"cpu\",\n  kernelFunc: atan22\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atanh.js\nvar atanh2 = unaryKernelFunc(Atanh, (xi) => Math.atanh(xi));\nvar atanhConfig = {\n  kernelName: Atanh,\n  backendName: \"cpu\",\n  kernelFunc: atanh2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/pool_utils.js\nfunction pool2(xValues, xShape, dtype, strides, convInfo, poolType) {\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padLeft = convInfo.padInfo.left;\n  const initialValue = poolType === \"max\" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;\n  const output = buffer(convInfo.outShape, dtype);\n  const outputVals = output.values;\n  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3];\n  const outputRowStrides = convInfo.outShape[2] * convInfo.outShape[3];\n  const outputColStrides = convInfo.outShape[3];\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    const outputBatchOffset = b * outputBatchStrides;\n    const inputBatchOffset = b * strides[0];\n    for (let d = 0; d < convInfo.inChannels; ++d) {\n      for (let yR = 0; yR < convInfo.outHeight; ++yR) {\n        const xRCorner = yR * strideHeight - padTop;\n        const xRMin = Math.max(0, xRCorner);\n        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);\n        const outputRowOffset = outputBatchOffset + yR * outputRowStrides;\n        for (let yC = 0; yC < convInfo.outWidth; ++yC) {\n          const xCCorner = yC * strideWidth - padLeft;\n          const xCMin = Math.max(0, xCCorner);\n          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);\n          let minMaxValue = initialValue;\n          let avgValue = 0;\n          let count2 = 0;\n          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {\n            const xROffset = inputBatchOffset + xR * strides[1];\n            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {\n              const xCOffset = xROffset + xC * strides[2];\n              const pixel = xValues[xCOffset + d];\n              if (poolType === \"max\" && pixel > minMaxValue) {\n                minMaxValue = pixel;\n              } else if (poolType === \"avg\") {\n                avgValue += pixel;\n                count2++;\n              }\n            }\n            if (isNaN(minMaxValue)) {\n              break;\n            }\n          }\n          const outputOffset = outputRowOffset + yC * outputColStrides + d;\n          outputVals[outputOffset] = poolType === \"avg\" ? avgValue / count2 : minMaxValue;\n        }\n      }\n    }\n  }\n  return output;\n}\nfunction maxPoolPositions(xValues, xShape, dtype, convInfo, flattenPositions = false, includeBatchInIndex = false) {\n  const maxPositions = buffer(convInfo.outShape, \"int32\");\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padLeft = convInfo.padInfo.left;\n  const xBuf = buffer(xShape, dtype, xValues);\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    for (let d = 0; d < convInfo.inChannels; ++d) {\n      for (let yR = 0; yR < convInfo.outHeight; ++yR) {\n        const xRCorner = yR * strideHeight - padTop;\n        let xRMin = xRCorner;\n        while (xRMin < 0) {\n          xRMin += dilationHeight;\n        }\n        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);\n        for (let yC = 0; yC < convInfo.outWidth; ++yC) {\n          const xCCorner = yC * strideWidth - padLeft;\n          let xCMin = xCCorner;\n          while (xCMin < 0) {\n            xCMin += dilationWidth;\n          }\n          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);\n          let maxValue = Number.NEGATIVE_INFINITY;\n          let maxPosition = -1;\n          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {\n            const wR = xR - xRCorner;\n            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {\n              const wC = xC - xCCorner;\n              const pixel = xBuf.get(b, xR, xC, d);\n              if (pixel > maxValue) {\n                maxValue = pixel;\n                if (flattenPositions) {\n                  maxPosition = includeBatchInIndex ? ((b * convInfo.inHeight + xR) * convInfo.inWidth + xC) * convInfo.inChannels + d : (xR * convInfo.inWidth + xC) * convInfo.inChannels + d;\n                } else {\n                  maxPosition = wR * effectiveFilterWidth + wC;\n                }\n              }\n            }\n          }\n          maxPositions.set(maxPosition, b, yR, yC, d);\n        }\n      }\n    }\n  }\n  return maxPositions;\n}\nfunction pool3d2(xValues, xShape, dtype, strides, convInfo, poolType) {\n  const strideDepth = convInfo.strideDepth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationDepth = convInfo.dilationDepth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padFront = convInfo.padInfo.front;\n  const padTop = convInfo.padInfo.top;\n  const padLeft = convInfo.padInfo.left;\n  const initialValue = poolType === \"max\" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;\n  const output = buffer(convInfo.outShape, dtype);\n  const outputVals = output.values;\n  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];\n  const outputDepthStrides = convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];\n  const outputRowStrides = convInfo.outShape[3] * convInfo.outShape[4];\n  const outputColStrides = convInfo.outShape[4];\n  for (let batch = 0; batch < convInfo.batchSize; ++batch) {\n    const outputBatchOffset = batch * outputBatchStrides;\n    const inputBatchOffset = batch * strides[0];\n    for (let channel = 0; channel < convInfo.inChannels; ++channel) {\n      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {\n        const xDepthCorner = yDepth * strideDepth - padFront;\n        let xDepthMin = xDepthCorner;\n        while (xDepthMin < 0) {\n          xDepthMin += dilationDepth;\n        }\n        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);\n        const outputDepthOffset = outputBatchOffset + yDepth * outputDepthStrides;\n        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {\n          const xRowCorner = yRow * strideHeight - padTop;\n          let xRowMin = xRowCorner;\n          while (xRowMin < 0) {\n            xRowMin += dilationHeight;\n          }\n          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);\n          const outputRowOffset = outputDepthOffset + yRow * outputRowStrides;\n          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {\n            const xColCorner = yCol * strideWidth - padLeft;\n            let xColMin = xColCorner;\n            while (xColMin < 0) {\n              xColMin += dilationWidth;\n            }\n            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);\n            const outputColOffset = outputRowOffset + yCol * outputColStrides;\n            let minMaxValue = initialValue;\n            let avgValue = 0;\n            let count2 = 0;\n            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {\n              const xDepthOffset = inputBatchOffset + xDepth * strides[1];\n              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {\n                const xRowOffset = xDepthOffset + xRow * strides[2];\n                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {\n                  const xColOffset = xRowOffset + xCol * strides[3];\n                  const pixel = xValues[xColOffset + channel];\n                  if (poolType === \"max\" && pixel > minMaxValue) {\n                    minMaxValue = pixel;\n                  } else if (poolType === \"avg\") {\n                    avgValue += pixel;\n                    count2++;\n                  }\n                  if (isNaN(minMaxValue)) {\n                    break;\n                  }\n                }\n                if (isNaN(minMaxValue)) {\n                  break;\n                }\n              }\n              if (isNaN(minMaxValue)) {\n                break;\n              }\n            }\n            const outputOffset = outputColOffset + channel;\n            outputVals[outputOffset] = poolType === \"avg\" ? avgValue / count2 : minMaxValue;\n          }\n        }\n      }\n    }\n  }\n  return output;\n}\nfunction maxPool3dPositions(xBuf, convInfo) {\n  const maxPositions = buffer(convInfo.outShape, \"int32\");\n  const strideDepth = convInfo.strideDepth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationDepth = convInfo.dilationDepth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padFront = convInfo.padInfo.front;\n  const padTop = convInfo.padInfo.top;\n  const padLeft = convInfo.padInfo.left;\n  for (let batch = 0; batch < convInfo.batchSize; ++batch) {\n    for (let channel = 0; channel < convInfo.inChannels; ++channel) {\n      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {\n        const xDepthCorner = yDepth * strideDepth - padFront;\n        let xDepthMin = xDepthCorner;\n        while (xDepthMin < 0) {\n          xDepthMin += dilationDepth;\n        }\n        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);\n        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {\n          const xRowCorner = yRow * strideHeight - padTop;\n          let xRowMin = xRowCorner;\n          while (xRowMin < 0) {\n            xRowMin += dilationHeight;\n          }\n          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);\n          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {\n            const xColCorner = yCol * strideWidth - padLeft;\n            let xColMin = xColCorner;\n            while (xColMin < 0) {\n              xColMin += dilationWidth;\n            }\n            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);\n            let maxValue = Number.NEGATIVE_INFINITY;\n            let maxPosition = -1;\n            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {\n              const wDepth = xDepth - xDepthCorner;\n              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {\n                const wRow = xRow - xRowCorner;\n                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {\n                  const wCol = xCol - xColCorner;\n                  const pixel = xBuf.get(batch, xDepth, xRow, xCol, channel);\n                  if (pixel >= maxValue) {\n                    maxValue = pixel;\n                    maxPosition = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterHeight + wCol;\n                  }\n                }\n              }\n            }\n            maxPositions.set(maxPosition, batch, yDepth, yRow, yCol, channel);\n          }\n        }\n      }\n    }\n  }\n  return maxPositions;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool.js\nfunction avgPool2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  assertNotComplex(x, \"avgPool\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = 1;\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  let res;\n  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {\n    res = identity2({ inputs: { x }, backend: backend2 });\n  } else {\n    const xValues = backend2.data.get(x.dataId).values;\n    const strides2 = util_exports.computeStrides(x.shape);\n    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, \"avg\");\n    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);\n  }\n  return res;\n}\nvar avgPoolConfig = {\n  kernelName: AvgPool,\n  backendName: \"cpu\",\n  kernelFunc: avgPool2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool3D.js\nfunction avgPool3D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;\n  assertNotComplex(x, \"avgPool3d\");\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);\n  const xValues = backend2.data.get(x.dataId).values;\n  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, \"avg\");\n  return backend2.makeTensorInfo(outBuf.shape, \"float32\", outBuf.values);\n}\nvar avgPool3DConfig = {\n  kernelName: AvgPool3D,\n  backendName: \"cpu\",\n  kernelFunc: avgPool3D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool3DGrad.js\nfunction avgPool3DGrad(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  assertNotComplex([dy, input2], \"avgPool3DGrad\");\n  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const strideDepth = convInfo.strideDepth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const filterDepth = convInfo.filterDepth;\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const dilationDepth = convInfo.dilationDepth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;\n  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n  const dx = buffer(input2.shape, \"float32\");\n  const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);\n  const dyBuf = backend2.bufferSync(dy);\n  for (let batch = 0; batch < convInfo.batchSize; ++batch) {\n    for (let channel = 0; channel < convInfo.inChannels; ++channel) {\n      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {\n        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {\n          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {\n            const dyDepthCorner = dxDepth - padFront;\n            const dyRowCorner = dxRow - padTop;\n            const dyColCorner = dxCol - padLeft;\n            let dotProd = 0;\n            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {\n              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;\n              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {\n                continue;\n              }\n              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {\n                const dyRow = (dyRowCorner + wRow) / strideHeight;\n                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {\n                  continue;\n                }\n                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {\n                  const dyCol = (dyColCorner + wCol) / strideWidth;\n                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {\n                    continue;\n                  }\n                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);\n                  dotProd += pixel;\n                }\n              }\n            }\n            dx.set(dotProd * avgMultiplier, batch, dxDepth, dxRow, dxCol, channel);\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar avgPool3DGradConfig2 = {\n  kernelName: AvgPool3DGrad,\n  backendName: \"cpu\",\n  kernelFunc: avgPool3DGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPoolGrad.js\nfunction avgPoolGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const x = input2;\n  assertNotComplex([dy, input2], \"avgPoolGrad\");\n  const { filterSize, strides, pad: pad3 } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n  const dx = buffer(x.shape, \"float32\");\n  const avgMultiplier = 1 / (filterHeight * filterWidth);\n  const dyData = backend2.data.get(dy.dataId).values;\n  const dyBuf = buffer(dy.shape, \"float32\", dyData);\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    for (let d = 0; d < convInfo.inChannels; ++d) {\n      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {\n        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {\n          const dyRCorner = dxR - padTop;\n          const dyCCorner = dxC - padLeft;\n          let dotProd = 0;\n          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {\n            const dyR = (dyRCorner + wR) / strideHeight;\n            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {\n              continue;\n            }\n            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {\n              const dyC = (dyCCorner + wC) / strideWidth;\n              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {\n                continue;\n              }\n              const pixel = dyBuf.get(b, dyR, dyC, d);\n              dotProd += pixel;\n            }\n          }\n          dx.set(dotProd * avgMultiplier, b, dxR, dxC, d);\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar avgPoolGradConfig2 = {\n  kernelName: AvgPoolGrad,\n  backendName: \"cpu\",\n  kernelFunc: avgPoolGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchNorm.js\nfunction batchNorm2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, scale: scale2, offset, mean: mean4, variance } = inputs;\n  util_exports.assert(mean4.shape.length === variance.shape.length, () => \"Batch normalization gradient requires mean and variance to have equal ranks.\");\n  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => \"Batch normalization gradient requires mean and offset to have equal ranks.\");\n  util_exports.assert(scale2 == null || mean4.shape.length === scale2.shape.length, () => \"Batch normalization gradient requires mean and scale to have equal ranks.\");\n  assertNotComplex([x, mean4, variance, scale2, offset], \"batchNorm\");\n  let { varianceEpsilon } = attrs;\n  if (varianceEpsilon == null) {\n    varianceEpsilon = 1e-3;\n  }\n  const xVals = backend2.data.get(x.dataId).values;\n  const mVals = backend2.data.get(mean4.dataId).values;\n  const varVals = backend2.data.get(variance.dataId).values;\n  const sVals = scale2 ? backend2.data.get(scale2.dataId).values : new Float32Array([1]);\n  const offVals = offset ? backend2.data.get(offset.dataId).values : new Float32Array([0]);\n  const outVals = new Float32Array(xVals.length);\n  const offValsLength = offVals.length;\n  const sValsLength = sVals.length;\n  const varValsLength = varVals.length;\n  const mValsLength = mVals.length;\n  let offi = 0;\n  let mi = 0;\n  let si = 0;\n  let vi = 0;\n  for (let i = 0; i < xVals.length; ++i) {\n    outVals[i] = offVals[offi++] + (xVals[i] - mVals[mi++]) * sVals[si++] / Math.sqrt(varVals[vi++] + varianceEpsilon);\n    if (offi >= offValsLength) {\n      offi = 0;\n    }\n    if (mi >= mValsLength) {\n      mi = 0;\n    }\n    if (si >= sValsLength) {\n      si = 0;\n    }\n    if (vi >= varValsLength) {\n      vi = 0;\n    }\n  }\n  return backend2.makeTensorInfo(x.shape, x.dtype, outVals);\n}\nvar batchNormConfig = {\n  kernelName: FusedBatchNorm,\n  backendName: \"cpu\",\n  kernelFunc: batchNorm2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchToSpaceND.js\nfunction batchToSpaceND2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, crops } = attrs;\n  assertNotComplex([x], \"batchToSpaceND\");\n  const prod5 = blockShape.reduce((a, b) => a * b);\n  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);\n  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);\n  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);\n  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);\n  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);\n  const xReshaped = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });\n  const xTransposed = transpose2({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });\n  const xTransposedReshaped = reshape3({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });\n  const result = slice2({\n    inputs: { x: xTransposedReshaped },\n    backend: backend2,\n    attrs: { begin: sliceBeginCoords, size: sliceSize }\n  });\n  backend2.disposeIntermediateTensorInfo(xReshaped);\n  backend2.disposeIntermediateTensorInfo(xTransposed);\n  backend2.disposeIntermediateTensorInfo(xTransposedReshaped);\n  return result;\n}\nvar batchToSpaceNDConfig = {\n  kernelName: BatchToSpaceND,\n  backendName: \"cpu\",\n  kernelFunc: batchToSpaceND2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Bincount.js\nfunction bincount2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, weights } = inputs;\n  const { size } = attrs;\n  const xVals = backend2.data.get(x.dataId).values;\n  const weightsVals = backend2.data.get(weights.dataId).values;\n  const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);\n  return backend2.makeTensorInfo([size], weights.dtype, outVals);\n}\nvar bincountConfig = {\n  kernelName: Bincount,\n  backendName: \"cpu\",\n  kernelFunc: bincount2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BroadcastArgs.js\nfunction broadcastArgs2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { s0, s1 } = inputs;\n  const s0Vals = backend2.data.get(s0.dataId).values;\n  const s1Vals = backend2.data.get(s1.dataId).values;\n  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));\n  return backend2.makeTensorInfo([broadcastShape.length], \"int32\", Int32Array.from(broadcastShape));\n}\nvar broadcastArgsConfig = {\n  kernelName: BroadcastArgs,\n  backendName: \"cpu\",\n  kernelFunc: broadcastArgs2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ClipByValue.js\nvar clipByValue2 = unaryKernelFunc(ClipByValue, (xi, attrs) => {\n  const clipAttrs = attrs;\n  if (xi > clipAttrs.clipValueMax) {\n    return clipAttrs.clipValueMax;\n  }\n  return xi < clipAttrs.clipValueMin ? clipAttrs.clipValueMin : xi;\n});\nvar clipByValueConfig = {\n  kernelName: ClipByValue,\n  backendName: \"cpu\",\n  kernelFunc: clipByValue2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ComplexAbs.js\nvar complexAbs = (args) => {\n  const { x } = args.inputs;\n  const cpuBackend = args.backend;\n  const resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));\n  const complexVals = cpuBackend.data.get(x.dataId);\n  const real4 = complexVals.complexTensorInfos.real;\n  const imag4 = complexVals.complexTensorInfos.imag;\n  const realVals = cpuBackend.data.get(real4.dataId).values;\n  const imagVals = cpuBackend.data.get(imag4.dataId).values;\n  for (let i = 0; i < realVals.length; i++) {\n    const real5 = realVals[i];\n    const imag5 = imagVals[i];\n    resultValues[i] = Math.hypot(real5, imag5);\n  }\n  return cpuBackend.makeOutput(resultValues, x.shape, \"float32\");\n};\nvar complexAbsConfig = {\n  kernelName: ComplexAbs,\n  backendName: \"cpu\",\n  kernelFunc: complexAbs\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Imag.js\nfunction imag2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const imag4 = backend2.data.get(input2.dataId).complexTensorInfos.imag;\n  const imagVal = backend2.data.get(imag4.dataId).values;\n  return backend2.makeTensorInfo(imag4.shape, imag4.dtype, imagVal);\n}\nvar imagConfig = {\n  kernelName: Imag,\n  backendName: \"cpu\",\n  kernelFunc: imag2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Concat.js\nfunction concat2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { axis } = attrs;\n  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];\n  const shapes = inputs.map((t) => t.shape);\n  backend_util_exports.assertParamsConsistent(shapes, $axis);\n  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);\n  if (util_exports.sizeFromShape(outShape) === 0) {\n    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);\n  }\n  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);\n  if ($inputs.length === 1) {\n    return identity2({ inputs: { x: $inputs[0] }, backend: backend2 });\n  }\n  if ($inputs[0].dtype === \"complex64\") {\n    const reals = $inputs.map((t) => real2({ inputs: { input: t }, backend: backend2 }));\n    const imags = $inputs.map((t) => imag2({ inputs: { input: t }, backend: backend2 }));\n    const realConcated = concat2({ inputs: reals, backend: backend2, attrs: { axis: $axis } });\n    const imagConcated = concat2({ inputs: imags, backend: backend2, attrs: { axis: $axis } });\n    const result = complex2({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });\n    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));\n    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));\n    backend2.disposeIntermediateTensorInfo(realConcated);\n    backend2.disposeIntermediateTensorInfo(imagConcated);\n    return result;\n  }\n  const inputs2D = $inputs.map((t) => {\n    const innerSize = util_exports.sizeFromShape(t.shape.slice($axis));\n    const shape = [-1, innerSize];\n    return reshape3({ inputs: { x: t }, backend: backend2, attrs: { shape } });\n  });\n  const inputsValShapes = inputs2D.map((t) => {\n    return { vals: backend2.data.get(t.dataId).values, shape: t.shape };\n  });\n  outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);\n  const simplyConcat = inputs2D[0].shape[0] === 1;\n  const outVals = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);\n  const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), $axis);\n  const outInfo = backend2.makeTensorInfo(finalOutShape, inputs[0].dtype, outVals);\n  inputs2D.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return outInfo;\n}\nvar concatConfig = {\n  kernelName: Concat,\n  backendName: \"cpu\",\n  kernelFunc: concat2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2D.js\nfunction conv2D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;\n  assertNotComplex([x, filter], \"conv2d\");\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const padLeft = convInfo.padInfo.left;\n  const padTop = convInfo.padInfo.top;\n  const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n  const y = new TensorBuffer(convInfo.outShape, x.dtype);\n  const xStrides = util_exports.computeStrides(x.shape);\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  const xBatchStride = xStrides[0];\n  const xRowStride = isChannelsLast ? xStrides[1] : xStrides[2];\n  const xColStride = isChannelsLast ? xStrides[2] : 1;\n  const xChannelStride = isChannelsLast ? 1 : xStrides[1];\n  const yBatchStride = y.strides[0];\n  const yRowStride = isChannelsLast ? y.strides[1] : y.strides[2];\n  const yColStride = isChannelsLast ? y.strides[2] : 1;\n  const yChannelStride = isChannelsLast ? 1 : y.strides[1];\n  const xVals = backend2.data.get(x.dataId).values;\n  const wVals = backend2.data.get(filter.dataId).values;\n  const yVals = y.values;\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    const xOffset1 = b * xBatchStride;\n    const yOffset1 = b * yBatchStride;\n    for (let yR = 0; yR < convInfo.outHeight; ++yR) {\n      const yOffset2 = yOffset1 + yR * yRowStride;\n      const xRCorner = yR * convInfo.strideHeight - padTop;\n      for (let wR = 0; wR < filterHeight; ++wR) {\n        const xR = xRCorner + wR * dilationHeight;\n        if (xR < 0 || xR >= convInfo.inHeight) {\n          continue;\n        }\n        const wOffset1 = wR * filterStrides[0];\n        const xOffset2 = xOffset1 + xR * xRowStride;\n        for (let yC = 0; yC < convInfo.outWidth; ++yC) {\n          const yOffset3 = yOffset2 + yC * yColStride;\n          const xCCorner = yC * convInfo.strideWidth - padLeft;\n          for (let wC = 0; wC < filterWidth; ++wC) {\n            const xC = xCCorner + wC * dilationWidth;\n            if (xC < 0 || xC >= convInfo.inWidth) {\n              continue;\n            }\n            const wOffset2 = wOffset1 + wC * filterStrides[1];\n            const xOffset3 = xOffset2 + xC * xColStride;\n            let wOffset3 = wOffset2;\n            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {\n              const xVal = xVals[xOffset3 + d1 * xChannelStride];\n              for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {\n                yVals[yOffset3 + d2 * yChannelStride] += xVal * wVals[wOffset3 + d2];\n              }\n              wOffset3 += convInfo.outChannels;\n            }\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(y.shape, y.dtype, yVals);\n}\nvar conv2DConfig = {\n  kernelName: Conv2D,\n  backendName: \"cpu\",\n  kernelFunc: conv2D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2DBackpropFilter.js\nfunction conv2DBackpropFilter2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;\n  assertNotComplex([x, dy], \"conv2dBackpropFilter\");\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);\n  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;\n  const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n  const dW = new TensorBuffer(convInfo.filterShape, \"float32\");\n  const leftPad = convInfo.padInfo.left;\n  const topPad = convInfo.padInfo.top;\n  const xVals = backend2.data.get(x.dataId).values;\n  const dyVals = backend2.data.get(dy.dataId).values;\n  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);\n  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);\n  for (let wR = 0; wR < filterHeight; ++wR) {\n    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));\n    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);\n    for (let wC = 0; wC < filterWidth; ++wC) {\n      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));\n      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);\n      for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {\n        for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {\n          let dotProd = 0;\n          for (let b = 0; b < convInfo.batchSize; ++b) {\n            for (let yR = yRMin; yR < yRMax; ++yR) {\n              const xR = wR + yR * strideHeight - topPad;\n              for (let yC = yCMin; yC < yCMax; ++yC) {\n                const xC = wC + yC * strideWidth - leftPad;\n                if (isChannelsLast) {\n                  dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);\n                } else {\n                  dotProd += xBuf.get(b, d1, xR, xC) * dyBuf.get(b, d2, yR, yC);\n                }\n              }\n            }\n          }\n          dW.set(dotProd, wR, wC, d1, d2);\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);\n}\nvar conv2DBackpropFilterConfig = {\n  kernelName: Conv2DBackpropFilter,\n  backendName: \"cpu\",\n  kernelFunc: conv2DBackpropFilter2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2DBackpropInput.js\nfunction conv2DBackpropInput2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;\n  assertNotComplex([dy, filter], \"conv2dBackpropInput\");\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  let $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);\n  const dx = new TensorBuffer(convInfo.inShape, \"float32\");\n  const dxValues = dx.values;\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const fltValues = backend2.data.get(filter.dataId).values;\n  const [fltS0, fltS1, fltS2] = filterStrides;\n  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;\n  $dataFormat = convInfo.dataFormat;\n  const topPad = filterHeight - 1 - convInfo.padInfo.top;\n  const leftPad = filterWidth - 1 - convInfo.padInfo.left;\n  const isChannelsLast = $dataFormat === \"channelsLast\";\n  const xBatchStride = dx.strides[0];\n  const xRowStride = isChannelsLast ? dx.strides[1] : dx.strides[2];\n  const xColStride = isChannelsLast ? dx.strides[2] : 1;\n  const xChannelStride = isChannelsLast ? 1 : dx.strides[1];\n  const yBatchStride = dyStrides[0];\n  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];\n  const yColStride = isChannelsLast ? dyStrides[2] : 1;\n  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];\n  for (let b = 0; b < batchSize; ++b) {\n    for (let d1 = 0; d1 < inChannels; ++d1) {\n      for (let xR = 0; xR < inHeight; ++xR) {\n        const xRCorner = xR - topPad;\n        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));\n        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);\n        for (let xC = 0; xC < inWidth; ++xC) {\n          const xCCorner = xC - leftPad;\n          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));\n          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);\n          let dotProd = 0;\n          for (let yR = xRMin; yR < yRMax; ++yR) {\n            const wR = yR * strideHeight - xRCorner;\n            for (let yC = xCMin; yC < yCMax; ++yC) {\n              const wC = yC * strideWidth - xCCorner;\n              const dyOffset = yBatchStride * b + yRowStride * yR + yColStride * yC;\n              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;\n              for (let d2 = 0; d2 < outChannels; ++d2) {\n                const pixel = dyValues[dyOffset + yChannelStride * d2];\n                const weight = fltValues[fltOffset + d2];\n                dotProd += pixel * weight;\n              }\n            }\n          }\n          const dxOffset = xBatchStride * b + xRowStride * xR + xColStride * xC + xChannelStride * d1;\n          dxValues[dxOffset] = dotProd;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar conv2DBackpropInputConfig = {\n  kernelName: Conv2DBackpropInput,\n  backendName: \"cpu\",\n  kernelFunc: conv2DBackpropInput2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3D.js\nfunction conv3D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dilations } = attrs;\n  assertNotComplex([x, filter], \"conv3d\");\n  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);\n  const { filterDepth, filterHeight, filterWidth, dilationDepth, dilationHeight, dilationWidth, padInfo } = convInfo;\n  const padFront = padInfo.front;\n  const padLeft = padInfo.left;\n  const padTop = padInfo.top;\n  const y = new TensorBuffer(convInfo.outShape, x.dtype);\n  const xVals = backend2.data.get(x.dataId).values;\n  const wVals = backend2.data.get(filter.dataId).values;\n  const yVals = y.values;\n  const xStrides = util_exports.computeStrides(x.shape);\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    const xOffset1 = b * xStrides[0];\n    const yOffset1 = b * y.strides[0];\n    for (let yF = 0; yF < convInfo.outDepth; ++yF) {\n      const yOffset2 = yOffset1 + yF * y.strides[1];\n      const xFCorner = yF * convInfo.strideDepth - padFront;\n      for (let wF = 0; wF < filterDepth; ++wF) {\n        const xF = xFCorner + wF * dilationDepth;\n        if (xF < 0 || xF >= convInfo.inDepth) {\n          continue;\n        }\n        const wOffset1 = wF * filterStrides[0];\n        const xOffset2 = xOffset1 + xF * xStrides[1];\n        for (let yR = 0; yR < convInfo.outHeight; ++yR) {\n          const yOffset3 = yOffset2 + yR * y.strides[2];\n          const xRCorner = yR * convInfo.strideHeight - padTop;\n          for (let wR = 0; wR < filterHeight; ++wR) {\n            const xR = xRCorner + wR * dilationHeight;\n            if (xR < 0 || xR >= convInfo.inHeight) {\n              continue;\n            }\n            const wOffset2 = wOffset1 + wR * filterStrides[1];\n            const xOffset3 = xOffset2 + xR * xStrides[2];\n            for (let yC = 0; yC < convInfo.outWidth; ++yC) {\n              const yOffset4 = yOffset3 + yC * convInfo.outChannels;\n              const xCCorner = yC * convInfo.strideWidth - padLeft;\n              for (let wC = 0; wC < filterWidth; ++wC) {\n                const xC = xCCorner + wC * dilationWidth;\n                if (xC < 0 || xC >= convInfo.inWidth) {\n                  continue;\n                }\n                const wOffset3 = wOffset2 + wC * filterStrides[2];\n                const xOffset4 = xOffset3 + xC * convInfo.inChannels;\n                let wOffset4 = wOffset3;\n                for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {\n                  const xVal = xVals[xOffset4 + d1];\n                  for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {\n                    yVals[yOffset4 + d2] += xVal * wVals[wOffset4 + d2];\n                  }\n                  wOffset4 += convInfo.outChannels;\n                }\n              }\n            }\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);\n}\nvar conv3DConfig = {\n  kernelName: Conv3D,\n  backendName: \"cpu\",\n  kernelFunc: conv3D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3DBackpropFilterV2.js\nfunction conv3DBackpropFilterV2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, pad: pad3, filterShape } = attrs;\n  assertNotComplex([x, dy], \"conv3dBackpropFilterV2\");\n  const xStrides = util_exports.computeStrides(x.shape);\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);\n  const strideDepth = convInfo.strideDepth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const filterDepth = convInfo.filterDepth;\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const dw = new TensorBuffer(convInfo.filterShape, \"float32\");\n  const dwValues = dw.values;\n  const [dwS0, dwS1, dwS2, dwS3] = dw.strides;\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;\n  const xValues = backend2.data.get(x.dataId).values;\n  const [xS0, xS1, xS2, xS3] = xStrides;\n  const frontPad = convInfo.padInfo.front;\n  const leftPad = convInfo.padInfo.left;\n  const topPad = convInfo.padInfo.top;\n  for (let wF = 0; wF < filterDepth; ++wF) {\n    const yFMin = Math.max(0, Math.ceil((frontPad - wF) / strideDepth));\n    const yFMax = Math.min(convInfo.outDepth, (convInfo.inDepth + frontPad - wF) / strideDepth);\n    const wOffset1 = wF * dwS0;\n    for (let wR = 0; wR < filterHeight; ++wR) {\n      const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));\n      const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);\n      const wOffset2 = wR * dwS1 + wOffset1;\n      for (let wC = 0; wC < filterWidth; ++wC) {\n        const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));\n        const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);\n        const wOffset3 = wC * dwS2 + wOffset2;\n        for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {\n          const wOffset4 = d1 * dwS3 + wOffset3;\n          for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {\n            let dotProd = 0;\n            for (let b = 0; b < convInfo.batchSize; ++b) {\n              const xOffset1 = b * xS0;\n              const yOffset1 = b * dyS0;\n              for (let yF = yFMin; yF < yFMax; ++yF) {\n                const xF = wF + yF * strideDepth - frontPad;\n                const xOffset2 = xF * xS1 + xOffset1;\n                const yOffset2 = yF * dyS1 + yOffset1;\n                for (let yR = yRMin; yR < yRMax; ++yR) {\n                  const xR = wR + yR * strideHeight - topPad;\n                  const xOffset3 = xR * xS2 + xOffset2;\n                  const yOffset3 = yR * dyS2 + yOffset2;\n                  for (let yC = yCMin; yC < yCMax; ++yC) {\n                    const xC = wC + yC * strideWidth - leftPad;\n                    const xOffset4 = xC * xS3 + xOffset3;\n                    const yOffset4 = yC * dyS3 + yOffset3;\n                    dotProd += xValues[xOffset4 + d1] * dyValues[yOffset4 + d2];\n                  }\n                }\n              }\n            }\n            dwValues[wOffset4 + d2] = dotProd;\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dw.shape, dw.dtype, dw.values);\n}\nvar conv3DBackpropFilterV2Config = {\n  kernelName: Conv3DBackpropFilterV2,\n  backendName: \"cpu\",\n  kernelFunc: conv3DBackpropFilterV2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3DBackpropInputV2.js\nfunction conv3DBackpropInputV2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { pad: pad3, strides, inputShape } = attrs;\n  assertNotComplex([dy], \"conv3dBackpropInputV2\");\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);\n  const dx = new TensorBuffer(convInfo.inShape, \"float32\");\n  const dxValues = dx.values;\n  const [dxS0, dxS1, dxS2, dxS3] = dx.strides;\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;\n  const fltValues = backend2.data.get(filter.dataId).values;\n  const [fltS0, fltS1, fltS2, fltS3] = filterStrides;\n  const { batchSize, filterDepth, filterHeight, filterWidth, inChannels, inDepth, inHeight, inWidth, outChannels, outDepth, outHeight, outWidth, strideDepth, strideHeight, strideWidth } = convInfo;\n  const frontPad = filterDepth - 1 - convInfo.padInfo.front;\n  const topPad = filterHeight - 1 - convInfo.padInfo.top;\n  const leftPad = filterWidth - 1 - convInfo.padInfo.left;\n  for (let b = 0; b < batchSize; ++b) {\n    for (let d1 = 0; d1 < inChannels; ++d1) {\n      for (let xF = 0; xF < inDepth; ++xF) {\n        const xFCorner = xF - frontPad;\n        const xFMin = Math.max(0, Math.ceil(xFCorner / strideDepth));\n        const yFMax = Math.min(outDepth, (filterDepth + xFCorner) / strideDepth);\n        for (let xR = 0; xR < inHeight; ++xR) {\n          const xRCorner = xR - topPad;\n          const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));\n          const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);\n          for (let xC = 0; xC < inWidth; ++xC) {\n            const xCCorner = xC - leftPad;\n            const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));\n            const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);\n            let dotProd = 0;\n            for (let yF = xFMin; yF < yFMax; ++yF) {\n              const wF = yF * strideDepth - xFCorner;\n              for (let yR = xRMin; yR < yRMax; ++yR) {\n                const wR = yR * strideHeight - xRCorner;\n                for (let yC = xCMin; yC < yCMax; ++yC) {\n                  const wC = yC * strideWidth - xCCorner;\n                  const dyOffset = dyS0 * b + dyS1 * yF + dyS2 * yR + dyS3 * yC;\n                  const fltOffset = fltS0 * (filterDepth - 1 - wF) + fltS1 * (filterHeight - 1 - wR) + fltS2 * (filterWidth - 1 - wC) + fltS3 * d1;\n                  for (let d2 = 0; d2 < outChannels; ++d2) {\n                    const pixel = dyValues[dyOffset + d2];\n                    const weight = fltValues[fltOffset + d2];\n                    dotProd += pixel * weight;\n                  }\n                }\n              }\n            }\n            dxValues[dxS0 * b + dxS1 * xF + dxS2 * xR + dxS3 * xC + d1] = dotProd;\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar conv3DBackpropInputV2Config = {\n  kernelName: Conv3DBackpropInputV2,\n  backendName: \"cpu\",\n  kernelFunc: conv3DBackpropInputV2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cos.js\nvar cos2 = unaryKernelFunc(Cos, (xi) => Math.cos(xi));\nvar cosConfig = {\n  kernelName: Cos,\n  backendName: \"cpu\",\n  kernelFunc: cos2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cosh.js\nvar cosh2 = unaryKernelFunc(Cosh, (xi) => Math.cosh(xi));\nvar coshConfig = {\n  kernelName: Cosh,\n  backendName: \"cpu\",\n  kernelFunc: cosh2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/CropAndResize.js\nfunction cropAndResize2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { image: image2, boxes, boxInd } = inputs;\n  const { cropSize, method, extrapolationValue } = attrs;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  const numBoxes = boxes.shape[0];\n  const [cropHeight, cropWidth] = cropSize;\n  const output = buffer([numBoxes, cropHeight, cropWidth, numChannels], \"float32\");\n  const boxVals = backend2.data.get(boxes.dataId).values;\n  const boxIndVals = backend2.data.get(boxInd.dataId).values;\n  const imageVals = backend2.data.get(image2.dataId).values;\n  const inStride = util_exports.computeStrides(image2.shape);\n  const outStride = util_exports.computeStrides(output.shape);\n  for (let b = 0; b < numBoxes; b++) {\n    const startInd = b * 4;\n    const y1 = boxVals[startInd];\n    const x1 = boxVals[startInd + 1];\n    const y2 = boxVals[startInd + 2];\n    const x2 = boxVals[startInd + 3];\n    const bInd = boxIndVals[b];\n    if (bInd >= batch) {\n      continue;\n    }\n    const heightScale = cropHeight > 1 ? (y2 - y1) * (imageHeight - 1) / (cropHeight - 1) : 0;\n    const widthScale = cropWidth > 1 ? (x2 - x1) * (imageWidth - 1) / (cropWidth - 1) : 0;\n    for (let y = 0; y < cropHeight; y++) {\n      const yInd = cropHeight > 1 ? y1 * (imageHeight - 1) + y * heightScale : 0.5 * (y1 + y2) * (imageHeight - 1);\n      if (yInd < 0 || yInd > imageHeight - 1) {\n        for (let x = 0; x < cropWidth; x++) {\n          for (let c = 0; c < numChannels; c++) {\n            const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];\n            output.values[ind] = extrapolationValue;\n          }\n        }\n        continue;\n      }\n      if (method === \"bilinear\") {\n        const topInd = Math.floor(yInd);\n        const bottomInd = Math.ceil(yInd);\n        const yLerp = yInd - topInd;\n        for (let x = 0; x < cropWidth; x++) {\n          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);\n          if (xInd < 0 || xInd > imageWidth - 1) {\n            for (let c = 0; c < numChannels; c++) {\n              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];\n              output.values[ind] = extrapolationValue;\n            }\n            continue;\n          }\n          const leftInd = Math.floor(xInd);\n          const rightInd = Math.ceil(xInd);\n          const xLerp = xInd - leftInd;\n          for (let c = 0; c < numChannels; c++) {\n            let ind = c + leftInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];\n            const topLeft = imageVals[ind];\n            ind = c + rightInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];\n            const topRight = imageVals[ind];\n            ind = c + leftInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];\n            const bottomLeft = imageVals[ind];\n            ind = c + rightInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];\n            const bottomRight = imageVals[ind];\n            const top = topLeft + (topRight - topLeft) * xLerp;\n            const bottom = bottomLeft + (bottomRight - bottomLeft) * xLerp;\n            ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];\n            output.values[ind] = top + (bottom - top) * yLerp;\n          }\n        }\n      } else {\n        for (let x = 0; x < cropWidth; ++x) {\n          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);\n          if (xInd < 0 || xInd > imageWidth - 1) {\n            for (let c = 0; c < numChannels; c++) {\n              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];\n              output.values[ind] = extrapolationValue;\n            }\n            continue;\n          }\n          const closestX = Math.round(xInd);\n          const closestY = Math.round(yInd);\n          for (let c = 0; c < numChannels; c++) {\n            const inInd = c + closestX * inStride[2] + closestY * inStride[1] + bInd * inStride[0];\n            const outInd = c + x * outStride[2] + y * outStride[1] + b * outStride[0];\n            output.values[outInd] = imageVals[inInd];\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(output.shape, output.dtype, output.values);\n}\nvar cropAndResizeConfig = {\n  kernelName: CropAndResize,\n  backendName: \"cpu\",\n  kernelFunc: cropAndResize2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cumprod.js\nfunction cumprod2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  assertNotComplex(x, \"cumprod\");\n  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);\n  let $x = x;\n  if (permutation != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });\n  }\n  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];\n  if (permutedAxis !== $x.shape.length - 1) {\n    throw new Error(`backend.cumprod in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);\n  }\n  const resultDtype = upcastType($x.dtype, \"int32\");\n  const vals = util_exports.makeOnesTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);\n  const aVals = backend2.data.get($x.dataId).values;\n  const finalDim = $x.shape[$x.shape.length - 1];\n  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;\n  for (let i = 0; i < aVals.length; i += finalDim) {\n    for (let j = 0; j < finalDim; j++) {\n      const idx = indexAdjuster(i, j);\n      if (j === 0) {\n        vals[idx] = exclusive ? 1 : aVals[idx];\n      } else {\n        const prevIdx = indexAdjuster(i, j - 1);\n        vals[idx] = exclusive ? aVals[prevIdx] * vals[prevIdx] : aVals[idx] * vals[prevIdx];\n      }\n    }\n  }\n  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);\n  if (permutation != null) {\n    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);\n    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });\n    backend2.disposeIntermediateTensorInfo(result);\n    backend2.disposeIntermediateTensorInfo($x);\n    return reverseTransposedResult;\n  }\n  return result;\n}\nvar cumprodConfig = {\n  kernelName: Cumprod,\n  backendName: \"cpu\",\n  kernelFunc: cumprod2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cumsum.js\nfunction cumsum2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  assertNotComplex(x, \"cumsum\");\n  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);\n  let $x = x;\n  if (permutation != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });\n  }\n  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];\n  if (permutedAxis !== $x.shape.length - 1) {\n    throw new Error(`backend.cumsum in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);\n  }\n  const resultDtype = upcastType($x.dtype, \"int32\");\n  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);\n  const aVals = backend2.data.get($x.dataId).values;\n  const finalDim = $x.shape[$x.shape.length - 1];\n  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;\n  for (let i = 0; i < aVals.length; i += finalDim) {\n    for (let j = 0; j < finalDim; j++) {\n      const idx = indexAdjuster(i, j);\n      if (j === 0) {\n        vals[idx] = exclusive ? 0 : aVals[idx];\n      } else {\n        const prevIdx = indexAdjuster(i, j - 1);\n        vals[idx] = exclusive ? aVals[prevIdx] + vals[prevIdx] : aVals[idx] + vals[prevIdx];\n      }\n    }\n  }\n  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);\n  if (permutation != null) {\n    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);\n    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });\n    backend2.disposeIntermediateTensorInfo(result);\n    backend2.disposeIntermediateTensorInfo($x);\n    return reverseTransposedResult;\n  }\n  return result;\n}\nvar cumsumConfig = {\n  kernelName: Cumsum,\n  backendName: \"cpu\",\n  kernelFunc: cumsum2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DenseBincount.js\nfunction denseBincount2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, weights } = inputs;\n  const { size, binaryOutput } = attrs;\n  if (x.shape.length === 1) {\n    const xVals = backend2.data.get(x.dataId).values;\n    const weightsVals = backend2.data.get(weights.dataId).values;\n    const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);\n    return backend2.makeTensorInfo([size], weights.dtype, outVals);\n  } else if (x.shape.length === 2) {\n    const xBuf = backend2.bufferSync(x);\n    const weightsBuf = backend2.bufferSync(weights);\n    const outBuf = bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput);\n    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);\n  }\n  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);\n}\nvar denseBincountConfig = {\n  kernelName: DenseBincount,\n  backendName: \"cpu\",\n  kernelFunc: denseBincount2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthToSpace.js\nfunction depthToSpace2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockSize, dataFormat } = attrs;\n  util_exports.assert(dataFormat === \"NHWC\", () => `Only NHWC dataFormat supported on CPU for depthToSpace. Got ${dataFormat}`);\n  const batchSize = x.shape[0];\n  const inputHeight = x.shape[1];\n  const inputWidth = x.shape[2];\n  const inputDepth = x.shape[3];\n  const outputHeight = inputHeight * blockSize;\n  const outputWidth = inputWidth * blockSize;\n  const outputDepth = inputDepth / (blockSize * blockSize);\n  const xValues = backend2.data.get(x.dataId).values;\n  const result = new Float32Array(batchSize * outputHeight * outputWidth * outputDepth);\n  let outputIdx = 0;\n  for (let b = 0; b < batchSize; ++b) {\n    for (let h = 0; h < outputHeight; ++h) {\n      const inH = Math.floor(h / blockSize);\n      const offsetH = h % blockSize;\n      for (let w = 0; w < outputWidth; ++w) {\n        const inW = Math.floor(w / blockSize);\n        const offsetW = w % blockSize;\n        const offsetD = (offsetH * blockSize + offsetW) * outputDepth;\n        for (let d = 0; d < outputDepth; ++d) {\n          const inD = d + offsetD;\n          const inputIdx = inD + inputDepth * (inW + inputWidth * (inH + inputHeight * b));\n          result[outputIdx++] = xValues[inputIdx];\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo([batchSize, outputHeight, outputWidth, outputDepth], x.dtype, result);\n}\nvar depthToSpaceConfig = {\n  kernelName: DepthToSpace,\n  backendName: \"cpu\",\n  kernelFunc: depthToSpace2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNative.js\nfunction depthwiseConv2dNative(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;\n  assertNotComplex([x, filter], \"depthwiseConv2DNative\");\n  const xStrides = util_exports.computeStrides(x.shape);\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  let $dilations = dilations;\n  if ($dilations == null) {\n    $dilations = [1, 1];\n  }\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);\n  const { filterHeight, filterWidth, dilationHeight, dilationWidth, padInfo } = convInfo;\n  const padLeft = padInfo.left;\n  const padTop = padInfo.top;\n  const chMul = convInfo.outChannels / convInfo.inChannels;\n  const y = new TensorBuffer(convInfo.outShape, x.dtype);\n  const xVals = backend2.data.get(x.dataId).values;\n  const wVals = backend2.data.get(filter.dataId).values;\n  const yVals = y.values;\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    const xOffset1 = b * xStrides[0];\n    const yOffset1 = b * y.strides[0];\n    for (let yR = 0; yR < convInfo.outHeight; ++yR) {\n      const yOffset2 = yOffset1 + yR * y.strides[1];\n      const xRCorner = yR * convInfo.strideHeight - padTop;\n      for (let wR = 0; wR < filterHeight; ++wR) {\n        const xR = xRCorner + wR * dilationHeight;\n        if (xR < 0 || xR >= convInfo.inHeight) {\n          continue;\n        }\n        const wOffset1 = wR * filterStrides[0];\n        const xOffset2 = xOffset1 + xR * xStrides[1];\n        for (let yC = 0; yC < convInfo.outWidth; ++yC) {\n          const yOffset3 = yOffset2 + yC * y.strides[2];\n          const xCCorner = yC * convInfo.strideWidth - padLeft;\n          for (let wC = 0; wC < filterWidth; ++wC) {\n            const xC = xCCorner + wC * dilationWidth;\n            if (xC < 0 || xC >= convInfo.inWidth) {\n              continue;\n            }\n            const wOffset2 = wOffset1 + wC * filterStrides[1];\n            const xOffset3 = xOffset2 + xC * convInfo.inChannels;\n            let yOffset4 = yOffset3;\n            let wOffset3 = wOffset2;\n            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {\n              const xVal = xVals[xOffset3 + d1];\n              for (let q = 0; q < chMul; ++q) {\n                yVals[yOffset4 + q] += xVal * wVals[wOffset3 + q];\n              }\n              yOffset4 += chMul;\n              wOffset3 += chMul;\n            }\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);\n}\nvar depthwiseConv2dNativeConfig = {\n  kernelName: DepthwiseConv2dNative,\n  backendName: \"cpu\",\n  kernelFunc: depthwiseConv2dNative\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNativeBackpropFilter.js\nfunction depthwiseConv2dNativeBackpropFilter2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;\n  assertNotComplex([x, dy], \"depthwiseConv2dNativeBackpropFilter\");\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);\n  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;\n  const dW = new TensorBuffer(convInfo.filterShape, \"float32\");\n  const leftPad = convInfo.padInfo.left;\n  const topPad = convInfo.padInfo.top;\n  const chMul = convInfo.outChannels / convInfo.inChannels;\n  const xVals = backend2.data.get(x.dataId).values;\n  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);\n  const dyVals = backend2.data.get(dy.dataId).values;\n  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);\n  for (let wR = 0; wR < filterHeight; ++wR) {\n    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));\n    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);\n    for (let wC = 0; wC < filterWidth; ++wC) {\n      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));\n      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);\n      for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {\n        const d1 = Math.trunc(d2 / chMul);\n        const dm = d2 % chMul;\n        let dotProd = 0;\n        for (let b = 0; b < convInfo.batchSize; ++b) {\n          for (let yR = yRMin; yR < yRMax; ++yR) {\n            const xR = wR + yR * strideHeight - topPad;\n            for (let yC = yCMin; yC < yCMax; ++yC) {\n              const xC = wC + yC * strideWidth - leftPad;\n              dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);\n            }\n          }\n        }\n        dW.set(dotProd, wR, wC, d1, dm);\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);\n}\nvar depthwiseConv2dNativeBackpropFilterConfig = {\n  kernelName: DepthwiseConv2dNativeBackpropFilter,\n  backendName: \"cpu\",\n  kernelFunc: depthwiseConv2dNativeBackpropFilter2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNativeBackpropInput.js\nfunction depthwiseConv2dNativeBackpropInput2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;\n  assertNotComplex([dy, filter], \"depthwiseConv2DNativeBackpropInput\");\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  const filterStrides = util_exports.computeStrides(filter.shape);\n  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);\n  const dx = new TensorBuffer(convInfo.inShape, \"float32\");\n  const dxValues = dx.values;\n  const [dxS0, dxS1, dxS2] = dx.strides;\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const [dyS0, dyS1, dyS2] = dyStrides;\n  const fltValues = backend2.data.get(filter.dataId).values;\n  const [fltS0, fltS1, fltS2] = filterStrides;\n  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;\n  const topPad = filterHeight - 1 - convInfo.padInfo.top;\n  const leftPad = filterWidth - 1 - convInfo.padInfo.left;\n  const chMul = outChannels / inChannels;\n  for (let b = 0; b < batchSize; ++b) {\n    for (let d1 = 0; d1 < inChannels; ++d1) {\n      for (let xR = 0; xR < inHeight; ++xR) {\n        const xRCorner = xR - topPad;\n        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));\n        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);\n        for (let xC = 0; xC < inWidth; ++xC) {\n          const xCCorner = xC - leftPad;\n          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));\n          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);\n          let dotProd = 0;\n          for (let yR = xRMin; yR < yRMax; ++yR) {\n            const wR = yR * strideHeight - xRCorner;\n            for (let yC = xCMin; yC < yCMax; ++yC) {\n              const wC = yC * strideWidth - xCCorner;\n              const dyOffset = dyS0 * b + dyS1 * yR + dyS2 * yC;\n              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;\n              for (let dm = 0; dm < chMul; ++dm) {\n                const d2 = d1 * chMul + dm;\n                const pixel = dyValues[dyOffset + d2];\n                const weight = fltValues[fltOffset + dm];\n                dotProd += pixel * weight;\n              }\n            }\n          }\n          dxValues[dxS0 * b + dxS1 * xR + dxS2 * xC + d1] = dotProd;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar depthwiseConv2dNativeBackpropInputConfig = {\n  kernelName: DepthwiseConv2dNativeBackpropInput,\n  backendName: \"cpu\",\n  kernelFunc: depthwiseConv2dNativeBackpropInput2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Diag.js\nfunction diag2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const xVals = backend2.data.get(x.dataId).values;\n  const outBuf = buffer([xSize, xSize], x.dtype);\n  const vals = outBuf.values;\n  for (let i = 0; i < xVals.length; i++) {\n    vals[i * xSize + i] = xVals[i];\n  }\n  const outShape = [...x.shape, ...x.shape];\n  return backend2.makeTensorInfo(outShape, outBuf.dtype, outBuf.values);\n}\nvar diagConfig = {\n  kernelName: Diag,\n  backendName: \"cpu\",\n  kernelFunc: diag2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2D.js\nvar dilation2DConfig = {\n  kernelName: Dilation2D,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, backend: backend2, attrs }) => {\n    const { x, filter } = inputs;\n    const { strides, pad: pad3, dilations } = attrs;\n    const cpuBackend = backend2;\n    const xVals = cpuBackend.data.get(x.dataId).values;\n    const xRank = x.shape.length;\n    const filterVals = cpuBackend.data.get(filter.dataId).values;\n    const filterRank = filter.shape.length;\n    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, \"NHWC\", dilations);\n    const outSize = util_exports.sizeFromShape(outShape);\n    const outRank = outShape.length;\n    const outputVals = util_exports.getArrayFromDType(x.dtype, outSize);\n    for (let b = 0; b < batchSize; ++b) {\n      for (let hOut = 0; hOut < outHeight; ++hOut) {\n        const hBeg = hOut * strideHeight - padInfo.top;\n        for (let wOut = 0; wOut < outWidth; ++wOut) {\n          const wBeg = wOut * strideWidth - padInfo.left;\n          for (let d = 0; d < inChannels; ++d) {\n            let curVal = Number.MIN_SAFE_INTEGER;\n            for (let h = 0; h < filterHeight; ++h) {\n              const hIn = hBeg + h * dilationHeight;\n              if (hIn >= 0 && hIn < inHeight) {\n                for (let w = 0; w < filterWidth; ++w) {\n                  const wIn = wBeg + w * dilationWidth;\n                  if (wIn >= 0 && wIn < inWidth) {\n                    const xIndex = util_exports.locToIndex([b, hIn, wIn, d], xRank, util_exports.computeStrides(x.shape));\n                    const filterIndex = util_exports.locToIndex([h, w, d], filterRank, util_exports.computeStrides(filter.shape));\n                    const val = xVals[xIndex] + filterVals[filterIndex];\n                    if (val > curVal) {\n                      curVal = val;\n                    }\n                  }\n                }\n              }\n            }\n            const outputIndex = util_exports.locToIndex([b, hOut, wOut, d], outRank, util_exports.computeStrides(outShape));\n            outputVals[outputIndex] = curVal;\n          }\n        }\n      }\n    }\n    const dataId = cpuBackend.write(util_exports.toTypedArray(outputVals, x.dtype), outShape, x.dtype);\n    return { dataId, shape: outShape, dtype: x.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2DBackpropFilter.js\nvar dilation2DBackpropFilterConfig = {\n  kernelName: Dilation2DBackpropFilter,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, backend: backend2, attrs }) => {\n    const { x, filter, dy } = inputs;\n    const { strides, pad: pad3, dilations } = attrs;\n    const cpuBackend = backend2;\n    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);\n    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);\n    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, \"NHWC\", dilations);\n    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropFilter}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);\n    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);\n    const gradients = util_exports.makeZerosNestedTypedArray(filter.shape, filter.dtype);\n    for (let b = 0; b < batchSize; ++b) {\n      for (let hOut = 0; hOut < outHeight; ++hOut) {\n        const hBeg = hOut * strideHeight - padInfo.top;\n        for (let wOut = 0; wOut < outWidth; ++wOut) {\n          const wBeg = wOut * strideWidth - padInfo.left;\n          for (let d = 0; d < inChannels; ++d) {\n            let curVal = Number.MIN_SAFE_INTEGER;\n            let hMax = 0;\n            let wMax = 0;\n            for (let h = 0; h < filterHeight; ++h) {\n              const hIn = hBeg + h * dilationHeight;\n              if (hIn >= 0 && hIn < inHeight) {\n                for (let w = 0; w < filterWidth; ++w) {\n                  const wIn = wBeg + w * dilationWidth;\n                  if (wIn >= 0 && wIn < inWidth) {\n                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];\n                    if (val > curVal) {\n                      curVal = val;\n                      hMax = h;\n                      wMax = w;\n                    }\n                  }\n                }\n              }\n            }\n            gradients[hMax][wMax][d] += $dy[b][hOut][wOut][d];\n          }\n        }\n      }\n    }\n    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), filter.shape, filter.dtype);\n    return { dataId, shape: filter.shape, dtype: filter.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2DBackpropInput.js\nvar dilation2DBackpropInputConfig = {\n  kernelName: Dilation2DBackpropInput,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, backend: backend2, attrs }) => {\n    const { x, filter, dy } = inputs;\n    const { strides, pad: pad3, dilations } = attrs;\n    const cpuBackend = backend2;\n    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);\n    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);\n    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, \"NHWC\", dilations);\n    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropInput}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);\n    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);\n    const gradients = util_exports.makeZerosNestedTypedArray(x.shape, x.dtype);\n    for (let b = 0; b < batchSize; ++b) {\n      for (let hOut = 0; hOut < outHeight; ++hOut) {\n        const hBeg = hOut * strideHeight - padInfo.top;\n        for (let wOut = 0; wOut < outWidth; ++wOut) {\n          const wBeg = wOut * strideWidth - padInfo.left;\n          for (let d = 0; d < inChannels; ++d) {\n            let curVal = Number.MIN_SAFE_INTEGER;\n            let hInMax = hBeg < 0 ? 0 : hBeg;\n            let wInMax = wBeg < 0 ? 0 : wBeg;\n            for (let h = 0; h < filterHeight; ++h) {\n              const hIn = hBeg + h * dilationHeight;\n              if (hIn >= 0 && hIn < inHeight) {\n                for (let w = 0; w < filterWidth; ++w) {\n                  const wIn = wBeg + w * dilationWidth;\n                  if (wIn >= 0 && wIn < inWidth) {\n                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];\n                    if (val > curVal) {\n                      curVal = val;\n                      hInMax = hIn;\n                      wInMax = wIn;\n                    }\n                  }\n                }\n              }\n            }\n            gradients[b][hInMax][wInMax][d] += $dy[b][hOut][wOut][d];\n          }\n        }\n      }\n    }\n    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), x.shape, x.dtype);\n    return { dataId, shape: x.shape, dtype: x.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sum.js\nfunction sum3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  assertNotComplex(x, \"sum\");\n  let $x;\n  if (x.dtype === \"bool\") {\n    $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: \"int32\" } });\n  } else {\n    $x = identity2({ inputs: { x }, backend: backend2 });\n  }\n  const xRank = $x.shape.length;\n  const axes = util_exports.parseAxisParam(axis, $x.shape);\n  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);\n  let reductionAxes = axes;\n  let permutedX = $x;\n  if (permutation != null) {\n    permutedX = transpose2({ inputs: { x: $x }, backend: backend2, attrs: { perm: permutation } });\n    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"sum\", reductionAxes, permutedX.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, reductionAxes);\n  const resultDtype = backend_util_exports.upcastType(permutedX.dtype, \"int32\");\n  let result = zeros3(backend2, outShape, resultDtype);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const vals = backend2.data.get(result.dataId).values;\n  const aVals = backend2.data.get(permutedX.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let sum6 = 0;\n    for (let j = 0; j < reduceSize; ++j) {\n      sum6 += aVals[offset + j];\n    }\n    vals[i] = sum6;\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(result.shape, axes);\n    const oldResult = result;\n    result = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: newShape } });\n    backend2.disposeIntermediateTensorInfo(oldResult);\n  }\n  backend2.disposeIntermediateTensorInfo($x);\n  if (permutation != null) {\n    backend2.disposeIntermediateTensorInfo(permutedX);\n  }\n  return result;\n}\nvar sumConfig = {\n  kernelName: Sum,\n  backendName: \"cpu\",\n  kernelFunc: sum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Einsum.js\nfunction einsum2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { equation } = attrs;\n  const tensors = inputs;\n  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);\n  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);\n  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);\n  const nSteps = steps.length;\n  let out = null;\n  let numDimsRemaining = allDims.length;\n  const tensorsToDispose = [];\n  for (let i = 0; i < nSteps; ++i) {\n    for (const idTerm of steps[i]) {\n      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);\n      let x;\n      if (backend_util_exports.isIdentityPermutation(perm)) {\n        x = tensors[idTerm];\n      } else {\n        x = transpose2({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });\n        tensorsToDispose.push(x);\n      }\n      const targetShape = x.shape.slice();\n      for (let k = 0; k < dimsToExpand.length; ++k) {\n        targetShape.splice(dimsToExpand[k], 0, 1);\n      }\n      if (!util_exports.arraysEqual(x.shape, targetShape)) {\n        x = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });\n        tensorsToDispose.push(x);\n      }\n      if (out === null) {\n        out = x;\n      } else {\n        out = multiply2({ inputs: { a: x, b: out }, backend: backend2 });\n        tensorsToDispose.push(out);\n      }\n    }\n    if (i < nSteps - 1) {\n      if (path[i] >= 0) {\n        out = sum3({\n          inputs: { x: out },\n          backend: backend2,\n          attrs: {\n            axis: path[i] - (allDims.length - numDimsRemaining),\n            keepDims: false\n          }\n        });\n        tensorsToDispose.push(out);\n      }\n      numDimsRemaining--;\n    }\n  }\n  for (const tensorInfo of tensorsToDispose) {\n    if (tensorInfo === out) {\n      continue;\n    }\n    backend2.disposeIntermediateTensorInfo(tensorInfo);\n  }\n  return out;\n}\nvar einsumConfig = {\n  kernelName: Einsum,\n  backendName: \"cpu\",\n  kernelFunc: einsum2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/EluGrad.js\nfunction eluGrad(args) {\n  const { inputs, backend: backend2 } = args;\n  const { dy, y } = inputs;\n  assertNotComplex([dy, y], \"eluGrad\");\n  const resultValues = new Float32Array(util_exports.sizeFromShape(y.shape));\n  const values = backend2.data.get(y.dataId).values;\n  const dyValues = backend2.data.get(dy.dataId).values;\n  for (let i = 0; i < values.length; ++i) {\n    const v = values[i];\n    if (v >= 1) {\n      resultValues[i] = dyValues[i];\n    } else {\n      resultValues[i] = dyValues[i] * (v + 1);\n    }\n  }\n  return backend2.makeTensorInfo(y.shape, \"float32\", resultValues);\n}\nvar eluGradConfig2 = {\n  kernelName: EluGrad,\n  backendName: \"cpu\",\n  kernelFunc: eluGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Erf.js\nvar p = backend_util_exports.ERF_P;\nvar a1 = backend_util_exports.ERF_A1;\nvar a2 = backend_util_exports.ERF_A2;\nvar a3 = backend_util_exports.ERF_A3;\nvar a4 = backend_util_exports.ERF_A4;\nvar a5 = backend_util_exports.ERF_A5;\nvar erf2 = unaryKernelFunc(Erf, (xi) => {\n  const sign4 = Math.sign(xi);\n  const v = Math.abs(xi);\n  const t = 1 / (1 + p * v);\n  return sign4 * (1 - ((((a5 * t + a4) * t + a3) * t + a2) * t + a1) * t * Math.exp(-v * v));\n});\nvar erfConfig = {\n  kernelName: Erf,\n  backendName: \"cpu\",\n  kernelFunc: erf2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ExpandDims.js\nfunction expandDims3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { input: input2 } = inputs;\n  const { dim } = attrs;\n  const inputRank = input2.shape.length;\n  const newShape = input2.shape.slice();\n  let $dim = dim;\n  if (dim < 0) {\n    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);\n    $dim = inputRank + dim + 1;\n  }\n  newShape.splice($dim, 0, 1);\n  return reshape3({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });\n}\nvar expandDimsConfig = {\n  kernelName: ExpandDims,\n  backendName: \"cpu\",\n  kernelFunc: expandDims3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RealDiv.js\nvar realDivImpl = createSimpleBinaryKernelImpl((a, b) => a / b);\nvar div2 = binaryKernelFunc(RealDiv, realDivImpl);\nvar realDivConfig = {\n  kernelName: RealDiv,\n  backendName: \"cpu\",\n  kernelFunc: div2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/fft_utils.js\nfunction fftBatch(input2, inverse, cpuBackend) {\n  const inputShape = input2.shape;\n  const batch = inputShape[0];\n  const innerDim = inputShape[1];\n  const inputVals = cpuBackend.data.get(input2.dataId);\n  const real2D = inputVals.complexTensorInfos.real;\n  const imag2D = inputVals.complexTensorInfos.imag;\n  const resultShape = [batch, innerDim];\n  const resultSize = util_exports.sizeFromShape(resultShape);\n  const resultReal = util_exports.getTypedArrayFromDType(\"float32\", resultSize);\n  const resultImag = util_exports.getTypedArrayFromDType(\"float32\", resultSize);\n  for (let b = 0; b < batch; b++) {\n    const r = slice2({\n      inputs: { x: real2D },\n      backend: cpuBackend,\n      attrs: { begin: [b, 0], size: [1, innerDim] }\n    });\n    const i = slice2({\n      inputs: { x: imag2D },\n      backend: cpuBackend,\n      attrs: { begin: [b, 0], size: [1, innerDim] }\n    });\n    const input3 = complex2({ inputs: { real: r, imag: i }, backend: cpuBackend });\n    const { real: real4, imag: imag4 } = fftImpl(input3, inverse, cpuBackend);\n    const res = backend_util_exports.mergeRealAndImagArrays(real4, imag4);\n    for (let d = 0; d < innerDim; d++) {\n      const c = backend_util_exports.getComplexWithIndex(res, d);\n      resultReal[b * innerDim + d] = c.real;\n      resultImag[b * innerDim + d] = c.imag;\n    }\n    cpuBackend.disposeIntermediateTensorInfo(r);\n    cpuBackend.disposeIntermediateTensorInfo(i);\n    cpuBackend.disposeIntermediateTensorInfo(input3);\n  }\n  const $realInfo = cpuBackend.makeTensorInfo(resultShape, \"float32\", resultReal);\n  const $imagInfo = cpuBackend.makeTensorInfo(resultShape, \"float32\", resultImag);\n  const result = complex2({ inputs: { real: $realInfo, imag: $imagInfo }, backend: cpuBackend });\n  cpuBackend.disposeIntermediateTensorInfo($realInfo);\n  cpuBackend.disposeIntermediateTensorInfo($imagInfo);\n  return result;\n}\nfunction fftImpl(input2, inverse, cpuBackend) {\n  const inputSize = util_exports.sizeFromShape(input2.shape);\n  const inputVals = cpuBackend.data.get(input2.dataId);\n  const realVals = cpuBackend.data.get(inputVals.complexTensorInfos.real.dataId).values;\n  const imagVals = cpuBackend.data.get(inputVals.complexTensorInfos.imag.dataId).values;\n  if (isExponentOf2(inputSize)) {\n    const result = fftRadix2(realVals, imagVals, inputSize, inverse, cpuBackend);\n    const resultShape = [input2.shape[0], input2.shape[1]];\n    if (inverse) {\n      const realInfo = cpuBackend.makeTensorInfo(resultShape, \"float32\", result.real);\n      const imagInfo = cpuBackend.makeTensorInfo(resultShape, \"float32\", result.imag);\n      const sizeInfo = cpuBackend.makeTensorInfo([], \"float32\", util_exports.createScalarValue(inputSize, \"float32\"));\n      const sizeInfoCopy = identity2({ inputs: { x: sizeInfo }, backend: cpuBackend });\n      const divRealInfo = realDivConfig.kernelFunc({ inputs: { a: realInfo, b: sizeInfo }, backend: cpuBackend });\n      const divImagInfo = realDivConfig.kernelFunc({ inputs: { a: imagInfo, b: sizeInfoCopy }, backend: cpuBackend });\n      const divRealVals = cpuBackend.data.get(divRealInfo.dataId).values;\n      const divImagVals = cpuBackend.data.get(divImagInfo.dataId).values;\n      cpuBackend.disposeIntermediateTensorInfo(realInfo);\n      cpuBackend.disposeIntermediateTensorInfo(imagInfo);\n      cpuBackend.disposeIntermediateTensorInfo(sizeInfo);\n      cpuBackend.disposeIntermediateTensorInfo(sizeInfoCopy);\n      cpuBackend.disposeIntermediateTensorInfo(divRealInfo);\n      cpuBackend.disposeIntermediateTensorInfo(divImagInfo);\n      return { real: divRealVals, imag: divImagVals };\n    }\n    return result;\n  } else {\n    const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);\n    const rawOutput = fourierTransformByMatmul(data, inputSize, inverse);\n    return backend_util_exports.splitRealAndImagArrays(rawOutput);\n  }\n}\nfunction isExponentOf2(size) {\n  return (size & size - 1) === 0;\n}\nfunction fftRadix2(realVals, imagVals, size, inverse, cpuBackend) {\n  if (size === 1) {\n    return { real: realVals, imag: imagVals };\n  }\n  const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);\n  const half = size / 2;\n  const evenComplex = backend_util_exports.complexWithEvenIndex(data);\n  const evenRealVals = evenComplex.real;\n  const evenImagVals = evenComplex.imag;\n  const evenShape = [evenRealVals.length];\n  const evenRealInfo = cpuBackend.makeTensorInfo(evenShape, \"float32\", evenRealVals);\n  const evenImagInfo = cpuBackend.makeTensorInfo(evenShape, \"float32\", evenImagVals);\n  const evenTensorInfo = complex2({ inputs: { real: evenRealInfo, imag: evenImagInfo }, backend: cpuBackend });\n  const oddComplex = backend_util_exports.complexWithOddIndex(data);\n  const oddRealVals = oddComplex.real;\n  const oddImagVals = oddComplex.imag;\n  const oddShape = [oddRealVals.length];\n  const oddRealInfo = cpuBackend.makeTensorInfo(oddShape, \"float32\", oddRealVals);\n  const oddImagInfo = cpuBackend.makeTensorInfo(oddShape, \"float32\", oddImagVals);\n  const oddTensorInfo = complex2({ inputs: { real: oddRealInfo, imag: oddImagInfo }, backend: cpuBackend });\n  const $evenComplex = fftRadix2(evenRealVals, evenImagVals, half, inverse, cpuBackend);\n  const $evenRealVals = $evenComplex.real;\n  const $evenImagVals = $evenComplex.imag;\n  const $evenShape = [$evenRealVals.length];\n  const $evenRealInfo = cpuBackend.makeTensorInfo($evenShape, \"float32\", $evenRealVals);\n  const $evenImagInfo = cpuBackend.makeTensorInfo($evenShape, \"float32\", $evenImagVals);\n  const $evenTensorInfo = complex2({\n    inputs: { real: $evenRealInfo, imag: $evenImagInfo },\n    backend: cpuBackend\n  });\n  const $oddComplex = fftRadix2(oddRealVals, oddImagVals, half, inverse, cpuBackend);\n  const $oddRealVals = $oddComplex.real;\n  const $oddImagVals = $oddComplex.imag;\n  const $oddShape = [$oddRealVals.length];\n  const $oddRealInfo = cpuBackend.makeTensorInfo($oddShape, \"float32\", $oddRealVals);\n  const $oddImagInfo = cpuBackend.makeTensorInfo($oddShape, \"float32\", $oddImagVals);\n  const $oddTensorInfo = complex2({ inputs: { real: $oddRealInfo, imag: $oddImagInfo }, backend: cpuBackend });\n  const e = backend_util_exports.exponents(size, inverse);\n  const eShape = [e.real.length];\n  const eRealInfo = cpuBackend.makeTensorInfo(eShape, \"float32\", e.real);\n  const eImagInfo = cpuBackend.makeTensorInfo(eShape, \"float32\", e.imag);\n  const complexInfo = complex2({ inputs: { real: eRealInfo, imag: eImagInfo }, backend: cpuBackend });\n  const exponentInfo = multiply2({ inputs: { a: complexInfo, b: $oddTensorInfo }, backend: cpuBackend });\n  const addPart = add4({\n    inputs: { a: $evenTensorInfo, b: exponentInfo },\n    backend: cpuBackend\n  });\n  const subPart = sub2({\n    inputs: { a: $evenTensorInfo, b: exponentInfo },\n    backend: cpuBackend\n  });\n  const addPartReal = real2({ inputs: { input: addPart }, backend: cpuBackend });\n  const subPartReal = real2({ inputs: { input: subPart }, backend: cpuBackend });\n  const addPartImag = imag2({ inputs: { input: addPart }, backend: cpuBackend });\n  const subPartImag = imag2({ inputs: { input: subPart }, backend: cpuBackend });\n  const $real = concat2({\n    inputs: [addPartReal, subPartReal],\n    backend: cpuBackend,\n    attrs: { axis: 0 }\n  });\n  const $imag = concat2({\n    inputs: [addPartImag, subPartImag],\n    backend: cpuBackend,\n    attrs: { axis: 0 }\n  });\n  const $realVals = cpuBackend.data.get($real.dataId).values;\n  const $imagVals = cpuBackend.data.get($imag.dataId).values;\n  cpuBackend.disposeIntermediateTensorInfo(evenRealInfo);\n  cpuBackend.disposeIntermediateTensorInfo(evenImagInfo);\n  cpuBackend.disposeIntermediateTensorInfo(evenTensorInfo);\n  cpuBackend.disposeIntermediateTensorInfo(oddRealInfo);\n  cpuBackend.disposeIntermediateTensorInfo(oddImagInfo);\n  cpuBackend.disposeIntermediateTensorInfo(oddTensorInfo);\n  cpuBackend.disposeIntermediateTensorInfo($evenRealInfo);\n  cpuBackend.disposeIntermediateTensorInfo($evenImagInfo);\n  cpuBackend.disposeIntermediateTensorInfo($evenTensorInfo);\n  cpuBackend.disposeIntermediateTensorInfo($oddRealInfo);\n  cpuBackend.disposeIntermediateTensorInfo($oddImagInfo);\n  cpuBackend.disposeIntermediateTensorInfo($oddTensorInfo);\n  cpuBackend.disposeIntermediateTensorInfo(eRealInfo);\n  cpuBackend.disposeIntermediateTensorInfo(eImagInfo);\n  cpuBackend.disposeIntermediateTensorInfo(complexInfo);\n  cpuBackend.disposeIntermediateTensorInfo(exponentInfo);\n  cpuBackend.disposeIntermediateTensorInfo(addPart);\n  cpuBackend.disposeIntermediateTensorInfo(subPart);\n  cpuBackend.disposeIntermediateTensorInfo(addPartReal);\n  cpuBackend.disposeIntermediateTensorInfo(addPartImag);\n  cpuBackend.disposeIntermediateTensorInfo(subPartReal);\n  cpuBackend.disposeIntermediateTensorInfo(subPartImag);\n  cpuBackend.disposeIntermediateTensorInfo($real);\n  cpuBackend.disposeIntermediateTensorInfo($imag);\n  return { real: $realVals, imag: $imagVals };\n}\nfunction fourierTransformByMatmul(data, size, inverse) {\n  const ret = new Float32Array(size * 2);\n  for (let r = 0; r < size; r++) {\n    let real4 = 0;\n    let imag4 = 0;\n    for (let c = 0; c < size; c++) {\n      const e = backend_util_exports.exponent(r * c, size, inverse);\n      const term = backend_util_exports.getComplexWithIndex(data, c);\n      real4 += term.real * e.real - term.imag * e.imag;\n      imag4 += term.real * e.imag + term.imag * e.real;\n    }\n    if (inverse) {\n      real4 /= size;\n      imag4 /= size;\n    }\n    backend_util_exports.assignToTypedArray(ret, real4, imag4, r);\n  }\n  return ret;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FFT.js\nfunction fft2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const inputSize = util_exports.sizeFromShape(input2.shape);\n  const innerDimensionSize = input2.shape[input2.shape.length - 1];\n  const batch = inputSize / innerDimensionSize;\n  const input2D = reshape3({\n    inputs: { x: input2 },\n    backend: backend2,\n    attrs: { shape: [batch, innerDimensionSize] }\n  });\n  const result = fftBatch(input2D, false, backend2);\n  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });\n  backend2.disposeIntermediateTensorInfo(input2D);\n  backend2.disposeIntermediateTensorInfo(result);\n  return resultReshaped;\n}\nvar fftConfig = {\n  kernelName: FFT,\n  backendName: \"cpu\",\n  kernelFunc: fft2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Fill.js\nfunction fill2(args) {\n  const { backend: backend2, attrs } = args;\n  const { shape, value, dtype } = attrs;\n  const $dtype = dtype || util_exports.inferDtype(value);\n  const values = util_exports.getArrayFromDType($dtype, util_exports.sizeFromShape(shape));\n  fillValues(values, value, $dtype);\n  return backend2.makeTensorInfo(shape, $dtype, values);\n}\nvar fillConfig = {\n  kernelName: Fill,\n  backendName: \"cpu\",\n  kernelFunc: fill2\n};\nfunction fillValues(values, value, dtype) {\n  if (dtype === \"string\") {\n    values.fill(value);\n  } else {\n    values.fill(value);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FlipLeftRight.js\nvar flipLeftRightConfig = {\n  kernelName: FlipLeftRight,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { image: image2 } = inputs;\n    const cpuBackend = backend2;\n    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));\n    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n    const imageVals = cpuBackend.data.get(image2.dataId).values;\n    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {\n      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;\n      for (let row = 0; row < imageHeight; row++) {\n        const rowOffset = row * (imageWidth * numChannels);\n        for (let col = 0; col < imageWidth; col++) {\n          const colOffset = col * numChannels;\n          for (let channel = 0; channel < numChannels; channel++) {\n            const coordX = Math.round(imageWidth - col - 1);\n            const outIdx = batchOffset + rowOffset + colOffset + channel;\n            let outputValue = imageVals[outIdx];\n            if (coordX >= 0 && coordX < imageWidth) {\n              const rotatedColOffset = coordX * numChannels;\n              const imageIdx = batchOffset + rowOffset + rotatedColOffset + channel;\n              outputValue = imageVals[imageIdx];\n            }\n            output[outIdx] = outputValue;\n          }\n        }\n      }\n    }\n    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);\n    return { dataId, shape: image2.shape, dtype: image2.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FloorDiv.js\nvar floorDivImpl = createSimpleBinaryKernelImpl((a, b) => Math.floor(a / b));\nvar floorDiv2 = binaryKernelFunc(FloorDiv, floorDivImpl, null, \"int32\");\nvar floorDivConfig = {\n  kernelName: FloorDiv,\n  backendName: \"cpu\",\n  kernelFunc: floorDiv2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FusedConv2D.js\nfunction fusedConv2D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  let result = conv2D({\n    inputs: { x, filter },\n    backend: backend2,\n    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }\n  });\n  if (bias) {\n    const resultOld = result;\n    if (dataFormat === \"NCHW\" && bias.shape.length === 1 && bias.shape[0] !== 1) {\n      const reshapedBias = reshape3({ inputs: { x: bias }, backend: backend2, attrs: { shape: [bias.shape[0], 1, 1] } });\n      result = add4({ inputs: { a: result, b: reshapedBias }, backend: backend2 });\n      backend2.disposeIntermediateTensorInfo(reshapedBias);\n    } else {\n      result = add4({ inputs: { a: result, b: bias }, backend: backend2 });\n    }\n    backend2.disposeIntermediateTensorInfo(resultOld);\n  }\n  if (activation2) {\n    const resultOld = result;\n    if (dataFormat === \"NCHW\" && activation2 === \"prelu\" && preluActivationWeights.shape.length === 1 && preluActivationWeights.shape[0] !== 1) {\n      const reshapedAlpha = reshape3({\n        inputs: { x: preluActivationWeights },\n        backend: backend2,\n        attrs: { shape: [preluActivationWeights.shape[0], 1, 1] }\n      });\n      result = applyActivation2(backend2, result, activation2, reshapedAlpha, leakyreluAlpha);\n      backend2.disposeIntermediateTensorInfo(reshapedAlpha);\n    } else {\n      result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);\n    }\n    backend2.disposeIntermediateTensorInfo(resultOld);\n  }\n  return result;\n}\nvar fusedConv2DConfig = {\n  kernelName: FusedConv2D,\n  backendName: \"cpu\",\n  kernelFunc: fusedConv2D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FusedDepthwiseConv2D.js\nfunction fusedDepthwiseConv2D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  let result = depthwiseConv2dNative({\n    inputs: { x, filter },\n    backend: backend2,\n    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }\n  });\n  if (bias) {\n    const oldResult = result;\n    result = add4({ inputs: { a: result, b: bias }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(oldResult);\n  }\n  if (activation2) {\n    const oldResult = result;\n    result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);\n    backend2.disposeIntermediateTensorInfo(oldResult);\n  }\n  return result;\n}\nvar fusedDepthwiseConv2DConfig = {\n  kernelName: FusedDepthwiseConv2D,\n  backendName: \"cpu\",\n  kernelFunc: fusedDepthwiseConv2D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherNd.js\nfunction gatherNd(args) {\n  const { inputs, backend: backend2 } = args;\n  const { params, indices } = inputs;\n  const paramsSize = util_exports.sizeFromShape(params.shape);\n  const indicesShape = indices.shape;\n  const sliceRank = indicesShape[indicesShape.length - 1];\n  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);\n  if (numSlices === 0) {\n    return backend2.makeTensorInfo(resultShape, params.dtype, []);\n  }\n  const indicesData = backend2.data.get(indices.dataId).values;\n  const paramsBuf = backend2.bufferSync(params);\n  const outBuf = gatherNdImpl(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);\n  return backend2.makeTensorInfo(resultShape, params.dtype, outBuf.values);\n}\nvar gatherNdConfig = {\n  kernelName: GatherNd,\n  backendName: \"cpu\",\n  kernelFunc: gatherNd\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherV2.js\nfunction gatherV2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, indices } = inputs;\n  const { axis, batchDims } = attrs;\n  assertNotComplex([x, indices], \"gatherV2\");\n  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];\n  const indicesVals = backend2.data.get(indices.dataId).values;\n  const axisDim = x.shape[parsedAxis];\n  for (let i = 0; i < indicesVals.length; ++i) {\n    const index = indicesVals[i];\n    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);\n  }\n  let $batchDims = batchDims;\n  if (batchDims == null) {\n    $batchDims = 0;\n  }\n  const indicesSize = util_exports.sizeFromShape(indices.shape);\n  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, $batchDims);\n  const flattenX = reshape3({\n    inputs: { x },\n    backend: backend2,\n    attrs: {\n      shape: [\n        shapeInfo.batchSize,\n        shapeInfo.outerSize,\n        shapeInfo.dimSize,\n        shapeInfo.sliceSize\n      ]\n    }\n  });\n  const flattenIndex = reshape3({\n    inputs: { x: indices },\n    backend: backend2,\n    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }\n  });\n  const flattenOutputShape = [\n    shapeInfo.batchSize,\n    shapeInfo.outerSize,\n    indicesSize / shapeInfo.batchSize,\n    shapeInfo.sliceSize\n  ];\n  const indicesBuf = backend2.bufferSync(flattenIndex);\n  const xBuf = backend2.bufferSync(flattenX);\n  const outBuf = gatherV2Impl(xBuf, indicesBuf, flattenOutputShape);\n  backend2.disposeIntermediateTensorInfo(flattenX);\n  backend2.disposeIntermediateTensorInfo(flattenIndex);\n  return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);\n}\nvar gatherV2Config = {\n  kernelName: GatherV2,\n  backendName: \"cpu\",\n  kernelFunc: gatherV2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IFFT.js\nfunction ifft2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const inputSize = util_exports.sizeFromShape(input2.shape);\n  const innerDimensionSize = input2.shape[input2.shape.length - 1];\n  const batch = inputSize / innerDimensionSize;\n  const input2D = reshape3({\n    inputs: { x: input2 },\n    backend: backend2,\n    attrs: { shape: [batch, innerDimensionSize] }\n  });\n  const result = fftBatch(input2D, true, backend2);\n  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });\n  backend2.disposeIntermediateTensorInfo(input2D);\n  backend2.disposeIntermediateTensorInfo(result);\n  return resultReshaped;\n}\nvar ifftConfig = {\n  kernelName: IFFT,\n  backendName: \"cpu\",\n  kernelFunc: ifft2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsFinite.js\nvar isFinite3 = unaryKernelFunc(IsFinite, (xi) => Number.isFinite(xi) ? 1 : 0, \"bool\");\nvar isFiniteConfig = {\n  kernelName: IsFinite,\n  backendName: \"cpu\",\n  kernelFunc: isFinite3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsInf.js\nvar isInf2 = unaryKernelFunc(IsInf, (xi) => Math.abs(xi) === Infinity ? 1 : 0, \"bool\");\nvar isInfConfig = {\n  kernelName: IsInf,\n  backendName: \"cpu\",\n  kernelFunc: isInf2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsNaN.js\nvar isNaN3 = unaryKernelFunc(IsNan, (xi) => Number.isNaN(xi) ? 1 : 0, \"bool\");\nvar isNaNConfig = {\n  kernelName: IsNan,\n  backendName: \"cpu\",\n  kernelFunc: isNaN3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LinSpace.js\nfunction linSpace(args) {\n  const { backend: backend2, attrs } = args;\n  const { start, stop, num } = attrs;\n  const outVals = linSpaceImpl(start, stop, num);\n  return backend2.makeTensorInfo([outVals.length], \"float32\", outVals);\n}\nvar linSpaceConfig = {\n  kernelName: LinSpace,\n  backendName: \"cpu\",\n  kernelFunc: linSpace\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Log1p.js\nvar log1p2 = unaryKernelFunc(Log1p, (xi) => Math.log1p(xi));\nvar log1pConfig = {\n  kernelName: Log1p,\n  backendName: \"cpu\",\n  kernelFunc: log1p2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalAnd.js\nvar logicalAndImpl = createSimpleBinaryKernelImpl((a, b) => a && b);\nvar logicalAnd2 = binaryKernelFunc(LogicalAnd, logicalAndImpl, null, \"bool\");\nvar logicalAndConfig = {\n  kernelName: LogicalAnd,\n  backendName: \"cpu\",\n  kernelFunc: logicalAnd2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalNot.js\nvar logicalNot2 = unaryKernelFunc(LogicalNot, (xi) => xi ? 0 : 1, \"bool\");\nvar logicalNotConfig = {\n  kernelName: LogicalNot,\n  backendName: \"cpu\",\n  kernelFunc: logicalNot2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalOr.js\nvar logicalOrImpl = createSimpleBinaryKernelImpl((a, b) => a || b);\nvar logicalOr2 = binaryKernelFunc(LogicalOr, logicalOrImpl, null, \"bool\");\nvar logicalOrConfig = {\n  kernelName: LogicalOr,\n  backendName: \"cpu\",\n  kernelFunc: logicalOr2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LRN.js\nfunction lRN(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { depthRadius, bias, alpha, beta } = attrs;\n  assertNotComplex(x, \"LRN\");\n  const channels = x.shape[3];\n  const maxD = channels - 1;\n  const xValues = backend2.data.get(x.dataId).values;\n  const size = util_exports.sizeFromShape(x.shape);\n  const result = new Float32Array(size);\n  function sumAcrossChannels(offset) {\n    const currentChannel = offset % channels;\n    let beginSumOffset = offset - currentChannel + Math.max(0, currentChannel - depthRadius);\n    const endSumOffset = offset - currentChannel + Math.min(currentChannel + depthRadius, maxD);\n    let sum6 = 0;\n    for (; beginSumOffset <= endSumOffset; beginSumOffset++) {\n      const z = xValues[beginSumOffset];\n      sum6 += z * z;\n    }\n    return sum6;\n  }\n  for (let offset = 0; offset < size; offset++) {\n    const sum6 = sumAcrossChannels(offset);\n    const val = xValues[offset] * Math.pow(bias + alpha * sum6, -beta);\n    result[offset] = val;\n  }\n  return backend2.makeTensorInfo(x.shape, x.dtype, result);\n}\nvar LRNConfig = {\n  kernelName: LRN,\n  backendName: \"cpu\",\n  kernelFunc: lRN\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LRNGrad.js\nfunction lRNGrad(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, y, dy } = inputs;\n  const { depthRadius, bias, alpha, beta } = attrs;\n  assertNotComplex(dy, \"LRNGrad\");\n  const dySize = util_exports.sizeFromShape(dy.shape);\n  const channels = dy.shape[3];\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const xValues = backend2.data.get(x.dataId).values;\n  const yValues = backend2.data.get(y.dataId).values;\n  const result = new Float32Array(dySize);\n  const size = dySize;\n  for (let offset = 0; offset < size; offset++) {\n    const currentChannel = offset % channels;\n    const depthBegin = offset - currentChannel + Math.max(0, currentChannel - depthRadius);\n    const depthEnd = offset - currentChannel + Math.min(channels, currentChannel + depthRadius + 1);\n    let norm2 = 0;\n    for (let k = depthBegin; k < depthEnd; k++) {\n      norm2 += Math.pow(xValues[k], 2);\n    }\n    norm2 = alpha * norm2 + bias;\n    for (let k = depthBegin; k < depthEnd; k++) {\n      let dyi = -2 * alpha * beta * xValues[k] * yValues[offset] / norm2;\n      if (offset === k) {\n        dyi += Math.pow(norm2, -beta);\n      }\n      dyi *= dyValues[offset];\n      result[k] += dyi;\n    }\n  }\n  return backend2.makeTensorInfo(dy.shape, x.dtype, result);\n}\nvar LRNGradConfig = {\n  kernelName: LRNGrad,\n  backendName: \"cpu\",\n  kernelFunc: lRNGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Max.js\nfunction max3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { reductionIndices, keepDims } = attrs;\n  const cpuBackend = backend2;\n  let xShape = x.shape;\n  const xRank = xShape.length;\n  const origAxes = util_exports.parseAxisParam(reductionIndices, xShape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let xVals = cpuBackend.data.get(x.dataId).values;\n  if (permutedAxes != null) {\n    const newShape = new Array(xRank);\n    for (let i = 0; i < newShape.length; i++) {\n      newShape[i] = xShape[permutedAxes[i]];\n    }\n    xVals = transposeImpl(xVals, xShape, x.dtype, permutedAxes, newShape);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n    xShape = newShape;\n  }\n  assertNotComplex(x, \"max\");\n  backend_util_exports.assertAxesAreInnerMostDims(\"max\", axes, xRank);\n  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const result = maxImpl(xVals, reduceSize, maxOutShape, x.dtype);\n  const dataId = cpuBackend.write(result, maxOutShape, x.dtype);\n  let outShape = maxOutShape;\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);\n    outShape = newShape;\n  }\n  return { dataId, shape: outShape, dtype: x.dtype };\n}\nvar maxConfig = {\n  kernelName: Max,\n  backendName: \"cpu\",\n  kernelFunc: max3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool.js\nfunction maxPool2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  assertNotComplex(x, \"maxPool\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = 1;\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  let res;\n  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {\n    res = identity2({ inputs: { x }, backend: backend2 });\n  } else {\n    const xValues = backend2.data.get(x.dataId).values;\n    const strides2 = util_exports.computeStrides(x.shape);\n    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, \"max\");\n    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);\n  }\n  return res;\n}\nvar maxPoolConfig = {\n  kernelName: MaxPool,\n  backendName: \"cpu\",\n  kernelFunc: maxPool2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool3D.js\nfunction maxPool3D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;\n  assertNotComplex(x, \"maxPool3d\");\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);\n  const xValues = backend2.data.get(x.dataId).values;\n  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, \"max\");\n  return backend2.makeTensorInfo(outBuf.shape, \"float32\", outBuf.values);\n}\nvar maxPool3DConfig = {\n  kernelName: MaxPool3D,\n  backendName: \"cpu\",\n  kernelFunc: maxPool3D\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool3DGrad.js\nfunction maxPool3DGrad(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  assertNotComplex([dy, input2], \"maxPool3DGrad\");\n  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const inputBuf = backend2.bufferSync(input2);\n  const maxPosBuf = maxPool3dPositions(inputBuf, convInfo);\n  const strideDepth = convInfo.strideDepth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationDepth = convInfo.dilationDepth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;\n  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n  const dx = buffer(input2.shape, \"float32\");\n  const dyBuf = backend2.bufferSync(dy);\n  for (let batch = 0; batch < convInfo.batchSize; ++batch) {\n    for (let channel = 0; channel < convInfo.inChannels; ++channel) {\n      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {\n        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {\n          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {\n            const dyDepthCorner = dxDepth - padFront;\n            const dyRowCorner = dxRow - padTop;\n            const dyColCorner = dxCol - padLeft;\n            let dotProd = 0;\n            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {\n              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;\n              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {\n                continue;\n              }\n              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {\n                const dyRow = (dyRowCorner + wRow) / strideHeight;\n                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {\n                  continue;\n                }\n                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {\n                  const dyCol = (dyColCorner + wCol) / strideWidth;\n                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {\n                    continue;\n                  }\n                  const maxPos = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(batch, dyDepth, dyRow, dyCol, channel);\n                  const curPos = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterWidth + wCol;\n                  const mask = maxPos === curPos ? 1 : 0;\n                  if (mask === 0) {\n                    continue;\n                  }\n                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);\n                  dotProd += pixel * mask;\n                }\n              }\n            }\n            dx.set(dotProd, batch, dxDepth, dxRow, dxCol, channel);\n          }\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar maxPool3DGradConfig2 = {\n  kernelName: MaxPool3DGrad,\n  backendName: \"cpu\",\n  kernelFunc: maxPool3DGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolGrad.js\nfunction maxPoolGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2, output } = inputs;\n  const x = input2;\n  assertNotComplex([input2, output], \"maxPoolGrad\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const xValues = backend2.data.get(x.dataId).values;\n  const maxPosBuf = buffer(convInfo.outShape, x.dtype, maxPoolPositions(xValues, x.shape, x.dtype, convInfo).values);\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n  const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n  const dx = buffer(x.shape, \"float32\");\n  const dyData = backend2.data.get(dy.dataId).values;\n  const dyBuf = buffer(dy.shape, \"float32\", dyData);\n  for (let b = 0; b < convInfo.batchSize; ++b) {\n    for (let d = 0; d < convInfo.inChannels; ++d) {\n      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {\n        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {\n          const dyRCorner = dxR - padTop;\n          const dyCCorner = dxC - padLeft;\n          let dotProd = 0;\n          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {\n            const dyR = (dyRCorner + wR) / strideHeight;\n            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {\n              continue;\n            }\n            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {\n              const dyC = (dyCCorner + wC) / strideWidth;\n              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {\n                continue;\n              }\n              const maxPos = effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(b, dyR, dyC, d);\n              const curPos = wR * effectiveFilterWidth + wC;\n              const mask = maxPos === curPos ? 1 : 0;\n              if (mask === 0) {\n                continue;\n              }\n              const pixel = dyBuf.get(b, dyR, dyC, d);\n              dotProd += pixel * mask;\n            }\n          }\n          dx.set(dotProd, b, dxR, dxC, d);\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);\n}\nvar maxPoolGradConfig2 = {\n  kernelName: MaxPoolGrad,\n  backendName: \"cpu\",\n  kernelFunc: maxPoolGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolWithArgmax_impl.js\nfunction maxPoolWithArgmaxImpl(xValues, xShape, dtype, includeBatchInIndex, convInfo) {\n  const strides = util_exports.computeStrides(xShape);\n  const maxPools = pool2(xValues, xShape, dtype, strides, convInfo, \"max\");\n  const maxPositions = maxPoolPositions(xValues, xShape, dtype, convInfo, true, includeBatchInIndex);\n  return [maxPools.values, maxPositions.values];\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolWithArgmax.js\nvar maxPoolWithArgmaxConfig = {\n  kernelName: MaxPoolWithArgmax,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { x } = inputs;\n    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;\n    const cpuBackend = backend2;\n    assertNotComplex(x, \"MaxPoolWithArgmax\");\n    const values = cpuBackend.data.get(x.dataId).values;\n    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, [1, 1], pad3);\n    const [pooled, indexes] = maxPoolWithArgmaxImpl(values, x.shape, x.dtype, includeBatchInIndex, convInfo);\n    const pooledDataId = cpuBackend.write(pooled, convInfo.outShape, x.dtype);\n    const indexesDataId = cpuBackend.write(indexes, convInfo.outShape, x.dtype);\n    return [\n      { dataId: pooledDataId, shape: convInfo.outShape, dtype: x.dtype },\n      { dataId: indexesDataId, shape: convInfo.outShape, dtype: \"int32\" }\n    ];\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Mean.js\nfunction mean2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  const axes = util_exports.parseAxisParam(axis, x.shape);\n  const shapes = backend_util_exports.computeOutAndReduceShapes(x.shape, axes);\n  const reduceShape = shapes[1];\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const toDispose = [];\n  const reduceSizeScalar = backend2.makeTensorInfo([], \"float32\", new Float32Array([reduceSize]));\n  toDispose.push(reduceSizeScalar);\n  const $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: \"float32\" } });\n  toDispose.push($x);\n  const res = div2({ inputs: { a: $x, b: reduceSizeScalar }, backend: backend2 });\n  toDispose.push(res);\n  const result = sum3({ inputs: { x: res }, backend: backend2, attrs: { axis, keepDims } });\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar meanConfig = {\n  kernelName: Mean,\n  backendName: \"cpu\",\n  kernelFunc: mean2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Min.js\nfunction min3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  assertNotComplex(x, \"min\");\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  if (permutedAxes != null) {\n    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"min\", axes, $x.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);\n  const aVals = backend2.data.get($x.dataId).values;\n  for (let i = 0; i < vals.length; ++i) {\n    const offset = i * reduceSize;\n    let min6 = aVals[offset];\n    for (let j = 0; j < reduceSize; ++j) {\n      const value = aVals[offset + j];\n      if (Number.isNaN(value) || value < min6) {\n        min6 = value;\n      }\n    }\n    vals[i] = min6;\n  }\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo($x);\n  }\n  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);\n  if (keepDims) {\n    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });\n    backend2.disposeIntermediateTensorInfo(result);\n    return reshapedResult;\n  }\n  return result;\n}\nvar minConfig = {\n  kernelName: Min,\n  backendName: \"cpu\",\n  kernelFunc: min3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MirrorPad.js\nfunction mirrorPad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { paddings, mode } = attrs;\n  assertNotComplex(x, \"mirrorPad\");\n  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);\n  const start = paddings.map((p2) => p2[0]);\n  const end = paddings.map((p2, i) => p2[0] + x.shape[i]);\n  const offset = mode === \"reflect\" ? 0 : 1;\n  const xVals = backend2.data.get(x.dataId).values;\n  const xRank = x.shape.length;\n  const xStrides = util_exports.computeStrides(x.shape);\n  const resultSize = util_exports.sizeFromShape(outShape);\n  const resultRank = outShape.length;\n  const resultStrides = util_exports.computeStrides(outShape);\n  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);\n  for (let i = 0; i < resultSize; i++) {\n    let coords2 = util_exports.indexToLoc(i, resultRank, resultStrides);\n    for (let i2 = 0; i2 < resultRank; i2++) {\n      if (coords2[i2] < start[i2]) {\n        coords2[i2] = start[i2] * 2 - coords2[i2] - offset;\n      } else if (coords2[i2] >= end[i2]) {\n        coords2[i2] = (end[i2] - 1) * 2 - coords2[i2] + offset;\n      }\n    }\n    coords2 = coords2.map((c, i2) => c - start[i2]);\n    const inIndex = util_exports.locToIndex(coords2, xRank, xStrides);\n    resVals[i] = xVals[inIndex];\n  }\n  const outId = backend2.write(resVals, outShape, x.dtype);\n  return { dataId: outId, shape: outShape, dtype: x.dtype };\n}\nvar mirrorPadConfig = {\n  kernelName: MirrorPad,\n  backendName: \"cpu\",\n  kernelFunc: mirrorPad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Mod.js\nvar modImpl = createSimpleBinaryKernelImpl((aValue, bValue) => {\n  const rem = aValue % bValue;\n  if (aValue < 0 && bValue < 0 || aValue >= 0 && bValue >= 0) {\n    return rem;\n  } else {\n    return (rem + bValue) % bValue;\n  }\n});\nvar mod2 = binaryKernelFunc(Mod, modImpl);\nvar modConfig = {\n  kernelName: Mod,\n  backendName: \"cpu\",\n  kernelFunc: mod2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multinomial.js\nvar seedrandom4 = __toESM(require_seedrandom2());\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Softmax.js\nfunction softmax3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { logits } = inputs;\n  const { dim } = attrs;\n  const logitsRank = logits.shape.length;\n  let $dim = dim;\n  if ($dim === -1) {\n    $dim = logitsRank - 1;\n  }\n  if ($dim !== logitsRank - 1) {\n    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${logitsRank} and dim was ${$dim}`);\n  }\n  const axes = util_exports.parseAxisParam([$dim], logits.shape);\n  const maxLogit = max3({\n    inputs: { x: logits },\n    backend: backend2,\n    attrs: { reductionIndices: axes, keepDims: false }\n  });\n  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);\n  const maxLogitReshaped = reshape3({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });\n  const a = sub2({ inputs: { a: logits, b: maxLogitReshaped }, backend: backend2 });\n  const b = exp2({ inputs: { x: a }, backend: backend2 });\n  const sumExp = sum3({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });\n  const sumReshaped = reshape3({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });\n  const result = div2({ inputs: { a: b, b: sumReshaped }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(maxLogit);\n  backend2.disposeIntermediateTensorInfo(maxLogitReshaped);\n  backend2.disposeIntermediateTensorInfo(a);\n  backend2.disposeIntermediateTensorInfo(b);\n  backend2.disposeIntermediateTensorInfo(sumExp);\n  backend2.disposeIntermediateTensorInfo(sumReshaped);\n  return result;\n}\nvar softmaxConfig = {\n  kernelName: Softmax,\n  backendName: \"cpu\",\n  kernelFunc: softmax3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multinomial.js\nfunction multinomial2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { logits } = inputs;\n  const { numSamples, seed, normalized } = attrs;\n  assertNotComplex(logits, \"multinomial\");\n  const probabilities = normalized ? logits : softmax3({ inputs: { logits }, backend: backend2, attrs: { dim: -1 } });\n  const batchSize = probabilities.shape[0];\n  const numEvents = probabilities.shape[1];\n  const probVals = backend2.data.get(probabilities.dataId).values;\n  const resShape = [batchSize, numSamples];\n  const resVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(resShape), \"int32\");\n  for (let b = 0; b < batchSize; ++b) {\n    const offset = b * numEvents;\n    const cdf = new Float32Array(numEvents - 1);\n    cdf[0] = probVals[offset];\n    for (let event = 1; event < cdf.length; ++event) {\n      cdf[event] = cdf[event - 1] + probVals[offset + event];\n    }\n    const random = seedrandom4.alea(seed.toString());\n    const outOffset = b * numSamples;\n    for (let sampleId = 0; sampleId < numSamples; ++sampleId) {\n      const r = random();\n      resVals[outOffset + sampleId] = cdf.length;\n      for (let event = 0; event < cdf.length; event++) {\n        if (r < cdf[event]) {\n          resVals[outOffset + sampleId] = event;\n          break;\n        }\n      }\n    }\n  }\n  if (!normalized) {\n    backend2.disposeIntermediateTensorInfo(probabilities);\n  }\n  return backend2.makeTensorInfo(resShape, \"int32\", resVals);\n}\nvar multinomialConfig = {\n  kernelName: Multinomial,\n  backendName: \"cpu\",\n  kernelFunc: multinomial2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV3.js\nvar nonMaxSuppressionV3Impl2 = kernel_impls_exports.nonMaxSuppressionV3Impl;\nfunction nonMaxSuppressionV3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;\n  assertNotComplex(boxes, \"NonMaxSuppression\");\n  const boxesVals = backend2.data.get(boxes.dataId).values;\n  const scoresVals = backend2.data.get(scores.dataId).values;\n  const { selectedIndices } = nonMaxSuppressionV3Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);\n  return backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices));\n}\nvar nonMaxSuppressionV3Config = {\n  kernelName: NonMaxSuppressionV3,\n  backendName: \"cpu\",\n  kernelFunc: nonMaxSuppressionV3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV4.js\nvar nonMaxSuppressionV4Impl2 = kernel_impls_exports.nonMaxSuppressionV4Impl;\nfunction nonMaxSuppressionV4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;\n  assertNotComplex(boxes, \"NonMaxSuppressionPadded\");\n  const boxesVals = backend2.data.get(boxes.dataId).values;\n  const scoresVals = backend2.data.get(scores.dataId).values;\n  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);\n  return [\n    backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices)),\n    backend2.makeTensorInfo([], \"int32\", new Int32Array([validOutputs]))\n  ];\n}\nvar nonMaxSuppressionV4Config = {\n  kernelName: NonMaxSuppressionV4,\n  backendName: \"cpu\",\n  kernelFunc: nonMaxSuppressionV4\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV5.js\nvar nonMaxSuppressionV5Impl2 = kernel_impls_exports.nonMaxSuppressionV5Impl;\nfunction nonMaxSuppressionV5(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;\n  assertNotComplex(boxes, \"NonMaxSuppressionWithScore\");\n  const boxesVals = backend2.data.get(boxes.dataId).values;\n  const scoresVals = backend2.data.get(scores.dataId).values;\n  const maxOutputSizeVal = maxOutputSize;\n  const iouThresholdVal = iouThreshold;\n  const scoreThresholdVal = scoreThreshold;\n  const softNmsSigmaVal = softNmsSigma;\n  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl2(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);\n  return [\n    backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices)),\n    backend2.makeTensorInfo([selectedScores.length], \"float32\", new Float32Array(selectedScores))\n  ];\n}\nvar nonMaxSuppressionV5Config = {\n  kernelName: NonMaxSuppressionV5,\n  backendName: \"cpu\",\n  kernelFunc: nonMaxSuppressionV5\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/OneHot.js\nfunction oneHot2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { indices } = inputs;\n  const { dtype, depth, onValue, offValue } = attrs;\n  assertNotComplex(indices, \"oneHot\");\n  const indicesSize = util_exports.sizeFromShape(indices.shape);\n  const res = new Float32Array(indicesSize * depth);\n  res.fill(offValue);\n  const indicesVal = backend2.data.get(indices.dataId).values;\n  for (let event = 0; event < indicesSize; ++event) {\n    if (indicesVal[event] >= 0 && indicesVal[event] < depth) {\n      res[event * depth + indicesVal[event]] = onValue;\n    }\n  }\n  return backend2.makeTensorInfo([...indices.shape, depth], dtype, res);\n}\nvar oneHotConfig = {\n  kernelName: OneHot,\n  backendName: \"cpu\",\n  kernelFunc: oneHot2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ZerosLike.js\nfunction zerosLike2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (x.dtype === \"string\") {\n    throw new Error(\"zerosLike is not supported for string tensors\");\n  } else if (x.dtype === \"complex64\") {\n    const realPart = real2({ inputs: { input: x }, backend: backend2 });\n    const r = zerosLike2({ inputs: { x: realPart }, backend: backend2 });\n    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });\n    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });\n    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    backend2.disposeIntermediateTensorInfo(r);\n    backend2.disposeIntermediateTensorInfo(imagPart);\n    backend2.disposeIntermediateTensorInfo(i);\n    return result;\n  } else {\n    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 0, dtype: x.dtype } });\n  }\n}\nvar zerosLikeConfig = {\n  kernelName: ZerosLike,\n  backendName: \"cpu\",\n  kernelFunc: zerosLike2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/OnesLike.js\nfunction onesLike2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (x.dtype === \"string\") {\n    throw new Error(\"onesLike is not supported for string tensors\");\n  } else if (x.dtype === \"complex64\") {\n    const realPart = real2({ inputs: { input: x }, backend: backend2 });\n    const r = onesLike2({ inputs: { x: realPart }, backend: backend2 });\n    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });\n    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });\n    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    backend2.disposeIntermediateTensorInfo(r);\n    backend2.disposeIntermediateTensorInfo(imagPart);\n    backend2.disposeIntermediateTensorInfo(i);\n    return result;\n  } else {\n    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 1, dtype: x.dtype } });\n  }\n}\nvar onesLikeConfig = {\n  kernelName: OnesLike,\n  backendName: \"cpu\",\n  kernelFunc: onesLike2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Pack.js\nfunction pack(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { axis } = attrs;\n  if (inputs.length === 1) {\n    return expandDims3({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });\n  }\n  const shape = inputs[0].shape;\n  const dtype = inputs[0].dtype;\n  inputs.forEach((t) => {\n    util_exports.assertShapesMatch(shape, t.shape, \"All tensors passed to stack must have matching shapes\");\n    util_exports.assert(dtype === t.dtype, () => \"All tensors passed to stack must have matching dtypes\");\n  });\n  const intermediateTensorInfos = [];\n  const expandedTensors = inputs.map((t) => {\n    const expandedT = expandDims3({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });\n    intermediateTensorInfos.push(expandedT);\n    return expandedT;\n  });\n  const result = concat2({ inputs: expandedTensors, backend: backend2, attrs: { axis } });\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar packConfig = {\n  kernelName: Pack,\n  backendName: \"cpu\",\n  kernelFunc: pack\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/PadV2.js\nfunction padV2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { paddings, constantValue } = attrs;\n  assertNotComplex(x, \"pad\");\n  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);\n  const start = paddings.map((p2) => p2[0]);\n  const xVals = backend2.data.get(x.dataId).values;\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const xRank = x.shape.length;\n  const xStrides = util_exports.computeStrides(x.shape);\n  const resultSize = util_exports.sizeFromShape(outShape);\n  const resultRank = outShape.length;\n  const resultStrides = util_exports.computeStrides(outShape);\n  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);\n  if (constantValue !== 0) {\n    resVals.fill(constantValue);\n  }\n  for (let i = 0; i < xSize; i++) {\n    const coords2 = util_exports.indexToLoc(i, xRank, xStrides);\n    const outCoords = coords2.map((c, i2) => c + start[i2]);\n    const outIndex = util_exports.locToIndex(outCoords, resultRank, resultStrides);\n    resVals[outIndex] = xVals[i];\n  }\n  const outId = backend2.write(resVals, outShape, x.dtype);\n  return { dataId: outId, shape: outShape, dtype: x.dtype };\n}\nvar padV2Config = {\n  kernelName: PadV2,\n  backendName: \"cpu\",\n  kernelFunc: padV2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Pow.js\nvar powImpl = createSimpleBinaryKernelImpl((a, b) => Math.pow(a, b));\nvar pow2 = binaryKernelFunc(Pow, powImpl);\nvar powConfig = {\n  kernelName: Pow,\n  backendName: \"cpu\",\n  kernelFunc: pow2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedGather.js\nfunction raggedGather2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;\n  const { outputRaggedRank } = attrs;\n  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.data.get(t.dataId).values);\n  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);\n  const $paramsDenseValues = backend2.data.get(paramsDenseValues.dataId).values;\n  const $indices = backend2.data.get(indices.dataId).values;\n  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImpl($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);\n  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], \"int32\", splits));\n  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);\n  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);\n}\nvar raggedGatherConfig = {\n  kernelName: RaggedGather,\n  backendName: \"cpu\",\n  kernelFunc: raggedGather2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedRange.js\nfunction raggedRange2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { starts, limits, deltas } = inputs;\n  const $starts = backend2.data.get(starts.dataId).values;\n  const $limits = backend2.data.get(limits.dataId).values;\n  const $deltas = backend2.data.get(deltas.dataId).values;\n  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImpl($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);\n  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], \"int32\", rtNestedSplitsData);\n  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);\n  return [rtNestedSplits, rtDenseValues];\n}\nvar raggedRangeConfig = {\n  kernelName: RaggedRange,\n  backendName: \"cpu\",\n  kernelFunc: raggedRange2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedTensorToTensor.js\nfunction raggedTensorToTensor2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { shape, values, defaultValue, rowPartitionTensors } = inputs;\n  const { rowPartitionTypes } = attrs;\n  const $shape = backend2.data.get(shape.dataId).values;\n  const $values = backend2.data.get(values.dataId).values;\n  const $defaultValue = backend2.data.get(defaultValue.dataId).values;\n  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.data.get(t.dataId).values);\n  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);\n  const [outputShape, output] = raggedTensorToTensorImpl($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);\n  return backend2.makeTensorInfo(outputShape, values.dtype, output);\n}\nvar raggedTensorToTensorConfig = {\n  kernelName: RaggedTensorToTensor,\n  backendName: \"cpu\",\n  kernelFunc: raggedTensorToTensor2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Range.js\nfunction range3(args) {\n  const { backend: backend2, attrs } = args;\n  const { start, stop, dtype, step: step5 } = attrs;\n  const values = rangeImpl(start, stop, step5, dtype);\n  return backend2.makeTensorInfo([values.length], dtype, values);\n}\nvar rangeConfig = {\n  kernelName: Range,\n  backendName: \"cpu\",\n  kernelFunc: range3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reciprocal.js\nvar reciprocal2 = unaryKernelFunc(Reciprocal, (xi) => 1 / xi);\nvar reciprocalConfig = {\n  kernelName: Reciprocal,\n  backendName: \"cpu\",\n  kernelFunc: reciprocal2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeBilinear.js\nfunction resizeBilinear2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  assertNotComplex(images, \"resizeBilinear\");\n  const imagesStrides = util_exports.computeStrides(images.shape);\n  const [newHeight, newWidth] = size;\n  const [batch, oldHeight, oldWidth, numChannels] = images.shape;\n  const xValues = backend2.data.get(images.dataId).values;\n  const result = new Float32Array(util_exports.sizeFromShape([batch, newHeight, newWidth, numChannels]));\n  const effectiveInputSize = [\n    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n  ];\n  const effectiveOutputSize = [\n    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n  ];\n  let outputIdx = 0;\n  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];\n  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];\n  for (let b = 0; b < batch; b++) {\n    for (let r = 0; r < newHeight; r++) {\n      let sourceFracRow;\n      if (halfPixelCenters) {\n        sourceFracRow = effectiveRowSizeRatio * (r + 0.5) - 0.5;\n      } else {\n        sourceFracRow = effectiveRowSizeRatio * r;\n      }\n      const sourceRowFloor = Math.max(0, Math.floor(sourceFracRow));\n      const rowFrac = sourceFracRow - sourceRowFloor;\n      const sourceRowCeil = Math.min(oldHeight - 1, Math.ceil(sourceFracRow));\n      const topRowOffset = b * imagesStrides[0] + sourceRowFloor * imagesStrides[1];\n      const botRowOffset = b * imagesStrides[0] + sourceRowCeil * imagesStrides[1];\n      for (let c = 0; c < newWidth; c++) {\n        let sourceFracCol;\n        if (halfPixelCenters) {\n          sourceFracCol = effectiveColSizeRatio * (c + 0.5) - 0.5;\n        } else {\n          sourceFracCol = effectiveColSizeRatio * c;\n        }\n        const sourceColFloor = Math.max(0, Math.floor(sourceFracCol));\n        const colFrac = sourceFracCol - sourceColFloor;\n        const sourceColCeil = Math.min(oldWidth - 1, Math.ceil(sourceFracCol));\n        const topLeftOffest = topRowOffset + sourceColFloor * imagesStrides[2];\n        const botLeftOffset = botRowOffset + sourceColFloor * imagesStrides[2];\n        const topRightOffset = topRowOffset + sourceColCeil * imagesStrides[2];\n        const botRightOffest = botRowOffset + sourceColCeil * imagesStrides[2];\n        for (let d = 0; d < numChannels; d++) {\n          const topLeft = xValues[topLeftOffest + d];\n          const bottomLeft = xValues[botLeftOffset + d];\n          const topRight = xValues[topRightOffset + d];\n          const bottomRight = xValues[botRightOffest + d];\n          const top = topLeft + (topRight - topLeft) * colFrac;\n          const bottom = bottomLeft + (bottomRight - bottomLeft) * colFrac;\n          const newValue = top + (bottom - top) * rowFrac;\n          result[outputIdx++] = newValue;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], \"float32\", result);\n}\nvar resizeBilinearConfig = {\n  kernelName: ResizeBilinear,\n  backendName: \"cpu\",\n  kernelFunc: resizeBilinear2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeBilinearGrad.js\nfunction resizeBilinearGrad(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images, dy } = inputs;\n  const { alignCorners } = attrs;\n  assertNotComplex([dy, images], \"resizeBilinearGrad\");\n  const imagesStrides = util_exports.computeStrides(images.shape);\n  const [batch, xHeight, xWidth, depth] = images.shape;\n  const [, yHeight, yWidth] = dy.shape;\n  const output = new Float32Array(batch * xHeight * xWidth * depth);\n  const effectiveXSize = [\n    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,\n    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth\n  ];\n  const effectiveYSize = [\n    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,\n    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth\n  ];\n  const heightScale = effectiveXSize[0] / effectiveYSize[0];\n  const widthScale = effectiveXSize[1] / effectiveYSize[1];\n  const dyValues = backend2.data.get(dy.dataId).values;\n  let offset = 0;\n  for (let b = 0; b < batch; b++) {\n    const bOffset = b * imagesStrides[0];\n    for (let r = 0; r < yHeight; r++) {\n      const dxR = r * heightScale;\n      const topDxRIndex = Math.floor(dxR);\n      const bottomDxRIndex = Math.min(Math.ceil(dxR), xHeight - 1);\n      const topDxROffset = bOffset + topDxRIndex * imagesStrides[1];\n      const bottomDxROffset = bOffset + bottomDxRIndex * imagesStrides[1];\n      const dxRLerp = dxR - topDxRIndex;\n      const inverseDxRLerp = 1 - dxRLerp;\n      for (let c = 0; c < yWidth; c++) {\n        const dxC = c * widthScale;\n        const leftDxCIndex = Math.floor(dxC);\n        const rightDxCIndex = Math.min(Math.ceil(dxC), xWidth - 1);\n        const dxCLerp = dxC - leftDxCIndex;\n        const inverseDxCLerp = 1 - dxCLerp;\n        const topLeftRCOffset = topDxROffset + leftDxCIndex * imagesStrides[2];\n        const topRightRCOffset = topDxROffset + rightDxCIndex * imagesStrides[2];\n        const bottomLeftRCOffset = bottomDxROffset + leftDxCIndex * imagesStrides[2];\n        const bottomRightRCOffset = bottomDxROffset + rightDxCIndex * imagesStrides[2];\n        const inverseDxRLerpTimesInverseDxCLerp = inverseDxRLerp * inverseDxCLerp;\n        const inverseDxRLerpTimesDxCLerp = inverseDxRLerp * dxCLerp;\n        const dxRLerpTimesInverseDxCLerp = dxRLerp * inverseDxCLerp;\n        const dxRLerpTimesDxCLerp = dxRLerp * dxCLerp;\n        for (let d = 0; d < depth; d++) {\n          const dyVal = dyValues[offset++];\n          output[topLeftRCOffset + d] += dyVal * inverseDxRLerpTimesInverseDxCLerp;\n          output[topRightRCOffset + d] += dyVal * inverseDxRLerpTimesDxCLerp;\n          output[bottomLeftRCOffset + d] += dyVal * dxRLerpTimesInverseDxCLerp;\n          output[bottomRightRCOffset + d] += dyVal * dxRLerpTimesDxCLerp;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo([batch, xWidth, xHeight, depth], \"float32\", output);\n}\nvar resizeBilinearGradConfig2 = {\n  kernelName: ResizeBilinearGrad,\n  backendName: \"cpu\",\n  kernelFunc: resizeBilinearGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeNearestNeighbor.js\nfunction resizeNearestNeighbor2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  assertNotComplex(images, \"resizeNearestNeighbor\");\n  const imagesStrides = util_exports.computeStrides(images.shape);\n  const [newHeight, newWidth] = size;\n  const [batch, oldHeight, oldWidth, numChannels] = images.shape;\n  const xValues = backend2.data.get(images.dataId).values;\n  const output = new Float32Array(batch * newHeight * newWidth * numChannels);\n  const effectiveInputSize = [\n    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n  ];\n  const effectiveOutputSize = [\n    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n  ];\n  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];\n  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];\n  let outputOffset = 0;\n  for (let b = 0; b < batch; b++) {\n    const batchOffset = b * imagesStrides[0];\n    for (let r = 0; r < newHeight; r++) {\n      const sourceFracRow = halfPixelCenters ? effectiveRowSizeRatio * (r + 0.5) : effectiveRowSizeRatio * r;\n      let sourceNearestRow = Math.min(oldHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));\n      if (halfPixelCenters) {\n        sourceNearestRow = Math.max(0, sourceNearestRow);\n      }\n      const rowOffset = batchOffset + sourceNearestRow * imagesStrides[1];\n      for (let c = 0; c < newWidth; c++) {\n        const sourceFracCol = halfPixelCenters ? effectiveColSizeRatio * (c + 0.5) : effectiveColSizeRatio * c;\n        let sourceNearestCol = Math.min(oldWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));\n        if (halfPixelCenters) {\n          sourceNearestCol = Math.max(0, sourceNearestCol);\n        }\n        const colOffset = rowOffset + sourceNearestCol * imagesStrides[2];\n        for (let d = 0; d < numChannels; d++) {\n          const newVal = xValues[colOffset + d];\n          output[outputOffset++] = newVal;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], images.dtype, output);\n}\nvar resizeNearestNeighborConfig = {\n  kernelName: ResizeNearestNeighbor,\n  backendName: \"cpu\",\n  kernelFunc: resizeNearestNeighbor2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeNearestNeighborGrad.js\nfunction resizeNearestNeighborGrad(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images, dy } = inputs;\n  const { alignCorners } = attrs;\n  assertNotComplex([dy, images], \"resizeNearestNeighborGrad\");\n  const imagesStrides = util_exports.computeStrides(images.shape);\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  const [batch, xHeight, xWidth, depth] = images.shape;\n  const [, yHeight, yWidth] = dy.shape;\n  const output = new Float32Array(batch * xHeight * xWidth * depth);\n  const dyValues = backend2.data.get(dy.dataId).values;\n  const effectiveXSize = [\n    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,\n    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth\n  ];\n  const effectiveYSize = [\n    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,\n    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth\n  ];\n  const heightScale = effectiveXSize[0] / effectiveYSize[0];\n  const widthScale = effectiveXSize[1] / effectiveYSize[1];\n  const invHeightScale = 1 / heightScale;\n  const invWidthScale = 1 / widthScale;\n  const winHeight = Math.ceil(invHeightScale) * 2 + 2;\n  const winWidth = Math.ceil(invWidthScale) * 2 + 2;\n  for (let b = 0; b < batch; b++) {\n    const batchOffset = b * imagesStrides[0];\n    for (let r = 0; r < xHeight; r++) {\n      const rowOffset = batchOffset + r * imagesStrides[1];\n      const startRLerp = Math.floor(r * invHeightScale);\n      const startDyR = Math.floor(startRLerp - winHeight / 2);\n      for (let c = 0; c < xWidth; c++) {\n        const colOffset = rowOffset + c * imagesStrides[2];\n        const startCLerp = Math.floor(c * invWidthScale);\n        const startDyC = Math.floor(startCLerp - winWidth / 2);\n        for (let d = 0; d < depth; d++) {\n          let accum = 0;\n          for (let dyRIndex = 0; dyRIndex < winHeight; dyRIndex++) {\n            const dyR = dyRIndex + startDyR;\n            if (dyR < 0 || dyR >= yHeight) {\n              continue;\n            }\n            const dyROffset = batchOffset + dyR * dyStrides[1];\n            const sourceFracRow = dyR * heightScale;\n            const sourceNearestRow = Math.min(xHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));\n            if (r !== sourceNearestRow) {\n              continue;\n            }\n            for (let dyCIndex = 0; dyCIndex < winWidth; dyCIndex++) {\n              const dyC = dyCIndex + startDyC;\n              if (dyC < 0 || dyC >= yWidth) {\n                continue;\n              }\n              const dyCOffset = dyROffset + dyC * dyStrides[2];\n              const sourceFracCol = dyC * widthScale;\n              const sourceNearestCol = Math.min(xWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));\n              if (c === sourceNearestCol) {\n                accum += dyValues[dyCOffset + d];\n              }\n            }\n          }\n          output[colOffset + d] = accum;\n        }\n      }\n    }\n  }\n  return backend2.makeTensorInfo(images.shape, images.dtype, output);\n}\nvar resizeNearestNeighborGradConfig2 = {\n  kernelName: ResizeNearestNeighborGrad,\n  backendName: \"cpu\",\n  kernelFunc: resizeNearestNeighborGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reverse.js\nfunction reverse2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { dims } = attrs;\n  assertNotComplex(x, \"reverse\");\n  const xRank = x.shape.length;\n  const $dims = util_exports.parseAxisParam(dims, x.shape);\n  if (xRank === 0) {\n    return identity2({ inputs: { x }, backend: backend2 });\n  }\n  const outBuf = new TensorBuffer(x.shape, x.dtype);\n  const xBuf = backend2.bufferSync(x);\n  for (let i = 0; i < outBuf.size; i++) {\n    const outLoc = outBuf.indexToLoc(i);\n    const inLoc = outLoc.slice();\n    $dims.forEach((d) => inLoc[d] = x.shape[d] - 1 - inLoc[d]);\n    outBuf.set(xBuf.get(...inLoc), ...outLoc);\n  }\n  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);\n}\nvar reverseConfig = {\n  kernelName: Reverse,\n  backendName: \"cpu\",\n  kernelFunc: reverse2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RotateWithOffset.js\nvar rotateWithOffsetConfig = {\n  kernelName: RotateWithOffset,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { image: image2 } = inputs;\n    const { radians, fillValue, center } = attrs;\n    const cpuBackend = backend2;\n    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));\n    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n    const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);\n    const fullOpacityValue = 255;\n    const sinFactor = Math.sin(radians);\n    const cosFactor = Math.cos(radians);\n    const imageVals = cpuBackend.data.get(image2.dataId).values;\n    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {\n      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;\n      for (let row = 0; row < imageHeight; row++) {\n        const rowOffset = row * (imageWidth * numChannels);\n        for (let col = 0; col < imageWidth; col++) {\n          const colOffset = col * numChannels;\n          for (let channel = 0; channel < numChannels; channel++) {\n            const coords2 = [batch, row, col, channel];\n            const x = coords2[2];\n            const y = coords2[1];\n            let coordX = (x - centerX) * cosFactor - (y - centerY) * sinFactor;\n            let coordY = (x - centerX) * sinFactor + (y - centerY) * cosFactor;\n            coordX = Math.round(coordX + centerX);\n            coordY = Math.round(coordY + centerY);\n            let outputValue = fillValue;\n            if (typeof fillValue !== \"number\") {\n              if (channel === 3) {\n                outputValue = fullOpacityValue;\n              } else {\n                outputValue = fillValue[channel];\n              }\n            }\n            if (coordX >= 0 && coordX < imageWidth && coordY >= 0 && coordY < imageHeight) {\n              const rotatedRowOffset = coordY * (imageWidth * numChannels);\n              const rotatedColOffset = coordX * numChannels;\n              const imageIdx = batchOffset + rotatedRowOffset + rotatedColOffset + channel;\n              outputValue = imageVals[imageIdx];\n            }\n            const outIdx = batchOffset + rowOffset + colOffset + channel;\n            output[outIdx] = outputValue;\n          }\n        }\n      }\n    }\n    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);\n    return { dataId, shape: image2.shape, dtype: image2.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Round.js\nvar round3 = unaryKernelFunc(Round, (xi) => {\n  const base = Math.floor(xi);\n  if (xi - base < 0.5) {\n    return Math.floor(xi);\n  } else if (xi - base > 0.5) {\n    return Math.ceil(xi);\n  } else {\n    if (base % 2 === 0) {\n      return base;\n    } else {\n      return base + 1;\n    }\n  }\n});\nvar roundConfig = {\n  kernelName: Round,\n  backendName: \"cpu\",\n  kernelFunc: round3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ScatterNd.js\nfunction scatterNd(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { indices, updates } = inputs;\n  const { shape } = attrs;\n  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);\n  const sumDupeIndices = true;\n  const indicesBuf = backend2.bufferSync(indices);\n  const updatesBuf = backend2.bufferSync(updates);\n  const outBuf = scatterImpl(indicesBuf, updatesBuf, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, 0, sumDupeIndices);\n  return backend2.makeTensorInfo(shape, outBuf.dtype, outBuf.values);\n}\nvar scatterNdConfig = {\n  kernelName: ScatterNd,\n  backendName: \"cpu\",\n  kernelFunc: scatterNd\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SearchSorted_impl.js\nfunction lowerBound2(array2, value) {\n  let left = 0;\n  let right = array2.length;\n  let mid = 0;\n  while (left < right) {\n    mid = Math.floor((left + right) / 2);\n    if (array2[mid] < value) {\n      left = mid + 1;\n    } else {\n      right = mid;\n    }\n  }\n  return right;\n}\nfunction upperBound2(array2, value) {\n  let left = 0;\n  let right = array2.length;\n  let mid = 0;\n  while (left < right) {\n    mid = Math.floor((left + right) / 2);\n    if (array2[mid] <= value) {\n      left = mid + 1;\n    } else {\n      right = mid;\n    }\n  }\n  return right;\n}\nfunction searchSortedImpl(sortedInputs, values, batchSize, numInputs, numValues, side) {\n  const output = util_exports.getArrayFromDType(\"int32\", batchSize * numValues);\n  for (let b = 0; b < batchSize; ++b) {\n    const sortedInputsSlice = sortedInputs.slice(b * numInputs, (b + 1) * numInputs);\n    const outputOffset = b * numValues;\n    for (let i = 0; i < numValues; ++i) {\n      output[outputOffset + i] = side === \"left\" ? lowerBound2(sortedInputsSlice, values[i + outputOffset]) : upperBound2(sortedInputsSlice, values[i + outputOffset]);\n    }\n  }\n  return output;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SearchSorted.js\nfunction searchSorted2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { sortedSequence, values } = inputs;\n  const { side } = attrs;\n  const $sortedSequence = backend2.data.get(sortedSequence.dataId).values;\n  const $values = backend2.data.get(values.dataId).values;\n  const output = searchSortedImpl($sortedSequence, $values, sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);\n  return backend2.makeTensorInfo(values.shape, \"int32\", output);\n}\nvar searchSortedConfig = {\n  kernelName: SearchSorted,\n  backendName: \"cpu\",\n  kernelFunc: searchSorted2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Select.js\nfunction select2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { condition, t, e } = inputs;\n  assertNotComplex([condition, t, e], \"select\");\n  const conditionRank = condition.shape.length;\n  const values = backend2.data.get(condition.dataId).values;\n  const tValues = backend2.data.get(t.dataId).values;\n  const eValues = backend2.data.get(e.dataId).values;\n  const resultDtype = upcastType(t.dtype, e.dtype);\n  const newValues = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(t.shape), resultDtype);\n  let index = 0;\n  const offset = conditionRank === 0 || conditionRank > 1 || t.shape.length === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));\n  for (let i = 0; i < values.length; i++) {\n    for (let j = 0; j < offset; j++) {\n      if (values[i] === 1) {\n        newValues[index++] = tValues[i];\n      } else {\n        newValues[index++] = eValues[i];\n      }\n    }\n  }\n  return backend2.makeTensorInfo(t.shape, resultDtype, newValues);\n}\nvar selectConfig = {\n  kernelName: Select,\n  backendName: \"cpu\",\n  kernelFunc: select2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Selu.js\nvar scaleAlpha = backend_util_exports.SELU_SCALEALPHA;\nvar scale = backend_util_exports.SELU_SCALE;\nvar selu2 = unaryKernelFunc(Selu, (xi) => {\n  if (xi >= 0) {\n    return scale * xi;\n  } else {\n    return scaleAlpha * (Math.exp(xi) - 1);\n  }\n});\nvar seluConfig = {\n  kernelName: Selu,\n  backendName: \"cpu\",\n  kernelFunc: selu2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sign.js\nvar sign2 = unaryKernelFunc(Sign, (xi) => {\n  if (xi < 0) {\n    return -1;\n  } else if (xi > 0) {\n    return 1;\n  } else {\n    return 0;\n  }\n});\nvar signConfig = {\n  kernelName: Sign,\n  backendName: \"cpu\",\n  kernelFunc: sign2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sin.js\nvar sin2 = unaryKernelFunc(Sin, (xi) => Math.sin(xi));\nvar sinConfig = {\n  kernelName: Sin,\n  backendName: \"cpu\",\n  kernelFunc: sin2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sinh.js\nvar sinh2 = unaryKernelFunc(Sinh, (xi) => Math.sinh(xi));\nvar sinhConfig = {\n  kernelName: Sinh,\n  backendName: \"cpu\",\n  kernelFunc: sinh2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Softplus.js\nvar epsilon2 = 11920928955078125e-23;\nvar threshold2 = Math.log(epsilon2) + 2;\nvar softplus2 = unaryKernelFunc(Softplus, (xi) => {\n  const tooLarge = xi > -threshold2;\n  const tooSmall = xi < threshold2;\n  const expX = Math.exp(xi);\n  let result;\n  if (tooSmall) {\n    result = expX;\n  } else if (tooLarge) {\n    result = xi;\n  } else {\n    result = Math.log(1 + expX);\n  }\n  return result;\n});\nvar softplusConfig = {\n  kernelName: Softplus,\n  backendName: \"cpu\",\n  kernelFunc: softplus2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SpaceToBatchND.js\nfunction spaceToBatchND2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, paddings } = attrs;\n  assertNotComplex([x], \"spaceToBatchND\");\n  const prod5 = util_exports.sizeFromShape(blockShape);\n  const completePaddings = [[0, 0]];\n  completePaddings.push(...paddings);\n  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {\n    completePaddings.push([0, 0]);\n  }\n  const paddedX = padV2Config.kernelFunc({\n    inputs: { x },\n    backend: backend2,\n    attrs: { paddings: completePaddings, constantValue: 0 }\n  });\n  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);\n  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);\n  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);\n  const reshapeInputs = { x: paddedX };\n  const reshapeAttrs = { shape: reshapedPaddedShape };\n  const paddedXReshaped = reshape3({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });\n  const transposeInputs = { x: paddedXReshaped };\n  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };\n  const paddedXT = transpose2({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });\n  const resultReshapeInputs = { x: paddedXT };\n  const resultReshapeAttrs = { shape: flattenShape };\n  const result = reshape3({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });\n  backend2.disposeIntermediateTensorInfo(paddedX);\n  backend2.disposeIntermediateTensorInfo(paddedXReshaped);\n  backend2.disposeIntermediateTensorInfo(paddedXT);\n  return result;\n}\nvar spaceToBatchNDConfig = {\n  kernelName: SpaceToBatchND,\n  backendName: \"cpu\",\n  kernelFunc: spaceToBatchND2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseFillEmptyRows.js\nfunction sparseFillEmptyRows2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { indices, values, denseShape, defaultValue } = inputs;\n  if (denseShape.shape.length !== 1) {\n    throw new Error(`Dense shape must be a vector, saw:\n        ${denseShape.shape}`);\n  }\n  if (indices.shape.length !== 2) {\n    throw new Error(`Indices must be a matrix, saw:\n        ${indices.shape}`);\n  }\n  if (values.shape.length !== 1) {\n    throw new Error(`Values must be a vector, saw:\n        ${values.shape}`);\n  }\n  if (defaultValue.shape.length !== 0) {\n    throw new Error(`Default value must be a scalar, saw:\n        ${defaultValue.shape}`);\n  }\n  const $indices = backend2.data.get(indices.dataId).values;\n  const $values = backend2.data.get(values.dataId).values;\n  const $denseShape = backend2.data.get(denseShape.dataId).values;\n  const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];\n  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImpl($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);\n  return [\n    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),\n    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),\n    backend2.makeTensorInfo([emptyRowIndicator.length], \"bool\", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),\n    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))\n  ];\n}\nvar sparseFillEmptyRowsConfig = {\n  kernelName: SparseFillEmptyRows,\n  backendName: \"cpu\",\n  kernelFunc: sparseFillEmptyRows2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseReshape.js\nfunction sparseReshape2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { inputIndices, inputShape, newShape } = inputs;\n  if (inputIndices.shape.length !== 2) {\n    throw new Error(`Input indices should be a matrix but received shape\n        ${inputIndices.shape}`);\n  }\n  if (inputShape.shape.length !== 1) {\n    throw new Error(`Input shape should be a vector but received shape\n        ${inputShape.shape}`);\n  }\n  if (newShape.shape.length !== 1) {\n    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);\n  }\n  const $inputShape = Array.from(backend2.data.get(inputShape.dataId).values);\n  const $inputIndices = backend2.data.get(inputIndices.dataId).values;\n  const targetShape = Array.from(backend2.data.get(newShape.dataId).values);\n  const [newIndices, indicesShape, outputShape] = sparseReshapeImpl($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);\n  return [\n    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),\n    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))\n  ];\n}\nvar sparseReshapeConfig = {\n  kernelName: SparseReshape,\n  backendName: \"cpu\",\n  kernelFunc: sparseReshape2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentMean.js\nfunction sparseSegmentMean2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { data, indices, segmentIds } = inputs;\n  if (data.shape.length < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if (indices.shape.length !== 1) {\n    throw new Error(`Indices should be a vector but received shape\n          ${indices.shape}`);\n  }\n  if (segmentIds.shape.length !== 1) {\n    throw new Error(`Segment ids should be a vector but received shape\n          ${segmentIds.shape}`);\n  }\n  if (indices.shape[0] !== segmentIds.shape[0]) {\n    throw new Error(`segmentIds and indices should have same size.`);\n  }\n  const $data = backend2.data.get(data.dataId).values;\n  const $indices = backend2.data.get(indices.dataId).values;\n  const $segmentIds = backend2.data.get(segmentIds.dataId).values;\n  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds, true);\n  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);\n}\nvar sparseSegmentMeanConfig = {\n  kernelName: SparseSegmentMean,\n  backendName: \"cpu\",\n  kernelFunc: sparseSegmentMean2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentSum.js\nfunction sparseSegmentSum2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { data, indices, segmentIds } = inputs;\n  if (data.shape.length < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if (indices.shape.length !== 1) {\n    throw new Error(`Indices should be a vector but received shape\n         ${indices.shape}`);\n  }\n  if (segmentIds.shape.length !== 1) {\n    throw new Error(`Segment ids should be a vector but received shape\n         ${segmentIds.shape}`);\n  }\n  if (indices.shape[0] !== segmentIds.shape[0]) {\n    throw new Error(`segmentIds and indices should have same size.`);\n  }\n  const $data = backend2.data.get(data.dataId).values;\n  const $indices = backend2.data.get(indices.dataId).values;\n  const $segmentIds = backend2.data.get(segmentIds.dataId).values;\n  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds);\n  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);\n}\nvar sparseSegmentSumConfig = {\n  kernelName: SparseSegmentSum,\n  backendName: \"cpu\",\n  kernelFunc: sparseSegmentSum2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseToDense.js\nfunction sparseToDense2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { sparseIndices, sparseValues, defaultValue } = inputs;\n  const { outputShape } = attrs;\n  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);\n  const sumDupeIndices = false;\n  const indicesBuf = backend2.bufferSync(sparseIndices);\n  let outBuf;\n  switch (sparseValues.dtype) {\n    case \"bool\": {\n      const updatesBuf = backend2.bufferSync(sparseValues);\n      const $defaultValue = Boolean(backend2.data.get(defaultValue.dataId).values[0]);\n      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);\n      break;\n    }\n    case \"float32\": {\n      const updatesBuf = backend2.bufferSync(sparseValues);\n      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];\n      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);\n      break;\n    }\n    case \"int32\": {\n      const updatesBuf = backend2.bufferSync(sparseValues);\n      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];\n      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);\n      break;\n    }\n    case \"string\": {\n      const updatesBuf = backend2.bufferSync(sparseValues);\n      const $defaultValue = util_exports.decodeString(backend2.data.get(defaultValue.dataId).values[0]);\n      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);\n      break;\n    }\n    default:\n      throw new Error(`Unsupported type ${sparseValues.dtype}`);\n  }\n  return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);\n}\nvar sparseToDenseConfig = {\n  kernelName: SparseToDense,\n  backendName: \"cpu\",\n  kernelFunc: sparseToDense2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SplitV.js\nfunction splitV(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { numOrSizeSplits, axis } = attrs;\n  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];\n  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);\n  const begin = new Array(x.shape.length).fill(0);\n  const size = x.shape.slice();\n  return splitSizes.map((s) => {\n    const sliceSize = [...size];\n    sliceSize[$axis] = s;\n    const sliceT = slice2({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });\n    begin[$axis] += s;\n    return sliceT;\n  });\n}\nvar splitVConfig = {\n  kernelName: SplitV,\n  backendName: \"cpu\",\n  kernelFunc: splitV\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Square.js\nvar squareConfig = {\n  kernelName: Square,\n  backendName: \"cpu\",\n  kernelFunc: ({ inputs, backend: backend2 }) => {\n    const { x } = inputs;\n    const cpuBackend = backend2;\n    assertNotComplex(x, \"square\");\n    const values = cpuBackend.data.get(x.dataId).values;\n    const newValues = new Float32Array(values.length);\n    for (let i = 0; i < values.length; ++i) {\n      const value = values[i];\n      newValues[i] = value * value;\n    }\n    const dataId = cpuBackend.write(newValues, x.shape, x.dtype);\n    return { dataId, shape: x.shape, dtype: x.dtype };\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Step.js\nvar step2 = unaryKernelFunc(Step, (xi, attrs) => {\n  const stepAttrs = attrs;\n  if (isNaN(xi)) {\n    return NaN;\n  } else {\n    return xi > 0 ? 1 : stepAttrs.alpha;\n  }\n});\nvar stepConfig = {\n  kernelName: Step,\n  backendName: \"cpu\",\n  kernelFunc: step2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StridedSlice.js\nfunction stridedSlice2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;\n  assertNotComplex(x, \"stridedSlice\");\n  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);\n  let result;\n  if (isIdentity) {\n    result = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });\n  } else if (sliceDim0 || isSimpleSlice) {\n    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);\n    const size = slice_util_exports.computeOutShape($begin, $end, $strides);\n    const sliced = slice2({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });\n    result = reshape3({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });\n    backend2.disposeIntermediateTensorInfo(sliced);\n  } else {\n    const xBuf = backend2.bufferSync(x);\n    const outBuf = stridedSliceImpl(finalShapeSparse, xBuf, $strides, $begin);\n    result = backend2.makeTensorInfo(finalShape, outBuf.dtype, outBuf.values);\n  }\n  return result;\n}\nvar stridedSliceConfig = {\n  kernelName: StridedSlice,\n  backendName: \"cpu\",\n  kernelFunc: stridedSlice2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringNGrams.js\nfunction stringNGrams2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;\n  const { data, dataSplits } = inputs;\n  const $data = backend2.data.get(data.dataId).values;\n  const $dataSplits = backend2.data.get(dataSplits.dataId).values;\n  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);\n  return [\n    backend2.makeTensorInfo([nGrams.length], \"string\", nGrams),\n    backend2.makeTensorInfo(dataSplits.shape, \"int32\", nGramsSplits)\n  ];\n}\nvar stringNGramsConfig = {\n  kernelName: StringNGrams,\n  backendName: \"cpu\",\n  kernelFunc: stringNGrams2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringSplit.js\nfunction stringSplit2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { skipEmpty } = attrs;\n  const { input: input2, delimiter } = inputs;\n  if (input2.dtype !== \"string\") {\n    throw new Error(\"Input must be of datatype string\");\n  }\n  if (input2.shape.length !== 1) {\n    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);\n  }\n  if (delimiter.shape.length !== 0) {\n    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);\n  }\n  const $input = backend2.data.get(input2.dataId).values;\n  const $delimiter = backend2.data.get(delimiter.dataId).values[0];\n  const [indices, values, shape] = stringSplitImpl($input, $delimiter, skipEmpty);\n  const outputSize = values.length;\n  return [\n    backend2.makeTensorInfo([outputSize, 2], \"int32\", indices),\n    backend2.makeTensorInfo([outputSize], \"string\", values),\n    backend2.makeTensorInfo([2], \"int32\", new Int32Array(shape))\n  ];\n}\nvar stringSplitConfig = {\n  kernelName: StringSplit,\n  backendName: \"cpu\",\n  kernelFunc: stringSplit2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringToHashBucketFast.js\nfunction stringToHashBucketFast2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { numBuckets } = attrs;\n  const { input: input2 } = inputs;\n  if (input2.dtype !== \"string\") {\n    throw new Error(\"Input must be of datatype string\");\n  }\n  if (numBuckets <= 0) {\n    throw new Error(`Number of buckets must be at least 1`);\n  }\n  const $input = backend2.data.get(input2.dataId).values;\n  const output = stringToHashBucketFastImpl($input, numBuckets);\n  return backend2.makeTensorInfo(input2.shape, \"int32\", output);\n}\nvar stringToHashBucketFastConfig = {\n  kernelName: StringToHashBucketFast,\n  backendName: \"cpu\",\n  kernelFunc: stringToHashBucketFast2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tan.js\nvar tan2 = unaryKernelFunc(Tan, (xi) => Math.tan(xi));\nvar tanConfig = {\n  kernelName: Tan,\n  backendName: \"cpu\",\n  kernelFunc: tan2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tanh.js\nvar tanh3 = unaryKernelFunc(Tanh, (xi) => Math.tanh(xi));\nvar tanhConfig = {\n  kernelName: Tanh,\n  backendName: \"cpu\",\n  kernelFunc: tanh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tile.js\nfunction tile3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { reps } = attrs;\n  assertNotComplex(x, \"tile\");\n  const outBuf = tileImpl(backend2.bufferSync(x), reps);\n  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);\n}\nvar tileConfig = {\n  kernelName: Tile,\n  backendName: \"cpu\",\n  kernelFunc: tile3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/TopK.js\nfunction topK(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { k, sorted } = attrs;\n  assertNotComplex(x, \"topk\");\n  const xVals = backend2.data.get(x.dataId).values;\n  const [allTopKVals, allTopKIndices] = topKImpl(xVals, x.shape, x.dtype, k, sorted);\n  return [\n    backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),\n    backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)\n  ];\n}\nvar topKConfig = {\n  kernelName: TopK,\n  backendName: \"cpu\",\n  kernelFunc: topK\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transform.js\nfunction transform2(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { image: image2, transforms } = inputs;\n  const { interpolation, fillMode, fillValue, outputShape } = attrs;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];\n  const outShape = [batch, outHeight, outWidth, numChannels];\n  const inStrides = util_exports.computeStrides(image2.shape);\n  const batchInStride = inStrides[0];\n  const rowInStride = inStrides[1];\n  const colInStride = inStrides[2];\n  const outStrides = util_exports.computeStrides(outShape);\n  const batchOutStride = outStrides[0];\n  const rowOutStride = outStrides[1];\n  const colOutStride = outStrides[2];\n  const outVals = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(outShape));\n  outVals.fill(fillValue);\n  const imageVals = backend2.data.get(image2.dataId).values;\n  const transformVals = backend2.data.get(transforms.dataId).values;\n  for (let b = 0; b < batch; ++b) {\n    const transform5 = transforms.shape[0] === 1 ? transformVals : transformVals.subarray(b * 8, b * 8 + 8);\n    for (let outY = 0; outY < outHeight; ++outY) {\n      for (let outX = 0; outX < outWidth; ++outX) {\n        for (let channel = 0; channel < numChannels; ++channel) {\n          let val;\n          const projection = transform5[6] * outX + transform5[7] * outY + 1;\n          if (projection === 0) {\n            continue;\n          }\n          const inX = (transform5[0] * outX + transform5[1] * outY + transform5[2]) / projection;\n          const inY = (transform5[3] * outX + transform5[4] * outY + transform5[5]) / projection;\n          const x = mapCoord(inX, imageWidth, fillMode);\n          const y = mapCoord(inY, imageHeight, fillMode);\n          switch (interpolation) {\n            case \"nearest\":\n              val = nearestInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);\n              break;\n            case \"bilinear\":\n              val = bilinearInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);\n              break;\n            default:\n              throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${interpolation}`);\n          }\n          const ind = b * batchOutStride + outY * rowOutStride + outX * colOutStride + channel;\n          outVals[ind] = val;\n        }\n      }\n    }\n    return backend2.makeTensorInfo(outShape, image2.dtype, outVals);\n  }\n  const dataId = backend2.write(outVals, outShape, image2.dtype);\n  return { dataId, shape: image2.shape, dtype: image2.dtype };\n}\nvar transformConfig = {\n  kernelName: Transform,\n  backendName: \"cpu\",\n  kernelFunc: transform2\n};\nfunction mapCoord(outCoord, len, mode) {\n  switch (mode) {\n    case \"reflect\":\n      return mapCoordReflect(outCoord, len);\n    case \"wrap\":\n      return mapCoordWrap(outCoord, len);\n    case \"nearest\":\n      return mapCoordNearest(outCoord, len);\n    case \"constant\":\n    default:\n      return mapCoordConstant(outCoord, len);\n  }\n}\nfunction mapCoordReflect(outCoord, len) {\n  let inCoord = outCoord;\n  if (inCoord < 0) {\n    if (len <= 1) {\n      inCoord = 0;\n    } else {\n      const sz2 = 2 * len;\n      if (inCoord < sz2) {\n        inCoord = sz2 * Math.trunc(-inCoord / sz2) + inCoord;\n      }\n      inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1;\n    }\n  } else if (inCoord > len - 1) {\n    if (len <= 1) {\n      inCoord = 0;\n    } else {\n      const sz2 = 2 * len;\n      inCoord -= sz2 * Math.trunc(inCoord / sz2);\n      if (inCoord >= len) {\n        inCoord = sz2 - inCoord - 1;\n      }\n    }\n  }\n  return util_exports.clamp(0, inCoord, len - 1);\n}\nfunction mapCoordWrap(outCoord, len) {\n  let inCoord = outCoord;\n  if (inCoord < 0) {\n    if (len <= 1) {\n      inCoord = 0;\n    } else {\n      const sz = len - 1;\n      inCoord += len * (Math.trunc(-inCoord / sz) + 1);\n    }\n  } else if (inCoord > len - 1) {\n    if (len <= 1) {\n      inCoord = 0;\n    } else {\n      const sz = len - 1;\n      inCoord -= len * Math.trunc(inCoord / sz);\n    }\n  }\n  return util_exports.clamp(0, inCoord, len - 1);\n}\nfunction mapCoordConstant(outCoord, len) {\n  return outCoord;\n}\nfunction mapCoordNearest(outCoord, len) {\n  return util_exports.clamp(0, outCoord, len - 1);\n}\nfunction readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {\n  const ind = batch * batchStride + y * rowStride + x * colStride + channel;\n  if (0 <= y && y < imageHeight && 0 <= x && x < imageWidth) {\n    return imageVals[ind];\n  } else {\n    return fillValue;\n  }\n}\nfunction nearestInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {\n  const $y = Math.round(y);\n  const $x = Math.round(x);\n  return readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, $y, $x, channel, fillValue);\n}\nfunction bilinearInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {\n  const yFloor = Math.floor(y);\n  const xFloor = Math.floor(x);\n  const yCeil = yFloor + 1;\n  const xCeil = xFloor + 1;\n  const valueYFloor = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xCeil, channel, fillValue);\n  const valueYCeil = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xCeil, channel, fillValue);\n  return (yCeil - y) * valueYFloor + (y - yFloor) * valueYCeil;\n}\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unique.js\nfunction unique3(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { axis } = attrs;\n  const { x } = inputs;\n  assertNotComplex(x, \"unique\");\n  const values = backend2.data.get(x.dataId).values;\n  const { outputValues, outputShape, indices } = uniqueImpl(values, axis, x.shape, x.dtype);\n  return [\n    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),\n    backend2.makeTensorInfo([indices.length], \"int32\", indices)\n  ];\n}\nvar uniqueConfig = {\n  kernelName: Unique,\n  backendName: \"cpu\",\n  kernelFunc: unique3\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unpack.js\nfunction unpack(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { value } = inputs;\n  let { axis } = attrs;\n  if (axis < 0) {\n    axis += value.shape.length;\n  }\n  const valueRank = value.shape.length;\n  const num = value.shape[axis];\n  const outShape = new Array(valueRank - 1);\n  let outIndex = 0;\n  for (let i = 0; i < valueRank; i++) {\n    if (i !== axis) {\n      outShape[outIndex++] = value.shape[i];\n    }\n  }\n  const begin = new Array(valueRank).fill(0);\n  const size = value.shape.slice();\n  size[axis] = 1;\n  const res = new Array(num);\n  for (let i = 0; i < res.length; i++) {\n    begin[axis] = i;\n    const tempRes = slice2({ inputs: { x: value }, backend: backend2, attrs: { begin, size } });\n    res[i] = reshape3({ inputs: { x: tempRes }, backend: backend2, attrs: { shape: outShape } });\n    backend2.disposeIntermediateTensorInfo(tempRes);\n  }\n  return res;\n}\nvar unpackConfig = {\n  kernelName: Unpack,\n  backendName: \"cpu\",\n  kernelFunc: unpack\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/UnsortedSegmentSum.js\nfunction unsortedSegmentSum2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, segmentIds } = inputs;\n  const { numSegments } = attrs;\n  assertNotComplex(x, \"unsortedSegmentSum\");\n  const xRank = x.shape.length;\n  const segmentIdsRank = segmentIds.shape.length;\n  const res = [];\n  const intermediates = [];\n  const numIters = xRank - segmentIdsRank;\n  let $segmentIds = segmentIds;\n  for (let i = 0; i < numIters; ++i) {\n    const expanded = expandDims3({ inputs: { input: $segmentIds }, backend: backend2, attrs: { dim: i + 1 } });\n    $segmentIds = expanded;\n    intermediates.push(expanded);\n  }\n  for (let i = 0; i < numSegments; ++i) {\n    const scalarValue = util_exports.createScalarValue(i, \"int32\");\n    const segmentId = backend2.makeTensorInfo([], \"int32\", scalarValue);\n    const mask = equal2({ inputs: { a: segmentId, b: $segmentIds }, backend: backend2 });\n    const maskCasted = cast3({ inputs: { x: mask }, backend: backend2, attrs: { dtype: \"float32\" } });\n    const mul2 = multiply2({ inputs: { a: maskCasted, b: x }, backend: backend2 });\n    const sumTensorInfo = sum3({ inputs: { x: mul2 }, backend: backend2, attrs: { axis: 0, keepDims: false } });\n    res.push(sumTensorInfo);\n    intermediates.push(segmentId);\n    intermediates.push(mask);\n    intermediates.push(maskCasted);\n    intermediates.push(mul2);\n    intermediates.push(sumTensorInfo);\n  }\n  const result = pack({ inputs: res, backend: backend2, attrs: { axis: 0 } });\n  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar unsortedSegmentSumConfig = {\n  kernelName: UnsortedSegmentSum,\n  backendName: \"cpu\",\n  kernelFunc: unsortedSegmentSum2\n};\n\n// node_modules/@tensorflow/tfjs-backend-cpu/dist/register_all_kernels.js\nvar kernelConfigs = [\n  _fusedMatMulConfig,\n  absConfig,\n  acosConfig,\n  acoshConfig,\n  addConfig,\n  addNConfig,\n  allConfig,\n  anyConfig,\n  argMaxConfig,\n  argMinConfig,\n  asinConfig,\n  asinhConfig,\n  atanConfig,\n  atan2Config,\n  atanhConfig,\n  avgPoolConfig,\n  avgPool3DConfig,\n  avgPool3DGradConfig2,\n  avgPoolGradConfig2,\n  batchMatMulConfig,\n  batchNormConfig,\n  batchToSpaceNDConfig,\n  bincountConfig,\n  broadcastArgsConfig,\n  castConfig,\n  ceilConfig,\n  clipByValueConfig,\n  complexConfig,\n  complexAbsConfig,\n  concatConfig,\n  conv2DConfig,\n  conv2DBackpropFilterConfig,\n  conv2DBackpropInputConfig,\n  conv3DConfig,\n  conv3DBackpropFilterV2Config,\n  conv3DBackpropInputV2Config,\n  cosConfig,\n  coshConfig,\n  cropAndResizeConfig,\n  cumprodConfig,\n  cumsumConfig,\n  denseBincountConfig,\n  depthToSpaceConfig,\n  depthwiseConv2dNativeConfig,\n  depthwiseConv2dNativeBackpropFilterConfig,\n  depthwiseConv2dNativeBackpropInputConfig,\n  diagConfig,\n  dilation2DConfig,\n  dilation2DBackpropFilterConfig,\n  dilation2DBackpropInputConfig,\n  einsumConfig,\n  eluConfig,\n  eluGradConfig2,\n  equalConfig,\n  erfConfig,\n  expConfig,\n  expandDimsConfig,\n  expm1Config,\n  fftConfig,\n  fillConfig,\n  flipLeftRightConfig,\n  floorConfig,\n  floorDivConfig,\n  fusedConv2DConfig,\n  fusedDepthwiseConv2DConfig,\n  gatherNdConfig,\n  gatherV2Config,\n  greaterConfig,\n  greaterEqualConfig,\n  identityConfig,\n  ifftConfig,\n  imagConfig,\n  isFiniteConfig,\n  isInfConfig,\n  isNaNConfig,\n  leakyReluConfig,\n  lessConfig,\n  lessEqualConfig,\n  linSpaceConfig,\n  logConfig,\n  log1pConfig,\n  logicalAndConfig,\n  logicalNotConfig,\n  logicalOrConfig,\n  LRNConfig,\n  LRNGradConfig,\n  maxConfig,\n  maximumConfig,\n  maxPoolConfig,\n  maxPool3DConfig,\n  maxPool3DGradConfig2,\n  maxPoolGradConfig2,\n  maxPoolWithArgmaxConfig,\n  meanConfig,\n  minConfig,\n  minimumConfig,\n  mirrorPadConfig,\n  modConfig,\n  multinomialConfig,\n  multiplyConfig,\n  negConfig,\n  nonMaxSuppressionV3Config,\n  nonMaxSuppressionV4Config,\n  nonMaxSuppressionV5Config,\n  notEqualConfig,\n  oneHotConfig,\n  onesLikeConfig,\n  packConfig,\n  padV2Config,\n  powConfig,\n  preluConfig,\n  prodConfig,\n  raggedGatherConfig,\n  raggedRangeConfig,\n  raggedTensorToTensorConfig,\n  rangeConfig,\n  realConfig,\n  realDivConfig,\n  reciprocalConfig,\n  reluConfig,\n  relu6Config,\n  reshapeConfig,\n  resizeBilinearConfig,\n  resizeBilinearGradConfig2,\n  resizeNearestNeighborConfig,\n  resizeNearestNeighborGradConfig2,\n  reverseConfig,\n  rotateWithOffsetConfig,\n  roundConfig,\n  rsqrtConfig,\n  scatterNdConfig,\n  searchSortedConfig,\n  selectConfig,\n  seluConfig,\n  sigmoidConfig,\n  signConfig,\n  sinConfig,\n  sinhConfig,\n  sliceConfig,\n  softmaxConfig,\n  softplusConfig,\n  spaceToBatchNDConfig,\n  sparseFillEmptyRowsConfig,\n  sparseReshapeConfig,\n  sparseSegmentMeanConfig,\n  sparseSegmentSumConfig,\n  sparseToDenseConfig,\n  splitVConfig,\n  sqrtConfig,\n  squareConfig,\n  squaredDifferenceConfig,\n  stepConfig,\n  stridedSliceConfig,\n  stringNGramsConfig,\n  stringSplitConfig,\n  stringToHashBucketFastConfig,\n  subConfig,\n  sumConfig,\n  tanConfig,\n  tanhConfig,\n  tileConfig,\n  topKConfig,\n  transformConfig,\n  transposeConfig,\n  uniqueConfig,\n  unpackConfig,\n  unsortedSegmentSumConfig,\n  zerosLikeConfig\n];\nfor (const kernelConfig of kernelConfigs) {\n  registerKernel(kernelConfig);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl_util.js\nvar webgl_util_exports = {};\n__export(webgl_util_exports, {\n  assertNotComplex: () => assertNotComplex2,\n  bindCanvasToFramebuffer: () => bindCanvasToFramebuffer,\n  bindColorTextureToFramebuffer: () => bindColorTextureToFramebuffer,\n  bindTextureToProgramUniformSampler: () => bindTextureToProgramUniformSampler,\n  bindTextureUnit: () => bindTextureUnit,\n  bindVertexBufferToProgramAttribute: () => bindVertexBufferToProgramAttribute,\n  callAndCheck: () => callAndCheck,\n  canBeRepresented: () => canBeRepresented,\n  createFragmentShader: () => createFragmentShader,\n  createFramebuffer: () => createFramebuffer,\n  createProgram: () => createProgram,\n  createStaticIndexBuffer: () => createStaticIndexBuffer,\n  createStaticVertexBuffer: () => createStaticVertexBuffer,\n  createTexture: () => createTexture,\n  createVertexShader: () => createVertexShader,\n  getBatchDim: () => getBatchDim,\n  getExtensionOrThrow: () => getExtensionOrThrow,\n  getFramebufferErrorMessage: () => getFramebufferErrorMessage,\n  getMaxTexturesInShader: () => getMaxTexturesInShader,\n  getNumChannels: () => getNumChannels,\n  getProgramUniformLocation: () => getProgramUniformLocation,\n  getProgramUniformLocationOrThrow: () => getProgramUniformLocationOrThrow,\n  getRowsCols: () => getRowsCols,\n  getShapeAs3D: () => getShapeAs3D,\n  getTextureShapeFromLogicalShape: () => getTextureShapeFromLogicalShape,\n  getWebGLDisjointQueryTimerVersion: () => getWebGLDisjointQueryTimerVersion,\n  getWebGLErrorMessage: () => getWebGLErrorMessage,\n  getWebGLMaxTextureSize: () => getWebGLMaxTextureSize,\n  hasExtension: () => hasExtension,\n  isCapableOfRenderingToFloatTexture: () => isCapableOfRenderingToFloatTexture,\n  isDownloadFloatTextureEnabled: () => isDownloadFloatTextureEnabled,\n  isReshapeFree: () => isReshapeFree,\n  isWebGLFenceEnabled: () => isWebGLFenceEnabled,\n  isWebGLVersionEnabled: () => isWebGLVersionEnabled,\n  linkProgram: () => linkProgram,\n  logShaderSourceAndInfoLog: () => logShaderSourceAndInfoLog,\n  resetMaxTextureSize: () => resetMaxTextureSize,\n  resetMaxTexturesInShader: () => resetMaxTexturesInShader,\n  unbindColorTextureFromFramebuffer: () => unbindColorTextureFromFramebuffer,\n  unbindTextureUnit: () => unbindTextureUnit,\n  validateFramebuffer: () => validateFramebuffer,\n  validateProgram: () => validateProgram,\n  validateTextureSize: () => validateTextureSize\n});\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/canvas_util.js\nvar contexts = {};\nvar WEBGL_ATTRIBUTES = {\n  alpha: false,\n  antialias: false,\n  premultipliedAlpha: false,\n  preserveDrawingBuffer: false,\n  depth: false,\n  stencil: false,\n  failIfMajorPerformanceCaveat: true\n};\nfunction setWebGLContext(webGLVersion, gl) {\n  contexts[webGLVersion] = gl;\n}\nfunction getWebGLContext(webGLVersion, customCanvas) {\n  if (!(webGLVersion in contexts) || customCanvas != null) {\n    const newCtx = getWebGLRenderingContext(webGLVersion, customCanvas);\n    if (newCtx !== null) {\n      contexts[webGLVersion] = newCtx;\n    } else {\n      console.log(\"Could not get context for WebGL version\", webGLVersion);\n      return null;\n    }\n  }\n  const gl = contexts[webGLVersion];\n  if (gl == null || gl.isContextLost()) {\n    delete contexts[webGLVersion];\n    return getWebGLContext(webGLVersion);\n  }\n  gl.disable(gl.DEPTH_TEST);\n  gl.disable(gl.STENCIL_TEST);\n  gl.disable(gl.BLEND);\n  gl.disable(gl.DITHER);\n  gl.disable(gl.POLYGON_OFFSET_FILL);\n  gl.disable(gl.SAMPLE_COVERAGE);\n  gl.enable(gl.SCISSOR_TEST);\n  gl.enable(gl.CULL_FACE);\n  gl.cullFace(gl.BACK);\n  return contexts[webGLVersion];\n}\nfunction createCanvas(webGLVersion) {\n  if (typeof OffscreenCanvas !== \"undefined\" && webGLVersion === 2) {\n    return new OffscreenCanvas(300, 150);\n  } else if (typeof document !== \"undefined\") {\n    return document.createElement(\"canvas\");\n  } else {\n    throw new Error(\"Cannot create a canvas in this context\");\n  }\n}\nfunction getWebGLRenderingContext(webGLVersion, customCanvas) {\n  if (webGLVersion !== 1 && webGLVersion !== 2) {\n    throw new Error(\"Cannot get WebGL rendering context, WebGL is disabled.\");\n  }\n  const canvas = customCanvas == null ? createCanvas(webGLVersion) : customCanvas;\n  canvas.addEventListener(\"webglcontextlost\", (ev) => {\n    ev.preventDefault();\n    delete contexts[webGLVersion];\n  }, false);\n  if (env().getBool(\"SOFTWARE_WEBGL_ENABLED\")) {\n    WEBGL_ATTRIBUTES.failIfMajorPerformanceCaveat = false;\n  }\n  if (webGLVersion === 1) {\n    return canvas.getContext(\"webgl\", WEBGL_ATTRIBUTES) || canvas.getContext(\"experimental-webgl\", WEBGL_ATTRIBUTES);\n  }\n  return canvas.getContext(\"webgl2\", WEBGL_ATTRIBUTES);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/tex_util.js\nvar PackingScheme;\n(function(PackingScheme2) {\n  PackingScheme2[PackingScheme2[\"DENSE\"] = 0] = \"DENSE\";\n  PackingScheme2[PackingScheme2[\"SHARED_BATCH\"] = 1] = \"SHARED_BATCH\";\n})(PackingScheme || (PackingScheme = {}));\nvar TextureUsage;\n(function(TextureUsage2) {\n  TextureUsage2[TextureUsage2[\"RENDER\"] = 0] = \"RENDER\";\n  TextureUsage2[TextureUsage2[\"UPLOAD\"] = 1] = \"UPLOAD\";\n  TextureUsage2[TextureUsage2[\"PIXELS\"] = 2] = \"PIXELS\";\n  TextureUsage2[TextureUsage2[\"DOWNLOAD\"] = 3] = \"DOWNLOAD\";\n})(TextureUsage || (TextureUsage = {}));\nvar PhysicalTextureType;\n(function(PhysicalTextureType2) {\n  PhysicalTextureType2[PhysicalTextureType2[\"UNPACKED_FLOAT16\"] = 0] = \"UNPACKED_FLOAT16\";\n  PhysicalTextureType2[PhysicalTextureType2[\"UNPACKED_FLOAT32\"] = 1] = \"UNPACKED_FLOAT32\";\n  PhysicalTextureType2[PhysicalTextureType2[\"PACKED_4X1_UNSIGNED_BYTE\"] = 2] = \"PACKED_4X1_UNSIGNED_BYTE\";\n  PhysicalTextureType2[PhysicalTextureType2[\"PACKED_2X2_FLOAT32\"] = 3] = \"PACKED_2X2_FLOAT32\";\n  PhysicalTextureType2[PhysicalTextureType2[\"PACKED_2X2_FLOAT16\"] = 4] = \"PACKED_2X2_FLOAT16\";\n})(PhysicalTextureType || (PhysicalTextureType = {}));\nfunction getUnpackedMatrixTextureShapeWidthHeight(rows, columns) {\n  return [columns, rows];\n}\nfunction getUnpackedArraySizeFromMatrixSize(matrixSize, channelsPerTexture) {\n  return matrixSize * channelsPerTexture;\n}\nfunction getDenseTexShape(shape) {\n  const size = util_exports.sizeFromShape(shape);\n  const texelsNeeded = Math.ceil(size / 4);\n  return util_exports.sizeToSquarishShape(texelsNeeded);\n}\nfunction getPackedMatrixTextureShapeWidthHeight(rows, columns) {\n  return [\n    Math.max(1, Math.ceil(columns / 2)),\n    Math.max(1, Math.ceil(rows / 2))\n  ];\n}\nfunction getPackedRGBAArraySizeFromMatrixShape(rows, columns) {\n  const [w, h] = getPackedMatrixTextureShapeWidthHeight(rows, columns);\n  return w * h * 4;\n}\nfunction getTextureConfig(gl, textureHalfFloatExtension) {\n  const glany = gl;\n  let internalFormatFloat;\n  let internalFormatHalfFloat;\n  let internalFormatPackedHalfFloat;\n  let internalFormatPackedFloat;\n  let textureFormatFloat;\n  let downloadTextureFormat;\n  let downloadUnpackNumChannels;\n  let defaultNumChannels;\n  let textureTypeHalfFloat;\n  let textureTypeFloat;\n  if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n    internalFormatFloat = glany.R32F;\n    internalFormatHalfFloat = glany.R16F;\n    internalFormatPackedHalfFloat = glany.RGBA16F;\n    internalFormatPackedFloat = glany.RGBA32F;\n    textureFormatFloat = glany.RED;\n    downloadUnpackNumChannels = 4;\n    defaultNumChannels = 1;\n    textureTypeHalfFloat = glany.HALF_FLOAT;\n    textureTypeFloat = glany.FLOAT;\n    downloadTextureFormat = glany.RGBA8;\n  } else {\n    internalFormatFloat = gl.RGBA;\n    internalFormatHalfFloat = gl.RGBA;\n    internalFormatPackedHalfFloat = gl.RGBA;\n    internalFormatPackedFloat = glany.RGBA;\n    textureFormatFloat = gl.RGBA;\n    downloadUnpackNumChannels = 4;\n    defaultNumChannels = 4;\n    textureTypeHalfFloat = textureHalfFloatExtension != null ? textureHalfFloatExtension.HALF_FLOAT_OES : null;\n    textureTypeFloat = gl.FLOAT;\n    downloadTextureFormat = gl.RGBA;\n  }\n  return {\n    internalFormatFloat,\n    internalFormatHalfFloat,\n    internalFormatPackedHalfFloat,\n    internalFormatPackedFloat,\n    textureFormatFloat,\n    downloadTextureFormat,\n    downloadUnpackNumChannels,\n    defaultNumChannels,\n    textureTypeHalfFloat,\n    textureTypeFloat\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl_util.js\nfunction callAndCheck(gl, func2) {\n  const returnValue = func2();\n  if (env().getBool(\"DEBUG\")) {\n    checkWebGLError(gl);\n  }\n  return returnValue;\n}\nfunction checkWebGLError(gl) {\n  const error = gl.getError();\n  if (error !== gl.NO_ERROR) {\n    throw new Error(\"WebGL Error: \" + getWebGLErrorMessage(gl, error));\n  }\n}\nvar MIN_FLOAT16 = 596e-10;\nvar MAX_FLOAT16 = 65504;\nfunction canBeRepresented(num) {\n  if (env().getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\") || num === 0 || MIN_FLOAT16 < Math.abs(num) && Math.abs(num) < MAX_FLOAT16) {\n    return true;\n  }\n  return false;\n}\nfunction getWebGLErrorMessage(gl, status) {\n  switch (status) {\n    case gl.NO_ERROR:\n      return \"NO_ERROR\";\n    case gl.INVALID_ENUM:\n      return \"INVALID_ENUM\";\n    case gl.INVALID_VALUE:\n      return \"INVALID_VALUE\";\n    case gl.INVALID_OPERATION:\n      return \"INVALID_OPERATION\";\n    case gl.INVALID_FRAMEBUFFER_OPERATION:\n      return \"INVALID_FRAMEBUFFER_OPERATION\";\n    case gl.OUT_OF_MEMORY:\n      return \"OUT_OF_MEMORY\";\n    case gl.CONTEXT_LOST_WEBGL:\n      return \"CONTEXT_LOST_WEBGL\";\n    default:\n      return `Unknown error code ${status}`;\n  }\n}\nfunction getExtensionOrThrow(gl, extensionName) {\n  return throwIfNull(gl, () => gl.getExtension(extensionName), 'Extension \"' + extensionName + '\" not supported on this browser.');\n}\nfunction createVertexShader(gl, vertexShaderSource) {\n  const vertexShader = throwIfNull(gl, () => gl.createShader(gl.VERTEX_SHADER), \"Unable to create vertex WebGLShader.\");\n  callAndCheck(gl, () => gl.shaderSource(vertexShader, vertexShaderSource));\n  callAndCheck(gl, () => gl.compileShader(vertexShader));\n  if (gl.getShaderParameter(vertexShader, gl.COMPILE_STATUS) === false) {\n    console.log(gl.getShaderInfoLog(vertexShader));\n    throw new Error(\"Failed to compile vertex shader.\");\n  }\n  return vertexShader;\n}\nfunction createFragmentShader(gl, fragmentShaderSource) {\n  const fragmentShader = throwIfNull(gl, () => gl.createShader(gl.FRAGMENT_SHADER), \"Unable to create fragment WebGLShader.\");\n  callAndCheck(gl, () => gl.shaderSource(fragmentShader, fragmentShaderSource));\n  callAndCheck(gl, () => gl.compileShader(fragmentShader));\n  if (env().get(\"ENGINE_COMPILE_ONLY\")) {\n    return fragmentShader;\n  }\n  if (gl.getShaderParameter(fragmentShader, gl.COMPILE_STATUS) === false) {\n    logShaderSourceAndInfoLog(fragmentShaderSource, gl.getShaderInfoLog(fragmentShader));\n    throw new Error(\"Failed to compile fragment shader.\");\n  }\n  return fragmentShader;\n}\nvar lineNumberRegex = /ERROR: [0-9]+:([0-9]+):/g;\nfunction logShaderSourceAndInfoLog(shaderSource, shaderInfoLog) {\n  const lineNumberRegexResult = lineNumberRegex.exec(shaderInfoLog);\n  if (lineNumberRegexResult == null) {\n    console.log(`Couldn't parse line number in error: ${shaderInfoLog}`);\n    console.log(shaderSource);\n    return;\n  }\n  const lineNumber = +lineNumberRegexResult[1];\n  const shaderLines = shaderSource.split(\"\\n\");\n  const pad3 = shaderLines.length.toString().length + 2;\n  const linesWithLineNumbers = shaderLines.map((line, lineNumber2) => util_exports.rightPad((lineNumber2 + 1).toString(), pad3) + line);\n  let maxLineLength = 0;\n  for (let i = 0; i < linesWithLineNumbers.length; i++) {\n    maxLineLength = Math.max(linesWithLineNumbers[i].length, maxLineLength);\n  }\n  const beforeErrorLines = linesWithLineNumbers.slice(0, lineNumber - 1);\n  const errorLine = linesWithLineNumbers.slice(lineNumber - 1, lineNumber);\n  const afterErrorLines = linesWithLineNumbers.slice(lineNumber);\n  console.log(beforeErrorLines.join(\"\\n\"));\n  console.log(shaderInfoLog.split(\"\\n\")[0]);\n  console.log(`%c ${util_exports.rightPad(errorLine[0], maxLineLength)}`, \"border:1px solid red; background-color:#e3d2d2; color:#a61717\");\n  console.log(afterErrorLines.join(\"\\n\"));\n}\nfunction createProgram(gl) {\n  return throwIfNull(gl, () => gl.createProgram(), \"Unable to create WebGLProgram.\");\n}\nfunction linkProgram(gl, program) {\n  callAndCheck(gl, () => gl.linkProgram(program));\n  if (env().get(\"ENGINE_COMPILE_ONLY\")) {\n    return;\n  }\n  if (gl.getProgramParameter(program, gl.LINK_STATUS) === false) {\n    console.log(gl.getProgramInfoLog(program));\n    throw new Error(\"Failed to link vertex and fragment shaders.\");\n  }\n}\nfunction validateProgram(gl, program) {\n  callAndCheck(gl, () => gl.validateProgram(program));\n  if (gl.getProgramParameter(program, gl.VALIDATE_STATUS) === false) {\n    console.log(gl.getProgramInfoLog(program));\n    throw new Error(\"Shader program validation failed.\");\n  }\n}\nfunction createStaticVertexBuffer(gl, data) {\n  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), \"Unable to create WebGLBuffer\");\n  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));\n  callAndCheck(gl, () => gl.bufferData(gl.ARRAY_BUFFER, data, gl.STATIC_DRAW));\n  return buffer2;\n}\nfunction createStaticIndexBuffer(gl, data) {\n  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), \"Unable to create WebGLBuffer\");\n  callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, buffer2));\n  callAndCheck(gl, () => gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, data, gl.STATIC_DRAW));\n  return buffer2;\n}\nfunction getNumChannels() {\n  if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n    return 1;\n  }\n  return 4;\n}\nfunction createTexture(gl) {\n  return throwIfNull(gl, () => gl.createTexture(), \"Unable to create WebGLTexture.\");\n}\nfunction validateTextureSize(width, height) {\n  const maxTextureSize = env().getNumber(\"WEBGL_MAX_TEXTURE_SIZE\");\n  if (width <= 0 || height <= 0) {\n    const requested = `[${width}x${height}]`;\n    throw new Error(\"Requested texture size \" + requested + \" is invalid.\");\n  }\n  if (width > maxTextureSize || height > maxTextureSize) {\n    const requested = `[${width}x${height}]`;\n    const max6 = `[${maxTextureSize}x${maxTextureSize}]`;\n    throw new Error(\"Requested texture size \" + requested + \" greater than WebGL maximum on this browser / GPU \" + max6 + \".\");\n  }\n}\nfunction createFramebuffer(gl) {\n  return throwIfNull(gl, () => gl.createFramebuffer(), \"Unable to create WebGLFramebuffer.\");\n}\nfunction bindVertexBufferToProgramAttribute(gl, program, attribute, buffer2, arrayEntriesPerItem, itemStrideInBytes, itemOffsetInBytes) {\n  const loc = gl.getAttribLocation(program, attribute);\n  if (loc === -1) {\n    return false;\n  }\n  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));\n  callAndCheck(gl, () => gl.vertexAttribPointer(loc, arrayEntriesPerItem, gl.FLOAT, false, itemStrideInBytes, itemOffsetInBytes));\n  callAndCheck(gl, () => gl.enableVertexAttribArray(loc));\n  return true;\n}\nfunction bindTextureUnit(gl, texture, textureUnit) {\n  validateTextureUnit(gl, textureUnit);\n  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));\n}\nfunction unbindTextureUnit(gl, textureUnit) {\n  validateTextureUnit(gl, textureUnit);\n  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));\n}\nfunction getProgramUniformLocationOrThrow(gl, program, uniformName) {\n  return throwIfNull(gl, () => gl.getUniformLocation(program, uniformName), 'uniform \"' + uniformName + '\" not present in program.');\n}\nfunction getProgramUniformLocation(gl, program, uniformName) {\n  return gl.getUniformLocation(program, uniformName);\n}\nfunction bindTextureToProgramUniformSampler(gl, texture, uniformSamplerLocation, textureUnit) {\n  callAndCheck(gl, () => bindTextureUnit(gl, texture, textureUnit));\n  callAndCheck(gl, () => gl.uniform1i(uniformSamplerLocation, textureUnit));\n}\nfunction bindCanvasToFramebuffer(gl) {\n  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));\n  callAndCheck(gl, () => gl.viewport(0, 0, gl.canvas.width, gl.canvas.height));\n  callAndCheck(gl, () => gl.scissor(0, 0, gl.canvas.width, gl.canvas.height));\n}\nfunction bindColorTextureToFramebuffer(gl, texture, framebuffer) {\n  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));\n  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0));\n}\nfunction unbindColorTextureFromFramebuffer(gl, framebuffer) {\n  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));\n  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, null, 0));\n}\nfunction validateFramebuffer(gl) {\n  const status = gl.checkFramebufferStatus(gl.FRAMEBUFFER);\n  if (status !== gl.FRAMEBUFFER_COMPLETE) {\n    throw new Error(\"Error binding framebuffer: \" + getFramebufferErrorMessage(gl, status));\n  }\n}\nfunction getFramebufferErrorMessage(gl, status) {\n  switch (status) {\n    case gl.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:\n      return \"FRAMEBUFFER_INCOMPLETE_ATTACHMENT\";\n    case gl.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:\n      return \"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT\";\n    case gl.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:\n      return \"FRAMEBUFFER_INCOMPLETE_DIMENSIONS\";\n    case gl.FRAMEBUFFER_UNSUPPORTED:\n      return \"FRAMEBUFFER_UNSUPPORTED\";\n    default:\n      return `unknown error ${status}`;\n  }\n}\nfunction throwIfNull(gl, returnTOrNull, failureMessage) {\n  const tOrNull = callAndCheck(gl, () => returnTOrNull());\n  if (tOrNull == null) {\n    throw new Error(failureMessage);\n  }\n  return tOrNull;\n}\nfunction validateTextureUnit(gl, textureUnit) {\n  const maxTextureUnit = gl.MAX_COMBINED_TEXTURE_IMAGE_UNITS - 1;\n  const glTextureUnit = textureUnit + gl.TEXTURE0;\n  if (glTextureUnit < gl.TEXTURE0 || glTextureUnit > maxTextureUnit) {\n    const textureUnitRange = `[gl.TEXTURE0, gl.TEXTURE${maxTextureUnit}]`;\n    throw new Error(`textureUnit must be in ${textureUnitRange}.`);\n  }\n}\nfunction getBatchDim(shape, dimsToSkip = 2) {\n  return util_exports.sizeFromShape(shape.slice(0, shape.length - dimsToSkip));\n}\nfunction getRowsCols(shape) {\n  if (shape.length === 0) {\n    throw Error(\"Cannot get rows and columns of an empty shape array.\");\n  }\n  return [\n    shape.length > 1 ? shape[shape.length - 2] : 1,\n    shape[shape.length - 1]\n  ];\n}\nfunction getShapeAs3D(shape) {\n  let shapeAs3D = [1, 1, 1];\n  const isScalar = shape.length === 0 || shape.length === 1 && shape[0] === 1;\n  if (!isScalar) {\n    shapeAs3D = [getBatchDim(shape), ...getRowsCols(shape)];\n  }\n  return shapeAs3D;\n}\nfunction getTextureShapeFromLogicalShape(logShape, isPacked = false) {\n  let maxTexSize = env().getNumber(\"WEBGL_MAX_TEXTURE_SIZE\");\n  let maxSizeForNarrowTex = env().getNumber(\"WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE\");\n  if (maxSizeForNarrowTex === Infinity && env().getBool(\"WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE\")) {\n    maxSizeForNarrowTex = maxTexSize / 2;\n  }\n  if (isPacked) {\n    maxTexSize = maxTexSize * 2;\n    maxSizeForNarrowTex = maxSizeForNarrowTex * 2;\n    logShape = logShape.map((d, i) => i >= logShape.length - 2 ? util_exports.nearestLargerEven(logShape[i]) : logShape[i]);\n    if (logShape.length === 1) {\n      logShape = [2, logShape[0]];\n    }\n  }\n  if (logShape.length !== 2) {\n    const squeezeResult = util_exports.squeezeShape(logShape);\n    logShape = squeezeResult.newShape;\n  }\n  let size = util_exports.sizeFromShape(logShape);\n  let textureShape = null;\n  if (logShape.length <= 1 && size <= maxTexSize) {\n    textureShape = [1, size];\n  } else if (logShape.length === 2 && logShape[0] <= maxTexSize && logShape[1] <= maxTexSize) {\n    textureShape = logShape;\n  } else if (logShape.length === 3 && logShape[0] * logShape[1] <= maxTexSize && logShape[2] <= maxTexSize) {\n    textureShape = [logShape[0] * logShape[1], logShape[2]];\n  } else if (logShape.length === 3 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] <= maxTexSize) {\n    textureShape = [logShape[0], logShape[1] * logShape[2]];\n  } else if (logShape.length === 4 && logShape[0] * logShape[1] * logShape[2] <= maxTexSize && logShape[3] <= maxTexSize) {\n    textureShape = [logShape[0] * logShape[1] * logShape[2], logShape[3]];\n  } else if (logShape.length === 4 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] * logShape[3] <= maxTexSize) {\n    textureShape = [logShape[0], logShape[1] * logShape[2] * logShape[3]];\n  }\n  const isLongNarrowTex = textureShape != null && Math.max(...textureShape) > maxSizeForNarrowTex && Math.min(...textureShape) <= (isPacked ? 2 : 1) && Math.min(...textureShape) > 0;\n  if (textureShape == null || isLongNarrowTex) {\n    if (isPacked) {\n      const batchDim = getBatchDim(logShape);\n      let rows = 2, cols = 2;\n      if (logShape.length) {\n        [rows, cols] = getRowsCols(logShape);\n      }\n      size = batchDim * (rows / 2) * (cols / 2);\n      textureShape = util_exports.sizeToSquarishShape(size).map((d) => d * 2);\n    } else {\n      textureShape = util_exports.sizeToSquarishShape(size);\n    }\n  }\n  return textureShape;\n}\nfunction isEven(n) {\n  return n % 2 === 0;\n}\nfunction isReshapeFree(shape1, shape2) {\n  shape1 = shape1.slice(-2);\n  shape2 = shape2.slice(-2);\n  if (util_exports.arraysEqual(shape1, shape2)) {\n    return true;\n  }\n  if (!shape1.length || !shape2.length) {\n    return true;\n  }\n  if (shape1[0] === 0 || shape1[1] === 0 || shape2[0] === 0 || shape2[1] === 0) {\n    return true;\n  }\n  if (shape1.length !== shape2.length) {\n    const shape1Cols = shape1.slice(-1)[0];\n    const shape2Cols = shape2.slice(-1)[0];\n    if (shape1Cols === shape2Cols) {\n      return true;\n    }\n    if (isEven(shape1Cols) && isEven(shape2Cols) && (shape1[0] === 1 || shape2[0] === 1)) {\n      return true;\n    }\n  }\n  return shape1[1] === shape2[1] && isEven(shape1[0]) && isEven(shape2[0]);\n}\nvar MAX_TEXTURE_SIZE;\nvar MAX_TEXTURES_IN_SHADER;\nfunction getWebGLMaxTextureSize(webGLVersion) {\n  if (MAX_TEXTURE_SIZE == null) {\n    const gl = getWebGLContext(webGLVersion);\n    MAX_TEXTURE_SIZE = gl.getParameter(gl.MAX_TEXTURE_SIZE);\n  }\n  return MAX_TEXTURE_SIZE;\n}\nfunction resetMaxTextureSize() {\n  MAX_TEXTURE_SIZE = null;\n}\nfunction resetMaxTexturesInShader() {\n  MAX_TEXTURES_IN_SHADER = null;\n}\nfunction getMaxTexturesInShader(webGLVersion) {\n  if (MAX_TEXTURES_IN_SHADER == null) {\n    const gl = getWebGLContext(webGLVersion);\n    MAX_TEXTURES_IN_SHADER = gl.getParameter(gl.MAX_TEXTURE_IMAGE_UNITS);\n  }\n  return Math.min(16, MAX_TEXTURES_IN_SHADER);\n}\nfunction getWebGLDisjointQueryTimerVersion(webGLVersion) {\n  if (webGLVersion === 0) {\n    return 0;\n  }\n  let queryTimerVersion;\n  const gl = getWebGLContext(webGLVersion);\n  if (hasExtension(gl, \"EXT_disjoint_timer_query_webgl2\") && webGLVersion === 2) {\n    queryTimerVersion = 2;\n  } else if (hasExtension(gl, \"EXT_disjoint_timer_query\")) {\n    queryTimerVersion = 1;\n  } else {\n    queryTimerVersion = 0;\n  }\n  return queryTimerVersion;\n}\nfunction hasExtension(gl, extensionName) {\n  const ext = gl.getExtension(extensionName);\n  return ext != null;\n}\nfunction isWebGLVersionEnabled(webGLVersion) {\n  try {\n    const gl = getWebGLContext(webGLVersion);\n    if (gl != null) {\n      return true;\n    }\n  } catch (e) {\n    console.log(\"Error when getting WebGL context: \", e);\n    return false;\n  }\n  return false;\n}\nfunction isCapableOfRenderingToFloatTexture(webGLVersion) {\n  if (webGLVersion === 0) {\n    return false;\n  }\n  const gl = getWebGLContext(webGLVersion);\n  if (webGLVersion === 1) {\n    if (!hasExtension(gl, \"OES_texture_float\")) {\n      return false;\n    }\n  } else {\n    if (!hasExtension(gl, \"EXT_color_buffer_float\")) {\n      return false;\n    }\n  }\n  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);\n  return isFrameBufferComplete;\n}\nfunction isDownloadFloatTextureEnabled(webGLVersion) {\n  if (webGLVersion === 0) {\n    return false;\n  }\n  const gl = getWebGLContext(webGLVersion);\n  if (webGLVersion === 1) {\n    if (!hasExtension(gl, \"OES_texture_float\")) {\n      return false;\n    }\n    if (!hasExtension(gl, \"WEBGL_color_buffer_float\")) {\n      return false;\n    }\n  } else {\n    if (hasExtension(gl, \"EXT_color_buffer_float\")) {\n      return createFloatTextureAndBindToFramebuffer(gl);\n    }\n    const COLOR_BUFFER_HALF_FLOAT = \"EXT_color_buffer_half_float\";\n    if (hasExtension(gl, COLOR_BUFFER_HALF_FLOAT)) {\n      const textureHalfFloatExtension = gl.getExtension(COLOR_BUFFER_HALF_FLOAT);\n      return createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension);\n    }\n    return false;\n  }\n  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);\n  return isFrameBufferComplete;\n}\nfunction createFloatTextureAndBindToFramebuffer(gl) {\n  const texConfig = getTextureConfig(gl);\n  const texture = gl.createTexture();\n  gl.bindTexture(gl.TEXTURE_2D, texture);\n  const width = 1;\n  const height = 1;\n  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeFloat, null);\n  const frameBuffer = gl.createFramebuffer();\n  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);\n  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);\n  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;\n  gl.bindTexture(gl.TEXTURE_2D, null);\n  gl.bindFramebuffer(gl.FRAMEBUFFER, null);\n  gl.deleteTexture(texture);\n  gl.deleteFramebuffer(frameBuffer);\n  return isFrameBufferComplete;\n}\nfunction createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension) {\n  const texConfig = getTextureConfig(gl, textureHalfFloatExtension);\n  const texture = gl.createTexture();\n  gl.bindTexture(gl.TEXTURE_2D, texture);\n  const width = 1;\n  const height = 1;\n  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatHalfFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeHalfFloat, null);\n  const frameBuffer = gl.createFramebuffer();\n  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);\n  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);\n  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;\n  gl.bindTexture(gl.TEXTURE_2D, null);\n  gl.bindFramebuffer(gl.FRAMEBUFFER, null);\n  gl.deleteTexture(texture);\n  gl.deleteFramebuffer(frameBuffer);\n  return isFrameBufferComplete;\n}\nfunction isWebGLFenceEnabled(webGLVersion) {\n  if (webGLVersion !== 2) {\n    return false;\n  }\n  const gl = getWebGLContext(webGLVersion);\n  const isEnabled = gl.fenceSync != null;\n  return isEnabled;\n}\nfunction assertNotComplex2(tensor2, opName) {\n  if (!Array.isArray(tensor2)) {\n    tensor2 = [tensor2];\n  }\n  tensor2.forEach((t) => {\n    if (t != null) {\n      util_exports.assert(t.dtype !== \"complex64\", () => `${opName} does not support complex64 tensors in the WebGL backend.`);\n    }\n  });\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/flags_webgl.js\nvar ENV5 = env();\nENV5.registerFlag(\"HAS_WEBGL\", () => ENV5.getNumber(\"WEBGL_VERSION\") > 0);\nENV5.registerFlag(\"WEBGL_VERSION\", () => {\n  if (isWebGLVersionEnabled(2)) {\n    return 2;\n  } else if (isWebGLVersionEnabled(1)) {\n    return 1;\n  }\n  return 0;\n});\nENV5.registerFlag(\"WEBGL_CHECK_NUMERICAL_PROBLEMS\", () => false);\nENV5.registerFlag(\"WEBGL_BUFFER_SUPPORTED\", () => ENV5.get(\"WEBGL_VERSION\") === 2);\nENV5.registerFlag(\"WEBGL_CPU_FORWARD\", () => true);\nENV5.registerFlag(\"WEBGL_FORCE_F16_TEXTURES\", () => false);\nENV5.registerFlag(\"WEBGL_PACK\", () => ENV5.getBool(\"HAS_WEBGL\"));\nENV5.registerFlag(\"WEBGL_PACK_NORMALIZATION\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_CLIP\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_DEPTHWISECONV\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_BINARY_OPERATIONS\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_UNARY_OPERATIONS\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_ARRAY_OPERATIONS\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_IMAGE_OPERATIONS\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_PACK_REDUCE\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_LAZILY_UNPACK\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_CONV_IM2COL\", () => ENV5.getBool(\"WEBGL_PACK\"));\nENV5.registerFlag(\"WEBGL_MAX_TEXTURE_SIZE\", () => getWebGLMaxTextureSize(ENV5.getNumber(\"WEBGL_VERSION\")));\nENV5.registerFlag(\"WEBGL_MAX_TEXTURES_IN_SHADER\", () => getMaxTexturesInShader(ENV5.getNumber(\"WEBGL_VERSION\")));\nENV5.registerFlag(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\", () => {\n  const webGLVersion = ENV5.getNumber(\"WEBGL_VERSION\");\n  if (webGLVersion === 0) {\n    return 0;\n  }\n  return getWebGLDisjointQueryTimerVersion(webGLVersion);\n});\nENV5.registerFlag(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\", () => ENV5.getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\") > 0 && !device_util_exports.isMobile());\nENV5.registerFlag(\"WEBGL_RENDER_FLOAT32_CAPABLE\", () => isCapableOfRenderingToFloatTexture(ENV5.getNumber(\"WEBGL_VERSION\")));\nENV5.registerFlag(\"WEBGL_RENDER_FLOAT32_ENABLED\", () => {\n  return ENV5.getBool(\"WEBGL_FORCE_F16_TEXTURES\") ? false : ENV5.getBool(\"WEBGL_RENDER_FLOAT32_CAPABLE\");\n});\nENV5.registerFlag(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\", () => isDownloadFloatTextureEnabled(ENV5.getNumber(\"WEBGL_VERSION\")));\nENV5.registerFlag(\"WEBGL_FENCE_API_ENABLED\", () => isWebGLFenceEnabled(ENV5.getNumber(\"WEBGL_VERSION\")));\nENV5.registerFlag(\"WEBGL_SIZE_UPLOAD_UNIFORM\", () => {\n  const useUniforms = ENV5.getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\");\n  return useUniforms ? 4 : 0;\n});\nENV5.registerFlag(\"WEBGL_DELETE_TEXTURE_THRESHOLD\", () => {\n  return -1;\n}, (threshold3) => {\n  if (threshold3 < 0 && threshold3 !== -1) {\n    throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${threshold3}.`);\n  }\n});\nENV5.registerFlag(\"WEBGL_FLUSH_THRESHOLD\", () => {\n  return device_util_exports.isMobile() ? 1 : -1;\n}, (threshold3) => {\n  if (threshold3 < 0 && threshold3 !== -1) {\n    throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${threshold3}.`);\n  }\n});\nENV5.registerFlag(\"CPU_HANDOFF_SIZE_THRESHOLD\", () => 128);\nENV5.registerFlag(\"WEBGL_USE_SHAPES_UNIFORMS\", () => false);\nENV5.registerFlag(\"TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD\", () => 1e5);\nENV5.registerFlag(\"TOPK_K_CPU_HANDOFF_THRESHOLD\", () => 128);\nENV5.registerFlag(\"WEBGL_EXP_CONV\", () => false);\nENV5.registerFlag(\"SOFTWARE_WEBGL_ENABLED\", () => ENV5.getBool(\"IS_TEST\"));\nENV5.registerFlag(\"WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE\", () => Infinity);\nENV5.registerFlag(\"WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE\", () => false);\nENV5.registerFlag(\"WEBGL2_ISNAN_CUSTOM\", () => false);\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/glsl_version.js\nfunction getGlslDifferences() {\n  let version10;\n  let attribute;\n  let varyingVs;\n  let varyingFs;\n  let texture2D;\n  let output;\n  let defineOutput;\n  let defineSpecialNaN;\n  let defineSpecialInf;\n  let defineRound;\n  if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n    version10 = \"#version 300 es\";\n    attribute = \"in\";\n    varyingVs = \"out\";\n    varyingFs = \"in\";\n    texture2D = \"texture\";\n    output = \"outputColor\";\n    defineOutput = \"out vec4 outputColor;\";\n    defineSpecialNaN = env().getBool(\"WEBGL2_ISNAN_CUSTOM\") ? `\n      bool isnan_custom(float val) {\n        uint floatToUint = floatBitsToUint(val);\n        return (floatToUint & 0x7fffffffu) > 0x7f800000u;\n      }\n\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan_custom(val.x),\n          isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));\n      }\n\n      #define isnan(value) isnan_custom(value)\n    ` : \"\";\n    defineSpecialInf = ``;\n    defineRound = `\n      #define round(value) newRound(value)\n      int newRound(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 newRound(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    `;\n  } else {\n    version10 = \"\";\n    attribute = \"attribute\";\n    varyingVs = \"varying\";\n    varyingFs = \"varying\";\n    texture2D = \"texture2D\";\n    output = \"gl_FragColor\";\n    defineOutput = \"\";\n    defineSpecialNaN = `\n      #define isnan(value) isnan_custom(value)\n      bool isnan_custom(float val) {\n        return (val > 0. || val < 1. || val == 0.) ? false : true;\n      }\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));\n      }\n    `;\n    defineSpecialInf = `\n      uniform float INFINITY;\n\n      bool isinf(float val) {\n        return abs(val) == INFINITY;\n      }\n      bvec4 isinf(vec4 val) {\n        return equal(abs(val), vec4(INFINITY));\n      }\n    `;\n    defineRound = `\n      int round(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 round(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    `;\n  }\n  return {\n    version: version10,\n    attribute,\n    varyingVs,\n    varyingFs,\n    texture2D,\n    output,\n    defineOutput,\n    defineSpecialNaN,\n    defineSpecialInf,\n    defineRound\n  };\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/shader_compiler_util.js\nfunction getLogicalCoordinatesFromFlatIndex(coords2, shape, index = \"index\") {\n  const strides = util_exports.computeStrides(shape);\n  return strides.map((stride, i) => {\n    const line1 = `int ${coords2[i]} = ${index} / ${stride}`;\n    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${stride}` : `index -= ${coords2[i]} * ${stride}`;\n    return `${line1}; ${line2};`;\n  }).join(\"\");\n}\nfunction getOutputLogicalCoordinatesFromFlatIndexByUniform(coords2, shape, index = \"index\") {\n  const strides = util_exports.computeStrides(shape);\n  return strides.map((_, i) => {\n    const line1 = `int ${coords2[i]} = ${index} / outShapeStrides[${i}]`;\n    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * outShapeStrides[${i}]` : `index -= ${coords2[i]} * outShapeStrides[${i}]`;\n    return `${line1}; ${line2};`;\n  }).join(\"\");\n}\nfunction symbolicallyComputeStrides(indicesArr, variableName) {\n  const numCoords = indicesArr.length;\n  const shape = indicesArr.map((d) => `${variableName}[${d}]`);\n  const strides = new Array(numCoords - 1);\n  strides[numCoords - 2] = shape[numCoords - 1];\n  for (let i = numCoords - 3; i >= 0; --i) {\n    strides[i] = `(${strides[i + 1]} * ${shape[i + 1]})`;\n  }\n  return strides;\n}\nfunction getLogicalCoordinatesFromFlatIndexByUniform(coords2, variableName, index = \"index\") {\n  const indicesArray = coords2.map((_, i) => i);\n  const strides = symbolicallyComputeStrides(indicesArray, variableName);\n  return strides.map((_, i) => {\n    const line1 = `int ${coords2[i]} = ${index} / ${strides[i]}`;\n    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${strides[i]}` : `index -= ${coords2[i]} * ${strides[i]}`;\n    return `${line1}; ${line2};`;\n  }).join(\"\");\n}\nfunction getFlatIndexFrom3D(shape) {\n  const strides = util_exports.computeStrides(shape).map((d) => d.toString());\n  return `\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * ${strides[0]} + coords.y * ${strides[1]} + coords.z;\n  }\n`;\n}\nfunction getFlatIndexFrom3DOutput() {\n  return `\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n`;\n}\nvar ENCODE_FLOAT_SNIPPET = `\n  const float FLOAT_MAX = 1.70141184e38;\n  const float FLOAT_MIN = 1.17549435e-38;\n\n  lowp vec4 encode_float(highp float v) {\n    if (isnan(v)) {\n      return vec4(255, 255, 255, 255);\n    }\n\n    highp float av = abs(v);\n\n    if(av < FLOAT_MIN) {\n      return vec4(0.0, 0.0, 0.0, 0.0);\n    } else if(v > FLOAT_MAX) {\n      return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;\n    } else if(v < -FLOAT_MAX) {\n      return vec4(0.0, 0.0,  128.0, 255.0) / 255.0;\n    }\n\n    highp vec4 c = vec4(0,0,0,0);\n\n    highp float e = floor(log2(av));\n    highp float m = exp2(fract(log2(av))) - 1.0;\n\n    c[2] = floor(128.0 * m);\n    m -= c[2] / 128.0;\n    c[1] = floor(32768.0 * m);\n    m -= c[1] / 32768.0;\n    c[0] = floor(8388608.0 * m);\n\n    highp float ebias = e + 127.0;\n    c[3] = floor(ebias / 2.0);\n    ebias -= c[3] * 2.0;\n    c[2] += floor(ebias) * 128.0;\n\n    c[3] += 128.0 * step(0.0, -v);\n\n    return c / 255.0;\n  }\n`;\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/shader_compiler.js\nvar { getBroadcastDims: getBroadcastDims2 } = backend_util_exports;\nfunction makeShader(inputsInfo, outputShape, program) {\n  const prefixSnippets = [];\n  inputsInfo.forEach((x) => {\n    const size = util_exports.sizeFromShape(x.shapeInfo.logicalShape);\n    if (x.shapeInfo.isUniform) {\n      prefixSnippets.push(`uniform float ${x.name}${size > 1 ? `[${size}]` : \"\"};`);\n    } else {\n      prefixSnippets.push(`uniform sampler2D ${x.name};`);\n      prefixSnippets.push(`uniform int offset${x.name};`);\n    }\n    if (program.enableShapeUniforms) {\n      const { uniformShape } = getUniformInfoFromShape(program.packedInputs, x.shapeInfo.logicalShape, x.shapeInfo.texShape);\n      switch (uniformShape.length) {\n        case 1:\n          prefixSnippets.push(`uniform int ${x.name}Shape;`);\n          break;\n        case 2:\n          prefixSnippets.push(`uniform ivec2 ${x.name}Shape;`);\n          break;\n        case 3:\n          prefixSnippets.push(`uniform ivec3 ${x.name}Shape;`);\n          break;\n        case 4:\n          prefixSnippets.push(`uniform ivec4 ${x.name}Shape;`);\n          break;\n        default:\n          break;\n      }\n      prefixSnippets.push(`uniform ivec2 ${x.name}TexShape;`);\n    }\n  });\n  if (program.enableShapeUniforms) {\n    switch (outputShape.logicalShape.length) {\n      case 1:\n        prefixSnippets.push(`uniform int outShape;`);\n        break;\n      case 2:\n        prefixSnippets.push(`uniform ivec2 outShape;`);\n        prefixSnippets.push(`uniform int outShapeStrides;`);\n        break;\n      case 3:\n        prefixSnippets.push(`uniform ivec3 outShape;`);\n        prefixSnippets.push(`uniform ivec2 outShapeStrides;`);\n        break;\n      case 4:\n        prefixSnippets.push(`uniform ivec4 outShape;`);\n        prefixSnippets.push(`uniform ivec3 outShapeStrides;`);\n        break;\n      default:\n        break;\n    }\n    prefixSnippets.push(`uniform ivec2 outTexShape;`);\n  }\n  if (program.customUniforms) {\n    program.customUniforms.forEach((d) => {\n      prefixSnippets.push(`uniform ${d.type} ${d.name}${d.arrayIndex ? `[${d.arrayIndex}]` : \"\"};`);\n    });\n  }\n  const inputPrefixSnippet = prefixSnippets.join(\"\\n\");\n  const inputSamplingSnippet = inputsInfo.map((x) => getInputSamplingSnippet(x, outputShape, program.packedInputs, program.enableShapeUniforms)).join(\"\\n\");\n  const outTexShape = outputShape.texShape;\n  const glsl = getGlslDifferences();\n  const floatTextureSampleSnippet = getFloatTextureSampleSnippet(glsl);\n  let outputSamplingSnippet;\n  let floatTextureSetOutputSnippet;\n  let shaderPrefix = getShaderPrefix(glsl);\n  if (outputShape.isPacked) {\n    outputSamplingSnippet = getPackedOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);\n    floatTextureSetOutputSnippet = getFloatTextureSetRGBASnippet(glsl);\n  } else {\n    outputSamplingSnippet = getOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);\n    floatTextureSetOutputSnippet = getFloatTextureSetRSnippet(glsl);\n  }\n  if (program.packedInputs) {\n    shaderPrefix += SHADER_PACKED_PREFIX;\n  }\n  const source = [\n    shaderPrefix,\n    floatTextureSampleSnippet,\n    floatTextureSetOutputSnippet,\n    inputPrefixSnippet,\n    outputSamplingSnippet,\n    inputSamplingSnippet,\n    program.userCode\n  ].join(\"\\n\");\n  return source;\n}\nfunction getSamplerFromInInfo(inInfo, enableShapeUniforms = false) {\n  const shape = inInfo.shapeInfo.logicalShape;\n  switch (shape.length) {\n    case 0:\n      return getSamplerScalar(inInfo, enableShapeUniforms);\n    case 1:\n      return getSampler1D(inInfo, enableShapeUniforms);\n    case 2:\n      return getSampler2D(inInfo, enableShapeUniforms);\n    case 3:\n      return getSampler3D(inInfo, enableShapeUniforms);\n    case 4:\n      return getSampler4D(inInfo, enableShapeUniforms);\n    case 5:\n      return getSampler5D(inInfo);\n    case 6:\n      return getSampler6D(inInfo);\n    default:\n      throw new Error(`${shape.length}-D input sampling is not yet supported`);\n  }\n}\nfunction getPackedSamplerFromInInfo(inInfo, enableShapeUniforms) {\n  const shape = inInfo.shapeInfo.logicalShape;\n  switch (shape.length) {\n    case 0:\n      return getPackedSamplerScalar(inInfo);\n    case 1:\n      return getPackedSampler1D(inInfo, enableShapeUniforms);\n    case 2:\n      return getPackedSampler2D(inInfo, enableShapeUniforms);\n    case 3:\n      return getPackedSampler3D(inInfo, enableShapeUniforms);\n    default:\n      return getPackedSamplerND(inInfo, enableShapeUniforms);\n  }\n}\nfunction getInputSamplingSnippet(inInfo, outShapeInfo, usesPackedTextures = false, enableShapeUniforms) {\n  let res = \"\";\n  if (usesPackedTextures) {\n    res += getPackedSamplerFromInInfo(inInfo, enableShapeUniforms);\n  } else {\n    res += getSamplerFromInInfo(inInfo, enableShapeUniforms);\n  }\n  const inShape = inInfo.shapeInfo.logicalShape;\n  const outShape = outShapeInfo.logicalShape;\n  if (inShape.length <= outShape.length) {\n    if (usesPackedTextures) {\n      res += getPackedSamplerAtOutputCoords(inInfo, outShapeInfo);\n    } else {\n      res += getSamplerAtOutputCoords(inInfo, outShapeInfo);\n    }\n  }\n  return res;\n}\nfunction getPackedOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {\n  switch (outShape.length) {\n    case 0:\n      return getOutputScalarCoords();\n    case 1:\n      return getOutputPacked1DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 2:\n      return getOutputPacked2DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 3:\n      return getOutputPacked3DCoords(outShape, outTexShape, enableShapeUniforms);\n    default:\n      return getOutputPackedNDCoords(outShape, outTexShape, enableShapeUniforms);\n  }\n}\nfunction getOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {\n  switch (outShape.length) {\n    case 0:\n      return getOutputScalarCoords();\n    case 1:\n      return getOutput1DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 2:\n      return getOutput2DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 3:\n      return getOutput3DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 4:\n      return getOutput4DCoords(outShape, outTexShape, enableShapeUniforms);\n    case 5:\n      return getOutput5DCoords(outShape, outTexShape);\n    case 6:\n      return getOutput6DCoords(outShape, outTexShape);\n    default:\n      throw new Error(`${outShape.length}-D output sampling is not yet supported`);\n  }\n}\nfunction getFloatTextureSampleSnippet(glsl) {\n  return `\n    float sampleTexture(sampler2D textureSampler, vec2 uv) {\n      return ${glsl.texture2D}(textureSampler, uv).r;\n    }\n  `;\n}\nfunction getFloatTextureSetRSnippet(glsl) {\n  return `\n    void setOutput(float val) {\n      ${glsl.output} = vec4(val, 0, 0, 0);\n    }\n  `;\n}\nfunction getFloatTextureSetRGBASnippet(glsl) {\n  return `\n    void setOutput(vec4 val) {\n      ${glsl.output} = val;\n    }\n  `;\n}\nfunction getShaderPrefix(glsl) {\n  const SHADER_PREFIX = `${glsl.version}\n    precision highp float;\n    precision highp int;\n    precision highp sampler2D;\n    ${glsl.varyingFs} vec2 resultUV;\n    ${glsl.defineOutput}\n    const vec2 halfCR = vec2(0.5, 0.5);\n\n    struct ivec5\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n    };\n\n    struct ivec6\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n      int v;\n    };\n\n    uniform float NAN;\n    ${glsl.defineSpecialNaN}\n    ${glsl.defineSpecialInf}\n    ${glsl.defineRound}\n\n    int imod(int x, int y) {\n      return x - y * (x / y);\n    }\n\n    int idiv(int a, int b, float sign) {\n      int res = a / b;\n      int mod = imod(a, b);\n      if (sign < 0. && mod != 0) {\n        res -= 1;\n      }\n      return res;\n    }\n\n    //Based on the work of Dave Hoskins\n    //https://www.shadertoy.com/view/4djSRW\n    #define HASHSCALE1 443.8975\n    float random(float seed){\n      vec2 p = resultUV * seed;\n      vec3 p3  = fract(vec3(p.xyx) * HASHSCALE1);\n      p3 += dot(p3, p3.yzx + 19.19);\n      return fract((p3.x + p3.y) * p3.z);\n    }\n\n    ${SAMPLE_1D_SNIPPET}\n    ${SAMPLE_2D_SNIPPET}\n    ${SAMPLE_3D_SNIPPET}\n  `;\n  return SHADER_PREFIX;\n}\nvar SAMPLE_1D_SNIPPET = `\nvec2 uvFromFlat(int texNumR, int texNumC, int index) {\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\nvec2 packedUVfrom1D(int texNumR, int texNumC, int index) {\n  int texelIndex = index / 2;\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`;\nvar SAMPLE_2D_SNIPPET = `\nvec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,\n  int texNumC, int row, int col) {\n  int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`;\nvar SAMPLE_3D_SNIPPET = `\nvec2 packedUVfrom3D(int texNumR, int texNumC,\n    int texelsInBatch, int texelsInLogicalRow, int b,\n    int row, int col) {\n  int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n`;\nvar SHADER_PACKED_PREFIX = `\n  float getChannel(vec4 frag, vec2 innerDims) {\n    vec2 modCoord = mod(innerDims, 2.);\n    return modCoord.x == 0. ?\n      (modCoord.y == 0. ? frag.r : frag.g) :\n      (modCoord.y == 0. ? frag.b : frag.a);\n  }\n  float getChannel(vec4 frag, int dim) {\n    float modCoord = mod(float(dim), 2.);\n    return modCoord == 0. ? frag.r : frag.g;\n  }\n`;\nfunction getOutputScalarCoords() {\n  return `\n    int getOutputCoords() {\n      return 0;\n    }\n  `;\n}\nfunction getOutputPacked1DCoords(shape, texShape, enableShapeUniforms) {\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  if (packedTexShape[0] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));\n      }\n    `;\n    }\n    return `\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ${packedTexShape[1]}.0);\n      }\n    `;\n  }\n  if (packedTexShape[1] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));\n      }\n    `;\n    }\n    return `\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ${packedTexShape[0]}.0);\n      }\n    `;\n  }\n  if (enableShapeUniforms) {\n    return `\n    int getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);\n    }\n  `;\n  }\n  return `\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));\n      return 2 * (resTexRC.x * ${packedTexShape[1]} + resTexRC.y);\n    }\n  `;\n}\nfunction getOutput1DCoords(shape, texShape, enableShapeUniforms) {\n  if (texShape[0] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      int getOutputCoords() {\n        return int(resultUV.x * float(outTexShape[1]));\n      }\n    `;\n    }\n    return `\n      int getOutputCoords() {\n        return int(resultUV.x * ${texShape[1]}.0);\n      }\n    `;\n  }\n  if (texShape[1] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      int getOutputCoords() {\n        return int(resultUV.y * float(outTexShape[0]));\n      }\n    `;\n    }\n    return `\n      int getOutputCoords() {\n        return int(resultUV.y * ${texShape[0]}.0);\n      }\n    `;\n  }\n  if (enableShapeUniforms) {\n    return `\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      return resTexRC.x * outTexShape[1] + resTexRC.y;\n    }\n  `;\n  }\n  return `\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${texShape[0]}, ${texShape[1]}));\n      return resTexRC.x * ${texShape[1]} + resTexRC.y;\n    }\n  `;\n}\nfunction getOutputPacked3DCoords(shape, texShape, enableShapeUniforms) {\n  if (enableShapeUniforms) {\n    return `\n    ivec3 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `;\n  }\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  const texelsInLogicalRow = Math.ceil(shape[2] / 2);\n  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[1] / 2);\n  return `\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));\n      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;\n\n      int b = index / ${texelsInBatch};\n      index -= b * ${texelsInBatch};\n\n      int r = 2 * (index / ${texelsInLogicalRow});\n      int c = imod(index, ${texelsInLogicalRow}) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `;\n}\nfunction getOutput3DCoords(shape, texShape, enableShapeUniforms) {\n  if (enableShapeUniforms) {\n    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform([\"r\", \"c\", \"d\"], shape);\n    return `\n  ivec3 getOutputCoords() {\n    ivec2 resTexRC = ivec2(resultUV.yx *\n                           vec2(outTexShape[0], outTexShape[1]));\n    int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n    ${coordsFromIndexSnippet2}\n    return ivec3(r, c, d);\n  }\n`;\n  }\n  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\"], shape);\n  return `\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${texShape[0]}, ${texShape[1]}));\n      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n      ${coordsFromIndexSnippet}\n      return ivec3(r, c, d);\n    }\n  `;\n}\nfunction getOutputPackedNDCoords(shape, texShape, enableShapeUniforms) {\n  if (enableShapeUniforms) {\n    return `\n    ivec4 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatchN = texelsInBatch * outShape[1];\n\n      int b2 = index / texelsInBatchN;\n      index -= b2 * texelsInBatchN;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec4(b2, b, r, c);\n    }\n  `;\n  }\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  const texelsInLogicalRow = Math.ceil(shape[shape.length - 1] / 2);\n  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[shape.length - 2] / 2);\n  let texelsInBatchN = texelsInBatch;\n  let batches = ``;\n  let coords2 = \"b, r, c\";\n  for (let b = 2; b < shape.length - 1; b++) {\n    texelsInBatchN *= shape[shape.length - b - 1];\n    batches = `\n      int b${b} = index / ${texelsInBatchN};\n      index -= b${b} * ${texelsInBatchN};\n    ` + batches;\n    coords2 = `b${b}, ` + coords2;\n  }\n  return `\n    ivec${shape.length} getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));\n      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;\n\n      ${batches}\n\n      int b = index / ${texelsInBatch};\n      index -= b * ${texelsInBatch};\n\n      int r = 2 * (index / ${texelsInLogicalRow});\n      int c = imod(index, ${texelsInLogicalRow}) * 2;\n\n      return ivec${shape.length}(${coords2});\n    }\n  `;\n}\nfunction getOutput4DCoords(shape, texShape, enableShapeUniforms) {\n  if (enableShapeUniforms) {\n    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform([\"r\", \"c\", \"d\", \"d2\"], shape);\n    return `\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      ${coordsFromIndexSnippet2}\n      return ivec4(r, c, d, d2);\n    }\n  `;\n  }\n  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\", \"d2\"], shape);\n  return `\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${texShape[0]}, ${texShape[1]}));\n      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n      ${coordsFromIndexSnippet}\n      return ivec4(r, c, d, d2);\n    }\n  `;\n}\nfunction getOutput5DCoords(shape, texShape) {\n  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\", \"d2\", \"d3\"], shape);\n  return `\n    ivec5 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${texShape[0]},\n                             ${texShape[1]}));\n\n      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n\n      ${coordsFromIndexSnippet}\n\n      ivec5 outShape = ivec5(r, c, d, d2, d3);\n      return outShape;\n    }\n  `;\n}\nfunction getOutput6DCoords(shape, texShape) {\n  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\", \"d2\", \"d3\", \"d4\"], shape);\n  return `\n    ivec6 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${texShape[0]}, ${texShape[1]}));\n      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n\n      ${coordsFromIndexSnippet}\n\n      ivec6 result = ivec6(r, c, d, d2, d3, d4);\n      return result;\n    }\n  `;\n}\nfunction getOutputPacked2DCoords(shape, texShape, enableShapeUniforms) {\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  if (util_exports.arraysEqual(shape, texShape)) {\n    if (enableShapeUniforms) {\n      return `\n      ivec2 getOutputCoords() {\n        ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n        return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));\n      }\n    `;\n    }\n    return `\n      ivec2 getOutputCoords() {\n        return 2 * ivec2(resultUV.yx * vec2(${packedTexShape[0]}, ${packedTexShape[1]}));\n      }\n    `;\n  }\n  const texelsInLogicalRow = Math.ceil(shape[1] / 2);\n  if (enableShapeUniforms) {\n    return `\n    ivec2 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec2(r, c);\n    }\n  `;\n  }\n  return `\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));\n\n      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;\n      int r = 2 * (index / ${texelsInLogicalRow});\n      int c = imod(index, ${texelsInLogicalRow}) * 2;\n\n      return ivec2(r, c);\n    }\n  `;\n}\nfunction getOutput2DCoords(shape, texShape, enableShapeUniforms) {\n  if (util_exports.arraysEqual(shape, texShape)) {\n    if (enableShapeUniforms) {\n      return `\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));\n      }\n    `;\n    }\n    return `\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(${texShape[0]}, ${texShape[1]}));\n      }\n    `;\n  }\n  if (shape[1] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `;\n    }\n    return `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${texShape[0]}, ${texShape[1]}));\n        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `;\n  }\n  if (shape[0] === 1) {\n    if (enableShapeUniforms) {\n      return `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `;\n    }\n    return `\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${texShape[0]}, ${texShape[1]}));\n        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `;\n  }\n  if (enableShapeUniforms) {\n    return `\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      int r = index / outShape[1];\n      int c = index - r * outShape[1];\n      return ivec2(r, c);\n    }\n  `;\n  }\n  return `\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${texShape[0]}, ${texShape[1]}));\n      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;\n      int r = index / ${shape[1]};\n      int c = index - r * ${shape[1]};\n      return ivec2(r, c);\n    }\n  `;\n}\nfunction getFlatOffsetUniformName(texName) {\n  return `offset${texName}`;\n}\nfunction getPackedSamplerScalar(inputInfo) {\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const glsl = getGlslDifferences();\n  return `\n    vec4 ${funcName}() {\n      return ${glsl.texture2D}(${texName}, halfCR);\n    }\n  `;\n}\nfunction getSamplerScalar(inputInfo, enableShapeUniforms) {\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  if (inputInfo.shapeInfo.isUniform) {\n    return `float ${funcName}() {return ${texName};}`;\n  }\n  const [texNumR, texNumC] = inputInfo.shapeInfo.texShape;\n  if (texNumR === 1 && texNumC === 1) {\n    return `\n      float ${funcName}() {\n        return sampleTexture(${texName}, halfCR);\n      }\n    `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  if (enableShapeUniforms) {\n    return `\n    float ${funcName}() {\n      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  const [tNumR, tNumC] = inputInfo.shapeInfo.texShape;\n  return `\n    float ${funcName}() {\n      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n}\nfunction getPackedSampler1D(inputInfo, enableShapeUniforms) {\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const texShape = inputInfo.shapeInfo.texShape;\n  const glsl = getGlslDifferences();\n  if (enableShapeUniforms) {\n    return `\n    vec4 ${funcName}(int index) {\n      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));\n      vec2 uv = packedUVfrom1D(\n        packedTexShape[0], packedTexShape[1], index);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n  }\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  return `\n    vec4 ${funcName}(int index) {\n      vec2 uv = packedUVfrom1D(\n        ${packedTexShape[0]}, ${packedTexShape[1]}, index);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler1D(inputInfo, enableShapeUniforms) {\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int index) {\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const texShape = inputInfo.shapeInfo.texShape;\n  const tNumR = texShape[0];\n  const tNumC = texShape[1];\n  if (tNumC === 1 && tNumR === 1) {\n    return `\n      float ${funcName}(int index) {\n        return sampleTexture(${texName}, halfCR);\n      }\n    `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  if (tNumC === 1) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / float(${texName}TexShape[0]));\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n      float ${funcName}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / ${tNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  if (tNumR === 1) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int index) {\n        vec2 uv = vec2((float(index + ${offset}) + 0.5) / float(${texName}TexShape[1]), 0.5);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n      float ${funcName}(int index) {\n        vec2 uv = vec2((float(index + ${offset}) + 0.5) / ${tNumC}.0, 0.5);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  if (enableShapeUniforms) {\n    return `\n    float ${funcName}(int index) {\n      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  return `\n    float ${funcName}(int index) {\n      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, index + ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n}\nfunction getPackedSampler2D(inputInfo, enableShapeUniforms) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const texShape = inputInfo.shapeInfo.texShape;\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  const glsl = getGlslDifferences();\n  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {\n    if (enableShapeUniforms) {\n      return `\n      vec4 ${funcName}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n\n        return ${glsl.texture2D}(${texName}, uv);\n      }\n    `;\n    }\n    return `\n      vec4 ${funcName}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);\n\n        return ${glsl.texture2D}(${texName}, uv);\n      }\n    `;\n  }\n  if (enableShapeUniforms) {\n    return `\n    vec4 ${funcName}(int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${texName}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n  }\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  const valuesPerRow = Math.ceil(shape[1] / 2);\n  return `\n    vec4 ${funcName}(int row, int col) {\n      vec2 uv = packedUVfrom2D(${valuesPerRow}, ${packedTexShape[0]}, ${packedTexShape[1]}, row, col);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler2D(inputInfo, enableShapeUniforms) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const texShape = inputInfo.shapeInfo.texShape;\n  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    const texNumR2 = texShape[0];\n    const texNumC2 = texShape[1];\n    return `\n    float ${funcName}(int row, int col) {\n      vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC2}.0, ${texNumR2}.0);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  const squeezedShape = newShape;\n  if (squeezedShape.length < shape.length) {\n    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);\n    const params = [\"row\", \"col\"];\n    return `\n      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}\n      float ${funcName}(int row, int col) {\n        return ${funcName}(${getSqueezedParams(params, keptDims)});\n      }\n    `;\n  }\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int row, int col) {\n        int index = round(dot(vec2(row, col), vec2(${shape[1]}, 1)));\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  const offset = getFlatOffsetUniformName(texName);\n  if (texNumC === 1) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col) {\n        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));\n        vec2 uv = vec2(0.5, (index + 0.5) / float(${texName}TexShape[0]));\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n    float ${funcName}(int row, int col) {\n      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));\n      vec2 uv = vec2(0.5, (index + 0.5) / ${texNumR}.0);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  if (texNumR === 1) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col) {\n        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));\n        vec2 uv = vec2((index + 0.5) / float(${texName}TexShape[1]), 0.5);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n    float ${funcName}(int row, int col) {\n      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));\n      vec2 uv = vec2((index + 0.5) / ${texNumC}.0, 0.5);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  if (enableShapeUniforms) {\n    return `\n      float ${funcName}(int row, int col) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${texName}Shape[1] + col + ${offset};\n        vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  return `\n  float ${funcName}(int row, int col) {\n    // Explicitly use integer operations as dot() only works on floats.\n    int index = row * ${shape[1]} + col + ${offset};\n    vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);\n    return sampleTexture(${texName}, uv);\n  }\n`;\n}\nfunction getPackedSampler3D(inputInfo, enableShapeUniforms) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const texShape = inputInfo.shapeInfo.texShape;\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  if (shape[0] === 1) {\n    const squeezedShape = shape.slice(1);\n    const keptDims = [1, 2];\n    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);\n    const params = [\"b\", \"row\", \"col\"];\n    return `\n        ${getPackedSamplerFromInInfo(newInputInfo, enableShapeUniforms)}\n        vec4 ${funcName}(int b, int row, int col) {\n          return ${funcName}(${getSqueezedParams(params, keptDims)});\n        }\n      `;\n  }\n  const glsl = getGlslDifferences();\n  if (enableShapeUniforms) {\n    return `\n    vec4 ${funcName}(int b, int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${texName}Shape[2]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom3D(\n        packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n  }\n  const texNumR = packedTexShape[0];\n  const texNumC = packedTexShape[1];\n  const valuesPerRow = Math.ceil(shape[2] / 2);\n  const texelsInBatch = valuesPerRow * Math.ceil(shape[1] / 2);\n  return `\n    vec4 ${funcName}(int b, int row, int col) {\n      vec2 uv = packedUVfrom3D(\n        ${texNumR}, ${texNumC}, ${texelsInBatch}, ${valuesPerRow}, b, row, col);\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler3D(inputInfo, enableShapeUniforms) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const stride0 = shape[1] * shape[2];\n  const stride1 = shape[2];\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  const squeezedShape = newShape;\n  if (squeezedShape.length < shape.length) {\n    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);\n    const params = [\"row\", \"col\", \"depth\"];\n    return `\n        ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}\n        float ${funcName}(int row, int col, int depth) {\n          return ${funcName}(${getSqueezedParams(params, keptDims)});\n        }\n      `;\n  }\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int row, int col, int depth) {\n        int index = round(dot(vec3(row, col, depth),\n                          vec3(${stride0}, ${stride1}, 1)));\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const texShape = inputInfo.shapeInfo.texShape;\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  const flatOffset = inputInfo.shapeInfo.flatOffset;\n  if (texNumC === stride0 && flatOffset == null) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col, int depth) {\n        int stride1 = ${texName}Shape[2];\n        float texR = float(row);\n        float texC = dot(vec2(col, depth), vec2(stride1, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n        float ${funcName}(int row, int col, int depth) {\n          float texR = float(row);\n          float texC = dot(vec2(col, depth), vec2(${stride1}, 1));\n          vec2 uv = (vec2(texC, texR) + halfCR) /\n                     vec2(${texNumC}.0, ${texNumR}.0);\n          return sampleTexture(${texName}, uv);\n        }\n      `;\n  }\n  if (texNumC === stride1 && flatOffset == null) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col, int depth) {\n        float texR = dot(vec2(row, col), vec2(${texName}Shape[1], 1));\n        float texC = float(depth);\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n    float ${funcName}(int row, int col, int depth) {\n      float texR = dot(vec2(row, col), vec2(${shape[1]}, 1));\n      float texC = float(depth);\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  if (enableShapeUniforms) {\n    return `\n    float ${funcName}(int row, int col, int depth) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int stride0 = ${texName}Shape[1] * ${texName}Shape[2];\n      int stride1 = ${texName}Shape[2];\n      int index = row * stride0 + col * stride1 + depth + ${offset};\n      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);\n      return sampleTexture(${texName}, uv);\n    }\n    `;\n  }\n  return `\n      float ${funcName}(int row, int col, int depth) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${stride0} + col * ${stride1} + depth + ${offset};\n        vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);\n        return sampleTexture(${texName}, uv);\n      }\n  `;\n}\nfunction getPackedSamplerND(inputInfo, enableShapeUniforms) {\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const glsl = getGlslDifferences();\n  if (enableShapeUniforms) {\n    return `\n    vec4 ${funcName}(int b2, int b, int row, int col) {\n      int valuesPerRow = int(ceil(float(${texName}Shape[3]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[2]) / 2.0));\n      int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);\n      texelsInBatch *= ${texName}Shape[1];\n      index = b2 * texelsInBatch + index;\n      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));\n      int texR = index / packedTexShape[1];\n      int texC = index - texR * packedTexShape[1];\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n  }\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const rank = shape.length;\n  const texShape = inputInfo.shapeInfo.texShape;\n  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];\n  const texNumR = packedTexShape[0];\n  const texNumC = packedTexShape[1];\n  const valuesPerRow = Math.ceil(shape[rank - 1] / 2);\n  let texelsInBatch = valuesPerRow * Math.ceil(shape[rank - 2] / 2);\n  let params = `int b, int row, int col`;\n  let index = `b * ${texelsInBatch} + (row / 2) * ${valuesPerRow} + (col / 2)`;\n  for (let b = 2; b < rank - 1; b++) {\n    params = `int b${b}, ` + params;\n    texelsInBatch *= shape[rank - b - 1];\n    index = `b${b} * ${texelsInBatch} + ` + index;\n  }\n  return `\n    vec4 ${funcName}(${params}) {\n      int index = ${index};\n      int texR = index / ${texNumC};\n      int texC = index - texR * ${texNumC};\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}, ${texNumR});\n      return ${glsl.texture2D}(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler4D(inputInfo, enableShapeUniforms) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const stride2 = shape[3];\n  const stride1 = shape[2] * stride2;\n  const stride0 = shape[1] * stride1;\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  if (newShape.length < shape.length) {\n    const newInputInfo = squeezeInputInfo(inputInfo, newShape);\n    const params = [\"row\", \"col\", \"depth\", \"depth2\"];\n    return `\n      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        return ${funcName}(${getSqueezedParams(params, keptDims)});\n      }\n    `;\n  }\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        int index = round(dot(vec4(row, col, depth, depth2),\n                          vec4(${stride0}, ${stride1}, ${stride2}, 1)));\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const flatOffset = inputInfo.shapeInfo.flatOffset;\n  const texShape = inputInfo.shapeInfo.texShape;\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  const stride2Str = `int stride2 = ${texName}Shape[3];`;\n  const stride1Str = `int stride1 = ${texName}Shape[2] * stride2;`;\n  const stride0Str = `int stride0 = ${texName}Shape[1] * stride1;`;\n  if (texNumC === stride0 && flatOffset == null) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        ${stride2Str}\n        ${stride1Str}\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(stride1, stride2, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(${stride1}, ${stride2}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  if (texNumC === stride2 && flatOffset == null) {\n    if (enableShapeUniforms) {\n      return `\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${texName}Shape[1] * ${texName}Shape[2], ${texName}Shape[2], 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${texName}TexShape[1], ${texName}TexShape[0]);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n    }\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${shape[1] * shape[2]}, ${shape[2]}, 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  if (enableShapeUniforms) {\n    return `\n    float ${funcName}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      ${stride2Str}\n      ${stride1Str}\n      ${stride0Str}\n      int index = row * stride0 + col * stride1 +\n          depth * stride2 + depth2;\n      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n  }\n  return `\n    float ${funcName}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${stride0} + col * ${stride1} +\n          depth * ${stride2} + depth2;\n      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index + ${offset});\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler5D(inputInfo) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const stride3 = shape[4];\n  const stride2 = shape[3] * stride3;\n  const stride1 = shape[2] * stride2;\n  const stride0 = shape[1] * stride1;\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  if (newShape.length < shape.length) {\n    const newInputInfo = squeezeInputInfo(inputInfo, newShape);\n    const params = [\"row\", \"col\", \"depth\", \"depth2\", \"depth3\"];\n    return `\n      ${getSamplerFromInInfo(newInputInfo)}\n      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {\n        return ${funcName}(${getSqueezedParams(params, keptDims)});\n      }\n    `;\n  }\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {\n        float index = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +\n          depth3;\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const flatOffset = inputInfo.shapeInfo.flatOffset;\n  const texShape = inputInfo.shapeInfo.texShape;\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  if (texNumC === stride0 && flatOffset == null) {\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n                         vec4(${stride1}, ${stride2}, ${stride3}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  if (texNumC === stride3 && flatOffset == null) {\n    return `\n      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {\n        float texR = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${shape[1] * shape[2] * shape[3]},\n               ${shape[2] * shape[3]}, ${shape[3]}, 1));\n        int texC = depth3;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  return `\n    float ${funcName}(int row, int col, int depth, int depth2, int depth3) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +\n          depth2 * ${stride3} + depth3 + ${offset};\n      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n}\nfunction getSampler6D(inputInfo) {\n  const shape = inputInfo.shapeInfo.logicalShape;\n  const texName = inputInfo.name;\n  const funcName = \"get\" + texName.charAt(0).toUpperCase() + texName.slice(1);\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  if (newShape.length < shape.length) {\n    const newInputInfo = squeezeInputInfo(inputInfo, newShape);\n    const params = [\"row\", \"col\", \"depth\", \"depth2\", \"depth3\", \"depth4\"];\n    return `\n      ${getSamplerFromInInfo(newInputInfo)}\n      float ${funcName}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        return ${funcName}(${getSqueezedParams(params, keptDims)});\n      }\n    `;\n  }\n  const stride4 = shape[5];\n  const stride3 = shape[4] * stride4;\n  const stride2 = shape[3] * stride3;\n  const stride1 = shape[2] * stride2;\n  const stride0 = shape[1] * stride1;\n  if (inputInfo.shapeInfo.isUniform) {\n    return `\n      float ${funcName}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n        int index = round(dot(\n          vec4(row, col, depth, depth2),\n          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +\n          dot(\n            vec2(depth3, depth4),\n            vec2(${stride4}, 1)));\n        ${getUniformSampler(inputInfo)}\n      }\n    `;\n  }\n  const flatOffset = inputInfo.shapeInfo.flatOffset;\n  const texShape = inputInfo.shapeInfo.texShape;\n  const texNumR = texShape[0];\n  const texNumC = texShape[1];\n  if (texNumC === stride0 && flatOffset == null) {\n    return `\n      float ${funcName}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n          vec4(${stride1}, ${stride2}, ${stride3}, ${stride4})) +\n               float(depth4);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  if (texNumC === stride4 && flatOffset == null) {\n    return `\n      float ${funcName}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        float texR = dot(vec4(row, col, depth, depth2),\n          vec4(${shape[1] * shape[2] * shape[3] * shape[4]},\n               ${shape[2] * shape[3] * shape[4]},\n               ${shape[3] * shape[4]},\n               ${shape[4]})) + float(depth3);\n        int texC = depth4;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${texNumC}.0, ${texNumR}.0);\n        return sampleTexture(${texName}, uv);\n      }\n    `;\n  }\n  const offset = getFlatOffsetUniformName(texName);\n  return `\n    float ${funcName}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +\n          depth2 * ${stride3} + depth3 * ${stride4} + depth4 + ${offset};\n      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);\n      return sampleTexture(${texName}, uv);\n    }\n  `;\n}\nfunction getUniformSampler(inputInfo) {\n  const texName = inputInfo.name;\n  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);\n  if (inSize < 2) {\n    return `return ${texName};`;\n  }\n  return `\n    for (int i = 0; i < ${inSize}; i++) {\n      if (i == index) {\n        return ${texName}[i];\n      }\n    }\n  `;\n}\nfunction getPackedSamplerAtOutputCoords(inputInfo, outShapeInfo) {\n  const texName = inputInfo.name;\n  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);\n  const funcName = \"get\" + texFuncSnippet + \"AtOutCoords\";\n  const inRank = inputInfo.shapeInfo.logicalShape.length;\n  const outRank = outShapeInfo.logicalShape.length;\n  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);\n  const type = getCoordsDataType(outRank);\n  const rankDiff = outRank - inRank;\n  let coordsSnippet;\n  const fields = [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"];\n  if (inRank === 0) {\n    coordsSnippet = \"\";\n  } else if (outRank < 2 && broadcastDims.length >= 1) {\n    coordsSnippet = \"coords = 0;\";\n  } else {\n    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join(\"\\n\");\n  }\n  let unpackedCoordsSnippet = \"\";\n  if (outRank < 2 && inRank > 0) {\n    unpackedCoordsSnippet = \"coords\";\n  } else {\n    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(\", \");\n  }\n  let output = `return outputValue;`;\n  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);\n  const isInputScalar = inSize === 1;\n  const outSize = util_exports.sizeFromShape(outShapeInfo.logicalShape);\n  const isOutputScalar = outSize === 1;\n  if (inRank === 1 && !isInputScalar && !isOutputScalar) {\n    output = `\n      return vec4(outputValue.xy, outputValue.xy);\n    `;\n  } else if (isInputScalar && !isOutputScalar) {\n    if (outRank === 1) {\n      output = `\n        return vec4(outputValue.x, outputValue.x, 0., 0.);\n      `;\n    } else {\n      output = `\n        return vec4(outputValue.x);\n      `;\n    }\n  } else if (broadcastDims.length) {\n    const rows = inRank - 2;\n    const cols = inRank - 1;\n    if (broadcastDims.indexOf(rows) > -1 && broadcastDims.indexOf(cols) > -1) {\n      output = `return vec4(outputValue.x);`;\n    } else if (broadcastDims.indexOf(rows) > -1) {\n      output = `return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);`;\n    } else if (broadcastDims.indexOf(cols) > -1) {\n      output = `return vec4(outputValue.xx, outputValue.zz);`;\n    }\n  }\n  return `\n    vec4 ${funcName}() {\n      ${type} coords = getOutputCoords();\n      ${coordsSnippet}\n      vec4 outputValue = get${texFuncSnippet}(${unpackedCoordsSnippet});\n      ${output}\n    }\n  `;\n}\nfunction getSamplerAtOutputCoords(inputInfo, outShapeInfo) {\n  const texName = inputInfo.name;\n  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);\n  const funcName = \"get\" + texFuncSnippet + \"AtOutCoords\";\n  const outTexShape = outShapeInfo.texShape;\n  const inTexShape = inputInfo.shapeInfo.texShape;\n  const inRank = inputInfo.shapeInfo.logicalShape.length;\n  const outRank = outShapeInfo.logicalShape.length;\n  if (!inputInfo.shapeInfo.isUniform && inRank === outRank && inputInfo.shapeInfo.flatOffset == null && util_exports.arraysEqual(inTexShape, outTexShape)) {\n    return `\n      float ${funcName}() {\n        return sampleTexture(${texName}, resultUV);\n      }\n    `;\n  }\n  const type = getCoordsDataType(outRank);\n  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);\n  const rankDiff = outRank - inRank;\n  let coordsSnippet;\n  const fields = [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"];\n  if (inRank === 0) {\n    coordsSnippet = \"\";\n  } else if (outRank < 2 && broadcastDims.length >= 1) {\n    coordsSnippet = \"coords = 0;\";\n  } else {\n    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join(\"\\n\");\n  }\n  let unpackedCoordsSnippet = \"\";\n  if (outRank < 2 && inRank > 0) {\n    unpackedCoordsSnippet = \"coords\";\n  } else {\n    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(\", \");\n  }\n  return `\n    float ${funcName}() {\n      ${type} coords = getOutputCoords();\n      ${coordsSnippet}\n      return get${texFuncSnippet}(${unpackedCoordsSnippet});\n    }\n  `;\n}\nfunction getCoordsDataType(rank) {\n  if (rank <= 1) {\n    return \"int\";\n  } else if (rank === 2) {\n    return \"ivec2\";\n  } else if (rank === 3) {\n    return \"ivec3\";\n  } else if (rank === 4) {\n    return \"ivec4\";\n  } else if (rank === 5) {\n    return \"ivec5\";\n  } else if (rank === 6) {\n    return \"ivec6\";\n  } else {\n    throw Error(`GPU for rank ${rank} is not yet supported`);\n  }\n}\nfunction getUniformInfoFromShape(isPacked, shape, texShape) {\n  const { newShape, keptDims } = util_exports.squeezeShape(shape);\n  const rank = shape.length;\n  const useSqueezePackedShape = isPacked && rank === 3 && shape[0] === 1;\n  const squeezeShape2 = useSqueezePackedShape ? shape.slice(1) : newShape;\n  const useSqueezeShape = !isPacked && rank > 1 && !util_exports.arraysEqual(shape, texShape) && newShape.length < rank || useSqueezePackedShape;\n  const uniformShape = useSqueezeShape ? squeezeShape2 : shape;\n  return { useSqueezeShape, uniformShape, keptDims };\n}\nfunction squeezeInputInfo(inInfo, squeezedShape) {\n  const newInputInfo = JSON.parse(JSON.stringify(inInfo));\n  newInputInfo.shapeInfo.logicalShape = squeezedShape;\n  return newInputInfo;\n}\nfunction getSqueezedParams(params, keptDims) {\n  return keptDims.map((d) => params[d]).join(\", \");\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_math.js\nfunction compileProgram(gpgpu, program, inputs, output) {\n  const inputInfos = inputs.map((input2, i) => {\n    const shapeInfo = {\n      logicalShape: input2.shape,\n      texShape: input2.isUniform ? null : input2.texData.texShape,\n      isUniform: input2.isUniform,\n      isPacked: input2.isUniform ? false : input2.texData.isPacked,\n      flatOffset: null\n    };\n    if (input2.texData != null && input2.texData.slice != null && input2.texData.slice.flatOffset > 0) {\n      shapeInfo.flatOffset = input2.texData.slice.flatOffset;\n    }\n    return { name: program.variableNames[i], shapeInfo };\n  });\n  const inShapeInfos = inputInfos.map((x) => x.shapeInfo);\n  const outShapeInfo = {\n    logicalShape: output.shape,\n    texShape: output.texData.texShape,\n    isUniform: false,\n    isPacked: output.texData.isPacked,\n    flatOffset: null\n  };\n  const source = makeShader(inputInfos, outShapeInfo, program);\n  const fragmentShader = createFragmentShader(gpgpu.gl, source);\n  const webGLProgram = gpgpu.createProgram(fragmentShader);\n  if (!env().get(\"ENGINE_COMPILE_ONLY\")) {\n    return Object.assign({\n      program,\n      fragmentShader,\n      source,\n      webGLProgram,\n      inShapeInfos,\n      outShapeInfo\n    }, getUniformLocations(gpgpu, program, webGLProgram));\n  } else {\n    return {\n      program,\n      fragmentShader,\n      source,\n      webGLProgram,\n      inShapeInfos,\n      outShapeInfo,\n      uniformLocations: null,\n      customUniformLocations: null,\n      infLoc: null,\n      nanLoc: null,\n      inShapesLocations: null,\n      inTexShapesLocations: null,\n      outShapeLocation: null,\n      outShapeStridesLocation: null,\n      outTexShapeLocation: null\n    };\n  }\n}\nfunction getUniformLocations(gpgpu, program, webGLProgram) {\n  const uniformLocations = {};\n  const inShapesLocations = {};\n  const inTexShapesLocations = {};\n  const customUniformLocations = [];\n  let outShapeLocation;\n  let outTexShapeLocation;\n  let outShapeStridesLocation;\n  let infLoc = null;\n  let nanLoc = null;\n  nanLoc = gpgpu.getUniformLocation(webGLProgram, \"NAN\", false);\n  if (env().getNumber(\"WEBGL_VERSION\") === 1) {\n    infLoc = gpgpu.getUniformLocation(webGLProgram, \"INFINITY\", false);\n  }\n  const shouldThrow = false;\n  for (let i = 0; i < program.variableNames.length; i++) {\n    const varName = program.variableNames[i];\n    uniformLocations[varName] = gpgpu.getUniformLocation(webGLProgram, varName, shouldThrow);\n    uniformLocations[`offset${varName}`] = gpgpu.getUniformLocation(webGLProgram, `offset${varName}`, shouldThrow);\n    if (program.enableShapeUniforms) {\n      inShapesLocations[`${varName}Shape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}Shape`, shouldThrow);\n      inTexShapesLocations[`${varName}TexShape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}TexShape`, shouldThrow);\n    }\n  }\n  if (program.enableShapeUniforms) {\n    outShapeLocation = gpgpu.getUniformLocation(webGLProgram, \"outShape\", shouldThrow);\n    outShapeStridesLocation = gpgpu.getUniformLocation(webGLProgram, \"outShapeStrides\", shouldThrow);\n    outTexShapeLocation = gpgpu.getUniformLocation(webGLProgram, \"outTexShape\", shouldThrow);\n  }\n  if (program.customUniforms) {\n    program.customUniforms.forEach((d, i) => {\n      customUniformLocations[i] = gpgpu.getUniformLocation(webGLProgram, d.name, shouldThrow);\n    });\n  }\n  return {\n    uniformLocations,\n    customUniformLocations,\n    infLoc,\n    nanLoc,\n    inShapesLocations,\n    inTexShapesLocations,\n    outShapeLocation,\n    outShapeStridesLocation,\n    outTexShapeLocation\n  };\n}\nfunction validateBinaryAndProgram(shapeInfos, inputs) {\n  if (shapeInfos.length !== inputs.length) {\n    throw Error(`Binary was compiled with ${shapeInfos.length} inputs, but was executed with ${inputs.length} inputs`);\n  }\n  shapeInfos.forEach((s, i) => {\n    const shapeA = s.logicalShape;\n    const input2 = inputs[i];\n    const shapeB = input2.shape;\n    if (!util_exports.arraysEqual(shapeA, shapeB)) {\n      throw Error(`Binary was compiled with different shapes than the current args. Shapes ${shapeA} and ${shapeB} must match`);\n    }\n    if (s.isUniform && input2.isUniform) {\n      return;\n    }\n    const texShapeA = s.texShape;\n    const texShapeB = input2.isUniform ? null : input2.texData.texShape;\n    if (!util_exports.arraysEqual(texShapeA, texShapeB)) {\n      throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${texShapeA} and ${texShapeB} must match`);\n    }\n  });\n}\nfunction runProgram(gpgpu, binary, inputs, output, customUniformValues) {\n  if (!binary.program.enableShapeUniforms) {\n    validateBinaryAndProgram(binary.inShapeInfos, inputs);\n    validateBinaryAndProgram([binary.outShapeInfo], [output]);\n  }\n  const outTex = output.texData.texture;\n  const outTexShape = output.texData.texShape;\n  if (output.texData.isPacked) {\n    gpgpu.setOutputPackedMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);\n  } else {\n    gpgpu.setOutputMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);\n  }\n  gpgpu.setProgram(binary.webGLProgram);\n  if (env().getNumber(\"WEBGL_VERSION\") === 1) {\n    if (binary.infLoc !== null) {\n      gpgpu.gl.uniform1f(binary.infLoc, Infinity);\n    }\n  }\n  if (binary.nanLoc !== null) {\n    gpgpu.gl.uniform1f(binary.nanLoc, NaN);\n  }\n  inputs.forEach((input2, i) => {\n    const varName = binary.program.variableNames[i];\n    const varLoc = binary.uniformLocations[varName];\n    const varOffsetLoc = binary.uniformLocations[`offset${varName}`];\n    const varShapeLoc = binary.inShapesLocations[`${varName}Shape`];\n    const varTexShapeLoc = binary.inTexShapesLocations[`${varName}TexShape`];\n    if (varShapeLoc) {\n      const { uniformShape } = getUniformInfoFromShape(binary.program.packedInputs, input2.shape, input2.texData.texShape);\n      switch (uniformShape.length) {\n        case 1:\n          gpgpu.gl.uniform1iv(varShapeLoc, new Int32Array(uniformShape));\n          break;\n        case 2:\n          gpgpu.gl.uniform2iv(varShapeLoc, new Int32Array(uniformShape));\n          break;\n        case 3:\n          gpgpu.gl.uniform3iv(varShapeLoc, new Int32Array(uniformShape));\n          break;\n        case 4:\n          gpgpu.gl.uniform4iv(varShapeLoc, new Int32Array(uniformShape));\n          break;\n        default:\n          break;\n      }\n    }\n    if (varTexShapeLoc) {\n      gpgpu.gl.uniform2i(varTexShapeLoc, input2.texData.texShape[0], input2.texData.texShape[1]);\n    }\n    if (varLoc == null) {\n      return;\n    }\n    if (input2.isUniform) {\n      if (util_exports.sizeFromShape(input2.shape) < 2) {\n        gpgpu.gl.uniform1f(varLoc, input2.uniformValues[0]);\n      } else {\n        let vals = input2.uniformValues;\n        if (!(vals instanceof Float32Array)) {\n          vals = new Float32Array(vals);\n        }\n        gpgpu.gl.uniform1fv(varLoc, vals);\n      }\n      return;\n    }\n    if (input2.texData.slice != null && varOffsetLoc != null) {\n      gpgpu.gl.uniform1i(varOffsetLoc, input2.texData.slice.flatOffset);\n    }\n    gpgpu.setInputMatrixTexture(input2.texData.texture.texture, varLoc, i);\n  });\n  const outShapeLoc = binary.outShapeLocation;\n  if (outShapeLoc) {\n    switch (output.shape.length) {\n      case 1:\n        gpgpu.gl.uniform1iv(outShapeLoc, new Int32Array(output.shape));\n        break;\n      case 2:\n        gpgpu.gl.uniform2iv(outShapeLoc, new Int32Array(output.shape));\n        break;\n      case 3:\n        gpgpu.gl.uniform3iv(outShapeLoc, new Int32Array(output.shape));\n        break;\n      case 4:\n        gpgpu.gl.uniform4iv(outShapeLoc, new Int32Array(output.shape));\n        break;\n      default:\n        break;\n    }\n  }\n  if (binary.outShapeStridesLocation) {\n    const strides = util_exports.computeStrides(output.shape);\n    switch (output.shape.length) {\n      case 2:\n        gpgpu.gl.uniform1iv(binary.outShapeStridesLocation, new Int32Array(strides));\n        break;\n      case 3:\n        gpgpu.gl.uniform2iv(binary.outShapeStridesLocation, new Int32Array(strides));\n        break;\n      case 4:\n        gpgpu.gl.uniform3iv(binary.outShapeStridesLocation, new Int32Array(strides));\n        break;\n      default:\n        break;\n    }\n  }\n  if (binary.outTexShapeLocation) {\n    gpgpu.gl.uniform2i(binary.outTexShapeLocation, output.texData.texShape[0], output.texData.texShape[1]);\n  }\n  if (binary.program.customUniforms && customUniformValues) {\n    binary.program.customUniforms.forEach((d, i) => {\n      const customLoc = binary.customUniformLocations[i];\n      const customValue = customUniformValues[i];\n      if (d.type === \"float\") {\n        gpgpu.gl.uniform1fv(customLoc, customValue);\n      } else if (d.type === \"vec2\") {\n        gpgpu.gl.uniform2fv(customLoc, customValue);\n      } else if (d.type === \"vec3\") {\n        gpgpu.gl.uniform3fv(customLoc, customValue);\n      } else if (d.type === \"vec4\") {\n        gpgpu.gl.uniform4fv(customLoc, customValue);\n      } else if (d.type === \"int\") {\n        gpgpu.gl.uniform1iv(customLoc, customValue);\n      } else if (d.type === \"ivec2\") {\n        gpgpu.gl.uniform2iv(customLoc, customValue);\n      } else if (d.type === \"ivec3\") {\n        gpgpu.gl.uniform3iv(customLoc, customValue);\n      } else if (d.type === \"ivec4\") {\n        gpgpu.gl.uniform4iv(customLoc, customValue);\n      } else {\n        throw Error(`uniform type ${d.type} is not supported yet.`);\n      }\n    });\n  }\n  gpgpu.executeProgram();\n}\nfunction makeShaderKey(program, inputs, output) {\n  let keyInputs = \"\";\n  inputs.concat(output).forEach((x) => {\n    const hasOffset = x.texData != null && x.texData.slice != null && x.texData.slice.flatOffset > 0;\n    if (program.enableShapeUniforms && !x.isUniform) {\n      const xTexShape = x.texData.texShape;\n      const { useSqueezeShape, uniformShape, keptDims } = getUniformInfoFromShape(program.packedInputs, x.shape, xTexShape);\n      let rank1 = \"\", rank2 = \"\", rank34 = \"\";\n      if (uniformShape.length === 1 && program.packedInputs) {\n        const packedTexShape = [Math.ceil(xTexShape[0] / 2), Math.ceil(xTexShape[1] / 2)];\n        rank1 = `${packedTexShape[0] > 1}_${packedTexShape[1] > 1}`;\n      } else if (uniformShape.length === 2 && !program.packedInputs) {\n        rank2 = `${uniformShape[0] > 1}_${uniformShape[1] > 1}`;\n      } else if (uniformShape.length > 2 && !program.packedInputs) {\n        const strides = util_exports.computeStrides(uniformShape);\n        rank34 = `${strides[0] === xTexShape[1]}_${strides[strides.length - 1] === xTexShape[1]}`;\n      }\n      const xRank = x.shape.length;\n      const isLogicalShapTexShapeEqual = uniformShape.length === 2 && util_exports.arraysEqual(x.shape, xTexShape);\n      const isScalar = util_exports.sizeFromShape(x.shape) === 1;\n      const broadcastDims = backend_util_exports.getBroadcastDims(x.shape, output.shape);\n      const isInOutTexShapeEqual = !program.packedInputs && xRank === output.shape.length && util_exports.arraysEqual(xTexShape, output.texData.texShape);\n      const isTexShapeGreaterThanOne = program.packedInputs || uniformShape.length > 2 ? \"\" : `${xTexShape[0] > 1}_${xTexShape[1] > 1}`;\n      keyInputs += `${xRank}_${isInOutTexShapeEqual}_${useSqueezeShape ? keptDims : \"\"}_${uniformShape.length}_${isScalar}_${broadcastDims}_${isLogicalShapTexShapeEqual}_${rank1}_${rank2}_${rank34}_${isTexShapeGreaterThanOne}_${hasOffset}`;\n    } else {\n      const texShape = x.isUniform ? \"uniform\" : x.texData.texShape;\n      keyInputs += `${x.shape}_${texShape}_${hasOffset}`;\n    }\n  });\n  const keyUserCode = program.userCode;\n  let key = program.constructor.name;\n  key += \"_\" + keyInputs + \"_\" + keyUserCode + `${env().getNumber(\"WEBGL_VERSION\")}`;\n  return key;\n}\nfunction useShapeUniforms(rank) {\n  return env().getBool(\"WEBGL_USE_SHAPES_UNIFORMS\") && rank <= 4;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/decode_matrix_gpu.js\nvar DecodeMatrixProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = false;\n    this.packedOutput = true;\n    this.outPackingScheme = PackingScheme.DENSE;\n    this.customUniforms = [{ name: \"texShape\", type: \"ivec2\" }];\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    this.userCode = `\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform([\"r\", \"c\", \"d\"], outputShape) : getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\"], outputShape)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getA(rc.x, rc.y, rc.z);\n        }\n\n        ${glsl.output} = result;\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/decode_matrix_packed_gpu.js\nvar DecodeMatrixPackedProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outPackingScheme = PackingScheme.DENSE;\n    this.customUniforms = [{ name: \"texShape\", type: \"ivec2\" }];\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    this.userCode = `\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform([\"r\", \"c\", \"d\"], outputShape) : getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\"], outputShape)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));\n        }\n\n        ${glsl.output} = result;\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_float_gpu.js\nvar EncodeFloatProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.outTexUsage = TextureUsage.DOWNLOAD;\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.userCode = `\n      ${ENCODE_FLOAT_SNIPPET}\n\n      void main() {\n        float x = getAAtOutCoords();\n        ${glsl.output} = encode_float(x);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_float_packed_gpu.js\nvar EncodeFloatPackedProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = false;\n    this.outTexUsage = TextureUsage.DOWNLOAD;\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.userCode = `\n      ${ENCODE_FLOAT_SNIPPET}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));\n        ${glsl.output} = encode_float(x);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_matrix_gpu.js\nvar CHANNEL_CHAR_TO_INDEX_MAP = {\n  \"R\": 0,\n  \"G\": 1,\n  \"B\": 2,\n  \"A\": 3\n};\nvar EncodeMatrixProgram = class {\n  constructor(outputShape, inputIsUnsignedByte = false, usedChannels = \"RGBA\") {\n    this.variableNames = [\"A\"];\n    this.customUniforms = [{ name: \"texShape\", type: \"ivec2\" }];\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    let output = `result`;\n    if (inputIsUnsignedByte) {\n      output = `floor(result * 255. + 0.5)`;\n    }\n    let mainLoop = \"\";\n    for (let usedChannelIndex = 0; usedChannelIndex < usedChannels.length; usedChannelIndex++) {\n      const curChannel = usedChannels[usedChannelIndex];\n      mainLoop += `\n          if(offset == ${usedChannelIndex}) {\n            result = values[${CHANNEL_CHAR_TO_INDEX_MAP[curChannel]}];\n          }`;\n    }\n    this.userCode = `\n      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int flatIndex = getFlatIndex(coords);\n        float result = 0.;\n        int offset = imod(flatIndex, ${usedChannels.length});\n\n        flatIndex = idiv(flatIndex, ${usedChannels.length}, 1.);\n\n        int r = flatIndex / texShape[1];\n        if (r < texShape[0]) {\n          int c = imod(flatIndex, texShape[1]);\n          vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n          vec4 values = ${glsl.texture2D}(A, uv);\n          ${mainLoop}\n        }\n        ${glsl.output} = vec4(${output}, 0., 0., 0.);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_matrix_packed_gpu.js\nvar EncodeMatrixPackedProgram = class {\n  constructor(outputShape, inputIsUnsignedByte = false) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = false;\n    this.packedOutput = true;\n    this.customUniforms = [{ name: \"texShape\", type: \"ivec2\" }];\n    const glsl = getGlslDifferences();\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    let mainLoop = \"\";\n    let output = \"result\";\n    if (inputIsUnsignedByte) {\n      output = \"floor(result * 255. + 0.5)\";\n    }\n    for (let row = 0; row <= 1; row++) {\n      for (let col = 0; col <= 1; col++) {\n        const channel = row * 2 + col;\n        mainLoop += `\n          localCoords = coords;\n          if(localCoords[2] + ${col} < ${this.enableShapeUniforms ? \"outShape[2]\" : `${outputShape[2]}`}) {\n          localCoords[2] += ${col};\n          if (localCoords[1] + ${row} < ${this.enableShapeUniforms ? \"outShape[1]\" : `${outputShape[1]}`}) {\n            localCoords[1] += ${row};\n\n            flatIndex = getFlatIndex(localCoords);\n            offset = imod(flatIndex, 4);\n\n            flatIndex = idiv(flatIndex, 4, 1.);\n\n            int r = flatIndex / texShape[1];\n            int c = imod(flatIndex, texShape[1]);\n            vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n            values = ${glsl.texture2D}(A, uv);\n\n            if (offset == 0) {\n              result[${channel}] = values[0];\n            } else if (offset == 1) {\n              result[${channel}] = values[1];\n            } else if (offset == 2) {\n              result[${channel}] = values[2];\n            } else {\n              result[${channel}] = values[3];\n            }\n          }\n        }\n        `;\n      }\n    }\n    this.userCode = `\n        ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}\n\n        void main() {\n          ivec3 coords = getOutputCoords();\n\n          vec4 result = vec4(0.);\n          int flatIndex, r, c, offset;\n          ivec3 localCoords;\n          vec2 uv;\n          vec4 values;\n\n          ${mainLoop}\n\n          ${glsl.output} = ${output};\n        }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_util.js\nvar gpgpu_util_exports = {};\n__export(gpgpu_util_exports, {\n  bindVertexProgramAttributeStreams: () => bindVertexProgramAttributeStreams,\n  createBufferFromOutputTexture: () => createBufferFromOutputTexture,\n  createFloat16MatrixTexture: () => createFloat16MatrixTexture,\n  createFloat16PackedMatrixTexture: () => createFloat16PackedMatrixTexture,\n  createFloat32MatrixTexture: () => createFloat32MatrixTexture,\n  createIndexBuffer: () => createIndexBuffer,\n  createPackedMatrixTexture: () => createPackedMatrixTexture,\n  createUnsignedBytesMatrixTexture: () => createUnsignedBytesMatrixTexture,\n  createVertexBuffer: () => createVertexBuffer,\n  createVertexShader: () => createVertexShader2,\n  downloadByteEncodedFloatMatrixFromOutputTexture: () => downloadByteEncodedFloatMatrixFromOutputTexture,\n  downloadFloat32MatrixFromBuffer: () => downloadFloat32MatrixFromBuffer,\n  downloadMatrixFromPackedOutputTexture: () => downloadMatrixFromPackedOutputTexture,\n  downloadPackedMatrixFromBuffer: () => downloadPackedMatrixFromBuffer,\n  getInternalFormatForFloat16MatrixTexture: () => getInternalFormatForFloat16MatrixTexture,\n  getInternalFormatForFloat16PackedMatrixTexture: () => getInternalFormatForFloat16PackedMatrixTexture,\n  getInternalFormatForFloat32MatrixTexture: () => getInternalFormatForFloat32MatrixTexture,\n  getInternalFormatForPackedMatrixTexture: () => getInternalFormatForPackedMatrixTexture,\n  getInternalFormatForUnsignedBytesMatrixTexture: () => getInternalFormatForUnsignedBytesMatrixTexture,\n  uploadDenseMatrixToTexture: () => uploadDenseMatrixToTexture,\n  uploadPixelDataToTexture: () => uploadPixelDataToTexture\n});\nfunction createVertexShader2(gl) {\n  const glsl = getGlslDifferences();\n  const vertexShaderSource = `${glsl.version}\n    precision highp float;\n    ${glsl.attribute} vec3 clipSpacePos;\n    ${glsl.attribute} vec2 uv;\n    ${glsl.varyingVs} vec2 resultUV;\n\n    void main() {\n      gl_Position = vec4(clipSpacePos, 1);\n      resultUV = uv;\n    }`;\n  return createVertexShader(gl, vertexShaderSource);\n}\nfunction createVertexBuffer(gl) {\n  const vertexArray = new Float32Array([-1, 1, 0, 0, 1, -1, -1, 0, 0, 0, 1, 1, 0, 1, 1, 1, -1, 0, 1, 0]);\n  return createStaticVertexBuffer(gl, vertexArray);\n}\nfunction createIndexBuffer(gl) {\n  const triangleVertexIndices = new Uint16Array([0, 1, 2, 2, 1, 3]);\n  return createStaticIndexBuffer(gl, triangleVertexIndices);\n}\nfunction createAndConfigureTexture(gl, width, height, internalFormat, textureFormat, textureType) {\n  validateTextureSize(width, height);\n  const texture = createTexture(gl);\n  const tex2d = gl.TEXTURE_2D;\n  callAndCheck(gl, () => gl.bindTexture(tex2d, texture));\n  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_S, gl.CLAMP_TO_EDGE));\n  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_T, gl.CLAMP_TO_EDGE));\n  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MIN_FILTER, gl.NEAREST));\n  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MAG_FILTER, gl.NEAREST));\n  if (env().getNumber(\"WEBGL_VERSION\") === 1) {\n    callAndCheck(gl, () => gl.texImage2D(tex2d, 0, internalFormat, width, height, 0, textureFormat, textureType, null));\n  } else {\n    callAndCheck(gl, () => gl.texStorage2D(tex2d, 1, internalFormat, width, height));\n  }\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));\n  return { texture, texShape: [height, width] };\n}\nfunction getInternalFormatForFloat32MatrixTexture(textureConfig) {\n  return textureConfig.internalFormatFloat;\n}\nfunction createFloat32MatrixTexture(gl, rows, columns, textureConfig) {\n  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);\n  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat32MatrixTexture(textureConfig), textureConfig.textureFormatFloat, gl.FLOAT);\n}\nfunction getInternalFormatForFloat16MatrixTexture(textureConfig) {\n  return textureConfig.internalFormatHalfFloat;\n}\nfunction createFloat16MatrixTexture(gl, rows, columns, textureConfig) {\n  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);\n  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16MatrixTexture(textureConfig), textureConfig.textureFormatFloat, textureConfig.textureTypeHalfFloat);\n}\nfunction getInternalFormatForUnsignedBytesMatrixTexture(textureConfig) {\n  return textureConfig.downloadTextureFormat;\n}\nfunction createUnsignedBytesMatrixTexture(gl, rows, columns, textureConfig) {\n  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);\n  return createAndConfigureTexture(gl, width, height, getInternalFormatForUnsignedBytesMatrixTexture(textureConfig), gl.RGBA, gl.UNSIGNED_BYTE);\n}\nfunction getInternalFormatForPackedMatrixTexture(textureConfig) {\n  return textureConfig.internalFormatPackedFloat;\n}\nfunction createPackedMatrixTexture(gl, rows, columns, textureConfig) {\n  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);\n  return createAndConfigureTexture(gl, width, height, getInternalFormatForPackedMatrixTexture(textureConfig), gl.RGBA, gl.FLOAT);\n}\nfunction getInternalFormatForFloat16PackedMatrixTexture(textureConfig) {\n  return textureConfig.internalFormatPackedHalfFloat;\n}\nfunction createFloat16PackedMatrixTexture(gl, rows, columns, textureConfig) {\n  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);\n  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16PackedMatrixTexture(textureConfig), gl.RGBA, textureConfig.textureTypeHalfFloat);\n}\nfunction bindVertexProgramAttributeStreams(gl, program, vertexBuffer) {\n  const posOffset = 0;\n  const uvOffset = 3 * 4;\n  const stride = 3 * 4 + 2 * 4;\n  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, vertexBuffer));\n  const success = bindVertexBufferToProgramAttribute(gl, program, \"clipSpacePos\", vertexBuffer, 3, stride, posOffset);\n  return success && bindVertexBufferToProgramAttribute(gl, program, \"uv\", vertexBuffer, 2, stride, uvOffset);\n}\nfunction uploadDenseMatrixToTexture(gl, texture, width, height, data, textureConfig) {\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));\n  let dataForUpload, texelDataType, internalFormat;\n  if (data instanceof Uint8Array) {\n    dataForUpload = new Uint8Array(width * height * 4);\n    texelDataType = gl.UNSIGNED_BYTE;\n    internalFormat = gl.RGBA;\n  } else {\n    dataForUpload = new Float32Array(width * height * 4);\n    texelDataType = gl.FLOAT;\n    internalFormat = textureConfig.internalFormatPackedFloat;\n  }\n  dataForUpload.set(data);\n  if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n    callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, width, height, gl.RGBA, texelDataType, dataForUpload));\n  } else {\n    callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, internalFormat, width, height, 0, gl.RGBA, texelDataType, dataForUpload));\n  }\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));\n}\nfunction uploadPixelDataToTexture(gl, texture, pixels) {\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));\n  if (pixels.data instanceof Uint8Array) {\n    if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, pixels.width, pixels.height, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));\n    } else {\n      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, pixels.width, pixels.height, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));\n    }\n  } else {\n    if (env().getNumber(\"WEBGL_VERSION\") === 2) {\n      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels));\n    } else {\n      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gl.RGBA, gl.UNSIGNED_BYTE, pixels));\n    }\n  }\n  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));\n}\nfunction createBufferFromOutputTexture(gl2, rows, columns, textureConfig) {\n  const buffer2 = gl2.createBuffer();\n  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2));\n  const bytesPerFloat = 4;\n  const valuesPerTexel = 4;\n  const bufferSizeBytes = bytesPerFloat * valuesPerTexel * rows * columns;\n  callAndCheck(gl2, () => gl2.bufferData(gl2.PIXEL_PACK_BUFFER, bufferSizeBytes, gl2.STREAM_READ));\n  callAndCheck(gl2, () => gl2.readPixels(0, 0, columns, rows, gl2.RGBA, gl2.FLOAT, 0));\n  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null));\n  return buffer2;\n}\nfunction downloadFloat32MatrixFromBuffer(gl, buffer2, size) {\n  const gl2 = gl;\n  const downloadTarget = new Float32Array(size);\n  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);\n  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);\n  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);\n  return downloadTarget;\n}\nfunction downloadByteEncodedFloatMatrixFromOutputTexture(gl, rows, columns, textureConfig) {\n  const [w, h] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);\n  const numChannels = 4;\n  const downloadTarget = new Uint8Array(getUnpackedArraySizeFromMatrixSize(rows * columns, numChannels));\n  callAndCheck(gl, () => gl.readPixels(0, 0, w, h, textureConfig.downloadTextureFormat, gl.UNSIGNED_BYTE, downloadTarget));\n  return new Float32Array(downloadTarget.buffer);\n}\nfunction downloadPackedMatrixFromBuffer(gl, buffer2, batch, rows, cols, physicalRows, physicalCols, textureConfig) {\n  const gl2 = gl;\n  const downloadTarget = new Float32Array(getPackedRGBAArraySizeFromMatrixShape(physicalRows, physicalCols));\n  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);\n  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);\n  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);\n  return downloadTarget;\n}\nfunction downloadMatrixFromPackedOutputTexture(gl, physicalRows, physicalCols) {\n  const packedRGBA = new Float32Array(physicalRows * physicalCols * 4);\n  callAndCheck(gl, () => gl.readPixels(0, 0, physicalCols, physicalRows, gl.RGBA, gl.FLOAT, packedRGBA));\n  return packedRGBA;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_context.js\nvar GPGPUContext = class {\n  constructor(gl) {\n    this.outputTexture = null;\n    this.program = null;\n    this.disposed = false;\n    this.vertexAttrsAreBound = false;\n    this.itemsToPoll = [];\n    const glVersion = env().getNumber(\"WEBGL_VERSION\");\n    if (gl != null) {\n      this.gl = gl;\n      setWebGLContext(glVersion, gl);\n    } else {\n      this.gl = getWebGLContext(glVersion);\n    }\n    let COLOR_BUFFER_FLOAT = \"WEBGL_color_buffer_float\";\n    const COLOR_BUFFER_HALF_FLOAT = \"EXT_color_buffer_half_float\";\n    this.parallelCompilationExtension = this.gl.getExtension(\"KHR_parallel_shader_compile\");\n    if (env().getNumber(\"WEBGL_VERSION\") === 1) {\n      const TEXTURE_FLOAT = \"OES_texture_float\";\n      const TEXTURE_HALF_FLOAT = \"OES_texture_half_float\";\n      this.textureFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_FLOAT);\n      if (hasExtension(this.gl, TEXTURE_HALF_FLOAT)) {\n        this.textureHalfFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_HALF_FLOAT);\n      } else if (env().get(\"WEBGL_FORCE_F16_TEXTURES\")) {\n        throw new Error(\"GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.\");\n      }\n      this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);\n      if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {\n        this.colorBufferHalfFloatExtension = getExtensionOrThrow(this.gl, COLOR_BUFFER_HALF_FLOAT);\n      } else if (env().get(\"WEBGL_FORCE_F16_TEXTURES\")) {\n        throw new Error(\"GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.\");\n      }\n    } else {\n      COLOR_BUFFER_FLOAT = \"EXT_color_buffer_float\";\n      if (hasExtension(this.gl, COLOR_BUFFER_FLOAT)) {\n        this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);\n      } else if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {\n        this.colorBufferHalfFloatExtension = this.gl.getExtension(COLOR_BUFFER_HALF_FLOAT);\n      } else {\n        throw new Error(\"GL context does not support color renderable floats\");\n      }\n    }\n    this.vertexBuffer = createVertexBuffer(this.gl);\n    this.indexBuffer = createIndexBuffer(this.gl);\n    this.framebuffer = createFramebuffer(this.gl);\n    this.textureConfig = getTextureConfig(this.gl, this.textureHalfFloatExtension);\n  }\n  get debug() {\n    return env().getBool(\"DEBUG\");\n  }\n  dispose() {\n    if (this.disposed) {\n      return;\n    }\n    if (this.program != null) {\n      console.warn(\"Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing.\");\n    }\n    if (this.outputTexture != null) {\n      console.warn(\"Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.\");\n    }\n    const gl = this.gl;\n    callAndCheck(gl, () => gl.finish());\n    callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));\n    callAndCheck(gl, () => gl.deleteFramebuffer(this.framebuffer));\n    callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, null));\n    callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, null));\n    callAndCheck(gl, () => gl.deleteBuffer(this.indexBuffer));\n    this.disposed = true;\n  }\n  createFloat32MatrixTexture(rows, columns) {\n    this.throwIfDisposed();\n    return createFloat32MatrixTexture(this.gl, rows, columns, this.textureConfig);\n  }\n  createFloat16MatrixTexture(rows, columns) {\n    this.throwIfDisposed();\n    return createFloat16MatrixTexture(this.gl, rows, columns, this.textureConfig);\n  }\n  createUnsignedBytesMatrixTexture(rows, columns) {\n    this.throwIfDisposed();\n    return createUnsignedBytesMatrixTexture(this.gl, rows, columns, this.textureConfig);\n  }\n  uploadPixelDataToTexture(texture, pixels) {\n    this.throwIfDisposed();\n    uploadPixelDataToTexture(this.gl, texture, pixels);\n  }\n  uploadDenseMatrixToTexture(texture, width, height, data) {\n    this.throwIfDisposed();\n    uploadDenseMatrixToTexture(this.gl, texture, width, height, data, this.textureConfig);\n  }\n  createFloat16PackedMatrixTexture(rows, columns) {\n    this.throwIfDisposed();\n    return createFloat16PackedMatrixTexture(this.gl, rows, columns, this.textureConfig);\n  }\n  createPackedMatrixTexture(rows, columns) {\n    this.throwIfDisposed();\n    return createPackedMatrixTexture(this.gl, rows, columns, this.textureConfig);\n  }\n  deleteMatrixTexture(texture) {\n    this.throwIfDisposed();\n    if (this.outputTexture === texture) {\n      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);\n      this.outputTexture = null;\n    }\n    callAndCheck(this.gl, () => this.gl.deleteTexture(texture));\n  }\n  downloadByteEncodedFloatMatrixFromOutputTexture(texture, rows, columns) {\n    return this.downloadMatrixDriver(texture, () => downloadByteEncodedFloatMatrixFromOutputTexture(this.gl, rows, columns, this.textureConfig));\n  }\n  downloadPackedMatrixFromBuffer(buffer2, batch, rows, columns, physicalRows, physicalCols) {\n    return downloadPackedMatrixFromBuffer(this.gl, buffer2, batch, rows, columns, physicalRows, physicalCols, this.textureConfig);\n  }\n  downloadFloat32MatrixFromBuffer(buffer2, size) {\n    return downloadFloat32MatrixFromBuffer(this.gl, buffer2, size);\n  }\n  createBufferFromTexture(texture, rows, columns) {\n    this.bindTextureToFrameBuffer(texture);\n    const result = createBufferFromOutputTexture(this.gl, rows, columns, this.textureConfig);\n    this.unbindTextureToFrameBuffer();\n    return result;\n  }\n  createAndWaitForFence() {\n    const fenceContext = this.createFence(this.gl);\n    return this.pollFence(fenceContext);\n  }\n  createFence(gl) {\n    let query;\n    let isFencePassed;\n    if (env().getBool(\"WEBGL_FENCE_API_ENABLED\")) {\n      const gl2 = gl;\n      const sync = gl2.fenceSync(gl2.SYNC_GPU_COMMANDS_COMPLETE, 0);\n      gl.flush();\n      isFencePassed = () => {\n        const status = gl2.clientWaitSync(sync, 0, 0);\n        return status === gl2.ALREADY_SIGNALED || status === gl2.CONDITION_SATISFIED;\n      };\n      query = sync;\n    } else if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\") > 0) {\n      query = this.beginQuery();\n      this.endQuery();\n      isFencePassed = () => this.isQueryAvailable(query, env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\"));\n    } else {\n      isFencePassed = () => true;\n    }\n    return { query, isFencePassed };\n  }\n  downloadMatrixFromPackedTexture(texture, physicalRows, physicalCols) {\n    return this.downloadMatrixDriver(texture, () => downloadMatrixFromPackedOutputTexture(this.gl, physicalRows, physicalCols));\n  }\n  createProgram(fragmentShader) {\n    this.throwIfDisposed();\n    const gl = this.gl;\n    if (this.vertexShader == null) {\n      this.vertexShader = createVertexShader2(gl);\n    }\n    const program = createProgram(gl);\n    callAndCheck(gl, () => gl.attachShader(program, this.vertexShader));\n    callAndCheck(gl, () => gl.attachShader(program, fragmentShader));\n    linkProgram(gl, program);\n    if (this.debug) {\n      validateProgram(gl, program);\n    }\n    if (!this.vertexAttrsAreBound) {\n      this.setProgram(program);\n      this.vertexAttrsAreBound = bindVertexProgramAttributeStreams(gl, this.program, this.vertexBuffer);\n    }\n    return program;\n  }\n  deleteProgram(program) {\n    this.throwIfDisposed();\n    if (program === this.program) {\n      this.program = null;\n    }\n    if (program != null) {\n      callAndCheck(this.gl, () => this.gl.deleteProgram(program));\n    }\n  }\n  setProgram(program) {\n    this.throwIfDisposed();\n    this.program = program;\n    if (this.program != null && this.debug) {\n      validateProgram(this.gl, this.program);\n    }\n    callAndCheck(this.gl, () => this.gl.useProgram(program));\n  }\n  getUniformLocation(program, uniformName, shouldThrow = true) {\n    this.throwIfDisposed();\n    if (shouldThrow) {\n      return getProgramUniformLocationOrThrow(this.gl, program, uniformName);\n    } else {\n      return getProgramUniformLocation(this.gl, program, uniformName);\n    }\n  }\n  getAttributeLocation(program, attribute) {\n    this.throwIfDisposed();\n    return callAndCheck(this.gl, () => this.gl.getAttribLocation(program, attribute));\n  }\n  getUniformLocationNoThrow(program, uniformName) {\n    this.throwIfDisposed();\n    return this.gl.getUniformLocation(program, uniformName);\n  }\n  setInputMatrixTexture(inputMatrixTexture, uniformLocation, textureUnit) {\n    this.throwIfDisposed();\n    this.throwIfNoProgram();\n    bindTextureToProgramUniformSampler(this.gl, inputMatrixTexture, uniformLocation, textureUnit);\n  }\n  setOutputMatrixTexture(outputMatrixTexture, rows, columns) {\n    this.setOutputMatrixTextureDriver(outputMatrixTexture, columns, rows);\n  }\n  setOutputPackedMatrixTexture(outputPackedMatrixTexture, rows, columns) {\n    this.throwIfDisposed();\n    const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);\n    this.setOutputMatrixTextureDriver(outputPackedMatrixTexture, width, height);\n  }\n  setOutputMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {\n    this.setOutputMatrixWriteRegionDriver(startColumn, startRow, numColumns, numRows);\n  }\n  setOutputPackedMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {\n    throw new Error(\"setOutputPackedMatrixWriteRegion not implemented.\");\n  }\n  debugValidate() {\n    if (this.program != null) {\n      validateProgram(this.gl, this.program);\n    }\n    validateFramebuffer(this.gl);\n  }\n  executeProgram() {\n    this.throwIfDisposed();\n    this.throwIfNoProgram();\n    const gl = this.gl;\n    if (this.debug) {\n      this.debugValidate();\n    }\n    callAndCheck(gl, () => gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 0));\n  }\n  blockUntilAllProgramsCompleted() {\n    this.throwIfDisposed();\n    callAndCheck(this.gl, () => this.gl.finish());\n  }\n  getQueryTimerExtension() {\n    if (this.disjointQueryTimerExtension == null) {\n      this.disjointQueryTimerExtension = getExtensionOrThrow(this.gl, env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\") === 2 ? \"EXT_disjoint_timer_query_webgl2\" : \"EXT_disjoint_timer_query\");\n    }\n    return this.disjointQueryTimerExtension;\n  }\n  getQueryTimerExtensionWebGL2() {\n    return this.getQueryTimerExtension();\n  }\n  getQueryTimerExtensionWebGL1() {\n    return this.getQueryTimerExtension();\n  }\n  beginQuery() {\n    if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\") === 2) {\n      const gl2 = this.gl;\n      const ext2 = this.getQueryTimerExtensionWebGL2();\n      const query2 = gl2.createQuery();\n      gl2.beginQuery(ext2.TIME_ELAPSED_EXT, query2);\n      return query2;\n    }\n    const ext = this.getQueryTimerExtensionWebGL1();\n    const query = ext.createQueryEXT();\n    ext.beginQueryEXT(ext.TIME_ELAPSED_EXT, query);\n    return query;\n  }\n  endQuery() {\n    if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\") === 2) {\n      const gl2 = this.gl;\n      const ext2 = this.getQueryTimerExtensionWebGL2();\n      gl2.endQuery(ext2.TIME_ELAPSED_EXT);\n      return;\n    }\n    const ext = this.getQueryTimerExtensionWebGL1();\n    ext.endQueryEXT(ext.TIME_ELAPSED_EXT);\n  }\n  async waitForQueryAndGetTime(query) {\n    await util_exports.repeatedTry(() => this.disposed || this.isQueryAvailable(query, env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\")));\n    return this.getQueryTime(query, env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION\"));\n  }\n  getQueryTime(query, queryTimerVersion) {\n    if (queryTimerVersion === 0) {\n      return null;\n    }\n    if (queryTimerVersion === 2) {\n      const gl2 = this.gl;\n      const timeElapsedNanos = gl2.getQueryParameter(query, gl2.QUERY_RESULT);\n      return timeElapsedNanos / 1e6;\n    } else {\n      const ext = this.getQueryTimerExtensionWebGL1();\n      const timeElapsedNanos = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_EXT);\n      return timeElapsedNanos / 1e6;\n    }\n  }\n  isQueryAvailable(query, queryTimerVersion) {\n    if (queryTimerVersion === 0) {\n      return true;\n    }\n    if (queryTimerVersion === 2) {\n      const gl2 = this.gl;\n      const ext = this.getQueryTimerExtensionWebGL2();\n      const available = gl2.getQueryParameter(query, gl2.QUERY_RESULT_AVAILABLE);\n      if (this.disjoint == null) {\n        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);\n      }\n      return available && !this.disjoint;\n    } else {\n      const ext = this.getQueryTimerExtensionWebGL1();\n      const available = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_AVAILABLE_EXT);\n      if (this.disjoint == null) {\n        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);\n      }\n      return available && !this.disjoint;\n    }\n  }\n  pollFence(fenceContext) {\n    return new Promise((resolve) => {\n      this.addItemToPoll(() => fenceContext.isFencePassed(), () => resolve());\n    });\n  }\n  pollItems() {\n    const index = linearSearchLastTrue(this.itemsToPoll.map((x) => x.isDoneFn));\n    for (let i = 0; i <= index; ++i) {\n      const { resolveFn } = this.itemsToPoll[i];\n      resolveFn();\n    }\n    this.itemsToPoll = this.itemsToPoll.slice(index + 1);\n  }\n  addItemToPoll(isDoneFn, resolveFn) {\n    this.itemsToPoll.push({ isDoneFn, resolveFn });\n    if (this.itemsToPoll.length > 1) {\n      return;\n    }\n    let scheduleFn = void 0;\n    if (\"setTimeoutCustom\" in env().platform) {\n      scheduleFn = env().platform.setTimeoutCustom.bind(env().platform);\n    }\n    util_exports.repeatedTry(() => {\n      this.pollItems();\n      return this.itemsToPoll.length === 0;\n    }, () => 0, null, scheduleFn);\n  }\n  bindTextureToFrameBuffer(texture) {\n    this.throwIfDisposed();\n    bindColorTextureToFramebuffer(this.gl, texture, this.framebuffer);\n    if (this.debug) {\n      validateFramebuffer(this.gl);\n    }\n  }\n  unbindTextureToFrameBuffer() {\n    if (this.outputTexture != null) {\n      bindColorTextureToFramebuffer(this.gl, this.outputTexture, this.framebuffer);\n      if (this.debug) {\n        validateFramebuffer(this.gl);\n      }\n    } else {\n      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);\n    }\n  }\n  downloadMatrixDriver(texture, downloadAndDecode) {\n    this.bindTextureToFrameBuffer(texture);\n    const result = downloadAndDecode();\n    this.unbindTextureToFrameBuffer();\n    return result;\n  }\n  setOutputMatrixTextureDriver(outputMatrixTextureMaybePacked, width, height) {\n    this.throwIfDisposed();\n    const gl = this.gl;\n    bindColorTextureToFramebuffer(gl, outputMatrixTextureMaybePacked, this.framebuffer);\n    if (this.debug) {\n      validateFramebuffer(gl);\n    }\n    this.outputTexture = outputMatrixTextureMaybePacked;\n    callAndCheck(gl, () => gl.viewport(0, 0, width, height));\n    callAndCheck(gl, () => gl.scissor(0, 0, width, height));\n  }\n  setOutputMatrixWriteRegionDriver(x, y, width, height) {\n    this.throwIfDisposed();\n    callAndCheck(this.gl, () => this.gl.scissor(x, y, width, height));\n  }\n  throwIfDisposed() {\n    if (this.disposed) {\n      throw new Error(\"Attempted to use disposed GPGPUContext.\");\n    }\n  }\n  throwIfNoProgram() {\n    if (this.program == null) {\n      throw new Error(\"No GPU program is currently set.\");\n    }\n  }\n};\nfunction linearSearchLastTrue(arr) {\n  let i = 0;\n  for (; i < arr.length; ++i) {\n    const isDone = arr[i]();\n    if (!isDone) {\n      break;\n    }\n  }\n  return i - 1;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/shared.js\nvar { addImpl: addImplCPU, bincountImpl: bincountImplCPU, bincountReduceImpl: bincountReduceImplCPU, castImpl: castImplCPU, ceilImpl: ceilImplCPU, concatImpl: concatImplCPU, equalImpl: equalImplCPU, expImpl: expImplCPU, expm1Impl: expm1ImplCPU, floorImpl: floorImplCPU, gatherNdImpl: gatherNdImplCPU, gatherV2Impl: gatherV2ImplCPU, greaterImpl: greaterImplCPU, greaterEqualImpl: greaterEqualImplCPU, lessImpl: lessImplCPU, lessEqualImpl: lessEqualImplCPU, linSpaceImpl: linSpaceImplCPU, logImpl: logImplCPU, maxImpl: maxImplCPU, maximumImpl: maximumImplCPU, minimumImpl: minimumImplCPU, multiplyImpl: multiplyImplCPU, negImpl: negImplCPU, notEqualImpl: notEqualImplCPU, prodImpl: prodImplCPU, raggedGatherImpl: raggedGatherImplCPU, raggedRangeImpl: raggedRangeImplCPU, raggedTensorToTensorImpl: raggedTensorToTensorImplCPU, rangeImpl: rangeImplCPU, rsqrtImpl: rsqrtImplCPU, scatterImpl: scatterImplCPU, sigmoidImpl: sigmoidImplCPU, simpleAbsImpl: simpleAbsImplCPU, sliceImpl: sliceImplCPU, sparseFillEmptyRowsImpl: sparseFillEmptyRowsImplCPU, sparseReshapeImpl: sparseReshapeImplCPU, sparseSegmentReductionImpl: sparseSegmentReductionImplCPU, sqrtImpl: sqrtImplCPU, stridedSliceImpl: stridedSliceImplCPU, stringNGramsImpl: stringNGramsImplCPU, stringSplitImpl: stringSplitImplCPU, stringToHashBucketFastImpl: stringToHashBucketFastImplCPU, subImpl: subImplCPU, tileImpl: tileImplCPU, topKImpl: topKImplCPU, transposeImpl: transposeImplCPU, uniqueImpl: uniqueImplCPU } = shared_exports;\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/packing_util.js\nfunction getVecChannels(name, rank) {\n  return [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"].slice(0, rank).map((d) => `${name}.${d}`);\n}\nfunction getChannels(name, rank) {\n  if (rank === 1) {\n    return [name];\n  }\n  return getVecChannels(name, rank);\n}\nfunction getSourceCoords(rank, dims) {\n  if (rank === 1) {\n    return \"rc\";\n  }\n  let coords2 = \"\";\n  for (let i = 0; i < rank; i++) {\n    coords2 += dims[i];\n    if (i < rank - 1) {\n      coords2 += \",\";\n    }\n  }\n  return coords2;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/pack_gpu.js\nvar PackProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = false;\n    this.packedOutput = true;\n    this.outputShape = outputShape;\n    this.rank = outputShape.length;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    if (this.rank === 0) {\n      this.userCode = `\n        void main() {\n          setOutput(vec4(getA(), 0., 0., 0.));\n        }\n      `;\n    } else {\n      const channels = getChannels(\"rc\", this.rank);\n      const dtype = getCoordsDataType(this.rank);\n      const outOfBoundsCondition = this.getOutOfBoundsCondition(channels);\n      const setup51 = this.getSetup(channels);\n      const output = this.getOutput(channels);\n      this.userCode = `\n        void main() {\n          ${dtype} rc = getOutputCoords();\n\n          if(${outOfBoundsCondition}) {\n            setOutput(vec4(0));\n          } else {\n            ${setup51}\n\n            setOutput(vec4(${output}));\n          }\n        }\n      `;\n    }\n  }\n  getSourceCoordsArr(dims) {\n    const coords2 = [];\n    for (let row = 0; row <= 1; row++) {\n      for (let col = 0; col <= 1; col++) {\n        let coord = `${row === 0 ? \"r\" : \"rp1\"}, ${col === 0 ? \"c\" : \"cp1\"}`;\n        for (let d = 2; d < this.rank; d++) {\n          coord = `${dims[dims.length - 1 - d]},` + coord;\n        }\n        coords2.push(coord);\n      }\n    }\n    return coords2;\n  }\n  getOutOfBoundsCondition(dims) {\n    if (this.rank === 1) {\n      return `rc > ${this.enableShapeUniforms ? \"outShape\" : this.outputShape[0]}`;\n    }\n    let cond = \"\";\n    for (let i = this.rank - 2; i < this.rank; i++) {\n      cond += `${dims[i]} >= ${this.enableShapeUniforms ? `outShape[${i}]` : this.outputShape[i]}`;\n      if (i < this.rank - 1) {\n        cond += \"||\";\n      }\n    }\n    return cond;\n  }\n  getSetup(dims) {\n    if (this.rank === 1) {\n      return \"\";\n    }\n    const innerDims = dims.slice(-2);\n    const col = this.enableShapeUniforms ? `outShape[${this.rank} - 1]` : this.outputShape[this.rank - 1];\n    const row = this.enableShapeUniforms ? `outShape[${this.rank} - 2]` : this.outputShape[this.rank - 2];\n    return `\n      int r = ${innerDims[0]};\n      int c = ${innerDims[1]};\n      int rp1 = r + 1;\n      int cp1 = c + 1;\n\n      bool cEdge = cp1 >= ${col};\n      bool rEdge = rp1 >= ${row};\n    `;\n  }\n  getOutput(dims) {\n    const sourceCoords = this.getSourceCoordsArr(dims);\n    if (this.rank === 1) {\n      const outShape = this.enableShapeUniforms ? \"outShape\" : this.outputShape[0];\n      return `getA(rc), (rc + 1 >= ${outShape} ? 0. : getA(rc + 1)), 0, 0`;\n    }\n    return `getA(${sourceCoords[0]}),\n            cEdge ? 0. : getA(${sourceCoords[1]}),\n            rEdge ? 0. : getA(${sourceCoords[2]}),\n            rEdge || cEdge ? 0. : getA(${sourceCoords[3]})`;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/reshape_packed_gpu.js\nvar ReshapePackedProgram = class {\n  constructor(outputShape, inputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [{ name: \"inputShape\", type: \"ivec3\" }];\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    let mainLoop = ``;\n    for (let i = 0; i < 4; i++) {\n      let thisRC = `thisRC = rc;`;\n      if (i % 2 === 1) {\n        thisRC += `thisRC.z += 1;`;\n      }\n      if (i > 1) {\n        thisRC += `thisRC.y += 1;`;\n      }\n      mainLoop += `\n        ${thisRC}\n        ${i > 0 ? `if(thisRC.y < rows && thisRC.z < cols){` : \"\"}\n          int flatIndex = getFlatIndex(thisRC);\n\n          ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);\n          vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));\n\n          result[${i}] =\n            getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);\n        ${i > 0 ? \"}\" : \"\"}\n      `;\n    }\n    this.userCode = `\n      ${getReshapedInputCoords(inputShape, this.enableShapeUniforms)}\n      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0.);\n\n        ivec3 thisRC;\n        int rows = ${this.enableShapeUniforms ? \"outShape[1]\" : outputShape[1]};\n        int cols = ${this.enableShapeUniforms ? \"outShape[2]\" : outputShape[2]};\n\n        ${mainLoop}\n\n        setOutput(result);\n      }\n    `;\n  }\n};\nfunction getReshapedInputCoords(shape, enableShapeUniforms) {\n  const coordsFromIndexSnippet = enableShapeUniforms ? getLogicalCoordinatesFromFlatIndexByUniform([\"r\", \"c\", \"d\"], \"inputShape\") : getLogicalCoordinatesFromFlatIndex([\"r\", \"c\", \"d\"], shape);\n  return `\n    ivec3 inputCoordsFromReshapedOutCoords(int index) {\n      ${coordsFromIndexSnippet}\n      return ivec3(r, c, d);\n    }\n  `;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/texture_manager.js\nvar TextureManager = class {\n  constructor(gpgpu) {\n    this.gpgpu = gpgpu;\n    this.numUsedTextures = 0;\n    this.numFreeTextures = 0;\n    this._numBytesAllocated = 0;\n    this._numBytesFree = 0;\n    this.freeTextures = {};\n    this.logEnabled = false;\n    this.usedTextures = {};\n  }\n  acquireTexture(shapeRC, usage, isPacked) {\n    const physicalTexType = getPhysicalFromLogicalTextureType(usage, isPacked);\n    const shapeKey = getKeyFromTextureShape(shapeRC, physicalTexType, isPacked);\n    if (!(shapeKey in this.freeTextures)) {\n      this.freeTextures[shapeKey] = [];\n    }\n    if (!(shapeKey in this.usedTextures)) {\n      this.usedTextures[shapeKey] = [];\n    }\n    const texBytes = computeBytes(shapeRC, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);\n    if (this.freeTextures[shapeKey].length > 0) {\n      this.numFreeTextures--;\n      this.numUsedTextures++;\n      this._numBytesFree -= texBytes;\n      this.log();\n      const newTexture2 = this.freeTextures[shapeKey].shift();\n      this.usedTextures[shapeKey].push(newTexture2);\n      return newTexture2;\n    }\n    let newTexture;\n    if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT32) {\n      newTexture = this.gpgpu.createPackedMatrixTexture(shapeRC[0], shapeRC[1]);\n    } else if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT16) {\n      newTexture = this.gpgpu.createFloat16PackedMatrixTexture(shapeRC[0], shapeRC[1]);\n    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT32) {\n      newTexture = this.gpgpu.createFloat32MatrixTexture(shapeRC[0], shapeRC[1]);\n    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT16) {\n      newTexture = this.gpgpu.createFloat16MatrixTexture(shapeRC[0], shapeRC[1]);\n    } else if (physicalTexType === PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE) {\n      newTexture = this.gpgpu.createUnsignedBytesMatrixTexture(shapeRC[0], shapeRC[1]);\n    }\n    this.usedTextures[shapeKey].push(newTexture);\n    this.numUsedTextures++;\n    this._numBytesAllocated += texBytes;\n    this.log();\n    return newTexture;\n  }\n  releaseTexture(texture, shape, logicalTexType, isPacked) {\n    if (this.freeTextures == null) {\n      return;\n    }\n    const physicalTexType = getPhysicalFromLogicalTextureType(logicalTexType, isPacked);\n    const shapeKey = getKeyFromTextureShape(shape, physicalTexType, isPacked);\n    if (!(shapeKey in this.freeTextures)) {\n      this.freeTextures[shapeKey] = [];\n    }\n    const texBytes = computeBytes(shape, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);\n    const deleteTexThreshold = env().get(\"WEBGL_DELETE_TEXTURE_THRESHOLD\");\n    if (deleteTexThreshold !== -1 && this._numBytesAllocated > deleteTexThreshold) {\n      this.gpgpu.deleteMatrixTexture(texture.texture);\n      this._numBytesAllocated -= texBytes;\n    } else {\n      this.freeTextures[shapeKey].push(texture);\n      this.numFreeTextures++;\n      this._numBytesFree += texBytes;\n    }\n    this.numUsedTextures--;\n    const texList = this.usedTextures[shapeKey];\n    const texIndex = texList.indexOf(texture);\n    if (texIndex < 0) {\n      throw new Error(\"Cannot release a texture that was never provided by this texture manager\");\n    }\n    texList.splice(texIndex, 1);\n    this.log();\n  }\n  log() {\n    if (!this.logEnabled) {\n      return;\n    }\n    const total = this.numFreeTextures + this.numUsedTextures;\n    console.log(\"Free/Used\", `${this.numFreeTextures} / ${this.numUsedTextures}`, `(${total})`);\n    const freeRatio = this._numBytesFree / this._numBytesAllocated;\n    console.log(`Bytes allocated: ${this._numBytesAllocated}`);\n    console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100 * freeRatio)}%)`);\n  }\n  get numBytesAllocated() {\n    return this._numBytesAllocated;\n  }\n  get numBytesFree() {\n    return this._numBytesFree;\n  }\n  getNumUsedTextures() {\n    return this.numUsedTextures;\n  }\n  getNumFreeTextures() {\n    return this.numFreeTextures;\n  }\n  dispose() {\n    if (this.freeTextures == null) {\n      return;\n    }\n    for (const texShape in this.freeTextures) {\n      this.freeTextures[texShape].forEach((tex) => {\n        this.gpgpu.deleteMatrixTexture(tex.texture);\n      });\n    }\n    for (const texShape in this.usedTextures) {\n      this.usedTextures[texShape].forEach((tex) => {\n        this.gpgpu.deleteMatrixTexture(tex.texture);\n      });\n    }\n    this.freeTextures = null;\n    this.usedTextures = null;\n    this.numUsedTextures = 0;\n    this.numFreeTextures = 0;\n    this._numBytesAllocated = 0;\n    this._numBytesFree = 0;\n  }\n};\nfunction numBytesForInternalFormat(gl, internalFormat) {\n  const glany = gl;\n  if (internalFormat === glany.R32F) {\n    return 4;\n  } else if (internalFormat === glany.R16F) {\n    return 2;\n  } else if (internalFormat === glany.RGBA32F) {\n    return 16;\n  } else if (internalFormat === gl.RGBA) {\n    return 16;\n  } else if (internalFormat === glany.RGBA16F) {\n    return 8;\n  } else if (internalFormat === glany.RGBA8) {\n    return 4;\n  }\n  throw new Error(`Unknown internal format ${internalFormat}`);\n}\nfunction computeBytes(shape, physicalTexType, gl, textureConfig, isPacked) {\n  const internalFormat = internalFormatForPhysicalTexType(physicalTexType, textureConfig);\n  let numElements;\n  if (isPacked) {\n    const [packedWidth, packedHeight] = getPackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);\n    numElements = packedWidth * packedHeight;\n  } else {\n    const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);\n    numElements = width * height;\n  }\n  const bytesPerElement2 = numBytesForInternalFormat(gl, internalFormat);\n  return numElements * bytesPerElement2;\n}\nfunction internalFormatForPhysicalTexType(physicalTexType, textureConfig) {\n  switch (physicalTexType) {\n    case PhysicalTextureType.PACKED_2X2_FLOAT32:\n      return getInternalFormatForPackedMatrixTexture(textureConfig);\n    case PhysicalTextureType.PACKED_2X2_FLOAT16:\n      return getInternalFormatForFloat16PackedMatrixTexture(textureConfig);\n    case PhysicalTextureType.UNPACKED_FLOAT32:\n      return getInternalFormatForFloat32MatrixTexture(textureConfig);\n    case PhysicalTextureType.UNPACKED_FLOAT16:\n      return getInternalFormatForFloat16MatrixTexture(textureConfig);\n    case PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE:\n      return getInternalFormatForUnsignedBytesMatrixTexture(textureConfig);\n    default:\n      throw new Error(`Unknown physical texture type ${physicalTexType}`);\n  }\n}\nfunction getPhysicalTextureForRendering(isPacked) {\n  if (env().getBool(\"WEBGL_RENDER_FLOAT32_ENABLED\")) {\n    if (isPacked) {\n      return PhysicalTextureType.PACKED_2X2_FLOAT32;\n    }\n    return PhysicalTextureType.UNPACKED_FLOAT32;\n  }\n  if (isPacked) {\n    return PhysicalTextureType.PACKED_2X2_FLOAT16;\n  }\n  return PhysicalTextureType.UNPACKED_FLOAT16;\n}\nfunction getPhysicalFromLogicalTextureType(logicalTexType, isPacked) {\n  if (logicalTexType === TextureUsage.UPLOAD) {\n    return PhysicalTextureType.PACKED_2X2_FLOAT32;\n  } else if (logicalTexType === TextureUsage.RENDER || logicalTexType == null) {\n    return getPhysicalTextureForRendering(isPacked);\n  } else if (logicalTexType === TextureUsage.DOWNLOAD || logicalTexType === TextureUsage.PIXELS) {\n    return PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE;\n  }\n  throw new Error(`Unknown logical texture type ${logicalTexType}`);\n}\nfunction getKeyFromTextureShape(shapeRowsCol, physicalTexType, isPacked) {\n  return `${shapeRowsCol[0]}_${shapeRowsCol[1]}_${physicalTexType}_${isPacked}`;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/unaryop_gpu.js\nvar UnaryOpProgram = class {\n  constructor(aShape, opSnippet) {\n    this.variableNames = [\"A\"];\n    this.outputShape = aShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    this.userCode = `\n      float unaryOperation(float x) {\n        ${opSnippet}\n      }\n\n      void main() {\n        float x = getAAtOutCoords();\n        float y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `;\n  }\n};\nvar CHECK_NAN_SNIPPET = `if (isnan(x)) return x;`;\nvar LINEAR = `return x;`;\nvar ABS = `return abs(x);`;\nvar ELU2 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;\nvar RELU = CHECK_NAN_SNIPPET + `\n  return (x < 0.0) ? 0.0 : x;\n`;\nvar RELU6 = CHECK_NAN_SNIPPET + `\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n`;\nvar CLONE = \"return x;\";\nvar SIGMOID = `return 1.0 / (1.0 + exp(-1.0 * x));`;\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/unaryop_packed_gpu.js\nvar LINEAR2 = `return x;`;\nvar ELU3 = `\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n`;\nvar RELU2 = `\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar RELU62 = `\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar SIGMOID2 = `return 1.0 / (1.0 + exp(-1.0 * x));`;\nvar UnaryOpPackedProgram = class {\n  constructor(aShape, opSnippet) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = aShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    this.userCode = `\n      vec4 unaryOperation(vec4 x) {\n        ${opSnippet}\n      }\n\n      void main() {\n        vec4 x = getAAtOutCoords();\n        vec4 y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/unpack_gpu.js\nvar UnpackProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = false;\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const rank = outputShape.length;\n    const channels = getChannels(\"rc\", rank);\n    const dtype = getCoordsDataType(rank);\n    const sourceCoords = getSourceCoords(rank, channels);\n    const innerDims = channels.slice(-2);\n    const coords2 = rank <= 1 ? \"rc\" : `vec2(${innerDims.join(\",\")})`;\n    this.userCode = `\n      void main() {\n        ${dtype} rc = getOutputCoords();\n        vec4 packedInput = getA(${sourceCoords});\n\n        setOutput(getChannel(packedInput, ${coords2}));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/backend_webgl.js\nvar whereImpl3 = kernel_impls_exports.whereImpl;\nvar EPSILON_FLOAT322 = 1e-7;\nvar EPSILON_FLOAT162 = 1e-4;\nvar binaryCaches = {};\nfunction getBinaryCache(webGLVersion) {\n  if (webGLVersion in binaryCaches) {\n    return binaryCaches[webGLVersion];\n  }\n  binaryCaches[webGLVersion] = {};\n  return binaryCaches[webGLVersion];\n}\nvar CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber(\"CPU_HANDOFF_SIZE_THRESHOLD\");\nvar BEFORE_PAGING_CONSTANT = 600;\nfunction numMBBeforeWarning() {\n  if (env().global.screen == null) {\n    return 1024;\n  }\n  return env().global.screen.height * env().global.screen.width * window.devicePixelRatio * BEFORE_PAGING_CONSTANT / 1024 / 1024;\n}\nvar MathBackendWebGL = class extends KernelBackend {\n  constructor(gpuResource) {\n    super();\n    this.pendingRead = /* @__PURE__ */ new WeakMap();\n    this.pendingDisposal = /* @__PURE__ */ new WeakSet();\n    this.dataRefCount = /* @__PURE__ */ new WeakMap();\n    this.numBytesInGPU = 0;\n    this.uploadWaitMs = 0;\n    this.downloadWaitMs = 0;\n    this.lastGlFlushTime = 0;\n    this.warnedAboutMemory = false;\n    this.pendingDeletes = 0;\n    this.disposed = false;\n    if (!env().getBool(\"HAS_WEBGL\")) {\n      throw new Error(\"WebGL is not supported on this device\");\n    }\n    let newGPGPU;\n    if (gpuResource != null) {\n      if (gpuResource instanceof GPGPUContext) {\n        newGPGPU = gpuResource;\n      } else {\n        const gl = getWebGLContext(env().getNumber(\"WEBGL_VERSION\"), gpuResource);\n        newGPGPU = new GPGPUContext(gl);\n      }\n      this.binaryCache = {};\n      this.gpgpuCreatedLocally = false;\n    } else {\n      const gl = getWebGLContext(env().getNumber(\"WEBGL_VERSION\"));\n      newGPGPU = new GPGPUContext(gl);\n      this.binaryCache = getBinaryCache(env().getNumber(\"WEBGL_VERSION\"));\n      this.gpgpuCreatedLocally = true;\n    }\n    this.gpgpu = newGPGPU;\n    this.canvas = this.gpgpu.gl.canvas;\n    this.textureManager = new TextureManager(this.gpgpu);\n    this.numMBBeforeWarning = numMBBeforeWarning();\n    this.texData = new DataStorage(this, engine());\n  }\n  nextDataId() {\n    return MathBackendWebGL.nextDataId++;\n  }\n  numDataIds() {\n    return this.texData.numDataIds() - this.pendingDeletes;\n  }\n  writeTexture(texture, shape, dtype, texHeight, texWidth, channels) {\n    const input2 = this.makeTensorInfo(shape, dtype);\n    const inData = this.texData.get(input2.dataId);\n    inData.isPacked = false;\n    inData.texture = { texture, texShape: [texHeight, texWidth] };\n    inData.texShape = [texHeight, texWidth];\n    const shapeAs3D = getShapeAs3D(shape);\n    const program = new EncodeMatrixProgram(shapeAs3D, false, channels);\n    const output = this.runWebGLProgram(program, [input2], dtype, [[texHeight, texWidth]]);\n    output.shape = shape;\n    inData.texture = null;\n    this.disposeIntermediateTensorInfo(input2);\n    return output.dataId;\n  }\n  write(values, shape, dtype) {\n    if (env().getBool(\"WEBGL_CHECK_NUMERICAL_PROBLEMS\") || env().getBool(\"DEBUG\")) {\n      this.checkNumericalProblems(values);\n    }\n    if (dtype === \"complex64\" && values != null) {\n      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);\n    }\n    const dataId = { id: this.nextDataId() };\n    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount: 1 });\n    return dataId;\n  }\n  refCount(dataId) {\n    if (this.texData.has(dataId)) {\n      const tensorData = this.texData.get(dataId);\n      return tensorData.refCount;\n    }\n    return 0;\n  }\n  incRef(dataId) {\n    const texData = this.texData.get(dataId);\n    texData.refCount++;\n  }\n  decRef(dataId) {\n    if (this.texData.has(dataId)) {\n      const texData = this.texData.get(dataId);\n      texData.refCount--;\n    }\n  }\n  move(dataId, values, shape, dtype, refCount) {\n    if (env().getBool(\"DEBUG\")) {\n      this.checkNumericalProblems(values);\n    }\n    if (dtype === \"complex64\") {\n      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);\n    }\n    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount });\n  }\n  disposeIntermediateTensorInfo(tensorInfo) {\n    this.disposeData(tensorInfo.dataId);\n  }\n  readSync(dataId) {\n    const texData = this.texData.get(dataId);\n    const { values, dtype, complexTensorInfos, slice: slice5, shape, isPacked } = texData;\n    if (slice5 != null) {\n      let program;\n      if (isPacked) {\n        program = new UnaryOpPackedProgram(shape, CLONE);\n      } else {\n        program = new UnaryOpProgram(shape, CLONE);\n      }\n      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);\n      const data = this.readSync(res.dataId);\n      this.disposeIntermediateTensorInfo(res);\n      return data;\n    }\n    if (values != null) {\n      return this.convertAndCacheOnCPU(dataId);\n    }\n    if (dtype === \"string\") {\n      return values;\n    }\n    const shouldTimeProgram = this.activeTimers != null;\n    let start;\n    if (shouldTimeProgram) {\n      start = util_exports.now();\n    }\n    let result;\n    if (dtype === \"complex64\") {\n      const realValues = this.readSync(complexTensorInfos.real.dataId);\n      const imagValues = this.readSync(complexTensorInfos.imag.dataId);\n      result = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);\n    } else {\n      result = this.getValuesFromTexture(dataId);\n    }\n    if (shouldTimeProgram) {\n      this.downloadWaitMs += util_exports.now() - start;\n    }\n    return this.convertAndCacheOnCPU(dataId, result);\n  }\n  async read(dataId) {\n    if (this.pendingRead.has(dataId)) {\n      const subscribers2 = this.pendingRead.get(dataId);\n      return new Promise((resolve) => subscribers2.push(resolve));\n    }\n    const texData = this.texData.get(dataId);\n    const { values, shape, slice: slice5, dtype, complexTensorInfos, isPacked } = texData;\n    if (slice5 != null) {\n      let program;\n      if (isPacked) {\n        program = new UnaryOpPackedProgram(shape, CLONE);\n      } else {\n        program = new UnaryOpProgram(shape, CLONE);\n      }\n      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);\n      const data = this.read(res.dataId);\n      this.disposeIntermediateTensorInfo(res);\n      return data;\n    }\n    if (values != null) {\n      return this.convertAndCacheOnCPU(dataId);\n    }\n    if (env().getBool(\"DEBUG\")) {\n      if (!env().getBool(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\") && env().getNumber(\"WEBGL_VERSION\") === 2) {\n        throw new Error(`tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.`);\n      }\n    }\n    let buffer2 = null;\n    let tmpDownloadTarget;\n    if (dtype !== \"complex64\" && env().get(\"WEBGL_BUFFER_SUPPORTED\")) {\n      tmpDownloadTarget = this.decode(dataId);\n      const tmpData = this.texData.get(tmpDownloadTarget.dataId);\n      buffer2 = this.gpgpu.createBufferFromTexture(tmpData.texture.texture, ...getDenseTexShape(shape));\n    }\n    this.pendingRead.set(dataId, []);\n    if (dtype !== \"complex64\") {\n      await this.gpgpu.createAndWaitForFence();\n    }\n    let vals;\n    if (dtype === \"complex64\") {\n      const ps = await Promise.all([\n        this.read(complexTensorInfos.real.dataId),\n        this.read(complexTensorInfos.imag.dataId)\n      ]);\n      const realValues = ps[0];\n      const imagValues = ps[1];\n      vals = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);\n    } else if (buffer2 == null) {\n      vals = this.getValuesFromTexture(dataId);\n    } else {\n      const size = util_exports.sizeFromShape(shape);\n      vals = this.gpgpu.downloadFloat32MatrixFromBuffer(buffer2, size);\n    }\n    if (tmpDownloadTarget != null) {\n      this.disposeIntermediateTensorInfo(tmpDownloadTarget);\n    }\n    if (buffer2 != null) {\n      const gl = this.gpgpu.gl;\n      callAndCheck(gl, () => gl.deleteBuffer(buffer2));\n    }\n    const dTypeVals = this.convertAndCacheOnCPU(dataId, vals);\n    const subscribers = this.pendingRead.get(dataId);\n    this.pendingRead.delete(dataId);\n    subscribers.forEach((resolve) => resolve(dTypeVals));\n    if (this.pendingDisposal.has(dataId)) {\n      this.pendingDisposal.delete(dataId);\n      if (this.disposeData(dataId)) {\n        engine().removeDataId(dataId, this);\n      }\n      this.pendingDeletes--;\n    }\n    return dTypeVals;\n  }\n  readToGPU(dataId, options = {}) {\n    const texData = this.texData.get(dataId);\n    const { values, shape, slice: slice5, dtype, isPacked, texture } = texData;\n    if (dtype === \"complex64\") {\n      throw new Error(\"Does not support reading texture for complex64 dtype.\");\n    }\n    if (slice5 != null) {\n      let program;\n      if (isPacked) {\n        program = new UnaryOpPackedProgram(shape, CLONE);\n      } else {\n        program = new UnaryOpProgram(shape, CLONE);\n      }\n      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);\n      const gpuResouorce = this.readToGPU(res, options);\n      this.disposeIntermediateTensorInfo(res);\n      return gpuResouorce;\n    }\n    if (texture == null) {\n      if (values != null) {\n        throw new Error(\"Data is not on GPU but on CPU.\");\n      } else {\n        throw new Error(\"There is no data on GPU or CPU.\");\n      }\n    }\n    const tmpTarget = this.decode(dataId, options.customTexShape);\n    const tensorRef = engine().makeTensorFromTensorInfo(tmpTarget);\n    const tmpData = this.texData.get(tmpTarget.dataId);\n    return Object.assign({ tensorRef }, tmpData.texture);\n  }\n  bufferSync(t) {\n    const data = this.readSync(t.dataId);\n    if (t.dtype === \"string\") {\n      try {\n        const strings = data.map((d) => util_exports.decodeString(d));\n        return buffer(t.shape, t.dtype, strings);\n      } catch (_a) {\n        throw new Error(\"Failed to decode encoded string bytes into utf-8\");\n      }\n    }\n    return buffer(t.shape, t.dtype, data);\n  }\n  checkNumericalProblems(values) {\n    if (values == null) {\n      return;\n    }\n    for (let i = 0; i < values.length; i++) {\n      const num = values[i];\n      if (!canBeRepresented(num)) {\n        if (env().getBool(\"WEBGL_RENDER_FLOAT32_CAPABLE\")) {\n          throw Error(`The value ${num} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`);\n        }\n        throw Error(`The value ${num} cannot be represented on this device.`);\n      }\n    }\n  }\n  getValuesFromTexture(dataId) {\n    const { shape, dtype, isPacked } = this.texData.get(dataId);\n    const size = util_exports.sizeFromShape(shape);\n    if (env().getBool(\"WEBGL_DOWNLOAD_FLOAT_ENABLED\")) {\n      const tmpTarget = this.decode(dataId);\n      const tmpData2 = this.texData.get(tmpTarget.dataId);\n      const vals2 = this.gpgpu.downloadMatrixFromPackedTexture(tmpData2.texture.texture, ...getDenseTexShape(shape)).subarray(0, size);\n      this.disposeIntermediateTensorInfo(tmpTarget);\n      return vals2;\n    }\n    const shouldUsePackedProgram = env().getBool(\"WEBGL_PACK\") && isPacked === true;\n    const outputShape = shouldUsePackedProgram ? getShapeAs3D(shape) : shape;\n    const program = shouldUsePackedProgram ? new EncodeFloatPackedProgram(outputShape) : new EncodeFloatProgram(outputShape);\n    const output = this.runWebGLProgram(program, [{ shape: outputShape, dtype, dataId }], \"float32\");\n    const tmpData = this.texData.get(output.dataId);\n    const vals = this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(tmpData.texture.texture, tmpData.texShape[0], tmpData.texShape[1]).subarray(0, size);\n    this.disposeIntermediateTensorInfo(output);\n    return vals;\n  }\n  timerAvailable() {\n    return env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\") > 0;\n  }\n  time(f) {\n    const oldActiveTimers = this.activeTimers;\n    const newActiveTimers = [];\n    let outerMostTime = false;\n    if (this.programTimersStack == null) {\n      this.programTimersStack = newActiveTimers;\n      outerMostTime = true;\n    } else {\n      this.activeTimers.push(newActiveTimers);\n    }\n    this.activeTimers = newActiveTimers;\n    f();\n    const flattenedActiveTimerQueries = util_exports.flatten(this.activeTimers.map((d) => d.query)).filter((d) => d != null);\n    const flattenedActiveTimerNames = util_exports.flatten(this.activeTimers.map((d) => d.name)).filter((d) => d != null);\n    this.activeTimers = oldActiveTimers;\n    if (outerMostTime) {\n      this.programTimersStack = null;\n    }\n    const res = {\n      uploadWaitMs: this.uploadWaitMs,\n      downloadWaitMs: this.downloadWaitMs,\n      kernelMs: null,\n      wallMs: null\n    };\n    return (async () => {\n      if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\") > 0) {\n        const kernelMs = await Promise.all(flattenedActiveTimerQueries);\n        res[\"kernelMs\"] = util_exports.sum(kernelMs);\n        res[\"getExtraProfileInfo\"] = () => kernelMs.map((d, i) => ({ name: flattenedActiveTimerNames[i], ms: d })).map((d) => `${d.name}: ${d.ms}`).join(\", \");\n      } else {\n        res[\"kernelMs\"] = {\n          error: \"WebGL query timers are not supported in this environment.\"\n        };\n      }\n      this.uploadWaitMs = 0;\n      this.downloadWaitMs = 0;\n      return res;\n    })();\n  }\n  memory() {\n    return {\n      unreliable: false,\n      numBytesInGPU: this.numBytesInGPU,\n      numBytesInGPUAllocated: this.textureManager.numBytesAllocated,\n      numBytesInGPUFree: this.textureManager.numBytesFree\n    };\n  }\n  startTimer() {\n    if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\") > 0) {\n      return this.gpgpu.beginQuery();\n    }\n    return { startMs: util_exports.now(), endMs: null };\n  }\n  endTimer(query) {\n    if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\") > 0) {\n      this.gpgpu.endQuery();\n      return query;\n    }\n    query.endMs = util_exports.now();\n    return query;\n  }\n  async getQueryTime(query) {\n    if (env().getNumber(\"WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE\") > 0) {\n      return this.gpgpu.waitForQueryAndGetTime(query);\n    }\n    const timerQuery = query;\n    return timerQuery.endMs - timerQuery.startMs;\n  }\n  disposeData(dataId, force = false) {\n    if (this.pendingDisposal.has(dataId)) {\n      return false;\n    }\n    if (!this.texData.has(dataId)) {\n      return true;\n    }\n    if (force) {\n      this.texData.get(dataId).refCount = 0;\n    } else {\n      this.texData.get(dataId).refCount--;\n    }\n    if (!force && this.texData.get(dataId).refCount > 0) {\n      return false;\n    }\n    if (this.pendingRead.has(dataId)) {\n      this.pendingDisposal.add(dataId);\n      this.pendingDeletes++;\n      return false;\n    }\n    this.releaseGPUData(dataId);\n    const { complexTensorInfos } = this.texData.get(dataId);\n    if (complexTensorInfos != null) {\n      this.disposeData(complexTensorInfos.real.dataId, force);\n      this.disposeData(complexTensorInfos.imag.dataId, force);\n    }\n    this.texData.delete(dataId);\n    return true;\n  }\n  releaseGPUData(dataId) {\n    const { texture, dtype, texShape, usage, isPacked, slice: slice5 } = this.texData.get(dataId);\n    const key = slice5 && slice5.origDataId || dataId;\n    const refCount = this.dataRefCount.get(key);\n    if (refCount > 1) {\n      this.dataRefCount.set(key, refCount - 1);\n    } else {\n      this.dataRefCount.delete(key);\n      if (texture != null) {\n        this.numBytesInGPU -= this.computeBytes(texShape, dtype);\n        this.textureManager.releaseTexture(texture, texShape, usage, isPacked);\n      }\n    }\n    const texData = this.texData.get(dataId);\n    texData.texture = null;\n    texData.texShape = null;\n    texData.isPacked = false;\n    texData.slice = null;\n  }\n  getTexture(dataId) {\n    this.uploadToGPU(dataId);\n    return this.texData.get(dataId).texture.texture;\n  }\n  getDataInfo(dataId) {\n    return this.texData.get(dataId);\n  }\n  shouldExecuteOnCPU(inputs, sizeThreshold = CPU_HANDOFF_SIZE_THRESHOLD) {\n    return env().getBool(\"WEBGL_CPU_FORWARD\") && inputs.every((input2) => this.texData.get(input2.dataId).texture == null && util_exports.sizeFromShape(input2.shape) < sizeThreshold);\n  }\n  getGPGPUContext() {\n    return this.gpgpu;\n  }\n  where(condition) {\n    backend_util_exports.warn(\"tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead\");\n    const condVals = condition.dataSync();\n    return whereImpl3(condition.shape, condVals);\n  }\n  packedUnaryOp(x, op2, dtype) {\n    const program = new UnaryOpPackedProgram(x.shape, op2);\n    const outInfo = this.compileAndRun(program, [x], dtype);\n    return engine().makeTensorFromTensorInfo(outInfo);\n  }\n  abs(x) {\n    if (this.shouldExecuteOnCPU([x]) && x.dtype !== \"complex64\") {\n      const outValues = simpleAbsImplCPU(this.texData.get(x.dataId).values);\n      return this.makeOutput(x.shape, x.dtype, outValues);\n    }\n    if (env().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")) {\n      return this.packedUnaryOp(x, ABS, x.dtype);\n    }\n    const program = new UnaryOpProgram(x.shape, ABS);\n    const outInfo = this.compileAndRun(program, [x]);\n    return engine().makeTensorFromTensorInfo(outInfo);\n  }\n  makeTensorInfo(shape, dtype, values) {\n    let dataId;\n    if (dtype === \"string\" && values != null && values.length > 0 && util_exports.isString(values[0])) {\n      const encodedValues = values.map((d) => util_exports.encodeString(d));\n      dataId = this.write(encodedValues, shape, dtype);\n    } else {\n      dataId = this.write(values, shape, dtype);\n    }\n    this.texData.get(dataId).usage = null;\n    return { dataId, shape, dtype };\n  }\n  makeOutput(shape, dtype, values) {\n    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);\n  }\n  unpackTensor(input2) {\n    const program = new UnpackProgram(input2.shape);\n    return this.runWebGLProgram(program, [input2], input2.dtype);\n  }\n  packTensor(input2) {\n    const program = new PackProgram(input2.shape);\n    const preventEagerUnpackingOutput = true;\n    return this.runWebGLProgram(program, [input2], input2.dtype, null, preventEagerUnpackingOutput);\n  }\n  packedReshape(input2, afterShape) {\n    const input3DShape = [\n      getBatchDim(input2.shape),\n      ...getRowsCols(input2.shape)\n    ];\n    const input3D = {\n      dtype: input2.dtype,\n      shape: input3DShape,\n      dataId: input2.dataId\n    };\n    const afterShapeAs3D = [\n      getBatchDim(afterShape),\n      ...getRowsCols(afterShape)\n    ];\n    const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);\n    const preventEagerUnpackingOfOutput = true;\n    const customValues = [input3DShape];\n    const output = this.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);\n    return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };\n  }\n  decode(dataId, customTexShape) {\n    const texData = this.texData.get(dataId);\n    const { isPacked, shape, dtype } = texData;\n    if (customTexShape != null) {\n      const size = util_exports.sizeFromShape(shape);\n      const texSize = customTexShape[0] * customTexShape[1] * 4;\n      util_exports.assert(size <= texSize, () => \"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.\");\n    }\n    const shapeAs3D = getShapeAs3D(shape);\n    let program;\n    if (isPacked) {\n      program = new DecodeMatrixPackedProgram(shapeAs3D);\n    } else {\n      program = new DecodeMatrixProgram(shapeAs3D);\n    }\n    const preventEagerUnpackingOfOutput = true;\n    const customValues = [customTexShape != null ? customTexShape : getDenseTexShape(shapeAs3D)];\n    const out = this.runWebGLProgram(program, [{ shape: shapeAs3D, dtype, dataId }], dtype, customValues, preventEagerUnpackingOfOutput, customTexShape);\n    return { dtype, shape, dataId: out.dataId };\n  }\n  runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false, customTexShape) {\n    const output = this.makeTensorInfo(program.outputShape, outputDtype);\n    const outData = this.texData.get(output.dataId);\n    if (program.packedOutput) {\n      outData.isPacked = true;\n    }\n    if (program.outPackingScheme === PackingScheme.DENSE) {\n      const texelShape = customTexShape != null ? customTexShape : getDenseTexShape(program.outputShape);\n      outData.texShape = texelShape.map((d) => d * 2);\n    }\n    if (program.outTexUsage != null) {\n      outData.usage = program.outTexUsage;\n    }\n    if (util_exports.sizeFromShape(output.shape) === 0) {\n      outData.values = util_exports.getTypedArrayFromDType(output.dtype, 0);\n      return output;\n    }\n    const dataToDispose = [];\n    const inputsData = inputs.map((input2) => {\n      if (input2.dtype === \"complex64\") {\n        throw new Error(`GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.`);\n      }\n      let texData = this.texData.get(input2.dataId);\n      if (texData.texture == null) {\n        if (!program.packedInputs && util_exports.sizeFromShape(input2.shape) <= env().getNumber(\"WEBGL_SIZE_UPLOAD_UNIFORM\")) {\n          return {\n            shape: input2.shape,\n            texData: null,\n            isUniform: true,\n            uniformValues: texData.values\n          };\n        }\n        if (program.packedInputs) {\n          texData.isPacked = true;\n          texData.shape = input2.shape;\n        }\n      }\n      this.uploadToGPU(input2.dataId);\n      if (!!texData.isPacked !== !!program.packedInputs) {\n        input2 = texData.isPacked ? this.unpackTensor(input2) : this.packTensor(input2);\n        dataToDispose.push(input2);\n        texData = this.texData.get(input2.dataId);\n      } else if (texData.isPacked && !isReshapeFree(texData.shape, input2.shape)) {\n        const savedInput = input2;\n        const targetShape = input2.shape;\n        input2.shape = texData.shape;\n        input2 = this.packedReshape(input2, targetShape);\n        dataToDispose.push(input2);\n        texData = this.texData.get(input2.dataId);\n        savedInput.shape = targetShape;\n      }\n      return { shape: input2.shape, texData, isUniform: false };\n    });\n    this.uploadToGPU(output.dataId);\n    const outputData = { shape: output.shape, texData: outData, isUniform: false };\n    const key = makeShaderKey(program, inputsData, outputData);\n    const binary = this.getAndSaveBinary(key, () => {\n      return compileProgram(this.gpgpu, program, inputsData, outputData);\n    });\n    const shouldTimeProgram = this.activeTimers != null;\n    let query;\n    if (shouldTimeProgram) {\n      query = this.startTimer();\n    }\n    if (!env().get(\"ENGINE_COMPILE_ONLY\")) {\n      runProgram(this.gpgpu, binary, inputsData, outputData, customUniformValues);\n    }\n    dataToDispose.forEach((info) => this.disposeIntermediateTensorInfo(info));\n    if (shouldTimeProgram) {\n      query = this.endTimer(query);\n      this.activeTimers.push({ name: program.constructor.name, query: this.getQueryTime(query) });\n    }\n    const glFlushThreshold = env().get(\"WEBGL_FLUSH_THRESHOLD\");\n    if (glFlushThreshold > 0) {\n      const time2 = util_exports.now();\n      if (time2 - this.lastGlFlushTime > glFlushThreshold) {\n        this.gpgpu.gl.flush();\n        this.lastGlFlushTime = time2;\n      }\n    }\n    if (!env().getBool(\"WEBGL_LAZILY_UNPACK\") && outData.isPacked && preventEagerUnpackingOfOutput === false) {\n      const unpacked = this.unpackTensor(output);\n      this.disposeIntermediateTensorInfo(output);\n      return unpacked;\n    }\n    return output;\n  }\n  compileAndRun(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false) {\n    outputDtype = outputDtype || inputs[0].dtype;\n    const outInfo = this.runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput);\n    return outInfo;\n  }\n  getAndSaveBinary(key, getBinary) {\n    if (!(key in this.binaryCache)) {\n      this.binaryCache[key] = getBinary();\n    }\n    return this.binaryCache[key];\n  }\n  getTextureManager() {\n    return this.textureManager;\n  }\n  dispose() {\n    if (this.disposed) {\n      return;\n    }\n    if (!env().getBool(\"IS_TEST\")) {\n      const allKeys = Object.keys(this.binaryCache);\n      allKeys.forEach((key) => {\n        this.gpgpu.deleteProgram(this.binaryCache[key].webGLProgram);\n        delete this.binaryCache[key];\n      });\n    }\n    this.textureManager.dispose();\n    if (this.canvas != null && (typeof HTMLCanvasElement !== \"undefined\" && this.canvas instanceof HTMLCanvasElement)) {\n      this.canvas.remove();\n    } else {\n      this.canvas = null;\n    }\n    if (this.gpgpuCreatedLocally) {\n      this.gpgpu.program = null;\n      this.gpgpu.dispose();\n    }\n    this.disposed = true;\n  }\n  floatPrecision() {\n    if (this.floatPrecisionValue == null) {\n      this.floatPrecisionValue = tidy(() => {\n        if (!env().get(\"WEBGL_RENDER_FLOAT32_ENABLED\")) {\n          const debugFlag = env().getBool(\"DEBUG\");\n          env().set(\"DEBUG\", false);\n          const underflowCheckValue = this.abs(scalar(1e-8)).dataSync()[0];\n          env().set(\"DEBUG\", debugFlag);\n          if (underflowCheckValue > 0) {\n            return 32;\n          }\n        }\n        return 16;\n      });\n    }\n    return this.floatPrecisionValue;\n  }\n  epsilon() {\n    return this.floatPrecision() === 32 ? EPSILON_FLOAT322 : EPSILON_FLOAT162;\n  }\n  uploadToGPU(dataId) {\n    const texData = this.texData.get(dataId);\n    const { shape, dtype, values, texture, usage, isPacked } = texData;\n    if (texture != null) {\n      return;\n    }\n    const shouldTimeProgram = this.activeTimers != null;\n    let start;\n    if (shouldTimeProgram) {\n      start = util_exports.now();\n    }\n    let texShape = texData.texShape;\n    if (texShape == null) {\n      texShape = getTextureShapeFromLogicalShape(shape, isPacked);\n      texData.texShape = texShape;\n    }\n    if (values != null) {\n      const shapeAs3D = getShapeAs3D(shape);\n      let program;\n      let width = texShape[1], height = texShape[0];\n      const isByteArray = values instanceof Uint8Array || values instanceof Uint8ClampedArray;\n      if (isPacked || !isByteArray) {\n        [width, height] = getPackedMatrixTextureShapeWidthHeight(texShape[0], texShape[1]);\n      }\n      if (isPacked) {\n        program = new EncodeMatrixPackedProgram(shapeAs3D, isByteArray);\n      } else {\n        program = new EncodeMatrixProgram(shapeAs3D, isByteArray);\n      }\n      const tempDenseInputTexShape = isByteArray ? [height, width] : texShape;\n      const tempDenseInputHandle = this.makeTensorInfo(tempDenseInputTexShape, dtype);\n      const tempDenseInputTexData = this.texData.get(tempDenseInputHandle.dataId);\n      if (isByteArray) {\n        tempDenseInputTexData.usage = TextureUsage.PIXELS;\n      } else {\n        tempDenseInputTexData.usage = TextureUsage.UPLOAD;\n      }\n      tempDenseInputTexData.texShape = tempDenseInputTexShape;\n      this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(tempDenseInputHandle.dataId), width, height, values);\n      const customValues = [[height, width]];\n      const preventEagerUnpacking = true;\n      const encodedOutputTarget = this.runWebGLProgram(program, [tempDenseInputHandle], dtype, customValues, preventEagerUnpacking);\n      const outputTexData = this.texData.get(encodedOutputTarget.dataId);\n      texData.texShape = outputTexData.texShape;\n      texData.isPacked = outputTexData.isPacked;\n      texData.usage = outputTexData.usage;\n      if (!env().get(\"ENGINE_COMPILE_ONLY\")) {\n        texData.texture = outputTexData.texture;\n        texData.values = null;\n        this.texData.delete(encodedOutputTarget.dataId);\n      } else {\n        this.disposeData(encodedOutputTarget.dataId);\n      }\n      this.disposeIntermediateTensorInfo(tempDenseInputHandle);\n      if (shouldTimeProgram) {\n        this.uploadWaitMs += util_exports.now() - start;\n      }\n    } else {\n      const newTexture = this.acquireTexture(texShape, usage, dtype, isPacked);\n      texData.texture = newTexture;\n    }\n  }\n  convertAndCacheOnCPU(dataId, float32Values) {\n    const texData = this.texData.get(dataId);\n    const { dtype } = texData;\n    this.releaseGPUData(dataId);\n    if (float32Values != null) {\n      texData.values = float32ToTypedArray(float32Values, dtype);\n    }\n    return texData.values;\n  }\n  acquireTexture(texShape, texType, dtype, isPacked) {\n    this.numBytesInGPU += this.computeBytes(texShape, dtype);\n    if (!this.warnedAboutMemory && this.numBytesInGPU > this.numMBBeforeWarning * 1024 * 1024) {\n      const mb = (this.numBytesInGPU / 1024 / 1024).toFixed(2);\n      this.warnedAboutMemory = true;\n      console.warn(`High memory usage in GPU: ${mb} MB, most likely due to a memory leak`);\n    }\n    return this.textureManager.acquireTexture(texShape, texType, isPacked);\n  }\n  computeBytes(shape, dtype) {\n    return shape[0] * shape[1] * util_exports.bytesPerElement(dtype);\n  }\n  checkCompileCompletion() {\n    for (const [, binary] of Object.entries(this.binaryCache)) {\n      this.checkCompletion_(binary);\n    }\n  }\n  async checkCompileCompletionAsync() {\n    const ps = [];\n    if (this.gpgpu.parallelCompilationExtension) {\n      for (const [, binary] of Object.entries(this.binaryCache)) {\n        ps.push(this.checkCompletionAsync_(binary));\n      }\n      return Promise.all(ps);\n    } else {\n      for (const [, binary] of Object.entries(this.binaryCache)) {\n        const p2 = new Promise((resolve) => {\n          try {\n            this.checkCompletion_(binary);\n            resolve(true);\n          } catch (error) {\n            throw error;\n          }\n        });\n        ps.push(p2);\n      }\n      return Promise.all(ps);\n    }\n  }\n  async checkCompletionAsync_(binary) {\n    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)) {\n      return this.checkCompletion_(binary);\n    } else {\n      await nextFrame();\n      return this.checkCompletionAsync_(binary);\n    }\n  }\n  checkCompletion_(binary) {\n    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.gl.LINK_STATUS) === false) {\n      console.log(this.gpgpu.gl.getProgramInfoLog(binary.webGLProgram));\n      if (this.gpgpu.gl.getShaderParameter(binary.fragmentShader, this.gpgpu.gl.COMPILE_STATUS) === false) {\n        logShaderSourceAndInfoLog(binary.source, this.gpgpu.gl.getShaderInfoLog(binary.fragmentShader));\n        throw new Error(\"Failed to compile fragment shader.\");\n      }\n      throw new Error(\"Failed to link vertex and fragment shaders.\");\n    }\n    return true;\n  }\n  getUniformLocations() {\n    for (const [, binary] of Object.entries(this.binaryCache)) {\n      const { uniformLocations, customUniformLocations, infLoc, nanLoc, inShapesLocations, inTexShapesLocations, outShapeLocation, outShapeStridesLocation, outTexShapeLocation } = getUniformLocations(this.gpgpu, binary.program, binary.webGLProgram);\n      binary.uniformLocations = uniformLocations;\n      binary.customUniformLocations = customUniformLocations;\n      binary.infLoc = infLoc;\n      binary.nanLoc = nanLoc;\n      binary.inShapesLocations = inShapesLocations;\n      binary.inTexShapesLocations = inTexShapesLocations;\n      binary.outShapeLocation = outShapeLocation;\n      binary.outShapeStridesLocation = outShapeStridesLocation;\n      binary.outTexShapeLocation = outTexShapeLocation;\n    }\n  }\n  createTensorFromTexture(values, shape, dtype) {\n    const { texture, height, width, channels } = values;\n    const backend2 = engine().backend;\n    if (!backend2.gpgpu.gl.isTexture(texture)) {\n      throw new Error(`The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.`);\n    }\n    const dataId = backend2.writeTexture(texture, shape, dtype, height, width, channels);\n    return engine().makeTensorFromDataId(dataId, shape, dtype, backend2);\n  }\n};\nMathBackendWebGL.nextDataId = 0;\nfunction float32ToTypedArray(a, dtype) {\n  if (dtype === \"float32\" || dtype === \"complex64\") {\n    return a;\n  } else if (dtype === \"int32\" || dtype === \"bool\") {\n    const result = dtype === \"int32\" ? new Int32Array(a.length) : new Uint8Array(a.length);\n    for (let i = 0; i < result.length; ++i) {\n      result[i] = Math.round(a[i]);\n    }\n    return result;\n  } else {\n    throw new Error(`Unknown dtype ${dtype}`);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/version.js\nvar version6 = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl.js\nfunction forceHalfFloat() {\n  env().set(\"WEBGL_FORCE_F16_TEXTURES\", true);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/base.js\nif (device_util_exports.isBrowser()) {\n  registerBackend(\"webgl\", () => new MathBackendWebGL(), 2);\n}\nvar webgl = { forceHalfFloat };\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_gpu.js\nvar CHECK_NAN_SNIPPET2 = `\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n`;\nvar BinaryOpProgram = class {\n  constructor(op2, aShape, bShape) {\n    this.variableNames = [\"A\", \"B\"];\n    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    this.userCode = `\n      float binaryOperation(float a, float b) {\n        ${op2}\n      }\n\n      void main() {\n        float a = getAAtOutCoords();\n        float b = getBAtOutCoords();\n        setOutput(binaryOperation(a, b));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_packed_gpu.js\nvar CHECK_NAN_SNIPPET_PACKED = `\n  result.r = isNaN.r ? NAN : result.r;\n  result.g = isNaN.g ? NAN : result.g;\n  result.b = isNaN.b ? NAN : result.b;\n  result.a = isNaN.a ? NAN : result.a;\n`;\nvar BinaryOpPackedProgram = class {\n  constructor(op2, aShape, bShape, checkOutOfBounds = false) {\n    this.variableNames = [\"A\", \"B\"];\n    this.supportsBroadcasting = true;\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);\n    const rank = this.outputShape.length;\n    this.enableShapeUniforms = useShapeUniforms(rank);\n    let checkOutOfBoundsString = \"\";\n    if (checkOutOfBounds) {\n      if (rank === 0 || util_exports.sizeFromShape(this.outputShape) === 1) {\n        checkOutOfBoundsString = `\n          result.y = 0.;\n          result.z = 0.;\n          result.w = 0.;\n        `;\n      } else {\n        const dtype = getCoordsDataType(rank);\n        checkOutOfBoundsString = `\n          ${dtype} coords = getOutputCoords();\n        `;\n        if (rank === 1) {\n          if (this.enableShapeUniforms) {\n            checkOutOfBoundsString += `\n            result.y = (coords + 1) >= outShape ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;\n          } else {\n            checkOutOfBoundsString += `\n            result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;\n          }\n        } else {\n          const channels = getChannels(\"coords\", rank);\n          if (this.enableShapeUniforms) {\n            checkOutOfBoundsString += `\n            bool nextRowOutOfBounds =\n              (${channels[rank - 2]} + 1) >= outShape[${rank} - 2];\n            bool nextColOutOfBounds =\n              (${channels[rank - 1]} + 1) >= outShape[${rank} - 1];\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `;\n          } else {\n            checkOutOfBoundsString += `\n            bool nextRowOutOfBounds =\n              (${channels[rank - 2]} + 1) >= ${this.outputShape[rank - 2]};\n            bool nextColOutOfBounds =\n              (${channels[rank - 1]} + 1) >= ${this.outputShape[rank - 1]};\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `;\n          }\n        }\n      }\n    }\n    this.userCode = `\n      vec4 binaryOperation(vec4 a, vec4 b) {\n        ${op2}\n      }\n\n      void main() {\n        vec4 a = getAAtOutCoords();\n        vec4 b = getBAtOutCoords();\n\n        vec4 result = binaryOperation(a, b);\n        ${checkOutOfBoundsString}\n\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Identity.js\nfunction identity3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  backend2.incRef(x.dataId);\n  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };\n}\nvar identityConfig2 = {\n  kernelName: Identity,\n  backendName: \"webgl\",\n  kernelFunc: identity3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Complex.js\nfunction complex3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { real: real4, imag: imag4 } = inputs;\n  const complexInfo = backend2.makeTensorInfo(real4.shape, \"complex64\");\n  const complex4 = backend2.texData.get(complexInfo.dataId);\n  const realTensorInfo = identity3({ inputs: { x: real4 }, backend: backend2 });\n  const imagTensorInfo = identity3({ inputs: { x: imag4 }, backend: backend2 });\n  complex4.complexTensorInfos = { real: realTensorInfo, imag: imagTensorInfo };\n  return complexInfo;\n}\nvar complexConfig2 = {\n  kernelName: Complex,\n  backendName: \"webgl\",\n  kernelFunc: complex3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LeakyRelu.js\nvar LEAKYRELU = `return (a < 0.) ? b * a : a;`;\nvar LEAKYRELU_PACKED = `\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n`;\nfunction leakyRelu3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { alpha } = attrs;\n  const $alpha = backend2.makeTensorInfo([], \"float32\", util_exports.createScalarValue(alpha, \"float32\"));\n  const program = env().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\") ? new BinaryOpPackedProgram(LEAKYRELU_PACKED, x.shape, $alpha.shape) : new BinaryOpProgram(LEAKYRELU, x.shape, $alpha.shape);\n  const result = backend2.runWebGLProgram(program, [x, $alpha], \"float32\");\n  backend2.disposeIntermediateTensorInfo($alpha);\n  return result;\n}\nvar leakyReluConfig2 = {\n  kernelName: LeakyRelu,\n  backendName: \"webgl\",\n  kernelFunc: leakyRelu3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Prelu.js\nvar PRELU = `return (a < 0.) ? b * a : a;`;\nvar PRELU_PACKED = `\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n`;\nfunction prelu4(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x, alpha } = inputs;\n  const program = env().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\") ? new BinaryOpPackedProgram(PRELU_PACKED, x.shape, alpha.shape) : new BinaryOpProgram(PRELU, x.shape, alpha.shape);\n  return backend2.runWebGLProgram(program, [x, alpha], \"float32\");\n}\nvar preluConfig2 = {\n  kernelName: Prelu,\n  backendName: \"webgl\",\n  kernelFunc: prelu4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/kernel_funcs_utils.js\nvar CHECK_NAN_SNIPPET_UNARY = `if (isnan(x)) return x;`;\nfunction unaryKernelFunc2({ opSnippet, packedOpSnippet, cpuKernelImpl, dtype }) {\n  return ({ inputs, backend: backend2 }) => {\n    const { x } = inputs;\n    const webglBackend = backend2;\n    const $dtype = dtype || x.dtype;\n    if (webglBackend.shouldExecuteOnCPU([x]) && cpuKernelImpl != null) {\n      const xData = webglBackend.texData.get(x.dataId);\n      const outValues = cpuKernelImpl(xData.values, $dtype);\n      return webglBackend.makeTensorInfo(x.shape, $dtype, outValues);\n    }\n    const shouldUsePackedProgram = env().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\") && packedOpSnippet != null;\n    let program;\n    if (shouldUsePackedProgram) {\n      program = new UnaryOpPackedProgram(x.shape, packedOpSnippet);\n    } else {\n      program = new UnaryOpProgram(x.shape, opSnippet);\n    }\n    return webglBackend.runWebGLProgram(program, [x], $dtype);\n  };\n}\nfunction binaryKernelFunc2({ opSnippet, packedOpSnippet, checkOutOfBounds = false, supportsComplex = false, cpuKernelImpl, dtype }) {\n  return ({ inputs, backend: backend2 }) => {\n    const { a, b } = inputs;\n    const webglBackend = backend2;\n    if (supportsComplex && a.dtype === \"complex64\") {\n      const aData = webglBackend.texData.get(a.dataId);\n      const bData = webglBackend.texData.get(b.dataId);\n      const [real4, imag4] = [\n        [aData.complexTensorInfos.real, bData.complexTensorInfos.real],\n        [aData.complexTensorInfos.imag, bData.complexTensorInfos.imag]\n      ].map((complexParts) => {\n        const [aPart, bPart] = complexParts;\n        const aHandle = {\n          dataId: aPart.dataId,\n          dtype: aPart.dtype,\n          shape: a.shape\n        };\n        const bHandle = {\n          dataId: bPart.dataId,\n          dtype: bPart.dtype,\n          shape: b.shape\n        };\n        const program2 = new BinaryOpProgram(opSnippet, a.shape, b.shape);\n        return webglBackend.runWebGLProgram(program2, [aHandle, bHandle], upcastType(aPart.dtype, bPart.dtype));\n      });\n      const complexOutput = complex3({ inputs: { real: real4, imag: imag4 }, backend: webglBackend });\n      webglBackend.disposeIntermediateTensorInfo(real4);\n      webglBackend.disposeIntermediateTensorInfo(imag4);\n      return complexOutput;\n    }\n    const $dtype = dtype || upcastType(a.dtype, b.dtype);\n    if ((a.dtype === \"string\" || b.dtype === \"string\" || webglBackend.shouldExecuteOnCPU([a, b])) && cpuKernelImpl != null) {\n      const aVals = webglBackend.texData.get(a.dataId).values;\n      const bVals = webglBackend.texData.get(b.dataId).values;\n      const decodedAVals = a.dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;\n      const decodedBVals = a.dtype === \"string\" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;\n      const [outValues, outShape] = cpuKernelImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);\n      const out = webglBackend.makeTensorInfo(outShape, $dtype);\n      const outData = webglBackend.texData.get(out.dataId);\n      outData.values = outValues;\n      return out;\n    }\n    const shouldUsePackedProgram = env().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\") && packedOpSnippet != null;\n    let program;\n    if (shouldUsePackedProgram) {\n      program = new BinaryOpPackedProgram(packedOpSnippet, a.shape, b.shape, checkOutOfBounds);\n    } else {\n      program = new BinaryOpProgram(opSnippet, a.shape, b.shape);\n    }\n    return webglBackend.runWebGLProgram(program, [a, b], $dtype);\n  };\n}\nfunction mapActivationToShaderProgram(activation2, packed = false) {\n  if (activation2 === \"linear\") {\n    if (packed) {\n      return LINEAR2;\n    }\n    return LINEAR;\n  } else if (activation2 === \"relu\") {\n    if (packed) {\n      return RELU2;\n    }\n    return RELU;\n  } else if (activation2 === \"elu\") {\n    if (packed) {\n      return ELU3;\n    }\n    return ELU2;\n  } else if (activation2 === \"relu6\") {\n    if (packed) {\n      return RELU62;\n    }\n    return RELU6;\n  } else if (activation2 === \"prelu\") {\n    if (packed) {\n      return PRELU_PACKED;\n    }\n    return PRELU;\n  } else if (activation2 === \"leakyrelu\") {\n    if (packed) {\n      return LEAKYRELU_PACKED;\n    }\n    return LEAKYRELU;\n  } else if (activation2 === \"sigmoid\") {\n    if (packed) {\n      return SIGMOID2;\n    }\n    return SIGMOID;\n  }\n  throw new Error(`Activation ${activation2} has not been implemented for the WebGL backend.`);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/mulmat_packed_gpu.js\nvar MatMulPackedProgram = class {\n  constructor(aShape, bShape, outputShape, transposeA = false, transposeB = false, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyreluActivation = false) {\n    this.variableNames = [\"matrixA\", \"matrixB\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const sharedDim = transposeA ? aShape[1] : aShape[2];\n    const sharedDimensionPacked = Math.ceil(sharedDim / 2);\n    const aSample = transposeA ? \"i * 2, rc.y\" : \"rc.y, i * 2\";\n    const bSample = transposeB ? \"rc.z, i * 2\" : \"i * 2, rc.z\";\n    const aSwizzle = transposeA ? [\"a.xxyy\", \"a.zzww\"] : [\"a.xxzz\", \"a.yyww\"];\n    const bSwizzle = transposeB ? [\"b.xzxz\", \"b.ywyw\"] : [\"b.xyxy\", \"b.zwzw\"];\n    let activationSnippet = \"\", applyActivationSnippet = \"\";\n    if (activation2) {\n      if (hasPreluActivation) {\n        activationSnippet = `vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${activation2}\n        }`;\n      } else if (hasLeakyreluActivation) {\n        activationSnippet = `vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${activation2}\n        }`;\n      } else {\n        activationSnippet = `vec4 activation(vec4 x) {\n          ${activation2}\n        }`;\n      }\n      applyActivationSnippet = `result = activation(result);`;\n    }\n    const addBiasSnippet = addBias ? \"result += getBiasAtOutCoords();\" : \"\";\n    if (addBias) {\n      this.variableNames.push(\"bias\");\n    }\n    if (hasPreluActivation) {\n      this.variableNames.push(\"preluActivationWeights\");\n    }\n    if (hasLeakyreluActivation) {\n      this.variableNames.push(\"leakyreluAlpha\");\n    }\n    let batchASnippet = \"rc.x\";\n    let batchBSnippet = \"rc.x\";\n    if (aShape[0] < bShape[0]) {\n      batchASnippet = `int(min(float(rc.x), ${aShape[0] - 1}.))`;\n    } else if (bShape[0] < aShape[0]) {\n      batchBSnippet = `int(min(float(rc.x), ${bShape[0] - 1}.))`;\n    }\n    this.userCode = `\n      ${activationSnippet}\n      // Don't use uniform for sharedDimensionPacked for performance.\n      const float sharedDimension = ${sharedDimensionPacked}.0;\n\n      vec4 dot2x2ARowBCol(ivec3 rc) {\n        vec4 result = vec4(0);\n        for (int i = 0; i < ${sharedDimensionPacked}; i++) {\n          int batchA = ${batchASnippet};\n          int batchB = ${batchBSnippet};\n          vec4 a = getMatrixA(batchA, ${aSample});\n          vec4 b = getMatrixB(batchB, ${bSample});\n\n          // These swizzled products need to be separately added.\n          // See: https://github.com/tensorflow/tfjs/issues/1735\n          result += (${aSwizzle[0]} * ${bSwizzle[0]});\n          result += (${aSwizzle[1]} * ${bSwizzle[1]});\n        }\n        return result;\n      }\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n        vec4 result = dot2x2ARowBCol(rc);\n\n        ${addBiasSnippet}\n\n        ${applyActivationSnippet}\n\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_complex_gpu.js\nvar COMPLEX_MULTIPLY = {\n  REAL: \"return areal * breal - aimag * bimag;\",\n  IMAG: \"return areal * bimag + aimag * breal;\"\n};\nvar BinaryOpComplexProgram = class {\n  constructor(op2, aShape, bShape) {\n    this.variableNames = [\"AReal\", \"AImag\", \"BReal\", \"BImag\"];\n    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);\n    this.userCode = `\n      float binaryOpComplex(\n          float areal, float aimag, float breal, float bimag) {\n        ${op2}\n      }\n\n      void main() {\n        float areal = getARealAtOutCoords();\n        float aimag = getAImagAtOutCoords();\n        float breal = getBRealAtOutCoords();\n        float bimag = getBImagAtOutCoords();\n        setOutput(binaryOpComplex(areal, aimag, breal, bimag));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Multiply.js\nvar MUL = \"return a * b;\";\nfunction multiply3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { a, b } = inputs;\n  const dtype = backend_util_exports.upcastType(a.dtype, b.dtype);\n  if (a.dtype === \"complex64\") {\n    const aData = backend2.texData.get(a.dataId);\n    const bData = backend2.texData.get(b.dataId);\n    const realProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.REAL, a.shape, b.shape);\n    const imagProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.IMAG, a.shape, b.shape);\n    const inputs2 = [\n      {\n        dataId: aData.complexTensorInfos.real.dataId,\n        dtype: aData.complexTensorInfos.real.dtype,\n        shape: a.shape\n      },\n      {\n        dataId: aData.complexTensorInfos.imag.dataId,\n        dtype: aData.complexTensorInfos.imag.dtype,\n        shape: a.shape\n      },\n      {\n        dataId: bData.complexTensorInfos.real.dataId,\n        dtype: bData.complexTensorInfos.real.dtype,\n        shape: b.shape\n      },\n      {\n        dataId: bData.complexTensorInfos.imag.dataId,\n        dtype: bData.complexTensorInfos.imag.dtype,\n        shape: b.shape\n      }\n    ];\n    const realPart = backend2.runWebGLProgram(realProgram, inputs2, \"float32\");\n    const imagPart = backend2.runWebGLProgram(imagProgram, inputs2, \"float32\");\n    const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    backend2.disposeIntermediateTensorInfo(imagPart);\n    return complexOutput;\n  }\n  if (backend2.shouldExecuteOnCPU([a, b])) {\n    const aData = backend2.texData.get(a.dataId);\n    const bData = backend2.texData.get(b.dataId);\n    const [outValues, outShape] = multiplyImplCPU(a.shape, b.shape, aData.values, bData.values, dtype);\n    const out = backend2.makeTensorInfo(outShape, dtype);\n    const outData = backend2.texData.get(out.dataId);\n    outData.values = outValues;\n    return out;\n  }\n  let program;\n  if (env().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\")) {\n    program = new BinaryOpPackedProgram(MUL, a.shape, b.shape);\n  } else {\n    program = new BinaryOpProgram(MUL, a.shape, b.shape);\n  }\n  return backend2.runWebGLProgram(program, [a, b], dtype);\n}\nvar multiplyConfig2 = {\n  kernelName: Multiply,\n  backendName: \"webgl\",\n  kernelFunc: multiply3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/reshape.js\nfunction packedReshape(input2, afterShape, backend2) {\n  const input3DShape = [\n    getBatchDim(input2.shape),\n    ...getRowsCols(input2.shape)\n  ];\n  const input3D = {\n    dtype: input2.dtype,\n    shape: input3DShape,\n    dataId: input2.dataId\n  };\n  const afterShapeAs3D = [\n    getBatchDim(afterShape),\n    ...getRowsCols(afterShape)\n  ];\n  const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);\n  const preventEagerUnpackingOfOutput = true;\n  const customValues = [input3DShape];\n  const output = backend2.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);\n  return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reshape.js\nfunction reshape4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { shape } = attrs;\n  const webglBackend = backend2;\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const $shape = util_exports.inferFromImplicitShape(shape, xSize);\n  const $xSize = util_exports.sizeFromShape($shape);\n  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);\n  const xTexData = webglBackend.texData.get(x.dataId);\n  if (xTexData.isPacked && !isReshapeFree(x.shape, $shape) && !(xTexData.texture !== null && isReshapeFree(xTexData.shape, $shape))) {\n    return packedReshape(x, $shape, webglBackend);\n  }\n  webglBackend.incRef(x.dataId);\n  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };\n}\nvar reshapeConfig2 = {\n  kernelName: Reshape,\n  backendName: \"webgl\",\n  kernelFunc: reshape4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/mean_gpu.js\nvar MeanProgram = class {\n  constructor(reduceInfo, divisor) {\n    this.variableNames = [\"x\"];\n    const { windowSize, batchSize, inSize, outSize } = reduceInfo;\n    this.outputShape = [batchSize, outSize];\n    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;\n    const windowSizeVec4Remainder = windowSize % 4;\n    let updateSnippet = `sumValue += dot(values, ones);`;\n    if (divisor != null) {\n      const denominator = 1 / divisor;\n      updateSnippet = `sumValue += dot(values * ${util_exports.isInt(denominator) ? denominator.toPrecision(2) : denominator}, ones);`;\n    }\n    let checkOutOfBounds = \"\";\n    if (inSize % windowSize > 0) {\n      checkOutOfBounds = `\n        if (inIdx < 0 || inIdx >= ${inSize}) {\n          return 0.0;\n        }\n      `;\n    }\n    this.userCode = `\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${checkOutOfBounds}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${windowSize};\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${updateSnippet}\n        }\n\n        int inIdx = inOffset + ${windowSizeNearestVec4};\n        if (${windowSizeVec4Remainder === 1}) {\n          vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 2}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1), 0.0, 0.0);\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 3}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2), 0.0);\n\n          ${updateSnippet}\n        }\n        setOutput(sumValue);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/reduce_gpu.js\nvar ReduceProgram = class {\n  constructor(reduceInfo, reduceType) {\n    this.variableNames = [\"x\"];\n    const { windowSize, batchSize, inSize, outSize } = reduceInfo;\n    this.outputShape = [batchSize, outSize];\n    let initializationValue = \"0.0\";\n    let compareOp = ``;\n    if (reduceType === \"prod\") {\n      initializationValue = \"1.0\";\n    } else if (reduceType === \"min\") {\n      initializationValue = \"1.0 / 1e-20\";\n      compareOp = `min`;\n    } else if (reduceType === \"max\") {\n      initializationValue = \"-1.0 / 1e-20\";\n      compareOp = `max`;\n    }\n    let returnValue = `${reduceType}(${reduceType}(${reduceType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;\n    if (reduceType === \"sum\") {\n      returnValue = `sumValue`;\n    } else if (reduceType === \"prod\") {\n      returnValue = `prodValue`;\n    } else if (reduceType === \"all\") {\n      returnValue = `allValue`;\n    } else if (reduceType === \"any\") {\n      returnValue = `anyValue`;\n    }\n    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;\n    const windowSizeVec4Remainder = windowSize % 4;\n    let updateSnippet = `\n      if (${reduceType === \"sum\"}) {\n        sumValue += dot(values, ones);\n      } else if (${reduceType === \"prod\"}) {\n        vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);\n        prodValue *= tmp[0] * tmp[1];\n      } else {\n        minMaxValue = ${compareOp}(values, minMaxValue);\n        if (${reduceType === \"min\"} || ${reduceType === \"max\"}) {\n          minMaxValue = ${compareOp}(values, minMaxValue);\n          bvec4 isNaN = isnan(values);\n          if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {\n            minMaxValue = vec4(NAN);\n          }\n        }\n      }\n    `;\n    let vecType = `vec4`;\n    if (reduceType === \"all\") {\n      initializationValue = \"1.0\";\n      updateSnippet = `\n        bool reducedAllValue = all(values);\n        float floatedReducedAllValue = float(reducedAllValue);\n        allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);\n      `;\n      vecType = `bvec4`;\n    } else if (reduceType === \"any\") {\n      initializationValue = \"0.0\";\n      updateSnippet = `\n        bool reducedAnyValue = any(values);\n        float floatedReducedAnyValue = float(reducedAnyValue);\n        anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);\n      `;\n      vecType = `bvec4`;\n    }\n    let checkOutOfBounds = \"\";\n    if (inSize % windowSize > 0) {\n      checkOutOfBounds = `\n        if (inIdx < 0 || inIdx >= ${inSize}) {\n          return initializationValue;\n        }\n      `;\n    }\n    this.userCode = `\n      const float initializationValue = ${initializationValue};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${checkOutOfBounds}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${windowSize};\n\n        vec4 minMaxValue = vec4(${initializationValue});\n        float prodValue = 1.0;\n        float sumValue = 0.0;\n        float allValue = 1.0;\n        float anyValue = 0.0;\n\n        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {\n          int inIdx = inOffset + i;\n          ${vecType} values = ${vecType}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${updateSnippet}\n        }\n\n        int inIdx = inOffset + ${windowSizeNearestVec4};\n        if (${windowSizeVec4Remainder === 1}) {\n          ${vecType} values = ${vecType}(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 2}) {\n          ${vecType} values = ${vecType}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 3}) {\n          ${vecType} values = ${vecType}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          ${updateSnippet}\n        }\n        setOutput(${returnValue});\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/reduce.js\nfunction getReductionStages(inShape) {\n  const stages = [];\n  while (stages.length === 0 || stages[stages.length - 1].outSize !== 1) {\n    const outSize = stages.length ? stages[stages.length - 1].outSize : inShape[1];\n    const windowSize = backend_util_exports.computeOptimalWindowSize(outSize);\n    stages.push({\n      inSize: outSize,\n      windowSize,\n      outSize: Math.ceil(outSize / windowSize)\n    });\n  }\n  return stages;\n}\nfunction reduce(x, dtype, reductionType, backend2) {\n  const reductionStages = getReductionStages(x.shape);\n  let result = x;\n  for (let i = 0; i < reductionStages.length; i++) {\n    const { inSize, windowSize, outSize } = reductionStages[i];\n    let program;\n    let previousResult;\n    if (reductionType === \"mean\") {\n      program = i === 0 ? new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, inSize) : new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize });\n    } else {\n      program = new ReduceProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, reductionType);\n    }\n    previousResult = result;\n    result = backend2.runWebGLProgram(program, [result], dtype);\n    if (previousResult.dataId !== x.dataId) {\n      backend2.disposeIntermediateTensorInfo(previousResult);\n    }\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/transpose_gpu.js\nvar TransposeProgram = class {\n  constructor(aShape, newDim) {\n    this.variableNames = [\"A\"];\n    const outputShape = new Array(aShape.length);\n    for (let i = 0; i < outputShape.length; i++) {\n      outputShape[i] = aShape[newDim[i]];\n    }\n    this.outputShape = outputShape;\n    this.rank = outputShape.length;\n    const dtype = getCoordsDataType(this.rank);\n    const switched = getSwitchedCoords(newDim);\n    this.userCode = `\n    void main() {\n      ${dtype} resRC = getOutputCoords();\n      setOutput(getA(${switched}));\n    }\n    `;\n  }\n};\nfunction getSwitchedCoords(newDim) {\n  const rank = newDim.length;\n  if (rank > 6) {\n    throw Error(`Transpose for rank ${rank} is not yet supported`);\n  }\n  const originalOrder = [\"resRC.x\", \"resRC.y\", \"resRC.z\", \"resRC.w\", \"resRC.u\", \"resRC.v\"];\n  const switchedCoords = new Array(rank);\n  for (let i = 0; i < newDim.length; i++) {\n    switchedCoords[newDim[i]] = originalOrder[i];\n  }\n  return switchedCoords.join();\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/transpose_packed_gpu.js\nvar TransposePackedProgram = class {\n  constructor(aShape, newDim) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    const outputShape = new Array(aShape.length);\n    for (let i = 0; i < outputShape.length; i++) {\n      outputShape[i] = aShape[newDim[i]];\n    }\n    this.outputShape = outputShape;\n    this.rank = outputShape.length;\n    if (this.rank > 6) {\n      throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);\n    }\n    const dtype = getCoordsDataType(this.rank);\n    const outputOrder = getVecChannels(\"rc\", this.rank);\n    const switchedOrder = new Array(this.rank);\n    for (let i = 0; i < newDim.length; i++) {\n      switchedOrder[newDim[i]] = outputOrder[i];\n    }\n    const innerDims = `vec2(${switchedOrder.slice(-2).join()})`;\n    const nextColumn = `++${outputOrder[this.rank - 1]} < ${outputShape[this.rank - 1]}`;\n    const getc = `getChannel(getA(${switchedOrder.join()}), ${innerDims})`;\n    this.userCode = `\n    void main() {\n      ${dtype} rc = getOutputCoords();\n      vec4 result = vec4(0.);\n      result[0] = ${getc};\n      if(${nextColumn}) {\n        result[1] = ${getc};\n      }\n      --${outputOrder[this.rank - 1]};\n      if(++${outputOrder[this.rank - 2]} < ${outputShape[this.rank - 2]}) {\n        result[2] = ${getc};\n        if(${nextColumn}) {\n          result[3] = ${getc};\n        }\n      }\n      setOutput(result);\n    }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transpose_impl.js\nfunction transposeImpl2(x, perm, backend2) {\n  const program = env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") ? new TransposePackedProgram(x.shape, perm) : new TransposeProgram(x.shape, perm);\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sum_impl.js\nfunction sumImpl(x, axis, keepDims, backend2) {\n  const reductionIndices = axis;\n  const xRank = x.shape.length;\n  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  const sumInputIsTransposed = permutedAxes != null;\n  let sumInput = x;\n  if (sumInputIsTransposed) {\n    sumInput = transposeImpl2(x, permutedAxes, backend2);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"sum\", axes, xRank);\n  const [sumOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(sumInput.shape, axes);\n  let outShape = sumOutShape;\n  if (keepDims) {\n    outShape = backend_util_exports.expandShapeToKeepDim(sumOutShape, origAxes);\n  }\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const batchSize = xSize / inSize;\n  const reshapedInput = reshape4({ inputs: { x: sumInput }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });\n  const outType = sumOutType(x.dtype);\n  const reduced = reduce(reshapedInput, outType, \"sum\", backend2);\n  const out = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(reshapedInput);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  if (sumInputIsTransposed) {\n    backend2.disposeIntermediateTensorInfo(sumInput);\n  }\n  return out;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sum.js\nfunction sum4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  return sumImpl(x, axis, keepDims, backend2);\n}\nvar sumConfig2 = {\n  kernelName: Sum,\n  backendName: \"webgl\",\n  kernelFunc: sum4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transpose.js\nfunction transpose3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { perm } = attrs;\n  const webglBackend = backend2;\n  const xRank = x.shape.length;\n  const newShape = new Array(xRank);\n  for (let i = 0; i < newShape.length; i++) {\n    newShape[i] = x.shape[perm[i]];\n  }\n  let out;\n  if (webglBackend.shouldExecuteOnCPU([x])) {\n    const xTexData = webglBackend.texData.get(x.dataId);\n    const values = xTexData.values;\n    const outValues = transposeImplCPU(values, x.shape, x.dtype, perm, newShape);\n    out = webglBackend.makeTensorInfo(newShape, x.dtype);\n    const outData = webglBackend.texData.get(out.dataId);\n    outData.values = outValues;\n  } else {\n    out = transposeImpl2(x, perm, webglBackend);\n  }\n  return out;\n}\nvar transposeConfig2 = {\n  kernelName: Transpose,\n  backendName: \"webgl\",\n  kernelFunc: transpose3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchMatMul_impl.js\nvar MATMUL_SHARED_DIM_THRESHOLD = 1e3;\nfunction batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {\n  const aRank = a.shape.length;\n  const bRank = b.shape.length;\n  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];\n  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];\n  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];\n  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];\n  const outerDimsA = a.shape.slice(0, -2);\n  const outerDimsB = b.shape.slice(0, -2);\n  const batchDimA = util_exports.sizeFromShape(outerDimsA);\n  const batchDimB = util_exports.sizeFromShape(outerDimsB);\n  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));\n  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);\n  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);\n  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];\n  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];\n  const a3d = reshape4({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });\n  const b3d = reshape4({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });\n  const intermediates = [a3d, b3d];\n  const batchDim = Math.max(batchDimA, batchDimB);\n  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];\n  const hasBias = bias != null;\n  const hasPreluActivationWeights = preluActivationWeights != null;\n  const hasLeakyreluAlpha = activation2 === \"leakyrelu\";\n  const fusedActivation = activation2 != null ? mapActivationToShaderProgram(activation2, true) : null;\n  const containsFusedOps = hasBias || hasPreluActivationWeights || hasLeakyreluAlpha || fusedActivation != null;\n  let out;\n  if ((outerShapeA === 1 || outerShapeB === 1) && sharedDim > MATMUL_SHARED_DIM_THRESHOLD && containsFusedOps === false) {\n    let aVec = a3d;\n    let bVec = b3d;\n    if (transposeA) {\n      aVec = transpose3({ inputs: { x: a3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });\n      intermediates.push(aVec);\n    }\n    if (transposeB) {\n      bVec = transpose3({ inputs: { x: b3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });\n      intermediates.push(bVec);\n    }\n    const shouldReshapeA = outerShapeB !== 1;\n    const shouldReshapeB = outerShapeB === 1;\n    let aVec3d = aVec;\n    if (shouldReshapeA) {\n      aVec3d = reshape4({\n        inputs: { x: aVec },\n        backend: backend2,\n        attrs: { shape: [batchDim, sharedDim, 1] }\n      });\n      intermediates.push(aVec3d);\n    }\n    const axis = outerShapeB === 1 ? 2 : 1;\n    let bVec3d = bVec;\n    if (shouldReshapeB) {\n      bVec3d = reshape4({\n        inputs: { x: bVec },\n        backend: backend2,\n        attrs: { shape: [batchDim, 1, sharedDim] }\n      });\n      intermediates.push(bVec3d);\n    }\n    const product = multiply3({ inputs: { a: aVec3d, b: bVec3d }, backend: backend2 });\n    out = sum4({ inputs: { x: product }, backend: backend2, attrs: { axis, keepDims: true } });\n    intermediates.push(product);\n  } else {\n    const dtype = upcastType(a.dtype, b.dtype);\n    const program = new MatMulPackedProgram(a3dShape, b3dShape, [batchDim, outerShapeA, outerShapeB], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n    const inputs = [a3d, b3d];\n    if (bias != null) {\n      inputs.push(bias);\n    }\n    if (hasPreluActivationWeights) {\n      inputs.push(preluActivationWeights);\n    }\n    if (hasLeakyreluAlpha) {\n      const $leakyreluAlpha = backend2.makeTensorInfo([], \"float32\", util_exports.createScalarValue(leakyreluAlpha, \"float32\"));\n      inputs.push($leakyreluAlpha);\n      intermediates.push($leakyreluAlpha);\n    }\n    out = backend2.runWebGLProgram(program, inputs, dtype);\n  }\n  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: outShape } });\n  intermediates.push(out);\n  for (const i of intermediates) {\n    backend2.disposeIntermediateTensorInfo(i);\n  }\n  return outReshaped;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/_FusedMatMul.js\nfunction _fusedMatMul2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b, bias, preluActivationWeights } = inputs;\n  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;\n  return batchMatMulImpl({\n    a,\n    b,\n    transposeA,\n    transposeB,\n    backend: backend2,\n    bias,\n    preluActivationWeights,\n    leakyreluAlpha,\n    activation: activation2\n  });\n}\nvar _fusedMatMulConfig2 = {\n  kernelName: _FusedMatMul,\n  backendName: \"webgl\",\n  kernelFunc: _fusedMatMul2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Abs.js\nvar ABS2 = `return abs(x);`;\nfunction abs3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (backend2.shouldExecuteOnCPU([x]) && x.dtype !== \"complex64\") {\n    const xData = backend2.texData.get(x.dataId);\n    const outValues = simpleAbsImplCPU(xData.values);\n    return backend2.makeTensorInfo(x.shape, x.dtype, outValues);\n  }\n  let program;\n  if (env().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")) {\n    program = new UnaryOpPackedProgram(x.shape, ABS2);\n  } else {\n    program = new UnaryOpProgram(x.shape, ABS2);\n  }\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\nvar absConfig2 = {\n  kernelName: Abs,\n  backendName: \"webgl\",\n  kernelFunc: abs3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Acos.js\nvar ACOS = CHECK_NAN_SNIPPET + `\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return acos(x);\n`;\nvar acos3 = unaryKernelFunc2({ opSnippet: ACOS });\nvar acosConfig2 = {\n  kernelName: Acos,\n  backendName: \"webgl\",\n  kernelFunc: acos3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Acosh.js\nvar ACOSH = CHECK_NAN_SNIPPET + `\n  if (x < 1.0) return NAN;\nreturn log(x + sqrt(x * x - 1.0));`;\nvar acosh3 = unaryKernelFunc2({ opSnippet: ACOSH });\nvar acoshConfig2 = {\n  kernelName: Acosh,\n  backendName: \"webgl\",\n  kernelFunc: acosh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Add.js\nvar ADD = \"return a + b;\";\nvar addKernelFunc = binaryKernelFunc2({\n  opSnippet: ADD,\n  packedOpSnippet: ADD,\n  supportsComplex: true,\n  cpuKernelImpl: addImplCPU\n});\nvar addConfig2 = {\n  kernelName: Add,\n  backendName: \"webgl\",\n  kernelFunc: addKernelFunc\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/addn_gpu.js\nvar AddNProgram = class {\n  constructor(outputShape, shapes) {\n    this.outputShape = [];\n    this.outputShape = outputShape;\n    this.variableNames = shapes.map((_, i) => `T${i}`);\n    const snippets = [];\n    this.variableNames.forEach((variable2) => {\n      snippets.push(`float v${variable2} = get${variable2}AtOutCoords();`);\n    });\n    const operation = this.variableNames.map((variable2) => {\n      return `v${variable2}`;\n    }).join(\" + \");\n    this.userCode = `\n      void main() {\n        ${snippets.join(\"\\n        \")}\n\n        float result = ${operation};\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/addn_packed_gpu.js\nvar AddNPackedProgram = class {\n  constructor(outputShape, shapes) {\n    this.outputShape = [];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = outputShape;\n    this.variableNames = shapes.map((_, i) => `T${i}`);\n    const snippets = [];\n    this.variableNames.forEach((variable2) => {\n      snippets.push(`vec4 v${variable2} = get${variable2}AtOutCoords();`);\n    });\n    const operation = this.variableNames.map((variable2) => {\n      return `v${variable2}`;\n    }).join(\" + \");\n    this.userCode = `\n      void main() {\n        ${snippets.join(\"\\n        \")}\n\n        vec4 result = ${operation};\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AddN.js\nfunction addN3(args) {\n  const { inputs, backend: backend2 } = args;\n  const tensors = inputs;\n  if (tensors.length === 1) {\n    return identity3({ inputs: { x: tensors[0] }, backend: backend2 });\n  }\n  if (tensors.length > env().get(\"WEBGL_MAX_TEXTURES_IN_SHADER\")) {\n    const midIndex = Math.floor(tensors.length / 2);\n    const leftSide = addN3({ inputs: tensors.slice(0, midIndex), backend: backend2 });\n    const rightSide = addN3({ inputs: tensors.slice(midIndex), backend: backend2 });\n    return addN3({ inputs: [leftSide, rightSide], backend: backend2 });\n  }\n  const dtype = tensors.map((t) => t.dtype).reduce((d1, d2) => upcastType(d1, d2));\n  const shapes = tensors.map((t) => t.shape);\n  const usePackedOp = env().getBool(\"WEBGL_PACK\");\n  const program = usePackedOp ? new AddNPackedProgram(tensors[0].shape, shapes) : new AddNProgram(tensors[0].shape, shapes);\n  return backend2.runWebGLProgram(program, tensors, dtype);\n}\nvar addNConfig2 = {\n  kernelName: AddN,\n  backendName: \"webgl\",\n  kernelFunc: addN3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/All.js\nfunction all3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  const xRank = x.shape.length;\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let permutedX = x;\n  if (permutedAxes != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"all\", axes, xRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });\n  const reduced = reduce(a2D, a2D.dtype, \"all\", backend2);\n  let res;\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });\n  } else {\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });\n  }\n  backend2.disposeIntermediateTensorInfo(a2D);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo(permutedX);\n  }\n  return res;\n}\nvar allConfig2 = {\n  kernelName: All,\n  backendName: \"webgl\",\n  kernelFunc: all3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Any.js\nfunction any3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  const xRank = x.shape.length;\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let permutedX = x;\n  if (permutedAxes != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"any\", axes, xRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });\n  const reduced = reduce(a2D, a2D.dtype, \"any\", backend2);\n  let res;\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });\n  } else {\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });\n  }\n  backend2.disposeIntermediateTensorInfo(a2D);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo(permutedX);\n  }\n  return res;\n}\nvar anyConfig2 = {\n  kernelName: Any,\n  backendName: \"webgl\",\n  kernelFunc: any3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/argminmax_gpu.js\nvar ArgMinMaxProgram = class {\n  constructor(reduceInfo, op2, firstPass) {\n    this.variableNames = [\"A\"];\n    const { windowSize, batchSize, outSize } = reduceInfo;\n    if (!firstPass) {\n      this.variableNames.push(\"bestIndicesA\");\n    }\n    this.outputShape = [batchSize, outSize];\n    const compOp = op2 === \"max\" ? \">\" : \"<\";\n    const indexSnippet = firstPass ? \"inOffset + i;\" : \"round(getBestIndicesA(batch, inOffset + i));\";\n    this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${windowSize};\n\n        int bestIndex = inOffset;\n        float bestValue = getA(batch, bestIndex);\n\n        for (int i = 0; i < ${windowSize}; i++) {\n          int inIdx = ${indexSnippet};\n          float candidate = getA(batch, inIdx);\n          if (candidate ${compOp} bestValue) {\n            bestValue = candidate;\n            bestIndex = inIdx;\n          }\n        }\n        setOutput(float(bestIndex));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/argminmax_packed_gpu.js\nvar ArgMinMaxPackedProgram = class {\n  constructor(shape, windowSize, op2, firstPass) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    util_exports.assert(shape.length > 2, () => `Packed arg${op2.charAt(0).toUpperCase() + op2.slice(1)} supports only inputs with rank above 2.`);\n    const inSize = shape[shape.length - 1];\n    const outSize = Math.ceil(inSize / windowSize);\n    this.outputShape = shape.slice(0, -1);\n    if (outSize > 1) {\n      this.outputShape.push(outSize);\n    }\n    if (!firstPass) {\n      this.variableNames.push(\"bestIndicesA\");\n    }\n    const outShape = this.outputShape;\n    const rank = outShape.length;\n    const dtype = getCoordsDataType(rank);\n    const coords2 = getChannels(\"coords\", rank);\n    let sourceLocSetup;\n    let sourceRank;\n    if (outSize === 1) {\n      sourceRank = rank + 1;\n      const sourceLocDType = getCoordsDataType(sourceRank);\n      sourceLocSetup = `\n        ${sourceLocDType} sourceLocR = ${sourceLocDType}(${coords2.join()}, 0);\n        ++${coords2[rank - 1]};\n        ${sourceLocDType} sourceLocG = ${sourceLocDType}(${coords2.join()}, 0);\n        ++${coords2[rank - 2]};\n        ${sourceLocDType} sourceLocA = ${sourceLocDType}(${coords2.join()}, 0);\n        --${coords2[rank - 1]};\n        ${sourceLocDType} sourceLocB = ${sourceLocDType}(${coords2.join()}, 0);\n        --${coords2[rank - 2]};`;\n    } else {\n      sourceRank = rank;\n      sourceLocSetup = `\n        ${dtype} sourceLocR = coords;\n        ++${coords2[rank - 1]};\n        ${dtype} sourceLocG = coords;\n        ++${coords2[rank - 2]};\n        ${dtype} sourceLocA = coords;\n        --${coords2[rank - 1]};\n        ${dtype} sourceLocB = coords;\n        --${coords2[rank - 2]};`;\n    }\n    const channels = [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"].slice(0, sourceRank);\n    const inChannel = \".\" + channels[sourceRank - 1];\n    const intChannels = channels.map((x) => \"int \" + x);\n    const srcRCoords = getChannels(\"sourceLocR\", sourceRank - 1).concat(\"inIdx.r\");\n    const srcGCoords = getChannels(\"sourceLocG\", sourceRank - 1).concat(\"inIdx.g\");\n    const srcBCoords = getChannels(\"sourceLocB\", sourceRank - 1).concat(\"inIdx.b\");\n    const srcACoords = getChannels(\"sourceLocA\", sourceRank - 1).concat(\"inIdx.a\");\n    const compOp = op2 === \"max\" ? \"greaterThan\" : \"lessThan\";\n    const fetchCandidateIdx = firstPass ? \"\" : `\n          inIdx = round(vec4(getBestIndicesAChannel(${srcRCoords.join()}),\n                             getBestIndicesAChannel(${srcGCoords.join()}),\n                             getBestIndicesAChannel(${srcBCoords.join()}),\n                             getBestIndicesAChannel(${srcACoords.join()})));`;\n    const fetchValue = `vec4(\n            getAChannel(${srcRCoords.join()}),\n            hasNextCol ? getAChannel(${srcGCoords.join()}) : 0.,\n            hasNextRow ? getAChannel(${srcBCoords.join()}) : 0.,\n            hasNextRow && hasNextCol ? getAChannel(${srcACoords.join()}) : 0.)`;\n    const getBestIndicesAChannelSnippet = firstPass ? \"\" : `\n      float getBestIndicesAChannel(${intChannels.join()}) {\n        return getChannel(getBestIndicesA(${channels.join()}),\n                                          vec2(${channels.slice(-2).join()}));\n      }`;\n    this.userCode = `\n      float getAChannel(${intChannels.join()}) {\n        return getChannel(getA(${channels.join()}),\n                               vec2(${channels.slice(-2).join()}));\n      }\n      ${getBestIndicesAChannelSnippet}\n      void main() {\n        ${dtype} coords = getOutputCoords();\n        bool hasNextCol = ${coords2[rank - 1]} < ${outShape[rank - 1] - 1};\n        bool hasNextRow = ${coords2[rank - 2]} < ${outShape[rank - 2] - 1};\n        ${sourceLocSetup}\n        ivec4 srcIdx = ivec4(sourceLocR${inChannel}, sourceLocG${inChannel},\n          sourceLocB${inChannel}, sourceLocA${inChannel}) * ${windowSize};\n        ivec4 inIdx = srcIdx;\n        vec4 bestIndex = vec4(inIdx);\n        vec4 bestValue = ${fetchValue};\n\n        for (int i = 0; i < ${windowSize}; i++) {\n          inIdx = srcIdx;\n          ${fetchCandidateIdx}\n          vec4 candidate = ${fetchValue};\n          bvec4 nan = isnan(candidate);\n          bvec4 replace = bvec4(\n            vec4(${compOp}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));\n\n          bestValue = vec4(replace.x  ? candidate.x : bestValue.x,\n                           replace.y  ? candidate.y : bestValue.y,\n                           replace.z  ? candidate.z : bestValue.z,\n                           replace.w  ? candidate.w : bestValue.w);\n          bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));\n          srcIdx++;\n        }\n        setOutput(bestIndex);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/arg_min_max.js\nfunction argReduce(backend2, x, reduceType, bestIndicesA = null) {\n  let batchSize = x.shape[0];\n  let inSize = x.shape[1];\n  if (bestIndicesA != null) {\n    batchSize = bestIndicesA.shape[0];\n    inSize = bestIndicesA.shape[1];\n  }\n  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);\n  const reduceInfo = { windowSize, inSize, batchSize, outSize: Math.ceil(inSize / windowSize) };\n  const program = new ArgMinMaxProgram(reduceInfo, reduceType, bestIndicesA == null);\n  const inputs = [x];\n  if (bestIndicesA != null) {\n    inputs.push(bestIndicesA);\n  }\n  const output = backend2.runWebGLProgram(program, inputs, \"int32\");\n  if (output.shape[1] === 1) {\n    return output;\n  }\n  const result = argReduce(backend2, x, reduceType, output);\n  backend2.disposeIntermediateTensorInfo(output);\n  return result;\n}\nfunction argReducePacked(backend2, x, reduceType, bestIndicesA = null) {\n  const inShape = bestIndicesA != null ? bestIndicesA.shape : x.shape;\n  const inSize = inShape[inShape.length - 1];\n  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);\n  const program = new ArgMinMaxPackedProgram(inShape, windowSize, reduceType, bestIndicesA == null);\n  const inputs = bestIndicesA == null ? [x] : [x, bestIndicesA];\n  const output = backend2.runWebGLProgram(program, inputs, \"int32\");\n  if (output.shape.length === x.shape.length) {\n    const result = argReducePacked(backend2, x, reduceType, output);\n    backend2.disposeIntermediateTensorInfo(output);\n    return result;\n  }\n  return output;\n}\nfunction argMinMaxReduce(backend2, x, axis, reduceType) {\n  const axes = [axis];\n  backend_util_exports.assertAxesAreInnerMostDims(\"arg\" + reduceType.charAt(0).toUpperCase() + reduceType.slice(1), axes, x.shape.length);\n  if (!env().getBool(\"WEBGL_PACK_REDUCE\") || x.shape.length <= 2) {\n    const intermediateTensorInfos = [];\n    const xtexData = backend2.texData.get(x.dataId);\n    const xIsPacked = xtexData !== null && xtexData.isPacked;\n    let xUnPacked = x;\n    if (xIsPacked) {\n      xUnPacked = backend2.unpackTensor(x);\n      intermediateTensorInfos.push(xUnPacked);\n    }\n    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xUnPacked.shape, axes);\n    const inSize = util_exports.sizeFromShape(reduceShape);\n    const a2D = reshape4({ inputs: { x: xUnPacked }, backend: backend2, attrs: { shape: [-1, inSize] } });\n    intermediateTensorInfos.push(a2D);\n    const reduced = argReduce(backend2, a2D, reduceType);\n    intermediateTensorInfos.push(reduced);\n    const reshaped = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });\n    intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n    return reshaped;\n  }\n  return argReducePacked(backend2, x, reduceType);\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ArgMax.js\nfunction argMax3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis } = attrs;\n  let axes = util_exports.parseAxisParam(axis, x.shape);\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  const intermediateTensorInfos = [];\n  if (permutedAxes != null) {\n    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    intermediateTensorInfos.push($x);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"argMax\", [axes[0]], $x.shape.length);\n  const out = argMinMaxReduce(backend2, $x, axes[0], \"max\");\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return out;\n}\nvar argMaxConfig2 = {\n  kernelName: ArgMax,\n  backendName: \"webgl\",\n  kernelFunc: argMax3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ArgMin.js\nfunction argMin3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis } = attrs;\n  let axes = util_exports.parseAxisParam(axis, x.shape);\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);\n  let $x = x;\n  const intermediateTensorInfos = [];\n  if (permutedAxes != null) {\n    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    intermediateTensorInfos.push($x);\n    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"argMin\", [axes[0]], $x.shape.length);\n  const out = argMinMaxReduce(backend2, $x, axes[0], \"min\");\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return out;\n}\nvar argMinConfig2 = {\n  kernelName: ArgMin,\n  backendName: \"webgl\",\n  kernelFunc: argMin3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Asin.js\nvar ASIN = CHECK_NAN_SNIPPET + `\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return asin(x);\n`;\nvar asin3 = unaryKernelFunc2({ opSnippet: ASIN });\nvar asinConfig2 = {\n  kernelName: Asin,\n  backendName: \"webgl\",\n  kernelFunc: asin3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Asinh.js\nvar ASINH = CHECK_NAN_SNIPPET + `return log(x + sqrt(x * x + 1.0));`;\nvar asinh3 = unaryKernelFunc2({ opSnippet: ASINH });\nvar asinhConfig2 = {\n  kernelName: Asinh,\n  backendName: \"webgl\",\n  kernelFunc: asinh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atan.js\nvar ATAN = CHECK_NAN_SNIPPET + `\n  return atan(x);\n`;\nvar atan4 = unaryKernelFunc2({ opSnippet: ATAN });\nvar atanConfig2 = {\n  kernelName: Atan,\n  backendName: \"webgl\",\n  kernelFunc: atan4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atan2.js\nvar ATAN2 = CHECK_NAN_SNIPPET2 + `\n  return atan(a, b);\n`;\nvar ATAN2_PACKED = `\n  vec4 result = atan(a, b);\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  ` + CHECK_NAN_SNIPPET_PACKED + `\n  return result;\n`;\nvar atan23 = binaryKernelFunc2({ opSnippet: ATAN2, packedOpSnippet: ATAN2_PACKED });\nvar atan2Config2 = {\n  kernelName: Atan2,\n  backendName: \"webgl\",\n  kernelFunc: atan23\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atanh.js\nvar ATANH = CHECK_NAN_SNIPPET + `\n  if ((x < -1.0) || (x > 1.0)) return NAN;\nreturn (log(1.0 + x) - log(1.0 - x)) / 2.0;`;\nvar atanh3 = unaryKernelFunc2({ opSnippet: ATANH });\nvar atanhConfig2 = {\n  kernelName: Atanh,\n  backendName: \"webgl\",\n  kernelFunc: atanh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/pool_gpu.js\nvar Pool2DProgram = class {\n  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {\n    this.variableNames = [\"x\"];\n    if (poolType === \"avg\" && computePositions) {\n      throw new Error(\"Cannot compute positions for average pool.\");\n    }\n    const filterWidth = convInfo.filterWidth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    this.outputShape = convInfo.outShape;\n    const isAvgPool = poolType === \"avg\";\n    const batchFlattenPositionStr = `((batch  * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;\n    const flattenPositionStr = `(xR * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;\n    let initializationValue = \"0.0\";\n    if (!isAvgPool) {\n      initializationValue = \"-1.0 / 1e-20\";\n    }\n    if (computePositions) {\n      const compareOp2 = \">=\";\n      this.userCode = `\n        const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});\n        const ivec2 pads = ivec2(${padTop}, ${padLeft});\n\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int batch = coords[0];\n          int d = coords[3];\n\n          ivec2 xRCCorner = coords.yz * strides - pads;\n          int xRCorner = xRCCorner.x;\n          int xCCorner = xRCCorner.y;\n\n          // max/min x(?, ?, d) to get y(yR, yC, d).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n          float avgValue = 0.0;\n\n          for (int wR = 0; wR < ${effectiveFilterHeight};\n              wR += ${dilationHeight}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${convInfo.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${effectiveFilterWidth};\n                wC += ${dilationWidth}) {\n              int xC = xCCorner + wC;\n\n              if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                continue;\n              }\n\n              float value = getX(batch, xR, xC, d);\n\n              // If a min / max value has already been found, use it. If not,\n              // use the current value.\n              float currMinMaxValue = mix(\n                  value, minMaxValue, minMaxValueFound);\n              if (value ${compareOp2} currMinMaxValue) {\n                minMaxValue = value;\n                minMaxValueFound = 1.0;\n                minMaxPosition = ${flattenPositions ? includeBatchInIndex ? batchFlattenPositionStr : flattenPositionStr : `wR * ${effectiveFilterWidth} + wC`};\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `;\n      return;\n    }\n    const compareOp = \"max\";\n    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;\n    if (poolType === \"avg\") {\n      returnValue = `avgValue / count`;\n    }\n    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;\n    const filterWidthVec4Remainder = filterWidth % 4;\n    const updateSnippet = `\n      if (${isAvgPool}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = ${compareOp}(values, minMaxValue);\n      }\n    `;\n    this.userCode = `\n      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n      const float initializationValue = ${initializationValue};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xR, int xC, int d) {\n        if (xC < 0 || xC >= ${convInfo.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xR, xC, d);\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d = coords[3];\n\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // max/min x(?, ?, d) to get y(yR, yC, d).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${initializationValue});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wR = 0; wR < ${effectiveFilterHeight};\n            wR += ${dilationHeight}) {\n          int xR = xRCorner + wR;\n\n          if (xR < 0 || xR >= ${convInfo.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {\n            int xC = xCCorner + wC * ${dilationWidth};\n\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${dilationWidth}, d),\n              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),\n              getValue(batch, xR, xC + 3 * ${dilationWidth}, d)\n            );\n\n            ${updateSnippet}\n          }\n\n          int xC = xCCorner + ${filterWidthNearestVec4};\n          if (${filterWidthVec4Remainder === 1}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              initializationValue,\n              initializationValue,\n              initializationValue\n            );\n\n            ${updateSnippet}\n          } else if (${filterWidthVec4Remainder === 2}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${dilationWidth}, d),\n              initializationValue,\n              initializationValue\n            );\n\n            ${updateSnippet}\n          } else if (${filterWidthVec4Remainder === 3}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${dilationWidth}, d),\n              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),\n              initializationValue\n            );\n\n            ${updateSnippet}\n          }\n        }\n        setOutput(${returnValue});\n      }\n    `;\n  }\n};\nvar Pool3DProgram = class {\n  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {\n    this.variableNames = [\"x\"];\n    if (poolType === \"avg\" && computePositions) {\n      throw new Error(\"Cannot compute positions for average pool.\");\n    }\n    const filterWidth = convInfo.filterWidth;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationDepth = convInfo.dilationDepth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padFront = convInfo.padInfo.front;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    this.outputShape = convInfo.outShape;\n    const isAvgPool = poolType === \"avg\";\n    let initializationValue = \"0.0\";\n    if (!isAvgPool) {\n      initializationValue = \"-1.0 / 1e-20\";\n    }\n    if (computePositions) {\n      const compareOp2 = \">=\";\n      this.userCode = `\n        const ivec3 strides =\n            ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});\n        const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n\n        void main() {\n          ivec5 coords = getOutputCoords();\n          int batch = coords.x;\n          int ch = coords.u;\n\n          ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n          int xDCorner = xCorner.x;\n          int xRCorner = xCorner.y;\n          int xCCorner = xCorner.z;\n\n          // max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n\n          for (int wD = 0; wD < ${effectiveFilterDepth};\n              wD += ${dilationDepth}) {\n            int xD = xDCorner + wD;\n\n            if (xD < 0 || xD >= ${convInfo.inDepth}) {\n              continue;\n            }\n\n            for (int wR = 0; wR < ${effectiveFilterHeight};\n                wR += ${dilationHeight}) {\n              int xR = xRCorner + wR;\n\n              if (xR < 0 || xR >= ${convInfo.inHeight}) {\n                continue;\n              }\n\n              for (int wC = 0; wC < ${effectiveFilterWidth};\n                  wC += ${dilationWidth}) {\n                int xC = xCCorner + wC;\n\n                if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                  continue;\n                }\n\n                float value = getX(batch, xD, xR, xC, ch);\n\n                // If a min / max value has already been found, use it. If not,\n                // use the current value.\n                float currMinMaxValue = mix(\n                    value, minMaxValue, minMaxValueFound);\n                if (value ${compareOp2} currMinMaxValue) {\n                  minMaxValue = value;\n                  minMaxValueFound = 1.0;\n                  minMaxPosition = ${flattenPositions ? includeBatchInIndex ? `(((batch * ${convInfo.inDepth} + xD) * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `((xD * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +\n                      wR * ${effectiveFilterWidth} + wC`};\n                }\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `;\n      return;\n    }\n    const compareOp = \"max\";\n    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;\n    if (poolType === \"avg\") {\n      returnValue = `avgValue / count`;\n    }\n    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;\n    const filterWidthVec4Remainder = filterWidth % 4;\n    const updateSnippet = `\n      if (${isAvgPool}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = ${compareOp}(values, minMaxValue);\n      }\n    `;\n    this.userCode = `\n      const ivec3 strides =\n        ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});\n      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n      const float initializationValue = ${initializationValue};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xD, int xR, int xC, int ch) {\n        if (xC < 0 || xC >= ${convInfo.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xD, xR, xC, ch);\n      }\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xDCorner = xCorner.x;\n        int xRCorner = xCorner.y;\n        int xCCorner = xCorner.z;\n\n        // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${initializationValue});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wD = 0; wD < ${effectiveFilterDepth};\n            wD += ${dilationDepth}) {\n          int xD = xDCorner + wD;\n\n          if (xD < 0 || xD >= ${convInfo.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${effectiveFilterHeight};\n            wR += ${dilationHeight}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${convInfo.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {\n              int xC = xCCorner + wC * ${dilationWidth};\n\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),\n                getValue(batch, xD, xR, xC + 3 * ${dilationWidth}, ch)\n              );\n\n              ${updateSnippet}\n            }\n\n            int xC = xCCorner + ${filterWidthNearestVec4};\n            if (${filterWidthVec4Remainder === 1}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                initializationValue,\n                initializationValue,\n                initializationValue\n              );\n\n              ${updateSnippet}\n            } else if (${filterWidthVec4Remainder === 2}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),\n                initializationValue,\n                initializationValue\n              );\n\n              ${updateSnippet}\n            } else if (${filterWidthVec4Remainder === 3}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),\n                initializationValue\n              );\n\n              ${updateSnippet}\n            }\n          }\n          setOutput(${returnValue});\n        }\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool.js\nfunction avgPool3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  assertNotComplex2(x, \"avgPool\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = 1;\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {\n    return identity3({ inputs: { x }, backend: backend2 });\n  }\n  const avgPoolProgram = new Pool2DProgram(convInfo, \"avg\", false);\n  return backend2.runWebGLProgram(avgPoolProgram, [x], \"float32\");\n}\nvar avgPoolConfig2 = {\n  kernelName: AvgPool,\n  backendName: \"webgl\",\n  kernelFunc: avgPool3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool3D.js\nfunction avgPool3D2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;\n  const dilations = [1, 1, 1];\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);\n  const avgPoolProgram = new Pool3DProgram(convInfo, \"avg\", false);\n  return backend2.runWebGLProgram(avgPoolProgram, [x], \"float32\");\n}\nvar avgPool3DConfig2 = {\n  kernelName: AvgPool3D,\n  backendName: \"webgl\",\n  kernelFunc: avgPool3D2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/avg_pool_backprop_gpu.js\nvar AvgPool2DBackpropProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\"];\n    this.outputShape = convInfo.inShape;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n    const avgMultiplier = 1 / (filterHeight * filterWidth);\n    this.userCode = `\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n      const float avgMultiplier = float(${avgMultiplier});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${effectiveFilterHeight};\n            wR += ${dilationHeight}) {\n          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${effectiveFilterWidth};\n            wC+= ${dilationWidth}) {\n            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n\n            dotProd += dyValue * avgMultiplier;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar AvgPool3DBackpropProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\"];\n    this.outputShape = convInfo.inShape;\n    const filterDepth = convInfo.filterDepth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationDepth = convInfo.dilationDepth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;\n    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n    const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);\n    this.userCode = `\n      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n      const float avgMultiplier = float(${avgMultiplier});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${effectiveFilterDepth};\n            wD += ${dilationDepth}) {\n          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;\n\n          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${effectiveFilterHeight};\n              wR += ${dilationHeight}) {\n            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${effectiveFilterWidth};\n                wC += ${dilationWidth}) {\n              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n\n              dotProd += dyValue * avgMultiplier;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool3DGrad.js\nfunction avgPool3DGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const x = input2;\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = [1, 1, 1];\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  const avgPoolBackpropProgram = new AvgPool3DBackpropProgram(convInfo);\n  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);\n}\nvar avgPool3DGradConfig3 = {\n  kernelName: AvgPool3DGrad,\n  backendName: \"webgl\",\n  kernelFunc: avgPool3DGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPoolGrad.js\nfunction avgPoolGrad3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const x = input2;\n  assertNotComplex2([dy, input2], \"avgPoolGrad\");\n  const { filterSize, strides, pad: pad3 } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);\n  const avgPoolBackpropProgram = new AvgPool2DBackpropProgram(convInfo);\n  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);\n}\nvar avgPoolGradConfig3 = {\n  kernelName: AvgPoolGrad,\n  backendName: \"webgl\",\n  kernelFunc: avgPoolGrad3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchMatMul.js\nfunction batchMatMul2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b } = inputs;\n  const { transposeA, transposeB } = attrs;\n  return batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2 });\n}\nvar batchMatMulConfig2 = {\n  kernelName: BatchMatMul,\n  backendName: \"webgl\",\n  kernelFunc: batchMatMul2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/batchnorm_gpu.js\nvar BatchNormProgram = class {\n  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {\n    this.outputShape = [];\n    this.variableNames = [\"x\", \"mean\", \"variance\"];\n    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);\n    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);\n    let offsetSnippet = \"0.0\";\n    if (offsetShape != null) {\n      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);\n      this.variableNames.push(\"offset\");\n      offsetSnippet = \"getOffsetAtOutCoords()\";\n    }\n    let scaleSnippet = \"1.0\";\n    if (scaleShape != null) {\n      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);\n      this.variableNames.push(\"scale\");\n      scaleSnippet = \"getScaleAtOutCoords()\";\n    }\n    this.outputShape = xShape;\n    this.userCode = `\n      void main() {\n        float x = getXAtOutCoords();\n        float mean = getMeanAtOutCoords();\n        float variance = getVarianceAtOutCoords();\n        float offset = ${offsetSnippet};\n        float scale = ${scaleSnippet};\n        float inv = scale * inversesqrt(variance + float(${varianceEpsilon}));\n        setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/batchnorm_packed_gpu.js\nvar BatchNormPackedProgram = class {\n  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.variableNames = [\"x\", \"mean\", \"variance\"];\n    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);\n    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);\n    let offsetSnippet = \"vec4(0.0)\";\n    if (offsetShape != null) {\n      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);\n      this.variableNames.push(\"offset\");\n      offsetSnippet = \"getOffsetAtOutCoords()\";\n    }\n    let scaleSnippet = \"vec4(1.0)\";\n    if (scaleShape != null) {\n      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);\n      this.variableNames.push(\"scale\");\n      scaleSnippet = \"getScaleAtOutCoords()\";\n    }\n    this.outputShape = xShape;\n    this.userCode = `\n      void main() {\n        vec4 offset = ${offsetSnippet};\n        vec4 scale = ${scaleSnippet};\n\n        vec4 x = getXAtOutCoords();\n        vec4 mean = getMeanAtOutCoords();\n        vec4 variance = getVarianceAtOutCoords();\n\n        vec4 inv = scale * inversesqrt(variance + vec4(${varianceEpsilon}));\n\n        setOutput((x - mean) * inv + offset);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchNorm.js\nvar batchNorm3 = ({ inputs, backend: backend2, attrs }) => {\n  const { x, mean: mean4, variance, offset, scale: scale2 } = inputs;\n  util_exports.assert(mean4.shape.length === variance.shape.length, () => \"Batch normalization gradient requires mean and variance to have equal ranks.\");\n  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => \"Batch normalization gradient requires mean and offset to have equal ranks.\");\n  util_exports.assert(scale2 == null || mean4.shape.length === scale2.shape.length, () => \"Batch normalization gradient requires mean and scale to have equal ranks.\");\n  let { varianceEpsilon } = attrs;\n  if (varianceEpsilon == null) {\n    varianceEpsilon = 1e-3;\n  }\n  const finalInputs = [x, mean4, variance];\n  let offsetShape = null;\n  if (offset != null) {\n    offsetShape = offset.shape;\n    finalInputs.push(offset);\n  }\n  let scaleShape = null;\n  if (scale2 != null) {\n    scaleShape = scale2.shape;\n    finalInputs.push(scale2);\n  }\n  const program = env().getBool(\"WEBGL_PACK_NORMALIZATION\") ? new BatchNormPackedProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon) : new BatchNormProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon);\n  const output = backend2.runWebGLProgram(program, finalInputs, finalInputs[0].dtype);\n  return output;\n};\nvar batchNormConfig2 = {\n  kernelName: FusedBatchNorm,\n  backendName: \"webgl\",\n  kernelFunc: batchNorm3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/slice_gpu.js\nvar SliceProgram = class {\n  constructor(destSize) {\n    this.variableNames = [\"source\"];\n    this.outputShape = destSize;\n    this.rank = destSize.length;\n    const dtype = getCoordsDataType(this.rank);\n    this.customUniforms = [{ name: \"start\", arrayIndex: this.rank, type: \"int\" }];\n    const sourceCoords = getCoords(this.rank);\n    let body;\n    const coordSum = destSize.map((_, i) => {\n      return `sourceLoc.${coords[i]} = start[${i}] + coords.${coords[i]};`;\n    });\n    body = `\n        ${dtype} sourceLoc;\n        ${dtype} coords = getOutputCoords();\n        ${coordSum.join(\"\\n\")}\n      `;\n    this.userCode = `\n      void main() {\n        ${body}\n        setOutput(getSource(${sourceCoords}));\n      }\n    `;\n  }\n};\nvar coords = [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"];\nfunction getCoords(rank) {\n  if (rank === 1) {\n    return \"sourceLoc\";\n  } else if (rank <= 6) {\n    return coords.slice(0, rank).map((x) => \"sourceLoc.\" + x).join(\",\");\n  } else {\n    throw Error(`Slicing for rank ${rank} is not yet supported`);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/slice_packed_gpu.js\nvar SlicePackedProgram = class {\n  constructor(destSize) {\n    this.variableNames = [\"source\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = destSize;\n    this.rank = destSize.length;\n    this.customUniforms = [{ name: \"start\", arrayIndex: this.rank, type: \"int\" }];\n    const dtype = getCoordsDataType(this.rank);\n    const coords2 = getChannels(\"coords\", this.rank);\n    const sourceLoc = getChannels(\"sourceLoc\", this.rank);\n    const innerDims = this.rank === 1 ? \"sourceLoc\" : `vec2(${sourceLoc.slice(-2).join()})`;\n    const getChannel = `getChannel(getSource(${sourceLoc.join()}), ${innerDims})`;\n    const upperRow = `\n      result.x = ${getChannel};\n      if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {\n        ++${sourceLoc[this.rank - 1]};\n        result.y = ${getChannel};\n        --${sourceLoc[this.rank - 1]};\n      }\n    `;\n    const lowerRow = this.rank === 1 ? \"\" : `\n      --${coords2[this.rank - 1]};\n      if (++${coords2[this.rank - 2]} < ${destSize[this.rank - 2]}) {\n        ++${sourceLoc[this.rank - 2]};\n        result.z = ${getChannel};\n        if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {\n          ++${sourceLoc[this.rank - 1]};\n          result.w = ${getChannel};\n        }\n      }\n    `;\n    const sourceLocSetup = this.rank <= 4 ? `sourceLoc = coords +\n            ${dtype}(${destSize.map((_, i) => `start[${i}]`).join()});` : destSize.map((_, i) => `${sourceLoc[i]} = ${coords2[i]} + start[${i}];`).join(\"\\n\");\n    this.userCode = `\n      void main() {\n        ${dtype} coords = getOutputCoords();\n        ${dtype} sourceLoc;\n        ${sourceLocSetup}\n        vec4 result = vec4(0.);\n        ${upperRow}\n        ${lowerRow}\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Slice.js\nfunction shallowSlice(x, begin, size, backend2) {\n  const xTexData = backend2.texData.get(x.dataId);\n  const t = backend2.makeTensorInfo(size, x.dtype);\n  const newTexData = backend2.texData.get(t.dataId);\n  Object.assign(newTexData, xTexData);\n  newTexData.refCount = 1;\n  newTexData.shape = size;\n  newTexData.dtype = x.dtype;\n  let flatOffset = slice_util_exports.computeFlatOffset(begin, util_exports.computeStrides(x.shape));\n  if (xTexData.slice) {\n    flatOffset += xTexData.slice.flatOffset;\n  }\n  newTexData.slice = {\n    flatOffset,\n    origDataId: xTexData.slice && xTexData.slice.origDataId || x.dataId\n  };\n  const refCount = backend2.dataRefCount.get(newTexData.slice.origDataId) || 1;\n  backend2.dataRefCount.set(newTexData.slice.origDataId, refCount + 1);\n  return t;\n}\nfunction slice3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { begin, size } = attrs;\n  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);\n  slice_util_exports.assertParamsValid(x, $begin, $size);\n  if (util_exports.sizeFromShape($size) === 0) {\n    return backend2.makeTensorInfo($size, x.dtype, []);\n  }\n  if (backend2.shouldExecuteOnCPU([x]) || x.dtype === \"string\") {\n    const xTexData = backend2.texData.get(x.dataId);\n    const outValues = sliceImplCPU(xTexData.values, $begin, $size, x.shape, x.dtype);\n    return backend2.makeTensorInfo($size, x.dtype, outValues);\n  }\n  const { isPacked } = backend2.texData.get(x.dataId);\n  const isContinous = slice_util_exports.isSliceContinous(x.shape, $begin, $size);\n  if (isPacked || !isContinous) {\n    const program = env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") ? new SlicePackedProgram($size) : new SliceProgram($size);\n    const customValues = [$begin];\n    return backend2.runWebGLProgram(program, [x], x.dtype, customValues);\n  }\n  backend2.uploadToGPU(x.dataId);\n  return shallowSlice(x, $begin, $size, backend2);\n}\nvar sliceConfig2 = {\n  kernelName: Slice,\n  backendName: \"webgl\",\n  kernelFunc: slice3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchToSpaceND.js\nvar batchToSpaceND3 = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, crops } = attrs;\n  util_exports.assert(x.shape.length <= 4, () => \"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet\");\n  const prod5 = blockShape.reduce((a, b) => a * b);\n  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);\n  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);\n  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);\n  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);\n  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);\n  const toDispose = [];\n  const reshapedIntermediate = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });\n  const transposedIntermediate = transpose3({ inputs: { x: reshapedIntermediate }, backend: backend2, attrs: { perm: permuted } });\n  const reshapedIntermediate2 = reshape4({\n    inputs: { x: transposedIntermediate },\n    backend: backend2,\n    attrs: { shape: reshapedPermuted }\n  });\n  const sliced = slice3({\n    inputs: { x: reshapedIntermediate2 },\n    backend: backend2,\n    attrs: { begin: sliceBeginCoords, size: sliceSize }\n  });\n  toDispose.push(reshapedIntermediate);\n  toDispose.push(transposedIntermediate);\n  toDispose.push(reshapedIntermediate2);\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return sliced;\n};\nvar batchToSpaceNDConfig2 = {\n  kernelName: BatchToSpaceND,\n  backendName: \"webgl\",\n  kernelFunc: batchToSpaceND3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Bincount.js\nfunction bincount3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, weights } = inputs;\n  const { size } = attrs;\n  const xVals = backend2.readSync(x.dataId);\n  const weightsVals = backend2.readSync(weights.dataId);\n  const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);\n  return backend2.makeTensorInfo([size], weights.dtype, outVals);\n}\nvar bincountConfig2 = {\n  kernelName: Bincount,\n  backendName: \"webgl\",\n  kernelFunc: bincount3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BroadcastArgs.js\nfunction broadcastArgs3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { s0, s1 } = inputs;\n  const s0Vals = backend2.readSync(s0.dataId);\n  const s1Vals = backend2.readSync(s1.dataId);\n  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));\n  return backend2.makeTensorInfo([broadcastShape.length], \"int32\", Int32Array.from(broadcastShape));\n}\nvar broadcastArgsConfig2 = {\n  kernelName: BroadcastArgs,\n  backendName: \"webgl\",\n  kernelFunc: broadcastArgs3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NotEqual.js\nvar NOT_EQUAL = `return float(a != b);`;\nvar notEqual3 = binaryKernelFunc2({ opSnippet: NOT_EQUAL, cpuKernelImpl: notEqualImplCPU, dtype: \"bool\" });\nvar notEqualConfig2 = {\n  kernelName: NotEqual,\n  backendName: \"webgl\",\n  kernelFunc: notEqual3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Real.js\nfunction real3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const inputData = backend2.texData.get(input2.dataId);\n  return identity3({ inputs: { x: inputData.complexTensorInfos.real }, backend: backend2 });\n}\nvar realConfig2 = {\n  kernelName: Real,\n  backendName: \"webgl\",\n  kernelFunc: real3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/int.js\nvar TO_INT = `return float(int(x));`;\nfunction int(input2, backend2) {\n  const program = new UnaryOpProgram(input2.shape, TO_INT);\n  const output = backend2.runWebGLProgram(program, [input2], \"int32\");\n  return { dataId: output.dataId, shape: output.shape, dtype: output.dtype };\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cast.js\nfunction cast4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { dtype } = attrs;\n  if (dtype === \"complex64\") {\n    if (x.dtype === \"complex64\") {\n      return identity3({ inputs: { x }, backend: backend2 });\n    }\n    const zerosTensor = zeros(x.shape);\n    const floatX = cast4({ inputs: { x }, backend: backend2, attrs: { dtype: \"float32\" } });\n    const result = complex3({ inputs: { real: floatX, imag: zerosTensor }, backend: backend2 });\n    zerosTensor.dispose();\n    backend2.disposeIntermediateTensorInfo(floatX);\n    return result;\n  }\n  if (x.dtype === \"complex64\") {\n    const realPart = real3({ inputs: { input: x }, backend: backend2 });\n    const result = cast4({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    return result;\n  }\n  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {\n    const result = identity3({ inputs: { x }, backend: backend2 });\n    return { dataId: result.dataId, shape: result.shape, dtype };\n  }\n  if (backend2.shouldExecuteOnCPU([x])) {\n    const values = backend2.texData.get(x.dataId).values;\n    const [resultShape, resultType, resultData] = castImplCPU(values, x.shape, x.dtype, dtype);\n    return backend2.makeTensorInfo(resultShape, resultType, resultData);\n  }\n  if (dtype === \"int32\") {\n    return int(x, backend2);\n  }\n  if (dtype === \"bool\") {\n    const zerosTensorInfo = backend2.makeTensorInfo([], \"bool\", util_exports.getTypedArrayFromDType(\"bool\", 1));\n    const binaryInputs = { a: x, b: zerosTensorInfo };\n    const result = notEqual3({ inputs: binaryInputs, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);\n    return result;\n  }\n  throw new Error(`Error in Cast: failed to cast ${x.dtype} to ${dtype}`);\n}\nvar castConfig2 = {\n  kernelName: Cast,\n  backendName: \"webgl\",\n  kernelFunc: cast4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Ceil.js\nvar CEIL = `return ceil(x);`;\nvar ceil3 = unaryKernelFunc2({ opSnippet: CEIL, packedOpSnippet: CEIL, cpuKernelImpl: ceilImplCPU });\nvar ceilConfig2 = {\n  kernelName: Ceil,\n  backendName: \"webgl\",\n  kernelFunc: ceil3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/clip_gpu.js\nvar ClipProgram = class {\n  constructor(aShape) {\n    this.variableNames = [\"A\"];\n    this.customUniforms = [\n      { name: \"minVal\", type: \"float\" },\n      { name: \"maxVal\", type: \"float\" }\n    ];\n    this.outputShape = aShape;\n    this.userCode = `\n\n      void main() {\n        float value = getAAtOutCoords();\n        if (isnan(value)) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, minVal, maxVal));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/clip_packed_gpu.js\nvar ClipPackedProgram = class {\n  constructor(aShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [\n      { name: \"minVal\", type: \"float\" },\n      { name: \"maxVal\", type: \"float\" }\n    ];\n    this.outputShape = aShape;\n    this.userCode = `\n      void main() {\n        vec4 value = getAAtOutCoords();\n\n        if (any(isnan(value))) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, vec4(minVal), vec4(maxVal)));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ClipByValue.js\nfunction clipByValue3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { clipValueMin, clipValueMax } = attrs;\n  let program;\n  if (env().getBool(\"WEBGL_PACK_CLIP\")) {\n    program = new ClipPackedProgram(x.shape);\n  } else {\n    program = new ClipProgram(x.shape);\n  }\n  const customValues = [[clipValueMin], [clipValueMax]];\n  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);\n}\nvar clipByValueConfig2 = {\n  kernelName: ClipByValue,\n  backendName: \"webgl\",\n  kernelFunc: clipByValue3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/complex_abs_gpu.js\nvar ComplexAbsProgram = class {\n  constructor(shape) {\n    this.variableNames = [\"real\", \"imag\"];\n    this.outputShape = shape;\n    this.userCode = `\n      void main() {\n        float re = abs(getRealAtOutCoords());\n        float im = abs(getImagAtOutCoords());\n        float mx = max(re, im);\n\n        // sadly the length function in glsl is not underflow-safe\n        // (at least not on Intel GPUs). So the safe solution is\n        // to ensure underflow-safety in all cases.\n        setOutput(\n          mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))\n        );\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ComplexAbs.js\nfunction makeComplexComponentTensorInfo(complexTensor, complexPart) {\n  return {\n    dataId: complexPart.dataId,\n    dtype: complexPart.dtype,\n    shape: complexTensor.shape\n  };\n}\nfunction complexAbs2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  const xData = backend2.texData.get(x.dataId);\n  const program = new ComplexAbsProgram(x.shape);\n  const programInputs = [\n    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.real),\n    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.imag)\n  ];\n  return backend2.runWebGLProgram(program, programInputs, programInputs[0].dtype);\n}\nvar complexAbsConfig2 = {\n  kernelName: ComplexAbs,\n  backendName: \"webgl\",\n  kernelFunc: complexAbs2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/concat_gpu.js\nvar ConcatProgram = class {\n  constructor(shapes) {\n    this.outputShape = [];\n    this.outputShape = backend_util_exports.computeOutShape(shapes, 1);\n    this.variableNames = shapes.map((_, i) => `T${i}`);\n    const offsets = new Array(shapes.length - 1);\n    offsets[0] = shapes[0][1];\n    for (let i = 1; i < offsets.length; i++) {\n      offsets[i] = offsets[i - 1] + shapes[i][1];\n    }\n    const snippets = [`if (yC < ${offsets[0]}) setOutput(getT0(yR, yC));`];\n    for (let i = 1; i < offsets.length; i++) {\n      const shift = offsets[i - 1];\n      snippets.push(`else if (yC < ${offsets[i]}) setOutput(getT${i}(yR, yC-${shift}));`);\n    }\n    const lastIndex = offsets.length;\n    const lastShift = offsets[offsets.length - 1];\n    snippets.push(`else setOutput(getT${lastIndex}(yR, yC-${lastShift}));`);\n    this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int yR = coords.x;\n        int yC = coords.y;\n\n        ${snippets.join(\"\\n        \")}\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/concat_packed_gpu.js\nvar ConcatPackedProgram = class {\n  constructor(shapes, axis) {\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = [];\n    this.outputShape = backend_util_exports.computeOutShape(shapes, axis);\n    const shape = this.outputShape;\n    const rank = shape.length;\n    const dtype = getCoordsDataType(rank);\n    const coords2 = getChannels(\"coords\", rank);\n    const channels = [\"x\", \"y\", \"z\", \"w\", \"u\", \"v\"].slice(0, rank);\n    this.variableNames = shapes.map((_, i) => `T${i}`);\n    const offsets = new Array(shapes.length - 1);\n    offsets[0] = shapes[0][axis];\n    for (let i = 1; i < offsets.length; i++) {\n      offsets[i] = offsets[i - 1] + shapes[i][axis];\n    }\n    const channel = channels[axis];\n    const lastChannels = channels.slice(-2);\n    const allChannels = channels.join();\n    let getValueSnippet = `if (${channel} < ${offsets[0]}) {\n        return getChannel(\n            getT0(${allChannels}), vec2(${lastChannels.join()}));\n        }`;\n    for (let i = 1; i < offsets.length; i++) {\n      const shift2 = offsets[i - 1];\n      getValueSnippet += `\n        if (${channel} < ${offsets[i]}  && ${channel} >= ${offsets[i - 1]}) {\n          return getChannel(\n            getT${i}(${shiftedChannels(channels, channel, shift2)}),\n            vec2(${shiftedChannels(lastChannels, channel, shift2)}));\n        }`;\n    }\n    const lastIndex = offsets.length;\n    const shift = offsets[offsets.length - 1];\n    getValueSnippet += `\n        return getChannel(\n          getT${lastIndex}(${shiftedChannels(channels, channel, shift)}),\n          vec2(${shiftedChannels(lastChannels, channel, shift)}));`;\n    this.userCode = `\n      float getValue(${channels.map((x) => \"int \" + x)}) {\n        ${getValueSnippet}\n      }\n\n      void main() {\n        ${dtype} coords = getOutputCoords();\n        vec4 result = vec4(getValue(${coords2}), 0., 0., 0.);\n\n        ${coords2[rank - 1]} = ${coords2[rank - 1]} + 1;\n        if (${coords2[rank - 1]} < ${shape[rank - 1]}) {\n          result.g = getValue(${coords2});\n        }\n\n        ${coords2[rank - 2]} = ${coords2[rank - 2]} + 1;\n        if (${coords2[rank - 2]} < ${shape[rank - 2]}) {\n          result.a = getValue(${coords2});\n        }\n\n        ${coords2[rank - 1]} = ${coords2[rank - 1]} - 1;\n        if (${coords2[rank - 2]} < ${shape[rank - 2]} &&\n            ${coords2[rank - 1]} < ${shape[rank - 1]}) {\n          result.b = getValue(${coords2});\n        }\n        setOutput(result);\n      }\n    `;\n  }\n};\nfunction shiftedChannels(channels, channel, shift) {\n  const channelIdx = channels.indexOf(channel);\n  const res = channels.map((c, idx) => {\n    if (idx === channelIdx) {\n      return `${c} - ${shift}`;\n    } else {\n      return c;\n    }\n  });\n  return res.join();\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Imag.js\nfunction imag3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const inputData = backend2.texData.get(input2.dataId);\n  return identity3({ inputs: { x: inputData.complexTensorInfos.imag }, backend: backend2 });\n}\nvar imagConfig2 = {\n  kernelName: Imag,\n  backendName: \"webgl\",\n  kernelFunc: imag3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Concat_impl.js\nfunction concatImpl2(inputs, axis, backend2) {\n  const dtype = inputs[0].dtype;\n  if (dtype === \"complex64\") {\n    const reals = inputs.map((t) => real3({ inputs: { input: t }, backend: backend2 }));\n    const imags = inputs.map((t) => imag3({ inputs: { input: t }, backend: backend2 }));\n    const realConcated = concatImpl2(reals, axis, backend2);\n    const imagConcated = concatImpl2(imags, axis, backend2);\n    const result2 = complex3({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });\n    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));\n    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));\n    backend2.disposeIntermediateTensorInfo(realConcated);\n    backend2.disposeIntermediateTensorInfo(imagConcated);\n    return result2;\n  }\n  let runOnCpu = backend2.shouldExecuteOnCPU(inputs);\n  if (dtype === \"string\") {\n    runOnCpu = true;\n  }\n  if (runOnCpu) {\n    const tensors2D2 = inputs.map((t) => {\n      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));\n      const shape = [-1, innerSize];\n      return reshape4({ inputs: { x: t }, backend: backend2, attrs: { shape } });\n    });\n    const inputsValShapes = tensors2D2.map((t) => {\n      return { vals: backend2.readSync(t.dataId), shape: t.shape };\n    });\n    const outShape2 = backend_util_exports.computeOutShape(tensors2D2.map((t) => t.shape), 1);\n    const simplyConcat = tensors2D2[0].shape[0] === 1;\n    const outVals = concatImplCPU(inputsValShapes, outShape2, dtype, simplyConcat);\n    const finalOutShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);\n    const outInfo = backend2.makeTensorInfo(finalOutShape, dtype, outVals);\n    tensors2D2.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n    return outInfo;\n  }\n  const maxTexturesInShader = env().getNumber(\"WEBGL_MAX_TEXTURES_IN_SHADER\");\n  if (inputs.length > maxTexturesInShader) {\n    const reducedInputs = [];\n    for (let i = 0; i < inputs.length; i += maxTexturesInShader) {\n      const subArray = inputs.slice(i, i + maxTexturesInShader);\n      reducedInputs.push(concatImpl2(subArray, axis, backend2));\n    }\n    const result2 = concatImpl2(reducedInputs, axis, backend2);\n    for (const i of reducedInputs) {\n      backend2.disposeIntermediateTensorInfo(i);\n    }\n    return result2;\n  }\n  if (env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") && inputs[0].shape.length > 1) {\n    const program2 = new ConcatPackedProgram(inputs.map((t) => t.shape), axis);\n    return backend2.runWebGLProgram(program2, inputs, dtype);\n  }\n  const { tensors2D, outShape } = computeTensors2D(inputs, axis, backend2);\n  const program = new ConcatProgram(tensors2D.map((t) => t.shape));\n  const result = backend2.runWebGLProgram(program, tensors2D, dtype);\n  tensors2D.forEach((r) => backend2.disposeIntermediateTensorInfo(r));\n  const reshapedResult = reshape4({ inputs: { x: result }, attrs: { shape: outShape }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(result);\n  return reshapedResult;\n}\nfunction computeTensors2D(inputs, axis, backend2) {\n  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);\n  const tensors2D = inputs.map((x) => reshape4({\n    inputs: { x },\n    attrs: { shape: [-1, util_exports.sizeFromShape(x.shape.slice(axis))] },\n    backend: backend2\n  }));\n  return { tensors2D, outShape };\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Concat.js\nfunction concat3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { axis } = attrs;\n  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];\n  const shapes = inputs.map((t) => t.shape);\n  backend_util_exports.assertParamsConsistent(shapes, $axis);\n  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);\n  if (util_exports.sizeFromShape(outShape) === 0) {\n    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);\n  }\n  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);\n  if ($inputs.length === 1) {\n    return identity3({ inputs: { x: $inputs[0] }, backend: backend2 });\n  }\n  return concatImpl2($inputs, $axis, backend2);\n}\nvar concatConfig2 = {\n  kernelName: Concat,\n  backendName: \"webgl\",\n  kernelFunc: concat3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_gpu.js\nvar Conv2DProgram = class {\n  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivationWeights = false, hasLeakyreluAlpha = false) {\n    this.variableNames = [\"x\", \"W\"];\n    this.outputShape = convInfo.outShape;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;\n    const inputDepthVec4Remainder = convInfo.inChannels % 4;\n    const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n    const rowDim = isChannelsLast ? 1 : 2;\n    const colDim = isChannelsLast ? 2 : 3;\n    const channelDim = isChannelsLast ? 3 : 1;\n    let activationSnippet = \"\", applyActivationSnippet = \"\";\n    if (activation2) {\n      if (hasPreluActivationWeights) {\n        activationSnippet = `float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${activation2}\n        }`;\n      } else if (hasLeakyreluAlpha) {\n        activationSnippet = `float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${activation2}\n        }`;\n      } else {\n        activationSnippet = `\n          float activation(float x) {\n            ${activation2}\n          }\n        `;\n      }\n      applyActivationSnippet = `result = activation(result);`;\n    }\n    const addBiasSnippet = addBias ? \"result += getBiasAtOutCoords();\" : \"\";\n    if (addBias) {\n      this.variableNames.push(\"bias\");\n    }\n    if (hasPreluActivationWeights) {\n      this.variableNames.push(\"preluActivationWeights\");\n    }\n    if (hasLeakyreluAlpha) {\n      this.variableNames.push(\"leakyreluAlpha\");\n    }\n    this.userCode = `\n      ${activationSnippet}\n\n      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d2 = coords[${channelDim}];\n\n        ivec2 xRCCorner =\n            ivec2(coords[${rowDim}], coords[${colDim}]) * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${filterHeight}; wR++) {\n          int xR = xRCorner + wR * ${dilationHeight};\n\n          if (xR < 0 || xR >= ${convInfo.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${filterWidth}; wC++) {\n            int xC = xCCorner + wC * ${dilationWidth};\n\n            if (xC < 0 || xC >= ${convInfo.inWidth}) {\n              continue;\n            }\n\n            for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {\n              vec4 wValues = vec4(\n                getW(wR, wC, d1, d2),\n                getW(wR, wC, d1 + 1, d2),\n                getW(wR, wC, d1 + 2, d2),\n                getW(wR, wC, d1 + 3, d2)\n              );\n\n              if (${isChannelsLast}) {\n                vec4 xValues = vec4(\n                  getX(batch, xR, xC, d1),\n                  getX(batch, xR, xC, d1 + 1),\n                  getX(batch, xR, xC, d1 + 2),\n                  getX(batch, xR, xC, d1 + 3)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec4 xValues = vec4(\n                  getX(batch, d1, xR, xC),\n                  getX(batch, d1 + 1, xR, xC),\n                  getX(batch, d1 + 2, xR, xC),\n                  getX(batch, d1 + 3, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n\n            if (${inputDepthVec4Remainder === 1}) {\n\n              if (${isChannelsLast}) {\n                dotProd +=\n                    getX(batch, xR, xC, ${inputDepthNearestVec4}) *\n                    getW(wR, wC, ${inputDepthNearestVec4}, d2);\n              } else {\n                dotProd +=\n                    getX(batch, ${inputDepthNearestVec4}, xR, xC) *\n                    getW(wR, wC, ${inputDepthNearestVec4}, d2);\n              }\n\n            } else if (${inputDepthVec4Remainder === 2}) {\n              vec2 wValues = vec2(\n                getW(wR, wC, ${inputDepthNearestVec4}, d2),\n                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2)\n              );\n\n              if (${isChannelsLast}) {\n                vec2 xValues = vec2(\n                  getX(batch, xR, xC, ${inputDepthNearestVec4}),\n                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec2 xValues = vec2(\n                  getX(batch, ${inputDepthNearestVec4}, xR, xC),\n                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            } else if (${inputDepthVec4Remainder === 3}) {\n              vec3 wValues = vec3(\n                getW(wR, wC, ${inputDepthNearestVec4}, d2),\n                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2),\n                getW(wR, wC, ${inputDepthNearestVec4} + 2, d2)\n              );\n\n              if (${isChannelsLast}) {\n                vec3 xValues = vec3(\n                  getX(batch, xR, xC, ${inputDepthNearestVec4}),\n                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1),\n                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec3 xValues = vec3(\n                  getX(batch, ${inputDepthNearestVec4}, xR, xC),\n                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC),\n                  getX(batch, ${inputDepthNearestVec4} + 2, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            }\n          }\n        }\n\n        float result = dotProd;\n        ${addBiasSnippet}\n        ${applyActivationSnippet}\n        setOutput(result);\n      }\n    `;\n  }\n};\nvar Conv3DProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"x\", \"W\"];\n    this.outputShape = convInfo.outShape;\n    const padFront = convInfo.padInfo.front;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationDepth = convInfo.dilationDepth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const filterDepth = convInfo.filterDepth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;\n    const inputDepthVec4Remainder = convInfo.inChannels % 4;\n    this.userCode = `\n      const ivec3 strides = ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});\n      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d2 = coords.u;\n\n        ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xFCorner = xFRCCorner.x;\n        int xRCorner = xFRCCorner.y;\n        int xCCorner = xFRCCorner.z;\n\n        // Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get\n        // y(yF, yR, yC, d2). ? = to be determined. : = across all\n        // values in that axis.\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${filterDepth}; wF++) {\n          int xF = xFCorner + wF * ${dilationDepth};\n\n          if (xF < 0 || xF >= ${convInfo.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${filterHeight}; wR++) {\n            int xR = xRCorner + wR * ${dilationHeight};\n\n            if (xR < 0 || xR >= ${convInfo.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${filterWidth}; wC++) {\n              int xC = xCCorner + wC * ${dilationWidth};\n\n              if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                continue;\n              }\n\n              for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {\n                vec4 xValues = vec4(\n                  getX(batch, xF, xR, xC, d1),\n                  getX(batch, xF, xR, xC, d1 + 1),\n                  getX(batch, xF, xR, xC, d1 + 2),\n                  getX(batch, xF, xR, xC, d1 + 3)\n                );\n                vec4 wValues = vec4(\n                  getW(wF, wR, wC, d1, d2),\n                  getW(wF, wR, wC, d1 + 1, d2),\n                  getW(wF, wR, wC, d1 + 2, d2),\n                  getW(wF, wR, wC, d1 + 3, d2)\n                );\n\n                dotProd += dot(xValues, wValues);\n              }\n\n              if (${inputDepthVec4Remainder === 1}) {\n                dotProd +=\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}) *\n                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2);\n              } else if (${inputDepthVec4Remainder === 2}) {\n                vec2 xValues = vec2(\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1)\n                );\n                vec2 wValues = vec2(\n                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),\n                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else if (${inputDepthVec4Remainder === 3}) {\n                vec3 xValues = vec3(\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1),\n                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 2)\n                );\n                vec3 wValues = vec3(\n                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),\n                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2),\n                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 2, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_packed_gpu.js\nvar Conv2DPackedProgram = class {\n  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {\n    this.variableNames = [\"x\", \"W\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [\n      { name: \"pads\", type: \"ivec2\" },\n      { name: \"strides\", type: \"ivec2\" },\n      { name: \"dilations\", type: \"ivec2\" },\n      { name: \"inDims\", type: \"ivec2\" }\n    ];\n    this.outputShape = convInfo.outShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const padLeft = convInfo.padInfo.left;\n    const strideWidth = convInfo.strideWidth;\n    const dilationWidth = convInfo.dilationWidth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const texelsAcross = filterWidth;\n    let mainLoop = `\n       int xR; int xC; int xCOffset;\n       vec4 wTexel; vec4 previous; vec4 final;`;\n    for (let c = 0; c < filterWidth; c++) {\n      mainLoop += `\n           vec4 xTexelC${c * 2};\n           int xTexelC${c * 2}Ready;\n           vec4 xTexelC${c * 2 + 1};\n           int xTexelC${c * 2 + 1}Ready;\n           vec4 xC${c};`;\n    }\n    mainLoop += `\n     for (int r = 0; r < ${filterHeight}; r++) {\n      for (int d1 = 0; d1 < ${convInfo.inChannels}; d1 += 2) {\n       `;\n    for (let c = 0; c < filterWidth; c++) {\n      mainLoop += `\n           xTexelC${c * 2} = vec4(0.0);\n           xTexelC${c * 2}Ready = 0;\n           xTexelC${c * 2 + 1} = vec4(0.0);\n           xTexelC${c * 2 + 1}Ready = 0;\n           xC${c} = vec4(0.0);`;\n    }\n    mainLoop += `\n         xR = xRCorner + r * dilations[0];\n         if (xR >=0 && xR < inDims[0]) {\n       `;\n    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {\n      const colIndex = texelC * 2;\n      mainLoop += `\n           xC = xCCorner + ${colIndex * dilationWidth};\n           `;\n      if (strideWidth === 1) {\n        if (colIndex < filterWidth) {\n          if (padLeft % 2 === 1) {\n            mainLoop += `\n                 xCOffset = xC + 1;\n                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);\n\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${colIndex}.zw = vec2(0.0);\n                   }\n                   xTexelC${colIndex}Ready = 1;\n                 }\n               `;\n            if (dilationWidth === 1 && colIndex > 0) {\n              mainLoop += `\n                 xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);\n                 `;\n            } else {\n              mainLoop += `\n                   xCOffset = xC + 1 - 2;\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       previous.zw = vec2(0.0);\n                     }\n\n                     xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);\n                   } else {\n                     xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);\n                   }\n                   `;\n            }\n          } else {\n            mainLoop += `\n                 if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                   xTexelC${colIndex} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${colIndex}.zw = vec2(0.0);\n                   }\n                   xTexelC${colIndex}Ready = 1;\n                 }\n\n                 xC${colIndex} = xTexelC${colIndex};\n                 `;\n          }\n          if (colIndex + 1 < filterWidth) {\n            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;\n            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {\n              mainLoop += `\n                   xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                     xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       xTexelC${colIndex + 1}.zw = vec2(0.0);\n                     }\n                     xTexelC${colIndex + 1}Ready = 1;\n                   }\n                   `;\n              if (dilationWidth > 1) {\n                mainLoop += `\n                     xCOffset -= 2;\n                     if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                      previous = getX(batch, xR, xCOffset, d1);\n                      xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);\n                     } else {\n                      xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);\n                     }\n                     `;\n              } else {\n                mainLoop += `\n                     xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);\n                     `;\n              }\n            } else {\n              if (nextTexelOffset === 1) {\n                mainLoop += `\n                     xC${colIndex + 1} = xTexelC${colIndex};\n                     `;\n              } else {\n                mainLoop += `\n                     xCOffset = xC + ${nextTexelOffset};\n\n                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                       xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n                       if (xCOffset + 1 >= inDims[1]) {\n                         xTexelC${colIndex + 1}.zw = vec2(0.0);\n                       }\n                       xTexelC${colIndex + 1}Ready = 1;\n                     }\n\n                     xC${colIndex + 1} = xTexelC${colIndex + 1};\n                     `;\n              }\n            }\n          }\n        }\n      } else {\n        if (colIndex < filterWidth) {\n          if (padLeft % 2 === 1) {\n            mainLoop += `\n                 xCOffset = xC + 1 - strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${colIndex}.zw = vec2(0.0);\n                   }\n                   xTexelC${colIndex}Ready = 1;\n                 }\n\n                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                   xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xC + 2 >= inDims[1]) {\n                     xTexelC${colIndex + 1}.zw = vec2(0.0);\n                   }\n                   xTexelC${colIndex + 1}Ready = 1;\n                 }\n\n                 xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);\n               `;\n            if (colIndex + 1 < filterWidth) {\n              mainLoop += `\n                   final = vec4(0.0);\n                   xCOffset = xC + 1 + strides[1];\n                   if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                     final = getX(batch, xR, xCOffset, d1);\n                   }\n                   xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);\n                 `;\n            }\n          } else {\n            mainLoop += `\n                 if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                   xTexelC${colIndex} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${colIndex}.zw = vec2(0.0);\n                   }\n                   xTexelC${colIndex}Ready = 1;\n                 }\n\n                 xCOffset = xC + strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                   xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${colIndex + 1}.zw = vec2(0.);\n                   }\n                   xTexelC${colIndex + 1}Ready = 1;\n                 }\n\n                 xC${colIndex} = vec4(\n                   xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);\n               `;\n            if (colIndex + 1 < filterWidth) {\n              mainLoop += `\n                   xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);\n                 `;\n            }\n          }\n        }\n      }\n      if (colIndex < filterWidth) {\n        mainLoop += `\n             wTexel = getW(r, ${colIndex}, d1, d2);\n             dotProd += xC${colIndex}.xxzz * vec4(wTexel.xy, wTexel.xy);\n             if(d1 + 1 < ${convInfo.inChannels}) {\n               dotProd += xC${colIndex}.yyww * vec4(wTexel.zw, wTexel.zw);\n             }\n           `;\n        if (colIndex + 1 < filterWidth) {\n          mainLoop += `\n               wTexel = getW(r, ${colIndex + 1}, d1, d2);\n               dotProd += xC${colIndex + 1}.xxzz * vec4(wTexel.xy, wTexel.xy);\n               if(d1 + 1 < ${convInfo.inChannels}) {\n                 dotProd += xC${colIndex + 1}.yyww * vec4(wTexel.zw, wTexel.zw);\n               }\n             `;\n        }\n      }\n    }\n    mainLoop += `\n     }\n   `;\n    mainLoop += `\n     }\n   `;\n    mainLoop += `\n     }\n   `;\n    let activationSnippet = \"\", applyActivationSnippet = \"\";\n    if (activation2) {\n      if (hasPreluActivation) {\n        activationSnippet = `vec4 activation(vec4 a) {\n           vec4 b = getPreluActivationWeightsAtOutCoords();\n           ${activation2}\n         }`;\n      } else if (hasLeakyReluAlpha) {\n        activationSnippet = `vec4 activation(vec4 a) {\n           vec4 b = getLeakyreluAlphaAtOutCoords();\n           ${activation2}\n         }`;\n      } else {\n        activationSnippet = `vec4 activation(vec4 x) {\n           ${activation2}\n         }`;\n      }\n      applyActivationSnippet = `result = activation(result);`;\n    }\n    const addBiasSnippet = addBias ? \"result += getBiasAtOutCoords();\" : \"\";\n    if (addBias) {\n      this.variableNames.push(\"bias\");\n    }\n    if (hasPreluActivation) {\n      this.variableNames.push(\"preluActivationWeights\");\n    }\n    if (hasLeakyReluAlpha) {\n      this.variableNames.push(\"leakyreluAlpha\");\n    }\n    this.userCode = `\n       ${activationSnippet}\n\n       void main() {\n         ivec4 coords = getOutputCoords();\n         int batch = coords.x;\n         ivec2 xRCCorner = coords.yz * strides - pads;\n         int d2 = coords.w;\n         int xRCorner = xRCCorner.x;\n         int xCCorner = xRCCorner.y;\n\n         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n         vec4 dotProd = vec4(0.000000000000001);\n\n         ${mainLoop}\n\n         vec4 result = dotProd - vec4(0.000000000000001);\n         ${addBiasSnippet}\n         ${applyActivationSnippet}\n         setOutput(result);\n       }\n     `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/im2col_packed_gpu.js\nvar Im2ColPackedProgram = class {\n  constructor(outputShape, convInfo) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [\n      { name: \"inputShape\", type: \"ivec4\" },\n      { name: \"pad\", type: \"ivec2\" },\n      { name: \"stride\", type: \"ivec2\" },\n      { name: \"dilation\", type: \"ivec2\" },\n      { name: \"inChannels\", type: \"int\" },\n      { name: \"itemsPerBlockRow\", type: \"int\" },\n      { name: \"outWidth\", type: \"int\" }\n    ];\n    this.outputShape = outputShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const { dataFormat } = convInfo;\n    const glsl = getGlslDifferences();\n    const isChannelsLast = dataFormat === \"channelsLast\";\n    const rowDim = isChannelsLast ? 1 : 2;\n    const colDim = isChannelsLast ? 2 : 3;\n    const boundsCheckingSnippet = this.enableShapeUniforms ? \"if(blockIndex < outShape[2] && pos < outShape[1]) {\" : `if(blockIndex < ${outputShape[2]} && pos < ${outputShape[1]}) {`;\n    let unrolled = ``;\n    for (let row = 0; row <= 1; row++) {\n      for (let col = 0; col <= 1; col++) {\n        unrolled += `\n          blockIndex = rc.z + ${col};\n          pos = rc.y + ${row};\n\n          ${boundsCheckingSnippet}\n            offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];\n            d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);\n\n            if(d0 < inputShape[${rowDim}] && d0 >= 0) {\n              // Use custom imod instead mod. On Intel GPU, mod may generate\n              // unexpected value.\n              // https://github.com/tensorflow/tfjs/issues/5447\n              offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];\n              d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /\n                  inChannels);\n\n              if(d1 < inputShape[${colDim}] && d1 >= 0) {\n\n                ch = imod(pos, inChannels);\n\n                if (${isChannelsLast}) {\n                  innerDims = vec2(d1, ch);\n                  result[${row * 2 + col}] = getChannel(\n                    getA(rc.x, d0, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                } else {\n                  innerDims = vec2(d0, d1);\n                  result[${row * 2 + col}] = getChannel(\n                    getA(rc.x, ch, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                }\n              }\n            }\n          }\n        `;\n      }\n    }\n    this.userCode = `\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0);\n\n        int blockIndex, pos, offsetY, d0, offsetX, d1, ch;\n        vec2 innerDims;\n\n        ${unrolled}\n\n        ${glsl.output} = result;\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2D_impl.js\nfunction getShapeForBatchMatMul(shape, isChannelsLast) {\n  const length = shape.length;\n  if (length >= 3) {\n    return isChannelsLast ? [\n      ...shape.slice(0, -3),\n      shape[length - 3] * shape[length - 2],\n      shape[length - 1]\n    ] : [\n      ...shape.slice(0, -3),\n      shape[length - 3],\n      shape[length - 2] * shape[length - 1]\n    ];\n  } else if (!isChannelsLast && length === 1 && shape[0] > 1) {\n    return [shape[0], 1];\n  } else {\n    return null;\n  }\n}\nfunction conv2dByMatMul({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {\n  const xShape = x.shape;\n  const xTexData = backend2.texData.get(x.dataId);\n  const sharedMatMulDim = convInfo.inChannels;\n  const outerShapeX = xShape[0] * xShape[1] * xShape[2];\n  const outerShapeFilter = convInfo.outChannels;\n  const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n  const transposeA = false;\n  const transposeB = false;\n  let out;\n  const intermediates = [];\n  if (preluActivationWeights != null) {\n    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);\n    if (targetShape != null) {\n      preluActivationWeights = reshape4({\n        inputs: { x: preluActivationWeights },\n        backend: backend2,\n        attrs: { shape: targetShape }\n      });\n      intermediates.push(preluActivationWeights);\n    }\n  }\n  if (bias != null) {\n    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);\n    if (targetShape != null) {\n      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });\n      intermediates.push(bias);\n    }\n  }\n  const batchMatMulWillBeUnpacked = (outerShapeX === 1 || outerShapeFilter === 1) && sharedMatMulDim > MATMUL_SHARED_DIM_THRESHOLD;\n  const canOptimize = !batchMatMulWillBeUnpacked && xTexData.isPacked && isChannelsLast && xTexData.texture != null && xShape[2] % 2 !== 0 && util_exports.arraysEqual(xTexData.shape.slice(-3), xShape.slice(-3));\n  if (canOptimize) {\n    const targetShape = xShape[0] * xShape[1] * (xShape[2] + 1);\n    const xReshaped = {\n      dataId: x.dataId,\n      shape: [1, targetShape, convInfo.inChannels],\n      dtype: x.dtype\n    };\n    const originalXTexDataShape = xTexData.shape;\n    xTexData.shape = xTexData.shape.slice();\n    xTexData.shape[xTexData.shape.length - 2]++;\n    util_exports.assert(isReshapeFree(xTexData.shape, xReshaped.shape), () => `packed reshape ${xTexData.shape} to ${xReshaped.shape} isn't free`);\n    const filterReshaped = reshape4({\n      inputs: { x: filter },\n      backend: backend2,\n      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }\n    });\n    intermediates.push(filterReshaped);\n    const pointwiseConv = batchMatMulImpl({\n      a: xReshaped,\n      b: filterReshaped,\n      backend: backend2,\n      transposeA,\n      transposeB,\n      bias,\n      activation: activation2,\n      preluActivationWeights,\n      leakyreluAlpha\n    });\n    const pointwiseConvTexData = backend2.texData.get(pointwiseConv.dataId);\n    util_exports.assert(pointwiseConvTexData.isPacked, () => \"batchMatMul result is expected to be packed\");\n    xTexData.shape = originalXTexDataShape;\n    pointwiseConvTexData.shape = convInfo.outShape;\n    out = identity3({ inputs: { x: pointwiseConv }, backend: backend2 });\n    out.shape = convInfo.outShape;\n    intermediates.push(pointwiseConv);\n  } else {\n    const numCols = convInfo.outHeight * convInfo.outWidth;\n    const xReshaped = reshape4({\n      inputs: { x },\n      backend: backend2,\n      attrs: {\n        shape: isChannelsLast ? [convInfo.batchSize, numCols, convInfo.inChannels] : [convInfo.batchSize, convInfo.inChannels, numCols]\n      }\n    });\n    const filterReshaped = reshape4({\n      inputs: { x: filter },\n      backend: backend2,\n      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }\n    });\n    const result = batchMatMulImpl({\n      a: isChannelsLast ? xReshaped : filterReshaped,\n      b: isChannelsLast ? filterReshaped : xReshaped,\n      transposeA: !isChannelsLast,\n      transposeB,\n      backend: backend2,\n      bias,\n      activation: activation2,\n      preluActivationWeights,\n      leakyreluAlpha\n    });\n    out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: convInfo.outShape } });\n    intermediates.push(xReshaped);\n    intermediates.push(filterReshaped);\n    intermediates.push(result);\n  }\n  for (const i of intermediates) {\n    backend2.disposeIntermediateTensorInfo(i);\n  }\n  return out;\n}\nfunction conv2dWithIm2Row({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {\n  const { filterWidth, filterHeight, inChannels, outWidth, outHeight, dataFormat } = convInfo;\n  const isChannelsLast = dataFormat === \"channelsLast\";\n  const sharedDim = filterWidth * filterHeight * inChannels;\n  const numCols = outHeight * outWidth;\n  const x2ColShape = [convInfo.batchSize, sharedDim, numCols];\n  const transposeA = true;\n  const transposeB = false;\n  const intermediates = [];\n  if (preluActivationWeights != null) {\n    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);\n    if (targetShape != null) {\n      preluActivationWeights = reshape4({\n        inputs: { x: preluActivationWeights },\n        backend: backend2,\n        attrs: { shape: targetShape }\n      });\n      intermediates.push(preluActivationWeights);\n    }\n  }\n  if (bias != null) {\n    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);\n    if (targetShape != null) {\n      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });\n      intermediates.push(bias);\n    }\n  }\n  const w2Row = reshape4({\n    inputs: { x: filter },\n    backend: backend2,\n    attrs: { shape: [1, sharedDim, util_exports.sizeFromShape(filter.shape) / sharedDim] }\n  });\n  intermediates.push(w2Row);\n  const im2ColProgram = new Im2ColPackedProgram(x2ColShape, convInfo);\n  const customValues = [\n    x.shape,\n    [convInfo.padInfo.top, convInfo.padInfo.left],\n    [convInfo.strideHeight, convInfo.strideWidth],\n    [convInfo.dilationHeight, convInfo.dilationWidth],\n    [convInfo.inChannels],\n    [convInfo.filterWidth * convInfo.inChannels],\n    [convInfo.outWidth]\n  ];\n  const im2Col = backend2.runWebGLProgram(im2ColProgram, [x], \"float32\", customValues);\n  const im2ColReshaped = reshape4({ inputs: { x: im2Col }, backend: backend2, attrs: { shape: x2ColShape } });\n  intermediates.push(im2Col);\n  intermediates.push(im2ColReshaped);\n  const hasBias = bias != null;\n  const hasPreluActivationWeights = preluActivationWeights != null;\n  const hasLeakyreluAlpha = activation2 === \"leakyrelu\";\n  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;\n  const matmulProgram = new MatMulPackedProgram(isChannelsLast ? im2ColReshaped.shape : w2Row.shape, isChannelsLast ? w2Row.shape : im2ColReshaped.shape, isChannelsLast ? [convInfo.batchSize, numCols, convInfo.outChannels] : [convInfo.batchSize, convInfo.outChannels, numCols], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n  const inputs = isChannelsLast ? [im2ColReshaped, w2Row] : [w2Row, im2ColReshaped];\n  if (bias) {\n    inputs.push(bias);\n  }\n  if (hasPreluActivationWeights) {\n    inputs.push(preluActivationWeights);\n  }\n  if (hasLeakyreluAlpha) {\n    const $leakyreluAlpha = backend2.makeTensorInfo([], \"float32\", util_exports.createScalarValue(leakyreluAlpha, \"float32\"));\n    inputs.push($leakyreluAlpha);\n    intermediates.push($leakyreluAlpha);\n  }\n  const product = backend2.runWebGLProgram(matmulProgram, inputs, \"float32\");\n  const out = reshape4({ inputs: { x: product }, backend: backend2, attrs: { shape: convInfo.outShape } });\n  intermediates.push(product);\n  for (const i of intermediates) {\n    backend2.disposeIntermediateTensorInfo(i);\n  }\n  return out;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2D.js\nfunction conv2d4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);\n  let out;\n  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === \"SAME\" || convInfo.padInfo.type === \"VALID\")) {\n    out = conv2dByMatMul({ x, filter, convInfo, backend: backend2 });\n  } else if (convInfo.strideWidth <= 2 && $dataFormat === \"channelsLast\" && env().getBool(\"WEBGL_EXP_CONV\")) {\n    const program = new Conv2DPackedProgram(convInfo);\n    const customValues = [\n      [convInfo.padInfo.top, convInfo.padInfo.left],\n      [convInfo.strideHeight, convInfo.strideWidth],\n      [convInfo.dilationHeight, convInfo.dilationWidth],\n      [convInfo.inHeight, convInfo.inWidth]\n    ];\n    out = backend2.runWebGLProgram(program, [x, filter], \"float32\", customValues);\n  } else if (env().getBool(\"WEBGL_CONV_IM2COL\")) {\n    out = conv2dWithIm2Row({ x, filter, convInfo, backend: backend2 });\n  } else {\n    const program = new Conv2DProgram(convInfo);\n    out = backend2.runWebGLProgram(program, [x, filter], \"float32\");\n  }\n  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });\n  backend2.disposeIntermediateTensorInfo(out);\n  return outReshaped;\n}\nvar conv2DConfig2 = {\n  kernelName: Conv2D,\n  backendName: \"webgl\",\n  kernelFunc: conv2d4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_backprop_gpu.js\nvar Conv2DDerFilterProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"x\", \"dy\"];\n    this.outputShape = convInfo.filterShape;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int d2 = coords.w;\n\n        // Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${convInfo.batchSize}; b++) {\n          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {\n            int xR = wR + yR * ${strideHeight} - ${padTop};\n\n            if (xR < 0 || xR >= ${convInfo.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {\n              int xC = wC + yC * ${strideWidth} - ${padLeft};\n\n              if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                continue;\n              }\n\n              if (${isChannelsLast}) {\n                float dyValue = getDy(b, yR, yC, d2);\n                float xValue = getX(b, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              } else {\n                float dyValue = getDy(b, d2, yR, yC);\n                float xValue = getX(b, d1, xR, xC);\n                dotProd += (xValue * dyValue);\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar Conv2DDerInputProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\", \"W\"];\n    this.outputShape = convInfo.inShape;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n    const padTop = filterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = filterWidth - 1 - convInfo.padInfo.left;\n    const rowDim = isChannelsLast ? 1 : 2;\n    const colDim = isChannelsLast ? 2 : 3;\n    const channelDim = isChannelsLast ? 3 : 1;\n    this.userCode = `\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[${channelDim}];\n\n        ivec2 dyCorner = ivec2(coords[${rowDim}], coords[${colDim}]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${filterHeight}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${filterHeight} - 1 - wR;\n\n          for (int wC = 0; wC < ${filterWidth}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${filterWidth} - 1 - wC;\n\n            for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {\n\n              if (${isChannelsLast}) {\n                float xValue = getDy(batch, idyR, idyC, d2);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              } else {\n                float xValue = getDy(batch, d2, idyR, idyC);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar Conv3DDerFilterProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"x\", \"dy\"];\n    this.outputShape = convInfo.filterShape;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const padFront = convInfo.padInfo.front;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    this.userCode = `\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int wF = coords.x;\n        int wR = coords.y;\n        int wC = coords.z;\n        int d1 = coords.w;\n        int d2 = coords.u;\n\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${convInfo.batchSize}; b++) {\n          for (int yF = 0; yF < ${convInfo.outDepth}; yF++) {\n            int xF = wF + yF * ${strideDepth} - ${padFront};\n\n            if (xF < 0 || xF >= ${convInfo.inDepth}) {\n              continue;\n            }\n\n            for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {\n              int xR = wR + yR * ${strideHeight} - ${padTop};\n\n              if (xR < 0 || xR >= ${convInfo.inHeight}) {\n                continue;\n              }\n\n              for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {\n                int xC = wC + yC * ${strideWidth} - ${padLeft};\n\n                if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                  continue;\n                }\n\n                float dyValue = getDy(b, yF, yR, yC, d2);\n                float xValue = getX(b, xF, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar Conv3DDerInputProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\", \"W\"];\n    this.outputShape = convInfo.inShape;\n    const filterDepth = convInfo.filterDepth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const padFront = filterDepth - 1 - convInfo.padInfo.front;\n    const padTop = filterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = filterWidth - 1 - convInfo.padInfo.left;\n    this.userCode = `\n      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.u;\n\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyFCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${filterDepth}; wF++) {\n          float dyF = float(dyFCorner + wF) / ${strideDepth}.0;\n\n          if (dyF < 0.0 || dyF >= ${convInfo.outDepth}.0 || fract(dyF) > 0.0) {\n            continue;\n          }\n          int idyF = int(dyF);\n\n          int wFPerm = ${filterDepth} - 1 - wF;\n\n          for (int wR = 0; wR < ${filterHeight}; wR++) {\n            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||\n              fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            int wRPerm = ${filterHeight} - 1 - wR;\n\n            for (int wC = 0; wC < ${filterWidth}; wC++) {\n              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              int wCPerm = ${filterWidth} - 1 - wC;\n\n              for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {\n                float xValue = getDy(batch, idyF, idyR, idyC, d2);\n                float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2DBackpropFilter.js\nfunction conv2DBackpropFilter3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);\n  const program = new Conv2DDerFilterProgram(convInfo);\n  return backend2.runWebGLProgram(program, [x, dy], \"float32\");\n}\nvar conv2DBackpropFilterConfig2 = {\n  kernelName: Conv2DBackpropFilter,\n  backendName: \"webgl\",\n  kernelFunc: conv2DBackpropFilter3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2DBackpropInput.js\nfunction conv2DBackpropInput3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);\n  const program = new Conv2DDerInputProgram(convInfo);\n  return backend2.runWebGLProgram(program, [dy, filter], \"float32\");\n}\nvar conv2DBackpropInputConfig2 = {\n  kernelName: Conv2DBackpropInput,\n  backendName: \"webgl\",\n  kernelFunc: conv2DBackpropInput3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3D.js\nfunction conv3D2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dilations } = attrs;\n  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);\n  const program = new Conv3DProgram(convInfo);\n  return backend2.runWebGLProgram(program, [x, filter], \"float32\");\n}\nvar conv3DConfig2 = {\n  kernelName: Conv3D,\n  backendName: \"webgl\",\n  kernelFunc: conv3D2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3DBackpropFilterV2.js\nfunction conv3DBackpropFilterV22(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, pad: pad3, filterShape } = attrs;\n  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);\n  const program = new Conv3DDerFilterProgram(convInfo);\n  return backend2.runWebGLProgram(program, [x, dy], \"float32\");\n}\nvar conv3DBackpropFilterV2Config2 = {\n  kernelName: Conv3DBackpropFilterV2,\n  backendName: \"webgl\",\n  kernelFunc: conv3DBackpropFilterV22\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3DBackpropInputV2.js\nfunction conv3DBackpropInput2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { pad: pad3, strides, inputShape } = attrs;\n  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);\n  const program = new Conv3DDerInputProgram(convInfo);\n  return backend2.runWebGLProgram(program, [dy, filter], \"float32\");\n}\nvar conv3DBackpropInputConfig = {\n  kernelName: Conv3DBackpropInputV2,\n  backendName: \"webgl\",\n  kernelFunc: conv3DBackpropInput2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cos.js\nvar COS = CHECK_NAN_SNIPPET_UNARY + `\n  return cos(x);\n`;\nvar cos3 = unaryKernelFunc2({ opSnippet: COS });\nvar cosConfig2 = {\n  kernelName: Cos,\n  backendName: \"webgl\",\n  kernelFunc: cos3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cosh.js\nvar COSH = `\n  float e2x = exp(-x);\n  return (e2x + 1.0 / e2x) / 2.0;\n`;\nvar cosh3 = unaryKernelFunc2({ opSnippet: COSH });\nvar coshConfig2 = {\n  kernelName: Cosh,\n  backendName: \"webgl\",\n  kernelFunc: cosh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/crop_and_resize_gpu.js\nvar CropAndResizeProgram = class {\n  constructor(imageShape, boxShape, cropSize, method, extrapolationValue) {\n    this.variableNames = [\"Image\", \"Boxes\", \"BoxInd\"];\n    this.outputShape = [];\n    const [batch, imageHeight, imageWidth, depth] = imageShape;\n    const [numBoxes] = boxShape;\n    const [cropHeight, cropWidth] = cropSize;\n    this.outputShape = [numBoxes, cropHeight, cropWidth, depth];\n    const methodId = method === \"bilinear\" ? 1 : 0;\n    const [inputHeightFloat, inputWidthFloat] = [`${imageHeight - 1}.0`, `${imageWidth - 1}.0`];\n    const [heightRatio, heightScale, inY] = cropHeight > 1 ? [\n      `${(imageHeight - 1) / (cropHeight - 1)}`,\n      \"(y2-y1) * height_ratio\",\n      `y1*${inputHeightFloat} + float(y)*(height_scale)`\n    ] : [\n      \"0.0\",\n      \"0.0\",\n      `0.5 * (y1+y2) * ${inputHeightFloat}`\n    ];\n    const [widthRatio, widthScale, inX] = cropWidth > 1 ? [\n      `${(imageWidth - 1) / (cropWidth - 1)}`,\n      \"(x2-x1) * width_ratio\",\n      `x1*${inputWidthFloat} + float(x)*(width_scale)`\n    ] : [\n      \"0.0\",\n      \"0.0\",\n      `0.5 * (x1+x2) * ${inputWidthFloat}`\n    ];\n    this.userCode = `\n      const float height_ratio = float(${heightRatio});\n      const float width_ratio = float(${widthRatio});\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int y = coords[1];\n        int x = coords[2];\n        int d = coords[3];\n\n        // get box vals\n        float y1 = getBoxes(b,0);\n        float x1 = getBoxes(b,1);\n        float y2 = getBoxes(b,2);\n        float x2 = getBoxes(b,3);\n\n        // get image in batch index\n        int bInd = round(getBoxInd(b));\n        if(bInd < 0 || bInd >= ${batch}) {\n          return;\n        }\n\n        float height_scale = ${heightScale};\n        float width_scale = ${widthScale};\n\n        float in_y = ${inY};\n        if( in_y < 0.0 || in_y > ${inputHeightFloat} ) {\n          setOutput(float(${extrapolationValue}));\n          return;\n        }\n        float in_x = ${inX};\n        if( in_x < 0.0 || in_x > ${inputWidthFloat} ) {\n          setOutput(float(${extrapolationValue}));\n          return;\n        }\n\n        vec2 sourceFracIndexCR = vec2(in_x,in_y);\n        if(${methodId} == 1) {\n          // Compute the four integer indices.\n          ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);\n          ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));\n\n          float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);\n          float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);\n          float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);\n          float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);\n\n          vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);\n\n          float top = topLeft + (topRight - topLeft) * fracCR.x;\n          float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;\n          float newValue = top + (bottom - top) * fracCR.y;\n          setOutput(newValue);\n        } else {\n          // Compute the coordinators of nearest neighbor point.\n          ivec2 sourceNearestCR = ivec2(floor(\n            sourceFracIndexCR + vec2(0.5,0.5)));\n          float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);\n          setOutput(newValue);\n        }\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/CropAndResize.js\nvar cropAndResize3 = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { image: image2, boxes, boxInd } = inputs;\n  const { cropSize, method, extrapolationValue } = attrs;\n  const program = new CropAndResizeProgram(image2.shape, boxes.shape, cropSize, method, extrapolationValue);\n  return backend2.runWebGLProgram(program, [image2, boxes, boxInd], \"float32\");\n};\nvar cropAndResizeConfig2 = {\n  kernelName: CropAndResize,\n  backendName: \"webgl\",\n  kernelFunc: cropAndResize3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/cum_gpu.js\nvar CumOpType;\n(function(CumOpType2) {\n  CumOpType2[\"Prod\"] = \"*\";\n  CumOpType2[\"Sum\"] = \"+\";\n})(CumOpType || (CumOpType = {}));\nvar CumProgram = class {\n  constructor(op2, outputShape, exclusive, reverse5) {\n    this.op = op2;\n    this.outputShape = outputShape;\n    this.variableNames = [\"x\"];\n    this.customUniforms = [{ name: \"index\", type: \"float\" }];\n    const rank = this.outputShape.length;\n    const initVal = this.op === CumOpType.Prod ? \"1.0\" : \"0.0\";\n    const val = exclusive ? initVal : `getX(${getCoords2(rank, \"coords\", this.op)})`;\n    const length = this.outputShape[this.outputShape.length - 1];\n    let condition = \"\";\n    let idxString = \"\";\n    if (exclusive) {\n      condition = reverse5 ? `end != ${length - 1}` : \"end != 0\";\n      idxString = reverse5 ? \"end + 1\" : \"end - 1\";\n    } else {\n      condition = reverse5 ? `end + pow2 < ${length}` : \"end >= pow2\";\n      idxString = reverse5 ? \"end + pow2\" : \"end - pow2\";\n    }\n    this.userCode = `\n      void main() {\n        ${getCoordsDataType(rank)} coords = getOutputCoords();\n        int end = ${getFinalCoord(rank, \"coords\", this.op)};\n        float val = ${val};\n        int pow2 = int(pow(2.0, index));\n        if (${condition}) {\n          int idx = ${idxString};\n          ${getFinalCoord(rank, \"coords\", this.op)} = idx;\n          val ${this.op}= getX(${getCoords2(rank, \"coords\", this.op)});\n        }\n        setOutput(val);\n      }\n    `;\n  }\n};\nfunction getCoords2(rank, name, op2) {\n  if (rank === 1) {\n    return `${name}`;\n  } else if (rank === 2) {\n    return `${name}.x, ${name}.y`;\n  } else if (rank === 3) {\n    return `${name}.x, ${name}.y, ${name}.z`;\n  } else if (rank === 4) {\n    return `${name}.x, ${name}.y, ${name}.z, ${name}.w`;\n  } else {\n    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);\n  }\n}\nfunction getFinalCoord(rank, name, op2) {\n  if (rank === 1) {\n    return `${name}`;\n  } else if (rank === 2) {\n    return `${name}.y`;\n  } else if (rank === 3) {\n    return `${name}.z`;\n  } else if (rank === 4) {\n    return `${name}.w`;\n  } else {\n    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);\n  }\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cum_impl.js\nfunction cumImpl(op2, x, backend2, axis, exclusive, reverse5) {\n  const xRank = x.shape.length;\n  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);\n  let permutedX = x;\n  if (permutation != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });\n  }\n  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];\n  if (permutedAxis !== xRank - 1) {\n    throw new Error(`WebGL cumprod shader expects an inner-most axis=${x.shape.length - 1} but got axis=${axis}`);\n  }\n  const size = permutedX.shape[permutedAxis];\n  let result = identity3({ inputs: { x: permutedX }, backend: backend2 });\n  for (let i = 0; i <= Math.ceil(Math.log2(size)) - 1; i++) {\n    const program = new CumProgram(op2, permutedX.shape, false, reverse5);\n    const customValues = [[i]];\n    const prevResult = result;\n    result = backend2.runWebGLProgram(program, [result], result.dtype, customValues);\n    backend2.disposeIntermediateTensorInfo(prevResult);\n  }\n  if (exclusive) {\n    const program = new CumProgram(op2, permutedX.shape, exclusive, reverse5);\n    const prevResult = result;\n    result = backend2.runWebGLProgram(program, [result], result.dtype);\n    backend2.disposeIntermediateTensorInfo(prevResult);\n  }\n  if (permutation != null) {\n    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);\n    const reverseTransposedResult = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });\n    backend2.disposeIntermediateTensorInfo(result);\n    backend2.disposeIntermediateTensorInfo(permutedX);\n    return reverseTransposedResult;\n  }\n  return result;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cumprod.js\nfunction cumprod3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  return cumImpl(CumOpType.Prod, x, backend2, axis, exclusive, reverse5);\n}\nvar cumprodConfig2 = {\n  kernelName: Cumprod,\n  backendName: \"webgl\",\n  kernelFunc: cumprod3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cumsum.js\nfunction cumsum3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  return cumImpl(CumOpType.Sum, x, backend2, axis, exclusive, reverse5);\n}\nvar cumsumConfig2 = {\n  kernelName: Cumsum,\n  backendName: \"webgl\",\n  kernelFunc: cumsum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DenseBincount.js\nfunction denseBincount3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, weights } = inputs;\n  const { size, binaryOutput } = attrs;\n  if (x.shape.length === 1) {\n    const xVals = backend2.readSync(x.dataId);\n    const weightsVals = backend2.readSync(weights.dataId);\n    const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);\n    return backend2.makeTensorInfo([size], weights.dtype, outVals);\n  } else if (x.shape.length === 2) {\n    const xBuf = backend2.bufferSync(x);\n    const weightsBuf = backend2.bufferSync(weights);\n    const outBuf = bincountReduceImplCPU(xBuf, weightsBuf, size, binaryOutput);\n    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);\n  }\n  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);\n}\nvar denseBincountConfig2 = {\n  kernelName: DenseBincount,\n  backendName: \"webgl\",\n  kernelFunc: denseBincount3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/depth_to_space_gpu.js\nvar DepthToSpaceProgram = class {\n  constructor(outputShape, blockSize, dataFormat) {\n    this.variableNames = [\"x\"];\n    this.outputShape = [];\n    this.outputShape = outputShape;\n    this.blockSize = blockSize;\n    this.dataFormat = dataFormat;\n    this.userCode = `\n    void main() {\n      ivec4 coords = getOutputCoords();\n      int b = coords[0];\n      int h = ${this.getHeightCoordString()};\n      int w = ${this.getWidthCoordString()};\n      int d = ${this.getDepthCoordString()};\n\n      int in_h = h / ${blockSize};\n      int offset_h = imod(h, ${blockSize});\n      int in_w = w / ${blockSize};\n      int offset_w = imod(w, ${blockSize});\n      int offset_d = (offset_h * ${blockSize} + offset_w) *\n        ${this.getOutputDepthSize()};\n      int in_d = d + offset_d;\n\n      float result = ${this.getInputSamplingString()};\n      setOutput(result);\n    }\n  `;\n  }\n  getHeightCoordString() {\n    if (this.dataFormat === \"NHWC\") {\n      return `coords[1]`;\n    } else {\n      return `coords[2]`;\n    }\n  }\n  getWidthCoordString() {\n    if (this.dataFormat === \"NHWC\") {\n      return `coords[2]`;\n    } else {\n      return `coords[3]`;\n    }\n  }\n  getDepthCoordString() {\n    if (this.dataFormat === \"NHWC\") {\n      return `coords[3]`;\n    } else {\n      return `coords[1]`;\n    }\n  }\n  getOutputDepthSize() {\n    if (this.dataFormat === \"NHWC\") {\n      return this.outputShape[3];\n    } else {\n      return this.outputShape[1];\n    }\n  }\n  getInputSamplingString() {\n    if (this.dataFormat === \"NHWC\") {\n      return `getX(b, in_h, in_w, in_d)`;\n    } else {\n      return `getX(b, in_d, in_h, in_w)`;\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthToSpace.js\nfunction depthToSpace3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockSize, dataFormat } = attrs;\n  const batchSize = x.shape[0];\n  const inputHeight = dataFormat === \"NHWC\" ? x.shape[1] : x.shape[2];\n  const inputWidth = dataFormat === \"NHWC\" ? x.shape[2] : x.shape[3];\n  const inputDepth = dataFormat === \"NHWC\" ? x.shape[3] : x.shape[1];\n  const outputHeight = inputHeight * blockSize;\n  const outputWidth = inputWidth * blockSize;\n  const outputDepth = inputDepth / (blockSize * blockSize);\n  const outputShape = dataFormat === \"NHWC\" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];\n  const program = new DepthToSpaceProgram(outputShape, blockSize, dataFormat);\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\nvar depthToSpaceConfig2 = {\n  kernelName: DepthToSpace,\n  backendName: \"webgl\",\n  kernelFunc: depthToSpace3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_gpu_depthwise.js\nvar DepthwiseConv2DProgram = class {\n  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {\n    this.variableNames = [\"x\", \"W\"];\n    this.customUniforms = [\n      { name: \"pads\", type: \"ivec2\" },\n      { name: \"strides\", type: \"ivec2\" },\n      { name: \"dilations\", type: \"ivec2\" },\n      { name: \"inDims\", type: \"ivec2\" }\n    ];\n    this.outputShape = convInfo.outShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const channelMul = convInfo.outChannels / convInfo.inChannels;\n    let activationSnippet = \"\", applyActivationSnippet = \"\";\n    if (activation2) {\n      if (hasPreluActivation) {\n        activationSnippet = `float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${activation2}\n        }`;\n      } else if (hasLeakyReluAlpha) {\n        activationSnippet = `float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${activation2}\n        }`;\n      } else {\n        activationSnippet = `\n          float activation(float x) {\n            ${activation2}\n          }\n        `;\n      }\n      applyActivationSnippet = `result = activation(result);`;\n    }\n    const addBiasSnippet = addBias ? \"result += getBiasAtOutCoords();\" : \"\";\n    if (addBias) {\n      this.variableNames.push(\"bias\");\n    }\n    if (hasPreluActivation) {\n      this.variableNames.push(\"preluActivationWeights\");\n    }\n    if (hasLeakyReluAlpha) {\n      this.variableNames.push(\"leakyreluAlpha\");\n    }\n    this.userCode = `\n      ${activationSnippet}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${channelMul};\n        int q = d2 - d1 * ${channelMul};\n\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.\n        for (int wR = 0; wR < ${filterHeight}; wR++) {\n          int xR = xRCorner + wR * dilations[0];\n\n          if (xR < 0 || xR >= inDims[0]) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${filterWidth}; wC++) {\n            int xC = xCCorner + wC * dilations[1];\n\n            if (xC < 0 || xC >= inDims[1]) {\n              continue;\n            }\n\n            float xVal = getX(batch, xR, xC, d1);\n            float wVal = getW(wR, wC, d1, q);\n            dotProd += xVal * wVal;\n          }\n        }\n\n        float result = dotProd;\n        ${addBiasSnippet}\n        ${applyActivationSnippet}\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_packed_gpu_depthwise.js\nvar DepthwiseConvPacked2DProgram = class {\n  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {\n    this.variableNames = [\"x\", \"W\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [\n      { name: \"pads\", type: \"ivec2\" },\n      { name: \"strides\", type: \"ivec2\" },\n      { name: \"dilations\", type: \"ivec2\" },\n      { name: \"inDims\", type: \"ivec2\" }\n    ];\n    this.outputShape = convInfo.outShape;\n    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);\n    const channelMul = convInfo.outChannels / convInfo.inChannels;\n    const padLeft = convInfo.padInfo.left;\n    const strideWidth = convInfo.strideWidth;\n    const dilationWidth = convInfo.dilationWidth;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const texelsAcross = filterWidth;\n    let mainLoop = `\n      int xR; int xC; int xCOffset;\n      vec4 wTexel; vec4 previous; vec4 final;`;\n    for (let c = 0; c < filterWidth; c++) {\n      mainLoop += `\n          vec4 xTexelC${c * 2};\n          int xTexelC${c * 2}Ready;\n          vec4 xTexelC${c * 2 + 1};\n          int xTexelC${c * 2 + 1}Ready;\n          vec4 xC${c};`;\n    }\n    mainLoop += `\n    for (int r = 0; r < ${filterHeight}; r++) {\n      `;\n    for (let c = 0; c < filterWidth; c++) {\n      mainLoop += `\n          xTexelC${c * 2} = vec4(0.0);\n          xTexelC${c * 2}Ready = 0;\n          xTexelC${c * 2 + 1} = vec4(0.0);\n          xTexelC${c * 2 + 1}Ready = 0;\n          xC${c} = vec4(0.0);`;\n    }\n    mainLoop += `\n        xR = xRCorner + r * dilations[0];\n        if (xR >=0 && xR < inDims[0]) {\n      `;\n    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {\n      const colIndex = texelC * 2;\n      mainLoop += `\n          xC = xCCorner + ${colIndex * dilationWidth};\n          `;\n      if (strideWidth === 1) {\n        if (colIndex < filterWidth) {\n          if (padLeft % 2 === 1) {\n            mainLoop += `\n                xCOffset = xC + 1;\n                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);\n\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${colIndex}.zw = vec2(0.0);\n                  }\n                  xTexelC${colIndex}Ready = 1;\n                }\n              `;\n            if (dilationWidth === 1 && colIndex > 0) {\n              mainLoop += `\n                xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);\n                `;\n            } else {\n              mainLoop += `\n                  xCOffset = xC + 1 - 2;\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                    previous = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      previous.zw = vec2(0.0);\n                    }\n\n                    xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);\n                  } else {\n                    xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);\n                  }\n                  `;\n            }\n          } else {\n            mainLoop += `\n                if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                  xTexelC${colIndex} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${colIndex}.zw = vec2(0.0);\n                  }\n                  xTexelC${colIndex}Ready = 1;\n                }\n\n                xC${colIndex} = xTexelC${colIndex};\n                `;\n          }\n          if (colIndex + 1 < filterWidth) {\n            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;\n            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {\n              mainLoop += `\n                  xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                    xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      xTexelC${colIndex + 1}.zw = vec2(0.0);\n                    }\n                    xTexelC${colIndex + 1}Ready = 1;\n                  }\n                  `;\n              if (dilationWidth > 1) {\n                mainLoop += `\n                    xCOffset -= 2;\n                    if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n                     xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);\n                    } else {\n                     xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);\n                    }\n                    `;\n              } else {\n                mainLoop += `\n                    xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);\n                    `;\n              }\n            } else {\n              if (nextTexelOffset === 1) {\n                mainLoop += `\n                    xC${colIndex + 1} = xTexelC${colIndex};\n                    `;\n              } else {\n                mainLoop += `\n                    xCOffset = xC + ${nextTexelOffset};\n\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                      xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n                      if (xCOffset + 1 >= inDims[1]) {\n                        xTexelC${colIndex + 1}.zw = vec2(0.0);\n                      }\n                      xTexelC${colIndex + 1}Ready = 1;\n                    }\n\n                    xC${colIndex + 1} = xTexelC${colIndex + 1};\n                    `;\n              }\n            }\n          }\n        }\n      } else {\n        if (colIndex < filterWidth) {\n          if (padLeft % 2 === 1) {\n            mainLoop += `\n                xCOffset = xC + 1 - strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${colIndex}.zw = vec2(0.0);\n                  }\n                  xTexelC${colIndex}Ready = 1;\n                }\n\n                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                  xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xC + 2 >= inDims[1]) {\n                    xTexelC${colIndex + 1}.zw = vec2(0.0);\n                  }\n                  xTexelC${colIndex + 1}Ready = 1;\n                }\n\n                xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);\n              `;\n            if (colIndex + 1 < filterWidth) {\n              mainLoop += `\n                  final = vec4(0.0);\n                  xCOffset = xC + 1 + strides[1];\n                  if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                    final = getX(batch, xR, xCOffset, d1);\n                  }\n                  xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);\n                `;\n            }\n          } else {\n            mainLoop += `\n                if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {\n                  xTexelC${colIndex} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${colIndex}.zw = vec2(0.0);\n                  }\n                  xTexelC${colIndex}Ready = 1;\n                }\n\n                xCOffset = xC + strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {\n                  xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${colIndex + 1}.zw = vec2(0.);\n                  }\n                  xTexelC${colIndex + 1}Ready = 1;\n                }\n\n                xC${colIndex} = vec4(\n                  xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);\n              `;\n            if (colIndex + 1 < filterWidth) {\n              mainLoop += `\n                  xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);\n                `;\n            }\n          }\n        }\n      }\n      if (colIndex < filterWidth) {\n        mainLoop += `\n            wTexel = getW(r, ${colIndex}, d1, q);\n            dotProd += xC${colIndex} * vec4(wTexel.xz, wTexel.xz);\n          `;\n        if (colIndex + 1 < filterWidth) {\n          mainLoop += `\n              wTexel = getW(r, ${colIndex + 1}, d1, q);\n              dotProd += xC${colIndex + 1} * vec4(wTexel.xz, wTexel.xz);\n            `;\n        }\n      }\n    }\n    mainLoop += `\n    }\n  `;\n    mainLoop += `\n      }\n    `;\n    let activationSnippet = \"\", applyActivationSnippet = \"\";\n    if (activation2) {\n      if (hasPreluActivation) {\n        activationSnippet = `vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${activation2}\n        }`;\n      } else if (hasLeakyReluAlpha) {\n        activationSnippet = `vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${activation2}\n        }`;\n      } else {\n        activationSnippet = `vec4 activation(vec4 x) {\n          ${activation2}\n        }`;\n      }\n      applyActivationSnippet = `result = activation(result);`;\n    }\n    const addBiasSnippet = addBias ? \"result += getBiasAtOutCoords();\" : \"\";\n    if (addBias) {\n      this.variableNames.push(\"bias\");\n    }\n    if (hasPreluActivation) {\n      this.variableNames.push(\"preluActivationWeights\");\n    }\n    if (hasLeakyReluAlpha) {\n      this.variableNames.push(\"leakyreluAlpha\");\n    }\n    this.userCode = `\n      ${activationSnippet}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${channelMul};\n        int q = d2 - d1 * ${channelMul};\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n        vec4 dotProd = vec4(0.000000000000001);\n\n        ${mainLoop}\n\n        vec4 result = dotProd - vec4(0.000000000000001);\n        ${addBiasSnippet}\n        ${applyActivationSnippet}\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNative.js\nfunction depthwiseConv2dNative2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;\n  let $dilations = dilations;\n  if ($dilations == null) {\n    $dilations = [1, 1];\n  }\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);\n  let program;\n  if (env().getBool(\"WEBGL_PACK_DEPTHWISECONV\") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1) {\n    program = new DepthwiseConvPacked2DProgram(convInfo);\n  } else {\n    program = new DepthwiseConv2DProgram(convInfo);\n  }\n  const customValues = [\n    [convInfo.padInfo.top, convInfo.padInfo.left],\n    [convInfo.strideHeight, convInfo.strideWidth],\n    [convInfo.dilationHeight, convInfo.dilationWidth],\n    [convInfo.inHeight, convInfo.inWidth]\n  ];\n  return backend2.runWebGLProgram(program, [x, filter], \"float32\", customValues);\n}\nvar depthwiseConv2dNativeConfig2 = {\n  kernelName: DepthwiseConv2dNative,\n  backendName: \"webgl\",\n  kernelFunc: depthwiseConv2dNative2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_backprop_gpu_depthwise.js\nvar DepthwiseConv2DDerFilterProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"x\", \"dy\"];\n    this.outputShape = convInfo.filterShape;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const padTop = convInfo.padInfo.top;\n    const padLeft = convInfo.padInfo.left;\n    const channelMul = convInfo.outChannels / convInfo.inChannels;\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int dm = coords.w;\n        int d2 = d1 * ${channelMul} + dm;\n\n        float dotProd = 0.0;\n\n        // TO DO: Vec4 over the batch size\n        for (int b = 0; b < ${convInfo.batchSize}; b++) {\n          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {\n            int xR = wR + yR * ${strideHeight} - ${padTop};\n\n            if (xR < 0 || xR >= ${convInfo.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {\n              int xC = wC + yC * ${strideWidth} - ${padLeft};\n\n              if (xC < 0 || xC >= ${convInfo.inWidth}) {\n                continue;\n              }\n\n              float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar DepthwiseConv2DDerInputProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\", \"W\"];\n    this.outputShape = convInfo.inShape;\n    const filterHeight = convInfo.filterHeight;\n    const filterWidth = convInfo.filterWidth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const padTop = filterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = filterWidth - 1 - convInfo.padInfo.left;\n    const channelMul = convInfo.outChannels / convInfo.inChannels;\n    this.userCode = `\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n        ivec2 dyCorner = coords.yz - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        float dotProd = 0.0;\n\n        for (int wR = 0; wR < ${filterHeight}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${filterHeight} - 1 - wR;\n\n          for (int wC = 0; wC < ${filterWidth}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${filterWidth} - 1 - wC;\n\n            // TO DO: Vec4 over the channelMul\n            for (int dm = 0; dm < ${channelMul}; dm++) {\n              int d2 = d1 * ${channelMul} + dm;\n              float xValue = getDy(batch, idyR, idyC, d2);\n              float wValue = getW(wRPerm, wCPerm, d1, dm);\n              dotProd += xValue * wValue;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNativeBackpropFilter.js\nfunction depthwiseConv2dNativeBackpropFilter3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, dy } = inputs;\n  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);\n  const program = new DepthwiseConv2DDerFilterProgram(convInfo);\n  return backend2.runWebGLProgram(program, [x, dy], \"float32\");\n}\nvar depthwiseConv2dNativeBackpropFilterConfig2 = {\n  kernelName: DepthwiseConv2dNativeBackpropFilter,\n  backendName: \"webgl\",\n  kernelFunc: depthwiseConv2dNativeBackpropFilter3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNativeBackpropInput.js\nfunction depthwiseConv2dNativeBackpropInput3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, filter } = inputs;\n  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;\n  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);\n  const program = new DepthwiseConv2DDerInputProgram(convInfo);\n  return backend2.runWebGLProgram(program, [dy, filter], \"float32\");\n}\nvar depthwiseConv2dNativeBackpropInputConfig2 = {\n  kernelName: DepthwiseConv2dNativeBackpropInput,\n  backendName: \"webgl\",\n  kernelFunc: depthwiseConv2dNativeBackpropInput3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/diag_gpu.js\nvar DiagProgram = class {\n  constructor(size) {\n    this.variableNames = [\"X\"];\n    this.outputShape = [size, size];\n    this.userCode = `\n      void main() {\n          ivec2 coords = getOutputCoords();\n          float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;\n          setOutput(val);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Diag.js\nfunction diag3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  const outShape = [...x.shape, ...x.shape];\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const flat = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [xSize] } });\n  const program = new DiagProgram(xSize);\n  const res = backend2.runWebGLProgram(program, [flat], flat.dtype);\n  const out = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outShape } });\n  backend2.disposeIntermediateTensorInfo(flat);\n  backend2.disposeIntermediateTensorInfo(res);\n  return out;\n}\nvar diagConfig2 = {\n  kernelName: Diag,\n  backendName: \"webgl\",\n  kernelFunc: diag3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/dilation_gpu.js\nvar Dilation2DProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"x\", \"W\"];\n    this.outputShape = convInfo.outShape;\n    const { inHeight, inWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth } = convInfo;\n    const { top: padTop, left: padLeft } = padInfo;\n    this.userCode = `\n      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n      const float neg_infinity = -3.4e38;\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.w;\n        ivec2 outTopLeftCorner =\n            coords.yz * strides - pads;\n        int hBeg = outTopLeftCorner.x;\n        int wBeg = outTopLeftCorner.y;\n\n        float curVal = neg_infinity;\n        for (int h = 0; h < ${filterHeight}; h++) {\n          int hIn = hBeg + h * ${dilationHeight};\n\n          if (hIn >= 0 && hIn < ${inHeight}) {\n            for (int w = 0; w < ${filterWidth}; w++) {\n              int wIn = wBeg + w * ${dilationWidth};\n\n              if (wIn >= 0 && wIn < ${inWidth}) {\n                float xVal = getX(batch, hIn, wIn, d1);\n                float wVal = getW(h, w, d1);\n\n                float val = xVal + wVal;\n                if (val > curVal) {\n                  curVal = val;\n                }\n              }\n            }\n          }\n        }\n\n        float result = curVal;\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Dilation2D.js\nfunction dilation2D(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter } = inputs;\n  const { strides, pad: pad3, dilations } = attrs;\n  const convInfo = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, \"NHWC\", dilations);\n  let out;\n  const program = new Dilation2DProgram(convInfo);\n  out = backend2.runWebGLProgram(program, [x, filter], \"float32\");\n  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });\n  backend2.disposeIntermediateTensorInfo(out);\n  return outReshaped;\n}\nvar dilation2DConfig2 = {\n  kernelName: Dilation2D,\n  backendName: \"webgl\",\n  kernelFunc: dilation2D\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Einsum.js\nfunction einsum3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { equation } = attrs;\n  const tensors = inputs;\n  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);\n  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);\n  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);\n  const nSteps = steps.length;\n  let out = null;\n  let numDimsRemaining = allDims.length;\n  const tensorsToDispose = [];\n  for (let i = 0; i < nSteps; ++i) {\n    for (const idTerm of steps[i]) {\n      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);\n      let x;\n      if (backend_util_exports.isIdentityPermutation(perm)) {\n        x = tensors[idTerm];\n      } else {\n        x = transpose3({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });\n        tensorsToDispose.push(x);\n      }\n      const targetShape = x.shape.slice();\n      for (let k = 0; k < dimsToExpand.length; ++k) {\n        targetShape.splice(dimsToExpand[k], 0, 1);\n      }\n      if (!util_exports.arraysEqual(x.shape, targetShape)) {\n        x = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });\n        tensorsToDispose.push(x);\n      }\n      if (out === null) {\n        out = x;\n      } else {\n        out = multiply3({ inputs: { a: x, b: out }, backend: backend2 });\n        tensorsToDispose.push(out);\n      }\n    }\n    if (i < nSteps - 1) {\n      if (path[i] >= 0) {\n        out = sum4({\n          inputs: { x: out },\n          backend: backend2,\n          attrs: {\n            axis: path[i] - (allDims.length - numDimsRemaining),\n            keepDims: false\n          }\n        });\n        tensorsToDispose.push(out);\n      }\n      numDimsRemaining--;\n    }\n  }\n  for (const tensorInfo of tensorsToDispose) {\n    if (tensorInfo === out) {\n      continue;\n    }\n    backend2.disposeIntermediateTensorInfo(tensorInfo);\n  }\n  return out;\n}\nvar einsumConfig2 = {\n  kernelName: Einsum,\n  backendName: \"webgl\",\n  kernelFunc: einsum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Elu.js\nvar ELU4 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;\nvar ELU_PACKED = `\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n`;\nvar elu5 = unaryKernelFunc2({ opSnippet: ELU4, packedOpSnippet: ELU_PACKED });\nvar eluConfig2 = {\n  kernelName: Elu,\n  backendName: \"webgl\",\n  kernelFunc: elu5\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/EluGrad.js\nvar ELU_DER = `return (b >= 1.0) ? a : a * (b + 1.0);`;\nvar ELU_DER_PACKED = `\n  vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));\n  return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));\n`;\nvar eluGrad2 = (args) => {\n  const { inputs, backend: backend2 } = args;\n  const { dy, y } = inputs;\n  const program = env().getBool(\"WEBGL_PACK_BINARY_OPERATIONS\") ? new BinaryOpPackedProgram(ELU_DER_PACKED, dy.shape, y.shape) : new BinaryOpProgram(ELU_DER, dy.shape, y.shape);\n  return backend2.runWebGLProgram(program, [dy, y], dy.dtype);\n};\nvar eluGradConfig3 = {\n  kernelName: EluGrad,\n  backendName: \"webgl\",\n  kernelFunc: eluGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Equal.js\nvar PACKED_EQUAL = `\n  return vec4(equal(a, b));\n`;\nvar EQUAL = `return float(a == b);`;\nvar equal3 = binaryKernelFunc2({\n  opSnippet: EQUAL,\n  packedOpSnippet: PACKED_EQUAL,\n  dtype: \"bool\",\n  cpuKernelImpl: equalImplCPU\n});\nvar equalConfig2 = {\n  kernelName: Equal,\n  backendName: \"webgl\",\n  kernelFunc: equal3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Erf.js\nvar ERF = `\n  // Error function is calculated approximately with elementary function.\n  // See \"Handbook of Mathematical Functions with Formulas,\n  // Graphs, and Mathematical Tables\", Abramowitz and Stegun.\n  float p = ${backend_util_exports.ERF_P};\n  float a1 = ${backend_util_exports.ERF_A1};\n  float a2 = ${backend_util_exports.ERF_A2};\n  float a3 = ${backend_util_exports.ERF_A3};\n  float a4 = ${backend_util_exports.ERF_A4};\n  float a5 = ${backend_util_exports.ERF_A5};\n\n  float sign = sign(x);\n  x = abs(x);\n  float t = 1.0 / (1.0 + p * x);\n  return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));\n`;\nvar erf3 = unaryKernelFunc2({ opSnippet: ERF });\nvar erfConfig2 = {\n  kernelName: Erf,\n  backendName: \"webgl\",\n  kernelFunc: erf3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Exp.js\nvar EXP = CHECK_NAN_SNIPPET_UNARY + `\n  return exp(x);\n`;\nvar EXP_PACKED = `\n  vec4 result = exp(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar exp3 = unaryKernelFunc2({\n  opSnippet: EXP,\n  packedOpSnippet: EXP_PACKED,\n  cpuKernelImpl: expImplCPU,\n  dtype: \"float32\"\n});\nvar expConfig2 = {\n  kernelName: Exp,\n  backendName: \"webgl\",\n  kernelFunc: exp3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ExpandDims.js\nfunction expandDims4(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { dim } = attrs;\n  const { input: input2 } = inputs;\n  const inputRank = input2.shape.length;\n  const newShape = input2.shape.slice();\n  let $dim = dim;\n  if (dim < 0) {\n    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);\n    $dim = inputRank + dim + 1;\n  }\n  newShape.splice($dim, 0, 1);\n  return reshape4({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });\n}\nvar expandDimsConfig2 = {\n  kernelName: ExpandDims,\n  backendName: \"webgl\",\n  kernelFunc: expandDims4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Expm1.js\nvar EXPM1 = `return exp(x) - 1.0;`;\nvar expm13 = unaryKernelFunc2({ opSnippet: EXPM1, packedOpSnippet: EXPM1, cpuKernelImpl: expm1ImplCPU });\nvar expm1Config2 = {\n  kernelName: Expm1,\n  backendName: \"webgl\",\n  kernelFunc: expm13\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/fft_gpu.js\nvar FFTProgram = class {\n  constructor(component, inputShape, inverse) {\n    this.variableNames = [\"real\", \"imag\"];\n    const innerDim = inputShape[1];\n    this.outputShape = inputShape;\n    const exponentMultiplierSnippet = inverse ? `2.0 * ${Math.PI}` : `-2.0 * ${Math.PI}`;\n    const resultDenominator = inverse ? `${innerDim}.0` : \"1.0\";\n    let opString;\n    if (component === \"real\") {\n      opString = \"return real * expR - imag * expI;\";\n    } else if (component === \"imag\") {\n      opString = \"return real * expI + imag * expR;\";\n    } else {\n      throw new Error(`FFT component must be either \"real\" or \"imag\", got ${component}.`);\n    }\n    this.userCode = `\n      const float exponentMultiplier = ${exponentMultiplierSnippet};\n\n      float unaryOpComplex(float real, float expR, float imag, float expI) {\n        ${opString}\n      }\n\n      float mulMatDFT(int batch, int index) {\n        float indexRatio = float(index) / float(${innerDim});\n        float exponentMultiplierTimesIndexRatio =\n            exponentMultiplier * indexRatio;\n\n        float result = 0.0;\n\n        for (int i = 0; i < ${innerDim}; i++) {\n          // x = (-2|2 * PI / N) * index * i;\n          float x = exponentMultiplierTimesIndexRatio * float(i);\n          float expR = cos(x);\n          float expI = sin(x);\n          float real = getReal(batch, i);\n          float imag = getImag(batch, i);\n\n          result +=\n              unaryOpComplex(real, expR, imag, expI) / ${resultDenominator};\n        }\n\n        return result;\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        setOutput(mulMatDFT(coords[0], coords[1]));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FFT_impl.js\nfunction fftImpl2(x, inverse, backend2) {\n  const xData = backend2.texData.get(x.dataId);\n  const inputSize = util_exports.sizeFromShape(x.shape);\n  const innerDimensionSize = x.shape[x.shape.length - 1];\n  const batch = inputSize / innerDimensionSize;\n  const input2D = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [batch, innerDimensionSize] } });\n  const xShape = input2D.shape;\n  const realProgram = new FFTProgram(\"real\", xShape, inverse);\n  const imagProgram = new FFTProgram(\"imag\", xShape, inverse);\n  const inputs = [\n    {\n      dataId: xData.complexTensorInfos.real.dataId,\n      dtype: xData.complexTensorInfos.real.dtype,\n      shape: xShape\n    },\n    {\n      dataId: xData.complexTensorInfos.imag.dataId,\n      dtype: xData.complexTensorInfos.imag.dtype,\n      shape: xShape\n    }\n  ];\n  const realPart = backend2.runWebGLProgram(realProgram, inputs, \"float32\");\n  const imagPart = backend2.runWebGLProgram(imagProgram, inputs, \"float32\");\n  const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(realPart);\n  backend2.disposeIntermediateTensorInfo(imagPart);\n  const complexOutputReshaped = reshape4({ inputs: { x: complexOutput }, backend: backend2, attrs: { shape: x.shape } });\n  backend2.disposeIntermediateTensorInfo(input2D);\n  backend2.disposeIntermediateTensorInfo(complexOutput);\n  return complexOutputReshaped;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FFT.js\nfunction fft3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  return fftImpl2(input2, false, backend2);\n}\nvar fftConfig2 = {\n  kernelName: FFT,\n  backendName: \"webgl\",\n  kernelFunc: fft3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/fill_gpu.js\nvar FillProgram = class {\n  constructor(shape, value) {\n    this.outputShape = [];\n    this.customUniforms = [{ name: \"value\", type: \"float\" }];\n    this.variableNames = [\"x\"];\n    this.outputShape = shape;\n    this.userCode = `\n      void main() {\n        // Input can be obtained from uniform value.\n        setOutput(value);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Fill.js\nfunction fill3(args) {\n  const { backend: backend2, attrs } = args;\n  const { shape, value } = attrs;\n  let { dtype } = attrs;\n  dtype = dtype || util_exports.inferDtype(value);\n  if (dtype === \"string\") {\n    const values = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(shape));\n    values.fill(value);\n    return backend2.makeTensorInfo(shape, dtype, values);\n  } else {\n    const program = new FillProgram(shape, value);\n    const customValues = [[value]];\n    return backend2.runWebGLProgram(program, [], dtype, customValues);\n  }\n}\nvar fillConfig2 = {\n  kernelName: Fill,\n  backendName: \"webgl\",\n  kernelFunc: fill3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/flip_left_right_gpu.js\nvar FlipLeftRightProgram = class {\n  constructor(imageShape) {\n    this.variableNames = [\"Image\"];\n    this.outputShape = [];\n    const imageWidth = imageShape[2];\n    this.outputShape = imageShape;\n    this.userCode = `\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n\n          int coordX = ${imageWidth} - x - 1;\n          float outputValue;\n          if(coordX >= 0 && coordX < ${imageWidth}) {\n            outputValue = getImage(coords[0], coords[1], coordX, coords[3]);\n          } else {\n            outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FlipLeftRight.js\nvar flipLeftRightConfig2 = {\n  kernelName: FlipLeftRight,\n  backendName: \"webgl\",\n  kernelFunc: ({ inputs, backend: backend2 }) => {\n    const { image: image2 } = inputs;\n    const webglBackend = backend2;\n    const program = new FlipLeftRightProgram(image2.shape);\n    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype);\n    return output;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Floor.js\nvar FLOOR = `return floor(x);`;\nvar floor3 = unaryKernelFunc2({ opSnippet: FLOOR, packedOpSnippet: FLOOR, cpuKernelImpl: floorImplCPU });\nvar floorConfig2 = {\n  kernelName: Floor,\n  backendName: \"webgl\",\n  kernelFunc: floor3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FloorDiv.js\nvar INT_DIV = `\n  float s = sign(a) * sign(b);\n  int ia = round(a);\n  int ib = round(b);\n  if (ib != 0) {\n    // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n    return float(idiv(ia, ib, s));\n  } else {\n    return NAN;\n  }\n`;\nvar INT_DIV_PACKED = `\n  ivec4 ia = round(a);\n  ivec4 ib = round(b);\n  bvec4 cond = notEqual(ib, ivec4(0));\n  ivec4 result = ivec4(0);\n  vec4 s = sign(a) * sign(b);\n\n  // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n  if (cond[0]) {\n    result[0] = idiv(ia[0], ib[0], s[0]);\n  }\n  if (cond[1]) {\n    result[1] = idiv(ia[1], ib[1], s[1]);\n  }\n  if (cond[2]) {\n    result[2] = idiv(ia[2], ib[2], s[2]);\n  }\n  if (cond[3]) {\n    result[3] = idiv(ia[3], ib[3], s[3]);\n  }\n  return vec4(result);\n`;\nvar floorDiv3 = binaryKernelFunc2({ opSnippet: INT_DIV, packedOpSnippet: INT_DIV_PACKED, dtype: \"int32\" });\nvar floorDivConfig2 = {\n  kernelName: FloorDiv,\n  backendName: \"webgl\",\n  kernelFunc: floorDiv3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels_utils/from_pixels_gpu.js\nvar FromPixelsProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    const glsl = getGlslDifferences();\n    const [height, width] = outputShape;\n    this.outputShape = outputShape;\n    this.userCode = `\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${width}.0, ${height}.0);\n\n        vec4 values = ${glsl.texture2D}(A, uv);\n        float value;\n        if (depth == 0) {\n          value = values.r;\n        } else if (depth == 1) {\n          value = values.g;\n        } else if (depth == 2) {\n          value = values.b;\n        } else if (depth == 3) {\n          value = values.a;\n        }\n\n        setOutput(floor(value * 255.0 + 0.5));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels_utils/from_pixels_packed_gpu.js\nvar FromPixelsPackedProgram = class {\n  constructor(outputShape) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = false;\n    this.packedOutput = true;\n    const glsl = getGlslDifferences();\n    const [height, width] = outputShape;\n    this.outputShape = outputShape;\n    this.userCode = `\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n\n        vec4 result = vec4(0.);\n\n        for(int row=0; row<=1; row++) {\n          for(int col=0; col<=1; col++) {\n            texC = coords[1] + row;\n            depth = coords[2] + col;\n\n            vec2 uv = (vec2(texC, texR) + halfCR) /\n                       vec2(${width}.0, ${height}.0);\n            vec4 values = ${glsl.texture2D}(A, uv);\n            float value;\n            if (depth == 0) {\n              value = values.r;\n            } else if (depth == 1) {\n              value = values.g;\n            } else if (depth == 2) {\n              value = values.b;\n            } else if (depth == 3) {\n              value = values.a;\n            }\n\n            result[row * 2 + col] = floor(value * 255.0 + 0.5);\n          }\n        }\n\n        ${glsl.output} = result;\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels.js\nvar fromPixelsConfig = {\n  kernelName: FromPixels,\n  backendName: \"webgl\",\n  kernelFunc: fromPixels2\n};\nvar fromPixels2DContext2;\nvar willReadFrequently = env().getBool(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\");\nfunction fromPixels2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  let { pixels } = inputs;\n  const { numChannels } = attrs;\n  const isVideo = typeof HTMLVideoElement !== \"undefined\" && pixels instanceof HTMLVideoElement;\n  const isImage = typeof HTMLImageElement !== \"undefined\" && pixels instanceof HTMLImageElement;\n  const [width, height] = isVideo ? [\n    pixels.videoWidth,\n    pixels.videoHeight\n  ] : [pixels.width, pixels.height];\n  const texShape = [height, width];\n  const outShape = [height, width, numChannels];\n  if (isImage || isVideo) {\n    const newWillReadFrequently = env().getBool(\"CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU\");\n    if (fromPixels2DContext2 == null || newWillReadFrequently !== willReadFrequently) {\n      willReadFrequently = newWillReadFrequently;\n      fromPixels2DContext2 = document.createElement(\"canvas\").getContext(\"2d\", { willReadFrequently });\n    }\n    fromPixels2DContext2.canvas.width = width;\n    fromPixels2DContext2.canvas.height = height;\n    fromPixels2DContext2.drawImage(pixels, 0, 0, width, height);\n    pixels = fromPixels2DContext2.canvas;\n  }\n  const tempPixelHandle = backend2.makeTensorInfo(texShape, \"int32\");\n  backend2.texData.get(tempPixelHandle.dataId).usage = TextureUsage.PIXELS;\n  backend2.gpgpu.uploadPixelDataToTexture(backend2.getTexture(tempPixelHandle.dataId), pixels);\n  const program = env().getBool(\"WEBGL_PACK\") ? new FromPixelsPackedProgram(outShape) : new FromPixelsProgram(outShape);\n  const res = backend2.runWebGLProgram(program, [tempPixelHandle], \"int32\");\n  backend2.disposeData(tempPixelHandle.dataId);\n  return res;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FusedConv2D.js\nfunction fusedConv2d(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);\n  let out;\n  const intermediates = [];\n  const hasBias = bias != null;\n  const hasPreluActivationWeights = preluActivationWeights != null;\n  const hasLeakyreluAlpha = activation2 === \"leakyrelu\";\n  const prepareInputs = () => {\n    const inputs2 = [x, filter];\n    const alignInputWithDataFormat = (input2, dataFormat2) => {\n      if (dataFormat2 === \"NCHW\" && input2.shape.length === 1 && input2.shape[0] !== 1) {\n        const alignedInput = reshape4({\n          inputs: { x: input2 },\n          backend: backend2,\n          attrs: { shape: [input2.shape[0], 1, 1] }\n        });\n        intermediates.push(alignedInput);\n        return alignedInput;\n      }\n      return input2;\n    };\n    if (hasBias) {\n      inputs2.push(alignInputWithDataFormat(bias, dataFormat));\n    }\n    if (hasPreluActivationWeights) {\n      inputs2.push(alignInputWithDataFormat(preluActivationWeights, dataFormat));\n    }\n    if (hasLeakyreluAlpha) {\n      const $leakyreluAlpha = backend2.makeTensorInfo([], \"float32\", util_exports.createScalarValue(leakyreluAlpha, \"float32\"));\n      inputs2.push($leakyreluAlpha);\n      intermediates.push($leakyreluAlpha);\n    }\n    return inputs2;\n  };\n  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === \"SAME\" || convInfo.padInfo.type === \"VALID\")) {\n    out = conv2dByMatMul({\n      x,\n      filter,\n      convInfo,\n      backend: backend2,\n      bias,\n      activation: activation2,\n      preluActivationWeights,\n      leakyreluAlpha\n    });\n  } else if (convInfo.strideWidth <= 2 && $dataFormat === \"channelsLast\" && env().getBool(\"WEBGL_EXP_CONV\")) {\n    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;\n    const program = new Conv2DPackedProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n    const customValues = [\n      [convInfo.padInfo.top, convInfo.padInfo.left],\n      [convInfo.strideHeight, convInfo.strideWidth],\n      [convInfo.dilationHeight, convInfo.dilationWidth],\n      [convInfo.inHeight, convInfo.inWidth]\n    ];\n    const inputs2 = prepareInputs();\n    out = backend2.runWebGLProgram(program, inputs2, \"float32\", customValues);\n  } else if (env().getBool(\"WEBGL_CONV_IM2COL\")) {\n    out = conv2dWithIm2Row({\n      x,\n      filter,\n      convInfo,\n      backend: backend2,\n      bias,\n      activation: activation2,\n      preluActivationWeights,\n      leakyreluAlpha\n    });\n  } else {\n    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, false) : null;\n    const program = new Conv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n    const inputs2 = prepareInputs();\n    out = backend2.runWebGLProgram(program, inputs2, \"float32\");\n  }\n  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });\n  intermediates.push(out);\n  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return outReshaped;\n}\nvar fusedConv2DConfig2 = {\n  kernelName: FusedConv2D,\n  backendName: \"webgl\",\n  kernelFunc: fusedConv2d\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FusedDepthwiseConv2D.js\nfunction fusedDepthwiseConv2D2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  const intermediates = [];\n  let $dilations = dilations;\n  if ($dilations == null) {\n    $dilations = [1, 1];\n  }\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);\n  const shouldPackDepthwiseConv = env().getBool(\"WEBGL_PACK_DEPTHWISECONV\") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1;\n  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, shouldPackDepthwiseConv) : null;\n  const programInputs = [x, filter];\n  const hasBias = bias != null;\n  const hasPreluActivationWeights = preluActivationWeights != null;\n  const hasLeakyreluAlpha = activation2 === \"leakyrelu\";\n  if (hasBias) {\n    programInputs.push(bias);\n  }\n  if (hasPreluActivationWeights) {\n    programInputs.push(preluActivationWeights);\n  }\n  if (hasLeakyreluAlpha) {\n    const $leakyreluAlpha = backend2.makeTensorInfo([], \"float32\", util_exports.createScalarValue(leakyreluAlpha, \"float32\"));\n    programInputs.push($leakyreluAlpha);\n    intermediates.push($leakyreluAlpha);\n  }\n  let program;\n  if (shouldPackDepthwiseConv) {\n    program = new DepthwiseConvPacked2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n  } else {\n    program = new DepthwiseConv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);\n  }\n  const customValues = [\n    [convInfo.padInfo.top, convInfo.padInfo.left],\n    [convInfo.strideHeight, convInfo.strideWidth],\n    [convInfo.dilationHeight, convInfo.dilationWidth],\n    [convInfo.inHeight, convInfo.inWidth]\n  ];\n  const result = backend2.runWebGLProgram(program, programInputs, \"float32\", customValues);\n  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar fusedDepthwiseConv2DConfig2 = {\n  kernelName: FusedDepthwiseConv2D,\n  backendName: \"webgl\",\n  kernelFunc: fusedDepthwiseConv2D2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/gather_nd_gpu.js\nvar GatherNDProgram = class {\n  constructor(sliceDim, strides, shape, paramsShape) {\n    this.sliceDim = sliceDim;\n    this.strides = strides;\n    this.paramsShape = paramsShape;\n    this.variableNames = [\"x\", \"indices\"];\n    this.outputShape = shape;\n    const dtype = getCoordsDataType(shape.length);\n    let mainLoop = `\n    int index;`;\n    for (let j = 0; j < this.sliceDim; j++) {\n      mainLoop += `\n          index = round(getIndices(coords[0], ${j}));\n          out_of_bounds = out_of_bounds || index < 0;\n          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[j]};\n          flattenIndex += index * ${this.strides[j]};`;\n    }\n    this.userCode = `\n         void main() {\n          ${dtype} coords = getOutputCoords();\n          int flattenIndex = 0;\n          bool out_of_bounds = false;\n\n          ${mainLoop}\n\n          setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));\n        }\n      `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GatherNd.js\nfunction gatherNd2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { params, indices } = inputs;\n  const indicesShape = indices.shape;\n  const sliceRank = indicesShape[indicesShape.length - 1];\n  const paramsSize = util_exports.sizeFromShape(params.shape);\n  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);\n  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numSlices, sliceRank] } });\n  const flattenX = reshape4({\n    inputs: { x: params },\n    backend: backend2,\n    attrs: { shape: [util_exports.sizeFromShape(params.shape) / sliceSize, sliceSize] }\n  });\n  if (backend2.shouldExecuteOnCPU([params, indices]) || params.dtype === \"string\") {\n    const indicesData = backend2.readSync(indices.dataId);\n    const paramsBuf = backend2.bufferSync(params);\n    const outValue = gatherNdImplCPU(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);\n    return backend2.makeTensorInfo(resultShape, params.dtype, outValue.values);\n  }\n  const program = new GatherNDProgram(sliceRank, strides, [numSlices, sliceSize], params.shape);\n  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices], flattenX.dtype);\n  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: resultShape } });\n  backend2.disposeIntermediateTensorInfo(flattenIndices);\n  backend2.disposeIntermediateTensorInfo(flattenX);\n  backend2.disposeIntermediateTensorInfo(res);\n  return reshaped;\n}\nvar gatherNdConfig2 = {\n  kernelName: GatherNd,\n  backendName: \"webgl\",\n  kernelFunc: gatherNd2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/gather_gpu.js\nvar GatherProgram = class {\n  constructor(aShape, outputShape) {\n    this.variableNames = [\"A\", \"indices\"];\n    this.outputShape = outputShape;\n    this.rank = outputShape.length;\n    const dtype = getCoordsDataType(this.rank);\n    const sourceCoords = getSourceCoords2(aShape, 2);\n    this.userCode = `\n      void main() {\n        ${dtype} resRC = getOutputCoords();\n        int index = int(getIndices(resRC.x, resRC.z));\n        float inBounds = (index >= 0) && (index < ${aShape[2]}) ? 1.0 : 0.0;\n        setOutput(inBounds * getA(${sourceCoords}));\n      }\n    `;\n  }\n};\nfunction getSourceCoords2(aShape, axis) {\n  const currentCoords = [\"resRC.x\", \"resRC.y\", \"resRC.z\", \"resRC.w\"];\n  const sourceCoords = [];\n  for (let i = 0; i < aShape.length; i++) {\n    if (i === 2) {\n      sourceCoords.push(\"index\");\n    } else {\n      sourceCoords.push(`${currentCoords[i]}`);\n    }\n  }\n  return sourceCoords.join();\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GatherV2.js\nfunction gatherV22(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, indices } = inputs;\n  const { axis, batchDims } = attrs;\n  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];\n  if (env().get(\"DEBUG\")) {\n    const indicesVals = backend2.readSync(indices.dataId);\n    const axisDim = x.shape[parsedAxis];\n    for (let i = 0; i < indicesVals.length; ++i) {\n      const index = indicesVals[i];\n      util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);\n    }\n  }\n  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);\n  const indicesSize = util_exports.sizeFromShape(indices.shape);\n  const toDispose = [];\n  const flattenX = reshape4({\n    inputs: { x },\n    backend: backend2,\n    attrs: {\n      shape: [\n        shapeInfo.batchSize,\n        shapeInfo.outerSize,\n        shapeInfo.dimSize,\n        shapeInfo.sliceSize\n      ]\n    }\n  });\n  const flattenIndex = reshape4({\n    inputs: { x: indices },\n    backend: backend2,\n    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }\n  });\n  toDispose.push(flattenX);\n  toDispose.push(flattenIndex);\n  const flattenOutputShape = [\n    shapeInfo.batchSize,\n    shapeInfo.outerSize,\n    indicesSize / shapeInfo.batchSize,\n    shapeInfo.sliceSize\n  ];\n  if (backend2.shouldExecuteOnCPU([x, indices]) || x.dtype === \"string\") {\n    const indicesBuf = backend2.bufferSync(flattenIndex);\n    const xBuf = backend2.bufferSync(flattenX);\n    const outBuf = gatherV2ImplCPU(xBuf, indicesBuf, flattenOutputShape);\n    toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n    return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);\n  }\n  const program = new GatherProgram(flattenX.shape, flattenOutputShape);\n  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndex], flattenX.dtype);\n  toDispose.push(res);\n  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: shapeInfo.outputShape } });\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return reshaped;\n}\nvar gatherV2Config2 = {\n  kernelName: GatherV2,\n  backendName: \"webgl\",\n  kernelFunc: gatherV22\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Greater.js\nvar GREATER = `return float(a > b);`;\nvar GREATER_PACKED = `\n  return vec4(greaterThan(a, b));\n`;\nvar greater4 = binaryKernelFunc2({\n  opSnippet: GREATER,\n  packedOpSnippet: GREATER_PACKED,\n  cpuKernelImpl: greaterImplCPU,\n  dtype: \"bool\"\n});\nvar greaterConfig2 = {\n  kernelName: Greater,\n  backendName: \"webgl\",\n  kernelFunc: greater4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GreaterEqual.js\nvar GREATER_EQUAL = `return float(a >= b);`;\nvar GREATER_EQUAL_PACKED = `\n  return vec4(greaterThanEqual(a, b));\n`;\nvar greaterEqual3 = binaryKernelFunc2({\n  opSnippet: GREATER_EQUAL,\n  packedOpSnippet: GREATER_EQUAL_PACKED,\n  dtype: \"bool\",\n  cpuKernelImpl: greaterEqualImplCPU\n});\nvar greaterEqualConfig2 = {\n  kernelName: GreaterEqual,\n  backendName: \"webgl\",\n  kernelFunc: greaterEqual3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IFFT.js\nfunction ifft3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  return fftImpl2(input2, true, backend2);\n}\nvar ifftConfig2 = {\n  kernelName: IFFT,\n  backendName: \"webgl\",\n  kernelFunc: ifft3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsFinite.js\nvar IS_FINITE = `return float(!isnan(x) && !isinf(x));`;\nvar isFinite4 = unaryKernelFunc2({ opSnippet: IS_FINITE, dtype: \"bool\" });\nvar isFiniteConfig2 = {\n  kernelName: IsFinite,\n  backendName: \"webgl\",\n  kernelFunc: isFinite4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsInf.js\nvar IS_INF = `return float(isinf(x));`;\nvar isInf3 = unaryKernelFunc2({ opSnippet: IS_INF, dtype: \"bool\" });\nvar isInfConfig2 = {\n  kernelName: IsInf,\n  backendName: \"webgl\",\n  kernelFunc: isInf3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsNaN.js\nvar IS_NAN = `return float(isnan(x));`;\nvar isNaN4 = unaryKernelFunc2({ opSnippet: IS_NAN, dtype: \"bool\" });\nvar isNaNConfig2 = {\n  kernelName: IsNan,\n  backendName: \"webgl\",\n  kernelFunc: isNaN4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Less.js\nvar LESS = `return float(a < b);`;\nvar LESS_PACKED = `\n  return vec4(lessThan(a, b));\n`;\nvar less4 = binaryKernelFunc2({\n  opSnippet: LESS,\n  packedOpSnippet: LESS_PACKED,\n  cpuKernelImpl: lessImplCPU,\n  dtype: \"bool\"\n});\nvar lessConfig2 = {\n  kernelName: Less,\n  backendName: \"webgl\",\n  kernelFunc: less4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LessEqual.js\nvar LESS_EQUAL = `return float(a <= b);`;\nvar LESS_EQUAL_PACKED = `\n  return vec4(lessThanEqual(a, b));\n`;\nvar lessEqual3 = binaryKernelFunc2({\n  opSnippet: LESS_EQUAL,\n  packedOpSnippet: LESS_EQUAL_PACKED,\n  cpuKernelImpl: lessEqualImplCPU,\n  dtype: \"bool\"\n});\nvar lessEqualConfig2 = {\n  kernelName: LessEqual,\n  backendName: \"webgl\",\n  kernelFunc: lessEqual3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LinSpace.js\nfunction linSpace2(args) {\n  const { backend: backend2, attrs } = args;\n  const { start, stop, num } = attrs;\n  const outVals = linSpaceImplCPU(start, stop, num);\n  return backend2.makeTensorInfo([outVals.length], \"float32\", outVals);\n}\nvar linSpaceConfig2 = {\n  kernelName: LinSpace,\n  backendName: \"webgl\",\n  kernelFunc: linSpace2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Log.js\nvar LOG = CHECK_NAN_SNIPPET_UNARY + `\n  return x < 0.0 ? 0./0. : log(x);\n`;\nvar LOG_PACKED = `\n  vec4 result = log(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);\n  result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);\n  result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);\n  result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);\n  return result;\n`;\nvar log4 = unaryKernelFunc2({ opSnippet: LOG, packedOpSnippet: LOG_PACKED, cpuKernelImpl: logImplCPU });\nvar logConfig2 = {\n  kernelName: Log,\n  backendName: \"webgl\",\n  kernelFunc: log4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Log1p.js\nvar LOG1P = CHECK_NAN_SNIPPET_UNARY + `\n  return log(1.0 + x);\n`;\nvar log1p3 = unaryKernelFunc2({ opSnippet: LOG1P });\nvar log1pConfig2 = {\n  kernelName: Log1p,\n  backendName: \"webgl\",\n  kernelFunc: log1p3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalAnd.js\nvar LOGICAL_AND = `return float(a >= 1.0 && b >= 1.0);`;\nvar LOGICAL_AND_PACKED = `\n  return vec4(\n    vec4(greaterThanEqual(a, vec4(1.0))) *\n    vec4(greaterThanEqual(b, vec4(1.0))));\n`;\nvar logicalAnd3 = binaryKernelFunc2({\n  opSnippet: LOGICAL_AND,\n  packedOpSnippet: LOGICAL_AND_PACKED,\n  dtype: \"bool\"\n});\nvar logicalAndConfig2 = {\n  kernelName: LogicalAnd,\n  backendName: \"webgl\",\n  kernelFunc: logicalAnd3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalNot.js\nvar LOGICAL_NOT = `return float(!(x >= 1.0));`;\nvar logicalNot3 = unaryKernelFunc2({ opSnippet: LOGICAL_NOT });\nvar logicalNotConfig2 = {\n  kernelName: LogicalNot,\n  backendName: \"webgl\",\n  kernelFunc: logicalNot3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalOr.js\nvar LOGICAL_OR = `return float(a >= 1.0 || b >= 1.0);`;\nvar LOGICAL_OR_PACKED = `\n  return min(\n    vec4(greaterThanEqual(a, vec4(1.0))) +\n    vec4(greaterThanEqual(b, vec4(1.0))),\n    vec4(1.0));\n`;\nvar logicalOr3 = binaryKernelFunc2({ opSnippet: LOGICAL_OR, packedOpSnippet: LOGICAL_OR_PACKED, dtype: \"bool\" });\nvar logicalOrConfig2 = {\n  kernelName: LogicalOr,\n  backendName: \"webgl\",\n  kernelFunc: logicalOr3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_gpu.js\nvar LRNProgram = class {\n  constructor(xShape, radius, bias, alpha, beta) {\n    this.variableNames = [\"x\"];\n    this.outputShape = [];\n    const rad = radius;\n    const maxD = xShape[3] - 1;\n    this.outputShape = xShape;\n    let powOperator;\n    const basis = `float(${bias}) + float(${alpha}) * sum`;\n    if (beta === 0.5) {\n      powOperator = `inversesqrt(${basis})`;\n    } else if (beta === 1) {\n      powOperator = `1.0/(${basis})`;\n    } else {\n      powOperator = `exp(log(${basis}) * float(-${beta}));`;\n    }\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n        int d = coords[3];\n        float x = getX(b, r, c, d);\n        float sum = 0.0;\n        for (int j = -${rad}; j <= ${rad}; j++) {\n          int idx = d + j;\n          if (idx >= 0 && idx <=  ${maxD}) {\n            float z = getX(b, r, c, idx);\n            sum += z * z;\n          }\n        }\n        float val = x * ${powOperator};\n        setOutput(val);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_packed_gpu.js\nvar LRNPackedProgram = class {\n  constructor(xShape, radius, bias, alpha, beta) {\n    this.variableNames = [\"x\"];\n    this.outputShape = [];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    const rad = radius;\n    const maxD = xShape[3] - 1;\n    this.outputShape = xShape;\n    let powOperator;\n    const basis = `float(${bias}) + float(${alpha}) * sum`;\n    if (beta === 0.5) {\n      powOperator = `inversesqrt(${basis})`;\n    } else if (beta === 1) {\n      powOperator = `1.0/(${basis})`;\n    } else {\n      powOperator = `exp(log(${basis}) * float(-${beta}));`;\n    }\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords.x;\n        int r = coords.y;\n        int c = coords.z;\n        int d = coords.w;\n\n        bool hasNextCol = d < ${this.outputShape[3]};\n        bool hasNextRow = c < ${this.outputShape[2]};\n\n        vec4 sum = vec4(0.);\n        vec4 xFragAtOutputCoords = getX(b, r, c, d);\n\n        vec4 xAtOutputCoords = vec4(\n          getChannel(xFragAtOutputCoords, vec2(c, d)),\n          hasNextCol ?\n            getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,\n          hasNextRow ?\n            getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0\n        );\n\n        int firstChannel = d - ${rad};\n        vec2 cache = vec2(0.);\n        if(firstChannel >= 0){\n          vec4 firstChannelFrag = getX(b, r, c, firstChannel);\n          cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));\n            if(hasNextRow){\n              cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));\n            }\n        }\n\n        ivec2 depth = ivec2(d, d + 1);\n        for (int j = - ${rad}; j <= ${rad}; j++) {\n          ivec2 idx = depth + j;\n          bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));\n          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${maxD}));\n\n          bool depthInRange = aboveLowerBound.x && belowUpperBound.x;\n          bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;\n\n          if(depthInRange || depthPlusOneInRange){\n            vec4 z = vec4(0.);\n            vec4 xFragAtCurrentDepth;\n            z.xz = cache.xy;\n            if(depthPlusOneInRange && hasNextCol){\n              xFragAtCurrentDepth = idx.y != d ?\n                getX(b, r, c, idx.y) : xFragAtOutputCoords;\n              z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));\n              if(hasNextRow){\n                z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));\n              }\n            }\n            cache.xy = z.yw;\n            sum += z * z;\n          }\n        }\n        vec4 result = xAtOutputCoords * ${powOperator};\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LRN.js\nvar lrn = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { depthRadius, bias, alpha, beta } = attrs;\n  const program = env().getBool(\"WEBGL_PACK_NORMALIZATION\") ? new LRNPackedProgram(x.shape, depthRadius, bias, alpha, beta) : new LRNProgram(x.shape, depthRadius, bias, alpha, beta);\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n};\nvar LRNConfig2 = {\n  kernelName: LRN,\n  backendName: \"webgl\",\n  kernelFunc: lrn\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_grad_gpu.js\nvar LRNGradProgram = class {\n  constructor(inputShape, depthRadius, bias, alpha, beta) {\n    this.variableNames = [\"inputImage\", \"outputImage\", \"dy\"];\n    this.outputShape = [];\n    this.outputShape = inputShape;\n    this.depth = inputShape[3];\n    this.depthRadius = depthRadius;\n    this.bias = bias;\n    this.alpha = alpha;\n    this.beta = beta;\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n\n        float result = 0.0;\n        for (int d = 0; d < ${this.depth}; ++d) {\n          int depthBegin = int(max(0.0, float(d - ${depthRadius})));\n          int depthEnd = int(min(float(${this.depth}),\n              float(d + ${depthRadius} + 1)));\n\n          const int MIN_DEPTH_BEGIN = 0;\n          const int MAX_DEPTH_END = ${this.depth};\n\n          float norm = 0.0;\n          for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd) {\n              norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);\n            }\n            else {\n              break;\n            }\n          }\n\n          norm = float(${alpha}) * norm + float(${bias});\n\n          for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd){\n              float dyi = -2.0 * float(${alpha})\n                * float(${beta})\n                * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)\n                / norm;\n              if (k == d) {\n                dyi += pow(norm, -1.0 * ${beta});\n              }\n              if (k == coords[3]) {\n                dyi *= getDy(b, r, c, d);\n                result += dyi;\n              }\n            }\n            else {\n              break;\n            }\n          }\n      }\n      setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LRNGrad.js\nvar lrnGrad = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, y, dy } = inputs;\n  const { depthRadius, bias, alpha, beta } = attrs;\n  const program = new LRNGradProgram(x.shape, depthRadius, bias, alpha, beta);\n  return backend2.runWebGLProgram(program, [x, y, dy], x.dtype);\n};\nvar LRNGradConfig2 = {\n  kernelName: LRNGrad,\n  backendName: \"webgl\",\n  kernelFunc: lrnGrad\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Max_impl.js\nfunction maxImpl2(x, reduceShape, outShape, backend2) {\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const batchSize = xSize / inSize;\n  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });\n  const reduced = reduce(reshapedInput, x.dtype, \"max\", backend2);\n  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(reshapedInput);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  return reshapedOutput;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Max.js\nfunction max4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { reductionIndices, keepDims } = attrs;\n  const xRank = x.shape.length;\n  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  const maxInputIsTransposed = permutedAxes != null;\n  const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);\n  let maxInput = x;\n  if (maxInputIsTransposed) {\n    if (shouldExecuteOnCPU) {\n      const xTexData = backend2.texData.get(maxInput.dataId);\n      const values = xTexData.values;\n      const newShape = new Array(xRank);\n      for (let i = 0; i < newShape.length; i++) {\n        newShape[i] = x.shape[permutedAxes[i]];\n      }\n      const maxInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);\n      maxInput = backend2.makeTensorInfo(newShape, x.dtype);\n      const maxInputData = backend2.texData.get(maxInput.dataId);\n      maxInputData.values = maxInputValues;\n    } else {\n      maxInput = transposeImpl2(x, permutedAxes, backend2);\n    }\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"max\", axes, xRank);\n  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(maxInput.shape, axes);\n  let outShape = maxOutShape;\n  if (keepDims) {\n    outShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);\n  }\n  let out;\n  if (shouldExecuteOnCPU) {\n    const xTexData = backend2.texData.get(maxInput.dataId);\n    const values = xTexData.values;\n    const outValues = maxImplCPU(values, util_exports.sizeFromShape(reduceShape), outShape, x.dtype);\n    out = backend2.makeTensorInfo(outShape, x.dtype);\n    const outData = backend2.texData.get(out.dataId);\n    outData.values = outValues;\n  } else {\n    out = maxImpl2(maxInput, reduceShape, outShape, backend2);\n  }\n  if (maxInputIsTransposed) {\n    backend2.disposeIntermediateTensorInfo(maxInput);\n  }\n  return out;\n}\nvar maxConfig2 = {\n  kernelName: Max,\n  backendName: \"webgl\",\n  kernelFunc: max4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Maximum.js\nvar MAXIMUM = CHECK_NAN_SNIPPET2 + `\n  return max(a, b);\n`;\nvar MAXIMUM_PACKED = `\n  vec4 result = vec4(max(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  ` + CHECK_NAN_SNIPPET_PACKED + `\n  return result;\n`;\nvar maximum4 = binaryKernelFunc2({\n  opSnippet: MAXIMUM,\n  packedOpSnippet: MAXIMUM_PACKED,\n  cpuKernelImpl: maximumImplCPU\n});\nvar maximumConfig2 = {\n  kernelName: Maximum,\n  backendName: \"webgl\",\n  kernelFunc: maximum4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool.js\nfunction maxPool3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  assertNotComplex2(x, \"maxPool\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = 1;\n  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {\n    return identity3({ inputs: { x }, backend: backend2 });\n  }\n  const maxPoolProgram = new Pool2DProgram(convInfo, \"max\", false);\n  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);\n}\nvar maxPoolConfig2 = {\n  kernelName: MaxPool,\n  backendName: \"webgl\",\n  kernelFunc: maxPool3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool3D.js\nfunction maxPool3d2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { filterSize, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;\n  const dilations = [1, 1, 1];\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);\n  const maxPoolProgram = new Pool3DProgram(convInfo, \"max\", false);\n  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);\n}\nvar maxPool3DConfig2 = {\n  kernelName: MaxPool3D,\n  backendName: \"webgl\",\n  kernelFunc: maxPool3d2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/max_pool_backprop_gpu.js\nvar MaxPool2DBackpropProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\", \"maxPos\"];\n    this.outputShape = convInfo.inShape;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationHeight = convInfo.dilationHeight;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n    const lastIndex = effectiveFilterHeight * effectiveFilterWidth - 1;\n    this.userCode = `\n      const ivec2 pads = ivec2(${padTop}, ${padLeft});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${effectiveFilterHeight};\n          wR += ${dilationHeight}) {\n          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${effectiveFilterWidth}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n            int maxPosValue = ${lastIndex} - int(getMaxPos(b, idyR, idyC, d));\n\n            // Get the current value, check it against the value from the\n            // position matrix.\n            int curPosValue = wR * ${effectiveFilterWidth} + wC;\n            float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n            dotProd += dyValue * mask;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\nvar MaxPool3DBackpropProgram = class {\n  constructor(convInfo) {\n    this.variableNames = [\"dy\", \"maxPos\"];\n    this.outputShape = convInfo.inShape;\n    const strideDepth = convInfo.strideDepth;\n    const strideHeight = convInfo.strideHeight;\n    const strideWidth = convInfo.strideWidth;\n    const dilationDepth = convInfo.dilationDepth;\n    const dilationHeight = convInfo.dilationHeight;\n    const dilationWidth = convInfo.dilationWidth;\n    const effectiveFilterDepth = convInfo.effectiveFilterDepth;\n    const effectiveFilterHeight = convInfo.effectiveFilterHeight;\n    const effectiveFilterWidth = convInfo.effectiveFilterWidth;\n    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;\n    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;\n    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;\n    const lastIndex = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1;\n    this.userCode = `\n      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${effectiveFilterDepth};\n           wD += ${dilationDepth}) {\n          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;\n\n          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${effectiveFilterHeight};\n              wR += ${dilationHeight}) {\n            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;\n\n            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${effectiveFilterWidth};\n                wC += ${dilationWidth}) {\n              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;\n\n              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n              int maxPosValue = ${lastIndex} -\n                  int(getMaxPos(batch, idyD, idyR, idyC, ch));\n\n              // Get the current value, check it against the value from the\n              // position matrix.\n              int curPosValue =\n                  wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +\n                  wR * ${effectiveFilterWidth} + wC;\n              float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n              dotProd += dyValue * mask;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool3DGrad.js\nfunction maxPool3DGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2 } = inputs;\n  const x = input2;\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const dilations = [1, 1, 1];\n  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);\n  const maxPool3dPositionsProgram = new Pool3DProgram(convInfo, \"max\", true);\n  const maxPool3dPositions2 = backend2.runWebGLProgram(maxPool3dPositionsProgram, [x], x.dtype);\n  const maxPoolBackpropProgram = new MaxPool3DBackpropProgram(convInfo);\n  const result = backend2.runWebGLProgram(maxPoolBackpropProgram, [dy, maxPool3dPositions2], x.dtype);\n  backend2.disposeIntermediateTensorInfo(maxPool3dPositions2);\n  return result;\n}\nvar maxPool3DGradConfig3 = {\n  kernelName: MaxPool3DGrad,\n  backendName: \"webgl\",\n  kernelFunc: maxPool3DGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolGrad.js\nfunction maxPoolGrad3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { dy, input: input2, output } = inputs;\n  const x = input2;\n  assertNotComplex2([input2, output], \"maxPoolGrad\");\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const getPositions = true;\n  const maxPoolPositionsProgram = new Pool2DProgram(convInfo, \"max\", getPositions);\n  const maxPoolPositions2 = backend2.runWebGLProgram(maxPoolPositionsProgram, [x], x.dtype);\n  const maxPoolBackPropProgram = new MaxPool2DBackpropProgram(convInfo);\n  const result = backend2.runWebGLProgram(maxPoolBackPropProgram, [dy, maxPoolPositions2], x.dtype);\n  backend2.disposeIntermediateTensorInfo(maxPoolPositions2);\n  return result;\n}\nvar maxPoolGradConfig3 = {\n  kernelName: MaxPoolGrad,\n  backendName: \"webgl\",\n  kernelFunc: maxPoolGrad3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolWithArgmax_impl.js\nfunction maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, backend2) {\n  let program = new Pool2DProgram(convInfo, \"max\", false);\n  const poolOutput = backend2.runWebGLProgram(program, [x], \"float32\");\n  program = new Pool2DProgram(convInfo, \"max\", true, true, includeBatchInIndex);\n  const indexOutput = backend2.runWebGLProgram(program, [x], \"float32\");\n  return [poolOutput, indexOutput];\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolWithArgmax.js\nvar maxPoolWithArgmaxConfig2 = {\n  kernelName: MaxPoolWithArgmax,\n  backendName: \"webgl\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { x } = inputs;\n    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;\n    const webglBackend = backend2;\n    util_exports.assert(x.shape.length === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x.shape.length}.`);\n    const dilations = [1, 1];\n    util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);\n    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3);\n    const [result, indexes] = maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, webglBackend);\n    return [result, indexes];\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mean_impl.js\nfunction meanImpl(x, reduceShape, outShape, backend2) {\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const batchSize = xSize / inSize;\n  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });\n  const reduced = reduce(reshapedInput, \"float32\", \"mean\", backend2);\n  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(reshapedInput);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  return reshapedOutput;\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mean.js\nvar meanConfig2 = {\n  kernelName: Mean,\n  backendName: \"webgl\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { x } = inputs;\n    const { keepDims, axis } = attrs;\n    const webglBackend = backend2;\n    const xRank = x.shape.length;\n    const origAxes = util_exports.parseAxisParam(axis, x.shape);\n    let axes = origAxes;\n    const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n    const meanInputIsTransposed = permutedAxes != null;\n    const shouldExecuteOnCPU = webglBackend.shouldExecuteOnCPU([x]);\n    const intermediates = [];\n    let meanInput = x;\n    if (meanInputIsTransposed) {\n      if (shouldExecuteOnCPU) {\n        const xTexData = webglBackend.texData.get(meanInput.dataId);\n        const values = xTexData.values;\n        const newShape = new Array(xRank);\n        for (let i = 0; i < newShape.length; i++) {\n          newShape[i] = x.shape[permutedAxes[i]];\n        }\n        const meanInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);\n        meanInput = webglBackend.makeTensorInfo(newShape, x.dtype);\n        const meanInputData = webglBackend.texData.get(meanInput.dataId);\n        meanInputData.values = meanInputValues;\n      } else {\n        meanInput = transposeImpl2(x, permutedAxes, webglBackend);\n      }\n      intermediates.push(meanInput);\n      axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n    }\n    backend_util_exports.assertAxesAreInnerMostDims(\"sum\", axes, xRank);\n    const [meanOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(meanInput.shape, axes);\n    let outShape = meanOutShape;\n    if (keepDims) {\n      outShape = backend_util_exports.expandShapeToKeepDim(meanOutShape, origAxes);\n    }\n    const out = meanImpl(meanInput, reduceShape, outShape, webglBackend);\n    for (const i of intermediates) {\n      webglBackend.disposeIntermediateTensorInfo(i);\n    }\n    return out;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Min.js\nfunction min4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  const xRank = x.shape.length;\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let permutedX = x;\n  if (permutedAxes != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"min\", axes, xRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);\n  const inSize = util_exports.sizeFromShape(reduceShape);\n  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });\n  const reduced = reduce(a2D, a2D.dtype, \"min\", backend2);\n  let res;\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });\n  } else {\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });\n  }\n  backend2.disposeIntermediateTensorInfo(a2D);\n  backend2.disposeIntermediateTensorInfo(reduced);\n  if (permutedAxes != null) {\n    backend2.disposeIntermediateTensorInfo(permutedX);\n  }\n  return res;\n}\nvar minConfig2 = {\n  kernelName: Min,\n  backendName: \"webgl\",\n  kernelFunc: min4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Minimum.js\nvar MINIMUM = CHECK_NAN_SNIPPET2 + `\n  return min(a, b);\n`;\nvar MINIMUM_PACKED = `\n  vec4 result = vec4(min(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  ` + CHECK_NAN_SNIPPET_PACKED + `\n  return result;\n`;\nvar minimum4 = binaryKernelFunc2({\n  opSnippet: MINIMUM,\n  packedOpSnippet: MINIMUM_PACKED,\n  cpuKernelImpl: minimumImplCPU\n});\nvar minimumConfig2 = {\n  kernelName: Minimum,\n  backendName: \"webgl\",\n  kernelFunc: minimum4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/mirror_pad_gpu.js\nvar MirrorPadProgram = class {\n  constructor(xShape, paddings, mode) {\n    this.variableNames = [\"x\"];\n    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);\n    const rank = xShape.length;\n    const dtype = getCoordsDataType(rank);\n    const start = paddings.map((p2) => p2[0]).join(\",\");\n    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(\",\");\n    const unpackedCoords = [\"coords[0]\", \"coords[1]\", \"coords[2]\", \"coords[3]\"].slice(0, rank);\n    const offset = mode === \"reflect\" ? 0 : 1;\n    if (rank === 1) {\n      this.userCode = `\n        int start = ${start};\n        int end = ${end};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start) {\n            outC = start * 2 - outC - ${offset};\n          } else if(outC >= end) {\n            outC = (end - 1) * 2 - outC + ${offset};\n          }\n          setOutput(getX(outC - start));\n        }\n      `;\n      return;\n    }\n    this.userCode = `\n      ${dtype} start = ${dtype}(${start});\n      ${dtype} end = ${dtype}(${end});\n\n      void main() {\n        ${dtype} outC = getOutputCoords();\n        for (int i = 0; i < ${rank}; i++) {\n          if (outC[i] < start[i]) {\n            outC[i] = start[i] * 2 - outC[i] - ${offset};\n          } else if(outC[i] >= end[i]) {\n            outC[i] = (end[i] - 1) * 2 - outC[i] + ${offset};\n          }\n        }\n        ${dtype} coords = outC - start;\n        setOutput(getX(${unpackedCoords}));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/mirror_pad_packed_gpu.js\nvar MirrorPadPackedProgram = class {\n  constructor(xShape, paddings, mode) {\n    this.variableNames = [\"x\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);\n    const rank = xShape.length;\n    const dtype = getCoordsDataType(rank);\n    const start = paddings.map((p2) => p2[0]).join(\",\");\n    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(\",\");\n    const coords2 = getChannels(\"rc\", rank);\n    const source = getChannels(\"source\", rank);\n    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;\n    const innerDims = rank === 1 ? \"source\" : `vec2(${source.slice(-2).join()})`;\n    const offset = mode === \"reflect\" ? 0 : 1;\n    let mainLoop = \"\";\n    if (rank === 1) {\n      const padSetup = `\n        ${dtype} source = rc;\n        if (source < start) {\n          source = start * 2 - source - ${offset};\n        } else if (source >= end) {\n          source = (end - 1) * 2 - source + ${offset};\n        }\n        source -= start;\n      `;\n      mainLoop = `\n        ${dtype} rc = outputLoc;\n        ${padSetup}\n        result[0] = getChannel(getX(${source.join()}), ${innerDims});\n        ${coords2[rank - 1]} += 1;\n        if(${cLimit}) {\n          ${padSetup}\n          result[1] = getChannel(getX(${source.join()}), ${innerDims});\n        }\n      `;\n    } else {\n      const padSetup = `\n        ${dtype} source = rc;\n        ${dtype} lt = ${dtype}(lessThan(source, start));\n        ${dtype} gte = ${dtype}(greaterThanEqual(source, end));\n        ${dtype} orig = 1 - (lt + gte);\n        source = orig * source +\n                lt * (start * 2 - source - ${offset}) +\n                gte * ((end - 1) * 2 - source + ${offset});\n        source -= start;\n      `;\n      mainLoop = `\n        ${dtype} rc = outputLoc;\n        ${padSetup}\n        result[0] = getChannel(getX(${source.join()}), ${innerDims});\n        ${coords2[rank - 1]} += 1;\n        if(${cLimit}) {\n          ${padSetup}\n          result[1] = getChannel(getX(${source.join()}), ${innerDims});\n        }\n        rc = outputLoc;\n        ${coords2[rank - 2]} += 1;\n        if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {\n          ${padSetup}\n          result[2] = getChannel(getX(${source.join()}), ${innerDims});\n          ${coords2[rank - 1]} += 1;\n          if(${cLimit}) {\n            ${padSetup}\n            result[3] = getChannel(getX(${source.join()}), ${innerDims});\n          }\n        }\n      `;\n    }\n    this.userCode = `\n      const ${dtype} start = ${dtype}(${start});\n      const ${dtype} end = ${dtype}(${end});\n\n      void main() {\n        ${dtype} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${mainLoop}\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MirrorPad.js\nvar mirrorPadKernelFunc = ({ inputs, backend: backend2, attrs }) => {\n  const { x } = inputs;\n  const { paddings, mode } = attrs;\n  const program = env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") ? new MirrorPadPackedProgram(x.shape, paddings, mode) : new MirrorPadProgram(x.shape, paddings, mode);\n  const output = backend2.runWebGLProgram(program, [x], x.dtype);\n  return output;\n};\nvar mirrorPadConfig2 = {\n  kernelName: MirrorPad,\n  backendName: \"webgl\",\n  kernelFunc: mirrorPadKernelFunc\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mod.js\nvar MOD = `if (b == 0.0) return NAN;\n  return mod(a, b);`;\nvar MOD_PACKED = `\n  vec4 result = mod(a, b);\n  bvec4 isNaN = equal(b, vec4(0.0));\n  ` + CHECK_NAN_SNIPPET_PACKED + `\n  return result;\n`;\nvar mod3 = binaryKernelFunc2({\n  opSnippet: MOD,\n  packedOpSnippet: MOD_PACKED\n});\nvar modConfig2 = {\n  kernelName: Mod,\n  backendName: \"webgl\",\n  kernelFunc: mod3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/multinomial_gpu.js\nvar MultinomialProgram = class {\n  constructor(batchSize, numOutcomes, numSamples) {\n    this.variableNames = [\"probs\"];\n    this.customUniforms = [{ name: \"seed\", type: \"float\" }];\n    this.outputShape = [batchSize, numSamples];\n    this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n\n        float r = random(seed);\n        float cdf = 0.0;\n\n        for (int i = 0; i < ${numOutcomes - 1}; i++) {\n          cdf += getProbs(batch, i);\n\n          if (r < cdf) {\n            setOutput(float(i));\n            return;\n          }\n        }\n\n        // If no other event happened, last event happened.\n        setOutput(float(${numOutcomes - 1}));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RealDiv.js\nvar DIV = `\nif (a == b) {\n  return 1.0;\n};\nreturn a / b;`;\nvar DIV_PACKED = `\n  // vec4 one = vec4(equal(a, b));\n  // return one + (vec4(1.0) - one) * a / b;\n  vec4 result = a / b;\n  if(a.x == b.x) {\n    result.x = 1.;\n  }\n  if(a.y == b.y) {\n    result.y = 1.;\n  }\n  if(a.z == b.z) {\n    result.z = 1.;\n  }\n  if(a.w == b.w) {\n    result.w = 1.;\n  }\n\n  return result;\n`;\nvar realDiv = binaryKernelFunc2({ opSnippet: DIV, packedOpSnippet: DIV_PACKED, checkOutOfBounds: true });\nvar realDivConfig2 = {\n  kernelName: RealDiv,\n  backendName: \"webgl\",\n  kernelFunc: realDiv\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sub.js\nvar SUB = \"return a - b;\";\nvar sub3 = binaryKernelFunc2({\n  opSnippet: SUB,\n  packedOpSnippet: SUB,\n  supportsComplex: true,\n  cpuKernelImpl: subImplCPU\n});\nvar subConfig2 = {\n  kernelName: Sub,\n  backendName: \"webgl\",\n  kernelFunc: sub3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Softmax.js\nfunction softmax4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { logits } = inputs;\n  const { dim } = attrs;\n  const axes = util_exports.parseAxisParam([dim], logits.shape);\n  const maxLogit = max4({\n    inputs: { x: logits },\n    backend: backend2,\n    attrs: { reductionIndices: axes, keepDims: false }\n  });\n  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);\n  const maxLogitsReshaped = reshape4({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });\n  const a = sub3({ inputs: { a: logits, b: maxLogitsReshaped }, backend: backend2 });\n  const b = exp3({ inputs: { x: a }, backend: backend2 });\n  const sumExp = sum4({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });\n  const sumExpReshaped = reshape4({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });\n  const res = realDiv({ inputs: { a: b, b: sumExpReshaped }, backend: backend2 });\n  backend2.disposeIntermediateTensorInfo(maxLogit);\n  backend2.disposeIntermediateTensorInfo(maxLogitsReshaped);\n  backend2.disposeIntermediateTensorInfo(a);\n  backend2.disposeIntermediateTensorInfo(b);\n  backend2.disposeIntermediateTensorInfo(sumExp);\n  backend2.disposeIntermediateTensorInfo(sumExpReshaped);\n  return res;\n}\nvar softmaxConfig2 = {\n  kernelName: Softmax,\n  backendName: \"webgl\",\n  kernelFunc: softmax4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Multinomial.js\nfunction multinomial3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { logits } = inputs;\n  const { numSamples, seed, normalized } = attrs;\n  const probs = normalized ? logits : softmax4({ inputs: { logits }, backend: backend2, attrs: { dim: logits.shape.length - 1 } });\n  const batchSize = probs.shape[0];\n  const numOutcomes = probs.shape[1];\n  const program = new MultinomialProgram(batchSize, numOutcomes, numSamples);\n  const customValues = [[seed]];\n  const res = backend2.runWebGLProgram(program, [probs], \"int32\", customValues);\n  if (!normalized) {\n    backend2.disposeIntermediateTensorInfo(probs);\n  }\n  return res;\n}\nvar multinomialConfig2 = {\n  kernelName: Multinomial,\n  backendName: \"webgl\",\n  kernelFunc: multinomial3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Neg.js\nvar NEG = CHECK_NAN_SNIPPET + `\n  return -x;\n`;\nvar NEG_PACKED = `\n  vec4 result = -x;\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nfunction neg3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (backend2.shouldExecuteOnCPU([x])) {\n    const xData = backend2.texData.get(x.dataId);\n    const [outValues, newShape] = negImplCPU(xData.values, x.shape, x.dtype);\n    return backend2.makeTensorInfo(newShape, x.dtype, outValues);\n  }\n  let program;\n  if (env().getBool(\"WEBGL_PACK_UNARY_OPERATIONS\")) {\n    program = new UnaryOpPackedProgram(x.shape, NEG_PACKED);\n  } else {\n    program = new UnaryOpProgram(x.shape, NEG);\n  }\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\nvar negConfig2 = {\n  kernelName: Neg,\n  backendName: \"webgl\",\n  kernelFunc: neg3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV3.js\nvar nonMaxSuppressionV3Impl3 = kernel_impls_exports.nonMaxSuppressionV3Impl;\nfunction nonMaxSuppressionV32(args) {\n  backend_util_exports.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;\n  const boxesVals = backend2.readSync(boxes.dataId);\n  const scoresVals = backend2.readSync(scores.dataId);\n  const { selectedIndices } = nonMaxSuppressionV3Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);\n  return backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices));\n}\nvar nonMaxSuppressionV3Config2 = {\n  kernelName: NonMaxSuppressionV3,\n  backendName: \"webgl\",\n  kernelFunc: nonMaxSuppressionV32\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV4.js\nvar nonMaxSuppressionV4Impl3 = kernel_impls_exports.nonMaxSuppressionV4Impl;\nfunction nonMaxSuppressionV42(args) {\n  backend_util_exports.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;\n  const boxesVals = backend2.readSync(boxes.dataId);\n  const scoresVals = backend2.readSync(scores.dataId);\n  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);\n  return [\n    backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices)),\n    backend2.makeTensorInfo([], \"int32\", new Int32Array([validOutputs]))\n  ];\n}\nvar nonMaxSuppressionV4Config2 = {\n  kernelName: NonMaxSuppressionV4,\n  backendName: \"webgl\",\n  kernelFunc: nonMaxSuppressionV42\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV5.js\nvar nonMaxSuppressionV5Impl3 = kernel_impls_exports.nonMaxSuppressionV5Impl;\nfunction nonMaxSuppressionV52(args) {\n  backend_util_exports.warn(\"tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead\");\n  const { inputs, backend: backend2, attrs } = args;\n  const { boxes, scores } = inputs;\n  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;\n  const boxesVals = backend2.readSync(boxes.dataId);\n  const scoresVals = backend2.readSync(scores.dataId);\n  const maxOutputSizeVal = maxOutputSize;\n  const iouThresholdVal = iouThreshold;\n  const scoreThresholdVal = scoreThreshold;\n  const softNmsSigmaVal = softNmsSigma;\n  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl3(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);\n  return [\n    backend2.makeTensorInfo([selectedIndices.length], \"int32\", new Int32Array(selectedIndices)),\n    backend2.makeTensorInfo([selectedScores.length], \"float32\", new Float32Array(selectedScores))\n  ];\n}\nvar nonMaxSuppressionV5Config2 = {\n  kernelName: NonMaxSuppressionV5,\n  backendName: \"webgl\",\n  kernelFunc: nonMaxSuppressionV52\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/onehot_gpu.js\nvar OneHotProgram = class {\n  constructor(numIndices, depth, onValue, offValue) {\n    this.variableNames = [\"indices\"];\n    this.outputShape = [numIndices, depth];\n    this.userCode = `\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int index = round(getIndices(coords.x));\n        setOutput(mix(float(${offValue}), float(${onValue}),\n                      float(index == coords.y)));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/OneHot.js\nvar oneHot3 = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { indices } = inputs;\n  const { dtype, depth, onValue, offValue } = attrs;\n  const indicesSize = util_exports.sizeFromShape(indices.shape);\n  const program = new OneHotProgram(indicesSize, depth, onValue, offValue);\n  const reshaped = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [indicesSize] } });\n  const result = backend2.runWebGLProgram(program, [reshaped], dtype);\n  backend2.disposeIntermediateTensorInfo(reshaped);\n  const outShape = [...indices.shape, depth];\n  const out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: outShape } });\n  backend2.disposeIntermediateTensorInfo(result);\n  return out;\n};\nvar oneHotConfig2 = {\n  kernelName: OneHot,\n  backendName: \"webgl\",\n  kernelFunc: oneHot3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ZerosLike.js\nfunction zerosLike3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (x.dtype === \"complex64\") {\n    const realPart = real3({ inputs: { input: x }, backend: backend2 });\n    const r = zerosLike3({ inputs: { x: realPart }, backend: backend2 });\n    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });\n    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });\n    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    backend2.disposeIntermediateTensorInfo(r);\n    backend2.disposeIntermediateTensorInfo(imagPart);\n    backend2.disposeIntermediateTensorInfo(i);\n    return result;\n  } else {\n    return fill3({\n      attrs: {\n        shape: x.shape,\n        dtype: x.dtype,\n        value: x.dtype === \"string\" ? \"\" : 0\n      },\n      backend: backend2\n    });\n  }\n}\nvar zerosLikeConfig2 = {\n  kernelName: ZerosLike,\n  backendName: \"webgl\",\n  kernelFunc: zerosLike3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/OnesLike.js\nfunction onesLike3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x } = inputs;\n  if (x.dtype === \"string\") {\n    throw new Error(\"onesLike is not supported under string dtype\");\n  } else if (x.dtype === \"complex64\") {\n    const realPart = real3({ inputs: { input: x }, backend: backend2 });\n    const r = onesLike3({ inputs: { x: realPart }, backend: backend2 });\n    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });\n    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });\n    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });\n    backend2.disposeIntermediateTensorInfo(realPart);\n    backend2.disposeIntermediateTensorInfo(r);\n    backend2.disposeIntermediateTensorInfo(imagPart);\n    backend2.disposeIntermediateTensorInfo(i);\n    return result;\n  } else {\n    return fill3({ attrs: { shape: x.shape, dtype: x.dtype, value: 1 }, backend: backend2 });\n  }\n}\nvar onesLikeConfig2 = {\n  kernelName: OnesLike,\n  backendName: \"webgl\",\n  kernelFunc: onesLike3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Pack.js\nfunction pack2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { axis } = attrs;\n  if (inputs.length === 1) {\n    return expandDims4({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });\n  }\n  const shape = inputs[0].shape;\n  const dtype = inputs[0].dtype;\n  inputs.forEach((t) => {\n    util_exports.assertShapesMatch(shape, t.shape, \"All tensors passed to stack must have matching shapes\");\n    util_exports.assert(dtype === t.dtype, () => \"All tensors passed to stack must have matching dtypes\");\n  });\n  const intermediateTensorInfos = [];\n  const expandedTensors = inputs.map((t) => {\n    const expandedT = expandDims4({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });\n    intermediateTensorInfos.push(expandedT);\n    return expandedT;\n  });\n  const result = concat3({ inputs: expandedTensors, backend: backend2, attrs: { axis } });\n  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar packConfig2 = {\n  kernelName: Pack,\n  backendName: \"webgl\",\n  kernelFunc: pack2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/pad_gpu.js\nvar PadProgram = class {\n  constructor(xShape, paddings, constantValue) {\n    this.variableNames = [\"x\"];\n    this.customUniforms = [{ name: \"value\", type: \"float\" }];\n    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);\n    const rank = xShape.length;\n    const type = getCoordsDataType(rank);\n    const start = paddings.map((p2) => p2[0]).join(\",\");\n    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(\",\");\n    const unpackedCoords = [\"coords[0]\", \"coords[1]\", \"coords[2]\", \"coords[3]\"].slice(0, rank);\n    if (rank === 1) {\n      this.userCode = `\n        int start = ${start};\n        int end = ${end};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start || outC >= end) {\n            setOutput(value);\n          } else {\n            setOutput(getX(outC - start));\n          }\n        }\n      `;\n      return;\n    }\n    this.userCode = `\n      ${type} start = ${type}(${start});\n      ${type} end = ${type}(${end});\n\n      void main() {\n        ${type} outC = getOutputCoords();\n        if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {\n          setOutput(value);\n        } else {\n          ${type} coords = outC - start;\n          setOutput(getX(${unpackedCoords}));\n        }\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/pad_packed_gpu.js\nvar PadPackedProgram = class {\n  constructor(xShape, paddings, constantValue) {\n    this.variableNames = [\"x\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.customUniforms = [{ name: \"value\", type: \"float\" }];\n    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);\n    const rank = xShape.length;\n    const dtype = getCoordsDataType(rank);\n    const start = paddings.map((p2) => p2[0]).join(\",\");\n    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(\",\");\n    const coords2 = getChannels(\"rc\", rank);\n    const source = getChannels(\"source\", rank);\n    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;\n    const innerDims = rank === 1 ? \"source\" : `vec2(${source.slice(-2).join()})`;\n    const componentSetup = [\n      `${dtype} rc = outputLoc;`,\n      `${coords2[rank - 1]} += 1;\n       if(${cLimit}) {\n      `,\n      rank === 1 ? \"\" : `}\n       rc = outputLoc;\n       ${coords2[rank - 2]} += 1;\n       if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {`,\n      rank === 1 ? \"\" : `  ${coords2[rank - 1]} += 1;\n         if(${cLimit}) {`\n    ];\n    const paddingArea = rank === 1 ? \"rc < start || rc >= end\" : \"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))\";\n    let mainLoop = \"\";\n    for (let i = 0, j = rank === 1 ? 2 : 4; i < j; i++) {\n      mainLoop += `\n        ${componentSetup[i]}\n        if (${paddingArea}) {\n          result[${i}] = float(value);\n        } else {\n          ${dtype} source = rc - start;\n          result[${i}] = getChannel(getX(${source.join()}), ${innerDims});\n        }\n      `;\n    }\n    mainLoop += rank === 1 ? `} ` : `}}`;\n    this.userCode = `\n      const ${dtype} start = ${dtype}(${start});\n      const ${dtype} end = ${dtype}(${end});\n\n      void main() {\n        ${dtype} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${mainLoop}\n        setOutput(result);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/PadV2.js\nvar padV22 = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { paddings, constantValue } = attrs;\n  if (util_exports.sizeFromShape(x.shape) === 0) {\n    const outputShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);\n    return fill3({\n      backend: backend2,\n      attrs: { shape: outputShape, value: constantValue, dtype: x.dtype }\n    });\n  }\n  const program = env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") ? new PadPackedProgram(x.shape, paddings, constantValue) : new PadProgram(x.shape, paddings, constantValue);\n  const customValues = [[constantValue]];\n  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);\n};\nvar padV2Config2 = {\n  kernelName: PadV2,\n  backendName: \"webgl\",\n  kernelFunc: padV22\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Pow.js\nvar POW = `\n  if(a < 0.0 && floor(b) < b){\n    return NAN;\n  }\n  if (b == 0.0) {\n    return 1.0;\n  }\n  return (round(mod(b, 2.0)) != 1) ?\n      pow(abs(a), b) : sign(a) * pow(abs(a), b);\n`;\nvar POW_PACKED = `\n  // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.\n  vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));\n  vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);\n  vec4 result = multiplier * pow(abs(a), b);\n\n  // Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS\n  bvec4 isExpZero = equal(b, vec4(0.0));\n  result.r = isExpZero.r ? 1.0 : result.r;\n  result.g = isExpZero.g ? 1.0 : result.g;\n  result.b = isExpZero.b ? 1.0 : result.b;\n  result.a = isExpZero.a ? 1.0 : result.a;\n\n  bvec4 isNaN1 = lessThan(a, vec4(0.0));\n  bvec4 isNaN2 = lessThan(floor(b), b);\n  bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);\n  ` + CHECK_NAN_SNIPPET_PACKED + `\n  return result;\n`;\nvar pow3 = binaryKernelFunc2({ opSnippet: POW, packedOpSnippet: POW_PACKED });\nvar powConfig2 = {\n  kernelName: Pow,\n  backendName: \"webgl\",\n  kernelFunc: pow3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Prod.js\nfunction prod3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, keepDims } = attrs;\n  const xRank = x.shape.length;\n  const toDispose = [];\n  const origAxes = util_exports.parseAxisParam(axis, x.shape);\n  let axes = origAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let permutedX = x;\n  if (permutedAxes != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n    toDispose.push(permutedX);\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"prod\", axes, xRank);\n  let res;\n  if (backend2.shouldExecuteOnCPU([permutedX])) {\n    const xVals = backend2.texData.get(permutedX.dataId).values;\n    const { outVals, outShape, outDtype } = prodImplCPU(permutedX.shape, permutedX.dtype, xVals, axes);\n    res = backend2.makeTensorInfo(outShape, outDtype, outVals);\n  } else {\n    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);\n    const inSize = util_exports.sizeFromShape(reduceShape);\n    const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });\n    const outputDType = sumOutType(x.dtype);\n    const reduced = reduce(a2D, outputDType, \"prod\", backend2);\n    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });\n    toDispose.push(a2D);\n    toDispose.push(reduced);\n  }\n  if (keepDims) {\n    toDispose.push(res);\n    const newShape = backend_util_exports.expandShapeToKeepDim(res.shape, origAxes);\n    res = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: newShape } });\n  }\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return res;\n}\nvar prodConfig2 = {\n  kernelName: Prod,\n  backendName: \"webgl\",\n  kernelFunc: prod3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedGather.js\nfunction raggedGather3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;\n  const { outputRaggedRank } = attrs;\n  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.readSync(t.dataId));\n  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);\n  const $paramsDenseValues = backend2.readSync(paramsDenseValues.dataId);\n  const $indices = backend2.readSync(indices.dataId);\n  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImplCPU($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);\n  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], \"int32\", splits));\n  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);\n  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);\n}\nvar raggedGatherConfig2 = {\n  kernelName: RaggedGather,\n  backendName: \"webgl\",\n  kernelFunc: raggedGather3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedRange.js\nfunction raggedRange3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { starts, limits, deltas } = inputs;\n  const $starts = backend2.readSync(starts.dataId);\n  const $limits = backend2.readSync(limits.dataId);\n  const $deltas = backend2.readSync(deltas.dataId);\n  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImplCPU($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);\n  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], \"int32\", rtNestedSplitsData);\n  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);\n  return [rtNestedSplits, rtDenseValues];\n}\nvar raggedRangeConfig2 = {\n  kernelName: RaggedRange,\n  backendName: \"webgl\",\n  kernelFunc: raggedRange3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedTensorToTensor.js\nfunction raggedTensorToTensor3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { shape, values, defaultValue, rowPartitionTensors } = inputs;\n  const { rowPartitionTypes } = attrs;\n  const $shape = backend2.readSync(shape.dataId);\n  const $values = backend2.readSync(values.dataId);\n  const $defaultValue = backend2.readSync(defaultValue.dataId);\n  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.readSync(t.dataId));\n  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);\n  const [outputShape, output] = raggedTensorToTensorImplCPU($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);\n  return backend2.makeTensorInfo(outputShape, values.dtype, output);\n}\nvar raggedTensorToTensorConfig2 = {\n  kernelName: RaggedTensorToTensor,\n  backendName: \"webgl\",\n  kernelFunc: raggedTensorToTensor3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Range.js\nvar range4 = (args) => {\n  const { backend: backend2, attrs } = args;\n  const { start, stop, step: step5, dtype } = attrs;\n  const values = rangeImplCPU(start, stop, step5, dtype);\n  return backend2.makeTensorInfo([values.length], dtype, values);\n};\nvar rangeConfig2 = {\n  kernelName: Range,\n  backendName: \"webgl\",\n  kernelFunc: range4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reciprocal.js\nvar RECIPROCAL = `return 1.0 / x;`;\nvar reciprocal3 = unaryKernelFunc2({ opSnippet: RECIPROCAL });\nvar reciprocalConfig2 = {\n  kernelName: Reciprocal,\n  backendName: \"webgl\",\n  kernelFunc: reciprocal3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Relu.js\nvar RELU3 = CHECK_NAN_SNIPPET + `\n  return (x < 0.0) ? 0.0 : x;\n`;\nvar RELU_PACKED = `\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar relu3 = unaryKernelFunc2({ opSnippet: RELU3, packedOpSnippet: RELU_PACKED });\nvar reluConfig2 = {\n  kernelName: Relu,\n  backendName: \"webgl\",\n  kernelFunc: relu3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Relu6.js\nvar RELU63 = CHECK_NAN_SNIPPET + `\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n`;\nvar RELU6_PACKED = `\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar relu63 = unaryKernelFunc2({ opSnippet: RELU63, packedOpSnippet: RELU6_PACKED });\nvar relu6Config2 = {\n  kernelName: Relu6,\n  backendName: \"webgl\",\n  kernelFunc: relu63\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_gpu.js\nvar ResizeBilinearProgram = class {\n  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {\n    this.variableNames = [\"A\"];\n    this.outputShape = [];\n    const [batch, oldHeight, oldWidth, depth] = inputShape;\n    this.outputShape = [batch, newHeight, newWidth, depth];\n    const effectiveInSize = [\n      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n    ];\n    const effectiveOutSize = [\n      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n    ];\n    let sourceFracIndexRC;\n    if (halfPixelCenters) {\n      sourceFracIndexRC = `(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)`;\n    } else {\n      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;\n    }\n    this.userCode = `\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${effectiveInSize[0] / effectiveOutSize[0]},\n          ${effectiveInSize[1] / effectiveOutSize[1]});\n      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${sourceFracIndexRC};\n\n        // Compute the four integer indices.\n        ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));\n        ivec2 sourceCeilRC = ivec2(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);\n        float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);\n        float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);\n        float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);\n\n        vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);\n\n        float top = topLeft + (topRight - topLeft) * fracRC.y;\n        float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;\n        float newValue = top + (bottom - top) * fracRC.x;\n\n        setOutput(newValue);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_packed_gpu.js\nvar ResizeBilinearPackedProgram = class {\n  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = [];\n    const [batch, oldHeight, oldWidth, depth] = inputShape;\n    this.outputShape = [batch, newHeight, newWidth, depth];\n    const effectiveInSize = [\n      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n    ];\n    const effectiveOutSize = [\n      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n    ];\n    let sourceFracIndexRC;\n    if (halfPixelCenters) {\n      sourceFracIndexRC = `(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)`;\n    } else {\n      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;\n    }\n    this.userCode = `\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${effectiveInSize[0] / effectiveOutSize[0]},\n          ${effectiveInSize[1] / effectiveOutSize[1]},\n          ${effectiveInSize[1] / effectiveOutSize[1]});\n      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,\n                                     ${oldWidth}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${sourceFracIndexRC};\n\n        // Compute the four integer indices.\n        ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));\n        ivec3 sourceCeilRC = ivec3(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${depth - 1};\n        bool hasNextRow = coords.z < ${newWidth - 1};\n\n        // In parallel, construct four corners for all four components in\n        // packed 2x2 cell.\n        vec4 topLeft = vec4(\n          getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 bottomLeft = vec4(\n          getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 topRight = vec4(\n          getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec4 bottomRight = vec4(\n          getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);\n\n        vec4 top = mix(topLeft, topRight, fracRC.yyzz);\n        vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);\n        vec4 newValue = mix(top, bottom, fracRC.x);\n\n        setOutput(newValue);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeBilinear.js\nfunction resizeBilinear3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  const [newHeight, newWidth] = size;\n  const program = env().getBool(\"WEBGL_PACK_IMAGE_OPERATIONS\") ? new ResizeBilinearPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeBilinearProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);\n  return backend2.runWebGLProgram(program, [images], \"float32\");\n}\nvar resizeBilinearConfig2 = {\n  kernelName: ResizeBilinear,\n  backendName: \"webgl\",\n  kernelFunc: resizeBilinear3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_backprop_gpu.js\nvar ResizeBilinearBackpropProgram = class {\n  constructor(dyShape, inputShape, alignCorners) {\n    this.variableNames = [\"dy\"];\n    this.outputShape = [];\n    this.outputShape = inputShape;\n    const [, xHeight, xWidth] = inputShape;\n    const [, yHeight, yWidth] = dyShape;\n    const effectiveXSize = [\n      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,\n      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth\n    ];\n    const effectiveYSize = [\n      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,\n      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth\n    ];\n    const heightScale = effectiveXSize[0] / effectiveYSize[0];\n    const widthScale = effectiveXSize[1] / effectiveYSize[1];\n    const invHeightScale = 1 / heightScale;\n    const invWidthScale = 1 / widthScale;\n    const winHeight = Math.ceil(invHeightScale) * 2 + 2;\n    const winWidth = Math.ceil(invWidthScale) * 2 + 2;\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${heightScale});\n        const float widthScale = float(${widthScale});\n\n        const float invHeightScale = float(${invHeightScale});\n        const float invWidthScale = float(${invWidthScale});\n\n        const int winHeight = int(${winHeight});\n        const int winWidth = int(${winWidth});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(startRLerp - float(winHeight / 2));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(startCLerp - float(winWidth / 2));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${yHeight}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${yWidth}) {\n              continue;\n            }\n\n            float dxR = float(dyR) * heightScale;\n            int topDxRIndex = int(floor(dxR));\n            int bottomDxRIndex = int(min(ceil(dxR), ${xHeight - 1}.0));\n            float dxRLerp = dxR - float(topDxRIndex);\n            float inverseDxRLerp = 1.0 - dxRLerp;\n\n            float dxC = float(dyC) * widthScale;\n            int leftDxCIndex = int(floor(dxC));\n            int rightDxCIndex = int(min(ceil(dxC), ${xWidth - 1}.0));\n            float dxCLerp = dxC - float(leftDxCIndex);\n            float inverseDxCLerp = 1.0 - dxCLerp;\n\n            if (r == topDxRIndex && c == leftDxCIndex) {\n              // topLeft\n              accumulator +=\n                getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;\n            }\n\n            if (r == topDxRIndex && c == rightDxCIndex) {\n              // topRight\n              accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == leftDxCIndex) {\n              // bottomLeft\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == rightDxCIndex) {\n              // bottomRight\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeBilinearGrad.js\nfunction resizeBilinearGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images, dy } = inputs;\n  const { alignCorners } = attrs;\n  const program = new ResizeBilinearBackpropProgram(dy.shape, images.shape, alignCorners);\n  return backend2.runWebGLProgram(program, [dy], dy.dtype);\n}\nvar resizeBilinearGradConfig3 = {\n  kernelName: ResizeBilinearGrad,\n  backendName: \"webgl\",\n  kernelFunc: resizeBilinearGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_gpu.js\nvar ResizeNearestNeighborProgram = class {\n  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {\n    this.variableNames = [\"A\"];\n    this.outputShape = [];\n    const [batch, oldHeight, oldWidth, depth] = inputShape;\n    this.outputShape = [batch, newHeight, newWidth, depth];\n    const effectiveInSize = [\n      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n    ];\n    const effectiveOutSize = [\n      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n    ];\n    const roundBase = alignCorners ? \"0.5\" : \"0.0\";\n    let sourceFracIndexRC;\n    if (halfPixelCenters) {\n      sourceFracIndexRC = `max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))`;\n    } else {\n      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;\n    }\n    this.userCode = `\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${effectiveInSize[0] / effectiveOutSize[0]},\n          ${effectiveInSize[1] / effectiveOutSize[1]});\n      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${sourceFracIndexRC};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec2 sourceNearestRC = ivec2(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));\n        float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);\n\n        setOutput(newValue);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_packed_gpu.js\nvar ResizeNearestNeighborPackedProgram = class {\n  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {\n    this.variableNames = [\"A\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    this.outputShape = [];\n    const [batch, oldHeight, oldWidth, depth] = inputShape;\n    this.outputShape = [batch, newHeight, newWidth, depth];\n    const effectiveInSize = [\n      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,\n      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth\n    ];\n    const effectiveOutSize = [\n      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,\n      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth\n    ];\n    const roundBase = alignCorners ? \"0.5\" : \"0.0\";\n    let sourceFracIndexRC;\n    if (halfPixelCenters) {\n      sourceFracIndexRC = `max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))`;\n    } else {\n      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;\n    }\n    this.userCode = `\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${effectiveInSize[0] / effectiveOutSize[0]},\n          ${effectiveInSize[1] / effectiveOutSize[1]},\n          ${effectiveInSize[1] / effectiveOutSize[1]});\n      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,\n                                     ${oldWidth}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${sourceFracIndexRC};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec3 sourceNearestRC = ivec3(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${depth - 1};\n        bool hasNextRow = coords.z < ${newWidth - 1};\n\n        vec4 newValue = vec4(\n          getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),\n          hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);\n\n        setOutput(newValue);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeNearestNeighbor.js\nfunction resizeNearestNeighbor3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  const [newHeight, newWidth] = size;\n  const program = env().getBool(\"WEBGL_PACK_IMAGE_OPERATIONS\") ? new ResizeNearestNeighborPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeNearestNeighborProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);\n  return backend2.runWebGLProgram(program, [images], images.dtype);\n}\nvar resizeNearestNeighborConfig2 = {\n  kernelName: ResizeNearestNeighbor,\n  backendName: \"webgl\",\n  kernelFunc: resizeNearestNeighbor3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_backprop_gpu.js\nvar ResizeNearestNeigborBackpropProgram = class {\n  constructor(dyShape, inputShape, alignCorners) {\n    this.variableNames = [\"dy\"];\n    this.outputShape = [];\n    this.outputShape = inputShape;\n    const [, xHeight, xWidth] = inputShape;\n    const [, yHeight, yWidth] = dyShape;\n    const effectiveXSize = [\n      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,\n      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth\n    ];\n    const effectiveYSize = [\n      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,\n      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth\n    ];\n    const heightScale = effectiveXSize[0] / effectiveYSize[0];\n    const widthScale = effectiveXSize[1] / effectiveYSize[1];\n    const invHeightScale = 1 / heightScale;\n    const invWidthScale = 1 / widthScale;\n    const winHeight = Math.ceil(invHeightScale) * 2 + 2;\n    const winWidth = Math.ceil(invWidthScale) * 2 + 2;\n    this.userCode = `\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${heightScale});\n        const float widthScale = float(${widthScale});\n\n        const float invHeightScale = float(${invHeightScale});\n        const float invWidthScale = float(${invWidthScale});\n\n        const int winHeight = int(${winHeight});\n        const int winWidth = int(${winWidth});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(floor(startRLerp - float(winHeight / 2)));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(floor(startCLerp - float(winWidth / 2)));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${yHeight}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${yWidth}) {\n              continue;\n            }\n\n            float sourceFracRow =\n              float(${effectiveXSize[0]}) *\n                (float(dyR) / float(${effectiveYSize[0]}));\n\n            float sourceFracCol =\n                float(${effectiveXSize[1]}) *\n                  (float(dyC) / float(${effectiveYSize[1]}));\n\n            int sourceNearestRow = int(min(\n                float(int(${xHeight}) - 1),\n                ${alignCorners} ? float(round(sourceFracRow)) :\n                                  float(floor(sourceFracRow))));\n\n            int sourceNearestCol = int(min(\n                float(int(${xWidth}) - 1),\n                ${alignCorners} ? float(round(sourceFracCol)) :\n                                  float(floor(sourceFracCol))));\n\n            if (r == sourceNearestRow && c == sourceNearestCol) {\n              accumulator += getDy(b, dyR, dyC, d);\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeNearestNeighborGrad.js\nfunction resizeNearestNeighborGrad2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { images, dy } = inputs;\n  const { alignCorners } = attrs;\n  const program = new ResizeNearestNeigborBackpropProgram(dy.shape, images.shape, alignCorners);\n  return backend2.runWebGLProgram(program, [dy], dy.dtype);\n}\nvar resizeNearestNeighborGradConfig3 = {\n  kernelName: ResizeNearestNeighborGrad,\n  backendName: \"webgl\",\n  kernelFunc: resizeNearestNeighborGrad2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/reverse_gpu.js\nvar ReverseProgram = class {\n  constructor(xShape, axis) {\n    this.variableNames = [\"x\"];\n    const rank = xShape.length;\n    if (rank > 4) {\n      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);\n    }\n    this.outputShape = xShape;\n    if (rank === 1) {\n      this.userCode = `\n        void main() {\n          int coord = getOutputCoords();\n          setOutput(getX(${xShape[0]} - coord - 1));\n        }\n      `;\n      return;\n    }\n    const getInCoord = (i) => {\n      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {\n        return `${xShape[i]} - coords[${i}] - 1`;\n      }\n      return `coords[${i}]`;\n    };\n    const inCoords = xShape.map((_, i) => getInCoord(i)).join(\",\");\n    const type = getCoordsDataType(rank);\n    this.userCode = `\n      void main() {\n        ${type} coords = getOutputCoords();\n        setOutput(getX(${inCoords}));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/reverse_packed_gpu.js\nvar ReversePackedProgram = class {\n  constructor(xShape, axis) {\n    this.variableNames = [\"x\"];\n    this.packedInputs = true;\n    this.packedOutput = true;\n    const rank = xShape.length;\n    if (rank > 4) {\n      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);\n    }\n    this.outputShape = xShape;\n    const channels = getChannels(\"rc\", rank);\n    const nextColumn = `${channels[rank - 1]} + 1 < ${this.outputShape[rank - 1]}`;\n    const nextRow = `${channels[rank - 2]} + 1 < ${this.outputShape[rank - 2]}`;\n    const type = getCoordsDataType(rank);\n    if (rank === 1) {\n      this.userCode = `\n        void main(){\n          int rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = getChannel(getX(${xShape[0]} - rc - 1),\n            ${xShape[0]} - rc - 1);\n          if(${nextColumn}){\n              result.g = getChannel(getX(${xShape[0]} - (rc  + 1) - 1),\n                ${xShape[0]} - (rc  + 1) - 1);\n          }\n          setOutput(result);\n        }\n      `;\n    } else {\n      this.userCode = `\n        void main() {\n          ${type} rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = ${getR(channels.slice())};\n          if(${nextColumn}){\n            result.g = ${getG(channels.slice())};\n          }\n          if(${nextRow}) {\n            result.b = ${getB(channels.slice())};\n            if(${nextColumn}) {\n              result.a = ${getA(channels.slice())};\n            }\n          }\n          setOutput(result);\n        }\n    `;\n    }\n    function getR(channels2) {\n      return getChannel(channels2);\n    }\n    function getG(channels2) {\n      channels2[rank - 1] = \"(\" + channels2[rank - 1] + ` + 1)`;\n      return getChannel(channels2);\n    }\n    function getB(channels2) {\n      channels2[rank - 2] = \"(\" + channels2[rank - 2] + ` + 1)`;\n      return getChannel(channels2);\n    }\n    function getA(channels2) {\n      channels2[rank - 1] = \"(\" + channels2[rank - 1] + ` + 1)`;\n      channels2[rank - 2] = \"(\" + channels2[rank - 2] + ` + 1)`;\n      return getChannel(channels2);\n    }\n    function getChannel(channels2) {\n      const inCoordsArray = xShape.map((_, i) => getInCoord(i, channels2));\n      const inCoords = inCoordsArray.join(\",\");\n      const innerDims = inCoordsArray.slice(-2).join(\",\");\n      return `getChannel(getX(${inCoords}), vec2(${innerDims}))`;\n    }\n    function getInCoord(i, channels1) {\n      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {\n        return `${xShape[i]} - ${channels1[i]} - 1`;\n      } else {\n        return `${channels1[i]}`;\n      }\n    }\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reverse.js\nfunction reverse3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { dims } = attrs;\n  const xRank = x.shape.length;\n  const $dims = util_exports.parseAxisParam(dims, x.shape);\n  if (xRank === 0) {\n    return identity3({ inputs: { x }, backend: backend2 });\n  }\n  const program = env().getBool(\"WEBGL_PACK_ARRAY_OPERATIONS\") ? new ReversePackedProgram(x.shape, $dims) : new ReverseProgram(x.shape, $dims);\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\nvar reverseConfig2 = {\n  kernelName: Reverse,\n  backendName: \"webgl\",\n  kernelFunc: reverse3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/rotate_gpu.js\nvar RotateProgram = class {\n  constructor(imageShape, fillValue) {\n    this.variableNames = [\"Image\"];\n    this.outputShape = [];\n    this.customUniforms = [{ name: \"params\", type: \"vec4\" }];\n    const imageHeight = imageShape[1];\n    const imageWidth = imageShape[2];\n    this.outputShape = imageShape;\n    let fillSnippet = \"\";\n    if (typeof fillValue === \"number\") {\n      fillSnippet = `float outputValue = ${fillValue.toFixed(2)};`;\n    } else {\n      fillSnippet = `\n        vec3 fill = vec3(${fillValue.join(\",\")});\n        float outputValue = fill[coords[3]];`;\n    }\n    this.userCode = `\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n          int y = coords[1];\n          float coordXFloat = (float(x) - params[0]) * params[3] -\n            (float(y) - params[1]) * params[2];\n          float coordYFloat = (float(x) - params[0]) * params[2] +\n            (float(y) - params[1]) * params[3];\n          int coordX = int(round(coordXFloat + params[0]));\n          int coordY = int(round(coordYFloat + params[1]));\n          ${fillSnippet}\n          if(coordX >= 0 && coordX < ${imageWidth} && coordY >= 0 && coordY < ${imageHeight}) {\n            outputValue = getImage(coords[0], coordY, coordX, coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RotateWithOffset.js\nvar rotateWithOffsetConfig2 = {\n  kernelName: RotateWithOffset,\n  backendName: \"webgl\",\n  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {\n    const { image: image2 } = inputs;\n    const { radians, fillValue, center } = attrs;\n    const webglBackend = backend2;\n    const program = new RotateProgram(image2.shape, fillValue);\n    const [centerX, centerY] = backend_util_exports.getImageCenter(center, image2.shape[1], image2.shape[2]);\n    const customValues = [[centerX, centerY, Math.sin(radians), Math.cos(radians)]];\n    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype, customValues);\n    return output;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Round.js\nvar ROUND = `\n  // OpenGL ES does not support round function.\n  // The algorithm is based on banker's rounding.\n  float base = floor(x);\n  if ((x - base) < 0.5) {\n    return floor(x);\n  } else if ((x - base) > 0.5) {\n    return ceil(x);\n  } else {\n    if (mod(base, 2.0) == 0.0) {\n      return base;\n    } else {\n      return base + 1.0;\n    }\n  }\n`;\nvar round4 = unaryKernelFunc2({ opSnippet: ROUND });\nvar roundConfig2 = {\n  kernelName: Round,\n  backendName: \"webgl\",\n  kernelFunc: round4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Rsqrt.js\nvar RSQRT = `return inversesqrt(x);`;\nvar rsqrt3 = unaryKernelFunc2({ opSnippet: RSQRT, cpuKernelImpl: rsqrtImplCPU });\nvar rsqrtConfig2 = {\n  kernelName: Rsqrt,\n  backendName: \"webgl\",\n  kernelFunc: rsqrt3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/scatter_gpu.js\nvar ScatterProgram = class {\n  constructor(updateSize, sliceDim, indicesRank, updatesRank, strides, shape, summingDupeIndex = true) {\n    this.variableNames = [\"updates\", \"indices\", \"defaultValue\"];\n    this.outputShape = shape;\n    const stridesType = getCoordsDataType(strides.length);\n    const dtype = getCoordsDataType(shape.length);\n    let indicesString = \"\";\n    if (indicesRank === 1) {\n      indicesString = \"i\";\n    } else if (indicesRank === 2) {\n      indicesString = \"i, j\";\n    }\n    const indicesSnippet = `getIndices(${indicesString})`;\n    let updatesString = \"\";\n    if (updatesRank === 1) {\n      updatesString = \"i\";\n    } else if (updatesRank === 2) {\n      updatesString = \"i, coords[1]\";\n    }\n    const updatesSnippet = `getUpdates(${updatesString})`;\n    const strideString = sliceDim > 1 ? \"strides[j]\" : \"strides\";\n    this.userCode = `\n        ${stridesType} strides = ${stridesType}(${strides});\n\n        void main() {\n          ${dtype} coords = getOutputCoords();\n          float sum = 0.0;\n          bool found = false;\n          for (int i = 0; i < ${updateSize}; i++) {\n            int flattenedIndex = 0;\n            for (int j = 0; j < ${sliceDim}; j++) {\n              int index = round(${indicesSnippet});\n              flattenedIndex += index * ${strideString};\n            }\n            if (flattenedIndex == coords[0]) {\n              sum += ${updatesSnippet};\n              found = true;\n            }\n          }\n          setOutput(mix(getDefaultValue(), sum, float(found)));\n        }\n      `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ScatterNd.js\nfunction scatterNd2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { indices, updates } = inputs;\n  const { shape } = attrs;\n  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);\n  const flattenShape = [outputSize / sliceSize, sliceSize];\n  if (outputSize === 0) {\n    return backend2.makeTensorInfo(shape, indices.dtype);\n  }\n  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numUpdates, sliceRank] } });\n  const flattenX = reshape4({ inputs: { x: updates }, backend: backend2, attrs: { shape: [numUpdates, sliceSize] } });\n  const defaultValue = backend2.makeTensorInfo([], \"float32\", new Float32Array([0]));\n  const program = new ScatterProgram(numUpdates, sliceRank, flattenIndices.shape.length, flattenX.shape.length, strides, flattenShape);\n  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices, defaultValue], flattenX.dtype);\n  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape } });\n  backend2.disposeIntermediateTensorInfo(flattenIndices);\n  backend2.disposeIntermediateTensorInfo(flattenX);\n  backend2.disposeIntermediateTensorInfo(res);\n  backend2.disposeIntermediateTensorInfo(defaultValue);\n  return reshaped;\n}\nvar scatterNdConfig2 = {\n  kernelName: ScatterNd,\n  backendName: \"webgl\",\n  kernelFunc: scatterNd2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/search_sorted_gpu.js\nvar SearchSortedProgram = class {\n  constructor(batchSize, numInputs, numValues, side) {\n    this.variableNames = [\"sortedSequence\", \"values\"];\n    this.customUniforms = [{ name: \"numInputs\", type: \"int\" }];\n    this.outputShape = [batchSize, numValues];\n    const webGL2LoopHead = \"while (left < right) {\";\n    const webGL1LoopHead = `for (int i = 0; i < ${Math.ceil(Math.log2(numInputs + 1))}; ++i) { if (left >= right) break;`;\n    const loopHead = env().getNumber(\"WEBGL_VERSION\") === 2 ? webGL2LoopHead : webGL1LoopHead;\n    const boundComparator = side === \"left\" ? \"<\" : \"<=\";\n    this.userCode = `\n       int findBound(int batch, float value) {\n         int left = 0;\n         int right = numInputs;\n         int mid;\n         ${loopHead}\n           mid = (left + right) / 2;\n           if (getSortedSequence(batch, mid) ${boundComparator} value) {\n             left = mid + 1;\n           } else {\n             right = mid;\n           }\n         }\n         return right;\n       }\n\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int valueIndex = coords[1];\n\n         float value = getValues(batch, valueIndex);\n\n         setOutput(float(findBound(batch, value)));\n       }\n     `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SearchSorted.js\nfunction searchSorted3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { sortedSequence, values } = inputs;\n  const { side } = attrs;\n  const program = new SearchSortedProgram(sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);\n  const customValues = [[sortedSequence.shape[1]]];\n  return backend2.runWebGLProgram(program, [sortedSequence, values], \"int32\", customValues);\n}\nvar searchSortedConfig2 = {\n  kernelName: SearchSorted,\n  backendName: \"webgl\",\n  kernelFunc: searchSorted3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/select_gpu.js\nvar SelectProgram = class {\n  constructor(cRank, shape, rank) {\n    this.variableNames = [\"c\", \"a\", \"b\"];\n    this.outputShape = shape;\n    let cCoords;\n    let abCoords;\n    if (rank > 4) {\n      throw Error(`Where for rank ${rank} is not yet supported`);\n    }\n    if (rank === 1) {\n      abCoords = `resRC`;\n      cCoords = `resRC`;\n    } else {\n      const currentCoords = [\"resRC.x\", \"resRC.y\", \"resRC.z\", \"resRC.w\"];\n      const cCoordVars = [];\n      const abCoordVars = [];\n      for (let i = 0; i < shape.length; i++) {\n        abCoordVars.push(`${currentCoords[i]}`);\n        if (i < cRank) {\n          cCoordVars.push(`${currentCoords[i]}`);\n        }\n      }\n      cCoords = cCoordVars.join();\n      abCoords = abCoordVars.join();\n    }\n    const dtype = getCoordsDataType(rank);\n    this.userCode = `\n      void main() {\n        ${dtype} resRC = getOutputCoords();\n        float cVal = getC(${cCoords});\n        if (cVal >= 1.0) {\n          setOutput(getA(${abCoords}));\n        } else {\n          setOutput(getB(${abCoords}));\n        }\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Select.js\nfunction select3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { condition, t, e } = inputs;\n  const program = new SelectProgram(condition.shape.length, t.shape, t.shape.length);\n  return backend2.runWebGLProgram(program, [condition, t, e], upcastType(t.dtype, e.dtype));\n}\nvar selectConfig2 = {\n  kernelName: Select,\n  backendName: \"webgl\",\n  kernelFunc: select3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Selu.js\nvar SELU = `\n  // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.\n  // see: https://arxiv.org/abs/1706.02515\n  float scaleAlpha = ${backend_util_exports.SELU_SCALEALPHA};\n  float scale = ${backend_util_exports.SELU_SCALE};\n  return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);\n`;\nvar selu3 = unaryKernelFunc2({ opSnippet: SELU });\nvar seluConfig2 = {\n  kernelName: Selu,\n  backendName: \"webgl\",\n  kernelFunc: selu3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sigmoid.js\nvar SIGMOID3 = CHECK_NAN_SNIPPET_UNARY + `\n  return 1.0 / (1.0 + exp(-1.0 * x));\n`;\nvar SIGMOID_PACKED = `\n  vec4 result = 1.0 / (1.0 + exp(-1.0 * x));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n`;\nvar sigmoid3 = unaryKernelFunc2({\n  opSnippet: SIGMOID3,\n  packedOpSnippet: SIGMOID_PACKED,\n  cpuKernelImpl: sigmoidImplCPU\n});\nvar sigmoidConfig2 = {\n  kernelName: Sigmoid,\n  backendName: \"webgl\",\n  kernelFunc: sigmoid3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sign.js\nvar SIGN = `\n  if (isnan(x)) { return 0.0; }\n  return sign(x);\n`;\nvar sign3 = unaryKernelFunc2({ opSnippet: SIGN });\nvar signConfig2 = {\n  kernelName: Sign,\n  backendName: \"webgl\",\n  kernelFunc: sign3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sin.js\nvar SIN = CHECK_NAN_SNIPPET_UNARY + `\n  return sin(x);\n`;\nvar sin3 = unaryKernelFunc2({ opSnippet: SIN });\nvar sinConfig2 = {\n  kernelName: Sin,\n  backendName: \"webgl\",\n  kernelFunc: sin3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sinh.js\nvar SINH = `\n  float e2x = exp(x);\n  return (e2x - 1.0 / e2x) / 2.0;\n`;\nvar sinh3 = unaryKernelFunc2({ opSnippet: SINH });\nvar sinhConfig2 = {\n  kernelName: Sinh,\n  backendName: \"webgl\",\n  kernelFunc: sinh3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Softplus.js\nvar SOFTPLUS = `\n  float epsilon = 1.1920928955078125e-7;\n  float threshold = log(epsilon) + 2.0;\n\n  bool too_large = x > -threshold;\n  bool too_small = x < threshold;\n\n  float result;\n  float exp_x = exp(x);\n\n  if (too_large){\n    result = x;\n  }\n  else if (too_small){\n    result = exp_x;\n  }\n  else{\n    result = log(exp_x + 1.0);\n  }\n  return result;\n`;\nvar softplus3 = unaryKernelFunc2({ opSnippet: SOFTPLUS });\nvar softplusConfig2 = {\n  kernelName: Softplus,\n  backendName: \"webgl\",\n  kernelFunc: softplus3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SpaceToBatchND.js\nvar spaceToBatchND3 = (args) => {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, paddings } = attrs;\n  util_exports.assert(x.shape.length <= 4, () => \"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet\");\n  const prod5 = blockShape.reduce((a, b) => a * b);\n  const completePaddings = [[0, 0]];\n  completePaddings.push(...paddings);\n  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {\n    completePaddings.push([0, 0]);\n  }\n  const toDispose = [];\n  const paddedX = padV22({\n    inputs: { x },\n    backend: backend2,\n    attrs: { paddings: completePaddings, constantValue: 0 }\n  });\n  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);\n  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);\n  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);\n  const reshapedPaddedX = reshape4({ inputs: { x: paddedX }, backend: backend2, attrs: { shape: reshapedPaddedShape } });\n  const paddedXT = transpose3({\n    inputs: { x: reshapedPaddedX },\n    backend: backend2,\n    attrs: { perm: permutedReshapedPaddedPermutation }\n  });\n  const result = reshape4({ inputs: { x: paddedXT }, backend: backend2, attrs: { shape: flattenShape } });\n  toDispose.push(paddedX);\n  toDispose.push(reshapedPaddedX);\n  toDispose.push(paddedXT);\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n};\nvar spaceToBatchNDConfig2 = {\n  kernelName: SpaceToBatchND,\n  backendName: \"webgl\",\n  kernelFunc: spaceToBatchND3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseFillEmptyRows.js\nfunction sparseFillEmptyRows3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { indices, values, denseShape, defaultValue } = inputs;\n  if (denseShape.shape.length !== 1) {\n    throw new Error(`Dense shape must be a vector, saw:\n         ${denseShape.shape}`);\n  }\n  if (indices.shape.length !== 2) {\n    throw new Error(`Indices must be a matrix, saw:\n         ${indices.shape}`);\n  }\n  if (values.shape.length !== 1) {\n    throw new Error(`Values must be a vector, saw:\n         ${values.shape}`);\n  }\n  if (defaultValue.shape.length !== 0) {\n    throw new Error(`Default value must be a scalar, saw:\n        ${defaultValue.shape}`);\n  }\n  const $indices = backend2.readSync(indices.dataId);\n  const $values = backend2.readSync(values.dataId);\n  const $denseShape = backend2.readSync(denseShape.dataId);\n  const $defaultValue = backend2.readSync(defaultValue.dataId)[0];\n  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImplCPU($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);\n  return [\n    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),\n    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),\n    backend2.makeTensorInfo([emptyRowIndicator.length], \"bool\", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),\n    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))\n  ];\n}\nvar sparseFillEmptyRowsConfig2 = {\n  kernelName: SparseFillEmptyRows,\n  backendName: \"webgl\",\n  kernelFunc: sparseFillEmptyRows3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseReshape.js\nfunction sparseReshape3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { inputIndices, inputShape, newShape } = inputs;\n  if (inputIndices.shape.length !== 2) {\n    throw new Error(`Input indices should be a matrix but received shape ${inputIndices.shape}`);\n  }\n  if (inputShape.shape.length !== 1) {\n    throw new Error(`Input shape should be a vector but received shape ${inputShape.shape}`);\n  }\n  if (newShape.shape.length !== 1) {\n    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);\n  }\n  const $inputShape = Array.from(backend2.readSync(inputShape.dataId));\n  const $inputIndices = backend2.readSync(inputIndices.dataId);\n  const targetShape = Array.from(backend2.readSync(newShape.dataId));\n  const [newIndices, indicesShape, outputShape] = sparseReshapeImplCPU($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);\n  return [\n    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),\n    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))\n  ];\n}\nvar sparseReshapeConfig2 = {\n  kernelName: SparseReshape,\n  backendName: \"webgl\",\n  kernelFunc: sparseReshape3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseSegmentMean.js\nfunction sparseSegmentMean3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { data, indices, segmentIds } = inputs;\n  if (data.shape.length < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if (indices.shape.length !== 1) {\n    throw new Error(`Indices should be a vector but received shape\n              ${indices.shape}`);\n  }\n  if (segmentIds.shape.length !== 1) {\n    throw new Error(`Segment ids should be a vector but received shape\n              ${segmentIds.shape}`);\n  }\n  const $data = backend2.readSync(data.dataId);\n  const $indices = backend2.readSync(indices.dataId);\n  const $segmentIds = backend2.readSync(segmentIds.dataId);\n  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds, true);\n  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);\n}\nvar sparseSegmentMeanConfig2 = {\n  kernelName: SparseSegmentMean,\n  backendName: \"webgl\",\n  kernelFunc: sparseSegmentMean3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseSegmentSum.js\nfunction sparseSegmentSum3(args) {\n  const { inputs, backend: backend2 } = args;\n  const { data, indices, segmentIds } = inputs;\n  if (data.shape.length < 1) {\n    throw new Error(`Data should be at least 1 dimensional but received scalar`);\n  }\n  if (indices.shape.length !== 1) {\n    throw new Error(`Indices should be a vector but received shape\n             ${indices.shape}`);\n  }\n  if (segmentIds.shape.length !== 1) {\n    throw new Error(`Segment ids should be a vector but received shape\n             ${segmentIds.shape}`);\n  }\n  const $data = backend2.readSync(data.dataId);\n  const $indices = backend2.readSync(indices.dataId);\n  const $segmentIds = backend2.readSync(segmentIds.dataId);\n  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds);\n  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);\n}\nvar sparseSegmentSumConfig2 = {\n  kernelName: SparseSegmentSum,\n  backendName: \"webgl\",\n  kernelFunc: sparseSegmentSum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseToDense.js\nfunction sparseToDense3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { sparseIndices, sparseValues, defaultValue } = inputs;\n  const { outputShape } = attrs;\n  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);\n  const sumDupeIndices = false;\n  if (sparseValues.dtype === \"string\") {\n    const indicesBuf = backend2.bufferSync(sparseIndices);\n    const updatesBuf = backend2.bufferSync(sparseValues);\n    const $defaultValue = util_exports.decodeString(backend2.readSync(defaultValue.dataId)[0]);\n    const outBuf = scatterImplCPU(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);\n    return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);\n  }\n  const program = new ScatterProgram(numUpdates, sliceRank, sparseIndices.shape.length, sparseValues.shape.length, strides, [outputSize, 1], sumDupeIndices);\n  const res = backend2.runWebGLProgram(program, [sparseValues, sparseIndices, defaultValue], sparseValues.dtype);\n  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outputShape } });\n  backend2.disposeIntermediateTensorInfo(res);\n  return reshaped;\n}\nvar sparseToDenseConfig2 = {\n  kernelName: SparseToDense,\n  backendName: \"webgl\",\n  kernelFunc: sparseToDense3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SplitV.js\nfunction splitV2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { numOrSizeSplits, axis } = attrs;\n  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];\n  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);\n  const xRank = x.shape.length;\n  const begin = new Array(xRank).fill(0);\n  const size = x.shape.slice();\n  return splitSizes.map((s) => {\n    const sliceSize = [...size];\n    sliceSize[$axis] = s;\n    const sliceT = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });\n    begin[$axis] += s;\n    return sliceT;\n  });\n}\nvar splitVConfig2 = {\n  kernelName: SplitV,\n  backendName: \"webgl\",\n  kernelFunc: splitV2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sqrt.js\nvar SQRT = `return sqrt(x);`;\nvar sqrt3 = unaryKernelFunc2({ opSnippet: SQRT, packedOpSnippet: SQRT, cpuKernelImpl: sqrtImplCPU });\nvar sqrtConfig2 = {\n  kernelName: Sqrt,\n  backendName: \"webgl\",\n  kernelFunc: sqrt3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Square.js\nvar SQUARE = `return x * x;`;\nvar square3 = unaryKernelFunc2({ opSnippet: SQUARE });\nvar squareConfig2 = {\n  kernelName: Square,\n  backendName: \"webgl\",\n  kernelFunc: square3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SquaredDifference.js\nvar SQUARED_DIFFERENCE = \"return (a - b) * (a - b);\";\nvar squaredDifference3 = binaryKernelFunc2({ opSnippet: SQUARED_DIFFERENCE, packedOpSnippet: SQUARED_DIFFERENCE });\nvar squaredDifferenceConfig2 = {\n  kernelName: SquaredDifference,\n  backendName: \"webgl\",\n  kernelFunc: squaredDifference3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Step.js\nfunction step3({ inputs, attrs, backend: backend2 }) {\n  const { x } = inputs;\n  const opSnippet = CHECK_NAN_SNIPPET + `\n    return x > 0.0 ? 1.0 : float(${attrs.alpha});\n  `;\n  const program = new UnaryOpProgram(x.shape, opSnippet);\n  return backend2.runWebGLProgram(program, [x], x.dtype);\n}\nvar stepConfig2 = {\n  kernelName: Step,\n  backendName: \"webgl\",\n  kernelFunc: step3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/strided_slice_gpu.js\nvar StridedSliceProgram = class {\n  constructor(begin, strides, size) {\n    this.variableNames = [\"x\"];\n    this.outputShape = size;\n    const rank = size.length;\n    const inputDtype = getCoordsDataType(size.length);\n    const dtype = getCoordsDataType(size.length);\n    let newCoords = \"\";\n    if (rank === 1) {\n      newCoords = \"coords * strides + begin\";\n    } else {\n      let outputAxis = 0;\n      newCoords = size.map((_, i) => {\n        outputAxis++;\n        return size.length === 1 ? `coords * strides[${i}] + begin[${i}]` : `coords[${outputAxis - 1}] * strides[${i}] + begin[${i}]`;\n      }).join(\",\");\n    }\n    this.userCode = `\n      ${inputDtype} begin = ${inputDtype}(${begin});\n      ${inputDtype} strides = ${inputDtype}(${strides});\n\n      void main() {\n        ${dtype} coords = getOutputCoords();\n        setOutput(getX(${newCoords}));\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StridedSlice.js\nfunction stridedSlice3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;\n  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);\n  let result;\n  if (isIdentity) {\n    result = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });\n  } else if (sliceDim0 || isSimpleSlice) {\n    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);\n    const size = slice_util_exports.computeOutShape($begin, $end, $strides);\n    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });\n    result = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });\n    backend2.disposeIntermediateTensorInfo(sliced);\n  } else {\n    const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);\n    if (shouldExecuteOnCPU) {\n      const values = backend2.readSync(x.dataId);\n      const xBuf = buffer(x.shape, x.dtype, values);\n      const resultValues = stridedSliceImplCPU(finalShapeSparse, xBuf, $strides, $begin);\n      result = backend2.makeTensorInfo(finalShape, x.dtype, resultValues.values);\n    } else {\n      const program = new StridedSliceProgram($begin, $strides, finalShapeSparse);\n      result = backend2.runWebGLProgram(program, [x], x.dtype);\n    }\n  }\n  const resultReshaped = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: finalShape } });\n  backend2.disposeIntermediateTensorInfo(result);\n  return resultReshaped;\n}\nvar stridedSliceConfig2 = {\n  kernelName: StridedSlice,\n  backendName: \"webgl\",\n  kernelFunc: stridedSlice3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringNGrams.js\nfunction stringNGrams3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;\n  const { data, dataSplits } = inputs;\n  const $data = backend2.readSync(data.dataId);\n  const $dataSplits = backend2.readSync(dataSplits.dataId);\n  const [nGrams, nGramsSplits] = stringNGramsImplCPU($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);\n  return [\n    backend2.makeTensorInfo([nGrams.length], \"string\", nGrams),\n    backend2.makeTensorInfo(dataSplits.shape, \"int32\", nGramsSplits)\n  ];\n}\nvar stringNGramsConfig2 = {\n  kernelName: StringNGrams,\n  backendName: \"webgl\",\n  kernelFunc: stringNGrams3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringSplit.js\nfunction stringSplit3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { skipEmpty } = attrs;\n  const { input: input2, delimiter } = inputs;\n  if (input2.dtype !== \"string\") {\n    throw new Error(\"Input must be of datatype string\");\n  }\n  if (input2.shape.length !== 1) {\n    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);\n  }\n  if (delimiter.shape.length !== 0) {\n    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);\n  }\n  const $input = backend2.readSync(input2.dataId);\n  const $delimiter = backend2.readSync(delimiter.dataId)[0];\n  const [indices, values, shape] = stringSplitImplCPU($input, $delimiter, skipEmpty);\n  const outputSize = values.length;\n  return [\n    backend2.makeTensorInfo([outputSize, 2], \"int32\", indices),\n    backend2.makeTensorInfo([outputSize], \"string\", values),\n    backend2.makeTensorInfo([2], \"int32\", new Int32Array(shape))\n  ];\n}\nvar stringSplitConfig2 = {\n  kernelName: StringSplit,\n  backendName: \"webgl\",\n  kernelFunc: stringSplit3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringToHashBucketFast.js\nfunction stringToHashBucketFast3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { numBuckets } = attrs;\n  const { input: input2 } = inputs;\n  if (input2.dtype !== \"string\") {\n    throw new Error(\"Input must be of datatype string\");\n  }\n  if (numBuckets <= 0) {\n    throw new Error(`Number of buckets must be at least 1`);\n  }\n  const $input = backend2.readSync(input2.dataId);\n  const output = stringToHashBucketFastImplCPU($input, numBuckets);\n  return backend2.makeTensorInfo(input2.shape, \"int32\", output);\n}\nvar stringToHashBucketFastConfig2 = {\n  kernelName: StringToHashBucketFast,\n  backendName: \"webgl\",\n  kernelFunc: stringToHashBucketFast3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tan.js\nvar TAN = `return tan(x);`;\nvar tan3 = unaryKernelFunc2({ opSnippet: TAN });\nvar tanConfig2 = {\n  kernelName: Tan,\n  backendName: \"webgl\",\n  kernelFunc: tan3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tanh.js\nvar TANH = `\n  float e2x = exp(-2.0 * abs(x));\n  return sign(x) * (1.0 - e2x) / (1.0 + e2x);\n`;\nvar tanh4 = unaryKernelFunc2({ opSnippet: TANH });\nvar tanhConfig2 = {\n  kernelName: Tanh,\n  backendName: \"webgl\",\n  kernelFunc: tanh4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/tile_gpu.js\nvar TileProgram = class {\n  constructor(aShape, reps) {\n    this.variableNames = [\"A\"];\n    const outputShape = new Array(aShape.length);\n    for (let i = 0; i < outputShape.length; i++) {\n      outputShape[i] = aShape[i] * reps[i];\n    }\n    this.outputShape = outputShape;\n    this.rank = outputShape.length;\n    const dtype = getCoordsDataType(this.rank);\n    const sourceCoords = getSourceCoords3(aShape);\n    this.userCode = `\n      void main() {\n        ${dtype} resRC = getOutputCoords();\n        setOutput(getA(${sourceCoords}));\n      }\n    `;\n  }\n};\nfunction getSourceCoords3(aShape) {\n  const rank = aShape.length;\n  if (rank > 5) {\n    throw Error(`Tile for rank ${rank} is not yet supported`);\n  }\n  if (rank === 1) {\n    return `imod(resRC, ${aShape[0]})`;\n  }\n  const currentCoords = [\"resRC.x\", \"resRC.y\", \"resRC.z\", \"resRC.w\", \"resRC.u\"];\n  const sourceCoords = [];\n  for (let i = 0; i < aShape.length; i++) {\n    sourceCoords.push(`imod(${currentCoords[i]}, ${aShape[i]})`);\n  }\n  return sourceCoords.join();\n}\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tile.js\nfunction tile4(params) {\n  const { inputs, backend: backend2, attrs } = params;\n  const { x } = inputs;\n  const { reps } = attrs;\n  if (x.dtype === \"string\" || x.shape.length > 5) {\n    const data = backend2.readSync(x.dataId);\n    const value = x.dtype === \"string\" ? data.map((d) => util_exports.decodeString(d)) : data;\n    const buf = buffer(x.shape, x.dtype, value);\n    const outBuf = tileImplCPU(buf, reps);\n    return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);\n  }\n  const program = new TileProgram(x.shape, reps);\n  const output = backend2.runWebGLProgram(program, [x], x.dtype);\n  return output;\n}\nvar tileConfig2 = {\n  kernelName: Tile,\n  backendName: \"webgl\",\n  kernelFunc: tile4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/top_k_gpu.js\nvar SwapProgram = class {\n  constructor(shape) {\n    this.variableNames = [\"x\", \"indices\"];\n    this.customUniforms = [\n      { name: \"n\", type: \"int\" },\n      { name: \"firstPass\", type: \"int\" },\n      { name: \"negativeInf\", type: \"float\" },\n      { name: \"dir\", type: \"int\" },\n      { name: \"inc\", type: \"int\" }\n    ];\n    this.outputShape = shape;\n    this.userCode = `\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // We compare elements pair-wise within a group of size 2 * inc.\n         // The comparing rule for each group alternates between ascending\n         // and descending. Within each group, we compare each pair at\n         // positions i and i+inc. To decide whether an element at position i\n         // is x0 or x1, we mod it by 2 * inc, if the result is smaller than\n         // inc, it is in the first half of the group, we denote it as x0,\n         // otherwise we denote it as x1.\n         // For example, as shown in the Bitonic top K paper referenced above,\n         // Figure5(a) shows that element[1] is in the\n         // second half of the group when group size is 2, but it is in the\n         // first half of the group when group size is 4.\n\n         bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;\n         int i = isFirstInPair ? elemIdx : elemIdx - inc;\n\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));\n         float x0 = i0 < n ? getX(batch, i0) : negativeInf;\n         float x1 = i1 < n ? getX(batch, i1) : negativeInf;\n\n         // Denotes which direction indices are in (ascending or descending).\n         bool reverse = imod(elemIdx, 2 * dir) >= dir;\n         bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);\n         if (reverse == isGreater) { // Elements in opposite order of direction\n           int iTemp = i0;\n           i0 = i1;\n           i1 = iTemp;\n         }\n         if (isFirstInPair) {\n            setOutput(float(i0));\n         } else {\n            setOutput(float(i1));\n         }\n       }\n     `;\n  }\n};\nvar MergeProgram = class {\n  constructor(shape) {\n    this.variableNames = [\"x\", \"indices\"];\n    this.customUniforms = [\n      { name: \"n\", type: \"int\" },\n      { name: \"firstPass\", type: \"int\" },\n      { name: \"k\", type: \"int\" }\n    ];\n    this.outputShape = shape;\n    this.userCode = `\n    void main() {\n         // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // The output size is half of the previous size.\n         // If the previous sequence is | | | | _ _ _ _  | | | |  _ _ _ _ (k=4),\n         // we only need to output the indices at positions |, the indices at\n         // positions _ can be thrown away, see Figure5(b) After Phase 2\n         // (Merge phase) in the Bitonic Top K paper referenced above.\n         // For example, the paper shows we only need to output the orange bars.\n         // The output sequence should look like this | | | | | | | |.\n         // Because the sequence is halved, to map the output index back\n         // to the previous sequence to find the corresponding value,\n         // we need to double the index. When we double the index,\n         // we basically interpolate a position, so 2i looks like\n         // | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position\n         // of each 2k positions by - elemIdx % k. E.g. for output at\n         // index 4,5,6,7, we want to get the corresponding element at\n         // original index 8,9,10,11, for output at index 8,9,10,11,\n         // we want to get the corresponding element at original index\n         // 16,17,18,19, so on and so forth.\n\n         int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));\n\n         float x0 = getX(batch, i0);\n         float x1 = i1 < n ? getX(batch, i1) : x0;\n\n         setOutput(x0 >= x1 ? float(i0) : float(i1));\n       }\n     `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/TopK.js\nfunction disposeIntermediateTensorInfoOrNull(backend2, tensorInfo) {\n  if (tensorInfo !== null) {\n    backend2.disposeIntermediateTensorInfo(tensorInfo);\n  }\n}\nfunction roundUpToPow2(num) {\n  let pow22 = 1;\n  while (pow22 < num) {\n    pow22 *= 2;\n  }\n  return pow22;\n}\nfunction topK2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { k, sorted } = attrs;\n  const TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber(\"TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD\");\n  const TOPK_K_CPU_HANDOFF_THRESHOLD = env().getNumber(\"TOPK_K_CPU_HANDOFF_THRESHOLD\");\n  const xShape = x.shape;\n  const lastDim = xShape[xShape.length - 1];\n  if (backend2.shouldExecuteOnCPU([x]) || lastDim < TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD || k > TOPK_K_CPU_HANDOFF_THRESHOLD) {\n    const xVals = backend2.readSync(x.dataId);\n    const [allTopKVals, allTopKIndices] = topKImplCPU(xVals, xShape, x.dtype, k, sorted);\n    return [\n      backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),\n      backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)\n    ];\n  }\n  if (k === 0) {\n    xShape[xShape.length - 1] = 0;\n    return [\n      backend2.makeTensorInfo(xShape, x.dtype, []),\n      backend2.makeTensorInfo(xShape, \"int32\", [])\n    ];\n  }\n  if (lastDim === 1) {\n    return [\n      x,\n      fill3({ attrs: { shape: xShape, dtype: \"int32\", value: 0 }, backend: backend2 })\n    ];\n  }\n  const xtexData = backend2.texData.get(x.dataId);\n  const xIsPacked = xtexData !== null && xtexData.isPacked;\n  const xUnPacked = xIsPacked ? backend2.unpackTensor(x) : x;\n  const xSize = util_exports.sizeFromShape(xShape);\n  const batch = xSize / lastDim;\n  const x2D = reshape4({ inputs: { x: xUnPacked }, attrs: { shape: [batch, lastDim] }, backend: backend2 });\n  if (xIsPacked) {\n    disposeIntermediateTensorInfoOrNull(backend2, xUnPacked);\n  }\n  const kPow2 = roundUpToPow2(k);\n  const lastDimPow2 = roundUpToPow2(lastDim);\n  let indices = null;\n  const getInputs = () => indices === null ? [x2D, x2D] : [x2D, indices];\n  const runSwap = (dir, inc, shape) => {\n    const inputs2 = getInputs();\n    const program = new SwapProgram(shape);\n    const fistPass = indices === null ? 1 : 0;\n    const customValues = [[lastDim], [fistPass], [Number.NEGATIVE_INFINITY], [dir], [inc]];\n    const prevIndices2 = indices;\n    indices = backend2.runWebGLProgram(program, inputs2, \"int32\", customValues);\n    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);\n  };\n  for (let len = 1; len < kPow2; len *= 2) {\n    const dir = len * 2;\n    for (let inc = len; inc >= 1; inc /= 2) {\n      runSwap(dir, inc, [batch, lastDimPow2]);\n    }\n  }\n  for (let indicesSize = lastDimPow2; indicesSize > kPow2; indicesSize /= 2) {\n    const inputs2 = getInputs();\n    const mergeProgram = new MergeProgram([batch, indicesSize / 2]);\n    const firstPass = indices === null ? 1 : 0;\n    const customValues = [[lastDim], [firstPass], [kPow2]];\n    const prevIndices2 = indices;\n    indices = backend2.runWebGLProgram(mergeProgram, inputs2, \"int32\", customValues);\n    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);\n    const len = kPow2 / 2;\n    const dir = len * 2;\n    for (let inc = len; inc >= 1; inc /= 2) {\n      runSwap(dir, inc, indices.shape);\n    }\n  }\n  let prevIndices = indices;\n  indices = slice3({ inputs: { x: indices }, backend: backend2, attrs: { begin: 0, size: [batch, k] } });\n  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);\n  let values = gatherV22({ inputs: { x: x2D, indices }, backend: backend2, attrs: { axis: 1, batchDims: 1 } });\n  disposeIntermediateTensorInfoOrNull(backend2, x2D);\n  const newShape = xShape.slice(0, -1);\n  newShape.push(k);\n  prevIndices = indices;\n  indices = reshape4({ inputs: { x: indices }, attrs: { shape: newShape }, backend: backend2 });\n  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);\n  const prevValues = values;\n  values = reshape4({ inputs: { x: values }, attrs: { shape: newShape }, backend: backend2 });\n  disposeIntermediateTensorInfoOrNull(backend2, prevValues);\n  return [values, indices];\n}\nvar topKConfig2 = {\n  kernelName: TopK,\n  backendName: \"webgl\",\n  kernelFunc: topK2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/transform_gpu.js\nvar TransformProgram = class {\n  constructor(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape) {\n    this.variableNames = [\"Image\", \"Transforms\"];\n    this.outputShape = outShape;\n    const interpolationModeId = interpolation === \"nearest\" ? 1 : 2;\n    let fillModeId;\n    switch (fillMode) {\n      case \"constant\":\n        fillModeId = 1;\n        break;\n      case \"reflect\":\n        fillModeId = 2;\n        break;\n      case \"wrap\":\n        fillModeId = 3;\n        break;\n      case \"nearest\":\n        fillModeId = 4;\n        break;\n      default:\n        fillModeId = 1;\n        break;\n    }\n    this.userCode = `\n            float mapCoord(float outCoord, float len) {\n              float inCoord = outCoord;\n              if(${fillModeId} == 2) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    if (inCoord < sz2) {\n                      inCoord = sz2 * float(int(float(-inCoord / sz2))) +\n                      inCoord;\n                    }\n                    inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    inCoord -= sz2 * float(int(float(inCoord / sz2)));\n                    if (inCoord >= len) {\n                      inCoord = sz2 - inCoord - 1.0;\n                    }\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${fillModeId} == 3) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord -= len * float(int(float(inCoord / sz)));\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${fillModeId} == 4) {\n                return clamp(outCoord, 0.0, len - 1.0);\n              } else {\n                return outCoord;\n              }\n            }\n\n            float readWithFillValue(int batch, int coordY, int coordX,\n              int channel) {\n              float outputValue;\n              if (0 <= coordY && coordY < ${imageHeight} && 0 <= coordX && coordX < ${imageWidth}) {\n                  outputValue = getImage(batch, coordY, coordX, channel);\n              } else {\n                outputValue = float(${fillValue});\n              }\n              return outputValue;\n            }\n\n            void main() {\n              ivec4 coords = getOutputCoords();\n              float outputValue;\n              int batch = coords[0];\n              int x = coords[2];\n              int y = coords[1];\n              int channel = coords[3];\n              float xf = float(x);\n              float yf = float(y);\n              float a1 = getTransforms(batch, 0);\n              float a2 = getTransforms(batch, 1);\n              float a3 = getTransforms(batch, 2);\n              float b1 = getTransforms(batch, 3);\n              float b2 = getTransforms(batch, 4);\n              float b3 = getTransforms(batch, 5);\n              float c1 = getTransforms(batch, 6);\n              float c2 = getTransforms(batch, 7);\n              float projection = c1 * xf + c2 * yf + 1.0;\n              if (projection == 0.0) {\n                outputValue = float(${fillValue});\n              } else {\n                float inX = (a1 * xf + a2 * yf + a3) / projection;\n                float inY = (b1 * xf + b2 * yf + b3) / projection;\n                float mapX = mapCoord(inX, float(${imageWidth}));\n                float mapY = mapCoord(inY, float(${imageHeight}));\n\n                if (${interpolationModeId} == 1) {\n                  int coordY = int(round(mapY));\n                  int coordX = int(round(mapX));\n                  outputValue = readWithFillValue(batch, coordY, coordX,\n                    channel);\n                } else {\n                  float yFloor = floor(mapY);\n                  float xFloor = floor(mapX);\n                  float yCeil = yFloor + 1.0;\n                  float xCeil = xFloor + 1.0;\n                  float valueYFloor = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yFloor), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yFloor), int(xCeil), channel);\n                  float valueYCeil = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yCeil), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yCeil), int(xCeil), channel);\n                  outputValue = (yCeil - mapY) * valueYFloor +\n                  (mapY - yFloor) * valueYCeil;\n                }\n              }\n              setOutput(outputValue);\n            }\n        `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transform.js\nfunction transform3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { image: image2, transforms } = inputs;\n  const { interpolation, fillMode, fillValue, outputShape } = attrs;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];\n  const outShape = [\n    batch,\n    outHeight,\n    outWidth,\n    numChannels\n  ];\n  const program = new TransformProgram(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape);\n  return backend2.runWebGLProgram(program, [image2, transforms], \"float32\");\n}\nvar transformConfig2 = {\n  kernelName: Transform,\n  backendName: \"webgl\",\n  kernelFunc: transform3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Unique.js\nfunction unique4(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { axis } = attrs;\n  const { x } = inputs;\n  assertNotComplex2(x, \"unique\");\n  console.warn(\"WARNING: \", \"UI might be locked temporarily as data is being downloaded\");\n  const values = backend2.readSync(x.dataId);\n  const { outputValues, outputShape, indices } = uniqueImplCPU(values, axis, x.shape, x.dtype);\n  return [\n    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),\n    backend2.makeTensorInfo([indices.length], \"int32\", indices)\n  ];\n}\nvar uniqueConfig2 = {\n  kernelName: Unique,\n  backendName: \"webgl\",\n  kernelFunc: unique4\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Unpack.js\nfunction unpack2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { value } = inputs;\n  let { axis } = attrs;\n  if (axis < 0) {\n    axis += value.shape.length;\n  }\n  const x = value;\n  const xRank = x.shape.length;\n  const num = value.shape[axis];\n  const outShape = new Array(xRank - 1);\n  let outIndex = 0;\n  for (let i = 0; i < xRank; i++) {\n    if (i !== axis) {\n      outShape[outIndex++] = x.shape[i];\n    }\n  }\n  const toDispose = [];\n  const begin = new Array(xRank).fill(0);\n  const size = x.shape.slice();\n  size[axis] = 1;\n  const res = new Array(num);\n  for (let i = 0; i < res.length; i++) {\n    begin[axis] = i;\n    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size } });\n    const reshaped = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: outShape } });\n    res[i] = reshaped;\n    toDispose.push(sliced);\n  }\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return res;\n}\nvar unpackConfig2 = {\n  kernelName: Unpack,\n  backendName: \"webgl\",\n  kernelFunc: unpack2\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/segment_gpu.js\nvar SegmentOpProgram = class {\n  constructor(segOpInfo, segOpType) {\n    this.variableNames = [\"x\", \"segmentIds\"];\n    const windowSize = segOpInfo.windowSize;\n    const batchSize = segOpInfo.batchSize;\n    const inSize = segOpInfo.inSize;\n    const numSegments = segOpInfo.numSegments;\n    const outSize = numSegments * Math.ceil(inSize / windowSize);\n    this.outputShape = [batchSize, outSize];\n    const initializationValue = \"0.0\";\n    const returnValue = `sumValue`;\n    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;\n    const windowSizeVec4Remainder = windowSize % 4;\n    const updateSnippet = `\n        sumValue += dot(values, segFilter);\n    `;\n    let checkValueOutOfBounds = \"\";\n    if (inSize % windowSize > 0) {\n      checkValueOutOfBounds = `\n        if (inIdx < 0 || inIdx >= ${inSize}) {\n          return initializationValue;\n        }\n      `;\n    }\n    let checkSegmentIdOutOfBounds = \"\";\n    if (inSize % windowSize > 0) {\n      checkSegmentIdOutOfBounds = `\n        if (inIdx < 0 || inIdx >= ${inSize}) {\n          return -1.0;\n        }\n      `;\n    }\n    this.userCode = `\n      const float initializationValue = ${initializationValue};\n\n      float getValue(int batch, int inIdx) {\n        ${checkValueOutOfBounds}\n        return getX(batch, inIdx);\n      }\n\n      float getSegmentIdAtIndex(int inIdx) {\n        ${checkSegmentIdOutOfBounds}\n        return getSegmentIds(inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = int(floor(float(outIdx) / float(\n          ${numSegments})) * float(${windowSize}));\n        int currentSeg = int(mod(float(outIdx), float(${numSegments})));\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0\n          );\n\n          ${updateSnippet}\n        }\n\n        int inIdx = inOffset + ${windowSizeNearestVec4};\n        if (${windowSizeVec4Remainder === 1}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          int inIdxSeg = int(getSegmentIdAtIndex(inIdx));\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            0,\n            0,\n            0\n          );\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 2}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n              0,\n              0\n          );\n\n          ${updateSnippet}\n        } else if (${windowSizeVec4Remainder === 3}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            0\n          );\n\n          ${updateSnippet}\n        }\n        setOutput(${returnValue});\n      }\n    `;\n  }\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/UnsortedSegmentSum.js\nfunction unsortedSegmentSum3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x, segmentIds } = inputs;\n  const { numSegments } = attrs;\n  const xRank = x.shape.length;\n  const toDispose = [];\n  let axis = 0;\n  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);\n  let permutedX = x;\n  if (permutation != null) {\n    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });\n    toDispose.push(permutedX);\n    axis = backend_util_exports.getInnerMostAxes(1, xRank)[0];\n  }\n  const outShape = backend_util_exports.segment_util.computeOutShape(permutedX.shape, axis, numSegments);\n  const inSize = util_exports.sizeFromShape([permutedX.shape[axis]]);\n  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });\n  toDispose.push(a2D);\n  const outputDType = sumOutType(x.dtype);\n  const segOpCompute = (x2, segOpType, segmentIds2, dtype, numSegments2) => {\n    const batchSize = x2.shape[0];\n    const inSize2 = x2.shape[1];\n    const windowSize = backend_util_exports.segment_util.segOpComputeOptimalWindowSize(inSize2, numSegments2);\n    const segOpInfo = { windowSize, inSize: inSize2, batchSize, numSegments: numSegments2 };\n    const program = new SegmentOpProgram(segOpInfo, segOpType);\n    const output = backend2.compileAndRun(program, [x2, segmentIds2], dtype);\n    toDispose.push(output);\n    if (output.shape[1] === numSegments2) {\n      return output;\n    }\n    const rangeInfo = range4({\n      backend: backend2,\n      attrs: { start: 0, stop: numSegments2, step: 1, dtype: \"float32\" }\n    });\n    const tileInfo = tile4({\n      inputs: { x: rangeInfo },\n      backend: backend2,\n      attrs: { reps: [inSize2 / windowSize] }\n    });\n    toDispose.push(rangeInfo);\n    toDispose.push(tileInfo);\n    const result2 = segOpCompute(output, segOpType, tileInfo, dtype, numSegments2);\n    return result2;\n  };\n  const segOpResult = segOpCompute(a2D, \"unsortedSegmentSum\", segmentIds, outputDType, numSegments);\n  const reshaped = reshape4({ inputs: { x: segOpResult }, backend: backend2, attrs: { shape: outShape } });\n  let result = reshaped;\n  if (permutation != null) {\n    toDispose.push(reshaped);\n    const perm = backend_util_exports.getUndoAxesPermutation(permutation);\n    result = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm } });\n  }\n  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));\n  return result;\n}\nvar unsortedSegmentSumConfig2 = {\n  kernelName: UnsortedSegmentSum,\n  backendName: \"webgl\",\n  kernelFunc: unsortedSegmentSum3\n};\n\n// node_modules/@tensorflow/tfjs-backend-webgl/dist/register_all_kernels.js\nvar kernelConfigs2 = [\n  _fusedMatMulConfig2,\n  absConfig2,\n  acosConfig2,\n  acoshConfig2,\n  addConfig2,\n  addNConfig2,\n  allConfig2,\n  anyConfig2,\n  argMaxConfig2,\n  argMinConfig2,\n  asinConfig2,\n  asinhConfig2,\n  atanConfig2,\n  atan2Config2,\n  atanhConfig2,\n  avgPoolConfig2,\n  avgPool3DConfig2,\n  avgPool3DGradConfig3,\n  avgPoolGradConfig3,\n  batchMatMulConfig2,\n  batchNormConfig2,\n  batchToSpaceNDConfig2,\n  bincountConfig2,\n  broadcastArgsConfig2,\n  castConfig2,\n  ceilConfig2,\n  clipByValueConfig2,\n  complexConfig2,\n  complexAbsConfig2,\n  concatConfig2,\n  conv2DConfig2,\n  conv2DBackpropFilterConfig2,\n  conv2DBackpropInputConfig2,\n  conv3DConfig2,\n  conv3DBackpropFilterV2Config2,\n  conv3DBackpropInputConfig,\n  cosConfig2,\n  coshConfig2,\n  cropAndResizeConfig2,\n  cumprodConfig2,\n  cumsumConfig2,\n  denseBincountConfig2,\n  depthToSpaceConfig2,\n  depthwiseConv2dNativeConfig2,\n  depthwiseConv2dNativeBackpropFilterConfig2,\n  depthwiseConv2dNativeBackpropInputConfig2,\n  diagConfig2,\n  dilation2DConfig2,\n  einsumConfig2,\n  eluConfig2,\n  eluGradConfig3,\n  equalConfig2,\n  erfConfig2,\n  expConfig2,\n  expandDimsConfig2,\n  expm1Config2,\n  fftConfig2,\n  fillConfig2,\n  flipLeftRightConfig2,\n  floorConfig2,\n  floorDivConfig2,\n  fromPixelsConfig,\n  fusedConv2DConfig2,\n  fusedDepthwiseConv2DConfig2,\n  gatherNdConfig2,\n  gatherV2Config2,\n  greaterConfig2,\n  greaterEqualConfig2,\n  identityConfig2,\n  ifftConfig2,\n  imagConfig2,\n  isFiniteConfig2,\n  isInfConfig2,\n  isNaNConfig2,\n  leakyReluConfig2,\n  lessConfig2,\n  lessEqualConfig2,\n  linSpaceConfig2,\n  logConfig2,\n  log1pConfig2,\n  logicalAndConfig2,\n  logicalNotConfig2,\n  logicalOrConfig2,\n  LRNConfig2,\n  LRNGradConfig2,\n  maxConfig2,\n  maximumConfig2,\n  maxPoolConfig2,\n  maxPool3DConfig2,\n  maxPool3DGradConfig3,\n  maxPoolGradConfig3,\n  maxPoolWithArgmaxConfig2,\n  meanConfig2,\n  minConfig2,\n  minimumConfig2,\n  mirrorPadConfig2,\n  modConfig2,\n  multinomialConfig2,\n  multiplyConfig2,\n  negConfig2,\n  nonMaxSuppressionV3Config2,\n  nonMaxSuppressionV4Config2,\n  nonMaxSuppressionV5Config2,\n  notEqualConfig2,\n  oneHotConfig2,\n  onesLikeConfig2,\n  packConfig2,\n  padV2Config2,\n  powConfig2,\n  preluConfig2,\n  prodConfig2,\n  raggedGatherConfig2,\n  raggedRangeConfig2,\n  raggedTensorToTensorConfig2,\n  rangeConfig2,\n  realConfig2,\n  realDivConfig2,\n  reciprocalConfig2,\n  reluConfig2,\n  relu6Config2,\n  reshapeConfig2,\n  resizeBilinearConfig2,\n  resizeBilinearGradConfig3,\n  resizeNearestNeighborConfig2,\n  resizeNearestNeighborGradConfig3,\n  reverseConfig2,\n  rotateWithOffsetConfig2,\n  roundConfig2,\n  rsqrtConfig2,\n  scatterNdConfig2,\n  searchSortedConfig2,\n  selectConfig2,\n  seluConfig2,\n  sigmoidConfig2,\n  signConfig2,\n  sinConfig2,\n  sinhConfig2,\n  sliceConfig2,\n  softmaxConfig2,\n  softplusConfig2,\n  spaceToBatchNDConfig2,\n  sparseFillEmptyRowsConfig2,\n  sparseReshapeConfig2,\n  sparseSegmentMeanConfig2,\n  sparseSegmentSumConfig2,\n  sparseToDenseConfig2,\n  splitVConfig2,\n  sqrtConfig2,\n  squareConfig2,\n  squaredDifferenceConfig2,\n  stepConfig2,\n  stridedSliceConfig2,\n  stringNGramsConfig2,\n  stringSplitConfig2,\n  stringToHashBucketFastConfig2,\n  subConfig2,\n  sumConfig2,\n  tanConfig2,\n  tanhConfig2,\n  tileConfig2,\n  topKConfig2,\n  transformConfig2,\n  transposeConfig2,\n  uniqueConfig2,\n  unpackConfig2,\n  unsortedSegmentSumConfig2,\n  zerosLikeConfig2\n];\nfor (const kernelConfig of kernelConfigs2) {\n  registerKernel(kernelConfig);\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/types.js\nvar CppDType;\n(function(CppDType2) {\n  CppDType2[CppDType2[\"float32\"] = 0] = \"float32\";\n  CppDType2[CppDType2[\"int32\"] = 1] = \"int32\";\n  CppDType2[CppDType2[\"bool\"] = 2] = \"bool\";\n  CppDType2[CppDType2[\"string\"] = 3] = \"string\";\n  CppDType2[CppDType2[\"complex64\"] = 4] = \"complex64\";\n})(CppDType || (CppDType = {}));\nvar FusableActivation;\n(function(FusableActivation2) {\n  FusableActivation2[FusableActivation2[\"linear\"] = 0] = \"linear\";\n  FusableActivation2[FusableActivation2[\"relu\"] = 1] = \"relu\";\n  FusableActivation2[FusableActivation2[\"relu6\"] = 2] = \"relu6\";\n  FusableActivation2[FusableActivation2[\"prelu\"] = 3] = \"prelu\";\n  FusableActivation2[FusableActivation2[\"leakyrelu\"] = 4] = \"leakyrelu\";\n  FusableActivation2[FusableActivation2[\"sigmoid\"] = 5] = \"sigmoid\";\n  FusableActivation2[FusableActivation2[\"elu\"] = 6] = \"elu\";\n})(FusableActivation || (FusableActivation = {}));\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/_FusedMatMul.js\nvar wasmFusedMatMul;\nfunction setup(backend2) {\n  wasmFusedMatMul = backend2.wasm.cwrap(_FusedMatMul, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction fusedBatchMatMul(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b, bias, preluActivationWeights } = inputs;\n  if (a.dtype !== \"float32\" || b.dtype !== \"float32\") {\n    throw new Error(`_FusedMatMul for non non-float32 tensors not yet supported.`);\n  }\n  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;\n  const aId = backend2.dataIdMap.get(a.dataId).id;\n  const bId = backend2.dataIdMap.get(b.dataId).id;\n  let biasId = 0;\n  if (bias != null) {\n    const biasData = backend2.dataIdMap.get(bias.dataId);\n    if (biasData.shape.length !== 1) {\n      throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${biasData.shape.length}.`);\n    }\n    biasId = biasData.id;\n  }\n  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;\n  const fusedActivation = FusableActivation[activation2];\n  if (fusedActivation == null) {\n    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);\n  }\n  const leftDim = transposeA ? a.shape[2] : a.shape[1];\n  const rightDim = transposeB ? b.shape[1] : b.shape[2];\n  const batchDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));\n  const out = backend2.makeOutput([...batchDims, leftDim, rightDim], a.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);\n  const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);\n  wasmFusedMatMul(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, transposeA, transposeB, fusedActivation, biasId, preluActivationWeightsId, leakyreluAlpha || 0, outId);\n  return out;\n}\nvar _fusedMatMulConfig3 = {\n  kernelName: _FusedMatMul,\n  backendName: \"wasm\",\n  setupFunc: setup,\n  kernelFunc: fusedBatchMatMul\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/unary_kernel.js\nfunction createUnaryKernelConfig(kernelName, outType) {\n  let wasmFunc9;\n  function setupFunc3(backend2) {\n    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [\n      \"number\",\n      \"number\",\n      \"number\"\n    ]);\n  }\n  function kernelFunc3(args) {\n    const { backend: backend2, inputs: { x } } = args;\n    const xId = backend2.dataIdMap.get(x.dataId).id;\n    const out = backend2.makeOutput(x.shape, outType || x.dtype);\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    if (util_exports.sizeFromShape(out.shape) === 0) {\n      return out;\n    }\n    wasmFunc9(xId, CppDType[x.dtype], outId);\n    return out;\n  }\n  return { kernelName, backendName: \"wasm\", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Abs.js\nvar absConfig3 = createUnaryKernelConfig(Abs);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/binary_kernel.js\nfunction createBinaryKernelConfig(kernelName, supportsFullBroadcast19, dtype) {\n  let wasmFunc9;\n  function setupFunc3(backend2) {\n    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [\n      \"number\",\n      \"array\",\n      \"number\",\n      \"number\",\n      \"array\",\n      \"number\",\n      \"number\",\n      \"number\"\n    ]);\n  }\n  function kernelFunc3(args) {\n    const { backend: backend2, inputs } = args;\n    const { a, b } = inputs;\n    const aId = backend2.dataIdMap.get(a.dataId).id;\n    const bId = backend2.dataIdMap.get(b.dataId).id;\n    const outputType = dtype != null ? dtype : a.dtype;\n    const newShape = backend_util_exports.assertAndGetBroadcastShape(a.shape, b.shape);\n    const out = backend2.makeOutput(newShape, outputType);\n    if (util_exports.sizeFromShape(newShape) === 0) {\n      return out;\n    }\n    const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);\n    const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    const kernelFunc4 = () => wasmFunc9(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, CppDType[a.dtype], outId);\n    kernelFunc4();\n    return out;\n  }\n  return { kernelName, backendName: \"wasm\", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Add.js\nvar supportsFullBroadcast = true;\nvar addConfig3 = createBinaryKernelConfig(Add, supportsFullBroadcast);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/AddN.js\nvar wasmFunc;\nfunction setupFunc(backend2) {\n  wasmFunc = backend2.wasm.cwrap(AddN, null, [\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction addn(args) {\n  const { inputs, backend: backend2 } = args;\n  const out = backend2.makeOutput(inputs[0].shape, inputs[0].dtype);\n  if (util_exports.sizeFromShape(out.shape) === 0) {\n    return out;\n  }\n  const inputIds = inputs.map((x) => backend2.dataIdMap.get(x.dataId).id);\n  const inputIdsBytes = new Uint8Array(new Int32Array(inputIds).buffer);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmFunc(inputIdsBytes, inputIds.length, CppDType[out.dtype], outId);\n  return out;\n}\nvar addNConfig3 = {\n  kernelName: AddN,\n  backendName: \"wasm\",\n  setupFunc,\n  kernelFunc: addn\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Identity.js\nfunction identity4(args) {\n  const { inputs: { x }, backend: backend2 } = args;\n  if (x.dtype === \"string\") {\n    return tensor(backend2.readSync(x.dataId), x.shape, x.dtype);\n  }\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const inVals = backend2.typedArrayFromHeap(x);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.set(inVals);\n  return out;\n}\nvar identityConfig3 = {\n  kernelName: Identity,\n  backendName: \"wasm\",\n  kernelFunc: identity4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Transpose.js\nvar wasmTranspose;\nfunction setup2(backend2) {\n  wasmTranspose = backend2.wasm.cwrap(Transpose, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\"\n  ]);\n}\nfunction transpose4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const [reducedShape, perm] = removeOneSizeDims(inputs.x.shape, attrs.perm);\n  let permIsNoOp = true;\n  for (let i = 0; i < perm.length; i++) {\n    if (perm[i] !== i) {\n      permIsNoOp = false;\n    }\n  }\n  const outShape = computeOutShape4(inputs.x.shape, attrs.perm);\n  const x = {\n    dataId: inputs.x.dataId,\n    shape: reducedShape,\n    dtype: inputs.x.dtype\n  };\n  if (permIsNoOp) {\n    const cloned = identity4({ inputs, backend: backend2 });\n    cloned.shape = outShape;\n    return cloned;\n  }\n  const out = backend2.makeOutput(outShape, x.dtype);\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const permBytes = new Uint8Array(new Int32Array(perm).buffer);\n  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  wasmTranspose(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], outId, permBytes, perm.length);\n  return out;\n}\nfunction computeOutShape4(inShape, perm) {\n  const outShape = new Array(inShape.length);\n  for (let i = 0; i < outShape.length; i++) {\n    outShape[i] = inShape[perm[i]];\n  }\n  return outShape;\n}\nfunction removeOneSizeDims(shape, perm) {\n  const newShape = [];\n  const newPerm = [];\n  for (let i = 0; i < shape.length; ++i) {\n    if (shape[i] !== 1) {\n      newShape.push(shape[i]);\n    }\n    if (shape[perm[i]] !== 1) {\n      newPerm.push(perm[i]);\n    }\n  }\n  for (let i = 0; i < newPerm.length; ++i) {\n    let minValIdx = -1;\n    for (let j = 0; j < newPerm.length; ++j) {\n      if (newPerm[j] >= i && (minValIdx === -1 || newPerm[minValIdx] > newPerm[j])) {\n        minValIdx = j;\n      }\n    }\n    newPerm[minValIdx] = i;\n  }\n  return [newShape, newPerm];\n}\nvar transposeConfig3 = {\n  kernelName: Transpose,\n  backendName: \"wasm\",\n  kernelFunc: transpose4,\n  setupFunc: setup2\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/kernel_utils.js\nfunction permuteAxesAndTranspose(x, axis, backend2) {\n  const xShape = x.shape;\n  const xRank = x.shape.length;\n  const originalAxes = util_exports.parseAxisParam(axis, xShape);\n  let axes = originalAxes;\n  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);\n  let xTransposed = null;\n  let inputWasTransposed = false;\n  if (permutedAxes != null) {\n    const newShape = new Array(xRank);\n    for (let i = 0; i < newShape.length; i++) {\n      newShape[i] = xShape[permutedAxes[i]];\n    }\n    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);\n    xTransposed = transpose4({ inputs: { x }, attrs: { perm: permutedAxes }, backend: backend2 });\n    const xId = backend2.dataIdMap.get(x.dataId).id;\n    const transposedId = backend2.dataIdMap.get(xTransposed.dataId).id;\n    if (transposedId !== xId) {\n      inputWasTransposed = true;\n    }\n  }\n  return { transposed: xTransposed, originalAxes, axes, inputWasTransposed };\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/All.js\nvar wasmAll;\nfunction setup3(backend2) {\n  wasmAll = backend2.wasm.cwrap(All, null, [\"number, number, number\"]);\n}\nfunction all4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    input2 = transposed;\n    inputId = transposedId;\n  }\n  const inputRank = input2.shape.length;\n  backend_util_exports.assertAxesAreInnerMostDims(\"all\", axes, inputRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, x.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmAll(inputId, reduceSize, outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar allConfig3 = {\n  kernelName: All,\n  backendName: \"wasm\",\n  setupFunc: setup3,\n  kernelFunc: all4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Any.js\nvar wasmAny;\nfunction setup4(backend2) {\n  wasmAny = backend2.wasm.cwrap(Any, null, [\"number, number, number\"]);\n}\nfunction any4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    input2 = transposed;\n    inputId = transposedId;\n  }\n  const inputRank = input2.shape.length;\n  backend_util_exports.assertAxesAreInnerMostDims(\"any\", axes, inputRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, x.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmAny(inputId, reduceSize, outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar anyConfig3 = {\n  kernelName: Any,\n  backendName: \"wasm\",\n  setupFunc: setup4,\n  kernelFunc: any4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ArgMax.js\nvar wasmFunc2;\nfunction setup5(backend2) {\n  wasmFunc2 = backend2.wasm.cwrap(ArgMax, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction argmax(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    if (transposedId !== xId) {\n      input2 = transposed;\n      inputId = transposedId;\n    }\n  }\n  const outShape = input2.shape.slice(0, -1);\n  const out = backend2.makeOutput(outShape, \"int32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const outerSize = util_exports.sizeFromShape(out.shape);\n  const innerSize = input2.shape[axes[0]];\n  wasmFunc2(inputId, CppDType[input2.dtype], outerSize, innerSize, outId);\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  return out;\n}\nvar argMaxConfig3 = {\n  kernelName: ArgMax,\n  backendName: \"wasm\",\n  kernelFunc: argmax,\n  setupFunc: setup5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/AvgPool.js\nvar wasmAvgPool;\nfunction setup6(backend2) {\n  wasmAvgPool = backend2.wasm.cwrap(AvgPool, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction avgPool4(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const x = inputs.x;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const channels = convInfo.inChannels;\n  if (convInfo.dataFormat !== \"channelsLast\") {\n    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);\n  }\n  if (convInfo.dilationWidth !== 1 || convInfo.dilationHeight !== 1) {\n    throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${convInfo.dilationHeight}, ${convInfo.dilationWidth}].`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmAvgPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, strideHeight, strideWidth, channels, outId);\n  return out;\n}\nvar avgPoolConfig3 = {\n  kernelName: AvgPool,\n  backendName: \"wasm\",\n  setupFunc: setup6,\n  kernelFunc: avgPool4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Reshape.js\nfunction reshape5(args) {\n  const { inputs, attrs } = args;\n  const { x } = inputs;\n  const { shape } = attrs;\n  const xSize = util_exports.sizeFromShape(x.shape);\n  const $shape = util_exports.inferFromImplicitShape(shape, xSize);\n  util_exports.assert(xSize === util_exports.sizeFromShape($shape), () => `new shape: ${$shape}, old shape: ${x.shape}. New shape and old shape must have the same number of elements.`);\n  args.backend.incRef(x.dataId);\n  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };\n}\nvar reshapeConfig3 = {\n  kernelName: Reshape,\n  backendName: \"wasm\",\n  kernelFunc: reshape5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/BatchMatMul.js\nvar wasmBatchMatMul;\nfunction setup7(backend2) {\n  wasmBatchMatMul = backend2.wasm.cwrap(BatchMatMul, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction batchMatMul3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { a, b } = inputs;\n  const { transposeA, transposeB } = attrs;\n  if (a.dtype !== \"float32\" || b.dtype !== \"float32\") {\n    throw new Error(`BatchMatMul for non non-float32 tensors not yet supported.`);\n  }\n  const aRank = a.shape.length;\n  const bRank = b.shape.length;\n  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];\n  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];\n  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];\n  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];\n  const outerDimsA = a.shape.slice(0, -2);\n  const outerDimsB = b.shape.slice(0, -2);\n  const batchDimA = util_exports.sizeFromShape(outerDimsA);\n  const batchDimB = util_exports.sizeFromShape(outerDimsB);\n  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));\n  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);\n  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);\n  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];\n  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];\n  const a3d = reshape5({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });\n  const b3d = reshape5({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });\n  const a3dId = backend2.dataIdMap.get(a3d.dataId).id;\n  const b3dId = backend2.dataIdMap.get(b3d.dataId).id;\n  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];\n  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];\n  const batchDim = Math.max(batchDimA, batchDimB);\n  const out = backend2.makeOutput([batchDim, leftDim, rightDim], a3d.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const aShapeBytes = new Uint8Array(new Int32Array(a3d.shape).buffer);\n  const bShapeBytes = new Uint8Array(new Int32Array(b3d.shape).buffer);\n  wasmBatchMatMul(a3dId, aShapeBytes, a3d.shape.length, b3dId, bShapeBytes, b3d.shape.length, transposeA, transposeB, outId);\n  backend2.disposeData(a3d.dataId);\n  backend2.disposeData(b3d.dataId);\n  out.shape = outShape;\n  return out;\n}\nvar batchMatMulConfig3 = {\n  kernelName: BatchMatMul,\n  backendName: \"wasm\",\n  setupFunc: setup7,\n  kernelFunc: batchMatMul3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Slice.js\nfunction slice4(args) {\n  const { inputs: { x }, attrs: { begin, size }, backend: backend2 } = args;\n  const [begin_, size_] = slice_util_exports.parseSliceParams(x, begin, size);\n  const isContinous = slice_util_exports.isSliceContinous(x.shape, begin_, size_);\n  const xVals = backend2.readSync(x.dataId);\n  const out = backend2.makeOutput(size_, x.dtype);\n  const xStrides = util_exports.computeStrides(x.shape);\n  const outData = backend2.dataIdMap.get(out.dataId);\n  if (isContinous) {\n    const flatOffset = slice_util_exports.computeFlatOffset(begin_, xStrides);\n    if (x.dtype === \"string\") {\n      outData.stringBytes = xVals.slice(flatOffset, flatOffset + util_exports.sizeFromShape(size_));\n    } else {\n      const outVals2 = backend2.typedArrayFromHeap(out);\n      outVals2.set(xVals.subarray(flatOffset, flatOffset + util_exports.sizeFromShape(size_)));\n    }\n    return out;\n  }\n  if (x.dtype === \"string\") {\n    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);\n    outData.stringBytes = res;\n    return out;\n  }\n  const outVals = backend2.typedArrayFromHeap(out);\n  const rank = x.shape.length;\n  if (rank === 2) {\n    slice2d2(xVals, xStrides[0], outVals, begin_, size_);\n  } else if (rank === 3) {\n    slice3d2(xVals, xStrides[0], xStrides[1], outVals, begin_, size_);\n  } else if (rank === 4) {\n    slice4d2(xVals, xStrides[0], xStrides[1], xStrides[2], outVals, begin_, size_);\n  } else {\n    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);\n    outVals.set(res);\n  }\n  return out;\n}\nfunction slice2d2(xVals, xStride, outVals, begin, size) {\n  let outOffset = 0;\n  const beginI = begin[0];\n  const beginJ = begin[1];\n  const endI = beginI + size[0];\n  for (let i = beginI; i < endI; i++) {\n    const xOffset = i * xStride + beginJ;\n    outVals.set(xVals.subarray(xOffset, xOffset + size[1]), outOffset);\n    outOffset += size[1];\n  }\n}\nfunction slice3d2(xVals, xStride1, xStride2, outVals, begin, size) {\n  let outOffset = 0;\n  const beginI = begin[0];\n  const beginJ = begin[1];\n  const beginK = begin[2];\n  const endI = beginI + size[0];\n  const endJ = beginJ + size[1];\n  for (let i = beginI; i < endI; i++) {\n    for (let j = beginJ; j < endJ; j++) {\n      const xOffset = i * xStride1 + j * xStride2 + beginK;\n      outVals.set(xVals.subarray(xOffset, xOffset + size[2]), outOffset);\n      outOffset += size[2];\n    }\n  }\n}\nfunction slice4d2(xVals, xStride1, xStride2, xStride3, outVals, begin, size) {\n  let outOffset = 0;\n  const beginI = begin[0];\n  const beginJ = begin[1];\n  const beginK = begin[2];\n  const endI = beginI + size[0];\n  const endJ = beginJ + size[1];\n  const endK = beginK + size[2];\n  const beginL = begin[3];\n  for (let i = beginI; i < endI; i++) {\n    for (let j = beginJ; j < endJ; j++) {\n      for (let k = beginK; k < endK; k++) {\n        const xOffset = i * xStride1 + j * xStride2 + k * xStride3 + beginL;\n        outVals.set(xVals.subarray(xOffset, xOffset + size[3]), outOffset);\n        outOffset += size[3];\n      }\n    }\n  }\n}\nvar sliceConfig3 = {\n  kernelName: Slice,\n  backendName: \"wasm\",\n  kernelFunc: slice4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/BatchToSpaceND.js\nfunction batchToSpaceND4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, crops } = attrs;\n  const prod5 = blockShape.reduce((a, b) => a * b);\n  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);\n  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);\n  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);\n  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);\n  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);\n  const xReshaped = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });\n  const xTransposed = transpose4({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });\n  const xTransposedReshaped = reshape5({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });\n  const result = slice4({\n    inputs: { x: xTransposedReshaped },\n    backend: backend2,\n    attrs: { begin: sliceBeginCoords, size: sliceSize }\n  });\n  backend2.disposeData(xReshaped.dataId);\n  backend2.disposeData(xTransposed.dataId);\n  backend2.disposeData(xReshaped.dataId);\n  return result;\n}\nvar batchToSpaceNDConfig3 = {\n  kernelName: BatchToSpaceND,\n  backendName: \"wasm\",\n  kernelFunc: batchToSpaceND4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cast.js\nfunction cast5(args) {\n  const { inputs: { x }, attrs: { dtype }, backend: backend2 } = args;\n  const out = backend2.makeOutput(x.shape, dtype);\n  const inVals = backend2.typedArrayFromHeap(x);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.set(inVals);\n  return out;\n}\nvar castConfig3 = {\n  kernelName: Cast,\n  backendName: \"wasm\",\n  kernelFunc: cast5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Ceil.js\nvar ceilConfig3 = createUnaryKernelConfig(Ceil);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ClipByValue.js\nvar wasmClip;\nfunction setup8(backend2) {\n  wasmClip = backend2.wasm.cwrap(ClipByValue, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction clip(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { clipValueMin, clipValueMax } = attrs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmClip(xId, clipValueMin, clipValueMax, outId);\n  return out;\n}\nvar clipByValueConfig3 = {\n  kernelName: ClipByValue,\n  backendName: \"wasm\",\n  setupFunc: setup8,\n  kernelFunc: clip\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Concat.js\nfunction concat4(args) {\n  const { inputs, backend: backend2 } = args;\n  const axis = util_exports.parseAxisParam(args.attrs.axis, inputs[0].shape)[0];\n  const shapes = inputs.map((t) => t.shape);\n  backend_util_exports.assertParamsConsistent(shapes, axis);\n  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);\n  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);\n  if ($inputs.length === 1) {\n    return identity4({ inputs: { x: $inputs[0] }, backend: backend2 });\n  }\n  const out = backend2.makeOutput(outShape, inputs[0].dtype);\n  if (util_exports.sizeFromShape(outShape) === 0) {\n    return out;\n  }\n  if ($inputs[0].dtype === \"string\") {\n    const inputs2D = $inputs.map((t) => {\n      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));\n      const shape = [-1, innerSize];\n      return reshape5({ inputs: { x: t }, backend: backend2, attrs: { shape } });\n    });\n    const inputsValShapes = inputs2D.map((t) => {\n      return { vals: backend2.readSync(t.dataId), shape: t.shape };\n    });\n    outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);\n    const simplyConcat = inputs2D[0].shape[0] === 1;\n    const outVals2 = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);\n    const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), axis);\n    out.shape = finalOutShape;\n    const outData = backend2.dataIdMap.get(out.dataId);\n    outData.stringBytes = backend_util_exports.fromStringArrayToUint8(outVals2);\n    inputs2D.forEach((t) => backend2.disposeData(t.dataId));\n    return out;\n  }\n  const batchDim = util_exports.sizeFromShape($inputs[0].shape.slice(0, axis));\n  let sumInnerDims = 0;\n  const innerDims = $inputs.map((input2) => {\n    const innerDim = util_exports.sizeFromShape(input2.shape.slice(axis));\n    sumInnerDims += innerDim;\n    return innerDim;\n  });\n  const inVals = $inputs.map((input2) => backend2.typedArrayFromHeap(input2));\n  const outVals = backend2.typedArrayFromHeap(out);\n  for (let b = 0; b < batchDim; b++) {\n    let outOffset = b * sumInnerDims;\n    for (let i = 0; i < inVals.length; i++) {\n      const innerDim = innerDims[i];\n      const inOffset = b * innerDim;\n      const vals = inVals[i].subarray(inOffset, inOffset + innerDim);\n      outVals.set(vals, outOffset);\n      outOffset += innerDim;\n    }\n  }\n  return out;\n}\nvar concatConfig3 = {\n  kernelName: Concat,\n  backendName: \"wasm\",\n  kernelFunc: concat4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Conv2D.js\nvar wasmConv2d;\nfunction setup9(backend2) {\n  wasmConv2d = backend2.wasm.cwrap(Conv2D, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction conv2d5(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x, filter } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const filterId = backend2.dataIdMap.get(filter.dataId).id;\n  const { strides, dilations, pad: pad3, dimRoundingMode, dataFormat } = attrs;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const inputChannels = convInfo.inChannels;\n  const outputChannels = convInfo.outChannels;\n  const isSamePad = convInfo.padInfo.type === \"SAME\" ? 1 : 0;\n  if (convInfo.dataFormat !== \"channelsLast\") {\n    throw new Error(`wasm backend Conv2D does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);\n  return out;\n}\nvar conv2DConfig3 = {\n  kernelName: Conv2D,\n  backendName: \"wasm\",\n  setupFunc: setup9,\n  kernelFunc: conv2d5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Conv2DBackpropInput.js\nvar wasmConv2DBackpropInput;\nfunction setup10(backend2) {\n  wasmConv2DBackpropInput = backend2.wasm.cwrap(Conv2DBackpropInput, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction conv2DBackpropInput4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { dy, filter } = inputs;\n  const { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape } = attrs;\n  const dilations = 1;\n  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);\n  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);\n  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;\n  const topPad = filterHeight - 1 - convInfo.padInfo.top;\n  const leftPad = filterWidth - 1 - convInfo.padInfo.left;\n  const isChannelsLast = convInfo.dataFormat === \"channelsLast\";\n  const dxStrides = util_exports.computeStrides(convInfo.inShape);\n  const dyStrides = util_exports.computeStrides(dy.shape);\n  const [fltS0, fltS1, fltS2] = util_exports.computeStrides(filter.shape);\n  const xBatchStride = dxStrides[0];\n  const xRowStride = isChannelsLast ? dxStrides[1] : dxStrides[2];\n  const xColStride = isChannelsLast ? dxStrides[2] : 1;\n  const xChannelStride = isChannelsLast ? 1 : dxStrides[1];\n  const yBatchStride = dyStrides[0];\n  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];\n  const yColStride = isChannelsLast ? dyStrides[2] : 1;\n  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];\n  const out = backend2.makeOutput(convInfo.inShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const dyId = backend2.dataIdMap.get(dy.dataId).id;\n  const filterId = backend2.dataIdMap.get(filter.dataId).id;\n  wasmConv2DBackpropInput(dyId, filterId, batchSize, filterHeight, filterWidth, inHeight, inWidth, inChannels, outHeight, outWidth, outChannels, strideHeight, strideWidth, topPad, leftPad, fltS0, fltS1, fltS2, xBatchStride, xRowStride, xColStride, xChannelStride, yBatchStride, yRowStride, yColStride, yChannelStride, outId);\n  return out;\n}\nvar conv2DBackpropInputConfig3 = {\n  kernelName: Conv2DBackpropInput,\n  backendName: \"wasm\",\n  setupFunc: setup10,\n  kernelFunc: conv2DBackpropInput4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cos.js\nvar cosConfig3 = createUnaryKernelConfig(Cos);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cosh.js\nvar coshConfig3 = createUnaryKernelConfig(Cosh);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/CropAndResize.js\nvar InterpolationMethod;\n(function(InterpolationMethod2) {\n  InterpolationMethod2[InterpolationMethod2[\"bilinear\"] = 0] = \"bilinear\";\n  InterpolationMethod2[InterpolationMethod2[\"nearest\"] = 1] = \"nearest\";\n})(InterpolationMethod || (InterpolationMethod = {}));\nvar wasmCropAndResize;\nfunction setup11(backend2) {\n  wasmCropAndResize = backend2.wasm.cwrap(CropAndResize, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction cropAndResize4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { method, extrapolationValue, cropSize } = attrs;\n  const { image: image2, boxes, boxInd } = inputs;\n  const numBoxes = boxes.shape[0];\n  const [cropHeight, cropWidth] = cropSize;\n  const outShape = [numBoxes, cropHeight, cropWidth, image2.shape[3]];\n  let imagesData = backend2.dataIdMap.get(image2.dataId);\n  let castedData;\n  if (image2.dtype !== \"float32\") {\n    castedData = cast5({ backend: backend2, inputs: { x: image2 }, attrs: { dtype: \"float32\" } });\n    imagesData = backend2.dataIdMap.get(castedData.dataId);\n  }\n  const imagesId = imagesData.id;\n  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;\n  const boxIndId = backend2.dataIdMap.get(boxInd.dataId).id;\n  const out = backend2.makeOutput(outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const imagesShapeBytes = new Uint8Array(new Int32Array(image2.shape).buffer);\n  wasmCropAndResize(imagesId, boxesId, boxIndId, numBoxes, imagesShapeBytes, cropHeight, cropWidth, InterpolationMethod[method], extrapolationValue, outId);\n  if (castedData != null) {\n    backend2.disposeData(castedData.dataId);\n  }\n  return out;\n}\nvar cropAndResizeConfig3 = {\n  kernelName: CropAndResize,\n  backendName: \"wasm\",\n  setupFunc: setup11,\n  kernelFunc: cropAndResize4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cumprod.js\nvar wasmCumprod;\nfunction setup12(backend2) {\n  wasmCumprod = backend2.wasm.cwrap(Cumprod, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction cumprod4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  const xRank = x.shape.length;\n  util_exports.assert(x.dtype === \"float32\" || x.dtype === \"int32\", () => `cumprod does not support ${x.dtype} tensors in the WASM backend`);\n  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);\n  let permutedX = x;\n  if (permutation !== null) {\n    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });\n  }\n  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];\n  backend_util_exports.assertAxesAreInnerMostDims(\"cumprod\", [permutedAxis], xRank);\n  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);\n  const finalDim = permutedX.shape[permutedAxis];\n  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;\n  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;\n  wasmCumprod(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);\n  let out = permutedOut;\n  if (permutation !== null) {\n    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);\n    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });\n    backend2.disposeData(permutedX.dataId);\n    backend2.disposeData(permutedOut.dataId);\n  }\n  return out;\n}\nvar cumprodConfig3 = {\n  kernelName: Cumprod,\n  backendName: \"wasm\",\n  setupFunc: setup12,\n  kernelFunc: cumprod4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cumsum.js\nvar wasmCumsum;\nfunction setup13(backend2) {\n  wasmCumsum = backend2.wasm.cwrap(Cumsum, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction cumsum4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { axis, exclusive, reverse: reverse5 } = attrs;\n  const xRank = x.shape.length;\n  util_exports.assert(x.dtype === \"float32\" || x.dtype === \"int32\", () => `cumsum does not support ${x.dtype} tensors in the WASM backend`);\n  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);\n  let permutedX = x;\n  if (permutation !== null) {\n    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });\n  }\n  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];\n  backend_util_exports.assertAxesAreInnerMostDims(\"cumsum\", [permutedAxis], xRank);\n  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);\n  const finalDim = permutedX.shape[permutedAxis];\n  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;\n  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;\n  wasmCumsum(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);\n  let out = permutedOut;\n  if (permutation !== null) {\n    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);\n    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });\n    backend2.disposeData(permutedX.dataId);\n    backend2.disposeData(permutedOut.dataId);\n  }\n  return out;\n}\nvar cumsumConfig3 = {\n  kernelName: Cumsum,\n  backendName: \"wasm\",\n  setupFunc: setup13,\n  kernelFunc: cumsum4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/DepthToSpace.js\nvar wasmDepthToSpace;\nfunction setup14(backend2) {\n  wasmDepthToSpace = backend2.wasm.cwrap(DepthToSpace, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"array\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction depthToSpace4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { x } = inputs;\n  const { blockSize, dataFormat } = attrs;\n  const batchSize = x.shape[0];\n  const inputHeight = dataFormat === \"NHWC\" ? x.shape[1] : x.shape[2];\n  const inputWidth = dataFormat === \"NHWC\" ? x.shape[2] : x.shape[3];\n  const inputDepth = dataFormat === \"NHWC\" ? x.shape[3] : x.shape[1];\n  const outputHeight = inputHeight * blockSize;\n  const outputWidth = inputWidth * blockSize;\n  const outputDepth = inputDepth / (blockSize * blockSize);\n  const outputShape = dataFormat === \"NHWC\" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];\n  const out = backend2.makeOutput(outputShape, \"float32\");\n  const xData = backend2.dataIdMap.get(x.dataId);\n  const xId = xData.id;\n  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);\n  const outputShapeBytes = new Uint8Array(new Int32Array(outputShape).buffer);\n  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(outputShape)).buffer);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const channelsLast = dataFormat === \"NHWC\" ? 1 : 0;\n  wasmDepthToSpace(xId, blockSize, channelsLast, xStridesBytes, x.shape.length - 1, outputShapeBytes, outStridesBytes, outputShape.length, outId);\n  return out;\n}\nvar depthToSpaceConfig3 = {\n  kernelName: DepthToSpace,\n  backendName: \"wasm\",\n  setupFunc: setup14,\n  kernelFunc: depthToSpace4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/DepthwiseConv2dNative.js\nvar wasmDepthwiseConv2d;\nfunction setup15(backend2) {\n  wasmDepthwiseConv2d = backend2.wasm.cwrap(DepthwiseConv2dNative, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction depthwiseConv2d5(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x, filter } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const filterId = backend2.dataIdMap.get(filter.dataId).id;\n  const { strides, dilations, pad: pad3, dimRoundingMode } = attrs;\n  const $dilations = dilations == null ? [1, 1] : dilations;\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const inputChannels = convInfo.inChannels;\n  const outputChannels = convInfo.outChannels;\n  const isSamePad = convInfo.padInfo.type === \"SAME\" ? 1 : 0;\n  if (convInfo.dataFormat !== \"channelsLast\") {\n    throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmDepthwiseConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);\n  return out;\n}\nvar depthwiseConv2dNativeConfig3 = {\n  kernelName: DepthwiseConv2dNative,\n  backendName: \"wasm\",\n  setupFunc: setup15,\n  kernelFunc: depthwiseConv2d5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Elu.js\nvar eluConfig3 = createUnaryKernelConfig(Elu);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Equal.js\nvar supportsFullBroadcast2 = false;\nvar equalConfig3 = createBinaryKernelConfig(Equal, supportsFullBroadcast2, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Exp.js\nvar expConfig3 = createUnaryKernelConfig(Exp, \"float32\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ExpandDims.js\nfunction expandDims5(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { input: input2 } = inputs;\n  const { dim } = attrs;\n  const inputRank = input2.shape.length;\n  const newShape = input2.shape.slice();\n  let $dim = dim;\n  if (dim < 0) {\n    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);\n    $dim = inputRank + dim + 1;\n  }\n  newShape.splice($dim, 0, 1);\n  return reshape5({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });\n}\nvar expandDimsConfig3 = {\n  kernelName: ExpandDims,\n  backendName: \"wasm\",\n  kernelFunc: expandDims5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Fill.js\nfunction fill4(args) {\n  const { attrs: { shape, value, dtype }, backend: backend2 } = args;\n  const out = backend2.makeOutput(shape, dtype);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.fill(value);\n  return out;\n}\nvar fillConfig3 = {\n  kernelName: Fill,\n  backendName: \"wasm\",\n  kernelFunc: fill4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FlipLeftRight.js\nvar wasmFlipLeftRight;\nfunction setup16(backend2) {\n  wasmFlipLeftRight = backend2.wasm.cwrap(FlipLeftRight, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction flipLeftRight2(args) {\n  const { inputs, backend: backend2 } = args;\n  const { image: image2 } = inputs;\n  const out = backend2.makeOutput(image2.shape, image2.dtype);\n  const imageId = backend2.dataIdMap.get(image2.dataId).id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  wasmFlipLeftRight(imageId, batch, imageHeight, imageWidth, numChannels, outId);\n  return out;\n}\nvar flipLeftRightConfig3 = {\n  kernelName: FlipLeftRight,\n  backendName: \"wasm\",\n  kernelFunc: flipLeftRight2,\n  setupFunc: setup16\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Floor.js\nvar floorConfig3 = createUnaryKernelConfig(Floor);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FloorDiv.js\nvar supportsFullBroadcast3 = false;\nvar floorDivConfig3 = createBinaryKernelConfig(FloorDiv, supportsFullBroadcast3);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedBatchNorm.js\nvar wasmBatchNorm;\nfunction setup17(backend2) {\n  wasmBatchNorm = backend2.wasm.cwrap(FusedBatchNorm, null, [\"number\", \"number\", \"number\", \"number\", \"number\", \"number\", \"number\"]);\n}\nfunction fusedBatchNorm(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { varianceEpsilon } = attrs;\n  const { x, mean: mean4, variance, offset, scale: scale2 } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const meanId = backend2.dataIdMap.get(mean4.dataId).id;\n  const varianceId = backend2.dataIdMap.get(variance.dataId).id;\n  const offsetId = offset != null ? backend2.dataIdMap.get(offset.dataId).id : 0;\n  const scaleId = scale2 != null ? backend2.dataIdMap.get(scale2.dataId).id : 0;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  if (util_exports.sizeFromShape(x.shape) === 0) {\n    return out;\n  }\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmBatchNorm(xId, meanId, varianceId, offsetId, scaleId, varianceEpsilon, outId);\n  return out;\n}\nvar fusedBatchNormConfig = {\n  kernelName: FusedBatchNorm,\n  backendName: \"wasm\",\n  setupFunc: setup17,\n  kernelFunc: fusedBatchNorm\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedConv2D.js\nvar wasmFusedConv2d;\nfunction setup18(backend2) {\n  wasmFusedConv2d = backend2.wasm.cwrap(FusedConv2D, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction fusedConv2d2(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode);\n  const fusedActivation = FusableActivation[activation2];\n  if (fusedActivation == null) {\n    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);\n  }\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const filterId = backend2.dataIdMap.get(filter.dataId).id;\n  const outputChannels = convInfo.outChannels;\n  let biasId = 0;\n  if (bias != null) {\n    const biasData = backend2.dataIdMap.get(bias.dataId);\n    if (biasData.shape.length !== 1) {\n      throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);\n    }\n    if (biasData.shape[0] !== outputChannels) {\n      throw new Error(`FusedConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);\n    }\n    biasId = biasData.id;\n  }\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const inputChannels = convInfo.inChannels;\n  const isSamePad = convInfo.padInfo.type === \"SAME\" ? 1 : 0;\n  const batchSize = convInfo.batchSize;\n  const inHeight = convInfo.inHeight;\n  const inWidth = convInfo.inWidth;\n  if (dataFormat !== \"NHWC\") {\n    throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;\n  wasmFusedConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);\n  return out;\n}\nvar fusedConv2DConfig3 = {\n  kernelName: FusedConv2D,\n  backendName: \"wasm\",\n  setupFunc: setup18,\n  kernelFunc: fusedConv2d2\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedDepthwiseConv2D.js\nvar wasmFusedDepthwiseConv2d;\nfunction setup19(backend2) {\n  wasmFusedDepthwiseConv2d = backend2.wasm.cwrap(FusedDepthwiseConv2D, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction fusedDepthwiseConv2d(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x, filter, bias, preluActivationWeights } = inputs;\n  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;\n  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);\n  const fusedActivation = FusableActivation[activation2];\n  if (fusedActivation == null) {\n    throw new Error(`${activation2} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);\n  }\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const filterId = backend2.dataIdMap.get(filter.dataId).id;\n  const outputChannels = convInfo.outChannels;\n  let biasId = 0;\n  if (bias != null) {\n    const biasData = backend2.dataIdMap.get(bias.dataId);\n    if (biasData.shape.length !== 1) {\n      throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);\n    }\n    if (biasData.shape[0] !== outputChannels) {\n      throw new Error(`FusedDepthwiseConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);\n    }\n    biasId = biasData.id;\n  }\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const inputChannels = convInfo.inChannels;\n  const isSamePad = convInfo.padInfo.type === \"SAME\" ? 1 : 0;\n  const batchSize = convInfo.batchSize;\n  const inHeight = convInfo.inHeight;\n  const inWidth = convInfo.inWidth;\n  if (dataFormat !== \"NHWC\") {\n    throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;\n  wasmFusedDepthwiseConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);\n  return out;\n}\nvar fusedDepthwiseConv2DConfig3 = {\n  kernelName: FusedDepthwiseConv2D,\n  backendName: \"wasm\",\n  setupFunc: setup19,\n  kernelFunc: fusedDepthwiseConv2d\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GatherNd.js\nvar wasmGatherNd;\nfunction setup20(backend2) {\n  wasmGatherNd = backend2.wasm.cwrap(GatherNd, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\"\n  ]);\n}\nfunction gatherNd3(args) {\n  const { backend: backend2, inputs } = args;\n  const { params, indices } = inputs;\n  const [resultShape, numSlices, sliceSize, strides] = gather_nd_util_exports.prepareAndValidate(params, indices);\n  const out = backend2.makeOutput(resultShape, params.dtype);\n  if (numSlices === 0) {\n    return out;\n  }\n  const indicesShape = indices.shape;\n  const sliceRank = indicesShape[indicesShape.length - 1];\n  const xData = backend2.dataIdMap.get(params.dataId);\n  const xId = xData.id;\n  const indicesData = backend2.dataIdMap.get(indices.dataId);\n  const indicesId = indicesData.id;\n  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmGatherNd(xId, CppDType[params.dtype], indicesId, numSlices, sliceRank, sliceSize, stridesBytes, outId);\n  return out;\n}\nvar gatherNdConfig3 = {\n  kernelName: GatherNd,\n  backendName: \"wasm\",\n  setupFunc: setup20,\n  kernelFunc: gatherNd3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GatherV2.js\nvar wasmGather;\nfunction setup21(backend2) {\n  wasmGather = backend2.wasm.cwrap(\"Gather\", null, [\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\"\n  ]);\n}\nfunction gatherV23(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { x, indices } = inputs;\n  const { axis, batchDims } = attrs;\n  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];\n  const indicesVals = backend2.readSync(indices.dataId);\n  const axisDim = x.shape[parsedAxis];\n  for (let i = 0; i < indicesVals.length; ++i) {\n    const index = indicesVals[i];\n    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);\n  }\n  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);\n  const flattenX = reshape5({\n    inputs: { x },\n    attrs: {\n      shape: [\n        shapeInfo.batchSize,\n        shapeInfo.outerSize,\n        shapeInfo.dimSize,\n        shapeInfo.sliceSize\n      ]\n    },\n    backend: backend2\n  });\n  const indicesSize = util_exports.sizeFromShape(indices.shape);\n  const flattenIndex = reshape5({\n    inputs: { x: indices },\n    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] },\n    backend: backend2\n  });\n  const flattenOutputShape = [\n    shapeInfo.batchSize,\n    shapeInfo.outerSize,\n    indicesSize / shapeInfo.batchSize,\n    shapeInfo.sliceSize\n  ];\n  const out = backend2.makeOutput(flattenOutputShape, x.dtype);\n  if (util_exports.sizeFromShape(x.shape) === 0) {\n    return out;\n  }\n  const stridesSize = flattenX.shape.length - 1;\n  const xData = backend2.dataIdMap.get(flattenX.dataId);\n  const xId = xData.id;\n  const indicesData = backend2.dataIdMap.get(flattenIndex.dataId);\n  const indicesId = indicesData.id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenX.shape)).buffer);\n  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenOutputShape)).buffer);\n  wasmGather(xId, CppDType[x.dtype], xStridesBytes, stridesSize, indicesId, shapeInfo.batchSize, outStridesBytes, outId);\n  backend2.disposeData(flattenX.dataId);\n  backend2.disposeData(flattenIndex.dataId);\n  out.shape = shapeInfo.outputShape;\n  return out;\n}\nvar gatherV2Config3 = {\n  kernelName: GatherV2,\n  backendName: \"wasm\",\n  setupFunc: setup21,\n  kernelFunc: gatherV23\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Greater.js\nvar supportsFullBroadcast4 = false;\nvar greaterConfig3 = createBinaryKernelConfig(Greater, supportsFullBroadcast4, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GreaterEqual.js\nvar supportsFullBroadcast5 = false;\nvar greaterEqualConfig3 = createBinaryKernelConfig(GreaterEqual, supportsFullBroadcast5, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LeakyRelu.js\nvar wasmFunc3;\nfunction setupFunc2(backend2) {\n  wasmFunc3 = backend2.wasm.cwrap(LeakyRelu, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction leakyRelu4(args) {\n  const { inputs: { x }, attrs: { alpha }, backend: backend2 } = args;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(x.shape, \"float32\");\n  if (util_exports.sizeFromShape(x.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmFunc3(xId, CppDType[x.dtype], alpha, outId);\n  }\n  return out;\n}\nvar leakyReluConfig3 = {\n  kernelName: LeakyRelu,\n  backendName: \"wasm\",\n  setupFunc: setupFunc2,\n  kernelFunc: leakyRelu4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Less.js\nvar supportsFullBroadcast6 = false;\nvar lessConfig3 = createBinaryKernelConfig(Less, supportsFullBroadcast6, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LessEqual.js\nvar supportsFullBroadcast7 = false;\nvar lessEqualConfig3 = createBinaryKernelConfig(LessEqual, supportsFullBroadcast7, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Log.js\nvar logConfig3 = createUnaryKernelConfig(Log);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalAnd.js\nvar supportsFullBroadcast8 = false;\nvar logicalAndConfig3 = createBinaryKernelConfig(LogicalAnd, supportsFullBroadcast8, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalNot.js\nvar logicalNotConfig3 = createUnaryKernelConfig(LogicalNot);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalOr.js\nvar supportsFullBroadcast9 = false;\nvar logicalOrConfig3 = createBinaryKernelConfig(LogicalOr, supportsFullBroadcast9, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalXor.js\nvar supportsFullBroadcast10 = false;\nvar logicalXorConfig = createBinaryKernelConfig(LogicalXor, supportsFullBroadcast10, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Max.js\nvar wasmMax;\nfunction setup22(backend2) {\n  wasmMax = backend2.wasm.cwrap(Max, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction max5(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { reductionIndices: axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    input2 = transposed;\n    inputId = transposedId;\n  }\n  const inputRank = input2.shape.length;\n  backend_util_exports.assertAxesAreInnerMostDims(\"max\", axes, inputRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, x.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmMax(inputId, CppDType[x.dtype], reduceSize, outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar maxConfig3 = {\n  kernelName: Max,\n  backendName: \"wasm\",\n  setupFunc: setup22,\n  kernelFunc: max5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Maximum.js\nvar supportsFullBroadcast11 = false;\nvar maximumConfig3 = createBinaryKernelConfig(Maximum, supportsFullBroadcast11);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/MaxPool.js\nvar wasmMaxPool;\nfunction setup23(backend2) {\n  wasmMaxPool = backend2.wasm.cwrap(MaxPool, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction maxPool4(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const x = inputs.x;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  util_exports.assert(x.dtype === \"float32\", () => `Error in MaxPool: only float32 input is supported. Got ${x.dtype}.`);\n  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;\n  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);\n  const filterHeight = convInfo.filterHeight;\n  const filterWidth = convInfo.filterWidth;\n  const padTop = convInfo.padInfo.top;\n  const padRight = convInfo.padInfo.right;\n  const padBottom = convInfo.padInfo.bottom;\n  const padLeft = convInfo.padInfo.left;\n  const dilationHeight = convInfo.dilationHeight;\n  const dilationWidth = convInfo.dilationWidth;\n  const strideHeight = convInfo.strideHeight;\n  const strideWidth = convInfo.strideWidth;\n  const inputChannels = convInfo.inChannels;\n  const outputChannels = convInfo.outChannels;\n  if (convInfo.dataFormat !== \"channelsLast\") {\n    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);\n  }\n  const out = backend2.makeOutput(convInfo.outShape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmMaxPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);\n  return out;\n}\nvar maxPoolConfig3 = {\n  kernelName: MaxPool,\n  backendName: \"wasm\",\n  setupFunc: setup23,\n  kernelFunc: maxPool4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Mean.js\nvar wasmMean;\nfunction setup24(backend2) {\n  wasmMean = backend2.wasm.cwrap(Mean, null, [\"number, number, number\"]);\n}\nfunction mean3(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  let reductionAxes = axes;\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    if (transposedId !== xId) {\n      input2 = transposed;\n      inputId = transposedId;\n      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);\n    }\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"mean\", reductionAxes, input2.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  let castedInput = input2;\n  if (input2.dtype !== \"float32\") {\n    castedInput = cast5({ backend: backend2, inputs: { x: input2 }, attrs: { dtype: \"float32\" } });\n    inputId = backend2.dataIdMap.get(castedInput.dataId).id;\n  }\n  const out = backend2.makeOutput(outShape, \"float32\");\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmMean(inputId, reduceSize, outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  if (input2.dtype !== \"float32\") {\n    backend2.disposeData(castedInput.dataId);\n  }\n  return out;\n}\nvar meanConfig3 = {\n  kernelName: Mean,\n  backendName: \"wasm\",\n  setupFunc: setup24,\n  kernelFunc: mean3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Min.js\nvar wasmMin;\nfunction setup25(backend2) {\n  wasmMin = backend2.wasm.cwrap(Min, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction min5(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    if (transposedId !== xId) {\n      input2 = transposed;\n      inputId = transposedId;\n    }\n  }\n  const inputRank = input2.shape.length;\n  backend_util_exports.assertAxesAreInnerMostDims(\"min\", axes, inputRank);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, input2.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmMin(inputId, CppDType[x.dtype], reduceSize, outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar minConfig3 = {\n  kernelName: Min,\n  backendName: \"wasm\",\n  setupFunc: setup25,\n  kernelFunc: min5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Minimum.js\nvar supportsFullBroadcast12 = false;\nvar minimumConfig3 = createBinaryKernelConfig(Minimum, supportsFullBroadcast12);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/MirrorPad.js\nvar MirrorPaddingMode;\n(function(MirrorPaddingMode2) {\n  MirrorPaddingMode2[MirrorPaddingMode2[\"reflect\"] = 0] = \"reflect\";\n  MirrorPaddingMode2[MirrorPaddingMode2[\"symmetric\"] = 1] = \"symmetric\";\n})(MirrorPaddingMode || (MirrorPaddingMode = {}));\nvar wasmMirrorPad;\nfunction setup26(backend2) {\n  wasmMirrorPad = backend2.wasm.cwrap(MirrorPad, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction mirrorPad3(args) {\n  const { inputs: { x }, backend: backend2, attrs: { paddings, mode } } = args;\n  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(outShape, x.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);\n  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);\n  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);\n  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);\n  wasmMirrorPad(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, MirrorPaddingMode[mode], outId);\n  return out;\n}\nvar mirrorPadConfig3 = {\n  kernelName: MirrorPad,\n  backendName: \"wasm\",\n  kernelFunc: mirrorPad3,\n  setupFunc: setup26\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Multiply.js\nvar supportsFullBroadcast13 = true;\nvar multiplyConfig3 = createBinaryKernelConfig(Multiply, supportsFullBroadcast13);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Neg.js\nvar negConfig3 = createUnaryKernelConfig(Neg);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppression_util.js\nfunction parseResultStruct(backend2, resOffset) {\n  const result = new Int32Array(backend2.wasm.HEAPU8.buffer, resOffset, 4);\n  const pSelectedIndices = result[0];\n  const selectedSize = result[1];\n  const pSelectedScores = result[2];\n  const pValidOutputs = result[3];\n  backend2.wasm._free(resOffset);\n  return { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs };\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV3.js\nvar wasmFunc4;\nfunction setup27(backend2) {\n  wasmFunc4 = backend2.wasm.cwrap(\n    NonMaxSuppressionV3,\n    \"number\",\n    [\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\"\n    ]\n  );\n}\nfunction kernelFunc(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { iouThreshold, maxOutputSize, scoreThreshold } = attrs;\n  const { boxes, scores } = inputs;\n  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;\n  const scoresId = backend2.dataIdMap.get(scores.dataId).id;\n  const resOffset = wasmFunc4(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold);\n  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);\n  backend2.wasm._free(pSelectedScores);\n  backend2.wasm._free(pValidOutputs);\n  const selectedIndicesTensor = backend2.makeOutput([selectedSize], \"int32\", pSelectedIndices);\n  return selectedIndicesTensor;\n}\nvar nonMaxSuppressionV3Config3 = {\n  kernelName: NonMaxSuppressionV3,\n  backendName: \"wasm\",\n  setupFunc: setup27,\n  kernelFunc\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV4.js\nvar wasmFunc5;\nfunction setup28(backend2) {\n  wasmFunc5 = backend2.wasm.cwrap(\n    NonMaxSuppressionV4,\n    \"number\",\n    [\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\",\n      \"bool\"\n    ]\n  );\n}\nfunction nonMaxSuppressionV43(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { iouThreshold, maxOutputSize, scoreThreshold, padToMaxOutputSize } = attrs;\n  const { boxes, scores } = inputs;\n  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;\n  const scoresId = backend2.dataIdMap.get(scores.dataId).id;\n  const resOffset = wasmFunc5(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);\n  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);\n  backend2.wasm._free(pSelectedScores);\n  const selectedIndicesTensor = backend2.makeOutput([selectedSize], \"int32\", pSelectedIndices);\n  const validOutputsTensor = backend2.makeOutput([], \"int32\", pValidOutputs);\n  return [selectedIndicesTensor, validOutputsTensor];\n}\nvar nonMaxSuppressionV4Config3 = {\n  kernelName: NonMaxSuppressionV4,\n  backendName: \"wasm\",\n  setupFunc: setup28,\n  kernelFunc: nonMaxSuppressionV43\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV5.js\nvar wasmFunc6;\nfunction setup29(backend2) {\n  wasmFunc6 = backend2.wasm.cwrap(\n    NonMaxSuppressionV5,\n    \"number\",\n    [\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\",\n      \"number\"\n    ]\n  );\n}\nfunction kernelFunc2(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { iouThreshold, maxOutputSize, scoreThreshold, softNmsSigma } = attrs;\n  const { boxes, scores } = inputs;\n  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;\n  const scoresId = backend2.dataIdMap.get(scores.dataId).id;\n  const resOffset = wasmFunc6(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);\n  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);\n  backend2.wasm._free(pValidOutputs);\n  const selectedIndicesTensor = backend2.makeOutput([selectedSize], \"int32\", pSelectedIndices);\n  const selectedScoresTensor = backend2.makeOutput([selectedSize], \"float32\", pSelectedScores);\n  return [selectedIndicesTensor, selectedScoresTensor];\n}\nvar nonMaxSuppressionV5Config3 = {\n  kernelName: NonMaxSuppressionV5,\n  backendName: \"wasm\",\n  setupFunc: setup29,\n  kernelFunc: kernelFunc2\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NotEqual.js\nvar supportsFullBroadcast14 = false;\nvar notEqualConfig3 = createBinaryKernelConfig(NotEqual, supportsFullBroadcast14, \"bool\");\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/OneHot.js\nvar wasmOneHot;\nfunction setup30(backend2) {\n  wasmOneHot = backend2.wasm.cwrap(OneHot, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction oneHot4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { indices } = inputs;\n  const { dtype, depth, onValue, offValue } = attrs;\n  const out = backend2.makeOutput([...indices.shape, depth], dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const indicesData = backend2.dataIdMap.get(indices.dataId);\n  const indicesId = indicesData.id;\n  wasmOneHot(indicesId, depth, onValue, offValue, outId);\n  return out;\n}\nvar oneHotConfig3 = {\n  kernelName: OneHot,\n  backendName: \"wasm\",\n  setupFunc: setup30,\n  kernelFunc: oneHot4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/OnesLike.js\nfunction onesLike4(args) {\n  const { inputs: { x }, backend: backend2 } = args;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.fill(1);\n  return out;\n}\nvar onesLikeConfig3 = {\n  kernelName: OnesLike,\n  backendName: \"wasm\",\n  kernelFunc: onesLike4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Pack.js\nfunction pack3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { axis } = attrs;\n  if (inputs.length === 1) {\n    return expandDims5({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });\n  }\n  const shape = inputs[0].shape;\n  const dtype = inputs[0].dtype;\n  inputs.forEach((t) => {\n    util_exports.assertShapesMatch(shape, t.shape, \"All tensors passed to stack must have matching shapes\");\n    util_exports.assert(dtype === t.dtype, () => \"All tensors passed to stack must have matching dtypes\");\n  });\n  const intermediateTensorInfos = [];\n  const expandedTensors = inputs.map((t) => {\n    const expandedT = expandDims5({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });\n    intermediateTensorInfos.push(expandedT);\n    return expandedT;\n  });\n  const result = concat4({ inputs: expandedTensors, backend: backend2, attrs: { axis } });\n  intermediateTensorInfos.forEach((t) => backend2.disposeData(t.dataId));\n  return result;\n}\nvar packConfig3 = {\n  kernelName: Pack,\n  backendName: \"wasm\",\n  kernelFunc: pack3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/PadV2.js\nvar wasmPadV2;\nfunction setup31(backend2) {\n  wasmPadV2 = backend2.wasm.cwrap(PadV2, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction pad2(args) {\n  const { inputs: { x }, backend: backend2, attrs: { paddings, constantValue } } = args;\n  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);\n  if (util_exports.sizeFromShape(x.shape) === 0) {\n    return fill4({\n      backend: backend2,\n      attrs: { shape: outShape, value: constantValue, dtype: x.dtype }\n    });\n  }\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(outShape, x.dtype);\n  const outTensorData = backend2.dataIdMap.get(out.dataId);\n  const outId = outTensorData.id;\n  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);\n  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);\n  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);\n  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);\n  wasmPadV2(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, constantValue, outId);\n  return out;\n}\nvar padV2Config3 = {\n  kernelName: PadV2,\n  backendName: \"wasm\",\n  kernelFunc: pad2,\n  setupFunc: setup31\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Pow.js\nvar supportsFullBroadcast15 = false;\nvar powConfig3 = createBinaryKernelConfig(Pow, supportsFullBroadcast15);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Prelu.js\nvar wasmPrelu;\nfunction setup32(backend2) {\n  wasmPrelu = backend2.wasm.cwrap(Prelu, null, [\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction prelu5(args) {\n  const { inputs, backend: backend2 } = args;\n  const { x, alpha } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const weightsId = backend2.dataIdMap.get(alpha.dataId).id;\n  let inputId = xId;\n  const input2 = x;\n  let castedInput = input2;\n  if (input2.dtype !== \"float32\") {\n    castedInput = cast5({ backend: backend2, inputs: { x }, attrs: { dtype: \"float32\" } });\n    inputId = backend2.dataIdMap.get(castedInput.dataId).id;\n  }\n  const out = backend2.makeOutput(x.shape, \"float32\");\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmPrelu(inputId, weightsId, outId);\n  if (input2.dtype !== \"float32\") {\n    backend2.disposeData(castedInput.dataId);\n  }\n  return out;\n}\nvar preluConfig3 = {\n  kernelName: Prelu,\n  backendName: \"wasm\",\n  setupFunc: setup32,\n  kernelFunc: prelu5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Prod.js\nvar wasmProd;\nfunction setup33(backend2) {\n  wasmProd = backend2.wasm.cwrap(Prod, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction prod4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  let reductionAxes = axes;\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    if (transposedId !== xId) {\n      input2 = transposed;\n      inputId = transposedId;\n      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);\n    }\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"prod\", reductionAxes, input2.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, input2.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmProd(inputId, reduceSize, CppDType[out.dtype], outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar prodConfig3 = {\n  kernelName: Prod,\n  backendName: \"wasm\",\n  setupFunc: setup33,\n  kernelFunc: prod4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Range.js\nvar range5 = (args) => {\n  const { backend: backend2, attrs } = args;\n  const { start, stop, step: step5, dtype } = attrs;\n  const values = rangeImpl(start, stop, step5, dtype);\n  const out = backend2.makeOutput([values.length], dtype);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.set(values);\n  return out;\n};\nvar rangeConfig3 = {\n  kernelName: Range,\n  backendName: \"wasm\",\n  kernelFunc: range5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/RealDiv.js\nvar supportsFullBroadcast16 = true;\nvar realDivConfig3 = createBinaryKernelConfig(RealDiv, supportsFullBroadcast16);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Relu.js\nvar reluConfig3 = createUnaryKernelConfig(Relu);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Relu6.js\nvar relu6Config3 = createUnaryKernelConfig(Relu6);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ResizeBilinear.js\nvar wasmResizeBilinear;\nfunction setup34(backend2) {\n  wasmResizeBilinear = backend2.wasm.cwrap(ResizeBilinear, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction resizeBilinear4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  const [newHeight, newWidth] = size;\n  const [batch, oldHeight, oldWidth, numChannels] = images.shape;\n  const outShape = [batch, newHeight, newWidth, numChannels];\n  let xData = backend2.dataIdMap.get(images.dataId);\n  let castedData;\n  if (xData.dtype !== \"float32\") {\n    castedData = cast5({ backend: backend2, inputs: { x: images }, attrs: { dtype: \"float32\" } });\n    xData = backend2.dataIdMap.get(castedData.dataId);\n  }\n  const xId = xData.id;\n  const out = backend2.makeOutput(outShape, \"float32\");\n  if (util_exports.sizeFromShape(images.shape) === 0) {\n    return out;\n  }\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmResizeBilinear(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);\n  if (castedData != null) {\n    backend2.disposeData(castedData.dataId);\n  }\n  return out;\n}\nvar resizeBilinearConfig3 = {\n  kernelName: ResizeBilinear,\n  backendName: \"wasm\",\n  setupFunc: setup34,\n  kernelFunc: resizeBilinear4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ResizeNearestNeighbor.js\nvar wasmResizeNearestNeighbor;\nfunction setup35(backend2) {\n  wasmResizeNearestNeighbor = backend2.wasm.cwrap(ResizeNearestNeighbor, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction resizeNearestNeighbor4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { images } = inputs;\n  const { alignCorners, halfPixelCenters, size } = attrs;\n  const [newHeight, newWidth] = size;\n  const [batch, oldHeight, oldWidth, numChannels] = images.shape;\n  const outShape = [batch, newHeight, newWidth, numChannels];\n  const out = backend2.makeOutput(outShape, \"float32\");\n  if (util_exports.sizeFromShape(images.shape) === 0) {\n    return out;\n  }\n  let xData = backend2.dataIdMap.get(images.dataId);\n  let castedData;\n  if (xData.dtype !== \"float32\") {\n    castedData = cast5({\n      backend: backend2,\n      inputs: { x: images },\n      attrs: { dtype: \"float32\" }\n    });\n    xData = backend2.dataIdMap.get(castedData.dataId);\n  }\n  const xId = xData.id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmResizeNearestNeighbor(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);\n  if (castedData != null) {\n    backend2.disposeData(castedData.dataId);\n  }\n  return out;\n}\nvar resizeNearestNeighborConfig3 = {\n  kernelName: ResizeNearestNeighbor,\n  backendName: \"wasm\",\n  setupFunc: setup35,\n  kernelFunc: resizeNearestNeighbor4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Reverse.js\nvar wasmReverse;\nfunction setup36(backend2) {\n  wasmReverse = backend2.wasm.cwrap(Reverse, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction reverse4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { dims } = attrs;\n  const axes = util_exports.parseAxisParam(dims, x.shape);\n  if (x.shape.length === 0) {\n    return identity4({ inputs: { x }, backend: backend2 });\n  }\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const axesBytes = new Uint8Array(new Int32Array(axes).buffer);\n  const outShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  wasmReverse(xId, axesBytes, axes.length, outShapeBytes, x.shape.length, outId);\n  const reshaped = reshape5({ inputs: { x: out }, attrs: { shape: x.shape }, backend: backend2 });\n  backend2.disposeData(out.dataId);\n  return reshaped;\n}\nvar reverseConfig3 = {\n  kernelName: Reverse,\n  backendName: \"wasm\",\n  kernelFunc: reverse4,\n  setupFunc: setup36\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/RotateWithOffset.js\nvar wasmRotate;\nfunction setup37(backend2) {\n  wasmRotate = backend2.wasm.cwrap(RotateWithOffset, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction rotateWithOffset2(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { image: image2 } = inputs;\n  const { radians, fillValue, center } = attrs;\n  const out = backend2.makeOutput(image2.shape, image2.dtype);\n  const imageId = backend2.dataIdMap.get(image2.dataId).id;\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);\n  const fillIsBlack = fillValue === 0;\n  const fullOpacityValue = 255;\n  const fillValues2 = typeof fillValue === \"number\" ? [fillValue, fillValue, fillValue, fillIsBlack ? 0 : fullOpacityValue] : [...fillValue, fullOpacityValue];\n  const fillBytes = new Uint8Array(new Int32Array(fillValues2).buffer);\n  wasmRotate(imageId, batch, imageHeight, imageWidth, numChannels, radians, centerX, centerY, fillBytes, fillValues2.length, outId);\n  return out;\n}\nvar rotateWithOffsetConfig3 = {\n  kernelName: RotateWithOffset,\n  backendName: \"wasm\",\n  kernelFunc: rotateWithOffset2,\n  setupFunc: setup37\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Round.js\nvar roundConfig3 = createUnaryKernelConfig(Round);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Rsqrt.js\nvar rsqrtConfig3 = createUnaryKernelConfig(Rsqrt);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ScatterNd.js\nvar wasmScatterNd;\nfunction setup38(backend2) {\n  wasmScatterNd = backend2.wasm.cwrap(ScatterNd, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction scatterNd3(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { indices, updates } = inputs;\n  const { shape } = attrs;\n  const out = backend2.makeOutput(shape, updates.dtype);\n  if (util_exports.sizeFromShape(shape) === 0) {\n    return out;\n  }\n  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = scatter_nd_util_exports.calculateShapes(updates, indices, shape);\n  const indicesData = backend2.dataIdMap.get(indices.dataId);\n  const indicesId = indicesData.id;\n  const updatesData = backend2.dataIdMap.get(updates.dataId);\n  const updatesId = updatesData.id;\n  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmScatterNd(indicesId, updatesId, CppDType[updates.dtype], sliceRank, numUpdates, sliceSize, stridesBytes, outputSize, outId);\n  return out;\n}\nvar scatterNdConfig3 = {\n  kernelName: ScatterNd,\n  backendName: \"wasm\",\n  setupFunc: setup38,\n  kernelFunc: scatterNd3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Select.js\nvar wasmSelect;\nfunction setup39(backend2) {\n  wasmSelect = backend2.wasm.cwrap(\"SelectV2\", null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction select4(args) {\n  const { inputs, backend: backend2 } = args;\n  const { condition, t, e } = inputs;\n  const conditionId = backend2.dataIdMap.get(condition.dataId).id;\n  const tId = backend2.dataIdMap.get(t.dataId).id;\n  const eId = backend2.dataIdMap.get(e.dataId).id;\n  const out = backend2.makeOutput(t.shape, t.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const cRank = condition.shape.length;\n  const tRank = t.shape.length;\n  const offset = cRank === 0 || cRank > 1 || tRank === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));\n  wasmSelect(conditionId, tId, eId, offset, outId);\n  return out;\n}\nvar selectConfig3 = {\n  kernelName: Select,\n  backendName: \"wasm\",\n  kernelFunc: select4,\n  setupFunc: setup39\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sigmoid.js\nvar wasmFunc7;\nfunction setup40(backend2) {\n  wasmFunc7 = backend2.wasm.cwrap(Sigmoid, null, [\"number\", \"number\"]);\n}\nfunction sigmoid4(args) {\n  const { backend: backend2, inputs: { x } } = args;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  if (util_exports.sizeFromShape(out.shape) === 0) {\n    return out;\n  }\n  wasmFunc7(xId, outId);\n  return out;\n}\nvar sigmoidConfig3 = {\n  kernelName: \"Sigmoid\",\n  backendName: \"wasm\",\n  setupFunc: setup40,\n  kernelFunc: sigmoid4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sin.js\nvar sinConfig3 = createUnaryKernelConfig(Sin);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Softmax.js\nvar wasmFunc8;\nfunction setup41(backend2) {\n  wasmFunc8 = backend2.wasm.cwrap(Softmax, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction softmax5(args) {\n  const { backend: backend2, inputs: { logits }, attrs: { dim } } = args;\n  const xId = backend2.dataIdMap.get(logits.dataId).id;\n  const out = backend2.makeOutput(logits.shape, logits.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const channels = logits.shape[dim];\n  const batch = util_exports.sizeFromShape(logits.shape) / channels;\n  if (util_exports.sizeFromShape(out.shape) === 0) {\n    return out;\n  }\n  wasmFunc8(xId, outId, channels, batch);\n  return out;\n}\nvar softmaxConfig3 = {\n  kernelName: Softmax,\n  backendName: \"wasm\",\n  setupFunc: setup41,\n  kernelFunc: softmax5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SpaceToBatchND.js\nfunction spaceToBatchND4(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const { blockShape, paddings } = attrs;\n  const prod5 = util_exports.sizeFromShape(blockShape);\n  const completePaddings = [[0, 0]];\n  completePaddings.push(...paddings);\n  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {\n    completePaddings.push([0, 0]);\n  }\n  const paddedX = padV2Config3.kernelFunc({\n    inputs: { x },\n    backend: backend2,\n    attrs: { paddings: completePaddings, constantValue: 0 }\n  });\n  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);\n  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);\n  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);\n  const reshapeInputs = { x: paddedX };\n  const reshapeAttrs = { shape: reshapedPaddedShape };\n  const paddedXReshaped = reshape5({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });\n  const transposeInputs = { x: paddedXReshaped };\n  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };\n  const paddedXT = transpose4({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });\n  const resultReshapeInputs = { x: paddedXT };\n  const resultReshapeAttrs = { shape: flattenShape };\n  const result = reshape5({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });\n  backend2.disposeData(paddedX.dataId);\n  backend2.disposeData(paddedXReshaped.dataId);\n  backend2.disposeData(paddedXT.dataId);\n  return result;\n}\nvar spaceToBatchNDConfig3 = {\n  kernelName: SpaceToBatchND,\n  backendName: \"wasm\",\n  kernelFunc: spaceToBatchND4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseFillEmptyRows.js\nvar wasmSparseFillEmptyRows;\nfunction setup42(backend2) {\n  wasmSparseFillEmptyRows = backend2.wasm.cwrap(\"SparseFillEmptyRows\", \"number\", [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction sparseFillEmptyRows4(args) {\n  const { backend: backend2, inputs } = args;\n  const { indices, values, denseShape, defaultValue } = inputs;\n  const indicesCount = indices.shape[0];\n  const rank = indices.shape[1];\n  const denseRows = backend2.readSync(denseShape.dataId)[0];\n  const maxOutputIndicesShape = [indicesCount + denseRows, rank];\n  const indicesId = backend2.dataIdMap.get(indices.dataId).id;\n  const valuesId = backend2.dataIdMap.get(values.dataId).id;\n  const defaultValueId = backend2.dataIdMap.get(defaultValue.dataId).id;\n  const outputIndices = backend2.makeOutput(maxOutputIndicesShape, indices.dtype);\n  const outputIndicesId = backend2.dataIdMap.get(outputIndices.dataId).id;\n  const outputValues = backend2.makeOutput(maxOutputIndicesShape.slice(0, 1), values.dtype);\n  const outputValuesId = backend2.dataIdMap.get(outputValues.dataId).id;\n  const emptyRowIndicator = backend2.makeOutput([denseRows], \"bool\");\n  const emptyRowIndicatorId = backend2.dataIdMap.get(emptyRowIndicator.dataId).id;\n  const reverseIndexMap = backend2.makeOutput([indicesCount], indices.dtype);\n  const reverseIndexMapId = backend2.dataIdMap.get(reverseIndexMap.dataId).id;\n  const exceptionValues = backend2.makeOutput([4], \"int32\");\n  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;\n  const outputRows = wasmSparseFillEmptyRows(indicesId, valuesId, CppDType[values.dtype], indicesCount, denseRows, rank, defaultValueId, outputIndicesId, outputValuesId, emptyRowIndicatorId, reverseIndexMapId, exceptionValuesId);\n  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);\n  let exceptionMessage;\n  switch (exceptionValuesArray[0]) {\n    case 1: {\n      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(exceptionValuesArray[1]);\n      break;\n    }\n    case 2: {\n      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);\n      break;\n    }\n    case 3:\n      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);\n      break;\n    default:\n      exceptionMessage = \"\";\n  }\n  backend2.disposeData(exceptionValues.dataId);\n  if (exceptionMessage) {\n    backend2.disposeData(outputIndices.dataId);\n    backend2.disposeData(outputValues.dataId);\n    backend2.disposeData(emptyRowIndicator.dataId);\n    backend2.disposeData(reverseIndexMap.dataId);\n    throw new Error(exceptionMessage);\n  }\n  let resizedIndices = outputIndices;\n  let resizedValues = outputValues;\n  if (outputRows !== maxOutputIndicesShape[0]) {\n    resizedIndices = slice4({\n      inputs: { x: outputIndices },\n      attrs: { begin: 0, size: [outputRows, rank] },\n      backend: backend2\n    });\n    resizedValues = slice4({\n      inputs: { x: outputValues },\n      attrs: { begin: 0, size: outputRows },\n      backend: backend2\n    });\n    backend2.disposeData(outputIndices.dataId);\n    backend2.disposeData(outputValues.dataId);\n  }\n  return [resizedIndices, resizedValues, emptyRowIndicator, reverseIndexMap];\n}\nvar sparseFillEmptyRowsConfig3 = {\n  kernelName: SparseFillEmptyRows,\n  backendName: \"wasm\",\n  setupFunc: setup42,\n  kernelFunc: sparseFillEmptyRows4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseReshape.js\nvar wasmSparseReshape;\nfunction setup43(backend2) {\n  wasmSparseReshape = backend2.wasm.cwrap(SparseReshape, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction sparseReshape4(args) {\n  const { backend: backend2, inputs } = args;\n  const { inputIndices, inputShape, newShape } = inputs;\n  if (inputIndices.shape.length !== 2) {\n    throw new Error(`Input indices should be a matrix but received shape\n        ${inputIndices.shape}`);\n  }\n  if (inputShape.shape.length !== 1) {\n    throw new Error(`Input shape should be a vector but received shape\n        ${inputShape.shape}`);\n  }\n  if (newShape.shape.length !== 1) {\n    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);\n  }\n  const inputIndicesId = backend2.dataIdMap.get(inputIndices.dataId).id;\n  const inputShapeId = backend2.dataIdMap.get(inputShape.dataId).id;\n  const newShapeId = backend2.dataIdMap.get(newShape.dataId).id;\n  const nnz = inputIndices.shape[0];\n  const outputRank = util_exports.sizeFromShape(newShape.shape);\n  const newIndices = backend2.makeOutput([nnz, outputRank], inputIndices.dtype);\n  const newIndicesId = backend2.dataIdMap.get(newIndices.dataId).id;\n  const outputShape = backend2.makeOutput([outputRank], newShape.dtype);\n  const outputShapeId = backend2.dataIdMap.get(outputShape.dataId).id;\n  const exceptionValues = backend2.makeOutput([3], \"int32\");\n  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;\n  wasmSparseReshape(inputIndicesId, inputShapeId, newShapeId, nnz, newIndicesId, outputShapeId, exceptionValuesId);\n  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);\n  let exceptionMessage;\n  switch (exceptionValuesArray[0]) {\n    case 0: {\n      exceptionMessage = backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);\n      break;\n    }\n    case 1: {\n      exceptionMessage = backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);\n      break;\n    }\n    case 2:\n      exceptionMessage = backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();\n      break;\n    case 3: {\n      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));\n      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShapeValues, outputShapeValues);\n      break;\n    }\n    case 4: {\n      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));\n      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShapeValues, outputShapeValues);\n      break;\n    }\n    default:\n      exceptionMessage = \"\";\n  }\n  backend2.disposeData(exceptionValues.dataId);\n  if (exceptionMessage) {\n    backend2.disposeData(newIndices.dataId);\n    backend2.disposeData(outputShape.dataId);\n    throw new Error(exceptionMessage);\n  }\n  return [newIndices, outputShape];\n}\nvar sparseReshapeConfig3 = {\n  kernelName: SparseReshape,\n  backendName: \"wasm\",\n  setupFunc: setup43,\n  kernelFunc: sparseReshape4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentReduction.js\nvar wasmSparseSegmentReduction;\nfunction setup44(backend2) {\n  wasmSparseSegmentReduction = backend2.wasm.cwrap(\"SparseSegmentReduction\", null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction sparseSegmentReduction(args, isMean) {\n  const { backend: backend2, inputs } = args;\n  const { data, indices, segmentIds } = inputs;\n  const numIndices = indices.shape[0];\n  const segmentIdsBack = backend2.readSync(segmentIds.dataId, numIndices - 1, numIndices)[0];\n  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIdsBack + 1 : 0;\n  const outputRows = lastSegmentIdPlusOne;\n  if (outputRows < 0) {\n    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());\n  }\n  const outputShape = data.shape.slice();\n  outputShape[0] = outputRows;\n  const dataId = backend2.dataIdMap.get(data.dataId).id;\n  const indicesId = backend2.dataIdMap.get(indices.dataId).id;\n  const segmentIdsId = backend2.dataIdMap.get(segmentIds.dataId).id;\n  const output = backend2.makeOutput(outputShape, data.dtype);\n  const outputId = backend2.dataIdMap.get(output.dataId).id;\n  const exceptionValues = backend2.makeOutput([4], \"int32\");\n  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;\n  wasmSparseSegmentReduction(dataId, CppDType[data.dtype], data.shape[0], indicesId, segmentIdsId, outputId, exceptionValuesId, isMean, 0);\n  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);\n  let exceptionMessage;\n  switch (exceptionValuesArray[0]) {\n    case 0: {\n      exceptionMessage = backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();\n      break;\n    }\n    case 1: {\n      exceptionMessage = backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();\n      break;\n    }\n    case 2:\n      exceptionMessage = backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);\n      break;\n    case 3:\n      exceptionMessage = backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);\n      break;\n    default:\n      exceptionMessage = \"\";\n  }\n  backend2.disposeData(exceptionValues.dataId);\n  if (exceptionMessage) {\n    backend2.disposeData(output.dataId);\n    throw new Error(exceptionMessage);\n  }\n  return output;\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentMean.js\nfunction sparseSegmentMean4(args) {\n  return sparseSegmentReduction(args, true);\n}\nvar sparseSegmentMeanConfig3 = {\n  kernelName: SparseSegmentMean,\n  backendName: \"wasm\",\n  setupFunc: setup44,\n  kernelFunc: sparseSegmentMean4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentSum.js\nfunction sparseSegmentSum4(args) {\n  return sparseSegmentReduction(args, false);\n}\nvar sparseSegmentSumConfig3 = {\n  kernelName: SparseSegmentSum,\n  backendName: \"wasm\",\n  setupFunc: setup44,\n  kernelFunc: sparseSegmentSum4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SplitV.js\nfunction splitV3(args) {\n  const { inputs, attrs, backend: backend2 } = args;\n  const { x } = inputs;\n  const { numOrSizeSplits, axis } = attrs;\n  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];\n  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);\n  const begin = new Array(x.shape.length).fill(0);\n  const size = x.shape.slice();\n  return splitSizes.map((s) => {\n    const xSliceSize = [...size];\n    xSliceSize[$axis] = s;\n    const xSlice = slice4({ inputs: { x }, attrs: { begin, size: xSliceSize }, backend: backend2 });\n    begin[$axis] += s;\n    return xSlice;\n  });\n}\nvar splitVConfig3 = {\n  kernelName: SplitV,\n  backendName: \"wasm\",\n  kernelFunc: splitV3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sqrt.js\nvar sqrtConfig3 = createUnaryKernelConfig(Sqrt);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Square.js\nvar squareConfig3 = createUnaryKernelConfig(Square);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SquaredDifference.js\nvar supportsFullBroadcast17 = true;\nvar squaredDifferenceConfig3 = createBinaryKernelConfig(SquaredDifference, supportsFullBroadcast17);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Step.js\nvar wasmStep;\nfunction setup45(backend2) {\n  wasmStep = backend2.wasm.cwrap(Step, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction step4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { alpha } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmStep(xId, alpha, CppDType[x.dtype], outId);\n  return out;\n}\nvar stepConfig3 = {\n  kernelName: Step,\n  backendName: \"wasm\",\n  setupFunc: setup45,\n  kernelFunc: step4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StridedSlice.js\nvar wasmStridedSlice;\nfunction setup46(backend2) {\n  wasmStridedSlice = backend2.wasm.cwrap(StridedSlice, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"array\",\n    \"array\",\n    \"array\",\n    \"array\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction stridedSlice4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { x } = inputs;\n  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;\n  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);\n  let result;\n  if (isIdentity) {\n    result = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });\n  } else if (sliceDim0 || isSimpleSlice) {\n    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);\n    const size = slice_util_exports.computeOutShape($begin, $end, $strides);\n    const sliced = slice4({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });\n    result = reshape5({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });\n    backend2.disposeData(sliced.dataId);\n  } else {\n    const out = backend2.makeOutput(finalShapeSparse, \"float32\");\n    const xId = backend2.dataIdMap.get(x.dataId).id;\n    const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);\n    const beginBytes = new Uint8Array(new Int32Array($begin).buffer);\n    const endBytes = new Uint8Array(new Int32Array($end).buffer);\n    const stridesBytes = new Uint8Array(new Int32Array($strides).buffer);\n    const outputShapeBytes = new Uint8Array(new Int32Array(finalShapeSparse).buffer);\n    const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(finalShapeSparse)).buffer);\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmStridedSlice(xId, xStridesBytes, x.shape.length, beginBytes, endBytes, stridesBytes, outputShapeBytes, outStridesBytes, finalShapeSparse.length, outId);\n    result = reshape5({ inputs: { x: out }, backend: backend2, attrs: { shape: finalShape } });\n    backend2.disposeData(out.dataId);\n  }\n  return result;\n}\nvar stridedSliceConfig3 = {\n  kernelName: StridedSlice,\n  backendName: \"wasm\",\n  setupFunc: setup46,\n  kernelFunc: stridedSlice4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringNGrams.js\nfunction stringNGrams4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { data, dataSplits } = inputs;\n  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;\n  const $data = backend2.readSync(data.dataId);\n  const $dataSplits = backend2.readSync(dataSplits.dataId);\n  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);\n  const nGramsOut = backend2.makeOutput([nGrams.length], \"string\");\n  const nGramsOutData = backend2.dataIdMap.get(nGramsOut.dataId);\n  nGramsOutData.stringBytes = nGrams;\n  const nGramsSplitsOut = backend2.makeOutput(dataSplits.shape, \"int32\");\n  const nGramsSplitsOutVals = backend2.typedArrayFromHeap(nGramsSplitsOut);\n  nGramsSplitsOutVals.set(nGramsSplits);\n  return [nGramsOut, nGramsSplitsOut];\n}\nvar stringNGramsConfig3 = {\n  kernelName: StringNGrams,\n  backendName: \"wasm\",\n  kernelFunc: stringNGrams4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringSplit.js\nfunction stringSplit4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { input: input2, delimiter } = inputs;\n  const { skipEmpty } = attrs;\n  const inputVals = backend2.readSync(input2.dataId);\n  const delimiterVals = backend2.readSync(delimiter.dataId);\n  const [indices, values, shape] = stringSplitImpl(inputVals, delimiterVals[0], skipEmpty);\n  const outputSize = values.length;\n  const indicesOut = backend2.makeOutput([outputSize, 2], \"int32\");\n  const indicesOutVals = backend2.typedArrayFromHeap(indicesOut);\n  indicesOutVals.set(indices);\n  const valuesOut = backend2.makeOutput([outputSize], \"string\");\n  const valuesOutData = backend2.dataIdMap.get(valuesOut.dataId);\n  valuesOutData.stringBytes = values;\n  const shapeOut = backend2.makeOutput([2], \"int32\");\n  const shapeOutVals = backend2.typedArrayFromHeap(shapeOut);\n  shapeOutVals.set(shape);\n  return [indicesOut, valuesOut, shapeOut];\n}\nvar stringSplitConfig3 = {\n  kernelName: StringSplit,\n  backendName: \"wasm\",\n  kernelFunc: stringSplit4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringToHashBucketFast.js\nfunction stringToHashBucketFast4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { input: input2 } = inputs;\n  const { numBuckets } = attrs;\n  const inputVals = backend2.readSync(input2.dataId);\n  const values = stringToHashBucketFastImpl(inputVals, numBuckets);\n  const out = backend2.makeOutput(input2.shape, \"int32\");\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.set(values);\n  return out;\n}\nvar stringToHashBucketFastConfig3 = {\n  kernelName: StringToHashBucketFast,\n  backendName: \"wasm\",\n  kernelFunc: stringToHashBucketFast4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sub.js\nvar supportsFullBroadcast18 = true;\nvar subConfig3 = createBinaryKernelConfig(Sub, supportsFullBroadcast18);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sum.js\nvar wasmSum;\nfunction setup47(backend2) {\n  wasmSum = backend2.wasm.cwrap(Sum, null, [\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction sum5(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { axis, keepDims } = attrs;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  let inputId = xId;\n  let input2 = x;\n  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);\n  let reductionAxes = axes;\n  if (inputWasTransposed) {\n    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;\n    if (transposedId !== xId) {\n      input2 = transposed;\n      inputId = transposedId;\n      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);\n    }\n  }\n  backend_util_exports.assertAxesAreInnerMostDims(\"sum\", reductionAxes, input2.shape.length);\n  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);\n  const reduceSize = util_exports.sizeFromShape(reduceShape);\n  const out = backend2.makeOutput(outShape, input2.dtype);\n  if (util_exports.sizeFromShape(input2.shape) !== 0) {\n    const outId = backend2.dataIdMap.get(out.dataId).id;\n    wasmSum(inputId, reduceSize, CppDType[out.dtype], outId);\n  }\n  if (inputWasTransposed) {\n    backend2.disposeData(transposed.dataId);\n  }\n  if (keepDims) {\n    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);\n    out.shape = newShape;\n  }\n  return out;\n}\nvar sumConfig3 = {\n  kernelName: Sum,\n  backendName: \"wasm\",\n  setupFunc: setup47,\n  kernelFunc: sum5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tan.js\nvar tanConfig3 = createUnaryKernelConfig(Tan);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tanh.js\nvar tanhConfig3 = createUnaryKernelConfig(Tanh);\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tile.js\nvar wasmTile;\nfunction setup48(backend2) {\n  wasmTile = backend2.wasm.cwrap(Tile, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction tile5(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { x } = inputs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const { reps } = attrs;\n  const newShape = new Array(x.shape.length);\n  for (let i = 0; i < newShape.length; i++) {\n    newShape[i] = x.shape[i] * reps[i];\n  }\n  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  const newShapeBytes = new Uint8Array(new Int32Array(newShape).buffer);\n  const out = backend2.makeOutput(newShape, x.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  wasmTile(xId, xShapeBytes, x.shape.length, newShapeBytes, newShape.length, CppDType[out.dtype], outId);\n  return out;\n}\nvar tileConfig3 = {\n  kernelName: Tile,\n  backendName: \"wasm\",\n  setupFunc: setup48,\n  kernelFunc: tile5\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/TopK.js\nvar wasmTopK;\nfunction setup49(backend2) {\n  wasmTopK = backend2.wasm.cwrap(TopK, null, [\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"bool\",\n    \"number\",\n    \"number\"\n  ]);\n}\nvar topk2 = ({ inputs, backend: backend2, attrs }) => {\n  const { x } = inputs;\n  const { k, sorted } = attrs;\n  const xId = backend2.dataIdMap.get(x.dataId).id;\n  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);\n  const outputShape = x.shape.slice();\n  outputShape[outputShape.length - 1] = k;\n  const outValues = backend2.makeOutput(outputShape, x.dtype);\n  const outValuesId = backend2.dataIdMap.get(outValues.dataId).id;\n  const outIndices = backend2.makeOutput(outputShape, \"int32\");\n  const outIndicesId = backend2.dataIdMap.get(outIndices.dataId).id;\n  wasmTopK(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], k, sorted, outValuesId, outIndicesId);\n  return [outValues, outIndices];\n};\nvar topKConfig3 = {\n  kernelName: TopK,\n  backendName: \"wasm\",\n  setupFunc: setup49,\n  kernelFunc: topk2\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Transform.js\nvar wasmTransform;\nfunction setup50(backend2) {\n  wasmTransform = backend2.wasm.cwrap(Transform, null, [\n    \"number\",\n    \"number\",\n    \"bool\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"array\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\",\n    \"number\"\n  ]);\n}\nfunction transform4(args) {\n  const { backend: backend2, inputs, attrs } = args;\n  const { image: image2, transforms } = inputs;\n  const { interpolation, fillMode, fillValue, outputShape } = attrs;\n  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;\n  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];\n  const outShape = [\n    batch,\n    outHeight,\n    outWidth,\n    numChannels\n  ];\n  const inputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(image2.shape)).buffer);\n  const outputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(outShape)).buffer);\n  const out = backend2.makeOutput(outShape, image2.dtype);\n  const outId = backend2.dataIdMap.get(out.dataId).id;\n  const imageData = backend2.dataIdMap.get(image2.dataId);\n  const imageId = imageData.id;\n  const transformsData = backend2.dataIdMap.get(transforms.dataId);\n  const transformsId = transformsData.id;\n  const interpolationModeId = interpolation === \"nearest\" ? 1 : 2;\n  let fillModeId;\n  switch (fillMode) {\n    case \"constant\":\n      fillModeId = 1;\n      break;\n    case \"reflect\":\n      fillModeId = 2;\n      break;\n    case \"wrap\":\n      fillModeId = 3;\n      break;\n    case \"nearest\":\n      fillModeId = 4;\n      break;\n    default:\n      fillModeId = 1;\n      break;\n  }\n  wasmTransform(imageId, transformsId, transforms.shape[0] > 1, batch, outHeight, outWidth, numChannels, imageWidth, imageHeight, inputStrides, image2.shape.length - 1, outputStrides, outShape.length - 1, interpolationModeId, fillModeId, fillValue, outId);\n  return out;\n}\nvar transformConfig3 = {\n  kernelName: Transform,\n  backendName: \"wasm\",\n  setupFunc: setup50,\n  kernelFunc: transform4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Unpack.js\nfunction unpack3(args) {\n  const { inputs, backend: backend2, attrs } = args;\n  const { value } = inputs;\n  let { axis } = attrs;\n  if (axis < 0) {\n    axis += value.shape.length;\n  }\n  const numOutputs = value.shape[axis];\n  const rank = value.shape.length;\n  const outShape = new Array(rank - 1);\n  let outIndex = 0;\n  for (let i = 0; i < rank; i++) {\n    if (i !== axis) {\n      outShape[outIndex++] = value.shape[i];\n    }\n  }\n  const outs = new Array(numOutputs);\n  const begin = new Array(rank).fill(0);\n  const size = value.shape.slice();\n  size[axis] = 1;\n  for (let i = 0; i < outs.length; i++) {\n    begin[axis] = i;\n    outs[i] = slice4({ inputs: { x: value }, attrs: { begin, size }, backend: backend2 });\n  }\n  return outs.map(({ dataId, dtype }) => ({ dataId, dtype, shape: outShape }));\n}\nvar unpackConfig3 = {\n  kernelName: Unpack,\n  backendName: \"wasm\",\n  kernelFunc: unpack3\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ZerosLike.js\nfunction zerosLike4(args) {\n  const { inputs: { x }, backend: backend2 } = args;\n  const out = backend2.makeOutput(x.shape, x.dtype);\n  const outVals = backend2.typedArrayFromHeap(out);\n  outVals.fill(0);\n  return out;\n}\nvar zerosLikeConfig3 = {\n  kernelName: ZerosLike,\n  backendName: \"wasm\",\n  kernelFunc: zerosLike4\n};\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/register_all_kernels.js\nvar kernelConfigs3 = [\n  _fusedMatMulConfig3,\n  absConfig3,\n  addConfig3,\n  addNConfig3,\n  allConfig3,\n  anyConfig3,\n  argMaxConfig3,\n  avgPoolConfig3,\n  batchMatMulConfig3,\n  batchToSpaceNDConfig3,\n  castConfig3,\n  ceilConfig3,\n  clipByValueConfig3,\n  concatConfig3,\n  conv2DConfig3,\n  conv2DBackpropInputConfig3,\n  cosConfig3,\n  coshConfig3,\n  cropAndResizeConfig3,\n  cumprodConfig3,\n  cumsumConfig3,\n  depthToSpaceConfig3,\n  depthwiseConv2dNativeConfig3,\n  eluConfig3,\n  equalConfig3,\n  expConfig3,\n  expandDimsConfig3,\n  fillConfig3,\n  flipLeftRightConfig3,\n  floorConfig3,\n  floorDivConfig3,\n  fusedBatchNormConfig,\n  fusedConv2DConfig3,\n  fusedDepthwiseConv2DConfig3,\n  gatherNdConfig3,\n  gatherV2Config3,\n  greaterConfig3,\n  greaterEqualConfig3,\n  identityConfig3,\n  leakyReluConfig3,\n  lessConfig3,\n  lessEqualConfig3,\n  logConfig3,\n  logicalAndConfig3,\n  logicalNotConfig3,\n  logicalOrConfig3,\n  logicalXorConfig,\n  maxConfig3,\n  maximumConfig3,\n  maxPoolConfig3,\n  meanConfig3,\n  minConfig3,\n  minimumConfig3,\n  mirrorPadConfig3,\n  multiplyConfig3,\n  negConfig3,\n  nonMaxSuppressionV3Config3,\n  nonMaxSuppressionV4Config3,\n  nonMaxSuppressionV5Config3,\n  notEqualConfig3,\n  oneHotConfig3,\n  onesLikeConfig3,\n  packConfig3,\n  padV2Config3,\n  powConfig3,\n  preluConfig3,\n  prodConfig3,\n  rangeConfig3,\n  realDivConfig3,\n  reluConfig3,\n  relu6Config3,\n  reshapeConfig3,\n  resizeBilinearConfig3,\n  resizeNearestNeighborConfig3,\n  reverseConfig3,\n  rotateWithOffsetConfig3,\n  roundConfig3,\n  rsqrtConfig3,\n  scatterNdConfig3,\n  selectConfig3,\n  sigmoidConfig3,\n  sinConfig3,\n  sliceConfig3,\n  softmaxConfig3,\n  spaceToBatchNDConfig3,\n  sparseFillEmptyRowsConfig3,\n  sparseReshapeConfig3,\n  sparseSegmentMeanConfig3,\n  sparseSegmentSumConfig3,\n  splitVConfig3,\n  sqrtConfig3,\n  squareConfig3,\n  squaredDifferenceConfig3,\n  stepConfig3,\n  stridedSliceConfig3,\n  stringNGramsConfig3,\n  stringSplitConfig3,\n  stringToHashBucketFastConfig3,\n  subConfig3,\n  sumConfig3,\n  tanConfig3,\n  tanhConfig3,\n  tileConfig3,\n  topKConfig3,\n  transformConfig3,\n  transposeConfig3,\n  unpackConfig3,\n  zerosLikeConfig3\n];\nfor (const kernelConfig of kernelConfigs3) {\n  registerKernel(kernelConfig);\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/flags_wasm.js\nvar ENV6 = env();\nENV6.registerFlag(\"WASM_HAS_SIMD_SUPPORT\", async () => {\n  try {\n    return WebAssembly.validate(new Uint8Array([\n      0,\n      97,\n      115,\n      109,\n      1,\n      0,\n      0,\n      0,\n      1,\n      4,\n      1,\n      96,\n      0,\n      0,\n      3,\n      2,\n      1,\n      0,\n      10,\n      9,\n      1,\n      7,\n      0,\n      65,\n      0,\n      253,\n      15,\n      26,\n      11\n    ]));\n  } catch (e) {\n    return false;\n  }\n});\nENV6.registerFlag(\"WASM_HAS_MULTITHREAD_SUPPORT\", async () => {\n  if (ENV6.get(\"IS_NODE\")) {\n    return false;\n  }\n  try {\n    new MessageChannel().port1.postMessage(new SharedArrayBuffer(1));\n    return WebAssembly.validate(new Uint8Array([\n      0,\n      97,\n      115,\n      109,\n      1,\n      0,\n      0,\n      0,\n      1,\n      4,\n      1,\n      96,\n      0,\n      0,\n      3,\n      2,\n      1,\n      0,\n      5,\n      4,\n      1,\n      3,\n      1,\n      1,\n      10,\n      11,\n      1,\n      9,\n      0,\n      65,\n      0,\n      254,\n      16,\n      2,\n      0,\n      26,\n      11\n    ]));\n  } catch (e) {\n    return false;\n  }\n});\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/backend_wasm.js\nvar wasmFactoryThreadedSimd_import = __toESM(require_tfjs_backend_wasm_threaded_simd());\nvar import_tfjs_backend_wasm_threaded_simd_worker = __toESM(require_tfjs_backend_wasm_threaded_simd_worker());\nvar wasmFactory_import = __toESM(require_tfjs_backend_wasm());\nvar wasmFactoryThreadedSimd = wasmFactoryThreadedSimd_import.default || wasmFactoryThreadedSimd_import;\nvar wasmFactory = wasmFactory_import.default || wasmFactory_import;\nvar BackendWasm = class extends KernelBackend {\n  constructor(wasm) {\n    super();\n    this.wasm = wasm;\n    this.dataIdNextNumber = 1;\n    this.wasm.tfjs.initWithThreadsCount(threadsCount);\n    actualThreadsCount = this.wasm.tfjs.getThreadsCount();\n    this.dataIdMap = new DataStorage(this, engine());\n  }\n  write(values, shape, dtype) {\n    const dataId = { id: this.dataIdNextNumber++ };\n    this.move(dataId, values, shape, dtype, 1);\n    return dataId;\n  }\n  numDataIds() {\n    return this.dataIdMap.numDataIds();\n  }\n  async time(f) {\n    const start = util_exports.now();\n    f();\n    const kernelMs = util_exports.now() - start;\n    return { kernelMs };\n  }\n  move(dataId, values, shape, dtype, refCount) {\n    const id = this.dataIdNextNumber++;\n    if (dtype === \"string\") {\n      const stringBytes = values;\n      this.dataIdMap.set(dataId, { id, stringBytes, shape, dtype, memoryOffset: null, refCount });\n      return;\n    }\n    const size = util_exports.sizeFromShape(shape);\n    const numBytes = size * util_exports.bytesPerElement(dtype);\n    const memoryOffset = this.wasm._malloc(numBytes);\n    this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount });\n    this.wasm.tfjs.registerTensor(id, size, memoryOffset);\n    if (values != null) {\n      this.wasm.HEAPU8.set(new Uint8Array(values.buffer, values.byteOffset, numBytes), memoryOffset);\n    }\n  }\n  async read(dataId) {\n    return this.readSync(dataId);\n  }\n  readSync(dataId, start, end) {\n    const { memoryOffset, dtype, shape, stringBytes } = this.dataIdMap.get(dataId);\n    if (dtype === \"string\") {\n      if ((start == null || start === 0) && (end == null || end >= stringBytes.length)) {\n        return stringBytes;\n      }\n      return stringBytes.slice(start, end);\n    }\n    start = start || 0;\n    end = end || util_exports.sizeFromShape(shape);\n    const bytesPerElement2 = util_exports.bytesPerElement(dtype);\n    const bytes = this.wasm.HEAPU8.slice(memoryOffset + start * bytesPerElement2, memoryOffset + end * bytesPerElement2);\n    return typedArrayFromBuffer(bytes.buffer, dtype);\n  }\n  disposeData(dataId, force = false) {\n    if (this.dataIdMap.has(dataId)) {\n      const data = this.dataIdMap.get(dataId);\n      data.refCount--;\n      if (!force && data.refCount > 0) {\n        return false;\n      }\n      this.wasm._free(data.memoryOffset);\n      this.wasm.tfjs.disposeData(data.id);\n      this.dataIdMap.delete(dataId);\n    }\n    return true;\n  }\n  refCount(dataId) {\n    if (this.dataIdMap.has(dataId)) {\n      const tensorData = this.dataIdMap.get(dataId);\n      return tensorData.refCount;\n    }\n    return 0;\n  }\n  incRef(dataId) {\n    const data = this.dataIdMap.get(dataId);\n    if (data != null) {\n      data.refCount++;\n    }\n  }\n  floatPrecision() {\n    return 32;\n  }\n  getMemoryOffset(dataId) {\n    return this.dataIdMap.get(dataId).memoryOffset;\n  }\n  dispose() {\n    this.wasm.tfjs.dispose();\n    if (\"PThread\" in this.wasm) {\n      this.wasm.PThread.terminateAllThreads();\n    }\n    this.wasm = null;\n  }\n  memory() {\n    return { unreliable: false };\n  }\n  makeOutput(shape, dtype, memoryOffset) {\n    let dataId;\n    if (memoryOffset == null) {\n      dataId = this.write(null, shape, dtype);\n    } else {\n      const id = this.dataIdNextNumber++;\n      dataId = { id };\n      this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount: 1 });\n      const size = util_exports.sizeFromShape(shape);\n      this.wasm.tfjs.registerTensor(id, size, memoryOffset);\n    }\n    return { dataId, shape, dtype };\n  }\n  typedArrayFromHeap({ shape, dtype, dataId }) {\n    const buffer2 = this.wasm.HEAPU8.buffer;\n    const { memoryOffset } = this.dataIdMap.get(dataId);\n    const size = util_exports.sizeFromShape(shape);\n    switch (dtype) {\n      case \"float32\":\n        return new Float32Array(buffer2, memoryOffset, size);\n      case \"int32\":\n        return new Int32Array(buffer2, memoryOffset, size);\n      case \"bool\":\n        return new Uint8Array(buffer2, memoryOffset, size);\n      default:\n        throw new Error(`Unknown dtype ${dtype}`);\n    }\n  }\n};\nfunction createInstantiateWasmFunc(path) {\n  return (imports, callback) => {\n    util_exports.fetch(path, { credentials: \"same-origin\" }).then((response) => {\n      if (!response[\"ok\"]) {\n        imports.env.a(`failed to load wasm binary file at '${path}'`);\n      }\n      response.arrayBuffer().then((binary) => {\n        WebAssembly.instantiate(binary, imports).then((output) => {\n          callback(output.instance, output.module);\n        });\n      });\n    });\n    return {};\n  };\n}\nfunction getPathToWasmBinary(simdSupported, threadsSupported, wasmModuleFolder) {\n  if (wasmPath != null) {\n    return wasmPath;\n  }\n  let path = \"tfjs-backend-wasm.wasm\";\n  if (simdSupported && threadsSupported) {\n    path = \"tfjs-backend-wasm-threaded-simd.wasm\";\n  } else if (simdSupported) {\n    path = \"tfjs-backend-wasm-simd.wasm\";\n  }\n  if (wasmFileMap != null) {\n    if (wasmFileMap[path] != null) {\n      return wasmFileMap[path];\n    }\n  }\n  return wasmModuleFolder + path;\n}\nasync function init() {\n  const [simdSupported, threadsSupported] = await Promise.all([\n    env().getAsync(\"WASM_HAS_SIMD_SUPPORT\"),\n    env().getAsync(\"WASM_HAS_MULTITHREAD_SUPPORT\")\n  ]);\n  return new Promise((resolve, reject) => {\n    const factoryConfig = {};\n    factoryConfig.locateFile = (path, prefix) => {\n      if (path.endsWith(\".worker.js\")) {\n        const response = import_tfjs_backend_wasm_threaded_simd_worker.wasmWorkerContents.replace(/\\n/g, \"\\\\n\");\n        const blob = new Blob([response], { type: \"application/javascript\" });\n        return URL.createObjectURL(blob);\n      }\n      if (path.endsWith(\".wasm\")) {\n        return getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : prefix);\n      }\n      return prefix + path;\n    };\n    if (customFetch) {\n      factoryConfig.instantiateWasm = createInstantiateWasmFunc(getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : \"\"));\n    }\n    let initialized = false;\n    factoryConfig.onAbort = () => {\n      if (initialized) {\n        return;\n      }\n      if (initAborted) {\n        return;\n      }\n      initAborted = true;\n      const rejectMsg = \"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers\";\n      reject({ message: rejectMsg });\n    };\n    let wasm;\n    if (threadsSupported && simdSupported && wasmPath == null) {\n      factoryConfig.mainScriptUrlOrBlob = new Blob([`var WasmBackendModuleThreadedSimd = ` + wasmFactoryThreadedSimd.toString()], { type: \"text/javascript\" });\n      wasm = wasmFactoryThreadedSimd(factoryConfig);\n    } else {\n      wasm = wasmFactory(factoryConfig);\n    }\n    wasm.then((module) => {\n      initialized = true;\n      initAborted = false;\n      const voidReturnType = null;\n      module.tfjs = {\n        init: module.cwrap(\"init\", null, []),\n        initWithThreadsCount: module.cwrap(\"init_with_threads_count\", null, [\"number\"]),\n        getThreadsCount: module.cwrap(\"get_threads_count\", \"number\", []),\n        registerTensor: module.cwrap(\"register_tensor\", null, [\n          \"number\",\n          \"number\",\n          \"number\"\n        ]),\n        disposeData: module.cwrap(\"dispose_data\", voidReturnType, [\"number\"]),\n        dispose: module.cwrap(\"dispose\", voidReturnType, [])\n      };\n      resolve({ wasm: module });\n    }).catch(reject);\n  });\n}\nfunction typedArrayFromBuffer(buffer2, dtype) {\n  switch (dtype) {\n    case \"float32\":\n      return new Float32Array(buffer2);\n    case \"int32\":\n      return new Int32Array(buffer2);\n    case \"bool\":\n      return new Uint8Array(buffer2);\n    default:\n      throw new Error(`Unknown dtype ${dtype}`);\n  }\n}\nvar wasmBinaryNames = [\n  \"tfjs-backend-wasm.wasm\",\n  \"tfjs-backend-wasm-simd.wasm\",\n  \"tfjs-backend-wasm-threaded-simd.wasm\"\n];\nvar wasmPath = null;\nvar wasmPathPrefix = null;\nvar wasmFileMap = {};\nvar initAborted = false;\nvar customFetch = false;\nfunction setWasmPath(path, usePlatformFetch = false) {\n  deprecationWarn(\"setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release.\");\n  if (initAborted) {\n    throw new Error(\"The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`\");\n  }\n  wasmPath = path;\n  customFetch = usePlatformFetch;\n}\nfunction setWasmPaths(prefixOrFileMap, usePlatformFetch = false) {\n  if (initAborted) {\n    throw new Error(\"The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`\");\n  }\n  if (typeof prefixOrFileMap === \"string\") {\n    wasmPathPrefix = prefixOrFileMap;\n  } else {\n    wasmFileMap = prefixOrFileMap;\n    const missingPaths = wasmBinaryNames.filter((name) => wasmFileMap[name] == null);\n    if (missingPaths.length > 0) {\n      throw new Error(`There were no entries found for the following binaries: ${missingPaths.join(\",\")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`);\n    }\n  }\n  customFetch = usePlatformFetch;\n}\nvar threadsCount = -1;\nvar actualThreadsCount = -1;\nfunction setThreadsCount(numThreads) {\n  threadsCount = numThreads;\n}\nfunction getThreadsCount() {\n  if (actualThreadsCount === -1) {\n    throw new Error(`WASM backend not initialized.`);\n  }\n  return actualThreadsCount;\n}\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/version.js\nvar version8 = \"4.0.0\";\n\n// node_modules/@tensorflow/tfjs-backend-wasm/dist/base.js\nvar WASM_PRIORITY = 2;\nregisterBackend(\"wasm\", async () => {\n  const { wasm } = await init();\n  return new BackendWasm(wasm);\n}, WASM_PRIORITY);\n\n// dist/tfjs.version.js\nvar version9 = \"4.0.0\";\nvar version22 = \"4.0.0\";\nvar version32 = \"4.0.0\";\nvar version42 = \"4.0.0\";\nvar version52 = \"4.0.0\";\nvar version62 = {\n  tfjs: version9,\n  \"tfjs-core\": version9,\n  \"tfjs-converter\": version22,\n  \"tfjs-backend-cpu\": version32,\n  \"tfjs-backend-webgl\": version42,\n  \"tfjs-backend-wasm\": version52\n};\nexport {\n  Abs,\n  Acos,\n  Acosh,\n  AdadeltaOptimizer,\n  AdagradOptimizer,\n  AdamOptimizer,\n  AdamaxOptimizer,\n  Add,\n  AddN,\n  All,\n  Any,\n  ArgMax,\n  ArgMin,\n  Asin,\n  Asinh,\n  Atan,\n  Atan2,\n  Atanh,\n  AvgPool,\n  AvgPool3D,\n  AvgPool3DGrad,\n  AvgPoolGrad,\n  BackendWasm,\n  BatchMatMul,\n  BatchToSpaceND,\n  Bincount,\n  BroadcastArgs,\n  BroadcastTo,\n  Callback,\n  CallbackList,\n  Cast,\n  Ceil,\n  ClipByValue,\n  Complex,\n  ComplexAbs,\n  Concat,\n  Conv2D,\n  Conv2DBackpropFilter,\n  Conv2DBackpropInput,\n  Conv3D,\n  Conv3DBackpropFilterV2,\n  Conv3DBackpropInputV2,\n  Cos,\n  Cosh,\n  CropAndResize,\n  Cumprod,\n  Cumsum,\n  CustomCallback,\n  DataStorage,\n  DenseBincount,\n  DepthToSpace,\n  DepthwiseConv2dNative,\n  DepthwiseConv2dNativeBackpropFilter,\n  DepthwiseConv2dNativeBackpropInput,\n  Diag,\n  Dilation2D,\n  Dilation2DBackpropFilter,\n  Dilation2DBackpropInput,\n  ENV,\n  EarlyStopping,\n  Einsum,\n  Elu,\n  EluGrad,\n  Environment,\n  Equal,\n  Erf,\n  Exp,\n  ExpandDims,\n  Expm1,\n  FFT,\n  Fill,\n  FlipLeftRight,\n  Floor,\n  FloorDiv,\n  FromPixels,\n  FusedBatchNorm,\n  FusedConv2D,\n  FusedDepthwiseConv2D,\n  GPGPUContext,\n  GatherNd,\n  GatherV2,\n  GraphModel,\n  Greater,\n  GreaterEqual,\n  History,\n  IFFT,\n  Identity,\n  Imag,\n  InputSpec,\n  IsFinite,\n  IsInf,\n  IsNan,\n  KernelBackend,\n  LRN,\n  LRNGrad,\n  LayerVariable,\n  LayersModel,\n  LeakyRelu,\n  Less,\n  LessEqual,\n  LinSpace,\n  Log,\n  Log1p,\n  LogSoftmax,\n  LogicalAnd,\n  LogicalNot,\n  LogicalOr,\n  LogicalXor,\n  LowerBound,\n  MathBackendWebGL,\n  Max,\n  MaxPool,\n  MaxPool3D,\n  MaxPool3DGrad,\n  MaxPoolGrad,\n  MaxPoolWithArgmax,\n  Maximum,\n  Mean,\n  Min,\n  Minimum,\n  MirrorPad,\n  Mod,\n  MomentumOptimizer,\n  Multinomial,\n  Multiply,\n  Neg,\n  NonMaxSuppressionV3,\n  NonMaxSuppressionV4,\n  NonMaxSuppressionV5,\n  NotEqual,\n  OP_SCOPE_SUFFIX,\n  OneHot,\n  OnesLike,\n  Optimizer,\n  OptimizerConstructors,\n  Pack,\n  PadV2,\n  Pool,\n  Pow,\n  Prelu,\n  Prod,\n  RMSPropOptimizer,\n  RNN,\n  RaggedGather,\n  RaggedRange,\n  RaggedTensorToTensor,\n  Range,\n  Rank,\n  Real,\n  RealDiv,\n  Reciprocal,\n  Reduction,\n  Relu,\n  Relu6,\n  Reshape,\n  ResizeBilinear,\n  ResizeBilinearGrad,\n  ResizeNearestNeighbor,\n  ResizeNearestNeighborGrad,\n  Reverse,\n  RotateWithOffset,\n  Round,\n  Rsqrt,\n  SGDOptimizer,\n  ScatterNd,\n  SearchSorted,\n  Select,\n  Selu,\n  Sequential,\n  Sigmoid,\n  Sign,\n  Sin,\n  Sinh,\n  Slice,\n  Softmax,\n  Softplus,\n  SpaceToBatchND,\n  SparseFillEmptyRows,\n  SparseReshape,\n  SparseSegmentMean,\n  SparseSegmentSum,\n  SparseToDense,\n  SplitV,\n  Sqrt,\n  Square,\n  SquaredDifference,\n  Step,\n  StridedSlice,\n  StringNGrams,\n  StringSplit,\n  StringToHashBucketFast,\n  Sub,\n  Sum,\n  SymbolicTensor,\n  Tan,\n  Tanh,\n  Tensor,\n  TensorBuffer,\n  Tile,\n  TopK,\n  Transform,\n  Transpose,\n  Unique,\n  Unpack,\n  UnsortedSegmentSum,\n  UpperBound,\n  Variable,\n  ZerosLike,\n  _FusedMatMul,\n  abs,\n  acos,\n  acosh,\n  add2 as add,\n  addN,\n  all,\n  any,\n  argMax,\n  argMin,\n  asin,\n  asinh,\n  atan,\n  atan2,\n  atanh,\n  avgPool,\n  avgPool3d,\n  backend,\n  backend_util_exports as backend_util,\n  basicLSTMCell,\n  batchNorm,\n  batchNorm2d,\n  batchNorm3d,\n  batchNorm4d,\n  batchToSpaceND,\n  bincount,\n  booleanMaskAsync,\n  broadcastArgs,\n  broadcastTo,\n  broadcast_util_exports as broadcast_util,\n  browser_exports as browser,\n  buffer,\n  callbacks,\n  cast,\n  ceil,\n  clipByValue,\n  clone,\n  complex,\n  concat,\n  concat1d,\n  concat2d,\n  concat3d,\n  concat4d,\n  exports_constraints_exports as constraints,\n  conv1d,\n  conv2d,\n  conv2dTranspose,\n  conv3d,\n  conv3dTranspose,\n  copyRegisteredKernels,\n  cos,\n  cosh,\n  cosineWindow,\n  cumprod,\n  cumsum,\n  customGrad,\n  dist_exports2 as data,\n  denseBincount,\n  deprecationWarn,\n  depthToSpace,\n  depthwiseConv2d,\n  deregisterOp,\n  device_util_exports as device_util,\n  diag,\n  dilation2d,\n  disableDeprecationWarnings,\n  dispose,\n  disposeVariables,\n  div,\n  divNoNan,\n  dot,\n  dropout,\n  einsum,\n  elu,\n  enableDebugMode,\n  enableProdMode,\n  enclosingPowerOfTwo,\n  engine,\n  env,\n  equal,\n  erf,\n  euclideanNorm,\n  exp,\n  expandDims,\n  expm1,\n  eye,\n  fft,\n  fill,\n  findBackend,\n  findBackendFactory,\n  floor,\n  floorDiv,\n  forceHalfFloat,\n  fused_ops_exports as fused,\n  gather,\n  gatherND,\n  gather_nd_util_exports as gather_util,\n  getBackend,\n  getGradient,\n  getKernel,\n  getKernelsForBackend,\n  getThreadsCount,\n  gpgpu_util_exports as gpgpu_util,\n  grad,\n  grads,\n  greater,\n  greaterEqual,\n  ifft,\n  imag,\n  image,\n  inTopKAsync,\n  exports_initializers_exports as initializers,\n  input,\n  io_exports as io,\n  irfft,\n  isFinite2 as isFinite,\n  isInf,\n  isNaN2 as isNaN,\n  keep,\n  kernel_impls_exports as kernel_impls,\n  exports_layers_exports as layers,\n  leakyRelu,\n  less,\n  lessEqual,\n  linalg,\n  linspace,\n  loadGraphModel,\n  loadGraphModelSync,\n  loadLayersModel,\n  localResponseNormalization,\n  log2 as log,\n  log1p,\n  logSigmoid,\n  logSoftmax,\n  logSumExp,\n  logicalAnd,\n  logicalNot,\n  logicalOr,\n  logicalXor,\n  losses,\n  lowerBound,\n  matMul,\n  math_exports as math,\n  max,\n  maxPool,\n  maxPool3d,\n  maxPoolWithArgmax,\n  maximum,\n  mean,\n  memory,\n  meshgrid,\n  exports_metrics_exports as metrics,\n  min,\n  minimum,\n  mirrorPad,\n  mod,\n  model,\n  exports_models_exports as models,\n  moments,\n  movingAverage,\n  mul,\n  multiRNNCell,\n  multinomial,\n  neg,\n  nextFrame,\n  norm,\n  notEqual,\n  oneHot,\n  ones2 as ones,\n  onesLike,\n  op,\n  outerProduct,\n  pad,\n  pad1d,\n  pad2d,\n  pad3d,\n  pad4d,\n  pool,\n  pow,\n  prelu,\n  print,\n  prod,\n  profile,\n  raggedGather,\n  raggedRange,\n  raggedTensorToTensor,\n  rand,\n  randomGamma,\n  randomNormal,\n  randomStandardNormal,\n  randomUniform,\n  range,\n  ready,\n  real,\n  reciprocal,\n  registerBackend,\n  registerCallbackConstructor,\n  registerGradient,\n  registerKernel,\n  registerOp,\n  exports_regularizers_exports as regularizers,\n  relu,\n  relu6,\n  removeBackend,\n  reshape,\n  reverse,\n  reverse1d,\n  reverse2d,\n  reverse3d,\n  reverse4d,\n  rfft,\n  round2 as round,\n  rsqrt,\n  scalar,\n  scatterND,\n  scatter_nd_util_exports as scatter_util,\n  searchSorted,\n  selu,\n  separableConv2d,\n  sequential,\n  serialization_exports as serialization,\n  setBackend,\n  setPlatform,\n  setThreadsCount,\n  setWasmPath,\n  setWasmPaths,\n  setWebGLContext,\n  setdiff1dAsync,\n  sigmoid,\n  sign,\n  signal,\n  sin,\n  sinh,\n  slice,\n  slice1d,\n  slice2d,\n  slice3d,\n  slice4d,\n  slice_util_exports as slice_util,\n  softmax,\n  softplus,\n  spaceToBatchND,\n  sparse,\n  sparseToDense,\n  spectral,\n  split,\n  sqrt,\n  square,\n  squaredDifference,\n  squeeze,\n  stack,\n  step,\n  stridedSlice,\n  string,\n  sub,\n  sum2 as sum,\n  sumOutType,\n  tan,\n  tanh2 as tanh,\n  tensor,\n  tensor1d,\n  tensor2d,\n  tensor3d,\n  tensor4d,\n  tensor5d,\n  tensor6d,\n  tensor_util_exports as tensor_util,\n  test_util_exports as test_util,\n  tidy,\n  tile,\n  time,\n  topk,\n  train,\n  transpose,\n  truncatedNormal,\n  unique,\n  unregisterGradient,\n  unregisterKernel,\n  unsortedSegmentSum,\n  unstack,\n  upcastType,\n  upperBound,\n  util_exports as util,\n  valueAndGrad,\n  valueAndGrads,\n  variable,\n  variableGrads,\n  version62 as version,\n  version3 as version_converter,\n  version as version_core,\n  version2 as version_layers,\n  version8 as version_wasm,\n  version6 as version_webgl,\n  webgl,\n  webgl_util_exports as webgl_util,\n  where,\n  whereAsync,\n  zeros,\n  zerosLike\n};\n", "export * from './drawContour';\nexport * from './drawDetections';\nexport * from './drawFaceExpressions';\nexport * from './DrawBox';\nexport * from './DrawFaceLandmarks';\nexport * from './DrawTextField';\n", "import { Point } from '../classes/index';\n\nexport function drawContour(\n  ctx: CanvasRenderingContext2D,\n  points: Point[],\n  isClosed = false,\n) {\n  ctx.beginPath();\n\n  points.slice(1).forEach(({ x, y }, prevIdx) => {\n    const from = points[prevIdx];\n    ctx.moveTo(from.x, from.y);\n    ctx.lineTo(x, y);\n  });\n\n  if (isClosed) {\n    const from = points[points.length - 1];\n    const to = points[0];\n    if (!from || !to) {\n      return;\n    }\n\n    ctx.moveTo(from.x, from.y);\n    ctx.lineTo(to.x, to.y);\n  }\n\n  ctx.stroke();\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Point } from '../classes/index';\nimport { Dimensions, IDimensions } from '../classes/Dimensions';\n\nexport function isTensor(tensor: any, dim: number) {\n  return tensor instanceof tf.Tensor && tensor.shape.length === dim;\n}\n\nexport function isTensor1D(tensor: any): tensor is tf.Tensor1D {\n  return isTensor(tensor, 1);\n}\n\nexport function isTensor2D(tensor: any): tensor is tf.Tensor2D {\n  return isTensor(tensor, 2);\n}\n\nexport function isTensor3D(tensor: any): tensor is tf.Tensor3D {\n  return isTensor(tensor, 3);\n}\n\nexport function isTensor4D(tensor: any): tensor is tf.Tensor4D {\n  return isTensor(tensor, 4);\n}\n\nexport function isFloat(num: number) {\n  return num % 1 !== 0;\n}\n\nexport function isEven(num: number) {\n  return num % 2 === 0;\n}\n\nexport function round(num: number, prec = 2) {\n  const f = 10 ** prec;\n  return Math.floor(num * f) / f;\n}\n\nexport function isDimensions(obj: any): boolean {\n  return obj && obj.width && obj.height;\n}\n\nexport function computeReshapedDimensions({ width, height }: IDimensions, inputSize: number) {\n  const scale = inputSize / Math.max(height, width);\n  return new Dimensions(Math.round(width * scale), Math.round(height * scale));\n}\n\nexport function getCenterPoint(pts: Point[]): Point {\n  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0))\n    .div(new Point(pts.length, pts.length));\n}\n\nexport function range(num: number, start: number, step: number): number[] {\n  return Array(num).fill(0).map((_, i) => start + (i * step));\n}\n\nexport function isValidNumber(num: any) {\n  return !!num && (num !== Infinity) && (num !== -Infinity) && !Number.isNaN(num) || num === 0;\n}\n\nexport function isValidProbablitiy(num: any) {\n  return isValidNumber(num) && num >= 0 && num <= 1.0;\n}\n", "import { isValidNumber } from '../utils/index';\n\nexport interface IDimensions {\n  width: number\n  height: number\n}\n\nexport class Dimensions implements IDimensions {\n  private _width: number;\n\n  private _height: number;\n\n  constructor(width: number, height: number) {\n    if (!isValidNumber(width) || !isValidNumber(height)) {\n      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);\n    }\n\n    this._width = width;\n    this._height = height;\n  }\n\n  public get width(): number { return this._width; }\n\n  public get height(): number { return this._height; }\n\n  public reverse(): Dimensions {\n    return new Dimensions(1 / this.width, 1 / this.height);\n  }\n}\n", "export interface IPoint {\n  x: number\n  y: number\n}\n\nexport class Point implements IPoint {\n  private _x: number;\n\n  private _y: number;\n\n  constructor(x: number, y: number) {\n    this._x = x;\n    this._y = y;\n  }\n\n  get x(): number { return this._x; }\n\n  get y(): number { return this._y; }\n\n  public add(pt: IPoint): Point {\n    return new Point(this.x + pt.x, this.y + pt.y);\n  }\n\n  public sub(pt: IPoint): Point {\n    return new Point(this.x - pt.x, this.y - pt.y);\n  }\n\n  public mul(pt: IPoint): Point {\n    return new Point(this.x * pt.x, this.y * pt.y);\n  }\n\n  public div(pt: IPoint): Point {\n    return new Point(this.x / pt.x, this.y / pt.y);\n  }\n\n  public abs(): Point {\n    return new Point(Math.abs(this.x), Math.abs(this.y));\n  }\n\n  public magnitude(): number {\n    return Math.sqrt((this.x ** 2) + (this.y ** 2));\n  }\n\n  public floor(): Point {\n    return new Point(Math.floor(this.x), Math.floor(this.y));\n  }\n}\n", "import { isDimensions, isValidNumber } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { IDimensions } from './Dimensions';\nimport { Point } from './Point';\nimport { IRect } from './Rect';\n\nexport class Box<BoxType = any> implements IBoundingBox, IRect {\n  public static isRect(rect: any): boolean {\n    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);\n  }\n\n  public static assertIsValidBox(box: any, callee: string, allowNegativeDimensions = false) {\n    if (!Box.isRect(box)) {\n      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);\n    }\n\n    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {\n      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);\n    }\n  }\n\n  private _x: number;\n\n  private _y: number;\n\n  private _width: number;\n\n  private _height: number;\n\n  constructor(_box: IBoundingBox | IRect, allowNegativeDimensions = true) {\n    const box = (_box || {}) as any;\n\n    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);\n    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);\n\n    if (!isRect && !isBbox) {\n      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);\n    }\n\n    const [x, y, width, height] = isRect\n      ? [box.x, box.y, box.width, box.height]\n      : [box.left, box.top, box.right - box.left, box.bottom - box.top];\n\n    Box.assertIsValidBox({\n      x, y, width, height,\n    }, 'Box.constructor', allowNegativeDimensions);\n\n    this._x = x;\n    this._y = y;\n    this._width = width;\n    this._height = height;\n  }\n\n  public get x(): number { return this._x; }\n\n  public get y(): number { return this._y; }\n\n  public get width(): number { return this._width; }\n\n  public get height(): number { return this._height; }\n\n  public get left(): number { return this.x; }\n\n  public get top(): number { return this.y; }\n\n  public get right(): number { return this.x + this.width; }\n\n  public get bottom(): number { return this.y + this.height; }\n\n  public get area(): number { return this.width * this.height; }\n\n  public get topLeft(): Point { return new Point(this.left, this.top); }\n\n  public get topRight(): Point { return new Point(this.right, this.top); }\n\n  public get bottomLeft(): Point { return new Point(this.left, this.bottom); }\n\n  public get bottomRight(): Point { return new Point(this.right, this.bottom); }\n\n  public round(): Box<BoxType> {\n    const [x, y, width, height] = [this.x, this.y, this.width, this.height]\n      .map((val) => Math.round(val));\n    return new Box({\n      x, y, width, height,\n    });\n  }\n\n  public floor(): Box<BoxType> {\n    const [x, y, width, height] = [this.x, this.y, this.width, this.height]\n      .map((val) => Math.floor(val));\n    return new Box({\n      x, y, width, height,\n    });\n  }\n\n  public toSquare(): Box<BoxType> {\n    let {\n      x, y, width, height,\n    } = this;\n    const diff = Math.abs(width - height);\n    if (width < height) {\n      x -= (diff / 2);\n      width += diff;\n    }\n    if (height < width) {\n      y -= (diff / 2);\n      height += diff;\n    }\n\n    return new Box({ x, y, width, height });\n  }\n\n  public rescale(s: IDimensions | number): Box<BoxType> {\n    const scaleX = isDimensions(s) ? (s as IDimensions).width : s as number;\n    const scaleY = isDimensions(s) ? (s as IDimensions).height : s as number;\n    return new Box({\n      x: this.x * scaleX,\n      y: this.y * scaleY,\n      width: this.width * scaleX,\n      height: this.height * scaleY,\n    });\n  }\n\n  public pad(padX: number, padY: number): Box<BoxType> {\n    const [x, y, width, height] = [\n      this.x - (padX / 2),\n      this.y - (padY / 2),\n      this.width + padX,\n      this.height + padY,\n    ];\n    return new Box({ x, y, width, height });\n  }\n\n  public clipAtImageBorders(imgWidth: number, imgHeight: number): Box<BoxType> {\n    const { x, y, right, bottom } = this;\n    const clippedX = Math.max(x, 0);\n    const clippedY = Math.max(y, 0);\n\n    const newWidth = right - clippedX;\n    const newHeight = bottom - clippedY;\n    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);\n    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);\n\n    return (new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight })).floor();\n  }\n\n  public shift(sx: number, sy: number): Box<BoxType> {\n    const { width, height } = this;\n    const x = this.x + sx;\n    const y = this.y + sy;\n\n    return new Box({ x, y, width, height });\n  }\n\n  public padAtBorders(imageHeight: number, imageWidth: number) {\n    const w = this.width + 1;\n    const h = this.height + 1;\n\n    const dx = 1;\n    const dy = 1;\n    let edx = w;\n    let edy = h;\n\n    let x = this.left;\n    let y = this.top;\n    let ex = this.right;\n    let ey = this.bottom;\n\n    if (ex > imageWidth) {\n      edx = -ex + imageWidth + w;\n      ex = imageWidth;\n    }\n    if (ey > imageHeight) {\n      edy = -ey + imageHeight + h;\n      ey = imageHeight;\n    }\n    if (x < 1) {\n      edy = 2 - x;\n      x = 1;\n    }\n    if (y < 1) {\n      edy = 2 - y;\n      y = 1;\n    }\n\n    return { dy, edy, dx, edx, y, ey, x, ex, w, h };\n  }\n\n  public calibrate(region: Box) {\n    return new Box({\n      left: this.left + (region.left * this.width),\n      top: this.top + (region.top * this.height),\n      right: this.right + (region.right * this.width),\n      bottom: this.bottom + (region.bottom * this.height),\n    }).toSquare().round();\n  }\n}\n", "import { Box } from './Box';\n\nexport interface IBoundingBox {\n  left: number\n  top: number\n  right: number\n  bottom: number\n}\n\nexport class BoundingBox extends Box implements IBoundingBox {\n  constructor(left: number, top: number, right: number, bottom: number, allowNegativeDimensions = false) {\n    super({ left, top, right, bottom }, allowNegativeDimensions);\n  }\n}\n", "import { Box } from './Box';\nimport { Dimensions, IDimensions } from './Dimensions';\nimport { IRect, Rect } from './Rect';\n\nexport class ObjectDetection {\n  private _score: number;\n\n  private _classScore: number;\n\n  private _className: string;\n\n  private _box: Rect;\n\n  private _imageDims: Dimensions;\n\n  constructor(\n    score: number,\n    classScore: number,\n    className: string,\n    relativeBox: IRect,\n    imageDims: IDimensions,\n  ) {\n    this._imageDims = new Dimensions(imageDims.width, imageDims.height);\n    this._score = score;\n    this._classScore = classScore;\n    this._className = className;\n    this._box = new Box(relativeBox).rescale(this._imageDims);\n  }\n\n  public get score(): number { return this._score; }\n\n  public get classScore(): number { return this._classScore; }\n\n  public get className(): string { return this._className; }\n\n  public get box(): Box { return this._box; }\n\n  public get imageDims(): Dimensions { return this._imageDims; }\n\n  public get imageWidth(): number { return this.imageDims.width; }\n\n  public get imageHeight(): number { return this.imageDims.height; }\n\n  public get relativeBox(): Box { return new Box(this._box).rescale(this.imageDims.reverse()); }\n\n  public forSize(width: number, height: number): ObjectDetection {\n    return new ObjectDetection(\n      this.score,\n      this.classScore,\n      this.className,\n      this.relativeBox,\n      { width, height },\n    );\n  }\n}\n", "import { Box } from './Box';\nimport { IDimensions } from './Dimensions';\nimport { ObjectDetection } from './ObjectDetection';\nimport { Rect } from './Rect';\n\nexport interface IFaceDetecion {\n  score: number\n  box: Box\n}\n\nexport class FaceDetection extends ObjectDetection implements IFaceDetecion {\n  constructor(\n    score: number,\n    relativeBox: Rect,\n    imageDims: IDimensions,\n  ) {\n    super(score, score, '', relativeBox, imageDims);\n  }\n\n  public override forSize(width: number, height: number): FaceDetection {\n    const { score, relativeBox, imageDims } = super.forSize(width, height);\n    return new FaceDetection(score, relativeBox, imageDims);\n  }\n}\n", "import { Box } from '../classes/Box';\n\nexport function iou(box1: Box, box2: Box, isIOU = true) {\n  const width = Math.max(0.0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));\n  const height = Math.max(0.0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));\n  const interSection = width * height;\n\n  return isIOU\n    ? interSection / (box1.area + box2.area - interSection)\n    : interSection / Math.min(box1.area, box2.area);\n}\n", "import { BoundingBox, IPoint } from '../classes/index';\n\nexport function minBbox(pts: IPoint[]): BoundingBox {\n  const xs = pts.map((pt) => pt.x);\n  const ys = pts.map((pt) => pt.y);\n  const minX = xs.reduce((min, x) => (x < min ? x : min), Infinity);\n  const minY = ys.reduce((min, y) => (y < min ? y : min), Infinity);\n  const maxX = xs.reduce((max, x) => (max < x ? x : max), 0);\n  const maxY = ys.reduce((max, y) => (max < y ? y : max), 0);\n\n  return new BoundingBox(minX, minY, maxX, maxY);\n}\n", "import { Box } from '../classes/Box';\nimport { iou } from './iou';\n\nexport function nonMaxSuppression(\n  boxes: Box[],\n  scores: number[],\n  iouThreshold: number,\n  isIOU = true,\n): number[] {\n  let indicesSortedByScore = scores\n    .map((score, boxIndex) => ({ score, boxIndex }))\n    .sort((c1, c2) => c1.score - c2.score)\n    .map((c) => c.boxIndex);\n\n  const pick: number[] = [];\n\n  while (indicesSortedByScore.length > 0) {\n    const curr = indicesSortedByScore.pop() as number;\n    pick.push(curr);\n\n    const indices = indicesSortedByScore;\n\n    const outputs: number[] = [];\n    for (let i = 0; i < indices.length; i++) {\n      const idx = indices[i];\n\n      const currBox = boxes[curr];\n      const idxBox = boxes[idx];\n\n      outputs.push(iou(currBox, idxBox, isIOU));\n    }\n\n    indicesSortedByScore = indicesSortedByScore.filter(\n      (_, j) => outputs[j] <= iouThreshold,\n    );\n  }\n\n  return pick;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function normalize(x: tf.Tensor4D, meanRgb: number[]): tf.Tensor4D {\n  return tf.tidy(() => {\n    const [r, g, b] = meanRgb;\n    const avg_r = tf.fill([...x.shape.slice(0, 3), 1], r, 'float32');\n    const avg_g = tf.fill([...x.shape.slice(0, 3), 1], g, 'float32');\n    const avg_b = tf.fill([...x.shape.slice(0, 3), 1], b, 'float32');\n    const avg_rgb = tf.concat([avg_r, avg_g, avg_b], 3);\n\n    return tf.sub(x, avg_rgb);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\n/**\n * Pads the smaller dimension of an image tensor with zeros, such that width === height.\n *\n * @param imgTensor The image tensor.\n * @param isCenterImage (optional, default: false) If true, add an equal amount of padding on\n * both sides of the minor dimension oof the image.\n * @returns The padded tensor with width === height.\n */\nexport function padToSquare(imgTensor: tf.Tensor4D, isCenterImage = false): tf.Tensor4D {\n  return tf.tidy(() => {\n    const [height, width] = imgTensor.shape.slice(1);\n    if (height === width) return imgTensor;\n    const dimDiff = Math.abs(height - width);\n    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));\n    const paddingAxis = height > width ? 2 : 1;\n    const createPaddingTensor = (paddingAmountLocal: number): tf.Tensor => {\n      const paddingTensorShape = imgTensor.shape.slice();\n      paddingTensorShape[paddingAxis] = paddingAmountLocal;\n      return tf.fill(paddingTensorShape, 0, 'float32');\n    };\n    const paddingTensorAppend = createPaddingTensor(paddingAmount);\n    const remainingPaddingAmount = dimDiff - (paddingTensorAppend.shape[paddingAxis] as number);\n    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;\n    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend]\n      .filter((t) => !!t)\n      .map((t) => tf.cast(t as tf.Tensor4D, 'float32')) as tf.Tensor4D[];\n    return tf.concat(tensorsToStack, paddingAxis);\n  });\n}\n", "export function shuffleArray(inputArray: any[]) {\n  const array = inputArray.slice();\n  for (let i = array.length - 1; i > 0; i--) {\n    const j = Math.floor(Math.random() * (i + 1));\n    const x = array[i];\n    array[i] = array[j];\n    array[j] = x;\n  }\n  return array;\n}\n", "export * from './iou';\nexport * from './minBbox';\nexport * from './nonMaxSuppression';\nexport * from './normalize';\nexport * from './padToSquare';\nexport * from './shuffleArray';\n\nexport function sigmoid(x: number) {\n  return 1 / (1 + Math.exp(-x));\n}\n\nexport function inverseSigmoid(x: number) {\n  return Math.log(x / (1 - x));\n}\n", "import { Box } from './Box';\n\nexport interface IRect {\n  x: number\n  y: number\n  width: number\n  height: number\n}\n\nexport class Rect extends Box implements IRect {\n  constructor(x: number, y: number, width: number, height: number, allowNegativeDimensions = false) {\n    super({ x, y, width, height }, allowNegativeDimensions);\n  }\n}\n", "import { minBbox } from '../ops/index';\nimport { getCenterPoint } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { Box } from './Box';\nimport { Dimensions, IDimensions } from './Dimensions';\nimport { FaceDetection } from './FaceDetection';\nimport { Point } from './Point';\nimport { IRect, Rect } from './Rect';\n\n// face alignment constants\nconst relX = 0.5;\nconst relY = 0.43;\nconst relScale = 0.45;\n\nexport interface IFaceLandmarks {\n  positions: Point[]\n  shift: Point\n}\n\nexport class FaceLandmarks implements IFaceLandmarks {\n  protected _shift: Point;\n\n  protected _positions: Point[];\n\n  protected _imgDims: Dimensions;\n\n  constructor(\n    relativeFaceLandmarkPositions: Point[],\n    imgDims: IDimensions,\n    shift: Point = new Point(0, 0),\n  ) {\n    const { width, height } = imgDims;\n    this._imgDims = new Dimensions(width, height);\n    this._shift = shift;\n    this._positions = relativeFaceLandmarkPositions.map(\n      (pt) => pt.mul(new Point(width, height)).add(shift),\n    );\n  }\n\n  public get shift(): Point { return new Point(this._shift.x, this._shift.y); }\n\n  public get imageWidth(): number { return this._imgDims.width; }\n\n  public get imageHeight(): number { return this._imgDims.height; }\n\n  public get positions(): Point[] { return this._positions; }\n\n  public get relativePositions(): Point[] {\n    return this._positions.map(\n      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight)),\n    );\n  }\n\n  public forSize<T extends FaceLandmarks>(width: number, height: number): T {\n    return new (this.constructor as any)(\n      this.relativePositions,\n      { width, height },\n    );\n  }\n\n  public shiftBy<T extends FaceLandmarks>(x: number, y: number): T {\n    return new (this.constructor as any)(\n      this.relativePositions,\n      this._imgDims,\n      new Point(x, y),\n    );\n  }\n\n  public shiftByPoint<T extends FaceLandmarks>(pt: Point): T {\n    return this.shiftBy(pt.x, pt.y);\n  }\n\n  /**\n   * Aligns the face landmarks after face detection from the relative positions of the faces\n   * bounding box, or it's current shift. This function should be used to align the face images\n   * after face detection has been performed, before they are passed to the face recognition net.\n   * This will make the computed face descriptor more accurate.\n   *\n   * @param detection (optional) The bounding box of the face or the face detection result. If\n   * no argument was passed the position of the face landmarks are assumed to be relative to\n   * it's current shift.\n   * @returns The bounding box of the aligned face.\n   */\n  public align(\n    detection?: FaceDetection | IRect | IBoundingBox | null,\n    options: { useDlibAlignment?: boolean, minBoxPadding?: number } = { },\n  ): Box {\n    if (detection) {\n      const box = detection instanceof FaceDetection\n        ? detection.box.floor()\n        : new Box(detection);\n\n      return this.shiftBy(box.x, box.y).align(null, options);\n    }\n\n    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };\n\n    if (useDlibAlignment) {\n      return this.alignDlib();\n    }\n\n    return this.alignMinBbox(minBoxPadding);\n  }\n\n  private alignDlib(): Box {\n    const centers = this.getRefPointsForAlignment();\n\n    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;\n    const distToMouth = (pt: Point) => mouthCenter.sub(pt).magnitude();\n    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;\n\n    const size = Math.floor(eyeToMouthDist / relScale);\n\n    const refPoint = getCenterPoint(centers);\n    // TODO: pad in case rectangle is out of image bounds\n    const x = Math.floor(Math.max(0, refPoint.x - (relX * size)));\n    const y = Math.floor(Math.max(0, refPoint.y - (relY * size)));\n\n    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));\n  }\n\n  private alignMinBbox(padding: number): Box {\n    const box = minBbox(this.positions);\n    return box.pad(box.width * padding, box.height * padding);\n  }\n\n  protected getRefPointsForAlignment(): Point[] {\n    throw new Error('getRefPointsForAlignment not implemented by base class');\n  }\n}\n", "import { getCenterPoint } from '../utils/index';\nimport { FaceLandmarks } from './FaceLandmarks';\nimport { Point } from './Point';\n\nexport class FaceLandmarks5 extends FaceLandmarks {\n  protected override getRefPointsForAlignment(): Point[] {\n    const pts = this.positions;\n    return [\n      pts[0],\n      pts[1],\n      getCenterPoint([pts[3], pts[4]]),\n    ];\n  }\n}\n", "import { getCenterPoint } from '../utils/index';\nimport { FaceLandmarks } from './FaceLandmarks';\nimport { Point } from './Point';\n\nexport class FaceLandmarks68 extends FaceLandmarks {\n  public getJawOutline(): Point[] {\n    return this.positions.slice(0, 17);\n  }\n\n  public getLeftEyeBrow(): Point[] {\n    return this.positions.slice(17, 22);\n  }\n\n  public getRightEyeBrow(): Point[] {\n    return this.positions.slice(22, 27);\n  }\n\n  public getNose(): Point[] {\n    return this.positions.slice(27, 36);\n  }\n\n  public getLeftEye(): Point[] {\n    return this.positions.slice(36, 42);\n  }\n\n  public getRightEye(): Point[] {\n    return this.positions.slice(42, 48);\n  }\n\n  public getMouth(): Point[] {\n    return this.positions.slice(48, 68);\n  }\n\n  protected override getRefPointsForAlignment(): Point[] {\n    return [\n      this.getLeftEye(),\n      this.getRightEye(),\n      this.getMouth(),\n    ].map(getCenterPoint);\n  }\n}\n", "import { round } from '../utils/index';\n\nexport interface IFaceMatch {\n  label: string\n  distance: number\n}\n\nexport class FaceMatch implements IFaceMatch {\n  private _label: string;\n  private _distance: number;\n\n  constructor(label: string, distance: number) {\n    this._label = label;\n    this._distance = distance;\n  }\n\n  public get label(): string { return this._label; }\n\n  public get distance(): number { return this._distance; }\n\n  public toString(withDistance = true): string {\n    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ''}`;\n  }\n}\n", "import { isValidNumber } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { Box } from './Box';\nimport { IRect } from './Rect';\n\nexport class LabeledBox extends Box {\n  public static assertIsValidLabeledBox(box: any, callee: string) {\n    Box.assertIsValidBox(box, callee);\n    if (!isValidNumber(box.label)) {\n      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);\n    }\n  }\n\n  private _label: number;\n\n  constructor(box: IBoundingBox | IRect | any, label: number) {\n    super(box);\n    this._label = label;\n  }\n\n  public get label(): number { return this._label; }\n}\n", "export class LabeledFaceDescriptors {\n  private _label: string;\n\n  private _descriptors: Float32Array[];\n\n  constructor(label: string, descriptors: Float32Array[]) {\n    if (!(typeof label === 'string')) {\n      throw new Error('LabeledFaceDescriptors - constructor expected label to be a string');\n    }\n\n    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {\n      throw new Error('LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array');\n    }\n\n    this._label = label;\n    this._descriptors = descriptors;\n  }\n\n  public get label(): string { return this._label; }\n\n  public get descriptors(): Float32Array[] { return this._descriptors; }\n\n  public toJSON(): any {\n    return {\n      label: this.label,\n      descriptors: this.descriptors.map((d) => Array.from(d)),\n    };\n  }\n\n  public static fromJSON(json: any): LabeledFaceDescriptors {\n    const descriptors = json.descriptors.map((d: any) => new Float32Array(d));\n    return new LabeledFaceDescriptors(json.label, descriptors);\n  }\n}\n", "import { isValidProbablitiy } from '../utils/index';\nimport { IBoundingBox } from './BoundingBox';\nimport { LabeledBox } from './LabeledBox';\nimport { IRect } from './Rect';\n\nexport class PredictedBox extends LabeledBox {\n  public static assertIsValidPredictedBox(box: any, callee: string) {\n    LabeledBox.assertIsValidLabeledBox(box, callee);\n\n    if (\n      !isValidProbablitiy(box.score)\n      || !isValidProbablitiy(box.classScore)\n    ) {\n      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);\n    }\n  }\n\n  private _score: number;\n\n  private _classScore: number;\n\n  constructor(box: IBoundingBox | IRect | any, label: number, score: number, classScore: number) {\n    super(box, label);\n    this._score = score;\n    this._classScore = classScore;\n  }\n\n  public get score(): number { return this._score; }\n\n  public get classScore(): number { return this._classScore; }\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\n\nexport type WithFaceDetection<TSource> = TSource & {\n  detection: FaceDetection\n}\n\nexport function isWithFaceDetection(obj: any): obj is WithFaceDetection<{}> {\n  return obj.detection instanceof FaceDetection;\n}\n\nexport function extendWithFaceDetection<TSource>(sourceObj: TSource, detection: FaceDetection): WithFaceDetection<TSource> {\n  const extension = { detection };\n  return { ...sourceObj, ...extension };\n}\n", "import { Environment } from './types';\n\nexport function createBrowserEnv(): Environment {\n  const fetch = window.fetch;\n  if (!fetch) throw new Error('fetch - missing fetch implementation for browser environment');\n\n  const readFile = () => {\n    throw new Error('readFile - filesystem not available for browser environment');\n  };\n\n  return {\n    Canvas: HTMLCanvasElement,\n    CanvasRenderingContext2D,\n    Image: HTMLImageElement,\n    ImageData,\n    Video: HTMLVideoElement,\n    createCanvasElement: () => document.createElement('canvas'),\n    createImageElement: () => document.createElement('img'),\n    createVideoElement: () => document.createElement('video'),\n    fetch,\n    readFile,\n  };\n}\n", "export function isNodejs(): boolean {\n  return typeof global === 'object'\n    && typeof process !== 'undefined'\n    && process.versions != null\n    && process.versions.node != null;\n}\n", "import { FileSystem } from './types';\nimport { isNodejs } from './isNodejs';\n\nexport function createFileSystem(fs?: any): FileSystem {\n  let requireFsError = '';\n  if (!fs && isNodejs()) {\n    try {\n      // eslint-disable-next-line global-require\n      fs = require('fs');\n    } catch (err) {\n      requireFsError = (err as any).toString();\n    }\n  }\n\n  const readFile = fs\n    ? (filePath: string) => new Promise((resolve, reject) => { fs.readFile(filePath, (err: any, buffer) => (err ? reject(err) : resolve(buffer))); })\n    : () => { throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`); };\n  return { readFile };\n}\n", "/* eslint-disable max-classes-per-file */\nimport { createFileSystem } from './createFileSystem';\nimport { Environment } from './types';\n\nexport function createNodejsEnv(): Environment {\n  // eslint-disable-next-line dot-notation\n  const Canvas = global['Canvas'] || global.HTMLCanvasElement;\n  const Image = global.Image || global.HTMLImageElement;\n  // eslint-disable-next-line dot-notation\n  const Video = global['Video'] || global.HTMLVideoElement;\n\n  const createCanvasElement = () => {\n    if (Canvas) return new Canvas();\n    throw new Error('createCanvasElement - missing Canvas implementation for nodejs environment');\n  };\n\n  const createImageElement = () => {\n    if (Image) return new Image();\n    throw new Error('createImageElement - missing Image implementation for nodejs environment');\n  };\n\n  const createVideoElement = () => {\n    if (Video) return new Video();\n    throw new Error('createVideoElement - missing Video implementation for nodejs environment');\n  };\n\n  const fetch = global.fetch;\n  // if (!fetch) throw new Error('fetch - missing fetch implementation for nodejs environment');\n\n  const fileSystem = createFileSystem();\n\n  return {\n    Canvas: Canvas || class {},\n    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {},\n    Image: Image || class {},\n    ImageData: global.ImageData || class {},\n    Video: global.HTMLVideoElement || class {},\n    createCanvasElement,\n    createImageElement,\n    createVideoElement,\n    fetch,\n    ...fileSystem,\n  };\n}\n", "export function isBrowser(): boolean {\n  return typeof window === 'object'\n    && typeof document !== 'undefined'\n    && typeof HTMLImageElement !== 'undefined'\n    && typeof HTMLCanvasElement !== 'undefined'\n    && typeof HTMLVideoElement !== 'undefined'\n    && typeof ImageData !== 'undefined'\n    && typeof CanvasRenderingContext2D !== 'undefined';\n}\n", "import { createBrowserEnv } from './createBrowserEnv';\nimport { createFileSystem } from './createFileSystem';\nimport { createNodejsEnv } from './createNodejsEnv';\nimport { isBrowser } from './isBrowser';\nimport { isNodejs } from './isNodejs';\nimport { Environment } from './types';\n\nlet environment: Environment | null;\n\nfunction getEnv(): Environment {\n  if (!environment) {\n    throw new Error('getEnv - environment is not defined, check isNodejs() and isBrowser()');\n  }\n  return environment;\n}\n\nfunction setEnv(env: Environment) {\n  environment = env;\n}\n\nfunction initialize() {\n  // check for isBrowser() first to prevent electron renderer process\n  // to be initialized with wrong environment due to isNodejs() returning true\n  if (isBrowser()) return setEnv(createBrowserEnv());\n  if (isNodejs()) return setEnv(createNodejsEnv());\n  return null;\n}\n\nfunction monkeyPatch(env: Partial<Environment>) {\n  if (!environment) {\n    initialize();\n  }\n\n  if (!environment) {\n    throw new Error('monkeyPatch - environment is not defined, check isNodejs() and isBrowser()');\n  }\n\n  const { Canvas = environment.Canvas, Image = environment.Image } = env;\n  environment.Canvas = Canvas;\n  environment.Image = Image;\n  environment.createCanvasElement = env.createCanvasElement || (() => new Canvas());\n  environment.createImageElement = env.createImageElement || (() => new Image());\n\n  environment.ImageData = env.ImageData || environment.ImageData;\n  environment.Video = env.Video || environment.Video;\n  environment.fetch = env.fetch || environment.fetch;\n  environment.readFile = env.readFile || environment.readFile;\n}\n\nexport const env = {\n  getEnv,\n  setEnv,\n  initialize,\n  createBrowserEnv,\n  createFileSystem,\n  createNodejsEnv,\n  monkeyPatch,\n  isBrowser,\n  isNodejs,\n};\n\ninitialize();\n\nexport * from './types';\n", "import { env } from '../env/index';\n\nexport function resolveInput(arg: string | any) {\n  if (!env.isNodejs() && typeof arg === 'string') {\n    return document.getElementById(arg);\n  }\n  return arg;\n}\n", "import { env } from '../env/index';\nimport { resolveInput } from './resolveInput';\n\nexport function getContext2dOrThrow(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D): CanvasRenderingContext2D {\n  const { Canvas, CanvasRenderingContext2D } = env.getEnv();\n\n  if (canvasArg instanceof CanvasRenderingContext2D) {\n    return canvasArg;\n  }\n\n  const canvas = resolveInput(canvasArg);\n\n  if (!(canvas instanceof Canvas)) {\n    throw new Error('resolveContext2d - expected canvas to be of instance of Canvas');\n  }\n\n  const ctx = canvas.getContext('2d');\n  if (!ctx) {\n    throw new Error('resolveContext2d - canvas 2d context is null');\n  }\n\n  return ctx;\n}\n", "/* eslint-disable max-classes-per-file */\nimport { IDimensions, IPoint } from '../classes/index';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { resolveInput } from '../dom/resolveInput';\n\n// eslint-disable-next-line no-shadow\nexport enum AnchorPosition {\n  // eslint-disable-next-line no-unused-vars\n  TOP_LEFT = 'TOP_LEFT',\n  // eslint-disable-next-line no-unused-vars\n  TOP_RIGHT = 'TOP_RIGHT',\n  // eslint-disable-next-line no-unused-vars\n  BOTTOM_LEFT = 'BOTTOM_LEFT',\n  // eslint-disable-next-line no-unused-vars\n  BOTTOM_RIGHT = 'BOTTOM_RIGHT'\n}\n\nexport interface IDrawTextFieldOptions {\n  anchorPosition?: AnchorPosition\n  backgroundColor?: string\n  fontColor?: string\n  fontSize?: number\n  fontStyle?: string\n  padding?: number\n}\n\nexport class DrawTextFieldOptions implements IDrawTextFieldOptions {\n  public anchorPosition: AnchorPosition;\n\n  public backgroundColor: string;\n\n  public fontColor: string;\n\n  public fontSize: number;\n\n  public fontStyle: string;\n\n  public padding: number;\n\n  constructor(options: IDrawTextFieldOptions = {}) {\n    const {\n      anchorPosition, backgroundColor, fontColor, fontSize, fontStyle, padding,\n    } = options;\n    this.anchorPosition = anchorPosition || AnchorPosition.TOP_LEFT;\n    this.backgroundColor = backgroundColor || 'rgba(0, 0, 0, 0.5)';\n    this.fontColor = fontColor || 'rgba(255, 255, 255, 1)';\n    this.fontSize = fontSize || 14;\n    this.fontStyle = fontStyle || 'Georgia';\n    this.padding = padding || 4;\n  }\n}\n\nexport class DrawTextField {\n  public text: string[];\n\n  public anchor : IPoint;\n\n  public options: DrawTextFieldOptions;\n\n  constructor(\n    text: string | string[] | DrawTextField,\n    anchor: IPoint,\n    options: IDrawTextFieldOptions = {},\n  ) {\n    // eslint-disable-next-line no-nested-ternary\n    this.text = typeof text === 'string'\n      ? [text]\n      : (text instanceof DrawTextField ? text.text : text);\n    this.anchor = anchor;\n    this.options = new DrawTextFieldOptions(options);\n  }\n\n  measureWidth(ctx: CanvasRenderingContext2D): number {\n    const { padding } = this.options;\n    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => (w0 < w1 ? w1 : w0), 0) + (2 * padding);\n  }\n\n  measureHeight(): number {\n    const { fontSize, padding } = this.options;\n    return this.text.length * fontSize + (2 * padding);\n  }\n\n  getUpperLeft(ctx: CanvasRenderingContext2D, canvasDims?: IDimensions): IPoint {\n    const { anchorPosition } = this.options;\n    const isShiftLeft = anchorPosition === AnchorPosition.BOTTOM_RIGHT || anchorPosition === AnchorPosition.TOP_RIGHT;\n    const isShiftTop = anchorPosition === AnchorPosition.BOTTOM_LEFT || anchorPosition === AnchorPosition.BOTTOM_RIGHT;\n\n    const textFieldWidth = this.measureWidth(ctx);\n    const textFieldHeight = this.measureHeight();\n    const x = (isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x);\n    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;\n\n    // adjust anchor if text box exceeds canvas borders\n    if (canvasDims) {\n      const { width, height } = canvasDims;\n      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);\n      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);\n      return { x: newX, y: newY };\n    }\n    return { x, y };\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const canvas = resolveInput(canvasArg);\n    const ctx = getContext2dOrThrow(canvas);\n\n    const {\n      backgroundColor, fontColor, fontSize, fontStyle, padding,\n    } = this.options;\n\n    ctx.font = `${fontSize}px ${fontStyle}`;\n    const maxTextWidth = this.measureWidth(ctx);\n    const textHeight = this.measureHeight();\n\n    ctx.fillStyle = backgroundColor;\n    const upperLeft = this.getUpperLeft(ctx, canvas);\n    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);\n\n    ctx.fillStyle = fontColor;\n    this.text.forEach((textLine, i) => {\n      const x = padding + upperLeft.x;\n      const y = padding + upperLeft.y + ((i + 1) * fontSize);\n      ctx.fillText(textLine, x, y);\n    });\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { Box, IBoundingBox, IRect } from '../classes/index';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { AnchorPosition, DrawTextField, DrawTextFieldOptions, IDrawTextFieldOptions } from './DrawTextField';\n\nexport interface IDrawBoxOptions {\n  boxColor?: string\n  lineWidth?: number\n  drawLabelOptions?: IDrawTextFieldOptions\n  label?: string\n}\n\nexport class DrawBoxOptions {\n  public boxColor: string;\n\n  public lineWidth: number;\n\n  public drawLabelOptions: DrawTextFieldOptions;\n\n  public label?: string;\n\n  constructor(options: IDrawBoxOptions = {}) {\n    const {\n      boxColor, lineWidth, label, drawLabelOptions,\n    } = options;\n    this.boxColor = boxColor || 'rgba(0, 0, 255, 1)';\n    this.lineWidth = lineWidth || 2;\n    this.label = label;\n\n    const defaultDrawLabelOptions = {\n      anchorPosition: AnchorPosition.BOTTOM_LEFT,\n      backgroundColor: this.boxColor,\n    };\n    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });\n  }\n}\n\nexport class DrawBox {\n  public box: Box;\n\n  public options: DrawBoxOptions;\n\n  constructor(\n    box: IBoundingBox | IRect,\n    options: IDrawBoxOptions = {},\n  ) {\n    this.box = new Box(box);\n    this.options = new DrawBoxOptions(options);\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const ctx = getContext2dOrThrow(canvasArg);\n\n    const { boxColor, lineWidth } = this.options;\n\n    const {\n      x, y, width, height,\n    } = this.box;\n    ctx.strokeStyle = boxColor;\n    ctx.lineWidth = lineWidth;\n    ctx.strokeRect(x, y, width, height);\n\n    const { label } = this.options;\n    if (label) {\n      new DrawTextField([label], { x: x - (lineWidth / 2), y }, this.options.drawLabelOptions).draw(canvasArg);\n    }\n  }\n}\n", "import { Box, IBoundingBox, IRect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { isWithFaceDetection, WithFaceDetection } from '../factories/WithFaceDetection';\nimport { round } from '../utils/index';\nimport { DrawBox } from './DrawBox';\n\nexport type TDrawDetectionsInput = IRect | IBoundingBox | FaceDetection | WithFaceDetection<{}>\n\nexport function drawDetections(\n  canvasArg: string | HTMLCanvasElement,\n  detections: TDrawDetectionsInput | Array<TDrawDetectionsInput>,\n) {\n  const detectionsArray = Array.isArray(detections) ? detections : [detections];\n\n  detectionsArray.forEach((det) => {\n    // eslint-disable-next-line no-nested-ternary\n    const score = det instanceof FaceDetection\n      ? det.score\n      : (isWithFaceDetection(det) ? det.detection.score : undefined);\n\n    // eslint-disable-next-line no-nested-ternary\n    const box = det instanceof FaceDetection\n      ? det.box\n      : (isWithFaceDetection(det) ? det.detection.box : new Box(det));\n\n    const label = score ? `${round(score)}` : undefined;\n    new DrawBox(box, { label }).draw(canvasArg);\n  });\n}\n", "import { env } from '../env/index';\n\nexport function isMediaLoaded(media: HTMLImageElement | HTMLVideoElement) : boolean {\n  const { Image, Video } = env.getEnv();\n\n  return (media instanceof Image && media.complete)\n    || (media instanceof Video && media.readyState >= 3);\n}\n", "import { env } from '../env/index';\nimport { isMediaLoaded } from './isMediaLoaded';\n\nexport function awaitMediaLoaded(media: HTMLImageElement | HTMLVideoElement | HTMLCanvasElement) {\n  // eslint-disable-next-line consistent-return\n  return new Promise((resolve, reject) => {\n    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media)) resolve(null);\n\n    function onError(e: Event) {\n      if (!e.currentTarget) return;\n      // eslint-disable-next-line no-use-before-define\n      e.currentTarget.removeEventListener('load', onLoad);\n      e.currentTarget.removeEventListener('error', onError);\n      reject(e);\n    }\n\n    function onLoad(e: Event) {\n      if (!e.currentTarget) return;\n      e.currentTarget.removeEventListener('load', onLoad);\n      e.currentTarget.removeEventListener('error', onError);\n      resolve(e);\n    }\n\n    media.addEventListener('load', onLoad);\n    media.addEventListener('error', onError);\n  });\n}\n", "import { env } from '../env/index';\n\nexport function bufferToImage(buf: Blob): Promise<HTMLImageElement> {\n  return new Promise((resolve, reject) => {\n    if (!(buf instanceof Blob)) reject(new Error('bufferToImage - expected buf to be of type: Blob'));\n    const reader = new FileReader();\n    reader.onload = () => {\n      if (typeof reader.result !== 'string') reject(new Error('bufferToImage - expected reader.result to be a string, in onload'));\n      const img = env.getEnv().createImageElement();\n      img.onload = () => resolve(img);\n      img.onerror = reject;\n      img.src = reader.result as string;\n    };\n    reader.onerror = reject;\n    reader.readAsDataURL(buf);\n  });\n}\n", "import { Dimensions, IDimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\n\nexport function getMediaDimensions(input: HTMLImageElement | HTMLCanvasElement | HTMLVideoElement | IDimensions): Dimensions {\n  const { Image, Video } = env.getEnv();\n\n  if (input instanceof Image) {\n    return new Dimensions(input.naturalWidth, input.naturalHeight);\n  }\n  if (input instanceof Video) {\n    return new Dimensions(input.videoWidth, input.videoHeight);\n  }\n  return new Dimensions(input.width, input.height);\n}\n", "import { IDimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { getMediaDimensions } from './getMediaDimensions';\nimport { isMediaLoaded } from './isMediaLoaded';\n\nexport function createCanvas({ width, height }: IDimensions): HTMLCanvasElement {\n  const { createCanvasElement } = env.getEnv();\n  const canvas = createCanvasElement();\n  canvas.width = width;\n  canvas.height = height;\n  return canvas;\n}\n\nexport function createCanvasFromMedia(media: HTMLImageElement | HTMLVideoElement | ImageData, dims?: IDimensions): HTMLCanvasElement {\n  const { ImageData } = env.getEnv();\n\n  if (!(media instanceof ImageData) && !isMediaLoaded(media)) {\n    throw new Error('createCanvasFromMedia - media has not finished loading yet');\n  }\n\n  const { width, height } = dims || getMediaDimensions(media);\n  const canvas = createCanvas({ width, height });\n\n  if (media instanceof ImageData) {\n    getContext2dOrThrow(canvas).putImageData(media, 0, 0);\n  } else {\n    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);\n  }\n  return canvas;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { env } from '../env/index';\nimport { isTensor4D } from '../utils/index';\n\nexport async function imageTensorToCanvas(\n  imgTensor: tf.Tensor,\n  canvas?: HTMLCanvasElement,\n): Promise<HTMLCanvasElement> {\n  const targetCanvas = canvas || env.getEnv().createCanvasElement();\n\n  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);\n  const imgTensor3D = tf.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());\n  await tf['browser'].toPixels(imgTensor3D, targetCanvas);\n\n  imgTensor3D.dispose();\n\n  return targetCanvas;\n}\n", "import { env } from '../env/index';\n\nexport function isMediaElement(input: any) {\n  const { Image, Canvas, Video } = env.getEnv();\n\n  return input instanceof Image\n    || input instanceof Canvas\n    || input instanceof Video;\n}\n", "import { env } from '../env/index';\nimport { createCanvas, createCanvasFromMedia } from './createCanvas';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { getMediaDimensions } from './getMediaDimensions';\n\nexport function imageToSquare(input: HTMLImageElement | HTMLCanvasElement, inputSize: number, centerImage = false) {\n  const { Image, Canvas } = env.getEnv();\n\n  if (!(input instanceof Image || input instanceof Canvas)) {\n    throw new Error('imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement');\n  }\n\n  if (inputSize <= 0) return createCanvas({ width: 1, height: 1 });\n  const dims = getMediaDimensions(input);\n  const scale = inputSize / Math.max(dims.height, dims.width);\n  const width = scale * dims.width;\n  const height = scale * dims.height;\n\n  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });\n  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);\n\n  const offset = Math.abs(width - height) / 2;\n  const dx = centerImage && width < height ? offset : 0;\n  const dy = centerImage && height < width ? offset : 0;\n  if (inputCanvas.width > 0 && inputCanvas.height > 0) getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);\n\n  return targetCanvas;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Dimensions } from '../classes/Dimensions';\nimport { env } from '../env/index';\nimport { padToSquare } from '../ops/padToSquare';\nimport { computeReshapedDimensions, isTensor3D, isTensor4D, range } from '../utils/index';\nimport { createCanvasFromMedia } from './createCanvas';\nimport { imageToSquare } from './imageToSquare';\nimport { TResolvedNetInput } from './types';\n\nexport class NetInput {\n  private _imageTensors: Array<tf.Tensor3D | tf.Tensor4D> = [];\n\n  private _canvases: HTMLCanvasElement[] = [];\n\n  private _batchSize: number;\n\n  private _treatAsBatchInput = false;\n\n  private _inputDimensions: number[][] = [];\n\n  private _inputSize = 0;\n\n  constructor(inputs: Array<TResolvedNetInput>, treatAsBatchInput = false) {\n    if (!Array.isArray(inputs)) {\n      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);\n    }\n\n    this._treatAsBatchInput = treatAsBatchInput;\n    this._batchSize = inputs.length;\n\n    inputs.forEach((input, idx) => {\n      if (isTensor3D(input)) {\n        this._imageTensors[idx] = input;\n        this._inputDimensions[idx] = input.shape;\n        return;\n      }\n\n      if (isTensor4D(input)) {\n        const batchSize = (input as any).shape[0];\n        if (batchSize !== 1) {\n          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);\n        }\n\n        this._imageTensors[idx] = input;\n        this._inputDimensions[idx] = (input as any).shape.slice(1);\n        return;\n      }\n\n      // @ts-ignore\n      const canvas = (input as any) instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);\n      this._canvases[idx] = canvas as HTMLCanvasElement;\n      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];\n    });\n  }\n\n  public get imageTensors(): Array<tf.Tensor3D | tf.Tensor4D> {\n    return this._imageTensors;\n  }\n\n  public get canvases(): HTMLCanvasElement[] {\n    return this._canvases;\n  }\n\n  public get isBatchInput(): boolean {\n    return this.batchSize > 1 || this._treatAsBatchInput;\n  }\n\n  public get batchSize(): number {\n    return this._batchSize;\n  }\n\n  public get inputDimensions(): number[][] {\n    return this._inputDimensions;\n  }\n\n  public get inputSize(): number | undefined {\n    return this._inputSize;\n  }\n\n  public get reshapedInputDimensions(): Dimensions[] {\n    return range(this.batchSize, 0, 1).map(\n      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx),\n    );\n  }\n\n  public getInput(batchIdx: number): tf.Tensor3D | tf.Tensor4D | HTMLCanvasElement {\n    return this.canvases[batchIdx] || this.imageTensors[batchIdx];\n  }\n\n  public getInputDimensions(batchIdx: number): number[] {\n    return this._inputDimensions[batchIdx];\n  }\n\n  public getInputHeight(batchIdx: number): number {\n    return this._inputDimensions[batchIdx][0];\n  }\n\n  public getInputWidth(batchIdx: number): number {\n    return this._inputDimensions[batchIdx][1];\n  }\n\n  public getReshapedInputDimensions(batchIdx: number): Dimensions {\n    if (typeof this.inputSize !== 'number') {\n      throw new Error('getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet');\n    }\n\n    const width = this.getInputWidth(batchIdx);\n    const height = this.getInputHeight(batchIdx);\n    return computeReshapedDimensions({ width, height }, this.inputSize);\n  }\n\n  /**\n   * Create a batch tensor from all input canvases and tensors\n   * with size [batchSize, inputSize, inputSize, 3].\n   *\n   * @param inputSize Height and width of the tensor.\n   * @param isCenterImage (optional, default: false) If true, add an equal amount of padding on\n   * both sides of the minor dimension oof the image.\n   * @returns The batch tensor.\n   */\n  public toBatchTensor(inputSize: number, isCenterInputs = true): tf.Tensor4D {\n    this._inputSize = inputSize;\n\n    return tf.tidy(() => {\n      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {\n        const input = this.getInput(batchIdx);\n\n        if (input instanceof tf.Tensor) {\n          let imgTensor = isTensor4D(input) ? input : tf.expandDims(input);\n          imgTensor = padToSquare(imgTensor as tf.Tensor4D, isCenterInputs);\n\n          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {\n            imgTensor = tf['image'].resizeBilinear(imgTensor as tf.Tensor4D, [inputSize, inputSize], false, false);\n          }\n\n          return imgTensor.as3D(inputSize, inputSize, 3);\n        }\n\n        if (input instanceof env.getEnv().Canvas) {\n          return tf['browser'].fromPixels(imageToSquare(input, inputSize, isCenterInputs));\n        }\n\n        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);\n      });\n\n      const batchTensor = tf.stack(inputTensors.map((t) => tf.cast(t, 'float32'))).as4D(this.batchSize, inputSize, inputSize, 3);\n      // const batchTensor = tf.stack(inputTensors.map((t) => tf.cast(t, 'float32'))) as tf.Tensor4D;\n\n      return batchTensor;\n    });\n  }\n}\n", "import { isTensor3D, isTensor4D } from '../utils/index';\nimport { awaitMediaLoaded } from './awaitMediaLoaded';\nimport { isMediaElement } from './isMediaElement';\nimport { NetInput } from './NetInput';\nimport { resolveInput } from './resolveInput';\nimport { TNetInput } from './types';\n\n/**\n * Validates the input to make sure, they are valid net inputs and awaits all media elements\n * to be finished loading.\n *\n * @param input The input, which can be a media element or an array of different media elements.\n * @returns A NetInput instance, which can be passed into one of the neural networks.\n */\nexport async function toNetInput(inputs: TNetInput): Promise<NetInput> {\n  if (inputs instanceof NetInput) return inputs;\n  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];\n  if (!inputArgArray.length) throw new Error('toNetInput - empty array passed as input');\n  const getIdxHint = (idx: number) => (Array.isArray(inputs) ? ` at input index ${idx}:` : '');\n  const inputArray = inputArgArray.map(resolveInput);\n  inputArray.forEach((input, i) => {\n    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {\n      if (typeof inputArgArray[i] === 'string') throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);\n      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);\n    }\n    if (isTensor4D(input)) {\n      // if tf.Tensor4D is passed in the input array, the batch size has to be 1\n      const batchSize = input.shape[0];\n      if (batchSize !== 1) throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);\n    }\n  });\n  // wait for all media elements being loaded\n  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));\n  return new NetInput(inputArray, Array.isArray(inputs));\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\nimport { Rect } from '../classes/Rect';\nimport { env } from '../env/index';\nimport { createCanvas } from './createCanvas';\nimport { getContext2dOrThrow } from './getContext2dOrThrow';\nimport { imageTensorToCanvas } from './imageTensorToCanvas';\nimport { toNetInput } from './toNetInput';\nimport { TNetInput } from './types';\n\n/**\n * Extracts the image regions containing the detected faces.\n *\n * @param input The image that face detection has been performed on.\n * @param detections The face detection results or face bounding boxes for that image.\n * @returns The Canvases of the corresponding image region for each detected face.\n */\nexport async function extractFaces(input: TNetInput, detections: Array<FaceDetection | Rect>): Promise<HTMLCanvasElement[]> {\n  const { Canvas } = env.getEnv();\n  let canvas = input as HTMLCanvasElement;\n  if (!(input instanceof Canvas)) {\n    const netInput = await toNetInput(input);\n    if (netInput.batchSize > 1) throw new Error('extractFaces - batchSize > 1 not supported');\n    const tensorOrCanvas = netInput.getInput(0);\n    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);\n  }\n  const ctx = getContext2dOrThrow(canvas);\n  const boxes = detections\n    .map((det) => (det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det))\n    .map((box) => box.clipAtImageBorders(canvas.width, canvas.height));\n  return boxes.map(({ x, y, width, height }) => {\n    const faceImg = createCanvas({ width, height });\n    if (width > 0 && height > 0) getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);\n    return faceImg;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Rect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { isTensor3D, isTensor4D } from '../utils/index';\n\n/**\n * Extracts the tensors of the image regions containing the detected faces.\n * Useful if you want to compute the face descriptors for the face images.\n * Using this method is faster then extracting a canvas for each face and\n * converting them to tensors individually.\n *\n * @param imageTensor The image tensor that face detection has been performed on.\n * @param detections The face detection results or face bounding boxes for that image.\n * @returns Tensors of the corresponding image region for each detected face.\n */\nexport async function extractFaceTensors(imageTensor: tf.Tensor3D | tf.Tensor4D, detections: Array<FaceDetection | Rect>): Promise<tf.Tensor3D[]> {\n  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {\n    throw new Error('extractFaceTensors - expected image tensor to be 3D or 4D');\n  }\n\n  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {\n    throw new Error('extractFaceTensors - batchSize > 1 not supported');\n  }\n\n  return tf.tidy(() => {\n    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);\n    const boxes = detections.map((det) => (det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det))\n      .map((box) => box.clipAtImageBorders(imgWidth, imgHeight));\n    const faceTensors = boxes\n      .filter((box) => box.width > 0 && box.height > 0)\n      .map(({ x, y, width, height }) => tf.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));\n    return faceTensors;\n  });\n}\n", "import { env } from '../env/index';\n\nexport async function fetchOrThrow(\n  url: string,\n  // eslint-disable-next-line no-undef\n  init?: RequestInit,\n): Promise<Response> {\n  const { fetch } = env.getEnv();\n  const res = await fetch(url, init);\n  if (!(res.status < 400)) {\n    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);\n  }\n  return res;\n}\n", "import { bufferToImage } from './bufferToImage';\nimport { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchImage(uri: string): Promise<HTMLImageElement> {\n  const res = await fetchOrThrow(uri);\n  const blob = await (res).blob();\n\n  if (!blob.type.startsWith('image/')) {\n    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);\n  }\n  return bufferToImage(blob);\n}\n", "import { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchJson<T>(uri: string): Promise<T> {\n  return (await fetchOrThrow(uri)).json();\n}\n", "import { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchNetWeights(uri: string): Promise<Float32Array> {\n  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());\n}\n", "import { env } from '../env/index';\n\nexport function bufferToVideo(buf: Blob): Promise<HTMLVideoElement> {\n  return new Promise((resolve, reject) => {\n    if (!(buf instanceof Blob)) reject(new Error('bufferToVideo - expected buf to be of type: Blob'));\n\n    const video = env.getEnv().createVideoElement();\n    video.oncanplay = () => resolve(video);\n    video.onerror = reject;\n    video.playsInline = true;\n    video.muted = true;\n    video.src = URL.createObjectURL(buf);\n    video.play();\n  });\n}\n", "import { bufferToVideo } from './bufferToVideo';\nimport { fetchOrThrow } from './fetchOrThrow';\n\nexport async function fetchVideo(uri: string): Promise<HTMLVideoElement> {\n  const res = await fetchOrThrow(uri);\n  const blob = await (res).blob();\n\n  if (!blob.type.startsWith('video/')) {\n    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);\n  }\n  return bufferToVideo(blob);\n}\n", "export function getModelUris(uri: string | undefined, defaultModelName: string) {\n  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;\n\n  if (!uri) {\n    return {\n      modelBaseUri: '',\n      manifestUri: defaultManifestFilename,\n    };\n  }\n\n  if (uri === '/') {\n    return {\n      modelBaseUri: '/',\n      manifestUri: `/${defaultManifestFilename}`,\n    };\n  }\n  // eslint-disable-next-line no-nested-ternary\n  const protocol = uri.startsWith('http://') ? 'http://' : uri.startsWith('https://') ? 'https://' : '';\n  uri = uri.replace(protocol, '');\n\n  const parts = uri.split('/').filter((s) => s);\n\n  const manifestFile = uri.endsWith('.json')\n    ? parts[parts.length - 1]\n    : defaultManifestFilename;\n\n  let modelBaseUri = protocol + (uri.endsWith('.json') ? parts.slice(0, parts.length - 1) : parts).join('/');\n  modelBaseUri = uri.startsWith('/') ? `/${modelBaseUri}` : modelBaseUri;\n\n  return {\n    modelBaseUri,\n    manifestUri: modelBaseUri === '/' ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { getModelUris } from '../common/getModelUris';\nimport { fetchJson } from './fetchJson';\n\nexport async function loadWeightMap(\n  uri: string | undefined,\n  defaultModelName: string,\n): Promise<tf.NamedTensorMap> {\n  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);\n  // @ts-ignore\n  const manifest = await fetchJson<tf.io.WeightsManifestConfig>(manifestUri);\n  // if (manifest['weightsManifest']) manifest = manifest['weightsManifest'];\n  return tf['io'].loadWeights(manifest, modelBaseUri);\n}\n", "import { IDimensions } from '../classes/index';\nimport { getMediaDimensions } from './getMediaDimensions';\n\nexport function matchDimensions(input: IDimensions, reference: IDimensions, useMediaDimensions = false) {\n  const { width, height } = useMediaDimensions\n    ? getMediaDimensions(reference)\n    : reference;\n  input.width = width;\n  input.height = height;\n  return { width, height };\n}\n", "import * as tf from '../dist/tfjs.esm';\n\nimport { ParamMapping } from './common/index';\nimport { getModelUris } from './common/getModelUris';\nimport { loadWeightMap } from './dom/index';\nimport { env } from './env/index';\n\nexport abstract class NeuralNetwork<TNetParams> {\n  constructor(name: string) {\n    this._name = name;\n  }\n\n  protected _params: TNetParams | undefined = undefined;\n\n  protected _paramMappings: ParamMapping[] = [];\n\n  public _name: any;\n\n  public get params(): TNetParams | undefined { return this._params; }\n\n  public get paramMappings(): ParamMapping[] { return this._paramMappings; }\n\n  public get isLoaded(): boolean { return !!this.params; }\n\n  public getParamFromPath(paramPath: string): tf.Tensor {\n    const { obj, objProp } = this.traversePropertyPath(paramPath);\n    return obj[objProp];\n  }\n\n  public reassignParamFromPath(paramPath: string, tensor: tf.Tensor) {\n    const { obj, objProp } = this.traversePropertyPath(paramPath);\n    obj[objProp].dispose();\n    obj[objProp] = tensor;\n  }\n\n  public getParamList() {\n    return this._paramMappings.map(({ paramPath }) => ({\n      path: paramPath,\n      tensor: this.getParamFromPath(paramPath),\n    }));\n  }\n\n  public getTrainableParams() {\n    return this.getParamList().filter((param) => param.tensor instanceof tf.Variable);\n  }\n\n  public getFrozenParams() {\n    return this.getParamList().filter((param) => !(param.tensor instanceof tf.Variable));\n  }\n\n  public variable() {\n    this.getFrozenParams().forEach(({ path, tensor }) => {\n      this.reassignParamFromPath(path, tensor.variable());\n    });\n  }\n\n  public freeze() {\n    this.getTrainableParams().forEach(({ path, tensor: variable }) => {\n      const tensor = tf.tensor(variable.dataSync());\n      variable.dispose();\n      this.reassignParamFromPath(path, tensor);\n    });\n  }\n\n  public dispose(throwOnRedispose = true) {\n    this.getParamList().forEach((param) => {\n      if (throwOnRedispose && param.tensor.isDisposed) {\n        throw new Error(`param tensor has already been disposed for path ${param.path}`);\n      }\n      param.tensor.dispose();\n    });\n    this._params = undefined;\n  }\n\n  public serializeParams(): Float32Array {\n    return new Float32Array(\n      this.getParamList()\n        .map(({ tensor }) => Array.from(tensor.dataSync()) as number[])\n        .reduce((flat, arr) => flat.concat(arr)),\n    );\n  }\n\n  public async load(weightsOrUrl: Float32Array | string | undefined): Promise<void> {\n    if (weightsOrUrl instanceof Float32Array) {\n      this.extractWeights(weightsOrUrl);\n      return;\n    }\n    await this.loadFromUri(weightsOrUrl);\n  }\n\n  public async loadFromUri(uri: string | undefined) {\n    if (uri && typeof uri !== 'string') {\n      throw new Error(`${this._name}.loadFromUri - expected model uri`);\n    }\n    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());\n    this.loadFromWeightMap(weightMap);\n  }\n\n  public async loadFromDisk(filePath: string | undefined) {\n    if (filePath && typeof filePath !== 'string') {\n      throw new Error(`${this._name}.loadFromDisk - expected model file path`);\n    }\n    const { readFile } = env.getEnv();\n    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());\n    const fetchWeightsFromDisk = (filePaths: string[]) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));\n    const loadWeights = tf['io'].weightsLoaderFactory(fetchWeightsFromDisk);\n    const manifest = JSON.parse((await readFile(manifestUri)).toString());\n    const weightMap = await loadWeights(manifest, modelBaseUri);\n    this.loadFromWeightMap(weightMap);\n  }\n\n  public loadFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);\n    this._paramMappings = paramMappings;\n    this._params = params;\n  }\n\n  public extractWeights(weights: Float32Array) {\n    const { paramMappings, params } = this.extractParams(weights);\n    this._paramMappings = paramMappings;\n    this._params = params;\n  }\n\n  private traversePropertyPath(paramPath: string) {\n    if (!this.params) {\n      throw new Error('traversePropertyPath - model has no loaded params');\n    }\n\n    const result = paramPath.split('/').reduce((res: { nextObj: any, obj?: any, objProp?: string }, objProp) => {\n      // eslint-disable-next-line no-prototype-builtins\n      if (!res.nextObj.hasOwnProperty(objProp)) {\n        throw new Error(`traversePropertyPath - object does not have property ${objProp}, for path ${paramPath}`);\n      }\n      return { obj: res.nextObj, objProp, nextObj: res.nextObj[objProp] };\n    }, { nextObj: this.params });\n\n    const { obj, objProp } = result;\n    if (!obj || !objProp || !(obj[objProp] instanceof tf.Tensor)) {\n      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);\n    }\n\n    return { obj, objProp };\n  }\n\n  protected abstract getDefaultModelName(): string\n\n  // eslint-disable-next-line no-unused-vars\n  protected abstract extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TNetParams, paramMappings: ParamMapping[] }\n\n  // eslint-disable-next-line no-unused-vars\n  protected abstract extractParams(weights: Float32Array): { params: TNetParams, paramMappings: ParamMapping[] }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { SeparableConvParams } from './types';\n\nexport function depthwiseSeparableConv(\n  x: tf.Tensor4D,\n  params: SeparableConvParams,\n  stride: [number, number],\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, 'same');\n    out = tf.add(out, params.bias);\n    return out;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, SeparableConvParams } from '../common/index';\nimport { depthwiseSeparableConv } from '../common/depthwiseSeparableConv';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function denseBlock3(\n  x: tf.Tensor4D,\n  denseBlockParams: DenseBlock3Params,\n  isFirstLayer = false,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out1 = tf.relu(\n      isFirstLayer\n        ? tf.add(\n          tf.conv2d(x, (denseBlockParams.conv0 as ConvParams).filters, [2, 2], 'same'),\n          denseBlockParams.conv0.bias,\n        )\n        : depthwiseSeparableConv(x, denseBlockParams.conv0 as SeparableConvParams, [2, 2]),\n    ) as tf.Tensor4D;\n    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);\n\n    const in3 = tf.relu(tf.add(out1, out2)) as tf.Tensor4D;\n    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);\n\n    return tf.relu(tf.add(out1, tf.add(out2, out3))) as tf.Tensor4D;\n  });\n}\n\nexport function denseBlock4(\n  x: tf.Tensor4D,\n  denseBlockParams: DenseBlock4Params,\n  isFirstLayer = false,\n  isScaleDown = true,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out1 = tf.relu(\n      isFirstLayer\n        ? tf.add(\n          tf.conv2d(x, (denseBlockParams.conv0 as ConvParams).filters, isScaleDown ? [2, 2] : [1, 1], 'same'),\n          denseBlockParams.conv0.bias,\n        )\n        : depthwiseSeparableConv(x, denseBlockParams.conv0 as SeparableConvParams, isScaleDown ? [2, 2] : [1, 1]),\n    ) as tf.Tensor4D;\n    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);\n\n    const in3 = tf.relu(tf.add(out1, out2)) as tf.Tensor4D;\n    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);\n\n    const in4 = tf.relu(tf.add(out1, tf.add(out2, out3))) as tf.Tensor4D;\n    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);\n\n    return tf.relu(tf.add(out1, tf.add(out2, tf.add(out3, out4)))) as tf.Tensor4D;\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from './types';\n\nexport function convLayer(\n  x: tf.Tensor4D,\n  params: ConvParams,\n  padding: 'valid' | 'same' = 'same',\n  withRelu = false,\n): tf.Tensor4D {\n  return tf.tidy(() => {\n    const out = tf.add(\n      tf.conv2d(x, params.filters, [1, 1], padding),\n      params.bias,\n    ) as tf.Tensor4D;\n\n    return withRelu ? tf.relu(out) : out;\n  });\n}\n", "import { ParamMapping } from './types';\n\nexport function disposeUnusedWeightTensors(weightMap: any, paramMappings: ParamMapping[]) {\n  Object.keys(weightMap).forEach((path) => {\n    if (!paramMappings.some((pm) => pm.originalPath === path)) {\n      weightMap[path].dispose();\n    }\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, ExtractWeightsFunction, ParamMapping } from './types';\n\nexport function extractConvParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvParams => {\n    const filters = tf.tensor4d(\n      extractWeights(channelsIn * channelsOut * filterSize * filterSize),\n      [filterSize, filterSize, channelsIn, channelsOut],\n    );\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return { filters, bias };\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, FCParams, ParamMapping } from './types';\n\nexport function extractFCParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (\n    channelsIn: number,\n    channelsOut: number,\n    mappedPrefix: string,\n  ): FCParams => {\n    const fc_weights = tf.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);\n    const fc_bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/weights` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return {\n      weights: fc_weights,\n      bias: fc_bias,\n    };\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\n// eslint-disable-next-line no-unused-vars\nexport type ExtractWeightsFunction = (numWeights: number) => Float32Array\n\nexport type ParamMapping = {\n  originalPath?: string\n  paramPath: string\n}\n\nexport type ConvParams = {\n  filters: tf.Tensor4D\n  bias: tf.Tensor1D\n}\n\nexport type FCParams = {\n  weights: tf.Tensor2D\n  bias: tf.Tensor1D\n}\n\nexport class SeparableConvParams {\n  // eslint-disable-next-line no-useless-constructor\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    public depthwise_filter: tf.Tensor4D,\n    // eslint-disable-next-line no-unused-vars\n    public pointwise_filter: tf.Tensor4D,\n    // eslint-disable-next-line no-unused-vars\n    public bias: tf.Tensor1D,\n  // eslint-disable-next-line no-empty-function\n  ) {}\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, ParamMapping, SeparableConvParams } from './types';\n\nexport function extractSeparableConvParamsFactory(\n  extractWeights: ExtractWeightsFunction,\n  paramMappings: ParamMapping[],\n) {\n  return (channelsIn: number, channelsOut: number, mappedPrefix: string): SeparableConvParams => {\n    const depthwise_filter = tf.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);\n    const pointwise_filter = tf.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/depthwise_filter` },\n      { paramPath: `${mappedPrefix}/pointwise_filter` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return new SeparableConvParams(\n      depthwise_filter,\n      pointwise_filter,\n      bias,\n    );\n  };\n}\n\nexport function loadSeparableConvParamsFactory(\n  // eslint-disable-next-line no-unused-vars\n  extractWeightEntry: <T>(originalPath: string, paramRank: number) => T,\n) {\n  return (prefix: string): SeparableConvParams => {\n    const depthwise_filter = extractWeightEntry<tf.Tensor4D>(`${prefix}/depthwise_filter`, 4);\n    const pointwise_filter = extractWeightEntry<tf.Tensor4D>(`${prefix}/pointwise_filter`, 4);\n    const bias = extractWeightEntry<tf.Tensor1D>(`${prefix}/bias`, 1);\n\n    return new SeparableConvParams(\n      depthwise_filter,\n      pointwise_filter,\n      bias,\n    );\n  };\n}\n", "import { isTensor } from '../utils/index';\nimport { ParamMapping } from './types';\n\nexport function extractWeightEntryFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  return (originalPath: string, paramRank: number, mappedPath?: string) => {\n    const tensor = weightMap[originalPath];\n\n    if (!isTensor(tensor, paramRank)) {\n      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor}`);\n    }\n\n    paramMappings.push(\n      { originalPath, paramPath: mappedPath || originalPath },\n    );\n\n    return tensor;\n  };\n}\n", "export function extractWeightsFactory(weights: Float32Array) {\n  let remainingWeights = weights;\n\n  function extractWeights(numWeights: number): Float32Array {\n    const ret = remainingWeights.slice(0, numWeights);\n    remainingWeights = remainingWeights.slice(numWeights);\n    return ret;\n  }\n\n  function getRemainingWeights(): Float32Array {\n    return remainingWeights;\n  }\n\n  return {\n    extractWeights,\n    getRemainingWeights,\n  };\n}\n", "import { extractConvParamsFactory, extractSeparableConvParamsFactory, ExtractWeightsFunction, ParamMapping } from '../common/index';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  function extractDenseBlock3Params(channelsIn: number, channelsOut: number, mappedPrefix: string, isFirstLayer = false): DenseBlock3Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`)\n      : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);\n    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);\n    const conv2 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);\n\n    return { conv0, conv1, conv2 };\n  }\n\n  function extractDenseBlock4Params(channelsIn: number, channelsOut: number, mappedPrefix: string, isFirstLayer = false): DenseBlock4Params {\n    const { conv0, conv1, conv2 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);\n    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);\n\n    return {\n      conv0, conv1, conv2, conv3,\n    };\n  }\n\n  return {\n    extractDenseBlock3Params,\n    extractDenseBlock4Params,\n  };\n}\n", "import { extractWeightsFactory, ParamMapping } from '../common/index';\nimport { extractorsFactory } from './extractorsFactory';\nimport { FaceFeatureExtractorParams } from './types';\n\nexport function extractParams(weights: Float32Array): { params: FaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractDenseBlock4Params,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const dense0 = extractDenseBlock4Params(3, 32, 'dense0', true);\n  const dense1 = extractDenseBlock4Params(32, 64, 'dense1');\n  const dense2 = extractDenseBlock4Params(64, 128, 'dense2');\n  const dense3 = extractDenseBlock4Params(128, 256, 'dense3');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: {\n      dense0, dense1, dense2, dense3,\n    },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from './types';\n\n// eslint-disable-next-line no-unused-vars\nexport function loadConvParamsFactory(extractWeightEntry: <T>(originalPath: string, paramRank: number) => T) {\n  return (prefix: string): ConvParams => {\n    const filters = extractWeightEntry<tf.Tensor4D>(`${prefix}/filters`, 4);\n    const bias = extractWeightEntry<tf.Tensor1D>(`${prefix}/bias`, 1);\n\n    return { filters, bias };\n  };\n}\n", "import { extractWeightEntryFactory, loadSeparableConvParamsFactory, ParamMapping } from '../common/index';\nimport { loadConvParamsFactory } from '../common/loadConvParamsFactory';\nimport { DenseBlock3Params, DenseBlock4Params } from './types';\n\nexport function loadParamsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  const extractConvParams = loadConvParamsFactory(extractWeightEntry);\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n\n  function extractDenseBlock3Params(prefix: string, isFirstLayer = false): DenseBlock3Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(`${prefix}/conv0`)\n      : extractSeparableConvParams(`${prefix}/conv0`);\n    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);\n    const conv2 = extractSeparableConvParams(`${prefix}/conv2`);\n\n    return { conv0, conv1, conv2 };\n  }\n\n  function extractDenseBlock4Params(prefix: string, isFirstLayer = false): DenseBlock4Params {\n    const conv0 = isFirstLayer\n      ? extractConvParams(`${prefix}/conv0`)\n      : extractSeparableConvParams(`${prefix}/conv0`);\n    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);\n    const conv2 = extractSeparableConvParams(`${prefix}/conv2`);\n    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);\n\n    return {\n      conv0, conv1, conv2, conv3,\n    };\n  }\n\n  return {\n    extractDenseBlock3Params,\n    extractDenseBlock4Params,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, ParamMapping } from '../common/index';\nimport { loadParamsFactory } from './loadParamsFactory';\nimport { FaceFeatureExtractorParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: FaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractDenseBlock4Params,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const params = {\n    dense0: extractDenseBlock4Params('dense0', true),\n    dense1: extractDenseBlock4Params('dense1'),\n    dense2: extractDenseBlock4Params('dense2'),\n    dense3: extractDenseBlock4Params('dense3'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { denseBlock4 } from './denseBlock';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { FaceFeatureExtractorParams, IFaceFeatureExtractor } from './types';\n\nexport class FaceFeatureExtractor extends NeuralNetwork<FaceFeatureExtractorParams> implements IFaceFeatureExtractor<FaceFeatureExtractorParams> {\n  constructor() {\n    super('FaceFeatureExtractor');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('FaceFeatureExtractor - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = denseBlock4(normalized, params.dense0, true);\n      out = denseBlock4(out, params.dense1);\n      out = denseBlock4(out, params.dense2);\n      out = denseBlock4(out, params.dense3);\n      out = tf.avgPool(out, [7, 7], [2, 2], 'valid');\n\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_feature_extractor_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FCParams } from './types';\n\nexport function fullyConnectedLayer(\n  x: tf.Tensor2D,\n  params: FCParams,\n): tf.Tensor2D {\n  return tf.tidy(() => tf.add(\n    tf.matMul(x, params.weights),\n    params.bias,\n  ));\n}\n", "import { extractFCParamsFactory, extractWeightsFactory, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParams(weights: Float32Array, channelsIn: number, channelsOut: number): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);\n\n  const fc = extractFCParams(channelsIn, channelsOut, 'fc');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { fc },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, FCParams, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractFcParams(prefix: string): FCParams {\n    const weights = extractWeightEntry(`${prefix}/weights`, 2);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { weights, bias };\n  }\n\n  const params = {\n    fc: extractFcParams('fc'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function seperateWeightMaps(weightMap: tf.NamedTensorMap) {\n  const featureExtractorMap: tf.NamedTensorMap = {};\n  const classifierMap: tf.NamedTensorMap = {};\n\n  Object.keys(weightMap).forEach((key) => {\n    const map = key.startsWith('fc') ? classifierMap : featureExtractorMap;\n    map[key] = weightMap[key];\n  });\n\n  return { featureExtractorMap, classifierMap };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { fullyConnectedLayer } from '../common/fullyConnectedLayer';\nimport { NetInput } from '../dom/index';\nimport { FaceFeatureExtractorParams, IFaceFeatureExtractor, TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { NetParams } from './types';\nimport { seperateWeightMaps } from './util';\n\nexport abstract class FaceProcessor<\n  TExtractorParams extends FaceFeatureExtractorParams | TinyFaceFeatureExtractorParams\n>\n  extends NeuralNetwork<NetParams> {\n  protected _faceFeatureExtractor: IFaceFeatureExtractor<TExtractorParams>;\n\n  constructor(_name: string, faceFeatureExtractor: IFaceFeatureExtractor<TExtractorParams>) {\n    super(_name);\n    this._faceFeatureExtractor = faceFeatureExtractor;\n  }\n\n  public get faceFeatureExtractor(): IFaceFeatureExtractor<TExtractorParams> {\n    return this._faceFeatureExtractor;\n  }\n\n  protected abstract override getDefaultModelName(): string\n\n  protected abstract getClassifierChannelsIn(): number\n\n  protected abstract getClassifierChannelsOut(): number\n\n  public runNet(input: NetInput | tf.Tensor4D): tf.Tensor2D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error(`${this._name} - load model before inference`);\n    }\n\n    return tf.tidy(() => {\n      const bottleneckFeatures = input instanceof NetInput\n        ? this.faceFeatureExtractor.forwardInput(input)\n        : input;\n      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);\n    });\n  }\n\n  public override dispose(throwOnRedispose = true) {\n    this.faceFeatureExtractor.dispose(throwOnRedispose);\n    super.dispose(throwOnRedispose);\n  }\n\n  public loadClassifierParams(weights: Float32Array) {\n    const { params, paramMappings } = this.extractClassifierParams(weights);\n    this._params = params;\n    this._paramMappings = paramMappings;\n  }\n\n  public extractClassifierParams(weights: Float32Array) {\n    return extractParams(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);\n\n    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);\n\n    return extractParamsFromWeightMap(classifierMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const cIn = this.getClassifierChannelsIn();\n    const cOut = this.getClassifierChannelsOut();\n    const classifierWeightSize = (cOut * cIn) + cOut;\n\n    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);\n    const classifierWeights = weights.slice(weights.length - classifierWeightSize);\n\n    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);\n    return this.extractClassifierParams(classifierWeights);\n  }\n}\n", "export const FACE_EXPRESSION_LABELS = ['neutral', 'happy', 'sad', 'angry', 'fearful', 'disgusted', 'surprised'];\n\nexport class FaceExpressions {\n  public neutral = 0;\n  public happy = 0;\n  public sad = 0;\n  public angry = 0;\n  public fearful = 0;\n  public disgusted = 0;\n  public surprised = 0;\n\n  constructor(probabilities: number[] | Float32Array) {\n    if (probabilities.length !== 7) {\n      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);\n    }\n\n    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {\n      this[expression] = probabilities[idx];\n    });\n  }\n\n  asSortedArray() {\n    return FACE_EXPRESSION_LABELS\n      .map((expression) => ({ expression, probability: this[expression] as number }))\n      .sort((e0, e1) => e1.probability - e0.probability);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { FaceFeatureExtractor } from '../faceFeatureExtractor/FaceFeatureExtractor';\nimport { FaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceProcessor } from '../faceProcessor/FaceProcessor';\nimport { FaceExpressions } from './FaceExpressions';\n\nexport class FaceExpressionNet extends FaceProcessor<FaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: FaceFeatureExtractor = new FaceFeatureExtractor()) {\n    super('FaceExpressionNet', faceFeatureExtractor);\n  }\n\n  public forwardInput(input: NetInput | tf.Tensor4D): tf.Tensor2D {\n    return tf.tidy(() => tf.softmax(this.runNet(input)));\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async predictExpressions(input: TNetInput) {\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput);\n    const probabilitesByBatch = await Promise.all(tf.unstack(out).map(async (t) => {\n      const data = t.dataSync();\n      t.dispose();\n      return data;\n    }));\n    out.dispose();\n\n    const predictionsByBatch = probabilitesByBatch\n      .map((probabilites) => new FaceExpressions(probabilites as Float32Array));\n\n    return netInput.isBatchInput\n      ? predictionsByBatch\n      : predictionsByBatch[0];\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_expression_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 256;\n  }\n\n  protected getClassifierChannelsOut(): number {\n    return 7;\n  }\n}\n", "import { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\n\nexport type WithFaceExpressions<TSource> = TSource & { expressions: FaceExpressions }\n\nexport function isWithFaceExpressions(obj: any): obj is WithFaceExpressions<{}> {\n  return obj.expressions instanceof FaceExpressions;\n}\n\nexport function extendWithFaceExpressions<TSource>(sourceObj: TSource, expressions: FaceExpressions): WithFaceExpressions<TSource> {\n  const extension = { expressions };\n  return { ...sourceObj, ...extension };\n}\n", "import { IPoint, Point } from '../classes/index';\nimport { FaceExpressions } from '../faceExpressionNet/index';\nimport { isWithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceExpressions, WithFaceExpressions } from '../factories/WithFaceExpressions';\nimport { round } from '../utils/index';\nimport { DrawTextField } from './DrawTextField';\n\nexport type DrawFaceExpressionsInput = FaceExpressions | WithFaceExpressions<{}>\n\nexport function drawFaceExpressions(canvasArg: string | HTMLCanvasElement, faceExpressions: DrawFaceExpressionsInput | Array<DrawFaceExpressionsInput>, minConfidence = 0.1, textFieldAnchor?: IPoint) {\n  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];\n\n  faceExpressionsArray.forEach((e) => {\n    // eslint-disable-next-line no-nested-ternary\n    const expr = e instanceof FaceExpressions\n      ? e\n      : (isWithFaceExpressions(e) ? e.expressions : undefined);\n    if (!expr) {\n      throw new Error('drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof');\n    }\n\n    const sorted = expr.asSortedArray();\n    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);\n\n    const anchor = isWithFaceDetection(e)\n      ? e.detection.box.bottomLeft\n      : (textFieldAnchor || new Point(0, 0));\n\n    const drawTextField = new DrawTextField(\n      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),\n      anchor,\n    );\n    drawTextField.draw(canvasArg);\n  });\n}\n", "import { FaceDetection } from '../classes/FaceDetection';\nimport { FaceLandmarks } from '../classes/FaceLandmarks';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { isWithFaceDetection, WithFaceDetection } from './WithFaceDetection';\n\nexport type WithFaceLandmarks<\n  TSource extends WithFaceDetection<{}>,\n  TFaceLandmarks extends FaceLandmarks = FaceLandmarks68\n> = TSource & {\n  landmarks: TFaceLandmarks;\n  unshiftedLandmarks: TFaceLandmarks;\n  alignedRect: FaceDetection;\n  angle: {\n    roll: number | undefined;\n    pitch: number | undefined;\n    yaw: number | undefined;\n  };\n};\n\nexport function isWithFaceLandmarks(\n  obj: any,\n): obj is WithFaceLandmarks<WithFaceDetection<{}>, FaceLandmarks> {\n  return (\n    isWithFaceDetection(obj)\n    // eslint-disable-next-line dot-notation\n    && obj['landmarks'] instanceof FaceLandmarks\n    // eslint-disable-next-line dot-notation\n    && obj['unshiftedLandmarks'] instanceof FaceLandmarks\n    // eslint-disable-next-line dot-notation\n    && obj['alignedRect'] instanceof FaceDetection\n  );\n}\n\nfunction calculateFaceAngle(mesh) {\n  // Helper to convert radians to degrees\n  // eslint-disable-next-line no-unused-vars, @typescript-eslint/no-unused-vars\n  const degrees = (radians) => (radians * 180) / Math.PI;\n  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);\n\n  const angle = {\n    roll: <number | undefined>undefined,\n    pitch: <number | undefined>undefined,\n    yaw: <number | undefined>undefined,\n  };\n\n  const calcYaw = (leftPoint, midPoint, rightPoint) => {\n    // Calc x-distance from left side of the face (\"ear\") to facial midpoint (\"nose\")\n    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);\n    // Calc x-distance from facial midpoint (\"nose\") to the right side of the face (\"ear\")\n    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);\n    // Difference in distances coincidentally approximates to angles\n    return leftToMidpoint - rightToMidpoint;\n  };\n\n  const calcRoll = (lever, pivot) => {\n    // When rolling, the head seems to pivot from the nose/lips/chin area.\n    // So, we'll choose any two points from the facial midline, where the first point should be the pivot, and the other \"lever\"\n    // Plan/Execution: get the hypotenuse & opposite sides of a 90deg triangle ==> Calculate angle in radians\n    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);\n    const opposite = pivot._y - lever._y;\n    const angleInRadians = Math.asin(opposite / hypotenuse);\n    const angleInDegrees = degrees(angleInRadians);\n    const normalizeAngle = Math.floor(90 - angleInDegrees);\n    // If lever more to the left of the pivot, then we're tilting left\n    // \"-\" is negative direction. \"+\", or absence of a sign is positive direction\n    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;\n    const result = normalizeAngle * tiltDirection;\n    return result;\n  };\n\n  const calcPitch = (leftPoint, midPoint, rightPoint) => {\n    // Theory: While pitching, the nose is the most salient point --> That's what we'll use to make a trianle.\n    // The \"base\" is between point that don't move when we pitch our head (i.e. an imaginary line running ear to ear through the nose).\n    // Executuin: Get the opposite & adjacent lengths of the triangle from the ear's perspective. Use it to get angle.\n\n    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);\n    // adjecent is base/2 technically.\n    const baseCoords = {\n      _x: (leftPoint._x + rightPoint._x) / 2,\n      _y: (leftPoint._y + rightPoint._y) / 2,\n    };\n    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);\n    const angleInRadians = Math.atan(midToBaseLength / base);\n    const angleInDegrees = Math.floor(degrees(angleInRadians));\n    // Account for directionality.\n    // pitch forwards (_i.e. tilting your head forwards) is positive (or no sign); backward is negative.\n    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;\n    const result = angleInDegrees * direction;\n    return result;\n  };\n\n  if (!mesh || !mesh._positions || mesh._positions.length !== 68) return angle;\n  const pt = mesh._positions;\n  angle.roll = calcRoll(pt[27], pt[66]);\n  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);\n  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);\n  return angle;\n}\n\nexport function extendWithFaceLandmarks<TSource extends WithFaceDetection<{}>, TFaceLandmarks extends FaceLandmarks = FaceLandmarks68>(\n  sourceObj: TSource,\n  unshiftedLandmarks: TFaceLandmarks,\n): WithFaceLandmarks<TSource, TFaceLandmarks> {\n  const { box: shift } = sourceObj.detection;\n  const landmarks = unshiftedLandmarks.shiftBy<TFaceLandmarks>(shift.x, shift.y);\n  const rect = landmarks.align();\n  const { imageDims } = sourceObj.detection;\n  const alignedRect = new FaceDetection(\n    sourceObj.detection.score,\n    rect.rescale(imageDims.reverse()),\n    imageDims,\n  );\n  const angle = calculateFaceAngle(unshiftedLandmarks);\n  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };\n  return { ...sourceObj, ...extension };\n}\n", "/* eslint-disable max-classes-per-file */\nimport { IPoint } from '../classes/index';\nimport { FaceLandmarks } from '../classes/FaceLandmarks';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { getContext2dOrThrow } from '../dom/getContext2dOrThrow';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { drawContour } from './drawContour';\n\nexport interface IDrawFaceLandmarksOptions {\n  drawLines?: boolean\n  drawPoints?: boolean\n  lineWidth?: number\n  pointSize?: number\n  lineColor?: string\n  pointColor?: string\n}\n\nexport class DrawFaceLandmarksOptions {\n  public drawLines: boolean;\n\n  public drawPoints: boolean;\n\n  public lineWidth: number;\n\n  public pointSize: number;\n\n  public lineColor: string;\n\n  public pointColor: string;\n\n  constructor(options: IDrawFaceLandmarksOptions = {}) {\n    const {\n      drawLines = true, drawPoints = true, lineWidth, lineColor, pointSize, pointColor,\n    } = options;\n    this.drawLines = drawLines;\n    this.drawPoints = drawPoints;\n    this.lineWidth = lineWidth || 1;\n    this.pointSize = pointSize || 2;\n    this.lineColor = lineColor || 'rgba(0, 255, 255, 1)';\n    this.pointColor = pointColor || 'rgba(255, 0, 255, 1)';\n  }\n}\n\nexport class DrawFaceLandmarks {\n  public faceLandmarks: FaceLandmarks;\n\n  public options: DrawFaceLandmarksOptions;\n\n  constructor(\n    faceLandmarks: FaceLandmarks,\n    options: IDrawFaceLandmarksOptions = {},\n  ) {\n    this.faceLandmarks = faceLandmarks;\n    this.options = new DrawFaceLandmarksOptions(options);\n  }\n\n  draw(canvasArg: string | HTMLCanvasElement | CanvasRenderingContext2D) {\n    const ctx = getContext2dOrThrow(canvasArg);\n\n    const {\n      drawLines, drawPoints, lineWidth, lineColor, pointSize, pointColor,\n    } = this.options;\n\n    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {\n      ctx.strokeStyle = lineColor;\n      ctx.lineWidth = lineWidth;\n      drawContour(ctx, this.faceLandmarks.getJawOutline());\n      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());\n      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());\n      drawContour(ctx, this.faceLandmarks.getNose());\n      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);\n      drawContour(ctx, this.faceLandmarks.getRightEye(), true);\n      drawContour(ctx, this.faceLandmarks.getMouth(), true);\n    }\n\n    if (drawPoints) {\n      ctx.strokeStyle = pointColor;\n      ctx.fillStyle = pointColor;\n\n      const drawPoint = (pt: IPoint) => {\n        ctx.beginPath();\n        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);\n        ctx.fill();\n      };\n      this.faceLandmarks.positions.forEach(drawPoint);\n    }\n  }\n}\n\nexport type DrawFaceLandmarksInput = FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>>\n\nexport function drawFaceLandmarks(\n  canvasArg: string | HTMLCanvasElement,\n  faceLandmarks: DrawFaceLandmarksInput | Array<DrawFaceLandmarksInput>,\n) {\n  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];\n  faceLandmarksArray.forEach((f) => {\n    // eslint-disable-next-line no-nested-ternary\n    const landmarks = f instanceof FaceLandmarks\n      ? f\n      : (isWithFaceLandmarks(f) ? f.landmarks : undefined);\n    if (!landmarks) {\n      throw new Error('drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof');\n    }\n\n    new DrawFaceLandmarks(landmarks).draw(canvasArg);\n  });\n}\n", "import { extractConvParamsFactory, extractSeparableConvParamsFactory, extractWeightsFactory } from '../common/index';\nimport { ExtractWeightsFunction, ParamMapping } from '../common/types';\nimport { range } from '../utils/index';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  function extractReductionBlockParams(channelsIn: number, channelsOut: number, mappedPrefix: string): ReductionBlockParams {\n    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);\n    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);\n\n    return { separable_conv0, separable_conv1, expansion_conv };\n  }\n\n  function extractMainBlockParams(channels: number, mappedPrefix: string): MainBlockParams {\n    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);\n    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);\n\n    return { separable_conv0, separable_conv1, separable_conv2 };\n  }\n\n  return {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array, numMainBlocks: number): { params: TinyXceptionParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const entry_flow_conv_in = extractConvParams(3, 32, 3, 'entry_flow/conv_in');\n  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, 'entry_flow/reduction_block_0');\n  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, 'entry_flow/reduction_block_1');\n\n  const entry_flow = {\n    conv_in: entry_flow_conv_in,\n    reduction_block_0: entry_flow_reduction_block_0,\n    reduction_block_1: entry_flow_reduction_block_1,\n  };\n\n  const middle_flow = {};\n  range(numMainBlocks, 0, 1).forEach((idx) => {\n    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);\n  });\n\n  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, 'exit_flow/reduction_block');\n  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, 'exit_flow/separable_conv');\n\n  const exit_flow = {\n    reduction_block: exit_flow_reduction_block,\n    separable_conv: exit_flow_separable_conv,\n  };\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { entry_flow, middle_flow, exit_flow },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, loadSeparableConvParamsFactory, ParamMapping } from '../common/index';\nimport { loadConvParamsFactory } from '../common/loadConvParamsFactory';\nimport { range } from '../utils/index';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction loadParamsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  const extractConvParams = loadConvParamsFactory(extractWeightEntry);\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n\n  function extractReductionBlockParams(mappedPrefix: string): ReductionBlockParams {\n    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);\n    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);\n\n    return { separable_conv0, separable_conv1, expansion_conv };\n  }\n\n  function extractMainBlockParams(mappedPrefix: string): MainBlockParams {\n    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);\n    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);\n    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);\n\n    return { separable_conv0, separable_conv1, separable_conv2 };\n  }\n\n  return {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n  numMainBlocks: number,\n): { params: TinyXceptionParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvParams,\n    extractSeparableConvParams,\n    extractReductionBlockParams,\n    extractMainBlockParams,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const entry_flow_conv_in = extractConvParams('entry_flow/conv_in');\n  const entry_flow_reduction_block_0 = extractReductionBlockParams('entry_flow/reduction_block_0');\n  const entry_flow_reduction_block_1 = extractReductionBlockParams('entry_flow/reduction_block_1');\n\n  const entry_flow = {\n    conv_in: entry_flow_conv_in,\n    reduction_block_0: entry_flow_reduction_block_0,\n    reduction_block_1: entry_flow_reduction_block_1,\n  };\n\n  const middle_flow = {};\n  range(numMainBlocks, 0, 1).forEach((idx) => {\n    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);\n  });\n\n  const exit_flow_reduction_block = extractReductionBlockParams('exit_flow/reduction_block');\n  const exit_flow_separable_conv = extractSeparableConvParams('exit_flow/separable_conv');\n\n  const exit_flow = {\n    reduction_block: exit_flow_reduction_block,\n    separable_conv: exit_flow_separable_conv,\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, depthwiseSeparableConv } from '../common/index';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { range } from '../utils/index';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { MainBlockParams, ReductionBlockParams, TinyXceptionParams } from './types';\n\nfunction conv(x: tf.Tensor4D, params: ConvParams, stride: [number, number]): tf.Tensor4D {\n  return tf.add(tf.conv2d(x, params.filters, stride, 'same'), params.bias);\n}\n\nfunction reductionBlock(x: tf.Tensor4D, params: ReductionBlockParams, isActivateInput = true): tf.Tensor4D {\n  let out = isActivateInput ? tf.relu(x) : x;\n  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv1, [1, 1]);\n  out = tf.maxPool(out, [3, 3], [2, 2], 'same');\n  out = tf.add(out, conv(x, params.expansion_conv, [2, 2]));\n  return out;\n}\n\nfunction mainBlock(x: tf.Tensor4D, params: MainBlockParams): tf.Tensor4D {\n  let out = depthwiseSeparableConv(tf.relu(x), params.separable_conv0, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv1, [1, 1]);\n  out = depthwiseSeparableConv(tf.relu(out), params.separable_conv2, [1, 1]);\n  out = tf.add(out, x);\n  return out;\n}\n\nexport class TinyXception extends NeuralNetwork<TinyXceptionParams> {\n  private _numMainBlocks: number;\n\n  constructor(numMainBlocks: number) {\n    super('TinyXception');\n    this._numMainBlocks = numMainBlocks;\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n    if (!params) {\n      throw new Error('TinyXception - load model before inference');\n    }\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n      let out = tf.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));\n      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);\n      out = reductionBlock(out, params.entry_flow.reduction_block_1);\n      range(this._numMainBlocks, 0, 1).forEach((idx) => {\n        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);\n      });\n      out = reductionBlock(out, params.exit_flow.reduction_block);\n      out = tf.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'tiny_xception_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap, this._numMainBlocks);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights, this._numMainBlocks);\n  }\n}\n", "import { extractFCParamsFactory, extractWeightsFactory, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);\n\n  const age = extractFCParams(512, 1, 'fc/age');\n  const gender = extractFCParams(512, 2, 'fc/gender');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { fc: { age, gender } },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, FCParams, ParamMapping } from '../common/index';\nimport { NetParams } from './types';\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractFcParams(prefix: string): FCParams {\n    const weights = extractWeightEntry(`${prefix}/weights`, 2);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { weights, bias };\n  }\n\n  const params = {\n    fc: {\n      age: extractFcParams('fc/age'),\n      gender: extractFcParams('fc/gender'),\n    },\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FCParams } from '../common/index';\n\n// eslint-disable-next-line no-shadow\nexport enum Gender {\n  // eslint-disable-next-line no-unused-vars\n  FEMALE = 'female',\n  // eslint-disable-next-line no-unused-vars\n  MALE = 'male'\n}\n\nexport type AgeAndGenderPrediction = {\n  age: number\n  gender: Gender\n  genderProbability: number\n}\n\nexport type NetOutput = { age: tf.Tensor1D, gender: tf.Tensor2D }\n\nexport type NetParams = {\n  fc: {\n    age: FCParams\n    gender: FCParams\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm.js';\nimport { fullyConnectedLayer } from '../common/fullyConnectedLayer';\nimport { seperateWeightMaps } from '../faceProcessor/util';\nimport { TinyXception } from '../xception/TinyXception';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { AgeAndGenderPrediction, Gender, NetOutput, NetParams } from './types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\n\nexport class AgeGenderNet extends NeuralNetwork<NetParams> {\n  private _faceFeatureExtractor: TinyXception;\n\n  constructor(faceFeatureExtractor: TinyXception = new TinyXception(2)) {\n    super('AgeGenderNet');\n    this._faceFeatureExtractor = faceFeatureExtractor;\n  }\n\n  public get faceFeatureExtractor(): TinyXception {\n    return this._faceFeatureExtractor;\n  }\n\n  public runNet(input: NetInput | tf.Tensor4D): NetOutput {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error(`${this._name} - load model before inference`);\n    }\n\n    return tf.tidy(() => {\n      const bottleneckFeatures = input instanceof NetInput\n        ? this.faceFeatureExtractor.forwardInput(input)\n        : input;\n\n      const pooled = tf.avgPool(bottleneckFeatures, [7, 7], [2, 2], 'valid').as2D(bottleneckFeatures.shape[0], -1);\n      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();\n      const gender = fullyConnectedLayer(pooled, params.fc.gender);\n      return { age, gender };\n    });\n  }\n\n  public forwardInput(input: NetInput | tf.Tensor4D): NetOutput {\n    return tf.tidy(() => {\n      const { age, gender } = this.runNet(input);\n      return { age, gender: tf.softmax(gender) };\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<NetOutput> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async predictAgeAndGender(input: TNetInput): Promise<AgeAndGenderPrediction | AgeAndGenderPrediction[]> {\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput);\n\n    const ages = tf.unstack(out.age);\n    const genders = tf.unstack(out.gender);\n    const ageAndGenderTensors = ages.map((ageTensor, i) => ({\n      ageTensor,\n      genderTensor: genders[i],\n    }));\n\n    const predictionsByBatch = await Promise.all(\n      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {\n        const age = (ageTensor.dataSync())[0];\n        const probMale = (genderTensor.dataSync())[0];\n        const isMale = probMale > 0.5;\n        const gender = isMale ? Gender.MALE : Gender.FEMALE;\n        const genderProbability = isMale ? probMale : (1 - probMale);\n\n        ageTensor.dispose();\n        genderTensor.dispose();\n        return { age, gender, genderProbability };\n      }),\n    );\n    out.age.dispose();\n    out.gender.dispose();\n\n    return netInput.isBatchInput ? predictionsByBatch as AgeAndGenderPrediction[] : predictionsByBatch[0] as AgeAndGenderPrediction;\n  }\n\n  protected getDefaultModelName(): string {\n    return 'age_gender_model';\n  }\n\n  public override dispose(throwOnRedispose = true) {\n    this.faceFeatureExtractor.dispose(throwOnRedispose);\n    super.dispose(throwOnRedispose);\n  }\n\n  public loadClassifierParams(weights: Float32Array) {\n    const { params, paramMappings } = this.extractClassifierParams(weights);\n    this._params = params;\n    this._paramMappings = paramMappings;\n  }\n\n  public extractClassifierParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);\n\n    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);\n\n    return extractParamsFromWeightMap(classifierMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const classifierWeightSize = (512 * 1 + 1) + (512 * 2 + 2);\n\n    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);\n    const classifierWeights = weights.slice(weights.length - classifierWeightSize);\n\n    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);\n    return this.extractClassifierParams(classifierWeights);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { IDimensions, Point } from '../classes/index';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { FaceFeatureExtractorParams, TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceProcessor } from '../faceProcessor/FaceProcessor';\nimport { isEven } from '../utils/index';\n\nexport abstract class FaceLandmark68NetBase<\n  TExtractorParams extends FaceFeatureExtractorParams | TinyFaceFeatureExtractorParams\n>\n  extends FaceProcessor<TExtractorParams> {\n  public postProcess(output: tf.Tensor2D, inputSize: number, originalDimensions: IDimensions[]): tf.Tensor2D {\n    const inputDimensions = originalDimensions.map(({ width, height }) => {\n      const scale = inputSize / Math.max(height, width);\n      return {\n        width: width * scale,\n        height: height * scale,\n      };\n    });\n\n    const batchSize = inputDimensions.length;\n\n    return tf.tidy(() => {\n      const createInterleavedTensor = (fillX: number, fillY: number) => tf.stack([tf.fill([68], fillX, 'float32'), tf.fill([68], fillY, 'float32')], 1).as2D(1, 136).as1D();\n\n      // eslint-disable-next-line no-unused-vars\n      const getPadding = (batchIdx: number, cond: (w: number, h: number) => boolean): number => {\n        const { width, height } = inputDimensions[batchIdx];\n        return cond(width, height) ? Math.abs(width - height) / 2 : 0;\n      };\n\n      const getPaddingX = (batchIdx: number) => getPadding(batchIdx, (w, h) => w < h);\n      const getPaddingY = (batchIdx: number) => getPadding(batchIdx, (w, h) => h < w);\n\n      const landmarkTensors = output\n        .mul(tf.fill([batchSize, 136], inputSize, 'float32'))\n        .sub(tf.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(\n          getPaddingX(batchIdx),\n          getPaddingY(batchIdx),\n        ))))\n        .div(tf.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(\n          inputDimensions[batchIdx].width,\n          inputDimensions[batchIdx].height,\n        ))));\n\n      return landmarkTensors as tf.Tensor2D;\n    });\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor2D {\n    return tf.tidy(() => {\n      const out = this.runNet(input);\n      return this.postProcess(\n        out,\n        input.inputSize as number,\n        input.inputDimensions.map(([height, width]) => ({ height, width })),\n      );\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async detectLandmarks(input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> {\n    const netInput = await toNetInput(input);\n    const landmarkTensors = tf.tidy(\n      () => tf.unstack(this.forwardInput(netInput)),\n    );\n\n    const landmarksForBatch = await Promise.all(landmarkTensors.map(\n      async (landmarkTensor, batchIdx) => {\n        const landmarksArray = Array.from(landmarkTensor.dataSync());\n        const xCoords = landmarksArray.filter((_, i) => isEven(i));\n        const yCoords = landmarksArray.filter((_, i) => !isEven(i));\n\n        return new FaceLandmarks68(\n          Array(68).fill(0).map((_, i) => new Point(xCoords[i] as number, yCoords[i] as number)),\n          {\n            height: netInput.getInputHeight(batchIdx),\n            width: netInput.getInputWidth(batchIdx),\n          },\n        );\n      },\n    ));\n\n    landmarkTensors.forEach((t) => t.dispose());\n\n    return netInput.isBatchInput ? landmarksForBatch as FaceLandmarks68[] : landmarksForBatch[0] as FaceLandmarks68;\n  }\n\n  protected getClassifierChannelsOut(): number {\n    return 136;\n  }\n}\n", "import { FaceFeatureExtractor } from '../faceFeatureExtractor/FaceFeatureExtractor';\nimport { FaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceLandmark68NetBase } from './FaceLandmark68NetBase';\n\nexport class FaceLandmark68Net extends FaceLandmark68NetBase<FaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: FaceFeatureExtractor = new FaceFeatureExtractor()) {\n    super('FaceLandmark68Net', faceFeatureExtractor);\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_landmark_68_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 256;\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, ParamMapping } from '../common/index';\nimport { loadParamsFactory } from './loadParamsFactory';\nimport { TinyFaceFeatureExtractorParams } from './types';\n\nexport function extractParamsFromWeightMapTiny(\n  weightMap: tf.NamedTensorMap,\n): { params: TinyFaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractDenseBlock3Params,\n  } = loadParamsFactory(weightMap, paramMappings);\n\n  const params = {\n    dense0: extractDenseBlock3Params('dense0', true),\n    dense1: extractDenseBlock3Params('dense1'),\n    dense2: extractDenseBlock3Params('dense2'),\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import { extractWeightsFactory, ParamMapping } from '../common/index';\nimport { extractorsFactory } from './extractorsFactory';\nimport { TinyFaceFeatureExtractorParams } from './types';\n\nexport function extractParamsTiny(weights: Float32Array): { params: TinyFaceFeatureExtractorParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const {\n    extractDenseBlock3Params,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const dense0 = extractDenseBlock3Params(3, 32, 'dense0', true);\n  const dense1 = extractDenseBlock3Params(32, 64, 'dense1');\n  const dense2 = extractDenseBlock3Params(64, 128, 'dense2');\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    paramMappings,\n    params: { dense0, dense1, dense2 },\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { denseBlock3 } from './denseBlock';\nimport { extractParamsFromWeightMapTiny } from './extractParamsFromWeightMapTiny';\nimport { extractParamsTiny } from './extractParamsTiny';\nimport { IFaceFeatureExtractor, TinyFaceFeatureExtractorParams } from './types';\n\nexport class TinyFaceFeatureExtractor extends NeuralNetwork<TinyFaceFeatureExtractorParams> implements IFaceFeatureExtractor<TinyFaceFeatureExtractorParams> {\n  constructor() {\n    super('TinyFaceFeatureExtractor');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('TinyFaceFeatureExtractor - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(112, true), 'float32');\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = denseBlock3(normalized, params.dense0, true);\n      out = denseBlock3(out, params.dense1);\n      out = denseBlock3(out, params.dense2);\n      out = tf.avgPool(out, [14, 14], [2, 2], 'valid');\n\n      return out;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_feature_extractor_tiny_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMapTiny(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParamsTiny(weights);\n  }\n}\n", "import { TinyFaceFeatureExtractor } from '../faceFeatureExtractor/TinyFaceFeatureExtractor';\nimport { TinyFaceFeatureExtractorParams } from '../faceFeatureExtractor/types';\nimport { FaceLandmark68NetBase } from './FaceLandmark68NetBase';\n\nexport class FaceLandmark68TinyNet extends FaceLandmark68NetBase<TinyFaceFeatureExtractorParams> {\n  constructor(faceFeatureExtractor: TinyFaceFeatureExtractor = new TinyFaceFeatureExtractor()) {\n    super('FaceLandmark68TinyNet', faceFeatureExtractor);\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_landmark_68_tiny_model';\n  }\n\n  protected getClassifierChannelsIn(): number {\n    return 128;\n  }\n}\n", "import { FaceLandmark68Net } from './FaceLandmark68Net';\n\nexport * from './FaceLandmark68Net';\nexport * from './FaceLandmark68TinyNet';\nexport class FaceLandmarkNet extends FaceLandmark68Net {}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ScaleLayerParams } from './types';\n\nexport function scale(x: tf.Tensor4D, params: ScaleLayerParams): tf.Tensor4D {\n  return tf.add(tf.mul(x, params.weights), params.biases);\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { scale } from './scaleLayer';\nimport { ConvLayerParams } from './types';\n\nfunction convLayer(\n  x: tf.Tensor4D,\n  params: ConvLayerParams,\n  strides: [number, number],\n  withRelu: boolean,\n  padding: 'valid' | 'same' = 'same',\n): tf.Tensor4D {\n  const { filters, bias } = params.conv;\n\n  let out = tf.conv2d(x, filters, strides, padding);\n  out = tf.add(out, bias);\n  out = scale(out, params.scale);\n  return withRelu ? tf.relu(out) : out;\n}\n\nexport function conv(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [1, 1], true);\n}\n\nexport function convNoRelu(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [1, 1], false);\n}\n\nexport function convDown(x: tf.Tensor4D, params: ConvLayerParams) {\n  return convLayer(x, params, [2, 2], true, 'valid');\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, extractWeightsFactory, ExtractWeightsFunction, ParamMapping } from '../common/index';\nimport { isFloat } from '../utils/index';\nimport { ConvLayerParams, NetParams, ResidualLayerParams, ScaleLayerParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  function extractFilterValues(numFilterValues: number, numFilters: number, filterSize: number): tf.Tensor4D {\n    const weights = extractWeights(numFilterValues);\n    const depth = weights.length / (numFilters * filterSize * filterSize);\n\n    if (isFloat(depth)) {\n      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);\n    }\n\n    return tf.tidy(\n      () => tf.transpose(\n        tf.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),\n        [2, 3, 1, 0],\n      ),\n    );\n  }\n\n  function extractConvParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvParams {\n    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);\n    const bias = tf.tensor1d(extractWeights(numFilters));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/bias` },\n    );\n\n    return { filters, bias };\n  }\n\n  function extractScaleLayerParams(numWeights: number, mappedPrefix: string): ScaleLayerParams {\n    const weights = tf.tensor1d(extractWeights(numWeights));\n    const biases = tf.tensor1d(extractWeights(numWeights));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/weights` },\n      { paramPath: `${mappedPrefix}/biases` },\n    );\n\n    return {\n      weights,\n      biases,\n    };\n  }\n\n  function extractConvLayerParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): ConvLayerParams {\n    const conv = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);\n    const scale = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);\n\n    return { conv, scale };\n  }\n\n  function extractResidualLayerParams(\n    numFilterValues: number,\n    numFilters: number,\n    filterSize: number,\n    mappedPrefix: string,\n    isDown = false,\n  ): ResidualLayerParams {\n    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);\n    const conv2 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);\n\n    return { conv1, conv2 };\n  }\n\n  return {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n\n  const conv32_down = extractConvLayerParams(4704, 32, 7, 'conv32_down');\n  const conv32_1 = extractResidualLayerParams(9216, 32, 3, 'conv32_1');\n  const conv32_2 = extractResidualLayerParams(9216, 32, 3, 'conv32_2');\n  const conv32_3 = extractResidualLayerParams(9216, 32, 3, 'conv32_3');\n\n  const conv64_down = extractResidualLayerParams(36864, 64, 3, 'conv64_down', true);\n  const conv64_1 = extractResidualLayerParams(36864, 64, 3, 'conv64_1');\n  const conv64_2 = extractResidualLayerParams(36864, 64, 3, 'conv64_2');\n  const conv64_3 = extractResidualLayerParams(36864, 64, 3, 'conv64_3');\n\n  const conv128_down = extractResidualLayerParams(147456, 128, 3, 'conv128_down', true);\n  const conv128_1 = extractResidualLayerParams(147456, 128, 3, 'conv128_1');\n  const conv128_2 = extractResidualLayerParams(147456, 128, 3, 'conv128_2');\n\n  const conv256_down = extractResidualLayerParams(589824, 256, 3, 'conv256_down', true);\n  const conv256_1 = extractResidualLayerParams(589824, 256, 3, 'conv256_1');\n  const conv256_2 = extractResidualLayerParams(589824, 256, 3, 'conv256_2');\n  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, 'conv256_down_out');\n\n  const fc = tf.tidy(\n    () => tf.transpose(tf.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0]),\n  );\n  paramMappings.push({ paramPath: 'fc' });\n\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  const params = {\n    conv32_down,\n    conv32_1,\n    conv32_2,\n    conv32_3,\n    conv64_down,\n    conv64_1,\n    conv64_2,\n    conv64_3,\n    conv128_down,\n    conv128_1,\n    conv128_2,\n    conv256_down,\n    conv256_1,\n    conv256_2,\n    conv256_down_out,\n    fc,\n  };\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { disposeUnusedWeightTensors, extractWeightEntryFactory, ParamMapping } from '../common/index';\nimport { isTensor2D } from '../utils/index';\nimport { ConvLayerParams, NetParams, ResidualLayerParams, ScaleLayerParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractScaleLayerParams(prefix: string): ScaleLayerParams {\n    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);\n    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);\n\n    return { weights, biases };\n  }\n\n  function extractConvLayerParams(prefix: string): ConvLayerParams {\n    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);\n    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);\n    const scale = extractScaleLayerParams(prefix);\n\n    return { conv: { filters, bias }, scale };\n  }\n\n  function extractResidualLayerParams(prefix: string): ResidualLayerParams {\n    return {\n      conv1: extractConvLayerParams(`${prefix}/conv1`),\n      conv2: extractConvLayerParams(`${prefix}/conv2`),\n    };\n  }\n\n  return {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvLayerParams,\n    extractResidualLayerParams,\n  } = extractorsFactory(weightMap, paramMappings);\n\n  const conv32_down = extractConvLayerParams('conv32_down');\n  const conv32_1 = extractResidualLayerParams('conv32_1');\n  const conv32_2 = extractResidualLayerParams('conv32_2');\n  const conv32_3 = extractResidualLayerParams('conv32_3');\n\n  const conv64_down = extractResidualLayerParams('conv64_down');\n  const conv64_1 = extractResidualLayerParams('conv64_1');\n  const conv64_2 = extractResidualLayerParams('conv64_2');\n  const conv64_3 = extractResidualLayerParams('conv64_3');\n\n  const conv128_down = extractResidualLayerParams('conv128_down');\n  const conv128_1 = extractResidualLayerParams('conv128_1');\n  const conv128_2 = extractResidualLayerParams('conv128_2');\n\n  const conv256_down = extractResidualLayerParams('conv256_down');\n  const conv256_1 = extractResidualLayerParams('conv256_1');\n  const conv256_2 = extractResidualLayerParams('conv256_2');\n  const conv256_down_out = extractResidualLayerParams('conv256_down_out');\n\n  const { fc } = weightMap;\n  paramMappings.push({ originalPath: 'fc', paramPath: 'fc' });\n\n  if (!isTensor2D(fc)) {\n    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);\n  }\n\n  const params = {\n    conv32_down,\n    conv32_1,\n    conv32_2,\n    conv32_3,\n    conv64_down,\n    conv64_1,\n    conv64_2,\n    conv64_3,\n    conv128_down,\n    conv128_1,\n    conv128_2,\n    conv256_down,\n    conv256_1,\n    conv256_2,\n    conv256_down_out,\n    fc,\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { conv, convDown, convNoRelu } from './convLayer';\nimport { ResidualLayerParams } from './types';\n\nexport function residual(x: tf.Tensor4D, params: ResidualLayerParams): tf.Tensor4D {\n  let out = conv(x, params.conv1);\n  out = convNoRelu(out, params.conv2);\n  out = tf.add(out, x);\n  out = tf.relu(out);\n  return out;\n}\n\nexport function residualDown(x: tf.Tensor4D, params: ResidualLayerParams): tf.Tensor4D {\n  let out = convDown(x, params.conv1);\n  out = convNoRelu(out, params.conv2);\n\n  let pooled = tf.avgPool(x, 2, 2, 'valid') as tf.Tensor4D;\n  const zeros = tf.zeros<tf.Rank.R4>(pooled.shape);\n  const isPad = pooled.shape[3] !== out.shape[3];\n  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];\n\n  if (isAdjustShape) {\n    const padShapeX = [...out.shape] as [number, number, number, number];\n    padShapeX[1] = 1;\n    const zerosW = tf.zeros<tf.Rank.R4>(padShapeX);\n    out = tf.concat([out, zerosW], 1);\n\n    const padShapeY = [...out.shape] as [number, number, number, number];\n    padShapeY[2] = 1;\n    const zerosH = tf.zeros<tf.Rank.R4>(padShapeY);\n    out = tf.concat([out, zerosH], 2);\n  }\n\n  pooled = isPad ? tf.concat([pooled, zeros], 3) : pooled;\n  out = tf.add(pooled, out) as tf.Tensor4D;\n\n  out = tf.relu(out);\n  return out;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { normalize } from '../ops/index';\nimport { convDown } from './convLayer';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { residual, residualDown } from './residualLayer';\nimport { NetParams } from './types';\n\nexport class FaceRecognitionNet extends NeuralNetwork<NetParams> {\n  constructor() {\n    super('FaceRecognitionNet');\n  }\n\n  public forwardInput(input: NetInput): tf.Tensor2D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('FaceRecognitionNet - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(150, true), 'float32');\n\n      const meanRgb = [122.782, 117.001, 104.298];\n      const normalized = normalize(batchTensor, meanRgb).div(255) as tf.Tensor4D;\n\n      let out = convDown(normalized, params.conv32_down);\n      out = tf.maxPool(out, 3, 2, 'valid');\n\n      out = residual(out, params.conv32_1);\n      out = residual(out, params.conv32_2);\n      out = residual(out, params.conv32_3);\n\n      out = residualDown(out, params.conv64_down);\n      out = residual(out, params.conv64_1);\n      out = residual(out, params.conv64_2);\n      out = residual(out, params.conv64_3);\n\n      out = residualDown(out, params.conv128_down);\n      out = residual(out, params.conv128_1);\n      out = residual(out, params.conv128_2);\n\n      out = residualDown(out, params.conv256_down);\n      out = residual(out, params.conv256_1);\n      out = residual(out, params.conv256_2);\n      out = residualDown(out, params.conv256_down_out);\n\n      const globalAvg = out.mean([1, 2]) as tf.Tensor2D;\n      const fullyConnected = tf.matMul(globalAvg, params.fc);\n\n      return fullyConnected as tf.Tensor2D;\n    });\n  }\n\n  public async forward(input: TNetInput): Promise<tf.Tensor2D> {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async computeFaceDescriptor(input: TNetInput): Promise<Float32Array|Float32Array[]> {\n    // @ts-ignore\n    if (input?.shape?.some((dim) => dim <= 0)) return new Float32Array(128);\n    const netInput = await toNetInput(input);\n    const faceDescriptorTensors = tf.tidy(() => tf.unstack(this.forwardInput(netInput)));\n    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data())) as Float32Array[];\n    faceDescriptorTensors.forEach((t) => t.dispose());\n    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];\n  }\n\n  protected getDefaultModelName(): string {\n    return 'face_recognition_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import { FaceRecognitionNet } from './FaceRecognitionNet';\n\nexport * from './FaceRecognitionNet';\n\nexport function createFaceRecognitionNet(weights: Float32Array) {\n  const net = new FaceRecognitionNet();\n  net.extractWeights(weights);\n  return net;\n}\n", "export type WithFaceDescriptor<TSource> = TSource & {\n  descriptor: Float32Array\n}\n\nexport function extendWithFaceDescriptor<\n  TSource\n>(\n  sourceObj: TSource,\n  descriptor: Float32Array,\n): WithFaceDescriptor<TSource> {\n  const extension = { descriptor };\n  return { ...sourceObj, ...extension };\n}\n", "export type WithAge<TSource> = TSource & {\n  age: number\n}\n\nexport function isWithAge(obj: any): obj is WithAge<{}> {\n  return typeof obj.age === 'number';\n}\n\nexport function extendWithAge<\n  TSource\n>(\n  sourceObj: TSource,\n  age: number,\n): WithAge<TSource> {\n  const extension = { age };\n  return { ...sourceObj, ...extension };\n}\n", "import { Gender } from '../ageGenderNet/types';\nimport { isValidProbablitiy } from '../utils/index';\n\nexport type WithGender<TSource> = TSource & {\n  gender: Gender\n  genderProbability: number\n}\n\nexport function isWithGender(obj: any): obj is WithGender<{}> {\n  return (obj.gender === Gender.MALE || obj.gender === Gender.FEMALE)\n    && isValidProbablitiy(obj.genderProbability);\n}\n\nexport function extendWithGender<\n  TSource\n>(\n  sourceObj: TSource,\n  gender: Gender,\n  genderProbability: number,\n): WithGender<TSource> {\n  const extension = { gender, genderProbability };\n  return { ...sourceObj, ...extension };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ExtractWeightsFunction, ParamMapping, ConvParams, extractWeightsFactory } from '../common/index';\nimport { MobileNetV1, NetParams, PointwiseConvParams, PredictionLayerParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  function extractDepthwiseConvParams(numChannels: number, mappedPrefix: string): MobileNetV1.DepthwiseConvParams {\n    const filters = tf.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);\n    const batch_norm_scale = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_offset = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_mean = tf.tensor1d(extractWeights(numChannels));\n    const batch_norm_variance = tf.tensor1d(extractWeights(numChannels));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/batch_norm_scale` },\n      { paramPath: `${mappedPrefix}/batch_norm_offset` },\n      { paramPath: `${mappedPrefix}/batch_norm_mean` },\n      { paramPath: `${mappedPrefix}/batch_norm_variance` },\n    );\n\n    return {\n      filters,\n      batch_norm_scale,\n      batch_norm_offset,\n      batch_norm_mean,\n      batch_norm_variance,\n    };\n  }\n\n  function extractConvParams(\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n    isPointwiseConv?: boolean,\n  ): ConvParams {\n    const filters = tf.tensor4d(\n      extractWeights(channelsIn * channelsOut * filterSize * filterSize),\n      [filterSize, filterSize, channelsIn, channelsOut],\n    );\n    const bias = tf.tensor1d(extractWeights(channelsOut));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/filters` },\n      { paramPath: `${mappedPrefix}/${isPointwiseConv ? 'batch_norm_offset' : 'bias'}` },\n    );\n\n    return { filters, bias };\n  }\n\n  function extractPointwiseConvParams(\n    channelsIn: number,\n    channelsOut: number,\n    filterSize: number,\n    mappedPrefix: string,\n  ): PointwiseConvParams {\n    const {\n      filters,\n      bias,\n    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);\n\n    return {\n      filters,\n      batch_norm_offset: bias,\n    };\n  }\n\n  function extractConvPairParams(\n    channelsIn: number,\n    channelsOut: number,\n    mappedPrefix: string,\n  ): MobileNetV1.ConvPairParams {\n    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);\n    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);\n\n    return { depthwise_conv, pointwise_conv };\n  }\n\n  function extractMobilenetV1Params(): MobileNetV1.Params {\n    const conv_0 = extractPointwiseConvParams(3, 32, 3, 'mobilenetv1/conv_0');\n    const conv_1 = extractConvPairParams(32, 64, 'mobilenetv1/conv_1');\n    const conv_2 = extractConvPairParams(64, 128, 'mobilenetv1/conv_2');\n    const conv_3 = extractConvPairParams(128, 128, 'mobilenetv1/conv_3');\n    const conv_4 = extractConvPairParams(128, 256, 'mobilenetv1/conv_4');\n    const conv_5 = extractConvPairParams(256, 256, 'mobilenetv1/conv_5');\n    const conv_6 = extractConvPairParams(256, 512, 'mobilenetv1/conv_6');\n    const conv_7 = extractConvPairParams(512, 512, 'mobilenetv1/conv_7');\n    const conv_8 = extractConvPairParams(512, 512, 'mobilenetv1/conv_8');\n    const conv_9 = extractConvPairParams(512, 512, 'mobilenetv1/conv_9');\n    const conv_10 = extractConvPairParams(512, 512, 'mobilenetv1/conv_10');\n    const conv_11 = extractConvPairParams(512, 512, 'mobilenetv1/conv_11');\n    const conv_12 = extractConvPairParams(512, 1024, 'mobilenetv1/conv_12');\n    const conv_13 = extractConvPairParams(1024, 1024, 'mobilenetv1/conv_13');\n    return {\n      conv_0,\n      conv_1,\n      conv_2,\n      conv_3,\n      conv_4,\n      conv_5,\n      conv_6,\n      conv_7,\n      conv_8,\n      conv_9,\n      conv_10,\n      conv_11,\n      conv_12,\n      conv_13,\n    };\n  }\n\n  function extractPredictionLayerParams(): PredictionLayerParams {\n    const conv_0 = extractPointwiseConvParams(1024, 256, 1, 'prediction_layer/conv_0');\n    const conv_1 = extractPointwiseConvParams(256, 512, 3, 'prediction_layer/conv_1');\n    const conv_2 = extractPointwiseConvParams(512, 128, 1, 'prediction_layer/conv_2');\n    const conv_3 = extractPointwiseConvParams(128, 256, 3, 'prediction_layer/conv_3');\n    const conv_4 = extractPointwiseConvParams(256, 128, 1, 'prediction_layer/conv_4');\n    const conv_5 = extractPointwiseConvParams(128, 256, 3, 'prediction_layer/conv_5');\n    const conv_6 = extractPointwiseConvParams(256, 64, 1, 'prediction_layer/conv_6');\n    const conv_7 = extractPointwiseConvParams(64, 128, 3, 'prediction_layer/conv_7');\n    const box_encoding_0_predictor = extractConvParams(512, 12, 1, 'prediction_layer/box_predictor_0/box_encoding_predictor');\n    const class_predictor_0 = extractConvParams(512, 9, 1, 'prediction_layer/box_predictor_0/class_predictor');\n    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, 'prediction_layer/box_predictor_1/box_encoding_predictor');\n    const class_predictor_1 = extractConvParams(1024, 18, 1, 'prediction_layer/box_predictor_1/class_predictor');\n    const box_encoding_2_predictor = extractConvParams(512, 24, 1, 'prediction_layer/box_predictor_2/box_encoding_predictor');\n    const class_predictor_2 = extractConvParams(512, 18, 1, 'prediction_layer/box_predictor_2/class_predictor');\n    const box_encoding_3_predictor = extractConvParams(256, 24, 1, 'prediction_layer/box_predictor_3/box_encoding_predictor');\n    const class_predictor_3 = extractConvParams(256, 18, 1, 'prediction_layer/box_predictor_3/class_predictor');\n    const box_encoding_4_predictor = extractConvParams(256, 24, 1, 'prediction_layer/box_predictor_4/box_encoding_predictor');\n    const class_predictor_4 = extractConvParams(256, 18, 1, 'prediction_layer/box_predictor_4/class_predictor');\n    const box_encoding_5_predictor = extractConvParams(128, 24, 1, 'prediction_layer/box_predictor_5/box_encoding_predictor');\n    const class_predictor_5 = extractConvParams(128, 18, 1, 'prediction_layer/box_predictor_5/class_predictor');\n\n    const box_predictor_0 = {\n      box_encoding_predictor: box_encoding_0_predictor,\n      class_predictor: class_predictor_0,\n    };\n    const box_predictor_1 = {\n      box_encoding_predictor: box_encoding_1_predictor,\n      class_predictor: class_predictor_1,\n    };\n    const box_predictor_2 = {\n      box_encoding_predictor: box_encoding_2_predictor,\n      class_predictor: class_predictor_2,\n    };\n    const box_predictor_3 = {\n      box_encoding_predictor: box_encoding_3_predictor,\n      class_predictor: class_predictor_3,\n    };\n    const box_predictor_4 = {\n      box_encoding_predictor: box_encoding_4_predictor,\n      class_predictor: class_predictor_4,\n    };\n    const box_predictor_5 = {\n      box_encoding_predictor: box_encoding_5_predictor,\n      class_predictor: class_predictor_5,\n    };\n    return {\n      conv_0,\n      conv_1,\n      conv_2,\n      conv_3,\n      conv_4,\n      conv_5,\n      conv_6,\n      conv_7,\n      box_predictor_0,\n      box_predictor_1,\n      box_predictor_2,\n      box_predictor_3,\n      box_predictor_4,\n      box_predictor_5,\n    };\n  }\n\n  return {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  };\n}\n\nexport function extractParams(weights: Float32Array): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n  const {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n  const mobilenetv1 = extractMobilenetV1Params();\n  const prediction_layer = extractPredictionLayerParams();\n  const extra_dim = tf.tensor3d(\n    extractWeights(5118 * 4),\n    [1, 5118, 4],\n  );\n  const output_layer = {\n    extra_dim,\n  };\n  paramMappings.push({ paramPath: 'output_layer/extra_dim' });\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n\n  return {\n    params: {\n      mobilenetv1,\n      prediction_layer,\n      output_layer,\n    },\n    paramMappings,\n  };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams, disposeUnusedWeightTensors, extractWeightEntryFactory, ParamMapping } from '../common/index';\nimport { isTensor3D } from '../utils/index';\nimport { BoxPredictionParams, MobileNetV1, NetParams, PointwiseConvParams, PredictionLayerParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractPointwiseConvParams(prefix: string, idx: number, mappedPrefix: string): PointwiseConvParams {\n    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);\n    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);\n    return { filters, batch_norm_offset };\n  }\n\n  function extractConvPairParams(idx: number): MobileNetV1.ConvPairParams {\n    const mappedPrefix = `mobilenetv1/conv_${idx}`;\n    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;\n    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;\n    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;\n\n    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);\n    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);\n    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);\n    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);\n    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);\n\n    return {\n      depthwise_conv: {\n        filters,\n        batch_norm_scale,\n        batch_norm_offset,\n        batch_norm_mean,\n        batch_norm_variance,\n      },\n      pointwise_conv: extractPointwiseConvParams('MobilenetV1', idx, mappedPrefixPointwiseConv),\n    };\n  }\n\n  function extractMobilenetV1Params(): MobileNetV1.Params {\n    return {\n      conv_0: extractPointwiseConvParams('MobilenetV1', 0, 'mobilenetv1/conv_0'),\n      conv_1: extractConvPairParams(1),\n      conv_2: extractConvPairParams(2),\n      conv_3: extractConvPairParams(3),\n      conv_4: extractConvPairParams(4),\n      conv_5: extractConvPairParams(5),\n      conv_6: extractConvPairParams(6),\n      conv_7: extractConvPairParams(7),\n      conv_8: extractConvPairParams(8),\n      conv_9: extractConvPairParams(9),\n      conv_10: extractConvPairParams(10),\n      conv_11: extractConvPairParams(11),\n      conv_12: extractConvPairParams(12),\n      conv_13: extractConvPairParams(13),\n    };\n  }\n\n  function extractConvParams(prefix: string, mappedPrefix: string): ConvParams {\n    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);\n    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);\n    return { filters, bias };\n  }\n\n  function extractBoxPredictorParams(idx: number): BoxPredictionParams {\n    const box_encoding_predictor = extractConvParams(\n      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,\n      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`,\n    );\n    const class_predictor = extractConvParams(\n      `Prediction/BoxPredictor_${idx}/ClassPredictor`,\n      `prediction_layer/box_predictor_${idx}/class_predictor`,\n    );\n    return { box_encoding_predictor, class_predictor };\n  }\n\n  function extractPredictionLayerParams(): PredictionLayerParams {\n    return {\n      conv_0: extractPointwiseConvParams('Prediction', 0, 'prediction_layer/conv_0'),\n      conv_1: extractPointwiseConvParams('Prediction', 1, 'prediction_layer/conv_1'),\n      conv_2: extractPointwiseConvParams('Prediction', 2, 'prediction_layer/conv_2'),\n      conv_3: extractPointwiseConvParams('Prediction', 3, 'prediction_layer/conv_3'),\n      conv_4: extractPointwiseConvParams('Prediction', 4, 'prediction_layer/conv_4'),\n      conv_5: extractPointwiseConvParams('Prediction', 5, 'prediction_layer/conv_5'),\n      conv_6: extractPointwiseConvParams('Prediction', 6, 'prediction_layer/conv_6'),\n      conv_7: extractPointwiseConvParams('Prediction', 7, 'prediction_layer/conv_7'),\n      box_predictor_0: extractBoxPredictorParams(0),\n      box_predictor_1: extractBoxPredictorParams(1),\n      box_predictor_2: extractBoxPredictorParams(2),\n      box_predictor_3: extractBoxPredictorParams(3),\n      box_predictor_4: extractBoxPredictorParams(4),\n      box_predictor_5: extractBoxPredictorParams(5),\n    };\n  }\n\n  return {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n): { params: NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractMobilenetV1Params,\n    extractPredictionLayerParams,\n  } = extractorsFactory(weightMap, paramMappings);\n  const extra_dim = weightMap['Output/extra_dim'];\n  paramMappings.push({ originalPath: 'Output/extra_dim', paramPath: 'output_layer/extra_dim' });\n  if (!isTensor3D(extra_dim)) {\n    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);\n  }\n\n  const params = {\n    mobilenetv1: extractMobilenetV1Params(),\n    prediction_layer: extractPredictionLayerParams(),\n    output_layer: {\n      extra_dim,\n    },\n  };\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { PointwiseConvParams } from './types';\n\nexport function pointwiseConvLayer(x: tf.Tensor4D, params: PointwiseConvParams, strides: [number, number]) {\n  return tf.tidy(() => {\n    let out = tf.conv2d(x, params.filters, strides, 'same');\n    out = tf.add(out, params.batch_norm_offset);\n    return tf.clipByValue(out, 0, 6);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { pointwiseConvLayer } from './pointwiseConvLayer';\nimport { MobileNetV1 } from './types';\n\nconst epsilon = 0.0010000000474974513;\n\nfunction depthwiseConvLayer(x: tf.Tensor4D, params: MobileNetV1.DepthwiseConvParams, strides: [number, number]) {\n  return tf.tidy(() => {\n    let out = tf.depthwiseConv2d(x, params.filters, strides, 'same');\n    out = tf.batchNorm<tf.Rank.R4>(\n      out,\n      params.batch_norm_mean,\n      params.batch_norm_variance,\n      params.batch_norm_offset,\n      params.batch_norm_scale,\n      epsilon,\n    );\n    return tf.clipByValue(out, 0, 6);\n  });\n}\n\nfunction getStridesForLayerIdx(layerIdx: number): [number, number] {\n  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];\n}\n\nexport function mobileNetV1(x: tf.Tensor4D, params: MobileNetV1.Params) {\n  return tf.tidy(() => {\n    let conv11;\n    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);\n\n    const convPairParams = [\n      params.conv_1,\n      params.conv_2,\n      params.conv_3,\n      params.conv_4,\n      params.conv_5,\n      params.conv_6,\n      params.conv_7,\n      params.conv_8,\n      params.conv_9,\n      params.conv_10,\n      params.conv_11,\n      params.conv_12,\n      params.conv_13,\n    ];\n\n    convPairParams.forEach((param, i) => {\n      const layerIdx = i + 1;\n      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);\n      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);\n      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);\n      if (layerIdx === 11) conv11 = out;\n    });\n\n    if (conv11 === null) {\n      throw new Error('mobileNetV1 - output of conv layer 11 is null');\n    }\n\n    return {\n      out,\n      conv11: conv11 as any,\n    };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nfunction IOU(boxes: tf.Tensor2D, i: number, j: number) {\n  const boxesData = boxes.arraySync();\n  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);\n  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);\n  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);\n  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);\n  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);\n  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);\n  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);\n  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);\n  const areaI = (ymaxI - yminI) * (xmaxI - xminI);\n  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);\n  if (areaI <= 0 || areaJ <= 0) return 0.0;\n  const intersectionYmin = Math.max(yminI, yminJ);\n  const intersectionXmin = Math.max(xminI, xminJ);\n  const intersectionYmax = Math.min(ymaxI, ymaxJ);\n  const intersectionXmax = Math.min(xmaxI, xmaxJ);\n  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0.0) * Math.max(intersectionXmax - intersectionXmin, 0.0);\n  return intersectionArea / (areaI + areaJ - intersectionArea);\n}\n\nexport function nonMaxSuppression(\n  boxes: tf.Tensor2D,\n  scores: number[],\n  maxOutputSize: number,\n  iouThreshold: number,\n  scoreThreshold: number,\n): number[] {\n  const numBoxes = boxes.shape[0];\n  const outputSize = Math.min(maxOutputSize, numBoxes);\n\n  const candidates = scores\n    .map((score, boxIndex) => ({ score, boxIndex }))\n    .filter((c) => c.score > scoreThreshold)\n    .sort((c1, c2) => c2.score - c1.score);\n\n  const suppressFunc = (x: number) => (x <= iouThreshold ? 1 : 0);\n  const selected: number[] = [];\n\n  candidates.forEach((c) => {\n    if (selected.length >= outputSize) return;\n    const originalScore = c.score;\n    for (let j = selected.length - 1; j >= 0; --j) {\n      const iou = IOU(boxes, c.boxIndex, selected[j]);\n      if (iou === 0.0) continue;\n      c.score *= suppressFunc(iou);\n      if (c.score <= scoreThreshold) break;\n    }\n    if (originalScore === c.score) {\n      selected.push(c.boxIndex);\n    }\n  });\n  return selected;\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { OutputLayerParams } from './types';\n\nfunction getCenterCoordinatesAndSizesLayer(x: tf.Tensor2D) {\n  const vec = tf.unstack(tf.transpose(x, [1, 0]));\n\n  const sizes = [\n    tf.sub(vec[2], vec[0]),\n    tf.sub(vec[3], vec[1]),\n  ];\n  const centers = [\n    tf.add(vec[0], tf.div(sizes[0], 2)),\n    tf.add(vec[1], tf.div(sizes[1], 2)),\n  ];\n  return { sizes, centers };\n}\n\nfunction decodeBoxesLayer(x0: tf.Tensor2D, x1: tf.Tensor2D) {\n  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);\n\n  const vec = tf.unstack(tf.transpose(x1, [1, 0]));\n  const div0_out = tf.div(tf.mul(tf.exp(tf.div(vec[2], 5)), sizes[0]), 2);\n  const add0_out = tf.add(tf.mul(tf.div(vec[0], 10), sizes[0]), centers[0]);\n  const div1_out = tf.div(tf.mul(tf.exp(tf.div(vec[3], 5)), sizes[1]), 2);\n  const add1_out = tf.add(tf.mul(tf.div(vec[1], 10), sizes[1]), centers[1]);\n\n  return tf.transpose(\n    tf.stack([\n      tf.sub(add0_out, div0_out),\n      tf.sub(add1_out, div1_out),\n      tf.add(add0_out, div0_out),\n      tf.add(add1_out, div1_out),\n    ]),\n    [1, 0],\n  );\n}\n\nexport function outputLayer(boxPredictions: tf.Tensor4D, classPredictions: tf.Tensor4D, params: OutputLayerParams) {\n  return tf.tidy(() => {\n    const batchSize = boxPredictions.shape[0];\n\n    let boxes = decodeBoxesLayer(\n      tf.reshape(tf.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]) as tf.Tensor2D,\n      tf.reshape(boxPredictions, [-1, 4]) as tf.Tensor2D,\n    );\n    boxes = tf.reshape(boxes, [batchSize, (boxes.shape[0] / batchSize), 4]);\n\n    const scoresAndClasses = tf.sigmoid(tf.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));\n    let scores = tf.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]) as tf.Tensor;\n\n    scores = tf.reshape(scores, [batchSize, scores.shape[1] as number]);\n\n    const boxesByBatch = tf.unstack(boxes) as tf.Tensor2D[];\n    const scoresByBatch = tf.unstack(scores) as tf.Tensor1D[];\n\n    return { boxes: boxesByBatch, scores: scoresByBatch };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { convLayer } from '../common/index';\nimport { BoxPredictionParams } from './types';\n\nexport function boxPredictionLayer(\n  x: tf.Tensor4D,\n  params: BoxPredictionParams,\n) {\n  return tf.tidy(() => {\n    const batchSize = x.shape[0];\n    const boxPredictionEncoding = tf.reshape(\n      convLayer(x, params.box_encoding_predictor),\n      [batchSize, -1, 1, 4],\n    );\n    const classPrediction = tf.reshape(\n      convLayer(x, params.class_predictor),\n      [batchSize, -1, 3],\n    );\n    return { boxPredictionEncoding, classPrediction };\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { boxPredictionLayer } from './boxPredictionLayer';\nimport { pointwiseConvLayer } from './pointwiseConvLayer';\nimport { PredictionLayerParams } from './types';\n\nexport function predictionLayer(\n  x: tf.Tensor4D,\n  conv11: tf.Tensor4D,\n  params: PredictionLayerParams,\n) {\n  return tf.tidy(() => {\n    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);\n    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);\n    const conv2 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);\n    const conv3 = pointwiseConvLayer(conv2, params.conv_3, [2, 2]);\n    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);\n    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);\n    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);\n    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);\n\n    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);\n    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);\n    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);\n    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);\n    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);\n    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);\n\n    const boxPredictions = tf.concat([\n      boxPrediction0.boxPredictionEncoding,\n      boxPrediction1.boxPredictionEncoding,\n      boxPrediction2.boxPredictionEncoding,\n      boxPrediction3.boxPredictionEncoding,\n      boxPrediction4.boxPredictionEncoding,\n      boxPrediction5.boxPredictionEncoding,\n    ], 1) as tf.Tensor4D;\n\n    const classPredictions = tf.concat([\n      boxPrediction0.classPrediction,\n      boxPrediction1.classPrediction,\n      boxPrediction2.classPrediction,\n      boxPrediction3.classPrediction,\n      boxPrediction4.classPrediction,\n      boxPrediction5.classPrediction,\n    ], 1) as tf.Tensor4D;\n\n    return {\n      boxPredictions,\n      classPredictions,\n    };\n  });\n}\n", "export interface ISsdMobilenetv1Options {\n  minConfidence?: number\n  maxResults?: number\n}\n\nexport class SsdMobilenetv1Options {\n  protected _name = 'SsdMobilenetv1Options';\n\n  private _minConfidence: number;\n\n  private _maxResults: number;\n\n  constructor({ minConfidence, maxResults }: ISsdMobilenetv1Options = {}) {\n    this._minConfidence = minConfidence || 0.5;\n    this._maxResults = maxResults || 100;\n\n    if (typeof this._minConfidence !== 'number' || this._minConfidence <= 0 || this._minConfidence >= 1) {\n      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);\n    }\n\n    if (typeof this._maxResults !== 'number') {\n      throw new Error(`${this._name} - expected maxResults to be a number`);\n    }\n  }\n\n  get minConfidence(): number { return this._minConfidence; }\n\n  get maxResults(): number { return this._maxResults; }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { Rect } from '../classes/index';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { NetInput, TNetInput, toNetInput } from '../dom/index';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { mobileNetV1 } from './mobileNetV1';\nimport { nonMaxSuppression } from './nonMaxSuppression';\nimport { outputLayer } from './outputLayer';\nimport { predictionLayer } from './predictionLayer';\nimport { ISsdMobilenetv1Options, SsdMobilenetv1Options } from './SsdMobilenetv1Options';\nimport { NetParams } from './types';\n\nexport class SsdMobilenetv1 extends NeuralNetwork<NetParams> {\n  constructor() {\n    super('SsdMobilenetv1');\n  }\n\n  public forwardInput(input: NetInput) {\n    const { params } = this;\n    if (!params) throw new Error('SsdMobilenetv1 - load model before inference');\n    return tf.tidy(() => {\n      const batchTensor = tf.cast(input.toBatchTensor(512, false), 'float32');\n      const x = tf.sub(tf.div(batchTensor, 127.5), 1) as tf.Tensor4D; // input is normalized -1..1\n      const features = mobileNetV1(x, params.mobilenetv1);\n      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);\n      return outputLayer(boxPredictions, classPredictions, params.output_layer);\n    });\n  }\n\n  public async forward(input: TNetInput) {\n    return this.forwardInput(await toNetInput(input));\n  }\n\n  public async locateFaces(input: TNetInput, options: ISsdMobilenetv1Options = {}): Promise<FaceDetection[]> {\n    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);\n    const netInput = await toNetInput(input);\n    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);\n    const boxes = _boxes[0];\n    const scores = _scores[0];\n    for (let i = 1; i < _boxes.length; i++) {\n      _boxes[i].dispose();\n      _scores[i].dispose();\n    }\n    const scoresData = Array.from(scores.dataSync());\n    const iouThreshold = 0.5;\n    const indices = nonMaxSuppression(boxes, scoresData as number[], maxResults, iouThreshold, minConfidence);\n    const reshapedDims = netInput.getReshapedInputDimensions(0);\n    const inputSize = netInput.inputSize as number;\n    const padX = inputSize / reshapedDims.width;\n    const padY = inputSize / reshapedDims.height;\n    const boxesData = boxes.arraySync();\n    const results = indices\n      .map((idx) => {\n        const [top, bottom] = [\n          Math.max(0, boxesData[idx][0]),\n          Math.min(1.0, boxesData[idx][2]),\n        ].map((val) => val * padY);\n        const [left, right] = [\n          Math.max(0, boxesData[idx][1]),\n          Math.min(1.0, boxesData[idx][3]),\n        ].map((val) => val * padX);\n        return new FaceDetection(\n          scoresData[idx] as number,\n          new Rect(left, top, right - left, bottom - top),\n          { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) },\n        );\n      });\n    boxes.dispose();\n    scores.dispose();\n    return results;\n  }\n\n  protected getDefaultModelName(): string {\n    return 'ssd_mobilenetv1_model';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    return extractParams(weights);\n  }\n}\n", "import { SsdMobilenetv1 } from './SsdMobilenetv1';\n\nexport * from './SsdMobilenetv1';\nexport * from './SsdMobilenetv1Options';\n\nexport function createSsdMobilenetv1(weights: Float32Array) {\n  const net = new SsdMobilenetv1();\n  net.extractWeights(weights);\n  return net;\n}\n\nexport function createFaceDetectionNet(weights: Float32Array) {\n  return createSsdMobilenetv1(weights);\n}\n\n// alias for backward compatibily\nexport class FaceDetectionNet extends SsdMobilenetv1 {}\n", "import { Point } from '../classes/index';\n\nexport const IOU_THRESHOLD = 0.4;\n\nexport const BOX_ANCHORS = [\n  new Point(0.738768, 0.874946),\n  new Point(2.42204, 2.65704),\n  new Point(4.30971, 7.04493),\n  new Point(10.246, 4.59428),\n  new Point(12.6868, 11.8741),\n];\n\nexport const BOX_ANCHORS_SEPARABLE = [\n  new Point(1.603231, 2.094468),\n  new Point(6.041143, 7.080126),\n  new Point(2.882459, 3.518061),\n  new Point(4.266906, 5.178857),\n  new Point(9.041765, 10.66308),\n];\n\nexport const MEAN_RGB_SEPARABLE: [number, number, number] = [117.001, 114.697, 97.404];\n\nexport const DEFAULT_MODEL_NAME = 'tiny_yolov2_model';\nexport const DEFAULT_MODEL_NAME_SEPARABLE_CONV = 'tiny_yolov2_separable_conv_model';\n", "import { Point } from '../classes/Point';\n\nexport type TinyYolov2Config = {\n  withSeparableConvs: boolean\n  iouThreshold: number\n  anchors: Point[]\n  classes: string[]\n  meanRgb?: [number, number, number]\n  withClassScores?: boolean,\n  filterSizes?: number[]\n  isFirstLayerConv2d?: boolean\n}\n\nconst isNumber = (arg: any) => typeof arg === 'number';\n\nexport function validateConfig(config: any) {\n  if (!config) {\n    throw new Error(`invalid config: ${config}`);\n  }\n\n  if (typeof config.withSeparableConvs !== 'boolean') {\n    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);\n  }\n\n  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1.0) {\n    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);\n  }\n\n  if (\n    !Array.isArray(config.classes)\n    || !config.classes.length\n    || !config.classes.every((c: any) => typeof c === 'string')\n  ) {\n    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);\n  }\n\n  if (\n    !Array.isArray(config.anchors)\n    || !config.anchors.length\n    || !config.anchors.map((a: any) => a || {}).every((a: any) => isNumber(a.x) && isNumber(a.y))\n  ) {\n    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);\n  }\n\n  if (config.meanRgb && (\n    !Array.isArray(config.meanRgb)\n    || config.meanRgb.length !== 3\n    || !config.meanRgb.every(isNumber)\n  )) {\n    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nexport function leaky(x: tf.Tensor4D): tf.Tensor4D {\n  return tf.tidy(() => {\n    const min = tf.mul(x, tf.scalar(0.10000000149011612));\n    return tf.add(tf.relu(tf.sub(x, min)), min);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { leaky } from './leaky';\nimport { ConvWithBatchNorm } from './types';\n\nexport function convWithBatchNorm(x: tf.Tensor4D, params: ConvWithBatchNorm): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]) as tf.Tensor4D;\n    out = tf.conv2d(out, params.conv.filters, [1, 1], 'valid');\n    out = tf.sub(out, params.bn.sub);\n    out = tf.mul(out, params.bn.truediv);\n    out = tf.add(out, params.conv.bias);\n    return leaky(out);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { SeparableConvParams } from '../common/types';\nimport { leaky } from './leaky';\n\nexport function depthwiseSeparableConv(x: tf.Tensor4D, params: SeparableConvParams): tf.Tensor4D {\n  return tf.tidy(() => {\n    let out = tf.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]) as tf.Tensor4D;\n    out = tf.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], 'valid');\n    out = tf.add(out, params.bias);\n    return leaky(out);\n  });\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { extractConvParamsFactory } from '../common/index';\nimport { extractSeparableConvParamsFactory } from '../common/extractSeparableConvParamsFactory';\nimport { extractWeightsFactory } from '../common/extractWeightsFactory';\nimport { ExtractWeightsFunction, ParamMapping } from '../common/types';\nimport { TinyYolov2Config } from './config';\nimport { BatchNorm, ConvWithBatchNorm, TinyYolov2NetParams } from './types';\n\nfunction extractorsFactory(extractWeights: ExtractWeightsFunction, paramMappings: ParamMapping[]) {\n  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);\n\n  function extractBatchNormParams(size: number, mappedPrefix: string): BatchNorm {\n    const sub = tf.tensor1d(extractWeights(size));\n    const truediv = tf.tensor1d(extractWeights(size));\n\n    paramMappings.push(\n      { paramPath: `${mappedPrefix}/sub` },\n      { paramPath: `${mappedPrefix}/truediv` },\n    );\n    return { sub, truediv };\n  }\n\n  function extractConvWithBatchNormParams(channelsIn: number, channelsOut: number, mappedPrefix: string): ConvWithBatchNorm {\n    const conv = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);\n    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);\n    return { conv, bn };\n  }\n  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);\n\n  return {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  };\n}\n\nexport function extractParams(\n  weights: Float32Array,\n  config: TinyYolov2Config,\n  boxEncodingSize: number,\n  filterSizes: number[],\n): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n  const {\n    extractWeights,\n    getRemainingWeights,\n  } = extractWeightsFactory(weights);\n\n  const paramMappings: ParamMapping[] = [];\n  const {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  } = extractorsFactory(extractWeights, paramMappings);\n  let params: TinyYolov2NetParams;\n\n  if (config.withSeparableConvs) {\n    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;\n    const conv0 = config.isFirstLayerConv2d\n      ? extractConvParams(s0, s1, 3, 'conv0')\n      : extractSeparableConvParams(s0, s1, 'conv0');\n    const conv1 = extractSeparableConvParams(s1, s2, 'conv1');\n    const conv2 = extractSeparableConvParams(s2, s3, 'conv2');\n    const conv3 = extractSeparableConvParams(s3, s4, 'conv3');\n    const conv4 = extractSeparableConvParams(s4, s5, 'conv4');\n    const conv5 = extractSeparableConvParams(s5, s6, 'conv5');\n    const conv6 = s7 ? extractSeparableConvParams(s6, s7, 'conv6') : undefined;\n    const conv7 = s8 ? extractSeparableConvParams(s7, s8, 'conv7') : undefined;\n    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, 'conv8');\n    params = {\n      conv0, conv1, conv2, conv3, conv4, conv5, conv6, conv7, conv8,\n    };\n  } else {\n    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;\n    const conv0 = extractConvWithBatchNormParams(s0, s1, 'conv0');\n    const conv1 = extractConvWithBatchNormParams(s1, s2, 'conv1');\n    const conv2 = extractConvWithBatchNormParams(s2, s3, 'conv2');\n    const conv3 = extractConvWithBatchNormParams(s3, s4, 'conv3');\n    const conv4 = extractConvWithBatchNormParams(s4, s5, 'conv4');\n    const conv5 = extractConvWithBatchNormParams(s5, s6, 'conv5');\n    const conv6 = extractConvWithBatchNormParams(s6, s7, 'conv6');\n    const conv7 = extractConvWithBatchNormParams(s7, s8, 'conv7');\n    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, 'conv8');\n    params = {\n      conv0, conv1, conv2, conv3, conv4, conv5, conv6, conv7, conv8,\n    };\n  }\n  if (getRemainingWeights().length !== 0) {\n    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);\n  }\n  return { params, paramMappings };\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { ConvParams } from '../common/index';\nimport { disposeUnusedWeightTensors } from '../common/disposeUnusedWeightTensors';\nimport { loadSeparableConvParamsFactory } from '../common/extractSeparableConvParamsFactory';\nimport { extractWeightEntryFactory } from '../common/extractWeightEntryFactory';\nimport { ParamMapping } from '../common/types';\nimport { TinyYolov2Config } from './config';\nimport { BatchNorm, ConvWithBatchNorm, TinyYolov2NetParams } from './types';\n\nfunction extractorsFactory(weightMap: any, paramMappings: ParamMapping[]) {\n  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);\n\n  function extractBatchNormParams(prefix: string): BatchNorm {\n    const sub = extractWeightEntry(`${prefix}/sub`, 1);\n    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);\n    return { sub, truediv };\n  }\n\n  function extractConvParams(prefix: string): ConvParams {\n    const filters = extractWeightEntry(`${prefix}/filters`, 4);\n    const bias = extractWeightEntry(`${prefix}/bias`, 1);\n    return { filters, bias };\n  }\n\n  function extractConvWithBatchNormParams(prefix: string): ConvWithBatchNorm {\n    const conv = extractConvParams(`${prefix}/conv`);\n    const bn = extractBatchNormParams(`${prefix}/bn`);\n    return { conv, bn };\n  }\n\n  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);\n  return {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  };\n}\n\nexport function extractParamsFromWeightMap(\n  weightMap: tf.NamedTensorMap,\n  config: TinyYolov2Config,\n): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n  const paramMappings: ParamMapping[] = [];\n\n  const {\n    extractConvParams,\n    extractConvWithBatchNormParams,\n    extractSeparableConvParams,\n  } = extractorsFactory(weightMap, paramMappings);\n\n  let params: TinyYolov2NetParams;\n\n  if (config.withSeparableConvs) {\n    // eslint-disable-next-line no-mixed-operators\n    const numFilters = (config.filterSizes && config.filterSizes.length || 9);\n    params = {\n      conv0: config.isFirstLayerConv2d ? extractConvParams('conv0') : extractSeparableConvParams('conv0'),\n      conv1: extractSeparableConvParams('conv1'),\n      conv2: extractSeparableConvParams('conv2'),\n      conv3: extractSeparableConvParams('conv3'),\n      conv4: extractSeparableConvParams('conv4'),\n      conv5: extractSeparableConvParams('conv5'),\n      conv6: numFilters > 7 ? extractSeparableConvParams('conv6') : undefined,\n      conv7: numFilters > 8 ? extractSeparableConvParams('conv7') : undefined,\n      conv8: extractConvParams('conv8'),\n    };\n  } else {\n    params = {\n      conv0: extractConvWithBatchNormParams('conv0'),\n      conv1: extractConvWithBatchNormParams('conv1'),\n      conv2: extractConvWithBatchNormParams('conv2'),\n      conv3: extractConvWithBatchNormParams('conv3'),\n      conv4: extractConvWithBatchNormParams('conv4'),\n      conv5: extractConvWithBatchNormParams('conv5'),\n      conv6: extractConvWithBatchNormParams('conv6'),\n      conv7: extractConvWithBatchNormParams('conv7'),\n      conv8: extractConvParams('conv8'),\n    };\n  }\n\n  disposeUnusedWeightTensors(weightMap, paramMappings);\n  return { params, paramMappings };\n}\n", "export interface ITinyYolov2Options {\n  inputSize?: number\n  scoreThreshold?: number\n}\n\nexport class TinyYolov2Options {\n  protected _name = 'TinyYolov2Options';\n\n  private _inputSize: number;\n\n  private _scoreThreshold: number;\n\n  constructor({ inputSize, scoreThreshold }: ITinyYolov2Options = {}) {\n    this._inputSize = inputSize || 416;\n    this._scoreThreshold = scoreThreshold || 0.5;\n\n    if (typeof this._inputSize !== 'number' || this._inputSize % 32 !== 0) {\n      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);\n    }\n\n    if (typeof this._scoreThreshold !== 'number' || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {\n      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);\n    }\n  }\n\n  get inputSize(): number { return this._inputSize; }\n\n  get scoreThreshold(): number { return this._scoreThreshold; }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { BoundingBox } from '../classes/BoundingBox';\nimport { Dimensions } from '../classes/Dimensions';\nimport { ObjectDetection } from '../classes/ObjectDetection';\nimport { convLayer } from '../common/index';\nimport { ConvParams, SeparableConvParams } from '../common/types';\nimport { toNetInput } from '../dom/index';\nimport { NetInput } from '../dom/NetInput';\nimport { TNetInput } from '../dom/types';\nimport { NeuralNetwork } from '../NeuralNetwork';\nimport { sigmoid } from '../ops/index';\nimport { nonMaxSuppression } from '../ops/nonMaxSuppression';\nimport { normalize } from '../ops/normalize';\nimport { TinyYolov2Config, validateConfig } from './config';\nimport { convWithBatchNorm } from './convWithBatchNorm';\nimport { depthwiseSeparableConv } from './depthwiseSeparableConv';\nimport { extractParams } from './extractParams';\nimport { extractParamsFromWeightMap } from './extractParamsFromWeightMap';\nimport { leaky } from './leaky';\nimport { ITinyYolov2Options, TinyYolov2Options } from './TinyYolov2Options';\nimport { DefaultTinyYolov2NetParams, MobilenetParams, TinyYolov2NetParams } from './types';\n\nexport class TinyYolov2Base extends NeuralNetwork<TinyYolov2NetParams> {\n  public static DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];\n\n  private _config: TinyYolov2Config;\n\n  constructor(config: TinyYolov2Config) {\n    super('TinyYolov2');\n    validateConfig(config);\n    this._config = config;\n  }\n\n  public get config(): TinyYolov2Config {\n    return this._config;\n  }\n\n  public get withClassScores(): boolean {\n    return this.config.withClassScores || this.config.classes.length > 1;\n  }\n\n  public get boxEncodingSize(): number {\n    return 5 + (this.withClassScores ? this.config.classes.length : 0);\n  }\n\n  public runTinyYolov2(x: tf.Tensor4D, params: DefaultTinyYolov2NetParams): tf.Tensor4D {\n    let out = convWithBatchNorm(x, params.conv0);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv1);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv2);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv3);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv4);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = convWithBatchNorm(out, params.conv5);\n    out = tf.maxPool(out, [2, 2], [1, 1], 'same');\n    out = convWithBatchNorm(out, params.conv6);\n    out = convWithBatchNorm(out, params.conv7);\n    return convLayer(out, params.conv8, 'valid', false);\n  }\n\n  public runMobilenet(x: tf.Tensor4D, params: MobilenetParams): tf.Tensor4D {\n    let out = this.config.isFirstLayerConv2d\n      ? leaky(convLayer(x, params.conv0 as ConvParams, 'valid', false))\n      : depthwiseSeparableConv(x, params.conv0 as SeparableConvParams);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv1);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv2);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv3);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv4);\n    out = tf.maxPool(out, [2, 2], [2, 2], 'same');\n    out = depthwiseSeparableConv(out, params.conv5);\n    out = tf.maxPool(out, [2, 2], [1, 1], 'same');\n    out = params.conv6 ? depthwiseSeparableConv(out, params.conv6) : out;\n    out = params.conv7 ? depthwiseSeparableConv(out, params.conv7) : out;\n    return convLayer(out, params.conv8, 'valid', false);\n  }\n\n  public forwardInput(input: NetInput, inputSize: number): tf.Tensor4D {\n    const { params } = this;\n\n    if (!params) {\n      throw new Error('TinyYolov2 - load model before inference');\n    }\n\n    return tf.tidy(() => {\n      let batchTensor = tf.cast(input.toBatchTensor(inputSize, false), 'float32');\n      batchTensor = this.config.meanRgb\n        ? normalize(batchTensor, this.config.meanRgb)\n        : batchTensor;\n      batchTensor = batchTensor.div(255) as tf.Tensor4D;\n      return this.config.withSeparableConvs\n        ? this.runMobilenet(batchTensor, params as MobilenetParams)\n        : this.runTinyYolov2(batchTensor, params as DefaultTinyYolov2NetParams);\n    });\n  }\n\n  public async forward(input: TNetInput, inputSize: number): Promise<tf.Tensor4D> {\n    return this.forwardInput(await toNetInput(input), inputSize);\n  }\n\n  public async detect(input: TNetInput, forwardParams: ITinyYolov2Options = {}): Promise<ObjectDetection[]> {\n    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);\n    const netInput = await toNetInput(input);\n    const out = await this.forwardInput(netInput, inputSize);\n    const out0 = tf.tidy(() => tf.unstack(out)[0].expandDims()) as tf.Tensor4D;\n    const inputDimensions = {\n      width: netInput.getInputWidth(0),\n      height: netInput.getInputHeight(0),\n    };\n\n    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);\n    out.dispose();\n    out0.dispose();\n\n    const boxes = results.map((res) => res.box);\n    const scores = results.map((res) => res.score);\n    const classScores = results.map((res) => res.classScore);\n    const classNames = results.map((res) => this.config.classes[res.label]);\n\n    const indices = nonMaxSuppression(\n      boxes.map((box) => box.rescale(inputSize)),\n      scores,\n      this.config.iouThreshold,\n      true,\n    );\n\n    const detections = indices.map((idx) => new ObjectDetection(\n      scores[idx],\n      classScores[idx],\n      classNames[idx],\n      boxes[idx],\n      inputDimensions,\n    ));\n    return detections;\n  }\n\n  protected getDefaultModelName(): string {\n    return '';\n  }\n\n  protected extractParamsFromWeightMap(weightMap: tf.NamedTensorMap) {\n    return extractParamsFromWeightMap(weightMap, this.config);\n  }\n\n  protected extractParams(weights: Float32Array) {\n    const filterSizes = this.config.filterSizes || TinyYolov2Base.DEFAULT_FILTER_SIZES;\n\n    const numFilters = filterSizes ? filterSizes.length : undefined;\n    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {\n      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);\n    }\n    return extractParams(weights, this.config, this.boxEncodingSize, filterSizes);\n  }\n\n  protected async extractBoxes(\n    outputTensor: tf.Tensor4D,\n    inputBlobDimensions: Dimensions,\n    scoreThreshold?: number,\n  ) {\n    const { width, height } = inputBlobDimensions;\n    const inputSize = Math.max(width, height);\n    const correctionFactorX = inputSize / width;\n    const correctionFactorY = inputSize / height;\n\n    const numCells = outputTensor.shape[1];\n    const numBoxes = this.config.anchors.length;\n\n    const [boxesTensor, scoresTensor, classScoresTensor] = tf.tidy(() => {\n      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);\n\n      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);\n      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);\n      const classScores = this.withClassScores\n        ? tf.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3)\n        : tf.scalar(0);\n      return [boxes, scores, classScores];\n    });\n\n    const results = [] as any;\n    const scoresData = await scoresTensor.array();\n    const boxesData = await boxesTensor.array();\n    for (let row = 0; row < numCells; row++) {\n      for (let col = 0; col < numCells; col++) {\n        for (let anchor = 0; anchor < numBoxes; anchor++) {\n          const score = sigmoid(scoresData[row][col][anchor][0]);\n          if (!scoreThreshold || score > scoreThreshold) {\n            const ctX = ((col + sigmoid(boxesData[row][col][anchor][0])) / numCells) * correctionFactorX;\n            const ctY = ((row + sigmoid(boxesData[row][col][anchor][1])) / numCells) * correctionFactorY;\n            const widthLocal = ((Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x) / numCells) * correctionFactorX;\n            const heightLocal = ((Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y) / numCells) * correctionFactorY;\n            const x = (ctX - (widthLocal / 2));\n            const y = (ctY - (heightLocal / 2));\n            const pos = { row, col, anchor };\n            const { classScore, label } = this.withClassScores\n              ? await this.extractPredictedClass(classScoresTensor as tf.Tensor4D, pos)\n              : { classScore: 1, label: 0 };\n            results.push({\n              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),\n              score,\n              classScore: score * classScore,\n              label,\n              ...pos,\n            });\n          }\n        }\n      }\n    }\n\n    boxesTensor.dispose();\n    scoresTensor.dispose();\n    classScoresTensor.dispose();\n    return results;\n  }\n\n  private async extractPredictedClass(classesTensor: tf.Tensor4D, pos: { row: number, col: number, anchor: number }) {\n    const { row, col, anchor } = pos;\n    const classesData = await classesTensor.array();\n    return Array(this.config.classes.length).fill(0)\n      .map((_, i) => classesData[row][col][anchor][i])\n      .map((classScore, label) => ({\n        classScore,\n        label,\n      }))\n      .reduce((max, curr) => (max.classScore > curr.classScore ? max : curr));\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection, Point } from '../classes/index';\nimport { ParamMapping } from '../common/types';\nimport { TNetInput } from '../dom/types';\nimport {\n  BOX_ANCHORS,\n  BOX_ANCHORS_SEPARABLE,\n  DEFAULT_MODEL_NAME,\n  DEFAULT_MODEL_NAME_SEPARABLE_CONV,\n  IOU_THRESHOLD,\n  MEAN_RGB_SEPARABLE,\n} from './const';\nimport { TinyYolov2Base } from './TinyYolov2Base';\nimport { ITinyYolov2Options } from './TinyYolov2Options';\nimport { TinyYolov2NetParams } from './types';\n\nexport class TinyYolov2 extends TinyYolov2Base {\n  constructor(withSeparableConvs = true) {\n    const config = {\n      withSeparableConvs,\n      iouThreshold: IOU_THRESHOLD,\n      classes: ['face'],\n      ...(withSeparableConvs\n        ? {\n          anchors: BOX_ANCHORS_SEPARABLE,\n          meanRgb: MEAN_RGB_SEPARABLE,\n        }\n        : {\n          anchors: BOX_ANCHORS,\n          withClassScores: true,\n        }),\n    };\n\n    super(config);\n  }\n\n  public get withSeparableConvs(): boolean {\n    return this.config.withSeparableConvs;\n  }\n\n  public get anchors(): Point[] {\n    return this.config.anchors;\n  }\n\n  public async locateFaces(input: TNetInput, forwardParams: ITinyYolov2Options): Promise<FaceDetection[]> {\n    const objectDetections = await this.detect(input, forwardParams);\n    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));\n  }\n\n  protected override getDefaultModelName(): string {\n    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;\n  }\n\n  protected override extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n    return super.extractParamsFromWeightMap(weightMap);\n  }\n}\n", "import { TinyYolov2 } from './TinyYolov2';\n\nexport * from './TinyYolov2Options';\nexport * from './config';\nexport * from './types';\nexport { TinyYolov2 };\n\nexport function createTinyYolov2(weights: Float32Array, withSeparableConvs = true) {\n  const net = new TinyYolov2(withSeparableConvs);\n  net.extractWeights(weights);\n  return net;\n}\n", "import { ITinyYolov2Options, TinyYolov2Options } from '../tinyYolov2/index';\n\nexport type ITinyFaceDetectorOptions = ITinyYolov2Options\n\nexport class TinyFaceDetectorOptions extends TinyYolov2Options {\n  protected override _name = 'TinyFaceDetectorOptions';\n}\n", "export class ComposableTask<T> {\n  // eslint-disable-next-line no-unused-vars\n  public async then(onfulfilled: (value: T) => T | PromiseLike<T>): Promise<T> {\n    return onfulfilled(await this.run());\n  }\n\n  public async run(): Promise<T> {\n    throw new Error('ComposableTask - run is not implemented');\n  }\n}\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { extractFaces, extractFaceTensors, TNetInput } from '../dom/index';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { isWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\n\nexport async function extractAllFacesAndComputeResults<TSource extends WithFaceDetection<{}>, TResult>(\n  parentResults: TSource[],\n  input: TNetInput,\n  // eslint-disable-next-line no-unused-vars\n  computeResults: (faces: Array<HTMLCanvasElement | tf.Tensor3D>) => Promise<TResult>,\n  extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D> | null,\n  // eslint-disable-next-line no-unused-vars\n  getRectForAlignment: (parentResult: WithFaceLandmarks<TSource, any>) => FaceDetection = ({ alignedRect }) => alignedRect,\n) {\n  const faceBoxes = parentResults.map((parentResult) => (isWithFaceLandmarks(parentResult)\n    ? getRectForAlignment(parentResult)\n    : parentResult.detection));\n  const faces: Array<HTMLCanvasElement | tf.Tensor3D> = extractedFaces || (\n    input instanceof tf.Tensor\n      ? await extractFaceTensors(input, faceBoxes)\n      : await extractFaces(input, faceBoxes)\n  );\n  const results = await computeResults(faces);\n  faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n  return results;\n}\n\nexport async function extractSingleFaceAndComputeResult<TSource extends WithFaceDetection<{}>, TResult>(\n  parentResult: TSource,\n  input: TNetInput,\n  // eslint-disable-next-line no-unused-vars\n  computeResult: (face: HTMLCanvasElement | tf.Tensor3D) => Promise<TResult>,\n  extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D> | null,\n  // eslint-disable-next-line no-unused-vars\n  getRectForAlignment?: (parentResultLocal: WithFaceLandmarks<TSource, any>) => FaceDetection,\n) {\n  return extractAllFacesAndComputeResults<TSource, TResult>(\n    [parentResult],\n    input,\n    async (faces) => computeResult(faces[0]),\n    extractedFaces,\n    getRectForAlignment,\n  );\n}\n", "import { Point } from '../classes/index';\n\nexport const IOU_THRESHOLD = 0.4;\n\nexport const BOX_ANCHORS = [\n  new Point(1.603231, 2.094468),\n  new Point(6.041143, 7.080126),\n  new Point(2.882459, 3.518061),\n  new Point(4.266906, 5.178857),\n  new Point(9.041765, 10.66308),\n];\n\nexport const MEAN_RGB: [number, number, number] = [117.001, 114.697, 97.404];\n", "import * as tf from '../../dist/tfjs.esm';\n\nimport { FaceDetection, Point } from '../classes/index';\nimport { ParamMapping } from '../common/index';\nimport { TNetInput } from '../dom/index';\nimport { ITinyYolov2Options } from '../tinyYolov2/index';\nimport { TinyYolov2Base } from '../tinyYolov2/TinyYolov2Base';\nimport { TinyYolov2NetParams } from '../tinyYolov2/types';\nimport { BOX_ANCHORS, IOU_THRESHOLD, MEAN_RGB } from './const';\n\nexport class TinyFaceDetector extends TinyYolov2Base {\n  constructor() {\n    const config = {\n      withSeparableConvs: true,\n      iouThreshold: IOU_THRESHOLD,\n      classes: ['face'],\n      anchors: BOX_ANCHORS,\n      meanRgb: MEAN_RGB,\n      isFirstLayerConv2d: true,\n      filterSizes: [3, 16, 32, 64, 128, 256, 512],\n    };\n\n    super(config);\n  }\n\n  public get anchors(): Point[] {\n    return this.config.anchors;\n  }\n\n  public async locateFaces(input: TNetInput, forwardParams: ITinyYolov2Options): Promise<FaceDetection[]> {\n    const objectDetections = await this.detect(input, forwardParams);\n    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));\n  }\n\n  protected override getDefaultModelName(): string {\n    return 'tiny_face_detector_model';\n  }\n\n  protected override extractParamsFromWeightMap(weightMap: tf.NamedTensorMap): { params: TinyYolov2NetParams, paramMappings: ParamMapping[] } {\n    return super.extractParamsFromWeightMap(weightMap);\n  }\n}\n", "import { AgeGenderNet } from '../ageGenderNet/AgeGenderNet';\nimport { AgeAndGenderPrediction } from '../ageGenderNet/types';\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { TNetInput } from '../dom/index';\nimport { FaceExpressionNet } from '../faceExpressionNet/FaceExpressionNet';\nimport { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\nimport { FaceLandmark68Net } from '../faceLandmarkNet/FaceLandmark68Net';\nimport { FaceLandmark68TinyNet } from '../faceLandmarkNet/FaceLandmark68TinyNet';\nimport { FaceRecognitionNet } from '../faceRecognitionNet/FaceRecognitionNet';\nimport { SsdMobilenetv1 } from '../ssdMobilenetv1/SsdMobilenetv1';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { TinyFaceDetector } from '../tinyFaceDetector/TinyFaceDetector';\nimport { TinyFaceDetectorOptions } from '../tinyFaceDetector/TinyFaceDetectorOptions';\nimport { ITinyYolov2Options, TinyYolov2 } from '../tinyYolov2/index';\n\nexport const nets = {\n  ssdMobilenetv1: new SsdMobilenetv1(),\n  tinyFaceDetector: new TinyFaceDetector(),\n  tinyYolov2: new TinyYolov2(),\n  faceLandmark68Net: new FaceLandmark68Net(),\n  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),\n  faceRecognitionNet: new FaceRecognitionNet(),\n  faceExpressionNet: new FaceExpressionNet(),\n  ageGenderNet: new AgeGenderNet(),\n};\n\n/**\n * Attempts to detect all faces in an image using SSD Mobilenetv1 Network.\n *\n * @param input The input image.\n * @param options (optional, default: see SsdMobilenetv1Options constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const ssdMobilenetv1 = (input: TNetInput, options: SsdMobilenetv1Options): Promise<FaceDetection[]> => nets.ssdMobilenetv1.locateFaces(input, options);\n\n/**\n * Attempts to detect all faces in an image using the Tiny Face Detector.\n *\n * @param input The input image.\n * @param options (optional, default: see TinyFaceDetectorOptions constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const tinyFaceDetector = (input: TNetInput, options: TinyFaceDetectorOptions): Promise<FaceDetection[]> => nets.tinyFaceDetector.locateFaces(input, options);\n\n/**\n * Attempts to detect all faces in an image using the Tiny Yolov2 Network.\n *\n * @param input The input image.\n * @param options (optional, default: see TinyYolov2Options constructor for default parameters).\n * @returns Bounding box of each face with score.\n */\nexport const tinyYolov2 = (input: TNetInput, options: ITinyYolov2Options): Promise<FaceDetection[]> => nets.tinyYolov2.locateFaces(input, options);\n\n/**\n * Detects the 68 point face landmark positions of the face shown in an image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns 68 point face landmarks or array thereof in case of batch input.\n */\nexport const detectFaceLandmarks = (input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> => nets.faceLandmark68Net.detectLandmarks(input);\n\n/**\n * Detects the 68 point face landmark positions of the face shown in an image\n * using a tinier version of the 68 point face landmark model, which is slightly\n * faster at inference, but also slightly less accurate.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns 68 point face landmarks or array thereof in case of batch input.\n */\nexport const detectFaceLandmarksTiny = (input: TNetInput): Promise<FaceLandmarks68 | FaceLandmarks68[]> => nets.faceLandmark68TinyNet.detectLandmarks(input);\n\n/**\n * Computes a 128 entry vector (face descriptor / face embeddings) from the face shown in an image,\n * which uniquely represents the features of that persons face. The computed face descriptor can\n * be used to measure the similarity between faces, by computing the euclidean distance of two\n * face descriptors.\n *\n * @param inputs The face image extracted from the aligned bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Face descriptor with 128 entries or array thereof in case of batch input.\n */\nexport const computeFaceDescriptor = (input: TNetInput): Promise<Float32Array | Float32Array[]> => nets.faceRecognitionNet.computeFaceDescriptor(input);\n\n/**\n * Recognizes the facial expressions from a face image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Facial expressions with corresponding probabilities or array thereof in case of batch input.\n */\nexport const recognizeFaceExpressions = (input: TNetInput): Promise<FaceExpressions | FaceExpressions[]> => nets.faceExpressionNet.predictExpressions(input);\n\n/**\n * Predicts age and gender from a face image.\n *\n * @param inputs The face image extracted from the bounding box of a face. Can\n * also be an array of input images, which will be batch processed.\n * @returns Predictions with age, gender and gender probability or array thereof in case of batch input.\n */\nexport const predictAgeAndGender = (input: TNetInput): Promise<AgeAndGenderPrediction | AgeAndGenderPrediction[]> => nets.ageGenderNet.predictAgeAndGender(input);\n\nexport const loadSsdMobilenetv1Model = (url: string) => nets.ssdMobilenetv1.load(url);\nexport const loadTinyFaceDetectorModel = (url: string) => nets.tinyFaceDetector.load(url);\nexport const loadTinyYolov2Model = (url: string) => nets.tinyYolov2.load(url);\nexport const loadFaceLandmarkModel = (url: string) => nets.faceLandmark68Net.load(url);\nexport const loadFaceLandmarkTinyModel = (url: string) => nets.faceLandmark68TinyNet.load(url);\nexport const loadFaceRecognitionModel = (url: string) => nets.faceRecognitionNet.load(url);\nexport const loadFaceExpressionModel = (url: string) => nets.faceExpressionNet.load(url);\nexport const loadAgeGenderModel = (url: string) => nets.ageGenderNet.load(url);\n\n// backward compatibility\nexport const loadFaceDetectionModel = loadSsdMobilenetv1Model;\nexport const locateFaces = ssdMobilenetv1;\nexport const detectLandmarks = detectFaceLandmarks;\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { TNetInput } from '../dom/index';\nimport { FaceExpressions } from '../faceExpressionNet/FaceExpressions';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { extendWithFaceExpressions, WithFaceExpressions } from '../factories/WithFaceExpressions';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderTask, PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\n\nexport class PredictFaceExpressionsTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D>,\n  ) {\n    super();\n  }\n}\n\nexport class PredictAllFaceExpressionsTask<TSource extends WithFaceDetection<{}>> extends PredictFaceExpressionsTaskBase<WithFaceExpressions<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceExpressions<TSource>[]> {\n    const parentResults = await this.parentTask;\n\n    const faceExpressionsByFace = await extractAllFacesAndComputeResults<TSource, FaceExpressions[]>(\n      parentResults,\n      this.input,\n      async (faces) => Promise.all(\n        faces.map((face) => nets.faceExpressionNet.predictExpressions(face) as Promise<FaceExpressions>),\n      ),\n      this.extractedFaces,\n    );\n\n    return parentResults.map(\n      (parentResult, i) => extendWithFaceExpressions<TSource>(parentResult, faceExpressionsByFace[i]),\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderTask(this, this.input);\n  }\n}\n\nexport class PredictSingleFaceExpressionsTask<TSource extends WithFaceDetection<{}>> extends PredictFaceExpressionsTaskBase<WithFaceExpressions<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceExpressions<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) {\n      return undefined;\n    }\n\n    const faceExpressions = await extractSingleFaceAndComputeResult<TSource, FaceExpressions>(\n      parentResult,\n      this.input,\n      (face) => nets.faceExpressionNet.predictExpressions(face) as Promise<FaceExpressions>,\n      this.extractedFaces,\n    );\n\n    return extendWithFaceExpressions(parentResult, faceExpressions);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderTask(this, this.input);\n  }\n}\n\nexport class PredictAllFaceExpressionsWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictAllFaceExpressionsTask<TSource> {\n  override withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleFaceExpressionsWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictSingleFaceExpressionsTask<TSource> {\n  override withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { AgeAndGenderPrediction } from '../ageGenderNet/types';\nimport { TNetInput } from '../dom/index';\nimport { extendWithAge, WithAge } from '../factories/WithAge';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { extendWithGender, WithGender } from '../factories/WithGender';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllFaceExpressionsTask, PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class PredictAgeAndGenderTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected extractedFaces?: Array<HTMLCanvasElement | tf.Tensor3D>,\n  ) {\n    super();\n  }\n}\n\nexport class PredictAllAgeAndGenderTask<TSource extends WithFaceDetection<{}>> extends PredictAgeAndGenderTaskBase<WithAge<WithGender<TSource>>[], TSource[]> {\n  public override async run(): Promise<WithAge<WithGender<TSource>>[]> {\n    const parentResults = await this.parentTask;\n    const ageAndGenderByFace = await extractAllFacesAndComputeResults<TSource, AgeAndGenderPrediction[]>(\n      parentResults,\n      this.input,\n      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face) as Promise<AgeAndGenderPrediction>)),\n      this.extractedFaces,\n    );\n    return parentResults.map((parentResult, i) => {\n      const { age, gender, genderProbability } = ageAndGenderByFace[i];\n      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);\n    });\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleAgeAndGenderTask<TSource extends WithFaceDetection<{}>> extends PredictAgeAndGenderTaskBase<WithAge<WithGender<TSource>> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithAge<WithGender<TSource>> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) return undefined;\n    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult<TSource, AgeAndGenderPrediction>(\n      parentResult,\n      this.input,\n      (face) => nets.ageGenderNet.predictAgeAndGender(face) as Promise<AgeAndGenderPrediction>,\n      this.extractedFaces,\n    );\n    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsTask(this, this.input);\n  }\n}\n\nexport class PredictAllAgeAndGenderWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictAllAgeAndGenderTask<TSource> {\n  override withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class PredictSingleAgeAndGenderWithFaceAlignmentTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends PredictSingleAgeAndGenderTask<TSource> {\n  override withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { TNetInput } from '../dom/index';\nimport { extendWithFaceDescriptor, WithFaceDescriptor } from '../factories/WithFaceDescriptor';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { extractAllFacesAndComputeResults, extractSingleFaceAndComputeResult } from './extractFacesAndComputeResults';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class ComputeFaceDescriptorsTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n  ) {\n    super();\n  }\n}\n\nexport class ComputeAllFaceDescriptorsTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends ComputeFaceDescriptorsTaskBase<WithFaceDescriptor<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceDescriptor<TSource>[]> {\n    const parentResults = await this.parentTask;\n    const descriptors = await extractAllFacesAndComputeResults<TSource, Float32Array[]>(\n      parentResults,\n      this.input,\n      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face) as Promise<Float32Array>)),\n      null,\n      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true }),\n    );\n    return descriptors.map((descriptor, i) => extendWithFaceDescriptor<TSource>(parentResults[i], descriptor));\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n}\n\nexport class ComputeSingleFaceDescriptorTask<TSource extends WithFaceLandmarks<WithFaceDetection<{}>>> extends ComputeFaceDescriptorsTaskBase<WithFaceDescriptor<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceDescriptor<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) return undefined;\n    const descriptor = await extractSingleFaceAndComputeResult<TSource, Float32Array>(\n      parentResult,\n      this.input,\n      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face) as Promise<Float32Array>,\n      null,\n      // eslint-disable-next-line no-shadow, @typescript-eslint/no-shadow\n      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true }),\n    );\n    return extendWithFaceDescriptor(parentResult, descriptor);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport * as tf from '../../dist/tfjs.esm';\n\nimport { FaceLandmarks68 } from '../classes/FaceLandmarks68';\nimport { extractFaces, extractFaceTensors, TNetInput } from '../dom/index';\nimport { FaceLandmark68Net } from '../faceLandmarkNet/FaceLandmark68Net';\nimport { FaceLandmark68TinyNet } from '../faceLandmarkNet/FaceLandmark68TinyNet';\nimport { WithFaceDetection } from '../factories/WithFaceDetection';\nimport { extendWithFaceLandmarks, WithFaceLandmarks } from '../factories/WithFaceLandmarks';\nimport { ComposableTask } from './ComposableTask';\nimport { ComputeAllFaceDescriptorsTask, ComputeSingleFaceDescriptorTask } from './ComputeFaceDescriptorsTasks';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderWithFaceAlignmentTask, PredictSingleAgeAndGenderWithFaceAlignmentTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsWithFaceAlignmentTask, PredictSingleFaceExpressionsWithFaceAlignmentTask } from './PredictFaceExpressionsTask';\n\nexport class DetectFaceLandmarksTaskBase<TReturn, TParentReturn> extends ComposableTask<TReturn> {\n  constructor(\n    // eslint-disable-next-line no-unused-vars\n    protected parentTask: ComposableTask<TParentReturn> | Promise<TParentReturn>,\n    // eslint-disable-next-line no-unused-vars\n    protected input: TNetInput,\n    // eslint-disable-next-line no-unused-vars\n    protected useTinyLandmarkNet: boolean,\n  ) {\n    super();\n  }\n\n  protected get landmarkNet(): FaceLandmark68Net | FaceLandmark68TinyNet {\n    return this.useTinyLandmarkNet\n      ? nets.faceLandmark68TinyNet\n      : nets.faceLandmark68Net;\n  }\n}\n\nexport class DetectAllFaceLandmarksTask<TSource extends WithFaceDetection<{}>> extends DetectFaceLandmarksTaskBase<WithFaceLandmarks<TSource>[], TSource[]> {\n  public override async run(): Promise<WithFaceLandmarks<TSource>[]> {\n    const parentResults = await this.parentTask;\n    const detections = parentResults.map((res) => res.detection);\n    const faces: Array<HTMLCanvasElement | tf.Tensor3D> = this.input instanceof tf.Tensor\n      ? await extractFaceTensors(this.input, detections)\n      : await extractFaces(this.input, detections);\n    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face))) as FaceLandmarks68[];\n    faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n    const result = parentResults\n      .filter((_parentResult, i) => faceLandmarksByFace[i])\n      .map((parentResult, i) => extendWithFaceLandmarks<TSource>(parentResult, faceLandmarksByFace[i]));\n    return result;\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptors() {\n    return new ComputeAllFaceDescriptorsTask(this, this.input);\n  }\n}\n\nexport class DetectSingleFaceLandmarksTask<TSource extends WithFaceDetection<{}>> extends DetectFaceLandmarksTaskBase<WithFaceLandmarks<TSource> | undefined, TSource | undefined> {\n  public override async run(): Promise<WithFaceLandmarks<TSource> | undefined> {\n    const parentResult = await this.parentTask;\n    if (!parentResult) {\n      return undefined;\n    }\n    const { detection } = parentResult;\n    const faces: Array<HTMLCanvasElement | tf.Tensor3D> = this.input instanceof tf.Tensor\n      ? await extractFaceTensors(this.input, [detection])\n      : await extractFaces(this.input, [detection]);\n    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]) as FaceLandmarks68;\n    faces.forEach((f) => f instanceof tf.Tensor && f.dispose());\n    return extendWithFaceLandmarks<TSource>(parentResult, landmarks);\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);\n  }\n\n  withFaceDescriptor() {\n    return new ComputeSingleFaceDescriptorTask(this, this.input);\n  }\n}\n", "/* eslint-disable max-classes-per-file */\nimport { FaceDetection } from '../classes/FaceDetection';\nimport { TNetInput } from '../dom/index';\nimport { extendWithFaceDetection, WithFaceDetection } from '../factories/WithFaceDetection';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { TinyFaceDetectorOptions } from '../tinyFaceDetector/TinyFaceDetectorOptions';\nimport { TinyYolov2Options } from '../tinyYolov2/index';\nimport { ComposableTask } from './ComposableTask';\nimport { DetectAllFaceLandmarksTask, DetectSingleFaceLandmarksTask } from './DetectFaceLandmarksTasks';\nimport { nets } from './nets';\nimport { PredictAllAgeAndGenderTask, PredictSingleAgeAndGenderTask } from './PredictAgeAndGenderTask';\nimport { PredictAllFaceExpressionsTask, PredictSingleFaceExpressionsTask } from './PredictFaceExpressionsTask';\nimport { FaceDetectionOptions } from './types';\n\nexport class DetectFacesTaskBase<TReturn> extends ComposableTask<TReturn> {\n  // eslint-disable-next-line no-unused-vars\n  constructor(protected input: TNetInput, protected options: FaceDetectionOptions = new SsdMobilenetv1Options()) {\n    super();\n  }\n}\n\nexport class DetectAllFacesTask extends DetectFacesTaskBase<FaceDetection[]> {\n  public override async run(): Promise<FaceDetection[]> {\n    const { input, options } = this;\n    let result;\n    if (options instanceof TinyFaceDetectorOptions) result = nets.tinyFaceDetector.locateFaces(input, options);\n    else if (options instanceof SsdMobilenetv1Options) result = nets.ssdMobilenetv1.locateFaces(input, options);\n    else if (options instanceof TinyYolov2Options) result = nets.tinyYolov2.locateFaces(input, options);\n    else throw new Error('detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options');\n    return result;\n  }\n\n  private runAndExtendWithFaceDetections(): Promise<WithFaceDetection<{}>[]> {\n    return new Promise<WithFaceDetection<{}>[]>((resolve, reject) => {\n      this.run()\n        .then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection))))\n        .catch((err) => reject(err));\n    });\n  }\n\n  withFaceLandmarks(useTinyLandmarkNet = false) {\n    return new DetectAllFaceLandmarksTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n      useTinyLandmarkNet,\n    );\n  }\n\n  withFaceExpressions() {\n    return new PredictAllFaceExpressionsTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictAllAgeAndGenderTask(\n      this.runAndExtendWithFaceDetections(),\n      this.input,\n    );\n  }\n}\n\nexport class DetectSingleFaceTask extends DetectFacesTaskBase<FaceDetection | undefined> {\n  public override async run(): Promise<FaceDetection | undefined> {\n    const faceDetections = await new DetectAllFacesTask(this.input, this.options);\n    let faceDetectionWithHighestScore = faceDetections[0];\n    faceDetections.forEach((faceDetection) => {\n      if (faceDetection.score > faceDetectionWithHighestScore.score) faceDetectionWithHighestScore = faceDetection;\n    });\n    return faceDetectionWithHighestScore;\n  }\n\n  private runAndExtendWithFaceDetection(): Promise<WithFaceDetection<{}> | undefined> {\n    // eslint-disable-next-line no-async-promise-executor\n    return new Promise<WithFaceDetection<{}> | undefined>(async (resolve) => {\n      const detection = await this.run();\n      resolve(detection ? extendWithFaceDetection<{}>({}, detection) : undefined);\n    });\n  }\n\n  withFaceLandmarks(useTinyLandmarkNet = false) {\n    return new DetectSingleFaceLandmarksTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n      useTinyLandmarkNet,\n    );\n  }\n\n  withFaceExpressions() {\n    return new PredictSingleFaceExpressionsTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n    );\n  }\n\n  withAgeAndGender() {\n    return new PredictSingleAgeAndGenderTask(\n      this.runAndExtendWithFaceDetection(),\n      this.input,\n    );\n  }\n}\n", "import { TNetInput } from '../dom/index';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/SsdMobilenetv1Options';\nimport { DetectAllFacesTask, DetectSingleFaceTask } from './DetectFacesTasks';\nimport { FaceDetectionOptions } from './types';\n\nexport function detectSingleFace(input: TNetInput, options: FaceDetectionOptions = new SsdMobilenetv1Options()): DetectSingleFaceTask {\n  return new DetectSingleFaceTask(input, options);\n}\n\nexport function detectAllFaces(input: TNetInput, options: FaceDetectionOptions = new SsdMobilenetv1Options()): DetectAllFacesTask {\n  return new DetectAllFacesTask(input, options);\n}\n", "import { TNetInput } from '../dom/index';\nimport { WithFaceDescriptor, WithFaceDetection, WithFaceLandmarks } from '../factories/index';\nimport { SsdMobilenetv1Options } from '../ssdMobilenetv1/index';\nimport { ITinyYolov2Options, TinyYolov2Options } from '../tinyYolov2/index';\nimport { detectAllFaces } from './detectFaces';\n\nexport async function allFacesSsdMobilenetv1(input: TNetInput, minConfidence?: number): Promise<WithFaceDescriptor<WithFaceLandmarks<WithFaceDetection<{}>>>[]> {\n  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {}))\n    .withFaceLandmarks()\n    .withFaceDescriptors();\n}\n\nexport async function allFacesTinyYolov2(input: TNetInput, forwardParams: ITinyYolov2Options = {}): Promise<WithFaceDescriptor<WithFaceLandmarks<WithFaceDetection<{}>>>[]> {\n  return detectAllFaces(input, new TinyYolov2Options(forwardParams))\n    .withFaceLandmarks()\n    .withFaceDescriptors();\n}\n\nexport const allFaces = allFacesSsdMobilenetv1;\n", "export function euclideanDistance(arr1: number[] | Float32Array, arr2: number[] | Float32Array) {\n  if (arr1.length !== arr2.length) throw new Error('euclideanDistance: arr1.length !== arr2.length');\n  const desc1 = Array.from(arr1);\n  const desc2 = Array.from(arr2);\n  return Math.sqrt(\n    desc1\n      .map((val, i) => val - desc2[i])\n      .reduce((res, diff) => res + (diff * diff), 0),\n  );\n}\n", "import { FaceMatch } from '../classes/FaceMatch';\nimport { LabeledFaceDescriptors } from '../classes/LabeledFaceDescriptors';\nimport { euclideanDistance } from '../euclideanDistance';\nimport { WithFaceDescriptor } from '../factories/index';\n\nexport class FaceMatcher {\n  private _labeledDescriptors: LabeledFaceDescriptors[];\n  private _distanceThreshold: number;\n\n  constructor(inputs: LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>, distanceThreshold = 0.6) {\n    this._distanceThreshold = distanceThreshold;\n    const inputArray = Array.isArray(inputs) ? inputs : [inputs];\n    if (!inputArray.length) throw new Error('FaceRecognizer.constructor - expected atleast one input');\n    let count = 1;\n    const createUniqueLabel = () => `person ${count++}`;\n    this._labeledDescriptors = inputArray.map((desc) => {\n      if (desc instanceof LabeledFaceDescriptors) return desc;\n      if (desc instanceof Float32Array) return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);\n      if (desc.descriptor && desc.descriptor instanceof Float32Array) return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);\n      throw new Error('FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>');\n    });\n  }\n\n  public get labeledDescriptors(): LabeledFaceDescriptors[] { return this._labeledDescriptors; }\n\n  public get distanceThreshold(): number { return this._distanceThreshold; }\n\n  public computeMeanDistance(queryDescriptor: Float32Array, descriptors: Float32Array[]): number {\n    return descriptors\n      .map((d) => euclideanDistance(d, queryDescriptor))\n      .reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);\n  }\n\n  public matchDescriptor(queryDescriptor: Float32Array): FaceMatch {\n    return this.labeledDescriptors\n      .map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors)))\n      .reduce((best, curr) => (best.distance < curr.distance ? best : curr));\n  }\n\n  public findBestMatch(queryDescriptor: Float32Array): FaceMatch {\n    const bestMatch = this.matchDescriptor(queryDescriptor);\n    return (bestMatch.distance < this._distanceThreshold) ? bestMatch : new FaceMatch('unknown', bestMatch.distance);\n  }\n\n  public toJSON(): any {\n    return {\n      distanceThreshold: this._distanceThreshold,\n      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON()),\n    };\n  }\n\n  public static fromJSON(json: any): FaceMatcher {\n    const labeledDescriptors = json.labeledDescriptors.map((ld: any) => LabeledFaceDescriptors.fromJSON(ld));\n    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);\n  }\n}\n", "import { TinyFaceDetector } from './TinyFaceDetector';\n\nexport * from './TinyFaceDetector';\nexport * from './TinyFaceDetectorOptions';\n\nexport function createTinyFaceDetector(weights: Float32Array) {\n  const net = new TinyFaceDetector();\n  net.extractWeights(weights);\n  return net;\n}\n", "import { Dimensions, IDimensions } from './classes/index';\nimport { FaceDetection } from './classes/FaceDetection';\nimport { FaceLandmarks } from './classes/FaceLandmarks';\nimport { extendWithFaceDetection, isWithFaceDetection } from './factories/WithFaceDetection';\nimport { extendWithFaceLandmarks, isWithFaceLandmarks } from './factories/WithFaceLandmarks';\n\nexport function resizeResults<T>(results: T, dimensions: IDimensions): T {\n  const { width, height } = new Dimensions(dimensions.width, dimensions.height);\n\n  if (width <= 0 || height <= 0) {\n    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);\n  }\n\n  if (Array.isArray(results)) {\n    // return results.map(obj => resizeResults(obj, { width, height })) as any as T\n    return (results as Array<any>).map((obj) => resizeResults(obj, { width, height } as IDimensions)) as any as T;\n  }\n\n  if (isWithFaceLandmarks(results)) {\n    const resizedDetection = results.detection.forSize(width, height);\n    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);\n    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);\n  }\n\n  if (isWithFaceDetection(results)) {\n    return extendWithFaceDetection(results, results.detection.forSize(width, height));\n  }\n\n  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {\n    return (results as any).forSize(width, height);\n  }\n\n  return results;\n}\n", "import * as tf from '../dist/tfjs.esm';\nimport * as draw from './draw/index';\nimport * as utils from './utils/index';\nimport * as pkg from '../package.json';\n\nexport { tf, draw, utils };\n\nexport * from './ageGenderNet/index';\nexport * from './classes/index';\nexport * from './dom/index';\nexport * from './env/index';\nexport * from './faceExpressionNet/index';\nexport * from './faceLandmarkNet/index';\nexport * from './faceRecognitionNet/index';\nexport * from './factories/index';\nexport * from './globalApi/index';\nexport * from './ops/index';\nexport * from './ssdMobilenetv1/index';\nexport * from './tinyFaceDetector/index';\nexport * from './tinyYolov2/index';\nexport * from './euclideanDistance';\nexport * from './NeuralNetwork';\nexport * from './resizeResults';\n\nexport const version = pkg.version as string;\n\n// set webgl defaults\n// if (browser) tf.ENV.set('WEBGL_USE_SHAPES_UNIFORMS', true);\n"],
+  "mappings": ";;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAMA,IAAI,WAAW,OAAO;AACtB,IAAIA,aAAY,OAAO;AACvB,IAAI,mBAAmB,OAAO;AAC9B,IAAI,oBAAoB,OAAO;AAC/B,IAAI,eAAe,OAAO;AAC1B,IAAI,eAAe,OAAO,UAAU;AACpC,IAAIC,cAA6B,CAAC,MAAM,OAAO,cAAY,cAAc,YAAU,OAAO,UAAU,cAAc,IAAI,MAAM,GAAG;AAAA,EAC7H,KAAK,CAAC,GAAG,OAAO,OAAO,cAAY,cAAc,YAAU,GAAG;AAChE,CAAC,IAAI,GAAG,SAAS,GAAG;AAClB,MAAI,OAAO,cAAY;AACrB,WAAO,UAAQ,MAAM,MAAM,SAAS;AACtC,QAAM,IAAI,MAAM,yBAAyB,IAAI,oBAAoB;AACnE,CAAC;AACD,IAAI,aAAa,CAAC,IAAI,SAAS,SAASC,cAAa;AACnD,SAAO,SAAS,GAAG,GAAG,kBAAkB,EAAE,EAAE,MAAM,OAAO,EAAE,SAAS,CAAC,EAAE,GAAG,SAAS,IAAI,GAAG,KAAK;AACjG;AACA,IAAIC,YAAW,CAAC,QAAQ,SAAS;AAC/B,WAAS,QAAQ;AACf,IAAAH,WAAU,QAAQ,MAAM,EAAE,KAAK,KAAK,OAAO,YAAY,KAAK,CAAC;AACjE;AACA,IAAI,cAAc,CAAC,IAAI,MAAM,QAAQ,SAAS;AAC5C,MAAI,QAAQ,OAAO,SAAS,YAAY,OAAO,SAAS,YAAY;AAClE,aAAS,OAAO,kBAAkB,IAAI;AACpC,UAAI,CAAC,aAAa,KAAK,IAAI,GAAG,KAAK,QAAQ;AACzC,QAAAA,WAAU,IAAI,KAAK,EAAE,KAAK,MAAM,KAAK,MAAM,YAAY,EAAE,OAAO,iBAAiB,MAAM,GAAG,MAAM,KAAK,WAAW,CAAC;AAAA,EACvH;AACA,SAAO;AACT;AACA,IAAI,UAAU,CAAC,MAAM,YAAY,YAAY,SAAS,QAAQ,OAAO,SAAS,aAAa,IAAI,CAAC,IAAI,CAAC,GAAG;AAAA,EACtG,cAAc,CAAC,QAAQ,CAAC,KAAK,aAAaA,WAAU,QAAQ,WAAW,EAAE,OAAO,MAAM,YAAY,KAAK,CAAC,IAAI;AAAA,EAC5G;AACF;AAGA,IAAI,eAAe,WAAW;AAAA,EAC5B,gCAAgC,SAAS,QAAQ;AAC/C,WAAO,UAAU;AACjB,QAAI,OAAO;AACX,QAAI;AACF,aAAO,IAAI,YAAY,SAAS,IAAI,YAAY,OAAO,IAAI,WAAW;AAAA,QACpE;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC,CAAC,GAAG,CAAC,CAAC,EAAE;AAAA,IACX,SAAS,GAAP;AAAA,IACF;AACA,aAAS,MAAM,KAAK,MAAM,UAAU;AAClC,WAAK,MAAM,MAAM;AACjB,WAAK,OAAO,OAAO;AACnB,WAAK,WAAW,CAAC,CAAC;AAAA,IACpB;AACA,UAAM,UAAU;AAChB,WAAO,eAAe,MAAM,WAAW,cAAc,EAAE,OAAO,KAAK,CAAC;AACpE,aAAS,OAAO,KAAK;AACnB,cAAQ,OAAO,IAAI,mBAAmB;AAAA,IACxC;AACA,UAAM,SAAS;AACf,QAAI,YAAY,CAAC;AACjB,QAAI,aAAa,CAAC;AAClB,aAAS,QAAQ,OAAO,UAAU;AAChC,UAAI,KAAK,WAAW;AACpB,UAAI,UAAU;AACZ,mBAAW;AACX,YAAI,QAAQ,KAAK,SAAS,QAAQ,KAAK;AACrC,sBAAY,WAAW;AACvB,cAAI;AACF,mBAAO;AAAA,QACX;AACA,cAAM,SAAS,QAAQ,QAAQ,KAAK,IAAI,KAAK,GAAG,IAAI;AACpD,YAAI;AACF,qBAAW,SAAS;AACtB,eAAO;AAAA,MACT,OAAO;AACL,iBAAS;AACT,YAAI,QAAQ,QAAQ,SAAS,QAAQ,KAAK;AACxC,sBAAY,UAAU;AACtB,cAAI;AACF,mBAAO;AAAA,QACX;AACA,cAAM,SAAS,OAAO,QAAQ,IAAI,KAAK,GAAG,KAAK;AAC/C,YAAI;AACF,oBAAU,SAAS;AACrB,eAAO;AAAA,MACT;AAAA,IACF;AACA,UAAM,UAAU;AAChB,aAAS,WAAW,OAAO,UAAU;AACnC,UAAI,MAAM,KAAK;AACb,eAAO,WAAW,QAAQ;AAC5B,UAAI,UAAU;AACZ,YAAI,QAAQ;AACV,iBAAO;AACT,YAAI,SAAS;AACX,iBAAO;AAAA,MACX,OAAO;AACL,YAAI,SAAS,CAAC;AACZ,iBAAO;AACT,YAAI,QAAQ,KAAK;AACf,iBAAO;AAAA,MACX;AACA,UAAI,QAAQ;AACV,eAAO,WAAW,CAAC,OAAO,QAAQ,EAAE,IAAI;AAC1C,aAAO,SAAS,QAAQ,iBAAiB,GAAG,QAAQ,iBAAiB,GAAG,QAAQ;AAAA,IAClF;AACA,UAAM,aAAa;AACnB,aAAS,SAAS,SAAS,UAAU,UAAU;AAC7C,aAAO,IAAI,MAAM,SAAS,UAAU,QAAQ;AAAA,IAC9C;AACA,UAAM,WAAW;AACjB,QAAI,UAAU,KAAK;AACnB,aAAS,WAAW,KAAK,UAAU,OAAO;AACxC,UAAI,IAAI,WAAW;AACjB,cAAM,MAAM,cAAc;AAC5B,UAAI,QAAQ,SAAS,QAAQ,cAAc,QAAQ,eAAe,QAAQ;AACxE,eAAO;AACT,UAAI,OAAO,aAAa,UAAU;AAChC,gBAAQ,UAAU,WAAW;AAAA,MAC/B,OAAO;AACL,mBAAW,CAAC,CAAC;AAAA,MACf;AACA,cAAQ,SAAS;AACjB,UAAI,QAAQ,KAAK,KAAK;AACpB,cAAM,WAAW,OAAO;AAC1B,UAAI;AACJ,WAAK,KAAK,IAAI,QAAQ,GAAG,KAAK;AAC5B,cAAM,MAAM,iBAAiB;AAAA,eACtB,OAAO,GAAG;AACjB,eAAO,WAAW,IAAI,UAAU,CAAC,GAAG,UAAU,KAAK,EAAE,IAAI;AAAA,MAC3D;AACA,UAAI,eAAe,WAAW,QAAQ,OAAO,CAAC,CAAC;AAC/C,UAAI,SAAS;AACb,eAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK,GAAG;AACtC,YAAI,OAAO,KAAK,IAAI,GAAG,IAAI,SAAS,CAAC,GAAG,QAAQ,SAAS,IAAI,UAAU,GAAG,IAAI,IAAI,GAAG,KAAK;AAC1F,YAAI,OAAO,GAAG;AACZ,cAAI,QAAQ,WAAW,QAAQ,OAAO,IAAI,CAAC;AAC3C,mBAAS,OAAO,IAAI,KAAK,EAAE,IAAI,WAAW,KAAK,CAAC;AAAA,QAClD,OAAO;AACL,mBAAS,OAAO,IAAI,YAAY;AAChC,mBAAS,OAAO,IAAI,WAAW,KAAK,CAAC;AAAA,QACvC;AAAA,MACF;AACA,aAAO,WAAW;AAClB,aAAO;AAAA,IACT;AACA,UAAM,aAAa;AACnB,aAAS,UAAU,KAAK,UAAU;AAChC,UAAI,OAAO,QAAQ;AACjB,eAAO,WAAW,KAAK,QAAQ;AACjC,UAAI,OAAO,QAAQ;AACjB,eAAO,WAAW,KAAK,QAAQ;AACjC,aAAO,SAAS,IAAI,KAAK,IAAI,MAAM,OAAO,aAAa,YAAY,WAAW,IAAI,QAAQ;AAAA,IAC5F;AACA,UAAM,YAAY;AAClB,QAAI,iBAAiB,KAAK;AAC1B,QAAI,iBAAiB,KAAK;AAC1B,QAAI,iBAAiB,iBAAiB;AACtC,QAAI,iBAAiB,iBAAiB;AACtC,QAAI,iBAAiB,iBAAiB;AACtC,QAAI,aAAa,QAAQ,cAAc;AACvC,QAAI,OAAO,QAAQ,CAAC;AACpB,UAAM,OAAO;AACb,QAAI,QAAQ,QAAQ,GAAG,IAAI;AAC3B,UAAM,QAAQ;AACd,QAAI,MAAM,QAAQ,CAAC;AACnB,UAAM,MAAM;AACZ,QAAI,OAAO,QAAQ,GAAG,IAAI;AAC1B,UAAM,OAAO;AACb,QAAI,UAAU,QAAQ,EAAE;AACxB,UAAM,UAAU;AAChB,QAAI,YAAY,SAAS,aAAa,GAAG,aAAa,GAAG,KAAK;AAC9D,UAAM,YAAY;AAClB,QAAI,qBAAqB,SAAS,aAAa,GAAG,aAAa,GAAG,IAAI;AACtE,UAAM,qBAAqB;AAC3B,QAAI,YAAY,SAAS,GAAG,aAAa,GAAG,KAAK;AACjD,UAAM,YAAY;AAClB,QAAI,gBAAgB,MAAM;AAC1B,kBAAc,QAAQ,SAAS,QAAQ;AACrC,aAAO,KAAK,WAAW,KAAK,QAAQ,IAAI,KAAK;AAAA,IAC/C;AACA,kBAAc,WAAW,SAAS,WAAW;AAC3C,UAAI,KAAK;AACP,gBAAQ,KAAK,SAAS,KAAK,kBAAkB,KAAK,QAAQ;AAC5D,aAAO,KAAK,OAAO,kBAAkB,KAAK,QAAQ;AAAA,IACpD;AACA,kBAAc,WAAW,SAAS,SAAS,OAAO;AAChD,cAAQ,SAAS;AACjB,UAAI,QAAQ,KAAK,KAAK;AACpB,cAAM,WAAW,OAAO;AAC1B,UAAI,KAAK,OAAO;AACd,eAAO;AACT,UAAI,KAAK,WAAW,GAAG;AACrB,YAAI,KAAK,GAAG,SAAS,GAAG;AACtB,cAAI,YAAY,WAAW,KAAK,GAAG,OAAO,KAAK,IAAI,SAAS,GAAG,OAAO,KAAK,IAAI,SAAS,EAAE,IAAI,IAAI;AAClG,iBAAO,KAAK,SAAS,KAAK,IAAI,KAAK,MAAM,EAAE,SAAS,KAAK;AAAA,QAC3D;AACE,iBAAO,MAAM,KAAK,IAAI,EAAE,SAAS,KAAK;AAAA,MAC1C;AACA,UAAI,eAAe,WAAW,QAAQ,OAAO,CAAC,GAAG,KAAK,QAAQ,GAAG,MAAM;AACvE,UAAI,SAAS;AACb,aAAO,MAAM;AACX,YAAI,SAAS,IAAI,IAAI,YAAY,GAAG,SAAS,IAAI,IAAI,OAAO,IAAI,YAAY,CAAC,EAAE,MAAM,MAAM,GAAG,SAAS,OAAO,SAAS,KAAK;AAC5H,cAAM;AACN,YAAI,IAAI,OAAO;AACb,iBAAO,SAAS;AAAA,aACb;AACH,iBAAO,OAAO,SAAS;AACrB,qBAAS,MAAM;AACjB,mBAAS,KAAK,SAAS;AAAA,QACzB;AAAA,MACF;AAAA,IACF;AACA,kBAAc,cAAc,SAAS,cAAc;AACjD,aAAO,KAAK;AAAA,IACd;AACA,kBAAc,sBAAsB,SAAS,sBAAsB;AACjE,aAAO,KAAK,SAAS;AAAA,IACvB;AACA,kBAAc,aAAa,SAAS,aAAa;AAC/C,aAAO,KAAK;AAAA,IACd;AACA,kBAAc,qBAAqB,SAAS,qBAAqB;AAC/D,aAAO,KAAK,QAAQ;AAAA,IACtB;AACA,kBAAc,gBAAgB,SAAS,gBAAgB;AACrD,UAAI,KAAK,WAAW;AAClB,eAAO,KAAK,GAAG,SAAS,IAAI,KAAK,KAAK,IAAI,EAAE,cAAc;AAC5D,UAAI,MAAM,KAAK,QAAQ,IAAI,KAAK,OAAO,KAAK;AAC5C,eAAS,MAAM,IAAI,MAAM,GAAG;AAC1B,aAAK,MAAM,KAAK,QAAQ;AACtB;AACJ,aAAO,KAAK,QAAQ,IAAI,MAAM,KAAK,MAAM;AAAA,IAC3C;AACA,kBAAc,SAAS,SAAS,SAAS;AACvC,aAAO,KAAK,SAAS,KAAK,KAAK,QAAQ;AAAA,IACzC;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,aAAa,SAAS,aAAa;AAC/C,aAAO,CAAC,KAAK,YAAY,KAAK,OAAO;AAAA,IACvC;AACA,kBAAc,aAAa,SAAS,aAAa;AAC/C,aAAO,KAAK,YAAY,KAAK,QAAQ;AAAA,IACvC;AACA,kBAAc,QAAQ,SAAS,QAAQ;AACrC,cAAQ,KAAK,MAAM,OAAO;AAAA,IAC5B;AACA,kBAAc,SAAS,SAASI,WAAU;AACxC,cAAQ,KAAK,MAAM,OAAO;AAAA,IAC5B;AACA,kBAAc,SAAS,SAAS,OAAO,OAAO;AAC5C,UAAI,CAAC,OAAO,KAAK;AACf,gBAAQ,UAAU,KAAK;AACzB,UAAI,KAAK,aAAa,MAAM,YAAY,KAAK,SAAS,OAAO,KAAK,MAAM,SAAS,OAAO;AACtF,eAAO;AACT,aAAO,KAAK,SAAS,MAAM,QAAQ,KAAK,QAAQ,MAAM;AAAA,IACxD;AACA,kBAAc,KAAK,cAAc;AACjC,kBAAc,YAAY,SAAS,UAAU,OAAO;AAClD,aAAO,CAAC,KAAK,GAAG,KAAK;AAAA,IACvB;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,KAAK,cAAc;AACjC,kBAAc,WAAW,SAAS,SAAS,OAAO;AAChD,aAAO,KAAK,KAAK,KAAK,IAAI;AAAA,IAC5B;AACA,kBAAc,KAAK,cAAc;AACjC,kBAAc,kBAAkB,SAAS,gBAAgB,OAAO;AAC9D,aAAO,KAAK,KAAK,KAAK,KAAK;AAAA,IAC7B;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,KAAK,cAAc;AACjC,kBAAc,cAAc,SAAS,YAAY,OAAO;AACtD,aAAO,KAAK,KAAK,KAAK,IAAI;AAAA,IAC5B;AACA,kBAAc,KAAK,cAAc;AACjC,kBAAc,qBAAqB,SAAS,mBAAmB,OAAO;AACpE,aAAO,KAAK,KAAK,KAAK,KAAK;AAAA,IAC7B;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,KAAK,cAAc;AACjC,kBAAc,UAAU,SAAS,QAAQ,OAAO;AAC9C,UAAI,CAAC,OAAO,KAAK;AACf,gBAAQ,UAAU,KAAK;AACzB,UAAI,KAAK,GAAG,KAAK;AACf,eAAO;AACT,UAAI,UAAU,KAAK,WAAW,GAAG,WAAW,MAAM,WAAW;AAC7D,UAAI,WAAW,CAAC;AACd,eAAO;AACT,UAAI,CAAC,WAAW;AACd,eAAO;AACT,UAAI,CAAC,KAAK;AACR,eAAO,KAAK,IAAI,KAAK,EAAE,WAAW,IAAI,KAAK;AAC7C,aAAO,MAAM,SAAS,IAAI,KAAK,SAAS,KAAK,MAAM,SAAS,KAAK,QAAQ,MAAM,QAAQ,IAAI,KAAK,QAAQ,IAAI,KAAK;AAAA,IACnH;AACA,kBAAc,OAAO,cAAc;AACnC,kBAAc,SAAS,SAAS,SAAS;AACvC,UAAI,CAAC,KAAK,YAAY,KAAK,GAAG,SAAS;AACrC,eAAO;AACT,aAAO,KAAK,IAAI,EAAE,IAAI,GAAG;AAAA,IAC3B;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,MAAM,SAAS,KAAK,QAAQ;AACxC,UAAI,CAAC,OAAO,MAAM;AAChB,iBAAS,UAAU,MAAM;AAC3B,UAAI,MAAM,KAAK,SAAS;AACxB,UAAI,MAAM,KAAK,OAAO;AACtB,UAAI,MAAM,KAAK,QAAQ;AACvB,UAAI,MAAM,KAAK,MAAM;AACrB,UAAI,MAAM,OAAO,SAAS;AAC1B,UAAI,MAAM,OAAO,OAAO;AACxB,UAAI,MAAM,OAAO,QAAQ;AACzB,UAAI,MAAM,OAAO,MAAM;AACvB,UAAI,MAAM,GAAG,MAAM,GAAG,MAAM,GAAG,MAAM;AACrC,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO;AACP,aAAO,SAAS,OAAO,KAAK,KAAK,OAAO,KAAK,KAAK,KAAK,QAAQ;AAAA,IACjE;AACA,kBAAc,WAAW,SAAS,SAAS,YAAY;AACrD,UAAI,CAAC,OAAO,UAAU;AACpB,qBAAa,UAAU,UAAU;AACnC,aAAO,KAAK,IAAI,WAAW,IAAI,CAAC;AAAA,IAClC;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,WAAW,SAAS,UAAU,YAAY;AACtD,UAAI,KAAK,OAAO;AACd,eAAO;AACT,UAAI,CAAC,OAAO,UAAU;AACpB,qBAAa,UAAU,UAAU;AACnC,UAAI,MAAM;AACR,YAAI,MAAM,KAAK;AAAA,UACb,KAAK;AAAA,UACL,KAAK;AAAA,UACL,WAAW;AAAA,UACX,WAAW;AAAA,QACb;AACA,eAAO,SAAS,KAAK,KAAK,SAAS,GAAG,KAAK,QAAQ;AAAA,MACrD;AACA,UAAI,WAAW,OAAO;AACpB,eAAO;AACT,UAAI,KAAK,GAAG,SAAS;AACnB,eAAO,WAAW,MAAM,IAAI,YAAY;AAC1C,UAAI,WAAW,GAAG,SAAS;AACzB,eAAO,KAAK,MAAM,IAAI,YAAY;AACpC,UAAI,KAAK,WAAW,GAAG;AACrB,YAAI,WAAW,WAAW;AACxB,iBAAO,KAAK,IAAI,EAAE,IAAI,WAAW,IAAI,CAAC;AAAA;AAEtC,iBAAO,KAAK,IAAI,EAAE,IAAI,UAAU,EAAE,IAAI;AAAA,MAC1C,WAAW,WAAW,WAAW;AAC/B,eAAO,KAAK,IAAI,WAAW,IAAI,CAAC,EAAE,IAAI;AACxC,UAAI,KAAK,GAAG,UAAU,KAAK,WAAW,GAAG,UAAU;AACjD,eAAO,WAAW,KAAK,SAAS,IAAI,WAAW,SAAS,GAAG,KAAK,QAAQ;AAC1E,UAAI,MAAM,KAAK,SAAS;AACxB,UAAI,MAAM,KAAK,OAAO;AACtB,UAAI,MAAM,KAAK,QAAQ;AACvB,UAAI,MAAM,KAAK,MAAM;AACrB,UAAI,MAAM,WAAW,SAAS;AAC9B,UAAI,MAAM,WAAW,OAAO;AAC5B,UAAI,MAAM,WAAW,QAAQ;AAC7B,UAAI,MAAM,WAAW,MAAM;AAC3B,UAAI,MAAM,GAAG,MAAM,GAAG,MAAM,GAAG,MAAM;AACrC,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM;AACb,aAAO,QAAQ;AACf,aAAO;AACP,aAAO,MAAM,MAAM,MAAM,MAAM,MAAM,MAAM,MAAM;AACjD,aAAO;AACP,aAAO,SAAS,OAAO,KAAK,KAAK,OAAO,KAAK,KAAK,KAAK,QAAQ;AAAA,IACjE;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,SAAS,SAAS,OAAO,SAAS;AAC9C,UAAI,CAAC,OAAO,OAAO;AACjB,kBAAU,UAAU,OAAO;AAC7B,UAAI,QAAQ,OAAO;AACjB,cAAM,MAAM,kBAAkB;AAChC,UAAI,MAAM;AACR,YAAI,CAAC,KAAK,YAAY,KAAK,SAAS,eAAe,QAAQ,QAAQ,MAAM,QAAQ,SAAS,IAAI;AAC5F,iBAAO;AAAA,QACT;AACA,YAAI,OAAO,KAAK,WAAW,KAAK,QAAQ,KAAK;AAAA,UAC3C,KAAK;AAAA,UACL,KAAK;AAAA,UACL,QAAQ;AAAA,UACR,QAAQ;AAAA,QACV;AACA,eAAO,SAAS,KAAK,KAAK,SAAS,GAAG,KAAK,QAAQ;AAAA,MACrD;AACA,UAAI,KAAK,OAAO;AACd,eAAO,KAAK,WAAW,QAAQ;AACjC,UAAI,QAAQ,KAAK;AACjB,UAAI,CAAC,KAAK,UAAU;AAClB,YAAI,KAAK,GAAG,SAAS,GAAG;AACtB,cAAI,QAAQ,GAAG,GAAG,KAAK,QAAQ,GAAG,OAAO;AACvC,mBAAO;AAAA,mBACA,QAAQ,GAAG,SAAS;AAC3B,mBAAO;AAAA,eACJ;AACH,gBAAI,WAAW,KAAK,IAAI,CAAC;AACzB,qBAAS,SAAS,IAAI,OAAO,EAAE,IAAI,CAAC;AACpC,gBAAI,OAAO,GAAG,IAAI,GAAG;AACnB,qBAAO,QAAQ,WAAW,IAAI,MAAM;AAAA,YACtC,OAAO;AACL,oBAAM,KAAK,IAAI,QAAQ,IAAI,MAAM,CAAC;AAClC,oBAAM,OAAO,IAAI,IAAI,IAAI,OAAO,CAAC;AACjC,qBAAO;AAAA,YACT;AAAA,UACF;AAAA,QACF,WAAW,QAAQ,GAAG,SAAS;AAC7B,iBAAO,KAAK,WAAW,QAAQ;AACjC,YAAI,KAAK,WAAW,GAAG;AACrB,cAAI,QAAQ,WAAW;AACrB,mBAAO,KAAK,IAAI,EAAE,IAAI,QAAQ,IAAI,CAAC;AACrC,iBAAO,KAAK,IAAI,EAAE,IAAI,OAAO,EAAE,IAAI;AAAA,QACrC,WAAW,QAAQ,WAAW;AAC5B,iBAAO,KAAK,IAAI,QAAQ,IAAI,CAAC,EAAE,IAAI;AACrC,cAAM;AAAA,MACR,OAAO;AACL,YAAI,CAAC,QAAQ;AACX,oBAAU,QAAQ,WAAW;AAC/B,YAAI,QAAQ,GAAG,IAAI;AACjB,iBAAO;AACT,YAAI,QAAQ,GAAG,KAAK,KAAK,CAAC,CAAC;AACzB,iBAAO;AACT,cAAM;AAAA,MACR;AACA,YAAM;AACN,aAAO,IAAI,IAAI,OAAO,GAAG;AACvB,iBAAS,KAAK,IAAI,GAAG,KAAK,MAAM,IAAI,SAAS,IAAI,QAAQ,SAAS,CAAC,CAAC;AACpE,YAAI,QAAQ,KAAK,KAAK,KAAK,IAAI,MAAM,IAAI,KAAK,GAAG,GAAG,QAAQ,SAAS,KAAK,IAAI,QAAQ,GAAG,QAAQ,EAAE,GAAG,YAAY,WAAW,MAAM,GAAG,YAAY,UAAU,IAAI,OAAO;AACvK,eAAO,UAAU,WAAW,KAAK,UAAU,GAAG,GAAG,GAAG;AAClD,oBAAU;AACV,sBAAY,WAAW,QAAQ,KAAK,QAAQ;AAC5C,sBAAY,UAAU,IAAI,OAAO;AAAA,QACnC;AACA,YAAI,UAAU,OAAO;AACnB,sBAAY;AACd,cAAM,IAAI,IAAI,SAAS;AACvB,cAAM,IAAI,IAAI,SAAS;AAAA,MACzB;AACA,aAAO;AAAA,IACT;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,SAAS,SAAS,OAAO,SAAS;AAC9C,UAAI,CAAC,OAAO,OAAO;AACjB,kBAAU,UAAU,OAAO;AAC7B,UAAI,MAAM;AACR,YAAI,OAAO,KAAK,WAAW,KAAK,QAAQ,KAAK;AAAA,UAC3C,KAAK;AAAA,UACL,KAAK;AAAA,UACL,QAAQ;AAAA,UACR,QAAQ;AAAA,QACV;AACA,eAAO,SAAS,KAAK,KAAK,SAAS,GAAG,KAAK,QAAQ;AAAA,MACrD;AACA,aAAO,KAAK,IAAI,KAAK,IAAI,OAAO,EAAE,IAAI,OAAO,CAAC;AAAA,IAChD;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,MAAM,cAAc;AAClC,kBAAc,MAAM,SAAS,MAAM;AACjC,aAAO,SAAS,CAAC,KAAK,KAAK,CAAC,KAAK,MAAM,KAAK,QAAQ;AAAA,IACtD;AACA,kBAAc,MAAM,SAAS,IAAI,OAAO;AACtC,UAAI,CAAC,OAAO,KAAK;AACf,gBAAQ,UAAU,KAAK;AACzB,aAAO,SAAS,KAAK,MAAM,MAAM,KAAK,KAAK,OAAO,MAAM,MAAM,KAAK,QAAQ;AAAA,IAC7E;AACA,kBAAc,KAAK,SAAS,GAAG,OAAO;AACpC,UAAI,CAAC,OAAO,KAAK;AACf,gBAAQ,UAAU,KAAK;AACzB,aAAO,SAAS,KAAK,MAAM,MAAM,KAAK,KAAK,OAAO,MAAM,MAAM,KAAK,QAAQ;AAAA,IAC7E;AACA,kBAAc,MAAM,SAAS,IAAI,OAAO;AACtC,UAAI,CAAC,OAAO,KAAK;AACf,gBAAQ,UAAU,KAAK;AACzB,aAAO,SAAS,KAAK,MAAM,MAAM,KAAK,KAAK,OAAO,MAAM,MAAM,KAAK,QAAQ;AAAA,IAC7E;AACA,kBAAc,YAAY,SAAS,UAAU,SAAS;AACpD,UAAI,OAAO,OAAO;AAChB,kBAAU,QAAQ,MAAM;AAC1B,WAAK,WAAW,QAAQ;AACtB,eAAO;AAAA,eACA,UAAU;AACjB,eAAO,SAAS,KAAK,OAAO,SAAS,KAAK,QAAQ,UAAU,KAAK,QAAQ,KAAK,SAAS,KAAK,QAAQ;AAAA;AAEpG,eAAO,SAAS,GAAG,KAAK,OAAO,UAAU,IAAI,KAAK,QAAQ;AAAA,IAC9D;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,aAAa,SAAS,WAAW,SAAS;AACtD,UAAI,OAAO,OAAO;AAChB,kBAAU,QAAQ,MAAM;AAC1B,WAAK,WAAW,QAAQ;AACtB,eAAO;AAAA,eACA,UAAU;AACjB,eAAO,SAAS,KAAK,QAAQ,UAAU,KAAK,QAAQ,KAAK,SAAS,KAAK,QAAQ,SAAS,KAAK,QAAQ;AAAA;AAErG,eAAO,SAAS,KAAK,QAAQ,UAAU,IAAI,KAAK,QAAQ,IAAI,IAAI,IAAI,KAAK,QAAQ;AAAA,IACrF;AACA,kBAAc,MAAM,cAAc;AAClC,kBAAc,qBAAqB,SAAS,mBAAmB,SAAS;AACtE,UAAI,OAAO,OAAO;AAChB,kBAAU,QAAQ,MAAM;AAC1B,iBAAW;AACX,UAAI,YAAY;AACd,eAAO;AAAA,WACJ;AACH,YAAI,OAAO,KAAK;AAChB,YAAI,UAAU,IAAI;AAChB,cAAI,MAAM,KAAK;AACf,iBAAO,SAAS,QAAQ,UAAU,QAAQ,KAAK,SAAS,SAAS,SAAS,KAAK,QAAQ;AAAA,QACzF,WAAW,YAAY;AACrB,iBAAO,SAAS,MAAM,GAAG,KAAK,QAAQ;AAAA;AAEtC,iBAAO,SAAS,SAAS,UAAU,IAAI,GAAG,KAAK,QAAQ;AAAA,MAC3D;AAAA,IACF;AACA,kBAAc,OAAO,cAAc;AACnC,kBAAc,QAAQ,cAAc;AACpC,kBAAc,WAAW,SAAS,WAAW;AAC3C,UAAI,CAAC,KAAK;AACR,eAAO;AACT,aAAO,SAAS,KAAK,KAAK,KAAK,MAAM,KAAK;AAAA,IAC5C;AACA,kBAAc,aAAa,SAAS,aAAa;AAC/C,UAAI,KAAK;AACP,eAAO;AACT,aAAO,SAAS,KAAK,KAAK,KAAK,MAAM,IAAI;AAAA,IAC3C;AACA,kBAAc,UAAU,SAAS,QAAQ,IAAI;AAC3C,aAAO,KAAK,KAAK,UAAU,IAAI,KAAK,UAAU;AAAA,IAChD;AACA,kBAAc,YAAY,SAAS,YAAY;AAC7C,UAAI,KAAK,KAAK,MAAM,KAAK,KAAK;AAC9B,aAAO;AAAA,QACL,KAAK;AAAA,QACL,OAAO,IAAI;AAAA,QACX,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,QACP,KAAK;AAAA,QACL,OAAO,IAAI;AAAA,QACX,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,MACT;AAAA,IACF;AACA,kBAAc,YAAY,SAAS,YAAY;AAC7C,UAAI,KAAK,KAAK,MAAM,KAAK,KAAK;AAC9B,aAAO;AAAA,QACL,OAAO;AAAA,QACP,OAAO,KAAK;AAAA,QACZ,OAAO,IAAI;AAAA,QACX,KAAK;AAAA,QACL,OAAO;AAAA,QACP,OAAO,KAAK;AAAA,QACZ,OAAO,IAAI;AAAA,QACX,KAAK;AAAA,MACP;AAAA,IACF;AACA,UAAM,YAAY,SAAS,UAAU,OAAO,UAAU,IAAI;AACxD,aAAO,KAAK,MAAM,YAAY,OAAO,QAAQ,IAAI,MAAM,YAAY,OAAO,QAAQ;AAAA,IACpF;AACA,UAAM,cAAc,SAAS,YAAY,OAAO,UAAU;AACxD,aAAO,IAAI;AAAA,QACT,MAAM,KAAK,MAAM,MAAM,IAAI,MAAM,MAAM,KAAK,MAAM,MAAM;AAAA,QACxD,MAAM,KAAK,MAAM,MAAM,IAAI,MAAM,MAAM,KAAK,MAAM,MAAM;AAAA,QACxD;AAAA,MACF;AAAA,IACF;AACA,UAAM,cAAc,SAAS,YAAY,OAAO,UAAU;AACxD,aAAO,IAAI;AAAA,QACT,MAAM,MAAM,KAAK,MAAM,MAAM,KAAK,MAAM,MAAM,IAAI,MAAM;AAAA,QACxD,MAAM,MAAM,KAAK,MAAM,MAAM,KAAK,MAAM,MAAM,IAAI,MAAM;AAAA,QACxD;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF,CAAC;AAGD,IAAI,kBAAkB,WAAW;AAAA,EAC/B,kDAAkD;AAAA,EAClD;AACF,CAAC;AAGD,IAAI,eAAe,WAAW;AAAA,EAC5B,oBAAoB;AAAA,EACpB;AACF,CAAC;AAGD,IAAI,eAAe,WAAW;AAAA,EAC5B,sCAAsC,SAAS,QAAQ;AACrD,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,KAAK,MAAM;AAClB,YAAI,KAAK,MAAM,OAAO,KAAK;AAC3B,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,UAAU,GAAG,KAAK,GAAG,IAAI;AACjC,aAAG,KAAK,GAAG;AACX,aAAG,KAAK,GAAG;AACX,iBAAO,GAAG,KAAK,KAAK,GAAG,IAAI,IAAI;AAAA,QACjC;AACA,WAAG,IAAI;AACP,WAAG,KAAK,KAAK,GAAG;AAChB,WAAG,KAAK,KAAK,GAAG;AAChB,WAAG,KAAK,KAAK,GAAG;AAChB,WAAG,MAAM,KAAK,IAAI;AAClB,YAAI,GAAG,KAAK,GAAG;AACb,aAAG,MAAM;AAAA,QACX;AACA,WAAG,MAAM,KAAK,IAAI;AAClB,YAAI,GAAG,KAAK,GAAG;AACb,aAAG,MAAM;AAAA,QACX;AACA,WAAG,MAAM,KAAK,IAAI;AAClB,YAAI,GAAG,KAAK,GAAG;AACb,aAAG,MAAM;AAAA,QACX;AACA,eAAO;AAAA,MACT;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,KAAK,EAAE;AACT,UAAE,KAAK,EAAE;AACT,UAAE,KAAK,EAAE;AACT,eAAO;AAAA,MACT;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,KAAK,IAAI,KAAK,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,GAAG;AAC/D,aAAK,QAAQ,WAAW;AACtB,iBAAO,GAAG,KAAK,IAAI,aAAa;AAAA,QAClC;AACA,aAAK,SAAS,WAAW;AACvB,iBAAO,KAAK,KAAK,KAAK,IAAI,UAAU,KAAK;AAAA,QAC3C;AACA,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,OAAO,SAAS;AAClB,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,eAAS,OAAO;AACd,YAAI,IAAI;AACR,YAAI,OAAO,SAAS,MAAM;AACxB,iBAAO,OAAO,IAAI;AAClB,mBAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,iBAAK,KAAK,WAAW,CAAC;AACtB,gBAAI,IAAI,sBAAsB;AAC9B,gBAAI,MAAM;AACV,iBAAK;AACL,iBAAK;AACL,gBAAI,MAAM;AACV,iBAAK;AACL,iBAAK,IAAI;AAAA,UACX;AACA,kBAAQ,MAAM,KAAK;AAAA,QACrB;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,OAAO;AAAA,MACd;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,iBAAiB,WAAW;AAAA,EAC9B,wCAAwC,SAAS,QAAQ;AACvD,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,OAAO,MAAM;AACpB,YAAI,KAAK,MAAM,UAAU;AACzB,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,GAAG,IAAI,GAAG,KAAK;AACvB,aAAG,IAAI,GAAG;AACV,aAAG,IAAI,GAAG;AACV,aAAG,IAAI,GAAG;AACV,iBAAO,GAAG,KAAK,GAAG,MAAM,KAAK,IAAI,MAAM;AAAA,QACzC;AACA,YAAI,UAAU,OAAO,IAAI;AACvB,aAAG,IAAI;AAAA,QACT,OAAO;AACL,qBAAW;AAAA,QACb;AACA,iBAAS,IAAI,GAAG,IAAI,QAAQ,SAAS,IAAI,KAAK;AAC5C,aAAG,KAAK,QAAQ,WAAW,CAAC,IAAI;AAChC,aAAG,KAAK;AAAA,QACV;AAAA,MACF;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,eAAO;AAAA,MACT;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,KAAK,IAAI,OAAO,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,WAAW;AACvE,kBAAQ,GAAG,KAAK,MAAM,KAAK;AAAA,QAC7B;AACA,aAAK,SAAS,WAAW;AACvB,aAAG;AACD,gBAAI,MAAM,GAAG,KAAK,MAAM,IAAI,OAAO,GAAG,KAAK,MAAM,KAAK,YAAY,UAAU,MAAM,QAAQ,KAAK;AAAA,UACjG,SAAS,WAAW;AACpB,iBAAO;AAAA,QACT;AACA,aAAK,QAAQ,GAAG;AAChB,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,OAAO,SAAS;AAClB,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,SAAS;AAAA,MAChB;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,iBAAiB,WAAW;AAAA,EAC9B,wCAAwC,SAAS,QAAQ;AACvD,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,OAAO,MAAM;AACpB,YAAI,KAAK,MAAM,UAAU;AACzB,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,GAAG,IAAI,GAAG,MAAM;AACxB,aAAG,IAAI,GAAG;AACV,aAAG,IAAI,GAAG;AACV,aAAG,IAAI,GAAG;AACV,aAAG,IAAI,GAAG;AACV,kBAAQ,GAAG,IAAI,GAAG,IAAI,SAAS,MAAM,GAAG,IAAI,GAAG,IAAI,GAAG,KAAK,KAAK,IAAI,KAAK,MAAM;AAAA,QACjF;AACA,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI;AACP,YAAI,UAAU,OAAO,IAAI;AACvB,aAAG,IAAI;AAAA,QACT,OAAO;AACL,qBAAW;AAAA,QACb;AACA,iBAAS,IAAI,GAAG,IAAI,QAAQ,SAAS,IAAI,KAAK;AAC5C,aAAG,KAAK,QAAQ,WAAW,CAAC,IAAI;AAChC,cAAI,KAAK,QAAQ,QAAQ;AACvB,eAAG,IAAI,GAAG,KAAK,KAAK,GAAG,MAAM;AAAA,UAC/B;AACA,aAAG,KAAK;AAAA,QACV;AAAA,MACF;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,eAAO;AAAA,MACT;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,KAAK,IAAI,OAAO,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,WAAW;AACvE,kBAAQ,GAAG,KAAK,MAAM,KAAK;AAAA,QAC7B;AACA,aAAK,SAAS,WAAW;AACvB,aAAG;AACD,gBAAI,MAAM,GAAG,KAAK,MAAM,IAAI,OAAO,GAAG,KAAK,MAAM,KAAK,YAAY,UAAU,MAAM,QAAQ,KAAK;AAAA,UACjG,SAAS,WAAW;AACpB,iBAAO;AAAA,QACT;AACA,aAAK,QAAQ,GAAG;AAChB,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,OAAO,SAAS;AAClB,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,SAAS;AAAA,MAChB;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,oBAAoB,WAAW;AAAA,EACjC,2CAA2C,SAAS,QAAQ;AAC1D,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,OAAO,MAAM;AACpB,YAAI,KAAK;AACT,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,GAAG,GAAG;AAC9B,cAAI,EAAE;AACN,eAAK,MAAM;AACX,cAAI,IAAI,KAAK;AACb,cAAI,EAAE,IAAI,IAAI;AACd,eAAK,IAAI,MAAM;AACf,cAAI,EAAE,IAAI,IAAI;AACd,eAAK,IAAI,MAAM;AACf,cAAI,EAAE,IAAI,IAAI;AACd,eAAK,IAAI,KAAK;AACd,cAAI,EAAE,IAAI,IAAI;AACd,cAAI,IAAI,KAAK;AACb,eAAK,IAAI,KAAK;AACd,YAAE,KAAK;AACP,aAAG,IAAI,IAAI,IAAI;AACf,iBAAO;AAAA,QACT;AACA,iBAAS,MAAM,KAAK,OAAO;AACzB,cAAI,GAAG,GAAG,IAAI,CAAC;AACf,cAAI,WAAW,QAAQ,IAAI;AACzB,gBAAI,EAAE,KAAK;AAAA,UACb,OAAO;AACL,oBAAQ,KAAK;AACb,iBAAK,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACjC,gBAAE,IAAI,KAAK,EAAE,IAAI,MAAM,KAAK,MAAM,WAAW,CAAC,IAAI,EAAE,IAAI,IAAI,MAAM;AAAA,YACpE;AAAA,UACF;AACA,iBAAO,EAAE,SAAS;AAChB,cAAE,KAAK,CAAC;AACV,eAAK,IAAI,GAAG,IAAI,KAAK,EAAE,OAAO,GAAG,EAAE;AACjC;AACF,cAAI,KAAK;AACP,gBAAI,EAAE,KAAK;AAAA;AAEX,gBAAI,EAAE;AACR,cAAI,IAAI;AACR,cAAI,IAAI;AACR,eAAK,IAAI,KAAK,IAAI,GAAG,EAAE,GAAG;AACxB,gBAAI,KAAK;AAAA,UACX;AAAA,QACF;AACA,cAAM,IAAI,IAAI;AAAA,MAChB;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE,EAAE,MAAM;AAChB,UAAE,IAAI,EAAE;AACR,eAAO;AAAA,MACT;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,QAAQ;AACV,iBAAO,CAAC,IAAI,KAAK;AACnB,YAAI,KAAK,IAAI,OAAO,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,WAAW;AACvE,kBAAQ,GAAG,KAAK,MAAM,KAAK;AAAA,QAC7B;AACA,aAAK,SAAS,WAAW;AACvB,aAAG;AACD,gBAAI,MAAM,GAAG,KAAK,MAAM,IAAI,OAAO,GAAG,KAAK,MAAM,KAAK,YAAY,UAAU,MAAM,QAAQ,KAAK;AAAA,UACjG,SAAS,WAAW;AACpB,iBAAO;AAAA,QACT;AACA,aAAK,QAAQ,GAAG;AAChB,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,MAAM;AACR,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,YAAY;AAAA,MACnB;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,kBAAkB,WAAW;AAAA,EAC/B,yCAAyC,SAAS,QAAQ;AACxD,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,OAAO,MAAM;AACpB,YAAI,KAAK;AACT,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,GAAG;AACrC,aAAG,IAAI,IAAI,IAAI,aAAa;AAC5B,cAAI,EAAE,IAAI,KAAK;AACf,cAAI,EAAE,IAAI,IAAI,IAAI;AAClB,eAAK,KAAK;AACV,eAAK,KAAK;AACV,eAAK,MAAM;AACX,eAAK,MAAM;AACX,cAAI,EAAE,KAAK,IAAI;AACf,aAAG,IAAI;AACP,iBAAO,KAAK,IAAI,MAAM,MAAM;AAAA,QAC9B;AACA,iBAAS,MAAM,KAAK,OAAO;AACzB,cAAI,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,CAAC,GAAG,QAAQ;AACnC,cAAI,WAAW,QAAQ,IAAI;AACzB,gBAAI;AACJ,oBAAQ;AAAA,UACV,OAAO;AACL,oBAAQ,QAAQ;AAChB,gBAAI;AACJ,oBAAQ,KAAK,IAAI,OAAO,MAAM,MAAM;AAAA,UACtC;AACA,eAAK,IAAI,GAAG,IAAI,KAAK,IAAI,OAAO,EAAE,GAAG;AACnC,gBAAI;AACF,mBAAK,MAAM,YAAY,IAAI,MAAM,MAAM,MAAM;AAC/C,gBAAI,MAAM;AACR,kBAAI;AACN,iBAAK,KAAK;AACV,iBAAK,MAAM;AACX,iBAAK,KAAK;AACV,iBAAK,MAAM;AACX,gBAAI,KAAK,GAAG;AACV,kBAAI,IAAI,aAAa;AACrB,kBAAI,EAAE,IAAI,QAAQ,IAAI;AACtB,kBAAI,KAAK,IAAI,IAAI,IAAI;AAAA,YACvB;AAAA,UACF;AACA,cAAI,KAAK,KAAK;AACZ,eAAG,SAAS,MAAM,UAAU,KAAK,OAAO;AAAA,UAC1C;AACA,cAAI;AACJ,eAAK,IAAI,IAAI,KAAK,IAAI,GAAG,EAAE,GAAG;AAC5B,gBAAI,EAAE,IAAI,KAAK;AACf,gBAAI,EAAE,IAAI,IAAI,IAAI;AAClB,iBAAK,KAAK;AACV,iBAAK,KAAK;AACV,iBAAK,MAAM;AACX,iBAAK,MAAM;AACX,cAAE,KAAK,IAAI;AAAA,UACb;AACA,cAAI,IAAI;AACR,cAAI,IAAI;AACR,cAAI,IAAI;AAAA,QACV;AACA,cAAM,IAAI,IAAI;AAAA,MAChB;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE,EAAE,MAAM;AAChB,eAAO;AAAA,MACT;AACA;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,QAAQ;AACV,iBAAO,CAAC,IAAI,KAAK;AACnB,YAAI,KAAK,IAAI,OAAO,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,WAAW;AACvE,kBAAQ,GAAG,KAAK,MAAM,KAAK;AAAA,QAC7B;AACA,aAAK,SAAS,WAAW;AACvB,aAAG;AACD,gBAAI,MAAM,GAAG,KAAK,MAAM,IAAI,OAAO,GAAG,KAAK,MAAM,KAAK,YAAY,UAAU,MAAM,QAAQ,KAAK;AAAA,UACjG,SAAS,WAAW;AACpB,iBAAO;AAAA,QACT;AACA,aAAK,QAAQ,GAAG;AAChB,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,MAAM;AACR,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,UAAU;AAAA,MACjB;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,iBAAiB,WAAW;AAAA,EAC9B,wCAAwC,SAAS,QAAQ;AACvD,KAAC,SAAS,SAAS,SAAS,SAAS;AACnC,eAAS,OAAO,MAAM;AACpB,YAAI,KAAK,MAAM,UAAU;AACzB,WAAG,OAAO,WAAW;AACnB,cAAI,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG,GAAG,IAAI,GAAG;AACzC,cAAI,KAAK,KAAK,MAAM,IAAI;AACxB,cAAI,IAAI,IAAI;AACZ,cAAI,KAAK,KAAK,MAAM,IAAI;AACxB,cAAI,IAAI,IAAI;AACZ,aAAG,IAAI,IAAI,KAAK,KAAK,MAAM,KAAK;AAChC,aAAG,IAAI,IAAI,IAAI,IAAI;AACnB,aAAG,IAAI,KAAK,KAAK,MAAM,KAAK;AAC5B,iBAAO,GAAG,IAAI,IAAI,IAAI;AAAA,QACxB;AACA,WAAG,IAAI;AACP,WAAG,IAAI;AACP,WAAG,IAAI,aAAa;AACpB,WAAG,IAAI;AACP,YAAI,SAAS,KAAK,MAAM,IAAI,GAAG;AAC7B,aAAG,IAAI,OAAO,aAAa;AAC3B,aAAG,IAAI,OAAO;AAAA,QAChB,OAAO;AACL,qBAAW;AAAA,QACb;AACA,iBAAS,IAAI,GAAG,IAAI,QAAQ,SAAS,IAAI,KAAK;AAC5C,aAAG,KAAK,QAAQ,WAAW,CAAC,IAAI;AAChC,aAAG,KAAK;AAAA,QACV;AAAA,MACF;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,eAAO;AAAA,MACT;AACA;AACA,eAAS,KAAK,MAAM,MAAM;AACxB,YAAI,KAAK,IAAI,OAAO,IAAI,GAAG,QAAQ,QAAQ,KAAK,OAAO,OAAO,WAAW;AACvE,kBAAQ,GAAG,KAAK,MAAM,KAAK;AAAA,QAC7B;AACA,aAAK,SAAS,WAAW;AACvB,aAAG;AACD,gBAAI,MAAM,GAAG,KAAK,MAAM,IAAI,OAAO,GAAG,KAAK,MAAM,KAAK,YAAY,UAAU,MAAM,QAAQ,KAAK;AAAA,UACjG,SAAS,WAAW;AACpB,iBAAO;AAAA,QACT;AACA,aAAK,QAAQ,GAAG;AAChB,aAAK,QAAQ;AACb,YAAI,OAAO;AACT,cAAI,OAAO,SAAS;AAClB,iBAAK,OAAO,EAAE;AAChB,eAAK,QAAQ,WAAW;AACtB,mBAAO,KAAK,IAAI,CAAC,CAAC;AAAA,UACpB;AAAA,QACF;AACA,eAAO;AAAA,MACT;AACA,UAAI,WAAW,QAAQ,SAAS;AAC9B,gBAAQ,UAAU;AAAA,MACpB,WAAW,WAAW,QAAQ,KAAK;AACjC,gBAAQ,WAAW;AACjB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,SAAS;AAAA,MAChB;AAAA,IACF;AAAA,MACE;AAAA,MACA,OAAO,UAAU,YAAY;AAAA,MAC7B,OAAO,UAAU,cAAc;AAAA,IACjC;AAAA,EACF;AACF,CAAC;AAGD,IAAI,iBAAiB,WAAW;AAAA,EAC9B,sBAAsB;AAAA,EACtB;AACF,CAAC;AAGD,IAAI,qBAAqB,WAAW;AAAA,EAClC,wCAAwC,SAAS,QAAQ;AACvD,KAAC,SAAS,SAAS,OAAO,MAAM;AAC9B,UAAI,QAAQ,KAAK,SAAS,GAAG,SAAS,IAAI,UAAU,UAAU,aAAa,KAAK,IAAI,OAAO,MAAM,GAAG,eAAe,KAAK,IAAI,GAAG,MAAM,GAAG,WAAW,eAAe,GAAG,OAAO,QAAQ,GAAG;AACvL,eAAS,YAAY,MAAM,SAAS,UAAU;AAC5C,YAAI,MAAM,CAAC;AACX,kBAAU,WAAW,OAAO,EAAE,SAAS,KAAK,IAAI,WAAW,CAAC;AAC5D,YAAI,YAAY,OAAO;AAAA,UACrB,QAAQ,UAAU,CAAC,MAAM,SAAS,KAAK,CAAC,IAAI,QAAQ,OAAO,SAAS,IAAI;AAAA,UACxE;AAAA,QACF,GAAG,GAAG;AACN,YAAI,OAAO,IAAI,KAAK,GAAG;AACvB,YAAI,OAAO,WAAW;AACpB,cAAI,IAAI,KAAK,EAAE,MAAM,GAAG,IAAI,YAAY,IAAI;AAC5C,iBAAO,IAAI,cAAc;AACvB,iBAAK,IAAI,KAAK;AACd,iBAAK;AACL,gBAAI,KAAK,EAAE,CAAC;AAAA,UACd;AACA,iBAAO,KAAK,UAAU;AACpB,iBAAK;AACL,iBAAK;AACL,mBAAO;AAAA,UACT;AACA,kBAAQ,IAAI,KAAK;AAAA,QACnB;AACA,aAAK,QAAQ,WAAW;AACtB,iBAAO,KAAK,EAAE,CAAC,IAAI;AAAA,QACrB;AACA,aAAK,QAAQ,WAAW;AACtB,iBAAO,KAAK,EAAE,CAAC,IAAI;AAAA,QACrB;AACA,aAAK,SAAS;AACd,eAAO,SAAS,KAAK,CAAC,GAAG,KAAK;AAC9B,gBAAQ,QAAQ,QAAQ,YAAY,SAAS,OAAO,OAAO,cAAc,OAAO;AAC9E,cAAI,OAAO;AACT,gBAAI,MAAM,GAAG;AACX,mBAAK,OAAO,IAAI;AAAA,YAClB;AACA,kBAAM,QAAQ,WAAW;AACvB,qBAAO,KAAK,MAAM,CAAC,CAAC;AAAA,YACtB;AAAA,UACF;AACA,cAAI,cAAc;AAChB,iBAAK,WAAW;AAChB,mBAAO;AAAA,UACT;AACE,mBAAO;AAAA,QACX;AAAA,UACE;AAAA,UACA;AAAA,UACA,YAAY,UAAU,QAAQ,SAAS,QAAQ;AAAA,UAC/C,QAAQ;AAAA,QACV;AAAA,MACF;AACA,eAAS,KAAK,KAAK;AACjB,YAAI,GAAG,SAAS,IAAI,QAAQ,KAAK,MAAM,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,GAAG,IAAI,CAAC;AAC/E,YAAI,CAAC,QAAQ;AACX,gBAAM,CAAC,QAAQ;AAAA,QACjB;AACA,eAAO,IAAI,OAAO;AAChB,YAAE,KAAK;AAAA,QACT;AACA,aAAK,IAAI,GAAG,IAAI,OAAO,KAAK;AAC1B,YAAE,KAAK,EAAE,IAAI,OAAO,IAAI,IAAI,IAAI,WAAW,IAAI,EAAE;AACjD,YAAE,KAAK;AAAA,QACT;AACA,SAAC,GAAG,IAAI,SAAS,QAAQ;AACvB,cAAI,IAAI,IAAI,GAAG,KAAK,GAAG,GAAG,KAAK,GAAG,GAAG,KAAK,GAAG;AAC7C,iBAAO,UAAU;AACf,iBAAK,GAAG,KAAK,OAAO,KAAK;AACzB,gBAAI,IAAI,QAAQ,GAAG,QAAQ,GAAG,MAAM,GAAG,KAAK,OAAO,KAAK,QAAQ,GAAG,MAAM;AAAA,UAC3E;AACA,aAAG,IAAI;AACP,aAAG,IAAI;AACP,iBAAO;AAAA,QACT,GAAG,KAAK;AAAA,MACV;AACA,eAAS,KAAK,GAAG,GAAG;AAClB,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE;AACR,UAAE,IAAI,EAAE,EAAE,MAAM;AAChB,eAAO;AAAA,MACT;AACA;AACA,eAAS,SAAS,KAAK,OAAO;AAC5B,YAAI,SAAS,CAAC,GAAG,MAAM,OAAO,KAAK;AACnC,YAAI,SAAS,OAAO,UAAU;AAC5B,eAAK,QAAQ,KAAK;AAChB,gBAAI;AACF,qBAAO,KAAK,SAAS,IAAI,OAAO,QAAQ,CAAC,CAAC;AAAA,YAC5C,SAAS,GAAP;AAAA,YACF;AAAA,UACF;AAAA,QACF;AACA,eAAO,OAAO,SAAS,SAAS,OAAO,WAAW,MAAM,MAAM;AAAA,MAChE;AACA,eAAS,OAAO,MAAM,KAAK;AACzB,YAAI,aAAa,OAAO,IAAI,OAAO,IAAI;AACvC,eAAO,IAAI,WAAW,QAAQ;AAC5B,cAAI,OAAO,KAAK,QAAQ,SAAS,IAAI,OAAO,KAAK,MAAM,WAAW,WAAW,GAAG;AAAA,QAClF;AACA,eAAO,SAAS,GAAG;AAAA,MACrB;AACA,eAAS,WAAW;AAClB,YAAI;AACF,cAAI;AACJ,cAAI,eAAe,MAAM,WAAW,cAAc;AAChD,kBAAM,IAAI,KAAK;AAAA,UACjB,OAAO;AACL,kBAAM,IAAI,WAAW,KAAK;AAC1B,aAAC,QAAQ,UAAU,QAAQ,UAAU,gBAAgB,GAAG;AAAA,UAC1D;AACA,iBAAO,SAAS,GAAG;AAAA,QACrB,SAAS,GAAP;AACA,cAAI,UAAU,QAAQ,WAAW,UAAU,WAAW,QAAQ;AAC9D,iBAAO,CAAC,CAAC,IAAI,KAAK,GAAG,SAAS,SAAS,QAAQ,QAAQ,SAAS,KAAK,CAAC;AAAA,QACxE;AAAA,MACF;AACA,eAAS,SAAS,GAAG;AACnB,eAAO,OAAO,aAAa,MAAM,GAAG,CAAC;AAAA,MACvC;AACA,aAAO,KAAK,OAAO,GAAG,KAAK;AAC3B,UAAI,OAAO,UAAU,YAAY,OAAO,SAAS;AAC/C,eAAO,UAAU;AACjB,YAAI;AACF,uBAAa,eAAe;AAAA,QAC9B,SAAS,IAAP;AAAA,QACF;AAAA,MACF,WAAW,OAAO,UAAU,cAAc,OAAO,KAAK;AACpD,eAAO,WAAW;AAChB,iBAAO;AAAA,QACT,CAAC;AAAA,MACH,OAAO;AACL,aAAK,SAAS,WAAW;AAAA,MAC3B;AAAA,IACF;AAAA,MACE,OAAO,SAAS,cAAc,OAAO;AAAA,MACrC,CAAC;AAAA,MACD;AAAA,IACF;AAAA,EACF;AACF,CAAC;AAGD,IAAI,sBAAsB,WAAW;AAAA,EACnC,mCAAmC,SAAS,QAAQ;AAClD,QAAI,QAAQ,aAAa;AACzB,QAAI,SAAS,eAAe;AAC5B,QAAI,SAAS,eAAe;AAC5B,QAAI,YAAY,kBAAkB;AAClC,QAAI,UAAU,gBAAgB;AAC9B,QAAI,SAAS,eAAe;AAC5B,QAAI,KAAK,mBAAmB;AAC5B,OAAG,OAAO;AACV,OAAG,SAAS;AACZ,OAAG,SAAS;AACZ,OAAG,YAAY;AACf,OAAG,UAAU;AACb,OAAG,SAAS;AACZ,WAAO,UAAU;AAAA,EACnB;AACF,CAAC;AAGD,IAAI,yBAAyB,WAAW;AAAA,EACtC,iEAAiE;AAAA,EACjE;AACF,CAAC;AAGD,IAAI,aAAa,WAAW;AAAA,EAC1B,kBAAkB;AAAA,EAClB;AACF,CAAC;AAGD,IAAI,eAAe,WAAW;AAAA,EAC5B,oBAAoB;AAAA,EACpB;AACF,CAAC;AAGD,IAAI,yBAAyB,WAAW;AAAA,EACtC,8BAA8B;AAAA,EAC9B;AACF,CAAC;AAGD,IAAI,qBAAqB,WAAW;AAAA,EAClC,0BAA0B;AAAA,EAC1B;AACF,CAAC;AAGD,IAAI,aAAa,WAAW;AAAA,EAC1B,kBAAkB;AAAA,EAClB;AACF,CAAC;AAGD,IAAI,0CAA0C,WAAW;AAAA,EACvD,yFAAyF,SAAS,QAAQ;AACxG,QAAI,kCAAkC,MAAM;AAC1C,UAAI,aAAa,OAAO,aAAa,eAAe,SAAS,gBAAgB,SAAS,cAAc,MAAM;AAC1G,UAAI,OAAO,eAAe;AACxB,qBAAa,cAAc;AAC7B,aAAO,SAAS,gCAAgC;AAC9C,yCAAiC,kCAAkC,CAAC;AACpE,iBAAS,mBAAmB;AAC1B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,mBAAmB;AAC1B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,oBAAoB;AAC3B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,oBAAoB;AAC3B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,oBAAoB;AAC3B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,oBAAoB;AAC3B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,oBAAoB;AAC3B,cAAI,WAAW,UAAU,SAAS;AAChC,uCAA2B,WAAW,MAAM;AAAA,UAC9C;AACA,iBAAO;AAAA,QACT;AACA,YAAI,SAAS,OAAO,kCAAkC,cAAc,iCAAiC,CAAC;AACtG,YAAI,qBAAqB;AACzB,eAAO,WAAW,IAAI,QAAQ,SAAS,SAAS,QAAQ;AACtD,gCAAsB;AACtB,+BAAqB;AAAA,QACvB,CAAC;AACD,YAAI;AACJ,YAAI,OAAO,YAAY,eAAe,QAAQ,WAAW;AACvD,4BAAkB,EAAE,mBAAmB,QAAQ,UAAU,mBAAmB,GAAG,oBAAoB,QAAQ,UAAU,oBAAoB,EAAE;AAAA,QAC7I;AACA,YAAI,kBAAkB,OAAO,OAAO,CAAC,GAAG,MAAM;AAC9C,YAAI,aAAa,CAAC;AAClB,YAAI,cAAc;AAClB,YAAI,QAAQ,CAAC,QAAQ,YAAY;AAC/B,gBAAM;AAAA,QACR;AACA,YAAI,qBAAqB,OAAO,UAAU;AAC1C,YAAI,wBAAwB,OAAO,iBAAiB;AACpD,YAAI,sBAAsB,OAAO,WAAW,YAAY,OAAO,QAAQ,YAAY,YAAY,OAAO,QAAQ,SAAS,QAAQ;AAC/H,YAAI,yBAAyB,OAAO,6BAA6B;AACjE,YAAI,kBAAkB;AACtB,iBAAS,WAAW,MAAM;AACxB,cAAI,OAAO,eAAe;AACxB,mBAAO,OAAO,cAAc,MAAM,eAAe;AAAA,UACnD;AACA,iBAAO,kBAAkB;AAAA,QAC3B;AACA,YAAI,OAAO,WAAW,YAAY;AAClC,iBAAS,mBAAmB,GAAG;AAC7B,cAAI,aAAa;AACf;AACF,cAAI,QAAQ;AACZ,cAAI,+BAA+B,KAAK;AAAA,QAC1C;AACA,YAAI,qBAAqB;AACvB,cAAI,uBAAuB;AACzB,8BAAkB,aAAa,EAAE,QAAQ,eAAe,IAAI;AAAA,UAC9D,OAAO;AACL,8BAAkB,YAAY;AAAA,UAChC;AACA,cAAI,IAAI;AACR,cAAI,OAAOH,eAAc,YAAY;AACnC,iBAAK,WAAW;AAChB,uBAAW,aAAa;AAAA,UAC1B;AACA,kBAAQ,CAAC,UAAU,WAAW;AAC5B,uBAAW,SAAS,aAAa,QAAQ;AACzC,mBAAO,GAAG,aAAa,UAAU,SAAS,SAAS,MAAM;AAAA,UAC3D;AACA,uBAAa,CAAC,aAAa;AACzB,gBAAI,MAAM,MAAM,UAAU,IAAI;AAC9B,gBAAI,CAAC,IAAI,QAAQ;AACf,oBAAM,IAAI,WAAW,GAAG;AAAA,YAC1B;AACA,mBAAO;AAAA,UACT;AACA,sBAAY,CAAC,UAAU,QAAQ,YAAY;AACzC,uBAAW,SAAS,aAAa,QAAQ;AACzC,eAAG,SAAS,UAAU,SAAS,MAAM,MAAM;AACzC,kBAAI;AACF,wBAAQ,IAAI;AAAA;AAEZ,uBAAO,KAAK,MAAM;AAAA,YACtB,CAAC;AAAA,UACH;AACA,cAAI,QAAQ,QAAQ,SAAS,GAAG;AAC9B,0BAAc,QAAQ,QAAQ,GAAG,QAAQ,OAAO,GAAG;AAAA,UACrD;AACA,uBAAa,QAAQ,QAAQ,MAAM,CAAC;AACpC,kBAAQ,MAAM,qBAAqB,SAAS,IAAI;AAC9C,gBAAI,EAAE,cAAc,aAAa;AAC/B,oBAAM;AAAA,YACR;AAAA,UACF,CAAC;AACD,kBAAQ,MAAM,sBAAsB,SAAS,QAAQ;AACnD,kBAAM;AAAA,UACR,CAAC;AACD,kBAAQ,CAAC,QAAQ,YAAY;AAC3B,gBAAI,iBAAiB,GAAG;AACtB,sBAAQ,cAAc;AACtB,oBAAM;AAAA,YACR;AACA,+BAAmB,OAAO;AAC1B,oBAAQ,QAAQ,MAAM;AAAA,UACxB;AACA,iBAAO,aAAa,WAAW;AAC7B,mBAAO;AAAA,UACT;AACA,cAAI;AACJ,cAAI;AACF,gCAAoB,uBAAuB;AAAA,UAC7C,SAAS,GAAP;AACA,oBAAQ,MAAM,yGAAyG;AACvH,kBAAM;AAAA,UACR;AACA,iBAAO,SAAS,kBAAkB;AAAA,QACpC,WAAW,sBAAsB,uBAAuB;AACtD,cAAI,uBAAuB;AACzB,8BAAkB,KAAK,SAAS;AAAA,UAClC,WAAW,OAAO,YAAY,eAAe,SAAS,eAAe;AACnE,8BAAkB,SAAS,cAAc;AAAA,UAC3C;AACA,cAAI,OAAO,eAAe,eAAe,YAAY;AACnD,8BAAkB;AAAA,UACpB;AACA,cAAI,gBAAgB,QAAQ,OAAO,MAAM,GAAG;AAC1C,8BAAkB,gBAAgB,OAAO,GAAG,gBAAgB,QAAQ,UAAU,EAAE,EAAE,YAAY,GAAG,IAAI,CAAC;AAAA,UACxG,OAAO;AACL,8BAAkB;AAAA,UACpB;AACA,cAAI,CAAC,qBAAqB;AACxB,oBAAQ,CAAC,QAAQ;AACf,kBAAI,MAAM,IAAI,eAAe;AAC7B,kBAAI,KAAK,OAAO,KAAK,KAAK;AAC1B,kBAAI,KAAK,IAAI;AACb,qBAAO,IAAI;AAAA,YACb;AACA,gBAAI,uBAAuB;AACzB,2BAAa,CAAC,QAAQ;AACpB,oBAAI,MAAM,IAAI,eAAe;AAC7B,oBAAI,KAAK,OAAO,KAAK,KAAK;AAC1B,oBAAI,eAAe;AACnB,oBAAI,KAAK,IAAI;AACb,uBAAO,IAAI,WAAW,IAAI,QAAQ;AAAA,cACpC;AAAA,YACF;AACA,wBAAY,CAAC,KAAK,QAAQ,YAAY;AACpC,kBAAI,MAAM,IAAI,eAAe;AAC7B,kBAAI,KAAK,OAAO,KAAK,IAAI;AACzB,kBAAI,eAAe;AACnB,kBAAI,SAAS,MAAM;AACjB,oBAAI,IAAI,UAAU,OAAO,IAAI,UAAU,KAAK,IAAI,UAAU;AACxD,yBAAO,IAAI,QAAQ;AACnB;AAAA,gBACF;AACA,wBAAQ;AAAA,cACV;AACA,kBAAI,UAAU;AACd,kBAAI,KAAK,IAAI;AAAA,YACf;AAAA,UACF;AACA,2BAAiB,CAAC,UAAU,SAAS,QAAQ;AAAA,QAC/C,OAAO;AAAA,QACP;AACA,YAAI,qBAAqB;AACvB,cAAI,OAAO,eAAe,aAAa;AACrC,mBAAO,cAAc,mBAAmB,EAAE;AAAA,UAC5C;AAAA,QACF;AACA,YAAI,eAAe,QAAQ,IAAI,KAAK,OAAO;AAC3C,YAAI,kBAAkB,QAAQ,KAAK,KAAK,OAAO;AAC/C,YAAI,qBAAqB;AACvB,yBAAe,CAAC,QAAQ,GAAG,UAAU,GAAG,MAAM,IAAI;AAClD,4BAAkB,CAAC,QAAQ,GAAG,UAAU,GAAG,MAAM,IAAI;AAAA,QACvD;AACA,YAAI,MAAM,OAAO,YAAY;AAC7B,YAAI,MAAM,OAAO,eAAe;AAChC,eAAO,OAAO,QAAQ,eAAe;AACrC,0BAAkB;AAClB,YAAI,OAAO;AACT,uBAAa,OAAO;AACtB,YAAI,OAAO;AACT,wBAAc,OAAO;AACvB,YAAI,OAAO;AACT,kBAAQ,OAAO;AACjB,YAAI,eAAe;AACnB,YAAI,eAAe,QAAQ;AAC3B,YAAI,gBAAgB,QAAQ;AAC5B,YAAI,0BAA0B,QAAQ;AACtC,YAAI;AACJ,YAAI,OAAO;AACT,uBAAa,OAAO;AACtB,YAAI,gBAAgB,OAAO,oBAAoB;AAC/C,YAAI,OAAO,eAAe,UAAU;AAClC,gBAAM,iCAAiC;AAAA,QACzC;AACA,YAAI;AACJ,YAAI;AACJ,YAAI,QAAQ;AACZ,YAAI;AACJ,iBAAS,QAAQ,WAAW,MAAM;AAChC,cAAI,CAAC,WAAW;AACd,kBAAM,IAAI;AAAA,UACZ;AAAA,QACF;AACA,YAAI,cAAc,OAAO,eAAe,cAAc,IAAI,YAAY,MAAM,IAAI;AAChF,iBAAS,kBAAkB,aAAa,KAAK,gBAAgB;AAC3D,cAAI,SAAS,MAAM;AACnB,cAAI,SAAS;AACb,iBAAO,YAAY,WAAW,EAAE,UAAU;AACxC,cAAE;AACJ,cAAI,SAAS,MAAM,MAAM,YAAY,UAAU,aAAa;AAC1D,mBAAO,YAAY,OAAO,YAAY,kBAAkB,oBAAoB,YAAY,MAAM,KAAK,MAAM,IAAI,YAAY,SAAS,KAAK,MAAM,CAAC;AAAA,UAChJ;AACA,cAAI,MAAM;AACV,iBAAO,MAAM,QAAQ;AACnB,gBAAI,KAAK,YAAY;AACrB,gBAAI,EAAE,KAAK,MAAM;AACf,qBAAO,OAAO,aAAa,EAAE;AAC7B;AAAA,YACF;AACA,gBAAI,KAAK,YAAY,SAAS;AAC9B,iBAAK,KAAK,QAAQ,KAAK;AACrB,qBAAO,OAAO,cAAc,KAAK,OAAO,IAAI,EAAE;AAC9C;AAAA,YACF;AACA,gBAAI,KAAK,YAAY,SAAS;AAC9B,iBAAK,KAAK,QAAQ,KAAK;AACrB,oBAAM,KAAK,OAAO,KAAK,MAAM,IAAI;AAAA,YACnC,OAAO;AACL,oBAAM,KAAK,MAAM,KAAK,MAAM,KAAK,MAAM,IAAI,YAAY,SAAS;AAAA,YAClE;AACA,gBAAI,KAAK,OAAO;AACd,qBAAO,OAAO,aAAa,EAAE;AAAA,YAC/B,OAAO;AACL,kBAAI,KAAK,KAAK;AACd,qBAAO,OAAO,aAAa,QAAQ,MAAM,IAAI,QAAQ,KAAK,IAAI;AAAA,YAChE;AAAA,UACF;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,aAAa,KAAK,gBAAgB;AACzC,iBAAO,MAAM,kBAAkB,iBAAiB,GAAG,KAAK,cAAc,IAAI;AAAA,QAC5E;AACA,iBAAS,kBAAkB,KAAK,MAAM,QAAQ,iBAAiB;AAC7D,cAAI,EAAE,kBAAkB;AACtB,mBAAO;AACT,cAAI,WAAW;AACf,cAAI,SAAS,SAAS,kBAAkB;AACxC,mBAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,EAAE,GAAG;AACnC,gBAAI,IAAI,IAAI,WAAW,CAAC;AACxB,gBAAI,KAAK,SAAS,KAAK,OAAO;AAC5B,kBAAI,KAAK,IAAI,WAAW,EAAE,CAAC;AAC3B,kBAAI,UAAU,IAAI,SAAS,MAAM,KAAK;AAAA,YACxC;AACA,gBAAI,KAAK,KAAK;AACZ,kBAAI,UAAU;AACZ;AACF,mBAAK,YAAY;AAAA,YACnB,WAAW,KAAK,MAAM;AACpB,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B,WAAW,KAAK,OAAO;AACrB,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,KAAK,IAAI;AAChC,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B,OAAO;AACL,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,KAAK,KAAK;AACjC,mBAAK,YAAY,MAAM,KAAK,IAAI;AAChC,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B;AAAA,UACF;AACA,eAAK,UAAU;AACf,iBAAO,SAAS;AAAA,QAClB;AACA,iBAAS,aAAa,KAAK,QAAQ,iBAAiB;AAClD,iBAAO,kBAAkB,KAAK,iBAAiB,GAAG,QAAQ,eAAe;AAAA,QAC3E;AACA,YAAI,SAAS,OAAO,QAAQ,QAAQ,SAAS,QAAQ,SAAS,SAAS;AACvE,YAAI,wBAAwB;AAC1B,oBAAU,OAAO;AAAA,QACnB;AACA,iBAAS,2BAA2B,KAAK;AACvC,oBAAU;AACV,iBAAO,WAAW,QAAQ,IAAI,UAAU,GAAG;AAC3C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,aAAa,UAAU,IAAI,YAAY,GAAG;AACjD,iBAAO,aAAa,UAAU,IAAI,YAAY,GAAG;AACjD,iBAAO,aAAa,UAAU,IAAI,aAAa,GAAG;AAClD,iBAAO,aAAa,UAAU,IAAI,aAAa,GAAG;AAAA,QACpD;AACA,YAAI,iBAAiB,OAAO,qBAAqB;AACjD,YAAI,wBAAwB;AAC1B,uBAAa,OAAO;AACpB,oBAAU,OAAO;AAAA,QACnB,OAAO;AACL,cAAI,OAAO,eAAe;AACxB,yBAAa,OAAO;AAAA,UACtB,OAAO;AACL,yBAAa,IAAI,YAAY,OAAO,EAAE,WAAW,iBAAiB,OAAO,WAAW,aAAa,OAAO,UAAU,KAAK,CAAC;AACxH,gBAAI,EAAE,WAAW,kBAAkB,oBAAoB;AACrD,kBAAI,6NAA6N;AACjO,kBAAI,qBAAqB;AACvB,wBAAQ,IAAI,mHAAmH;AAAA,cACjI;AACA,oBAAM,MAAM,YAAY;AAAA,YAC1B;AAAA,UACF;AAAA,QACF;AACA,YAAI,YAAY;AACd,oBAAU,WAAW;AAAA,QACvB;AACA,yBAAiB,QAAQ;AACzB,mCAA2B,OAAO;AAClC,YAAI;AACJ,YAAI,eAAe,CAAC;AACpB,YAAI,aAAa,CAAC;AAClB,YAAI,gBAAgB,CAAC;AACrB,YAAI,qBAAqB;AACzB,iBAAS,mBAAmB;AAC1B,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS;AAChB,cAAI,OAAO,WAAW;AACpB,gBAAI,OAAO,OAAO,aAAa;AAC7B,qBAAO,YAAY,CAAC,OAAO,SAAS;AACtC,mBAAO,OAAO,UAAU,QAAQ;AAC9B,0BAAY,OAAO,UAAU,MAAM,CAAC;AAAA,YACtC;AAAA,UACF;AACA,+BAAqB,YAAY;AAAA,QACnC;AACA,iBAAS,cAAc;AACrB,+BAAqB;AACrB,cAAI;AACF;AACF,+BAAqB,UAAU;AAAA,QACjC;AACA,iBAAS,UAAU;AACjB,cAAI;AACF;AACF,cAAI,OAAO,YAAY;AACrB,gBAAI,OAAO,OAAO,cAAc;AAC9B,qBAAO,aAAa,CAAC,OAAO,UAAU;AACxC,mBAAO,OAAO,WAAW,QAAQ;AAC/B,2BAAa,OAAO,WAAW,MAAM,CAAC;AAAA,YACxC;AAAA,UACF;AACA,+BAAqB,aAAa;AAAA,QACpC;AACA,iBAAS,YAAY,IAAI;AACvB,uBAAa,QAAQ,EAAE;AAAA,QACzB;AACA,iBAAS,UAAU,IAAI;AACrB,qBAAW,QAAQ,EAAE;AAAA,QACvB;AACA,iBAAS,aAAa,IAAI;AACxB,wBAAc,QAAQ,EAAE;AAAA,QAC1B;AACA,YAAI,kBAAkB;AACtB,YAAI,uBAAuB;AAC3B,YAAI,wBAAwB;AAC5B,iBAAS,iBAAiB,IAAI;AAC5B;AACA,cAAI,OAAO,2BAA2B;AACpC,mBAAO,0BAA0B,eAAe;AAAA,UAClD;AAAA,QACF;AACA,iBAAS,oBAAoB,IAAI;AAC/B;AACA,cAAI,OAAO,2BAA2B;AACpC,mBAAO,0BAA0B,eAAe;AAAA,UAClD;AACA,cAAI,mBAAmB,GAAG;AACxB,gBAAI,yBAAyB,MAAM;AACjC,4BAAc,oBAAoB;AAClC,qCAAuB;AAAA,YACzB;AACA,gBAAI,uBAAuB;AACzB,kBAAI,WAAW;AACf,sCAAwB;AACxB,uBAAS;AAAA,YACX;AAAA,UACF;AAAA,QACF;AACA,iBAAS,MAAM,MAAM;AACnB,cAAI,wBAAwB;AAC1B,wBAAY,EAAE,OAAO,WAAW,OAAO,KAAK,CAAC;AAAA,UAC/C,OAAO;AACL,gBAAI,OAAO,YAAY;AACrB,qBAAO,WAAW,IAAI;AAAA,YACxB;AAAA,UACF;AACA,iBAAO,aAAa,OAAO;AAC3B,cAAI,IAAI;AACR,kBAAQ;AACR,uBAAa;AACb,kBAAQ;AACR,cAAI,IAAI,IAAI,YAAY,aAAa,IAAI;AACzC,6BAAmB,CAAC;AACpB,gBAAM;AAAA,QACR;AACA,YAAI,gBAAgB;AACpB,iBAAS,UAAU,UAAU;AAC3B,iBAAO,SAAS,WAAW,aAAa;AAAA,QAC1C;AACA,iBAAS,UAAU,UAAU;AAC3B,iBAAO,SAAS,WAAW,SAAS;AAAA,QACtC;AACA,YAAI;AACJ,yBAAiB;AACjB,YAAI,CAAC,UAAU,cAAc,GAAG;AAC9B,2BAAiB,WAAW,cAAc;AAAA,QAC5C;AACA,iBAAS,UAAU,MAAM;AACvB,cAAI;AACF,gBAAI,QAAQ,kBAAkB,YAAY;AACxC,qBAAO,IAAI,WAAW,UAAU;AAAA,YAClC;AACA,gBAAI,YAAY;AACd,qBAAO,WAAW,IAAI;AAAA,YACxB;AACA,kBAAM;AAAA,UACR,SAAS,MAAP;AACA,kBAAM,IAAI;AAAA,UACZ;AAAA,QACF;AACA,iBAAS,mBAAmB;AAC1B,cAAI,CAAC,eAAe,sBAAsB,wBAAwB;AAChE,gBAAI,OAAO,SAAS,cAAc,CAAC,UAAU,cAAc,GAAG;AAC5D,qBAAO,MAAM,gBAAgB,EAAE,aAAa,cAAc,CAAC,EAAE,KAAK,SAAS,UAAU;AACnF,oBAAI,CAAC,SAAS,OAAO;AACnB,wBAAM,yCAAyC,iBAAiB;AAAA,gBAClE;AACA,uBAAO,SAAS,eAAe;AAAA,cACjC,CAAC,EAAE,MAAM,WAAW;AAClB,uBAAO,UAAU,cAAc;AAAA,cACjC,CAAC;AAAA,YACH,OAAO;AACL,kBAAI,WAAW;AACb,uBAAO,IAAI,QAAQ,SAAS,SAAS,QAAQ;AAC3C,4BAAU,gBAAgB,SAAS,UAAU;AAC3C,4BAAQ,IAAI,WAAW,QAAQ,CAAC;AAAA,kBAClC,GAAG,MAAM;AAAA,gBACX,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF;AACA,iBAAO,QAAQ,QAAQ,EAAE,KAAK,WAAW;AACvC,mBAAO,UAAU,cAAc;AAAA,UACjC,CAAC;AAAA,QACH;AACA,iBAAS,aAAa;AACpB,cAAI,OAAO,EAAE,OAAO,eAAe,0BAA0B,cAAc;AAC3E,mBAAS,gBAAgB,UAAU,SAAS;AAC1C,gBAAI,WAAW,SAAS;AACxB,mBAAO,SAAS;AAChB,4BAAgB,OAAO,OAAO,uBAAuB;AACrD,wBAAY,OAAO,OAAO;AAC1B,sBAAU,OAAO,OAAO,oBAAoB;AAC5C,yBAAa;AACb,gBAAI,CAAC,wBAAwB;AAC3B,kBAAI,mBAAmB,QAAQ,cAAc;AAC7C,sBAAQ,cAAc,QAAQ,SAAS,GAAG;AACxC,wBAAQ,uBAAuB,GAAG,WAAW;AAC3C,sBAAI,CAAC,EAAE;AACL,wCAAoB,kBAAkB;AAAA,gBAC1C,CAAC;AAAA,cACH,CAAC;AAAA,YACH;AAAA,UACF;AACA,cAAI,CAAC,wBAAwB;AAC3B,6BAAiB,kBAAkB;AAAA,UACrC;AACA,mBAAS,2BAA2B,QAAQ;AAC1C,4BAAgB,OAAO,aAAa,OAAO,SAAS;AAAA,UACtD;AACA,mBAAS,uBAAuB,UAAU;AACxC,mBAAO,iBAAiB,EAAE,KAAK,SAAS,QAAQ;AAC9C,qBAAO,YAAY,YAAY,QAAQ,IAAI;AAAA,YAC7C,CAAC,EAAE,KAAK,SAAS,UAAU;AACzB,qBAAO;AAAA,YACT,CAAC,EAAE,KAAK,UAAU,SAAS,QAAQ;AACjC,kBAAI,4CAA4C,MAAM;AACtD,oBAAM,MAAM;AAAA,YACd,CAAC;AAAA,UACH;AACA,mBAAS,mBAAmB;AAC1B,gBAAI,CAAC,cAAc,OAAO,YAAY,wBAAwB,cAAc,CAAC,UAAU,cAAc,KAAK,CAAC,UAAU,cAAc,KAAK,CAAC,uBAAuB,OAAO,SAAS,YAAY;AAC1L,qBAAO,MAAM,gBAAgB,EAAE,aAAa,cAAc,CAAC,EAAE,KAAK,SAAS,UAAU;AACnF,oBAAI,SAAS,YAAY,qBAAqB,UAAU,IAAI;AAC5D,uBAAO,OAAO,KAAK,4BAA4B,SAAS,QAAQ;AAC9D,sBAAI,oCAAoC,MAAM;AAC9C,sBAAI,2CAA2C;AAC/C,yBAAO,uBAAuB,0BAA0B;AAAA,gBAC1D,CAAC;AAAA,cACH,CAAC;AAAA,YACH,OAAO;AACL,qBAAO,uBAAuB,0BAA0B;AAAA,YAC1D;AAAA,UACF;AACA,cAAI,OAAO,oBAAoB;AAC7B,gBAAI;AACF,kBAAI,WAAW,OAAO,mBAAmB,MAAM,eAAe;AAC9D,qBAAO;AAAA,YACT,SAAS,GAAP;AACA,kBAAI,wDAAwD,CAAC;AAC7D,iCAAmB,CAAC;AAAA,YACtB;AAAA,UACF;AACA,2BAAiB,EAAE,MAAM,kBAAkB;AAC3C,iBAAO,CAAC;AAAA,QACV;AACA,YAAI;AACJ,YAAI;AACJ,YAAI,aAAa,CAAC;AAClB,iBAAS,WAAW,QAAQ;AAC1B,eAAK,OAAO;AACZ,eAAK,UAAU,kCAAkC,SAAS;AAC1D,eAAK,SAAS;AAAA,QAChB;AACA,iBAAS,WAAW,aAAa;AAC/B,cAAI,SAAS,QAAQ,SAAS;AAC9B,iBAAO,QAAQ,SAAS;AACxB,iBAAO,UAAU;AACjB,wCAA8B,WAAW;AACzC,kBAAQ,eAAe,OAAO,QAAQ,eAAe,QAAQ,MAAM,GAAG,CAAC;AACvE,iBAAO,cAAc;AAAA,QACvB;AACA,iBAAS,aAAa,aAAa;AACjC,cAAI,SAAS,QAAQ,SAAS;AAC9B,iBAAO,YAAY,EAAE,OAAO,SAAS,CAAC;AAAA,QACxC;AACA,iBAAS,cAAc,aAAa;AAClC,cAAI,SAAS,QAAQ,SAAS;AAC9B,kBAAQ,MAAM;AACd,kBAAQ,mBAAmB,MAAM;AAAA,QACnC;AACA,iBAAS,YAAY,cAAc;AACjC,cAAI,SAAS,QAAQ,aAAa;AAClC,cAAI,CAAC,QAAQ;AACX,mBAAO;AAAA,UACT;AACA,kBAAQ,eAAe,KAAK,MAAM;AAClC,kBAAQ,SAAS,aAAa,eAAe;AAC7C,iBAAO,cAAc,aAAa;AAClC,cAAI,MAAM,EAAE,OAAO,OAAO,iBAAiB,aAAa,cAAc,OAAO,aAAa,KAAK,eAAe,aAAa,YAAY;AACvI,iBAAO,aAAa,MAAM;AACxB,gBAAI,OAAO,YAAY,IAAI;AAC3B,mBAAO,YAAY,KAAK,aAAa,YAAY;AAAA,UACnD;AACA,cAAI,OAAO,QAAQ;AACjB,mBAAO,WAAW;AAClB,mBAAO,OAAO;AAAA,UAChB;AACA,iBAAO;AAAA,QACT;AACA,YAAI,WAAW,EAAE,SAAS,QAAQ,KAAK,WAAW;AAChD,mBAAS,WAAW;AACpB,cAAI,MAAM,kBAAkB,EAAE,SAAS,UAAU,KAAK;AACtD,iBAAO;AAAA,QACT,GAAG,QAAQ,SAAS,KAAK;AACvB,cAAI,MAAM,aAAa,GAAG;AAC1B,iBAAO;AAAA,QACT,EAAE;AACF,iBAAS,WAAW,MAAM;AACxB,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,IAAI;AACvD,uBAAa;AACb,cAAI,CAAC,iBAAiB,GAAG;AACvB,oBAAQ,oBAAoB;AAC5B,gBAAI,OAAO;AACT,qBAAO,UAAU,IAAI;AACvB,oBAAQ;AAAA,UACV;AACA,gBAAM,MAAM,IAAI,WAAW,IAAI,CAAC;AAAA,QAClC;AACA,iBAAS,OAAO,QAAQ,UAAU;AAChC,uBAAa;AACb,cAAI,CAAC,UAAU;AACb,gBAAI,wBAAwB;AAC1B,+BAAiB,MAAM;AACvB,oBAAM;AAAA,YACR,OAAO;AAAA,YACP;AAAA,UACF;AACA,qBAAW,MAAM;AAAA,QACnB;AACA,YAAI,QAAQ;AACZ,iBAAS,gBAAgB,GAAG;AAC1B,cAAI,aAAa,cAAc,KAAK,UAAU;AAC5C,mBAAO;AAAA,UACT;AACA,gBAAM,GAAG,CAAC;AAAA,QACZ;AACA,YAAI,UAAU,EAAE,eAAe,CAAC,GAAG,gBAAgB,CAAC,GAAG,kBAAkB,CAAC,GAAG,UAAU,CAAC,GAAG,MAAM,WAAW;AAC1G,cAAI,wBAAwB;AAC1B,oBAAQ,WAAW;AAAA,UACrB,OAAO;AACL,oBAAQ,eAAe;AAAA,UACzB;AAAA,QACF,GAAG,gBAAgB,WAAW;AAC5B,cAAI,kBAAkB;AACtB,iBAAO,mBAAmB;AACxB,oBAAQ,qBAAqB;AAAA,UAC/B;AAAA,QACF,GAAG,YAAY,WAAW;AACxB,0BAAgB;AAAA,QAClB,GAAG,eAAe,SAAS,QAAQ;AACjC,uBAAa;AAAA,QACf,GAAG,qBAAqB,WAAW;AACjC,mBAAS,UAAU,OAAO,OAAO,QAAQ,QAAQ,GAAG;AAClD,oBAAQ,mBAAmB,MAAM;AAAA,UACnC;AACA,mBAAS,UAAU,QAAQ,eAAe;AACxC,mBAAO,UAAU;AAAA,UACnB;AACA,kBAAQ,gBAAgB,CAAC;AAAA,QAC3B,GAAG,oBAAoB,SAAS,QAAQ;AACtC,cAAI,cAAc,OAAO;AACzB,iBAAO,QAAQ,SAAS;AACxB,kBAAQ,cAAc,KAAK,MAAM;AACjC,kBAAQ,eAAe,OAAO,QAAQ,eAAe,QAAQ,MAAM,GAAG,CAAC;AACvE,iBAAO,cAAc;AACrB,wCAA8B,WAAW;AAAA,QAC3C,GAAG,uBAAuB,SAAS,MAAM;AAAA,QACzC,GAAG,eAAe,WAAW;AAC3B,kBAAQ,iBAAiB,QAAQ,CAAC,MAAM,EAAE,CAAC;AAAA,QAC7C,GAAG,wBAAwB,SAAS,QAAQ,mBAAmB;AAC7D,iBAAO,YAAY,CAAC,MAAM;AACxB,gBAAI,IAAI,EAAE;AACV,gBAAI,MAAM,EAAE;AACZ,gBAAI,OAAO;AACT,sBAAQ,sCAAsC,OAAO;AACvD,gBAAI,EAAE,mBAAmB,EAAE,mBAAmB,cAAc,GAAG;AAC7D,kBAAI,eAAe,QAAQ,SAAS,EAAE;AACtC,kBAAI,cAAc;AAChB,6BAAa,YAAY,GAAG,EAAE,eAAe;AAAA,cAC/C,OAAO;AACL,oBAAI,4CAA4C,MAAM,yBAAyB,EAAE,kBAAkB,qCAAqC;AAAA,cAC1I;AACA,sBAAQ,sCAAsC;AAC9C;AAAA,YACF;AACA,gBAAI,QAAQ,wBAAwB;AAClC,2CAA6B,EAAE,QAAQ;AAAA,YACzC,WAAW,QAAQ,eAAe;AAChC,0BAAY,CAAC;AAAA,YACf,WAAW,QAAQ,iBAAiB;AAClC,4BAAc,EAAE,SAAS;AAAA,YAC3B,WAAW,QAAQ,cAAc;AAC/B,yBAAW,EAAE,SAAS;AAAA,YACxB,WAAW,QAAQ,gBAAgB;AACjC,2BAAa,EAAE,SAAS;AAAA,YAC1B,WAAW,QAAQ,UAAU;AAC3B,qBAAO,SAAS;AAChB,kBAAI;AACF,kCAAkB,MAAM;AAC1B,kBAAI,OAAO,YAAY;AACrB,uBAAO,WAAW;AAClB,uBAAO,OAAO;AAAA,cAChB;AAAA,YACF,WAAW,QAAQ,SAAS;AAC1B,kBAAI,YAAY,EAAE,cAAc,OAAO,EAAE,OAAO;AAAA,YAClD,WAAW,QAAQ,YAAY;AAC7B,kBAAI,YAAY,EAAE,cAAc,OAAO,EAAE,OAAO;AAAA,YAClD,WAAW,QAAQ,SAAS;AAC1B,oBAAM,YAAY,EAAE,cAAc,OAAO,EAAE,OAAO;AAAA,YACpD,WAAW,EAAE,WAAW,gBAAgB;AACtC,qBAAO,YAAY,CAAC;AAAA,YACtB,WAAW,QAAQ,WAAW;AAC5B,kBAAI,OAAO,YAAY;AACrB,uBAAO,WAAW,EAAE,MAAM;AAAA,cAC5B;AAAA,YACF,WAAW,KAAK;AACd,kBAAI,oCAAoC,GAAG;AAAA,YAC7C;AACA,oBAAQ,sCAAsC;AAAA,UAChD;AACA,iBAAO,UAAU,CAAC,MAAM;AACtB,gBAAI,UAAU;AACd,gBAAI,UAAU,MAAM,EAAE,WAAW,MAAM,EAAE,SAAS,OAAO,EAAE,OAAO;AAClE,kBAAM;AAAA,UACR;AACA,cAAI,qBAAqB;AACvB,mBAAO,GAAG,WAAW,SAAS,MAAM;AAClC,qBAAO,UAAU,EAAE,KAAK,CAAC;AAAA,YAC3B,CAAC;AACD,mBAAO,GAAG,SAAS,SAAS,GAAG;AAC7B,qBAAO,QAAQ,CAAC;AAAA,YAClB,CAAC;AACD,mBAAO,GAAG,gBAAgB,WAAW;AAAA,YACrC,CAAC;AAAA,UACH;AACA,iBAAO,YAAY,EAAE,OAAO,QAAQ,aAAa,OAAO,0BAA0B,YAAY,cAAc,YAAY,cAAc,WAAW,CAAC;AAAA,QACpJ,GAAG,sBAAsB,WAAW;AAClC,cAAI,gBAAgB,WAAW,2CAA2C;AAC1E,kBAAQ,cAAc,KAAK,IAAI,OAAO,aAAa,CAAC;AAAA,QACtD,GAAG,cAAc,WAAW;AAC1B,cAAI,QAAQ,cAAc,UAAU,GAAG;AACrC,oBAAQ,qBAAqB;AAC7B,oBAAQ,uBAAuB,QAAQ,cAAc,EAAE;AAAA,UACzD;AACA,iBAAO,QAAQ,cAAc,IAAI;AAAA,QACnC,EAAE;AACF,eAAO,aAAa;AACpB,iBAAS,qBAAqB,YAAY;AACxC,iBAAO,WAAW,SAAS,GAAG;AAC5B,uBAAW,MAAM,EAAE,MAAM;AAAA,UAC3B;AAAA,QACF;AACA,iBAAS,cAAc,GAAG;AACxB,cAAI,SAAS,UAAU;AACvB,cAAI,MAAM,EAAE;AACZ,uBAAa,MAAM;AACnB,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS,OAAO;AACvB,iBAAO;AAAA,QACT;AACA,iBAAS,YAAY,MAAM;AACzB,cAAI,QAAQ;AACZ,iBAAO,KAAK,QAAQ,OAAO,SAAS,GAAG;AACrC,gBAAI,IAAI,SAAS,CAAC;AAClB,mBAAO,MAAM,IAAI,IAAI,IAAI,OAAO,IAAI;AAAA,UACtC,CAAC;AAAA,QACH;AACA,iBAAS,sBAAsB;AAC7B,cAAI,cAAc,cAAc;AAChC,cAAI,WAAW,kBAAkB,EAAE,cAAc,MAAM;AACvD,cAAI,YAAY,kBAAkB,EAAE,cAAc,MAAM;AACxD,cAAI,WAAW,WAAW;AAC1B,uCAA6B,UAAU,QAAQ;AAC/C,uBAAa,QAAQ;AAAA,QACvB;AACA,eAAO,yBAAyB;AAChC,iBAAS,iBAAiB,YAAY;AACpC,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,UAAU;AAC7D,cAAI;AACF,kBAAM,UAAU;AAAA,UAClB,SAAS,GAAP;AACA,4BAAgB,CAAC;AAAA,UACnB;AAAA,QACF;AACA,YAAI,kBAAkB,CAAC;AACvB,iBAAS,kBAAkB,SAAS;AAClC,cAAI,QAAQ,gBAAgB;AAC5B,cAAI,CAAC,OAAO;AACV,gBAAI,WAAW,gBAAgB;AAC7B,8BAAgB,SAAS,UAAU;AACrC,4BAAgB,WAAW,QAAQ,UAAU,IAAI,OAAO;AAAA,UAC1D;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,iBAAiB,KAAK,KAAK;AAClC,cAAI,SAAS,kBAAkB,GAAG,EAAE,GAAG;AACvC,cAAI,iBAAiB,GAAG;AACtB,oBAAQ,cAAc,MAAM;AAAA,UAC9B,OAAO;AACL,qCAAyB,MAAM;AAAA,UACjC;AAAA,QACF;AACA,eAAO,sBAAsB;AAC7B,iBAAS,eAAe;AACtB,cAAI,QAAQ,IAAI,MAAM;AACtB,cAAI,CAAC,MAAM,OAAO;AAChB,gBAAI;AACF,oBAAM,IAAI,MAAM;AAAA,YAClB,SAAS,GAAP;AACA,sBAAQ;AAAA,YACV;AACA,gBAAI,CAAC,MAAM,OAAO;AAChB,qBAAO;AAAA,YACT;AAAA,UACF;AACA,iBAAO,MAAM,MAAM,SAAS;AAAA,QAC9B;AACA,iBAAS,gBAAgB,aAAa;AACpC,kBAAQ,iBAAiB,KAAK,WAAW;AAAA,QAC3C;AACA,iBAAS,mBAAmB,QAAQ,SAAS;AAC3C,2BAAiB,EAAE,IAAI,QAAQ,OAAO;AAAA,QACxC;AACA,iBAAS,kCAAkC,IAAI;AAC7C,mCAAyB,IAAI,CAAC,uBAAuB,GAAG,CAAC,kBAAkB;AAC3E,kBAAQ,cAAc;AAAA,QACxB;AACA,iBAAS,6BAA6B,QAAQ;AAC5C,cAAI,CAAC;AACH,0BAAc,MAAM;AAAA;AAEpB,wBAAY,EAAE,OAAO,iBAAiB,UAAU,OAAO,CAAC;AAAA,QAC5D;AACA,iBAAS,qBAAqB,aAAa,MAAM,cAAc,KAAK;AAClE,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,aAAa,MAAM,cAAc,GAAG;AACvF,iBAAO,qBAAqB,aAAa,MAAM,cAAc,GAAG;AAAA,QAClE;AACA,iBAAS,qBAAqB,aAAa,MAAM,cAAc,KAAK;AAClE,cAAI,OAAO,qBAAqB,aAAa;AAC3C,gBAAI,qFAAqF;AACzF,mBAAO;AAAA,UACT;AACA,cAAI,eAAe,CAAC;AACpB,cAAI,QAAQ;AACZ,cAAI,2BAA2B,aAAa,WAAW,KAAK,QAAQ;AAClE,mBAAO,qBAAqB,aAAa,MAAM,cAAc,GAAG;AAAA,UAClE;AACA,cAAI;AACF,mBAAO;AACT,cAAI,eAAe,EAAE,cAAc,aAAa,KAAK,aAAa;AAClE,cAAI,wBAAwB;AAC1B,yBAAa,MAAM;AACnB,wBAAY,cAAc,YAAY;AACtC,mBAAO;AAAA,UACT;AACA,iBAAO,YAAY,YAAY;AAAA,QACjC;AACA,iBAAS,0CAA0C;AACjD,iBAAO;AAAA,QACT;AACA,YAAI,iBAAiB;AACrB,iBAAS,oCAAoC;AAC3C,iBAAO;AAAA,QACT;AACA,iBAAS,6BAA6B,OAAO;AAC3C,kBAAQ,MAAM,kBAAkB,GAAG,SAAS,GAAG,CAAC;AAChD,cAAI,cAAc,GAAG;AACnB,kDAAsC,KAAK;AAAA,UAC7C;AACA,kBAAQ,gBAAgB,kBAAkB,GAAG,SAAS,GAAG,GAAG,CAAC;AAAA,QAC/D;AACA,eAAO,kCAAkC;AACzC,iBAAS,+BAA+B,gBAAgB,cAAc,cAAc,OAAO;AACzF,cAAI,kBAAkB,cAAc;AAClC,uBAAW,MAAM,6BAA6B,KAAK,CAAC;AAAA,UACtD,WAAW,wBAAwB;AACjC,wBAAY,EAAE,gBAAgB,gBAAgB,OAAO,wBAAwB,SAAS,MAAM,CAAC;AAAA,UAC/F,OAAO;AACL,gBAAI,SAAS,QAAQ,SAAS;AAC9B,gBAAI,CAAC,QAAQ;AACX;AAAA,YACF;AACA,mBAAO,YAAY,EAAE,OAAO,wBAAwB,SAAS,MAAM,CAAC;AAAA,UACtE;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,sCAAsC,QAAQ,OAAO,QAAQ;AACpE,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS;AAChB,gBAAM,EAAE;AAAA,QACV;AACA,iBAAS,SAAS,MAAM;AACtB,cAAI,CAAC,SAAS;AACZ,qBAAS,QAAQ,CAAC;AACpB,cAAI,CAAC,SAAS,MAAM,OAAO;AACzB,qBAAS,MAAM,QAAQ;AACvB,gBAAI;AACF,qBAAO,cAAc;AACvB,gBAAI,IAAI;AAAA,UACV;AAAA,QACF;AACA,iBAAS,qCAAqC;AAC5C,cAAI;AACF;AACF,cAAI;AACF;AACF,mBAAS,0IAA0I;AAAA,QACrJ;AACA,iBAAS,uBAAuB;AAC9B,iBAAO,KAAK,IAAI;AAAA,QAClB;AACA,iBAAS,aAAa;AACpB,iBAAO;AAAA,QACT;AACA,iBAAS,2BAA2B;AAClC,iBAAO,WAAW;AAAA,QACpB;AACA,YAAI;AACJ,YAAI,qBAAqB;AACvB,gCAAsB,MAAM;AAC1B,gBAAI,IAAI,QAAQ,UAAU;AAC1B,mBAAO,EAAE,KAAK,MAAM,EAAE,KAAK;AAAA,UAC7B;AAAA,QACF,WAAW,wBAAwB;AACjC,gCAAsB,MAAM,YAAY,IAAI,IAAI,OAAO;AAAA,QACzD;AACE,gCAAsB,MAAM,YAAY,IAAI;AAC9C,iBAAS,uBAAuB,MAAM,KAAK,KAAK;AAC9C,2BAAiB,EAAE,WAAW,MAAM,KAAK,MAAM,GAAG;AAAA,QACpD;AACA,iBAAS,gCAAgC;AACvC,cAAI;AACF,mBAAO,WAAW,EAAE,KAAK,EAAE;AAC7B,iBAAO,UAAU;AAAA,QACnB;AACA,iBAAS,oCAAoC,OAAO,MAAM;AACxD,cAAI,cAAc,UAAU,SAAS;AACrC,cAAI,YAAY;AAChB,iBAAO,cAAc,MAAM;AACzB,gBAAI,wBAAwB;AAC5B,gBAAI,OAAO,WAAW,wBAAwB,CAAC;AAC/C,gBAAI,IAAI,QAAQ;AAChB,qBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAI,MAAM,UAAU,IAAI;AACxB,gCAAkB,EAAE,IAAI,KAAK;AAAA,YAC/B;AACA,mBAAO,0CAA0C,OAAO,uBAAuB,MAAM,IAAI;AAAA,UAC3F,CAAC;AAAA,QACH;AACA,YAAI,iDAAiD,CAAC;AACtD,iBAAS,sCAAsC,OAAO,aAAa,MAAM;AACvE,yDAA+C,SAAS;AACxD,cAAI,IAAI,QAAQ;AAChB,mBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,2DAA+C,KAAK,kBAAkB,EAAE,IAAI;AAAA,UAC9E;AACA,cAAI,eAAe,QAAQ;AAC3B,cAAI,QAAQ,CAAC,eAAe,qBAAqB,SAAS,WAAW,CAAC,QAAQ;AAC9E,iBAAO,MAAM,MAAM,MAAM,8CAA8C;AAAA,QACzE;AACA,iBAAS,0BAA0B,MAAM;AACvC,cAAI;AACF,uBAAW,KAAK,OAAO,QAAQ,aAAa,UAAU,EAAE;AACxD,uCAA2B,WAAW,MAAM;AAC5C,mBAAO;AAAA,UACT,SAAS,GAAP;AAAA,UACF;AAAA,QACF;AACA,iBAAS,wBAAwB,eAAe;AAC9C,cAAI,UAAU,iBAAiB,EAAE;AACjC,0BAAgB,kBAAkB;AAClC,cAAI,iBAAiB,SAAS;AAC5B,mBAAO;AAAA,UACT;AACA,cAAI,cAAc,WAAW;AAC7B,cAAI,gBAAgB,aAAa;AAC/B,mBAAO;AAAA,UACT;AACA,cAAI,UAAU,CAAC,GAAG,aAAa,KAAK,WAAW,IAAI,YAAY;AAC/D,mBAAS,UAAU,GAAG,WAAW,GAAG,WAAW,GAAG;AAChD,gBAAI,oBAAoB,WAAW,IAAI,MAAM;AAC7C,gCAAoB,KAAK,IAAI,mBAAmB,gBAAgB,SAAS;AACzE,gBAAI,UAAU,KAAK,IAAI,aAAa,QAAQ,KAAK,IAAI,eAAe,iBAAiB,GAAG,KAAK,CAAC;AAC9F,gBAAI,cAAc,0BAA0B,OAAO;AACnD,gBAAI,aAAa;AACf,qBAAO;AAAA,YACT;AAAA,UACF;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,sCAAsC;AAC7C,gBAAM;AAAA,QACR;AACA,iBAAS,UAAU,IAAI;AACrB,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,EAAE;AACrD,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS,IAAI,YAAY,aAAa,QAAQ,WAAW;AAChE,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,IAAI,YAAY,aAAa,QAAQ,SAAS;AACjG,iBAAO;AAAA,QACT;AACA,YAAI,mBAAmB,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;AACpC,iBAAS,UAAU,QAAQ,MAAM;AAC/B,cAAI,UAAU,iBAAiB;AAC/B,cAAI,SAAS,KAAK,SAAS,IAAI;AAC7B,aAAC,WAAW,IAAI,MAAM,KAAK,kBAAkB,SAAS,CAAC,CAAC;AACxD,oBAAQ,SAAS;AAAA,UACnB,OAAO;AACL,oBAAQ,KAAK,IAAI;AAAA,UACnB;AAAA,QACF;AACA,iBAAS,UAAU,IAAI,KAAK,QAAQ,MAAM;AACxC,cAAI;AACF,mBAAO,oCAAoC,GAAG,GAAG,IAAI,KAAK,QAAQ,IAAI;AACxE,cAAI,MAAM;AACV,mBAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,gBAAI,MAAM,kBAAkB,EAAE,OAAO;AACrC,gBAAI,MAAM,kBAAkB,EAAE,MAAM,KAAK;AACzC,mBAAO;AACP,qBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,wBAAU,IAAI,iBAAiB,EAAE,MAAM,EAAE;AAAA,YAC3C;AACA,mBAAO;AAAA,UACT;AACA,4BAAkB,EAAE,QAAQ,KAAK;AACjC,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS,OAAO;AACvB,cAAI,QAAQ,OAAO,MAAM;AACzB,iBAAO;AAAA,QACT;AACA,iBAAS,MAAM,OAAO,YAAY,UAAU,MAAM,MAAM;AACtD,cAAI,MAAM,EAAE,UAAU,CAAC,QAAQ;AAC7B,gBAAI,OAAO;AACX,gBAAI,QAAQ,QAAQ,QAAQ,UAAU,QAAQ,GAAG;AAC/C,kBAAI,OAAO,IAAI,UAAU,KAAK;AAC9B,qBAAO,WAAW,GAAG;AACrB,2BAAa,KAAK,MAAM,GAAG;AAAA,YAC7B;AACA,mBAAO;AAAA,UACT,GAAG,SAAS,CAAC,QAAQ;AACnB,gBAAI,OAAO,WAAW,IAAI,MAAM;AAChC,+BAAmB,KAAK,IAAI;AAC5B,mBAAO;AAAA,UACT,EAAE;AACF,mBAAS,mBAAmB,MAAM;AAChC,gBAAI,eAAe,UAAU;AAC3B,qBAAO,aAAa,IAAI;AAAA,YAC1B;AACA,gBAAI,eAAe;AACjB,qBAAO,QAAQ,IAAI;AACrB,mBAAO;AAAA,UACT;AACA,cAAI,QAAQ,SAAS,KAAK;AAC1B,cAAI,QAAQ,CAAC;AACb,cAAI,SAAS;AACb,cAAI,MAAM;AACR,qBAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,kBAAI,YAAY,IAAI,SAAS;AAC7B,kBAAI,WAAW;AACb,oBAAI,WAAW;AACb,2BAAS,UAAU;AACrB,sBAAM,KAAK,UAAU,KAAK,EAAE;AAAA,cAC9B,OAAO;AACL,sBAAM,KAAK,KAAK;AAAA,cAClB;AAAA,YACF;AAAA,UACF;AACA,cAAI,MAAM,MAAM,MAAM,MAAM,KAAK;AACjC,mBAAS,OAAO,MAAM;AACpB,gBAAI,WAAW;AACb,2BAAa,MAAM;AACrB,mBAAO,mBAAmB,IAAI;AAAA,UAChC;AACA,gBAAM,OAAO,GAAG;AAChB,iBAAO;AAAA,QACT;AACA,iBAAS,MAAM,OAAO,YAAY,UAAU,MAAM;AAChD,qBAAW,YAAY,CAAC;AACxB,cAAI,cAAc,SAAS,MAAM,CAAC,SAAS,SAAS,YAAY,SAAS,SAAS;AAClF,cAAI,aAAa,eAAe;AAChC,cAAI,cAAc,eAAe,CAAC,MAAM;AACtC,mBAAO,SAAS,KAAK;AAAA,UACvB;AACA,iBAAO,WAAW;AAChB,mBAAO,MAAM,OAAO,YAAY,UAAU,WAAW,IAAI;AAAA,UAC3D;AAAA,QACF;AACA,gBAAQ,KAAK;AACb,YAAI,uBAAuB,CAAC,MAAM,YAAY,kBAAkB,sBAAsB,WAAW,UAAU,SAAS;AACpH,YAAI,gBAAgB,EAAE,oCAAoC,mCAAmC,+BAA+B,8BAA8B,uBAAuB,sBAAsB,0CAA0C,yCAAyC,oCAAoC,mCAAmC,iCAAiC,gCAAgC,wCAAwC,uCAAuC,SAAS,QAAQ,qCAAqC,oCAAoC,uBAAuB,sBAAsB,2BAA2B,0BAA0B,sBAAsB,qBAAqB,yBAAyB,wBAAwB,gCAAgC,+BAA+B,wCAAwC,uCAAuC,0BAA0B,yBAAyB,sCAAsC,qCAAqC,QAAQ,OAAO,YAAY,WAAW,WAAW,UAAU,YAAY,WAAW,UAAU,cAAc,OAAO,cAAc;AACppC,YAAI,MAAM,WAAW;AACrB,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,2BAA2B,OAAO,8BAA8B,WAAW;AAC7E,kBAAQ,2BAA2B,OAAO,8BAA8B,OAAO,OAAO,4BAA4B,MAAM,MAAM,SAAS;AAAA,QACzI;AACA,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,mBAAmB,OAAO,sBAAsB,WAAW;AAC7D,kBAAQ,mBAAmB,OAAO,sBAAsB,OAAO,OAAO,oBAAoB,MAAM,MAAM,SAAS;AAAA,QACjH;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,yBAAyB,OAAO,4BAA4B,WAAW;AACzE,kBAAQ,yBAAyB,OAAO,4BAA4B,OAAO,OAAO,0BAA0B,MAAM,MAAM,SAAS;AAAA,QACnI;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,kBAAkB,OAAO,qBAAqB,WAAW;AAC3D,kBAAQ,kBAAkB,OAAO,qBAAqB,OAAO,OAAO,mBAAmB,MAAM,MAAM,SAAS;AAAA,QAC9G;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,wBAAwB,OAAO,2BAA2B,WAAW;AACvE,kBAAQ,wBAAwB,OAAO,2BAA2B,OAAO,OAAO,yBAAyB,MAAM,MAAM,SAAS;AAAA,QAChI;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,kBAAkB,OAAO,qBAAqB,WAAW;AAC3D,kBAAQ,kBAAkB,OAAO,qBAAqB,OAAO,OAAO,mBAAmB,MAAM,MAAM,SAAS;AAAA,QAC9G;AACA,YAAI,yBAAyB,OAAO,4BAA4B,WAAW;AACzE,kBAAQ,yBAAyB,OAAO,4BAA4B,OAAO,OAAO,0BAA0B,MAAM,MAAM,SAAS;AAAA,QACnI;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,oBAAoB,OAAO,uBAAuB,WAAW;AAC/D,kBAAQ,oBAAoB,OAAO,uBAAuB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QACpH;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,0BAA0B,OAAO,6BAA6B,WAAW;AAC3E,kBAAQ,0BAA0B,OAAO,6BAA6B,OAAO,OAAO,2BAA2B,MAAM,MAAM,SAAS;AAAA,QACtI;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,wBAAwB,OAAO,2BAA2B,WAAW;AACvE,kBAAQ,wBAAwB,OAAO,2BAA2B,OAAO,OAAO,yBAAyB,MAAM,MAAM,SAAS;AAAA,QAChI;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,oBAAoB,OAAO,uBAAuB,WAAW;AAC/D,kBAAQ,oBAAoB,OAAO,uBAAuB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QACpH;AACA,YAAI,2BAA2B,OAAO,8BAA8B,WAAW;AAC7E,kBAAQ,2BAA2B,OAAO,8BAA8B,OAAO,OAAO,4BAA4B,MAAM,MAAM,SAAS;AAAA,QACzI;AACA,YAAI,8BAA8B,OAAO,iCAAiC,WAAW;AACnF,kBAAQ,8BAA8B,OAAO,iCAAiC,OAAO,OAAO,+BAA+B,MAAM,MAAM,SAAS;AAAA,QAClJ;AACA,YAAI,+CAA+C,OAAO,kDAAkD,WAAW;AACrH,kBAAQ,+CAA+C,OAAO,kDAAkD,OAAO,OAAO,gDAAgD,MAAM,MAAM,SAAS;AAAA,QACrM;AACA,YAAI,qCAAqC,OAAO,wCAAwC,WAAW;AACjG,kBAAQ,qCAAqC,OAAO,wCAAwC,OAAO,OAAO,sCAAsC,MAAM,MAAM,SAAS;AAAA,QACvK;AACA,YAAI,4CAA4C,OAAO,+CAA+C,WAAW;AAC/G,kBAAQ,4CAA4C,OAAO,+CAA+C,OAAO,OAAO,6CAA6C,MAAM,MAAM,SAAS;AAAA,QAC5L;AACA,YAAI,kCAAkC,OAAO,qCAAqC,WAAW;AAC3F,kBAAQ,kCAAkC,OAAO,qCAAqC,OAAO,OAAO,mCAAmC,MAAM,MAAM,SAAS;AAAA,QAC9J;AACA,YAAI,wCAAwC,OAAO,2CAA2C,WAAW;AACvG,kBAAQ,wCAAwC,OAAO,2CAA2C,OAAO,OAAO,yCAAyC,MAAM,MAAM,SAAS;AAAA,QAChL;AACA,YAAI,gCAAgC,OAAO,mCAAmC,WAAW;AACvF,kBAAQ,gCAAgC,OAAO,mCAAmC,OAAO,OAAO,iCAAiC,MAAM,MAAM,SAAS;AAAA,QACxJ;AACA,YAAI,2BAA2B,OAAO,8BAA8B,WAAW;AAC7E,kBAAQ,2BAA2B,OAAO,8BAA8B,OAAO,OAAO,4BAA4B,MAAM,MAAM,SAAS;AAAA,QACzI;AACA,YAAI,+BAA+B,OAAO,kCAAkC,WAAW;AACrF,kBAAQ,+BAA+B,OAAO,kCAAkC,OAAO,OAAO,gCAAgC,MAAM,MAAM,SAAS;AAAA,QACrJ;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC7F;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACtG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAChG;AACA,YAAI,mBAAmB,OAAO,sBAAsB,WAAW;AAC7D,kBAAQ,mBAAmB,OAAO,sBAAsB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QAClH;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACtG;AACA,eAAO,sBAAsB;AAC7B,eAAO,gBAAgB;AACvB,eAAO,WAAW;AAClB,eAAO,gBAAgB;AACvB,eAAO,aAAa;AACpB,YAAI;AACJ,gCAAwB,SAAS,YAAY;AAC3C,cAAI,CAAC;AACH,gBAAI;AACN,cAAI,CAAC;AACH,oCAAwB;AAAA,QAC5B;AACA,iBAAS,IAAI,MAAM;AACjB,iBAAO,QAAQ;AACf,cAAI,kBAAkB,GAAG;AACvB;AAAA,UACF;AACA,cAAI,wBAAwB;AAC1B,gCAAoB,MAAM;AAC1B,wBAAY;AACZ,wBAAY,EAAE,OAAO,SAAS,CAAC;AAC/B;AAAA,UACF;AACA,iBAAO;AACP,cAAI,kBAAkB,GAAG;AACvB;AAAA,UACF;AACA,mBAAS,QAAQ;AACf,gBAAI;AACF;AACF,wBAAY;AACZ,mBAAO,eAAe;AACtB,gBAAI;AACF;AACF,wBAAY;AACZ,gCAAoB,MAAM;AAC1B,gBAAI,OAAO;AACT,qBAAO,wBAAwB;AACjC,oBAAQ;AAAA,UACV;AACA,cAAI,OAAO,cAAc;AACvB,mBAAO,aAAa,YAAY;AAChC,uBAAW,WAAW;AACpB,yBAAW,WAAW;AACpB,uBAAO,aAAa,EAAE;AAAA,cACxB,GAAG,CAAC;AACJ,oBAAM;AAAA,YACR,GAAG,CAAC;AAAA,UACN,OAAO;AACL,kBAAM;AAAA,UACR;AAAA,QACF;AACA,YAAI,OAAO,YAAY;AACrB,cAAI,OAAO,OAAO,cAAc;AAC9B,mBAAO,aAAa,CAAC,OAAO,UAAU;AACxC,iBAAO,OAAO,WAAW,SAAS,GAAG;AACnC,mBAAO,WAAW,IAAI,EAAE;AAAA,UAC1B;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACJ,YAAI,iBAAiB;AACnB,2BAAiB,EAAE,mBAAmB,QAAQ,UAAU,mBAAmB,EAAE,OAAO,SAAS,UAAU;AACrG,mBAAO,CAAC,gBAAgB,kBAAkB,QAAQ,QAAQ,IAAI;AAAA,UAChE,CAAC,GAAG,oBAAoB,QAAQ,UAAU,oBAAoB,EAAE,OAAO,SAAS,UAAU;AACxF,mBAAO,CAAC,gBAAgB,mBAAmB,QAAQ,QAAQ,IAAI;AAAA,UACjE,CAAC,EAAE;AAAA,QACL;AACA,YAAI;AACJ,YAAI,OAAO,sBAAsB,aAAa;AAC5C,yBAAe;AAAA,QACjB,WAAW,OAAO,mCAAmC,aAAa;AAChE,yBAAe;AAAA,QACjB,OAAO;AACL,gBAAM,IAAI,MAAM,uCAAuC;AAAA,QACzD;AACA,YAAI,gBAAgB;AAClB,cAAI,aAAa,aAAa;AAC9B,uBAAa,cAAc,WAAW;AACpC,uBAAW;AACX,2BAAe,kBAAkB,QAAQ,SAAS,UAAU;AAC1D,sBAAQ,eAAe,qBAAqB,QAAQ;AAAA,YACtD,CAAC;AACD,2BAAe,mBAAmB,QAAQ,SAAS,UAAU;AAC3D,sBAAQ,eAAe,sBAAsB,QAAQ;AAAA,YACvD,CAAC;AAAA,UACH;AAAA,QACF;AACA,eAAO,+BAA+B;AAAA,MACxC;AAAA,IACF,GAAG;AACH,QAAI,OAAO,YAAY,YAAY,OAAO,WAAW;AACnD,aAAO,UAAU;AAAA,aACV,OAAO,WAAW,cAAc,OAAO;AAC9C,aAAO,CAAC,GAAG,WAAW;AACpB,eAAO;AAAA,MACT,CAAC;AAAA,aACM,OAAO,YAAY;AAC1B,cAAQ,mCAAmC;AAAA,EAC/C;AACF,CAAC;AAGD,IAAI,iDAAiD,WAAW;AAAA,EAC9D,gGAAgG,SAAS,QAAQ;AAC/G,WAAO,QAAQ,qBAAqB;AAAA;AAAA,EAEtC;AACF,CAAC;AAGD,IAAI,4BAA4B,WAAW;AAAA,EACzC,2EAA2E,SAAS,QAAQ;AAC1F,QAAI,sBAAsB,MAAM;AAC9B,UAAI,aAAa,OAAO,aAAa,eAAe,SAAS,gBAAgB,SAAS,cAAc,MAAM;AAC1G,UAAI,OAAO,eAAe;AACxB,qBAAa,cAAc;AAC7B,aAAO,SAAS,oBAAoB;AAClC,6BAAqB,sBAAsB,CAAC;AAC5C,YAAI,SAAS,OAAO,sBAAsB,cAAc,qBAAqB,CAAC;AAC9E,YAAI,qBAAqB;AACzB,eAAO,WAAW,IAAI,QAAQ,SAAS,SAAS,QAAQ;AACtD,gCAAsB;AACtB,+BAAqB;AAAA,QACvB,CAAC;AACD,YAAI;AACJ,YAAI,OAAO,YAAY,eAAe,QAAQ,WAAW;AACvD,4BAAkB,EAAE,mBAAmB,QAAQ,UAAU,mBAAmB,GAAG,oBAAoB,QAAQ,UAAU,oBAAoB,EAAE;AAAA,QAC7I;AACA,YAAI,kBAAkB,OAAO,OAAO,CAAC,GAAG,MAAM;AAC9C,YAAI,aAAa,CAAC;AAClB,YAAI,cAAc;AAClB,YAAI,QAAQ,CAAC,QAAQ,YAAY;AAC/B,gBAAM;AAAA,QACR;AACA,YAAI,qBAAqB,OAAO,UAAU;AAC1C,YAAI,wBAAwB,OAAO,iBAAiB;AACpD,YAAI,sBAAsB,OAAO,WAAW,YAAY,OAAO,QAAQ,YAAY,YAAY,OAAO,QAAQ,SAAS,QAAQ;AAC/H,YAAI,kBAAkB;AACtB,iBAAS,WAAW,MAAM;AACxB,cAAI,OAAO,eAAe;AACxB,mBAAO,OAAO,cAAc,MAAM,eAAe;AAAA,UACnD;AACA,iBAAO,kBAAkB;AAAA,QAC3B;AACA,YAAI,OAAO,WAAW,YAAY;AAClC,iBAAS,mBAAmB,GAAG;AAC7B,cAAI,aAAa;AACf;AACF,cAAI,QAAQ;AACZ,cAAI,+BAA+B,KAAK;AAAA,QAC1C;AACA,YAAI,qBAAqB;AACvB,cAAI,uBAAuB;AACzB,8BAAkB,aAAa,EAAE,QAAQ,eAAe,IAAI;AAAA,UAC9D,OAAO;AACL,8BAAkB,YAAY;AAAA,UAChC;AACA,cAAI,IAAI;AACR,cAAI,OAAOA,eAAc,YAAY;AACnC,iBAAK,WAAW;AAChB,uBAAW,aAAa;AAAA,UAC1B;AACA,kBAAQ,CAAC,UAAU,WAAW;AAC5B,uBAAW,SAAS,aAAa,QAAQ;AACzC,mBAAO,GAAG,aAAa,UAAU,SAAS,SAAS,MAAM;AAAA,UAC3D;AACA,uBAAa,CAAC,aAAa;AACzB,gBAAI,MAAM,MAAM,UAAU,IAAI;AAC9B,gBAAI,CAAC,IAAI,QAAQ;AACf,oBAAM,IAAI,WAAW,GAAG;AAAA,YAC1B;AACA,mBAAO;AAAA,UACT;AACA,sBAAY,CAAC,UAAU,QAAQ,YAAY;AACzC,uBAAW,SAAS,aAAa,QAAQ;AACzC,eAAG,SAAS,UAAU,SAAS,MAAM,MAAM;AACzC,kBAAI;AACF,wBAAQ,IAAI;AAAA;AAEZ,uBAAO,KAAK,MAAM;AAAA,YACtB,CAAC;AAAA,UACH;AACA,cAAI,QAAQ,QAAQ,SAAS,GAAG;AAC9B,0BAAc,QAAQ,QAAQ,GAAG,QAAQ,OAAO,GAAG;AAAA,UACrD;AACA,uBAAa,QAAQ,QAAQ,MAAM,CAAC;AACpC,kBAAQ,MAAM,qBAAqB,SAAS,IAAI;AAC9C,gBAAI,EAAE,cAAc,aAAa;AAC/B,oBAAM;AAAA,YACR;AAAA,UACF,CAAC;AACD,kBAAQ,MAAM,sBAAsB,SAAS,QAAQ;AACnD,kBAAM;AAAA,UACR,CAAC;AACD,kBAAQ,CAAC,QAAQ,YAAY;AAC3B,gBAAI,iBAAiB,GAAG;AACtB,sBAAQ,cAAc;AACtB,oBAAM;AAAA,YACR;AACA,+BAAmB,OAAO;AAC1B,oBAAQ,QAAQ,MAAM;AAAA,UACxB;AACA,iBAAO,aAAa,WAAW;AAC7B,mBAAO;AAAA,UACT;AAAA,QACF,WAAW,sBAAsB,uBAAuB;AACtD,cAAI,uBAAuB;AACzB,8BAAkB,KAAK,SAAS;AAAA,UAClC,WAAW,OAAO,YAAY,eAAe,SAAS,eAAe;AACnE,8BAAkB,SAAS,cAAc;AAAA,UAC3C;AACA,cAAI,YAAY;AACd,8BAAkB;AAAA,UACpB;AACA,cAAI,gBAAgB,QAAQ,OAAO,MAAM,GAAG;AAC1C,8BAAkB,gBAAgB,OAAO,GAAG,gBAAgB,QAAQ,UAAU,EAAE,EAAE,YAAY,GAAG,IAAI,CAAC;AAAA,UACxG,OAAO;AACL,8BAAkB;AAAA,UACpB;AACA;AACE,oBAAQ,CAAC,QAAQ;AACf,kBAAI,MAAM,IAAI,eAAe;AAC7B,kBAAI,KAAK,OAAO,KAAK,KAAK;AAC1B,kBAAI,KAAK,IAAI;AACb,qBAAO,IAAI;AAAA,YACb;AACA,gBAAI,uBAAuB;AACzB,2BAAa,CAAC,QAAQ;AACpB,oBAAI,MAAM,IAAI,eAAe;AAC7B,oBAAI,KAAK,OAAO,KAAK,KAAK;AAC1B,oBAAI,eAAe;AACnB,oBAAI,KAAK,IAAI;AACb,uBAAO,IAAI,WAAW,IAAI,QAAQ;AAAA,cACpC;AAAA,YACF;AACA,wBAAY,CAAC,KAAK,QAAQ,YAAY;AACpC,kBAAI,MAAM,IAAI,eAAe;AAC7B,kBAAI,KAAK,OAAO,KAAK,IAAI;AACzB,kBAAI,eAAe;AACnB,kBAAI,SAAS,MAAM;AACjB,oBAAI,IAAI,UAAU,OAAO,IAAI,UAAU,KAAK,IAAI,UAAU;AACxD,yBAAO,IAAI,QAAQ;AACnB;AAAA,gBACF;AACA,wBAAQ;AAAA,cACV;AACA,kBAAI,UAAU;AACd,kBAAI,KAAK,IAAI;AAAA,YACf;AAAA,UACF;AACA,2BAAiB,CAAC,UAAU,SAAS,QAAQ;AAAA,QAC/C,OAAO;AAAA,QACP;AACA,YAAI,MAAM,OAAO,YAAY,QAAQ,IAAI,KAAK,OAAO;AACrD,YAAI,MAAM,OAAO,eAAe,QAAQ,KAAK,KAAK,OAAO;AACzD,eAAO,OAAO,QAAQ,eAAe;AACrC,0BAAkB;AAClB,YAAI,OAAO;AACT,uBAAa,OAAO;AACtB,YAAI,OAAO;AACT,wBAAc,OAAO;AACvB,YAAI,OAAO;AACT,kBAAQ,OAAO;AACjB,YAAI,eAAe;AACnB,YAAI;AACJ,YAAI,OAAO;AACT,uBAAa,OAAO;AACtB,YAAI,gBAAgB,OAAO,oBAAoB;AAC/C,YAAI,OAAO,eAAe,UAAU;AAClC,gBAAM,iCAAiC;AAAA,QACzC;AACA,YAAI;AACJ,YAAI,QAAQ;AACZ,YAAI;AACJ,iBAAS,QAAQ,WAAW,MAAM;AAChC,cAAI,CAAC,WAAW;AACd,kBAAM,IAAI;AAAA,UACZ;AAAA,QACF;AACA,YAAI,cAAc,OAAO,eAAe,cAAc,IAAI,YAAY,MAAM,IAAI;AAChF,iBAAS,kBAAkB,aAAa,KAAK,gBAAgB;AAC3D,cAAI,SAAS,MAAM;AACnB,cAAI,SAAS;AACb,iBAAO,YAAY,WAAW,EAAE,UAAU;AACxC,cAAE;AACJ,cAAI,SAAS,MAAM,MAAM,YAAY,UAAU,aAAa;AAC1D,mBAAO,YAAY,OAAO,YAAY,SAAS,KAAK,MAAM,CAAC;AAAA,UAC7D;AACA,cAAI,MAAM;AACV,iBAAO,MAAM,QAAQ;AACnB,gBAAI,KAAK,YAAY;AACrB,gBAAI,EAAE,KAAK,MAAM;AACf,qBAAO,OAAO,aAAa,EAAE;AAC7B;AAAA,YACF;AACA,gBAAI,KAAK,YAAY,SAAS;AAC9B,iBAAK,KAAK,QAAQ,KAAK;AACrB,qBAAO,OAAO,cAAc,KAAK,OAAO,IAAI,EAAE;AAC9C;AAAA,YACF;AACA,gBAAI,KAAK,YAAY,SAAS;AAC9B,iBAAK,KAAK,QAAQ,KAAK;AACrB,oBAAM,KAAK,OAAO,KAAK,MAAM,IAAI;AAAA,YACnC,OAAO;AACL,oBAAM,KAAK,MAAM,KAAK,MAAM,KAAK,MAAM,IAAI,YAAY,SAAS;AAAA,YAClE;AACA,gBAAI,KAAK,OAAO;AACd,qBAAO,OAAO,aAAa,EAAE;AAAA,YAC/B,OAAO;AACL,kBAAI,KAAK,KAAK;AACd,qBAAO,OAAO,aAAa,QAAQ,MAAM,IAAI,QAAQ,KAAK,IAAI;AAAA,YAChE;AAAA,UACF;AACA,iBAAO;AAAA,QACT;AACA,iBAAS,aAAa,KAAK,gBAAgB;AACzC,iBAAO,MAAM,kBAAkB,QAAQ,KAAK,cAAc,IAAI;AAAA,QAChE;AACA,iBAAS,kBAAkB,KAAK,MAAM,QAAQ,iBAAiB;AAC7D,cAAI,EAAE,kBAAkB;AACtB,mBAAO;AACT,cAAI,WAAW;AACf,cAAI,SAAS,SAAS,kBAAkB;AACxC,mBAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,EAAE,GAAG;AACnC,gBAAI,IAAI,IAAI,WAAW,CAAC;AACxB,gBAAI,KAAK,SAAS,KAAK,OAAO;AAC5B,kBAAI,KAAK,IAAI,WAAW,EAAE,CAAC;AAC3B,kBAAI,UAAU,IAAI,SAAS,MAAM,KAAK;AAAA,YACxC;AACA,gBAAI,KAAK,KAAK;AACZ,kBAAI,UAAU;AACZ;AACF,mBAAK,YAAY;AAAA,YACnB,WAAW,KAAK,MAAM;AACpB,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B,WAAW,KAAK,OAAO;AACrB,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,KAAK,IAAI;AAChC,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B,OAAO;AACL,kBAAI,SAAS,KAAK;AAChB;AACF,mBAAK,YAAY,MAAM,KAAK;AAC5B,mBAAK,YAAY,MAAM,KAAK,KAAK;AACjC,mBAAK,YAAY,MAAM,KAAK,IAAI;AAChC,mBAAK,YAAY,MAAM,IAAI;AAAA,YAC7B;AAAA,UACF;AACA,eAAK,UAAU;AACf,iBAAO,SAAS;AAAA,QAClB;AACA,iBAAS,aAAa,KAAK,QAAQ,iBAAiB;AAClD,iBAAO,kBAAkB,KAAK,QAAQ,QAAQ,eAAe;AAAA,QAC/D;AACA,YAAI,SAAS,OAAO,QAAQ,QAAQ,SAAS,QAAQ,SAAS,SAAS;AACvE,iBAAS,2BAA2B,KAAK;AACvC,oBAAU;AACV,iBAAO,WAAW,QAAQ,IAAI,UAAU,GAAG;AAC3C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,YAAY,SAAS,IAAI,WAAW,GAAG;AAC9C,iBAAO,aAAa,UAAU,IAAI,YAAY,GAAG;AACjD,iBAAO,aAAa,UAAU,IAAI,YAAY,GAAG;AACjD,iBAAO,aAAa,UAAU,IAAI,aAAa,GAAG;AAClD,iBAAO,aAAa,UAAU,IAAI,aAAa,GAAG;AAAA,QACpD;AACA,YAAI,iBAAiB,OAAO,qBAAqB;AACjD,YAAI;AACJ,YAAI,eAAe,CAAC;AACpB,YAAI,aAAa,CAAC;AAClB,YAAI,gBAAgB,CAAC;AACrB,YAAI,qBAAqB;AACzB,iBAAS,mBAAmB;AAC1B,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS;AAChB,cAAI,OAAO,WAAW;AACpB,gBAAI,OAAO,OAAO,aAAa;AAC7B,qBAAO,YAAY,CAAC,OAAO,SAAS;AACtC,mBAAO,OAAO,UAAU,QAAQ;AAC9B,0BAAY,OAAO,UAAU,MAAM,CAAC;AAAA,YACtC;AAAA,UACF;AACA,+BAAqB,YAAY;AAAA,QACnC;AACA,iBAAS,cAAc;AACrB,+BAAqB;AACrB,+BAAqB,UAAU;AAAA,QACjC;AACA,iBAAS,UAAU;AACjB,cAAI,OAAO,YAAY;AACrB,gBAAI,OAAO,OAAO,cAAc;AAC9B,qBAAO,aAAa,CAAC,OAAO,UAAU;AACxC,mBAAO,OAAO,WAAW,QAAQ;AAC/B,2BAAa,OAAO,WAAW,MAAM,CAAC;AAAA,YACxC;AAAA,UACF;AACA,+BAAqB,aAAa;AAAA,QACpC;AACA,iBAAS,YAAY,IAAI;AACvB,uBAAa,QAAQ,EAAE;AAAA,QACzB;AACA,iBAAS,UAAU,IAAI;AACrB,qBAAW,QAAQ,EAAE;AAAA,QACvB;AACA,iBAAS,aAAa,IAAI;AACxB,wBAAc,QAAQ,EAAE;AAAA,QAC1B;AACA,YAAI,kBAAkB;AACtB,YAAI,uBAAuB;AAC3B,YAAI,wBAAwB;AAC5B,iBAAS,iBAAiB,IAAI;AAC5B;AACA,cAAI,OAAO,2BAA2B;AACpC,mBAAO,0BAA0B,eAAe;AAAA,UAClD;AAAA,QACF;AACA,iBAAS,oBAAoB,IAAI;AAC/B;AACA,cAAI,OAAO,2BAA2B;AACpC,mBAAO,0BAA0B,eAAe;AAAA,UAClD;AACA,cAAI,mBAAmB,GAAG;AACxB,gBAAI,yBAAyB,MAAM;AACjC,4BAAc,oBAAoB;AAClC,qCAAuB;AAAA,YACzB;AACA,gBAAI,uBAAuB;AACzB,kBAAI,WAAW;AACf,sCAAwB;AACxB,uBAAS;AAAA,YACX;AAAA,UACF;AAAA,QACF;AACA,iBAAS,MAAM,MAAM;AACnB;AACE,gBAAI,OAAO,YAAY;AACrB,qBAAO,WAAW,IAAI;AAAA,YACxB;AAAA,UACF;AACA,iBAAO,aAAa,OAAO;AAC3B,cAAI,IAAI;AACR,kBAAQ;AACR,uBAAa;AACb,kBAAQ;AACR,cAAI,IAAI,IAAI,YAAY,aAAa,IAAI;AACzC,6BAAmB,CAAC;AACpB,gBAAM;AAAA,QACR;AACA,YAAI,gBAAgB;AACpB,iBAAS,UAAU,UAAU;AAC3B,iBAAO,SAAS,WAAW,aAAa;AAAA,QAC1C;AACA,iBAAS,UAAU,UAAU;AAC3B,iBAAO,SAAS,WAAW,SAAS;AAAA,QACtC;AACA,YAAI;AACJ,yBAAiB;AACjB,YAAI,CAAC,UAAU,cAAc,GAAG;AAC9B,2BAAiB,WAAW,cAAc;AAAA,QAC5C;AACA,iBAAS,UAAU,MAAM;AACvB,cAAI;AACF,gBAAI,QAAQ,kBAAkB,YAAY;AACxC,qBAAO,IAAI,WAAW,UAAU;AAAA,YAClC;AACA,gBAAI,YAAY;AACd,qBAAO,WAAW,IAAI;AAAA,YACxB;AACA,kBAAM;AAAA,UACR,SAAS,MAAP;AACA,kBAAM,IAAI;AAAA,UACZ;AAAA,QACF;AACA,iBAAS,mBAAmB;AAC1B,cAAI,CAAC,eAAe,sBAAsB,wBAAwB;AAChE,gBAAI,OAAO,SAAS,cAAc,CAAC,UAAU,cAAc,GAAG;AAC5D,qBAAO,MAAM,gBAAgB,EAAE,aAAa,cAAc,CAAC,EAAE,KAAK,SAAS,UAAU;AACnF,oBAAI,CAAC,SAAS,OAAO;AACnB,wBAAM,yCAAyC,iBAAiB;AAAA,gBAClE;AACA,uBAAO,SAAS,eAAe;AAAA,cACjC,CAAC,EAAE,MAAM,WAAW;AAClB,uBAAO,UAAU,cAAc;AAAA,cACjC,CAAC;AAAA,YACH,OAAO;AACL,kBAAI,WAAW;AACb,uBAAO,IAAI,QAAQ,SAAS,SAAS,QAAQ;AAC3C,4BAAU,gBAAgB,SAAS,UAAU;AAC3C,4BAAQ,IAAI,WAAW,QAAQ,CAAC;AAAA,kBAClC,GAAG,MAAM;AAAA,gBACX,CAAC;AAAA,cACH;AAAA,YACF;AAAA,UACF;AACA,iBAAO,QAAQ,QAAQ,EAAE,KAAK,WAAW;AACvC,mBAAO,UAAU,cAAc;AAAA,UACjC,CAAC;AAAA,QACH;AACA,iBAAS,aAAa;AACpB,cAAI,OAAO,EAAE,OAAO,eAAe,0BAA0B,cAAc;AAC3E,mBAAS,gBAAgB,UAAU,SAAS;AAC1C,gBAAI,WAAW,SAAS;AACxB,mBAAO,SAAS;AAChB,yBAAa,OAAO,OAAO;AAC3B,uCAA2B,WAAW,MAAM;AAC5C,wBAAY,OAAO,OAAO;AAC1B,sBAAU,OAAO,OAAO,oBAAoB;AAC5C,gCAAoB,kBAAkB;AAAA,UACxC;AACA,2BAAiB,kBAAkB;AACnC,mBAAS,2BAA2B,QAAQ;AAC1C,4BAAgB,OAAO,WAAW;AAAA,UACpC;AACA,mBAAS,uBAAuB,UAAU;AACxC,mBAAO,iBAAiB,EAAE,KAAK,SAAS,QAAQ;AAC9C,qBAAO,YAAY,YAAY,QAAQ,IAAI;AAAA,YAC7C,CAAC,EAAE,KAAK,SAAS,UAAU;AACzB,qBAAO;AAAA,YACT,CAAC,EAAE,KAAK,UAAU,SAAS,QAAQ;AACjC,kBAAI,4CAA4C,MAAM;AACtD,oBAAM,MAAM;AAAA,YACd,CAAC;AAAA,UACH;AACA,mBAAS,mBAAmB;AAC1B,gBAAI,CAAC,cAAc,OAAO,YAAY,wBAAwB,cAAc,CAAC,UAAU,cAAc,KAAK,CAAC,UAAU,cAAc,KAAK,CAAC,uBAAuB,OAAO,SAAS,YAAY;AAC1L,qBAAO,MAAM,gBAAgB,EAAE,aAAa,cAAc,CAAC,EAAE,KAAK,SAAS,UAAU;AACnF,oBAAI,SAAS,YAAY,qBAAqB,UAAU,IAAI;AAC5D,uBAAO,OAAO,KAAK,4BAA4B,SAAS,QAAQ;AAC9D,sBAAI,oCAAoC,MAAM;AAC9C,sBAAI,2CAA2C;AAC/C,yBAAO,uBAAuB,0BAA0B;AAAA,gBAC1D,CAAC;AAAA,cACH,CAAC;AAAA,YACH,OAAO;AACL,qBAAO,uBAAuB,0BAA0B;AAAA,YAC1D;AAAA,UACF;AACA,cAAI,OAAO,oBAAoB;AAC7B,gBAAI;AACF,kBAAI,WAAW,OAAO,mBAAmB,MAAM,eAAe;AAC9D,qBAAO;AAAA,YACT,SAAS,GAAP;AACA,kBAAI,wDAAwD,CAAC;AAC7D,iCAAmB,CAAC;AAAA,YACtB;AAAA,UACF;AACA,2BAAiB,EAAE,MAAM,kBAAkB;AAC3C,iBAAO,CAAC;AAAA,QACV;AACA,YAAI;AACJ,YAAI;AACJ,iBAAS,WAAW,QAAQ;AAC1B,eAAK,OAAO;AACZ,eAAK,UAAU,kCAAkC,SAAS;AAC1D,eAAK,SAAS;AAAA,QAChB;AACA,iBAAS,qBAAqB,YAAY;AACxC,iBAAO,WAAW,SAAS,GAAG;AAC5B,uBAAW,MAAM,EAAE,MAAM;AAAA,UAC3B;AAAA,QACF;AACA,iBAAS,SAAS,OAAO;AACvB,iBAAO;AAAA,QACT;AACA,iBAAS,YAAY,MAAM;AACzB,cAAI,QAAQ;AACZ,iBAAO,KAAK,QAAQ,OAAO,SAAS,GAAG;AACrC,gBAAI,IAAI,SAAS,CAAC;AAClB,mBAAO,MAAM,IAAI,IAAI,IAAI,OAAO,IAAI;AAAA,UACtC,CAAC;AAAA,QACH;AACA,iBAAS,eAAe;AACtB,cAAI,QAAQ,IAAI,MAAM;AACtB,cAAI,CAAC,MAAM,OAAO;AAChB,gBAAI;AACF,oBAAM,IAAI,MAAM;AAAA,YAClB,SAAS,GAAP;AACA,sBAAQ;AAAA,YACV;AACA,gBAAI,CAAC,MAAM,OAAO;AAChB,qBAAO;AAAA,YACT;AAAA,UACF;AACA,iBAAO,MAAM,MAAM,SAAS;AAAA,QAC9B;AACA,iBAAS,mBAAmB,QAAQ,SAAS;AAC3C,gBAAM,IAAI,QAAQ,OAAO;AAAA,QAC3B;AACA,iBAAS,SAAS;AAChB,gBAAM,EAAE;AAAA,QACV;AACA,iBAAS,aAAa;AACpB,iBAAO;AAAA,QACT;AACA,iBAAS,2BAA2B;AAClC,iBAAO,WAAW;AAAA,QACpB;AACA,iBAAS,uBAAuB,MAAM,KAAK,KAAK;AAC9C,iBAAO,WAAW,MAAM,KAAK,MAAM,GAAG;AAAA,QACxC;AACA,iBAAS,0BAA0B,MAAM;AACvC,cAAI;AACF,uBAAW,KAAK,OAAO,QAAQ,aAAa,UAAU,EAAE;AACxD,uCAA2B,WAAW,MAAM;AAC5C,mBAAO;AAAA,UACT,SAAS,GAAP;AAAA,UACF;AAAA,QACF;AACA,iBAAS,wBAAwB,eAAe;AAC9C,cAAI,UAAU,OAAO;AACrB,0BAAgB,kBAAkB;AAClC,cAAI,cAAc,WAAW;AAC7B,cAAI,gBAAgB,aAAa;AAC/B,mBAAO;AAAA,UACT;AACA,cAAI,UAAU,CAAC,GAAG,aAAa,KAAK,WAAW,IAAI,YAAY;AAC/D,mBAAS,UAAU,GAAG,WAAW,GAAG,WAAW,GAAG;AAChD,gBAAI,oBAAoB,WAAW,IAAI,MAAM;AAC7C,gCAAoB,KAAK,IAAI,mBAAmB,gBAAgB,SAAS;AACzE,gBAAI,UAAU,KAAK,IAAI,aAAa,QAAQ,KAAK,IAAI,eAAe,iBAAiB,GAAG,KAAK,CAAC;AAC9F,gBAAI,cAAc,0BAA0B,OAAO;AACnD,gBAAI,aAAa;AACf,qBAAO;AAAA,YACT;AAAA,UACF;AACA,iBAAO;AAAA,QACT;AACA,YAAI,WAAW,EAAE,SAAS,QAAQ,KAAK,WAAW;AAChD,mBAAS,WAAW;AACpB,cAAI,MAAM,OAAO,SAAS,UAAU,KAAK;AACzC,iBAAO;AAAA,QACT,GAAG,QAAQ,SAAS,KAAK;AACvB,cAAI,MAAM,aAAa,GAAG;AAC1B,iBAAO;AAAA,QACT,EAAE;AACF,iBAAS,UAAU,IAAI;AACrB,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS,IAAI,YAAY,aAAa,QAAQ,WAAW;AAChE,iBAAO;AAAA,QACT;AACA,YAAI,mBAAmB,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;AACpC,iBAAS,UAAU,QAAQ,MAAM;AAC/B,cAAI,UAAU,iBAAiB;AAC/B,cAAI,SAAS,KAAK,SAAS,IAAI;AAC7B,aAAC,WAAW,IAAI,MAAM,KAAK,kBAAkB,SAAS,CAAC,CAAC;AACxD,oBAAQ,SAAS;AAAA,UACnB,OAAO;AACL,oBAAQ,KAAK,IAAI;AAAA,UACnB;AAAA,QACF;AACA,iBAAS,UAAU,IAAI,KAAK,QAAQ,MAAM;AACxC,cAAI,MAAM;AACV,mBAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,gBAAI,MAAM,QAAQ,OAAO;AACzB,gBAAI,MAAM,QAAQ,MAAM,KAAK;AAC7B,mBAAO;AACP,qBAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,wBAAU,IAAI,OAAO,MAAM,EAAE;AAAA,YAC/B;AACA,mBAAO;AAAA,UACT;AACA,kBAAQ,QAAQ,KAAK;AACrB,iBAAO;AAAA,QACT;AACA,iBAAS,SAAS,OAAO;AACvB,cAAI,QAAQ,OAAO,MAAM;AACzB,iBAAO;AAAA,QACT;AACA,iBAAS,MAAM,OAAO,YAAY,UAAU,MAAM,MAAM;AACtD,cAAI,MAAM,EAAE,UAAU,CAAC,QAAQ;AAC7B,gBAAI,OAAO;AACX,gBAAI,QAAQ,QAAQ,QAAQ,UAAU,QAAQ,GAAG;AAC/C,kBAAI,OAAO,IAAI,UAAU,KAAK;AAC9B,qBAAO,WAAW,GAAG;AACrB,2BAAa,KAAK,MAAM,GAAG;AAAA,YAC7B;AACA,mBAAO;AAAA,UACT,GAAG,SAAS,CAAC,QAAQ;AACnB,gBAAI,OAAO,WAAW,IAAI,MAAM;AAChC,+BAAmB,KAAK,IAAI;AAC5B,mBAAO;AAAA,UACT,EAAE;AACF,mBAAS,mBAAmB,MAAM;AAChC,gBAAI,eAAe,UAAU;AAC3B,qBAAO,aAAa,IAAI;AAAA,YAC1B;AACA,gBAAI,eAAe;AACjB,qBAAO,QAAQ,IAAI;AACrB,mBAAO;AAAA,UACT;AACA,cAAI,QAAQ,SAAS,KAAK;AAC1B,cAAI,QAAQ,CAAC;AACb,cAAI,SAAS;AACb,cAAI,MAAM;AACR,qBAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,kBAAI,YAAY,IAAI,SAAS;AAC7B,kBAAI,WAAW;AACb,oBAAI,WAAW;AACb,2BAAS,UAAU;AACrB,sBAAM,KAAK,UAAU,KAAK,EAAE;AAAA,cAC9B,OAAO;AACL,sBAAM,KAAK,KAAK;AAAA,cAClB;AAAA,YACF;AAAA,UACF;AACA,cAAI,MAAM,MAAM,MAAM,MAAM,KAAK;AACjC,mBAAS,OAAO,MAAM;AACpB,gBAAI,WAAW;AACb,2BAAa,MAAM;AACrB,mBAAO,mBAAmB,IAAI;AAAA,UAChC;AACA,gBAAM,OAAO,GAAG;AAChB,iBAAO;AAAA,QACT;AACA,iBAAS,MAAM,OAAO,YAAY,UAAU,MAAM;AAChD,qBAAW,YAAY,CAAC;AACxB,cAAI,cAAc,SAAS,MAAM,CAAC,SAAS,SAAS,YAAY,SAAS,SAAS;AAClF,cAAI,aAAa,eAAe;AAChC,cAAI,cAAc,eAAe,CAAC,MAAM;AACtC,mBAAO,SAAS,KAAK;AAAA,UACvB;AACA,iBAAO,WAAW;AAChB,mBAAO,MAAM,OAAO,YAAY,UAAU,WAAW,IAAI;AAAA,UAC3D;AAAA,QACF;AACA,YAAI,gBAAgB,EAAE,SAAS,QAAQ,2BAA2B,0BAA0B,yBAAyB,wBAAwB,0BAA0B,yBAAyB,YAAY,WAAW,WAAW,UAAU,YAAY,UAAU;AAClQ,YAAI,MAAM,WAAW;AACrB,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,2BAA2B,OAAO,8BAA8B,WAAW;AAC7E,kBAAQ,2BAA2B,OAAO,8BAA8B,OAAO,OAAO,4BAA4B,MAAM,MAAM,SAAS;AAAA,QACzI;AACA,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,mBAAmB,OAAO,sBAAsB,WAAW;AAC7D,kBAAQ,mBAAmB,OAAO,sBAAsB,OAAO,OAAO,oBAAoB,MAAM,MAAM,SAAS;AAAA,QACjH;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,yBAAyB,OAAO,4BAA4B,WAAW;AACzE,kBAAQ,yBAAyB,OAAO,4BAA4B,OAAO,OAAO,0BAA0B,MAAM,MAAM,SAAS;AAAA,QACnI;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,kBAAkB,OAAO,qBAAqB,WAAW;AAC3D,kBAAQ,kBAAkB,OAAO,qBAAqB,OAAO,OAAO,mBAAmB,MAAM,MAAM,SAAS;AAAA,QAC9G;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,gBAAgB,MAAM,MAAM,SAAS;AAAA,QACrG;AACA,YAAI,wBAAwB,OAAO,2BAA2B,WAAW;AACvE,kBAAQ,wBAAwB,OAAO,2BAA2B,OAAO,OAAO,yBAAyB,MAAM,MAAM,SAAS;AAAA,QAChI;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,cAAc,OAAO,iBAAiB,WAAW;AACnD,kBAAQ,cAAc,OAAO,iBAAiB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAClG;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,kBAAkB,OAAO,qBAAqB,WAAW;AAC3D,kBAAQ,kBAAkB,OAAO,qBAAqB,OAAO,OAAO,mBAAmB,MAAM,MAAM,SAAS;AAAA,QAC9G;AACA,YAAI,yBAAyB,OAAO,4BAA4B,WAAW;AACzE,kBAAQ,yBAAyB,OAAO,4BAA4B,OAAO,OAAO,0BAA0B,MAAM,MAAM,SAAS;AAAA,QACnI;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,oBAAoB,OAAO,uBAAuB,WAAW;AAC/D,kBAAQ,oBAAoB,OAAO,uBAAuB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QACpH;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,SAAS,OAAO,YAAY,WAAW;AACzC,kBAAQ,SAAS,OAAO,YAAY,OAAO,OAAO,UAAU,MAAM,MAAM,SAAS;AAAA,QACnF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,aAAa,MAAM,MAAM,SAAS;AAAA,QAC5F;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,WAAW,OAAO,cAAc,WAAW;AAC7C,kBAAQ,WAAW,OAAO,cAAc,OAAO,OAAO,YAAY,MAAM,MAAM,SAAS;AAAA,QACzF;AACA,YAAI,uBAAuB,OAAO,0BAA0B,WAAW;AACrE,kBAAQ,uBAAuB,OAAO,0BAA0B,OAAO,OAAO,wBAAwB,MAAM,MAAM,SAAS;AAAA,QAC7H;AACA,YAAI,iBAAiB,OAAO,oBAAoB,WAAW;AACzD,kBAAQ,iBAAiB,OAAO,oBAAoB,OAAO,OAAO,kBAAkB,MAAM,MAAM,SAAS;AAAA,QAC3G;AACA,YAAI,0BAA0B,OAAO,6BAA6B,WAAW;AAC3E,kBAAQ,0BAA0B,OAAO,6BAA6B,OAAO,OAAO,2BAA2B,MAAM,MAAM,SAAS;AAAA,QACtI;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,qBAAqB,OAAO,wBAAwB,WAAW;AACjE,kBAAQ,qBAAqB,OAAO,wBAAwB,OAAO,OAAO,sBAAsB,MAAM,MAAM,SAAS;AAAA,QACvH;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,OAAO,OAAO,UAAU,WAAW;AACrC,kBAAQ,OAAO,OAAO,UAAU,OAAO,OAAO,QAAQ,MAAM,MAAM,SAAS;AAAA,QAC7E;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC/F;AACA,YAAI,gBAAgB,OAAO,mBAAmB,WAAW;AACvD,kBAAQ,gBAAgB,OAAO,mBAAmB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACxG;AACA,YAAI,UAAU,OAAO,aAAa,WAAW;AAC3C,kBAAQ,UAAU,OAAO,aAAa,OAAO,OAAO,WAAW,MAAM,MAAM,SAAS;AAAA,QACtF;AACA,YAAI,QAAQ,OAAO,WAAW,WAAW;AACvC,kBAAQ,QAAQ,OAAO,WAAW,OAAO,OAAO,SAAS,MAAM,MAAM,SAAS;AAAA,QAChF;AACA,YAAI,oBAAoB,OAAO,uBAAuB,WAAW;AAC/D,kBAAQ,oBAAoB,OAAO,uBAAuB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QACpH;AACA,YAAI,YAAY,OAAO,eAAe,WAAW;AAC/C,kBAAQ,YAAY,OAAO,eAAe,OAAO,OAAO,cAAc,MAAM,MAAM,SAAS;AAAA,QAC7F;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACtG;AACA,YAAI,aAAa,OAAO,gBAAgB,WAAW;AACjD,kBAAQ,aAAa,OAAO,gBAAgB,OAAO,OAAO,eAAe,MAAM,MAAM,SAAS;AAAA,QAChG;AACA,YAAI,mBAAmB,OAAO,sBAAsB,WAAW;AAC7D,kBAAQ,mBAAmB,OAAO,sBAAsB,OAAO,OAAO,qBAAqB,MAAM,MAAM,SAAS;AAAA,QAClH;AACA,YAAI,eAAe,OAAO,kBAAkB,WAAW;AACrD,kBAAQ,eAAe,OAAO,kBAAkB,OAAO,OAAO,iBAAiB,MAAM,MAAM,SAAS;AAAA,QACtG;AACA,eAAO,WAAW;AAClB,YAAI;AACJ,gCAAwB,SAAS,YAAY;AAC3C,cAAI,CAAC;AACH,gBAAI;AACN,cAAI,CAAC;AACH,oCAAwB;AAAA,QAC5B;AACA,iBAAS,IAAI,MAAM;AACjB,iBAAO,QAAQ;AACf,cAAI,kBAAkB,GAAG;AACvB;AAAA,UACF;AACA,iBAAO;AACP,cAAI,kBAAkB,GAAG;AACvB;AAAA,UACF;AACA,mBAAS,QAAQ;AACf,gBAAI;AACF;AACF,wBAAY;AACZ,mBAAO,eAAe;AACtB,gBAAI;AACF;AACF,wBAAY;AACZ,gCAAoB,MAAM;AAC1B,gBAAI,OAAO;AACT,qBAAO,wBAAwB;AACjC,oBAAQ;AAAA,UACV;AACA,cAAI,OAAO,cAAc;AACvB,mBAAO,aAAa,YAAY;AAChC,uBAAW,WAAW;AACpB,yBAAW,WAAW;AACpB,uBAAO,aAAa,EAAE;AAAA,cACxB,GAAG,CAAC;AACJ,oBAAM;AAAA,YACR,GAAG,CAAC;AAAA,UACN,OAAO;AACL,kBAAM;AAAA,UACR;AAAA,QACF;AACA,YAAI,OAAO,YAAY;AACrB,cAAI,OAAO,OAAO,cAAc;AAC9B,mBAAO,aAAa,CAAC,OAAO,UAAU;AACxC,iBAAO,OAAO,WAAW,SAAS,GAAG;AACnC,mBAAO,WAAW,IAAI,EAAE;AAAA,UAC1B;AAAA,QACF;AACA,YAAI;AACJ,YAAI;AACJ,YAAI,iBAAiB;AACnB,2BAAiB,EAAE,mBAAmB,QAAQ,UAAU,mBAAmB,EAAE,OAAO,SAAS,UAAU;AACrG,mBAAO,CAAC,gBAAgB,kBAAkB,QAAQ,QAAQ,IAAI;AAAA,UAChE,CAAC,GAAG,oBAAoB,QAAQ,UAAU,oBAAoB,EAAE,OAAO,SAAS,UAAU;AACxF,mBAAO,CAAC,gBAAgB,mBAAmB,QAAQ,QAAQ,IAAI;AAAA,UACjE,CAAC,EAAE;AAAA,QACL;AACA,YAAI;AACJ,YAAI,OAAO,uBAAuB,aAAa;AAC7C,yBAAe;AAAA,QACjB,WAAW,OAAO,kCAAkC,aAAa;AAC/D,yBAAe;AAAA,QACjB,OAAO;AACL,gBAAM,IAAI,MAAM,uCAAuC;AAAA,QACzD;AACA,YAAI,gBAAgB;AAClB,cAAI,aAAa,aAAa;AAC9B,uBAAa,cAAc,WAAW;AACpC,uBAAW;AACX,2BAAe,kBAAkB,QAAQ,SAAS,UAAU;AAC1D,sBAAQ,eAAe,qBAAqB,QAAQ;AAAA,YACtD,CAAC;AACD,2BAAe,mBAAmB,QAAQ,SAAS,UAAU;AAC3D,sBAAQ,eAAe,sBAAsB,QAAQ;AAAA,YACvD,CAAC;AAAA,UACH;AAAA,QACF;AACA,eAAO,mBAAmB;AAAA,MAC5B;AAAA,IACF,GAAG;AACH,QAAI,OAAO,YAAY,YAAY,OAAO,WAAW;AACnD,aAAO,UAAU;AAAA,aACV,OAAO,WAAW,cAAc,OAAO;AAC9C,aAAO,CAAC,GAAG,WAAW;AACpB,eAAO;AAAA,MACT,CAAC;AAAA,aACM,OAAO,YAAY;AAC1B,cAAQ,uBAAuB;AAAA,EACnC;AACF,CAAC;AAGD,IAAI,kBAAkB;AACtB,IAAI,kBAAkB;AACtB,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,UAAU,WAAW;AAC/B,SAAK,UAAU;AACf,SAAK,YAAY;AACjB,SAAK,OAAuB,oBAAI,QAAQ;AACxC,SAAK,eAAe;AAAA,EACtB;AAAA,EACA,IAAI,QAAQ;AACV,QAAI,CAAC,KAAK,KAAK,IAAI,MAAM,GAAG;AAC1B,WAAK,UAAU,SAAS,KAAK,SAAS,MAAM;AAAA,IAC9C;AACA,WAAO,KAAK,KAAK,IAAI,MAAM;AAAA,EAC7B;AAAA,EACA,IAAI,QAAQ,OAAO;AACjB,SAAK;AACL,SAAK,KAAK,IAAI,QAAQ,KAAK;AAAA,EAC7B;AAAA,EACA,IAAI,QAAQ;AACV,WAAO,KAAK,KAAK,IAAI,MAAM;AAAA,EAC7B;AAAA,EACA,OAAO,QAAQ;AACb,SAAK;AACL,WAAO,KAAK,KAAK,OAAO,MAAM;AAAA,EAChC;AAAA,EACA,aAAa;AACX,WAAO,KAAK;AAAA,EACd;AACF;AACA,IAAI,gBAAgB,MAAM;AAAA,EACxB,SAAS,QAAQ;AACf,WAAO,kBAAkB,UAAU;AAAA,EACrC;AAAA,EACA,OAAO,QAAQ;AACb,WAAO,kBAAkB,QAAQ;AAAA,EACnC;AAAA,EACA,iBAAiB;AACf,WAAO;AAAA,EACT;AAAA,EACA,KAAK,GAAG;AACN,WAAO,kBAAkB,MAAM;AAAA,EACjC;AAAA,EACA,KAAK,QAAQ;AACX,WAAO,kBAAkB,MAAM;AAAA,EACjC;AAAA,EACA,SAAS,QAAQ;AACf,WAAO,kBAAkB,UAAU;AAAA,EACrC;AAAA,EACA,UAAU,QAAQ,SAAS;AACzB,WAAO,kBAAkB,WAAW;AAAA,EACtC;AAAA,EACA,aAAa;AACX,WAAO,kBAAkB,YAAY;AAAA,EACvC;AAAA,EACA,YAAY,QAAQ,OAAO;AACzB,WAAO,kBAAkB,aAAa;AAAA,EACxC;AAAA,EACA,MAAM,QAAQ,OAAO,OAAO;AAC1B,WAAO,kBAAkB,OAAO;AAAA,EAClC;AAAA,EACA,KAAK,QAAQ,QAAQ,OAAO,OAAO,UAAU;AAC3C,WAAO,kBAAkB,MAAM;AAAA,EACjC;AAAA,EACA,wBAAwB,QAAQ,OAAO,OAAO;AAC5C,WAAO,kBAAkB,yBAAyB;AAAA,EACpD;AAAA,EACA,SAAS;AACP,WAAO,kBAAkB,QAAQ;AAAA,EACnC;AAAA,EACA,iBAAiB;AACf,WAAO,kBAAkB,gBAAgB;AAAA,EAC3C;AAAA,EACA,UAAU;AACR,WAAO,KAAK,eAAe,MAAM,KAAK,kBAAkB;AAAA,EAC1D;AAAA,EACA,UAAU;AACR,WAAO,kBAAkB,SAAS;AAAA,EACpC;AACF;AACA,SAAS,kBAAkB,YAAY;AACrC,QAAM,IAAI,MAAM,IAAI,oIAAoI;AAC1J;AAGA,SAAS,QAAQ,QAAQ;AACvB,MAAI,UAAU,OAAO;AACrB,MAAI,QAAQ;AACZ,SAAO,UAAU,GAAG;AAClB,YAAQ,KAAK,OAAO,IAAI,UAAU;AAClC;AACA,SAAK,QAAQ,SAAS,KAAK;AAAA,EAC7B;AACF;AACA,SAAS,aAAa,QAAQ,SAAS;AACrC,MAAI,OAAO,WAAW,QAAQ,QAAQ;AACpC,UAAM,IAAI,MAAM,yEAAyE,OAAO,iCAAiC,QAAQ,QAAQ;AAAA,EACnJ;AACA,MAAI,UAAU,OAAO;AACrB,MAAI,QAAQ;AACZ,SAAO,UAAU,GAAG;AAClB,YAAQ,KAAK,OAAO,IAAI,UAAU;AAClC;AACA,SAAK,QAAQ,SAAS,KAAK;AAC3B,SAAK,SAAS,SAAS,KAAK;AAAA,EAC9B;AACF;AACA,SAAS,MAAM,MAAM,GAAG,MAAM;AAC5B,SAAO,KAAK,IAAI,MAAM,KAAK,IAAI,GAAG,IAAI,CAAC;AACzC;AACA,SAAS,kBAAkB,KAAK;AAC9B,SAAO,MAAM,MAAM,IAAI,MAAM,MAAM;AACrC;AACA,SAAS,KAAK,QAAQ,MAAM,OAAO;AACjC,QAAM,OAAO,OAAO;AACpB,SAAO,QAAQ,OAAO;AACtB,SAAO,SAAS;AAClB;AACA,SAAS,IAAI,KAAK;AAChB,MAAI,OAAO;AACX,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,YAAQ,IAAI;AAAA,EACd;AACA,SAAO;AACT;AACA,SAAS,YAAY,GAAG,GAAG;AACzB,QAAM,IAAI,KAAK,OAAO;AACtB,SAAO,IAAI,KAAK,IAAI,KAAK;AAC3B;AACA,SAAS,YAAY,GAAG,GAAG;AACzB,MAAI,SAAS;AACb,WAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,UAAM,OAAO,OAAO,EAAE,EAAE,IAAI,OAAO,EAAE,EAAE;AACvC,cAAU,OAAO;AAAA,EACnB;AACA,SAAO;AACT;AACA,SAAS,OAAO,MAAM,KAAK;AACzB,MAAI,CAAC,MAAM;AACT,UAAM,IAAI,MAAM,OAAO,QAAQ,WAAW,MAAM,IAAI,CAAC;AAAA,EACvD;AACF;AACA,SAAS,kBAAkB,QAAQ,QAAQ,qBAAqB,IAAI;AAClE,SAAO,YAAY,QAAQ,MAAM,GAAG,MAAM,qBAAqB,WAAW,cAAc,mBAAmB;AAC7G;AACA,SAAS,cAAc,GAAG;AACxB,SAAO,KAAK,MAAM,MAAM,+DAA+D;AACzF;AACA,SAAS,QAAQ,KAAK,SAAS,CAAC,GAAG,iBAAiB,OAAO;AACzD,MAAI,UAAU,MAAM;AAClB,aAAS,CAAC;AAAA,EACZ;AACA,MAAI,MAAM,QAAQ,GAAG,KAAK,aAAa,GAAG,KAAK,CAAC,gBAAgB;AAC9D,aAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,EAAE,GAAG;AACnC,cAAQ,IAAI,IAAI,QAAQ,cAAc;AAAA,IACxC;AAAA,EACF,OAAO;AACL,WAAO,KAAK,GAAG;AAAA,EACjB;AACA,SAAO;AACT;AACA,SAAS,cAAc,OAAO;AAC5B,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,EACT;AACA,MAAI,OAAO,MAAM;AACjB,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,YAAQ,MAAM;AAAA,EAChB;AACA,SAAO;AACT;AACA,SAAS,cAAc,OAAO;AAC5B,SAAO,MAAM,WAAW;AAC1B;AACA,SAAS,YAAY,IAAI,IAAI;AAC3B,MAAI,OAAO,IAAI;AACb,WAAO;AAAA,EACT;AACA,MAAI,MAAM,QAAQ,MAAM,MAAM;AAC5B,WAAO;AAAA,EACT;AACA,MAAI,GAAG,WAAW,GAAG,QAAQ;AAC3B,WAAO;AAAA,EACT;AACA,WAAS,IAAI,GAAG,IAAI,GAAG,QAAQ,KAAK;AAClC,QAAI,GAAG,OAAO,GAAG,IAAI;AACnB,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,MAAM,GAAG;AAChB,SAAO,IAAI,MAAM;AACnB;AACA,SAAS,KAAK,GAAG;AACf,MAAI,KAAK,QAAQ,MAAM;AACrB,WAAO,KAAK,KAAK,CAAC;AAAA,EACpB;AACA,MAAI,MAAM,UAAU;AAClB,WAAO;AAAA,EACT,WAAW,MAAM,WAAW;AAC1B,WAAO;AAAA,EACT,OAAO;AACL,UAAM,MAAM,KAAK,IAAI,IAAI,CAAC;AAC1B,YAAQ,MAAM,MAAM,MAAM;AAAA,EAC5B;AACF;AACA,SAAS,oBAAoB,MAAM;AACjC,QAAM,QAAQ,KAAK,KAAK,KAAK,KAAK,IAAI,CAAC;AACvC,SAAO,CAAC,OAAO,KAAK,KAAK,OAAO,KAAK,CAAC;AACxC;AACA,SAAS,sBAAsB,GAAG;AAChC,QAAM,kBAAkB,IAAI,YAAY,CAAC;AACzC,WAAS,IAAI,GAAG,IAAI,GAAG,EAAE,GAAG;AAC1B,oBAAgB,KAAK;AAAA,EACvB;AACA,UAAQ,eAAe;AACvB,SAAO;AACT;AACA,SAAS,SAAS,GAAG,MAAM;AACzB,MAAI,QAAQ,EAAE,QAAQ;AACpB,WAAO;AAAA,EACT;AACA,SAAO,IAAI,IAAI,OAAO,OAAO,EAAE,MAAM;AACvC;AACA,SAAS,YAAY,SAAS,UAAU,CAAC,YAAY,GAAG,YAAY,YAAY;AAC9E,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,QAAI,WAAW;AACf,UAAM,QAAQ,MAAM;AAClB,UAAI,QAAQ,GAAG;AACb,gBAAQ;AACR;AAAA,MACF;AACA;AACA,YAAM,cAAc,QAAQ,QAAQ;AACpC,UAAI,cAAc,QAAQ,YAAY,YAAY;AAChD,eAAO;AACP;AAAA,MACF;AACA,UAAI,cAAc,MAAM;AACtB,mBAAW,OAAO,WAAW;AAAA,MAC/B,OAAO;AACL,mBAAW,OAAO,WAAW;AAAA,MAC/B;AAAA,IACF;AACA,UAAM;AAAA,EACR,CAAC;AACH;AACA,SAAS,uBAAuB,OAAO,MAAM;AAC3C,MAAI,YAAY;AAChB,MAAI,cAAc;AAClB,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,QAAI,MAAM,MAAM,GAAG;AACjB,mBAAa,MAAM;AAAA,IACrB,WAAW,MAAM,OAAO,IAAI;AAC1B,UAAI,gBAAgB,IAAI;AACtB,cAAM,MAAM,yDAAyD,uBAAuB,GAAG;AAAA,MACjG;AACA,oBAAc;AAAA,IAChB,WAAW,MAAM,KAAK,GAAG;AACvB,YAAM,MAAM,gCAAgC,MAAM,aAAa,GAAG;AAAA,IACpE;AAAA,EACF;AACA,MAAI,gBAAgB,IAAI;AACtB,QAAI,OAAO,KAAK,SAAS,WAAW;AAClC,YAAM,MAAM,QAAQ,yCAAyC,OAAO;AAAA,IACtE;AACA,WAAO;AAAA,EACT;AACA,MAAI,cAAc,GAAG;AACnB,UAAM,MAAM,qCAAqC,kCAAkC;AAAA,EACrF;AACA,MAAI,OAAO,cAAc,GAAG;AAC1B,UAAM,MAAM,wDAAwD,UAAU,WAAW;AAAA,EAC3F;AACA,QAAM,WAAW,MAAM,MAAM;AAC7B,WAAS,eAAe,OAAO;AAC/B,SAAO;AACT;AACA,SAAS,eAAe,MAAM,OAAO;AACnC,QAAM,OAAO,MAAM;AACnB,SAAO,QAAQ,OAAO,MAAM,IAAI,CAAC,GAAG,MAAM,CAAC,IAAI,CAAC,EAAE,OAAO,IAAI;AAC7D,SAAO,KAAK,MAAM,CAAC,OAAO,MAAM,CAAC,QAAQ,KAAK,IAAI,GAAG,MAAM,+CAA+C,SAAS,sBAAsB,MAAM;AAC/I,SAAO,KAAK,MAAM,CAAC,OAAO,MAAM,EAAE,CAAC,GAAG,MAAM,0DAA0D,MAAM;AAC5G,SAAO,KAAK,IAAI,CAAC,MAAM,IAAI,IAAI,OAAO,IAAI,CAAC;AAC7C;AACA,SAAS,aAAa,OAAO,MAAM;AACjC,QAAM,WAAW,CAAC;AAClB,QAAM,WAAW,CAAC;AAClB,QAAM,eAAe,QAAQ,QAAQ,MAAM,QAAQ,IAAI,KAAK,KAAK,WAAW;AAC5E,QAAM,OAAO,QAAQ,QAAQ,eAAe,OAAO,eAAe,MAAM,KAAK,EAAE,KAAK;AACpF,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,QAAI,QAAQ,MAAM;AAChB,UAAI,KAAK,OAAO,KAAK,MAAM,OAAO,GAAG;AACnC,cAAM,IAAI,MAAM,sBAAsB,oBAAoB,MAAM,cAAc;AAAA,MAChF;AACA,WAAK,KAAK,MAAM,QAAQ,KAAK,KAAK,MAAM,MAAM,OAAO,GAAG;AACtD,iBAAS,KAAK,MAAM,EAAE;AACtB,iBAAS,KAAK,CAAC;AAAA,MACjB;AACA,UAAI,KAAK,MAAM,GAAG;AAChB;AAAA,MACF;AAAA,IACF;AACA,QAAI,MAAM,OAAO,GAAG;AAClB,eAAS,KAAK,MAAM,EAAE;AACtB,eAAS,KAAK,CAAC;AAAA,IACjB;AAAA,EACF;AACA,SAAO,EAAE,UAAU,SAAS;AAC9B;AACA,SAAS,uBAAuB,OAAO,MAAM;AAC3C,MAAI,SAAS;AACb,MAAI,SAAS,QAAQ,UAAU,WAAW;AACxC,aAAS,IAAI,aAAa,IAAI;AAAA,EAChC,WAAW,UAAU,SAAS;AAC5B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,WAAW,UAAU,QAAQ;AAC3B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,OAAO,MAAM;AACtC,MAAI,SAAS;AACb,MAAI,SAAS,QAAQ,UAAU,WAAW;AACxC,aAAS,IAAI,aAAa,IAAI;AAAA,EAChC,WAAW,UAAU,SAAS;AAC5B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,WAAW,UAAU,QAAQ;AAC3B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,WAAW,UAAU,UAAU;AAC7B,aAAS,IAAI,MAAM,IAAI;AAAA,EACzB,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACA,SAAO;AACT;AACA,SAAS,yBAAyB,MAAM,OAAO;AAC7C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,MAAM,KAAK;AACjB,QAAI,MAAM,GAAG,KAAK,CAAC,SAAS,GAAG,GAAG;AAChC,YAAM,MAAM,oBAAoB,iCAAiC,MAAM;AAAA,IACzE;AAAA,EACF;AACF;AACA,SAAS,aAAa,OAAO;AAC3B,SAAO,UAAU,UAAU,UAAU,eAAe,UAAU,aAAa,UAAU,WAAW,UAAU;AAC5G;AACA,SAAS,gBAAgB,SAAS,SAAS;AACzC,MAAI,YAAY,aAAa;AAC3B,WAAO;AAAA,EACT;AACA,MAAI,YAAY,aAAa,YAAY,aAAa;AACpD,WAAO;AAAA,EACT;AACA,MAAI,YAAY,WAAW,YAAY,aAAa,YAAY,aAAa;AAC3E,WAAO;AAAA,EACT;AACA,MAAI,YAAY,UAAU,YAAY,QAAQ;AAC5C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,aAAa,GAAG;AACvB,SAAO,aAAa,gBAAgB,aAAa,cAAc,aAAa,cAAc,aAAa;AACzG;AACA,SAAS,gBAAgB,OAAO;AAC9B,MAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,WAAO;AAAA,EACT,WAAW,UAAU,aAAa;AAChC,WAAO;AAAA,EACT,WAAW,UAAU,QAAQ;AAC3B,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,iBAAiB,OAAO;AAAA,EAC1C;AACF;AACA,SAAS,qBAAqB,KAAK;AACjC,MAAI,OAAO,MAAM;AACf,WAAO;AAAA,EACT;AACA,MAAI,QAAQ;AACZ,MAAI,QAAQ,CAAC,MAAM,SAAS,EAAE,MAAM;AACpC,SAAO;AACT;AACA,SAAS,SAAS,OAAO;AACvB,SAAO,OAAO,UAAU,YAAY,iBAAiB;AACvD;AACA,SAAS,UAAU,OAAO;AACxB,SAAO,OAAO,UAAU;AAC1B;AACA,SAAS,SAAS,OAAO;AACvB,SAAO,OAAO,UAAU;AAC1B;AACA,SAAS,WAAW,QAAQ;AAC1B,MAAI,MAAM,QAAQ,MAAM,GAAG;AACzB,WAAO,WAAW,OAAO,EAAE;AAAA,EAC7B;AACA,MAAI,kBAAkB,cAAc;AAClC,WAAO;AAAA,EACT,WAAW,kBAAkB,cAAc,kBAAkB,cAAc,kBAAkB,mBAAmB;AAC9G,WAAO;AAAA,EACT,WAAW,SAAS,MAAM,GAAG;AAC3B,WAAO;AAAA,EACT,WAAW,SAAS,MAAM,GAAG;AAC3B,WAAO;AAAA,EACT,WAAW,UAAU,MAAM,GAAG;AAC5B,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,WAAW,GAAG;AACrB,SAAO,CAAC,EAAE,KAAK,EAAE,eAAe,EAAE,QAAQ,EAAE;AAC9C;AACA,SAAS,eAAe,MAAM,OAAO;AACnC,WAAS,IAAI,OAAO,IAAI,MAAM,EAAE,GAAG;AACjC,QAAI,OAAO,MAAM,GAAG;AAClB,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,eAAe,OAAO;AAC7B,QAAM,OAAO,MAAM;AACnB,MAAI,OAAO,GAAG;AACZ,WAAO,CAAC;AAAA,EACV;AACA,QAAM,UAAU,IAAI,MAAM,OAAO,CAAC;AAClC,UAAQ,OAAO,KAAK,MAAM,OAAO;AACjC,WAAS,IAAI,OAAO,GAAG,KAAK,GAAG,EAAE,GAAG;AAClC,YAAQ,KAAK,QAAQ,IAAI,KAAK,MAAM,IAAI;AAAA,EAC1C;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,QAAQ,OAAO,GAAG,YAAY,OAAO;AAC9D,QAAM,MAAM,IAAI,MAAM;AACtB,MAAI,MAAM,WAAW,GAAG;AACtB,UAAM,IAAI,MAAM,MAAM,YAAY,IAAI;AACtC,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAI,KAAK,EAAE,SAAS;AAAA,IACtB;AAAA,EACF,OAAO;AACL,UAAM,IAAI,MAAM;AAChB,UAAM,OAAO,MAAM,MAAM,CAAC;AAC1B,UAAM,MAAM,KAAK,OAAO,CAAC,KAAK,MAAM,MAAM,CAAC,KAAK,YAAY,IAAI;AAChE,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAI,KAAK,kBAAkB,SAAS,IAAI,KAAK,MAAM,GAAG,SAAS;AAAA,IACjE;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,cAAc,OAAO,GAAG,YAAY,OAAO;AAClD,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO,EAAE;AAAA,EACX;AACA,QAAM,OAAO,MAAM,OAAO,CAAC,KAAK,MAAM,MAAM,CAAC,KAAK,YAAY,IAAI;AAClE,MAAI,SAAS,GAAG;AACd,WAAO,CAAC;AAAA,EACV;AACA,MAAI,SAAS,EAAE,QAAQ;AACrB,UAAM,IAAI,MAAM,IAAI,wCAAwC,EAAE,SAAS,YAAY,0BAA0B,KAAK;AAAA,EACpH;AACA,SAAO,kBAAkB,GAAG,OAAO,GAAG,SAAS;AACjD;AACA,SAAS,mBAAmB,MAAM,OAAO;AACvC,QAAM,SAAS,oBAAoB,MAAM,KAAK;AAC9C,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,WAAO,KAAK;AAAA,EACd;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,MAAM,OAAO;AACxC,MAAI,SAAS,QAAQ,UAAU,aAAa,UAAU,aAAa;AACjE,WAAO,IAAI,aAAa,IAAI;AAAA,EAC9B,WAAW,UAAU,SAAS;AAC5B,WAAO,IAAI,WAAW,IAAI;AAAA,EAC5B,WAAW,UAAU,QAAQ;AAC3B,WAAO,IAAI,WAAW,IAAI;AAAA,EAC5B,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACF;AACA,SAAS,0BAA0B,OAAO,OAAO;AAC/C,QAAM,OAAO,MAAM,OAAO,CAAC,MAAM,SAAS,OAAO,MAAM,CAAC;AACxD,MAAI,SAAS,QAAQ,UAAU,WAAW;AACxC,WAAO,cAAc,OAAO,IAAI,aAAa,IAAI,CAAC;AAAA,EACpD,WAAW,UAAU,SAAS;AAC5B,WAAO,cAAc,OAAO,IAAI,WAAW,IAAI,CAAC;AAAA,EAClD,WAAW,UAAU,QAAQ;AAC3B,WAAO,cAAc,OAAO,IAAI,WAAW,IAAI,CAAC;AAAA,EAClD,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACF;AACA,SAAS,mCAAmC,OAAO;AACjD,QAAM,QAAQ,CAAC,YAAY;AACzB,WAAO,OAAO,UAAU,OAAO,KAAK,WAAW,GAAG,MAAM,0EAA0E,SAAS;AAAA,EAC7I,CAAC;AACH;AACA,SAAS,WAAW,MAAM,MAAM,SAAS;AACvC,MAAI,SAAS,GAAG;AACd,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO,KAAK;AAAA,EACd;AACA,MAAI,QAAQ,KAAK,KAAK,SAAS;AAC/B,WAAS,IAAI,GAAG,IAAI,KAAK,SAAS,GAAG,EAAE,GAAG;AACxC,aAAS,QAAQ,KAAK,KAAK;AAAA,EAC7B;AACA,SAAO;AACT;AACA,SAAS,WAAW,OAAO,MAAM,SAAS;AACxC,MAAI,SAAS,GAAG;AACd,WAAO,CAAC;AAAA,EACV,WAAW,SAAS,GAAG;AACrB,WAAO,CAAC,KAAK;AAAA,EACf;AACA,QAAM,OAAO,IAAI,MAAM,IAAI;AAC3B,WAAS,IAAI,GAAG,IAAI,KAAK,SAAS,GAAG,EAAE,GAAG;AACxC,SAAK,KAAK,KAAK,MAAM,QAAQ,QAAQ,EAAE;AACvC,aAAS,KAAK,KAAK,QAAQ;AAAA,EAC7B;AACA,OAAK,KAAK,SAAS,KAAK;AACxB,SAAO;AACT;AACA,SAAS,UAAU,QAAQ;AACzB,SAAO,UAAU,OAAO,QAAQ,OAAO,OAAO,SAAS;AACzD;AAGA,IAAI,4BAA4B;AAChC,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,SAAS;AACnB,SAAK,SAAS;AACd,SAAK,QAAQ,CAAC;AACd,SAAK,eAAe,CAAC;AACrB,SAAK,WAAW,CAAC;AACjB,SAAK,iBAAiB;AACtB,SAAK,iBAAiB;AAAA,EACxB;AAAA,EACA,YAAY,cAAc,UAAU;AAClC,QAAI,KAAK,YAAY,MAAM;AACzB,UAAI,EAAE,IAAI,EAAE,QAAQ,SAAS,KAAK,IAAI,EAAE,QAAQ,MAAM,IAAI;AACxD,gBAAQ,KAAK,YAAY,KAAK,oEAAoE,eAAe;AAAA,MACnH;AAAA,IACF;AACA,SAAK,eAAe;AACpB,SAAK,WAAW;AAAA,EAClB;AAAA,EACA,aAAa,UAAU,cAAc,SAAS;AAC5C,SAAK,aAAa,YAAY,EAAE,cAAc,QAAQ;AACtD,QAAI,KAAK,SAAS,aAAa,MAAM;AACnC,YAAM,YAAY,KAAK,SAAS;AAChC,UAAI,EAAE,IAAI,EAAE,QAAQ,SAAS,KAAK,IAAI,EAAE,QAAQ,MAAM,IAAI;AACxD,gBAAQ,KAAK,qCAAqC,aAAa,YAAY;AAAA,MAC7E;AACA,WAAK,IAAI,UAAU,SAAS;AAAA,IAC9B;AAAA,EACF;AAAA,EACA,MAAM,SAAS,UAAU;AACvB,QAAI,YAAY,KAAK,OAAO;AAC1B,aAAO,KAAK,MAAM;AAAA,IACpB;AACA,SAAK,MAAM,YAAY,MAAM,KAAK,aAAa,QAAQ;AACvD,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,IAAI,UAAU;AACZ,QAAI,YAAY,KAAK,OAAO;AAC1B,aAAO,KAAK,MAAM;AAAA,IACpB;AACA,UAAM,YAAY,KAAK,aAAa,QAAQ;AAC5C,QAAI,UAAU,SAAS,GAAG;AACxB,YAAM,IAAI,MAAM,QAAQ,4EAA4E;AAAA,IACtG;AACA,SAAK,MAAM,YAAY;AACvB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,UAAU,UAAU;AAClB,WAAO,KAAK,IAAI,QAAQ;AAAA,EAC1B;AAAA,EACA,QAAQ,UAAU;AAChB,WAAO,KAAK,IAAI,QAAQ;AAAA,EAC1B;AAAA,EACA,WAAW;AACT,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,WAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,OAAO;AACnB,QAAI,KAAK,aAAa,aAAa,MAAM;AACvC,YAAM,IAAI,MAAM,mBAAmB,yCAAyC;AAAA,IAC9E;AACA,SAAK,MAAM,YAAY;AACvB,QAAI,KAAK,aAAa,UAAU,WAAW,MAAM;AAC/C,WAAK,aAAa,UAAU,QAAQ,KAAK;AAAA,IAC3C;AAAA,EACF;AAAA,EACA,aAAa,UAAU;AACrB,QAAI,KAAK,aAAa,aAAa,MAAM;AACvC,YAAM,IAAI,MAAM,yBAAyB,0CAA0C;AAAA,IACrF;AACA,WAAO,KAAK,aAAa,UAAU,aAAa;AAAA,EAClD;AAAA,EACA,SAAS,OAAO;AACd,SAAK,QAAQ,OAAO,OAAO,CAAC,GAAG,KAAK;AAAA,EACtC;AAAA,EACA,QAAQ;AACN,SAAK,QAAQ,CAAC;AACd,SAAK,WAAW,CAAC;AACjB,SAAK,iBAAiB;AAAA,EACxB;AAAA,EACA,mBAAmB;AACjB,QAAI,OAAO,KAAK,WAAW,eAAe,OAAO,KAAK,OAAO,aAAa,eAAe,OAAO,KAAK,OAAO,SAAS,WAAW,aAAa;AAC3I;AAAA,IACF;AACA,UAAM,YAAY,KAAK,eAAe,KAAK,OAAO,SAAS,MAAM;AACjE,QAAI,6BAA6B,WAAW;AAC1C,YAAM,YAAY,UAAU,2BAA2B,MAAM,GAAG;AAChE,gBAAU,QAAQ,CAAC,aAAa;AAC9B,cAAM,CAAC,KAAK,KAAK,IAAI,SAAS,MAAM,GAAG;AACvC,aAAK,SAAS,OAAO,WAAW,KAAK,KAAK;AAAA,MAC5C,CAAC;AAAA,IACH;AAAA,EACF;AACF;AACA,SAAS,eAAe,aAAa;AACnC,QAAM,SAAS,CAAC;AAChB,cAAY,QAAQ,+BAA+B,CAAC,MAAM,MAAM;AAC9D,gBAAY,QAAQ,EAAE,IAAI,EAAE,EAAE;AAC9B,WAAO,EAAE,KAAK,GAAG;AAAA,EACnB,CAAC;AACD,SAAO;AACT;AACA,SAAS,YAAY,QAAQ,MAAM,OAAO;AACxC,SAAO,mBAAmB,IAAI,KAAK,mBAAmB,SAAS,EAAE;AACnE;AACA,SAAS,WAAW,UAAU,OAAO;AACnC,UAAQ,MAAM,YAAY;AAC1B,MAAI,UAAU,UAAU,UAAU,SAAS;AACzC,WAAO,UAAU;AAAA,EACnB,WAAW,GAAG,CAAC,YAAY,OAAO;AAChC,WAAO,CAAC;AAAA,EACV;AACA,QAAM,IAAI,MAAM,oCAAoC,kBAAkB,WAAW;AACnF;AACA,SAAS,MAAM;AACb,SAAO;AACT;AACA,IAAI,MAAM;AACV,SAAS,qBAAqBI,cAAa;AACzC,QAAMA;AACR;AAGA,IAAI;AACJ,SAAS,qBAAqB;AAC5B,MAAI,mBAAmB,MAAM;AAC3B,QAAI;AACJ,QAAI,OAAO,WAAW,aAAa;AACjC,WAAK;AAAA,IACP,WAAW,OAAO,WAAW,aAAa;AACxC,WAAK;AAAA,IACP,WAAW,OAAO,YAAY,aAAa;AACzC,WAAK;AAAA,IACP,WAAW,OAAO,SAAS,aAAa;AACtC,WAAK;AAAA,IACP,OAAO;AACL,YAAM,IAAI,MAAM,gCAAgC;AAAA,IAClD;AACA,sBAAkB;AAAA,EACpB;AACA,SAAO;AACT;AACA,SAAS,eAAe;AACtB,QAAM,KAAK,mBAAmB;AAC9B,MAAI,GAAG,cAAc,MAAM;AACzB,OAAG,aAA6B,oBAAI,IAAI;AAAA,EAC1C;AACA,SAAO,GAAG;AACZ;AACA,SAAS,UAAU,KAAK,OAAO;AAC7B,QAAM,YAAY,aAAa;AAC/B,MAAI,UAAU,IAAI,GAAG,GAAG;AACtB,WAAO,UAAU,IAAI,GAAG;AAAA,EAC1B,OAAO;AACL,UAAM,YAAY,MAAM;AACxB,cAAU,IAAI,KAAK,SAAS;AAC5B,WAAO,UAAU,IAAI,GAAG;AAAA,EAC1B;AACF;AAGA,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,QAAQ;AACZ,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,MAAM;AACV,IAAI,MAAM;AACV,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,OAAO;AACX,IAAI,QAAQ;AACZ,IAAI,OAAO;AACX,IAAI,QAAQ;AACZ,IAAI,QAAQ;AACZ,IAAI,UAAU;AACd,IAAI,cAAc;AAClB,IAAI,YAAY;AAChB,IAAI,gBAAgB;AACpB,IAAI,cAAc;AAClB,IAAI,iBAAiB;AACrB,IAAI,WAAW;AACf,IAAI,cAAc;AAClB,IAAI,gBAAgB;AACpB,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,cAAc;AAClB,IAAI,UAAU;AACd,IAAI,aAAa;AACjB,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,uBAAuB;AAC3B,IAAI,sBAAsB;AAC1B,IAAI,SAAS;AACb,IAAI,yBAAyB;AAC7B,IAAI,wBAAwB;AAC5B,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,UAAU;AACd,IAAI,SAAS;AACb,IAAI,gBAAgB;AACpB,IAAI,gBAAgB;AACpB,IAAI,eAAe;AACnB,IAAI,wBAAwB;AAC5B,IAAI,sCAAsC;AAC1C,IAAI,qCAAqC;AACzC,IAAI,OAAO;AACX,IAAI,aAAa;AACjB,IAAI,0BAA0B;AAC9B,IAAI,2BAA2B;AAC/B,IAAI,UAAU;AACd,IAAI,SAAS;AACb,IAAI,MAAM;AACV,IAAI,UAAU;AACd,IAAI,MAAM;AACV,IAAI,QAAQ;AACZ,IAAI,MAAM;AACV,IAAI,aAAa;AACjB,IAAI,QAAQ;AACZ,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,gBAAgB;AACpB,IAAI,QAAQ;AACZ,IAAI,WAAW;AACf,IAAI,iBAAiB;AACrB,IAAI,WAAW;AACf,IAAI,WAAW;AACf,IAAI,UAAU;AACd,IAAI,eAAe;AACnB,IAAI,WAAW;AACf,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,WAAW;AACf,IAAI,QAAQ;AACZ,IAAI,QAAQ;AACZ,IAAI,YAAY;AAChB,IAAI,OAAO;AACX,IAAI,YAAY;AAChB,IAAI,WAAW;AACf,IAAI,MAAM;AACV,IAAI,QAAQ;AACZ,IAAI,aAAa;AACjB,IAAI,aAAa;AACjB,IAAI,YAAY;AAChB,IAAI,aAAa;AACjB,IAAI,aAAa;AACjB,IAAI,aAAa;AACjB,IAAI,MAAM;AACV,IAAI,UAAU;AACd,IAAI,MAAM;AACV,IAAI,UAAU;AACd,IAAI,UAAU;AACd,IAAI,cAAc;AAClB,IAAI,YAAY;AAChB,IAAI,gBAAgB;AACpB,IAAI,oBAAoB;AACxB,IAAI,OAAO;AACX,IAAI,MAAM;AACV,IAAI,UAAU;AACd,IAAI,YAAY;AAChB,IAAI,MAAM;AACV,IAAI,cAAc;AAClB,IAAI,WAAW;AACf,IAAI,MAAM;AACV,IAAI,WAAW;AACf,IAAI,sBAAsB;AAC1B,IAAI,sBAAsB;AAC1B,IAAI,sBAAsB;AAC1B,IAAI,WAAW;AACf,IAAI,SAAS;AACb,IAAI,OAAO;AACX,IAAI,QAAQ;AACZ,IAAI,OAAO;AACX,IAAI,MAAM;AACV,IAAI,QAAQ;AACZ,IAAI,OAAO;AACX,IAAI,eAAe;AACnB,IAAI,cAAc;AAClB,IAAI,uBAAuB;AAC3B,IAAI,QAAQ;AACZ,IAAI,OAAO;AACX,IAAI,aAAa;AACjB,IAAI,OAAO;AACX,IAAI,UAAU;AACd,IAAI,wBAAwB;AAC5B,IAAI,4BAA4B;AAChC,IAAI,iBAAiB;AACrB,IAAI,qBAAqB;AACzB,IAAI,QAAQ;AACZ,IAAI,UAAU;AACd,IAAI,QAAQ;AACZ,IAAI,QAAQ;AACZ,IAAI,YAAY;AAChB,IAAI,eAAe;AACnB,IAAI,SAAS;AACb,IAAI,OAAO;AACX,IAAI,QAAQ;AACZ,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,UAAU;AACd,IAAI,WAAW;AACf,IAAI,OAAO;AACX,IAAI,MAAM;AACV,IAAI,iBAAiB;AACrB,IAAI,SAAS;AACb,IAAI,UAAU;AACd,IAAI,sBAAsB;AAC1B,IAAI,gBAAgB;AACpB,IAAI,oBAAoB;AACxB,IAAI,mBAAmB;AACvB,IAAI,gBAAgB;AACpB,IAAI,oBAAoB;AACxB,IAAI,SAAS;AACb,IAAI,eAAe;AACnB,IAAI,eAAe;AACnB,IAAI,cAAc;AAClB,IAAI,yBAAyB;AAC7B,IAAI,MAAM;AACV,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,qBAAqB;AACzB,IAAI,aAAa;AACjB,IAAI,YAAY;AAChB,IAAI,OAAO;AACX,IAAI,aAAa;AACjB,IAAI,mBAAmB;AACvB,IAAI,eAAe;AACnB,IAAI,cAAc;AAClB,IAAI,uBAAuB;AAG3B,SAAS,QAAQ,KAAK;AACpB,MAAI,EAAE,IAAI,EAAE,QAAQ,SAAS,KAAK,IAAI,EAAE,QAAQ,MAAM,IAAI;AACxD,YAAQ,KAAK,GAAG,GAAG;AAAA,EACrB;AACF;AACA,SAAS,OAAO,KAAK;AACnB,MAAI,EAAE,IAAI,EAAE,QAAQ,SAAS,KAAK,IAAI,EAAE,QAAQ,MAAM,IAAI;AACxD,YAAQ,IAAI,GAAG,GAAG;AAAA,EACpB;AACF;AAGA,IAAI,iBAAiB,UAAU,kBAAkB,MAAsB,oBAAI,IAAI,CAAC;AAChF,IAAI,eAAe,UAAU,gBAAgB,MAAsB,oBAAI,IAAI,CAAC;AAC5E,SAAS,UAAU,YAAY,aAAa;AAC1C,QAAM,MAAM,QAAQ,YAAY,WAAW;AAC3C,SAAO,eAAe,IAAI,GAAG;AAC/B;AACA,SAAS,YAAY,YAAY;AAC/B,SAAO,aAAa,IAAI,UAAU;AACpC;AACA,SAAS,qBAAqB,aAAa;AACzC,QAAM,KAAK,eAAe,QAAQ;AAClC,QAAM,SAAS,CAAC;AAChB,SAAO,MAAM;AACX,UAAM,EAAE,MAAM,MAAM,IAAI,GAAG,KAAK;AAChC,QAAI,MAAM;AACR;AAAA,IACF;AACA,UAAM,CAAC,KAAK,MAAM,IAAI;AACtB,UAAM,CAAC,QAAQ,IAAI,IAAI,MAAM,GAAG;AAChC,QAAI,aAAa,aAAa;AAC5B,aAAO,KAAK,MAAM;AAAA,IACpB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,eAAe,QAAQ;AAC9B,QAAM,EAAE,YAAY,YAAY,IAAI;AACpC,QAAM,MAAM,QAAQ,YAAY,WAAW;AAC3C,MAAI,eAAe,IAAI,GAAG,GAAG;AAC3B,SAAK,eAAe,4BAA4B,oCAAoC;AAAA,EACtF;AACA,iBAAe,IAAI,KAAK,MAAM;AAChC;AACA,SAAS,iBAAiB,QAAQ;AAChC,QAAM,EAAE,WAAW,IAAI;AACvB,MAAI,aAAa,IAAI,UAAU,GAAG;AAChC,QAAI,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC1B,WAAK,gCAAgC,aAAa;AAAA,IACpD;AAAA,EACF;AACA,eAAa,IAAI,YAAY,MAAM;AACrC;AACA,SAAS,iBAAiB,YAAY,aAAa;AACjD,QAAM,MAAM,QAAQ,YAAY,WAAW;AAC3C,MAAI,CAAC,eAAe,IAAI,GAAG,GAAG;AAC5B,UAAM,IAAI,MAAM,eAAe,4BAA4B,gCAAgC;AAAA,EAC7F;AACA,iBAAe,OAAO,GAAG;AAC3B;AACA,SAAS,mBAAmB,YAAY;AACtC,MAAI,CAAC,aAAa,IAAI,UAAU,GAAG;AACjC,UAAM,IAAI,MAAM,iBAAiB,2CAA2C;AAAA,EAC9E;AACA,eAAa,OAAO,UAAU;AAChC;AACA,SAAS,sBAAsB,uBAAuB,gBAAgB;AACpE,QAAM,UAAU,qBAAqB,qBAAqB;AAC1D,UAAQ,QAAQ,CAAC,iBAAiB;AAChC,UAAM,kBAAkB,OAAO,OAAO,CAAC,GAAG,cAAc,EAAE,aAAa,eAAe,CAAC;AACvF,mBAAe,eAAe;AAAA,EAChC,CAAC;AACH;AACA,SAAS,QAAQ,YAAY,aAAa;AACxC,SAAO,GAAG,eAAe;AAC3B;AAGA,IAAI,eAAe,CAAC;AACpBF,UAAS,cAAc;AAAA,EACrB,aAAa,MAAM;AAAA,EACnB,QAAQ,MAAM;AAAA,EACd,oCAAoC,MAAM;AAAA,EAC1C,eAAe,MAAM;AAAA,EACrB,mBAAmB,MAAM;AAAA,EACzB,sBAAsB,MAAM;AAAA,EAC5B,iBAAiB,MAAM;AAAA,EACvB,0BAA0B,MAAM;AAAA,EAChC,OAAO,MAAM;AAAA,EACb,gBAAgB,MAAM;AAAA,EACtB,mBAAmB,MAAM;AAAA,EACzB,uBAAuB,MAAM;AAAA,EAC7B,cAAc,MAAM;AAAA,EACpB,aAAa,MAAM;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,OAAO,MAAM;AAAA,EACb,eAAe,MAAM;AAAA,EACrB,SAAS,MAAM;AAAA,EACf,mBAAmB,MAAM;AAAA,EACzB,wBAAwB,MAAM;AAAA,EAC9B,iBAAiB,MAAM;AAAA,EACvB,WAAW,MAAM;AAAA,EACjB,YAAY,MAAM;AAAA,EAClB,YAAY,MAAM;AAAA,EAClB,wBAAwB,MAAM;AAAA,EAC9B,WAAW,MAAM;AAAA,EACjB,YAAY,MAAM;AAAA,EAClB,OAAO,MAAM;AAAA,EACb,UAAU,MAAM;AAAA,EAChB,WAAW,MAAM;AAAA,EACjB,eAAe,MAAM;AAAA,EACrB,UAAU,MAAM;AAAA,EAChB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,YAAY,MAAM;AAAA,EAClB,oBAAoB,MAAM;AAAA,EAC1B,2BAA2B,MAAM;AAAA,EACjC,qBAAqB,MAAM;AAAA,EAC3B,gBAAgB,MAAM;AAAA,EACtB,mBAAmB,MAAM;AAAA,EACzB,KAAK,MAAM;AAAA,EACX,gBAAgB,MAAM;AAAA,EACtB,aAAa,MAAM;AAAA,EACnB,aAAa,MAAM;AAAA,EACnB,UAAU,MAAM;AAAA,EAChB,SAAS,MAAM;AAAA,EACf,cAAc,MAAM;AAAA,EACpB,eAAe,MAAM;AAAA,EACrB,qBAAqB,MAAM;AAAA,EAC3B,cAAc,MAAM;AAAA,EACpB,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,MAAM,MAAM;AAAA,EACZ,eAAe,MAAM;AAAA,EACrB,cAAc,MAAM;AACtB,CAAC;AAGD,IAAI,cAAc,QAAQ,aAAa,CAAC;AACxC,IAAI,OAAO,YAAY,WAAW;AAClC,SAAS,UAAU,KAAK;AACtB,SAAO,KAAK,WAAW,KAAK,MAAM,EAAE;AACtC;AACA,IAAI,KAAK,UAAU,kBAAkB;AACrC,IAAI,KAAK,UAAU,kBAAkB;AACrC,IAAI,KAAK,UAAU,kBAAkB;AACrC,SAAS,SAAS,KAAK;AACrB,SAAO,IAAI,IAAI,IAAI,KAAK,EAAE,CAAC;AAC7B;AACA,SAAS,OAAO,GAAG,QAAQ,UAAU;AACnC,QAAM,QAAQ,EAAE,MAAM,QAAQ,SAAS,QAAQ;AAC/C,SAAO,KAAK,UAAU,MAAM,KAAK,KAAK,GAAG,MAAM,IAAI;AACrD;AACA,SAAS,QAAQ,GAAG,QAAQ;AAC1B,SAAO,OAAO,GAAG,QAAQ,CAAC;AAC5B;AACA,SAAS,QAAQ,GAAG,QAAQ;AAC1B,SAAO,OAAO,GAAG,QAAQ,CAAC;AAC5B;AACA,SAAS,SAAS,KAAK,OAAO;AAC5B,SAAO,UAAU,IAAI,MAAM,IAAI,KAAK,KAAK,EAAE,GAAG,IAAI,IAAI,KAAK,KAAK,CAAC;AACnE;AACA,SAAS,UAAU,GAAG,GAAG,OAAO,UAAU,kBAAkB,GAAG;AAC7D,MAAI,IAAI,EAAE,IAAI,CAAC,EAAE,IAAI,IAAI;AACzB,MAAI,EAAE,IAAI,EAAE,KAAK,EAAE,CAAC;AACpB,MAAI,IAAI,EAAE,IAAI,CAAC,EAAE,IAAI,IAAI;AACzB,MAAI,EAAE,IAAI,EAAE,KAAK,EAAE,CAAC;AACpB,MAAI,EAAE,IAAI,IAAI;AACd,SAAO;AACT;AACA,SAAS,uBAAuB,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG;AAChD,MAAI,EAAE,IAAI,CAAC;AACX,MAAI,SAAS,EAAE,IAAI,CAAC,EAAE,IAAI,CAAC,GAAG,EAAE;AAChC,QAAM,IAAI;AACV,MAAI,EAAE,IAAI,CAAC;AACX,MAAI,EAAE,IAAI,CAAC;AACX,MAAI,EAAE,IAAI,SAAS,GAAG,EAAE,CAAC;AACzB,SAAO,CAAC,EAAE,IAAI,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC;AAC5B;AACA,SAAS,0BAA0B,GAAG,QAAQ,GAAG,GAAG;AAClD,SAAO,uBAAuB,QAAQ,GAAG,MAAM,GAAG,QAAQ,GAAG,SAAS,CAAC,GAAG,QAAQ,GAAG,SAAS,EAAE,GAAG,QAAQ,GAAG,SAAS,EAAE,GAAG,GAAG,CAAC;AAClI;AACA,SAAS,aAAa,GAAG,MAAM,EAAE,QAAQ;AACvC,MAAI,OAAO,GAAG;AACZ,UAAM,OAAO,GAAG,IAAI,MAAM,CAAC;AAC3B,UAAM,IAAI,QAAQ,GAAG,CAAC,EAAE,IAAI,EAAE;AAC9B,UAAM,IAAI,QAAQ,GAAG,MAAM,CAAC;AAC5B,UAAM,IAAI,SAAS,GAAG,EAAE,EAAE,IAAI,IAAI,EAAE,IAAI,CAAC;AACzC,UAAM,IAAI,SAAS,GAAG,EAAE,EAAE,IAAI,CAAC,EAAE,IAAI,IAAI;AACzC,WAAO,UAAU,GAAG,GAAG,IAAI;AAAA,EAC7B;AACA,MAAI,OAAO,GAAG;AACZ,UAAM,OAAO,GAAG,IAAI,MAAM,CAAC;AAC3B,UAAM,IAAI,QAAQ,GAAG,CAAC;AACtB,WAAO,UAAU,EAAE,IAAI,CAAC,EAAE,IAAI,GAAG,GAAG,QAAQ,GAAG,MAAM,CAAC,GAAG,IAAI;AAAA,EAC/D;AACA,MAAI,MAAM,GAAG;AACX,UAAM,IAAI,EAAE;AACZ,UAAM,IAAI,EAAE,OAAO;AACnB,UAAM,IAAI,EAAE,MAAM;AAClB,UAAM,IAAI,KAAK,KAAK;AACpB,UAAM,IAAI,OAAO,KAAK;AACtB,WAAO,SAAS,GAAG,IAAI,CAAC,EAAE,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE;AAAA,EAClD;AACA,SAAO;AACT;AACA,SAAS,cAAc,GAAG,MAAM,EAAE,QAAQ;AACxC,QAAM,OAAO,GAAG,IAAI,MAAM,CAAC;AAC3B,QAAM,IAAI,QAAQ,GAAG,CAAC,EAAE,IAAI,EAAE;AAC9B,QAAM,IAAI,QAAQ,GAAG,CAAC;AACtB,QAAM,IAAI,QAAQ,GAAG,MAAM,CAAC,EAAE,IAAI,IAAI;AACtC,QAAM,IAAI,QAAQ,GAAG,MAAM,EAAE,EAAE,IAAI,EAAE;AACrC,SAAO,UAAU,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,IAAI,SAAS,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,EAAE,IAAI,SAAS,EAAE,IAAI,EAAE,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,IAAI;AAClH;AACA,SAAS,cAAc,GAAG,MAAM,EAAE,QAAQ;AACxC,QAAM,OAAO,GAAG,IAAI,MAAM,CAAC;AAC3B,QAAM,IAAI,QAAQ,GAAG,CAAC,EAAE,IAAI,EAAE;AAC9B,QAAM,IAAI,QAAQ,GAAG,CAAC;AACtB,QAAM,IAAI,QAAQ,GAAG,MAAM,CAAC,EAAE,IAAI,IAAI;AACtC,QAAM,IAAI,QAAQ,GAAG,MAAM,EAAE,EAAE,IAAI,EAAE;AACrC,QAAM,IAAI,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,IAAI,SAAS,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC;AAC3D,QAAM,IAAI,UAAU,GAAG,EAAE,IAAI,SAAS,EAAE,IAAI,EAAE,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,IAAI;AAClE,QAAM,IAAI,QAAQ,GAAG,EAAE,EAAE,IAAI,IAAI;AACjC,QAAM,IAAI,QAAQ,GAAG,EAAE;AACvB,QAAM,IAAI,EAAE,IAAI,QAAQ,GAAG,MAAM,EAAE,CAAC,EAAE,IAAI,IAAI;AAC9C,QAAM,IAAI,EAAE,IAAI,QAAQ,GAAG,MAAM,EAAE,CAAC,EAAE,IAAI,IAAI;AAC9C,SAAO,UAAU,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,IAAI,SAAS,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,EAAE,IAAI,SAAS,EAAE,IAAI,CAAC,GAAG,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,IAAI;AACjH;AACA,SAAS,cAAc,GAAG,MAAM,EAAE,QAAQ;AACxC,QAAM,OAAO,KAAK,WAAW,IAAI,IAAI;AACrC,MAAI,OAAO,IAAI;AACb,QAAI,OAAO,IAAI;AACb,aAAO,aAAa,GAAG,GAAG;AAAA,IAC5B,OAAO;AACL,aAAO,cAAc,GAAG,GAAG;AAAA,IAC7B;AAAA,EACF,WAAW,OAAO,IAAI;AACpB,WAAO,cAAc,GAAG,GAAG;AAAA,EAC7B;AACA,MAAI,IAAI;AACR,MAAI,IAAI,KAAK,IAAI,EAAE,EAAE,IAAI,GAAG;AAC5B,MAAI,IAAI,SAAS,EAAE,IAAI,EAAE,EAAE,IAAI,GAAG,CAAC,EAAE,IAAI,EAAE;AAC3C,MAAI,IAAI,CAAC,KAAK,OAAO,KAAK,KAAK;AAC/B,MAAI,IAAI,CAAC,KAAK,OAAO,KAAK,KAAK;AAC/B,MAAI,EAAE,IAAI,EAAE,EAAE,IAAI,QAAQ,GAAG,CAAC,CAAC;AAC/B,MAAI,SAAS;AACb,QAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,QAAM,SAAS,OAAO,MAAM,IAAI,MAAM;AACtC,KAAG;AACD,QAAI,SAAS,EAAE,IAAI,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,IAAI,QAAQ,GAAG,SAAS,CAAC,CAAC,GAAG,EAAE,EAAE,IAAI,EAAE;AACvE,QAAI,SAAS,EAAE,IAAI,EAAE,EAAE,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC,GAAG,EAAE,EAAE,IAAI,EAAE;AACjE,QAAI,EAAE,IAAI,EAAE,EAAE;AACd,QAAI,EAAE,IAAI,EAAE,EAAE,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC;AAC3C,QAAI,SAAS,EAAE,IAAI,EAAE,EAAE,GAAG,EAAE,EAAE,IAAI,EAAE;AACpC,QAAI,0BAA0B,GAAG,QAAQ,EAAE,GAAG,IAAI,EAAE,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC;AAClE,QAAI,0BAA0B,GAAG,SAAS,IAAI,EAAE,IAAI,EAAE,EAAE,GAAG,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC,CAAC;AACzF,KAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AACd,cAAU;AAAA,EACZ,SAAS,WAAW;AACpB,QAAM,OAAO,GAAG,IAAI,EAAE,IAAI,GAAG,EAAE,IAAI,CAAC,CAAC;AACrC,WAAS;AACT,IAAE,KAAK,EAAE,GAAG,IAAI,MAAM,IAAI,EAAE;AAC5B,IAAE,KAAK,EAAE,GAAG,IAAI,EAAE,EAAE;AACpB,IAAE,KAAK,EAAE,GAAG,IAAI,EAAE,EAAE;AACpB,MAAI,SAAS,EAAE,IAAI,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,IAAI,QAAQ,GAAG,SAAS,CAAC,CAAC,GAAG,EAAE,EAAE,IAAI,IAAI;AACzE,MAAI,SAAS,EAAE,IAAI,EAAE,EAAE,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC,GAAG,EAAE,EAAE,IAAI,IAAI;AACnE,MAAI,EAAE,IAAI,EAAE,GAAG,IAAI,CAAC,CAAC;AACrB,MAAI,EAAE,IAAI,EAAE,GAAG,IAAI,CAAC,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC,CAAC;AAClD,MAAI,SAAS,EAAE,IAAI,EAAE,EAAE,GAAG,EAAE,EAAE,IAAI,IAAI;AACtC,MAAI,0BAA0B,GAAG,QAAQ,EAAE,GAAG,IAAI,IAAI,GAAG,EAAE,IAAI,EAAE,EAAE,CAAC;AACpE,MAAI,0BAA0B,GAAG,SAAS,IAAI,EAAE,IAAI,EAAE,EAAE,GAAG,EAAE,IAAI,QAAQ,GAAG,SAAS,EAAE,CAAC,CAAC;AACzF,GAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AACd,SAAO,UAAU,UAAU,EAAE,IAAI,EAAE,IAAI,IAAI,EAAE,IAAI,SAAS,CAAC,EAAE,IAAI,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,UAAU,EAAE,IAAI,EAAE,IAAI,IAAI,EAAE,IAAI,CAAC,GAAG,IAAI;AACxH;AAGA,SAAS,kBAAkB,OAAO,OAAO;AACvC,MAAI,UAAU,UAAU;AACtB,WAAO,aAAa,KAAK;AAAA,EAC3B;AACA,SAAO,aAAa,CAAC,KAAK,GAAG,KAAK;AACpC;AACA,SAAS,mBAAmB,GAAG,OAAO;AACpC,SAAO,aAAa,gBAAgB,UAAU,aAAa,aAAa,cAAc,UAAU,WAAW,aAAa,cAAc,UAAU;AAClJ;AACA,SAAS,aAAa,GAAG,OAAO;AAC9B,MAAI,UAAU,UAAU;AACtB,UAAM,IAAI,MAAM,2CAA2C;AAAA,EAC7D;AACA,MAAI,MAAM,QAAQ,CAAC,GAAG;AACpB,QAAI,QAAQ,CAAC;AAAA,EACf;AACA,MAAI,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC1B,6BAAyB,GAAG,KAAK;AAAA,EACnC;AACA,MAAI,mBAAmB,GAAG,KAAK,GAAG;AAChC,WAAO;AAAA,EACT;AACA,MAAI,SAAS,QAAQ,UAAU,aAAa,UAAU,aAAa;AACjE,WAAO,IAAI,aAAa,CAAC;AAAA,EAC3B,WAAW,UAAU,SAAS;AAC5B,WAAO,IAAI,WAAW,CAAC;AAAA,EACzB,WAAW,UAAU,QAAQ;AAC3B,UAAM,OAAO,IAAI,WAAW,EAAE,MAAM;AACpC,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAI,KAAK,MAAM,EAAE,EAAE,MAAM,GAAG;AAC1B,aAAK,KAAK;AAAA,MACZ;AAAA,IACF;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACF;AACA,SAAS,MAAM;AACb,SAAO,IAAI,EAAE,SAAS,IAAI;AAC5B;AACA,SAAS,OAAO,MAAM,cAAc;AAClC,SAAO,IAAI,EAAE,SAAS,MAAM,MAAM,YAAY;AAChD;AACA,SAAS,aAAa,GAAG,WAAW,SAAS;AAC3C,aAAW,YAAY;AACvB,SAAO,IAAI,EAAE,SAAS,OAAO,GAAG,QAAQ;AAC1C;AACA,SAAS,aAAa,OAAO,WAAW,SAAS;AAC/C,aAAW,YAAY;AACvB,SAAO,IAAI,EAAE,SAAS,OAAO,OAAO,QAAQ;AAC9C;AAGA,IAAI,WAAW,MAAM;AAAA,EACnB,YAAY,cAAc,QAAQ;AAChC,SAAK,eAAe;AACpB,SAAK,SAAS;AACd,QAAI,UAAU,MAAM;AAClB,WAAK,SAAS,IAAI,OAAO;AAAA,IAC3B;AAAA,EACF;AAAA,EACA,cAAc,YAAY,QAAQ,GAAG;AACnC,QAAI;AACJ,UAAM,sBAAsB,MAAM;AAChC,gBAAU,EAAE;AAAA,IACd;AACA,QAAI;AACJ,UAAM,QAAQ,IAAI;AAClB,QAAI,KAAK,aAAa,eAAe,GAAG;AACtC,cAAQ,KAAK,aAAa,KAAK,mBAAmB;AAAA,IACpD,OAAO;AACL,0BAAoB;AACpB,iBAAW,UAAU,SAAS;AAC5B,eAAO,SAAS;AAAA,MAClB;AACA,cAAQ,QAAQ,QAAQ,EAAE,UAAU,IAAI,IAAI,MAAM,CAAC;AAAA,IACrD;AACA,QAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,eAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,cAAM,SAAS,QAAQ;AACvB,eAAO,KAAK,EAAE,KAAK,CAAC,eAAe;AACjC,oCAA0B,YAAY,OAAO,OAAO,UAAU;AAAA,QAChE,CAAC;AAAA,MACH;AAAA,IACF;AACA,UAAM,gBAAgB;AAAA,MACpB;AAAA,MACA;AAAA,MACA;AAAA,MACA,QAAQ,MAAM,KAAK,CAAC,WAAW,OAAO,QAAQ;AAAA,MAC9C,WAAW,MAAM,KAAK,CAAC,WAAW,OAAO,uBAAuB,OAAO,OAAO,oBAAoB,IAAI,EAAE;AAAA,IAC1G;AACA,WAAO;AAAA,EACT;AAAA,EACA,iBAAiB,eAAe;AAC9B,UAAM,EAAE,YAAY,SAAS,QAAQ,QAAQ,UAAU,IAAI;AAC3D,YAAQ,QAAQ,CAAC,WAAW;AAC1B,cAAQ,IAAI,CAAC,OAAO,KAAK,GAAG,QAAQ,SAAS,CAAC,EAAE,KAAK,CAAC,mBAAmB;AACvE,aAAK,OAAO,iBAAiB,YAAY,QAAQ,eAAe,IAAI,eAAe,IAAI,QAAQ,eAAe,EAAE;AAAA,MAClH,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AACF;AACA,SAAS,0BAA0B,MAAM,OAAO,YAAY;AAC1D,MAAI,UAAU,WAAW;AACvB,WAAO;AAAA,EACT;AACA,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,MAAM,KAAK;AACjB,QAAI,MAAM,GAAG,KAAK,CAAC,SAAS,GAAG,GAAG;AAChC,cAAQ,KAAK,SAAS,yBAAyB,aAAa;AAC5D,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,IAAI,SAAS,MAAM;AAAA,EACjB,iBAAiB,MAAM,QAAQ,MAAM,QAAQ,QAAQ,WAAW;AAC9D,UAAM,QAAQ,OAAO,WAAW,WAAW,SAAS,GAAG,YAAY,CAAC,IAAI,OAAO;AAC/E,UAAM,aAAa,SAAS,MAAM,EAAE;AACpC,UAAM,OAAO,OAAO;AACpB,UAAM,OAAO,OAAO;AACpB,UAAM,QAAQ,SAAS,OAAO,MAAM,SAAS,GAAG,EAAE;AAClD,QAAI,yBAAyB;AAC7B,eAAW,SAAS,QAAQ;AAC1B,YAAM,SAAS,OAAO;AACtB,UAAI,UAAU,MAAM;AAClB,cAAM,aAAa,OAAO,SAAS,OAAO;AAC1C,cAAM,YAAY,WAAW;AAC7B,kCAA0B,GAAG,UAAU,cAAc,YAAY,IAAI,aAAa;AAAA,MACpF;AAAA,IACF;AACA,YAAQ,IAAI,KAAK,gBAAgB,WAAW,SAAS,WAAW,UAAU,4BAA4B,aAAa,oBAAoB,aAAa,cAAc,iBAAiB,gBAAgB,kBAAkB;AAAA,EACvN;AACF;AAGA,SAAS,qBAAqB,MAAM,IAAI,GAAG;AACzC,QAAM,eAAe,CAAC;AACtB,QAAM,aAAa,CAAC;AACpB,WAAS,IAAI,GAAG,IAAI,GAAG,QAAQ,KAAK;AAClC,iBAAa,GAAG,GAAG,MAAM;AAAA,EAC3B;AACA,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,OAAO,KAAK;AAClB,UAAM,aAAa,KAAK;AACxB,eAAW,aAAa,YAAY;AAClC,YAAM,SAAS,WAAW;AAC1B,UAAI,gBAAgB;AACpB,eAAS,IAAI,GAAG,IAAI,GAAG,QAAQ,KAAK;AAClC,YAAI,aAAa,OAAO,KAAK;AAC3B,eAAK,QAAQ,QAAQ,CAAC,WAAW,aAAa,OAAO,MAAM,IAAI;AAC/D,0BAAgB;AAChB,qBAAW,KAAK,MAAM;AACtB;AAAA,QACF;AAAA,MACF;AACA,UAAI,eAAe;AACjB;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,QAAM,iBAAiB,CAAC;AACxB,iBAAe,EAAE,MAAM;AACvB,QAAM,WAAW,CAAC;AAClB,WAAS,IAAI,KAAK,SAAS,GAAG,KAAK,GAAG,KAAK;AACzC,UAAM,OAAO,KAAK;AAClB,UAAM,aAAa,KAAK;AACxB,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,QAAQ,KAAK;AAC5C,UAAI,eAAe,KAAK,QAAQ,GAAG,KAAK;AACtC,mBAAW,aAAa,YAAY;AAClC,yBAAe,WAAW,WAAW,MAAM;AAC3C,mBAAS,KAAK,MAAM;AAAA,QACtB;AACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,QAAM,eAAe,CAAC;AACtB,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,OAAO,KAAK;AAClB,QAAI,WAAW,KAAK,OAAO,SAAS,KAAK,KAAK;AAC5C,YAAM,eAAe,CAAC;AACtB,iBAAW,aAAa,KAAK,QAAQ;AACnC,cAAM,YAAY,KAAK,OAAO;AAC9B,YAAI,aAAa,UAAU,KAAK;AAC9B,uBAAa,aAAa;AAAA,QAC5B;AAAA,MACF;AACA,YAAM,aAAa,OAAO,OAAO,CAAC,GAAG,IAAI;AACzC,iBAAW,SAAS;AACpB,iBAAW,UAAU,KAAK;AAC1B,mBAAa,KAAK,UAAU;AAAA,IAC9B;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,uBAAuB,8BAA8B,cAAc,OAAO,MAAM;AACvF,WAAS,IAAI,aAAa,SAAS,GAAG,KAAK,GAAG,KAAK;AACjD,UAAM,OAAO,aAAa;AAC1B,UAAM,MAAM,CAAC;AACb,SAAK,QAAQ,QAAQ,CAAC,MAAM;AAC1B,YAAM,aAAa,6BAA6B,EAAE;AAClD,UAAI,cAAc,MAAM;AACtB,YAAI,KAAK,UAAU;AAAA,MACrB,OAAO;AACL,YAAI,KAAK,IAAI;AAAA,MACf;AAAA,IACF,CAAC;AACD,QAAI,KAAK,YAAY,MAAM;AACzB,YAAM,IAAI,MAAM,4DAA4D,KAAK,aAAa;AAAA,IAChG;AACA,UAAM,iBAAiB,KAAK,SAAS,GAAG;AACxC,eAAW,aAAa,KAAK,QAAQ;AACnC,UAAI,EAAE,aAAa,iBAAiB;AAClC,cAAM,IAAI,MAAM,iCAAiC,yCAAyC,OAAO,KAAK,cAAc,IAAI;AAAA,MAC1H;AACA,YAAM,KAAK,MAAM,MAAM,eAAe,WAAW,CAAC;AAClD,UAAI,GAAG,UAAU,WAAW;AAC1B,cAAM,IAAI,MAAM,4BAA4B,KAAK,qCAAqC,iDAAiD,GAAG,QAAQ;AAAA,MACpJ;AACA,YAAM,IAAI,KAAK,OAAO;AACtB,UAAI,CAAC,YAAY,GAAG,OAAO,EAAE,KAAK,GAAG;AACnC,cAAM,IAAI,MAAM,4BAA4B,KAAK,sCAAsC,yBAAyB,GAAG,wDAAwD,EAAE,QAAQ;AAAA,MACvL;AACA,UAAI,6BAA6B,EAAE,OAAO,MAAM;AAC9C,qCAA6B,EAAE,MAAM;AAAA,MACvC,OAAO;AACL,cAAM,cAAc,6BAA6B,EAAE;AACnD,qCAA6B,EAAE,MAAM,KAAK,aAAa,EAAE;AACzD,oBAAY,QAAQ;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,wBAAwB;AAC5B,IAAI,6BAA6B;AACjC,IAAI,wBAAwB;AAC5B,SAAS,eAAe,MAAM,OAAO,OAAO,SAAS;AACnD,QAAM,UAAU,eAAe,KAAK;AACpC,QAAM,YAAY,wBAAwB,MAAM,OAAO,OAAO,OAAO;AACrE,QAAM,OAAO,MAAM;AACnB,QAAM,YAAY,kBAAkB,MAAM,OAAO,OAAO,SAAS,SAAS;AAC1E,QAAM,QAAQ,CAAC,QAAQ;AACvB,MAAI,SAAS;AACX,UAAM,KAAK,YAAY,OAAO;AAC9B,UAAM,KAAK,WAAW,MAAM;AAC5B,UAAM,KAAK,aAAa,QAAQ;AAChC,UAAM,KAAK,WAAW;AAAA,EACxB;AACA,QAAM,KAAK,UAAU,IAAI,CAAC,MAAM,SAAS,CAAC,EAAE,KAAK,IAAI,CAAC;AACtD,SAAO,MAAM,KAAK,IAAI;AACxB;AACA,SAAS,wBAAwB,MAAM,OAAO,OAAO,SAAS;AAC5D,QAAM,IAAI,cAAc,KAAK;AAC7B,QAAM,UAAU,QAAQ,QAAQ,SAAS;AACzC,QAAM,YAAY,IAAI,MAAM,OAAO,EAAE,KAAK,CAAC;AAC3C,QAAM,OAAO,MAAM;AACnB,QAAM,iBAAiB,UAAU,cAAc,oBAAoB,IAAI,IAAI;AAC3E,MAAI,OAAO,GAAG;AACZ,aAAS,MAAM,GAAG,MAAM,IAAI,SAAS,OAAO;AAC1C,YAAM,SAAS,MAAM;AACrB,eAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,kBAAU,KAAK,KAAK,IAAI,UAAU,IAAI,YAAY,eAAe,SAAS,IAAI,GAAG,KAAK,EAAE,MAAM;AAAA,MAChG;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,YAAY,KAAK,MAAM,OAAO;AACrC,MAAI;AACJ,MAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,aAAS,GAAG,WAAW,IAAI,GAAG,QAAQ,qBAAqB,CAAC,OAAO,WAAW,IAAI,GAAG,QAAQ,qBAAqB,CAAC;AAAA,EACrH,WAAW,SAAS,GAAG,GAAG;AACxB,aAAS,IAAI;AAAA,EACf,WAAW,UAAU,QAAQ;AAC3B,aAAS,gBAAgB,GAAG;AAAA,EAC9B,OAAO;AACL,aAAS,WAAW,IAAI,QAAQ,qBAAqB,CAAC,EAAE,SAAS;AAAA,EACnE;AACA,SAAO,SAAS,QAAQ,IAAI;AAC9B;AACA,SAAS,gBAAgB,GAAG;AAC1B,SAAO,MAAM,IAAI,UAAU;AAC7B;AACA,SAAS,kBAAkB,MAAM,OAAO,OAAO,SAAS,WAAW,SAAS,MAAM;AAChF,QAAM,oBAAoB,UAAU,cAAc,IAAI;AACtD,QAAM,OAAO,MAAM;AACnB,QAAM,OAAO,MAAM;AACnB,MAAI,SAAS,GAAG;AACd,QAAI,UAAU,aAAa;AACzB,YAAM,eAAe,oBAAoB,IAAI;AAC7C,aAAO,CAAC,YAAY,aAAa,IAAI,GAAG,KAAK,CAAC;AAAA,IAChD;AACA,QAAI,UAAU,QAAQ;AACpB,aAAO,CAAC,gBAAgB,KAAK,EAAE,CAAC;AAAA,IAClC;AACA,WAAO,CAAC,KAAK,GAAG,SAAS,CAAC;AAAA,EAC5B;AACA,MAAI,SAAS,GAAG;AACd,QAAI,OAAO,uBAAuB;AAChC,YAAM,gBAAgB,6BAA6B;AACnD,UAAI,YAAY,MAAM,KAAK,KAAK,MAAM,GAAG,aAAa,CAAC;AACvD,UAAI,WAAW,MAAM,KAAK,KAAK,OAAO,OAAO,8BAA8B,mBAAmB,OAAO,iBAAiB,CAAC;AACvH,UAAI,UAAU,aAAa;AACzB,oBAAY,oBAAoB,SAAS;AACzC,mBAAW,oBAAoB,QAAQ;AAAA,MACzC;AACA,aAAO;AAAA,QACL,MAAM,UAAU,IAAI,CAAC,GAAG,MAAM,YAAY,GAAG,UAAU,IAAI,KAAK,CAAC,EAAE,KAAK,IAAI,IAAI,YAAY,SAAS,IAAI,CAAC,GAAG,MAAM,YAAY,GAAG,UAAU,OAAO,6BAA6B,IAAI,KAAK,CAAC,EAAE,KAAK,IAAI,IAAI;AAAA,MAC3M;AAAA,IACF;AACA,UAAM,cAAc,UAAU,cAAc,oBAAoB,IAAI,IAAI,MAAM,KAAK,IAAI;AACvF,WAAO;AAAA,MACL,MAAM,YAAY,IAAI,CAAC,GAAG,MAAM,YAAY,GAAG,UAAU,IAAI,KAAK,CAAC,EAAE,KAAK,IAAI,IAAI;AAAA,IACpF;AAAA,EACF;AACA,QAAM,WAAW,MAAM,MAAM,CAAC;AAC9B,QAAM,aAAa,QAAQ,MAAM,CAAC;AAClC,QAAM,SAAS,QAAQ,KAAK;AAC5B,QAAM,QAAQ,CAAC;AACf,MAAI,OAAO,uBAAuB;AAChC,aAAS,IAAI,GAAG,IAAI,4BAA4B,KAAK;AACnD,YAAM,QAAQ,IAAI;AAClB,YAAM,MAAM,QAAQ;AACpB,YAAM,KAAK,GAAG,kBAAkB,KAAK,MAAM,OAAO,GAAG,GAAG,UAAU,OAAO,YAAY,WAAW,KAAK,CAAC;AAAA,IACxG;AACA,UAAM,KAAK,KAAK;AAChB,aAAS,IAAI,OAAO,4BAA4B,IAAI,MAAM,KAAK;AAC7D,YAAM,QAAQ,IAAI;AAClB,YAAM,MAAM,QAAQ;AACpB,YAAM,KAAK,GAAG,kBAAkB,KAAK,MAAM,OAAO,GAAG,GAAG,UAAU,OAAO,YAAY,WAAW,MAAM,OAAO,CAAC,CAAC;AAAA,IACjH;AAAA,EACF,OAAO;AACL,aAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,YAAM,QAAQ,IAAI;AAClB,YAAM,MAAM,QAAQ;AACpB,YAAM,KAAK,GAAG,kBAAkB,KAAK,MAAM,OAAO,GAAG,GAAG,UAAU,OAAO,YAAY,WAAW,MAAM,OAAO,CAAC,CAAC;AAAA,IACjH;AAAA,EACF;AACA,QAAM,MAAM,SAAS,IAAI,MAAM;AAC/B,QAAM,KAAK,MAAM,MAAM,KAAK;AAC5B,WAAS,IAAI,GAAG,IAAI,MAAM,SAAS,GAAG,KAAK;AACzC,UAAM,KAAK,MAAM,MAAM,KAAK;AAAA,EAC9B;AACA,MAAI,aAAa;AACjB,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,kBAAc;AAAA,EAChB;AACA,QAAM,MAAM,SAAS,KAAK,MAAM,MAAM,MAAM,SAAS,KAAK,OAAO,SAAS,KAAK;AAC/E,SAAO;AACT;AACA,SAAS,oBAAoB,MAAM;AACjC,QAAM,gBAAgB,CAAC;AACvB,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,GAAG;AACvC,kBAAc,KAAK,CAAC,KAAK,IAAI,KAAK,IAAI,EAAE,CAAC;AAAA,EAC3C;AACA,SAAO;AACT;AAGA,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,OAAO,OAAO,QAAQ;AAChC,SAAK,QAAQ;AACb,SAAK,QAAQ,MAAM,MAAM;AACzB,SAAK,OAAO,cAAc,KAAK;AAC/B,QAAI,UAAU,MAAM;AAClB,YAAM,IAAI,OAAO;AACjB,aAAO,MAAM,KAAK,MAAM,MAAM,qBAAqB,qDAAqD,KAAK,QAAQ;AAAA,IACvH;AACA,QAAI,UAAU,aAAa;AACzB,YAAM,IAAI,MAAM,4JAA4J;AAAA,IAC9K;AACA,SAAK,SAAS,UAAU,kBAAkB,OAAO,KAAK,IAAI;AAC1D,SAAK,UAAU,eAAe,KAAK;AAAA,EACrC;AAAA,EACA,IAAI,UAAU,MAAM;AAClB,QAAI,KAAK,WAAW,GAAG;AACrB,aAAO,CAAC,CAAC;AAAA,IACX;AACA,WAAO,KAAK,WAAW,KAAK,MAAM,MAAM,uCAAuC,KAAK,gCAAgC,KAAK,OAAO;AAChI,UAAM,QAAQ,KAAK,WAAW,IAAI;AAClC,SAAK,OAAO,SAAS;AAAA,EACvB;AAAA,EACA,OAAO,MAAM;AACX,QAAI,KAAK,WAAW,GAAG;AACrB,aAAO,CAAC,CAAC;AAAA,IACX;AACA,QAAI,IAAI;AACR,eAAW,OAAO,MAAM;AACtB,UAAI,MAAM,KAAK,OAAO,KAAK,MAAM,IAAI;AACnC,cAAM,MAAM,qCAAqC,wBAAwB,KAAK;AAC9E,cAAM,IAAI,MAAM,GAAG;AAAA,MACrB;AACA;AAAA,IACF;AACA,QAAI,QAAQ,KAAK,KAAK,SAAS;AAC/B,aAAS,KAAK,GAAG,KAAK,KAAK,SAAS,GAAG,EAAE,IAAI;AAC3C,eAAS,KAAK,QAAQ,MAAM,KAAK;AAAA,IACnC;AACA,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EACA,WAAW,MAAM;AACf,QAAI,KAAK,SAAS,GAAG;AACnB,aAAO;AAAA,IACT,WAAW,KAAK,SAAS,GAAG;AAC1B,aAAO,KAAK;AAAA,IACd;AACA,QAAI,QAAQ,KAAK,KAAK,SAAS;AAC/B,aAAS,IAAI,GAAG,IAAI,KAAK,SAAS,GAAG,EAAE,GAAG;AACxC,eAAS,KAAK,QAAQ,KAAK,KAAK;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EACA,WAAW,OAAO;AAChB,QAAI,KAAK,SAAS,GAAG;AACnB,aAAO,CAAC;AAAA,IACV,WAAW,KAAK,SAAS,GAAG;AAC1B,aAAO,CAAC,KAAK;AAAA,IACf;AACA,UAAM,OAAO,IAAI,MAAM,KAAK,MAAM,MAAM;AACxC,aAAS,IAAI,GAAG,IAAI,KAAK,SAAS,GAAG,EAAE,GAAG;AACxC,WAAK,KAAK,KAAK,MAAM,QAAQ,KAAK,QAAQ,EAAE;AAC5C,eAAS,KAAK,KAAK,KAAK,QAAQ;AAAA,IAClC;AACA,SAAK,KAAK,SAAS,KAAK;AACxB,WAAO;AAAA,EACT;AAAA,EACA,IAAI,OAAO;AACT,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,WAAW;AACT,WAAO,UAAU,EAAE,WAAW,KAAK,QAAQ,KAAK,OAAO,KAAK,KAAK;AAAA,EACnE;AACF;AACA,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,IAAI,uBAAuB;AAC3B,SAAS,iBAAiB,IAAI;AAC5B,cAAY;AACd;AACA,SAAS,aAAa,SAAS;AAC7B,cAAY;AACd;AACA,SAAS,wBAAwB,IAAI;AACnC,yBAAuB;AACzB;AACA,IAAI,SAAS,MAAM;AAAA,EACjB,YAAY,OAAO,OAAO,QAAQ,IAAI;AACpC,SAAK,OAAO;AACZ,SAAK,qBAAqB;AAC1B,SAAK,QAAQ,MAAM,MAAM;AACzB,SAAK,QAAQ,SAAS;AACtB,SAAK,OAAO,cAAc,KAAK;AAC/B,SAAK,UAAU,eAAe,KAAK;AACnC,SAAK,SAAS;AACd,SAAK,KAAK;AACV,SAAK,WAAW,KAAK,OAAO,IAAI,KAAK,KAAK,SAAS,IAAI;AAAA,EACzD;AAAA,EACA,IAAI,OAAO;AACT,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,MAAM,SAAS;AACb,UAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,WAAO,UAAU,OAAO,KAAK,OAAO,KAAK,OAAO,IAAI;AAAA,EACtD;AAAA,EACA,aAAa;AACX,WAAO,UAAU,OAAO,KAAK,OAAO,KAAK,OAAO,KAAK,SAAS,CAAC;AAAA,EACjE;AAAA,EACA,MAAM,QAAQ;AACZ,UAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,WAAO,cAAc,KAAK,OAAO,MAAM,KAAK,UAAU,WAAW;AAAA,EACnE;AAAA,EACA,YAAY;AACV,WAAO,cAAc,KAAK,OAAO,KAAK,SAAS,GAAG,KAAK,UAAU,WAAW;AAAA,EAC9E;AAAA,EACA,MAAM,OAAO;AACX,SAAK,gBAAgB;AACrB,UAAM,OAAO,UAAU,EAAE,KAAK,KAAK,MAAM;AACzC,QAAI,KAAK,UAAU,UAAU;AAC3B,YAAM,QAAQ,MAAM;AACpB,UAAI;AACF,eAAO,MAAM,IAAI,CAAC,MAAM,aAAa,CAAC,CAAC;AAAA,MACzC,SAAS,IAAP;AACA,cAAM,IAAI,MAAM,+FAA+F;AAAA,MACjH;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,UAAU,SAAS;AACjB,SAAK,gBAAgB;AACrB,WAAO,UAAU,EAAE,UAAU,KAAK,QAAQ,OAAO;AAAA,EACnD;AAAA,EACA,WAAW;AACT,SAAK,gBAAgB;AACrB,UAAM,OAAO,UAAU,EAAE,SAAS,KAAK,MAAM;AAC7C,QAAI,KAAK,UAAU,UAAU;AAC3B,UAAI;AACF,eAAO,KAAK,IAAI,CAAC,MAAM,aAAa,CAAC,CAAC;AAAA,MACxC,SAAS,IAAP;AACA,cAAM,IAAI,MAAM,+FAA+F;AAAA,MACjH;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,QAAQ;AACZ,SAAK,gBAAgB;AACrB,UAAM,OAAO,MAAM,UAAU,EAAE,KAAK,KAAK,MAAM;AAC/C,QAAI,KAAK,UAAU,UAAU;AAC3B,aAAO;AAAA,IACT,OAAO;AACL,aAAO,IAAI,WAAW,KAAK,MAAM;AAAA,IACnC;AAAA,EACF;AAAA,EACA,UAAU;AACR,QAAI,KAAK,YAAY;AACnB;AAAA,IACF;AACA,cAAU,EAAE,cAAc,IAAI;AAC9B,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EACA,IAAI,aAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EACA,kBAAkB;AAChB,QAAI,KAAK,YAAY;AACnB,YAAM,IAAI,MAAM,qBAAqB;AAAA,IACvC;AAAA,EACF;AAAA,EACA,MAAM,UAAU,OAAO;AACrB,WAAO,UAAU,MAAM,MAAM,OAAO;AAAA,EACtC;AAAA,EACA,QAAQ;AACN,SAAK,gBAAgB;AACrB,WAAO,UAAU,MAAM,IAAI;AAAA,EAC7B;AAAA,EACA,SAAS,UAAU,OAAO;AACxB,UAAM,OAAO,KAAK,SAAS;AAC3B,WAAO,eAAe,MAAM,KAAK,OAAO,KAAK,OAAO,OAAO;AAAA,EAC7D;AAAA,EACA,KAAK,OAAO;AACV,SAAK,gBAAgB;AACrB,WAAO,UAAU,KAAK,MAAM,KAAK;AAAA,EACnC;AAAA,EACA,SAAS,YAAY,MAAM,MAAM,OAAO;AACtC,SAAK,gBAAgB;AACrB,WAAO,UAAU,EAAE,aAAa,MAAM,WAAW,MAAM,KAAK;AAAA,EAC9D;AACF;AACA,OAAO,eAAe,QAAQ,OAAO,aAAa;AAAA,EAChD,OAAO,CAAC,aAAa;AACnB,WAAO,CAAC,CAAC,YAAY,SAAS,QAAQ,QAAQ,SAAS,YAAY,QAAQ,SAAS,mBAAmB;AAAA,EACzG;AACF,CAAC;AACD,SAAS,uBAAuB;AAC9B,SAAO,UAAU,UAAU,MAAM;AAC/B,WAAO;AAAA,EACT,CAAC;AACH;AACA,qBAAqB;AACrB,IAAI,WAAW,cAAc,OAAO;AAAA,EAClC,YAAY,cAAc,WAAW,MAAM,UAAU;AACnD,UAAM,aAAa,OAAO,aAAa,OAAO,aAAa,QAAQ,QAAQ;AAC3E,SAAK,YAAY;AACjB,SAAK,OAAO;AAAA,EACd;AAAA,EACA,OAAO,UAAU;AACf,QAAI,SAAS,UAAU,KAAK,OAAO;AACjC,YAAM,IAAI,MAAM,2BAA2B,SAAS,8BAA8B,KAAK,mBAAmB;AAAA,IAC5G;AACA,QAAI,CAAC,YAAY,SAAS,OAAO,KAAK,KAAK,GAAG;AAC5C,YAAM,IAAI,MAAM,2BAA2B,SAAS,8BAA8B,KAAK,mBAAmB;AAAA,IAC5G;AACA,cAAU,EAAE,cAAc,IAAI;AAC9B,SAAK,SAAS,SAAS;AACvB,cAAU,EAAE,OAAO,MAAM,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU;AACR,cAAU,EAAE,gBAAgB,IAAI;AAChC,SAAK,qBAAqB;AAAA,EAC5B;AACF;AACA,OAAO,eAAe,UAAU,OAAO,aAAa;AAAA,EAClD,OAAO,CAAC,aAAa;AACnB,WAAO,oBAAoB,UAAU,SAAS,UAAU,QAAQ,SAAS,kBAAkB;AAAA,EAC7F;AACF,CAAC;AAGD,IAAI,sBAAsB,CAAC;AAC3BA,UAAS,qBAAqB;AAAA,EAC5B,kBAAkB,MAAM;AAAA,EACxB,uBAAuB,MAAM;AAAA,EAC7B,gBAAgB,MAAM;AAAA,EACtB,gBAAgB,MAAM;AACxB,CAAC;AAGD,IAAI;AAAA,CACH,SAAS,OAAO;AACf,QAAM,QAAQ;AACd,QAAM,QAAQ;AACd,QAAM,QAAQ;AACd,QAAM,QAAQ;AACd,QAAM,QAAQ;AACd,QAAM,QAAQ;AACd,QAAM,QAAQ;AAChB,GAAG,SAAS,OAAO,CAAC,EAAE;AACtB,IAAI;AAAA,CACH,SAAS,oBAAoB;AAC5B,qBAAmB,aAAa;AAChC,qBAAmB,WAAW;AAC9B,qBAAmB,UAAU;AAC7B,qBAAmB,eAAe;AACpC,GAAG,sBAAsB,oBAAoB,CAAC,EAAE;AAChD,IAAI;AAAA,CACH,SAAS,mBAAmB;AAC3B,oBAAkB,aAAa;AAC/B,oBAAkB,WAAW;AAC7B,oBAAkB,UAAU;AAC5B,oBAAkB,eAAe;AACnC,GAAG,qBAAqB,mBAAmB,CAAC,EAAE;AAC9C,IAAI;AAAA,CACH,SAAS,sBAAsB;AAC9B,uBAAqB,aAAa;AAClC,uBAAqB,WAAW;AAChC,uBAAqB,UAAU;AAC/B,uBAAqB,eAAe;AACtC,GAAG,wBAAwB,sBAAsB,CAAC,EAAE;AACpD,IAAI;AAAA,CACH,SAAS,wBAAwB;AAChC,yBAAuB,aAAa;AACpC,yBAAuB,WAAW;AAClC,yBAAuB,UAAU;AACjC,yBAAuB,eAAe;AACxC,GAAG,0BAA0B,wBAAwB,CAAC,EAAE;AACxD,IAAI,gBAAgB;AAAA,EAClB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,aAAa;AACf;AACA,SAAS,WAAW,OAAO,OAAO;AAChC,MAAI,UAAU,YAAY,UAAU,UAAU;AAC5C,QAAI,UAAU,YAAY,UAAU,UAAU;AAC5C,aAAO;AAAA,IACT;AACA,UAAM,IAAI,MAAM,kBAAkB,cAAc,OAAO;AAAA,EACzD;AACA,SAAO,cAAc,OAAO;AAC9B;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,WAAW,MAAM,OAAO;AACjC;AAGA,SAAS,eAAe,GAAG,GAAG;AAC5B,MAAI,EAAE,UAAU,EAAE,OAAO;AACvB,WAAO,CAAC,GAAG,CAAC;AAAA,EACd;AACA,QAAM,QAAQ,WAAW,EAAE,OAAO,EAAE,KAAK;AACzC,SAAO,CAAC,EAAE,KAAK,KAAK,GAAG,EAAE,KAAK,KAAK,CAAC;AACtC;AACA,SAAS,iBAAiB,GAAG,GAAG;AAC9B,SAAO,EAAE,UAAU,EAAE,OAAO,MAAM,2BAA2B,EAAE,qBAAqB,EAAE,yBAAyB;AACjH;AACA,SAAS,eAAe,SAAS,YAAY;AAC3C,SAAO,WAAW,KAAK,CAAC,MAAM,EAAE,OAAO,QAAQ,EAAE;AACnD;AACA,SAAS,sBAAsB,QAAQ;AACrC,QAAM,OAAO,CAAC;AACd,QAAM,OAAuB,oBAAI,IAAI;AACrC,sBAAoB,QAAQ,MAAM,IAAI;AACtC,SAAO;AACT;AACA,SAAS,oBAAoB,WAAW,MAAM,MAAM;AAClD,MAAI,aAAa,MAAM;AACrB;AAAA,EACF;AACA,MAAI,qBAAqB,QAAQ;AAC/B,SAAK,KAAK,SAAS;AACnB;AAAA,EACF;AACA,MAAI,CAAC,WAAW,SAAS,GAAG;AAC1B;AAAA,EACF;AACA,QAAM,WAAW;AACjB,aAAW,KAAK,UAAU;AACxB,UAAM,MAAM,SAAS;AACrB,QAAI,CAAC,KAAK,IAAI,GAAG,GAAG;AAClB,WAAK,IAAI,GAAG;AACZ,0BAAoB,KAAK,MAAM,IAAI;AAAA,IACrC;AAAA,EACF;AACF;AACA,SAAS,WAAW,KAAK;AACvB,SAAO,MAAM,QAAQ,GAAG,KAAK,OAAO,QAAQ;AAC9C;AAGA,SAAS,6BAA6B,kBAAkB;AACtD,SAAO,iBAAiB,cAAc;AACxC;AACA,IAAI,cAAc,MAAM;AAAA,EACtB,cAAc;AACZ,SAAK,sBAAsB,CAAC;AAC5B,SAAK,iBAAiB;AACtB,SAAK,WAAW;AAChB,SAAK,aAAa;AAClB,SAAK,mBAAmB;AACxB,SAAK,iBAAiB;AACtB,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,aAAa,CAAC;AACnB,SAAK,oBAAoB,CAAC;AAC1B,SAAK,cAAc;AACnB,SAAK,aAA6B,oBAAI,QAAQ;AAC9C,SAAK,YAAY;AACjB,SAAK,gBAAgB;AAAA,MACnB,UAAU;AAAA,MACV,YAAY;AAAA,MACZ,WAAW;AAAA,MACX,SAAS,CAAC;AAAA,MACV,QAAQ;AAAA,MACR,IAAI,cAAc;AAChB,eAAO,MAAM,KAAK,IAAI,IAAI,KAAK,QAAQ,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;AAAA,MAC5D;AAAA,IACF;AAAA,EACF;AAAA,EACA,UAAU;AACR,eAAW,gBAAgB,KAAK,qBAAqB;AACnD,WAAK,oBAAoB,cAAc,QAAQ;AAAA,IACjD;AAAA,EACF;AACF;AACA,IAAI,SAAS,MAAM;AAAA,EACjB,YAAY,MAAM;AAChB,SAAK,MAAM;AACX,SAAK,WAAW,CAAC;AACjB,SAAK,kBAAkB,CAAC;AACxB,SAAK,uBAAuB;AAC5B,SAAK,QAAQ,IAAI,YAAY;AAAA,EAC/B;AAAA,EACA,MAAM,QAAQ;AACZ,QAAI,KAAK,sBAAsB,MAAM;AACnC,aAAO,KAAK,mBAAmB,KAAK,MAAM;AAAA,MAC1C,CAAC;AAAA,IACH;AACA,QAAI,KAAK,mBAAmB,MAAM;AAChC;AAAA,IACF;AACA,UAAM,iBAAiB,KAAK,kBAAkB;AAC9C,aAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,YAAM,cAAc,eAAe;AACnC,YAAM,UAAU,MAAM,KAAK,kBAAkB,WAAW,EAAE;AAC1D,UAAI,SAAS;AACX,cAAM,KAAK,WAAW,WAAW;AACjC;AAAA,MACF;AAAA,IACF;AACA,UAAM,IAAI,MAAM,wEAAwE;AAAA,EAC1F;AAAA,EACA,IAAI,UAAU;AACZ,QAAI,KAAK,sBAAsB,MAAM;AACnC,YAAM,IAAI,MAAM,YAAY,KAAK,gIAAgI;AAAA,IACnK;AACA,QAAI,KAAK,mBAAmB,MAAM;AAChC,YAAM,EAAE,MAAM,UAAU,IAAI,KAAK,gCAAgC;AACjE,UAAI,WAAW;AACb,cAAM,IAAI,MAAM,iCAAiC,yHAAyH;AAAA,MAC5K;AACA,WAAK,WAAW,IAAI;AAAA,IACtB;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EACA,eAAe;AACb,WAAO,OAAO,KAAK,KAAK,eAAe;AAAA,EACzC;AAAA,EACA,YAAY,aAAa;AACvB,QAAI,EAAE,eAAe,KAAK,WAAW;AACnC,UAAI,eAAe,KAAK,iBAAiB;AACvC,cAAM,EAAE,UAAU,IAAI,KAAK,kBAAkB,WAAW;AACxD,YAAI,WAAW;AACb,iBAAO;AAAA,QACT;AAAA,MACF,OAAO;AACL,eAAO;AAAA,MACT;AAAA,IACF;AACA,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,mBAAmB,aAAa;AAC9B,QAAI,EAAE,eAAe,KAAK,kBAAkB;AAC1C,aAAO;AAAA,IACT;AACA,WAAO,KAAK,gBAAgB,aAAa;AAAA,EAC3C;AAAA,EACA,gBAAgB,aAAa,SAAS,WAAW,GAAG;AAClD,QAAI,eAAe,KAAK,iBAAiB;AACvC,WAAK,GAAG,+EAA+E;AACvF,aAAO;AAAA,IACT;AACA,SAAK,gBAAgB,eAAe,EAAE,SAAS,SAAS;AACxD,WAAO;AAAA,EACT;AAAA,EACA,MAAM,WAAW,aAAa;AAC5B,QAAI,KAAK,gBAAgB,gBAAgB,MAAM;AAC7C,YAAM,IAAI,MAAM,iBAAiB,oCAAoC;AAAA,IACvE;AACA,SAAK,cAAc;AACnB,QAAI,KAAK,SAAS,gBAAgB,MAAM;AACtC,WAAK,kBAAkB;AACvB,YAAM,EAAE,SAAS,UAAU,IAAI,KAAK,kBAAkB,WAAW;AACjE,YAAM,SAAS,YAAY,MAAM,UAAU;AAC3C,UAAI,CAAC,QAAQ;AACX,eAAO;AAAA,MACT;AAAA,IACF;AACA,SAAK,kBAAkB,KAAK,SAAS;AACrC,SAAK,uBAAuB;AAC5B,SAAK,WAAW,IAAI,SAAS,KAAK,eAAe;AACjD,WAAO;AAAA,EACT;AAAA,EACA,yBAAyB;AACvB,UAAM,UAAU,qBAAqB,KAAK,WAAW;AACrD,YAAQ,QAAQ,CAAC,WAAW;AAC1B,UAAI,OAAO,aAAa,MAAM;AAC5B,eAAO,UAAU,KAAK,eAAe;AAAA,MACvC;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,yBAAyB,aAAa;AACpC,UAAM,UAAU,qBAAqB,WAAW;AAChD,YAAQ,QAAQ,CAAC,WAAW;AAC1B,UAAI,OAAO,eAAe,MAAM;AAC9B,eAAO,YAAY,KAAK,SAAS,YAAY;AAAA,MAC/C;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,kBAAkB,aAAa;AAC7B,UAAM,uBAAuB,KAAK,gBAAgB;AAClD,QAAI,wBAAwB,MAAM;AAChC,YAAM,IAAI,MAAM,6BAA6B,qCAAqC;AAAA,IACpF;AACA,QAAI;AACF,YAAM,WAAW,qBAAqB,QAAQ;AAC9C,UAAI,YAAY,EAAE,oBAAoB,kBAAkB,OAAO,SAAS,SAAS,YAAY;AAC3F,cAAM,YAAY,EAAE,KAAK;AACzB,cAAM,UAAU,SAAS,KAAK,CAAC,oBAAoB;AACjD,cAAI,YAAY,KAAK,sBAAsB;AACzC,mBAAO;AAAA,UACT;AACA,eAAK,SAAS,eAAe;AAC7B,eAAK,qBAAqB;AAC1B,iBAAO;AAAA,QACT,CAAC,EAAE,MAAM,CAAC,QAAQ;AAChB,cAAI,YAAY,KAAK,sBAAsB;AACzC,mBAAO;AAAA,UACT;AACA,eAAK,qBAAqB;AAC1B,eAAK,6BAA6B,oBAAoB;AACtD,eAAK,IAAI,SAAS,IAAI,OAAO;AAC7B,iBAAO;AAAA,QACT,CAAC;AACD,aAAK,qBAAqB;AAC1B,eAAO,EAAE,SAAS,WAAW,KAAK;AAAA,MACpC,OAAO;AACL,aAAK,SAAS,eAAe;AAC7B,eAAO,EAAE,SAAS,MAAM,WAAW,MAAM;AAAA,MAC3C;AAAA,IACF,SAAS,KAAP;AACA,WAAK,6BAA6B,oBAAoB;AACtD,WAAK,IAAI,SAAS,IAAI,OAAO;AAC7B,aAAO,EAAE,SAAS,OAAO,WAAW,MAAM;AAAA,IAC5C;AAAA,EACF;AAAA,EACA,cAAc,aAAa;AACzB,QAAI,EAAE,eAAe,KAAK,kBAAkB;AAC1C,YAAM,IAAI,MAAM,GAAG,2CAA2C;AAAA,IAChE;AACA,QAAI,KAAK,gBAAgB,eAAe,KAAK,sBAAsB,MAAM;AACvE,WAAK;AAAA,IACP;AACA,QAAI,eAAe,KAAK,UAAU;AAChC,WAAK,yBAAyB,WAAW;AACzC,WAAK,SAAS,aAAa,QAAQ;AACnC,aAAO,KAAK,SAAS;AAAA,IACvB;AACA,WAAO,KAAK,gBAAgB;AAC5B,QAAI,KAAK,gBAAgB,aAAa;AACpC,WAAK,qBAAqB;AAC1B,WAAK,cAAc;AACnB,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EACA,oBAAoB;AAClB,QAAI,OAAO,KAAK,KAAK,eAAe,EAAE,WAAW,GAAG;AAClD,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AACA,WAAO,OAAO,KAAK,KAAK,eAAe,EAAE,KAAK,CAAC,GAAG,MAAM;AACtD,aAAO,KAAK,gBAAgB,GAAG,WAAW,KAAK,gBAAgB,GAAG;AAAA,IACpE,CAAC;AAAA,EACH;AAAA,EACA,kCAAkC;AAChC,UAAM,iBAAiB,KAAK,kBAAkB;AAC9C,aAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,YAAM,cAAc,eAAe;AACnC,YAAM,EAAE,SAAS,UAAU,IAAI,KAAK,kBAAkB,WAAW;AACjE,UAAI,aAAa,SAAS;AACxB,eAAO,EAAE,MAAM,aAAa,UAAU;AAAA,MACxC;AAAA,IACF;AACA,UAAM,IAAI,MAAM,wEAAwE;AAAA,EAC1F;AAAA,EACA,SAAS,UAAU,QAAQ;AACzB,UAAM,OAAO,KAAK,MAAM,WAAW,IAAI,MAAM;AAC7C,UAAM,aAAa,KAAK;AACxB,UAAM,SAAS,KAAK,SAAS,MAAM;AACnC,UAAM,WAAW,WAAW,SAAS,MAAM;AAC3C,eAAW,YAAY,QAAQ,IAAI;AACnC,SAAK,UAAU;AACf,aAAS,KAAK,QAAQ,QAAQ,KAAK,OAAO,KAAK,OAAO,QAAQ;AAC9D,QAAI,KAAK,uBAAuB,GAAG;AACjC,WAAK,MAAM,kBAAkB,KAAK,MAAM,kBAAkB,SAAS;AAAA,IACrE;AAAA,EACF;AAAA,EACA,KAAK,UAAU,IAAI;AACjB,QAAI,OAAO;AACX,QAAI,MAAM,MAAM;AACd,UAAI,OAAO,aAAa,YAAY;AAClC,cAAM,IAAI,MAAM,qCAAqC;AAAA,MACvD;AACA,WAAK;AAAA,IACP,OAAO;AACL,UAAI,OAAO,aAAa,YAAY,EAAE,oBAAoB,SAAS;AACjE,cAAM,IAAI,MAAM,gFAAgF;AAAA,MAClG;AACA,UAAI,OAAO,OAAO,YAAY;AAC5B,cAAM,IAAI,MAAM,gFAAgF;AAAA,MAClG;AACA,aAAO;AAAA,IACT;AACA,QAAI;AACJ,WAAO,KAAK,UAAU,MAAM,KAAK,WAAW,IAAI,GAAG,MAAM,KAAK,SAAS,MAAM,GAAG,MAAM;AACpF,eAAS,GAAG;AACZ,UAAI,kBAAkB,SAAS;AAC7B,gBAAQ,MAAM,yCAAyC;AAAA,MACzD;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,UAAU,OAAO,KAAK,GAAG;AACvB,UAAM;AACN,QAAI;AACF,YAAM,MAAM,EAAE;AACd,UAAI;AACJ,aAAO;AAAA,IACT,SAAS,IAAP;AACA,UAAI;AACJ,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EACA,eAAe;AACb,WAAO,OAAO;AAAA,EAChB;AAAA,EACA,iBAAiB;AACf,WAAO,OAAO;AAAA,EAChB;AAAA,EACA,MAAM,GAAG;AACP,UAAM,IAAI,OAAO,UAAU,UAAU,EAAE,EAAE,CAAC;AAC1C,UAAM,SAAS,EAAE,EAAE;AACnB,UAAM,QAAQ,CAAC,QAAQ;AAAA,MACrB,GAAG,MAAM;AACP,cAAM,QAAQ;AACd,cAAM,aAAa,EAAE,GAAG,GAAG;AAC3B,cAAM,QAAQ,EAAE,MAAM;AACtB,eAAO,OAAO;AAAA,UACZ;AAAA,UACA;AAAA,UACA;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,QAAQ,CAAC;AACf,SAAK,YAAY,KAAK,MAAM,YAAY,MAAM,QAAQ,CAAC,CAAC,GAAG,OAAO,OAAO,CAAC,CAAC;AAC3E,WAAO;AAAA,EACT;AAAA,EACA,UAAU,YAAY,QAAQ,OAAO;AACnC,QAAI,KAAK,eAAe,MAAM;AAC5B,WAAK;AAAA,IACP;AACA,UAAM,YAAY,UAAU,YAAY,KAAK,WAAW,KAAK;AAC7D,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,WAAW,2CAA2C,KAAK,cAAc;AAAA,IAC3F;AACA,WAAO,KAAK,cAAc,EAAE,YAAY,QAAQ,MAAM,CAAC;AAAA,EACzD;AAAA,EACA,yBAAyB;AACvB,WAAO,KAAK,IAAI,QAAQ,SAAS;AAAA,EACnC;AAAA,EACA,sBAAsB,YAAY,kBAAkB,UAAU;AAC5D,UAAM,kBAAkB,KAAK,QAAQ,WAAW;AAChD,QAAI,mBAAmB;AACvB,aAAS,QAAQ,CAAC,SAAS;AACzB,0BAAoB,KAAK,UAAU,cAAc,IAAI;AAAA,IACvD,CAAC;AACD,UAAM,WAAW,KAAK,MAAM,kBAAkB,KAAK,MAAM,kBAAkB,SAAS;AACpF,UAAM,gBAAgB,kBAAkB,mBAAmB,mBAAmB;AAC9E,QAAI,gBAAgB,GAAG;AACrB,YAAM,IAAI,MAAM,YAAY,KAAK,6CAA6C,0CAA0C,aAAa;AAAA,IACvI;AAAA,EACF;AAAA,EACA,cAAc,cAAc;AAC1B,QAAI;AACJ,QAAI,QAAQ,CAAC;AACb,UAAM,WAAW,KAAK,SAAS;AAC/B,UAAM,oBAAoB,KAAK,MAAM;AACrC,UAAM,qBAAqB,KAAK,MAAM;AACtC,QAAI,KAAK,uBAAuB,GAAG;AACjC,WAAK,MAAM,kBAAkB,KAAK,CAAC;AAAA,IACrC;AACA,QAAI;AACJ,QAAI,KAAK,eAAe,MAAM;AAC5B,WAAK;AAAA,IACP;AACA,QAAI;AACJ,UAAM,oBAAoB,6BAA6B,YAAY,IAAI,aAAa,aAAa,KAAK,MAAM,eAAe,OAAO,KAAK,MAAM,YAAY,OAAO;AAChK,QAAI,6BAA6B,YAAY,GAAG;AAC9C,YAAM,EAAE,YAAY,QAAQ,SAAS,OAAO,OAAO,IAAI;AACvD,UAAI,KAAK,eAAe,MAAM;AAC5B,aAAK;AAAA,MACP;AACA,YAAM,SAAS,UAAU,YAAY,KAAK,WAAW;AACrD,aAAO,UAAU,MAAM,MAAM,kCAAkC,4BAA4B,KAAK,cAAc;AAC9G,oBAAc,MAAM;AAClB,cAAM,mBAAmB,KAAK,QAAQ,WAAW;AACjD,cAAM,OAAO,WAAW,EAAE,QAAQ,SAAS,OAAO,QAAQ,SAAS,KAAK,QAAQ,CAAC;AACjF,cAAM,WAAW,MAAM,QAAQ,GAAG,IAAI,MAAM,CAAC,GAAG;AAChD,YAAI,KAAK,uBAAuB,GAAG;AACjC,eAAK,sBAAsB,YAAY,kBAAkB,QAAQ;AAAA,QACnE;AACA,cAAM,aAAa,SAAS,IAAI,CAAC,YAAY;AAC3C,cAAI,QAAQ,QAAQ,MAAM;AACxB,mBAAO;AAAA,UACT;AACA,iBAAO,KAAK,yBAAyB,OAAO;AAAA,QAC9C,CAAC;AACD,YAAI,UAAU;AACZ,gBAAM,gBAAgB,KAAK,sBAAsB,YAAY,SAAS,UAAU;AAChF,kBAAQ,KAAK,2BAA2B,aAAa;AAAA,QACvD;AACA,eAAO;AAAA,MACT;AAAA,IACF,OAAO;AACL,YAAM,EAAE,YAAY,IAAI;AACxB,YAAM,WAAW,CAAC,YAAY;AAC5B,YAAI,CAAC,UAAU;AACb;AAAA,QACF;AACA,gBAAQ,QAAQ,IAAI,CAAC,YAAY,KAAK,KAAK,KAAK,MAAM,OAAO,CAAC,CAAC;AAAA,MACjE;AACA,oBAAc,MAAM;AAClB,cAAM,mBAAmB,KAAK,QAAQ,WAAW;AACjD,cAAM,KAAK,KAAK,MAAM,YAAY,KAAK,SAAS,QAAQ,CAAC;AACzD,cAAM,OAAO,MAAM,QAAQ,GAAG,IAAI,MAAM,CAAC,GAAG;AAC5C,YAAI,KAAK,uBAAuB,GAAG;AACjC,eAAK,sBAAsB,mBAAmB,kBAAkB,IAAI;AAAA,QACtE;AACA,eAAO;AAAA,MACT;AAAA,IACF;AACA,UAAM,EAAE,QAAQ,MAAM,IAAI;AAC1B,UAAM,gBAAgB,6BAA6B,YAAY,IAAI,OAAO,aAAa;AACvF,QAAI;AACJ,SAAK;AAAA,MACH,MAAM,KAAK,MAAM;AAAA,MACjB,MAAM,KAAK,MAAM;AAAA,MACjB,MAAM;AACJ,YAAI,CAAC,KAAK,IAAI,QAAQ,OAAO,KAAK,CAAC,KAAK,MAAM,WAAW;AACvD,oBAAU,YAAY;AAAA,QACxB,OAAO;AACL,0BAAgB,KAAK,SAAS,cAAc,mBAAmB,QAAQ,MAAM,YAAY,CAAC;AAC1F,cAAI,KAAK,IAAI,QAAQ,OAAO,GAAG;AAC7B,iBAAK,SAAS,iBAAiB,aAAa;AAAA,UAC9C;AACA,oBAAU,cAAc;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AACA,QAAI,UAAU;AACZ,WAAK,YAAY,mBAAmB,QAAQ,SAAS,eAAe,OAAO,KAAK;AAAA,IAClF;AACA,QAAI,KAAK,MAAM,WAAW;AACxB,WAAK,MAAM,cAAc,QAAQ,KAAK;AAAA,QACpC,MAAM;AAAA,QACN,YAAY,KAAK,MAAM,WAAW;AAAA,QAClC,oBAAoB,KAAK,MAAM;AAAA,QAC/B,cAAc,KAAK,MAAM,aAAa;AAAA,QACtC,sBAAsB,KAAK,MAAM;AAAA,QACjC,aAAa,OAAO,KAAK,MAAM,EAAE,IAAI,CAAC,QAAQ,OAAO,QAAQ,OAAO,OAAO,KAAK,QAAQ,IAAI;AAAA,QAC5F,cAAc,QAAQ,IAAI,CAAC,SAAS,KAAK,KAAK;AAAA,QAC9C,cAAc,cAAc;AAAA,QAC5B,WAAW,cAAc;AAAA,MAC3B,CAAC;AAAA,IACH;AACA,WAAO,MAAM,QAAQ,GAAG,IAAI,UAAU,QAAQ;AAAA,EAChD;AAAA,EACA,2BAA2B,SAAS;AAClC,UAAM,QAAQ,QAAQ,IAAI,CAAC,YAAY,KAAK,KAAK,KAAK,MAAM,OAAO,CAAC,CAAC;AACrE,WAAO;AAAA,EACT;AAAA,EACA,sBAAsB,YAAY,QAAQ,SAAS;AACjD,UAAM,aAAa,YAAY,UAAU;AACzC,QAAI,cAAc,MAAM;AACtB,YAAM,eAAe,WAAW,gBAAgB,CAAC;AACjD,YAAM,gBAAgB,WAAW,iBAAiB,CAAC;AACnD,UAAI;AACJ,UAAI,WAAW,eAAe;AAC5B,eAAO,MAAM,QAAQ,MAAM,GAAG,MAAM,wDAAwD;AAC5F,6BAAqB,OAAO,KAAK,MAAM,EAAE,IAAI,CAAC,QAAQ,OAAO,IAAI;AAAA,MACnE,OAAO;AACL,6BAAqB,aAAa,IAAI,CAAC,cAAc,OAAO,UAAU;AAAA,MACxE;AACA,YAAM,sBAAsB,QAAQ,OAAO,CAAC,GAAG,MAAM,cAAc,EAAE;AACrE,aAAO,mBAAmB,OAAO,mBAAmB;AAAA,IACtD;AACA,WAAO,CAAC;AAAA,EACV;AAAA,EACA,WAAW,QAAQ,OAAO,OAAO,UAAU;AACzC,QAAI,UAAU,MAAM;AAClB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,YAAQ,SAAS;AACjB,eAAW,YAAY,KAAK;AAC5B,QAAI,cAAc;AAClB,QAAI,UAAU,YAAY,SAAS,OAAO,EAAE,GAAG;AAC7C,oBAAc,OAAO,IAAI,CAAC,MAAM,aAAa,CAAC,CAAC;AAAA,IACjD;AACA,UAAM,SAAS,SAAS,MAAM,aAAa,OAAO,KAAK;AACvD,UAAM,IAAI,IAAI,OAAO,OAAO,OAAO,QAAQ,KAAK,aAAa,CAAC;AAC9D,SAAK,YAAY,GAAG,QAAQ;AAC5B,QAAI,UAAU,UAAU;AACtB,YAAM,OAAO,KAAK,MAAM,WAAW,IAAI,MAAM;AAC7C,YAAM,WAAW,qBAAqB,WAAW;AACjD,WAAK,MAAM,YAAY,WAAW,KAAK;AACvC,WAAK,QAAQ;AAAA,IACf;AACA,WAAO;AAAA,EACT;AAAA,EACA,qBAAqB,QAAQ,OAAO,OAAO,UAAU;AACnD,YAAQ,SAAS;AACjB,UAAM,aAAa,EAAE,QAAQ,OAAO,MAAM;AAC1C,WAAO,KAAK,yBAAyB,YAAY,QAAQ;AAAA,EAC3D;AAAA,EACA,yBAAyB,YAAY,UAAU;AAC7C,UAAM,EAAE,QAAQ,OAAO,MAAM,IAAI;AACjC,UAAM,IAAI,IAAI,OAAO,OAAO,OAAO,QAAQ,KAAK,aAAa,CAAC;AAC9D,SAAK,YAAY,GAAG,QAAQ;AAC5B,WAAO;AAAA,EACT;AAAA,EACA,aAAa,cAAc,YAAY,MAAM,MAAM,OAAO;AACxD,WAAO,QAAQ,KAAK,eAAe,EAAE,SAAS;AAC9C,QAAI,SAAS,QAAQ,UAAU,aAAa,OAAO;AACjD,qBAAe,aAAa,KAAK,KAAK;AAAA,IACxC;AACA,UAAM,IAAI,IAAI,SAAS,cAAc,WAAW,MAAM,KAAK,aAAa,CAAC;AACzE,QAAI,KAAK,MAAM,oBAAoB,EAAE,SAAS,MAAM;AAClD,YAAM,IAAI,MAAM,sBAAsB,EAAE,6BAA6B;AAAA,IACvE;AACA,SAAK,MAAM,oBAAoB,EAAE,QAAQ;AACzC,SAAK,OAAO,GAAG,KAAK,OAAO;AAC3B,WAAO;AAAA,EACT;AAAA,EACA,YAAY,GAAG,UAAU;AACvB,SAAK,MAAM;AACX,QAAI,EAAE,UAAU,UAAU;AACxB,WAAK,MAAM;AAAA,IACb;AACA,QAAI,QAAQ;AACZ,QAAI,EAAE,UAAU,eAAe,EAAE,UAAU,UAAU;AACnD,cAAQ,EAAE,OAAO,gBAAgB,EAAE,KAAK;AAAA,IAC1C;AACA,SAAK,MAAM,YAAY;AACvB,QAAI,CAAC,KAAK,MAAM,WAAW,IAAI,EAAE,MAAM,GAAG;AACxC,WAAK,MAAM;AACX,WAAK,MAAM,WAAW,IAAI,EAAE,QAAQ;AAAA,QAClC,SAAS,YAAY,KAAK;AAAA,QAC1B,OAAO,EAAE;AAAA,QACT,OAAO,EAAE;AAAA,QACT;AAAA,MACF,CAAC;AAAA,IACH;AACA,QAAI,EAAE,aAAa,WAAW;AAC5B,WAAK,MAAM,CAAC;AAAA,IACd;AAAA,EACF;AAAA,EACA,OAAO,GAAG,UAAU;AAClB,SAAK,YAAY,GAAG,QAAQ;AAC5B,SAAK,QAAQ,OAAO,EAAE,MAAM;AAAA,EAC9B;AAAA,EACA,aAAa,QAAQ,UAAU;AAC7B,QAAI,KAAK,MAAM,WAAW,IAAI,MAAM,KAAK,KAAK,MAAM,WAAW,IAAI,MAAM,EAAE,YAAY,UAAU;AAC/F,WAAK,MAAM,WAAW,OAAO,MAAM;AACnC,WAAK,MAAM;AAAA,IACb;AAAA,EACF;AAAA,EACA,cAAc,GAAG;AACf,QAAI,CAAC,KAAK,MAAM,WAAW,IAAI,EAAE,MAAM,GAAG;AACxC;AAAA,IACF;AACA,UAAM,OAAO,KAAK,MAAM,WAAW,IAAI,EAAE,MAAM;AAC/C,SAAK,MAAM;AACX,QAAI,EAAE,UAAU,UAAU;AACxB,WAAK,MAAM;AACX,WAAK,MAAM,YAAY,KAAK;AAAA,IAC9B;AACA,QAAI,EAAE,UAAU,eAAe,EAAE,UAAU,UAAU;AACnD,YAAM,QAAQ,EAAE,OAAO,gBAAgB,EAAE,KAAK;AAC9C,WAAK,MAAM,YAAY;AAAA,IACzB;AACA,QAAI,KAAK,QAAQ,YAAY,EAAE,MAAM,GAAG;AACtC,WAAK,aAAa,EAAE,QAAQ,KAAK,OAAO;AAAA,IAC1C;AAAA,EACF;AAAA,EACA,mBAAmB;AACjB,eAAW,WAAW,KAAK,MAAM,qBAAqB;AACpD,YAAM,IAAI,KAAK,MAAM,oBAAoB;AACzC,WAAK,gBAAgB,CAAC;AAAA,IACxB;AAAA,EACF;AAAA,EACA,gBAAgB,GAAG;AACjB,SAAK,cAAc,CAAC;AACpB,QAAI,KAAK,MAAM,oBAAoB,EAAE,SAAS,MAAM;AAClD,aAAO,KAAK,MAAM,oBAAoB,EAAE;AAAA,IAC1C;AAAA,EACF;AAAA,EACA,SAAS;AACP,UAAM,OAAO,KAAK,QAAQ,OAAO;AACjC,SAAK,aAAa,KAAK,MAAM;AAC7B,SAAK,iBAAiB,KAAK,MAAM;AACjC,SAAK,WAAW,KAAK,MAAM;AAC3B,QAAI,KAAK,MAAM,mBAAmB,GAAG;AACnC,WAAK,aAAa;AAClB,UAAI,KAAK,WAAW,MAAM;AACxB,aAAK,UAAU,CAAC;AAAA,MAClB;AACA,WAAK,QAAQ,KAAK,uEAAuE;AAAA,IAC3F;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,QAAQ,OAAO;AACnB,SAAK,MAAM,YAAY;AACvB,UAAM,aAAa,KAAK,MAAM;AAC9B,UAAM,kBAAkB,KAAK,MAAM;AACnC,SAAK,MAAM,cAAc,UAAU,CAAC;AACpC,SAAK,MAAM,cAAc,SAAS,MAAM,MAAM;AAC9C,SAAK,MAAM,YAAY;AACvB,SAAK,MAAM,cAAc,YAAY,KAAK,IAAI,GAAG,KAAK,MAAM,cAAc,QAAQ,IAAI,CAAC,MAAM,EAAE,kBAAkB,CAAC;AAClH,SAAK,MAAM,cAAc,WAAW,KAAK,MAAM,WAAW;AAC1D,SAAK,MAAM,cAAc,aAAa,KAAK,MAAM,aAAa;AAC9D,eAAW,UAAU,KAAK,MAAM,cAAc,SAAS;AACrD,aAAO,eAAe,MAAM,OAAO;AACnC,aAAO,YAAY,MAAM,OAAO;AAAA,IAClC;AACA,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,WAAW;AACT,WAAO,KAAK,MAAM,gBAAgB,KAAK,KAAK,MAAM,gBAAgB;AAAA,EACpE;AAAA,EACA,YAAY,YAAY,QAAQ,SAAS,eAAe,OAAO,OAAO;AACpE,UAAM,WAAW,EAAE,IAAI,KAAK,MAAM,kBAAkB,YAAY,QAAQ,SAAS,MAAM;AACvF,UAAM,aAAa,YAAY,UAAU;AACzC,QAAI,cAAc,MAAM;AACtB,sBAAgB,WAAW;AAAA,IAC7B;AACA,QAAI,iBAAiB,MAAM;AACzB,eAAS,WAAW,CAAC,QAAQ;AAC3B,cAAM,IAAI,IAAI,CAAC,IAAI,MAAM;AACvB,cAAI,MAAM,MAAM;AACd,kBAAM,SAAS,QAAQ;AACvB,kBAAM,OAAO,oBAAoB,OAAO,MAAM,OAAO,KAAK;AAC1D,mBAAO,KAAK,WAAW,MAAM,OAAO,OAAO,OAAO,KAAK;AAAA,UACzD;AACA,iBAAO;AAAA,QACT,CAAC;AACD,eAAO,cAAc,IAAI,SAAS,IAAI,MAAM,IAAI,IAAI,OAAO,KAAK;AAAA,MAClE;AAAA,IACF;AACA,SAAK,MAAM,WAAW,KAAK,QAAQ;AAAA,EACrC;AAAA,EACA,KAAK,QAAQ;AACX,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,QAAI,KAAK,MAAM,kBAAkB,GAAG;AAClC,WAAK,MAAM,aAAa,CAAC;AAAA,IAC3B;AACA,SAAK,MAAM;AAAA,EACb;AAAA,EACA,UAAU;AACR,SAAK,MAAM;AAAA,EACb;AAAA,EACA,WAAW,MAAM;AACf,UAAM,YAAY;AAAA,MAChB,OAAO,CAAC;AAAA,MACR,MAAM;AAAA,MACN,IAAI,KAAK,MAAM;AAAA,IACjB;AACA,QAAI,MAAM;AACR,gBAAU,OAAO;AAAA,IACnB;AACA,SAAK,MAAM,WAAW,KAAK,SAAS;AACpC,SAAK,MAAM,cAAc;AAAA,EAC3B;AAAA,EACA,SAAS,QAAQ;AACf,UAAM,yBAAyB,sBAAsB,MAAM;AAC3D,UAAM,4BAA4B,IAAI,IAAI,uBAAuB,IAAI,CAAC,MAAM,EAAE,EAAE,CAAC;AACjF,aAAS,IAAI,GAAG,IAAI,KAAK,MAAM,YAAY,MAAM,QAAQ,KAAK;AAC5D,YAAM,UAAU,KAAK,MAAM,YAAY,MAAM;AAC7C,UAAI,CAAC,QAAQ,QAAQ,CAAC,0BAA0B,IAAI,QAAQ,EAAE,GAAG;AAC/D,gBAAQ,QAAQ;AAAA,MAClB;AAAA,IACF;AACA,UAAM,WAAW,KAAK,MAAM,WAAW,IAAI;AAC3C,SAAK,MAAM,cAAc,KAAK,MAAM,WAAW,WAAW,IAAI,OAAO,KAAK,MAAM,WAAW,KAAK,MAAM,WAAW,SAAS;AAC1H,2BAAuB,QAAQ,CAAC,YAAY;AAC1C,UAAI,CAAC,QAAQ,QAAQ,QAAQ,YAAY,SAAS,IAAI;AACpD,aAAK,MAAM,OAAO;AAAA,MACpB;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,UAAU,GAAG,IAAI,IAAI,mBAAmB,OAAO;AAC7C,WAAO,GAAG,SAAS,GAAG,MAAM,2CAA2C;AACvE,QAAI,MAAM,QAAQ,GAAG,UAAU,WAAW;AACxC,YAAM,IAAI,MAAM,0CAA0C,GAAG,QAAQ;AAAA,IACvE;AACA,UAAM,IAAI,KAAK,UAAU,MAAM,KAAK,UAAU,GAAG,MAAM,KAAK,QAAQ,GAAG,MAAM,KAAK,KAAK,WAAW,CAAC,CAAC;AACpG,WAAO,aAAa,QAAQ,MAAM,gDAAgD;AAClF,UAAM,eAAe,qBAAqB,KAAK,MAAM,YAAY,IAAI,CAAC;AACtE,QAAI,CAAC,oBAAoB,aAAa,WAAW,KAAK,GAAG,SAAS,GAAG;AACnE,YAAM,IAAI,MAAM,qIAAqI;AAAA,IACvJ;AACA,WAAO,KAAK,KAAK,YAAY,MAAM;AACjC,YAAM,yBAAyB,CAAC;AAChC,6BAAuB,EAAE,MAAM,MAAM,OAAO,KAAK,EAAE,KAAK,IAAI;AAC5D;AAAA,QACE;AAAA,QACA;AAAA,QACA,CAAC,OAAO,KAAK,KAAK,EAAE;AAAA,QACpB;AAAA,MACF;AACA,YAAM,SAAS,GAAG,IAAI,CAAC,MAAM,uBAAuB,EAAE,GAAG;AACzD,UAAI,KAAK,MAAM,kBAAkB,GAAG;AAClC,aAAK,MAAM,WAAW,QAAQ,CAAC,SAAS;AACtC,qBAAW,WAAW,KAAK,OAAO;AAChC,oBAAQ,QAAQ;AAAA,UAClB;AAAA,QACF,CAAC;AACD,aAAK,MAAM,aAAa;AAAA,MAC1B;AACA,aAAO,EAAE,OAAO,GAAG,OAAO,OAAO;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,WAAW,GAAG;AACZ,WAAO,WAAW,CAAC,GAAG,MAAM,mDAAmD;AAC/E,WAAO,IAAI,WAAW;AACpB,aAAO,OAAO,MAAM,CAAC,MAAM,aAAa,MAAM,GAAG,MAAM,kEAAkE;AACzH,UAAI;AACJ,YAAM,WAAW,CAAC;AAClB,aAAO,QAAQ,CAAC,QAAQ,MAAM;AAC5B,iBAAS,KAAK;AAAA,MAChB,CAAC;AACD,YAAM,cAAc,CAAC,GAAG,SAAS;AAC/B,cAAM,EAAE,GAAG,CAAC,GAAG,QAAQ,IAAI,CAAC;AAC5B,eAAO,IAAI,iBAAiB,QAAQ,MAAM,4FAA4F;AACtI,eAAO,WAAW,IAAI,QAAQ,GAAG,MAAM,kGAAkG;AACzI,eAAO,IAAI;AAAA,MACb;AACA,YAAM,gBAAgB,CAAC,IAAI,UAAU;AACnC,cAAM,UAAU,IAAI,SAAS,IAAI,KAAK;AACtC,cAAM,SAAS,MAAM,QAAQ,OAAO,IAAI,UAAU,CAAC,OAAO;AAC1D,eAAO,OAAO,WAAW,OAAO,QAAQ,MAAM,qKAAqK;AACnN,eAAO,OAAO,MAAM,CAAC,MAAM,aAAa,MAAM,GAAG,MAAM,sIAAsI;AAC7L,cAAM,UAAU,CAAC;AACjB,eAAO,QAAQ,CAAC,OAAO,MAAM;AAC3B,kBAAQ,KAAK,MAAM;AAAA,QACrB,CAAC;AACD,eAAO;AAAA,MACT;AACA,aAAO,KAAK,cAAc;AAAA,QACxB;AAAA,QACA;AAAA,QACA,QAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EACA,SAAS,QAAQ;AACf,UAAM,OAAO,KAAK,MAAM,WAAW,IAAI,MAAM;AAC7C,WAAO,KAAK,QAAQ,SAAS,MAAM;AAAA,EACrC;AAAA,EACA,KAAK,QAAQ;AACX,UAAM,OAAO,KAAK,MAAM,WAAW,IAAI,MAAM;AAC7C,WAAO,KAAK,QAAQ,KAAK,MAAM;AAAA,EACjC;AAAA,EACA,UAAU,QAAQ,SAAS;AACzB,UAAM,OAAO,KAAK,MAAM,WAAW,IAAI,MAAM;AAC7C,WAAO,KAAK,QAAQ,UAAU,QAAQ,OAAO;AAAA,EAC/C;AAAA,EACA,MAAM,KAAK,OAAO;AAChB,UAAM,QAAQ,IAAI;AAClB,UAAM,aAAa,MAAM,KAAK,QAAQ,KAAK,KAAK;AAChD,eAAW,SAAS,IAAI,IAAI;AAC5B,WAAO;AAAA,EACT;AAAA,EACA,MAAM,QAAQ;AACZ,QAAI,KAAK,MAAM,eAAe,MAAM;AAClC,aAAO,UAAU,KAAK,MAAM,YAAY;AACxC,WAAK,MAAM,YAAY,MAAM,KAAK,MAAM;AAAA,IAC1C;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,sBAAsB;AACxB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,QAAQ;AACN,SAAK;AACL,SAAK,MAAM,QAAQ;AACnB,SAAK,IAAI,MAAM;AACf,SAAK,QAAQ,IAAI,YAAY;AAC7B,eAAW,eAAe,KAAK,UAAU;AACvC,WAAK,yBAAyB,WAAW;AACzC,WAAK,SAAS,aAAa,QAAQ;AACnC,aAAO,KAAK,SAAS;AAAA,IACvB;AACA,SAAK,cAAc;AACnB,SAAK,kBAAkB;AACvB,SAAK,qBAAqB;AAAA,EAC5B;AACF;AACA,OAAO,eAAe;AACtB,OAAO,iBAAiB;AACxB,SAAS,KAAK,OAAO;AACnB,QAAM,SAAS,mBAAmB,cAAc,KAAK,GAAG,SAAS;AACjE,SAAO,OAAO,WAAW,QAAQ,OAAO,SAAS;AACnD;AACA,SAAS,kBAAkB;AACzB,QAAM,KAAK,mBAAmB;AAC9B,MAAI,GAAG,aAAa,MAAM;AACxB,UAAME,eAAc,IAAI,YAAY,EAAE;AACtC,OAAG,YAAY,IAAI,OAAOA,YAAW;AAAA,EACvC;AACA,uBAAqB,GAAG,UAAU,GAAG;AACrC,mBAAiB,MAAM,GAAG,SAAS;AACnC,SAAO,GAAG;AACZ;AACA,IAAI,SAAS,gBAAgB;AAC7B,SAAS,IAAI,GAAG,GAAG;AACjB,QAAM,SAAS,EAAE,GAAG,EAAE;AACtB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AAGA,IAAI,sBAAsB,CAAC;AAC3BF,UAAS,qBAAqB;AAAA,EAC5B,WAAW,MAAM;AAAA,EACjB,UAAU,MAAM;AAAA,EAChB,cAAc,MAAM;AACtB,CAAC;AACD,SAAS,sBAAsB;AAC7B,SAAO,OAAO,cAAc,eAAe,aAAa;AAC1D;AACA,IAAI;AACJ,SAAS,aAAa,OAAO;AAC3B,sBAAoB;AACtB;AACA,SAAS,SAAS,KAAK;AACrB,MAAI,sBAAsB,QAAQ;AAChC,WAAO;AAAA,EACT;AACA,MAAI,OAAO,oBAAoB,GAAG;AAChC,QAAI,CAAC,KAAK;AACR,YAAM;AAAA,IACR;AACA,QAAI,IAAI,YAAY,eAAe;AACjC,aAAO;AAAA,IACT;AACA,UAAM,IAAI,IAAI,aAAa,IAAI,WAAW,OAAO,WAAW,cAAc,OAAO,QAAQ;AACzF,QAAI,CAAC,GAAG;AACN,YAAM,SAAS;AACf,aAAO,OAAO,iBAAiB,OAAO,cAAc;AAAA,IACtD;AACA,WAAO,2TAA2T,KAAK,CAAC,KAAK,0kDAA0kD,KAAK,EAAE,OAAO,GAAG,CAAC,CAAC;AAAA,EAC56D;AACA,SAAO;AACT;AACA,SAAS,YAAY;AACnB,SAAO,OAAO,WAAW,eAAe,OAAO,YAAY,QAAQ,OAAO,sBAAsB;AAClG;AAGA,IAAI,OAAO,IAAI;AACf,KAAK,aAAa,SAAS,MAAM,OAAO,CAAC,eAAe;AACtD,MAAI,YAAY;AACd,YAAQ,KAAK,6IAA6I;AAAA,EAC5J;AACF,CAAC;AACD,KAAK,aAAa,cAAc,MAAM,UAAU,CAAC;AACjD,KAAK,aAAa,WAAW,MAAM,OAAO,YAAY,eAAe,OAAO,QAAQ,aAAa,eAAe,OAAO,QAAQ,SAAS,SAAS,WAAW;AAC5J,KAAK,aAAa,aAAa,MAAM,OAAO,cAAc,eAAe,aAAa,QAAQ,UAAU,aAAa,QAAQ,SAAS,KAAK,UAAU,SAAS,KAAK,aAAa,KAAK,UAAU,MAAM,CAAC;AACtM,KAAK,aAAa,QAAQ,MAAM,KAAK;AACrC,KAAK,aAAa,sCAAsC,MAAM,KAAK,QAAQ,OAAO,CAAC;AACnF,KAAK,aAAa,gCAAgC,MAAM,IAAI;AAC5D,KAAK,aAAa,WAAW,MAAM,KAAK;AACxC,KAAK,aAAa,gCAAgC,MAAM,IAAI;AAC5D,KAAK,aAAa,uBAAuB,MAAM,KAAK;AACpD,KAAK,aAAa,uBAAuB,MAAM,KAAK;AACpD,KAAK,aAAa,yCAAyC,MAAM,KAAK;AACtE,KAAK,aAAa,wBAAwB,MAAM,KAAK;AAGrD,SAAS,WAAW,KAAK,OAAO;AAC9B,MAAI,YAAY;AAChB,MAAI,aAAa,GAAG,GAAG;AACrB,WAAO,UAAU,WAAW,CAAC,IAAI,CAAC,IAAI,MAAM;AAAA,EAC9C;AACA,MAAI,OAAO,QAAQ,YAAY,aAAa,KAAK;AAC/C,UAAM,eAAe,IAAI,YAAY;AACrC,WAAO,CAAC,IAAI,QAAQ,IAAI,QAAQ,aAAa,MAAM;AAAA,EACrD;AACA,MAAI,CAAC,MAAM,QAAQ,GAAG,GAAG;AACvB,WAAO,CAAC;AAAA,EACV;AACA,QAAM,QAAQ,CAAC;AACf,SAAO,MAAM,QAAQ,SAAS,KAAK,aAAa,SAAS,KAAK,UAAU,UAAU;AAChF,UAAM,KAAK,UAAU,MAAM;AAC3B,gBAAY,UAAU;AAAA,EACxB;AACA,MAAI,MAAM,QAAQ,GAAG,KAAK,IAAI,EAAE,QAAQ,oCAAoC,GAAG;AAC7E,+BAA2B,KAAK,OAAO,CAAC,CAAC;AAAA,EAC3C;AACA,SAAO;AACT;AACA,SAAS,2BAA2B,KAAK,OAAO,SAAS;AACvD,YAAU,WAAW,CAAC;AACtB,MAAI,CAAC,MAAM,QAAQ,GAAG,KAAK,CAAC,aAAa,GAAG,GAAG;AAC7C,WAAO,MAAM,WAAW,GAAG,MAAM,eAAe,QAAQ,KAAK,IAAI,2DAA2D,MAAM,aAAa;AAC/I;AAAA,EACF;AACA,SAAO,MAAM,SAAS,GAAG,MAAM,eAAe,QAAQ,KAAK,IAAI,gDAAgD,IAAI,iBAAiB;AACpI,SAAO,IAAI,WAAW,MAAM,IAAI,MAAM,eAAe,QAAQ,KAAK,IAAI,kBAAkB,MAAM,wBAAwB,IAAI,iBAAiB;AAC3I,QAAM,WAAW,MAAM,MAAM,CAAC;AAC9B,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,EAAE,GAAG;AACnC,+BAA2B,IAAI,IAAI,UAAU,QAAQ,OAAO,CAAC,CAAC;AAAA,EAChE;AACF;AACA,SAAS,YAAY,eAAe,aAAa,SAAS,cAAc;AACtE,MAAI,kBAAkB,qBAAqB;AACzC;AAAA,EACF;AACA,MAAI,iBAAiB,MAAM;AACzB,UAAM,IAAI,MAAM,gCAAgC;AAAA,EAClD;AACA,MAAI,kBAAkB,aAAa,kBAAkB,eAAe,kBAAkB,aAAa,gBAAgB,UAAU;AAC3H,UAAM,IAAI,MAAM,aAAa,uBAAuB,yBAAyB,iCAAiC,oBAAoB;AAAA,EACpI;AACF;AACA,SAAS,gBAAgB,GAAG,SAAS,cAAc,eAAe,WAAW;AAC3E,MAAI,aAAa,QAAQ;AACvB,gBAAY,cAAc,EAAE,OAAO,SAAS,YAAY;AACxD,WAAO;AAAA,EACT;AACA,MAAI,gBAAgB,WAAW,CAAC;AAChC,MAAI,kBAAkB,YAAY,CAAC,QAAQ,SAAS,SAAS,EAAE,QAAQ,YAAY,KAAK,GAAG;AACzF,oBAAgB;AAAA,EAClB;AACA,cAAY,cAAc,eAAe,SAAS,YAAY;AAC9D,MAAI,KAAK,QAAQ,CAAC,aAAa,CAAC,KAAK,CAAC,MAAM,QAAQ,CAAC,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,aAAa,OAAO,MAAM,UAAU;AAClI,UAAM,OAAO,KAAK,OAAO,SAAS,EAAE,YAAY;AAChD,UAAM,IAAI,MAAM,aAAa,uBAAuB,0DAA0D,OAAO;AAAA,EACvH;AACA,QAAM,gBAAgB,WAAW,GAAG,aAAa;AACjD,MAAI,CAAC,aAAa,CAAC,KAAK,CAAC,MAAM,QAAQ,CAAC,GAAG;AACzC,QAAI,CAAC,CAAC;AAAA,EACR;AACA,QAAM,iBAAiB;AACvB,QAAM,SAAS,kBAAkB,WAAW,aAAa,GAAG,aAAa,IAAI,QAAQ,GAAG,CAAC,GAAG,cAAc;AAC1G,SAAO,OAAO,WAAW,QAAQ,eAAe,aAAa;AAC/D;AACA,SAAS,qBAAqB,KAAK,SAAS,cAAc,eAAe,WAAW;AAClF,MAAI,CAAC,MAAM,QAAQ,GAAG,GAAG;AACvB,UAAM,IAAI,MAAM,YAAY,qBAAqB,yDAAyD;AAAA,EAC5G;AACA,QAAM,UAAU;AAChB,SAAO,QAAQ,IAAI,CAAC,GAAG,MAAM,gBAAgB,GAAG,GAAG,WAAW,MAAM,cAAc,YAAY,CAAC;AACjG;AAGA,IAAI,kBAAkB;AACtB,SAAS,GAAG,GAAG;AACb,QAAM,OAAO,OAAO,KAAK,CAAC;AAC1B,MAAI,KAAK,WAAW,GAAG;AACrB,UAAM,IAAI,MAAM,yGAAyG,KAAK,cAAc;AAAA,EAC9I;AACA,MAAI,SAAS,KAAK;AAClB,QAAM,KAAK,EAAE;AACb,MAAI,OAAO,SAAS,GAAG,GAAG;AACxB,aAAS,OAAO,UAAU,GAAG,OAAO,SAAS,CAAC;AAAA,EAChD;AACA,WAAS,SAAS;AAClB,QAAM,KAAK,IAAI,SAAS;AACtB,WAAO,WAAW,MAAM;AACxB,QAAI;AACF,YAAM,SAAS,GAAG,GAAG,IAAI;AACzB,UAAI,UAAU,MAAM,GAAG;AACrB,gBAAQ,MAAM,yCAAyC;AAAA,MACzD;AACA,aAAO,SAAS,MAAM;AACtB,aAAO;AAAA,IACT,SAAS,IAAP;AACA,aAAO,SAAS,IAAI;AACpB,YAAM;AAAA,IACR;AAAA,EACF;AACA,SAAO,eAAe,IAAI,QAAQ,EAAE,OAAO,QAAQ,cAAc,KAAK,CAAC;AACvE,SAAO;AACT;AAGA,SAAS,SAAS,OAAO,OAAO;AAC9B,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,SAAS;AACtD,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,SAAS;AACtD,oBAAkB,MAAM,OAAO,MAAM,OAAO,yBAAyB,MAAM,aAAa,MAAM,4CAA4C;AAC1I,QAAM,SAAS,EAAE,MAAM,OAAO,MAAM,MAAM;AAC1C,SAAO,OAAO,UAAU,SAAS,MAAM;AACzC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,QAAQ,OAAO,eAAe,OAAO;AACvD,MAAI,SAAS,MAAM;AACjB,YAAQ,WAAW,MAAM;AAAA,EAC3B;AACA,MAAI,UAAU,aAAa;AACzB,UAAM,IAAI,MAAM,kFAAkF;AAAA,EACpG;AACA,MAAI,OAAO,WAAW,YAAY,aAAa,QAAQ;AACrD,QAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,YAAM,IAAI,MAAM,0FAA0F,QAAQ;AAAA,IACpH;AACA,WAAO,WAAW,OAAO,YAAY;AACrC,WAAO,OAAO,QAAQ,wBAAwB,QAAQ,SAAS,eAAe,KAAK;AAAA,EACrF;AACA,MAAI,CAAC,aAAa,MAAM,KAAK,CAAC,MAAM,QAAQ,MAAM,KAAK,OAAO,WAAW,YAAY,OAAO,WAAW,aAAa,OAAO,WAAW,UAAU;AAC9I,UAAM,IAAI,MAAM,0HAA0H;AAAA,EAC5I;AACA,MAAI,SAAS,MAAM;AACjB,uCAAmC,KAAK;AACxC,UAAM,eAAe,cAAc,KAAK;AACxC,UAAM,eAAe,cAAc,aAAa;AAChD,WAAO,iBAAiB,cAAc,MAAM,iCAAiC,kCAAkC,+BAA+B,cAAc;AAC5J,aAAS,IAAI,GAAG,IAAI,cAAc,QAAQ,EAAE,GAAG;AAC7C,YAAM,WAAW,cAAc;AAC/B,YAAM,oBAAoB,MAAM,cAAc,SAAS,IAAI,aAAa,cAAc,MAAM,MAAM,CAAC,CAAC,IAAI;AACxG,aAAO,cAAc,OAAO,MAAM,MAAM,CAAC,mBAAmB,MAAM,gDAAgD,qDAAqD,UAAU;AAAA,IACnL;AAAA,EACF;AACA,MAAI,CAAC,aAAa,MAAM,KAAK,CAAC,MAAM,QAAQ,MAAM,GAAG;AACnD,aAAS,CAAC,MAAM;AAAA,EAClB;AACA,UAAQ,SAAS;AACjB,WAAS,UAAU,WAAW,aAAa,QAAQ,KAAK,IAAI,QAAQ,QAAQ,CAAC,GAAG,IAAI;AACpF,SAAO,OAAO,WAAW,QAAQ,OAAO,KAAK;AAC/C;AAGA,SAAS,OAAO,QAAQ,OAAO,OAAO;AACpC,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,IAAI,uBAAuB;AAAA,EACzB,WAAW;AAAA,EACX,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AAAA,EACV,SAAS;AAAA,EACT,QAAQ;AAAA,EACR,aAAa;AACf;AAGA,IAAI,0BAA0B;AAC9B,eAAe,cAAc,SAAS,OAAO;AAC3C,QAAM,QAAQ,CAAC;AACf,QAAM,eAAe,CAAC;AACtB,QAAM,QAAQ,MAAM,QAAQ,OAAO,IAAI,QAAQ,IAAI,CAAC,YAAY,QAAQ,IAAI,IAAI,OAAO,KAAK,OAAO;AACnG,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,UAAM,OAAO,MAAM;AACnB,UAAM,IAAI,MAAM,QAAQ,OAAO,IAAI,QAAQ,GAAG,SAAS,QAAQ;AAC/D,QAAI,EAAE,UAAU,aAAa,EAAE,UAAU,WAAW,EAAE,UAAU,UAAU,EAAE,UAAU,YAAY,EAAE,UAAU,aAAa;AACzH,YAAM,IAAI,MAAM,gCAAgC,UAAU,EAAE,OAAO;AAAA,IACrE;AACA,UAAM,OAAO,EAAE,MAAM,OAAO,EAAE,OAAO,OAAO,EAAE,MAAM;AACpD,QAAI,EAAE,UAAU,UAAU;AACxB,YAAM,YAAY,IAAI,QAAQ,OAAO,YAAY;AAC/C,cAAM,OAAO,MAAM,EAAE,MAAM;AAC3B,cAAM,gBAAgB,KAAK,OAAO,CAAC,IAAI,MAAM,KAAK,EAAE,QAAQ,CAAC,IAAI,0BAA0B,KAAK;AAChG,cAAM,QAAQ,IAAI,WAAW,aAAa;AAC1C,YAAI,SAAS;AACb,iBAAS,KAAK,GAAG,KAAK,KAAK,QAAQ,MAAM;AACvC,gBAAM,MAAM,KAAK;AACjB,gBAAM,gBAAgB,IAAI,WAAW,IAAI,YAAY,CAAC,IAAI,MAAM,CAAC,EAAE,MAAM;AACzE,gBAAM,IAAI,eAAe,MAAM;AAC/B,oBAAU;AACV,gBAAM,IAAI,KAAK,MAAM;AACrB,oBAAU,IAAI;AAAA,QAChB;AACA,gBAAQ,KAAK;AAAA,MACf,CAAC;AACD,mBAAa,KAAK,SAAS;AAAA,IAC7B,OAAO;AACL,mBAAa,KAAK,EAAE,KAAK,CAAC;AAAA,IAC5B;AACA,QAAI,SAAS,MAAM;AACjB,WAAK,QAAQ;AAAA,IACf;AACA,UAAM,KAAK,IAAI;AAAA,EACjB;AACA,QAAM,eAAe,MAAM,QAAQ,IAAI,YAAY;AACnD,SAAO,EAAE,MAAM,uBAAuB,YAAY,GAAG,MAAM;AAC7D;AACA,SAAS,cAAc,SAAS,OAAO;AACrC,QAAM,MAAM,CAAC;AACb,MAAI;AACJ,MAAI,SAAS;AACb,aAAW,QAAQ,OAAO;AACxB,UAAM,OAAO,KAAK;AAClB,UAAM,QAAQ,KAAK;AACnB,UAAM,QAAQ,KAAK;AACnB,UAAM,OAAO,cAAc,KAAK;AAChC,QAAI;AACJ,QAAI,kBAAkB,MAAM;AAC1B,YAAM,eAAe,KAAK;AAC1B,UAAI,aAAa,UAAU,WAAW,aAAa,UAAU,UAAU;AACrE,YAAI,EAAE,SAAS,gBAAgB,WAAW,eAAe;AACvD,gBAAM,IAAI,MAAM,UAAU,KAAK,0BAA0B,aAAa,0DAA0D;AAAA,QAClI;AAAA,MACF,WAAW,aAAa,UAAU,WAAW;AAC3C,YAAI,UAAU,WAAW;AACvB,gBAAM,IAAI,MAAM,UAAU,KAAK,0BAA0B,aAAa,yDAAyD,QAAQ;AAAA,QACzI;AAAA,MACF,OAAO;AACL,cAAM,IAAI,MAAM,UAAU,KAAK,uCAAuC,aAAa,6EAA6E;AAAA,MAClK;AACA,YAAM,yBAAyB,qBAAqB,aAAa;AACjE,YAAM,aAAa,QAAQ,MAAM,QAAQ,SAAS,OAAO,sBAAsB;AAC/E,YAAM,iBAAiB,aAAa,UAAU,UAAU,IAAI,WAAW,UAAU,IAAI,IAAI,YAAY,UAAU;AAC/G,UAAI,UAAU,WAAW;AACvB,YAAI,aAAa,UAAU,WAAW,aAAa,UAAU,UAAU;AACrE,mBAAS,IAAI,aAAa,eAAe,MAAM;AAC/C,mBAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,kBAAM,IAAI,eAAe;AACzB,mBAAO,KAAK,IAAI,aAAa,QAAQ,aAAa;AAAA,UACpD;AAAA,QACF,WAAW,aAAa,UAAU,WAAW;AAC3C,cAAI,kBAAkB,QAAQ;AAC5B,4BAAgB,kBAAkB;AAAA,UACpC;AACA,mBAAS,cAAc,cAAc;AAAA,QACvC,OAAO;AACL,gBAAM,IAAI,MAAM,iCAAiC,aAAa,gCAAgC;AAAA,QAChG;AAAA,MACF,WAAW,UAAU,SAAS;AAC5B,YAAI,aAAa,UAAU,WAAW,aAAa,UAAU,UAAU;AACrE,gBAAM,IAAI,MAAM,iCAAiC,aAAa,8BAA8B;AAAA,QAC9F;AACA,iBAAS,IAAI,WAAW,eAAe,MAAM;AAC7C,iBAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,gBAAM,IAAI,eAAe;AACzB,iBAAO,KAAK,KAAK,MAAM,IAAI,aAAa,QAAQ,aAAa,GAAG;AAAA,QAClE;AAAA,MACF,OAAO;AACL,cAAM,IAAI,MAAM,gCAAgC,UAAU,OAAO;AAAA,MACnE;AACA,gBAAU,OAAO;AAAA,IACnB,WAAW,UAAU,UAAU;AAC7B,YAAM,QAAQ,cAAc,KAAK,KAAK;AACtC,eAAS,CAAC;AACV,eAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,cAAM,aAAa,IAAI,YAAY,QAAQ,MAAM,QAAQ,SAAS,uBAAuB,CAAC,EAAE;AAC5F,kBAAU;AACV,cAAM,QAAQ,IAAI,WAAW,QAAQ,MAAM,QAAQ,SAAS,UAAU,CAAC;AACvE,eAAO,KAAK,KAAK;AACjB,kBAAU;AAAA,MACZ;AAAA,IACF,OAAO;AACL,YAAM,cAAc,qBAAqB;AACzC,YAAM,aAAa,QAAQ,MAAM,QAAQ,SAAS,OAAO,WAAW;AACpE,UAAI,UAAU,WAAW;AACvB,iBAAS,IAAI,aAAa,UAAU;AAAA,MACtC,WAAW,UAAU,SAAS;AAC5B,iBAAS,IAAI,WAAW,UAAU;AAAA,MACpC,WAAW,UAAU,QAAQ;AAC3B,iBAAS,IAAI,WAAW,UAAU;AAAA,MACpC,WAAW,UAAU,aAAa;AAChC,iBAAS,IAAI,aAAa,UAAU;AACpC,cAAM,QAAQ,IAAI,aAAa,OAAO,SAAS,CAAC;AAChD,cAAM,SAAS,IAAI,aAAa,OAAO,SAAS,CAAC;AACjD,iBAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,gBAAM,KAAK,OAAO,IAAI;AACtB,iBAAO,KAAK,OAAO,IAAI,IAAI;AAAA,QAC7B;AACA,cAAM,aAAa,OAAO,OAAO,OAAO,SAAS;AACjD,cAAM,cAAc,OAAO,QAAQ,OAAO,SAAS;AACnD,YAAI,QAAQ,QAAQ,YAAY,WAAW;AAC3C,mBAAW,QAAQ;AACnB,oBAAY,QAAQ;AAAA,MACtB,OAAO;AACL,cAAM,IAAI,MAAM,gCAAgC,UAAU,OAAO;AAAA,MACnE;AACA,gBAAU,OAAO;AAAA,IACnB;AACA,QAAI,UAAU,aAAa;AACzB,UAAI,QAAQ,OAAO,QAAQ,OAAO,KAAK;AAAA,IACzC;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,uBAAuB,IAAI;AAClC,MAAI,OAAO,MAAM;AACf,UAAM,IAAI,MAAM,wBAAwB,KAAK,UAAU,EAAE,GAAG;AAAA,EAC9D;AACA,MAAI,kBAAkB;AACtB,QAAM,eAAe,CAAC;AACtB,KAAG,QAAQ,CAAC,MAAM;AAChB,uBAAmB,EAAE;AACrB,iBAAa,KAAK,EAAE,eAAe,EAAE,OAAO,aAAa,IAAI,IAAI,EAAE,YAAY,CAAC,CAAC;AACjF,QAAI,EAAE,aAAa,gBAAgB,aAAa,cAAc,aAAa,aAAa;AACtF,YAAM,IAAI,MAAM,mCAAmC,EAAE,YAAY,MAAM;AAAA,IACzE;AAAA,EACF,CAAC;AACD,QAAM,IAAI,IAAI,WAAW,eAAe;AACxC,MAAI,SAAS;AACb,eAAa,QAAQ,CAAC,MAAM;AAC1B,MAAE,IAAI,IAAI,WAAW,EAAE,MAAM,GAAG,MAAM;AACtC,cAAU,EAAE;AAAA,EACd,CAAC;AACD,SAAO,EAAE;AACX;AACA,IAAI,gBAAgB,OAAO,WAAW,gBAAgB,OAAO,SAAS,eAAe,OAAO,SAAS,eAAe,OAAO,SAAS;AACpI,SAAS,iBAAiB,KAAK;AAC7B,MAAI,eAAe;AACjB,WAAO,OAAO,WAAW,GAAG;AAAA,EAC9B;AACA,SAAO,IAAI,KAAK,CAAC,GAAG,CAAC,EAAE;AACzB;AACA,SAAS,0BAA0B,SAAS;AAC1C,MAAI,eAAe;AACjB,WAAO,OAAO,KAAK,OAAO,EAAE,SAAS,QAAQ;AAAA,EAC/C;AACA,QAAM,MAAM,IAAI,WAAW,OAAO;AAClC,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,IAAI,GAAG,KAAK;AAC1C,SAAK,OAAO,aAAa,IAAI,EAAE;AAAA,EACjC;AACA,SAAO,KAAK,CAAC;AACf;AACA,SAAS,0BAA0B,KAAK;AACtC,MAAI,eAAe;AACjB,UAAM,MAAM,OAAO,KAAK,KAAK,QAAQ;AACrC,WAAO,IAAI,OAAO,MAAM,IAAI,YAAY,IAAI,aAAa,IAAI,UAAU;AAAA,EACzE;AACA,QAAM,IAAI,KAAK,GAAG;AAClB,QAAM,UAAU,IAAI,WAAW,EAAE,MAAM;AACvC,WAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,EAAE,GAAG;AACjC,YAAQ,IAAI,CAAC,EAAE,WAAW,CAAC,CAAC,GAAG,CAAC;AAAA,EAClC;AACA,SAAO,QAAQ;AACjB;AACA,SAAS,wBAAwB,SAAS;AACxC,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,QAAQ;AAAA,EACjB;AACA,MAAI,kBAAkB;AACtB,UAAQ,QAAQ,CAAC,YAAY;AAC3B,uBAAmB,QAAQ;AAAA,EAC7B,CAAC;AACD,QAAM,OAAO,IAAI,WAAW,eAAe;AAC3C,MAAI,SAAS;AACb,UAAQ,QAAQ,CAAC,YAAY;AAC3B,SAAK,IAAI,IAAI,WAAW,OAAO,GAAG,MAAM;AACxC,cAAU,QAAQ;AAAA,EACpB,CAAC;AACD,SAAO,KAAK;AACd;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,YAAY;AAClB,SAAO,KAAK,KAAK;AACjB,SAAO,KAAK,SAAS,SAAS,GAAG;AAC/B,WAAO,KAAK,MAAM,GAAG,KAAK,SAAS,CAAC;AAAA,EACtC;AACA,QAAM,QAAQ,KAAK,MAAM,SAAS;AAClC,SAAO,MAAM,MAAM,SAAS;AAC9B;AACA,SAAS,8BAA8B,WAAW,UAAU;AAC1D,QAAM,SAAS;AAAA,IACb,eAAe,UAAU;AAAA,IACzB,QAAQ,UAAU;AAAA,IAClB,aAAa,UAAU;AAAA,IACvB,aAAa,UAAU;AAAA,IACvB,iBAAiB;AAAA,EACnB;AACA,MAAI,UAAU,aAAa,MAAM;AAC/B,WAAO,YAAY,UAAU;AAAA,EAC/B;AACA,MAAI,UAAU,uBAAuB,MAAM;AACzC,WAAO,sBAAsB,UAAU;AAAA,EACzC;AACA,MAAI,UAAU,oBAAoB,MAAM;AACtC,WAAO,mBAAmB,UAAU;AAAA,EACtC;AACA,MAAI,UAAU,wBAAwB,MAAM;AAC1C,WAAO,uBAAuB,UAAU;AAAA,EAC1C;AACA,MAAI,UAAU,kBAAkB,MAAM;AACpC,WAAO,iBAAiB,UAAU;AAAA,EACpC;AACA,SAAO;AACT;AACA,SAAS,6BAA6B,WAAW,aAAa,YAAY;AACxE,QAAM,iBAAiB;AAAA,IACrB,eAAe,UAAU;AAAA,IACzB,QAAQ,UAAU;AAAA,IAClB,aAAa,UAAU;AAAA,IACvB,aAAa,UAAU;AAAA,EACzB;AACA,MAAI,UAAU,kBAAkB,MAAM;AACpC,mBAAe,iBAAiB,UAAU;AAAA,EAC5C;AACA,MAAI,UAAU,mBAAmB,MAAM;AACrC,QAAI,CAAC,aAAa;AAChB,YAAM,IAAI,MAAM,uDAAuD;AAAA,IACzE;AACA,QAAI,CAAC,YAAY;AACf,YAAM,IAAI,MAAM,sDAAsD;AAAA,IACxE;AACA,mBAAe,cAAc;AAC7B,mBAAe,aAAa;AAAA,EAC9B;AACA,MAAI,UAAU,aAAa,MAAM;AAC/B,mBAAe,YAAY,UAAU;AAAA,EACvC;AACA,MAAI,UAAU,uBAAuB,MAAM;AACzC,mBAAe,sBAAsB,UAAU;AAAA,EACjD;AACA,MAAI,UAAU,oBAAoB,MAAM;AACtC,mBAAe,mBAAmB,UAAU;AAAA,EAC9C;AACA,MAAI,UAAU,wBAAwB,MAAM;AAC1C,mBAAe,uBAAuB,UAAU;AAAA,EAClD;AACA,SAAO;AACT;AACA,eAAe,yBAAyB,WAAW,cAAc;AAC/D,MAAI;AACJ,MAAI;AACJ,MAAI,UAAU,mBAAmB,MAAM;AACrC,KAAC,aAAa,UAAU,IAAI,MAAM,aAAa,UAAU,eAAe;AAAA,EAC1E;AACA,SAAO,6BAA6B,WAAW,aAAa,UAAU;AACxE;AACA,SAAS,6BAA6B,gBAAgB;AACpD,MAAI,eAAe,yBAAyB,aAAa;AACvD,UAAM,IAAI,MAAM,qDAAqD;AAAA,EACvE;AACA,SAAO;AAAA,IACL,WAAW,IAAI,KAAK;AAAA,IACpB,mBAAmB;AAAA,IACnB,oBAAoB,eAAe,iBAAiB,OAAO,IAAI,iBAAiB,KAAK,UAAU,eAAe,aAAa,CAAC;AAAA,IAC5H,kBAAkB,eAAe,eAAe,OAAO,IAAI,iBAAiB,KAAK,UAAU,eAAe,WAAW,CAAC;AAAA,IACtH,iBAAiB,eAAe,cAAc,OAAO,IAAI,eAAe,WAAW;AAAA,EACrF;AACF;AACA,SAAS,eAAe,iBAAiB;AACvC,QAAM,cAAc,CAAC;AACrB,aAAW,SAAS,iBAAiB;AACnC,gBAAY,KAAK,GAAG,MAAM,OAAO;AAAA,EACnC;AACA,SAAO;AACT;AACA,SAAS,6BAA6B;AACpC,QAAM,kBAAkB,CAAC,MAAM;AAC7B,QAAI,IAAI,KAAK;AACb,QAAI,IAAI;AACR,YAAQ,IAAI,aAAa,GAAG;AAC1B,WAAK;AACL,YAAM;AAAA,IACR;AACA,SAAK,CAAC;AACN,SAAK;AACL,WAAO,IAAI;AAAA,EACb;AACA,QAAM,eAAe,IAAI,YAAY,IAAI;AACzC,eAAa,KAAK;AAClB,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,iBAAa,KAAK,gBAAgB,CAAC;AAAA,EACrC;AACA,WAAS,IAAI,MAAM,IAAI,MAAM,KAAK;AAChC,iBAAa,KAAK,aAAa,IAAI,QAAQ;AAAA,EAC7C;AACA,SAAO;AACT;AACA,SAAS,8BAA8B;AACrC,QAAM,gBAAgB,IAAI,YAAY,EAAE;AACxC,gBAAc,KAAK;AACnB,gBAAc,MAAM;AACpB,gBAAc,MAAM;AACpB,gBAAc,MAAM;AACpB,WAAS,IAAI,GAAG,IAAI,IAAI,KAAK;AAC3B,kBAAc,KAAK,KAAK;AAAA,EAC1B;AACA,WAAS,IAAI,IAAI,IAAI,IAAI,KAAK;AAC5B,kBAAc,KAAK,cAAc,IAAI,MAAM;AAAA,EAC7C;AACA,SAAO;AACT;AACA,SAAS,4BAA4B;AACnC,QAAM,cAAc,IAAI,YAAY,EAAE;AACtC,WAAS,IAAI,GAAG,IAAI,IAAI,KAAK;AAC3B,gBAAY,KAAK;AAAA,EACnB;AACA,cAAY,KAAK,YAAY,MAAM;AACnC,SAAO;AACT;AACA,SAAS,oBAAoB;AAC3B,QAAM,eAAe,2BAA2B;AAChD,QAAM,gBAAgB,4BAA4B;AAClD,QAAM,cAAc,0BAA0B;AAC9C,SAAO,CAAC,mBAAmB;AACzB,UAAM,UAAU,IAAI,YAAY,IAAI,eAAe,MAAM;AACzD,UAAM,mBAAmB,IAAI,YAAY,OAAO;AAChD,aAAS,QAAQ,GAAG,QAAQ,eAAe,QAAQ,SAAS;AAC1D,YAAM,cAAc,eAAe;AACnC,YAAM,cAAc,aAAa,YAAY,eAAe,OAAO,cAAc,SAAS,cAAc,eAAe;AACvH,uBAAiB,SAAS;AAAA,IAC5B;AACA,WAAO,IAAI,aAAa,OAAO;AAAA,EACjC;AACF;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,cAAc;AACZ,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc,CAAC;AAAA,EACtB;AAAA,EACA,OAAO,cAAc;AACnB,QAAI,iBAAiB,YAAY,MAAM;AACrC,uBAAiB,WAAW,IAAI,iBAAiB;AAAA,IACnD;AACA,WAAO,iBAAiB;AAAA,EAC1B;AAAA,EACA,OAAO,mBAAmB,YAAY;AACpC,qBAAiB,YAAY,EAAE,YAAY,KAAK,UAAU;AAAA,EAC5D;AAAA,EACA,OAAO,mBAAmB,YAAY;AACpC,qBAAiB,YAAY,EAAE,YAAY,KAAK,UAAU;AAAA,EAC5D;AAAA,EACA,OAAO,gBAAgB,KAAK;AAC1B,WAAO,iBAAiB,YAAY,KAAK,MAAM;AAAA,EACjD;AAAA,EACA,OAAO,gBAAgB,KAAK,aAAa;AACvC,WAAO,iBAAiB,YAAY,KAAK,QAAQ,WAAW;AAAA,EAC9D;AAAA,EACA,OAAO,YAAY,KAAK,aAAa,aAAa;AAChD,UAAM,gBAAgB,CAAC;AACvB,UAAM,UAAU,gBAAgB,SAAS,iBAAiB,YAAY,EAAE,cAAc,iBAAiB,YAAY,EAAE;AACrH,YAAQ,QAAQ,CAAC,WAAW;AAC1B,YAAM,UAAU,OAAO,KAAK,WAAW;AACvC,UAAI,YAAY,MAAM;AACpB,sBAAc,KAAK,OAAO;AAAA,MAC5B;AAAA,IACF,CAAC;AACD,WAAO;AAAA,EACT;AACF;AACA,IAAI,qBAAqB,CAAC,eAAe,iBAAiB,mBAAmB,UAAU;AACvF,IAAI,qBAAqB,CAAC,eAAe,iBAAiB,mBAAmB,UAAU;AACvF,IAAI,kBAAkB,CAAC,QAAQ,iBAAiB,gBAAgB,GAAG;AACnE,IAAI,kBAAkB,CAAC,KAAK,gBAAgB,iBAAiB,gBAAgB,KAAK,WAAW;AAG7F,IAAI,gBAAgB;AACpB,IAAI,mBAAmB;AACvB,IAAI,mBAAmB;AACvB,IAAI,kBAAkB;AACtB,SAAS,sBAAsB;AAC7B,MAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,GAAG;AAChC,UAAM,IAAI,MAAM,yFAAyF;AAAA,EAC3G;AACA,QAAM,YAAY,OAAO,WAAW,cAAc,OAAO;AACzD,QAAM,UAAU,UAAU,aAAa,UAAU,gBAAgB,UAAU,mBAAmB,UAAU,eAAe,UAAU;AACjI,MAAI,WAAW,MAAM;AACnB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,SAAO;AACT;AACA,SAAS,cAAc,aAAa;AAClC,QAAM,KAAK,YAAY;AACvB,KAAG,kBAAkB,kBAAkB,EAAE,SAAS,YAAY,CAAC;AAC/D,KAAG,kBAAkB,iBAAiB,EAAE,SAAS,YAAY,CAAC;AAChE;AACA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,WAAW;AACrB,SAAK,YAAY,oBAAoB;AACrC,QAAI,aAAa,QAAQ,CAAC,WAAW;AACnC,YAAM,IAAI,MAAM,gEAAgE;AAAA,IAClF;AACA,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,MAAM,KAAK,gBAAgB;AACzB,QAAI,eAAe,yBAAyB,aAAa;AACvD,YAAM,IAAI,MAAM,0FAA0F;AAAA,IAC5G;AACA,WAAO,KAAK,eAAe,KAAK,WAAW,cAAc;AAAA,EAC3D;AAAA,EACA,MAAM,OAAO;AACX,WAAO,KAAK,eAAe,KAAK,SAAS;AAAA,EAC3C;AAAA,EACA,eAAe,WAAW,gBAAgB;AACxC,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,cAAc,KAAK,UAAU,KAAK,eAAe,gBAAgB;AACvE,kBAAY,kBAAkB,MAAM,cAAc,WAAW;AAC7D,kBAAY,YAAY,MAAM;AAC5B,cAAM,KAAK,YAAY;AACvB,YAAI,kBAAkB,MAAM;AAC1B,gBAAM,UAAU,GAAG,YAAY,kBAAkB,UAAU;AAC3D,gBAAM,aAAa,QAAQ,YAAY,gBAAgB;AACvD,gBAAM,aAAa,WAAW,IAAI,KAAK,SAAS;AAChD,qBAAW,YAAY,MAAM;AAC3B,gBAAI,WAAW,UAAU,MAAM;AAC7B,iBAAG,MAAM;AACT,qBAAO,OAAO,IAAI,MAAM,gCAAgC,KAAK,0BAA0B,CAAC;AAAA,YAC1F,OAAO;AACL,sBAAQ,WAAW,OAAO,cAAc;AAAA,YAC1C;AAAA,UACF;AACA,qBAAW,UAAU,CAAC,UAAU;AAC9B,eAAG,MAAM;AACT,mBAAO,OAAO,WAAW,KAAK;AAAA,UAChC;AACA,kBAAQ,aAAa,MAAM,GAAG,MAAM;AAAA,QACtC,OAAO;AACL,gBAAM,qBAAqB,6BAA6B,cAAc;AACtE,gBAAM,SAAS,GAAG,YAAY,iBAAiB,WAAW;AAC1D,cAAI,YAAY,OAAO,YAAY,eAAe;AAClD,gBAAM,iBAAiB,UAAU,IAAI,EAAE,WAAW,KAAK,WAAW,mBAAmB,CAAC;AACtF,cAAI;AACJ,yBAAe,YAAY,MAAM;AAC/B,sBAAU,GAAG,YAAY,kBAAkB,WAAW;AACtD,kBAAM,aAAa,QAAQ,YAAY,gBAAgB;AACvD,kBAAM,kBAAkB,WAAW,IAAI;AAAA,cACrC,WAAW,KAAK;AAAA,cAChB;AAAA,cACA;AAAA,YACF,CAAC;AACD,4BAAgB,YAAY,MAAM,QAAQ,EAAE,mBAAmB,CAAC;AAChE,4BAAgB,UAAU,CAAC,UAAU;AACnC,0BAAY,OAAO,YAAY,eAAe;AAC9C,oBAAM,oBAAoB,UAAU,OAAO,KAAK,SAAS;AACzD,gCAAkB,YAAY,MAAM;AAClC,mBAAG,MAAM;AACT,uBAAO,OAAO,gBAAgB,KAAK;AAAA,cACrC;AACA,gCAAkB,UAAU,CAAC,WAAW;AACtC,mBAAG,MAAM;AACT,uBAAO,OAAO,gBAAgB,KAAK;AAAA,cACrC;AAAA,YACF;AAAA,UACF;AACA,yBAAe,UAAU,CAAC,UAAU;AAClC,eAAG,MAAM;AACT,mBAAO,OAAO,eAAe,KAAK;AAAA,UACpC;AACA,iBAAO,aAAa,MAAM;AACxB,gBAAI,WAAW,MAAM;AACnB,iBAAG,MAAM;AAAA,YACX,OAAO;AACL,sBAAQ,aAAa,MAAM,GAAG,MAAM;AAAA,YACtC;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,kBAAY,UAAU,CAAC,UAAU,OAAO,YAAY,KAAK;AAAA,IAC3D,CAAC;AAAA,EACH;AACF;AACA,iBAAiB,aAAa;AAC9B,IAAI,kBAAkB,CAAC,QAAQ;AAC7B,MAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,GAAG;AAChC,WAAO;AAAA,EACT,OAAO;AACL,QAAI,CAAC,MAAM,QAAQ,GAAG,KAAK,IAAI,WAAW,iBAAiB,UAAU,GAAG;AACtE,aAAO,iBAAiB,IAAI,MAAM,iBAAiB,WAAW,MAAM,CAAC;AAAA,IACvE,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AACF;AACA,iBAAiB,mBAAmB,eAAe;AACnD,iBAAiB,mBAAmB,eAAe;AACnD,SAAS,iBAAiB,WAAW;AACnC,SAAO,IAAI,iBAAiB,SAAS;AACvC;AACA,SAAS,iBAAiB,KAAK;AAC7B,SAAO,IAAI,WAAW,iBAAiB,UAAU,IAAI,IAAI,MAAM,iBAAiB,WAAW,MAAM,IAAI;AACvG;AACA,IAAI,0BAA0B,MAAM;AAAA,EAClC,cAAc;AACZ,SAAK,YAAY,oBAAoB;AAAA,EACvC;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,cAAc,KAAK,UAAU,KAAK,eAAe,gBAAgB;AACvE,kBAAY,kBAAkB,MAAM,cAAc,WAAW;AAC7D,kBAAY,YAAY,MAAM;AAC5B,cAAM,KAAK,YAAY;AACvB,cAAM,KAAK,GAAG,YAAY,iBAAiB,UAAU;AACrD,cAAM,QAAQ,GAAG,YAAY,eAAe;AAC5C,cAAM,oBAAoB,MAAM,OAAO;AACvC,0BAAkB,YAAY,MAAM;AAClC,gBAAM,MAAM,CAAC;AACb,qBAAW,QAAQ,kBAAkB,QAAQ;AAC3C,gBAAI,KAAK,aAAa,KAAK;AAAA,UAC7B;AACA,kBAAQ,GAAG;AAAA,QACb;AACA,0BAAkB,UAAU,CAAC,UAAU;AACrC,aAAG,MAAM;AACT,iBAAO,OAAO,kBAAkB,KAAK;AAAA,QACvC;AACA,WAAG,aAAa,MAAM,GAAG,MAAM;AAAA,MACjC;AACA,kBAAY,UAAU,CAAC,UAAU,OAAO,YAAY,KAAK;AAAA,IAC3D,CAAC;AAAA,EACH;AAAA,EACA,MAAM,YAAY,MAAM;AACtB,WAAO,iBAAiB,IAAI;AAC5B,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,cAAc,KAAK,UAAU,KAAK,eAAe,gBAAgB;AACvE,kBAAY,kBAAkB,MAAM,cAAc,WAAW;AAC7D,kBAAY,YAAY,MAAM;AAC5B,cAAM,KAAK,YAAY;AACvB,cAAM,SAAS,GAAG,YAAY,iBAAiB,WAAW;AAC1D,cAAM,YAAY,OAAO,YAAY,eAAe;AACpD,cAAM,iBAAiB,UAAU,IAAI,IAAI;AACzC,YAAI;AACJ,uBAAe,YAAY,MAAM;AAC/B,cAAI,eAAe,UAAU,MAAM;AACjC,eAAG,MAAM;AACT,mBAAO,OAAO,IAAI,MAAM,gCAAgC,qBAAqB,CAAC;AAAA,UAChF,OAAO;AACL,kBAAM,oBAAoB,UAAU,OAAO,IAAI;AAC/C,kBAAM,kBAAkB,MAAM;AAC5B,wBAAU,GAAG,YAAY,kBAAkB,WAAW;AACtD,oBAAM,aAAa,QAAQ,YAAY,gBAAgB;AACvD,oBAAM,qBAAqB,WAAW,OAAO,IAAI;AACjD,iCAAmB,YAAY,MAAM,QAAQ,eAAe,OAAO,kBAAkB;AACrF,iCAAmB,UAAU,CAAC,UAAU,OAAO,eAAe,KAAK;AAAA,YACrE;AACA,8BAAkB,YAAY;AAC9B,8BAAkB,UAAU,CAAC,UAAU;AACrC,8BAAgB;AAChB,iBAAG,MAAM;AACT,qBAAO,OAAO,eAAe,KAAK;AAAA,YACpC;AAAA,UACF;AAAA,QACF;AACA,uBAAe,UAAU,CAAC,UAAU;AAClC,aAAG,MAAM;AACT,iBAAO,OAAO,eAAe,KAAK;AAAA,QACpC;AACA,eAAO,aAAa,MAAM;AACxB,cAAI,WAAW,MAAM;AACnB,eAAG,MAAM;AAAA,UACX,OAAO;AACL,oBAAQ,aAAa,MAAM,GAAG,MAAM;AAAA,UACtC;AAAA,QACF;AAAA,MACF;AACA,kBAAY,UAAU,CAAC,UAAU,OAAO,YAAY,KAAK;AAAA,IAC3D,CAAC;AAAA,EACH;AACF;AAGA,IAAI,iBAAiB;AACrB,IAAI,cAAc;AAClB,IAAI,cAAc;AAClB,IAAI,wBAAwB;AAC5B,IAAI,sBAAsB;AAC1B,IAAI,qBAAqB;AACzB,IAAI,wBAAwB;AAC5B,SAAS,aAAa,MAAM;AAC1B,SAAO;AAAA,IACL,MAAM,CAAC,aAAa,MAAM,WAAW,EAAE,KAAK,cAAc;AAAA,IAC1D,UAAU,CAAC,aAAa,MAAM,qBAAqB,EAAE,KAAK,cAAc;AAAA,IACxE,aAAa,CAAC,aAAa,MAAM,mBAAmB,EAAE,KAAK,cAAc;AAAA,IACzE,YAAY,CAAC,aAAa,MAAM,kBAAkB,EAAE,KAAK,cAAc;AAAA,IACvE,eAAe,CAAC,aAAa,MAAM,qBAAqB,EAAE,KAAK,cAAc;AAAA,EAC/E;AACF;AACA,SAAS,YAAY,MAAM;AACzB,aAAW,OAAO,OAAO,OAAO,IAAI,GAAG;AACrC,WAAO,aAAa,WAAW,GAAG;AAAA,EACpC;AACF;AACA,SAAS,oBAAoB,KAAK;AAChC,QAAM,QAAQ,IAAI,MAAM,cAAc;AACtC,MAAI,MAAM,SAAS,GAAG;AACpB,UAAM,IAAI,MAAM,uBAAuB,KAAK;AAAA,EAC9C;AACA,SAAO,MAAM,MAAM,GAAG,MAAM,SAAS,CAAC,EAAE,KAAK,cAAc;AAC7D;AACA,SAAS,kBAAkB,KAAK;AAC9B,SAAO,IAAI,WAAW,oBAAoB,UAAU,IAAI,IAAI,MAAM,oBAAoB,WAAW,MAAM,IAAI;AAC7G;AACA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,WAAW;AACrB,QAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,KAAK,OAAO,WAAW,eAAe,OAAO,OAAO,iBAAiB,aAAa;AAC/G,YAAM,IAAI,MAAM,yDAAyD;AAAA,IAC3E;AACA,SAAK,KAAK,OAAO;AACjB,QAAI,aAAa,QAAQ,CAAC,WAAW;AACnC,YAAM,IAAI,MAAM,oEAAoE;AAAA,IACtF;AACA,SAAK,YAAY;AACjB,SAAK,OAAO,aAAa,KAAK,SAAS;AAAA,EACzC;AAAA,EACA,MAAM,KAAK,gBAAgB;AACzB,QAAI,eAAe,yBAAyB,aAAa;AACvD,YAAM,IAAI,MAAM,0FAA0F;AAAA,IAC5G,OAAO;AACL,YAAM,WAAW,KAAK,UAAU,eAAe,aAAa;AAC5D,YAAM,cAAc,KAAK,UAAU,eAAe,WAAW;AAC7D,YAAM,qBAAqB,6BAA6B,cAAc;AACtE,UAAI;AACF,aAAK,GAAG,QAAQ,KAAK,KAAK,MAAM,KAAK,UAAU,kBAAkB,CAAC;AAClE,aAAK,GAAG,QAAQ,KAAK,KAAK,UAAU,QAAQ;AAC5C,aAAK,GAAG,QAAQ,KAAK,KAAK,aAAa,WAAW;AAClD,aAAK,GAAG,QAAQ,KAAK,KAAK,YAAY,0BAA0B,eAAe,UAAU,CAAC;AAC1F,cAAM,WAAW;AAAA,UACf,QAAQ,eAAe;AAAA,UACvB,aAAa,eAAe;AAAA,UAC5B,aAAa,eAAe;AAAA,UAC5B,WAAW,eAAe,aAAa,OAAO,eAAe,YAAY;AAAA,UACzE,qBAAqB,eAAe,uBAAuB,OAAO,eAAe,sBAAsB;AAAA,UACvG,kBAAkB,eAAe,oBAAoB,OAAO,eAAe,mBAAmB;AAAA,UAC9F,sBAAsB,eAAe,wBAAwB,OAAO,eAAe,uBAAuB;AAAA,UAC1G,gBAAgB,eAAe,kBAAkB,OAAO,eAAe,iBAAiB;AAAA,QAC1F;AACA,aAAK,GAAG,QAAQ,KAAK,KAAK,eAAe,KAAK,UAAU,QAAQ,CAAC;AACjE,eAAO,EAAE,mBAAmB;AAAA,MAC9B,SAAS,KAAP;AACA,oBAAY,KAAK,IAAI;AACrB,cAAM,IAAI,MAAM,yBAAyB,KAAK,kHAAkH,mBAAmB,wCAAwC,mBAAmB,qCAAqC,mBAAmB,kBAAkB;AAAA,MAC1T;AAAA,IACF;AAAA,EACF;AAAA,EACA,MAAM,OAAO;AACX,UAAM,OAAO,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,IAAI,CAAC;AACvD,QAAI,QAAQ,MAAM;AAChB,YAAM,IAAI,MAAM,kDAAkD,KAAK,YAAY;AAAA,IACrF;AACA,QAAI,KAAK,sBAAsB,QAAQ;AACrC,YAAM,IAAI,MAAM,2EAA2E;AAAA,IAC7F;AACA,UAAM,MAAM,CAAC;AACb,UAAM,WAAW,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,QAAQ,CAAC;AAC/D,QAAI,YAAY,MAAM;AACpB,YAAM,IAAI,MAAM,4CAA4C,KAAK,wBAAwB;AAAA,IAC3F;AACA,QAAI,gBAAgB;AACpB,UAAM,cAAc,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,KAAK,WAAW,CAAC;AACrE,QAAI,eAAe,MAAM;AACvB,YAAM,IAAI,MAAM,gDAAgD,KAAK,yBAAyB;AAAA,IAChG;AACA,QAAI,cAAc;AAClB,UAAM,iBAAiB,KAAK,GAAG,QAAQ,KAAK,KAAK,aAAa;AAC9D,QAAI,kBAAkB,MAAM;AAC1B,YAAM,WAAW,KAAK,MAAM,cAAc;AAC1C,UAAI,SAAS,SAAS;AACtB,UAAI,cAAc,SAAS;AAC3B,UAAI,cAAc,SAAS;AAC3B,UAAI,SAAS,aAAa,MAAM;AAC9B,YAAI,YAAY,SAAS;AAAA,MAC3B;AACA,UAAI,SAAS,uBAAuB,MAAM;AACxC,YAAI,sBAAsB,SAAS;AAAA,MACrC;AACA,UAAI,SAAS,oBAAoB,MAAM;AACrC,YAAI,mBAAmB,SAAS;AAAA,MAClC;AACA,UAAI,SAAS,wBAAwB,MAAM;AACzC,YAAI,uBAAuB,SAAS;AAAA,MACtC;AACA,UAAI,SAAS,kBAAkB,MAAM;AACnC,YAAI,iBAAiB,SAAS;AAAA,MAChC;AAAA,IACF;AACA,UAAM,mBAAmB,KAAK,GAAG,QAAQ,KAAK,KAAK,UAAU;AAC7D,QAAI,oBAAoB,MAAM;AAC5B,YAAM,IAAI,MAAM,wDAAwD,KAAK,yBAAyB;AAAA,IACxG;AACA,QAAI,aAAa,0BAA0B,gBAAgB;AAC3D,WAAO;AAAA,EACT;AACF;AACA,oBAAoB,aAAa;AACjC,IAAI,qBAAqB,CAAC,QAAQ;AAChC,MAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,GAAG;AAChC,WAAO;AAAA,EACT,OAAO;AACL,QAAI,CAAC,MAAM,QAAQ,GAAG,KAAK,IAAI,WAAW,oBAAoB,UAAU,GAAG;AACzE,aAAO,oBAAoB,IAAI,MAAM,oBAAoB,WAAW,MAAM,CAAC;AAAA,IAC7E,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AACF;AACA,iBAAiB,mBAAmB,kBAAkB;AACtD,iBAAiB,mBAAmB,kBAAkB;AACtD,SAAS,oBAAoB,WAAW;AACtC,SAAO,IAAI,oBAAoB,SAAS;AAC1C;AACA,IAAI,6BAA6B,MAAM;AAAA,EACrC,cAAc;AACZ,WAAO,IAAI,EAAE,QAAQ,YAAY,GAAG,MAAM,0CAA0C;AACpF,WAAO,OAAO,WAAW,eAAe,OAAO,OAAO,iBAAiB,aAAa,MAAM,yDAAyD;AACnJ,SAAK,KAAK,OAAO;AAAA,EACnB;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,MAAM,CAAC;AACb,UAAM,SAAS,cAAc;AAC7B,UAAM,SAAS,iBAAiB;AAChC,aAAS,IAAI,GAAG,IAAI,KAAK,GAAG,QAAQ,EAAE,GAAG;AACvC,YAAM,MAAM,KAAK,GAAG,IAAI,CAAC;AACzB,UAAI,IAAI,WAAW,MAAM,KAAK,IAAI,SAAS,MAAM,GAAG;AAClD,cAAM,YAAY,oBAAoB,GAAG;AACzC,YAAI,aAAa,KAAK,MAAM,KAAK,GAAG,QAAQ,GAAG,CAAC;AAAA,MAClD;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,YAAY,MAAM;AACtB,WAAO,kBAAkB,IAAI;AAC7B,UAAM,OAAO,aAAa,IAAI;AAC9B,QAAI,KAAK,GAAG,QAAQ,KAAK,IAAI,KAAK,MAAM;AACtC,YAAM,IAAI,MAAM,8BAA8B,OAAO;AAAA,IACvD;AACA,UAAM,OAAO,KAAK,MAAM,KAAK,GAAG,QAAQ,KAAK,IAAI,CAAC;AAClD,gBAAY,IAAI;AAChB,WAAO;AAAA,EACT;AACF;AAGA,IAAI,oBAAoB;AACxB,IAAI,4BAA4B,MAAM;AAAA,EACpC,cAAc;AACZ,SAAK,WAAW,CAAC;AAAA,EACnB;AAAA,EACA,OAAO,cAAc;AACnB,QAAI,0BAA0B,YAAY,MAAM;AAC9C,gCAA0B,WAAW,IAAI,0BAA0B;AAAA,IACrE;AACA,WAAO,0BAA0B;AAAA,EACnC;AAAA,EACA,OAAO,gBAAgB,QAAQ,SAAS;AACtC,WAAO,UAAU,MAAM,MAAM,uCAAuC;AACpE,QAAI,OAAO,SAAS,iBAAiB,GAAG;AACtC,eAAS,OAAO,MAAM,GAAG,OAAO,QAAQ,iBAAiB,CAAC;AAAA,IAC5D;AACA,WAAO,OAAO,SAAS,GAAG,MAAM,qCAAqC;AACrE,UAAM,WAAW,0BAA0B,YAAY;AACvD,WAAO,SAAS,SAAS,WAAW,MAAM,MAAM,2DAA2D,UAAU;AACrH,aAAS,SAAS,UAAU;AAAA,EAC9B;AAAA,EACA,OAAO,WAAW,QAAQ;AACxB,UAAM,UAAU,0BAA0B,YAAY,EAAE,SAAS;AACjE,QAAI,WAAW,MAAM;AACnB,YAAM,IAAI,MAAM,yCAAyC,SAAS;AAAA,IACpE;AACA,WAAO;AAAA,EACT;AAAA,EACA,OAAO,aAAa;AAClB,WAAO,OAAO,KAAK,0BAA0B,YAAY,EAAE,QAAQ;AAAA,EACrE;AACF;AACA,SAAS,SAAS,KAAK;AACrB,MAAI,IAAI,QAAQ,iBAAiB,MAAM,IAAI;AACzC,UAAM,IAAI,MAAM,6EAA6E,0BAA0B,WAAW,EAAE,KAAK,GAAG,GAAG;AAAA,EACjJ;AACA,SAAO;AAAA,IACL,QAAQ,IAAI,MAAM,iBAAiB,EAAE;AAAA,IACrC,MAAM,IAAI,MAAM,iBAAiB,EAAE;AAAA,EACrC;AACF;AACA,eAAe,mBAAmB,WAAW,SAAS,eAAe,OAAO;AAC1E,SAAO,cAAc,SAAS,MAAM,wCAAwC,YAAY;AACxF,QAAM,eAAe,iBAAiB,gBAAgB,SAAS;AAC/D,SAAO,aAAa,SAAS,GAAG,MAAM,kEAAkE,YAAY;AACpH,SAAO,aAAa,SAAS,GAAG,MAAM,yCAAyC,aAAa,wCAAwC,YAAY;AAChJ,QAAM,cAAc,aAAa;AACjC,QAAM,eAAe,iBAAiB,gBAAgB,OAAO;AAC7D,SAAO,aAAa,SAAS,GAAG,MAAM,uEAAuE,UAAU;AACvH,SAAO,aAAa,SAAS,GAAG,MAAM,yCAAyC,aAAa,6CAA6C,UAAU;AACnJ,QAAM,cAAc,aAAa;AACjC,QAAM,eAAe,SAAS,SAAS,EAAE;AACzC,QAAM,aAAa,SAAS,SAAS,EAAE;AACvC,QAAM,aAAa,iBAAiB,SAAS,SAAS,EAAE;AACxD,QAAM,iBAAiB,MAAM,YAAY,KAAK;AAC9C,MAAI,gBAAgB,YAAY;AAC9B,UAAM,0BAA0B,WAAW,YAAY,EAAE,YAAY,UAAU;AAAA,EACjF;AACA,QAAM,aAAa,MAAM,YAAY,KAAK,cAAc;AACxD,MAAI,gBAAgB,CAAC,YAAY;AAC/B,UAAM,0BAA0B,WAAW,YAAY,EAAE,YAAY,UAAU;AAAA,EACjF;AACA,SAAO,WAAW;AACpB;AACA,eAAe,aAAa;AAC1B,QAAM,UAAU,0BAA0B,WAAW;AACrD,QAAM,MAAM,CAAC;AACb,aAAW,UAAU,SAAS;AAC5B,UAAM,YAAY,MAAM,0BAA0B,WAAW,MAAM,EAAE,WAAW;AAChF,eAAW,QAAQ,WAAW;AAC5B,YAAM,MAAM,SAAS,oBAAoB;AACzC,UAAI,OAAO,UAAU;AAAA,IACvB;AAAA,EACF;AACA,SAAO;AACT;AACA,eAAe,YAAY,KAAK;AAC9B,QAAM,gBAAgB,SAAS,GAAG;AAClC,QAAM,UAAU,0BAA0B,WAAW,cAAc,MAAM;AACzE,SAAO,QAAQ,YAAY,cAAc,IAAI;AAC/C;AACA,eAAe,UAAU,WAAW,SAAS;AAC3C,QAAM,eAAe;AACrB,SAAO,mBAAmB,WAAW,SAAS,YAAY;AAC5D;AACA,eAAe,UAAU,WAAW,SAAS;AAC3C,QAAM,eAAe;AACrB,SAAO,mBAAmB,WAAW,SAAS,YAAY;AAC5D;AAGA,IAAI,kBAAkB,MAAM;AAAA,EAC1B,cAAc;AACZ,SAAK,cAAc;AACnB,SAAK,eAAe,CAAC;AACrB,SAAK,sBAAsB;AAC3B,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EACA,MAAM,MAAM,OAAO;AACjB,WAAO,MAAM,MAAM,KAAK;AAAA,EAC1B;AAAA,EACA,MAAM;AACJ,WAAO,YAAY,IAAI;AAAA,EACzB;AAAA,EACA,OAAO,MAAM,UAAU;AACrB,QAAI,aAAa,WAAW,aAAa,QAAQ;AAC/C,YAAM,IAAI,MAAM,kDAAkD,UAAU;AAAA,IAC9E;AACA,QAAI,KAAK,eAAe,MAAM;AAC5B,WAAK,cAAc,IAAI,YAAY;AAAA,IACrC;AACA,WAAO,KAAK,YAAY,OAAO,IAAI;AAAA,EACrC;AAAA,EACA,OAAO,OAAO,UAAU;AACtB,WAAO,IAAI,YAAY,QAAQ,EAAE,OAAO,KAAK;AAAA,EAC/C;AAAA,EACA,iBAAiB,aAAa,OAAO;AACnC,QAAI,OAAO,WAAW,eAAe,CAAC,IAAI,EAAE,QAAQ,sBAAsB,GAAG;AAC3E,iBAAW,aAAa,KAAK;AAC7B;AAAA,IACF;AACA,SAAK,aAAa,KAAK,WAAW;AAClC,eAAW,MAAM;AACf,aAAO,YAAY,EAAE,MAAM,KAAK,aAAa,OAAO,KAAK,aAAa,SAAS,EAAE,GAAG,GAAG;AAAA,IACzF,GAAG,KAAK;AACR,QAAI,CAAC,KAAK,kBAAkB;AAC1B,WAAK,mBAAmB;AACxB,aAAO,iBAAiB,WAAW,CAAC,UAAU;AAC5C,YAAI,MAAM,WAAW,UAAU,MAAM,KAAK,SAAS,KAAK,aAAa;AACnE,gBAAM,gBAAgB;AACtB,gBAAM,eAAe,KAAK,aAAa,MAAM,KAAK;AAClD,uBAAa;AACb,eAAK;AACL,cAAI,KAAK,wBAAwB,KAAK,aAAa,QAAQ;AACzD,iBAAK,eAAe,CAAC;AACrB,iBAAK,sBAAsB;AAAA,UAC7B;AAAA,QACF;AAAA,MACF,GAAG,IAAI;AAAA,IACT;AAAA,EACF;AACF;AACA,IAAI,IAAI,EAAE,IAAI,YAAY,GAAG;AAC3B,MAAI,EAAE,YAAY,WAAW,IAAI,gBAAgB,CAAC;AAClD,MAAI;AACF,8BAA0B,gBAAgB,oBAAoB,YAAY,IAAI,2BAA2B,CAAC;AAAA,EAC5G,SAAS,KAAP;AAAA,EACF;AACA,MAAI;AACF,8BAA0B,gBAAgB,iBAAiB,YAAY,IAAI,wBAAwB,CAAC;AAAA,EACtG,SAAS,KAAP;AAAA,EACF;AACF;AAGA,IAAI,eAAe;AAAA,EACjB,aAAa,MAAM,gBAAgB;AACrC;AACA,IAAI;AACJ,IAAI,eAAe,MAAM;AAAA,EACvB,cAAc;AACZ,SAAK,OAAO,aAAa;AACzB,SAAK,cAAc,IAAI,KAAK,KAAK,YAAY;AAAA,EAC/C;AAAA,EACA,MAAM,MAAM,cAAc;AACxB,QAAI,IAAI,EAAE,OAAO,SAAS,MAAM;AAC9B,aAAO,IAAI,EAAE,OAAO,MAAM,MAAM,YAAY;AAAA,IAC9C;AACA,QAAI,eAAe,MAAM;AACvB,oBAAc,aAAa,YAAY;AAAA,IACzC;AACA,WAAO,YAAY,MAAM,YAAY;AAAA,EACvC;AAAA,EACA,MAAM;AACJ,UAAM,QAAQ,QAAQ,OAAO;AAC7B,WAAO,MAAM,KAAK,MAAM,MAAM,KAAK;AAAA,EACrC;AAAA,EACA,OAAO,MAAM,UAAU;AACrB,QAAI,aAAa,WAAW,aAAa,QAAQ;AAC/C,YAAM,IAAI,MAAM,sDAAsD,UAAU;AAAA,IAClF;AACA,WAAO,KAAK,YAAY,OAAO,IAAI;AAAA,EACrC;AAAA,EACA,OAAO,OAAO,UAAU;AACtB,QAAI,MAAM,WAAW,GAAG;AACtB,aAAO;AAAA,IACT;AACA,WAAO,IAAI,KAAK,KAAK,YAAY,QAAQ,EAAE,OAAO,KAAK;AAAA,EACzD;AACF;AACA,IAAI,IAAI,EAAE,IAAI,SAAS,KAAK,CAAC,IAAI,EAAE,IAAI,YAAY,GAAG;AACpD,MAAI,EAAE,YAAY,QAAQ,IAAI,aAAa,CAAC;AAC9C;AAGA,SAAS,OAAO,OAAO,QAAQ,WAAW,QAAQ;AAChD,UAAQ,SAAS;AACjB,qCAAmC,KAAK;AACxC,SAAO,IAAI,aAAa,OAAO,OAAO,MAAM;AAC9C;AAGA,SAAS,MAAM,GAAG,OAAO;AACvB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,MAAI,CAAC,aAAa,KAAK,GAAG;AACxB,UAAM,IAAI,MAAM,mCAAmC,OAAO;AAAA,EAC5D;AACA,MAAI,UAAU,YAAY,GAAG,UAAU,YAAY,UAAU,YAAY,GAAG,UAAU,UAAU;AAC9F,UAAM,IAAI,MAAM,uCAAuC;AAAA,EACzD;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM;AACtB,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS,mBAAmB;AAC/D,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG,UAAU,OAAO;AACjC,UAAQ,IAAI,EAAE,SAAS,OAAO,CAAC;AACjC;AAGA,gBAAgB;AAChB,IAAI,aAAa;AAAA,EACf;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,aAAa,UAAU;AAGvB,IAAI,aAAa,CAAC;AAClBA,UAAS,YAAY;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,oBAAoB,MAAM;AAAA,EAC1B,yBAAyB,MAAM;AAAA,EAC/B,WAAW,MAAM;AAAA,EACjB,eAAe,MAAM;AAAA,EACrB,eAAe,MAAM;AAAA,EACrB,YAAY,MAAM;AAAA,EAClB,gBAAgB,MAAM;AAAA,EACtB,iBAAiB,MAAM;AAAA,EACvB,0BAA0B,MAAM;AAAA,EAChC,8BAA8B,MAAM;AAAA,EACpC,8BAA8B,MAAM;AAAA,EACpC,iBAAiB,MAAM;AAAA,EACvB,gBAAgB,MAAM;AAAA,EACtB,MAAM,MAAM;AAAA,EACZ,cAAc,MAAM;AAAA,EACpB,YAAY,MAAM;AAAA,EAClB,aAAa,MAAM;AAAA,EACnB,WAAW,MAAM;AAAA,EACjB,oBAAoB,MAAM;AAAA,EAC1B,oBAAoB,MAAM;AAAA,EAC1B,aAAa,MAAM;AAAA,EACnB,sBAAsB,MAAM;AAAA,EAC5B,iBAAiB,MAAM;AAAA,EACvB,qBAAqB,MAAM;AAC7B,CAAC;AAGD,IAAI,2BAA2B;AAC/B,IAAI,8BAA8B;AAClC,IAAI,qCAAqC;AACzC,SAAS,MAAM,GAAG;AAChB,SAAO,IAAI,QAAQ,CAAC,YAAY,WAAW,OAAO,CAAC,EAAE,KAAK,CAAC;AAC7D;AACA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,gBAAgB;AAC1B,QAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,GAAG;AAChC,YAAM,IAAI,MAAM,qFAAqF;AAAA,IACvG;AACA,QAAI,eAAe,WAAW,iBAAiB,UAAU,GAAG;AAC1D,uBAAiB,eAAe,MAAM,iBAAiB,WAAW,MAAM;AAAA,IAC1E;AACA,QAAI,kBAAkB,QAAQ,eAAe,WAAW,GAAG;AACzD,uBAAiB;AAAA,IACnB;AACA,SAAK,oBAAoB,iBAAiB;AAC1C,SAAK,qBAAqB,iBAAiB;AAAA,EAC7C;AAAA,EACA,MAAM,KAAK,gBAAgB;AACzB,QAAI,OAAO,aAAa,aAAa;AACnC,YAAM,IAAI,MAAM,yFAAyF;AAAA,IAC3G;AACA,UAAM,aAAa,OAAO,IAAI,gBAAgB,IAAI,KAAK,CAAC,eAAe,UAAU,GAAG,EAAE,MAAM,2BAA2B,CAAC,CAAC;AACzH,QAAI,eAAe,yBAAyB,aAAa;AACvD,YAAM,IAAI,MAAM,uFAAuF;AAAA,IACzG,OAAO;AACL,YAAM,kBAAkB,CAAC;AAAA,QACvB,OAAO,CAAC,OAAO,KAAK,kBAAkB;AAAA,QACtC,SAAS,eAAe;AAAA,MAC1B,CAAC;AACD,YAAM,YAAY,8BAA8B,gBAAgB,eAAe;AAC/E,YAAM,eAAe,OAAO,IAAI,gBAAgB,IAAI,KAAK,CAAC,KAAK,UAAU,SAAS,CAAC,GAAG,EAAE,MAAM,mBAAmB,CAAC,CAAC;AACnH,YAAM,aAAa,KAAK,mBAAmB,OAAO,SAAS,cAAc,GAAG,IAAI,KAAK;AACrF,iBAAW,WAAW,KAAK;AAC3B,iBAAW,OAAO;AAClB,YAAM,MAAM,MAAM,WAAW,cAAc,IAAI,WAAW,OAAO,CAAC,CAAC;AACnE,UAAI,eAAe,cAAc,MAAM;AACrC,cAAM,mBAAmB,KAAK,oBAAoB,OAAO,SAAS,cAAc,GAAG,IAAI,KAAK;AAC5F,yBAAiB,WAAW,KAAK;AACjC,yBAAiB,OAAO;AACxB,cAAM,MAAM,MAAM,iBAAiB,cAAc,IAAI,WAAW,OAAO,CAAC,CAAC;AAAA,MAC3E;AACA,aAAO,EAAE,oBAAoB,6BAA6B,cAAc,EAAE;AAAA,IAC5E;AAAA,EACF;AACF;AACA,iBAAiB,aAAa;AAC9B,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,OAAO;AACjB,QAAI,SAAS,QAAQ,MAAM,SAAS,GAAG;AACrC,YAAM,IAAI,MAAM,wEAAwE,OAAO;AAAA,IACjG;AACA,SAAK,WAAW,MAAM;AACtB,SAAK,eAAe,MAAM,MAAM,CAAC;AAAA,EACnC;AAAA,EACA,MAAM,OAAO;AACX,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,aAAa,IAAI,WAAW;AAClC,iBAAW,SAAS,CAAC,UAAU;AAC7B,cAAM,YAAY,KAAK,MAAM,MAAM,OAAO,MAAM;AAChD,cAAM,gBAAgB,UAAU;AAChC,YAAI,iBAAiB,MAAM;AACzB,iBAAO,IAAI,MAAM,4CAA4C,KAAK,SAAS,MAAM,CAAC;AAClF;AAAA,QACF;AACA,cAAM,kBAAkB,UAAU;AAClC,YAAI,mBAAmB,MAAM;AAC3B,iBAAO,IAAI,MAAM,6CAA6C,KAAK,SAAS,MAAM,CAAC;AACnF;AAAA,QACF;AACA,YAAI,KAAK,aAAa,WAAW,GAAG;AAClC,kBAAQ,EAAE,cAAc,CAAC;AACzB;AAAA,QACF;AACA,cAAM,wBAAwB,yBAAyB,WAAW,CAAC,qBAAqB,KAAK,YAAY,gBAAgB,CAAC;AAC1H,gBAAQ,qBAAqB;AAAA,MAC/B;AACA,iBAAW,UAAU,CAAC,UAAU,OAAO,sEAAsE,KAAK,SAAS,2EAA2E;AACtM,iBAAW,WAAW,KAAK,QAAQ;AAAA,IACrC,CAAC;AAAA,EACH;AAAA,EACA,YAAY,iBAAiB;AAC3B,UAAM,cAAc,CAAC;AACrB,UAAM,QAAQ,CAAC;AACf,eAAW,SAAS,iBAAiB;AACnC,kBAAY,KAAK,GAAG,MAAM,OAAO;AACjC,YAAM,KAAK,GAAG,MAAM,KAAK;AAAA,IAC3B;AACA,UAAM,aAAa,KAAK,4BAA4B,eAAe;AACnE,UAAM,WAAW,MAAM,IAAI,CAAC,SAAS,KAAK,gBAAgB,MAAM,WAAW,KAAK,CAAC;AACjF,WAAO,QAAQ,IAAI,QAAQ,EAAE,KAAK,CAAC,YAAY,CAAC,aAAa,wBAAwB,OAAO,CAAC,CAAC;AAAA,EAChG;AAAA,EACA,gBAAgB,MAAM,MAAM;AAC1B,WAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,YAAM,mBAAmB,IAAI,WAAW;AACxC,uBAAiB,SAAS,CAAC,UAAU;AACnC,cAAM,aAAa,MAAM,OAAO;AAChC,gBAAQ,UAAU;AAAA,MACpB;AACA,uBAAiB,UAAU,CAAC,UAAU,OAAO,6CAA6C,QAAQ;AAClG,uBAAiB,kBAAkB,IAAI;AAAA,IACzC,CAAC;AAAA,EACH;AAAA,EACA,4BAA4B,UAAU;AACpC,UAAM,YAAY,CAAC;AACnB,UAAM,YAAY,KAAK,aAAa,IAAI,CAAC,SAAS,SAAS,KAAK,IAAI,CAAC;AACrE,UAAM,aAAa,CAAC;AACpB,eAAW,SAAS,UAAU;AAC5B,YAAM,MAAM,QAAQ,CAAC,SAAS;AAC5B,cAAM,eAAe,SAAS,IAAI;AAClC,YAAI,UAAU,QAAQ,YAAY,MAAM,IAAI;AAC1C,gBAAM,IAAI,MAAM,uDAAuD,eAAe;AAAA,QACxF;AACA,kBAAU,KAAK,YAAY;AAC3B,YAAI,UAAU,QAAQ,YAAY,MAAM,IAAI;AAC1C,gBAAM,IAAI,MAAM,8BAA8B,gCAAgC;AAAA,QAChF,OAAO;AACL,qBAAW,QAAQ,KAAK,aAAa,UAAU,QAAQ,YAAY;AAAA,QACrE;AAAA,MACF,CAAC;AAAA,IACH;AACA,QAAI,UAAU,WAAW,KAAK,aAAa,QAAQ;AACjD,YAAM,IAAI,MAAM,wDAAwD,UAAU,oDAAoD,KAAK,aAAa,UAAU;AAAA,IACpK;AACA,WAAO;AAAA,EACT;AACF;AACA,IAAI,yBAAyB,CAAC,QAAQ;AACpC,MAAI,CAAC,IAAI,EAAE,QAAQ,YAAY,GAAG;AAChC,WAAO;AAAA,EACT,OAAO;AACL,QAAI,CAAC,MAAM,QAAQ,GAAG,KAAK,IAAI,WAAW,iBAAiB,UAAU,GAAG;AACtE,aAAO,iBAAiB,IAAI,MAAM,iBAAiB,WAAW,MAAM,CAAC;AAAA,IACvE,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AACF;AACA,iBAAiB,mBAAmB,sBAAsB;AAC1D,SAAS,iBAAiB,iBAAiB,SAAS;AAClD,SAAO,IAAI,iBAAiB,cAAc;AAC5C;AACA,SAAS,aAAa,OAAO;AAC3B,SAAO,IAAI,aAAa,KAAK;AAC/B;AAGA,SAAS,wBAAwB,UAAU,YAAY,eAAe,aAAa;AACjF,gBAAc,QAAQ;AACtB,kBAAgB,iBAAiB,OAAO,IAAI;AAC5C,gBAAc,eAAe,OAAO,IAAI;AACxC,gBAAc,eAAe,WAAW;AACxC,MAAI,kBAAkB;AACtB,QAAM,kBAAkB,CAAC,YAAY;AACnC,YAAQ,KAAK,CAAC,UAAU;AACtB,YAAM,WAAW,gBAAgB,EAAE,kBAAkB,SAAS,UAAU,cAAc;AACtF,iBAAW,QAAQ;AACnB,aAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AACA,WAAS,cAAc,WAAW;AAChC,WAAO,aAAa,QAAQ,MAAM,QAAQ,SAAS,KAAK,UAAU,SAAS,GAAG,MAAM,qCAAqC;AAAA,EAC3H;AACA,WAAS,cAAc,gBAAgB,cAAc;AACnD,WAAO,kBAAkB,KAAK,kBAAkB,GAAG,MAAM,oEAAoE,gBAAgB;AAC7I,WAAO,gBAAgB,KAAK,gBAAgB,GAAG,MAAM,kEAAkE,cAAc;AACrI,WAAO,gBAAgB,gBAAgB,MAAM,yEAAyE,kCAAkC,cAAc;AAAA,EACxK;AACA,SAAO,QAAQ,IAAI,SAAS,IAAI,eAAe,CAAC;AAClD;AAGA,eAAe,yBAAyB,WAAW,aAAa;AAC9D,MAAI,eAAe,MAAM;AACvB,kBAAc,CAAC;AAAA,EACjB;AACA,QAAM,YAAY,YAAY,aAAa,OAAO,IAAI,EAAE,SAAS,QAAQ,YAAY;AACrF,QAAM,WAAW,UAAU,IAAI,CAAC,aAAa,UAAU,UAAU,YAAY,aAAa,EAAE,UAAU,KAAK,CAAC,CAAC;AAC7G,QAAM,qBAAqB;AAC3B,QAAM,mBAAmB;AACzB,QAAM,YAAY,YAAY,cAAc,OAAO,MAAM,QAAQ,IAAI,QAAQ,IAAI,MAAM,wBAAwB,UAAU,YAAY,YAAY,oBAAoB,gBAAgB;AACrL,QAAM,iBAAiB,UAAU,IAAI,CAAC,aAAa,SAAS,YAAY,CAAC;AACzE,QAAM,sBAAsB;AAC5B,QAAM,oBAAoB;AAC1B,QAAM,UAAU,YAAY,cAAc,OAAO,MAAM,QAAQ,IAAI,cAAc,IAAI,MAAM,wBAAwB,gBAAgB,YAAY,YAAY,qBAAqB,iBAAiB;AACjM,SAAO;AACT;AACA,eAAe,YAAY,UAAU,iBAAiB,IAAI,aAAa,aAAa;AAClF,QAAM,eAAe,CAAC,cAAc,yBAAyB,WAAW,EAAE,YAAY,CAAC;AACvF,QAAM,eAAe,qBAAqB,YAAY;AACtD,SAAO,aAAa,UAAU,gBAAgB,WAAW;AAC3D;AACA,SAAS,qBAAqB,sBAAsB;AAClD,SAAO,OAAO,UAAU,iBAAiB,IAAI,gBAAgB;AAC3D,UAAM,yBAAyB,SAAS,IAAI,MAAM,KAAK;AACvD,UAAM,sBAAsB,CAAC;AAC7B,UAAM,eAAe,eAAe,OAAO,YAAY,IAAI,MAAM,KAAK,IAAI,CAAC;AAC3E,UAAM,yBAAyB,CAAC;AAChC,aAAS,QAAQ,CAAC,qBAAqB,eAAe;AACpD,UAAI,cAAc;AAClB,0BAAoB,QAAQ,QAAQ,CAAC,iBAAiB;AACpD,cAAM,WAAW,kBAAkB,eAAe,aAAa,aAAa,QAAQ,aAAa;AACjG,cAAM,eAAe,qBAAqB,YAAY,cAAc,aAAa,KAAK;AACtF,cAAM,8BAA8B,MAAM;AACxC,iCAAuB,cAAc;AACrC,cAAI,oBAAoB,eAAe,MAAM;AAC3C,gCAAoB,cAAc,CAAC;AAAA,UACrC;AACA,8BAAoB,YAAY,KAAK;AAAA,YACnC,eAAe;AAAA,YACf;AAAA,YACA,WAAW;AAAA,UACb,CAAC;AAAA,QACH;AACA,YAAI,eAAe,MAAM;AACvB,sBAAY,QAAQ,CAAC,YAAY,gBAAgB;AAC/C,gBAAI,eAAe,aAAa,MAAM;AACpC,0CAA4B;AAC5B,2BAAa,eAAe;AAAA,YAC9B;AAAA,UACF,CAAC;AAAA,QACH,OAAO;AACL,sCAA4B;AAAA,QAC9B;AACA,+BAAuB,KAAK,aAAa,IAAI;AAC7C,uBAAe;AAAA,MACjB,CAAC;AAAA,IACH,CAAC;AACD,QAAI,CAAC,aAAa,MAAM,CAAC,UAAU,KAAK,GAAG;AACzC,YAAM,kBAAkB,YAAY,OAAO,CAAC,GAAG,MAAM,CAAC,aAAa,EAAE;AACrE,YAAM,IAAI,MAAM,kDAAkD,gBAAgB,KAAK,IAAI;AAAA,wCACzD,uBAAuB,KAAK,IAAI,IAAI;AAAA,IACxE;AACA,UAAM,sBAAsB,uBAAuB,OAAO,CAAC,aAAa,aAAa,MAAM;AACzF,UAAI,aAAa;AACf,oBAAY,KAAK,CAAC;AAAA,MACpB;AACA,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AACL,UAAM,YAAY,CAAC;AACnB,wBAAoB,QAAQ,CAAC,MAAM;AACjC,eAAS,GAAG,MAAM,QAAQ,CAAC,aAAa;AACtC,cAAM,WAAW,kBAAkB,CAAC,eAAe,SAAS,GAAG,IAAI,MAAM,MAAM;AAC/E,kBAAU,KAAK,QAAQ;AAAA,MACzB,CAAC;AAAA,IACH,CAAC;AACD,UAAM,UAAU,MAAM,qBAAqB,SAAS;AACpD,UAAM,mBAAmB,CAAC;AAC1B,QAAI,oBAAoB;AACxB,wBAAoB,QAAQ,CAAC,MAAM;AACjC,YAAM,aAAa,SAAS,GAAG,MAAM;AACrC,UAAI,aAAa;AACjB,eAAS,KAAK,GAAG,KAAK,YAAY,MAAM;AACtC,sBAAc,QAAQ,oBAAoB,IAAI;AAAA,MAChD;AACA,YAAM,cAAc,IAAI,YAAY,UAAU;AAC9C,YAAM,kBAAkB,IAAI,WAAW,WAAW;AAClD,UAAI,oBAAoB;AACxB,eAAS,KAAK,GAAG,KAAK,YAAY,MAAM;AACtC,cAAM,UAAU,IAAI,WAAW,QAAQ,oBAAoB,GAAG;AAC9D,wBAAgB,IAAI,SAAS,iBAAiB;AAC9C,6BAAqB,QAAQ;AAAA,MAC/B;AACA,YAAM,iBAAiB,oBAAoB;AAC3C,qBAAe,QAAQ,CAAC,iBAAiB;AACvC,cAAM,aAAa,YAAY,MAAM,aAAa,aAAa,aAAa,cAAc,aAAa,SAAS;AAChH,cAAM,kBAAkB,cAAc,YAAY,CAAC,aAAa,aAAa,CAAC;AAC9E,mBAAW,QAAQ,iBAAiB;AAClC,2BAAiB,QAAQ,gBAAgB;AAAA,QAC3C;AAAA,MACF,CAAC;AACD,2BAAqB;AAAA,IACvB,CAAC;AACD,WAAO;AAAA,EACT;AACF;AAGA,IAAI,yBAAyB;AAC7B,IAAI,YAAY;AAChB,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,MAAM,aAAa;AAC7B,SAAK,iBAAiB;AACtB,QAAI,eAAe,MAAM;AACvB,oBAAc,CAAC;AAAA,IACjB;AACA,SAAK,mBAAmB,YAAY;AACpC,SAAK,aAAa,YAAY;AAC9B,SAAK,qBAAqB,YAAY;AACtC,QAAI,YAAY,aAAa,MAAM;AACjC,aAAO,OAAO,YAAY,cAAc,YAAY,MAAM,6HAA6H;AACvL,WAAK,QAAQ,YAAY;AAAA,IAC3B,OAAO;AACL,WAAK,QAAQ,IAAI,EAAE,SAAS;AAAA,IAC9B;AACA,WAAO,QAAQ,QAAQ,KAAK,SAAS,GAAG,MAAM,yDAAyD;AACvG,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,aAAO,KAAK,WAAW,GAAG,MAAM,iEAAiE,KAAK,UAAU;AAAA,IAClH;AACA,SAAK,OAAO;AACZ,QAAI,YAAY,eAAe,QAAQ,YAAY,YAAY,QAAQ,MAAM;AAC3E,YAAM,IAAI,MAAM,oEAAoE;AAAA,IACtF;AACA,SAAK,cAAc,YAAY,eAAe,CAAC;AAAA,EACjD;AAAA,EACA,MAAM,KAAK,gBAAgB;AACzB,QAAI,eAAe,yBAAyB,aAAa;AACvD,YAAM,IAAI,MAAM,yFAAyF;AAAA,IAC3G;AACA,UAAM,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,eAAe,GAAG,KAAK,WAAW;AAC7E,UAAM,OAAO,IAAI,SAAS;AAC1B,UAAM,kBAAkB,CAAC;AAAA,MACvB,OAAO,CAAC,qBAAqB;AAAA,MAC7B,SAAS,eAAe;AAAA,IAC1B,CAAC;AACD,UAAM,iCAAiC,8BAA8B,gBAAgB,eAAe;AACpG,UAAM,KAAK,OAAO,cAAc,IAAI,KAAK,CAAC,KAAK,UAAU,8BAA8B,CAAC,GAAG,EAAE,MAAM,UAAU,CAAC,GAAG,YAAY;AAC7H,QAAI,eAAe,cAAc,MAAM;AACrC,YAAM,KAAK,OAAO,qBAAqB,IAAI,KAAK,CAAC,eAAe,UAAU,GAAG,EAAE,MAAM,uBAAuB,CAAC,GAAG,mBAAmB;AAAA,IACrI;AACA,UAAM,WAAW,MAAM,KAAK,MAAM,KAAK,MAAM,KAAK;AAClD,QAAI,SAAS,IAAI;AACf,aAAO;AAAA,QACL,oBAAoB,6BAA6B,cAAc;AAAA,QAC/D,WAAW,CAAC,QAAQ;AAAA,MACtB;AAAA,IACF,OAAO;AACL,YAAM,IAAI,MAAM,gEAAgE,SAAS,SAAS;AAAA,IACpG;AAAA,EACF;AAAA,EACA,MAAM,OAAO;AACX,UAAM,qBAAqB,MAAM,KAAK,MAAM,KAAK,MAAM,KAAK,WAAW;AACvE,QAAI,CAAC,mBAAmB,IAAI;AAC1B,YAAM,IAAI,MAAM,cAAc,KAAK,gCAAgC,mBAAmB,+EAA+E;AAAA,IACvK;AACA,QAAI;AACJ,QAAI;AACF,kBAAY,MAAM,mBAAmB,KAAK;AAAA,IAC5C,SAAS,GAAP;AACA,UAAI,UAAU,+CAA+C,KAAK;AAClE,UAAI,KAAK,KAAK,SAAS,KAAK,GAAG;AAC7B,mBAAW;AAAA,MACb,OAAO;AACL,mBAAW;AAAA,MACb;AACA,YAAM,IAAI,MAAM,OAAO;AAAA,IACzB;AACA,UAAM,gBAAgB,UAAU;AAChC,UAAM,kBAAkB,UAAU;AAClC,QAAI,iBAAiB,QAAQ,mBAAmB,MAAM;AACpD,YAAM,IAAI,MAAM,2BAA2B,KAAK,+DAA+D;AAAA,IACjH;AACA,WAAO,yBAAyB,WAAW,CAAC,qBAAqB,KAAK,YAAY,gBAAgB,CAAC;AAAA,EACrG;AAAA,EACA,MAAM,YAAY,iBAAiB;AACjC,UAAM,aAAa,MAAM,QAAQ,KAAK,IAAI,IAAI,KAAK,KAAK,KAAK,KAAK;AAClE,UAAM,CAAC,QAAQ,MAAM,IAAI,SAAS,UAAU;AAC5C,UAAM,aAAa,KAAK,oBAAoB;AAC5C,UAAM,cAAc,eAAe,eAAe;AAClD,UAAM,YAAY,CAAC;AACnB,UAAM,cAAc,CAAC;AACrB,eAAW,gBAAgB,iBAAiB;AAC1C,iBAAW,QAAQ,aAAa,OAAO;AACrC,YAAI,KAAK,sBAAsB,MAAM;AACnC,sBAAY,KAAK,KAAK,mBAAmB,IAAI,CAAC;AAAA,QAChD,OAAO;AACL,oBAAU,KAAK,aAAa,OAAO,MAAM;AAAA,QAC3C;AAAA,MACF;AAAA,IACF;AACA,QAAI,KAAK,oBAAoB;AAC3B,gBAAU,KAAK,GAAG,MAAM,QAAQ,IAAI,WAAW,CAAC;AAAA,IAClD;AACA,UAAM,UAAU,MAAM,yBAAyB,WAAW;AAAA,MACxD,aAAa,KAAK;AAAA,MAClB,WAAW,KAAK;AAAA,MAChB,YAAY,KAAK;AAAA,IACnB,CAAC;AACD,WAAO,CAAC,aAAa,wBAAwB,OAAO,CAAC;AAAA,EACvD;AACF;AACA,YAAY,mBAAmB;AAC/B,SAAS,SAAS,KAAK;AACrB,QAAM,YAAY,IAAI,YAAY,GAAG;AACrC,QAAM,kBAAkB,IAAI,YAAY,GAAG;AAC3C,QAAM,SAAS,IAAI,UAAU,GAAG,SAAS;AACzC,QAAM,SAAS,kBAAkB,YAAY,IAAI,UAAU,eAAe,IAAI;AAC9E,SAAO,CAAC,SAAS,KAAK,MAAM;AAC9B;AACA,SAAS,aAAa,KAAK;AACzB,SAAO,IAAI,MAAM,YAAY,gBAAgB,KAAK;AACpD;AACA,IAAI,aAAa,CAAC,KAAK,gBAAgB;AACrC,MAAI,OAAO,UAAU,gBAAgB,eAAe,QAAQ,YAAY,aAAa,OAAO;AAC1F,WAAO;AAAA,EACT,OAAO;AACL,QAAI,SAAS;AACb,QAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,eAAS,IAAI,MAAM,CAAC,YAAY,aAAa,OAAO,CAAC;AAAA,IACvD,OAAO;AACL,eAAS,aAAa,GAAG;AAAA,IAC3B;AACA,QAAI,QAAQ;AACV,aAAO,KAAK,KAAK,WAAW;AAAA,IAC9B;AAAA,EACF;AACA,SAAO;AACT;AACA,iBAAiB,mBAAmB,UAAU;AAC9C,iBAAiB,mBAAmB,UAAU;AAC9C,SAAS,KAAK,MAAM,aAAa;AAC/B,SAAO,IAAI,YAAY,MAAM,WAAW;AAC1C;AACA,SAAS,mBAAmB,MAAM,aAAa;AAC7C,SAAO,KAAK,MAAM,WAAW;AAC/B;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,gBAAgB;AAC1B,SAAK,iBAAiB;AAAA,EACxB;AAAA,EACA,OAAO;AACL,WAAO,KAAK;AAAA,EACd;AACF;AACA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,aAAa;AACvB,SAAK,cAAc;AAAA,EACrB;AAAA,EACA,KAAK,gBAAgB;AACnB,WAAO,KAAK,YAAY,cAAc;AAAA,EACxC;AACF;AACA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,SAAS;AACnB,QAAI,QAAQ,MAAM;AAChB,WAAK,OAAO,MAAM,QAAQ,QAAQ,QAAQ,KAAK,CAAC;AAAA,IAClD;AACA,QAAI,QAAQ,MAAM;AAChB,WAAK,OAAO,CAAC,mBAAmB,QAAQ,QAAQ,QAAQ,KAAK,cAAc,CAAC;AAAA,IAC9E;AAAA,EACF;AACF;AACA,SAAS,WAAW,gBAAgB,aAAa,YAAY,gBAAgB;AAC3E,QAAM,OAAO;AACb,SAAO,IAAI,iBAAiB,eAAe,GAAG,IAAI,CAAC;AACrD;AACA,SAAS,eAAe,gBAAgB,aAAa,YAAY,gBAAgB;AAC/E,MAAI,UAAU,WAAW,GAAG;AAC1B,UAAM,mBAAmB,eAAe,iBAAiB,QAAQ,eAAe,eAAe;AAC/F,QAAI,kBAAkB;AACpB,aAAO,IAAI,kBAAkB,cAAc;AAAA,IAC7C,OAAO;AACL,cAAQ,KAAK,uNAAuN;AACpO,aAAO,IAAI,kBAAkB,EAAE,eAAe,eAAe,CAAC;AAAA,IAChE;AAAA,EACF,OAAO;AACL,YAAQ,KAAK,uNAAuN;AACpO,WAAO,IAAI,kBAAkB;AAAA,MAC3B,eAAe;AAAA,MACf;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AACF;AACA,SAAS,gBAAgB,aAAa;AACpC,SAAO,IAAI,iBAAiB,WAAW;AACzC;AACA,SAAS,oBAAoB,aAAa;AACxC,SAAO,IAAI,iBAAiB,WAAW;AACzC;AAGA,IAAI,eAAe,CAAC;AACpBA,UAAS,cAAc;AAAA,EACrB,iBAAiB,MAAM;AACzB,CAAC;AAGD,SAAS,QAAQ,GAAG,GAAG,aAAa,OAAO,aAAa,OAAO;AAC7D,MAAI,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AACzC,MAAI,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AACzC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,QAAM,QAAQ,EAAE,YAAY,WAAW;AACvC,SAAO,OAAO,UAAU,aAAa,QAAQ,KAAK;AACpD;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,QAAQ,SAAS,OAAO,UAAU,GAAG,WAAW,GAAG,QAAQ,SAAS;AAC3E,MAAI,QAAQ,GAAG;AACb,UAAM,IAAI,MAAM,iDAAiD,OAAO;AAAA,EAC1E;AACA,QAAM,WAAW,gBAAgB,SAAS,WAAW,UAAU,OAAO;AACtE,QAAM,SAAS,EAAE,SAAS,SAAS;AACnC,QAAM,QAAQ,EAAE,OAAO,OAAO,SAAS,SAAS;AAChD,SAAO,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAC/C;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,iBAAiB;AACxB,MAAI,EAAE,IAAI,QAAQ,IAAI;AACxB;AACA,SAAS,kBAAkB;AACzB,MAAI,EAAE,IAAI,SAAS,IAAI;AACzB;AACA,SAAS,6BAA6B;AACpC,MAAI,EAAE,IAAI,gCAAgC,KAAK;AAC/C,UAAQ,KAAK,wDAAwD;AACvE;AACA,SAAS,gBAAgB,KAAK;AAC5B,MAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,YAAQ,KAAK,MAAM,6EAA6E;AAAA,EAClG;AACF;AACA,wBAAwB,eAAe;AACvC,SAAS,mBAAmB;AAC1B,SAAO,iBAAiB;AAC1B;AACA,SAAS,SAAS;AAChB,SAAO;AACT;AACA,SAAS,SAAS;AAChB,SAAO,OAAO,OAAO;AACvB;AACA,SAAS,QAAQ,GAAG;AAClB,SAAO,OAAO,QAAQ,CAAC;AACzB;AACA,SAAS,KAAK,UAAU,IAAI;AAC1B,SAAO,OAAO,KAAK,UAAU,EAAE;AACjC;AACA,SAAS,QAAQ,WAAW;AAC1B,QAAM,UAAU,sBAAsB,SAAS;AAC/C,UAAQ,QAAQ,CAAC,YAAY,QAAQ,QAAQ,CAAC;AAChD;AACA,SAAS,KAAK,QAAQ;AACpB,SAAO,OAAO,KAAK,MAAM;AAC3B;AACA,SAAS,KAAK,GAAG;AACf,SAAO,OAAO,KAAK,CAAC;AACtB;AACA,SAAS,WAAW,aAAa;AAC/B,SAAO,OAAO,WAAW,WAAW;AACtC;AACA,SAAS,QAAQ;AACf,SAAO,OAAO,MAAM;AACtB;AACA,SAAS,aAAa;AACpB,SAAO,OAAO;AAChB;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,cAAc,IAAI;AAC3B;AACA,SAAS,YAAY,MAAM;AACzB,SAAO,OAAO,YAAY,IAAI;AAChC;AACA,SAAS,mBAAmB,MAAM;AAChC,SAAO,OAAO,mBAAmB,IAAI;AACvC;AACA,SAAS,gBAAgB,MAAM,SAAS,WAAW,GAAG;AACpD,SAAO,OAAO,gBAAgB,MAAM,SAAS,QAAQ;AACvD;AACA,SAAS,UAAU;AACjB,SAAO,OAAO;AAChB;AACA,SAAS,YAAY,cAAc,UAAU;AAC3C,MAAI,EAAE,YAAY,cAAc,QAAQ;AAC1C;AAGA,SAAS,MAAM,QAAQ;AACrB,QAAM,SAAS,gBAAgB,QAAQ,SAAS,MAAM;AACtD,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,MAAM,QAAQ;AACrB,QAAM,SAAS,gBAAgB,QAAQ,SAAS,MAAM;AACtD,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,WAAW,GAAG,MAAM,WAAW;AACtC,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,MAAI,QAAQ,MAAM;AAChB,WAAO,GAAG,MAAM,IAAI,CAAC,GAAG,MAAM,CAAC,EAAE,QAAQ;AAAA,EAC3C;AACA,SAAO,GAAG,SAAS,KAAK,QAAQ,MAAM,qCAAqC,GAAG,kCAAkC,OAAO;AACvH,OAAK,QAAQ,CAAC,SAAS;AACrB,WAAO,QAAQ,KAAK,OAAO,GAAG,MAAM,MAAM,+CAA+C,GAAG,OAAO,aAAa,MAAM;AAAA,EACxH,CAAC;AACD,MAAI,GAAG,QAAQ,GAAG;AAChB,WAAO,GAAG,MAAM;AAAA,EAClB;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,KAAK;AACrB,MAAI,GAAG,UAAU,aAAa;AAC5B,WAAO,KAAK,MAAM;AAChB,UAAI,QAAQ,KAAK,EAAE;AACnB,UAAI,QAAQ,KAAK,EAAE;AACnB,cAAQ,OAAO,UAAU,WAAW,EAAE,GAAG,MAAM,GAAG,KAAK;AACvD,cAAQ,OAAO,UAAU,WAAW,EAAE,GAAG,MAAM,GAAG,KAAK;AACvD,UAAI,WAAW;AACb,gBAAQ,IAAI,KAAK;AAAA,MACnB;AACA,aAAO,QAAQ,OAAO,KAAK;AAAA,IAC7B,CAAC;AAAA,EACH;AACA,SAAO,OAAO,UAAU,WAAW,QAAQ,KAAK;AAClD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,iBAAiB,QAAQ,aAAa,YAAY;AACzD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,iBAAiB;AACnE,QAAM,eAAe,gBAAgB,aAAa,eAAe,iBAAiB;AAClF,SAAO,cAAc,QAAQ,aAAa,KAAK,OAAO,UAAU,UAAU,GAAG,MAAM,+DAA+D,YAAY;AAC9J,SAAO,QAAQ,SAAS,GAAG,MAAM,gDAAgD,QAAQ,MAAM;AAC/F,SAAO,aAAa,SAAS,GAAG,MAAM,qDAAqD,aAAa,MAAM;AAC9G,SAAO,QAAQ,MAAM,OAAO,aAAa,MAAM,IAAI,MAAM,uCAAuC,QAAQ,MAAM,UAAU,aAAa,MAAM,qEAAqE;AAChN,SAAO,aAAa,KAAK,OAAO,UAAU,UAAU,GAAG,MAAM,4DAA4D,YAAY;AACrI,QAAM,eAAe,OAAO,KAAK,SAAS,OAAO,GAAG,UAAU;AAC9D,QAAM,oBAAoB,OAAO,KAAK,cAAc,OAAO,GAAG,UAAU;AACxE,QAAM,gBAAgB,UAAU,YAAY;AAC5C,QAAM,UAAU,OAAO,eAAe,iBAAiB;AACvD,SAAO,KAAK,SAAS,OAAO;AAC9B;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,IAAI,yBAAyB,CAAC;AAC9BA,UAAS,wBAAwB;AAAA,EAC/B,4BAA4B,MAAM;AAAA,EAClC,kBAAkB,MAAM;AAAA,EACxB,kBAAkB,MAAM;AAC1B,CAAC;AACD,SAAS,iBAAiB,SAAS,UAAU;AAC3C,QAAM,SAAS,QAAQ;AACvB,QAAM,OAAO,CAAC;AACd,WAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,UAAM,MAAM,SAAS,IAAI;AACzB,UAAM,IAAI,QAAQ,QAAQ;AAC1B,UAAM,IAAI,SAAS,SAAS,SAAS,IAAI,MAAM;AAC/C,QAAI,IAAI,KAAK,MAAM,GAAG;AACpB,WAAK,QAAQ,GAAG;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,SAAS,UAAU;AAC3C,QAAM,SAAS,CAAC;AAChB,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAM,QAAQ,QAAQ,QAAQ,SAAS,IAAI;AAC3C,UAAM,UAAU,SAAS,SAAS,IAAI;AACtC,UAAM,SAAS,SAAS;AACxB,QAAI,SAAS,QAAQ,UAAU,KAAK,SAAS,GAAG;AAC9C,aAAO,QAAQ,OAAO;AAAA,IACxB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,2BAA2B,QAAQ,QAAQ;AAClD,QAAM,SAAS,CAAC;AAChB,QAAM,IAAI,KAAK,IAAI,OAAO,QAAQ,OAAO,MAAM;AAC/C,WAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,QAAI,IAAI,OAAO,OAAO,SAAS,IAAI;AACnC,QAAI,KAAK,MAAM;AACb,UAAI;AAAA,IACN;AACA,QAAI,IAAI,OAAO,OAAO,SAAS,IAAI;AACnC,QAAI,KAAK,MAAM;AACb,UAAI;AAAA,IACN;AACA,QAAI,MAAM,GAAG;AACX,aAAO,QAAQ,CAAC;AAAA,IAClB,WAAW,MAAM,GAAG;AAClB,aAAO,QAAQ,CAAC;AAAA,IAClB,WAAW,MAAM,GAAG;AAClB,YAAM,SAAS,wDAAwD,cAAc;AACrF,YAAM,MAAM,MAAM;AAAA,IACpB,OAAO;AACL,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,kBAAkB,CAAC;AACvBA,UAAS,iBAAiB;AAAA,EACxB,YAAY,MAAM;AAAA,EAClB,iBAAiB,MAAM;AAAA,EACvB,UAAU,MAAM;AAClB,CAAC;AAGD,SAAS,SAAS,QAAQ,OAAO,OAAO;AACtC,gBAAc,MAAM;AACpB,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,iDAAiD;AAAA,EACnE;AACA,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,KAAK,cAAc,WAAW,GAAG;AAC5D,UAAM,IAAI,MAAM,kEAAkE;AAAA,EACpF;AACA,MAAI,cAAc,WAAW,KAAK,SAAS,MAAM;AAC/C,UAAM,IAAI,MAAM,yEAAyE;AAAA,EAC3F;AACA,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,IAAI;AACJ,SAAS,YAAY,QAAQ,cAAc,GAAG;AAC5C,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AACA,MAAI,UAAU,MAAM;AAClB,UAAM,IAAI,MAAM,0DAA0D;AAAA,EAC5E;AACA,MAAI,eAAe;AACnB,MAAI,cAAc;AAClB,MAAI,UAAU;AACd,MAAI,UAAU;AACd,MAAI,eAAe;AACnB,MAAI,gBAAgB;AACpB,MAAI,OAAO,gBAAgB,YAAY;AACrC,mBAAe;AAAA,EACjB,WAAW,OAAO,cAAc,eAAe,kBAAkB,WAAW;AAC1E,kBAAc;AAAA,EAChB,WAAW,OAAO,qBAAqB,eAAe,kBAAkB,kBAAkB;AACxF,cAAU;AAAA,EACZ,WAAW,OAAO,qBAAqB,eAAe,kBAAkB,kBAAkB;AACxF,cAAU;AAAA,EACZ,WAAW,OAAO,cAAc,MAAM;AACpC,mBAAe;AAAA,EACjB,WAAW,OAAO,gBAAgB,eAAe,kBAAkB,aAAa;AAC9E,oBAAgB;AAAA,EAClB,OAAO;AACL,UAAM,IAAI,MAAM,qPAAqP,OAAO,YAAY,MAAM;AAAA,EAChS;AACA,QAAM,SAAS,UAAU,YAAY,OAAO,WAAW;AACvD,MAAI,UAAU,MAAM;AAClB,UAAM,SAAS,EAAE,OAAO;AACxB,UAAM,QAAQ,EAAE,YAAY;AAC5B,WAAO,OAAO,UAAU,YAAY,QAAQ,KAAK;AAAA,EACnD;AACA,QAAM,CAAC,OAAO,MAAM,IAAI,UAAU;AAAA,IAChC,OAAO;AAAA,IACP,OAAO;AAAA,EACT,IAAI,CAAC,OAAO,OAAO,OAAO,MAAM;AAChC,MAAI;AACJ,MAAI,cAAc;AAChB,WAAO,OAAO,WAAW,IAAI,EAAE,aAAa,GAAG,GAAG,OAAO,MAAM,EAAE;AAAA,EACnE,WAAW,eAAe,cAAc;AACtC,WAAO,OAAO;AAAA,EAChB,WAAW,WAAW,WAAW,eAAe;AAC9C,QAAI,uBAAuB,MAAM;AAC/B,UAAI,OAAO,aAAa,aAAa;AACnC,YAAI,OAAO,oBAAoB,eAAe,OAAO,sCAAsC,aAAa;AACtG,gCAAsB,IAAI,gBAAgB,GAAG,CAAC,EAAE,WAAW,IAAI;AAAA,QACjE,OAAO;AACL,gBAAM,IAAI,MAAM,sGAAsG;AAAA,QACxH;AAAA,MACF,OAAO;AACL,8BAAsB,SAAS,cAAc,QAAQ,EAAE,WAAW,MAAM,EAAE,oBAAoB,KAAK,CAAC;AAAA,MACtG;AAAA,IACF;AACA,wBAAoB,OAAO,QAAQ;AACnC,wBAAoB,OAAO,SAAS;AACpC,wBAAoB,UAAU,QAAQ,GAAG,GAAG,OAAO,MAAM;AACzD,WAAO,oBAAoB,aAAa,GAAG,GAAG,OAAO,MAAM,EAAE;AAAA,EAC/D;AACA,MAAI;AACJ,MAAI,gBAAgB,GAAG;AACrB,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,OAAO;AACL,UAAM,YAAY,QAAQ;AAC1B,aAAS,IAAI,WAAW,YAAY,WAAW;AAC/C,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,eAAS,UAAU,GAAG,UAAU,aAAa,EAAE,SAAS;AACtD,eAAO,IAAI,cAAc,WAAW,KAAK,IAAI,IAAI;AAAA,MACnD;AAAA,IACF;AAAA,EACF;AACA,QAAM,WAAW,CAAC,QAAQ,OAAO,WAAW;AAC5C,SAAO,SAAS,QAAQ,UAAU,OAAO;AAC3C;AACA,SAAS,YAAY,QAAQ;AAC3B,SAAO,UAAU,QAAQ,OAAO,gBAAgB;AAClD;AACA,SAAS,8BAA8B;AACrC,SAAO,OAAO,WAAW,eAAe,OAAO,gBAAgB,eAAe,OAAO,eAAe,mBAAmB;AACzH;AACA,SAAS,iBAAiB,QAAQ;AAChC,SAAO,UAAU,QAAQ,OAAO,UAAU,KAAK,OAAO,WAAW;AACnE;AACA,SAAS,2BAA2B,QAAQ;AAC1C,SAAO,4BAA4B,KAAK,EAAE,kBAAkB,gBAAgB,iBAAiB,MAAM,KAAK,CAAC,YAAY,MAAM;AAC7H;AACA,eAAe,gBAAgB,QAAQ,cAAc,GAAG;AACtD,MAAI,SAAS;AACb,MAAI,IAAI,EAAE,QAAQ,qBAAqB,KAAK,2BAA2B,MAAM,GAAG;AAC9E,QAAI;AACJ,QAAI;AACF,oBAAc,MAAM,kBAAkB,QAAQ,EAAE,kBAAkB,OAAO,CAAC;AAAA,IAC5E,SAAS,GAAP;AACA,oBAAc;AAAA,IAChB;AACA,QAAI,eAAe,QAAQ,YAAY,UAAU,OAAO,SAAS,YAAY,WAAW,OAAO,QAAQ;AACrG,eAAS;AAAA,IACX,OAAO;AACL,eAAS;AAAA,IACX;AAAA,EACF,OAAO;AACL,aAAS;AAAA,EACX;AACA,SAAO,YAAY,QAAQ,WAAW;AACxC;AACA,eAAe,SAAS,KAAK,QAAQ;AACnC,MAAI,OAAO,gBAAgB,KAAK,OAAO,UAAU;AACjD,MAAI,EAAE,eAAe,SAAS;AAC5B,UAAM,oBAAoB;AAC1B,WAAO,KAAK,mBAAmB,OAAO;AACtC,sBAAkB,QAAQ;AAAA,EAC5B;AACA,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,GAAG;AACtC,UAAM,IAAI,MAAM,wDAAwD,KAAK,OAAO;AAAA,EACtF;AACA,QAAM,CAAC,QAAQ,KAAK,IAAI,KAAK,MAAM,MAAM,GAAG,CAAC;AAC7C,QAAM,QAAQ,KAAK,SAAS,IAAI,IAAI,KAAK,MAAM;AAC/C,MAAI,QAAQ,KAAK,UAAU,GAAG;AAC5B,UAAM,IAAI,MAAM,0DAA0D,OAAO;AAAA,EACnF;AACA,MAAI,KAAK,UAAU,aAAa,KAAK,UAAU,SAAS;AACtD,UAAM,IAAI,MAAM,kCAAkC,KAAK,6CAA6C;AAAA,EACtG;AACA,QAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,QAAM,aAAa,KAAK,UAAU,YAAY,MAAM;AACpD,QAAM,QAAQ,IAAI,kBAAkB,QAAQ,SAAS,CAAC;AACtD,WAAS,IAAI,GAAG,IAAI,SAAS,OAAO,EAAE,GAAG;AACvC,UAAM,OAAO,CAAC,GAAG,GAAG,GAAG,GAAG;AAC1B,aAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,YAAM,QAAQ,KAAK,IAAI,QAAQ;AAC/B,UAAI,KAAK,UAAU,WAAW;AAC5B,YAAI,QAAQ,KAAK,QAAQ,GAAG;AAC1B,gBAAM,IAAI,MAAM,mFAAmF,QAAQ;AAAA,QAC7G;AAAA,MACF,WAAW,KAAK,UAAU,SAAS;AACjC,YAAI,QAAQ,KAAK,QAAQ,KAAK;AAC5B,gBAAM,IAAI,MAAM,mFAAmF,QAAQ;AAAA,QAC7G;AAAA,MACF;AACA,UAAI,UAAU,GAAG;AACf,aAAK,KAAK,QAAQ;AAClB,aAAK,KAAK,QAAQ;AAClB,aAAK,KAAK,QAAQ;AAAA,MACpB,OAAO;AACL,aAAK,KAAK,QAAQ;AAAA,MACpB;AAAA,IACF;AACA,UAAM,IAAI,IAAI;AACd,UAAM,IAAI,KAAK,KAAK,MAAM,KAAK,EAAE;AACjC,UAAM,IAAI,KAAK,KAAK,MAAM,KAAK,EAAE;AACjC,UAAM,IAAI,KAAK,KAAK,MAAM,KAAK,EAAE;AACjC,UAAM,IAAI,KAAK,KAAK,MAAM,KAAK,EAAE;AAAA,EACnC;AACA,MAAI,UAAU,MAAM;AAClB,WAAO,QAAQ;AACf,WAAO,SAAS;AAChB,UAAM,MAAM,OAAO,WAAW,IAAI;AAClC,UAAM,YAAY,IAAI,UAAU,OAAO,OAAO,MAAM;AACpD,QAAI,aAAa,WAAW,GAAG,CAAC;AAAA,EAClC;AACA,MAAI,SAAS,KAAK;AAChB,SAAK,QAAQ;AAAA,EACf;AACA,SAAO;AACT;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,IAAI,yBAAyB,CAAC;AAC9BA,UAAS,wBAAwB;AAAA,EAC/B,oBAAoB,MAAM;AAC5B,CAAC;AACD,SAAS,mBAAmB,SAAS,SAAS;AAC5C,QAAM,aAAa,QAAQ,MAAM;AACjC,QAAM,cAAc,QAAQ,MAAM;AAClC,MAAI,aAAa,GAAG;AAClB,UAAM,IAAI,MAAM,4EAA4E,aAAa;AAAA,EAC3G;AACA,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,8EAA8E,cAAc;AAAA,EAC9G;AACA,MAAI,QAAQ,UAAU,SAAS;AAC7B,UAAM,IAAI,MAAM,yEAAyE,QAAQ,QAAQ;AAAA,EAC3G;AACA,MAAI,QAAQ,MAAM,cAAc,KAAK,YAAY;AAC/C,UAAM,IAAI,MAAM,iEAAiE,QAAQ,MAAM,cAAc,UAAU,YAAY;AAAA,EACrI;AACA,MAAI,cAAc,QAAQ,KAAK,MAAM,GAAG;AACtC,UAAM,IAAI,MAAM,mEAAmE,QAAQ,QAAQ;AAAA,EACrG;AACA,QAAM,eAAe,QAAQ;AAC7B,QAAM,YAAY,aAAa,aAAa,SAAS;AACrD,MAAI,UAAU;AACd,WAAS,IAAI,GAAG,IAAI,aAAa,SAAS,GAAG,EAAE,GAAG;AAChD,eAAW,aAAa;AAAA,EAC1B;AACA,QAAM,aAAa,QAAQ;AAC3B,QAAM,cAAc,aAAa,MAAM;AACvC,cAAY,IAAI;AAChB,MAAI,YAAY;AAChB,WAAS,IAAI,WAAW,IAAI,YAAY,EAAE,GAAG;AAC3C,iBAAa,WAAW;AACxB,gBAAY,KAAK,WAAW,EAAE;AAAA,EAChC;AACA,QAAM,UAAU;AAAA,IACd,GAAG,eAAe,QAAQ,KAAK,EAAE,IAAI,CAAC,WAAW,SAAS,SAAS;AAAA,IACnE;AAAA,EACF,EAAE,MAAM,GAAG,SAAS;AACpB,SAAO,CAAC,aAAa,SAAS,WAAW,OAAO;AAClD;AAGA,IAAI,0BAA0B,CAAC;AAC/BA,UAAS,yBAAyB;AAAA,EAChC,iBAAiB,MAAM;AAAA,EACvB,eAAe,MAAM;AAAA,EACrB,qBAAqB,MAAM;AAC7B,CAAC;AACD,SAAS,oBAAoB,OAAO,SAAS,SAAS;AACpD,QAAM,WAAW,QAAQ,OAAO,IAAI,QAAQ,MAAM,QAAQ,OAAO,KAAK;AACtE,QAAM,WAAW,QAAQ,OAAO,IAAI,QAAQ,OAAO,IAAI;AACvD,QAAM,aAAa,6FAA6F,QAAQ,yBAAyB,QAAQ,iBAAiB,oBAAoB,2BAA2B;AACzN,MAAI,QAAQ,OAAO,UAAU;AAC3B,UAAM,IAAI,MAAM,aAAa,kBAAkB,YAAY;AAAA,EAC7D;AACA,MAAI,MAAM,SAAS,YAAY,QAAQ,OAAO,WAAW;AACvD,UAAM,IAAI,MAAM,aAAa,0BAA0B,YAAY,QAAQ,OAAO,WAAW;AAAA,EAC/F;AACA,MAAI,QAAQ,SAAS,WAAW,MAAM,SAAS,UAAU;AACvD,UAAM,IAAI,MAAM,aAAa,mBAAmB,WAAW,MAAM,SAAS,UAAU;AAAA,EACtF;AACA,WAAS,IAAI,GAAG,IAAI,UAAU,EAAE,GAAG;AACjC,QAAI,QAAQ,MAAM,OAAO,QAAQ,MAAM,IAAI;AACzC,YAAM,IAAI,MAAM,aAAa,kBAAkB,OAAO,QAAQ,MAAM,wBAAwB,OAAO,QAAQ,MAAM,MAAM;AAAA,IACzH;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,OAAO,UAAU,EAAE,GAAG;AAChD,QAAI,QAAQ,MAAM,IAAI,cAAc,MAAM,IAAI,WAAW;AACvD,YAAM,IAAI,MAAM,aAAa,kBAAkB,IAAI,cAAc,QAAQ,MAAM,IAAI,uBAAuB,IAAI,cAAc,MAAM,IAAI,YAAY;AAAA,IACpJ;AAAA,EACF;AACF;AACA,SAAS,cAAc,SAAS,SAAS,OAAO;AAC9C,MAAI,QAAQ,OAAO,GAAG;AACpB,UAAM,IAAI,MAAM,+EAA+E,QAAQ,OAAO;AAAA,EAChH;AACA,MAAI,QAAQ,OAAO,GAAG;AACpB,UAAM,IAAI,MAAM,+EAA+E,QAAQ,OAAO;AAAA,EAChH;AACA,MAAI,QAAQ,UAAU,SAAS;AAC7B,UAAM,IAAI,MAAM,0DAA0D,QAAQ,OAAO;AAAA,EAC3F;AACA,MAAI,MAAM,SAAS,GAAG;AACpB,UAAM,IAAI,MAAM,6DAA6D,OAAO;AAAA,EACtF;AACA,MAAI,MAAM,WAAW,GAAG;AACtB,QAAI,QAAQ,SAAS,GAAG;AACtB,YAAM,IAAI,MAAM,sDAAsD,QAAQ,OAAO;AAAA,IACvF;AACA,QAAI,QAAQ,SAAS,GAAG;AACtB,YAAM,IAAI,MAAM,sDAAsD,QAAQ,OAAO;AAAA,IACvF;AAAA,EACF;AACA,sBAAoB,OAAO,SAAS,OAAO;AAC7C;AACA,SAAS,gBAAgB,SAAS,SAAS,OAAO;AAChD,QAAM,cAAc,QAAQ,MAAM;AAClC,QAAM,YAAY,cAAc,IAAI,QAAQ,MAAM,cAAc,KAAK;AACrE,QAAM,UAAU,MAAM;AACtB,MAAI,YAAY;AAChB,WAAS,IAAI,WAAW,IAAI,SAAS,EAAE,GAAG;AACxC,iBAAa,MAAM;AAAA,EACrB;AACA,QAAM,eAAe,YAAY,IAAI,IAAI;AACzC,QAAM,aAAa,cAAc,QAAQ,KAAK,IAAI;AAClD,QAAM,UAAU,CAAC,GAAG,eAAe,MAAM,MAAM,GAAG,SAAS,CAAC,GAAG,CAAC;AAChE,QAAM,aAAa,cAAc,KAAK;AACtC,SAAO,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW;AACjE;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,mBAAmB,MAAM;AAAA,EACzB,mBAAmB,MAAM;AAAA,EACzB,iBAAiB,MAAM;AAAA,EACvB,mBAAmB,MAAM;AAAA,EACzB,kBAAkB,MAAM;AAAA,EACxB,YAAY,MAAM;AAAA,EAClB,kBAAkB,MAAM;AAAA,EACxB,WAAW,MAAM;AAAA,EACjB,cAAc,MAAM;AAAA,EACpB,4BAA4B,MAAM;AAAA,EAClC,aAAa,MAAM;AAAA,EACnB,2BAA2B,MAAM;AAAA,EACjC,gBAAgB,MAAM;AAAA,EACtB,uBAAuB,MAAM;AAC/B,CAAC;AACD,IAAI,WAAW;AACf,IAAI,cAAc;AAClB,SAAS,kBAAkB,QAAQ,OAAO,MAAM;AAC9C,QAAM,YAAY,OAAO,MAAM;AAC/B,SAAO,cAAc,MAAM,QAAQ,MAAM,iBAAiB,+BAA+B,2CAA2C,aAAa;AACjJ,SAAO,cAAc,KAAK,QAAQ,MAAM,iBAAiB,8BAA8B,0CAA0C,aAAa;AAC9I,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,WAAO,MAAM,KAAK,KAAK,MAAM,OAAO,MAAM,IAAI,MAAM,iBAAiB,qBAAqB,aAAa,OAAO,MAAM,KAAK,KAAK,kCAAkC,OAAO,OAAO,MAAM,KAAK;AAAA,EAC3L;AACF;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,OAAO,CAAC;AACd,MAAI,OAAO;AACX,SAAO,OAAO,GAAG;AACf,QAAI,OAAO,GAAG;AACZ,WAAK,KAAK,IAAI;AAAA,IAChB;AACA,YAAQ;AACR;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,gBAAgB,OAAO,KAAK,SAAS;AAC5C,QAAM,OAAO,CAAC;AACd,WAAS,OAAO,GAAG,OAAO,MAAM,QAAQ,QAAQ;AAC9C,SAAK,QAAQ,KAAK,MAAM,IAAI,QAAQ,MAAM,SAAS,QAAQ,KAAK;AAAA,EAClE;AACA,SAAO;AACT;AACA,SAAS,sBAAsB,SAAS,wBAAwB,eAAe,YAAY;AACzF,QAAM,aAAa,CAAC,GAAG,OAAO;AAC9B,WAAS,IAAI,WAAW,QAAQ,IAAI,WAAW,QAAQ,KAAK;AAC1D,eAAW,KAAK,CAAC;AAAA,EACnB;AACA,WAAS,IAAI,GAAG,IAAI,eAAe,KAAK;AACtC,QAAI,MAAM,GAAG;AACX,iBAAW,0BAA0B;AAAA,IACvC,OAAO;AACL,iBAAW,OAAO,wBAAwB,GAAG,CAAC;AAC9C,iBAAW,IAAI;AAAA,IACjB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,gBAAgB,wBAAwB,eAAe,gBAAgB;AAC9E,MAAI,kBAAkB,wBAAwB;AAC5C,WAAO;AAAA,EACT;AACA,SAAO,kBAAkB,gBAAgB;AAC3C;AACA,SAAS,cAAc,eAAe,wBAAwB;AAC5D,QAAM,aAAa,CAAC;AACpB,WAAS,IAAI,GAAG,IAAI,eAAe,KAAK;AACtC,eAAW,KAAK,yBAAyB,CAAC;AAAA,EAC5C;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,YAAY,cAAc,qBAAqB,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc;AAC/H,QAAM,YAAY,WAAW;AAC7B,MAAI,kBAAkB,IAAI,MAAM,SAAS,GAAG,gBAAgB,IAAI,MAAM,SAAS,GAAG,oBAAoB,IAAI,MAAM,SAAS;AACzH,MAAI,aAAa,UAAU,sBAAsB,GAAG;AAClD,UAAM,YAAY,aAAa;AAC/B,UAAM,gBAAgB,sBAAsB;AAC5C,sBAAkB,2BAA2B,WAAW,WAAW,eAAe,OAAO,UAAU;AACnG,oBAAgB,0BAA0B,SAAS,WAAW,eAAe,KAAK,UAAU;AAC5F,wBAAoB,sBAAsB,SAAS,WAAW,eAAe,UAAU;AAAA,EACzF,OAAO;AACL,aAAS,OAAO,GAAG,OAAO,WAAW,QAAQ;AAC3C,sBAAgB,QAAQ,aAAa,WAAW,OAAO,SAAS,YAAY,MAAM,YAAY;AAC9F,oBAAc,QAAQ,YAAY,SAAS,KAAK,SAAS,YAAY,MAAM,YAAY;AACvF,wBAAkB,QAAQ,eAAe,SAAS,MAAM,YAAY;AAAA,IACtE;AAAA,EACF;AACA,SAAO;AAAA,IACL,OAAO;AAAA,IACP,KAAK;AAAA,IACL,SAAS;AAAA,EACX;AACF;AACA,SAAS,2BAA2B,WAAW,wBAAwB,eAAe,eAAe,YAAY;AAC/G,QAAM,aAAa,CAAC,GAAG,UAAU;AACjC,QAAM,aAAa,cAAc,eAAe,sBAAsB;AACtE,WAAS,OAAO,GAAG,OAAO,WAAW,QAAQ,QAAQ;AACnD,QAAI,WAAW,QAAQ,IAAI,IAAI,IAAI;AACjC,iBAAW,QAAQ;AAAA,IACrB,OAAO;AACL,YAAM,eAAe,gBAAgB,wBAAwB,eAAe,IAAI;AAChF,UAAI,gBAAgB,cAAc;AAClC,UAAI,YAAY,KAAK,cAAc;AACjC,wBAAgB;AAAA,MAClB;AACA,iBAAW,QAAQ;AAAA,IACrB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,0BAA0B,SAAS,wBAAwB,eAAe,aAAa,YAAY;AAC1G,QAAM,aAAa,CAAC,GAAG,UAAU;AACjC,QAAM,aAAa,cAAc,eAAe,sBAAsB;AACtE,WAAS,OAAO,GAAG,OAAO,WAAW,QAAQ,QAAQ;AACnD,QAAI,WAAW,QAAQ,IAAI,IAAI,IAAI;AACjC,iBAAW,QAAQ,OAAO;AAAA,IAC5B,OAAO;AACL,YAAM,eAAe,gBAAgB,wBAAwB,eAAe,IAAI;AAChF,UAAI,gBAAgB,YAAY;AAChC,UAAI,UAAU,KAAK,cAAc;AAC/B,wBAAgB,OAAO;AAAA,MACzB;AACA,iBAAW,QAAQ;AAAA,IACrB;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,KAAK;AAC1C,UAAM,WAAW,WAAW;AAC5B,QAAI,WAAW,KAAK,GAAG;AACrB,iBAAW,MAAM;AAAA,IACnB;AACA,eAAW,KAAK,MAAM,GAAG,WAAW,IAAI,WAAW,EAAE;AAAA,EACvD;AACA,SAAO;AACT;AACA,SAAS,eAAe,SAAS,MAAM,cAAc;AACnD,MAAI,SAAS,QAAQ;AACrB,MAAI,eAAe,KAAK,QAAQ,UAAU,MAAM;AAC9C,aAAS;AAAA,EACX;AACA,SAAO;AACT;AACA,SAAS,aAAa,WAAW,cAAc,SAAS,YAAY,MAAM,cAAc;AACtF,MAAI,QAAQ,aAAa;AACzB,QAAM,SAAS,QAAQ,SAAS;AAChC,MAAI,YAAY,KAAK,QAAQ,eAAe,KAAK,QAAQ,SAAS,MAAM;AACtE,QAAI,SAAS,GAAG;AACd,cAAQ,OAAO;AAAA,IACjB,OAAO;AACL,cAAQ,OAAO;AAAA,IACjB;AAAA,EACF;AACA,QAAM,WAAW,WAAW;AAC5B,MAAI,QAAQ,GAAG;AACb,aAAS;AAAA,EACX;AACA,UAAQ,MAAM,GAAG,OAAO,WAAW,CAAC;AACpC,SAAO;AACT;AACA,SAAS,YAAY,SAAS,aAAa,SAAS,YAAY,MAAM,cAAc;AAClF,MAAI,OAAO,YAAY;AACvB,QAAM,SAAS,QAAQ,SAAS;AAChC,MAAI,UAAU,KAAK,QAAQ,eAAe,KAAK,QAAQ,QAAQ,MAAM;AACnE,QAAI,SAAS,GAAG;AACd,aAAO,OAAO;AAAA,IAChB,OAAO;AACL,aAAO,OAAO;AAAA,IAChB;AAAA,EACF;AACA,QAAM,WAAW,WAAW;AAC5B,MAAI,OAAO,GAAG;AACZ,YAAQ;AAAA,EACV;AACA,MAAI,SAAS,GAAG;AACd,WAAO,MAAM,GAAG,MAAM,QAAQ;AAAA,EAChC,OAAO;AACL,WAAO,MAAM,IAAI,MAAM,WAAW,CAAC;AAAA,EACrC;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,OAAO,OAAO,MAAM;AAC5C,MAAI,kBAAkB,KAAK;AAC3B,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,QAAI,KAAK,KAAK,GAAG;AACf,wBAAkB;AAClB;AAAA,IACF;AAAA,EACF;AACA,WAAS,IAAI,kBAAkB,GAAG,IAAI,KAAK,QAAQ,KAAK;AACtD,QAAI,MAAM,KAAK,KAAK,KAAK,OAAO,MAAM,IAAI;AACxC,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,OAAO,SAAS;AACzC,MAAI,aAAa,MAAM,SAAS,IAAI,MAAM,MAAM,SAAS,KAAK;AAC9D,WAAS,IAAI,GAAG,IAAI,MAAM,SAAS,GAAG,KAAK;AACzC,kBAAc,MAAM,KAAK,QAAQ;AAAA,EACnC;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,GAAG,OAAO,MAAM;AACxC,MAAI;AACJ,QAAM,QAAQ,EAAE,MAAM;AACtB,MAAI,OAAO,UAAU,UAAU;AAC7B,aAAS,CAAC,OAAO,GAAG,IAAI,MAAM,QAAQ,CAAC,EAAE,KAAK,CAAC,CAAC;AAAA,EAClD,WAAW,MAAM,SAAS,OAAO;AAC/B,aAAS,MAAM,OAAO,IAAI,MAAM,QAAQ,MAAM,MAAM,EAAE,KAAK,CAAC,CAAC;AAAA,EAC/D,OAAO;AACL,aAAS,MAAM,MAAM;AAAA,EACvB;AACA,SAAO,QAAQ,CAAC,MAAM;AACpB,WAAO,MAAM,IAAI,MAAM,mDAAmD;AAAA,EAC5E,CAAC;AACD,MAAI;AACJ,MAAI,QAAQ,MAAM;AAChB,YAAQ,IAAI,MAAM,KAAK,EAAE,KAAK,EAAE;AAAA,EAClC,WAAW,OAAO,SAAS,UAAU;AACnC,YAAQ,CAAC,MAAM,GAAG,IAAI,MAAM,QAAQ,CAAC,EAAE,KAAK,EAAE,CAAC;AAAA,EACjD,WAAW,KAAK,SAAS,OAAO;AAC9B,YAAQ,KAAK,OAAO,IAAI,MAAM,QAAQ,KAAK,MAAM,EAAE,KAAK,EAAE,CAAC;AAAA,EAC7D,OAAO;AACL,YAAQ;AAAA,EACV;AACA,UAAQ,MAAM,IAAI,CAAC,GAAG,MAAM;AAC1B,QAAI,KAAK,GAAG;AACV,aAAO;AAAA,IACT,OAAO;AACL,aAAO,MAAM,IAAI,MAAM,qDAAqD,mCAAmC,IAAI;AACnH,aAAO,EAAE,MAAM,KAAK,OAAO;AAAA,IAC7B;AAAA,EACF,CAAC;AACD,SAAO,CAAC,QAAQ,KAAK;AACvB;AACA,SAAS,UAAU,QAAQ,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,gBAAgB;AAC7G,MAAI;AACJ,MAAI,WAAW,MAAM;AACnB,qBAAiB,IAAI,MAAM,MAAM,MAAM;AACvC,mBAAe,KAAK,CAAC;AAAA,EACvB,OAAO;AACL,qBAAiB;AAAA,EACnB;AACA,MAAI,gBAAgB,SAAS,eAAe,eAAe,OAAO,GAAG;AACnE,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AACA,MAAI,eAAe;AACnB,QAAM,aAAa;AAAA,IACjB,MAAM,eAAe;AAAA,IACrB,yBAAyB;AAAA,IACzB,OAAO,MAAM,MAAM;AAAA,IACnB,KAAK,IAAI,MAAM;AAAA,IACf,SAAS,eAAe,MAAM;AAAA,IAC9B;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,WAAW,MAAM,KAAK;AACxC,QAAI,iBAAiB,KAAK,IAAI,iBAAiB,GAAG;AAChD,iBAAW;AAAA,IACb;AACA,QAAI,KAAK,IAAI,cAAc;AACzB,qBAAe;AAAA,IACjB;AAAA,EACF;AACA,MAAI,CAAC,cAAc;AACjB,eAAW,gBAAgB,KAAK,WAAW;AAC3C,eAAW;AAAA,EACb;AACA,QAAM,YAAY;AAAA,IAChB,MAAM,OAAO;AAAA,IACb,WAAW;AAAA,IACX,SAAS;AAAA,IACT,YAAY;AAAA,IACZ,UAAU;AAAA,EACZ;AACA,iBAAe,YAAY,SAAS;AACpC,MAAI,aAAa;AACjB,MAAI,YAAY;AAChB,MAAI,gBAAgB;AACpB,QAAM,kBAAkB,CAAC;AACzB,QAAM,aAAa,CAAC;AACpB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,QAAI,UAAU,QAAQ,OAAO,GAAG;AAC9B,YAAM,MAAM,WAAW,qBAAqB;AAAA,IAC9C;AACA,UAAM,UAAU,CAAC,EAAE,UAAU,iBAAiB,KAAK;AACnD,UAAM,OAAO,OAAO;AACpB,QAAI,SAAS,IAAI;AACf,sBAAgB,KAAK,UAAU,IAAI,EAAE;AACrC;AAAA,IACF;AACA,UAAM,QAAQ,CAAC,UAAU,YAAY,KAAK,GAAG,UAAU,UAAU,KAAK,CAAC;AACvE,UAAM,aAAa;AAAA,MACjB,UAAU,QAAQ,KAAK,IAAI,IAAI;AAAA,MAC/B,UAAU,QAAQ,KAAK,IAAI,OAAO,OAAO;AAAA,IAC3C;AACA,QAAI,WAAW,UAAU,QAAQ,MAAM,GAAG;AACxC,YAAM,MAAM,8CAA8C;AAAA,IAC5D;AACA,oBAAgB,iBAAiB,UAAU,QAAQ,OAAO;AAC1D,UAAM,oBAAoB,CAAC,EAAE,UAAU,YAAY,KAAK,KAAK,UAAU,UAAU,KAAK;AACtF,QAAI,UAAU,cAAc,UAAU,UAAU;AAC9C,UAAI,SAAS;AACX,cAAM,OAAO,UAAU,MAAM,KAAK,IAAI,OAAO,UAAU,MAAM,KAAK,UAAU,MAAM;AAClF,kBAAU,MAAM,KAAK;AACrB,kBAAU,IAAI,KAAK,UAAU,MAAM,KAAK;AACxC,YAAI,OAAO,KAAK,QAAQ,MAAM;AAC5B,gBAAM,MAAM,eAAe,UAAU,MAAM,mBAAmB,kBAAkB;AAAA,QAClF;AAAA,MACF,OAAO;AACL,kBAAU,MAAM,KAAK,UAAU,UAAU,MAAM,IAAI,GAAG,UAAU,QAAQ,IAAI,MAAM,OAAO,UAAU;AACnG,kBAAU,IAAI,KAAK,UAAU,UAAU,IAAI,IAAI,GAAG,UAAU,QAAQ,IAAI,MAAM,OAAO,UAAU;AAAA,MACjG;AACA,YAAM,qBAAqB,UAAU,QAAQ,OAAO,KAAK,UAAU,MAAM,OAAO,KAAK,UAAU,IAAI,OAAO;AAC1G,mBAAa,cAAc;AAC3B,kBAAY,cAAc,MAAM,KAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,IACrE,OAAO;AACL,mBAAa,eAAe,UAAU,QAAQ,OAAO,KAAK;AAC1D,kBAAY,cAAc,MAAM,KAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,IACrE;AACA,QAAI;AACJ,QAAI,gBAAgB;AACpB,QAAI,UAAU,cAAc,UAAU,UAAU;AAC9C,uBAAiB,UAAU,IAAI,KAAK,UAAU,MAAM;AACpD,sBAAgB;AAAA,IAClB,WAAW,SAAS;AAClB,uBAAiB;AACjB,sBAAgB;AAAA,IAClB,WAAW,mBAAmB;AAC5B,UAAI,QAAQ,GAAG;AACb,YAAI,UAAU,QAAQ,KAAK,GAAG;AAC5B,2BAAiB,CAAC;AAAA,QACpB,OAAO;AACL,2BAAiB;AAAA,QACnB;AACA,wBAAgB;AAAA,MAClB;AAAA,IACF;AACA,QAAI,eAAe;AACjB,UAAI;AACJ,UAAI,mBAAmB,KAAK,iBAAiB,MAAM,UAAU,QAAQ,KAAK,GAAG;AAC3E,gBAAQ;AAAA,MACV,OAAO;AACL,gBAAQ,KAAK,MAAM,iBAAiB,UAAU,QAAQ,EAAE,KAAK,iBAAiB,UAAU,QAAQ,OAAO,IAAI,IAAI;AAAA,MACjH;AACA,sBAAgB,KAAK,KAAK;AAAA,IAC5B,OAAO;AACL,sBAAgB,KAAK,EAAE;AAAA,IACzB;AAAA,EACF;AACA,WAAS,WAAW,GAAG,WAAW,UAAU,wBAAwB,QAAQ,EAAE,UAAU;AACtF,UAAM,cAAc,UAAU,wBAAwB;AACtD,QAAI,eAAe,GAAG;AACpB,iBAAW,KAAK,gBAAgB,YAAY;AAAA,IAC9C,WAAW,gBAAgB,UAAU;AACnC,iBAAW,KAAK,CAAC;AAAA,IACnB;AAAA,EACF;AACA,QAAM,mBAAmB,WAAW,OAAO,CAAC,KAAK,MAAM,UAAU,wBAAwB,OAAO,QAAQ;AACxG,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,OAAO,UAAU;AAAA,IACjB,KAAK,UAAU;AAAA,IACf,SAAS,UAAU;AAAA,EACrB;AACF;AACA,SAAS,eAAe,SAAS,QAAQ;AACvC,SAAO,YAAY;AACnB,SAAO,UAAU;AACjB,SAAO,iBAAiB;AACxB,MAAI,YAAY;AAChB,SAAO,aAAa,QAAQ,SAAS;AACrC,SAAO,WAAW,QAAQ,OAAO;AACjC,SAAO,QAAQ,IAAI,MAAM,OAAO,IAAI;AACpC,SAAO,MAAM,IAAI,MAAM,OAAO,IAAI;AAClC,SAAO,UAAU,IAAI,MAAM,OAAO,IAAI;AACtC,SAAO,0BAA0B,CAAC;AAClC,SAAO,gCAAgC,CAAC;AACxC,SAAO,gCAAgC,IAAI,MAAM,OAAO,IAAI;AAC5D,WAAS,IAAI,GAAG,IAAI,QAAQ,MAAM,KAAK;AACrC,QAAI,KAAK,IAAI,QAAQ,cAAc;AACjC,YAAM,YAAY,KAAK,IAAI,OAAO,QAAQ,QAAQ,OAAO,KAAK,IAAI,QAAQ,yBAAyB,OAAO,IAAI;AAC9G,aAAO,YAAY,WAAW,aAAa;AACzC,eAAO,MAAM,aAAa;AAC1B,eAAO,IAAI,aAAa;AACxB,eAAO,QAAQ,aAAa;AAC5B,eAAO,aAAa,KAAK;AACzB,eAAO,WAAW,KAAK;AACvB,eAAO,wBAAwB,KAAK,SAAS;AAC7C,eAAO,8BAA8B,KAAK,EAAE;AAC5C,eAAO,8BAA8B,aAAa;AAAA,MACpD;AAAA,IACF,WAAW,KAAK,IAAI,QAAQ,aAAa;AACvC,aAAO,wBAAwB,KAAK,QAAQ;AAC5C,aAAO,8BAA8B,KAAK,EAAE;AAAA,IAC9C,OAAO;AACL,UAAI,cAAc,OAAO,MAAM,QAAQ;AACrC,cAAM,MAAM,sCAAsC,6BAA6B,OAAO,cAAc,OAAO,MAAM,SAAS;AAAA,MAC5H;AACA,UAAI,QAAQ,SAAS,MAAM;AACzB,eAAO,MAAM,aAAa,QAAQ,MAAM;AAAA,MAC1C;AACA,UAAI,QAAQ,OAAO,MAAM;AACvB,eAAO,IAAI,aAAa,QAAQ,IAAI;AAAA,MACtC;AACA,aAAO,QAAQ,aAAa,QAAQ,QAAQ;AAC5C,UAAI,QAAQ,YAAY,KAAK,GAAG;AAC9B,eAAO,aAAa,KAAK;AAAA,MAC3B;AACA,UAAI,QAAQ,UAAU,KAAK,GAAG;AAC5B,eAAO,WAAW,KAAK;AAAA,MACzB;AACA,UAAI,QAAQ,iBAAiB,KAAK,GAAG;AACnC,eAAO,wBAAwB,KAAK,WAAW;AAC/C,eAAO,8BAA8B,KAAK,EAAE;AAC5C,eAAO,kBAAkB,KAAK;AAAA,MAChC,OAAO;AACL,eAAO,wBAAwB,KAAK,SAAS;AAC7C,eAAO,8BAA8B,KAAK,CAAC;AAAA,MAC7C;AACA,aAAO,8BAA8B,aAAa;AAClD;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,UAAU,GAAG,GAAG,SAAS,MAAM,OAAO,YAAY;AACzD,MAAI,MAAM,IAAI;AACZ,WAAO,UAAU,IAAI,WAAW,KAAK,WAAW,IAAI,IAAI;AAAA,EAC1D,OAAO;AACL,UAAM,OAAO,IAAI,IAAI,OAAO,IAAI;AAChC,WAAO,OAAO,WAAW,KAAK,WAAW,KAAK,OAAO,WAAW,KAAK,WAAW,KAAK;AAAA,EACvF;AACF;AAGA,IAAI,wBAAwB,CAAC;AAC7BA,UAAS,uBAAuB;AAAA,EAC9B,cAAc,MAAM;AAAA,EACpB,kBAAkB,MAAM;AAAA,EACxB,eAAe,MAAM;AACvB,CAAC;AACD,IAAI,eAAe,MAAM;AAAA,EACvB,eAAe;AACb,WAAO,KAAK,YAAY;AAAA,EAC1B;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,MAAM;AAAA,EACvB;AACF;AACA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,cAAc;AACZ,SAAK,eAAe,CAAC;AAAA,EACvB;AAAA,EACA,OAAO,SAAS;AACd,QAAI,iBAAiB,YAAY,MAAM;AACrC,uBAAiB,WAAW,IAAI,iBAAiB;AAAA,IACnD;AACA,WAAO,iBAAiB;AAAA,EAC1B;AAAA,EACA,OAAO,SAAS,KAAK;AACnB,qBAAiB,OAAO,EAAE,aAAa,IAAI,aAAa,CAAC,KAAK,IAAI,UAAU;AAAA,EAC9E;AACF;AACA,SAAS,cAAc,KAAK;AAC1B,SAAO,IAAI,aAAa,MAAM,MAAM,6EAA6E;AACjH,SAAO,OAAO,IAAI,cAAc,UAAU,MAAM,wDAAwD,OAAO,IAAI,SAAS;AAC5H,SAAO,IAAI,UAAU,SAAS,GAAG,MAAM,mFAAmF;AAC1H,mBAAiB,SAAS,GAAG;AAC/B;AAGA,IAAI,oBAAoB,CAAC;AACzBA,UAAS,mBAAmB;AAAA,EAC1B,sBAAsB,MAAM;AAAA,EAC5B,oBAAoB,MAAM;AAAA,EAC1B,eAAe,MAAM;AAAA,EACrB,yBAAyB,MAAM;AAAA,EAC/B,mBAAmB,MAAM;AAAA,EACzB,mBAAmB,MAAM;AAAA,EACzB,oBAAoB,MAAM;AAAA,EAC1B,qBAAqB,MAAM;AAAA,EAC3B,qBAAqB,MAAM;AAAA,EAC3B,MAAM,MAAM;AAAA,EACZ,aAAa,MAAM;AACrB,CAAC;AACD,IAAI,uBAAuB;AAC3B,IAAI,uBAAuB;AAC3B,SAAS,kBAAkB,QAAQ,UAAUG,WAAU;AACrD,MAAIA,aAAY,MAAM;AACpB,IAAAA,YAAW,YAAY;AAAA,EACzB;AACA,SAAO,sBAAsB,QAAQ,UAAU,CAAC,GAAG,MAAM,SAAS,GAAG,GAAGA,SAAQ,CAAC;AACnF;AACA,SAAS,cAAc;AACrB,SAAO,OAAO,QAAQ,eAAe,MAAM,KAAK,uBAAuB;AACzE;AACA,SAAS,sBAAsB,QAAQ,UAAU,WAAW;AAC1D,MAAI,iBAAiB;AACrB,MAAI,aAAa,MAAM,KAAK,aAAa,QAAQ,GAAG;AAClD,qBAAiB;AAAA,EACnB;AACA,MAAI,aAAa,MAAM,KAAK,aAAa,QAAQ,GAAG;AAClD,qBAAiB;AAAA,EACnB;AACA,MAAI,gBAAgB;AAClB,UAAM,QAAQ,OAAO,YAAY;AACjC,UAAM,QAAQ,SAAS,YAAY;AACnC,QAAI,UAAU,OAAO;AACnB,YAAM,IAAI,MAAM,yCAAyC,oBAAoB,OAAO;AAAA,IACtF;AAAA,EACF;AACA,MAAI,MAAM,QAAQ,MAAM,KAAK,MAAM,QAAQ,QAAQ,GAAG;AACpD,UAAM,cAAc,WAAW,MAAM;AACrC,UAAM,gBAAgB,WAAW,QAAQ;AACzC,QAAI,CAAC,YAAY,aAAa,aAAa,GAAG;AAC5C,YAAM,IAAI,MAAM,0CAA0C,4BAA4B,gBAAgB;AAAA,IACxG;AAAA,EACF;AACA,QAAM,aAAa,aAAa,MAAM,IAAI,SAAS,QAAQ,MAAM;AACjE,QAAM,eAAe,aAAa,QAAQ,IAAI,WAAW,QAAQ,QAAQ;AACzE,MAAI,WAAW,WAAW,aAAa,QAAQ;AAC7C,UAAM,IAAI,MAAM,yCAAyC,WAAW,uBAAuB,aAAa;AAAA,YAChG;AAAA,YACA,eAAe;AAAA,EACzB;AACA,WAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,UAAM,IAAI,WAAW;AACrB,UAAM,IAAI,aAAa;AACvB,QAAI,CAAC,UAAU,GAAG,CAAC,GAAG;AACpB,YAAM,IAAI,MAAM,yBAAyB,QAAQ,eAAe,QAAQ;AAAA,YAClE;AAAA,YACA,eAAe;AAAA,IACvB;AAAA,EACF;AACA,MAAI,OAAO,WAAW,aAAa;AACjC,WAAO,EAAE,QAAQ;AAAA,EACnB;AACF;AACA,SAAS,oBAAoB,IAAI,MAAM;AACrC,KAAG,EAAE,KAAK,MAAM,KAAK,KAAK,GAAG,MAAM,KAAK,CAAC;AACzC,MAAI,OAAO,WAAW,aAAa;AACjC,WAAO,EAAE,QAAQ;AAAA,EACnB;AACF;AACA,SAAS,kBAAkB,QAAQ,UAAU;AAC3C,QAAM,OAAO,OAAO,aAAa,YAAY,OAAO,aAAa,YAAY,OAAO,aAAa,YAAY,CAAC,QAAQ,IAAI;AAC1H,MAAI,SAAS,MAAM,KAAK,SAAS,OAAO,EAAE,KAAK,SAAS,QAAQ,KAAK,SAAS,SAAS,EAAE,GAAG;AAC1F,WAAO,sBAAsB,QAAQ,MAAM,CAAC,GAAG,MAAM,KAAK,CAAC;AAAA,EAC7D;AACA,SAAO,sBAAsB,QAAQ,UAAU,CAAC,GAAG,MAAM,SAAS,GAAG,GAAG,CAAC,CAAC;AAC5E;AACA,SAAS,mBAAmB,GAAG,GAAGA,WAAU;AAC1C,MAAIA,aAAY,MAAM;AACpB,IAAAA,YAAW,YAAY;AAAA,EACzB;AACA,MAAI,CAAC,SAAS,GAAG,GAAGA,SAAQ,GAAG;AAC7B,UAAM,IAAI,MAAM,8BAA8B,mBAAmB,GAAG;AAAA,EACtE;AACA,MAAI,OAAO,WAAW,aAAa;AACjC,WAAO,EAAE,QAAQ;AAAA,EACnB;AACF;AACA,SAAS,SAAS,GAAG,GAAGA,WAAU;AAChC,MAAI,CAAC,SAAS,CAAC,KAAK,CAAC,SAAS,CAAC,GAAG;AAChC,WAAO;AAAA,EACT;AACA,MAAI,MAAM,CAAC,KAAK,MAAM,CAAC,KAAK,KAAK,IAAI,IAAI,CAAC,IAAIA,WAAU;AACtD,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,QAAQ,KAAK,MAAM;AAC9C,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,QAAI,OAAO,KAAK,OAAO,OAAO,KAAK,MAAM;AACvC,YAAM,IAAI,MAAM,sBAAsB,OAAO,WAAW,cAAc,MAAM;AAAA,IAC9E;AAAA,EACF;AACF;AACA,SAAS,wBAAwB,QAAQ,UAAU;AACjD,QAAM,cAAc,IAAI,aAAa,MAAM;AAC3C,QAAM,gBAAgB,IAAI,aAAa,QAAQ;AAC/C,MAAI,YAAY,WAAW,cAAc,QAAQ;AAC/C,UAAM,IAAI,MAAM,wCAAwC,cAAc,sBAAsB,YAAY,QAAQ;AAAA,EAClH;AACA,WAAS,IAAI,GAAG,IAAI,cAAc,QAAQ,KAAK;AAC7C,QAAI,YAAY,OAAO,cAAc,IAAI;AACvC,YAAM,IAAI,MAAM,iCAAiC,WAAW,cAAc,cAAc,YAAY,YAAY;AAAA,IAClH;AAAA,EACF;AACF;AACA,SAAS,cAAc,GAAG;AACxB,WAAS,IAAI,GAAG,IAAI,EAAE,QAAQ,KAAK;AACjC,UAAM,MAAM,EAAE;AACd,QAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,oBAAc,GAAG;AAAA,IACnB,OAAO;AACL,QAAE,KAAK,aAAa,GAAG;AAAA,IACzB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,mBAAmB,QAAQ;AAClC,QAAM,QAAQ,SAAS,cAAc,OAAO;AAC5C,MAAI,iBAAiB,OAAO;AAC1B,UAAM,cAAc;AAAA,EACtB;AACA,QAAM,QAAQ;AACd,QAAM,OAAO;AACb,QAAM,MAAM,WAAW;AACvB,QAAM,MAAM,OAAO;AACnB,QAAM,MAAM,MAAM;AAClB,QAAM,UAAU;AAChB,QAAM,YAAY,MAAM;AACxB,SAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,UAAM,iBAAiB,cAAc,CAAC,MAAM,QAAQ,KAAK,CAAC;AAC1D,UAAM,KAAK;AAAA,EACb,CAAC;AACH;AACA,eAAe,KAAK,OAAO;AACzB,QAAM,MAAM,KAAK;AACjB,MAAI,+BAA+B,OAAO;AACxC,UAAM,IAAI,QAAQ,CAAC,YAAY;AAC7B,YAAM,0BAA0B,OAAO;AAAA,IACzC,CAAC;AAAA,EACH;AACF;AAGA,IAAI,UAAU;AAGd,SAAS,KAAK,GAAG,GAAG;AAClB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,OAAO,GAAG,EAAE,KAAK,CAAC;AAGtB,SAAS,UAAU,GAAG,GAAG;AACvB,MAAI,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC3C,MAAI,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC3C,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,KAAK,GAAG,GAAG;AAClB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,MAAI,GAAG,UAAU,WAAW,GAAG,UAAU,SAAS;AAChD,WAAO,SAAS,IAAI,EAAE;AAAA,EACxB;AACA,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,QAAM,QAAQ,CAAC;AACf,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,GAAG,GAAG;AAClB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,MAAI,GAAG,UAAU,aAAa;AAC5B,UAAM,SAAS,EAAE,GAAG,GAAG;AACvB,WAAO,OAAO,UAAU,YAAY,MAAM;AAAA,EAC5C,OAAO;AACL,UAAM,SAAS,EAAE,GAAG,GAAG;AACvB,WAAO,OAAO,UAAU,KAAK,MAAM;AAAA,EACrC;AACF;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,SAAS;AACtB,SAAO,MAAM,QAAQ,OAAO,GAAG,MAAM,4DAA4D;AACjG,SAAO,QAAQ,UAAU,GAAG,MAAM,uDAAuD,QAAQ,QAAQ;AACzG,QAAM,WAAW,QAAQ,IAAI,CAAC,GAAG,MAAM,gBAAgB,GAAG,UAAU,KAAK,MAAM,CAAC;AAChF,QAAM,cAAc,SAAS;AAC7B,WAAS,QAAQ,CAAC,MAAM;AACtB,QAAI,EAAE,UAAU,YAAY,OAAO;AACjC,YAAM,IAAI,MAAM,0DAA0D;AAAA,IAC5E;AAAA,EACF,CAAC;AACD,WAAS,QAAQ,CAAC,MAAM;AACtB,QAAI,CAAC,YAAY,EAAE,OAAO,YAAY,KAAK,GAAG;AAC5C,YAAM,IAAI,MAAM,0DAA0D;AAAA,IAC5E;AAAA,EACF,CAAC;AACD,QAAM,SAAS;AACf,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,KAAK,GAAG,OAAO,MAAM,WAAW,OAAO;AAC9C,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,MAAM;AAChD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC5C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,GAAG,OAAO,MAAM,WAAW,OAAO;AAC9C,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,MAAM;AAChD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC5C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,QAAQ,GAAG,OAAO,GAAG;AAC5B,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAC/C;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,QAAQ,GAAG,OAAO,GAAG;AAC5B,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAC/C;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG,GAAG;AACpB,MAAI,KAAK,gBAAgB,GAAG,KAAK,OAAO;AACxC,MAAI,KAAK,gBAAgB,GAAG,KAAK,OAAO;AACxC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,sBAAsB,YAAY,aAAa,SAAS,MAAM,aAAa,QAAQ,WAAW;AACrG,QAAM,gBAAgB,WAAW;AACjC,QAAM,eAAe,CAAC,GAAG,aAAa,aAAa;AACnD,QAAM,cAAc,wBAAwB,UAAU;AACtD,SAAO,kBAAkB,YAAY,cAAc,SAAS,WAAW,MAAM,MAAM,MAAM,WAAW;AACtG;AACA,SAAS,kBAAkB,SAAS,YAAY,SAAS,WAAW,MAAM,cAAc,aAAa,gBAAgB;AACnH,QAAM,CAAC,cAAc,WAAW,IAAI,gBAAgB,UAAU;AAC9D,MAAI;AACJ,MAAI,eAAe,gBAAgB;AACjC,kBAAc,CAAC,cAAc,aAAa,QAAQ,IAAI,QAAQ,EAAE;AAAA,EAClE,WAAW,eAAe,iBAAiB;AACzC,kBAAc,CAAC,cAAc,aAAa,QAAQ,IAAI,QAAQ,EAAE;AAAA,EAClE,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB,YAAY;AAAA,EACpD;AACA,SAAO,kBAAkB,SAAS,aAAa,SAAS,WAAW,MAAM,cAAc,OAAO,UAAU;AAC1G;AACA,SAAS,kBAAkB,SAAS,YAAY,SAAS,WAAW,MAAM,cAAc,aAAa,SAAS;AAC5G,QAAM,CAAC,aAAa,cAAc,WAAW,IAAI,iBAAiB,UAAU;AAC5E,MAAI;AACJ,MAAI;AACJ,MAAI,eAAe,SAAS;AAC1B,kBAAc;AACd,kBAAc,CAAC,aAAa,cAAc,aAAa,QAAQ,IAAI,QAAQ,EAAE;AAAA,EAC/E,WAAW,eAAe,SAAS;AACjC,kBAAc;AACd,kBAAc,CAAC,aAAa,cAAc,aAAa,QAAQ,IAAI,QAAQ,EAAE;AAAA,EAC/E,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB,YAAY;AAAA,EACpD;AACA,SAAO,kBAAkB,SAAS,aAAa,SAAS,WAAW,MAAM,OAAO,aAAa,YAAY;AAC3G;AACA,SAAS,kBAAkB,SAAS,aAAa,SAAS,WAAW,MAAM,cAAc,YAAY,OAAO,aAAa,gBAAgB;AACvI,MAAI,CAAC,WAAW,UAAU,SAAS,UAAU,IAAI,CAAC,IAAI,IAAI,IAAI,EAAE;AAChE,MAAI,eAAe,gBAAgB;AACjC,KAAC,WAAW,UAAU,SAAS,UAAU,IAAI;AAAA,EAC/C,WAAW,eAAe,iBAAiB;AACzC,KAAC,WAAW,YAAY,UAAU,OAAO,IAAI;AAAA,EAC/C,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB,YAAY;AAAA,EACpD;AACA,QAAM,CAAC,cAAc,aAAa,EAAE,cAAc,IAAI;AACtD,QAAM,CAAC,cAAc,WAAW,IAAI,gBAAgB,OAAO;AAC3D,QAAM,CAAC,gBAAgB,aAAa,IAAI,gBAAgB,SAAS;AACjE,QAAM,wBAAwB,uBAAuB,cAAc,cAAc;AACjF,QAAM,uBAAuB,uBAAuB,aAAa,aAAa;AAC9E,QAAM,EAAE,SAAS,WAAW,SAAS,IAAI,iBAAiB,MAAM,UAAU,SAAS,cAAc,aAAa,uBAAuB,sBAAsB,cAAc,UAAU;AACnL,QAAM,cAAc,YAAY,iBAAiB,aAAa;AAC9D,MAAI;AACJ,MAAI,eAAe,iBAAiB;AAClC,eAAW,CAAC,WAAW,aAAa,WAAW,QAAQ;AAAA,EACzD,WAAW,eAAe,gBAAgB;AACxC,eAAW,CAAC,WAAW,WAAW,UAAU,WAAW;AAAA,EACzD;AACA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,SAAS,kBAAkB,SAAS,aAAa,SAAS,WAAW,MAAM,YAAY,OAAO,aAAa,gBAAgB,cAAc;AACvI,MAAI,CAAC,WAAW,SAAS,UAAU,SAAS,UAAU,IAAI,CAAC,IAAI,IAAI,IAAI,IAAI,EAAE;AAC7E,MAAI,eAAe,gBAAgB;AACjC,KAAC,WAAW,SAAS,UAAU,SAAS,UAAU,IAAI;AAAA,EACxD,WAAW,eAAe,iBAAiB;AACzC,KAAC,WAAW,YAAY,SAAS,UAAU,OAAO,IAAI;AAAA,EACxD,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB,YAAY;AAAA,EACpD;AACA,QAAM,CAAC,aAAa,cAAc,aAAa,EAAE,cAAc,IAAI;AACnE,QAAM,CAAC,aAAa,cAAc,WAAW,IAAI,iBAAiB,OAAO;AACzE,QAAM,CAAC,eAAe,gBAAgB,aAAa,IAAI,iBAAiB,SAAS;AACjF,QAAM,uBAAuB,uBAAuB,aAAa,aAAa;AAC9E,QAAM,wBAAwB,uBAAuB,cAAc,cAAc;AACjF,QAAM,uBAAuB,uBAAuB,aAAa,aAAa;AAC9E,QAAM,EAAE,SAAS,UAAU,WAAW,SAAS,IAAI,mBAAmB,MAAM,SAAS,UAAU,SAAS,aAAa,cAAc,aAAa,sBAAsB,uBAAuB,sBAAsB,YAAY;AAC/N,QAAM,cAAc,YAAY,iBAAiB,aAAa;AAC9D,MAAI;AACJ,MAAI,eAAe,iBAAiB;AAClC,eAAW,CAAC,WAAW,aAAa,UAAU,WAAW,QAAQ;AAAA,EACnE,WAAW,eAAe,gBAAgB;AACxC,eAAW,CAAC,WAAW,UAAU,WAAW,UAAU,WAAW;AAAA,EACnE;AACA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,SAAS,qBAAqB,SAAS,WAAW,QAAQ,SAAS,cAAc;AAC/E,MAAI,WAAW,MAAM;AACnB,cAAU,kBAAkB,SAAS,WAAW,MAAM;AAAA,EACxD;AACA,QAAM,YAAY,QAAQ;AAC1B,QAAM,YAAY,QAAQ;AAC1B,QAAM,aAAa,OAAO,YAAY,YAAY,IAAI,WAAW,SAAS,GAAG,YAAY;AACzF,QAAM,aAAa,OAAO,YAAY,YAAY,IAAI,WAAW,SAAS,GAAG,YAAY;AACzF,SAAO,CAAC,YAAY,UAAU;AAChC;AACA,SAAS,qBAAqB,SAAS,WAAW,aAAa,QAAQ,SAAS,cAAc;AAC5F,MAAI,WAAW,MAAM;AACnB,cAAU,kBAAkB,SAAS,WAAW,MAAM;AAAA,EACxD;AACA,QAAM,aAAa,QAAQ;AAC3B,QAAM,YAAY,QAAQ;AAC1B,QAAM,YAAY,QAAQ;AAC1B,QAAM,eAAe,OAAO,aAAa,YAAY,IAAI,WAAW,SAAS,GAAG,YAAY;AAC5F,QAAM,aAAa,OAAO,YAAY,YAAY,IAAI,WAAW,SAAS,GAAG,YAAY;AACzF,QAAM,aAAa,OAAO,YAAY,YAAY,IAAI,WAAW,SAAS,GAAG,YAAY;AACzF,SAAO,CAAC,cAAc,YAAY,YAAY,WAAW;AAC3D;AACA,SAAS,kBAAkB,YAAY,WAAW,QAAQ,WAAW,GAAG;AACtE,QAAM,qBAAqB,uBAAuB,WAAW,QAAQ;AACrE,SAAO,KAAK,OAAO,WAAW,MAAM,SAAS,KAAK,SAAS,sBAAsB,CAAC;AACpF;AACA,SAAS,gBAAgB,OAAO;AAC9B,MAAI,OAAO,UAAU,UAAU;AAC7B,WAAO,CAAC,OAAO,OAAO,KAAK;AAAA,EAC7B;AACA,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO,CAAC,MAAM,IAAI,MAAM,IAAI,CAAC;AAAA,EAC/B;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,OAAO;AAC/B,SAAO,OAAO,UAAU,WAAW,CAAC,OAAO,OAAO,KAAK,IAAI;AAC7D;AACA,SAAS,uBAAuB,YAAY,UAAU;AACpD,MAAI,YAAY,GAAG;AACjB,WAAO;AAAA,EACT;AACA,SAAO,cAAc,aAAa,MAAM,WAAW;AACrD;AACA,SAAS,iBAAiB,MAAM,UAAU,SAAS,cAAc,aAAa,cAAc,aAAa,cAAc,YAAY;AACjI,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI,OAAO,SAAS,UAAU;AAC5B,UAAM,UAAU,SAAS,IAAI,UAAU;AACvC,cAAU,EAAE,KAAK,MAAM,QAAQ,MAAM,MAAM,MAAM,OAAO,MAAM,MAAM,QAAQ;AAC5E,UAAM,WAAW,qBAAqB,CAAC,UAAU,OAAO,GAAG,cAAc,cAAc,MAAM,YAAY;AACzG,gBAAY,SAAS;AACrB,eAAW,SAAS;AAAA,EACtB,WAAW,SAAS,QAAQ;AAC1B,gBAAY,KAAK,KAAK,WAAW,YAAY;AAC7C,eAAW,KAAK,KAAK,UAAU,WAAW;AAC1C,UAAM,iBAAiB,KAAK,IAAI,IAAI,YAAY,KAAK,eAAe,eAAe,QAAQ;AAC3F,UAAM,gBAAgB,KAAK,IAAI,IAAI,WAAW,KAAK,cAAc,cAAc,OAAO;AACtF,UAAM,MAAM,KAAK,MAAM,iBAAiB,CAAC;AACzC,UAAM,SAAS,iBAAiB;AAChC,UAAM,OAAO,KAAK,MAAM,gBAAgB,CAAC;AACzC,UAAM,QAAQ,gBAAgB;AAC9B,cAAU,EAAE,KAAK,QAAQ,MAAM,OAAO,MAAM,OAAO;AAAA,EACrD,WAAW,SAAS,SAAS;AAC3B,cAAU,EAAE,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,MAAM,QAAQ;AAChE,gBAAY,KAAK,MAAM,WAAW,eAAe,KAAK,YAAY;AAClE,eAAW,KAAK,MAAM,UAAU,cAAc,KAAK,WAAW;AAAA,EAChE,WAAW,OAAO,SAAS,UAAU;AACnC,UAAM,MAAM,eAAe,iBAAiB,KAAK,GAAG,KAAK,KAAK,GAAG;AACjE,UAAM,SAAS,eAAe,iBAAiB,KAAK,GAAG,KAAK,KAAK,GAAG;AACpE,UAAM,OAAO,eAAe,iBAAiB,KAAK,GAAG,KAAK,KAAK,GAAG;AAClE,UAAM,QAAQ,eAAe,iBAAiB,KAAK,GAAG,KAAK,KAAK,GAAG;AACnE,UAAM,UAAU,QAAQ,KAAK,WAAW,KAAK,SAAS,KAAK,UAAU,IAAI,UAAU;AACnF,cAAU,EAAE,KAAK,QAAQ,MAAM,OAAO,MAAM,QAAQ;AACpD,gBAAY,OAAO,WAAW,eAAe,MAAM,UAAU,eAAe,GAAG,YAAY;AAC3F,eAAW,OAAO,UAAU,cAAc,OAAO,SAAS,cAAc,GAAG,YAAY;AAAA,EACzF,OAAO;AACL,UAAM,MAAM,8BAA8B,MAAM;AAAA,EAClD;AACA,SAAO,EAAE,SAAS,WAAW,SAAS;AACxC;AACA,SAAS,mBAAmB,MAAM,SAAS,UAAU,SAAS,aAAa,cAAc,aAAa,aAAa,cAAc,aAAa,cAAc;AAC1J,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI,OAAO,SAAS,UAAU;AAC5B,UAAM,UAAU,SAAS,IAAI,UAAU;AACvC,cAAU;AAAA,MACR,KAAK;AAAA,MACL,QAAQ;AAAA,MACR,MAAM;AAAA,MACN,OAAO;AAAA,MACP,OAAO;AAAA,MACP,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AACA,UAAM,WAAW,qBAAqB,CAAC,SAAS,UAAU,SAAS,CAAC,GAAG,aAAa,GAAG,aAAa,MAAM,YAAY;AACtH,eAAW,SAAS;AACpB,gBAAY,SAAS;AACrB,eAAW,SAAS;AAAA,EACtB,WAAW,SAAS,QAAQ;AAC1B,eAAW,KAAK,KAAK,UAAU,WAAW;AAC1C,gBAAY,KAAK,KAAK,WAAW,YAAY;AAC7C,eAAW,KAAK,KAAK,UAAU,WAAW;AAC1C,UAAM,iBAAiB,WAAW,KAAK,cAAc,cAAc;AACnE,UAAM,kBAAkB,YAAY,KAAK,eAAe,eAAe;AACvE,UAAM,iBAAiB,WAAW,KAAK,cAAc,cAAc;AACnE,UAAM,QAAQ,KAAK,MAAM,gBAAgB,CAAC;AAC1C,UAAM,OAAO,gBAAgB;AAC7B,UAAM,MAAM,KAAK,MAAM,iBAAiB,CAAC;AACzC,UAAM,SAAS,iBAAiB;AAChC,UAAM,OAAO,KAAK,MAAM,gBAAgB,CAAC;AACzC,UAAM,QAAQ,gBAAgB;AAC9B,cAAU,EAAE,KAAK,QAAQ,MAAM,OAAO,OAAO,MAAM,MAAM,OAAO;AAAA,EAClE,WAAW,SAAS,SAAS;AAC3B,cAAU;AAAA,MACR,KAAK;AAAA,MACL,QAAQ;AAAA,MACR,MAAM;AAAA,MACN,OAAO;AAAA,MACP,OAAO;AAAA,MACP,MAAM;AAAA,MACN,MAAM;AAAA,IACR;AACA,eAAW,KAAK,MAAM,UAAU,cAAc,KAAK,WAAW;AAC9D,gBAAY,KAAK,MAAM,WAAW,eAAe,KAAK,YAAY;AAClE,eAAW,KAAK,MAAM,UAAU,cAAc,KAAK,WAAW;AAAA,EAChE,OAAO;AACL,UAAM,MAAM,8BAA8B,MAAM;AAAA,EAClD;AACA,SAAO,EAAE,SAAS,UAAU,WAAW,SAAS;AAClD;AACA,SAAS,MAAM,OAAO,cAAc;AAClC,MAAI,CAAC,cAAc;AACjB,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AACA,UAAQ,cAAc;AAAA,IACpB,KAAK;AACH,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB,KAAK;AACH,aAAO,KAAK,KAAK,KAAK;AAAA,IACxB,KAAK;AACH,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACE,YAAM,IAAI,MAAM,wBAAwB,cAAc;AAAA,EAC1D;AACF;AACA,SAAS,kBAAkB,OAAO;AAChC,QAAM,CAAC,MAAM,MAAM,IAAI,IAAI,gBAAgB,KAAK;AAChD,SAAO,SAAS,KAAK,SAAS,KAAK,SAAS;AAC9C;AACA,SAAS,+BAA+B,SAAS,WAAW;AAC1D,SAAO,kBAAkB,OAAO,KAAK,kBAAkB,SAAS;AAClE;AACA,SAAS,wBAAwB,YAAY;AAC3C,MAAI,eAAe,QAAQ;AACzB,WAAO;AAAA,EACT,WAAW,eAAe,QAAQ;AAChC,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB,YAAY;AAAA,EACpD;AACF;AACA,SAAS,0BAA0B,QAAQ,MAAM,iBAAiB;AAChE,MAAI,mBAAmB,MAAM;AAC3B,QAAI,OAAO,SAAS,UAAU;AAC5B,YAAM,MAAM,YAAY,6DAA6D,+BAA+B,OAAO;AAAA,IAC7H,WAAW,OAAO,SAAS,UAAU;AACnC,aAAO,MAAM,IAAI,GAAG,MAAM,YAAY,6DAA6D,+BAA+B,OAAO;AAAA,IAC3I,WAAW,OAAO,SAAS,UAAU;AACnC,WAAK,QAAQ,CAAC,OAAO;AACnB,WAAG,QAAQ,CAAC,MAAM;AAChB,iBAAO,MAAM,CAAC,GAAG,MAAM,YAAY,6DAA6D,+BAA+B,IAAI;AAAA,QACrI,CAAC;AAAA,MACH,CAAC;AAAA,IACH,OAAO;AACL,YAAM,MAAM,YAAY,sCAAsC,MAAM;AAAA,IACtE;AAAA,EACF;AACF;AAGA,SAAS,SAAS,GAAG,OAAO;AAC1B,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AACjE,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM;AACtB,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,SAAS,GAAG,YAAY,SAAS,MAAM,iBAAiB;AAC/D,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,SAAS;AACvD,QAAM,YAAY;AAClB,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AAC/K,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,mDAAmD,IAAI,OAAO;AAC3F,4BAA0B,WAAW,MAAM,eAAe;AAC1D,QAAM,SAAS,EAAE,GAAG,IAAI;AACxB,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB;AAChE,MAAI,MAAM,OAAO,UAAU,SAAS,QAAQ,KAAK;AACjD,QAAM,KAAK,KAAK,GAAG,KAAK;AACxB,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,GAAG,YAAY,SAAS,MAAM,iBAAiB,aAAa,SAAS;AACvF,QAAM,KAAK,gBAAgB,GAAG,KAAK,aAAa,SAAS;AACzD,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC3E;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,qDAAqD,IAAI,OAAO;AAC7F,SAAO,eAAe,SAAS,MAAM,gFAAgF,YAAY;AACjI,4BAA0B,aAAa,MAAM,eAAe;AAC5D,QAAM,SAAS,EAAE,GAAG,IAAI;AACxB,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,iBAAiB,WAAW;AAC5E,MAAI,MAAM,OAAO,UAAU,WAAW,QAAQ,KAAK;AACnD,QAAM,KAAK,KAAK,IAAI,KAAK;AACzB,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,QAAQ,SAAS,OAAO,GAAG;AAClC,SAAO,QAAQ,UAAU,GAAG,MAAM,oCAAoC;AACtE,QAAM,WAAW,qBAAqB,SAAS,WAAW,UAAU,mBAAmB;AACvF,MAAI,SAAS,GAAG,UAAU,aAAa;AACrC,aAAS,QAAQ,CAAC,YAAY;AAC5B,UAAI,QAAQ,UAAU,aAAa;AACjC,cAAM,IAAI,MAAM;AAAA,uBACD,QAAQ,SAAS;AAAA,MAClC;AAAA,IACF,CAAC;AAAA,EACH;AACA,MAAI,SAAS,WAAW,GAAG;AACzB,WAAO,MAAM,SAAS,EAAE;AAAA,EAC1B;AACA,QAAM,SAAS;AACf,QAAM,OAAO,EAAE,KAAK;AACpB,SAAO,OAAO,UAAU,QAAQ,QAAQ,IAAI;AAC9C;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,SAAS,GAAG;AACnB,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,SAAS;AACvD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,SAAS,MAAM;AACzC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,OAAO,GAAG,OAAO,MAAM;AAC9B,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS,mBAAmB;AAC/D,MAAI,GAAG,SAAS,GAAG;AACjB,UAAM,IAAI,MAAM,gCAAgC;AAAA,EAClD;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,OAAO,KAAK;AAC5B,SAAO,OAAO,UAAU,OAAO,QAAQ,KAAK;AAC9C;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ,SAAS;AACpD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,QAAQ,GAAG,EAAE,MAAM,CAAC;AAGxB,SAAS,eAAe,YAAY,YAAY,UAAU,MAAM,GAAG,GAAG;AACpE,QAAM,cAAc,gBAAgB,YAAY,cAAc,eAAe;AAC7E,QAAM,cAAc,gBAAgB,YAAY,cAAc,eAAe;AAC7E,QAAM,YAAY,gBAAgB,UAAU,YAAY,eAAe;AACvE,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,eAAe;AAC3D,QAAM,KAAK,gBAAgB,GAAG,KAAK,eAAe;AAClD,QAAM,KAAK,gBAAgB,GAAG,KAAK,eAAe;AAClD,QAAM,WAAW,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC;AACtC,QAAM,WAAW,OAAO,UAAU,WAAW;AAC7C,QAAM,MAAM,KAAK,UAAU,SAAS;AACpC,QAAM,YAAY,IAAI,MAAM;AAC5B,QAAM,YAAY,IAAI,MAAM,KAAK;AACjC,QAAM,YAAY,CAAC,WAAW,SAAS;AACvC,QAAM,IAAI,MAAM,KAAK,CAAC,GAAG,CAAC,GAAG,SAAS;AACtC,QAAM,IAAI,MAAM,KAAK,CAAC,GAAG,SAAS,GAAG,SAAS;AAC9C,QAAM,IAAI,MAAM,KAAK,CAAC,GAAG,YAAY,CAAC,GAAG,SAAS;AAClD,QAAM,IAAI,MAAM,KAAK,CAAC,GAAG,YAAY,CAAC,GAAG,SAAS;AAClD,QAAM,OAAO,KAAK,IAAI,QAAQ,CAAC,GAAG,MAAM,CAAC,CAAC,GAAG,IAAI,IAAI,QAAQ,KAAK,aAAa,CAAC,CAAC,CAAC,CAAC;AACnF,QAAM,OAAO,IAAI,MAAM,IAAI,GAAG,QAAQ,CAAC,CAAC;AACxC,SAAO,CAAC,MAAM,IAAI;AACpB;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,gBAAgB,GAAG,YAAY,OAAO;AAC7C,QAAM,KAAK,gBAAgB,GAAG,KAAK,gBAAgB;AACnD,QAAM,QAAQ,WAAW,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAC/C,SAAO,GAAG,QAAQ,IAAI,WAAW,QAAQ,MAAM,iBAAiB,GAAG,+CAA+C,WAAW,QAAQ;AACrI,SAAO,MAAM,WAAW,WAAW,QAAQ,MAAM,mBAAmB,MAAM,oDAAoD,WAAW,QAAQ;AACjJ,SAAO,GAAG,MAAM,KAAK,UAAU,GAAG,MAAM,yBAAyB,GAAG,MAAM,wEAAwE,WAAW,KAAK,KAAK,SAAS,OAAO;AACvL,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,YAAY,MAAM;AAClC,SAAO,OAAO,UAAU,gBAAgB,QAAQ,KAAK;AACvD;AACA,IAAI,iBAAiB,GAAG,EAAE,gBAAgB,CAAC;AAG3C,SAAS,MAAM,GAAG;AAChB,MAAI;AACJ,MAAI,EAAE,SAAS,KAAK,EAAE,SAAS,GAAG;AAChC,UAAM,QAAQ,GAAG,CAAC,GAAG,GAAG,GAAG,EAAE,IAAI,CAAC;AAAA,EACpC,WAAW,EAAE,SAAS,GAAG;AACvB,UAAM,QAAQ,GAAG,CAAC,GAAG,GAAG,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC;AAAA,EACjD,WAAW,EAAE,SAAS,GAAG;AACvB,UAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC;AAAA,EAC1D,OAAO;AACL,UAAM;AAAA,EACR;AACA,SAAO;AACT;AAGA,SAAS,WAAW,GAAG,OAAO,UAAU,QAAQC,SAAQ,iBAAiB;AACvE,MAAI,mBAAmB,MAAM;AAC3B,sBAAkB;AAAA,EACpB;AACA,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,WAAW;AACxD,QAAM,YAAY,gBAAgB,UAAU,YAAY,WAAW;AACnE,MAAI;AACJ,MAAIA,WAAU,MAAM;AAClB,aAAS,gBAAgBA,SAAQ,SAAS,WAAW;AAAA,EACvD;AACA,MAAI;AACJ,MAAI,UAAU,MAAM;AAClB,cAAU,gBAAgB,QAAQ,UAAU,WAAW;AAAA,EACzD;AACA,SAAO,MAAM,SAAS,UAAU,MAAM,MAAM,8EAA8E;AAC1H,SAAO,WAAW,QAAQ,MAAM,SAAS,QAAQ,MAAM,MAAM,4EAA4E;AACzI,SAAO,UAAU,QAAQ,MAAM,SAAS,OAAO,MAAM,MAAM,2EAA2E;AACtI,QAAM,MAAM,MAAM,EAAE;AACpB,QAAM,SAAS;AAAA,IACb,GAAG;AAAA,IACH,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,UAAU;AAAA,EACZ;AACA,QAAM,QAAQ,EAAE,gBAAgB;AAChC,QAAM,MAAM,OAAO,UAAU,gBAAgB,QAAQ,KAAK;AAC1D,SAAO,QAAQ,KAAK,GAAG,KAAK;AAC9B;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,aAAa,GAAG,OAAO,UAAU,QAAQA,SAAQ,iBAAiB;AACzE,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,WAAW;AACxD,QAAM,YAAY,gBAAgB,UAAU,YAAY,WAAW;AACnE,MAAI;AACJ,MAAIA,WAAU,MAAM;AAClB,aAAS,gBAAgBA,SAAQ,SAAS,WAAW;AAAA,EACvD;AACA,MAAI;AACJ,MAAI,UAAU,MAAM;AAClB,cAAU,gBAAgB,QAAQ,UAAU,WAAW;AAAA,EACzD;AACA,SAAO,GAAG,SAAS,GAAG,MAAM,uDAAuD,GAAG,OAAO;AAC7F,SAAO,MAAM,SAAS,KAAK,MAAM,SAAS,GAAG,MAAM,oEAAoE,MAAM,OAAO;AACpI,SAAO,UAAU,SAAS,KAAK,UAAU,SAAS,GAAG,MAAM,wEAAwE,UAAU,OAAO;AACpJ,MAAI,UAAU,MAAM;AAClB,WAAO,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG,MAAM,qEAAqE,OAAO,OAAO;AAAA,EAC1I;AACA,MAAI,WAAW,MAAM;AACnB,WAAO,QAAQ,SAAS,KAAK,QAAQ,SAAS,GAAG,MAAM,sEAAsE,QAAQ,OAAO;AAAA,EAC9I;AACA,SAAO,UAAU,IAAI,OAAO,WAAW,SAAS,QAAQ,eAAe;AACzE;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,aAAa,GAAG,OAAO,UAAU,QAAQA,SAAQ,iBAAiB;AACzE,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,WAAW;AACxD,QAAM,YAAY,gBAAgB,UAAU,YAAY,WAAW;AACnE,MAAI;AACJ,MAAIA,WAAU,MAAM;AAClB,aAAS,gBAAgBA,SAAQ,SAAS,WAAW;AAAA,EACvD;AACA,MAAI;AACJ,MAAI,UAAU,MAAM;AAClB,cAAU,gBAAgB,QAAQ,UAAU,WAAW;AAAA,EACzD;AACA,SAAO,GAAG,SAAS,GAAG,MAAM,uDAAuD,GAAG,OAAO;AAC7F,SAAO,MAAM,SAAS,KAAK,MAAM,SAAS,GAAG,MAAM,oEAAoE,MAAM,OAAO;AACpI,SAAO,UAAU,SAAS,KAAK,UAAU,SAAS,GAAG,MAAM,wEAAwE,UAAU,OAAO;AACpJ,MAAI,UAAU,MAAM;AAClB,WAAO,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG,MAAM,qEAAqE,OAAO,OAAO;AAAA,EAC1I;AACA,MAAI,WAAW,MAAM;AACnB,WAAO,QAAQ,SAAS,KAAK,QAAQ,SAAS,GAAG,MAAM,sEAAsE,QAAQ,OAAO;AAAA,EAC9I;AACA,SAAO,UAAU,IAAI,OAAO,WAAW,SAAS,QAAQ,eAAe;AACzE;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,aAAa,GAAG,OAAO,UAAU,QAAQA,SAAQ,iBAAiB;AACzE,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,QAAQ,gBAAgB,OAAO,QAAQ,WAAW;AACxD,QAAM,YAAY,gBAAgB,UAAU,YAAY,WAAW;AACnE,MAAI;AACJ,MAAIA,WAAU,MAAM;AAClB,aAAS,gBAAgBA,SAAQ,SAAS,WAAW;AAAA,EACvD;AACA,MAAI;AACJ,MAAI,UAAU,MAAM;AAClB,cAAU,gBAAgB,QAAQ,UAAU,WAAW;AAAA,EACzD;AACA,SAAO,GAAG,SAAS,GAAG,MAAM,uDAAuD,GAAG,OAAO;AAC7F,SAAO,MAAM,SAAS,KAAK,MAAM,SAAS,GAAG,MAAM,oEAAoE,MAAM,OAAO;AACpI,SAAO,UAAU,SAAS,KAAK,UAAU,SAAS,GAAG,MAAM,wEAAwE,UAAU,OAAO;AACpJ,MAAI,UAAU,MAAM;AAClB,WAAO,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG,MAAM,qEAAqE,OAAO,OAAO;AAAA,EAC1I;AACA,MAAI,WAAW,MAAM;AACnB,WAAO,QAAQ,SAAS,KAAK,QAAQ,SAAS,GAAG,MAAM,sEAAsE,QAAQ,OAAO;AAAA,EAC9I;AACA,SAAO,UAAU,IAAI,OAAO,WAAW,SAAS,QAAQ,eAAe;AACzE;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,UAAU,GAAG,SAAS,MAAM;AACnC,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC7C,QAAM,WAAW,gBAAgB,SAAS,WAAW,UAAU;AAC/D,SAAO,GAAG,UAAU,SAAS,MAAM,yDAAyD,GAAG,OAAO;AACtG,SAAO,QAAQ,GAAG,MAAM,sCAAsC,OAAO;AACrE,SAAO,SAAS,SAAS,GAAG,QAAQ,SAAS,SAAS,GAAG,MAAM,gGAAgG,GAAG,yBAAyB,SAAS,QAAQ;AAC5M,QAAM,SAAS,EAAE,GAAG,IAAI,SAAS,SAAS;AAC1C,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,UAAU,QAAQ,KAAK;AACjD;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,eAAe,IAAI,IAAI;AAC9B,QAAM,cAAc,gBAAgB,IAAI,MAAM,iBAAiB,OAAO;AACtE,QAAM,cAAc,gBAAgB,IAAI,MAAM,iBAAiB,OAAO;AACtE,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,oEAAoE,YAAY,MAAM;AAAA,EACxG;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,qEAAqE,YAAY,MAAM;AAAA,EACzG;AACA,QAAM,SAAS,EAAE,IAAI,aAAa,IAAI,YAAY;AAClD,SAAO,OAAO,UAAU,eAAe,MAAM;AAC/C;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,aAAa,GAAG,OAAO;AAC9B,MAAI,SAAS,gBAAgB,GAAG,eAAe,GAAG;AAClD,QAAM,SAAS,OAAO;AACtB,MAAI,MAAM,KAAK,CAAC,MAAM,EAAE,IAAI,MAAM,IAAI,MAAM,CAAC,GAAG;AAC9C,UAAM,IAAI,MAAM,2CAA2C,SAAS;AAAA,EACtE;AACA,MAAI,MAAM,SAAS,OAAO,MAAM;AAC9B,UAAM,IAAI,MAAM,+BAA+B,MAAM,uBAAuB,OAAO,OAAO;AAAA,EAC5F;AACA,MAAI,MAAM,SAAS,OAAO,MAAM;AAC9B,UAAM,WAAW,OAAO,MAAM,MAAM;AACpC,WAAO,SAAS,SAAS,MAAM,QAAQ;AACrC,eAAS,QAAQ,CAAC;AAAA,IACpB;AACA,aAAS,QAAQ,QAAQ,QAAQ;AAAA,EACnC;AACA,QAAM,aAAa,OAAO;AAC1B,QAAM,OAAO,MAAM,KAAK,KAAK;AAC7B,WAAS,IAAI,MAAM,SAAS,GAAG,KAAK,GAAG,KAAK;AAC1C,QAAI,WAAW,OAAO,MAAM,IAAI;AAC9B,WAAK,KAAK;AAAA,IACZ,WAAW,OAAO,MAAM,OAAO,GAAG;AAChC,YAAM,IAAI,MAAM,mBAAmB,mCAAmC,SAAS;AAAA,IACjF;AAAA,EACF;AACA,QAAM,OAAO,KAAK,IAAI,CAAC,GAAG,MAAM,IAAI,IAAI,IAAI,EAAE,EAAE,OAAO,CAAC,MAAM,KAAK,CAAC;AACpE,MAAI,KAAK,WAAW,GAAG;AACrB,WAAO,MAAM,MAAM;AAAA,EACrB;AACA,QAAM,SAAS,EAAE,GAAG,OAAO;AAC3B,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ,SAAS;AACpD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,KAAK,OAAO,OAAO,OAAO;AACjC,QAAM,QAAQ,EAAE,OAAO,OAAO,MAAM;AACpC,SAAO,OAAO,UAAU,MAAM,CAAC,GAAG,KAAK;AACzC;AAGA,SAAS,aAAa,GAAG,cAAc,cAAc;AACnD,QAAM,KAAK,gBAAgB,GAAG,KAAK,aAAa;AAChD,SAAO,gBAAgB,cAAc,MAAM,uBAAuB,oDAAoD,gBAAgB;AACtI,MAAI,iBAAiB,cAAc;AACjC,WAAO,KAAK,GAAG,OAAO,cAAc,GAAG,KAAK;AAAA,EAC9C;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,cAAc,aAAa;AAC3C,SAAO,OAAO,UAAU,aAAa,QAAQ,KAAK;AACpD;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,UAAU,SAAS;AAC1B,SAAO,OAAO,SAAS,CAAC;AAC1B;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,UAAU,SAAS,MAAM;AAChC,SAAO,OAAO,SAAS,IAAI;AAC7B;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,UAAU,SAAS,MAAM;AAChC,SAAO,OAAO,SAAS,IAAI;AAC7B;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,UAAU,SAAS,MAAM;AAChC,SAAO,OAAO,SAAS,IAAI;AAC7B;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,QAAQ,GAAG,QAAQ,SAAS,MAAM,aAAa,QAAQ,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB;AACnG,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU,SAAS;AACtD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,UAAU,SAAS;AACrE,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,uDAAuD,IAAI,OAAO;AAC/F,SAAO,QAAQ,SAAS,GAAG,MAAM,wDAAwD,QAAQ,OAAO;AACxG,4BAA0B,UAAU,MAAM,eAAe;AACzD,QAAM,UAAU,eAAe,SAAS,IAAI,MAAM,KAAK,IAAI,MAAM;AACjE,SAAO,YAAY,QAAQ,MAAM,IAAI,MAAM,oCAAoC,8CAA8C,QAAQ,MAAM,KAAK;AAChJ,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,uEAAuE,0BAA0B,YAAY;AAC9K,QAAM,SAAS,EAAE,GAAG,KAAK,QAAQ,QAAQ;AACzC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB;AAC3E,QAAM,MAAM,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAClD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,QAAQ,GAAG,QAAQ,QAAQ,MAAM,aAAa,OAAO,WAAW,GAAG,iBAAiB;AAC3F,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,UAAU,gBAAgB,QAAQ,UAAU,QAAQ;AAC1D,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EACjD;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,uDAAuD,IAAI,OAAO;AAC/F,SAAO,QAAQ,SAAS,GAAG,MAAM,wDAAwD,QAAQ,OAAO;AACxG,4BAA0B,UAAU,MAAM,eAAe;AACzD,SAAO,IAAI,MAAM,OAAO,QAAQ,MAAM,IAAI,MAAM,oCAAoC,IAAI,MAAM,yCAAyC,QAAQ,MAAM,KAAK;AAC1J,SAAO,+BAA+B,QAAQ,QAAQ,GAAG,MAAM,oEAAoE,wBAAwB,WAAW;AACtK,SAAO,eAAe,OAAO,MAAM,sCAAsC,iDAAiD;AAC1H,QAAM,WAAW,QAAQ,SAAS,CAAC,GAAG,QAAQ,MAAM,IAAI,QAAQ,MAAM,IAAI,QAAQ,MAAM,EAAE,CAAC;AAC3F,QAAM,UAAU,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,GAAG,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAC1E,QAAM,UAAU,CAAC,GAAG,MAAM;AAC1B,QAAM,YAAY,CAAC,GAAG,QAAQ;AAC9B,QAAM,mBAAmB;AACzB,QAAM,MAAM,OAAO,SAAS,UAAU,SAAS,MAAM,kBAAkB,WAAW,eAAe;AACjG,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAClD;AACA,SAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAChE;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,qBAAqB,QAAQ,IAAI,QAAQ,SAAS,MAAM,aAAa,QAAQ,iBAAiB;AACrG,SAAO,OAAO,WAAW,GAAG,MAAM,MAAM,sBAAsB,OAAO,2BAA2B,GAAG,kBAAkB;AACrH,MAAI,WAAW;AACf,MAAI,OAAO;AACX,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAC7D,eAAW,CAAC,GAAG,OAAO,IAAI,OAAO,IAAI,OAAO,EAAE;AAAA,EAChD;AACA,SAAO,SAAS,WAAW,GAAG,MAAM,qEAAqE,SAAS,SAAS;AAC3H,SAAO,KAAK,SAAS,GAAG,MAAM,4DAA4D,KAAK,MAAM;AACrG,SAAO,OAAO,SAAS,GAAG,MAAM,gEAAgE,OAAO,MAAM;AAC7G,QAAM,UAAU,eAAe,SAAS,SAAS,KAAK,SAAS;AAC/D,QAAM,WAAW,eAAe,SAAS,KAAK,MAAM,KAAK,KAAK,MAAM;AACpE,SAAO,YAAY,OAAO,MAAM,IAAI,MAAM,4CAA4C,8CAA8C,OAAO,MAAM,KAAK;AACtJ,SAAO,aAAa,OAAO,MAAM,IAAI,MAAM,6CAA6C,gDAAgD,OAAO,MAAM,KAAK;AAC1J,4BAA0B,kBAAkB,MAAM,eAAe;AACjE,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO;AAClC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY,iBAAiB,YAAY,SAAS;AACtF,QAAM,MAAM,OAAO,UAAU,qBAAqB,QAAQ,KAAK;AAC/D,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,iBAAiB,GAAG,QAAQ,aAAa,SAAS,MAAM,iBAAiB;AAChF,QAAM,KAAK,gBAAgB,GAAG,KAAK,iBAAiB;AACpD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,iBAAiB;AACnE,SAAO,oBAAoB,aAAa,IAAI,SAAS,SAAS,MAAM,QAAQ,eAAe;AAC7F;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,SAAS,QAAQ,GAAG,QAAQ,SAAS,MAAM,aAAa,SAAS,YAAY,CAAC,GAAG,GAAG,CAAC,GAAG;AACtF,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,UAAU,gBAAgB,QAAQ,UAAU,QAAQ;AAC1D,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC3E;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,uDAAuD,IAAI,OAAO;AAC/F,SAAO,QAAQ,SAAS,GAAG,MAAM,wDAAwD,QAAQ,OAAO;AACxG,SAAO,IAAI,MAAM,OAAO,QAAQ,MAAM,IAAI,MAAM,oCAAoC,IAAI,MAAM,yCAAyC,QAAQ,MAAM,KAAK;AAC1J,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,uEAAuE,0BAA0B,YAAY;AAC9K,SAAO,eAAe,SAAS,MAAM,sCAAsC,mDAAmD;AAC9H,QAAM,SAAS,EAAE,GAAG,KAAK,QAAQ,QAAQ;AACzC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY,UAAU;AAC1D,QAAM,MAAM,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAClD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,qBAAqB,QAAQ,IAAI,QAAQ,SAAS,MAAM;AAC/D,SAAO,OAAO,WAAW,GAAG,MAAM,MAAM,sBAAsB,OAAO,2BAA2B,GAAG,kBAAkB;AACrH,MAAI,WAAW;AACf,MAAI,OAAO;AACX,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAC1E,eAAW,CAAC,GAAG,OAAO,IAAI,OAAO,IAAI,OAAO,IAAI,OAAO,EAAE;AAAA,EAC3D;AACA,QAAM,UAAU,SAAS;AACzB,QAAM,WAAW,KAAK,MAAM;AAC5B,SAAO,SAAS,WAAW,GAAG,MAAM,qEAAqE,SAAS,SAAS;AAC3H,SAAO,KAAK,SAAS,GAAG,MAAM,4DAA4D,KAAK,MAAM;AACrG,SAAO,OAAO,SAAS,GAAG,MAAM,gEAAgE,OAAO,MAAM;AAC7G,SAAO,YAAY,OAAO,MAAM,IAAI,MAAM,4CAA4C,8CAA8C,OAAO,MAAM,KAAK;AACtJ,SAAO,aAAa,OAAO,MAAM,IAAI,MAAM,6CAA6C,gDAAgD,OAAO,MAAM,KAAK;AAC1J,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO;AAClC,QAAM,QAAQ,EAAE,KAAK,MAAM,SAAS,YAAY,SAAS;AACzD,QAAM,MAAM,OAAO,UAAU,uBAAuB,QAAQ,KAAK;AACjE,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,iBAAiB,GAAG,QAAQ,aAAa,SAAS,MAAM;AAC/D,QAAM,KAAK,gBAAgB,GAAG,KAAK,iBAAiB;AACpD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,iBAAiB;AACnE,SAAO,oBAAoB,aAAa,IAAI,SAAS,SAAS,IAAI;AACpE;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,SAAS;AACnD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ,SAAS;AACpD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,SAAS,GAAG,OAAO,GAAG,YAAY,OAAO,WAAW,OAAO;AAClE,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,WAAW,SAAS,SAAS;AACnD,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,QAAQ,GAAG,OAAO,GAAG,YAAY,OAAO,WAAW,OAAO;AACjE,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,WAAW,SAAS,SAAS;AACnD,SAAO,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAC/C;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,eAAe,GAAG,SAAS,MAAM,eAAe,OAAO;AAC9D,QAAM,KAAK,gBAAgB,GAAG,KAAK,eAAe;AAClD,QAAM,WAAW,gBAAgB,SAAS,WAAW,eAAe;AACpE,SAAO,GAAG,UAAU,SAAS,MAAM,8DAA8D,GAAG,OAAO;AAC3G,SAAO,GAAG,QAAQ,GAAG,MAAM,sEAAsE,GAAG,OAAO;AAC3G,SAAO,QAAQ,GAAG,MAAM,sCAAsC,OAAO;AACrE,SAAO,SAAS,SAAS,GAAG,QAAQ,SAAS,SAAS,GAAG,MAAM,+FAA+F,GAAG,yBAAyB,SAAS,QAAQ;AAC3M,QAAM,SAAS,EAAE,GAAG,IAAI,SAAS,SAAS;AAC1C,QAAM,QAAQ,EAAE,MAAM,aAAa;AACnC,SAAO,OAAO,UAAU,eAAe,QAAQ,KAAK;AACtD;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,cAAc,GAAG,WAAW,aAAa,QAAQ;AACxD,QAAM,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,SAAS;AAC5D,QAAM,cAAc,eAAe,SAAS,GAAG,MAAM,KAAK,GAAG,MAAM;AACnE,QAAM,aAAa,eAAe,SAAS,GAAG,MAAM,KAAK,GAAG,MAAM;AAClE,QAAM,aAAa,eAAe,SAAS,GAAG,MAAM,KAAK,GAAG,MAAM;AAClE,SAAO,YAAY,GAAG,MAAM,sDAAsD,WAAW;AAC7F,SAAO,cAAc,aAAa,GAAG,MAAM;AAAA,MACvC,mBAAmB;AAAA,MACnB,GAAG,OAAO;AACd,SAAO,aAAa,aAAa,GAAG,MAAM;AAAA,MACtC,kBAAkB;AAAA,UACd,GAAG,OAAO;AAClB,SAAO,cAAc,YAAY,eAAe,GAAG,MAAM,8CAA8C,YAAY,oBAAoB,gDAAgD,GAAG,OAAO;AACjM,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,WAAW,WAAW;AACtC,SAAO,OAAO,UAAU,cAAc,QAAQ,KAAK;AACrD;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,iBAAiB,GAAG,QAAQ,SAAS,MAAM,aAAa,QAAQ,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB;AAC5G,QAAM,KAAK,gBAAgB,GAAG,KAAK,mBAAmB,SAAS;AAC/D,QAAM,UAAU,gBAAgB,QAAQ,UAAU,mBAAmB,SAAS;AAC9E,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,gEAAgE,IAAI,OAAO;AACxG,SAAO,QAAQ,SAAS,GAAG,MAAM,iEAAiE,QAAQ,OAAO;AACjH,QAAM,aAAa,eAAe,SAAS,IAAI,MAAM,KAAK,IAAI,MAAM;AACpE,SAAO,eAAe,QAAQ,MAAM,IAAI,MAAM,uDAAuD,6DAA6D,QAAQ,MAAM,KAAK;AACrL,4BAA0B,mBAAmB,MAAM,eAAe;AAClE,QAAM,SAAS,EAAE,GAAG,KAAK,QAAQ,QAAQ;AACzC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB;AAC3E,QAAM,MAAM,OAAO,UAAU,uBAAuB,QAAQ,KAAK;AACjE,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,YAAY,GAAG,QAAQ,SAAS,MAAM,YAAY,CAAC,GAAG,CAAC,GAAG,aAAa,QAAQ;AACtF,QAAM,KAAK,gBAAgB,GAAG,KAAK,YAAY;AAC/C,QAAM,UAAU,gBAAgB,QAAQ,UAAU,YAAY;AAC9D,SAAO,GAAG,SAAS,KAAK,GAAG,SAAS,GAAG,MAAM,gEAAgE,GAAG,OAAO;AACvH,SAAO,QAAQ,SAAS,GAAG,MAAM,4DAA4D,QAAQ,OAAO;AAC5G,SAAO,eAAe,QAAQ,MAAM,gFAAgF,YAAY;AAChI,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAC5D,mBAAe;AAAA,EACjB;AACA,QAAM,SAAS,EAAE,GAAG,KAAK,QAAQ,QAAQ;AACzC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,UAAU;AAC9C,QAAM,MAAM,OAAO,UAAU,YAAY,QAAQ,KAAK;AACtD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,OAAO,GAAG,GAAG;AACpB,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS,mBAAmB;AAC7D,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS,mBAAmB;AAC7D,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,WAAW,GAAG,GAAG;AAC/B,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,aAAa,gBAAgB,WAAW,aAAa,SAAS,MAAM;AAC1E,QAAM,iBAAiB,2BAA2B,2BAA2B,WAAW,OAAO,GAAG,KAAK,GAAG,GAAG,KAAK;AAClH,QAAM,wBAAwB,YAAY,YAAY,cAAc;AACpE,QAAM,gBAAgB,YAAY,IAAI,cAAc;AACpD,QAAM,gBAAgB,YAAY,IAAI,cAAc;AACpD,QAAM,SAAS;AAAA,IACb,WAAW;AAAA,IACX,GAAG;AAAA,IACH,GAAG;AAAA,EACL;AACA,SAAO,OAAO,UAAU,QAAQ,MAAM;AACxC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,WAAW,GAAG;AACrB,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,WAAW,MAAM;AAC3C;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,UAAU,GAAG,GAAG;AACvB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,YAAY,IAAI,IAAI,EAAE;AAC5B,QAAM,SAAS,UAAU,SAAS;AAClC,QAAM,cAAc,MAAM,IAAI,MAAM;AACpC,SAAO,MAAM,aAAa,QAAQ,SAAS;AAC7C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,KAAK,IAAI,IAAI;AACpB,QAAM,MAAM,gBAAgB,IAAI,MAAM,KAAK;AAC3C,QAAM,MAAM,gBAAgB,IAAI,MAAM,KAAK;AAC3C,UAAQ,IAAI,SAAS,KAAK,IAAI,SAAS,OAAO,IAAI,SAAS,KAAK,IAAI,SAAS,IAAI,MAAM,+DAA+D,IAAI,YAAY,IAAI,OAAO;AACjL,QAAM,UAAU,IAAI,SAAS,IAAI,IAAI,OAAO,IAAI,MAAM;AACtD,QAAM,UAAU,IAAI,SAAS,IAAI,IAAI,OAAO,IAAI,MAAM;AACtD,SAAO,YAAY,SAAS,MAAM,gEAAgE,eAAe,UAAU;AAC3H,MAAI,IAAI,SAAS,KAAK,IAAI,SAAS,GAAG;AACpC,UAAM,OAAO,QAAQ,KAAK,CAAC,GAAG,EAAE,CAAC;AACjC,UAAM,OAAO,QAAQ,KAAK,CAAC,IAAI,CAAC,CAAC;AACjC,UAAM,OAAO,OAAO,MAAM,IAAI;AAC9B,WAAO,QAAQ,MAAM,CAAC,CAAC;AAAA,EACzB,WAAW,IAAI,SAAS,KAAK,IAAI,SAAS,GAAG;AAC3C,UAAM,OAAO,QAAQ,KAAK,CAAC,GAAG,EAAE,CAAC;AACjC,UAAM,OAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AACtD,UAAM,OAAO,OAAO,MAAM,IAAI;AAC9B,WAAO,QAAQ,MAAM,CAAC,KAAK,IAAI,CAAC;AAAA,EAClC,WAAW,IAAI,SAAS,KAAK,IAAI,SAAS,GAAG;AAC3C,UAAM,OAAO,QAAQ,KAAK,CAAC,IAAI,CAAC,CAAC;AACjC,UAAM,OAAO,OAAO,KAAK,IAAI;AAC7B,WAAO,QAAQ,MAAM,CAAC,KAAK,IAAI,CAAC;AAAA,EAClC,OAAO;AACL,UAAM,OAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AACtD,UAAM,OAAO,OAAO,KAAK,IAAI;AAC7B,WAAO;AAAA,EACT;AACF;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,QAAQ,aAAa,SAAS;AACrC,QAAM,WAAW,QAAQ,IAAI,CAAC,GAAG,MAAM,gBAAgB,GAAG,UAAU,KAAK,QAAQ,CAAC;AAClF,QAAM,QAAQ,EAAE,SAAS;AACzB,SAAO,OAAO,UAAU,QAAQ,UAAU,KAAK;AACjD;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,SAAS;AACnD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,GAAG;AACf,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,SAAO,GAAG,UAAU,WAAW,GAAG,UAAU,WAAW,MAAM,2CAA2C;AACxG,MAAI,GAAG,UAAU,SAAS;AACxB,SAAK,KAAK,IAAI,SAAS;AAAA,EACzB;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,qBAAqB,MAAM,MAAM;AACxC,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,QAAI,KAAK,KAAK,SAAS,IAAI,OAAO,OAAO,IAAI,GAAG;AAC9C,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,WAAW,WAAW,MAAM;AACpD,QAAM,OAAO,UAAU,SAAS,UAAU;AAC1C,QAAM,MAAM,CAAC;AACb,MAAI,SAAS;AACb,MAAI,YAAY;AAChB,WAAS,MAAM,GAAG,MAAM,MAAM,OAAO;AACnC,QAAI,KAAK,QAAQ,GAAG,MAAM,IAAI;AAC5B,UAAI,KAAK,UAAU,SAAS;AAAA,IAC9B,OAAO;AACL,UAAI,KAAK,UAAU,YAAY;AAAA,IACjC;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,0BAA0B,QAAQ,MAAM;AAC/C,QAAM,WAAW,CAAC;AAClB,QAAM,OAAO,OAAO;AACpB,WAAS,MAAM,GAAG,MAAM,MAAM,OAAO;AACnC,QAAI,KAAK,QAAQ,GAAG,MAAM,IAAI;AAC5B,eAAS,KAAK,OAAO,IAAI;AAAA,IAC3B;AAAA,EACF;AACA,QAAM,cAAc,KAAK,IAAI,CAAC,QAAQ,OAAO,IAAI;AACjD,SAAO,CAAC,UAAU,WAAW;AAC/B;AACA,SAAS,qBAAqB,OAAO,MAAM;AACzC,QAAM,iBAAiB,KAAK,IAAI,CAAC,MAAM,CAAC;AACxC,SAAO,iBAAiB,OAAO,gBAAgB,IAAI;AACrD;AACA,SAAS,2BAA2B,KAAK,MAAM,MAAM;AACnD,SAAO,qBAAqB,MAAM,IAAI,GAAG,MAAM,GAAG,uDAAuD,iBAAiB,aAAa;AACzI;AACA,SAAS,mBAAmB,MAAM,MAAM;AACtC,MAAI,qBAAqB,MAAM,IAAI,GAAG;AACpC,WAAO;AAAA,EACT;AACA,QAAM,SAAS,CAAC;AAChB,WAAS,IAAI,GAAG,IAAI,MAAM,EAAE,GAAG;AAC7B,QAAI,KAAK,QAAQ,CAAC,MAAM,IAAI;AAC1B,aAAO,KAAK,CAAC;AAAA,IACf;AAAA,EACF;AACA,OAAK,QAAQ,CAAC,SAAS,OAAO,KAAK,IAAI,CAAC;AACxC,SAAO;AACT;AACA,SAAS,uBAAuB,MAAM;AACpC,SAAO,KAAK,IAAI,CAAC,MAAM,MAAM,CAAC,GAAG,IAAI,CAAC,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,KAAK,EAAE,EAAE,EAAE,IAAI,CAAC,MAAM,EAAE,EAAE;AACrF;AACA,SAAS,iBAAiB,SAAS,MAAM;AACvC,QAAM,MAAM,CAAC;AACb,WAAS,IAAI,OAAO,SAAS,IAAI,MAAM,EAAE,GAAG;AAC1C,QAAI,KAAK,CAAC;AAAA,EACZ;AACA,SAAO;AACT;AAGA,SAAS,KAAK,GAAG,OAAO,MAAM,WAAW,OAAO;AAC9C,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,kBAAkB,MAAM,SAAS;AACjD,SAAO,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC5C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,GAAG,OAAO,MAAM,WAAW,OAAO;AAC9C,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC5C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,KAAK,MAAM,MAAM;AACxB,MAAI,QAAQ,gBAAgB,MAAM,QAAQ,KAAK;AAC/C,MAAI,OAAO,gBAAgB,MAAM,OAAO,KAAK;AAC7C,GAAC,OAAO,IAAI,IAAI,eAAe,OAAO,IAAI;AAC1C,QAAM,SAAS,EAAE,GAAG,OAAO,GAAG,KAAK;AACnC,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,OAAO,OAAO,OAAO;AAC5B,OAAK,aAAa,KAAK,KAAK,UAAU,YAAY,MAAM,QAAQ,KAAK,MAAM,UAAU,aAAa;AAChG,UAAM,IAAI,MAAM,gFAAgF;AAAA,EAClG;AACA,MAAI,UAAU,YAAY,aAAa,KAAK,KAAK,EAAE,iBAAiB,aAAa;AAC/E,UAAM,IAAI,MAAM,2EAA2E;AAAA,EAC7F;AACA,QAAM,QAAQ,CAAC;AACf,QAAM,gBAAgB,CAAC;AACvB,SAAO,WAAW,OAAO,OAAO,eAAe,KAAK;AACtD;AAGA,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ,SAAS;AACpD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,QAAQ,GAAG;AAClB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,QAAQ,CAAC;AACf,SAAO,OAAO,UAAU,UAAU,EAAE,GAAG,GAAG,GAAG,KAAK;AACpD;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,KAAK,GAAG,OAAO,MAAM,WAAW,OAAO;AAC9C,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,GAAG,UAAU,QAAQ;AACvB,SAAK,KAAK,IAAI,OAAO;AAAA,EACvB;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC5C;AACA,IAAI,OAAO,GAAG,EAAE,KAAK,CAAC;AAGtB,SAAS,MAAM,GAAG,MAAM,aAAa,OAAO,MAAM,WAAW,OAAO;AAClE,MAAI,gBAAgB,GAAG,KAAK,MAAM;AAClC,QAAM,QAAQ,SAAS,GAAG,KAAK,IAAI;AACnC,MAAI,gBAAgB,MAAM;AAC1B,MAAI,UAAU;AACZ,UAAM,OAAO,eAAe,MAAM,EAAE,KAAK;AACzC,oBAAgB,qBAAqB,MAAM,OAAO,IAAI;AAAA,EACxD;AACA,SAAO,QAAQ,OAAO,aAAa;AACrC;AACA,SAAS,SAAS,GAAG,IAAI,OAAO,MAAM;AACpC,MAAI,EAAE,SAAS,GAAG;AAChB,WAAO,IAAI,CAAC;AAAA,EACd;AACA,MAAI,EAAE,SAAS,KAAK,SAAS,MAAM;AACjC,WAAO,SAAS,QAAQ,GAAG,CAAC,EAAE,CAAC,GAAG,IAAI,IAAI;AAAA,EAC5C;AACA,MAAI,EAAE,SAAS,KAAK,OAAO,SAAS,YAAY,MAAM,QAAQ,IAAI,KAAK,KAAK,WAAW,GAAG;AACxF,QAAI,OAAO,GAAG;AACZ,aAAO,KAAK,IAAI,CAAC,GAAG,IAAI;AAAA,IAC1B;AACA,QAAI,OAAO,UAAU;AACnB,aAAO,IAAI,IAAI,CAAC,GAAG,IAAI;AAAA,IACzB;AACA,QAAI,OAAO,WAAW;AACpB,aAAO,IAAI,IAAI,CAAC,GAAG,IAAI;AAAA,IACzB;AACA,QAAI,OAAO,eAAe,OAAO,GAAG;AAClC,aAAO,KAAK,KAAK,IAAI,IAAI,CAAC,GAAG,OAAO,GAAG,OAAO,CAAC,GAAG,IAAI,CAAC;AAAA,IACzD;AACA,UAAM,IAAI,MAAM,qCAAqC,IAAI;AAAA,EAC3D;AACA,MAAI,MAAM,QAAQ,IAAI,KAAK,KAAK,WAAW,GAAG;AAC5C,QAAI,OAAO,GAAG;AACZ,aAAO,IAAI,KAAK,IAAI,CAAC,GAAG,KAAK,EAAE,GAAG,KAAK,KAAK,CAAC;AAAA,IAC/C;AACA,QAAI,OAAO,UAAU;AACnB,aAAO,IAAI,KAAK,IAAI,CAAC,GAAG,KAAK,EAAE,GAAG,KAAK,EAAE;AAAA,IAC3C;AACA,QAAI,OAAO,WAAW;AACpB,aAAO,IAAI,KAAK,IAAI,CAAC,GAAG,KAAK,EAAE,GAAG,KAAK,EAAE;AAAA,IAC3C;AACA,QAAI,OAAO,SAAS,OAAO,aAAa;AACtC,aAAO,KAAK,KAAK,OAAO,CAAC,GAAG,IAAI,CAAC;AAAA,IACnC;AACA,UAAM,IAAI,MAAM,qCAAqC,IAAI;AAAA,EAC3D;AACA,QAAM,IAAI,MAAM,gCAAgC,MAAM;AACxD;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,eAAe,GAAG,OAAO,MAAM,WAAW,OAAO;AACxD,SAAO,KAAK,GAAG,aAAa,MAAM,QAAQ;AAC5C;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,YAAY,GAAG,OAAO,GAAG;AAChC,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,mBAAmB;AACpE,SAAO,QAAQ,GAAG,MAAM,MAAM,oCAAoC;AAClE,QAAM,SAAS,EAAE,OAAO,GAAG;AAC3B,QAAM,QAAQ,EAAE,KAAK,KAAK;AAC1B,SAAO,OAAO,UAAU,YAAY,QAAQ,KAAK;AACnD;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG,MAAM;AACtB,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ,mBAAmB;AAC9D,SAAO,GAAG,SAAS,KAAK,QAAQ,MAAM,qCAAqC,GAAG,kCAAkC,OAAO;AACvH,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,KAAK,SAAS,YAAY,YAAY,QAAQ,WAAW;AAChE,MAAI,cAAc,MAAM;AACtB,iBAAa;AAAA,EACf;AACA,QAAM,OAAO,OAAO,CAAC,SAAS,UAAU,GAAG,KAAK;AAChD,QAAM,IAAI,WAAW,aAAa,UAAU;AAC5C,WAAS,IAAI,GAAG,IAAI,GAAG,EAAE,GAAG;AAC1B,SAAK,IAAI,GAAG,GAAG,CAAC;AAAA,EAClB;AACA,QAAM,MAAM,QAAQ,KAAK,SAAS,GAAG,CAAC,SAAS,UAAU,CAAC;AAC1D,MAAI,cAAc,MAAM;AACtB,WAAO;AAAA,EACT,OAAO;AACL,QAAI,WAAW,WAAW,GAAG;AAC3B,aAAO,KAAK,WAAW,KAAK,CAAC,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,CAAC;AAAA,IACvD,WAAW,WAAW,WAAW,GAAG;AAClC,aAAO,KAAK,WAAW,WAAW,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,WAAW,IAAI,WAAW,IAAI,GAAG,CAAC,CAAC;AAAA,IACrF,WAAW,WAAW,WAAW,GAAG;AAClC,aAAO,KAAK,WAAW,WAAW,WAAW,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG;AAAA,QAC5D,WAAW;AAAA,QACX,WAAW;AAAA,QACX,WAAW;AAAA,QACX;AAAA,QACA;AAAA,MACF,CAAC;AAAA,IACH,OAAO;AACL,YAAM,IAAI,MAAM,qEAAqE,WAAW,UAAU;AAAA,IAC5G;AAAA,EACF;AACF;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS,SAAS;AACrD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,QAAQ,GAAG,SAAS,OAAO,GAAG,YAAY,GAAG;AACpD,QAAM,KAAK,gBAAgB,GAAG,KAAK,QAAQ;AAC3C,QAAM,WAAW,gBAAgB,SAAS,WAAW,UAAU,OAAO;AACtE,QAAM,SAAS,EAAE,GAAG,IAAI,SAAS,SAAS;AAC1C,QAAM,QAAQ,EAAE,MAAM,UAAU;AAChC,SAAO,OAAO,UAAU,UAAU,QAAQ,KAAK;AACjD;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,SAAS,GAAG,GAAG;AACtB,MAAI,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AAC/D,MAAI,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AAC/D,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,SAAS,MAAM;AACzC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,cAAc,GAAG,GAAG;AAC3B,MAAI,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,mBAAmB;AACpE,MAAI,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,mBAAmB;AACpE,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,cAAc,MAAM;AAC9C;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,UAAU,GAAG;AACpB,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC7C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,YAAY,GAAG,EAAE,UAAU,CAAC;AAGhC,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,SAAS,GAAG,EAAE,OAAO,CAAC;AAG1B,SAAS,WAAW,GAAG,QAAQ,KAAK;AAClC,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM;AACtB,SAAO,OAAO,UAAU,WAAW,QAAQ,KAAK;AAClD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,MAAM,GAAG,GAAG;AACnB,MAAI,KAAK,gBAAgB,GAAG,KAAK,QAAQ,mBAAmB;AAC5D,MAAI,KAAK,gBAAgB,GAAG,KAAK,QAAQ,mBAAmB;AAC5D,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,WAAW,GAAG,GAAG;AACxB,MAAI,KAAK,gBAAgB,GAAG,KAAK,aAAa,mBAAmB;AACjE,MAAI,KAAK,gBAAgB,GAAG,KAAK,aAAa,mBAAmB;AACjE,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,WAAW,MAAM;AAC3C;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,SAAS,OAAO,MAAM,KAAK;AAClC,MAAI,OAAO,GAAG;AACZ,UAAM,IAAI,MAAM,0CAA0C;AAAA,EAC5D;AACA,QAAM,QAAQ,EAAE,OAAO,MAAM,IAAI;AACjC,SAAO,OAAO,UAAU,UAAU,CAAC,GAAG,KAAK;AAC7C;AAGA,SAAS,4BAA4B,GAAG,cAAc,GAAG,OAAO,GAAG,QAAQ,GAAG,OAAO,KAAK;AACxF,QAAM,KAAK,gBAAgB,GAAG,KAAK,4BAA4B;AAC/D,SAAO,GAAG,SAAS,KAAK,GAAG,SAAS,GAAG,MAAM;AAAA,sBACzB,GAAG,OAAO;AAC9B,SAAO,MAAM,WAAW,GAAG,MAAM,2FAA2F,cAAc;AAC1I,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,QAAM,SAAS,EAAE,GAAG,IAAI;AACxB,QAAM,QAAQ,EAAE,aAAa,MAAM,OAAO,KAAK;AAC/C,QAAM,MAAM,OAAO,UAAU,KAAK,QAAQ,KAAK;AAC/C,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,IAAI,6BAA6B,GAAG,EAAE,4BAA4B,CAAC;AAGnE,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,SAAS;AACnD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,OAAO,GAAG,EAAE,KAAK,CAAC;AAGtB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,KAAK,GAAG;AACf,SAAO,WAAW,CAAC,GAAG,MAAM,4CAA4C;AACxE,SAAO,CAAC,GAAG,OAAO;AAChB,UAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AACjE,UAAM,MAAM,MAAM,OAAO,gBAAgB,IAAI,MAAM,SAAS,IAAI;AAChE,WAAO,OAAO,KAAK,MAAM;AACvB,YAAM,EAAE,OAAO,OAAO,OAAO,IAAI,OAAO,UAAU,MAAM,EAAE,EAAE,GAAG,CAAC,EAAE,GAAG,GAAG;AACxE,UAAI,OAAO,MAAM;AACf,0BAAkB,MAAM,OAAO,IAAI,OAAO,gFAAgF;AAAA,MAC5H;AACA,iBAAW,MAAM;AACjB,aAAO,OAAO;AAAA,IAChB,CAAC;AAAA,EACH;AACF;AACA,SAAS,MAAM,GAAG;AAChB,SAAO,WAAW,CAAC,GAAG,MAAM,6CAA6C;AACzE,SAAO,CAAC,MAAM,OAAO;AACnB,WAAO,MAAM,QAAQ,IAAI,GAAG,MAAM,kFAAkF;AACpH,UAAM,QAAQ,qBAAqB,MAAM,QAAQ,YAAY,mBAAmB;AAChF,UAAM,MAAM,MAAM,OAAO,gBAAgB,IAAI,MAAM,UAAU,IAAI;AACjE,WAAO,OAAO,KAAK,MAAM;AACvB,YAAM,EAAE,OAAO,OAAO,OAAO,IAAI,OAAO,UAAU,MAAM,EAAE,GAAG,KAAK,GAAG,OAAO,GAAG;AAC/E,UAAI,OAAO,MAAM;AACf,0BAAkB,MAAM,OAAO,IAAI,OAAO,+FAA+F;AAAA,MAC3I;AACA,iBAAW,MAAM;AACjB,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,SAAS,aAAa,GAAG;AACvB,SAAO,WAAW,CAAC,GAAG,MAAM,oDAAoD;AAChF,SAAO,CAAC,GAAG,OAAO;AAChB,WAAO,aAAa,QAAQ,MAAM,qDAAqD;AACvF,WAAO,MAAM,QAAQ,cAAc,QAAQ,MAAM,0DAA0D;AAC3G,UAAM,EAAE,OAAO,QAAQ,MAAM,IAAI,OAAO,UAAU,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE;AACrE,eAAW,MAAM;AACjB,WAAO,EAAE,MAAM,OAAO,IAAI,MAAM;AAAA,EAClC;AACF;AACA,SAAS,cAAc,GAAG;AACxB,SAAO,WAAW,CAAC,GAAG,MAAM,qDAAqD;AACjF,SAAO,CAAC,MAAM,OAAO;AACnB,WAAO,MAAM,QAAQ,IAAI,KAAK,KAAK,MAAM,CAAC,QAAQ,eAAe,MAAM,GAAG,MAAM,oEAAoE;AACpJ,WAAO,MAAM,QAAQ,cAAc,QAAQ,MAAM,8DAA8D;AAC/G,UAAM,MAAM,OAAO,UAAU,MAAM,EAAE,GAAG,IAAI,GAAG,MAAM,EAAE;AACvD,QAAI,MAAM,MAAM;AACd,wBAAkB,IAAI,MAAM,OAAO,GAAG,OAAO,uGAAuG;AAAA,IACtJ;AACA,eAAW,IAAI,KAAK;AACpB,WAAO;AAAA,EACT;AACF;AACA,SAAS,cAAc,GAAG,SAAS;AACjC,SAAO,WAAW,CAAC,GAAG,MAAM,qDAAqD;AACjF,SAAO,WAAW,QAAQ,MAAM,QAAQ,OAAO,KAAK,QAAQ,MAAM,CAAC,MAAM,aAAa,QAAQ,GAAG,MAAM,+EAA+E;AACtL,QAAM,mBAAmB,WAAW;AACpC,MAAI,CAAC,kBAAkB;AACrB,cAAU,CAAC;AACX,eAAW,WAAW,OAAO,qBAAqB;AAChD,cAAQ,KAAK,OAAO,oBAAoB,QAAQ;AAAA,IAClD;AAAA,EACF;AACA,QAAM,wBAAwB,mBAAmB,QAAQ,OAAO,CAAC,cAAc,CAAC,UAAU,SAAS,IAAI;AACvG,QAAM,mBAAmB,QAAQ;AACjC,YAAU,QAAQ,OAAO,CAAC,cAAc,UAAU,SAAS;AAC3D,SAAO,QAAQ,SAAS,GAAG,MAAM,gGAAgG,0CAA0C;AAC3K,QAAM,mBAAmB;AACzB,QAAM,EAAE,OAAO,OAAO,OAAO,IAAI,OAAO,UAAU,GAAG,SAAS,MAAM,gBAAgB;AACpF,SAAO,OAAO,KAAK,CAAC,MAAM,KAAK,IAAI,GAAG,MAAM,8LAA8L;AAC1O,SAAO,MAAM,SAAS,GAAG,MAAM,iFAAiF,MAAM,aAAa;AACnI,QAAM,aAAa,CAAC;AACpB,UAAQ,QAAQ,CAAC,GAAG,MAAM;AACxB,QAAI,OAAO,MAAM,MAAM;AACrB,iBAAW,EAAE,QAAQ,OAAO;AAAA,IAC9B;AAAA,EACF,CAAC;AACD,MAAI,yBAAyB,MAAM;AACjC,0BAAsB,QAAQ,CAAC,MAAM,WAAW,EAAE,QAAQ,IAAI;AAAA,EAChE;AACA,SAAO,EAAE,OAAO,OAAO,WAAW;AACpC;AACA,SAAS,WAAW,GAAG;AACrB,SAAO,OAAO,WAAW,CAAC;AAC5B;AACA,SAAS,WAAW,QAAQ;AAC1B,QAAM,mBAAmB,OAAO,OAAO,CAAC,MAAM,KAAK,IAAI,EAAE;AACzD,MAAI,mBAAmB,GAAG;AACxB,UAAM,IAAI,MAAM;AAAA,oEACgD;AAAA,EAClE;AACF;AAGA,SAAS,UAAU,GAAG;AACpB,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC7C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,YAAY,GAAG;AACtB,QAAM,KAAK,gBAAgB,GAAG,KAAK,YAAY;AAC/C,QAAM,WAAW,WAAW,CAAC,OAAO;AAClC,UAAM,QAAQ,IAAI,SAAS,IAAI,EAAE,CAAC,CAAC;AACnC,UAAM,WAAW,CAAC,OAAO;AACvB,YAAM,OAAO,IAAI,IAAI,QAAQ,IAAI,EAAE,CAAC,CAAC;AACrC,aAAO;AAAA,IACT;AACA,WAAO,EAAE,OAAO,SAAS;AAAA,EAC3B,CAAC;AACD,SAAO,SAAS,EAAE;AACpB;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,KAAK,GAAG,GAAG;AAClB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,YAAY,QAAQ,OAAO,IAAI;AACtC,QAAM,UAAU,gBAAgB,QAAQ,UAAU,YAAY;AAC9D,MAAI,SAAS,IAAI;AACf,WAAO,QAAQ,OAAO;AAAA,EACxB;AACA,MAAI,SAAS,QAAQ,OAAO,GAAG;AAC7B,UAAM,MAAM,gFAAgF,QAAQ,qBAAqB,MAAM;AAAA,EACjI;AACA,QAAM,WAAW,WAAW,CAAC,SAAS,SAAS;AAC7C,UAAM,WAAW;AACjB,UAAM,OAAO,IAAI,SAAS,MAAM,IAAI;AACpC,UAAM,UAAU,IAAI,SAAS,IAAI;AACjC,UAAM,QAAQ,IAAI,KAAK,SAAS,SAAS,GAAG,KAAK,KAAK,IAAI,OAAO,GAAG,MAAM,QAAQ,CAAC,CAAC;AACpF,SAAK,CAAC,KAAK,CAAC;AACZ,UAAM,WAAW,CAAC,IAAI,UAAU;AAC9B,YAAM,CAAC,MAAM,IAAI;AACjB,YAAM,YAAY;AAClB,YAAM,WAAW,IAAI,MAAM;AAC3B,aAAO,IAAI,IAAI,IAAI,KAAK,IAAI,MAAM,SAAS,GAAG,QAAQ,CAAC;AAAA,IACzD;AACA,WAAO,EAAE,OAAO,SAAS;AAAA,EAC3B,CAAC;AACD,SAAO,SAAS,OAAO;AACzB;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,WAAW,GAAG,OAAO,MAAM,WAAW,OAAO;AACpD,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,OAAO,eAAe,MAAM,GAAG,KAAK;AAC1C,QAAM,OAAO,IAAI,IAAI,MAAM,IAAI;AAC/B,QAAM,IAAI,IAAI,IAAI,IAAI;AACtB,QAAM,IAAI,IAAI,CAAC;AACf,QAAM,IAAI,KAAK,GAAG,IAAI;AACtB,QAAM,IAAI,KAAK,CAAC;AAChB,QAAM,MAAM,KAAK,QAAQ,MAAM,EAAE,KAAK,GAAG,CAAC;AAC1C,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,IAAI,OAAO,IAAI;AACrD,WAAO,QAAQ,KAAK,QAAQ;AAAA,EAC9B;AACA,SAAO;AACT;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,YAAY,GAAG,GAAG;AACzB,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,MAAM;AACvD,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,MAAM;AACvD,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,YAAY,MAAM;AAC5C;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,YAAY,GAAG;AACtB,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,MAAM;AACvD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,YAAY,MAAM;AAC5C;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,WAAW,GAAG,GAAG;AACxB,QAAM,KAAK,gBAAgB,GAAG,KAAK,aAAa,MAAM;AACtD,QAAM,KAAK,gBAAgB,GAAG,KAAK,aAAa,MAAM;AACtD,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,WAAW,MAAM;AAC3C;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,YAAY,GAAG,GAAG;AACzB,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,MAAM;AACvD,QAAM,KAAK,gBAAgB,GAAG,KAAK,cAAc,MAAM;AACvD,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,SAAO,WAAW,UAAU,GAAG,CAAC,GAAG,WAAW,WAAW,GAAG,CAAC,CAAC,CAAC;AACjE;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,IAAI,YAAY;AAChB,SAAS,cAAc,gBAAgB,QAAQ,OAAO,QAAQ;AAC5D,QAAM,kBAAkB,gBAAgB,gBAAgB,kBAAkB,cAAc;AACxF,QAAM,UAAU,gBAAgB,QAAQ,UAAU,cAAc;AAChE,QAAM,eAAe,gBAAgB,MAAM,gBAAgB,MAAM,SAAS;AAC1E,QAAM,aAAa,QAAQ,MAAM,QAAQ,MAAM,SAAS;AACxD,QAAM,oBAAoB,QAAQ,iBAAiB,CAAC,IAAI,YAAY,CAAC;AACrE,QAAM,YAAY,QAAQ,SAAS,CAAC,IAAI,UAAU,CAAC;AACnD,MAAI,kBAAkB,OAAO,GAAG;AAC9B,UAAM,IAAI,MAAM,sDAAsD;AAAA,EACxE;AACA,MAAI,kBAAkB,MAAM,OAAO,UAAU,MAAM,IAAI;AACrD,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AACA,MAAI,cAAc,UAAU,KAAK,KAAK,WAAW;AAC/C,UAAM,IAAI,MAAM,qCAAqC,WAAW;AAAA,EAClE;AACA,MAAI,kBAAkB,MAAM,MAAM,WAAW;AAC3C,UAAM,IAAI,MAAM,oCAAoC,wCAAwC,kBAAkB,MAAM,IAAI;AAAA,EAC1H;AACA,QAAM,SAAS;AAAA,IACb,gBAAgB;AAAA,IAChB,QAAQ;AAAA,EACV;AACA,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,cAAc,QAAQ,KAAK;AACrD;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,WAAW,gBAAgB,QAAQ;AAC1C,SAAO,aAAa,gBAAgB,QAAQ,MAAM;AACpD;AAGA,SAAS,SAAS,GAAG,YAAY,SAAS,MAAM,iBAAiB;AAC/D,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,QAAM,YAAY;AAClB,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,uDAAuD,IAAI,OAAO;AAC/F,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AAC/K,4BAA0B,WAAW,MAAM,eAAe;AAC1D,QAAM,SAAS,EAAE,GAAG,IAAI;AACxB,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB;AAChE,QAAM,MAAM,OAAO,UAAU,SAAS,QAAQ,KAAK;AACnD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,GAAG,aAAa,CAAC,GAAG,GAAG,CAAC,GAAG,SAAS,MAAM,iBAAiB,aAAa,SAAS;AACnG,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC3E;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,qDAAqD,IAAI,OAAO;AAC7F,SAAO,eAAe,SAAS,MAAM,gFAAgF,YAAY;AACjI,4BAA0B,aAAa,MAAM,eAAe;AAC5D,QAAM,SAAS,EAAE,GAAG,IAAI;AACxB,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,iBAAiB,WAAW;AAC5E,QAAM,MAAM,OAAO,UAAU,WAAW,QAAQ,KAAK;AACrD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,mBAAmB,GAAG,YAAY,SAAS,MAAM,sBAAsB,OAAO;AACrF,QAAM,KAAK,gBAAgB,GAAG,KAAK,mBAAmB;AACtD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,oBAAoB;AACpE,QAAM,SAAS,OAAO,UAAU,mBAAmB,QAAQ,KAAK;AAChE,SAAO,EAAE,QAAQ,OAAO,IAAI,SAAS,OAAO,GAAG;AACjD;AACA,IAAI,oBAAoB,GAAG,EAAE,mBAAmB,CAAC;AAGjD,SAAS,SAAS,GAAG,GAAG;AACtB,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC1C,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC1C,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,MAAI,GAAG,UAAU,QAAQ;AACvB,SAAK,KAAK,IAAI,OAAO;AACrB,SAAK,KAAK,IAAI,OAAO;AAAA,EACvB;AACA,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,SAAS,MAAM;AACzC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,MAAM,GAAG,OAAO,MAAM,WAAW,OAAO;AAC/C,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,MAAM,OAAO,QAAQ,WAAW;AACvC,MAAI,UAAU,aAAa;AACzB,UAAM,QAAQ,MAAM,OAAO,SAAS;AACpC,UAAM,QAAQ,MAAM,OAAO,SAAS;AACpC,WAAO,QAAQ,OAAO,KAAK;AAAA,EAC7B;AACA,QAAM,SAAS,oBAAoB,cAAc,KAAK,GAAG,KAAK;AAC9D,SAAO,OAAO,WAAW,QAAQ,OAAO,KAAK;AAC/C;AAGA,SAAS,MAAM,OAAO,QAAQ,WAAW;AACvC,MAAI,UAAU,aAAa;AACzB,UAAM,QAAQ,MAAM,OAAO,SAAS;AACpC,UAAM,QAAQ,MAAM,OAAO,SAAS;AACpC,WAAO,QAAQ,OAAO,KAAK;AAAA,EAC7B;AACA,QAAM,SAAS,mBAAmB,cAAc,KAAK,GAAG,KAAK;AAC7D,SAAO,OAAO,WAAW,QAAQ,OAAO,KAAK;AAC/C;AAGA,SAAS,SAAS,GAAG,GAAG,EAAE,WAAW,KAAK,IAAI,CAAC,GAAG;AAChD,MAAI,aAAa,QAAQ,aAAa,MAAM;AAC1C,UAAM,IAAI,UAAU,GAAG,oDAAoD;AAAA,EAC7E;AACA,MAAI,MAAM,QAAQ;AAChB,WAAO,CAAC;AAAA,EACV;AACA,MAAI,KAAK,gBAAgB,GAAG,KAAK,YAAY,aAAa,SAAS,EAAE,QAAQ,SAAS;AACtF,MAAI,MAAM,QAAQ;AAChB,WAAO,CAAC,EAAE;AAAA,EACZ;AACA,MAAI,KAAK,gBAAgB,GAAG,KAAK,YAAY,aAAa,SAAS,EAAE,QAAQ,SAAS;AACtF,QAAM,IAAI,cAAc,GAAG,KAAK;AAChC,QAAM,IAAI,cAAc,GAAG,KAAK;AAChC,MAAI,aAAa,MAAM;AACrB,SAAK,QAAQ,IAAI,CAAC,GAAG,EAAE,CAAC;AACxB,SAAK,QAAQ,IAAI,CAAC,IAAI,CAAC,CAAC;AACxB,WAAO;AAAA,MACL,OAAO,MAAM,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK,GAAG,EAAE;AAAA,MAClC,OAAO,IAAI,MAAM,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK,CAAC;AAAA,IACpC;AAAA,EACF;AACA,OAAK,QAAQ,IAAI,CAAC,IAAI,CAAC,CAAC;AACxB,OAAK,QAAQ,IAAI,CAAC,GAAG,EAAE,CAAC;AACxB,SAAO;AAAA,IACL,OAAO,IAAI,MAAM,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK,CAAC;AAAA,IAClC,OAAO,MAAM,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK,GAAG,EAAE;AAAA,EACpC;AACF;AAGA,SAAS,SAAS,GAAG,GAAG;AACtB,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC1C,MAAI,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC1C,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,MAAI,GAAG,UAAU,QAAQ;AACvB,SAAK,KAAK,IAAI,OAAO;AACrB,SAAK,KAAK,IAAI,OAAO;AAAA,EACvB;AACA,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,SAAS,MAAM;AACzC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,GAAG,UAAU,MAAM;AACrC,SAAO,SAAS,aAAa,SAAS,aAAa,MAAM,+DAA+D,OAAO;AAC/H,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,MAAI,GAAG,SAAS,GAAG;AACjB,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AACA,SAAO,SAAS,WAAW,GAAG,MAAM,MAAM,wCAAwC,GAAG,aAAa,SAAS,SAAS;AACpH,QAAM,cAAc,SAAS,YAAY,IAAI;AAC7C,WAAS,IAAI,GAAG,IAAI,GAAG,MAAM,KAAK;AAChC,WAAO,SAAS,GAAG,WAAW,GAAG,MAAM,uDAAuD;AAC9F,WAAO,SAAS,GAAG,MAAM,KAAK,SAAS,GAAG,MAAM,GAAG,MAAM,KAAK,eAAe,SAAS,GAAG,MAAM,KAAK,SAAS,GAAG,MAAM,GAAG,MAAM,KAAK,aAAa,MAAM,wBAAwB,wCAAwC,GAAG,MAAM,KAAK,iDAAiD,GAAG,OAAO;AAAA,EAClS;AACA,QAAM,QAAQ,EAAE,UAAU,KAAK;AAC/B,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,WAAW,QAAQ,KAAK;AAClD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,KAAK,GAAG,GAAG;AAClB,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,MAAI,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACtC,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,SAAS,GAAG,OAAO,MAAM,WAAW,OAAO;AAClD,MAAI,gBAAgB,GAAG,KAAK,SAAS;AACrC,QAAM,OAAO,eAAe,MAAM,EAAE,KAAK;AACzC,QAAM,QAAQ,KAAK,GAAG,MAAM,QAAQ;AACpC,MAAI,gBAAgB,MAAM;AAC1B,MAAI,CAAC,UAAU;AACb,oBAAgB,qBAAqB,MAAM,OAAO,IAAI;AAAA,EACxD;AACA,QAAM,aAAa,OAAO,IAAI,KAAK,GAAG,SAAS,GAAG,QAAQ,OAAO,aAAa,CAAC,CAAC;AAChF,QAAM,WAAW,KAAK,YAAY,MAAM,QAAQ;AAChD,SAAO,EAAE,MAAM,OAAO,SAAS;AACjC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,cAAc,WAAW,MAAM,GAAG,GAAG;AAC5C,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,cAAc;AAC1D,QAAM,KAAK,qBAAqB,GAAG,KAAK,cAAc;AACtD,QAAM,KAAK,qBAAqB,GAAG,KAAK,cAAc;AACtD,MAAI,SAAS;AACb,QAAM,YAAY,CAAC;AACnB,WAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,KAAK;AACzC,UAAM,SAAS,UAAU,GAAG,QAAQ,GAAG,IAAI,GAAG,EAAE;AAChD,cAAU,KAAK,OAAO,EAAE;AACxB,cAAU,KAAK,OAAO,EAAE;AACxB,aAAS,OAAO;AAAA,EAClB;AACA,QAAM,OAAO,CAAC;AACd,QAAM,OAAO,CAAC;AACd,WAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,KAAK,GAAG;AAC5C,SAAK,KAAK,UAAU,EAAE;AACtB,SAAK,KAAK,UAAU,IAAI,EAAE;AAAA,EAC5B;AACA,SAAO,CAAC,MAAM,IAAI;AACpB;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,aAAa,QAAQ,YAAY,MAAM,aAAa,OAAO;AAClE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,aAAa;AAC/D,QAAM,cAAc,QAAQ;AAC5B,QAAM,WAAW,QAAQ;AACzB,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,+DAA+D,cAAc;AAAA,EAC/F;AACA,MAAI,WAAW,GAAG;AAChB,UAAM,IAAI,MAAM,gDAAgD,UAAU;AAAA,EAC5E;AACA,SAAO,QAAQ,KAAK,OAAO;AAC3B,QAAM,WAAW,aAAa,IAAI,QAAQ,SAAS,CAAC,GAAG,EAAE,CAAC,IAAI;AAC9D,QAAM,SAAS,EAAE,QAAQ,SAAS;AAClC,QAAM,QAAQ,EAAE,YAAY,MAAM,WAAW;AAC7C,QAAM,MAAM,OAAO,UAAU,aAAa,QAAQ,KAAK;AACvD,SAAO,aAAa,IAAI,QAAQ,KAAK,CAAC,IAAI,IAAI,CAAC,IAAI;AACrD;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,UAAU,GAAG,GAAG;AACvB,MAAI,KAAK,gBAAgB,GAAG,KAAK,YAAY,mBAAmB;AAChE,MAAI,KAAK,gBAAgB,GAAG,KAAK,YAAY,mBAAmB;AAChE,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,UAAU,GAAG;AACpB,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC7C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,cAAc,IAAI,IAAI;AAC7B,QAAM,MAAM,gBAAgB,IAAI,MAAM,cAAc;AACpD,QAAM,MAAM,gBAAgB,IAAI,MAAM,cAAc;AACpD,SAAO,IAAI,SAAS,KAAK,IAAI,SAAS,GAAG,MAAM,+DAA+D,IAAI,YAAY,IAAI,OAAO;AACzI,QAAM,OAAO,QAAQ,KAAK,CAAC,IAAI,CAAC,CAAC;AACjC,QAAM,OAAO,QAAQ,KAAK,CAAC,GAAG,EAAE,CAAC;AACjC,SAAO,OAAO,MAAM,IAAI;AAC1B;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,KAAK,GAAG,UAAU,gBAAgB,GAAG;AAC5C,QAAM,KAAK,gBAAgB,GAAG,KAAK,KAAK;AACxC,MAAI,GAAG,SAAS,GAAG;AACjB,UAAM,IAAI,MAAM,oDAAoD;AAAA,EACtE;AACA,QAAM,QAAQ,EAAE,UAAU,cAAc;AACxC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,QAAQ,KAAK;AAC9C;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,OAAO,GAAG,UAAU,gBAAgB,GAAG;AAC9C,SAAO,SAAS,WAAW,GAAG,MAAM,kDAAkD;AACtF,SAAO,IAAI,GAAG,CAAC,QAAQ,GAAG,aAAa;AACzC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG,UAAU,gBAAgB,GAAG;AAC9C,SAAO,SAAS,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,GAAG,MAAM,uDAAuD;AACnJ,SAAO,IAAI,GAAG,UAAU,aAAa;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG,UAAU,gBAAgB,GAAG;AAC9C,SAAO,SAAS,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,GAAG,MAAM,uDAAuD;AAC/K,SAAO,IAAI,GAAG,UAAU,aAAa;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG,UAAU,gBAAgB,GAAG;AAC9C,SAAO,SAAS,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,KAAK,SAAS,GAAG,WAAW,GAAG,MAAM,uDAAuD;AAC3M,SAAO,IAAI,GAAG,UAAU,aAAa;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,gBAAgB,GAAG,YAAY,UAAU;AAChD,QAAM,KAAK,gBAAgB,GAAG,KAAK,gBAAgB;AACnD,SAAO,GAAG,QAAQ,IAAI,WAAW,QAAQ,MAAM,cAAc,GAAG,sCAAsC,WAAW,QAAQ;AACzH,SAAO,SAAS,WAAW,WAAW,QAAQ,MAAM,qBAAqB,SAAS,wCAAwC,WAAW,QAAQ;AAC7I,SAAO,GAAG,MAAM,OAAO,CAAC,GAAG,GAAG,MAAM;AAClC,QAAI,IAAI,KAAK,KAAK,WAAW,QAAQ;AACnC,aAAO,MAAM,IAAI,SAAS,IAAI,GAAG,KAAK,SAAS,IAAI,GAAG,MAAM,WAAW,IAAI,OAAO;AAAA,IACpF;AACA,WAAO;AAAA,EACT,GAAG,IAAI,GAAG,MAAM,4BAA4B,GAAG,MAAM,MAAM,CAAC,mBAAmB,SAAS,SAAS,sCAAsC,WAAW,SAAS,GAAG;AAC9J,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,YAAY,SAAS;AACrC,SAAO,OAAO,UAAU,gBAAgB,QAAQ,KAAK;AACvD;AACA,IAAI,iBAAiB,GAAG,EAAE,gBAAgB,CAAC;AAG3C,SAAS,MAAM,QAAQ,aAAa,aAAa,MAAM,WAAW,SAAS,iBAAiB;AAC1F,MAAI,aAAa,MAAM;AACrB,gBAAY,CAAC,GAAG,CAAC;AAAA,EACnB;AACA,MAAI,WAAW,MAAM;AACnB,cAAU;AAAA,EACZ;AACA,MAAI,SAAS,GAAG;AACd,WAAO;AAAA,EACT;AACA,QAAM,KAAK,gBAAgB,QAAQ,KAAK,SAAS;AACjD,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,qEAAqE,0BAA0B,YAAY;AAC5K,QAAM,WAAW,kBAAkB,IAAI,OAAO,aAAa,SAAS,WAAW,IAAI;AACnF,QAAM,WAAW,CAAC,SAAS,gBAAgB,SAAS,aAAa;AACjE,MAAI;AACJ,MAAI,SAAS,QAAQ;AACnB,kBAAc,6BAA6B,CAAC,SAAS,cAAc,SAAS,WAAW,GAAG,QAAQ;AAAA,EACpG,OAAO;AACL,kBAAc,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,EAC/B;AACA,QAAM,gBAAgB,SAAS,OAAO,KAAK,SAAS,OAAO;AAC3D,QAAM,CAAC,iBAAiB,aAAa,IAAI,6BAA6B,CAAC,SAAS,UAAU,SAAS,OAAO,GAAG,UAAU,WAAW;AAClI,QAAM,eAAe,gBAAgB,OAAO;AAC5C,QAAM,aAAa,gBAAgB,MAAM,eAAe,KAAK,UAAU,eAAe;AACtF,QAAM,YAAY,gBAAgB,QAAQ,MAAM,QAAQ,YAAY,aAAa,SAAS,cAAc,eAAe,IAAI,MAAM,QAAQ,YAAY,aAAa,SAAS,cAAc,eAAe;AACxM,QAAM,IAAI,UAAU;AACpB,QAAM,MAAM,gBAAgB,IAAI,eAAe,GAAG,UAAU,aAAa;AACzE,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,SAAS,6BAA6B,YAAY,YAAY,aAAa;AACzE,QAAM,WAAW,YAAY,IAAI,CAAC,MAAM,EAAE,EAAE;AAC5C,QAAM,aAAa,YAAY,IAAI,CAAC,MAAM,EAAE,EAAE;AAC9C,QAAM,iBAAiB,WAAW,OAAO,UAAU,UAAU;AAC7D,QAAM,cAAc,WAAW,IAAI,CAAC,GAAG,OAAO,IAAI,eAAe,KAAK,KAAK,CAAC;AAC5E,QAAM,SAAS,WAAW,IAAI,CAAC,GAAG,MAAM,IAAI,YAAY,EAAE;AAC1D,QAAM,WAAW,WAAW,IAAI,CAAC,GAAG,MAAM,CAAC,SAAS,IAAI,OAAO,EAAE,CAAC;AAClE,QAAM,QAAQ,WAAW,IAAI,CAAC,GAAG,MAAM,CAAC,GAAG,YAAY,EAAE,CAAC;AAC1D,SAAO,CAAC,UAAU,KAAK;AACzB;AACA,SAAS,6BAA6B,aAAa,UAAU;AAC3D,QAAM,qBAAqB,YAAY,IAAI,CAAC,GAAG,MAAM;AACnD,WAAO,KAAK,IAAI,MAAM,SAAS,KAAK;AAAA,EACtC,CAAC;AACD,QAAM,gBAAgB,mBAAmB,IAAI,CAAC,MAAM,IAAI,CAAC;AACzD,QAAM,gBAAgB,cAAc,IAAI,CAAC,MAAM,KAAK,MAAM,IAAI,CAAC,CAAC;AAChE,QAAM,cAAc,cAAc,IAAI,CAAC,GAAG,MAAM,IAAI,cAAc,EAAE;AACpE,SAAO,cAAc,IAAI,CAAC,GAAG,MAAM;AACjC,WAAO,CAAC,cAAc,IAAI,YAAY,EAAE;AAAA,EAC1C,CAAC;AACH;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG,OAAO;AACxB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,gBAAgB,OAAO,SAAS,OAAO;AACtD,QAAM,SAAS,EAAE,GAAG,IAAI,OAAO,OAAO;AACtC,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG,OAAO,MAAM,WAAW,OAAO;AAC/C,MAAI,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACvC,MAAI,GAAG,UAAU,QAAQ;AACvB,SAAK,KAAK,IAAI,OAAO;AAAA,EACvB;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,SAAS;AAC/B,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,cAAc,oBAAoB,mBAAmB,SAAS,kBAAkB;AACvF,QAAM,sBAAsB,mBAAmB,IAAI,CAAC,GAAG,MAAM,gBAAgB,GAAG,UAAU,KAAK,gBAAgB,OAAO,CAAC;AACvH,QAAM,qBAAqB,gBAAgB,mBAAmB,qBAAqB,cAAc;AACjG,QAAM,WAAW,gBAAgB,SAAS,WAAW,gBAAgB,OAAO;AAC5E,QAAM,SAAS;AAAA,IACb,oBAAoB;AAAA,IACpB,mBAAmB;AAAA,IACnB,SAAS;AAAA,EACX;AACA,QAAM,QAAQ,EAAE,iBAAiB;AACjC,QAAM,SAAS,OAAO,UAAU,cAAc,QAAQ,KAAK;AAC3D,SAAO;AAAA,IACL,oBAAoB,OAAO,MAAM,GAAG,OAAO,SAAS,CAAC;AAAA,IACrD,mBAAmB,OAAO,OAAO,SAAS;AAAA,EAC5C;AACF;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,aAAa,QAAQ,QAAQ,QAAQ;AAC5C,QAAM,UAAU,gBAAgB,QAAQ,UAAU,aAAa;AAC/D,QAAM,UAAU,gBAAgB,QAAQ,UAAU,eAAe,QAAQ,KAAK;AAC9E,QAAM,UAAU,gBAAgB,QAAQ,UAAU,eAAe,QAAQ,KAAK;AAC9E,QAAM,SAAS;AAAA,IACb,QAAQ;AAAA,IACR,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV;AACA,QAAM,SAAS,OAAO,UAAU,aAAa,MAAM;AACnD,SAAO;AAAA,IACL,gBAAgB,OAAO;AAAA,IACvB,eAAe,OAAO;AAAA,EACxB;AACF;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,sBAAsB,OAAO,QAAQ,cAAc,qBAAqB,mBAAmB;AAClG,QAAM,SAAS,gBAAgB,OAAO,SAAS,wBAAwB,OAAO;AAC9E,QAAM,UAAU,gBAAgB,QAAQ,UAAU,sBAAsB;AACxE,QAAM,gBAAgB,gBAAgB,cAAc,gBAAgB,wBAAwB,QAAQ,KAAK;AACzG,QAAM,uBAAuB,oBAAoB,IAAI,CAAC,GAAG,MAAM,gBAAgB,GAAG,UAAU,KAAK,wBAAwB,OAAO,CAAC;AACjI,QAAM,SAAS;AAAA,IACb,OAAO;AAAA,IACP,QAAQ;AAAA,IACR,cAAc;AAAA,IACd,qBAAqB;AAAA,EACvB;AACA,QAAM,QAAQ,EAAE,kBAAkB;AAClC,SAAO,OAAO,UAAU,sBAAsB,QAAQ,KAAK;AAC7D;AACA,IAAI,uBAAuB,GAAG,EAAE,sBAAsB,CAAC;AAGvD,SAAS,MAAM,OAAO,cAAc,OAAO;AACzC,QAAM,OAAO,cAAc,KAAK;AAChC,MAAI,SAAS;AACb,MAAI,SAAS,QAAQ,UAAU,WAAW;AACxC,aAAS,IAAI,aAAa,IAAI;AAAA,EAChC,WAAW,UAAU,SAAS;AAC5B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,WAAW,UAAU,QAAQ;AAC3B,aAAS,IAAI,WAAW,IAAI;AAAA,EAC9B,OAAO;AACL,UAAM,IAAI,MAAM,qBAAqB,OAAO;AAAA,EAC9C;AACA,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,WAAO,KAAK,aAAa;AAAA,EAC3B;AACA,SAAO,OAAO,WAAW,QAAQ,OAAO,KAAK;AAC/C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,IAAI,aAAa,QAAQ,oBAAoB,CAAC;AAC9C,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,OAAO,cAAc,OAAO,WAAW,MAAM;AACvD,SAAK,OAAO;AACZ,SAAK,SAAS;AACd,SAAK,QAAQ;AACb,SAAK,UAAU;AACf,SAAK,YAAY;AACjB,QAAI,KAAK,WAAW;AAClB,WAAK,QAAQ,KAAK,OAAO,KAAK,SAAS;AACvC,WAAK,QAAQ,KAAK,OAAO,KAAK,SAAS;AAAA,IACzC;AACA,UAAM,YAAY,OAAO,OAAO,KAAK,OAAO;AAC5C,SAAK,SAAS,WAAW,KAAK,UAAU,SAAS,CAAC;AAAA,EACpD;AAAA,EACA,YAAY;AACV,QAAI,CAAC,MAAM,KAAK,OAAO,GAAG;AACxB,YAAM,QAAQ,KAAK;AACnB,WAAK,UAAU;AACf,aAAO;AAAA,IACT;AACA,QAAI,SAAS;AACb,QAAI,UAAU;AACd,WAAO,CAAC,SAAS;AACf,UAAI,IAAI,IAAI;AACZ,SAAG;AACD,aAAK,IAAI,KAAK,OAAO,IAAI;AACzB,aAAK,IAAI,KAAK,OAAO,IAAI;AACzB,YAAI,KAAK,KAAK,KAAK;AAAA,MACrB,SAAS,KAAK,KAAK,MAAM;AACzB,YAAM,OAAO,KAAK,KAAK,KAAK,KAAK,IAAI,CAAC,IAAI,CAAC;AAC3C,gBAAU,KAAK,OAAO,KAAK,SAAS,KAAK;AACzC,gBAAU,KAAK,OAAO,KAAK,SAAS,KAAK;AACzC,UAAI,CAAC,KAAK,aAAa,KAAK,iBAAiB,OAAO,GAAG;AACrD,kBAAU;AAAA,MACZ;AAAA,IACF;AACA,QAAI,CAAC,KAAK,aAAa,KAAK,iBAAiB,OAAO,GAAG;AACrD,WAAK,UAAU,KAAK,aAAa,OAAO;AAAA,IAC1C;AACA,WAAO,KAAK,aAAa,OAAO;AAAA,EAClC;AAAA,EACA,aAAa,OAAO;AAClB,QAAI,KAAK,SAAS,QAAQ,KAAK,UAAU,WAAW;AAClD,aAAO;AAAA,IACT;AACA,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA,EACA,iBAAiB,OAAO;AACtB,WAAO,SAAS,KAAK,SAAS,SAAS,KAAK;AAAA,EAC9C;AACF;AACA,IAAI,YAAY,MAAM;AAAA,EACpB,YAAY,OAAO,MAAM,OAAO,MAAM;AACpC,SAAK,QAAQ;AACb,SAAK,OAAO,IAAI;AAChB,SAAK,QAAQ;AACb,UAAM,YAAY,OAAO,OAAO,KAAK,OAAO;AAC5C,SAAK,QAAQ,WAAW,KAAK,UAAU,SAAS,CAAC;AACjD,SAAK,QAAQ,IAAI,YAAY,GAAG,GAAG,OAAO,OAAO,KAAK,MAAM,CAAC;AAC7D,QAAI,QAAQ,GAAG;AACb,WAAK,IAAI,QAAQ,IAAI;AAAA,IACvB,OAAO;AACL,WAAK,IAAI,QAAQ,IAAI;AAAA,IACvB;AACA,SAAK,IAAI,IAAI,KAAK,KAAK,IAAI,KAAK,CAAC;AAAA,EACnC;AAAA,EACA,YAAY;AACV,QAAI,IAAI,IAAI,IAAI,GAAG,GAAG;AACtB,WAAO,MAAM;AACX,SAAG;AACD,YAAI,KAAK,MAAM,UAAU;AACzB,YAAI,IAAI,KAAK,IAAI;AAAA,MACnB,SAAS,KAAK;AACd,WAAK,IAAI;AACT,WAAK,IAAI;AACT,WAAK,IAAI,QAAQ,KAAK;AACtB,WAAK,MAAM,KAAK,KAAK,KAAK,IAAI,IAAI,KAAK,IAAI,CAAC;AAC5C,UAAI,KAAK,MAAM;AACf,UAAI,IAAI,MAAM,KAAK,IAAI,CAAC,IAAI,IAAI;AAC9B;AAAA,MACF;AAAA,IACF;AACA,QAAI,IAAI,KAAK,OAAO,KAAK,IAAI;AAC7B,QAAI,KAAK,QAAQ,GAAG;AAClB,WAAK,KAAK,IAAI,KAAK,MAAM,GAAG,IAAI,KAAK,KAAK;AAAA,IAC5C;AACA,WAAO,KAAK,aAAa,CAAC;AAAA,EAC5B;AAAA,EACA,aAAa,OAAO;AAClB,QAAI,KAAK,UAAU,WAAW;AAC5B,aAAO;AAAA,IACT;AACA,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AACF;AACA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,OAAO,GAAG,OAAO,GAAG,OAAO,MAAM;AAC3C,SAAK,iBAAiB,MAAM,KAAK,SAAS,QAAQ,KAAK,UAAU;AACjE,SAAK,MAAM;AACX,SAAK,QAAQ,OAAO;AACpB,SAAK,QAAQ;AACb,QAAI,QAAQ,MAAM;AAChB,aAAO,KAAK,OAAO;AAAA,IACrB;AACA,QAAI,OAAO,SAAS,UAAU;AAC5B,aAAO,KAAK,SAAS;AAAA,IACvB;AACA,QAAI,CAAC,KAAK,eAAe,KAAK,KAAK,SAAS,GAAG;AAC7C,YAAM,IAAI,MAAM,0BAA0B,UAAU,kCAAkC;AAAA,IACxF;AACA,SAAK,SAAS,WAAW,KAAK,IAAI;AAAA,EACpC;AAAA,EACA,aAAa,OAAO;AAClB,QAAI,KAAK,eAAe,GAAG;AACzB,aAAO;AAAA,IACT;AACA,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB;AAAA,EACA,YAAY;AACV,WAAO,KAAK,aAAa,KAAK,MAAM,KAAK,QAAQ,KAAK,OAAO,CAAC;AAAA,EAChE;AACF;AAGA,SAAS,aAAa,OAAO,OAAO,OAAO,GAAG,QAAQ,WAAW,MAAM;AACrE,MAAI,QAAQ,MAAM;AAChB,WAAO;AAAA,EACT;AACA,MAAI,SAAS,MAAM;AACjB,YAAQ;AAAA,EACV;AACA,MAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,UAAM,IAAI,MAAM,yBAAyB,OAAO;AAAA,EAClD;AACA,QAAM,SAAS,IAAI,UAAU,OAAO,MAAM,OAAO,IAAI;AACrD,QAAM,MAAM,OAAO,OAAO,KAAK;AAC/B,WAAS,IAAI,GAAG,IAAI,IAAI,OAAO,QAAQ,KAAK;AAC1C,QAAI,OAAO,KAAK,OAAO,UAAU;AAAA,EACnC;AACA,SAAO,IAAI,SAAS;AACtB;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,cAAc,OAAO,QAAQ,GAAG,SAAS,GAAG,OAAO,MAAM;AAChE,MAAI,SAAS,QAAQ,UAAU,QAAQ;AACrC,UAAM,IAAI,MAAM,yBAAyB,OAAO;AAAA,EAClD;AACA,QAAM,YAAY,IAAI,YAAY,OAAO,QAAQ,OAAO,OAAO,IAAI;AACnE,QAAM,MAAM,OAAO,OAAO,KAAK;AAC/B,WAAS,IAAI,GAAG,IAAI,IAAI,OAAO,QAAQ,KAAK;AAC1C,QAAI,OAAO,KAAK,UAAU,UAAU;AAAA,EACtC;AACA,SAAO,IAAI,SAAS;AACtB;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,sBAAsB,OAAO,OAAO,MAAM;AACjD,MAAI,SAAS,QAAQ,UAAU,QAAQ;AACrC,UAAM,IAAI,MAAM,yBAAyB,OAAO;AAAA,EAClD;AACA,SAAO,aAAa,OAAO,GAAG,GAAG,OAAO,IAAI;AAC9C;AACA,IAAI,uBAAuB,GAAG,EAAE,sBAAsB,CAAC;AAGvD,SAAS,eAAe,OAAO,SAAS,GAAG,SAAS,GAAG,QAAQ,WAAW,MAAM;AAC9E,QAAM,MAAM,OAAO,OAAO,KAAK;AAC/B,QAAM,SAAS,IAAI,cAAc,QAAQ,QAAQ,MAAM,IAAI;AAC3D,WAAS,IAAI,GAAG,IAAI,IAAI,OAAO,QAAQ,KAAK;AAC1C,QAAI,OAAO,KAAK,OAAO,UAAU;AAAA,EACnC;AACA,SAAO,IAAI,SAAS;AACtB;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,MAAM,OAAO,MAAM,QAAQ,GAAG,QAAQ,WAAW;AACxD,MAAI,UAAU,GAAG;AACf,UAAM,IAAI,MAAM,4BAA4B;AAAA,EAC9C;AACA,QAAM,QAAQ,EAAE,OAAO,MAAM,MAAM,OAAO,MAAM;AAChD,SAAO,OAAO,UAAU,OAAO,CAAC,GAAG,KAAK;AAC1C;AAGA,SAAS,YAAY,GAAG;AACtB,QAAM,KAAK,gBAAgB,GAAG,KAAK,YAAY;AAC/C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,YAAY,MAAM;AAC5C;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,SAAS,GAAG,MAAM;AACzB,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM,KAAK;AAC3B,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,GAAG;AACrB,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,qDAAqD,GAAG,OAAO;AAC3F,SAAO,QAAQ,IAAI,CAAC;AACtB;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,WAAW,GAAG,MAAM;AAC3B,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,qDAAqD,GAAG,OAAO;AAC3F,SAAO,QAAQ,IAAI,IAAI;AACzB;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,WAAW,GAAG,MAAM;AAC3B,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,qDAAqD,GAAG,OAAO;AAC3F,SAAO,QAAQ,IAAI,IAAI;AACzB;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,WAAW,GAAG,MAAM;AAC3B,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,qDAAqD,GAAG,OAAO;AAC3F,SAAO,QAAQ,IAAI,IAAI;AACzB;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,SAAS,GAAG,EAAE,OAAO,CAAC;AAG1B,SAAS,OAAO,GAAG;AACjB,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS,SAAS;AACrD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,OAAO,MAAM;AACvC;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,iBAAiB,GAAG,iBAAiB,iBAAiB,SAAS,MAAM,WAAW,CAAC,GAAG,CAAC,GAAG,aAAa,QAAQ;AACpH,QAAM,KAAK,gBAAgB,GAAG,KAAK,iBAAiB;AACpD,QAAM,mBAAmB,gBAAgB,iBAAiB,mBAAmB,iBAAiB;AAC9F,QAAM,mBAAmB,gBAAgB,iBAAiB,mBAAmB,iBAAiB;AAC9F,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,MAAI,eAAe,QAAQ;AACzB,UAAM,IAAI,MAAM,oFAAoF;AAAA,EACtG;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,gEAAgE,IAAI,OAAO;AACxG,SAAO,iBAAiB,SAAS,GAAG,MAAM,2EAA2E,iBAAiB,OAAO;AAC7I,SAAO,iBAAiB,SAAS,GAAG,MAAM,2EAA2E,iBAAiB,OAAO;AAC7I,SAAO,iBAAiB,MAAM,OAAO,GAAG,MAAM,yFAAyF,iBAAiB,MAAM,KAAK;AACnK,SAAO,iBAAiB,MAAM,OAAO,GAAG,MAAM,yFAAyF,iBAAiB,MAAM,KAAK;AACnK,QAAM,aAAa,iBAAiB,MAAM;AAC1C,QAAM,oBAAoB,iBAAiB,MAAM;AACjD,SAAO,iBAAiB,MAAM,OAAO,aAAa,mBAAmB,MAAM,6EAA6E,aAAa,8BAA8B,iBAAiB,MAAM,KAAK;AAC/N,QAAM,YAAY,gBAAgB,KAAK,kBAAkB,SAAS,MAAM,YAAY,QAAQ;AAC5F,QAAM,kBAAkB;AACxB,QAAM,MAAM,OAAO,WAAW,kBAAkB,iBAAiB,SAAS,UAAU;AACpF,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,eAAe,gBAAgB,GAAG,GAAG;AACnC,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW;AAC9C,SAAO,GAAG,UAAU,GAAG,OAAO,MAAM,kDAAkD,GAAG,iBAAiB,GAAG,SAAS;AACtH,SAAO,GAAG,SAAS,GAAG,MAAM,qCAAqC,GAAG,SAAS;AAC7E,SAAO,GAAG,SAAS,GAAG,MAAM,qCAAqC,GAAG,SAAS;AAC7E,QAAM,QAAQ,MAAM,GAAG,KAAK;AAC5B,QAAM,QAAQ,MAAM,GAAG,KAAK;AAC5B,QAAM,OAAO,IAAI,IAAI,KAAK;AAC1B,MAAI,aAAa;AACjB,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,QAAI,CAAC,KAAK,IAAI,MAAM,EAAE,GAAG;AACvB;AAAA,IACF;AAAA,EACF;AACA,QAAM,UAAU,IAAI,aAAa,CAAC,UAAU,GAAG,GAAG,KAAK;AACvD,QAAM,UAAU,IAAI,aAAa,CAAC,UAAU,GAAG,OAAO;AACtD,WAAS,IAAI,GAAG,KAAK,GAAG,IAAI,MAAM,QAAQ,KAAK;AAC7C,QAAI,CAAC,KAAK,IAAI,MAAM,EAAE,GAAG;AACvB,cAAQ,OAAO,MAAM,MAAM;AAC3B,cAAQ,OAAO,MAAM;AACrB;AAAA,IACF;AAAA,EACF;AACA,SAAO,CAAC,QAAQ,SAAS,GAAG,QAAQ,SAAS,CAAC;AAChD;AACA,IAAI,iBAAiB;AAGrB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,SAAS;AACnD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,MAAM,GAAG;AAChB,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,SAAS,GAAG,OAAO,MAAM;AAChC,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,mDAAmD,GAAG,aAAa;AAC/F,SAAO,MAAM,IAAI,CAAC,KAAK,GAAG,CAAC,IAAI,CAAC;AAClC;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,SAAS,GAAG,OAAO,MAAM;AAChC,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,mDAAmD,GAAG,aAAa;AAC/F,SAAO,MAAM,IAAI,OAAO,IAAI;AAC9B;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,SAAS,GAAG,OAAO,MAAM;AAChC,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,mDAAmD,GAAG,aAAa;AAC/F,SAAO,MAAM,IAAI,OAAO,IAAI;AAC9B;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,SAAS,GAAG,OAAO,MAAM;AAChC,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,SAAS,GAAG,MAAM,mDAAmD,GAAG,aAAa;AAC/F,SAAO,MAAM,IAAI,OAAO,IAAI;AAC9B;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,SAAS,QAAQ,MAAM,IAAI;AAClC,QAAM,UAAU,gBAAgB,QAAQ,UAAU,WAAW,SAAS;AACtE,MAAI,QAAQ,IAAI;AACd,UAAM,QAAQ,OAAO;AAAA,EACvB;AACA,MAAI,QAAQ,QAAQ,OAAO,GAAG;AAC5B,UAAM,MAAM,4EAA4E,QAAQ,oBAAoB,KAAK;AAAA,EAC3H;AACA,QAAM,SAAS,EAAE,QAAQ,QAAQ;AACjC,QAAM,QAAQ,EAAE,IAAI;AACpB,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,KAAK,QAAQ;AACpB,SAAO,OAAO,UAAU,aAAa,MAAM,6DAA6D,OAAO,QAAQ;AACvH,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,MAAM,QAAQ;AACrB,SAAO,OAAO,UAAU,aAAa,MAAM,8DAA8D,OAAO,QAAQ;AACxH,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,SAAO,OAAO,UAAU,MAAM,MAAM;AACtC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,OAAO,QAAQ;AACtB,QAAM,qBAAqB,OAAO,MAAM,OAAO,MAAM,SAAS;AAC9D,QAAM,QAAQ,OAAO,OAAO;AAC5B,MAAI;AACJ,MAAI,sBAAsB,GAAG;AAC3B,UAAM,eAAe,QAAQ,QAAQ,CAAC,OAAO,kBAAkB,CAAC;AAChE,UAAM,KAAK,YAAY;AAAA,EACzB,OAAO;AACL,UAAM,cAAc,CAAC,OAAO,KAAK,qBAAqB,EAAE;AACxD,UAAM,YAAY,QAAQ,KAAK,MAAM,GAAG,CAAC,OAAO,kBAAkB,CAAC;AACnE,UAAM,YAAY,QAAQ,KAAK,MAAM,GAAG,CAAC,OAAO,kBAAkB,CAAC;AACnE,UAAM,gBAAgB,QAAQ,MAAM,WAAW,CAAC,GAAG,CAAC,GAAG,CAAC,OAAO,qBAAqB,CAAC,CAAC,GAAG,CAAC;AAC1F,UAAM,gBAAgB,IAAI,QAAQ,MAAM,WAAW,CAAC,GAAG,CAAC,GAAG,CAAC,OAAO,qBAAqB,CAAC,CAAC,GAAG,CAAC,GAAG,OAAO,EAAE,CAAC;AAC3G,UAAM,IAAI,OAAO,CAAC,WAAW,aAAa,GAAG,CAAC;AAC9C,UAAM,IAAI,OAAO,CAAC,WAAW,aAAa,GAAG,CAAC;AAC9C,UAAM,eAAe,QAAQ,QAAQ,GAAG,CAAC,GAAG,CAAC,YAAY,IAAI,YAAY,EAAE,CAAC;AAC5E,UAAM,KAAK,YAAY;AAAA,EACzB;AACA,QAAM,KAAK,GAAG;AACd,MAAI,OAAO,SAAS,KAAK,OAAO,MAAM,OAAO,GAAG;AAC9C,UAAM,OAAO;AACb,UAAM,SAAS,OAAO,MAAM;AAC5B,UAAM,QAAQ,KAAK,CAAC,QAAQ,IAAI,MAAM,KAAK,QAAQ,IAAI,MAAM,EAAE,CAAC;AAChE,SAAK,QAAQ;AAAA,EACf;AACA,SAAO;AACT;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,OAAO,GAAG,iBAAiB,OAAO,GAAG;AAC5C,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO;AAC1C,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,OAAO,EAAE,iBAAiB,KAAK;AACrC,SAAO,OAAO,UAAU,QAAQ,QAAQ,IAAI;AAC9C;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,QAAQ,WAAW;AAChC,SAAO,OAAO,UAAU,WAAW,MAAM,mDAAmD,OAAO,OAAO;AAC1G,MAAI,qBAAqB,OAAO,MAAM,OAAO,MAAM,SAAS;AAC5D,QAAM,QAAQ,OAAO,OAAO;AAC5B,MAAI;AACJ,MAAI,aAAa,QAAQ,YAAY,oBAAoB;AACvD,UAAM,QAAQ,OAAO,MAAM,IAAI,CAAC,MAAM,CAAC;AACvC,UAAM,OAAO,OAAO,MAAM,IAAI,CAAC,MAAM,CAAC;AACtC,SAAK,OAAO,MAAM,SAAS,KAAK;AAChC,oBAAgB,MAAM,QAAQ,OAAO,IAAI;AACzC,yBAAqB;AAAA,EACvB,WAAW,aAAa,QAAQ,YAAY,oBAAoB;AAC9D,UAAM,aAAa,OAAO,MAAM,IAAI,CAAC,MAAM,CAAC;AAC5C,eAAW,OAAO,MAAM,SAAS,KAAK,YAAY;AAClD,oBAAgB,OAAO,CAAC,QAAQ,MAAM,UAAU,CAAC,GAAG,OAAO,MAAM,SAAS,CAAC;AAC3E,yBAAqB;AAAA,EACvB,OAAO;AACL,oBAAgB;AAAA,EAClB;AACA,QAAM,aAAa,UAAU,aAAa;AAC1C,QAAM,eAAe,QAAQ,QAAQ,eAAe,UAAU,GAAG,CAAC,OAAO,kBAAkB,CAAC;AAC5F,QAAM,MAAM,IAAI,YAAY;AAC5B,QAAM,OAAO,KAAK,MAAM,qBAAqB,CAAC,IAAI;AAClD,QAAM,aAAa,KAAK,GAAG;AAC3B,QAAM,aAAa,KAAK,GAAG;AAC3B,QAAM,uBAAuB,MAAM,YAAY,CAAC,MAAM,qBAAqB,IAAI,GAAG,WAAW,MAAM,SAAS,CAAC;AAC7G,QAAM,uBAAuB,MAAM,YAAY,CAAC,MAAM,qBAAqB,IAAI,GAAG,WAAW,MAAM,SAAS,CAAC;AAC7G,QAAM,cAAc,cAAc,MAAM,MAAM;AAC9C,cAAY,cAAc,MAAM,SAAS,KAAK;AAC9C,SAAO,QAAQ,QAAQ,qBAAqB,IAAI,qBAAqB,EAAE,GAAG,WAAW;AACvF;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,mBAAmB,GAAG,GAAG;AAChC,MAAI,KAAK,gBAAgB,GAAG,KAAK,mBAAmB;AACpD,MAAI,KAAK,gBAAgB,GAAG,KAAK,mBAAmB;AACpD,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,6BAA2B,GAAG,OAAO,GAAG,KAAK;AAC7C,QAAM,SAAS,EAAE,GAAG,IAAI,GAAG,GAAG;AAC9B,QAAM,QAAQ,CAAC;AACf,SAAO,OAAO,UAAU,mBAAmB,QAAQ,KAAK;AAC1D;AACA,IAAI,oBAAoB,GAAG,EAAE,mBAAmB,CAAC;AAGjD,SAAS,SAAS,GAAG,MAAM;AACzB,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AACjE,SAAO,QAAQ,IAAI,aAAa,GAAG,OAAO,IAAI,EAAE,QAAQ;AAC1D;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,OAAO,SAAS,OAAO,GAAG;AACjC,QAAM,WAAW,qBAAqB,SAAS,WAAW,SAAS,mBAAmB;AACtF,SAAO,SAAS,UAAU,GAAG,MAAM,sCAAsC;AACzE,MAAI,SAAS,SAAS,GAAG;AACvB,WAAO,QAAQ,SAAS,GAAG,MAAM,MAAM,oCAAoC;AAAA,EAC7E;AACA,QAAM,SAAS;AACf,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,GAAG,QAAQ,GAAG;AAC3B,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,MAAM;AACtB,SAAO,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC7C;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,cAAc,GAAG,OAAO,KAAK,SAAS,YAAY,GAAG,UAAU,GAAG,eAAe,GAAG,cAAc,GAAG,iBAAiB,GAAG;AAChI,QAAM,KAAK,gBAAgB,GAAG,KAAK,gBAAgB,mBAAmB;AACtE,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,SAAO,OAAO,UAAU,cAAc,QAAQ,KAAK;AACrD;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,KAAK,GAAG;AACf,QAAM,KAAK,gBAAgB,GAAG,KAAK,OAAO,SAAS;AACnD,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,SAAO,OAAO,UAAU,KAAK,MAAM;AACrC;AACA,IAAI,MAAM,GAAG,EAAE,KAAK,CAAC;AAGrB,SAAS,SAAS,QAAQ,OAAO;AAC/B,gBAAc,MAAM;AACpB,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM,oDAAoD;AAAA,EACtE;AACA,QAAM,QAAQ;AACd,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,SAAS,SAAS,QAAQ,OAAO,OAAO;AACtC,gBAAc,MAAM;AACpB,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,KAAK,cAAc,WAAW,GAAG;AAC5D,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AACA,MAAI,cAAc,WAAW,KAAK,SAAS,MAAM;AAC/C,UAAM,IAAI,MAAM,8EAA8E;AAAA,EAChG;AACA,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,SAAS,SAAS,QAAQ,OAAO,OAAO;AACtC,gBAAc,MAAM;AACpB,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,gDAAgD;AAAA,EAClE;AACA,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,KAAK,cAAc,WAAW,GAAG;AAC5D,UAAM,IAAI,MAAM,oEAAoE;AAAA,EACtF;AACA,MAAI,cAAc,WAAW,KAAK,SAAS,MAAM;AAC/C,UAAM,IAAI,MAAM,yEAAyE;AAAA,EAC3F;AACA,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,SAAS,SAAS,QAAQ,OAAO,OAAO;AACtC,gBAAc,MAAM;AACpB,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,gDAAgD;AAAA,EAClE;AACA,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,KAAK,cAAc,WAAW,GAAG;AAC5D,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AACA,MAAI,cAAc,WAAW,KAAK,SAAS,MAAM;AAC/C,UAAM,IAAI,MAAM,yEAAyE;AAAA,EAC3F;AACA,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,SAAS,SAAS,QAAQ,OAAO,OAAO;AACtC,gBAAc,MAAM;AACpB,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,gBAAgB,WAAW,QAAQ,KAAK;AAC9C,MAAI,cAAc,WAAW,KAAK,cAAc,WAAW,GAAG;AAC5D,UAAM,IAAI,MAAM,wEAAwE;AAAA,EAC1F;AACA,MAAI,cAAc,WAAW,KAAK,SAAS,MAAM;AAC/C,UAAM,IAAI,MAAM,yEAAyE;AAAA,EAC3F;AACA,UAAQ,SAAS;AACjB,SAAO,WAAW,QAAQ,OAAO,eAAe,KAAK;AACvD;AAGA,SAAS,MAAM,GAAG,IAAI,GAAG,SAAS,MAAM;AACtC,QAAM,KAAK,gBAAgB,GAAG,KAAK,MAAM;AACzC,MAAI,GAAG,SAAS,GAAG;AACjB,UAAM,IAAI,MAAM,oDAAoD;AAAA,EACtE;AACA,QAAM,UAAU,GAAG,MAAM,GAAG,MAAM,SAAS;AAC3C,MAAI,IAAI,GAAG;AACT,UAAM,IAAI,MAAM,6CAA6C,GAAG;AAAA,EAClE;AACA,MAAI,IAAI,SAAS;AACf,UAAM,IAAI,MAAM,uDAAuD,oBAAoB,GAAG;AAAA,EAChG;AACA,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,GAAG,OAAO;AAC1B,QAAM,CAAC,QAAQ,OAAO,IAAI,OAAO,UAAU,MAAM,QAAQ,KAAK;AAC9D,SAAO,EAAE,QAAQ,QAAQ;AAC3B;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,iBAAiB,OAAO,QAAQ,GAAG,SAAS,GAAG,OAAO,MAAM;AACnE,MAAI,SAAS,QAAQ,UAAU,QAAQ;AACrC,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACA,QAAM,YAAY,IAAI,YAAY,OAAO,QAAQ,OAAO,MAAM,IAAI;AAClE,QAAM,MAAM,OAAO,OAAO,KAAK;AAC/B,WAAS,IAAI,GAAG,IAAI,IAAI,OAAO,QAAQ,KAAK;AAC1C,QAAI,OAAO,KAAK,UAAU,UAAU;AAAA,EACtC;AACA,SAAO,IAAI,SAAS;AACtB;AACA,IAAI,kBAAkB,GAAG,EAAE,iBAAiB,CAAC;AAG7C,SAAS,QAAQ,GAAG,OAAO,GAAG;AAC5B,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU,mBAAmB;AAChE,SAAO,GAAG,OAAO,GAAG,MAAM,sCAAsC;AAChE,QAAM,SAAS,EAAE,GAAG,GAAG;AACvB,QAAM,QAAQ,EAAE,KAAK;AACrB,QAAM,CAAC,QAAQ,OAAO,IAAI,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAChE,SAAO,EAAE,QAAQ,QAAQ;AAC3B;AACA,IAAI,SAAS,GAAG,EAAE,QAAQ,CAAC;AAG3B,SAAS,oBAAoB,GAAG,YAAY,aAAa;AACvD,QAAM,KAAK,gBAAgB,GAAG,KAAK,oBAAoB;AACvD,QAAM,cAAc,gBAAgB,YAAY,cAAc,sBAAsB,OAAO;AAC3F,SAAO,MAAM,WAAW,GAAG,MAAM,kCAAkC;AACnE,QAAM,SAAS,EAAE,GAAG,IAAI,YAAY,YAAY;AAChD,QAAM,QAAQ,EAAE,YAAY;AAC5B,SAAO,OAAO,UAAU,oBAAoB,QAAQ,KAAK;AAC3D;AACA,IAAI,qBAAqB,GAAG,EAAE,oBAAoB,CAAC;AAGnD,SAAS,SAAS,GAAG,OAAO,GAAG;AAC7B,QAAM,KAAK,gBAAgB,GAAG,KAAK,WAAW,mBAAmB;AACjE,SAAO,QAAQ,CAAC,GAAG,MAAM,UAAU,OAAO,GAAG,MAAM,QAAQ,MAAM,UAAU,oBAAoB,GAAG,MAAM,WAAW,GAAG,MAAM,SAAS;AACrI,QAAM,SAAS,EAAE,OAAO,GAAG;AAC3B,QAAM,QAAQ,EAAE,KAAK;AACrB,SAAO,OAAO,UAAU,QAAQ,QAAQ,KAAK;AAC/C;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,WAAW,gBAAgB,QAAQ;AAC1C,SAAO,aAAa,gBAAgB,QAAQ,OAAO;AACrD;AAGA,SAAS,SAAS,cAAc,YAAY,MAAM,MAAM,OAAO;AAC7D,SAAO,OAAO,aAAa,cAAc,WAAW,MAAM,KAAK;AACjE;AAGA,SAAS,UAAU,WAAW,UAAU;AACtC,QAAM,UAAU,CAAC;AACjB,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,QAAI,SAAS,IAAI;AACf,cAAQ,KAAK,CAAC;AAAA,IAChB;AAAA,EACF;AACA,QAAM,WAAW,OAAO,WAAW,OAAO;AAC1C,QAAM,MAAM,OAAO,CAAC,QAAQ,QAAQ,UAAU,MAAM,GAAG,OAAO;AAC9D,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,UAAM,MAAM,SAAS,WAAW,QAAQ,EAAE;AAC1C,UAAM,SAAS,IAAI,UAAU;AAC7B,QAAI,OAAO,IAAI,KAAK,MAAM;AAAA,EAC5B;AACA,SAAO,IAAI,SAAS;AACtB;AAGA,eAAe,YAAY,WAAW;AACpC,QAAM,aAAa,gBAAgB,WAAW,aAAa,cAAc,MAAM;AAC/E,QAAM,OAAO,MAAM,WAAW,KAAK;AACnC,QAAM,MAAM,UAAU,WAAW,OAAO,IAAI;AAC5C,MAAI,cAAc,YAAY;AAC5B,eAAW,QAAQ;AAAA,EACrB;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAGjB,eAAe,kBAAkB,SAAS,MAAM,MAAM;AACpD,QAAM,UAAU,gBAAgB,SAAS,UAAU,UAAU;AAC7D,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,YAAY,MAAM;AAC9D,QAAM,WAAW,QAAQ,OAAO,IAAI;AACpC,QAAM,UAAU,MAAM;AACtB,QAAM,cAAc,QAAQ;AAC5B,SAAO,UAAU,GAAG,MAAM,uBAAuB;AACjD,oBAAkB,YAAY,MAAM,UAAU,WAAW,OAAO,GAAG,MAAM,OAAO,mEAAmE;AACnJ,MAAI,cAAc;AAClB,WAAS,IAAI,UAAU,IAAI,WAAW,SAAS,KAAK;AAClD,mBAAe,YAAY;AAAA,EAC7B;AACA,QAAM,oBAAoB,YAAY,MAAM,GAAG,QAAQ,EAAE,OAAO,CAAC,WAAW,GAAG,YAAY,MAAM,WAAW,OAAO,CAAC;AACpH,QAAM,iBAAiB,QAAQ,SAAS,iBAAiB;AACzD,QAAM,eAAe,QAAQ,OAAO,CAAC,EAAE,CAAC;AACxC,QAAM,oBAAoB,MAAM,WAAW,YAAY;AACvD,QAAM,UAAU,QAAQ,mBAAmB,CAAC,CAAC,CAAC;AAC9C,QAAM,MAAM,OAAO,gBAAgB,SAAS,QAAQ;AACpD,MAAI,YAAY,SAAS;AACvB,YAAQ,QAAQ;AAAA,EAClB;AACA,MAAI,SAAS,OAAO;AAClB,UAAM,QAAQ;AAAA,EAChB;AACA,UAAQ,QAAQ;AAChB,iBAAe,QAAQ;AACvB,eAAa,QAAQ;AACrB,oBAAkB,QAAQ;AAC1B,SAAO;AACT;AACA,IAAI,mBAAmB;AAGvB,SAAS,eAAe,GAAG,GAAG,OAAO,OAAO,aAAa,MAAM;AAC7D,QAAM,KAAK,gBAAgB,GAAG,KAAK,eAAe;AAClD,QAAM,KAAK,gBAAgB,GAAG,KAAK,eAAe;AAClD,QAAM,SAAS,gBAAgB,OAAO,SAAS,eAAe;AAC9D,mBAAiB,IAAI,EAAE;AACvB,SAAO,YAAY,GAAG,OAAO,GAAG,KAAK,GAAG,MAAM,2BAA2B;AACzE,QAAM,MAAM,OAAO,CAAC;AACpB,QAAM,gBAAgB,IAAI,KAAK,MAAM;AACrC,MAAI,SAAS,IAAI,IAAI,IAAI,EAAE,GAAG,aAAa;AAC3C,MAAI,YAAY;AACd,WAAO,SAAS,MAAM,MAAM,gDAAgD;AAC5E,UAAM,QAAQ,gBAAgB,OAAO,QAAQ,eAAe;AAC5D,aAAS,IAAI,QAAQ,IAAI,KAAK,IAAI,QAAQ,KAAK,CAAC,CAAC;AAAA,EACnD;AACA,SAAO,KAAK,IAAI,MAAM;AACxB;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,WAAW,SAAS,SAAS,OAAO;AAC3C,QAAM,WAAW,gBAAgB,SAAS,WAAW,aAAa,OAAO;AACzE,QAAM,WAAW,gBAAgB,SAAS,WAAW,WAAW;AAChE,gBAAc,UAAU,UAAU,KAAK;AACvC,QAAM,SAAS,EAAE,SAAS,UAAU,SAAS,SAAS;AACtD,QAAM,QAAQ,EAAE,MAAM;AACtB,SAAO,OAAO,UAAU,WAAW,QAAQ,KAAK;AAClD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,eAAe,eAAe,cAAc,aAAa,eAAe;AAC/E,MAAI,cAAc,UAAU,SAAS;AACnC,UAAM,IAAI,MAAM,8EAA8E,cAAc,QAAQ;AAAA,EACtH;AACA,MAAI,cAAc,OAAO,GAAG;AAC1B,UAAM,IAAI,MAAM,sEAAsE,cAAc,QAAQ;AAAA,EAC9G;AACA,QAAM,WAAW,cAAc,OAAO,IAAI,cAAc,MAAM,KAAK;AACnE,QAAM,UAAU,cAAc,OAAO,IAAI,cAAc,MAAM,KAAK;AAClE,MAAI,YAAY,WAAW,SAAS;AAClC,UAAM,IAAI,MAAM,kDAAkD,YAAY,sBAAsB,UAAU;AAAA,EAChH;AACA,QAAM,YAAY,aAAa;AAC/B,MAAI,EAAE,aAAa,SAAS,KAAK,aAAa,SAAS,KAAK,cAAc,WAAW;AACnF,UAAM,IAAI,MAAM,oCAAoC,aAAa,2BAA2B,WAAW;AAAA,EACzG;AACA,MAAI,aAAa,UAAU,cAAc,OAAO;AAC9C,UAAM,IAAI,MAAM,mDAAmD;AAAA,EACrE;AACF;AAGA,SAAS,eAAe,eAAe,cAAc,aAAa,eAAe,GAAG;AAClF,QAAM,iBAAiB,gBAAgB,eAAe,iBAAiB,iBAAiB,OAAO;AAC/F,QAAM,gBAAgB,gBAAgB,cAAc,gBAAgB,iBAAiB,mBAAmB;AACxG,QAAM,gBAAgB,gBAAgB,cAAc,gBAAgB,iBAAiB,cAAc,KAAK;AACxG,iBAAe,gBAAgB,eAAe,aAAa,aAAa;AACxE,QAAM,SAAS;AAAA,IACb,eAAe;AAAA,IACf,cAAc;AAAA,IACd,cAAc;AAAA,EAChB;AACA,QAAM,QAAQ,EAAE,YAAY;AAC5B,SAAO,OAAO,UAAU,eAAe,QAAQ,KAAK;AACtD;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,UAAU,GAAG,SAAS;AAC7B,QAAM,WAAW,gBAAgB,SAAS,WAAW,YAAY,OAAO;AACxE,QAAM,KAAK,gBAAgB,GAAG,KAAK,YAAY,mBAAmB;AAClE,QAAM,SAAS,EAAE,QAAQ,IAAI,SAAS,SAAS;AAC/C,SAAO,OAAO,UAAU,UAAU,MAAM;AAC1C;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,cAAc,GAAG,YAAY;AACpC,MAAI,cAAc,MAAM;AACtB,WAAO,EAAE,MAAM,MAAM;AAAA,EACvB;AACA,MAAI,YAAY,EAAE,OAAO,UAAU,GAAG;AACpC,WAAO;AAAA,EACT;AACA,MAAI,EAAE,MAAM,WAAW,WAAW,QAAQ;AACxC,UAAM,eAAe,CAAC;AACtB,aAAS,IAAI,GAAG,IAAI,EAAE,MAAM,QAAQ,KAAK;AACvC,UAAI,WAAW,MAAM,QAAQ,EAAE,MAAM,MAAM,MAAM;AAC/C,qBAAa,KAAK,EAAE,MAAM,EAAE;AAAA,MAC9B,OAAO;AACL,qBAAa,KAAK,WAAW,EAAE;AAAA,MACjC;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAGA,SAAS,SAAS,GAAG,MAAM,YAAY,MAAM;AAC3C,QAAM,KAAK,gBAAgB,GAAG,KAAK,SAAS;AAC5C,SAAO,GAAG,UAAU,WAAW,MAAM,gFAAgF,GAAG,uBAAuB;AAC/I,SAAO,QAAQ,KAAK,OAAO,GAAG,MAAM,qDAAqD,OAAO;AAChG,MAAI,SAAS,GAAG;AACd,WAAO,aAAa,SAAS,GAAG,MAAM,IAAI;AAAA,EAC5C;AACA,QAAM,cAAc,cAAc,IAAI,UAAU;AAChD,QAAM,WAAW,IAAI;AACrB,QAAM,aAAa,IAAI,MAAM,KAAK,cAAc,aAAa,GAAG,GAAG,WAAW,IAAI,GAAG,QAAQ,CAAC,GAAG,QAAQ;AACzG,SAAO,IAAI,IAAI,UAAU;AAC3B;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,oBAAoB,OAAO;AAClC,SAAO,KAAK,MAAM,KAAK,IAAI,GAAG,KAAK,KAAK,KAAK,IAAI,KAAK,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC;AACzE;AACA,SAAS,aAAa,cAAc,GAAG,GAAG;AACxC,QAAM,OAAO,IAAI,eAAe;AAChC,QAAM,YAAY,IAAI,aAAa,YAAY;AAC/C,WAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,UAAM,SAAS,IAAI,KAAK,KAAK,KAAK,eAAe,OAAO;AACxD,cAAU,KAAK,IAAI,IAAI,KAAK,IAAI,MAAM;AAAA,EACxC;AACA,SAAO,SAAS,WAAW,SAAS;AACtC;AAGA,eAAe,aAAa,aAAa,SAAS,IAAI,GAAG;AACvD,QAAM,eAAe,gBAAgB,aAAa,eAAe,QAAQ;AACzE,QAAM,WAAW,gBAAgB,SAAS,WAAW,QAAQ;AAC7D,SAAO,aAAa,OAAO,GAAG,MAAM,uEAAuE,aAAa,MAAM;AAC9H,SAAO,aAAa,OAAO,MAAM,SAAS,MAAM,MAAM,mFAAmF,aAAa,yBAAyB,SAAS,MAAM;AAC9L,oBAAkB,aAAa,MAAM,MAAM,GAAG,aAAa,MAAM,SAAS,CAAC,GAAG,SAAS,OAAO,yFAAyF;AACvL,QAAM,UAAU,aAAa,MAAM,aAAa,MAAM,SAAS;AAC/D,SAAO,IAAI,KAAK,KAAK,SAAS,MAAM,4EAA4E,qBAAqB,GAAG;AACxI,QAAM,kBAAkB,MAAM,aAAa,KAAK;AAChD,QAAM,cAAc,MAAM,SAAS,KAAK;AACxC,QAAM,CAAC,OAAO,IAAI,IAAI,CAAC,gBAAgB,SAAS,SAAS,OAAO;AAChE,QAAM,aAAa,uBAAuB,QAAQ,KAAK;AACvD,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,SAAS,IAAI;AACnB,UAAM,OAAO,gBAAgB,SAAS,QAAQ,SAAS,IAAI;AAC3D,UAAM,YAAY,CAAC;AACnB,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,gBAAU,KAAK,EAAE,OAAO,KAAK,IAAI,OAAO,EAAE,CAAC;AAAA,IAC7C;AACA,cAAU,KAAK,CAAC,GAAG,OAAO,GAAG,QAAQ,EAAE,KAAK;AAC5C,eAAW,KAAK;AAChB,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAI,UAAU,GAAG,UAAU,YAAY,IAAI;AACzC,mBAAW,KAAK;AAChB;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,MAAI,gBAAgB,cAAc;AAChC,iBAAa,QAAQ;AAAA,EACvB;AACA,MAAI,YAAY,UAAU;AACxB,aAAS,QAAQ;AAAA,EACnB;AACA,SAAO,OAAO,YAAY,SAAS,OAAO,MAAM;AAClD;AACA,IAAI,cAAc;AAGlB,IAAI,oBAAoB,CAAC;AACzBJ,UAAS,mBAAmB;AAAA,EAC1B,QAAQ,MAAM;AAAA,EACd,iBAAiB,MAAM;AAAA,EACvB,QAAQ,MAAM;AAChB,CAAC;AAGD,SAAS,sBAAsB,GAAG,IAAI,aAAa,SAAS,MAAM,aAAa,QAAQ,iBAAiB;AACtG,MAAI,MAAM;AACV,MAAI,EAAE,SAAS,GAAG;AAChB,UAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC;AAAA,EAC1D;AACA,MAAI,OAAO;AACX,MAAI,KAAK,SAAS,GAAG;AACnB,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC/D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,iEAAiE,IAAI,QAAQ;AAC1G,SAAO,KAAK,SAAS,GAAG,MAAM,8DAA8D,KAAK,QAAQ;AACzG,SAAO,YAAY,WAAW,GAAG,MAAM,mEAAmE,cAAc;AACxH,QAAM,UAAU,eAAe,SAAS,IAAI,MAAM,KAAK,IAAI,MAAM;AACjE,QAAM,WAAW,eAAe,SAAS,KAAK,MAAM,KAAK,KAAK,MAAM;AACpE,SAAO,YAAY,YAAY,IAAI,MAAM,4CAA4C,8CAA8C,YAAY,KAAK;AACpJ,SAAO,aAAa,YAAY,IAAI,MAAM,0CAA0C,iDAAiD,YAAY,MAAM;AACvJ,4BAA0B,mBAAmB,MAAM,eAAe;AAClE,QAAM,SAAS,EAAE,GAAG,KAAK,IAAI,KAAK;AAClC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY,iBAAiB,YAAY;AAC7E,SAAO,OAAO,UAAU,sBAAsB,QAAQ,KAAK;AAC7D;AACA,IAAI,uBAAuB,GAAG,EAAE,sBAAsB,CAAC;AAGvD,SAAS,qBAAqB,IAAI,GAAG,aAAa;AAChD,MAAI,eAAe,QAAQ,gBAAgB,UAAU;AACnD,WAAO;AAAA,EACT;AACA,MAAI,gBAAgB,QAAQ;AAC1B,WAAO,IAAI,IAAI,KAAK,CAAC,CAAC;AAAA,EACxB;AACA,QAAM,IAAI,MAAM,gDAAgD,cAAc;AAChF;AACA,SAAS,qBAAqB,MAAM,cAAc;AAChD,MAAI,MAAM;AACV,QAAM,aAAa,iBAAiB,KAAK,OAAO,aAAa,KAAK;AAClE,MAAI,WAAW,SAAS,GAAG;AACzB,UAAM,KAAK,KAAK,UAAU;AAAA,EAC5B;AACA,SAAO,QAAQ,KAAK,KAAK,KAAK;AAChC;AACA,SAAS,gBAAgB,GAAG,aAAa,wBAAwB,gBAAgB;AAC/E,MAAI,gBAAgB,UAAU;AAC5B,WAAO;AAAA,EACT,WAAW,gBAAgB,QAAQ;AACjC,WAAO,KAAK,CAAC;AAAA,EACf,WAAW,gBAAgB,OAAO;AAChC,WAAO,IAAI,CAAC;AAAA,EACd,WAAW,gBAAgB,SAAS;AAClC,WAAO,MAAM,CAAC;AAAA,EAChB,WAAW,gBAAgB,SAAS;AAClC,WAAO,MAAM,GAAG,sBAAsB;AAAA,EACxC,WAAW,gBAAgB,aAAa;AACtC,WAAO,UAAU,GAAG,cAAc;AAAA,EACpC,WAAW,gBAAgB,WAAW;AACpC,WAAO,QAAQ,CAAC;AAAA,EAClB;AACA,QAAM,IAAI,MAAM,4BAA4B,cAAc;AAC5D;AACA,IAAI,aAAa,CAAC,eAAe,gBAAgB;AAC/C,QAAM,eAAe,gBAAgB;AACrC,SAAO,CAAC,gBAAgB,gBAAgB;AAC1C;AAGA,SAAS,aAAa,EAAE,GAAG,QAAQ,SAAS,KAAK,MAAM,aAAa,QAAQ,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB,MAAM,YAAY,cAAc,UAAU,wBAAwB,eAAe,GAAG;AACnM,gBAAc,eAAe;AAC7B,MAAI,WAAW,OAAO,MAAM,eAAe,WAAW,MAAM,OAAO;AACjE,WAAO,eAAe,QAAQ,MAAM,4CAA4C,uHAAuH;AACvM,QAAI,SAAS,OAAO,GAAG,QAAQ,SAAS,MAAM,YAAY,WAAW,eAAe;AACpF,QAAI,QAAQ,MAAM;AAChB,eAAS,KAAK,QAAQ,IAAI;AAAA,IAC5B;AACA,WAAO,gBAAgB,QAAQ,aAAa,wBAAwB,cAAc;AAAA,EACpF;AACA,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU,SAAS;AACtD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,UAAU,SAAS;AACrE,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,6DAA6D,IAAI,OAAO;AACrG,SAAO,QAAQ,SAAS,GAAG,MAAM,8DAA8D,QAAQ,OAAO;AAC9G,4BAA0B,gBAAgB,MAAM,eAAe;AAC/D,QAAM,gBAAgB,eAAe,SAAS,IAAI,MAAM,KAAK,IAAI,MAAM;AACvE,SAAO,QAAQ,MAAM,OAAO,eAAe,MAAM,oCAAoC,oDAAoD,QAAQ,MAAM,KAAK;AAC5J,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,uEAAuE,0BAA0B,YAAY;AAC9K,QAAM,WAAW,kBAAkB,IAAI,OAAO,QAAQ,OAAO,SAAS,WAAW,MAAM,eAAe;AACtG,MAAI;AACJ,MAAI,QAAQ,MAAM;AAChB,YAAQ,gBAAgB,MAAM,QAAQ,cAAc;AACpD,KAAC,KAAK,IAAI,eAAe,OAAO,EAAE;AAClC,QAAI,eAAe,QAAQ;AACzB,iCAA2B,SAAS,UAAU,MAAM,KAAK;AAAA,IAC3D,OAAO;AACL,aAAO,MAAM,MAAM,UAAU,GAAG,MAAM,2GAA2G,MAAM,MAAM,SAAS;AACtK,aAAO,MAAM,MAAM,WAAW,KAAK,MAAM,MAAM,OAAO,SAAS,eAAe,MAAM,MAAM,OAAO,GAAG,MAAM,sCAAsC,MAAM,gEAAgE,SAAS,cAAc;AAAA,IAC/O;AAAA,EACF;AACA,MAAI;AACJ,MAAI,0BAA0B,MAAM;AAClC,UAAM,aAAa,uBAAuB;AAC1C,WAAO,WAAW,UAAU,KAAK,WAAW,WAAW,GAAG,MAAM,2HAA2H,WAAW,SAAS;AAC/M,QAAI,WAAW,WAAW,GAAG;AAC3B,aAAO,WAAW,OAAO,KAAK,WAAW,OAAO,SAAS,aAAa,MAAM,oDAAoD,qEAAqE,SAAS,eAAe;AAAA,IAC/N,WAAW,WAAW,WAAW,GAAG;AAClC,UAAI;AACF,mCAA2B,YAAY,SAAS,QAAQ;AAAA,MAC1D,SAAS,GAAP;AACA,cAAM,SAAS,oDAAoD,sEAAsE,SAAS;AAClJ,cAAM,MAAM,MAAM;AAAA,MACpB;AAAA,IACF;AACA,8BAA0B,gBAAgB,wBAAwB,iBAAiB,cAAc;AAAA,EACnG;AACA,QAAM,QAAQ,CAAC,IAAI,UAAU;AAC3B,WAAO,eAAe,QAAQ,MAAM,wDAAwD,kDAAkD;AAC9I,UAAM,CAAC,UAAU,MAAM,GAAG,MAAM,IAAI;AACpC,UAAM,eAAe,qBAAqB,IAAI,GAAG,WAAW;AAC5D,WAAO,kBAAkB,SAAS,GAAG,MAAM,uHAAuH,YAAY;AAC9K,UAAM,OAAO,oBAAoB,KAAK,OAAO,cAAc,UAAU,SAAS,IAAI;AAClF,UAAM,YAAY,qBAAqB,MAAM,cAAc,SAAS,OAAO,SAAS,IAAI;AACxF,UAAM,MAAM,CAAC,MAAM,SAAS;AAC5B,QAAI,UAAU,MAAM;AAClB,YAAM,UAAU,qBAAqB,QAAQ,YAAY;AACzD,UAAI,KAAK,OAAO;AAAA,IAClB;AACA,WAAO;AAAA,EACT;AACA,QAAM,SAAS;AAAA,IACb,GAAG;AAAA,IACH,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,wBAAwB;AAAA,EAC1B;AACA,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA,KAAK;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA,YAAY;AAAA,IACZ;AAAA,EACF;AACA,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,WAAW,CAAC,MAAM,SAAS,SAAS;AACnD,UAAI,MAAM,OAAO,UAAU,aAAa,QAAQ,KAAK;AACrD,WAAK,CAAC,SAAS,MAAM,GAAG,CAAC;AACzB,UAAI,cAAc;AAChB,cAAM,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,MAC/D;AACA,aAAO,EAAE,OAAO,KAAK,UAAU,MAAM;AAAA,IACvC,CAAC;AACD,WAAO,SAAS,KAAK,OAAO;AAAA,EAC9B,OAAO;AACL,UAAM,mBAAmB,WAAW,CAAC,MAAM,SAAS,OAAO,SAAS;AAClE,UAAI,MAAM,OAAO,UAAU,aAAa,QAAQ,KAAK;AACrD,WAAK,CAAC,SAAS,MAAM,KAAK,KAAK,CAAC;AAChC,UAAI,cAAc;AAChB,cAAM,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,MAC/D;AACA,aAAO,EAAE,OAAO,KAAK,UAAU,MAAM;AAAA,IACvC,CAAC;AACD,WAAO,iBAAiB,KAAK,SAAS,KAAK;AAAA,EAC7C;AACF;AACA,IAAI,UAAU,GAAG,EAAE,aAAa,CAAC;AAGjC,SAAS,qCAAqC,GAAG,IAAI,aAAa,SAAS,MAAM,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB;AACpH,MAAI,MAAM;AACV,MAAI,EAAE,SAAS,GAAG;AAChB,UAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC;AAAA,EAC1D;AACA,MAAI,OAAO;AACX,MAAI,KAAK,SAAS,GAAG;AACnB,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC/D;AACA,QAAM,SAAS,EAAE,GAAG,KAAK,IAAI,KAAK;AAClC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,iBAAiB,WAAW,YAAY;AAC5E,SAAO,OAAO,UAAU,qCAAqC,QAAQ,KAAK;AAC5E;AACA,IAAI,sCAAsC,GAAG,EAAE,qCAAqC,CAAC;AAGrF,SAAS,oCAAoC,QAAQ,IAAI,QAAQ,SAAS,MAAM,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB;AACnH,MAAI,OAAO;AACX,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC/D;AACA,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO;AAClC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,iBAAiB,WAAW,YAAY,OAAO;AACnF,QAAM,MAAM,OAAO,UAAU,oCAAoC,QAAQ,KAAK;AAC9E,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,qCAAqC,GAAG,EAAE,oCAAoC,CAAC;AAGnF,SAAS,sBAAsB,EAAE,GAAG,QAAQ,SAAS,KAAK,MAAM,aAAa,QAAQ,YAAY,CAAC,GAAG,CAAC,GAAG,iBAAiB,MAAM,YAAY,cAAc,UAAU,wBAAwB,eAAe,GAAG;AAC5M,MAAI,WAAW,OAAO,MAAM,eAAe,WAAW,MAAM,OAAO;AACjE,QAAI,SAAS,gBAAgB,GAAG,QAAQ,SAAS,MAAM,YAAY,WAAW,eAAe;AAC7F,QAAI,QAAQ,MAAM;AAChB,eAAS,KAAK,QAAQ,IAAI;AAAA,IAC5B;AACA,WAAO,gBAAgB,QAAQ,aAAa,wBAAwB,cAAc;AAAA,EACpF;AACA,QAAM,KAAK,gBAAgB,GAAG,KAAK,mBAAmB,SAAS;AAC/D,QAAM,UAAU,gBAAgB,QAAQ,UAAU,mBAAmB,SAAS;AAC9E,MAAI,MAAM;AACV,MAAI,eAAe;AACnB,MAAI,GAAG,SAAS,GAAG;AACjB,mBAAe;AACf,UAAM,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC9D;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,sEAAsE,IAAI,OAAO;AAC9G,SAAO,QAAQ,SAAS,GAAG,MAAM,uEAAuE,QAAQ,OAAO;AACvH,SAAO,IAAI,MAAM,OAAO,QAAQ,MAAM,IAAI,MAAM,6DAA6D,IAAI,MAAM,qDAAqD,QAAQ,MAAM,KAAK;AAC/L,MAAI,aAAa,MAAM;AACrB,gBAAY,CAAC,GAAG,CAAC;AAAA,EACnB;AACA,SAAO,+BAA+B,SAAS,SAAS,GAAG,MAAM,sFAAsF,0BAA0B,YAAY;AAC7L,4BAA0B,yBAAyB,MAAM,eAAe;AACxE,QAAM,WAAW,kBAAkB,IAAI,OAAO,QAAQ,OAAO,SAAS,WAAW,MAAM,iBAAiB,IAAI;AAC5G,MAAI;AACJ,MAAI,QAAQ,MAAM;AAChB,YAAQ,gBAAgB,MAAM,QAAQ,cAAc;AACpD,KAAC,KAAK,IAAI,eAAe,OAAO,EAAE;AAClC,+BAA2B,SAAS,UAAU,MAAM,KAAK;AAAA,EAC3D;AACA,MAAI;AACJ,MAAI,0BAA0B,MAAM;AAClC,8BAA0B,gBAAgB,wBAAwB,iBAAiB,uBAAuB;AAAA,EAC5G;AACA,QAAM,QAAQ,CAAC,IAAI,UAAU;AAC3B,WAAO,kBAAkB,SAAS,GAAG,MAAM,mHAAmH,YAAY;AAC1K,UAAM,CAAC,UAAU,MAAM,GAAG,KAAK,IAAI;AACnC,UAAM,eAAe,qBAAqB,IAAI,GAAG,WAAW;AAC5D,UAAM,OAAO,mCAAmC,KAAK,OAAO,cAAc,UAAU,SAAS,MAAM,WAAW,eAAe;AAC7H,UAAM,YAAY,oCAAoC,MAAM,cAAc,SAAS,OAAO,SAAS,MAAM,WAAW,eAAe;AACnI,QAAI,SAAS,MAAM;AACjB,YAAM,UAAU,qBAAqB,OAAO,YAAY;AACxD,aAAO,CAAC,MAAM,WAAW,OAAO;AAAA,IAClC;AACA,WAAO,CAAC,MAAM,SAAS;AAAA,EACzB;AACA,QAAM,SAAS;AAAA,IACb,GAAG;AAAA,IACH,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,wBAAwB;AAAA,EAC1B;AACA,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA,KAAK;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA,YAAY;AAAA,IACZ;AAAA,EACF;AACA,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,WAAW,CAAC,MAAM,SAAS,SAAS;AACnD,UAAI,MAAM,OAAO,UAAU,sBAAsB,QAAQ,KAAK;AAC9D,WAAK,CAAC,SAAS,MAAM,GAAG,CAAC;AACzB,UAAI,cAAc;AAChB,cAAM,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,MAC/D;AACA,aAAO,EAAE,OAAO,KAAK,UAAU,MAAM;AAAA,IACvC,CAAC;AACD,WAAO,SAAS,KAAK,OAAO;AAAA,EAC9B,OAAO;AACL,UAAM,mBAAmB,WAAW,CAAC,MAAM,SAAS,OAAO,SAAS;AAClE,UAAI,MAAM,OAAO,UAAU,sBAAsB,QAAQ,KAAK;AAC9D,WAAK,CAAC,SAAS,MAAM,KAAK,KAAK,CAAC;AAChC,UAAI,cAAc;AAChB,cAAM,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,MAC/D;AACA,aAAO,EAAE,OAAO,KAAK,UAAU,MAAM;AAAA,IACvC,CAAC;AACD,WAAO,iBAAiB,KAAK,SAAS,KAAK;AAAA,EAC7C;AACF;AACA,IAAI,mBAAmB,GAAG,EAAE,sBAAsB,CAAC;AAGnD,SAAS,aAAa,EAAE,GAAG,GAAG,aAAa,OAAO,aAAa,OAAO,MAAM,YAAY,cAAc,UAAU,wBAAwB,iBAAiB,IAAI,GAAG;AAC9J,MAAI,WAAW,OAAO,MAAM,eAAe,WAAW,MAAM,OAAO;AACjE,QAAI,SAAS,OAAO,GAAG,GAAG,YAAY,UAAU;AAChD,QAAI,QAAQ,MAAM;AAChB,eAAS,KAAK,QAAQ,IAAI;AAAA,IAC5B;AACA,WAAO,gBAAgB,QAAQ,aAAa,wBAAwB,cAAc;AAAA,EACpF;AACA,MAAI,KAAK,gBAAgB,GAAG,KAAK,cAAc;AAC/C,MAAI,KAAK,gBAAgB,GAAG,KAAK,cAAc;AAC/C,GAAC,IAAI,EAAE,IAAI,eAAe,IAAI,EAAE;AAChC,QAAM,cAAc,aAAa,GAAG,MAAM,GAAG,OAAO,KAAK,GAAG,MAAM,GAAG,OAAO;AAC5E,QAAM,cAAc,aAAa,GAAG,MAAM,GAAG,OAAO,KAAK,GAAG,MAAM,GAAG,OAAO;AAC5E,QAAM,cAAc,aAAa,GAAG,MAAM,GAAG,OAAO,KAAK,GAAG,MAAM,GAAG,OAAO;AAC5E,QAAM,cAAc,aAAa,GAAG,MAAM,GAAG,OAAO,KAAK,GAAG,MAAM,GAAG,OAAO;AAC5E,QAAM,aAAa,GAAG,MAAM,MAAM,GAAG,EAAE;AACvC,QAAM,aAAa,GAAG,MAAM,MAAM,GAAG,EAAE;AACvC,QAAM,YAAY,cAAc,UAAU;AAC1C,QAAM,YAAY,cAAc,UAAU;AAC1C,SAAO,gBAAgB,aAAa,MAAM,wCAAwC,qBAAqB,uCAAuC,GAAG,aAAa,GAAG,wBAAwB,6BAA6B,wBAAwB;AAC9O,QAAM,oBAAoB,2BAA2B,GAAG,MAAM,MAAM,GAAG,EAAE,GAAG,GAAG,MAAM,MAAM,GAAG,EAAE,CAAC;AACjG,QAAM,WAAW,kBAAkB,OAAO,CAAC,aAAa,WAAW,CAAC;AACpE,QAAM,MAAM,aAAa,QAAQ,IAAI,CAAC,WAAW,aAAa,WAAW,CAAC,IAAI,QAAQ,IAAI,CAAC,WAAW,aAAa,WAAW,CAAC;AAC/H,QAAM,MAAM,aAAa,QAAQ,IAAI,CAAC,WAAW,aAAa,WAAW,CAAC,IAAI,QAAQ,IAAI,CAAC,WAAW,aAAa,WAAW,CAAC;AAC/H,MAAI;AACJ,MAAI,QAAQ,MAAM;AAChB,YAAQ,gBAAgB,MAAM,QAAQ,cAAc;AACpD,KAAC,KAAK,IAAI,eAAe,OAAO,EAAE;AAClC,+BAA2B,UAAU,MAAM,KAAK;AAAA,EAClD;AACA,MAAI;AACJ,MAAI,0BAA0B,MAAM;AAClC,8BAA0B,gBAAgB,wBAAwB,iBAAiB,cAAc;AAAA,EACnG;AACA,QAAM,QAAQ,CAAC,IAAI,UAAU;AAC3B,UAAM,CAAC,MAAM,MAAM,GAAG,MAAM,IAAI;AAChC,UAAM,eAAe,qBAAqB,QAAQ,IAAI,EAAE,KAAK,GAAG,GAAG,WAAW;AAC9E,QAAI;AACJ,QAAI;AACJ,QAAI,CAAC,cAAc,CAAC,YAAY;AAC9B,aAAO,OAAO,cAAc,MAAM,OAAO,IAAI;AAC7C,aAAO,OAAO,MAAM,cAAc,MAAM,KAAK;AAAA,IAC/C,WAAW,CAAC,cAAc,YAAY;AACpC,aAAO,OAAO,cAAc,MAAM,OAAO,KAAK;AAC9C,aAAO,OAAO,cAAc,MAAM,MAAM,KAAK;AAAA,IAC/C,WAAW,cAAc,CAAC,YAAY;AACpC,aAAO,OAAO,MAAM,cAAc,OAAO,IAAI;AAC7C,aAAO,OAAO,MAAM,cAAc,OAAO,KAAK;AAAA,IAChD,OAAO;AACL,aAAO,OAAO,MAAM,cAAc,MAAM,IAAI;AAC5C,aAAO,OAAO,cAAc,MAAM,MAAM,IAAI;AAAA,IAC9C;AACA,QAAI,QAAQ,MAAM;AAChB,YAAM,UAAU,qBAAqB,QAAQ,YAAY;AACzD,aAAO,CAAC,MAAM,MAAM,OAAO;AAAA,IAC7B,OAAO;AACL,aAAO,CAAC,MAAM,IAAI;AAAA,IACpB;AAAA,EACF;AACA,QAAM,SAAS;AAAA,IACb,GAAG;AAAA,IACH,GAAG;AAAA,IACH,MAAM;AAAA,IACN,wBAAwB;AAAA,EAC1B;AACA,QAAM,QAAQ,EAAE,YAAY,YAAY,YAAY,aAAa,eAAe;AAChF,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,WAAW,CAAC,MAAM,MAAM,SAAS;AAChD,YAAM,MAAM,OAAO,UAAU,cAAc,QAAQ,KAAK;AACxD,WAAK,CAAC,MAAM,MAAM,GAAG,CAAC;AACtB,aAAO,EAAE,OAAO,QAAQ,KAAK,QAAQ,GAAG,UAAU,MAAM;AAAA,IAC1D,CAAC;AACD,WAAO,SAAS,KAAK,GAAG;AAAA,EAC1B,OAAO;AACL,UAAM,mBAAmB,WAAW,CAAC,MAAM,MAAM,QAAQ,SAAS;AAChE,YAAM,MAAM,OAAO,UAAU,cAAc,QAAQ,KAAK;AACxD,WAAK,CAAC,MAAM,MAAM,KAAK,MAAM,CAAC;AAC9B,aAAO,EAAE,OAAO,QAAQ,KAAK,QAAQ,GAAG,UAAU,MAAM;AAAA,IAC1D,CAAC;AACD,WAAO,iBAAiB,KAAK,KAAK,KAAK;AAAA,EACzC;AACF;AACA,IAAI,UAAU,GAAG,EAAE,aAAa,CAAC;AAGjC,SAAS,eAAe,cAAc;AACpC,SAAO,aAAa,cAAc,MAAM,IAAI;AAC9C;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,YAAY,cAAc;AACjC,SAAO,aAAa,cAAc,KAAK,GAAG;AAC5C;AACA,IAAI,aAAa,GAAG,EAAE,YAAY,CAAC;AAGnC,SAAS,OAAO,SAAS,aAAa,WAAW,SAAS,OAAO,WAAW,GAAG;AAC7E,MAAI,QAAQ;AACZ,QAAM,SAAS,CAAC;AAChB,SAAO,QAAQ,eAAe,QAAQ,MAAM;AAC1C,WAAO,KAAK,MAAM,SAAS,OAAO,WAAW,CAAC;AAC9C,aAAS;AAAA,EACX;AACA,MAAI,QAAQ;AACV,WAAO,QAAQ,QAAQ,MAAM;AAC3B,YAAM,SAAS,QAAQ,cAAc,QAAQ;AAC7C,YAAM,OAAO,OAAO;AAAA,QAClB,MAAM,SAAS,OAAO,cAAc,MAAM;AAAA,QAC1C,KAAK,CAAC,MAAM,GAAG,QAAQ;AAAA,MACzB,CAAC;AACD,aAAO,KAAK,IAAI;AAChB,eAAS;AAAA,IACX;AAAA,EACF;AACA,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,SAAS,CAAC,GAAG,CAAC,GAAG,WAAW,CAAC;AAAA,EACtC;AACA,SAAO,QAAQ,OAAO,MAAM,GAAG,CAAC,OAAO,QAAQ,WAAW,CAAC;AAC7D;AACA,IAAI,QAAQ,GAAG,EAAE,OAAO,CAAC;AAGzB,SAAS,MAAM,SAAS,aAAa,WAAW,WAAW,WAAW,YAAY;AAChF,MAAI,aAAa,MAAM;AACrB,gBAAY,oBAAoB,WAAW;AAAA,EAC7C;AACA,QAAM,eAAe,MAAM,SAAS,aAAa,SAAS;AAC1D,QAAM,iBAAiB,IAAI,cAAc,SAAS,WAAW,CAAC;AAC9D,SAAO,KAAK,gBAAgB,SAAS;AACvC;AACA,IAAI,OAAO,GAAG,EAAE,MAAM,CAAC;AAGvB,SAAS,eAAe,QAAQ,OAAO,QAAQ,UAAU,SAAS,YAAY,qBAAqB,GAAG;AACpG,QAAM,SAAS,gBAAgB,QAAQ,SAAS,eAAe;AAC/D,QAAM,SAAS,gBAAgB,OAAO,SAAS,iBAAiB,SAAS;AACzE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,iBAAiB,OAAO;AAC1E,QAAM,WAAW,OAAO,MAAM;AAC9B,SAAO,OAAO,SAAS,GAAG,MAAM,6DAA6D,OAAO,OAAO;AAC3G,SAAO,OAAO,SAAS,KAAK,OAAO,MAAM,OAAO,GAAG,MAAM,oDAAoD,6BAA6B,OAAO,QAAQ;AACzJ,SAAO,QAAQ,SAAS,KAAK,QAAQ,MAAM,OAAO,UAAU,MAAM,qDAAqD,2BAA2B,OAAO,QAAQ;AACjK,SAAO,SAAS,WAAW,GAAG,MAAM,wEAAwE,SAAS,SAAS;AAC9H,SAAO,SAAS,MAAM,KAAK,SAAS,MAAM,GAAG,MAAM,2CAA2C,UAAU;AACxG,SAAO,WAAW,cAAc,WAAW,WAAW,MAAM,+CAA+C,QAAQ;AACnH,QAAM,SAAS,EAAE,OAAO,QAAQ,OAAO,QAAQ,QAAQ,QAAQ;AAC/D,QAAM,QAAQ,EAAE,QAAQ,oBAAoB,SAAS;AACrD,QAAM,MAAM,OAAO,UAAU,eAAe,QAAQ,KAAK;AACzD,SAAO;AACT;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,eAAe,QAAQ;AAC9B,QAAM,SAAS,gBAAgB,QAAQ,SAAS,iBAAiB,SAAS;AAC1E,SAAO,OAAO,SAAS,GAAG,MAAM,6DAA6D,OAAO,OAAO;AAC3G,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,QAAM,MAAM,OAAO,UAAU,eAAe,QAAQ,CAAC,CAAC;AACtD,SAAO;AACT;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,gBAAgB,QAAQ;AAC/B,QAAM,SAAS,gBAAgB,QAAQ,SAAS,gBAAgB;AAChE,QAAM,cAAc,OAAO,OAAO;AAClC,QAAM,WAAW,OAAO,MAAM;AAC9B,SAAO,OAAO,QAAQ,GAAG,MAAM,yEAAyE,OAAO,OAAO;AACtH,SAAO,aAAa,GAAG,MAAM,+FAA+F,WAAW;AACvI,QAAM,OAAO,IAAI,MAAM,OAAO,IAAI;AAClC,OAAK,KAAK,GAAG,GAAG,WAAW;AAC3B,OAAK,eAAe;AACpB,SAAO,KAAK,QAAQ,IAAI;AAC1B;AACA,IAAI,iBAAiB,GAAG,EAAE,gBAAgB,CAAC;AAG3C,SAAS,kBAAkB,QAAQ,SAAS,YAAY,GAAG,SAAS,KAAK;AACvE,QAAM,SAAS,gBAAgB,QAAQ,SAAS,oBAAoB,SAAS;AAC7E,SAAO,OAAO,SAAS,GAAG,MAAM,gEAAgE,OAAO,OAAO;AAC9G,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,QAAM,QAAQ,EAAE,SAAS,WAAW,OAAO;AAC3C,QAAM,MAAM,OAAO,UAAU,kBAAkB,QAAQ,KAAK;AAC5D,SAAO;AACT;AACA,IAAI,mBAAmB,GAAG,EAAE,kBAAkB,CAAC;AAG/C,SAAS,sBAAsB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,cAAc;AACvG,MAAI,gBAAgB,MAAM;AACxB,mBAAe;AAAA,EACjB;AACA,MAAI,kBAAkB,MAAM;AAC1B,qBAAiB,OAAO;AAAA,EAC1B;AACA,MAAI,gBAAgB,MAAM;AACxB,mBAAe;AAAA,EACjB;AACA,QAAM,WAAW,MAAM,MAAM;AAC7B,kBAAgB,KAAK,IAAI,eAAe,QAAQ;AAChD,SAAO,KAAK,gBAAgB,gBAAgB,GAAG,MAAM,4CAA4C,eAAe;AAChH,SAAO,MAAM,SAAS,GAAG,MAAM,+CAA+C,MAAM,OAAO;AAC3F,SAAO,MAAM,MAAM,OAAO,GAAG,MAAM,oDAAoD,MAAM,MAAM,IAAI;AACvG,SAAO,OAAO,SAAS,GAAG,MAAM,4BAA4B;AAC5D,SAAO,OAAO,MAAM,OAAO,UAAU,MAAM,sDAAsD,qBAAqB,OAAO,MAAM,IAAI;AACvI,SAAO,KAAK,gBAAgB,gBAAgB,GAAG,MAAM,4CAA4C,eAAe;AAChH,SAAO,EAAE,eAAe,cAAc,gBAAgB,aAAa;AACrE;AAGA,SAAS,mBAAmB,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB;AACvH,QAAM,SAAS,gBAAgB,OAAO,SAAS,qBAAqB,SAAS;AAC7E,QAAM,UAAU,gBAAgB,QAAQ,UAAU,qBAAqB,SAAS;AAChF,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,cAAc;AACjG,kBAAgB,OAAO;AACvB,iBAAe,OAAO;AACtB,mBAAiB,OAAO;AACxB,QAAM,QAAQ,EAAE,eAAe,cAAc,eAAe;AAC5D,SAAO,OAAO,UAAU,qBAAqB,EAAE,OAAO,QAAQ,QAAQ,QAAQ,GAAG,KAAK;AACxF;AACA,IAAI,oBAAoB,GAAG,EAAE,mBAAmB,CAAC;AAGjD,SAAS,aAAa,KAAK,SAAS,YAAY;AAC9C,QAAM,QAAQ,aAAa,KAAK,SAAS,UAAU;AACnD,QAAM,iBAAiB,QAAQ,IAAI,EAAE,QAAQ,KAAK;AAClD,MAAI,OAAO,gBAAgB,GAAG,OAAO;AACvC;AACA,SAAS,aAAa,KAAK,QAAQ,YAAY;AAC7C,SAAO,cAAc,KAAK,QAAQ,cAAc,iBAAiB;AACnE;AACA,SAAS,kBAAkB,GAAG,GAAG;AAC/B,SAAO,IAAI,IAAI,IAAI,IAAI,IAAI,KAAK;AAClC;AACA,SAAS,cAAc,KAAK,QAAQ,YAAY;AAC9C,MAAI,OAAO;AACX,MAAI,QAAQ,IAAI;AAChB,MAAI,SAAS;AACb,MAAI,QAAQ;AACZ,SAAO,OAAO,OAAO;AACnB,aAAS,QAAQ,QAAQ,SAAS;AAClC,UAAM,gBAAgB,WAAW,QAAQ,IAAI,OAAO;AACpD,QAAI,gBAAgB,GAAG;AACrB,aAAO,SAAS;AAAA,IAClB,OAAO;AACL,cAAQ;AACR,cAAQ,CAAC;AAAA,IACX;AAAA,EACF;AACA,SAAO,QAAQ,OAAO,CAAC,OAAO;AAChC;AAGA,SAAS,wBAAwB,OAAO,QAAQ,eAAe,cAAc,gBAAgB;AAC3F,SAAO,uBAAuB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,CAAC;AAC7F;AACA,SAAS,wBAAwB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,oBAAoB;AAC/G,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,SAAS,wBAAwB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,cAAc;AACzG,SAAO,uBAAuB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,cAAc,IAAI;AAC9G;AACA,SAAS,uBAAuB,OAAO,QAAQ,eAAe,cAAc,gBAAgB,cAAc,qBAAqB,OAAO,qBAAqB,OAAO,qBAAqB,OAAO;AAC5L,QAAM,aAAa,CAAC;AACpB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,QAAI,OAAO,KAAK,gBAAgB;AAC9B,iBAAW,KAAK,EAAE,OAAO,OAAO,IAAI,UAAU,GAAG,oBAAoB,EAAE,CAAC;AAAA,IAC1E;AAAA,EACF;AACA,aAAW,KAAK,mBAAmB;AACnC,QAAMI,UAAS,eAAe,IAAI,OAAO,eAAe;AACxD,QAAM,kBAAkB,CAAC;AACzB,QAAM,iBAAiB,CAAC;AACxB,SAAO,gBAAgB,SAAS,iBAAiB,WAAW,SAAS,GAAG;AACtE,UAAM,YAAY,WAAW,IAAI;AACjC,UAAM,EAAE,OAAO,eAAe,UAAU,mBAAmB,IAAI;AAC/D,QAAI,gBAAgB,gBAAgB;AAClC;AAAA,IACF;AACA,QAAI,kBAAkB;AACtB,aAAS,IAAI,gBAAgB,SAAS,GAAG,KAAK,oBAAoB,EAAE,GAAG;AACrE,YAAMC,OAAM,sBAAsB,OAAO,UAAU,gBAAgB,EAAE;AACrE,UAAIA,QAAO,cAAc;AACvB,0BAAkB;AAClB;AAAA,MACF;AACA,gBAAU,QAAQ,UAAU,QAAQ,eAAe,cAAcD,SAAQC,IAAG;AAC5E,UAAI,UAAU,SAAS,gBAAgB;AACrC;AAAA,MACF;AAAA,IACF;AACA,cAAU,qBAAqB,gBAAgB;AAC/C,QAAI,CAAC,iBAAiB;AACpB,UAAI,UAAU,UAAU,eAAe;AACrC,wBAAgB,KAAK,QAAQ;AAC7B,uBAAe,KAAK,UAAU,KAAK;AAAA,MACrC,WAAW,UAAU,QAAQ,gBAAgB;AAC3C,qBAAa,YAAY,WAAW,mBAAmB;AAAA,MACzD;AAAA,IACF;AAAA,EACF;AACA,QAAM,eAAe,gBAAgB;AACrC,QAAM,aAAa,gBAAgB;AACnC,MAAI,sBAAsB,aAAa,GAAG;AACxC,oBAAgB,KAAK,GAAG,IAAI,MAAM,UAAU,EAAE,KAAK,CAAC,CAAC;AACrD,mBAAe,KAAK,GAAG,IAAI,MAAM,UAAU,EAAE,KAAK,CAAC,CAAC;AAAA,EACtD;AACA,QAAM,SAAS,EAAE,gBAAgB;AACjC,MAAI,oBAAoB;AACtB,WAAO,oBAAoB;AAAA,EAC7B;AACA,MAAI,oBAAoB;AACtB,WAAO,kBAAkB;AAAA,EAC3B;AACA,SAAO;AACT;AACA,SAAS,sBAAsB,OAAO,GAAG,GAAG;AAC1C,QAAM,SAAS,MAAM,SAAS,IAAI,GAAG,IAAI,IAAI,CAAC;AAC9C,QAAM,SAAS,MAAM,SAAS,IAAI,GAAG,IAAI,IAAI,CAAC;AAC9C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,QAAQ,KAAK,IAAI,OAAO,IAAI,OAAO,EAAE;AAC3C,QAAM,SAAS,QAAQ,UAAU,QAAQ;AACzC,QAAM,SAAS,QAAQ,UAAU,QAAQ;AACzC,MAAI,SAAS,KAAK,SAAS,GAAG;AAC5B,WAAO;AAAA,EACT;AACA,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,mBAAmB,kBAAkB,CAAC,IAAI,KAAK,IAAI,mBAAmB,kBAAkB,CAAC;AAC3H,SAAO,oBAAoB,QAAQ,QAAQ;AAC7C;AACA,SAAS,eAAe,cAAcD,SAAQC,MAAK;AACjD,QAAM,SAAS,KAAK,IAAID,UAASC,OAAMA,IAAG;AAC1C,SAAOA,QAAO,eAAe,SAAS;AACxC;AACA,SAAS,oBAAoB,IAAI,IAAI;AACnC,SAAO,GAAG,QAAQ,GAAG,SAAS,GAAG,UAAU,GAAG,SAAS,GAAG,WAAW,GAAG;AAC1E;AAGA,eAAe,wBAAwB,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB;AAClI,QAAM,SAAS,gBAAgB,OAAO,SAAS,wBAAwB;AACvE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,wBAAwB;AAC1E,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,cAAc;AACjG,kBAAgB,OAAO;AACvB,iBAAe,OAAO;AACtB,mBAAiB,OAAO;AACxB,QAAM,iBAAiB,MAAM,QAAQ,IAAI,CAAC,OAAO,KAAK,GAAG,QAAQ,KAAK,CAAC,CAAC;AACxE,QAAM,YAAY,eAAe;AACjC,QAAM,aAAa,eAAe;AAClC,QAAM,EAAE,gBAAgB,IAAI,wBAAwB,WAAW,YAAY,eAAe,cAAc,cAAc;AACtH,MAAI,WAAW,OAAO;AACpB,WAAO,QAAQ;AAAA,EACjB;AACA,MAAI,YAAY,QAAQ;AACtB,YAAQ,QAAQ;AAAA,EAClB;AACA,SAAO,SAAS,iBAAiB,OAAO;AAC1C;AACA,IAAI,yBAAyB;AAG7B,SAAS,4BAA4B,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB,eAAe,GAAG;AAClJ,QAAM,SAAS,gBAAgB,OAAO,SAAS,mBAAmB;AAClE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,mBAAmB;AACrE,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,gBAAgB,YAAY;AAC/G,kBAAgB,OAAO;AACvB,iBAAe,OAAO;AACtB,mBAAiB,OAAO;AACxB,iBAAe,OAAO;AACtB,QAAM,SAAS,EAAE,OAAO,QAAQ,QAAQ,QAAQ;AAChD,QAAM,QAAQ,EAAE,eAAe,cAAc,gBAAgB,aAAa;AAC1E,QAAM,SAAS,OAAO,UAAU,qBAAqB,QAAQ,KAAK;AAClE,SAAO,EAAE,iBAAiB,OAAO,IAAI,gBAAgB,OAAO,GAAG;AACjE;AACA,IAAI,6BAA6B,GAAG,EAAE,4BAA4B,CAAC;AAGnE,eAAe,iCAAiC,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB,eAAe,GAAG;AAC7J,QAAM,SAAS,gBAAgB,OAAO,SAAS,wBAAwB;AACvE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,wBAAwB;AAC1E,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,gBAAgB,YAAY;AAC/G,kBAAgB,OAAO;AACvB,iBAAe,OAAO;AACtB,mBAAiB,OAAO;AACxB,iBAAe,OAAO;AACtB,QAAM,iBAAiB,MAAM,QAAQ,IAAI,CAAC,OAAO,KAAK,GAAG,QAAQ,KAAK,CAAC,CAAC;AACxE,QAAM,YAAY,eAAe;AACjC,QAAM,aAAa,eAAe;AAClC,QAAM,EAAE,iBAAiB,eAAe,IAAI,wBAAwB,WAAW,YAAY,eAAe,cAAc,gBAAgB,YAAY;AACpJ,MAAI,WAAW,OAAO;AACpB,WAAO,QAAQ;AAAA,EACjB;AACA,MAAI,YAAY,QAAQ;AACtB,YAAQ,QAAQ;AAAA,EAClB;AACA,SAAO;AAAA,IACL,iBAAiB,SAAS,iBAAiB,OAAO;AAAA,IAClD,gBAAgB,SAAS,cAAc;AAAA,EACzC;AACF;AACA,IAAI,kCAAkC;AAGtC,SAAS,yBAAyB,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB,qBAAqB,OAAO;AACzJ,QAAM,SAAS,gBAAgB,OAAO,SAAS,mBAAmB;AAClE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,mBAAmB;AACrE,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,gBAAgB,IAAI;AACvG,QAAM,iBAAiB,OAAO;AAC9B,QAAM,gBAAgB,OAAO;AAC7B,QAAM,kBAAkB,OAAO;AAC/B,QAAM,SAAS,EAAE,OAAO,QAAQ,QAAQ,QAAQ;AAChD,QAAM,QAAQ;AAAA,IACZ,eAAe;AAAA,IACf,cAAc;AAAA,IACd,gBAAgB;AAAA,IAChB;AAAA,EACF;AACA,QAAM,SAAS,OAAO,UAAU,qBAAqB,QAAQ,KAAK;AAClE,SAAO,EAAE,iBAAiB,OAAO,IAAI,cAAc,OAAO,GAAG;AAC/D;AACA,IAAI,0BAA0B,GAAG,EAAE,yBAAyB,CAAC;AAG7D,eAAe,8BAA8B,OAAO,QAAQ,eAAe,eAAe,KAAK,iBAAiB,OAAO,mBAAmB,qBAAqB,OAAO;AACpK,QAAM,SAAS,gBAAgB,OAAO,SAAS,wBAAwB;AACvE,QAAM,UAAU,gBAAgB,QAAQ,UAAU,wBAAwB;AAC1E,QAAM,SAAS,sBAAsB,QAAQ,SAAS,eAAe,cAAc,gBAAgB,IAAI;AACvG,QAAM,iBAAiB,OAAO;AAC9B,QAAM,gBAAgB,OAAO;AAC7B,QAAM,kBAAkB,OAAO;AAC/B,QAAM,CAAC,WAAW,UAAU,IAAI,MAAM,QAAQ,IAAI,CAAC,OAAO,KAAK,GAAG,QAAQ,KAAK,CAAC,CAAC;AACjF,QAAM,EAAE,iBAAiB,aAAa,IAAI,wBAAwB,WAAW,YAAY,gBAAgB,eAAe,iBAAiB,kBAAkB;AAC3J,MAAI,WAAW,OAAO;AACpB,WAAO,QAAQ;AAAA,EACjB;AACA,MAAI,YAAY,QAAQ;AACtB,YAAQ,QAAQ;AAAA,EAClB;AACA,SAAO;AAAA,IACL,iBAAiB,SAAS,iBAAiB,OAAO;AAAA,IAClD,cAAc,OAAO,cAAc,OAAO;AAAA,EAC5C;AACF;AACA,IAAI,+BAA+B;AAGnC,SAAS,gBAAgB,QAAQ,MAAM,eAAe,OAAO,mBAAmB,OAAO;AACrF,QAAM,UAAU,gBAAgB,QAAQ,UAAU,gBAAgB;AAClE,SAAO,QAAQ,SAAS,KAAK,QAAQ,SAAS,GAAG,MAAM,gEAAgE,QAAQ,OAAO;AACtI,SAAO,KAAK,WAAW,GAAG,MAAM,6DAA6D,OAAO;AACpG,SAAO,qBAAqB,SAAS,iBAAiB,OAAO,MAAM,mFAAmF;AACtJ,MAAI,cAAc;AAClB,MAAI,eAAe;AACnB,MAAI,QAAQ,SAAS,GAAG;AACtB,mBAAe;AACf,kBAAc,QAAQ,SAAS,CAAC,GAAG,QAAQ,MAAM,IAAI,QAAQ,MAAM,IAAI,QAAQ,MAAM,EAAE,CAAC;AAAA,EAC1F;AACA,QAAM,CAAC,IAAI;AACX,QAAM,SAAS,EAAE,QAAQ,YAAY;AACrC,QAAM,QAAQ,EAAE,cAAc,kBAAkB,KAAK;AACrD,QAAM,MAAM,OAAO,UAAU,gBAAgB,QAAQ,KAAK;AAC1D,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,iBAAiB,GAAG,EAAE,gBAAgB,CAAC;AAG3C,SAAS,uBAAuB,QAAQ,MAAM,eAAe,OAAO,mBAAmB,OAAO;AAC5F,QAAM,UAAU,gBAAgB,QAAQ,UAAU,uBAAuB;AACzE,SAAO,QAAQ,SAAS,KAAK,QAAQ,SAAS,GAAG,MAAM,uEAAuE,QAAQ,OAAO;AAC7I,SAAO,KAAK,WAAW,GAAG,MAAM,oEAAoE,OAAO;AAC3G,SAAO,QAAQ,UAAU,aAAa,QAAQ,UAAU,SAAS,MAAM,kDAAkD;AACzH,SAAO,qBAAqB,SAAS,iBAAiB,OAAO,MAAM,0FAA0F;AAC7J,MAAI,cAAc;AAClB,MAAI,eAAe;AACnB,MAAI,QAAQ,SAAS,GAAG;AACtB,mBAAe;AACf,kBAAc,QAAQ,SAAS,CAAC,GAAG,QAAQ,MAAM,IAAI,QAAQ,MAAM,IAAI,QAAQ,MAAM,EAAE,CAAC;AAAA,EAC1F;AACA,QAAM,CAAC,IAAI;AACX,QAAM,SAAS,EAAE,QAAQ,YAAY;AACrC,QAAM,QAAQ,EAAE,cAAc,kBAAkB,KAAK;AACrD,QAAM,MAAM,OAAO,UAAU,uBAAuB,QAAQ,KAAK;AACjE,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,wBAAwB,GAAG,EAAE,uBAAuB,CAAC;AAGzD,SAAS,WAAW,QAAQ,SAAS,UAAU,WAAW,OAAO,cAAc,KAAK;AAClF,QAAM,SAAS,gBAAgB,QAAQ,SAAS,WAAW;AAC3D,QAAM,qBAAqB;AAC3B,QAAM,uBAAuB;AAC7B,QAAM,sBAAsB;AAC5B,QAAM,qBAAqB,OAAO,MAAM,KAAK,OAAO,MAAM;AAC1D,MAAI,aAAa,IAAI,SAAS,CAAC,WAAW,CAAC,GAAG,GAAG;AACjD,MAAI,GAAG,GAAG,GAAG;AACb,SAAO,OAAO,SAAS,GAAG,MAAM,yDAAyD,OAAO,OAAO;AACvG,SAAO,OAAO,MAAM,OAAO,KAAK,OAAO,MAAM,OAAO,GAAG,MAAM,0EAA0E,OAAO,MAAM,KAAK;AACzJ,SAAO,OAAO,UAAU,WAAW,OAAO,UAAU,WAAW,MAAM,sEAAsE,OAAO,QAAQ;AAC1J,SAAO,WAAW,UAAU,WAAW,UAAU,MAAM,0CAA0C,QAAQ;AACzG,MAAI,OAAO,MAAM,OAAO,GAAG;AACzB,KAAC,GAAG,GAAG,CAAC,IAAI,MAAM,QAAQ,CAAC,GAAG,GAAG,CAAC,GAAG,EAAE;AACvC,UAAM,KAAK,IAAI,GAAG,kBAAkB;AACpC,UAAM,KAAK,IAAI,GAAG,oBAAoB;AACtC,UAAM,KAAK,IAAI,GAAG,mBAAmB;AACrC,gBAAY,KAAK,KAAK,IAAI,EAAE,GAAG,EAAE;AAAA,EACnC,OAAO;AACL,gBAAY;AAAA,EACd;AACA,MAAI,WAAW,QAAQ;AACrB,UAAM,aAAa,SAAS,KAAK,OAAO,SAAS,GAAG,OAAO,GAAG,OAAO,CAAC,CAAC,GAAG,GAAG;AAC7E,iBAAa,KAAK,YAAY,kBAAkB;AAAA,EAClD;AACA,QAAM,eAAe,WAAW,UAAU,WAAW,UAAU,IAAI,QAAQ,WAAW,UAAU;AAChG,QAAM,SAAS,KAAK,IAAI,cAAc,GAAG,GAAG,OAAO;AACnD,SAAO;AACT;AACA,SAAS,KAAK,WAAW,OAAO;AAC9B,MAAI,aAAa,SAAS,CAAC,EAAE,CAAC;AAC9B,MAAI,eAAe,SAAS,CAAC,CAAC,CAAC;AAC/B,MAAI,YAAY,SAAS,CAAC,CAAC,CAAC;AAC5B,MAAI,YAAY,aAAa,WAAW,SAAS,kBAAkB;AACnE,WAAS,QAAQ,GAAG,QAAQ,UAAU,OAAO,GAAG,SAAS;AACvD,iBAAa,MAAM,WAAW,GAAG,QAAQ,CAAC;AAC1C,kBAAc,MAAM,WAAW,QAAQ,CAAC;AACxC,uBAAmB,IAAI,KAAK,UAAU,GAAG,KAAK;AAC9C,iBAAa,IAAI,KAAK,WAAW,GAAG,KAAK;AACzC,UAAM,gBAAgB,KAAK,IAAI,YAAY,MAAM,GAAG,WAAW,IAAI,CAAC,CAAC;AACrE,gBAAY,IAAI,eAAe,KAAK,UAAU,CAAC;AAC/C,UAAM,cAAc,KAAK,YAAY,OAAO,WAAW,IAAI;AAC3D,UAAM,aAAa,KAAK,MAAM,GAAG,YAAY,IAAI,GAAG,WAAW;AAC/D,UAAM,aAAa,IAAI,aAAa,UAAU;AAC9C,cAAU,IAAI,KAAK,UAAU,GAAG,KAAK,WAAW,CAAC;AACjD,UAAM,gBAAgB,IAAI,WAAW,OAAO;AAC5C,UAAM,gBAAgB,IAAI,WAAW,OAAO;AAC5C,UAAM,eAAe,IAAI,kBAAkB,UAAU;AACrD,gBAAY,IAAI,IAAI,cAAc,aAAa,GAAG,aAAa;AAC/D,UAAM,YAAY,QAAQ,WAAW,YAAY;AACjD,mBAAe,MAAM,WAAW,WAAW,YAAY;AACvD,iBAAa,MAAM,WAAW,SAAS,CAAC,KAAK,CAAC,GAAG,UAAU;AAAA,EAC7D;AACA,SAAO;AACT;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,WAAW,QAAQ,YAAY,gBAAgB,WAAW,WAAW,YAAY,YAAY,GAAG,aAAa;AACpH,QAAM,SAAS,gBAAgB,QAAQ,SAAS,aAAa,SAAS;AACtE,QAAM,cAAc,gBAAgB,YAAY,cAAc,aAAa,SAAS;AACpF,SAAO,OAAO,SAAS,GAAG,MAAM,yDAAyD,OAAO,OAAO;AACvG,SAAO,YAAY,SAAS,MAAM,YAAY,MAAM,OAAO,OAAO,MAAM,MAAM,YAAY,MAAM,OAAO,MAAM,YAAY,MAAM,OAAO,GAAG,MAAM,kEAAkE;AACjN,SAAO,eAAe,QAAQ,YAAY,WAAW,GAAG,MAAM,4EAA4E,cAAc;AACxJ,QAAM,SAAS,EAAE,OAAO,QAAQ,YAAY,YAAY;AACxD,QAAM,QAAQ,EAAE,eAAe,UAAU,WAAW,YAAY;AAChE,SAAO,OAAO,UAAU,WAAW,QAAQ,KAAK;AAClD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,UAAU,GAAG,UAAU,UAAU;AACxC,SAAO,WAAW,MAAM,GAAG,MAAM,gDAAgD,WAAW;AAC5F,SAAO,WAAW,MAAM,GAAG,MAAM,gDAAgD,WAAW;AAC5F,QAAM,KAAK,gBAAgB,GAAG,KAAK,UAAU;AAC7C,SAAO,GAAG,QAAQ,GAAG,MAAM,4CAA4C,GAAG,OAAO;AACjF,QAAM,QAAQ,GAAG;AACjB,QAAM,CAAC,GAAG,CAAC,IAAI,GAAG,MAAM,MAAM,EAAE;AAChC,MAAI,EAAE,YAAY,IAAI;AACpB,UAAM,IAAI,MAAM,yBAAyB,0DAA0D,KAAK;AAAA,EAC1G;AACA,MAAI,EAAE,YAAY,IAAI;AACpB,UAAM,IAAI,MAAM,yBAAyB,6DAA6D,KAAK;AAAA,EAC7G;AACA,MAAI,WAAW,GAAG;AAChB,eAAW;AAAA,EACb;AACA,MAAI,WAAW,GAAG;AAChB,eAAW;AAAA,EACb;AACA,QAAM,IAAI,QAAQ,MAAM,GAAG,GAAG,GAAG,OAAO,GAAG,CAAC,IAAI,CAAC,CAAC;AAClD,QAAM,IAAI,MAAM,GAAG,GAAG,GAAG,OAAO;AAChC,QAAM,KAAK,IAAI,GAAG,CAAC;AACnB,QAAM,SAAS,WAAW,UAAU,IAAI,OAAO,CAAC,UAAU,OAAO,CAAC,GAAG,aAAa,IAAI,OAAO,CAAC,UAAU,OAAO,CAAC,CAAC;AACjH,QAAM,OAAO,MAAM,CAAC,GAAG,CAAC,GAAG,GAAG,KAAK;AACnC,SAAO,QAAQ,MAAM,QAAQ,QAAQ,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,QAAQ,MAAM,QAAQ,KAAK,IAAI,CAAC,CAAC,GAAG,KAAK;AACtG;AACA,IAAI,WAAW,GAAG,EAAE,UAAU,CAAC;AAG/B,SAAS,aAAa,IAAI;AACxB,MAAI;AACJ,MAAI,MAAM,QAAQ,EAAE,GAAG;AACrB,sBAAkB;AAClB,WAAO,MAAM,QAAQ,GAAG,SAAS,GAAG,MAAM,mEAAmE;AAC7G,UAAM,MAAM,GAAG,GAAG,MAAM;AACxB,aAAS,IAAI,GAAG,IAAI,GAAG,QAAQ,EAAE,GAAG;AAClC,aAAO,GAAG,GAAG,MAAM,OAAO,KAAK,MAAM,iEAAiE,GAAG,GAAG,MAAM,UAAU,MAAM;AAAA,IACpI;AAAA,EACF,OAAO;AACL,sBAAkB;AAClB,SAAK,MAAM,IAAI,GAAG,MAAM,IAAI,CAAC,EAAE,IAAI,CAAC,MAAM,QAAQ,GAAG,CAAC,CAAC,CAAC,CAAC;AAAA,EAC3D;AACA,SAAO,GAAG,UAAU,GAAG,GAAG,MAAM,IAAI,MAAM,oCAAoC,GAAG,yCAAyC,GAAG,GAAG,MAAM,MAAM;AAC5I,QAAM,KAAK,CAAC;AACZ,QAAM,OAAO;AACb,WAAS,IAAI,GAAG,IAAI,GAAG,QAAQ,EAAE,GAAG;AAClC,OAAG,KAAK,OAAO,KAAK,MAAM;AACxB,UAAI,IAAI,KAAK;AACb,UAAI,IAAI,GAAG;AACT,iBAAS,IAAI,GAAG,IAAI,GAAG,EAAE,GAAG;AAC1B,gBAAM,OAAO,IAAI,KAAK,IAAI,GAAG,IAAI,CAAC,CAAC,GAAG,GAAG,EAAE;AAC3C,cAAI,IAAI,GAAG,IAAI;AAAA,QACjB;AAAA,MACF;AACA,aAAO,IAAI,GAAG,KAAK,GAAG,WAAW,CAAC;AAAA,IACpC,CAAC,CAAC;AAAA,EACJ;AACA,MAAI,iBAAiB;AACnB,WAAO,MAAM,IAAI,CAAC;AAAA,EACpB,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,IAAI,GAAG,eAAe,OAAO;AACpC,SAAO,EAAE,QAAQ,GAAG,MAAM,gEAAgE,EAAE,MAAM;AAClG,MAAI,EAAE,SAAS,GAAG;AAChB,WAAO,KAAK,GAAG,YAAY;AAAA,EAC7B,OAAO;AACL,UAAM,gBAAgB,EAAE,MAAM,MAAM,GAAG,EAAE,MAAM,SAAS,CAAC,EAAE,OAAO,CAAC,OAAO,SAAS,QAAQ,IAAI;AAC/F,UAAM,OAAO,QAAQ,QAAQ,GAAG;AAAA,MAC9B;AAAA,MACA,EAAE,MAAM,EAAE,MAAM,SAAS;AAAA,MACzB,EAAE,MAAM,EAAE,MAAM,SAAS;AAAA,IAC3B,CAAC,GAAG,CAAC;AACL,UAAM,OAAO,CAAC;AACd,UAAM,OAAO,CAAC;AACd,SAAK,QAAQ,CAAC,QAAQ;AACpB,YAAM,CAAC,KAAK,GAAG,IAAI,KAAK,KAAK,YAAY;AACzC,WAAK,KAAK,GAAG;AACb,WAAK,KAAK,GAAG;AAAA,IACf,CAAC;AACD,UAAM,IAAI,QAAQ,MAAM,MAAM,CAAC,GAAG,EAAE,KAAK;AACzC,UAAM,IAAI,QAAQ,MAAM,MAAM,CAAC,GAAG,EAAE,KAAK;AACzC,WAAO,CAAC,GAAG,CAAC;AAAA,EACd;AACF;AACA,SAAS,KAAK,GAAG,eAAe,OAAO;AACrC,SAAO,OAAO,KAAK,MAAM;AACvB,WAAO,EAAE,MAAM,WAAW,GAAG,MAAM,0CAA0C,EAAE,MAAM,iBAAiB;AACtG,UAAM,IAAI,EAAE,MAAM;AAClB,UAAM,IAAI,EAAE,MAAM;AAClB,QAAI,IAAI,IAAI,CAAC;AACb,QAAI,IAAI,MAAM,CAAC;AACf,UAAM,QAAQ,SAAS,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AACpC,QAAI,IAAI,MAAM,KAAK;AACnB,UAAM,QAAQ,KAAK,IAAI,IAAI;AAC3B,aAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,YAAM,QAAQ;AACd,YAAM,QAAQ;AACd,YAAM,QAAQ;AACd,OAAC,GAAG,GAAG,CAAC,IAAI,OAAO,KAAK,MAAM;AAC5B,cAAM,SAAS,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,GAAG,CAAC,CAAC;AAC1C,cAAM,QAAQ,KAAK,MAAM;AACzB,cAAM,MAAM,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AACnC,cAAM,IAAI,MAAM,QAAQ,KAAK,CAAC,GAAG,SAAS,CAAC,CAAC,EAAE,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAClE,cAAM,KAAK,IAAI,KAAK,IAAI,GAAG,KAAK,CAAC;AACjC,cAAM,OAAO,IAAI,QAAQ,EAAE;AAC3B,YAAI,KAAK,MAAM,OAAO,GAAG;AACvB,cAAI,MAAM,KAAK;AAAA,QACjB,OAAO;AACL,cAAI,OAAO;AAAA,YACT;AAAA,YACA,MAAM,MAAM,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,MAAM,KAAK,GAAG,KAAK,MAAM,EAAE,CAAC;AAAA,UACxD,GAAG,CAAC;AAAA,QACN;AACA,cAAM,MAAM,IAAI,IAAI,OAAO,GAAG,EAAE,GAAG,KAAK,CAAC;AACzC,cAAM,WAAW,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,GAAG,CAAC,CAAC;AAC5C,cAAM,YAAY,IAAI,KAAK,CAAC;AAC5B,cAAM,KAAK,UAAU,CAAC;AACtB,YAAI,MAAM,GAAG;AACX,cAAI,IAAI,UAAU,OAAO,WAAW,OAAO,IAAI,QAAQ,CAAC,CAAC;AAAA,QAC3D,OAAO;AACL,gBAAM,YAAY,IAAI,UAAU,OAAO,WAAW,OAAO,IAAI,QAAQ,CAAC,CAAC;AACvE,cAAI,OAAO,CAAC,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,GAAG,SAAS,GAAG,CAAC;AAAA,QACrD;AACA,cAAM,aAAa,UAAU,SAAS;AACtC,cAAM,WAAW,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,EAAE,MAAM,KAAK,CAAC,CAAC;AACrD,YAAI,MAAM,GAAG;AACX,cAAI,IAAI,UAAU,OAAO,OAAO,UAAU,CAAC,GAAG,UAAU,CAAC;AAAA,QAC3D,OAAO;AACL,gBAAM,YAAY,IAAI,UAAU,OAAO,OAAO,UAAU,CAAC,GAAG,UAAU,CAAC;AACvE,cAAI,OAAO,CAAC,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,GAAG,SAAS,GAAG,CAAC;AAAA,QACrD;AACA,eAAO,CAAC,GAAG,GAAG,CAAC;AAAA,MACjB,CAAC;AACD,cAAQ,CAAC,OAAO,OAAO,KAAK,CAAC;AAAA,IAC/B;AACA,QAAI,CAAC,gBAAgB,IAAI,GAAG;AAC1B,UAAI,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAC3B,UAAI,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,IAC7B;AACA,WAAO,CAAC,GAAG,CAAC;AAAA,EACd,CAAC;AACH;AACA,IAAI,KAAK,GAAG,EAAE,IAAI,CAAC;AAGnB,IAAI;AAAA,CACH,SAAS,YAAY;AACpB,aAAW,WAAW,UAAU,KAAK;AACrC,aAAW,WAAW,UAAU,KAAK;AACrC,aAAW,WAAW,SAAS,KAAK;AACpC,aAAW,WAAW,4BAA4B,KAAK;AACzD,GAAG,cAAc,YAAY,CAAC,EAAE;AAGhC,SAAS,qBAAqB,SAAS,SAAS,YAAY,UAAU,wBAAwB;AAC5F,QAAM,UAAU,gBAAgB,SAAS,UAAU,qBAAqB;AACxE,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,qBAAqB;AAAA,EACtE;AACA,QAAM,eAAe,YAAY,OAAO,UAAU,IAAI,SAAS,QAAQ;AACvE,MAAI,cAAc,UAAU,MAAM;AAChC,WAAO;AAAA,EACT;AACA,MAAI,cAAc,UAAU,KAAK;AAC/B,WAAO,KAAK,YAAY;AAAA,EAC1B;AACA,MAAI,cAAc,UAAU,MAAM;AAChC,QAAI,YAAY,MAAM;AACpB,aAAO,KAAK,YAAY;AAAA,IAC1B,OAAO;AACL,YAAM,kBAAkB,QAAQ,OAAO,SAAS;AAChD,YAAM,SAAS,IAAI,KAAK,YAAY,GAAG,KAAK,QAAQ,CAAC;AACrD,aAAO,kBAAkB,IAAI,IAAI,QAAQ,OAAO,eAAe,CAAC,IAAI;AAAA,IACtE;AAAA,EACF;AACA,MAAI,cAAc,UAAU,wBAAwB;AAClD,QAAI,YAAY,MAAM;AACpB,aAAO,IAAI,KAAK,YAAY,GAAG,OAAO,QAAQ,IAAI,CAAC;AAAA,IACrD,OAAO;AACL,YAAM,qBAAqB,IAAI,UAAU,MAAM,QAAQ,KAAK,CAAC;AAC7D,YAAM,cAAc,KAAK,KAAK,SAAS,oBAAoB,OAAO,CAAC,CAAC,CAAC,GAAG,SAAS;AACjF,aAAO,IAAI,KAAK,YAAY,GAAG,WAAW;AAAA,IAC5C;AAAA,EACF;AACA,QAAM,MAAM,sBAAsB,WAAW;AAC/C;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,oBAAoB,QAAQ,aAAa,SAAS,YAAY,UAAU,wBAAwB;AACvG,QAAM,UAAU,gBAAgB,QAAQ,UAAU,oBAAoB;AACtE,QAAM,eAAe,gBAAgB,aAAa,eAAe,oBAAoB;AACrF,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,oBAAoB;AAAA,EACrE;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,+BAA+B;AACpF,QAAM,UAAU,IAAI,IAAI,SAAS,YAAY,CAAC;AAC9C,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,qBAAqB,GAAG,EAAE,oBAAoB,CAAC;AAGnD,SAAS,gBAAgB,QAAQ,aAAa,MAAM,SAAS,YAAY,UAAU,wBAAwB;AACzG,QAAM,UAAU,gBAAgB,QAAQ,UAAU,gBAAgB;AAClE,QAAM,eAAe,gBAAgB,aAAa,eAAe,gBAAgB;AACjF,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,gBAAgB;AAAA,EACjE;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,2BAA2B;AAChF,QAAM,MAAM,OAAO,CAAC;AACpB,QAAM,UAAU,IAAI,KAAK,KAAK,IAAI,SAAS,YAAY,GAAG,MAAM,IAAI,CAAC;AACrE,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,iBAAiB,GAAG,EAAE,gBAAgB,CAAC;AAG3C,SAAS,WAAW,QAAQ,aAAa,SAAS,YAAY,UAAU,wBAAwB;AAC9F,MAAI,UAAU,gBAAgB,QAAQ,UAAU,WAAW;AAC3D,QAAM,eAAe,gBAAgB,aAAa,eAAe,WAAW;AAC5E,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,WAAW;AAAA,EAC5D;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,sBAAsB;AAC3E,QAAM,MAAM,OAAO,CAAC;AACpB,YAAU,IAAI,IAAI,OAAO,CAAC,GAAG,OAAO,GAAG,GAAG;AAC1C,QAAM,UAAU,KAAK,IAAI,KAAK,IAAI,SAAS,YAAY,CAAC,CAAC;AACzD,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,WAAW,QAAQ,aAAa,SAAS,QAAQ,GAAG,YAAY,UAAU,wBAAwB;AACzG,QAAM,UAAU,gBAAgB,QAAQ,UAAU,WAAW;AAC7D,QAAM,eAAe,gBAAgB,aAAa,eAAe,WAAW;AAC5E,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,WAAW;AAAA,EAC5D;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,sBAAsB;AAC3E,QAAM,cAAc,OAAO,KAAK;AAChC,QAAM,QAAQ,IAAI,IAAI,cAAc,OAAO,CAAC;AAC5C,QAAM,YAAY,QAAQ,OAAO,WAAW;AAC5C,QAAM,SAAS,IAAI,OAAO,SAAS;AACnC,QAAM,UAAU,KAAK,IAAI,OAAO,GAAG,GAAG,OAAO,SAAS,CAAC,GAAG,IAAI,aAAa,MAAM,CAAC;AAClF,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,YAAY,GAAG,EAAE,WAAW,CAAC;AAGjC,SAAS,SAAS,QAAQ,aAAa,SAASF,YAAW,MAAM,YAAY,UAAU,wBAAwB;AAC7G,QAAM,UAAU,gBAAgB,QAAQ,UAAU,SAAS;AAC3D,QAAM,eAAe,gBAAgB,aAAa,eAAe,SAAS;AAC1E,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,SAAS;AAAA,EAC1D;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,oBAAoB;AACzE,QAAM,MAAM,OAAO,CAAC;AACpB,QAAM,gBAAgB,OAAOA,SAAQ;AACrC,QAAM,MAAM,IAAI,IAAI,SAAS,KAAK,KAAK,cAAc,aAAa,CAAC,CAAC,CAAC;AACrE,QAAM,MAAM,IAAI,IAAI,KAAK,OAAO,GAAG,KAAK,KAAK,IAAI,KAAK,YAAY,GAAG,aAAa,CAAC,CAAC;AACpF,QAAM,UAAU,IAAI,KAAK,GAAG;AAC5B,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,UAAU,GAAG,EAAE,SAAS,CAAC;AAG7B,SAAS,kBAAkB,QAAQ,aAAa,SAAS,YAAY,UAAU,wBAAwB;AACrG,QAAM,UAAU,gBAAgB,QAAQ,UAAU,kBAAkB;AACpE,QAAM,eAAe,gBAAgB,aAAa,eAAe,kBAAkB;AACnF,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,kBAAkB;AAAA,EACnE;AACA,oBAAkB,QAAQ,OAAO,aAAa,OAAO,6BAA6B;AAClF,QAAM,UAAU,kBAAkB,SAAS,YAAY;AACvD,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,mBAAmB,GAAG,EAAE,kBAAkB,CAAC;AAG/C,SAAS,+BAA+B,QAAQ,QAAQ;AACtD,QAAM,UAAU,gBAAgB,QAAQ,UAAU,+BAA+B;AACjF,QAAM,UAAU,gBAAgB,QAAQ,UAAU,+BAA+B;AACjF,oBAAkB,QAAQ,OAAO,QAAQ,OAAO,0CAA0C;AAC1F,QAAM,YAAY,KAAK,OAAO;AAC9B,QAAM,gBAAgB,IAAI,SAAS,OAAO;AAC1C,QAAM,gBAAgB,MAAM,IAAI,IAAI,IAAI,OAAO,CAAC,CAAC,CAAC;AAClD,SAAO,KAAK,IAAI,WAAW,aAAa,GAAG,aAAa;AAC1D;AACA,SAAS,qBAAqB,kBAAkB,QAAQ,SAAS,iBAAiB,GAAG,YAAY,UAAU,wBAAwB;AACjI,MAAI,oBAAoB,gBAAgB,kBAAkB,oBAAoB,qBAAqB;AACnG,QAAM,UAAU,gBAAgB,QAAQ,UAAU,qBAAqB;AACvE,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,qBAAqB;AAAA,EACtE;AACA,oBAAkB,kBAAkB,OAAO,QAAQ,OAAO,gCAAgC;AAC1F,MAAI,iBAAiB,GAAG;AACtB,UAAM,uBAAuB,OAAO,cAAc;AAClD,UAAM,MAAM,OAAO,CAAC;AACpB,UAAM,OAAO,OAAO,GAAG;AACvB,wBAAoB,KAAK,IAAI,mBAAmB,IAAI,KAAK,oBAAoB,CAAC,GAAG,IAAI,MAAM,oBAAoB,CAAC;AAAA,EAClH;AACA,QAAM,UAAU,+BAA+B,mBAAmB,OAAO;AACzE,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,+BAA+B,QAAQ,QAAQ,MAAM,IAAI;AAChE,MAAI,QAAQ,IAAI;AACd,UAAM,OAAO,OAAO;AAAA,EACtB;AACA,MAAI,QAAQ,OAAO,OAAO,GAAG;AAC3B,UAAM,MAAM,mGAAmG,OAAO,oBAAoB,KAAK;AAAA,EACjJ;AACA,QAAM,WAAW,WAAW,CAAC,SAAS,SAAS,SAAS;AACtD,UAAM,WAAW;AACjB,UAAM,MAAM,UAAU,SAAS,CAAC,GAAG,GAAG,QAAQ;AAC9C,UAAM,YAAY,IAAI,KAAK,SAAS,SAAS,GAAG,GAAG;AACnD,SAAK,CAAC,SAAS,SAAS,CAAC;AACzB,UAAM,aAAa,IAAI,IAAI,WAAW,OAAO,CAAC;AAC9C,UAAM,QAAQ,KAAK,YAAY,CAAC,GAAG,CAAC;AACpC,UAAM,WAAW,CAAC,IAAI,UAAU;AAC9B,YAAM,CAAC,SAAS,UAAU,IAAI;AAC9B,YAAM,UAAU,qBAAqB,GAAG,OAAO,CAAC,GAAG,CAAC;AACpD,aAAO;AAAA,QACL,IAAI,QAAQ,IAAI,OAAO,GAAG,IAAI,KAAK,SAAS,SAAS,GAAG,IAAI,UAAU,CAAC,CAAC;AAAA,QACxE,IAAI,QAAQ,IAAI,OAAO,GAAG,IAAI,IAAI,UAAU,GAAG,KAAK,SAAS,SAAS,CAAC,CAAC;AAAA,MAC1E;AAAA,IACF;AACA,WAAO,EAAE,OAAO,SAAS;AAAA,EAC3B,CAAC;AACD,SAAO,SAAS,QAAQ,MAAM;AAChC;AACA,SAAS,qBAAqB,cAAc,QAAQ,SAAS,iBAAiB,GAAG,YAAY,UAAU,wBAAwB;AAC7H,MAAI,gBAAgB,gBAAgB,cAAc,gBAAgB,qBAAqB;AACvF,QAAM,UAAU,gBAAgB,QAAQ,UAAU,qBAAqB;AACvE,MAAI,WAAW;AACf,MAAI,WAAW,MAAM;AACnB,eAAW,gBAAgB,SAAS,WAAW,qBAAqB;AAAA,EACtE;AACA,oBAAkB,cAAc,OAAO,QAAQ,OAAO,gCAAgC;AACtF,MAAI,iBAAiB,GAAG;AACtB,UAAM,uBAAuB,OAAO,cAAc;AAClD,UAAM,MAAM,OAAO,CAAC;AACpB,UAAM,aAAa,OAAO,cAAc,MAAM,EAAE;AAChD,oBAAgB,KAAK,IAAI,eAAe,IAAI,KAAK,oBAAoB,CAAC,GAAG,IAAI,sBAAsB,UAAU,CAAC;AAAA,EAChH;AACA,QAAM,UAAU,+BAA+B,eAAe,OAAO;AACrE,SAAO,oBAAoB,SAAS,UAAU,SAAS;AACzD;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,qBAAqB,SAAS,QAAQ,YAAY,cAAc;AACvE,QAAM,WAAW,gBAAgB,SAAS,WAAW,uBAAuB,OAAO;AACnF,QAAM,UAAU,gBAAgB,QAAQ,UAAU,qBAAqB;AACvE,QAAM,cAAc,gBAAgB,YAAY,cAAc,uBAAuB,OAAO;AAC5F,QAAM,gBAAgB,gBAAgB,cAAc,gBAAgB,uBAAuB,QAAQ,KAAK;AACxG,MAAI,SAAS,SAAS,GAAG;AACvB,UAAM,IAAI,MAAM;AAAA,UACV,SAAS,OAAO;AAAA,EACxB;AACA,MAAI,QAAQ,SAAS,GAAG;AACtB,UAAM,IAAI,MAAM,gDAAgD,QAAQ,OAAO;AAAA,EACjF;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,qDAAqD,YAAY,OAAO;AAAA,EAC1F;AACA,MAAI,cAAc,SAAS,GAAG;AAC5B,UAAM,IAAI,MAAM,uDAAuD,cAAc,OAAO;AAAA,EAC9F;AACA,QAAM,SAAS;AAAA,IACb,SAAS;AAAA,IACT,QAAQ;AAAA,IACR,YAAY;AAAA,IACZ,cAAc;AAAA,EAChB;AACA,QAAM,SAAS,OAAO,UAAU,qBAAqB,MAAM;AAC3D,SAAO;AAAA,IACL,eAAe,OAAO;AAAA,IACtB,cAAc,OAAO;AAAA,IACrB,mBAAmB,OAAO;AAAA,IAC1B,iBAAiB,OAAO;AAAA,EAC1B;AACF;AACA,IAAI,sBAAsB,GAAG,EAAE,qBAAqB,CAAC;AAGrD,SAAS,eAAe,cAAc,YAAY,UAAU;AAC1D,QAAM,gBAAgB,gBAAgB,cAAc,gBAAgB,iBAAiB,OAAO;AAC5F,QAAM,cAAc,gBAAgB,YAAY,cAAc,iBAAiB,OAAO;AACtF,QAAM,YAAY,gBAAgB,UAAU,YAAY,iBAAiB,OAAO;AAChF,MAAI,cAAc,SAAS,GAAG;AAC5B,UAAM,IAAI,MAAM;AAAA,UACV,cAAc,OAAO;AAAA,EAC7B;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,qDAAqD,YAAY,OAAO;AAAA,EAC1F;AACA,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,IAAI,MAAM,mDAAmD,UAAU,OAAO;AAAA,EACtF;AACA,QAAM,SAAS;AAAA,IACb,cAAc;AAAA,IACd,YAAY;AAAA,IACZ,UAAU;AAAA,EACZ;AACA,QAAM,SAAS,OAAO,UAAU,eAAe,MAAM;AACrD,SAAO,EAAE,eAAe,OAAO,IAAI,aAAa,OAAO,GAAG;AAC5D;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,SAAS,mBAAmB,MAAM,SAAS,YAAY;AACrD,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,mBAAmB;AAC/D,QAAM,WAAW,gBAAgB,SAAS,WAAW,qBAAqB,OAAO;AACjF,QAAM,cAAc,gBAAgB,YAAY,cAAc,qBAAqB,OAAO;AAC1F,MAAI,MAAM,OAAO,GAAG;AAClB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,SAAS,SAAS,GAAG;AACvB,UAAM,IAAI,MAAM;AAAA,YACR,SAAS,OAAO;AAAA,EAC1B;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM;AAAA,YACR,YAAY,OAAO;AAAA,EAC7B;AACA,QAAM,SAAS;AAAA,IACb,MAAM;AAAA,IACN,SAAS;AAAA,IACT,YAAY;AAAA,EACd;AACA,SAAO,OAAO,UAAU,mBAAmB,MAAM;AACnD;AACA,IAAI,oBAAoB,GAAG,EAAE,mBAAmB,CAAC;AAGjD,SAAS,kBAAkB,MAAM,SAAS,YAAY;AACpD,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,kBAAkB;AAC9D,QAAM,WAAW,gBAAgB,SAAS,WAAW,oBAAoB,OAAO;AAChF,QAAM,cAAc,gBAAgB,YAAY,cAAc,oBAAoB,OAAO;AACzF,MAAI,MAAM,OAAO,GAAG;AAClB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,SAAS,SAAS,GAAG;AACvB,UAAM,IAAI,MAAM;AAAA,WACT,SAAS,OAAO;AAAA,EACzB;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM;AAAA,WACT,YAAY,OAAO;AAAA,EAC5B;AACA,QAAM,SAAS;AAAA,IACb,MAAM;AAAA,IACN,SAAS;AAAA,IACT,YAAY;AAAA,EACd;AACA,SAAO,OAAO,UAAU,kBAAkB,MAAM;AAClD;AACA,IAAI,mBAAmB,GAAG,EAAE,kBAAkB,CAAC;AAG/C,SAAS,cAAc,MAAM,YAAY,WAAW,aAAa,SAAS,WAAW,UAAU,wBAAwB;AACrH,QAAM,QAAQ,gBAAgB,MAAM,QAAQ,gBAAgB,QAAQ;AACpE,MAAI,MAAM,UAAU,UAAU;AAC5B,UAAM,IAAI,MAAM,iCAAiC;AAAA,EACnD;AACA,MAAI,MAAM,MAAM,WAAW,GAAG;AAC5B,UAAM,IAAI,MAAM,+BAA+B,MAAM,OAAO;AAAA,EAC9D;AACA,QAAM,cAAc,gBAAgB,YAAY,cAAc,cAAc;AAC5E,MAAI,YAAY,UAAU,SAAS;AACjC,UAAM,IAAI,MAAM,uCAAuC;AAAA,EACzD;AACA,QAAM,QAAQ;AAAA,IACZ;AAAA,IACA;AAAA,IACA;AAAA,IACA,UAAU;AAAA,IACV;AAAA,IACA;AAAA,EACF;AACA,QAAM,SAAS,EAAE,MAAM,OAAO,YAAY,YAAY;AACtD,QAAM,SAAS,OAAO,UAAU,cAAc,QAAQ,KAAK;AAC3D,SAAO,EAAE,QAAQ,OAAO,IAAI,cAAc,OAAO,GAAG;AACtD;AACA,IAAI,eAAe,GAAG,EAAE,cAAc,CAAC;AAGvC,SAAS,aAAa,QAAQ,WAAW,YAAY,MAAM;AACzD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,eAAe,QAAQ;AACvE,QAAM,aAAa,gBAAgB,WAAW,aAAa,eAAe,QAAQ;AAClF,MAAI,OAAO,SAAS,GAAG;AACrB,UAAM,IAAI,MAAM,+CAA+C,OAAO,OAAO;AAAA,EAC/E;AACA,MAAI,WAAW,SAAS,GAAG;AACzB,UAAM,IAAI,MAAM,mDAAmD,WAAW,OAAO;AAAA,EACvF;AACA,QAAM,QAAQ,EAAE,UAAU;AAC1B,QAAM,SAAS,EAAE,OAAO,QAAQ,WAAW,WAAW;AACtD,QAAM,SAAS,OAAO,UAAU,aAAa,QAAQ,KAAK;AAC1D,SAAO,EAAE,SAAS,OAAO,IAAI,QAAQ,OAAO,IAAI,OAAO,OAAO,GAAG;AACnE;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,SAAS,wBAAwB,QAAQ,YAAY;AACnD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,0BAA0B,QAAQ;AAClF,QAAM,QAAQ,EAAE,WAAW;AAC3B,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,QAAM,SAAS,EAAE,OAAO,OAAO;AAC/B,SAAO,OAAO,UAAU,wBAAwB,QAAQ,KAAK;AAC/D;AACA,IAAI,yBAAyB,GAAG,EAAE,wBAAwB,CAAC;AAG3D,IAAI,WAAW;AAAA,EACb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cAAc,aAAa;AAAA,EACzC,SAAS,GAAG,aAAa,OAAO,SAAS;AACvC,UAAM,EAAE,OAAO,OAAO,OAAO,IAAI,KAAK,iBAAiB,GAAG,OAAO;AACjE,QAAI,WAAW,MAAM;AACnB,YAAM,YAAY,QAAQ,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,QAAQ,OAAO,EAAE,MAAM,EAAE;AAC/E,WAAK,eAAe,SAAS;AAAA,IAC/B,OAAO;AACL,WAAK,eAAe,MAAM;AAAA,IAC5B;AACA,YAAQ,MAAM;AACd,QAAI,YAAY;AACd,aAAO;AAAA,IACT,OAAO;AACL,YAAM,QAAQ;AACd,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,IAAI,aAAa;AACf,QAAI,KAAK,eAAe,MAAM;AAC5B,WAAK,cAAc;AAAA,IACrB;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EACA,sBAAsB;AACpB,SAAK,cAAc,KAAK,aAAa;AAAA,EACvC;AAAA,EACA,iBAAiB,GAAG,SAAS;AAC3B,WAAO,cAAc,GAAG,OAAO;AAAA,EACjC;AAAA,EACA,UAAU;AACR,QAAI,KAAK,eAAe,MAAM;AAC5B,cAAQ,KAAK,WAAW;AAAA,IAC1B;AAAA,EACF;AAAA,EACA,MAAM,iBAAiB;AACrB,QAAI,KAAK,eAAe,MAAM;AAC5B,WAAK,cAAc;AAAA,IACrB;AACA,WAAO;AAAA,MACL,MAAM;AAAA,MACN,QAAQ,OAAO,KAAK,aAAa,OAAO;AAAA,IAC1C;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,IAAI,MAAM,yDAAyD;AAAA,EAC3E;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,UAAM,IAAI,MAAM,4DAA4D,KAAK,aAAa,GAAG;AAAA,EACnG;AAAA,EACA,MAAM,kBAAkB,cAAc;AACpC,SAAK,eAAe,MAAM,aAAa,GAAG,OAAO,KAAK,GAAG;AACzD,WAAO,aAAa,MAAM,CAAC;AAAA,EAC7B;AACF;AACA,OAAO,eAAe,WAAW,OAAO,aAAa;AAAA,EACnD,OAAO,CAAC,aAAa;AACnB,WAAO,SAAS,YAAY,QAAQ,SAAS,oBAAoB,QAAQ,SAAS,kBAAkB;AAAA,EACtG;AACF,CAAC;AAGD,IAAI,oBAAoB,cAAc,UAAU;AAAA,EAC9C,YAAY,cAAc,KAAKA,YAAW,MAAM;AAC9C,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,MAAM;AACX,SAAK,UAAUA;AACf,SAAK,mBAAmB,CAAC;AACzB,SAAK,qBAAqB,CAAC;AAC3B,QAAIA,aAAY,MAAM;AACpB,WAAK,UAAU,OAAO,QAAQ,QAAQ;AAAA,IACxC;AAAA,EACF;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,gBAAgB,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACnI,kBAAc,QAAQ,CAAC,MAAM,MAAM;AACjC,YAAM,QAAQ,OAAO,oBAAoB;AACzC,YAAM,YAAY;AAClB,UAAI,KAAK,iBAAiB,MAAM,MAAM;AACpC,aAAK,iBAAiB,KAAK;AAAA,UACzB,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,UAAI,KAAK,mBAAmB,MAAM,MAAM;AACtC,aAAK,mBAAmB,KAAK;AAAA,UAC3B,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,YAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,UAAI,YAAY,MAAM;AACpB;AAAA,MACF;AACA,YAAM,kBAAkB,KAAK,iBAAiB,GAAG;AACjD,YAAM,oBAAoB,KAAK,mBAAmB,GAAG;AACrD,WAAK,MAAM;AACT,cAAM,qBAAqB,KAAK,IAAI,iBAAiB,KAAK,GAAG,GAAG,IAAI,OAAO,QAAQ,GAAG,IAAI,KAAK,GAAG,CAAC;AACnG,cAAM,UAAU,IAAI,IAAI,KAAK,KAAK,mBAAmB,KAAK,OAAO,CAAC,GAAG,KAAK,KAAK,iBAAiB,KAAK,OAAO,CAAC,CAAC,GAAG,QAAQ;AACzH,cAAM,uBAAuB,KAAK,IAAI,mBAAmB,KAAK,GAAG,GAAG,IAAI,OAAO,OAAO,GAAG,IAAI,KAAK,GAAG,CAAC;AACtG,wBAAgB,OAAO,kBAAkB;AACzC,0BAAkB,OAAO,oBAAoB;AAC7C,cAAM,WAAW,KAAK,IAAI,SAAS,CAAC,KAAK,YAAY,GAAG,KAAK;AAC7D,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AAAA,IACH,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,QAAI,KAAK,sBAAsB,MAAM;AACnC,cAAQ,KAAK,iBAAiB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AACpD,cAAQ,KAAK,mBAAmB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IACxD;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,YAAY,CAAC,GAAG,KAAK,kBAAkB,GAAG,KAAK,kBAAkB;AACvE,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,UAAU,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,cAAc,QAAQ,EAAE,SAAS,EAAE,CAAC;AAAA,EAClH;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,UAAM,gBAAgB,aAAa,SAAS;AAC5C,UAAM,YAAY;AAClB,SAAK,mBAAmB,aAAa,MAAM,GAAG,aAAa,EAAE,IAAI,CAAC,OAAO;AAAA,MACvE,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AACF,SAAK,qBAAqB,aAAa,MAAM,eAAe,gBAAgB,CAAC,EAAE,IAAI,CAAC,OAAO;AAAA,MACzF,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AAAA,EACJ;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,OAAO,KAAK;AAAA,MACZ,WAAW,KAAK;AAAA,IAClB;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,QAAQ,OAAO,UAAU;AAAA,EACzE;AACF;AACA,kBAAkB,YAAY;AAC9B,cAAc,iBAAiB;AAG/B,IAAI,mBAAmB,cAAc,UAAU;AAAA,EAC7C,YAAY,cAAc,0BAA0B,KAAK;AACvD,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,0BAA0B;AAC/B,SAAK,mBAAmB,CAAC;AAAA,EAC3B;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,gBAAgB,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACnI,kBAAc,QAAQ,CAAC,MAAM,MAAM;AACjC,YAAM,QAAQ,OAAO,oBAAoB;AACzC,UAAI,KAAK,iBAAiB,MAAM,MAAM;AACpC,cAAM,YAAY;AAClB,aAAK,iBAAiB,KAAK;AAAA,UACzB,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,KAAK,MAAM,OAAO,KAAK,uBAAuB,EAAE,SAAS,SAAS,CAAC;AAAA,QAC1F;AAAA,MACF;AACA,YAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,UAAI,YAAY,MAAM;AACpB;AAAA,MACF;AACA,YAAM,kBAAkB,KAAK,iBAAiB,GAAG;AACjD,WAAK,MAAM;AACT,cAAM,qBAAqB,KAAK,iBAAiB,OAAO,QAAQ,CAAC;AACjE,wBAAgB,OAAO,kBAAkB;AACzC,cAAM,WAAW,KAAK,IAAI,IAAI,UAAU,KAAK,KAAK,oBAAoB,OAAO,QAAQ,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,KAAK,YAAY,GAAG,KAAK;AAC7H,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AAAA,IACH,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,QAAI,KAAK,oBAAoB,MAAM;AACjC,cAAQ,KAAK,iBAAiB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IACtD;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,KAAK,iBAAiB,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,cAAc,QAAQ,EAAE,SAAS,EAAE,CAAC;AAAA,EAC9H;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,UAAM,YAAY;AAClB,SAAK,mBAAmB,aAAa,IAAI,CAAC,OAAO,EAAE,cAAc,EAAE,MAAM,UAAU,EAAE,OAAO,SAAS,SAAS,EAAE,EAAE;AAAA,EACpH;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,2BAA2B,KAAK;AAAA,IAClC;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,0BAA0B;AAAA,EAC1E;AACF;AACA,iBAAiB,YAAY;AAC7B,cAAc,gBAAgB;AAG9B,IAAI,gBAAgB,cAAc,UAAU;AAAA,EAC1C,YAAY,cAAc,OAAO,OAAOA,YAAW,MAAM;AACvD,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,UAAUA;AACf,SAAK,yBAAyB,CAAC;AAC/B,SAAK,0BAA0B,CAAC;AAChC,SAAK,MAAM;AACT,WAAK,WAAW,OAAO,KAAK,EAAE,SAAS;AACvC,WAAK,WAAW,OAAO,KAAK,EAAE,SAAS;AAAA,IACzC,CAAC;AACD,QAAIA,aAAY,MAAM;AACpB,WAAK,UAAU,OAAO,QAAQ,QAAQ;AAAA,IACxC;AAAA,EACF;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,MAAM,EAAE,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACxH,SAAK,MAAM;AACT,YAAM,mBAAmB,IAAI,GAAG,KAAK,QAAQ;AAC7C,YAAM,mBAAmB,IAAI,GAAG,KAAK,QAAQ;AAC7C,eAAS,QAAQ,CAAC,MAAM,MAAM;AAC5B,cAAM,QAAQ,OAAO,oBAAoB;AACzC,cAAM,YAAY;AAClB,YAAI,KAAK,uBAAuB,MAAM,MAAM;AAC1C,eAAK,uBAAuB,KAAK;AAAA,YAC/B,cAAc,GAAG;AAAA,YACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,UAC3D;AAAA,QACF;AACA,YAAI,KAAK,wBAAwB,MAAM,MAAM;AAC3C,eAAK,wBAAwB,KAAK;AAAA,YAChC,cAAc,GAAG;AAAA,YACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,UAC3D;AAAA,QACF;AACA,cAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,YAAI,YAAY,MAAM;AACpB;AAAA,QACF;AACA,cAAM,cAAc,KAAK,uBAAuB,GAAG;AACnD,cAAM,eAAe,KAAK,wBAAwB,GAAG;AACrD,cAAM,iBAAiB,KAAK,IAAI,aAAa,KAAK,KAAK,GAAG,IAAI,UAAU,IAAI,KAAK,KAAK,CAAC;AACvF,cAAM,kBAAkB,KAAK,IAAI,cAAc,KAAK,KAAK,GAAG,IAAI,OAAO,QAAQ,GAAG,IAAI,KAAK,KAAK,CAAC;AACjG,cAAM,2BAA2B,IAAI,gBAAgB,gBAAgB;AACrE,cAAM,4BAA4B,IAAI,iBAAiB,gBAAgB;AACvE,oBAAY,OAAO,cAAc;AACjC,qBAAa,OAAO,eAAe;AACnC,cAAM,WAAW,KAAK,IAAI,IAAI,0BAA0B,KAAK,KAAK,yBAAyB,GAAG,KAAK,OAAO,CAAC,GAAG,CAAC,KAAK,YAAY,GAAG,KAAK;AACxI,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AACD,WAAK,SAAS,OAAO,IAAI,KAAK,UAAU,KAAK,KAAK,CAAC;AACnD,WAAK,SAAS,OAAO,IAAI,KAAK,UAAU,KAAK,KAAK,CAAC;AAAA,IACrD,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,SAAK,SAAS,QAAQ;AACtB,SAAK,SAAS,QAAQ;AACtB,QAAI,KAAK,0BAA0B,MAAM;AACvC,cAAQ,KAAK,uBAAuB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAC5D;AACA,QAAI,KAAK,2BAA2B,MAAM;AACxC,cAAQ,KAAK,wBAAwB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAC7D;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,YAAY,CAAC,GAAG,KAAK,wBAAwB,GAAG,KAAK,uBAAuB;AAClF,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,UAAU,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,cAAc,QAAQ,EAAE,SAAS,EAAE,CAAC;AAAA,EAClH;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,SAAK,MAAM;AACT,WAAK,SAAS,OAAO,IAAI,KAAK,OAAO,KAAK,cAAc,CAAC,CAAC;AAC1D,WAAK,SAAS,OAAO,IAAI,KAAK,OAAO,KAAK,cAAc,CAAC,CAAC;AAAA,IAC5D,CAAC;AACD,UAAM,gBAAgB,aAAa,SAAS;AAC5C,UAAM,YAAY;AAClB,SAAK,yBAAyB,aAAa,MAAM,GAAG,aAAa,EAAE,IAAI,CAAC,OAAO;AAAA,MAC7E,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AACF,SAAK,0BAA0B,aAAa,MAAM,eAAe,gBAAgB,CAAC,EAAE,IAAI,CAAC,OAAO;AAAA,MAC9F,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AAAA,EACJ;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,MACd,WAAW,KAAK;AAAA,IAClB;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,UAAU,OAAO,UAAU,OAAO,UAAU;AAAA,EAC5F;AACF;AACA,cAAc,YAAY;AAC1B,cAAc,aAAa;AAG3B,IAAI,kBAAkB,cAAc,UAAU;AAAA,EAC5C,YAAY,cAAc,OAAO,OAAOA,YAAW,MAAM,QAAQ,GAAG;AAClE,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,UAAUA;AACf,SAAK,QAAQ;AACb,SAAK,yBAAyB,CAAC;AAC/B,SAAK,6BAA6B,CAAC;AACnC,SAAK,MAAM;AACT,WAAK,YAAY,OAAO,CAAC,EAAE,SAAS;AACpC,WAAK,WAAW,OAAO,KAAK,EAAE,SAAS;AAAA,IACzC,CAAC;AACD,QAAIA,aAAY,MAAM;AACpB,WAAK,UAAU,OAAO,QAAQ,QAAQ;AAAA,IACxC;AAAA,EACF;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,gBAAgB,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACnI,SAAK,MAAM;AACT,YAAM,mBAAmB,IAAI,GAAG,KAAK,QAAQ;AAC7C,YAAM,KAAK,IAAI,CAAC,KAAK,cAAc,KAAK,IAAI,KAAK,WAAW,KAAK,KAAK,GAAG,CAAC,CAAC;AAC3E,oBAAc,QAAQ,CAAC,MAAM,MAAM;AACjC,cAAM,QAAQ,OAAO,oBAAoB;AACzC,cAAM,YAAY;AAClB,YAAI,KAAK,uBAAuB,MAAM,MAAM;AAC1C,eAAK,uBAAuB,KAAK;AAAA,YAC/B,cAAc,GAAG;AAAA,YACjB,UAAU,UAAU,KAAK,EAAE,SAAS,SAAS;AAAA,UAC/C;AAAA,QACF;AACA,YAAI,KAAK,2BAA2B,MAAM,MAAM;AAC9C,eAAK,2BAA2B,KAAK;AAAA,YACnC,cAAc,GAAG;AAAA,YACjB,UAAU,UAAU,KAAK,EAAE,SAAS,SAAS;AAAA,UAC/C;AAAA,QACF;AACA,cAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,YAAI,YAAY,MAAM;AACpB;AAAA,QACF;AACA,cAAM,cAAc,KAAK,uBAAuB,GAAG;AACnD,cAAM,kBAAkB,KAAK,2BAA2B,GAAG;AAC3D,cAAM,iBAAiB,KAAK,IAAI,aAAa,KAAK,KAAK,GAAG,IAAI,UAAU,IAAI,KAAK,KAAK,CAAC;AACvF,cAAM,MAAM,IAAI,iBAAiB,KAAK,KAAK;AAC3C,cAAM,MAAM,IAAI,QAAQ;AACxB,cAAM,qBAAqB,QAAQ,KAAK,GAAG;AAC3C,oBAAY,OAAO,cAAc;AACjC,wBAAgB,OAAO,kBAAkB;AACzC,cAAM,WAAW,KAAK,IAAI,IAAI,IAAI,gBAAgB,GAAG,IAAI,gBAAgB,KAAK,oBAAoB,KAAK,OAAO,CAAC,CAAC,GAAG,KAAK;AACxH,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AACD,WAAK,UAAU,OAAO,KAAK,KAAK,WAAW,CAAC,CAAC;AAC7C,WAAK,SAAS,OAAO,IAAI,KAAK,UAAU,KAAK,KAAK,CAAC;AAAA,IACrD,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,SAAK,SAAS,QAAQ;AACtB,SAAK,UAAU,QAAQ;AACvB,QAAI,KAAK,0BAA0B,MAAM;AACvC,cAAQ,KAAK,uBAAuB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAC5D;AACA,QAAI,KAAK,8BAA8B,MAAM;AAC3C,cAAQ,KAAK,2BAA2B,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAChE;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,IAAI,MAAM,iDAAiD;AAAA,EACnE;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,UAAM,IAAI,MAAM,iDAAiD;AAAA,EACnE;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,MACd,WAAW,KAAK;AAAA,MAChB,SAAS,KAAK;AAAA,IAChB;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,UAAU,OAAO,UAAU,OAAO,YAAY,OAAO,QAAQ;AAAA,EAC7G;AACF;AACA,gBAAgB,YAAY;AAC5B,cAAc,eAAe;AAG7B,IAAI,eAAe,cAAc,UAAU;AAAA,EACzC,YAAY,cAAc;AACxB,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,gBAAgB,YAAY;AAAA,EACnC;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,MAAM,EAAE,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACxH,aAAS,QAAQ,CAAC,MAAM,MAAM;AAC5B,YAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,UAAI,YAAY,MAAM;AACpB;AAAA,MACF;AACA,YAAM,QAAQ,OAAO,oBAAoB;AACzC,WAAK,MAAM;AACT,cAAM,WAAW,KAAK,IAAI,KAAK,GAAG,QAAQ,GAAG,KAAK;AAClD,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AAAA,IACH,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,gBAAgB,cAAc;AAC5B,SAAK,eAAe;AACpB,QAAI,KAAK,KAAK,MAAM;AAClB,WAAK,EAAE,QAAQ;AAAA,IACjB;AACA,SAAK,IAAI,KAAK,OAAO,CAAC,YAAY,CAAC;AAAA,EACrC;AAAA,EACA,UAAU;AACR,SAAK,EAAE,QAAQ;AAAA,EACjB;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC;AAAA,EACrC;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,QAAI,aAAa,WAAW,GAAG;AAC7B,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AAAA,EACF;AAAA,EACA,YAAY;AACV,WAAO,EAAE,gBAAgB,KAAK,aAAa;AAAA,EAC7C;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,eAAe;AAAA,EACvC;AACF;AACA,aAAa,YAAY;AACzB,cAAc,YAAY;AAG1B,IAAI,oBAAoB,cAAc,aAAa;AAAA,EACjD,YAAY,cAAc,UAAU,cAAc,OAAO;AACvD,UAAM,YAAY;AAClB,SAAK,eAAe;AACpB,SAAK,WAAW;AAChB,SAAK,cAAc;AACnB,SAAK,gBAAgB,CAAC;AACtB,SAAK,IAAI,OAAO,KAAK,QAAQ;AAAA,EAC/B;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,gBAAgB,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACnI,kBAAc,QAAQ,CAAC,MAAM,MAAM;AACjC,YAAM,QAAQ,OAAO,oBAAoB;AACzC,UAAI,KAAK,cAAc,MAAM,MAAM;AACjC,cAAM,YAAY;AAClB,aAAK,cAAc,KAAK;AAAA,UACtB,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,YAAM,eAAe,KAAK,cAAc,GAAG;AAC3C,YAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,UAAI,YAAY,MAAM;AACpB;AAAA,MACF;AACA,WAAK,MAAM;AACT,YAAI;AACJ,cAAM,kBAAkB,KAAK,IAAI,KAAK,GAAG,YAAY,GAAG,QAAQ;AAChE,YAAI,KAAK,aAAa;AACpB,qBAAW,KAAK,IAAI,KAAK,GAAG,KAAK,UAAU,IAAI,iBAAiB,KAAK,CAAC,CAAC,CAAC,GAAG,KAAK;AAAA,QAClF,OAAO;AACL,qBAAW,KAAK,IAAI,KAAK,GAAG,eAAe,GAAG,KAAK;AAAA,QACrD;AACA,qBAAa,OAAO,eAAe;AACnC,cAAM,OAAO,QAAQ;AAAA,MACvB,CAAC;AAAA,IACH,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,SAAK,EAAE,QAAQ;AACf,QAAI,KAAK,iBAAiB,MAAM;AAC9B,cAAQ,KAAK,cAAc,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IACnD;AAAA,EACF;AAAA,EACA,YAAY,UAAU;AACpB,SAAK,WAAW;AAAA,EAClB;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,KAAK,cAAc,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,cAAc,QAAQ,EAAE,SAAS,EAAE,CAAC;AAAA,EAC3H;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,UAAM,YAAY;AAClB,SAAK,gBAAgB,aAAa,IAAI,CAAC,OAAO,EAAE,cAAc,EAAE,MAAM,UAAU,EAAE,OAAO,SAAS,SAAS,EAAE,EAAE;AAAA,EACjH;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,YAAY,KAAK;AAAA,MACjB,eAAe,KAAK;AAAA,IACtB;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,aAAa,OAAO,cAAc;AAAA,EAClF;AACF;AACA,kBAAkB,YAAY;AAC9B,cAAc,iBAAiB;AAG/B,IAAI,mBAAmB,cAAc,UAAU;AAAA,EAC7C,YAAY,cAAc,QAAQ,KAAK,WAAW,GAAGA,YAAW,MAAM,WAAW,OAAO;AACtF,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,WAAW;AAChB,SAAK,UAAUA;AACf,SAAK,yBAAyB,CAAC;AAC/B,SAAK,qBAAqB,CAAC;AAC3B,SAAK,uBAAuB,CAAC;AAC7B,SAAK,WAAW;AAChB,QAAIA,aAAY,MAAM;AACpB,WAAK,UAAU,OAAO,QAAQ,QAAQ;AAAA,IACxC;AACA,QAAI,gBAAgB,MAAM;AACxB,YAAM,IAAI,MAAM,oDAAoD;AAAA,IACtE;AAAA,EACF;AAAA,EACA,eAAe,mBAAmB;AAChC,UAAM,gBAAgB,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,OAAO,KAAK,iBAAiB;AACnI,kBAAc,QAAQ,CAAC,MAAM,MAAM;AACjC,YAAM,QAAQ,OAAO,oBAAoB;AACzC,YAAM,YAAY;AAClB,UAAI,KAAK,uBAAuB,MAAM,MAAM;AAC1C,aAAK,uBAAuB,KAAK;AAAA,UAC/B,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,UAAI,KAAK,mBAAmB,MAAM,MAAM;AACtC,aAAK,mBAAmB,KAAK;AAAA,UAC3B,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,UAAI,KAAK,qBAAqB,MAAM,QAAQ,KAAK,UAAU;AACzD,aAAK,qBAAqB,KAAK;AAAA,UAC7B,cAAc,GAAG;AAAA,UACjB,UAAU,KAAK,MAAM,UAAU,KAAK,EAAE,SAAS,SAAS,CAAC;AAAA,QAC3D;AAAA,MACF;AACA,YAAM,WAAW,MAAM,QAAQ,iBAAiB,IAAI,kBAAkB,GAAG,SAAS,kBAAkB;AACpG,UAAI,YAAY,MAAM;AACpB;AAAA,MACF;AACA,YAAM,wBAAwB,KAAK,uBAAuB,GAAG;AAC7D,YAAM,qBAAqB,KAAK,mBAAmB,GAAG;AACtD,WAAK,MAAM;AACT,cAAM,2BAA2B,KAAK,IAAI,uBAAuB,KAAK,KAAK,GAAG,IAAI,OAAO,QAAQ,GAAG,IAAI,KAAK,KAAK,CAAC;AACnH,YAAI,KAAK,UAAU;AACjB,gBAAM,sBAAsB,KAAK,qBAAqB,GAAG;AACzD,gBAAM,yBAAyB,KAAK,IAAI,qBAAqB,KAAK,KAAK,GAAG,IAAI,UAAU,IAAI,KAAK,KAAK,CAAC;AACvG,gBAAM,mBAAmB,IAAI,IAAI,UAAU,KAAK,YAAY,GAAG,KAAK,IAAI,0BAA0B,KAAK,OAAO,sBAAsB,GAAG,KAAK,OAAO,CAAC,CAAC,CAAC;AACtJ,gBAAM,wBAAwB,KAAK,IAAI,oBAAoB,KAAK,QAAQ,GAAG,gBAAgB;AAC3F,gCAAsB,OAAO,wBAAwB;AACrD,8BAAoB,OAAO,sBAAsB;AACjD,6BAAmB,OAAO,qBAAqB;AAC/C,gBAAM,WAAW,IAAI,OAAO,qBAAqB;AACjD,gBAAM,OAAO,QAAQ;AAAA,QACvB,OAAO;AACL,gBAAM,4BAA4B,KAAK,IAAI,uBAAuB,KAAK,KAAK,GAAG,IAAI,OAAO,QAAQ,GAAG,IAAI,KAAK,KAAK,CAAC;AACpH,gBAAM,wBAAwB,KAAK,IAAI,oBAAoB,KAAK,QAAQ,GAAG,IAAI,IAAI,UAAU,KAAK,YAAY,GAAG,KAAK,KAAK,2BAA2B,KAAK,OAAO,CAAC,CAAC,CAAC;AACrK,gCAAsB,OAAO,yBAAyB;AACtD,6BAAmB,OAAO,qBAAqB;AAC/C,gBAAM,WAAW,IAAI,OAAO,qBAAqB;AACjD,gBAAM,OAAO,QAAQ;AAAA,QACvB;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AACD,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,QAAI,KAAK,0BAA0B,MAAM;AACvC,cAAQ,KAAK,uBAAuB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAC5D;AACA,QAAI,KAAK,wBAAwB,QAAQ,KAAK,UAAU;AACtD,cAAQ,KAAK,qBAAqB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IAC1D;AACA,QAAI,KAAK,sBAAsB,MAAM;AACnC,cAAQ,KAAK,mBAAmB,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAAA,IACxD;AAAA,EACF;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,YAAY,CAAC,GAAG,KAAK,wBAAwB,GAAG,KAAK,kBAAkB;AAC7E,QAAI,KAAK,UAAU;AACjB,gBAAU,KAAK,GAAG,KAAK,oBAAoB;AAAA,IAC7C;AACA,WAAO,CAAC,MAAM,KAAK,eAAe,CAAC,EAAE,OAAO,UAAU,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,cAAc,QAAQ,EAAE,SAAS,EAAE,CAAC;AAAA,EAClH;AAAA,EACA,MAAM,WAAW,cAAc;AAC7B,mBAAe,MAAM,KAAK,kBAAkB,YAAY;AACxD,UAAM,gBAAgB,KAAK,WAAW,aAAa,SAAS,IAAI,aAAa,SAAS;AACtF,UAAM,YAAY;AAClB,SAAK,yBAAyB,aAAa,MAAM,GAAG,aAAa,EAAE,IAAI,CAAC,OAAO;AAAA,MAC7E,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AACF,SAAK,qBAAqB,aAAa,MAAM,eAAe,gBAAgB,CAAC,EAAE,IAAI,CAAC,OAAO;AAAA,MACzF,cAAc,EAAE;AAAA,MAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,IACvC,EAAE;AACF,QAAI,KAAK,UAAU;AACjB,WAAK,uBAAuB,aAAa,MAAM,gBAAgB,GAAG,gBAAgB,CAAC,EAAE,IAAI,CAAC,OAAO;AAAA,QAC/F,cAAc,EAAE;AAAA,QAChB,UAAU,EAAE,OAAO,SAAS,SAAS;AAAA,MACvC,EAAE;AAAA,IACJ;AAAA,EACF;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,gBAAgB,KAAK;AAAA,MACrB,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,MACjB,WAAW,KAAK;AAAA,MAChB,YAAY,KAAK;AAAA,IACnB;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,OAAO,iBAAiB,OAAO,UAAU,OAAO,aAAa,OAAO,YAAY,OAAO,WAAW;AAAA,EACnH;AACF;AACA,iBAAiB,YAAY;AAC7B,cAAc,gBAAgB;AAG9B,IAAI,wBAAwB,MAAM;AAAA,EAChC,OAAO,IAAI,cAAc;AACvB,WAAO,IAAI,aAAa,YAAY;AAAA,EACtC;AAAA,EACA,OAAO,SAAS,cAAc,UAAU,cAAc,OAAO;AAC3D,WAAO,IAAI,kBAAkB,cAAc,UAAU,WAAW;AAAA,EAClE;AAAA,EACA,OAAO,QAAQ,cAAc,QAAQ,KAAK,WAAW,GAAGA,YAAW,MAAM,WAAW,OAAO;AACzF,WAAO,IAAI,iBAAiB,cAAc,OAAO,UAAUA,WAAU,QAAQ;AAAA,EAC/E;AAAA,EACA,OAAO,KAAK,eAAe,MAAM,QAAQ,KAAK,QAAQ,OAAOA,YAAW,MAAM;AAC5E,WAAO,IAAI,cAAc,cAAc,OAAO,OAAOA,SAAQ;AAAA,EAC/D;AAAA,EACA,OAAO,SAAS,eAAe,MAAM,MAAM,MAAMA,YAAW,MAAM;AAChE,WAAO,IAAI,kBAAkB,cAAc,KAAKA,SAAQ;AAAA,EAC1D;AAAA,EACA,OAAO,OAAO,eAAe,MAAM,QAAQ,KAAK,QAAQ,OAAOA,YAAW,MAAM,QAAQ,GAAG;AACzF,WAAO,IAAI,gBAAgB,cAAc,OAAO,OAAOA,WAAU,KAAK;AAAA,EACxE;AAAA,EACA,OAAO,QAAQ,cAAc,0BAA0B,KAAK;AAC1D,WAAO,IAAI,iBAAiB,cAAc,uBAAuB;AAAA,EACnE;AACF;AAGA,IAAI,QAAQ;AAAA,EACV,KAAK,sBAAsB;AAAA,EAC3B,UAAU,sBAAsB;AAAA,EAChC,UAAU,sBAAsB;AAAA,EAChC,SAAS,sBAAsB;AAAA,EAC/B,SAAS,sBAAsB;AAAA,EAC/B,QAAQ,sBAAsB;AAAA,EAC9B,MAAM,sBAAsB;AAC9B;AAGA,IAAI,iBAAiB,MAAM;AACzB,MAAI,OAAO,0BAA0B,aAAa;AAChD,WAAO;AAAA,EACT,WAAW,OAAO,iBAAiB,aAAa;AAC9C,WAAO;AAAA,EACT;AACA,SAAO,CAAC,MAAM,EAAE;AAClB,GAAG;AACH,SAAS,YAAY;AACnB,SAAO,IAAI,QAAQ,CAAC,YAAY,cAAc,MAAM,QAAQ,CAAC,CAAC;AAChE;AAGA,IAAI,uBAAuB,CAAC;AAC5BH,UAAS,sBAAsB;AAAA,EAC7B,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,OAAO,MAAM;AAAA,EACb,uBAAuB,MAAM;AAAA,EAC7B,kBAAkB,MAAM;AAAA,EACxB,YAAY,MAAM;AAAA,EAClB,iBAAiB,MAAM;AAAA,EACvB,iBAAiB,MAAM;AAAA,EACvB,4BAA4B,MAAM;AAAA,EAClC,4BAA4B,MAAM;AAAA,EAClC,wBAAwB,MAAM;AAAA,EAC9B,oBAAoB,MAAM;AAAA,EAC1B,sBAAsB,MAAM;AAAA,EAC5B,iBAAiB,MAAM;AAAA,EACvB,qBAAqB,MAAM;AAAA,EAC3B,2BAA2B,MAAM;AAAA,EACjC,kBAAkB,MAAM;AAAA,EACxB,mCAAmC,MAAM;AAAA,EACzC,sBAAsB,MAAM;AAAA,EAC5B,qBAAqB,MAAM;AAAA,EAC3B,mBAAmB,MAAM;AAAA,EACzB,mBAAmB,MAAM;AAAA,EACzB,mBAAmB,MAAM;AAAA,EACzB,uBAAuB,MAAM;AAAA,EAC7B,0BAA0B,MAAM;AAAA,EAChC,2BAA2B,MAAM;AAAA,EACjC,iBAAiB,MAAM;AAAA,EACvB,mBAAmB,MAAM;AAAA,EACzB,mBAAmB,MAAM;AAAA,EACzB,yBAAyB,MAAM;AAAA,EAC/B,sBAAsB,MAAM;AAAA,EAC5B,gCAAgC,MAAM;AAAA,EACtC,sBAAsB,MAAM;AAAA,EAC5B,UAAU,MAAM;AAAA,EAChB,WAAW,MAAM;AAAA,EACjB,wBAAwB,MAAM;AAAA,EAC9B,wBAAwB,MAAM;AAAA,EAC9B,oBAAoB,MAAM;AAAA,EAC1B,kBAAkB,MAAM;AAAA,EACxB,qBAAqB,MAAM;AAAA,EAC3B,sBAAsB,MAAM;AAAA,EAC5B,sBAAsB,MAAM;AAAA,EAC5B,sBAAsB,MAAM;AAAA,EAC5B,sBAAsB,MAAM;AAAA,EAC5B,gBAAgB,MAAM;AAAA,EACtB,kBAAkB,MAAM;AAAA,EACxB,aAAa,MAAM;AAAA,EACnB,eAAe,MAAM;AAAA,EACrB,kBAAkB,MAAM;AAAA,EACxB,aAAa,MAAM;AAAA,EACnB,qBAAqB,MAAM;AAAA,EAC3B,4BAA4B,MAAM;AAAA,EAClC,qBAAqB,MAAM;AAAA,EAC3B,cAAc,MAAM;AAAA,EACpB,iDAAiD,MAAM;AAAA,EACvD,iDAAiD,MAAM;AAAA,EACvD,mDAAmD,MAAM;AAAA,EACzD,sDAAsD,MAAM;AAAA,EAC5D,iDAAiD,MAAM;AAAA,EACvD,iDAAiD,MAAM;AAAA,EACvD,0DAA0D,MAAM;AAAA,EAChE,+CAA+C,MAAM;AAAA,EACrD,wDAAwD,MAAM;AAAA,EAC9D,yDAAyD,MAAM;AAAA,EAC/D,8DAA8D,MAAM;AAAA,EACpE,0DAA0D,MAAM;AAAA,EAChE,wBAAwB,MAAM;AAAA,EAC9B,uBAAuB,MAAM;AAAA,EAC7B,KAAK,MAAM;AAAA,EACX,wBAAwB,MAAM;AAAA,EAC9B,oBAAoB,MAAM;AAAA,EAC1B,kBAAkB,MAAM;AAAA,EACxB,cAAc,MAAM;AAAA,EACpB,YAAY,MAAM;AAAA,EAClB,YAAY,MAAM;AAAA,EAClB,wBAAwB,MAAM;AAAA,EAC9B,mBAAmB,MAAM;AAAA,EACzB,YAAY,MAAM;AAAA,EAClB,2BAA2B,MAAM;AAAA,EACjC,eAAe,MAAM;AAAA,EACrB,qBAAqB,MAAM;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AAGD,SAAS,uBAAuB,QAAQ,MAAM;AAC5C,QAAM,OAAO,OAAO,GAAG;AACvB,SAAO,QAAQ,CAAC,OAAO,MAAM;AAC3B,WAAO,MAAM,WAAW,MAAM,MAAM,kBAAkB,0BAA0B,gDAAgD,OAAO;AAAA,EACzI,CAAC;AACD,SAAO,QAAQ,KAAK,OAAO,MAAM,MAAM,kBAAkB,qCAAqC,OAAO,IAAI;AACzG,QAAM,aAAa,OAAO;AAC1B,SAAO,QAAQ,CAAC,OAAO,MAAM;AAC3B,aAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,aAAO,MAAM,QAAQ,MAAM,OAAO,WAAW,IAAI,MAAM,kBAAkB,2BAA2B,OAAO,gDAAgD,+CAA+C,IAAI;AAAA,IAChN;AAAA,EACF,CAAC;AACH;AACA,SAAS,iBAAiB,QAAQ,MAAM;AACtC,QAAM,cAAc,OAAO,GAAG,MAAM;AACpC,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,gBAAY,SAAS,OAAO,GAAG;AAAA,EACjC;AACA,SAAO;AACT;AAGA,IAAI;AAAA,CACH,SAAS,mBAAmB;AAC3B,oBAAkB,kBAAkB,oBAAoB,KAAK;AAC7D,oBAAkB,kBAAkB,kBAAkB,KAAK;AAC3D,oBAAkB,kBAAkB,iBAAiB,KAAK;AAC1D,oBAAkB,kBAAkB,gBAAgB,KAAK;AACzD,oBAAkB,kBAAkB,gBAAgB,KAAK;AACzD,oBAAkB,kBAAkB,gBAAgB,KAAK;AAC3D,GAAG,qBAAqB,mBAAmB,CAAC,EAAE;AAC9C,SAAS,kCAAkC,YAAY,OAAO,YAAY;AACxE,MAAI,cAAc,IAAI,MAAM;AAC5B,MAAI,cAAc,QAAQ,SAAS,MAAM;AACvC,WAAO;AAAA,EACT;AACA,MAAI,SAAS,MAAM;AACjB,WAAO,YAAY,SAAS,aAAa,WAAW,QAAQ;AAC1D,kBAAY,KAAK,EAAE;AAAA,IACrB;AAAA,EACF,OAAO;AACL,kBAAc,MAAM,MAAM;AAAA,EAC5B;AACA,MAAI,cAAc,MAAM;AACtB,WAAO;AAAA,EACT;AACA,MAAI,aAAa,WAAW,WAAW,YAAY,QAAQ;AACzD,UAAM,IAAI,MAAM,4BAA4B,2CAA2C,aAAa,WAAW,4BAA4B,YAAY,QAAQ;AAAA,EACjK;AACA,WAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,UAAM,WAAW,WAAW;AAC5B,UAAM,sBAAsB,YAAY,YAAY,SAAS,WAAW,SAAS;AACjF,UAAM,iBAAiB,YAAY;AACnC,QAAI,YAAY,GAAG;AACjB,UAAI,kBAAkB,GAAG;AACvB,YAAI,mBAAmB,UAAU;AAC/B,gBAAM,IAAI,MAAM,4BAA4B,0CAA0C,IAAI,iBAAiB,sBAAsB,IAAI,iBAAiB,gBAAgB;AAAA,QACxK;AAAA,MACF,OAAO;AACL,oBAAY,uBAAuB;AAAA,MACrC;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,2BAA2B,yBAAyB;AAC3D,QAAM,eAAe;AAAA,IACnB,kBAAkB,iBAAiB;AAAA,IACnC,gBAAgB,iBAAiB;AAAA,IACjC,eAAe,iBAAiB;AAAA,IAChC,cAAc,iBAAiB;AAAA,IAC/B,cAAc,iBAAiB;AAAA,IAC/B,cAAc,iBAAiB;AAAA,EACjC;AACA,QAAM,SAAS,CAAC;AAChB,aAAW,WAAW,yBAAyB;AAC7C,QAAI,WAAW,cAAc;AAC3B,aAAO,KAAK,aAAa,QAAQ;AAAA,IACnC,OAAO;AACL;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,cAAc,mBAAmB;AACxC,MAAI,kBAAkB,WAAW,GAAG;AAClC,WAAO;AAAA,EACT;AACA,MAAI,kBAAkB,OAAO,iBAAiB,gBAAgB;AAC5D,WAAO,kBAAkB,SAAS;AAAA,EACpC;AACA,SAAO,kBAAkB;AAC3B;AACA,SAAS,0BAA0B,mBAAmB,YAAY;AAChE,MAAI,qBAAqB,QAAQ,cAAc,MAAM;AACnD;AAAA,EACF;AACA,QAAM,eAAe,kBAAkB;AACvC,QAAM,cAAc,WAAW;AAC/B,MAAI,gBAAgB,aAAa;AAC/B,UAAM,IAAI,MAAM,sBAAsB,wDAAwD,qDAAqD,wEAAwE,cAAc;AAAA,EAC3O;AACA,WAAS,IAAI,GAAG,IAAI,KAAK,IAAI,cAAc,cAAc,CAAC,GAAG,EAAE,GAAG;AAChE,UAAM,aAAa,kBAAkB;AACrC,UAAM,WAAW,WAAW,IAAI;AAChC,QAAI,cAAc,KAAK,YAAY,KAAK,eAAe,KAAK,eAAe,UAAU;AACnF,YAAM,IAAI,MAAM,sBAAsB,+DAA+D,mDAAmD,IAAI,kBAAkB,aAAa,uDAAuD,IAAI,kBAAkB,aAAa,UAAU;AAAA,IACjS;AAAA,EACF;AACF;AAGA,IAAI,wBAAwB;AAC5B,SAAS,yBAAyB,QAAQ;AACxC,MAAI,UAAU,uBAAuB;AACnC,WAAO;AAAA,EACT;AACA,SAAO,eAAe,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,CAAC,CAAC;AAC7D;AAGA,SAAS,eAAe,QAAQ,aAAa,YAAY;AACvD,QAAM,UAAU,cAAc,OAAO,WAAW,WAAW,SAAS,OAAO;AAC3E,QAAM,UAAU,eAAe,OAAO,WAAW,WAAW,SAAS,OAAO;AAC5E,SAAO,CAAC,SAAS,OAAO;AAC1B;AAGA,SAAS,YAAY,YAAY,YAAY,OAAO,eAAe,MAAM;AACvE,MAAI,WAAW,CAAC;AAChB,MAAI,cAAc;AAChB,eAAW,SAAS,OAAO,WAAW,MAAM,CAAC,CAAC;AAC9C,aAAS,KAAK,WAAW,KAAK,KAAK;AACnC,eAAW,SAAS,OAAO,WAAW,MAAM,CAAC,CAAC;AAAA,EAChD,OAAO;AACL,eAAW,SAAS,OAAO,WAAW,EAAE;AACxC,UAAM,gBAAgB,WAAW;AACjC,aAAS,IAAI,GAAG,IAAI,eAAe,EAAE,GAAG;AACtC,iBAAW,SAAS,OAAO,CAAC,WAAW,IAAI,KAAK,WAAW,IAAI,WAAW,EAAE,CAAC;AAAA,IAC/E;AACA,eAAW,SAAS,OAAO,WAAW,MAAM,gBAAgB,CAAC,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,SAAS,YAAY,cAAc,gBAAgB,eAAe,MAAM;AACtE,QAAM,WAAW,CAAC;AAClB,MAAI,cAAc;AAChB,aAAS,KAAK,cAAc;AAC5B,aAAS,IAAI,iBAAiB,GAAG,IAAI,cAAc,EAAE,GAAG;AACtD,UAAI,KAAK,IAAI,gBAAgB;AAC3B,iBAAS,KAAK,CAAC;AACf,iBAAS,KAAK,KAAK,iBAAiB,EAAE;AAAA,MACxC,OAAO;AACL,iBAAS,KAAK,CAAC;AAAA,MACjB;AAAA,IACF;AAAA,EACF,OAAO;AACL,UAAM,sBAAsB,CAAC;AAC7B,UAAM,qBAAqB,CAAC;AAC5B,aAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,UAAI,KAAK,iBAAiB,IAAI,KAAK,IAAI,MAAM,GAAG;AAC9C,2BAAmB,KAAK,CAAC;AAAA,MAC3B,OAAO;AACL,4BAAoB,KAAK,CAAC;AAAA,MAC5B;AAAA,IACF;AACA,aAAS,KAAK,GAAG,mBAAmB;AACpC,aAAS,KAAK,CAAC;AACf,aAAS,KAAK,GAAG,kBAAkB;AAAA,EACrC;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,YAAY,YAAY,OAAO,eAAe,MAAM;AAC/E,QAAM,mBAAmB,CAAC;AAC1B,MAAI,cAAc;AAChB,qBAAiB,KAAK,WAAW,KAAK,KAAK;AAAA,EAC7C,OAAO;AACL,qBAAiB,KAAK,WAAW,KAAK,KAAK;AAAA,EAC7C;AACA,WAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,QAAI,KAAK,WAAW,QAAQ;AAC1B,UAAI,cAAc;AAChB,yBAAiB,KAAK,WAAW,IAAI,KAAK,WAAW,EAAE;AAAA,MACzD,OAAO;AACL,yBAAiB,KAAK,WAAW,KAAK,WAAW,IAAI,EAAE;AAAA,MACzD;AAAA,IACF,OAAO;AACL,uBAAiB,KAAK,WAAW,EAAE;AAAA,IACrC;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,OAAO,YAAY;AAC9C,QAAM,mBAAmB,CAAC,CAAC;AAC3B,WAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,qBAAiB,KAAK,MAAM,GAAG,EAAE;AAAA,EACnC;AACA,SAAO;AACT;AACA,SAAS,aAAa,gBAAgB,OAAO,YAAY;AACvD,QAAM,YAAY,eAAe,MAAM,GAAG,CAAC;AAC3C,WAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,cAAU,KAAK,eAAe,IAAI,KAAK,MAAM,GAAG,KAAK,MAAM,GAAG,EAAE;AAAA,EAClE;AACA,SAAO;AACT;AAGA,IAAI,kBAAkB;AACtB,IAAI,aAAa;AAGjB,IAAI,QAAQ;AACZ,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,SAAS;AACb,IAAI,SAAS;AAGb,SAAS,uBAAuB,OAAO,OAAO;AAC5C,MAAI,MAAM,WAAW,MAAM,QAAQ;AACjC,UAAM,IAAI,MAAM,gEAAgE,MAAM,iBAAiB,MAAM,SAAS;AAAA,EACxH;AACA,QAAM,SAAS,IAAI,aAAa,MAAM,SAAS,CAAC;AAChD,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,GAAG;AACzC,WAAO,KAAK,MAAM,IAAI;AACtB,WAAO,IAAI,KAAK,MAAM,IAAI;AAAA,EAC5B;AACA,SAAO;AACT;AACA,SAAS,uBAAuB,UAAU;AACxC,QAAM,QAAQ,IAAI,aAAa,SAAS,SAAS,CAAC;AAClD,QAAM,QAAQ,IAAI,aAAa,SAAS,SAAS,CAAC;AAClD,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK,GAAG;AAC3C,UAAM,IAAI,KAAK,SAAS;AACxB,UAAM,IAAI,KAAK,SAAS,IAAI;AAAA,EAC9B;AACA,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AACA,SAAS,qBAAqB,UAAU;AACtC,QAAM,MAAM,KAAK,KAAK,SAAS,SAAS,CAAC;AACzC,QAAM,QAAQ,IAAI,aAAa,GAAG;AAClC,QAAM,QAAQ,IAAI,aAAa,GAAG;AAClC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK,GAAG;AAC3C,UAAM,KAAK,MAAM,IAAI,CAAC,KAAK,SAAS;AACpC,UAAM,KAAK,MAAM,IAAI,CAAC,KAAK,SAAS,IAAI;AAAA,EAC1C;AACA,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AACA,SAAS,oBAAoB,UAAU;AACrC,QAAM,MAAM,KAAK,MAAM,SAAS,SAAS,CAAC;AAC1C,QAAM,QAAQ,IAAI,aAAa,GAAG;AAClC,QAAM,QAAQ,IAAI,aAAa,GAAG;AAClC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK,GAAG;AAC3C,UAAM,KAAK,MAAM,IAAI,CAAC,KAAK,SAAS;AACpC,UAAM,KAAK,MAAM,IAAI,CAAC,KAAK,SAAS,IAAI;AAAA,EAC1C;AACA,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AACA,SAAS,oBAAoB,UAAU,OAAO;AAC5C,QAAM,QAAQ,SAAS,QAAQ;AAC/B,QAAM,QAAQ,SAAS,QAAQ,IAAI;AACnC,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AACA,SAAS,mBAAmB,MAAM,OAAO,OAAO,OAAO;AACrD,OAAK,QAAQ,KAAK;AAClB,OAAK,QAAQ,IAAI,KAAK;AACxB;AACA,SAAS,UAAU,GAAG,SAAS;AAC7B,QAAM,QAAQ,IAAI,aAAa,IAAI,CAAC;AACpC,QAAM,QAAQ,IAAI,aAAa,IAAI,CAAC;AACpC,WAAS,IAAI,GAAG,IAAI,KAAK,KAAK,IAAI,CAAC,GAAG,KAAK;AACzC,UAAM,KAAK,UAAU,IAAI,MAAM,KAAK,MAAM,IAAI;AAC9C,UAAM,KAAK,KAAK,IAAI,CAAC;AACrB,UAAM,KAAK,KAAK,IAAI,CAAC;AAAA,EACvB;AACA,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AACA,SAAS,SAAS,GAAG,GAAG,SAAS;AAC/B,QAAM,KAAK,UAAU,IAAI,MAAM,KAAK,MAAM,IAAI;AAC9C,QAAM,QAAQ,KAAK,IAAI,CAAC;AACxB,QAAM,QAAQ,KAAK,IAAI,CAAC;AACxB,SAAO,EAAE,MAAM,OAAO,MAAM,MAAM;AACpC;AAGA,IAAI,QAAQ;AACZ,IAAI,cAAc;AAClB,IAAI,QAAQ;AACZ,IAAI,WAAW;AACf,SAAS,qBAAqB,UAAU,YAAY;AAClD,aAAW,SAAS,QAAQ,OAAO,EAAE;AACrC,QAAM,aAAa,SAAS,SAAS,SAAS,QAAQ,aAAa,EAAE,EAAE,UAAU,MAAM;AACvF,MAAI,YAAY,GAAG;AACjB,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE,WAAW,YAAY,GAAG;AACxB,UAAM,IAAI,MAAM,6CAA6C,UAAU;AAAA,EACzE;AACA,QAAM,CAAC,aAAa,YAAY,IAAI,SAAS,MAAM,KAAK;AACxD,SAAO,YAAY,QAAQ,QAAQ,MAAM,IAAI,MAAM,2BAA2B,kCAAkC;AAChH,QAAM,aAAa,YAAY,MAAM,KAAK;AAC1C,QAAM,YAAY,WAAW;AAC7B,MAAI,eAAe,WAAW;AAC5B,UAAM,IAAI,MAAM,YAAY,qCAAqC,YAAY;AAAA,EAC/E;AACA,MAAI,YAAY,GAAG;AACjB,UAAM,IAAI,MAAM,+DAA+D;AAAA,EACjF;AACA,QAAM,UAAU,CAAC;AACjB,WAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,UAAM,UAAU,aAAa;AAC7B,QAAI,CAAC,WAAW,KAAK,CAAC,cAAc,UAAU,QAAQ,OAAO,MAAM,EAAE,GAAG;AACtE,YAAM,IAAI,MAAM,uCAAuC,8CAA8C;AAAA,IACvG;AACA,QAAI,QAAQ,QAAQ,OAAO,MAAM,IAAI;AACnC,cAAQ,KAAK,OAAO;AAAA,IACtB;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,UAAM,UAAU,YAAY;AAC5B,QAAI,QAAQ,QAAQ,OAAO,MAAM,MAAM,YAAY,OAAO;AACxD,cAAQ,KAAK,OAAO;AAAA,IACtB;AAAA,EACF;AACA,QAAM,SAAS,IAAI,MAAM,WAAW,MAAM;AAC1C,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,QAAI,IAAI,IAAI,WAAW,GAAG,MAAM,EAAE,CAAC,EAAE,SAAS,WAAW,GAAG,QAAQ;AAClE,YAAM,IAAI,MAAM,2CAA2C,WAAW,iEAAiE;AAAA,IACzI;AACA,WAAO,KAAK,CAAC;AACb,aAAS,IAAI,GAAG,IAAI,WAAW,GAAG,QAAQ,EAAE,GAAG;AAC7C,aAAO,GAAG,KAAK,QAAQ,QAAQ,WAAW,GAAG,EAAE,CAAC;AAAA,IAClD;AAAA,EACF;AACA,QAAM,UAAU,QAAQ;AACxB,QAAM,aAAa,aAAa;AAChC,QAAM,aAAa,CAAC;AACpB,WAAS,IAAI,YAAY,IAAI,SAAS,EAAE,GAAG;AACzC,eAAW,KAAK,CAAC;AAAA,EACnB;AACA,SAAO,EAAE,SAAS,YAAY,OAAO;AACvC;AACA,SAAS,qBAAqB,OAAO,QAAQ;AAC3C,MAAI,qBAAqB,IAAI,MAAM,KAAK;AACxC,qBAAmB,KAAK,EAAE;AAC1B,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,uBAAmB,OAAO,MAAM;AAAA,EAClC;AACA,QAAM,cAAc,CAAC;AACrB,WAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,QAAI,mBAAmB,OAAO,IAAI;AAChC,kBAAY,KAAK,CAAC;AAAA,IACpB;AAAA,EACF;AACA,uBAAqB,mBAAmB,OAAO,CAAC,MAAM,MAAM,EAAE;AAC9D,SAAO,EAAE,oBAAoB,YAAY,YAAY;AACvD;AACA,SAAS,oBAAoB,OAAO,QAAQ,SAAS;AACnD,QAAM,WAAW,IAAI,MAAM,KAAK;AAChC,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAM,QAAQ,QAAQ,GAAG;AACzB,aAAS,IAAI,GAAG,IAAI,OAAO,GAAG,QAAQ,EAAE,GAAG;AACzC,UAAI,SAAS,OAAO,GAAG,QAAQ,QAAQ;AACrC,iBAAS,OAAO,GAAG,MAAM,MAAM;AAAA,MACjC,OAAO;AACL,eAAO,SAAS,OAAO,GAAG,QAAQ,MAAM,IAAI,MAAM,sBAAsB,SAAS,OAAO,GAAG,eAAe,qBAAqB,KAAK,UAAU,KAAK,wBAAwB,MAAM,IAAI;AAAA,MACvL;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,qBAAqB,YAAY,QAAQ;AAChD,QAAM,OAAO;AACb,QAAM,QAAQ,CAAC;AACf,MAAI,SAAS;AACb,MAAI,WAAW,WAAW,GAAG;AAC3B,SAAK,KAAK,EAAE;AAAA,EACd;AACA,WAAS,WAAW,SAAS;AAC7B,WAAS,IAAI,GAAG,IAAI,QAAQ,EAAE,GAAG;AAC/B,UAAM,KAAK,CAAC,CAAC;AAAA,EACf;AACA,QAAM,sBAAsB,CAAC;AAC7B,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,YAAY,KAAK;AACvB,UAAM,cAAc,iBAAiB,QAAQ,SAAS;AACtD,eAAW,aAAa,aAAa;AACnC,UAAI,oBAAoB,QAAQ,SAAS,MAAM,IAAI;AACjD,cAAM,GAAG,KAAK,SAAS;AACvB,4BAAoB,KAAK,SAAS;AAAA,MACpC;AAAA,IACF;AAAA,EACF;AACA,SAAO,EAAE,MAAM,MAAM;AACvB;AACA,SAAS,sBAAsB,MAAM;AACnC,SAAO,KAAK,MAAM,CAAC,KAAK,UAAU,QAAQ,KAAK;AACjD;AACA,SAAS,iBAAiB,QAAQ,KAAK;AACrC,QAAM,cAAc,CAAC;AACrB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,QAAI,OAAO,GAAG,WAAW,KAAK,OAAO,GAAG,QAAQ,GAAG,MAAM,MAAM,QAAQ,IAAI;AACzE,kBAAY,KAAK,CAAC;AAAA,IACpB;AAAA,EACF;AACA,SAAO;AACT;AAGA,SAAS,iBAAiB,GAAG,iBAAiB,OAAO,GAAG;AACtD,MAAI,aAAa,CAAC;AAClB,MAAI,OAAO,oBAAoB,UAAU;AACvC,WAAO,EAAE,MAAM,QAAQ,oBAAoB,GAAG,MAAM,+CAA+C;AACnG,iBAAa,IAAI,MAAM,eAAe,EAAE,KAAK,EAAE,MAAM,QAAQ,eAAe;AAAA,EAC9E,OAAO;AACL,UAAM,YAAY,gBAAgB,OAAO,CAAC,QAAQ,UAAU;AAC1D,UAAI,UAAU,IAAI;AAChB,kBAAU;AAAA,MACZ;AACA,aAAO;AAAA,IACT,GAAG,CAAC;AACJ,WAAO,aAAa,GAAG,MAAM,yDAAyD;AACtF,UAAM,WAAW,gBAAgB,QAAQ,EAAE;AAC3C,QAAI,aAAa,IAAI;AACnB,YAAM,QAAQ,gBAAgB,OAAO,CAAC,GAAG,MAAM,IAAI,IAAI,IAAI,IAAI,CAAC;AAChE,sBAAgB,YAAY,EAAE,MAAM,QAAQ;AAAA,IAC9C;AACA,WAAO,EAAE,MAAM,UAAU,gBAAgB,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC,GAAG,MAAM,6DAA6D;AACrI,iBAAa;AAAA,EACf;AACA,SAAO;AACT;AAGA,SAAS,gDAAgD,eAAe;AACtE,SAAO;AAAA,uBACc;AACvB;AACA,SAAS,gDAAgD,OAAO,OAAO;AACrE,SAAO,WAAW,yBAAyB;AAC7C;AACA,SAAS,kDAAkD,OAAO,OAAO,OAAO;AAC9E,SAAO,WAAW,yBAAyB,YAAY;AACzD;AAGA,SAAS,yDAAyD,MAAM,MAAM;AAC5E,SAAO,iDAAiD,YAAY;AACtE;AACA,SAAS,8CAA8C,KAAK,OAAO;AACjE,SAAO,QAAQ,iCAAiC;AAClD;AACA,SAAS,uDAAuD;AAC9D,SAAO;AACT;AACA,SAAS,gDAAgD,YAAY,aAAa;AAChF,QAAM,YAAY,cAAc,UAAU;AAC1C,QAAM,aAAa,cAAc,WAAW;AAC5C,SAAO,2CAA2C;AAAA,iEACa,0BAA0B,2BAA2B;AACtH;AACA,SAAS,gDAAgD,YAAY,aAAa;AAChF,QAAM,YAAY,cAAc,UAAU;AAC1C,QAAM,aAAa,cAAc,WAAW;AAC5C,SAAO,qCAAqC,uDAAuD,0BAA0B,0BAA0B;AACzJ;AAGA,SAAS,0DAA0D;AACjE,SAAO;AACT;AACA,SAAS,+DAA+D;AACtE,SAAO;AACT;AACA,SAAS,yDAAyD,WAAW,YAAY;AACvF,SAAO,cAAc,8BAA8B;AACrD;AACA,SAAS,uDAAuD,OAAO,YAAY,WAAW;AAC5F,SAAO,gBAAgB,aAAa,+BAA+B;AACrE;AAGA,IAAI,uBAAuB,CAAC;AAC5BA,UAAS,sBAAsB;AAAA,EAC7B,0BAA0B,MAAM;AAAA,EAChC,iBAAiB,MAAM;AAAA,EACvB,+BAA+B,MAAM;AACvC,CAAC;AACD,SAAS,8BAA8B,QAAQ,aAAa;AAC1D,MAAI,OAAO;AACX,MAAI;AACJ,MAAI,UAAU,uBAAuB;AACnC,UAAM;AACN,WAAO;AAAA,EACT,OAAO;AACL,UAAM,eAAe,QAAQ,KAAK,MAAM,KAAK,KAAK,MAAM,CAAC,CAAC;AAAA,EAC5D;AACA,SAAO,CAAC,MAAM;AACZ,QAAI,MAAM,eAAe,QAAQ,QAAQ;AACvC,aAAO;AAAA,IACT,OAAO;AACL,YAAM,eAAe,QAAQ,MAAM,CAAC;AAAA,IACtC;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,QAAQ,MAAM,aAAa;AACnD,QAAM,WAAW,CAAC;AAClB,QAAM,OAAO,OAAO;AACpB,WAAS,MAAM,GAAG,MAAM,MAAM,OAAO;AACnC,QAAI,QAAQ,MAAM;AAChB,eAAS,KAAK,OAAO,IAAI;AAAA,IAC3B,OAAO;AACL,eAAS,KAAK,WAAW;AAAA,IAC3B;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,yBAAyB,GAAG,SAAS,MAAM,WAAW;AAC7D,QAAM,cAAc,QAAQ,MAAM;AAClC,QAAM,QAAQ,EAAE,MAAM;AACtB,MAAI,cAAc,GAAG;AACnB,QAAI,YAAY,CAAC,eAAe,YAAY,aAAa;AACvD,YAAM,IAAI,MAAM,sCAAsC,gBAAgB,yBAAyB,WAAW;AAAA,IAC5G;AAAA,EACF;AACA,MAAI,YAAY,GAAG;AACjB,iBAAa;AAAA,EACf;AACA,MAAI,YAAY,OAAO;AACrB,UAAM,IAAI,MAAM,cAAc;AAAA,MAC5B,SAAS;AAAA,EACb;AACA,MAAI,OAAO,WAAW;AACpB,UAAM,IAAI,MAAM,cAAc,kDAAkD,QAAQ;AAAA,EAC1F;AACA,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,QAAI,EAAE,MAAM,OAAO,QAAQ,MAAM,IAAI;AACnC,YAAM,IAAI,MAAM,WAAW,OAAO,EAAE,MAAM,uCAAuC,OAAO,QAAQ,MAAM,KAAK;AAAA,IAC7G;AAAA,EACF;AACA,QAAM,UAAU,EAAE,MAAM;AACxB,QAAM,cAAc,CAAC;AACrB,MAAI,YAAY;AAChB,MAAI,YAAY;AAChB,MAAI,YAAY;AAChB,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,gBAAY,KAAK,EAAE,MAAM,EAAE;AAC3B,iBAAa,EAAE,MAAM;AAAA,EACvB;AACA,WAAS,IAAI,WAAW,IAAI,MAAM,KAAK;AACrC,gBAAY,KAAK,EAAE,MAAM,EAAE;AAC3B,iBAAa,EAAE,MAAM;AAAA,EACvB;AACA,WAAS,IAAI,WAAW,IAAI,aAAa,KAAK;AAC5C,gBAAY,KAAK,QAAQ,MAAM,EAAE;AAAA,EACnC;AACA,WAAS,IAAI,OAAO,GAAG,IAAI,OAAO,KAAK;AACrC,gBAAY,KAAK,EAAE,MAAM,EAAE;AAC3B,iBAAa,EAAE,MAAM;AAAA,EACvB;AACA,SAAO,EAAE,WAAW,WAAW,WAAW,SAAS,YAAY;AACjE;AAGA,SAAS,uBAAuB,MAAM;AACpC,MAAI;AACF,WAAO,KAAK,IAAI,CAAC,QAAQ,aAAa,GAAG,CAAC;AAAA,EAC5C,SAAS,KAAP;AACA,UAAM,IAAI,MAAM,4DAA4D,KAAK;AAAA,EACnF;AACF;AACA,SAAS,uBAAuB,SAAS;AACvC,SAAO,QAAQ,IAAI,CAAC,MAAM,aAAa,CAAC,CAAC;AAC3C;AAGA,IAAI,uBAAuB,CAAC;AAC5BA,UAAS,sBAAsB;AAAA,EAC7B,yBAAyB,MAAM;AAAA,EAC/B,yBAAyB,MAAM;AAAA,EAC/B,yBAAyB,MAAM;AAAA,EAC/B,WAAW,MAAM;AACnB,CAAC;AAGD,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,KAAK,GAAG,SAAS,GAAG,EAAE,CAAC,EAAE;AAAA,EAC1D;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO;AAAA,MACL,GAAG,MAAM;AACP,cAAM,IAAI,OAAO,KAAK,GAAG,SAAS,CAAC;AACnC,cAAM,IAAI,KAAK,IAAI,OAAO,CAAC,GAAG,CAAC,CAAC;AAChC,eAAO,IAAI,IAAI,IAAI,CAAC,CAAC;AAAA,MACvB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO;AAAA,MACL,GAAG,MAAM;AACP,cAAM,IAAI,KAAK,IAAI,OAAO,KAAK,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC;AACjD,eAAO,IAAI,IAAI,CAAC;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM;AACV,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,EAAE,KAAK;AAAA,IAC7B;AACA,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM;AACV,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,EAAE,KAAK;AAAA,IAC7B;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,eAAe;AAAA,EACf,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,OAAO,CAAC;AACd,UAAM,QAAQ,CAAC,GAAG,MAAM;AACtB,WAAK,KAAK,MAAM,GAAG,MAAM;AAAA,IAC3B,CAAC;AACD,WAAO;AAAA,EACT;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,UAAU,CAAC,EAAE;AAAA,EACjC;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,UAAU,CAAC,EAAE;AAAA,EACjC;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,IAAI,OAAO,CAAC,GAAG,OAAO,KAAK,GAAG,SAAS,CAAC,CAAC,CAAC,CAAC,EAAE;AAAA,EAC9E;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO;AAAA,MACL,GAAG,MAAM;AACP,cAAM,IAAI,KAAK,KAAK,OAAO,CAAC,GAAG,OAAO,KAAK,GAAG,SAAS,CAAC,CAAC,CAAC;AAC1D,eAAO,IAAI,IAAI,CAAC;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,YAAM,IAAI,KAAK,OAAO,CAAC,GAAG,OAAO,CAAC,CAAC;AACnC,UAAI,MAAM,IAAI,IAAI,IAAI,GAAG,CAAC,CAAC;AAC3B,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,EAAE,KAAK;AAAA,IAC7B;AACA,UAAM,OAAO,MAAM;AACjB,YAAM,IAAI,KAAK,OAAO,CAAC,GAAG,OAAO,CAAC,CAAC;AACnC,UAAI,MAAM,IAAI,IAAI,IAAI,IAAI,GAAG,CAAC,CAAC,CAAC;AAChC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,EAAE,KAAK;AAAA,IAC7B;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,OAAO,KAAK,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC,EAAE;AAAA,EACjE;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,OAAO,CAAC,GAAG,OAAO,KAAK,GAAG,SAAS,CAAC,CAAC,CAAC,EAAE;AAAA,EACxE;AACF;AAGA,SAAS,eAAe,IAAI,QAAQ,YAAY,SAAS,MAAM,iBAAiB;AAC9E,QAAM,MAAM,gBAAgB,IAAI,MAAM,eAAe;AACrD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,eAAe;AAC/D,MAAI,OAAO;AACX,MAAI,UAAU;AACd,MAAI,eAAe;AACnB,MAAI,OAAO,SAAS,GAAG;AACrB,mBAAe;AACf,WAAO,QAAQ,KAAK,CAAC,GAAG,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAC/E,cAAU,QAAQ,QAAQ;AAAA,MACxB;AAAA,MACA,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,IACf,CAAC;AAAA,EACH;AACA,SAAO,KAAK,SAAS,GAAG,MAAM,0DAA0D,KAAK,OAAO;AACpG,SAAO,QAAQ,SAAS,GAAG,MAAM,6DAA6D,QAAQ,OAAO;AAC7G,4BAA0B,iBAAiB,MAAM,eAAe;AAChE,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO,QAAQ;AAC1C,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB;AAChE,QAAM,MAAM,OAAO,UAAU,eAAe,QAAQ,KAAK;AACzD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,WAAO;AAAA,MACL,GAAG,MAAM,cAAc,IAAI,GAAG,YAAY,SAAS,MAAM,eAAe;AAAA,IAC1E;AAAA,EACF;AACF;AAGA,SAAS,aAAa,IAAI,QAAQ,YAAY,SAAS,MAAM;AAC3D,QAAM,MAAM,gBAAgB,IAAI,MAAM,aAAa;AACnD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,aAAa;AAC7D,SAAO,OAAO,SAAS,IAAI,MAAM,MAAM,kBAAkB,OAAO,oCAAoC,IAAI,OAAO;AAC/G,MAAI,UAAU;AACd,MAAI,OAAO;AACX,MAAI,eAAe;AACnB,MAAI,OAAO,SAAS,GAAG;AACrB,mBAAe;AACf,cAAU,QAAQ,QAAQ,CAAC,GAAG,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE,CAAC;AAChF,WAAO,QAAQ,KAAK,CAAC,GAAG,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EACnE;AACA,SAAO,KAAK,SAAS,GAAG,MAAM,wDAAwD,KAAK,OAAO;AAClG,SAAO,QAAQ,SAAS,GAAG,MAAM,2DAA2D,QAAQ,OAAO;AAC3G,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO,QAAQ;AAC1C,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,KAAK;AAC/C,QAAM,MAAM,OAAO,UAAU,aAAa,QAAQ,KAAK;AACvD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,YAAY,SAAS,KAAK,KAAK,IAAI;AAC3C,WAAO,EAAE,GAAG,MAAM,YAAY,IAAI,GAAG,YAAY,SAAS,IAAI,EAAE;AAAA,EAClE;AACF;AAGA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,EAAE,YAAY,WAAW,IAAI;AACnC,QAAI,CAAC,cAAc,CAAC,YAAY;AAC9B,aAAO;AAAA,QACL,GAAG,MAAM,OAAO,IAAI,GAAG,OAAO,IAAI;AAAA,QAClC,GAAG,MAAM,OAAO,GAAG,IAAI,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,WAAW,CAAC,cAAc,YAAY;AACpC,aAAO;AAAA,QACL,GAAG,MAAM,OAAO,IAAI,GAAG,OAAO,KAAK;AAAA,QACnC,GAAG,MAAM,OAAO,IAAI,GAAG,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,WAAW,cAAc,CAAC,YAAY;AACpC,aAAO;AAAA,QACL,GAAG,MAAM,OAAO,GAAG,IAAI,OAAO,IAAI;AAAA,QAClC,GAAG,MAAM,OAAO,GAAG,IAAI,OAAO,KAAK;AAAA,MACrC;AAAA,IACF,OAAO;AACL,aAAO;AAAA,QACL,GAAG,MAAM,OAAO,GAAG,IAAI,MAAM,IAAI;AAAA,QACjC,GAAG,MAAM,OAAO,IAAI,GAAG,MAAM,IAAI;AAAA,MACnC;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,YAAY,MAAM,IAAI;AAC9B,WAAO,EAAE,GAAG,MAAM,eAAe,IAAI,YAAY,KAAK,EAAE;AAAA,EAC1D;AACF;AAGA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,mBAAmB;AACzB,UAAM,aAAa,iBAAiB;AACpC,UAAM,cAAc,iBAAiB;AACrC,UAAM,OAAO,MAAM,KAAK,WAAW;AACnC,aAAS,IAAI,WAAW,SAAS,GAAG,KAAK,GAAG,KAAK;AAC/C,UAAI,WAAW,OAAO,YAAY,IAAI;AACpC,aAAK,KAAK;AAAA,MACZ,WAAW,WAAW,OAAO,GAAG;AAC9B,cAAM,IAAI,MAAM,mBAAmB,uCAAuC,eAAe;AAAA,MAC3F;AAAA,IACF;AACA,UAAM,OAAO,CAAC;AACd,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAI,KAAK,KAAK,GAAG;AACf,aAAK,KAAK,CAAC;AAAA,MACb;AAAA,IACF;AACA,WAAO,EAAE,GAAG,MAAM,KAAK,IAAI,MAAM,IAAI,EAAE;AAAA,EACzC;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,GAAG,MAAM,EAAE;AAAA,EAC/B;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,cAAc,aAAa,IAAI;AACvC,WAAO;AAAA,MACL,GAAG,MAAM,MAAM,WAAW,aAAa,GAAG,YAAY,GAAG,UAAU,GAAG,YAAY,CAAC,GAAG,IAAI,UAAU,EAAE,CAAC;AAAA,IACzG;AAAA,EACF;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,cAAc;AAC1B;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,eAAe;AAAA,EACf,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,SAAS,MAAM,IAAI,CAAC,MAAM,EAAE,KAAK;AACvC,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,QAAQ,eAAe,MAAM,MAAM,GAAG,KAAK,EAAE;AACnD,UAAM,aAAa,OAAO,IAAI,CAAC,MAAM,EAAE,MAAM;AAC7C,UAAM,aAAa,MAAM,IAAI,YAAY,KAAK;AAC9C,WAAO,WAAW,IAAI,CAAC,MAAM,MAAM,CAAC;AAAA,EACtC;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,QAAQ;AAAA,EAC5B,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,KAAK,OAAO,IAAI;AACvB,UAAM,EAAE,WAAW,SAAS,KAAK,MAAM,WAAW,IAAI;AACtD,WAAO,kBAAkB,SAAS,GAAG,MAAM,iHAAiH,YAAY;AACxK,WAAO;AAAA,MACL,GAAG,MAAM,oBAAoB,IAAI,OAAO,IAAI,SAAS,SAAS,MAAM,UAAU;AAAA,MAC9E,QAAQ,MAAM,qBAAqB,KAAK,IAAI,QAAQ,OAAO,SAAS,MAAM,UAAU;AAAA,IACtF;AAAA,EACF;AACF;AAGA,IAAI,gCAAgC;AAAA,EAClC,YAAY;AAAA,EACZ,cAAc,CAAC,MAAM,QAAQ;AAAA,EAC7B,UAAU,CAAC,KAAK,OAAO,UAAU;AAC/B,UAAM,CAAC,IAAI,MAAM,IAAI;AACrB,UAAM,EAAE,SAAS,KAAK,MAAM,YAAY,gBAAgB,IAAI;AAC5D,WAAO;AAAA,MACL,IAAI,MAAM,OAAO,KAAK,QAAQ,SAAS,MAAM,YAAY,GAAG,eAAe;AAAA,MAC3E,QAAQ,MAAM,qBAAqB,KAAK,IAAI,OAAO,OAAO,SAAS,MAAM,YAAY,eAAe;AAAA,IACtG;AAAA,EACF;AACF;AAGA,SAAS,sBAAsB,GAAG,IAAI,aAAa,SAAS,MAAM;AAChE,MAAI,MAAM;AACV,MAAI,EAAE,SAAS,GAAG;AAChB,UAAM,QAAQ,GAAG,CAAC,GAAG,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC;AAAA,EACtE;AACA,MAAI,OAAO;AACX,MAAI,KAAK,SAAS,GAAG;AACnB,WAAO,QAAQ,IAAI,CAAC,GAAG,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC;AAAA,EAC5E;AACA,SAAO,IAAI,SAAS,GAAG,MAAM,iEAAiE,IAAI,QAAQ;AAC1G,SAAO,KAAK,SAAS,GAAG,MAAM,8DAA8D,KAAK,QAAQ;AACzG,SAAO,YAAY,WAAW,GAAG,MAAM,mEAAmE,cAAc;AACxH,SAAO,IAAI,MAAM,OAAO,YAAY,IAAI,MAAM,4CAA4C,IAAI,MAAM,yCAAyC,YAAY,KAAK;AAC9J,SAAO,KAAK,MAAM,OAAO,YAAY,IAAI,MAAM,0CAA0C,KAAK,MAAM,2CAA2C,YAAY,MAAM;AACjK,QAAM,SAAS,EAAE,GAAG,KAAK,IAAI,KAAK;AAClC,QAAM,QAAQ,EAAE,SAAS,KAAK,MAAM,YAAY;AAChD,SAAO,OAAO,UAAU,wBAAwB,QAAQ,KAAK;AAC/D;AACA,IAAI,uBAAuB,GAAG,EAAE,sBAAsB,CAAC;AAGvD,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,QAAQ;AAAA,EAC5B,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,WAAW,SAAS,KAAK,KAAK,IAAI;AAC1C,WAAO,kBAAkB,SAAS,GAAG,MAAM,iHAAiH,YAAY;AACxK,UAAM,CAAC,KAAK,OAAO,IAAI;AACvB,WAAO;AAAA,MACL,GAAG,MAAM,oBAAoB,IAAI,OAAO,IAAI,SAAS,SAAS,IAAI;AAAA,MAClE,QAAQ,MAAM,qBAAqB,KAAK,IAAI,QAAQ,OAAO,SAAS,IAAI;AAAA,IAC1E;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC,CAAC,GAAG,EAAE,EAAE;AAAA,EAC1D;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,KAAK,KAAK,GAAG,SAAS,CAAC,GAAG,EAAE,EAAE;AAAA,EACtD;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,WAAO;AAAA,MACL,GAAG,MAAM;AACP,cAAM,cAAc,mBAAmB,CAAC,IAAI,GAAG,EAAE,IAAI;AACrD,YAAI,MAAM,OAAO,IAAI,MAAM,WAAW,CAAC,QAAQ;AAC/C,YAAI,eAAe,MAAM;AACvB,gBAAM,UAAU,KAAK,WAAW;AAAA,QAClC;AACA,eAAO;AAAA,MACT;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kCAAkC;AAAA,EACpC,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,QAAQ;AAAA,EAC5B,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,WAAW,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC3D,UAAM,aAAa,aAAa,OAAO,CAAC,GAAG,CAAC,IAAI;AAChD,WAAO,kBAAkB,UAAU,GAAG,MAAM,mHAAmH,aAAa;AAC5K,UAAM,CAAC,GAAG,MAAM,IAAI;AACpB,WAAO,EAAE,SAAS,GAAG,MAAM,kFAAkF,EAAE,OAAO;AACtH,WAAO,OAAO,SAAS,GAAG,MAAM,mFAAmF,OAAO,OAAO;AACjI,WAAO,EAAE,MAAM,OAAO,OAAO,MAAM,IAAI,MAAM,mEAAmE,EAAE,MAAM,qDAAqD,OAAO,MAAM,KAAK;AAC/L,WAAO,+BAA+B,SAAS,UAAU,GAAG,MAAM,6FAA6F,0BAA0B,cAAc;AACvM,8BAA0B,mBAAmB,MAAM,eAAe;AAClE,WAAO;AAAA,MACL,GAAG,MAAM,mCAAmC,EAAE,OAAO,IAAI,QAAQ,SAAS,MAAM,YAAY,eAAe;AAAA,MAC3G,QAAQ,MAAM,oCAAoC,GAAG,IAAI,OAAO,OAAO,SAAS,MAAM,YAAY,eAAe;AAAA,IACnH;AAAA,EACF;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,QAAQ;AAAA,EAC5B,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,MAAM,IAAI;AACpB,UAAM,cAAc,EAAE,GAAG,QAAQ,GAAG;AACpC,UAAM,eAAe,EAAE,GAAG,QAAQ,GAAG;AACrC,WAAO;AAAA,MACL,GAAG,MAAM,OAAO,UAAU,yBAAyB,aAAa,KAAK;AAAA,MACrE,QAAQ,MAAM,OAAO,UAAU,0BAA0B,cAAc,KAAK;AAAA,IAC9E;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,SAAS,EAAE,IAAI,EAAE;AACvB,WAAO,EAAE,GAAG,MAAM,OAAO,UAAU,SAAS,MAAM,EAAE;AAAA,EACtD;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,IAAI,IAAI,IAAI,IAAI,OAAO,CAAC,CAAC,CAAC,GAAG,IAAI,KAAK,KAAK,KAAK,EAAE,CAAC;AACzD,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,CAAC,EAAE;AAAA,EAC/B;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,CAAC,EAAE;AAAA,EAC/B;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,cAAc,CAAC,OAAO;AAAA,EACtB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,MAAM,IAAI;AACjB,WAAO,EAAE,OAAO,MAAM,QAAQ,IAAI,OAAO,KAAK,EAAE;AAAA,EAClD;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,CAAC,CAAC,EAAE;AAAA,EACpC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,YAAM,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACtC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC/C;AACA,aAAO;AAAA,IACT;AACA,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACpC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC9C;AACA,YAAM,MAAM,OAAO,CAAC;AACpB,aAAO,IAAI,IAAI,KAAK,KAAK,KAAK,SAAS,CAAC,CAAC;AAAA,IAC3C;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,QAAQ,YAAY,OAAO;AAAA,EAC/C,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,gBAAgB,IAAI;AAC5B,UAAM,CAAC,GAAG,OAAO,UAAUI,OAAM,IAAI;AACrC,UAAM,aAAaA,WAAU,OAAO,OAAO,CAAC,IAAIA;AAChD,UAAM,gBAAgB,iBAAiB,MAAM,OAAO,EAAE,KAAK;AAC3D,UAAM,YAAY,CAAC;AACnB,QAAI,MAAM,SAAS,GAAG;AACpB,eAAS,IAAI,GAAG,IAAI,EAAE,MAAM,SAAS,GAAG,EAAE,GAAG;AAC3C,kBAAU,KAAK,EAAE,MAAM,EAAE;AAAA,MAC3B;AACA,gBAAU,KAAK,CAAC;AAAA,IAClB;AACA,UAAM,aAAa,IAAI,GAAG,KAAK;AAC/B,UAAM,oBAAoB,IAAI,IAAI,UAAU;AAC5C,UAAM,sBAAsB,MAAM,KAAK,UAAU,OAAO,eAAe,CAAC,CAAC;AACzE,UAAM,iBAAiB,IAAI,IAAI,IAAI,qBAAqB,mBAAmB,GAAG,mBAAmB,GAAG,OAAO,IAAI,CAAC;AAChH,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM,SAAS,GAAG;AACpB,eAAO,QAAQ,IAAI,IAAI,IAAI,KAAK,QAAQ,qBAAqB,CAAC,GAAG,GAAG,GAAG,MAAM,MAAM,EAAE,CAAC,GAAG,SAAS,CAAC,GAAG,UAAU,GAAG,EAAE,KAAK;AAAA,MAC5H,OAAO;AACL,eAAO,QAAQ,IAAI,IAAI,IAAI,mBAAmB,GAAG,UAAU,GAAG,EAAE,KAAK;AAAA,MACvE;AAAA,IACF;AACA,UAAM,UAAU,MAAM;AACpB,UAAI,UAAU,IAAI,IAAI,qBAAqB,OAAO,EAAE,CAAC,GAAG,iBAAiB;AACzE,UAAI,MAAM,SAAS,GAAG;AACpB,kBAAU,KAAK,SAAS,aAAa;AAAA,MACvC;AACA,aAAO,QAAQ,SAAS,MAAM,KAAK;AAAA,IACrC;AACA,UAAM,cAAc,MAAM;AACxB,UAAI,cAAc,IAAI,IAAI,gBAAgB,UAAU,GAAG,iBAAiB;AACxE,UAAI,MAAM,SAAS,GAAG;AACpB,sBAAc,KAAK,aAAa,aAAa;AAAA,MAC/C;AACA,aAAO,QAAQ,aAAa,MAAM,KAAK;AAAA,IACzC;AACA,UAAM,WAAW,MAAM;AACrB,YAAM,wBAAwB,IAAI,YAAY,mBAAmB;AACjE,UAAI,WAAW,IAAI,IAAI,qBAAqB;AAC5C,UAAI,MAAM,SAAS,GAAG;AACpB,mBAAW,KAAK,UAAU,aAAa;AAAA,MACzC;AACA,aAAO,QAAQ,UAAU,MAAM,KAAK;AAAA,IACtC;AACA,UAAM,YAAY,MAAM;AACtB,UAAI,YAAY;AAChB,UAAI,MAAM,SAAS,GAAG;AACpB,oBAAY,KAAK,WAAW,aAAa;AAAA,MAC3C;AACA,aAAO,QAAQ,WAAW,MAAM,KAAK;AAAA,IACvC;AACA,WAAO;AAAA,MACL,GAAG;AAAA,MACH,MAAM;AAAA,MACN,UAAU;AAAA,MACV,OAAO;AAAA,MACP,QAAQ;AAAA,IACV;AAAA,EACF;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,SAAS;AAAA,EAC7B,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,OAAO,IAAI;AACrB,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AACjD,UAAM,OAAO,MAAM;AACjB,YAAM,cAAc,EAAE;AACtB,YAAM,cAAc,QAAQ;AAC5B,YAAM,aAAa,YAAY,MAAM,GAAG,UAAU;AAClD,YAAM,YAAY,WAAW;AAC7B,YAAM,aAAa,YAAY,MAAM,MAAM,YAAY,MAAM,EAAE,MAAM,CAAC;AACtE,YAAM,YAAY,WAAW;AAC7B,YAAM,mBAAmB,WAAW,GAAG,SAAS;AAChD,YAAM,mBAAmB,WAAW,YAAY,GAAG,YAAY,IAAI,SAAS;AAC5E,YAAM,cAAc,YAAY,CAAC,YAAY,CAAC,WAAW,GAAG,UAAU,CAAC;AACvE,YAAM,SAAS,QAAQ,IAAI,WAAW;AACtC,YAAM,kBAAkB,QAAQ,SAAS,CAAC,WAAW,CAAC;AACtD,YAAM,gBAAgB,YAAY,CAAC,CAAC,SAAS,GAAG,kBAAkB,gBAAgB,CAAC;AACnF,YAAM,kBAAkB,UAAU,QAAQ,aAAa;AACvD,UAAI,aAAa,mBAAmB,iBAAiB,iBAAiB,EAAE,MAAM,WAAW;AACzF,YAAM,sBAAsB,uBAAuB,aAAa;AAChE,mBAAa,UAAU,YAAY,mBAAmB;AACtD,aAAO;AAAA,IACT;AACA,WAAO,EAAE,GAAG,MAAM,SAAS,MAAM,QAAQ;AAAA,EAC3C;AACF;AACA,SAAS,WAAW,OAAO,MAAM;AAC/B,QAAM,SAAS,CAAC;AAChB,WAAS,IAAI,OAAO,IAAI,MAAM,EAAE,GAAG;AACjC,WAAO,KAAK,CAAC;AAAA,EACf;AACA,SAAO;AACT;AACA,SAAS,YAAY,QAAQ;AAC3B,QAAM,SAAS,CAAC;AAChB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,aAAS,IAAI,GAAG,IAAI,OAAO,GAAG,QAAQ,EAAE,GAAG;AACzC,aAAO,KAAK,OAAO,GAAG,EAAE;AAAA,IAC1B;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,yBAAyB;AAAA,EAC3B,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,WAAO,EAAE,GAAG,MAAM,UAAU,CAAC,GAAG,GAAG,MAAM,UAAU,CAAC,EAAE;AAAA,EACxD;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,KAAK,IAAI,SAAS,EAAE;AAAA,EACxC;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,MAAM,IAAI;AAClB,UAAM,OAAO,QAAQ,GAAG,CAAC;AACzB,WAAO,EAAE,GAAG,MAAM,MAAM,MAAM,IAAI,IAAI,IAAI,KAAK,CAAC,EAAE;AAAA,EACpD;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,GAAG,CAAC,CAAC,EAAE;AAAA,EACxC;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC,EAAE;AAAA,EAChD;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,cAAc,CAAC;AAAA,EACf,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,KAAK,IAAI;AAChB,UAAM,EAAE,KAAK,IAAI;AACjB,WAAO;AAAA,MACL,QAAQ,MAAM;AACZ,cAAM,WAAW;AACjB,cAAM,WAAW,IAAI,KAAK;AAC1B,eAAO,IAAI,IAAI,IAAI,KAAK,IAAI,MAAM,QAAQ,GAAG,QAAQ,CAAC;AAAA,MACxD;AAAA,IACF;AAAA,EACF;AACF;AAGA,SAAS,oCAAoC,GAAG,GAAG,IAAI,cAAc,GAAG,OAAO,GAAG,QAAQ,GAAG,OAAO,KAAK;AACvG,QAAM,SAAS,EAAE,GAAG,GAAG,GAAG;AAC1B,QAAM,QAAQ,EAAE,aAAa,MAAM,OAAO,KAAK;AAC/C,SAAO,OAAO,UAAU,SAAS,QAAQ,KAAK;AAChD;AACA,IAAI,qCAAqC,GAAG,EAAE,oCAAoC,CAAC;AAGnF,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,EAAE,aAAa,MAAM,OAAO,KAAK,IAAI;AAC3C,WAAO;AAAA,MACL,GAAG,MAAM,mCAAmC,GAAG,GAAG,IAAI,aAAa,MAAM,OAAO,IAAI;AAAA,IACtF;AAAA,EACF;AACF;AAGA,SAAS,iBAAiB,IAAI,GAAG,OAAO,UAAU;AAChD,MAAI,EAAE,OAAO,MAAM,MAAM;AACvB,QAAI,QAAQ,GAAG,qBAAqB,EAAE,OAAO,QAAQ,CAAC;AAAA,EACxD;AACA,MAAI,GAAG,OAAO,MAAM,MAAM;AACxB,SAAK,QAAQ,IAAI,qBAAqB,GAAG,OAAO,QAAQ,CAAC;AAAA,EAC3D;AACA,SAAO;AAAA,IACL,GAAG,MAAM;AACP,YAAM,KAAK,IAAI,IAAI,KAAK,MAAM,OAAO,CAAC,GAAG,GAAG,KAAK,CAAC;AAClD,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,WAAW;AACjB,UAAM,EAAE,iBAAiB,IAAI;AAC7B,UAAM,IAAI,MAAM;AAChB,UAAM,IAAI,MAAM;AAChB,UAAM,WAAW,eAAe,kBAAkB,EAAE,KAAK;AACzD,UAAM,UAAU,iBAAiB,IAAI,GAAG,GAAG,QAAQ;AACnD,WAAO;AAAA,MACL,GAAG,MAAM;AACP,eAAO,QAAQ,KAAK;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,OAAO,MAAM,IAAI,IAAI,KAAK,aAAa,GAAG,CAAC,GAAG,SAAS,CAAC;AAC9D,UAAM,OAAO,MAAM,IAAI,IAAI,KAAK,KAAK,GAAG,CAAC,GAAG,SAAS,CAAC;AACtD,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,SAAS,eAAe,IAAI,QAAQ,QAAQ,YAAY,SAAS,MAAM,iBAAiB;AACtF,QAAM,MAAM,gBAAgB,IAAI,MAAM,eAAe;AACrD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,eAAe;AAC/D,QAAM,UAAU,gBAAgB,QAAQ,UAAU,eAAe;AACjE,MAAI,OAAO;AACX,MAAI,UAAU;AACd,MAAI,WAAW;AACf,MAAI,eAAe;AACnB,MAAI,OAAO,SAAS,GAAG;AACrB,mBAAe;AACf,WAAO,QAAQ,KAAK,CAAC,GAAG,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAC/E,cAAU,QAAQ,QAAQ;AAAA,MACxB;AAAA,MACA,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,MACb,OAAO,MAAM;AAAA,IACf,CAAC;AACD,eAAW,QAAQ,SAAS;AAAA,MAC1B;AAAA,MACA,QAAQ,MAAM;AAAA,MACd,QAAQ,MAAM;AAAA,MACd,QAAQ,MAAM;AAAA,MACd,QAAQ,MAAM;AAAA,IAChB,CAAC;AAAA,EACH;AACA,SAAO,KAAK,SAAS,GAAG,MAAM,0DAA0D,KAAK,OAAO;AACpG,SAAO,QAAQ,SAAS,GAAG,MAAM,6DAA6D,QAAQ,OAAO;AAC7G,SAAO,SAAS,SAAS,GAAG,MAAM,8DAA8D,SAAS,OAAO;AAChH,4BAA0B,iBAAiB,MAAM,eAAe;AAChE,QAAM,SAAS,EAAE,IAAI,MAAM,OAAO,SAAS,QAAQ,SAAS;AAC5D,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB;AAChE,QAAM,MAAM,OAAO,UAAU,eAAe,QAAQ,KAAK;AACzD,MAAI,cAAc;AAChB,WAAO,QAAQ,KAAK,CAAC,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;AAAA,EAC9E;AACA,SAAO;AACT;AACA,IAAI,gBAAgB,GAAG,EAAE,eAAe,CAAC;AAGzC,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,WAAO;AAAA,MACL,GAAG,MAAM,cAAc,IAAI,GAAG,GAAG,YAAY,SAAS,MAAM,eAAe;AAAA,IAC7E;AAAA,EACF;AACF;AAGA,SAAS,aAAa,IAAI,QAAQ,QAAQ,YAAY,SAAS,MAAM,iBAAiB;AACpF,QAAM,MAAM,gBAAgB,IAAI,MAAM,aAAa;AACnD,QAAM,SAAS,gBAAgB,QAAQ,SAAS,aAAa;AAC7D,QAAM,UAAU,gBAAgB,QAAQ,UAAU,aAAa;AAC/D,SAAO,OAAO,SAAS,IAAI,MAAM,MAAM,kBAAkB,OAAO,oCAAoC,IAAI,OAAO;AAC/G,SAAO,IAAI,SAAS,GAAG,MAAM,wDAAwD,IAAI,OAAO;AAChG,SAAO,OAAO,SAAS,GAAG,MAAM,2DAA2D,OAAO,OAAO;AACzG,4BAA0B,eAAe,MAAM,eAAe;AAC9D,QAAM,SAAS,EAAE,IAAI,KAAK,OAAO,QAAQ,QAAQ,QAAQ;AACzD,QAAM,QAAQ,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB;AAChE,SAAO,OAAO,UAAU,aAAa,QAAQ,KAAK;AACpD;AACA,IAAI,cAAc,GAAG,EAAE,aAAa,CAAC;AAGrC,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,EAAE,YAAY,SAAS,KAAK,KAAK,IAAI;AAC3C,WAAO;AAAA,MACL,GAAG,MAAM,YAAY,IAAI,GAAG,GAAG,YAAY,SAAS,IAAI;AAAA,IAC1D;AAAA,EACF;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,OAAO,eAAe,MAAM,EAAE,KAAK;AACzC,UAAM,SAAS,0BAA0B,EAAE,OAAO,IAAI;AACtD,UAAM,cAAc,OAAO;AAC3B,UAAM,aAAa,cAAc,WAAW;AAC5C,UAAM,OAAO,MAAM;AACjB,YAAM,kBAAkB,EAAE,MAAM,MAAM;AACtC,WAAK,QAAQ,CAAC,UAAU;AACtB,wBAAgB,SAAS;AAAA,MAC3B,CAAC;AACD,YAAM,aAAa,QAAQ,IAAI,eAAe;AAC9C,YAAM,MAAM,IAAI,IAAI,YAAY,MAAM,EAAE,OAAO,SAAS,CAAC,GAAG,UAAU;AACtE,aAAO;AAAA,IACT;AACA,WAAO,EAAE,GAAG,KAAK;AAAA,EACnB;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,WAAW;AACjB,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,eAAe,MAAM,EAAE,KAAK;AAC7C,UAAM,UAAU,iBAAiB,IAAI,GAAG,GAAG,QAAQ;AACnD,WAAO;AAAA,MACL,GAAG,MAAM;AACP,eAAO,QAAQ,KAAK;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,OAAO,MAAM,IAAI,IAAI,KAAK,UAAU,GAAG,CAAC,GAAG,SAAS,CAAC;AAC3D,UAAM,OAAO,MAAM,IAAI,IAAI,KAAK,QAAQ,GAAG,CAAC,GAAG,SAAS,CAAC;AACzD,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,IAAI,MAAM;AAChB,UAAM,EAAE,SAAS,IAAI;AACrB,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE;AACxC,WAAO,EAAE,GAAG,MAAM,MAAM,IAAI,OAAO,EAAE,KAAK,EAAE;AAAA,EAC9C;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE,KAAK;AAAA,MAC9C;AACA,aAAO;AAAA,IACT;AACA,UAAM,OAAO,MAAM;AACjB,YAAM,MAAM,IAAI,IAAI,IAAI,MAAM,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;AACzC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC/C;AACA,aAAO;AAAA,IACT;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,YAAM,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACtC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC/C;AACA,aAAO;AAAA,IACT;AACA,UAAM,OAAO,MAAM;AACjB,YAAM,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACtC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC/C;AACA,aAAO;AAAA,IACT;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,IAAI,EAAE,EAAE;AAAA,EAC5B;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,SAAS;AAAA,EACxB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,UAAU,MAAM;AACtB,WAAO,EAAE,SAAS,MAAM,MAAM,QAAQ,OAAO,SAAS,EAAE;AAAA,EAC1D;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,eAAe;AAAA,EACf,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,aAAa,QAAQ,IAAI,IAAI;AACnC,WAAO,WAAW,IAAI,CAAC,MAAM,MAAM,CAAC;AAAA,EACtC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,IAAI,MAAM;AAChB,UAAM,EAAE,SAAS,IAAI;AACrB,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE;AACxC,WAAO,EAAE,GAAG,MAAM,MAAM,IAAI,OAAO,EAAE,KAAK,EAAE;AAAA,EAC9C;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,GAAG,CAAC,IAAI;AAClB,UAAM,OAAO;AACb,UAAM,OAAO;AACb,UAAM,WAAW,2BAA2B,KAAK,OAAO,KAAK,KAAK;AAClE,UAAM,UAAU,MAAM;AACpB,YAAM,WAAW,KAAK,MAAM,SAAS;AACrC,UAAI,MAAM,IAAI,IAAI,IAAI,UAAU,IAAI,MAAM,IAAI,UAAU,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC;AACpE,YAAM,aAAa,iBAAiB,KAAK,OAAO,QAAQ;AACxD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,KAAK,KAAK;AAAA,IAChC;AACA,UAAM,SAAS,MAAM;AACnB,YAAM,YAAY,QAAQ,MAAM,CAAC;AACjC,YAAM,UAAU,MAAM,WAAW,KAAK,IAAI,GAAG,UAAU,IAAI,CAAC;AAC5D,UAAI,MAAM,IAAI,IAAI,IAAI,GAAG,OAAO,CAAC;AACjC,YAAM,aAAa,iBAAiB,KAAK,OAAO,QAAQ;AACxD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,KAAK,KAAK;AAAA,IAChC;AACA,WAAO,EAAE,GAAG,SAAS,GAAG,OAAO;AAAA,EACjC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,OAAO;AAAA,EAC3B,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,KAAK,IAAI;AACnB,UAAM,OAAO,QAAQ,GAAG,CAAC;AACzB,WAAO;AAAA,MACL,GAAG,MAAM,MAAM,MAAM,IAAI,IAAI,IAAI,KAAK,CAAC;AAAA,MACvC,OAAO,MAAM;AACX,YAAI,MAAM,MAAM,MAAM,UAAU,EAAE,GAAG,IAAI,IAAI,CAAC,CAAC;AAC/C,cAAM,aAAa,iBAAiB,MAAM,OAAO,GAAG,KAAK;AACzD,YAAI,WAAW,SAAS,GAAG;AACzB,gBAAM,KAAK,KAAK,UAAU;AAAA,QAC5B;AACA,eAAO,QAAQ,KAAK,MAAM,KAAK;AAAA,MACjC;AAAA,IACF;AAAA,EACF;AACF;AAGA,SAAS,YAAY,GAAG,IAAI,MAAM;AAChC,QAAM,iBAAiB,EAAE,MAAM,MAAM;AACrC,iBAAe,QAAQ;AACvB,QAAM,aAAa,QAAQ,IAAI,cAAc;AAC7C,QAAM,WAAW,QAAQ,GAAG,MAAM,MAAM,KAAK;AAC7C,QAAM,cAAc,QAAQ,GAAG,MAAM,MAAM,IAAI;AAC/C,QAAM,KAAK,IAAI,UAAU,WAAW;AACpC,SAAO,IAAI,YAAY,EAAE;AAC3B;AACA,SAAS,aAAa,GAAG,IAAI,MAAM;AACjC,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,gBAAgB,QAAQ,KAAK;AACnC,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,UAAU,GAAG,YAAY;AAAA,EACvC;AACA,QAAM,WAAW,UAAU,MAAM,MAAM;AACvC,QAAM,eAAe,SAAS,OAAO,QAAQ,KAAK,QAAQ,KAAK,MAAM;AACrE,QAAM,eAAe,aAAa,OAAO,CAAC,IAAI,MAAM,KAAK,GAAG,CAAC;AAC7D,WAAS,KAAK,YAAY;AAC1B,QAAM,oBAAoB,UAAU,QAAQ,QAAQ;AACpD,MAAI,WAAW,YAAY,mBAAmB,IAAI,aAAa;AAC/D,aAAW,SAAS,QAAQ,UAAU,KAAK;AAC3C,MAAI,gBAAgB,MAAM;AACxB,UAAM,kBAAkB,qBAAqB,uBAAuB,YAAY;AAChF,eAAW,UAAU,UAAU,eAAe;AAAA,EAChD;AACA,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,KAAK,IAAI;AACjB,QAAI,UAAU,CAAC;AACf,QAAI,SAAS,UAAU,SAAS,MAAM;AACpC,gBAAU,EAAE,MAAM,IAAI,CAAC,GAAG,MAAM,CAAC;AAAA,IACnC,WAAW,OAAO,SAAS,UAAU;AACnC,gBAAU,CAAC,IAAI;AAAA,IACjB,OAAO;AACL,gBAAU;AAAA,IACZ;AACA,WAAO,EAAE,GAAG,MAAM,aAAa,GAAG,IAAI,OAAO,EAAE;AAAA,EACjD;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,YAAM,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACtC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,eAAO,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC/C;AACA,aAAO;AAAA,IACT;AACA,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC;AACpC,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,QAAQ,KAAK,KAAK,UAAU,GAAG,EAAE,KAAK;AAAA,MAC9C;AACA,YAAM,MAAM,OAAO,CAAC;AACpB,aAAO,IAAI,IAAI,KAAK,KAAK,KAAK,SAAS,CAAC,CAAC;AAAA,IAC3C;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,OAAO,CAAC,CAAC,CAAC,EAAE;AAAA,EAC5C;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,OAAO,IAAI,UAAU,GAAG,CAAC,GAAG,KAAK,CAAC,CAAC;AACzC,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,MAAM,SAAS,CAAC,EAAE;AAAA,EACnD;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,KAAK,CAAC,GAAG,SAAS,CAAC,EAAE;AAAA,EACtD;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,QAAQ,IAAI,EAAE,KAAK,EAAE;AAAA,EACzC;AACF;AAGA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,cAAc,CAAC,QAAQ;AAAA,EACvB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,MAAM,IAAI;AACjB,UAAM,SAAS,EAAE,IAAI,OAAO;AAC5B,UAAM,YAAY,MAAM,OAAO,UAAU,oBAAoB,QAAQ,KAAK;AAC1E,WAAO,EAAE,QAAQ,UAAU;AAAA,EAC7B;AACF;AAGA,IAAI,kCAAkC;AAAA,EACpC,YAAY;AAAA,EACZ,cAAc,CAAC,QAAQ;AAAA,EACvB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,MAAM,IAAI;AACjB,UAAM,SAAS,EAAE,IAAI,OAAO;AAC5B,UAAM,YAAY,MAAM,OAAO,UAAU,2BAA2B,QAAQ,KAAK;AACjF,WAAO,EAAE,QAAQ,UAAU;AAAA,EAC7B;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,OAAO,eAAe,MAAM,GAAG,KAAK;AAC1C,WAAO,EAAE,GAAG,MAAM,QAAQ,IAAI,IAAI,EAAE;AAAA,EACtC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,IAAI,IAAI,GAAG,GAAG,GAAG,CAAC,CAAC,CAAC,EAAE;AAAA,EACtD;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,WAAW;AAAA,EAC1B,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,SAAS,IAAI;AACpB,WAAO;AAAA,MACL,WAAW,MAAM,KAAK,UAAU,SAAS,GAAG,SAAS;AAAA,MACrD,GAAG,MAAM,IAAI,IAAI,KAAK,WAAW,GAAG,KAAK,CAAC;AAAA,MAC1C,GAAG,MAAM,IAAI,IAAI,KAAK,WAAW,SAAS,GAAG,GAAG,KAAK,CAAC;AAAA,IACxD;AAAA,EACF;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO;AAAA,MACL,GAAG,MAAM;AACP,cAAM,OAAO,QAAQ,GAAG,OAAO,CAAC,CAAC;AACjC,cAAM,cAAc,OAAO,eAAe;AAC1C,cAAMA,UAAS,OAAO,UAAU;AAChC,cAAM,qBAAqB,IAAI,IAAIA,OAAM;AACzC,cAAM,mBAAmB,IAAI,IAAI,IAAI,WAAW,GAAG,IAAI,KAAK,GAAG,SAAS,CAAC,CAAC;AAC1E,eAAO,MAAM,MAAM,oBAAoB,gBAAgB;AAAA,MACzD;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,GAAG,IAAI,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE;AAAA,EACvD;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,KAAK,GAAG,SAAS,CAAC,GAAG,EAAE,EAAE;AAAA,EACrD;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,KAAK,KAAK,GAAG,SAAS,CAAC,GAAG,EAAE,EAAE;AAAA,EACtD;AACF;AAGA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,OAAO,KAAK,IAAI;AACxB,UAAM,aAAa,EAAE;AACrB,UAAM,CAAC,QAAQ,KAAK,IAAI,iBAAiB,GAAG,OAAO,IAAI;AACvD,UAAM,WAAW,CAAC;AAClB,aAAS,IAAI,GAAG,IAAI,GAAG,MAAM,KAAK;AAChC,eAAS,KAAK,CAAC,OAAO,IAAI,WAAW,KAAK,OAAO,KAAK,MAAM,EAAE,CAAC;AAAA,IACjE;AACA,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,QAAQ,EAAE;AAAA,EACtC;AACF;AAGA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,IAAI,IAAI;AAChB,UAAM,WAAW;AACjB,UAAM,WAAW,IAAI,IAAI,CAAC;AAC1B,WAAO;AAAA,MACL,QAAQ,MAAM,IAAI,UAAU,IAAI,KAAK,UAAU,CAAC,GAAG,GAAG,QAAQ,GAAG,CAAC,CAAC;AAAA,IACrE;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,QAAQ,CAAC,CAAC,EAAE;AAAA,EACxC;AACF;AAGA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,YAAY,SAAS,IAAI;AACjC,WAAO,EAAE,GAAG,MAAM,eAAe,IAAI,YAAY,QAAQ,EAAE;AAAA,EAC7D;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,EAAE,KAAK,IAAI;AACjB,WAAO,EAAE,GAAG,MAAM,OAAO,IAAI,IAAI,EAAE;AAAA,EACrC;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,KAAK,KAAK,GAAG,SAAS,CAAC,GAAG,CAAC,CAAC,EAAE;AAAA,EAC9D;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,IAAI,KAAK,GAAG,SAAS,GAAG,CAAC,CAAC,EAAE;AAAA,EACxD;AACF;AAGA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,MAAM,OAAO,CAAC;AACpB,UAAM,OAAO,MAAM,IAAI,IAAI,IAAI,KAAK,IAAI,GAAG,CAAC,CAAC,CAAC;AAC9C,UAAM,OAAO,MAAM,IAAI,IAAI,IAAI,KAAK,IAAI,GAAG,CAAC,CAAC,CAAC;AAC9C,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,KAAK,GAAG;AAAA,EACvB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,GAAG,CAAC,IAAI;AACf,UAAM,WAAW,2BAA2B,EAAE,OAAO,EAAE,KAAK;AAC5D,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM;AACV,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,KAAK,EAAE,KAAK;AAAA,IAC7B;AACA,UAAM,OAAO,MAAM;AACjB,UAAI,MAAM;AACV,YAAM,aAAa,iBAAiB,EAAE,OAAO,QAAQ;AACrD,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,KAAK,KAAK,UAAU;AAAA,MAC5B;AACA,aAAO,QAAQ,IAAI,GAAG,GAAG,EAAE,KAAK;AAAA,IAClC;AACA,WAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,EAC5B;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,kBAAkB,EAAE,MAAM,MAAM;AACtC,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,OAAO,eAAe,MAAM,EAAE,KAAK;AACzC,SAAK,QAAQ,CAAC,UAAU;AACtB,sBAAgB,SAAS;AAAA,IAC3B,CAAC;AACD,UAAM,aAAa,QAAQ,IAAI,eAAe;AAC9C,UAAM,OAAO,IAAI,YAAY,MAAM,EAAE,OAAO,SAAS,CAAC;AACtD,WAAO,EAAE,GAAG,MAAM,KAAK;AAAA,EACzB;AACF;AAGA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,OAAO,IAAI,CAAC,CAAC,CAAC,EAAE;AAAA,EAC5C;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,eAAe,CAAC,IAAI;AAAA,EACpB,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,CAAC,IAAI;AACZ,WAAO,EAAE,GAAG,MAAM,IAAI,IAAI,OAAO,CAAC,GAAG,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE;AAAA,EACvD;AACF;AAGA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,cAAc,CAAC,GAAG;AAAA,EAClB,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,CAAC,CAAC,IAAI;AACZ,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,OAAO,MAAM;AACjB,UAAI,QAAQ,UAAU,CAAC;AACvB,UAAI,EAAE,SAAS,GAAG;AAChB,iBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,kBAAQ,KAAK,OAAO,MAAM,IAAI,CAAC,IAAI,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,QAC/D;AAAA,MACF,WAAW,EAAE,SAAS,GAAG;AACvB,iBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,mBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,oBAAQ,KAAK,OAAO,MAAM,IAAI,CAAC,IAAI,EAAE,MAAM,IAAI,IAAI,EAAE,MAAM,EAAE,GAAG;AAAA,cAC9D,EAAE,MAAM;AAAA,cACR,EAAE,MAAM;AAAA,YACV,CAAC,CAAC;AAAA,UACJ;AAAA,QACF;AAAA,MACF,WAAW,EAAE,SAAS,GAAG;AACvB,iBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,mBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,qBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,sBAAQ,KAAK,OAAO,MAAM,IAAI,CAAC,IAAI,EAAE,MAAM,IAAI,IAAI,EAAE,MAAM,IAAI,IAAI,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,YACvH;AAAA,UACF;AAAA,QACF;AAAA,MACF,WAAW,EAAE,SAAS,GAAG;AACvB,iBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,mBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,qBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,uBAAS,IAAI,GAAG,IAAI,KAAK,IAAI,EAAE,GAAG;AAChC,wBAAQ,KAAK,OAAO,MAAM,IAAI;AAAA,kBAC5B,IAAI,EAAE,MAAM;AAAA,kBACZ,IAAI,EAAE,MAAM;AAAA,kBACZ,IAAI,EAAE,MAAM;AAAA,kBACZ,IAAI,EAAE,MAAM;AAAA,gBACd,GAAG,CAAC,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,cACtD;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF,OAAO;AACL,cAAM,IAAI,MAAM,2DAA2D,EAAE,mBAAmB;AAAA,MAClG;AACA,aAAO;AAAA,IACT;AACA,WAAO,EAAE,GAAG,KAAK;AAAA,EACnB;AACF;AAGA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,iBAAiB;AACvB,UAAM,EAAE,KAAK,IAAI;AACjB,UAAM,WAAW,uBAAuB,IAAI;AAC5C,WAAO,EAAE,GAAG,MAAM,UAAU,IAAI,QAAQ,EAAE;AAAA,EAC5C;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,UAAU,CAAC,IAAI,OAAO,UAAU;AAC9B,UAAM,cAAc;AACpB,UAAM,EAAE,KAAK,IAAI;AACjB,WAAO,EAAE,OAAO,MAAM,MAAM,IAAI,IAAI,EAAE;AAAA,EACxC;AACF;AAGA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,cAAc,CAAC,YAAY;AAAA,EAC3B,UAAU,CAAC,IAAI,UAAU;AACvB,UAAM,CAAC,UAAU,IAAI;AACrB,UAAM,OAAO,MAAM;AACjB,aAAO,oBAAoB,IAAI,UAAU;AAAA,IAC3C;AACA,WAAO,EAAE,GAAG,KAAK;AAAA,EACnB;AACF;AACA,SAAS,oBAAoB,GAAG,SAAS;AACvC,QAAM,qBAAqB,QAAQ,SAAS,UAAU,OAAO,CAAC;AAC9D,QAAM,WAAW,OAAO,GAAG,kBAAkB;AAC7C,MAAI,aAAa,aAAa,SAAS,OAAO,GAAG,OAAO,CAAC;AACzD,QAAM,WAAW,SAAS,OAAO,WAAW;AAC5C,WAAS,IAAI,GAAG,IAAI,UAAU,EAAE,GAAG;AACjC,iBAAa,WAAW,YAAY,IAAI,CAAC;AAAA,EAC3C;AACA,eAAa,WAAW,YAAY,MAAM,SAAS,OAAO,MAAM,CAAC;AACjE,QAAM,YAAY,UAAU,QAAQ;AACpC,SAAO,MAAM,YAAY,UAAU,SAAS;AAC9C;AAGA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,UAAU,CAAC,OAAO;AAChB,WAAO,EAAE,GAAG,MAAM,UAAU,EAAE,EAAE;AAAA,EAClC;AACF;AAGA,IAAI,cAAc;AAAA,EAChB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,WAAW,kBAAkB,aAAa;AACxC,mBAAiB,cAAc;AACjC;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,CAAC;AACrB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM,UAAU;AAC9D,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,MAAM,QAAQ;AACjC;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM,UAAU;AAC9D,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,MAAM,QAAQ;AACjC;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,MAAM;AACvD,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,IAAI;AAC1B;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,MAAM;AACvD,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,IAAI;AAC1B;AAGA,qBAAqB,EAAE,UAAU,WAAW,WAAW;AACrD,OAAK,gBAAgB;AACrB,SAAO,KAAK,SAAS,GAAG,MAAM,qCAAqC;AACnE,SAAO,QAAQ,MAAM,CAAC,CAAC;AACzB;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,OAAO;AACxD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,KAAK;AACzB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,KAAK,IAAI,CAAC;AAClC;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,SAAS;AAC9D,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,MAAM,OAAO,CAAC;AACtC;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,SAAS,OAAO;AACrE,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,MAAM,SAAS,KAAK,CAAC;AAC7C;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,SAAS,OAAO,QAAQ;AAC7E,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,MAAM,SAAS,OAAO,MAAM,CAAC;AACrD;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,SAAS,OAAO,QAAQ,QAAQ;AACrF,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,MAAM,SAAS,OAAO,QAAQ,MAAM,CAAC;AAC7D;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,GAAG;AACnD,OAAK,gBAAgB;AACrB,SAAO,MAAM,MAAM,CAAC;AACtB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,YAAY,SAAS,MAAM,iBAAiB;AAC9F,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,YAAY,SAAS,MAAM,eAAe;AACjE;AAGA,qBAAqB,EAAE,UAAU,iBAAiB,SAAS,YAAY,OAAO;AAC5E,OAAK,gBAAgB;AACrB,SAAO,eAAe,MAAM,YAAY,KAAK;AAC/C;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,OAAO,UAAU,QAAQA,SAAQ,iBAAiB;AACtG,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,OAAO,UAAU,QAAQA,SAAQ,eAAe;AACzE;AAGA,qBAAqB,EAAE,UAAU,cAAc,SAAS,OAAO;AAC7D,OAAK,gBAAgB;AACrB,SAAO,YAAY,MAAM,KAAK;AAChC;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,OAAO;AACtD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,KAAK;AACzB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,cAAc,SAAS,MAAM,MAAM;AAClE,OAAK,gBAAgB;AACrB,SAAO,YAAY,MAAM,MAAM,IAAI;AACrC;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,GAAG,MAAM;AAC1D,OAAK,gBAAgB;AACrB,MAAI,aAAa,QAAQ;AACvB,QAAI,CAAC,CAAC;AAAA,EACR;AACA,SAAO,OAAO,CAAC,MAAM,GAAG,CAAC,GAAG,IAAI;AAClC;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,QAAQ,QAAQ,MAAM,YAAY,UAAU,iBAAiB;AAC9G,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,QAAQ,QAAQ,MAAM,YAAY,UAAU,eAAe;AACjF;AAGA,qBAAqB,EAAE,UAAU,kBAAkB,SAAS,QAAQ,aAAa,SAAS,MAAM,iBAAiB;AAC/G,OAAK,gBAAgB;AACrB,SAAO,gBAAgB,MAAM,QAAQ,aAAa,SAAS,MAAM,eAAe;AAClF;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,QAAQ,SAAS,MAAM,YAAY,WAAW,iBAAiB;AAChH,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,QAAQ,SAAS,MAAM,YAAY,WAAW,eAAe;AACnF;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,MAAM,WAAW,UAAU;AAC7E,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,MAAM,WAAW,QAAQ;AAChD;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,MAAM,WAAW,UAAU;AAC5E,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,MAAM,WAAW,QAAQ;AAC/C;AAGA,qBAAqB,EAAE,UAAU,eAAe,SAAS,WAAW,YAAY;AAC9E,OAAK,gBAAgB;AACrB,SAAO,aAAa,MAAM,WAAW,UAAU;AACjD;AAGA,qBAAqB,EAAE,UAAU,kBAAkB,SAAS,QAAQ,SAAS,MAAM,YAAY,WAAW,iBAAiB;AACzH,OAAK,gBAAgB;AACrB,SAAO,gBAAgB,MAAM,QAAQ,SAAS,MAAM,YAAY,WAAW,eAAe;AAC5F;AAGA,qBAAqB,EAAE,UAAU,aAAa,SAAS,QAAQ,SAAS,MAAM,WAAW,YAAY;AACnG,OAAK,gBAAgB;AACrB,SAAO,WAAW,MAAM,QAAQ,SAAS,MAAM,WAAW,UAAU;AACtE;AAGA,qBAAqB,EAAE,UAAU,WAAW,SAAS,GAAG;AACtD,OAAK,gBAAgB;AACrB,SAAO,SAAS,MAAM,CAAC;AACzB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,CAAC;AACpB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,CAAC;AACpB;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,GAAG;AACnD,OAAK,gBAAgB;AACrB,SAAO,MAAM,MAAM,CAAC;AACtB;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,gBAAgB,SAAS,MAAM,UAAU;AACxE,OAAK,gBAAgB;AACrB,SAAO,cAAc,MAAM,MAAM,QAAQ;AAC3C;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,aAAa,SAAS,MAAM;AAC3D,OAAK,gBAAgB;AACrB,SAAO,WAAW,MAAM,IAAI;AAC9B;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,UAAU,WAAW;AACpD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC,KAAK,IAAI,CAAC;AAClC;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,WAAW,SAAS,GAAG;AACtD,OAAK,gBAAgB;AACrB,SAAO,SAAS,MAAM,CAAC;AACzB;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,SAAS,MAAM;AAChE,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,SAAS,IAAI;AACnC;AAGA,qBAAqB,EAAE,UAAU,eAAe,SAAS,GAAG;AAC1D,OAAK,gBAAgB;AACrB,SAAO,aAAa,MAAM,CAAC;AAC7B;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,GAAG;AACrD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC;AACxB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,WAAW,WAAW;AACrD,OAAK,gBAAgB;AACrB,SAAO,UAAU,IAAI;AACvB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,OAAO,IAAI;AACpB;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,OAAO;AAC3D,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,KAAK;AAC9B;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,GAAG;AACvD,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,CAAC;AAC1B;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,GAAG;AAClD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,CAAC;AACrB;AAGA,qBAAqB,EAAE,UAAU,6BAA6B,SAAS,aAAa,MAAM,OAAO,MAAM;AACrG,OAAK,gBAAgB;AACrB,SAAO,2BAA2B,MAAM,aAAa,MAAM,OAAO,IAAI;AACxE;AAGA,qBAAqB,EAAE,UAAU,aAAa,WAAW;AACvD,OAAK,gBAAgB;AACrB,SAAO,WAAW,IAAI;AACxB;AAGA,qBAAqB,EAAE,UAAU,aAAa,SAAS,MAAM;AAC3D,OAAK,gBAAgB;AACrB,SAAO,WAAW,MAAM,IAAI;AAC9B;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,MAAM,UAAU;AACpE,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,MAAM,QAAQ;AACvC;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,aAAa,SAAS,GAAG;AACxD,OAAK,gBAAgB;AACrB,SAAO,WAAW,MAAM,CAAC;AAC3B;AAGA,qBAAqB,EAAE,UAAU,aAAa,WAAW;AACvD,OAAK,gBAAgB;AACrB,SAAO,WAAW,IAAI;AACxB;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,GAAG;AACvD,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,CAAC;AAC1B;AAGA,qBAAqB,EAAE,UAAU,aAAa,SAAS,GAAG;AACxD,OAAK,gBAAgB;AACrB,SAAO,WAAW,MAAM,CAAC;AAC3B;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,GAAG,YAAY,YAAY;AAC5E,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,GAAG,YAAY,UAAU;AAC/C;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,YAAY,SAAS,MAAM,iBAAiB;AAC9F,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,YAAY,SAAS,MAAM,eAAe;AACjE;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM,UAAU;AAC9D,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,MAAM,QAAQ;AACjC;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,GAAG;AACrD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC;AACxB;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,UAAU;AAC/D,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,MAAM,QAAQ;AAClC;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM,UAAU;AAC9D,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,MAAM,QAAQ;AACjC;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,GAAG;AACrD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,CAAC;AACxB;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,UAAU,MAAM;AACpE,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,UAAU,IAAI;AACvC;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,CAAC;AACpB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,CAAC;AACpB;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,KAAK,MAAM,UAAU;AACpE,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,KAAK,MAAM,QAAQ;AACvC;AAGA,qBAAqB,EAAE,UAAU,WAAW,SAAS,GAAG;AACtD,OAAK,gBAAgB;AACrB,SAAO,SAAS,MAAM,CAAC;AACzB;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,OAAO,UAAU,GAAG,WAAW,GAAG;AACnF,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,OAAO,SAAS,QAAQ;AAC9C;AAGA,qBAAqB,EAAE,UAAU,WAAW,WAAW;AACrD,OAAK,gBAAgB;AACrB,SAAO,SAAS,IAAI;AACtB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,UAAU,eAAe;AACvE,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,UAAU,aAAa;AAC1C;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,aAAa,aAAa,SAAS,cAAc,SAAS,iBAAiB;AAC1H,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,aAAa,aAAa,SAAS,cAAc,SAAS,eAAe;AAC7F;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM;AACpD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,IAAI;AACvB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,OAAO;AACvD,OAAK,gBAAgB;AACrB,SAAO,MAAM,MAAM,KAAK;AAC1B;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM,UAAU;AAC/D,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,MAAM,QAAQ;AAClC;AAGA,qBAAqB,EAAE,UAAU,aAAa,WAAW;AACvD,OAAK,gBAAgB;AACrB,SAAO,WAAW,IAAI;AACxB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,GAAG;AACvD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,EAAE,KAAK;AAC9B;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,OAAO;AACzD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,KAAK;AAC5B;AAGA,qBAAqB,EAAE,UAAU,iBAAiB,SAAS,YAAY,cAAc,kBAAkB;AACrG,OAAK,gBAAgB;AACrB,SAAO,eAAe,MAAM,YAAY,cAAc,gBAAgB;AACxE;AAGA,qBAAqB,EAAE,UAAU,wBAAwB,SAAS,YAAY,cAAc,kBAAkB;AAC5G,OAAK,gBAAgB;AACrB,SAAO,sBAAsB,MAAM,YAAY,cAAc,gBAAgB;AAC/E;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,MAAM;AACxD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,IAAI;AAC3B;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,OAAO,IAAI;AACpB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,kBAAkB,SAAS,iBAAiB,iBAAiB,SAAS,MAAM,UAAU,YAAY;AACjI,OAAK,gBAAgB;AACrB,SAAO,gBAAgB,MAAM,iBAAiB,iBAAiB,SAAS,MAAM,UAAU,UAAU;AACpG;AAGA,qBAAqB,EAAE,UAAU,UAAU,WAAW;AACpD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,IAAI;AACrB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,OAAO,MAAM;AAC7D,OAAK,gBAAgB;AACrB,SAAO,MAAM,MAAM,OAAO,IAAI;AAChC;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,KAAK;AACvD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,GAAG;AAC1B;AAGA,qBAAqB,EAAE,UAAU,WAAW,WAAW;AACrD,OAAK,gBAAgB;AACrB,SAAO,SAAS,IAAI;AACtB;AAGA,qBAAqB,EAAE,UAAU,iBAAiB,SAAS,YAAY,UAAU;AAC/E,OAAK,gBAAgB;AACrB,SAAO,eAAe,MAAM,YAAY,QAAQ;AAClD;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,iBAAiB,MAAM;AACvE,OAAK,gBAAgB;AACrB,SAAO,MAAM,MAAM,iBAAiB,IAAI;AAC1C;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,KAAK,IAAI;AAClB;AAGA,qBAAqB,EAAE,UAAU,SAAS,WAAW;AACnD,OAAK,gBAAgB;AACrB,SAAO,OAAO,IAAI;AACpB;AAGA,qBAAqB,EAAE,UAAU,oBAAoB,SAAS,GAAG;AAC/D,OAAK,gBAAgB;AACrB,SAAO,kBAAkB,MAAM,CAAC;AAClC;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,MAAM;AACxD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,IAAI;AAC3B;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,GAAG,MAAM;AACzD,OAAK,gBAAgB;AACrB,QAAM,qBAAqB,aAAa,SAAS,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC;AACxE,SAAO,MAAM,oBAAoB,IAAI;AACvC;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,OAAO;AACtD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,KAAK;AACzB;AAGA,qBAAqB,EAAE,UAAU,eAAe,SAAS,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,gBAAgB;AAC3I,OAAK,gBAAgB;AACrB,SAAO,aAAa,MAAM,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,cAAc;AAC9G;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,GAAG;AACjD,OAAK,gBAAgB;AACrB,SAAO,IAAI,MAAM,CAAC;AACpB;AAGA,qBAAqB,EAAE,UAAU,MAAM,SAAS,MAAM,UAAU;AAC9D,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,MAAM,QAAQ;AAClC;AAGA,qBAAqB,EAAE,UAAU,MAAM,WAAW;AAChD,OAAK,gBAAgB;AACrB,SAAO,IAAI,IAAI;AACjB;AAGA,qBAAqB,EAAE,UAAU,OAAO,WAAW;AACjD,OAAK,gBAAgB;AACrB,SAAO,MAAM,IAAI;AACnB;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,MAAM;AACrD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,IAAI;AACxB;AAGA,qBAAqB,EAAE,UAAU,SAAS,WAAW;AACnD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,MAAM;AAC1B;AAGA,qBAAqB,EAAE,UAAU,UAAU,WAAW;AACpD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,SAAS;AAC7B;AAGA,qBAAqB,EAAE,UAAU,QAAQ,WAAW;AAClD,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,OAAO;AAC3B;AAGA,qBAAqB,EAAE,UAAU,OAAO,SAAS,GAAG,QAAQ;AAC1D,OAAK,gBAAgB;AACrB,SAAO,KAAK,MAAM,GAAG,MAAM;AAC7B;AAGA,qBAAqB,EAAE,UAAU,YAAY,SAAS,MAAM;AAC1D,OAAK,gBAAgB;AACrB,SAAO,UAAU,MAAM,IAAI;AAC7B;AAGA,qBAAqB,EAAE,UAAU,SAAS,SAAS,MAAM;AACvD,OAAK,gBAAgB;AACrB,SAAO,OAAO,MAAM,IAAI;AAC1B;AAGA,qBAAqB,EAAE,UAAU,qBAAqB,SAAS,YAAY,aAAa;AACtF,OAAK,gBAAgB;AACrB,SAAO,mBAAmB,MAAM,YAAY,WAAW;AACzD;AAGA,qBAAqB,EAAE,UAAU,UAAU,SAAS,MAAM;AACxD,OAAK,gBAAgB;AACrB,SAAO,QAAQ,MAAM,IAAI;AAC3B;AAGA,qBAAqB,EAAE,UAAU,QAAQ,SAAS,WAAW,GAAG;AAC9D,OAAK,gBAAgB;AACrB,SAAO,MAAM,WAAW,MAAM,CAAC;AACjC;AAGA,qBAAqB,EAAE,UAAU,YAAY,WAAW;AACtD,OAAK,gBAAgB;AACrB,SAAO,UAAU,IAAI;AACvB;AAGA,IAAI,iBAAiB,cAAc,MAAM;AAAA,EACvC,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,WAAO,eAAe,MAAM,eAAe,SAAS;AAAA,EACtD;AACF;AACA,IAAI,eAAe,cAAc,MAAM;AAAA,EACrC,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,WAAO,eAAe,MAAM,aAAa,SAAS;AAAA,EACpD;AACF;AACA,IAAI,aAAa,cAAc,MAAM;AAAA,EACnC,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,WAAO,eAAe,MAAM,WAAW,SAAS;AAAA,EAClD;AACF;AACA,IAAI,sBAAsB,cAAc,MAAM;AAAA,EAC5C,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,WAAO,eAAe,MAAM,oBAAoB,SAAS;AAAA,EAC3D;AACF;AACA,IAAI,iBAAiB,cAAc,MAAM;AAAA,EACvC,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,WAAO,eAAe,MAAM,eAAe,SAAS;AAAA,EACtD;AACF;AAGA,IAAI,WAAW,MAAM;AAAA,EACnB,YAAY,YAAY;AACtB,SAAK,aAAa,cAAc;AAChC,SAAK,QAAwB,oBAAI,IAAI;AAAA,EACvC;AAAA,EACA,IAAI,KAAK;AACP,QAAI;AACJ,QAAI,KAAK,MAAM,IAAI,GAAG,GAAG;AACvB,cAAQ,KAAK,MAAM,IAAI,GAAG;AAC1B,WAAK,MAAM,OAAO,GAAG;AACrB,WAAK,MAAM,IAAI,KAAK,KAAK;AAAA,IAC3B;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,KAAK,OAAO;AACd,QAAI,KAAK,MAAM,IAAI,GAAG,GAAG;AACvB,WAAK,MAAM,OAAO,GAAG;AAAA,IACvB,WAAW,KAAK,MAAM,QAAQ,KAAK,YAAY;AAC7C,YAAM,cAAc,KAAK,MAAM,KAAK,EAAE,KAAK,EAAE;AAC7C,WAAK,MAAM,OAAO,WAAW;AAAA,IAC/B;AACA,SAAK,MAAM,IAAI,KAAK,KAAK;AAAA,EAC3B;AAAA,EACA,gBAAgB;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,cAAc,YAAY;AACxB,QAAI,aAAa,GAAG;AAClB,YAAM,IAAI,MAAM,4DAA4D,aAAa;AAAA,IAC3F;AACA,QAAI,KAAK,aAAa,YAAY;AAChC,eAAS,IAAI,GAAG,IAAI,KAAK,aAAa,YAAY,KAAK;AACrD,cAAM,cAAc,KAAK,MAAM,KAAK,EAAE,KAAK,EAAE;AAC7C,aAAK,MAAM,OAAO,WAAW;AAAA,MAC/B;AAAA,IACF;AACA,SAAK,aAAa;AAAA,EACpB;AACF;AAGA,SAAS,aAAa,OAAO,WAAW;AACtC,MAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,QAAI,WAAW,CAAC;AAChB,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,iBAAW,SAAS,OAAO,KAAK;AAAA,IAClC;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,WAAW,IAAI,MAAM,SAAS;AACpC,aAAS,KAAK,KAAK;AACnB,WAAO;AAAA,EACT;AACF;AACA,SAAS,QAAQ,KAAK,SAAS;AAC7B,MAAI,CAAC,KAAK;AACR,UAAM,IAAI,eAAe,OAAO;AAAA,EAClC;AACF;AACA,SAAS,MAAM,QAAQ,UAAU;AAC/B,MAAI,UAAU;AACd,aAAW,QAAQ,QAAQ;AACzB,QAAI,SAAS,UAAU;AACrB;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,IAAI;AAC5B,MAAI,GAAG,WAAW,GAAG;AACnB,WAAO,GAAG;AAAA,EACZ;AACA,SAAO;AACT;AACA,SAAS,OAAO,GAAG;AACjB,MAAI,MAAM,QAAQ,CAAC,GAAG;AACpB,WAAO;AAAA,EACT;AACA,SAAO,CAAC,CAAC;AACX;AACA,SAAS,YAAY,MAAM;AACzB,QAAM,eAAe,KAAK,QAAQ,wBAAwB,OAAO;AACjE,QAAM,WAAW,aAAa,QAAQ,mBAAmB,OAAO,EAAE,YAAY;AAC9E,MAAI,SAAS,OAAO,KAAK;AACvB,WAAO;AAAA,EACT;AACA,SAAO,YAAY;AACrB;AACA,SAAS,YAAY,YAAY;AAC/B,MAAI,WAAW,UAAU,GAAG;AAC1B,WAAO;AAAA,EACT;AACA,MAAI,WAAW,QAAQ,GAAG,MAAM,IAAI;AAClC,WAAO;AAAA,EACT;AACA,SAAO,WAAW,QAAQ,eAAe,CAAC,GAAG,OAAO,GAAG,YAAY,CAAC;AACtE;AACA,IAAI,yBAAyB,CAAC;AAC9B,SAAS,qBAAqB,UAAU;AACtC,MAAI,aAAa,QAAQ,aAAa,QAAQ;AAC5C,WAAO;AAAA,EACT;AACA,QAAM,OAAO,CAAC;AACd,OAAK,eAAe,SAAS,aAAa;AAC1C,OAAK,YAAY,SAAS,UAAU;AACpC,SAAO;AACT;AACA,SAAS,8BAA8B,QAAQ;AAC7C,MAAI,UAAU,QAAQ,OAAO,WAAW,UAAU;AAChD;AAAA,EACF,WAAW,MAAM,QAAQ,MAAM,GAAG;AAChC,WAAO,QAAQ,CAAC,eAAe,8BAA8B,UAAU,CAAC;AAAA,EAC1E,OAAO;AACL,UAAM,SAAS,OAAO,KAAK,MAAM;AACjC,eAAW,SAAS,QAAQ;AAC1B,YAAM,QAAQ,OAAO;AACrB,UAAI,SAAS,QAAQ,OAAO,UAAU,UAAU;AAC9C,YAAI,CAAC,MAAM,QAAQ,KAAK,KAAK,MAAM,YAAY,aAAa,OAAO,MAAM,aAAa,UAAU;AAC9F,iBAAO,SAAS,MAAM;AAAA,QACxB,OAAO;AACL,wCAA8B,KAAK;AAAA,QACrC;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,uBAAuB,YAAY,gBAAgB,CAAC,GAAG,gBAAgB,CAAC,GAAG,sBAAsB,UAAU,iBAAiB,OAAO;AAC1I,MAAI,OAAO,eAAe,UAAU;AAClC,UAAM,eAAe;AACrB,QAAI;AACJ,QAAI,gBAAgB,eAAe;AACjC,WAAK,cAAc;AAAA,IACrB,WAAW,gBAAgB,wBAAwB;AACjD,WAAK,uBAAuB;AAAA,IAC9B,OAAO;AACL,WAAK,cAAc;AACnB,UAAI,MAAM,MAAM;AACd,cAAM,IAAI,WAAW,WAAW,wBAAwB;AAAA,SACvD;AAAA,gBACO,qHAAqH;AAAA,MAC/H;AAAA,IACF;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,SAAS;AACf,QAAI,OAAO,gBAAgB,QAAQ,OAAO,aAAa,MAAM;AAC3D,YAAM,IAAI,WAAW,GAAG,gDAAgD,KAAK,UAAU,MAAM;AAAA,mCAChE;AAAA,IAC/B;AACA,UAAM,YAAY,OAAO;AACzB,QAAI,KAAK;AACT,QAAI,aAAa,eAAe;AAC9B,OAAC,KAAK,UAAU,IAAI,cAAc;AAAA,IACpC,WAAW,aAAa,wBAAwB;AAC9C,OAAC,KAAK,UAAU,IAAI,uBAAuB;AAAA,IAC7C,WAAW,aAAa,eAAe;AACrC,OAAC,KAAK,UAAU,IAAI,cAAc;AAAA,IACpC;AACA,QAAI,OAAO,MAAM;AACf,YAAM,IAAI,WAAW,WAAW,wBAAwB;AAAA,SACrD;AAAA,gBACO,qHAAqH;AAAA,IACjI;AACA,QAAI,cAAc,MAAM;AACtB,YAAM,wBAAwB,CAAC;AAC/B,iBAAW,OAAO,OAAO,KAAK,sBAAsB,GAAG;AACrD,8BAAsB,OAAO,uBAAuB;AAAA,MACtD;AACA,iBAAW,OAAO,OAAO,KAAK,aAAa,GAAG;AAC5C,8BAAsB,OAAO,cAAc;AAAA,MAC7C;AACA,YAAM,eAAe,OAAO;AAC5B,mBAAa,mBAAmB;AAChC,YAAM,sBAAsB,OAAO,OAAO,CAAC,GAAG,sBAAsB;AACpE,iBAAW,OAAO,OAAO,KAAK,aAAa,GAAG;AAC5C,+BAAuB,OAAO,cAAc;AAAA,MAC9C;AACA,oCAA8B,OAAO,SAAS;AAC9C,YAAM,YAAY,WAAW,KAAK,OAAO,WAAW,eAAe,cAAc;AACjF,+BAAyB,OAAO,OAAO,CAAC,GAAG,mBAAmB;AAC9D,aAAO;AAAA,IACT,OAAO;AACL,YAAM,sBAAsB,OAAO,OAAO,CAAC,GAAG,sBAAsB;AACpE,iBAAW,OAAO,OAAO,KAAK,aAAa,GAAG;AAC5C,+BAAuB,OAAO,cAAc;AAAA,MAC9C;AACA,YAAM,YAAY,IAAI,IAAI,OAAO,SAAS;AAC1C,+BAAyB,OAAO,OAAO,CAAC,GAAG,mBAAmB;AAC9D,aAAO;AAAA,IACT;AAAA,EACF;AACF;AACA,SAAS,cAAc,GAAG,GAAG;AAC3B,SAAO,IAAI,IAAI,KAAK,IAAI,IAAI,IAAI;AAClC;AACA,SAAS,qBAAqB,GAAG,GAAG;AAClC,SAAO,KAAK,cAAc,GAAG,CAAC;AAChC;AACA,SAAS,QAAQ,IAAI;AACnB,MAAI,MAAM,MAAM;AACd,WAAO;AAAA,EACT;AACA,QAAM,MAAM,CAAC;AACb,aAAW,KAAK,IAAI;AAClB,QAAI,IAAI,QAAQ,CAAC,MAAM,IAAI;AACzB,UAAI,KAAK,CAAC;AAAA,IACZ;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,cAAc,KAAK;AAC1B,MAAI,OAAO,MAAM;AACf,UAAM,IAAI,WAAW,yBAAyB,KAAK,UAAU,GAAG,GAAG;AAAA,EACrE;AACA,aAAW,OAAO,KAAK;AACrB,QAAI,IAAI,eAAe,GAAG,GAAG;AAC3B,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,0BAA0B,QAAQ,OAAO,OAAO;AACvD,MAAI,SAAS,MAAM;AACjB;AAAA,EACF;AACA,MAAI,OAAO,QAAQ,KAAK,IAAI,GAAG;AAC7B,UAAM,IAAI,WAAW,GAAG,wBAAwB,4BAA4B,2BAA2B;AAAA,EACzG;AACF;AACA,SAAS,wBAAwB,GAAG,cAAc,YAAY,GAAG,YAAY,UAAU;AACrF,UAAQ,aAAa,CAAC;AACtB,UAAQ,aAAa,SAAS;AAC9B,SAAO,MAAM,QAAQ,CAAC,KAAK,EAAE,UAAU,aAAa,EAAE,UAAU,aAAa,EAAE,MAAM,CAAC,MAAM,OAAO,MAAM,YAAY;AACvH;AACA,SAAS,sBAAsB,OAAO,MAAM;AAC1C,MAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,iBAAa,OAAO,MAAM,SAAS,GAAG,MAAM,GAAG,sCAAsC;AACrF,UAAM,QAAQ,CAAC,GAAG,MAAM,sBAAsB,GAAG,WAAW,IAAI,QAAQ,MAAM,CAAC;AAAA,EACjF,OAAO;AACL,iBAAa,OAAO,OAAO,UAAU,KAAK,KAAK,QAAQ,GAAG,MAAM,YAAY,0CAA0C,uBAAuB,KAAK,IAAI;AAAA,EACxJ;AACF;AACA,SAAS,uBAAuB,OAAO;AACrC,MAAI,UAAU,MAAM;AAClB,WAAO;AAAA,EACT,WAAW,MAAM,QAAQ,KAAK,GAAG;AAC/B,WAAO,MAAM,MAAM,IAAI,CAAC,MAAM,uBAAuB,CAAC,CAAC,EAAE,KAAK,GAAG,IAAI;AAAA,EACvE,WAAW,OAAO,UAAU,UAAU;AACpC,WAAO,IAAI;AAAA,EACb,OAAO;AACL,WAAO,GAAG;AAAA,EACZ;AACF;AACA,SAAS,SAAS,GAAG,QAAQ,SAAS;AACpC,MAAI,WAAW,WAAW,OAAO,QAAQ,IAAI,aAAa,IAAI;AAC9D,MAAI;AACJ,QAAM,KAAK,IAAI,SAAS;AACtB,UAAM,OAAO,WAAW,OAAO,QAAQ,IAAI,aAAa,IAAI;AAC5D,QAAI,OAAO,WAAW,QAAQ;AAC5B,aAAO;AAAA,IACT;AACA,eAAW;AACX,iBAAa,EAAE,GAAG,IAAI;AACtB,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,2BAA2B,gBAAgB;AAClD,MAAI,mBAAmB,QAAQ;AAC7B,WAAO;AAAA,EACT;AACA,MAAI,mBAAmB,UAAU;AAC/B,WAAO;AAAA,EACT;AACA,MAAI,mBAAmB,OAAO;AAC5B,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAGA,IAAI,sBAAsB;AAC1B,SAAS,wBAAwB;AAC/B,SAAO;AACT;AACA,IAAI,eAAe,CAAC;AACpB,SAAS,OAAO,SAAS,IAAI;AAC3B,MAAI,EAAE,UAAU,eAAe;AAC7B,iBAAa,UAAU;AAAA,EACzB;AACA,eAAa,WAAW;AACxB,SAAO,SAAS,aAAa,QAAQ,SAAS;AAChD;AAGA,IAAI,2BAA2B,CAAC,iBAAiB,cAAc;AAC/D,IAAI,oCAAoC,CAAC,WAAW,UAAU;AAC9D,IAAI,4BAA4B,CAAC,SAAS,QAAQ,QAAQ;AAC1D,IAAI,yBAAyB,CAAC,OAAO,KAAK;AAC1C,IAAI,kCAAkC,CAAC,OAAO,OAAO,UAAU,KAAK;AAGpE,IAAI,UAA0B,oBAAI,IAAI;AACtC,SAAS,gBAAgB,OAAO;AAC9B,4BAA0B,0BAA0B,cAAc,KAAK;AACzE;AACA,SAAS,yBAAyB,OAAO;AACvC,4BAA0B,mCAAmC,uBAAuB,KAAK;AAC3F;AACA,SAAS,iBAAiB,OAAO;AAC/B,4BAA0B,2BAA2B,eAAe,KAAK;AAC3E;AACA,SAAS,cAAc,OAAO;AAC5B,4BAA0B,wBAAwB,YAAY,KAAK;AACrE;AACA,IAAI,kBAAkB,CAAC;AACvB,IAAI,oBAAoB;AACxB,SAAS,UAAU,MAAM,IAAI;AAC3B,kBAAgB,KAAK,IAAI;AACzB,MAAI;AACF,UAAM,MAAM,GAAG;AACf,oBAAgB,IAAI;AACpB,WAAO;AAAA,EACT,SAAS,GAAP;AACA,oBAAgB,IAAI;AACpB,UAAM;AAAA,EACR;AACF;AACA,SAAS,yBAAyB;AAChC,MAAI,gBAAgB,WAAW,GAAG;AAChC,WAAO;AAAA,EACT,OAAO;AACL,WAAO,gBAAgB,KAAK,iBAAiB,IAAI;AAAA,EACnD;AACF;AACA,SAAS,oBAAoB,YAAY;AACvC,MAAI,CAAC,kBAAkB,UAAU,GAAG;AAClC,UAAM,IAAI,MAAM,+BAA+B,aAAa,GAAG;AAAA,EACjE;AACA,SAAO,uBAAuB,IAAI;AACpC;AACA,SAAS,oBAAoB,YAAY;AACvC,MAAI,CAAC,kBAAkB,UAAU,GAAG;AAClC,UAAM,IAAI,MAAM,+BAA+B,aAAa,GAAG;AAAA,EACjE;AACA,MAAI,CAAC,QAAQ,IAAI,UAAU,GAAG;AAC5B,YAAQ,IAAI,YAAY,CAAC;AAAA,EAC3B;AACA,QAAM,QAAQ,QAAQ,IAAI,UAAU;AACpC,UAAQ,IAAI,YAAY,QAAQ,IAAI,UAAU,IAAI,CAAC;AACnD,MAAI,QAAQ,GAAG;AACb,UAAM,SAAS,GAAG,cAAc;AAChC,YAAQ,IAAI,QAAQ,CAAC;AACrB,WAAO;AAAA,EACT,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,IAAI,kBAAkB,IAAI,OAAO,iCAAiC;AAClE,SAAS,kBAAkB,MAAM;AAC/B,SAAO,CAAC,CAAC,KAAK,MAAM,eAAe;AACrC;AAGA,SAAS,UAAU,GAAG;AACpB,SAAO,MAAM,SAAS,EAAE,SAAS,GAAG,EAAE;AACxC;AACA,SAAS,UAAU,QAAQ,OAAO,KAAK;AACrC,MAAI,SAAS,MAAM;AACjB,YAAQ;AAAA,EACV;AACA,MAAI,OAAO,MAAM;AACf,UAAM,OAAO;AAAA,EACf;AACA,MAAI,QAAQ;AACZ,WAAS,IAAI,OAAO,IAAI,KAAK,EAAE,GAAG;AAChC,aAAS,OAAO;AAAA,EAClB;AACA,SAAO;AACT;AACA,SAAS,KAAK,QAAQ;AACpB,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,OAAO;AAAA,EAChB;AACA,MAAI,OAAO,OAAO;AAClB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,UAAM,QAAQ,OAAO;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,KAAK,QAAQ;AACpB,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,OAAO;AAAA,EAChB;AACA,MAAI,OAAO,OAAO;AAClB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,UAAM,QAAQ,OAAO;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,OAAO,OAAO,KAAK;AAC1B,MAAI,MAAM,OAAO;AACf,UAAM,IAAI,WAAW,QAAQ,iBAAiB,sBAAsB;AAAA,EACtE;AACA,QAAM,MAAM,CAAC;AACb,WAAS,IAAI,OAAO,IAAI,KAAK,EAAE,GAAG;AAChC,QAAI,KAAK,CAAC;AAAA,EACZ;AACA,SAAO;AACT;AAGA,IAAI;AACJ,SAAS,UAAU;AACjB,MAAI,YAAY,MAAM;AACpB,eAAW,QAAQ,EAAE,QAAQ;AAAA,EAC/B;AACA,SAAO;AACT;AACA,SAAS,kBAAkB;AACzB,SAAO;AACT;AAGA,SAAS,MAAM,GAAG,OAAO;AACvB,SAAO,KAAK,GAAG,KAAK;AACtB;AACA,SAAS,YAAY,GAAG,OAAO,IAAI;AACjC,QAAM,WAAW,EAAE,MAAM,MAAM;AAC/B,MAAI,OAAO,GAAG;AACZ,WAAO,SAAS,SAAS,OAAO;AAAA,EAClC;AACA,WAAS,OAAO,MAAM,GAAG,CAAC;AAC1B,SAAO,QAAQ,GAAG,QAAQ;AAC5B;AACA,SAAS,OAAO,GAAG,GAAG;AACpB,SAAO,KAAK,MAAM;AAChB,QAAI,EAAE,MAAM,WAAW,GAAG;AACxB,YAAM,IAAI,WAAW,yDAAyD,EAAE,MAAM,gBAAgB;AAAA,IACxG;AACA,UAAM,IAAI,YAAY,GAAG,CAAC;AAC1B,WAAO,MAAM,GAAG,CAAC,GAAG,GAAG,CAAC,CAAC;AAAA,EAC3B,CAAC;AACH;AACA,SAAS,SAAS,GAAG;AACnB,QAAM,WAAW,CAAC,UAAU,EAAE,KAAK,CAAC;AACpC,SAAO,QAAQ,GAAG,QAAQ;AAC5B;AACA,SAAS,aAAa,GAAG;AACvB,MAAI,EAAE,QAAQ,GAAG;AACf,UAAM,IAAI,WAAW,wDAAwD,EAAE,OAAO;AAAA,EACxF;AACA,QAAM,WAAW,CAAC,EAAE,MAAM,IAAI,UAAU,EAAE,OAAO,CAAC,CAAC;AACnD,SAAO,QAAQ,GAAG,QAAQ;AAC5B;AACA,SAAS,oBAAoB,QAAQ,OAAO,MAAM;AAChD,SAAO,KAAK,MAAM;AAChB,YAAQ,OAAO,MAAM;AAAA,MACnB,KAAK;AACH,eAAO,QAAQ,QAAQ,OAAO,IAAI;AAAA,MACpC,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,OAAO,CAAC,GAAG,CAAC,MAAM,OAAO,MAAM,EAAE,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,OAAO,GAAG,CAAC,GAAG,CAAC,MAAM,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE,CAAC;AAAA,MAChF,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,OAAO,GAAG,GAAG,CAAC,GAAG,CAAC,MAAM,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE,CAAC;AAAA,MACpG,KAAK;AACH,eAAO,MAAM,QAAQ,CAAC,OAAO,GAAG,GAAG,GAAG,CAAC,GAAG;AAAA,UACxC;AAAA,UACA,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,QACf,CAAC;AAAA,MACH,KAAK;AACH,eAAO,MAAM,QAAQ,CAAC,OAAO,GAAG,GAAG,GAAG,GAAG,CAAC,GAAG;AAAA,UAC3C;AAAA,UACA,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,UACb,OAAO,MAAM;AAAA,QACf,CAAC;AAAA,MACH;AACE,cAAM,IAAI,WAAW,8DAA8D,OAAO,MAAM;AAAA,IACpG;AAAA,EACF,CAAC;AACH;AACA,SAAS,mBAAmB,QAAQ,OAAO,MAAM;AAC/C,SAAO,KAAK,MAAM;AAChB,YAAQ,OAAO,MAAM;AAAA,MACnB,KAAK;AACH,eAAO,QAAQ,QAAQ,OAAO,IAAI;AAAA,MACpC,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,GAAG,KAAK,GAAG,CAAC,OAAO,MAAM,IAAI,IAAI,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,GAAG,GAAG,KAAK,GAAG,CAAC,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,IAAI,CAAC;AAAA,MAChF,KAAK;AACH,eAAO,QAAQ,QAAQ,CAAC,GAAG,GAAG,GAAG,KAAK,GAAG,CAAC,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,IAAI,CAAC;AAAA,MACpG;AACE,cAAM,IAAI,WAAW,6DAA6D,OAAO,MAAM;AAAA,IACnG;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,QAAQ,OAAO,MAAM,MAAM;AACjD,SAAO,KAAK,MAAM;AAChB,YAAQ,OAAO,MAAM;AAAA,MACnB,KAAK;AACH,eAAO,QAAQ,QAAQ,OAAO,IAAI;AAAA,MACpC,KAAK;AACH,gBAAQ,MAAM;AAAA,UACZ,KAAK;AACH,mBAAO,oBAAoB,QAAQ,OAAO,IAAI;AAAA,UAChD,KAAK;AACH,mBAAO,mBAAmB,QAAQ,OAAO,IAAI;AAAA,UAC/C;AACE,kBAAM,IAAI,WAAW,iDAAiD,MAAM;AAAA,QAChF;AAAA,MACF,KAAK;AACH,gBAAQ,MAAM;AAAA,UACZ,KAAK;AACH,mBAAO,oBAAoB,QAAQ,OAAO,IAAI;AAAA,UAChD,KAAK;AACH,mBAAO,QAAQ,QAAQ,CAAC,GAAG,OAAO,CAAC,GAAG,CAAC,OAAO,MAAM,IAAI,MAAM,OAAO,MAAM,EAAE,CAAC;AAAA,UAChF,KAAK;AACH,mBAAO,mBAAmB,QAAQ,OAAO,IAAI;AAAA,UAC/C;AACE,kBAAM,IAAI,WAAW,iDAAiD,MAAM;AAAA,QAChF;AAAA,MACF,KAAK;AACH,gBAAQ,MAAM;AAAA,UACZ,KAAK;AACH,mBAAO,oBAAoB,QAAQ,OAAO,IAAI;AAAA,UAChD,KAAK;AACH,mBAAO,QAAQ,QAAQ,CAAC,GAAG,OAAO,GAAG,CAAC,GAAG,CAAC,OAAO,MAAM,IAAI,MAAM,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE,CAAC;AAAA,UACpG,KAAK;AACH,mBAAO,QAAQ,QAAQ,CAAC,GAAG,GAAG,OAAO,CAAC,GAAG,CAAC,OAAO,MAAM,IAAI,OAAO,MAAM,IAAI,MAAM,OAAO,MAAM,EAAE,CAAC;AAAA,UACpG,KAAK;AACH,mBAAO,mBAAmB,QAAQ,OAAO,IAAI;AAAA,UAC/C;AACE,kBAAM,IAAI,WAAW,iDAAiD,MAAM;AAAA,QAChF;AAAA,MACF;AACE,cAAM,IAAI,WAAW,6DAA6D,OAAO,MAAM;AAAA,IACnG;AAAA,EACF,CAAC;AACH;AACA,SAAS,YAAY,SAAS,OAAO,IAAI;AACvC,MAAI;AACJ,MAAI,OAAO,GAAG;AACZ,WAAO,QAAQ,GAAG;AAClB,QAAI,SAAS,GAAG;AACd,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AACA,MAAI,SAAS,QAAQ,GAAG,MAAM;AAC5B,WAAO;AAAA,EACT;AACA,SAAO,OAAO,SAAS,IAAI;AAC7B;AACA,SAAS,qBAAqB,GAAG,GAAG;AAClC,UAAQ,EAAE,MAAM;AAAA,IACd,KAAK;AACH,aAAO,SAAS,CAAC,GAAG,CAAC,CAAC;AAAA,IACxB,KAAK;AACH,aAAO,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC;AAAA,IAC3B,KAAK;AACH,aAAO,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC;AAAA,IAC3B,KAAK;AACH,aAAO,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC;AAAA,IAC3B;AACE,YAAM,IAAI,WAAW,+DAA+D,EAAE,MAAM;AAAA,EAChG;AACF;AACA,SAAS,MAAM,GAAG,GAAG;AACnB,MAAI,CAAC,MAAM,QAAQ,CAAC,GAAG;AACrB,QAAI,CAAC,CAAC;AAAA,EACR;AACA,MAAI,EAAE,SAAS,EAAE,QAAQ;AACvB,UAAM,IAAI,WAAW,0BAA0B,EAAE,+DAA+D,EAAE,OAAO;AAAA,EAC3H;AACA,SAAO,KAAK,GAAG,CAAC;AAClB;AACA,SAAS,cAAc,OAAO,QAAQ,GAAG,SAAS,GAAG,OAAO,MAAM;AAChE,SAAO,aAAa,OAAO,OAAO,QAAQ,OAAO,IAAI;AACvD;AACA,SAAS,KAAK,GAAG,GAAG,aAAa,MAAM;AACrC,MAAI,EAAE,OAAO,KAAK,EAAE,OAAO,GAAG;AAC5B,UAAM,IAAI,oBAAoB,8DAA8D,EAAE,uBAAuB,EAAE,OAAO;AAAA,EAChI;AACA,MAAI,EAAE,QAAQ,GAAG;AACf,UAAM,WAAW,EAAE,MAAM,MAAM,EAAE,EAAE;AACnC,UAAM,iBAAiB,EAAE,MAAM,MAAM,EAAE,EAAE;AACzC,QAAI,aAAa,gBAAgB;AAC/B,YAAM,IAAI,oBAAoB,gGAAgG,EAAE,wBAAwB,EAAE,OAAO;AAAA,IACnK;AAAA,EACF;AACA,MAAI,EAAE,SAAS,KAAK,EAAE,SAAS,GAAG;AAChC,UAAM,aAAa;AACnB,UAAM,aAAa;AACnB,WAAO,kBAAkB,OAAO;AAAA,MAC9B;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,MAAM,OAAO,YAAY,EAAE,MAAM,MAAM,gBAAgB,CAAC,IAAI;AAAA,MAC5D,YAAY;AAAA,IACd,CAAC;AAAA,EACH,OAAO;AACL,UAAM,aAAa,EAAE,MAAM,MAAM;AACjC,UAAM,WAAW,WAAW,IAAI;AAChC,QAAI,QAAQ,GAAG,CAAC,IAAI,QAAQ,CAAC;AAC7B,UAAM,SAAS,EAAE,MAAM,MAAM;AAC7B,UAAM,WAAW,OAAO,IAAI;AAC5B,UAAM,iBAAiB,OAAO,IAAI;AAClC,UAAM,aAAa,CAAC,GAAG,QAAQ,QAAQ;AACvC,UAAM,OAAO,MAAM,KAAK,EAAE,QAAQ,EAAE,KAAK,GAAG,CAAC,GAAG,MAAM;AACpD,UAAI,MAAM,GAAG;AACX,eAAO,EAAE,OAAO;AAAA,MAClB,WAAW,KAAK,EAAE,OAAO,GAAG;AAC1B,eAAO,IAAI;AAAA,MACb;AACA,aAAO;AAAA,IACT,CAAC;AACD,QAAI,QAAQ,UAAU,GAAG,IAAI,GAAG,CAAC,gBAAgB,EAAE,CAAC;AACpD,UAAM,cAAc,CAAC,GAAG,YAAY,GAAG,UAAU;AACjD,UAAM,aAAa;AACnB,UAAM,aAAa;AACnB,WAAO,QAAQ,kBAAkB,OAAO;AAAA,MACtC;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,MAAM,OAAO,YAAY,EAAE,MAAM,MAAM,gBAAgB,CAAC,IAAI;AAAA,MAC5D,YAAY;AAAA,IACd,CAAC,GAAG,WAAW;AAAA,EACjB;AACF;AACA,SAAS,QAAQ,WAAW,SAAS,MAAM;AACzC,SAAO,KAAK,MAAM;AAChB,QAAI,MAAM,QAAQ,OAAO,GAAG;AAC1B,gBAAU,SAAS,SAAS,OAAO;AAAA,IACrC,OAAO;AACL,gBAAU,KAAK,SAAS,OAAO;AAAA,IACjC;AACA,WAAO,OAAO,WAAW,SAAS,IAAI;AAAA,EACxC,CAAC;AACH;AACA,SAAS,QAAQ,GAAG;AAClB,SAAO,IAAI,GAAG,CAAC;AACjB;AACA,SAAS,YAAY,OAAO,MAAM,YAAY;AAC5C,QAAM,YAAY,KAAK;AACvB,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,OAAO;AAC1C,UAAM,IAAI,WAAW,+BAA+B,KAAK,gCAAgC,OAAO;AAAA,EAClG;AACA,MAAI,UAAU,GAAG;AACf,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,GAAG,GAAG,CAAC,CAAC;AAAA,MACjD,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,UAAU,IAAI,UAAU,IAAI,UAAU,EAAE,CAAC;AAAA,MAClF;AAAA,IACF,WAAW,eAAe,gBAAgB;AACxC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,GAAG,GAAG,GAAG,UAAU,EAAE,CAAC;AAAA,MACjD,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,CAAC,EAAE,OAAO,SAAS,CAAC;AAAA,MAC5C;AAAA,IACF;AAAA,EACF,WAAW,UAAU,GAAG;AACtB,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,GAAG,CAAC,CAAC;AAAA,MAC9C,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,UAAU,IAAI,UAAU,EAAE,CAAC;AAAA,MACpE;AAAA,IACF,WAAW,eAAe,gBAAgB;AACxC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,GAAG,GAAG,UAAU,EAAE,CAAC;AAAA,MAC9C,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,CAAC,EAAE,OAAO,SAAS,CAAC;AAAA,MAC5C;AAAA,IACF;AAAA,EACF,WAAW,UAAU,GAAG;AACtB,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,CAAC,CAAC;AAAA,MAC3C,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,GAAG,UAAU,IAAI,UAAU,EAAE,CAAC;AAAA,MACtD;AAAA,IACF,WAAW,eAAe,gBAAgB;AACxC,UAAI,UAAU,WAAW,GAAG;AAC1B,eAAO,QAAQ,MAAM,CAAC,GAAG,GAAG,UAAU,EAAE,CAAC;AAAA,MAC3C,OAAO;AACL,eAAO,QAAQ,MAAM,CAAC,CAAC,EAAE,OAAO,SAAS,CAAC;AAAA,MAC5C;AAAA,IACF;AAAA,EACF,WAAW,QAAQ,GAAG;AACpB,WAAO;AAAA,EACT;AACA,QAAM,IAAI,WAAW,sCAAsC,KAAK,MAAM;AACxE;AACA,SAAS,QAAQ,GAAG,MAAM,YAAY;AACpC,SAAO,KAAK,MAAM;AAChB,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,oBAAgB,UAAU;AAC1B,WAAO,KAAK,GAAG,YAAY,EAAE,MAAM,MAAM,UAAU,CAAC;AAAA,EACtD,CAAC;AACH;AACA,SAAS,KAAK,GAAG,QAAQ,GAAG;AAC1B,MAAI,UAAU,GAAG;AACf,UAAM,IAAI,oBAAoB,0CAA0C,gCAAgC;AAAA,EAC1G;AACA,SAAO,IAAI,CAAC;AACd;AACA,SAAS,SAAS,GAAG;AACnB,SAAO,KAAK,MAAM,IAAI,GAAG,KAAK,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC;AAC3C;AACA,SAAS,SAAS,GAAG,OAAO,YAAY,MAAM;AAC5C,SAAO,KAAK,MAAM,QAAQ,GAAG,OAAO,YAAY,IAAI,CAAC;AACvD;AACA,SAAS,YAAY,GAAG;AACtB,SAAO,KAAK,MAAM;AAChB,UAAM,IAAI,KAAK,KAAK,IAAI,KAAK,CAAC,CAAC;AAC/B,WAAO,YAAY,GAAG,GAAG,CAAC;AAAA,EAC5B,CAAC;AACH;AACA,SAAS,aAAa,GAAG,KAAK,WAAW,OAAO;AAC9C,SAAO,WAAW,EAAE,IAAI,IAAI;AAC9B;AAGA,IAAI,wBAAwB,CAAC,SAAS,UAAU,QAAQ;AACxD,IAAI,4BAA4B,CAAC,UAAU,WAAW,iBAAiB;AAGvE,SAAS,aAAa,OAAO;AAC3B,4BAA0B,uBAAuB,WAAW,KAAK;AACnE;AACA,SAAS,kBAAkB,OAAO;AAChC,4BAA0B,2BAA2B,gBAAgB,KAAK;AAC5E;AACA,IAAI,cAAc,cAAc,sBAAsB,aAAa;AAAA,EACjE,8BAA8B;AAC5B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,WAAO,CAAC;AAAA,EACV;AACF;AACA,IAAI,QAAQ,cAAc,YAAY;AAAA,EACpC,MAAM,OAAO,OAAO;AAClB,WAAO,MAAM,OAAO,KAAK;AAAA,EAC3B;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,OAAO,cAAc,YAAY;AAAA,EACnC,MAAM,OAAO,OAAO;AAClB,WAAO,MAAM,OAAO,KAAK;AAAA,EAC3B;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,IAAI,WAAW,cAAc,YAAY;AAAA,EACvC,YAAY,MAAM;AAChB,UAAM;AACN,QAAI,OAAO,SAAS,UAAU;AAC5B,YAAM,IAAI,WAAW,oDAAoD,MAAM;AAAA,IACjF;AACA,QAAI,KAAK,UAAU,QAAQ;AACzB,YAAM,IAAI,WAAW,sCAAsC,MAAM;AAAA,IACnE;AACA,SAAK,QAAQ,KAAK;AAAA,EACpB;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,WAAO,KAAK,MAAM,IAAI,OAAO,KAAK,KAAK,GAAG,MAAM,OAAO,KAAK,CAAC,CAAC;AAAA,EAChE;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,OAAO,KAAK;AAAA,IACd;AAAA,EACF;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,gBAAgB,cAAc,YAAY;AAAA,EAC5C,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,iBAAiB;AACtB,SAAK,iBAAiB;AACtB,SAAK,SAAS,KAAK,UAAU,KAAK;AAClC,SAAK,SAAS,KAAK,UAAU,KAAK;AAClC,SAAK,OAAO,KAAK;AAAA,EACnB;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,WAAO,cAAc,OAAO,KAAK,QAAQ,KAAK,QAAQ,KAAK;AAAA,EAC7D;AAAA,EACA,YAAY;AACV,WAAO,EAAE,QAAQ,KAAK,QAAQ,QAAQ,KAAK,QAAQ,MAAM,KAAK,KAAK;AAAA,EACrE;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AACjD,IAAI,eAAe,cAAc,YAAY;AAAA,EAC3C,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,OAAO,KAAK,QAAQ,KAAK;AAC9B,SAAK,SAAS,KAAK,UAAU,KAAK;AAClC,SAAK,OAAO,KAAK;AAAA,EACnB;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,YAAQ,SAAS;AACjB,QAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,YAAM,IAAI,oBAAoB,uCAAuC,QAAQ;AAAA,IAC/E;AACA,WAAO,cAAc,OAAO,KAAK,MAAM,KAAK,QAAQ,OAAO,KAAK,IAAI;AAAA,EACtE;AAAA,EACA,YAAY;AACV,WAAO,EAAE,MAAM,KAAK,MAAM,QAAQ,KAAK,QAAQ,MAAM,KAAK,KAAK;AAAA,EACjE;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,kBAAkB,cAAc,YAAY;AAAA,EAC9C,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,OAAO,KAAK,QAAQ,KAAK;AAC9B,SAAK,SAAS,KAAK,UAAU,KAAK;AAClC,SAAK,OAAO,KAAK;AAAA,EACnB;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,YAAQ,SAAS;AACjB,QAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,YAAM,IAAI,oBAAoB,0CAA0C,QAAQ;AAAA,IAClF;AACA,WAAO,gBAAgB,OAAO,KAAK,MAAM,KAAK,QAAQ,OAAO,KAAK,IAAI;AAAA,EACxE;AAAA,EACA,YAAY;AACV,WAAO,EAAE,MAAM,KAAK,MAAM,QAAQ,KAAK,QAAQ,MAAM,KAAK,KAAK;AAAA,EACjE;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,YAAY,cAAc,YAAY;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO;AAAA,EAC9C;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,WAAO,KAAK,MAAM;AAChB,UAAI,MAAM,WAAW,KAAK,MAAM,OAAO,MAAM,IAAI;AAC/C,cAAM,IAAI,WAAW,sEAAsE;AAAA,MAC7F,OAAO;AACL,eAAO,IAAI,KAAK,MAAM,IAAI,MAAM,EAAE,CAAC;AAAA,MACrC;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,WAAO,EAAE,MAAM,KAAK,KAAK;AAAA,EAC3B;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,SAAS,YAAY,OAAO,aAAa,gBAAgB;AACvD,MAAI;AACJ,MAAI;AACJ,kBAAgB,UAAU;AAC1B,MAAI,MAAM,WAAW,GAAG;AACtB,YAAQ,MAAM;AACd,aAAS,MAAM;AAAA,EACjB,WAAW,CAAC,GAAG,GAAG,CAAC,EAAE,QAAQ,MAAM,MAAM,MAAM,IAAI;AACjD,QAAI,eAAe,iBAAiB;AAClC,YAAM,qBAAqB,UAAU,OAAO,CAAC;AAC7C,cAAQ,MAAM,KAAK;AACnB,eAAS,MAAM,KAAK;AAAA,IACtB,WAAW,eAAe,gBAAgB;AACxC,YAAM,qBAAqB,UAAU,OAAO,GAAG,MAAM,SAAS,CAAC;AAC/D,cAAQ,MAAM,MAAM,SAAS,KAAK;AAClC,eAAS,MAAM,MAAM,SAAS,KAAK;AAAA,IACrC;AAAA,EACF,OAAO;AACL,UAAM,YAAY,UAAU,KAAK;AACjC,YAAQ,KAAK,KAAK,SAAS;AAC3B,aAAS,KAAK,KAAK,SAAS;AAAA,EAC9B;AACA,SAAO,CAAC,OAAO,MAAM;AACvB;AACA,IAAI,kBAAkB,cAAc,YAAY;AAAA,EAC9C,YAAY,MAAM;AAChB,UAAM;AACN,QAAI,KAAK,QAAQ,GAAG;AAClB,YAAM,IAAI,WAAW,wCAAwC,KAAK,OAAO;AAAA,IAC3E;AACA,SAAK,QAAQ,KAAK,SAAS,OAAO,IAAI,KAAK;AAC3C,SAAK,OAAO,KAAK,QAAQ,OAAO,UAAU,KAAK;AAC/C,iBAAa,KAAK,IAAI;AACtB,SAAK,eAAe,KAAK,gBAAgB,OAAO,WAAW,KAAK;AAChE,sBAAkB,KAAK,YAAY;AACnC,SAAK,OAAO,KAAK;AAAA,EACnB;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,UAAM,OAAO,YAAY,KAAK;AAC9B,UAAM,QAAQ,KAAK;AACnB,UAAM,SAAS,KAAK;AACpB,QAAIA,UAAS,KAAK;AAClB,QAAI,KAAK,SAAS,SAAS;AACzB,MAAAA,WAAU,KAAK,IAAI,GAAG,KAAK;AAAA,IAC7B,WAAW,KAAK,SAAS,UAAU;AACjC,MAAAA,WAAU,KAAK,IAAI,GAAG,MAAM;AAAA,IAC9B,OAAO;AACL,MAAAA,WAAU,KAAK,IAAI,IAAI,QAAQ,UAAU,CAAC;AAAA,IAC5C;AACA,QAAI,KAAK,iBAAiB,UAAU;AAClC,YAAM,SAAS,KAAK,KAAKA,OAAM;AAC/B,cAAQ,SAAS;AACjB,UAAI,UAAU,aAAa,UAAU,SAAS;AAC5C,cAAM,IAAI,oBAAoB,GAAG,KAAK,aAAa,4BAA4B,QAAQ;AAAA,MACzF;AACA,aAAO,gBAAgB,OAAO,GAAG,QAAQ,OAAO,KAAK,IAAI;AAAA,IAC3D,OAAO;AACL,YAAM,QAAQ,KAAK,KAAK,IAAIA,OAAM;AAClC,aAAO,cAAc,OAAO,CAAC,OAAO,OAAO,KAAK;AAAA,IAClD;AAAA,EACF;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,OAAO,KAAK;AAAA,MACZ,MAAM,KAAK;AAAA,MACX,cAAc,KAAK;AAAA,MACnB,MAAM,KAAK;AAAA,IACb;AAAA,EACF;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,gBAAgB,cAAc,gBAAgB;AAAA,EAChD,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AACjD,IAAI,eAAe,cAAc,gBAAgB;AAAA,EAC/C,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,WAAW,cAAc,gBAAgB;AAAA,EAC3C,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,YAAY,cAAc,gBAAgB;AAAA,EAC5C,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,IAAI,cAAc,cAAc,gBAAgB;AAAA,EAC9C,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,IAAI,eAAe,cAAc,gBAAgB;AAAA,EAC/C,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO;AAAA,MACP,MAAM;AAAA,MACN,cAAc;AAAA,MACd,MAAM,QAAQ,OAAO,OAAO,KAAK;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,eAAe;AACb,WAAO,gBAAgB;AAAA,EACzB;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,aAAa,cAAc,YAAY;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,eAAe,KAAK;AACzD,SAAK,OAAO,KAAK;AACjB,QAAI,KAAK,QAAQ,MAAM;AACrB,YAAM,IAAI,oBAAoB,gEAAgE;AAAA,IAChG;AAAA,EACF;AAAA,EACA,MAAM,OAAO,OAAO;AAClB,WAAO,KAAK,MAAM;AAChB,UAAI,MAAM,SAAS,GAAG;AACpB,cAAM,IAAI,oBAAoB,4BAA4B;AAAA,MAC5D;AACA,UAAI,MAAM,KAAK,MAAM,KAAK,KAAK;AAC7B,gBAAQ,KAAK,2EAA2E,MAAM,KAAK,MAAM,oCAAoC;AAAA,MAC/I;AACA,YAAM,kBAAkB,MAAM,KAAK,MAAM,KAAK,CAAC,MAAM,IAAI,MAAM,EAAE,IAAI;AACrE,YAAM,IAAI,cAAc,iBAAiB,GAAG,GAAG,SAAS;AACxD,UAAI,IAAI,OAAO,YAAY,CAAC;AAC5B,UAAI,MAAM,KAAK,MAAM,IAAI;AACvB,YAAI,UAAU,CAAC;AAAA,MACjB;AACA,aAAO,IAAI,KAAK,MAAM,CAAC;AAAA,IACzB,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,MAAM,KAAK;AAAA,MACX,MAAM,KAAK;AAAA,IACb;AAAA,EACF;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAC9C,IAAI,6CAA6C;AAAA,EAC/C,YAAY;AAAA,EACZ,gBAAgB;AAAA,EAChB,iBAAiB;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,eAAe;AAAA,EACf,gBAAgB;AAAA,EAChB,QAAQ;AAAA,EACR,cAAc;AAAA,EACd,gBAAgB;AAAA,EAChB,iBAAiB;AAAA,EACjB,mBAAmB;AAAA,EACnB,mBAAmB;AAAA,EACnB,SAAS;AACX;AACA,SAAS,uBAAuB,QAAQ,gBAAgB,CAAC,GAAG;AAC1D,SAAO,uBAAuB,QAAQ,sBAAsB,iBAAiB,OAAO,EAAE,cAAc,eAAe,aAAa;AAClI;AACA,SAAS,qBAAqB,aAAa;AACzC,SAAO,qBAAqB,WAAW;AACzC;AACA,SAAS,eAAe,YAAY;AAClC,MAAI,OAAO,eAAe,UAAU;AAClC,UAAM,YAAY,cAAc,6CAA6C,2CAA2C,cAAc;AACtI,QAAI,cAAc,gBAAgB;AAChC,aAAO,IAAI,aAAa;AAAA,IAC1B,WAAW,cAAc,iBAAiB;AACxC,aAAO,IAAI,cAAc;AAAA,IAC3B,WAAW,cAAc,YAAY;AACnC,aAAO,IAAI,SAAS;AAAA,IACtB,WAAW,cAAc,aAAa;AACpC,aAAO,IAAI,UAAU;AAAA,IACvB,WAAW,cAAc,eAAe;AACtC,aAAO,IAAI,YAAY;AAAA,IACzB,WAAW,cAAc,gBAAgB;AACvC,aAAO,IAAI,aAAa;AAAA,IAC1B,OAAO;AACL,YAAM,SAAS,CAAC;AAChB,aAAO,eAAe;AACtB,aAAO,YAAY,CAAC;AACpB,aAAO,uBAAuB,MAAM;AAAA,IACtC;AAAA,EACF,WAAW,sBAAsB,aAAa;AAC5C,WAAO;AAAA,EACT,OAAO;AACL,WAAO,uBAAuB,UAAU;AAAA,EAC1C;AACF;AAGA,SAAS,gBAAgB,GAAG;AAC1B,SAAO,MAAM,QAAQ,CAAC,KAAK,MAAM,QAAQ,EAAE,EAAE;AAC/C;AACA,SAAS,mBAAmB,GAAG;AAC7B,MAAI,EAAE,WAAW,GAAG;AAClB,WAAO,CAAC;AAAA,EACV;AACA,MAAI,CAAC,MAAM,QAAQ,EAAE,EAAE,GAAG;AACxB,WAAO,CAAC,CAAC;AAAA,EACX;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,IAAI;AAC/B,MAAI;AACJ,MAAI,MAAM,QAAQ,EAAE,GAAG;AACrB,QAAI,GAAG,WAAW,GAAG;AACnB,YAAM,IAAI,WAAW,uCAAuC,GAAG,QAAQ;AAAA,IACzE;AACA,QAAI,GAAG;AAAA,EACT,OAAO;AACL,QAAI;AAAA,EACN;AACA,SAAO;AACT;AACA,SAAS,mBAAmB,QAAQ;AAClC,MAAI,MAAM,QAAQ,MAAM,KAAK,MAAM,QAAQ,OAAO,EAAE,GAAG;AACrD,QAAI,OAAO,WAAW,GAAG;AACvB,eAAS;AACT,aAAO,OAAO;AAAA,IAChB,OAAO;AACL,YAAM,IAAI,WAAW,iCAAiC,OAAO,QAAQ;AAAA,IACvE;AAAA,EACF,OAAO;AACL,WAAO;AAAA,EACT;AACF;AAGA,SAAS,qBAAqB,SAAS;AACrC,MAAI,SAAS;AACb,aAAW,UAAU,SAAS;AAC5B,QAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,gBAAU;AAAA,IACZ,OAAO;AACL,gBAAU,OAAO,MAAM,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAAA,IAC/C;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,+BAA+B;AACnC,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,KAAK,QAAQ,WAAW,OAAO,8BAA8B,YAAY,MAAM,aAAa,MAAM;AAC5G,SAAK,QAAQ,SAAS,OAAO,YAAY;AACzC,SAAK,QAAQ,IAAI;AACjB,SAAK,KAAK,sBAAsB;AAChC,WAAO,QAAQ,OAAO,+BAA+B;AACrD,SAAK,eAAe,oBAAoB,IAAI;AAC5C,SAAK,OAAO,oBAAoB,KAAK,YAAY;AACjD,SAAK,aAAa;AAClB,SAAK,aAAa;AAClB,SAAK,MAAM,SAAS,KAAK,KAAK,YAAY,KAAK,MAAM,KAAK,KAAK;AAAA,EACjE;AAAA,EACA,OAAO;AACL,SAAK,kBAAkB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,QAAQ;AACZ,SAAK,kBAAkB;AACvB,qBAAiB,KAAK,KAAK,MAAM;AACjC,QAAI,KAAK,IAAI,OAAO,OAAO,IAAI;AAC7B,WAAK,IAAI,OAAO,MAAM;AACtB,UAAI,KAAK,cAAc,MAAM;AAC3B,aAAK,IAAI,OAAO,KAAK,WAAW,MAAM,KAAK,GAAG,CAAC;AAAA,MACjD;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,UAAU;AACR,SAAK,kBAAkB;AACvB,SAAK,IAAI,QAAQ;AAAA,EACnB;AAAA,EACA,oBAAoB;AAClB,QAAI,KAAK,IAAI,YAAY;AACvB,YAAM,IAAI,MAAM,kBAAkB,KAAK,2BAA2B;AAAA,IACpE;AAAA,EACF;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,SAAK,aAAa;AAClB,SAAK,IAAI,YAAY;AAAA,EACvB;AACF;AACA,SAAS,iBAAiB,GAAG,GAAG;AAC9B,MAAI,EAAE,MAAM,SAAS,MAAM,EAAE,MAAM,SAAS,GAAG;AAC7C,UAAM,IAAI,MAAM,qBAAqB,KAAK,UAAU,EAAE,KAAK,IAAI,UAAU,KAAK,UAAU,EAAE,KAAK,CAAC;AAAA,EAClG;AACF;AACA,SAAS,cAAc,IAAI;AACzB,SAAO,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC;AAC/B;AACA,SAAS,cAAc,oBAAoB;AACzC,qBAAmB,QAAQ,CAAC,qBAAqB;AAC/C,UAAM,YAAY,iBAAiB;AACnC,cAAU,MAAM,iBAAiB,EAAE;AAAA,EACrC,CAAC;AACH;AAGA,IAAI,YAAY,MAAM;AAAA,EACpB,YAAY,MAAM;AAChB,SAAK,QAAQ,KAAK;AAClB,SAAK,QAAQ,KAAK;AAClB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,OAAO,KAAK,MAAM;AAAA,IACzB,OAAO;AACL,WAAK,OAAO,KAAK;AAAA,IACnB;AACA,SAAK,UAAU,KAAK;AACpB,SAAK,UAAU,KAAK;AACpB,SAAK,OAAO,KAAK,QAAQ,CAAC;AAAA,EAC5B;AACF;AACA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,OAAO,OAAO,aAAa,QAAQ,UAAU,MAAM,mBAAmB;AAChF,SAAK,QAAQ;AACb,SAAK,QAAQ;AACb,SAAK,cAAc;AACnB,SAAK,SAAS;AACd,SAAK,WAAW;AAChB,SAAK,oBAAoB;AACzB,SAAK,KAAK,sBAAsB;AAChC,QAAI,QAAQ,MAAM;AAChB,WAAK,eAAe,oBAAoB,IAAI;AAC5C,WAAK,OAAO,oBAAoB,KAAK,YAAY;AAAA,IACnD;AACA,SAAK,OAAO,MAAM;AAAA,EACpB;AACF;AACA,IAAI,cAAc;AAClB,IAAI,OAAO,MAAM;AAAA,EACf,YAAY,MAAM,UAAU;AAC1B,SAAK,WAAW;AAChB,SAAK,KAAK;AACV,SAAK,gBAAgB,KAAK;AAC1B,SAAK,gBAAgB,KAAK;AAC1B,SAAK,cAAc,KAAK;AACxB,SAAK,gBAAgB,KAAK;AAC1B,SAAK,eAAe,KAAK;AACzB,SAAK,gBAAgB,KAAK;AAC1B,SAAK,aAAa,KAAK;AACvB,SAAK,cAAc,KAAK;AACxB,SAAK,cAAc,KAAK;AACxB,SAAK,eAAe,KAAK;AACzB,eAAW,SAAS,KAAK,eAAe;AACtC,UAAI,SAAS,MAAM;AACjB,cAAM,cAAc,KAAK,IAAI;AAAA,MAC/B;AAAA,IACF;AACA,SAAK,cAAc,aAAa,KAAK,IAAI;AAAA,EAC3C;AAAA,EACA,YAAY;AACV,UAAM,eAAe,CAAC;AACtB,eAAW,SAAS,KAAK,eAAe;AACtC,UAAI,SAAS,MAAM;AACjB,qBAAa,KAAK,MAAM,IAAI;AAAA,MAC9B,OAAO;AACL,qBAAa,KAAK,IAAI;AAAA,MACxB;AAAA,IACF;AACA,WAAO;AAAA,MACL,eAAe,KAAK,gBAAgB,KAAK,cAAc,OAAO;AAAA,MAC9D,eAAe;AAAA,MACf,aAAa,KAAK;AAAA,MAClB,eAAe,KAAK;AAAA,IACtB;AAAA,EACF;AACF;AACA,IAAI,eAAe;AACnB,IAAI,QAAQ,cAAc,sBAAsB,aAAa;AAAA,EAC3D,YAAY,OAAO,CAAC,GAAG;AACrB,UAAM;AACN,SAAK,YAAY;AACjB,SAAK,oBAAoB,CAAC;AAC1B,SAAK,YAAY;AACjB,SAAK,KAAK;AACV,SAAK,sBAAsB;AAC3B,SAAK,YAAY;AACjB,SAAK,kBAAkB;AACvB,SAAK,oBAAoB,CAAC;AAC1B,SAAK,uBAAuB,CAAC;AAC7B,SAAK,UAAU,CAAC;AAChB,SAAK,WAAW,CAAC;AACjB,SAAK,SAAS;AACd,SAAK,eAAe,CAAC;AACrB,SAAK,gBAAgB,CAAC;AACtB,QAAI,OAAO,KAAK;AAChB,QAAI,CAAC,MAAM;AACT,YAAM,SAAS,KAAK,aAAa;AACjC,aAAO,YAAY,MAAM,IAAI,MAAM,OAAO,MAAM;AAAA,IAClD;AACA,SAAK,OAAO;AACZ,SAAK,aAAa,KAAK,aAAa,OAAO,OAAO,KAAK;AACvD,QAAI,KAAK,cAAc,QAAQ,KAAK,mBAAmB,MAAM;AAC3D,UAAI;AACJ,UAAI,KAAK,mBAAmB,MAAM;AAChC,0BAAkB,KAAK;AAAA,MACzB,WAAW,KAAK,cAAc,MAAM;AAClC,YAAI,YAAY;AAChB,YAAI,KAAK,aAAa,MAAM;AAC1B,sBAAY,KAAK;AAAA,QACnB;AACA,0BAAkB,CAAC,SAAS,EAAE,OAAO,KAAK,UAAU;AAAA,MACtD;AACA,WAAK,kBAAkB;AACvB,UAAI,QAAQ,KAAK;AACjB,UAAI,SAAS,MAAM;AACjB,gBAAQ,KAAK;AAAA,MACf;AACA,UAAI,SAAS,MAAM;AACjB,gBAAQ;AAAA,MACV;AACA,WAAK,QAAQ;AAAA,IACf;AACA,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,iBAAiB,KAAK;AAAA,IAC7B,OAAO;AACL,WAAK,iBAAiB;AAAA,IACxB;AACA,SAAK,YAAY;AACjB,SAAK,4BAA4B;AAAA,EACnC;AAAA,EACA,OAAO,QAAQ,OAAO,WAAW;AAC/B,WAAO,MAAM,OAAO,SAAS,UAAU,SAAS;AAAA,EAClD;AAAA,EACA,eAAe,WAAW,UAAU;AAClC,QAAI,KAAK,aAAa,WAAW,GAAG;AAClC,YAAM,IAAI,aAAa,2DAA2D,WAAW;AAAA,IAC/F;AACA,QAAI,KAAK,aAAa,UAAU,WAAW;AACzC,YAAM,IAAI,WAAW,gBAAgB,oBAAoB,qCAAqC,KAAK,aAAa,uBAAuB;AAAA,IACzI;AACA,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EACA,WAAW,WAAW;AACpB,WAAO,iBAAiB,KAAK,eAAe,WAAW,OAAO,EAAE,YAAY;AAAA,EAC9E;AAAA,EACA,YAAY,WAAW;AACrB,WAAO,iBAAiB,KAAK,eAAe,WAAW,QAAQ,EAAE,aAAa;AAAA,EAChF;AAAA,EACA,IAAI,QAAQ;AACV,QAAI,KAAK,aAAa,SAAS,GAAG;AAChC,YAAM,IAAI,eAAe,SAAS,KAAK,2HAA2H;AAAA,IACpK,WAAW,KAAK,aAAa,WAAW,GAAG;AACzC,YAAM,IAAI,eAAe,SAAS,KAAK,4CAA4C;AAAA,IACrF;AACA,WAAO,iBAAiB,KAAK,eAAe,GAAG,OAAO,EAAE,YAAY;AAAA,EACtE;AAAA,EACA,IAAI,SAAS;AACX,QAAI,KAAK,aAAa,WAAW,GAAG;AAClC,YAAM,IAAI,eAAe,SAAS,KAAK,4BAA4B;AAAA,IACrE;AACA,QAAI,KAAK,aAAa,SAAS,GAAG;AAChC,YAAM,IAAI,eAAe,SAAS,KAAK,6HAA6H;AAAA,IACtK;AACA,WAAO,iBAAiB,KAAK,eAAe,GAAG,QAAQ,EAAE,aAAa;AAAA,EACxE;AAAA,EACA,IAAI,SAAS;AACX,WAAO,KAAK;AAAA,EACd;AAAA,EACA,kBAAkB;AAChB,WAAO,KAAK,OAAO,IAAI,CAAC,WAAW,OAAO,CAAC;AAAA,EAC7C;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,QAAQ;AACV,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,MAAM,OAAO;AACf,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,SAAK,kBAAkB,QAAQ,CAAC,MAAM,EAAE,YAAY,SAAS;AAC7D,SAAK,aAAa;AAAA,EACpB;AAAA,EACA,IAAI,mBAAmB;AACrB,QAAI,KAAK,YAAY;AACnB,aAAO,KAAK,kBAAkB,OAAO,CAAC,MAAM,EAAE,SAAS;AAAA,IACzD,OAAO;AACL,aAAO,CAAC;AAAA,IACV;AAAA,EACF;AAAA,EACA,IAAI,iBAAiB,SAAS;AAC5B,SAAK,oBAAoB;AAAA,EAC3B;AAAA,EACA,IAAI,sBAAsB;AACxB,QAAI,KAAK,WAAW;AAClB,aAAO,KAAK,kBAAkB,OAAO,CAAC,MAAM,CAAC,EAAE,SAAS,EAAE,OAAO,KAAK,oBAAoB;AAAA,IAC5F,OAAO;AACL,aAAO,KAAK,kBAAkB,OAAO,KAAK,oBAAoB;AAAA,IAChE;AAAA,EACF;AAAA,EACA,IAAI,oBAAoB,SAAS;AAC/B,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,iBAAiB,OAAO,KAAK,mBAAmB;AAAA,EAC9D;AAAA,EACA,IAAI,WAAW;AACb,WAAO,KAAK;AAAA,EACd;AAAA,EACA,cAAc;AACZ,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,sEAAsE;AAAA,IACxF;AAAA,EACF;AAAA,EACA,yBAAyB,QAAQ;AAC/B,aAAS,OAAO,MAAM;AACtB,QAAI,KAAK,aAAa,QAAQ,KAAK,UAAU,WAAW,GAAG;AACzD;AAAA,IACF;AACA,UAAM,YAAY,OAAO,KAAK,SAAS;AACvC,QAAI,OAAO,WAAW,UAAU,QAAQ;AACtC,YAAM,IAAI,WAAW,SAAS,KAAK,gBAAgB,UAAU,kCAAkC,OAAO,yCAAyC,QAAQ;AAAA,IACzJ;AACA,aAAS,aAAa,GAAG,aAAa,OAAO,QAAQ,cAAc;AACjE,YAAM,IAAI,OAAO;AACjB,YAAM,OAAO,UAAU;AACvB,UAAI,QAAQ,MAAM;AAChB;AAAA,MACF;AACA,YAAM,OAAO,EAAE;AACf,UAAI,KAAK,QAAQ,MAAM;AACrB,YAAI,SAAS,KAAK,MAAM;AACtB,gBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,uBAAuB,KAAK,oBAAoB,MAAM;AAAA,QACpI;AAAA,MACF;AACA,UAAI,KAAK,WAAW,MAAM;AACxB,YAAI,OAAO,KAAK,SAAS;AACvB,gBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,2BAA2B,KAAK,uBAAuB,MAAM;AAAA,QAC3I;AAAA,MACF;AACA,UAAI,KAAK,WAAW,MAAM;AACxB,YAAI,OAAO,KAAK,SAAS;AACvB,gBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,2BAA2B,KAAK,uBAAuB,OAAO;AAAA,QAC5I;AAAA,MACF;AACA,UAAI,KAAK,SAAS,MAAM;AACtB,YAAI,EAAE,UAAU,KAAK,OAAO;AAC1B,gBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,yBAAyB,KAAK,sBAAsB,EAAE,QAAQ;AAAA,QAC5I;AAAA,MACF;AACA,UAAI,KAAK,MAAM;AACb,cAAM,SAAS,EAAE;AACjB,mBAAW,OAAO,KAAK,MAAM;AAC3B,gBAAM,OAAO,OAAO,GAAG;AACvB,gBAAM,QAAQ,KAAK,KAAK;AACxB,gBAAM,eAAe,QAAQ,IAAI,OAAO,QAAQ,OAAO,OAAO,SAAS;AACvE,cAAI,SAAS,QAAQ,CAAC,OAAO,IAAI,EAAE,QAAQ,YAAY,MAAM,IAAI;AAC/D,kBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,uBAAuB,qCAAqC,uBAAuB,SAAS;AAAA,UAC1K;AAAA,QACF;AAAA,MACF;AACA,UAAI,KAAK,SAAS,MAAM;AACtB,iBAAS,IAAI,GAAG,IAAI,KAAK,MAAM,QAAQ,EAAE,GAAG;AAC1C,gBAAM,UAAU,KAAK,MAAM;AAC3B,gBAAM,MAAM,EAAE,MAAM;AACpB,cAAI,WAAW,QAAQ,OAAO,MAAM;AAClC,gBAAI,YAAY,KAAK;AACnB,oBAAM,IAAI,WAAW,SAAS,yCAAyC,KAAK,wBAAwB,KAAK,sBAAsB,EAAE,QAAQ;AAAA,YAC3I;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO;AAAA,EACT;AAAA,EACA,eAAe,QAAQ,QAAQ;AAC7B,QAAI,KAAK,aAAa,MAAM;AAC1B,WAAK,UAAU,QAAQ,MAAM;AAAA,IAC/B;AAAA,EACF;AAAA,EACA,YAAY,UAAU;AACpB,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,gBAAgB;AACd,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,MAAM,QAAQ,QAAQ;AACpB,aAAS,UAAU,CAAC;AACpB,SAAK,kBAAkB;AACvB,UAAM,aAAa,OAAO,MAAM;AAChC,QAAI,iBAAiB;AACrB,eAAW,UAAU,YAAY;AAC/B,UAAI,EAAE,kBAAkB,iBAAiB;AACvC,yBAAiB;AACjB;AAAA,MACF;AAAA,IACF;AACA,QAAI,kBAAkB;AACtB,eAAW,UAAU,YAAY;AAC/B,UAAI,kBAAkB,gBAAgB;AACpC,0BAAkB;AAClB;AAAA,MACF;AAAA,IACF;AACA,QAAI,mBAAmB,iBAAiB;AACtC,YAAM,IAAI,WAAW,iEAAiE;AAAA,IACxF;AACA,WAAO,UAAU,KAAK,MAAM,MAAM;AAChC,UAAI,CAAC,KAAK,OAAO;AACf,aAAK,yBAAyB,MAAM;AACpC,cAAM,cAAc,CAAC;AACrB,mBAAW,SAAS,OAAO,MAAM,GAAG;AAClC,sBAAY,KAAK,MAAM,KAAK;AAAA,QAC9B;AACA,aAAK,MAAM,iBAAiB,WAAW,CAAC;AACxC,aAAK,QAAQ;AACb,YAAI,KAAK,gBAAgB;AACvB,eAAK,WAAW,KAAK,cAAc;AAAA,QACrC;AACA,YAAI,KAAK,cAAc,QAAQ,iBAAiB;AAC9C,eAAK,YAAY;AAAA,QACnB;AAAA,MACF;AACA,WAAK,yBAAyB,MAAM;AACpC,UAAI,iBAAiB;AACnB,YAAI,SAAS,KAAK,KAAK,QAAQ,MAAM;AACrC,cAAM,aAAa,OAAO,MAAM;AAChC,cAAM,iBAAiB,CAAC;AACxB,iBAAS,KAAK,YAAY;AACxB,cAAI,WAAW,QAAQ,CAAC,MAAM,IAAI;AAChC,gBAAI,EAAE,MAAM;AAAA,UACd;AACA,yBAAe,KAAK,CAAC;AAAA,QACvB;AACA,iBAAS,iBAAiB,cAAc;AACxC,YAAI,KAAK,uBAAuB,MAAM;AACpC,gBAAM,IAAI,oBAAoB,mFAAmF;AAAA,QACnH;AACA,eAAO;AAAA,MACT,OAAO;AACL,cAAM,aAAa,kBAAkB,MAAM;AAC3C,cAAM,cAAc,KAAK,mBAAmB,UAAU;AACtD,YAAI;AACJ,cAAM,cAAc,iBAAiB,MAAM;AAC3C,aAAK,6BAA6B,MAAM,QAAQ,MAAM,IAAI,WAAW,KAAK,UAAU;AACpF,YAAI,eAAe,QAAQ,YAAY,SAAS,KAAK,MAAM,QAAQ,YAAY,EAAE,GAAG;AAClF,mBAAS,YAAY,IAAI,CAAC,OAAO,UAAU,IAAI,eAAe,aAAa,OAAO,MAAM,OAAO,MAAM,GAAG,QAAQ,KAAK,MAAM,KAAK,CAAC;AAAA,QACnI,OAAO;AACL,mBAAS,IAAI,eAAe,aAAa,aAAa,MAAM,OAAO,MAAM,GAAG,QAAQ,KAAK,IAAI;AAAA,QAC/F;AACA,aAAK,eAAe,QAAQ,QAAQ,MAAM,MAAM,YAAY,aAAa,MAAM;AAC/E,aAAK;AACL,YAAI,KAAK,uBAAuB,MAAM;AACpC,gBAAM,IAAI,oBAAoB,mFAAmF;AAAA,QACnH;AACA,eAAO;AAAA,MACT;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,6BAA6B,YAAY;AACvC,QAAI,KAAK,mBAAmB,MAAM;AAChC;AAAA,IACF,WAAW,WAAW,WAAW,KAAK,gBAAgB,QAAQ;AAC5D,cAAQ,KAAK,iDAAiD,KAAK,UAAU,UAAU,kDAAkD,KAAK,UAAU,KAAK,eAAe,mBAAmB,KAAK,MAAM;AAAA,IAC5M,OAAO;AACL,UAAI,cAAc;AAClB,WAAK,gBAAgB,QAAQ,CAAC,WAAW,MAAM;AAC7C,YAAI,aAAa,QAAQ,WAAW,MAAM,QAAQ,WAAW,OAAO,WAAW;AAC7E,wBAAc;AAAA,QAChB;AAAA,MACF,CAAC;AACD,UAAI,aAAa;AACf,gBAAQ,KAAK,kCAAkC,KAAK,UAAU,UAAU,8CAA8C,KAAK,SAAS,KAAK,UAAU,KAAK,eAAe,GAAG;AAAA,MAC5K;AAAA,IACF;AAAA,EACF;AAAA,EACA,IAAI,cAAc;AAChB,QAAI,KAAK,gBAAgB,QAAQ,KAAK,aAAa,WAAW,GAAG;AAC/D,YAAM,IAAI,eAAe,aAAa,KAAK,kEAAkE;AAAA,IAC/G;AACA,UAAM,kBAAkB,CAAC;AACzB,eAAW,QAAQ,KAAK,cAAc;AACpC,YAAM,cAAc,KAAK,UAAU,KAAK,YAAY;AACpD,UAAI,gBAAgB,QAAQ,WAAW,MAAM,IAAI;AAC/C,wBAAgB,KAAK,WAAW;AAAA,MAClC;AAAA,IACF;AACA,QAAI,gBAAgB,WAAW,GAAG;AAChC,YAAM,eAAe,KAAK,aAAa,GAAG;AAC1C,UAAI,MAAM,QAAQ,YAAY,KAAK,MAAM,QAAQ,aAAa,EAAE,KAAK,aAAa,WAAW,GAAG;AAC9F,eAAO,aAAa;AAAA,MACtB,OAAO;AACL,eAAO;AAAA,MACT;AAAA,IACF,OAAO;AACL,YAAM,IAAI,eAAe,aAAa,KAAK,gIAAgI;AAAA,IAC7K;AAAA,EACF;AAAA,EACA,cAAc;AACZ,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,aAAa,sCAAsC,KAAK,yFAAyF;AAAA,IAC7J;AACA,WAAO,qBAAqB,KAAK,OAAO;AAAA,EAC1C;AAAA,EACA,MAAM,YAAY;AAChB,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,WAAW,gBAAgB,OAAO;AAChC,WAAO,cAAc,gBAAgB,KAAK,mBAAmB,KAAK,OAAO;AAAA,EAC3E;AAAA,EACA,WAAW,SAAS;AAClB,SAAK,MAAM;AACT,YAAM,SAAS,KAAK;AACpB,UAAI,OAAO,WAAW,QAAQ,QAAQ;AACpC,cAAM,IAAI,WAAW,4CAA4C,KAAK,sCAAsC,QAAQ,uCAAuC,OAAO,qCAAqC,YAAY;AAAA,MACrN;AACA,UAAI,OAAO,WAAW,GAAG;AACvB;AAAA,MACF;AACA,YAAM,oBAAoB,CAAC;AAC3B,YAAM,cAAc,cAAc,MAAM;AACxC,eAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,cAAM,KAAK,YAAY;AACvB,cAAM,KAAK,OAAO;AAClB,cAAM,IAAI,QAAQ;AAClB,YAAI,CAAC,aAAa,YAAY,GAAG,OAAO,EAAE,KAAK,GAAG;AAChD,gBAAM,IAAI,WAAW,sBAAsB,GAAG,mDAAmD,EAAE,OAAO;AAAA,QAC5G;AACA,0BAAkB,KAAK,CAAC,IAAI,CAAC,CAAC;AAAA,MAChC;AACA,oBAAc,iBAAiB;AAAA,IACjC,CAAC;AAAA,EACH;AAAA,EACA,UAAU,MAAM,OAAO,OAAO,aAAa,aAAa,WAAW,YAAY,oBAAoB;AACjG,QAAI,KAAK,kBAAkB,QAAQ,IAAI,MAAM,IAAI;AAC/C,YAAM,IAAI,WAAW,yBAAyB,kBAAkB,KAAK,MAAM;AAAA,IAC7E;AACA,SAAK,kBAAkB,KAAK,IAAI;AAChC,QAAI,SAAS,MAAM;AACjB,cAAQ;AAAA,IACV;AACA,QAAI,KAAK,2BAA2B;AAClC,oBAAc,sBAAsB,OAAO,mBAAmB,IAAI,eAAe,OAAO;AAAA,IAC1F;AACA,UAAM,YAAY,YAAY,MAAM,OAAO,KAAK;AAChD,UAAM,SAAS,IAAI,cAAc,WAAW,OAAO,MAAM,WAAW,UAAU;AAC9E,cAAU,QAAQ;AAClB,QAAI,eAAe,MAAM;AACvB,WAAK,QAAQ,MAAM,YAAY,MAAM,OAAO,KAAK,CAAC,CAAC;AAAA,IACrD;AACA,QAAI,aAAa,MAAM;AACrB,kBAAY;AAAA,IACd;AACA,QAAI,WAAW;AACb,WAAK,kBAAkB,KAAK,MAAM;AAAA,IACpC,OAAO;AACL,WAAK,qBAAqB,KAAK,MAAM;AAAA,IACvC;AACA,WAAO;AAAA,EACT;AAAA,EACA,6BAA6B,OAAO;AAClC,SAAK,4BAA4B;AAAA,EACnC;AAAA,EACA,QAAQ,SAAS;AACf,QAAI,WAAW,QAAQ,MAAM,QAAQ,OAAO,KAAK,QAAQ,WAAW,GAAG;AACrE;AAAA,IACF;AACA,cAAU,OAAO,OAAO;AACxB,QAAI,KAAK,YAAY,UAAU,KAAK,YAAY,MAAM;AACpD,WAAK,OAAO,KAAK,GAAG,OAAO;AAAA,IAC7B;AAAA,EACF;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,QAAI,CAAC,KAAK,iBAAiB;AACzB,UAAI,QAAQ,MAAM;AAChB,YAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,eAAK,QAAQ,CAAC,gBAAgB;AAC5B,gBAAI,eAAe,MAAM;AACvB,oBAAM,IAAI,UAAU,SAAS,KAAK,6DAA6D;AAAA,YACjG;AAAA,UACF,CAAC;AAAA,QACH,OAAO;AACL,gBAAM,IAAI,UAAU,SAAS,KAAK,6DAA6D;AAAA,QACjG;AAAA,MACF;AACA,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAAA,EACA,eAAe,cAAc,eAAe,YAAY,aAAa,aAAa,cAAc,SAAS,MAAM;AAC7G,UAAM,kBAAkB,OAAO,YAAY;AAC3C,oBAAgB,OAAO,aAAa;AACpC,iBAAa,OAAO,UAAU;AAC9B,kBAAc,OAAO,WAAW;AAChC,kBAAc,mBAAmB,WAAW;AAC5C,mBAAe,mBAAmB,YAAY;AAC9C,UAAM,gBAAgB,CAAC;AACvB,UAAM,cAAc,CAAC;AACrB,UAAM,gBAAgB,CAAC;AACvB,eAAW,KAAK,iBAAiB;AAC/B,oBAAc,KAAK,EAAE,WAAW;AAChC,kBAAY,KAAK,EAAE,SAAS;AAC5B,oBAAc,KAAK,EAAE,WAAW;AAAA,IAClC;AACA,QAAI,KAAK;AAAA,MACP,eAAe;AAAA,MACf;AAAA,MACA;AAAA,MACA;AAAA,MACA,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,GAAG,MAAM;AACT,aAAS,IAAI,GAAG,IAAI,cAAc,QAAQ,KAAK;AAC7C,oBAAc,GAAG,cAAc;AAC/B,oBAAc,GAAG,YAAY,KAAK,aAAa,SAAS;AACxD,oBAAc,GAAG,cAAc;AAAA,IACjC;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,MAAM,KAAK,MAAM,WAAW,KAAK,UAAU;AAC5D,QAAI,KAAK,mBAAmB,MAAM;AAChC,aAAO,qBAAqB,KAAK;AAAA,IACnC;AACA,QAAI,KAAK,SAAS,MAAM;AACtB,aAAO,WAAW,KAAK;AAAA,IACzB;AACA,WAAO;AAAA,EACT;AAAA,EACA,iBAAiB;AACf,SAAK,QAAQ,QAAQ,CAAC,WAAW,OAAO,QAAQ,CAAC;AACjD,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EACA,oBAAoB;AAClB,QAAI,KAAK,cAAc,GAAG;AACxB,YAAM,IAAI,MAAM,UAAU,KAAK,4BAA4B;AAAA,IAC7D;AAAA,EACF;AAAA,EACA,UAAU;AACR,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,wBAAwB,KAAK,yCAAyC;AAAA,IACxF;AACA,QAAI,KAAK,cAAc,MAAM;AAC3B,YAAM,IAAI,MAAM,wBAAwB,KAAK,wCAAwC;AAAA,IACvF;AACA,SAAK,kBAAkB;AACvB,QAAI,uBAAuB;AAC3B,QAAI,EAAE,KAAK,cAAc,GAAG;AAC1B,6BAAuB,KAAK,eAAe;AAAA,IAC7C;AACA,WAAO,EAAE,sBAAsB,KAAK,WAAW,qBAAqB;AAAA,EACtE;AACF;AACA,SAAS,kBAAkB,cAAc;AACvC,iBAAe,OAAO,YAAY;AAClC,QAAM,SAAS,CAAC;AAChB,aAAW,KAAK,cAAc;AAC5B,WAAO,KAAK,EAAE,KAAK;AAAA,EACrB;AACA,SAAO,iBAAiB,MAAM;AAChC;AACA,SAAS,iBAAiB,cAAc;AACtC,SAAO;AACT;AACA,SAAS,gBAAgB,SAAS,OAAO,WAAW;AAClD,MAAI,SAAS,QAAQ,aAAa,QAAQ,YAAY,GAAG;AACvD,YAAQ,QAAQ;AAChB,gBAAY,QAAQ;AAAA,EACtB;AACA,MAAI,MAAM,aAAa,WAAW,GAAG;AACnC,WAAO,CAAC,OAAO;AAAA,EACjB,OAAO;AACL,UAAM,OAAO,MAAM,aAAa;AAChC,QAAI,KAAK,cAAc,WAAW,GAAG;AACnC,aAAO,KAAK;AAAA,IACd,OAAO;AACL,YAAM,gBAAgB,CAAC;AACvB,eAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,KAAK;AAClD,cAAM,IAAI,KAAK,aAAa;AAC5B,cAAM,SAAS,KAAK,cAAc;AAClC,cAAM,aAAa,KAAK,YAAY;AACpC,cAAM,kBAAkB,gBAAgB,GAAG,QAAQ,UAAU;AAC7D,mBAAW,MAAM,iBAAiB;AAChC,cAAI,cAAc,QAAQ,EAAE,MAAM,IAAI;AACpC,0BAAc,KAAK,EAAE;AAAA,UACvB;AAAA,QACF;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAGA,IAAI,aAAa,cAAc,MAAM;AAAA,EACnC,YAAY,MAAM;AAChB,UAAM;AAAA,MACJ,OAAO,KAAK;AAAA,MACZ,MAAM,KAAK,QAAQ,OAAO,KAAK,OAAO,OAAO,OAAO,EAAE,SAAS;AAAA,IACjE,CAAC;AACD,QAAI,KAAK,aAAa,MAAM;AAC1B,WAAK,YAAY;AAAA,IACnB;AACA,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,SAAS;AAAA,IAChB;AACA,SAAK,YAAY;AACjB,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK;AACnB,QAAI,KAAK,cAAc,QAAQ,KAAK,mBAAmB,MAAM;AAC3D,YAAM,IAAI,WAAW,mGAAmG;AAAA,IAC1H;AACA,QAAI,kBAAkB,KAAK;AAC3B,QAAI,mBAAmB,MAAM;AAC3B,UAAI,KAAK,cAAc,MAAM;AAC3B,cAAM,IAAI,WAAW,+EAA+E;AAAA,MACtG,OAAO;AACL,0BAAkB,CAAC,KAAK,SAAS,EAAE,OAAO,KAAK,UAAU;AAAA,MAC3D;AAAA,IACF,OAAO;AACL,UAAI,KAAK,aAAa,MAAM;AAC1B,cAAM,IAAI,WAAW,uFAAuF;AAAA,MAC9G;AAAA,IACF;AACA,UAAM,QAAQ,KAAK,SAAS;AAC5B,SAAK,kBAAkB;AACvB,SAAK,QAAQ;AACb,SAAK,YAAY,CAAC,EAAE,OAAO,gBAAgB,CAAC;AAC5C,UAAM,cAAc,IAAI,eAAe,KAAK,OAAO,KAAK,iBAAiB,MAAM,CAAC,GAAG,CAAC,GAAG,KAAK,IAAI;AAChG,gBAAY,YAAY;AACxB,gBAAY,cAAc;AAC1B,QAAI,KAAK;AAAA,MACP,eAAe;AAAA,MACf,eAAe,CAAC;AAAA,MAChB,aAAa,CAAC;AAAA,MACd,eAAe,CAAC;AAAA,MAChB,cAAc,CAAC,WAAW;AAAA,MAC1B,eAAe,CAAC,WAAW;AAAA,MAC3B,YAAY,CAAC,IAAI;AAAA,MACjB,aAAa,CAAC,IAAI;AAAA,MAClB,aAAa,CAAC,eAAe;AAAA,MAC7B,cAAc,CAAC,eAAe;AAAA,IAChC,CAAC;AAAA,EACH;AAAA,EACA,MAAM,QAAQ,QAAQ;AACpB,UAAM,IAAI,WAAW,6EAA6E,KAAK,MAAM;AAAA,EAC/G;AAAA,EACA,UAAU;AACR,WAAO,EAAE,sBAAsB,KAAK,WAAW,sBAAsB,EAAE;AAAA,EACzE;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,iBAAiB,KAAK;AAAA,MACtB,OAAO,KAAK;AAAA,MACZ,QAAQ,KAAK;AAAA,MACb,MAAM,KAAK;AAAA,IACb;AAAA,EACF;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAC9C,SAAS,MAAM,QAAQ;AACrB,MAAI,OAAO,cAAc,QAAQ,OAAO,SAAS,MAAM;AACrD,UAAM,IAAI,MAAM,8HAA8H;AAAA,EAChJ;AACA,MAAI,OAAO,cAAc,QAAQ,OAAO,SAAS,MAAM;AACrD,UAAM,IAAI,WAAW,kFAAkF;AAAA,EACzG;AACA,MAAI,aAAa,OAAO;AACxB,MAAI,OAAO,SAAS,QAAQ,cAAc,MAAM;AAC9C,iBAAa,CAAC,IAAI,EAAE,OAAO,OAAO,KAAK;AAAA,EACzC;AACA,MAAI,QAAQ,OAAO;AACnB,MAAI,SAAS,MAAM;AACjB,YAAQ;AAAA,EACV;AACA,QAAM,cAAc,IAAI,WAAW;AAAA,IACjC,iBAAiB;AAAA,IACjB,MAAM,OAAO;AAAA,IACb;AAAA,IACA,QAAQ,OAAO;AAAA,EACjB,CAAC;AACD,QAAM,UAAU,YAAY,aAAa,GAAG;AAC5C,SAAO,QAAQ;AACjB;AAGA,SAAS,wBAAwB,KAAK,KAAK;AACzC,MAAI,IAAI,SAAS,QAAQ,IAAI,UAAU,IAAI,OAAO;AAChD,WAAO;AAAA,EACT;AACA,MAAI;AACF,WAAO,KAAK,KAAK,IAAI,KAAK;AAAA,EAC5B,SAAS,KAAP;AACA,UAAM,IAAI,WAAW,0BAA0B,IAAI,mDAAmD,IAAI,UAAU,IAAI,SAAS;AAAA,EACnI;AACF;AACA,IAAI,WAAW,MAAM;AAAA,EACnB,YAAY,OAAO;AACjB,SAAK,WAAW,CAAC;AACjB,SAAK,UAAU,CAAC;AAChB,SAAK,UAAU,CAAC;AAChB,QAAI,iBAAiB,UAAU;AAC7B,iBAAW,MAAM,MAAM,UAAU;AAC/B,aAAK,SAAS,MAAM,MAAM,SAAS;AACnC,YAAI,MAAM,MAAM,SAAS;AACvB,eAAK,QAAQ,MAAM,MAAM,QAAQ;AAAA,QACnC;AAAA,MACF;AAAA,IACF,OAAO;AACL,UAAI,SAAS,MAAM;AACjB;AAAA,MACF;AACA,iBAAW,QAAQ,OAAO;AACxB,aAAK,IAAI,KAAK,KAAK,KAAK,KAAK;AAAA,MAC/B;AAAA,IACF;AAAA,EACF;AAAA,EACA,IAAI,KAAK,OAAO,MAAM;AACpB,QAAI,KAAK,SAAS,IAAI,OAAO,MAAM;AACjC,WAAK,SAAS,IAAI,MAAM,wBAAwB,KAAK,KAAK;AAC1D,WAAK,QAAQ,IAAI,QAAQ,IAAI;AAC7B,UAAI,QAAQ,MAAM;AAChB,aAAK,QAAQ,IAAI,MAAM;AAAA,MACzB;AAAA,IACF,OAAO;AACL,YAAM,IAAI,WAAW,uBAAuB,IAAI,YAAY,IAAI,IAAI;AAAA,IACtE;AACA,WAAO;AAAA,EACT;AAAA,EACA,QAAQ,MAAM;AACZ,SAAK,IAAI,KAAK,KAAK,KAAK,KAAK;AAAA,EAC/B;AAAA,EACA,OAAO,KAAK;AACV,WAAO,KAAK,SAAS,IAAI,OAAO;AAAA,EAClC;AAAA,EACA,QAAQ;AACN,WAAO,OAAO,KAAK,KAAK,OAAO;AAAA,EACjC;AAAA,EACA,SAAS,KAAK;AACZ,QAAI,eAAe,gBAAgB;AACjC,UAAI,KAAK,SAAS,IAAI,OAAO,MAAM;AACjC,cAAM,IAAI,WAAW,oBAAoB,IAAI,MAAM;AAAA,MACrD,OAAO;AACL,eAAO,KAAK,SAAS,IAAI;AAAA,MAC3B;AAAA,IACF,OAAO;AACL,YAAM,KAAK,KAAK,QAAQ;AACxB,UAAI,MAAM,MAAM;AACd,cAAM,IAAI,WAAW,yCAAyC,KAAK;AAAA,MACrE;AACA,aAAO,KAAK,SAAS;AAAA,IACvB;AAAA,EACF;AAAA,EACA,QAAQ,KAAK;AACX,QAAI,eAAe,gBAAgB;AACjC,UAAI,KAAK,SAAS,IAAI,OAAO,MAAM;AACjC,cAAM,IAAI,WAAW,oBAAoB,IAAI,MAAM;AAAA,MACrD,OAAO;AACL,eAAO,KAAK,QAAQ,IAAI;AAAA,MAC1B;AAAA,IACF,OAAO;AACL,YAAM,KAAK,KAAK,QAAQ;AACxB,UAAI,MAAM,MAAM;AACd,cAAM,IAAI,WAAW,yCAAyC,KAAK;AAAA,MACrE;AACA,aAAO,KAAK,QAAQ;AAAA,IACtB;AAAA,EACF;AAAA,EACA,eAAe;AACb,QAAI,KAAK,WAAW,MAAM;AACxB,cAAQ,KAAK,OAAO;AAAA,IACtB;AAAA,EACF;AACF;AACA,IAAI,eAAe,IAAI,SAAS;AAChC,IAAI,wBAAwB,IAAI,SAAS;AACzC,SAAS,sBAAsB,YAAY;AACzC,MAAI,gBAAgB,MAAM;AACxB,iBAAa,cAAc,UAAU;AAAA,EACvC;AACA,MAAI,yBAAyB,MAAM;AACjC,0BAAsB,cAAc,UAAU;AAAA,EAChD;AACF;AACA,SAAS,QAAQ,SAAS,UAAU,QAAQ,OAAO;AACjD,QAAM,WAAW,UAAU,OAAO,QAAQ,OAAO;AACjD,QAAM,eAAe,MAAM,QAAQ,OAAO;AAC1C,QAAM,aAAa,eAAe,UAAU,CAAC,OAAO;AACpD,QAAM,cAAc,WAAW,IAAI,CAAC,MAAM,EAAE,IAAI;AAChD,QAAM,eAAe,CAAC;AACtB,QAAM,YAAY,SAAS,MAAM;AACjC,aAAW,cAAc,aAAa;AACpC,QAAI,UAAU,QAAQ,UAAU,MAAM,IAAI;AACxC,mBAAa,KAAK,SAAS,SAAS,UAAU,CAAC;AAAA,IACjD,OAAO;AACL,mBAAa,KAAK,IAAI;AAAA,IACxB;AAAA,EACF;AACA,MAAI,SAAS,MAAM;AACjB,UAAM,gBAAgB;AACtB,UAAM,gBAAgB;AAAA,EACxB;AACA,QAAM,kBAAkB,YAAY,KAAK,GAAG,IAAI,MAAM,SAAS,MAAM,EAAE,KAAK,EAAE,KAAK,GAAG;AACtF,MAAI,SAAS,aAAa,IAAI,eAAe;AAC7C,MAAI;AACJ,MAAI,UAAU,MAAM;AAClB,UAAM,MAAM,qCAAqC,YAAY,QAAQ;AACrE,aAAS,IAAI;AACb,sBAAkB,IAAI;AACtB,iBAAa,IAAI,iBAAiB,MAAM;AACxC,0BAAsB,IAAI,iBAAiB,eAAe;AAAA,EAC5D;AACA,oBAAkB,CAAC;AACnB,MAAI,CAAC,UAAU;AACb,WAAO,OAAO,iBAAiB,sBAAsB,IAAI,eAAe,CAAC;AAAA,EAC3E;AACA,QAAM,mBAAmB,IAAI,SAAS,QAAQ;AAC9C,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,QAAI,SAAS,MAAM;AACjB,YAAM,aAAa,OAAO,EAAE;AAC5B,UAAI,aAAa,MAAM,eAAe;AACpC,cAAM,gBAAgB;AAAA,MACxB;AACA,UAAI,aAAa,MAAM,eAAe;AACpC,cAAM,gBAAgB;AAAA,MACxB;AAAA,IACF;AACA,UAAM,WAAW,OAAO;AACxB,UAAM,WAAW,SAAS;AAC1B,QAAI,oBAAoB,YAAY;AAClC;AAAA,IACF;AACA,UAAM,cAAc,CAAC;AACrB,UAAM,aAAa,CAAC;AACpB,UAAM,mBAAmB,CAAC;AAC1B,QAAI,aAAa;AACjB,eAAW,UAAU,SAAS,QAAQ;AACpC,YAAM,QAAQ,iBAAiB,SAAS,MAAM;AAC9C,YAAM,OAAO,iBAAiB,QAAQ,MAAM;AAC5C,kBAAY,KAAK,KAAK;AACtB,iBAAW,KAAK,IAAI;AACpB,UAAI,QAAQ,MAAM;AAChB,qBAAa;AAAA,MACf;AACA,UAAI,CAAC,UAAU;AACb,wBAAgB,OAAO;AACvB,YAAI,gBAAgB,OAAO,UAAU,KAAK,CAAC,SAAS,OAAO,MAAM,KAAK,YAAY,QAAQ,OAAO,IAAI,MAAM,MAAM,CAAC,MAAM,cAAc,OAAO,YAAY,aAAa,MAAM;AAC1K,2BAAiB,KAAK,KAAK;AAAA,QAC7B;AAAA,MACF;AAAA,IACF;AACA,QAAI,YAAY;AACd,eAAS,UAAU,CAAC;AACpB,aAAO,UAAU,WAAW;AAAA,IAC9B;AACA,UAAM,gBAAgB,OAAO,SAAS,MAAM,aAAa,MAAM,CAAC;AAChE,QAAI,aAAa;AACjB,QAAI,SAAS,iBAAiB;AAC5B,mBAAa,SAAS,YAAY,aAAa,UAAU;AAAA,IAC3D;AACA,UAAM,eAAe,eAAe,QAAQ;AAC5C,UAAM,wBAAwB,MAAM,QAAQ,YAAY,IAAI,eAAe,CAAC,YAAY;AACxF,aAAS,KAAK,GAAG,KAAK,sBAAsB,QAAQ,EAAE,IAAI;AACxD,UAAI,CAAC,iBAAiB,OAAO,sBAAsB,GAAG,GAAG;AACvD,yBAAiB,IAAI,sBAAsB,KAAK,cAAc,KAAK,MAAM,QAAQ,UAAU,IAAI,WAAW,KAAK,UAAU;AAAA,MAC3H;AACA,YAAM,QAAQ,YAAY,QAAQ,sBAAsB,IAAI,IAAI;AAChE,UAAI,UAAU,IAAI;AAChB,qBAAa,SAAS,cAAc;AAAA,MACtC;AAAA,IACF;AACA,QAAI,CAAC,UAAU;AACb,cAAQ,gBAAgB;AAAA,IAC1B;AAAA,EACF;AACA,mBAAiB,aAAa;AAC9B,SAAO,eAAe,eAAe,aAAa;AACpD;AACA,SAAS,qCAAqC,SAAS,UAAU;AAC/D,eAAa,OAAO,WAAW,QAAQ,QAAQ,SAAS,GAAG,MAAM,uCAAuC;AACxG,MAAI,cAAc,CAAC;AACnB,MAAI,oBAAoB,CAAC;AACzB,MAAI,QAAQ,WAAW,GAAG;AACxB,UAAM,MAAM,gDAAgD,QAAQ,IAAI,QAAQ;AAChF,kBAAc,IAAI;AAClB,wBAAoB,IAAI;AAAA,EAC1B,OAAO;AACL,UAAM,UAA0B,oBAAI,IAAI;AACxC,eAAW,UAAU,SAAS;AAC5B,YAAM,EAAE,QAAQ,aAAa,IAAI,gDAAgD,QAAQ,QAAQ;AACjG,iBAAW,kBAAkB,QAAQ;AACnC,YAAI,CAAC,QAAQ,IAAI,eAAe,IAAI,GAAG;AACrC,sBAAY,KAAK,cAAc;AAC/B,kBAAQ,IAAI,eAAe,IAAI;AAAA,QACjC;AAAA,MACF;AACA,iBAAW,QAAQ,cAAc;AAC/B,YAAI,kBAAkB,SAAS,MAAM;AACnC,4BAAkB,QAAwB,oBAAI,IAAI;AAAA,QACpD;AACA,qBAAa,MAAM,QAAQ,CAAC,cAAc,kBAAkB,MAAM,IAAI,SAAS,CAAC;AAAA,MAClF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,iBAAiB,oBAAoB,iBAAiB;AAAA,EACxD;AACF;AACA,SAAS,oBAAoB,cAAc;AACzC,QAAM,kBAAkB,CAAC;AACzB,aAAW,QAAQ,cAAc;AAC/B,oBAAgB,QAAQ,aAAa,MAAM;AAAA,EAC7C;AACA,SAAO;AACT;AACA,SAAS,gDAAgD,QAAQ,UAAU;AACzE,QAAM,UAA0B,oBAAI,IAAI;AACxC,QAAM,SAAS,CAAC;AAChB,QAAM,eAAe,CAAC;AACtB,aAAW,OAAO,SAAS,MAAM,GAAG;AAClC,YAAQ,IAAI,GAAG;AAAA,EACjB;AACA,QAAM,SAAS,CAAC;AAChB,QAAM,QAAQ,CAAC;AACf,SAAO,KAAK,MAAM;AAClB,SAAO,OAAO,SAAS,GAAG;AACxB,UAAM,MAAM,OAAO,OAAO,SAAS;AACnC,QAAI,QAAQ,IAAI,IAAI,IAAI,GAAG;AACzB,aAAO,IAAI;AACX;AAAA,IACF;AACA,UAAM,cAAc,MAAM,MAAM,SAAS,OAAO,OAAO,SAAS;AAChE,QAAI,IAAI,OAAO,WAAW,KAAK,aAAa;AAC1C,aAAO,IAAI;AACX,aAAO,KAAK,GAAG;AACf,cAAQ,IAAI,IAAI,IAAI;AACpB,UAAI,aAAa;AACf,cAAM,IAAI;AAAA,MACZ;AAAA,IACF,OAAO;AACL,YAAM,KAAK,OAAO,SAAS,CAAC;AAC5B,iBAAW,UAAU,IAAI,QAAQ;AAC/B,YAAI,aAAa,OAAO,SAAS,MAAM;AACrC,uBAAa,OAAO,QAAwB,oBAAI,IAAI;AAAA,QACtD;AACA,qBAAa,OAAO,MAAM,IAAI,IAAI,IAAI;AACtC,YAAI,QAAQ,IAAI,OAAO,IAAI,GAAG;AAC5B;AAAA,QACF;AACA,eAAO,KAAK,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AACA,SAAO,EAAE,QAAQ,aAAa;AAChC;AACA,SAAS,eAAe,QAAQ;AAC9B,MAAI;AACJ,MAAI,OAAO,YAAY,aAAa,WAAW,GAAG;AAChD,mBAAe,OAAO,YAAY;AAAA,EACpC,OAAO;AACL,QAAI,YAAY;AAChB,aAAS,IAAI,GAAG,IAAI,OAAO,YAAY,aAAa,QAAQ,EAAE,GAAG;AAC/D,iBAAW,gBAAgB,OAAO,YAAY,aAAa,GAAG,eAAe;AAC3E,YAAI,aAAa,OAAO,OAAO,IAAI;AACjC,sBAAY;AACZ;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,mBAAe,OAAO,YAAY,YAAY,SAAS;AAAA,EACzD;AACA,SAAO;AACT;AAGA,IAAI,OAAO,IAAI;AACf,KAAK,aAAa,sCAAsC,MAAM,KAAK,qBAAqB;AAGxF,IAAI,8BAA8B,CAAC;AACnCJ,UAAS,6BAA6B;AAAA,EACpC,SAAS,MAAM;AAAA,EACf,YAAY,MAAM;AAAA,EAClB,QAAQ,MAAM;AAAA,EACd,UAAU,MAAM;AAClB,CAAC;AAGD,SAAS,YAAY,GAAG,MAAM;AAC5B,SAAO,KAAK,MAAM,KAAK,KAAK,IAAI,GAAG,CAAC,GAAG,MAAM,IAAI,CAAC,CAAC;AACrD;AACA,IAAI,aAAa,cAAc,sBAAsB,aAAa;AAAA,EAChE,YAAY;AACV,WAAO,CAAC;AAAA,EACV;AACF;AACA,IAAI,UAAU,cAAc,WAAW;AAAA,EACrC,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,kBAAkB;AACvB,SAAK,cAAc;AACnB,SAAK,WAAW,KAAK,YAAY,OAAO,KAAK,WAAW,KAAK;AAC7D,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,KAAK;AAAA,EACnD;AAAA,EACA,MAAM,GAAG;AACP,WAAO,KAAK,MAAM;AAChB,YAAM,QAAQ,YAAY,GAAG,KAAK,IAAI;AACtC,YAAM,UAAU,YAAY,OAAO,GAAG,KAAK,QAAQ;AACnD,aAAO,IAAI,GAAG,IAAI,SAAS,KAAK,QAAQ,GAAG,KAAK,CAAC,CAAC;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,WAAO,EAAE,UAAU,KAAK,UAAU,MAAM,KAAK,KAAK;AAAA,EACpD;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,WAAW,cAAc,WAAW;AAAA,EACtC,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,cAAc;AACnB,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,KAAK;AAAA,EACnD;AAAA,EACA,MAAM,GAAG;AACP,WAAO,KAAK,MAAM,IAAI,GAAG,KAAK,QAAQ,GAAG,YAAY,GAAG,KAAK,IAAI,CAAC,CAAC,CAAC;AAAA,EACtE;AAAA,EACA,YAAY;AACV,WAAO,EAAE,MAAM,KAAK,KAAK;AAAA,EAC3B;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,SAAS,cAAc,WAAW;AAAA,EACpC,MAAM,GAAG;AACP,WAAO,KAAK,CAAC;AAAA,EACf;AACF;AACA,OAAO,YAAY;AACnB,sBAAsB,cAAc,MAAM;AAC1C,IAAI,aAAa,cAAc,WAAW;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AACvB,SAAK,cAAc;AACnB,SAAK,cAAc;AACnB,SAAK,WAAW,KAAK,YAAY,OAAO,KAAK,WAAW,KAAK;AAC7D,SAAK,WAAW,KAAK,YAAY,OAAO,KAAK,WAAW,KAAK;AAC7D,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,KAAK;AACjD,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,KAAK;AAAA,EACnD;AAAA,EACA,MAAM,GAAG;AACP,WAAO,KAAK,MAAM;AAChB,YAAM,QAAQ,YAAY,GAAG,KAAK,IAAI;AACtC,YAAM,UAAU,KAAK,IAAI,KAAK,MAAM,YAAY,OAAO,KAAK,UAAU,KAAK,QAAQ,CAAC,GAAG,IAAI,IAAI,KAAK,MAAM,KAAK,CAAC;AAChH,aAAO,IAAI,GAAG,IAAI,SAAS,KAAK,QAAQ,GAAG,KAAK,CAAC,CAAC;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,WAAO;AAAA,MACL,UAAU,KAAK;AAAA,MACf,UAAU,KAAK;AAAA,MACf,MAAM,KAAK;AAAA,MACX,MAAM,KAAK;AAAA,IACb;AAAA,EACF;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAC9C,IAAI,4CAA4C;AAAA,EAC9C,WAAW;AAAA,EACX,cAAc;AAAA,EACd,UAAU;AAAA,EACV,YAAY;AACd;AACA,SAAS,oBAAoB,YAAY;AACvC,SAAO,qBAAqB,UAAU;AACxC;AACA,SAAS,sBAAsB,QAAQ,gBAAgB,CAAC,GAAG;AACzD,SAAO,uBAAuB,QAAQ,sBAAsB,iBAAiB,OAAO,EAAE,cAAc,eAAe,YAAY;AACjI;AACA,SAAS,cAAc,YAAY;AACjC,MAAI,cAAc,MAAM;AACtB,WAAO;AAAA,EACT;AACA,MAAI,OAAO,eAAe,UAAU;AAClC,UAAM,YAAY,cAAc,4CAA4C,0CAA0C,cAAc;AACpI,UAAM,SAAS,EAAE,WAAW,QAAQ,CAAC,EAAE;AACvC,WAAO,sBAAsB,MAAM;AAAA,EACrC,WAAW,sBAAsB,YAAY;AAC3C,WAAO;AAAA,EACT,OAAO;AACL,WAAO,sBAAsB,UAAU;AAAA,EACzC;AACF;AAGA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,SAAS;AAChB,SAAO,IAAI,OAAO;AACpB;AACA,SAAS,WAAW,QAAQ;AAC1B,SAAO,IAAI,WAAW,MAAM;AAC9B;AAGA,IAAI,+BAA+B,CAAC;AACpCA,UAAS,8BAA8B;AAAA,EACrC,UAAU,MAAM;AAAA,EAChB,cAAc,MAAM;AAAA,EACpB,eAAe,MAAM;AAAA,EACrB,UAAU,MAAM;AAAA,EAChB,WAAW,MAAM;AAAA,EACjB,UAAU,MAAM;AAAA,EAChB,aAAa,MAAM;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,MAAM,MAAM;AAAA,EACZ,YAAY,MAAM;AAAA,EAClB,cAAc,MAAM;AAAA,EACpB,eAAe,MAAM;AAAA,EACrB,iBAAiB,MAAM;AAAA,EACvB,iBAAiB,MAAM;AAAA,EACvB,OAAO,MAAM;AACf,CAAC;AACD,SAAS,SAAS;AAChB,SAAO,IAAI,MAAM;AACnB;AACA,SAAS,QAAQ;AACf,SAAO,IAAI,KAAK;AAClB;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,eAAe,MAAM;AAC5B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,gBAAgB,QAAQ;AAC/B,SAAO,IAAI,gBAAgB,MAAM;AACnC;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,YAAY,MAAM;AACzB,SAAO,IAAI,YAAY,IAAI;AAC7B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,IAAI,WAAW,IAAI;AAC5B;AAGA,IAAI,yBAAyB,CAAC;AAC9BA,UAAS,wBAAwB;AAAA,EAC/B,OAAO,MAAM;AAAA,EACb,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,YAAY,MAAM;AAAA,EAClB,KAAK,MAAM;AAAA,EACX,cAAc,MAAM;AAAA,EACpB,SAAS,MAAM;AAAA,EACf,kBAAkB,MAAM;AAAA,EACxB,kBAAkB,MAAM;AAAA,EACxB,kBAAkB,MAAM;AAAA,EACxB,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,oBAAoB,MAAM;AAAA,EAC1B,eAAe,MAAM;AAAA,EACrB,kBAAkB,MAAM;AAAA,EACxB,aAAa,MAAM;AAAA,EACnB,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,iBAAiB,MAAM;AAAA,EACvB,QAAQ,MAAM;AAAA,EACd,iBAAiB,MAAM;AAAA,EACvB,YAAY,MAAM;AAAA,EAClB,gBAAgB,MAAM;AAAA,EACtB,YAAY,MAAM;AAAA,EAClB,OAAO,MAAM;AAAA,EACb,iBAAiB,MAAM;AAAA,EACvB,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,KAAK,MAAM;AAAA,EACX,WAAW,MAAM;AAAA,EACjB,SAAS,MAAM;AAAA,EACf,iBAAiB,MAAM;AAAA,EACvB,eAAe,MAAM;AAAA,EACrB,wBAAwB,MAAM;AAAA,EAC9B,wBAAwB,MAAM;AAAA,EAC9B,iBAAiB,MAAM;AAAA,EACvB,iBAAiB,MAAM;AAAA,EACvB,oBAAoB,MAAM;AAAA,EAC1B,oBAAoB,MAAM;AAAA,EAC1B,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,OAAO,MAAM;AAAA,EACb,YAAY,MAAM;AAAA,EAClB,oBAAoB,MAAM;AAAA,EAC1B,WAAW,MAAM;AAAA,EACjB,MAAM,MAAM;AAAA,EACZ,UAAU,MAAM;AAAA,EAChB,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,UAAU,MAAM;AAAA,EAChB,SAAS,MAAM;AAAA,EACf,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,cAAc,MAAM;AAAA,EACpB,WAAW,MAAM;AAAA,EACjB,SAAS,MAAM;AAAA,EACf,UAAU,MAAM;AAAA,EAChB,KAAK,MAAM;AAAA,EACX,iBAAiB,MAAM;AAAA,EACvB,WAAW,MAAM;AAAA,EACjB,eAAe,MAAM;AAAA,EACrB,SAAS,MAAM;AAAA,EACf,kBAAkB,MAAM;AAAA,EACxB,iBAAiB,MAAM;AAAA,EACvB,iBAAiB,MAAM;AAAA,EACvB,iBAAiB,MAAM;AAAA,EACvB,cAAc,MAAM;AAAA,EACpB,eAAe,MAAM;AACvB,CAAC;AAGD,eAAe,qBAAqB,MAAM;AACxC,MAAI,QAAQ,MAAM;AAChB;AAAA,EACF;AACA,QAAM,WAAW,CAAC;AAClB,QAAM,OAAO,CAAC;AACd,QAAM,mBAAmB,CAAC;AAC1B,aAAW,OAAO,MAAM;AACtB,UAAM,QAAQ,KAAK;AACnB,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,cAAc;AACpB,eAAS,KAAK,YAAY,KAAK,CAAC;AAChC,WAAK,KAAK,GAAG;AACb,uBAAiB,KAAK,WAAW;AAAA,IACnC;AAAA,EACF;AACA,MAAI,SAAS,SAAS,GAAG;AACvB,UAAM,SAAS,MAAM,QAAQ,IAAI,QAAQ;AACzC,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,WAAK,KAAK,MAAM,OAAO,GAAG;AAAA,IAC5B;AACA,YAAQ,gBAAgB;AAAA,EAC1B;AACF;AACA,SAAS,qBAAqB,MAAM;AAClC,MAAI,QAAQ,MAAM;AAChB;AAAA,EACF;AACA,aAAW,OAAO,MAAM;AACtB,UAAM,QAAQ,KAAK;AACnB,QAAI,OAAO,UAAU,UAAU;AAC7B,YAAM,QAAQ;AAAA,IAChB;AAAA,EACF;AACF;AAGA,IAAI;AAAA,CACH,SAAS,wBAAwB;AAChC,yBAAuB,uBAAuB,YAAY,KAAK;AAC/D,yBAAuB,uBAAuB,aAAa,KAAK;AAClE,GAAG,0BAA0B,wBAAwB,CAAC,EAAE;AACxD,IAAI,yBAAyB;AAC7B,IAAI,eAAe,MAAM;AAAA,EACvB,cAAc;AACZ,SAAK,iBAAiB;AAAA,EACxB;AAAA,EACA,UAAU,QAAQ;AAChB,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAAA,EAChC;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAAA,EAC9B;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAAA,EAChC;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAAA,EAC9B;AAAA,EACA,MAAM,aAAa,MAAM;AAAA,EACzB;AAAA,EACA,MAAM,WAAW,MAAM;AAAA,EACvB;AAAA,EACA,SAAS,QAAQ;AAAA,EACjB;AACF;AACA,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,YAAY,cAAc,IAAI;AACxC,QAAI,cAAc,MAAM;AACtB,mBAAa,CAAC;AAAA,IAChB;AACA,SAAK,YAAY;AACjB,SAAK,cAAc;AAAA,EACrB;AAAA,EACA,OAAO,UAAU;AACf,SAAK,UAAU,KAAK,QAAQ;AAAA,EAC9B;AAAA,EACA,UAAU,QAAQ;AAChB,eAAW,YAAY,KAAK,WAAW;AACrC,eAAS,UAAU,MAAM;AAAA,IAC3B;AAAA,EACF;AAAA,EACA,SAAS,QAAQ;AACf,eAAW,YAAY,KAAK,WAAW;AACrC,eAAS,SAAS,MAAM;AAAA,IAC1B;AAAA,EACF;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAC9B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,aAAa,OAAO,IAAI;AAAA,IACzC;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,WAAW,OAAO,IAAI;AAAA,IACvC;AAAA,EACF;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAC9B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,aAAa,OAAO,IAAI;AAAA,IACzC;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,WAAW,OAAO,IAAI;AAAA,IACvC;AAAA,EACF;AAAA,EACA,MAAM,aAAa,MAAM;AACvB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,aAAa,IAAI;AAAA,IAClC;AAAA,EACF;AAAA,EACA,MAAM,WAAW,MAAM;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,eAAW,YAAY,KAAK,WAAW;AACrC,YAAM,SAAS,WAAW,IAAI;AAAA,IAChC;AAAA,EACF;AACF;AACA,IAAI,aAAa,cAAc,aAAa;AAAA,EAC1C,cAAc;AACZ,UAAM;AAAA,EACR;AAAA,EACA,MAAM,aAAa,OAAO;AACxB,SAAK,OAAO;AACZ,SAAK,SAAS,CAAC;AAAA,EACjB;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,YAAY,KAAK,WAAW,OAAO,IAAI,KAAK;AAClD,SAAK,QAAQ;AACb,eAAW,OAAO,MAAM;AACtB,YAAM,QAAQ,KAAK;AACnB,UAAI,OAAO,UAAU,UAAU;AAC7B,YAAI,CAAC,KAAK,OAAO,eAAe,GAAG,GAAG;AACpC,eAAK,OAAO,OAAO;AAAA,QACrB;AACA,aAAK,OAAO,OAAO,KAAK,OAAO,OAAO,QAAQ;AAAA,MAChD,OAAO;AACL,YAAI;AACJ,YAAI,OAAO,KAAK,QAAQ;AACtB,+BAAqB,KAAK,OAAO;AAAA,QACnC,OAAO;AACL,eAAK,OAAO,OAAO;AAAA,QACrB;AACA,cAAM,QAAQ,KAAK,MAAM,KAAK,KAAK,OAAO,MAAM,IAAI,OAAO,SAAS,CAAC,CAAC;AACtE,aAAK,OAAO,OAAO;AACnB,YAAI,sBAAsB,MAAM;AAC9B,6BAAmB,QAAQ;AAAA,QAC7B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,QAAI,QAAQ,MAAM;AAChB,iBAAW,OAAO,KAAK,OAAO,YAAY;AACxC,YAAI,KAAK,OAAO,QAAQ,MAAM;AAC5B;AAAA,QACF;AACA,YAAI,OAAO,KAAK,OAAO,SAAS,UAAU;AACxC,eAAK,OAAO,KAAK,OAAO,OAAO,KAAK;AAAA,QACtC,OAAO;AACL,eAAK,MAAM;AACT,kBAAM,OAAO,IAAI,IAAI,GAAG,KAAK,IAAI,GAAG,KAAK,OAAO,IAAI;AACpD,iBAAK,OAAO;AACZ,iBAAK,OAAO,KAAK,QAAQ;AACzB,iBAAK,KAAK,IAAI;AAAA,UAChB,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AACA,IAAI,UAAU,cAAc,aAAa;AAAA,EACvC,MAAM,aAAa,MAAM;AACvB,SAAK,QAAQ,CAAC;AACd,SAAK,UAAU,CAAC;AAAA,EAClB;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,MAAM,KAAK,KAAK;AACrB,eAAW,OAAO,MAAM;AACtB,UAAI,KAAK,QAAQ,QAAQ,MAAM;AAC7B,aAAK,QAAQ,OAAO,CAAC;AAAA,MACvB;AACA,WAAK,QAAQ,KAAK,KAAK,KAAK,IAAI;AAAA,IAClC;AAAA,EACF;AAAA,EACA,MAAM,WAAW;AACf,UAAM,WAAW,CAAC;AAClB,UAAM,OAAO,CAAC;AACd,UAAM,UAAU,CAAC;AACjB,eAAW,OAAO,KAAK,SAAS;AAC9B,YAAM,aAAa,KAAK,QAAQ;AAChC,eAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,YAAI,OAAO,WAAW,OAAO,UAAU;AACrC,gBAAM,cAAc,WAAW;AAC/B,mBAAS,KAAK,YAAY,KAAK,CAAC;AAChC,eAAK,KAAK,GAAG;AACb,kBAAQ,KAAK,CAAC;AAAA,QAChB;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,MAAM,QAAQ,IAAI,QAAQ;AACzC,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,YAAM,kBAAkB,KAAK,QAAQ,KAAK,IAAI,QAAQ;AACtD,sBAAgB,QAAQ;AACxB,WAAK,QAAQ,KAAK,IAAI,QAAQ,MAAM,OAAO,GAAG;AAAA,IAChD;AAAA,EACF;AACF;AACA,IAAI,iBAAiB,cAAc,aAAa;AAAA,EAC9C,YAAY,MAAM,YAAY;AAC5B,UAAM;AACN,SAAK,eAAe;AACpB,SAAK,UAAU,KAAK;AACpB,SAAK,gBAAgB,KAAK,iBAAiB;AAC3C,SAAK,aAAa,cAAc;AAChC,QAAI,KAAK,eAAe,QAAQ;AAC9B,WAAK,aAAa;AAAA,IACpB;AACA,QAAI,KAAK,eAAe,WAAW,KAAK,WAAW,MAAM;AACvD,YAAM,IAAI,MAAM,iHAAiH;AAAA,IACnI;AACA,QAAI,aAAa,SAAS,KAAK,UAAU,GAAG;AAC1C,WAAK,YAAY,SAAS,KAAK,UAAU,KAAK,IAAI,GAAG,KAAK,YAAY,KAAK,OAAO;AAAA,IACpF;AACA,SAAK,aAAa,KAAK;AACvB,SAAK,WAAW,KAAK;AACrB,SAAK,aAAa,KAAK;AACvB,SAAK,WAAW,KAAK;AACrB,SAAK,aAAa,KAAK;AACvB,SAAK,WAAW,KAAK;AACrB,SAAK,QAAQ,KAAK;AAAA,EACpB;AAAA,EACA,MAAM,UAAU,OAAO,OAAO,MAAM;AAClC,UAAM,KAAK,CAAC;AACZ,QAAI,KAAK,SAAS,MAAM;AACtB,YAAM,qBAAqB,IAAI;AAC/B,SAAG,KAAK,KAAK,MAAM,OAAO,OAAO,IAAI,CAAC;AAAA,IACxC;AACA,OAAG,KAAK,KAAK,cAAc,CAAC;AAC5B,UAAM,QAAQ,IAAI,EAAE;AAAA,EACtB;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAC9B,SAAK,eAAe;AACpB,QAAI,KAAK,cAAc,MAAM;AAC3B,YAAM,qBAAqB,IAAI;AAC/B,YAAM,KAAK,WAAW,OAAO,IAAI;AAAA,IACnC;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,UAAM,KAAK,CAAC;AACZ,QAAI,KAAK,YAAY,MAAM;AACzB,YAAM,qBAAqB,IAAI;AAC/B,SAAG,KAAK,KAAK,SAAS,OAAO,IAAI,CAAC;AAAA,IACpC;AACA,QAAI,KAAK,eAAe,SAAS;AAC/B,SAAG,KAAK,KAAK,cAAc,CAAC;AAAA,IAC9B;AACA,UAAM,QAAQ,IAAI,EAAE;AAAA,EACtB;AAAA,EACA,MAAM,aAAa,OAAO,MAAM;AAC9B,QAAI,KAAK,cAAc,MAAM;AAC3B,YAAM,qBAAqB,IAAI;AAC/B,YAAM,KAAK,WAAW,OAAO,IAAI;AAAA,IACnC;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,UAAM,KAAK,CAAC;AACZ,QAAI,KAAK,YAAY,MAAM;AACzB,YAAM,qBAAqB,IAAI;AAC/B,SAAG,KAAK,KAAK,SAAS,OAAO,IAAI,CAAC;AAAA,IACpC;AACA,QAAI,KAAK,eAAe,SAAS;AAC/B,SAAG,KAAK,KAAK,cAAc,CAAC;AAAA,IAC9B,WAAW,aAAa,SAAS,KAAK,UAAU,GAAG;AACjD,SAAG,KAAK,KAAK,UAAU,KAAK,cAAc,OAAO,IAAI,CAAC;AAAA,IACxD;AACA,UAAM,QAAQ,IAAI,EAAE;AAAA,EACtB;AAAA,EACA,MAAM,aAAa,MAAM;AACvB,QAAI,KAAK,cAAc,MAAM;AAC3B,YAAM,qBAAqB,IAAI;AAC/B,YAAM,KAAK,WAAW,IAAI;AAAA,IAC5B;AAAA,EACF;AAAA,EACA,MAAM,WAAW,MAAM;AACrB,QAAI,KAAK,YAAY,MAAM;AACzB,YAAM,qBAAqB,IAAI;AAC/B,YAAM,KAAK,SAAS,IAAI;AAAA,IAC1B;AAAA,EACF;AACF;AACA,SAAS,qBAAqB,YAAY,YAAY;AACpD,MAAI,cAAc,MAAM;AACtB,iBAAa,CAAC;AAAA,EAChB;AACA,MAAI,sBAAsB,cAAc;AACtC,WAAO,CAAC,UAAU;AAAA,EACpB;AACA,MAAI,MAAM,QAAQ,UAAU,KAAK,WAAW,cAAc,cAAc;AACtE,WAAO;AAAA,EACT;AACA,QAAM,kBAAkB,OAAO,UAAU;AACzC,SAAO,gBAAgB,IAAI,CAAC,mBAAmB,IAAI,eAAe,gBAAgB,UAAU,CAAC;AAC/F;AACA,IAAI,8BAA8B,MAAM;AAAA,EACtC,cAAc;AAAA,EACd;AAAA,EACA,OAAO,4BAA4B,gBAAgB,qBAAqB;AACtE,iBAAa,OAAO,kBAAkB,KAAK,OAAO,UAAU,cAAc,GAAG,MAAM,8DAA8D,gBAAgB;AACjK,gCAA4B,kBAAkB,mBAAmB;AACjE,QAAI,4BAA4B,aAAa,mBAAmB,MAAM;AACpE,kCAA4B,aAAa,kBAAkB,CAAC;AAAA,IAC9D;AACA,gCAA4B,aAAa,gBAAgB,KAAK,mBAAmB;AAAA,EACnF;AAAA,EACA,OAAO,kBAAkB,qBAAqB;AAC5C,eAAW,aAAa,4BAA4B,cAAc;AAChE,YAAM,eAAe,4BAA4B,aAAa,CAAC;AAC/D,mBAAa,QAAQ,CAAC,SAAS;AAC7B,YAAI,SAAS,qBAAqB;AAChC,gBAAM,IAAI,WAAW,iCAAiC;AAAA,QACxD;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EACA,OAAO,QAAQ;AACb,gCAA4B,eAAe,CAAC;AAAA,EAC9C;AAAA,EACA,OAAO,gBAAgB,gBAAgB;AACrC,UAAM,eAAe,CAAC;AACtB,eAAW,aAAa,4BAA4B,cAAc;AAChE,YAAM,QAAQ,CAAC;AACf,UAAI,kBAAkB,OAAO;AAC3B,qBAAa,KAAK,GAAG,4BAA4B,aAAa,MAAM;AAAA,MACtE;AAAA,IACF;AACA,WAAO,aAAa,IAAI,CAAC,SAAS,IAAI,KAAK,CAAC;AAAA,EAC9C;AACF;AACA,4BAA4B,eAAe,CAAC;AAC5C,SAAS,mBAAmB,YAAY,SAAS,QAAQ,cAAc,iBAAiB,eAAe,WAAW,cAAc,iBAAiB;AAC/I,QAAM,UAAU,IAAI,QAAQ;AAC5B,QAAM,kBAAkB;AAAA,IACtB,IAAI,WAAW;AAAA,IACf,GAAG,4BAA4B,gBAAgB,OAAO;AAAA,EACxD;AACA,MAAI,cAAc,MAAM;AACtB,oBAAgB,KAAK,GAAG,UAAU;AAAA,EACpC;AACA,kBAAgB,KAAK,OAAO;AAC5B,QAAM,eAAe,IAAI,aAAa,eAAe;AACrD,eAAa,UAAU;AAAA,IACrB;AAAA,IACA;AAAA,IACA,SAAS;AAAA,IACT,OAAO;AAAA,IACP;AAAA,IACA;AAAA,IACA;AAAA,IACA,SAAS;AAAA,EACX,CAAC;AACD,SAAO,EAAE,cAAc,QAAQ;AACjC;AAGA,SAAS,YAAY,QAAQ,gBAAgB,CAAC,GAAG,iBAAiB,OAAO;AACvE,SAAO,uBAAuB,QAAQ,sBAAsB,iBAAiB,OAAO,EAAE,cAAc,eAAe,SAAS,cAAc;AAC5I;AAGA,SAAS,YAAY,GAAG,MAAM;AAC5B,SAAO,KAAK,MAAM;AAChB,QAAI,EAAE,UAAU,WAAW;AACzB,UAAI,KAAK,GAAG,SAAS;AAAA,IACvB;AACA,UAAM,YAAY,KAAK,QAAQ,CAAC,GAAG,MAAM,IAAI;AAC7C,UAAM,gBAAgB,KAAK,UAAU,OAAO,QAAQ,CAAC;AACrD,UAAM,QAAQ,KAAK,QAAQ,WAAW,aAAa,CAAC;AACpD,WAAO,IAAI,GAAG,KAAK;AAAA,EACrB,CAAC;AACH;AACA,SAAS,kBAAkB,OAAO,OAAO;AACvC,SAAO,KAAK,MAAM,KAAK,QAAQ,IAAI,OAAO,KAAK,CAAC,GAAG,EAAE,CAAC;AACxD;AACA,SAAS,kBAAkB,OAAO,OAAO;AACvC,SAAO,KAAK,MAAM,KAAK,IAAI,IAAI,OAAO,KAAK,CAAC,GAAG,EAAE,CAAC;AACpD;AACA,SAAS,4BAA4B,OAAO,OAAO;AACjD,SAAO,KAAK,MAAM;AAChB,UAAM,OAAO,IAAI,OAAO,KAAK;AAC7B,UAAM,cAAc,YAAY,IAAI,KAAK,GAAG,QAAQ,GAAG,OAAO,SAAS;AACvE,UAAM,YAAY,IAAI,IAAI,MAAM,WAAW,CAAC;AAC5C,WAAO,IAAI,KAAK,KAAK,WAAW,EAAE,CAAC;AAAA,EACrC,CAAC;AACH;AACA,SAAS,4BAA4B,OAAO,OAAO;AACjD,SAAO,KAAK,MAAM;AAChB,UAAM,cAAc,YAAY,OAAO,QAAQ,GAAG,OAAO,SAAS;AAClE,UAAM,WAAW,KAAK,KAAK,GAAG,WAAW,CAAC;AAC1C,UAAM,cAAc,YAAY,OAAO,QAAQ,GAAG,OAAO,SAAS;AAClE,UAAM,YAAY,KAAK,KAAK,GAAG,WAAW,CAAC;AAC3C,WAAO,KAAK,QAAQ,IAAI,UAAU,SAAS,CAAC,GAAG,EAAE;AAAA,EACnD,CAAC;AACH;AACA,SAAS,aAAa,OAAO,OAAO;AAClC,SAAO,KAAK,MAAM;AAChB,UAAM,YAAY,QAAQ,GAAG,IAAI,GAAG,IAAI,OAAO,KAAK,CAAC,CAAC;AACtD,WAAO,KAAK,QAAQ,SAAS,GAAG,EAAE;AAAA,EACpC,CAAC;AACH;AACA,SAAS,MAAM,OAAO,OAAO;AAC3B,SAAO,KAAK,MAAM;AAChB,UAAM,YAAY,QAAQ,GAAG,IAAI,GAAG,IAAI,OAAO,KAAK,CAAC,CAAC;AACtD,WAAO,KAAK,WAAW,EAAE;AAAA,EAC3B,CAAC;AACH;AACA,SAAS,iBAAiB,OAAO,OAAO;AACtC,SAAO,KAAK,MAAM;AAChB,UAAM,MAAM,KAAK,IAAI,OAAO,KAAK,GAAG,EAAE;AACtC,UAAM,OAAO,IAAI,IAAI,IAAI,GAAG,KAAK,GAAG,KAAK,GAAG,EAAE;AAC9C,WAAO,QAAQ,GAAG,KAAK,GAAG,IAAI,MAAM,GAAG,CAAC,CAAC;AAAA,EAC3C,CAAC;AACH;AACA,SAAS,QAAQ,OAAO,OAAO;AAC7B,SAAO,KAAK,MAAM;AAChB,UAAM,QAAQ,KAAK,IAAI,CAAC;AACxB,UAAM,iBAAiB,IAAI,OAAO,KAAK;AACvC,UAAM,gBAAgB,IAAI,KAAK,gBAAgB,SAAS,IAAI,IAAI,cAAc,CAAC,CAAC,GAAG,KAAK;AACxF,WAAO,KAAK,eAAe,EAAE;AAAA,EAC/B,CAAC;AACH;AACA,SAAS,wBAAwB,QAAQ,QAAQ,aAAa,OAAO;AACnE,SAAO,KAAK,MAAM;AAChB,QAAI,YAAY;AACd,eAAS,QAAQ,MAAM;AAAA,IACzB,OAAO;AACL,YAAM,YAAY,KAAK,QAAQ,OAAO,MAAM,SAAS,GAAG,IAAI;AAC5D,eAAS,IAAI,QAAQ,SAAS;AAAA,IAChC;AACA,aAAS,YAAY,QAAQ,QAAQ,GAAG,IAAI,QAAQ,CAAC;AACrD,WAAO,IAAI,KAAK,IAAI,KAAK,QAAQ,SAAS,GAAG,KAAK,MAAM,CAAC,GAAG,OAAO,MAAM,SAAS,CAAC,CAAC;AAAA,EACtF,CAAC;AACH;AACA,SAAS,8BAA8B,QAAQ,QAAQ,aAAa,OAAO;AACzE,SAAO,KAAK,MAAM;AAChB,UAAM,aAAa,KAAK,MAAM,SAAS,MAAM,CAAC,GAAG,OAAO;AACxD,aAAS,YAAY,QAAQ,QAAQ,GAAG,IAAI,QAAQ,CAAC;AACrD,UAAM,cAAc,OAAO;AAC3B,UAAM,eAAe,QAAQ,OAAO,YAAY,YAAY,YAAY,SAAS,EAAE,GAAG,WAAW;AACjG,WAAO,wBAAwB,cAAc,QAAQ,UAAU;AAAA,EACjE,CAAC;AACH;AACA,SAAS,8BAA8B,QAAQ,QAAQ;AACrD,MAAI,CAAC,aAAa,YAAY,OAAO,OAAO,OAAO,KAAK,GAAG;AACzD,UAAM,IAAI,WAAW,8DAA8D,KAAK,UAAU,OAAO,KAAK,SAAS,KAAK,UAAU,OAAO,KAAK,GAAG;AAAA,EACvJ;AACA,SAAO,KAAK,MAAM;AAChB,UAAM,aAAa,KAAK,MAAM;AAC9B,UAAM,eAAe,IAAI,IAAI,MAAM,CAAC;AACpC,WAAO,KAAK,IAAI,YAAY,IAAI,QAAQ,MAAM,CAAC,GAAG,MAAM,IAAI,YAAY,CAAC,CAAC;AAAA,EAC5E,CAAC;AACH;AACA,SAAS,mBAAmB,OAAO,OAAO;AACxC,SAAO,KAAK,MAAM;AAChB,QAAI;AACJ,QAAI,YAAY,OAAO,QAAQ,GAAG,IAAI,QAAQ,CAAC;AAC/C,QAAI,KAAK,IAAI,GAAG,IAAI,GAAG,CAAC,CAAC,CAAC;AAC1B,WAAO,KAAK,8BAA8B,OAAO,CAAC,GAAG,EAAE;AAAA,EACzD,CAAC;AACH;AACA,SAAS,0BAA0B,OAAO,OAAO;AAC/C,SAAO,KAAK,MAAM;AAChB,UAAM,cAAc,YAAY,OAAO,QAAQ,GAAG,CAAC;AACnD,UAAM,cAAc,YAAY,OAAO,QAAQ,GAAG,CAAC;AACnD,WAAO,KAAK,IAAI,OAAO,KAAK,IAAI,aAAa,WAAW,CAAC,CAAC,GAAG,EAAE;AAAA,EACjE,CAAC;AACH;AACA,SAAS,QAAQ,OAAO,OAAO;AAC7B,SAAO,KAAK,MAAM;AAChB,UAAM,UAAU,KAAK,KAAK,QAAQ,GAAG,KAAK,CAAC;AAC3C,WAAO,KAAK,IAAI,OAAO,IAAI,OAAO,OAAO,CAAC,GAAG,EAAE;AAAA,EACjD,CAAC;AACH;AACA,SAAS,gBAAgB,OAAO,OAAO;AACrC,SAAO,KAAK,MAAM;AAChB,UAAM,iBAAiB,YAAY,OAAO,EAAE;AAC5C,UAAM,iBAAiB,YAAY,OAAO,EAAE;AAC5C,UAAM,YAAY,IAAI,gBAAgB,cAAc;AACpD,WAAO,IAAI,KAAK,WAAW,EAAE,CAAC;AAAA,EAChC,CAAC;AACH;AACA,IAAI,YAAY;AAAA,EACd,kBAAkB;AAAA,EAClB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,SAAS,IAAI,gBAAgB;AAC3B,MAAI,OAAO,mBAAmB,UAAU;AACtC,QAAI,kBAAkB,WAAW;AAC/B,aAAO,UAAU;AAAA,IACnB;AACA,QAAI,SAAS,gBAAgB;AAC7B,QAAI,eAAe,YAAY,EAAE,SAAS,qBAAqB,GAAG;AAChE,eAAS,gBAAgB;AAAA,IAC3B;AACA,UAAM,IAAI,WAAW,MAAM;AAAA,EAC7B,OAAO;AACL,WAAO;AAAA,EACT;AACF;AAGA,SAAS,eAAe,OAAO,OAAO;AACpC,SAAO,KAAK,MAAM;AAChB,UAAM,aAAa,IAAI,KAAK,SAAS,KAAK,CAAC;AAC3C,UAAM,mBAAmB,MAAM,QAAQ,OAAO,UAAU,GAAG,MAAM,KAAK;AACtE,WAAO,KAAK,MAAM,OAAO,gBAAgB,GAAG,EAAE;AAAA,EAChD,CAAC;AACH;AACA,SAAS,oBAAoB,OAAO,OAAO;AACzC,SAAO,KAAK,MAAM,MAAM,MAAM,OAAO,OAAO,EAAE,GAAG,OAAO,OAAO,EAAE,CAAC,GAAG,SAAS,CAAC;AACjF;AACA,SAAS,cAAc,OAAO,OAAO;AACnC,SAAO,KAAK,MAAM;AAChB,WAAO,KAAK,KAAK,WAAW,MAAM,OAAO,CAAC,GAAG,MAAM,OAAO,CAAC,CAAC,CAAC,GAAG,SAAS;AAAA,EAC3E,CAAC;AACH;AACA,SAAS,eAAe,OAAO,OAAO;AACpC,SAAO,KAAK,MAAM;AAChB,WAAO,KAAK,KAAK,WAAW,MAAM,OAAO,CAAC,GAAG,MAAM,OAAO,CAAC,CAAC,CAAC,GAAG,SAAS;AAAA,EAC3E,CAAC;AACH;AACA,SAAS,eAAe,OAAO,OAAO;AACpC,SAAO,KAAK,MAAM;AAChB,WAAO,KAAK,KAAK,WAAW,MAAM,OAAO,CAAC,GAAG,MAAM,OAAO,CAAC,CAAC,CAAC,GAAG,SAAS;AAAA,EAC3E,CAAC;AACH;AACA,SAAS,UAAU,OAAO,OAAO;AAC/B,SAAO,KAAK,MAAM;AAChB,UAAM,KAAK,cAAc,OAAO,KAAK;AACrC,UAAM,KAAK,eAAe,OAAO,KAAK;AACtC,UAAM,cAAc,KAAK,IAAI,EAAE;AAC/B,WAAO,KAAK,MAAM,QAAQ,aAAa,CAAC,GAAG,IAAI,IAAI,WAAW,GAAG,CAAC,GAAG,SAAS;AAAA,EAChF,CAAC;AACH;AACA,SAAS,OAAO,OAAO,OAAO;AAC5B,SAAO,KAAK,MAAM;AAChB,UAAM,KAAK,cAAc,OAAO,KAAK;AACrC,UAAM,KAAK,eAAe,OAAO,KAAK;AACtC,UAAM,cAAc,KAAK,IAAI,EAAE;AAC/B,WAAO,KAAK,MAAM,QAAQ,aAAa,CAAC,GAAG,IAAI,IAAI,WAAW,GAAG,CAAC,GAAG,SAAS;AAAA,EAChF,CAAC;AACH;AACA,SAAS,oBAAoB,OAAO,OAAO;AACzC,SAAO,mBAAmB,OAAO,KAAK;AACxC;AACA,SAAS,0BAA0B,OAAO,OAAO;AAC/C,MAAI,MAAM,SAAS,MAAM,MAAM;AAC7B,YAAQ,QAAQ,OAAO,CAAC,MAAM,OAAO,CAAC,CAAC;AAAA,EACzC;AACA,UAAQ,OAAO,OAAO,EAAE;AACxB,MAAI,MAAM,UAAU,MAAM,OAAO;AAC/B,YAAQ,KAAK,OAAO,MAAM,KAAK;AAAA,EACjC;AACA,SAAO,KAAK,MAAM,OAAO,KAAK,GAAG,SAAS;AAC5C;AACA,IAAI,MAAM;AACV,IAAI,MAAM;AACV,IAAI,MAAM;AACV,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,OAAO;AACX,IAAI,2BAA2B;AAC/B,IAAI,SAAS;AACb,IAAI,iCAAiC;AACrC,IAAI,aAAa;AAAA,EACf;AAAA,EACA;AAAA,EACA;AAAA,EACA,yBAAyB;AAAA,EACzB,+BAA+B;AAAA,EAC/B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,SAAS,KAAK,YAAY;AACxB,MAAI,OAAO,eAAe,YAAY,cAAc,YAAY;AAC9D,WAAO,WAAW;AAAA,EACpB,WAAW,OAAO,eAAe,YAAY,cAAc,MAAM;AAC/D,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,WAAW,kBAAkB,YAAY;AAAA,EACrD;AACF;AACA,SAAS,oBAAoB,IAAI;AAC/B,UAAQ,OAAO,MAAM,0BAA0B,IAAI;AACnD,MAAI,OAAO,OAAO,UAAU;AAC1B,WAAO;AAAA,EACT,OAAO;AACL,QAAI;AACJ,eAAW,OAAO,OAAO,KAAK,SAAS,GAAG;AACxC,UAAI,UAAU,SAAS,IAAI;AACzB,iBAAS;AACT;AAAA,MACF;AAAA,IACF;AACA,QAAI,WAAW,QAAQ;AACrB,aAAO;AAAA,IACT;AACA,eAAW,OAAO,OAAO,KAAK,UAAU,GAAG;AACzC,UAAI,WAAW,SAAS,IAAI;AAC1B,iBAAS;AACT;AAAA,MACF;AAAA,IACF;AACA,QAAI,WAAW,QAAQ;AACrB,aAAO;AAAA,IACT;AACA,WAAO,GAAG;AAAA,EACZ;AACF;AAGA,SAAS,aAAa,YAAY;AAChC,QAAM,eAAe;AAAA,IACnB,WAAW,MAAM,MAAM,QAAQ,IAAI;AAAA,IACnC,YAAY,MAAM,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC;AAAA,IACnD,QAAQ,MAAM,MAAM,KAAK,MAAM,KAAK,OAAO,QAAQ,CAAC;AAAA,IACpD,UAAU,MAAM,MAAM,OAAO,MAAM,KAAK,OAAO,QAAQ,GAAG,CAAC;AAAA,IAC3D,WAAW,MAAM,MAAM,QAAQ,MAAM,KAAK,GAAG,QAAQ,CAAC;AAAA,IACtD,OAAO,MAAM,MAAM,IAAI,IAAI;AAAA,EAC7B;AACA,eAAa,aAAa,aAAa;AACvC,eAAa,cAAc,aAAa;AACxC,eAAa,UAAU,aAAa;AACpC,eAAa,YAAY,aAAa;AACtC,eAAa,aAAa,aAAa;AACvC,eAAa,SAAS,aAAa;AACnC,MAAI,cAAc,cAAc;AAC9B,WAAO,aAAa,YAAY;AAAA,EAClC;AACA,QAAM,IAAI,WAAW,qBAAqB,YAAY;AACxD;AAGA,IAAI,8CAA8C,IAAI,OAAO;AAC7D,SAAS,yBAAyB,qBAAqB,WAAW,YAAY,OAAO;AACnF,MAAI,uBAAuB,QAAQ,OAAO,wBAAwB,YAAY,OAAO,eAAe,mBAAmB,MAAM,OAAO,aAAa,CAAC,iBAAiB,mBAAmB,GAAG;AACvL,UAAM,IAAI,MAAM,oEAAoE;AAAA,EACtF;AACA,MAAI,WAAW;AACb,UAAM,MAAM,KAAK,UAAU,mBAAmB;AAC9C,QAAI,IAAI,SAAS,6CAA6C;AAC5D,cAAQ,KAAK,mCAAmC,2CAA2C,IAAI,qJAAqJ,8CAA8C;AAAA,IACpS;AAAA,EACF;AACF;AACA,SAAS,iBAAiB,GAAG;AAC3B,MAAI,MAAM,MAAM;AACd,WAAO;AAAA,EACT,WAAW,OAAO,MAAM,UAAU;AAChC,QAAI,OAAO,eAAe,CAAC,MAAM,OAAO,WAAW;AACjD,YAAM,OAAO,OAAO,KAAK,CAAC;AAC1B,iBAAW,OAAO,MAAM;AACtB,YAAI,OAAO,QAAQ,UAAU;AAC3B,iBAAO;AAAA,QACT;AACA,YAAI,CAAC,iBAAiB,EAAE,IAAI,GAAG;AAC7B,iBAAO;AAAA,QACT;AAAA,MACF;AACA,aAAO;AAAA,IACT,OAAO;AACL,UAAI,MAAM,QAAQ,CAAC,GAAG;AACpB,mBAAW,QAAQ,GAAG;AACpB,cAAI,CAAC,iBAAiB,IAAI,GAAG;AAC3B,mBAAO;AAAA,UACT;AAAA,QACF;AACA,eAAO;AAAA,MACT,OAAO;AACL,eAAO;AAAA,MACT;AAAA,IACF;AAAA,EACF,OAAO;AACL,UAAM,QAAQ,OAAO;AACrB,WAAO,UAAU,YAAY,UAAU,YAAY,UAAU;AAAA,EAC/D;AACF;AAGA,SAAS,aAAa,QAAQ,YAAY,WAAW,UAAU,QAAQ,KAAK;AAC1E,QAAM,iBAAiB,sBAAsB,MAAM;AACnD,QAAM,YAAY,CAAC,gBAAgB,eAAe,gBAAgB,SAAS;AAC3E,MAAI,gBAAgB;AAClB,iBAAa,cAAc;AAC3B,gBAAY,aAAa,CAAC,MAAM,MAAM,MAAM,CAAC;AAAA,EAC/C,OAAO;AACL,iBAAa,cAAc;AAC3B,gBAAY,aAAa,CAAC,MAAM,MAAM,KAAK,KAAK,CAAC;AAAA,EACnD;AACA,MAAI,UAAU,UAAU,SAAS,MAAM,GAAG;AACxC,gBAAY,UAAU,IAAI,CAAC,OAAO,KAAK,MAAM,aAAa,EAAE,CAAC;AAAA,EAC/D;AACA,MAAI;AACJ,MAAI,CAAC,gBAAgB;AACnB,cAAU,KAAK,iBAAiB;AAChC,oBAAgB,CAAC;AACjB,eAAW,SAAS,OAAO,cAAc;AACvC,oBAAc,KAAK,GAAG,OAAO,aAAa,MAAM;AAAA,IAClD;AAAA,EACF;AACA,UAAQ,IAAI,OAAO,UAAU,CAAC;AAC9B,WAAS,WAAW,WAAW,OAAO;AACtC,UAAQ,IAAI,OAAO,UAAU,CAAC;AAC9B,QAAM,SAAS,OAAO;AACtB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,QAAI,gBAAgB;AAClB,wBAAkB,OAAO,IAAI,WAAW,OAAO;AAAA,IACjD,OAAO;AACL,uCAAiC,OAAO,IAAI,WAAW,eAAe,OAAO;AAAA,IAC/E;AACA,aAAS,MAAM,OAAO,SAAS,IAAI,MAAM,KAAK,OAAO,UAAU,CAAC;AAAA,EAClE;AACA,SAAO,iCAAiC;AACxC,QAAM,iBAAiB,qBAAqB,MAAM;AAClD,QAAM,oBAAoB,qBAAqB,OAAO,mBAAmB;AACzE,UAAQ,iBAAiB,iBAAiB,mBAAmB;AAC7D,UAAQ,qBAAqB,gBAAgB;AAC7C,UAAQ,yBAAyB,mBAAmB;AACpD,UAAQ,IAAI,OAAO,UAAU,CAAC;AAChC;AACA,SAAS,qBAAqB,QAAQ;AACpC,MAAI;AACJ,MAAI,OAAO,6BAA6B,MAAM;AAC5C,qBAAiB,qBAAqB,OAAO,yBAAyB;AAAA,EACxE,OAAO;AACL,qBAAiB,qBAAqB,OAAO,gBAAgB;AAAA,EAC/D;AACA,SAAO;AACT;AACA,SAAS,sBAAsB,QAAQ;AACrC,MAAI,iBAAiB;AACrB,QAAM,eAAe,CAAC;AACtB,QAAM,QAAQ,CAAC;AACf,aAAW,SAAS,OAAO,cAAc;AACvC,iBAAa,KAAK,OAAO,aAAa,MAAM;AAAA,EAC9C;AACA,aAAW,cAAc,cAAc;AACrC,QAAI,WAAW,SAAS,KAAK,WAAW,WAAW,KAAK,WAAW,GAAG,cAAc,SAAS,GAAG;AAC9F,uBAAiB;AACjB;AAAA,IACF;AACA,UAAM,KAAK,GAAG,UAAU;AAAA,EAC1B;AACA,MAAI,gBAAgB;AAClB,eAAW,SAAS,OAAO,QAAQ;AACjC,UAAI,OAAO;AACX,iBAAW,QAAQ,MAAM,cAAc;AACrC,YAAI,MAAM,QAAQ,IAAI,MAAM,IAAI;AAC9B,cAAI,MAAM;AACR,6BAAiB;AACjB;AAAA,UACF,OAAO;AACL,mBAAO;AAAA,UACT;AAAA,QACF;AAAA,MACF;AACA,UAAI,CAAC,gBAAgB;AACnB;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,SAAS,QAAQ,WAAW,UAAU,QAAQ,KAAK;AAC1D,MAAI,OAAO;AACX,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,QAAI,IAAI,GAAG;AACT,aAAO,KAAK,MAAM,GAAG,KAAK,SAAS,CAAC,IAAI;AAAA,IAC1C;AACA,YAAQ,OAAO;AACf,WAAO,KAAK,MAAM,GAAG,UAAU,EAAE;AACjC,YAAQ,IAAI,OAAO,UAAU,KAAK,KAAK,MAAM;AAAA,EAC/C;AACA,UAAQ,IAAI;AACd;AACA,SAAS,kBAAkB,OAAO,WAAW,SAAS;AACpD,MAAI;AACJ,MAAI;AACJ,MAAI;AACF,iBAAa,MAAM,aAAa,IAAI,CAAC,MAAM,KAAK,UAAU,EAAE,WAAW,CAAC,EAAE,KAAK,GAAG;AAAA,EACpF,SAAS,KAAP;AACA,iBAAa;AAAA,EACf;AACA,MAAI;AACF,kBAAc,KAAK,UAAU,MAAM,WAAW;AAAA,EAChD,SAAS,KAAP;AACA,kBAAc;AAAA,EAChB;AACA,QAAM,OAAO,MAAM;AACnB,QAAM,YAAY,MAAM,aAAa;AACrC,QAAM,SAAS;AAAA,IACb,GAAG,SAAS;AAAA,IACZ;AAAA,IACA;AAAA,IACA,MAAM,YAAY,EAAE,SAAS;AAAA,EAC/B;AACA,WAAS,QAAQ,WAAW,OAAO;AACrC;AACA,SAAS,iCAAiC,OAAO,WAAW,eAAe,SAAS;AAClF,MAAI;AACJ,MAAI;AACJ,MAAI;AACF,iBAAa,MAAM,aAAa,IAAI,CAAC,MAAM,KAAK,UAAU,EAAE,WAAW,CAAC,EAAE,KAAK,GAAG;AAAA,EACpF,SAAS,KAAP;AACA,iBAAa;AAAA,EACf;AACA,MAAI;AACF,kBAAc,KAAK,UAAU,MAAM,WAAW;AAAA,EAChD,SAAS,KAAP;AACA,kBAAc;AAAA,EAChB;AACA,QAAM,cAAc,CAAC;AACrB,aAAW,QAAQ,MAAM,cAAc;AACrC,QAAI,iBAAiB,QAAQ,cAAc,SAAS,KAAK,cAAc,QAAQ,IAAI,MAAM,IAAI;AAC3F;AAAA,IACF;AACA,aAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,EAAE,GAAG;AAClD,YAAM,eAAe,KAAK,cAAc,GAAG;AAC3C,YAAM,oBAAoB,KAAK,YAAY;AAC3C,YAAM,qBAAqB,KAAK,cAAc;AAC9C,kBAAY,KAAK,GAAG,gBAAgB,sBAAsB,qBAAqB;AAAA,IACjF;AAAA,EACF;AACA,QAAM,OAAO,MAAM;AACnB,QAAM,YAAY,MAAM,aAAa;AACrC,QAAM,kBAAkB,YAAY,WAAW,IAAI,KAAK,YAAY;AACpE,QAAM,SAAS;AAAA,IACb,GAAG,SAAS;AAAA,IACZ;AAAA,IACA;AAAA,IACA,MAAM,YAAY,EAAE,SAAS;AAAA,IAC7B;AAAA,EACF;AACA,WAAS,QAAQ,WAAW,OAAO;AACnC,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,aAAS,CAAC,IAAI,IAAI,IAAI,IAAI,YAAY,EAAE,GAAG,WAAW,OAAO;AAAA,EAC/D;AACF;AAGA,SAAS,6BAA6B,KAAK,OAAO,OAAO;AACvD,UAAQ,QAAQ,kBAAkB,QAAQ,kBAAkB,QAAQ,kBAAkB,UAAU,KAAK,OAAO,UAAU;AACxH;AACA,SAAS,oBAAoB,gBAAgB,KAAK;AAChD,MAAI,mBAAmB,MAAM;AAC3B,WAAO;AAAA,EACT,WAAW,OAAO,mBAAmB,UAAU;AAC7C,WAAO,YAAY,cAAc;AAAA,EACnC,WAAW,OAAO,mBAAmB,YAAY,OAAO,mBAAmB,WAAW;AACpF,WAAO;AAAA,EACT,WAAW,0BAA0B,OAAO;AAC1C,UAAM,UAAU,CAAC;AACjB,UAAM,cAAc,eAAe;AACnC,aAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,YAAM,OAAO,eAAe;AAC5B,UAAI,6BAA6B,KAAK,GAAG,IAAI,GAAG;AAC9C,gBAAQ,KAAK,IAAI;AAAA,MACnB,OAAO;AACL,gBAAQ,KAAK,oBAAoB,MAAM,GAAG,CAAC;AAAA,MAC7C;AAAA,IACF;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,SAAS,CAAC;AAChB,eAAW,eAAe,OAAO,KAAK,cAAc,GAAG;AACrD,YAAM,gBAAgB,eAAe;AACrC,UAAI,gBAAgB,UAAU,OAAO,kBAAkB,UAAU;AAC/D,eAAO,eAAe;AAAA,MACxB,OAAO;AACL,cAAM,QAAQ,YAAY,WAAW;AACrC,eAAO,SAAS,oBAAoB,eAAe,KAAK;AAAA,MAC1D;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;AACA,SAAS,oBAAoB,UAAU,KAAK;AAC1C,MAAI,aAAa,QAAQ,aAAa,QAAQ;AAC5C,WAAO;AAAA,EACT,WAAW,OAAO,aAAa,UAAU;AACvC,WAAO,YAAY,QAAQ;AAAA,EAC7B,WAAW,OAAO,aAAa,YAAY,OAAO,aAAa,WAAW;AACxE,WAAO;AAAA,EACT,WAAW,oBAAoB,OAAO;AACpC,UAAM,UAAU,CAAC;AACjB,UAAM,cAAc,SAAS;AAC7B,aAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,YAAM,OAAO,SAAS;AACtB,UAAI,6BAA6B,KAAK,GAAG,IAAI,GAAG;AAC9C,gBAAQ,KAAK,IAAI;AAAA,MACnB,OAAO;AACL,gBAAQ,KAAK,oBAAoB,MAAM,GAAG,CAAC;AAAA,MAC7C;AAAA,IACF;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,SAAS,CAAC;AAChB,eAAW,SAAS,OAAO,KAAK,QAAQ,GAAG;AACzC,YAAM,UAAU,SAAS;AACzB,YAAM,QAAQ,YAAY,KAAK;AAC/B,WAAK,UAAU,UAAU,UAAU,gBAAgB,OAAO,YAAY,UAAU;AAC9E,eAAO,SAAS;AAAA,MAClB,OAAO;AACL,eAAO,SAAS,oBAAoB,SAAS,KAAK;AAAA,MACpD;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;AAGA,IAAI,WAAW;AAGf,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,CAAC,CAAC;AACR,SAAK,iBAAiC,oBAAI,IAAI;AAC9C,SAAK,OAAO,KAAK;AACjB,QAAI,KAAK,QAAQ,MAAM;AACrB,YAAM,SAAS,KAAK,aAAa,EAAE,YAAY;AAC/C,WAAK,OAAO,OAAO,MAAM;AAAA,IAC3B;AACA,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAClB,QAAI,MAAM,QAAQ,KAAK,MAAM,GAAG;AAC9B,WAAK,SAAS,KAAK,OAAO,MAAM;AAAA,IAClC,OAAO;AACL,WAAK,SAAS,CAAC,KAAK,MAAM;AAAA,IAC5B;AACA,QAAI,MAAM,QAAQ,KAAK,OAAO,GAAG;AAC/B,WAAK,UAAU,KAAK,QAAQ,MAAM;AAAA,IACpC,OAAO;AACL,WAAK,UAAU,CAAC,KAAK,OAAO;AAAA,IAC9B;AACA,QAAI,QAAQ,KAAK,MAAM,EAAE,WAAW,KAAK,OAAO,QAAQ;AACtD,YAAM,IAAI,WAAW,mGAAmG,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,IAAI,GAAG;AAAA,IAC1J;AACA,QAAI,QAAQ,KAAK,OAAO,EAAE,WAAW,KAAK,QAAQ,QAAQ;AACxD,cAAQ,KAAK,qGAAqG,KAAK,QAAQ,IAAI,CAAC,MAAM,EAAE,IAAI,GAAG;AAAA,IACrJ;AACA,SAAK,cAAc,CAAC;AACpB,SAAK,yBAAyB,CAAC;AAC/B,SAAK,2BAA2B,CAAC;AACjC,SAAK,eAAe,CAAC;AACrB,SAAK,0BAA0B,CAAC;AAChC,SAAK,4BAA4B,CAAC;AAClC,SAAK,SAAS,CAAC;AACf,SAAK,wBAAwB,CAAC;AAC9B,eAAW,KAAK,KAAK,SAAS;AAC5B,YAAM,QAAQ,EAAE;AAChB,YAAM,YAAY,EAAE;AACpB,YAAM,cAAc,EAAE;AACtB,WAAK,aAAa,KAAK,KAAK;AAC5B,WAAK,wBAAwB,KAAK,SAAS;AAC3C,WAAK,0BAA0B,KAAK,WAAW;AAAA,IACjD;AACA,eAAW,KAAK,KAAK,QAAQ;AAC3B,YAAM,QAAQ,EAAE;AAChB,YAAM,YAAY,EAAE;AACpB,YAAM,cAAc,EAAE;AACtB,cAAQ,cAAc,GAAG,0BAA0B;AACnD,cAAQ,gBAAgB,GAAG,4BAA4B;AACvD,WAAK,YAAY,KAAK,KAAK;AAC3B,WAAK,uBAAuB,KAAK,SAAS;AAC1C,WAAK,yBAAyB,KAAK,WAAW;AAAA,IAChD;AACA,SAAK,aAAa,CAAC;AACnB,SAAK,cAAc,CAAC;AACpB,SAAK,kBAAkB,CAAC;AACxB,SAAK,iBAAiB,CAAC;AACvB,SAAK,kBAAkB,CAAC;AACxB,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,QAAQ,KAAK;AAChD,YAAM,QAAQ,KAAK,YAAY;AAC/B,UAAI,EAAE,iBAAiB,aAAa;AAClC,cAAM,IAAI,UAAU,8EAA8E,KAAK,iBAAiB,0CAA0C,MAAM,aAAa,IAAI;AAAA,MAC3L;AACA,WAAK,WAAW,KAAK,MAAM,IAAI;AAC/B,WAAK,gBAAgB,KAAK,MAAM,eAAe;AAC/C,WAAK,eAAe,KAAK,MAAM,IAAI;AAAA,IACrC;AACA,eAAW,SAAS,KAAK,cAAc;AACrC,WAAK,YAAY,KAAK,MAAM,IAAI;AAAA,IAClC;AACA,SAAK,sBAAsB,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AACzD,SAAK,uBAAuB,KAAK,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK;AAC3D,UAAM,cAAc,CAAC;AACrB,UAAM,eAAe,CAAC;AACtB,UAAM,eAAe,CAAC;AACtB,UAAM,iBAAiB,CAAC;AACxB,UAAM,eAAe,CAAC;AACtB,UAAM,yBAAyB,CAAC;AAChC,UAAM,kBAAkB,CAAC,SAAS,gBAAgB,kBAAkB,OAAO,WAAW,gBAAgB;AACpG,UAAI,SAAS,QAAQ,aAAa,QAAQ,eAAe,MAAM;AAC7D,gBAAQ,QAAQ;AAChB,oBAAY,QAAQ;AACpB,sBAAc,QAAQ;AAAA,MACxB;AACA,YAAM,OAAO,MAAM,aAAa;AAChC,UAAI,iBAAiB,QAAQ,IAAI,MAAM,IAAI;AACzC,cAAM,IAAI,aAAa,cAAc,QAAQ,kBAAkB,MAAM,2BAA2B;AAAA,MAClG;AACA,UAAI,eAAe,QAAQ,IAAI,MAAM,IAAI;AACvC;AAAA,MACF;AACA,WAAK,eAAe,IAAI,UAAU,QAAQ,OAAO,SAAS,CAAC;AAC3D,UAAI,EAAE,MAAM,MAAM,eAAe;AAC/B,qBAAa,MAAM,MAAM,OAAO,KAAK,YAAY,EAAE;AAAA,MACrD;AACA,UAAI,iBAAiB,QAAQ,IAAI,MAAM,IAAI;AACzC,yBAAiB,KAAK,IAAI;AAAA,MAC5B;AACA,YAAM,mBAAmB,KAAK,cAAc;AAC5C,eAAS,IAAI,GAAG,IAAI,kBAAkB,KAAK;AACzC,cAAM,IAAI,KAAK,aAAa;AAC5B,cAAM,SAAS,KAAK,cAAc;AAClC,cAAM,aAAa,KAAK,YAAY;AACpC,cAAM,eAAe,KAAK,cAAc;AACxC,wBAAgB,GAAG,gBAAgB,kBAAkB,QAAQ,YAAY,YAAY;AAAA,MACvF;AACA,qBAAe,KAAK,IAAI;AACxB,aAAO,iBAAiB,QAAQ,IAAI,KAAK,GAAG;AAC1C,yBAAiB,OAAO,iBAAiB,QAAQ,IAAI,GAAG,CAAC;AAAA,MAC3D;AACA,6BAAuB,KAAK,IAAI;AAAA,IAClC;AACA,UAAM,gBAAgB,CAAC;AACvB,UAAM,kBAAkB,CAAC;AACzB,eAAW,KAAK,KAAK,SAAS;AAC5B,sBAAgB,GAAG,eAAe,eAAe;AAAA,IACnD;AACA,UAAM,iCAAiC,uBAAuB,MAAM,EAAE,QAAQ;AAC9E,eAAW,QAAQ,gCAAgC;AACjD,mBAAa,KAAK,MAAM;AACxB,UAAI,EAAE,KAAK,MAAM,cAAc;AAC7B,oBAAY,KAAK,MAAM;AAAA,MACzB;AACA,UAAI,QAAQ,YAAY,KAAK;AAC7B,YAAM,gBAAgB,aAAa,KAAK,cAAc,OAAO,OAAO,IAAI,aAAa,KAAK,cAAc;AACxG,cAAQ,KAAK,IAAI,OAAO,aAAa;AACrC,mBAAa,KAAK,cAAc,MAAM;AACtC,qBAAe,KAAK,cAAc,MAAM,KAAK;AAC7C,kBAAY,KAAK,MAAM;AACvB,eAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,KAAK;AAClD,cAAM,eAAe,KAAK,cAAc;AACxC,cAAM,YAAY,KAAK,YAAY;AACnC,cAAM,cAAc,aAAa,aAAa;AAC9C,cAAM,iBAAiB,YAAY,YAAY,OAAO,OAAO,IAAI,YAAY,YAAY;AACzF,oBAAY,YAAY,MAAM,KAAK,IAAI,QAAQ,GAAG,cAAc;AAChE,qBAAa,YAAY,MAAM;AAAA,MACjC;AAAA,IACF;AACA,UAAM,eAAe,CAAC;AACtB,eAAW,UAAU,aAAa;AAChC,YAAM,QAAQ,YAAY;AAC1B,UAAI,EAAE,SAAS,eAAe;AAC5B,qBAAa,SAAS,CAAC;AAAA,MACzB;AACA,mBAAa,OAAO,KAAK,aAAa,OAAO;AAAA,IAC/C;AACA,UAAM,gBAAgB,CAAC;AACvB,eAAW,WAAW,cAAc;AAClC,YAAM,QAAQ,aAAa;AAC3B,UAAI,EAAE,SAAS,gBAAgB;AAC7B,sBAAc,SAAS,CAAC;AAAA,MAC1B;AACA,oBAAc,OAAO,KAAK,eAAe,QAAQ;AAAA,IACnD;AACA,QAAI,YAAY,OAAO,KAAK,aAAa,EAAE,IAAI,CAAC,MAAM,SAAS,GAAG,EAAE,CAAC,EAAE,KAAK,oBAAoB;AAChG,SAAK,SAAS,CAAC;AACf,eAAW,SAAS,WAAW;AAC7B,YAAM,iBAAiB,cAAc;AACrC,qBAAe,KAAK,CAAC,GAAG,MAAM;AAC5B,cAAM,SAAS,aAAa,EAAE;AAC9B,cAAM,SAAS,aAAa,EAAE;AAC9B,YAAI,SAAS,QAAQ;AACnB,iBAAO;AAAA,QACT;AACA,YAAI,SAAS,QAAQ;AACnB,iBAAO;AAAA,QACT;AACA,eAAO;AAAA,MACT,CAAC;AACD,iBAAW,SAAS,gBAAgB;AAClC,YAAI,iBAAiB,WAAW;AAC9B,eAAK,sBAAsB,KAAK,KAAK;AAAA,QACvC;AACA,aAAK,OAAO,KAAK,KAAK;AAAA,MACxB;AAAA,IACF;AACA,SAAK,gBAAgB;AACrB,gBAAY,OAAO,KAAK,YAAY,EAAE,IAAI,CAAC,MAAM,SAAS,GAAG,EAAE,CAAC,EAAE,KAAK,oBAAoB;AAC3F,UAAM,oBAAoB,KAAK,OAAO,MAAM;AAC5C,UAAM,0BAA0B,CAAC;AACjC,eAAW,SAAS,WAAW;AAC7B,iBAAW,QAAQ,aAAa,QAAQ;AACtC,cAAM,QAAQ,KAAK;AACnB,YAAI,SAAS,MAAM;AACjB,qBAAW,KAAK,KAAK,cAAc;AACjC,gBAAI,kBAAkB,QAAQ,CAAC,MAAM,IAAI;AACvC,oBAAM,IAAI,aAAa,sDAAsD,eAAe,MAAM,qEAAqE,yBAAyB;AAAA,YAClM;AAAA,UACF;AACA,qBAAW,KAAK,KAAK,eAAe;AAClC,8BAAkB,KAAK,CAAC;AAAA,UAC1B;AACA,kCAAwB,KAAK,MAAM,IAAI;AAAA,QACzC;AAAA,MACF;AAAA,IACF;AACA,SAAK,eAAe;AACpB,UAAM,WAAW,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,IAAI;AAC9C,eAAW,QAAQ,UAAU;AAC3B,YAAM,iBAAiB,SAAS,OAAO,CAAC,MAAM,MAAM,IAAI,EAAE;AAC1D,UAAI,mBAAmB,GAAG;AACxB,cAAM,IAAI,aAAa,aAAa,iBAAiB,uFAAuF,KAAK,UAAU,QAAQ,CAAC;AAAA,MACtK;AAAA,IACF;AACA,SAAK,gBAAgB,CAAC;AACtB,SAAK,eAAe,CAAC;AACrB,QAAI,KAAK;AAAA,MACP,eAAe;AAAA,MACf,eAAe,CAAC;AAAA,MAChB,aAAa,CAAC;AAAA,MACd,eAAe,CAAC;AAAA,MAChB,cAAc,KAAK;AAAA,MACnB,eAAe,KAAK;AAAA,MACpB,YAAY,KAAK,OAAO,IAAI,CAAC,MAAM,IAAI;AAAA,MACvC,aAAa,KAAK,QAAQ,IAAI,CAAC,MAAM,IAAI;AAAA,MACzC,aAAa,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AAAA,MAC3C,cAAc,KAAK,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK;AAAA,IAC/C,CAAC;AACD,SAAK,QAAQ;AACb,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,oBAAoB;AAClB,QAAI,KAAK,cAAc,GAAG;AACxB,YAAM,IAAI,MAAM,cAAc,KAAK,4BAA4B;AAAA,IACjE;AAAA,EACF;AAAA,EACA,UAAU;AACR,SAAK,kBAAkB;AACvB,UAAM,SAAS,EAAE,sBAAsB,MAAM,sBAAsB,EAAE;AACrE,QAAI,EAAE,KAAK,cAAc,GAAG;AAC1B,iBAAW,SAAS,KAAK,QAAQ;AAC/B,eAAO,wBAAwB,MAAM,QAAQ,EAAE;AAAA,MACjD;AACA,iBAAW,aAAa,KAAK,uBAAuB;AAClD,eAAO,wBAAwB,UAAU,QAAQ,EAAE;AAAA,MACrD;AAAA,IACF;AACA,WAAO,uBAAuB,KAAK;AACnC,WAAO;AAAA,EACT;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,SAAK,OAAO,QAAQ,CAAC,UAAU;AAC7B,YAAM,kBAAkB,QAAQ,CAAC,MAAM,EAAE,YAAY,SAAS;AAAA,IAChE,CAAC;AACD,SAAK,aAAa;AAAA,EACpB;AAAA,EACA,IAAI,mBAAmB;AACrB,QAAI,KAAK,kBAAkB,SAAS,GAAG;AACrC,YAAM,IAAI,WAAW,sNAAsN;AAAA,IAC7O;AACA,QAAI,CAAC,KAAK,WAAW;AACnB,aAAO,CAAC;AAAA,IACV;AACA,QAAI,UAAU,CAAC;AACf,eAAW,SAAS,KAAK,QAAQ;AAC/B,gBAAU,QAAQ,OAAO,MAAM,gBAAgB;AAAA,IACjD;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,sBAAsB;AACxB,UAAM,UAAU,CAAC;AACjB,eAAW,SAAS,KAAK,QAAQ;AAC/B,cAAQ,KAAK,GAAG,MAAM,mBAAmB;AAAA,IAC3C;AACA,QAAI,CAAC,KAAK,WAAW;AACnB,YAAM,mBAAmB,CAAC;AAC1B,iBAAW,SAAS,KAAK,QAAQ;AAC/B,yBAAiB,KAAK,GAAG,MAAM,gBAAgB;AAAA,MACjD;AACA,aAAO,iBAAiB,OAAO,OAAO;AAAA,IACxC;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,iBAAiB,OAAO,KAAK,mBAAmB;AAAA,EAC9D;AAAA,EACA,YAAY,SAAS,SAAS,MAAM;AAClC,UAAM,eAAe,CAAC;AACtB,QAAI,oBAAoB;AACxB,eAAW,SAAS,KAAK,QAAQ;AAC/B,iBAAW,UAAU,MAAM,SAAS;AAClC,YAAI,aAAa,OAAO,iBAAiB,MAAM;AAC7C,gBAAM,IAAI,WAAW,0BAA0B,OAAO,cAAc;AAAA,QACtE;AACA,qBAAa,OAAO,gBAAgB;AACpC;AAAA,MACF;AAAA,IACF;AACA,UAAM,oBAAoB,CAAC;AAC3B,eAAW,QAAQ,SAAS;AAC1B,UAAI,gBAAgB;AACpB,UAAI,aAAa,SAAS,MAAM;AAC9B,cAAM,SAAS,KAAK,MAAM,GAAG;AAC7B,cAAM,mBAAmB,OAAO,MAAM,GAAG,EAAE,EAAE,OAAO,CAAC,OAAO,OAAO,SAAS,EAAE,CAAC;AAC/E,wBAAgB,iBAAiB,KAAK,GAAG;AAAA,MAC3C;AACA,UAAI,aAAa,kBAAkB,MAAM;AACvC,0BAAkB,KAAK,CAAC,aAAa,gBAAgB,QAAQ,KAAK,CAAC;AAAA,MACrE,WAAW,QAAQ;AACjB,cAAM,IAAI,WAAW,gDAAgD,MAAM;AAAA,MAC7E;AACA,aAAO,aAAa;AAAA,IACtB;AACA,QAAI,QAAQ;AACV,YAAM,aAAa,CAAC;AACpB,iBAAW,QAAQ,cAAc;AAC/B,mBAAW,KAAK,IAAI;AAAA,MACtB;AACA,UAAI,WAAW,SAAS,GAAG;AACzB,cAAM,IAAI,WAAW,GAAG,WAAW,aAAa,0CAA0C,YAAY;AAAA,MACxG;AAAA,IACF;AACA,kBAAc,iBAAiB;AAAA,EACjC;AAAA,EACA,gBAAgB;AACd,UAAM,YAAY,KAAK,UAAU;AACjC,UAAM,cAAc,CAAC;AACrB,gBAAY,eAAe,KAAK,aAAa;AAC7C,gBAAY,YAAY;AACxB,gBAAY,kBAAkB,eAAe;AAC7C,gBAAY,aAAa;AACzB,WAAO;AAAA,EACT;AAAA,EACA,OAAO,QAAQ,eAAe,MAAM;AAClC,UAAM,cAAc,oBAAoB,KAAK,cAAc,CAAC;AAC5D,WAAO,eAAe,KAAK,UAAU,WAAW,IAAI;AAAA,EACtD;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,OAAO,MAAM;AACtB,YAAM,WAAW,IAAI,SAAS;AAC9B,eAAS,IAAI,GAAG,IAAI,KAAK,OAAO,QAAQ,EAAE,GAAG;AAC3C,iBAAS,IAAI,KAAK,OAAO,IAAI,OAAO,EAAE;AAAA,MACxC;AACA,aAAO,QAAQ,KAAK,SAAS,UAAU,MAAM;AAAA,IAC/C,CAAC;AAAA,EACH;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,WAAO,KAAK,MAAM;AAChB,eAAS,OAAO,MAAM;AACtB,UAAI;AACJ,UAAI,QAAQ,MAAM;AAChB,gBAAQ,aAAa,MAAM,OAAO,MAAM;AAAA,MAC1C,OAAO;AACL,gBAAQ,OAAO,IAAI;AAAA,MACrB;AACA,aAAO,KAAK,iBAAiB,QAAQ,KAAK,EAAE;AAAA,IAC9C,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,UAAM,cAAc,mBAAmB,UAAU;AACjD,QAAI,YAAY,WAAW,KAAK,YAAY,QAAQ;AAClD,YAAM,IAAI,WAAW,+BAA+B,yBAAyB,KAAK,YAAY,uBAAuB;AAAA,IACvH;AACA,UAAM,uBAAuB,CAAC;AAC9B,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,YAAM,QAAQ,KAAK,YAAY;AAC/B,YAAM,cAAc,YAAY;AAChC,YAAM,WAAW,MAAM,OAAO;AAC9B,2BAAqB,YAAY;AAAA,IACnC;AACA,UAAM,YAAY,OAAO,KAAK,KAAK,YAAY,EAAE,IAAI,CAAC,MAAM,SAAS,GAAG,EAAE,CAAC,EAAE,KAAK,oBAAoB;AACtG,QAAI,UAAU,SAAS,GAAG;AACxB,iBAAW,SAAS,WAAW;AAC7B,cAAM,QAAQ,KAAK,aAAa;AAChC,mBAAW,QAAQ,OAAO;AACxB,gBAAM,QAAQ,KAAK;AACnB,cAAI,KAAK,YAAY,IAAI,CAAC,MAAM,EAAE,EAAE,EAAE,QAAQ,MAAM,EAAE,MAAM,IAAI;AAC9D;AAAA,UACF;AACA,gBAAM,eAAe,CAAC;AACtB,mBAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,KAAK;AAClD,kBAAM,eAAe,KAAK,cAAc;AACxC,kBAAM,aAAa,KAAK,YAAY;AACpC,kBAAM,cAAc,KAAK,cAAc;AACvC,kBAAM,WAAW,GAAG,aAAa,QAAQ,cAAc;AACvD,kBAAM,cAAc,qBAAqB;AACzC,yBAAa,KAAK,WAAW;AAAA,UAC/B;AACA,gBAAM,cAAc,MAAM,mBAAmB,iBAAiB,YAAY,CAAC;AAC3E,gBAAM,gBAAgB,mBAAmB,WAAW;AACpD,gBAAM,YAAY,MAAM,aAAa,QAAQ,IAAI;AACjD,mBAAS,IAAI,GAAG,IAAI,cAAc,QAAQ,KAAK;AAC7C,kBAAM,WAAW,GAAG,MAAM,QAAQ,aAAa;AAC/C,iCAAqB,YAAY,cAAc;AAAA,UACjD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,eAAe,CAAC;AACtB,UAAM,kBAAkB,CAAC;AACzB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,QAAQ,KAAK;AACjD,YAAM,QAAQ,KAAK,aAAa;AAChC,YAAM,YAAY,KAAK,wBAAwB;AAC/C,YAAM,cAAc,KAAK,0BAA0B;AACnD,YAAM,WAAW,GAAG,MAAM,QAAQ,aAAa;AAC/C,sBAAgB,KAAK,QAAQ;AAAA,IAC/B;AACA,aAAS,IAAI,GAAG,IAAI,gBAAgB,QAAQ,KAAK;AAC/C,YAAM,MAAM,gBAAgB;AAC5B,cAAQ,OAAO,oBAAoB;AACnC,mBAAa,KAAK,qBAAqB,IAAI;AAAA,IAC7C;AACA,WAAO,iBAAiB,YAAY;AAAA,EACtC;AAAA,EACA,iBAAiB,QAAQ,OAAO;AAC9B,QAAI,SAAS,MAAM;AACjB,cAAQ,aAAa,MAAM,OAAO,MAAM;AAAA,IAC1C;AACA,UAAM,YAAY,CAAC;AACnB,aAAS,IAAI,GAAG,IAAI,KAAK,OAAO,QAAQ,EAAE,GAAG;AAC3C,YAAM,IAAI,KAAK,OAAO;AACtB,YAAM,IAAI,OAAO;AACjB,YAAM,OAAO,MAAM;AACnB,gBAAU,EAAE,MAAM,CAAC,GAAG,IAAI;AAAA,IAC5B;AACA,UAAM,YAAY,OAAO,KAAK,KAAK,YAAY,EAAE,IAAI,CAAC,MAAM,SAAS,GAAG,EAAE,CAAC,EAAE,KAAK,oBAAoB;AACtG,eAAW,SAAS,WAAW;AAC7B,YAAM,QAAQ,KAAK,aAAa;AAChC,iBAAW,QAAQ,OAAO;AACxB,cAAM,QAAQ,KAAK;AACnB,cAAM,wBAAwB,KAAK;AACnC,cAAM,yBAAyB,KAAK;AACpC,cAAM,eAAe,IAAI,MAAM;AAC/B,mBAAW,KAAK,uBAAuB;AACrC,cAAI,EAAE,MAAM,WAAW;AACrB,yBAAa,KAAK,UAAU,EAAE,GAAG;AAAA,UACnC;AAAA,QACF;AACA,YAAI,aAAa,WAAW,sBAAsB,QAAQ;AACxD,cAAI,SAAS,CAAC;AACd,cAAI;AACJ,cAAI;AACJ,cAAI;AACJ,cAAI;AACJ,cAAI,KAAK,YAAY,MAAM;AACzB,qBAAS,KAAK;AAAA,UAChB;AACA,cAAI,aAAa,WAAW,GAAG;AAC7B,kBAAM,CAAC,gBAAgB,YAAY,IAAI,aAAa;AACpD,gBAAI,OAAO,WAAW,MAAM;AAC1B,qBAAO,UAAU;AAAA,YACnB;AACA,6BAAiB,OAAO,MAAM,KAAK,gBAAgB,MAAM,CAAC;AAC1D,2BAAe,OAAO,MAAM,YAAY,gBAAgB,YAAY,CAAC;AACrE,8BAAkB,CAAC,cAAc;AACjC,4BAAgB,CAAC,YAAY;AAAA,UAC/B,OAAO;AACL,8BAAkB,aAAa,IAAI,CAAC,MAAM,EAAE,EAAE;AAC9C,4BAAgB,aAAa,IAAI,CAAC,MAAM,EAAE,EAAE;AAC5C,gBAAI,OAAO,WAAW,MAAM;AAC1B,qBAAO,UAAU;AAAA,YACnB;AACA,6BAAiB,OAAO,MAAM,KAAK,iBAAiB,MAAM,CAAC;AAC3D,2BAAe,OAAO,MAAM,YAAY,iBAAiB,aAAa,CAAC;AAAA,UACzE;AACA,cAAI,MAAM,qBAAqB;AAC7B,kBAAM,IAAI,oBAAoB,uHAAuH;AAAA,UACvJ;AACA,mBAAS,IAAI,GAAG,IAAI,uBAAuB,QAAQ,EAAE,GAAG;AACtD,kBAAM,IAAI,uBAAuB;AACjC,kBAAM,IAAI,eAAe;AACzB,kBAAM,OAAO,aAAa;AAC1B,sBAAU,EAAE,MAAM,CAAC,GAAG,IAAI;AAAA,UAC5B;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,gBAAgB,CAAC;AACvB,UAAM,cAAc,CAAC;AACrB,UAAM,eAAe,CAAC;AACtB,eAAW,KAAK,KAAK,SAAS;AAC5B,cAAQ,EAAE,MAAM,WAAW,4BAA4B,EAAE,UAAU,EAAE,IAAI;AACzE,YAAM,CAAC,SAAS,IAAI,IAAI,UAAU,EAAE;AACpC,mBAAa,KAAK,QAAQ,KAAK;AAC/B,oBAAc,KAAK,OAAO;AAC1B,kBAAY,KAAK,IAAI;AAAA,IACvB;AACA,WAAO,CAAC,eAAe,aAAa,YAAY;AAAA,EAClD;AAAA,EACA,uBAAuB,QAAQ;AAC7B,UAAM,oBAAoB,CAAC;AAC3B,QAAI;AACJ,eAAW,SAAS,KAAK,QAAQ;AAC/B,kBAAY,iBAAiB,YAAY,IAAI;AAC7C,eAAS,oBAAoB,GAAG,oBAAoB,MAAM,aAAa,QAAQ,qBAAqB;AAClG,cAAM,UAAU,UAAU,QAAQ,OAAO,iBAAiB;AAC1D,YAAI,KAAK,eAAe,IAAI,OAAO,GAAG;AACpC,4BAAkB,WAAW;AAC7B,uBAAa;AAAA,QACf;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,SAAS,MAAM,OAAO;AACpB,QAAI,SAAS,MAAM;AACjB,UAAI,KAAK,OAAO,UAAU,OAAO;AAC/B,cAAM,IAAI,WAAW,wCAAwC,6BAA6B,KAAK,OAAO,kBAAkB;AAAA,MAC1H,OAAO;AACL,eAAO,KAAK,OAAO;AAAA,MACrB;AAAA,IACF,OAAO;AACL,UAAI,QAAQ,MAAM;AAChB,cAAM,IAAI,WAAW,4CAA4C;AAAA,MACnE;AAAA,IACF;AACA,eAAW,SAAS,KAAK,QAAQ;AAC/B,UAAI,MAAM,SAAS,MAAM;AACvB,eAAO;AAAA,MACT;AAAA,IACF;AACA,UAAM,IAAI,WAAW,kBAAkB,MAAM;AAAA,EAC/C;AAAA,EACA,kBAAkB;AAChB,WAAO,KAAK,MAAM;AAChB,YAAM,UAAU,CAAC;AACjB,iBAAW,SAAS,KAAK,QAAQ;AAC/B,iBAAS,YAAY,GAAG,YAAY,MAAM,aAAa,QAAQ,EAAE,WAAW;AAC1E,gBAAM,UAAU,UAAU,QAAQ,OAAO,SAAS;AAClD,cAAI,KAAK,eAAe,IAAI,OAAO,GAAG;AACpC,oBAAQ,KAAK,GAAG,MAAM,gBAAgB,CAAC;AAAA,UACzC;AAAA,QACF;AAAA,MACF;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,MAAM,KAAK,KAAK;AACjC,UAAM,oBAAoB,KAAK,uBAAuB,KAAK,MAAM;AACjE,UAAM,eAAe,CAAC;AACtB,eAAW,SAAS,KAAK,QAAQ;AAC/B,YAAM,iBAAiB,MAAM,aAAa;AAC1C,YAAM,cAAc,MAAM,UAAU;AACpC,YAAM,uBAAuB,CAAC;AAC9B,eAAS,oBAAoB,GAAG,oBAAoB,MAAM,aAAa,QAAQ,qBAAqB;AAClG,cAAM,OAAO,MAAM,aAAa;AAChC,cAAM,UAAU,UAAU,QAAQ,OAAO,iBAAiB;AAC1D,YAAI,SAAS,CAAC;AACd,YAAI,KAAK,eAAe,IAAI,OAAO,GAAG;AACpC,cAAI,KAAK,UAAU;AACjB,gBAAI;AACF,mBAAK,UAAU,KAAK,QAAQ;AAC5B,uBAAS,KAAK;AAAA,YAChB,SAAS,KAAP;AACA,sBAAQ,KAAK,SAAS,MAAM,uDAAuD,KAAK,iHAAiH;AACzM,uBAAS,CAAC;AAAA,YACZ;AAAA,UACF;AACA,cAAI,KAAK,cAAc,SAAS,GAAG;AACjC,kBAAM,WAAW,CAAC;AAClB,qBAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,KAAK;AAClD,oBAAM,eAAe,KAAK,cAAc;AACxC,oBAAM,YAAY,KAAK,YAAY;AACnC,oBAAM,cAAc,KAAK,cAAc;AACvC,oBAAM,WAAW,UAAU,QAAQ,cAAc,SAAS;AAC1D,kBAAI,eAAe,kBAAkB;AACrC,kBAAI,gBAAgB,MAAM;AACxB,+BAAe;AAAA,cACjB;AACA,uBAAS,KAAK,CAAC,aAAa,MAAM,cAAc,aAAa,MAAM,CAAC;AAAA,YACtE;AACA,iCAAqB,KAAK,QAAQ;AAAA,UACpC;AAAA,QACF;AAAA,MACF;AACA,YAAM,OAAO,CAAC;AACd,WAAK,UAAU,MAAM;AACrB,WAAK,eAAe;AACpB,WAAK,YAAY;AACjB,WAAK,kBAAkB;AACvB,mBAAa,KAAK,IAAI;AAAA,IACxB;AACA,WAAO,YAAY;AACnB,UAAM,cAAc,CAAC;AACrB,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,QAAQ,KAAK;AAChD,YAAM,QAAQ,KAAK,YAAY;AAC/B,YAAM,YAAY,KAAK,uBAAuB;AAC9C,YAAM,UAAU,UAAU,QAAQ,OAAO,SAAS;AAClD,UAAI,CAAC,KAAK,eAAe,IAAI,OAAO,GAAG;AACrC;AAAA,MACF;AACA,UAAI,eAAe,kBAAkB;AACrC,UAAI,iBAAiB,QAAQ,iBAAiB,QAAQ;AACpD,uBAAe;AAAA,MACjB;AACA,YAAM,cAAc,KAAK,yBAAyB;AAClD,kBAAY,KAAK,CAAC,MAAM,MAAM,cAAc,WAAW,CAAC;AAAA,IAC1D;AACA,WAAO,iBAAiB;AACxB,UAAM,eAAe,CAAC;AACtB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,QAAQ,KAAK;AACjD,YAAM,QAAQ,KAAK,aAAa;AAChC,YAAM,YAAY,KAAK,wBAAwB;AAC/C,YAAM,UAAU,UAAU,QAAQ,OAAO,SAAS;AAClD,UAAI,CAAC,KAAK,eAAe,IAAI,OAAO,GAAG;AACrC;AAAA,MACF;AACA,UAAI,eAAe,kBAAkB;AACrC,UAAI,iBAAiB,QAAQ,iBAAiB,QAAQ;AACpD,uBAAe;AAAA,MACjB;AACA,YAAM,cAAc,KAAK,0BAA0B;AACnD,mBAAa,KAAK,CAAC,MAAM,MAAM,cAAc,WAAW,CAAC;AAAA,IAC3D;AACA,WAAO,kBAAkB;AACzB,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ,gBAAgB,CAAC,GAAG,iBAAiB,OAAO;AACzE,UAAM,gBAAgB,CAAC;AACvB,UAAM,mBAAmB,CAAC;AAC1B,aAAS,mBAAmB,OAAO,UAAU;AAC3C,UAAI,EAAE,MAAM,QAAQ,mBAAmB;AACrC,yBAAiB,MAAM,QAAQ,CAAC,QAAQ;AAAA,MAC1C,OAAO;AACL,yBAAiB,MAAM,MAAM,KAAK,QAAQ;AAAA,MAC5C;AAAA,IACF;AACA,aAAS,YAAY,OAAO,UAAU;AACpC,YAAM,gBAAgB,CAAC;AACvB,UAAI;AACJ,iBAAW,aAAa,UAAU;AAChC,cAAM,mBAAmB,UAAU;AACnC,cAAM,mBAAmB,UAAU;AACnC,cAAM,qBAAqB,UAAU;AACrC,iBAAS,UAAU,MAAM,OAAO,CAAC,IAAI,UAAU;AAC/C,YAAI,EAAE,oBAAoB,gBAAgB;AACxC,6BAAmB,OAAO,QAAQ;AAClC;AAAA,QACF;AACA,cAAM,eAAe,cAAc;AACnC,YAAI,aAAa,aAAa,UAAU,kBAAkB;AACxD,6BAAmB,OAAO,QAAQ;AAClC;AAAA,QACF;AACA,cAAM,cAAc,aAAa,aAAa;AAC9C,sBAAc,KAAK,YAAY,cAAc,mBAAmB;AAAA,MAClE;AACA,UAAI,cAAc,SAAS,GAAG;AAC5B,cAAM,MAAM,iBAAiB,aAAa,GAAG,MAAM;AAAA,MACrD;AAAA,IACF;AACA,aAAS,aAAa,WAAW;AAC/B,YAAM,YAAY,UAAU;AAC5B,YAAM,QAAQ,YAAY,WAAW,OAAO,oBAAoB,OAAO,OAAO,mBAAmB,CAAC,CAAC;AACnG,YAAM,6BAA6B,cAAc;AACjD,oBAAc,aAAa;AAC3B,YAAM,mBAAmB,UAAU;AACnC,uBAAiB,QAAQ,CAAC,aAAa;AACrC,YAAI,EAAE,oBAAoB,QAAQ;AAChC,gBAAM,IAAI,WAAW,yDAAyD,UAAU;AAAA,QAC1F;AACA,2BAAmB,OAAO,QAAQ;AAAA,MACpC,CAAC;AAAA,IACH;AACA,UAAM,OAAO,OAAO;AACpB,UAAM,mBAAmB,OAAO;AAChC,eAAW,aAAa,kBAAkB;AACxC,mBAAa,SAAS;AAAA,IACxB;AACA,WAAO,CAAC,cAAc,gBAAgB,GAAG;AACvC,iBAAW,aAAa,kBAAkB;AACxC,cAAM,QAAQ,cAAc,UAAU;AACtC,YAAI,MAAM,QAAQ,kBAAkB;AAClC,gBAAM,kCAAkC,iBAAiB,MAAM;AAC/D,iBAAO,iBAAiB,MAAM;AAC9B,qBAAW,YAAY,iCAAiC;AACtD,wBAAY,OAAO,QAAQ;AAAA,UAC7B;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,eAAe,CAAC;AACtB,UAAM,gBAAgB,CAAC;AACvB,UAAM,wBAAwB,OAAO;AACrC,eAAW,aAAa,uBAAuB;AAC7C,YAAM,YAAY,UAAU;AAC5B,YAAM,YAAY,UAAU;AAC5B,YAAM,cAAc,UAAU;AAC9B,cAAQ,aAAa,aAAa;AAClC,YAAM,QAAQ,cAAc;AAC5B,YAAM,qBAAqB,MAAM,aAAa,WAAW;AACzD,mBAAa,KAAK,mBAAmB,YAAY;AAAA,IACnD;AACA,UAAM,yBAAyB,OAAO;AACtC,eAAW,aAAa,wBAAwB;AAC9C,YAAM,YAAY,UAAU;AAC5B,YAAM,YAAY,UAAU;AAC5B,YAAM,cAAc,UAAU;AAC9B,cAAQ,aAAa,aAAa;AAClC,YAAM,QAAQ,cAAc;AAC5B,YAAM,qBAAqB,MAAM,aAAa,WAAW;AACzD,oBAAc,KAAK,mBAAmB,YAAY;AAAA,IACpD;AACA,WAAO,IAAI,IAAI,EAAE,QAAQ,cAAc,SAAS,eAAe,KAAK,CAAC;AAAA,EACvE;AAAA,EACA,IAAI,WAAW;AACb,QAAI,KAAK,WAAW;AAClB,YAAM,IAAI,WAAW,sLAAsL;AAAA,IAC7M;AACA,eAAW,SAAS,KAAK,QAAQ;AAC/B,UAAI,MAAM,UAAU;AAClB,eAAO;AAAA,MACT;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,cAAc;AACZ,SAAK,MAAM;AACT,WAAK,OAAO,QAAQ,CAAC,UAAU;AAC7B,YAAI,MAAM,UAAU;AAClB,gBAAM,YAAY;AAAA,QACpB;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AACF;AAGA,SAAS,gCAAgC,SAAS,aAAa,YAAY;AACzE,QAAM,aAAa,YAAY;AAC/B,MAAI,WAAW,QAAQ,MAAM,QAAQ,OAAO,KAAK,QAAQ,WAAW,GAAG;AACrE,WAAO,YAAY,IAAI,CAAC,SAAS,IAAI;AAAA,EACvC;AACA,MAAI,eAAe,GAAG;AACpB,QAAI,MAAM,QAAQ,OAAO,KAAK,QAAQ,WAAW,GAAG;AAClD,aAAO;AAAA,IACT,WAAW,OAAO,YAAY,YAAY,YAAY,MAAM,SAAS;AACnE,aAAO,CAAC,QAAQ,YAAY,GAAG;AAAA,IACjC,OAAO;AACL,aAAO,CAAC,OAAO;AAAA,IACjB;AAAA,EACF;AACA,MAAI,MAAM,QAAQ,OAAO,GAAG;AAC1B,QAAI,QAAQ,WAAW,YAAY;AACjC,YAAM,IAAI,MAAM,YAAY,6BAA6B,QAAQ,wCAAwC,mFAAmF;AAAA,IAC9L;AACA,WAAO;AAAA,EACT,WAAW,OAAO,YAAY,YAAY,OAAO,KAAK,OAAO,EAAE,SAAS,KAAK,OAAO,QAAQ,OAAO,KAAK,OAAO,EAAE,QAAQ,UAAU;AACjI,UAAM,SAAS,CAAC;AAChB,gBAAY,QAAQ,CAAC,eAAe;AAClC,UAAI,cAAc,SAAS;AACzB,eAAO,KAAK,QAAQ,WAAW;AAAA,MACjC,OAAO;AACL,eAAO,KAAK,IAAI;AAAA,MAClB;AAAA,IACF,CAAC;AACD,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,2BAA2B,2BAA2B,2CAA2C,yCAAyC,8BAA8B,8BAA8B,KAAK,UAAU,OAAO,GAAG;AAAA,EACjP;AACF;AACA,SAAS,wBAAwB,aAAa,aAAa;AACzD,SAAO,gCAAgC,aAAa,aAAa,aAAa;AAChF;AACA,eAAe,mBAAmB,GAAG,cAAc,aAAa,kBAAkB;AAChF,MAAI,gBAAgB,QAAQ,oBAAoB,MAAM;AACpD,UAAM,IAAI,MAAM,6CAA6C;AAAA,EAC/D;AACA,MAAI,eAAe,MAAM;AACvB,UAAM,WAAW,KAAK,MAAM;AAC1B,UAAI,EAAE,MAAM,WAAW,GAAG;AACxB,eAAO,MAAM,CAAC;AAAA,MAChB,WAAW,EAAE,MAAM,WAAW,GAAG;AAC/B,YAAI,EAAE,MAAM,KAAK,GAAG;AAClB,gBAAM,OAAO;AACb,iBAAO,OAAO,GAAG,IAAI;AAAA,QACvB,WAAW,EAAE,MAAM,OAAO,GAAG;AAC3B,iBAAO,QAAQ,GAAG,CAAC,EAAE,MAAM,EAAE,CAAC;AAAA,QAChC,OAAO;AACL,gBAAM,IAAI,MAAM,+CAA+C,EAAE,MAAM,wEAAwE;AAAA,QACjJ;AAAA,MACF,OAAO;AACL,cAAM,IAAI,MAAM,yCAAyC,EAAE,4EAA4E;AAAA,MACzI;AAAA,IACF,CAAC;AACD,UAAM,gBAAgB,MAAM,KAAK,MAAM,SAAS,KAAK,CAAC;AACtD,YAAQ,QAAQ;AAChB,UAAM,oBAAoB,CAAC;AAC3B,kBAAc,QAAQ,CAAC,eAAe;AACpC,UAAI,YAAY,eAAe,MAAM;AACnC,cAAM,IAAI,MAAM,wEAAwE,sDAAsD;AAAA,MAChJ,OAAO;AACL,0BAAkB,KAAK,YAAY,WAAW;AAAA,MAChD;AAAA,IACF,CAAC;AACD,WAAO,SAAS,mBAAmB,SAAS;AAAA,EAC9C,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,SAAS,qBAAqB,SAAS,eAAe;AACpD,SAAO,IAAI,SAAS,aAAa;AACnC;AAGA,IAAI,gCAAgC;AACpC,SAAS,8BAA8B,QAAQ,aAAa;AAC1D,MAAI;AACJ,MAAI;AACJ,QAAM,iBAAiB;AACvB,OAAK,eAAe;AACpB,OAAK,eAAe;AACpB,eAAa,OAAO,MAAM,QAAQ,MAAM,MAAM,MAAM,mPAAmP,aAAa;AACpT,QAAM,cAAc,0BAA0B,SAAS,OAAO,YAAY,EAAE;AAC5E,QAAM,cAAc,0BAA0B,UAAU,OAAO,aAAa,EAAE;AAC9E,QAAM,YAAY,YAAY,GAAG,MAAM;AACvC,eAAa,OAAO,YAAY,WAAW,OAAO,OAAO,QAAQ,MAAM,mBAAmB,OAAO,OAAO,2CAA2C,YAAY,yCAAyC,KAAK,UAAU,OAAO,UAAU,IAAI;AAC5O,eAAa,OAAO,YAAY,WAAW,OAAO,QAAQ,QAAQ,MAAM,mBAAmB,OAAO,QAAQ,4CAA4C,YAAY,2CAA2C,KAAK,UAAU,OAAO,WAAW,IAAI;AAClP,WAAS,SAAS,GAAG,SAAS,YAAY,QAAQ,UAAU;AAC1D,iBAAa,OAAO,YAAY,QAAQ,MAAM,OAAO,WAAW,MAAM,8BAA8B,OAAO,WAAW,eAAe,YAAY,QAAQ,MAAM,iBAAiB,4BAA4B,OAAO,WAAW,KAAK;AAAA,EACrO;AACA,WAAS,SAAS,GAAG,SAAS,YAAY,QAAQ,UAAU;AAC1D,iBAAa,OAAO,YAAY,QAAQ,MAAM,OAAO,WAAW,MAAM,+BAA+B,OAAO,YAAY,eAAe,YAAY,QAAQ,MAAM,iBAAiB,4BAA4B,OAAO,WAAW,KAAK;AAAA,EACvO;AACA,SAAO,EAAE,IAAI,aAAa,IAAI,YAAY;AAC5C;AACA,SAAS,0BAA0B,eAAe,OAAO,QAAQ;AAC/D,MAAI,kBAAkB,QAAQ;AAC5B,WAAO,CAAC,MAAM;AAAA,EAChB,WAAW,MAAM,QAAQ,MAAM,GAAG;AAChC,iBAAa,OAAO,OAAO,WAAW,MAAM,QAAQ,MAAM,wBAAwB,OAAO,gCAAgC,MAAM,uBAAuB,sBAAsB,QAAQ;AACpL,WAAO;AAAA,EACT,OAAO;AACL,UAAM,SAAS,CAAC;AAChB,eAAW,QAAQ,OAAO;AACxB,UAAI,OAAO,SAAS,MAAM;AACxB,cAAM,IAAI,WAAW,gEAAgE,sBAAsB,QAAQ;AAAA,MACrH;AACA,aAAO,KAAK,OAAO,KAAK;AAAA,IAC1B;AACA,WAAO;AAAA,EACT;AACF;AACA,SAAS,gCAAgC,MAAM;AAC7C,MAAI,KAAK,WAAW,GAAG;AACrB,UAAM,IAAI,oBAAoB,wDAAwD;AAAA,EACxF;AACA,SAAO,EAAE,IAAI,KAAK,IAAI,IAAI,KAAK,GAAG;AACpC;AACA,eAAe,WAAW,QAAQ,SAAS,MAAM;AAC/C,QAAM,qBAAqB,KAAK,mBAAmB;AACnD,eAAa,OAAO,OAAO,aAAa,MAAM,MAAM,gGAAgG;AACpJ,eAAa,OAAO,QAAQ,MAAM,MAAM,+FAA+F;AACvI,eAAa,OAAO,KAAK,UAAU,QAAQ,KAAK,SAAS,KAAK,OAAO,UAAU,KAAK,MAAM,GAAG,MAAM,iFAAiF,KAAK,QAAQ;AACjM,eAAa,OAAO,CAAC,sBAAsB,KAAK,kBAAkB,KAAK,OAAO,UAAU,KAAK,eAAe,GAAG,MAAM,uGAAuG,KAAK,iBAAiB;AAClP,eAAa;AAAA,IACX,KAAK,sBAAsB;AAAA,IAC3B,MAAM;AAAA,EACR;AACA,MAAI,OAAO,YAAY;AACrB,UAAM,IAAI,MAAM,8DAA8D;AAAA,EAChF;AACA,SAAO,aAAa;AACpB,MAAI;AACF,UAAM,eAAe,KAAK,kBAAkB;AAC5C,QAAI;AACJ,QAAI;AACJ,QAAI,cAAc;AAChB,UAAI,gBAAgB,KAAK,cAAc,GAAG;AACxC,qBAAa,OAAO,KAAK,qBAAqB,QAAQ,KAAK,oBAAoB,KAAK,OAAO,UAAU,KAAK,iBAAiB,GAAG,MAAM,iJAAiJ,KAAK,mBAAmB;AAAA,MAC/S,OAAO;AACL,cAAM,iBAAiB,gCAAgC,KAAK,cAAc;AAC1E,gBAAQ,eAAe;AACvB,gBAAQ,eAAe;AAAA,MACzB;AAAA,IACF;AACA,UAAM,gBAAgB,OAAO,kBAAkB;AAC/C,UAAM,YAAY,OAAO,uBAAuB;AAChD,QAAI;AACJ,QAAI,cAAc;AAChB,wBAAkB,UAAU,MAAM,EAAE,OAAO,UAAU,IAAI,CAAC,MAAM,SAAS,CAAC,CAAC;AAAA,IAC7E,OAAO;AACL,wBAAkB,UAAU,MAAM;AAAA,IACpC;AACA,UAAM,aAAa,qBAAqB,KAAK,WAAW,KAAK,UAAU;AACvE,UAAM,UAAU,KAAK,WAAW,OAAO,IAAI,KAAK;AAChD,UAAM,EAAE,cAAc,QAAQ,IAAI;AAAA,MAChC;AAAA,MACA;AAAA,MACA,KAAK;AAAA,MACL;AAAA,MACA;AAAA,MACA,iBAAiB,SAAS,IAAI;AAAA,MAC9B;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,iBAAa,SAAS,MAAM;AAC5B,WAAO,UAAU;AACjB,UAAM,aAAa,aAAa;AAChC,WAAO,gBAAgB;AACvB,QAAI,QAAQ,KAAK,gBAAgB,OAAO,IAAI,KAAK;AACjD,QAAI,eAAe,MAAM,QAAQ,SAAS;AAC1C,WAAO,QAAQ,KAAK,QAAQ;AAC1B,YAAM,YAAY,CAAC;AACnB,YAAM,aAAa,aAAa,KAAK;AACrC,UAAI,YAAY;AAChB,UAAI,aAAa;AACjB,UAAI,CAAC,oBAAoB;AACvB,uBAAe,MAAM,QAAQ,SAAS;AAAA,MACxC;AACA,aAAO,qBAAqB,YAAY,KAAK,kBAAkB,MAAM;AACnE,cAAM,cAAc,MAAM,aAAa,KAAK;AAC5C,YAAI,sBAAsB,YAAY,MAAM;AAC1C,kBAAQ,KAAK,uCAAuC,KAAK,oEAAoE,mJAAmJ,KAAK,kBAAkB,KAAK,wFAAwF;AACpY;AAAA,QACF;AACA,YAAI,YAAY,SAAS,MAAM;AAC7B,gBAAM,EAAE,IAAI,GAAG,IAAI,8BAA8B,QAAQ,YAAY,KAAK;AAC1E,gBAAM,YAAY,CAAC;AACnB,oBAAU,WAAW;AACrB,oBAAU,UAAU,GAAG,GAAG,MAAM;AAChC,gBAAM,aAAa,aAAa,YAAY,SAAS;AACrD,gBAAM,gBAAgB,CAAC;AACvB,cAAI,KAAK,eAAe,MAAM;AAC5B,kBAAM,uBAAuB,wBAAwB,KAAK,aAAa,OAAO,WAAW;AACzF,qBAAS,IAAI,GAAG,IAAI,qBAAqB,QAAQ,EAAE,GAAG;AACpD,4BAAc,KAAK,MAAM,mBAAmB,GAAG,IAAI,MAAM,qBAAqB,EAAE,CAAC;AAAA,YACnF;AAAA,UACF;AACA,gBAAM,MAAM,GAAG,OAAO,EAAE,EAAE,OAAO,aAAa;AAC9C,gBAAM,OAAO,cAAc,GAAG;AAC9B,kBAAQ,GAAG;AACX,mBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,kBAAM,QAAQ,UAAU;AACxB,kBAAM,MAAM,KAAK;AACjB,sBAAU,SAAS;AACnB,iBAAK,GAAG;AAAA,UACV;AACA,gBAAM,aAAa,WAAW,YAAY,SAAS;AACnD,+BAAqB,SAAS;AAC9B;AACA;AAAA,QACF;AACA,YAAI,qBAAqB,aAAa,KAAK,kBAAkB,YAAY,MAAM;AAC7E,cAAI,cAAc;AAChB,gBAAI;AACJ,gBAAI,gBAAgB,KAAK,cAAc,GAAG;AACxC,wBAAU,OAAO,MAAM,OAAO,gBAAgB,KAAK,gBAAgB,EAAE,SAAS,KAAK,kBAAkB,CAAC,CAAC;AAAA,YACzG,OAAO;AACL,wBAAU,OAAO,OAAO,SAAS,OAAO,OAAO;AAAA,gBAC7C,WAAW,KAAK,uBAAuB,OAAO,gCAAgC,KAAK;AAAA,gBACnF,SAAS;AAAA,cACX,CAAC,CAAC;AAAA,YACJ;AACA,qBAAS,IAAI,GAAG,IAAI,OAAO,aAAa,QAAQ,EAAE,GAAG;AACnD,wBAAU,OAAO,OAAO,aAAa,QAAQ,QAAQ;AAAA,YACvD;AAAA,UACF;AACA;AAAA,QACF;AACA,YAAI,OAAO,eAAe;AACxB;AAAA,QACF;AAAA,MACF;AACA,YAAM,aAAa,WAAW,OAAO,SAAS;AAC9C;AACA,UAAI,OAAO,eAAe;AACxB;AAAA,MACF;AAAA,IACF;AACA,UAAM,aAAa,WAAW;AAC9B,UAAM,OAAO,QAAQ,SAAS;AAC9B,WAAO,OAAO;AAAA,EAChB,UAAE;AACA,WAAO,aAAa;AAAA,EACtB;AACF;AACA,SAAS,iBAAiB,SAAS,MAAM;AACvC,MAAI,gBAAgB;AACpB,MAAI,KAAK,mBAAmB,MAAM;AAChC,oBAAgB,KAAK;AAAA,EACvB,WAAW,OAAO,SAAS,QAAQ,IAAI,GAAG;AACxC,oBAAgB,QAAQ;AAAA,EAC1B;AACA,SAAO;AACT;AACA,SAAS,gBAAgB,SAAS;AAChC,SAAO,OAAO,QAAQ,aAAa;AACrC;AACA,SAAS,qBAAqB,UAAU;AACtC,SAAO,OAAO,SAAS,SAAS;AAClC;AACA,eAAe,gBAAgB,QAAQ,SAAS,MAAM;AACpD,SAAO,QAAQ,CAAC;AAChB,QAAM,aAAa,KAAK,WAAW;AACnC,QAAM,IAAI,OAAO;AACjB,MAAI,OAAO,CAAC;AACZ,MAAI,KAAK,UAAU,GAAG;AACpB,UAAM,IAAI,oBAAoB,sCAAsC;AAAA,EACtE;AACA,eAAa,OAAO,CAAC,cAAc,KAAK,UAAU,KAAK,OAAO,UAAU,KAAK,OAAO,GAAG,MAAM,wEAAwE,KAAK,UAAU,KAAK,OAAO,GAAG;AACnM,QAAM,eAAe,qBAAqB,OAAO,IAAI,UAAU,MAAM,QAAQ,SAAS;AACtF,MAAI,cAAc;AAClB,MAAI,QAAQ;AACZ,SAAO,aAAa,QAAQ,KAAK,UAAU,MAAM;AAC/C,UAAM,cAAc,MAAM,aAAa,KAAK;AAC5C,WAAO,KAAK,MAAM;AAChB,UAAI,YAAY,OAAO;AACrB,cAAM,EAAE,IAAI,GAAG,IAAI,8BAA8B,QAAQ,YAAY,KAAK;AAC1E,cAAM,UAAU,GAAG,OAAO,EAAE;AAC5B,cAAM,YAAY,KAAK,MAAM,EAAE,OAAO,CAAC;AACvC,gBAAQ,OAAO;AACf,YAAI,UAAU,GAAG;AACf,mBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,iBAAK,KAAK,OAAO,CAAC,CAAC;AAAA,UACrB;AAAA,QACF;AACA,cAAM,YAAY,QAAQ,GAAG,MAAM;AACnC,iBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,gBAAM,WAAW,UAAU;AAC3B,gBAAM,YAAY,KAAK;AACvB,eAAK,KAAK,KAAK,MAAM,KAAK,KAAK,IAAI,IAAI,WAAW,QAAQ,CAAC,CAAC;AAC5D,cAAI,QAAQ,GAAG;AACb,oBAAQ,SAAS;AAAA,UACnB;AAAA,QACF;AACA,gBAAQ,SAAS;AACjB,uBAAe;AACf,UAAE;AAAA,MACJ;AACA,aAAO;AAAA,IACT,CAAC;AACD,QAAI,YAAY,MAAM;AACpB,UAAI,YAAY;AACd,gBAAQ,KAAK,gLAAgL,KAAK,yFAAyF;AAAA,MAC7R;AACA;AAAA,IACF;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,YAAY,KAAK;AACvB,SAAK,KAAK,IAAI,KAAK,IAAI,WAAW;AAClC,YAAQ,SAAS;AAAA,EACnB;AACA,SAAO,iBAAiB,IAAI;AAC9B;AAGA,SAAS,eAAe,WAAW;AACjC,eAAa,OAAO,YAAY,KAAK,OAAO,UAAU,SAAS,GAAG,MAAM,2DAA2D,WAAW;AAChJ;AACA,SAAS,YAAY,QAAQ,OAAO,MAAM;AACxC,MAAI,UAAU,MAAM;AAClB,WAAO,CAAC,IAAI;AAAA,EACd,WAAW,MAAM,QAAQ,MAAM,GAAG;AAChC,WAAO,OAAO,IAAI,CAAC,WAAW,oBAAoB,QAAQ,OAAO,OAAO,KAAK,CAAC;AAAA,EAChF,OAAO;AACL,WAAO,oBAAoB,QAAQ,OAAO,OAAO,KAAK;AAAA,EACxD;AACF;AACA,SAAS,qBAAqB,QAAQ,SAAS;AAC7C,SAAO,KAAK,MAAM;AAChB,QAAI,UAAU,MAAM;AAClB,aAAO;AAAA,IACT,WAAW,MAAM,QAAQ,MAAM,GAAG;AAChC,aAAO,OAAO,IAAI,CAAC,WAAW,qBAAqB,QAAQ,OAAO,CAAC;AAAA,IACrE,OAAO;AACL,aAAO,QAAQ,QAAQ,QAAQ,UAAU,UAAU,UAAU,KAAK,SAAS,OAAO,CAAC;AAAA,IACrF;AAAA,EACF,CAAC;AACH;AACA,SAAS,YAAY,MAAM,WAAW;AACpC,QAAM,SAAS,CAAC;AAChB,MAAI,aAAa;AACjB,MAAI,WAAW;AACf,SAAO,aAAa,MAAM;AACxB,eAAW,aAAa;AACxB,QAAI,YAAY,MAAM;AACpB,iBAAW;AAAA,IACb;AACA,WAAO,KAAK,CAAC,YAAY,QAAQ,CAAC;AAClC,iBAAa;AAAA,EACf;AACA,SAAO;AACT;AACA,eAAe,QAAQ,QAAQ,GAAG,KAAK,WAAW,WAAW,QAAQ,SAAS,YAAY,MAAM,QAAQ,UAAU,iBAAiB,cAAc,eAAe,iBAAiB;AAC/K,MAAI,aAAa,MAAM;AACrB,gBAAY;AAAA,EACd;AACA,MAAI,UAAU,MAAM;AAClB,aAAS;AAAA,EACX;AACA,MAAI,YAAY,MAAM;AACpB,eAAW;AAAA,EACb;AACA,MAAI,gBAAgB,MAAM;AACxB,mBAAe;AAAA,EACjB;AACA,MAAI,eAAe;AACnB,MAAI,QAAQ,QAAQ,UAAU,MAAM;AAClC,mBAAe;AAAA,EACjB;AACA,MAAI,mBAAmB,MAAM;AAC3B,mBAAe;AACf,QAAI,iBAAiB,MAAM;AACzB,YAAM,IAAI,WAAW,kGAAkG;AAAA,IACzH;AAAA,EACF;AACA,QAAM,kBAAkB,OAAO,gBAAgB,KAAK,WAAW,eAAe,iBAAiB;AAC/F,MAAI;AACJ,MAAI,mBAAmB,MAAM;AAC3B,iBAAa,OAAO,GAAG,eAAe;AAAA,EACxC;AACA,MAAI,WAAW,MAAM;AACnB,cAAU;AAAA,EACZ;AACA,QAAM,EAAE,cAAc,QAAQ,IAAI,mBAAmB,YAAY,SAAS,QAAQ,cAAc,iBAAiB,eAAe,WAAW,cAAc,eAAe;AACxK,eAAa,SAAS,MAAM;AAC5B,SAAO,UAAU;AACjB,QAAM,aAAa,aAAa;AAChC,SAAO,gBAAgB;AACvB,WAAS,QAAQ,cAAc,QAAQ,QAAQ,EAAE,OAAO;AACtD,UAAM,aAAa,aAAa,KAAK;AACrC,UAAM,YAAY,CAAC;AACnB,QAAI,iBAAiB,MAAM;AACzB,YAAM,IAAI,oBAAoB,4CAA4C;AAAA,IAC5E,OAAO;AACL,UAAI,aAAa,SAAS;AACxB,cAAM,IAAI,oBAAoB,wCAAwC;AAAA,MACxE,WAAW,UAAU;AACnB,qBAAa,QAAQ,UAAU;AAAA,MACjC;AACA,YAAM,oBAAoB,SAAS,UAAU;AAC7C,YAAM,UAAU,YAAY,iBAAiB,SAAS;AACtD,eAAS,aAAa,GAAG,aAAa,QAAQ,QAAQ,EAAE,YAAY;AAClE,cAAM,YAAY,CAAC;AACnB,cAAM,aAAa,aAAa,YAAY,SAAS;AACrD,aAAK,MAAM;AACT,gBAAM,aAAa,QAAQ,YAAY;AACvC,gBAAM,WAAW,QAAQ,YAAY;AACrC,gBAAM,WAAW,oBAAoB,mBAAmB,YAAY,WAAW,UAAU;AACzF,oBAAU,WAAW;AACrB,oBAAU,UAAU,WAAW;AAC/B,gBAAM,WAAW,qBAAqB,KAAK,QAAQ;AACnD,gBAAM,OAAO,EAAE,QAAQ;AACvB,mBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,kBAAM,QAAQ,UAAU;AACxB,kBAAM,MAAM,KAAK;AACjB,sBAAU,SAAS;AACnB,iBAAK,GAAG;AAAA,UACV;AACA,cAAI,eAAe,QAAQ,SAAS,GAAG;AACrC,gBAAI,cAAc;AAChB,oBAAM,UAAU,OAAO,SAAS,MAAM,QAAQ,SAAS;AACvD,uBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,sBAAM,QAAQ,UAAU;AACxB,sBAAM,MAAM,QAAQ;AACpB,qBAAK,GAAG;AACR,0BAAU,SAAS,SAAS;AAAA,cAC9B;AAAA,YACF;AAAA,UACF;AAAA,QACF,CAAC;AACD,cAAM,aAAa,WAAW,YAAY,SAAS;AACnD,6BAAqB,SAAS;AAC9B,YAAI,OAAO,eAAe;AACxB;AAAA,QACF;AAAA,MACF;AACA,wBAAkB,QAAQ;AAAA,IAC5B;AACA,UAAM,aAAa,WAAW,OAAO,SAAS;AAC9C,QAAI,OAAO,eAAe;AACxB;AAAA,IACF;AAAA,EACF;AACA,QAAM,aAAa,WAAW;AAC9B,QAAM,OAAO,QAAQ,SAAS;AAC9B,SAAO,OAAO;AAChB;AACA,eAAe,WAAW,QAAQ,GAAG,GAAG,OAAO,CAAC,GAAG;AACjD,MAAI,OAAO,YAAY;AACrB,UAAM,IAAI,MAAM,8DAA8D;AAAA,EAChF;AACA,SAAO,aAAa;AACpB,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACF,UAAM,YAAY,KAAK,aAAa,OAAO,KAAK,KAAK;AACrD,mBAAe,SAAS;AACxB,UAAM,iBAAiB;AACvB,UAAM,mBAAmB,MAAM,OAAO,oBAAoB,GAAG,GAAG,KAAK,cAAc,KAAK,aAAa,gBAAgB,SAAS;AAC9H,aAAS,iBAAiB;AAC1B,cAAU,iBAAiB;AAC3B,oBAAgB,iBAAiB;AACjC,QAAI,eAAe;AACnB,QAAI;AACJ,QAAI,KAAK,kBAAkB,QAAQ,KAAK,eAAe,SAAS,GAAG;AACjE,qBAAe;AACf,UAAI,KAAK,eAAe,WAAW,GAAG;AACpC,oBAAY,KAAK,eAAe;AAChC,oBAAY,KAAK,eAAe;AAAA,MAClC,WAAW,KAAK,eAAe,WAAW,GAAG;AAC3C,cAAM,IAAI,oBAAoB,+DAA+D;AAAA,MAC/F,OAAO;AACL,cAAM,IAAI,WAAW,0GAA0G,KAAK,4BAA4B;AAAA,MAClK;AACA,YAAM,kBAAkB;AACxB,YAAM,kBAAkB,MAAM,OAAO,oBAAoB,WAAW,WAAW,MAAM,MAAM,iBAAiB,SAAS;AACrH,aAAO,gBAAgB;AACvB,aAAO,gBAAgB;AACvB,eAAS,KAAK,OAAO,IAAI;AAAA,IAC3B,WAAW,KAAK,mBAAmB,QAAQ,KAAK,kBAAkB,KAAK,KAAK,kBAAkB,GAAG;AAC/F,qBAAe;AACf,YAAM,UAAU,KAAK,MAAM,OAAO,GAAG,MAAM,MAAM,IAAI,KAAK,gBAAgB;AAC1E,YAAM,oBAAoB,OAAO,GAAG,MAAM;AAC1C,aAAO,YAAY,QAAQ,SAAS,iBAAiB;AACrD,uBAAiB;AACjB,eAAS,YAAY,QAAQ,GAAG,OAAO;AACvC,aAAO,YAAY,SAAS,SAAS,iBAAiB;AACtD,wBAAkB;AAClB,gBAAU,YAAY,SAAS,GAAG,OAAO;AACzC,eAAS,KAAK,OAAO,IAAI;AAAA,IAC3B,WAAW,KAAK,mBAAmB,MAAM;AACvC,qBAAe;AAAA,IACjB;AACA,UAAM,MAAM,OAAO,OAAO,OAAO,EAAE,OAAO,aAAa;AACvD,WAAO,iCAAiC;AACxC,UAAM,gBAAgB,OAAO,kBAAkB;AAC/C,UAAM,YAAY,OAAO,uBAAuB;AAChD,QAAI;AACJ,QAAI;AACJ,QAAI,cAAc;AAChB,aAAO,iBAAiB;AACxB,oBAAc,OAAO;AACrB,wBAAkB,UAAU,MAAM,EAAE,OAAO,UAAU,IAAI,CAAC,MAAM,SAAS,CAAC,CAAC;AAAA,IAC7E,OAAO;AACL,oBAAc;AACd,eAAS,CAAC;AACV,wBAAkB,UAAU,MAAM;AAAA,IACpC;AACA,UAAM,aAAa,qBAAqB,KAAK,WAAW,KAAK,UAAU;AACvE,UAAM,MAAM,MAAM,QAAQ,QAAQ,eAAe,KAAK,WAAW,WAAW,KAAK,QAAQ,KAAK,SAAS,YAAY,aAAa,QAAQ,KAAK,SAAS,iBAAiB,KAAK,cAAc,MAAM,IAAI;AACpM,WAAO;AAAA,EACT,UAAE;AACA,WAAO,aAAa;AACpB,sBAAkB,QAAQ,CAAC;AAC3B,sBAAkB,SAAS,CAAC;AAC5B,sBAAkB,gBAAgB,CAAC;AACnC,sBAAkB,iBAAiB,CAAC;AACpC,sBAAkB,MAAM,SAAS;AACjC,sBAAkB,MAAM,SAAS;AACjC,QAAI,iBAAiB,MAAM;AACzB,cAAQ,aAAa;AAAA,IACvB;AAAA,EACF;AACF;AACA,SAAS,2BAA2B,SAAS;AAC3C,QAAM,OAAO,CAAC;AACd,MAAI,mBAAmB,QAAQ;AAC7B,cAAU,CAAC,OAAO;AAAA,EACpB;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAM,UAAU,QAAQ;AACxB,QAAI,QAAQ,SAAS,GAAG;AACtB,WAAK,KAAK,YAAY,SAAS,CAAC,CAAC;AAAA,IACnC,WAAW,QAAQ,SAAS,GAAG;AAC7B,YAAM,IAAI,MAAM,uEAAuE;AAAA,IACzF,OAAO;AACL,WAAK,KAAK,OAAO;AAAA,IACnB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,SAAS,YAAY;AAC9C,MAAI,WAAW,MAAM;AACnB;AAAA,EACF;AACA,QAAM,eAAe,CAAC;AACtB,MAAI,sBAAsB,QAAQ;AAChC,iBAAa,KAAK,WAAW,EAAE;AAAA,EACjC,WAAW,MAAM,QAAQ,UAAU,GAAG;AACpC,eAAW,QAAQ,CAAC,MAAM,aAAa,KAAK,EAAE,EAAE,CAAC;AAAA,EACnD,WAAW,cAAc,MAAM;AAC7B,eAAW,QAAQ,YAAY;AAC7B,YAAM,YAAY,WAAW;AAC7B,mBAAa,KAAK,UAAU,EAAE;AAAA,IAChC;AAAA,EACF;AACA,QAAM,mBAAmB,CAAC;AAC1B,MAAI,mBAAmB,QAAQ;AAC7B,QAAI,aAAa,QAAQ,QAAQ,EAAE,MAAM,IAAI;AAC3C,uBAAiB,KAAK,OAAO;AAAA,IAC/B;AAAA,EACF,WAAW,MAAM,QAAQ,OAAO,GAAG;AACjC,YAAQ,QAAQ,CAAC,MAAM;AACrB,UAAI,aAAa,QAAQ,EAAE,EAAE,MAAM,IAAI;AACrC,yBAAiB,KAAK,CAAC;AAAA,MACzB;AAAA,IACF,CAAC;AAAA,EACH,WAAW,WAAW,MAAM;AAC1B,eAAW,QAAQ,SAAS;AAC1B,YAAM,UAAU,QAAQ;AACxB,UAAI,aAAa,QAAQ,QAAQ,EAAE,MAAM,IAAI;AAC3C,yBAAiB,KAAK,OAAO;AAAA,MAC/B;AAAA,IACF;AAAA,EACF;AACA,mBAAiB,QAAQ,CAAC,MAAM;AAC9B,QAAI,CAAC,EAAE,YAAY;AACjB,QAAE,QAAQ;AAAA,IACZ;AAAA,EACF,CAAC;AACH;AAGA,SAAS,aAAa,GAAG;AACvB,SAAO,aAAa;AACtB;AACA,SAAS,YAAY,GAAG;AACtB,SAAO,MAAM,QAAQ,CAAC;AACxB;AACA,SAAS,WAAW,GAAG;AACrB,SAAO,CAAC,aAAa,CAAC,KAAK,CAAC,YAAY,CAAC;AAC3C;AACA,SAAS,qBAAqB,MAAM,OAAO,QAAQ,iBAAiB,MAAM,kBAAkB,IAAI;AAC9F,MAAI,SAAS,QAAQ,MAAM,WAAW,GAAG;AACvC,QAAI,QAAQ,MAAM;AAChB,UAAI,oBAAoB;AACxB,UAAI,YAAY,IAAI,KAAK,KAAK,SAAS,GAAG;AACxC,4BAAoB;AAAA,MACtB,WAAW,WAAW,IAAI,GAAG;AAC3B,mBAAW,OAAO,MAAM;AACtB,cAAI,KAAK,eAAe,GAAG,GAAG;AAC5B,gCAAoB;AACpB;AAAA,UACF;AAAA,QACF;AAAA,MACF,OAAO;AACL,4BAAoB;AAAA,MACtB;AACA,UAAI,mBAAmB;AACrB,cAAM,IAAI,WAAW,6BAA6B,6CAA6C,MAAM;AAAA,MACvG;AAAA,IACF;AACA,WAAO,CAAC;AAAA,EACV;AACA,MAAI,QAAQ,MAAM;AAChB,WAAO,MAAM,IAAI,CAAC,SAAS,IAAI;AAAA,EACjC;AACA,MAAI;AACJ,MAAI,WAAW,IAAI,GAAG;AACpB,WAAO;AACP,aAAS,CAAC;AACV,eAAW,QAAQ,OAAO;AACxB,UAAI,KAAK,SAAS,MAAM;AACtB,cAAM,IAAI,WAAW,yBAAyB,qCAAqC,OAAO;AAAA,MAC5F;AACA,aAAO,KAAK,KAAK,KAAK;AAAA,IACxB;AAAA,EACF,WAAW,YAAY,IAAI,GAAG;AAC5B,WAAO;AACP,QAAI,KAAK,WAAW,MAAM,QAAQ;AAChC,YAAM,IAAI,WAAW,6BAA6B,gIAAgI,MAAM,sEAAsE,MAAM;AAAA,IACtQ;AACA,aAAS;AAAA,EACX,OAAO;AACL,WAAO;AACP,QAAI,MAAM,SAAS,GAAG;AACpB,YAAM,IAAI,WAAW,aAAa,2BAA2B,MAAM,4EAA4E,KAAK,OAAO;AAAA,IAC7J;AACA,aAAS,CAAC,IAAI;AAAA,EAChB;AACA,WAAS,2BAA2B,MAAM;AAC1C,MAAI,UAAU,MAAM;AAClB,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,UAAI,OAAO,MAAM,MAAM;AACrB;AAAA,MACF;AACA,YAAM,SAAS,OAAO;AACtB,UAAI,OAAO,MAAM,WAAW,OAAO,GAAG,QAAQ;AAC5C,cAAM,IAAI,WAAW,uBAAuB,6BAA6B,MAAM,cAAc,OAAO,GAAG,iDAAiD,OAAO,OAAO;AAAA,MACxK;AACA,eAAS,IAAI,GAAG,IAAI,OAAO,GAAG,QAAQ,EAAE,GAAG;AACzC,YAAI,MAAM,KAAK,CAAC,gBAAgB;AAC9B;AAAA,QACF;AACA,cAAM,MAAM,OAAO,MAAM;AACzB,cAAM,SAAS,OAAO,GAAG;AACzB,YAAI,UAAU,QAAQ,UAAU,KAAK,QAAQ,QAAQ;AACnD,gBAAM,IAAI,WAAW,GAAG,8EAA8E,OAAO,GAAG,MAAM,GAAG,OAAO,GAAG,MAAM,4BAA4B,OAAO,GAAG,MAAM,GAAG,OAAO,GAAG,MAAM,eAAe,0CAA0C,OAAO,MAAM,iCAAiC,OAAO,MAAM,MAAM,GAAG,OAAO,MAAM,MAAM,qBAAqB,OAAO,SAAS;AAAA,QAC/X;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,QAAQ,SAAS,SAAS;AACnD,QAAM,OAAO,QAAQ,OAAO,IAAI,CAAC,WAAW,OAAO,MAAM,EAAE,CAAC;AAC5D,OAAK,KAAK;AACV,QAAM,OAAO,QAAQ,QAAQ,IAAI,CAAC,WAAW,OAAO,MAAM,EAAE,CAAC;AAC7D,OAAK,KAAK;AACV,MAAI,KAAK,SAAS,GAAG;AACnB,UAAM,IAAI,WAAW,mFAAmF,KAAK,UAAU,OAAO,IAAI,CAAC,WAAW,OAAO,KAAK,CAAC,GAAG;AAAA,EAChK;AACA,MAAI,KAAK,SAAS,GAAG;AACnB,UAAM,IAAI,WAAW,oFAAoF,KAAK,UAAU,QAAQ,IAAI,CAAC,WAAW,OAAO,KAAK,CAAC,GAAG;AAAA,EAClK;AACA,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,KAAK,CAAC,aAAa,YAAY,MAAM,IAAI,GAAG;AAC/E,UAAM,IAAI,WAAW,iFAAiF,KAAK,0BAA0B,KAAK,sBAAsB;AAAA,EAClK;AACF;AACA,SAAS,gCAAgC,SAAS,SAAS,cAAc;AACvE,QAAM,YAAY;AAAA,IAChB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAM,IAAI,QAAQ;AAClB,UAAM,OAAO,QAAQ;AACrB,UAAM,QAAQ,aAAa;AAC3B,QAAI,QAAQ,MAAM;AAChB;AAAA,IACF;AACA,QAAI,SAAS,yBAAyB;AACpC,UAAI,EAAE,MAAM,EAAE,MAAM,SAAS,OAAO,GAAG;AACrC,cAAM,IAAI,WAAW,2CAA2C,EAAE,+JAA+J;AAAA,MACnO;AAAA,IACF;AACA,QAAI,UAAU,QAAQ,IAAI,MAAM,IAAI;AAClC,YAAM,eAAe,EAAE,MAAM,MAAM,CAAC;AACpC,YAAM,cAAc,MAAM,MAAM,CAAC;AACjC,eAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,cAAM,YAAY,aAAa;AAC/B,cAAM,SAAS,YAAY;AAC3B,YAAI,UAAU,QAAQ,cAAc,QAAQ;AAC1C,gBAAM,IAAI,WAAW,8BAA8B,EAAE,2CAA2C,+FAA+F;AAAA,QACjM;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,eAAe,MAAM,OAAO,QAAQ,iBAAiB,MAAM,kBAAkB,IAAI;AACxF,MAAI;AACJ,MAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,QAAI,KAAK,WAAW,MAAM,QAAQ;AAChC,YAAM,IAAI,WAAW,6BAA6B,oIAAoI,MAAM,qCAAqC,KAAK,oBAAoB;AAAA,IAC5P;AACA,aAAS;AAAA,EACX,OAAO;AACL,QAAI,MAAM,SAAS,GAAG;AACpB,YAAM,IAAI,WAAW,qBAAqB,MAAM,UAAU,kFAAkF,KAAK,UAAU,KAAK,KAAK,IAAI;AAAA,IAC3K;AACA,aAAS,CAAC,IAAI;AAAA,EAChB;AACA,MAAI,UAAU,MAAM;AAClB,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,UAAI,OAAO,MAAM,MAAM;AACrB;AAAA,MACF;AACA,YAAM,SAAS,OAAO;AACtB,UAAI,OAAO,MAAM,WAAW,OAAO,GAAG,QAAQ;AAC5C,cAAM,IAAI,WAAW,uBAAuB,6BAA6B,MAAM,cAAc,OAAO,GAAG,iDAAiD,KAAK,UAAU,OAAO,KAAK,GAAG;AAAA,MACxL;AACA,eAAS,IAAI,GAAG,IAAI,OAAO,GAAG,QAAQ,EAAE,GAAG;AACzC,YAAI,MAAM,KAAK,CAAC,gBAAgB;AAC9B;AAAA,QACF;AACA,cAAM,MAAM,OAAO,MAAM;AACzB,cAAM,SAAS,OAAO,GAAG;AACzB,YAAI,UAAU,MAAM;AAClB,cAAI,WAAW,KAAK;AAClB,kBAAM,IAAI,WAAW,uBAAuB,6BAA6B,MAAM,oBAAoB,KAAK,UAAU,OAAO,EAAE,8BAA8B,KAAK,UAAU,OAAO,KAAK,IAAI;AAAA,UAC1L;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,eAAe,SAAS,aAAa;AAC5C,MAAI,WAAW,QAAQ,MAAM,QAAQ,OAAO,KAAK,QAAQ,WAAW,GAAG;AACrE,WAAO,YAAY,IAAI,CAAC,SAAS,CAAC,CAAC;AAAA,EACrC;AACA,MAAI;AACJ,MAAI,OAAO,YAAY,YAAY,OAAO,YAAY,YAAY;AAChE,qBAAiB,CAAC,OAAO;AAAA,EAC3B,WAAW,MAAM,QAAQ,OAAO,KAAK,OAAO,YAAY,UAAU;AAChE,qBAAiB;AAAA,EACnB,OAAO;AACL,UAAM,IAAI,UAAU,kGAAkG,SAAS;AAAA,EACjI;AACA,MAAI,MAAM,QAAQ,cAAc,GAAG;AACjC,WAAO,YAAY,IAAI,CAAC,SAAS,cAAc;AAAA,EACjD,OAAO;AACL,UAAM,gBAAgB,CAAC;AACvB,eAAW,QAAQ,aAAa;AAC9B,UAAI,gBAAgB,eAAe,eAAe,IAAI,IAAI,eAAe,QAAQ,CAAC;AAClF,UAAI,CAAC,MAAM,QAAQ,aAAa,GAAG;AACjC,wBAAgB,CAAC,aAAa;AAAA,MAChC;AACA,oBAAc,KAAK,aAAa;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AACF;AACA,IAAI,2BAA2B;AAC/B,IAAI,cAAc,cAAc,UAAU;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,aAAa;AAAA,EACpB;AAAA,EACA,QAAQ,YAAY,WAAW,UAAU,QAAQ,KAAK;AACpD,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,WAAW,8KAA8K;AAAA,IACrM;AACA,iBAAa,MAAM,YAAY,WAAW,OAAO;AAAA,EACnD;AAAA,EACA,QAAQ,MAAM;AACZ,QAAI,KAAK,QAAQ,MAAM;AACrB,WAAK,OAAO,CAAC;AAAA,IACf;AACA,SAAK,OAAO,KAAK;AACjB,QAAI,OAAO,KAAK,cAAc,UAAU;AACtC,WAAK,aAAa,aAAa,KAAK,SAAS;AAC7C,WAAK,mBAAmB;AAAA,IAC1B,OAAO;AACL,UAAI,EAAE,KAAK,qBAAqB,YAAY;AAC1C,cAAM,IAAI,WAAW,6DAA6D;AAAA,MACpF;AACA,WAAK,aAAa,KAAK;AACvB,WAAK,mBAAmB;AAAA,IAC1B;AACA,QAAI,gBAAgB,CAAC;AACrB,QAAI,CAAC,MAAM,QAAQ,KAAK,IAAI,KAAK,OAAO,KAAK,SAAS,YAAY,OAAO,KAAK,SAAS,YAAY;AACjG,WAAK,OAAO,KAAK;AACjB,iBAAW,QAAQ,KAAK,MAAM;AAC5B,YAAI,KAAK,YAAY,QAAQ,IAAI,MAAM,IAAI;AACzC,gBAAM,IAAI,WAAW,sCAAsC,4CAA4C,KAAK,aAAa;AAAA,QAC3H;AAAA,MACF;AACA,iBAAW,QAAQ,KAAK,aAAa;AACnC,YAAI,KAAK,KAAK,SAAS,MAAM;AAC3B,kBAAQ,KAAK,WAAW,gIAAgI,sBAAsB;AAAA,QAChL;AACA,sBAAc,KAAK,IAAI,KAAK,KAAK,KAAK,CAAC;AAAA,MACzC;AAAA,IACF,WAAW,MAAM,QAAQ,KAAK,IAAI,GAAG;AACnC,UAAI,KAAK,KAAK,WAAW,KAAK,QAAQ,QAAQ;AAC5C,cAAM,IAAI,WAAW,2FAA2F,KAAK,QAAQ,yCAAyC,KAAK,OAAO;AAAA,MACpL;AACA,YAAM,YAAY,KAAK;AACvB,sBAAgB,UAAU,IAAI,CAAC,MAAM,IAAI,CAAC,CAAC;AAAA,IAC7C,OAAO;AACL,YAAM,eAAe,IAAI,KAAK,IAAI;AAClC,WAAK,QAAQ,QAAQ,CAAC,MAAM;AAC1B,sBAAc,KAAK,YAAY;AAAA,MACjC,CAAC;AAAA,IACH;AACA,SAAK,gBAAgB;AACrB,SAAK,kBAAkB,CAAC;AACxB,SAAK,mBAAmB,CAAC;AACzB,SAAK,cAAc,CAAC;AACpB,aAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,QAAQ,EAAE,GAAG;AAC5C,YAAM,QAAQ,KAAK,qBAAqB;AACxC,YAAM,OAAO,KAAK,YAAY;AAC9B,WAAK,gBAAgB,KAAK,IAAI;AAC9B,WAAK,iBAAiB,KAAK,KAAK;AAChC,WAAK,YAAY,KAAK,KAAK,cAAc,EAAE;AAAA,IAC7C;AACA,UAAM,oBAAoB,CAAC;AAC3B,SAAK,UAAU,KAAK;AACpB,SAAK,eAAe,CAAC,MAAM;AAC3B,SAAK,iBAAiB,CAAC;AACvB,cAAU,QAAQ,MAAM;AACtB,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,QAAQ,EAAE,GAAG;AAC5C,YAAI,kBAAkB,QAAQ,CAAC,MAAM,IAAI;AACvC;AAAA,QACF;AACA,cAAM,eAAe,KAAK,cAAc;AACxC,YAAI,KAAK,QAAQ,SAAS,GAAG;AAC3B,eAAK,eAAe,KAAK,CAAC,cAAc,CAAC,CAAC;AAC1C,eAAK,aAAa,KAAK,KAAK,YAAY,KAAK,OAAO;AAAA,QACtD;AAAA,MACF;AAAA,IACF,CAAC;AACD,UAAM,gBAAgB,eAAe,KAAK,SAAS,KAAK,WAAW;AACnE,UAAM,eAAe,CAAC,aAAa,YAAY,iBAAiB;AAC9D,UAAI,KAAK,YAAY,SAAS,GAAG;AAC/B,qBAAa,KAAK,YAAY,eAAe,MAAM;AAAA,MACrD;AACA,WAAK,aAAa,KAAK,UAAU;AACjC,WAAK,eAAe,KAAK,CAAC,cAAc,WAAW,CAAC;AAAA,IACtD;AACA,cAAU,UAAU,MAAM;AACxB,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,QAAQ,EAAE,GAAG;AAC5C,YAAI,kBAAkB,QAAQ,CAAC,MAAM,IAAI;AACvC;AAAA,QACF;AACA,cAAM,gBAAgB,cAAc;AACpC,cAAM,gBAAgB,CAAC,YAAY;AACjC,gBAAM,mBAAmB;AACzB,cAAI;AACJ,cAAI;AACJ,cAAI;AACJ,qBAAW,UAAU,SAAS;AAC5B,gBAAI,OAAO,WAAW,YAAY,CAAC,YAAY,OAAO,gBAAgB,IAAI,EAAE,QAAQ,MAAM,MAAM,IAAI;AAClG,oBAAM,cAAc,KAAK,qBAAqB;AAC9C,kBAAI,YAAY,YAAY,SAAS,OAAO,KAAK,KAAK,cAAc,OAAO,oBAAoB;AAC7F,oBAAI,CAAC,YAAY,KAAK,EAAE,QAAQ,MAAM,MAAM,IAAI;AAC9C,0BAAQ;AAAA,gBACV,WAAW,CAAC,gBAAgB,IAAI,EAAE,QAAQ,MAAM,MAAM,IAAI;AACxD,0BAAQ;AAAA,gBACV;AAAA,cACF,WAAW,KAAK,cAAc,OAAO,+BAA+B;AAClE,oBAAI,CAAC,YAAY,KAAK,EAAE,QAAQ,MAAM,MAAM,IAAI;AAC9C,0BAAQ;AAAA,gBACV,WAAW,CAAC,gBAAgB,IAAI,EAAE,QAAQ,MAAM,MAAM,IAAI;AACxD,0BAAQ;AAAA,gBACV;AAAA,cACF,OAAO;AACL,oBAAI,CAAC,YAAY,KAAK,EAAE,QAAQ,MAAM,MAAM,IAAI;AAC9C,0BAAQ;AAAA,gBACV,WAAW,CAAC,gBAAgB,IAAI,EAAE,QAAQ,MAAM,MAAM,IAAI;AACxD,0BAAQ;AAAA,gBACV;AAAA,cACF;AACA,kBAAI;AACJ,kBAAI,CAAC,YAAY,KAAK,EAAE,QAAQ,MAAM,MAAM,IAAI;AAC9C,yBAAS;AAAA,cACX,WAAW,CAAC,gBAAgB,IAAI,EAAE,QAAQ,MAAM,MAAM,IAAI;AACxD,yBAAS;AAAA,cACX;AACA,iCAAmB;AACnB,2BAAa,mBAAmB;AAAA,YAClC,OAAO;AACL,oBAAM,WAAW,KAAK,MAAM;AAC5B,iCAAmB;AACnB,2BAAa,mBAAmB,oBAAoB,MAAM;AAAA,YAC5D;AACA,gBAAI;AACJ,sBAAU,YAAY,MAAM;AAC1B,6BAAe;AAAA,YACjB,CAAC;AACD,yBAAa,GAAG,YAAY,YAAY;AAAA,UAC1C;AAAA,QACF;AACA,sBAAc,aAAa;AAAA,MAC7B;AAAA,IACF,CAAC;AACD,SAAK,4BAA4B,KAAK;AAAA,EACxC;AAAA,EACA,mCAAmC;AACjC,QAAI,KAAK,6BAA6B,MAAM;AAC1C;AAAA,IACF;AACA,QAAI,KAAK,iBAAiB,WAAW,KAAK,0BAA0B,QAAQ;AAC1E,cAAQ,KAAK,mJAAmJ;AAAA,IAClK;AAAA,EACF;AAAA,EACA,SAAS,GAAG,GAAG,OAAO,CAAC,GAAG;AACxB,UAAM,YAAY,KAAK,aAAa,OAAO,KAAK,KAAK;AACrD,mBAAe,SAAS;AACxB,UAAM,iBAAiB;AACvB,UAAM,mBAAmB,KAAK,sBAAsB,GAAG,GAAG,gBAAgB,SAAS;AACnF,QAAI;AACF,YAAM,MAAM,iBAAiB,GAAG,OAAO,iBAAiB,EAAE;AAC1D,WAAK,iBAAiB;AACtB,YAAM,IAAI,KAAK;AACf,YAAM,WAAW,KAAK,SAAS,GAAG,KAAK,WAAW,KAAK,SAAS,KAAK,KAAK;AAC1E,aAAO,iBAAiB,QAAQ;AAAA,IAClC,UAAE;AACA,wBAAkB,iBAAiB,IAAI,CAAC;AACxC,wBAAkB,iBAAiB,IAAI,CAAC;AAAA,IAC1C;AAAA,EACF;AAAA,EACA,MAAM,gBAAgB,SAAS,MAAM;AACnC,SAAK,iBAAiB;AACtB,WAAO,gBAAgB,MAAM,SAAS,IAAI;AAAA,EAC5C;AAAA,EACA,gBAAgB,KAAK,WAAW,OAAO,YAAY,SAAS;AAC1D,QAAI;AACJ,QAAI,SAAS,MAAM;AACjB,mBAAa;AACb,UAAI,aAAa,MAAM;AACrB,cAAM,IAAI,WAAW,MAAM,yEAAyE,WAAW;AAAA,MACjH;AAAA,IACF,WAAW,OAAO,MAAM;AACtB,UAAI,MAAM,QAAQ,GAAG,GAAG;AACtB,qBAAa,IAAI,GAAG,MAAM;AAAA,MAC5B,OAAO;AACL,qBAAa,IAAI,MAAM;AAAA,MACzB;AAAA,IACF,OAAO;AACL,YAAM,IAAI,WAAW,yDAAyD,+BAA+B;AAAA,IAC/G;AACA,WAAO;AAAA,EACT;AAAA,EACA,QAAQ,QAAQ,SAAS;AACvB,QAAI,MAAM,QAAQ,OAAO,KAAK,QAAQ,WAAW,GAAG;AAClD,YAAM,IAAI,WAAW,oDAAoD;AAAA,IAC3E;AACA,UAAM,iBAAiB,MAAM,QAAQ,OAAO;AAC5C,UAAM,cAAc,iBAAiB,UAAU,CAAC,OAAO;AACvD,UAAM,wBAAwB,KAAK,wBAAwB,WAAW;AACtE,UAAM,WAAW,IAAI,SAAS;AAC9B,QAAI,kBAAkB,QAAQ;AAC5B,eAAS,CAAC,MAAM;AAAA,IAClB;AACA,QAAI,MAAM,QAAQ,MAAM,GAAG;AACzB,UAAI,OAAO,WAAW,KAAK,OAAO,QAAQ;AACxC,cAAM,IAAI,WAAW,kCAAkC,OAAO,8DAA8D,KAAK,OAAO,UAAU;AAAA,MACpJ;AACA,eAAS,IAAI,GAAG,IAAI,KAAK,OAAO,QAAQ,EAAE,GAAG;AAC3C,iBAAS,IAAI,KAAK,OAAO,IAAI,OAAO,EAAE;AAAA,MACxC;AAAA,IACF,OAAO;AACL,iBAAW,UAAU,KAAK,QAAQ;AAChC,cAAM,cAAc,OAAO,OAAO;AAClC,YAAI,eAAe,MAAM;AACvB,gBAAM,IAAI,WAAW,8CAA8C,OAAO,MAAM;AAAA,QAClF;AACA,iBAAS,IAAI,QAAQ,WAAW;AAAA,MAClC;AAAA,IACF;AACA,UAAM,iBAAiB,QAAQ,uBAAuB,QAAQ;AAC9D,WAAO,iBAAiB,iBAAiB,eAAe;AAAA,EAC1D;AAAA,EACA,wBAAwB,qBAAqB;AAC3C,UAAM,wBAAwB,aAAa,MAAM,oBAAoB,MAAM;AAC3E,QAAI,mBAAmB,oBAAoB;AAC3C,eAAW,SAAS,KAAK,QAAQ;AAC/B,YAAM,eAAe,MAAM,QAAQ,MAAM,MAAM,IAAI,MAAM,SAAS,CAAC,MAAM,MAAM;AAC/E,YAAM,mBAAmB,aAAa,IAAI,CAAC,WAAW,OAAO,IAAI;AACjE,eAAS,IAAI,GAAG,IAAI,oBAAoB,QAAQ,EAAE,GAAG;AACnD,cAAM,QAAQ,iBAAiB,QAAQ,oBAAoB,EAAE;AAC7D,YAAI,UAAU,IAAI;AAChB,gCAAsB,KAAK,aAAa;AACxC;AAAA,QACF;AACA,YAAI,qBAAqB,GAAG;AAC1B;AAAA,QACF;AAAA,MACF;AACA,UAAI,qBAAqB,GAAG;AAC1B;AAAA,MACF;AAAA,IACF;AACA,QAAI,mBAAmB,GAAG;AACxB,YAAM,iBAAiB,CAAC;AACxB,4BAAsB,QAAQ,CAAC,SAAS,MAAM;AAC5C,YAAI,WAAW,MAAM;AACnB,yBAAe,KAAK,oBAAoB,EAAE;AAAA,QAC5C;AAAA,MACF,CAAC;AACD,YAAM,IAAI,WAAW,mDAAmD,KAAK,UAAU,cAAc,GAAG;AAAA,IAC1G;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,KAAK,YAAY,IAAI,UAAU,OAAO;AAChD,WAAO,KAAK,MAAM;AAChB,YAAM,aAAa,KAAK,gBAAgB,GAAG;AAC3C,UAAI,SAAS;AACX,cAAM,IAAI,oBAAoB,+CAA+C;AAAA,MAC/E;AACA,YAAM,UAAU,YAAY,YAAY,SAAS;AACjD,YAAM,cAAc,KAAK,QAAQ,IAAI,CAAC,WAAW,CAAC,CAAC;AACnD,eAAS,aAAa,GAAG,aAAa,QAAQ,QAAQ,EAAE,YAAY;AAClE,cAAM,YAAY,KAAK,MAAM;AAC3B,gBAAM,aAAa,QAAQ,YAAY;AACvC,gBAAM,WAAW,QAAQ,YAAY;AACrC,gBAAM,WAAW,YAAY,KAAK,YAAY,QAAQ;AACtD,gBAAM,QAAQ,CAAC;AACf,cAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,qBAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,EAAE,GAAG;AACxC,oBAAM,KAAK,EAAE,KAAK,KAAK,OAAO,IAAI,OAAO,SAAS,GAAG,CAAC;AAAA,YACxD;AAAA,UACF,OAAO;AACL,kBAAM,KAAK,EAAE,KAAK,KAAK,OAAO,IAAI,OAAO,SAAS,CAAC;AAAA,UACrD;AACA,gBAAM,WAAW,IAAI,SAAS,KAAK;AACnC,iBAAO,QAAQ,KAAK,SAAS,QAAQ;AAAA,QACvC,CAAC;AACD,kBAAU,QAAQ,CAAC,UAAU,MAAM,YAAY,GAAG,KAAK,QAAQ,CAAC;AAAA,MAClE;AACA,aAAO,iBAAiB,YAAY,IAAI,CAAC,aAAa,OAAO,UAAU,CAAC,CAAC,CAAC;AAAA,IAC5E,CAAC;AAAA,EACH;AAAA,EACA,QAAQ,GAAG,OAAO,CAAC,GAAG;AACpB,UAAM,kBAAkB,2BAA2B,CAAC;AACpD,mBAAe,iBAAiB,KAAK,YAAY,KAAK,iBAAiB,KAAK;AAC5E,QAAI;AACF,YAAM,YAAY,KAAK,aAAa,OAAO,KAAK,KAAK;AACrD,qBAAe,SAAS;AACxB,aAAO,KAAK,YAAY,iBAAiB,SAAS;AAAA,IACpD,UAAE;AACA,wBAAkB,iBAAiB,CAAC;AAAA,IACtC;AAAA,EACF;AAAA,EACA,eAAe,GAAG;AAChB,mBAAe,GAAG,KAAK,YAAY,KAAK,iBAAiB,IAAI;AAC7D,UAAM,aAAa,MAAM,QAAQ,CAAC,IAAI,EAAE,KAAK,GAAG,MAAM;AACtD,WAAO,KAAK,YAAY,GAAG,SAAS;AAAA,EACtC;AAAA,EACA,sBAAsB,GAAG,GAAG,iBAAiB,MAAM,WAAW;AAC5D,QAAI,KAAK,cAAc,MAAM;AAC3B,YAAM,IAAI,aAAa,8FAA8F;AAAA,IACvH;AACA,UAAM,eAAe,CAAC;AACtB,aAAS,IAAI,GAAG,IAAI,KAAK,iBAAiB,QAAQ,EAAE,GAAG;AACrD,YAAM,cAAc,KAAK,iBAAiB;AAC1C,YAAM,SAAS,KAAK,YAAY;AAChC,UAAI,WAAW,+BAA+B;AAC5C,qBAAa,KAAK,YAAY,MAAM,GAAG,YAAY,SAAS,CAAC,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC;AAAA,MAC5E,OAAO;AACL,qBAAa,KAAK,WAAW;AAAA,MAC/B;AAAA,IACF;AACA,QAAI,qBAAqB,GAAG,KAAK,gBAAgB,KAAK,iBAAiB,OAAO,OAAO;AACrF,QAAI,qBAAqB,GAAG,KAAK,iBAAiB,cAAc,OAAO,QAAQ;AAC/E,sBAAkB,GAAG,GAAG,IAAI;AAC5B,oCAAgC,GAAG,KAAK,aAAa,KAAK,gBAAgB;AAC1E,QAAI,KAAK,YAAY,aAAa,QAAQ,YAAY,GAAG;AACvD,UAAI,EAAE,GAAG,MAAM,KAAK,cAAc,GAAG;AACnC,cAAM,IAAI,WAAW,mHAAmH,qBAAqB,EAAE,GAAG,MAAM,eAAe;AAAA,MACzL;AAAA,IACF;AACA,WAAO,CAAC,GAAG,CAAC;AAAA,EACd;AAAA,EACA,MAAM,oBAAoB,GAAG,GAAG,cAAc,aAAa,iBAAiB,MAAM,WAAW;AAC3F,UAAM,CAAC,YAAY,UAAU,IAAI,KAAK,sBAAsB,GAAG,GAAG,gBAAgB,SAAS;AAC3F,QAAI,gBAAgB,MAAM;AACxB,YAAM,IAAI,MAAM,qCAAqC;AAAA,IACvD;AACA,QAAI,wBAAwB;AAC5B,QAAI,eAAe,MAAM;AACvB,YAAM,eAAe,wBAAwB,aAAa,KAAK,WAAW;AAC1E,8BAAwB,CAAC;AACzB,eAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,8BAAsB,KAAK,MAAM,mBAAmB,WAAW,IAAI,MAAM,aAAa,EAAE,CAAC;AAAA,MAC3F;AAAA,IACF;AACA,WAAO,CAAC,YAAY,YAAY,qBAAqB;AAAA,EACvD;AAAA,EACA,SAAS,GAAG,KAAK,WAAW,UAAU,GAAG,OAAO;AAC9C,WAAO,KAAK,MAAM;AAChB,YAAM,aAAa,KAAK,gBAAgB,KAAK,WAAW,OAAO,OAAO;AACtE,YAAM,OAAO,CAAC;AACd,UAAI,UAAU,GAAG;AACf,cAAM,IAAI,oBAAoB,sCAAsC;AAAA,MACtE;AACA,UAAI,SAAS,MAAM;AACjB,cAAM,IAAI,oBAAoB,iDAAiD;AAAA,MACjF,OAAO;AACL,cAAM,UAAU,YAAY,YAAY,SAAS;AACjD,cAAM,aAAa,SAAS,OAAO,GAAG,UAAU,CAAC;AACjD,iBAAS,aAAa,GAAG,aAAa,QAAQ,QAAQ,EAAE,YAAY;AAClE,gBAAM,aAAa,QAAQ,YAAY;AACvC,gBAAM,WAAW,QAAQ,YAAY;AACrC,gBAAM,WAAW,oBAAoB,YAAY,YAAY,WAAW,UAAU;AAClF,gBAAM,WAAW,qBAAqB,KAAK,QAAQ;AACnD,gBAAM,YAAY,EAAE,QAAQ;AAC5B,cAAI,eAAe,GAAG;AACpB,qBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,mBAAK,KAAK,OAAO,CAAC,CAAC;AAAA,YACrB;AAAA,UACF;AACA,mBAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,kBAAM,WAAW,UAAU;AAC3B,iBAAK,KAAK,KAAK,KAAK,IAAI,IAAI,WAAW,YAAY,QAAQ,CAAC;AAAA,UAC9D;AAAA,QACF;AACA,iBAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,eAAK,KAAK,IAAI,KAAK,IAAI,UAAU;AAAA,QACnC;AAAA,MACF;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,yBAAyB;AACvB,UAAM,YAAY,KAAK;AACvB,UAAM,mBAAmB,CAAC;AAC1B,aAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,YAAM,QAAQ,UAAU;AACxB,UAAI,WAAW;AACf,UAAI,MAAM,WAAW,KAAK,IAAI,GAAG;AAC/B,cAAM,WAAW,MAAM,UAAU,MAAM,GAAG,CAAC,GAAG,KAAK;AACnD,oBAAY,IAAI;AAAA,MAClB;AACA,uBAAiB,KAAK,QAAQ;AAAA,IAChC;AACA,WAAO;AAAA,EACT;AAAA,EACA,oBAAoB;AAClB,WAAO,CAAC,SAAS;AACf,YAAM,aAAa,CAAC;AACpB,YAAM,SAAS,KAAK,MAAM,GAAG,KAAK,OAAO,MAAM;AAC/C,YAAM,UAAU,KAAK,MAAM,KAAK,OAAO,QAAQ,KAAK,OAAO,SAAS,KAAK,QAAQ,MAAM;AACvF,YAAM,gBAAgB,KAAK,MAAM,KAAK,OAAO,SAAS,KAAK,QAAQ,QAAQ,KAAK,OAAO,SAAS,KAAK,QAAQ,SAAS,CAAC;AACvH,YAAM,gBAAgB,CAAC;AACvB,YAAM,oBAAoB,MAAM;AAC9B,cAAM,QAAQ,CAAC;AACf,iBAAS,IAAI,GAAG,IAAI,KAAK,OAAO,QAAQ,EAAE,GAAG;AAC3C,gBAAM,KAAK,EAAE,KAAK,KAAK,OAAO,IAAI,OAAO,OAAO,GAAG,CAAC;AAAA,QACtD;AACA,cAAM,WAAW,IAAI,SAAS,KAAK;AACnC,cAAM,UAAU,QAAQ,KAAK,SAAS,UAAU,EAAE,YAAY,KAAK,CAAC;AACpE,YAAI;AACJ,iBAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,EAAE,GAAG;AAClD,gBAAM,eAAe,KAAK,cAAc;AACxC,cAAI,OAAO,aAAa,QAAQ,IAAI,QAAQ,EAAE;AAC9C,cAAI,cAAc,MAAM,MAAM;AAC5B,mBAAO,qBAAqB,MAAM,cAAc,EAAE;AAAA,UACpD;AACA,gBAAM,WAAW,KAAK,IAAI;AAC1B,qBAAW,KAAK,QAAQ;AACxB,cAAI,MAAM,GAAG;AACX,wBAAY;AAAA,UACd,OAAO;AACL,wBAAY,KAAK,WAAW,IAAI;AAAA,UAClC;AAAA,QACF;AACA,iBAAS,IAAI,GAAG,IAAI,KAAK,eAAe,QAAQ,EAAE,GAAG;AACnD,cAAI;AACJ,cAAI,KAAK,QAAQ,SAAS,KAAK,IAAI,KAAK,QAAQ,QAAQ;AACtD,6BAAiB,WAAW;AAAA,UAC9B,OAAO;AACL,kBAAM,SAAS,KAAK,eAAe,GAAG;AACtC,kBAAM,cAAc,KAAK,eAAe,GAAG;AAC3C,6BAAiB,KAAK,OAAO,QAAQ,cAAc,QAAQ,YAAY,CAAC;AAAA,UAC1E;AACA,eAAK,cAAc;AACnB,wBAAc,KAAK,cAAc;AAAA,QACnC;AACA,oBAAY,KAAK,SAAS;AAC1B,aAAK,gBAAgB,EAAE,QAAQ,CAAC,oBAAoB;AAClD,sBAAY,KAAK,WAAW,eAAe;AAAA,QAC7C,CAAC;AACD,eAAO;AAAA,MACT;AACA,YAAM,YAAY,KAAK,0BAA0B,IAAI,CAAC,UAAU,MAAM,KAAK,CAAC;AAC5E,YAAM,aAAa;AACnB,YAAM,iBAAiB,KAAK,WAAW,SAAS,mBAAmB,YAAY,SAAS;AACxF,aAAO,CAAC,cAAc,EAAE,OAAO,aAAa;AAAA,IAC9C;AAAA,EACF;AAAA,EACA,mBAAmB;AACjB,SAAK,eAAe,CAAC,SAAS;AAC5B,aAAO,KAAK,MAAM;AAChB,cAAM,aAAa,CAAC;AACpB,YAAI;AACJ,cAAM,SAAS,KAAK,MAAM,GAAG,KAAK,OAAO,MAAM;AAC/C,cAAM,UAAU,KAAK,MAAM,KAAK,OAAO,QAAQ,KAAK,OAAO,SAAS,KAAK,QAAQ,MAAM;AACvF,cAAM,QAAQ,CAAC;AACf,iBAAS,IAAI,GAAG,IAAI,KAAK,OAAO,QAAQ,EAAE,GAAG;AAC3C,gBAAM,KAAK,EAAE,KAAK,KAAK,OAAO,IAAI,OAAO,OAAO,GAAG,CAAC;AAAA,QACtD;AACA,cAAM,WAAW,IAAI,SAAS,KAAK;AACnC,cAAM,UAAU,QAAQ,KAAK,SAAS,QAAQ;AAC9C,iBAAS,IAAI,GAAG,IAAI,KAAK,cAAc,QAAQ,EAAE,GAAG;AAClD,gBAAM,eAAe,KAAK,cAAc;AACxC,gBAAM,OAAO,KAAK,aAAa,QAAQ,IAAI,QAAQ,EAAE,CAAC;AACtD,cAAI,MAAM,GAAG;AACX,wBAAY;AAAA,UACd,OAAO;AACL,wBAAY,KAAK,WAAW,IAAI;AAAA,UAClC;AACA,qBAAW,KAAK,SAAS;AAAA,QAC3B;AACA,iBAAS,IAAI,GAAG,IAAI,KAAK,eAAe,QAAQ,EAAE,GAAG;AACnD,gBAAM,SAAS,KAAK,eAAe,GAAG;AACtC,gBAAM,cAAc,KAAK,eAAe,GAAG;AAC3C,gBAAM,aAAa,KAAK,OAAO,QAAQ,cAAc,QAAQ,YAAY,CAAC;AAC1E,qBAAW,KAAK,UAAU;AAAA,QAC5B;AACA,eAAO;AAAA,MACT,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EACA,MAAM,IAAI,GAAG,GAAG,OAAO,CAAC,GAAG;AACzB,WAAO,WAAW,MAAM,GAAG,GAAG,IAAI;AAAA,EACpC;AAAA,EACA,MAAM,WAAW,SAAS,MAAM;AAC9B,WAAO,WAAW,MAAM,SAAS,IAAI;AAAA,EACvC;AAAA,EACA,MAAM,aAAa,GAAG,GAAG;AACvB,UAAM,iBAAiB,MAAM,KAAK,oBAAoB,GAAG,CAAC;AAC1D,UAAM,SAAS,eAAe;AAC9B,UAAM,UAAU,eAAe;AAC/B,UAAM,gBAAgB,KAAK,kBAAkB;AAC7C,UAAM,UAAU,cAAc,OAAO,OAAO,OAAO,CAAC;AACpD,UAAM,aAAa,CAAC;AACpB,eAAW,QAAQ,SAAS;AAC1B,YAAM,IAAI,MAAM,KAAK,KAAK;AAC1B,iBAAW,KAAK,EAAE,EAAE;AAAA,IACtB;AACA,YAAQ,OAAO;AACf,sBAAkB,eAAe,IAAI,CAAC;AACtC,sBAAkB,eAAe,IAAI,CAAC;AACtC,WAAO,iBAAiB,UAAU;AAAA,EACpC;AAAA,EACA,gBAAgB,QAAQ;AACtB,UAAM,eAAe,CAAC;AACtB,UAAM,gBAAgB,UAAU,QAAQ,OAAO;AAC/C,UAAM,UAAU,gBAAgB,KAAK,mBAAmB,KAAK;AAC7D,UAAM,eAAe,KAAK,WAAW,aAAa;AAClD,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAI,iBAAiB,CAAC,QAAQ,GAAG,WAAW;AAC1C;AAAA,MACF;AACA,mBAAa,KAAK,EAAE,MAAM,QAAQ,GAAG,cAAc,QAAQ,aAAa,GAAG,CAAC;AAAA,IAC9E;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,aAAa,MAAM;AACrB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EACA,IAAI,eAAe;AACjB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,QAAI,KAAK,eAAe,WAAW;AACjC,WAAK,aAAa;AAClB,WAAK,mBAAmB;AAAA,IAC1B;AAAA,EACF;AAAA,EACA,UAAU;AACR,UAAM,SAAS,MAAM,QAAQ;AAC7B,QAAI,OAAO,yBAAyB,KAAK,KAAK,aAAa,QAAQ,KAAK,kBAAkB;AACxF,YAAM,mCAAmC,OAAO,EAAE;AAClD,WAAK,WAAW,QAAQ;AACxB,aAAO,wBAAwB,mCAAmC,OAAO,EAAE;AAAA,IAC7E;AACA,WAAO;AAAA,EACT;AAAA,EACA,qBAAqB;AACnB,QAAI;AACJ,QAAI,OAAO,KAAK,SAAS,UAAU;AACjC,kBAAY,YAAY,KAAK,IAAI;AAAA,IACnC,WAAW,MAAM,QAAQ,KAAK,IAAI,GAAG;AACnC,iBAAW,QAAQ,KAAK,MAAM;AAC5B,YAAI,OAAO,SAAS,UAAU;AAC5B,gBAAM,IAAI,MAAM,oDAAoD;AAAA,QACtE;AAAA,MACF;AACA,kBAAY,KAAK,KAAK,IAAI,CAAC,SAAS,YAAY,IAAI,CAAC;AAAA,IACvD,OAAO;AACL,YAAM,cAAc,OAAO,KAAK,KAAK,IAAI;AACzC,kBAAY,CAAC;AACb,YAAM,UAAU,KAAK;AACrB,iBAAW,cAAc,aAAa;AACpC,YAAI,OAAO,QAAQ,gBAAgB,UAAU;AAC3C,oBAAU,cAAc,YAAY,QAAQ,WAAW;AAAA,QACzD,OAAO;AACL,gBAAM,IAAI,MAAM,oDAAoD;AAAA,QACtE;AAAA,MACF;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,uBAAuB;AACrB,QAAI,OAAO,KAAK,YAAY,YAAY,OAAO,KAAK,YAAY,YAAY;AAC1E,aAAO,CAAC,YAAY,oBAAoB,KAAK,OAAO,CAAC,CAAC;AAAA,IACxD,WAAW,MAAM,QAAQ,KAAK,OAAO,GAAG;AACtC,aAAO,KAAK,QAAQ,IAAI,CAAC,WAAW,YAAY,oBAAoB,MAAM,CAAC,CAAC;AAAA,IAC9E,OAAO;AACL,YAAM,qBAAqB,CAAC;AAC5B,iBAAW,OAAO,KAAK,SAAS;AAC9B,2BAAmB,OAAO,YAAY,oBAAoB,KAAK,QAAQ,IAAI,CAAC;AAAA,MAC9E;AACA,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,oBAAoB;AAClB,WAAO;AAAA,MACL,MAAM,KAAK,mBAAmB;AAAA,MAC9B,SAAS,KAAK,qBAAqB;AAAA,MACnC,kBAAkB;AAAA,QAChB,YAAY,KAAK,UAAU,aAAa;AAAA,QACxC,QAAQ,KAAK,UAAU,UAAU;AAAA,MACnC;AAAA,IACF;AAAA,EACF;AAAA,EACA,mBAAmB,gBAAgB;AACjC,QAAI,eAAe,oBAAoB,MAAM;AAC3C,YAAM,IAAI,MAAM,8CAA8C;AAAA,IAChE;AACA,QAAI,eAAe,gBAAgB,MAAM;AACvC,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,QAAI,eAAe,sBAAsB,MAAM;AAC7C,YAAM,IAAI,MAAM,kDAAkD;AAAA,IACpE;AACA,UAAM,WAAW,oBAAoB,eAAe,gBAAgB;AACpE,UAAM,YAAY,YAAY,QAAQ;AACtC,QAAI;AACJ,QAAI,OAAO,eAAe,SAAS,UAAU;AAC3C,aAAO,YAAY,eAAe,IAAI;AAAA,IACxC,WAAW,MAAM,QAAQ,eAAe,IAAI,GAAG;AAC7C,aAAO,eAAe,KAAK,IAAI,CAAC,cAAc,YAAY,SAAS,CAAC;AAAA,IACtE,WAAW,eAAe,QAAQ,MAAM;AACtC,aAAO,CAAC;AACR,iBAAW,OAAO,eAAe,MAAM;AACrC,aAAK,OAAO,YAAY,eAAe,KAAK,IAAI;AAAA,MAClD;AAAA,IACF;AACA,QAAI;AACJ,QAAI,MAAM,QAAQ,eAAe,OAAO,GAAG;AACzC,gBAAU,eAAe,QAAQ,IAAI,CAAC,WAAW,YAAY,MAAM,CAAC;AAAA,IACtE,WAAW,eAAe,WAAW,MAAM;AACzC,gBAAU,CAAC;AACX,iBAAW,OAAO,eAAe,SAAS;AACxC,gBAAQ,OAAO,YAAY,eAAe,QAAQ,IAAI;AAAA,MACxD;AAAA,IACF;AACA,SAAK,QAAQ,EAAE,MAAM,SAAS,UAAU,CAAC;AAAA,EAC3C;AAAA,EACA,MAAM,KAAK,cAAc,QAAQ;AAC/B,QAAI,OAAO,iBAAiB,UAAU;AACpC,YAAM,WAAW,WAAW,gBAAgB,YAAY;AACxD,UAAI,SAAS,WAAW,GAAG;AACzB,cAAM,IAAI,WAAW,0CAA0C,eAAe;AAAA,MAChF,WAAW,SAAS,SAAS,GAAG;AAC9B,cAAM,IAAI,WAAW,wBAAwB,SAAS,kCAAkC,eAAe;AAAA,MACzG;AACA,qBAAe,SAAS;AAAA,IAC1B;AACA,QAAI,aAAa,QAAQ,MAAM;AAC7B,YAAM,IAAI,WAAW,8GAA8G;AAAA,IACrI;AACA,UAAM,qBAAqB,MAAM,WAAW,cAAc,KAAK,gBAAgB,MAAM,CAAC;AACtF,UAAM,eAAe;AACrB,UAAM,YAAY;AAClB,UAAM,cAAc,KAAK,OAAO,WAAW,YAAY;AACvD,UAAM,iBAAiB;AAAA,MACrB,eAAe;AAAA,MACf,QAAQ;AAAA,MACR,aAAa,8BAA8B;AAAA,MAC3C,aAAa;AAAA,IACf;AACA,UAAM,mBAAmB,UAAU,OAAO,QAAQ,OAAO;AACzD,QAAI,oBAAoB,KAAK,aAAa,MAAM;AAC9C,qBAAe,iBAAiB,KAAK,kBAAkB;AACvD,YAAM,aAAa;AACnB,YAAM,EAAE,MAAM,qBAAqB,OAAO,qBAAqB,IAAI,MAAM,WAAW,cAAc,MAAM,KAAK,UAAU,WAAW,GAAG,UAAU;AAC/I,yBAAmB,MAAM,KAAK,GAAG,oBAAoB;AACrD,yBAAmB,OAAO,WAAW,wBAAwB,CAAC,mBAAmB,MAAM,mBAAmB,CAAC;AAAA,IAC7G;AACA,QAAI,KAAK,uBAAuB,MAAM;AACpC,YAAM,YAAY;AAClB,+BAAyB,KAAK,qBAAqB,KAAK,MAAM,SAAS;AACvE,qBAAe,sBAAsB,KAAK;AAAA,IAC5C;AACA,mBAAe,aAAa,mBAAmB;AAC/C,mBAAe,cAAc,mBAAmB;AAChD,WAAO,aAAa,KAAK,cAAc;AAAA,EACzC;AAAA,EACA,uBAAuB,qBAAqB;AAC1C,6BAAyB,qBAAqB,KAAK,IAAI;AACvD,SAAK,sBAAsB;AAAA,EAC7B;AAAA,EACA,yBAAyB;AACvB,WAAO,KAAK;AAAA,EACd;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,IAAI,aAAa,cAAc,YAAY;AAC3C;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAG9C,eAAe,cAAc,uBAAuB,eAAe;AACjE,MAAI,EAAE,mBAAmB,wBAAwB;AAC/C,4BAAwB,EAAE,eAAe,sBAAsB;AAAA,EACjE;AACA,0BAAwB;AACxB,MAAI,gBAAgB,sBAAsB;AAC1C,MAAI,cAAc,mBAAmB,MAAM;AACzC,oBAAgB,cAAc;AAAA,EAChC;AACA,QAAM,WAAW,oBAAoB,aAAa;AAClD,QAAM,SAAS,YAAY,UAAU,aAAa;AAClD,MAAI,sBAAsB,mBAAmB,MAAM;AACjD,UAAM,eAAe,MAAM,WAAW,YAAY,sBAAsB,iBAAiB,sBAAsB,YAAY,OAAO,QAAQ,IAAI,CAAC,WAAW,OAAO,YAAY,CAAC;AAC9K,UAAM,qBAAqB,CAAC;AAC5B,eAAW,UAAU,OAAO,SAAS;AACnC,yBAAmB,OAAO,gBAAgB,aAAa,OAAO;AAAA,IAChE;AACA,WAAO,YAAY,kBAAkB;AACrC,YAAQ,YAAY;AAAA,EACtB;AACA,SAAO;AACT;AACA,eAAe,gBAAgB,iBAAiB,SAAS;AACvD,MAAI,WAAW,MAAM;AACnB,cAAU,CAAC;AAAA,EACb;AACA,MAAI,OAAO,oBAAoB,UAAU;AACvC,UAAM,WAAW,WAAW,gBAAgB,iBAAiB,OAAO;AACpE,QAAI,SAAS,WAAW,GAAG;AACzB,eAAS,KAAK,WAAW,mBAAmB,iBAAiB,OAAO,CAAC;AAAA,IACvE,WAAW,SAAS,SAAS,GAAG;AAC9B,YAAM,IAAI,WAAW,wBAAwB,SAAS,kCAAkC,kBAAkB;AAAA,IAC5G;AACA,sBAAkB,SAAS;AAAA,EAC7B;AACA,SAAO,6BAA6B,iBAAiB,QAAQ,OAAO;AACtE;AACA,eAAe,6BAA6B,SAAS,eAAe,SAAS;AAC3E,MAAI,WAAW,MAAM;AACnB,cAAU,CAAC;AAAA,EACb;AACA,MAAI,QAAQ,QAAQ,MAAM;AACxB,UAAM,IAAI,WAAW,+GAA+G;AAAA,EACtI;AACA,QAAM,YAAY,MAAM,QAAQ,KAAK;AACrC,MAAI,gBAAgB,UAAU;AAC9B,MAAI,cAAc,mBAAmB,MAAM;AACzC,oBAAgB,cAAc;AAAA,EAChC;AACA,QAAM,SAAS,QAAQ,UAAU,OAAO,OAAO,QAAQ;AACvD,QAAM,iBAAiB,UAAU,cAAc,QAAQ,UAAU,eAAe,QAAQ;AACxF,QAAM,SAAS,YAAY,oBAAoB,aAAa,GAAG,eAAe,cAAc;AAC5F,QAAM,iBAAiB,UAAU;AACjC,MAAI,kBAAkB,MAAM;AAC1B,WAAO,mBAAmB,cAAc;AAAA,EAC1C;AACA,MAAI,UAAU,uBAAuB,MAAM;AACzC,WAAO,uBAAuB,UAAU,mBAAmB;AAAA,EAC7D;AACA,MAAI,UAAU,cAAc,MAAM;AAChC,QAAI,UAAU,eAAe,MAAM;AACjC,YAAM,IAAI,WAAW,gHAAgH;AAAA,IACvI;AACA,UAAM,EAAE,cAAc,iBAAiB,IAAI,+BAA+B,UAAU,YAAY,UAAU,WAAW;AACrH,WAAO,YAAY,cAAc,MAAM;AACvC,QAAI,OAAO,aAAa,QAAQ,iBAAiB,SAAS,GAAG;AAC3D,YAAM,OAAO,UAAU,WAAW,gBAAgB;AAAA,IACpD;AACA,YAAQ,YAAY;AACpB,YAAQ,iBAAiB,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAAA,EAC/C;AACA,SAAO;AACT;AACA,SAAS,+BAA+B,SAAS,OAAO;AACtD,QAAM,cAAc,WAAW,cAAc,SAAS,KAAK;AAC3D,QAAM,eAAe,CAAC;AACtB,QAAM,mBAAmB,CAAC;AAC1B,QAAM,QAAQ,CAAC,SAAS;AACtB,QAAI,KAAK,UAAU,aAAa;AAC9B,uBAAiB,KAAK,EAAE,MAAM,KAAK,MAAM,QAAQ,YAAY,KAAK,MAAM,CAAC;AAAA,IAC3E,OAAO;AACL,mBAAa,KAAK,QAAQ,YAAY,KAAK;AAAA,IAC7C;AAAA,EACF,CAAC;AACD,SAAO,EAAE,cAAc,iBAAiB;AAC1C;AACA,IAAI,aAAa,cAAc,YAAY;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM,EAAE,QAAQ,CAAC,GAAG,SAAS,CAAC,EAAE,CAAC;AACjC,WAAO,QAAQ,CAAC;AAChB,SAAK,YAAY;AACjB,SAAK,QAAQ;AACb,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,OAAO,OAAO,aAAa;AAChE,QAAI,KAAK,UAAU,MAAM;AACvB,iBAAW,SAAS,KAAK,QAAQ;AAC/B,aAAK,IAAI,KAAK;AAAA,MAChB;AAAA,IACF;AAAA,EACF;AAAA,EACA,WAAW,OAAO;AAChB,UAAM,QAAQ,MAAM,aAAa,GAAG,cAAc,GAAG;AACrD,QAAI,MAAM,KAAK,CAAC,MAAM,IAAI,CAAC,GAAG;AAC5B,YAAM,IAAI,WAAW,kDAAkD,MAAM,0BAA0B,MAAM,aAAa,GAAG,aAAa,GAAG,QAAQ;AAAA,IACvJ;AAAA,EACF;AAAA,EACA,IAAI,OAAO;AACT,UAAM,uBAAuB,iBAAiB,cAAc,iBAAiB;AAC7E,QAAI;AACJ,QAAI,sBAAsB;AACxB,mBAAa;AACb,UAAI,WAAW,QAAQ,WAAW,GAAG;AACnC,cAAM,IAAI,WAAW,uHAAuH;AAAA,MAC9I;AACA,UAAI,WAAW,OAAO,WAAW,GAAG;AAClC,cAAM,IAAI,WAAW,qHAAqH;AAAA,MAC5I;AAAA,IACF;AACA,QAAI,KAAK,QAAQ,WAAW,GAAG;AAC7B,UAAI,MAAM,aAAa,WAAW,GAAG;AACnC,YAAI,MAAM,mBAAmB,MAAM;AACjC,gBAAM,IAAI,WAAW,+FAA+F;AAAA,QACtH;AACA,cAAM,IAAI,MAAM;AAAA,UACd,YAAY,MAAM;AAAA,UAClB,OAAO,MAAM;AAAA,UACb,MAAM,MAAM,OAAO;AAAA,QACrB,CAAC;AACD,cAAM,MAAM,CAAC;AAAA,MACf;AACA,UAAI,sBAAsB;AACxB,aAAK,UAAU,WAAW;AAC1B,aAAK,SAAS,WAAW;AAAA,MAC3B,OAAO;AACL,YAAI,MAAM,aAAa,WAAW,GAAG;AACnC,gBAAM,IAAI,WAAW,gHAAgH,MAAM,kBAAkB,MAAM,aAAa,0CAA0C;AAAA,QAC5N;AACA,YAAI,MAAM,aAAa,GAAG,cAAc,WAAW,GAAG;AACpD,gBAAM,IAAI,WAAW,uHAAuH;AAAA,QAC9I;AACA,aAAK,WAAW,KAAK;AACrB,aAAK,UAAU,CAAC,MAAM,aAAa,GAAG,cAAc,EAAE;AACtD,aAAK,SAAS,gBAAgB,KAAK,QAAQ,EAAE;AAAA,MAC/C;AACA,WAAK,eAAe,CAAC;AACrB,UAAI,KAAK;AAAA,QACP,eAAe;AAAA,QACf,eAAe,CAAC;AAAA,QAChB,aAAa,CAAC;AAAA,QACd,eAAe,CAAC;AAAA,QAChB,cAAc,KAAK;AAAA,QACnB,eAAe,KAAK;AAAA,QACpB,YAAY,aAAa,MAAM,KAAK,OAAO,MAAM;AAAA,QACjD,aAAa,CAAC,IAAI;AAAA,QAClB,aAAa,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AAAA,QAC3C,cAAc,KAAK,QAAQ,GAAG;AAAA,MAChC,CAAC;AAAA,IACH,OAAO;AACL,YAAM,eAAe,MAAM,MAAM,KAAK,QAAQ,EAAE;AAChD,UAAI,MAAM,QAAQ,YAAY,GAAG;AAC/B,cAAM,IAAI,UAAU,uHAAuH;AAAA,MAC7I;AACA,WAAK,WAAW,KAAK;AACrB,WAAK,UAAU,CAAC,YAAY;AAC5B,WAAK,aAAa,GAAG,gBAAgB,KAAK;AAC1C,WAAK,aAAa,GAAG,eAAe,CAAC,KAAK,QAAQ,GAAG,KAAK;AAAA,IAC5D;AACA,SAAK,OAAO,KAAK,KAAK;AACtB,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,MAAM;AACJ,QAAI,KAAK,OAAO,WAAW,GAAG;AAC5B,YAAM,IAAI,UAAU,mCAAmC;AAAA,IACzD;AACA,SAAK,OAAO,IAAI;AAChB,QAAI,KAAK,OAAO,WAAW,GAAG;AAC5B,WAAK,UAAU,CAAC;AAChB,WAAK,eAAe,CAAC;AACrB,WAAK,gBAAgB,CAAC;AAAA,IACxB,OAAO;AACL,YAAM,iBAAiB,KAAK,OAAO,SAAS;AAC5C,WAAK,OAAO,gBAAgB,gBAAgB,CAAC;AAC7C,WAAK,UAAU,CAAC,KAAK,OAAO,gBAAgB,MAAM;AAClD,WAAK,aAAa,GAAG,gBAAgB,KAAK;AAC1C,WAAK,aAAa,GAAG,eAAe,CAAC,KAAK,QAAQ,GAAG,KAAK;AAAA,IAC5D;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM;AAAA,IACb;AACA,WAAO,KAAK,MAAM,KAAK,QAAQ,MAAM;AAAA,EACvC;AAAA,EACA,MAAM,YAAY;AAChB,uBAAmB,UAAU;AAC7B,QAAI,KAAK,OAAO,WAAW,KAAK,KAAK,QAAQ,WAAW,GAAG;AACzD,YAAM,IAAI,UAAU,0EAA0E;AAAA,IAChG;AACA,SAAK,QAAQ,IAAI,YAAY;AAAA,MAC3B,QAAQ,KAAK;AAAA,MACb,SAAS,KAAK,QAAQ;AAAA,MACtB,MAAM,KAAK,OAAO;AAAA,IACpB,CAAC;AACD,SAAK,MAAM,YAAY,KAAK;AAC5B,SAAK,kBAAkB,KAAK,MAAM;AAClC,SAAK,cAAc,KAAK,MAAM;AAC9B,SAAK,yBAAyB,KAAK,MAAM;AACzC,SAAK,2BAA2B,KAAK,MAAM;AAC3C,SAAK,eAAe,KAAK,MAAM;AAC/B,SAAK,0BAA0B,KAAK,MAAM;AAC1C,SAAK,4BAA4B,KAAK,MAAM;AAC5C,SAAK,eAAe,KAAK,MAAM;AAC/B,SAAK,iBAAiB,KAAK,MAAM;AACjC,SAAK,cAAc,KAAK,MAAM;AAC9B,SAAK,aAAa,KAAK,MAAM;AAC7B,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,cAAc;AACZ,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,MAAM;AAAA,IACb;AACA,WAAO,MAAM,YAAY;AAAA,EAC3B;AAAA,EACA,QAAQ,YAAY,WAAW,UAAU,QAAQ,KAAK;AACpD,QAAI,CAAC,KAAK,OAAO;AACf,WAAK,MAAM;AAAA,IACb;AACA,UAAM,QAAQ,YAAY,WAAW,OAAO;AAAA,EAC9C;AAAA,EACA,WAAW,SAAS;AAClB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM;AAAA,IACb;AACA,SAAK,MAAM,WAAW,OAAO;AAAA,EAC/B;AAAA,EACA,SAAS,GAAG,GAAG,OAAO,CAAC,GAAG;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,aAAa,mDAAmD;AAAA,IAC5E;AACA,WAAO,KAAK,MAAM,SAAS,GAAG,GAAG,IAAI;AAAA,EACvC;AAAA,EACA,MAAM,gBAAgB,SAAS,MAAM;AACnC,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,aAAa,mDAAmD;AAAA,IAC5E;AACA,WAAO,KAAK,MAAM,gBAAgB,SAAS,IAAI;AAAA,EACjD;AAAA,EACA,QAAQ,GAAG,OAAO,CAAC,GAAG;AACpB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM;AAAA,IACb;AACA,WAAO,KAAK,MAAM,QAAQ,GAAG,IAAI;AAAA,EACnC;AAAA,EACA,eAAe,GAAG;AAChB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM;AAAA,IACb;AACA,WAAO,KAAK,MAAM,eAAe,CAAC;AAAA,EACpC;AAAA,EACA,QAAQ,MAAM;AACZ,SAAK,MAAM;AACX,SAAK,MAAM,QAAQ,IAAI;AACvB,SAAK,aAAa,KAAK,MAAM;AAC7B,SAAK,mBAAmB,KAAK,MAAM;AACnC,SAAK,OAAO,KAAK,MAAM;AACvB,SAAK,UAAU,KAAK,MAAM;AAC1B,SAAK,iBAAiB,KAAK,MAAM;AACjC,SAAK,eAAe,KAAK,MAAM;AAAA,EACjC;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK,SAAS,OAAO,SAAS,KAAK,MAAM;AAAA,EAClD;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,SAAK,MAAM,YAAY;AAAA,EACzB;AAAA,EACA,MAAM,IAAI,GAAG,GAAG,OAAO,CAAC,GAAG;AACzB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,aAAa,mDAAmD;AAAA,IAC5E;AACA,WAAO,KAAK,MAAM,IAAI,GAAG,GAAG,IAAI;AAAA,EAClC;AAAA,EACA,MAAM,WAAW,SAAS,MAAM;AAC9B,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,aAAa,mDAAmD;AAAA,IAC5E;AACA,WAAO,KAAK,MAAM,WAAW,SAAS,IAAI;AAAA,EAC5C;AAAA,EACA,MAAM,aAAa,GAAG,GAAG;AACvB,WAAO,KAAK,MAAM,aAAa,GAAG,CAAC;AAAA,EACrC;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ,gBAAgB,CAAC,GAAG,iBAAiB,OAAO;AACzE,QAAI;AACJ,QAAI,mBAAmB,CAAC;AACxB,QAAI,kBAAkB,OAAO;AAC3B,UAAI,EAAE,OAAO,GAAG,aAAa,SAAS,OAAO,GAAG,iBAAiB,SAAS;AACxE,cAAM,IAAI,WAAW,gDAAgD;AAAA,MACvE;AACA,oBAAc;AAAA,IAChB,OAAO;AACL,mBAAa,OAAO,OAAO,aAAa,MAAM,MAAM,qHAAqH;AACzK,oBAAc,OAAO;AACrB,aAAO,OAAO;AACd,yBAAmB;AAAA,IACrB;AACA,UAAM,SAAS,IAAI,IAAI,gBAAgB;AACvC,QAAI,EAAE,kBAAkB,aAAa;AACnC,YAAM,IAAI,oBAAoB,yDAAyD,QAAQ;AAAA,IACjG;AACA,eAAW,QAAQ,aAAa;AAC9B,YAAM,iBAAiB;AACvB,YAAM,QAAQ,YAAY,MAAM,gBAAgB,cAAc;AAC9D,UAAI,gBAAgB;AAClB,cAAM,6BAA6B,IAAI;AAAA,MACzC;AACA,aAAO,IAAI,KAAK;AAAA,IAClB;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,aAAa,MAAM;AACrB,QAAI,KAAK,SAAS,MAAM;AACtB,YAAM,IAAI,WAAW,mFAAmF;AAAA,IAC1G;AACA,SAAK,MAAM,eAAe;AAAA,EAC5B;AAAA,EACA,IAAI,eAAe;AACjB,QAAI,KAAK,SAAS,MAAM;AACtB,YAAM,IAAI,WAAW,mFAAmF;AAAA,IAC1G;AACA,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,YAAY;AACV,UAAM,SAAS,CAAC;AAChB,eAAW,SAAS,KAAK,QAAQ;AAC/B,YAAM,OAAO,CAAC;AACd,WAAK,eAAe,MAAM,aAAa;AACvC,WAAK,YAAY,MAAM,UAAU;AACjC,aAAO,KAAK,IAAI;AAAA,IAClB;AACA,WAAO,EAAE,MAAM,KAAK,MAAM,OAAO;AAAA,EACnC;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAG9C,SAAS,MAAM,MAAM;AACnB,SAAO,IAAI,YAAY,IAAI;AAC7B;AACA,SAAS,WAAW,QAAQ;AAC1B,SAAO,IAAI,WAAW,MAAM;AAC9B;AACA,SAAS,MAAM,QAAQ;AACrB,SAAO,MAAM,MAAM;AACrB;AACA,SAAS,4BAA4B,gBAAgB,qBAAqB;AACxE,8BAA4B,4BAA4B,gBAAgB,mBAAmB;AAC7F;AAGA,IAAI,aAAa,cAAc,sBAAsB,aAAa;AAAA,EAChE,YAAY;AACV,WAAO,CAAC;AAAA,EACV;AACF;AACA,IAAI,OAAO,cAAc,WAAW;AAAA,EAClC,MAAM,GAAG,QAAQ,GAAG;AAClB,WAAO,KAAK,GAAG,KAAK;AAAA,EACtB;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,IAAI,QAAQ,cAAc,WAAW;AAAA,EACnC,MAAM,GAAG;AACP,WAAO,KAAK,CAAC;AAAA,EACf;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,QAAQ,cAAc,WAAW;AAAA,EACnC,MAAM,GAAG;AACP,WAAO,KAAK,CAAC;AAAA,EACf;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,SAAS,cAAc,WAAW;AAAA,EACpC,MAAM,GAAG;AACP,WAAO,KAAK,MAAM,QAAQ,GAAG,KAAK,CAAC,CAAC,CAAC;AAAA,EACvC;AACF;AACA,OAAO,YAAY;AACnB,sBAAsB,cAAc,MAAM;AAC1C,IAAI,SAAS,cAAc,WAAW;AAAA,EACpC,MAAM,GAAG;AACP,WAAO;AAAA,EACT;AACF;AACA,OAAO,YAAY;AACnB,sBAAsB,cAAc,MAAM;AAC1C,IAAI,WAAW,cAAc,WAAW;AAAA,EACtC,MAAM,GAAG;AACP,WAAO,QAAQ,CAAC;AAAA,EAClB;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,cAAc,cAAc,WAAW;AAAA,EACzC,MAAM,GAAG;AACP,WAAO,YAAY,CAAC;AAAA,EACtB;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,IAAI,YAAY,cAAc,WAAW;AAAA,EACvC,MAAM,GAAG;AACP,WAAO,SAAS,CAAC;AAAA,EACnB;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,IAAI,WAAW,cAAc,WAAW;AAAA,EACtC,MAAM,GAAG;AACP,WAAO,SAAS,CAAC;AAAA,EACnB;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,QAAQ,cAAc,WAAW;AAAA,EACnC,MAAM,GAAG;AACP,WAAO,MAAM,CAAC;AAAA,EAChB;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,WAAW,cAAc,WAAW;AAAA,EACtC,MAAM,GAAG,OAAO,IAAI;AAClB,WAAO,QAAQ,GAAG,IAAI;AAAA,EACxB;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,cAAc,cAAc,WAAW;AAAA,EACzC,MAAM,GAAG,OAAO,IAAI;AAClB,WAAO,WAAW,GAAG,IAAI;AAAA,EAC3B;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,IAAI,QAAQ,cAAc,WAAW;AAAA,EACnC,MAAM,GAAG,QAAQ,GAAG;AAClB,WAAO,KAAK,MAAM,IAAI,QAAQ,IAAI,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC;AAAA,EAClD;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,OAAO,cAAc,WAAW;AAAA,EAClC,MAAM,GAAG;AACP,WAAO,KAAK,MAAM,IAAI,GAAG,MAAM,SAAS,CAAC,CAAC,CAAC,CAAC;AAAA,EAC9C;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,SAAS,oBAAoB,aAAa;AACxC,SAAO,YAAY,aAAa;AAClC;AACA,SAAS,sBAAsB,QAAQ,gBAAgB,CAAC,GAAG;AACzD,SAAO,uBAAuB,QAAQ,sBAAsB,iBAAiB,OAAO,EAAE,cAAc,eAAe,YAAY;AACjI;AACA,SAAS,cAAc,YAAY;AACjC,MAAI,cAAc,MAAM;AACtB,UAAM,SAAS,CAAC;AAChB,WAAO,eAAe;AACtB,WAAO,YAAY,CAAC;AACpB,WAAO,sBAAsB,MAAM;AAAA,EACrC;AACA,MAAI,OAAO,eAAe,UAAU;AAClC,UAAM,SAAS,CAAC;AAChB,WAAO,eAAe;AACtB,WAAO,YAAY,CAAC;AACpB,WAAO,sBAAsB,MAAM;AAAA,EACrC,WAAW,sBAAsB,YAAY;AAC3C,WAAO;AAAA,EACT,OAAO;AACL,WAAO,sBAAsB,UAAU;AAAA,EACzC;AACF;AAGA,SAAS,iBAAiB,MAAM;AAC9B,MAAI,QAAQ,QAAQ,OAAO,SAAS,UAAU;AAC5C,UAAM,IAAI,MAAM,yFAAyF,MAAM;AAAA,EACjH;AACF;AACA,IAAI,cAAc,cAAc,sBAAsB,aAAa;AACnE;AACA,IAAI,OAAO,cAAc,YAAY;AAAA,EACnC,YAAY,MAAM;AAChB,UAAM;AACN,qBAAiB,IAAI;AACrB,SAAK,KAAK,QAAQ,QAAQ,KAAK,MAAM,OAAO,OAAO,KAAK;AACxD,SAAK,KAAK,QAAQ,QAAQ,KAAK,MAAM,OAAO,OAAO,KAAK;AACxD,SAAK,QAAQ,KAAK,OAAO;AACzB,SAAK,QAAQ,KAAK,OAAO;AAAA,EAC3B;AAAA,EACA,MAAM,GAAG;AACP,WAAO,KAAK,MAAM;AAChB,UAAI,iBAAiB,MAAM,CAAC,CAAC,CAAC;AAC9B,UAAI,KAAK,OAAO;AACd,yBAAiB,KAAK,gBAAgB,KAAK,IAAI,KAAK,IAAI,IAAI,CAAC,CAAC,CAAC,CAAC;AAAA,MAClE;AACA,UAAI,KAAK,OAAO;AACd,yBAAiB,KAAK,gBAAgB,KAAK,IAAI,KAAK,IAAI,QAAQ,CAAC,CAAC,CAAC,CAAC;AAAA,MACtE;AACA,aAAO,QAAQ,gBAAgB,CAAC,CAAC;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,WAAO,EAAE,MAAM,KAAK,IAAI,MAAM,KAAK,GAAG;AAAA,EACxC;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,EAAE,IAAI,OAAO,OAAO,IAAI,OAAO,MAAM,CAAC;AAAA,EACvD;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,SAAS,GAAG,MAAM;AAChB,mBAAiB,IAAI;AACrB,SAAO,IAAI,KAAK,EAAE,IAAI,QAAQ,OAAO,KAAK,KAAK,MAAM,IAAI,EAAE,CAAC;AAC9D;AACA,SAAS,GAAG,MAAM;AAChB,mBAAiB,IAAI;AACrB,SAAO,IAAI,KAAK,EAAE,IAAI,QAAQ,OAAO,KAAK,KAAK,MAAM,IAAI,EAAE,CAAC;AAC9D;AACA,IAAI,6CAA6C;AAAA,EAC/C,QAAQ;AACV;AACA,SAAS,qBAAqB,YAAY;AACxC,SAAO,qBAAqB,UAAU;AACxC;AACA,SAAS,uBAAuB,QAAQ,gBAAgB,CAAC,GAAG;AAC1D,SAAO,uBAAuB,QAAQ,sBAAsB,iBAAiB,OAAO,EAAE,cAAc,eAAe,aAAa;AAClI;AACA,SAAS,eAAe,YAAY;AAClC,MAAI,cAAc,MAAM;AACtB,WAAO;AAAA,EACT;AACA,MAAI,OAAO,eAAe,UAAU;AAClC,UAAM,YAAY,cAAc,6CAA6C,2CAA2C,cAAc;AACtI,UAAM,SAAS,EAAE,WAAW,QAAQ,CAAC,EAAE;AACvC,WAAO,uBAAuB,MAAM;AAAA,EACtC,WAAW,sBAAsB,aAAa;AAC5C,WAAO;AAAA,EACT,OAAO;AACL,WAAO,uBAAuB,UAAU;AAAA,EAC1C;AACF;AAGA,IAAI,OAAO,cAAc,MAAM;AAAA,EAC7B,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,kBAAkB;AACvB,QAAI,QAAQ,MAAM;AAChB,WAAK,WAAW,KAAK;AAAA,IACvB;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,aAAS,oBAAoB,MAAM;AACnC,QAAI,SAAS,KAAK,MAAM;AACxB,QAAI,KAAK,YAAY,MAAM;AACzB,eAAS,YAAY,QAAQ,GAAG,KAAK,QAAQ;AAAA,IAC/C;AACA,WAAO;AAAA,EACT;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,UAAU,KAAK,SAAS;AACzC,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,gBAAgB;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,QAAQ,KAAK,SAAS,OAAO,KAAK,gBAAgB,KAAK;AAAA,EAC9D;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB,MAAM;AACpC,WAAO,UAAU,GAAG,KAAK,KAAK;AAAA,EAChC;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,OAAO,KAAK,MAAM;AACnC,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,IAAI,QAAQ,cAAc,MAAM;AAAA,EAC9B,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,4BAA4B;AACjC,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,kBAAkB;AACvB,SAAK,mBAAmB,eAAe,KAAK,oBAAoB,KAAK,yBAAyB;AAC9F,SAAK,mBAAmB,eAAe,KAAK,gBAAgB;AAC5D,SAAK,kBAAkB,cAAc,KAAK,eAAe;AACzD,QAAI,KAAK,cAAc,MAAM;AAC3B,WAAK,aAAa;AAAA,IACpB,WAAW,MAAM,QAAQ,KAAK,UAAU,GAAG;AACzC,WAAK,aAAa,KAAK;AAAA,IACzB,WAAW,OAAO,KAAK,eAAe,UAAU;AAC9C,WAAK,aAAa,CAAC,KAAK,UAAU;AAAA,IACpC,OAAO;AACL,YAAM,IAAI,WAAW,sEAAsE,KAAK,YAAY;AAAA,IAC9G;AAAA,EACF;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,aAAa,WAAW,MAAM,CAAC;AACrC,QAAI,KAAK,cAAc,MAAM;AAC3B,iBAAW,KAAK,KAAK,YAAY;AAC/B,mBAAW,IAAI,KAAK;AAAA,MACtB;AAAA,IACF;AACA,SAAK,QAAQ,KAAK,UAAU,SAAS,YAAY,WAAW,KAAK,kBAAkB,KAAK,kBAAkB,MAAM,KAAK,eAAe;AACpI,UAAM,OAAO,CAAC;AACd,QAAI,KAAK,cAAc,MAAM;AAC3B,eAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,aAAK,KAAK,WAAW;AAAA,MACvB;AAAA,IACF;AACA,SAAK,YAAY,CAAC,IAAI,UAAU;AAAA,MAC9B,MAAM,WAAW;AAAA,MACjB;AAAA,IACF,CAAC,CAAC;AACF,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,aAAS,oBAAoB,MAAM;AACnC,WAAO,MAAM,QAAQ,KAAK,MAAM,KAAK,CAAC;AAAA,EACxC;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,MAC5D,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,MAC5D,iBAAiB,oBAAoB,KAAK,eAAe;AAAA,MACzD,YAAY,KAAK;AAAA,IACnB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,MAAM,cAAc,MAAM;AAAA,EAC5B,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,gBAAgB;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,QAAI,KAAK,SAAS,QAAQ,KAAK,UAAU,KAAK,eAAe;AAC3D,YAAM,IAAI,oBAAoB,4BAA4B,KAAK,+CAA+C;AAAA,IAChH;AACA,SAAK,QAAQ,KAAK,SAAS,OAAO,KAAK,gBAAgB,KAAK;AAAA,EAC9D;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB,MAAM;AACpC,WAAO,IAAI,CAAC;AAAA,EACd;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,OAAO,KAAK,MAAM;AACnC,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,YAAY;AAChB,sBAAsB,cAAc,GAAG;AACvC,IAAI,kBAAkB,cAAc,MAAM;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,gBAAgB;AACrB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,QAAQ,KAAK,SAAS,OAAO,KAAK,gBAAgB,KAAK;AAAA,EAC9D;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB,MAAM;AACpC,WAAO,IAAI,GAAG,KAAK,QAAQ,GAAG,KAAK,KAAK,GAAG,SAAS,CAAC;AAAA,EACvD;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,OAAO,KAAK,MAAM;AACnC,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,eAAe;AACpB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,UAAU,IAAI,SAAS,EAAE;AAC9B,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,eAAe,KAAK;AAAA,EAC3D;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB,MAAM;AACpC,WAAO,KAAK,QAAQ,GAAG,KAAK,IAAI;AAAA,EAClC;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,MAAM,KAAK,KAAK;AACjC,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAG5C,SAAS,eAAe,OAAO,GAAG,MAAM;AACtC,MAAI,OAAO,UAAU,UAAU;AAC7B,WAAO,aAAa,OAAO,CAAC;AAAA,EAC9B,OAAO;AACL,QAAI,MAAM,WAAW,GAAG;AACtB,YAAM,IAAI,WAAW,OAAO,gDAAgD,yBAAyB,MAAM,kBAAkB;AAAA,IAC/H;AACA,aAAS,IAAI,GAAG,IAAI,GAAG,EAAE,GAAG;AAC1B,YAAM,cAAc,MAAM;AAC1B,UAAI,CAAC,UAAU,WAAW,GAAG;AAC3B,cAAM,IAAI,WAAW,OAAO,gDAAgD,yBAAyB,KAAK,UAAU,KAAK,oCAAoC,aAAa;AAAA,MAC5K;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;AACA,SAAS,iBAAiB,aAAa,YAAY,SAAS,QAAQ,WAAW,GAAG;AAChF,MAAI,eAAe,MAAM;AACvB,WAAO;AAAA,EACT;AACA,QAAM,oBAAoB,cAAc,aAAa,MAAM,WAAW;AACtE,MAAI;AACJ,MAAI,YAAY,QAAQ;AACtB,mBAAe;AAAA,EACjB,OAAO;AACL,mBAAe,cAAc,oBAAoB;AAAA,EACnD;AACA,SAAO,KAAK,OAAO,eAAe,SAAS,KAAK,MAAM;AACxD;AACA,SAAS,aAAa,SAAS,YAAY,YAAY,SAAS;AAC9D,MAAI,WAAW,MAAM;AACnB,WAAO;AAAA,EACT;AACA,MAAI,YAAY,SAAS;AACvB,cAAU,UAAU,aAAa,KAAK,CAAC,aAAa,YAAY,CAAC,CAAC;AAAA,EACpE,WAAW,YAAY,QAAQ;AAC7B,cAAU,UAAU;AAAA,EACtB,OAAO;AACL,UAAM,IAAI,WAAW,2BAA2B,UAAU;AAAA,EAC5D;AACA,SAAO;AACT;AAGA,SAAS,sBAAsB,GAAG,YAAY;AAC5C,SAAO,KAAK,MAAM;AAChB,oBAAgB,UAAU;AAC1B,QAAI,eAAe,iBAAiB;AAClC,aAAO,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAClC,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF,CAAC;AACH;AACA,SAAS,sBAAsB,GAAG,YAAY;AAC5C,SAAO,KAAK,MAAM;AAChB,oBAAgB,UAAU;AAC1B,QAAI,eAAe,iBAAiB;AAClC,aAAO,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IACrC,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,GAAG,QAAQ,MAAM,UAAU,GAAG,UAAU,SAAS,YAAY,eAAe,GAAG;AACrG,SAAO,KAAK,MAAM;AAChB,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,oBAAgB,UAAU;AAC1B,QAAI,EAAE,MAAM,WAAW,GAAG;AACxB,YAAM,IAAI,WAAW,+DAA+D,EAAE,MAAM,iBAAiB;AAAA,IAC/G;AACA,QAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,YAAM,IAAI,WAAW,iEAAiE,OAAO,MAAM,gBAAgB;AAAA,IACrH;AACA,QAAI,QAAQ,QAAQ,KAAK,MAAM,WAAW,GAAG;AAC3C,YAAM,IAAI,WAAW,+DAA+D,OAAO,MAAM,gBAAgB;AAAA,IACnH;AACA,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,CAAC,CAAC;AAAA,IAC5B;AACA,QAAI,YAAY,UAAU;AACxB,YAAM,IAAI,oBAAoB,+EAA+E;AAAA,IAC/G;AACA,QAAI,IAAI,OAAO,GAAG,QAAQ,SAAS,YAAY,SAAS,SAAS,SAAS,OAAO,YAAY;AAC7F,QAAI,QAAQ,MAAM;AAChB,UAAI,QAAQ,GAAG,IAAI;AAAA,IACrB;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,SAAS,yBAAyB,GAAG,QAAQ,MAAM,UAAU,CAAC,GAAG,CAAC,GAAG,UAAU,SAAS,YAAY,cAAc,cAAc,MAAM;AACpI,SAAO,KAAK,MAAM;AAChB,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,oBAAgB,UAAU;AAC1B,QAAI,EAAE,SAAS,KAAK,EAAE,SAAS,GAAG;AAChC,YAAM,IAAI,WAAW,6EAA6E,EAAE,OAAO;AAAA,IAC7G;AACA,QAAI,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG;AAC1C,YAAM,IAAI,WAAW,8EAA8E,EAAE,OAAO;AAAA,IAC9G;AACA,QAAI,IAAI,sBAAsB,GAAG,UAAU;AAC3C,QAAI,YAAY,UAAU;AACxB,YAAM,IAAI,oBAAoB,+EAA+E;AAAA,IAC/G;AACA,QAAI,kBAAkB,OAAO;AAAA,MAC3B,GAAG;AAAA,MACH,QAAQ;AAAA,MACR;AAAA,MACA,KAAK,YAAY,SAAS,SAAS;AAAA,MACnC,WAAW;AAAA,MACX,YAAY;AAAA,MACZ;AAAA,MACA,YAAY;AAAA,IACd,CAAC;AACD,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAC/B;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,SAAS,eAAe,GAAG,QAAQ,MAAM,UAAU,CAAC,GAAG,GAAG,CAAC,GAAG,UAAU,SAAS,YAAY,cAAc;AACzG,SAAO,KAAK,MAAM;AAChB,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,oBAAgB,UAAU;AAC1B,QAAI,EAAE,SAAS,KAAK,EAAE,SAAS,GAAG;AAChC,YAAM,IAAI,WAAW,mEAAmE,EAAE,OAAO;AAAA,IACnG;AACA,QAAI,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG;AAC1C,YAAM,IAAI,WAAW,oEAAoE,EAAE,OAAO;AAAA,IACpG;AACA,QAAI,IAAI,sBAAsB,GAAG,UAAU;AAC3C,QAAI,YAAY,UAAU;AACxB,YAAM,IAAI,oBAAoB,+EAA+E;AAAA,IAC/G;AACA,QAAI,OAAO,GAAG,QAAQ,SAAS,YAAY,SAAS,SAAS,SAAS,SAAS,YAAY;AAC3F,QAAI,QAAQ,MAAM;AAChB,UAAI,QAAQ,GAAG,IAAI;AAAA,IACrB;AACA,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAClC;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM,MAAM;AACtB,UAAM,IAAI;AACV,SAAK,OAAO;AACZ,SAAK,6BAA6B;AAClC,SAAK,2BAA2B;AAChC,aAAS,WAAW,IAAI;AACxB,SAAK,OAAO;AACZ,0BAAsB,KAAK,MAAM,MAAM;AACvC,QAAI,KAAK,SAAS,KAAK,KAAK,SAAS,KAAK,KAAK,SAAS,GAAG;AACzD,YAAM,IAAI,oBAAoB,qDAAqD,KAAK,+BAA+B;AAAA,IACzH;AACA,SAAK,aAAa,eAAe,KAAK,YAAY,MAAM,YAAY;AACpE,SAAK,UAAU,eAAe,KAAK,WAAW,OAAO,IAAI,KAAK,SAAS,MAAM,SAAS;AACtF,SAAK,UAAU,KAAK,WAAW,OAAO,UAAU,KAAK;AACrD,qBAAiB,KAAK,OAAO;AAC7B,SAAK,aAAa,KAAK,cAAc,OAAO,iBAAiB,KAAK;AAClE,oBAAgB,KAAK,UAAU;AAC/B,SAAK,aAAa,cAAc,KAAK,UAAU;AAC/C,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,KAAK,wBAAwB;AAC3F,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,sBAAsB,eAAe,KAAK,mBAAmB;AAClE,SAAK,eAAe,eAAe,KAAK,gBAAgB,OAAO,IAAI,KAAK,cAAc,MAAM,cAAc;AAC1G,QAAI,KAAK,SAAS,MAAM,MAAM,QAAQ,KAAK,YAAY,KAAK,KAAK,aAAa,WAAW,IAAI;AAC3F,YAAM,IAAI,WAAW,iGAAiG,KAAK,UAAU,KAAK,YAAY,GAAG;AAAA,IAC3J,WAAW,KAAK,SAAS,GAAG;AAC1B,UAAI,OAAO,KAAK,iBAAiB,UAAU;AACzC,aAAK,eAAe,CAAC,KAAK,cAAc,KAAK,YAAY;AAAA,MAC3D,WAAW,KAAK,aAAa,WAAW,GAAG;AACzC,cAAM,IAAI,WAAW,0FAA0F,KAAK,UAAU,KAAK,YAAY,GAAG;AAAA,MACpJ;AAAA,IACF,WAAW,KAAK,SAAS,GAAG;AAC1B,UAAI,OAAO,KAAK,iBAAiB,UAAU;AACzC,aAAK,eAAe,CAAC,KAAK,cAAc,KAAK,cAAc,KAAK,YAAY;AAAA,MAC9E,WAAW,KAAK,aAAa,WAAW,GAAG;AACzC,cAAM,IAAI,WAAW,4FAA4F,KAAK,UAAU,KAAK,YAAY,GAAG;AAAA,MACtJ;AAAA,IACF;AAAA,EACF;AAAA,EACA,OAAO,WAAW,MAAM;AACtB,YAAQ,gBAAgB,MAAM,yCAAyC;AACvE,QAAI,OAAO,KAAK,eAAe,YAAY,CAAC,wBAAwB,KAAK,YAAY,UAAU,GAAG,CAAC,GAAG;AACpG,YAAM,IAAI,WAAW,oGAAoG,KAAK,UAAU,KAAK,UAAU,IAAI;AAAA,IAC7J;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,YAAY,KAAK;AAAA,MACjB,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,MACjB,cAAc,KAAK;AAAA,MACnB,YAAY,oBAAoB,KAAK,UAAU;AAAA,MAC/C,SAAS,KAAK;AAAA,MACd,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,IACzD;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,OAAO,cAAc,SAAS;AAAA,EAChC,YAAY,MAAM,MAAM;AACtB,UAAM,MAAM,IAAI;AAChB,SAAK,SAAS;AACd,SAAK,WAAW,IAAI;AACpB,SAAK,UAAU,KAAK;AACpB,0BAAsB,KAAK,SAAS,SAAS;AAC7C,SAAK,oBAAoB,eAAe,KAAK,qBAAqB,KAAK,0BAA0B;AACjG,SAAK,mBAAmB,cAAc,KAAK,gBAAgB;AAC3D,SAAK,oBAAoB,eAAe,KAAK,iBAAiB;AAAA,EAChE;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI,WAAW,SAAS;AAClF,QAAI,WAAW,gBAAgB,MAAM;AACnC,YAAM,IAAI,WAAW,+DAA+D,WAAW,cAAc;AAAA,IAC/G;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,cAAc,KAAK,WAAW,OAAO,CAAC,UAAU,KAAK,OAAO,CAAC;AACnE,SAAK,SAAS,KAAK,UAAU,UAAU,aAAa,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACrI,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,OAAO,GAAG,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IAChI;AACA,SAAK,YAAY,CAAC,EAAE,MAAM,KAAK,OAAO,GAAG,MAAM,EAAE,CAAC,cAAc,SAAS,EAAE,CAAC;AAC5E,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI;AACJ,YAAM,YAAY,KAAK,QAAQ,OAAO,OAAO,KAAK,KAAK,KAAK;AAC5D,YAAM,sBAAsB,2BAA2B,KAAK,WAAW,aAAa,CAAC;AACrF,UAAI,uBAAuB,QAAQ,KAAK,SAAS,GAAG;AAClD,kBAAU,yBAAyB,QAAQ,KAAK,OAAO,KAAK,GAAG,WAAW,KAAK,SAAS,KAAK,SAAS,KAAK,YAAY,KAAK,cAAc,mBAAmB;AAAA,MAC/J,OAAO;AACL,YAAI,KAAK,SAAS,GAAG;AACnB,oBAAU,eAAe,QAAQ,KAAK,OAAO,KAAK,GAAG,WAAW,KAAK,QAAQ,IAAI,KAAK,SAAS,KAAK,YAAY,KAAK,aAAa,EAAE;AAAA,QACtI,WAAW,KAAK,SAAS,GAAG;AAC1B,oBAAU,yBAAyB,QAAQ,KAAK,OAAO,KAAK,GAAG,WAAW,KAAK,SAAS,KAAK,SAAS,KAAK,YAAY,KAAK,YAAY;AAAA,QAC1I,WAAW,KAAK,SAAS,GAAG;AAC1B,oBAAU,eAAe,QAAQ,KAAK,OAAO,KAAK,GAAG,WAAW,KAAK,SAAS,KAAK,SAAS,KAAK,YAAY,KAAK,YAAY;AAAA,QAChI,OAAO;AACL,gBAAM,IAAI,oBAAoB,uDAAuD;AAAA,QACvF;AACA,YAAI,KAAK,cAAc,MAAM;AAC3B,oBAAU,KAAK,WAAW,MAAM,OAAO;AAAA,QACzC;AAAA,MACF;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,WAAW,CAAC;AAClB,UAAM,QAAQ,KAAK,eAAe,iBAAiB,WAAW,MAAM,GAAG,WAAW,SAAS,CAAC,IAAI,WAAW,MAAM,CAAC;AAClH,aAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,YAAM,SAAS,iBAAiB,MAAM,IAAI,KAAK,WAAW,IAAI,KAAK,SAAS,KAAK,QAAQ,IAAI,OAAO,KAAK,iBAAiB,WAAW,KAAK,eAAe,KAAK,aAAa,EAAE;AAC7K,eAAS,KAAK,MAAM;AAAA,IACtB;AACA,QAAI,cAAc,CAAC,WAAW,EAAE;AAChC,QAAI,KAAK,eAAe,gBAAgB;AACtC,oBAAc,YAAY,OAAO,QAAQ;AACzC,kBAAY,KAAK,KAAK,OAAO;AAAA,IAC/B,OAAO;AACL,kBAAY,KAAK,KAAK,OAAO;AAC7B,oBAAc,YAAY,OAAO,QAAQ;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,SAAS,KAAK;AAAA,MACd,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,kBAAkB,oBAAoB,KAAK,gBAAgB;AAAA,IAC7D;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,MAAM;AACtB,QAAI,EAAE,aAAa,SAAS,OAAO,KAAK,YAAY,YAAY,KAAK,UAAU,GAAG;AAChF,YAAM,IAAI,WAAW,0EAA0E,KAAK,UAAU,KAAK,OAAO,GAAG;AAAA,IAC/H;AAAA,EACF;AACF;AACA,IAAI,UAAU,cAAc,KAAK;AAAA,EAC/B,YAAY,MAAM;AAChB,UAAM,GAAG,IAAI;AACb,YAAQ,WAAW,IAAI;AAAA,EACzB;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,MAAM;AACtB,QAAI,OAAO,KAAK,eAAe,YAAY,CAAC,wBAAwB,KAAK,YAAY,UAAU,GAAG,CAAC,GAAG;AACpG,YAAM,IAAI,WAAW,8FAA8F,KAAK,UAAU,KAAK,UAAU,IAAI;AAAA,IACvJ;AAAA,EACF;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,UAAU,cAAc,KAAK;AAAA,EAC/B,YAAY,MAAM;AAChB,UAAM,GAAG,IAAI;AACb,YAAQ,WAAW,IAAI;AAAA,EACzB;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,MAAM;AACtB,QAAI,OAAO,KAAK,eAAe,UAAU;AACvC,UAAI,EAAE,MAAM,QAAQ,KAAK,UAAU,MAAM,KAAK,WAAW,WAAW,KAAK,KAAK,WAAW,WAAW,KAAK;AACvG,cAAM,IAAI,WAAW,2FAA2F,KAAK,UAAU,KAAK,UAAU,IAAI;AAAA,MACpJ;AAAA,IACF;AAAA,EACF;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,kBAAkB,cAAc,QAAQ;AAAA,EAC1C,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAC5C,QAAI,KAAK,YAAY,UAAU,KAAK,YAAY,SAAS;AACvD,YAAM,IAAI,WAAW,uGAAuG,KAAK,SAAS;AAAA,IAC5I;AAAA,EACF;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,WAAW,WAAW,GAAG;AAC3B,YAAM,IAAI,WAAW,qDAAqD,KAAK,UAAU,UAAU,CAAC;AAAA,IACtG;AACA,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI,WAAW,SAAS;AAClF,QAAI,WAAW,gBAAgB,MAAM;AACnC,YAAM,IAAI,WAAW,sEAAsE;AAAA,IAC7F;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,cAAc,KAAK,WAAW,OAAO,CAAC,KAAK,SAAS,QAAQ,CAAC;AACnE,SAAK,SAAS,KAAK,UAAU,UAAU,aAAa,WAAW,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AAC1I,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,OAAO,GAAG,WAAW,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IACrI;AACA,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC,cAAc,SAAS,EAAE,CAAC,CAAC;AAC/E,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,oBAAoB,MAAM;AACvC,UAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,cAAM,IAAI,WAAW,2FAA2F,OAAO,MAAM,QAAQ;AAAA,MACvI;AACA,YAAM,aAAa,OAAO;AAC1B,YAAM,YAAY,WAAW;AAC7B,UAAI;AACJ,UAAI;AACJ,UAAI,KAAK,eAAe,iBAAiB;AACvC,gBAAQ;AACR,gBAAQ;AAAA,MACV,OAAO;AACL,gBAAQ;AACR,gBAAQ;AAAA,MACV;AACA,YAAM,SAAS,WAAW;AAC1B,YAAM,QAAQ,WAAW;AACzB,YAAM,UAAU,KAAK,WAAW;AAChC,YAAM,UAAU,KAAK,WAAW;AAChC,YAAM,UAAU,KAAK,QAAQ;AAC7B,YAAM,UAAU,KAAK,QAAQ;AAC7B,YAAM,YAAY,aAAa,QAAQ,SAAS,SAAS,KAAK,OAAO;AACrE,YAAM,WAAW,aAAa,OAAO,SAAS,SAAS,KAAK,OAAO;AACnE,YAAM,cAAc,CAAC,WAAW,WAAW,UAAU,KAAK,OAAO;AACjE,UAAI,KAAK,eAAe,gBAAgB;AACtC,iBAAS,UAAU,QAAQ,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MACzC;AACA,UAAI,UAAU,gBAAgB,QAAQ,KAAK,OAAO,KAAK,GAAG,aAAa,KAAK,SAAS,KAAK,OAAO;AACjG,UAAI,KAAK,eAAe,gBAAgB;AACtC,kBAAU,UAAU,SAAS,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MAC3C;AACA,UAAI,KAAK,QAAQ,MAAM;AACrB,kBAAU,QAAQ,SAAS,KAAK,KAAK,KAAK,GAAG,KAAK,UAAU;AAAA,MAC9D;AACA,UAAI,KAAK,cAAc,MAAM;AAC3B,kBAAU,KAAK,WAAW,MAAM,OAAO;AAAA,MACzC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,WAAW,MAAM;AACrC,QAAI;AACJ,QAAI;AACJ,QAAI;AACJ,QAAI,KAAK,eAAe,iBAAiB;AACvC,oBAAc;AACd,mBAAa;AACb,kBAAY;AAAA,IACd,OAAO;AACL,oBAAc;AACd,mBAAa;AACb,kBAAY;AAAA,IACd;AACA,UAAM,UAAU,KAAK,WAAW;AAChC,UAAM,UAAU,KAAK,WAAW;AAChC,UAAM,UAAU,KAAK,QAAQ;AAC7B,UAAM,UAAU,KAAK,QAAQ;AAC7B,gBAAY,eAAe,KAAK;AAChC,gBAAY,cAAc,aAAa,YAAY,aAAa,SAAS,SAAS,KAAK,OAAO;AAC9F,gBAAY,aAAa,aAAa,YAAY,YAAY,SAAS,SAAS,KAAK,OAAO;AAC5F,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,kBAAkB,cAAc,QAAQ;AAAA,EAC1C,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAC5C,QAAI,KAAK,YAAY,UAAU,KAAK,YAAY,SAAS;AACvD,YAAM,IAAI,WAAW,uGAAuG,KAAK,SAAS;AAAA,IAC5I;AAAA,EACF;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,WAAW,WAAW,GAAG;AAC3B,YAAM,IAAI,WAAW,qDAAqD,KAAK,UAAU,UAAU,CAAC;AAAA,IACtG;AACA,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI,WAAW,SAAS;AAClF,QAAI,WAAW,gBAAgB,MAAM;AACnC,YAAM,IAAI,WAAW,sEAAsE;AAAA,IAC7F;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,cAAc,KAAK,WAAW,OAAO,CAAC,KAAK,SAAS,QAAQ,CAAC;AACnE,SAAK,SAAS,KAAK,UAAU,UAAU,aAAa,WAAW,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AAC1I,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,OAAO,GAAG,WAAW,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IACrI;AACA,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC,cAAc,SAAS,EAAE,CAAC,CAAC;AAC/E,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,oBAAoB,MAAM;AACvC,UAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,cAAM,IAAI,WAAW,2FAA2F,OAAO,MAAM,QAAQ;AAAA,MACvI;AACA,YAAM,aAAa,OAAO;AAC1B,YAAM,YAAY,WAAW;AAC7B,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI,KAAK,eAAe,iBAAiB;AACvC,gBAAQ;AACR,gBAAQ;AACR,gBAAQ;AAAA,MACV,OAAO;AACL,gBAAQ;AACR,gBAAQ;AACR,gBAAQ;AAAA,MACV;AACA,YAAM,QAAQ,WAAW;AACzB,YAAM,SAAS,WAAW;AAC1B,YAAM,QAAQ,WAAW;AACzB,YAAM,UAAU,KAAK,WAAW;AAChC,YAAM,UAAU,KAAK,WAAW;AAChC,YAAM,UAAU,KAAK,WAAW;AAChC,YAAM,UAAU,KAAK,QAAQ;AAC7B,YAAM,UAAU,KAAK,QAAQ;AAC7B,YAAM,UAAU,KAAK,QAAQ;AAC7B,YAAM,WAAW,aAAa,OAAO,SAAS,SAAS,KAAK,OAAO;AACnE,YAAM,YAAY,aAAa,QAAQ,SAAS,SAAS,KAAK,OAAO;AACrE,YAAM,WAAW,aAAa,OAAO,SAAS,SAAS,KAAK,OAAO;AACnE,YAAM,cAAc,CAAC,WAAW,UAAU,WAAW,UAAU,KAAK,OAAO;AAC3E,UAAI,KAAK,eAAe,gBAAgB;AACtC,iBAAS,UAAU,QAAQ,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MAC5C;AACA,UAAI,UAAU,gBAAgB,QAAQ,KAAK,OAAO,KAAK,GAAG,aAAa,KAAK,SAAS,KAAK,OAAO;AACjG,UAAI,KAAK,eAAe,gBAAgB;AACtC,kBAAU,UAAU,SAAS,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MAC9C;AACA,UAAI,KAAK,SAAS,MAAM;AACtB,kBAAU,QAAQ,SAAS,KAAK,KAAK,KAAK,GAAG,KAAK,UAAU;AAAA,MAC9D;AACA,UAAI,KAAK,eAAe,MAAM;AAC5B,kBAAU,KAAK,WAAW,MAAM,OAAO;AAAA,MACzC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,WAAW,MAAM;AACrC,QAAI;AACJ,QAAI;AACJ,QAAI;AACJ,QAAI;AACJ,QAAI,KAAK,eAAe,iBAAiB;AACvC,oBAAc;AACd,kBAAY;AACZ,mBAAa;AACb,kBAAY;AAAA,IACd,OAAO;AACL,oBAAc;AACd,kBAAY;AACZ,mBAAa;AACb,kBAAY;AAAA,IACd;AACA,UAAM,UAAU,KAAK,WAAW;AAChC,UAAM,UAAU,KAAK,WAAW;AAChC,UAAM,UAAU,KAAK,WAAW;AAChC,UAAM,UAAU,KAAK,QAAQ;AAC7B,UAAM,UAAU,KAAK,QAAQ;AAC7B,UAAM,UAAU,KAAK,QAAQ;AAC7B,gBAAY,eAAe,KAAK;AAChC,gBAAY,aAAa,aAAa,YAAY,YAAY,SAAS,SAAS,KAAK,OAAO;AAC5F,gBAAY,cAAc,aAAa,YAAY,aAAa,SAAS,SAAS,KAAK,OAAO;AAC9F,gBAAY,aAAa,aAAa,YAAY,YAAY,SAAS,SAAS,KAAK,OAAO;AAC5F,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,gBAAgB,cAAc,KAAK;AAAA,EACrC,YAAY,MAAM,QAAQ;AACxB,UAAM,MAAM,MAAM;AAClB,SAAK,gCAAgC;AACrC,SAAK,gCAAgC;AACrC,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AACvB,QAAI,OAAO,WAAW,MAAM;AAC1B,YAAM,IAAI,WAAW,qFAAqF;AAAA,IAC5G;AACA,QAAI,OAAO,qBAAqB,QAAQ,OAAO,qBAAqB,QAAQ,OAAO,oBAAoB,MAAM;AAC3G,YAAM,IAAI,WAAW,oPAAoP;AAAA,IAC3Q;AACA,QAAI,OAAO,WAAW,QAAQ,OAAO,YAAY,UAAU,OAAO,YAAY,SAAS;AACrF,YAAM,IAAI,WAAW,gBAAgB,KAAK,uEAAuE,KAAK,UAAU,OAAO,OAAO,GAAG;AAAA,IACnJ;AACA,SAAK,kBAAkB,OAAO,mBAAmB,OAAO,IAAI,OAAO;AACnE,SAAK,uBAAuB,eAAe,OAAO,wBAAwB,KAAK,6BAA6B;AAC5G,SAAK,uBAAuB,eAAe,OAAO,oBAAoB;AACtE,SAAK,sBAAsB,cAAc,OAAO,mBAAmB;AACnE,SAAK,uBAAuB,eAAe,OAAO,wBAAwB,KAAK,6BAA6B;AAC5G,SAAK,uBAAuB,eAAe,OAAO,oBAAoB;AACtE,SAAK,sBAAsB,cAAc,OAAO,mBAAmB;AAAA,EACrE;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,WAAW,SAAS,KAAK,OAAO,GAAG;AACrC,YAAM,IAAI,WAAW,0BAA0B,KAAK,0BAA0B,KAAK,OAAO,gCAAgC,KAAK,UAAU,UAAU,GAAG;AAAA,IACxJ;AACA,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI,WAAW,SAAS;AAClF,QAAI,WAAW,gBAAgB,QAAQ,WAAW,eAAe,GAAG;AAClE,YAAM,IAAI,WAAW,oEAAoE,KAAK,UAAU,WAAW,YAAY,GAAG;AAAA,IACpI;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,uBAAuB,KAAK,WAAW,OAAO,CAAC,UAAU,KAAK,eAAe,CAAC;AACpF,UAAM,uBAAuB,CAAC;AAC9B,aAAS,IAAI,GAAG,IAAI,KAAK,MAAM,EAAE,GAAG;AAClC,2BAAqB,KAAK,CAAC;AAAA,IAC7B;AACA,yBAAqB,KAAK,WAAW,KAAK,iBAAiB,KAAK,OAAO;AACvE,UAAM,YAAY;AAClB,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,sBAAsB,WAAW,KAAK,sBAAsB,KAAK,sBAAsB,WAAW,KAAK,mBAAmB;AACpL,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,sBAAsB,WAAW,KAAK,sBAAsB,KAAK,sBAAsB,WAAW,KAAK,mBAAmB;AACpL,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,OAAO,GAAG,WAAW,KAAK,iBAAiB,KAAK,iBAAiB,WAAW,KAAK,cAAc;AAAA,IAC1I,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,KAAK,OAAO,GAAG,MAAM,EAAE,CAAC,cAAc,SAAS,EAAE,CAAC,CAAC;AAC3F,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI;AACJ,UAAI,KAAK,SAAS,GAAG;AACnB,cAAM,IAAI,oBAAoB,kDAAkD;AAAA,MAClF,WAAW,KAAK,SAAS,GAAG;AAC1B,YAAI,KAAK,eAAe,iBAAiB;AACvC,mBAAS,UAAU,QAAQ,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,QACzC;AACA,iBAAS,gBAAgB,QAAQ,KAAK,gBAAgB,KAAK,GAAG,KAAK,gBAAgB,KAAK,GAAG,KAAK,SAAS,KAAK,SAAS,KAAK,cAAc,MAAM;AAAA,MAClJ;AACA,UAAI,KAAK,SAAS;AAChB,iBAAS,QAAQ,QAAQ,KAAK,KAAK,KAAK,GAAG,KAAK,UAAU;AAAA,MAC5D;AACA,UAAI,KAAK,cAAc,MAAM;AAC3B,iBAAS,KAAK,WAAW,MAAM,MAAM;AAAA,MACvC;AACA,UAAI,KAAK,eAAe,iBAAiB;AACvC,iBAAS,UAAU,QAAQ,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MACzC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO,OAAO;AACd,WAAO,OAAO;AACd,WAAO,OAAO;AACd,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,yBAAyB,oBAAoB,KAAK,mBAAmB;AAC5E,WAAO,yBAAyB,oBAAoB,KAAK,mBAAmB;AAC5E,WAAO;AAAA,EACT;AACF;AACA,cAAc,YAAY;AAC1B,IAAI,kBAAkB,cAAc,cAAc;AAAA,EAChD,YAAY,MAAM;AAChB,UAAM,GAAG,IAAI;AAAA,EACf;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,SAAS,cAAc,KAAK;AAAA,EAC9B,YAAY,MAAM;AAChB,UAAM,GAAG,IAAI;AACb,WAAO,WAAW,IAAI;AACtB,SAAK,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;AAAA,EAC/B;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,OAAO;AACd,WAAO,OAAO;AACd,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,MAAM;AACtB,QAAI,OAAO,KAAK,eAAe,YAAY,CAAC,wBAAwB,KAAK,YAAY,UAAU,GAAG,CAAC,GAAG;AACpG,YAAM,IAAI,WAAW,yFAAyF,KAAK,UAAU,KAAK,UAAU,IAAI;AAAA,IAClJ;AAAA,EACF;AACF;AACA,OAAO,YAAY;AACnB,sBAAsB,cAAc,MAAM;AAC1C,IAAI,aAAa,cAAc,MAAM;AAAA,EACnC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,QAAI,OAAO,KAAK,aAAa,UAAU;AACrC,WAAK,WAAW,CAAC,CAAC,KAAK,UAAU,KAAK,QAAQ,GAAG,CAAC,KAAK,UAAU,KAAK,QAAQ,CAAC;AAAA,IACjF,WAAW,OAAO,KAAK,SAAS,OAAO,UAAU;AAC/C,WAAK,WAAW;AAAA,QACd,CAAC,KAAK,SAAS,IAAI,KAAK,SAAS,EAAE;AAAA,QACnC,CAAC,KAAK,SAAS,IAAI,KAAK,SAAS,EAAE;AAAA,MACrC;AAAA,IACF,OAAO;AACL,WAAK,WAAW,KAAK;AAAA,IACvB;AACA,SAAK,aAAa,KAAK,eAAe,SAAS,iBAAiB,KAAK;AACrE,SAAK,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;AAAA,EAC/B;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,KAAK,eAAe,iBAAiB;AACvC,aAAO;AAAA,QACL,WAAW;AAAA,QACX,WAAW;AAAA,QACX,WAAW,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG;AAAA,QACvD,WAAW,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG;AAAA,MACzD;AAAA,IACF,OAAO;AACL,aAAO;AAAA,QACL,WAAW;AAAA,QACX,WAAW,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG;AAAA,QACvD,WAAW,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG;AAAA,QACvD,WAAW;AAAA,MACb;AAAA,IACF;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI,KAAK,eAAe,gBAAgB;AACtC,cAAM,UAAU,eAAe,QAAQ,KAAK,SAAS,GAAG,IAAI,OAAO,MAAM,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG,IAAI,CAAC;AAC1H,eAAO,eAAe,SAAS,KAAK,SAAS,GAAG,IAAI,OAAO,MAAM,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG,IAAI,CAAC;AAAA,MACpH,OAAO;AACL,cAAM,UAAU,eAAe,QAAQ,KAAK,SAAS,GAAG,IAAI,OAAO,MAAM,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG,IAAI,CAAC;AAC1H,eAAO,eAAe,SAAS,KAAK,SAAS,GAAG,IAAI,OAAO,MAAM,KAAK,KAAK,SAAS,GAAG,KAAK,KAAK,SAAS,GAAG,IAAI,CAAC;AAAA,MACpH;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,UAAU,KAAK,UAAU,YAAY,KAAK,WAAW;AACtE,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAC9C,IAAI,eAAe,cAAc,MAAM;AAAA,EACrC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,eAAe,CAAC,GAAG,CAAC;AACzB,SAAK,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;AAC7B,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,eAAe,KAAK;AACzD,SAAK,aAAa,KAAK,cAAc,OAAO,iBAAiB,KAAK;AAClE,oBAAgB,KAAK,UAAU;AAC/B,SAAK,gBAAgB,KAAK,iBAAiB,OAAO,YAAY,KAAK;AACnE,6BAAyB,KAAK,aAAa;AAAA,EAC7C;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,KAAK,eAAe,iBAAiB;AACvC,YAAM,SAAS,WAAW,MAAM,OAAO,OAAO,KAAK,KAAK,KAAK,WAAW;AACxE,YAAM,QAAQ,WAAW,MAAM,OAAO,OAAO,KAAK,KAAK,KAAK,WAAW;AACvE,aAAO,CAAC,WAAW,IAAI,WAAW,IAAI,QAAQ,KAAK;AAAA,IACrD,OAAO;AACL,YAAM,SAAS,WAAW,MAAM,OAAO,OAAO,KAAK,KAAK,KAAK,WAAW;AACxE,YAAM,QAAQ,WAAW,MAAM,OAAO,OAAO,KAAK,KAAK,KAAK,WAAW;AACvE,aAAO,CAAC,WAAW,IAAI,QAAQ,OAAO,WAAW,EAAE;AAAA,IACrD;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,oBAAoB,MAAM;AACvC,YAAM,aAAa,OAAO;AAC1B,UAAI,KAAK,eAAe,iBAAiB;AACvC,iBAAS,UAAU,QAAQ,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AACvC,cAAM,SAAS,KAAK,KAAK,KAAK,WAAW;AACzC,cAAM,QAAQ,KAAK,KAAK,KAAK,WAAW;AACxC,cAAM,UAAU,KAAK,kBAAkB,YAAY,MAAM,sBAAsB,QAAQ,CAAC,QAAQ,KAAK,CAAC,IAAI,MAAM,eAAe,QAAQ,CAAC,QAAQ,KAAK,CAAC;AACtJ,eAAO,UAAU,SAAS,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,MACxC,OAAO;AACL,cAAM,SAAS,KAAK,KAAK,KAAK,WAAW;AACzC,cAAM,QAAQ,KAAK,KAAK,KAAK,WAAW;AACxC,eAAO,KAAK,kBAAkB,YAAY,MAAM,sBAAsB,QAAQ,CAAC,QAAQ,KAAK,CAAC,IAAI,MAAM,eAAe,QAAQ,CAAC,QAAQ,KAAK,CAAC;AAAA,MAC/I;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,MAAM,KAAK;AAAA,MACX,YAAY,KAAK;AAAA,MACjB,eAAe,KAAK;AAAA,IACtB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAGhD,SAAS,iBAAiB,GAAG,iBAAiB,UAAU,CAAC,GAAG,CAAC,GAAG,UAAU,SAAS,YAAY,cAAc;AAC3G,SAAO,KAAK,MAAM;AAChB,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,oBAAgB,UAAU;AAC1B,QAAI,IAAI,sBAAsB,GAAG,UAAU;AAC3C,QAAI,EAAE,SAAS,GAAG;AAChB,YAAM,IAAI,WAAW,mEAAmE,EAAE,QAAQ;AAAA,IACpG;AACA,QAAI,gBAAgB,SAAS,GAAG;AAC9B,YAAM,IAAI,WAAW,yDAAyD,gBAAgB,QAAQ;AAAA,IACxG;AACA,QAAI,gBAAgB,GAAG,iBAAiB,SAAS,YAAY,SAAS,SAAS,SAAS,QAAQ,YAAY;AAC5G,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAC/B;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,kBAAkB,cAAc,SAAS;AAAA,EAC3C,YAAY,MAAM;AAChB,UAAM,GAAG,IAAI;AACb,SAAK,kBAAkB;AACvB,SAAK,kBAAkB,KAAK,mBAAmB,OAAO,IAAI,KAAK;AAC/D,SAAK,uBAAuB,eAAe,KAAK,wBAAwB,KAAK,0BAA0B;AACvG,SAAK,sBAAsB,cAAc,KAAK,mBAAmB;AACjE,SAAK,uBAAuB,eAAe,KAAK,oBAAoB;AAAA,EACtE;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,IAAI,WAAW,uEAAuE,KAAK,UAAU,UAAU,IAAI;AAAA,IAC3H;AACA,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI;AAC9D,QAAI,WAAW,gBAAgB,QAAQ,WAAW,eAAe,GAAG;AAClE,YAAM,IAAI,WAAW,yFAAyF,WAAW,gBAAgB;AAAA,IAC3I;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,uBAAuB;AAAA,MAC3B,KAAK,WAAW;AAAA,MAChB,KAAK,WAAW;AAAA,MAChB;AAAA,MACA,KAAK;AAAA,IACP;AACA,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,sBAAsB,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,MAAM,KAAK,mBAAmB;AAC1K,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,WAAW,KAAK,eAAe,GAAG,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IACnJ,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI,UAAU,iBAAiB,QAAQ,KAAK,gBAAgB,KAAK,GAAG,KAAK,SAAS,KAAK,SAAS,KAAK,YAAY,IAAI;AACrH,UAAI,KAAK,SAAS;AAChB,kBAAU,QAAQ,SAAS,KAAK,KAAK,KAAK,GAAG,KAAK,UAAU;AAAA,MAC9D;AACA,UAAI,KAAK,cAAc,MAAM;AAC3B,kBAAU,KAAK,WAAW,MAAM,OAAO;AAAA,MACzC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC9E,UAAM,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC9E,UAAM,aAAa,KAAK,eAAe,kBAAkB,WAAW,KAAK,KAAK,kBAAkB,WAAW,KAAK,KAAK;AACrH,UAAM,UAAU,iBAAiB,MAAM,KAAK,WAAW,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AACxF,UAAM,UAAU,iBAAiB,MAAM,KAAK,WAAW,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AACxF,QAAI,KAAK,eAAe,iBAAiB;AACvC,aAAO,CAAC,WAAW,IAAI,YAAY,SAAS,OAAO;AAAA,IACrD,OAAO;AACL,aAAO,CAAC,WAAW,IAAI,SAAS,SAAS,UAAU;AAAA,IACrD;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS,MAAM,UAAU;AAC/B,WAAO,qBAAqB,KAAK;AACjC,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,0BAA0B,qBAAqB,KAAK,oBAAoB;AAC/E,WAAO,yBAAyB,oBAAoB,KAAK,oBAAoB;AAC7E,WAAO;AAAA,EACT;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AAGnD,SAAS,gBAAgB,QAAQ,cAAc,WAAW,cAAc;AACtE,MAAI,MAAM,QAAQ,MAAM,GAAG;AACzB,QAAI,gBAAgB,QAAQ,aAAa,MAAM;AAC7C,YAAM,IAAI,WAAW,+EAA+E;AAAA,IACtG;AACA,QAAI,gBAAgB,MAAM;AACxB,kBAAY,OAAO,MAAM,OAAO,SAAS,cAAc,OAAO,MAAM;AACpE,eAAS,OAAO,MAAM,GAAG,OAAO,SAAS,YAAY;AAAA,IACvD;AACA,QAAI,OAAO,SAAS,GAAG;AACrB,qBAAe,OAAO,MAAM,GAAG,OAAO,MAAM;AAAA,IAC9C;AACA,aAAS,OAAO;AAAA,EAClB;AACA,WAAS,aAAa,GAAG;AACvB,QAAI,KAAK,QAAQ,MAAM,QAAQ,CAAC,GAAG;AACjC,aAAO;AAAA,IACT,OAAO;AACL,aAAO,CAAC,CAAC;AAAA,IACX;AAAA,EACF;AACA,iBAAe,aAAa,YAAY;AACxC,cAAY,aAAa,SAAS;AAClC,SAAO,EAAE,QAAQ,cAAc,UAAU;AAC3C;AACA,SAAS,IAAI,cAAc,QAAQ,eAAe,cAAc,OAAO,MAAM,WAAW,SAAS,OAAO,qBAAqB,OAAO;AAClI,SAAO,KAAK,MAAM;AAChB,UAAM,OAAO,OAAO,MAAM;AAC1B,QAAI,OAAO,GAAG;AACZ,YAAM,IAAI,WAAW,uCAAuC,QAAQ;AAAA,IACtE;AACA,UAAM,OAAO,CAAC,GAAG,CAAC,EAAE,OAAO,OAAO,GAAG,IAAI,CAAC;AAC1C,aAAS,UAAU,QAAQ,IAAI;AAC/B,QAAI,aAAa,MAAM;AACrB,YAAM,IAAI,oBAAoB,gFAAgF;AAAA,IAChH;AACA,QAAI,QAAQ;AACV,cAAQ,KAAK,mGAAmG;AAAA,IAClH;AACA,QAAI,QAAQ,MAAM;AAChB,aAAO,KAAK,KAAK,MAAM,MAAM,GAAG,SAAS;AACzC,UAAI,KAAK,SAAS,OAAO,GAAG;AAC1B,eAAO,WAAW,MAAM,EAAE;AAAA,MAC5B;AACA,aAAO,UAAU,MAAM,IAAI;AAAA,IAC7B;AACA,QAAI,aAAa;AACf,eAAS,QAAQ,QAAQ,CAAC;AAC1B,UAAI,QAAQ,MAAM;AAChB,eAAO,QAAQ,MAAM,CAAC;AAAA,MACxB;AAAA,IACF;AACA,UAAM,iBAAiB,CAAC;AACxB,QAAI;AACJ,QAAI,SAAS;AACb,UAAM,YAAY,OAAO,MAAM;AAC/B,UAAM,gBAAgB,QAAQ,MAAM;AACpC,QAAI;AACJ,QAAI,QAAQ,MAAM;AAChB,qBAAe,QAAQ,IAAI;AAAA,IAC7B;AACA,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,YAAM,eAAe,cAAc;AACnC,YAAM,cAAc,KAAK,MAAM,aAAa,cAAc,MAAM,CAAC;AACjE,UAAI,QAAQ,MAAM;AAChB,qBAAa,YAAY;AACzB,iBAAS,YAAY;AAAA,MACvB,OAAO;AACL,cAAM,gBAAgB,KAAK,MAAM;AAC/B,gBAAM,WAAW,aAAa;AAC9B,gBAAM,cAAc,IAAI,SAAS,QAAQ,GAAG,QAAQ;AACpD,gBAAM,SAAS,KAAK,IAAI,YAAY,IAAI,QAAQ,GAAG,IAAI,OAAO,IAAI,WAAW,CAAC;AAC9E,gBAAM,YAAY,OAAO,IAAI,CAAC,OAAO,MAAM;AACzC,mBAAO,KAAK,IAAI,YAAY,GAAG,IAAI,QAAQ,GAAG,IAAI,OAAO,WAAW,CAAC;AAAA,UACvE,CAAC;AACD,iBAAO,EAAE,QAAQ,UAAU;AAAA,QAC7B,CAAC;AACD,qBAAa,cAAc;AAC3B,iBAAS,cAAc;AAAA,MACzB;AACA,UAAI,oBAAoB;AACtB,uBAAe,KAAK,UAAU;AAAA,MAChC;AAAA,IACF;AACA,QAAI;AACJ,QAAI,oBAAoB;AACtB,YAAM,OAAO;AACb,gBAAU,MAAM,gBAAgB,IAAI;AAAA,IACtC;AACA,WAAO,CAAC,YAAY,SAAS,MAAM;AAAA,EACrC,CAAC;AACH;AACA,IAAI,MAAM,cAAc,MAAM;AAAA,EAC5B,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,QAAI;AACJ,QAAI,KAAK,QAAQ,MAAM;AACrB,YAAM,IAAI,WAAW,sDAAsD;AAAA,IAC7E,WAAW,MAAM,QAAQ,KAAK,IAAI,GAAG;AACnC,aAAO,IAAI,gBAAgB,EAAE,OAAO,KAAK,KAAK,CAAC;AAAA,IACjD,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AACA,QAAI,KAAK,aAAa,MAAM;AAC1B,YAAM,IAAI,WAAW,mGAAmG;AAAA,IAC1H;AACA,SAAK,OAAO;AACZ,SAAK,kBAAkB,KAAK,mBAAmB,OAAO,QAAQ,KAAK;AACnE,SAAK,cAAc,KAAK,eAAe,OAAO,QAAQ,KAAK;AAC3D,SAAK,cAAc,KAAK,eAAe,OAAO,QAAQ,KAAK;AAC3D,SAAK,YAAY,KAAK,YAAY,OAAO,QAAQ,KAAK;AACtD,SAAK,SAAS,KAAK,UAAU,OAAO,QAAQ,KAAK;AACjD,SAAK,kBAAkB;AACvB,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAC5C,SAAK,YAAY;AACjB,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,aAAa,CAAC;AAAA,EACrB;AAAA,EACA,YAAY;AACV,QAAI,KAAK,WAAW,MAAM;AACxB,YAAM,YAAY,MAAM,QAAQ,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,UAAU,SAAS;AACpF,aAAO,OAAO,GAAG,SAAS,EAAE,IAAI,CAAC,MAAM,IAAI;AAAA,IAC7C,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AAAA,EACF;AAAA,EACA,UAAU,QAAQ;AAChB,SAAK,UAAU;AAAA,EACjB;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,gBAAgB,UAAU,GAAG;AAC/B,mBAAa,WAAW;AAAA,IAC1B;AACA,iBAAa;AACb,QAAI,YAAY,KAAK,KAAK;AAC1B,QAAI,CAAC,MAAM,QAAQ,SAAS,GAAG;AAC7B,kBAAY,CAAC,SAAS;AAAA,IACxB;AACA,UAAM,YAAY,UAAU;AAC5B,QAAI;AACJ,QAAI,KAAK,iBAAiB;AACxB,oBAAc,CAAC,WAAW,IAAI,WAAW,IAAI,SAAS;AAAA,IACxD,OAAO;AACL,oBAAc,CAAC,WAAW,IAAI,SAAS;AAAA,IACzC;AACA,QAAI,KAAK,aAAa;AACpB,YAAM,aAAa,CAAC;AACpB,iBAAW,OAAO,WAAW;AAC3B,mBAAW,KAAK,CAAC,WAAW,IAAI,GAAG,CAAC;AAAA,MACtC;AACA,aAAO,CAAC,WAAW,EAAE,OAAO,UAAU;AAAA,IACxC,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,WAAO,KAAK,MAAM;AAChB,UAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,eAAO,KAAK;AAAA,MACd;AACA,YAAM,aAAa,KAAK,kBAAkB,OAAO;AACjD,UAAI,KAAK,aAAa;AACpB,cAAM,YAAY,KAAK,OAAO,IAAI,CAAC,MAAM,IAAI;AAC7C,eAAO,CAAC,UAAU,EAAE,OAAO,SAAS;AAAA,MACtC,OAAO;AACL,eAAO;AAAA,MACT;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,IAAI,SAAS;AACX,QAAI,KAAK,WAAW,MAAM;AACxB,YAAM,YAAY,MAAM,QAAQ,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,UAAU,SAAS;AACpF,YAAM,SAAS,CAAC;AAChB,eAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,eAAO,KAAK,IAAI;AAAA,MAClB;AACA,aAAO;AAAA,IACT,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AAAA,EACF;AAAA,EACA,IAAI,OAAO,GAAG;AACZ,SAAK,UAAU;AAAA,EACjB;AAAA,EACA,MAAM,YAAY;AAChB,UAAM,gBAAgB;AACtB,QAAI,KAAK,gBAAgB,MAAM;AAC7B,YAAM,IAAI,oBAAoB,kDAAkD;AAAA,IAClF;AACA,QAAI,gBAAgB,UAAU,GAAG;AAC/B,mBAAa,WAAW;AAAA,IAC1B;AACA,iBAAa;AACb,UAAM,YAAY,KAAK,WAAW,WAAW,KAAK;AAClD,UAAM,WAAW,WAAW,MAAM,CAAC;AACnC,SAAK,UAAU,KAAK,IAAI,UAAU,EAAE,OAAO,CAAC,WAAW,MAAM,GAAG,QAAQ,EAAE,CAAC;AAC3E,UAAM,iBAAiB,CAAC,WAAW,EAAE,EAAE,OAAO,WAAW,MAAM,CAAC,CAAC;AACjE,QAAI,iBAAiB,MAAM;AACzB,YAAM,IAAI,oBAAoB,kDAAkD;AAAA,IAClF,OAAO;AACL,WAAK,KAAK,MAAM,cAAc;AAAA,IAChC;AACA,QAAI;AACJ,QAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,kBAAY,KAAK,KAAK;AAAA,IACxB,OAAO;AACL,kBAAY,CAAC,KAAK,KAAK,SAAS;AAAA,IAClC;AACA,QAAI,KAAK,aAAa,MAAM;AAC1B,UAAI,CAAC,aAAa,YAAY,KAAK,UAAU,IAAI,CAAC,SAAS,KAAK,MAAM,KAAK,MAAM,SAAS,EAAE,GAAG,SAAS,GAAG;AACzG,cAAM,IAAI,WAAW,6FAA6F,KAAK,wCAAwC,KAAK,KAAK,WAAW;AAAA,MACtL;AAAA,IACF,OAAO;AACL,WAAK,YAAY,UAAU,IAAI,CAAC,QAAQ,IAAI,UAAU,EAAE,OAAO,CAAC,MAAM,GAAG,EAAE,CAAC,CAAC;AAAA,IAC/E;AACA,QAAI,KAAK,UAAU;AACjB,WAAK,YAAY;AAAA,IACnB;AAAA,EACF;AAAA,EACA,YAAY,QAAQ,WAAW,OAAO;AACpC,SAAK,MAAM;AACT,UAAI,CAAC,KAAK,UAAU;AAClB,cAAM,IAAI,eAAe,iEAAiE;AAAA,MAC5F;AACA,YAAM,YAAY,KAAK,UAAU,GAAG,MAAM;AAC1C,UAAI,aAAa,MAAM;AACrB,cAAM,IAAI,WAAW,uUAAuU;AAAA,MAC9V;AACA,UAAI,KAAK,WAAW,MAAM;AACxB,YAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,eAAK,UAAU,KAAK,KAAK,UAAU,IAAI,CAAC,QAAQ,MAAM,CAAC,WAAW,GAAG,CAAC,CAAC;AAAA,QACzE,OAAO;AACL,eAAK,UAAU,CAAC,MAAM,CAAC,WAAW,KAAK,KAAK,SAAS,CAAC,CAAC;AAAA,QACzD;AAAA,MACF,WAAW,UAAU,MAAM;AACzB,gBAAQ,KAAK,OAAO;AACpB,YAAI,KAAK,cAAc,MAAM;AAC3B,kBAAQ,KAAK,UAAU;AACvB,eAAK,aAAa,CAAC;AAAA,QACrB;AACA,YAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,eAAK,UAAU,KAAK,KAAK,UAAU,IAAI,CAAC,QAAQ,MAAM,CAAC,WAAW,GAAG,CAAC,CAAC;AAAA,QACzE,OAAO;AACL,eAAK,QAAQ,KAAK,MAAM,CAAC,WAAW,KAAK,KAAK,SAAS,CAAC;AAAA,QAC1D;AAAA,MACF,OAAO;AACL,YAAI,CAAC,MAAM,QAAQ,MAAM,GAAG;AAC1B,mBAAS,CAAC,MAAM;AAAA,QAClB;AACA,YAAI,OAAO,WAAW,KAAK,QAAQ,QAAQ;AACzC,gBAAM,IAAI,WAAW,SAAS,KAAK,gBAAgB,KAAK,QAAQ,oCAAoC,OAAO,0CAA0C,QAAQ;AAAA,QAC/J;AACA,YAAI,aAAa,MAAM;AACrB,eAAK,WAAW,KAAK,KAAK,QAAQ,MAAM,CAAC;AAAA,QAC3C,OAAO;AACL,kBAAQ,KAAK,OAAO;AAAA,QACtB;AACA,iBAAS,QAAQ,GAAG,QAAQ,KAAK,QAAQ,QAAQ,EAAE,OAAO;AACxD,gBAAM,QAAQ,OAAO;AACrB,gBAAM,MAAM,MAAM,QAAQ,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,UAAU,SAAS,KAAK,KAAK;AACxF,gBAAM,gBAAgB,CAAC,WAAW,GAAG;AACrC,cAAI,CAAC,aAAa,YAAY,MAAM,OAAO,aAAa,GAAG;AACzD,kBAAM,IAAI,WAAW,SAAS,oCAAoC,KAAK,wBAAwB,iCAAiC,MAAM,OAAO;AAAA,UAC/I;AACA,eAAK,QAAQ,SAAS;AAAA,QACxB;AAAA,MACF;AACA,WAAK,UAAU,KAAK,QAAQ,IAAI,CAAC,UAAU,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,IAChE,CAAC;AAAA,EACH;AAAA,EACA,MAAM,QAAQ,QAAQ;AACpB,QAAI,eAAe,UAAU,OAAO,OAAO,OAAO;AAClD,QAAI,YAAY,UAAU,OAAO,OAAO,OAAO;AAC/C,QAAI,UAAU,MAAM;AAClB,eAAS,CAAC;AAAA,IACZ;AACA,UAAM,eAAe,gBAAgB,QAAQ,cAAc,WAAW,KAAK,YAAY;AACvF,aAAS,aAAa;AACtB,mBAAe,aAAa;AAC5B,gBAAY,aAAa;AACzB,QAAI,mBAAmB,CAAC;AACxB,QAAI,kBAAkB,CAAC;AACvB,QAAI,gBAAgB,MAAM;AACxB,aAAO,kBAAkB;AACzB,yBAAmB,iBAAiB,OAAO,YAAY;AACvD,WAAK,YAAY,CAAC;AAClB,iBAAW,SAAS,cAAc;AAChC,aAAK,UAAU,KAAK,IAAI,UAAU,EAAE,OAAO,MAAM,MAAM,CAAC,CAAC;AAAA,MAC3D;AACA,wBAAkB,gBAAgB,OAAO,KAAK,SAAS;AAAA,IACzD;AACA,QAAI,aAAa,MAAM;AACrB,aAAO,eAAe;AACtB,yBAAmB,iBAAiB,OAAO,SAAS;AACpD,WAAK,eAAe,UAAU;AAAA,IAChC;AACA,UAAMM,YAAW,iBAAiB,cAAc;AAChD,QAAIA,WAAU;AACZ,YAAM,YAAY,CAAC,MAAM,EAAE,OAAO,gBAAgB;AAClD,YAAM,gBAAgB,KAAK,UAAU,OAAO,eAAe;AAC3D,YAAM,oBAAoB,KAAK;AAC/B,WAAK,YAAY;AACjB,YAAM,SAAS,MAAM,MAAM,WAAW,MAAM;AAC5C,WAAK,YAAY;AACjB,aAAO;AAAA,IACT,OAAO;AACL,aAAO,MAAM,MAAM,QAAQ,MAAM;AAAA,IACnC;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,OAAO,UAAU,OAAO,OAAO,OAAO;AAC5C,YAAM,WAAW,UAAU,OAAO,OAAO,OAAO;AAChD,UAAI,eAAe,UAAU,OAAO,OAAO,OAAO;AAClD,eAAS,oBAAoB,MAAM;AACnC,UAAI,gBAAgB,MAAM;AACxB,YAAI,KAAK,UAAU;AACjB,yBAAe,KAAK;AAAA,QACtB,OAAO;AACL,yBAAe,KAAK,gBAAgB,MAAM;AAAA,QAC5C;AAAA,MACF;AACA,YAAM,YAAY,MAAM,QAAQ,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,UAAU,SAAS;AACpF,UAAI,aAAa,WAAW,WAAW;AACrC,cAAM,IAAI,WAAW,iBAAiB,qCAAqC,aAAa,0BAA0B;AAAA,MACpH;AACA,UAAI,KAAK,QAAQ;AACf,gBAAQ,KAAK,kEAAkE;AAAA,MACjF;AACA,YAAM,iBAAiB,EAAE,SAAS;AAClC,YAAM,QAAQ,CAAC,SAAS,YAAY;AAClC,cAAM,WAAW,KAAK,KAAK,KAAK,CAAC,OAAO,EAAE,OAAO,OAAO,GAAG,cAAc;AACzE,eAAO,CAAC,SAAS,IAAI,SAAS,MAAM,CAAC,CAAC;AAAA,MACxC;AACA,YAAM,aAAa,IAAI,OAAO,QAAQ,cAAc,KAAK,aAAa,MAAM,MAAM,KAAK,QAAQ,KAAK,eAAe;AACnH,YAAM,aAAa,WAAW;AAC9B,YAAM,UAAU,WAAW;AAC3B,YAAM,SAAS,WAAW;AAC1B,UAAI,KAAK,UAAU;AACjB,aAAK,YAAY,QAAQ,QAAQ;AAAA,MACnC;AACA,YAAM,SAAS,KAAK,kBAAkB,UAAU;AAChD,UAAI,KAAK,aAAa;AACpB,eAAO,CAAC,MAAM,EAAE,OAAO,MAAM;AAAA,MAC/B,OAAO;AACL,eAAO;AAAA,MACT;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,gBAAgB,QAAQ;AACtB,WAAO,KAAK,MAAM;AAChB,UAAI,eAAe,MAAM,OAAO,KAAK;AACrC,qBAAe,KAAK,cAAc,CAAC,GAAG,CAAC,CAAC;AACxC,qBAAe,YAAY,YAAY;AACvC,UAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,eAAO,KAAK,KAAK,UAAU,IAAI,CAAC,QAAQ,MAAM,IAAI,MAAM,cAAc,CAAC,GAAG,GAAG,CAAC,IAAI,YAAY;AAAA,MAChG,OAAO;AACL,eAAO,KAAK,KAAK,YAAY,IAAI,CAAC,MAAM,cAAc,CAAC,GAAG,KAAK,KAAK,SAAS,CAAC,CAAC,IAAI,CAAC,YAAY;AAAA,MAClG;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,IAAI,mBAAmB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,aAAO,CAAC;AAAA,IACV;AACA,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EACA,IAAI,sBAAsB;AACxB,QAAI,CAAC,KAAK,WAAW;AACnB,aAAO,KAAK,KAAK;AAAA,IACnB;AACA,WAAO,KAAK,KAAK;AAAA,EACnB;AAAA,EACA,6BAA6B,OAAO;AAClC,UAAM,6BAA6B,KAAK;AACxC,QAAI,KAAK,QAAQ,MAAM;AACrB,WAAK,KAAK,6BAA6B,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS;AAAA,MACb,iBAAiB,KAAK;AAAA,MACtB,aAAa,KAAK;AAAA,MAClB,aAAa,KAAK;AAAA,MAClB,UAAU,KAAK;AAAA,MACf,QAAQ,KAAK;AAAA,IACf;AACA,QAAI,KAAK,gBAAgB,MAAM;AAC7B,aAAO,kBAAkB,KAAK;AAAA,IAChC;AACA,UAAM,aAAa,KAAK,KAAK,UAAU;AACvC,QAAI,KAAK,aAAa,MAAM,IAAI,WAAW;AACzC,aAAO,UAAU;AAAA,QACf,aAAa,KAAK,KAAK,aAAa;AAAA,QACpC,UAAU;AAAA,MACZ;AAAA,IACF;AACA,WAAO,OAAO,OAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,UAAU,GAAG,MAAM;AAAA,EACvF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ,gBAAgB,CAAC,GAAG;AACjD,UAAM,aAAa,OAAO;AAC1B,UAAM,OAAO,YAAY,YAAY,aAAa;AAClD,WAAO,IAAI,IAAI,OAAO,OAAO,QAAQ,EAAE,KAAK,CAAC,CAAC;AAAA,EAChD;AACF;AACA,IAAI,YAAY;AAChB,sBAAsB,cAAc,GAAG;AACvC,IAAI,UAAU,cAAc,MAAM;AAClC;AACA,IAAI,gBAAgB,cAAc,QAAQ;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,qBAAqB;AAC1B,SAAK,6BAA6B;AAClC,SAAK,gCAAgC;AACrC,SAAK,2BAA2B;AAChC,SAAK,QAAQ,KAAK;AAClB,0BAAsB,KAAK,OAAO,OAAO;AACzC,SAAK,aAAa,cAAc,KAAK,cAAc,OAAO,KAAK,qBAAqB,KAAK,UAAU;AACnG,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,oBAAoB,eAAe,KAAK,qBAAqB,KAAK,0BAA0B;AACjG,SAAK,uBAAuB,eAAe,KAAK,wBAAwB,KAAK,6BAA6B;AAC1G,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,KAAK,wBAAwB;AAC3F,SAAK,oBAAoB,eAAe,KAAK,iBAAiB;AAC9D,SAAK,uBAAuB,eAAe,KAAK,oBAAoB;AACpE,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,mBAAmB,cAAc,KAAK,gBAAgB;AAC3D,SAAK,sBAAsB,cAAc,KAAK,mBAAmB;AACjE,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,UAAU,KAAK,CAAC,GAAG,KAAK,CAAC,GAAG,KAAK,WAAW,OAAO,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC;AAC3E,SAAK,mBAAmB,KAAK;AAAA,MAC3B;AAAA,MACA,KAAK,CAAC,GAAG,KAAK,oBAAoB,OAAO,IAAI,KAAK,gBAAgB,CAAC;AAAA,IACrE,CAAC;AACD,SAAK,cAAc,KAAK;AACxB,SAAK,YAAY,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,SAAK,SAAS,KAAK,UAAU,UAAU,CAAC,WAAW,WAAW,SAAS,IAAI,KAAK,KAAK,GAAG,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACzK,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,CAAC,KAAK,OAAO,KAAK,KAAK,GAAG,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,MAAM,KAAK,mBAAmB;AAC9K,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,KAAK,GAAG,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IAC9H,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS;AACT,UAAI,OAAO,WAAW,GAAG;AACvB,cAAM,IAAI,WAAW,8CAA8C,OAAO,SAAS;AAAA,MACrF;AACA,UAAI,aAAa,OAAO;AACxB,eAAS,OAAO;AAChB,YAAM,WAAW,OAAO,eAAe,OAAO,QAAQ,OAAO;AAC7D,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,eAAe,MAAM;AACpE,aAAK,cAAc,oBAAoB;AAAA,UACrC,MAAM,MAAM,SAAS,MAAM;AAAA,UAC3B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,KAAK,KAAK,wBAAwB,MAAM;AAC/F,aAAK,uBAAuB,oBAAoB;AAAA,UAC9C,MAAM,MAAM,SAAS,UAAU;AAAA,UAC/B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,UAAI;AACJ,YAAM,SAAS,KAAK;AACpB,YAAM,YAAY,KAAK;AACvB,UAAI,UAAU,MAAM;AAClB,YAAI,KAAK,IAAI,QAAQ,MAAM,GAAG,KAAK,OAAO,KAAK,CAAC;AAAA,MAClD,OAAO;AACL,YAAI,KAAK,QAAQ,KAAK,OAAO,KAAK,CAAC;AAAA,MACrC;AACA,UAAI,KAAK,QAAQ,MAAM;AACrB,YAAI,QAAQ,GAAG,KAAK,KAAK,KAAK,CAAC;AAAA,MACjC;AACA,UAAI,aAAa,MAAM;AACrB,qBAAa,IAAI,YAAY,SAAS;AAAA,MACxC;AACA,UAAI,SAAS,KAAK,GAAG,KAAK,YAAY,KAAK,gBAAgB,KAAK,CAAC,CAAC;AAClE,UAAI,KAAK,cAAc,MAAM;AAC3B,iBAAS,KAAK,WAAW,MAAM,MAAM;AAAA,MACvC;AACA,aAAO,CAAC,QAAQ,MAAM;AAAA,IACxB,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,YAAY,oBAAoB,KAAK,UAAU;AAAA,MAC/C,SAAS,KAAK;AAAA,MACd,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,kBAAkB,oBAAoB,KAAK,gBAAgB;AAAA,MAC3D,qBAAqB,oBAAoB,KAAK,mBAAmB;AAAA,MACjE,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,MACvD,SAAS,KAAK;AAAA,MACd,kBAAkB,KAAK;AAAA,IACzB;AACA,WAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,MAAM;AAAA,EAC5D;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AACjD,IAAI,YAAY,cAAc,IAAI;AAAA,EAChC,YAAY,MAAM;AAChB,SAAK,OAAO,IAAI,cAAc,IAAI;AAClC,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,KAAK,KAAK,eAAe,MAAM;AACjC,gBAAQ,KAAK,KAAK,WAAW;AAC7B,aAAK,KAAK,cAAc;AAAA,MAC1B;AACA,UAAI,KAAK,KAAK,wBAAwB,MAAM;AAC1C,gBAAQ,KAAK,KAAK,oBAAoB;AACtC,aAAK,KAAK,uBAAuB;AAAA,MACnC;AACA,YAAM,OAAO,UAAU,OAAO,OAAO,OAAO;AAC5C,YAAM,WAAW,UAAU,OAAO,OAAO,OAAO;AAChD,YAAM,eAAe,UAAU,OAAO,OAAO,OAAO;AACpD,aAAO,MAAM,KAAK,QAAQ,EAAE,MAAM,UAAU,aAAa,CAAC;AAAA,IAC5D,CAAC;AAAA,EACH;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,MAAM;AAAA,EACvB;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,IAAI,UAAU,cAAc,QAAQ;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,qBAAqB;AAC1B,SAAK,+BAA+B;AACpC,SAAK,6BAA6B;AAClC,SAAK,gCAAgC;AACrC,SAAK,2BAA2B;AAChC,QAAI,KAAK,YAAY;AACnB,YAAM,IAAI,WAAW,6DAA6D;AAAA,IACpF;AACA,SAAK,QAAQ,KAAK;AAClB,0BAAsB,KAAK,OAAO,OAAO;AACzC,SAAK,aAAa,cAAc,KAAK,eAAe,SAAS,KAAK,qBAAqB,KAAK,UAAU;AACtG,SAAK,sBAAsB,cAAc,KAAK,wBAAwB,SAAS,KAAK,+BAA+B,KAAK,mBAAmB;AAC3I,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,oBAAoB,eAAe,KAAK,qBAAqB,KAAK,0BAA0B;AACjG,SAAK,uBAAuB,eAAe,KAAK,wBAAwB,KAAK,6BAA6B;AAC1G,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,KAAK,wBAAwB;AAC3F,SAAK,oBAAoB,eAAe,KAAK,iBAAiB;AAC9D,SAAK,uBAAuB,eAAe,KAAK,oBAAoB;AACpE,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,mBAAmB,cAAc,KAAK,gBAAgB;AAC3D,SAAK,sBAAsB,cAAc,KAAK,mBAAmB;AACjE,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,UAAU,KAAK,CAAC,GAAG,KAAK,CAAC,GAAG,KAAK,WAAW,OAAO,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC;AAC3E,SAAK,mBAAmB,KAAK;AAAA,MAC3B;AAAA,MACA,KAAK,CAAC,GAAG,KAAK,oBAAoB,OAAO,IAAI,KAAK,gBAAgB,CAAC;AAAA,IACrE,CAAC;AACD,SAAK,cAAc,KAAK;AACxB,SAAK,iBAAiB,KAAK;AAC3B,SAAK,YAAY,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,WAAW,WAAW,WAAW,SAAS;AAChD,SAAK,SAAS,KAAK,UAAU,UAAU,CAAC,UAAU,KAAK,QAAQ,CAAC,GAAG,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACpJ,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,CAAC,KAAK,OAAO,KAAK,QAAQ,CAAC,GAAG,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,MAAM,KAAK,mBAAmB;AAClL,QAAI,KAAK,SAAS;AAChB,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,QAAQ,CAAC,GAAG,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IAClI,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS;AACT,UAAI,OAAO,WAAW,GAAG;AACvB,cAAM,IAAI,WAAW,uDAAuD,OAAO,SAAS;AAAA,MAC9F;AACA,YAAM,WAAW,OAAO,eAAe,OAAO,QAAQ,OAAO;AAC7D,UAAI,WAAW,OAAO;AACtB,eAAS,OAAO;AAChB,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,eAAe,MAAM;AACpE,aAAK,cAAc,oBAAoB;AAAA,UACrC,MAAM,MAAM,SAAS,MAAM;AAAA,UAC3B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,KAAK,KAAK,wBAAwB,MAAM;AAC/F,aAAK,uBAAuB,oBAAoB;AAAA,UAC9C,MAAM,MAAM,SAAS,QAAQ;AAAA,UAC7B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,YAAM,SAAS,KAAK;AACpB,YAAM,YAAY,KAAK;AACvB,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,GAAG;AACxC,iBAAS,IAAI,QAAQ,OAAO,EAAE;AAAA,MAChC;AACA,UAAI,UAAU,KAAK,QAAQ,KAAK,OAAO,KAAK,CAAC;AAC7C,UAAI,KAAK,SAAS;AAChB,kBAAU,QAAQ,SAAS,KAAK,KAAK,KAAK,CAAC;AAAA,MAC7C;AACA,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,GAAG;AAC1D,mBAAW,IAAI,UAAU,UAAU,EAAE;AAAA,MACvC;AACA,YAAM,uBAAuB,KAAK,gBAAgB,KAAK;AACvD,YAAM,CAAC,KAAK,GAAG,IAAI,MAAM,sBAAsB,CAAC,IAAI,KAAK,OAAO,KAAK,KAAK,GAAG,qBAAqB,OAAO,CAAC;AAC1G,YAAM,cAAc,KAAK,UAAU,GAAG;AACtC,YAAM,CAAC,IAAI,IAAI,EAAE,IAAI,MAAM,SAAS,GAAG,QAAQ,OAAO,CAAC;AACvD,YAAM,CAAC,YAAY,UAAU,IAAI,MAAM,aAAa,GAAG,YAAY,OAAO,CAAC;AAC3E,UAAI,KAAK,oBAAoB,MAAM,KAAK,IAAI,UAAU,CAAC;AACvD,UAAI,KAAK,oBAAoB,MAAM,KAAK,IAAI,UAAU,CAAC;AACvD,YAAM,aAAa,KAAK,IAAI,GAAG,QAAQ,GAAG,GAAG;AAC7C,WAAK,KAAK,WAAW,MAAM,KAAK,IAAI,UAAU,CAAC;AAC/C,YAAM,IAAI,KAAK,IAAI,GAAG,QAAQ,GAAG,IAAI,KAAK,GAAG,IAAI,CAAC,CAAC,GAAG,EAAE,CAAC;AACzD,aAAO,CAAC,GAAG,CAAC;AAAA,IACd,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,YAAY,oBAAoB,KAAK,UAAU;AAAA,MAC/C,qBAAqB,oBAAoB,KAAK,mBAAmB;AAAA,MACjE,SAAS,KAAK;AAAA,MACd,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,kBAAkB,oBAAoB,KAAK,gBAAgB;AAAA,MAC3D,qBAAqB,oBAAoB,KAAK,mBAAmB;AAAA,MACjE,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,MACvD,SAAS,KAAK;AAAA,MACd,kBAAkB,KAAK;AAAA,MACvB,gBAAgB,KAAK;AAAA,MACrB,YAAY;AAAA,IACd;AACA,WAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,MAAM;AAAA,EAC5D;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,MAAM,cAAc,IAAI;AAAA,EAC1B,YAAY,MAAM;AAChB,QAAI,KAAK,mBAAmB,GAAG;AAC7B,cAAQ,KAAK,gHAAgH;AAAA,IAC/H;AACA,SAAK,OAAO,IAAI,QAAQ,IAAI;AAC5B,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,KAAK,KAAK,eAAe,MAAM;AACjC,gBAAQ,KAAK,KAAK,WAAW;AAC7B,aAAK,KAAK,cAAc;AAAA,MAC1B;AACA,UAAI,KAAK,KAAK,wBAAwB,MAAM;AAC1C,gBAAQ,KAAK,KAAK,oBAAoB;AACtC,aAAK,KAAK,uBAAuB;AAAA,MACnC;AACA,YAAM,OAAO,UAAU,OAAO,OAAO,OAAO;AAC5C,YAAM,WAAW,UAAU,OAAO,OAAO,OAAO;AAChD,YAAM,eAAe,UAAU,OAAO,OAAO,OAAO;AACpD,aAAO,MAAM,KAAK,QAAQ,EAAE,MAAM,UAAU,aAAa,CAAC;AAAA,IAC5D,CAAC;AAAA,EACH;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,QAAI,OAAO,qBAAqB,GAAG;AACjC,aAAO,oBAAoB;AAAA,IAC7B;AACA,WAAO,IAAI,IAAI,MAAM;AAAA,EACvB;AACF;AACA,IAAI,YAAY;AAChB,sBAAsB,cAAc,GAAG;AACvC,IAAI,WAAW,cAAc,QAAQ;AAAA,EACnC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,qBAAqB;AAC1B,SAAK,+BAA+B;AACpC,SAAK,6BAA6B;AAClC,SAAK,gCAAgC;AACrC,SAAK,2BAA2B;AAChC,SAAK,QAAQ,KAAK;AAClB,0BAAsB,KAAK,OAAO,OAAO;AACzC,SAAK,aAAa,cAAc,KAAK,eAAe,SAAS,KAAK,qBAAqB,KAAK,UAAU;AACtG,SAAK,sBAAsB,cAAc,KAAK,wBAAwB,SAAS,KAAK,+BAA+B,KAAK,mBAAmB;AAC3I,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,oBAAoB,eAAe,KAAK,qBAAqB,KAAK,0BAA0B;AACjG,SAAK,uBAAuB,eAAe,KAAK,wBAAwB,KAAK,6BAA6B;AAC1G,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,KAAK,wBAAwB;AAC3F,SAAK,iBAAiB,KAAK;AAC3B,SAAK,oBAAoB,eAAe,KAAK,iBAAiB;AAC9D,SAAK,uBAAuB,eAAe,KAAK,oBAAoB;AACpE,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,mBAAmB,cAAc,KAAK,gBAAgB;AAC3D,SAAK,sBAAsB,cAAc,KAAK,mBAAmB;AACjE,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,UAAU,KAAK,CAAC,GAAG,KAAK,CAAC,GAAG,KAAK,WAAW,OAAO,IAAI,KAAK,OAAO,CAAC,CAAC,CAAC;AAC3E,SAAK,mBAAmB,KAAK;AAAA,MAC3B;AAAA,MACA,KAAK,CAAC,GAAG,KAAK,oBAAoB,OAAO,IAAI,KAAK,gBAAgB,CAAC;AAAA,IACrE,CAAC;AACD,SAAK,cAAc,KAAK;AACxB,SAAK,iBAAiB,KAAK;AAC3B,SAAK,YAAY,CAAC,KAAK,OAAO,KAAK,KAAK;AACxC,SAAK,cAAc;AACnB,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EACA,MAAM,YAAY;AAChB,QAAI;AACJ,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,WAAW,WAAW,WAAW,SAAS;AAChD,SAAK,SAAS,KAAK,UAAU,UAAU,CAAC,UAAU,KAAK,QAAQ,CAAC,GAAG,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACpJ,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,CAAC,KAAK,OAAO,KAAK,QAAQ,CAAC,GAAG,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,MAAM,KAAK,mBAAmB;AAClL,QAAI;AACJ,QAAI,KAAK,SAAS;AAChB,UAAI,KAAK,gBAAgB;AACvB,cAAM,mBAAmB,KAAK;AAC9B,cAAM,gBAAgB,KAAK;AAC3B,0BAAkB,KAAK,KAAK,MAAM,mBAAmB,YAAY;AAAA,UAC/D,MAAM,OAAO,OAAO;AAClB,kBAAM,KAAK,iBAAiB,MAAM,CAAC,aAAa,CAAC;AACjD,kBAAM,KAAK,IAAI,KAAK,EAAE,MAAM,CAAC,aAAa,CAAC;AAC3C,kBAAM,SAAS,iBAAiB,MAAM,CAAC,gBAAgB,CAAC,CAAC;AACzD,mBAAO,qBAAqB,qBAAqB,IAAI,EAAE,GAAG,MAAM;AAAA,UAClE;AAAA,QACF,GAAG,GAAG,YAAY,cAAc,IAAI;AAAA,MACtC,OAAO;AACL,0BAAkB,KAAK;AAAA,MACzB;AACA,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,QAAQ,CAAC,GAAG,MAAM,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IAC7H,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,WAAW,OAAO,eAAe,OAAO,QAAQ,OAAO;AAC7D,eAAS;AACT,UAAI,OAAO,WAAW,GAAG;AACvB,cAAM,IAAI,WAAW,wDAAwD,OAAO,SAAS;AAAA,MAC/F;AACA,UAAI,WAAW,OAAO;AACtB,YAAM,WAAW,OAAO;AACxB,eAAS,OAAO;AAChB,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,eAAe,MAAM;AACpE,aAAK,cAAc,oBAAoB;AAAA,UACrC,MAAM,MAAM,SAAS,MAAM;AAAA,UAC3B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,KAAK,KAAK,wBAAwB,MAAM;AAC/F,aAAK,uBAAuB,oBAAoB;AAAA,UAC9C,MAAM,MAAM,SAAS,QAAQ;AAAA,UAC7B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,YAAM,SAAS,KAAK;AACpB,YAAM,YAAY,KAAK;AACvB,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,GAAG;AACxC,iBAAS,IAAI,QAAQ,OAAO,EAAE;AAAA,MAChC;AACA,UAAI,IAAI,KAAK,QAAQ,KAAK,OAAO,KAAK,CAAC;AACvC,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,GAAG;AAC1D,mBAAW,IAAI,UAAU,UAAU,EAAE;AAAA,MACvC;AACA,UAAI,KAAK,GAAG,KAAK,UAAU,KAAK,gBAAgB,KAAK,CAAC,CAAC;AACvD,UAAI,KAAK,SAAS;AAChB,YAAI,QAAQ,GAAG,KAAK,KAAK,KAAK,CAAC;AAAA,MACjC;AACA,YAAM,CAAC,IAAI,IAAI,IAAI,EAAE,IAAI,MAAM,GAAG,GAAG,EAAE,OAAO,CAAC;AAC/C,UAAI,KAAK,oBAAoB,MAAM,EAAE;AACrC,UAAI,KAAK,oBAAoB,MAAM,EAAE;AACrC,UAAI,KAAK,IAAI,GAAG,QAAQ,GAAG,IAAI,GAAG,KAAK,WAAW,MAAM,EAAE,CAAC,CAAC;AAC5D,UAAI,KAAK,oBAAoB,MAAM,EAAE;AACrC,YAAM,IAAI,IAAI,GAAG,KAAK,WAAW,MAAM,CAAC,CAAC;AACzC,aAAO,CAAC,GAAG,GAAG,CAAC;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,YAAY,oBAAoB,KAAK,UAAU;AAAA,MAC/C,qBAAqB,oBAAoB,KAAK,mBAAmB;AAAA,MACjE,SAAS,KAAK;AAAA,MACd,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,gBAAgB,KAAK;AAAA,MACrB,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,sBAAsB,qBAAqB,KAAK,oBAAoB;AAAA,MACpE,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,kBAAkB,oBAAoB,KAAK,gBAAgB;AAAA,MAC3D,qBAAqB,oBAAoB,KAAK,mBAAmB;AAAA,MACjE,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,MACvD,SAAS,KAAK;AAAA,MACd,kBAAkB,KAAK;AAAA,MACvB,gBAAgB,KAAK;AAAA,IACvB;AACA,WAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,MAAM;AAAA,EAC5D;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,OAAO,cAAc,IAAI;AAAA,EAC3B,YAAY,MAAM;AAChB,QAAI,KAAK,mBAAmB,GAAG;AAC7B,cAAQ,KAAK,gHAAgH;AAAA,IAC/H;AACA,SAAK,OAAO,IAAI,SAAS,IAAI;AAC7B,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,KAAK,KAAK,eAAe,MAAM;AACjC,gBAAQ,KAAK,KAAK,WAAW;AAC7B,aAAK,KAAK,cAAc;AAAA,MAC1B;AACA,UAAI,KAAK,KAAK,wBAAwB,MAAM;AAC1C,gBAAQ,KAAK,KAAK,oBAAoB;AACtC,aAAK,KAAK,uBAAuB;AAAA,MACnC;AACA,YAAM,OAAO,UAAU,OAAO,OAAO,OAAO;AAC5C,YAAM,WAAW,UAAU,OAAO,OAAO,OAAO;AAChD,YAAM,eAAe,UAAU,OAAO,OAAO,OAAO;AACpD,aAAO,MAAM,KAAK,QAAQ,EAAE,MAAM,UAAU,aAAa,CAAC;AAAA,IAC5D,CAAC;AAAA,EACH;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,QAAI,OAAO,qBAAqB,GAAG;AACjC,aAAO,oBAAoB;AAAA,IAC7B;AACA,WAAO,IAAI,IAAI,MAAM;AAAA,EACvB;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,IAAI,kBAAkB,cAAc,QAAQ;AAAA,EAC1C,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,QAAQ,KAAK;AAAA,EACpB;AAAA,EACA,IAAI,YAAY;AACd,UAAM,YAAY,CAAC;AACnB,eAAW,QAAQ,KAAK,MAAM,MAAM,EAAE,QAAQ,GAAG;AAC/C,UAAI,MAAM,QAAQ,KAAK,SAAS,GAAG;AACjC,kBAAU,KAAK,GAAG,KAAK,SAAS;AAAA,MAClC,OAAO;AACL,kBAAU,KAAK,KAAK,SAAS;AAAA,MAC/B;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS;AACT,UAAI,SAAS,OAAO,MAAM,CAAC;AAC3B,YAAM,eAAe,CAAC;AACtB,iBAAW,QAAQ,KAAK,MAAM,MAAM,EAAE,QAAQ,GAAG;AAC/C,YAAI,MAAM,QAAQ,KAAK,SAAS,GAAG;AACjC,uBAAa,KAAK,OAAO,OAAO,GAAG,KAAK,UAAU,MAAM,CAAC;AAAA,QAC3D,OAAO;AACL,uBAAa,KAAK,OAAO,OAAO,GAAG,CAAC,CAAC;AAAA,QACvC;AAAA,MACF;AACA,mBAAa,QAAQ;AACrB,YAAM,kBAAkB,CAAC;AACzB,UAAI;AACJ,eAAS,IAAI,GAAG,IAAI,KAAK,MAAM,QAAQ,EAAE,GAAG;AAC1C,cAAM,OAAO,KAAK,MAAM;AACxB,iBAAS,aAAa;AACtB,YAAI,MAAM,GAAG;AACX,uBAAa,CAAC,OAAO,EAAE,EAAE,OAAO,MAAM;AAAA,QACxC,OAAO;AACL,uBAAa,CAAC,WAAW,EAAE,EAAE,OAAO,MAAM;AAAA,QAC5C;AACA,qBAAa,KAAK,KAAK,YAAY,MAAM;AACzC,wBAAgB,KAAK,WAAW,MAAM,CAAC,CAAC;AAAA,MAC1C;AACA,eAAS,CAAC;AACV,iBAAW,cAAc,gBAAgB,MAAM,EAAE,QAAQ,GAAG;AAC1D,eAAO,KAAK,GAAG,UAAU;AAAA,MAC3B;AACA,aAAO,CAAC,WAAW,EAAE,EAAE,OAAO,MAAM;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EACA,MAAM,YAAY;AAChB,QAAI,gBAAgB,UAAU,GAAG;AAC/B,mBAAa,WAAW;AAAA,IAC1B;AACA,iBAAa;AACb,QAAI;AACJ,SAAK,MAAM,QAAQ,CAAC,MAAM,MAAM;AAC9B,gBAAU,WAAW,KAAK,MAAM;AAC9B,aAAK,MAAM,UAAU;AACrB,YAAI,MAAM,QAAQ,KAAK,SAAS,GAAG;AACjC,sBAAY,KAAK,UAAU;AAAA,QAC7B,OAAO;AACL,sBAAY,KAAK;AAAA,QACnB;AACA,qBAAa,CAAC,WAAW,IAAI,SAAS;AAAA,MACxC,CAAC;AAAA,IACH,CAAC;AACD,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,gBAAgB,CAAC,SAAS;AAC9B,aAAO;AAAA,QACL,aAAa,KAAK,aAAa;AAAA,QAC/B,UAAU,KAAK,UAAU;AAAA,MAC3B;AAAA,IACF;AACA,UAAM,cAAc,KAAK,MAAM,IAAI,aAAa;AAChD,UAAM,SAAS,EAAE,SAAS,YAAY;AACtC,WAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,MAAM;AAAA,EAC5D;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ,gBAAgB,CAAC,GAAG;AACjD,UAAM,QAAQ,CAAC;AACf,eAAW,cAAc,OAAO,UAAU;AACxC,YAAM,KAAK,YAAY,YAAY,aAAa,CAAC;AAAA,IACnD;AACA,WAAO,IAAI,IAAI,EAAE,MAAM,CAAC;AAAA,EAC1B;AAAA,EACA,IAAI,mBAAmB;AACrB,QAAI,CAAC,KAAK,WAAW;AACnB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,UAAU,CAAC;AACjB,eAAW,QAAQ,KAAK,OAAO;AAC7B,cAAQ,KAAK,GAAG,KAAK,gBAAgB;AAAA,IACvC;AACA,WAAO;AAAA,EACT;AAAA,EACA,IAAI,sBAAsB;AACxB,UAAM,UAAU,CAAC;AACjB,eAAW,QAAQ,KAAK,OAAO;AAC7B,cAAQ,KAAK,GAAG,KAAK,mBAAmB;AAAA,IAC1C;AACA,QAAI,CAAC,KAAK,WAAW;AACnB,YAAM,mBAAmB,CAAC;AAC1B,iBAAW,QAAQ,KAAK,OAAO;AAC7B,yBAAiB,KAAK,GAAG,KAAK,gBAAgB;AAAA,MAChD;AACA,aAAO,iBAAiB,OAAO,OAAO;AAAA,IACxC;AACA,WAAO;AAAA,EACT;AAAA,EACA,aAAa;AACX,UAAM,UAAU,CAAC;AACjB,eAAW,QAAQ,KAAK,OAAO;AAC7B,cAAQ,KAAK,GAAG,KAAK,OAAO;AAAA,IAC9B;AACA,WAAO,cAAc,OAAO;AAAA,EAC9B;AAAA,EACA,WAAW,SAAS;AAClB,UAAM,SAAS,CAAC;AAChB,eAAW,QAAQ,KAAK,OAAO;AAC7B,YAAM,YAAY,KAAK,QAAQ;AAC/B,YAAM,eAAe,QAAQ,OAAO,SAAS;AAC7C,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,QAAQ,EAAE,GAAG;AAC5C,eAAO,KAAK,CAAC,KAAK,QAAQ,IAAI,aAAa,EAAE,CAAC;AAAA,MAChD;AAAA,IACF;AACA,kBAAc,MAAM;AAAA,EACtB;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,MAAM,OAAO,MAAM,WAAW,OAAO,OAAO,SAAS,GAAG,YAAY,IAAI;AAChF,QAAM,gBAAgB,MAAM,eAAe,OAAO,YAAY,MAAM,GAAG,IAAI,IAAI,SAAS,MAAM,GAAG,IAAI;AACrG,QAAM,aAAa,MAAM,aAAa,eAAe,OAAO,QAAQ;AACpE,MAAI,CAAC,UAAU,UAAU,GAAG;AAC1B,WAAO,KAAK,WAAW,EAAE,MAAM,CAAC;AAAA,EAClC;AACA,QAAM,QAAQ,MAAM,MAAM,EAAE,KAAK,MAAM,EAAE,IAAI,UAAU;AACvD,SAAO,MAAM,IAAI,CAAC,MAAM,KAAK,EAAE,MAAM,CAAC,CAAC;AACzC;AAGA,IAAI,SAAS,SAAS,GAAG,GAAG;AAC1B,MAAI,IAAI,CAAC;AACT,WAAS,MAAM;AACb,QAAI,OAAO,UAAU,eAAe,KAAK,GAAG,EAAE,KAAK,EAAE,QAAQ,EAAE,IAAI;AACjE,QAAE,MAAM,EAAE;AACd,MAAI,KAAK,QAAQ,OAAO,OAAO,0BAA0B;AACvD,aAAS,IAAI,GAAG,KAAK,OAAO,sBAAsB,CAAC,GAAG,IAAI,GAAG,QAAQ,KAAK;AACxE,UAAI,EAAE,QAAQ,GAAG,EAAE,IAAI,KAAK,OAAO,UAAU,qBAAqB,KAAK,GAAG,GAAG,EAAE;AAC7E,UAAE,GAAG,MAAM,EAAE,GAAG;AAAA,IACpB;AACF,SAAO;AACT;AACA,IAAI,YAAY,cAAc,IAAI;AAAA,EAChC,YAAY,MAAM;AAChB,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,oBAAoB,oDAAoD;AAAA,IACpF;AACA,QAAI,MAAM,QAAQ,KAAK,IAAI,GAAG;AAC5B,YAAM,IAAI,oBAAoB,gEAAgE;AAAA,IAChG;AACA,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,KAAK,KAAK,eAAe,MAAM;AACjC,gBAAQ,KAAK,KAAK,WAAW;AAC7B,aAAK,KAAK,cAAc;AAAA,MAC1B;AACA,UAAI,KAAK,KAAK,wBAAwB,MAAM;AAC1C,gBAAQ,KAAK,KAAK,oBAAoB;AACtC,aAAK,KAAK,uBAAuB;AAAA,MACnC;AACA,UAAI,UAAU,OAAO,cAAc;AACjC,cAAM,IAAI,WAAW,2CAA2C;AAAA,MAClE;AACA,YAAM,OAAO,UAAU,OAAO,OAAO,OAAO;AAC5C,YAAM,WAAW,UAAU,OAAO,OAAO,OAAO;AAChD,YAAM,eAAe,UAAU,OAAO,OAAO,OAAO;AACpD,aAAO,MAAM,KAAK,QAAQ,EAAE,MAAM,UAAU,aAAa,CAAC;AAAA,IAC5D,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,WAAW,KAAK,yBAAyB,UAAU;AACvD,QAAI,CAAC,KAAK,iBAAiB;AACzB,iBAAW,CAAC,SAAS,IAAI,GAAG,SAAS,MAAM,CAAC,CAAC;AAAA,IAC/C;AACA,QAAI,KAAK,aAAa;AACpB,iBAAW,CAAC,UAAU,GAAG,MAAM,CAAC,EAAE,KAAK,CAAC,WAAW,IAAI,GAAG,SAAS,MAAM,EAAE,CAAC,CAAC,CAAC;AAAA,IAChF;AACA,WAAO;AAAA,EACT;AAAA,EACA,gBAAgB,QAAQ;AACtB,WAAO,KAAK,MAAM;AAChB,YAAM,EAAE,UAAU,IAAI,KAAK;AAC3B,YAAM,aAAa,OAAO;AAC1B,YAAM,cAAc,KAAK,yBAAyB,UAAU;AAC5D,YAAM,aAAa,CAAC,YAAY,IAAI,GAAG,YAAY,MAAM,CAAC,CAAC;AAC3D,YAAM,eAAe,MAAM,UAAU;AACrC,UAAI,MAAM,QAAQ,SAAS,GAAG;AAC5B,eAAO,MAAM,UAAU,MAAM,EAAE,KAAK,YAAY;AAAA,MAClD;AACA,aAAO,CAAC,YAAY;AAAA,IACtB,CAAC;AAAA,EACH;AAAA,EACA,YAAY,QAAQ,WAAW,OAAO;AACpC,SAAK,MAAM;AACT,UAAI,CAAC,KAAK,UAAU;AAClB,cAAM,IAAI,eAAe,iEAAiE;AAAA,MAC5F;AACA,YAAM,aAAa,KAAK,UAAU,GAAG;AACrC,YAAM,cAAc,KAAK,yBAAyB,UAAU;AAC5D,YAAM,aAAa,CAAC,YAAY,IAAI,GAAG,YAAY,MAAM,CAAC,CAAC;AAC3D,YAAM,YAAY,WAAW;AAC7B,UAAI,aAAa,MAAM;AACrB,cAAM,IAAI,WAAW,uUAAuU;AAAA,MAC9V;AACA,UAAI,KAAK,UAAU,KAAK,MAAM;AAC5B,YAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,eAAK,UAAU,KAAK,KAAK,UAAU,IAAI,MAAM,MAAM,UAAU,CAAC;AAAA,QAChE,OAAO;AACL,eAAK,UAAU,CAAC,MAAM,UAAU,CAAC;AAAA,QACnC;AAAA,MACF,WAAW,UAAU,MAAM;AACzB,gBAAQ,KAAK,OAAO;AACpB,YAAI,KAAK,cAAc,MAAM;AAC3B,kBAAQ,KAAK,UAAU;AACvB,eAAK,aAAa,CAAC;AAAA,QACrB;AACA,YAAI,MAAM,QAAQ,KAAK,KAAK,SAAS,GAAG;AACtC,eAAK,UAAU,KAAK,KAAK,UAAU,IAAI,MAAM,MAAM,UAAU,CAAC;AAAA,QAChE,OAAO;AACL,eAAK,QAAQ,KAAK,MAAM,UAAU;AAAA,QACpC;AAAA,MACF,OAAO;AACL,YAAI,CAAC,MAAM,QAAQ,MAAM,GAAG;AAC1B,mBAAS,CAAC,MAAM;AAAA,QAClB;AACA,YAAI,OAAO,WAAW,KAAK,QAAQ,QAAQ;AACzC,gBAAM,IAAI,WAAW,SAAS,KAAK,gBAAgB,KAAK,QAAQ,oCAAoC,OAAO,0CAA0C,QAAQ;AAAA,QAC/J;AACA,YAAI,UAAU;AACZ,eAAK,WAAW,KAAK,KAAK,QAAQ,MAAM,CAAC;AAAA,QAC3C,OAAO;AACL,kBAAQ,KAAK,OAAO;AAAA,QACtB;AACA,iBAAS,QAAQ,GAAG,QAAQ,KAAK,QAAQ,QAAQ,EAAE,OAAO;AACxD,gBAAM,QAAQ,OAAO;AACrB,gBAAM,gBAAgB;AACtB,cAAI,CAAC,aAAa,YAAY,MAAM,OAAO,aAAa,GAAG;AACzD,kBAAM,IAAI,WAAW,SAAS,oCAAoC,KAAK,wBAAwB,iCAAiC,MAAM,OAAO;AAAA,UAC/I;AACA,eAAK,QAAQ,SAAS;AAAA,QACxB;AAAA,MACF;AACA,WAAK,UAAU,KAAK,QAAQ,IAAI,CAAC,UAAU,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,IAChE,CAAC;AAAA,EACH;AAAA,EACA,yBAAyB,YAAY;AACnC,UAAM,EAAE,YAAY,SAAS,YAAY,SAAS,SAAS,aAAa,IAAI,KAAK;AACjF,UAAM,kBAAkB,eAAe;AACvC,UAAM,IAAI,WAAW,kBAAkB,IAAI;AAC3C,UAAM,IAAI,WAAW,kBAAkB,IAAI;AAC3C,UAAM,OAAO,iBAAiB,GAAG,WAAW,IAAI,SAAS,QAAQ,IAAI,aAAa,EAAE;AACpF,UAAM,OAAO,iBAAiB,GAAG,WAAW,IAAI,SAAS,QAAQ,IAAI,aAAa,EAAE;AACpF,UAAM,WAAW;AAAA,MACf,GAAG,WAAW,MAAM,GAAG,CAAC;AAAA,MACxB,GAAG,kBAAkB,CAAC,SAAS,MAAM,IAAI,IAAI,CAAC,MAAM,MAAM,OAAO;AAAA,IACnE;AACA,WAAO;AAAA,EACT;AACF;AACA,UAAU,YAAY;AACtB,IAAI,iBAAiB,cAAc,SAAS;AAAA,EAC1C,YAAY,MAAM;AAChB,UAAM,EAAE,SAAS,YAAY,SAAS,SAAS,YAAY,aAAa,IAAI;AAC5E,UAAM,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,IAAI,GAAG,EAAE,OAAO,QAAQ,CAAC,CAAC;AAChE,SAAK,UAAU;AACf,0BAAsB,KAAK,SAAS,SAAS;AAC7C,SAAK,aAAa,eAAe,YAAY,GAAG,YAAY;AAC5D,SAAK,WAAW,QAAQ,CAAC,SAAS,sBAAsB,MAAM,YAAY,CAAC;AAC3E,SAAK,UAAU,eAAe,WAAW,GAAG,GAAG,SAAS;AACxD,SAAK,QAAQ,QAAQ,CAAC,WAAW,sBAAsB,QAAQ,SAAS,CAAC;AACzE,SAAK,UAAU,WAAW;AAC1B,qBAAiB,KAAK,OAAO;AAC7B,SAAK,aAAa,cAAc;AAChC,oBAAgB,KAAK,UAAU;AAC/B,SAAK,eAAe,eAAe,gBAAgB,GAAG,GAAG,cAAc;AACvE,SAAK,aAAa,QAAQ,CAAC,SAAS,sBAAsB,MAAM,cAAc,CAAC;AAAA,EACjF;AAAA,EACA,MAAM,YAAY;AAChB,QAAI;AACJ,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,KAAK,eAAe,kBAAkB,IAAI,WAAW,SAAS;AAClF,QAAI,WAAW,gBAAgB,MAAM;AACnC,YAAM,IAAI,WAAW,+DAA+D,WAAW,cAAc;AAAA,IAC/G;AACA,UAAM,WAAW,WAAW;AAC5B,UAAM,eAAe;AACrB,UAAM,cAAc,KAAK,WAAW,OAAO,CAAC,UAAU,KAAK,UAAU,YAAY,CAAC;AAClF,SAAK,SAAS,KAAK,UAAU,UAAU,aAAa,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACrI,UAAM,uBAAuB,KAAK,WAAW,OAAO,CAAC,KAAK,SAAS,KAAK,UAAU,YAAY,CAAC;AAC/F,SAAK,kBAAkB,KAAK,UAAU,oBAAoB,sBAAsB,MAAM,KAAK,sBAAsB,KAAK,sBAAsB,MAAM,KAAK,mBAAmB;AAC1K,QAAI,KAAK,SAAS;AAChB,UAAI;AACJ,UAAI,KAAK,gBAAgB;AACvB,cAAM,QAAQ,KAAK;AACnB,cAAM,UAAU,KAAK;AACrB,0BAAkB,KAAK,KAAK,MAAM,mBAAmB,YAAY;AAAA,UAC/D,MAAM,OAAO,OAAO;AAClB,kBAAM,QAAQ,MAAM,MAAM,CAAC,OAAO,CAAC;AACnC,kBAAM,QAAQ,MAAM,CAAC,OAAO,CAAC;AAC7B,kBAAM,YAAY,MAAM,MAAM,CAAC,UAAU,CAAC,CAAC;AAC3C,mBAAO,YAAY,CAAC,OAAO,OAAO,SAAS,CAAC;AAAA,UAC9C;AAAA,QACF,GAAG,GAAG,YAAY,cAAc,IAAI;AAAA,MACtC,OAAO;AACL,0BAAkB,KAAK;AAAA,MACzB;AACA,WAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,UAAU,YAAY,GAAG,MAAM,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IAC1I;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,OAAO,WAAW,GAAG;AACvB,cAAM,IAAI,WAAW,8DAA8D,OAAO,SAAS;AAAA,MACrG;AACA,YAAM,WAAW,OAAO,eAAe;AACvC,YAAM,IAAI,OAAO;AACjB,YAAM,WAAW,OAAO;AACxB,YAAM,WAAW,OAAO;AACxB,YAAM,eAAe;AACrB,UAAI,IAAI,KAAK,WAAW,KAAK,UAAU,KAAK,KAAK,eAAe,MAAM;AACpE,aAAK,cAAc,oBAAoB;AAAA,UACrC,MAAM,MAAM,SAAS,CAAC;AAAA,UACtB,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,YAAM,cAAc,KAAK;AACzB,YAAM,eAAe,CAAC,IAAI,MAAM,UAAU;AACxC,YAAI,CAAC,QAAQ,CAAC,KAAK,QAAQ;AACzB,iBAAO;AAAA,QACT;AACA,eAAO,IAAI,KAAK,QAAQ,EAAE;AAAA,MAC5B;AACA,UAAI,KAAK,aAAa,GAAG,aAAa,CAAC;AACvC,UAAI,KAAK,aAAa,GAAG,aAAa,CAAC;AACvC,UAAI,KAAK,aAAa,GAAG,aAAa,CAAC;AACvC,UAAI,KAAK,aAAa,GAAG,aAAa,CAAC;AACvC,UAAI,IAAI,KAAK,oBAAoB,KAAK,mBAAmB,KAAK,KAAK,wBAAwB,MAAM;AAC/F,aAAK,uBAAuB,oBAAoB;AAAA,UAC9C,MAAM,MAAM,SAAS,QAAQ;AAAA,UAC7B,MAAM,KAAK;AAAA,UACX;AAAA,UACA,OAAO;AAAA,UACP,aAAa,KAAK;AAAA,QACpB,CAAC;AAAA,MACH;AACA,YAAM,iBAAiB,KAAK;AAC5B,UAAI,KAAK,aAAa,UAAU,gBAAgB,CAAC;AACjD,UAAI,KAAK,aAAa,UAAU,gBAAgB,CAAC;AACjD,UAAI,KAAK,aAAa,UAAU,gBAAgB,CAAC;AACjD,UAAI,KAAK,aAAa,UAAU,gBAAgB,CAAC;AACjD,YAAM,oBAAoB;AAC1B,YAAM,CAAC,SAAS,SAAS,SAAS,OAAO,IAAI,MAAM,KAAK,OAAO,KAAK,GAAG,cAAc,iBAAiB;AACtG,YAAM,CAAC,OAAO,OAAO,OAAO,KAAK,IAAI,KAAK,UAAU,MAAM,KAAK,KAAK,KAAK,GAAG,YAAY,IAAI,CAAC,MAAM,MAAM,MAAM,IAAI;AACnH,WAAK,KAAK,UAAU,IAAI,SAAS,OAAO,KAAK,OAAO;AACpD,WAAK,KAAK,UAAU,IAAI,SAAS,OAAO,KAAK,OAAO;AACpD,WAAK,KAAK,UAAU,IAAI,SAAS,OAAO,KAAK,OAAO;AACpD,WAAK,KAAK,UAAU,IAAI,SAAS,OAAO,KAAK,OAAO;AACpD,YAAM,CAAC,YAAY,YAAY,YAAY,UAAU,IAAI,MAAM,KAAK,gBAAgB,KAAK,GAAG,cAAc,iBAAiB;AAC3H,WAAK,KAAK,cAAc,IAAI,UAAU;AACtC,WAAK,KAAK,cAAc,IAAI,UAAU;AACtC,WAAK,KAAK,cAAc,IAAI,UAAU;AACtC,WAAK,KAAK,cAAc,IAAI,UAAU;AACtC,YAAM,IAAI,KAAK,oBAAoB,MAAM,KAAK,IAAI,EAAE,CAAC;AACrD,YAAM,IAAI,KAAK,oBAAoB,MAAM,KAAK,IAAI,EAAE,CAAC;AACrD,YAAM,IAAI,KAAK,IAAI,GAAG,QAAQ,GAAG,IAAI,GAAG,KAAK,WAAW,MAAM,KAAK,IAAI,EAAE,CAAC,CAAC,CAAC;AAC5E,YAAM,IAAI,IAAI,KAAK,oBAAoB,MAAM,KAAK,IAAI,EAAE,CAAC,GAAG,KAAK,WAAW,MAAM,CAAC,CAAC;AACpF,aAAO,CAAC,GAAG,GAAG,CAAC;AAAA,IACjB,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,KAAK,MAAM,UAAU,GAAG,EAAE,SAAS,EAAE,IAAI,IAAI,aAAa,OAAO,IAAI,CAAC,OAAO,CAAC;AACpF,UAAM,SAAS;AAAA,MACb,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,MACjB,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,MACjB,cAAc,KAAK;AAAA,MACnB,SAAS,KAAK;AAAA,IAChB;AACA,WAAO,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,UAAU,GAAG,MAAM;AAAA,EAC5D;AAAA,EACA,UAAU,GAAG,GAAG,GAAG,SAAS;AAC1B,UAAM,MAAM,OAAO,GAAG,GAAG,KAAK,SAAS,WAAW,SAAS,KAAK,eAAe,kBAAkB,SAAS,QAAQ,KAAK,YAAY;AACnI,QAAI,GAAG;AACL,aAAO,QAAQ,KAAK,GAAG,KAAK,UAAU;AAAA,IACxC;AACA,WAAO;AAAA,EACT;AAAA,EACA,cAAc,GAAG,GAAG;AAClB,UAAM,UAAU;AAChB,WAAO,OAAO,GAAG,GAAG,SAAS,QAAQ,KAAK,eAAe,kBAAkB,SAAS,MAAM;AAAA,EAC5F;AACF;AACA,eAAe,YAAY;AAC3B,sBAAsB,cAAc,cAAc;AAClD,IAAI,aAAa,cAAc,UAAU;AAAA,EACvC,YAAY,MAAM;AAChB,UAAM,OAAO,IAAI,eAAe,IAAI;AACpC,UAAM,OAAO,OAAO,OAAO,OAAO,CAAC,GAAG,IAAI,GAAG,EAAE,KAAK,CAAC,CAAC;AAAA,EACxD;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,WAAO,IAAI,IAAI,MAAM;AAAA,EACvB;AACF;AACA,WAAW,YAAY;AACvB,sBAAsB,cAAc,UAAU;AAG9C,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,OAAO,KAAK,IAAI,KAAK,IAAI,KAAK,MAAM,CAAC,GAAG,CAAC;AAC9C,SAAK,aAAa,KAAK;AACvB,SAAK,OAAO,KAAK;AACjB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,cAAc,QAAQ;AACpB,QAAI,KAAK,cAAc,MAAM;AAC3B,aAAO,KAAK;AAAA,IACd;AACA,UAAM,aAAa,OAAO;AAC1B,UAAM,aAAa,CAAC;AACpB,aAAS,IAAI,GAAG,IAAI,KAAK,WAAW,QAAQ,EAAE,GAAG;AAC/C,iBAAW,KAAK,KAAK,WAAW,MAAM,OAAO,WAAW,KAAK,KAAK,WAAW,EAAE;AAAA,IACjF;AACA,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,UAAI,IAAI,KAAK,QAAQ,KAAK,OAAO,GAAG;AAClC,cAAM,WAAW,OAAO,eAAe,OAAO,QAAQ,OAAO;AAC7D,cAAM,aAAa,KAAK,cAAc,MAAM;AAC5C,cAAM,SAAS,aAAa,MAAM,SAAS,QAAQ,KAAK,MAAM,YAAY,KAAK,IAAI,GAAG,MAAM,QAAQ,QAAQ;AAC5G,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,MAAM,KAAK;AAAA,MACX,YAAY,KAAK;AAAA,MACjB,MAAM,KAAK;AAAA,IACb;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,UAAU;AACR,WAAO,MAAM,QAAQ;AAAA,EACvB;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,mBAAmB,cAAc,QAAQ;AAAA,EAC3C,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;AAAA,EAC/B;AAAA,EACA,cAAc,QAAQ;AACpB,UAAM,aAAa,OAAO;AAC1B,WAAO,CAAC,WAAW,IAAI,GAAG,WAAW,EAAE;AAAA,EACzC;AACF;AACA,iBAAiB,YAAY;AAC7B,sBAAsB,cAAc,gBAAgB;AACpD,IAAI,QAAQ,cAAc,MAAM;AAAA,EAC9B,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,aAAa;AAClB,SAAK,UAAU;AACf,SAAK,SAAS;AACd,SAAK,OAAO;AACZ,SAAK,6BAA6B;AAClC,SAAK,2BAA2B;AAChC,QAAI,KAAK,mBAAmB,QAAQ,KAAK,cAAc,QAAQ,KAAK,YAAY,MAAM;AACpF,UAAI,YAAY;AAChB,UAAI,KAAK,aAAa,MAAM;AAC1B,oBAAY,KAAK;AAAA,MACnB;AACA,WAAK,kBAAkB,CAAC,WAAW,KAAK,QAAQ;AAAA,IAClD;AACA,SAAK,QAAQ,KAAK;AAClB,0BAAsB,KAAK,OAAO,OAAO;AACzC,SAAK,aAAa,cAAc,KAAK,UAAU;AAC/C,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,UAAU,KAAK;AAAA,IACtB;AACA,SAAK,oBAAoB,eAAe,KAAK,qBAAqB,KAAK,0BAA0B;AACjG,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,KAAK,wBAAwB;AAC3F,SAAK,mBAAmB,cAAc,KAAK,gBAAgB;AAC3D,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,oBAAoB,eAAe,KAAK,iBAAiB;AAC9D,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,sBAAsB,eAAe,KAAK,mBAAmB;AAClE,SAAK,kBAAkB;AACvB,SAAK,YAAY,CAAC,EAAE,SAAS,EAAE,CAAC;AAAA,EAClC;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,eAAe,WAAW,WAAW,SAAS;AACpD,QAAI,KAAK,UAAU,MAAM;AACvB,WAAK,SAAS,KAAK,UAAU,UAAU,CAAC,cAAc,KAAK,KAAK,GAAG,MAAM,KAAK,mBAAmB,KAAK,mBAAmB,MAAM,KAAK,gBAAgB;AACpJ,UAAI,KAAK,SAAS;AAChB,aAAK,OAAO,KAAK,UAAU,QAAQ,CAAC,KAAK,KAAK,GAAG,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,MAC9H;AAAA,IACF;AACA,SAAK,YAAY,CAAC,EAAE,SAAS,GAAG,MAAM,EAAE,CAAC,KAAK,aAAa,EAAE,CAAC;AAC9D,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,WAAW,MAAM;AACrC,gBAAY,YAAY,SAAS,KAAK,KAAK;AAC3C,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,sBAAsB,2BAA2B,KAAK,WAAW,aAAa,CAAC;AACrF,UAAI;AACJ,UAAI,uBAAuB,MAAM;AAC/B,iBAAS,KAAK,QAAQ,KAAK,OAAO,KAAK,GAAG,qBAAqB,KAAK,OAAO,KAAK,KAAK,KAAK,IAAI,IAAI;AAAA,MACpG,OAAO;AACL,iBAAS,KAAK,QAAQ,KAAK,OAAO,KAAK,CAAC;AACxC,YAAI,KAAK,QAAQ,MAAM;AACrB,mBAAS,QAAQ,QAAQ,KAAK,KAAK,KAAK,CAAC;AAAA,QAC3C;AACA,YAAI,KAAK,cAAc,MAAM;AAC3B,mBAAS,KAAK,WAAW,MAAM,MAAM;AAAA,QACvC;AAAA,MACF;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,YAAY,oBAAoB,KAAK,UAAU;AAAA,MAC/C,SAAS,KAAK;AAAA,MACd,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,mBAAmB,qBAAqB,KAAK,iBAAiB;AAAA,MAC9D,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,kBAAkB,oBAAoB,KAAK,gBAAgB;AAAA,MAC3D,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,IACzD;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,MAAM,YAAY;AAClB,sBAAsB,cAAc,KAAK;AACzC,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,WAAO,QAAQ,CAAC;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,EAAE,SAAS,EAAE,CAAC;AAChC,SAAK,aAAa,KAAK;AAAA,EACzB;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,eAAW,OAAO,WAAW,MAAM,CAAC,GAAG;AACrC,UAAI,OAAO,MAAM;AACf,cAAM,IAAI,WAAW,iEAAiE,WAAW,MAAM,CAAC,kHAAkH;AAAA,MAC5N;AAAA,IACF;AACA,WAAO,CAAC,WAAW,IAAI,UAAU,YAAY,CAAC,CAAC;AAAA,EACjD;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,UAAI,SAAS,oBAAoB,MAAM;AACvC,UAAI,KAAK,eAAe,mBAAmB,OAAO,OAAO,GAAG;AAC1D,cAAM,cAAc,CAAC,CAAC;AACtB,iBAAS,IAAI,GAAG,IAAI,OAAO,MAAM,EAAE,GAAG;AACpC,sBAAY,KAAK,CAAC;AAAA,QACpB;AACA,oBAAY,KAAK,CAAC;AAClB,iBAAS,UAAU,QAAQ,WAAW;AAAA,MACxC;AACA,aAAO,aAAa,MAAM;AAAA,IAC5B,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS,CAAC;AAChB,QAAI,KAAK,cAAc,MAAM;AAC3B,aAAO,gBAAgB,KAAK;AAAA,IAC9B;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,cAAc,cAAc,MAAM;AAAA,EACpC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,kBAAkB;AACvB,SAAK,aAAa,cAAc,KAAK,UAAU;AAAA,EACjD;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,aAAO,KAAK,WAAW,MAAM,MAAM;AAAA,IACrC,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,YAAY,oBAAoB,KAAK,UAAU,EAAE;AAClE,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,IAAI,eAAe,cAAc,MAAM;AAAA,EACrC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,IAAI,KAAK;AACd,SAAK,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;AAAA,EAC/B;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO,CAAC,WAAW,IAAI,KAAK,GAAG,WAAW,EAAE;AAAA,EAC9C;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,aAAO,OAAO,QAAQ,KAAK,CAAC;AAAA,IAC9B,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,GAAG,KAAK;AAAA,IACV;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,cAAc,KAAK;AACxB,aAAS,IAAI,GAAG,IAAI,KAAK,YAAY,QAAQ,EAAE,GAAG;AAChD,UAAI,KAAK,UAAU,KAAK,YAAY,EAAE,GAAG;AACvC,aAAK,YAAY,KAAK;AAAA,MACxB;AAAA,IACF;AAAA,EACF;AAAA,EACA,UAAU,KAAK;AACb,WAAO,MAAM,KAAK,OAAO;AAAA,EAC3B;AAAA,EACA,oBAAoB,YAAY,aAAa;AAC3C,UAAM,WAAW;AACjB,UAAM,aAAa,YAAY,MAAM;AACrC,QAAI,QAAQ;AACZ,QAAI,UAAU;AACd,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,YAAM,MAAM,WAAW;AACvB,UAAI,KAAK,UAAU,GAAG,GAAG;AACvB,YAAI,YAAY,MAAM;AACpB,oBAAU;AAAA,QACZ,OAAO;AACL,gBAAM,IAAI,WAAW,0CAA0C;AAAA,QACjE;AAAA,MACF,OAAO;AACL,iBAAS;AAAA,MACX;AAAA,IACF;AACA,UAAM,eAAe,UAAU,UAAU;AACzC,QAAI,YAAY,MAAM;AACpB,UAAI,UAAU,KAAK,eAAe,UAAU,GAAG;AAC7C,cAAM,IAAI,WAAW,QAAQ;AAAA,MAC/B;AACA,iBAAW,WAAW,eAAe;AAAA,IACvC,WAAW,iBAAiB,OAAO;AACjC,YAAM,IAAI,WAAW,QAAQ;AAAA,IAC/B;AACA,WAAO;AAAA,EACT;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,iBAAiB;AACrB,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,UAAI,KAAK,UAAU,WAAW,EAAE,GAAG;AACjC,yBAAiB;AACjB;AAAA,MACF;AAAA,IACF;AACA,QAAI,gBAAgB;AAClB,aAAO,WAAW,MAAM,GAAG,CAAC,EAAE,OAAO,KAAK,WAAW;AAAA,IACvD,OAAO;AACL,aAAO,WAAW,MAAM,GAAG,CAAC,EAAE,OAAO,KAAK,oBAAoB,WAAW,MAAM,CAAC,GAAG,KAAK,WAAW,CAAC;AAAA,IACtG;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,aAAa,OAAO;AAC1B,YAAM,cAAc,WAAW,MAAM,GAAG,CAAC,EAAE,OAAO,KAAK,oBAAoB,WAAW,MAAM,CAAC,GAAG,KAAK,WAAW,CAAC;AACjH,aAAO,QAAQ,QAAQ,WAAW;AAAA,IACpC,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,aAAa,KAAK;AAAA,IACpB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,QAAI,KAAK,QAAQ,MAAM;AACrB,YAAM,IAAI,MAAM,iFAAiF;AAAA,IACnG;AACA,QAAI,CAAC,MAAM,QAAQ,KAAK,IAAI,GAAG;AAC7B,YAAM,IAAI,MAAM,sEAAsE,KAAK,eAAe;AAAA,IAC5G;AACA,UAAM,wBAAwB,OAAO,GAAG,KAAK,KAAK,SAAS,CAAC;AAC5D,QAAI,CAAC,aAAa,YAAY,KAAK,KAAK,MAAM,EAAE,KAAK,GAAG,qBAAqB,GAAG;AAC9E,YAAM,IAAI,MAAM,iCAAiC,KAAK,UAAU,KAAK,IAAI,IAAI,4DAA4D;AAAA,IAC3I;AACA,SAAK,OAAO,KAAK;AACjB,SAAK,qBAAqB,CAAC,CAAC,EAAE,OAAO,KAAK,IAAI;AAC9C,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,KAAK,KAAK,SAAS,EAAE,CAAC,CAAC;AAAA,EACjE;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,WAAW,MAAM;AACrC,SAAK,KAAK,QAAQ,CAAC,KAAK,MAAM;AAC5B,kBAAY,IAAI,KAAK,WAAW;AAAA,IAClC,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,UAAU,oBAAoB,MAAM,GAAG,KAAK,kBAAkB;AAAA,EACvE;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,MAAM,KAAK;AAAA,IACb;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,UAAM,QAAQ,OAAO,CAAC,IAAI,IAAI;AAC9B,SAAK,kBAAkB;AACvB,QAAI,QAAQ,MAAM;AAChB,WAAK,YAAY,KAAK,aAAa,OAAO,IAAI,KAAK;AAAA,IACrD,OAAO;AACL,WAAK,YAAY;AAAA,IACnB;AAAA,EACF;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS,EAAE,WAAW,KAAK,UAAU;AAC3C,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,UAAM,SAAS,oBAAoB,MAAM;AACzC,UAAM,OAAO;AACb,WAAO,IAAI,SAAS,QAAQ,KAAK,SAAS,GAAG,IAAI;AAAA,EACnD;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,OAAO;AACb,YAAM,WAAW;AACjB,YAAM,cAAc,IAAI,SAAS,QAAQ,KAAK,SAAS,GAAG,MAAM,QAAQ;AACxE,YAAM,SAAS,IAAI,QAAQ,KAAK,aAAa,OAAO,KAAK,CAAC;AAC1D,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAG3C,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,aAAa;AAClB,SAAK,iCAAiC;AACtC,QAAI,KAAK,mBAAmB,QAAQ,KAAK,cAAc,MAAM;AAC3D,UAAI,YAAY;AAChB,UAAI,KAAK,aAAa,MAAM;AAC1B,oBAAY,KAAK;AAAA,MACnB;AACA,UAAI,KAAK,eAAe,MAAM;AAC5B,aAAK,kBAAkB,CAAC,WAAW,IAAI;AAAA,MACzC,OAAO;AACL,aAAK,kBAAkB,CAAC,SAAS,EAAE,OAAO,OAAO,KAAK,WAAW,CAAC;AAAA,MACpE;AAAA,IACF;AACA,SAAK,WAAW,KAAK;AACrB,0BAAsB,KAAK,UAAU,UAAU;AAC/C,SAAK,YAAY,KAAK;AACtB,0BAAsB,KAAK,WAAW,WAAW;AACjD,SAAK,wBAAwB,eAAe,KAAK,yBAAyB,KAAK,8BAA8B;AAC7G,SAAK,wBAAwB,eAAe,KAAK,qBAAqB;AACtE,SAAK,sBAAsB,eAAe,KAAK,mBAAmB;AAClE,SAAK,uBAAuB,cAAc,KAAK,oBAAoB;AACnE,SAAK,WAAW,KAAK;AACrB,SAAK,kBAAkB,KAAK;AAC5B,SAAK,cAAc,KAAK;AAAA,EAC1B;AAAA,EACA,MAAM,YAAY;AAChB,SAAK,aAAa,KAAK,UAAU,cAAc,CAAC,KAAK,UAAU,KAAK,SAAS,GAAG,KAAK,OAAO,KAAK,uBAAuB,KAAK,uBAAuB,MAAM,KAAK,oBAAoB;AACnL,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,6BAA6B,YAAY;AAAA,EACzC;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,WAAO,KAAK,MAAM;AAChB,UAAI,CAAC,KAAK,UAAU;AAClB,eAAO;AAAA,MACT,OAAO;AACL,iBAAS,oBAAoB,MAAM;AACnC,eAAO,SAAS,QAAQ,UAAU,MAAM,CAAC;AAAA,MAC3C;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,KAAK,eAAe,MAAM;AAC5B,aAAO,CAAC,GAAG,YAAY,KAAK,SAAS;AAAA,IACvC;AACA,UAAM,SAAS,OAAO,KAAK,WAAW;AACtC,QAAI,OAAO,WAAW,WAAW,SAAS,GAAG;AAC3C,YAAM,IAAI,WAAW,oBAAoB,KAAK,mDAAmD,YAAY;AAAA,IAC/G,OAAO;AACL,UAAI,IAAI;AACR,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,cAAM,KAAK,OAAO;AAClB,cAAM,KAAK,WAAW,IAAI;AAC1B,YAAI,MAAM,QAAQ,MAAM,QAAQ,OAAO,IAAI;AACzC,gBAAM,IAAI,WAAW,oBAAoB,KAAK,mDAAmD,YAAY;AAAA,QAC/G,WAAW,MAAM,MAAM;AACrB,iBAAO,KAAK;AAAA,QACd;AACA;AAAA,MACF;AAAA,IACF;AACA,WAAO,CAAC,WAAW,IAAI,GAAG,QAAQ,KAAK,SAAS;AAAA,EAClD;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,UAAI,SAAS,oBAAoB,MAAM;AACvC,UAAI,OAAO,UAAU,SAAS;AAC5B,iBAAS,MAAM,QAAQ,OAAO;AAAA,MAChC;AACA,YAAM,SAAS,QAAQ,KAAK,WAAW,KAAK,GAAG,QAAQ,QAAQ,CAAC,OAAO,IAAI,CAAC,CAAC;AAC7E,aAAO,QAAQ,QAAQ,mBAAmB,KAAK,mBAAmB,OAAO,KAAK,CAAC,CAAC;AAAA,IAClF,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,UAAU,KAAK;AAAA,MACf,WAAW,KAAK;AAAA,MAChB,uBAAuB,qBAAqB,KAAK,qBAAqB;AAAA,MACtE,uBAAuB,qBAAqB,KAAK,qBAAqB;AAAA,MACtE,qBAAqB,qBAAqB,KAAK,mBAAmB;AAAA,MAClE,sBAAsB,oBAAoB,KAAK,oBAAoB;AAAA,MACnE,UAAU,KAAK;AAAA,MACf,aAAa,KAAK;AAAA,IACpB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAG7C,IAAI,QAAQ,cAAc,MAAM;AAAA,EAC9B,YAAY,MAAM;AAChB,UAAM,QAAQ,CAAC,CAAC;AAChB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,cAAc,QAAQ;AACpB,UAAM,IAAI,oBAAoB;AAAA,EAChC;AAAA,EACA,gCAAgC,QAAQ,QAAQ;AAC9C,QAAI,UAAU,QAAQ,UAAU,MAAM;AACpC,aAAO;AAAA,IACT,WAAW,OAAO,SAAS,OAAO,QAAQ;AACxC,aAAO,KAAK,gCAAgC,QAAQ,MAAM;AAAA,IAC5D,WAAW,OAAO,WAAW,GAAG;AAC9B,aAAO;AAAA,IACT;AACA,UAAM,cAAc,OAAO,MAAM,GAAG,OAAO,SAAS,OAAO,MAAM;AACjE,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,YAAM,IAAI,OAAO,OAAO,SAAS,OAAO,SAAS;AACjD,YAAM,IAAI,OAAO;AACjB,UAAI,KAAK,QAAQ,KAAK,QAAQ,IAAI,KAAK,IAAI,GAAG;AAC5C,oBAAY,KAAK,IAAI;AAAA,MACvB,WAAW,MAAM,GAAG;AAClB,oBAAY,KAAK,CAAC;AAAA,MACpB,WAAW,MAAM,GAAG;AAClB,oBAAY,KAAK,CAAC;AAAA,MACpB,OAAO;AACL,YAAI,MAAM,GAAG;AACX,gBAAM,IAAI,WAAW,0DAA0D,KAAK,UAAU,MAAM,IAAI,MAAM,KAAK,UAAU,MAAM,CAAC;AAAA,QACtI;AACA,oBAAY,KAAK,CAAC;AAAA,MACpB;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,YAAY;AAChB,QAAI,MAAM,QAAQ,UAAU,KAAK,CAAC,MAAM,QAAQ,WAAW,EAAE,GAAG;AAC9D,mBAAa,CAAC,mBAAmB,UAAU,CAAC;AAAA,IAC9C;AACA,iBAAa;AACb,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,IAAI,WAAW,wEAAwE,WAAW,kBAAkB;AAAA,IAC5H;AACA,QAAI,aAAa,CAAC;AAClB,eAAW,SAAS,YAAY;AAC9B,UAAI,SAAS,QAAQ,MAAM,OAAO,MAAM;AACtC,mBAAW,KAAK,MAAM,EAAE;AAAA,MAC1B;AAAA,IACF;AACA,iBAAa,QAAQ,UAAU;AAC/B,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,IAAI,WAAW,8EAA8E,KAAK,UAAU,UAAU,IAAI;AAAA,IAClI;AACA,QAAI,cAAc,WAAW,MAAM,OAAO,OAAO,WAAW,GAAG,MAAM,CAAC;AACtE,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,YAAM,QAAQ,WAAW,MAAM,OAAO,OAAO,WAAW,GAAG,MAAM,CAAC;AAClE,oBAAc,KAAK,gCAAgC,aAAa,KAAK;AAAA,IACvE;AACA,UAAM,WAAW,WAAW,IAAI,CAAC,UAAU,MAAM,MAAM;AACvD,QAAI,WAAW,QAAQ,IAAI,MAAM,MAAM,QAAQ,QAAQ,EAAE,WAAW,GAAG;AACrE,WAAK,kBAAkB;AAAA,IACzB,OAAO;AACL,WAAK,kBAAkB;AAAA,IACzB;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS;AACT,UAAI,KAAK,iBAAiB;AACxB,cAAM,iBAAiB,CAAC;AACxB,cAAM,YAAY,OAAO,IAAI,CAAC,WAAW,OAAO,IAAI;AACpD,YAAI,UAAU,QAAQ,IAAI,MAAM,IAAI;AAClC,gBAAM,UAAU,KAAK,SAAS;AAC9B,mBAAS,KAAK,QAAQ;AACpB,kBAAM,QAAQ,EAAE;AAChB,qBAAS,IAAI,GAAG,IAAI,UAAU,OAAO,EAAE,GAAG;AACxC,kBAAI,YAAY,GAAG,CAAC;AAAA,YACtB;AACA,2BAAe,KAAK,CAAC;AAAA,UACvB;AACA,iBAAO,KAAK,cAAc,cAAc;AAAA,QAC1C,OAAO;AACL,cAAI,aAAa;AACjB,qBAAW,KAAK,QAAQ;AACtB,kBAAM,QAAQ,EAAE;AAChB,gBAAI,SAAS,MAAM;AACjB,oBAAM,SAAS,EAAE;AACjB,oBAAM,YAAY,OAAO;AACzB,oBAAM,WAAW,OAAO,MAAM,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC;AACnD,kBAAI,cAAc,QAAQ,GAAG,CAAC,SAAS,EAAE,OAAO,UAAU,OAAO,MAAM,CAAC,CAAC,CAAC,CAAC;AAC3E,4BAAc,UAAU,aAAa,CAAC,GAAG,CAAC,CAAC;AAC3C,4BAAc,QAAQ,aAAa,QAAQ;AAC3C,6BAAe,KAAK,WAAW;AAC/B,2BAAa;AAAA,YACf,WAAW,QAAQ,GAAG;AACpB,oBAAM,OAAO,OAAO,GAAG,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC;AACxC,6BAAe,KAAK,UAAU,GAAG,IAAI,CAAC;AACtC,2BAAa;AAAA,YACf,OAAO;AACL,6BAAe,KAAK,CAAC;AAAA,YACvB;AAAA,UACF;AACA,cAAI,IAAI,KAAK,cAAc,cAAc;AACzC,gBAAM,QAAQ,EAAE;AAChB,cAAI,YAAY;AACd,gBAAI,SAAS,MAAM;AACjB,oBAAM,SAAS,EAAE;AACjB,oBAAM,SAAS,OAAO;AACtB,oBAAM,YAAY,OAAO,SAAS;AAClC,oBAAM,WAAW,CAAC,SAAS,EAAE,OAAO,OAAO,MAAM,GAAG,OAAO,SAAS,CAAC,CAAC;AACtE,kBAAI,QAAQ,UAAU,QAAQ,GAAG,CAAC,IAAI,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,GAAG,QAAQ;AAAA,YACtE,WAAW,QAAQ,GAAG;AACpB,oBAAM,OAAO,CAAC,QAAQ,CAAC,EAAE,OAAO,OAAO,GAAG,QAAQ,CAAC,CAAC;AACpD,kBAAI,UAAU,GAAG,IAAI;AAAA,YACvB;AAAA,UACF;AACA,iBAAO;AAAA,QACT;AAAA,MACF,OAAO;AACL,eAAO,KAAK,cAAc,MAAM;AAAA,MAClC;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa;AACb,QAAI;AACJ,QAAI,WAAW,MAAM,MAAM;AACzB,oBAAc;AAAA,IAChB,OAAO;AACL,oBAAc,WAAW,GAAG,MAAM,CAAC;AAAA,IACrC;AACA,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,YAAM,QAAQ,WAAW,MAAM,OAAO,OAAO,WAAW,GAAG,MAAM,CAAC;AAClE,oBAAc,KAAK,gCAAgC,aAAa,KAAK;AAAA,IACvE;AACA,QAAI,aAAa,CAAC;AAClB,eAAW,SAAS,YAAY;AAC9B,UAAI,SAAS,QAAQ,MAAM,OAAO,MAAM;AACtC,mBAAW,KAAK,MAAM,EAAE;AAAA,MAC1B;AAAA,IACF;AACA,iBAAa,QAAQ,UAAU;AAC/B,QAAI,WAAW,WAAW,GAAG;AAC3B,oBAAc,WAAW,OAAO,WAAW;AAAA,IAC7C,OAAO;AACL,oBAAc,CAAC,IAAI,EAAE,OAAO,WAAW;AAAA,IACzC;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,WAAO,KAAK,MAAM;AAChB,UAAI,QAAQ,MAAM;AAChB,eAAO;AAAA,MACT;AACA,UAAI,CAAC,MAAM,QAAQ,IAAI,GAAG;AACxB,cAAM,IAAI,WAAW,2BAA2B;AAAA,MAClD;AACA,UAAI,CAAC,MAAM,QAAQ,MAAM,GAAG;AAC1B,cAAM,IAAI,WAAW,6BAA6B;AAAA,MACpD;AACA,UAAI,KAAK,WAAW,OAAO,QAAQ;AACjC,cAAM,IAAI,WAAW,mGAAmG,OAAO,aAAa,KAAK,SAAS;AAAA,MAC5J;AACA,UAAI,KAAK,MAAM,CAAC,MAAM,KAAK,IAAI,GAAG;AAChC,eAAO;AAAA,MACT;AACA,aAAO,KAAK,IAAI,CAAC,MAAM,KAAK,OAAO,IAAI,WAAW,GAAG,CAAC,CAAC;AACvD,UAAI,SAAS,KAAK;AAClB,eAAS,IAAI,GAAG,IAAI,KAAK,SAAS,GAAG,EAAE,GAAG;AACxC,iBAAS,WAAW,QAAQ,KAAK,EAAE;AAAA,MACrC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,IAAI,OAAO,cAAc,MAAM;AAAA,EAC7B,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,OAAO,GAAG,MAAM;AAC7B,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,iBAAS,KAAK,QAAQ,OAAO,EAAE;AAAA,MACjC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,KAAK,YAAY;AACjB,sBAAsB,cAAc,IAAI;AACxC,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,OAAO,GAAG,MAAM;AAC7B,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,iBAAS,IAAI,QAAQ,OAAO,EAAE;AAAA,MAChC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAC7C,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,OAAO,GAAG,MAAM;AAC7B,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,iBAAS,KAAK,QAAQ,OAAO,EAAE;AAAA,MACjC;AACA,aAAO,IAAI,IAAI,OAAO,QAAQ,MAAM;AAAA,IACtC,CAAC;AAAA,EACH;AACF;AACA,QAAQ,YAAY;AACpB,sBAAsB,cAAc,OAAO;AAC3C,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,OAAO;AACpB,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,iBAAS,QAAQ,QAAQ,OAAO,EAAE;AAAA,MACpC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,UAAI,SAAS,OAAO;AACpB,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,iBAAS,QAAQ,QAAQ,OAAO,EAAE;AAAA,MACpC;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAC5C,IAAI,cAAc,cAAc,MAAM;AAAA,EACpC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,eAAe;AACpB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,eAAe,KAAK;AACzD,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,MAAM,YAAY;AAChB,QAAI,EAAE,MAAM,QAAQ,UAAU,KAAK,MAAM,QAAQ,WAAW,EAAE,MAAM,WAAW,WAAW,GAAG;AAC3F,YAAM,IAAI,WAAW,uEAAuE;AAAA,IAC9F;AACA,iBAAa;AACb,QAAI,eAAe;AACnB,eAAW,SAAS,YAAY;AAC9B,UAAI,SAAS,MAAM;AACjB,uBAAe;AACf;AAAA,MACF;AAAA,IACF;AACA,QAAI,cAAc;AAChB;AAAA,IACF;AACA,UAAM,WAAW,CAAC;AAClB,aAAS,IAAI,GAAG,IAAI,WAAW,QAAQ,EAAE,GAAG;AAC1C,YAAM,yBAAyB,WAAW,GAAG,MAAM;AACnD,6BAAuB,OAAO,KAAK,MAAM,CAAC;AAC1C,UAAI,SAAS;AACb,iBAAW,SAAS,UAAU;AAC5B,YAAI,aAAa,YAAY,OAAO,sBAAsB,GAAG;AAC3D,mBAAS;AACT;AAAA,QACF;AAAA,MACF;AACA,UAAI,CAAC,QAAQ;AACX,iBAAS,KAAK,sBAAsB;AAAA,MACtC;AAAA,IACF;AACA,QAAI,SAAS,SAAS,GAAG;AACvB,YAAM,IAAI,WAAW,8GAA8G,KAAK,UAAU,UAAU,CAAC;AAAA,IAC/J;AAAA,EACF;AAAA,EACA,cAAc,QAAQ;AACpB,WAAO,KAAK,MAAM;AAChB,aAAO,YAAY,QAAQ,KAAK,IAAI;AAAA,IACtC,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,EAAE,MAAM,QAAQ,UAAU,KAAK,MAAM,QAAQ,WAAW,EAAE,IAAI;AAChE,YAAM,IAAI,WAAW,6DAA6D;AAAA,IACpF;AACA,UAAM,cAAc;AACpB,UAAM,cAAc,YAAY,GAAG,MAAM;AACzC,UAAM,OAAO,KAAK,OAAO,IAAI,YAAY,SAAS,KAAK,OAAO,KAAK;AACnE,eAAW,SAAS,YAAY,MAAM,CAAC,GAAG;AACxC,UAAI,YAAY,SAAS,QAAQ,MAAM,SAAS,MAAM;AACpD,oBAAY,QAAQ;AACpB;AAAA,MACF;AACA,kBAAY,SAAS,MAAM;AAAA,IAC7B;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,QAAI,QAAQ,MAAM;AAChB,aAAO;AAAA,IACT;AACA,QAAI,CAAC,MAAM,QAAQ,IAAI,GAAG;AACxB,YAAM,IAAI,WAAW,2CAA2C;AAAA,IAClE;AACA,QAAI,CAAC,MAAM,QAAQ,MAAM,GAAG;AAC1B,YAAM,IAAI,WAAW,6CAA6C;AAAA,IACpE;AACA,QAAI,KAAK,WAAW,OAAO,QAAQ;AACjC,YAAM,IAAI,WAAW,mCAAmC,KAAK,qCAAqC,OAAO,SAAS;AAAA,IACpH;AACA,WAAO,KAAK,MAAM;AAChB,UAAI,eAAe;AACnB,WAAK,QAAQ,CAAC,MAAM;AAClB,YAAI,KAAK,MAAM;AACb,yBAAe;AACf;AAAA,QACF;AAAA,MACF,CAAC;AACD,UAAI,cAAc;AAChB,eAAO;AAAA,MACT;AACA,YAAM,cAAc,CAAC;AACrB,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,YAAI,KAAK,MAAM,MAAM;AACnB,sBAAY,KAAK,KAAK,SAAS,OAAO,EAAE,GAAG,MAAM,CAAC;AAAA,QACpD,WAAW,KAAK,GAAG,OAAO,OAAO,GAAG,MAAM;AACxC,sBAAY,KAAK,WAAW,KAAK,IAAI,EAAE,CAAC;AAAA,QAC1C,OAAO;AACL,sBAAY,KAAK,KAAK,EAAE;AAAA,QAC1B;AAAA,MACF;AACA,YAAM,oBAAoB,OAAO,aAAa,KAAK,IAAI;AACvD,aAAO,IAAI,mBAAmB,IAAI,KAAK;AAAA,IACzC,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,QAAQ,KAAK;AAAA,IACf;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,YAAY,YAAY;AACxB,sBAAsB,cAAc,WAAW;AAC/C,SAAS,cAAc,MAAM,KAAK;AAChC,SAAO,OAAO,GAAG;AACf,YAAQ;AAAA,EACV;AACA,SAAO;AACT;AACA,SAAS,SAAS,GAAG,GAAG,MAAM;AAC5B,MAAI,EAAE,MAAM,SAAS,KAAK,EAAE,MAAM,SAAS,GAAG;AAC5C,UAAM,IAAI,oBAAoB,kEAAkE;AAAA,EAClG;AACA,eAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,uDAAuD,EAAE,MAAM,QAAQ;AACtH,eAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,uDAAuD,EAAE,MAAM,QAAQ;AACtH,MAAI,OAAO,SAAS,UAAU;AAC5B,WAAO,CAAC,MAAM,IAAI;AAAA,EACpB;AACA,MAAI,EAAE,UAAU,eAAe,EAAE,UAAU,aAAa;AACtD,UAAM,IAAI,oBAAoB,6DAA6D;AAAA,EAC7F;AACA,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,EAAE,MAAM;AACtB,MAAI,QAAQ,MAAM;AAChB,WAAO,CAAC,QAAQ,GAAG,QAAQ,CAAC;AAAA,EAC9B;AACA,QAAM,YAAY;AAClB,SAAO,KAAK,MAAM;AAChB,QAAI;AACJ,QAAI,QAAQ,OAAO;AACjB,aAAO,QAAQ;AACf,YAAM,YAAY,CAAC;AACnB,eAAS,IAAI,GAAG,IAAI,MAAM,EAAE,GAAG;AAC7B,kBAAU,KAAK,CAAC;AAAA,MAClB;AACA,UAAI,QAAQ,GAAG,EAAE,MAAM,OAAO,SAAS,CAAC;AAAA,IAC1C,WAAW,QAAQ,OAAO;AACxB,aAAO,QAAQ;AACf,YAAM,YAAY,CAAC;AACnB,eAAS,IAAI,GAAG,IAAI,MAAM,EAAE,GAAG;AAC7B,kBAAU,KAAK,CAAC;AAAA,MAClB;AACA,UAAI,QAAQ,GAAG,EAAE,MAAM,OAAO,SAAS,CAAC;AAAA,IAC1C,OAAO;AACL,aAAO;AAAA,IACT;AACA,QAAI;AACJ,QAAI,EAAE,MAAM,WAAW,KAAK,EAAE,MAAM,WAAW,GAAG;AAChD,UAAI,UAAU,OAAO,UAAU,IAAI;AACjC,cAAM,KAAK,IAAI,GAAG,CAAC,GAAG,UAAU,EAAE;AAAA,MACpC,OAAO;AACL,cAAM,KAAK,IAAI,UAAU,GAAG,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,GAAG,UAAU,EAAE;AAAA,MACvD;AAAA,IACF,OAAO;AACL,YAAM,OAAO,UAAU,OAAO,EAAE,MAAM,SAAS;AAC/C,YAAM,OAAO,UAAU,OAAO,EAAE,MAAM,SAAS;AAC/C,YAAM,OAAO,GAAG,GAAG,MAAM,IAAI;AAAA,IAC/B;AACA,QAAI,OAAO,GAAG;AACZ,UAAI;AACJ,UAAI,QAAQ,OAAO;AACjB,cAAM,QAAQ,QAAQ;AAAA,MACxB,OAAO;AACL,cAAM,QAAQ;AAAA,MAChB;AACA,YAAM,cAAc,CAAC;AACrB,eAAS,IAAI,KAAK,IAAI,MAAM,MAAM,EAAE,GAAG;AACrC,oBAAY,KAAK,CAAC;AAAA,MACpB;AACA,YAAM,QAAQ,KAAK,WAAW;AAAA,IAChC;AACA,QAAI,IAAI,MAAM,WAAW,GAAG;AAC1B,YAAM,WAAW,KAAK,CAAC;AAAA,IACzB;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,MAAM,cAAc,MAAM;AAAA,EAC5B,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,OAAO,KAAK;AACjB,SAAK,YAAY,KAAK,aAAa,OAAO,QAAQ,KAAK;AACvD,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,OAAO,MAAM,QAAQ,UAAU,KAAK,WAAW,WAAW,KAAK,MAAM,QAAQ,WAAW,EAAE,KAAK,MAAM,QAAQ,WAAW,EAAE,GAAG,MAAM,+DAA+D;AAC/M,UAAM,SAAS,WAAW;AAC1B,UAAM,SAAS,WAAW;AAC1B,QAAI,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG;AAC1C,YAAM,IAAI,oBAAoB,8DAA8D;AAAA,IAC9F;AACA,UAAM,OAAO,KAAK,cAAc,QAAQ,MAAM;AAC9C,QAAI,OAAO,KAAK,QAAQ,OAAO,KAAK,KAAK;AACvC,YAAM,IAAI,WAAW,8BAA8B,OAAO,KAAK,WAAW,OAAO,KAAK,KAAK;AAAA,IAC7F;AAAA,EACF;AAAA,EACA,cAAc,QAAQ;AACpB,QAAI,OAAO,WAAW,GAAG;AACvB,YAAM,IAAI,WAAW,oEAAoE,OAAO,kBAAkB;AAAA,IACpH;AACA,QAAI,KAAK,OAAO;AAChB,QAAI,KAAK,OAAO;AAChB,QAAI;AACJ,QAAI,CAAC,MAAM,QAAQ,KAAK,IAAI,GAAG;AAC7B,aAAO;AAAA,QACL,cAAc,KAAK,MAAM,GAAG,MAAM,MAAM;AAAA,QACxC,cAAc,KAAK,MAAM,GAAG,MAAM,MAAM;AAAA,MAC1C;AAAA,IACF,OAAO;AACL,aAAO,KAAK,KAAK,IAAI,CAAC,MAAM,MAAM,cAAc,MAAM,OAAO,GAAG,MAAM,MAAM,CAAC;AAAA,IAC/E;AACA,QAAI,KAAK,WAAW;AAClB,WAAK,YAAY,IAAI,KAAK,EAAE;AAC5B,WAAK,YAAY,IAAI,KAAK,EAAE;AAAA,IAC9B;AACA,WAAO,SAAS,IAAI,IAAI,IAAI;AAAA,EAC9B;AAAA,EACA,cAAc,QAAQ,QAAQ;AAC5B,QAAI;AACJ,QAAI,CAAC,MAAM,QAAQ,KAAK,IAAI,GAAG;AAC7B,aAAO;AAAA,QACL,cAAc,KAAK,MAAM,OAAO,MAAM;AAAA,QACtC,cAAc,KAAK,MAAM,OAAO,MAAM;AAAA,MACxC;AAAA,IACF,OAAO;AACL,aAAO,KAAK;AAAA,IACd;AACA,WAAO;AAAA,EACT;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,OAAO,MAAM,QAAQ,UAAU,KAAK,WAAW,WAAW,KAAK,MAAM,QAAQ,WAAW,EAAE,KAAK,MAAM,QAAQ,WAAW,EAAE,GAAG,MAAM,+DAA+D;AAC/M,UAAM,SAAS,WAAW,GAAG,MAAM;AACnC,UAAM,SAAS,WAAW,GAAG,MAAM;AACnC,QAAI,OAAO,SAAS,KAAK,OAAO,SAAS,GAAG;AAC1C,YAAM,IAAI,oBAAoB,8DAA8D;AAAA,IAC9F;AACA,UAAM,OAAO,KAAK,cAAc,QAAQ,MAAM;AAC9C,WAAO,OAAO,KAAK,IAAI,CAAC;AACxB,WAAO,OAAO,KAAK,IAAI,CAAC;AACxB,WAAO,OAAO,GAAG,CAAC;AAClB,UAAM,cAAc,OAAO,OAAO,MAAM;AACxC,QAAI,YAAY,WAAW,GAAG;AAC5B,kBAAY,KAAK,CAAC;AAAA,IACpB;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,QAAQ,KAAK;AAAA,MACb,aAAa,KAAK;AAAA,IACpB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,YAAY;AAChB,sBAAsB,cAAc,GAAG;AAGvC,IAAI,gBAAgB,cAAc,MAAM;AAAA,EACtC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,kBAAkB;AACvB,SAAK,SAAS,KAAK;AAAA,EACrB;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS,EAAE,QAAQ,KAAK,OAAO;AACrC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,SAAS,MAAM,KAAK,cAAc,OAAO,OAAO,GAAG,KAAK,MAAM,GAAG,MAAM;AAC7E,YAAM,SAAS,aAAa,QAAQ,MAAM,QAAQ,OAAO,eAAe,KAAK;AAC7E,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AACjD,IAAI,kBAAkB,cAAc,MAAM;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,kBAAkB;AACvB,SAAK,OAAO,KAAK;AAAA,EACnB;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS,EAAE,MAAM,KAAK,KAAK;AACjC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,YAAM,SAAS,oBAAoB,MAAM;AACzC,UAAI,KAAK,OAAO,KAAK,KAAK,OAAO,GAAG;AAClC,cAAM,SAAS,MAAM;AACnB,gBAAM,SAAS,KAAK,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK;AACpD,iBAAO,IAAI,QAAQ,cAAc,OAAO,OAAO,GAAG,MAAM,CAAC;AAAA,QAC3D;AACA,eAAO,aAAa,QAAQ,MAAM,QAAQ,OAAO,eAAe,KAAK;AAAA,MACvE;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,IAAI,eAAe,cAAc,MAAM;AAAA,EACrC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,kBAAkB;AACvB,SAAK,OAAO,KAAK;AACjB,SAAK,aAAa,KAAK;AAAA,EACzB;AAAA,EACA,eAAe,QAAQ;AACrB,WAAO,KAAK,cAAc,oBAAoB,MAAM,EAAE;AAAA,EACxD;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO;AAAA,EACT;AAAA,EACA,YAAY;AACV,UAAM,aAAa,MAAM,UAAU;AACnC,UAAM,SAAS,EAAE,MAAM,KAAK,KAAK;AACjC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,UAAI,KAAK,OAAO,KAAK,KAAK,OAAO,GAAG;AAClC,cAAM,aAAa,KAAK,eAAe,MAAM;AAC7C,cAAM,gBAAgB,MAAM;AAC1B,gBAAM,SAAS,oBAAoB,MAAM;AACzC,gBAAM,QAAQ;AACd,gBAAMF,UAAS;AACf,gBAAM,SAAS,CAAC,QAAQA;AACxB,cAAI,UAAU,aAAa,cAAc,UAAU,GAAG,KAAK,IAAI;AAC/D,oBAAU,MAAM,SAAS,SAAS;AAClC,gBAAM,MAAM,IAAI,KAAK,SAAS,IAAI,KAAK,OAAO,UAAU,OAAO;AAC/D,gBAAM,IAAI,CAAC,IAAI,SAAS,KAAK;AAC7B,gBAAM,IAAI,KAAK,IAAI,QAAQ,OAAO,GAAG,IAAI,KAAK,SAAS,EAAE,GAAG,MAAM,CAAC;AACnE,iBAAO,KAAK,IAAI,GAAG,CAAC,GAAG,CAAC;AAAA,QAC1B;AACA,eAAO,aAAa,eAAe,MAAM,oBAAoB,MAAM,GAAG,OAAO,eAAe,KAAK;AAAA,MACnG;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAGhD,SAAS,mBAAmB,GAAG,OAAO,UAAU,MAAM,OAAOD,YAAW,MAAM;AAC5E,MAAI;AACJ,MAAI,EAAE,SAAS,GAAG;AAChB,UAAM,YAAY,GAAG,OAAO,UAAU,MAAM,OAAOA,SAAQ;AAAA,EAC7D,WAAW,EAAE,SAAS,GAAG;AACvB,UAAM,YAAY,GAAG,OAAO,UAAU,MAAM,OAAOA,SAAQ;AAAA,EAC7D,WAAW,EAAE,SAAS,GAAG;AACvB,UAAM,YAAY,GAAG,OAAO,UAAU,MAAM,OAAOA,SAAQ;AAAA,EAC7D,OAAO;AACL,UAAM,IAAI,oBAAoB,2DAA2D,EAAE,UAAU;AAAA,EACvG;AACA,SAAO;AACT;AACA,SAAS,gCAAgC,GAAG,OAAO,MAAM,eAAeA,YAAW,MAAM;AACvF,SAAO,KAAK,MAAM;AAChB,UAAM,kBAAkB,QAAQ,GAAG,aAAa;AAChD,UAAM,QAAQ,gBAAgB;AAC9B,UAAM,WAAW,gBAAgB;AACjC,UAAM,SAAS,mBAAmB,GAAG,OAAO,UAAU,MAAM,OAAOA,SAAQ;AAC3E,WAAO,CAAC,QAAQ,OAAO,QAAQ;AAAA,EACjC,CAAC;AACH;AACA,SAAS,kCAAkC,GAAG,OAAO,MAAM,eAAeA,YAAW,MAAM;AACzF,SAAO,KAAK,MAAM;AAChB,UAAM,kBAAkB,QAAQ,GAAG,aAAa;AAChD,UAAM,QAAQ,gBAAgB;AAC9B,UAAM,WAAW,gBAAgB;AACjC,UAAM,cAAc,CAAC;AACrB,eAAW,QAAQ,OAAO,GAAG,EAAE,IAAI,GAAG;AACpC,UAAI,cAAc,QAAQ,IAAI,MAAM,IAAI;AACtC,oBAAY,KAAK,CAAC;AAAA,MACpB,OAAO;AACL,oBAAY,KAAK,EAAE,MAAM,KAAK;AAAA,MAChC;AAAA,IACF;AACA,UAAM,gBAAgB,QAAQ,OAAO,WAAW;AAChD,UAAM,oBAAoB,QAAQ,UAAU,WAAW;AACvD,UAAM,iBAAiB,SAAS,OAAO,OAAO,QAAQ,OAAO,WAAW;AACxE,UAAM,gBAAgB,QAAQ,OAAO,OAAO,QAAQ,MAAM,WAAW;AACrE,UAAM,SAAS,mBAAmB,GAAG,eAAe,mBAAmB,eAAe,gBAAgBA,SAAQ;AAC9G,WAAO,CAAC,QAAQ,OAAO,QAAQ;AAAA,EACjC,CAAC;AACH;AACA,SAAS,yBAAyB,GAAG,OAAO,MAAM,eAAeA,YAAW,MAAM;AAChF,MAAI,aAAa,YAAY,cAAc,MAAM,EAAE,KAAK,GAAG,OAAO,GAAG,EAAE,OAAO,CAAC,CAAC,GAAG;AACjF,WAAO,gCAAgC,GAAG,OAAO,MAAM,eAAeA,SAAQ;AAAA,EAChF,OAAO;AACL,WAAO,kCAAkC,GAAG,OAAO,MAAM,eAAeA,SAAQ;AAAA,EAClF;AACF;AACA,IAAI,qBAAqB,cAAc,MAAM;AAAA,EAC3C,YAAY,MAAM;AAChB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,IAAI;AACV,SAAK,kBAAkB;AACvB,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,KAAK;AAC1C,SAAK,WAAW,KAAK,YAAY,OAAO,OAAO,KAAK;AACpD,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,SAAS,KAAK,UAAU,OAAO,OAAO,KAAK;AAChD,SAAK,QAAQ,KAAK,SAAS,OAAO,OAAO,KAAK;AAC9C,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,OAAO;AACrE,SAAK,mBAAmB,eAAe,KAAK,oBAAoB,MAAM;AACtE,SAAK,wBAAwB,eAAe,KAAK,yBAAyB,OAAO;AACjF,SAAK,4BAA4B,eAAe,KAAK,6BAA6B,MAAM;AACxF,SAAK,iBAAiB,cAAc,KAAK,cAAc;AACvD,SAAK,kBAAkB,cAAc,KAAK,eAAe;AACzD,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,mBAAmB,eAAe,KAAK,gBAAgB;AAAA,EAC9D;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,OAAO,KAAK,QAAQ,IAAI,KAAK,OAAO,KAAK,OAAO,WAAW;AACjE,UAAM,MAAM,WAAW;AACvB,QAAI,OAAO,MAAM;AACf,YAAM,IAAI,WAAW,QAAQ,mGAAmG,KAAK,UAAU,UAAU,IAAI;AAAA,IAC/J;AACA,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,WAAW,QAAQ,MAAM,EAAE,CAAC,OAAO,IAAI,EAAE,CAAC,CAAC;AACnF,UAAM,QAAQ,CAAC,GAAG;AAClB,QAAI,KAAK,OAAO;AACd,WAAK,QAAQ,KAAK,UAAU,SAAS,OAAO,MAAM,KAAK,kBAAkB,KAAK,kBAAkB,MAAM,KAAK,eAAe;AAAA,IAC5H;AACA,QAAI,KAAK,QAAQ;AACf,WAAK,OAAO,KAAK,UAAU,QAAQ,OAAO,MAAM,KAAK,iBAAiB,KAAK,iBAAiB,MAAM,KAAK,cAAc;AAAA,IACvH;AACA,SAAK,aAAa,KAAK,UAAU,eAAe,OAAO,MAAM,KAAK,uBAAuB,MAAM,KAAK;AACpG,SAAK,iBAAiB,KAAK,UAAU,mBAAmB,OAAO,MAAM,KAAK,2BAA2B,MAAM,KAAK;AAChH,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,WAAW,OAAO,eAAe,OAAO,QAAQ,OAAO;AAC7D,YAAM,SAAS,oBAAoB,MAAM;AACzC,YAAM,aAAa,OAAO;AAC1B,YAAM,OAAO,WAAW;AACxB,YAAM,gBAAgB,OAAO,GAAG,IAAI;AACpC,YAAM,OAAO,KAAK,QAAQ,IAAI,KAAK,OAAO,KAAK,OAAO;AACtD,oBAAc,OAAO,MAAM,CAAC;AAC5B,YAAM,iBAAiB,aAAa,GAAG,IAAI;AAC3C,qBAAe,QAAQ,WAAW;AAClC,YAAM,sBAAsB,cAAc,MAAM;AAChD,0BAAoB,KAAK;AACzB,YAAM,oBAAoB,CAAC,aAAa,YAAY,qBAAqB,OAAO,GAAG,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,CAAC;AAC3G,YAAM,qBAAqB,MAAM;AAC/B,YAAI,mBAAmB;AACrB,gBAAM,sBAAsB,QAAQ,KAAK,WAAW,KAAK,GAAG,cAAc;AAC1E,gBAAM,0BAA0B,QAAQ,KAAK,eAAe,KAAK,GAAG,cAAc;AAClF,gBAAM,gBAAgB,KAAK,SAAS,QAAQ,KAAK,KAAK,KAAK,GAAG,cAAc,IAAI;AAChF,gBAAM,iBAAiB,KAAK,QAAQ,QAAQ,KAAK,MAAM,KAAK,GAAG,cAAc,IAAI;AACjF,iBAAO,mBAAmB,QAAQ,qBAAqB,yBAAyB,eAAe,gBAAgB,KAAK,OAAO;AAAA,QAC7H,OAAO;AACL,iBAAO,mBAAmB,QAAQ,KAAK,WAAW,KAAK,GAAG,KAAK,eAAe,KAAK,GAAG,KAAK,QAAQ,OAAO,OAAO,KAAK,KAAK,KAAK,GAAG,KAAK,SAAS,OAAO,OAAO,KAAK,MAAM,KAAK,GAAG,KAAK,OAAO;AAAA,QAChM;AAAA,MACF;AACA,UAAI,CAAC,UAAU;AACb,eAAO,mBAAmB;AAAA,MAC5B;AACA,YAAM,CAAC,gBAAgB,OAAO,QAAQ,IAAI,yBAAyB,QAAQ,KAAK,MAAM,KAAK,GAAG,KAAK,KAAK,KAAK,GAAG,eAAe,KAAK,OAAO;AAC3I,YAAM,kBAAkB,CAAC,WAAW,OAAO,aAAa;AACtD,aAAK,MAAM;AACT,gBAAM,QAAQ,IAAI;AAClB,gBAAM,YAAY,UAAU,KAAK;AACjC,gBAAM,cAAc,IAAI,IAAI,WAAW,KAAK,GAAG,KAAK;AACpD,oBAAU,MAAM,IAAI,WAAW,WAAW,CAAC;AAAA,QAC7C,CAAC;AAAA,MACH;AACA,YAAM,8BAA8B,MAAM;AACxC,wBAAgB,KAAK,YAAY,OAAO,KAAK,QAAQ;AACrD,wBAAgB,KAAK,gBAAgB,UAAU,KAAK,QAAQ;AAAA,MAC9D;AACA,kCAA4B;AAC5B,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,MAAM,KAAK;AAAA,MACX,UAAU,KAAK;AAAA,MACf,SAAS,KAAK;AAAA,MACd,QAAQ,KAAK;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,MAC5D,uBAAuB,qBAAqB,KAAK,qBAAqB;AAAA,MACtE,2BAA2B,qBAAqB,KAAK,yBAAyB;AAAA,MAC9E,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,MAC5D,gBAAgB,oBAAoB,KAAK,cAAc;AAAA,MACvD,iBAAiB,oBAAoB,KAAK,eAAe;AAAA,IAC3D;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,mBAAmB,YAAY;AAC/B,sBAAsB,cAAc,kBAAkB;AACtD,IAAI,qBAAqB,cAAc,MAAM;AAAA,EAC3C,YAAY,MAAM;AAChB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,IAAI;AACV,SAAK,OAAO,KAAK,QAAQ,OAAO,KAAK,KAAK;AAC1C,QAAI,OAAO,KAAK,SAAS,UAAU;AACjC,UAAI,CAAC,OAAO,UAAU,KAAK,IAAI,GAAG;AAChC,cAAM,IAAI,MAAM,gDAAgD,KAAK,MAAM;AAAA,MAC7E;AAAA,IACF,WAAW,MAAM,QAAQ,KAAK,IAAI,GAAG;AACnC,iBAAW,QAAQ,KAAK,MAAM;AAC5B,YAAI,CAAC,OAAO,UAAU,IAAI,GAAG;AAC3B,gBAAM,IAAI,MAAM,0DAA0D,KAAK,UAAU,KAAK,IAAI,GAAG;AAAA,QACvG;AAAA,MACF;AAAA,IACF,OAAO;AACL,YAAM,IAAI,MAAM,wEAAwE,KAAK,UAAU,KAAK,IAAI,GAAG;AAAA,IACrH;AACA,SAAK,UAAU,KAAK,WAAW,OAAO,OAAO,KAAK;AAClD,SAAK,SAAS,KAAK,UAAU,OAAO,OAAO,KAAK;AAChD,SAAK,QAAQ,KAAK,SAAS,OAAO,OAAO,KAAK;AAC9C,SAAK,kBAAkB,eAAe,KAAK,mBAAmB,OAAO;AACrE,SAAK,mBAAmB,eAAe,KAAK,oBAAoB,MAAM;AACtE,SAAK,kBAAkB,eAAe,KAAK,eAAe;AAC1D,SAAK,mBAAmB,eAAe,KAAK,gBAAgB;AAC5D,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,QAAQ,WAAW;AACzB,QAAI,OAAO,KAAK,SAAS,UAAU;AACjC,WAAK,OAAO,CAAC,KAAK,IAAI;AAAA,IACxB;AACA,aAAS,IAAI,GAAG,IAAI,KAAK,KAAK,QAAQ,EAAE,GAAG;AACzC,UAAI,KAAK,KAAK,KAAK,GAAG;AACpB,aAAK,KAAK,MAAM;AAAA,MAClB;AAAA,IACF;AACA,eAAW,QAAQ,KAAK,MAAM;AAC5B,UAAI,OAAO,KAAK,QAAQ,OAAO;AAC7B,cAAM,IAAI,MAAM,iBAAiB,MAAM;AAAA,MACzC;AAAA,IACF;AACA,QAAI,KAAK,KAAK,WAAW,QAAQ,KAAK,IAAI,EAAE,QAAQ;AAClD,YAAM,IAAI,MAAM,4BAA4B,KAAK,MAAM;AAAA,IACzD;AACA,UAAM,aAAa,KAAK,KAAK,IAAI,CAAC,SAAS,WAAW,KAAK;AAC3D,UAAM,YAAY;AAClB,QAAI,KAAK,OAAO;AACd,WAAK,QAAQ,KAAK,UAAU,SAAS,YAAY,WAAW,KAAK,kBAAkB,KAAK,kBAAkB,SAAS;AAAA,IACrH,OAAO;AACL,WAAK,QAAQ;AAAA,IACf;AACA,QAAI,KAAK,QAAQ;AACf,WAAK,OAAO,KAAK,UAAU,QAAQ,YAAY,WAAW,KAAK,iBAAiB,KAAK,iBAAiB,SAAS;AAAA,IACjH,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AACA,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,SAAS,oBAAoB,MAAM;AACzC,UAAM,aAAa,OAAO;AAC1B,UAAM,QAAQ,WAAW;AACzB,WAAO,KAAK,MAAM;AAChB,YAAM,WAAW;AACjB,UAAI,EAAE,MAAM,OAAO,SAAS,IAAI,QAAQ,QAAQ,KAAK,MAAM,QAAQ;AACnE,YAAM,iBAAiB,aAAa,GAAG,KAAK;AAC5C,iBAAW,OAAO,KAAK,MAAM;AAC3B,uBAAe,OAAO,WAAW;AAAA,MACnC;AACA,YAAM,YAAY,CAAC,MAAM;AACvB,YAAI,KAAK,QAAQ,EAAE,MAAM,WAAW,OAAO;AACzC,iBAAO,QAAQ,GAAG,cAAc;AAAA,QAClC,OAAO;AACL,iBAAO;AAAA,QACT;AAAA,MACF;AACA,UAAIC,UAAS,KAAK,QAAQ,UAAU,KAAK,MAAM,KAAK,CAAC,IAAI;AACzD,UAAI,SAAS,KAAK,SAAS,UAAU,KAAK,KAAK,KAAK,CAAC,IAAI;AACzD,YAAM,gBAAgB,CAAC;AACvB,YAAM,oBAAoB,CAAC;AAC3B,eAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,YAAI,KAAK,KAAK,QAAQ,CAAC,MAAM,IAAI;AAC/B,wBAAc,KAAK,WAAW,EAAE;AAChC,4BAAkB,KAAK,CAAC;AAAA,QAC1B,OAAO;AACL,wBAAc,KAAK,CAAC;AACpB,4BAAkB,KAAK,WAAW,EAAE;AAAA,QACtC;AAAA,MACF;AACA,cAAQ,KAAK,OAAO,aAAa;AACjC,iBAAW,KAAK,UAAU,aAAa;AACvC,UAAIA,WAAU,MAAM;AAClB,QAAAA,UAAS,KAAKA,SAAQ,iBAAiB;AAAA,MACzC;AACA,UAAI,UAAU,MAAM;AAClB,iBAAS,KAAK,QAAQ,iBAAiB;AAAA,MACzC;AACA,aAAO,mBAAmB,QAAQ,OAAO,UAAU,QAAQA,SAAQ,KAAK,OAAO;AAAA,IACjF,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,MAAM,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,MACd,QAAQ,KAAK;AAAA,MACb,OAAO,KAAK;AAAA,MACZ,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,MAC5D,iBAAiB,qBAAqB,KAAK,eAAe;AAAA,MAC1D,kBAAkB,qBAAqB,KAAK,gBAAgB;AAAA,IAC9D;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,mBAAmB,YAAY;AAC/B,sBAAsB,cAAc,kBAAkB;AAGtD,SAAS,iBAAiB,GAAG,SAAS,YAAY;AAChD,SAAO,KAAK,MAAM;AAChB,QAAI,EAAE,SAAS,GAAG;AAChB,YAAM,IAAI,WAAW,kEAAkE,EAAE,gBAAgB;AAAA,IAC3G;AACA,QAAI,WAAW,MAAM;AACnB,gBAAU,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,IAC3B;AACA,QAAI,QAAQ,WAAW,KAAK,QAAQ,GAAG,WAAW,KAAK,QAAQ,GAAG,WAAW,GAAG;AAC9E,YAAM,IAAI,WAAW,6GAA6G;AAAA,IACpI;AACA,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,QAAI,eAAe,kBAAkB,eAAe,iBAAiB;AACnE,YAAM,IAAI,WAAW,wBAAwB,2EAA2E;AAAA,IAC1H;AACA,QAAI;AACJ,QAAI,eAAe,iBAAiB;AAClC,gBAAU,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,QAAQ,IAAI,QAAQ,EAAE;AAAA,IACnD,OAAO;AACL,gBAAU,CAAC,CAAC,GAAG,CAAC,GAAG,QAAQ,IAAI,QAAQ,IAAI,CAAC,GAAG,CAAC,CAAC;AAAA,IACnD;AACA,WAAO,IAAI,GAAG,OAAO;AAAA,EACvB,CAAC;AACH;AACA,IAAI,gBAAgB,cAAc,MAAM;AAAA,EACtC,YAAY,MAAM;AAChB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,IAAI;AACV,SAAK,aAAa,KAAK,cAAc,OAAO,gBAAgB,IAAI,KAAK;AACrE,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,UAAU,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,IAChC,WAAW,OAAO,KAAK,YAAY,UAAU;AAC3C,WAAK,UAAU,CAAC,CAAC,KAAK,SAAS,KAAK,OAAO,GAAG,CAAC,KAAK,SAAS,KAAK,OAAO,CAAC;AAAA,IAC5E,OAAO;AACL,WAAK,UAAU,KAAK;AACpB,UAAI,KAAK,QAAQ,WAAW,GAAG;AAC7B,cAAM,IAAI,WAAW,+EAA+E,KAAK,QAAQ,eAAe;AAAA,MAClI;AACA,UAAI;AACJ,UAAI;AACJ,UAAI,OAAO,KAAK,QAAQ,OAAO,UAAU;AACvC,wBAAgB,CAAC,KAAK,QAAQ,IAAI,KAAK,QAAQ,EAAE;AACjD,uBAAe,CAAC,KAAK,QAAQ,IAAI,KAAK,QAAQ,EAAE;AAAA,MAClD,OAAO;AACL,aAAK,UAAU,KAAK;AACpB,YAAI,KAAK,QAAQ,GAAG,WAAW,GAAG;AAChC,gBAAM,IAAI,WAAW,sFAAsF,KAAK,QAAQ,GAAG,eAAe;AAAA,QAC5I;AACA,wBAAgB,KAAK,QAAQ;AAC7B,YAAI,KAAK,QAAQ,GAAG,WAAW,GAAG;AAChC,gBAAM,IAAI,WAAW,qFAAqF,KAAK,QAAQ,GAAG,eAAe;AAAA,QAC3I;AACA,uBAAe,KAAK,QAAQ;AAAA,MAC9B;AACA,WAAK,UAAU,CAAC,eAAe,YAAY;AAAA,IAC7C;AACA,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,QAAI;AACJ,QAAI;AACJ,QAAI,KAAK,eAAe,iBAAiB;AACvC,UAAI,WAAW,MAAM,QAAQ,WAAW,MAAM,GAAG;AAC/C,eAAO,WAAW,KAAK,KAAK,QAAQ,GAAG,KAAK,KAAK,QAAQ,GAAG;AAAA,MAC9D,OAAO;AACL,eAAO;AAAA,MACT;AACA,UAAI,WAAW,MAAM,QAAQ,WAAW,MAAM,GAAG;AAC/C,eAAO,WAAW,KAAK,KAAK,QAAQ,GAAG,KAAK,KAAK,QAAQ,GAAG;AAAA,MAC9D,OAAO;AACL,eAAO;AAAA,MACT;AACA,aAAO,CAAC,WAAW,IAAI,WAAW,IAAI,MAAM,IAAI;AAAA,IAClD,OAAO;AACL,UAAI,WAAW,MAAM,QAAQ,WAAW,MAAM,GAAG;AAC/C,eAAO,WAAW,KAAK,KAAK,QAAQ,GAAG,KAAK,KAAK,QAAQ,GAAG;AAAA,MAC9D,OAAO;AACL,eAAO;AAAA,MACT;AACA,UAAI,WAAW,MAAM,QAAQ,WAAW,MAAM,GAAG;AAC/C,eAAO,WAAW,KAAK,KAAK,QAAQ,GAAG,KAAK,KAAK,QAAQ,GAAG;AAAA,MAC9D,OAAO;AACL,eAAO;AAAA,MACT;AACA,aAAO,CAAC,WAAW,IAAI,MAAM,MAAM,WAAW,EAAE;AAAA,IAClD;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM,iBAAiB,oBAAoB,MAAM,GAAG,KAAK,SAAS,KAAK,UAAU,CAAC;AAAA,EAChG;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,IACnB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AAGjD,SAAS,OAAO,GAAG,UAAU,SAAS,SAAS,YAAY,UAAU;AACnE,SAAO,KAAK,MAAM;AAChB,oBAAgB,UAAU;AAC1B,kBAAc,QAAQ;AACtB,qBAAiB,OAAO;AACxB,QAAI,WAAW,MAAM;AACnB,gBAAU,CAAC,GAAG,CAAC;AAAA,IACjB;AACA,QAAI,WAAW,MAAM;AACnB,gBAAU;AAAA,IACZ;AACA,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,QAAI,YAAY,MAAM;AACpB,iBAAW;AAAA,IACb;AACA,QAAI,sBAAsB,GAAG,UAAU;AACvC,QAAI;AACJ,UAAM,gBAAgB,YAAY,SAAS,SAAS;AACpD,QAAI,aAAa,OAAO;AACtB,UAAI,QAAQ,GAAG,UAAU,SAAS,aAAa;AAAA,IACjD,OAAO;AACL,UAAI;AAAA,QACF;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAC/B;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,SAAS,OAAO,GAAG,UAAU,SAAS,SAAS,YAAY,UAAU;AACnE,SAAO,KAAK,MAAM;AAChB,oBAAgB,UAAU;AAC1B,kBAAc,QAAQ;AACtB,qBAAiB,OAAO;AACxB,QAAI,WAAW,MAAM;AACnB,gBAAU,CAAC,GAAG,GAAG,CAAC;AAAA,IACpB;AACA,QAAI,WAAW,MAAM;AACnB,gBAAU;AAAA,IACZ;AACA,QAAI,cAAc,MAAM;AACtB,mBAAa,gBAAgB;AAAA,IAC/B;AACA,QAAI,YAAY,MAAM;AACpB,iBAAW;AAAA,IACb;AACA,QAAI,sBAAsB,GAAG,UAAU;AACvC,QAAI;AACJ,UAAM,gBAAgB,YAAY,SAAS,SAAS;AACpD,QAAI,aAAa,OAAO;AACtB,UAAI,UAAU,GAAG,UAAU,SAAS,aAAa;AAAA,IACnD,OAAO;AACL,UAAI,UAAU,GAAG,UAAU,SAAS,aAAa;AAAA,IACnD;AACA,QAAI,eAAe,iBAAiB;AAClC,UAAI,UAAU,GAAG,CAAC,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAAA,IAClC;AACA,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,QAAI,KAAK,YAAY,MAAM;AACzB,WAAK,WAAW;AAAA,IAClB;AACA,UAAM,IAAI;AACV,QAAI,OAAO,KAAK,aAAa,UAAU;AACrC,WAAK,WAAW,CAAC,KAAK,QAAQ;AAAA,IAChC,WAAW,MAAM,QAAQ,KAAK,QAAQ,KAAK,KAAK,SAAS,WAAW,KAAK,OAAO,KAAK,SAAS,OAAO,UAAU;AAC7G,WAAK,WAAW,KAAK;AAAA,IACvB,OAAO;AACL,YAAM,IAAI,WAAW,qGAAqG,KAAK,UAAU,KAAK,QAAQ,GAAG;AAAA,IAC3J;AACA,0BAAsB,KAAK,UAAU,UAAU;AAC/C,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,UAAU,KAAK;AAAA,IACtB,OAAO;AACL,UAAI,OAAO,KAAK,YAAY,UAAU;AACpC,aAAK,UAAU,CAAC,KAAK,OAAO;AAAA,MAC9B,WAAW,MAAM,QAAQ,KAAK,OAAO,KAAK,KAAK,QAAQ,WAAW,KAAK,OAAO,KAAK,QAAQ,OAAO,UAAU;AAC1G,aAAK,UAAU,KAAK;AAAA,MACtB,OAAO;AACL,cAAM,IAAI,WAAW,oGAAoG,KAAK,UAAU,KAAK,OAAO,GAAG;AAAA,MACzJ;AAAA,IACF;AACA,0BAAsB,KAAK,SAAS,SAAS;AAC7C,SAAK,UAAU,KAAK,WAAW,OAAO,UAAU,KAAK;AACrD,qBAAiB,KAAK,OAAO;AAC7B,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,SAAS,iBAAiB,WAAW,IAAI,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AAC9F,WAAO,CAAC,WAAW,IAAI,QAAQ,WAAW,EAAE;AAAA,EAC9C;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,eAAS,YAAY,oBAAoB,MAAM,GAAG,CAAC;AACnD,YAAM,SAAS,KAAK,gBAAgB,oBAAoB,MAAM,GAAG,CAAC,KAAK,SAAS,IAAI,CAAC,GAAG,CAAC,KAAK,QAAQ,IAAI,CAAC,GAAG,KAAK,SAAS,cAAc;AAC1I,aAAO,QAAQ,QAAQ,CAAC,CAAC,CAAC;AAAA,IAC5B,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,UAAU,KAAK;AAAA,MACf,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,IAChB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,eAAe,cAAc,UAAU;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,mBAAmB,cAAc,UAAU;AAAA,EAC7C,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,iBAAiB,YAAY;AAC7B,sBAAsB,cAAc,gBAAgB;AACpD,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,QAAI,KAAK,YAAY,MAAM;AACzB,WAAK,WAAW,CAAC,GAAG,CAAC;AAAA,IACvB;AACA,UAAM,IAAI;AACV,SAAK,WAAW,MAAM,QAAQ,KAAK,QAAQ,IAAI,KAAK,WAAW,CAAC,KAAK,UAAU,KAAK,QAAQ;AAC5F,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,UAAU,KAAK;AAAA,IACtB,WAAW,MAAM,QAAQ,KAAK,OAAO,GAAG;AACtC,UAAI,KAAK,QAAQ,WAAW,GAAG;AAC7B,cAAM,IAAI,WAAW,wHAAwH,KAAK,QAAQ,SAAS;AAAA,MACrK;AACA,WAAK,UAAU,KAAK;AAAA,IACtB,OAAO;AACL,WAAK,UAAU,CAAC,KAAK,SAAS,KAAK,OAAO;AAAA,IAC5C;AACA,0BAAsB,KAAK,UAAU,UAAU;AAC/C,0BAAsB,KAAK,SAAS,SAAS;AAC7C,SAAK,UAAU,KAAK,WAAW,OAAO,UAAU,KAAK;AACrD,SAAK,aAAa,KAAK,cAAc,OAAO,iBAAiB,KAAK;AAClE,oBAAgB,KAAK,UAAU;AAC/B,qBAAiB,KAAK,OAAO;AAC7B,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC5E,QAAI,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC5E,WAAO,iBAAiB,MAAM,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AAC7E,WAAO,iBAAiB,MAAM,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AAC7E,QAAI,KAAK,eAAe,iBAAiB;AACvC,aAAO,CAAC,WAAW,IAAI,WAAW,IAAI,MAAM,IAAI;AAAA,IAClD,OAAO;AACL,aAAO,CAAC,WAAW,IAAI,MAAM,MAAM,WAAW,EAAE;AAAA,IAClD;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,aAAO,KAAK,gBAAgB,oBAAoB,MAAM,GAAG,KAAK,UAAU,KAAK,SAAS,KAAK,SAAS,KAAK,UAAU;AAAA,IACrH,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,UAAU,KAAK;AAAA,MACf,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,IACnB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,eAAe,cAAc,UAAU;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,mBAAmB,cAAc,UAAU;AAAA,EAC7C,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,iBAAiB,YAAY;AAC7B,sBAAsB,cAAc,gBAAgB;AACpD,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,QAAI,KAAK,YAAY,MAAM;AACzB,WAAK,WAAW,CAAC,GAAG,GAAG,CAAC;AAAA,IAC1B;AACA,UAAM,IAAI;AACV,SAAK,WAAW,MAAM,QAAQ,KAAK,QAAQ,IAAI,KAAK,WAAW,CAAC,KAAK,UAAU,KAAK,UAAU,KAAK,QAAQ;AAC3G,QAAI,KAAK,WAAW,MAAM;AACxB,WAAK,UAAU,KAAK;AAAA,IACtB,WAAW,MAAM,QAAQ,KAAK,OAAO,GAAG;AACtC,UAAI,KAAK,QAAQ,WAAW,GAAG;AAC7B,cAAM,IAAI,WAAW,wHAAwH,KAAK,QAAQ,SAAS;AAAA,MACrK;AACA,WAAK,UAAU,KAAK;AAAA,IACtB,OAAO;AACL,WAAK,UAAU,CAAC,KAAK,SAAS,KAAK,SAAS,KAAK,OAAO;AAAA,IAC1D;AACA,0BAAsB,KAAK,UAAU,UAAU;AAC/C,0BAAsB,KAAK,SAAS,SAAS;AAC7C,SAAK,UAAU,KAAK,WAAW,OAAO,UAAU,KAAK;AACrD,SAAK,aAAa,KAAK,cAAc,OAAO,iBAAiB,KAAK;AAClE,oBAAgB,KAAK,UAAU;AAC/B,qBAAiB,KAAK,OAAO;AAC7B,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,SAAS,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC9E,QAAI,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC5E,QAAI,OAAO,KAAK,eAAe,kBAAkB,WAAW,KAAK,WAAW;AAC5E,aAAS,iBAAiB,QAAQ,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AACjF,WAAO,iBAAiB,MAAM,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AAC7E,WAAO,iBAAiB,MAAM,KAAK,SAAS,IAAI,KAAK,SAAS,KAAK,QAAQ,EAAE;AAC7E,QAAI,KAAK,eAAe,iBAAiB;AACvC,aAAO,CAAC,WAAW,IAAI,WAAW,IAAI,QAAQ,MAAM,IAAI;AAAA,IAC1D,OAAO;AACL,aAAO,CAAC,WAAW,IAAI,QAAQ,MAAM,MAAM,WAAW,EAAE;AAAA,IAC1D;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,WAAK,eAAe,QAAQ,MAAM;AAClC,aAAO,KAAK,gBAAgB,oBAAoB,MAAM,GAAG,KAAK,UAAU,KAAK,SAAS,KAAK,SAAS,KAAK,UAAU;AAAA,IACrH,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,UAAU,KAAK;AAAA,MACf,SAAS,KAAK;AAAA,MACd,SAAS,KAAK;AAAA,MACd,YAAY,KAAK;AAAA,IACnB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,eAAe,cAAc,UAAU;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,aAAa,YAAY;AACzB,sBAAsB,cAAc,YAAY;AAChD,IAAI,mBAAmB,cAAc,UAAU;AAAA,EAC7C,YAAY,MAAM;AAChB,UAAM,IAAI;AAAA,EACZ;AAAA,EACA,gBAAgB,QAAQ,UAAU,SAAS,SAAS,YAAY;AAC9D,oBAAgB,UAAU;AAC1B,qBAAiB,OAAO;AACxB,WAAO,OAAO,QAAQ,UAAU,SAAS,SAAS,YAAY,KAAK;AAAA,EACrE;AACF;AACA,iBAAiB,YAAY;AAC7B,sBAAsB,cAAc,gBAAgB;AACpD,IAAI,kBAAkB,cAAc,MAAM;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,WAAO,CAAC,WAAW,IAAI,WAAW,EAAE;AAAA,EACtC;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB;AAAA,EAChC;AACF;AACA,IAAI,yBAAyB,cAAc,gBAAgB;AAAA,EACzD,YAAY,MAAM;AAChB,UAAM,QAAQ,CAAC,CAAC;AAAA,EAClB;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,SAAS,oBAAoB,MAAM;AACzC,aAAO,KAAK,QAAQ,CAAC;AAAA,IACvB,CAAC;AAAA,EACH;AACF;AACA,uBAAuB,YAAY;AACnC,sBAAsB,cAAc,sBAAsB;AAC1D,IAAI,qBAAqB,cAAc,gBAAgB;AAAA,EACrD,YAAY,MAAM;AAChB,UAAM,QAAQ,CAAC,CAAC;AAAA,EAClB;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,SAAS,oBAAoB,MAAM;AACzC,aAAO,IAAI,QAAQ,CAAC;AAAA,IACtB,CAAC;AAAA,EACH;AACF;AACA,mBAAmB,YAAY;AAC/B,sBAAsB,cAAc,kBAAkB;AACtD,IAAI,kBAAkB,cAAc,MAAM;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,aAAa,KAAK,cAAc,OAAO,iBAAiB,KAAK;AAClE,oBAAgB,KAAK,UAAU;AAC/B,SAAK,YAAY,CAAC,IAAI,UAAU,EAAE,MAAM,EAAE,CAAC,CAAC;AAAA,EAC9C;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa;AACb,QAAI,KAAK,eAAe,gBAAgB;AACtC,aAAO,CAAC,WAAW,IAAI,WAAW,EAAE;AAAA,IACtC,OAAO;AACL,aAAO,CAAC,WAAW,IAAI,WAAW,EAAE;AAAA,IACtC;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,UAAM,IAAI,oBAAoB;AAAA,EAChC;AAAA,EACA,YAAY;AACV,UAAM,SAAS,EAAE,YAAY,KAAK,WAAW;AAC7C,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AACF;AACA,IAAI,yBAAyB,cAAc,gBAAgB;AAAA,EACzD,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,SAAS,oBAAoB,MAAM;AACzC,UAAI,KAAK,eAAe,gBAAgB;AACtC,eAAO,KAAK,QAAQ,CAAC,GAAG,CAAC,CAAC;AAAA,MAC5B,OAAO;AACL,eAAO,KAAK,QAAQ,CAAC,GAAG,CAAC,CAAC;AAAA,MAC5B;AAAA,IACF,CAAC;AAAA,EACH;AACF;AACA,uBAAuB,YAAY;AACnC,sBAAsB,cAAc,sBAAsB;AAC1D,IAAI,qBAAqB,cAAc,gBAAgB;AAAA,EACrD,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,SAAS,oBAAoB,MAAM;AACzC,UAAI,KAAK,eAAe,gBAAgB;AACtC,eAAO,IAAI,QAAQ,CAAC,GAAG,CAAC,CAAC;AAAA,MAC3B,OAAO;AACL,eAAO,IAAI,QAAQ,CAAC,GAAG,CAAC,CAAC;AAAA,MAC3B;AAAA,IACF,CAAC;AAAA,EACH;AACF;AACA,mBAAmB,YAAY;AAC/B,sBAAsB,cAAc,kBAAkB;AAGtD,IAAI,UAAU,cAAc,MAAM;AAAA,EAChC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,QAAQ,KAAK;AAAA,EACpB;AAAA,EACA,MAAM,YAAY;AAChB,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,IAAI,YAAY;AACd,QAAI,KAAK,SAAS,MAAM;AACtB,aAAO,KAAK,MAAM;AAAA,IACpB,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,IAAI,UAAU,OAAO;AACnB,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM,YAAY;AAAA,IACzB;AAAA,EACF;AAAA,EACA,IAAI,mBAAmB;AACrB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,IAAI,sBAAsB;AACxB,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,IAAI,SAAS;AACX,WAAO,KAAK,MAAM;AAAA,EACpB;AAAA,EACA,aAAa;AACX,WAAO,KAAK,MAAM,WAAW;AAAA,EAC/B;AAAA,EACA,WAAW,SAAS;AAClB,SAAK,MAAM,WAAW,OAAO;AAAA,EAC/B;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,SAAS;AAAA,QACP,aAAa,KAAK,MAAM,aAAa;AAAA,QACrC,UAAU,KAAK,MAAM,UAAU;AAAA,MACjC;AAAA,IACF;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,6BAA6B,OAAO;AAClC,UAAM,6BAA6B,KAAK;AACxC,QAAI,KAAK,SAAS,MAAM;AACtB,WAAK,MAAM,6BAA6B,KAAK;AAAA,IAC/C;AAAA,EACF;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ,gBAAgB,CAAC,GAAG;AACjD,UAAM,cAAc,OAAO;AAC3B,UAAM,QAAQ,YAAY,aAAa,aAAa;AACpD,WAAO,OAAO;AACd,UAAM,YAAY,EAAE,MAAM;AAC1B,WAAO,OAAO,WAAW,MAAM;AAC/B,WAAO,IAAI,IAAI,SAAS;AAAA,EAC1B;AACF;AACA,IAAI,kBAAkB,cAAc,QAAQ;AAAA,EAC1C,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,MAAM,YAAY;AAChB,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,IAAI,WAAW,gFAAgF,KAAK,UAAU,UAAU,GAAG;AAAA,IACnI;AACA,SAAK,YAAY,CAAC,EAAE,OAAO,WAAW,CAAC;AACvC,UAAM,kBAAkB,CAAC,WAAW,EAAE,EAAE,OAAO,WAAW,MAAM,CAAC,CAAC;AAClE,QAAI,CAAC,KAAK,MAAM,OAAO;AACrB,WAAK,MAAM,MAAM,eAAe;AAChC,WAAK,MAAM,QAAQ;AAAA,IACrB;AACA,UAAM,MAAM,UAAU;AAAA,EACxB;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,kBAAkB,CAAC,WAAW,EAAE,EAAE,OAAO,WAAW,MAAM,CAAC,CAAC;AAClE,UAAM,mBAAmB,KAAK,MAAM,mBAAmB,eAAe;AACtE,UAAM,YAAY,WAAW;AAC7B,WAAO,CAAC,iBAAiB,IAAI,SAAS,EAAE,OAAO,iBAAiB,MAAM,CAAC,CAAC;AAAA,EAC1E;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,YAAM,QAAQ,CAAC,SAAS,WAAW;AACjC,cAAM,SAAS,oBAAoB,KAAK,MAAM,KAAK,SAAS,MAAM,CAAC;AACnE,eAAO,CAAC,QAAQ,CAAC,CAAC;AAAA,MACpB;AACA,YAAM,aAAa,IAAI,OAAO,QAAQ,CAAC,GAAG,OAAO,MAAM,MAAM,OAAO,IAAI;AACxE,YAAM,IAAI,WAAW;AACrB,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AACA,gBAAgB,YAAY;AAC5B,sBAAsB,cAAc,eAAe;AACnD,SAAS,4BAA4B,OAAO;AAC1C,4BAA0B,iCAAiC,0BAA0B,KAAK;AAC5F;AACA,IAAI,mCAAmC;AACvC,IAAI,gBAAgB,cAAc,QAAQ;AAAA,EACxC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,UAAM,cAAc,KAAK,MAAM,UAAU;AACzC,UAAM,WAAW,CAAC;AAClB,aAAS,eAAe,KAAK,MAAM,aAAa;AAChD,aAAS,YAAY;AACrB,SAAK,eAAe,YAAY,QAAQ;AACxC,gBAAY,iBAAiB,YAAY,mBAAmB,OAAO,QAAQ;AAC3E,UAAM,WAAW,CAAC;AAClB,aAAS,eAAe,KAAK,MAAM,aAAa;AAChD,aAAS,YAAY;AACrB,SAAK,gBAAgB,YAAY,QAAQ;AACzC,SAAK,aAAa,OAAO,aAAa,KAAK,aAAa;AACxD,SAAK,cAAc,OAAO,cAAc,KAAK,cAAc;AAC3D,SAAK,YAAY,KAAK,cAAc,SAAS,mCAAmC,KAAK;AACrF,gCAA4B,KAAK,SAAS;AAC1C,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,oBAAoB,iEAAiE;AAAA,IACjG;AACA,SAAK,YAAY,KAAK,MAAM;AAC5B,SAAK,kBAAkB,KAAK,MAAM;AAClC,SAAK,cAAc,KAAK,MAAM;AAC9B,SAAK,kBAAkB;AACvB,SAAK,aAAa;AAClB,SAAK,YAAY,KAAK,MAAM;AAC5B,SAAK,eAAe;AAAA,EACtB;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,UAAU,OAAO;AACnB,SAAK,aAAa;AAClB,QAAI,KAAK,gBAAgB,MAAM;AAC7B,WAAK,aAAa,YAAY;AAAA,IAChC;AACA,QAAI,KAAK,iBAAiB,MAAM;AAC9B,WAAK,cAAc,YAAY;AAAA,IACjC;AAAA,EACF;AAAA,EACA,aAAa;AACX,WAAO,KAAK,aAAa,WAAW,EAAE,OAAO,KAAK,cAAc,WAAW,CAAC;AAAA,EAC9E;AAAA,EACA,WAAW,SAAS;AAClB,UAAM,aAAa,QAAQ;AAC3B,UAAM,iBAAiB,KAAK,MAAM,aAAa,CAAC;AAChD,SAAK,aAAa,WAAW,QAAQ,MAAM,GAAG,cAAc,CAAC;AAC7D,SAAK,cAAc,WAAW,QAAQ,MAAM,cAAc,CAAC;AAAA,EAC7D;AAAA,EACA,mBAAmB,YAAY;AAC7B,QAAI,cAAc,KAAK,aAAa,mBAAmB,UAAU;AACjE,QAAI,EAAE,MAAM,QAAQ,WAAW,KAAK,MAAM,QAAQ,YAAY,EAAE,IAAI;AAClE,oBAAc,CAAC,WAAW;AAAA,IAC5B;AACA,kBAAc;AACd,QAAI;AACJ,QAAI;AACJ,QAAI;AACJ,QAAI,KAAK,aAAa;AACpB,mBAAa,YAAY,MAAM,CAAC;AAChC,oBAAc,YAAY;AAAA,IAC5B,OAAO;AACL,oBAAc,YAAY;AAAA,IAC5B;AACA,kBAAc;AACd,QAAI,KAAK,cAAc,UAAU;AAC/B,kBAAY,YAAY,SAAS,MAAM;AACvC,qBAAe,CAAC,WAAW;AAAA,IAC7B,WAAW,KAAK,aAAa,MAAM;AACjC,qBAAe,CAAC,aAAa,YAAY,MAAM,CAAC;AAAA,IAClD,OAAO;AACL,qBAAe,CAAC,WAAW;AAAA,IAC7B;AACA,QAAI,KAAK,aAAa;AACpB,UAAI,KAAK,aAAa,MAAM;AAC1B,eAAO,aAAa,OAAO,UAAU,EAAE,OAAO,WAAW,MAAM,CAAC;AAAA,MAClE;AACA,aAAO,CAAC,WAAW,EAAE,OAAO,UAAU,EAAE,OAAO,WAAW,MAAM,CAAC;AAAA,IACnE;AACA,WAAO,iBAAiB,YAAY;AAAA,EACtC;AAAA,EACA,MAAM,QAAQ,QAAQ;AACpB,QAAI,eAAe,UAAU,OAAO,OAAO,OAAO;AAClD,QAAI,YAAY,UAAU,OAAO,OAAO,OAAO;AAC/C,QAAI,UAAU,MAAM;AAClB,eAAS,CAAC;AAAA,IACZ;AACA,UAAM,eAAe,gBAAgB,QAAQ,cAAc,WAAW,KAAK,YAAY;AACvF,aAAS,aAAa;AACtB,mBAAe,aAAa;AAC5B,gBAAY,aAAa;AACzB,QAAI,MAAM,QAAQ,MAAM,GAAG;AACzB,qBAAe,OAAO,MAAM,CAAC;AAC7B,eAAS,OAAO;AAAA,IAClB;AACA,SAAK,gBAAgB,QAAQ,aAAa,WAAW,MAAM,aAAa,MAAM;AAC5E,aAAO,MAAM,MAAM,QAAQ,MAAM;AAAA,IACnC;AACA,UAAM,mBAAmB,CAAC;AAC1B,UAAM,kBAAkB,CAAC;AACzB,QAAI,gBAAgB,MAAM;AACxB,YAAM,YAAY,aAAa;AAC/B,UAAI,YAAY,IAAI,GAAG;AACrB,cAAM,IAAI,WAAW,+HAA+H;AAAA,MACtJ;AACA,aAAO,kBAAkB;AACzB,uBAAiB,KAAK,GAAG,YAAY;AACrC,YAAM,aAAa,aAAa,IAAI,CAAC,UAAU,IAAI,UAAU,EAAE,OAAO,MAAM,MAAM,CAAC,CAAC;AACpF,WAAK,aAAa,YAAY,WAAW,MAAM,GAAG,YAAY,CAAC;AAC/D,WAAK,cAAc,YAAY,WAAW,MAAM,YAAY,CAAC;AAC7D,sBAAgB,KAAK,GAAG,UAAU;AAAA,IACpC;AACA,QAAI,aAAa,MAAM;AACrB,YAAM,IAAI,oBAAoB,uEAAuE;AAAA,IACvG;AACA,UAAM,mBAAmB,iBAAiB,cAAc;AACxD,eAAW,WAAW,kBAAkB;AACtC,UAAI,mBAAmB,mBAAmB,kBAAkB;AAC1D,cAAM,IAAI,WAAW,8GAA8G;AAAA,MACrI;AAAA,IACF;AACA,QAAI,kBAAkB;AACpB,YAAM,YAAY,CAAC,MAAM,EAAE,OAAO,gBAAgB;AAClD,YAAM,gBAAgB,KAAK,UAAU,OAAO,eAAe;AAC3D,YAAM,oBAAoB,KAAK;AAC/B,WAAK,YAAY;AACjB,YAAM,SAAS,MAAM,MAAM,WAAW,MAAM;AAC5C,WAAK,YAAY;AACjB,aAAO;AAAA,IACT,OAAO;AACL,aAAO,MAAM,MAAM,QAAQ,MAAM;AAAA,IACnC;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,eAAe,OAAO;AAC5B,UAAI;AACJ,UAAI;AACJ,UAAI,gBAAgB,MAAM;AACxB,YAAI,KAAK,aAAa,KAAK,QAAQ,MAAM;AACzC,eAAO,KAAK,cAAc,KAAK,QAAQ,MAAM;AAAA,MAC/C,OAAO;AACL,cAAM,eAAe,aAAa,MAAM,GAAG,aAAa,SAAS,CAAC;AAClE,cAAM,gBAAgB,aAAa,MAAM,aAAa,SAAS,CAAC;AAChE,YAAI,KAAK,aAAa,KAAK,QAAQ,OAAO,OAAO,QAAQ,EAAE,cAAc,aAAa,CAAC,CAAC;AACxF,eAAO,KAAK,cAAc,KAAK,QAAQ,OAAO,OAAO,QAAQ,EAAE,cAAc,cAAc,CAAC,CAAC;AAAA,MAC/F;AACA,UAAI;AACJ,UAAI,KAAK,aAAa;AACpB,YAAI,MAAM,QAAQ,CAAC,GAAG;AACpB,mBAAS,EAAE,MAAM,CAAC,EAAE,OAAO,KAAK,MAAM,CAAC,CAAC;AAAA,QAC1C,OAAO;AAAA,QACP;AACA,YAAI,EAAE;AACN,eAAO,KAAK;AAAA,MACd;AACA,UAAI,KAAK,iBAAiB;AACxB,eAAO,QAAQ,MAAM,CAAC;AAAA,MACxB;AACA,UAAI;AACJ,UAAI,KAAK,cAAc,UAAU;AAC/B,iBAAS,YAAY,CAAC,GAAG,IAAI,CAAC;AAAA,MAChC,WAAW,KAAK,cAAc,OAAO;AACnC,iBAAS,KAAK,GAAG,IAAI;AAAA,MACvB,WAAW,KAAK,cAAc,OAAO;AACnC,iBAAS,IAAI,KAAK,KAAK,GAAG,IAAI,CAAC;AAAA,MACjC,WAAW,KAAK,cAAc,OAAO;AACnC,iBAAS,IAAI,GAAG,IAAI;AAAA,MACtB,WAAW,KAAK,aAAa,MAAM;AACjC,iBAAS,CAAC,GAAG,IAAI;AAAA,MACnB;AACA,UAAI,KAAK,aAAa;AACpB,YAAI,KAAK,aAAa,MAAM;AAC1B,iBAAO,OAAO,OAAO,MAAM;AAAA,QAC7B;AACA,eAAO,CAAC,MAAM,EAAE,OAAO,MAAM;AAAA,MAC/B;AACA,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EACA,YAAY,QAAQ;AAClB,SAAK,aAAa,YAAY;AAC9B,SAAK,cAAc,YAAY;AAAA,EACjC;AAAA,EACA,MAAM,YAAY;AAChB,cAAU,KAAK,aAAa,MAAM,MAAM;AACtC,WAAK,aAAa,MAAM,UAAU;AAAA,IACpC,CAAC;AACD,cAAU,KAAK,cAAc,MAAM,MAAM;AACvC,WAAK,cAAc,MAAM,UAAU;AAAA,IACrC,CAAC;AACD,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,YAAY,QAAQ,MAAM;AACxB,QAAI,MAAM,QAAQ,IAAI,GAAG;AACvB,aAAO,KAAK;AAAA,IACd;AACA,QAAI;AACJ,QAAI,KAAK,iBAAiB;AACxB,UAAI,KAAK,aAAa,MAAM;AAC1B,qBAAa,CAAC,MAAM,IAAI;AAAA,MAC1B,OAAO;AACL,qBAAa;AAAA,MACf;AAAA,IACF,OAAO;AACL,UAAI,KAAK,aAAa,MAAM;AAC1B,qBAAa,CAAC,MAAM,IAAI;AAAA,MAC1B,OAAO;AACL,qBAAa;AAAA,MACf;AAAA,IACF;AACA,QAAI,KAAK,aAAa;AACpB,YAAM,SAAS,KAAK,aAAa;AACjC,YAAM,YAAY,OAAO,IAAI,CAAC,UAAU,IAAI;AAC5C,UAAI,MAAM,QAAQ,UAAU,GAAG;AAC7B,eAAO,WAAW,OAAO,SAAS,EAAE,OAAO,SAAS;AAAA,MACtD,OAAO;AACL,eAAO,CAAC,UAAU,EAAE,OAAO,SAAS,EAAE,OAAO,SAAS;AAAA,MACxD;AAAA,IACF,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,IAAI,mBAAmB;AACrB,WAAO,KAAK,aAAa,iBAAiB,OAAO,KAAK,cAAc,gBAAgB;AAAA,EACtF;AAAA,EACA,IAAI,sBAAsB;AACxB,WAAO,KAAK,aAAa,oBAAoB,OAAO,KAAK,cAAc,mBAAmB;AAAA,EAC5F;AAAA,EACA,6BAA6B,OAAO;AAClC,UAAM,6BAA6B,KAAK;AACxC,QAAI,KAAK,gBAAgB,MAAM;AAC7B,WAAK,aAAa,6BAA6B,KAAK;AAAA,IACtD;AACA,QAAI,KAAK,iBAAiB,MAAM;AAC9B,WAAK,cAAc,6BAA6B,KAAK;AAAA,IACvD;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,aAAa,KAAK;AAAA,IACpB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,OAAO,WAAW,KAAK,QAAQ;AAC7B,UAAM,WAAW,YAAY,OAAO,QAAQ;AAC5C,WAAO,OAAO;AACd,QAAI,OAAO,mBAAmB,MAAM;AAClC,YAAM,IAAI,oBAAoB,0FAA0F;AAAA,IAC1H;AACA,UAAM,YAAY;AAClB,cAAU,WAAW;AACrB,WAAO,IAAI,IAAI,SAAS;AAAA,EAC1B;AACF;AACA,cAAc,YAAY;AAC1B,sBAAsB,cAAc,aAAa;AAGjD,IAAI,YAAY,cAAc,MAAM;AAAA,EAClC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,QAAQ,KAAK;AAClB,QAAI,KAAK,QAAQ;AACf,WAAK,SAAS,KAAK;AAAA,IACrB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,SAAS,KAAK;AAAA,MACd,UAAU,KAAK;AAAA,IACjB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI,OAAO,UAAU,WAAW;AAC9B,iBAAS,MAAM,QAAQ,SAAS;AAAA,MAClC;AACA,aAAO,KAAK,IAAI,QAAQ,KAAK,KAAK,GAAG,KAAK,MAAM;AAAA,IAClD,CAAC;AAAA,EACH;AACF;AACA,UAAU,YAAY;AACtB,sBAAsB,cAAc,SAAS;AAG7C,IAAI,qBAAqB,CAAC,YAAY,SAAS;AAC/C,IAAI,wBAAwB,IAAI,IAAI,kBAAkB;AACtD,IAAI,WAAW,cAAc,MAAM;AAAA,EACjC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,SAAS,KAAK;AACnB,SAAK,QAAQ,KAAK;AAClB,QAAI,KAAK,eAAe;AACtB,UAAI,sBAAsB,IAAI,KAAK,aAAa,GAAG;AACjD,aAAK,gBAAgB,KAAK;AAAA,MAC5B,OAAO;AACL,cAAM,IAAI,WAAW,oCAAoC,KAAK,kCAAkC;AAAA,MAClG;AAAA,IACF,OAAO;AACL,WAAK,gBAAgB;AAAA,IACvB;AACA,SAAK,oBAAoB,QAAQ,KAAK,iBAAiB;AAAA,EACzD;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,UAAM,cAAc,WAAW;AAC/B,WAAO,CAAC,KAAK,QAAQ,KAAK,OAAO,WAAW;AAAA,EAC9C;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,UAAU,KAAK;AAAA,MACf,SAAS,KAAK;AAAA,MACd,iBAAiB,KAAK;AAAA,MACtB,qBAAqB,KAAK;AAAA,IAC5B;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,YAAM,OAAO,CAAC,KAAK,QAAQ,KAAK,KAAK;AACrC,UAAI,KAAK,kBAAkB,YAAY;AACrC,eAAO,MAAM,eAAe,QAAQ,MAAM,CAAC,KAAK,iBAAiB;AAAA,MACnE,WAAW,KAAK,kBAAkB,WAAW;AAC3C,eAAO,MAAM,sBAAsB,QAAQ,MAAM,CAAC,KAAK,iBAAiB;AAAA,MAC1E,OAAO;AACL,cAAM,IAAI,MAAM,oBAAoB,KAAK,0BAA0B,CAAC,GAAG,qBAAqB,iBAAiB;AAAA,MAC/G;AAAA,IACF,CAAC;AAAA,EACH;AACF;AACA,SAAS,YAAY;AACrB,sBAAsB,cAAc,QAAQ;AAG5C,SAAS,wBAAwB,QAAQ,YAAY,OAAO,SAAS;AACnE,MAAI,SAAS,oBAAoB,MAAM;AACvC,MAAI,OAAO,UAAU,SAAS;AAC5B,aAAS,MAAM,QAAQ,OAAO;AAAA,EAChC;AACA,MAAI,eAAe,OAAO;AACxB,WAAO;AAAA,EACT;AACA,QAAM,gBAAgB,OAAO;AAC7B,MAAI,OAAO,SAAS,GAAG;AACrB,aAAS,WAAW,QAAQ,EAAE;AAAA,EAChC;AACA,MAAI,eAAe,UAAU;AAC3B,QAAI,OAAO,MAAM,OAAO,MAAM,SAAS,OAAO,GAAG;AAC/C,eAAS,WAAW,QAAQ,EAAE;AAAA,IAChC;AAAA,EACF;AACA,MAAI,OAAO,OAAO,GAAG;AACnB,UAAM,IAAI,WAAW,4EAA4E,8BAA8B,mDAAmD,OAAO,OAAO;AAAA,EAClM;AACA,QAAM,eAAe,CAAC,YAAY,QAAQ,EAAE,SAAS,UAAU;AAC/D,QAAM,qBAAqB;AAC3B,MAAI;AACJ,MAAI,OAAO,YAAY,eAAe,eAAe,SAAS;AAC5D,gBAAY,cAAc,oBAAoB,SAAS,OAAO,YAAY;AAAA,EAC5E,OAAO;AACL,gBAAY,cAAc,oBAAoB,CAAC,GAAG,OAAO,YAAY;AAAA,EACvE;AACA,MAAI,eAAe,SAAS;AAC1B,WAAO;AAAA,EACT;AACA,MAAI,SAAS;AACX,WAAO,IAAI,WAAW,OAAO;AAAA,EAC/B,OAAO;AACL,UAAM,IAAI,WAAW,uDAAuD;AAAA,EAC9E;AACF;AAGA,IAAI,mBAAmB,cAAc,MAAM;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM,IAAI;AACV,SAAK,YAAY,KAAK;AACtB,QAAI,KAAK,YAAY;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB,OAAO;AACL,WAAK,aAAa;AAAA,IACpB;AAAA,EACF;AAAA,EACA,YAAY;AACV,UAAM,SAAS;AAAA,MACb,aAAa,KAAK;AAAA,MAClB,cAAc,KAAK;AAAA,IACrB;AACA,UAAM,aAAa,MAAM,UAAU;AACnC,WAAO,OAAO,QAAQ,UAAU;AAChC,WAAO;AAAA,EACT;AAAA,EACA,mBAAmB,YAAY;AAC7B,iBAAa,mBAAmB,UAAU;AAC1C,QAAI,cAAc,MAAM;AACtB,aAAO,CAAC,KAAK,SAAS;AAAA,IACxB;AACA,QAAI,KAAK,eAAe,YAAY,WAAW,WAAW,SAAS,OAAO,GAAG;AAC3E,iBAAW,KAAK,KAAK,SAAS;AAC9B,aAAO;AAAA,IACT;AACA,eAAW,WAAW,SAAS,KAAK,KAAK;AACzC,WAAO;AAAA,EACT;AAAA,EACA,KAAK,QAAQ,QAAQ;AACnB,WAAO,KAAK,MAAM;AAChB,eAAS,oBAAoB,MAAM;AACnC,UAAI,OAAO,UAAU,SAAS;AAC5B,iBAAS,MAAM,QAAQ,OAAO;AAAA,MAChC;AACA,UAAI;AACJ,UAAI,OAAO,OAAO,oBAAoB,aAAa;AACjD,YAAI,KAAK,eAAe,SAAS;AAC/B,gBAAM,IAAI,WAAW;AAAA,sCACO,OAAO,iBAAiB;AAAA,QACtD;AACA,uBAAe,oBAAoB,OAAO,eAAe;AAAA,MAC3D;AACA,YAAM,WAAW,IAAI,MAAM;AAC3B,YAAM,WAAW,IAAI,MAAM;AAC3B,YAAM,kBAAkB,QAAQ,KAAK,WAAW,QAAQ,EAAE,WAAW,EAAE,IAAI,CAAC;AAC5E,YAAM,aAAa,aAAa,UAAU,CAAC,EAAE,WAAW,EAAE,IAAI,CAAC;AAC/D,UAAI,EAAE,mBAAmB,aAAa;AACpC,cAAM,IAAI,WAAW,uEAAuE,KAAK,WAAW;AAAA,MAC9G;AACA,aAAO,wBAAwB,QAAQ,KAAK,YAAY,KAAK,WAAW,YAAY;AAAA,IACtF,CAAC;AAAA,EACH;AACF;AACA,iBAAiB,YAAY;AAC7B,sBAAsB,cAAc,gBAAgB;AAGpD,SAAS,WAAW,MAAM;AACxB,SAAO,IAAI,WAAW,IAAI;AAC5B;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,IAAI,IAAI;AACrB;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,KAAK,IAAI;AACtB;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,OAAO,MAAM;AACpB,SAAO,IAAI,MAAM,IAAI;AACvB;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,gBAAgB,MAAM;AAC7B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,OAAO,IAAI;AACxB;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,IAAI,WAAW,IAAI;AAC5B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,IAAI,YAAY,IAAI;AAC7B;AACA,SAAS,MAAM,MAAM;AACnB,SAAO,IAAI,MAAM,IAAI;AACvB;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,iBAAiB,IAAI;AAClC;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,KAAK,IAAI;AACtB;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,YAAY,IAAI;AAC7B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,IAAI,IAAI;AACrB;AACA,SAAS,oBAAoB,MAAM;AACjC,SAAO,IAAI,mBAAmB,IAAI;AACpC;AACA,SAAS,mBAAmB,MAAM;AAChC,SAAO,IAAI,mBAAmB,IAAI;AACpC;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,iBAAiB,IAAI;AAClC;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,iBAAiB,IAAI;AAClC;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,iBAAiB,IAAI;AAClC;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,iBAAiB,IAAI;AAC9B;AACA,SAAS,uBAAuB,MAAM;AACpC,SAAO,IAAI,uBAAuB,IAAI;AACxC;AACA,SAAS,uBAAuB,MAAM;AACpC,SAAO,IAAI,uBAAuB,IAAI;AACxC;AACA,SAAS,mBAAmB,MAAM;AAChC,SAAO,IAAI,mBAAmB,IAAI;AACpC;AACA,SAAS,mBAAmB,MAAM;AAChC,SAAO,IAAI,mBAAmB,IAAI;AACpC;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,IAAI,MAAM;AACjB,SAAO,IAAI,IAAI,IAAI;AACrB;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,KAAK,IAAI;AACtB;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,WAAW,MAAM;AACxB,SAAO,IAAI,WAAW,IAAI;AAC5B;AACA,SAAS,eAAe,MAAM;AAC5B,SAAO,IAAI,eAAe,IAAI;AAChC;AACA,SAAS,KAAK,MAAM;AAClB,SAAO,IAAI,IAAI,IAAI;AACrB;AACA,SAAS,gBAAgB,MAAM;AAC7B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,gBAAgB,MAAM;AAC7B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,IAAI,kBAAkB;AACtB,IAAI,kBAAkB;AACtB,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,SAAS,gBAAgB,MAAM;AAC7B,SAAO,IAAI,gBAAgB,IAAI;AACjC;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,IAAI,aAAa,IAAI;AAC9B;AACA,SAAS,QAAQ,MAAM;AACrB,SAAO,IAAI,QAAQ,IAAI;AACzB;AACA,SAAS,UAAU,MAAM;AACvB,SAAO,IAAI,UAAU,IAAI;AAC3B;AACA,SAAS,SAAS,MAAM;AACtB,SAAO,IAAI,SAAS,IAAI;AAC1B;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,iBAAiB,IAAI;AAClC;AAGA,IAAI,0BAA0B,CAAC;AAC/BJ,UAAS,yBAAyB;AAAA,EAChC,MAAM,MAAM;AAAA,EACZ,KAAK,MAAM;AAAA,EACX,gBAAgB,MAAM;AAAA,EACtB,oBAAoB,MAAM;AAAA,EAC1B,qBAAqB,MAAM;AAAA,EAC3B,yBAAyB,MAAM;AAAA,EAC/B,iBAAiB,MAAM;AAAA,EACvB,MAAM,MAAM;AAAA,EACZ,mBAAmB,MAAM;AAAA,EACzB,6BAA6B,MAAM;AAAA,EACnC,kBAAkB,MAAM;AAAA,EACxB,KAAK,MAAM;AAAA,EACX,WAAW,MAAM;AAAA,EACjB,QAAQ,MAAM;AAAA,EACd,2BAA2B,MAAM;AACnC,CAAC;AACD,SAAS,gBAAgB,OAAO,OAAO;AACrC,SAAO,eAAe,OAAO,KAAK;AACpC;AACA,SAAS,oBAAoB,OAAO,OAAO;AACzC,SAAO,oBAAoB,OAAO,KAAK;AACzC;AACA,SAAS,2BAA2B,OAAO,OAAO;AAChD,SAAO,0BAA0B,OAAO,KAAK;AAC/C;AACA,SAAS,qBAAqB,OAAO,OAAO;AAC1C,SAAO,oBAAoB,OAAO,KAAK;AACzC;AACA,SAAS,yBAAyB,OAAO,OAAO;AAC9C,SAAO,yBAAyB,OAAO,KAAK;AAC9C;AACA,SAAS,WAAW,OAAO,OAAO;AAChC,SAAO,UAAU,OAAO,KAAK;AAC/B;AACA,SAAS,QAAQ,OAAO,OAAO;AAC7B,SAAO,OAAO,OAAO,KAAK;AAC5B;AACA,SAAS,iBAAiB,OAAO,OAAO;AACtC,SAAO,gBAAgB,OAAO,KAAK;AACrC;AACA,SAAS,mBAAmB,OAAO,OAAO;AACxC,SAAO,kBAAkB,OAAO,KAAK;AACvC;AACA,SAAS,6BAA6B,OAAO,OAAO;AAClD,SAAO,4BAA4B,OAAO,KAAK;AACjD;AACA,SAAS,MAAM,OAAO,OAAO;AAC3B,SAAO,4BAA4B,OAAO,KAAK;AACjD;AACA,SAAS,MAAM,OAAO,OAAO;AAC3B,SAAO,4BAA4B,OAAO,KAAK;AACjD;AACA,SAAS,kBAAkB,OAAO,OAAO;AACvC,SAAO,kBAAkB,OAAO,KAAK;AACvC;AACA,SAAS,KAAK,OAAO,OAAO;AAC1B,SAAO,kBAAkB,OAAO,KAAK;AACvC;AACA,SAAS,KAAK,OAAO,OAAO;AAC1B,SAAO,kBAAkB,OAAO,KAAK;AACvC;AAGA,IAAI,yBAAyB,CAAC;AAC9BA,UAAS,wBAAwB;AAAA,EAC/B,eAAe,MAAM;AACvB,CAAC;AAGD,IAAI,+BAA+B,CAAC;AACpCA,UAAS,8BAA8B;AAAA,EACrC,IAAI,MAAM;AAAA,EACV,MAAM,MAAM;AAAA,EACZ,IAAI,MAAM;AACZ,CAAC;AACD,SAAS,KAAK,QAAQ;AACpB,SAAO,IAAI,KAAK,MAAM;AACxB;AACA,SAAS,IAAI,QAAQ;AACnB,SAAO,GAAG,MAAM;AAClB;AACA,SAAS,IAAI,QAAQ;AACnB,SAAO,GAAG,MAAM;AAClB;AAGA,IAAI,WAAW,cAAc,aAAa;AAAA,EACxC,cAAc;AACZ,UAAM,GAAG,SAAS;AAClB,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,SAAS,QAAQ;AACf,QAAI,EAAE,kBAAkB,cAAc;AACpC,YAAM,IAAI,MAAM,uDAAuD;AAAA,IACzE;AACA,SAAK,QAAQ;AAAA,EACf;AACF;AACA,SAAS,MAAM,SAAS,SAAS;AAC/B,SAAO,UAAU;AACnB;AACA,SAAS,SAAS,SAAS,SAAS;AAClC,SAAO,UAAU;AACnB;AACA,IAAI,gBAAgB,cAAc,SAAS;AAAA,EACzC,YAAY,MAAM;AAChB,UAAM;AACN,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,QAAI,KAAK,oBAAoB;AAC3B,YAAM,IAAI,oBAAoB,oEAAoE;AAAA,IACpG;AACA,SAAK,UAAU,KAAK,WAAW;AAC/B,SAAK,WAAW,KAAK,IAAI,KAAK,YAAY,CAAC;AAC3C,SAAK,WAAW,KAAK,YAAY;AACjC,SAAK,UAAU,KAAK,WAAW;AAC/B,SAAK,OAAO,KAAK,QAAQ;AACzB,SAAK,WAAW,KAAK;AACrB,QAAI,CAAC,QAAQ,OAAO,KAAK,EAAE,QAAQ,KAAK,IAAI,MAAM,IAAI;AACpD,cAAQ,KAAK,uBAAuB,KAAK,gDAAgD;AACzF,WAAK,OAAO;AAAA,IACd;AACA,QAAI,KAAK,SAAS,OAAO;AACvB,WAAK,cAAc;AAAA,IACrB,WAAW,KAAK,SAAS,OAAO;AAC9B,WAAK,cAAc;AAAA,IACrB,OAAO;AACL,UAAI,KAAK,QAAQ,QAAQ,KAAK,MAAM,IAAI;AACtC,aAAK,cAAc;AAAA,MACrB,OAAO;AACL,aAAK,cAAc;AAAA,MACrB;AAAA,IACF;AACA,QAAI,KAAK,gBAAgB,OAAO;AAC9B,WAAK,YAAY;AAAA,IACnB;AAAA,EACF;AAAA,EACA,MAAM,aAAa,MAAM;AACvB,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,QAAI,KAAK,YAAY,MAAM;AACzB,WAAK,OAAO,KAAK;AAAA,IACnB,OAAO;AACL,WAAK,OAAO,KAAK,gBAAgB,QAAQ,WAAW;AAAA,IACtD;AAAA,EACF;AAAA,EACA,MAAM,WAAW,OAAO,MAAM;AAC5B,UAAM,qBAAqB,IAAI;AAC/B,UAAM,UAAU,KAAK,gBAAgB,IAAI;AACzC,QAAI,WAAW,MAAM;AACnB;AAAA,IACF;AACA,QAAI,KAAK,YAAY,UAAU,KAAK,UAAU,KAAK,IAAI,GAAG;AACxD,WAAK,OAAO;AACZ,WAAK,OAAO;AAAA,IACd,OAAO;AACL,WAAK;AACL,UAAI,KAAK,QAAQ,KAAK,UAAU;AAC9B,aAAK,eAAe;AACpB,aAAK,MAAM,eAAe;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EACA,MAAM,WAAW,MAAM;AACrB,QAAI,KAAK,eAAe,KAAK,KAAK,SAAS;AACzC,cAAQ,IAAI,SAAS,KAAK,+BAA+B;AAAA,IAC3D;AAAA,EACF;AAAA,EACA,gBAAgB,MAAM;AACpB,QAAI,QAAQ,MAAM;AAChB,aAAO,CAAC;AAAA,IACV;AACA,UAAM,eAAe,KAAK,KAAK;AAC/B,QAAI,gBAAgB,MAAM;AACxB,cAAQ,KAAK,4BAA4B,KAAK,oDAAoD,OAAO,KAAK,IAAI,GAAG;AAAA,IACvH;AACA,WAAO;AAAA,EACT;AACF;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,IAAI,cAAc,IAAI;AAC/B;AACA,IAAI,YAAY,EAAE,cAAc;AAGhC,IAAI,OAAO,IAAI;AACf,KAAK,aAAa,6BAA6B,MAAM,OAAO,CAAC,eAAe;AAC1E,MAAI,YAAY;AACd,YAAQ,KAAK,+OAA+O;AAAA,EAC9P;AACF,CAAC;AAGD,IAAI;AAAA,CACH,SAAS,WAAW;AACnB,YAAU,UAAU,gBAAgB,KAAK;AACzC,YAAU,UAAU,cAAc,KAAK;AACvC,YAAU,UAAU,eAAe,KAAK;AACxC,YAAU,UAAU,cAAc,KAAK;AACvC,YAAU,UAAU,cAAc,KAAK;AACvC,YAAU,UAAU,cAAc,KAAK;AACvC,YAAU,UAAU,aAAa,KAAK;AACtC,YAAU,UAAU,eAAe,KAAK;AACxC,YAAU,UAAU,kBAAkB,KAAK;AAC3C,YAAU,UAAU,cAAc,KAAK;AACvC,YAAU,UAAU,aAAa,MAAM;AACvC,YAAU,UAAU,cAAc,MAAM;AACxC,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,iBAAiB,MAAM;AAC3C,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,gBAAgB,MAAM;AAC1C,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,mBAAmB,MAAM;AAC7C,YAAU,UAAU,aAAa,MAAM;AACvC,YAAU,UAAU,iBAAiB,MAAM;AAC3C,YAAU,UAAU,gBAAgB,MAAM;AAC1C,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,eAAe,MAAM;AACzC,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,iBAAiB,OAAO;AAC5C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,sBAAsB,OAAO;AACjD,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,iBAAiB,OAAO;AAC5C,YAAU,UAAU,kBAAkB,OAAO;AAC7C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,qBAAqB,OAAO;AAChD,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,oBAAoB,OAAO;AAC/C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,uBAAuB,OAAO;AAClD,YAAU,UAAU,iBAAiB,OAAO;AAC5C,YAAU,UAAU,qBAAqB,OAAO;AAChD,YAAU,UAAU,oBAAoB,OAAO;AAC/C,YAAU,UAAU,mBAAmB,OAAO;AAC9C,YAAU,UAAU,mBAAmB,OAAO;AAChD,GAAG,aAAa,WAAW,CAAC,EAAE;AAC9B,IAAI;AAAA,CACH,SAAS,WAAW;AACnB,MAAI;AACJ,GAAC,SAAS,0BAA0B;AAClC,6BAAyB,yBAAyB,YAAY,KAAK;AACnE,6BAAyB,yBAAyB,QAAQ,KAAK;AAC/D,6BAAyB,yBAAyB,QAAQ,KAAK;AAAA,EACjE,GAAG,0BAA0B,UAAU,4BAA4B,UAAU,0BAA0B,CAAC,EAAE;AAC5G,GAAG,aAAa,WAAW,CAAC,EAAE;AAG9B,IAAI,aAAa,CAAC;AAClB,SAAS,WAAW,MAAM,QAAQ;AAChC,QAAM,WAAW;AAAA,IACf,UAAU;AAAA,IACV,UAAU;AAAA,IACV,QAAQ,CAAC;AAAA,IACT,OAAO,CAAC;AAAA,IACR,gBAAgB;AAAA,EAClB;AACA,aAAW,QAAQ;AACrB;AACA,SAAS,gBAAgB,MAAM;AAC7B,SAAO,WAAW;AACpB;AACA,SAAS,aAAa,MAAM;AAC1B,SAAO,WAAW;AACpB;AAGA,SAAS,cAAc,WAAW,MAAM,WAAW,SAAS,iBAAiB;AAC3E,QAAM,aAAa,KAAK,YAAY;AACpC,MAAI,cAAc,WAAW,oBAAoB,QAAQ;AACvD,UAAM,QAAQ,WAAW;AACzB,UAAM,MAAM,WAAW,kBAAkB,IAAI,SAAS,WAAW,kBAAkB,SAAS,QAAQ,IAAI,WAAW;AACnH,QAAI,WAAW,SAAS,UAAU;AAChC,aAAO,UAAU,KAAK,WAAW,WAAW,kBAAkB,WAAW,SAAS,eAAe;AAAA,IACnG;AACA,QAAI,WAAW,SAAS,WAAW;AACjC,YAAM,SAAS,KAAK,WAAW,MAAM,OAAO,GAAG;AAC/C,aAAO,OAAO,IAAI,CAAC,SAAS,UAAU,MAAM,WAAW,SAAS,eAAe,CAAC;AAAA,IAClF;AACA,UAAM,UAAU,UAAU,KAAK,WAAW,MAAM,KAAK,EAAE,IAAI,WAAW,SAAS,eAAe;AAC9F,UAAM,OAAO,QAAQ,SAAS;AAC9B,WAAO,WAAW,SAAS,WAAW,KAAK,KAAK,aAAa,cAAc,QAAQ,OAAO,IAAI;AAAA,EAChG;AACA,QAAM,YAAY,KAAK,WAAW;AAClC,SAAO,aAAa,UAAU;AAChC;AACA,SAAS,UAAU,MAAM,YAAY,SAAS,iBAAiB;AAC7D,QAAM,CAAC,UAAU,KAAK,IAAI,cAAc,IAAI;AAC5C,MAAI,mBAAmB,MAAM;AAC3B,UAAM,UAAU,gBAAgB,yBAAyB,QAAQ;AACjE,QAAI,WAAW,MAAM;AACnB,aAAO;AAAA,IACT;AAAA,EACF;AACA,QAAM,YAAY,QAAQ,kBAAkB,KAAK,CAAC,eAAe;AAC/D,WAAO,CAAC,CAAC,WAAW,yBAAyB,UAAU,UAAU;AAAA,EACnE,CAAC;AACD,SAAO,cAAc,SAAS,WAAW,yBAAyB,UAAU,SAAS,GAAG,SAAS;AACnG;AACA,SAAS,6BAA6B,MAAM,YAAY,SAAS;AAC/D,SAAO,WAAW,yBAAyB,MAAM,QAAQ,gBAAgB;AAC3E;AACA,SAAS,oBAAoB,WAAW,SAAS;AAC/C,QAAM,CAAC,UAAU,OAAO,UAAU,IAAI,cAAc,SAAS;AAC7D,SAAO;AAAA,IACL,yBAAyB,UAAU,WAAW,QAAQ,gBAAgB;AAAA,IACtE;AAAA,IACA;AAAA,EACF;AACF;AACA,SAAS,yBAAyB,MAAM,WAAW;AACjD,SAAO,CAAC,CAAC,YAAY,GAAG,QAAQ,cAAc;AAChD;AACA,SAAS,cAAc,MAAM;AAC3B,QAAM,QAAQ,KAAK,MAAM,GAAG;AAC5B,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO,CAAC,MAAM,GAAG,MAAM;AAAA,EACzB;AACA,QAAM,WAAW,MAAM;AACvB,QAAM,aAAa,MAAM,WAAW,IAAI,MAAM,KAAK;AACnD,QAAM,QAAQ,OAAO,MAAM,MAAM,SAAS,EAAE;AAC5C,SAAO,CAAC,UAAU,OAAO,UAAU;AACrC;AACA,SAAS,WAAW,MAAM,WAAW,SAAS;AAC5C,MAAI,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AACxD,MAAI,SAAS,YAAY;AACvB,WAAO,cAAc,oBAAoB,MAAM,WAAW,OAAO;AACjE,UAAM,kBAAkB,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AACvD,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,sBAAgB,GAAG,KAAK,KAAK,IAAI;AACjC,sBAAgB,GAAG,KAAK,KAAK,IAAI,IAAI;AAAA,IACvC;AACA,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,YAAY,SAAS;AAC5B,SAAO,QAAQ,OAAO,UAAU,MAAM,OAAO;AAC/C;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,OAAO;AAAA,EACT;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kBAAkB,CAAC;AACvBA,UAAS,iBAAiB;AAAA,EACxB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,sBAAsB,CAAC;AAC3BA,UAAS,qBAAqB;AAAA,EAC5B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,QACjB,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,UACd;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,QACF;AAAA,MACF;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,UACd;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,QACF;AAAA,MACF;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,mBAAmB,CAAC;AACxBA,UAAS,kBAAkB;AAAA,EACzB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kBAAkB,CAAC;AACvBA,UAAS,iBAAiB;AAAA,EACxB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB,CAAC;AACrBA,UAAS,eAAe;AAAA,EACtB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,EACd;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU,CAAC;AAAA,EACb;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,QAAQ;AAAA,EACV;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU,CAAC;AAAA,IACX,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU,CAAC;AAAA,IACX,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,gBAAgB,CAAC;AACrBA,UAAS,eAAe;AAAA,EACtB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,kBAAkB,CAAC;AACvBA,UAAS,iBAAiB;AAAA,EACxB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,mBAAmB,CAAC;AACxBA,UAAS,kBAAkB;AAAA,EACzB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB,CAAC;AAAA,MACnB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,wBAAwB,CAAC;AAC7BA,UAAS,uBAAuB;AAAA,EAC9B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,oBAAoB,CAAC;AACzBA,UAAS,mBAAmB;AAAA,EAC1B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,OAAO;AAAA,QACP,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,QAChB,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,iBAAiB,CAAC;AACtBA,UAAS,gBAAgB;AAAA,EACvB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,mBAAmB,CAAC;AACxBA,UAAS,kBAAkB;AAAA,EACzB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,iBAAiB,CAAC;AACtBA,UAAS,gBAAgB;AAAA,EACvB,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,WAAW;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AACF;AAGA,IAAI,yBAAyB,CAAC;AAC9BA,UAAS,wBAAwB;AAAA,EAC/B,MAAM,MAAM;AACd,CAAC;AACD,IAAI,SAAS;AAAA,EACX;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,QACR,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,oBAAoB;AAAA,QACpB,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS;AAAA,MACP;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,UAAU;AAAA,QACV,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS,CAAC;AAAA,EACZ;AAAA,EACA;AAAA,IACE,YAAY;AAAA,IACZ,YAAY;AAAA,IACZ,UAAU;AAAA,MACR;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,MACA;AAAA,QACE,SAAS;AAAA,QACT,QAAQ;AAAA,QACR,QAAQ;AAAA,MACV;AAAA,IACF;AAAA,IACA,SAAS,CAAC;AAAA,EACZ;AACF;AAGA,IAAI,kBAAkB,MAAM;AAAA,EAC1B,cAAc;AACZ,UAAM,MAAM;AAAA,MACV;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AACA,UAAM,cAAc,CAAC,EAAE,OAAO,GAAG,IAAI,IAAI,CAAC,QAAQ,IAAI,IAAI,CAAC;AAC3D,SAAK,YAAY,YAAY,OAAO,CAAC,KAAK,WAAW;AACnD,UAAI,OAAO,YAAY;AACvB,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,WAAW,WAAW;AACpB,WAAO,KAAK,cAAc,KAAK,YAAY,IAAI,KAAK;AAAA,EACtD;AAAA,EACA,eAAe,OAAO,YAAY,CAAC,GAAG;AACpC,UAAM,UAAU,MAAM;AACtB,UAAM,eAAe,CAAC;AACtB,UAAM,UAAU,CAAC;AACjB,UAAM,YAAY,CAAC;AACnB,UAAM,QAAQ,QAAQ,OAAO,CAAC,KAAK,SAAS;AAC1C,UAAI,KAAK,QAAQ,KAAK,QAAQ,IAAI;AAClC,UAAI,KAAK,GAAG,WAAW,aAAa,GAAG;AACrC,qBAAa,KAAK,IAAI,KAAK,KAAK;AAAA,MAClC,WAAW,KAAK,OAAO,SAAS;AAC9B,gBAAQ,KAAK,IAAI,KAAK,KAAK;AAAA,MAC7B,WAAW,KAAK,SAAS,QAAQ,KAAK,MAAM,WAAW,GAAG;AACxD,kBAAU,KAAK,IAAI,KAAK,KAAK;AAAA,MAC/B;AACA,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AACL,QAAI,SAAS,CAAC;AACd,UAAM,UAAU,CAAC;AACjB,QAAI,qBAAqB,CAAC;AAC1B,QAAI,sBAAsB,CAAC;AAC3B,QAAI,aAAa,MAAM;AACrB,2BAAqB,KAAK,oBAAoB,UAAU,MAAM;AAC9D,4BAAsB,KAAK,oBAAoB,UAAU,OAAO;AAAA,IAClE;AACA,UAAM,WAAW,OAAO,KAAK,KAAK;AAClC,aAAS,QAAQ,CAAC,QAAQ;AACxB,YAAM,OAAO,MAAM;AACnB,WAAK,WAAW,QAAQ,CAAC,MAAM,UAAU;AACvC,cAAM,CAAC,UAAU,EAAE,UAAU,IAAI,oBAAoB,IAAI;AACzD,cAAM,YAAY,MAAM;AACxB,YAAI,UAAU,WAAW,MAAM;AAC7B,gBAAM,cAAc,UAAU,QAAQ,QAAQ,UAAU;AACxD,cAAI,gBAAgB,IAAI;AACtB,kBAAM,YAAY,GAAG,YAAY;AACjC,iBAAK,WAAW,SAAS;AAAA,UAC3B;AAAA,QACF;AACA,aAAK,OAAO,KAAK,SAAS;AAC1B,kBAAU,SAAS,KAAK,IAAI;AAAA,MAC9B,CAAC;AAAA,IACH,CAAC;AACD,QAAI,OAAO,KAAK,mBAAmB,EAAE,WAAW,GAAG;AACjD,eAAS,QAAQ,CAAC,QAAQ;AACxB,cAAM,OAAO,MAAM;AACnB,YAAI,KAAK,SAAS,WAAW,GAAG;AAC9B,kBAAQ,KAAK,IAAI;AAAA,QACnB;AAAA,MACF,CAAC;AAAA,IACH,OAAO;AACL,aAAO,KAAK,mBAAmB,EAAE,QAAQ,CAAC,SAAS;AACjD,cAAM,CAAC,QAAQ,IAAI,oBAAoB,IAAI;AAC3C,cAAM,OAAO,MAAM;AACnB,YAAI,QAAQ,MAAM;AAChB,eAAK,eAAe,oBAAoB;AACxC,kBAAQ,KAAK,IAAI;AAAA,QACnB;AAAA,MACF,CAAC;AAAA,IACH;AACA,QAAI,OAAO,KAAK,kBAAkB,EAAE,SAAS,GAAG;AAC9C,aAAO,KAAK,kBAAkB,EAAE,QAAQ,CAAC,SAAS;AAChD,cAAM,CAAC,QAAQ,IAAI,oBAAoB,IAAI;AAC3C,cAAM,OAAO,MAAM;AACnB,YAAI,MAAM;AACR,eAAK,eAAe,mBAAmB;AACvC,iBAAO,KAAK,IAAI;AAAA,QAClB;AAAA,MACF,CAAC;AAAA,IACH,OAAO;AACL,eAAS;AAAA,IACX;AACA,QAAI,YAAY,CAAC;AACjB,QAAI,MAAM,WAAW,QAAQ,MAAM,QAAQ,YAAY,MAAM;AAC3D,kBAAY,MAAM,QAAQ,SAAS,OAAO,CAAC,YAAY,UAAU;AAC/D,mBAAW,MAAM,UAAU,QAAQ,KAAK,YAAY,KAAK;AACzD,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AACA,UAAM,SAAS,EAAE,OAAO,QAAQ,SAAS,SAAS,cAAc,WAAW,UAAU;AACrF,QAAI,UAAU,SAAS,GAAG;AACxB,aAAO,YAAY;AAAA,IACrB;AACA,WAAO;AAAA,EACT;AAAA,EACA,oBAAoB,SAAS;AAC3B,WAAO,OAAO,KAAK,WAAW,CAAC,CAAC,EAAE,OAAO,CAAC,MAAM,SAAS;AACvD,WAAK,QAAQ,MAAM,QAAQ;AAC3B,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,QAAQ,MAAM;AACZ,UAAM,SAAS,gBAAgB,KAAK,EAAE,KAAK,KAAK,UAAU,KAAK,OAAO,CAAC;AACvE,QAAI,KAAK,QAAQ,MAAM;AACrB,WAAK,OAAO,CAAC;AAAA,IACf;AACA,UAAM,UAAU;AAAA,MACd,MAAM,KAAK;AAAA,MACX,IAAI,KAAK;AAAA,MACT,UAAU,OAAO;AAAA,MACjB,aAAa,KAAK,SAAS,CAAC,GAAG,IAAI,CAAC,WAAW,OAAO,WAAW,GAAG,IAAI,OAAO,MAAM,CAAC,IAAI,MAAM;AAAA,MAChG,QAAQ,CAAC;AAAA,MACT,UAAU,CAAC;AAAA,MACX,aAAa,CAAC;AAAA,MACd,YAAY,CAAC;AAAA,MACb,UAAU,KAAK;AAAA,MACf,SAAS,OAAO;AAAA,IAClB;AACA,QAAI,OAAO,UAAU,MAAM;AACzB,cAAQ,cAAc,OAAO,OAAO,OAAO,CAAC,KAAK,UAAU;AACzD,YAAI,MAAM,QAAQ;AAAA,UAChB,MAAM,MAAM;AAAA,UACZ,iBAAiB,MAAM;AAAA,UACvB,eAAe,MAAM;AAAA,QACvB;AACA,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AACA,QAAI,OAAO,SAAS,MAAM;AACxB,cAAQ,aAAa,OAAO,MAAM,OAAO,CAAC,KAAK,UAAU;AACvD,cAAM,OAAO,MAAM;AACnB,YAAI,QAAQ;AACZ,gBAAQ,MAAM,MAAM;AAAA,UAClB,KAAK;AACH,oBAAQ,eAAe,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AAClE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,eAAe,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAC9E;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,oBAAoB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACvE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,oBAAoB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YACnF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,eAAe,KAAK,MAAM,MAAM,QAAQ,MAAM,gBAAgB,CAAC;AACvE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,eAAe,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAC9E;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,qBAAqB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACxE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,qBAAqB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YACpF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,aAAa,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AAChE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,aAAa,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAC5E;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,kBAAkB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACrE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,kBAAkB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YACjF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,oBAAoB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACvE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,oBAAoB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YACnF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,yBAAyB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AAC5E,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,yBAAyB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YACxF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,cAAc,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACjE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,cAAc,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAC7E;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,mBAAmB,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AACtE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,mBAAmB,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAClF;AACA;AAAA,UACF,KAAK;AACH,oBAAQ,aAAa,KAAK,MAAM,MAAM,QAAQ,MAAM,YAAY;AAChE,gBAAI,UAAU,UAAU,CAAC,CAAC,MAAM,kBAAkB;AAChD,sBAAQ,aAAa,KAAK,MAAM,MAAM,kBAAkB,MAAM,YAAY;AAAA,YAC5E;AACA;AAAA,UACF,KAAK;AAAA,UACL,KAAK;AACH;AAAA,UACF;AACE,kBAAM,IAAI,MAAM,2BAA2B,MAAM,gBAAgB,KAAK,IAAI;AAAA,QAC9E;AACA,YAAI,MAAM,QAAQ,EAAE,OAAO,KAAK;AAChC,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,aAAa;AACvB,UAAM,UAAU,YAAY;AAC5B,UAAM,eAAe,CAAC;AACtB,UAAM,UAAU,CAAC;AACjB,QAAI,QAAQ,CAAC;AACb,QAAI,WAAW,MAAM;AACnB,cAAQ,QAAQ,OAAO,CAAC,KAAK,SAAS;AACpC,YAAI,KAAK,QAAQ,KAAK,QAAQ,IAAI;AAClC,YAAI,KAAK,OAAO,SAAS;AACvB,kBAAQ,KAAK,IAAI,KAAK,KAAK;AAAA,QAC7B;AACA,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AACA,UAAM,SAAS,CAAC;AAChB,UAAM,UAAU,CAAC;AACjB,gBAAY,UAAU,SAAS,QAAQ,CAAC,QAAQ;AAC9C,YAAM,CAAC,QAAQ,IAAI,oBAAoB,IAAI,IAAI;AAC/C,YAAM,OAAO;AAAA,QACX,MAAM;AAAA,QACN,IAAI;AAAA,QACJ,QAAQ,CAAC;AAAA,QACT,YAAY,CAAC;AAAA,QACb,UAAU;AAAA,QACV,aAAa,CAAC;AAAA,QACd,YAAY,EAAE,OAAO,EAAE,OAAO,gBAAgB,IAAI,IAAI,GAAG,MAAM,QAAQ,EAAE;AAAA,QACzE,UAAU,CAAC;AAAA,MACb;AACA,WAAK,eAAe,IAAI;AACxB,aAAO,KAAK,IAAI;AAChB,YAAM,YAAY;AAAA,IACpB,CAAC;AACD,UAAM,WAAW,OAAO,KAAK,KAAK;AAClC,aAAS,QAAQ,CAAC,QAAQ;AACxB,YAAM,OAAO,MAAM;AACnB,WAAK,WAAW,QAAQ,CAAC,MAAM,UAAU;AACvC,cAAM,CAAC,UAAU,EAAE,UAAU,IAAI,oBAAoB,IAAI;AACzD,cAAM,YAAY,MAAM;AACxB,YAAI,UAAU,WAAW,MAAM;AAC7B,gBAAM,cAAc,UAAU,QAAQ,QAAQ,UAAU;AACxD,cAAI,gBAAgB,IAAI;AACtB,kBAAM,YAAY,GAAG,YAAY;AACjC,iBAAK,WAAW,SAAS;AAAA,UAC3B;AAAA,QACF;AACA,aAAK,OAAO,KAAK,SAAS;AAC1B,kBAAU,SAAS,KAAK,IAAI;AAAA,MAC9B,CAAC;AAAA,IACH,CAAC;AACD,UAAM,gBAAgB,YAAY;AAClC,gBAAY,UAAU,UAAU,QAAQ,CAAC,WAAW;AAClD,YAAM,CAAC,UAAU,KAAK,IAAI,oBAAoB,cAAc,OAAO,KAAK;AACxE,YAAM,OAAO,MAAM;AACnB,UAAI,QAAQ,MAAM;AAChB,aAAK,gBAAgB;AACrB,gBAAQ,KAAK,IAAI;AAAA,MACnB;AAAA,IACF,CAAC;AACD,UAAM,YAAY,KAAK,mBAAmB,WAAW;AACrD,WAAO,EAAE,OAAO,QAAQ,SAAS,SAAS,cAAc,UAAU;AAAA,EACpE;AAAA,EACA,mBAAmB,aAAa;AAC9B,WAAO;AAAA,MACL,YAAY,YAAY,UAAU;AAAA,MAClC,QAAQ,YAAY,UAAU,SAAS,OAAO,CAAC,KAAK,QAAQ;AAC1D,YAAI,IAAI,QAAQ,KAAK,mBAAmB,GAAG;AAC3C,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,MACL,SAAS,YAAY,UAAU,UAAU,OAAO,CAAC,KAAK,QAAQ;AAC5D,YAAI,IAAI,QAAQ,KAAK,mBAAmB,KAAK,YAAY,GAAG;AAC5D,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AAAA,EACF;AAAA,EACA,mBAAmB,KAAK,UAAU;AAChC,QAAI,OAAO,IAAI;AACf,QAAI,YAAY,MAAM;AACpB,aAAO,SAAS;AAAA,IAClB;AACA,WAAO,EAAE,MAAM,OAAO,IAAI,KAAK;AAAA,EACjC;AACF;AACA,SAAS,aAAa,MAAM;AAC1B,QAAM,UAAU,IAAI,EAAE;AACtB,MAAI,OAAO,QAAQ,SAAS,aAAa;AACvC,WAAO,QAAQ,KAAK,IAAI;AAAA,EAC1B,WAAW,OAAO,WAAW,aAAa;AACxC,WAAO,IAAI,OAAO,MAAM,QAAQ,EAAE,SAAS;AAAA,EAC7C,OAAO;AACL,UAAM,IAAI,MAAM,kFAAkF;AAAA,EACpG;AACF;AACA,SAAS,iBAAiB,GAAG,UAAU;AACrC,QAAM,QAAQ,MAAM,QAAQ,CAAC,IAAI,OAAO,aAAa,MAAM,MAAM,CAAC,IAAI,aAAa,CAAC;AACpF,SAAO,WAAW,QAAQ,MAAM,YAAY;AAC9C;AACA,SAAS,eAAe,OAAO,MAAM,KAAK,WAAW,OAAO;AAC1D,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM;AACjB,WAAO,iBAAiB,MAAM,GAAG,QAAQ;AAAA,EAC3C;AACA,SAAO;AACT;AACA,SAAS,aAAa,OAAO,MAAM,KAAK;AACtC,QAAM,QAAQ,MAAM;AACpB,SAAO,QAAQ,MAAM,IAAI;AAC3B;AACA,SAAS,eAAe,OAAO,MAAM,KAAK;AACxC,QAAM,QAAQ,MAAM,SAAS,CAAC;AAC9B,QAAM,QAAQ,MAAM,QAAQ,OAAO,MAAM,OAAO,MAAM,QAAQ,OAAO,MAAM,OAAO;AAClF,SAAO,OAAO,UAAU,WAAW,QAAQ,SAAS,OAAO,EAAE;AAC/D;AACA,SAAS,gBAAgB,OAAO;AAC9B,MAAI,OAAO,UAAU,UAAU;AAC7B,YAAQ,SAAS;AAAA,EACnB;AACA,UAAQ,OAAO;AAAA,IACb,KAAK,SAAS;AAAA,IACd,KAAK,SAAS;AACZ,aAAO;AAAA,IACT,KAAK,SAAS;AAAA,IACd,KAAK,SAAS;AAAA,IACd,KAAK,SAAS;AAAA,IACd,KAAK,SAAS;AACZ,aAAO;AAAA,IACT,KAAK,SAAS;AACZ,aAAO;AAAA,IACT,KAAK,SAAS;AACZ,aAAO;AAAA,IACT,KAAK,SAAS;AACZ,aAAO;AAAA,IACT;AACE,aAAO;AAAA,EACX;AACF;AACA,SAAS,aAAa,OAAO,MAAM,KAAK;AACtC,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,MAAM;AACvB,WAAO,MAAM,KAAK;AAAA,EACpB;AACA,SAAO;AACT;AACA,SAAS,cAAc,OAAO,MAAM,KAAK;AACvC,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,MAAM;AACvB,WAAO,gBAAgB,MAAM,IAAI;AAAA,EACnC;AACA,SAAO;AACT;AACA,SAAS,mBAAmB,OAAO,MAAM,KAAK;AAC5C,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,QAAQ,MAAM,KAAK,MAAM;AAC1C,WAAO,MAAM,KAAK,KAAK,IAAI,CAAC,MAAM,gBAAgB,CAAC,CAAC;AAAA,EACtD;AACA,SAAO;AACT;AACA,SAAS,sBAAsB,OAAO;AACpC,MAAI,MAAM,aAAa;AACrB,WAAO;AAAA,EACT;AACA,MAAI,MAAM,OAAO,MAAM;AACrB,WAAO,MAAM,IAAI,IAAI,CAAC,QAAQ,OAAO,IAAI,SAAS,WAAW,IAAI,OAAO,SAAS,IAAI,MAAM,EAAE,CAAC;AAAA,EAChG;AACA,SAAO,CAAC;AACV;AACA,SAAS,oBAAoB,OAAO,MAAM,KAAK;AAC7C,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,OAAO;AACxB,WAAO,sBAAsB,MAAM,KAAK;AAAA,EAC1C;AACA,SAAO;AACT;AACA,SAAS,qBAAqB,OAAO,MAAM,KAAK;AAC9C,QAAM,QAAQ,MAAM;AACpB,MAAI,OAAO;AACT,aAAS,MAAM,KAAK,KAAK,MAAM,KAAK,EAAE,SAAS,MAAM,KAAK,IAAI,MAAM,KAAK,MAAM,CAAC,GAAG,IAAI,CAAC,MAAM,OAAO,MAAM,WAAW,IAAI,SAAS,GAAG,EAAE,CAAC;AAAA,EAC3I;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,OAAO,MAAM,KAAK,WAAW,OAAO;AAC/D,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,QAAQ,MAAM,KAAK,GAAG;AACvC,WAAO,MAAM,KAAK,EAAE,IAAI,CAAC,MAAM;AAC7B,aAAO,iBAAiB,GAAG,QAAQ;AAAA,IACrC,CAAC;AAAA,EACH;AACA,SAAO;AACT;AACA,SAAS,yBAAyB,OAAO,MAAM,KAAK;AAClD,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,QAAQ,MAAM,KAAK,OAAO;AAC3C,WAAO,MAAM,KAAK,MAAM,IAAI,CAAC,MAAM;AACjC,aAAO,sBAAsB,CAAC;AAAA,IAChC,CAAC;AAAA,EACH;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,OAAO,MAAM,KAAK;AAC3C,QAAM,QAAQ,MAAM;AACpB,MAAI,SAAS,MAAM,QAAQ,MAAM,KAAK,GAAG;AACvC,WAAO,MAAM,KAAK;AAAA,EACpB;AACA,SAAO;AACT;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,MAAM,WAAW,SAAS;AACpC,SAAK,OAAO;AACZ,SAAK,YAAY;AACjB,SAAK,UAAU;AACf,SAAK,SAAS,CAAC;AACf,SAAK,QAAQ,CAAC;AACd,SAAK,SAAS,KAAK,WAAW,IAAI,CAAC,SAAS,KAAK,SAAS,IAAI,CAAC;AAC/D,QAAI,KAAK,YAAY,MAAM;AACzB,WAAK,QAAQ,OAAO,KAAK,KAAK,QAAQ,EAAE,OAAO,CAAC,OAAO,QAAQ;AAC7D,cAAM,OAAO,KAAK,QAAQ,GAAG;AAC7B,eAAO;AAAA,MACT,GAAG,CAAC,CAAC;AAAA,IACP;AAAA,EACF;AAAA,EACA,SAAS,MAAM;AACb,WAAO,UAAU,MAAM,KAAK,WAAW,KAAK,OAAO;AAAA,EACrD;AAAA,EACA,QAAQ,MAAM,cAAc;AAC1B,UAAM,QAAQ,KAAK,KAAK,SAAS;AACjC,QAAI,MAAM,UAAU,MAAM;AACxB,aAAO,UAAU,MAAM,KAAK,WAAW,KAAK,OAAO;AAAA,IACrD;AACA,QAAI,MAAM,KAAK,QAAQ,MAAM,KAAK,MAAM;AACtC,aAAO,eAAe,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,IAC9D;AACA,QAAI,MAAM,KAAK,MAAM;AACnB,aAAO,eAAe,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,IAC9D;AACA,QAAI,MAAM,KAAK,MAAM;AACnB,aAAO,aAAa,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,IAC5D;AACA,QAAI,MAAM,SAAS,MAAM;AACvB,aAAO,oBAAoB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,IACnE;AACA,QAAI,MAAM,QAAQ,MAAM;AACtB,aAAO,cAAc,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,IAC7D;AACA,QAAI,MAAM,QAAQ,MAAM;AACtB,UAAI,MAAM,KAAK,KAAK,QAAQ,MAAM,KAAK,KAAK,MAAM;AAChD,eAAO,qBAAqB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,MACpE;AACA,UAAI,MAAM,KAAK,KAAK,MAAM;AACxB,eAAO,oBAAoB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,MACnE;AACA,UAAI,MAAM,KAAK,SAAS,MAAM;AAC5B,eAAO,yBAAyB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,MACxE;AACA,UAAI,MAAM,KAAK,KAAK,MAAM;AACxB,eAAO,kBAAkB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,MACjE;AACA,UAAI,MAAM,KAAK,QAAQ,MAAM;AAC3B,eAAO,mBAAmB,KAAK,KAAK,UAAU,MAAM,YAAY;AAAA,MAClE;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;AAGA,IAAI,4BAA4B,CAAC;AACjCA,UAAS,2BAA2B;AAAA,EAClC,iBAAiB,MAAM;AAAA,EACvB,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,KAAK,MAAM;AAAA,EACX,KAAK,MAAM;AAAA,EACX,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,eAAe,MAAM;AAAA,EACrB,WAAW,MAAM;AAAA,EACjB,aAAa,MAAM;AAAA,EACnB,aAAa,MAAM;AAAA,EACnB,aAAa,MAAM;AAAA,EACnB,gBAAgB,MAAM;AAAA,EACtB,UAAU,MAAM;AAAA,EAChB,kBAAkB,MAAM;AAAA,EACxB,eAAe,MAAM;AAAA,EACrB,aAAa,MAAM;AAAA,EACnB,QAAQ,MAAM;AAAA,EACd,MAAM,MAAM;AAAA,EACZ,MAAM,MAAM;AAAA,EACZ,aAAa,MAAM;AAAA,EACnB,OAAO,MAAM;AAAA,EACb,SAAS,MAAM;AAAA,EACf,QAAQ,MAAM;AAAA,EACd,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,QAAQ,MAAM;AAAA,EACd,QAAQ,MAAM;AAAA,EACd,iBAAiB,MAAM;AAAA,EACvB,QAAQ,MAAM;AAAA,EACd,iBAAiB,MAAM;AAAA,EACvB,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,cAAc,MAAM;AAAA,EACpB,SAAS,MAAM;AAAA,EACf,QAAQ,MAAM;AAAA,EACd,eAAe,MAAM;AAAA,EACrB,cAAc,MAAM;AAAA,EACpB,iBAAiB,MAAM;AAAA,EACvB,MAAM,MAAM;AAAA,EACZ,YAAY,MAAM;AAAA,EAClB,KAAK,MAAM;AAAA,EACX,UAAU,MAAM;AAAA,EAChB,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,QAAQ,MAAM;AAAA,EACd,KAAK,MAAM;AAAA,EACX,qBAAqB,MAAM;AAAA,EAC3B,OAAO,MAAM;AAAA,EACb,KAAK,MAAM;AAAA,EACX,eAAe,MAAM;AAAA,EACrB,KAAK,MAAM;AAAA,EACX,YAAY,MAAM;AAAA,EAClB,OAAO,MAAM;AAAA,EACb,KAAK,MAAM;AAAA,EACX,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,UAAU,MAAM;AAAA,EAChB,OAAO,MAAM;AAAA,EACb,QAAQ,MAAM;AAAA,EACd,UAAU,MAAM;AAAA,EAChB,SAAS,MAAM;AAAA,EACf,cAAc,MAAM;AAAA,EACpB,MAAM,MAAM;AAAA,EACZ,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,aAAa,MAAM;AAAA,EACnB,OAAO,MAAM;AAAA,EACb,UAAU,MAAM;AAAA,EAChB,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,WAAW,MAAM;AAAA,EACjB,MAAM,MAAM;AAAA,EACZ,WAAW,MAAM;AAAA,EACjB,QAAQ,MAAM;AAAA,EACd,UAAU,MAAM;AAAA,EAChB,4BAA4B,MAAM;AAAA,EAClC,KAAK,MAAM;AAAA,EACX,OAAO,MAAM;AAAA,EACb,YAAY,MAAM;AAAA,EAClB,YAAY,MAAM;AAAA,EAClB,WAAW,MAAM;AAAA,EACjB,YAAY,MAAM;AAAA,EAClB,YAAY,MAAM;AAAA,EAClB,WAAW,MAAM;AAAA,EACjB,YAAY,MAAM;AAAA,EAClB,QAAQ,MAAM;AAAA,EACd,YAAY,MAAM;AAAA,EAClB,QAAQ,MAAM;AAAA,EACd,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,mBAAmB,MAAM;AAAA,EACzB,SAAS,MAAM;AAAA,EACf,MAAM,MAAM;AAAA,EACZ,UAAU,MAAM;AAAA,EAChB,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,KAAK,MAAM;AAAA,EACX,SAAS,MAAM;AAAA,EACf,eAAe,MAAM;AAAA,EACrB,KAAK,MAAM;AAAA,EACX,cAAc,MAAM;AAAA,EACpB,aAAa,MAAM;AAAA,EACnB,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,UAAU,MAAM;AAAA,EAChB,QAAQ,MAAM;AAAA,EACd,MAAM,MAAM;AAAA,EACZ,UAAU,MAAM;AAAA,EAChB,IAAI,MAAM;AAAA,EACV,cAAc,MAAM;AAAA,EACpB,KAAK,MAAM;AAAA,EACX,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,KAAK,MAAM;AAAA,EACX,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,cAAc,MAAM;AAAA,EACpB,aAAa,MAAM;AAAA,EACnB,sBAAsB,MAAM;AAAA,EAC5B,MAAM,MAAM;AAAA,EACZ,aAAa,MAAM;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,sBAAsB,MAAM;AAAA,EAC5B,eAAe,MAAM;AAAA,EACrB,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,YAAY,MAAM;AAAA,EAClB,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,OAAO,MAAM;AAAA,EACb,QAAQ,MAAM;AAAA,EACd,WAAW,MAAM;AAAA,EACjB,cAAc,MAAM;AAAA,EACpB,MAAM,MAAM;AAAA,EACZ,iBAAiB,MAAM;AAAA,EACvB,gBAAgB,MAAM;AAAA,EACtB,SAAS,MAAM;AAAA,EACf,MAAM,MAAM;AAAA,EACZ,QAAQ,MAAM;AAAA,EACd,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,OAAO,MAAM;AAAA,EACb,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,UAAU,MAAM;AAAA,EAChB,gBAAgB,MAAM;AAAA,EACtB,QAAQ,MAAM;AAAA,EACd,eAAe,MAAM;AAAA,EACrB,UAAU,MAAM;AAAA,EAChB,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,QAAQ,MAAM;AAAA,EACd,mBAAmB,MAAM;AAAA,EACzB,SAAS,MAAM;AAAA,EACf,OAAO,MAAM;AAAA,EACb,MAAM,MAAM;AAAA,EACZ,cAAc,MAAM;AAAA,EACpB,QAAQ,MAAM;AAAA,EACd,KAAK,MAAM;AAAA,EACX,KAAK,MAAM;AAAA,EACX,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,QAAQ,MAAM;AAAA,EACd,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,MAAM,MAAM;AAAA,EACZ,MAAM,MAAM;AAAA,EACZ,WAAW,MAAM;AAAA,EACjB,iBAAiB,MAAM;AAAA,EACvB,QAAQ,MAAM;AAAA,EACd,oBAAoB,MAAM;AAAA,EAC1B,SAAS,MAAM;AAAA,EACf,YAAY,MAAM;AAAA,EAClB,UAAU,MAAM;AAAA,EAChB,OAAO,MAAM;AAAA,EACb,YAAY,MAAM;AAAA,EAClB,OAAO,MAAM;AAAA,EACb,WAAW,MAAM;AACnB,CAAC;AAGD,IAAI,YAAY,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC7E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,WAAW,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE;AAAA,IACA,KAAK;AAAA,IACL,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G,KAAK;AAAA,IACL,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G;AAAA,IACA,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAClH;AAAA,IACA,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAClH;AAAA,IACA,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G;AAAA,IACA,KAAK,WAAW;AACd,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjH;AAAA,IACA,KAAK,WAAW;AACd,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjH;AAAA,IACA,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7G;AAAA,IACA,KAAK,qBAAqB;AACxB,aAAO,CAAC,IAAI,kBAAkB,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC3H;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,aAAa,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC9E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/G,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,QAAQ,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACvH,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE;AAAA,IACA,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE;AAAA,IACA,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK,cAAc;AACjB,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE;AAAA,IACA,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE;AAAA,IACA,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE,KAAK;AACH,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACnE,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACpE;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,UAAU;AACb,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAClE;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK;AACH,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D,KAAK;AACH,aAAO,CAAC,IAAI,YAAY,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,gBAAgB,MAAM,WAAW,OAAO,GAAG,cAAc,gBAAgB,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACzL,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,UAAU,KAAK,WAAW,IAAI,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE,KAAK;AACH,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjH,KAAK;AACH,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACvH,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACnH,KAAK;AACH,aAAO,CAAC,IAAI,MAAM,UAAU,KAAK,WAAW,IAAI,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,SAAS,oCAAoC,QAAQ,QAAQ,qBAAqB,IAAI;AACpF,MAAI,OAAO,WAAW,YAAY,OAAO,WAAW,UAAU;AAC5D;AAAA,EACF;AACA,eAAa,OAAO,OAAO,WAAW,OAAO,QAAQ,MAAM,qBAAqB,WAAW,cAAc,mBAAmB;AAC5H,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,UAAM,OAAO,OAAO;AACpB,UAAM,OAAO,OAAO;AACpB,iBAAa,OAAO,OAAO,KAAK,OAAO,KAAK,SAAS,MAAM,MAAM,qBAAqB,WAAW,cAAc,mBAAmB;AAAA,EACpI;AACF;AACA,SAAS,iBAAiB,cAAc;AACtC,MAAI,OAAO,iBAAiB,YAAY,aAAa,KAAK,CAAC,QAAQ,MAAM,CAAC,GAAG;AAC3E,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,kBAAkB,SAAS,cAAc;AAClE,MAAI,eAAe,kBAAkB,kBAAkB,YAAY;AACnE,QAAM,sBAAsB,CAAC,iBAAiB,YAAY;AAC1D,MAAI,uBAAuB,QAAQ,WAAW,GAAG;AAC/C,UAAM,IAAI,MAAM,qFAAqF,cAAc;AAAA,EACrH;AACA,MAAI,qBAAqB;AACvB,YAAQ,QAAQ,CAAC,YAAY;AAC3B,qBAAe,kBAAkB,QAAQ,OAAO,YAAY;AAAA,IAC9D,CAAC;AAAA,EACH;AACA,MAAI,CAAC,iBAAiB,YAAY,GAAG;AACnC,UAAM,IAAI,MAAM,mCAAmC,cAAc;AAAA,EACnE;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,eAAe,eAAe;AACvD,MAAI,OAAO,kBAAkB,UAAU;AACrC,WAAO;AAAA,EACT;AACA,MAAI,OAAO,kBAAkB,UAAU;AACrC,WAAO;AAAA,EACT;AACA,MAAI,cAAc,WAAW,cAAc,QAAQ;AACjD,UAAM,IAAI,MAAM,oCAAoC,qBAAqB,eAAe;AAAA,EAC1F;AACA,QAAM,SAAS,CAAC;AAChB,WAAS,IAAI,GAAG,IAAI,cAAc,QAAQ,EAAE,GAAG;AAC7C,UAAM,OAAO,cAAc;AAC3B,UAAM,OAAO,cAAc;AAC3B,QAAI,QAAQ,KAAK,QAAQ,KAAK,SAAS,MAAM;AAC3C,YAAM,IAAI,MAAM,oCAAoC,qBAAqB,eAAe;AAAA,IAC1F;AACA,WAAO,KAAK,QAAQ,IAAI,OAAO;AAAA,EACjC;AACA,SAAO;AACT;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,MAAM,OAAO,SAAS,cAAc,wBAAwB,aAAa,gBAAgB;AACnG,SAAK,OAAO;AACZ,SAAK,QAAQ;AACb,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,yBAAyB;AAC9B,SAAK,cAAc;AACnB,SAAK,iBAAiB;AACtB,SAAK,UAAU,CAAC;AAChB,SAAK,UAAU;AACf,SAAK,WAAW,OAAO,CAAC;AACxB,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EACA,IAAI,KAAK;AACP,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,SAAS;AACX,WAAO,KAAK;AAAA,EACd;AAAA,EACA,cAAc,SAAS;AACrB,SAAK,QAAQ,QAAQ,CAAC,YAAY;AAChC,UAAI,WAAW,QAAQ,CAAC,QAAQ,IAAI,QAAQ,OAAO,EAAE,GAAG;AACtD,gBAAQ,OAAO,QAAQ;AAAA,MACzB;AAAA,IACF,CAAC;AACD,SAAK,UAAU,CAAC;AAChB,SAAK,UAAU;AACf,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EACA,OAAO;AACL,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EACA,KAAK,OAAO;AACV,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,eAAe,KAAK,+BAA+B;AAAA,IACrE;AACA,QAAI,QAAQ,KAAK,SAAS,KAAK,KAAK,GAAG;AACrC,YAAM,IAAI,MAAM,4BAA4B,6BAA6B,KAAK,KAAK,GAAG;AAAA,IACxF;AACA,UAAM,kBAAkB,KAAK,QAAQ;AACrC,QAAI,gBAAgB,SAAS;AAC3B,YAAM,IAAI,MAAM,eAAe,KAAK,8BAA8B,2GAA2G;AAAA,IAC/K;AACA,QAAI,KAAK,gBAAgB;AACvB,sBAAgB,UAAU;AAAA,IAC5B;AACA,oBAAgB,OAAO;AACvB,WAAO,gBAAgB;AAAA,EACzB;AAAA,EACA,SAAS,SAAS;AAChB,WAAO,QAAQ,IAAI,CAAC,UAAU,KAAK,KAAK,KAAK,CAAC;AAAA,EAChD;AAAA,EACA,MAAM,OAAO,SAAS;AACpB,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,eAAe,KAAK,+BAA+B;AAAA,IACrE;AACA,QAAI,QAAQ,KAAK,CAAC,KAAK,eAAe,SAAS,KAAK,SAAS;AAC3D,YAAM,IAAI,MAAM,2BAA2B,mDAAmD,KAAK,SAAS;AAAA,IAC9G;AACA,UAAM,IAAI,KAAK,QAAQ,UAAU,CAAC;AAClC,QAAI,QAAQ,UAAU,KAAK,OAAO;AAChC,YAAM,IAAI,MAAM,eAAe,KAAK,8CAA8C;AAAA,uCACjD,QAAQ,mCAAmC,KAAK,QAAQ;AAAA,IAC3F;AACA,QAAI,KAAK,KAAK,MAAM,MAAM,KAAK,gBAAgB,QAAQ,KAAK,aAAa,WAAW,IAAI;AACtF,WAAK,eAAe,QAAQ;AAAA,IAC9B;AACA,wCAAoC,KAAK,cAAc,QAAQ,OAAO,eAAe,KAAK,8CAA8C,QAAQ;AAChJ,QAAI,EAAE,MAAM;AACV,YAAM,IAAI,MAAM,eAAe,KAAK,8CAA8C,0CAA0C;AAAA,IAC9H;AACA,QAAI,EAAE,SAAS;AACb,YAAM,IAAI,MAAM,eAAe,KAAK,8CAA8C,6CAA6C;AAAA,IACjI;AACA,MAAE,SAAS;AACX,SAAK,OAAO;AACZ,MAAE,UAAU;AACZ,SAAK,QAAQ,SAAS;AAAA,EACxB;AAAA,EACA,UAAU,SAAS,SAAS;AAC1B,QAAI,QAAQ,WAAW,QAAQ,QAAQ;AACrC,YAAM,IAAI,MAAM,eAAe,KAAK,kEAAkE,QAAQ,2CAA2C,QAAQ,SAAS;AAAA,IAC5K;AACA,YAAQ,QAAQ,CAAC,GAAG,UAAU,KAAK,MAAM,GAAG,QAAQ,MAAM,CAAC;AAAA,EAC7D;AAAA,EACA,OAAO,SAAS,OAAO;AACrB,QAAI,CAAC,CAAC,SAAS,UAAU,KAAK,OAAO;AACnC,YAAM,IAAI,MAAM,wBAAwB,KAAK,oCAAoC,OAAO;AAAA,IAC1F;AACA,QAAI,CAAC,SAAS;AACZ,gBAAU,CAAC;AACX,eAAS,IAAI,GAAG,IAAI,KAAK,KAAK,GAAG,KAAK;AACpC,gBAAQ,KAAK,CAAC;AAAA,MAChB;AAAA,IACF,OAAO;AACL,gBAAU,QAAQ,MAAM,GAAG,KAAK,KAAK,CAAC;AAAA,IACxC;AACA,QAAI,QAAQ,WAAW,GAAG;AACxB,aAAO,OAAO,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,KAAK,YAAY,CAAC;AAAA,IACjD;AACA,UAAM,UAAU,KAAK,SAAS,OAAO;AACrC,wCAAoC,KAAK,cAAc,QAAQ,GAAG,OAAO,8BAA8B;AACvG,WAAO,MAAM,SAAS,CAAC;AAAA,EACzB;AAAA,EACA,OAAO,OAAO;AACZ,QAAI,CAAC,CAAC,SAAS,UAAU,KAAK,OAAO;AACnC,YAAM,IAAI,MAAM,wBAAwB,KAAK,oCAAoC,OAAO;AAAA,IAC1F;AACA,QAAI,KAAK,KAAK,MAAM,GAAG;AACrB,aAAO,OAAO,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,KAAK,YAAY,CAAC;AAAA,IACjD;AACA,UAAM,UAAU,CAAC;AACjB,aAAS,IAAI,GAAG,IAAI,KAAK,KAAK,GAAG,KAAK;AACpC,cAAQ,KAAK,CAAC;AAAA,IAChB;AACA,UAAM,UAAU,KAAK,SAAS,OAAO;AACrC,wCAAoC,KAAK,cAAc,QAAQ,GAAG,OAAO,mDAAmD,KAAK,wCAAwC,QAAQ,GAAG,QAAQ;AAC5L,WAAO,OAAO,SAAS,CAAC;AAAA,EAC1B;AAAA,EACA,QAAQ,SAAS,SAAS;AACxB,QAAI,QAAQ,UAAU,KAAK,OAAO;AAChC,YAAM,IAAI,MAAM,wBAAwB,KAAK,8BAA8B,QAAQ,OAAO;AAAA,IAC5F;AACA,QAAI,QAAQ,WAAW,QAAQ,MAAM,IAAI;AACvC,YAAM,IAAI,MAAM,sDAAsD,QAAQ,cAAc,QAAQ,MAAM,IAAI;AAAA,IAChH;AACA,UAAM,WAAW,KAAK,IAAI,GAAG,OAAO;AACpC,QAAI,CAAC,KAAK,eAAe,YAAY,KAAK,SAAS;AACjD,YAAM,IAAI,MAAM,mCAAmC,iBAAiB,KAAK,UAAU;AAAA,IACrF;AACA,SAAK,UAAU,SAAS,QAAQ,SAAS,CAAC,CAAC;AAAA,EAC7C;AAAA,EACA,MAAM,QAAQ,SAAS;AACrB,QAAI,QAAQ,UAAU,KAAK,OAAO;AAChC,YAAM,IAAI,MAAM,wBAAwB,KAAK,8BAA8B,QAAQ,OAAO;AAAA,IAC5F;AACA,QAAI,cAAc;AAClB,UAAM,oBAAoB,OAAO,IAAI,CAAC,QAAQ;AAC5C,qBAAe;AACf,aAAO;AAAA,IACT,CAAC;AACD,QAAI,gBAAgB,QAAQ,MAAM,IAAI;AACpC,YAAM,IAAI,MAAM;AAAA;AAAA,UAEZ,uCAAuC,QAAQ,OAAO;AAAA,IAC5D;AACA,QAAI,CAAC,KAAK,eAAe,OAAO,WAAW,KAAK,SAAS;AACvD,YAAM,IAAI,MAAM,2DAA2D,KAAK,eAAe,OAAO,sEAAsE;AAAA,IAC9K;AACA,UAAM,gBAAgB,gBAAgB,IAAI,IAAI,QAAQ,OAAO;AAC7D,UAAM,UAAU,CAAC;AACjB,SAAK,MAAM;AACT,gBAAU,QAAQ,SAAS,CAAC,GAAG,aAAa,aAAa,CAAC;AAC1D,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,cAAM,iBAAiB,MAAM,IAAI,IAAI,kBAAkB,IAAI;AAC3D,cAAM,WAAW,CAAC,GAAG,gBAAgB,CAAC;AACtC,cAAM,QAAQ,CAAC,GAAG,OAAO,IAAI,aAAa;AAC1C,gBAAQ,KAAK,QAAQ,MAAM,SAAS,UAAU,KAAK,GAAG,KAAK,YAAY;AAAA,MACzE;AACA,aAAO;AAAA,IACT,CAAC;AACD,UAAM,UAAU,CAAC;AACjB,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,cAAQ,KAAK;AAAA,IACf;AACA,SAAK,UAAU,SAAS,OAAO;AAAA,EACjC;AACF;AAGA,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,SAAS,cAAc,cAAc,iBAAiB,IAAI;AACpE,SAAK,UAAU;AACf,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,QAAI,WAAW,MAAM;AACnB,cAAQ,QAAQ,CAAC,YAAY;AAC3B,YAAI,iBAAiB,QAAQ,OAAO;AAClC,gBAAM,IAAI,MAAM,mCAAmC,mCAAmC,QAAQ,OAAO;AAAA,QACvG;AACA,4CAAoC,cAAc,QAAQ,OAAO,6BAA6B;AAC9F,aAAK,OAAO;AAAA,MACd,CAAC;AAAA,IACH;AACA,SAAK,WAAW,OAAO,CAAC;AACxB,SAAK,iBAAiB;AACtB,SAAK,KAAK,QAAQ;AAAA,EACpB;AAAA,EACA,IAAI,KAAK;AACP,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,OAAO;AACL,WAAO,IAAI,WAAW,CAAC,GAAG,KAAK,OAAO,GAAG,KAAK,cAAc,KAAK,YAAY;AAAA,EAC/E;AAAA,EACA,cAAc,SAAS;AACrB,SAAK,QAAQ,QAAQ,CAAC,YAAY;AAChC,UAAI,WAAW,QAAQ,CAAC,QAAQ,IAAI,QAAQ,EAAE,GAAG;AAC/C,gBAAQ,QAAQ;AAAA,MAClB;AAAA,IACF,CAAC;AACD,SAAK,QAAQ,SAAS;AACtB,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EACA,OAAO;AACL,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EACA,MAAM,cAAc,cAAc,cAAc,IAAI;AAClD,QAAI,iBAAiB,KAAK,cAAc;AACtC,YAAM,IAAI,MAAM,mCAAmC,mCAAmC,KAAK,cAAc;AAAA,IAC3G;AACA,QAAI,gBAAgB,MAAM,KAAK,QAAQ,WAAW,aAAa;AAC7D,YAAM,IAAI,MAAM,kCAAkC,4CAA4C,KAAK,QAAQ,kBAAkB;AAAA,IAC/H;AACA,wCAAoC,cAAc,KAAK,cAAc,6BAA6B;AAClG,UAAM,qBAAqB,kBAAkB,KAAK,cAAc,KAAK,SAAS,YAAY;AAC1F,WAAO,KAAK,MAAM;AAChB,YAAM,kBAAkB,KAAK,QAAQ,IAAI,CAAC,YAAY,QAAQ,SAAS,kBAAkB,CAAC;AAC1F,aAAO,MAAM,iBAAiB,CAAC;AAAA,IACjC,CAAC;AAAA,EACH;AAAA,EACA,QAAQ,cAAc,cAAc;AAClC,QAAI,iBAAiB,KAAK,cAAc;AACtC,YAAM,IAAI,MAAM,mCAAmC,mCAAmC,KAAK,cAAc;AAAA,IAC3G;AACA,QAAI,KAAK,KAAK,MAAM,GAAG;AACrB,YAAM,IAAI,MAAM,mCAAmC;AAAA,IACrD;AACA,UAAM,qBAAqB,kBAAkB,KAAK,cAAc,KAAK,SAAS,YAAY;AAC1F,UAAM,UAAU,KAAK,QAAQ,IAAI;AACjC,YAAQ,OAAO;AACf,wCAAoC,QAAQ,OAAO,cAAc,6BAA6B;AAC9F,WAAO,QAAQ,SAAS,kBAAkB;AAAA,EAC5C;AAAA,EACA,SAAS,SAAS;AAChB,QAAI,QAAQ,UAAU,KAAK,cAAc;AACvC,YAAM,IAAI,MAAM,mCAAmC,QAAQ,4BAA4B,KAAK,cAAc;AAAA,IAC5G;AACA,wCAAoC,QAAQ,OAAO,KAAK,cAAc,6BAA6B;AACnG,QAAI,KAAK,mBAAmB,KAAK,KAAK,GAAG;AACvC,YAAM,IAAI,MAAM,0CAA0C;AAAA,IAC5D;AACA,SAAK,OAAO;AACZ,SAAK,QAAQ,KAAK,OAAO;AAAA,EAC3B;AAAA,EACA,OAAO,MAAM;AACX,QAAI,OAAO,GAAG;AACZ,YAAM,IAAI,MAAM,0DAA0D,MAAM;AAAA,IAClF;AACA,QAAI,KAAK,mBAAmB,MAAM,OAAO,KAAK,gBAAgB;AAC5D,YAAM,IAAI,MAAM,+BAA+B,iCAAiC,KAAK,iBAAiB;AAAA,IACxG;AACA,UAAM,iBAAiB,IAAI,WAAW,CAAC,GAAG,KAAK,cAAc,KAAK,cAAc,KAAK,cAAc;AACnG,mBAAe,QAAQ,SAAS;AAChC,aAAS,IAAI,GAAG,IAAI,KAAK,IAAI,KAAK,QAAQ,QAAQ,IAAI,GAAG,EAAE,GAAG;AAC5D,qBAAe,QAAQ,KAAK,KAAK,QAAQ;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAAA,EACA,QAAQ,cAAc,cAAc,cAAc;AAChD,QAAI,iBAAiB,KAAK,cAAc;AACtC,YAAM,IAAI,MAAM,mCAAmC,mCAAmC,KAAK,cAAc;AAAA,IAC3G;AACA,QAAI,eAAe,KAAK,eAAe,KAAK,QAAQ,QAAQ;AAC1D,YAAM,IAAI,MAAM,4BAA4B,+BAA+B,KAAK,QAAQ,kBAAkB;AAAA,IAC5G;AACA,QAAI,KAAK,QAAQ,iBAAiB,MAAM;AACtC,YAAM,IAAI,MAAM,oBAAoB,uBAAuB;AAAA,IAC7D;AACA,wCAAoC,KAAK,QAAQ,cAAc,OAAO,cAAc,6BAA6B;AACjH,UAAM,qBAAqB,kBAAkB,KAAK,cAAc,KAAK,SAAS,YAAY;AAC1F,WAAO,QAAQ,KAAK,QAAQ,eAAe,kBAAkB;AAAA,EAC/D;AAAA,EACA,QAAQ,cAAc,SAAS;AAC7B,QAAI,QAAQ,UAAU,KAAK,cAAc;AACvC,YAAM,IAAI,MAAM,mCAAmC,QAAQ,4BAA4B,KAAK,cAAc;AAAA,IAC5G;AACA,QAAI,eAAe,KAAK,KAAK,mBAAmB,MAAM,gBAAgB,KAAK,gBAAgB;AACzF,YAAM,IAAI,MAAM,yBAAyB,mCAAmC,KAAK,0BAA0B;AAAA,IAC7G;AACA,wCAAoC,KAAK,cAAc,QAAQ,OAAO,6BAA6B;AACnG,SAAK,OAAO;AACZ,QAAI,KAAK,QAAQ,iBAAiB,MAAM;AACtC,WAAK,QAAQ,cAAc,OAAO;AAAA,IACpC;AACA,SAAK,QAAQ,gBAAgB;AAAA,EAC/B;AAAA,EACA,OAAO,SAAS,cAAc,cAAc;AAC1C,QAAI,iBAAiB,KAAK,cAAc;AACtC,YAAM,IAAI,MAAM,mCAAmC,mCAAmC,KAAK,cAAc;AAAA,IAC3G;AACA,wCAAoC,KAAK,cAAc,cAAc,6BAA6B;AAClG,cAAU,QAAQ,MAAM,GAAG,KAAK,KAAK,CAAC;AACtC,UAAM,qBAAqB,kBAAkB,KAAK,cAAc,KAAK,SAAS,YAAY;AAC1F,QAAI,QAAQ,WAAW,GAAG;AACxB,aAAO,OAAO,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,kBAAkB,CAAC;AAAA,IAClD;AACA,WAAO,KAAK,MAAM;AAChB,YAAM,UAAU,QAAQ,IAAI,CAAC,MAAM,QAAQ,KAAK,QAAQ,IAAI,kBAAkB,CAAC;AAC/E,aAAO,MAAM,SAAS,CAAC;AAAA,IACzB,CAAC;AAAA,EACH;AAAA,EACA,OAAO,cAAc,cAAc;AACjC,QAAI,CAAC,CAAC,gBAAgB,iBAAiB,KAAK,cAAc;AACxD,YAAM,IAAI,MAAM,uBAAuB,KAAK,2CAA2C,cAAc;AAAA,IACvG;AACA,wCAAoC,KAAK,cAAc,cAAc,6BAA6B;AAClG,UAAM,qBAAqB,kBAAkB,KAAK,cAAc,KAAK,SAAS,YAAY;AAC1F,QAAI,KAAK,KAAK,MAAM,GAAG;AACrB,aAAO,OAAO,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,kBAAkB,CAAC;AAAA,IAClD;AACA,WAAO,KAAK,MAAM;AAChB,YAAM,UAAU,KAAK,QAAQ,IAAI,CAAC,MAAM,QAAQ,GAAG,kBAAkB,CAAC;AACtE,aAAO,OAAO,SAAS,CAAC;AAAA,IAC1B,CAAC;AAAA,EACH;AACF;AACA,SAAS,WAAW,SAAS,cAAc,cAAc;AACvD,QAAM,QAAQ,QAAQ;AACtB,MAAI,QAAQ,MAAM,SAAS,GAAG;AAC5B,UAAM,IAAI,MAAM,oDAAoD,QAAQ,OAAO;AAAA,EACrF;AACA,MAAI,QAAQ,UAAU,cAAc;AAClC,UAAM,IAAI,MAAM,mCAAmC,QAAQ,4BAA4B,cAAc;AAAA,EACvG;AACA,QAAM,qBAAqB,QAAQ,MAAM,MAAM,CAAC;AAChD,sCAAoC,oBAAoB,cAAc,6BAA6B;AACnG,QAAM,aAAa,QAAQ,OAAO;AAClC,SAAO,IAAI,WAAW,YAAY,cAAc,KAAK;AACvD;AACA,SAAS,QAAQ,cAAc,cAAc,aAAa,gBAAgB;AACxE,SAAO,IAAI,WAAW,CAAC,GAAG,cAAc,cAAc,cAAc;AACtE;AACA,SAAS,QAAQ,SAAS,SAAS,cAAc,aAAa;AAC5D,MAAI,QAAQ,WAAW,QAAQ,MAAM,IAAI;AACvC,UAAM,IAAI,MAAM,sDAAsD,QAAQ,cAAc,QAAQ,MAAM,IAAI;AAAA,EAChH;AACA,QAAM,WAAW,KAAK,IAAI,GAAG,OAAO;AACpC,MAAI,eAAe,QAAQ,gBAAgB,MAAM,YAAY,aAAa;AACxE,UAAM,IAAI,MAAM,mCAAmC,iBAAiB,cAAc;AAAA,EACpF;AACA,QAAM,OAAO,IAAI,WAAW,CAAC,GAAG,cAAc,QAAQ,OAAO,WAAW;AACxE,QAAM,UAAU,QAAQ,SAAS,CAAC;AAClC,UAAQ,QAAQ,CAAC,OAAO,UAAU;AAChC,SAAK,QAAQ,OAAO,QAAQ,MAAM;AAAA,EACpC,CAAC;AACD,SAAO;AACT;AACA,SAAS,OAAO,SAAS,QAAQ,cAAc;AAC7C,MAAI,cAAc;AAClB,QAAM,oBAAoB,OAAO,IAAI,CAAC,QAAQ;AAC5C,mBAAe;AACf,WAAO;AAAA,EACT,CAAC;AACD,MAAI,gBAAgB,QAAQ,MAAM,IAAI;AACpC,UAAM,IAAI,MAAM;AAAA;AAAA,UAEV,uCAAuC,QAAQ,OAAO;AAAA,EAC9D;AACA,QAAM,uBAAuB,QAAQ,MAAM,MAAM,CAAC;AAClD,QAAM,qBAAqB,kBAAkB,sBAAsB,YAAY;AAC/E,QAAM,gBAAgB,gBAAgB,IAAI,IAAI,QAAQ,OAAO;AAC7D,QAAM,UAAU,KAAK,MAAM;AACzB,UAAM,WAAW,CAAC;AAClB,cAAU,QAAQ,SAAS,CAAC,GAAG,aAAa,aAAa,CAAC;AAC1D,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,YAAM,iBAAiB,MAAM,IAAI,IAAI,kBAAkB,IAAI;AAC3D,YAAM,UAAU,CAAC,GAAG,gBAAgB,CAAC;AACrC,YAAM,QAAQ,CAAC,GAAG,OAAO,IAAI,aAAa;AAC1C,eAAS,KAAK,QAAQ,MAAM,SAAS,SAAS,KAAK,GAAG,kBAAkB;AAAA,IAC1E;AACA,YAAQ,QAAQ;AAChB,WAAO;AAAA,EACT,CAAC;AACD,QAAM,OAAO,IAAI,WAAW,CAAC,GAAG,cAAc,QAAQ,OAAO,OAAO,MAAM;AAC1E,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,SAAK,QAAQ,GAAG,QAAQ,EAAE;AAAA,EAC5B;AACA,SAAO;AACT;AAGA,IAAI,aAAa,OAAO,MAAM,WAAW,YAAY;AACnD,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK,eAAe;AAClB,YAAM,WAAW,cAAc,cAAc,MAAM,WAAW,OAAO;AACrE,YAAM,WAAW,cAAc,cAAc,MAAM,WAAW,OAAO;AACrE,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,YAAY,MAAM,KAAK,KAAK;AAClC,UAAI,UAAU,IAAI;AAChB,eAAO,QAAQ,YAAY,UAAU,qBAAqB,MAAM,QAAQ,gBAAgB,QAAQ,aAAa;AAAA,MAC/G,OAAO;AACL,eAAO,QAAQ,YAAY,UAAU,qBAAqB,MAAM,QAAQ,gBAAgB,QAAQ,aAAa;AAAA,MAC/G;AAAA,IACF;AAAA,IACA,KAAK;AAAA,IACL,KAAK,kBAAkB;AACrB,YAAM,WAAW,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC/D,YAAM,WAAW,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC/D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,aAAa,MAAM,QAAQ,YAAY,UAAU,qBAAqB,MAAM,QAAQ,gBAAgB,QAAQ,aAAa;AAC/H,YAAM,SAAS,KAAK,IAAI,CAAC,YAAY,QAAQ,EAAE;AAC/C,UAAI,YAAY,MAAM,WAAW,GAAG,KAAK;AACzC,iBAAW,QAAQ,CAAC,YAAY;AAC9B,YAAI,CAAC,QAAQ,QAAQ,OAAO,QAAQ,QAAQ,EAAE,MAAM,IAAI;AACtD,kBAAQ,QAAQ;AAAA,QAClB;AAAA,MACF,CAAC;AACD,UAAI,SAAS;AACb,aAAO,UAAU,IAAI;AACnB,cAAM,aAAa;AACnB,iBAAS,MAAM,QAAQ,YAAY,UAAU,qBAAqB,QAAQ,QAAQ,gBAAgB,QAAQ,aAAa;AACvH,cAAM,YAAY,OAAO,IAAI,CAAC,YAAY,QAAQ,EAAE;AACpD,mBAAW,QAAQ,CAAC,YAAY;AAC9B,cAAI,CAAC,QAAQ,QAAQ,OAAO,QAAQ,QAAQ,EAAE,MAAM,MAAM,UAAU,QAAQ,QAAQ,EAAE,MAAM,IAAI;AAC9F,oBAAQ,QAAQ;AAAA,UAClB;AAAA,QACF,CAAC;AACD,cAAM,cAAc,MAAM,QAAQ,YAAY,UAAU,qBAAqB,QAAQ,QAAQ,gBAAgB,QAAQ,aAAa;AAClI,oBAAY,MAAM,YAAY,GAAG,KAAK;AACtC,oBAAY,QAAQ,CAAC,YAAY;AAC/B,cAAI,CAAC,QAAQ,QAAQ,OAAO,QAAQ,QAAQ,EAAE,MAAM,MAAM,UAAU,QAAQ,QAAQ,EAAE,MAAM,IAAI;AAC9F,oBAAQ,QAAQ;AAAA,UAClB;AAAA,QACF,CAAC;AAAA,MACH;AACA,aAAO;AAAA,IACT;AAAA,IACA,KAAK,YAAY;AACf,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,YAAY,IAAI,CAAC;AAAA,IAC3B;AAAA,IACA,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,UAAI,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AACzD,UAAI,CAAC,KAAK,MAAM;AACd,eAAO,YAAY,IAAI;AAAA,MACzB;AACA,cAAQ,MAAM,KAAK,KAAK,GAAG,KAAK,CAAC,QAAQ,IAAI,IAAI,CAAC,MAAM,MAAM;AAAA,IAChE;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,YAAY,KAAK,WAAW,KAAK,CAAC,SAAS,UAAU,MAAM,WAAW,OAAO,MAAM,MAAM;AAC/F,UAAI,WAAW;AACb,cAAM,OAAO,UAAU,WAAW,WAAW,OAAO;AACpD,eAAO,CAAC,YAAY,IAAI,CAAC;AAAA,MAC3B;AACA,aAAO;AAAA,IACT;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,UAAU,cAAc,aAAa,MAAM,WAAW,OAAO;AACnE,YAAM,OAAO,cAAc,UAAU,MAAM,WAAW,OAAO;AAC7D,cAAQ,WAAW,OAAO;AAC1B,aAAO,CAAC,YAAY,IAAI,CAAC;AAAA,IAC3B;AAAA,IACA,KAAK,QAAQ;AACX,YAAM,OAAO,cAAc,UAAU,MAAM,WAAW,OAAO;AAC7D,cAAQ,UAAU;AAClB,aAAO,CAAC,YAAY,IAAI,CAAC;AAAA,IAC3B;AAAA,IACA,KAAK,iBAAiB;AACpB,YAAM,OAAO,cAAc,UAAU,MAAM,WAAW,OAAO;AAC7D,cAAQ,cAAc;AACtB,aAAO,CAAC,YAAY,IAAI,CAAC;AAAA,IAC3B;AAAA,IACA,KAAK,iBAAiB;AACpB,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,cAAc,cAAc,eAAe,MAAM,WAAW,OAAO;AACzE,YAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,YAAM,yBAAyB,cAAc,0BAA0B,MAAM,WAAW,OAAO;AAC/F,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,cAAc,IAAI,YAAY,MAAM,OAAO,MAAM,cAAc,wBAAwB,aAAa,cAAc;AACxH,cAAQ,eAAe,WAAW;AAClC,aAAO,CAAC,YAAY,UAAU,OAAO,CAAC,CAAC;AAAA,IACzC;AAAA,IACA,KAAK,sBAAsB;AACzB,YAAM,KAAK,cAAc,iBAAiB,MAAM,WAAW,OAAO;AAClE,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,cAAc,cAAc,UAAU,MAAM,WAAW,OAAO;AACpE,YAAM,mBAAmB,QAAQ,eAAe,GAAG,EAAE;AACrD,uBAAiB,MAAM,OAAO,WAAW;AACzC,aAAO,CAAC,iBAAiB,QAAQ;AAAA,IACnC;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,SAAS,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACtE,YAAM,YAAY,cAAc,SAAS,MAAM,WAAW,OAAO;AACjE,YAAM,kBAAkB,QAAQ,eAAe,OAAO,EAAE;AACxD,aAAO,CAAC,gBAAgB,KAAK,SAAS,CAAC;AAAA,IACzC;AAAA,IACA,KAAK,uBAAuB;AAC1B,YAAM,WAAW,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACxE,YAAM,gBAAgB,cAAc,WAAW,MAAM,WAAW,OAAO;AACvE,YAAM,cAAc,cAAc,SAAS,MAAM,WAAW,OAAO;AACnE,YAAM,oBAAoB,QAAQ,eAAe,SAAS,EAAE;AAC5D,aAAO,CAAC,kBAAkB,OAAO,eAAe,WAAW,CAAC;AAAA,IAC9D;AAAA,IACA,KAAK,wBAAwB;AAC3B,YAAM,YAAY,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACzE,YAAM,iBAAiB,cAAc,WAAW,MAAM,WAAW,OAAO;AACxE,YAAM,gBAAgB,cAAc,UAAU,MAAM,WAAW,OAAO;AACtE,YAAM,qBAAqB,QAAQ,eAAe,UAAU,EAAE;AAC9D,yBAAmB,QAAQ,gBAAgB,aAAa;AACxD,aAAO,CAAC,mBAAmB,QAAQ;AAAA,IACrC;AAAA,IACA,KAAK,uBAAuB;AAC1B,YAAM,WAAW,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACxE,YAAM,oBAAoB,QAAQ,eAAe,SAAS,EAAE;AAC5D,YAAM,cAAc,cAAc,SAAS,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,kBAAkB,OAAO,WAAW,CAAC;AAAA,IAC/C;AAAA,IACA,KAAK,sBAAsB;AACzB,YAAM,UAAU,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACvE,YAAM,cAAc,cAAc,UAAU,MAAM,WAAW,OAAO;AACpE,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,mBAAmB,QAAQ,eAAe,QAAQ,EAAE;AAC1D,uBAAiB,MAAM,SAAS,WAAW;AAC3C,aAAO,CAAC,iBAAiB,QAAQ;AAAA,IACnC;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,SAAS,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACtE,YAAM,kBAAkB,QAAQ,eAAe,OAAO,EAAE;AACxD,aAAO,CAAC,OAAO,gBAAgB,KAAK,GAAG,OAAO,CAAC;AAAA,IACjD;AAAA,IACA,KAAK,sBAAsB;AACzB,YAAM,UAAU,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACvE,YAAM,mBAAmB,QAAQ,eAAe,QAAQ,EAAE;AAC1D,uBAAiB,cAAc;AAC/B,aAAO,CAAC,iBAAiB,QAAQ;AAAA,IACnC;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,cAAc,cAAc,UAAU,MAAM,WAAW,OAAO;AACpE,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,iBAAW,QAAQ,OAAO,WAAW;AACrC,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,YAAY,cAAc,SAAS,MAAM,WAAW,OAAO;AACjE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,aAAO,CAAC,WAAW,QAAQ,WAAW,cAAc,YAAY,CAAC;AAAA,IACnE;AAAA,IACA,KAAK;AAAA,IACL,KAAK,qBAAqB;AACxB,YAAM,iBAAiB,cAAc,WAAW,MAAM,WAAW,OAAO;AACxE,YAAM,gBAAgB,cAAc,UAAU,MAAM,WAAW,OAAO;AACtE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,cAAc,cAAc,eAAe,MAAM,WAAW,OAAO;AACzE,YAAM,aAAa,QAAQ,eAAe,gBAAgB,cAAc,WAAW;AACnF,cAAQ,cAAc,UAAU;AAChC,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK;AAAA,IACL,KAAK,mBAAmB;AACtB,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,UAAI;AACJ,UAAI,KAAK,OAAO,qBAAqB;AACnC,2BAAmB;AAAA,MACrB,OAAO;AACL,2BAAmB;AAAA,MACrB;AACA,YAAM,cAAc,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC5E,YAAM,iBAAiB,KAAK,OAAO,sBAAsB,KAAK;AAC9D,YAAM,aAAa,QAAQ,cAAc,cAAc,aAAa,cAAc;AAClF,cAAQ,cAAc,UAAU;AAChC,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK,oBAAoB;AACvB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,gBAAgB,cAAc,WAAW,MAAM,WAAW,OAAO;AACvE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,aAAO,CAAC,WAAW,OAAO,eAAe,cAAc,YAAY,CAAC;AAAA,IACtE;AAAA,IACA,KAAK,mBAAmB;AACtB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,cAAc,cAAc,eAAe,MAAM,WAAW,OAAO;AACzE,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,aAAO,CAAC,WAAW,MAAM,cAAc,cAAc,WAAW,CAAC;AAAA,IACnE;AAAA,IACA,KAAK,wBAAwB;AAC3B,YAAM,UAAU,cAAc,UAAU,MAAM,WAAW,OAAO;AAChE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,aAAa,WAAW,SAAS,cAAc,YAAY;AACjE,cAAQ,cAAc,UAAU;AAChC,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK;AAAA,IACL,KAAK,sBAAsB;AACzB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,YAAM,cAAc,cAAc,SAAS,MAAM,WAAW,OAAO;AACnE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,aAAO,CAAC,WAAW,OAAO,aAAa,YAAY,CAAC;AAAA,IACtD;AAAA,IACA,KAAK,sBAAsB;AACzB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,cAAc,cAAc,UAAU,MAAM,WAAW,OAAO;AACpE,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,iBAAW,SAAS,WAAW;AAC/B,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,aAAO,CAAC,WAAW,QAAQ,cAAc,YAAY,CAAC;AAAA,IACxD;AAAA,IACA,KAAK,mBAAmB;AACtB,YAAM,cAAc,cAAc,UAAU,MAAM,WAAW,OAAO;AACpE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,aAAa,OAAO,aAAa,SAAS,YAAY;AAC5D,cAAQ,cAAc,UAAU;AAChC,aAAO,CAAC,WAAW,QAAQ;AAAA,IAC7B;AAAA,IACA,KAAK,oBAAoB;AACvB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,aAAa,QAAQ,cAAc,SAAS,EAAE;AACpD,aAAO,CAAC,OAAO,WAAW,KAAK,GAAG,OAAO,CAAC;AAAA,IAC5C;AAAA,IACA,KAAK,oBAAoB;AACvB,YAAM,WAAW,cAAc,gBAAgB,MAAM,WAAW,OAAO;AACvE,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,gBAAgB,QAAQ,cAAc,SAAS,EAAE;AACvD,YAAM,iBAAiB,cAAc,OAAO,IAAI;AAChD,cAAQ,cAAc,cAAc;AACpC,aAAO,CAAC,eAAe,QAAQ;AAAA,IACjC;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,SAAS,4BAA4B,MAAM,WAAW,SAAS;AAC7D,QAAM,CAAC,SAAS,cAAc,IAAI,cAAc,YAAY,MAAM,WAAW,OAAO;AACpF,QAAM,YAAY,YAAY;AAC9B,QAAM,YAAY,CAAC;AACnB,QAAM,UAAU,mBAAmB;AACnC,QAAM,cAAc,YAAY;AAChC,QAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,MAAI,WAAW;AACb,QAAI,WAAW,YAAY,GAAG;AAC5B,YAAM,IAAI,MAAM,uGAAuG;AAAA,IACzH;AACA,QAAI,CAAC,WAAW,aAAa,YAAY,GAAG;AAC1C,YAAM,IAAI,MAAM,kFAAkF;AAAA,IACpG;AAAA,EACF;AACA,MAAI,aAAa;AACf,UAAM,IAAI,MAAM,sEAAsE;AAAA,EACxF;AACA,QAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,QAAM,OAAO,WAAW,MAAM,WAAW,OAAO;AAChD,QAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,QAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,MAAI,CAAC,SAAS,QAAQ,IAAI,cAAc,QAAQ,MAAM,WAAW,OAAO;AACxE,MAAI,WAAW;AACb,eAAW;AACX,cAAU;AAAA,EACZ;AACA,QAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,SAAO;AAAA,IACL;AAAA,IACA,KAAK;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,IAAI,aAAa,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC9E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,UAAU;AACb,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,QAAQ,MAAM,YAAY,QAAQ,CAAC;AAAA,IACzJ;AAAA,IACA,KAAK,UAAU;AACb,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,WAAW,MAAM,WAAW,OAAO;AAChD,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,MAAM,YAAY,CAAC,UAAU,IAAI,UAAU,EAAE,CAAC,CAAC;AAAA,IAC7L;AAAA,IACA,KAAK,gBAAgB;AACnB,YAAM,EAAE,QAAQ,KAAK,MAAM,YAAY,WAAW,SAAS,UAAU,gBAAgB,eAAe,IAAI,4BAA4B,MAAM,WAAW,OAAO;AAC5J,aAAO,CAAC,IAAI,MAAM,OAAO;AAAA,QACvB,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO;AAAA,QAC9C,QAAQ,cAAc,UAAU,MAAM,WAAW,OAAO;AAAA,QACxD,SAAS,CAAC,OAAO,IAAI,OAAO,EAAE;AAAA,QAC9B,KAAK;AAAA,QACL;AAAA,QACA,WAAW,CAAC,UAAU,IAAI,UAAU,EAAE;AAAA,QACtC,MAAM;AAAA,QACN,YAAY;AAAA,QACZ,wBAAwB;AAAA,QACxB;AAAA,MACF,CAAC,CAAC;AAAA,IACJ;AAAA,IACA,KAAK,8BAA8B;AACjC,YAAM,EAAE,QAAQ,KAAK,MAAM,YAAY,WAAW,SAAS,UAAU,gBAAgB,eAAe,IAAI,4BAA4B,MAAM,WAAW,OAAO;AAC5J,aAAO,CAAC,IAAI,MAAM,gBAAgB;AAAA,QAChC,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO;AAAA,QAC9C,QAAQ,cAAc,UAAU,MAAM,WAAW,OAAO;AAAA,QACxD,SAAS,CAAC,OAAO,IAAI,OAAO,EAAE;AAAA,QAC9B,KAAK;AAAA,QACL;AAAA,QACA,WAAW,CAAC,UAAU,IAAI,UAAU,EAAE;AAAA,QACtC,MAAM;AAAA,QACN,YAAY;AAAA,QACZ,wBAAwB;AAAA,QACxB;AAAA,MACF,CAAC,CAAC;AAAA,IACJ;AAAA,IACA,KAAK;AAAA,IACL,KAAK,mBAAmB;AACtB,YAAM,QAAQ,cAAc,eAAe,MAAM,WAAW,OAAO;AACnE,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,WAAW,MAAM,WAAW,OAAO;AAChD,aAAO,CAAC,IAAI,gBAAgB,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,OAAO,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,IAAI,CAAC;AAAA,IACnK;AAAA,IACA,KAAK;AAAA,IACL,KAAK,mBAAmB;AACtB,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,WAAW,MAAM,WAAW,OAAO;AAChD,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,aAAO,CAAC,IAAI,gBAAgB,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,MAAM,YAAY,CAAC,UAAU,IAAI,UAAU,EAAE,CAAC,CAAC;AAAA,IAC1M;AAAA,IACA,KAAK,UAAU;AACb,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,CAAC,OAAO,IAAI,OAAO,IAAI,OAAO,EAAE,GAAG,MAAM,YAAY,CAAC,UAAU,IAAI,UAAU,IAAI,UAAU,EAAE,CAAC,CAAC;AAAA,IACtN;AAAA,IACA,KAAK,WAAW;AACd,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,CAAC,WAAW,IAAI,WAAW,EAAE,GAAG,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,IAAI,CAAC;AAAA,IACjI;AAAA,IACA,KAAK,WAAW;AACd,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,CAAC,WAAW,IAAI,WAAW,EAAE,GAAG,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,IAAI,CAAC;AAAA,IACjI;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,YAAM,sBAAsB,cAAc,uBAAuB,MAAM,WAAW,OAAO;AACzF,YAAM,EAAE,QAAQ,QAAQ,IAAI,IAAI,kBAAkB,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,CAAC,WAAW,IAAI,WAAW,EAAE,GAAG,CAAC,OAAO,IAAI,OAAO,EAAE,GAAG,MAAM,mBAAmB;AACjL,aAAO,CAAC,QAAQ,OAAO;AAAA,IACzB;AAAA,IACA,KAAK,aAAa;AAChB,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,CAAC,WAAW,IAAI,WAAW,IAAI,WAAW,EAAE,GAAG,CAAC,OAAO,IAAI,OAAO,IAAI,OAAO,EAAE,GAAG,IAAI,CAAC;AAAA,IAC7J;AAAA,IACA,KAAK,aAAa;AAChB,YAAM,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAChE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,CAAC,WAAW,IAAI,WAAW,IAAI,WAAW,EAAE,GAAG,CAAC,OAAO,IAAI,OAAO,IAAI,OAAO,EAAE,GAAG,IAAI,CAAC;AAAA,IAC7J;AAAA,IACA,KAAK,cAAc;AACjB,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,OAAO,cAAc,OAAO,MAAM,WAAW,OAAO;AAC1D,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,eAAe,QAAQ;AAC7B,YAAM,cAAc,QAAQ;AAC5B,YAAM,iBAAiB,UAAU;AACjC,YAAM,gBAAgB,UAAU;AAChC,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,CAAC,cAAc,WAAW,GAAG,MAAM,CAAC,gBAAgB,aAAa,GAAG,MAAM,CAAC;AAAA,IACrM;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,aAAa,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC9E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,QAAQ;AACX,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,aAAO,CAAC,IAAI,KAAK,OAAO,OAAO,KAAK,CAAC;AAAA,IACvC;AAAA,IACA,KAAK,YAAY;AACf,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,MAAM,cAAc,OAAO,MAAM,WAAW,OAAO;AACzD,aAAO,CAAC,IAAI,SAAS,OAAO,MAAM,GAAG,CAAC;AAAA,IACxC;AAAA,IACA,KAAK,eAAe;AAClB,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,YAAY,QAAQ,YAAY,IAAI,CAAC;AAAA,IACnD;AAAA,IACA,KAAK,UAAU;AACb,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,aAAO,CAAC,IAAI,OAAO,SAAS,OAAO,SAAS,UAAU,KAAK,CAAC;AAAA,IAC9D;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtH;AAAA,IACA,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACpE;AAAA,IACA,KAAK,wBAAwB;AAC3B,aAAO,CAAC,IAAI,qBAAqB,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACvL;AAAA,IACA,KAAK,iBAAiB;AACpB,aAAO,CAAC,IAAI;AAAA,QACV,cAAc,SAAS,MAAM,WAAW,OAAO;AAAA,QAC/C,cAAc,UAAU,MAAM,WAAW,OAAO;AAAA,QAChD,cAAc,UAAU,MAAM,WAAW,OAAO;AAAA,QAChD,cAAc,SAAS,MAAM,WAAW,OAAO;AAAA,MACjD,CAAC;AAAA,IACH;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,QAAQ,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC5D,aAAO,CAAC,IAAI,MAAM,OAAO,MAAM,OAAO,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACzF;AAAA,IACA,KAAK,mBAAmB;AACtB,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,QAAQ,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC5D,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,gBAAgB,OAAO,OAAO,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IAC3G;AAAA,IACA,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACvH;AAAA,IACA,KAAK,aAAa;AAChB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACrE;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,SAAS,UAAU,MAAM,WAAW,SAAS;AAC3C,QAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,QAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,QAAM,gBAAgB,cAAc,iBAAiB,MAAM,WAAW,OAAO;AAC7E,QAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,QAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,QAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,IAAI,aAAa,OAAO,MAAM,WAAW,SAAS,iBAAiB,MAAM,8BAA8B;AACrG,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,uBAAuB;AAC1B,YAAM,EAAE,OAAO,QAAQ,eAAe,cAAc,gBAAgB,aAAa,IAAI,UAAU,MAAM,WAAW,OAAO;AACvH,YAAM,SAAS,MAAM,IAAI,MAAM,gCAAgC,OAAO,QAAQ,eAAe,cAAc,gBAAgB,YAAY;AACvI,aAAO,CAAC,OAAO,iBAAiB,OAAO,cAAc;AAAA,IACvD;AAAA,IACA,KAAK,uBAAuB;AAC1B,YAAM,EAAE,OAAO,QAAQ,eAAe,cAAc,eAAe,IAAI,UAAU,MAAM,WAAW,OAAO;AACzG,YAAM,qBAAqB,cAAc,sBAAsB,MAAM,WAAW,OAAO;AACvF,YAAM,SAAS,MAAM,IAAI,MAAM,6BAA6B,OAAO,QAAQ,eAAe,cAAc,gBAAgB,kBAAkB;AAC1I,aAAO,CAAC,OAAO,iBAAiB,OAAO,YAAY;AAAA,IACrD;AAAA,IACA,KAAK;AAAA,IACL,KAAK,uBAAuB;AAC1B,YAAM,EAAE,OAAO,QAAQ,eAAe,cAAc,eAAe,IAAI,UAAU,MAAM,WAAW,OAAO;AACzG,aAAO,CAAC,MAAM,IAAI,MAAM,uBAAuB,OAAO,QAAQ,eAAe,cAAc,cAAc,CAAC;AAAA,IAC5G;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,YAAY,IAAI,KAAK,cAAc,aAAa,MAAM,WAAW,OAAO,GAAG,MAAM;AACvF,YAAM,SAAS,CAAC,MAAM,IAAI,WAAW,SAAS,CAAC;AAC/C,gBAAU,QAAQ;AAClB,aAAO;AAAA,IACT;AAAA,IACA,KAAK,YAAY;AACf,aAAO,IAAI,eAAe,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC;AAAA,IACtH;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,aAAa,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC9E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,cAAc;AACjB,YAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,aAAO,CAAC,IAAI,WAAW,gBAAgB,MAAM,CAAC;AAAA,IAChD;AAAA,IACA,KAAK,UAAU;AACb,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,SAAS,IAAI,KAAK,GAAG,GAAG,MAAM;AACpC,aAAO,CAAC,OAAO,QAAQ,OAAO,OAAO;AAAA,IACvC;AAAA,IACA,KAAK,cAAc;AACjB,YAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,aAAO,CAAC,IAAI,WAAW,gBAAgB,MAAM,CAAC;AAAA,IAChD;AAAA,IACA,KAAK,UAAU;AACb,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,SAAS,IAAI,OAAO,CAAC;AAC3B,aAAO,CAAC,OAAO,QAAQ,OAAO,OAAO;AAAA,IACvC;AAAA,IACA,KAAK,YAAY;AACf,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,SAAS,IAAI,OAAO,GAAG,IAAI;AACjC,aAAO,CAAC,OAAO,QAAQ,OAAO,OAAO;AAAA,IACvC;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,aAAa,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC9E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,SAAS;AACZ,aAAO,UAAU,KAAK;AAAA,IACxB;AAAA,IACA,KAAK;AACH,YAAM,MAAM,cAAc,WAAW,MAAM,WAAW,OAAO;AAC7D,aAAO,CAAC,UAAU,KAAK,MAAM,WAAW,OAAO,KAAK,GAAG;AAAA,IACzD,KAAK;AACH,aAAO,CAAC,UAAU,KAAK,MAAM,WAAW,OAAO,CAAC;AAAA,IAClD,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK,2BAA2B;AAC9B,YAAM,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO;AACzD,aAAO,CAAC,YAAY,KAAK,CAAC;AAAA,IAC5B;AAAA,IACA,KAAK;AACH,aAAO,cAAc,KAAK,MAAM,WAAW,OAAO,EAAE,IAAI,CAAC,MAAM,YAAY,CAAC,CAAC;AAAA,IAC/E,KAAK;AACH,YAAM,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO;AAC5D,aAAO,CAAC,YAAY,QAAQ,CAAC;AAAA,IAC/B,KAAK;AACH,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,EAAE,OAAO,OAAO,CAAC;AAAA,IACnF,KAAK;AACH,aAAO,cAAc,KAAK,MAAM,WAAW,OAAO,EAAE,IAAI,CAAC,MAAM,IAAI,SAAS,EAAE,KAAK,CAAC;AAAA,IACtF,KAAK;AACH,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,EAAE,MAAM,OAAO,CAAC;AAAA,IAChF,KAAK;AACH,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,EAAE,MAAM,OAAO,CAAC;AAAA,IAChF,KAAK;AACH,aAAO,CAAC,IAAI,OAAO,CAAC,CAAC;AAAA,IACvB,KAAK;AACH,YAAM,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO;AAC1D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,cAAQ,KAAK,gGAAgG;AAC7G,cAAQ,IAAI,OAAO;AACnB,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,gBAAQ,IAAI,MAAM,UAAU,MAAM,KAAK,KAAK,GAAG,SAAS,CAAC,EAAE,MAAM,GAAG,SAAS,CAAC;AAAA,MAChF;AACA,aAAO,CAAC,MAAM;AAAA,IAChB;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,YAAY,MAAM;AAAA,EACpB,YAAY,UAAU,YAAY;AAChC,SAAK,WAAW;AAChB,SAAK,aAAa;AAClB,SAAK,SAAS,OAAO,CAAC;AACtB,SAAK,YAA4B,oBAAI,IAAI;AACzC,SAAK,KAAK,MAAM;AAAA,EAClB;AAAA,EACA,IAAI,KAAK;AACP,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EACA,gBAAgB;AACd,SAAK,UAAU,QAAQ,CAAC,UAAU,MAAM,QAAQ,CAAC;AACjD,SAAK,UAAU,MAAM;AACrB,SAAK,OAAO,QAAQ;AAAA,EACtB;AAAA,EACA,OAAO;AACL,WAAO,KAAK,UAAU;AAAA,EACxB;AAAA,EACA,aAAa;AACX,WAAO,OAAO,KAAK,KAAK,GAAG,OAAO;AAAA,EACpC;AAAA,EACA,MAAM,OAAO,MAAM,QAAQ;AACzB,SAAK,uBAAuB,MAAM,MAAM;AACxC,UAAM,QAAQ,MAAM,KAAK,KAAK;AAC9B,SAAK,UAAU,QAAQ,CAAC,UAAU,MAAM,QAAQ,CAAC;AACjD,SAAK,UAAU,MAAM;AACrB,WAAO,KAAK,MAAM;AAChB,YAAM,UAAU,QAAQ,MAAM;AAC9B,YAAM,aAAa,MAAM;AACzB,YAAM,eAAe,QAAQ;AAC7B,mBAAa,OAAO,eAAe,cAAc,MAAM,kDAAkD,uCAAuC,wBAAwB;AACxK,eAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,cAAM,MAAM,MAAM;AAClB,cAAM,QAAQ,QAAQ;AACtB,aAAK,KAAK;AACV,aAAK,UAAU,IAAI,KAAK,KAAK;AAAA,MAC/B;AACA,aAAO,KAAK;AAAA,IACd,CAAC;AAAA,EACH;AAAA,EACA,MAAM,KAAK,MAAM,cAAc;AAC7B,SAAK,uBAAuB,MAAM,YAAY;AAC9C,UAAM,QAAQ,MAAM,KAAK,KAAK;AAC9B,WAAO,KAAK,MAAM;AAChB,YAAM,SAAS,CAAC;AAChB,eAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,cAAM,MAAM,MAAM;AAClB,cAAM,QAAQ,KAAK,gBAAgB,KAAK,YAAY;AACpD,eAAO,KAAK,KAAK;AAAA,MACnB;AACA,aAAO,MAAM,MAAM;AAAA,IACrB,CAAC;AAAA,EACH;AAAA,EACA,gBAAgB,KAAK,cAAc;AACjC,UAAM,SAAS,KAAK,UAAU,IAAI,GAAG;AACrC,WAAO,UAAU,OAAO,SAAS;AAAA,EACnC;AAAA,EACA,uBAAuB,KAAK,OAAO;AACjC,QAAI,IAAI,UAAU,KAAK,UAAU;AAC/B,YAAM,IAAI,MAAM,oBAAoB,KAAK,qBAAqB,IAAI,OAAO;AAAA,IAC3E;AACA,QAAI,MAAM,UAAU,KAAK,YAAY;AACnC,YAAM,IAAI,MAAM,sBAAsB,KAAK,uBAAuB,MAAM,OAAO;AAAA,IACjF;AAAA,EACF;AACF;AAGA,IAAI,aAAa,OAAO,MAAM,WAAW,SAAS,oBAAoB;AACpE,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK,eAAe;AAClB,YAAM,sBAAsB,gBAAgB,yBAAyB,KAAK,IAAI;AAC9E,UAAI,uBAAuB,MAAM;AAC/B,eAAO,CAAC,mBAAmB;AAAA,MAC7B,OAAO;AACL,cAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,cAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,cAAM,YAAY,IAAI,UAAU,UAAU,UAAU;AACpD,wBAAgB,aAAa,KAAK,MAAM,SAAS;AACjD,eAAO,CAAC,UAAU,MAAM;AAAA,MAC1B;AAAA,IACF;AAAA,IACA,KAAK;AAAA,IACL,KAAK,uBAAuB;AAC1B,YAAM,SAAS,cAAc,eAAe,MAAM,WAAW,SAAS,eAAe;AACrF,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,YAAY,gBAAgB,iBAAiB,OAAO,EAAE;AAC5D,aAAO,CAAC,MAAM,UAAU,OAAO,MAAM,MAAM,CAAC;AAAA,IAC9C;AAAA,IACA,KAAK;AAAA,IACL,KAAK,qBAAqB;AACxB,YAAM,SAAS,cAAc,eAAe,MAAM,WAAW,SAAS,eAAe;AACrF,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,YAAY,gBAAgB,iBAAiB,OAAO,EAAE;AAC5D,aAAO,CAAC,MAAM,UAAU,KAAK,MAAM,YAAY,CAAC;AAAA,IAClD;AAAA,IACA,KAAK;AAAA,IACL,KAAK,qBAAqB;AACxB,YAAM,SAAS,cAAc,eAAe,MAAM,WAAW,SAAS,eAAe;AACrF,YAAM,YAAY,gBAAgB,iBAAiB,OAAO,EAAE;AAC5D,aAAO,CAAC,UAAU,WAAW,CAAC;AAAA,IAChC;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,kBAAkB;AACrB,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,mBAAmB,cAAc,oBAAoB,MAAM,WAAW,OAAO;AACnF,aAAO,CAAC,IAAI,MAAM,eAAe,QAAQ,CAAC,KAAK,IAAI,KAAK,EAAE,GAAG,cAAc,gBAAgB,CAAC;AAAA,IAC9F;AAAA,IACA,KAAK,yBAAyB;AAC5B,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,mBAAmB,cAAc,oBAAoB,MAAM,WAAW,OAAO;AACnF,aAAO,CAAC,IAAI,MAAM,sBAAsB,QAAQ,CAAC,KAAK,IAAI,KAAK,EAAE,GAAG,cAAc,gBAAgB,CAAC;AAAA,IACrG;AAAA,IACA,KAAK,iBAAiB;AACpB,YAAM,SAAS,cAAc,SAAS,MAAM,WAAW,OAAO;AAC9D,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,qBAAqB,cAAc,sBAAsB,MAAM,WAAW,OAAO;AACvF,aAAO,CAAC,IAAI,MAAM,cAAc,QAAQ,OAAO,QAAQ,UAAU,QAAQ,kBAAkB,CAAC;AAAA,IAC9F;AAAA,IACA,KAAK,8BAA8B;AACjC,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,YAAM,cAAc,cAAc,eAAe,MAAM,WAAW,OAAO;AACzE,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,gBAAgB,cAAc,iBAAiB,MAAM,WAAW,OAAO;AAC7E,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,MAAM,UAAU,QAAQ,YAAY,cAAc,YAAY,GAAG,SAAS,YAAY,GAAG,WAAW,WAAW,CAAC;AAAA,IAC9H;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/G;AAAA,IACA,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAClH;AAAA,IACA,KAAK,WAAW;AACd,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjH;AAAA,IACA,KAAK,gBAAgB;AACnB,aAAO,CAAC,IAAI,aAAa,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtH;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC9G;AAAA,IACA,KAAK,aAAa;AAChB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACnH;AAAA,IACA,KAAK,cAAc;AACjB,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACpH;AAAA,IACA,KAAK,cAAc;AACjB,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE;AAAA,IACA,KAAK,aAAa;AAChB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACnH;AAAA,IACA,KAAK;AAAA,IACL,KAAK,YAAY;AACf,aAAO,CAAC,IAAI,MAAM,cAAc,aAAa,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACrK;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AACH,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC9N,KAAK;AACH,aAAO,CAAC,IAAI,OAAO,cAAc,YAAY,MAAM,WAAW,OAAO,GAAG,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChI,KAAK;AACH,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtH,KAAK;AACH,YAAM,CAAC,SAAS,cAAc,IAAI,cAAc,YAAY,MAAM,WAAW,OAAO;AACpF,YAAM,YAAY,YAAY;AAC9B,YAAM,UAAU,mBAAmB;AACnC,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,UAAI,WAAW;AACb,YAAI,WAAW,YAAY,GAAG;AAC5B,gBAAM,IAAI,MAAM,oFAAoF;AAAA,QACtG;AACA,YAAI,CAAC,WAAW,YAAY,GAAG;AAC7B,gBAAM,IAAI,MAAM,+DAA+D;AAAA,QACjF;AAAA,MACF;AACA,YAAM,CAAC,SAAS,QAAQ,IAAI,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC1E,aAAO,CAAC,IAAI,MAAM,OAAO;AAAA,QACvB,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO;AAAA,QAC9C,GAAG,cAAc,KAAK,MAAM,WAAW,OAAO;AAAA,QAC9C,YAAY,cAAc,cAAc,MAAM,WAAW,OAAO;AAAA,QAChE,YAAY,cAAc,cAAc,MAAM,WAAW,OAAO;AAAA,QAChE,MAAM;AAAA,QACN,YAAY;AAAA,QACZ,wBAAwB;AAAA,QACxB;AAAA,MACF,CAAC,CAAC;AAAA,IACJ;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AACH,aAAO,CAAC,IAAI,cAAc,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/K,KAAK;AAAA,IACL,KAAK,oBAAoB;AACvB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACpU;AAAA,IACA,KAAK,oBAAoB;AACvB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACpU;AAAA,IACA,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,2BAA2B,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7R;AAAA,IACA,KAAK,WAAW;AACd,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACnE;AAAA,IACA,KAAK,cAAc;AACjB,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACtE;AAAA,IACA,KAAK,iBAAiB;AACpB,aAAO,CAAC,IAAI,cAAc,cAAc,iBAAiB,MAAM,WAAW,OAAO,GAAG,cAAc,eAAe,MAAM,WAAW,OAAO,GAAG,cAAc,gBAAgB,MAAM,WAAW,OAAO,GAAG,cAAc,gBAAgB,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/P;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,OAAO;AACV,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAC/E;AAAA,IACA,KAAK,QAAQ;AACX,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAChF;AAAA,IACA,KAAK,OAAO;AACV,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAC/E;AAAA,IACA,KAAK,OAAO;AACV,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAC/E;AAAA,IACA,KAAK,OAAO;AACV,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAC/E;AAAA,IACA,KAAK,OAAO;AACV,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAC/E;AAAA,IACA,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IACxE;AAAA,IACA,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IACxE;AAAA,IACA,KAAK,QAAQ;AACX,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,QAAQ,CAAC;AAAA,IAChF;AAAA,IACA,KAAK,WAAW;AACd,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,WAAW,cAAc,WAAW,MAAM,WAAW,OAAO;AAClE,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,WAAW,QAAQ,CAAC;AAAA,IAC9F;AAAA,IACA,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,WAAW,cAAc,WAAW,MAAM,WAAW,OAAO;AAClE,aAAO,CAAC,IAAI,OAAO,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,MAAM,WAAW,QAAQ,CAAC;AAAA,IAC7F;AAAA,IACA,KAAK;AACH,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,SAAS,GAAG,SAAS,IAAI,CAAC;AAAA,IACxC,KAAK,iBAAiB;AACpB,YAAM,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO;AACtD,YAAM,WAAW,cAAc,WAAW,MAAM,WAAW,OAAO;AAClE,YAAM,QAAQ,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC5D,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,aAAO,CAAC,IAAI,cAAc,IAAI,UAAU,OAAO,YAAY,CAAC;AAAA,IAC9D;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK;AAAA,IACL,KAAK,UAAU;AACb,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,UAAI,SAAS,cAAc,WAAW,MAAM,WAAW,OAAO;AAC9D,eAAS,OAAO,MAAM,GAAG,CAAC;AAC1B,aAAO,CAAC,IAAI,OAAO,QAAQ,IAAI,CAAC;AAAA,IAClC;AAAA,IACA,KAAK,UAAU;AACb,YAAM,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO;AAC1D,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,aAAO,CAAC,IAAI,OAAO,QAAQ,IAAI,KAAK,SAAS,OAAO,GAAG,CAAC,CAAC;AAAA,IAC3D;AAAA,IACA,KAAK,YAAY;AACf,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO;AAC1D,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,aAAO,CAAC,IAAI,OAAO,QAAQ,IAAI,KAAK,SAAS,OAAO,GAAG,MAAM,SAAS,CAAC;AAAA,IACzE;AAAA,IACA,KAAK,WAAW;AACd,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,OAAO,CAAC;AACd,eAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,YAAI,KAAK,IAAI;AACX,eAAK,KAAK,CAAC;AAAA,QACb;AAAA,MACF;AACA,YAAM,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO;AAC1D,aAAO,CAAC,IAAI,QAAQ,QAAQ,IAAI,CAAC;AAAA,IACnC;AAAA,IACA,KAAK,aAAa;AAChB,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,SAAS,cAAc,KAAK,MAAM,WAAW,OAAO;AAC1D,aAAO,CAAC,IAAI,QAAQ,QAAQ,IAAI,CAAC;AAAA,IACnC;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,OAAO,IAAI,CAAC;AAAA,IAC9E;AAAA,IACA,KAAK,gBAAgB;AACnB,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,YAAM,MAAM,cAAc,OAAO,MAAM,WAAW,OAAO;AACzD,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,cAAc,cAAc,eAAe,MAAM,WAAW,OAAO;AACzE,YAAM,iBAAiB,cAAc,kBAAkB,MAAM,WAAW,OAAO;AAC/E,YAAM,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,aAAa,SAAS,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,cAAc,CAAC;AAAA,IACvH;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,KAAK,MAAM;AAChB,cAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,cAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,cAAM,QAAQ,QAAQ,GAAG;AACzB,cAAM,gBAAgB,IAAI,QAAQ,QAAQ,EAAE,EAAE;AAC9C,cAAM,SAAS,QAAQ,IAAI,CAAC,YAAY;AACtC,gBAAM,YAAY,aAAa,YAAY,QAAQ,OAAO,KAAK;AAC/D,cAAI,CAAC,aAAa,CAAC,aAAa,YAAY,IAAI,QAAQ,OAAO,EAAE,OAAO,aAAa,GAAG;AACtF,kBAAM,IAAI,MAAM,wCAAwC;AAAA,UAC1D;AACA,iBAAO,YAAY,UAAU,IAAI,QAAQ,SAAS,KAAK;AAAA,QACzD,CAAC;AACD,eAAO,CAAC,IAAI,MAAM,QAAQ,IAAI,CAAC;AAAA,MACjC,CAAC;AAAA,IACH;AAAA,IACA,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,UAAU,cAAc,UAAU,MAAM,WAAW,OAAO;AAChE,aAAO,IAAI,QAAQ,SAAS,IAAI;AAAA,IAClC;AAAA,IACA,KAAK,QAAQ;AACX,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IACtE;AAAA,IACA,KAAK;AAAA,IACL,KAAK,UAAU;AACb,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,YAAM,kBAAkB,cAAc,mBAAmB,MAAM,WAAW,OAAO;AACjF,YAAM,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO;AAC3D,aAAO,IAAI,MAAM,SAAS,iBAAiB,IAAI;AAAA,IACjD;AAAA,IACA,KAAK,aAAa;AAChB,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,YAAM,SAAS,cAAc,UAAU,MAAM,WAAW,OAAO;AAC/D,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,aAAO,CAAC,IAAI,UAAU,SAAS,QAAQ,KAAK,CAAC;AAAA,IAC/C;AAAA,IACA,KAAK,YAAY;AACf,YAAM,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO;AACrD,YAAM,UAAU,cAAc,WAAW,MAAM,WAAW,OAAO;AACjE,aAAO,CAAC,IAAI,SAAS,GAAG,OAAO,CAAC;AAAA,IAClC;AAAA,IACA,KAAK,iBAAiB;AACpB,YAAM,UAAU,cAAc,iBAAiB,MAAM,WAAW,OAAO;AACvE,YAAM,QAAQ,cAAc,eAAe,MAAM,WAAW,OAAO;AACnE,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,YAAM,eAAe,cAAc,gBAAgB,MAAM,WAAW,OAAO;AAC3E,aAAO,CAAC,IAAI,cAAc,SAAS,cAAc,OAAO,aAAa,UAAU,aAAa,QAAQ,eAAe,IAAI,KAAK,cAAc,aAAa,KAAK,CAAC,CAAC;AAAA,IAChK;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,uBAAuB;AAC1B,YAAM,EAAE,eAAe,cAAc,mBAAmB,gBAAgB,IAAI,IAAI,OAAO,oBAAoB,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,UAAU,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,GAAG,cAAc,gBAAgB,MAAM,WAAW,OAAO,CAAC;AAChU,aAAO;AAAA,QACL;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,IACA,KAAK,iBAAiB;AACpB,YAAM,EAAE,eAAe,YAAY,IAAI,IAAI,OAAO,cAAc,cAAc,gBAAgB,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,CAAC;AACnO,aAAO,CAAC,eAAe,WAAW;AAAA,IACpC;AAAA,IACA,KAAK,qBAAqB;AACxB,YAAM,aAAa,IAAI,OAAO,kBAAkB,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,CAAC;AAC1M,aAAO,CAAC,UAAU;AAAA,IACpB;AAAA,IACA,KAAK,oBAAoB;AACvB,YAAM,aAAa,IAAI,OAAO,iBAAiB,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,CAAC;AACzM,aAAO,CAAC,UAAU;AAAA,IACpB;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC/D;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAChE;AAAA,IACA,KAAK,SAAS;AACZ,aAAO,CAAC,IAAI,MAAM,cAAc,KAAK,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACjE;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,gBAAgB;AACnB,YAAM,EAAE,QAAQ,aAAa,IAAI,IAAI,OAAO,aAAa,cAAc,QAAQ,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,GAAG,cAAc,aAAa,MAAM,WAAW,OAAO,GAAG,cAAc,eAAe,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,GAAG,cAAc,YAAY,MAAM,WAAW,OAAO,GAAG,cAAc,0BAA0B,MAAM,WAAW,OAAO,CAAC;AAC9e,aAAO,CAAC,QAAQ,YAAY;AAAA,IAC9B;AAAA,IACA,KAAK,eAAe;AAClB,YAAM,EAAE,SAAS,QAAQ,MAAM,IAAI,IAAI,OAAO,YAAY,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,aAAa,MAAM,WAAW,OAAO,GAAG,cAAc,aAAa,MAAM,WAAW,OAAO,CAAC;AACtN,aAAO,CAAC,SAAS,QAAQ,KAAK;AAAA,IAChC;AAAA,IACA,KAAK,0BAA0B;AAC7B,YAAM,SAAS,IAAI,OAAO,uBAAuB,cAAc,SAAS,MAAM,WAAW,OAAO,GAAG,cAAc,cAAc,MAAM,WAAW,OAAO,CAAC;AACxJ,aAAO,CAAC,MAAM;AAAA,IAChB;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,IAAI,cAAc,CAAC,MAAM,WAAW,SAAS,MAAM,8BAA8B;AAC/E,UAAQ,KAAK,IAAI;AAAA,IACf,KAAK,QAAQ;AACX,aAAO,CAAC,IAAI,KAAK,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAClH;AAAA,IACA,KAAK,cAAc;AACjB,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,WAAW,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IAC5E;AAAA,IACA,KAAK,WAAW;AACd,YAAM,OAAO,cAAc,QAAQ,MAAM,WAAW,OAAO;AAC3D,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,IAAI,CAAC;AAAA,IACzE;AAAA,IACA,KAAK,WAAW;AACd,aAAO,CAAC,IAAI,QAAQ,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACrH;AAAA,IACA,KAAK,aAAa;AAChB,aAAO,CAAC,IAAI,UAAU,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,QAAQ,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC1K;AAAA,IACA,KAAK;AAAA,IACL,KAAK,OAAO;AACV,aAAO,CAAC,IAAI,IAAI,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,WAAW,MAAM,WAAW,OAAO,GAAG,cAAc,iBAAiB,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IAC7K;AAAA,IACA,KAAK,kBAAkB;AACrB,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,YAAM,WAAW,cAAc,YAAY,MAAM,WAAW,OAAO;AACnE,aAAO,CAAC,IAAI,eAAe,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,YAAY,QAAQ,CAAC;AAAA,IAChG;AAAA,IACA,KAAK,kBAAkB;AACrB,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO;AACvE,YAAM,QAAQ,cAAc,SAAS,MAAM,WAAW,OAAO;AAC7D,aAAO,CAAC,IAAI,eAAe,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,YAAY,KAAK,CAAC;AAAA,IAC7F;AAAA,IACA,KAAK,gBAAgB;AACnB,YAAM,YAAY,cAAc,aAAa,MAAM,WAAW,OAAO;AACrE,YAAM,aAAa,cAAc,cAAc,MAAM,WAAW,OAAO,EAAE,YAAY;AACrF,aAAO,CAAC,IAAI,aAAa,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,WAAW,UAAU,CAAC;AAAA,IAC/F;AAAA,IACA,KAAK,eAAe;AAClB,aAAO,CAAC,IAAI,YAAY,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG,cAAc,SAAS,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACzH;AAAA,IACA,KAAK,iBAAiB;AACpB,aAAO,CAAC,IAAI,cAAc,cAAc,MAAM,MAAM,WAAW,OAAO,GAAG,cAAc,MAAM,MAAM,WAAW,OAAO,CAAC,CAAC;AAAA,IACzH;AAAA,IACA;AACE,YAAM,UAAU,aAAa,KAAK,uBAAuB;AAAA,EAC7D;AACF;AAGA,SAAS,YAAY,MAAM,WAAW,SAAS,iBAAiB,QAAQ,MAAM;AAC5E,QAAM,SAAS,CAAC,OAAO,YAAY,aAAa;AAC9C,YAAQ,MAAM,UAAU;AAAA,MACtB,KAAK;AACH,eAAO,MAAM,MAAM,UAAU,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC3D,KAAK;AACH,eAAO,MAAM,MAAM,WAAW,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,WAAW,OAAO,YAAY,QAAQ;AAAA,MAC/C,KAAK;AACH,eAAO,MAAM,MAAM,WAAW,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,MAAM,MAAM,WAAW,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,WAAW,OAAO,YAAY,QAAQ;AAAA,MAC/C,KAAK;AACH,eAAO,MAAM,MAAM,WAAW,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,WAAW,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC5D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,MAAM,MAAM,YAAY,OAAO,YAAY,QAAQ,CAAC;AAAA,MAC7D,KAAK;AACH,eAAO,WAAW,OAAO,YAAY,UAAU,eAAe;AAAA,MAChE,KAAK;AACH,cAAM,WAAW,gBAAgB,MAAM,EAAE;AACzC,YAAI,YAAY,SAAS,gBAAgB;AACvC,iBAAO,SAAS,eAAe,IAAI,cAAc,OAAO,YAAY,QAAQ,CAAC;AAAA,QAC/E,OAAO;AACL,gBAAM,UAAU,aAAa,MAAM,uBAAuB;AAAA,QAC5D;AAAA,MACF;AACE,cAAM,UAAU,eAAe,MAAM,uIAAuI;AAAA,IAChL;AAAA,EACF,GAAG,MAAM,WAAW,OAAO;AAC3B,MAAI,aAAa,UAAU,KAAK,GAAG;AACjC,WAAO,MAAM,KAAK,CAAC,SAAS,CAAC,EAAE,OAAO,IAAI,CAAC;AAAA,EAC7C;AACA,SAAO,CAAC,EAAE,OAAO,KAAK;AACxB;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,YAAY,CAAC,GAAG,iBAAiB,CAAC,GAAG,gBAAgB,CAAC,GAAG,cAAc,CAAC,GAAG;AACrF,SAAK,YAAY;AACjB,SAAK,iBAAiB;AACtB,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,cAAc,EAAE,IAAI,GAAG,WAAW,IAAI,aAAa,EAAE;AAC1D,SAAK,WAAW,CAAC,KAAK,WAAW;AACjC,SAAK,SAAS;AACd,SAAK,0BAA0B;AAAA,EACjC;AAAA,EACA,SAAS,IAAI,WAAW;AACtB,WAAO,EAAE,IAAI,WAAW,aAAa,EAAE;AAAA,EACzC;AAAA,EACA,IAAI,eAAe,WAAW;AAC5B,QAAI,KAAK,aAAa,WAAW;AAC/B,WAAK,WAAW;AAChB,WAAK,0BAA0B;AAAA,IACjC;AAAA,EACF;AAAA,EACA,IAAI,iBAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,mBAAmB;AACrB,WAAO,KAAK,mBAAmB;AAAA,EACjC;AAAA,EACA,IAAI,oBAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,4BAA4B;AAC1B,UAAM,QAAQ,CAAC;AACf,aAAS,IAAI,GAAG,IAAI,KAAK,SAAS,SAAS,GAAG,KAAK;AACjD,YAAM,YAAY,KAAK,SAAS,MAAM,GAAG,KAAK,SAAS,SAAS,CAAC;AACjE,YAAM,KAAK,KAAK,qBAAqB,SAAS,CAAC;AAAA,IACjD;AACA,UAAM,KAAK,EAAE;AACb,SAAK,qBAAqB;AAAA,EAC5B;AAAA,EACA,qBAAqB,WAAW;AAC9B,WAAO,YAAY,UAAU,IAAI,CAAC,YAAY,QAAQ,OAAO,KAAK,QAAQ,gBAAgB,IAAI,KAAK,GAAG,QAAQ,aAAa,QAAQ,aAAa,EAAE,KAAK,GAAG,IAAI;AAAA,EAChK;AAAA,EACA,WAAW,SAAS;AAClB,QAAI,KAAK,UAAU;AACjB,WAAK;AACL,WAAK,WAAW,KAAK,SAAS,MAAM;AACpC,WAAK,SAAS,KAAK,KAAK,SAAS,KAAK,QAAQ,OAAO,CAAC;AACtD,WAAK,mBAAmB,QAAQ,KAAK,qBAAqB,KAAK,QAAQ,CAAC;AAAA,IAC1E;AAAA,EACF;AAAA,EACA,YAAY;AACV,QAAI,KAAK,YAAY,KAAK,SAAS,SAAS,GAAG;AAC7C,WAAK,WAAW,KAAK,SAAS,MAAM;AACpC,WAAK,SAAS,OAAO,EAAE;AACvB,WAAK,kBAAkB,MAAM;AAAA,IAC/B,OAAO;AACL,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AAAA,EACA,gBAAgB;AACd,QAAI,KAAK,YAAY,KAAK,SAAS,SAAS,GAAG;AAC7C,WAAK,WAAW,KAAK,SAAS,MAAM;AACpC,WAAK;AACL,YAAM,UAAU,OAAO,OAAO,CAAC,GAAG,KAAK,SAAS,KAAK,SAAS,SAAS,EAAE;AACzE,cAAQ,eAAe;AACvB,cAAQ,KAAK,KAAK;AAClB,WAAK,SAAS,OAAO,IAAI,GAAG,OAAO;AACnC,WAAK,mBAAmB,OAAO,GAAG,GAAG,KAAK,qBAAqB,KAAK,QAAQ,CAAC;AAAA,IAC/E,OAAO;AACL,YAAM,IAAI,MAAM,uDAAuD;AAAA,IACzE;AAAA,EACF;AAAA,EACA,UAAU,MAAM;AACd,WAAO,KAAK,UAAU;AAAA,EACxB;AAAA,EACA,eAAe,aAAa;AAC1B,SAAK,eAAe,YAAY,MAAM;AAAA,EACxC;AAAA,EACA,eAAe,IAAI;AACjB,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EACA,cAAc,YAAY;AACxB,SAAK,cAAc,WAAW,MAAM;AAAA,EACtC;AAAA,EACA,cAAc,IAAI;AAChB,WAAO,KAAK,cAAc;AAAA,EAC5B;AAAA,EACA,QAAQ,SAAS;AACf,eAAW,OAAO,KAAK,gBAAgB;AACrC,WAAK,eAAe,KAAK,cAAc,OAAO;AAAA,IAChD;AACA,eAAW,OAAO,KAAK,eAAe;AACpC,WAAK,cAAc,KAAK,cAAc,OAAO;AAAA,IAC/C;AAAA,EACF;AACF;AAGA,SAAS,qBAAqB,QAAQ,SAAS,WAAW,WAAW;AACnE,QAAM,YAA4B,oBAAI,IAAI;AAC1C,QAAM,gBAAgB,CAAC;AACvB,MAAI,cAAc;AAClB,MAAI,aAAa;AACjB,QAAM,OAAuB,oBAAI,IAAI;AACrC,QAAM,iBAAiB,OAAO,KAAK,MAAM,EAAE,IAAI,CAAC,SAAS,cAAc,IAAI,EAAE,EAAE;AAC/E,MAAI,gBAAgB,CAAC;AACrB,MAAI,aAAa,MAAM;AACrB,oBAAgB,UAAU,IAAI,CAAC,SAAS,cAAc,KAAK,IAAI,EAAE,EAAE;AAAA,EACrE;AACA,QAAM,WAAW,CAAC,GAAG,OAAO;AAC5B,SAAO,SAAS,SAAS,GAAG;AAC1B,UAAM,OAAO,SAAS,IAAI;AAC1B,QAAI,cAAc,IAAI,KAAK,eAAe,IAAI,KAAK,YAAY,IAAI,GAAG;AACpE,UAAI,eAAe,MAAM;AACvB,sBAAc;AACd,qBAAa,YAAY,SAAS,IAAI,CAAC,UAAU,MAAM,IAAI,EAAE,OAAO,CAAC,SAAS,UAAU,IAAI,IAAI,CAAC;AAAA,MACnG;AAAA,IACF;AACA,cAAU,IAAI,KAAK,IAAI;AACvB,QAAI,UAAU,KAAK,SAAS,MAAM;AAChC;AAAA,IACF;AACA,QAAI,eAAe,QAAQ,KAAK,IAAI,MAAM,IAAI;AAC5C;AAAA,IACF;AACA,QAAI,cAAc,QAAQ,KAAK,IAAI,MAAM,IAAI;AAC3C;AAAA,IACF;AACA,QAAI,KAAK,OAAO,WAAW,GAAG;AAC5B,oBAAc,KAAK,KAAK,IAAI;AAC5B;AAAA,IACF;AACA,SAAK,OAAO,QAAQ,CAAC,WAAW;AAC9B,UAAI,KAAK,IAAI,OAAO,IAAI,GAAG;AACzB;AAAA,MACF;AACA,WAAK,IAAI,OAAO,IAAI;AACpB,eAAS,KAAK,MAAM;AAAA,IACtB,CAAC;AAAA,EACH;AACA,SAAO,EAAE,QAAQ,SAAS,WAAW,eAAe,aAAa,WAAW;AAC9E;AACA,SAAS,2BAA2B,OAAO,WAAW,eAAe;AACnE,QAAM,EAAE,WAAW,OAAO,IAAI;AAC9B,QAAM,WAAW,CAAC;AAClB,QAAM,aAAa,OAAO,KAAK,MAAM,EAAE,IAAI,CAAC,SAAS,cAAc,IAAI,EAAE,EAAE,EAAE,IAAI,CAAC,SAAS,MAAM,MAAM,KAAK;AAC5G,QAAM,YAAY,MAAM;AACxB,aAAW,QAAQ,CAAC,WAAW;AAC7B,QAAI,UAAU,IAAI,OAAO,IAAI,GAAG;AAC9B,eAAS,KAAK,MAAM;AAAA,IACtB;AAAA,EACF,CAAC;AACD,QAAM,QAAQ,QAAQ,CAAC,WAAW;AAChC,QAAI,UAAU,IAAI,OAAO,IAAI,GAAG;AAC9B,eAAS,KAAK,MAAM;AAAA,IACtB;AAAA,EACF,CAAC;AACD,MAAI,aAAa,MAAM;AACrB,cAAU,QAAQ,CAAC,SAAS;AAC1B,UAAI,UAAU,IAAI,KAAK,IAAI,GAAG;AAC5B,iBAAS,KAAK,IAAI;AAAA,MACpB;AAAA,IACF,CAAC;AAAA,EACH;AACA,QAAM,OAAuB,oBAAI,IAAI;AACrC,QAAM,eAAe,CAAC;AACtB,SAAO,SAAS,SAAS,GAAG;AAC1B,UAAM,OAAO,SAAS,IAAI;AAC1B,SAAK,IAAI,KAAK,IAAI;AAClB,QAAI,CAAC,UAAU,KAAK,OAAO;AACzB,mBAAa,KAAK,IAAI;AAAA,IACxB;AACA,SAAK,SAAS,QAAQ,CAAC,UAAU;AAC/B,UAAI,CAAC,KAAK,IAAI,MAAM,IAAI,KAAK,UAAU,IAAI,MAAM,IAAI,KAAK,MAAM,OAAO,MAAM,CAAC,WAAW,KAAK,IAAI,OAAO,IAAI,CAAC,GAAG;AAC/G,iBAAS,KAAK,KAAK;AAAA,MACrB;AAAA,IACF,CAAC;AAAA,EACH;AACA,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,IAAI,oBAAoB;AAAA,EACtB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,IAAI,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,SAAS,cAAc,MAAM;AAC3B,SAAO,iBAAiB,QAAQ,KAAK,EAAE,KAAK;AAC9C;AACA,SAAS,eAAe,MAAM;AAC5B,SAAO,kBAAkB,QAAQ,KAAK,EAAE,KAAK;AAC/C;AACA,SAAS,YAAY,MAAM;AACzB,SAAO,eAAe,QAAQ,KAAK,EAAE,KAAK;AAC5C;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,OAAO,QAAQ;AACzB,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,cAA8B,oBAAI,IAAI;AAC3C,SAAK,aAAa,CAAC;AACnB,SAAK,YAAY;AACjB,SAAK,aAAa,CAAC;AACnB,SAAK,uBAAuB,CAAC;AAC7B,SAAK,sBAAsB,CAAC;AAC5B,SAAK,qBAAqB;AAC1B,SAAK,WAAW,MAAM;AACtB,SAAK,UAAU,MAAM;AACrB,SAAK,aAAa,MAAM;AACxB,SAAK,aAAa,MAAM;AACxB,SAAK,aAAa,MAAM;AACxB,QAAI,MAAM,aAAa,MAAM;AAC3B,aAAO,KAAK,MAAM,SAAS,EAAE,QAAQ,CAAC,SAAS;AAC7C,aAAK,qBAAqB,QAAQ,IAAI,cAAc,MAAM,UAAU,OAAO,IAAI;AAAA,MACjF,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK,SAAS,KAAK,OAAO,YAAY,KAAK;AAAA,EACpD;AAAA,EACA,IAAI,sBAAsB;AACxB,WAAO,KAAK,SAAS,KAAK,OAAO,sBAAsB,KAAK;AAAA,EAC9D;AAAA,EACA,IAAI,YAAY;AACd,WAAO,KAAK,SAAS,KAAK,OAAO,YAAY,KAAK;AAAA,EACpD;AAAA,EACA,IAAI,UAAU,WAAW;AACvB,UAAM,YAAY,OAAO,KAAK,SAAS,EAAE,IAAI,CAAC,QAAQ,UAAU,KAAK,IAAI,CAAC,YAAY,QAAQ,EAAE,CAAC;AACjG,SAAK,aAAa,CAAC,EAAE,OAAO,GAAG,SAAS;AACxC,SAAK,aAAa;AAAA,EACpB;AAAA,EACA,IAAI,gBAAgB,iBAAiB;AACnC,SAAK,mBAAmB;AAAA,EAC1B;AAAA,EACA,IAAI,SAAS;AACX,WAAO,KAAK,QAAQ,IAAI,CAAC,SAAS;AAChC,aAAO;AAAA,QACL,MAAM,KAAK;AAAA,QACX,OAAO,KAAK,WAAW,WAAW,KAAK,WAAW,SAAS,QAAQ;AAAA,QACnE,OAAO,KAAK,WAAW,WAAW,KAAK,WAAW,SAAS,QAAQ;AAAA,MACrE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,SAAS,IAAI,CAAC,SAAS;AACjC,aAAO;AAAA,QACL,MAAM,KAAK;AAAA,QACX,OAAO,KAAK,WAAW,WAAW,KAAK,WAAW,SAAS,QAAQ;AAAA,QACnE,OAAO,KAAK,WAAW,WAAW,KAAK,WAAW,SAAS,QAAQ;AAAA,MACrE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,IAAI,aAAa;AACf,WAAO,KAAK,QAAQ,IAAI,CAAC,SAAS,KAAK,gBAAgB,KAAK,IAAI;AAAA,EAClE;AAAA,EACA,IAAI,cAAc;AAChB,WAAO,KAAK,SAAS,IAAI,CAAC,SAAS;AACjC,YAAM,OAAO,KAAK,gBAAgB,KAAK;AACvC,aAAO,KAAK,gBAAgB,GAAG,QAAQ,KAAK,kBAAkB;AAAA,IAChE,CAAC;AAAA,EACH;AAAA,EACA,IAAI,YAAY;AACd,WAAO,OAAO,KAAK,KAAK,UAAU,EAAE,OAAO,CAAC,KAAK,QAAQ;AACvD,UAAI,OAAO,KAAK,WAAW,KAAK;AAChC,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,kBAAkB,QAAQ,SAAS;AACjC,UAAM,eAAe,OAAO,IAAI,CAAC,SAAS,KAAK,IAAI,EAAE,KAAK;AAC1D,UAAM,gBAAgB,QAAQ,IAAI,CAAC,SAAS,KAAK,IAAI,EAAE,KAAK;AAC5D,WAAO,aAAa,KAAK,KAAK,SAAS,IAAI,OAAO,cAAc,KAAK,KAAK,SAAS;AAAA,EACrF;AAAA,EACA,QAAQ,QAAQ,SAAS;AACvB,UAAM,gBAAgB,qBAAqB,QAAQ,SAAS,KAAK,WAAW,KAAK,UAAU;AAC3F,UAAM,EAAE,eAAe,aAAa,WAAW,IAAI;AACnD,QAAI,eAAe,MAAM;AACvB,YAAM,IAAI,MAAM,qCAAqC,YAAY,oCAAoC,YAAY,8GAA8G,aAAa;AAAA,IAC9O;AACA,QAAI,cAAc,SAAS,GAAG;AAC5B,YAAM,WAAW,QAAQ,IAAI,CAAC,MAAM,EAAE,IAAI;AAC1C,YAAM,UAAU,OAAO,KAAK,MAAM;AAClC,YAAM,IAAI,MAAM,+BAA+B,uCAAuC,4CAA4C,gBAAgB;AAAA,IACpJ;AACA,WAAO,2BAA2B,KAAK,OAAO,KAAK,WAAW,aAAa;AAAA,EAC7E;AAAA,EACA,QAAQ,QAAQ,SAAS;AACvB,aAAS,KAAK,UAAU,MAAM;AAC9B,UAAM,QAAQ,OAAO,KAAK,MAAM,EAAE,KAAK;AACvC,SAAK,YAAY,MAAM;AACvB,SAAK,uBAAuB,MAAM;AAClC,cAAU,KAAK,WAAW,OAAO;AACjC,SAAK,aAAa,OAAO;AACzB,UAAM,aAAa,MAAM,IAAI,CAAC,SAAS,KAAK,MAAM,MAAM,cAAc,IAAI,EAAE,GAAG;AAC/E,UAAM,kBAAkB,QAAQ,IAAI,CAAC,SAAS,cAAc,IAAI,EAAE,EAAE;AACpE,QAAI,cAAc,gBAAgB,IAAI,CAAC,SAAS,KAAK,MAAM,MAAM,KAAK;AACtE,SAAK,yBAAyB;AAC9B,QAAI,YAAY,WAAW,GAAG;AAC5B,oBAAc,KAAK;AAAA,IACrB;AACA,UAAM,iBAAiB,KAAK,kBAAkB,YAAY,WAAW;AACrE,QAAI,eAAe,KAAK,YAAY,IAAI,cAAc;AACtD,QAAI,gBAAgB,MAAM;AACxB,qBAAe,KAAK,QAAQ,QAAQ,WAAW;AAC/C,WAAK,YAAY,IAAI,gBAAgB,YAAY;AAAA,IACnD;AACA,UAAM,iBAAiB,CAAC;AACxB,UAAM,gBAAgB,CAAC;AACvB,WAAO,KAAK,MAAM;AAChB,YAAM,UAAU,IAAI,iBAAiB,KAAK,WAAW,gBAAgB,eAAe,KAAK,mBAAmB;AAC5G,YAAM,aAAa,OAAO,OAAO,CAAC,GAAG,KAAK,SAAS;AACnD,aAAO,KAAK,MAAM,EAAE,QAAQ,CAAC,SAAS;AACpC,cAAM,CAAC,UAAU,KAAK,IAAI,cAAc,IAAI;AAC5C,cAAM,UAAU,CAAC;AACjB,gBAAQ,SAAS,OAAO;AACxB,mBAAW,YAAY;AAAA,MACzB,CAAC;AACD,YAAM,gBAAgB,KAAK,mBAAmB,UAAU;AACxD,YAAM,kCAAkC,CAAC;AACzC,eAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,KAAK;AAC5C,cAAM,OAAO,aAAa;AAC1B,YAAI,CAAC,WAAW,KAAK,OAAO;AAC1B,gBAAM,UAAU,YAAY,MAAM,YAAY,SAAS,KAAK,gBAAgB;AAC5E,cAAI,aAAa,UAAU,OAAO,GAAG;AACnC,kBAAM,IAAI,MAAM,4BAA4B,KAAK,kEAAkE;AAAA,UACrH;AACA,qBAAW,KAAK,QAAQ;AACxB,eAAK,uBAAuB,KAAK,MAAM,MAAM,YAAY,SAAS,eAAe,iBAAiB,+BAA+B;AAAA,QACnI;AAAA,MACF;AACA,UAAI,KAAK,UAAU,MAAM;AACvB,gBAAQ,QAAQ,aAAa;AAAA,MAC/B;AACA,aAAO,QAAQ,IAAI,CAAC,SAAS,UAAU,MAAM,YAAY,OAAO,CAAC;AAAA,IACnE,CAAC;AAAA,EACH;AAAA,EACA,mBAAmB,WAAW;AAC5B,UAAM,MAAM,CAAC,EAAE,OAAO,MAAM,CAAC,GAAG,OAAO,KAAK,SAAS,EAAE,IAAI,CAAC,QAAQ,UAAU,IAAI,EAAE,IAAI,CAAC,YAAY,QAAQ,IAAI,CAAC,YAAY,QAAQ,EAAE,CAAC,CAAC;AAC1I,WAAO,IAAI,IAAI,GAAG;AAAA,EACpB;AAAA,EACA,uBAAuB,UAAU,MAAM,WAAW,SAAS,eAAe,aAAa,iCAAiC;AACtH,QAAI,KAAK,aAAa,aAAa,YAAY,QAAQ,QAAQ,MAAM,IAAI;AACvE;AAAA,IACF;AACA,cAAU,UAAU,QAAQ,CAAC,YAAY;AACvC,UAAI,WAAW,MAAM;AACnB,wCAAgC,QAAQ,OAAO,gCAAgC,QAAQ,OAAO,KAAK,KAAK,SAAS;AAAA,MACnH;AAAA,IACF,CAAC;AACD,SAAK,OAAO,QAAQ,CAAC,WAAW;AAC9B,UAAI,OAAO,aAAa,WAAW;AACjC,cAAM,UAAU,6BAA6B,OAAO,MAAM,WAAW,OAAO;AAC5E,YAAI,WAAW,MAAM;AACnB,kBAAQ,QAAQ,CAAC,YAAY;AAC3B,gBAAI,WAAW,CAAC,QAAQ,QAAQ,CAAC,cAAc,IAAI,QAAQ,EAAE,GAAG;AAC9D,oBAAM,SAAS,gCAAgC,QAAQ;AACvD,kBAAI,WAAW,GAAG;AAChB,oBAAI,CAAC,KAAK,oBAAoB;AAC5B,0BAAQ,QAAQ;AAAA,gBAClB,OAAO;AACL,wBAAM,CAAC,WAAW,KAAK,IAAI,oBAAoB,KAAK,MAAM,OAAO;AACjE,sBAAI,KAAK,oBAAoB,YAAY;AACvC,yBAAK,oBAAoB,WAAW,SAAS;AAAA,kBAC/C,OAAO;AACL,yBAAK,oBAAoB,aAAa,CAAC;AACvC,yBAAK,oBAAoB,WAAW,SAAS;AAAA,kBAC/C;AAAA,gBACF;AACA,uBAAO,gCAAgC,QAAQ;AAAA,cACjD,WAAW,UAAU,MAAM;AACzB,gDAAgC,QAAQ;AAAA,cAC1C;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,MAAM,aAAa,QAAQ,SAAS;AAClC,WAAO,KAAK,cAAc,QAAQ,OAAO;AAAA,EAC3C;AAAA,EACA,6BAA6B;AAC3B,QAAI,CAAC,KAAK,qBAAqB;AAC7B;AAAA,IACF;AACA,WAAO,KAAK,KAAK,mBAAmB,EAAE,QAAQ,CAAC,QAAQ,KAAK,oBAAoB,KAAK,QAAQ,CAAC,YAAY,QAAQ,QAAQ,CAAC,CAAC;AAC5H,SAAK,kBAAkB;AAAA,EACzB;AAAA,EACA,oBAAoB;AAClB,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AACA,WAAO,KAAK,KAAK,UAAU,EAAE,QAAQ,CAAC,QAAQ;AAC5C,YAAM,cAAc,KAAK,WAAW;AACpC,kBAAY,QAAQ,CAAC,YAAY;AAC/B,YAAI,WAAW,CAAC,QAAQ,QAAQ,CAAC,QAAQ,cAAc,CAAC,KAAK,QAAQ,IAAI,QAAQ,EAAE,GAAG;AACpF,kBAAQ,QAAQ;AAAA,QAClB;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAAA,EACH;AAAA,EACA,yBAAyB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,2BAA2B;AACzB,eAAW,OAAO,KAAK,qBAAqB;AAC1C,WAAK,oBAAoB,KAAK,QAAQ,CAAC,YAAY,QAAQ,QAAQ,CAAC;AACpE,aAAO,KAAK,oBAAoB;AAAA,IAClC;AAAA,EACF;AAAA,EACA,MAAM,cAAc,QAAQ,SAAS,sBAAsB,OAAO,iBAAiB,CAAC,GAAG,gBAAgB,CAAC,GAAG;AACzG,QAAI,CAAC,qBAAqB;AACxB,eAAS,KAAK,UAAU,MAAM;AAC9B,WAAK,YAAY,MAAM;AACvB,WAAK,uBAAuB,MAAM;AAClC,gBAAU,KAAK,WAAW,OAAO;AACjC,WAAK,aAAa,OAAO;AAAA,IAC3B;AACA,QAAI;AACF,WAAK,qBAAqB,IAAI,EAAE,QAAQ,2BAA2B;AAAA,IACrE,SAAS,GAAP;AACA,cAAQ,KAAK,EAAE,OAAO;AAAA,IACxB;AACA,SAAK,yBAAyB;AAC9B,UAAM,UAAU,IAAI,iBAAiB,KAAK,WAAW,gBAAgB,eAAe,KAAK,mBAAmB;AAC5G,SAAK,aAAa,MAAM,KAAK,uBAAuB,QAAQ,SAAS,SAAS,mBAAmB;AACjG,UAAM,UAAU,QAAQ,IAAI,CAAC,SAAS,UAAU,MAAM,KAAK,YAAY,OAAO,CAAC;AAC/E,UAAM,YAAY,QAAQ,IAAI,CAAC,MAAM,EAAE,EAAE;AACzC,UAAM,WAAW,OAAO,KAAK,MAAM,EAAE,IAAI,CAAC,SAAS,OAAO,MAAM,EAAE;AAClE,SAAK,UAA0B,oBAAI,IAAI,CAAC,GAAG,WAAW,GAAG,UAAU,GAAG,KAAK,SAAS,CAAC;AACrF,QAAI,CAAC,KAAK,oBAAoB;AAC5B,WAAK,kBAAkB;AAAA,IACzB;AACA,QAAI,KAAK,UAAU,MAAM;AACvB,cAAQ,QAAQ,KAAK,OAAO;AAAA,IAC9B;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,qBAAqB,QAAQ,gBAAgB,eAAe;AAChE,UAAM,eAAe,OAAO,OAAO,CAAC,KAAK,SAAS,UAAU;AAC1D,UAAI,KAAK,OAAO,OAAO,QAAQ;AAC/B,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AACL,WAAO,KAAK,cAAc,cAAc,KAAK,aAAa,MAAM,gBAAgB,aAAa;AAAA,EAC/F;AAAA,EACA,MAAM,uBAAuB,QAAQ,SAAS,aAAa,qBAAqB;AAC9E,UAAM,QAAQ,OAAO,KAAK,MAAM;AAChC,UAAM,aAAa,MAAM,IAAI,CAAC,SAAS,KAAK,MAAM,MAAM,cAAc,IAAI,EAAE,GAAG;AAC/E,UAAM,kBAAkB,YAAY,IAAI,CAAC,SAAS,cAAc,IAAI,EAAE,EAAE;AACxE,QAAI,cAAc,gBAAgB,IAAI,CAAC,SAAS,KAAK,MAAM,MAAM,KAAK;AACtE,QAAI,YAAY,WAAW,GAAG;AAC5B,oBAAc,KAAK;AAAA,IACrB;AACA,UAAM,EAAE,WAAW,eAAe,aAAa,WAAW,IAAI,qBAAqB,QAAQ,aAAa,KAAK,WAAW,KAAK,UAAU;AACvI,UAAM,SAAS;AAAA,MACb,GAAG;AAAA,MACH,GAAG,KAAK,MAAM;AAAA,MACd,GAAG,KAAK,cAAc,CAAC;AAAA,IACzB,EAAE,IAAI,CAAC,SAAS;AACd,aAAO,EAAE,MAAM,UAAU,QAAQ,eAAe;AAAA,IAClD,CAAC;AACD,UAAM,aAAa,OAAO,OAAO,CAAC,GAAG,KAAK,SAAS;AACnD,WAAO,KAAK,MAAM,EAAE,QAAQ,CAAC,SAAS;AACpC,YAAM,CAAC,UAAU,KAAK,IAAI,cAAc,IAAI;AAC5C,YAAM,UAAU,CAAC;AACjB,cAAQ,SAAS,OAAO;AACxB,iBAAW,YAAY;AAAA,IACzB,CAAC;AACD,UAAM,kCAAkC,CAAC;AACzC,UAAM,gBAAgB,KAAK,mBAAmB,UAAU;AACxD,UAAM,QAAQ,CAAC;AACf,WAAO,OAAO,SAAS,GAAG;AACxB,YAAM,WAAW,KAAK,aAAa,YAAY,QAAQ,SAAS,YAAY,OAAO,eAAe,iBAAiB,iCAAiC,SAAS;AAC7J,YAAM,QAAQ,IAAI,QAAQ;AAAA,IAC5B;AACA,QAAI,eAAe,QAAQ,CAAC,qBAAqB;AAC/C,cAAQ,KAAK,iIAAiI;AAAA,IAChJ;AACA,UAAM,iBAAiB,YAAY,OAAO,CAAC,SAAS,CAAC,cAAc,IAAI,KAAK,CAAC,UAAU,KAAK,MAAM,YAAY,OAAO,CAAC,EAAE,IAAI,CAAC,SAAS,KAAK,IAAI;AAC/I,QAAI,eAAe,SAAS,GAAG;AAC7B,UAAI,iBAAiB;AACrB,UAAI,eAAe,MAAM;AACvB,yBAAiB,wFAAwF;AAAA,MAC3G;AACA,YAAM,IAAI,MAAM,+BAA+B,6CAA6C,qDAAqD,mBAAmB,gBAAgB;AAAA,IACtL;AACA,WAAO;AAAA,EACT;AAAA,EACA,aAAa,YAAY,QAAQ,SAAS,WAAW,OAAO,eAAe,aAAa,iCAAiC,WAAW;AAClI,UAAM,WAAW,CAAC;AAClB,WAAO,OAAO,SAAS,GAAG;AACxB,YAAM,OAAO,OAAO,IAAI;AACxB,cAAQ,iBAAiB,KAAK;AAC9B,UAAI,WAAW;AACf,UAAI,KAAK,KAAK,OAAO,WAAW,cAAc,cAAc,KAAK,MAAM,WAAW,OAAO,GAAG;AAC1F,SAAC,QAAQ,IAAI,oBAAoB,KAAK,KAAK,MAAM,OAAO;AAAA,MAC1D;AACA,UAAI,UAAU,KAAK,KAAK,SAAS,MAAM;AACrC,cAAM,UAAU,YAAY,KAAK,MAAM,WAAW,SAAS,KAAK,gBAAgB;AAChF,YAAI,CAAC,UAAU;AACb,WAAC,QAAQ,IAAI,oBAAoB,KAAK,KAAK,MAAM,OAAO;AAAA,QAC1D;AACA,cAAM,iBAAiB,QAAQ;AAC/B,YAAI,aAAa,UAAU,OAAO,GAAG;AACnC,mBAAS,KAAK,QAAQ,KAAK,CAAC,MAAM;AAChC,sBAAU,YAAY;AACtB,oBAAQ,iBAAiB;AACzB,iBAAK,uBAAuB,UAAU,KAAK,MAAM,WAAW,SAAS,eAAe,aAAa,+BAA+B;AAChI,iBAAK,kBAAkB,KAAK,MAAM,QAAQ,SAAS,WAAW,OAAO,SAAS;AAC9E,mBAAO;AAAA,UACT,CAAC,CAAC;AAAA,QACJ,OAAO;AACL,oBAAU,YAAY;AACtB,eAAK,uBAAuB,UAAU,KAAK,MAAM,WAAW,SAAS,eAAe,aAAa,+BAA+B;AAChI,eAAK,kBAAkB,KAAK,MAAM,QAAQ,SAAS,WAAW,OAAO,SAAS;AAAA,QAChF;AAAA,MACF,OAAO;AACL,aAAK,kBAAkB,KAAK,MAAM,QAAQ,SAAS,WAAW,OAAO,SAAS;AAAA,MAChF;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,kBAAkB,MAAM,QAAQ,SAAS,WAAW,OAAO,WAAW;AACpE,SAAK,SAAS,QAAQ,CAAC,cAAc;AACnC,YAAM,CAAC,QAAQ,IAAI,oBAAoB,UAAU,MAAM,OAAO;AAC9D,UAAI,MAAM,aAAa,CAAC,UAAU,IAAI,UAAU,IAAI,GAAG;AACrD;AAAA,MACF;AACA,UAAI,UAAU,OAAO,SAAS;AAC5B,YAAI,UAAU,WAAW,KAAK,CAAC,SAAS;AACtC,iBAAO,CAAC,CAAC,UAAU,MAAM,WAAW,OAAO;AAAA,QAC7C,CAAC,GAAG;AACF,gBAAM,YAAY;AAClB,iBAAO,KAAK,EAAE,UAAU,QAAQ,gBAAgB,MAAM,UAAU,CAAC;AAAA,QACnE;AAAA,MACF,WAAW,UAAU,WAAW,MAAM,CAAC,SAAS;AAC9C,eAAO,CAAC,CAAC,UAAU,MAAM,WAAW,OAAO;AAAA,MAC7C,CAAC,GAAG;AACF,cAAM,YAAY;AAClB,eAAO,KAAK,EAAE,UAAU,QAAQ,gBAAgB,MAAM,UAAU,CAAC;AAAA,MACnE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,UAAU;AACR,WAAO,KAAK,KAAK,SAAS,EAAE,QAAQ,CAAC,QAAQ,KAAK,UAAU,KAAK,QAAQ,CAAC,YAAY,QAAQ,QAAQ,CAAC,CAAC;AAAA,EAC1G;AAAA,EACA,uBAAuB,QAAQ;AAC7B,WAAO,KAAK,MAAM,EAAE,QAAQ,CAAC,SAAS;AACpC,YAAM,SAAS,OAAO;AACtB,YAAM,CAAC,QAAQ,IAAI,cAAc,IAAI;AACrC,YAAM,OAAO,KAAK,MAAM,MAAM;AAC9B,UAAI,KAAK,WAAW,YAAY,KAAK,WAAW,SAAS,OAAO;AAC9D,cAAM,QAAQ,KAAK,WAAW,SAAS;AACvC,cAAM,QAAQ,MAAM,WAAW,OAAO,MAAM,UAAU,OAAO,MAAM,MAAM,CAAC,KAAK,UAAU,MAAM,WAAW,MAAM,MAAM,WAAW,GAAG;AACpI,qBAAa,OAAO,OAAO,MAAM,sBAAsB,KAAK,mDAAmD,oBAAoB,OAAO,QAAQ;AAAA,MACpJ;AACA,UAAI,KAAK,WAAW,YAAY,KAAK,WAAW,SAAS,OAAO;AAC9D,qBAAa,OAAO,OAAO,UAAU,KAAK,WAAW,SAAS,OAAO,MAAM,sBAAsB,KAAK,kDAAkD,KAAK,WAAW,SAAS,kBAAkB,OAAO,OAAO;AAAA,MACnN;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,UAAU,QAAQ;AAChB,UAAM,SAAS,CAAC;AAChB,eAAW,aAAa,QAAQ;AAC9B,UAAI,KAAK,cAAc,QAAQ,KAAK,WAAW,UAAU,QAAQ,KAAK,WAAW,OAAO,cAAc,MAAM;AAC1G,cAAM,UAAU,KAAK,WAAW,OAAO;AACvC,eAAO,QAAQ,QAAQ,OAAO;AAAA,MAChC,OAAO;AACL,eAAO,aAAa,OAAO;AAAA,MAC7B;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,YAAY,QAAQ;AAClB,UAAM,aAAa,OAAO,KAAK,MAAM,EAAE,OAAO,CAAC,SAAS;AACtD,YAAM,CAAC,QAAQ,IAAI,cAAc,IAAI;AACrC,aAAO,KAAK,MAAM,MAAM,aAAa;AAAA,IACvC,CAAC;AACD,QAAI,WAAW,SAAS,GAAG;AACzB,YAAM,IAAI,MAAM,uDAAuD,wCAAwC;AAAA,IACjH;AAAA,EACF;AAAA,EACA,WAAW,SAAS;AAClB,WAAO,QAAQ,IAAI,CAAC,SAAS;AAC3B,UAAI,KAAK,cAAc,QAAQ,KAAK,WAAW,WAAW,QAAQ,KAAK,WAAW,QAAQ,SAAS,MAAM;AACvG,cAAM,UAAU,KAAK,WAAW,QAAQ;AACxC,eAAO,QAAQ;AAAA,MACjB;AACA,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,aAAa,SAAS;AACpB,YAAQ,QAAQ,CAAC,SAAS;AACxB,YAAM,CAAC,cAAc,IAAI,cAAc,IAAI;AAC3C,UAAI,CAAC,KAAK,MAAM,MAAM,iBAAiB;AACrC,cAAM,IAAI,MAAM,eAAe,iCAAiC;AAAA,MAClE;AAAA,IACF,CAAC;AAAA,EACH;AACF;AAGA,IAAI,kBAAkB,MAAM;AAAA,EAC1B,YAAY,wBAAwB,CAAC,GAAG,eAAe,CAAC,GAAG;AACzD,SAAK,wBAAwB;AAC7B,SAAK,eAAe;AAAA,EACtB;AAAA,EACA,aAAa,MAAM,WAAW;AAC5B,SAAK,sBAAsB,QAAQ,UAAU;AAC7C,SAAK,aAAa,UAAU,MAAM;AAAA,EACpC;AAAA,EACA,yBAAyB,MAAM;AAC7B,WAAO,KAAK,sBAAsB;AAAA,EACpC;AAAA,EACA,iBAAiB,IAAI;AACnB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EACA,UAAU;AACR,eAAW,OAAO,KAAK,cAAc;AACnC,WAAK,aAAa,KAAK,cAAc;AACrC,aAAO,KAAK,aAAa;AAAA,IAC3B;AACA,eAAW,QAAQ,KAAK,uBAAuB;AAC7C,WAAK,sBAAsB,MAAM,QAAQ;AACzC,aAAO,KAAK,sBAAsB;AAAA,IACpC;AAAA,EACF;AACF;AAGA,IAAI,qBAAqB;AACzB,IAAI,qBAAqB;AACzB,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,UAAU,cAAc,CAAC,GAAG,OAAO,YAAY;AACzD,SAAK,WAAW;AAChB,SAAK,cAAc;AACnB,SAAK,UAAU;AACf,SAAK,KAAK;AACV,QAAI,eAAe,MAAM;AACvB,WAAK,cAAc,CAAC;AAAA,IACtB;AACA,SAAK,kBAAkB,IAAI,gBAAgB;AAAA,EAC7C;AAAA,EACA,IAAI,eAAe;AACjB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,aAAa;AACf,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,cAAc;AAChB,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,SAAS;AACX,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,UAAU;AACZ,WAAO,KAAK,SAAS;AAAA,EACvB;AAAA,EACA,IAAI,WAAW;AACb,WAAO,KAAK,UAAU;AAAA,EACxB;AAAA,EACA,IAAI,iBAAiB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,4BAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EACA,gBAAgB;AACd,UAAM,OAAO,KAAK;AAClB,QAAI,KAAK,QAAQ,MAAM;AACrB,WAAK,UAAU;AAAA,IACjB,WAAW,KAAK,YAAY,eAAe,MAAM;AAC/C,WAAK,UAAU,KAAK,GAAG,mBAAmB,MAAM,KAAK,WAAW;AAAA,IAClE,OAAO;AACL,YAAM,WAAW,KAAK,GAAG,gBAAgB,MAAM,KAAK,WAAW;AAC/D,UAAI,SAAS,WAAW,GAAG;AACzB,iBAAS,KAAK,KAAK,GAAG,mBAAmB,MAAM,KAAK,WAAW,CAAC;AAAA,MAClE,WAAW,SAAS,SAAS,GAAG;AAC9B,cAAM,IAAI,MAAM,wBAAwB,SAAS,kCAAkC,CAAC,IAAI,IAAI;AAAA,MAC9F;AACA,WAAK,UAAU,SAAS;AAAA,IAC1B;AAAA,EACF;AAAA,EACA,OAAO;AACL,SAAK,cAAc;AACnB,QAAI,KAAK,QAAQ,QAAQ,MAAM;AAC7B,YAAM,IAAI,MAAM,+GAA+G;AAAA,IACjI;AACA,UAAM,aAAa,KAAK,QAAQ,KAAK;AACrC,QAAI,aAAa,UAAU,UAAU,GAAG;AACtC,aAAO,WAAW,KAAK,CAAC,cAAc,KAAK,SAAS,SAAS,CAAC;AAAA,IAChE;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EACA,SAAS,WAAW;AAClB,SAAK,YAAY;AACjB,UAAM,QAAQ,KAAK,UAAU;AAC7B,QAAI,YAAY,KAAK,UAAU;AAC/B,QAAI,KAAK,UAAU,uBAAuB,MAAM;AAC9C,YAAM,WAAW,KAAK,UAAU;AAChC,UAAI,SAAS,aAAa,MAAM;AAC9B,oBAAY,SAAS;AAAA,MACvB;AACA,UAAI,SAAS,wBAAwB,MAAM;AACzC,aAAK,uBAAuB,SAAS;AAAA,MACvC;AAAA,IACF;AACA,SAAK,YAAY;AACjB,SAAK,UAAU,GAAG,MAAM,SAAS,YAAY,MAAM,SAAS;AAC5D,UAAM,YAAY,KAAK,GAAG,cAAc,KAAK,UAAU,YAAY,KAAK,UAAU,WAAW;AAC7F,SAAK,WAAW,IAAI,cAAc,gBAAgB,SAAS,eAAe,OAAO,KAAK,SAAS,CAAC;AAChG,SAAK,SAAS,YAAY,KAAK,6BAA6B,SAAS;AACrE,SAAK,SAAS,kBAAkB,KAAK;AACrC,QAAI,UAAU,oBAAoB,QAAQ,UAAU,iBAAiB,QAAQ,MAAM;AACjF,YAAM,cAAc,gBAAgB,SAAS,eAAe,UAAU,gBAAgB;AACtF,WAAK,cAAc,IAAI,cAAc,WAAW;AAChD,WAAK,YAAY,YAAY,KAAK,SAAS;AAC3C,WAAK,YAAY,kBAAkB,KAAK;AACxC,WAAK,uBAAuB,UAAU;AAAA,IACxC;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,KAAK,cAAc,QAAQ;AAC/B,QAAI,OAAO,iBAAiB,UAAU;AACpC,YAAM,WAAW,KAAK,GAAG,gBAAgB,YAAY;AACrD,UAAI,SAAS,WAAW,GAAG;AACzB,cAAM,IAAI,MAAM,0CAA0C,eAAe;AAAA,MAC3E,WAAW,SAAS,SAAS,GAAG;AAC9B,cAAM,IAAI,MAAM,wBAAwB,SAAS,kCAAkC,eAAe;AAAA,MACpG;AACA,qBAAe,SAAS;AAAA,IAC1B;AACA,QAAI,aAAa,QAAQ,MAAM;AAC7B,YAAM,IAAI,MAAM,6GAA6G;AAAA,IAC/H;AACA,WAAO,aAAa,KAAK,KAAK,SAAS;AAAA,EACzC;AAAA,EACA,QAAQ,QAAQ,QAAQ;AACtB,UAAM,gBAAgB,KAAK,QAAQ,QAAQ,KAAK,WAAW;AAC3D,QAAI,KAAK,sBAAsB;AAC7B,YAAM,qBAAqB,yBAAyB,SAAS,CAAC,aAAa,IAAI;AAC/E,YAAM,kBAAkB,CAAC;AACzB,yBAAmB,QAAQ,CAAC,cAAc,MAAM,gBAAgB,KAAK,qBAAqB,MAAM,YAAY;AAC5G,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAAA,EACA,gBAAgB,QAAQ;AACtB,QAAI,EAAE,kBAAkB,WAAW,CAAC,MAAM,QAAQ,MAAM,GAAG;AACzD,UAAI,KAAK,aAAa,QAAQ,KAAK,UAAU,UAAU,MAAM;AAC3D,mBAAW,UAAU,KAAK,UAAU,QAAQ;AAC1C,gBAAM,UAAU,KAAK,UAAU,OAAO;AACtC,cAAI,QAAQ,cAAc,MAAM;AAC9B,mBAAO,UAAU,KAAK,0BAA0B,QAAQ;AAAA,UAC1D;AAAA,QACF;AAAA,MACF;AACA,aAAO;AAAA,IACT;AACA,aAAS,MAAM,QAAQ,MAAM,IAAI,SAAS,CAAC,MAAM;AACjD,UAAM,oBAAoB,OAAO,KAAK,KAAK,yBAAyB,EAAE;AACtE,QAAI,OAAO,SAAS,sBAAsB,KAAK,WAAW,QAAQ;AAChE,YAAM,IAAI,MAAM,oDAAoD,KAAK,WAAW,SAAS,gEAAgE,OAAO,gCAAgC;AAAA,IACtM;AACA,QAAI,aAAa;AACjB,WAAO,KAAK,WAAW,OAAO,CAAC,KAAK,cAAc;AAChD,YAAM,YAAY,KAAK,YAAY,KAAK,UAAU,OAAO,aAAa;AACtE,UAAI,aAAa,QAAQ,UAAU,cAAc,MAAM;AACrD,YAAI,aAAa,KAAK,0BAA0B,UAAU;AAAA,MAC5D,OAAO;AACL,YAAI,aAAa,OAAO;AAAA,MAC1B;AACA,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,iBAAiB,SAAS;AACxB,cAAU,WAAW,KAAK;AAC1B,WAAO,CAAC,MAAM,QAAQ,OAAO,IAAI,CAAC,OAAO,IAAI;AAAA,EAC/C;AAAA,EACA,0BAA0B;AACxB,QAAI,KAAK,eAAe,MAAM;AAC5B,aAAO,CAAC;AAAA,IACV;AACA,QAAI,KAAK,wBAAwB,MAAM;AACrC,aAAO,KAAK,YAAY,QAAQ,CAAC,GAAG,CAAC,CAAC;AAAA,IACxC,OAAO;AACL,aAAO,KAAK,YAAY,QAAQ,CAAC,GAAG,OAAO,KAAK,KAAK,qBAAqB,OAAO,CAAC;AAAA,IACpF;AAAA,EACF;AAAA,EACA,MAAM,+BAA+B;AACnC,QAAI,KAAK,eAAe,MAAM;AAC5B,aAAO,CAAC;AAAA,IACV;AACA,QAAI,KAAK,wBAAwB,MAAM;AACrC,aAAO,KAAK,YAAY,aAAa,CAAC,GAAG,CAAC,CAAC;AAAA,IAC7C,OAAO;AACL,aAAO,KAAK,YAAY,aAAa,CAAC,GAAG,OAAO,KAAK,KAAK,qBAAqB,OAAO,CAAC;AAAA,IACzF;AAAA,EACF;AAAA,EACA,6BAA6B,SAAS;AACpC,SAAK,4BAA4B,CAAC;AAClC,QAAI,KAAK,sBAAsB;AAC7B,YAAM,cAAc,OAAO,KAAK,KAAK,qBAAqB,OAAO;AACjE,eAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,cAAM,aAAa,YAAY;AAC/B,cAAM,aAAa,KAAK,qBAAqB,QAAQ;AACrD,aAAK,0BAA0B,WAAW,cAAc,QAAQ;AAAA,MAClE;AAAA,IACF;AAAA,EACF;AAAA,EACA,QAAQ,QAAQ,SAAS;AACvB,QAAI,KAAK,6BAA6B,MAAM;AAC1C,WAAK,6BAA6B,KAAK,wBAAwB,CAAC;AAAA,IAClE;AACA,aAAS,KAAK,gBAAgB,MAAM;AACpC,cAAU,KAAK,iBAAiB,OAAO;AACvC,UAAM,SAAS,KAAK,SAAS,QAAQ,QAAQ,OAAO;AACpD,WAAO,OAAO,SAAS,IAAI,SAAS,OAAO;AAAA,EAC7C;AAAA,EACA,MAAM,aAAa,QAAQ,SAAS;AAClC,QAAI,KAAK,6BAA6B,MAAM;AAC1C,WAAK,6BAA6B,MAAM,KAAK,6BAA6B,CAAC;AAAA,IAC7E;AACA,aAAS,KAAK,gBAAgB,MAAM;AACpC,cAAU,KAAK,iBAAiB,OAAO;AACvC,UAAM,SAAS,MAAM,KAAK,SAAS,aAAa,QAAQ,OAAO;AAC/D,WAAO,OAAO,SAAS,IAAI,SAAS,OAAO;AAAA,EAC7C;AAAA,EACA,yBAAyB;AACvB,WAAO,KAAK,SAAS,uBAAuB;AAAA,EAC9C;AAAA,EACA,6BAA6B;AAC3B,SAAK,SAAS,2BAA2B;AAAA,EAC3C;AAAA,EACA,6BAA6B,KAAK;AAChC,WAAO,OAAO,KAAK,GAAG,EAAE,OAAO,CAAC,QAAQ,QAAQ;AAC9C,aAAO,OAAO,CAAC,IAAI,IAAI;AACvB,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AAAA,EACP;AAAA,EACA,UAAU;AACR,SAAK,SAAS,QAAQ;AACtB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,QAAQ;AACzB,UAAI,KAAK,2BAA2B;AAClC,gBAAQ,KAAK,yBAAyB;AAAA,MACxC;AAAA,IACF;AACA,SAAK,gBAAgB,QAAQ;AAAA,EAC/B;AACF;AACA,eAAe,eAAe,UAAU,UAAU,CAAC,GAAG,OAAO,YAAY;AACvE,MAAI,YAAY,MAAM;AACpB,UAAM,IAAI,MAAM,wGAAwG;AAAA,EAC1H;AACA,MAAI,WAAW,MAAM;AACnB,cAAU,CAAC;AAAA,EACb;AACA,MAAI,QAAQ,aAAa,OAAO,aAAa,UAAU;AACrD,eAAW,YAAY,QAAQ;AAAA,EACjC;AACA,QAAM,SAAS,IAAI,WAAW,UAAU,SAAS,IAAI;AACrD,QAAM,OAAO,KAAK;AAClB,SAAO;AACT;AACA,SAAS,mBAAmB,aAAa;AACvC,MAAI,eAAe,MAAM;AACvB,UAAM,IAAI,MAAM,sHAAsH;AAAA,EACxI;AACA,MAAI;AACJ,MAAI,uBAAuB,OAAO;AAChC,UAAM,CAAC,WAAW,OAAO,IAAI;AAC7B,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,kDAAkD;AAAA,IACpE;AACA,QAAI,CAAC,WAAW,EAAE,mBAAmB,cAAc;AACjD,YAAM,IAAI,MAAM,mEAAmE;AAAA,IACrF;AACA,QAAI,EAAE,mBAAmB,YAAY;AACnC,YAAM,IAAI,MAAM,uCAAuC;AAAA,IACzD;AACA,QAAI,EAAE,qBAAqB,YAAY;AACrC,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AACA,UAAM,cAAc,WAAW,eAAe,UAAU,eAAe;AACvE,UAAM,iBAAiB,WAAW,6BAA6B,WAAW,aAAa,OAAO;AAC9F,gBAAY,WAAW,eAAe,cAAc;AAAA,EACtD,WAAW,UAAU,aAAa;AAChC,gBAAY;AAAA,EACd,WAAW,mBAAmB,eAAe,iBAAiB,eAAe,gBAAgB,aAAa;AACxG,gBAAY,WAAW,eAAe,WAAW;AAAA,EACnD,OAAO;AACL,UAAM,IAAI,MAAM,sBAAsB;AAAA,EACxC;AACA,QAAM,SAAS,IAAI,WAAW,SAAS;AACvC,SAAO,KAAK;AACZ,SAAO;AACT;AACA,SAAS,YAAY,UAAU;AAC7B,MAAI,CAAC,SAAS,SAAS,GAAG,GAAG;AAC3B,eAAW,WAAW;AAAA,EACxB;AACA,SAAO,GAAG,WAAW,qBAAqB;AAC5C;AAGA,IAAI,WAAW;AAGf,IAAI,gBAAgB,CAAC;AACrBA,UAAS,eAAe;AAAA,EACtB,YAAY,MAAM;AAAA,EAClB,SAAS,MAAM;AAAA,EACf,gBAAgB,MAAM;AAAA,EACtB,iBAAiB,MAAM;AAAA,EACvB,eAAe,MAAM;AAAA,EACrB,OAAO,MAAM;AAAA,EACb,KAAK,MAAM;AAAA,EACX,MAAM,MAAM;AAAA,EACZ,WAAW,MAAM;AAAA,EACjB,YAAY,MAAM;AAAA,EAClB,cAAc,MAAM;AAAA,EACpB,QAAQ,MAAM;AAAA,EACd,KAAK,MAAM;AACb,CAAC;AAGD,IAAI,cAAc,QAAQ,oBAAoB,CAAC;AAG/C,IAAI,cAAc,QAAQ,oBAAoB,CAAC;AAG/C,SAAS,QAAQ,QAAQ,OAAO;AAC9B,SAAO,gBAAgB,QAAQ,KAAK;AACtC;AACA,SAAS,gBAAgB,QAAQ,OAAO,OAAuB,oBAAI,IAAI,GAAG,cAA8B,oBAAI,IAAI,GAAG;AACjH,MAAI,UAAU,MAAM;AAClB,WAAO;AAAA,EACT;AACA,MAAI,OAAO,SAAS,cAAc,kBAAkB,MAAM;AACxD,WAAO,OAAO,MAAM;AAAA,EACtB;AACA,MAAI,YAAY,IAAI,MAAM,GAAG;AAC3B,UAAM,IAAI,MAAM,wCAAwC;AAAA,EAC1D;AACA,MAAI,KAAK,IAAI,MAAM,GAAG;AACpB,WAAO,KAAK,IAAI,MAAM;AAAA,EACxB;AACA,QAAM,SAAS,MAAM,MAAM;AAC3B,MAAI,OAAO,WAAW,OAAO,UAAU,MAAM;AAC3C,UAAM,IAAI,MAAM,mEAAmE;AAAA,EACrF;AACA,MAAI,CAAC,OAAO,SAAS;AACnB,SAAK,IAAI,QAAQ,OAAO,KAAK;AAC7B,WAAO,OAAO;AAAA,EAChB,WAAW,YAAY,MAAM,GAAG;AAC9B,UAAM,iBAAiB,MAAM,QAAQ,MAAM,IAAI,CAAC,IAAI,CAAC;AACrD,gBAAY,IAAI,MAAM;AACtB,eAAW,KAAK,QAAQ;AACtB,YAAM,QAAQ,OAAO;AACrB,YAAM,cAAc,gBAAgB,OAAO,OAAO,MAAM,WAAW;AACnE,qBAAe,KAAK;AAAA,IACtB;AACA,gBAAY,OAAO,MAAM;AACzB,QAAI,OAAO,WAAW;AACpB,qBAAe,YAAY,OAAO;AAAA,IACpC;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,yCAAyC,QAAQ;AAAA,EACnE;AACF;AACA,SAAS,QAAQ,QAAQ,QAAQ,WAAW;AAC1C,SAAO,gBAAgB,QAAQ,KAAK;AACtC;AACA,SAAS,gBAAgB,QAAQ,OAAO,cAA8B,oBAAI,IAAI,GAAG;AAC/E,QAAM,SAAS,OAAO;AACtB,MAAI,YAAY,IAAI,MAAM,GAAG;AAC3B,UAAM,IAAI,MAAM,wCAAwC;AAAA,EAC1D;AACA,QAAM,SAAS,MAAM,MAAM;AAC3B,MAAI,OAAO,WAAW,OAAO,UAAU,MAAM;AAC3C,UAAM,IAAI,MAAM,mEAAmE;AAAA,EACrF;AACA,MAAI,CAAC,OAAO,SAAS;AACnB,WAAO,OAAO;AAAA,EAChB,WAAW,YAAY,MAAM,GAAG;AAC9B,UAAM,iBAAiB,MAAM,QAAQ,MAAM,IAAI,CAAC,IAAI,CAAC;AACrD,gBAAY,IAAI,MAAM;AACtB,eAAW,KAAK,QAAQ;AACtB,YAAM,WAAW,OAAO,IAAI,CAAC,MAAM,EAAE,EAAE;AACvC,YAAM,cAAc,gBAAgB,UAAU,OAAO,WAAW;AAChE,qBAAe,KAAK;AAAA,IACtB;AACA,gBAAY,OAAO,MAAM;AACzB,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,yCAAyC,QAAQ;AAAA,EACnE;AACF;AACA,SAAS,UAAU,GAAG;AACpB,MAAI,MAAM,MAAM;AACd,WAAO;AAAA,EACT;AACA,MAAI,YAAY,EAAE,EAAE,GAAG;AACrB,WAAO,EAAE,OAAO,MAAM,SAAS,KAAK;AAAA,EACtC,OAAO;AACL,WAAO,EAAE,OAAO,GAAG,SAAS,MAAM;AAAA,EACpC;AACF;AACA,eAAe,mBAAmB,QAAQ,OAAO;AAC/C,QAAM,OAAuB,oBAAI,IAAI;AACrC,kBAAgB,QAAQ,OAAO,IAAI;AACnC,aAAW,OAAO,MAAM,KAAK,KAAK,KAAK,CAAC,GAAG;AACzC,UAAM,QAAQ,KAAK,IAAI,GAAG;AAC1B,QAAI,aAAa,UAAU,KAAK,GAAG;AACjC,YAAM,cAAc,MAAM;AAC1B,WAAK,IAAI,KAAK,WAAW;AAAA,IAC3B;AAAA,EACF;AACA,QAAM,SAAS,gBAAgB,QAAQ,OAAO,IAAI;AAClD,SAAO;AACT;AACA,SAAS,YAAY,KAAK;AACxB,MAAI,gBAAgB;AACpB,MAAI,IAAI,EAAE,IAAI,YAAY,GAAG;AAC3B,oBAAgB,eAAe;AAAA,EACjC,OAAO;AACL,UAAM,EAAE,cAAc,IAAI,uBAAuB;AACjD,oBAAgB,eAAe;AAAA,EACjC;AACA,SAAO,OAAO,QAAQ,CAAC,YAAY,OAAO,GAAG,MAAM,MAAM,QAAQ,GAAG,KAAK,OAAO,QAAQ,YAAY,EAAE,eAAe,WAAW,EAAE,eAAe,YAAY,CAAC;AAChK;AACA,SAAS,aAAa,KAAK;AACzB,SAAO,OAAO,QAAQ,YAAY,GAAG,KAAK,MAAM,QAAQ,GAAG,KAAK,OAAO,QAAQ,YAAY,eAAe,UAAU,aAAa,aAAa,GAAG;AACnJ;AACA,SAAS,YAAY,OAAO;AAC1B,SAAO,UAAU,QAAQ,OAAO,UAAU,YAAY,OAAO,UAAU;AACzE;AAGA,SAAS,UAAU,WAAW;AAC5B,SAAO,QAAQ,WAAW,aAAa;AACzC;AACA,SAAS,cAAc,MAAM;AAC3B,MAAI,gBAAgB,QAAQ;AAC1B,WAAO,EAAE,OAAO,KAAK,MAAM,GAAG,SAAS,MAAM;AAAA,EAC/C,WAAW,YAAY,IAAI,GAAG;AAC5B,WAAO,EAAE,OAAO,MAAM,SAAS,KAAK;AAAA,EACtC,OAAO;AACL,WAAO,EAAE,OAAO,MAAM,SAAS,MAAM;AAAA,EACvC;AACF;AAGA,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,UAAU;AACpB,SAAK,WAAW;AAChB,SAAK,QAAQ;AACb,SAAK,MAAM;AACX,QAAI,YAAY,MAAM;AACpB,YAAM,IAAI,WAAW,iDAAiD;AAAA,IACxE;AACA,QAAI,WAAW,GAAG;AAChB,YAAM,IAAI,WAAW,2CAA2C;AAAA,IAClE;AACA,SAAK,OAAO,IAAI,MAAM,QAAQ;AAC9B,SAAK,kBAAkB,IAAI;AAAA,EAC7B;AAAA,EACA,KAAK,OAAO;AACV,WAAO,QAAQ,GAAG;AAChB,eAAS,KAAK;AAAA,IAChB;AACA,WAAO,QAAQ,KAAK;AAAA,EACtB;AAAA,EACA,IAAI,OAAO;AACT,QAAI,QAAQ,GAAG;AACb,YAAM,IAAI,WAAW,qCAAqC;AAAA,IAC5D;AACA,WAAO,KAAK,KAAK,QAAQ,KAAK;AAAA,EAChC;AAAA,EACA,IAAI,OAAO,OAAO;AAChB,QAAI,QAAQ,GAAG;AACb,YAAM,IAAI,WAAW,qCAAqC;AAAA,IAC5D;AACA,SAAK,KAAK,QAAQ,KAAK,YAAY;AAAA,EACrC;AAAA,EACA,SAAS;AACP,QAAI,SAAS,KAAK,MAAM,KAAK;AAC7B,QAAI,SAAS,GAAG;AACd,eAAS,KAAK,kBAAkB;AAAA,IAClC;AACA,WAAO;AAAA,EACT;AAAA,EACA,SAAS;AACP,WAAO,KAAK,OAAO,MAAM,KAAK;AAAA,EAChC;AAAA,EACA,UAAU;AACR,WAAO,KAAK,OAAO,MAAM;AAAA,EAC3B;AAAA,EACA,KAAK,OAAO;AACV,QAAI,KAAK,OAAO,GAAG;AACjB,YAAM,IAAI,WAAW,sBAAsB;AAAA,IAC7C;AACA,SAAK,IAAI,KAAK,KAAK,KAAK;AACxB,SAAK,MAAM,KAAK,KAAK,KAAK,MAAM,CAAC;AAAA,EACnC;AAAA,EACA,QAAQ,QAAQ;AACd,eAAW,SAAS,QAAQ;AAC1B,WAAK,KAAK,KAAK;AAAA,IACjB;AAAA,EACF;AAAA,EACA,MAAM;AACJ,QAAI,KAAK,QAAQ,GAAG;AAClB,YAAM,IAAI,WAAW,uBAAuB;AAAA,IAC9C;AACA,SAAK,MAAM,KAAK,KAAK,KAAK,MAAM,CAAC;AACjC,UAAM,SAAS,KAAK,IAAI,KAAK,GAAG;AAChC,SAAK,IAAI,KAAK,KAAK,MAAM;AACzB,WAAO;AAAA,EACT;AAAA,EACA,QAAQ,OAAO;AACb,QAAI,KAAK,OAAO,GAAG;AACjB,YAAM,IAAI,WAAW,sBAAsB;AAAA,IAC7C;AACA,SAAK,QAAQ,KAAK,KAAK,KAAK,QAAQ,CAAC;AACrC,SAAK,IAAI,KAAK,OAAO,KAAK;AAAA,EAC5B;AAAA,EACA,QAAQ;AACN,QAAI,KAAK,QAAQ,GAAG;AAClB,YAAM,IAAI,WAAW,uBAAuB;AAAA,IAC9C;AACA,UAAM,SAAS,KAAK,IAAI,KAAK,KAAK;AAClC,SAAK,IAAI,KAAK,OAAO,MAAM;AAC3B,SAAK,QAAQ,KAAK,KAAK,KAAK,QAAQ,CAAC;AACrC,WAAO;AAAA,EACT;AAAA,EACA,cAAc,eAAe;AAC3B,QAAI,KAAK,QAAQ,GAAG;AAClB,YAAM,IAAI,WAAW,uBAAuB;AAAA,IAC9C;AACA,UAAM,QAAQ,KAAK,KAAK,KAAK,QAAQ,aAAa;AAClD,UAAM,SAAS,KAAK,IAAI,KAAK;AAC7B,SAAK,IAAI,OAAO,KAAK,IAAI,CAAC;AAC1B,WAAO;AAAA,EACT;AACF;AAGA,IAAI,oBAAoB,cAAc,WAAW;AAAA,EAC/C,cAAc;AACZ,UAAM,kBAAkB,gBAAgB;AAAA,EAC1C;AAAA,EACA,SAAS;AACP,WAAO;AAAA,EACT;AAAA,EACA,KAAK,OAAO;AACV,QAAI,MAAM,OAAO,GAAG;AAClB,WAAK,OAAO;AAAA,IACd;AACA,UAAM,KAAK,KAAK;AAAA,EAClB;AAAA,EACA,QAAQ,OAAO;AACb,QAAI,MAAM,OAAO,GAAG;AAClB,WAAK,OAAO;AAAA,IACd;AACA,UAAM,QAAQ,KAAK;AAAA,EACrB;AAAA,EACA,SAAS;AACP,UAAM,cAAc,KAAK,WAAW;AACpC,UAAM,UAAU,IAAI,MAAM,WAAW;AACrC,UAAM,MAAM,KAAK,OAAO;AACxB,aAAS,IAAI,GAAG,IAAI,KAAK,KAAK;AAC5B,cAAQ,KAAK,KAAK,IAAI,KAAK,KAAK,KAAK,QAAQ,CAAC,CAAC;AAAA,IACjD;AACA,SAAK,OAAO;AACZ,SAAK,WAAW;AAChB,SAAK,kBAAkB,IAAI,KAAK;AAChC,SAAK,QAAQ;AACb,SAAK,MAAM;AAAA,EACb;AACF;AACA,kBAAkB,mBAAmB;AAGrC,SAAS,kBAAkB,OAAO;AAChC,SAAO,IAAI,cAAc,KAAK;AAChC;AACA,SAAS,qBAAqB,OAAO;AACnC,SAAO,IAAI,qBAAqB,KAAK;AACvC;AACA,SAAS,yBAAyB,eAAe,kBAAkB;AACjE,SAAO,IAAI,gBAAgB,eAAe,gBAAgB;AAC5D;AACA,SAAS,mBAAmB,WAAW,eAAe,gBAAgB,MAAM;AAC1E,SAAO,IAAI,YAAY,WAAW,YAAY;AAChD;AACA,IAAI,eAAe,MAAM;AAAA,EACvB,MAAM,UAAU;AACd,UAAM,SAAS,CAAC;AAChB,QAAI,IAAI,MAAM,KAAK,KAAK;AACxB,WAAO,CAAC,EAAE,MAAM;AACd,aAAO,KAAK,EAAE,KAAK;AACnB,UAAI,MAAM,KAAK,KAAK;AAAA,IACtB;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,iBAAiB;AACrB,UAAM,SAAS,KAAK,SAAS,GAAG;AAChC,UAAM,SAAS,CAAC;AAChB,QAAI,IAAI,MAAM,OAAO,KAAK;AAC1B,WAAO,CAAC,EAAE,MAAM;AACd,aAAO,KAAK,EAAE,KAAK;AACnB,UAAI,MAAM,OAAO,KAAK;AAAA,IACxB;AACA,WAAO;AAAA,EACT;AAAA,EACA,MAAM,eAAe;AACnB,QAAI,IAAI,MAAM,KAAK,KAAK;AACxB,WAAO,CAAC,EAAE,MAAM;AACd,UAAI,MAAM,KAAK,KAAK;AAAA,IACtB;AAAA,EACF;AAAA,EACA,MAAM,aAAa,WAAW;AAC5B,QAAI,IAAI,MAAM,KAAK,KAAK;AACxB,QAAI,iBAAiB,UAAU,EAAE,KAAK;AACtC,WAAO,CAAC,EAAE,QAAQ,gBAAgB;AAChC,UAAI,MAAM,KAAK,KAAK;AACpB,uBAAiB,UAAU,EAAE,KAAK;AAAA,IACpC;AAAA,EACF;AAAA,EACA,aAAa,SAAS;AACpB,WAAO,IAAI,0BAA0B,MAAM,OAAO;AAAA,EACpD;AAAA,EACA,OAAO,WAAW;AAChB,WAAO,IAAI,eAAe,MAAM,SAAS;AAAA,EAC3C;AAAA,EACA,IAAI,YAAY;AACd,WAAO,IAAI,YAAY,MAAM,UAAU;AAAA,EACzC;AAAA,EACA,SAAS,YAAY;AACnB,WAAO,IAAI,iBAAiB,MAAM,UAAU;AAAA,EAC9C;AAAA,EACA,eAAe,YAAY;AACzB,WAAO,IAAI,iBAAiB,MAAM,UAAU,EAAE,OAAO;AAAA,EACvD;AAAA,EACA,QAAQ,YAAY;AAClB,WAAO,IAAI,gBAAgB,MAAM,UAAU;AAAA,EAC7C;AAAA,EACA,MAAM,aAAa,GAAG;AACpB,WAAO,KAAK,IAAI,CAAC,EAAE,aAAa;AAAA,EAClC;AAAA,EACA,MAAM,cAAc,GAAG;AACrB,WAAO,KAAK,eAAe,CAAC,EAAE,aAAa,CAAC,MAAM,MAAM,IAAI;AAAA,EAC9D;AAAA,EACA,cAAc,WAAW,iBAAiB,MAAM;AAC9C,WAAO,IAAI,sBAAsB,MAAM,WAAW,cAAc;AAAA,EAClE;AAAA,EACA,iBAAiB,WAAW,iBAAiB,MAAM,QAAQ,WAAW;AACpE,UAAM,aAAa,KAAK,cAAc,WAAW,cAAc;AAC/D,WAAO,WAAW,IAAI,CAAC,MAAM,QAAQ,GAAG,KAAK,CAAC;AAAA,EAChD;AAAA,EACA,YAAY,UAAU,kBAAkB;AACtC,WAAO,IAAI,gBAAgB,kBAAkB,CAAC,MAAM,QAAQ,CAAC,GAAG,gBAAgB;AAAA,EAClF;AAAA,EACA,KAAK,QAAQ;AACX,QAAI,SAAS,KAAK,UAAU,MAAM;AAChC,aAAO;AAAA,IACT;AACA,WAAO,IAAI,aAAa,MAAM,MAAM;AAAA,EACtC;AAAA,EACA,KAAK,QAAQ;AACX,QAAI,SAAS,KAAK,UAAU,MAAM;AAChC,aAAO;AAAA,IACT;AACA,WAAO,IAAI,aAAa,MAAM,MAAM;AAAA,EACtC;AAAA,EACA,SAAS,YAAY;AACnB,WAAO,IAAI,iBAAiB,MAAM,UAAU;AAAA,EAC9C;AAAA,EACA,QAAQ,YAAY,MAAM;AACxB,WAAO,IAAI,gBAAgB,MAAM,YAAY,IAAI;AAAA,EACnD;AAAA,EACA,SAAS;AACP,WAAO,IAAI,eAAe,IAAI;AAAA,EAChC;AACF;AACA,IAAI,gBAAgB,cAAc,aAAa;AAAA,EAC7C,YAAY,OAAO;AACjB,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,OAAO;AAAA,EACd;AAAA,EACA,UAAU;AACR,WAAO,YAAY,KAAK,MAAM;AAAA,EAChC;AAAA,EACA,MAAM,OAAO;AACX,QAAI,KAAK,QAAQ,KAAK,MAAM,QAAQ;AAClC,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,UAAM,OAAO,KAAK,MAAM,KAAK;AAC7B,SAAK;AACL,WAAO,EAAE,OAAO,UAAU,IAAI,GAAG,MAAM,MAAM;AAAA,EAC/C;AACF;AACA,IAAI,uBAAuB,cAAc,aAAa;AAAA,EACpD,YAAY,QAAQ;AAClB,UAAM;AACN,SAAK,SAAS;AAAA,EAChB;AAAA,EACA,UAAU;AACR,WAAO;AAAA,EACT;AAAA,EACA,MAAM,OAAO;AACX,QAAI;AACF,aAAO,KAAK,OAAO;AAAA,IACrB,SAAS,GAAP;AACA,QAAE,UAAU,mDAAmD,EAAE;AACjE,YAAM;AAAA,IACR;AAAA,EACF;AACF;AACA,IAAI,iBAAiB,cAAc,aAAa;AAAA,EAC9C,YAAY,UAAU;AACpB,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AACF;AACA,IAAI,eAAe,cAAc,aAAa;AAAA,EAC5C,YAAY,UAAU,UAAU;AAC9B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,WAAW;AAChB,SAAK,QAAQ;AACb,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,KAAK,UAAU,KAAK,UAAU;AACnC,YAAM,UAAU,MAAM,KAAK,SAAS,KAAK;AACzC,UAAI,QAAQ,MAAM;AAChB,eAAO;AAAA,MACT;AACA,cAAQ,QAAQ,KAAK;AAAA,IACvB;AACA,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AACF;AACA,IAAI,eAAe,cAAc,aAAa;AAAA,EAC5C,YAAY,UAAU,UAAU;AAC9B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,WAAW;AAChB,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,QAAI,KAAK,WAAW,KAAK,UAAU;AACjC,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AACF;AACA,IAAI,wBAAwB,cAAc,aAAa;AAAA,EACrD,YAAY,UAAU,WAAW,uBAAuB,MAAM;AAC5D,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AACjB,SAAK,uBAAuB;AAC5B,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,UAAM,QAAQ,CAAC;AACf,WAAO,MAAM,SAAS,KAAK,WAAW;AACpC,YAAM,OAAO,MAAM,KAAK,SAAS,KAAK;AACtC,UAAI,KAAK,MAAM;AACb,YAAI,KAAK,wBAAwB,MAAM,SAAS,GAAG;AACjD,iBAAO,EAAE,OAAO,OAAO,MAAM,MAAM;AAAA,QACrC;AACA,eAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,MACnC;AACA,YAAM,KAAK,KAAK,KAAK;AAAA,IACvB;AACA,WAAO,EAAE,OAAO,OAAO,MAAM,MAAM;AAAA,EACrC;AACF;AACA,IAAI,iBAAiB,cAAc,aAAa;AAAA,EAC9C,YAAY,UAAU,WAAW;AAC/B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AACjB,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,MAAM;AACX,YAAM,OAAO,MAAM,KAAK,SAAS,KAAK;AACtC,UAAI,KAAK,QAAQ,KAAK,UAAU,KAAK,KAAK,GAAG;AAC3C,eAAO;AAAA,MACT;AACA,cAAQ,KAAK,KAAK;AAAA,IACpB;AAAA,EACF;AACF;AACA,IAAI,cAAc,cAAc,aAAa;AAAA,EAC3C,YAAY,UAAU,YAAY;AAChC,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,UAAM,OAAO,MAAM,KAAK,SAAS,KAAK;AACtC,QAAI,KAAK,MAAM;AACb,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,UAAM,eAAe,oBAAoB,sBAAsB,KAAK,KAAK;AACzE,UAAM,SAAS,KAAK,UAAU,KAAK,KAAK;AACxC,UAAM,gBAAgB,oBAAoB,sBAAsB,MAAM;AACtE,eAAW,KAAK,cAAc;AAC5B,UAAI,CAAC,oBAAoB,eAAe,GAAG,aAAa,GAAG;AACzD,UAAE,QAAQ;AAAA,MACZ;AAAA,IACF;AACA,WAAO,EAAE,OAAO,QAAQ,MAAM,MAAM;AAAA,EACtC;AACF;AACA,IAAI,4BAA4B,cAAc,aAAa;AAAA,EACzD,YAAY,UAAU,SAAS;AAC7B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,QAAQ;AACb,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,MAAM;AACX,UAAI;AACF,eAAO,MAAM,KAAK,SAAS,KAAK;AAAA,MAClC,SAAS,GAAP;AACA,YAAI,CAAC,KAAK,QAAQ,CAAC,GAAG;AACpB,iBAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,QACnC;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AACA,IAAI,mBAAmB,cAAc,aAAa;AAAA,EAChD,YAAY,UAAU,YAAY;AAChC,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,UAAM,OAAO,MAAM,KAAK,SAAS,KAAK;AACtC,QAAI,KAAK,MAAM;AACb,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,UAAM,eAAe,oBAAoB,sBAAsB,KAAK,KAAK;AACzE,UAAM,SAAS,MAAM,KAAK,UAAU,KAAK,KAAK;AAC9C,UAAM,gBAAgB,oBAAoB,sBAAsB,MAAM;AACtE,eAAW,KAAK,cAAc;AAC5B,UAAI,CAAC,oBAAoB,eAAe,GAAG,aAAa,GAAG;AACzD,UAAE,QAAQ;AAAA,MACZ;AAAA,IACF;AACA,WAAO,EAAE,OAAO,QAAQ,MAAM,MAAM;AAAA,EACtC;AACF;AACA,IAAI,oBAAoB,cAAc,aAAa;AAAA,EACjD,cAAc;AACZ,UAAM;AACN,SAAK,cAAc,IAAI,kBAAkB;AACzC,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,aAAa;AACjB,WAAO,KAAK,YAAY,OAAO,MAAM,GAAG;AACtC,UAAI,CAAC,MAAM,KAAK,KAAK,GAAG;AACtB,eAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,MACnC;AAAA,IACF;AACA,WAAO,EAAE,OAAO,KAAK,YAAY,MAAM,GAAG,MAAM,MAAM;AAAA,EACxD;AACF;AACA,IAAI,kBAAkB,cAAc,kBAAkB;AAAA,EACpD,YAAY,UAAU,YAAY;AAChC,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,UAAM,OAAO,MAAM,KAAK,SAAS,KAAK;AACtC,QAAI,KAAK,MAAM;AACb,aAAO;AAAA,IACT;AACA,UAAM,eAAe,oBAAoB,sBAAsB,KAAK,KAAK;AACzE,UAAM,cAAc,KAAK,UAAU,KAAK,KAAK;AAC7C,UAAM,gBAAgB,oBAAoB,sBAAsB,WAAW;AAC3E,SAAK,YAAY,QAAQ,WAAW;AACpC,eAAW,KAAK,cAAc;AAC5B,UAAI,CAAC,oBAAoB,eAAe,GAAG,aAAa,GAAG;AACzD,UAAE,QAAQ;AAAA,MACZ;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;AACA,IAAI,kBAAkB,cAAc,aAAa;AAAA,EAC/C,YAAY,WAAW,kBAAkB;AACvC,UAAM;AACN,SAAK,mBAAmB;AACxB,SAAK,WAAW;AAChB,SAAK,WAAW;AAChB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EACA,UAAU;AACR,UAAM,oBAAoB;AAC1B,WAAO,GAAG;AAAA,EACZ;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,cAAc,KAAK,QAAQ;AAChD,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,cAAc,UAAU;AAC5B,UAAM;AACN,QAAI,KAAK,YAAY,MAAM;AACzB,YAAM,iBAAiB,MAAM,KAAK,cAAc,KAAK;AACrD,UAAI,eAAe,MAAM;AACvB,eAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,MACnC;AACA,WAAK,WAAW,eAAe;AAC/B,UAAI,KAAK,oBAAoB,MAAM;AACjC,aAAK,WAAW,KAAK,SAAS,aAAa,KAAK,gBAAgB;AAAA,MAClE;AAAA,IACF;AACA,UAAM,aAAa,MAAM,KAAK,SAAS,KAAK;AAC5C,QAAI,WAAW,MAAM;AACnB,WAAK,WAAW;AAChB,aAAO,KAAK,cAAc,QAAQ;AAAA,IACpC;AACA,WAAO;AAAA,EACT;AACF;AACA,IAAI;AAAA,CACH,SAAS,kBAAkB;AAC1B,mBAAiB,iBAAiB,UAAU,KAAK;AACjD,mBAAiB,iBAAiB,cAAc,KAAK;AACrD,mBAAiB,iBAAiB,aAAa,KAAK;AACtD,GAAG,oBAAoB,kBAAkB,CAAC,EAAE;AAC5C,IAAI,cAAc,cAAc,aAAa;AAAA,EAC3C,YAAY,WAAW,eAAe,gBAAgB,MAAM;AAC1D,UAAM;AACN,SAAK,YAAY;AACjB,SAAK,eAAe;AACpB,SAAK,QAAQ;AACb,SAAK,iBAAiB;AAAA,EACxB;AAAA,EACA,UAAU;AACR,UAAM,oBAAoB;AAC1B,WAAO,IAAI;AAAA,EACb;AAAA,EACA,MAAM,UAAU,YAAY;AAC1B,UAAM;AACN,QAAI,eAAe;AACnB,QAAI,gBAAgB;AACpB,aAAS,QAAQ,WAAW;AAC1B,UAAI,qBAAqB,cAAc;AACrC,cAAM,SAAS,UAAU,KAAK;AAC9B,eAAO;AAAA,UACL,OAAO,OAAO,KAAK,CAAC,MAAM;AACxB;AACA,gBAAI,EAAE,MAAM;AACV;AAAA,YACF;AACA,mBAAO,EAAE;AAAA,UACX,CAAC;AAAA,UACD,SAAS;AAAA,QACX;AAAA,MACF,OAAO;AACL,eAAO,EAAE,OAAO,MAAM,SAAS,KAAK;AAAA,MACtC;AAAA,IACF;AACA,UAAM,SAAS,MAAM,mBAAmB,KAAK,WAAW,OAAO;AAC/D,QAAI,iBAAiB,eAAe;AAClC,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,QAAI,gBAAgB,GAAG;AACrB,cAAQ,KAAK,cAAc;AAAA,QACzB,KAAK,gBAAgB;AACnB,gBAAM,IAAI,MAAM,qEAAqE,KAAK,QAAQ;AAAA,QACpG,KAAK,gBAAgB;AACnB,iBAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,QACnC,KAAK,gBAAgB;AAAA,QACrB;AAAA,MACF;AAAA,IACF;AACA,SAAK;AACL,WAAO,EAAE,OAAO,QAAQ,MAAM,MAAM;AAAA,EACtC;AAAA,EACA,MAAM,OAAO;AACX,SAAK,iBAAiB,KAAK,UAAU,KAAK,cAAc;AACxD,WAAO,KAAK;AAAA,EACd;AACF;AACA,IAAI,mBAAmB,cAAc,aAAa;AAAA,EAChD,YAAY,UAAU,YAAY;AAChC,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,aAAa;AAClB,SAAK,SAAS,IAAI,WAAW,UAAU;AAAA,EACzC;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,SAAS;AACP,WAAO,CAAC,KAAK,OAAO,OAAO,GAAG;AAC5B,YAAM,IAAI,KAAK,SAAS,KAAK;AAC7B,WAAK,OAAO,KAAK,CAAC;AAAA,IACpB;AAAA,EACF;AAAA,EACA,OAAO;AACL,SAAK,OAAO;AACZ,WAAO,KAAK,OAAO,MAAM;AAAA,EAC3B;AACF;AACA,IAAI,kBAAkB,cAAc,iBAAiB;AAAA,EACnD,YAAY,UAAU,YAAY,MAAM;AACtC,UAAM,UAAU,UAAU;AAC1B,SAAK,WAAW;AAChB,SAAK,aAAa;AAClB,SAAK,oBAAoB;AACzB,SAAK,SAAS,YAAY,KAAK,QAAQ,aAAa,IAAI,EAAE,SAAS,CAAC;AACpE,SAAK,WAAW,QAAQ,QAAQ,EAAE,OAAO,MAAM,MAAM,MAAM,CAAC;AAAA,EAC9D;AAAA,EACA,MAAM,OAAO;AACX,SAAK,WAAW,KAAK,SAAS,KAAK,MAAM,KAAK,WAAW,CAAC;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EACA,UAAU,MAAM;AACd,WAAO,KAAK,MAAM,KAAK,OAAO,IAAI,IAAI;AAAA,EACxC;AAAA,EACA,cAAc;AACZ,WAAO,KAAK,UAAU,KAAK,OAAO,OAAO,CAAC;AAAA,EAC5C;AAAA,EACA,MAAM,aAAa;AACjB,QAAI,CAAC,KAAK,mBAAmB;AAC3B,WAAK,OAAO;AAAA,IACd;AACA,WAAO,CAAC,KAAK,OAAO,QAAQ,GAAG;AAC7B,YAAM,cAAc,KAAK,YAAY;AACrC,YAAM,SAAS,MAAM,KAAK,OAAO,cAAc,WAAW;AAC1D,UAAI,OAAO,MAAM;AACf,aAAK,oBAAoB;AAAA,MAC3B,OAAO;AACL,aAAK,OAAO;AACZ,eAAO;AAAA,MACT;AAAA,IACF;AACA,WAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,EACnC;AACF;AAGA,IAAI,UAAU,MAAM;AAAA,EAClB,cAAc;AACZ,SAAK,OAAO;AAAA,EACd;AAAA,EACA,MAAM,WAAW,iBAAiB,MAAM;AACtC,UAAM,OAAO;AACb,iBAAa,OAAO,YAAY,GAAG,MAAM;AAAA,QACrC,WAAW;AACf,QAAI;AACJ,QAAI,KAAK,SAAS,YAAY,KAAK,QAAQ,MAAM;AAC/C,aAAO,KAAK;AAAA,IACd,WAAW,gBAAgB;AACzB,aAAO,KAAK,KAAK,KAAK,OAAO,SAAS;AAAA,IACxC,OAAO;AACL,aAAO,KAAK,MAAM,KAAK,OAAO,SAAS;AAAA,IACzC;AACA,WAAO,sBAAsB,YAAY;AACvC,cAAQ,MAAM,KAAK,SAAS,GAAG,iBAAiB,WAAW,gBAAgB,eAAe;AAAA,IAC5F,GAAG,IAAI;AAAA,EACT;AAAA,EACA,YAAY,SAAS;AACnB,UAAM,OAAO;AACb,QAAI;AACJ,QAAI,KAAK,SAAS,YAAY,QAAQ,SAAS,UAAU;AACvD,aAAO;AAAA,IACT,WAAW,KAAK,QAAQ,QAAQ,QAAQ,QAAQ,MAAM;AACpD,aAAO,KAAK,OAAO,QAAQ;AAAA,IAC7B,OAAO;AACL,aAAO;AAAA,IACT;AACA,WAAO,sBAAsB,aAAa,MAAM,KAAK,SAAS,GAAG,YAAY,MAAM,QAAQ,SAAS,CAAC,GAAG,IAAI;AAAA,EAC9G;AAAA,EACA,OAAO,WAAW;AAChB,UAAM,OAAO;AACb,QAAI;AACJ,QAAI,KAAK,SAAS,UAAU;AAC1B,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AACA,WAAO,sBAAsB,YAAY;AACvC,cAAQ,MAAM,KAAK,SAAS,GAAG,OAAO,CAAC,MAAM,KAAK,MAAM,UAAU,CAAC,CAAC,CAAC;AAAA,IACvE,GAAG,IAAI;AAAA,EACT;AAAA,EACA,MAAM,aAAa,GAAG;AACpB,YAAQ,MAAM,KAAK,SAAS,GAAG,aAAa,CAAC;AAAA,EAC/C;AAAA,EACA,IAAI,YAAY;AACd,UAAM,OAAO;AACb,WAAO,sBAAsB,YAAY;AACvC,cAAQ,MAAM,KAAK,SAAS,GAAG,IAAI,CAAC,MAAM,KAAK,MAAM,WAAW,CAAC,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,IAAI;AAAA,EACd;AAAA,EACA,SAAS,YAAY;AACnB,UAAM,OAAO;AACb,WAAO,sBAAsB,YAAY;AACvC,cAAQ,MAAM,KAAK,SAAS,GAAG,SAAS,UAAU;AAAA,IACpD,GAAG,KAAK,IAAI;AAAA,EACd;AAAA,EACA,SAAS,YAAY;AACnB,QAAI,cAAc,MAAM;AACtB,YAAM,IAAI,WAAW,2DAA2D;AAAA,IAClF;AACA,UAAM,OAAO;AACb,WAAO,sBAAsB,aAAa,MAAM,KAAK,SAAS,GAAG,SAAS,UAAU,GAAG,KAAK,IAAI;AAAA,EAClG;AAAA,EACA,OAAO,QAAQ;AACb,UAAM,OAAO;AACb,QAAI;AACJ,QAAI,KAAK,QAAQ,QAAQ,SAAS,GAAG;AACnC,aAAO,KAAK,OAAO;AAAA,IACrB,WAAW,WAAW,GAAG;AACvB,aAAO;AAAA,IACT,WAAW,KAAK,QAAQ,SAAS,WAAW,UAAU,SAAS,IAAI;AACjE,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AACA,WAAO,sBAAsB,YAAY;AACvC,YAAM,mBAAmB,qBAAqB,aAAa,EAAE,OAAO,MAAM,KAAK,SAAS,GAAG,MAAM,MAAM,EAAE;AACzG,aAAO,yBAAyB,iBAAiB,KAAK,MAAM,CAAC;AAAA,IAC/D,GAAG,IAAI;AAAA,EACT;AAAA,EACA,KAAK,QAAQ;AACX,UAAM,OAAO;AACb,QAAI;AACJ,QAAI,KAAK,QAAQ,QAAQ,UAAU,KAAK,KAAK,QAAQ,QAAQ;AAC3D,aAAO,KAAK,OAAO;AAAA,IACrB,WAAW,KAAK,QAAQ,SAAS,KAAK,OAAO,UAAU,WAAW,UAAU,SAAS,IAAI;AACvF,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AACA,WAAO,sBAAsB,aAAa,MAAM,KAAK,SAAS,GAAG,KAAK,MAAM,GAAG,IAAI;AAAA,EACrF;AAAA,EACA,QAAQ,YAAY,MAAM,yBAAyB,MAAM;AACvD,QAAI,cAAc,QAAQ,aAAa,GAAG;AACxC,UAAI,KAAK,QAAQ,MAAM;AACrB,cAAM,IAAI,WAAW,0DAA0D;AAAA,MACjF,OAAO;AACL,cAAM,IAAI,WAAW,mNAAmN,KAAK,gBAAgB;AAAA,MAC/P;AAAA,IACF;AACA,UAAM,OAAO;AACb,UAAM,SAAS,YAAY,KAAK,QAAQ,aAAa,IAAI,EAAE,SAAS,CAAC;AACrE,WAAO,sBAAsB,YAAY;AACvC,UAAI,QAAQ,OAAO,MAAM;AACzB,UAAI,wBAAwB;AAC1B,iBAAS,OAAO,MAAM;AAAA,MACxB;AACA,cAAQ,MAAM,KAAK,SAAS,GAAG,QAAQ,YAAY,MAAM,SAAS,CAAC;AAAA,IACrE,GAAG,KAAK,IAAI;AAAA,EACd;AAAA,EACA,KAAK,QAAQ;AACX,UAAM,OAAO;AACb,QAAI;AACJ,QAAI,KAAK,QAAQ,QAAQ,KAAK,OAAO,QAAQ;AAC3C,aAAO;AAAA,IACT,WAAW,KAAK,QAAQ,QAAQ,KAAK,QAAQ,QAAQ;AACnD,aAAO,KAAK;AAAA,IACd,OAAO;AACL,aAAO;AAAA,IACT;AACA,WAAO,sBAAsB,aAAa,MAAM,KAAK,SAAS,GAAG,KAAK,MAAM,GAAG,IAAI;AAAA,EACrF;AAAA,EACA,MAAM,UAAU;AACd,QAAI,KAAK,SAAS,UAAU;AAC1B,YAAM,IAAI,MAAM,gDAAgD;AAAA,IAClE;AACA,YAAQ,MAAM,KAAK,SAAS,GAAG,QAAQ;AAAA,EACzC;AAAA,EACA,MAAM,iBAAiB;AACrB,QAAI,KAAK,SAAS,UAAU;AAC1B,YAAM,IAAI,MAAM,gDAAgD;AAAA,IAClE;AACA,YAAQ,MAAM,KAAK,SAAS,GAAG,eAAe;AAAA,EAChD;AACF;AACA,QAAQ,kBAAkB;AAC1B,SAAS,sBAAsB,YAAY,OAAO,MAAM;AACtD,SAAO,IAAI,cAAc,QAAQ;AAAA,IAC/B,cAAc;AACZ,YAAM,GAAG,SAAS;AAClB,WAAK,OAAO;AAAA,IACd;AAAA,IACA,MAAM,WAAW;AACf,aAAO,WAAW;AAAA,IACpB;AAAA,EACF,EAAE;AACJ;AACA,SAAS,MAAM,OAAO;AACpB,SAAO,sBAAsB,YAAY,kBAAkB,KAAK,GAAG,MAAM,MAAM;AACjF;AACA,SAAS,IAAI,UAAU;AACrB,MAAI,CAAC,YAAY,QAAQ,GAAG;AAC1B,UAAM,IAAI,MAAM,mDAAmD;AAAA,EACrE;AACA,MAAI;AACJ,MAAI,MAAM,QAAQ,QAAQ,GAAG;AAC3B,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAO,QAAQ,OAAO,SAAS,GAAG,OAAO,KAAK,IAAI,MAAM,SAAS,GAAG,IAAI;AAAA,IAC1E;AAAA,EACF,WAAW,oBAAoB,QAAQ;AACrC,eAAW,MAAM,UAAU;AACzB,aAAO,QAAQ,OAAO,SAAS,IAAI,OAAO,KAAK,IAAI,MAAM,SAAS,IAAI,IAAI;AAAA,IAC5E;AAAA,EACF;AACA,SAAO,sBAAsB,YAAY;AACvC,UAAM,UAAU,MAAM,mBAAmB,UAAU,CAAC,MAAM;AACxD,UAAI,aAAa,SAAS;AACxB,eAAO,EAAE,OAAO,EAAE,SAAS,GAAG,SAAS,MAAM;AAAA,MAC/C,WAAW,YAAY,CAAC,GAAG;AACzB,eAAO,EAAE,OAAO,MAAM,SAAS,KAAK;AAAA,MACtC,OAAO;AACL,cAAM,IAAI,MAAM,2EAA2E;AAAA,MAC7F;AAAA,IACF,CAAC;AACD,WAAO,mBAAmB,SAAS,gBAAgB,QAAQ;AAAA,EAC7D,GAAG,IAAI;AACT;AACA,SAAS,gBAAgB,MAAM;AAC7B,MAAI,SAAS,MAAM;AACjB,WAAO;AAAA,EACT;AACA,QAAM,aAAa,KAAK;AACxB,MAAI,aAAa,UAAU,GAAG;AAC5B,UAAM,QAAQ,YAAY,IAAI;AAC9B,WAAO,EAAE,OAAO,SAAS,MAAM;AAAA,EACjC;AACA,SAAO,EAAE,OAAO,MAAM,SAAS,KAAK;AACtC;AACA,SAAS,YAAY,QAAQ;AAC3B,MAAI,OAAO,WAAW,GAAG;AACvB,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,MAAI,OAAO,cAAc,QAAQ;AAC/B,WAAO,MAAM,MAAM;AAAA,EACrB,OAAO;AACL,WAAO,OAAO,MAAM;AAAA,EACtB;AACF;AAGA,IAAI,kBAAkB,cAAc,QAAQ;AAAA,EAC1C,YAAY,QAAQ;AAClB,UAAM;AACN,SAAK,QAAQ;AAAA,EACf;AAAA,EACA,MAAM,WAAW;AACf,UAAM,gBAAgB,MAAM,KAAK,MAAM,SAAS;AAChD,UAAM,eAAe,cAAc,WAAW;AAC9C,UAAM,eAAe,aAAa,MAAM,IAAI,EAAE,IAAI,CAAC,SAAS;AAC1D,UAAI,KAAK,SAAS,IAAI,GAAG;AACvB,eAAO,KAAK,MAAM,GAAG,EAAE;AAAA,MACzB;AACA,aAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AACF;AAGA,IAAI,aAAa;AACjB,IAAI,YAAY,OAAO,KAAK;AAC5B,IAAI,cAAc,OAAO,OAAO;AAChC,IAAI,cAAc,OAAO,OAAO;AAChC,IAAI,0BAA0B,OAAO,iBAAiB;AACtD,IAAI,8BAA8B,OAAO,cAAc;AACvD,IAAI,aAAa,cAAc,QAAQ;AAAA,EACrC,YAAY,QAAQ,WAAW;AAC7B,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,YAAY;AACjB,SAAK,kBAAkB;AACvB,SAAK,uBAAuB;AAC5B,SAAK,gBAAgB;AACrB,SAAK,wBAAwB;AAC7B,SAAK,YAAY;AACjB,SAAK,kBAAkB;AACvB,SAAK,OAAO,IAAI,gBAAgB,MAAM;AACtC,QAAI,CAAC,WAAW;AACd,kBAAY,CAAC;AAAA,IACf;AACA,SAAK,YAAY,UAAU,cAAc,QAAQ,QAAQ;AACzD,SAAK,kBAAkB,UAAU;AACjC,SAAK,gBAAgB,UAAU;AAC/B,SAAK,wBAAwB,UAAU;AACvC,QAAI,UAAU,iBAAiB;AAC7B,mBAAa,OAAO,UAAU,aAAa,MAAM,MAAM,gEAAgE;AACvH,WAAK,kBAAkB;AACvB,WAAK,YAAY;AAAA,IACnB,OAAO;AACL,WAAK,YAAY,UAAU,YAAY,UAAU,YAAY;AAAA,IAC/D;AAAA,EACF;AAAA,EACA,MAAM,cAAc;AAClB,QAAI,CAAC,KAAK,sBAAsB;AAC9B,YAAM,KAAK,eAAe;AAAA,IAC5B;AACA,WAAO,KAAK,wBAAwB,OAAO,KAAK,KAAK,aAAa,IAAI,KAAK;AAAA,EAC7E;AAAA,EACA,MAAM,iBAAiB;AACrB,UAAM,sBAAsB,MAAM,KAAK,oBAAoB;AAC3D,QAAI,CAAC,KAAK,mBAAmB,CAAC,qBAAqB;AACjD,YAAM,IAAI,MAAM,2DAA2D;AAAA,IAC7E,WAAW,KAAK,mBAAmB,qBAAqB;AACtD,mBAAa,OAAO,oBAAoB,WAAW,KAAK,gBAAgB,QAAQ,MAAM,yCAAyC,KAAK,gBAAgB,OAAO,SAAS,IAAI,oEAAoE,oBAAoB,OAAO,SAAS,IAAI,IAAI;AAAA,IAC1R;AACA,QAAI,CAAC,KAAK,iBAAiB;AACzB,WAAK,kBAAkB;AAAA,IACzB;AACA,UAAM,SAAS,KAAK,gBAAgB,OAAO,CAAC,UAAU,SAAS;AAC7D,eAAS,QAAQ,SAAS,QAAQ,KAAK;AACvC,aAAO;AAAA,IACT,GAAG,CAAC,CAAC;AACL,UAAM,iBAAiB,OAAO,KAAK,MAAM,EAAE,OAAO,CAAC,SAAS,OAAO,QAAQ,CAAC;AAC5E,iBAAa,OAAO,eAAe,WAAW,GAAG,MAAM,mCAAmC,eAAe,SAAS,CAAC;AACnH,QAAI,KAAK,eAAe;AACtB,iBAAW,OAAO,OAAO,KAAK,KAAK,aAAa,GAAG;AACjD,cAAM,QAAQ,KAAK,gBAAgB,QAAQ,GAAG;AAC9C,YAAI,UAAU,IAAI;AAChB,gBAAM,IAAI,MAAM,cAAc,MAAM,yEAAyE,KAAK,gBAAgB,SAAS,IAAI,IAAI;AAAA,QACrJ;AAAA,MACF;AAAA,IACF;AACA,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EACA,MAAM,sBAAsB;AAC1B,QAAI,KAAK,WAAW;AAClB,YAAM,OAAO,MAAM,KAAK,KAAK,SAAS;AACtC,YAAM,eAAe,MAAM,KAAK,KAAK;AACrC,UAAI,aAAa,MAAM;AACrB,cAAM,IAAI,MAAM,oCAAoC;AAAA,MACtD;AACA,YAAM,YAAY,aAAa;AAC/B,YAAM,UAAU,KAAK,SAAS,WAAW,KAAK;AAC9C,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,MAAM,WAAW;AACf,QAAI,CAAC,KAAK,sBAAsB;AAC9B,YAAM,KAAK,eAAe;AAAA,IAC5B;AACA,QAAI,QAAQ,MAAM,KAAK,KAAK,SAAS;AACrC,QAAI,KAAK,WAAW;AAClB,cAAQ,MAAM,KAAK,CAAC;AAAA,IACtB;AACA,WAAO,MAAM,IAAI,CAAC,MAAM,KAAK,gBAAgB,CAAC,CAAC;AAAA,EACjD;AAAA,EACA,gBAAgB,MAAM;AACpB,UAAM,SAAS,KAAK,SAAS,IAAI;AACjC,UAAM,WAAW,CAAC;AAClB,UAAM,SAAS,CAAC;AAChB,aAAS,IAAI,GAAG,IAAI,KAAK,gBAAgB,QAAQ,KAAK;AACpD,YAAM,MAAM,KAAK,gBAAgB;AACjC,YAAM,SAAS,KAAK,gBAAgB,KAAK,cAAc,OAAO;AAC9D,UAAI,KAAK,yBAAyB,CAAC,QAAQ;AACzC;AAAA,MACF,OAAO;AACL,cAAM,QAAQ,OAAO;AACrB,YAAI,cAAc;AAClB,YAAI,UAAU,IAAI;AAChB,cAAI,UAAU,OAAO,YAAY,QAAQ;AACvC,0BAAc,OAAO;AAAA,UACvB,WAAW,WAAW,OAAO,YAAY,OAAO,UAAU;AACxD,kBAAM,IAAI,MAAM,mBAAmB,8BAA8B,MAAM;AAAA,UACzE,OAAO;AACL,0BAAc;AAAA,UAChB;AAAA,QACF,OAAO;AACL,gBAAM,aAAa,OAAO,KAAK;AAC/B,cAAI,MAAM,UAAU,GAAG;AACrB,gBAAI,UAAU,OAAO,UAAU,QAAQ;AACrC,4BAAc,KAAK,WAAW,KAAK;AAAA,YACrC,OAAO;AACL,4BAAc;AAAA,YAChB;AAAA,UACF,WAAW,CAAC,UAAU,CAAC,OAAO,OAAO;AACnC,0BAAc;AAAA,UAChB,OAAO;AACL,oBAAQ,OAAO,OAAO;AAAA,cACpB,KAAK;AACH,8BAAc;AACd;AAAA,cACF,KAAK;AACH,8BAAc,KAAK,MAAM,UAAU;AACnC;AAAA,cACF,KAAK;AACH,8BAAc,KAAK,WAAW,KAAK;AACnC;AAAA,cACF;AACE,8BAAc;AAAA,YAClB;AAAA,UACF;AAAA,QACF;AACA,kBAAU,OAAO,UAAU,OAAO,OAAO,cAAc,SAAS,OAAO;AAAA,MACzE;AAAA,IACF;AACA,QAAI,OAAO,KAAK,MAAM,EAAE,WAAW,GAAG;AACpC,aAAO;AAAA,IACT,OAAO;AACL,aAAO,EAAE,IAAI,UAAU,IAAI,OAAO;AAAA,IACpC;AAAA,EACF;AAAA,EACA,WAAW,OAAO;AAChB,QAAI,UAAU,OAAO,MAAM,YAAY,MAAM,QAAQ;AACnD,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,SAAS,MAAM,uBAAuB,MAAM;AAC1C,UAAM,SAAS,CAAC;AAChB,QAAI,aAAa;AACjB,UAAM,aAAa,KAAK;AACxB,QAAI,eAAe;AACnB,aAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,cAAQ,cAAc;AAAA,QACpB,KAAK;AACH,kBAAQ,KAAK,OAAO,CAAC,GAAG;AAAA,YACtB,KAAK;AACH,2BAAa,IAAI;AACjB,6BAAe;AACf;AAAA,YACF,KAAK,KAAK;AACR,2BAAa,IAAI;AACjB,kBAAI,KAAK,cAAc,OAAO,KAAK,iBAAiB;AAClD;AAAA,cACF;AACA,qBAAO,KAAK,EAAE;AACd,6BAAe;AACf;AAAA,YACF;AACE,6BAAe;AACf,2BAAa;AACb;AAAA,UACJ;AACA;AAAA,QACF,KAAK;AACH,kBAAQ,KAAK,OAAO,CAAC,GAAG;AAAA,YACtB,KAAK,KAAK;AACR,qBAAO,KAAK,KAAK,UAAU,YAAY,CAAC,CAAC;AACzC,6BAAe;AACf,2BAAa,IAAI;AACjB;AAAA,YACF;AAAA,UACF;AACA;AAAA,QACF,KAAK;AACH,kBAAQ,KAAK,OAAO,CAAC,GAAG;AAAA,YACtB,KAAK;AACH,6BAAe;AACf;AAAA,YACF;AAAA,UACF;AACA;AAAA,QACF,KAAK;AACH,kBAAQ,KAAK,OAAO,CAAC,GAAG;AAAA,YACtB,KAAK,KAAK;AACR,qBAAO,KAAK,KAAK,UAAU,YAAY,IAAI,CAAC,CAAC;AAC7C,6BAAe;AACf,2BAAa,IAAI;AACjB;AAAA,YACF,KAAK;AACH,6BAAe;AACf;AAAA,YACF;AACE,6BAAe;AACf;AAAA,UACJ;AACA;AAAA,QACF,KAAK;AACH,kBAAQ,KAAK,OAAO,CAAC,GAAG;AAAA,YACtB,KAAK;AACH,6BAAe;AACf;AAAA,YACF;AAAA,UACF;AACA;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,QAAI,iBAAiB,yBAAyB;AAC5C,aAAO,KAAK,KAAK,UAAU,YAAY,aAAa,CAAC,CAAC;AAAA,IACxD,OAAO;AACL,aAAO,KAAK,KAAK,UAAU,UAAU,CAAC;AAAA,IACxC;AACA,QAAI,wBAAwB,OAAO,WAAW,KAAK,gBAAgB,QAAQ;AACzE,YAAM,IAAI,MAAM,wCAAwC,KAAK,gBAAgB,qCAAqC,QAAQ;AAAA,IAC5H;AACA,WAAO;AAAA,EACT;AACF;AAGA,IAAI,qBAAqB,cAAc,aAAa;AAAA,EAClD,YAAY,kBAAkB;AAC5B,UAAM;AACN,SAAK,mBAAmB;AACxB,SAAK,WAAW;AAChB,SAAK,UAAU,iBAAiB,WAAW;AAC3C,UAAM,cAAc,KAAK,KAAK,KAAK,OAAO;AAC1C,QAAI,KAAK,UAAU,KAAK,cAAc,KAAK,cAAc,MAAM,CAAC,OAAO,UAAU,WAAW,GAAG;AAC7F,YAAM,IAAI,MAAM,gFAAgF,KAAK,SAAS;AAAA,IAChH;AACA,SAAK,YAAY,iBAAiB,2BAA2B;AAC7D,SAAK,eAAe,iBAAiB;AACrC,SAAK,uBAAuB,iBAAiB,wBAAwB,KAAK;AAC1E,SAAK,wBAAwB,iBAAiB;AAC9C,SAAK,wBAAwB,iBAAiB,yBAAyB;AACvE,SAAK,qBAAqB,iBAAiB,uBAAuB,QAAQ,QAAQ;AAClF,SAAK,kBAAkB,iBAAiB,oBAAoB,OAAO,OAAO;AAC1E,QAAI,CAAC,KAAK,sBAAsB,CAAC,KAAK,iBAAiB;AACrD,YAAM,IAAI,MAAM,sGAAsG;AAAA,IACxH;AAAA,EACF;AAAA,EACA,UAAU;AACR,WAAO;AAAA,EACT;AAAA,EACA,aAAa,OAAO,mBAAmB,CAAC,GAAG;AACzC,QAAI,CAAC,IAAI,EAAE,IAAI,YAAY,GAAG;AAC5B,YAAM,IAAI,MAAM,0DAA0D;AAAA,IAC5E;AACA,UAAM,qBAAqB,IAAI,mBAAmB,gBAAgB;AAClE,UAAM,mBAAmB,MAAM;AAC/B,WAAO;AAAA,EACT;AAAA,EACA,MAAM,QAAQ;AACZ,QAAI;AACF,WAAK,SAAS,MAAM,UAAU,aAAa,aAAa;AAAA,QACtD,OAAO,KAAK,yBAAyB,OAAO,OAAO,KAAK;AAAA,QACxD,OAAO;AAAA,MACT,CAAC;AAAA,IACH,SAAS,GAAP;AACA,YAAM,IAAI,MAAM,iDAAiD,EAAE,SAAS;AAAA,IAC9E;AACA,QAAI,CAAC,KAAK,QAAQ;AAChB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AACA,UAAM,iBAAiB,OAAO,gBAAgB,OAAO;AACrD,SAAK,eAAe,IAAI,eAAe;AACvC,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,eAAe,KAAK,aAAa;AAAA,IACxC,WAAW,KAAK,aAAa,eAAe,KAAK,cAAc;AAC7D,YAAM,IAAI,MAAM,wCAAwC,KAAK,yBAAyB,KAAK,aAAa,YAAY;AAAA,IACtH;AACA,UAAM,eAAe,KAAK,aAAa,wBAAwB,KAAK,MAAM;AAC1E,SAAK,WAAW,KAAK,aAAa,eAAe;AACjD,SAAK,SAAS,UAAU,KAAK,UAAU;AACvC,SAAK,SAAS,wBAAwB,KAAK;AAC3C,iBAAa,QAAQ,KAAK,QAAQ;AAClC,SAAK,WAAW,IAAI,aAAa,KAAK,OAAO;AAC7C,SAAK,WAAW,IAAI,aAAa,KAAK,OAAO;AAC7C;AAAA,EACF;AAAA,EACA,MAAM,OAAO;AACX,QAAI,KAAK,UAAU;AACjB,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,QAAI;AACJ,QAAI;AACJ,UAAM,iBAAiB,MAAM,KAAK,aAAa;AAC/C,QAAI,KAAK,oBAAoB;AAC3B,YAAM,WAAW,KAAK,aAAa,eAAe,aAAa;AAC/D,0BAAoB,KAAK,4BAA4B,UAAU,CAAC,KAAK,WAAW,KAAK,sBAAsB,CAAC,CAAC;AAAA,IAC/G;AACA,QAAI,KAAK,iBAAiB;AACxB,YAAM,WAAW,KAAK,aAAa,eAAe,aAAa;AAC/D,uBAAiB,KAAK,4BAA4B,UAAU,CAAC,KAAK,YAAY,KAAK,SAAS,CAAC,CAAC;AAAA,IAChG;AACA,WAAO;AAAA,MACL,OAAO,EAAE,eAAe,mBAAmB,YAAY,eAAe;AAAA,MACtE,MAAM;AAAA,IACR;AAAA,EACF;AAAA,EACA,MAAM,UAAU;AACd,YAAQ,MAAM,KAAK,KAAK,GAAG;AAAA,EAC7B;AAAA,EACA,MAAM,eAAe;AACnB,UAAM,gBAAgB,CAAC;AACvB,UAAM,gBAAgB,CAAC;AACvB,QAAI,gBAAgB;AACpB,WAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,YAAM,aAAa,YAAY,MAAM;AACnC,YAAI,KAAK,oBAAoB;AAC3B,eAAK,SAAS,sBAAsB,KAAK,QAAQ;AACjD,cAAI,KAAK,SAAS,OAAO,WAAW;AAClC,oBAAQ,EAAE,eAAe,cAAc,CAAC;AAAA,UAC1C;AACA,wBAAc,KAAK,KAAK,SAAS,MAAM,GAAG,KAAK,oBAAoB,CAAC;AAAA,QACtE;AACA,YAAI,KAAK,iBAAiB;AACxB,eAAK,SAAS,uBAAuB,KAAK,QAAQ;AAClD,wBAAc,KAAK,KAAK,SAAS,MAAM,CAAC;AAAA,QAC1C;AACA,YAAI,EAAE,kBAAkB,KAAK,WAAW;AACtC,wBAAc,UAAU;AACxB,kBAAQ,EAAE,eAAe,cAAc,CAAC;AAAA,QAC1C;AAAA,MACF,GAAG,KAAK,UAAU,KAAK,eAAe,GAAG;AAAA,IAC3C,CAAC;AAAA,EACH;AAAA,EACA,OAAO;AACL,QAAI,CAAC,KAAK,UAAU;AAClB,WAAK,WAAW;AAChB,WAAK,SAAS,WAAW;AACzB,WAAK,aAAa,MAAM;AACxB,UAAI,KAAK,UAAU,QAAQ,KAAK,OAAO,UAAU,EAAE,SAAS,GAAG;AAC7D,aAAK,OAAO,UAAU,EAAE,GAAG,KAAK;AAAA,MAClC;AAAA,IACF;AAAA,EACF;AAAA,EACA,UAAU;AACR,UAAM,IAAI,MAAM,iDAAiD;AAAA,EACnE;AAAA,EACA,gBAAgB;AACd,WAAO,KAAK;AAAA,EACd;AAAA,EACA,aAAa,OAAO;AAClB,UAAM,YAAY,MAAM,GAAG;AAC3B,UAAM,WAAW,IAAI,aAAa,MAAM,SAAS,SAAS;AAC1D,UAAM,QAAQ,CAAC,MAAM,MAAM,SAAS,IAAI,MAAM,IAAI,SAAS,CAAC;AAC5D,WAAO;AAAA,EACT;AAAA,EACA,4BAA4B,UAAU,OAAO;AAC3C,UAAM,OAAO,IAAI,aAAa,aAAa,cAAc,KAAK,CAAC;AAC/D,SAAK,IAAI,UAAU,KAAK,SAAS,SAAS,MAAM;AAChD,WAAO,OAAO,MAAM,KAAK;AAAA,EAC3B;AACF;AAGA,IAAI,iBAAiB,cAAc,aAAa;AAAA,EAC9C,YAAY,oBAAoB,cAAc;AAC5C,UAAM;AACN,SAAK,qBAAqB;AAC1B,SAAK,eAAe;AACpB,SAAK,WAAW;AAChB,SAAK,SAAS;AACd,QAAI,KAAK,aAAa,GAAG;AACvB,WAAK,SAAS;AACd,WAAK,WAAW,CAAC,KAAK,aAAa,cAAc,KAAK,aAAa,WAAW;AAC9E,WAAK,aAAa,SAAS,CAAC,CAAC,GAAG,OAAO;AACvC,UAAI,KAAK,aAAa,YAAY;AAChC,cAAM,qBAAqB,KAAK,aAAa,cAAc,IAAI,KAAK,mBAAmB;AACvF,cAAM,sBAAsB,KAAK,aAAa,eAAe,IAAI,KAAK,mBAAmB;AACzF,cAAM,kBAAkB,IAAI,sBAAsB;AAClD,cAAM,mBAAmB,IAAI,uBAAuB;AACpD,cAAM,eAAe,iBAAiB;AACtC,cAAM,gBAAgB,sBAAsB;AAC5C,aAAK,UAAU,SAAS,CAAC,iBAAiB,gBAAgB,eAAe,YAAY,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,MAChG,OAAO;AACL,aAAK,UAAU,SAAS,CAAC,GAAG,GAAG,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,MAC9C;AAAA,IACF;AAAA,EACF;AAAA,EACA,UAAU;AACR,WAAO;AAAA,EACT;AAAA,EACA,aAAa,OAAO,oBAAoB,eAAe,CAAC,GAAG;AACzD,QAAI,CAAC,IAAI,EAAE,IAAI,YAAY,GAAG;AAC5B,YAAM,IAAI,MAAM,0DAA0D;AAAA,IAC5E;AACA,QAAI,CAAC,oBAAoB;AACvB,2BAAqB,SAAS,cAAc,OAAO;AACnD,UAAI,CAAC,aAAa,eAAe,CAAC,aAAa,cAAc;AAC3D,cAAM,IAAI,MAAM,wGAAwG;AAAA,MAC1H;AACA,yBAAmB,QAAQ,aAAa;AACxC,yBAAmB,SAAS,aAAa;AAAA,IAC3C;AACA,UAAM,iBAAiB,IAAI,eAAe,oBAAoB,YAAY;AAC1E,UAAM,eAAe,MAAM;AAC3B,WAAO;AAAA,EACT;AAAA,EACA,MAAM,QAAQ;AACZ,QAAI,KAAK,aAAa,YAAY;AAChC,mBAAa,OAAO,KAAK,aAAa,eAAe,UAAU,KAAK,aAAa,eAAe,eAAe,MAAM,+BAA+B,KAAK,aAAa,oDAAoD;AAAA,IAC5N;AACA,QAAI;AACF,WAAK,SAAS,MAAM,UAAU,aAAa,aAAa;AAAA,QACtD,OAAO;AAAA,UACL,UAAU,KAAK,aAAa;AAAA,UAC5B,YAAY,KAAK,aAAa,aAAa,KAAK,aAAa,aAAa;AAAA,UAC1E,OAAO,KAAK,mBAAmB;AAAA,UAC/B,QAAQ,KAAK,mBAAmB;AAAA,QAClC;AAAA,MACF,CAAC;AAAA,IACH,SAAS,GAAP;AACA,QAAE,UAAU,iDAAiD,EAAE;AAC/D,YAAM;AAAA,IACR;AACA,QAAI,CAAC,KAAK,QAAQ;AAChB,YAAM,IAAI,MAAM,qCAAqC;AAAA,IACvD;AACA,QAAI;AACF,WAAK,mBAAmB,YAAY,KAAK;AAAA,IAC3C,SAAS,OAAP;AACA,cAAQ,IAAI,KAAK;AACjB,WAAK,mBAAmB,MAAM,OAAO,IAAI,gBAAgB,KAAK,MAAM;AAAA,IACtE;AACA,SAAK,mBAAmB,KAAK;AAC7B,SAAK,WAAW;AAChB,WAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,WAAK,mBAAmB,mBAAmB,MAAM;AAC/C,gBAAQ;AAAA,MACV;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EACA,MAAM,OAAO;AACX,QAAI,KAAK,UAAU;AACjB,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,QAAI;AACJ,QAAI;AACF,YAAM,gBAAgB,WAAW,KAAK,kBAAkB;AAAA,IAC1D,SAAS,GAAP;AACA,YAAM,IAAI,MAAM,4CAA4C,KAAK,UAAU,CAAC,GAAG;AAAA,IACjF;AACA,QAAI,KAAK,QAAQ;AACf,UAAI;AACF,eAAO,EAAE,OAAO,KAAK,mBAAmB,GAAG,GAAG,MAAM,MAAM;AAAA,MAC5D,SAAS,GAAP;AACA,cAAM,IAAI,MAAM,oCAAoC,EAAE,SAAS;AAAA,MACjE,UAAE;AACA,YAAI,QAAQ;AAAA,MACd;AAAA,IACF,OAAO;AACL,aAAO,EAAE,OAAO,KAAK,MAAM,MAAM;AAAA,IACnC;AAAA,EACF;AAAA,EACA,eAAe;AACb,QAAI,KAAK,aAAa,eAAe,KAAK,aAAa,iBAAiB,KAAK,mBAAmB,UAAU,KAAK,aAAa,eAAe,KAAK,mBAAmB,WAAW,KAAK,aAAa,eAAe;AAC7M,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAAA,EACA,mBAAmB,KAAK;AACtB,WAAO,KAAK,MAAM;AAChB,YAAM,gBAAgB,WAAW,KAAK,KAAK,SAAS,GAAG,CAAC;AACxD,UAAI;AACJ,qBAAe,MAAM,cAAc,eAAe,KAAK,SAAS,KAAK,YAAY,KAAK,UAAU,UAAU;AAC1G,YAAM,QAAQ,aAAa;AAC3B,aAAO,QAAQ,cAAc,MAAM,MAAM,CAAC,CAAC;AAAA,IAC7C,CAAC;AAAA,EACH;AAAA,EACA,MAAM,UAAU;AACd,YAAQ,MAAM,KAAK,KAAK,GAAG;AAAA,EAC7B;AAAA,EACA,OAAO;AACL,UAAM,SAAS,KAAK,OAAO,UAAU;AACrC,WAAO,QAAQ,CAAC,UAAU,MAAM,KAAK,CAAC;AACtC,QAAI;AACF,WAAK,mBAAmB,YAAY;AAAA,IACtC,SAAS,OAAP;AACA,cAAQ,IAAI,KAAK;AACjB,WAAK,mBAAmB,MAAM;AAAA,IAChC;AACA,SAAK,WAAW;AAAA,EAClB;AAAA,EACA,UAAU;AACR,UAAM,IAAI,MAAM,iDAAiD;AAAA,EACnE;AACF;AAGA,IAAI,aAAa,MAAM;AACvB;AAGA,IAAI,iBAAiB,cAAc,aAAa;AAAA,EAC9C,MAAM,WAAW;AACf,WAAO,IAAI,cAAc,MAAM,SAAS;AAAA,EAC1C;AACF;AACA,IAAI,gBAAgB,cAAc,eAAe;AAAA,EAC/C,YAAY,UAAU,WAAW;AAC/B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,OAAO,IAAI,kBAAkB,UAAU,SAAS;AAAA,EACvD;AAAA,EACA,UAAU;AACR,WAAO,KAAK,KAAK,QAAQ;AAAA,EAC3B;AAAA,EACA,MAAM,OAAO;AACX,WAAO,KAAK,KAAK,KAAK;AAAA,EACxB;AACF;AACA,IAAI,oBAAoB,cAAc,kBAAkB;AAAA,EACtD,YAAY,UAAU,WAAW;AAC/B,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,YAAY;AACjB,SAAK,YAAY;AAAA,EACnB;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ,eAAe,KAAK;AAAA,EACtD;AAAA,EACA,MAAM,OAAO;AACX,UAAM,cAAc,MAAM,KAAK,SAAS,KAAK;AAC7C,QAAI,YAAY,MAAM;AACpB,UAAI,KAAK,cAAc,IAAI;AACzB,eAAO;AAAA,MACT;AACA,WAAK,YAAY,KAAK,KAAK,SAAS;AACpC,WAAK,YAAY;AACjB,aAAO;AAAA,IACT;AACA,UAAM,QAAQ,YAAY,MAAM,MAAM,KAAK,SAAS;AACpD,UAAM,KAAK,KAAK,YAAY,MAAM;AAClC,eAAW,QAAQ,MAAM,MAAM,GAAG,EAAE,GAAG;AACrC,WAAK,YAAY,KAAK,IAAI;AAAA,IAC5B;AACA,SAAK,YAAY,MAAM,MAAM,SAAS;AACtC,WAAO;AAAA,EACT;AACF;AAGA,IAAI,oBAAoB,cAAc,aAAa;AAAA,EACjD,aAAa;AACX,WAAO,IAAI,aAAa,IAAI;AAAA,EAC9B;AACF;AACA,IAAI,eAAe,cAAc,eAAe;AAAA,EAC9C,YAAY,UAAU;AACpB,UAAM;AACN,SAAK,WAAW;AAChB,SAAK,OAAO,IAAI,iBAAiB,QAAQ;AAAA,EAC3C;AAAA,EACA,UAAU;AACR,WAAO,KAAK,KAAK,QAAQ;AAAA,EAC3B;AAAA,EACA,MAAM,OAAO;AACX,WAAO,KAAK,KAAK,KAAK;AAAA,EACxB;AACF;AACA,IAAI,mBAAmB,cAAc,kBAAkB;AAAA,EACrD,YAAY,UAAU;AACpB,UAAM;AACN,SAAK,WAAW;AAChB,QAAI,IAAI,EAAE,IAAI,YAAY,GAAG;AAC3B,WAAK,UAAU,IAAI,YAAY,OAAO;AAAA,IACxC,OAAO;AACL,YAAM,EAAE,cAAc,IAAI,uBAAuB;AACjD,WAAK,UAAU,IAAI,cAAc,MAAM;AAAA,IACzC;AAAA,EACF;AAAA,EACA,UAAU;AACR,WAAO,GAAG,KAAK,SAAS,QAAQ;AAAA,EAClC;AAAA,EACA,MAAM,OAAO;AACX,UAAM,cAAc,MAAM,KAAK,SAAS,KAAK;AAC7C,QAAI;AACJ,QAAI,YAAY,MAAM;AACpB,aAAO;AAAA,IACT,OAAO;AACL,cAAQ,YAAY;AAAA,IACtB;AACA,QAAI;AACJ,QAAI,IAAI,EAAE,IAAI,YAAY,GAAG;AAC3B,aAAO,KAAK,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAAA,IACpD,OAAO;AACL,aAAO,KAAK,QAAQ,MAAM,OAAO,KAAK,MAAM,MAAM,CAAC;AAAA,IACrD;AACA,SAAK,YAAY,KAAK,IAAI;AAC1B,WAAO;AAAA,EACT;AACF;AAGA,IAAI,oBAAoB,cAAc,kBAAkB;AAAA,EACtD,YAAY,MAAM,UAAU,CAAC,GAAG;AAC9B,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,UAAU;AACf,iBAAa,OAAO,gBAAgB,eAAe,IAAI,EAAE,IAAI,YAAY,IAAI,gBAAgB,QAAQ,gBAAgB,OAAO,QAAQ,MAAM,sEAAsE;AAChN,SAAK,SAAS,QAAQ,UAAU;AAChC,SAAK,YAAY,QAAQ,aAAa,OAAO;AAAA,EAC/C;AAAA,EACA,UAAU;AACR,WAAO,cAAc,KAAK;AAAA,EAC5B;AAAA,EACA,MAAM,OAAO;AACX,QAAI,KAAK,WAAW,KAAK,gBAAgB,aAAa,KAAK,KAAK,aAAa,KAAK,KAAK,OAAO;AAC5F,aAAO,EAAE,OAAO,MAAM,MAAM,KAAK;AAAA,IACnC;AACA,UAAM,QAAQ,IAAI,QAAQ,CAAC,SAAS,WAAW;AAC7C,YAAM,MAAM,KAAK,SAAS,KAAK;AAC/B,UAAI,KAAK,gBAAgB,YAAY;AACnC,gBAAQ,IAAI,WAAW,KAAK,KAAK,MAAM,KAAK,QAAQ,GAAG,CAAC,CAAC;AAAA,MAC3D,OAAO;AACL,cAAM,aAAa,IAAI,WAAW;AAClC,mBAAW,SAAS,CAAC,UAAU;AAC7B,cAAI,OAAO,WAAW;AACtB,cAAI,gBAAgB,aAAa;AAC/B,mBAAO,IAAI,WAAW,IAAI;AAAA,UAC5B;AACA,cAAI,EAAE,gBAAgB,aAAa;AACjC,mBAAO,OAAO,IAAI,UAAU,mCAAmC,CAAC;AAAA,UAClE;AACA,kBAAQ,IAAI;AAAA,QACd;AACA,mBAAW,UAAU,CAAC,UAAU;AAC9B,iBAAO,OAAO,IAAI,MAAM,SAAS,CAAC;AAAA,QACpC;AACA,mBAAW,UAAU,CAAC,UAAU;AAC9B,iBAAO,OAAO,IAAI,MAAM,MAAM,IAAI,CAAC;AAAA,QACrC;AACA,cAAM,SAAS,KAAK,KAAK,MAAM,KAAK,QAAQ,GAAG;AAC/C,mBAAW,kBAAkB,MAAM;AAAA,MACrC;AACA,WAAK,SAAS;AAAA,IAChB,CAAC;AACD,WAAO,EAAE,OAAO,MAAM,OAAO,MAAM,MAAM;AAAA,EAC3C;AACF;AAGA,eAAe,iBAAiB,KAAK,UAAU,CAAC,GAAG,WAAW;AAC5D,MAAI;AACJ,MAAI;AACJ,MAAI,OAAO,QAAQ,UAAU;AAC3B,gBAAY;AAAA,EACd,OAAO;AACL,gBAAY,IAAI;AAChB,kBAAc,0BAA0B,GAAG;AAAA,EAC7C;AACA,QAAM,WAAW,OAAO,aAAa,aAAa,OAAO,WAAW,WAAW;AAC/E,MAAI,SAAS,IAAI;AACf,UAAM,aAAa,IAAI,WAAW,MAAM,SAAS,YAAY,CAAC;AAC9D,WAAO,IAAI,kBAAkB,YAAY,OAAO;AAAA,EAClD,OAAO;AACL,UAAM,IAAI,MAAM,SAAS,UAAU;AAAA,EACrC;AACF;AACA,IAAI,4BAA4B,CAAC,YAAY;AAC3C,QAAM,QAAQ;AAAA,IACZ,QAAQ,QAAQ;AAAA,IAChB,SAAS,QAAQ;AAAA,IACjB,MAAM,QAAQ;AAAA,IACd,MAAM,QAAQ;AAAA,IACd,aAAa,QAAQ;AAAA,IACrB,OAAO,QAAQ;AAAA,IACf,UAAU,QAAQ;AAAA,IAClB,UAAU,QAAQ;AAAA,IAClB,WAAW,QAAQ;AAAA,EACrB;AACA,SAAO;AACT;AAGA,SAAS,YAAY,QAAQ;AAC3B,SAAO,OAAO,WAAW,YAAY,OAAO,MAAM,GAAG,CAAC,MAAM;AAC9D;AAGA,IAAI,iBAAiB,cAAc,WAAW;AAAA,EAC5C,YAAY,QAAQ,UAAU,CAAC,GAAG;AAChC,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,UAAU;AAAA,EACjB;AAAA,EACA,MAAM,WAAW;AACf,QAAI,YAAY,KAAK,KAAK,KAAK,IAAI,EAAE,IAAI,SAAS,GAAG;AACnD,YAAM,KAAK,WAAW;AACtB,WAAK,QAAQ,GAAG,aAAa,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,IAClD;AACA,WAAO,IAAI,kBAAkB,KAAK,OAAO,KAAK,OAAO;AAAA,EACvD;AACF;AAGA,IAAI,gBAAgB,cAAc,WAAW;AAAA,EAC3C,YAAY,KAAK,cAAc,CAAC,GAAG;AACjC,UAAM;AACN,SAAK,MAAM;AACX,SAAK,cAAc;AAAA,EACrB;AAAA,EACA,MAAM,WAAW;AACf,QAAI,YAAY,KAAK,GAAG,GAAG;AACzB,aAAO,IAAI,eAAe,KAAK,KAAK,KAAK,WAAW,EAAE,SAAS;AAAA,IACjE,OAAO;AACL,aAAO,iBAAiB,KAAK,KAAK,KAAK,WAAW;AAAA,IACpD;AAAA,EACF;AACF;AAGA,SAAS,IAAI,QAAQ,YAAY,CAAC,GAAG;AACnC,SAAO,IAAI,WAAW,IAAI,cAAc,MAAM,GAAG,SAAS;AAC5D;AACA,SAAS,KAAK,GAAG;AACf,QAAM,OAAO,qBAAqB,CAAC;AACnC,SAAO,sBAAsB,YAAY,IAAI;AAC/C;AACA,SAAS,UAAU,YAAY;AAC7B,SAAO,sBAAsB,YAAY;AACvC,UAAM,MAAM,MAAM,WAAW;AAC7B,WAAO,qBAAqB,MAAM,IAAI,KAAK,CAAC;AAAA,EAC9C,CAAC;AACH;AACA,eAAe,OAAO,oBAAoB,cAAc;AACtD,SAAO,eAAe,OAAO,oBAAoB,YAAY;AAC/D;AACA,eAAe,WAAW,kBAAkB;AAC1C,SAAO,mBAAmB,OAAO,gBAAgB;AACnD;AAGA,IAAI,WAAW;AAGf,SAAS,iBAAiB,SAAS,QAAQ;AACzC,MAAI,CAAC,MAAM,QAAQ,OAAO,GAAG;AAC3B,cAAU,CAAC,OAAO;AAAA,EACpB;AACA,UAAQ,QAAQ,CAAC,MAAM;AACrB,QAAI,KAAK,MAAM;AACb,mBAAa,OAAO,EAAE,UAAU,aAAa,MAAM,GAAG,+DAA+D;AAAA,IACvH;AAAA,EACF,CAAC;AACH;AAGA,IAAI,aAAa,qBAAqB;AACtC,IAAI,iBAAiB,cAAc,cAAc;AAAA,EAC/C,cAAc;AACZ,UAAM;AACN,SAAK,YAAY;AACjB,SAAK,WAAW;AAChB,SAAK,OAAO,IAAI,YAAY,MAAM,OAAO,CAAC;AAAA,EAC5C;AAAA,EACA,aAAa;AACX,WAAO,eAAe;AAAA,EACxB;AAAA,EACA,MAAM,QAAQ,OAAO,OAAO;AAC1B,QAAI,KAAK,UAAU;AACjB,WAAK,WAAW;AAChB,UAAI,IAAI,EAAE,IAAI,SAAS,GAAG;AACxB,6BAAqB,KAAK,oPAAoP;AAAA,MAChR;AAAA,IACF;AACA,UAAM,SAAS,EAAE,IAAI,KAAK,WAAW,EAAE;AACvC,SAAK,KAAK,IAAI,QAAQ,EAAE,QAAQ,OAAO,UAAU,EAAE,CAAC;AACpD,WAAO;AAAA,EACT;AAAA,EACA,eAAe,OAAO,OAAO,QAAQ;AACnC,QAAI;AACJ,QAAI,UAAU,YAAY,UAAU,QAAQ,OAAO,SAAS,KAAK,aAAa,SAAS,OAAO,EAAE,GAAG;AACjG,YAAM,gBAAgB,OAAO,IAAI,CAAC,MAAM,aAAa,aAAa,CAAC,CAAC;AACpE,cAAQ,KAAK,MAAM,eAAe,OAAO,KAAK;AAAA,IAChD,OAAO;AACL,cAAQ,KAAK,MAAM,QAAQ,OAAO,KAAK;AAAA,IACzC;AACA,WAAO,EAAE,QAAQ,OAAO,OAAO,MAAM;AAAA,EACvC;AAAA,EACA,SAAS,QAAQ;AACf,QAAI,KAAK,KAAK,IAAI,MAAM,GAAG;AACzB,YAAM,aAAa,KAAK,KAAK,IAAI,MAAM;AACvC,aAAO,WAAW;AAAA,IACpB;AACA,WAAO;AAAA,EACT;AAAA,EACA,OAAO,QAAQ;AACb,UAAM,aAAa,KAAK,KAAK,IAAI,MAAM;AACvC,eAAW;AAAA,EACb;AAAA,EACA,OAAO,QAAQ;AACb,QAAI,KAAK,KAAK,IAAI,MAAM,GAAG;AACzB,YAAM,aAAa,KAAK,KAAK,IAAI,MAAM;AACvC,iBAAW;AAAA,IACb;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ,OAAO,OAAO,UAAU;AAC3C,SAAK,KAAK,IAAI,QAAQ,EAAE,QAAQ,OAAO,SAAS,CAAC;AAAA,EACnD;AAAA,EACA,aAAa;AACX,WAAO,KAAK,KAAK,WAAW;AAAA,EAC9B;AAAA,EACA,MAAM,KAAK,QAAQ;AACjB,WAAO,KAAK,SAAS,MAAM;AAAA,EAC7B;AAAA,EACA,SAAS,QAAQ;AACf,UAAM,EAAE,OAAO,mBAAmB,IAAI,KAAK,KAAK,IAAI,MAAM;AAC1D,QAAI,UAAU,aAAa;AACzB,YAAM,aAAa,KAAK,SAAS,mBAAmB,KAAK,MAAM;AAC/D,YAAM,aAAa,KAAK,SAAS,mBAAmB,KAAK,MAAM;AAC/D,aAAO,qBAAqB,uBAAuB,YAAY,UAAU;AAAA,IAC3E;AACA,WAAO,KAAK,KAAK,IAAI,MAAM,EAAE;AAAA,EAC/B;AAAA,EACA,WAAW,GAAG;AACZ,UAAM,OAAO,KAAK,SAAS,EAAE,MAAM;AACnC,QAAI,EAAE,UAAU,UAAU;AACxB,UAAI;AACF,cAAM,UAAU,KAAK,IAAI,CAAC,MAAM,aAAa,aAAa,CAAC,CAAC;AAC5D,eAAO,OAAO,EAAE,OAAO,EAAE,OAAO,OAAO;AAAA,MACzC,SAAS,IAAP;AACA,cAAM,IAAI,MAAM,kDAAkD;AAAA,MACpE;AAAA,IACF;AACA,WAAO,OAAO,EAAE,OAAO,EAAE,OAAO,IAAI;AAAA,EACtC;AAAA,EACA,WAAW,QAAQ,OAAO,OAAO;AAC/B,WAAO,OAAO,EAAE,yBAAyB,KAAK,eAAe,OAAO,OAAO,MAAM,GAAG,IAAI;AAAA,EAC1F;AAAA,EACA,YAAY,QAAQ,QAAQ,OAAO;AACjC,QAAI,KAAK,KAAK,IAAI,MAAM,GAAG;AACzB,WAAK,KAAK,IAAI,MAAM,EAAE;AACtB,UAAI,CAAC,SAAS,KAAK,KAAK,IAAI,MAAM,EAAE,WAAW,GAAG;AAChD,eAAO;AAAA,MACT;AACA,YAAM,EAAE,mBAAmB,IAAI,KAAK,KAAK,IAAI,MAAM;AACnD,UAAI,sBAAsB,MAAM;AAC9B,aAAK,YAAY,mBAAmB,KAAK,QAAQ,IAAI;AACrD,aAAK,YAAY,mBAAmB,KAAK,QAAQ,IAAI;AAAA,MACvD;AACA,WAAK,KAAK,OAAO,MAAM;AAAA,IACzB;AACA,WAAO;AAAA,EACT;AAAA,EACA,8BAA8B,YAAY;AACxC,SAAK,YAAY,WAAW,MAAM;AAAA,EACpC;AAAA,EACA,MAAM,KAAK,GAAG;AACZ,UAAM,QAAQ,aAAa,IAAI;AAC/B,MAAE;AACF,UAAM,WAAW,aAAa,IAAI,IAAI;AACtC,WAAO,EAAE,SAAS;AAAA,EACpB;AAAA,EACA,SAAS;AACP,WAAO;AAAA,MACL,YAAY;AAAA,MACZ,SAAS,CAAC,oHAAoH;AAAA,IAChI;AAAA,EACF;AAAA,EACA,MAAM,WAAW;AACf,qBAAiB,CAAC,SAAS,GAAG,OAAO;AACrC,UAAM,WAAW,KAAK,SAAS,UAAU,MAAM;AAC/C,WAAO,WAAW,UAAU,OAAO,QAAQ;AAAA,EAC7C;AAAA,EACA,UAAU;AAAA,EACV;AAAA,EACA,iBAAiB;AACf,WAAO;AAAA,EACT;AAAA,EACA,UAAU;AACR,WAAO,MAAM,QAAQ;AAAA,EACvB;AACF;AACA,eAAe,aAAa;AAG5B,IAAI,iBAAiB,CAAC;AACtBA,UAAS,gBAAgB;AAAA,EACvB,SAAS,MAAM;AAAA,EACf,cAAc,MAAM;AAAA,EACpB,oBAAoB,MAAM;AAAA,EAC1B,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,YAAY,MAAM;AAAA,EAClB,WAAW,MAAM;AAAA,EACjB,SAAS,MAAM;AAAA,EACf,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,cAAc,MAAM;AAAA,EACpB,cAAc,MAAM;AAAA,EACpB,kBAAkB,MAAM;AAAA,EACxB,aAAa,MAAM;AAAA,EACnB,eAAe,MAAM;AAAA,EACrB,UAAU,MAAM;AAAA,EAChB,cAAc,MAAM;AAAA,EACpB,SAAS,MAAM;AAAA,EACf,SAAS,MAAM;AAAA,EACf,aAAa,MAAM;AAAA,EACnB,aAAa,MAAM;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,SAAS,MAAM;AAAA,EACf,cAAc,MAAM;AAAA,EACpB,UAAU,MAAM;AAAA,EAChB,kBAAkB,MAAM;AAAA,EACxB,iBAAiB,MAAM;AAAA,EACvB,0BAA0B,MAAM;AAAA,EAChC,WAAW,MAAM;AAAA,EACjB,WAAW,MAAM;AAAA,EACjB,aAAa,MAAM;AAAA,EACnB,aAAa,MAAM;AAAA,EACnB,eAAe,MAAM;AAAA,EACrB,WAAW,MAAM;AAAA,EACjB,yBAAyB,MAAM;AAAA,EAC/B,mBAAmB,MAAM;AAAA,EACzB,4BAA4B,MAAM;AAAA,EAClC,UAAU,MAAM;AAAA,EAChB,uBAAuB,MAAM;AAAA,EAC7B,kBAAkB,MAAM;AAAA,EACxB,kBAAkB,MAAM;AAAA,EACxB,iBAAiB,MAAM;AAAA,EACvB,4BAA4B,MAAM;AAAA,EAClC,SAAS,MAAM;AAAA,EACf,UAAU,MAAM;AAAA,EAChB,UAAU,MAAM;AAAA,EAChB,eAAe,MAAM;AAAA,EACrB,YAAY,MAAM;AACpB,CAAC;AAGD,SAAS,cAAc,MAAM;AAC3B,QAAM,eAAe,IAAI,aAAa,KAAK,MAAM;AACjD,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,iBAAa,KAAK,KAAK,IAAI,KAAK,EAAE;AAAA,EACpC;AACA,SAAO;AACT;AACA,IAAI,OAAO,CAAC,SAAS;AACnB,QAAM,EAAE,EAAE,IAAI,KAAK;AACnB,QAAM,aAAa,KAAK;AACxB,mBAAiB,GAAG,KAAK;AACzB,MAAI,eAAe,IAAI,aAAa,aAAa,cAAc,EAAE,KAAK,CAAC;AACvE,QAAM,SAAS,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC7C,iBAAe,cAAc,MAAM;AACnC,SAAO,WAAW,WAAW,cAAc,EAAE,OAAO,EAAE,KAAK;AAC7D;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,6BAA6B,KAAK;AACzC,SAAO,CAAC,QAAQ,QAAQ,OAAO,OAAO,UAAU;AAC9C,UAAM,WAAW,qBAAqB,2BAA2B,QAAQ,MAAM;AAC/E,UAAM,aAAa,SAAS;AAC5B,UAAM,gBAAgB,aAAa,eAAe,QAAQ;AAC1D,UAAM,aAAa,aAAa,cAAc,QAAQ;AACtD,UAAM,SAAS,aAAa,uBAAuB,OAAO,UAAU;AACpE,UAAM,QAAQ,OAAO;AACrB,UAAM,QAAQ,OAAO;AACrB,UAAM,WAAW,aAAa,eAAe,MAAM;AACnD,UAAM,WAAW,aAAa,eAAe,MAAM;AACnD,UAAM,iBAAiB,qBAAqB,iBAAiB,QAAQ,QAAQ;AAC7E,UAAM,iBAAiB,qBAAqB,iBAAiB,QAAQ,QAAQ;AAC7E,QAAI,eAAe,SAAS,eAAe,WAAW,GAAG;AACvD,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,eAAO,KAAK,IAAI,MAAM,IAAI,MAAM,SAAS,MAAM,IAAI,MAAM,OAAO;AAAA,MAClE;AAAA,IACF,OAAO;AACL,eAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,cAAM,MAAM,aAAa,WAAW,GAAG,YAAY,aAAa;AAChE,cAAM,OAAO,IAAI,MAAM,CAAC,KAAK;AAC7B,uBAAe,QAAQ,CAAC,MAAM,KAAK,KAAK,CAAC;AACzC,cAAM,SAAS,aAAa,WAAW,MAAM,OAAO,QAAQ;AAC5D,cAAM,OAAO,IAAI,MAAM,CAAC,KAAK;AAC7B,uBAAe,QAAQ,CAAC,MAAM,KAAK,KAAK,CAAC;AACzC,cAAM,SAAS,aAAa,WAAW,MAAM,OAAO,QAAQ;AAC5D,eAAO,KAAK,IAAI,MAAM,SAAS,MAAM,OAAO;AAAA,MAC9C;AAAA,IACF;AACA,WAAO,CAAC,QAAQ,QAAQ;AAAA,EAC1B;AACF;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,OAAO,MAAM,MAAM,IAAI;AACrC,QAAM,WAAW,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AACjD,QAAM,WAAW,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AACjD,QAAM,cAAc,SAAS,eAAe,MAAM,OAAO,WAAW;AACpE,QAAM,WAAW,SAAS,KAAK,IAAI,YAAY,MAAM;AACrD,WAAS,qBAAqB;AAAA,IAC5B,MAAM,SAAS,eAAe,MAAM,OAAO,WAAW,QAAQ;AAAA,IAC9D,MAAM,SAAS,eAAe,MAAM,OAAO,WAAW,QAAQ;AAAA,EAChE;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,UAAU,OAAO,QAAQ,WAAW;AAClD,MAAI,UAAU,aAAa;AACzB,UAAM,QAAQ,OAAO,UAAU,OAAO,SAAS;AAC/C,UAAM,QAAQ,OAAO,UAAU,OAAO,SAAS;AAC/C,WAAO,SAAS,EAAE,QAAQ,EAAE,MAAM,OAAO,MAAM,MAAM,GAAG,SAAS,SAAS,CAAC;AAAA,EAC7E;AACA,QAAM,SAAS,aAAa,oBAAoB,aAAa,cAAc,KAAK,GAAG,KAAK;AACxF,SAAO,SAAS,eAAe,OAAO,OAAO,MAAM;AACrD;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,WAAS,OAAO,EAAE,MAAM;AACxB,SAAO,EAAE,QAAQ,EAAE,QAAQ,OAAO,EAAE,OAAO,OAAO,EAAE,MAAM;AAC5D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,QAAQ,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE,mBAAmB;AAClE,QAAM,UAAU,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAChD,SAAO,SAAS,eAAe,MAAM,OAAO,MAAM,OAAO,OAAO;AAClE;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,QAAQ,OAAO,WAAW,OAAO;AACjD,MAAI,UAAU,SAAS;AACrB,UAAM,eAAe,WAAW,KAAK,MAAM;AAC3C,WAAO,CAAC,OAAO,SAAS,YAAY;AAAA,EACtC;AACA,MAAI,UAAU,QAAQ;AACpB,UAAM,OAAO,aAAa,aAAa,CAAC,CAAC,GAAG,SAAS;AACrD,UAAM,CAAC,YAAY,WAAW,IAAI,6BAA6B,CAAC,GAAG,MAAM,MAAM,IAAI,IAAI,CAAC,EAAE,OAAO,CAAC,GAAG,QAAQ,MAAM,MAAM;AACzH,WAAO,CAAC,aAAa,QAAQ,UAAU;AAAA,EACzC;AACA,QAAM,IAAI,MAAM,iCAAiC,gBAAgB,OAAO;AAC1E;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,MAAI,UAAU,aAAa;AACzB,QAAI,EAAE,UAAU,aAAa;AAC3B,aAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,IACvD;AACA,UAAM,kBAAkB,OAAO,UAAU,EAAE,OAAO,EAAE,KAAK;AACzD,UAAM,SAAS,MAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AACtF,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,QAAQ,MAAM,gBAAgB,GAAG,SAAS,SAAS,CAAC;AAC9F,aAAS,8BAA8B,eAAe;AACtD,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,MAAI,EAAE,UAAU,aAAa;AAC3B,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,SAAS,MAAM,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AACrF,aAAS,8BAA8B,QAAQ;AAC/C,WAAO;AAAA,EACT;AACA,MAAI,CAAC,aAAa,gBAAgB,EAAE,OAAO,KAAK,GAAG;AACjD,UAAM,SAAS,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAC7D,WAAO,EAAE,QAAQ,OAAO,QAAQ,OAAO,OAAO,OAAO,MAAM;AAAA,EAC7D;AACA,QAAM,SAAS,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC3C,QAAM,CAAC,aAAa,YAAY,UAAU,IAAI,SAAS,QAAQ,EAAE,OAAO,EAAE,OAAO,KAAK;AACtF,SAAO,SAAS,eAAe,aAAa,YAAY,UAAU;AACpE;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,iBAAiB,MAAM,YAAY,aAAa,OAAO;AAC9D,MAAI,eAAe,MAAM;AACvB,WAAO,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AACxC,YAAM,EAAE,GAAG,EAAE,IAAI;AACjB,YAAM,aAAa;AACnB,uBAAiB,CAAC,GAAG,CAAC,GAAG,IAAI;AAC7B,YAAM,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,YAAM,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,YAAM,eAAe,EAAE,UAAU,WAAW,qBAAqB,uBAAuB,KAAK,IAAI;AACjG,YAAM,eAAe,EAAE,UAAU,WAAW,qBAAqB,uBAAuB,KAAK,IAAI;AACjG,YAAM,SAAS,SAAS,EAAE;AAC1B,YAAM,CAAC,YAAY,WAAW,IAAI,WAAW,EAAE,OAAO,EAAE,OAAO,cAAc,cAAc,MAAM;AACjG,aAAO,WAAW,eAAe,aAAa,QAAQ,UAAU;AAAA,IAClE;AAAA,EACF;AACA,SAAO,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AACxC,UAAM,EAAE,GAAG,EAAE,IAAI;AACjB,UAAM,aAAa;AACnB,QAAI,EAAE,UAAU,eAAe,EAAE,UAAU,aAAa;AACtD,YAAM,YAAY,MAAM,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,YAAY,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAChG,YAAM,gBAAgB,WAAW,KAAK,IAAI,UAAU,MAAM;AAC1D,YAAM,QAAQ,cAAc,mBAAmB;AAC/C,YAAM,QAAQ,cAAc,mBAAmB;AAC/C,YAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,YAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,YAAM,YAAY,MAAM,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,YAAY,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAChG,YAAM,gBAAgB,WAAW,KAAK,IAAI,UAAU,MAAM;AAC1D,YAAM,QAAQ,cAAc,mBAAmB;AAC/C,YAAM,QAAQ,cAAc,mBAAmB;AAC/C,YAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,YAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,YAAM,CAAC,gBAAgB,gBAAgB,WAAW,IAAI,YAAY,EAAE,OAAO,EAAE,OAAO,WAAW,WAAW,WAAW,SAAS;AAC9H,YAAM,aAAa,WAAW,eAAe,aAAa,WAAW,cAAc;AACnF,YAAM,aAAa,WAAW,eAAe,aAAa,WAAW,cAAc;AACnF,YAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,YAAY,MAAM,WAAW,GAAG,SAAS,WAAW,CAAC;AAC/F,iBAAW,8BAA8B,SAAS;AAClD,iBAAW,8BAA8B,SAAS;AAClD,iBAAW,8BAA8B,UAAU;AACnD,iBAAW,8BAA8B,UAAU;AACnD,aAAO;AAAA,IACT,OAAO;AACL,YAAM,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,YAAM,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,YAAM,SAAS,SAAS,EAAE;AAC1B,YAAM,CAAC,YAAY,WAAW,IAAI,WAAW,EAAE,OAAO,EAAE,OAAO,OAAO,OAAO,MAAM;AACnF,aAAO,WAAW,eAAe,aAAa,QAAQ,UAAU;AAAA,IAClE;AAAA,EACF;AACF;AACA,SAAS,8BAA8B,KAAK;AAC1C,SAAO,CAAC,QAAQ,QAAQ,WAAW,WAAW,WAAW,cAAc;AACrE,UAAM,cAAc,qBAAqB,2BAA2B,QAAQ,MAAM;AAClF,UAAM,aAAa,aAAa,cAAc,WAAW;AACzD,UAAM,aAAa,YAAY;AAC/B,UAAM,gBAAgB,aAAa,eAAe,WAAW;AAC7D,UAAM,iBAAiB,aAAa,uBAAuB,WAAW,UAAU;AAChF,UAAM,iBAAiB,aAAa,uBAAuB,WAAW,UAAU;AAChF,UAAM,iBAAiB,qBAAqB,iBAAiB,QAAQ,WAAW;AAChF,UAAM,iBAAiB,qBAAqB,iBAAiB,QAAQ,WAAW;AAChF,UAAM,QAAQ,qBAAqB,uBAAuB,WAAW,SAAS;AAC9E,UAAM,QAAQ,qBAAqB,uBAAuB,WAAW,SAAS;AAC9E,UAAM,QAAQ,OAAO;AACrB,UAAM,WAAW,aAAa,eAAe,MAAM;AACnD,UAAM,QAAQ,OAAO;AACrB,UAAM,WAAW,aAAa,eAAe,MAAM;AACnD,QAAI,eAAe,SAAS,eAAe,WAAW,GAAG;AACvD,eAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,cAAM,OAAO,IAAI,MAAM;AACvB,cAAM,OAAO,IAAI,MAAM;AACvB,cAAM,SAAS,IAAI,MAAM,OAAO,IAAI,MAAM,OAAO,IAAI,IAAI,MAAM,OAAO,IAAI,MAAM,OAAO,IAAI,EAAE;AAC7F,uBAAe,KAAK,OAAO;AAC3B,uBAAe,KAAK,OAAO;AAAA,MAC7B;AAAA,IACF,OAAO;AACL,eAAS,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK;AAC9C,cAAM,MAAM,aAAa,WAAW,GAAG,YAAY,aAAa;AAChE,cAAM,OAAO,IAAI,MAAM,CAAC,KAAK;AAC7B,uBAAe,QAAQ,CAAC,MAAM,KAAK,KAAK,CAAC;AACzC,cAAM,SAAS,aAAa,WAAW,MAAM,OAAO,QAAQ;AAC5D,cAAM,OAAO,IAAI,MAAM,CAAC,KAAK;AAC7B,uBAAe,QAAQ,CAAC,MAAM,KAAK,KAAK,CAAC;AACzC,cAAM,SAAS,aAAa,WAAW,MAAM,OAAO,QAAQ;AAC5D,cAAM,WAAW,IAAI,MAAM,SAAS,IAAI,MAAM,SAAS,IAAI,IAAI,MAAM,SAAS,IAAI,MAAM,SAAS,IAAI,EAAE;AACvG,uBAAe,KAAK,SAAS;AAC7B,uBAAe,KAAK,SAAS;AAAA,MAC/B;AAAA,IACF;AACA,WAAO,CAAC,gBAAgB,gBAAgB,WAAW;AAAA,EACrD;AACF;AAGA,IAAI,UAAU,6BAA6B,CAAC,GAAG,MAAM,IAAI,CAAC;AAC1D,IAAI,iBAAiB,8BAA8B,CAAC,OAAO,OAAO,OAAO,UAAU;AACjF,SAAO,EAAE,MAAM,QAAQ,OAAO,MAAM,QAAQ,MAAM;AACpD,CAAC;AACD,IAAI,OAAO,iBAAiB,KAAK,SAAS,cAAc;AACxD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,OAAO,aAAa,cAAc,cAAc,MAAM;AAC1E,QAAM,cAAc,aAAa,cAAc,YAAY;AAC3D,QAAM,UAAU,aAAa,oBAAoB,MAAM,YAAY;AACnE,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,UAAM,QAAQ,MAAM;AACpB,QAAI,QAAQ,GAAG;AACb,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AACA,QAAI,SAAS,MAAM;AACjB;AAAA,IACF;AACA,QAAI,cAAc,GAAG;AACnB,cAAQ,UAAU,YAAY;AAAA,IAChC,OAAO;AACL,cAAQ,UAAU;AAAA,IACpB;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,mBAAmB,MAAM,YAAY,MAAM,eAAe,OAAO;AACxE,QAAM,UAAU,KAAK,MAAM;AAC3B,QAAM,UAAU,KAAK,MAAM;AAC3B,QAAM,SAAS,OAAO,CAAC,SAAS,IAAI,GAAG,WAAW,KAAK;AACvD,WAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,aAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,YAAM,QAAQ,KAAK,IAAI,GAAG,CAAC;AAC3B,UAAI,QAAQ,GAAG;AACb,cAAM,IAAI,MAAM,+BAA+B;AAAA,MACjD;AACA,UAAI,SAAS,MAAM;AACjB;AAAA,MACF;AACA,UAAI,cAAc;AAChB,eAAO,IAAI,GAAG,GAAG,KAAK;AAAA,MACxB,OAAO;AACL,YAAI,WAAW,OAAO,GAAG;AACvB,iBAAO,IAAI,OAAO,IAAI,GAAG,KAAK,IAAI,WAAW,IAAI,GAAG,CAAC,GAAG,GAAG,KAAK;AAAA,QAClE,OAAO;AACL,iBAAO,IAAI,OAAO,IAAI,GAAG,KAAK,IAAI,GAAG,GAAG,KAAK;AAAA,QAC/C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AAGA,SAAS,sBAAsB,KAAK;AAClC,SAAO,CAAC,QAAQ,OAAO,UAAU;AAC/B,UAAM,YAAY,aAAa,uBAAuB,OAAO,OAAO,MAAM;AAC1E,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,gBAAU,KAAK,IAAI,OAAO,IAAI,KAAK;AAAA,IACrC;AACA,WAAO;AAAA,EACT;AACF;AAGA,SAAS,gBAAgB,MAAM,KAAK,OAAO;AACzC,SAAO,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AAC/C,UAAM,EAAE,EAAE,IAAI;AACd,qBAAiB,GAAG,IAAI;AACxB,QAAI,EAAE,UAAU,YAAY,UAAU,UAAU;AAC9C,YAAM,IAAI,MAAM,sDAAsD;AAAA,IACxE;AACA,UAAM,aAAa;AACnB,UAAM,SAAS,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,UAAM,SAAS,SAAS,EAAE;AAC1B,UAAM,YAAY,aAAa,kBAAkB,QAAQ,KAAK;AAC9D,aAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,gBAAU,KAAK,IAAI,OAAO,IAAI,KAAK;AAAA,IACrC;AACA,WAAO,WAAW,eAAe,EAAE,OAAO,QAAQ,SAAS;AAAA,EAC7D;AACF;AACA,SAAS,wBAAwB,MAAM,WAAW,OAAO;AACvD,SAAO,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AAC/C,UAAM,EAAE,EAAE,IAAI;AACd,qBAAiB,GAAG,IAAI;AACxB,QAAI,EAAE,UAAU,YAAY,UAAU,UAAU;AAC9C,YAAM,IAAI,MAAM,sDAAsD;AAAA,IACxE;AACA,UAAM,aAAa;AACnB,UAAM,SAAS,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,SAAS,SAAS,EAAE;AAC1B,UAAM,YAAY,UAAU,QAAQ,QAAQ,KAAK;AACjD,WAAO,WAAW,eAAe,EAAE,OAAO,QAAQ,SAAS;AAAA,EAC7D;AACF;AAGA,IAAI,WAAW,sBAAsB,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AAC1D,IAAI,QAAQ,wBAAwB,MAAM,QAAQ;AAClD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,QAAQ,UAAU,OAAO,cAAc;AACzD,QAAM,UAAU,aAAa,kBAAkB,OAAO,aAAa,cAAc,QAAQ,CAAC;AAC1F,MAAI,gBAAgB,UAAU,UAAU;AACtC,QAAI,SAAS;AACb,WAAO,QAAQ,CAAC,WAAW;AACzB,YAAM,OAAO,aAAa,cAAc,OAAO,KAAK;AACpD,cAAQ,IAAI,OAAO,MAAM,MAAM;AAC/B,gBAAU;AAAA,IACZ,CAAC;AAAA,EACH,OAAO;AACL,QAAI,YAAY;AAChB,WAAO,QAAQ,CAAC,WAAW;AACzB,YAAM,cAAc,UAAU,WAAW,qBAAqB,uBAAuB,OAAO,IAAI,IAAI,OAAO;AAC3G,UAAI,OAAO;AACX,eAAS,MAAM,GAAG,MAAM,OAAO,MAAM,IAAI,EAAE,KAAK;AAC9C,cAAM,SAAS,MAAM,SAAS,KAAK;AACnC,iBAAS,MAAM,GAAG,MAAM,OAAO,MAAM,IAAI,EAAE,KAAK;AAC9C,kBAAQ,SAAS,OAAO,YAAY;AAAA,QACtC;AAAA,MACF;AACA,mBAAa,OAAO,MAAM;AAAA,IAC5B,CAAC;AAAA,EACH;AACA,SAAO;AACT;AAGA,IAAI,YAAY,6BAA6B,CAAC,GAAG,MAAM,MAAM,IAAI,IAAI,CAAC;AACtE,IAAI,SAAS,iBAAiB,OAAO,WAAW,MAAM,MAAM;AAC5D,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU,sBAAsB,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;AACxD,IAAI,OAAO,wBAAwB,KAAK,SAAS,SAAS;AAC1D,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY,sBAAsB,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC5D,IAAI,SAAS,wBAAwB,OAAO,SAAS;AACrD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY,sBAAsB,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC5D,IAAI,SAAS,wBAAwB,OAAO,SAAS;AACrD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,aAAa,WAAW,OAAO,WAAW,WAAW,WAAW,SAAS,aAAa,YAAY;AACtH,QAAM,SAAS,OAAO,CAAC,WAAW,SAAS,GAAG,KAAK;AACnD,WAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,UAAM,QAAQ,CAAC;AACf,QAAI,eAAe;AACnB,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,YAAM,MAAM,YAAY,IAAI,YAAY;AACxC,sBAAgB,MAAM,QAAQ;AAC9B,YAAM,KAAK,GAAG;AAAA,IAChB;AACA,QAAI,eAAe,KAAK,gBAAgB,aAAa,WAAW;AAC9D,YAAM,IAAI,MAAM,oBAAoB,6BAA6B,aAAa;AAAA,IAChF;AACA,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,aAAO,OAAO,IAAI,YAAY,KAAK,UAAU,IAAI,GAAG,UAAU,WAAW,eAAe,YAAY,CAAC,CAAC;AAAA,IACxG;AAAA,EACF;AACA,SAAO;AACT;AAGA,SAAS,aAAa,MAAM,YAAY,oBAAoB;AAC1D,QAAM,SAAS,OAAO,oBAAoB,KAAK,KAAK;AACpD,WAAS,IAAI,GAAG,IAAI,OAAO,MAAM,EAAE,GAAG;AACpC,UAAM,SAAS,OAAO,WAAW,CAAC;AAClC,UAAM,cAAc,OAAO,MAAM;AACjC,UAAM,WAAW,YAAY;AAC7B,UAAM,aAAa,YAAY;AAC/B,UAAM,eAAe,WAAW,WAAW,CAAC,UAAU,UAAU,CAAC;AACjE,gBAAY,KAAK,WAAW,OAAO;AACnC,UAAM,gBAAgB,KAAK,WAAW,WAAW;AACjD,QAAI,KAAK,iBAAiB,gBAAgB,KAAK,OAAO,QAAQ;AAC5D,aAAO,OAAO,KAAK,KAAK,OAAO;AAAA,IACjC;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,cAAc,6BAA6B,CAAC,GAAG,MAAM,IAAI,IAAI,IAAI,CAAC;AACtE,IAAI,WAAW,iBAAiB,SAAS,aAAa,MAAM,MAAM;AAClE,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,6BAA6B,CAAC,GAAG,MAAM,KAAK,IAAI,IAAI,CAAC;AAC5E,IAAI,gBAAgB,iBAAiB,cAAc,kBAAkB,MAAM,MAAM;AACjF,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,WAAW,6BAA6B,CAAC,GAAG,MAAM,IAAI,IAAI,IAAI,CAAC;AACnE,IAAI,QAAQ,iBAAiB,MAAM,UAAU,MAAM,MAAM;AACzD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,6BAA6B,CAAC,GAAG,MAAM,KAAK,IAAI,IAAI,CAAC;AACzE,IAAI,aAAa,iBAAiB,WAAW,eAAe,MAAM,MAAM;AACxE,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,OAAO,MAAM,KAAK;AACtC,QAAM,SAAS,OAAO,UAAU,MAAM;AACtC,QAAM,SAAS,aAAa,oBAAoB,KAAK,SAAS;AAC9D,SAAO,KAAK;AACZ,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,WAAO,KAAK,OAAO,IAAI,KAAK;AAAA,EAC9B;AACA,SAAO;AACT;AAGA,IAAI,UAAU,sBAAsB,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;AACxD,IAAI,OAAO,wBAAwB,KAAK,OAAO;AAC/C,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,OAAO,YAAY,UAAU,OAAO;AACnD,QAAM,OAAO,aAAa,uBAAuB,OAAO,aAAa,cAAc,QAAQ,CAAC;AAC5F,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO,MAAM;AACjB,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,UAAI,OAAO,MAAM,KAAK,KAAK,QAAQ,MAAM;AACvC,eAAO;AAAA,MACT;AAAA,IACF;AACA,SAAK,KAAK;AAAA,EACZ;AACA,SAAO;AACT;AAGA,IAAI,cAAc,6BAA6B,CAAC,QAAQ,WAAW,KAAK,IAAI,QAAQ,MAAM,CAAC;AAC3F,IAAI,WAAW,iBAAiB,SAAS,WAAW;AACpD,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,6BAA6B,CAAC,QAAQ,WAAW,KAAK,IAAI,QAAQ,MAAM,CAAC;AAC3F,IAAI,WAAW,iBAAiB,SAAS,WAAW;AACpD,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe,6BAA6B,CAAC,QAAQ,WAAW,SAAS,MAAM;AACnF,IAAI,sBAAsB,8BAA8B,CAAC,OAAO,OAAO,OAAO,UAAU;AACtF,SAAO;AAAA,IACL,MAAM,QAAQ,QAAQ,QAAQ;AAAA,IAC9B,MAAM,QAAQ,QAAQ,QAAQ;AAAA,EAChC;AACF,CAAC;AACD,IAAI,YAAY,iBAAiB,UAAU,cAAc,mBAAmB;AAC5E,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,OAAO,QAAQ,QAAQ;AACtC,QAAM,WAAW,aAAa,kBAAkB,IAAI,MAAM;AAC1D,SAAO,aAAa,CAAC,GAAG,QAAQ,UAAU,OAAO,MAAM;AACzD;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,mBAAiB,GAAG,KAAK;AACzB,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,CAAC,KAAK,QAAQ,IAAI,QAAQ,OAAO,EAAE,OAAO,EAAE,KAAK;AACvD,SAAO,SAAS,eAAe,UAAU,EAAE,OAAO,GAAG;AACvD;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe,6BAA6B,CAAC,GAAG,MAAM,MAAM,IAAI,IAAI,CAAC;AACzE,IAAI,YAAY,iBAAiB,UAAU,cAAc,MAAM,MAAM;AACrE,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,OAAO,QAAQ,OAAO,MAAM,UAAU;AAC3D,QAAM,QAAQ,OAAO;AACrB,QAAM,QAAQ,aAAa,cAAc,MAAM;AAC/C,QAAM,WAAW,aAAa,eAAe,MAAM;AACnD,QAAM,aAAa,aAAa,eAAe,QAAQ;AACvD,QAAM,SAAS,aAAa,uBAAuB,OAAO,aAAa,cAAc,QAAQ,CAAC;AAC9F,WAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,UAAM,MAAM,aAAa,WAAW,GAAG,OAAO,QAAQ;AACtD,UAAM,SAAS,IAAI,MAAM,IAAI,MAAM;AACnC,aAAS,KAAK,GAAG,KAAK,OAAO,QAAQ,MAAM;AACzC,aAAO,MAAM,IAAI,KAAK;AAAA,IACxB;AACA,UAAM,WAAW,aAAa,WAAW,QAAQ,OAAO,UAAU;AAClE,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,SAAO;AACT;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,mBAAiB,GAAG,WAAW;AAC/B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,IAAI,MAAM,KAAK;AAChC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAS,KAAK,EAAE,MAAM,KAAK;AAAA,EAC7B;AACA,QAAM,SAAS,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC3C,QAAM,SAAS,cAAc,QAAQ,EAAE,OAAO,EAAE,OAAO,MAAM,QAAQ;AACrE,QAAM,SAAS,SAAS,MAAM,QAAQ,UAAU,EAAE,KAAK;AACvD,SAAO,EAAE,QAAQ,OAAO,UAAU,OAAO,EAAE,MAAM;AACnD;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,QAAQ,QAAQ,OAAO,eAAe;AACtD,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,QAAQ,aAAa;AACpG,QAAM,WAAW,WAAW,QAAQ,OAAO;AAC3C,QAAM,UAAU,aAAa,oBAAoB,aAAa,cAAc,QAAQ,GAAG,QAAQ;AAC/F,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAM,SAAS,IAAI;AACnB,QAAI,QAAQ;AACZ,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,eAAS,MAAM,SAAS;AAAA,IAC1B;AACA,YAAQ,KAAK;AAAA,EACf;AACA,SAAO,EAAE,SAAS,UAAU,SAAS;AACvC;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,mBAAiB,GAAG,MAAM;AAC1B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACtD,QAAM,cAAc,qBAAqB,mBAAmB,MAAM,KAAK;AACvE,MAAI,gBAAgB;AACpB,MAAI,YAAY;AAChB,QAAM,0BAA0B,CAAC;AACjC,MAAI,eAAe,MAAM;AACvB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AACzF,4BAAwB,KAAK,SAAS;AACtC,oBAAgB,qBAAqB,iBAAiB,cAAc,QAAQ,KAAK;AAAA,EACnF;AACA,QAAM,QAAQ,SAAS,KAAK,IAAI,UAAU,MAAM,EAAE;AAClD,QAAM,EAAE,SAAS,UAAU,SAAS,IAAI,SAAS,UAAU,OAAO,UAAU,OAAO,OAAO,aAAa;AACvG,MAAI,cAAc;AAClB,MAAI,UAAU;AACZ,kBAAc,qBAAqB,qBAAqB,UAAU,IAAI;AAAA,EACxE;AACA,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO,SAAS,eAAe,aAAa,UAAU,OAAO;AAC/D;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,gBAAgB,SAAS,cAAc,WAAW;AACzD,UAAQ,QAAQ,CAAC,OAAO,MAAM;AAC5B,QAAI,QAAQ,KAAK,SAAS,WAAW;AACnC,YAAM,YAAY,aAAa,WAAW,GAAG,aAAa,QAAQ,aAAa,eAAe,YAAY,CAAC,EAAE,KAAK,GAAG;AACrH,YAAM,IAAI,MAAM,WAAW,gBAAgB,uBAAuB,YAAY;AAAA,IAChF;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,oBAAoB,sBAAsB;AAChE,WAAS,MAAM,GAAG,MAAM,mBAAmB,QAAQ,EAAE,KAAK;AACxD,UAAM,SAAS,mBAAmB;AAClC,UAAM,YAAY,QAAQ,mBAAmB,SAAS,IAAI,uBAAuB,mBAAmB,MAAM,GAAG;AAC7G,QAAI,OAAO,WAAW,GAAG;AACvB,YAAM,IAAI,MAAM,gCAAgC;AAAA,IAClD;AACA,QAAI,OAAO,KAAK,GAAG;AACjB,YAAM,IAAI,MAAM,oCAAoC;AAAA,IACtD;AACA,QAAI,OAAO,OAAO,SAAS,KAAK,WAAW;AACzC,YAAM,IAAI,MAAM,0CAA0C;AAAA,IAC5D;AACA,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,UAAI,OAAO,IAAI,KAAK,OAAO,IAAI;AAC7B,cAAM,IAAI,MAAM,iDAAiD;AAAA,MACnE;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,WAAW,SAAS,cAAc,oBAAoB,sBAAsB;AACnF,QAAM,cAAc,CAAC;AACrB,MAAI,YAAY;AAChB,QAAM,YAAY,aAAa,SAAS,IAAI,mBAAmB;AAC/D,QAAM,YAAY,IAAI,MAAM,SAAS,EAAE,KAAK,IAAI,EAAE,IAAI,MAAM,CAAC,CAAC,CAAC;AAC/D,iBAAe,oBAAoB,oBAAoB;AACvD,MAAI,QAAQ;AACZ,WAAS,MAAM,GAAG,MAAM,aAAa,SAAS,GAAG,EAAE,KAAK;AACtD,aAAS,aAAa;AACtB,UAAM,YAAY,aAAa,MAAM;AACrC,aAAS,IAAI,GAAG,IAAI,QAAQ,GAAG,EAAE,GAAG;AAClC,gBAAU,KAAK,KAAK,IAAI,SAAS;AAAA,IACnC;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,QAAI,QAAQ,QAAQ;AACpB,QAAI,QAAQ,QAAQ,KAAK;AACzB,aAAS,MAAM,GAAG,MAAM,mBAAmB,QAAQ,EAAE,KAAK;AACxD,YAAM,SAAS,mBAAmB;AAClC,YAAM,SAAS,MAAM,aAAa,SAAS;AAC3C,UAAI,UAAU,GAAG;AACf,cAAM,kBAAkB,UAAU;AAClC,cAAM,QAAQ,gBAAgB,gBAAgB,SAAS,KAAK,OAAO;AACnE,iBAAS,IAAI,OAAO,IAAI,OAAO,EAAE,GAAG;AAClC,oBAAU,QAAQ,KAAK,OAAO,IAAI,KAAK,KAAK;AAAA,QAC9C;AAAA,MACF;AACA,cAAQ,OAAO;AACf,cAAQ,OAAO;AAAA,IACjB;AACA,QAAI,UAAU,OAAO;AACnB,kBAAY,KAAK,CAAC,OAAO,KAAK,CAAC;AAC/B,mBAAa,QAAQ;AAAA,IACvB;AAAA,EACF;AACA,SAAO,EAAE,WAAW,aAAa,UAAU;AAC7C;AACA,SAAS,UAAU,WAAW;AAC5B,QAAM,YAAY,CAAC;AACnB,WAAS,IAAI,GAAG,IAAI,UAAU,QAAQ,EAAE,GAAG;AACzC,UAAM,YAAY,UAAU,GAAG;AAC/B,UAAM,SAAS,aAAa,kBAAkB,SAAS,SAAS;AAChE,cAAU,KAAK,MAAM;AACrB,cAAU,GAAG,QAAQ,CAAC,OAAO,MAAM,OAAO,KAAK,KAAK;AAAA,EACtD;AACA,SAAO;AACT;AACA,SAAS,qBAAqB,MAAM,YAAY;AAC9C,QAAM,UAAU,KAAK,MAAM,GAAG,UAAU;AACxC,SAAO,QAAQ,SAAS,YAAY;AAClC,YAAQ,KAAK,CAAC;AAAA,EAChB;AACA,WAAS,QAAQ,YAAY,QAAQ,KAAK,QAAQ,SAAS;AACzD,YAAQ,aAAa,MAAM,KAAK;AAAA,EAClC;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,mBAAmB,wBAAwB,aAAa,WAAW,QAAQ,aAAa;AAChH,QAAM,SAAS,qBAAqB,wBAAwB,CAAC,EAAE;AAC/D,QAAM,UAAU,qBAAqB,aAAa,CAAC,EAAE;AACrD,MAAI,SAAS;AACb,aAAW,UAAU,aAAa;AAChC,aAAS,IAAI,OAAO,IAAI,IAAI,OAAO,IAAI,EAAE,GAAG;AAC1C,eAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,eAAO,SAAS,UAAU,KAAK,kBAAkB,IAAI,SAAS;AAAA,MAChE;AACA,QAAE;AAAA,IACJ;AAAA,EACF;AACF;AACA,SAAS,UAAU,mBAAmB,wBAAwB,wBAAwB,aAAa,WAAW;AAC5G,QAAM,cAAc,uBAAuB,MAAM;AACjD,cAAY,KAAK;AACjB,QAAM,YAAY,aAAa,kBAAkB,wBAAwB,aAAa,cAAc,WAAW,CAAC;AAChH,QAAM,cAAc,kBAAkB;AACtC,QAAM,YAAY,gBAAgB,IAAI,IAAI,cAAc,uBAAuB;AAC/E,mBAAiB,mBAAmB,wBAAwB,aAAa,WAAW,WAAW,WAAW;AAC1G,SAAO,CAAC,WAAW,WAAW;AAChC;AACA,SAAS,iBAAiB,oBAAoB,0BAA0B,mBAAmB,wBAAwB,wBAAwB,SAAS,cAAc,kBAAkB;AAClL,MAAI,mBAAmB,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,MAAI,yBAAyB,GAAG,WAAW,GAAG;AAC5C,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACA,QAAM,YAAY,yBAAyB,GAAG,KAAK;AACnD,kBAAgB,SAAS,cAAc,SAAS;AAChD,MAAI,uBAAuB,WAAW,GAAG;AACvC,UAAM,IAAI,MAAM,6BAA6B;AAAA,EAC/C;AACA,QAAM,uBAAuB,uBAAuB;AACpD,QAAM,EAAE,WAAW,aAAa,UAAU,IAAI,WAAW,SAAS,cAAc,oBAAoB,oBAAoB;AACxH,QAAM,qBAAqB,UAAU,SAAS;AAC9C,QAAM,oBAAoB,UAAU,mBAAmB,wBAAwB,wBAAwB,aAAa,SAAS;AAC7H,SAAO,CAAC,oBAAoB,kBAAkB,IAAI,kBAAkB,EAAE;AACxE;AAGA,IAAI,aAAa;AACjB,SAAS,gBAAgB,QAAQ,aAAa,aAAa,QAAQ,aAAa,QAAQ,aAAa;AACnG,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACA,MAAI,YAAY,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACA,QAAM,kBAAkB,YAAY,WAAW;AAC/C,QAAM,kBAAkB,YAAY,WAAW;AAC/C,QAAM,kBAAkB,YAAY,WAAW;AAC/C,QAAM,UAAU,CAAC;AACjB,MAAI,CAAC,iBAAiB;AACpB,YAAQ,KAAK,YAAY,EAAE;AAAA,EAC7B;AACA,MAAI,CAAC,iBAAiB;AACpB,YAAQ,KAAK,YAAY,EAAE;AAAA,EAC7B;AACA,MAAI,CAAC,iBAAiB;AACpB,YAAQ,KAAK,YAAY,EAAE;AAAA,EAC7B;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,QAAI,QAAQ,OAAO,QAAQ,IAAI,IAAI;AACjC,YAAM,IAAI,MAAM,qDAAqD;AAAA,IACvE;AAAA,EACF;AACA,QAAM,QAAQ,QAAQ,WAAW,IAAI,IAAI,QAAQ;AACjD,QAAM,iBAAiB,aAAa,kBAAkB,SAAS,QAAQ,CAAC;AACxE,iBAAe,KAAK;AACpB,WAAS,MAAM,GAAG,MAAM,OAAO,EAAE,KAAK;AACpC,UAAM,QAAQ,kBAAkB,OAAO,KAAK,OAAO;AACnD,UAAM,QAAQ,kBAAkB,OAAO,KAAK,OAAO;AACnD,UAAM,QAAQ,kBAAkB,OAAO,KAAK,OAAO;AACnD,QAAI,UAAU,GAAG;AACf,YAAM,IAAI,MAAM,qBAAqB;AAAA,IACvC;AACA,QAAI;AACJ,QAAI,QAAQ,KAAK,QAAQ,SAAS,QAAQ,KAAK,QAAQ,OAAO;AAC5D,aAAO;AAAA,IACT,OAAO;AACL,aAAO,KAAK,KAAK,KAAK,KAAK,QAAQ,SAAS,KAAK,CAAC;AAClD,UAAI,OAAO,YAAY;AACrB,cAAM,IAAI,MAAM,yCAAyC,YAAY;AAAA,MACvE;AAAA,IACF;AACA,mBAAe,MAAM,KAAK,eAAe,OAAO;AAAA,EAClD;AACA,QAAM,QAAQ,eAAe;AAC7B,QAAM,gBAAgB,aAAa,kBAAkB,aAAa,KAAK;AACvE,MAAI,aAAa;AACjB,WAAS,MAAM,GAAG,MAAM,OAAO,EAAE,KAAK;AACpC,UAAM,UAAU,eAAe,MAAM,KAAK,eAAe;AACzD,QAAI,QAAQ,kBAAkB,OAAO,KAAK,OAAO;AACjD,UAAM,QAAQ,kBAAkB,OAAO,KAAK,OAAO;AACnD,aAAS,IAAI,GAAG,IAAI,SAAS,EAAE,GAAG;AAChC,oBAAc,gBAAgB;AAC9B,eAAS;AAAA,IACX;AAAA,EACF;AACA,SAAO,CAAC,gBAAgB,aAAa;AACvC;AAGA,IAAI,oBAAoB,qBAAqB;AAC7C,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,OAAO,YAAY,QAAQ,aAAa,aAAa,cAAc,mBAAmB,oBAAoB,0BAA0B,yBAAyB;AACvK,SAAK,QAAQ;AACb,SAAK,aAAa;AAClB,SAAK,SAAS;AACd,SAAK,cAAc;AACnB,SAAK,cAAc;AACnB,SAAK,eAAe;AACpB,SAAK,oBAAoB;AACzB,SAAK,qBAAqB;AAC1B,SAAK,2BAA2B;AAChC,SAAK,oBAAoB,qBAAqB,2BAA2B,uBAAuB;AAChG,SAAK,aAAa,qBAAqB,cAAc,KAAK,iBAAiB;AAAA,EAC7E;AAAA,EACA,+BAA+B,WAAW;AACxC,QAAI,KAAK,kBAAkB,OAAO,kBAAkB,gBAAgB;AAClE,aAAO,KAAK,kBAAkB,YAAY;AAAA,IAC5C,OAAO;AACL,aAAO,KAAK,kBAAkB;AAAA,IAChC;AAAA,EACF;AAAA,EACA,sBAAsB,WAAW;AAC/B,QAAI,KAAK,kBAAkB,OAAO,kBAAkB,gBAAgB;AAClE,aAAO,KAAK,mBAAmB,YAAY;AAAA,IAC7C,OAAO;AACL,aAAO,KAAK,mBAAmB;AAAA,IACjC;AAAA,EACF;AAAA,EACA,YAAY,WAAW;AACrB,UAAM,qBAAqB,KAAK,sBAAsB,YAAY,CAAC;AACnE,YAAQ,KAAK,+BAA+B,YAAY,CAAC,GAAG;AAAA,MAC1D,KAAK,kBAAkB;AACrB,eAAO,uBAAuB,sBAAsB,kBAAkB;AAAA,MACxE,KAAK,kBAAkB;AACrB,eAAO,uBAAuB,oBAAoB,kBAAkB;AAAA,MACtE;AACE,cAAM,IAAI,MAAM,gCAAgC,kBAAkB,KAAK,+BAA+B,YAAY,CAAC,IAAI;AAAA,IAC3H;AAAA,EACF;AAAA,EACA,OAAO,oBAAoB,UAAU;AACnC,UAAM,eAAe,SAAS;AAC9B,QAAI,iBAAiB,KAAK,iBAAiB,GAAG;AAC5C,aAAO;AAAA,IACT;AACA,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,eAAe,GAAG,EAAE,GAAG;AACzC,YAAM,eAAe,SAAS,IAAI,KAAK,SAAS;AAChD,UAAI,eAAe,UAAU;AAC3B,mBAAW;AAAA,MACb;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,OAAO,sBAAsB,aAAa;AACxC,UAAM,cAAc,YAAY;AAChC,QAAI,gBAAgB,GAAG;AACrB,aAAO;AAAA,IACT;AACA,QAAI,kBAAkB;AACtB,QAAI,uBAAuB,YAAY;AACvC,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,YAAM,QAAQ,YAAY;AAC1B,UAAI,UAAU,sBAAsB;AAClC,+BAAuB;AACvB,mBAAW,KAAK,IAAI,IAAI,iBAAiB,QAAQ;AACjD,0BAAkB;AAAA,MACpB;AAAA,IACF;AACA,WAAO,KAAK,IAAI,cAAc,iBAAiB,QAAQ;AAAA,EACzD;AAAA,EACA,sBAAsB,GAAG,QAAQ,YAAY,MAAM;AACjD,QAAI,OAAO,WAAW,GAAG;AACvB,UAAI,EAAE,OAAO,IAAI;AACf,eAAO,CAAC;AAAA,MACV;AACA,YAAM,IAAI,MAAM,gFAAgF;AAAA,IAClG;AACA,WAAO,UAAU,GAAG,SAAS;AAAA,EAC/B;AAAA,EACA,oBAAoB,UAAU;AAC5B,UAAM,aAAa,KAAK;AACxB,UAAM,oBAAoB,KAAK;AAC/B,yBAAqB,0BAA0B,mBAAmB,UAAU;AAC5E,UAAM,QAAQ,KAAK,sBAAsB,KAAK,OAAO,KAAK,UAAU;AACpE,UAAM,cAAc,qBAAqB,kCAAkC,KAAK,YAAY,OAAO,UAAU;AAC7G,UAAM,SAAS;AACf,QAAI,OAAO,KAAK,GAAG;AACjB,aAAO,KAAK;AAAA,IACd;AACA,aAAS,IAAI,GAAG,KAAK,KAAK,YAAY,EAAE,GAAG;AACzC,UAAI,OAAO,KAAK,GAAG;AACjB,eAAO,KAAK,KAAK,YAAY,CAAC;AAAA,MAChC;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,gCAAgC,gBAAgB,uBAAuB,sBAAsB;AAC3F,UAAM,eAAe,KAAK,IAAI,gBAAgB,oBAAoB;AAClE,UAAM,SAAS,CAAC;AAChB,QAAI,qBAAqB;AACzB,aAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG,sBAAsB,uBAAuB;AAClF,aAAO,KAAK,kBAAkB;AAAA,IAChC;AACA,aAAS,IAAI,cAAc,IAAI,gBAAgB,EAAE,GAAG;AAClD,aAAO,KAAK,EAAE;AAAA,IAChB;AACA,iBAAa,OAAO,OAAO,WAAW,gBAAgB,MAAM,yDAAyD;AACrH,WAAO;AAAA,EACT;AAAA,EACA,6BAA6B,UAAU,mBAAmB,uBAAuB,YAAY;AAC3F,UAAM,eAAe,SAAS;AAC9B,UAAM,SAAS,CAAC;AAChB,aAAS,IAAI,GAAG,IAAI,eAAe,GAAG,EAAE,GAAG;AACzC,YAAM,YAAY,SAAS,IAAI,KAAK,SAAS;AAC7C,UAAI,aAAa,KAAK,IAAI,YAAY,SAAS;AAC/C,UAAI,2BAA2B,kBAAkB;AACjD,UAAI,6BAA6B,IAAI;AACnC,qBAAa;AAAA,MACf;AACA,eAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,eAAO,KAAK,wBAAwB;AACpC,oCAA4B;AAAA,MAC9B;AACA,eAAS,IAAI,GAAG,IAAI,YAAY,YAAY,EAAE,GAAG;AAC/C,eAAO,KAAK,EAAE;AAAA,MAChB;AAAA,IACF;AACA,QAAI,eAAe,KAAK,OAAO,WAAW,SAAS,eAAe,IAAI;AACpE,YAAM,IAAI,MAAM,yBAAyB;AAAA,IAC3C;AACA,WAAO;AAAA,EACT;AAAA,EACA,+BAA+B,aAAa,mBAAmB,uBAAuB,YAAY;AAChG,UAAM,YAAY,YAAY;AAC9B,UAAM,SAAS,CAAC;AAChB,QAAI,cAAc,GAAG;AACnB,aAAO,CAAC;AAAA,IACV;AACA,QAAI,sBAAsB;AAC1B,QAAI,oBAAoB,YAAY;AACpC,QAAI,qBAAqB,kBAAkB,QAAQ;AACjD,YAAM,IAAI,MAAM,yBAAyB,6CAA6C,kBAAkB,QAAQ;AAAA,IAClH;AACA,QAAI,qBAAqB,kBAAkB;AAC3C,WAAO,KAAK,kBAAkB;AAC9B,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,YAAM,iBAAiB,YAAY;AACnC,UAAI,mBAAmB,mBAAmB;AACxC,YAAI,sBAAsB,GAAG;AAC3B,YAAE;AACF,cAAI,sBAAsB,YAAY;AACpC,kCAAsB;AAAA,UACxB,OAAO;AACL,iCAAqB;AAAA,UACvB;AAAA,QACF;AAAA,MACF,OAAO;AACL,8BAAsB;AACtB,4BAAoB;AACpB,YAAI,kBAAkB,kBAAkB,QAAQ;AAC9C,gBAAM,IAAI,MAAM,sBAAsB,yCAAyC,kBAAkB,QAAQ;AAAA,QAC3G;AACA,6BAAqB,kBAAkB;AAAA,MACzC;AACA,aAAO,KAAK,kBAAkB;AAAA,IAChC;AACA,QAAI,OAAO,WAAW,YAAY,QAAQ;AACxC,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,WAAO;AAAA,EACT;AAAA,EACA,qBAAqB,WAAW,mBAAmB,uBAAuB,YAAY;AACpF,UAAM,qBAAqB,KAAK,sBAAsB,SAAS;AAC/D,UAAM,gBAAgB,KAAK,+BAA+B,SAAS;AACnE,YAAQ,eAAe;AAAA,MACrB,KAAK,kBAAkB;AACrB,eAAO,KAAK,+BAA+B,oBAAoB,mBAAmB,uBAAuB,UAAU;AAAA,MACrH,KAAK,kBAAkB;AACrB,YAAI,mBAAmB,SAAS,IAAI,kBAAkB,QAAQ;AAC5D,gBAAM,IAAI,MAAM,mDAAmD,mBAAmB,SAAS,OAAO,kBAAkB,QAAQ;AAAA,QAClI;AACA,eAAO,KAAK,6BAA6B,oBAAoB,mBAAmB,uBAAuB,UAAU;AAAA,MACnH;AACE,cAAM,IAAI,MAAM,+BAA+B,kBAAkB,gBAAgB;AAAA,IACrF;AAAA,EACF;AAAA,EACA,wBAAwB;AACtB,UAAM,uBAAuB,KAAK,mBAAmB;AACrD,QAAI,KAAK,kBAAkB,WAAW,GAAG;AACvC,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AACA,UAAM,qBAAqB,KAAK,kBAAkB;AAClD,YAAQ,oBAAoB;AAAA,MAC1B,KAAK,kBAAkB;AACrB,eAAO,qBAAqB;AAAA,MAC9B,KAAK,kBAAkB;AACrB,cAAM,IAAI,MAAM,gDAAgD;AAAA,MAClE,KAAK,kBAAkB;AACrB,eAAO,KAAK,yBAAyB,GAAG,KAAK;AAAA,MAC/C;AACE,cAAM,IAAI,MAAM,sBAAsB,kBAAkB,qBAAqB;AAAA,IACjF;AAAA,EACF;AAAA,EACA,UAAU;AACR,UAAM,uBAAuB,KAAK,mBAAmB;AACrD,QAAI,qBAAqB,UAAU,GAAG;AACpC,YAAM,IAAI,MAAM,sEAAsE;AAAA,IACxF;AACA,UAAM,iBAAiB,KAAK,sBAAsB;AAClD,UAAM,aAAa,KAAK,oBAAoB,cAAc;AAC1D,UAAM,aAAa,IAAI,MAAM,KAAK,aAAa,CAAC;AAChD,eAAW,WAAW,SAAS,KAAK;AACpC,aAAS,IAAI,WAAW,SAAS,GAAG,KAAK,GAAG,EAAE,GAAG;AAC/C,iBAAW,KAAK,WAAW,IAAI,KAAK,WAAW,IAAI;AAAA,IACrD;AACA,UAAM,cAAc,UAAU,YAAY,KAAK;AAC/C,UAAM,eAAe,aAAa,kBAAkB,KAAK,aAAa,aAAa,cAAc,WAAW,CAAC;AAC7G,UAAM,WAAW,WAAW,KAAK,WAAW;AAC5C,QAAI,WAAW,GAAG;AAChB,UAAI,cAAc,KAAK,gCAAgC,gBAAgB,WAAW,IAAI,WAAW,EAAE;AACnG,eAAS,IAAI,GAAG,KAAK,KAAK,YAAY,EAAE,GAAG;AACzC,cAAM,iBAAiB,KAAK,qBAAqB,IAAI,GAAG,aAAa,WAAW,IAAI,WAAW,EAAE;AACjG,sBAAc;AAAA,MAChB;AACA,WAAK,UAAU,KAAK,YAAY,aAAa,cAAc,WAAW;AAAA,IACxE;AACA,WAAO,CAAC,aAAa,YAAY;AAAA,EACnC;AAAA,EACA,UAAU,YAAY,aAAa,cAAc,aAAa;AAC5D,QAAI,aAAa,WAAW,GAAG;AAC7B;AAAA,IACF;AACA,UAAM,aAAa,KAAK;AACxB,UAAM,aAAa;AACnB,QAAI,eAAe,YAAY,MAAM;AACrC,mBAAe,aAAa,MAAM,aAAa,CAAC;AAChD,UAAM,mBAAmB,aAAa,cAAc,YAAY;AAChE,UAAM,kBAAkB,YAAY;AACpC,QAAI,eAAe,KAAK;AACxB,QAAI,aAAa,WAAW,oBAAoB,aAAa,WAAW,GAAG;AACzE,YAAM,WAAW,KAAK;AACtB,WAAK,MAAM;AACT,cAAM,qBAAqB,QAAQ,cAAc,QAAQ;AACzD,cAAM,eAAe,YAAY,oBAAoB,YAAY;AACjE,uBAAe,aAAa,SAAS;AAAA,MACvC,CAAC;AAAA,IACH;AACA,QAAI,WAAW;AACf,QAAI,WAAW;AACf,QAAI,SAAS;AACb,aAAS,OAAO,GAAG,QAAQ,iBAAiB,EAAE,MAAM;AAClD,UAAI,OAAO,OAAO,kBAAkB,YAAY,QAAQ;AACxD,UAAI,SAAS,QAAQ;AACnB,UAAE;AACF;AAAA,MACF;AACA,UAAI,WAAW,QAAQ;AACrB,cAAM,MAAM,WAAW,SAAS,WAAW,gBAAgB;AAC3D,cAAM,MAAM,WAAW,SAAS,WAAW,gBAAgB;AAC3D,cAAM,SAAS,SAAS,YAAY;AACpC,kBAAU,KAAK,KAAK,KAAK;AAAA,MAC3B;AACA,UAAI,QAAQ,iBAAiB;AAC3B,cAAM,aAAa,aAAa;AAChC,eAAO,KAAK,MAAM,aAAa,gBAAgB;AAAA,MACjD;AACA,UAAI,OAAO,QAAQ;AACjB,YAAI,KAAK,aAAa,WAAW,GAAG;AAClC,qBAAW,SAAS,SAAS,kBAAkB,OAAO,gBAAgB,EAAE,KAAK,KAAK,aAAa,EAAE;AACjG,mBAAS;AAAA,QACX,OAAO;AACL,iBAAO,OAAO,QAAQ;AACpB,kBAAM,MAAM,WAAW,MAAM,SAAS,gBAAgB;AACtD,sBAAU,KAAK,cAAc,gBAAgB;AAC7C,cAAE;AAAA,UACJ;AAAA,QACF;AAAA,MACF;AACA,UAAI,OAAO,GAAG;AACZ,mBAAW,OAAO;AAClB,mBAAW;AAAA,MACb,OAAO;AACL,mBAAW;AACX,mBAAW;AACX,iBAAS,WAAW;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,UAAU,KAAK,KAAK,MAAM;AACjC,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,QAAI,KAAK,IAAI;AAAA,EACf;AACF;AACA,SAAS,UAAU,OAAO,WAAW;AACnC,QAAM,MAAM,CAAC;AACb,WAAS,OAAO,OAAO;AACrB,QAAI,MAAM,GAAG;AACX,UAAI,CAAC,WAAW;AACd,cAAM,IAAI,MAAM,aAAa,kBAAkB;AAAA,MACjD;AACA,UAAI,MAAM,IAAI;AACZ,cAAM,IAAI,MAAM,aAAa,mBAAmB;AAAA,MAClD;AACA,YAAM;AAAA,IACR;AACA,QAAI,KAAK,GAAG;AAAA,EACd;AACA,SAAO;AACT;AACA,SAAS,yBAAyB,OAAO,aAAa,QAAQ,aAAa,aAAa,cAAc,mBAAmB,oBAAoB,0BAA0B,mBAAmB;AACxL,SAAO,IAAI,uBAAuB,OAAO,aAAa,QAAQ,aAAa,aAAa,cAAc,mBAAmB,oBAAoB,0BAA0B,iBAAiB,EAAE,QAAQ;AACpM;AAGA,SAAS,UAAU,OAAO,MAAM,OAAO,OAAO;AAC5C,QAAM,gBAAgB,UAAU;AAChC,QAAM,8BAA8B,QAAQ,QAAQ,QAAQ;AAC5D,QAAM,8BAA8B,OAAO,SAAS,QAAQ;AAC5D,MAAI,iBAAiB,+BAA+B,6BAA6B;AAC/E,WAAO,aAAa,oBAAoB,GAAG,KAAK;AAAA,EAClD;AACA,QAAM,cAAc,KAAK,IAAI,KAAK,MAAM,OAAO,SAAS,KAAK,CAAC;AAC9D,QAAM,SAAS,aAAa,oBAAoB,aAAa,KAAK;AAClE,MAAI,OAAO,SAAS,UAAU,GAAG;AAC/B,YAAQ;AAAA,EACV;AACA,SAAO,KAAK;AACZ,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,WAAO,KAAK,OAAO,IAAI,KAAK;AAAA,EAC9B;AACA,SAAO;AACT;AAGA,IAAI,YAAY,sBAAsB,CAAC,OAAO,IAAI,KAAK,KAAK,EAAE,CAAC;AAC/D,IAAI,SAAS,wBAAwB,OAAO,SAAS;AACrD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,SAAS,SAAS,OAAO,YAAY,WAAW,YAAY,WAAW,SAAS,cAAc,gBAAgB;AACjI,QAAM,eAAe,CAAC,aAAa,WAAW,SAAS;AACvD,QAAM,cAAc,QAAQ;AAC5B,QAAM,cAAc,QAAQ;AAC5B,MAAI,eAAe,GAAG;AACpB,WAAO,OAAO,OAAO,QAAQ,KAAK;AAAA,EACpC;AACA,QAAM,SAAS,OAAO,cAAc,QAAQ,KAAK;AACjD,MAAI,OAAO,iBAAiB,UAAU;AACpC,WAAO,OAAO,KAAK,YAAY;AAAA,EACjC,WAAW,OAAO,iBAAiB,UAAU;AAC3C,WAAO,OAAO,KAAK,YAAY;AAAA,EACjC,WAAW,OAAO,iBAAiB,WAAW;AAC5C,WAAO,OAAO,KAAK,CAAC,YAAY;AAAA,EAClC;AACA,WAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,UAAM,QAAQ,CAAC;AACf,QAAI,eAAe;AACnB,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,YAAM,MAAM,YAAY,IAAI,YAAY;AACxC,YAAM,KAAK,GAAG;AACd,sBAAgB,MAAM,QAAQ;AAAA,IAChC;AACA,QAAI,eAAe,KAAK,gBAAgB,aAAa,WAAW;AAC9D,YAAM,IAAI,MAAM,oBAAoB,6BAA6B,OAAO;AAAA,IAC1E;AACA,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,UAAI,gBAAgB;AAClB,eAAO,OAAO,eAAe,YAAY,MAAM,YAAY,IAAI,YAAY;AAAA,MAC7E,OAAO;AACL,eAAO,OAAO,eAAe,YAAY,KAAK,QAAQ,SAAS,IAAI,YAAY,KAAK,YAAY,IAAI,YAAY;AAAA,MAClH;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,cAAc,sBAAsB,CAAC,OAAO,KAAK,IAAI,KAAK,IAAI,CAAC,EAAE,EAAE;AACvE,IAAI,WAAW,gBAAgB,SAAS,CAAC,OAAO,KAAK,IAAI,KAAK,IAAI,CAAC,EAAE,EAAE;AACvE,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM,OAAO,MAAM,OAAO,OAAO;AAClD,QAAM,cAAc,mBAAmB,iBAAiB,OAAO,OAAO,IAAI;AAC1E,QAAM,SAAS,aAAa,cAAc,IAAI;AAC9C,QAAM,WAAW,aAAa,eAAe,KAAK;AAClD,MAAI,aAAa;AACf,UAAM,aAAa,mBAAmB,kBAAkB,OAAO,QAAQ;AACvE,QAAI,UAAU,UAAU;AACtB,aAAO,KAAK,MAAM,YAAY,aAAa,MAAM;AAAA,IACnD;AACA,WAAO,KAAK,SAAS,YAAY,aAAa,MAAM;AAAA,EACtD;AACA,QAAM,cAAc,UAAU,WAAW,qBAAqB,uBAAuB,IAAI,IAAI;AAC7F,QAAM,QAAQ,OAAO,OAAO,OAAO,WAAW;AAC9C,QAAM,SAAS,OAAO,MAAM,KAAK;AACjC,WAAS,IAAI,GAAG,IAAI,OAAO,MAAM,EAAE,GAAG;AACpC,UAAM,SAAS,OAAO,WAAW,CAAC;AAClC,UAAM,QAAQ,OAAO,IAAI,CAAC,KAAK,MAAM,MAAM,MAAM,EAAE;AACnD,WAAO,IAAI,MAAM,IAAI,GAAG,KAAK,GAAG,GAAG,MAAM;AAAA,EAC3C;AACA,MAAI,UAAU,UAAU;AACtB,WAAO,qBAAqB,uBAAuB,OAAO,MAAM;AAAA,EAClE;AACA,SAAO,OAAO;AAChB;AACA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,OAAO,KAAK,IAAI;AACxB,mBAAiB,GAAG,OAAO;AAC3B,QAAM,CAAC,QAAQ,KAAK,IAAI,mBAAmB,iBAAiB,GAAG,OAAO,IAAI;AAC1E,qBAAmB,kBAAkB,GAAG,QAAQ,KAAK;AACrD,QAAM,OAAO,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AACzC,QAAM,UAAU,UAAU,MAAM,QAAQ,OAAO,EAAE,OAAO,EAAE,KAAK;AAC/D,SAAO,SAAS,eAAe,OAAO,EAAE,OAAO,OAAO;AACxD;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,wBAAwB,SAAS,cAAc,cAAc,QAAQ,aAAa,YAAY,cAAc;AACnH,QAAM,eAAe,aAAa;AAClC,QAAM,YAAY,WAAW;AAC7B,QAAM,oBAAoB,IAAI,MAAM,SAAS;AAC7C,QAAM,kBAAkB,IAAI,MAAM,YAAY;AAC9C,QAAM,OAAO,aAAa;AAC1B,MAAI,cAAc,GAAG;AACnB,QAAI,iBAAiB,GAAG;AACtB,YAAM,IAAI,MAAM,qBAAqB,gDAAgD,YAAY,CAAC;AAAA,IACpG;AACA,UAAM,gBAAgB,aAAa,kBAAkB,cAAc,CAAC;AACpE,UAAM,eAAe,aAAa,kBAAkB,aAAa,CAAC;AAClE,WAAO;AAAA,MACL;AAAA,MACA,CAAC,GAAG,IAAI;AAAA,MACR;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,MAAI,iBAAiB;AACrB,MAAI,iBAAiB;AACrB,QAAM,YAAY,IAAI,MAAM,SAAS,EAAE,KAAK,CAAC;AAC7C,WAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,UAAM,MAAM,QAAQ,IAAI;AACxB,QAAI,MAAM,GAAG;AACX,YAAM,IAAI,MAAM,qBAAqB,gDAAgD,GAAG,GAAG,CAAC;AAAA,IAC9F;AACA,QAAI,OAAO,WAAW;AACpB,YAAM,IAAI,MAAM,qBAAqB,kDAAkD,GAAG,KAAK,SAAS,CAAC;AAAA,IAC3G;AACA,MAAE,UAAU;AACZ,qBAAiB,kBAAkB,OAAO;AAC1C,qBAAiB;AAAA,EACnB;AACA,MAAI,cAAc;AAClB,WAAS,MAAM,GAAG,MAAM,WAAW,EAAE,KAAK;AACxC,UAAM,WAAW,UAAU,SAAS;AACpC,sBAAkB,OAAO;AACzB,kBAAc,eAAe,CAAC;AAC9B,cAAU,OAAO,KAAK,IAAI,UAAU,MAAM,CAAC;AAC3C,QAAI,MAAM,GAAG;AACX,gBAAU,QAAQ,UAAU,MAAM;AAAA,IACpC;AAAA,EACF;AACA,MAAI,eAAe,gBAAgB;AACjC,UAAM,gBAAgB;AACtB,UAAM,eAAe;AACrB,aAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,sBAAgB,KAAK;AAAA,IACvB;AACA,WAAO;AAAA,MACL;AAAA,MACA,CAAC,cAAc,IAAI;AAAA,MACnB;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF,OAAO;AACL,UAAM,mBAAmB,UAAU,YAAY;AAC/C,UAAM,gBAAgB,aAAa,kBAAkB,cAAc,mBAAmB,IAAI;AAC1F,UAAM,eAAe,aAAa,kBAAkB,aAAa,gBAAgB;AACjF,UAAM,cAAc,IAAI,MAAM,SAAS,EAAE,KAAK,CAAC;AAC/C,aAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,YAAM,MAAM,QAAQ,IAAI;AACxB,YAAM,SAAS,YAAY;AAC3B,YAAM,WAAW,QAAQ,IAAI,IAAI,UAAU,MAAM,MAAM;AACvD,kBAAY;AACZ,eAAS,IAAI,GAAG,IAAI,MAAM,EAAE,GAAG;AAC7B,sBAAc,UAAU,OAAO,KAAK,QAAQ,IAAI,OAAO;AAAA,MACzD;AACA,mBAAa,WAAW,OAAO;AAC/B,sBAAgB,KAAK;AAAA,IACvB;AACA,aAAS,MAAM,GAAG,MAAM,WAAW,EAAE,KAAK;AACxC,YAAM,WAAW,YAAY;AAC7B,UAAI,aAAa,GAAG;AAClB,cAAM,gBAAgB,QAAQ,IAAI,IAAI,UAAU,MAAM;AACtD,sBAAc,gBAAgB,OAAO,KAAK;AAC1C,iBAAS,MAAM,GAAG,MAAM,MAAM,EAAE,KAAK;AACnC,wBAAc,gBAAgB,OAAO,OAAO;AAAA,QAC9C;AACA,qBAAa,iBAAiB;AAAA,MAChC;AAAA,IACF;AACA,WAAO;AAAA,MACL;AAAA,MACA,CAAC,kBAAkB,IAAI;AAAA,MACvB;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AAGA,SAAS,kBAAkB,cAAc,mBAAmB,YAAY,YAAY,aAAa;AAC/F,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,MAAM,kBAAkB;AAC9B,QAAM,aAAa,YAAY;AAC/B,QAAM,cAAc,CAAC;AACrB,MAAI,UAAU;AACd,MAAI,eAAe;AACnB,WAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,UAAM,OAAO,YAAY;AACzB,QAAI,SAAS,IAAI;AACf,UAAI,iBAAiB,IAAI;AACvB,cAAM,IAAI,MAAM,qBAAqB,yDAAyD,cAAc,CAAC,CAAC;AAAA,MAChH;AACA,qBAAe;AACf,kBAAY,KAAK,CAAC;AAAA,IACpB,OAAO;AACL,UAAI,OAAO,GAAG;AACZ,cAAM,IAAI,MAAM,qBAAqB,8CAA8C,GAAG,IAAI,CAAC;AAAA,MAC7F;AACA,iBAAW;AACX,kBAAY,KAAK,IAAI;AAAA,IACvB;AAAA,EACF;AACA,MAAI,iBAAiB,IAAI;AACvB,QAAI,WAAW,GAAG;AAChB,YAAM,IAAI,MAAM,qBAAqB,qDAAqD,CAAC;AAAA,IAC7F;AACA,UAAM,UAAU,KAAK,MAAM,YAAY,OAAO;AAC9C,QAAI,UAAU,YAAY,WAAW;AACnC,YAAM,IAAI,MAAM,qBAAqB,gDAAgD,YAAY,WAAW,CAAC;AAAA,IAC/G;AACA,gBAAY,gBAAgB;AAAA,EAC9B;AACA,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,MAAI,eAAe,WAAW;AAC5B,UAAM,IAAI,MAAM,qBAAqB,gDAAgD,YAAY,WAAW,CAAC;AAAA,EAC/G;AACA,QAAM,YAAY,WAAW;AAC7B,QAAM,eAAe,CAAC;AACtB,MAAI,YAAY,GAAG;AACjB,iBAAa,YAAY,KAAK;AAC9B,aAAS,IAAI,YAAY,GAAG,KAAK,GAAG,EAAE,GAAG;AACvC,mBAAa,KAAK,aAAa,IAAI,KAAK,WAAW,IAAI;AAAA,IACzD;AAAA,EACF;AACA,QAAM,gBAAgB,CAAC;AACvB,MAAI,aAAa,GAAG;AAClB,kBAAc,aAAa,KAAK;AAChC,aAAS,IAAI,aAAa,GAAG,KAAK,GAAG,EAAE,GAAG;AACxC,oBAAc,KAAK,cAAc,IAAI,KAAK,YAAY,IAAI;AAAA,IAC5D;AAAA,EACF;AACA,QAAM,aAAa,aAAa,kBAAkB,YAAY,MAAM,UAAU;AAC9E,WAAS,IAAI,GAAG,IAAI,KAAK,EAAE,GAAG;AAC5B,QAAI,KAAK;AACT,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,YAAM,aAAa,IAAI,YAAY,KAAK,aAAa;AAAA,IACvD;AACA,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,iBAAW,IAAI,aAAa,KAAK,KAAK,MAAM,KAAK,cAAc,EAAE;AACjE,YAAM,cAAc;AAAA,IACtB;AAAA,EACF;AACA,SAAO,CAAC,YAAY,CAAC,KAAK,UAAU,GAAG,WAAW;AACpD;AAGA,SAAS,2BAA2B,QAAQ,YAAY,YAAY,SAAS,YAAY,SAAS,OAAO,eAAe,GAAG;AACzH,QAAM,aAAa,QAAQ;AAC3B,QAAM,YAAY,CAAC,WAAW,IAAI,OAAO,SAAS,WAAW,EAAE;AAC/D,QAAM,SAAS,UAAU;AACzB,QAAM,uBAAuB,aAAa,IAAI,WAAW,aAAa,KAAK,IAAI;AAC/E,QAAM,aAAa;AACnB,MAAI,aAAa,GAAG;AAClB,UAAM,IAAI,MAAM,qBAAqB,wDAAwD,CAAC;AAAA,EAChG;AACA,QAAM,cAAc,WAAW,MAAM;AACrC,cAAY,KAAK;AACjB,QAAM,eAAe,YAAY,OAAO,CAAC,SAAS,UAAU,UAAU,OAAO,CAAC;AAC9E,QAAM,SAAS,aAAa,kBAAkB,YAAY,YAAY;AACtE,MAAI,eAAe,GAAG;AACpB,QAAI,aAAa,GAAG;AAClB,aAAO,KAAK,YAAY;AAAA,IAC1B;AACA,WAAO,CAAC,QAAQ,WAAW;AAAA,EAC7B;AACA,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,qBAAqB,wDAAwD,CAAC;AAAA,EAChG;AACA,MAAI,QAAQ,GAAG,MAAM;AACrB,MAAI,qBAAqB;AACzB,MAAI,WAAW,WAAW;AAC1B,SAAO,MAAM;AACX,QAAI,YAAY;AAChB,QAAI,MAAM,YAAY;AACpB,kBAAY,WAAW;AACvB,UAAI,aAAa,WAAW;AAC1B,UAAE;AACF;AAAA,MACF;AACA,UAAI,YAAY,WAAW;AACzB,cAAM,IAAI,MAAM,qBAAqB,6DAA6D,CAAC;AAAA,MACrG;AAAA,IACF;AACA,QAAI,WAAW,KAAK,YAAY,YAAY;AAC1C,YAAM,IAAI,MAAM,qBAAqB,yDAAyD,UAAU,UAAU,CAAC;AAAA,IACrH;AACA,QAAI,WAAW,oBAAoB;AACjC,aAAO,KAAK,cAAc,qBAAqB,QAAQ,WAAW,MAAM;AAAA,IAC1E;AACA,aAAS,IAAI,OAAO,IAAI,KAAK,EAAE,GAAG;AAChC,YAAM,QAAQ,QAAQ;AACtB,UAAI,QAAQ,KAAK,SAAS,UAAU,IAAI;AACtC,cAAM,IAAI,MAAM,qBAAqB,uDAAuD,GAAG,QAAQ,IAAI,UAAU,EAAE,CAAC;AAAA,MAC1H;AACA,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,eAAO,WAAW,SAAS,MAAM,OAAO,QAAQ,SAAS;AAAA,MAC3D;AAAA,IACF;AACA,QAAI,QAAQ;AACV,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,eAAO,WAAW,SAAS,MAAM,MAAM;AAAA,MACzC;AAAA,IACF;AACA,YAAQ;AACR,MAAE;AACF,yBAAqB,WAAW;AAChC,eAAW;AACX,QAAI,MAAM,YAAY;AACpB;AAAA,IACF;AAAA,EACF;AACA,MAAI,qBAAqB,YAAY;AACnC,WAAO,KAAK,cAAc,qBAAqB,QAAQ,aAAa,MAAM;AAAA,EAC5E;AACA,SAAO,CAAC,QAAQ,WAAW;AAC7B;AAGA,IAAI,WAAW,sBAAsB,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AAC1D,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,wBAAwB,6BAA6B,CAAC,GAAG,MAAM;AACjE,QAAM,OAAO,IAAI;AACjB,SAAO,OAAO;AAChB,CAAC;AACD,IAAI,qBAAqB,iBAAiB,mBAAmB,qBAAqB;AAClF,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,iBAAiB,UAAU,MAAM,SAAS,OAAO;AACxD,QAAM,SAAS,OAAO,UAAU,KAAK,KAAK;AAC1C,WAAS,IAAI,GAAG,IAAI,OAAO,MAAM,KAAK;AACpC,UAAM,MAAM,OAAO,WAAW,CAAC;AAC/B,UAAM,SAAS,IAAI,MAAM,IAAI,MAAM;AACnC,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,aAAO,KAAK,IAAI,KAAK,QAAQ,KAAK,MAAM;AAAA,IAC1C;AACA,WAAO,IAAI,KAAK,IAAI,GAAG,MAAM,GAAG,GAAG,GAAG;AAAA,EACxC;AACA,SAAO;AACT;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,WAAW,aAAa,SAAS,WAAW,UAAU,wBAAwB;AACxF,SAAK,YAAY,aAAa,aAAa,SAAS;AACpD,SAAK,cAAc;AACnB,SAAK,UAAU,aAAa,aAAa,OAAO;AAChD,SAAK,WAAW,aAAa,aAAa,SAAS;AACnD,SAAK,WAAW;AAChB,SAAK,gBAAgB;AAAA,EACvB;AAAA,EACA,YAAY,YAAY;AACtB,WAAO,KAAK,IAAI,KAAK,WAAW,IAAI,aAAa,IAAI,KAAK,UAAU,aAAa,CAAC;AAAA,EACpF;AAAA,EACA,aAAa,QAAQ,YAAY;AAC/B,UAAM,WAAW,KAAK,YAAY,UAAU;AAC5C,WAAO,KAAK,IAAI,GAAG,SAAS,IAAI,WAAW,aAAa,CAAC;AAAA,EAC3D;AAAA,EACA,aAAa,MAAM,YAAY,QAAQ,kBAAkB,WAAW,YAAY;AAC9E,aAAS,aAAa,GAAG,aAAa,WAAW,EAAE,YAAY;AAC7D,YAAM,WAAW,KAAK,YAAY,UAAU;AAC5C,YAAM,cAAc,KAAK,IAAI,GAAG,WAAW,UAAU;AACrD,YAAM,eAAe,KAAK,IAAI,GAAG,YAAY,aAAa,aAAa,GAAG;AAC1E,YAAM,YAAY,cAAc,cAAc;AAC9C,YAAM,iBAAiB,cAAc,cAAc,IAAI,IAAI,aAAa;AACxE,UAAI,YAAY;AAChB,mBAAa,cAAc,KAAK,QAAQ;AACxC,eAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,qBAAa,KAAK,iBAAiB,GAAG;AAAA,MACxC;AACA,mBAAa,eAAe,KAAK,SAAS;AAC1C,YAAM,gBAAgB,cAAc,eAAe,YAAY;AAC/D,mBAAa,gBAAgB,KAAK,UAAU;AAC5C,aAAO,mBAAmB,cAAc,IAAI,WAAW,SAAS;AAChE,YAAM,QAAQ,OAAO,mBAAmB;AACxC,UAAI,iBAAiB;AACrB,YAAM,gBAAgB,CAAC,QAAQ,IAAI,QAAQ,CAAC,UAAU,MAAM,oBAAoB,KAAK;AACrF,eAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,sBAAc,KAAK,OAAO;AAC1B,sBAAc,KAAK,SAAS;AAAA,MAC9B;AACA,eAAS,IAAI,GAAG,IAAI,YAAY,GAAG,EAAE,GAAG;AACtC,sBAAc,KAAK,iBAAiB,EAAE;AACtC,sBAAc,KAAK,SAAS;AAAA,MAC9B;AACA,UAAI,YAAY,GAAG;AACjB,sBAAc,KAAK,iBAAiB,YAAY,EAAE;AAClD,iBAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,wBAAc,KAAK,SAAS;AAC5B,wBAAc,KAAK,QAAQ;AAAA,QAC7B;AAAA,MACF,OAAO;AACL,iBAAS,IAAI,GAAG,IAAI,eAAe,GAAG,EAAE,GAAG;AACzC,wBAAc,KAAK,QAAQ;AAC3B,wBAAc,KAAK,SAAS;AAAA,QAC9B;AACA,sBAAc,KAAK,QAAQ;AAAA,MAC7B;AAAA,IACF;AAAA,EACF;AAAA,EACA,QAAQ,MAAM,QAAQ;AACpB,UAAM,gBAAgB,KAAK;AAC3B,UAAM,aAAa,OAAO;AAC1B,QAAI,aAAa,GAAG;AAClB,UAAI,YAAY,OAAO;AACvB,UAAI,cAAc,GAAG;AACnB,cAAM,IAAI,MAAM,oCAAoC,WAAW;AAAA,MACjE;AACA,eAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAI,cAAc,OAAO,MAAM;AAC/B,sBAAc,eAAe,OAAO,MAAM;AAC1C,YAAI,CAAC,aAAa;AAChB,gBAAM,IAAI,MAAM,uBAAuB,OAAO,mBAAmB,cAAc,gBAAgB;AAAA,QACjG;AACA,oBAAY,OAAO;AAAA,MACrB;AACA,UAAI,cAAc,eAAe;AAC/B,cAAM,IAAI,MAAM,gDAAgD,sBAAsB,WAAW;AAAA,MACnG;AAAA,IACF;AACA,UAAM,gBAAgB,aAAa;AACnC,UAAM,eAAe,aAAa,kBAAkB,SAAS,UAAU;AACvE,QAAI,kBAAkB,KAAK,eAAe,GAAG;AAC3C,YAAM,QAAQ,IAAI,MAAM,aAAa;AACrC,eAAS,IAAI,GAAG,KAAK,eAAe,EAAE,GAAG;AACvC,qBAAa,KAAK;AAAA,MACpB;AACA,aAAO,CAAC,OAAO,YAAY;AAAA,IAC7B;AACA,iBAAa,KAAK;AAClB,aAAS,IAAI,GAAG,KAAK,eAAe,EAAE,GAAG;AACvC,YAAM,SAAS,OAAO,KAAK,OAAO,IAAI;AACtC,UAAI,YAAY;AAChB,WAAK,YAAY,QAAQ,CAAC,eAAe;AACvC,qBAAa,KAAK,aAAa,QAAQ,UAAU;AAAA,MACnD,CAAC;AACD,UAAI,KAAK,iBAAiB,SAAS,KAAK,cAAc,GAAG;AACvD,oBAAY;AAAA,MACd;AACA,mBAAa,KAAK,aAAa,IAAI,KAAK;AAAA,IAC1C;AACA,UAAM,SAAS,IAAI,MAAM,aAAa,cAAc;AACpD,aAAS,IAAI,GAAG,IAAI,eAAe,EAAE,GAAG;AACtC,YAAM,aAAa,OAAO;AAC1B,UAAI,iBAAiB,aAAa;AAClC,WAAK,YAAY,QAAQ,CAAC,eAAe;AACvC,cAAM,SAAS,OAAO,IAAI,KAAK,OAAO;AACtC,cAAM,YAAY,KAAK,aAAa,QAAQ,UAAU;AACtD,aAAK,aAAa,MAAM,YAAY,QAAQ,gBAAgB,WAAW,UAAU;AACjF,0BAAkB;AAAA,MACpB,CAAC;AACD,UAAI,KAAK,iBAAiB,mBAAmB,aAAa,IAAI;AAC5D,cAAM,aAAa,OAAO,IAAI,KAAK,OAAO;AAC1C,YAAI,eAAe,GAAG;AACpB;AAAA,QACF;AACA,cAAM,aAAa,aAAa,IAAI,KAAK;AACzC,cAAM,YAAY;AAClB,aAAK,aAAa,MAAM,YAAY,QAAQ,gBAAgB,WAAW,UAAU;AAAA,MACnF;AAAA,IACF;AACA,WAAO,CAAC,QAAQ,YAAY;AAAA,EAC9B;AACF;AACA,SAAS,iBAAiB,MAAM,YAAY,WAAW,aAAa,SAAS,WAAW,UAAU,wBAAwB;AACxH,SAAO,IAAI,eAAe,WAAW,aAAa,SAAS,WAAW,UAAU,sBAAsB,EAAE,QAAQ,MAAM,UAAU;AAClI;AAGA,SAAS,OAAO,KAAK,YAAY,WAAW,QAAQ;AAClD,MAAI,CAAC,IAAI,QAAQ;AACf;AAAA,EACF;AACA,MAAI,WAAW,WAAW,GAAG;AAC3B,aAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,EAAE,GAAG;AACnC,aAAO,KAAK,IAAI,SAAS,GAAG,IAAI,CAAC,CAAC;AAAA,IACpC;AACA;AAAA,EACF;AACA,MAAI,WAAW,WAAW,GAAG;AAC3B,UAAM,YAAY,WAAW;AAC7B,QAAI,IAAI,IAAI,QAAQ,SAAS;AAC7B,WAAO,MAAM,IAAI;AACf,YAAM,QAAQ,IAAI,SAAS,GAAG,CAAC;AAC/B,UAAI,CAAC,aAAa,MAAM,WAAW,GAAG;AACpC,eAAO,KAAK,KAAK;AAAA,MACnB;AACA,YAAM,IAAI,SAAS,IAAI,CAAC;AACxB,UAAI,IAAI,QAAQ,SAAS;AAAA,IAC3B;AACA,QAAI,CAAC,aAAa,IAAI,WAAW,GAAG;AAClC,aAAO,KAAK,GAAG;AAAA,IACjB;AACA;AAAA,EACF;AACA,MAAI,aAAa;AACjB,WAAS,IAAI,GAAG,IAAI,IAAI,SAAS,GAAG,KAAK;AACvC,QAAI,MAAM,IAAI,UAAU,WAAW,QAAQ,IAAI,EAAE,MAAM,IAAI;AACzD,YAAM,QAAQ,IAAI,SAAS,YAAY,CAAC;AACxC,UAAI,CAAC,aAAa,MAAM,WAAW,GAAG;AACpC,eAAO,KAAK,KAAK;AAAA,MACnB;AACA,mBAAa,IAAI;AAAA,IACnB;AAAA,EACF;AACF;AACA,SAAS,gBAAgB,QAAQ,WAAW,WAAW;AACrD,QAAM,YAAY,OAAO;AACzB,QAAM,SAAS,CAAC;AAChB,MAAI,aAAa;AACjB,MAAI,gBAAgB;AACpB,QAAM,aAAa,IAAI,MAAM,SAAS;AACtC,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,UAAM,mBAAmB,OAAO;AAChC,WAAO,OAAO,IAAI,WAAW,WAAW,MAAM;AAC9C,UAAM,WAAW,OAAO,SAAS;AACjC,eAAW,KAAK;AAChB,kBAAc;AACd,oBAAgB,KAAK,IAAI,eAAe,QAAQ;AAAA,EAClD;AACA,QAAM,UAAU,aAAa,kBAAkB,SAAS,aAAa,CAAC;AACtE,QAAM,SAAS,IAAI,MAAM,UAAU;AACnC,QAAM,QAAQ,CAAC,WAAW,aAAa;AACvC,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAS,IAAI,GAAG,IAAI,WAAW,IAAI,EAAE,GAAG;AACtC,cAAQ,IAAI,KAAK;AACjB,cAAQ,IAAI,IAAI,KAAK;AACrB,aAAO,KAAK,OAAO;AACnB,QAAE;AAAA,IACJ;AAAA,EACF;AACA,SAAO,CAAC,SAAS,QAAQ,KAAK;AAChC;AAGA,SAAS,2BAA2B,QAAQ,YAAY;AACtD,QAAM,SAAS,aAAa,kBAAkB,SAAS,OAAO,MAAM;AACpE,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,WAAO,KAAK,aAAa,cAAc,OAAO,EAAE,EAAE,OAAO,UAAU,EAAE,mBAAmB;AAAA,EAC1F;AACA,SAAO;AACT;AAGA,IAAI,UAAU,6BAA6B,CAAC,QAAQ,WAAW,SAAS,MAAM;AAC9E,IAAI,iBAAiB,8BAA8B,CAAC,OAAO,OAAO,OAAO,UAAU;AACjF,SAAO,EAAE,MAAM,QAAQ,OAAO,MAAM,QAAQ,MAAM;AACpD,CAAC;AACD,IAAI,OAAO,iBAAiB,KAAK,SAAS,cAAc;AACxD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM,MAAM;AAC5B,QAAM,WAAW,IAAI,MAAM,KAAK,IAAI;AACpC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAS,KAAK,KAAK,MAAM,KAAK,KAAK;AAAA,EACrC;AACA,QAAM,SAAS,OAAO,UAAU,KAAK,KAAK;AAC1C,WAAS,IAAI,GAAG,IAAI,OAAO,OAAO,QAAQ,EAAE,GAAG;AAC7C,UAAM,SAAS,OAAO,WAAW,CAAC;AAClC,UAAM,cAAc,IAAI,MAAM,KAAK,IAAI;AACvC,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,kBAAY,KAAK,OAAO,KAAK,KAAK,MAAM;AAAA,IAC1C;AACA,UAAM,gBAAgB,KAAK,WAAW,WAAW;AACjD,WAAO,OAAO,KAAK,KAAK,OAAO;AAAA,EACjC;AACA,SAAO;AACT;AAGA,IAAI,cAAc,CAAC,GAAG,MAAM;AAC1B,QAAM,YAAY,EAAE,QAAQ,EAAE;AAC9B,SAAO,cAAc,IAAI,EAAE,QAAQ,EAAE,QAAQ;AAC/C;AACA,SAAS,OAAO,QAAQ,GAAG,OAAO,GAAG,QAAQ,OAAO,SAAS,GAAG;AAC9D,SAAO,QAAQ,MAAM;AACnB,QAAI,QAAQ,OAAO,KAAK;AACtB,YAAM,IAAI,QAAQ,OAAO;AACzB,YAAM,KAAK,IAAI,OAAO;AACtB,YAAM,IAAI,KAAK,IAAI,CAAC;AACpB,YAAM,IAAI,MAAM,KAAK,IAAI,IAAI,IAAI,CAAC;AAClC,YAAM,KAAK,MAAM,KAAK,KAAK,IAAI,KAAK,IAAI,KAAK,CAAC,IAAI,KAAK,KAAK,KAAK,IAAI,CAAC;AACtE,YAAM,UAAU,KAAK,IAAI,MAAM,KAAK,MAAM,IAAI,KAAK,IAAI,IAAI,EAAE,CAAC;AAC9D,YAAM,WAAW,KAAK,IAAI,OAAO,KAAK,MAAM,KAAK,IAAI,MAAM,IAAI,IAAI,EAAE,CAAC;AACtE,aAAO,QAAQ,GAAG,SAAS,QAAQ;AAAA,IACrC;AACA,UAAM,IAAI,OAAO;AACjB,QAAI,IAAI;AACR,QAAI,IAAI;AACR,iBAAa,KAAK,QAAQ,MAAM,CAAC;AACjC,QAAI,YAAY,OAAO,QAAQ,CAAC,IAAI,GAAG;AACrC,mBAAa,KAAK,QAAQ,MAAM,KAAK;AAAA,IACvC;AACA,WAAO,IAAI,GAAG;AACZ,mBAAa,KAAK,QAAQ,GAAG,CAAC;AAC9B;AACA;AACA,aAAO,YAAY,OAAO,IAAI,CAAC,IAAI,GAAG;AACpC,YAAI,IAAI;AAAA,MACV;AACA,aAAO,YAAY,OAAO,IAAI,CAAC,IAAI,GAAG;AACpC,YAAI,IAAI;AAAA,MACV;AAAA,IACF;AACA,QAAI,YAAY,OAAO,OAAO,CAAC,MAAM,GAAG;AACtC,mBAAa,KAAK,QAAQ,MAAM,CAAC;AAAA,IACnC,OAAO;AACL,UAAI,IAAI;AACR,mBAAa,KAAK,QAAQ,GAAG,KAAK;AAAA,IACpC;AACA,QAAI,KAAK,GAAG;AACV,aAAO,IAAI;AAAA,IACb;AACA,QAAI,KAAK,GAAG;AACV,cAAQ,IAAI;AAAA,IACd;AAAA,EACF;AACF;AACA,SAAS,SAAS,GAAG,QAAQ,QAAQ,GAAG,QAAQ;AAC9C,QAAM,UAAU,OAAO,OAAO,SAAS;AACvC,QAAM,CAAC,OAAO,IAAI,IAAI,CAAC,EAAE,SAAS,SAAS,OAAO;AAClD,QAAM,cAAc,aAAa,uBAAuB,QAAQ,QAAQ,CAAC;AACzE,QAAM,iBAAiB,aAAa,uBAAuB,SAAS,QAAQ,CAAC;AAC7E,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,SAAS,IAAI;AACnB,UAAM,OAAO,EAAE,SAAS,QAAQ,SAAS,IAAI;AAC7C,QAAI,YAAY,IAAI,MAAM,KAAK,MAAM;AACrC,SAAK,QAAQ,CAAC,OAAO,UAAU,UAAU,SAAS,EAAE,OAAO,MAAM,CAAC;AAClE,QAAI,IAAI,UAAU,QAAQ;AACxB,aAAO,WAAW,CAAC;AACnB,kBAAY,UAAU,MAAM,GAAG,CAAC;AAAA,IAClC;AACA,QAAI,QAAQ;AACV,gBAAU,KAAK,WAAW;AAAA,IAC5B;AACA,UAAM,YAAY,IAAI;AACtB,UAAM,WAAW,YAAY,SAAS,WAAW,YAAY,CAAC;AAC9D,UAAM,cAAc,eAAe,SAAS,WAAW,YAAY,CAAC;AACpE,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,eAAS,KAAK,UAAU,GAAG;AAC3B,kBAAY,KAAK,UAAU,GAAG;AAAA,IAChC;AAAA,EACF;AACA,QAAM,cAAc,OAAO,MAAM;AACjC,cAAY,YAAY,SAAS,KAAK;AACtC,SAAO;AAAA,IACL,OAAO,aAAa,QAAQ,WAAW;AAAA,IACvC,OAAO,aAAa,SAAS,cAAc;AAAA,EAC7C;AACF;AAGA,SAAS,WAAW,QAAQ,MAAM,OAAO,OAAO;AAC9C,QAAM,QAAQ,aAAa,eAAe,MAAM,KAAK,EAAE;AACvD,QAAM,WAAW,CAAC,GAAG,MAAM,IAAI,CAAC;AAChC,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,aAAS,MAAM,MAAM;AAAA,EACvB;AACA,WAAS,KAAK,MAAM;AACpB,WAAS,IAAI,QAAQ,GAAG,IAAI,MAAM,QAAQ,KAAK;AAC7C,aAAS,MAAM,MAAM;AAAA,EACvB;AACA,QAAM,iBAAiB,CAAC;AACxB,QAAM,UAAU,IAAI,WAAW,MAAM,MAAM;AAC3C,QAAM,cAAc,IAAI,aAAa,UAAU,OAAO,MAAM;AAC5D,QAAM,gBAAgB,CAAC;AACvB,QAAM,aAAa,SAAS,OAAO,KAAK,SAAS,OAAO;AACxD,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,QAAI;AACJ,QAAI,YAAY;AACd,gBAAU,OAAO,GAAG,SAAS;AAAA,IAC/B,OAAO;AACL,YAAM,aAAa,CAAC;AACpB,eAAS,IAAI,GAAG,IAAI,SAAS,IAAI,KAAK;AACpC,iBAAS,IAAI,GAAG,IAAI,SAAS,IAAI,KAAK;AACpC,qBAAW,KAAK,YAAY,IAAI,GAAG,GAAG,CAAC,CAAC;AAAA,QAC1C;AAAA,MACF;AACA,gBAAU,WAAW,KAAK,GAAG;AAAA,IAC/B;AACA,QAAI,eAAe,aAAa,QAAQ;AACtC,cAAQ,KAAK,eAAe;AAAA,IAC9B,OAAO;AACL,YAAM,cAAc,OAAO,KAAK,cAAc,EAAE;AAChD,qBAAe,WAAW;AAC1B,cAAQ,KAAK;AACb,oBAAc,KAAK,CAAC;AAAA,IACtB;AAAA,EACF;AACA,QAAM,iBAAiB,SAAS,MAAM;AACtC,iBAAe,KAAK,OAAO,KAAK,cAAc,EAAE;AAChD,QAAM,eAAe,IAAI,aAAa,gBAAgB,KAAK;AAC3D,gBAAc,QAAQ,CAAC,oBAAoB,MAAM;AAC/C,aAAS,IAAI,GAAG,IAAI,SAAS,IAAI,KAAK;AACpC,eAAS,IAAI,GAAG,IAAI,SAAS,IAAI,KAAK;AACpC,qBAAa,IAAI,YAAY,IAAI,GAAG,oBAAoB,CAAC,GAAG,GAAG,GAAG,CAAC;AAAA,MACrE;AAAA,IACF;AAAA,EACF,CAAC;AACD,QAAM,cAAc,MAAM,MAAM;AAChC,cAAY,SAAS,eAAe;AACpC,SAAO;AAAA,IACL,cAAc,aAAa;AAAA,IAC3B;AAAA,IACA;AAAA,EACF;AACF;AAGA,gBAAgB,OAAO,MAAM,IAAI,eAAe,GAAG,CAAC;AAGpD,IAAI,OAAO,gBAAgB,KAAK,CAAC,OAAO,MAAM,IAAI,KAAK,KAAK,IAAI,EAAE,IAAI,CAAC;AACvE,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,mBAAiB,CAAC,CAAC,GAAG,WAAW;AACjC,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,UAAU,aAAa,uBAAuB,WAAW,KAAK;AACpE,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,YAAQ,KAAK,MAAM,KAAK,IAAI,QAAQ,MAAM,KAAK,MAAM;AAAA,EACvD;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,WAAW,OAAO;AAC5D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY,6BAA6B,CAAC,QAAQ,WAAW,SAAS,IAAI,SAAS,SAAS,MAAM;AACtG,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,GAAG,MAAM,IAAI;AACrB,mBAAiB,CAAC,GAAG,KAAK,GAAG,OAAO;AACpC,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAC9C,QAAM,CAAC,YAAY,WAAW,IAAI,UAAU,EAAE,OAAO,MAAM,OAAO,OAAO,OAAO,SAAS;AACzF,SAAO,SAAS,eAAe,aAAa,WAAW,UAAU;AACnE;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,IAAI,GAAG,EAAE,CAAC;AACzD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,IAAI,KAAK,IAAI,GAAG,EAAE,GAAG,CAAC,CAAC;AACxE,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,iBAAiB,UAAU,GAAG,aAAa,wBAAwB,gBAAgB;AAC1F,MAAI,gBAAgB,UAAU;AAC5B,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD,WAAW,gBAAgB,QAAQ;AACjC,WAAO,MAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACnD,WAAW,gBAAgB,OAAO;AAChC,WAAO,KAAK,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EAClD,WAAW,gBAAgB,SAAS;AAClC,WAAO,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACpD,WAAW,gBAAgB,SAAS;AAClC,WAAO,OAAO,EAAE,QAAQ,EAAE,GAAG,OAAO,uBAAuB,GAAG,SAAS,SAAS,CAAC;AAAA,EACnF,WAAW,gBAAgB,aAAa;AACtC,WAAO,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,eAAe,EAAE,CAAC;AAAA,EAC1F,WAAW,gBAAgB,WAAW;AACpC,WAAO,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACtD;AACA,QAAM,IAAI,MAAM,cAAc,2DAA2D;AAC3F;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,SAAS,aAAa,uBAAuB,OAAO,KAAK;AAC/D,QAAM,SAAS,aAAa,cAAc,MAAM;AAChD,eAAa,OAAO,UAAU,QAAQ,MAAM,kBAAkB,eAAe,sCAAsC,EAAE,cAAc,oFAAoF;AACvN,WAAS,OAAO,EAAE,MAAM;AACxB,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM;AACxC,MAAI,MAAM,sBAAsB,MAAM;AACpC,UAAM,QAAQ,MAAM,mBAAmB;AACvC,UAAM,QAAQ,MAAM,mBAAmB;AACvC,UAAM,QAAQ;AACd,UAAM,QAAQ;AAAA,EAChB;AACA,SAAO,EAAE,QAAQ,EAAE,QAAQ,OAAO,QAAQ,OAAO,EAAE,MAAM;AAC3D;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,EAAE,IAAI;AACjB,QAAM,EAAE,YAAY,WAAW,IAAI;AACnC,mBAAiB,CAAC,GAAG,CAAC,GAAG,QAAQ;AACjC,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,oBAAoB,uBAAuB,2BAA2B,EAAE,MAAM,MAAM,GAAG,EAAE,GAAG,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC;AACtH,QAAM,WAAW,kBAAkB,OAAO,CAAC,aAAa,WAAW,CAAC;AACpE,eAAa,OAAO,gBAAgB,aAAa,MAAM,kCAAkC,qBAAqB,uCAAuC,EAAE,aAAa,EAAE,wBAAwB,6BAA6B,wBAAwB;AACnP,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,YAAY,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACxD,QAAM,UAAU,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACtD,QAAM,WAAW,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACvD,QAAM,WAAW,KAAK,IAAI,WAAW,SAAS;AAC9C,QAAM,YAAY,SAAS,KAAK,IAAI,IAAI,MAAM,EAAE;AAChD,QAAM,YAAY,SAAS,KAAK,IAAI,IAAI,MAAM,EAAE;AAChD,QAAM,aAAa,aAAa,eAAe,IAAI,KAAK;AACxD,QAAM,aAAa,aAAa,eAAe,IAAI,KAAK;AACxD,QAAM,CAAC,QAAQ,YAAY,UAAU,IAAI,aAAa,CAAC,WAAW,IAAI,GAAG,WAAW,EAAE,IAAI,CAAC,WAAW,IAAI,WAAW,IAAI,CAAC;AAC1H,QAAM,CAAC,YAAY,YAAY,MAAM,IAAI,aAAa,CAAC,GAAG,WAAW,IAAI,WAAW,EAAE,IAAI,CAAC,WAAW,IAAI,GAAG,WAAW,EAAE;AAC1H,QAAM,OAAO,UAAU;AACvB,QAAM,SAAS,OAAO,CAAC,UAAU,SAAS,QAAQ,GAAG,IAAI,KAAK;AAC9D,QAAM,UAAU,OAAO;AACvB,QAAM,YAAY,SAAS;AAC3B,WAAS,KAAK,GAAG,KAAK,UAAU,MAAM;AACpC,aAAS,KAAK,GAAG,KAAK,SAAS,MAAM,WAAW;AAC9C,eAAS,KAAK,GAAG,KAAK,UAAU,MAAM,WAAW;AAC/C,iBAAS,MAAM,GAAG,MAAM,WAAW,OAAO,WAAW;AACnD,gBAAM,SAAS,KAAK,IAAI,KAAK,WAAW,OAAO;AAC/C,gBAAM,SAAS,KAAK,IAAI,KAAK,WAAW,QAAQ;AAChD,gBAAM,SAAS,KAAK,IAAI,MAAM,WAAW,SAAS;AAClD,mBAAS,IAAI,IAAI,IAAI,QAAQ,KAAK;AAChC,qBAAS,IAAI,IAAI,IAAI,QAAQ,KAAK;AAChC,kBAAI,OAAO;AACX,uBAAS,IAAI,KAAK,IAAI,QAAQ,KAAK;AACjC,sBAAM,eAAe,KAAK,IAAI,IAAI,YAAY,CAAC,IAAI;AACnD,sBAAM,eAAe,KAAK,IAAI,IAAI,YAAY,CAAC,IAAI;AACnD,sBAAM,OAAO,UAAU,eAAe,IAAI,aAAa,IAAI;AAC3D,sBAAM,OAAO,UAAU,IAAI,aAAa,IAAI,aAAa;AACzD,wBAAQ,OAAO;AAAA,cACjB;AACA,sBAAQ,KAAK,QAAQ,IAAI,WAAW,OAAO;AAAA,YAC7C;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,WAAS,8BAA8B,GAAG;AAC1C,WAAS,8BAA8B,GAAG;AAC1C,SAAO,SAAS,eAAe,UAAU,OAAO,OAAO,OAAO,MAAM;AACtE;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,MAAM,uBAAuB,IAAI;AAC/C,QAAM,EAAE,YAAY,YAAY,YAAY,aAAa,eAAe,IAAI;AAC5E,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,QAAM,gBAAgB,CAAC;AACvB,QAAM,YAAY,YAAY,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,OAAO,EAAE,YAAY,WAAW,GAAG,SAAS,SAAS,CAAC;AACxG,YAAU;AACV,MAAI,MAAM;AACR,aAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,KAAK,GAAG,SAAS,SAAS,CAAC;AACpE,kBAAc,KAAK,OAAO;AAC1B,cAAU;AAAA,EACZ;AACA,MAAI,aAAa;AACf,oBAAgB,iBAAiB,UAAU,SAAS,aAAa,wBAAwB,cAAc;AACvG,kBAAc,KAAK,OAAO;AAC1B,cAAU;AAAA,EACZ;AACA,aAAW,KAAK,eAAe;AAC7B,aAAS,8BAA8B,CAAC;AAAA,EAC1C;AACA,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC1D,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,UAAU;AAChB,mBAAiB,QAAQ,MAAM;AAC/B,QAAM,OAAO,QAAQ,IAAI,CAAC,MAAM,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM;AAClE,QAAM,SAAS,OAAO,QAAQ,GAAG,OAAO,QAAQ,GAAG,KAAK;AACxD,QAAM,UAAU,OAAO;AACvB,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,UAAM,WAAW,KAAK;AACtB,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,cAAQ,MAAM,SAAS;AAAA,IACzB;AAAA,EACF;AACA,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC1E;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,mBAAiB,GAAG,KAAK;AACzB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,EAC1E;AACA,uBAAqB,2BAA2B,OAAO,MAAM,GAAG,MAAM,MAAM;AAC5E,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,GAAG,OAAO,IAAI;AAC7F,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,OAAO,aAAa,oBAAoB,aAAa,cAAc,QAAQ,GAAG,GAAG,KAAK;AAC5F,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO,MAAM;AACjB,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,aAAO,QAAQ;AAAA,IACjB;AACA,SAAK,KAAK;AAAA,EACZ;AACA,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,EAAE;AAAA,EAC3C;AACA,QAAM,SAAS,SAAS,eAAe,UAAU,GAAG,OAAO,IAAI;AAC/D,MAAI,UAAU;AACZ,UAAM,gBAAgB,qBAAqB,qBAAqB,UAAU,QAAQ;AAClF,UAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAC7G,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,mBAAiB,GAAG,KAAK;AACzB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,EAC1E;AACA,uBAAqB,2BAA2B,OAAO,MAAM,GAAG,MAAM,MAAM;AAC5E,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,GAAG,OAAO,IAAI;AAC7F,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,OAAO,aAAa,oBAAoB,aAAa,cAAc,QAAQ,GAAG,GAAG,KAAK;AAC5F,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,SAAS,MAAM;AACnB,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,eAAS,UAAU;AAAA,IACrB;AACA,SAAK,KAAK;AAAA,EACZ;AACA,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,EAAE;AAAA,EAC3C;AACA,QAAM,SAAS,SAAS,eAAe,UAAU,GAAG,OAAO,IAAI;AAC/D,MAAI,UAAU;AACZ,UAAM,gBAAgB,qBAAqB,qBAAqB,UAAU,QAAQ;AAClF,UAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAC7G,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,mBAAiB,GAAG,QAAQ;AAC5B,MAAI,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACpD,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,QAAM,0BAA0B,CAAC;AACjC,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,4BAAwB,KAAK,EAAE;AAC/B,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,GAAG,MAAM,MAAM;AAAA,EAC3E;AACA,SAAO,CAAC,KAAK,EAAE;AACf,uBAAqB,2BAA2B,UAAU,MAAM,GAAG,MAAM,MAAM;AAC/E,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,GAAG,OAAO,IAAI;AAC7F,QAAM,UAAU,aAAa,cAAc,QAAQ;AACnD,QAAM,OAAO,aAAa,oBAAoB,SAAS,OAAO;AAC9D,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO,MAAM;AACjB,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,UAAI,QAAQ,MAAM;AAChB,eAAO;AACP,mBAAW;AAAA,MACb;AAAA,IACF;AACA,SAAK,KAAK;AAAA,EACZ;AACA,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO,SAAS,eAAe,UAAU,SAAS,IAAI;AACxD;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,mBAAiB,GAAG,QAAQ;AAC5B,MAAI,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACpD,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,QAAM,0BAA0B,CAAC;AACjC,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,4BAAwB,KAAK,EAAE;AAC/B,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,GAAG,MAAM,MAAM;AAAA,EAC3E;AACA,SAAO,CAAC,KAAK,EAAE;AACf,uBAAqB,2BAA2B,UAAU,MAAM,GAAG,MAAM,MAAM;AAC/E,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,GAAG,OAAO,IAAI;AAC7F,QAAM,UAAU,aAAa,cAAc,QAAQ;AACnD,QAAM,OAAO,aAAa,oBAAoB,SAAS,OAAO;AAC9D,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO,MAAM;AACjB,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,UAAI,QAAQ,MAAM;AAChB,eAAO;AACP,mBAAW;AAAA,MACb;AAAA,IACF;AACA,SAAK,KAAK;AAAA,EACZ;AACA,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO,SAAS,eAAe,UAAU,SAAS,IAAI;AACxD;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC1D,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY,6BAA6B,CAAC,QAAQ,WAAW,KAAK,MAAM,QAAQ,MAAM,CAAC;AAC3F,IAAI,SAAS,iBAAiB,OAAO,SAAS;AAC9C,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC1D,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,SAAS,QAAQ,OAAO,SAAS,UAAU,UAAU;AAClE,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,eAAe,aAAa,QAAQ,OAAO,oBAAoB,OAAO;AAC5E,QAAM,SAAS,OAAO,SAAS,UAAU,KAAK;AAC9C,QAAM,aAAa,OAAO;AAC1B,QAAM,qBAAqB,SAAS,SAAS,KAAK,SAAS,SAAS,KAAK,SAAS,SAAS;AAC3F,QAAM,mBAAmB,SAAS,SAAS,KAAK,SAAS,SAAS;AAClE,QAAM,mBAAmB,SAAS,SAAS;AAC3C,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,UAAM,oBAAoB,IAAI;AAC9B,UAAM,mBAAmB,IAAI,QAAQ;AACrC,aAAS,IAAI,GAAG,IAAI,SAAS,YAAY,EAAE,GAAG;AAC5C,eAAS,KAAK,GAAG,KAAK,SAAS,WAAW,EAAE,IAAI;AAC9C,cAAM,WAAW,KAAK,eAAe;AACrC,cAAM,QAAQ,KAAK,IAAI,GAAG,QAAQ;AAClC,cAAM,QAAQ,KAAK,IAAI,SAAS,UAAU,wBAAwB,QAAQ;AAC1E,cAAM,kBAAkB,oBAAoB,KAAK;AACjD,iBAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,gBAAM,WAAW,KAAK,cAAc;AACpC,gBAAM,QAAQ,KAAK,IAAI,GAAG,QAAQ;AAClC,gBAAM,QAAQ,KAAK,IAAI,SAAS,SAAS,uBAAuB,QAAQ;AACxE,cAAI,cAAc;AAClB,cAAI,WAAW;AACf,cAAI,SAAS;AACb,mBAAS,KAAK,OAAO,KAAK,OAAO,MAAM,gBAAgB;AACrD,kBAAM,WAAW,mBAAmB,KAAK,QAAQ;AACjD,qBAAS,KAAK,OAAO,KAAK,OAAO,MAAM,eAAe;AACpD,oBAAM,WAAW,WAAW,KAAK,QAAQ;AACzC,oBAAM,QAAQ,QAAQ,WAAW;AACjC,kBAAI,aAAa,SAAS,QAAQ,aAAa;AAC7C,8BAAc;AAAA,cAChB,WAAW,aAAa,OAAO;AAC7B,4BAAY;AACZ;AAAA,cACF;AAAA,YACF;AACA,gBAAI,MAAM,WAAW,GAAG;AACtB;AAAA,YACF;AAAA,UACF;AACA,gBAAM,eAAe,kBAAkB,KAAK,mBAAmB;AAC/D,qBAAW,gBAAgB,aAAa,QAAQ,WAAW,SAAS;AAAA,QACtE;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,SAAS,QAAQ,OAAO,UAAU,mBAAmB,OAAO,sBAAsB,OAAO;AACjH,QAAM,eAAe,OAAO,SAAS,UAAU,OAAO;AACtD,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,OAAO,OAAO,QAAQ,OAAO,OAAO;AAC1C,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,aAAS,IAAI,GAAG,IAAI,SAAS,YAAY,EAAE,GAAG;AAC5C,eAAS,KAAK,GAAG,KAAK,SAAS,WAAW,EAAE,IAAI;AAC9C,cAAM,WAAW,KAAK,eAAe;AACrC,YAAI,QAAQ;AACZ,eAAO,QAAQ,GAAG;AAChB,mBAAS;AAAA,QACX;AACA,cAAM,QAAQ,KAAK,IAAI,SAAS,UAAU,wBAAwB,QAAQ;AAC1E,iBAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,gBAAM,WAAW,KAAK,cAAc;AACpC,cAAI,QAAQ;AACZ,iBAAO,QAAQ,GAAG;AAChB,qBAAS;AAAA,UACX;AACA,gBAAM,QAAQ,KAAK,IAAI,SAAS,SAAS,uBAAuB,QAAQ;AACxE,cAAI,WAAW,OAAO;AACtB,cAAI,cAAc;AAClB,mBAAS,KAAK,OAAO,KAAK,OAAO,MAAM,gBAAgB;AACrD,kBAAM,KAAK,KAAK;AAChB,qBAAS,KAAK,OAAO,KAAK,OAAO,MAAM,eAAe;AACpD,oBAAM,KAAK,KAAK;AAChB,oBAAM,QAAQ,KAAK,IAAI,GAAG,IAAI,IAAI,CAAC;AACnC,kBAAI,QAAQ,UAAU;AACpB,2BAAW;AACX,oBAAI,kBAAkB;AACpB,gCAAc,wBAAwB,IAAI,SAAS,WAAW,MAAM,SAAS,UAAU,MAAM,SAAS,aAAa,KAAK,KAAK,SAAS,UAAU,MAAM,SAAS,aAAa;AAAA,gBAC9K,OAAO;AACL,gCAAc,KAAK,uBAAuB;AAAA,gBAC5C;AAAA,cACF;AAAA,YACF;AAAA,UACF;AACA,uBAAa,IAAI,aAAa,GAAG,IAAI,IAAI,CAAC;AAAA,QAC5C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,QAAQ,SAAS,QAAQ,OAAO,SAAS,UAAU,UAAU;AACpE,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,uBAAuB,SAAS;AACtC,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,eAAe,aAAa,QAAQ,OAAO,oBAAoB,OAAO;AAC5E,QAAM,SAAS,OAAO,SAAS,UAAU,KAAK;AAC9C,QAAM,aAAa,OAAO;AAC1B,QAAM,qBAAqB,SAAS,SAAS,KAAK,SAAS,SAAS,KAAK,SAAS,SAAS,KAAK,SAAS,SAAS;AAClH,QAAM,qBAAqB,SAAS,SAAS,KAAK,SAAS,SAAS,KAAK,SAAS,SAAS;AAC3F,QAAM,mBAAmB,SAAS,SAAS,KAAK,SAAS,SAAS;AAClE,QAAM,mBAAmB,SAAS,SAAS;AAC3C,WAAS,QAAQ,GAAG,QAAQ,SAAS,WAAW,EAAE,OAAO;AACvD,UAAM,oBAAoB,QAAQ;AAClC,UAAM,mBAAmB,QAAQ,QAAQ;AACzC,aAAS,UAAU,GAAG,UAAU,SAAS,YAAY,EAAE,SAAS;AAC9D,eAAS,SAAS,GAAG,SAAS,SAAS,UAAU,EAAE,QAAQ;AACzD,cAAM,eAAe,SAAS,cAAc;AAC5C,YAAI,YAAY;AAChB,eAAO,YAAY,GAAG;AACpB,uBAAa;AAAA,QACf;AACA,cAAM,YAAY,KAAK,IAAI,SAAS,SAAS,uBAAuB,YAAY;AAChF,cAAM,oBAAoB,oBAAoB,SAAS;AACvD,iBAAS,OAAO,GAAG,OAAO,SAAS,WAAW,EAAE,MAAM;AACpD,gBAAM,aAAa,OAAO,eAAe;AACzC,cAAI,UAAU;AACd,iBAAO,UAAU,GAAG;AAClB,uBAAW;AAAA,UACb;AACA,gBAAM,UAAU,KAAK,IAAI,SAAS,UAAU,wBAAwB,UAAU;AAC9E,gBAAM,kBAAkB,oBAAoB,OAAO;AACnD,mBAAS,OAAO,GAAG,OAAO,SAAS,UAAU,EAAE,MAAM;AACnD,kBAAM,aAAa,OAAO,cAAc;AACxC,gBAAI,UAAU;AACd,mBAAO,UAAU,GAAG;AAClB,yBAAW;AAAA,YACb;AACA,kBAAM,UAAU,KAAK,IAAI,SAAS,SAAS,uBAAuB,UAAU;AAC5E,kBAAM,kBAAkB,kBAAkB,OAAO;AACjD,gBAAI,cAAc;AAClB,gBAAI,WAAW;AACf,gBAAI,SAAS;AACb,qBAAS,SAAS,WAAW,SAAS,WAAW,UAAU,eAAe;AACxE,oBAAM,eAAe,mBAAmB,SAAS,QAAQ;AACzD,uBAAS,OAAO,SAAS,OAAO,SAAS,QAAQ,gBAAgB;AAC/D,sBAAM,aAAa,eAAe,OAAO,QAAQ;AACjD,yBAAS,OAAO,SAAS,OAAO,SAAS,QAAQ,eAAe;AAC9D,wBAAM,aAAa,aAAa,OAAO,QAAQ;AAC/C,wBAAM,QAAQ,QAAQ,aAAa;AACnC,sBAAI,aAAa,SAAS,QAAQ,aAAa;AAC7C,kCAAc;AAAA,kBAChB,WAAW,aAAa,OAAO;AAC7B,gCAAY;AACZ;AAAA,kBACF;AACA,sBAAI,MAAM,WAAW,GAAG;AACtB;AAAA,kBACF;AAAA,gBACF;AACA,oBAAI,MAAM,WAAW,GAAG;AACtB;AAAA,gBACF;AAAA,cACF;AACA,kBAAI,MAAM,WAAW,GAAG;AACtB;AAAA,cACF;AAAA,YACF;AACA,kBAAM,eAAe,kBAAkB;AACvC,uBAAW,gBAAgB,aAAa,QAAQ,WAAW,SAAS;AAAA,UACtE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,mBAAmB,MAAM,UAAU;AAC1C,QAAM,eAAe,OAAO,SAAS,UAAU,OAAO;AACtD,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,uBAAuB,SAAS;AACtC,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,UAAU,SAAS,QAAQ;AACjC,WAAS,QAAQ,GAAG,QAAQ,SAAS,WAAW,EAAE,OAAO;AACvD,aAAS,UAAU,GAAG,UAAU,SAAS,YAAY,EAAE,SAAS;AAC9D,eAAS,SAAS,GAAG,SAAS,SAAS,UAAU,EAAE,QAAQ;AACzD,cAAM,eAAe,SAAS,cAAc;AAC5C,YAAI,YAAY;AAChB,eAAO,YAAY,GAAG;AACpB,uBAAa;AAAA,QACf;AACA,cAAM,YAAY,KAAK,IAAI,SAAS,SAAS,uBAAuB,YAAY;AAChF,iBAAS,OAAO,GAAG,OAAO,SAAS,WAAW,EAAE,MAAM;AACpD,gBAAM,aAAa,OAAO,eAAe;AACzC,cAAI,UAAU;AACd,iBAAO,UAAU,GAAG;AAClB,uBAAW;AAAA,UACb;AACA,gBAAM,UAAU,KAAK,IAAI,SAAS,UAAU,wBAAwB,UAAU;AAC9E,mBAAS,OAAO,GAAG,OAAO,SAAS,UAAU,EAAE,MAAM;AACnD,kBAAM,aAAa,OAAO,cAAc;AACxC,gBAAI,UAAU;AACd,mBAAO,UAAU,GAAG;AAClB,yBAAW;AAAA,YACb;AACA,kBAAM,UAAU,KAAK,IAAI,SAAS,SAAS,uBAAuB,UAAU;AAC5E,gBAAI,WAAW,OAAO;AACtB,gBAAI,cAAc;AAClB,qBAAS,SAAS,WAAW,SAAS,WAAW,UAAU,eAAe;AACxE,oBAAM,SAAS,SAAS;AACxB,uBAAS,OAAO,SAAS,OAAO,SAAS,QAAQ,gBAAgB;AAC/D,sBAAM,OAAO,OAAO;AACpB,yBAAS,OAAO,SAAS,OAAO,SAAS,QAAQ,eAAe;AAC9D,wBAAM,OAAO,OAAO;AACpB,wBAAM,QAAQ,KAAK,IAAI,OAAO,QAAQ,MAAM,MAAM,OAAO;AACzD,sBAAI,SAAS,UAAU;AACrB,+BAAW;AACX,kCAAc,SAAS,wBAAwB,uBAAuB,OAAO,wBAAwB;AAAA,kBACvG;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,yBAAa,IAAI,aAAa,OAAO,QAAQ,MAAM,MAAM,OAAO;AAAA,UAClE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,mBAAiB,GAAG,SAAS;AAC7B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY;AAClB,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AACjN,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,MAAI;AACJ,MAAI,SAAS,gBAAgB,KAAK,SAAS,iBAAiB,KAAK,aAAa,YAAY,SAAS,SAAS,SAAS,QAAQ,GAAG;AAC9H,UAAM,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACtD,OAAO;AACL,UAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,UAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,UAAM,UAAU,MAAM,SAAS,EAAE,OAAO,EAAE,OAAO,UAAU,UAAU,KAAK;AAC1E,UAAM,SAAS,eAAe,SAAS,UAAU,EAAE,OAAO,QAAQ,MAAM;AAAA,EAC1E;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACxE,mBAAiB,GAAG,WAAW;AAC/B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,iBAAiB,UAAU;AAC1H,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,SAAS,QAAQ,SAAS,EAAE,OAAO,EAAE,OAAO,aAAa,eAAe,EAAE,KAAK,GAAG,UAAU,KAAK;AACvG,SAAO,SAAS,eAAe,OAAO,OAAO,WAAW,OAAO,MAAM;AACvE;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,mBAAiB,CAAC,IAAI,MAAM,GAAG,eAAe;AAC9C,QAAM,WAAW,qBAAqB,kBAAkB,OAAO,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AACnH,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,uBAAuB,SAAS;AACtC,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,WAAW,uBAAuB,IAAI,SAAS,QAAQ;AAC7D,QAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,QAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,QAAM,KAAK,OAAO,OAAO,OAAO,SAAS;AACzC,QAAM,gBAAgB,KAAK,cAAc,eAAe;AACxD,QAAM,QAAQ,SAAS,WAAW,EAAE;AACpC,WAAS,QAAQ,GAAG,QAAQ,SAAS,WAAW,EAAE,OAAO;AACvD,aAAS,UAAU,GAAG,UAAU,SAAS,YAAY,EAAE,SAAS;AAC9D,eAAS,UAAU,GAAG,UAAU,SAAS,SAAS,EAAE,SAAS;AAC3D,iBAAS,QAAQ,GAAG,QAAQ,SAAS,UAAU,EAAE,OAAO;AACtD,mBAAS,QAAQ,GAAG,QAAQ,SAAS,SAAS,EAAE,OAAO;AACrD,kBAAM,gBAAgB,UAAU;AAChC,kBAAM,cAAc,QAAQ;AAC5B,kBAAM,cAAc,QAAQ;AAC5B,gBAAI,UAAU;AACd,qBAAS,SAAS,GAAG,SAAS,sBAAsB,UAAU,eAAe;AAC3E,oBAAM,WAAW,gBAAgB,UAAU;AAC3C,kBAAI,UAAU,KAAK,WAAW,SAAS,YAAY,KAAK,MAAM,OAAO,MAAM,SAAS;AAClF;AAAA,cACF;AACA,uBAAS,OAAO,GAAG,OAAO,uBAAuB,QAAQ,gBAAgB;AACvE,sBAAM,SAAS,cAAc,QAAQ;AACrC,oBAAI,QAAQ,KAAK,SAAS,SAAS,aAAa,KAAK,MAAM,KAAK,MAAM,OAAO;AAC3E;AAAA,gBACF;AACA,yBAAS,OAAO,GAAG,OAAO,sBAAsB,QAAQ,eAAe;AACrE,wBAAM,SAAS,cAAc,QAAQ;AACrC,sBAAI,QAAQ,KAAK,SAAS,SAAS,YAAY,KAAK,MAAM,KAAK,MAAM,OAAO;AAC1E;AAAA,kBACF;AACA,wBAAM,QAAQ,MAAM,IAAI,OAAO,SAAS,OAAO,OAAO,OAAO;AAC7D,6BAAW;AAAA,gBACb;AAAA,cACF;AAAA,YACF;AACA,eAAG,IAAI,UAAU,eAAe,OAAO,SAAS,OAAO,OAAO,OAAO;AAAA,UACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,IAAI;AACV,mBAAiB,CAAC,IAAI,MAAM,GAAG,aAAa;AAC5C,QAAM,EAAE,YAAY,SAAS,KAAK,KAAK,IAAI;AAC3C,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,IAAI;AAC7F,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,QAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,QAAM,KAAK,OAAO,EAAE,OAAO,SAAS;AACpC,QAAM,gBAAgB,KAAK,eAAe;AAC1C,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,QAAQ,OAAO,GAAG,OAAO,WAAW,MAAM;AAChD,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,aAAS,IAAI,GAAG,IAAI,SAAS,YAAY,EAAE,GAAG;AAC5C,eAAS,MAAM,GAAG,MAAM,SAAS,UAAU,EAAE,KAAK;AAChD,iBAAS,MAAM,GAAG,MAAM,SAAS,SAAS,EAAE,KAAK;AAC/C,gBAAM,YAAY,MAAM;AACxB,gBAAM,YAAY,MAAM;AACxB,cAAI,UAAU;AACd,mBAAS,KAAK,GAAG,KAAK,uBAAuB,MAAM,gBAAgB;AACjE,kBAAM,OAAO,YAAY,MAAM;AAC/B,gBAAI,MAAM,KAAK,OAAO,SAAS,aAAa,KAAK,MAAM,GAAG,MAAM,KAAK;AACnE;AAAA,YACF;AACA,qBAAS,KAAK,GAAG,KAAK,sBAAsB,MAAM,eAAe;AAC/D,oBAAM,OAAO,YAAY,MAAM;AAC/B,kBAAI,MAAM,KAAK,OAAO,SAAS,YAAY,KAAK,MAAM,GAAG,MAAM,KAAK;AAClE;AAAA,cACF;AACA,oBAAM,QAAQ,MAAM,IAAI,GAAG,KAAK,KAAK,CAAC;AACtC,yBAAW;AAAA,YACb;AAAA,UACF;AACA,aAAG,IAAI,UAAU,eAAe,GAAG,KAAK,KAAK,CAAC;AAAA,QAChD;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAOI,SAAQ,QAAQ,MAAM,OAAO,SAAS,IAAI;AAC5D,eAAa,OAAO,MAAM,MAAM,WAAW,SAAS,MAAM,QAAQ,MAAM,8EAA8E;AACtJ,eAAa,OAAO,UAAU,QAAQ,MAAM,MAAM,WAAW,OAAO,MAAM,QAAQ,MAAM,4EAA4E;AACpK,eAAa,OAAOA,WAAU,QAAQ,MAAM,MAAM,WAAWA,QAAO,MAAM,QAAQ,MAAM,2EAA2E;AACnK,mBAAiB,CAAC,GAAG,OAAO,UAAUA,SAAQ,MAAM,GAAG,WAAW;AAClE,MAAI,EAAE,gBAAgB,IAAI;AAC1B,MAAI,mBAAmB,MAAM;AAC3B,sBAAkB;AAAA,EACpB;AACA,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAC9C,QAAM,UAAU,SAAS,KAAK,IAAI,SAAS,MAAM,EAAE;AACnD,QAAM,QAAQA,UAAS,SAAS,KAAK,IAAIA,QAAO,MAAM,EAAE,SAAS,IAAI,aAAa,CAAC,CAAC,CAAC;AACrF,QAAM,UAAU,SAAS,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE,SAAS,IAAI,aAAa,CAAC,CAAC,CAAC;AACvF,QAAM,UAAU,IAAI,aAAa,MAAM,MAAM;AAC7C,QAAM,gBAAgB,QAAQ;AAC9B,QAAM,cAAc,MAAM;AAC1B,QAAM,gBAAgB,QAAQ;AAC9B,QAAM,cAAc,MAAM;AAC1B,MAAI,OAAO;AACX,MAAI,KAAK;AACT,MAAI,KAAK;AACT,MAAI,KAAK;AACT,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,YAAQ,KAAK,QAAQ,WAAW,MAAM,KAAK,MAAM,SAAS,MAAM,QAAQ,KAAK,KAAK,QAAQ,QAAQ,eAAe;AACjH,QAAI,QAAQ,eAAe;AACzB,aAAO;AAAA,IACT;AACA,QAAI,MAAM,aAAa;AACrB,WAAK;AAAA,IACP;AACA,QAAI,MAAM,aAAa;AACrB,WAAK;AAAA,IACP;AACA,QAAI,MAAM,eAAe;AACvB,WAAK;AAAA,IACP;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,OAAO;AAC1D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,MAAM,IAAI;AAC9B,mBAAiB,CAAC,CAAC,GAAG,gBAAgB;AACtC,QAAM,QAAQ,WAAW,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAC/C,QAAM,WAAW,qBAAqB,YAAY,EAAE,OAAO,YAAY,KAAK;AAC5E,QAAM,WAAW,qBAAqB,YAAY,SAAS,QAAQ,WAAW,MAAM;AACpF,QAAM,mBAAmB,qBAAqB,oBAAoB,EAAE,OAAO,YAAY,KAAK;AAC5F,QAAM,mBAAmB,qBAAqB,oBAAoB,OAAO,WAAW,MAAM;AAC1F,QAAM,YAAY,qBAAqB,aAAa,kBAAkB,OAAO,WAAW,MAAM;AAC9F,QAAM,YAAY,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC3F,QAAM,cAAc,WAAW,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,SAAS,EAAE,CAAC;AACzG,QAAM,sBAAsB,SAAS,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,iBAAiB,EAAE,CAAC;AAC1H,QAAM,SAAS,OAAO;AAAA,IACpB,QAAQ,EAAE,GAAG,oBAAoB;AAAA,IACjC,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,kBAAkB,MAAM,UAAU;AAAA,EACpD,CAAC;AACD,WAAS,8BAA8B,SAAS;AAChD,WAAS,8BAA8B,WAAW;AAClD,WAAS,8BAA8B,mBAAmB;AAC1D,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,cAAc,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACtD,QAAM,UAAU,aAAa,OAAO,aAAa,QAAQ,OAAO,QAAQ,OAAO,IAAI;AACnF,SAAO,SAAS,eAAe,CAAC,IAAI,GAAG,QAAQ,OAAO,OAAO;AAC/D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,IAAI,GAAG,IAAI;AACnB,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,iBAAiB,qBAAqB,2BAA2B,MAAM,KAAK,MAAM,GAAG,MAAM,KAAK,MAAM,CAAC;AAC7G,SAAO,SAAS,eAAe,CAAC,eAAe,MAAM,GAAG,SAAS,WAAW,KAAK,cAAc,CAAC;AAClG;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe,gBAAgB,aAAa,CAAC,IAAI,UAAU;AAC7D,QAAM,YAAY;AAClB,MAAI,KAAK,UAAU,cAAc;AAC/B,WAAO,UAAU;AAAA,EACnB;AACA,SAAO,KAAK,UAAU,eAAe,UAAU,eAAe;AAChE,CAAC;AACD,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa,CAAC,SAAS;AACzB,QAAM,EAAE,EAAE,IAAI,KAAK;AACnB,QAAM,aAAa,KAAK;AACxB,QAAM,eAAe,IAAI,aAAa,aAAa,cAAc,EAAE,KAAK,CAAC;AACzE,QAAM,cAAc,WAAW,KAAK,IAAI,EAAE,MAAM;AAChD,QAAM,QAAQ,YAAY,mBAAmB;AAC7C,QAAM,QAAQ,YAAY,mBAAmB;AAC7C,QAAM,WAAW,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACnD,QAAM,WAAW,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACnD,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,UAAM,QAAQ,SAAS;AACvB,UAAM,QAAQ,SAAS;AACvB,iBAAa,KAAK,KAAK,MAAM,OAAO,KAAK;AAAA,EAC3C;AACA,SAAO,WAAW,WAAW,cAAc,EAAE,OAAO,SAAS;AAC/D;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,QAAQ,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE,mBAAmB;AAClE,QAAM,UAAU,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAChD,SAAO,SAAS,eAAe,MAAM,OAAO,MAAM,OAAO,OAAO;AAClE;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,QAAQ,aAAa,eAAe,MAAM,OAAO,GAAG,KAAK,EAAE;AACjE,QAAM,SAAS,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AACxC,uBAAqB,uBAAuB,QAAQ,KAAK;AACzD,MAAI,WAAW,qBAAqB,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,KAAK;AACrF,MAAI,aAAa,cAAc,QAAQ,MAAM,GAAG;AAC9C,WAAO,SAAS,eAAe,UAAU,OAAO,GAAG,OAAO,CAAC,CAAC;AAAA,EAC9D;AACA,QAAM,UAAU,OAAO,OAAO,CAAC,MAAM,aAAa,cAAc,EAAE,KAAK,IAAI,CAAC;AAC5E,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,GAAG,SAAS,SAAS,CAAC;AAAA,EACnE;AACA,MAAI,QAAQ,GAAG,UAAU,aAAa;AACpC,UAAM,QAAQ,QAAQ,IAAI,CAAC,MAAM,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC,CAAC;AACnF,UAAM,QAAQ,QAAQ,IAAI,CAAC,MAAM,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC,CAAC;AACnF,UAAM,eAAe,QAAQ,EAAE,QAAQ,OAAO,SAAS,UAAU,OAAO,EAAE,MAAM,MAAM,EAAE,CAAC;AACzF,UAAM,eAAe,QAAQ,EAAE,QAAQ,OAAO,SAAS,UAAU,OAAO,EAAE,MAAM,MAAM,EAAE,CAAC;AACzF,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,cAAc,MAAM,aAAa,GAAG,SAAS,SAAS,CAAC;AACjG,UAAM,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAC9D,UAAM,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAC9D,aAAS,8BAA8B,YAAY;AACnD,aAAS,8BAA8B,YAAY;AACnD,WAAO;AAAA,EACT;AACA,QAAM,WAAW,QAAQ,IAAI,CAAC,MAAM;AAClC,UAAM,YAAY,aAAa,cAAc,EAAE,MAAM,MAAM,KAAK,CAAC;AACjE,UAAM,QAAQ,CAAC,IAAI,SAAS;AAC5B,WAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AAAA,EAC3E,CAAC;AACD,QAAM,kBAAkB,SAAS,IAAI,CAAC,MAAM;AAC1C,WAAO,EAAE,MAAM,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE,QAAQ,OAAO,EAAE,MAAM;AAAA,EACpE,CAAC;AACD,aAAW,qBAAqB,gBAAgB,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAC/E,QAAM,eAAe,SAAS,GAAG,MAAM,OAAO;AAC9C,QAAM,UAAU,WAAW,iBAAiB,UAAU,OAAO,GAAG,OAAO,YAAY;AACnF,QAAM,gBAAgB,qBAAqB,gBAAgB,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,KAAK;AAC7F,QAAM,UAAU,SAAS,eAAe,eAAe,OAAO,GAAG,OAAO,OAAO;AAC/E,WAAS,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AACjE,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB,IAAI;AACvE,mBAAiB,CAAC,GAAG,MAAM,GAAG,QAAQ;AACtC,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,OAAO,WAAW;AAC5I,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,iBAAiB,SAAS,eAAe;AAC/C,QAAM,IAAI,IAAI,aAAa,SAAS,UAAU,EAAE,KAAK;AACrD,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,eAAe,SAAS;AAC9B,QAAM,aAAa,iBAAiB,SAAS,KAAK,SAAS;AAC3D,QAAM,aAAa,iBAAiB,SAAS,KAAK;AAClD,QAAM,iBAAiB,iBAAiB,IAAI,SAAS;AACrD,QAAM,eAAe,EAAE,QAAQ;AAC/B,QAAM,aAAa,iBAAiB,EAAE,QAAQ,KAAK,EAAE,QAAQ;AAC7D,QAAM,aAAa,iBAAiB,EAAE,QAAQ,KAAK;AACnD,QAAM,iBAAiB,iBAAiB,IAAI,EAAE,QAAQ;AACtD,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAC/C,QAAM,QAAQ,EAAE;AAChB,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,UAAM,WAAW,IAAI;AACrB,UAAM,WAAW,IAAI;AACrB,aAAS,KAAK,GAAG,KAAK,SAAS,WAAW,EAAE,IAAI;AAC9C,YAAM,WAAW,WAAW,KAAK;AACjC,YAAM,WAAW,KAAK,SAAS,eAAe;AAC9C,eAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,cAAM,KAAK,WAAW,KAAK;AAC3B,YAAI,KAAK,KAAK,MAAM,SAAS,UAAU;AACrC;AAAA,QACF;AACA,cAAM,WAAW,KAAK,cAAc;AACpC,cAAM,WAAW,WAAW,KAAK;AACjC,iBAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,gBAAM,WAAW,WAAW,KAAK;AACjC,gBAAM,WAAW,KAAK,SAAS,cAAc;AAC7C,mBAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,kBAAM,KAAK,WAAW,KAAK;AAC3B,gBAAI,KAAK,KAAK,MAAM,SAAS,SAAS;AACpC;AAAA,YACF;AACA,kBAAM,WAAW,WAAW,KAAK,cAAc;AAC/C,kBAAM,WAAW,WAAW,KAAK;AACjC,gBAAI,WAAW;AACf,qBAAS,KAAK,GAAG,KAAK,SAAS,YAAY,EAAE,IAAI;AAC/C,oBAAM,OAAO,MAAM,WAAW,KAAK;AACnC,uBAAS,KAAK,GAAG,KAAK,SAAS,aAAa,EAAE,IAAI;AAChD,sBAAM,WAAW,KAAK,mBAAmB,OAAO,MAAM,WAAW;AAAA,cACnE;AACA,0BAAY,SAAS;AAAA,YACvB;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,KAAK;AACxD;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,iBAAiB,YAAY,IAAI;AACzE,mBAAiB,CAAC,GAAG,EAAE,GAAG,sBAAsB;AAChD,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,GAAG,MAAM,iBAAiB,OAAO,WAAW;AACnI,QAAM,EAAE,cAAc,aAAa,cAAc,YAAY,IAAI;AACjE,QAAM,iBAAiB,SAAS,eAAe;AAC/C,QAAM,KAAK,IAAI,aAAa,SAAS,aAAa,SAAS;AAC3D,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,OAAO,IAAI,aAAa,EAAE,OAAO,EAAE,OAAO,KAAK;AACrD,QAAM,QAAQ,IAAI,aAAa,GAAG,OAAO,GAAG,OAAO,MAAM;AACzD,WAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,UAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,SAAS,MAAM,YAAY,CAAC;AACjE,UAAM,QAAQ,KAAK,IAAI,SAAS,YAAY,SAAS,WAAW,SAAS,MAAM,YAAY;AAC3F,aAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,YAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,UAAU,MAAM,WAAW,CAAC;AACjE,YAAM,QAAQ,KAAK,IAAI,SAAS,WAAW,SAAS,UAAU,UAAU,MAAM,WAAW;AACzF,eAAS,KAAK,GAAG,KAAK,SAAS,YAAY,EAAE,IAAI;AAC/C,iBAAS,KAAK,GAAG,KAAK,SAAS,aAAa,EAAE,IAAI;AAChD,cAAI,UAAU;AACd,mBAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,qBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,oBAAM,KAAK,KAAK,KAAK,eAAe;AACpC,uBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,sBAAM,KAAK,KAAK,KAAK,cAAc;AACnC,oBAAI,gBAAgB;AAClB,6BAAW,KAAK,IAAI,GAAG,IAAI,IAAI,EAAE,IAAI,MAAM,IAAI,GAAG,IAAI,IAAI,EAAE;AAAA,gBAC9D,OAAO;AACL,6BAAW,KAAK,IAAI,GAAG,IAAI,IAAI,EAAE,IAAI,MAAM,IAAI,GAAG,IAAI,IAAI,EAAE;AAAA,gBAC9D;AAAA,cACF;AAAA,YACF;AAAA,UACF;AACA,aAAG,IAAI,SAAS,IAAI,IAAI,IAAI,EAAE;AAAA,QAChC;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,YAAY,gBAAgB,IAAI;AACxE,mBAAiB,CAAC,IAAI,MAAM,GAAG,qBAAqB;AACpD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,MAAI,cAAc,qBAAqB,wBAAwB,UAAU;AACzE,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,GAAG,MAAM,iBAAiB,OAAO,WAAW;AACvI,QAAM,KAAK,IAAI,aAAa,SAAS,SAAS,SAAS;AACvD,QAAM,WAAW,GAAG;AACpB,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,YAAY,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACnD,QAAM,CAAC,OAAO,OAAO,KAAK,IAAI;AAC9B,QAAM,EAAE,WAAW,cAAc,aAAa,YAAY,UAAU,SAAS,aAAa,WAAW,UAAU,cAAc,YAAY,IAAI;AAC7I,gBAAc,SAAS;AACvB,QAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,QAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,QAAM,iBAAiB,gBAAgB;AACvC,QAAM,eAAe,GAAG,QAAQ;AAChC,QAAM,aAAa,iBAAiB,GAAG,QAAQ,KAAK,GAAG,QAAQ;AAC/D,QAAM,aAAa,iBAAiB,GAAG,QAAQ,KAAK;AACpD,QAAM,iBAAiB,iBAAiB,IAAI,GAAG,QAAQ;AACvD,QAAM,eAAe,UAAU;AAC/B,QAAM,aAAa,iBAAiB,UAAU,KAAK,UAAU;AAC7D,QAAM,aAAa,iBAAiB,UAAU,KAAK;AACnD,QAAM,iBAAiB,iBAAiB,IAAI,UAAU;AACtD,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAS,KAAK,GAAG,KAAK,YAAY,EAAE,IAAI;AACtC,eAAS,KAAK,GAAG,KAAK,UAAU,EAAE,IAAI;AACpC,cAAM,WAAW,KAAK;AACtB,cAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,YAAY,CAAC;AAC5D,cAAM,QAAQ,KAAK,IAAI,YAAY,eAAe,YAAY,YAAY;AAC1E,iBAAS,KAAK,GAAG,KAAK,SAAS,EAAE,IAAI;AACnC,gBAAM,WAAW,KAAK;AACtB,gBAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,WAAW,CAAC;AAC3D,gBAAM,QAAQ,KAAK,IAAI,WAAW,cAAc,YAAY,WAAW;AACvE,cAAI,UAAU;AACd,mBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,kBAAM,KAAK,KAAK,eAAe;AAC/B,qBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,oBAAM,KAAK,KAAK,cAAc;AAC9B,oBAAM,WAAW,eAAe,IAAI,aAAa,KAAK,aAAa;AACnE,oBAAM,YAAY,SAAS,eAAe,IAAI,MAAM,SAAS,cAAc,IAAI,MAAM,QAAQ;AAC7F,uBAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,sBAAM,QAAQ,SAAS,WAAW,iBAAiB;AACnD,sBAAM,SAAS,UAAU,YAAY;AACrC,2BAAW,QAAQ;AAAA,cACrB;AAAA,YACF;AAAA,UACF;AACA,gBAAM,WAAW,eAAe,IAAI,aAAa,KAAK,aAAa,KAAK,iBAAiB;AACzF,mBAAS,YAAY;AAAA,QACvB;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,mBAAiB,CAAC,GAAG,MAAM,GAAG,QAAQ;AACtC,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,IAAI;AACvG,QAAM,EAAE,aAAa,cAAc,aAAa,eAAe,gBAAgB,eAAe,QAAQ,IAAI;AAC1G,QAAM,WAAW,QAAQ;AACzB,QAAM,UAAU,QAAQ;AACxB,QAAM,SAAS,QAAQ;AACvB,QAAM,IAAI,IAAI,aAAa,SAAS,UAAU,EAAE,KAAK;AACrD,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAC/C,QAAM,QAAQ,EAAE;AAChB,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,UAAM,WAAW,IAAI,SAAS;AAC9B,UAAM,WAAW,IAAI,EAAE,QAAQ;AAC/B,aAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,YAAM,WAAW,WAAW,KAAK,EAAE,QAAQ;AAC3C,YAAM,WAAW,KAAK,SAAS,cAAc;AAC7C,eAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,cAAM,KAAK,WAAW,KAAK;AAC3B,YAAI,KAAK,KAAK,MAAM,SAAS,SAAS;AACpC;AAAA,QACF;AACA,cAAM,WAAW,KAAK,cAAc;AACpC,cAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,iBAAS,KAAK,GAAG,KAAK,SAAS,WAAW,EAAE,IAAI;AAC9C,gBAAM,WAAW,WAAW,KAAK,EAAE,QAAQ;AAC3C,gBAAM,WAAW,KAAK,SAAS,eAAe;AAC9C,mBAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,kBAAM,KAAK,WAAW,KAAK;AAC3B,gBAAI,KAAK,KAAK,MAAM,SAAS,UAAU;AACrC;AAAA,YACF;AACA,kBAAM,WAAW,WAAW,KAAK,cAAc;AAC/C,kBAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,qBAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,oBAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,oBAAM,WAAW,KAAK,SAAS,cAAc;AAC7C,uBAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,sBAAM,KAAK,WAAW,KAAK;AAC3B,oBAAI,KAAK,KAAK,MAAM,SAAS,SAAS;AACpC;AAAA,gBACF;AACA,sBAAM,WAAW,WAAW,KAAK,cAAc;AAC/C,sBAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,oBAAI,WAAW;AACf,yBAAS,KAAK,GAAG,KAAK,SAAS,YAAY,EAAE,IAAI;AAC/C,wBAAM,OAAO,MAAM,WAAW;AAC9B,2BAAS,KAAK,GAAG,KAAK,SAAS,aAAa,EAAE,IAAI;AAChD,0BAAM,WAAW,OAAO,OAAO,MAAM,WAAW;AAAA,kBAClD;AACA,8BAAY,SAAS;AAAA,gBACvB;AAAA,cACF;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM;AAC3D;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,uBAAuB,MAAM;AACpC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,IAAI;AAC5C,mBAAiB,CAAC,GAAG,EAAE,GAAG,wBAAwB;AAClD,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,GAAG,IAAI;AAC9F,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,KAAK,IAAI,aAAa,SAAS,aAAa,SAAS;AAC3D,QAAM,WAAW,GAAG;AACpB,QAAM,CAAC,MAAM,MAAM,MAAM,IAAI,IAAI,GAAG;AACpC,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,CAAC,MAAM,MAAM,MAAM,IAAI,IAAI;AACjC,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,CAAC,KAAK,KAAK,KAAK,GAAG,IAAI;AAC7B,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,SAAS,SAAS,QAAQ;AAChC,WAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,UAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,WAAW,MAAM,WAAW,CAAC;AAClE,UAAM,QAAQ,KAAK,IAAI,SAAS,WAAW,SAAS,UAAU,WAAW,MAAM,WAAW;AAC1F,UAAM,WAAW,KAAK;AACtB,aAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,YAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,SAAS,MAAM,YAAY,CAAC;AACjE,YAAM,QAAQ,KAAK,IAAI,SAAS,YAAY,SAAS,WAAW,SAAS,MAAM,YAAY;AAC3F,YAAM,WAAW,KAAK,OAAO;AAC7B,eAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,cAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,UAAU,MAAM,WAAW,CAAC;AACjE,cAAM,QAAQ,KAAK,IAAI,SAAS,WAAW,SAAS,UAAU,UAAU,MAAM,WAAW;AACzF,cAAM,WAAW,KAAK,OAAO;AAC7B,iBAAS,KAAK,GAAG,KAAK,SAAS,YAAY,EAAE,IAAI;AAC/C,gBAAM,WAAW,KAAK,OAAO;AAC7B,mBAAS,KAAK,GAAG,KAAK,SAAS,aAAa,EAAE,IAAI;AAChD,gBAAI,UAAU;AACd,qBAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,oBAAM,WAAW,IAAI;AACrB,oBAAM,WAAW,IAAI;AACrB,uBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,sBAAM,KAAK,KAAK,KAAK,cAAc;AACnC,sBAAM,WAAW,KAAK,MAAM;AAC5B,sBAAM,WAAW,KAAK,OAAO;AAC7B,yBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,wBAAM,KAAK,KAAK,KAAK,eAAe;AACpC,wBAAM,WAAW,KAAK,MAAM;AAC5B,wBAAM,WAAW,KAAK,OAAO;AAC7B,2BAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,0BAAM,KAAK,KAAK,KAAK,cAAc;AACnC,0BAAM,WAAW,KAAK,MAAM;AAC5B,0BAAM,WAAW,KAAK,OAAO;AAC7B,+BAAW,QAAQ,WAAW,MAAM,SAAS,WAAW;AAAA,kBAC1D;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,qBAAS,WAAW,MAAM;AAAA,UAC5B;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,KAAK,MAAM,SAAS,WAAW,IAAI;AAC3C,mBAAiB,CAAC,EAAE,GAAG,uBAAuB;AAC9C,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,GAAG,IAAI;AAClG,QAAM,KAAK,IAAI,aAAa,SAAS,SAAS,SAAS;AACvD,QAAM,WAAW,GAAG;AACpB,QAAM,CAAC,MAAM,MAAM,MAAM,IAAI,IAAI,GAAG;AACpC,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,CAAC,MAAM,MAAM,MAAM,IAAI,IAAI;AACjC,QAAM,YAAY,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACnD,QAAM,CAAC,OAAO,OAAO,OAAO,KAAK,IAAI;AACrC,QAAM,EAAE,WAAW,aAAa,cAAc,aAAa,YAAY,SAAS,UAAU,SAAS,aAAa,UAAU,WAAW,UAAU,aAAa,cAAc,YAAY,IAAI;AAC1L,QAAM,WAAW,cAAc,IAAI,SAAS,QAAQ;AACpD,QAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,QAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAS,KAAK,GAAG,KAAK,YAAY,EAAE,IAAI;AACtC,eAAS,KAAK,GAAG,KAAK,SAAS,EAAE,IAAI;AACnC,cAAM,WAAW,KAAK;AACtB,cAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,WAAW,CAAC;AAC3D,cAAM,QAAQ,KAAK,IAAI,WAAW,cAAc,YAAY,WAAW;AACvE,iBAAS,KAAK,GAAG,KAAK,UAAU,EAAE,IAAI;AACpC,gBAAM,WAAW,KAAK;AACtB,gBAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,YAAY,CAAC;AAC5D,gBAAM,QAAQ,KAAK,IAAI,YAAY,eAAe,YAAY,YAAY;AAC1E,mBAAS,KAAK,GAAG,KAAK,SAAS,EAAE,IAAI;AACnC,kBAAM,WAAW,KAAK;AACtB,kBAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,WAAW,CAAC;AAC3D,kBAAM,QAAQ,KAAK,IAAI,WAAW,cAAc,YAAY,WAAW;AACvE,gBAAI,UAAU;AACd,qBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,oBAAM,KAAK,KAAK,cAAc;AAC9B,uBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,sBAAM,KAAK,KAAK,eAAe;AAC/B,yBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,wBAAM,KAAK,KAAK,cAAc;AAC9B,wBAAM,WAAW,OAAO,IAAI,OAAO,KAAK,OAAO,KAAK,OAAO;AAC3D,wBAAM,YAAY,SAAS,cAAc,IAAI,MAAM,SAAS,eAAe,IAAI,MAAM,SAAS,cAAc,IAAI,MAAM,QAAQ;AAC9H,2BAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,0BAAM,QAAQ,SAAS,WAAW;AAClC,0BAAM,SAAS,UAAU,YAAY;AACrC,+BAAW,QAAQ;AAAA,kBACrB;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,qBAAS,OAAO,IAAI,OAAO,KAAK,OAAO,KAAK,OAAO,KAAK,MAAM;AAAA,UAChE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,gBAAgB,KAAK,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;AACpD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,OAAO,OAAO,IAAI;AACzC,QAAM,EAAE,UAAU,QAAQ,mBAAmB,IAAI;AACjD,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,QAAM,WAAW,MAAM,MAAM;AAC7B,QAAM,CAAC,YAAY,SAAS,IAAI;AAChC,QAAM,SAAS,OAAO,CAAC,UAAU,YAAY,WAAW,WAAW,GAAG,SAAS;AAC/E,QAAM,UAAU,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAChD,QAAM,aAAa,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACpD,QAAM,YAAY,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACnD,QAAM,WAAW,aAAa,eAAe,OAAO,KAAK;AACzD,QAAM,YAAY,aAAa,eAAe,OAAO,KAAK;AAC1D,WAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,UAAM,WAAW,IAAI;AACrB,UAAM,KAAK,QAAQ;AACnB,UAAM,KAAK,QAAQ,WAAW;AAC9B,UAAM,KAAK,QAAQ,WAAW;AAC9B,UAAM,KAAK,QAAQ,WAAW;AAC9B,UAAM,OAAO,WAAW;AACxB,QAAI,QAAQ,OAAO;AACjB;AAAA,IACF;AACA,UAAM,cAAc,aAAa,KAAK,KAAK,OAAO,cAAc,MAAM,aAAa,KAAK;AACxF,UAAM,aAAa,YAAY,KAAK,KAAK,OAAO,aAAa,MAAM,YAAY,KAAK;AACpF,aAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,YAAM,OAAO,aAAa,IAAI,MAAM,cAAc,KAAK,IAAI,cAAc,OAAO,KAAK,OAAO,cAAc;AAC1G,UAAI,OAAO,KAAK,OAAO,cAAc,GAAG;AACtC,iBAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,mBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAM,MAAM,IAAI,IAAI,UAAU,KAAK,IAAI,UAAU,KAAK,IAAI,UAAU;AACpE,mBAAO,OAAO,OAAO;AAAA,UACvB;AAAA,QACF;AACA;AAAA,MACF;AACA,UAAI,WAAW,YAAY;AACzB,cAAM,SAAS,KAAK,MAAM,IAAI;AAC9B,cAAM,YAAY,KAAK,KAAK,IAAI;AAChC,cAAM,QAAQ,OAAO;AACrB,iBAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,gBAAM,OAAO,YAAY,IAAI,MAAM,aAAa,KAAK,IAAI,aAAa,OAAO,KAAK,OAAO,aAAa;AACtG,cAAI,OAAO,KAAK,OAAO,aAAa,GAAG;AACrC,qBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,oBAAM,MAAM,IAAI,IAAI,UAAU,KAAK,IAAI,UAAU,KAAK,IAAI,UAAU;AACpE,qBAAO,OAAO,OAAO;AAAA,YACvB;AACA;AAAA,UACF;AACA,gBAAM,UAAU,KAAK,MAAM,IAAI;AAC/B,gBAAM,WAAW,KAAK,KAAK,IAAI;AAC/B,gBAAM,QAAQ,OAAO;AACrB,mBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,gBAAI,MAAM,IAAI,UAAU,SAAS,KAAK,SAAS,SAAS,KAAK,OAAO,SAAS;AAC7E,kBAAM,UAAU,UAAU;AAC1B,kBAAM,IAAI,WAAW,SAAS,KAAK,SAAS,SAAS,KAAK,OAAO,SAAS;AAC1E,kBAAM,WAAW,UAAU;AAC3B,kBAAM,IAAI,UAAU,SAAS,KAAK,YAAY,SAAS,KAAK,OAAO,SAAS;AAC5E,kBAAM,aAAa,UAAU;AAC7B,kBAAM,IAAI,WAAW,SAAS,KAAK,YAAY,SAAS,KAAK,OAAO,SAAS;AAC7E,kBAAM,cAAc,UAAU;AAC9B,kBAAM,MAAM,WAAW,WAAW,WAAW;AAC7C,kBAAM,SAAS,cAAc,cAAc,cAAc;AACzD,kBAAM,IAAI,IAAI,UAAU,KAAK,IAAI,UAAU,KAAK,IAAI,UAAU;AAC9D,mBAAO,OAAO,OAAO,OAAO,SAAS,OAAO;AAAA,UAC9C;AAAA,QACF;AAAA,MACF,OAAO;AACL,iBAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,gBAAM,OAAO,YAAY,IAAI,MAAM,aAAa,KAAK,IAAI,aAAa,OAAO,KAAK,OAAO,aAAa;AACtG,cAAI,OAAO,KAAK,OAAO,aAAa,GAAG;AACrC,qBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,oBAAM,MAAM,IAAI,IAAI,UAAU,KAAK,IAAI,UAAU,KAAK,IAAI,UAAU;AACpE,qBAAO,OAAO,OAAO;AAAA,YACvB;AACA;AAAA,UACF;AACA,gBAAM,WAAW,KAAK,MAAM,IAAI;AAChC,gBAAM,WAAW,KAAK,MAAM,IAAI;AAChC,mBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAM,QAAQ,IAAI,WAAW,SAAS,KAAK,WAAW,SAAS,KAAK,OAAO,SAAS;AACpF,kBAAM,SAAS,IAAI,IAAI,UAAU,KAAK,IAAI,UAAU,KAAK,IAAI,UAAU;AACvE,mBAAO,OAAO,UAAU,UAAU;AAAA,UACpC;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC1E;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,mBAAiB,GAAG,SAAS;AAC7B,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,EAAE,MAAM,MAAM;AAClF,MAAI,KAAK;AACT,MAAI,eAAe,MAAM;AACvB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AAAA,EACpF;AACA,QAAM,eAAe,qBAAqB,iBAAiB,GAAG,EAAE,MAAM,MAAM,EAAE;AAC9E,MAAI,iBAAiB,GAAG,MAAM,SAAS,GAAG;AACxC,UAAM,IAAI,MAAM,qDAAqD,GAAG,MAAM,SAAS,kBAAkB,cAAc;AAAA,EACzH;AACA,QAAM,cAAc,WAAW,GAAG,OAAO,OAAO;AAChD,QAAM,OAAO,aAAa,mBAAmB,aAAa,cAAc,GAAG,KAAK,GAAG,WAAW;AAC9F,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,QAAM,WAAW,GAAG,MAAM,GAAG,MAAM,SAAS;AAC5C,QAAM,gBAAgB,WAAW,CAAC,GAAG,MAAM,IAAI,WAAW,IAAI,IAAI,CAAC,GAAG,MAAM,IAAI;AAChF,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,UAAU;AAC/C,aAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,YAAM,MAAM,cAAc,GAAG,CAAC;AAC9B,UAAI,MAAM,GAAG;AACX,aAAK,OAAO,YAAY,IAAI,MAAM;AAAA,MACpC,OAAO;AACL,cAAM,UAAU,cAAc,GAAG,IAAI,CAAC;AACtC,aAAK,OAAO,YAAY,MAAM,WAAW,KAAK,WAAW,MAAM,OAAO,KAAK;AAAA,MAC7E;AAAA,IACF;AAAA,EACF;AACA,QAAM,SAAS,SAAS,eAAe,GAAG,OAAO,aAAa,IAAI;AAClE,MAAI,eAAe,MAAM;AACvB,UAAM,qBAAqB,qBAAqB,uBAAuB,WAAW;AAClF,UAAM,0BAA0B,WAAW,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,mBAAmB,EAAE,CAAC;AAC5H,aAAS,8BAA8B,MAAM;AAC7C,aAAS,8BAA8B,EAAE;AACzC,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,mBAAiB,GAAG,QAAQ;AAC5B,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,EAAE,MAAM,MAAM;AAClF,MAAI,KAAK;AACT,MAAI,eAAe,MAAM;AACvB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AAAA,EACpF;AACA,QAAM,eAAe,qBAAqB,iBAAiB,GAAG,EAAE,MAAM,MAAM,EAAE;AAC9E,MAAI,iBAAiB,GAAG,MAAM,SAAS,GAAG;AACxC,UAAM,IAAI,MAAM,oDAAoD,GAAG,MAAM,SAAS,kBAAkB,cAAc;AAAA,EACxH;AACA,QAAM,cAAc,WAAW,GAAG,OAAO,OAAO;AAChD,QAAM,OAAO,aAAa,oBAAoB,aAAa,cAAc,GAAG,KAAK,GAAG,WAAW;AAC/F,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,QAAM,WAAW,GAAG,MAAM,GAAG,MAAM,SAAS;AAC5C,QAAM,gBAAgB,WAAW,CAAC,GAAG,MAAM,IAAI,WAAW,IAAI,IAAI,CAAC,GAAG,MAAM,IAAI;AAChF,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,UAAU;AAC/C,aAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,YAAM,MAAM,cAAc,GAAG,CAAC;AAC9B,UAAI,MAAM,GAAG;AACX,aAAK,OAAO,YAAY,IAAI,MAAM;AAAA,MACpC,OAAO;AACL,cAAM,UAAU,cAAc,GAAG,IAAI,CAAC;AACtC,aAAK,OAAO,YAAY,MAAM,WAAW,KAAK,WAAW,MAAM,OAAO,KAAK;AAAA,MAC7E;AAAA,IACF;AAAA,EACF;AACA,QAAM,SAAS,SAAS,eAAe,GAAG,OAAO,aAAa,IAAI;AAClE,MAAI,eAAe,MAAM;AACvB,UAAM,qBAAqB,qBAAqB,uBAAuB,WAAW;AAClF,UAAM,0BAA0B,WAAW,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,mBAAmB,EAAE,CAAC;AAC5H,aAAS,8BAA8B,MAAM;AAC7C,aAAS,8BAA8B,EAAE;AACzC,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,MAAM,aAAa,IAAI;AAC/B,MAAI,EAAE,MAAM,WAAW,GAAG;AACxB,UAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,UAAM,cAAc,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACtD,UAAM,UAAU,aAAa,OAAO,aAAa,QAAQ,OAAO,QAAQ,OAAO,IAAI;AACnF,WAAO,SAAS,eAAe,CAAC,IAAI,GAAG,QAAQ,OAAO,OAAO;AAAA,EAC/D,WAAW,EAAE,MAAM,WAAW,GAAG;AAC/B,UAAM,OAAO,SAAS,WAAW,CAAC;AAClC,UAAM,aAAa,SAAS,WAAW,OAAO;AAC9C,UAAM,SAAS,mBAAmB,MAAM,YAAY,MAAM,YAAY;AACtE,WAAO,SAAS,eAAe,OAAO,OAAO,QAAQ,OAAO,OAAO,MAAM;AAAA,EAC3E;AACA,QAAM,IAAI,MAAM,qEAAqE,EAAE,MAAM,SAAS;AACxG;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,WAAW,WAAW,IAAI;AAClC,eAAa,OAAO,eAAe,QAAQ,MAAM,+DAA+D,YAAY;AAC5H,QAAM,YAAY,EAAE,MAAM;AAC1B,QAAM,cAAc,EAAE,MAAM;AAC5B,QAAM,aAAa,EAAE,MAAM;AAC3B,QAAM,aAAa,EAAE,MAAM;AAC3B,QAAM,eAAe,cAAc;AACnC,QAAM,cAAc,aAAa;AACjC,QAAM,cAAc,cAAc,YAAY;AAC9C,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,SAAS,IAAI,aAAa,YAAY,eAAe,cAAc,WAAW;AACpF,MAAI,YAAY;AAChB,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,YAAM,MAAM,KAAK,MAAM,IAAI,SAAS;AACpC,YAAM,UAAU,IAAI;AACpB,eAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,cAAM,MAAM,KAAK,MAAM,IAAI,SAAS;AACpC,cAAM,UAAU,IAAI;AACpB,cAAM,WAAW,UAAU,YAAY,WAAW;AAClD,iBAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,gBAAM,MAAM,IAAI;AAChB,gBAAM,WAAW,MAAM,cAAc,MAAM,cAAc,MAAM,cAAc;AAC7E,iBAAO,eAAe,QAAQ;AAAA,QAChC;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,CAAC,WAAW,cAAc,aAAa,WAAW,GAAG,EAAE,OAAO,MAAM;AACrG;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,WAAW,gBAAgB,IAAI;AAC3D,mBAAiB,CAAC,GAAG,MAAM,GAAG,uBAAuB;AACrD,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,MAAI,aAAa;AACjB,MAAI,cAAc,MAAM;AACtB,iBAAa,CAAC,GAAG,CAAC;AAAA,EACpB;AACA,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,UAAU,GAAG,MAAM,gFAAgF,0BAA0B,aAAa;AAC3N,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,YAAY,MAAM,iBAAiB,IAAI;AAC/H,QAAM,EAAE,cAAc,aAAa,gBAAgB,eAAe,QAAQ,IAAI;AAC9E,QAAM,UAAU,QAAQ;AACxB,QAAM,SAAS,QAAQ;AACvB,QAAM,QAAQ,SAAS,cAAc,SAAS;AAC9C,QAAM,IAAI,IAAI,aAAa,SAAS,UAAU,EAAE,KAAK;AACrD,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAC/C,QAAM,QAAQ,EAAE;AAChB,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,UAAM,WAAW,IAAI,SAAS;AAC9B,UAAM,WAAW,IAAI,EAAE,QAAQ;AAC/B,aAAS,KAAK,GAAG,KAAK,SAAS,WAAW,EAAE,IAAI;AAC9C,YAAM,WAAW,WAAW,KAAK,EAAE,QAAQ;AAC3C,YAAM,WAAW,KAAK,SAAS,eAAe;AAC9C,eAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,cAAM,KAAK,WAAW,KAAK;AAC3B,YAAI,KAAK,KAAK,MAAM,SAAS,UAAU;AACrC;AAAA,QACF;AACA,cAAM,WAAW,KAAK,cAAc;AACpC,cAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,iBAAS,KAAK,GAAG,KAAK,SAAS,UAAU,EAAE,IAAI;AAC7C,gBAAM,WAAW,WAAW,KAAK,EAAE,QAAQ;AAC3C,gBAAM,WAAW,KAAK,SAAS,cAAc;AAC7C,mBAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,kBAAM,KAAK,WAAW,KAAK;AAC3B,gBAAI,KAAK,KAAK,MAAM,SAAS,SAAS;AACpC;AAAA,YACF;AACA,kBAAM,WAAW,WAAW,KAAK,cAAc;AAC/C,kBAAM,WAAW,WAAW,KAAK,SAAS;AAC1C,gBAAI,WAAW;AACf,gBAAI,WAAW;AACf,qBAAS,KAAK,GAAG,KAAK,SAAS,YAAY,EAAE,IAAI;AAC/C,oBAAM,OAAO,MAAM,WAAW;AAC9B,uBAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,sBAAM,WAAW,MAAM,OAAO,MAAM,WAAW;AAAA,cACjD;AACA,0BAAY;AACZ,0BAAY;AAAA,YACd;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM;AAC3D;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qCAAqC,MAAM;AAClD,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,iBAAiB,YAAY,IAAI;AACxE,mBAAiB,CAAC,GAAG,EAAE,GAAG,qCAAqC;AAC/D,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,WAAW,MAAM,iBAAiB,IAAI;AAC7H,QAAM,EAAE,cAAc,aAAa,cAAc,YAAY,IAAI;AACjE,QAAM,KAAK,IAAI,aAAa,SAAS,aAAa,SAAS;AAC3D,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,QAAQ,SAAS,cAAc,SAAS;AAC9C,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,OAAO,IAAI,aAAa,EAAE,OAAO,EAAE,OAAO,KAAK;AACrD,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,QAAQ,IAAI,aAAa,GAAG,OAAO,GAAG,OAAO,MAAM;AACzD,WAAS,KAAK,GAAG,KAAK,cAAc,EAAE,IAAI;AACxC,UAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,SAAS,MAAM,YAAY,CAAC;AACjE,UAAM,QAAQ,KAAK,IAAI,SAAS,YAAY,SAAS,WAAW,SAAS,MAAM,YAAY;AAC3F,aAAS,KAAK,GAAG,KAAK,aAAa,EAAE,IAAI;AACvC,YAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,MAAM,UAAU,MAAM,WAAW,CAAC;AACjE,YAAM,QAAQ,KAAK,IAAI,SAAS,WAAW,SAAS,UAAU,UAAU,MAAM,WAAW;AACzF,eAAS,KAAK,GAAG,KAAK,SAAS,aAAa,EAAE,IAAI;AAChD,cAAM,KAAK,KAAK,MAAM,KAAK,KAAK;AAChC,cAAM,KAAK,KAAK;AAChB,YAAI,UAAU;AACd,iBAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,mBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,kBAAM,KAAK,KAAK,KAAK,eAAe;AACpC,qBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,oBAAM,KAAK,KAAK,KAAK,cAAc;AACnC,yBAAW,KAAK,IAAI,GAAG,IAAI,IAAI,EAAE,IAAI,MAAM,IAAI,GAAG,IAAI,IAAI,EAAE;AAAA,YAC9D;AAAA,UACF;AAAA,QACF;AACA,WAAG,IAAI,SAAS,IAAI,IAAI,IAAI,EAAE;AAAA,MAChC;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,4CAA4C;AAAA,EAC9C,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,oCAAoC,MAAM;AACjD,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACvE,mBAAiB,CAAC,IAAI,MAAM,GAAG,oCAAoC;AACnE,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,IAAI;AACjI,QAAM,KAAK,IAAI,aAAa,SAAS,SAAS,SAAS;AACvD,QAAM,WAAW,GAAG;AACpB,QAAM,CAAC,MAAM,MAAM,IAAI,IAAI,GAAG;AAC9B,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,CAAC,MAAM,MAAM,IAAI,IAAI;AAC3B,QAAM,YAAY,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACnD,QAAM,CAAC,OAAO,OAAO,KAAK,IAAI;AAC9B,QAAM,EAAE,WAAW,cAAc,aAAa,YAAY,UAAU,SAAS,aAAa,WAAW,UAAU,cAAc,YAAY,IAAI;AAC7I,QAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,QAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,QAAM,QAAQ,cAAc;AAC5B,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAS,KAAK,GAAG,KAAK,YAAY,EAAE,IAAI;AACtC,eAAS,KAAK,GAAG,KAAK,UAAU,EAAE,IAAI;AACpC,cAAM,WAAW,KAAK;AACtB,cAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,YAAY,CAAC;AAC5D,cAAM,QAAQ,KAAK,IAAI,YAAY,eAAe,YAAY,YAAY;AAC1E,iBAAS,KAAK,GAAG,KAAK,SAAS,EAAE,IAAI;AACnC,gBAAM,WAAW,KAAK;AACtB,gBAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,KAAK,WAAW,WAAW,CAAC;AAC3D,gBAAM,QAAQ,KAAK,IAAI,WAAW,cAAc,YAAY,WAAW;AACvE,cAAI,UAAU;AACd,mBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,kBAAM,KAAK,KAAK,eAAe;AAC/B,qBAAS,KAAK,OAAO,KAAK,OAAO,EAAE,IAAI;AACrC,oBAAM,KAAK,KAAK,cAAc;AAC9B,oBAAM,WAAW,OAAO,IAAI,OAAO,KAAK,OAAO;AAC/C,oBAAM,YAAY,SAAS,eAAe,IAAI,MAAM,SAAS,cAAc,IAAI,MAAM,QAAQ;AAC7F,uBAAS,KAAK,GAAG,KAAK,OAAO,EAAE,IAAI;AACjC,sBAAM,KAAK,KAAK,QAAQ;AACxB,sBAAM,QAAQ,SAAS,WAAW;AAClC,sBAAM,SAAS,UAAU,YAAY;AACrC,2BAAW,QAAQ;AAAA,cACrB;AAAA,YACF;AAAA,UACF;AACA,mBAAS,OAAO,IAAI,OAAO,KAAK,OAAO,KAAK,MAAM;AAAA,QACpD;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,2CAA2C;AAAA,EAC7C,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,SAAS,OAAO,CAAC,OAAO,KAAK,GAAG,EAAE,KAAK;AAC7C,QAAM,OAAO,OAAO;AACpB,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,SAAK,IAAI,QAAQ,KAAK,MAAM;AAAA,EAC9B;AACA,QAAM,WAAW,CAAC,GAAG,EAAE,OAAO,GAAG,EAAE,KAAK;AACxC,SAAO,SAAS,eAAe,UAAU,OAAO,OAAO,OAAO,MAAM;AACtE;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AACpD,UAAM,EAAE,GAAG,OAAO,IAAI;AACtB,UAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,UAAM,aAAa;AACnB,UAAM,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,UAAM,QAAQ,EAAE,MAAM;AACtB,UAAM,aAAa,WAAW,KAAK,IAAI,OAAO,MAAM,EAAE;AACtD,UAAM,aAAa,OAAO,MAAM;AAChC,UAAM,EAAE,WAAW,UAAU,SAAS,YAAY,WAAW,UAAU,SAAS,cAAc,aAAa,cAAc,aAAa,gBAAgB,eAAe,SAAS,IAAI,qBAAqB,sBAAsB,EAAE,OAAO,OAAO,OAAO,SAAS,MAAM,QAAQ,SAAS;AACpR,UAAM,UAAU,aAAa,cAAc,QAAQ;AACnD,UAAM,UAAU,SAAS;AACzB,UAAM,aAAa,aAAa,kBAAkB,EAAE,OAAO,OAAO;AAClE,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,eAAS,OAAO,GAAG,OAAO,WAAW,EAAE,MAAM;AAC3C,cAAM,OAAO,OAAO,eAAe,QAAQ;AAC3C,iBAAS,OAAO,GAAG,OAAO,UAAU,EAAE,MAAM;AAC1C,gBAAM,OAAO,OAAO,cAAc,QAAQ;AAC1C,mBAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,gBAAI,SAAS,OAAO;AACpB,qBAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,oBAAM,MAAM,OAAO,IAAI;AACvB,kBAAI,OAAO,KAAK,MAAM,UAAU;AAC9B,yBAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,wBAAM,MAAM,OAAO,IAAI;AACvB,sBAAI,OAAO,KAAK,MAAM,SAAS;AAC7B,0BAAM,SAAS,aAAa,WAAW,CAAC,GAAG,KAAK,KAAK,CAAC,GAAG,OAAO,aAAa,eAAe,EAAE,KAAK,CAAC;AACpG,0BAAM,cAAc,aAAa,WAAW,CAAC,GAAG,GAAG,CAAC,GAAG,YAAY,aAAa,eAAe,OAAO,KAAK,CAAC;AAC5G,0BAAM,MAAM,MAAM,UAAU,WAAW;AACvC,wBAAI,MAAM,QAAQ;AAChB,+BAAS;AAAA,oBACX;AAAA,kBACF;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,kBAAM,cAAc,aAAa,WAAW,CAAC,GAAG,MAAM,MAAM,CAAC,GAAG,SAAS,aAAa,eAAe,QAAQ,CAAC;AAC9G,uBAAW,eAAe;AAAA,UAC5B;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,WAAW,MAAM,aAAa,aAAa,YAAY,EAAE,KAAK,GAAG,UAAU,EAAE,KAAK;AACjG,WAAO,EAAE,QAAQ,OAAO,UAAU,OAAO,EAAE,MAAM;AAAA,EACnD;AACF;AAGA,IAAI,iCAAiC;AAAA,EACnC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AACpD,UAAM,EAAE,GAAG,QAAQ,GAAG,IAAI;AAC1B,UAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,UAAM,aAAa;AACnB,UAAM,KAAK,aAAa,cAAc,EAAE,OAAO,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM;AACnF,UAAM,UAAU,aAAa,cAAc,OAAO,OAAO,WAAW,KAAK,IAAI,OAAO,MAAM,EAAE,MAAM;AAClG,UAAM,EAAE,WAAW,UAAU,SAAS,YAAY,WAAW,UAAU,SAAS,cAAc,aAAa,cAAc,aAAa,gBAAgB,eAAe,SAAS,IAAI,qBAAqB,sBAAsB,EAAE,OAAO,OAAO,OAAO,SAAS,MAAM,QAAQ,SAAS;AACpR,iBAAa,OAAO,GAAG,SAAS,SAAS,QAAQ,MAAM,YAAY,kEAAkE,SAAS,mBAAmB,GAAG,MAAM;AAC1K,UAAM,MAAM,aAAa,cAAc,UAAU,WAAW,KAAK,IAAI,GAAG,MAAM,EAAE,MAAM;AACtF,UAAM,YAAY,aAAa,0BAA0B,OAAO,OAAO,OAAO,KAAK;AACnF,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,eAAS,OAAO,GAAG,OAAO,WAAW,EAAE,MAAM;AAC3C,cAAM,OAAO,OAAO,eAAe,QAAQ;AAC3C,iBAAS,OAAO,GAAG,OAAO,UAAU,EAAE,MAAM;AAC1C,gBAAM,OAAO,OAAO,cAAc,QAAQ;AAC1C,mBAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,gBAAI,SAAS,OAAO;AACpB,gBAAI,OAAO;AACX,gBAAI,OAAO;AACX,qBAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,oBAAM,MAAM,OAAO,IAAI;AACvB,kBAAI,OAAO,KAAK,MAAM,UAAU;AAC9B,yBAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,wBAAM,MAAM,OAAO,IAAI;AACvB,sBAAI,OAAO,KAAK,MAAM,SAAS;AAC7B,0BAAM,MAAM,GAAG,GAAG,KAAK,KAAK,KAAK,QAAQ,GAAG,GAAG;AAC/C,wBAAI,MAAM,QAAQ;AAChB,+BAAS;AACT,6BAAO;AACP,6BAAO;AAAA,oBACT;AAAA,kBACF;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,sBAAU,MAAM,MAAM,MAAM,IAAI,GAAG,MAAM,MAAM;AAAA,UACjD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,WAAW,MAAM,aAAa,aAAa,WAAW,EAAE,KAAK,GAAG,OAAO,OAAO,OAAO,KAAK;AACzG,WAAO,EAAE,QAAQ,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAAA,EAC5D;AACF;AAGA,IAAI,gCAAgC;AAAA,EAClC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AACpD,UAAM,EAAE,GAAG,QAAQ,GAAG,IAAI;AAC1B,UAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,UAAM,aAAa;AACnB,UAAM,KAAK,aAAa,cAAc,EAAE,OAAO,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM;AACnF,UAAM,UAAU,aAAa,cAAc,OAAO,OAAO,WAAW,KAAK,IAAI,OAAO,MAAM,EAAE,MAAM;AAClG,UAAM,EAAE,WAAW,UAAU,SAAS,YAAY,WAAW,UAAU,SAAS,cAAc,aAAa,cAAc,aAAa,gBAAgB,eAAe,SAAS,IAAI,qBAAqB,sBAAsB,EAAE,OAAO,OAAO,OAAO,SAAS,MAAM,QAAQ,SAAS;AACpR,iBAAa,OAAO,GAAG,SAAS,SAAS,QAAQ,MAAM,YAAY,iEAAiE,SAAS,mBAAmB,GAAG,MAAM;AACzK,UAAM,MAAM,aAAa,cAAc,UAAU,WAAW,KAAK,IAAI,GAAG,MAAM,EAAE,MAAM;AACtF,UAAM,YAAY,aAAa,0BAA0B,EAAE,OAAO,EAAE,KAAK;AACzE,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,eAAS,OAAO,GAAG,OAAO,WAAW,EAAE,MAAM;AAC3C,cAAM,OAAO,OAAO,eAAe,QAAQ;AAC3C,iBAAS,OAAO,GAAG,OAAO,UAAU,EAAE,MAAM;AAC1C,gBAAM,OAAO,OAAO,cAAc,QAAQ;AAC1C,mBAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,gBAAI,SAAS,OAAO;AACpB,gBAAI,SAAS,OAAO,IAAI,IAAI;AAC5B,gBAAI,SAAS,OAAO,IAAI,IAAI;AAC5B,qBAAS,IAAI,GAAG,IAAI,cAAc,EAAE,GAAG;AACrC,oBAAM,MAAM,OAAO,IAAI;AACvB,kBAAI,OAAO,KAAK,MAAM,UAAU;AAC9B,yBAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,wBAAM,MAAM,OAAO,IAAI;AACvB,sBAAI,OAAO,KAAK,MAAM,SAAS;AAC7B,0BAAM,MAAM,GAAG,GAAG,KAAK,KAAK,KAAK,QAAQ,GAAG,GAAG;AAC/C,wBAAI,MAAM,QAAQ;AAChB,+BAAS;AACT,+BAAS;AACT,+BAAS;AAAA,oBACX;AAAA,kBACF;AAAA,gBACF;AAAA,cACF;AAAA,YACF;AACA,sBAAU,GAAG,QAAQ,QAAQ,MAAM,IAAI,GAAG,MAAM,MAAM;AAAA,UACxD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,WAAW,MAAM,aAAa,aAAa,WAAW,EAAE,KAAK,GAAG,EAAE,OAAO,EAAE,KAAK;AAC/F,WAAO,EAAE,QAAQ,OAAO,EAAE,OAAO,OAAO,EAAE,MAAM;AAAA,EAClD;AACF;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,mBAAiB,GAAG,KAAK;AACzB,MAAI;AACJ,MAAI,EAAE,UAAU,QAAQ;AACtB,SAAK,MAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,QAAQ,EAAE,CAAC;AAAA,EAC5E,OAAO;AACL,SAAK,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACrD;AACA,QAAM,QAAQ,GAAG,MAAM;AACvB,QAAM,OAAO,aAAa,eAAe,MAAM,GAAG,KAAK;AACvD,QAAM,cAAc,qBAAqB,mBAAmB,MAAM,KAAK;AACvE,MAAI,gBAAgB;AACpB,MAAI,YAAY;AAChB,MAAI,eAAe,MAAM;AACvB,gBAAY,WAAW,EAAE,QAAQ,EAAE,GAAG,GAAG,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AAC7F,oBAAgB,qBAAqB,iBAAiB,cAAc,QAAQ,KAAK;AAAA,EACnF;AACA,uBAAqB,2BAA2B,OAAO,eAAe,UAAU,MAAM,MAAM;AAC5F,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,aAAa;AAC7G,QAAM,cAAc,qBAAqB,WAAW,UAAU,OAAO,OAAO;AAC5E,MAAI,SAAS,OAAO,UAAU,UAAU,WAAW;AACnD,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,OAAO,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAC9C,QAAM,QAAQ,SAAS,KAAK,IAAI,UAAU,MAAM,EAAE;AAClD,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO;AACX,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,cAAQ,MAAM,SAAS;AAAA,IACzB;AACA,SAAK,KAAK;AAAA,EACZ;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,OAAO,OAAO,IAAI;AAC7E,UAAM,YAAY;AAClB,aAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC1F,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,WAAS,8BAA8B,EAAE;AACzC,MAAI,eAAe,MAAM;AACvB,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,SAAS,IAAI;AACrB,QAAM,UAAU;AAChB,QAAM,EAAE,SAAS,YAAY,OAAO,IAAI,qBAAqB,qBAAqB,UAAU,QAAQ,MAAM;AAC1G,uBAAqB,oBAAoB,QAAQ,QAAQ,QAAQ,OAAO;AACxE,QAAM,EAAE,MAAM,MAAM,IAAI,qBAAqB,qBAAqB,YAAY,MAAM;AACpF,QAAM,SAAS,MAAM;AACrB,MAAI,MAAM;AACV,MAAI,mBAAmB,QAAQ;AAC/B,QAAM,mBAAmB,CAAC;AAC1B,WAAS,IAAI,GAAG,IAAI,QAAQ,EAAE,GAAG;AAC/B,eAAW,UAAU,MAAM,IAAI;AAC7B,YAAM,EAAE,oBAAoB,MAAM,YAAY,aAAa,IAAI,qBAAqB,qBAAqB,kBAAkB,OAAO,OAAO;AACzI,UAAI;AACJ,UAAI,qBAAqB,sBAAsB,IAAI,GAAG;AACpD,YAAI,QAAQ;AAAA,MACd,OAAO;AACL,YAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,QAAQ,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AACrF,yBAAiB,KAAK,CAAC;AAAA,MACzB;AACA,YAAM,cAAc,EAAE,MAAM,MAAM;AAClC,eAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,oBAAY,OAAO,aAAa,IAAI,GAAG,CAAC;AAAA,MAC1C;AACA,UAAI,CAAC,aAAa,YAAY,EAAE,OAAO,WAAW,GAAG;AACnD,YAAI,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAChF,yBAAiB,KAAK,CAAC;AAAA,MACzB;AACA,UAAI,QAAQ,MAAM;AAChB,cAAM;AAAA,MACR,OAAO;AACL,cAAM,UAAU,EAAE,QAAQ,EAAE,GAAG,GAAG,GAAG,IAAI,GAAG,SAAS,SAAS,CAAC;AAC/D,yBAAiB,KAAK,GAAG;AAAA,MAC3B;AAAA,IACF;AACA,QAAI,IAAI,SAAS,GAAG;AAClB,UAAI,KAAK,MAAM,GAAG;AAChB,cAAM,KAAK;AAAA,UACT,QAAQ,EAAE,GAAG,IAAI;AAAA,UACjB,SAAS;AAAA,UACT,OAAO;AAAA,YACL,MAAM,KAAK,MAAM,QAAQ,SAAS;AAAA,YAClC,UAAU;AAAA,UACZ;AAAA,QACF,CAAC;AACD,yBAAiB,KAAK,GAAG;AAAA,MAC3B;AACA;AAAA,IACF;AAAA,EACF;AACA,aAAW,cAAc,kBAAkB;AACzC,QAAI,eAAe,KAAK;AACtB;AAAA,IACF;AACA,aAAS,8BAA8B,UAAU;AAAA,EACnD;AACA,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,IAAI,EAAE,IAAI;AAClB,mBAAiB,CAAC,IAAI,CAAC,GAAG,SAAS;AACnC,QAAM,eAAe,IAAI,aAAa,aAAa,cAAc,EAAE,KAAK,CAAC;AACzE,QAAM,SAAS,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC3C,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,UAAM,IAAI,OAAO;AACjB,QAAI,KAAK,GAAG;AACV,mBAAa,KAAK,SAAS;AAAA,IAC7B,OAAO;AACL,mBAAa,KAAK,SAAS,MAAM,IAAI;AAAA,IACvC;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,WAAW,YAAY;AACjE;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,IAAI,qBAAqB;AAC7B,IAAI,KAAK,qBAAqB;AAC9B,IAAI,KAAK,qBAAqB;AAC9B,IAAI,KAAK,qBAAqB;AAC9B,IAAI,KAAK,qBAAqB;AAC9B,IAAI,KAAK,qBAAqB;AAC9B,IAAI,OAAO,gBAAgB,KAAK,CAAC,OAAO;AACtC,QAAM,QAAQ,KAAK,KAAK,EAAE;AAC1B,QAAM,IAAI,KAAK,IAAI,EAAE;AACrB,QAAM,IAAI,KAAK,IAAI,IAAI;AACvB,SAAO,SAAS,QAAQ,KAAK,IAAI,MAAM,IAAI,MAAM,IAAI,MAAM,IAAI,MAAM,IAAI,KAAK,IAAI,CAAC,IAAI,CAAC;AAC1F,CAAC;AACD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,IAAI,IAAI;AAChB,QAAM,YAAY,OAAO,MAAM;AAC/B,QAAM,WAAW,OAAO,MAAM,MAAM;AACpC,MAAI,OAAO;AACX,MAAI,MAAM,GAAG;AACX,iBAAa,OAAO,EAAE,YAAY,MAAM,KAAK,MAAM,iCAAiC,EAAE,YAAY,OAAO,YAAY;AACrH,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,WAAS,OAAO,MAAM,GAAG,CAAC;AAC1B,SAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC1F;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,6BAA6B,CAAC,GAAG,MAAM,IAAI,CAAC;AAC9D,IAAI,OAAO,iBAAiB,SAAS,WAAW;AAChD,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,QAAQ,SAAS,YAAY;AAC7C,QAAM,aAAa,OAAO;AAC1B,QAAM,QAAQ,WAAW;AACzB,QAAM,WAAW,WAAW;AAC5B,QAAM,YAAY,WAAW,KAAK,IAAI,OAAO,MAAM;AACnD,QAAM,SAAS,UAAU,mBAAmB;AAC5C,QAAM,SAAS,UAAU,mBAAmB;AAC5C,QAAM,cAAc,CAAC,OAAO,QAAQ;AACpC,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,aAAa,aAAa,uBAAuB,WAAW,UAAU;AAC5E,QAAM,aAAa,aAAa,uBAAuB,WAAW,UAAU;AAC5E,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,IAAI,OAAO;AAAA,MACf,QAAQ,EAAE,GAAG,OAAO;AAAA,MACpB,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,CAAC,GAAG,CAAC,GAAG,MAAM,CAAC,GAAG,QAAQ,EAAE;AAAA,IAC9C,CAAC;AACD,UAAM,IAAI,OAAO;AAAA,MACf,QAAQ,EAAE,GAAG,OAAO;AAAA,MACpB,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,CAAC,GAAG,CAAC,GAAG,MAAM,CAAC,GAAG,QAAQ,EAAE;AAAA,IAC9C,CAAC;AACD,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,SAAS,WAAW,CAAC;AAC7E,UAAM,EAAE,MAAM,OAAO,MAAM,MAAM,IAAI,QAAQ,QAAQ,SAAS,UAAU;AACxE,UAAM,MAAM,qBAAqB,uBAAuB,OAAO,KAAK;AACpE,aAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,YAAM,IAAI,qBAAqB,oBAAoB,KAAK,CAAC;AACzD,iBAAW,IAAI,WAAW,KAAK,EAAE;AACjC,iBAAW,IAAI,WAAW,KAAK,EAAE;AAAA,IACnC;AACA,eAAW,8BAA8B,CAAC;AAC1C,eAAW,8BAA8B,CAAC;AAC1C,eAAW,8BAA8B,MAAM;AAAA,EACjD;AACA,QAAM,YAAY,WAAW,eAAe,aAAa,WAAW,UAAU;AAC9E,QAAM,YAAY,WAAW,eAAe,aAAa,WAAW,UAAU;AAC9E,QAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,WAAW,MAAM,UAAU,GAAG,SAAS,WAAW,CAAC;AAC7F,aAAW,8BAA8B,SAAS;AAClD,aAAW,8BAA8B,SAAS;AAClD,SAAO;AACT;AACA,SAAS,QAAQ,QAAQ,SAAS,YAAY;AAC5C,QAAM,YAAY,aAAa,cAAc,OAAO,KAAK;AACzD,QAAM,YAAY,WAAW,KAAK,IAAI,OAAO,MAAM;AACnD,QAAM,WAAW,WAAW,KAAK,IAAI,UAAU,mBAAmB,KAAK,MAAM,EAAE;AAC/E,QAAM,WAAW,WAAW,KAAK,IAAI,UAAU,mBAAmB,KAAK,MAAM,EAAE;AAC/E,MAAI,cAAc,SAAS,GAAG;AAC5B,UAAM,SAAS,UAAU,UAAU,UAAU,WAAW,SAAS,UAAU;AAC3E,UAAM,cAAc,CAAC,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE;AACrD,QAAI,SAAS;AACX,YAAM,WAAW,WAAW,eAAe,aAAa,WAAW,OAAO,IAAI;AAC9E,YAAM,WAAW,WAAW,eAAe,aAAa,WAAW,OAAO,IAAI;AAC9E,YAAM,WAAW,WAAW,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,WAAW,SAAS,CAAC;AAC9G,YAAM,eAAe,UAAU,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,WAAW,CAAC;AAC/E,YAAM,cAAc,cAAc,WAAW,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,GAAG,SAAS,WAAW,CAAC;AAC1G,YAAM,cAAc,cAAc,WAAW,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,aAAa,GAAG,SAAS,WAAW,CAAC;AAC9G,YAAM,cAAc,WAAW,KAAK,IAAI,YAAY,MAAM,EAAE;AAC5D,YAAM,cAAc,WAAW,KAAK,IAAI,YAAY,MAAM,EAAE;AAC5D,iBAAW,8BAA8B,QAAQ;AACjD,iBAAW,8BAA8B,QAAQ;AACjD,iBAAW,8BAA8B,QAAQ;AACjD,iBAAW,8BAA8B,YAAY;AACrD,iBAAW,8BAA8B,WAAW;AACpD,iBAAW,8BAA8B,WAAW;AACpD,aAAO,EAAE,MAAM,aAAa,MAAM,YAAY;AAAA,IAChD;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,OAAO,qBAAqB,uBAAuB,UAAU,QAAQ;AAC3E,UAAM,YAAY,yBAAyB,MAAM,WAAW,OAAO;AACnE,WAAO,qBAAqB,uBAAuB,SAAS;AAAA,EAC9D;AACF;AACA,SAAS,cAAc,MAAM;AAC3B,UAAQ,OAAO,OAAO,OAAO;AAC/B;AACA,SAAS,UAAU,UAAU,UAAU,MAAM,SAAS,YAAY;AAChE,MAAI,SAAS,GAAG;AACd,WAAO,EAAE,MAAM,UAAU,MAAM,SAAS;AAAA,EAC1C;AACA,QAAM,OAAO,qBAAqB,uBAAuB,UAAU,QAAQ;AAC3E,QAAM,OAAO,OAAO;AACpB,QAAM,cAAc,qBAAqB,qBAAqB,IAAI;AAClE,QAAM,eAAe,YAAY;AACjC,QAAM,eAAe,YAAY;AACjC,QAAM,YAAY,CAAC,aAAa,MAAM;AACtC,QAAM,eAAe,WAAW,eAAe,WAAW,WAAW,YAAY;AACjF,QAAM,eAAe,WAAW,eAAe,WAAW,WAAW,YAAY;AACjF,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,MAAM,cAAc,MAAM,aAAa,GAAG,SAAS,WAAW,CAAC;AAC3G,QAAM,aAAa,qBAAqB,oBAAoB,IAAI;AAChE,QAAM,cAAc,WAAW;AAC/B,QAAM,cAAc,WAAW;AAC/B,QAAM,WAAW,CAAC,YAAY,MAAM;AACpC,QAAM,cAAc,WAAW,eAAe,UAAU,WAAW,WAAW;AAC9E,QAAM,cAAc,WAAW,eAAe,UAAU,WAAW,WAAW;AAC9E,QAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,MAAM,aAAa,MAAM,YAAY,GAAG,SAAS,WAAW,CAAC;AACxG,QAAM,eAAe,UAAU,cAAc,cAAc,MAAM,SAAS,UAAU;AACpF,QAAM,gBAAgB,aAAa;AACnC,QAAM,gBAAgB,aAAa;AACnC,QAAM,aAAa,CAAC,cAAc,MAAM;AACxC,QAAM,gBAAgB,WAAW,eAAe,YAAY,WAAW,aAAa;AACpF,QAAM,gBAAgB,WAAW,eAAe,YAAY,WAAW,aAAa;AACpF,QAAM,kBAAkB,SAAS;AAAA,IAC/B,QAAQ,EAAE,MAAM,eAAe,MAAM,cAAc;AAAA,IACnD,SAAS;AAAA,EACX,CAAC;AACD,QAAM,cAAc,UAAU,aAAa,aAAa,MAAM,SAAS,UAAU;AACjF,QAAM,eAAe,YAAY;AACjC,QAAM,eAAe,YAAY;AACjC,QAAM,YAAY,CAAC,aAAa,MAAM;AACtC,QAAM,eAAe,WAAW,eAAe,WAAW,WAAW,YAAY;AACjF,QAAM,eAAe,WAAW,eAAe,WAAW,WAAW,YAAY;AACjF,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,MAAM,cAAc,MAAM,aAAa,GAAG,SAAS,WAAW,CAAC;AAC3G,QAAM,IAAI,qBAAqB,UAAU,MAAM,OAAO;AACtD,QAAM,SAAS,CAAC,EAAE,KAAK,MAAM;AAC7B,QAAM,YAAY,WAAW,eAAe,QAAQ,WAAW,EAAE,IAAI;AACrE,QAAM,YAAY,WAAW,eAAe,QAAQ,WAAW,EAAE,IAAI;AACrE,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,MAAM,WAAW,MAAM,UAAU,GAAG,SAAS,WAAW,CAAC;AAClG,QAAM,eAAe,UAAU,EAAE,QAAQ,EAAE,GAAG,aAAa,GAAG,eAAe,GAAG,SAAS,WAAW,CAAC;AACrG,QAAM,UAAU,KAAK;AAAA,IACnB,QAAQ,EAAE,GAAG,iBAAiB,GAAG,aAAa;AAAA,IAC9C,SAAS;AAAA,EACX,CAAC;AACD,QAAM,UAAU,KAAK;AAAA,IACnB,QAAQ,EAAE,GAAG,iBAAiB,GAAG,aAAa;AAAA,IAC9C,SAAS;AAAA,EACX,CAAC;AACD,QAAM,cAAc,MAAM,EAAE,QAAQ,EAAE,OAAO,QAAQ,GAAG,SAAS,WAAW,CAAC;AAC7E,QAAM,cAAc,MAAM,EAAE,QAAQ,EAAE,OAAO,QAAQ,GAAG,SAAS,WAAW,CAAC;AAC7E,QAAM,cAAc,MAAM,EAAE,QAAQ,EAAE,OAAO,QAAQ,GAAG,SAAS,WAAW,CAAC;AAC7E,QAAM,cAAc,MAAM,EAAE,QAAQ,EAAE,OAAO,QAAQ,GAAG,SAAS,WAAW,CAAC;AAC7E,QAAM,QAAQ,QAAQ;AAAA,IACpB,QAAQ,CAAC,aAAa,WAAW;AAAA,IACjC,SAAS;AAAA,IACT,OAAO,EAAE,MAAM,EAAE;AAAA,EACnB,CAAC;AACD,QAAM,QAAQ,QAAQ;AAAA,IACpB,QAAQ,CAAC,aAAa,WAAW;AAAA,IACjC,SAAS;AAAA,IACT,OAAO,EAAE,MAAM,EAAE;AAAA,EACnB,CAAC;AACD,QAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,QAAM,YAAY,WAAW,KAAK,IAAI,MAAM,MAAM,EAAE;AACpD,aAAW,8BAA8B,YAAY;AACrD,aAAW,8BAA8B,YAAY;AACrD,aAAW,8BAA8B,cAAc;AACvD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,aAAa;AACtD,aAAW,8BAA8B,aAAa;AACtD,aAAW,8BAA8B,aAAa;AACtD,aAAW,8BAA8B,eAAe;AACxD,aAAW,8BAA8B,YAAY;AACrD,aAAW,8BAA8B,YAAY;AACrD,aAAW,8BAA8B,cAAc;AACvD,aAAW,8BAA8B,SAAS;AAClD,aAAW,8BAA8B,SAAS;AAClD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,YAAY;AACrD,aAAW,8BAA8B,OAAO;AAChD,aAAW,8BAA8B,OAAO;AAChD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,WAAW;AACpD,aAAW,8BAA8B,KAAK;AAC9C,aAAW,8BAA8B,KAAK;AAC9C,SAAO,EAAE,MAAM,WAAW,MAAM,UAAU;AAC5C;AACA,SAAS,yBAAyB,MAAM,MAAM,SAAS;AACrD,QAAM,MAAM,IAAI,aAAa,OAAO,CAAC;AACrC,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,QAAI,QAAQ;AACZ,QAAI,QAAQ;AACZ,aAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,YAAM,IAAI,qBAAqB,SAAS,IAAI,GAAG,MAAM,OAAO;AAC5D,YAAM,OAAO,qBAAqB,oBAAoB,MAAM,CAAC;AAC7D,eAAS,KAAK,OAAO,EAAE,OAAO,KAAK,OAAO,EAAE;AAC5C,eAAS,KAAK,OAAO,EAAE,OAAO,KAAK,OAAO,EAAE;AAAA,IAC9C;AACA,QAAI,SAAS;AACX,eAAS;AACT,eAAS;AAAA,IACX;AACA,yBAAqB,mBAAmB,KAAK,OAAO,OAAO,CAAC;AAAA,EAC9D;AACA,SAAO;AACT;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,YAAY,aAAa,cAAc,OAAO,KAAK;AACzD,QAAM,qBAAqB,OAAO,MAAM,OAAO,MAAM,SAAS;AAC9D,QAAM,QAAQ,YAAY;AAC1B,QAAM,UAAU,SAAS;AAAA,IACvB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,OAAO,kBAAkB,EAAE;AAAA,EAC9C,CAAC;AACD,QAAM,SAAS,SAAS,SAAS,OAAO,QAAQ;AAChD,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,OAAO,MAAM,EAAE,CAAC;AAC5G,WAAS,8BAA8B,OAAO;AAC9C,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,OAAO,MAAM,IAAI;AAChC,QAAM,SAAS,SAAS,aAAa,WAAW,KAAK;AACrD,QAAM,SAAS,aAAa,kBAAkB,QAAQ,aAAa,cAAc,KAAK,CAAC;AACvF,aAAW,QAAQ,OAAO,MAAM;AAChC,SAAO,SAAS,eAAe,OAAO,QAAQ,MAAM;AACtD;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AACA,SAAS,WAAW,QAAQ,OAAO,OAAO;AACxC,MAAI,UAAU,UAAU;AACtB,WAAO,KAAK,KAAK;AAAA,EACnB,OAAO;AACL,WAAO,KAAK,KAAK;AAAA,EACnB;AACF;AAGA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,aAAa;AACnB,UAAM,SAAS,aAAa,uBAAuB,OAAO,OAAO,aAAa,cAAc,OAAO,KAAK,CAAC;AACzG,UAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,UAAM,YAAY,WAAW,KAAK,IAAI,OAAO,MAAM,EAAE;AACrD,aAAS,WAAW,GAAG,WAAW,OAAO,YAAY;AACnD,YAAM,cAAc,WAAW,aAAa,cAAc;AAC1D,eAAS,MAAM,GAAG,MAAM,aAAa,OAAO;AAC1C,cAAM,YAAY,OAAO,aAAa;AACtC,iBAAS,MAAM,GAAG,MAAM,YAAY,OAAO;AACzC,gBAAM,YAAY,MAAM;AACxB,mBAAS,UAAU,GAAG,UAAU,aAAa,WAAW;AACtD,kBAAM,SAAS,KAAK,MAAM,aAAa,MAAM,CAAC;AAC9C,kBAAM,SAAS,cAAc,YAAY,YAAY;AACrD,gBAAI,cAAc,UAAU;AAC5B,gBAAI,UAAU,KAAK,SAAS,YAAY;AACtC,oBAAM,mBAAmB,SAAS;AAClC,oBAAM,WAAW,cAAc,YAAY,mBAAmB;AAC9D,4BAAc,UAAU;AAAA,YAC1B;AACA,mBAAO,UAAU;AAAA,UACnB;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,WAAW,MAAM,QAAQ,OAAO,OAAO,OAAO,KAAK;AAClE,WAAO,EAAE,QAAQ,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAAA,EAC5D;AACF;AAGA,IAAI,eAAe,6BAA6B,CAAC,GAAG,MAAM,KAAK,MAAM,IAAI,CAAC,CAAC;AAC3E,IAAI,YAAY,iBAAiB,UAAU,cAAc,MAAM,OAAO;AACtE,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,iBAAiB,YAAY,aAAa,eAAe,IAAI;AAChH,MAAI,SAAS,OAAO;AAAA,IAClB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB;AAAA,EACtE,CAAC;AACD,MAAI,MAAM;AACR,UAAM,YAAY;AAClB,QAAI,eAAe,UAAU,KAAK,MAAM,WAAW,KAAK,KAAK,MAAM,OAAO,GAAG;AAC3E,YAAM,eAAe,SAAS,EAAE,QAAQ,EAAE,GAAG,KAAK,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,KAAK,MAAM,IAAI,GAAG,CAAC,EAAE,EAAE,CAAC;AACjH,eAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,aAAa,GAAG,SAAS,SAAS,CAAC;AAC3E,eAAS,8BAA8B,YAAY;AAAA,IACrD,OAAO;AACL,eAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,KAAK,GAAG,SAAS,SAAS,CAAC;AAAA,IACrE;AACA,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,MAAI,aAAa;AACf,UAAM,YAAY;AAClB,QAAI,eAAe,UAAU,gBAAgB,WAAW,uBAAuB,MAAM,WAAW,KAAK,uBAAuB,MAAM,OAAO,GAAG;AAC1I,YAAM,gBAAgB,SAAS;AAAA,QAC7B,QAAQ,EAAE,GAAG,uBAAuB;AAAA,QACpC,SAAS;AAAA,QACT,OAAO,EAAE,OAAO,CAAC,uBAAuB,MAAM,IAAI,GAAG,CAAC,EAAE;AAAA,MAC1D,CAAC;AACD,eAAS,iBAAiB,UAAU,QAAQ,aAAa,eAAe,cAAc;AACtF,eAAS,8BAA8B,aAAa;AAAA,IACtD,OAAO;AACL,eAAS,iBAAiB,UAAU,QAAQ,aAAa,wBAAwB,cAAc;AAAA,IACjG;AACA,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,iBAAiB,YAAY,aAAa,eAAe,IAAI;AAChH,MAAI,SAAS,sBAAsB;AAAA,IACjC,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB;AAAA,EACtE,CAAC;AACD,MAAI,MAAM;AACR,UAAM,YAAY;AAClB,aAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,KAAK,GAAG,SAAS,SAAS,CAAC;AACnE,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,MAAI,aAAa;AACf,UAAM,YAAY;AAClB,aAAS,iBAAiB,UAAU,QAAQ,aAAa,wBAAwB,cAAc;AAC/F,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,QAAQ,QAAQ,IAAI;AAC5B,QAAM,aAAa,aAAa,cAAc,OAAO,KAAK;AAC1D,QAAM,eAAe,QAAQ;AAC7B,QAAM,YAAY,aAAa,aAAa,SAAS;AACrD,QAAM,CAAC,aAAa,WAAW,WAAW,OAAO,IAAI,qBAAqB,mBAAmB,QAAQ,OAAO;AAC5G,MAAI,cAAc,GAAG;AACnB,WAAO,SAAS,eAAe,aAAa,OAAO,OAAO,CAAC,CAAC;AAAA,EAC9D;AACA,QAAM,cAAc,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACtD,QAAM,YAAY,SAAS,WAAW,MAAM;AAC5C,QAAM,SAAS,aAAa,aAAa,WAAW,OAAO,OAAO,WAAW,WAAW,WAAW,SAAS,OAAO,OAAO,UAAU;AACpI,SAAO,SAAS,eAAe,aAAa,OAAO,OAAO,OAAO,MAAM;AACzE;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,MAAM,UAAU,IAAI;AAC5B,mBAAiB,CAAC,GAAG,OAAO,GAAG,UAAU;AACzC,QAAM,aAAa,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AAC9D,QAAM,cAAc,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACtD,QAAM,UAAU,EAAE,MAAM;AACxB,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,UAAM,QAAQ,YAAY;AAC1B,iBAAa,OAAO,SAAS,UAAU,KAAK,SAAS,GAAG,MAAM,6BAA6B,uBAAuB,UAAU,IAAI;AAAA,EAClI;AACA,MAAI,aAAa;AACjB,MAAI,aAAa,MAAM;AACrB,iBAAa;AAAA,EACf;AACA,QAAM,cAAc,aAAa,cAAc,QAAQ,KAAK;AAC5D,QAAM,YAAY,qBAAqB,aAAa,yBAAyB,GAAG,SAAS,YAAY,UAAU;AAC/G,QAAM,WAAW,SAAS;AAAA,IACxB,QAAQ,EAAE,EAAE;AAAA,IACZ,SAAS;AAAA,IACT,OAAO;AAAA,MACL,OAAO;AAAA,QACL,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,MACZ;AAAA,IACF;AAAA,EACF,CAAC;AACD,QAAM,eAAe,SAAS;AAAA,IAC5B,QAAQ,EAAE,GAAG,QAAQ;AAAA,IACrB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,UAAU,WAAW,cAAc,UAAU,SAAS,EAAE;AAAA,EAC3E,CAAC;AACD,QAAM,qBAAqB;AAAA,IACzB,UAAU;AAAA,IACV,UAAU;AAAA,IACV,cAAc,UAAU;AAAA,IACxB,UAAU;AAAA,EACZ;AACA,QAAM,aAAa,SAAS,WAAW,YAAY;AACnD,QAAM,OAAO,SAAS,WAAW,QAAQ;AACzC,QAAM,SAAS,aAAa,MAAM,YAAY,kBAAkB;AAChE,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,YAAY;AACnD,SAAO,SAAS,eAAe,UAAU,aAAa,OAAO,OAAO,OAAO,MAAM;AACnF;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,YAAY,aAAa,cAAc,OAAO,KAAK;AACzD,QAAM,qBAAqB,OAAO,MAAM,OAAO,MAAM,SAAS;AAC9D,QAAM,QAAQ,YAAY;AAC1B,QAAM,UAAU,SAAS;AAAA,IACvB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,OAAO,kBAAkB,EAAE;AAAA,EAC9C,CAAC;AACD,QAAM,SAAS,SAAS,SAAS,MAAM,QAAQ;AAC/C,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,OAAO,MAAM,EAAE,CAAC;AAC5G,WAAS,8BAA8B,OAAO;AAC9C,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY,gBAAgB,UAAU,CAAC,OAAO,OAAO,SAAS,EAAE,IAAI,IAAI,GAAG,MAAM;AACrF,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,IAAI,EAAE,MAAM,WAAW,IAAI,GAAG,MAAM;AACrF,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,OAAO,MAAM,EAAE,IAAI,IAAI,GAAG,MAAM;AAC5E,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,IAAI,IAAI;AAC7B,QAAM,UAAU,aAAa,OAAO,MAAM,GAAG;AAC7C,SAAO,SAAS,eAAe,CAAC,QAAQ,MAAM,GAAG,WAAW,OAAO;AACrE;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO,KAAK,MAAM,EAAE,CAAC;AAC1D,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB,6BAA6B,CAAC,GAAG,MAAM,KAAK,CAAC;AAClE,IAAI,cAAc,iBAAiB,YAAY,gBAAgB,MAAM,MAAM;AAC3E,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,gBAAgB,YAAY,CAAC,OAAO,KAAK,IAAI,GAAG,MAAM;AACxE,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,6BAA6B,CAAC,GAAG,MAAM,KAAK,CAAC;AACjE,IAAI,aAAa,iBAAiB,WAAW,eAAe,MAAM,MAAM;AACxE,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,IAAI,MAAM;AACjB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,aAAa,MAAM,OAAO,KAAK,IAAI;AAC3C,mBAAiB,GAAG,KAAK;AACzB,QAAM,WAAW,EAAE,MAAM;AACzB,QAAM,OAAO,WAAW;AACxB,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,OAAO,aAAa,cAAc,EAAE,KAAK;AAC/C,QAAM,SAAS,IAAI,aAAa,IAAI;AACpC,WAAS,kBAAkB,QAAQ;AACjC,UAAM,iBAAiB,SAAS;AAChC,QAAI,iBAAiB,SAAS,iBAAiB,KAAK,IAAI,GAAG,iBAAiB,WAAW;AACvF,UAAM,eAAe,SAAS,iBAAiB,KAAK,IAAI,iBAAiB,aAAa,IAAI;AAC1F,QAAI,OAAO;AACX,WAAO,kBAAkB,cAAc,kBAAkB;AACvD,YAAM,IAAI,QAAQ;AAClB,cAAQ,IAAI;AAAA,IACd;AACA,WAAO;AAAA,EACT;AACA,WAAS,SAAS,GAAG,SAAS,MAAM,UAAU;AAC5C,UAAM,OAAO,kBAAkB,MAAM;AACrC,UAAM,MAAM,QAAQ,UAAU,KAAK,IAAI,OAAO,QAAQ,MAAM,CAAC,IAAI;AACjE,WAAO,UAAU;AAAA,EACnB;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,MAAM;AACzD;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,GAAG,IAAI;AACrB,QAAM,EAAE,aAAa,MAAM,OAAO,KAAK,IAAI;AAC3C,mBAAiB,IAAI,SAAS;AAC9B,QAAM,SAAS,aAAa,cAAc,GAAG,KAAK;AAClD,QAAM,WAAW,GAAG,MAAM;AAC1B,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,SAAS,IAAI,aAAa,MAAM;AACtC,QAAM,OAAO;AACb,WAAS,SAAS,GAAG,SAAS,MAAM,UAAU;AAC5C,UAAM,iBAAiB,SAAS;AAChC,UAAM,aAAa,SAAS,iBAAiB,KAAK,IAAI,GAAG,iBAAiB,WAAW;AACrF,UAAM,WAAW,SAAS,iBAAiB,KAAK,IAAI,UAAU,iBAAiB,cAAc,CAAC;AAC9F,QAAI,QAAQ;AACZ,aAAS,IAAI,YAAY,IAAI,UAAU,KAAK;AAC1C,eAAS,KAAK,IAAI,QAAQ,IAAI,CAAC;AAAA,IACjC;AACA,YAAQ,QAAQ,QAAQ;AACxB,aAAS,IAAI,YAAY,IAAI,UAAU,KAAK;AAC1C,UAAI,MAAM,KAAK,QAAQ,OAAO,QAAQ,KAAK,QAAQ,UAAU;AAC7D,UAAI,WAAW,GAAG;AAChB,eAAO,KAAK,IAAI,OAAO,CAAC,IAAI;AAAA,MAC9B;AACA,aAAO,SAAS;AAChB,aAAO,MAAM;AAAA,IACf;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,EAAE,OAAO,MAAM;AAC1D;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,kBAAkB,SAAS,IAAI;AACvC,QAAM,aAAa;AACnB,MAAI,SAAS,EAAE;AACf,QAAM,QAAQ,OAAO;AACrB,QAAM,WAAW,aAAa,eAAe,kBAAkB,MAAM;AACrE,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,QAAQ,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,MAAI,gBAAgB,MAAM;AACxB,UAAM,WAAW,IAAI,MAAM,KAAK;AAChC,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,eAAS,KAAK,OAAO,aAAa;AAAA,IACpC;AACA,YAAQ,cAAc,OAAO,QAAQ,EAAE,OAAO,cAAc,QAAQ;AACpE,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAC/D,aAAS;AAAA,EACX;AACA,mBAAiB,GAAG,KAAK;AACzB,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,aAAa,WAAW,IAAI,qBAAqB,0BAA0B,QAAQ,IAAI;AAC9F,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,SAAS,QAAQ,OAAO,YAAY,aAAa,EAAE,KAAK;AAC9D,QAAM,SAAS,WAAW,MAAM,QAAQ,aAAa,EAAE,KAAK;AAC5D,MAAI,WAAW;AACf,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,aAAa,QAAQ;AAChF,eAAW;AAAA,EACb;AACA,SAAO,EAAE,QAAQ,OAAO,UAAU,OAAO,EAAE,MAAM;AACnD;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,mBAAiB,GAAG,SAAS;AAC7B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY;AAClB,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AACjN,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,MAAI;AACJ,MAAI,SAAS,gBAAgB,KAAK,SAAS,iBAAiB,KAAK,aAAa,YAAY,SAAS,SAAS,SAAS,QAAQ,GAAG;AAC9H,UAAM,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACtD,OAAO;AACL,UAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,UAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,UAAM,UAAU,MAAM,SAAS,EAAE,OAAO,EAAE,OAAO,UAAU,UAAU,KAAK;AAC1E,UAAM,SAAS,eAAe,SAAS,UAAU,EAAE,OAAO,QAAQ,MAAM;AAAA,EAC1E;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACxE,mBAAiB,GAAG,WAAW;AAC/B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,iBAAiB,UAAU;AAC1H,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,SAAS,QAAQ,SAAS,EAAE,OAAO,EAAE,OAAO,aAAa,eAAe,EAAE,KAAK,GAAG,UAAU,KAAK;AACvG,SAAO,SAAS,eAAe,OAAO,OAAO,WAAW,OAAO,MAAM;AACvE;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,mBAAiB,CAAC,IAAI,MAAM,GAAG,eAAe;AAC9C,QAAM,WAAW,qBAAqB,kBAAkB,OAAO,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AACnH,QAAM,WAAW,SAAS,WAAW,MAAM;AAC3C,QAAM,YAAY,mBAAmB,UAAU,QAAQ;AACvD,QAAM,cAAc,SAAS;AAC7B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,uBAAuB,SAAS;AACtC,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,WAAW,uBAAuB,IAAI,SAAS,QAAQ;AAC7D,QAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,QAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,QAAM,KAAK,OAAO,OAAO,OAAO,SAAS;AACzC,QAAM,QAAQ,SAAS,WAAW,EAAE;AACpC,WAAS,QAAQ,GAAG,QAAQ,SAAS,WAAW,EAAE,OAAO;AACvD,aAAS,UAAU,GAAG,UAAU,SAAS,YAAY,EAAE,SAAS;AAC9D,eAAS,UAAU,GAAG,UAAU,SAAS,SAAS,EAAE,SAAS;AAC3D,iBAAS,QAAQ,GAAG,QAAQ,SAAS,UAAU,EAAE,OAAO;AACtD,mBAAS,QAAQ,GAAG,QAAQ,SAAS,SAAS,EAAE,OAAO;AACrD,kBAAM,gBAAgB,UAAU;AAChC,kBAAM,cAAc,QAAQ;AAC5B,kBAAM,cAAc,QAAQ;AAC5B,gBAAI,UAAU;AACd,qBAAS,SAAS,GAAG,SAAS,sBAAsB,UAAU,eAAe;AAC3E,oBAAM,WAAW,gBAAgB,UAAU;AAC3C,kBAAI,UAAU,KAAK,WAAW,SAAS,YAAY,KAAK,MAAM,OAAO,MAAM,SAAS;AAClF;AAAA,cACF;AACA,uBAAS,OAAO,GAAG,OAAO,uBAAuB,QAAQ,gBAAgB;AACvE,sBAAM,SAAS,cAAc,QAAQ;AACrC,oBAAI,QAAQ,KAAK,SAAS,SAAS,aAAa,KAAK,MAAM,KAAK,MAAM,OAAO;AAC3E;AAAA,gBACF;AACA,yBAAS,OAAO,GAAG,OAAO,sBAAsB,QAAQ,eAAe;AACrE,wBAAM,SAAS,cAAc,QAAQ;AACrC,sBAAI,QAAQ,KAAK,SAAS,SAAS,YAAY,KAAK,MAAM,KAAK,MAAM,OAAO;AAC1E;AAAA,kBACF;AACA,wBAAM,SAAS,uBAAuB,wBAAwB,uBAAuB,IAAI,UAAU,IAAI,OAAO,SAAS,OAAO,OAAO,OAAO;AAC5I,wBAAM,SAAS,SAAS,wBAAwB,uBAAuB,OAAO,uBAAuB;AACrG,wBAAM,OAAO,WAAW,SAAS,IAAI;AACrC,sBAAI,SAAS,GAAG;AACd;AAAA,kBACF;AACA,wBAAM,QAAQ,MAAM,IAAI,OAAO,SAAS,OAAO,OAAO,OAAO;AAC7D,6BAAW,QAAQ;AAAA,gBACrB;AAAA,cACF;AAAA,YACF;AACA,eAAG,IAAI,SAAS,OAAO,SAAS,OAAO,OAAO,OAAO;AAAA,UACvD;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,QAAQ,OAAO,IAAI;AACtC,QAAM,IAAI;AACV,mBAAiB,CAAC,QAAQ,MAAM,GAAG,aAAa;AAChD,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AAC9G,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,YAAY,OAAO,SAAS,UAAU,EAAE,OAAO,iBAAiB,SAAS,EAAE,OAAO,EAAE,OAAO,QAAQ,EAAE,MAAM;AACjH,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,wBAAwB,SAAS;AACvC,QAAM,uBAAuB,SAAS;AACtC,QAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,QAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,QAAM,KAAK,OAAO,EAAE,OAAO,SAAS;AACpC,QAAM,SAAS,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC5C,QAAM,QAAQ,OAAO,GAAG,OAAO,WAAW,MAAM;AAChD,WAAS,IAAI,GAAG,IAAI,SAAS,WAAW,EAAE,GAAG;AAC3C,aAAS,IAAI,GAAG,IAAI,SAAS,YAAY,EAAE,GAAG;AAC5C,eAAS,MAAM,GAAG,MAAM,SAAS,UAAU,EAAE,KAAK;AAChD,iBAAS,MAAM,GAAG,MAAM,SAAS,SAAS,EAAE,KAAK;AAC/C,gBAAM,YAAY,MAAM;AACxB,gBAAM,YAAY,MAAM;AACxB,cAAI,UAAU;AACd,mBAAS,KAAK,GAAG,KAAK,uBAAuB,MAAM,gBAAgB;AACjE,kBAAM,OAAO,YAAY,MAAM;AAC/B,gBAAI,MAAM,KAAK,OAAO,SAAS,aAAa,KAAK,MAAM,GAAG,MAAM,KAAK;AACnE;AAAA,YACF;AACA,qBAAS,KAAK,GAAG,KAAK,sBAAsB,MAAM,eAAe;AAC/D,oBAAM,OAAO,YAAY,MAAM;AAC/B,kBAAI,MAAM,KAAK,OAAO,SAAS,YAAY,KAAK,MAAM,GAAG,MAAM,KAAK;AAClE;AAAA,cACF;AACA,oBAAM,SAAS,wBAAwB,uBAAuB,IAAI,UAAU,IAAI,GAAG,KAAK,KAAK,CAAC;AAC9F,oBAAM,SAAS,KAAK,uBAAuB;AAC3C,oBAAM,OAAO,WAAW,SAAS,IAAI;AACrC,kBAAI,SAAS,GAAG;AACd;AAAA,cACF;AACA,oBAAM,QAAQ,MAAM,IAAI,GAAG,KAAK,KAAK,CAAC;AACtC,yBAAW,QAAQ;AAAA,YACrB;AAAA,UACF;AACA,aAAG,IAAI,SAAS,GAAG,KAAK,KAAK,CAAC;AAAA,QAChC;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,GAAG,OAAO,GAAG,OAAO,GAAG,MAAM;AAC9D;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,SAAS,QAAQ,OAAO,qBAAqB,UAAU;AACpF,QAAM,UAAU,aAAa,eAAe,MAAM;AAClD,QAAM,WAAW,MAAM,SAAS,QAAQ,OAAO,SAAS,UAAU,KAAK;AACvE,QAAM,eAAe,iBAAiB,SAAS,QAAQ,OAAO,UAAU,MAAM,mBAAmB;AACjG,SAAO,CAAC,SAAS,QAAQ,aAAa,MAAM;AAC9C;AAGA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,EAAE,IAAI;AACd,UAAM,EAAE,YAAY,SAAS,KAAK,MAAM,oBAAoB,IAAI;AAChE,UAAM,aAAa;AACnB,qBAAiB,GAAG,mBAAmB;AACvC,UAAM,SAAS,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,CAAC,GAAG,CAAC,GAAG,IAAI;AAClG,UAAM,CAAC,QAAQ,OAAO,IAAI,sBAAsB,QAAQ,EAAE,OAAO,EAAE,OAAO,qBAAqB,QAAQ;AACvG,UAAM,eAAe,WAAW,MAAM,QAAQ,SAAS,UAAU,EAAE,KAAK;AACxE,UAAM,gBAAgB,WAAW,MAAM,SAAS,SAAS,UAAU,EAAE,KAAK;AAC1E,WAAO;AAAA,MACL,EAAE,QAAQ,cAAc,OAAO,SAAS,UAAU,OAAO,EAAE,MAAM;AAAA,MACjE,EAAE,QAAQ,eAAe,OAAO,SAAS,UAAU,OAAO,QAAQ;AAAA,IACpE;AAAA,EACF;AACF;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACtD,QAAM,SAAS,qBAAqB,0BAA0B,EAAE,OAAO,IAAI;AAC3E,QAAM,cAAc,OAAO;AAC3B,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,YAAY,CAAC;AACnB,QAAM,mBAAmB,SAAS,eAAe,CAAC,GAAG,WAAW,IAAI,aAAa,CAAC,UAAU,CAAC,CAAC;AAC9F,YAAU,KAAK,gBAAgB;AAC/B,QAAM,KAAK,MAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AAClF,YAAU,KAAK,EAAE;AACjB,QAAM,MAAM,KAAK,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,iBAAiB,GAAG,SAAS,SAAS,CAAC;AAC9E,YAAU,KAAK,GAAG;AAClB,QAAM,SAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,SAAS,EAAE,CAAC;AACxF,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,mBAAiB,GAAG,KAAK;AACzB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,EAC1E;AACA,uBAAqB,2BAA2B,OAAO,MAAM,GAAG,MAAM,MAAM;AAC5E,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,GAAG,OAAO,IAAI;AAC7F,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,OAAO,aAAa,oBAAoB,aAAa,cAAc,QAAQ,GAAG,GAAG,KAAK;AAC5F,QAAM,QAAQ,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC3C,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,EAAE,GAAG;AACpC,UAAM,SAAS,IAAI;AACnB,QAAI,OAAO,MAAM;AACjB,aAAS,IAAI,GAAG,IAAI,YAAY,EAAE,GAAG;AACnC,YAAM,QAAQ,MAAM,SAAS;AAC7B,UAAI,OAAO,MAAM,KAAK,KAAK,QAAQ,MAAM;AACvC,eAAO;AAAA,MACT;AAAA,IACF;AACA,SAAK,KAAK;AAAA,EACZ;AACA,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,EAAE;AAAA,EAC3C;AACA,QAAM,SAAS,SAAS,eAAe,UAAU,GAAG,OAAO,IAAI;AAC/D,MAAI,UAAU;AACZ,UAAM,gBAAgB,qBAAqB,qBAAqB,UAAU,QAAQ;AAClF,UAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAC7G,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,UAAU,KAAK,IAAI;AAC3B,mBAAiB,GAAG,WAAW;AAC/B,QAAM,WAAW,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,KAAK,GAAG,EAAE;AACnE,QAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE;AACxC,QAAM,MAAM,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,EAAE;AACtD,QAAM,SAAS,SAAS,YAAY,IAAI;AACxC,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,aAAa,aAAa,cAAc,QAAQ;AACtD,QAAM,aAAa,SAAS;AAC5B,QAAM,gBAAgB,aAAa,eAAe,QAAQ;AAC1D,QAAM,UAAU,aAAa,uBAAuB,EAAE,OAAO,UAAU;AACvE,WAAS,IAAI,GAAG,IAAI,YAAY,KAAK;AACnC,QAAI,UAAU,aAAa,WAAW,GAAG,YAAY,aAAa;AAClE,aAAS,KAAK,GAAG,KAAK,YAAY,MAAM;AACtC,UAAI,QAAQ,MAAM,MAAM,KAAK;AAC3B,gBAAQ,MAAM,MAAM,MAAM,IAAI,QAAQ,MAAM;AAAA,MAC9C,WAAW,QAAQ,OAAO,IAAI,KAAK;AACjC,gBAAQ,OAAO,IAAI,MAAM,KAAK,IAAI,QAAQ,MAAM;AAAA,MAClD;AAAA,IACF;AACA,cAAU,QAAQ,IAAI,CAAC,GAAG,OAAO,IAAI,MAAM,GAAG;AAC9C,UAAM,UAAU,aAAa,WAAW,SAAS,OAAO,QAAQ;AAChE,YAAQ,KAAK,MAAM;AAAA,EACrB;AACA,QAAM,QAAQ,SAAS,MAAM,SAAS,UAAU,EAAE,KAAK;AACvD,SAAO,EAAE,QAAQ,OAAO,OAAO,UAAU,OAAO,EAAE,MAAM;AAC1D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU,6BAA6B,CAAC,QAAQ,WAAW;AAC7D,QAAM,MAAM,SAAS;AACrB,MAAI,SAAS,KAAK,SAAS,KAAK,UAAU,KAAK,UAAU,GAAG;AAC1D,WAAO;AAAA,EACT,OAAO;AACL,YAAQ,MAAM,UAAU;AAAA,EAC1B;AACF,CAAC;AACD,IAAI,OAAO,iBAAiB,KAAK,OAAO;AACxC,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,QAAQ,oBAAoB,CAAC;AAG/C,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,IAAI,IAAI;AAChB,QAAM,aAAa,OAAO,MAAM;AAChC,MAAI,OAAO;AACX,MAAI,SAAS,IAAI;AACf,WAAO,aAAa;AAAA,EACtB;AACA,MAAI,SAAS,aAAa,GAAG;AAC3B,UAAM,MAAM,4EAA4E,0BAA0B,MAAM;AAAA,EAC1H;AACA,QAAM,OAAO,aAAa,eAAe,CAAC,IAAI,GAAG,OAAO,KAAK;AAC7D,QAAM,WAAW,KAAK;AAAA,IACpB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,kBAAkB,MAAM,UAAU,MAAM;AAAA,EACnD,CAAC;AACD,QAAM,gBAAgB,qBAAqB,qBAAqB,SAAS,OAAO,IAAI;AACpF,QAAM,mBAAmB,SAAS,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AACjH,QAAM,IAAI,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,iBAAiB,GAAG,SAAS,SAAS,CAAC;AAChF,QAAM,IAAI,KAAK,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,SAAS,CAAC;AACtD,QAAM,SAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,MAAM,UAAU,MAAM,EAAE,CAAC;AACnG,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAC1G,QAAM,SAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,GAAG,GAAG,YAAY,GAAG,SAAS,SAAS,CAAC;AAC3E,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,gBAAgB;AACvD,WAAS,8BAA8B,CAAC;AACxC,WAAS,8BAA8B,CAAC;AACxC,WAAS,8BAA8B,MAAM;AAC7C,WAAS,8BAA8B,WAAW;AAClD,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,YAAY,MAAM,WAAW,IAAI;AACzC,mBAAiB,QAAQ,aAAa;AACtC,QAAM,gBAAgB,aAAa,SAAS,SAAS,EAAE,QAAQ,EAAE,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,GAAG,EAAE,CAAC;AAClH,QAAM,YAAY,cAAc,MAAM;AACtC,QAAM,YAAY,cAAc,MAAM;AACtC,QAAM,WAAW,SAAS,KAAK,IAAI,cAAc,MAAM,EAAE;AACzD,QAAM,WAAW,CAAC,WAAW,UAAU;AACvC,QAAM,UAAU,aAAa,oBAAoB,aAAa,cAAc,QAAQ,GAAG,OAAO;AAC9F,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,UAAM,SAAS,IAAI;AACnB,UAAM,MAAM,IAAI,aAAa,YAAY,CAAC;AAC1C,QAAI,KAAK,SAAS;AAClB,aAAS,QAAQ,GAAG,QAAQ,IAAI,QAAQ,EAAE,OAAO;AAC/C,UAAI,SAAS,IAAI,QAAQ,KAAK,SAAS,SAAS;AAAA,IAClD;AACA,UAAM,SAAS,YAAY,KAAK,KAAK,SAAS,CAAC;AAC/C,UAAM,YAAY,IAAI;AACtB,aAAS,WAAW,GAAG,WAAW,YAAY,EAAE,UAAU;AACxD,YAAM,IAAI,OAAO;AACjB,cAAQ,YAAY,YAAY,IAAI;AACpC,eAAS,QAAQ,GAAG,QAAQ,IAAI,QAAQ,SAAS;AAC/C,YAAI,IAAI,IAAI,QAAQ;AAClB,kBAAQ,YAAY,YAAY;AAChC;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,MAAI,CAAC,YAAY;AACf,aAAS,8BAA8B,aAAa;AAAA,EACtD;AACA,SAAO,SAAS,eAAe,UAAU,SAAS,OAAO;AAC3D;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,eAAe,IAAI;AACxD,mBAAiB,OAAO,mBAAmB;AAC3C,QAAM,YAAY,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAClD,QAAM,aAAa,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACpD,QAAM,EAAE,gBAAgB,IAAI,yBAAyB,WAAW,YAAY,eAAe,cAAc,cAAc;AACvH,SAAO,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AACnG;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,gBAAgB,mBAAmB,IAAI;AAC5E,mBAAiB,OAAO,yBAAyB;AACjD,QAAM,YAAY,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAClD,QAAM,aAAa,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACpD,QAAM,EAAE,iBAAiB,aAAa,IAAI,yBAAyB,WAAW,YAAY,eAAe,cAAc,gBAAgB,kBAAkB;AACzJ,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AAAA,IAC1F,SAAS,eAAe,CAAC,GAAG,SAAS,IAAI,WAAW,CAAC,YAAY,CAAC,CAAC;AAAA,EACrE;AACF;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,gBAAgB,aAAa,IAAI;AACtE,mBAAiB,OAAO,4BAA4B;AACpD,QAAM,YAAY,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAClD,QAAM,aAAa,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACpD,QAAM,mBAAmB;AACzB,QAAM,kBAAkB;AACxB,QAAM,oBAAoB;AAC1B,QAAM,kBAAkB;AACxB,QAAM,EAAE,iBAAiB,eAAe,IAAI,yBAAyB,WAAW,YAAY,kBAAkB,iBAAiB,mBAAmB,eAAe;AACjK,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AAAA,IAC1F,SAAS,eAAe,CAAC,eAAe,MAAM,GAAG,WAAW,IAAI,aAAa,cAAc,CAAC;AAAA,EAC9F;AACF;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,IAAI;AACpB,QAAM,EAAE,OAAO,OAAO,SAAS,SAAS,IAAI;AAC5C,mBAAiB,SAAS,QAAQ;AAClC,QAAM,cAAc,aAAa,cAAc,QAAQ,KAAK;AAC5D,QAAM,MAAM,IAAI,aAAa,cAAc,KAAK;AAChD,MAAI,KAAK,QAAQ;AACjB,QAAM,aAAa,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACrD,WAAS,QAAQ,GAAG,QAAQ,aAAa,EAAE,OAAO;AAChD,QAAI,WAAW,UAAU,KAAK,WAAW,SAAS,OAAO;AACvD,UAAI,QAAQ,QAAQ,WAAW,UAAU;AAAA,IAC3C;AAAA,EACF;AACA,SAAO,SAAS,eAAe,CAAC,GAAG,QAAQ,OAAO,KAAK,GAAG,OAAO,GAAG;AACtE;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,EAAE,UAAU,UAAU;AACxB,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE,WAAW,EAAE,UAAU,aAAa;AAClC,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC3E,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,WAAO;AAAA,EACT,OAAO;AACL,WAAO,MAAM,EAAE,SAAS,UAAU,OAAO,EAAE,OAAO,EAAE,OAAO,OAAO,GAAG,OAAO,EAAE,MAAM,EAAE,CAAC;AAAA,EACzF;AACF;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,EAAE,UAAU,UAAU;AACxB,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE,WAAW,EAAE,UAAU,aAAa;AAClC,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,UAAU,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC3E,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,WAAO;AAAA,EACT,OAAO;AACL,WAAO,MAAM,EAAE,SAAS,UAAU,OAAO,EAAE,OAAO,EAAE,OAAO,OAAO,GAAG,OAAO,EAAE,MAAM,EAAE,CAAC;AAAA,EACzF;AACF;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,YAAY,EAAE,QAAQ,EAAE,OAAO,OAAO,GAAG,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAAA,EAC9F;AACA,QAAM,QAAQ,OAAO,GAAG;AACxB,QAAM,QAAQ,OAAO,GAAG;AACxB,SAAO,QAAQ,CAAC,MAAM;AACpB,iBAAa,kBAAkB,OAAO,EAAE,OAAO,uDAAuD;AACtG,iBAAa,OAAO,UAAU,EAAE,OAAO,MAAM,uDAAuD;AAAA,EACtG,CAAC;AACD,QAAM,0BAA0B,CAAC;AACjC,QAAM,kBAAkB,OAAO,IAAI,CAAC,MAAM;AACxC,UAAM,YAAY,YAAY,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAC/F,4BAAwB,KAAK,SAAS;AACtC,WAAO;AAAA,EACT,CAAC;AACD,QAAM,SAAS,QAAQ,EAAE,QAAQ,iBAAiB,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AACtF,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,UAAU,cAAc,IAAI;AACpC,mBAAiB,GAAG,KAAK;AACzB,QAAM,WAAW,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,KAAK,GAAG,EAAE;AACnE,QAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE;AACxC,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,aAAa,aAAa,cAAc,QAAQ;AACtD,QAAM,aAAa,SAAS;AAC5B,QAAM,gBAAgB,aAAa,eAAe,QAAQ;AAC1D,QAAM,UAAU,aAAa,uBAAuB,EAAE,OAAO,UAAU;AACvE,MAAI,kBAAkB,GAAG;AACvB,YAAQ,KAAK,aAAa;AAAA,EAC5B;AACA,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,UAAU,aAAa,WAAW,GAAG,OAAO,QAAQ;AAC1D,UAAM,YAAY,QAAQ,IAAI,CAAC,GAAG,OAAO,IAAI,MAAM,GAAG;AACtD,UAAM,WAAW,aAAa,WAAW,WAAW,YAAY,aAAa;AAC7E,YAAQ,YAAY,MAAM;AAAA,EAC5B;AACA,QAAM,QAAQ,SAAS,MAAM,SAAS,UAAU,EAAE,KAAK;AACvD,SAAO,EAAE,QAAQ,OAAO,OAAO,UAAU,OAAO,EAAE,MAAM;AAC1D;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU,6BAA6B,CAAC,GAAG,MAAM,KAAK,IAAI,GAAG,CAAC,CAAC;AACnE,IAAI,OAAO,iBAAiB,KAAK,OAAO;AACxC,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,oBAAoB,mBAAmB,QAAQ,IAAI;AAC3D,QAAM,EAAE,iBAAiB,IAAI;AAC7B,QAAM,sBAAsB,mBAAmB,IAAI,CAAC,MAAM,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM;AAC5F,QAAM,4BAA4B,mBAAmB,IAAI,CAAC,MAAM,EAAE,KAAK;AACvE,QAAM,qBAAqB,SAAS,KAAK,IAAI,kBAAkB,MAAM,EAAE;AACvE,QAAM,WAAW,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACnD,QAAM,CAAC,oBAAoB,mBAAmB,sBAAsB,IAAI,iBAAiB,qBAAqB,2BAA2B,oBAAoB,kBAAkB,OAAO,kBAAkB,OAAO,UAAU,QAAQ,OAAO,gBAAgB;AACxP,QAAM,4BAA4B,mBAAmB,IAAI,CAAC,WAAW,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,SAAS,MAAM,CAAC;AAC9H,QAAM,0BAA0B,SAAS,eAAe,wBAAwB,kBAAkB,OAAO,iBAAiB;AAC1H,SAAO,0BAA0B,OAAO,CAAC,uBAAuB,CAAC;AACnE;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,QAAQ,QAAQ,OAAO,IAAI;AACnC,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,CAAC,oBAAoB,iBAAiB,IAAI,gBAAgB,SAAS,OAAO,OAAO,OAAO,OAAO,SAAS,OAAO,OAAO,SAAS,OAAO,KAAK;AACjJ,QAAM,iBAAiB,SAAS,eAAe,CAAC,mBAAmB,MAAM,GAAG,SAAS,kBAAkB;AACvG,QAAM,gBAAgB,SAAS,eAAe,CAAC,kBAAkB,MAAM,GAAG,OAAO,OAAO,iBAAiB;AACzG,SAAO,CAAC,gBAAgB,aAAa;AACvC;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,cAAc,oBAAoB,IAAI;AAC7D,QAAM,EAAE,kBAAkB,IAAI;AAC9B,QAAM,SAAS,SAAS,KAAK,IAAI,MAAM,MAAM,EAAE;AAC/C,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,gBAAgB,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE;AAC7D,QAAM,sBAAsB,oBAAoB,IAAI,CAAC,MAAM,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE,MAAM;AAC7F,QAAM,2BAA2B,oBAAoB,IAAI,CAAC,MAAM,EAAE,KAAK;AACvE,QAAM,CAAC,aAAa,MAAM,IAAI,yBAAyB,QAAQ,MAAM,OAAO,SAAS,OAAO,OAAO,OAAO,OAAO,eAAe,aAAa,OAAO,qBAAqB,0BAA0B,iBAAiB;AACpN,SAAO,SAAS,eAAe,aAAa,OAAO,OAAO,MAAM;AAClE;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,OAAO,MAAM,MAAM,IAAI;AAC5C,QAAM,SAAS,UAAU,OAAO,MAAM,OAAO,KAAK;AAClD,SAAO,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,OAAO,MAAM;AAC/D;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,gBAAgB,YAAY,CAAC,OAAO,IAAI,EAAE;AAC5D,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,mBAAiB,QAAQ,gBAAgB;AACzC,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,CAAC,OAAO,WAAW,UAAU,WAAW,IAAI,OAAO;AACzD,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,SAAS,IAAI,aAAa,aAAa,cAAc,CAAC,OAAO,WAAW,UAAU,WAAW,CAAC,CAAC;AACrG,QAAM,qBAAqB;AAAA,IACzB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,IAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,EAChD;AACA,QAAM,sBAAsB;AAAA,IAC1B,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,IAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,EAChD;AACA,MAAI,YAAY;AAChB,QAAM,wBAAwB,mBAAmB,KAAK,oBAAoB;AAC1E,QAAM,wBAAwB,mBAAmB,KAAK,oBAAoB;AAC1E,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,UAAI;AACJ,UAAI,kBAAkB;AACpB,wBAAgB,yBAAyB,IAAI,OAAO;AAAA,MACtD,OAAO;AACL,wBAAgB,wBAAwB;AAAA,MAC1C;AACA,YAAM,iBAAiB,KAAK,IAAI,GAAG,KAAK,MAAM,aAAa,CAAC;AAC5D,YAAM,UAAU,gBAAgB;AAChC,YAAM,gBAAgB,KAAK,IAAI,YAAY,GAAG,KAAK,KAAK,aAAa,CAAC;AACtE,YAAM,eAAe,IAAI,cAAc,KAAK,iBAAiB,cAAc;AAC3E,YAAM,eAAe,IAAI,cAAc,KAAK,gBAAgB,cAAc;AAC1E,eAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,YAAI;AACJ,YAAI,kBAAkB;AACpB,0BAAgB,yBAAyB,IAAI,OAAO;AAAA,QACtD,OAAO;AACL,0BAAgB,wBAAwB;AAAA,QAC1C;AACA,cAAM,iBAAiB,KAAK,IAAI,GAAG,KAAK,MAAM,aAAa,CAAC;AAC5D,cAAM,UAAU,gBAAgB;AAChC,cAAM,gBAAgB,KAAK,IAAI,WAAW,GAAG,KAAK,KAAK,aAAa,CAAC;AACrE,cAAM,gBAAgB,eAAe,iBAAiB,cAAc;AACpE,cAAM,gBAAgB,eAAe,iBAAiB,cAAc;AACpE,cAAM,iBAAiB,eAAe,gBAAgB,cAAc;AACpE,cAAM,iBAAiB,eAAe,gBAAgB,cAAc;AACpE,iBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,gBAAM,UAAU,QAAQ,gBAAgB;AACxC,gBAAM,aAAa,QAAQ,gBAAgB;AAC3C,gBAAM,WAAW,QAAQ,iBAAiB;AAC1C,gBAAM,cAAc,QAAQ,iBAAiB;AAC7C,gBAAM,MAAM,WAAW,WAAW,WAAW;AAC7C,gBAAM,SAAS,cAAc,cAAc,cAAc;AACzD,gBAAM,WAAW,OAAO,SAAS,OAAO;AACxC,iBAAO,eAAe;AAAA,QACxB;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,CAAC,OAAO,WAAW,UAAU,WAAW,GAAG,WAAW,MAAM;AAC7F;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,mBAAmB,MAAM;AAChC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,GAAG,IAAI;AACvB,QAAM,EAAE,aAAa,IAAI;AACzB,mBAAiB,CAAC,IAAI,MAAM,GAAG,oBAAoB;AACnD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,CAAC,OAAO,SAAS,QAAQ,KAAK,IAAI,OAAO;AAC/C,QAAM,CAAC,EAAE,SAAS,MAAM,IAAI,GAAG;AAC/B,QAAM,SAAS,IAAI,aAAa,QAAQ,UAAU,SAAS,KAAK;AAChE,QAAM,iBAAiB;AAAA,IACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,IAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,EAC5C;AACA,QAAM,iBAAiB;AAAA,IACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,IAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,EAC5C;AACA,QAAM,cAAc,eAAe,KAAK,eAAe;AACvD,QAAM,aAAa,eAAe,KAAK,eAAe;AACtD,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,MAAI,SAAS;AACb,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,UAAU,IAAI,cAAc;AAClC,aAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,YAAM,MAAM,IAAI;AAChB,YAAM,cAAc,KAAK,MAAM,GAAG;AAClC,YAAM,iBAAiB,KAAK,IAAI,KAAK,KAAK,GAAG,GAAG,UAAU,CAAC;AAC3D,YAAM,eAAe,UAAU,cAAc,cAAc;AAC3D,YAAM,kBAAkB,UAAU,iBAAiB,cAAc;AACjE,YAAM,UAAU,MAAM;AACtB,YAAM,iBAAiB,IAAI;AAC3B,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,cAAM,MAAM,IAAI;AAChB,cAAM,eAAe,KAAK,MAAM,GAAG;AACnC,cAAM,gBAAgB,KAAK,IAAI,KAAK,KAAK,GAAG,GAAG,SAAS,CAAC;AACzD,cAAM,UAAU,MAAM;AACtB,cAAM,iBAAiB,IAAI;AAC3B,cAAM,kBAAkB,eAAe,eAAe,cAAc;AACpE,cAAM,mBAAmB,eAAe,gBAAgB,cAAc;AACtE,cAAM,qBAAqB,kBAAkB,eAAe,cAAc;AAC1E,cAAM,sBAAsB,kBAAkB,gBAAgB,cAAc;AAC5E,cAAM,oCAAoC,iBAAiB;AAC3D,cAAM,6BAA6B,iBAAiB;AACpD,cAAM,6BAA6B,UAAU;AAC7C,cAAM,sBAAsB,UAAU;AACtC,iBAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,gBAAM,QAAQ,SAAS;AACvB,iBAAO,kBAAkB,MAAM,QAAQ;AACvC,iBAAO,mBAAmB,MAAM,QAAQ;AACxC,iBAAO,qBAAqB,MAAM,QAAQ;AAC1C,iBAAO,sBAAsB,MAAM,QAAQ;AAAA,QAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,CAAC,OAAO,QAAQ,SAAS,KAAK,GAAG,WAAW,MAAM;AACnF;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,uBAAuB,MAAM;AACpC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,mBAAiB,QAAQ,uBAAuB;AAChD,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,CAAC,OAAO,WAAW,UAAU,WAAW,IAAI,OAAO;AACzD,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,SAAS,IAAI,aAAa,QAAQ,YAAY,WAAW,WAAW;AAC1E,QAAM,qBAAqB;AAAA,IACzB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,IAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,EAChD;AACA,QAAM,sBAAsB;AAAA,IAC1B,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,IAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,EAChD;AACA,QAAM,wBAAwB,mBAAmB,KAAK,oBAAoB;AAC1E,QAAM,wBAAwB,mBAAmB,KAAK,oBAAoB;AAC1E,MAAI,eAAe;AACnB,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,cAAc,IAAI,cAAc;AACtC,aAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,YAAM,gBAAgB,mBAAmB,yBAAyB,IAAI,OAAO,wBAAwB;AACrG,UAAI,mBAAmB,KAAK,IAAI,YAAY,GAAG,eAAe,KAAK,MAAM,aAAa,IAAI,KAAK,MAAM,aAAa,CAAC;AACnH,UAAI,kBAAkB;AACpB,2BAAmB,KAAK,IAAI,GAAG,gBAAgB;AAAA,MACjD;AACA,YAAM,YAAY,cAAc,mBAAmB,cAAc;AACjE,eAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,cAAM,gBAAgB,mBAAmB,yBAAyB,IAAI,OAAO,wBAAwB;AACrG,YAAI,mBAAmB,KAAK,IAAI,WAAW,GAAG,eAAe,KAAK,MAAM,aAAa,IAAI,KAAK,MAAM,aAAa,CAAC;AAClH,YAAI,kBAAkB;AACpB,6BAAmB,KAAK,IAAI,GAAG,gBAAgB;AAAA,QACjD;AACA,cAAM,YAAY,YAAY,mBAAmB,cAAc;AAC/D,iBAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,gBAAM,SAAS,QAAQ,YAAY;AACnC,iBAAO,kBAAkB;AAAA,QAC3B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,CAAC,OAAO,WAAW,UAAU,WAAW,GAAG,OAAO,OAAO,MAAM;AAChG;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,0BAA0B,MAAM;AACvC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,GAAG,IAAI;AACvB,QAAM,EAAE,aAAa,IAAI;AACzB,mBAAiB,CAAC,IAAI,MAAM,GAAG,2BAA2B;AAC1D,QAAM,gBAAgB,aAAa,eAAe,OAAO,KAAK;AAC9D,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,QAAM,CAAC,OAAO,SAAS,QAAQ,KAAK,IAAI,OAAO;AAC/C,QAAM,CAAC,EAAE,SAAS,MAAM,IAAI,GAAG;AAC/B,QAAM,SAAS,IAAI,aAAa,QAAQ,UAAU,SAAS,KAAK;AAChE,QAAM,WAAW,SAAS,KAAK,IAAI,GAAG,MAAM,EAAE;AAC9C,QAAM,iBAAiB;AAAA,IACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,IAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,EAC5C;AACA,QAAM,iBAAiB;AAAA,IACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,IAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,EAC5C;AACA,QAAM,cAAc,eAAe,KAAK,eAAe;AACvD,QAAM,aAAa,eAAe,KAAK,eAAe;AACtD,QAAM,iBAAiB,IAAI;AAC3B,QAAM,gBAAgB,IAAI;AAC1B,QAAM,YAAY,KAAK,KAAK,cAAc,IAAI,IAAI;AAClD,QAAM,WAAW,KAAK,KAAK,aAAa,IAAI,IAAI;AAChD,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,UAAM,cAAc,IAAI,cAAc;AACtC,aAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,YAAM,YAAY,cAAc,IAAI,cAAc;AAClD,YAAM,aAAa,KAAK,MAAM,IAAI,cAAc;AAChD,YAAM,WAAW,KAAK,MAAM,aAAa,YAAY,CAAC;AACtD,eAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,cAAM,YAAY,YAAY,IAAI,cAAc;AAChD,cAAM,aAAa,KAAK,MAAM,IAAI,aAAa;AAC/C,cAAM,WAAW,KAAK,MAAM,aAAa,WAAW,CAAC;AACrD,iBAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,cAAI,QAAQ;AACZ,mBAAS,WAAW,GAAG,WAAW,WAAW,YAAY;AACvD,kBAAM,MAAM,WAAW;AACvB,gBAAI,MAAM,KAAK,OAAO,SAAS;AAC7B;AAAA,YACF;AACA,kBAAM,YAAY,cAAc,MAAM,UAAU;AAChD,kBAAM,gBAAgB,MAAM;AAC5B,kBAAM,mBAAmB,KAAK,IAAI,UAAU,GAAG,eAAe,KAAK,MAAM,aAAa,IAAI,KAAK,MAAM,aAAa,CAAC;AACnH,gBAAI,MAAM,kBAAkB;AAC1B;AAAA,YACF;AACA,qBAAS,WAAW,GAAG,WAAW,UAAU,YAAY;AACtD,oBAAM,MAAM,WAAW;AACvB,kBAAI,MAAM,KAAK,OAAO,QAAQ;AAC5B;AAAA,cACF;AACA,oBAAM,YAAY,YAAY,MAAM,UAAU;AAC9C,oBAAM,gBAAgB,MAAM;AAC5B,oBAAM,mBAAmB,KAAK,IAAI,SAAS,GAAG,eAAe,KAAK,MAAM,aAAa,IAAI,KAAK,MAAM,aAAa,CAAC;AAClH,kBAAI,MAAM,kBAAkB;AAC1B,yBAAS,SAAS,YAAY;AAAA,cAChC;AAAA,YACF;AAAA,UACF;AACA,iBAAO,YAAY,KAAK;AAAA,QAC1B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,MAAM;AACnE;AACA,IAAI,mCAAmC;AAAA,EACrC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,mBAAiB,GAAG,SAAS;AAC7B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,aAAa,eAAe,MAAM,EAAE,KAAK;AACvD,MAAI,UAAU,GAAG;AACf,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD;AACA,QAAM,SAAS,IAAI,aAAa,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,OAAO,SAAS,WAAW,CAAC;AAClC,WAAS,IAAI,GAAG,IAAI,OAAO,MAAM,KAAK;AACpC,UAAM,SAAS,OAAO,WAAW,CAAC;AAClC,UAAM,QAAQ,OAAO,MAAM;AAC3B,UAAM,QAAQ,CAAC,MAAM,MAAM,KAAK,EAAE,MAAM,KAAK,IAAI,MAAM,EAAE;AACzD,WAAO,IAAI,KAAK,IAAI,GAAG,KAAK,GAAG,GAAG,MAAM;AAAA,EAC1C;AACA,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC1E;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,yBAAyB;AAAA,EAC3B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,EAAE,SAAS,WAAW,OAAO,IAAI;AACvC,UAAM,aAAa;AACnB,UAAM,SAAS,aAAa,uBAAuB,OAAO,OAAO,aAAa,cAAc,OAAO,KAAK,CAAC;AACzG,UAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,UAAM,CAAC,SAAS,OAAO,IAAI,qBAAqB,eAAe,QAAQ,aAAa,UAAU;AAC9F,UAAM,mBAAmB;AACzB,UAAM,YAAY,KAAK,IAAI,OAAO;AAClC,UAAM,YAAY,KAAK,IAAI,OAAO;AAClC,UAAM,YAAY,WAAW,KAAK,IAAI,OAAO,MAAM,EAAE;AACrD,aAAS,WAAW,GAAG,WAAW,OAAO,YAAY;AACnD,YAAM,cAAc,WAAW,aAAa,cAAc;AAC1D,eAAS,MAAM,GAAG,MAAM,aAAa,OAAO;AAC1C,cAAM,YAAY,OAAO,aAAa;AACtC,iBAAS,MAAM,GAAG,MAAM,YAAY,OAAO;AACzC,gBAAM,YAAY,MAAM;AACxB,mBAAS,UAAU,GAAG,UAAU,aAAa,WAAW;AACtD,kBAAM,UAAU,CAAC,OAAO,KAAK,KAAK,OAAO;AACzC,kBAAM,IAAI,QAAQ;AAClB,kBAAM,IAAI,QAAQ;AAClB,gBAAI,UAAU,IAAI,WAAW,aAAa,IAAI,WAAW;AACzD,gBAAI,UAAU,IAAI,WAAW,aAAa,IAAI,WAAW;AACzD,qBAAS,KAAK,MAAM,SAAS,OAAO;AACpC,qBAAS,KAAK,MAAM,SAAS,OAAO;AACpC,gBAAI,cAAc;AAClB,gBAAI,OAAO,cAAc,UAAU;AACjC,kBAAI,YAAY,GAAG;AACjB,8BAAc;AAAA,cAChB,OAAO;AACL,8BAAc,UAAU;AAAA,cAC1B;AAAA,YACF;AACA,gBAAI,UAAU,KAAK,SAAS,cAAc,UAAU,KAAK,SAAS,aAAa;AAC7E,oBAAM,mBAAmB,UAAU,aAAa;AAChD,oBAAM,mBAAmB,SAAS;AAClC,oBAAM,WAAW,cAAc,mBAAmB,mBAAmB;AACrE,4BAAc,UAAU;AAAA,YAC1B;AACA,kBAAM,SAAS,cAAc,YAAY,YAAY;AACrD,mBAAO,UAAU;AAAA,UACnB;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,UAAM,SAAS,WAAW,MAAM,QAAQ,OAAO,OAAO,OAAO,KAAK;AAClE,WAAO,EAAE,QAAQ,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAAA,EAC5D;AACF;AAGA,IAAI,SAAS,gBAAgB,OAAO,CAAC,OAAO;AAC1C,QAAM,OAAO,KAAK,MAAM,EAAE;AAC1B,MAAI,KAAK,OAAO,KAAK;AACnB,WAAO,KAAK,MAAM,EAAE;AAAA,EACtB,WAAW,KAAK,OAAO,KAAK;AAC1B,WAAO,KAAK,KAAK,EAAE;AAAA,EACrB,OAAO;AACL,QAAI,OAAO,MAAM,GAAG;AAClB,aAAO;AAAA,IACT,OAAO;AACL,aAAO,OAAO;AAAA,IAChB;AAAA,EACF;AACF,CAAC;AACD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,SAAS,QAAQ,IAAI;AAC7B,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW,IAAI,qBAAqB,gBAAgB,SAAS,SAAS,KAAK;AAC9H,QAAM,iBAAiB;AACvB,QAAM,aAAa,SAAS,WAAW,OAAO;AAC9C,QAAM,aAAa,SAAS,WAAW,OAAO;AAC9C,QAAM,SAAS,YAAY,YAAY,YAAY,OAAO,YAAY,WAAW,YAAY,WAAW,SAAS,GAAG,cAAc;AAClI,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,MAAM;AACnE;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,QAAQ,OAAO;AAClC,MAAI,OAAO;AACX,MAAI,QAAQ,OAAO;AACnB,MAAI,MAAM;AACV,SAAO,OAAO,OAAO;AACnB,UAAM,KAAK,OAAO,OAAO,SAAS,CAAC;AACnC,QAAI,OAAO,OAAO,OAAO;AACvB,aAAO,MAAM;AAAA,IACf,OAAO;AACL,cAAQ;AAAA,IACV;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,YAAY,QAAQ,OAAO;AAClC,MAAI,OAAO;AACX,MAAI,QAAQ,OAAO;AACnB,MAAI,MAAM;AACV,SAAO,OAAO,OAAO;AACnB,UAAM,KAAK,OAAO,OAAO,SAAS,CAAC;AACnC,QAAI,OAAO,QAAQ,OAAO;AACxB,aAAO,MAAM;AAAA,IACf,OAAO;AACL,cAAQ;AAAA,IACV;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,cAAc,QAAQ,WAAW,WAAW,WAAW,MAAM;AACrF,QAAM,SAAS,aAAa,kBAAkB,SAAS,YAAY,SAAS;AAC5E,WAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,UAAM,oBAAoB,aAAa,MAAM,IAAI,YAAY,IAAI,KAAK,SAAS;AAC/E,UAAM,eAAe,IAAI;AACzB,aAAS,IAAI,GAAG,IAAI,WAAW,EAAE,GAAG;AAClC,aAAO,eAAe,KAAK,SAAS,SAAS,YAAY,mBAAmB,OAAO,IAAI,aAAa,IAAI,YAAY,mBAAmB,OAAO,IAAI,aAAa;AAAA,IACjK;AAAA,EACF;AACA,SAAO;AACT;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,gBAAgB,OAAO,IAAI;AACnC,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,kBAAkB,SAAS,KAAK,IAAI,eAAe,MAAM,EAAE;AACjE,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,SAAS,iBAAiB,iBAAiB,SAAS,eAAe,MAAM,IAAI,eAAe,MAAM,IAAI,OAAO,MAAM,IAAI,IAAI;AACjI,SAAO,SAAS,eAAe,OAAO,OAAO,SAAS,MAAM;AAC9D;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,WAAW,GAAG,EAAE,IAAI;AAC5B,mBAAiB,CAAC,WAAW,GAAG,CAAC,GAAG,QAAQ;AAC5C,QAAM,gBAAgB,UAAU,MAAM;AACtC,QAAM,SAAS,SAAS,KAAK,IAAI,UAAU,MAAM,EAAE;AACnD,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,UAAU,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC5C,QAAM,cAAc,WAAW,EAAE,OAAO,EAAE,KAAK;AAC/C,QAAM,YAAY,aAAa,oBAAoB,aAAa,cAAc,EAAE,KAAK,GAAG,WAAW;AACnG,MAAI,QAAQ;AACZ,QAAM,SAAS,kBAAkB,KAAK,gBAAgB,KAAK,EAAE,MAAM,WAAW,IAAI,IAAI,aAAa,cAAc,EAAE,MAAM,MAAM,CAAC,CAAC;AACjI,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,aAAS,IAAI,GAAG,IAAI,QAAQ,KAAK;AAC/B,UAAI,OAAO,OAAO,GAAG;AACnB,kBAAU,WAAW,QAAQ;AAAA,MAC/B,OAAO;AACL,kBAAU,WAAW,QAAQ;AAAA,MAC/B;AAAA,IACF;AAAA,EACF;AACA,SAAO,SAAS,eAAe,EAAE,OAAO,aAAa,SAAS;AAChE;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa,qBAAqB;AACtC,IAAI,QAAQ,qBAAqB;AACjC,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO;AACxC,MAAI,MAAM,GAAG;AACX,WAAO,QAAQ;AAAA,EACjB,OAAO;AACL,WAAO,cAAc,KAAK,IAAI,EAAE,IAAI;AAAA,EACtC;AACF,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO;AACxC,MAAI,KAAK,GAAG;AACV,WAAO;AAAA,EACT,WAAW,KAAK,GAAG;AACjB,WAAO;AAAA,EACT,OAAO;AACL,WAAO;AAAA,EACT;AACF,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,gBAAgB,KAAK,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;AACpD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,WAAW;AACf,IAAI,aAAa,KAAK,IAAI,QAAQ,IAAI;AACtC,IAAI,YAAY,gBAAgB,UAAU,CAAC,OAAO;AAChD,QAAM,WAAW,KAAK,CAAC;AACvB,QAAM,WAAW,KAAK;AACtB,QAAM,OAAO,KAAK,IAAI,EAAE;AACxB,MAAI;AACJ,MAAI,UAAU;AACZ,aAAS;AAAA,EACX,WAAW,UAAU;AACnB,aAAS;AAAA,EACX,OAAO;AACL,aAAS,KAAK,IAAI,IAAI,IAAI;AAAA,EAC5B;AACA,SAAO;AACT,CAAC;AACD,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,IAAI;AACjC,mBAAiB,CAAC,CAAC,GAAG,gBAAgB;AACtC,QAAM,QAAQ,aAAa,cAAc,UAAU;AACnD,QAAM,mBAAmB,CAAC,CAAC,GAAG,CAAC,CAAC;AAChC,mBAAiB,KAAK,GAAG,QAAQ;AACjC,WAAS,IAAI,IAAI,WAAW,QAAQ,IAAI,EAAE,MAAM,QAAQ,EAAE,GAAG;AAC3D,qBAAiB,KAAK,CAAC,GAAG,CAAC,CAAC;AAAA,EAC9B;AACA,QAAM,UAAU,YAAY,WAAW;AAAA,IACrC,QAAQ,EAAE,EAAE;AAAA,IACZ,SAAS;AAAA,IACT,OAAO,EAAE,UAAU,kBAAkB,eAAe,EAAE;AAAA,EACxD,CAAC;AACD,QAAM,sBAAsB,qBAAqB,YAAY,QAAQ,OAAO,YAAY,OAAO,KAAK;AACpG,QAAM,oCAAoC,qBAAqB,YAAY,oBAAoB,QAAQ,WAAW,QAAQ,KAAK;AAC/H,QAAM,eAAe,qBAAqB,oBAAoB,QAAQ,OAAO,YAAY,OAAO,KAAK;AACrG,QAAM,gBAAgB,EAAE,GAAG,QAAQ;AACnC,QAAM,eAAe,EAAE,OAAO,oBAAoB;AAClD,QAAM,kBAAkB,SAAS,EAAE,QAAQ,eAAe,SAAS,UAAU,OAAO,aAAa,CAAC;AAClG,QAAM,kBAAkB,EAAE,GAAG,gBAAgB;AAC7C,QAAM,iBAAiB,EAAE,MAAM,kCAAkC;AACjE,QAAM,WAAW,WAAW,EAAE,QAAQ,iBAAiB,SAAS,UAAU,OAAO,eAAe,CAAC;AACjG,QAAM,sBAAsB,EAAE,GAAG,SAAS;AAC1C,QAAM,qBAAqB,EAAE,OAAO,aAAa;AACjD,QAAM,SAAS,SAAS,EAAE,QAAQ,qBAAqB,SAAS,UAAU,OAAO,mBAAmB,CAAC;AACrG,WAAS,8BAA8B,OAAO;AAC9C,WAAS,8BAA8B,eAAe;AACtD,WAAS,8BAA8B,QAAQ;AAC/C,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,SAAS,QAAQ,YAAY,aAAa,IAAI;AACtD,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,UACV,WAAW,OAAO;AAAA,EAC1B;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,UACV,QAAQ,OAAO;AAAA,EACvB;AACA,MAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,UAAM,IAAI,MAAM;AAAA,UACV,OAAO,OAAO;AAAA,EACtB;AACA,MAAI,aAAa,MAAM,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM;AAAA,UACV,aAAa,OAAO;AAAA,EAC5B;AACA,QAAM,WAAW,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACnD,QAAM,UAAU,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACjD,QAAM,cAAc,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE;AACzD,QAAM,gBAAgB,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE,OAAO;AACpE,QAAM,CAAC,eAAe,oBAAoB,cAAc,mBAAmB,eAAe,IAAI,wBAAwB,UAAU,QAAQ,OAAO,QAAQ,OAAO,SAAS,OAAO,OAAO,aAAa,aAAa;AAC/M,SAAO;AAAA,IACL,SAAS,eAAe,oBAAoB,QAAQ,OAAO,aAAa;AAAA,IACxE,SAAS,eAAe,CAAC,mBAAmB,EAAE,GAAG,OAAO,OAAO,YAAY;AAAA,IAC3E,SAAS,eAAe,CAAC,kBAAkB,MAAM,GAAG,QAAQ,IAAI,WAAW,kBAAkB,IAAI,CAAC,UAAU,OAAO,KAAK,CAAC,CAAC,CAAC;AAAA,IAC3H,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,QAAQ,OAAO,IAAI,WAAW,eAAe,CAAC;AAAA,EAClG;AACF;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,cAAc,YAAY,SAAS,IAAI;AAC/C,MAAI,aAAa,MAAM,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM;AAAA,UACV,aAAa,OAAO;AAAA,EAC5B;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,UACV,WAAW,OAAO;AAAA,EAC1B;AACA,MAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,UAAM,IAAI,MAAM,sDAAsD,SAAS,OAAO;AAAA,EACxF;AACA,QAAM,cAAc,MAAM,KAAK,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE,MAAM;AAC1E,QAAM,gBAAgB,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE;AAC7D,QAAM,cAAc,MAAM,KAAK,SAAS,KAAK,IAAI,SAAS,MAAM,EAAE,MAAM;AACxE,QAAM,CAAC,YAAY,cAAc,WAAW,IAAI,kBAAkB,eAAe,aAAa,OAAO,aAAa,OAAO,aAAa,WAAW;AACjJ,SAAO;AAAA,IACL,SAAS,eAAe,cAAc,aAAa,OAAO,UAAU;AAAA,IACpE,SAAS,eAAe,CAAC,YAAY,MAAM,GAAG,SAAS,OAAO,IAAI,WAAW,WAAW,CAAC;AAAA,EAC3F;AACF;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,mBAAmB,MAAM;AAChC,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,SAAS,WAAW,IAAI;AACtC,MAAI,KAAK,MAAM,SAAS,GAAG;AACzB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,YACR,QAAQ,OAAO;AAAA,EACzB;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,YACR,WAAW,OAAO;AAAA,EAC5B;AACA,MAAI,QAAQ,MAAM,OAAO,WAAW,MAAM,IAAI;AAC5C,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,QAAQ,SAAS,KAAK,IAAI,KAAK,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACnD,QAAM,cAAc,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE;AACzD,QAAM,CAAC,YAAY,eAAe,IAAI,2BAA2B,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,aAAa,IAAI;AAC3H,SAAO,SAAS,eAAe,iBAAiB,KAAK,OAAO,UAAU;AACxE;AACA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,kBAAkB,MAAM;AAC/B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,SAAS,WAAW,IAAI;AACtC,MAAI,KAAK,MAAM,SAAS,GAAG;AACzB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,WACT,QAAQ,OAAO;AAAA,EACxB;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,WACT,WAAW,OAAO;AAAA,EAC3B;AACA,MAAI,QAAQ,MAAM,OAAO,WAAW,MAAM,IAAI;AAC5C,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,QAAQ,SAAS,KAAK,IAAI,KAAK,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,KAAK,IAAI,QAAQ,MAAM,EAAE;AACnD,QAAM,cAAc,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE;AACzD,QAAM,CAAC,YAAY,eAAe,IAAI,2BAA2B,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,WAAW;AACrH,SAAO,SAAS,eAAe,iBAAiB,KAAK,OAAO,UAAU;AACxE;AACA,IAAI,yBAAyB;AAAA,EAC3B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,eAAe,cAAc,aAAa,IAAI;AACtD,QAAM,EAAE,YAAY,IAAI;AACxB,QAAM,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW,IAAI,qBAAqB,gBAAgB,cAAc,eAAe,WAAW;AAC/I,QAAM,iBAAiB;AACvB,QAAM,aAAa,SAAS,WAAW,aAAa;AACpD,MAAI;AACJ,UAAQ,aAAa,OAAO;AAAA,IAC1B,KAAK,QAAQ;AACX,YAAM,aAAa,SAAS,WAAW,YAAY;AACnD,YAAM,gBAAgB,QAAQ,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE,OAAO,EAAE;AAC9E,eAAS,YAAY,YAAY,YAAY,aAAa,YAAY,WAAW,YAAY,WAAW,SAAS,eAAe,cAAc;AAC9I;AAAA,IACF;AAAA,IACA,KAAK,WAAW;AACd,YAAM,aAAa,SAAS,WAAW,YAAY;AACnD,YAAM,gBAAgB,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE,OAAO;AACpE,eAAS,YAAY,YAAY,YAAY,aAAa,YAAY,WAAW,YAAY,WAAW,SAAS,eAAe,cAAc;AAC9I;AAAA,IACF;AAAA,IACA,KAAK,SAAS;AACZ,YAAM,aAAa,SAAS,WAAW,YAAY;AACnD,YAAM,gBAAgB,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE,OAAO;AACpE,eAAS,YAAY,YAAY,YAAY,aAAa,YAAY,WAAW,YAAY,WAAW,SAAS,eAAe,cAAc;AAC9I;AAAA,IACF;AAAA,IACA,KAAK,UAAU;AACb,YAAM,aAAa,SAAS,WAAW,YAAY;AACnD,YAAM,gBAAgB,aAAa,aAAa,SAAS,KAAK,IAAI,aAAa,MAAM,EAAE,OAAO,EAAE;AAChG,eAAS,YAAY,YAAY,YAAY,aAAa,YAAY,WAAW,YAAY,WAAW,SAAS,eAAe,cAAc;AAC9I;AAAA,IACF;AAAA,IACA;AACE,YAAM,IAAI,MAAM,oBAAoB,aAAa,OAAO;AAAA,EAC5D;AACA,SAAO,SAAS,eAAe,aAAa,OAAO,OAAO,OAAO,MAAM;AACzE;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,iBAAiB,KAAK,IAAI;AAClC,QAAM,QAAQ,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AACzD,QAAM,aAAa,qBAAqB,iBAAiB,GAAG,iBAAiB,KAAK;AAClF,QAAM,QAAQ,IAAI,MAAM,EAAE,MAAM,MAAM,EAAE,KAAK,CAAC;AAC9C,QAAM,OAAO,EAAE,MAAM,MAAM;AAC3B,SAAO,WAAW,IAAI,CAAC,MAAM;AAC3B,UAAM,YAAY,CAAC,GAAG,IAAI;AAC1B,cAAU,SAAS;AACnB,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,MAAM,UAAU,EAAE,CAAC;AAC7F,UAAM,UAAU;AAChB,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AAC7C,UAAM,EAAE,EAAE,IAAI;AACd,UAAM,aAAa;AACnB,qBAAiB,GAAG,QAAQ;AAC5B,UAAM,SAAS,WAAW,KAAK,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,YAAY,IAAI,aAAa,OAAO,MAAM;AAChD,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,YAAM,QAAQ,OAAO;AACrB,gBAAU,KAAK,QAAQ;AAAA,IACzB;AACA,UAAM,SAAS,WAAW,MAAM,WAAW,EAAE,OAAO,EAAE,KAAK;AAC3D,WAAO,EAAE,QAAQ,OAAO,EAAE,OAAO,OAAO,EAAE,MAAM;AAAA,EAClD;AACF;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,IAAI,UAAU;AAC/C,QAAM,YAAY;AAClB,MAAI,MAAM,EAAE,GAAG;AACb,WAAO;AAAA,EACT,OAAO;AACL,WAAO,KAAK,IAAI,IAAI,UAAU;AAAA,EAChC;AACF,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,eAAe,IAAI;AAC/F,mBAAiB,GAAG,cAAc;AAClC,QAAM,EAAE,kBAAkB,YAAY,YAAY,WAAW,eAAe,OAAO,QAAQ,KAAK,MAAM,SAAS,SAAS,IAAI,mBAAmB,UAAU,EAAE,OAAO,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,cAAc;AACpP,MAAI;AACJ,MAAI,YAAY;AACd,aAAS,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAAA,EACtF,WAAW,aAAa,eAAe;AACrC,iBAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,yCAAyC,EAAE,MAAM,QAAQ;AACxG,UAAM,OAAO,mBAAmB,gBAAgB,QAAQ,MAAM,QAAQ;AACtE,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,QAAQ,KAAK,EAAE,CAAC;AAC1F,aAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAC5F,aAAS,8BAA8B,MAAM;AAAA,EAC/C,OAAO;AACL,UAAM,OAAO,SAAS,WAAW,CAAC;AAClC,UAAM,SAAS,iBAAiB,kBAAkB,MAAM,UAAU,MAAM;AACxE,aAAS,SAAS,eAAe,YAAY,OAAO,OAAO,OAAO,MAAM;AAAA,EAC1E;AACA,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,WAAW,aAAa,SAAS,UAAU,WAAW,UAAU,uBAAuB,IAAI;AACnG,QAAM,EAAE,MAAM,WAAW,IAAI;AAC7B,QAAM,QAAQ,SAAS,KAAK,IAAI,KAAK,MAAM,EAAE;AAC7C,QAAM,cAAc,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE;AACzD,QAAM,CAAC,QAAQ,YAAY,IAAI,iBAAiB,OAAO,aAAa,WAAW,aAAa,SAAS,WAAW,UAAU,sBAAsB;AAChJ,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,UAAU,MAAM;AAAA,IACzD,SAAS,eAAe,WAAW,OAAO,SAAS,YAAY;AAAA,EACjE;AACF;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,UAAU,IAAI;AACtB,QAAM,EAAE,OAAO,QAAQ,UAAU,IAAI;AACrC,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AACA,MAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,UAAM,IAAI,MAAM,sCAAsC,OAAO,OAAO;AAAA,EACtE;AACA,MAAI,UAAU,MAAM,WAAW,GAAG;AAChC,UAAM,IAAI,MAAM,0CAA0C,UAAU,OAAO;AAAA,EAC7E;AACA,QAAM,SAAS,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAChD,QAAM,aAAa,SAAS,KAAK,IAAI,UAAU,MAAM,EAAE,OAAO;AAC9D,QAAM,CAAC,SAAS,QAAQ,KAAK,IAAI,gBAAgB,QAAQ,YAAY,SAAS;AAC9E,QAAM,aAAa,OAAO;AAC1B,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,YAAY,CAAC,GAAG,SAAS,OAAO;AAAA,IACzD,SAAS,eAAe,CAAC,UAAU,GAAG,UAAU,MAAM;AAAA,IACtD,SAAS,eAAe,CAAC,CAAC,GAAG,SAAS,IAAI,WAAW,KAAK,CAAC;AAAA,EAC7D;AACF;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,wBAAwB,MAAM;AACrC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,WAAW,IAAI;AACvB,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AACA,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,QAAM,SAAS,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AAChD,QAAM,SAAS,2BAA2B,QAAQ,UAAU;AAC5D,SAAO,SAAS,eAAe,OAAO,OAAO,SAAS,MAAM;AAC9D;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,gBAAgB,KAAK,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;AACpD,IAAI,YAAY;AAAA,EACd,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,gBAAgB,MAAM,CAAC,OAAO,KAAK,KAAK,EAAE,CAAC;AACvD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,mBAAiB,GAAG,MAAM;AAC1B,QAAM,SAAS,SAAS,SAAS,WAAW,CAAC,GAAG,IAAI;AACpD,SAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC1E;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,mBAAiB,GAAG,MAAM;AAC1B,QAAM,QAAQ,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC1C,QAAM,CAAC,aAAa,cAAc,IAAI,SAAS,OAAO,EAAE,OAAO,EAAE,OAAO,GAAG,MAAM;AACjF,SAAO;AAAA,IACL,SAAS,eAAe,YAAY,OAAO,YAAY,OAAO,YAAY,MAAM;AAAA,IAChF,SAAS,eAAe,eAAe,OAAO,eAAe,OAAO,eAAe,MAAM;AAAA,EAC3F;AACF;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,WAAW,IAAI;AACtC,QAAM,EAAE,eAAe,UAAU,WAAW,YAAY,IAAI;AAC5D,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,QAAM,CAAC,WAAW,QAAQ,IAAI,eAAe,OAAO,cAAc,CAAC,aAAa,UAAU;AAC1F,QAAM,WAAW,CAAC,OAAO,WAAW,UAAU,WAAW;AACzD,QAAM,YAAY,aAAa,eAAe,OAAO,KAAK;AAC1D,QAAM,gBAAgB,UAAU;AAChC,QAAM,cAAc,UAAU;AAC9B,QAAM,cAAc,UAAU;AAC9B,QAAM,aAAa,aAAa,eAAe,QAAQ;AACvD,QAAM,iBAAiB,WAAW;AAClC,QAAM,eAAe,WAAW;AAChC,QAAM,eAAe,WAAW;AAChC,QAAM,UAAU,aAAa,uBAAuB,OAAO,OAAO,aAAa,cAAc,QAAQ,CAAC;AACtG,UAAQ,KAAK,SAAS;AACtB,QAAM,YAAY,SAAS,KAAK,IAAI,OAAO,MAAM,EAAE;AACnD,QAAM,gBAAgB,SAAS,KAAK,IAAI,WAAW,MAAM,EAAE;AAC3D,WAAS,IAAI,GAAG,IAAI,OAAO,EAAE,GAAG;AAC9B,UAAM,aAAa,WAAW,MAAM,OAAO,IAAI,gBAAgB,cAAc,SAAS,IAAI,GAAG,IAAI,IAAI,CAAC;AACtG,aAAS,OAAO,GAAG,OAAO,WAAW,EAAE,MAAM;AAC3C,eAAS,OAAO,GAAG,OAAO,UAAU,EAAE,MAAM;AAC1C,iBAAS,UAAU,GAAG,UAAU,aAAa,EAAE,SAAS;AACtD,cAAI;AACJ,gBAAM,aAAa,WAAW,KAAK,OAAO,WAAW,KAAK,OAAO;AACjE,cAAI,eAAe,GAAG;AACpB;AAAA,UACF;AACA,gBAAM,OAAO,WAAW,KAAK,OAAO,WAAW,KAAK,OAAO,WAAW,MAAM;AAC5E,gBAAM,OAAO,WAAW,KAAK,OAAO,WAAW,KAAK,OAAO,WAAW,MAAM;AAC5E,gBAAM,IAAI,SAAS,KAAK,YAAY,QAAQ;AAC5C,gBAAM,IAAI,SAAS,KAAK,aAAa,QAAQ;AAC7C,kBAAQ,eAAe;AAAA,YACrB,KAAK;AACH,oBAAM,qBAAqB,WAAW,aAAa,YAAY,eAAe,aAAa,aAAa,GAAG,GAAG,GAAG,SAAS,SAAS;AACnI;AAAA,YACF,KAAK;AACH,oBAAM,sBAAsB,WAAW,aAAa,YAAY,eAAe,aAAa,aAAa,GAAG,GAAG,GAAG,SAAS,SAAS;AACpI;AAAA,YACF;AACE,oBAAM,IAAI,MAAM,+DAA+D,eAAe;AAAA,UAClG;AACA,gBAAM,MAAM,IAAI,iBAAiB,OAAO,eAAe,OAAO,eAAe;AAC7E,kBAAQ,OAAO;AAAA,QACjB;AAAA,MACF;AAAA,IACF;AACA,WAAO,SAAS,eAAe,UAAU,OAAO,OAAO,OAAO;AAAA,EAChE;AACA,QAAM,SAAS,SAAS,MAAM,SAAS,UAAU,OAAO,KAAK;AAC7D,SAAO,EAAE,QAAQ,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC5D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AACA,SAAS,SAAS,UAAU,KAAK,MAAM;AACrC,UAAQ,MAAM;AAAA,IACZ,KAAK;AACH,aAAO,gBAAgB,UAAU,GAAG;AAAA,IACtC,KAAK;AACH,aAAO,aAAa,UAAU,GAAG;AAAA,IACnC,KAAK;AACH,aAAO,gBAAgB,UAAU,GAAG;AAAA,IACtC,KAAK;AAAA,IACL;AACE,aAAO,iBAAiB,UAAU,GAAG;AAAA,EACzC;AACF;AACA,SAAS,gBAAgB,UAAU,KAAK;AACtC,MAAI,UAAU;AACd,MAAI,UAAU,GAAG;AACf,QAAI,OAAO,GAAG;AACZ,gBAAU;AAAA,IACZ,OAAO;AACL,YAAM,MAAM,IAAI;AAChB,UAAI,UAAU,KAAK;AACjB,kBAAU,MAAM,KAAK,MAAM,CAAC,UAAU,GAAG,IAAI;AAAA,MAC/C;AACA,gBAAU,UAAU,CAAC,MAAM,UAAU,MAAM,CAAC,UAAU;AAAA,IACxD;AAAA,EACF,WAAW,UAAU,MAAM,GAAG;AAC5B,QAAI,OAAO,GAAG;AACZ,gBAAU;AAAA,IACZ,OAAO;AACL,YAAM,MAAM,IAAI;AAChB,iBAAW,MAAM,KAAK,MAAM,UAAU,GAAG;AACzC,UAAI,WAAW,KAAK;AAClB,kBAAU,MAAM,UAAU;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AACA,SAAO,aAAa,MAAM,GAAG,SAAS,MAAM,CAAC;AAC/C;AACA,SAAS,aAAa,UAAU,KAAK;AACnC,MAAI,UAAU;AACd,MAAI,UAAU,GAAG;AACf,QAAI,OAAO,GAAG;AACZ,gBAAU;AAAA,IACZ,OAAO;AACL,YAAM,KAAK,MAAM;AACjB,iBAAW,OAAO,KAAK,MAAM,CAAC,UAAU,EAAE,IAAI;AAAA,IAChD;AAAA,EACF,WAAW,UAAU,MAAM,GAAG;AAC5B,QAAI,OAAO,GAAG;AACZ,gBAAU;AAAA,IACZ,OAAO;AACL,YAAM,KAAK,MAAM;AACjB,iBAAW,MAAM,KAAK,MAAM,UAAU,EAAE;AAAA,IAC1C;AAAA,EACF;AACA,SAAO,aAAa,MAAM,GAAG,SAAS,MAAM,CAAC;AAC/C;AACA,SAAS,iBAAiB,UAAU,KAAK;AACvC,SAAO;AACT;AACA,SAAS,gBAAgB,UAAU,KAAK;AACtC,SAAO,aAAa,MAAM,GAAG,UAAU,MAAM,CAAC;AAChD;AACA,SAAS,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,GAAG,GAAG,SAAS,WAAW;AACjI,QAAM,MAAM,QAAQ,cAAc,IAAI,YAAY,IAAI,YAAY;AAClE,MAAI,KAAK,KAAK,IAAI,eAAe,KAAK,KAAK,IAAI,YAAY;AACzD,WAAO,UAAU;AAAA,EACnB,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,SAAS,qBAAqB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,GAAG,GAAG,SAAS,WAAW;AACpI,QAAM,KAAK,KAAK,MAAM,CAAC;AACvB,QAAM,KAAK,KAAK,MAAM,CAAC;AACvB,SAAO,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,IAAI,IAAI,SAAS,SAAS;AACnI;AACA,SAAS,sBAAsB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,GAAG,GAAG,SAAS,WAAW;AACrI,QAAM,SAAS,KAAK,MAAM,CAAC;AAC3B,QAAM,SAAS,KAAK,MAAM,CAAC;AAC3B,QAAM,QAAQ,SAAS;AACvB,QAAM,QAAQ,SAAS;AACvB,QAAM,eAAe,QAAQ,KAAK,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,QAAQ,QAAQ,SAAS,SAAS,KAAK,IAAI,UAAU,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,QAAQ,OAAO,SAAS,SAAS;AACxT,QAAM,cAAc,QAAQ,KAAK,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,OAAO,QAAQ,SAAS,SAAS,KAAK,IAAI,UAAU,kBAAkB,WAAW,aAAa,YAAY,aAAa,WAAW,WAAW,OAAO,OAAO,OAAO,SAAS,SAAS;AACrT,UAAQ,QAAQ,KAAK,eAAe,IAAI,UAAU;AACpD;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,EAAE,EAAE,IAAI;AACd,mBAAiB,GAAG,QAAQ;AAC5B,QAAM,SAAS,SAAS,KAAK,IAAI,EAAE,MAAM,EAAE;AAC3C,QAAM,EAAE,cAAc,aAAa,QAAQ,IAAI,WAAW,QAAQ,MAAM,EAAE,OAAO,EAAE,KAAK;AACxF,SAAO;AAAA,IACL,SAAS,eAAe,aAAa,EAAE,OAAO,YAAY;AAAA,IAC1D,SAAS,eAAe,CAAC,QAAQ,MAAM,GAAG,SAAS,OAAO;AAAA,EAC5D;AACF;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,IAAI;AAClB,MAAI,EAAE,KAAK,IAAI;AACf,MAAI,OAAO,GAAG;AACZ,YAAQ,MAAM,MAAM;AAAA,EACtB;AACA,QAAM,YAAY,MAAM,MAAM;AAC9B,QAAM,MAAM,MAAM,MAAM;AACxB,QAAM,WAAW,IAAI,MAAM,YAAY,CAAC;AACxC,MAAI,WAAW;AACf,WAAS,IAAI,GAAG,IAAI,WAAW,KAAK;AAClC,QAAI,MAAM,MAAM;AACd,eAAS,cAAc,MAAM,MAAM;AAAA,IACrC;AAAA,EACF;AACA,QAAM,QAAQ,IAAI,MAAM,SAAS,EAAE,KAAK,CAAC;AACzC,QAAM,OAAO,MAAM,MAAM,MAAM;AAC/B,OAAK,QAAQ;AACb,QAAM,MAAM,IAAI,MAAM,GAAG;AACzB,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,UAAM,QAAQ;AACd,UAAM,UAAU,OAAO,EAAE,QAAQ,EAAE,GAAG,MAAM,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,KAAK,EAAE,CAAC;AAC1F,QAAI,KAAK,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC3F,aAAS,8BAA8B,OAAO;AAAA,EAChD;AACA,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,WAAW,IAAI;AAC1B,QAAM,EAAE,YAAY,IAAI;AACxB,mBAAiB,GAAG,oBAAoB;AACxC,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,iBAAiB,WAAW,MAAM;AACxC,QAAM,MAAM,CAAC;AACb,QAAM,gBAAgB,CAAC;AACvB,QAAM,WAAW,QAAQ;AACzB,MAAI,cAAc;AAClB,WAAS,IAAI,GAAG,IAAI,UAAU,EAAE,GAAG;AACjC,UAAM,WAAW,YAAY,EAAE,QAAQ,EAAE,OAAO,YAAY,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,IAAI,EAAE,EAAE,CAAC;AACzG,kBAAc;AACd,kBAAc,KAAK,QAAQ;AAAA,EAC7B;AACA,WAAS,IAAI,GAAG,IAAI,aAAa,EAAE,GAAG;AACpC,UAAM,cAAc,aAAa,kBAAkB,GAAG,OAAO;AAC7D,UAAM,YAAY,SAAS,eAAe,CAAC,GAAG,SAAS,WAAW;AAClE,UAAM,OAAO,OAAO,EAAE,QAAQ,EAAE,GAAG,WAAW,GAAG,YAAY,GAAG,SAAS,SAAS,CAAC;AACnF,UAAM,aAAa,MAAM,EAAE,QAAQ,EAAE,GAAG,KAAK,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AAChG,UAAM,OAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,EAAE,GAAG,SAAS,SAAS,CAAC;AAC7E,UAAM,gBAAgB,KAAK,EAAE,QAAQ,EAAE,GAAG,KAAK,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,GAAG,UAAU,MAAM,EAAE,CAAC;AAC1G,QAAI,KAAK,aAAa;AACtB,kBAAc,KAAK,SAAS;AAC5B,kBAAc,KAAK,IAAI;AACvB,kBAAc,KAAK,UAAU;AAC7B,kBAAc,KAAK,IAAI;AACvB,kBAAc,KAAK,aAAa;AAAA,EAClC;AACA,QAAM,SAAS,KAAK,EAAE,QAAQ,KAAK,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,EAAE,CAAC;AAC1E,gBAAc,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AACtE,SAAO;AACT;AACA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB;AAAA,EAClB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,WAAW,gBAAgB,eAAe;AACxC,iBAAe,YAAY;AAC7B;AAGA,IAAI,qBAAqB,CAAC;AAC1BJ,UAAS,oBAAoB;AAAA,EAC3B,kBAAkB,MAAM;AAAA,EACxB,yBAAyB,MAAM;AAAA,EAC/B,+BAA+B,MAAM;AAAA,EACrC,oCAAoC,MAAM;AAAA,EAC1C,iBAAiB,MAAM;AAAA,EACvB,oCAAoC,MAAM;AAAA,EAC1C,cAAc,MAAM;AAAA,EACpB,kBAAkB,MAAM;AAAA,EACxB,sBAAsB,MAAM;AAAA,EAC5B,mBAAmB,MAAM;AAAA,EACzB,eAAe,MAAM;AAAA,EACrB,yBAAyB,MAAM;AAAA,EAC/B,0BAA0B,MAAM;AAAA,EAChC,eAAe,MAAM;AAAA,EACrB,oBAAoB,MAAM;AAAA,EAC1B,aAAa,MAAM;AAAA,EACnB,qBAAqB,MAAM;AAAA,EAC3B,4BAA4B,MAAM;AAAA,EAClC,wBAAwB,MAAM;AAAA,EAC9B,gBAAgB,MAAM;AAAA,EACtB,2BAA2B,MAAM;AAAA,EACjC,kCAAkC,MAAM;AAAA,EACxC,aAAa,MAAM;AAAA,EACnB,cAAc,MAAM;AAAA,EACpB,iCAAiC,MAAM;AAAA,EACvC,mCAAmC,MAAM;AAAA,EACzC,sBAAsB,MAAM;AAAA,EAC5B,wBAAwB,MAAM;AAAA,EAC9B,cAAc,MAAM;AAAA,EACpB,oCAAoC,MAAM;AAAA,EAC1C,+BAA+B,MAAM;AAAA,EACrC,eAAe,MAAM;AAAA,EACrB,qBAAqB,MAAM;AAAA,EAC3B,uBAAuB,MAAM;AAAA,EAC7B,aAAa,MAAM;AAAA,EACnB,2BAA2B,MAAM;AAAA,EACjC,qBAAqB,MAAM;AAAA,EAC3B,0BAA0B,MAAM;AAAA,EAChC,mCAAmC,MAAM;AAAA,EACzC,mBAAmB,MAAM;AAAA,EACzB,qBAAqB,MAAM;AAAA,EAC3B,iBAAiB,MAAM;AAAA,EACvB,qBAAqB,MAAM;AAC7B,CAAC;AAGD,IAAI,WAAW,CAAC;AAChB,IAAI,mBAAmB;AAAA,EACrB,OAAO;AAAA,EACP,WAAW;AAAA,EACX,oBAAoB;AAAA,EACpB,uBAAuB;AAAA,EACvB,OAAO;AAAA,EACP,SAAS;AAAA,EACT,8BAA8B;AAChC;AACA,SAAS,gBAAgB,cAAc,IAAI;AACzC,WAAS,gBAAgB;AAC3B;AACA,SAAS,gBAAgB,cAAc,cAAc;AACnD,MAAI,EAAE,gBAAgB,aAAa,gBAAgB,MAAM;AACvD,UAAM,SAAS,yBAAyB,cAAc,YAAY;AAClE,QAAI,WAAW,MAAM;AACnB,eAAS,gBAAgB;AAAA,IAC3B,OAAO;AACL,cAAQ,IAAI,2CAA2C,YAAY;AACnE,aAAO;AAAA,IACT;AAAA,EACF;AACA,QAAM,KAAK,SAAS;AACpB,MAAI,MAAM,QAAQ,GAAG,cAAc,GAAG;AACpC,WAAO,SAAS;AAChB,WAAO,gBAAgB,YAAY;AAAA,EACrC;AACA,KAAG,QAAQ,GAAG,UAAU;AACxB,KAAG,QAAQ,GAAG,YAAY;AAC1B,KAAG,QAAQ,GAAG,KAAK;AACnB,KAAG,QAAQ,GAAG,MAAM;AACpB,KAAG,QAAQ,GAAG,mBAAmB;AACjC,KAAG,QAAQ,GAAG,eAAe;AAC7B,KAAG,OAAO,GAAG,YAAY;AACzB,KAAG,OAAO,GAAG,SAAS;AACtB,KAAG,SAAS,GAAG,IAAI;AACnB,SAAO,SAAS;AAClB;AACA,SAAS,aAAa,cAAc;AAClC,MAAI,OAAO,oBAAoB,eAAe,iBAAiB,GAAG;AAChE,WAAO,IAAI,gBAAgB,KAAK,GAAG;AAAA,EACrC,WAAW,OAAO,aAAa,aAAa;AAC1C,WAAO,SAAS,cAAc,QAAQ;AAAA,EACxC,OAAO;AACL,UAAM,IAAI,MAAM,wCAAwC;AAAA,EAC1D;AACF;AACA,SAAS,yBAAyB,cAAc,cAAc;AAC5D,MAAI,iBAAiB,KAAK,iBAAiB,GAAG;AAC5C,UAAM,IAAI,MAAM,wDAAwD;AAAA,EAC1E;AACA,QAAM,SAAS,gBAAgB,OAAO,aAAa,YAAY,IAAI;AACnE,SAAO,iBAAiB,oBAAoB,CAAC,OAAO;AAClD,OAAG,eAAe;AAClB,WAAO,SAAS;AAAA,EAClB,GAAG,KAAK;AACR,MAAI,IAAI,EAAE,QAAQ,wBAAwB,GAAG;AAC3C,qBAAiB,+BAA+B;AAAA,EAClD;AACA,MAAI,iBAAiB,GAAG;AACtB,WAAO,OAAO,WAAW,SAAS,gBAAgB,KAAK,OAAO,WAAW,sBAAsB,gBAAgB;AAAA,EACjH;AACA,SAAO,OAAO,WAAW,UAAU,gBAAgB;AACrD;AAGA,IAAI;AAAA,CACH,SAAS,gBAAgB;AACxB,iBAAe,eAAe,WAAW,KAAK;AAC9C,iBAAe,eAAe,kBAAkB,KAAK;AACvD,GAAG,kBAAkB,gBAAgB,CAAC,EAAE;AACxC,IAAI;AAAA,CACH,SAAS,eAAe;AACvB,gBAAc,cAAc,YAAY,KAAK;AAC7C,gBAAc,cAAc,YAAY,KAAK;AAC7C,gBAAc,cAAc,YAAY,KAAK;AAC7C,gBAAc,cAAc,cAAc,KAAK;AACjD,GAAG,iBAAiB,eAAe,CAAC,EAAE;AACtC,IAAI;AAAA,CACH,SAAS,sBAAsB;AAC9B,uBAAqB,qBAAqB,sBAAsB,KAAK;AACrE,uBAAqB,qBAAqB,sBAAsB,KAAK;AACrE,uBAAqB,qBAAqB,8BAA8B,KAAK;AAC7E,uBAAqB,qBAAqB,wBAAwB,KAAK;AACvE,uBAAqB,qBAAqB,wBAAwB,KAAK;AACzE,GAAG,wBAAwB,sBAAsB,CAAC,EAAE;AACpD,SAAS,yCAAyC,MAAM,SAAS;AAC/D,SAAO,CAAC,SAAS,IAAI;AACvB;AACA,SAAS,mCAAmC,YAAY,oBAAoB;AAC1E,SAAO,aAAa;AACtB;AACA,SAAS,iBAAiB,OAAO;AAC/B,QAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,QAAM,eAAe,KAAK,KAAK,OAAO,CAAC;AACvC,SAAO,aAAa,oBAAoB,YAAY;AACtD;AACA,SAAS,uCAAuC,MAAM,SAAS;AAC7D,SAAO;AAAA,IACL,KAAK,IAAI,GAAG,KAAK,KAAK,UAAU,CAAC,CAAC;AAAA,IAClC,KAAK,IAAI,GAAG,KAAK,KAAK,OAAO,CAAC,CAAC;AAAA,EACjC;AACF;AACA,SAAS,sCAAsC,MAAM,SAAS;AAC5D,QAAM,CAAC,GAAG,CAAC,IAAI,uCAAuC,MAAM,OAAO;AACnE,SAAO,IAAI,IAAI;AACjB;AACA,SAAS,iBAAiB,IAAI,2BAA2B;AACvD,QAAM,QAAQ;AACd,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,0BAAsB,MAAM;AAC5B,8BAA0B,MAAM;AAChC,oCAAgC,MAAM;AACtC,gCAA4B,MAAM;AAClC,yBAAqB,MAAM;AAC3B,gCAA4B;AAC5B,yBAAqB;AACrB,2BAAuB,MAAM;AAC7B,uBAAmB,MAAM;AACzB,4BAAwB,MAAM;AAAA,EAChC,OAAO;AACL,0BAAsB,GAAG;AACzB,8BAA0B,GAAG;AAC7B,oCAAgC,GAAG;AACnC,gCAA4B,MAAM;AAClC,yBAAqB,GAAG;AACxB,gCAA4B;AAC5B,yBAAqB;AACrB,2BAAuB,6BAA6B,OAAO,0BAA0B,iBAAiB;AACtG,uBAAmB,GAAG;AACtB,4BAAwB,GAAG;AAAA,EAC7B;AACA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAGA,SAAS,aAAa,IAAI,OAAO;AAC/B,QAAM,cAAc,MAAM;AAC1B,MAAI,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC1B,oBAAgB,EAAE;AAAA,EACpB;AACA,SAAO;AACT;AACA,SAAS,gBAAgB,IAAI;AAC3B,QAAM,QAAQ,GAAG,SAAS;AAC1B,MAAI,UAAU,GAAG,UAAU;AACzB,UAAM,IAAI,MAAM,kBAAkB,qBAAqB,IAAI,KAAK,CAAC;AAAA,EACnE;AACF;AACA,IAAI,cAAc;AAClB,IAAI,cAAc;AAClB,SAAS,iBAAiB,KAAK;AAC7B,MAAI,IAAI,EAAE,QAAQ,8BAA8B,KAAK,QAAQ,KAAK,cAAc,KAAK,IAAI,GAAG,KAAK,KAAK,IAAI,GAAG,IAAI,aAAa;AAC5H,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,qBAAqB,IAAI,QAAQ;AACxC,UAAQ,QAAQ;AAAA,IACd,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT;AACE,aAAO,sBAAsB;AAAA,EACjC;AACF;AACA,SAAS,oBAAoB,IAAI,eAAe;AAC9C,SAAO,YAAY,IAAI,MAAM,GAAG,aAAa,aAAa,GAAG,gBAAgB,gBAAgB,kCAAkC;AACjI;AACA,SAAS,mBAAmB,IAAI,oBAAoB;AAClD,QAAM,eAAe,YAAY,IAAI,MAAM,GAAG,aAAa,GAAG,aAAa,GAAG,sCAAsC;AACpH,eAAa,IAAI,MAAM,GAAG,aAAa,cAAc,kBAAkB,CAAC;AACxE,eAAa,IAAI,MAAM,GAAG,cAAc,YAAY,CAAC;AACrD,MAAI,GAAG,mBAAmB,cAAc,GAAG,cAAc,MAAM,OAAO;AACpE,YAAQ,IAAI,GAAG,iBAAiB,YAAY,CAAC;AAC7C,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AACA,SAAO;AACT;AACA,SAAS,qBAAqB,IAAI,sBAAsB;AACtD,QAAM,iBAAiB,YAAY,IAAI,MAAM,GAAG,aAAa,GAAG,eAAe,GAAG,wCAAwC;AAC1H,eAAa,IAAI,MAAM,GAAG,aAAa,gBAAgB,oBAAoB,CAAC;AAC5E,eAAa,IAAI,MAAM,GAAG,cAAc,cAAc,CAAC;AACvD,MAAI,IAAI,EAAE,IAAI,qBAAqB,GAAG;AACpC,WAAO;AAAA,EACT;AACA,MAAI,GAAG,mBAAmB,gBAAgB,GAAG,cAAc,MAAM,OAAO;AACtE,8BAA0B,sBAAsB,GAAG,iBAAiB,cAAc,CAAC;AACnF,UAAM,IAAI,MAAM,oCAAoC;AAAA,EACtD;AACA,SAAO;AACT;AACA,IAAI,kBAAkB;AACtB,SAAS,0BAA0B,cAAc,eAAe;AAC9D,QAAM,wBAAwB,gBAAgB,KAAK,aAAa;AAChE,MAAI,yBAAyB,MAAM;AACjC,YAAQ,IAAI,wCAAwC,eAAe;AACnE,YAAQ,IAAI,YAAY;AACxB;AAAA,EACF;AACA,QAAM,aAAa,CAAC,sBAAsB;AAC1C,QAAM,cAAc,aAAa,MAAM,IAAI;AAC3C,QAAM,OAAO,YAAY,OAAO,SAAS,EAAE,SAAS;AACpD,QAAM,uBAAuB,YAAY,IAAI,CAAC,MAAM,gBAAgB,aAAa,UAAU,cAAc,GAAG,SAAS,GAAG,IAAI,IAAI,IAAI;AACpI,MAAI,gBAAgB;AACpB,WAAS,IAAI,GAAG,IAAI,qBAAqB,QAAQ,KAAK;AACpD,oBAAgB,KAAK,IAAI,qBAAqB,GAAG,QAAQ,aAAa;AAAA,EACxE;AACA,QAAM,mBAAmB,qBAAqB,MAAM,GAAG,aAAa,CAAC;AACrE,QAAM,YAAY,qBAAqB,MAAM,aAAa,GAAG,UAAU;AACvE,QAAM,kBAAkB,qBAAqB,MAAM,UAAU;AAC7D,UAAQ,IAAI,iBAAiB,KAAK,IAAI,CAAC;AACvC,UAAQ,IAAI,cAAc,MAAM,IAAI,EAAE,EAAE;AACxC,UAAQ,IAAI,MAAM,aAAa,SAAS,UAAU,IAAI,aAAa,KAAK,+DAA+D;AACvI,UAAQ,IAAI,gBAAgB,KAAK,IAAI,CAAC;AACxC;AACA,SAAS,cAAc,IAAI;AACzB,SAAO,YAAY,IAAI,MAAM,GAAG,cAAc,GAAG,gCAAgC;AACnF;AACA,SAAS,YAAY,IAAI,SAAS;AAChC,eAAa,IAAI,MAAM,GAAG,YAAY,OAAO,CAAC;AAC9C,MAAI,IAAI,EAAE,IAAI,qBAAqB,GAAG;AACpC;AAAA,EACF;AACA,MAAI,GAAG,oBAAoB,SAAS,GAAG,WAAW,MAAM,OAAO;AAC7D,YAAQ,IAAI,GAAG,kBAAkB,OAAO,CAAC;AACzC,UAAM,IAAI,MAAM,6CAA6C;AAAA,EAC/D;AACF;AACA,SAAS,gBAAgB,IAAI,SAAS;AACpC,eAAa,IAAI,MAAM,GAAG,gBAAgB,OAAO,CAAC;AAClD,MAAI,GAAG,oBAAoB,SAAS,GAAG,eAAe,MAAM,OAAO;AACjE,YAAQ,IAAI,GAAG,kBAAkB,OAAO,CAAC;AACzC,UAAM,IAAI,MAAM,mCAAmC;AAAA,EACrD;AACF;AACA,SAAS,yBAAyB,IAAI,MAAM;AAC1C,QAAM,UAAU,YAAY,IAAI,MAAM,GAAG,aAAa,GAAG,8BAA8B;AACvF,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,cAAc,OAAO,CAAC;AAC9D,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,cAAc,MAAM,GAAG,WAAW,CAAC;AAC3E,SAAO;AACT;AACA,SAAS,wBAAwB,IAAI,MAAM;AACzC,QAAM,UAAU,YAAY,IAAI,MAAM,GAAG,aAAa,GAAG,8BAA8B;AACvF,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,sBAAsB,OAAO,CAAC;AACtE,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,sBAAsB,MAAM,GAAG,WAAW,CAAC;AACnF,SAAO;AACT;AACA,SAAS,iBAAiB;AACxB,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,cAAc,IAAI;AACzB,SAAO,YAAY,IAAI,MAAM,GAAG,cAAc,GAAG,gCAAgC;AACnF;AACA,SAAS,oBAAoB,OAAO,QAAQ;AAC1C,QAAM,iBAAiB,IAAI,EAAE,UAAU,wBAAwB;AAC/D,MAAI,SAAS,KAAK,UAAU,GAAG;AAC7B,UAAM,YAAY,IAAI,SAAS;AAC/B,UAAM,IAAI,MAAM,4BAA4B,YAAY,cAAc;AAAA,EACxE;AACA,MAAI,QAAQ,kBAAkB,SAAS,gBAAgB;AACrD,UAAM,YAAY,IAAI,SAAS;AAC/B,UAAM,OAAO,IAAI,kBAAkB;AACnC,UAAM,IAAI,MAAM,4BAA4B,YAAY,uDAAuD,OAAO,GAAG;AAAA,EAC3H;AACF;AACA,SAAS,kBAAkB,IAAI;AAC7B,SAAO,YAAY,IAAI,MAAM,GAAG,kBAAkB,GAAG,oCAAoC;AAC3F;AACA,SAAS,mCAAmC,IAAI,SAAS,WAAW,SAAS,qBAAqB,mBAAmB,mBAAmB;AACtI,QAAM,MAAM,GAAG,kBAAkB,SAAS,SAAS;AACnD,MAAI,QAAQ,IAAI;AACd,WAAO;AAAA,EACT;AACA,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,cAAc,OAAO,CAAC;AAC9D,eAAa,IAAI,MAAM,GAAG,oBAAoB,KAAK,qBAAqB,GAAG,OAAO,OAAO,mBAAmB,iBAAiB,CAAC;AAC9H,eAAa,IAAI,MAAM,GAAG,wBAAwB,GAAG,CAAC;AACtD,SAAO;AACT;AACA,SAAS,gBAAgB,IAAI,SAAS,aAAa;AACjD,sBAAoB,IAAI,WAAW;AACnC,eAAa,IAAI,MAAM,GAAG,cAAc,GAAG,WAAW,WAAW,CAAC;AAClE,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,OAAO,CAAC;AAC/D;AACA,SAAS,kBAAkB,IAAI,aAAa;AAC1C,sBAAoB,IAAI,WAAW;AACnC,eAAa,IAAI,MAAM,GAAG,cAAc,GAAG,WAAW,WAAW,CAAC;AAClE,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,IAAI,CAAC;AAC5D;AACA,SAAS,iCAAiC,IAAI,SAAS,aAAa;AAClE,SAAO,YAAY,IAAI,MAAM,GAAG,mBAAmB,SAAS,WAAW,GAAG,cAAc,cAAc,2BAA2B;AACnI;AACA,SAAS,0BAA0B,IAAI,SAAS,aAAa;AAC3D,SAAO,GAAG,mBAAmB,SAAS,WAAW;AACnD;AACA,SAAS,mCAAmC,IAAI,SAAS,wBAAwB,aAAa;AAC5F,eAAa,IAAI,MAAM,gBAAgB,IAAI,SAAS,WAAW,CAAC;AAChE,eAAa,IAAI,MAAM,GAAG,UAAU,wBAAwB,WAAW,CAAC;AAC1E;AACA,SAAS,wBAAwB,IAAI;AACnC,eAAa,IAAI,MAAM,GAAG,gBAAgB,GAAG,aAAa,IAAI,CAAC;AAC/D,eAAa,IAAI,MAAM,GAAG,SAAS,GAAG,GAAG,GAAG,OAAO,OAAO,GAAG,OAAO,MAAM,CAAC;AAC3E,eAAa,IAAI,MAAM,GAAG,QAAQ,GAAG,GAAG,GAAG,OAAO,OAAO,GAAG,OAAO,MAAM,CAAC;AAC5E;AACA,SAAS,8BAA8B,IAAI,SAAS,aAAa;AAC/D,eAAa,IAAI,MAAM,GAAG,gBAAgB,GAAG,aAAa,WAAW,CAAC;AACtE,eAAa,IAAI,MAAM,GAAG,qBAAqB,GAAG,aAAa,GAAG,mBAAmB,GAAG,YAAY,SAAS,CAAC,CAAC;AACjH;AACA,SAAS,kCAAkC,IAAI,aAAa;AAC1D,eAAa,IAAI,MAAM,GAAG,gBAAgB,GAAG,aAAa,WAAW,CAAC;AACtE,eAAa,IAAI,MAAM,GAAG,qBAAqB,GAAG,aAAa,GAAG,mBAAmB,GAAG,YAAY,MAAM,CAAC,CAAC;AAC9G;AACA,SAAS,oBAAoB,IAAI;AAC/B,QAAM,SAAS,GAAG,uBAAuB,GAAG,WAAW;AACvD,MAAI,WAAW,GAAG,sBAAsB;AACtC,UAAM,IAAI,MAAM,gCAAgC,2BAA2B,IAAI,MAAM,CAAC;AAAA,EACxF;AACF;AACA,SAAS,2BAA2B,IAAI,QAAQ;AAC9C,UAAQ,QAAQ;AAAA,IACd,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT,KAAK,GAAG;AACN,aAAO;AAAA,IACT;AACE,aAAO,iBAAiB;AAAA,EAC5B;AACF;AACA,SAAS,YAAY,IAAI,eAAe,gBAAgB;AACtD,QAAM,UAAU,aAAa,IAAI,MAAM,cAAc,CAAC;AACtD,MAAI,WAAW,MAAM;AACnB,UAAM,IAAI,MAAM,cAAc;AAAA,EAChC;AACA,SAAO;AACT;AACA,SAAS,oBAAoB,IAAI,aAAa;AAC5C,QAAM,iBAAiB,GAAG,mCAAmC;AAC7D,QAAM,gBAAgB,cAAc,GAAG;AACvC,MAAI,gBAAgB,GAAG,YAAY,gBAAgB,gBAAgB;AACjE,UAAM,mBAAmB,2BAA2B;AACpD,UAAM,IAAI,MAAM,0BAA0B,mBAAmB;AAAA,EAC/D;AACF;AACA,SAAS,YAAY,OAAO,aAAa,GAAG;AAC1C,SAAO,aAAa,cAAc,MAAM,MAAM,GAAG,MAAM,SAAS,UAAU,CAAC;AAC7E;AACA,SAAS,YAAY,OAAO;AAC1B,MAAI,MAAM,WAAW,GAAG;AACtB,UAAM,MAAM,sDAAsD;AAAA,EACpE;AACA,SAAO;AAAA,IACL,MAAM,SAAS,IAAI,MAAM,MAAM,SAAS,KAAK;AAAA,IAC7C,MAAM,MAAM,SAAS;AAAA,EACvB;AACF;AACA,SAAS,aAAa,OAAO;AAC3B,MAAI,YAAY,CAAC,GAAG,GAAG,CAAC;AACxB,QAAM,WAAW,MAAM,WAAW,KAAK,MAAM,WAAW,KAAK,MAAM,OAAO;AAC1E,MAAI,CAAC,UAAU;AACb,gBAAY,CAAC,YAAY,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC;AAAA,EACxD;AACA,SAAO;AACT;AACA,SAAS,gCAAgC,UAAU,WAAW,OAAO;AACnE,MAAI,aAAa,IAAI,EAAE,UAAU,wBAAwB;AACzD,MAAI,sBAAsB,IAAI,EAAE,UAAU,mCAAmC;AAC7E,MAAI,wBAAwB,YAAY,IAAI,EAAE,QAAQ,0CAA0C,GAAG;AACjG,0BAAsB,aAAa;AAAA,EACrC;AACA,MAAI,UAAU;AACZ,iBAAa,aAAa;AAC1B,0BAAsB,sBAAsB;AAC5C,eAAW,SAAS,IAAI,CAAC,GAAG,MAAM,KAAK,SAAS,SAAS,IAAI,aAAa,kBAAkB,SAAS,EAAE,IAAI,SAAS,EAAE;AACtH,QAAI,SAAS,WAAW,GAAG;AACzB,iBAAW,CAAC,GAAG,SAAS,EAAE;AAAA,IAC5B;AAAA,EACF;AACA,MAAI,SAAS,WAAW,GAAG;AACzB,UAAM,gBAAgB,aAAa,aAAa,QAAQ;AACxD,eAAW,cAAc;AAAA,EAC3B;AACA,MAAI,OAAO,aAAa,cAAc,QAAQ;AAC9C,MAAI,eAAe;AACnB,MAAI,SAAS,UAAU,KAAK,QAAQ,YAAY;AAC9C,mBAAe,CAAC,GAAG,IAAI;AAAA,EACzB,WAAW,SAAS,WAAW,KAAK,SAAS,MAAM,cAAc,SAAS,MAAM,YAAY;AAC1F,mBAAe;AAAA,EACjB,WAAW,SAAS,WAAW,KAAK,SAAS,KAAK,SAAS,MAAM,cAAc,SAAS,MAAM,YAAY;AACxG,mBAAe,CAAC,SAAS,KAAK,SAAS,IAAI,SAAS,EAAE;AAAA,EACxD,WAAW,SAAS,WAAW,KAAK,SAAS,MAAM,cAAc,SAAS,KAAK,SAAS,MAAM,YAAY;AACxG,mBAAe,CAAC,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE;AAAA,EACxD,WAAW,SAAS,WAAW,KAAK,SAAS,KAAK,SAAS,KAAK,SAAS,MAAM,cAAc,SAAS,MAAM,YAAY;AACtH,mBAAe,CAAC,SAAS,KAAK,SAAS,KAAK,SAAS,IAAI,SAAS,EAAE;AAAA,EACtE,WAAW,SAAS,WAAW,KAAK,SAAS,MAAM,cAAc,SAAS,KAAK,SAAS,KAAK,SAAS,MAAM,YAAY;AACtH,mBAAe,CAAC,SAAS,IAAI,SAAS,KAAK,SAAS,KAAK,SAAS,EAAE;AAAA,EACtE;AACA,QAAM,kBAAkB,gBAAgB,QAAQ,KAAK,IAAI,GAAG,YAAY,IAAI,uBAAuB,KAAK,IAAI,GAAG,YAAY,MAAM,WAAW,IAAI,MAAM,KAAK,IAAI,GAAG,YAAY,IAAI;AAClL,MAAI,gBAAgB,QAAQ,iBAAiB;AAC3C,QAAI,UAAU;AACZ,YAAM,WAAW,YAAY,QAAQ;AACrC,UAAI,OAAO,GAAG,OAAO;AACrB,UAAI,SAAS,QAAQ;AACnB,SAAC,MAAM,IAAI,IAAI,YAAY,QAAQ;AAAA,MACrC;AACA,aAAO,YAAY,OAAO,MAAM,OAAO;AACvC,qBAAe,aAAa,oBAAoB,IAAI,EAAE,IAAI,CAAC,MAAM,IAAI,CAAC;AAAA,IACxE,OAAO;AACL,qBAAe,aAAa,oBAAoB,IAAI;AAAA,IACtD;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,OAAO,GAAG;AACjB,SAAO,IAAI,MAAM;AACnB;AACA,SAAS,cAAc,QAAQ,QAAQ;AACrC,WAAS,OAAO,MAAM,EAAE;AACxB,WAAS,OAAO,MAAM,EAAE;AACxB,MAAI,aAAa,YAAY,QAAQ,MAAM,GAAG;AAC5C,WAAO;AAAA,EACT;AACA,MAAI,CAAC,OAAO,UAAU,CAAC,OAAO,QAAQ;AACpC,WAAO;AAAA,EACT;AACA,MAAI,OAAO,OAAO,KAAK,OAAO,OAAO,KAAK,OAAO,OAAO,KAAK,OAAO,OAAO,GAAG;AAC5E,WAAO;AAAA,EACT;AACA,MAAI,OAAO,WAAW,OAAO,QAAQ;AACnC,UAAM,aAAa,OAAO,MAAM,EAAE,EAAE;AACpC,UAAM,aAAa,OAAO,MAAM,EAAE,EAAE;AACpC,QAAI,eAAe,YAAY;AAC7B,aAAO;AAAA,IACT;AACA,QAAI,OAAO,UAAU,KAAK,OAAO,UAAU,MAAM,OAAO,OAAO,KAAK,OAAO,OAAO,IAAI;AACpF,aAAO;AAAA,IACT;AAAA,EACF;AACA,SAAO,OAAO,OAAO,OAAO,MAAM,OAAO,OAAO,EAAE,KAAK,OAAO,OAAO,EAAE;AACzE;AACA,IAAI;AACJ,IAAI;AACJ,SAAS,uBAAuB,cAAc;AAC5C,MAAI,oBAAoB,MAAM;AAC5B,UAAM,KAAK,gBAAgB,YAAY;AACvC,uBAAmB,GAAG,aAAa,GAAG,gBAAgB;AAAA,EACxD;AACA,SAAO;AACT;AACA,SAAS,sBAAsB;AAC7B,qBAAmB;AACrB;AACA,SAAS,2BAA2B;AAClC,2BAAyB;AAC3B;AACA,SAAS,uBAAuB,cAAc;AAC5C,MAAI,0BAA0B,MAAM;AAClC,UAAM,KAAK,gBAAgB,YAAY;AACvC,6BAAyB,GAAG,aAAa,GAAG,uBAAuB;AAAA,EACrE;AACA,SAAO,KAAK,IAAI,IAAI,sBAAsB;AAC5C;AACA,SAAS,kCAAkC,cAAc;AACvD,MAAI,iBAAiB,GAAG;AACtB,WAAO;AAAA,EACT;AACA,MAAI;AACJ,QAAM,KAAK,gBAAgB,YAAY;AACvC,MAAI,aAAa,IAAI,iCAAiC,KAAK,iBAAiB,GAAG;AAC7E,wBAAoB;AAAA,EACtB,WAAW,aAAa,IAAI,0BAA0B,GAAG;AACvD,wBAAoB;AAAA,EACtB,OAAO;AACL,wBAAoB;AAAA,EACtB;AACA,SAAO;AACT;AACA,SAAS,aAAa,IAAI,eAAe;AACvC,QAAM,MAAM,GAAG,aAAa,aAAa;AACzC,SAAO,OAAO;AAChB;AACA,SAAS,sBAAsB,cAAc;AAC3C,MAAI;AACF,UAAM,KAAK,gBAAgB,YAAY;AACvC,QAAI,MAAM,MAAM;AACd,aAAO;AAAA,IACT;AAAA,EACF,SAAS,GAAP;AACA,YAAQ,IAAI,sCAAsC,CAAC;AACnD,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,mCAAmC,cAAc;AACxD,MAAI,iBAAiB,GAAG;AACtB,WAAO;AAAA,EACT;AACA,QAAM,KAAK,gBAAgB,YAAY;AACvC,MAAI,iBAAiB,GAAG;AACtB,QAAI,CAAC,aAAa,IAAI,mBAAmB,GAAG;AAC1C,aAAO;AAAA,IACT;AAAA,EACF,OAAO;AACL,QAAI,CAAC,aAAa,IAAI,wBAAwB,GAAG;AAC/C,aAAO;AAAA,IACT;AAAA,EACF;AACA,QAAM,wBAAwB,uCAAuC,EAAE;AACvE,SAAO;AACT;AACA,SAAS,8BAA8B,cAAc;AACnD,MAAI,iBAAiB,GAAG;AACtB,WAAO;AAAA,EACT;AACA,QAAM,KAAK,gBAAgB,YAAY;AACvC,MAAI,iBAAiB,GAAG;AACtB,QAAI,CAAC,aAAa,IAAI,mBAAmB,GAAG;AAC1C,aAAO;AAAA,IACT;AACA,QAAI,CAAC,aAAa,IAAI,0BAA0B,GAAG;AACjD,aAAO;AAAA,IACT;AAAA,EACF,OAAO;AACL,QAAI,aAAa,IAAI,wBAAwB,GAAG;AAC9C,aAAO,uCAAuC,EAAE;AAAA,IAClD;AACA,UAAM,0BAA0B;AAChC,QAAI,aAAa,IAAI,uBAAuB,GAAG;AAC7C,YAAM,4BAA4B,GAAG,aAAa,uBAAuB;AACzE,aAAO,2CAA2C,IAAI,yBAAyB;AAAA,IACjF;AACA,WAAO;AAAA,EACT;AACA,QAAM,wBAAwB,uCAAuC,EAAE;AACvE,SAAO;AACT;AACA,SAAS,uCAAuC,IAAI;AAClD,QAAM,YAAY,iBAAiB,EAAE;AACrC,QAAM,UAAU,GAAG,cAAc;AACjC,KAAG,YAAY,GAAG,YAAY,OAAO;AACrC,QAAM,QAAQ;AACd,QAAM,SAAS;AACf,KAAG,WAAW,GAAG,YAAY,GAAG,UAAU,qBAAqB,OAAO,QAAQ,GAAG,UAAU,oBAAoB,UAAU,kBAAkB,IAAI;AAC/I,QAAM,cAAc,GAAG,kBAAkB;AACzC,KAAG,gBAAgB,GAAG,aAAa,WAAW;AAC9C,KAAG,qBAAqB,GAAG,aAAa,GAAG,mBAAmB,GAAG,YAAY,SAAS,CAAC;AACvF,QAAM,wBAAwB,GAAG,uBAAuB,GAAG,WAAW,MAAM,GAAG;AAC/E,KAAG,YAAY,GAAG,YAAY,IAAI;AAClC,KAAG,gBAAgB,GAAG,aAAa,IAAI;AACvC,KAAG,cAAc,OAAO;AACxB,KAAG,kBAAkB,WAAW;AAChC,SAAO;AACT;AACA,SAAS,2CAA2C,IAAI,2BAA2B;AACjF,QAAM,YAAY,iBAAiB,IAAI,yBAAyB;AAChE,QAAM,UAAU,GAAG,cAAc;AACjC,KAAG,YAAY,GAAG,YAAY,OAAO;AACrC,QAAM,QAAQ;AACd,QAAM,SAAS;AACf,KAAG,WAAW,GAAG,YAAY,GAAG,UAAU,yBAAyB,OAAO,QAAQ,GAAG,UAAU,oBAAoB,UAAU,sBAAsB,IAAI;AACvJ,QAAM,cAAc,GAAG,kBAAkB;AACzC,KAAG,gBAAgB,GAAG,aAAa,WAAW;AAC9C,KAAG,qBAAqB,GAAG,aAAa,GAAG,mBAAmB,GAAG,YAAY,SAAS,CAAC;AACvF,QAAM,wBAAwB,GAAG,uBAAuB,GAAG,WAAW,MAAM,GAAG;AAC/E,KAAG,YAAY,GAAG,YAAY,IAAI;AAClC,KAAG,gBAAgB,GAAG,aAAa,IAAI;AACvC,KAAG,cAAc,OAAO;AACxB,KAAG,kBAAkB,WAAW;AAChC,SAAO;AACT;AACA,SAAS,oBAAoB,cAAc;AACzC,MAAI,iBAAiB,GAAG;AACtB,WAAO;AAAA,EACT;AACA,QAAM,KAAK,gBAAgB,YAAY;AACvC,QAAM,YAAY,GAAG,aAAa;AAClC,SAAO;AACT;AACA,SAAS,kBAAkB,SAAS,QAAQ;AAC1C,MAAI,CAAC,MAAM,QAAQ,OAAO,GAAG;AAC3B,cAAU,CAAC,OAAO;AAAA,EACpB;AACA,UAAQ,QAAQ,CAAC,MAAM;AACrB,QAAI,KAAK,MAAM;AACb,mBAAa,OAAO,EAAE,UAAU,aAAa,MAAM,GAAG,iEAAiE;AAAA,IACzH;AAAA,EACF,CAAC;AACH;AAGA,IAAI,OAAO,IAAI;AACf,KAAK,aAAa,aAAa,MAAM,KAAK,UAAU,eAAe,IAAI,CAAC;AACxE,KAAK,aAAa,iBAAiB,MAAM;AACvC,MAAI,sBAAsB,CAAC,GAAG;AAC5B,WAAO;AAAA,EACT,WAAW,sBAAsB,CAAC,GAAG;AACnC,WAAO;AAAA,EACT;AACA,SAAO;AACT,CAAC;AACD,KAAK,aAAa,kCAAkC,MAAM,KAAK;AAC/D,KAAK,aAAa,0BAA0B,MAAM,KAAK,IAAI,eAAe,MAAM,CAAC;AACjF,KAAK,aAAa,qBAAqB,MAAM,IAAI;AACjD,KAAK,aAAa,4BAA4B,MAAM,KAAK;AACzD,KAAK,aAAa,cAAc,MAAM,KAAK,QAAQ,WAAW,CAAC;AAC/D,KAAK,aAAa,4BAA4B,MAAM,KAAK,QAAQ,YAAY,CAAC;AAC9E,KAAK,aAAa,mBAAmB,MAAM,KAAK,QAAQ,YAAY,CAAC;AACrE,KAAK,aAAa,4BAA4B,MAAM,KAAK,QAAQ,YAAY,CAAC;AAC9E,KAAK,aAAa,gCAAgC,MAAM,KAAK,QAAQ,YAAY,CAAC;AAClF,KAAK,aAAa,+BAA+B,MAAM,KAAK,QAAQ,YAAY,CAAC;AACjF,KAAK,aAAa,+BAA+B,MAAM,KAAK,QAAQ,YAAY,CAAC;AACjF,KAAK,aAAa,+BAA+B,MAAM,KAAK,QAAQ,YAAY,CAAC;AACjF,KAAK,aAAa,qBAAqB,MAAM,KAAK,QAAQ,YAAY,CAAC;AACvE,KAAK,aAAa,uBAAuB,MAAM,KAAK,QAAQ,YAAY,CAAC;AACzE,KAAK,aAAa,qBAAqB,MAAM,KAAK,QAAQ,YAAY,CAAC;AACvE,KAAK,aAAa,0BAA0B,MAAM,uBAAuB,KAAK,UAAU,eAAe,CAAC,CAAC;AACzG,KAAK,aAAa,gCAAgC,MAAM,uBAAuB,KAAK,UAAU,eAAe,CAAC,CAAC;AAC/G,KAAK,aAAa,gDAAgD,MAAM;AACtE,QAAM,eAAe,KAAK,UAAU,eAAe;AACnD,MAAI,iBAAiB,GAAG;AACtB,WAAO;AAAA,EACT;AACA,SAAO,kCAAkC,YAAY;AACvD,CAAC;AACD,KAAK,aAAa,iDAAiD,MAAM,KAAK,UAAU,8CAA8C,IAAI,KAAK,CAAC,oBAAoB,SAAS,CAAC;AAC9K,KAAK,aAAa,gCAAgC,MAAM,mCAAmC,KAAK,UAAU,eAAe,CAAC,CAAC;AAC3H,KAAK,aAAa,gCAAgC,MAAM;AACtD,SAAO,KAAK,QAAQ,0BAA0B,IAAI,QAAQ,KAAK,QAAQ,8BAA8B;AACvG,CAAC;AACD,KAAK,aAAa,gCAAgC,MAAM,8BAA8B,KAAK,UAAU,eAAe,CAAC,CAAC;AACtH,KAAK,aAAa,2BAA2B,MAAM,oBAAoB,KAAK,UAAU,eAAe,CAAC,CAAC;AACvG,KAAK,aAAa,6BAA6B,MAAM;AACnD,QAAM,cAAc,KAAK,QAAQ,8BAA8B;AAC/D,SAAO,cAAc,IAAI;AAC3B,CAAC;AACD,KAAK,aAAa,kCAAkC,MAAM;AACxD,SAAO;AACT,GAAG,CAAC,eAAe;AACjB,MAAI,aAAa,KAAK,eAAe,IAAI;AACvC,UAAM,IAAI,MAAM,8FAA8F,aAAa;AAAA,EAC7H;AACF,CAAC;AACD,KAAK,aAAa,yBAAyB,MAAM;AAC/C,SAAO,oBAAoB,SAAS,IAAI,IAAI;AAC9C,GAAG,CAAC,eAAe;AACjB,MAAI,aAAa,KAAK,eAAe,IAAI;AACvC,UAAM,IAAI,MAAM,2FAA2F,aAAa;AAAA,EAC1H;AACF,CAAC;AACD,KAAK,aAAa,8BAA8B,MAAM,GAAG;AACzD,KAAK,aAAa,6BAA6B,MAAM,KAAK;AAC1D,KAAK,aAAa,4CAA4C,MAAM,GAAG;AACvE,KAAK,aAAa,gCAAgC,MAAM,GAAG;AAC3D,KAAK,aAAa,kBAAkB,MAAM,KAAK;AAC/C,KAAK,aAAa,0BAA0B,MAAM,KAAK,QAAQ,SAAS,CAAC;AACzE,KAAK,aAAa,qCAAqC,MAAM,QAAQ;AACrE,KAAK,aAAa,4CAA4C,MAAM,KAAK;AACzE,KAAK,aAAa,uBAAuB,MAAM,KAAK;AAGpD,SAAS,qBAAqB;AAC5B,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,aAAS;AACT,mBAAe;AACf,uBAAmB,IAAI,EAAE,QAAQ,qBAAqB,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAYtD;AACJ,uBAAmB;AACnB,kBAAc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUhB,OAAO;AACL,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,gBAAY;AACZ,aAAS;AACT,mBAAe;AACf,uBAAmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AASnB,uBAAmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUnB,kBAAc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAShB;AACA,SAAO;AAAA,IACL,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAGA,SAAS,mCAAmC,SAAS,OAAO,QAAQ,SAAS;AAC3E,QAAM,UAAU,aAAa,eAAe,KAAK;AACjD,SAAO,QAAQ,IAAI,CAAC,QAAQ,MAAM;AAChC,UAAM,QAAQ,OAAO,QAAQ,QAAQ,WAAW;AAChD,UAAM,QAAQ,MAAM,QAAQ,SAAS,IAAI,OAAO,QAAQ,IAAI,QAAQ,WAAW,QAAQ,QAAQ,WAAW,YAAY,QAAQ,QAAQ;AACtI,WAAO,GAAG,UAAU;AAAA,EACtB,CAAC,EAAE,KAAK,EAAE;AACZ;AACA,SAAS,kDAAkD,SAAS,OAAO,QAAQ,SAAS;AAC1F,QAAM,UAAU,aAAa,eAAe,KAAK;AACjD,SAAO,QAAQ,IAAI,CAAC,GAAG,MAAM;AAC3B,UAAM,QAAQ,OAAO,QAAQ,QAAQ,2BAA2B;AAChE,UAAM,QAAQ,MAAM,QAAQ,SAAS,IAAI,OAAO,QAAQ,IAAI,QAAQ,WAAW,QAAQ,wBAAwB,OAAO,YAAY,QAAQ,wBAAwB;AAClK,WAAO,GAAG,UAAU;AAAA,EACtB,CAAC,EAAE,KAAK,EAAE;AACZ;AACA,SAAS,2BAA2B,YAAY,cAAc;AAC5D,QAAM,YAAY,WAAW;AAC7B,QAAM,QAAQ,WAAW,IAAI,CAAC,MAAM,GAAG,gBAAgB,IAAI;AAC3D,QAAM,UAAU,IAAI,MAAM,YAAY,CAAC;AACvC,UAAQ,YAAY,KAAK,MAAM,YAAY;AAC3C,WAAS,IAAI,YAAY,GAAG,KAAK,GAAG,EAAE,GAAG;AACvC,YAAQ,KAAK,IAAI,QAAQ,IAAI,QAAQ,MAAM,IAAI;AAAA,EACjD;AACA,SAAO;AACT;AACA,SAAS,4CAA4C,SAAS,cAAc,QAAQ,SAAS;AAC3F,QAAM,eAAe,QAAQ,IAAI,CAAC,GAAG,MAAM,CAAC;AAC5C,QAAM,UAAU,2BAA2B,cAAc,YAAY;AACrE,SAAO,QAAQ,IAAI,CAAC,GAAG,MAAM;AAC3B,UAAM,QAAQ,OAAO,QAAQ,QAAQ,WAAW,QAAQ;AACxD,UAAM,QAAQ,MAAM,QAAQ,SAAS,IAAI,OAAO,QAAQ,IAAI,QAAQ,WAAW,QAAQ,QAAQ,QAAQ,OAAO,YAAY,QAAQ,QAAQ,QAAQ;AAClJ,WAAO,GAAG,UAAU;AAAA,EACtB,CAAC,EAAE,KAAK,EAAE;AACZ;AACA,SAAS,mBAAmB,OAAO;AACjC,QAAM,UAAU,aAAa,eAAe,KAAK,EAAE,IAAI,CAAC,MAAM,EAAE,SAAS,CAAC;AAC1E,SAAO;AAAA;AAAA,wBAEe,QAAQ,mBAAmB,QAAQ;AAAA;AAAA;AAG3D;AACA,SAAS,2BAA2B;AAClC,SAAO;AAAA;AAAA;AAAA;AAAA;AAKT;AACA,IAAI,uBAAuB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AA0C3B,IAAI,EAAE,kBAAkB,kBAAkB,IAAI;AAC9C,SAAS,WAAW,YAAY,aAAa,SAAS;AACpD,QAAM,iBAAiB,CAAC;AACxB,aAAW,QAAQ,CAAC,MAAM;AACxB,UAAM,OAAO,aAAa,cAAc,EAAE,UAAU,YAAY;AAChE,QAAI,EAAE,UAAU,WAAW;AACzB,qBAAe,KAAK,iBAAiB,EAAE,OAAO,OAAO,IAAI,IAAI,UAAU,KAAK;AAAA,IAC9E,OAAO;AACL,qBAAe,KAAK,qBAAqB,EAAE,OAAO;AAClD,qBAAe,KAAK,qBAAqB,EAAE,OAAO;AAAA,IACpD;AACA,QAAI,QAAQ,qBAAqB;AAC/B,YAAM,EAAE,aAAa,IAAI,wBAAwB,QAAQ,cAAc,EAAE,UAAU,cAAc,EAAE,UAAU,QAAQ;AACrH,cAAQ,aAAa,QAAQ;AAAA,QAC3B,KAAK;AACH,yBAAe,KAAK,eAAe,EAAE,YAAY;AACjD;AAAA,QACF,KAAK;AACH,yBAAe,KAAK,iBAAiB,EAAE,YAAY;AACnD;AAAA,QACF,KAAK;AACH,yBAAe,KAAK,iBAAiB,EAAE,YAAY;AACnD;AAAA,QACF,KAAK;AACH,yBAAe,KAAK,iBAAiB,EAAE,YAAY;AACnD;AAAA,QACF;AACE;AAAA,MACJ;AACA,qBAAe,KAAK,iBAAiB,EAAE,eAAe;AAAA,IACxD;AAAA,EACF,CAAC;AACD,MAAI,QAAQ,qBAAqB;AAC/B,YAAQ,YAAY,aAAa,QAAQ;AAAA,MACvC,KAAK;AACH,uBAAe,KAAK,uBAAuB;AAC3C;AAAA,MACF,KAAK;AACH,uBAAe,KAAK,yBAAyB;AAC7C,uBAAe,KAAK,8BAA8B;AAClD;AAAA,MACF,KAAK;AACH,uBAAe,KAAK,yBAAyB;AAC7C,uBAAe,KAAK,gCAAgC;AACpD;AAAA,MACF,KAAK;AACH,uBAAe,KAAK,yBAAyB;AAC7C,uBAAe,KAAK,gCAAgC;AACpD;AAAA,MACF;AACE;AAAA,IACJ;AACA,mBAAe,KAAK,4BAA4B;AAAA,EAClD;AACA,MAAI,QAAQ,gBAAgB;AAC1B,YAAQ,eAAe,QAAQ,CAAC,MAAM;AACpC,qBAAe,KAAK,WAAW,EAAE,QAAQ,EAAE,OAAO,EAAE,aAAa,IAAI,EAAE,gBAAgB,KAAK;AAAA,IAC9F,CAAC;AAAA,EACH;AACA,QAAM,qBAAqB,eAAe,KAAK,IAAI;AACnD,QAAM,uBAAuB,WAAW,IAAI,CAAC,MAAM,wBAAwB,GAAG,aAAa,QAAQ,cAAc,QAAQ,mBAAmB,CAAC,EAAE,KAAK,IAAI;AACxJ,QAAM,cAAc,YAAY;AAChC,QAAM,OAAO,mBAAmB;AAChC,QAAM,4BAA4B,6BAA6B,IAAI;AACnE,MAAI;AACJ,MAAI;AACJ,MAAI,eAAe,gBAAgB,IAAI;AACvC,MAAI,YAAY,UAAU;AACxB,4BAAwB,+BAA+B,YAAY,cAAc,aAAa,QAAQ,mBAAmB;AACzH,mCAA+B,8BAA8B,IAAI;AAAA,EACnE,OAAO;AACL,4BAAwB,yBAAyB,YAAY,cAAc,aAAa,QAAQ,mBAAmB;AACnH,mCAA+B,2BAA2B,IAAI;AAAA,EAChE;AACA,MAAI,QAAQ,cAAc;AACxB,oBAAgB;AAAA,EAClB;AACA,QAAM,SAAS;AAAA,IACb;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,QAAQ;AAAA,EACV,EAAE,KAAK,IAAI;AACX,SAAO;AACT;AACA,SAAS,qBAAqB,QAAQ,sBAAsB,OAAO;AACjE,QAAM,QAAQ,OAAO,UAAU;AAC/B,UAAQ,MAAM,QAAQ;AAAA,IACpB,KAAK;AACH,aAAO,iBAAiB,QAAQ,mBAAmB;AAAA,IACrD,KAAK;AACH,aAAO,aAAa,QAAQ,mBAAmB;AAAA,IACjD,KAAK;AACH,aAAO,aAAa,QAAQ,mBAAmB;AAAA,IACjD,KAAK;AACH,aAAO,aAAa,QAAQ,mBAAmB;AAAA,IACjD,KAAK;AACH,aAAO,aAAa,QAAQ,mBAAmB;AAAA,IACjD,KAAK;AACH,aAAO,aAAa,MAAM;AAAA,IAC5B,KAAK;AACH,aAAO,aAAa,MAAM;AAAA,IAC5B;AACE,YAAM,IAAI,MAAM,GAAG,MAAM,8CAA8C;AAAA,EAC3E;AACF;AACA,SAAS,2BAA2B,QAAQ,qBAAqB;AAC/D,QAAM,QAAQ,OAAO,UAAU;AAC/B,UAAQ,MAAM,QAAQ;AAAA,IACpB,KAAK;AACH,aAAO,uBAAuB,MAAM;AAAA,IACtC,KAAK;AACH,aAAO,mBAAmB,QAAQ,mBAAmB;AAAA,IACvD,KAAK;AACH,aAAO,mBAAmB,QAAQ,mBAAmB;AAAA,IACvD,KAAK;AACH,aAAO,mBAAmB,QAAQ,mBAAmB;AAAA,IACvD;AACE,aAAO,mBAAmB,QAAQ,mBAAmB;AAAA,EACzD;AACF;AACA,SAAS,wBAAwB,QAAQ,cAAc,qBAAqB,OAAO,qBAAqB;AACtG,MAAI,MAAM;AACV,MAAI,oBAAoB;AACtB,WAAO,2BAA2B,QAAQ,mBAAmB;AAAA,EAC/D,OAAO;AACL,WAAO,qBAAqB,QAAQ,mBAAmB;AAAA,EACzD;AACA,QAAM,UAAU,OAAO,UAAU;AACjC,QAAM,WAAW,aAAa;AAC9B,MAAI,QAAQ,UAAU,SAAS,QAAQ;AACrC,QAAI,oBAAoB;AACtB,aAAO,+BAA+B,QAAQ,YAAY;AAAA,IAC5D,OAAO;AACL,aAAO,yBAAyB,QAAQ,YAAY;AAAA,IACtD;AAAA,EACF;AACA,SAAO;AACT;AACA,SAAS,+BAA+B,UAAU,aAAa,qBAAqB;AAClF,UAAQ,SAAS,QAAQ;AAAA,IACvB,KAAK;AACH,aAAO,sBAAsB;AAAA,IAC/B,KAAK;AACH,aAAO,wBAAwB,UAAU,aAAa,mBAAmB;AAAA,IAC3E,KAAK;AACH,aAAO,wBAAwB,UAAU,aAAa,mBAAmB;AAAA,IAC3E,KAAK;AACH,aAAO,wBAAwB,UAAU,aAAa,mBAAmB;AAAA,IAC3E;AACE,aAAO,wBAAwB,UAAU,aAAa,mBAAmB;AAAA,EAC7E;AACF;AACA,SAAS,yBAAyB,UAAU,aAAa,qBAAqB;AAC5E,UAAQ,SAAS,QAAQ;AAAA,IACvB,KAAK;AACH,aAAO,sBAAsB;AAAA,IAC/B,KAAK;AACH,aAAO,kBAAkB,UAAU,aAAa,mBAAmB;AAAA,IACrE,KAAK;AACH,aAAO,kBAAkB,UAAU,aAAa,mBAAmB;AAAA,IACrE,KAAK;AACH,aAAO,kBAAkB,UAAU,aAAa,mBAAmB;AAAA,IACrE,KAAK;AACH,aAAO,kBAAkB,UAAU,aAAa,mBAAmB;AAAA,IACrE,KAAK;AACH,aAAO,kBAAkB,UAAU,WAAW;AAAA,IAChD,KAAK;AACH,aAAO,kBAAkB,UAAU,WAAW;AAAA,IAChD;AACE,YAAM,IAAI,MAAM,GAAG,SAAS,+CAA+C;AAAA,EAC/E;AACF;AACA,SAAS,6BAA6B,MAAM;AAC1C,SAAO;AAAA;AAAA,eAEM,KAAK;AAAA;AAAA;AAGpB;AACA,SAAS,2BAA2B,MAAM;AACxC,SAAO;AAAA;AAAA,QAED,KAAK;AAAA;AAAA;AAGb;AACA,SAAS,8BAA8B,MAAM;AAC3C,SAAO;AAAA;AAAA,QAED,KAAK;AAAA;AAAA;AAGb;AACA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,gBAAgB,GAAG,KAAK;AAAA;AAAA;AAAA;AAAA,MAI1B,KAAK;AAAA,MACL,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAuBL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAyBL;AAAA,MACA;AAAA,MACA;AAAA;AAEJ,SAAO;AACT;AACA,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAaxB,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AASxB,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUxB,IAAI,uBAAuB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAY3B,SAAS,wBAAwB;AAC/B,SAAO;AAAA;AAAA;AAAA;AAAA;AAKT;AACA,SAAS,wBAAwB,OAAO,UAAU,qBAAqB;AACrE,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,MAAI,eAAe,OAAO,GAAG;AAC3B,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKT;AACA,WAAO;AAAA;AAAA,sCAE2B,eAAe;AAAA;AAAA;AAAA,EAGnD;AACA,MAAI,eAAe,OAAO,GAAG;AAC3B,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKT;AACA,WAAO;AAAA;AAAA,sCAE2B,eAAe;AAAA;AAAA;AAAA,EAGnD;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQT;AACA,SAAO;AAAA;AAAA;AAAA,oCAG2B,eAAe,OAAO,eAAe;AAAA,iCACxC,eAAe;AAAA;AAAA;AAGhD;AACA,SAAS,kBAAkB,OAAO,UAAU,qBAAqB;AAC/D,MAAI,SAAS,OAAO,GAAG;AACrB,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKT;AACA,WAAO;AAAA;AAAA,kCAEuB,SAAS;AAAA;AAAA;AAAA,EAGzC;AACA,MAAI,SAAS,OAAO,GAAG;AACrB,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKT;AACA,WAAO;AAAA;AAAA,kCAEuB,SAAS;AAAA;AAAA;AAAA,EAGzC;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOT;AACA,SAAO;AAAA;AAAA;AAAA,oCAG2B,SAAS,OAAO,SAAS;AAAA,4BACjC,SAAS;AAAA;AAAA;AAGrC;AACA,SAAS,wBAAwB,OAAO,UAAU,qBAAqB;AACrE,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkBT;AACA,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,QAAM,qBAAqB,KAAK,KAAK,MAAM,KAAK,CAAC;AACjD,QAAM,gBAAgB,qBAAqB,KAAK,KAAK,MAAM,KAAK,CAAC;AACjE,SAAO;AAAA;AAAA;AAAA,oCAG2B,eAAe,OAAO,eAAe;AAAA,iCACxC,eAAe;AAAA;AAAA,wBAExB;AAAA,qBACH;AAAA;AAAA,6BAEQ;AAAA,4BACD;AAAA;AAAA;AAAA;AAAA;AAK5B;AACA,SAAS,kBAAkB,OAAO,UAAU,qBAAqB;AAC/D,MAAI,qBAAqB;AACvB,UAAM,0BAA0B,kDAAkD,CAAC,KAAK,KAAK,GAAG,GAAG,KAAK;AACxG,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA,MAKL;AAAA;AAAA;AAAA;AAAA,EAIJ;AACA,QAAM,yBAAyB,mCAAmC,CAAC,KAAK,KAAK,GAAG,GAAG,KAAK;AACxF,SAAO;AAAA;AAAA;AAAA,oCAG2B,SAAS,OAAO,SAAS;AAAA,iCAC5B,SAAS;AAAA,QAClC;AAAA;AAAA;AAAA;AAIR;AACA,SAAS,wBAAwB,OAAO,UAAU,qBAAqB;AACrE,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAuBT;AACA,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,QAAM,qBAAqB,KAAK,KAAK,MAAM,MAAM,SAAS,KAAK,CAAC;AAChE,QAAM,gBAAgB,qBAAqB,KAAK,KAAK,MAAM,MAAM,SAAS,KAAK,CAAC;AAChF,MAAI,iBAAiB;AACrB,MAAI,UAAU;AACd,MAAI,UAAU;AACd,WAAS,IAAI,GAAG,IAAI,MAAM,SAAS,GAAG,KAAK;AACzC,sBAAkB,MAAM,MAAM,SAAS,IAAI;AAC3C,cAAU;AAAA,aACD,eAAe;AAAA,kBACV,OAAO;AAAA,QACjB;AACJ,cAAU,IAAI,QAAQ;AAAA,EACxB;AACA,SAAO;AAAA,UACC,MAAM;AAAA;AAAA,oCAEoB,eAAe,OAAO,eAAe;AAAA,iCACxC,eAAe;AAAA;AAAA,QAExC;AAAA;AAAA,wBAEgB;AAAA,qBACH;AAAA;AAAA,6BAEQ;AAAA,4BACD;AAAA;AAAA,mBAET,MAAM,UAAU;AAAA;AAAA;AAGnC;AACA,SAAS,kBAAkB,OAAO,UAAU,qBAAqB;AAC/D,MAAI,qBAAqB;AACvB,UAAM,0BAA0B,kDAAkD,CAAC,KAAK,KAAK,KAAK,IAAI,GAAG,KAAK;AAC9G,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA,QAKH;AAAA;AAAA;AAAA;AAAA,EAIN;AACA,QAAM,yBAAyB,mCAAmC,CAAC,KAAK,KAAK,KAAK,IAAI,GAAG,KAAK;AAC9F,SAAO;AAAA;AAAA;AAAA,eAGM,SAAS,OAAO,SAAS;AAAA,iCACP,SAAS;AAAA,QAClC;AAAA;AAAA;AAAA;AAIR;AACA,SAAS,kBAAkB,OAAO,UAAU;AAC1C,QAAM,yBAAyB,mCAAmC,CAAC,KAAK,KAAK,KAAK,MAAM,IAAI,GAAG,KAAK;AACpG,SAAO;AAAA;AAAA,kDAEyC,SAAS;AAAA,+BAC5B,SAAS;AAAA;AAAA,iCAEP,SAAS;AAAA;AAAA,QAElC;AAAA;AAAA;AAAA;AAAA;AAAA;AAMR;AACA,SAAS,kBAAkB,OAAO,UAAU;AAC1C,QAAM,yBAAyB,mCAAmC,CAAC,KAAK,KAAK,KAAK,MAAM,MAAM,IAAI,GAAG,KAAK;AAC1G,SAAO;AAAA;AAAA;AAAA,eAGM,SAAS,OAAO,SAAS;AAAA,iCACP,SAAS;AAAA;AAAA,QAElC;AAAA;AAAA;AAAA;AAAA;AAAA;AAMR;AACA,SAAS,wBAAwB,OAAO,UAAU,qBAAqB;AACrE,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,MAAI,aAAa,YAAY,OAAO,QAAQ,GAAG;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAMT;AACA,WAAO;AAAA;AAAA,8CAEmC,eAAe,OAAO,eAAe;AAAA;AAAA;AAAA,EAGjF;AACA,QAAM,qBAAqB,KAAK,KAAK,MAAM,KAAK,CAAC;AACjD,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAcT;AACA,SAAO;AAAA;AAAA;AAAA,oCAG2B,eAAe,OAAO,eAAe;AAAA;AAAA,iCAExC,eAAe;AAAA,6BACnB;AAAA,4BACD;AAAA;AAAA;AAAA;AAAA;AAK5B;AACA,SAAS,kBAAkB,OAAO,UAAU,qBAAqB;AAC/D,MAAI,aAAa,YAAY,OAAO,QAAQ,GAAG;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKT;AACA,WAAO;AAAA;AAAA,0CAE+B,SAAS,OAAO,SAAS;AAAA;AAAA;AAAA,EAGjE;AACA,MAAI,MAAM,OAAO,GAAG;AAClB,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAQT;AACA,WAAO;AAAA;AAAA;AAAA,sCAG2B,SAAS,OAAO,SAAS;AAAA,mCAC5B,SAAS;AAAA;AAAA;AAAA;AAAA,EAI1C;AACA,MAAI,MAAM,OAAO,GAAG;AAClB,QAAI,qBAAqB;AACvB,aAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAQT;AACA,WAAO;AAAA;AAAA;AAAA,sCAG2B,SAAS,OAAO,SAAS;AAAA,mCAC5B,SAAS;AAAA;AAAA;AAAA;AAAA,EAI1C;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUT;AACA,SAAO;AAAA;AAAA;AAAA,oCAG2B,SAAS,OAAO,SAAS;AAAA,iCAC5B,SAAS;AAAA,wBAClB,MAAM;AAAA,4BACF,MAAM;AAAA;AAAA;AAAA;AAIlC;AACA,SAAS,yBAAyB,SAAS;AACzC,SAAO,SAAS;AAClB;AACA,SAAS,uBAAuB,WAAW;AACzC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,OAAO,mBAAmB;AAChC,SAAO;AAAA,WACE;AAAA,eACI,KAAK,aAAa;AAAA;AAAA;AAGjC;AACA,SAAS,iBAAiB,WAAW,qBAAqB;AACxD,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO,SAAS,sBAAsB;AAAA,EACxC;AACA,QAAM,CAAC,SAAS,OAAO,IAAI,UAAU,UAAU;AAC/C,MAAI,YAAY,KAAK,YAAY,GAAG;AAClC,WAAO;AAAA,cACG;AAAA,+BACiB;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,MAAI,qBAAqB;AACvB,WAAO;AAAA,YACC;AAAA,6BACiB,uBAAuB,uBAAuB;AAAA,6BAC9C;AAAA;AAAA;AAAA,EAG3B;AACA,QAAM,CAAC,OAAO,KAAK,IAAI,UAAU,UAAU;AAC3C,SAAO;AAAA,YACG;AAAA,6BACiB,UAAU,UAAU;AAAA,6BACpB;AAAA;AAAA;AAG7B;AACA,SAAS,mBAAmB,WAAW,qBAAqB;AAC1D,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,OAAO,mBAAmB;AAChC,MAAI,qBAAqB;AACvB,WAAO;AAAA,WACA;AAAA,gDACqC,0CAA0C;AAAA;AAAA;AAAA,eAG3E,KAAK,aAAa;AAAA;AAAA;AAAA,EAG/B;AACA,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,SAAO;AAAA,WACE;AAAA;AAAA,UAED,eAAe,OAAO,eAAe;AAAA,eAChC,KAAK,aAAa;AAAA;AAAA;AAGjC;AACA,SAAS,aAAa,WAAW,qBAAqB;AACpD,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA,UACJ,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,QAAQ,SAAS;AACvB,QAAM,QAAQ,SAAS;AACvB,MAAI,UAAU,KAAK,UAAU,GAAG;AAC9B,WAAO;AAAA,cACG;AAAA,+BACiB;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,MAAI,UAAU,GAAG;AACf,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,6CAC+B,0BAA0B;AAAA,+BACxC;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,cACG;AAAA,6CAC+B,oBAAoB;AAAA,+BAClC;AAAA;AAAA;AAAA,EAG7B;AACA,MAAI,UAAU,GAAG;AACf,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,wCAC0B,0BAA0B;AAAA,+BACnC;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,cACG;AAAA,wCAC0B,oBAAoB;AAAA,+BAC7B;AAAA;AAAA;AAAA,EAG7B;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA,YACC;AAAA,6BACiB,uBAAuB,+BAA+B;AAAA,6BACtD;AAAA;AAAA;AAAA,EAG3B;AACA,SAAO;AAAA,YACG;AAAA,6BACiB,UAAU,kBAAkB;AAAA,6BAC5B;AAAA;AAAA;AAG7B;AACA,SAAS,mBAAmB,WAAW,qBAAqB;AAC1D,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,QAAM,OAAO,mBAAmB;AAChC,MAAI,YAAY,QAAQ,aAAa,YAAY,OAAO,QAAQ,GAAG;AACjE,QAAI,qBAAqB;AACvB,aAAO;AAAA,aACA;AAAA,qDACwC,uBAAuB;AAAA;AAAA,iBAE3D,KAAK,aAAa;AAAA;AAAA;AAAA,IAG/B;AACA,WAAO;AAAA,aACE;AAAA,qDACwC,cAAc;AAAA;AAAA,iBAElD,KAAK,aAAa;AAAA;AAAA;AAAA,EAGjC;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA,WACA;AAAA,gDACqC,0CAA0C;AAAA,0CAChD;AAAA;AAAA,eAE3B,KAAK,aAAa;AAAA;AAAA;AAAA,EAG/B;AACA,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,QAAM,eAAe,KAAK,KAAK,MAAM,KAAK,CAAC;AAC3C,SAAO;AAAA,WACE;AAAA,iCACsB,iBAAiB,eAAe,OAAO,eAAe;AAAA,eACxE,KAAK,aAAa;AAAA;AAAA;AAGjC;AACA,SAAS,aAAa,WAAW,qBAAqB;AACpD,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,WAAW,UAAU,UAAU;AACrC,MAAI,YAAY,QAAQ,aAAa,YAAY,OAAO,QAAQ,GAAG;AACjE,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,qDACuC,uBAAuB;AAAA,+BAC7C;AAAA;AAAA;AAAA,IAG3B;AACA,UAAM,WAAW,SAAS;AAC1B,UAAM,WAAW,SAAS;AAC1B,WAAO;AAAA,YACC;AAAA,mDACuC,eAAe;AAAA,6BACrC;AAAA;AAAA;AAAA,EAG3B;AACA,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,QAAM,gBAAgB;AACtB,MAAI,cAAc,SAAS,MAAM,QAAQ;AACvC,UAAM,eAAe,iBAAiB,WAAW,aAAa;AAC9D,UAAM,SAAS,CAAC,OAAO,KAAK;AAC5B,WAAO;AAAA,QACH,qBAAqB,cAAc,mBAAmB;AAAA,cAChD;AAAA,iBACG,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG7D;AACA,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA,qDACuC,MAAM;AAAA,UACjD,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,QAAM,SAAS,yBAAyB,OAAO;AAC/C,MAAI,YAAY,GAAG;AACjB,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,2CAC6B,iBAAiB;AAAA,oDACR;AAAA,+BACrB;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,YACC;AAAA,yCAC6B,iBAAiB,MAAM;AAAA,4CACpB;AAAA,6BACf;AAAA;AAAA;AAAA,EAG3B;AACA,MAAI,YAAY,GAAG;AACjB,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,2CAC6B,iBAAiB;AAAA,+CACb;AAAA,+BAChB;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,YACC;AAAA,yCAC6B,iBAAiB,MAAM;AAAA,uCACzB;AAAA,6BACV;AAAA;AAAA;AAAA,EAG3B;AACA,MAAI,qBAAqB;AACvB,WAAO;AAAA,cACG;AAAA;AAAA,4BAEc,2BAA2B;AAAA,+BACxB,uBAAuB;AAAA,+BACvB;AAAA;AAAA;AAAA,EAG7B;AACA,SAAO;AAAA,UACC;AAAA;AAAA,wBAEc,MAAM,cAAc;AAAA,2BACjB,YAAY;AAAA,2BACZ;AAAA;AAAA;AAG3B;AACA,SAAS,mBAAmB,WAAW,qBAAqB;AAC1D,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,MAAI,MAAM,OAAO,GAAG;AAClB,UAAM,gBAAgB,MAAM,MAAM,CAAC;AACnC,UAAM,WAAW,CAAC,GAAG,CAAC;AACtB,UAAM,eAAe,iBAAiB,WAAW,aAAa;AAC9D,UAAM,SAAS,CAAC,KAAK,OAAO,KAAK;AACjC,WAAO;AAAA,UACD,2BAA2B,cAAc,mBAAmB;AAAA,eACvD;AAAA,mBACI,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG/D;AACA,QAAM,OAAO,mBAAmB;AAChC,MAAI,qBAAqB;AACvB,WAAO;AAAA,WACA;AAAA,gDACqC,0CAA0C;AAAA,0CAChD;AAAA,0DACgB;AAAA;AAAA;AAAA,eAG3C,KAAK,aAAa;AAAA;AAAA;AAAA,EAG/B;AACA,QAAM,UAAU,eAAe;AAC/B,QAAM,UAAU,eAAe;AAC/B,QAAM,eAAe,KAAK,KAAK,MAAM,KAAK,CAAC;AAC3C,QAAM,gBAAgB,eAAe,KAAK,KAAK,MAAM,KAAK,CAAC;AAC3D,SAAO;AAAA,WACE;AAAA;AAAA,UAED,YAAY,YAAY,kBAAkB;AAAA,eACrC,KAAK,aAAa;AAAA;AAAA;AAGjC;AACA,SAAS,aAAa,WAAW,qBAAqB;AACpD,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,UAAU,MAAM,KAAK,MAAM;AACjC,QAAM,UAAU,MAAM;AACtB,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,QAAM,gBAAgB;AACtB,MAAI,cAAc,SAAS,MAAM,QAAQ;AACvC,UAAM,eAAe,iBAAiB,WAAW,aAAa;AAC9D,UAAM,SAAS,CAAC,OAAO,OAAO,OAAO;AACrC,WAAO;AAAA,UACD,qBAAqB,cAAc,mBAAmB;AAAA,gBAChD;AAAA,mBACG,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG/D;AACA,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA;AAAA,iCAEmB,YAAY;AAAA,UACnC,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,QAAM,aAAa,UAAU,UAAU;AACvC,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,wBACU;AAAA;AAAA;AAAA;AAAA,0BAIE,uBAAuB;AAAA,+BAClB;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,gBACK;AAAA;AAAA,oDAEoC;AAAA;AAAA,4BAExB,cAAc;AAAA,iCACT;AAAA;AAAA;AAAA,EAG/B;AACA,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,gDACkC;AAAA;AAAA,uDAEO,uBAAuB;AAAA,+BAC/C;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,YACC;AAAA,8CACkC,MAAM;AAAA;AAAA,qDAEC,cAAc;AAAA,6BACtC;AAAA;AAAA;AAAA,EAG3B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,MAAI,qBAAqB;AACvB,WAAO;AAAA,YACC;AAAA;AAAA,sBAEU,qBAAqB;AAAA,sBACrB;AAAA,4DACsC;AAAA,6BAC/B,uBAAuB;AAAA,6BACvB;AAAA;AAAA;AAAA,EAG3B;AACA,SAAO;AAAA,cACK;AAAA;AAAA,4BAEc,mBAAmB,qBAAqB;AAAA,+BACrC,YAAY;AAAA,+BACZ;AAAA;AAAA;AAG/B;AACA,SAAS,mBAAmB,WAAW,qBAAqB;AAC1D,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,OAAO,mBAAmB;AAChC,MAAI,qBAAqB;AACvB,WAAO;AAAA,WACA;AAAA,0CAC+B;AAAA,0DACgB;AAAA;AAAA,yBAEjC;AAAA;AAAA,gDAEuB,0CAA0C;AAAA;AAAA;AAAA,mGAGS,KAAK,aAAa;AAAA;AAAA;AAAA,EAGnH;AACA,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,OAAO,MAAM;AACnB,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,iBAAiB,CAAC,KAAK,KAAK,SAAS,KAAK,CAAC,GAAG,KAAK,KAAK,SAAS,KAAK,CAAC,CAAC;AAC9E,QAAM,UAAU,eAAe;AAC/B,QAAM,UAAU,eAAe;AAC/B,QAAM,eAAe,KAAK,KAAK,MAAM,OAAO,KAAK,CAAC;AAClD,MAAI,gBAAgB,eAAe,KAAK,KAAK,MAAM,OAAO,KAAK,CAAC;AAChE,MAAI,SAAS;AACb,MAAI,QAAQ,OAAO,+BAA+B;AAClD,WAAS,IAAI,GAAG,IAAI,OAAO,GAAG,KAAK;AACjC,aAAS,QAAQ,QAAQ;AACzB,qBAAiB,MAAM,OAAO,IAAI;AAClC,YAAQ,IAAI,OAAO,qBAAqB;AAAA,EAC1C;AACA,SAAO;AAAA,WACE,YAAY;AAAA,oBACH;AAAA,2BACO;AAAA,kCACO;AAAA,qDACmB,YAAY;AAAA,eAClD,KAAK,aAAa;AAAA;AAAA;AAGjC;AACA,SAAS,aAAa,WAAW,qBAAqB;AACpD,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,UAAU,MAAM;AACtB,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,MAAI,SAAS,SAAS,MAAM,QAAQ;AAClC,UAAM,eAAe,iBAAiB,WAAW,QAAQ;AACzD,UAAM,SAAS,CAAC,OAAO,OAAO,SAAS,QAAQ;AAC/C,WAAO;AAAA,QACH,qBAAqB,cAAc,mBAAmB;AAAA,cAChD;AAAA,iBACG,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG7D;AACA,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA;AAAA,iCAEmB,YAAY,YAAY;AAAA,UAC/C,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,aAAa,UAAU,UAAU;AACvC,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,QAAM,aAAa,iBAAiB;AACpC,QAAM,aAAa,iBAAiB;AACpC,QAAM,aAAa,iBAAiB;AACpC,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA,UACJ;AAAA,UACA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAMgB,uBAAuB;AAAA,+BAClB;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,cACG;AAAA;AAAA;AAAA;AAAA,uBAIS,YAAY;AAAA;AAAA,0BAET,cAAc;AAAA,+BACT;AAAA;AAAA;AAAA,EAG7B;AACA,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,QAAI,qBAAqB;AACvB,aAAO;AAAA,cACC;AAAA;AAAA,gCAEkB,qBAAqB,oBAAoB;AAAA;AAAA;AAAA,yBAGhD,uBAAuB;AAAA,+BACjB;AAAA;AAAA;AAAA,IAG3B;AACA,WAAO;AAAA,cACG;AAAA;AAAA,gCAEkB,MAAM,KAAK,MAAM,OAAO,MAAM;AAAA;AAAA;AAAA,yBAGrC,cAAc;AAAA,+BACR;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,MAAI,qBAAqB;AACvB,WAAO;AAAA,YACC;AAAA;AAAA,QAEJ;AAAA,QACA;AAAA,QACA;AAAA;AAAA;AAAA,6BAGqB,uBAAuB,+BAA+B;AAAA,6BACtD;AAAA;AAAA;AAAA,EAG3B;AACA,SAAO;AAAA,YACG;AAAA;AAAA,0BAEc,mBAAmB;AAAA,oBACzB;AAAA,6BACS,YAAY,oBAAoB;AAAA,6BAChC;AAAA;AAAA;AAG7B;AACA,SAAS,aAAa,WAAW;AAC/B,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,UAAU,MAAM;AACtB,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,MAAI,SAAS,SAAS,MAAM,QAAQ;AAClC,UAAM,eAAe,iBAAiB,WAAW,QAAQ;AACzD,UAAM,SAAS,CAAC,OAAO,OAAO,SAAS,UAAU,QAAQ;AACzD,WAAO;AAAA,QACH,qBAAqB,YAAY;AAAA,cAC3B;AAAA,iBACG,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG7D;AACA,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA;AAAA;AAAA,iBAGG,YAAY,YAAY,YAAY;AAAA;AAAA,UAE3C,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,aAAa,UAAU,UAAU;AACvC,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,WAAO;AAAA,cACG;AAAA;AAAA;AAAA,gCAGkB,YAAY,YAAY;AAAA;AAAA,0BAE9B,cAAc;AAAA,+BACT;AAAA;AAAA;AAAA,EAG7B;AACA,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,WAAO;AAAA,cACG;AAAA;AAAA;AAAA,iBAGG,MAAM,KAAK,MAAM,KAAK,MAAM;AAAA,iBAC5B,MAAM,KAAK,MAAM,OAAO,MAAM;AAAA;AAAA;AAAA,yBAGtB,cAAc;AAAA,+BACR;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,SAAO;AAAA,YACG;AAAA;AAAA,0BAEc,mBAAmB,qBAAqB;AAAA,qBAC7C,sBAAsB;AAAA,6BACd,YAAY;AAAA,6BACZ;AAAA;AAAA;AAG7B;AACA,SAAS,aAAa,WAAW;AAC/B,QAAM,QAAQ,UAAU,UAAU;AAClC,QAAM,UAAU,UAAU;AAC1B,QAAM,WAAW,QAAQ,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AAC1E,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,MAAI,SAAS,SAAS,MAAM,QAAQ;AAClC,UAAM,eAAe,iBAAiB,WAAW,QAAQ;AACzD,UAAM,SAAS,CAAC,OAAO,OAAO,SAAS,UAAU,UAAU,QAAQ;AACnE,WAAO;AAAA,QACH,qBAAqB,YAAY;AAAA,cAC3B;AAAA;AAAA,iBAEG,YAAY,kBAAkB,QAAQ,QAAQ;AAAA;AAAA;AAAA,EAG7D;AACA,QAAM,UAAU,MAAM;AACtB,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,QAAM,UAAU,MAAM,KAAK;AAC3B,MAAI,UAAU,UAAU,WAAW;AACjC,WAAO;AAAA,cACG;AAAA;AAAA;AAAA;AAAA,iBAIG,YAAY,YAAY,YAAY;AAAA;AAAA;AAAA,mBAGlC;AAAA,UACT,kBAAkB,SAAS;AAAA;AAAA;AAAA,EAGnC;AACA,QAAM,aAAa,UAAU,UAAU;AACvC,QAAM,WAAW,UAAU,UAAU;AACrC,QAAM,UAAU,SAAS;AACzB,QAAM,UAAU,SAAS;AACzB,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,WAAO;AAAA,cACG;AAAA;AAAA;AAAA;AAAA,iBAIG,YAAY,YAAY,YAAY;AAAA;AAAA;AAAA,0BAG3B,cAAc;AAAA,+BACT;AAAA;AAAA;AAAA,EAG7B;AACA,MAAI,YAAY,WAAW,cAAc,MAAM;AAC7C,WAAO;AAAA,cACG;AAAA;AAAA;AAAA,iBAGG,MAAM,KAAK,MAAM,KAAK,MAAM,KAAK,MAAM;AAAA,iBACvC,MAAM,KAAK,MAAM,KAAK,MAAM;AAAA,iBAC5B,MAAM,KAAK,MAAM;AAAA,iBACjB,MAAM;AAAA;AAAA;AAAA,yBAGE,cAAc;AAAA,+BACR;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,SAAS,yBAAyB,OAAO;AAC/C,SAAO;AAAA,YACG;AAAA;AAAA;AAAA,0BAGc,mBAAmB,qBAAqB;AAAA,qBAC7C,sBAAsB,sBAAsB;AAAA,6BACpC,YAAY;AAAA,6BACZ;AAAA;AAAA;AAG7B;AACA,SAAS,kBAAkB,WAAW;AACpC,QAAM,UAAU,UAAU;AAC1B,QAAM,SAAS,aAAa,cAAc,UAAU,UAAU,YAAY;AAC1E,MAAI,SAAS,GAAG;AACd,WAAO,UAAU;AAAA,EACnB;AACA,SAAO;AAAA,0BACiB;AAAA;AAAA,iBAET;AAAA;AAAA;AAAA;AAIjB;AACA,SAAS,+BAA+B,WAAW,cAAc;AAC/D,QAAM,UAAU,UAAU;AAC1B,QAAM,iBAAiB,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AACxE,QAAM,WAAW,QAAQ,iBAAiB;AAC1C,QAAM,SAAS,UAAU,UAAU,aAAa;AAChD,QAAM,UAAU,aAAa,aAAa;AAC1C,QAAM,gBAAgB,kBAAkB,UAAU,UAAU,cAAc,aAAa,YAAY;AACnG,QAAM,OAAO,kBAAkB,OAAO;AACtC,QAAM,WAAW,UAAU;AAC3B,MAAI;AACJ,QAAM,SAAS,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG;AAC5C,MAAI,WAAW,GAAG;AAChB,oBAAgB;AAAA,EAClB,WAAW,UAAU,KAAK,cAAc,UAAU,GAAG;AACnD,oBAAgB;AAAA,EAClB,OAAO;AACL,oBAAgB,cAAc,IAAI,CAAC,MAAM,UAAU,OAAO,IAAI,gBAAgB,EAAE,KAAK,IAAI;AAAA,EAC3F;AACA,MAAI,wBAAwB;AAC5B,MAAI,UAAU,KAAK,SAAS,GAAG;AAC7B,4BAAwB;AAAA,EAC1B,OAAO;AACL,4BAAwB,UAAU,UAAU,aAAa,IAAI,CAAC,GAAG,MAAM,UAAU,OAAO,IAAI,WAAW,EAAE,KAAK,IAAI;AAAA,EACpH;AACA,MAAI,SAAS;AACb,QAAM,SAAS,aAAa,cAAc,UAAU,UAAU,YAAY;AAC1E,QAAM,gBAAgB,WAAW;AACjC,QAAM,UAAU,aAAa,cAAc,aAAa,YAAY;AACpE,QAAM,iBAAiB,YAAY;AACnC,MAAI,WAAW,KAAK,CAAC,iBAAiB,CAAC,gBAAgB;AACrD,aAAS;AAAA;AAAA;AAAA,EAGX,WAAW,iBAAiB,CAAC,gBAAgB;AAC3C,QAAI,YAAY,GAAG;AACjB,eAAS;AAAA;AAAA;AAAA,IAGX,OAAO;AACL,eAAS;AAAA;AAAA;AAAA,IAGX;AAAA,EACF,WAAW,cAAc,QAAQ;AAC/B,UAAM,OAAO,SAAS;AACtB,UAAM,OAAO,SAAS;AACtB,QAAI,cAAc,QAAQ,IAAI,IAAI,MAAM,cAAc,QAAQ,IAAI,IAAI,IAAI;AACxE,eAAS;AAAA,IACX,WAAW,cAAc,QAAQ,IAAI,IAAI,IAAI;AAC3C,eAAS;AAAA,IACX,WAAW,cAAc,QAAQ,IAAI,IAAI,IAAI;AAC3C,eAAS;AAAA,IACX;AAAA,EACF;AACA,SAAO;AAAA,WACE;AAAA,QACH;AAAA,QACA;AAAA,8BACsB,kBAAkB;AAAA,QACxC;AAAA;AAAA;AAGR;AACA,SAAS,yBAAyB,WAAW,cAAc;AACzD,QAAM,UAAU,UAAU;AAC1B,QAAM,iBAAiB,QAAQ,OAAO,CAAC,EAAE,YAAY,IAAI,QAAQ,MAAM,CAAC;AACxE,QAAM,WAAW,QAAQ,iBAAiB;AAC1C,QAAM,cAAc,aAAa;AACjC,QAAM,aAAa,UAAU,UAAU;AACvC,QAAM,SAAS,UAAU,UAAU,aAAa;AAChD,QAAM,UAAU,aAAa,aAAa;AAC1C,MAAI,CAAC,UAAU,UAAU,aAAa,WAAW,WAAW,UAAU,UAAU,cAAc,QAAQ,aAAa,YAAY,YAAY,WAAW,GAAG;AACvJ,WAAO;AAAA,cACG;AAAA,+BACiB;AAAA;AAAA;AAAA,EAG7B;AACA,QAAM,OAAO,kBAAkB,OAAO;AACtC,QAAM,gBAAgB,kBAAkB,UAAU,UAAU,cAAc,aAAa,YAAY;AACnG,QAAM,WAAW,UAAU;AAC3B,MAAI;AACJ,QAAM,SAAS,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG;AAC5C,MAAI,WAAW,GAAG;AAChB,oBAAgB;AAAA,EAClB,WAAW,UAAU,KAAK,cAAc,UAAU,GAAG;AACnD,oBAAgB;AAAA,EAClB,OAAO;AACL,oBAAgB,cAAc,IAAI,CAAC,MAAM,UAAU,OAAO,IAAI,gBAAgB,EAAE,KAAK,IAAI;AAAA,EAC3F;AACA,MAAI,wBAAwB;AAC5B,MAAI,UAAU,KAAK,SAAS,GAAG;AAC7B,4BAAwB;AAAA,EAC1B,OAAO;AACL,4BAAwB,UAAU,UAAU,aAAa,IAAI,CAAC,GAAG,MAAM,UAAU,OAAO,IAAI,WAAW,EAAE,KAAK,IAAI;AAAA,EACpH;AACA,SAAO;AAAA,YACG;AAAA,QACJ;AAAA,QACA;AAAA,kBACU,kBAAkB;AAAA;AAAA;AAGpC;AACA,SAAS,kBAAkB,MAAM;AAC/B,MAAI,QAAQ,GAAG;AACb,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO;AAAA,EACT,WAAW,SAAS,GAAG;AACrB,WAAO;AAAA,EACT,OAAO;AACL,UAAM,MAAM,gBAAgB,2BAA2B;AAAA,EACzD;AACF;AACA,SAAS,wBAAwB,UAAU,OAAO,UAAU;AAC1D,QAAM,EAAE,UAAU,SAAS,IAAI,aAAa,aAAa,KAAK;AAC9D,QAAM,OAAO,MAAM;AACnB,QAAM,wBAAwB,YAAY,SAAS,KAAK,MAAM,OAAO;AACrE,QAAM,gBAAgB,wBAAwB,MAAM,MAAM,CAAC,IAAI;AAC/D,QAAM,kBAAkB,CAAC,YAAY,OAAO,KAAK,CAAC,aAAa,YAAY,OAAO,QAAQ,KAAK,SAAS,SAAS,QAAQ;AACzH,QAAM,eAAe,kBAAkB,gBAAgB;AACvD,SAAO,EAAE,iBAAiB,cAAc,SAAS;AACnD;AACA,SAAS,iBAAiB,QAAQ,eAAe;AAC/C,QAAM,eAAe,KAAK,MAAM,KAAK,UAAU,MAAM,CAAC;AACtD,eAAa,UAAU,eAAe;AACtC,SAAO;AACT;AACA,SAAS,kBAAkB,QAAQ,UAAU;AAC3C,SAAO,SAAS,IAAI,CAAC,MAAM,OAAO,EAAE,EAAE,KAAK,IAAI;AACjD;AAGA,SAAS,eAAe,OAAO,SAAS,QAAQ,QAAQ;AACtD,QAAM,aAAa,OAAO,IAAI,CAAC,QAAQ,MAAM;AAC3C,UAAM,YAAY;AAAA,MAChB,cAAc,OAAO;AAAA,MACrB,UAAU,OAAO,YAAY,OAAO,OAAO,QAAQ;AAAA,MACnD,WAAW,OAAO;AAAA,MAClB,UAAU,OAAO,YAAY,QAAQ,OAAO,QAAQ;AAAA,MACpD,YAAY;AAAA,IACd;AACA,QAAI,OAAO,WAAW,QAAQ,OAAO,QAAQ,SAAS,QAAQ,OAAO,QAAQ,MAAM,aAAa,GAAG;AACjG,gBAAU,aAAa,OAAO,QAAQ,MAAM;AAAA,IAC9C;AACA,WAAO,EAAE,MAAM,QAAQ,cAAc,IAAI,UAAU;AAAA,EACrD,CAAC;AACD,QAAM,eAAe,WAAW,IAAI,CAAC,MAAM,EAAE,SAAS;AACtD,QAAM,eAAe;AAAA,IACnB,cAAc,OAAO;AAAA,IACrB,UAAU,OAAO,QAAQ;AAAA,IACzB,WAAW;AAAA,IACX,UAAU,OAAO,QAAQ;AAAA,IACzB,YAAY;AAAA,EACd;AACA,QAAM,SAAS,WAAW,YAAY,cAAc,OAAO;AAC3D,QAAM,iBAAiB,qBAAqB,MAAM,IAAI,MAAM;AAC5D,QAAM,eAAe,MAAM,cAAc,cAAc;AACvD,MAAI,CAAC,IAAI,EAAE,IAAI,qBAAqB,GAAG;AACrC,WAAO,OAAO,OAAO;AAAA,MACnB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,GAAG,oBAAoB,OAAO,SAAS,YAAY,CAAC;AAAA,EACtD,OAAO;AACL,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,kBAAkB;AAAA,MAClB,wBAAwB;AAAA,MACxB,QAAQ;AAAA,MACR,QAAQ;AAAA,MACR,mBAAmB;AAAA,MACnB,sBAAsB;AAAA,MACtB,kBAAkB;AAAA,MAClB,yBAAyB;AAAA,MACzB,qBAAqB;AAAA,IACvB;AAAA,EACF;AACF;AACA,SAAS,oBAAoB,OAAO,SAAS,cAAc;AACzD,QAAM,mBAAmB,CAAC;AAC1B,QAAM,oBAAoB,CAAC;AAC3B,QAAM,uBAAuB,CAAC;AAC9B,QAAM,yBAAyB,CAAC;AAChC,MAAI;AACJ,MAAI;AACJ,MAAI;AACJ,MAAI,SAAS;AACb,MAAI,SAAS;AACb,WAAS,MAAM,mBAAmB,cAAc,OAAO,KAAK;AAC5D,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,aAAS,MAAM,mBAAmB,cAAc,YAAY,KAAK;AAAA,EACnE;AACA,QAAM,cAAc;AACpB,WAAS,IAAI,GAAG,IAAI,QAAQ,cAAc,QAAQ,KAAK;AACrD,UAAM,UAAU,QAAQ,cAAc;AACtC,qBAAiB,WAAW,MAAM,mBAAmB,cAAc,SAAS,WAAW;AACvF,qBAAiB,SAAS,aAAa,MAAM,mBAAmB,cAAc,SAAS,WAAW,WAAW;AAC7G,QAAI,QAAQ,qBAAqB;AAC/B,wBAAkB,GAAG,kBAAkB,MAAM,mBAAmB,cAAc,GAAG,gBAAgB,WAAW;AAC5G,2BAAqB,GAAG,qBAAqB,MAAM,mBAAmB,cAAc,GAAG,mBAAmB,WAAW;AAAA,IACvH;AAAA,EACF;AACA,MAAI,QAAQ,qBAAqB;AAC/B,uBAAmB,MAAM,mBAAmB,cAAc,YAAY,WAAW;AACjF,8BAA0B,MAAM,mBAAmB,cAAc,mBAAmB,WAAW;AAC/F,0BAAsB,MAAM,mBAAmB,cAAc,eAAe,WAAW;AAAA,EACzF;AACA,MAAI,QAAQ,gBAAgB;AAC1B,YAAQ,eAAe,QAAQ,CAAC,GAAG,MAAM;AACvC,6BAAuB,KAAK,MAAM,mBAAmB,cAAc,EAAE,MAAM,WAAW;AAAA,IACxF,CAAC;AAAA,EACH;AACA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AACA,SAAS,yBAAyB,YAAY,QAAQ;AACpD,MAAI,WAAW,WAAW,OAAO,QAAQ;AACvC,UAAM,MAAM,4BAA4B,WAAW,wCAAwC,OAAO,eAAe;AAAA,EACnH;AACA,aAAW,QAAQ,CAAC,GAAG,MAAM;AAC3B,UAAM,SAAS,EAAE;AACjB,UAAM,SAAS,OAAO;AACtB,UAAM,SAAS,OAAO;AACtB,QAAI,CAAC,aAAa,YAAY,QAAQ,MAAM,GAAG;AAC7C,YAAM,MAAM,2EAA2E,cAAc,mBAAmB;AAAA,IAC1H;AACA,QAAI,EAAE,aAAa,OAAO,WAAW;AACnC;AAAA,IACF;AACA,UAAM,YAAY,EAAE;AACpB,UAAM,YAAY,OAAO,YAAY,OAAO,OAAO,QAAQ;AAC3D,QAAI,CAAC,aAAa,YAAY,WAAW,SAAS,GAAG;AACnD,YAAM,MAAM,kFAAkF,iBAAiB,sBAAsB;AAAA,IACvI;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,OAAO,QAAQ,QAAQ,QAAQ,qBAAqB;AACtE,MAAI,CAAC,OAAO,QAAQ,qBAAqB;AACvC,6BAAyB,OAAO,cAAc,MAAM;AACpD,6BAAyB,CAAC,OAAO,YAAY,GAAG,CAAC,MAAM,CAAC;AAAA,EAC1D;AACA,QAAM,SAAS,OAAO,QAAQ;AAC9B,QAAM,cAAc,OAAO,QAAQ;AACnC,MAAI,OAAO,QAAQ,UAAU;AAC3B,UAAM,6BAA6B,OAAO,SAAS,YAAY,IAAI,YAAY,EAAE;AAAA,EACnF,OAAO;AACL,UAAM,uBAAuB,OAAO,SAAS,YAAY,IAAI,YAAY,EAAE;AAAA,EAC7E;AACA,QAAM,WAAW,OAAO,YAAY;AACpC,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,QAAI,OAAO,WAAW,MAAM;AAC1B,YAAM,GAAG,UAAU,OAAO,QAAQ,QAAQ;AAAA,IAC5C;AAAA,EACF;AACA,MAAI,OAAO,WAAW,MAAM;AAC1B,UAAM,GAAG,UAAU,OAAO,QAAQ,GAAG;AAAA,EACvC;AACA,SAAO,QAAQ,CAAC,QAAQ,MAAM;AAC5B,UAAM,UAAU,OAAO,QAAQ,cAAc;AAC7C,UAAM,SAAS,OAAO,iBAAiB;AACvC,UAAM,eAAe,OAAO,iBAAiB,SAAS;AACtD,UAAM,cAAc,OAAO,kBAAkB,GAAG;AAChD,UAAM,iBAAiB,OAAO,qBAAqB,GAAG;AACtD,QAAI,aAAa;AACf,YAAM,EAAE,aAAa,IAAI,wBAAwB,OAAO,QAAQ,cAAc,OAAO,OAAO,OAAO,QAAQ,QAAQ;AACnH,cAAQ,aAAa,QAAQ;AAAA,QAC3B,KAAK;AACH,gBAAM,GAAG,WAAW,aAAa,IAAI,WAAW,YAAY,CAAC;AAC7D;AAAA,QACF,KAAK;AACH,gBAAM,GAAG,WAAW,aAAa,IAAI,WAAW,YAAY,CAAC;AAC7D;AAAA,QACF,KAAK;AACH,gBAAM,GAAG,WAAW,aAAa,IAAI,WAAW,YAAY,CAAC;AAC7D;AAAA,QACF,KAAK;AACH,gBAAM,GAAG,WAAW,aAAa,IAAI,WAAW,YAAY,CAAC;AAC7D;AAAA,QACF;AACE;AAAA,MACJ;AAAA,IACF;AACA,QAAI,gBAAgB;AAClB,YAAM,GAAG,UAAU,gBAAgB,OAAO,QAAQ,SAAS,IAAI,OAAO,QAAQ,SAAS,EAAE;AAAA,IAC3F;AACA,QAAI,UAAU,MAAM;AAClB;AAAA,IACF;AACA,QAAI,OAAO,WAAW;AACpB,UAAI,aAAa,cAAc,OAAO,KAAK,IAAI,GAAG;AAChD,cAAM,GAAG,UAAU,QAAQ,OAAO,cAAc,EAAE;AAAA,MACpD,OAAO;AACL,YAAI,OAAO,OAAO;AAClB,YAAI,EAAE,gBAAgB,eAAe;AACnC,iBAAO,IAAI,aAAa,IAAI;AAAA,QAC9B;AACA,cAAM,GAAG,WAAW,QAAQ,IAAI;AAAA,MAClC;AACA;AAAA,IACF;AACA,QAAI,OAAO,QAAQ,SAAS,QAAQ,gBAAgB,MAAM;AACxD,YAAM,GAAG,UAAU,cAAc,OAAO,QAAQ,MAAM,UAAU;AAAA,IAClE;AACA,UAAM,sBAAsB,OAAO,QAAQ,QAAQ,SAAS,QAAQ,CAAC;AAAA,EACvE,CAAC;AACD,QAAM,cAAc,OAAO;AAC3B,MAAI,aAAa;AACf,YAAQ,OAAO,MAAM,QAAQ;AAAA,MAC3B,KAAK;AACH,cAAM,GAAG,WAAW,aAAa,IAAI,WAAW,OAAO,KAAK,CAAC;AAC7D;AAAA,MACF,KAAK;AACH,cAAM,GAAG,WAAW,aAAa,IAAI,WAAW,OAAO,KAAK,CAAC;AAC7D;AAAA,MACF,KAAK;AACH,cAAM,GAAG,WAAW,aAAa,IAAI,WAAW,OAAO,KAAK,CAAC;AAC7D;AAAA,MACF,KAAK;AACH,cAAM,GAAG,WAAW,aAAa,IAAI,WAAW,OAAO,KAAK,CAAC;AAC7D;AAAA,MACF;AACE;AAAA,IACJ;AAAA,EACF;AACA,MAAI,OAAO,yBAAyB;AAClC,UAAM,UAAU,aAAa,eAAe,OAAO,KAAK;AACxD,YAAQ,OAAO,MAAM,QAAQ;AAAA,MAC3B,KAAK;AACH,cAAM,GAAG,WAAW,OAAO,yBAAyB,IAAI,WAAW,OAAO,CAAC;AAC3E;AAAA,MACF,KAAK;AACH,cAAM,GAAG,WAAW,OAAO,yBAAyB,IAAI,WAAW,OAAO,CAAC;AAC3E;AAAA,MACF,KAAK;AACH,cAAM,GAAG,WAAW,OAAO,yBAAyB,IAAI,WAAW,OAAO,CAAC;AAC3E;AAAA,MACF;AACE;AAAA,IACJ;AAAA,EACF;AACA,MAAI,OAAO,qBAAqB;AAC9B,UAAM,GAAG,UAAU,OAAO,qBAAqB,OAAO,QAAQ,SAAS,IAAI,OAAO,QAAQ,SAAS,EAAE;AAAA,EACvG;AACA,MAAI,OAAO,QAAQ,kBAAkB,qBAAqB;AACxD,WAAO,QAAQ,eAAe,QAAQ,CAAC,GAAG,MAAM;AAC9C,YAAM,YAAY,OAAO,uBAAuB;AAChD,YAAM,cAAc,oBAAoB;AACxC,UAAI,EAAE,SAAS,SAAS;AACtB,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,QAAQ;AAC5B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,QAAQ;AAC5B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,QAAQ;AAC5B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,OAAO;AAC3B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,SAAS;AAC7B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,SAAS;AAC7B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,WAAW,EAAE,SAAS,SAAS;AAC7B,cAAM,GAAG,WAAW,WAAW,WAAW;AAAA,MAC5C,OAAO;AACL,cAAM,MAAM,gBAAgB,EAAE,4BAA4B;AAAA,MAC5D;AAAA,IACF,CAAC;AAAA,EACH;AACA,QAAM,eAAe;AACvB;AACA,SAAS,cAAc,SAAS,QAAQ,QAAQ;AAC9C,MAAI,YAAY;AAChB,SAAO,OAAO,MAAM,EAAE,QAAQ,CAAC,MAAM;AACnC,UAAM,YAAY,EAAE,WAAW,QAAQ,EAAE,QAAQ,SAAS,QAAQ,EAAE,QAAQ,MAAM,aAAa;AAC/F,QAAI,QAAQ,uBAAuB,CAAC,EAAE,WAAW;AAC/C,YAAM,YAAY,EAAE,QAAQ;AAC5B,YAAM,EAAE,iBAAiB,cAAc,SAAS,IAAI,wBAAwB,QAAQ,cAAc,EAAE,OAAO,SAAS;AACpH,UAAI,QAAQ,IAAI,QAAQ,IAAI,SAAS;AACrC,UAAI,aAAa,WAAW,KAAK,QAAQ,cAAc;AACrD,cAAM,iBAAiB,CAAC,KAAK,KAAK,UAAU,KAAK,CAAC,GAAG,KAAK,KAAK,UAAU,KAAK,CAAC,CAAC;AAChF,gBAAQ,GAAG,eAAe,KAAK,KAAK,eAAe,KAAK;AAAA,MAC1D,WAAW,aAAa,WAAW,KAAK,CAAC,QAAQ,cAAc;AAC7D,gBAAQ,GAAG,aAAa,KAAK,KAAK,aAAa,KAAK;AAAA,MACtD,WAAW,aAAa,SAAS,KAAK,CAAC,QAAQ,cAAc;AAC3D,cAAM,UAAU,aAAa,eAAe,YAAY;AACxD,iBAAS,GAAG,QAAQ,OAAO,UAAU,MAAM,QAAQ,QAAQ,SAAS,OAAO,UAAU;AAAA,MACvF;AACA,YAAM,QAAQ,EAAE,MAAM;AACtB,YAAM,6BAA6B,aAAa,WAAW,KAAK,aAAa,YAAY,EAAE,OAAO,SAAS;AAC3G,YAAM,WAAW,aAAa,cAAc,EAAE,KAAK,MAAM;AACzD,YAAM,gBAAgB,qBAAqB,iBAAiB,EAAE,OAAO,OAAO,KAAK;AACjF,YAAM,uBAAuB,CAAC,QAAQ,gBAAgB,UAAU,OAAO,MAAM,UAAU,aAAa,YAAY,WAAW,OAAO,QAAQ,QAAQ;AAClJ,YAAM,2BAA2B,QAAQ,gBAAgB,aAAa,SAAS,IAAI,KAAK,GAAG,UAAU,KAAK,KAAK,UAAU,KAAK;AAC9H,mBAAa,GAAG,SAAS,wBAAwB,kBAAkB,WAAW,MAAM,aAAa,UAAU,YAAY,iBAAiB,8BAA8B,SAAS,SAAS,UAAU,4BAA4B;AAAA,IAChO,OAAO;AACL,YAAM,WAAW,EAAE,YAAY,YAAY,EAAE,QAAQ;AACrD,mBAAa,GAAG,EAAE,SAAS,YAAY;AAAA,IACzC;AAAA,EACF,CAAC;AACD,QAAM,cAAc,QAAQ;AAC5B,MAAI,MAAM,QAAQ,YAAY;AAC9B,SAAO,MAAM,YAAY,MAAM,cAAc,GAAG,IAAI,EAAE,UAAU,eAAe;AAC/E,SAAO;AACT;AACA,SAAS,iBAAiB,MAAM;AAC9B,SAAO,IAAI,EAAE,QAAQ,2BAA2B,KAAK,QAAQ;AAC/D;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,mBAAmB,cAAc;AACtC,SAAK,iBAAiB,CAAC,EAAE,MAAM,YAAY,MAAM,QAAQ,CAAC;AAC1D,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,SAAK,WAAW;AAAA;AAAA,UAEV,KAAK,sBAAsB,kDAAkD,CAAC,KAAK,KAAK,GAAG,GAAG,WAAW,IAAI,mCAAmC,CAAC,KAAK,KAAK,GAAG,GAAG,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgB5K,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,IAAI,4BAA4B,MAAM;AAAA,EACpC,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,mBAAmB,cAAc;AACtC,SAAK,iBAAiB,CAAC,EAAE,MAAM,YAAY,MAAM,QAAQ,CAAC;AAC1D,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,SAAK,WAAW;AAAA;AAAA,UAEV,KAAK,sBAAsB,kDAAkD,CAAC,KAAK,KAAK,GAAG,GAAG,WAAW,IAAI,mCAAmC,CAAC,KAAK,KAAK,GAAG,GAAG,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgB5K,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,IAAI,qBAAqB,MAAM;AAAA,EAC7B,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,aAAa;AAChC,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA;AAAA;AAAA,UAIE,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,IAAI,2BAA2B,MAAM;AAAA,EACnC,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,aAAa;AAChC,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA;AAAA;AAAA;AAAA,UAKE,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,IAAI,4BAA4B;AAAA,EAC9B,KAAK;AAAA,EACL,KAAK;AAAA,EACL,KAAK;AAAA,EACL,KAAK;AACP;AACA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,aAAa,sBAAsB,OAAO,eAAe,QAAQ;AAC3E,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,iBAAiB,CAAC,EAAE,MAAM,YAAY,MAAM,QAAQ,CAAC;AAC1D,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,QAAI,SAAS;AACb,QAAI,qBAAqB;AACvB,eAAS;AAAA,IACX;AACA,QAAI,WAAW;AACf,aAAS,mBAAmB,GAAG,mBAAmB,aAAa,QAAQ,oBAAoB;AACzF,YAAM,aAAa,aAAa;AAChC,kBAAY;AAAA,yBACO;AAAA,8BACK,0BAA0B;AAAA;AAAA,IAEpD;AACA,SAAK,WAAW;AAAA,QACZ,KAAK,sBAAsB,yBAAyB,IAAI,mBAAmB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uCAMvD,aAAa;AAAA;AAAA,sCAEd,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAMzB,KAAK;AAAA,YACnB;AAAA;AAAA,UAEF,KAAK,iBAAiB;AAAA;AAAA;AAAA,EAG9B;AACF;AAGA,IAAI,4BAA4B,MAAM;AAAA,EACpC,YAAY,aAAa,sBAAsB,OAAO;AACpD,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB,CAAC,EAAE,MAAM,YAAY,MAAM,QAAQ,CAAC;AAC1D,UAAM,OAAO,mBAAmB;AAChC,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,QAAI,WAAW;AACf,QAAI,SAAS;AACb,QAAI,qBAAqB;AACvB,eAAS;AAAA,IACX;AACA,aAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,eAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,cAAM,UAAU,MAAM,IAAI;AAC1B,oBAAY;AAAA;AAAA,gCAEY,SAAS,KAAK,sBAAsB,gBAAgB,GAAG,YAAY;AAAA,8BACrE;AAAA,iCACG,SAAS,KAAK,sBAAsB,gBAAgB,GAAG,YAAY;AAAA,gCACpE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uBAUT,KAAK;AAAA;AAAA;AAAA,uBAGL;AAAA;AAAA,uBAEA;AAAA;AAAA,uBAEA;AAAA;AAAA,uBAEA;AAAA;AAAA;AAAA;AAAA;AAAA,MAKjB;AAAA,IACF;AACA,SAAK,WAAW;AAAA,UACV,KAAK,sBAAsB,yBAAyB,IAAI,mBAAmB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAWpF;AAAA;AAAA,YAEA,KAAK,YAAY;AAAA;AAAA;AAAA,EAG3B;AACF;AAGA,IAAI,qBAAqB,CAAC;AAC1BA,UAAS,oBAAoB;AAAA,EAC3B,mCAAmC,MAAM;AAAA,EACzC,+BAA+B,MAAM;AAAA,EACrC,4BAA4B,MAAM;AAAA,EAClC,kCAAkC,MAAM;AAAA,EACxC,4BAA4B,MAAM;AAAA,EAClC,mBAAmB,MAAM;AAAA,EACzB,2BAA2B,MAAM;AAAA,EACjC,kCAAkC,MAAM;AAAA,EACxC,oBAAoB,MAAM;AAAA,EAC1B,oBAAoB,MAAM;AAAA,EAC1B,iDAAiD,MAAM;AAAA,EACvD,iCAAiC,MAAM;AAAA,EACvC,uCAAuC,MAAM;AAAA,EAC7C,gCAAgC,MAAM;AAAA,EACtC,0CAA0C,MAAM;AAAA,EAChD,gDAAgD,MAAM;AAAA,EACtD,0CAA0C,MAAM;AAAA,EAChD,yCAAyC,MAAM;AAAA,EAC/C,gDAAgD,MAAM;AAAA,EACtD,4BAA4B,MAAM;AAAA,EAClC,0BAA0B,MAAM;AAClC,CAAC;AACD,SAAS,oBAAoB,IAAI;AAC/B,QAAM,OAAO,mBAAmB;AAChC,QAAM,qBAAqB,GAAG,KAAK;AAAA;AAAA,MAE/B,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAMT,SAAO,mBAAmB,IAAI,kBAAkB;AAClD;AACA,SAAS,mBAAmB,IAAI;AAC9B,QAAM,cAAc,IAAI,aAAa,CAAC,IAAI,GAAG,GAAG,GAAG,GAAG,IAAI,IAAI,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,IAAI,GAAG,GAAG,CAAC,CAAC;AACrG,SAAO,yBAAyB,IAAI,WAAW;AACjD;AACA,SAAS,kBAAkB,IAAI;AAC7B,QAAM,wBAAwB,IAAI,YAAY,CAAC,GAAG,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AAChE,SAAO,wBAAwB,IAAI,qBAAqB;AAC1D;AACA,SAAS,0BAA0B,IAAI,OAAO,QAAQ,gBAAgB,eAAe,aAAa;AAChG,sBAAoB,OAAO,MAAM;AACjC,QAAM,UAAU,cAAc,EAAE;AAChC,QAAM,QAAQ,GAAG;AACjB,eAAa,IAAI,MAAM,GAAG,YAAY,OAAO,OAAO,CAAC;AACrD,eAAa,IAAI,MAAM,GAAG,cAAc,OAAO,GAAG,gBAAgB,GAAG,aAAa,CAAC;AACnF,eAAa,IAAI,MAAM,GAAG,cAAc,OAAO,GAAG,gBAAgB,GAAG,aAAa,CAAC;AACnF,eAAa,IAAI,MAAM,GAAG,cAAc,OAAO,GAAG,oBAAoB,GAAG,OAAO,CAAC;AACjF,eAAa,IAAI,MAAM,GAAG,cAAc,OAAO,GAAG,oBAAoB,GAAG,OAAO,CAAC;AACjF,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,iBAAa,IAAI,MAAM,GAAG,WAAW,OAAO,GAAG,gBAAgB,OAAO,QAAQ,GAAG,eAAe,aAAa,IAAI,CAAC;AAAA,EACpH,OAAO;AACL,iBAAa,IAAI,MAAM,GAAG,aAAa,OAAO,GAAG,gBAAgB,OAAO,MAAM,CAAC;AAAA,EACjF;AACA,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,IAAI,CAAC;AAC1D,SAAO,EAAE,SAAS,UAAU,CAAC,QAAQ,KAAK,EAAE;AAC9C;AACA,SAAS,yCAAyC,eAAe;AAC/D,SAAO,cAAc;AACvB;AACA,SAAS,2BAA2B,IAAI,MAAM,SAAS,eAAe;AACpE,QAAM,CAAC,OAAO,MAAM,IAAI,yCAAyC,MAAM,OAAO;AAC9E,SAAO,0BAA0B,IAAI,OAAO,QAAQ,yCAAyC,aAAa,GAAG,cAAc,oBAAoB,GAAG,KAAK;AACzJ;AACA,SAAS,yCAAyC,eAAe;AAC/D,SAAO,cAAc;AACvB;AACA,SAAS,2BAA2B,IAAI,MAAM,SAAS,eAAe;AACpE,QAAM,CAAC,OAAO,MAAM,IAAI,yCAAyC,MAAM,OAAO;AAC9E,SAAO,0BAA0B,IAAI,OAAO,QAAQ,yCAAyC,aAAa,GAAG,cAAc,oBAAoB,cAAc,oBAAoB;AACnL;AACA,SAAS,+CAA+C,eAAe;AACrE,SAAO,cAAc;AACvB;AACA,SAAS,iCAAiC,IAAI,MAAM,SAAS,eAAe;AAC1E,QAAM,CAAC,OAAO,MAAM,IAAI,yCAAyC,MAAM,OAAO;AAC9E,SAAO,0BAA0B,IAAI,OAAO,QAAQ,+CAA+C,aAAa,GAAG,GAAG,MAAM,GAAG,aAAa;AAC9I;AACA,SAAS,wCAAwC,eAAe;AAC9D,SAAO,cAAc;AACvB;AACA,SAAS,0BAA0B,IAAI,MAAM,SAAS,eAAe;AACnE,QAAM,CAAC,OAAO,MAAM,IAAI,uCAAuC,MAAM,OAAO;AAC5E,SAAO,0BAA0B,IAAI,OAAO,QAAQ,wCAAwC,aAAa,GAAG,GAAG,MAAM,GAAG,KAAK;AAC/H;AACA,SAAS,+CAA+C,eAAe;AACrE,SAAO,cAAc;AACvB;AACA,SAAS,iCAAiC,IAAI,MAAM,SAAS,eAAe;AAC1E,QAAM,CAAC,OAAO,MAAM,IAAI,uCAAuC,MAAM,OAAO;AAC5E,SAAO,0BAA0B,IAAI,OAAO,QAAQ,+CAA+C,aAAa,GAAG,GAAG,MAAM,cAAc,oBAAoB;AAChK;AACA,SAAS,kCAAkC,IAAI,SAAS,cAAc;AACpE,QAAM,YAAY;AAClB,QAAM,WAAW,IAAI;AACrB,QAAM,SAAS,IAAI,IAAI,IAAI;AAC3B,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,cAAc,YAAY,CAAC;AACnE,QAAM,UAAU,mCAAmC,IAAI,SAAS,gBAAgB,cAAc,GAAG,QAAQ,SAAS;AAClH,SAAO,WAAW,mCAAmC,IAAI,SAAS,MAAM,cAAc,GAAG,QAAQ,QAAQ;AAC3G;AACA,SAAS,2BAA2B,IAAI,SAAS,OAAO,QAAQ,MAAM,eAAe;AACnF,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,OAAO,CAAC;AAC7D,MAAI,eAAe,eAAe;AAClC,MAAI,gBAAgB,YAAY;AAC9B,oBAAgB,IAAI,WAAW,QAAQ,SAAS,CAAC;AACjD,oBAAgB,GAAG;AACnB,qBAAiB,GAAG;AAAA,EACtB,OAAO;AACL,oBAAgB,IAAI,aAAa,QAAQ,SAAS,CAAC;AACnD,oBAAgB,GAAG;AACnB,qBAAiB,cAAc;AAAA,EACjC;AACA,gBAAc,IAAI,IAAI;AACtB,MAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,iBAAa,IAAI,MAAM,GAAG,cAAc,GAAG,YAAY,GAAG,GAAG,GAAG,OAAO,QAAQ,GAAG,MAAM,eAAe,aAAa,CAAC;AAAA,EACvH,OAAO;AACL,iBAAa,IAAI,MAAM,GAAG,WAAW,GAAG,YAAY,GAAG,gBAAgB,OAAO,QAAQ,GAAG,GAAG,MAAM,eAAe,aAAa,CAAC;AAAA,EACjI;AACA,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,IAAI,CAAC;AAC5D;AACA,SAAS,yBAAyB,IAAI,SAAS,QAAQ;AACrD,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,OAAO,CAAC;AAC7D,MAAI,OAAO,gBAAgB,YAAY;AACrC,QAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,mBAAa,IAAI,MAAM,GAAG,cAAc,GAAG,YAAY,GAAG,GAAG,GAAG,OAAO,OAAO,OAAO,QAAQ,GAAG,MAAM,GAAG,eAAe,OAAO,IAAI,CAAC;AAAA,IACtI,OAAO;AACL,mBAAa,IAAI,MAAM,GAAG,WAAW,GAAG,YAAY,GAAG,GAAG,MAAM,OAAO,OAAO,OAAO,QAAQ,GAAG,GAAG,MAAM,GAAG,eAAe,OAAO,IAAI,CAAC;AAAA,IACzI;AAAA,EACF,OAAO;AACL,QAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,mBAAa,IAAI,MAAM,GAAG,cAAc,GAAG,YAAY,GAAG,GAAG,GAAG,GAAG,MAAM,GAAG,eAAe,MAAM,CAAC;AAAA,IACpG,OAAO;AACL,mBAAa,IAAI,MAAM,GAAG,WAAW,GAAG,YAAY,GAAG,GAAG,MAAM,GAAG,MAAM,GAAG,eAAe,MAAM,CAAC;AAAA,IACpG;AAAA,EACF;AACA,eAAa,IAAI,MAAM,GAAG,YAAY,GAAG,YAAY,IAAI,CAAC;AAC5D;AACA,SAAS,8BAA8B,KAAK,MAAM,SAAS,eAAe;AACxE,QAAM,UAAU,IAAI,aAAa;AACjC,eAAa,KAAK,MAAM,IAAI,WAAW,IAAI,mBAAmB,OAAO,CAAC;AACtE,QAAM,gBAAgB;AACtB,QAAM,iBAAiB;AACvB,QAAM,kBAAkB,gBAAgB,iBAAiB,OAAO;AAChE,eAAa,KAAK,MAAM,IAAI,WAAW,IAAI,mBAAmB,iBAAiB,IAAI,WAAW,CAAC;AAC/F,eAAa,KAAK,MAAM,IAAI,WAAW,GAAG,GAAG,SAAS,MAAM,IAAI,MAAM,IAAI,OAAO,CAAC,CAAC;AACnF,eAAa,KAAK,MAAM,IAAI,WAAW,IAAI,mBAAmB,IAAI,CAAC;AACnE,SAAO;AACT;AACA,SAAS,gCAAgC,IAAI,SAAS,MAAM;AAC1D,QAAM,MAAM;AACZ,QAAM,iBAAiB,IAAI,aAAa,IAAI;AAC5C,MAAI,WAAW,IAAI,mBAAmB,OAAO;AAC7C,MAAI,iBAAiB,IAAI,mBAAmB,GAAG,cAAc;AAC7D,MAAI,WAAW,IAAI,mBAAmB,IAAI;AAC1C,SAAO;AACT;AACA,SAAS,gDAAgD,IAAI,MAAM,SAAS,eAAe;AACzF,QAAM,CAAC,GAAG,CAAC,IAAI,yCAAyC,MAAM,OAAO;AACrE,QAAM,cAAc;AACpB,QAAM,iBAAiB,IAAI,WAAW,mCAAmC,OAAO,SAAS,WAAW,CAAC;AACrG,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,GAAG,GAAG,GAAG,cAAc,uBAAuB,GAAG,eAAe,cAAc,CAAC;AACvH,SAAO,IAAI,aAAa,eAAe,MAAM;AAC/C;AACA,SAAS,+BAA+B,IAAI,SAAS,OAAO,MAAM,MAAM,cAAc,cAAc,eAAe;AACjH,QAAM,MAAM;AACZ,QAAM,iBAAiB,IAAI,aAAa,sCAAsC,cAAc,YAAY,CAAC;AACzG,MAAI,WAAW,IAAI,mBAAmB,OAAO;AAC7C,MAAI,iBAAiB,IAAI,mBAAmB,GAAG,cAAc;AAC7D,MAAI,WAAW,IAAI,mBAAmB,IAAI;AAC1C,SAAO;AACT;AACA,SAAS,sCAAsC,IAAI,cAAc,cAAc;AAC7E,QAAM,aAAa,IAAI,aAAa,eAAe,eAAe,CAAC;AACnE,eAAa,IAAI,MAAM,GAAG,WAAW,GAAG,GAAG,cAAc,cAAc,GAAG,MAAM,GAAG,OAAO,UAAU,CAAC;AACrG,SAAO;AACT;AAGA,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,IAAI;AACd,SAAK,gBAAgB;AACrB,SAAK,UAAU;AACf,SAAK,WAAW;AAChB,SAAK,sBAAsB;AAC3B,SAAK,cAAc,CAAC;AACpB,UAAM,YAAY,IAAI,EAAE,UAAU,eAAe;AACjD,QAAI,MAAM,MAAM;AACd,WAAK,KAAK;AACV,sBAAgB,WAAW,EAAE;AAAA,IAC/B,OAAO;AACL,WAAK,KAAK,gBAAgB,SAAS;AAAA,IACrC;AACA,QAAI,qBAAqB;AACzB,UAAM,0BAA0B;AAChC,SAAK,+BAA+B,KAAK,GAAG,aAAa,6BAA6B;AACtF,QAAI,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC1C,YAAM,gBAAgB;AACtB,YAAM,qBAAqB;AAC3B,WAAK,wBAAwB,oBAAoB,KAAK,IAAI,aAAa;AACvE,UAAI,aAAa,KAAK,IAAI,kBAAkB,GAAG;AAC7C,aAAK,4BAA4B,oBAAoB,KAAK,IAAI,kBAAkB;AAAA,MAClF,WAAW,IAAI,EAAE,IAAI,0BAA0B,GAAG;AAChD,cAAM,IAAI,MAAM,oHAAoH;AAAA,MACtI;AACA,WAAK,4BAA4B,KAAK,GAAG,aAAa,kBAAkB;AACxE,UAAI,aAAa,KAAK,IAAI,uBAAuB,GAAG;AAClD,aAAK,gCAAgC,oBAAoB,KAAK,IAAI,uBAAuB;AAAA,MAC3F,WAAW,IAAI,EAAE,IAAI,0BAA0B,GAAG;AAChD,cAAM,IAAI,MAAM,6HAA6H;AAAA,MAC/I;AAAA,IACF,OAAO;AACL,2BAAqB;AACrB,UAAI,aAAa,KAAK,IAAI,kBAAkB,GAAG;AAC7C,aAAK,4BAA4B,KAAK,GAAG,aAAa,kBAAkB;AAAA,MAC1E,WAAW,aAAa,KAAK,IAAI,uBAAuB,GAAG;AACzD,aAAK,gCAAgC,KAAK,GAAG,aAAa,uBAAuB;AAAA,MACnF,OAAO;AACL,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AAAA,IACF;AACA,SAAK,eAAe,mBAAmB,KAAK,EAAE;AAC9C,SAAK,cAAc,kBAAkB,KAAK,EAAE;AAC5C,SAAK,cAAc,kBAAkB,KAAK,EAAE;AAC5C,SAAK,gBAAgB,iBAAiB,KAAK,IAAI,KAAK,yBAAyB;AAAA,EAC/E;AAAA,EACA,IAAI,QAAQ;AACV,WAAO,IAAI,EAAE,QAAQ,OAAO;AAAA,EAC9B;AAAA,EACA,UAAU;AACR,QAAI,KAAK,UAAU;AACjB;AAAA,IACF;AACA,QAAI,KAAK,WAAW,MAAM;AACxB,cAAQ,KAAK,sKAAsK;AAAA,IACrL;AACA,QAAI,KAAK,iBAAiB,MAAM;AAC9B,cAAQ,KAAK,oMAAoM;AAAA,IACnN;AACA,UAAM,KAAK,KAAK;AAChB,iBAAa,IAAI,MAAM,GAAG,OAAO,CAAC;AAClC,iBAAa,IAAI,MAAM,GAAG,gBAAgB,GAAG,aAAa,IAAI,CAAC;AAC/D,iBAAa,IAAI,MAAM,GAAG,kBAAkB,KAAK,WAAW,CAAC;AAC7D,iBAAa,IAAI,MAAM,GAAG,WAAW,GAAG,cAAc,IAAI,CAAC;AAC3D,iBAAa,IAAI,MAAM,GAAG,WAAW,GAAG,sBAAsB,IAAI,CAAC;AACnE,iBAAa,IAAI,MAAM,GAAG,aAAa,KAAK,WAAW,CAAC;AACxD,SAAK,WAAW;AAAA,EAClB;AAAA,EACA,2BAA2B,MAAM,SAAS;AACxC,SAAK,gBAAgB;AACrB,WAAO,2BAA2B,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AAAA,EAC9E;AAAA,EACA,2BAA2B,MAAM,SAAS;AACxC,SAAK,gBAAgB;AACrB,WAAO,2BAA2B,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AAAA,EAC9E;AAAA,EACA,iCAAiC,MAAM,SAAS;AAC9C,SAAK,gBAAgB;AACrB,WAAO,iCAAiC,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AAAA,EACpF;AAAA,EACA,yBAAyB,SAAS,QAAQ;AACxC,SAAK,gBAAgB;AACrB,6BAAyB,KAAK,IAAI,SAAS,MAAM;AAAA,EACnD;AAAA,EACA,2BAA2B,SAAS,OAAO,QAAQ,MAAM;AACvD,SAAK,gBAAgB;AACrB,+BAA2B,KAAK,IAAI,SAAS,OAAO,QAAQ,MAAM,KAAK,aAAa;AAAA,EACtF;AAAA,EACA,iCAAiC,MAAM,SAAS;AAC9C,SAAK,gBAAgB;AACrB,WAAO,iCAAiC,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AAAA,EACpF;AAAA,EACA,0BAA0B,MAAM,SAAS;AACvC,SAAK,gBAAgB;AACrB,WAAO,0BAA0B,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AAAA,EAC7E;AAAA,EACA,oBAAoB,SAAS;AAC3B,SAAK,gBAAgB;AACrB,QAAI,KAAK,kBAAkB,SAAS;AAClC,wCAAkC,KAAK,IAAI,KAAK,WAAW;AAC3D,WAAK,gBAAgB;AAAA,IACvB;AACA,iBAAa,KAAK,IAAI,MAAM,KAAK,GAAG,cAAc,OAAO,CAAC;AAAA,EAC5D;AAAA,EACA,gDAAgD,SAAS,MAAM,SAAS;AACtE,WAAO,KAAK,qBAAqB,SAAS,MAAM,gDAAgD,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa,CAAC;AAAA,EAC7I;AAAA,EACA,+BAA+B,SAAS,OAAO,MAAM,SAAS,cAAc,cAAc;AACxF,WAAO,+BAA+B,KAAK,IAAI,SAAS,OAAO,MAAM,SAAS,cAAc,cAAc,KAAK,aAAa;AAAA,EAC9H;AAAA,EACA,gCAAgC,SAAS,MAAM;AAC7C,WAAO,gCAAgC,KAAK,IAAI,SAAS,IAAI;AAAA,EAC/D;AAAA,EACA,wBAAwB,SAAS,MAAM,SAAS;AAC9C,SAAK,yBAAyB,OAAO;AACrC,UAAM,SAAS,8BAA8B,KAAK,IAAI,MAAM,SAAS,KAAK,aAAa;AACvF,SAAK,2BAA2B;AAChC,WAAO;AAAA,EACT;AAAA,EACA,wBAAwB;AACtB,UAAM,eAAe,KAAK,YAAY,KAAK,EAAE;AAC7C,WAAO,KAAK,UAAU,YAAY;AAAA,EACpC;AAAA,EACA,YAAY,IAAI;AACd,QAAI;AACJ,QAAI;AACJ,QAAI,IAAI,EAAE,QAAQ,yBAAyB,GAAG;AAC5C,YAAM,MAAM;AACZ,YAAM,OAAO,IAAI,UAAU,IAAI,4BAA4B,CAAC;AAC5D,SAAG,MAAM;AACT,sBAAgB,MAAM;AACpB,cAAM,SAAS,IAAI,eAAe,MAAM,GAAG,CAAC;AAC5C,eAAO,WAAW,IAAI,oBAAoB,WAAW,IAAI;AAAA,MAC3D;AACA,cAAQ;AAAA,IACV,WAAW,IAAI,EAAE,UAAU,8CAA8C,IAAI,GAAG;AAC9E,cAAQ,KAAK,WAAW;AACxB,WAAK,SAAS;AACd,sBAAgB,MAAM,KAAK,iBAAiB,OAAO,IAAI,EAAE,UAAU,8CAA8C,CAAC;AAAA,IACpH,OAAO;AACL,sBAAgB,MAAM;AAAA,IACxB;AACA,WAAO,EAAE,OAAO,cAAc;AAAA,EAChC;AAAA,EACA,gCAAgC,SAAS,cAAc,cAAc;AACnE,WAAO,KAAK,qBAAqB,SAAS,MAAM,sCAAsC,KAAK,IAAI,cAAc,YAAY,CAAC;AAAA,EAC5H;AAAA,EACA,cAAc,gBAAgB;AAC5B,SAAK,gBAAgB;AACrB,UAAM,KAAK,KAAK;AAChB,QAAI,KAAK,gBAAgB,MAAM;AAC7B,WAAK,eAAe,oBAAoB,EAAE;AAAA,IAC5C;AACA,UAAM,UAAU,cAAc,EAAE;AAChC,iBAAa,IAAI,MAAM,GAAG,aAAa,SAAS,KAAK,YAAY,CAAC;AAClE,iBAAa,IAAI,MAAM,GAAG,aAAa,SAAS,cAAc,CAAC;AAC/D,gBAAY,IAAI,OAAO;AACvB,QAAI,KAAK,OAAO;AACd,sBAAgB,IAAI,OAAO;AAAA,IAC7B;AACA,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,WAAW,OAAO;AACvB,WAAK,sBAAsB,kCAAkC,IAAI,KAAK,SAAS,KAAK,YAAY;AAAA,IAClG;AACA,WAAO;AAAA,EACT;AAAA,EACA,cAAc,SAAS;AACrB,SAAK,gBAAgB;AACrB,QAAI,YAAY,KAAK,SAAS;AAC5B,WAAK,UAAU;AAAA,IACjB;AACA,QAAI,WAAW,MAAM;AACnB,mBAAa,KAAK,IAAI,MAAM,KAAK,GAAG,cAAc,OAAO,CAAC;AAAA,IAC5D;AAAA,EACF;AAAA,EACA,WAAW,SAAS;AAClB,SAAK,gBAAgB;AACrB,SAAK,UAAU;AACf,QAAI,KAAK,WAAW,QAAQ,KAAK,OAAO;AACtC,sBAAgB,KAAK,IAAI,KAAK,OAAO;AAAA,IACvC;AACA,iBAAa,KAAK,IAAI,MAAM,KAAK,GAAG,WAAW,OAAO,CAAC;AAAA,EACzD;AAAA,EACA,mBAAmB,SAAS,aAAa,cAAc,MAAM;AAC3D,SAAK,gBAAgB;AACrB,QAAI,aAAa;AACf,aAAO,iCAAiC,KAAK,IAAI,SAAS,WAAW;AAAA,IACvE,OAAO;AACL,aAAO,0BAA0B,KAAK,IAAI,SAAS,WAAW;AAAA,IAChE;AAAA,EACF;AAAA,EACA,qBAAqB,SAAS,WAAW;AACvC,SAAK,gBAAgB;AACrB,WAAO,aAAa,KAAK,IAAI,MAAM,KAAK,GAAG,kBAAkB,SAAS,SAAS,CAAC;AAAA,EAClF;AAAA,EACA,0BAA0B,SAAS,aAAa;AAC9C,SAAK,gBAAgB;AACrB,WAAO,KAAK,GAAG,mBAAmB,SAAS,WAAW;AAAA,EACxD;AAAA,EACA,sBAAsB,oBAAoB,iBAAiB,aAAa;AACtE,SAAK,gBAAgB;AACrB,SAAK,iBAAiB;AACtB,uCAAmC,KAAK,IAAI,oBAAoB,iBAAiB,WAAW;AAAA,EAC9F;AAAA,EACA,uBAAuB,qBAAqB,MAAM,SAAS;AACzD,SAAK,6BAA6B,qBAAqB,SAAS,IAAI;AAAA,EACtE;AAAA,EACA,6BAA6B,2BAA2B,MAAM,SAAS;AACrE,SAAK,gBAAgB;AACrB,UAAM,CAAC,OAAO,MAAM,IAAI,uCAAuC,MAAM,OAAO;AAC5E,SAAK,6BAA6B,2BAA2B,OAAO,MAAM;AAAA,EAC5E;AAAA,EACA,2BAA2B,UAAU,SAAS,aAAa,YAAY;AACrE,SAAK,iCAAiC,aAAa,UAAU,YAAY,OAAO;AAAA,EAClF;AAAA,EACA,iCAAiC,UAAU,SAAS,aAAa,YAAY;AAC3E,UAAM,IAAI,MAAM,mDAAmD;AAAA,EACrE;AAAA,EACA,gBAAgB;AACd,QAAI,KAAK,WAAW,MAAM;AACxB,sBAAgB,KAAK,IAAI,KAAK,OAAO;AAAA,IACvC;AACA,wBAAoB,KAAK,EAAE;AAAA,EAC7B;AAAA,EACA,iBAAiB;AACf,SAAK,gBAAgB;AACrB,SAAK,iBAAiB;AACtB,UAAM,KAAK,KAAK;AAChB,QAAI,KAAK,OAAO;AACd,WAAK,cAAc;AAAA,IACrB;AACA,iBAAa,IAAI,MAAM,GAAG,aAAa,GAAG,WAAW,GAAG,GAAG,gBAAgB,CAAC,CAAC;AAAA,EAC/E;AAAA,EACA,iCAAiC;AAC/B,SAAK,gBAAgB;AACrB,iBAAa,KAAK,IAAI,MAAM,KAAK,GAAG,OAAO,CAAC;AAAA,EAC9C;AAAA,EACA,yBAAyB;AACvB,QAAI,KAAK,+BAA+B,MAAM;AAC5C,WAAK,8BAA8B,oBAAoB,KAAK,IAAI,IAAI,EAAE,UAAU,8CAA8C,MAAM,IAAI,oCAAoC,0BAA0B;AAAA,IACxM;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EACA,+BAA+B;AAC7B,WAAO,KAAK,uBAAuB;AAAA,EACrC;AAAA,EACA,+BAA+B;AAC7B,WAAO,KAAK,uBAAuB;AAAA,EACrC;AAAA,EACA,aAAa;AACX,QAAI,IAAI,EAAE,UAAU,8CAA8C,MAAM,GAAG;AACzE,YAAM,MAAM,KAAK;AACjB,YAAM,OAAO,KAAK,6BAA6B;AAC/C,YAAM,SAAS,IAAI,YAAY;AAC/B,UAAI,WAAW,KAAK,kBAAkB,MAAM;AAC5C,aAAO;AAAA,IACT;AACA,UAAM,MAAM,KAAK,6BAA6B;AAC9C,UAAM,QAAQ,IAAI,eAAe;AACjC,QAAI,cAAc,IAAI,kBAAkB,KAAK;AAC7C,WAAO;AAAA,EACT;AAAA,EACA,WAAW;AACT,QAAI,IAAI,EAAE,UAAU,8CAA8C,MAAM,GAAG;AACzE,YAAM,MAAM,KAAK;AACjB,YAAM,OAAO,KAAK,6BAA6B;AAC/C,UAAI,SAAS,KAAK,gBAAgB;AAClC;AAAA,IACF;AACA,UAAM,MAAM,KAAK,6BAA6B;AAC9C,QAAI,YAAY,IAAI,gBAAgB;AAAA,EACtC;AAAA,EACA,MAAM,uBAAuB,OAAO;AAClC,UAAM,aAAa,YAAY,MAAM,KAAK,YAAY,KAAK,iBAAiB,OAAO,IAAI,EAAE,UAAU,8CAA8C,CAAC,CAAC;AACnJ,WAAO,KAAK,aAAa,OAAO,IAAI,EAAE,UAAU,8CAA8C,CAAC;AAAA,EACjG;AAAA,EACA,aAAa,OAAO,mBAAmB;AACrC,QAAI,sBAAsB,GAAG;AAC3B,aAAO;AAAA,IACT;AACA,QAAI,sBAAsB,GAAG;AAC3B,YAAM,MAAM,KAAK;AACjB,YAAM,mBAAmB,IAAI,kBAAkB,OAAO,IAAI,YAAY;AACtE,aAAO,mBAAmB;AAAA,IAC5B,OAAO;AACL,YAAM,MAAM,KAAK,6BAA6B;AAC9C,YAAM,mBAAmB,IAAI,kBAAkB,OAAO,IAAI,gBAAgB;AAC1E,aAAO,mBAAmB;AAAA,IAC5B;AAAA,EACF;AAAA,EACA,iBAAiB,OAAO,mBAAmB;AACzC,QAAI,sBAAsB,GAAG;AAC3B,aAAO;AAAA,IACT;AACA,QAAI,sBAAsB,GAAG;AAC3B,YAAM,MAAM,KAAK;AACjB,YAAM,MAAM,KAAK,6BAA6B;AAC9C,YAAM,YAAY,IAAI,kBAAkB,OAAO,IAAI,sBAAsB;AACzE,UAAI,KAAK,YAAY,MAAM;AACzB,aAAK,WAAW,KAAK,GAAG,aAAa,IAAI,gBAAgB;AAAA,MAC3D;AACA,aAAO,aAAa,CAAC,KAAK;AAAA,IAC5B,OAAO;AACL,YAAM,MAAM,KAAK,6BAA6B;AAC9C,YAAM,YAAY,IAAI,kBAAkB,OAAO,IAAI,0BAA0B;AAC7E,UAAI,KAAK,YAAY,MAAM;AACzB,aAAK,WAAW,KAAK,GAAG,aAAa,IAAI,gBAAgB;AAAA,MAC3D;AACA,aAAO,aAAa,CAAC,KAAK;AAAA,IAC5B;AAAA,EACF;AAAA,EACA,UAAU,cAAc;AACtB,WAAO,IAAI,QAAQ,CAAC,YAAY;AAC9B,WAAK,cAAc,MAAM,aAAa,cAAc,GAAG,MAAM,QAAQ,CAAC;AAAA,IACxE,CAAC;AAAA,EACH;AAAA,EACA,YAAY;AACV,UAAM,QAAQ,qBAAqB,KAAK,YAAY,IAAI,CAAC,MAAM,EAAE,QAAQ,CAAC;AAC1E,aAAS,IAAI,GAAG,KAAK,OAAO,EAAE,GAAG;AAC/B,YAAM,EAAE,UAAU,IAAI,KAAK,YAAY;AACvC,gBAAU;AAAA,IACZ;AACA,SAAK,cAAc,KAAK,YAAY,MAAM,QAAQ,CAAC;AAAA,EACrD;AAAA,EACA,cAAc,UAAU,WAAW;AACjC,SAAK,YAAY,KAAK,EAAE,UAAU,UAAU,CAAC;AAC7C,QAAI,KAAK,YAAY,SAAS,GAAG;AAC/B;AAAA,IACF;AACA,QAAI,aAAa;AACjB,QAAI,sBAAsB,IAAI,EAAE,UAAU;AACxC,mBAAa,IAAI,EAAE,SAAS,iBAAiB,KAAK,IAAI,EAAE,QAAQ;AAAA,IAClE;AACA,iBAAa,YAAY,MAAM;AAC7B,WAAK,UAAU;AACf,aAAO,KAAK,YAAY,WAAW;AAAA,IACrC,GAAG,MAAM,GAAG,MAAM,UAAU;AAAA,EAC9B;AAAA,EACA,yBAAyB,SAAS;AAChC,SAAK,gBAAgB;AACrB,kCAA8B,KAAK,IAAI,SAAS,KAAK,WAAW;AAChE,QAAI,KAAK,OAAO;AACd,0BAAoB,KAAK,EAAE;AAAA,IAC7B;AAAA,EACF;AAAA,EACA,6BAA6B;AAC3B,QAAI,KAAK,iBAAiB,MAAM;AAC9B,oCAA8B,KAAK,IAAI,KAAK,eAAe,KAAK,WAAW;AAC3E,UAAI,KAAK,OAAO;AACd,4BAAoB,KAAK,EAAE;AAAA,MAC7B;AAAA,IACF,OAAO;AACL,wCAAkC,KAAK,IAAI,KAAK,WAAW;AAAA,IAC7D;AAAA,EACF;AAAA,EACA,qBAAqB,SAAS,mBAAmB;AAC/C,SAAK,yBAAyB,OAAO;AACrC,UAAM,SAAS,kBAAkB;AACjC,SAAK,2BAA2B;AAChC,WAAO;AAAA,EACT;AAAA,EACA,6BAA6B,gCAAgC,OAAO,QAAQ;AAC1E,SAAK,gBAAgB;AACrB,UAAM,KAAK,KAAK;AAChB,kCAA8B,IAAI,gCAAgC,KAAK,WAAW;AAClF,QAAI,KAAK,OAAO;AACd,0BAAoB,EAAE;AAAA,IACxB;AACA,SAAK,gBAAgB;AACrB,iBAAa,IAAI,MAAM,GAAG,SAAS,GAAG,GAAG,OAAO,MAAM,CAAC;AACvD,iBAAa,IAAI,MAAM,GAAG,QAAQ,GAAG,GAAG,OAAO,MAAM,CAAC;AAAA,EACxD;AAAA,EACA,iCAAiC,GAAG,GAAG,OAAO,QAAQ;AACpD,SAAK,gBAAgB;AACrB,iBAAa,KAAK,IAAI,MAAM,KAAK,GAAG,QAAQ,GAAG,GAAG,OAAO,MAAM,CAAC;AAAA,EAClE;AAAA,EACA,kBAAkB;AAChB,QAAI,KAAK,UAAU;AACjB,YAAM,IAAI,MAAM,yCAAyC;AAAA,IAC3D;AAAA,EACF;AAAA,EACA,mBAAmB;AACjB,QAAI,KAAK,WAAW,MAAM;AACxB,YAAM,IAAI,MAAM,kCAAkC;AAAA,IACpD;AAAA,EACF;AACF;AACA,SAAS,qBAAqB,KAAK;AACjC,MAAI,IAAI;AACR,SAAO,IAAI,IAAI,QAAQ,EAAE,GAAG;AAC1B,UAAM,SAAS,IAAI,GAAG;AACtB,QAAI,CAAC,QAAQ;AACX;AAAA,IACF;AAAA,EACF;AACA,SAAO,IAAI;AACb;AAGA,IAAI,EAAE,SAAS,YAAY,cAAc,iBAAiB,oBAAoB,uBAAuB,UAAU,aAAa,UAAU,aAAa,YAAY,eAAe,WAAW,cAAc,SAAS,YAAY,WAAW,cAAc,WAAW,cAAc,cAAc,iBAAiB,cAAc,iBAAiB,aAAa,gBAAgB,kBAAkB,qBAAqB,UAAU,aAAa,eAAe,kBAAkB,cAAc,iBAAiB,SAAS,YAAY,SAAS,YAAY,aAAa,gBAAgB,aAAa,gBAAgB,cAAc,iBAAiB,SAAS,YAAY,cAAc,iBAAiB,UAAU,aAAa,kBAAkB,qBAAqB,iBAAiB,oBAAoB,0BAA0B,6BAA6B,WAAW,cAAc,WAAW,cAAc,aAAa,gBAAgB,aAAa,gBAAgB,eAAe,kBAAkB,WAAW,cAAc,yBAAyB,4BAA4B,mBAAmB,sBAAsB,4BAA4B,+BAA+B,UAAU,aAAa,kBAAkB,qBAAqB,kBAAkB,qBAAqB,iBAAiB,oBAAoB,4BAA4B,+BAA+B,SAAS,YAAY,UAAU,aAAa,UAAU,aAAa,eAAe,kBAAkB,YAAY,cAAc,IAAI;AAG97C,SAAS,eAAe,MAAM,MAAM;AAClC,SAAO,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG,EAAE,MAAM,GAAG,IAAI,EAAE,IAAI,CAAC,MAAM,GAAG,QAAQ,GAAG;AAChF;AACA,SAAS,YAAY,MAAM,MAAM;AAC/B,MAAI,SAAS,GAAG;AACd,WAAO,CAAC,IAAI;AAAA,EACd;AACA,SAAO,eAAe,MAAM,IAAI;AAClC;AACA,SAAS,gBAAgB,MAAM,MAAM;AACnC,MAAI,SAAS,GAAG;AACd,WAAO;AAAA,EACT;AACA,MAAI,UAAU;AACd,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,eAAW,KAAK;AAChB,QAAI,IAAI,OAAO,GAAG;AAChB,iBAAW;AAAA,IACb;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,OAAO,YAAY;AACxB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,QAAI,KAAK,SAAS,GAAG;AACnB,WAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,IAKlB,OAAO;AACL,YAAM,WAAW,YAAY,MAAM,KAAK,IAAI;AAC5C,YAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,YAAM,uBAAuB,KAAK,wBAAwB,QAAQ;AAClE,YAAM,UAAU,KAAK,SAAS,QAAQ;AACtC,YAAM,SAAS,KAAK,UAAU,QAAQ;AACtC,WAAK,WAAW;AAAA;AAAA,YAEV;AAAA;AAAA,eAEG;AAAA;AAAA;AAAA,cAGD;AAAA;AAAA,6BAEe;AAAA;AAAA;AAAA;AAAA,IAIzB;AAAA,EACF;AAAA,EACA,mBAAmB,MAAM;AACvB,UAAM,UAAU,CAAC;AACjB,aAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,eAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,YAAI,QAAQ,GAAG,QAAQ,IAAI,MAAM,UAAU,QAAQ,IAAI,MAAM;AAC7D,iBAAS,IAAI,GAAG,IAAI,KAAK,MAAM,KAAK;AAClC,kBAAQ,GAAG,KAAK,KAAK,SAAS,IAAI,QAAQ;AAAA,QAC5C;AACA,gBAAQ,KAAK,KAAK;AAAA,MACpB;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,wBAAwB,MAAM;AAC5B,QAAI,KAAK,SAAS,GAAG;AACnB,aAAO,QAAQ,KAAK,sBAAsB,aAAa,KAAK,YAAY;AAAA,IAC1E;AACA,QAAI,OAAO;AACX,aAAS,IAAI,KAAK,OAAO,GAAG,IAAI,KAAK,MAAM,KAAK;AAC9C,cAAQ,GAAG,KAAK,SAAS,KAAK,sBAAsB,YAAY,OAAO,KAAK,YAAY;AACxF,UAAI,IAAI,KAAK,OAAO,GAAG;AACrB,gBAAQ;AAAA,MACV;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EACA,SAAS,MAAM;AACb,QAAI,KAAK,SAAS,GAAG;AACnB,aAAO;AAAA,IACT;AACA,UAAM,YAAY,KAAK,MAAM,EAAE;AAC/B,UAAM,MAAM,KAAK,sBAAsB,YAAY,KAAK,cAAc,KAAK,YAAY,KAAK,OAAO;AACnG,UAAM,MAAM,KAAK,sBAAsB,YAAY,KAAK,cAAc,KAAK,YAAY,KAAK,OAAO;AACnG,WAAO;AAAA,gBACK,UAAU;AAAA,gBACV,UAAU;AAAA;AAAA;AAAA;AAAA,4BAIE;AAAA,4BACA;AAAA;AAAA,EAE1B;AAAA,EACA,UAAU,MAAM;AACd,UAAM,eAAe,KAAK,mBAAmB,IAAI;AACjD,QAAI,KAAK,SAAS,GAAG;AACnB,YAAM,WAAW,KAAK,sBAAsB,aAAa,KAAK,YAAY;AAC1E,aAAO,wBAAwB;AAAA,IACjC;AACA,WAAO,QAAQ,aAAa;AAAA,gCACA,aAAa;AAAA,gCACb,aAAa;AAAA,yCACJ,aAAa;AAAA,EACpD;AACF;AAGA,IAAI,uBAAuB,MAAM;AAAA,EAC/B,YAAY,aAAa,YAAY;AACnC,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB,CAAC,EAAE,MAAM,cAAc,MAAM,QAAQ,CAAC;AAC5D,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,UAAI,SAAS;AACb,UAAI,IAAI,MAAM,GAAG;AACf,kBAAU;AAAA,MACZ;AACA,UAAI,IAAI,GAAG;AACT,kBAAU;AAAA,MACZ;AACA,kBAAY;AAAA,UACR;AAAA,UACA,IAAI,IAAI,4CAA4C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mBAM3C;AAAA;AAAA,UAET,IAAI,IAAI,MAAM;AAAA;AAAA,IAEpB;AACA,SAAK,WAAW;AAAA,QACZ,uBAAuB,YAAY,KAAK,mBAAmB;AAAA,QAC3D,KAAK,sBAAsB,yBAAyB,IAAI,mBAAmB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,qBAQzE,KAAK,sBAAsB,gBAAgB,YAAY;AAAA,qBACvD,KAAK,sBAAsB,gBAAgB,YAAY;AAAA;AAAA,UAElE;AAAA;AAAA;AAAA;AAAA;AAAA,EAKR;AACF;AACA,SAAS,uBAAuB,OAAO,qBAAqB;AAC1D,QAAM,yBAAyB,sBAAsB,4CAA4C,CAAC,KAAK,KAAK,GAAG,GAAG,YAAY,IAAI,mCAAmC,CAAC,KAAK,KAAK,GAAG,GAAG,KAAK;AAC3L,SAAO;AAAA;AAAA,QAED;AAAA;AAAA;AAAA;AAIR;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,OAAO;AACjB,SAAK,QAAQ;AACb,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AACvB,SAAK,qBAAqB;AAC1B,SAAK,gBAAgB;AACrB,SAAK,eAAe,CAAC;AACrB,SAAK,aAAa;AAClB,SAAK,eAAe,CAAC;AAAA,EACvB;AAAA,EACA,eAAe,SAAS,OAAO,UAAU;AACvC,UAAM,kBAAkB,kCAAkC,OAAO,QAAQ;AACzE,UAAM,WAAW,uBAAuB,SAAS,iBAAiB,QAAQ;AAC1E,QAAI,EAAE,YAAY,KAAK,eAAe;AACpC,WAAK,aAAa,YAAY,CAAC;AAAA,IACjC;AACA,QAAI,EAAE,YAAY,KAAK,eAAe;AACpC,WAAK,aAAa,YAAY,CAAC;AAAA,IACjC;AACA,UAAM,WAAW,aAAa,SAAS,iBAAiB,KAAK,MAAM,IAAI,KAAK,MAAM,eAAe,QAAQ;AACzG,QAAI,KAAK,aAAa,UAAU,SAAS,GAAG;AAC1C,WAAK;AACL,WAAK;AACL,WAAK,iBAAiB;AACtB,WAAK,IAAI;AACT,YAAM,cAAc,KAAK,aAAa,UAAU,MAAM;AACtD,WAAK,aAAa,UAAU,KAAK,WAAW;AAC5C,aAAO;AAAA,IACT;AACA,QAAI;AACJ,QAAI,oBAAoB,oBAAoB,oBAAoB;AAC9D,mBAAa,KAAK,MAAM,0BAA0B,QAAQ,IAAI,QAAQ,EAAE;AAAA,IAC1E,WAAW,oBAAoB,oBAAoB,oBAAoB;AACrE,mBAAa,KAAK,MAAM,iCAAiC,QAAQ,IAAI,QAAQ,EAAE;AAAA,IACjF,WAAW,oBAAoB,oBAAoB,kBAAkB;AACnE,mBAAa,KAAK,MAAM,2BAA2B,QAAQ,IAAI,QAAQ,EAAE;AAAA,IAC3E,WAAW,oBAAoB,oBAAoB,kBAAkB;AACnE,mBAAa,KAAK,MAAM,2BAA2B,QAAQ,IAAI,QAAQ,EAAE;AAAA,IAC3E,WAAW,oBAAoB,oBAAoB,0BAA0B;AAC3E,mBAAa,KAAK,MAAM,iCAAiC,QAAQ,IAAI,QAAQ,EAAE;AAAA,IACjF;AACA,SAAK,aAAa,UAAU,KAAK,UAAU;AAC3C,SAAK;AACL,SAAK,sBAAsB;AAC3B,SAAK,IAAI;AACT,WAAO;AAAA,EACT;AAAA,EACA,eAAe,SAAS,OAAO,gBAAgB,UAAU;AACvD,QAAI,KAAK,gBAAgB,MAAM;AAC7B;AAAA,IACF;AACA,UAAM,kBAAkB,kCAAkC,gBAAgB,QAAQ;AAClF,UAAM,WAAW,uBAAuB,OAAO,iBAAiB,QAAQ;AACxE,QAAI,EAAE,YAAY,KAAK,eAAe;AACpC,WAAK,aAAa,YAAY,CAAC;AAAA,IACjC;AACA,UAAM,WAAW,aAAa,OAAO,iBAAiB,KAAK,MAAM,IAAI,KAAK,MAAM,eAAe,QAAQ;AACvG,UAAM,qBAAqB,IAAI,EAAE,IAAI,gCAAgC;AACrE,QAAI,uBAAuB,MAAM,KAAK,qBAAqB,oBAAoB;AAC7E,WAAK,MAAM,oBAAoB,QAAQ,OAAO;AAC9C,WAAK,sBAAsB;AAAA,IAC7B,OAAO;AACL,WAAK,aAAa,UAAU,KAAK,OAAO;AACxC,WAAK;AACL,WAAK,iBAAiB;AAAA,IACxB;AACA,SAAK;AACL,UAAM,UAAU,KAAK,aAAa;AAClC,UAAM,WAAW,QAAQ,QAAQ,OAAO;AACxC,QAAI,WAAW,GAAG;AAChB,YAAM,IAAI,MAAM,0EAA0E;AAAA,IAC5F;AACA,YAAQ,OAAO,UAAU,CAAC;AAC1B,SAAK,IAAI;AAAA,EACX;AAAA,EACA,MAAM;AACJ,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AACA,UAAM,QAAQ,KAAK,kBAAkB,KAAK;AAC1C,YAAQ,IAAI,aAAa,GAAG,KAAK,qBAAqB,KAAK,mBAAmB,IAAI,QAAQ;AAC1F,UAAM,YAAY,KAAK,gBAAgB,KAAK;AAC5C,YAAQ,IAAI,oBAAoB,KAAK,oBAAoB;AACzD,YAAQ,IAAI,iBAAiB,KAAK,kBAAkB,KAAK,MAAM,MAAM,SAAS,KAAK;AAAA,EACrF;AAAA,EACA,IAAI,oBAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,eAAe;AACjB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,qBAAqB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,qBAAqB;AACnB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,UAAU;AACR,QAAI,KAAK,gBAAgB,MAAM;AAC7B;AAAA,IACF;AACA,eAAW,YAAY,KAAK,cAAc;AACxC,WAAK,aAAa,UAAU,QAAQ,CAAC,QAAQ;AAC3C,aAAK,MAAM,oBAAoB,IAAI,OAAO;AAAA,MAC5C,CAAC;AAAA,IACH;AACA,eAAW,YAAY,KAAK,cAAc;AACxC,WAAK,aAAa,UAAU,QAAQ,CAAC,QAAQ;AAC3C,aAAK,MAAM,oBAAoB,IAAI,OAAO;AAAA,MAC5C,CAAC;AAAA,IACH;AACA,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,kBAAkB;AACvB,SAAK,kBAAkB;AACvB,SAAK,qBAAqB;AAC1B,SAAK,gBAAgB;AAAA,EACvB;AACF;AACA,SAAS,0BAA0B,IAAI,gBAAgB;AACrD,QAAM,QAAQ;AACd,MAAI,mBAAmB,MAAM,MAAM;AACjC,WAAO;AAAA,EACT,WAAW,mBAAmB,MAAM,MAAM;AACxC,WAAO;AAAA,EACT,WAAW,mBAAmB,MAAM,SAAS;AAC3C,WAAO;AAAA,EACT,WAAW,mBAAmB,GAAG,MAAM;AACrC,WAAO;AAAA,EACT,WAAW,mBAAmB,MAAM,SAAS;AAC3C,WAAO;AAAA,EACT,WAAW,mBAAmB,MAAM,OAAO;AACzC,WAAO;AAAA,EACT;AACA,QAAM,IAAI,MAAM,2BAA2B,gBAAgB;AAC7D;AACA,SAAS,aAAa,OAAO,iBAAiB,IAAI,eAAe,UAAU;AACzE,QAAM,iBAAiB,iCAAiC,iBAAiB,aAAa;AACtF,MAAI;AACJ,MAAI,UAAU;AACZ,UAAM,CAAC,aAAa,YAAY,IAAI,uCAAuC,MAAM,IAAI,MAAM,EAAE;AAC7F,kBAAc,cAAc;AAAA,EAC9B,OAAO;AACL,UAAM,CAAC,OAAO,MAAM,IAAI,yCAAyC,MAAM,IAAI,MAAM,EAAE;AACnF,kBAAc,QAAQ;AAAA,EACxB;AACA,QAAM,mBAAmB,0BAA0B,IAAI,cAAc;AACrE,SAAO,cAAc;AACvB;AACA,SAAS,iCAAiC,iBAAiB,eAAe;AACxE,UAAQ,iBAAiB;AAAA,IACvB,KAAK,oBAAoB;AACvB,aAAO,wCAAwC,aAAa;AAAA,IAC9D,KAAK,oBAAoB;AACvB,aAAO,+CAA+C,aAAa;AAAA,IACrE,KAAK,oBAAoB;AACvB,aAAO,yCAAyC,aAAa;AAAA,IAC/D,KAAK,oBAAoB;AACvB,aAAO,yCAAyC,aAAa;AAAA,IAC/D,KAAK,oBAAoB;AACvB,aAAO,+CAA+C,aAAa;AAAA,IACrE;AACE,YAAM,IAAI,MAAM,iCAAiC,iBAAiB;AAAA,EACtE;AACF;AACA,SAAS,+BAA+B,UAAU;AAChD,MAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,QAAI,UAAU;AACZ,aAAO,oBAAoB;AAAA,IAC7B;AACA,WAAO,oBAAoB;AAAA,EAC7B;AACA,MAAI,UAAU;AACZ,WAAO,oBAAoB;AAAA,EAC7B;AACA,SAAO,oBAAoB;AAC7B;AACA,SAAS,kCAAkC,gBAAgB,UAAU;AACnE,MAAI,mBAAmB,aAAa,QAAQ;AAC1C,WAAO,oBAAoB;AAAA,EAC7B,WAAW,mBAAmB,aAAa,UAAU,kBAAkB,MAAM;AAC3E,WAAO,+BAA+B,QAAQ;AAAA,EAChD,WAAW,mBAAmB,aAAa,YAAY,mBAAmB,aAAa,QAAQ;AAC7F,WAAO,oBAAoB;AAAA,EAC7B;AACA,QAAM,IAAI,MAAM,gCAAgC,gBAAgB;AAClE;AACA,SAAS,uBAAuB,cAAc,iBAAiB,UAAU;AACvE,SAAO,GAAG,aAAa,MAAM,aAAa,MAAM,mBAAmB;AACrE;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,QAAQ,WAAW;AAC7B,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUR;AACF;AACA,IAAI,oBAAoB;AACxB,IAAI,SAAS;AACb,IAAI,MAAM;AACV,IAAI,OAAO;AACX,IAAI,OAAO,oBAAoB;AAAA;AAAA;AAG/B,IAAI,QAAQ,oBAAoB;AAAA;AAAA;AAGhC,IAAI,QAAQ;AACZ,IAAI,UAAU;AAGd,IAAI,UAAU;AACd,IAAI,OAAO;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUX,IAAI,QAAQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWZ,IAAI,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWb,IAAI,WAAW;AACf,IAAI,uBAAuB,MAAM;AAAA,EAC/B,YAAY,QAAQ,WAAW;AAC7B,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUR;AACF;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,OAAO,YAAY;AACzB,UAAM,WAAW,YAAY,MAAM,IAAI;AACvC,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,eAAe,gBAAgB,MAAM,QAAQ;AACnD,UAAM,YAAY,SAAS,MAAM,EAAE;AACnC,UAAM,UAAU,QAAQ,IAAI,OAAO,QAAQ,UAAU,KAAK,GAAG;AAC7D,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,kCACwB;AAAA;AAAA,4CAEU;AAAA;AAAA;AAAA,EAG1C;AACF;AAGA,IAAI,aAAa,qBAAqB;AACtC,IAAI,mBAAmB;AACvB,IAAI,mBAAmB;AACvB,IAAI,eAAe,CAAC;AACpB,SAAS,eAAe,cAAc;AACpC,MAAI,gBAAgB,cAAc;AAChC,WAAO,aAAa;AAAA,EACtB;AACA,eAAa,gBAAgB,CAAC;AAC9B,SAAO,aAAa;AACtB;AACA,IAAI,6BAA6B,IAAI,EAAE,UAAU,4BAA4B;AAC7E,IAAI,yBAAyB;AAC7B,SAAS,qBAAqB;AAC5B,MAAI,IAAI,EAAE,OAAO,UAAU,MAAM;AAC/B,WAAO;AAAA,EACT;AACA,SAAO,IAAI,EAAE,OAAO,OAAO,SAAS,IAAI,EAAE,OAAO,OAAO,QAAQ,OAAO,mBAAmB,yBAAyB,OAAO;AAC5H;AACA,IAAI,mBAAmB,cAAc,cAAc;AAAA,EACjD,YAAY,aAAa;AACvB,UAAM;AACN,SAAK,cAA8B,oBAAI,QAAQ;AAC/C,SAAK,kBAAkC,oBAAI,QAAQ;AACnD,SAAK,eAA+B,oBAAI,QAAQ;AAChD,SAAK,gBAAgB;AACrB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AACtB,SAAK,kBAAkB;AACvB,SAAK,oBAAoB;AACzB,SAAK,iBAAiB;AACtB,SAAK,WAAW;AAChB,QAAI,CAAC,IAAI,EAAE,QAAQ,WAAW,GAAG;AAC/B,YAAM,IAAI,MAAM,uCAAuC;AAAA,IACzD;AACA,QAAI;AACJ,QAAI,eAAe,MAAM;AACvB,UAAI,uBAAuB,cAAc;AACvC,mBAAW;AAAA,MACb,OAAO;AACL,cAAM,KAAK,gBAAgB,IAAI,EAAE,UAAU,eAAe,GAAG,WAAW;AACxE,mBAAW,IAAI,aAAa,EAAE;AAAA,MAChC;AACA,WAAK,cAAc,CAAC;AACpB,WAAK,sBAAsB;AAAA,IAC7B,OAAO;AACL,YAAM,KAAK,gBAAgB,IAAI,EAAE,UAAU,eAAe,CAAC;AAC3D,iBAAW,IAAI,aAAa,EAAE;AAC9B,WAAK,cAAc,eAAe,IAAI,EAAE,UAAU,eAAe,CAAC;AAClE,WAAK,sBAAsB;AAAA,IAC7B;AACA,SAAK,QAAQ;AACb,SAAK,SAAS,KAAK,MAAM,GAAG;AAC5B,SAAK,iBAAiB,IAAI,eAAe,KAAK,KAAK;AACnD,SAAK,qBAAqB,mBAAmB;AAC7C,SAAK,UAAU,IAAI,YAAY,MAAM,OAAO,CAAC;AAAA,EAC/C;AAAA,EACA,aAAa;AACX,WAAO,iBAAiB;AAAA,EAC1B;AAAA,EACA,aAAa;AACX,WAAO,KAAK,QAAQ,WAAW,IAAI,KAAK;AAAA,EAC1C;AAAA,EACA,aAAa,SAAS,OAAO,OAAO,WAAW,UAAU,UAAU;AACjE,UAAM,SAAS,KAAK,eAAe,OAAO,KAAK;AAC/C,UAAM,SAAS,KAAK,QAAQ,IAAI,OAAO,MAAM;AAC7C,WAAO,WAAW;AAClB,WAAO,UAAU,EAAE,SAAS,UAAU,CAAC,WAAW,QAAQ,EAAE;AAC5D,WAAO,WAAW,CAAC,WAAW,QAAQ;AACtC,UAAM,YAAY,aAAa,KAAK;AACpC,UAAM,UAAU,IAAI,oBAAoB,WAAW,OAAO,QAAQ;AAClE,UAAM,SAAS,KAAK,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,CAAC,CAAC,WAAW,QAAQ,CAAC,CAAC;AACrF,WAAO,QAAQ;AACf,WAAO,UAAU;AACjB,SAAK,8BAA8B,MAAM;AACzC,WAAO,OAAO;AAAA,EAChB;AAAA,EACA,MAAM,QAAQ,OAAO,OAAO;AAC1B,QAAI,IAAI,EAAE,QAAQ,gCAAgC,KAAK,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC7E,WAAK,uBAAuB,MAAM;AAAA,IACpC;AACA,QAAI,UAAU,eAAe,UAAU,MAAM;AAC3C,YAAM,IAAI,MAAM,uEAAuE;AAAA,IACzF;AACA,UAAM,SAAS,EAAE,IAAI,KAAK,WAAW,EAAE;AACvC,SAAK,QAAQ,IAAI,QAAQ,EAAE,OAAO,OAAO,QAAQ,OAAO,aAAa,QAAQ,UAAU,EAAE,CAAC;AAC1F,WAAO;AAAA,EACT;AAAA,EACA,SAAS,QAAQ;AACf,QAAI,KAAK,QAAQ,IAAI,MAAM,GAAG;AAC5B,YAAM,aAAa,KAAK,QAAQ,IAAI,MAAM;AAC1C,aAAO,WAAW;AAAA,IACpB;AACA,WAAO;AAAA,EACT;AAAA,EACA,OAAO,QAAQ;AACb,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,YAAQ;AAAA,EACV;AAAA,EACA,OAAO,QAAQ;AACb,QAAI,KAAK,QAAQ,IAAI,MAAM,GAAG;AAC5B,YAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,cAAQ;AAAA,IACV;AAAA,EACF;AAAA,EACA,KAAK,QAAQ,QAAQ,OAAO,OAAO,UAAU;AAC3C,QAAI,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC1B,WAAK,uBAAuB,MAAM;AAAA,IACpC;AACA,QAAI,UAAU,aAAa;AACzB,YAAM,IAAI,MAAM,uEAAuE;AAAA,IACzF;AACA,SAAK,QAAQ,IAAI,QAAQ,EAAE,OAAO,OAAO,QAAQ,OAAO,aAAa,QAAQ,SAAS,CAAC;AAAA,EACzF;AAAA,EACA,8BAA8B,YAAY;AACxC,SAAK,YAAY,WAAW,MAAM;AAAA,EACpC;AAAA,EACA,SAAS,QAAQ;AACf,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,QAAQ,OAAO,oBAAoB,OAAO,QAAQ,OAAO,SAAS,IAAI;AAC9E,QAAI,UAAU,MAAM;AAClB,UAAI;AACJ,UAAI,UAAU;AACZ,kBAAU,IAAI,qBAAqB,OAAO,KAAK;AAAA,MACjD,OAAO;AACL,kBAAU,IAAI,eAAe,OAAO,KAAK;AAAA,MAC3C;AACA,YAAM,MAAM,KAAK,gBAAgB,SAAS,CAAC,EAAE,QAAQ,OAAO,MAAM,CAAC,GAAG,KAAK;AAC3E,YAAM,OAAO,KAAK,SAAS,IAAI,MAAM;AACrC,WAAK,8BAA8B,GAAG;AACtC,aAAO;AAAA,IACT;AACA,QAAI,UAAU,MAAM;AAClB,aAAO,KAAK,qBAAqB,MAAM;AAAA,IACzC;AACA,QAAI,UAAU,UAAU;AACtB,aAAO;AAAA,IACT;AACA,UAAM,oBAAoB,KAAK,gBAAgB;AAC/C,QAAI;AACJ,QAAI,mBAAmB;AACrB,cAAQ,aAAa,IAAI;AAAA,IAC3B;AACA,QAAI;AACJ,QAAI,UAAU,aAAa;AACzB,YAAM,aAAa,KAAK,SAAS,mBAAmB,KAAK,MAAM;AAC/D,YAAM,aAAa,KAAK,SAAS,mBAAmB,KAAK,MAAM;AAC/D,eAAS,qBAAqB,uBAAuB,YAAY,UAAU;AAAA,IAC7E,OAAO;AACL,eAAS,KAAK,qBAAqB,MAAM;AAAA,IAC3C;AACA,QAAI,mBAAmB;AACrB,WAAK,kBAAkB,aAAa,IAAI,IAAI;AAAA,IAC9C;AACA,WAAO,KAAK,qBAAqB,QAAQ,MAAM;AAAA,EACjD;AAAA,EACA,MAAM,KAAK,QAAQ;AACjB,QAAI,KAAK,YAAY,IAAI,MAAM,GAAG;AAChC,YAAM,eAAe,KAAK,YAAY,IAAI,MAAM;AAChD,aAAO,IAAI,QAAQ,CAAC,YAAY,aAAa,KAAK,OAAO,CAAC;AAAA,IAC5D;AACA,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,QAAQ,OAAO,OAAO,QAAQ,OAAO,oBAAoB,SAAS,IAAI;AAC9E,QAAI,UAAU,MAAM;AAClB,UAAI;AACJ,UAAI,UAAU;AACZ,kBAAU,IAAI,qBAAqB,OAAO,KAAK;AAAA,MACjD,OAAO;AACL,kBAAU,IAAI,eAAe,OAAO,KAAK;AAAA,MAC3C;AACA,YAAM,MAAM,KAAK,gBAAgB,SAAS,CAAC,EAAE,QAAQ,OAAO,MAAM,CAAC,GAAG,KAAK;AAC3E,YAAM,OAAO,KAAK,KAAK,IAAI,MAAM;AACjC,WAAK,8BAA8B,GAAG;AACtC,aAAO;AAAA,IACT;AACA,QAAI,UAAU,MAAM;AAClB,aAAO,KAAK,qBAAqB,MAAM;AAAA,IACzC;AACA,QAAI,IAAI,EAAE,QAAQ,OAAO,GAAG;AAC1B,UAAI,CAAC,IAAI,EAAE,QAAQ,8BAA8B,KAAK,IAAI,EAAE,UAAU,eAAe,MAAM,GAAG;AAC5F,cAAM,IAAI,MAAM,8FAA8F;AAAA,MAChH;AAAA,IACF;AACA,QAAI,UAAU;AACd,QAAI;AACJ,QAAI,UAAU,eAAe,IAAI,EAAE,IAAI,wBAAwB,GAAG;AAChE,0BAAoB,KAAK,OAAO,MAAM;AACtC,YAAM,UAAU,KAAK,QAAQ,IAAI,kBAAkB,MAAM;AACzD,gBAAU,KAAK,MAAM,wBAAwB,QAAQ,QAAQ,SAAS,GAAG,iBAAiB,KAAK,CAAC;AAAA,IAClG;AACA,SAAK,YAAY,IAAI,QAAQ,CAAC,CAAC;AAC/B,QAAI,UAAU,aAAa;AACzB,YAAM,KAAK,MAAM,sBAAsB;AAAA,IACzC;AACA,QAAI;AACJ,QAAI,UAAU,aAAa;AACzB,YAAM,KAAK,MAAM,QAAQ,IAAI;AAAA,QAC3B,KAAK,KAAK,mBAAmB,KAAK,MAAM;AAAA,QACxC,KAAK,KAAK,mBAAmB,KAAK,MAAM;AAAA,MAC1C,CAAC;AACD,YAAM,aAAa,GAAG;AACtB,YAAM,aAAa,GAAG;AACtB,aAAO,qBAAqB,uBAAuB,YAAY,UAAU;AAAA,IAC3E,WAAW,WAAW,MAAM;AAC1B,aAAO,KAAK,qBAAqB,MAAM;AAAA,IACzC,OAAO;AACL,YAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,aAAO,KAAK,MAAM,gCAAgC,SAAS,IAAI;AAAA,IACjE;AACA,QAAI,qBAAqB,MAAM;AAC7B,WAAK,8BAA8B,iBAAiB;AAAA,IACtD;AACA,QAAI,WAAW,MAAM;AACnB,YAAM,KAAK,KAAK,MAAM;AACtB,mBAAa,IAAI,MAAM,GAAG,aAAa,OAAO,CAAC;AAAA,IACjD;AACA,UAAM,YAAY,KAAK,qBAAqB,QAAQ,IAAI;AACxD,UAAM,cAAc,KAAK,YAAY,IAAI,MAAM;AAC/C,SAAK,YAAY,OAAO,MAAM;AAC9B,gBAAY,QAAQ,CAAC,YAAY,QAAQ,SAAS,CAAC;AACnD,QAAI,KAAK,gBAAgB,IAAI,MAAM,GAAG;AACpC,WAAK,gBAAgB,OAAO,MAAM;AAClC,UAAI,KAAK,YAAY,MAAM,GAAG;AAC5B,eAAO,EAAE,aAAa,QAAQ,IAAI;AAAA,MACpC;AACA,WAAK;AAAA,IACP;AACA,WAAO;AAAA,EACT;AAAA,EACA,UAAU,QAAQ,UAAU,CAAC,GAAG;AAC9B,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,QAAQ,OAAO,OAAO,QAAQ,OAAO,UAAU,QAAQ,IAAI;AACnE,QAAI,UAAU,aAAa;AACzB,YAAM,IAAI,MAAM,uDAAuD;AAAA,IACzE;AACA,QAAI,UAAU,MAAM;AAClB,UAAI;AACJ,UAAI,UAAU;AACZ,kBAAU,IAAI,qBAAqB,OAAO,KAAK;AAAA,MACjD,OAAO;AACL,kBAAU,IAAI,eAAe,OAAO,KAAK;AAAA,MAC3C;AACA,YAAM,MAAM,KAAK,gBAAgB,SAAS,CAAC,EAAE,QAAQ,OAAO,MAAM,CAAC,GAAG,KAAK;AAC3E,YAAM,eAAe,KAAK,UAAU,KAAK,OAAO;AAChD,WAAK,8BAA8B,GAAG;AACtC,aAAO;AAAA,IACT;AACA,QAAI,WAAW,MAAM;AACnB,UAAI,UAAU,MAAM;AAClB,cAAM,IAAI,MAAM,gCAAgC;AAAA,MAClD,OAAO;AACL,cAAM,IAAI,MAAM,iCAAiC;AAAA,MACnD;AAAA,IACF;AACA,UAAM,YAAY,KAAK,OAAO,QAAQ,QAAQ,cAAc;AAC5D,UAAM,YAAY,OAAO,EAAE,yBAAyB,SAAS;AAC7D,UAAM,UAAU,KAAK,QAAQ,IAAI,UAAU,MAAM;AACjD,WAAO,OAAO,OAAO,EAAE,UAAU,GAAG,QAAQ,OAAO;AAAA,EACrD;AAAA,EACA,WAAW,GAAG;AACZ,UAAM,OAAO,KAAK,SAAS,EAAE,MAAM;AACnC,QAAI,EAAE,UAAU,UAAU;AACxB,UAAI;AACF,cAAM,UAAU,KAAK,IAAI,CAAC,MAAM,aAAa,aAAa,CAAC,CAAC;AAC5D,eAAO,OAAO,EAAE,OAAO,EAAE,OAAO,OAAO;AAAA,MACzC,SAAS,IAAP;AACA,cAAM,IAAI,MAAM,kDAAkD;AAAA,MACpE;AAAA,IACF;AACA,WAAO,OAAO,EAAE,OAAO,EAAE,OAAO,IAAI;AAAA,EACtC;AAAA,EACA,uBAAuB,QAAQ;AAC7B,QAAI,UAAU,MAAM;AAClB;AAAA,IACF;AACA,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,YAAM,MAAM,OAAO;AACnB,UAAI,CAAC,iBAAiB,GAAG,GAAG;AAC1B,YAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,gBAAM,MAAM,aAAa,kJAAkJ;AAAA,QAC7K;AACA,cAAM,MAAM,aAAa,2CAA2C;AAAA,MACtE;AAAA,IACF;AAAA,EACF;AAAA,EACA,qBAAqB,QAAQ;AAC3B,UAAM,EAAE,OAAO,OAAO,SAAS,IAAI,KAAK,QAAQ,IAAI,MAAM;AAC1D,UAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,QAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAM,WAAW,KAAK,QAAQ,IAAI,UAAU,MAAM;AAClD,YAAM,QAAQ,KAAK,MAAM,gCAAgC,SAAS,QAAQ,SAAS,GAAG,iBAAiB,KAAK,CAAC,EAAE,SAAS,GAAG,IAAI;AAC/H,WAAK,8BAA8B,SAAS;AAC5C,aAAO;AAAA,IACT;AACA,UAAM,yBAAyB,IAAI,EAAE,QAAQ,YAAY,KAAK,aAAa;AAC3E,UAAM,cAAc,yBAAyB,aAAa,KAAK,IAAI;AACnE,UAAM,UAAU,yBAAyB,IAAI,yBAAyB,WAAW,IAAI,IAAI,mBAAmB,WAAW;AACvH,UAAM,SAAS,KAAK,gBAAgB,SAAS,CAAC,EAAE,OAAO,aAAa,OAAO,OAAO,CAAC,GAAG,SAAS;AAC/F,UAAM,UAAU,KAAK,QAAQ,IAAI,OAAO,MAAM;AAC9C,UAAM,OAAO,KAAK,MAAM,gDAAgD,QAAQ,QAAQ,SAAS,QAAQ,SAAS,IAAI,QAAQ,SAAS,EAAE,EAAE,SAAS,GAAG,IAAI;AAC3J,SAAK,8BAA8B,MAAM;AACzC,WAAO;AAAA,EACT;AAAA,EACA,iBAAiB;AACf,WAAO,IAAI,EAAE,UAAU,+CAA+C,IAAI;AAAA,EAC5E;AAAA,EACA,KAAK,GAAG;AACN,UAAM,kBAAkB,KAAK;AAC7B,UAAM,kBAAkB,CAAC;AACzB,QAAI,gBAAgB;AACpB,QAAI,KAAK,sBAAsB,MAAM;AACnC,WAAK,qBAAqB;AAC1B,sBAAgB;AAAA,IAClB,OAAO;AACL,WAAK,aAAa,KAAK,eAAe;AAAA,IACxC;AACA,SAAK,eAAe;AACpB,MAAE;AACF,UAAM,8BAA8B,aAAa,QAAQ,KAAK,aAAa,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,EAAE,OAAO,CAAC,MAAM,KAAK,IAAI;AACvH,UAAM,4BAA4B,aAAa,QAAQ,KAAK,aAAa,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,EAAE,OAAO,CAAC,MAAM,KAAK,IAAI;AACpH,SAAK,eAAe;AACpB,QAAI,eAAe;AACjB,WAAK,qBAAqB;AAAA,IAC5B;AACA,UAAM,MAAM;AAAA,MACV,cAAc,KAAK;AAAA,MACnB,gBAAgB,KAAK;AAAA,MACrB,UAAU;AAAA,MACV,QAAQ;AAAA,IACV;AACA,YAAQ,YAAY;AAClB,UAAI,IAAI,EAAE,UAAU,+CAA+C,IAAI,GAAG;AACxE,cAAM,WAAW,MAAM,QAAQ,IAAI,2BAA2B;AAC9D,YAAI,cAAc,aAAa,IAAI,QAAQ;AAC3C,YAAI,yBAAyB,MAAM,SAAS,IAAI,CAAC,GAAG,OAAO,EAAE,MAAM,0BAA0B,IAAI,IAAI,EAAE,EAAE,EAAE,IAAI,CAAC,MAAM,GAAG,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,IAAI;AAAA,MACvJ,OAAO;AACL,YAAI,cAAc;AAAA,UAChB,OAAO;AAAA,QACT;AAAA,MACF;AACA,WAAK,eAAe;AACpB,WAAK,iBAAiB;AACtB,aAAO;AAAA,IACT,GAAG;AAAA,EACL;AAAA,EACA,SAAS;AACP,WAAO;AAAA,MACL,YAAY;AAAA,MACZ,eAAe,KAAK;AAAA,MACpB,wBAAwB,KAAK,eAAe;AAAA,MAC5C,mBAAmB,KAAK,eAAe;AAAA,IACzC;AAAA,EACF;AAAA,EACA,aAAa;AACX,QAAI,IAAI,EAAE,UAAU,+CAA+C,IAAI,GAAG;AACxE,aAAO,KAAK,MAAM,WAAW;AAAA,IAC/B;AACA,WAAO,EAAE,SAAS,aAAa,IAAI,GAAG,OAAO,KAAK;AAAA,EACpD;AAAA,EACA,SAAS,OAAO;AACd,QAAI,IAAI,EAAE,UAAU,+CAA+C,IAAI,GAAG;AACxE,WAAK,MAAM,SAAS;AACpB,aAAO;AAAA,IACT;AACA,UAAM,QAAQ,aAAa,IAAI;AAC/B,WAAO;AAAA,EACT;AAAA,EACA,MAAM,aAAa,OAAO;AACxB,QAAI,IAAI,EAAE,UAAU,+CAA+C,IAAI,GAAG;AACxE,aAAO,KAAK,MAAM,uBAAuB,KAAK;AAAA,IAChD;AACA,UAAM,aAAa;AACnB,WAAO,WAAW,QAAQ,WAAW;AAAA,EACvC;AAAA,EACA,YAAY,QAAQ,QAAQ,OAAO;AACjC,QAAI,KAAK,gBAAgB,IAAI,MAAM,GAAG;AACpC,aAAO;AAAA,IACT;AACA,QAAI,CAAC,KAAK,QAAQ,IAAI,MAAM,GAAG;AAC7B,aAAO;AAAA,IACT;AACA,QAAI,OAAO;AACT,WAAK,QAAQ,IAAI,MAAM,EAAE,WAAW;AAAA,IACtC,OAAO;AACL,WAAK,QAAQ,IAAI,MAAM,EAAE;AAAA,IAC3B;AACA,QAAI,CAAC,SAAS,KAAK,QAAQ,IAAI,MAAM,EAAE,WAAW,GAAG;AACnD,aAAO;AAAA,IACT;AACA,QAAI,KAAK,YAAY,IAAI,MAAM,GAAG;AAChC,WAAK,gBAAgB,IAAI,MAAM;AAC/B,WAAK;AACL,aAAO;AAAA,IACT;AACA,SAAK,eAAe,MAAM;AAC1B,UAAM,EAAE,mBAAmB,IAAI,KAAK,QAAQ,IAAI,MAAM;AACtD,QAAI,sBAAsB,MAAM;AAC9B,WAAK,YAAY,mBAAmB,KAAK,QAAQ,KAAK;AACtD,WAAK,YAAY,mBAAmB,KAAK,QAAQ,KAAK;AAAA,IACxD;AACA,SAAK,QAAQ,OAAO,MAAM;AAC1B,WAAO;AAAA,EACT;AAAA,EACA,eAAe,QAAQ;AACrB,UAAM,EAAE,SAAS,OAAO,UAAU,OAAO,UAAU,OAAO,OAAO,IAAI,KAAK,QAAQ,IAAI,MAAM;AAC5F,UAAM,MAAM,UAAU,OAAO,cAAc;AAC3C,UAAM,WAAW,KAAK,aAAa,IAAI,GAAG;AAC1C,QAAI,WAAW,GAAG;AAChB,WAAK,aAAa,IAAI,KAAK,WAAW,CAAC;AAAA,IACzC,OAAO;AACL,WAAK,aAAa,OAAO,GAAG;AAC5B,UAAI,WAAW,MAAM;AACnB,aAAK,iBAAiB,KAAK,aAAa,UAAU,KAAK;AACvD,aAAK,eAAe,eAAe,SAAS,UAAU,OAAO,QAAQ;AAAA,MACvE;AAAA,IACF;AACA,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,YAAQ,UAAU;AAClB,YAAQ,WAAW;AACnB,YAAQ,WAAW;AACnB,YAAQ,QAAQ;AAAA,EAClB;AAAA,EACA,WAAW,QAAQ;AACjB,SAAK,YAAY,MAAM;AACvB,WAAO,KAAK,QAAQ,IAAI,MAAM,EAAE,QAAQ;AAAA,EAC1C;AAAA,EACA,YAAY,QAAQ;AAClB,WAAO,KAAK,QAAQ,IAAI,MAAM;AAAA,EAChC;AAAA,EACA,mBAAmB,QAAQ,gBAAgB,4BAA4B;AACrE,WAAO,IAAI,EAAE,QAAQ,mBAAmB,KAAK,OAAO,MAAM,CAAC,WAAW,KAAK,QAAQ,IAAI,OAAO,MAAM,EAAE,WAAW,QAAQ,aAAa,cAAc,OAAO,KAAK,IAAI,aAAa;AAAA,EACnL;AAAA,EACA,kBAAkB;AAChB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,MAAM,WAAW;AACf,yBAAqB,KAAK,uEAAuE;AACjG,UAAM,WAAW,UAAU,SAAS;AACpC,WAAO,WAAW,UAAU,OAAO,QAAQ;AAAA,EAC7C;AAAA,EACA,cAAc,GAAG,KAAK,OAAO;AAC3B,UAAM,UAAU,IAAI,qBAAqB,EAAE,OAAO,GAAG;AACrD,UAAM,UAAU,KAAK,cAAc,SAAS,CAAC,CAAC,GAAG,KAAK;AACtD,WAAO,OAAO,EAAE,yBAAyB,OAAO;AAAA,EAClD;AAAA,EACA,IAAI,GAAG;AACL,QAAI,KAAK,mBAAmB,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,aAAa;AAC3D,YAAM,YAAY,iBAAiB,KAAK,QAAQ,IAAI,EAAE,MAAM,EAAE,MAAM;AACpE,aAAO,KAAK,WAAW,EAAE,OAAO,EAAE,OAAO,SAAS;AAAA,IACpD;AACA,QAAI,IAAI,EAAE,QAAQ,6BAA6B,GAAG;AAChD,aAAO,KAAK,cAAc,GAAG,KAAK,EAAE,KAAK;AAAA,IAC3C;AACA,UAAM,UAAU,IAAI,eAAe,EAAE,OAAO,GAAG;AAC/C,UAAM,UAAU,KAAK,cAAc,SAAS,CAAC,CAAC,CAAC;AAC/C,WAAO,OAAO,EAAE,yBAAyB,OAAO;AAAA,EAClD;AAAA,EACA,eAAe,OAAO,OAAO,QAAQ;AACnC,QAAI;AACJ,QAAI,UAAU,YAAY,UAAU,QAAQ,OAAO,SAAS,KAAK,aAAa,SAAS,OAAO,EAAE,GAAG;AACjG,YAAM,gBAAgB,OAAO,IAAI,CAAC,MAAM,aAAa,aAAa,CAAC,CAAC;AACpE,eAAS,KAAK,MAAM,eAAe,OAAO,KAAK;AAAA,IACjD,OAAO;AACL,eAAS,KAAK,MAAM,QAAQ,OAAO,KAAK;AAAA,IAC1C;AACA,SAAK,QAAQ,IAAI,MAAM,EAAE,QAAQ;AACjC,WAAO,EAAE,QAAQ,OAAO,MAAM;AAAA,EAChC;AAAA,EACA,WAAW,OAAO,OAAO,QAAQ;AAC/B,WAAO,OAAO,EAAE,yBAAyB,KAAK,eAAe,OAAO,OAAO,MAAM,GAAG,IAAI;AAAA,EAC1F;AAAA,EACA,aAAa,QAAQ;AACnB,UAAM,UAAU,IAAI,cAAc,OAAO,KAAK;AAC9C,WAAO,KAAK,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,KAAK;AAAA,EAC7D;AAAA,EACA,WAAW,QAAQ;AACjB,UAAM,UAAU,IAAI,YAAY,OAAO,KAAK;AAC5C,UAAM,8BAA8B;AACpC,WAAO,KAAK,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,OAAO,MAAM,2BAA2B;AAAA,EAChG;AAAA,EACA,cAAc,QAAQ,YAAY;AAChC,UAAM,eAAe;AAAA,MACnB,YAAY,OAAO,KAAK;AAAA,MACxB,GAAG,YAAY,OAAO,KAAK;AAAA,IAC7B;AACA,UAAM,UAAU;AAAA,MACd,OAAO,OAAO;AAAA,MACd,OAAO;AAAA,MACP,QAAQ,OAAO;AAAA,IACjB;AACA,UAAM,iBAAiB;AAAA,MACrB,YAAY,UAAU;AAAA,MACtB,GAAG,YAAY,UAAU;AAAA,IAC3B;AACA,UAAM,UAAU,IAAI,qBAAqB,gBAAgB,YAAY;AACrE,UAAM,gCAAgC;AACtC,UAAM,eAAe,CAAC,YAAY;AAClC,UAAM,SAAS,KAAK,gBAAgB,SAAS,CAAC,OAAO,GAAG,OAAO,OAAO,cAAc,6BAA6B;AACjH,WAAO,EAAE,QAAQ,OAAO,QAAQ,OAAO,YAAY,OAAO,OAAO,MAAM;AAAA,EACzE;AAAA,EACA,OAAO,QAAQ,gBAAgB;AAC7B,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,UAAU,OAAO,MAAM,IAAI;AACnC,QAAI,kBAAkB,MAAM;AAC1B,YAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,YAAM,UAAU,eAAe,KAAK,eAAe,KAAK;AACxD,mBAAa,OAAO,QAAQ,SAAS,MAAM,2GAA2G;AAAA,IACxJ;AACA,UAAM,YAAY,aAAa,KAAK;AACpC,QAAI;AACJ,QAAI,UAAU;AACZ,gBAAU,IAAI,0BAA0B,SAAS;AAAA,IACnD,OAAO;AACL,gBAAU,IAAI,oBAAoB,SAAS;AAAA,IAC7C;AACA,UAAM,gCAAgC;AACtC,UAAM,eAAe,CAAC,kBAAkB,OAAO,iBAAiB,iBAAiB,SAAS,CAAC;AAC3F,UAAM,MAAM,KAAK,gBAAgB,SAAS,CAAC,EAAE,OAAO,WAAW,OAAO,OAAO,CAAC,GAAG,OAAO,cAAc,+BAA+B,cAAc;AACnJ,WAAO,EAAE,OAAO,OAAO,QAAQ,IAAI,OAAO;AAAA,EAC5C;AAAA,EACA,gBAAgB,SAAS,QAAQ,aAAa,qBAAqB,gCAAgC,OAAO,gBAAgB;AACxH,UAAM,SAAS,KAAK,eAAe,QAAQ,aAAa,WAAW;AACnE,UAAM,UAAU,KAAK,QAAQ,IAAI,OAAO,MAAM;AAC9C,QAAI,QAAQ,cAAc;AACxB,cAAQ,WAAW;AAAA,IACrB;AACA,QAAI,QAAQ,qBAAqB,cAAc,OAAO;AACpD,YAAM,aAAa,kBAAkB,OAAO,iBAAiB,iBAAiB,QAAQ,WAAW;AACjG,cAAQ,WAAW,WAAW,IAAI,CAAC,MAAM,IAAI,CAAC;AAAA,IAChD;AACA,QAAI,QAAQ,eAAe,MAAM;AAC/B,cAAQ,QAAQ,QAAQ;AAAA,IAC1B;AACA,QAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,cAAQ,SAAS,aAAa,uBAAuB,OAAO,OAAO,CAAC;AACpE,aAAO;AAAA,IACT;AACA,UAAM,gBAAgB,CAAC;AACvB,UAAM,aAAa,OAAO,IAAI,CAAC,WAAW;AACxC,UAAI,OAAO,UAAU,aAAa;AAChC,cAAM,IAAI,MAAM,iIAAiI;AAAA,MACnJ;AACA,UAAI,UAAU,KAAK,QAAQ,IAAI,OAAO,MAAM;AAC5C,UAAI,QAAQ,WAAW,MAAM;AAC3B,YAAI,CAAC,QAAQ,gBAAgB,aAAa,cAAc,OAAO,KAAK,KAAK,IAAI,EAAE,UAAU,2BAA2B,GAAG;AACrH,iBAAO;AAAA,YACL,OAAO,OAAO;AAAA,YACd,SAAS;AAAA,YACT,WAAW;AAAA,YACX,eAAe,QAAQ;AAAA,UACzB;AAAA,QACF;AACA,YAAI,QAAQ,cAAc;AACxB,kBAAQ,WAAW;AACnB,kBAAQ,QAAQ,OAAO;AAAA,QACzB;AAAA,MACF;AACA,WAAK,YAAY,OAAO,MAAM;AAC9B,UAAI,CAAC,CAAC,QAAQ,aAAa,CAAC,CAAC,QAAQ,cAAc;AACjD,iBAAS,QAAQ,WAAW,KAAK,aAAa,MAAM,IAAI,KAAK,WAAW,MAAM;AAC9E,sBAAc,KAAK,MAAM;AACzB,kBAAU,KAAK,QAAQ,IAAI,OAAO,MAAM;AAAA,MAC1C,WAAW,QAAQ,YAAY,CAAC,cAAc,QAAQ,OAAO,OAAO,KAAK,GAAG;AAC1E,cAAM,aAAa;AACnB,cAAM,cAAc,OAAO;AAC3B,eAAO,QAAQ,QAAQ;AACvB,iBAAS,KAAK,cAAc,QAAQ,WAAW;AAC/C,sBAAc,KAAK,MAAM;AACzB,kBAAU,KAAK,QAAQ,IAAI,OAAO,MAAM;AACxC,mBAAW,QAAQ;AAAA,MACrB;AACA,aAAO,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM;AAAA,IAC1D,CAAC;AACD,SAAK,YAAY,OAAO,MAAM;AAC9B,UAAM,aAAa,EAAE,OAAO,OAAO,OAAO,SAAS,SAAS,WAAW,MAAM;AAC7E,UAAM,MAAM,cAAc,SAAS,YAAY,UAAU;AACzD,UAAM,SAAS,KAAK,iBAAiB,KAAK,MAAM;AAC9C,aAAO,eAAe,KAAK,OAAO,SAAS,YAAY,UAAU;AAAA,IACnE,CAAC;AACD,UAAM,oBAAoB,KAAK,gBAAgB;AAC/C,QAAI;AACJ,QAAI,mBAAmB;AACrB,cAAQ,KAAK,WAAW;AAAA,IAC1B;AACA,QAAI,CAAC,IAAI,EAAE,IAAI,qBAAqB,GAAG;AACrC,iBAAW,KAAK,OAAO,QAAQ,YAAY,YAAY,mBAAmB;AAAA,IAC5E;AACA,kBAAc,QAAQ,CAAC,SAAS,KAAK,8BAA8B,IAAI,CAAC;AACxE,QAAI,mBAAmB;AACrB,cAAQ,KAAK,SAAS,KAAK;AAC3B,WAAK,aAAa,KAAK,EAAE,MAAM,QAAQ,YAAY,MAAM,OAAO,KAAK,aAAa,KAAK,EAAE,CAAC;AAAA,IAC5F;AACA,UAAM,mBAAmB,IAAI,EAAE,IAAI,uBAAuB;AAC1D,QAAI,mBAAmB,GAAG;AACxB,YAAM,QAAQ,aAAa,IAAI;AAC/B,UAAI,QAAQ,KAAK,kBAAkB,kBAAkB;AACnD,aAAK,MAAM,GAAG,MAAM;AACpB,aAAK,kBAAkB;AAAA,MACzB;AAAA,IACF;AACA,QAAI,CAAC,IAAI,EAAE,QAAQ,qBAAqB,KAAK,QAAQ,YAAY,kCAAkC,OAAO;AACxG,YAAM,WAAW,KAAK,aAAa,MAAM;AACzC,WAAK,8BAA8B,MAAM;AACzC,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AAAA,EACA,cAAc,SAAS,QAAQ,aAAa,qBAAqB,gCAAgC,OAAO;AACtG,kBAAc,eAAe,OAAO,GAAG;AACvC,UAAM,UAAU,KAAK,gBAAgB,SAAS,QAAQ,aAAa,qBAAqB,6BAA6B;AACrH,WAAO;AAAA,EACT;AAAA,EACA,iBAAiB,KAAK,WAAW;AAC/B,QAAI,EAAE,OAAO,KAAK,cAAc;AAC9B,WAAK,YAAY,OAAO,UAAU;AAAA,IACpC;AACA,WAAO,KAAK,YAAY;AAAA,EAC1B;AAAA,EACA,oBAAoB;AAClB,WAAO,KAAK;AAAA,EACd;AAAA,EACA,UAAU;AACR,QAAI,KAAK,UAAU;AACjB;AAAA,IACF;AACA,QAAI,CAAC,IAAI,EAAE,QAAQ,SAAS,GAAG;AAC7B,YAAM,UAAU,OAAO,KAAK,KAAK,WAAW;AAC5C,cAAQ,QAAQ,CAAC,QAAQ;AACvB,aAAK,MAAM,cAAc,KAAK,YAAY,KAAK,YAAY;AAC3D,eAAO,KAAK,YAAY;AAAA,MAC1B,CAAC;AAAA,IACH;AACA,SAAK,eAAe,QAAQ;AAC5B,QAAI,KAAK,UAAU,SAAS,OAAO,sBAAsB,eAAe,KAAK,kBAAkB,oBAAoB;AACjH,WAAK,OAAO,OAAO;AAAA,IACrB,OAAO;AACL,WAAK,SAAS;AAAA,IAChB;AACA,QAAI,KAAK,qBAAqB;AAC5B,WAAK,MAAM,UAAU;AACrB,WAAK,MAAM,QAAQ;AAAA,IACrB;AACA,SAAK,WAAW;AAAA,EAClB;AAAA,EACA,iBAAiB;AACf,QAAI,KAAK,uBAAuB,MAAM;AACpC,WAAK,sBAAsB,KAAK,MAAM;AACpC,YAAI,CAAC,IAAI,EAAE,IAAI,8BAA8B,GAAG;AAC9C,gBAAM,YAAY,IAAI,EAAE,QAAQ,OAAO;AACvC,cAAI,EAAE,IAAI,SAAS,KAAK;AACxB,gBAAM,sBAAsB,KAAK,IAAI,OAAO,IAAI,CAAC,EAAE,SAAS,EAAE;AAC9D,cAAI,EAAE,IAAI,SAAS,SAAS;AAC5B,cAAI,sBAAsB,GAAG;AAC3B,mBAAO;AAAA,UACT;AAAA,QACF;AACA,eAAO;AAAA,MACT,CAAC;AAAA,IACH;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EACA,UAAU;AACR,WAAO,KAAK,eAAe,MAAM,KAAK,mBAAmB;AAAA,EAC3D;AAAA,EACA,YAAY,QAAQ;AAClB,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,OAAO,OAAO,QAAQ,SAAS,OAAO,SAAS,IAAI;AAC3D,QAAI,WAAW,MAAM;AACnB;AAAA,IACF;AACA,UAAM,oBAAoB,KAAK,gBAAgB;AAC/C,QAAI;AACJ,QAAI,mBAAmB;AACrB,cAAQ,aAAa,IAAI;AAAA,IAC3B;AACA,QAAI,WAAW,QAAQ;AACvB,QAAI,YAAY,MAAM;AACpB,iBAAW,gCAAgC,OAAO,QAAQ;AAC1D,cAAQ,WAAW;AAAA,IACrB;AACA,QAAI,UAAU,MAAM;AAClB,YAAM,YAAY,aAAa,KAAK;AACpC,UAAI;AACJ,UAAI,QAAQ,SAAS,IAAI,SAAS,SAAS;AAC3C,YAAM,cAAc,kBAAkB,cAAc,kBAAkB;AACtE,UAAI,YAAY,CAAC,aAAa;AAC5B,SAAC,OAAO,MAAM,IAAI,uCAAuC,SAAS,IAAI,SAAS,EAAE;AAAA,MACnF;AACA,UAAI,UAAU;AACZ,kBAAU,IAAI,0BAA0B,WAAW,WAAW;AAAA,MAChE,OAAO;AACL,kBAAU,IAAI,oBAAoB,WAAW,WAAW;AAAA,MAC1D;AACA,YAAM,yBAAyB,cAAc,CAAC,QAAQ,KAAK,IAAI;AAC/D,YAAM,uBAAuB,KAAK,eAAe,wBAAwB,KAAK;AAC9E,YAAM,wBAAwB,KAAK,QAAQ,IAAI,qBAAqB,MAAM;AAC1E,UAAI,aAAa;AACf,8BAAsB,QAAQ,aAAa;AAAA,MAC7C,OAAO;AACL,8BAAsB,QAAQ,aAAa;AAAA,MAC7C;AACA,4BAAsB,WAAW;AACjC,WAAK,MAAM,2BAA2B,KAAK,WAAW,qBAAqB,MAAM,GAAG,OAAO,QAAQ,MAAM;AACzG,YAAM,eAAe,CAAC,CAAC,QAAQ,KAAK,CAAC;AACrC,YAAM,wBAAwB;AAC9B,YAAM,sBAAsB,KAAK,gBAAgB,SAAS,CAAC,oBAAoB,GAAG,OAAO,cAAc,qBAAqB;AAC5H,YAAM,gBAAgB,KAAK,QAAQ,IAAI,oBAAoB,MAAM;AACjE,cAAQ,WAAW,cAAc;AACjC,cAAQ,WAAW,cAAc;AACjC,cAAQ,QAAQ,cAAc;AAC9B,UAAI,CAAC,IAAI,EAAE,IAAI,qBAAqB,GAAG;AACrC,gBAAQ,UAAU,cAAc;AAChC,gBAAQ,SAAS;AACjB,aAAK,QAAQ,OAAO,oBAAoB,MAAM;AAAA,MAChD,OAAO;AACL,aAAK,YAAY,oBAAoB,MAAM;AAAA,MAC7C;AACA,WAAK,8BAA8B,oBAAoB;AACvD,UAAI,mBAAmB;AACrB,aAAK,gBAAgB,aAAa,IAAI,IAAI;AAAA,MAC5C;AAAA,IACF,OAAO;AACL,YAAM,aAAa,KAAK,eAAe,UAAU,OAAO,OAAO,QAAQ;AACvE,cAAQ,UAAU;AAAA,IACpB;AAAA,EACF;AAAA,EACA,qBAAqB,QAAQ,eAAe;AAC1C,UAAM,UAAU,KAAK,QAAQ,IAAI,MAAM;AACvC,UAAM,EAAE,MAAM,IAAI;AAClB,SAAK,eAAe,MAAM;AAC1B,QAAI,iBAAiB,MAAM;AACzB,cAAQ,SAAS,oBAAoB,eAAe,KAAK;AAAA,IAC3D;AACA,WAAO,QAAQ;AAAA,EACjB;AAAA,EACA,eAAe,UAAU,SAAS,OAAO,UAAU;AACjD,SAAK,iBAAiB,KAAK,aAAa,UAAU,KAAK;AACvD,QAAI,CAAC,KAAK,qBAAqB,KAAK,gBAAgB,KAAK,qBAAqB,OAAO,MAAM;AACzF,YAAM,MAAM,KAAK,gBAAgB,OAAO,MAAM,QAAQ,CAAC;AACvD,WAAK,oBAAoB;AACzB,cAAQ,KAAK,6BAA6B,yCAAyC;AAAA,IACrF;AACA,WAAO,KAAK,eAAe,eAAe,UAAU,SAAS,QAAQ;AAAA,EACvE;AAAA,EACA,aAAa,OAAO,OAAO;AACzB,WAAO,MAAM,KAAK,MAAM,KAAK,aAAa,gBAAgB,KAAK;AAAA,EACjE;AAAA,EACA,yBAAyB;AACvB,eAAW,CAAC,EAAE,MAAM,KAAK,OAAO,QAAQ,KAAK,WAAW,GAAG;AACzD,WAAK,iBAAiB,MAAM;AAAA,IAC9B;AAAA,EACF;AAAA,EACA,MAAM,8BAA8B;AAClC,UAAM,KAAK,CAAC;AACZ,QAAI,KAAK,MAAM,8BAA8B;AAC3C,iBAAW,CAAC,EAAE,MAAM,KAAK,OAAO,QAAQ,KAAK,WAAW,GAAG;AACzD,WAAG,KAAK,KAAK,sBAAsB,MAAM,CAAC;AAAA,MAC5C;AACA,aAAO,QAAQ,IAAI,EAAE;AAAA,IACvB,OAAO;AACL,iBAAW,CAAC,EAAE,MAAM,KAAK,OAAO,QAAQ,KAAK,WAAW,GAAG;AACzD,cAAM,KAAK,IAAI,QAAQ,CAAC,YAAY;AAClC,cAAI;AACF,iBAAK,iBAAiB,MAAM;AAC5B,oBAAQ,IAAI;AAAA,UACd,SAAS,OAAP;AACA,kBAAM;AAAA,UACR;AAAA,QACF,CAAC;AACD,WAAG,KAAK,EAAE;AAAA,MACZ;AACA,aAAO,QAAQ,IAAI,EAAE;AAAA,IACvB;AAAA,EACF;AAAA,EACA,MAAM,sBAAsB,QAAQ;AAClC,QAAI,KAAK,MAAM,GAAG,oBAAoB,OAAO,cAAc,KAAK,MAAM,6BAA6B,qBAAqB,GAAG;AACzH,aAAO,KAAK,iBAAiB,MAAM;AAAA,IACrC,OAAO;AACL,YAAM,UAAU;AAChB,aAAO,KAAK,sBAAsB,MAAM;AAAA,IAC1C;AAAA,EACF;AAAA,EACA,iBAAiB,QAAQ;AACvB,QAAI,KAAK,MAAM,GAAG,oBAAoB,OAAO,cAAc,KAAK,MAAM,GAAG,WAAW,MAAM,OAAO;AAC/F,cAAQ,IAAI,KAAK,MAAM,GAAG,kBAAkB,OAAO,YAAY,CAAC;AAChE,UAAI,KAAK,MAAM,GAAG,mBAAmB,OAAO,gBAAgB,KAAK,MAAM,GAAG,cAAc,MAAM,OAAO;AACnG,kCAA0B,OAAO,QAAQ,KAAK,MAAM,GAAG,iBAAiB,OAAO,cAAc,CAAC;AAC9F,cAAM,IAAI,MAAM,oCAAoC;AAAA,MACtD;AACA,YAAM,IAAI,MAAM,6CAA6C;AAAA,IAC/D;AACA,WAAO;AAAA,EACT;AAAA,EACA,sBAAsB;AACpB,eAAW,CAAC,EAAE,MAAM,KAAK,OAAO,QAAQ,KAAK,WAAW,GAAG;AACzD,YAAM,EAAE,kBAAkB,wBAAwB,QAAQ,QAAQ,mBAAmB,sBAAsB,kBAAkB,yBAAyB,oBAAoB,IAAI,oBAAoB,KAAK,OAAO,OAAO,SAAS,OAAO,YAAY;AACjP,aAAO,mBAAmB;AAC1B,aAAO,yBAAyB;AAChC,aAAO,SAAS;AAChB,aAAO,SAAS;AAChB,aAAO,oBAAoB;AAC3B,aAAO,uBAAuB;AAC9B,aAAO,mBAAmB;AAC1B,aAAO,0BAA0B;AACjC,aAAO,sBAAsB;AAAA,IAC/B;AAAA,EACF;AAAA,EACA,wBAAwB,QAAQ,OAAO,OAAO;AAC5C,UAAM,EAAE,SAAS,QAAQ,OAAO,SAAS,IAAI;AAC7C,UAAM,WAAW,OAAO,EAAE;AAC1B,QAAI,CAAC,SAAS,MAAM,GAAG,UAAU,OAAO,GAAG;AACzC,YAAM,IAAI,MAAM,+RAA+R;AAAA,IACjT;AACA,UAAM,SAAS,SAAS,aAAa,SAAS,OAAO,OAAO,QAAQ,OAAO,QAAQ;AACnF,WAAO,OAAO,EAAE,qBAAqB,QAAQ,OAAO,OAAO,QAAQ;AAAA,EACrE;AACF;AACA,iBAAiB,aAAa;AAC9B,SAAS,oBAAoB,GAAG,OAAO;AACrC,MAAI,UAAU,aAAa,UAAU,aAAa;AAChD,WAAO;AAAA,EACT,WAAW,UAAU,WAAW,UAAU,QAAQ;AAChD,UAAM,SAAS,UAAU,UAAU,IAAI,WAAW,EAAE,MAAM,IAAI,IAAI,WAAW,EAAE,MAAM;AACrF,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,EAAE,GAAG;AACtC,aAAO,KAAK,KAAK,MAAM,EAAE,EAAE;AAAA,IAC7B;AACA,WAAO;AAAA,EACT,OAAO;AACL,UAAM,IAAI,MAAM,iBAAiB,OAAO;AAAA,EAC1C;AACF;AAGA,IAAI,WAAW;AAGf,SAAS,iBAAiB;AACxB,MAAI,EAAE,IAAI,4BAA4B,IAAI;AAC5C;AAGA,IAAI,oBAAoB,UAAU,GAAG;AACnC,kBAAgB,SAAS,MAAM,IAAI,iBAAiB,GAAG,CAAC;AAC1D;AACA,IAAI,QAAQ,EAAE,eAAe;AAG7B,IAAI,qBAAqB;AAAA;AAAA;AAAA;AAIzB,IAAI,kBAAkB,MAAM;AAAA,EAC1B,YAAY,KAAK,QAAQ,QAAQ;AAC/B,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,cAAc,qBAAqB,2BAA2B,QAAQ,MAAM;AACjF,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASR;AACF;AAGA,IAAI,2BAA2B;AAAA;AAAA;AAAA;AAAA;AAAA;AAM/B,IAAI,wBAAwB,MAAM;AAAA,EAChC,YAAY,KAAK,QAAQ,QAAQ,mBAAmB,OAAO;AACzD,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,uBAAuB;AAC5B,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,qBAAqB,2BAA2B,QAAQ,MAAM;AACjF,UAAM,OAAO,KAAK,YAAY;AAC9B,SAAK,sBAAsB,iBAAiB,IAAI;AAChD,QAAI,yBAAyB;AAC7B,QAAI,kBAAkB;AACpB,UAAI,SAAS,KAAK,aAAa,cAAc,KAAK,WAAW,MAAM,GAAG;AACpE,iCAAyB;AAAA;AAAA;AAAA;AAAA;AAAA,MAK3B,OAAO;AACL,cAAM,QAAQ,kBAAkB,IAAI;AACpC,iCAAyB;AAAA,YACrB;AAAA;AAEJ,YAAI,SAAS,GAAG;AACd,cAAI,KAAK,qBAAqB;AAC5B,sCAA0B;AAAA;AAAA;AAAA;AAAA;AAAA,UAK5B,OAAO;AACL,sCAA0B;AAAA,yCACG,KAAK,YAAY;AAAA;AAAA;AAAA;AAAA,UAIhD;AAAA,QACF,OAAO;AACL,gBAAM,WAAW,YAAY,UAAU,IAAI;AAC3C,cAAI,KAAK,qBAAqB;AAC5B,sCAA0B;AAAA;AAAA,iBAErB,SAAS,OAAO,uBAAuB;AAAA;AAAA,iBAEvC,SAAS,OAAO,uBAAuB;AAAA;AAAA;AAAA;AAAA;AAAA,UAK9C,OAAO;AACL,sCAA0B;AAAA;AAAA,iBAErB,SAAS,OAAO,cAAc,KAAK,YAAY,OAAO;AAAA;AAAA,iBAEtD,SAAS,OAAO,cAAc,KAAK,YAAY,OAAO;AAAA;AAAA;AAAA;AAAA;AAAA,UAK7D;AAAA,QACF;AAAA,MACF;AAAA,IACF;AACA,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAQA;AAAA;AAAA;AAAA;AAAA;AAAA,EAKR;AACF;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,WAAS,OAAO,EAAE,MAAM;AACxB,SAAO,EAAE,QAAQ,EAAE,QAAQ,OAAO,EAAE,OAAO,OAAO,EAAE,MAAM;AAC5D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,OAAO,MAAM,MAAM,IAAI;AACrC,QAAM,cAAc,SAAS,eAAe,MAAM,OAAO,WAAW;AACpE,QAAM,WAAW,SAAS,QAAQ,IAAI,YAAY,MAAM;AACxD,QAAM,iBAAiB,UAAU,EAAE,QAAQ,EAAE,GAAG,MAAM,GAAG,SAAS,SAAS,CAAC;AAC5E,QAAM,iBAAiB,UAAU,EAAE,QAAQ,EAAE,GAAG,MAAM,GAAG,SAAS,SAAS,CAAC;AAC5E,WAAS,qBAAqB,EAAE,MAAM,gBAAgB,MAAM,eAAe;AAC3E,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY;AAChB,IAAI,mBAAmB;AAAA;AAAA;AAAA;AAIvB,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,SAAS,SAAS,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,OAAO,SAAS,CAAC;AACtG,QAAM,UAAU,IAAI,EAAE,QAAQ,8BAA8B,IAAI,IAAI,sBAAsB,kBAAkB,EAAE,OAAO,OAAO,KAAK,IAAI,IAAI,gBAAgB,WAAW,EAAE,OAAO,OAAO,KAAK;AACzL,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,SAAS;AACvE,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ;AACZ,IAAI,eAAe;AAAA;AAAA;AAAA;AAInB,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,GAAG,MAAM,IAAI;AACrB,QAAM,UAAU,IAAI,EAAE,QAAQ,8BAA8B,IAAI,IAAI,sBAAsB,cAAc,EAAE,OAAO,MAAM,KAAK,IAAI,IAAI,gBAAgB,OAAO,EAAE,OAAO,MAAM,KAAK;AAC/K,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,KAAK,GAAG,SAAS;AAChE;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,SAAS,iBAAiB,EAAE,WAAW,iBAAiB,eAAe,MAAM,GAAG;AAC9E,SAAO,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AACxC,UAAM,EAAE,EAAE,IAAI;AACd,UAAM,eAAe;AACrB,UAAM,SAAS,SAAS,EAAE;AAC1B,QAAI,aAAa,mBAAmB,CAAC,CAAC,CAAC,KAAK,iBAAiB,MAAM;AACjE,YAAM,QAAQ,aAAa,QAAQ,IAAI,EAAE,MAAM;AAC/C,YAAM,YAAY,cAAc,MAAM,QAAQ,MAAM;AACpD,aAAO,aAAa,eAAe,EAAE,OAAO,QAAQ,SAAS;AAAA,IAC/D;AACA,UAAM,yBAAyB,IAAI,EAAE,QAAQ,6BAA6B,KAAK,mBAAmB;AAClG,QAAI;AACJ,QAAI,wBAAwB;AAC1B,gBAAU,IAAI,qBAAqB,EAAE,OAAO,eAAe;AAAA,IAC7D,OAAO;AACL,gBAAU,IAAI,eAAe,EAAE,OAAO,SAAS;AAAA,IACjD;AACA,WAAO,aAAa,gBAAgB,SAAS,CAAC,CAAC,GAAG,MAAM;AAAA,EAC1D;AACF;AACA,SAAS,kBAAkB,EAAE,WAAW,iBAAiB,mBAAmB,OAAO,kBAAkB,OAAO,eAAe,MAAM,GAAG;AAClI,SAAO,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AACxC,UAAM,EAAE,GAAG,EAAE,IAAI;AACjB,UAAM,eAAe;AACrB,QAAI,mBAAmB,EAAE,UAAU,aAAa;AAC9C,YAAM,QAAQ,aAAa,QAAQ,IAAI,EAAE,MAAM;AAC/C,YAAM,QAAQ,aAAa,QAAQ,IAAI,EAAE,MAAM;AAC/C,YAAM,CAAC,OAAO,KAAK,IAAI;AAAA,QACrB,CAAC,MAAM,mBAAmB,MAAM,MAAM,mBAAmB,IAAI;AAAA,QAC7D,CAAC,MAAM,mBAAmB,MAAM,MAAM,mBAAmB,IAAI;AAAA,MAC/D,EAAE,IAAI,CAAC,iBAAiB;AACtB,cAAM,CAAC,OAAO,KAAK,IAAI;AACvB,cAAM,UAAU;AAAA,UACd,QAAQ,MAAM;AAAA,UACd,OAAO,MAAM;AAAA,UACb,OAAO,EAAE;AAAA,QACX;AACA,cAAM,UAAU;AAAA,UACd,QAAQ,MAAM;AAAA,UACd,OAAO,MAAM;AAAA,UACb,OAAO,EAAE;AAAA,QACX;AACA,cAAM,WAAW,IAAI,gBAAgB,WAAW,EAAE,OAAO,EAAE,KAAK;AAChE,eAAO,aAAa,gBAAgB,UAAU,CAAC,SAAS,OAAO,GAAG,WAAW,MAAM,OAAO,MAAM,KAAK,CAAC;AAAA,MACxG,CAAC;AACD,YAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,MAAM,OAAO,MAAM,MAAM,GAAG,SAAS,aAAa,CAAC;AAC9F,mBAAa,8BAA8B,KAAK;AAChD,mBAAa,8BAA8B,KAAK;AAChD,aAAO;AAAA,IACT;AACA,UAAM,SAAS,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AACnD,SAAK,EAAE,UAAU,YAAY,EAAE,UAAU,YAAY,aAAa,mBAAmB,CAAC,GAAG,CAAC,CAAC,MAAM,iBAAiB,MAAM;AACtH,YAAM,QAAQ,aAAa,QAAQ,IAAI,EAAE,MAAM,EAAE;AACjD,YAAM,QAAQ,aAAa,QAAQ,IAAI,EAAE,MAAM,EAAE;AACjD,YAAM,eAAe,EAAE,UAAU,WAAW,qBAAqB,uBAAuB,KAAK,IAAI;AACjG,YAAM,eAAe,EAAE,UAAU,WAAW,qBAAqB,uBAAuB,KAAK,IAAI;AACjG,YAAM,CAAC,WAAW,QAAQ,IAAI,cAAc,EAAE,OAAO,EAAE,OAAO,cAAc,cAAc,MAAM;AAChG,YAAM,MAAM,aAAa,eAAe,UAAU,MAAM;AACxD,YAAM,UAAU,aAAa,QAAQ,IAAI,IAAI,MAAM;AACnD,cAAQ,SAAS;AACjB,aAAO;AAAA,IACT;AACA,UAAM,yBAAyB,IAAI,EAAE,QAAQ,8BAA8B,KAAK,mBAAmB;AACnG,QAAI;AACJ,QAAI,wBAAwB;AAC1B,gBAAU,IAAI,sBAAsB,iBAAiB,EAAE,OAAO,EAAE,OAAO,gBAAgB;AAAA,IACzF,OAAO;AACL,gBAAU,IAAI,gBAAgB,WAAW,EAAE,OAAO,EAAE,KAAK;AAAA,IAC3D;AACA,WAAO,aAAa,gBAAgB,SAAS,CAAC,GAAG,CAAC,GAAG,MAAM;AAAA,EAC7D;AACF;AACA,SAAS,6BAA6B,aAAa,SAAS,OAAO;AACjE,MAAI,gBAAgB,UAAU;AAC5B,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,QAAQ;AACjC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,OAAO;AAChC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,SAAS;AAClC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,SAAS;AAClC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,aAAa;AACtC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT,WAAW,gBAAgB,WAAW;AACpC,QAAI,QAAQ;AACV,aAAO;AAAA,IACT;AACA,WAAO;AAAA,EACT;AACA,QAAM,IAAI,MAAM,cAAc,6DAA6D;AAC7F;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,QAAQ,QAAQ,aAAa,aAAa,OAAO,aAAa,OAAO,UAAU,OAAO,cAAc,MAAM,qBAAqB,OAAO,yBAAyB,OAAO;AAChL,SAAK,gBAAgB,CAAC,WAAW,SAAS;AAC1C,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,YAAY,aAAa,OAAO,KAAK,OAAO;AAClD,UAAM,wBAAwB,KAAK,KAAK,YAAY,CAAC;AACrD,UAAM,UAAU,aAAa,gBAAgB;AAC7C,UAAM,UAAU,aAAa,gBAAgB;AAC7C,UAAM,WAAW,aAAa,CAAC,UAAU,QAAQ,IAAI,CAAC,UAAU,QAAQ;AACxE,UAAM,WAAW,aAAa,CAAC,UAAU,QAAQ,IAAI,CAAC,UAAU,QAAQ;AACxE,QAAI,oBAAoB,IAAI,yBAAyB;AACrD,QAAI,aAAa;AACf,UAAI,oBAAoB;AACtB,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,WAAW,wBAAwB;AACjC,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,OAAO;AACL,4BAAoB;AAAA,YAChB;AAAA;AAAA,MAEN;AACA,+BAAyB;AAAA,IAC3B;AACA,UAAM,iBAAiB,UAAU,oCAAoC;AACrE,QAAI,SAAS;AACX,WAAK,cAAc,KAAK,MAAM;AAAA,IAChC;AACA,QAAI,oBAAoB;AACtB,WAAK,cAAc,KAAK,wBAAwB;AAAA,IAClD;AACA,QAAI,wBAAwB;AAC1B,WAAK,cAAc,KAAK,gBAAgB;AAAA,IAC1C;AACA,QAAI,gBAAgB;AACpB,QAAI,gBAAgB;AACpB,QAAI,OAAO,KAAK,OAAO,IAAI;AACzB,sBAAgB,wBAAwB,OAAO,KAAK;AAAA,IACtD,WAAW,OAAO,KAAK,OAAO,IAAI;AAChC,sBAAgB,wBAAwB,OAAO,KAAK;AAAA,IACtD;AACA,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA,sCAE8B;AAAA;AAAA;AAAA;AAAA,8BAIR;AAAA,yBACL;AAAA,yBACA;AAAA,wCACe;AAAA,wCACA;AAAA;AAAA;AAAA;AAAA,uBAIjB,SAAS,QAAQ,SAAS;AAAA,uBAC1B,SAAS,QAAQ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UASvC;AAAA;AAAA,UAEA;AAAA;AAAA;AAAA;AAAA;AAAA,EAKR;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,MAAM;AAAA,EACN,MAAM;AACR;AACA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,KAAK,QAAQ,QAAQ;AAC/B,SAAK,gBAAgB,CAAC,SAAS,SAAS,SAAS,OAAO;AACxD,SAAK,cAAc,qBAAqB,2BAA2B,QAAQ,MAAM;AACjF,SAAK,WAAW;AAAA;AAAA;AAAA,UAGV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWR;AACF;AAGA,IAAI,MAAM;AACV,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,GAAG,EAAE,IAAI;AACjB,QAAM,QAAQ,qBAAqB,WAAW,EAAE,OAAO,EAAE,KAAK;AAC9D,MAAI,EAAE,UAAU,aAAa;AAC3B,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,cAAc,IAAI,uBAAuB,iBAAiB,MAAM,EAAE,OAAO,EAAE,KAAK;AACtF,UAAM,cAAc,IAAI,uBAAuB,iBAAiB,MAAM,EAAE,OAAO,EAAE,KAAK;AACtF,UAAM,UAAU;AAAA,MACd;AAAA,QACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,QACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,QACrC,OAAO,EAAE;AAAA,MACX;AAAA,MACA;AAAA,QACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,QACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,QACrC,OAAO,EAAE;AAAA,MACX;AAAA,MACA;AAAA,QACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,QACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,QACrC,OAAO,EAAE;AAAA,MACX;AAAA,MACA;AAAA,QACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,QACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,QACrC,OAAO,EAAE;AAAA,MACX;AAAA,IACF;AACA,UAAM,WAAW,SAAS,gBAAgB,aAAa,SAAS,SAAS;AACzE,UAAM,WAAW,SAAS,gBAAgB,aAAa,SAAS,SAAS;AACzE,UAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,MAAM,UAAU,MAAM,SAAS,GAAG,SAAS,SAAS,CAAC;AAChG,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,QAAQ;AAC/C,WAAO;AAAA,EACT;AACA,MAAI,SAAS,mBAAmB,CAAC,GAAG,CAAC,CAAC,GAAG;AACvC,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,CAAC,WAAW,QAAQ,IAAI,gBAAgB,EAAE,OAAO,EAAE,OAAO,MAAM,QAAQ,MAAM,QAAQ,KAAK;AACjG,UAAM,MAAM,SAAS,eAAe,UAAU,KAAK;AACnD,UAAM,UAAU,SAAS,QAAQ,IAAI,IAAI,MAAM;AAC/C,YAAQ,SAAS;AACjB,WAAO;AAAA,EACT;AACA,MAAI;AACJ,MAAI,IAAI,EAAE,QAAQ,8BAA8B,GAAG;AACjD,cAAU,IAAI,sBAAsB,KAAK,EAAE,OAAO,EAAE,KAAK;AAAA,EAC3D,OAAO;AACL,cAAU,IAAI,gBAAgB,KAAK,EAAE,OAAO,EAAE,KAAK;AAAA,EACrD;AACA,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,CAAC,GAAG,KAAK;AACxD;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,QAAQ,YAAY,UAAU;AACnD,QAAM,eAAe;AAAA,IACnB,YAAY,OAAO,KAAK;AAAA,IACxB,GAAG,YAAY,OAAO,KAAK;AAAA,EAC7B;AACA,QAAM,UAAU;AAAA,IACd,OAAO,OAAO;AAAA,IACd,OAAO;AAAA,IACP,QAAQ,OAAO;AAAA,EACjB;AACA,QAAM,iBAAiB;AAAA,IACrB,YAAY,UAAU;AAAA,IACtB,GAAG,YAAY,UAAU;AAAA,EAC3B;AACA,QAAM,UAAU,IAAI,qBAAqB,gBAAgB,YAAY;AACrE,QAAM,gCAAgC;AACtC,QAAM,eAAe,CAAC,YAAY;AAClC,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,OAAO,GAAG,OAAO,OAAO,cAAc,6BAA6B;AACrH,SAAO,EAAE,QAAQ,OAAO,QAAQ,OAAO,YAAY,OAAO,OAAO,MAAM;AACzE;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,eAAe;AACrB,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,SAAS,aAAa,uBAAuB,OAAO,KAAK;AAC/D,QAAM,SAAS,aAAa,cAAc,MAAM;AAChD,eAAa,OAAO,UAAU,QAAQ,MAAM,kBAAkB,eAAe,sCAAsC,EAAE,cAAc,oFAAoF;AACvN,QAAM,WAAW,aAAa,QAAQ,IAAI,EAAE,MAAM;AAClD,MAAI,SAAS,YAAY,CAAC,cAAc,EAAE,OAAO,MAAM,KAAK,EAAE,SAAS,YAAY,QAAQ,cAAc,SAAS,OAAO,MAAM,IAAI;AACjI,WAAO,cAAc,GAAG,QAAQ,YAAY;AAAA,EAC9C;AACA,eAAa,OAAO,EAAE,MAAM;AAC5B,SAAO,EAAE,QAAQ,EAAE,QAAQ,OAAO,QAAQ,OAAO,EAAE,MAAM;AAC3D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,YAAY,SAAS;AAC/B,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,EAAE,YAAY,WAAW,QAAQ,QAAQ,IAAI;AACnD,SAAK,cAAc,CAAC,WAAW,OAAO;AACtC,UAAM,wBAAwB,KAAK,MAAM,aAAa,CAAC,IAAI;AAC3D,UAAM,0BAA0B,aAAa;AAC7C,QAAI,gBAAgB;AACpB,QAAI,WAAW,MAAM;AACnB,YAAM,cAAc,IAAI;AACxB,sBAAgB,4BAA4B,aAAa,MAAM,WAAW,IAAI,YAAY,YAAY,CAAC,IAAI;AAAA,IAC7G;AACA,QAAI,mBAAmB;AACvB,QAAI,SAAS,aAAa,GAAG;AAC3B,yBAAmB;AAAA,oCACW;AAAA;AAAA;AAAA;AAAA,IAIhC;AACA,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA,UAIV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAQwB;AAAA;AAAA;AAAA;AAAA,8BAIJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YASlB;AAAA;AAAA;AAAA,iCAGqB;AAAA,cACnB,4BAA4B;AAAA;AAAA;AAAA,YAG9B;AAAA,qBACS,4BAA4B;AAAA;AAAA;AAAA;AAAA;AAAA,YAKrC;AAAA,qBACS,4BAA4B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAMrC;AAAA;AAAA;AAAA;AAAA;AAAA,EAKV;AACF;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,YAAY,YAAY;AAClC,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,EAAE,YAAY,WAAW,QAAQ,QAAQ,IAAI;AACnD,SAAK,cAAc,CAAC,WAAW,OAAO;AACtC,QAAI,sBAAsB;AAC1B,QAAI,YAAY;AAChB,QAAI,eAAe,QAAQ;AACzB,4BAAsB;AAAA,IACxB,WAAW,eAAe,OAAO;AAC/B,4BAAsB;AACtB,kBAAY;AAAA,IACd,WAAW,eAAe,OAAO;AAC/B,4BAAsB;AACtB,kBAAY;AAAA,IACd;AACA,QAAI,cAAc,GAAG,cAAc,cAAc;AACjD,QAAI,eAAe,OAAO;AACxB,oBAAc;AAAA,IAChB,WAAW,eAAe,QAAQ;AAChC,oBAAc;AAAA,IAChB,WAAW,eAAe,OAAO;AAC/B,oBAAc;AAAA,IAChB,WAAW,eAAe,OAAO;AAC/B,oBAAc;AAAA,IAChB;AACA,UAAM,wBAAwB,KAAK,MAAM,aAAa,CAAC,IAAI;AAC3D,UAAM,0BAA0B,aAAa;AAC7C,QAAI,gBAAgB;AAAA,YACZ,eAAe;AAAA;AAAA,mBAER,eAAe;AAAA;AAAA;AAAA;AAAA,wBAIV;AAAA,cACV,eAAe,YAAY,eAAe;AAAA,0BAC9B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAQtB,QAAI,UAAU;AACd,QAAI,eAAe,OAAO;AACxB,4BAAsB;AACtB,sBAAgB;AAAA;AAAA;AAAA;AAAA;AAKhB,gBAAU;AAAA,IACZ,WAAW,eAAe,OAAO;AAC/B,4BAAsB;AACtB,sBAAgB;AAAA;AAAA;AAAA;AAAA;AAKhB,gBAAU;AAAA,IACZ;AACA,QAAI,mBAAmB;AACvB,QAAI,SAAS,aAAa,GAAG;AAC3B,yBAAmB;AAAA,oCACW;AAAA;AAAA;AAAA;AAAA,IAIhC;AACA,SAAK,WAAW;AAAA,0CACsB;AAAA;AAAA;AAAA;AAAA,UAIhC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAQwB;AAAA;AAAA,kCAEA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMJ;AAAA;AAAA,YAElB,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOpB;AAAA;AAAA;AAAA,iCAGqB;AAAA,cACnB,4BAA4B;AAAA,YAC9B,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOpB;AAAA,qBACS,4BAA4B;AAAA,YACrC,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOpB;AAAA,qBACS,4BAA4B;AAAA,YACrC,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAOpB;AAAA;AAAA,oBAEQ;AAAA;AAAA;AAAA,EAGlB;AACF;AAGA,SAAS,mBAAmB,SAAS;AACnC,QAAM,SAAS,CAAC;AAChB,SAAO,OAAO,WAAW,KAAK,OAAO,OAAO,SAAS,GAAG,YAAY,GAAG;AACrE,UAAM,UAAU,OAAO,SAAS,OAAO,OAAO,SAAS,GAAG,UAAU,QAAQ;AAC5E,UAAM,aAAa,qBAAqB,yBAAyB,OAAO;AACxE,WAAO,KAAK;AAAA,MACV,QAAQ;AAAA,MACR;AAAA,MACA,SAAS,KAAK,KAAK,UAAU,UAAU;AAAA,IACzC,CAAC;AAAA,EACH;AACA,SAAO;AACT;AACA,SAAS,OAAO,GAAG,OAAO,eAAe,UAAU;AACjD,QAAM,kBAAkB,mBAAmB,EAAE,KAAK;AAClD,MAAI,SAAS;AACb,WAAS,IAAI,GAAG,IAAI,gBAAgB,QAAQ,KAAK;AAC/C,UAAM,EAAE,QAAQ,YAAY,QAAQ,IAAI,gBAAgB;AACxD,QAAI;AACJ,QAAI;AACJ,QAAI,kBAAkB,QAAQ;AAC5B,gBAAU,MAAM,IAAI,IAAI,YAAY,EAAE,YAAY,QAAQ,WAAW,EAAE,MAAM,IAAI,QAAQ,GAAG,MAAM,IAAI,IAAI,YAAY,EAAE,YAAY,QAAQ,WAAW,EAAE,MAAM,IAAI,QAAQ,CAAC;AAAA,IAC9K,OAAO;AACL,gBAAU,IAAI,cAAc,EAAE,YAAY,QAAQ,WAAW,EAAE,MAAM,IAAI,QAAQ,GAAG,aAAa;AAAA,IACnG;AACA,qBAAiB;AACjB,aAAS,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,KAAK;AAC1D,QAAI,eAAe,WAAW,EAAE,QAAQ;AACtC,eAAS,8BAA8B,cAAc;AAAA,IACvD;AAAA,EACF;AACA,SAAO;AACT;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,QAAQ,QAAQ;AAC1B,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,cAAc,IAAI,MAAM,OAAO,MAAM;AAC3C,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,kBAAY,KAAK,OAAO,OAAO;AAAA,IACjC;AACA,SAAK,cAAc;AACnB,SAAK,OAAO,YAAY;AACxB,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,UAAM,WAAW,kBAAkB,MAAM;AACzC,SAAK,WAAW;AAAA;AAAA,QAEZ;AAAA,uBACe;AAAA;AAAA;AAAA,EAGrB;AACF;AACA,SAAS,kBAAkB,QAAQ;AACjC,QAAM,OAAO,OAAO;AACpB,MAAI,OAAO,GAAG;AACZ,UAAM,MAAM,sBAAsB,2BAA2B;AAAA,EAC/D;AACA,QAAM,gBAAgB,CAAC,WAAW,WAAW,WAAW,WAAW,WAAW,SAAS;AACvF,QAAM,iBAAiB,IAAI,MAAM,IAAI;AACrC,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,mBAAe,OAAO,MAAM,cAAc;AAAA,EAC5C;AACA,SAAO,eAAe,KAAK;AAC7B;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,QAAQ,QAAQ;AAC1B,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,UAAM,cAAc,IAAI,MAAM,OAAO,MAAM;AAC3C,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,kBAAY,KAAK,OAAO,OAAO;AAAA,IACjC;AACA,SAAK,cAAc;AACnB,SAAK,OAAO,YAAY;AACxB,QAAI,KAAK,OAAO,GAAG;AACjB,YAAM,MAAM,6BAA6B,KAAK,4BAA4B;AAAA,IAC5E;AACA,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,UAAM,cAAc,eAAe,MAAM,KAAK,IAAI;AAClD,UAAM,gBAAgB,IAAI,MAAM,KAAK,IAAI;AACzC,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,oBAAc,OAAO,MAAM,YAAY;AAAA,IACzC;AACA,UAAM,YAAY,QAAQ,cAAc,MAAM,EAAE,EAAE,KAAK;AACvD,UAAM,aAAa,KAAK,YAAY,KAAK,OAAO,QAAQ,YAAY,KAAK,OAAO;AAChF,UAAM,OAAO,mBAAmB,cAAc,KAAK,OAAO;AAC1D,SAAK,WAAW;AAAA;AAAA,QAEZ;AAAA;AAAA,oBAEY;AAAA,WACT;AAAA,sBACW;AAAA;AAAA,UAEZ,YAAY,KAAK,OAAO;AAAA,aACrB,YAAY,KAAK,OAAO,QAAQ,YAAY,KAAK,OAAO;AAAA,sBAC/C;AAAA,aACT;AAAA,wBACW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMtB;AACF;AAGA,SAAS,eAAe,GAAG,MAAM,UAAU;AACzC,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,uBAAuB,EAAE,OAAO,IAAI,IAAI,IAAI,iBAAiB,EAAE,OAAO,IAAI;AAC7I,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AAGA,SAAS,QAAQ,GAAG,MAAM,UAAU,UAAU;AAC5C,QAAM,mBAAmB;AACzB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,kBAAkB,EAAE,KAAK;AACtE,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,QAAM,uBAAuB,gBAAgB;AAC7C,MAAI,WAAW;AACf,MAAI,sBAAsB;AACxB,eAAW,eAAe,GAAG,cAAc,QAAQ;AACnD,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAAA,EACjE;AACA,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,aAAa,WAAW,IAAI,qBAAqB,0BAA0B,SAAS,OAAO,IAAI;AACtG,MAAI,WAAW;AACf,MAAI,UAAU;AACZ,eAAW,qBAAqB,qBAAqB,aAAa,QAAQ;AAAA,EAC5E;AACA,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,YAAY,QAAQ;AAC1B,QAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,OAAO,EAAE,OAAO,CAAC,WAAW,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AACpH,QAAM,UAAU,WAAW,EAAE,KAAK;AAClC,QAAM,UAAU,OAAO,eAAe,SAAS,OAAO,QAAQ;AAC9D,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AAC9F,WAAS,8BAA8B,aAAa;AACpD,WAAS,8BAA8B,OAAO;AAC9C,MAAI,sBAAsB;AACxB,aAAS,8BAA8B,QAAQ;AAAA,EACjD;AACA,SAAO;AACT;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,SAAO,QAAQ,GAAG,MAAM,UAAU,QAAQ;AAC5C;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,eAAe;AACrB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,IAAI,MAAM,KAAK;AAChC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAS,KAAK,EAAE,MAAM,KAAK;AAAA,EAC7B;AACA,MAAI;AACJ,MAAI,aAAa,mBAAmB,CAAC,CAAC,CAAC,GAAG;AACxC,UAAM,WAAW,aAAa,QAAQ,IAAI,EAAE,MAAM;AAClD,UAAM,SAAS,SAAS;AACxB,UAAM,YAAY,iBAAiB,QAAQ,EAAE,OAAO,EAAE,OAAO,MAAM,QAAQ;AAC3E,UAAM,aAAa,eAAe,UAAU,EAAE,KAAK;AACnD,UAAM,UAAU,aAAa,QAAQ,IAAI,IAAI,MAAM;AACnD,YAAQ,SAAS;AAAA,EACnB,OAAO;AACL,UAAM,eAAe,GAAG,MAAM,YAAY;AAAA,EAC5C;AACA,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,8BAA8B;AAClC,SAAS,gBAAgB,EAAE,GAAG,GAAG,YAAY,YAAY,SAAS,UAAU,OAAO,MAAM,yBAAyB,MAAM,iBAAiB,GAAG,YAAY,cAAc,KAAK,GAAG;AAC5K,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,oBAAoB,uBAAuB,2BAA2B,EAAE,MAAM,MAAM,GAAG,EAAE,GAAG,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC;AACtH,QAAM,WAAW,kBAAkB,OAAO,CAAC,aAAa,WAAW,CAAC;AACpE,eAAa,OAAO,gBAAgB,aAAa,MAAM,kCAAkC,qBAAqB,uCAAuC,EAAE,aAAa,EAAE,wBAAwB,6BAA6B,wBAAwB;AACnP,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,gBAAgB,CAAC,KAAK,GAAG;AAC/B,QAAM,WAAW,KAAK,IAAI,WAAW,SAAS;AAC9C,QAAM,YAAY,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACxD,QAAM,UAAU,QAAQ;AACxB,QAAM,4BAA4B,0BAA0B;AAC5D,QAAM,oBAAoB,gBAAgB;AAC1C,QAAM,kBAAkB,eAAe,OAAO,6BAA6B,aAAa,IAAI,IAAI;AAChG,QAAM,mBAAmB,WAAW,6BAA6B,qBAAqB,mBAAmB;AACzG,MAAI;AACJ,OAAK,gBAAgB,KAAK,gBAAgB,MAAM,YAAY,+BAA+B,qBAAqB,OAAO;AACrH,QAAI,OAAO;AACX,QAAI,OAAO;AACX,QAAI,YAAY;AACd,aAAO,WAAW,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,CAAC,GAAG,GAAG,CAAC,EAAE,EAAE,CAAC;AACvF,oBAAc,KAAK,IAAI;AAAA,IACzB;AACA,QAAI,YAAY;AACd,aAAO,WAAW,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,CAAC,GAAG,GAAG,CAAC,EAAE,EAAE,CAAC;AACvF,oBAAc,KAAK,IAAI;AAAA,IACzB;AACA,UAAM,iBAAiB,gBAAgB;AACvC,UAAM,iBAAiB,gBAAgB;AACvC,QAAI,SAAS;AACb,QAAI,gBAAgB;AAClB,eAAS,SAAS;AAAA,QAChB,QAAQ,EAAE,GAAG,KAAK;AAAA,QAClB,SAAS;AAAA,QACT,OAAO,EAAE,OAAO,CAAC,UAAU,WAAW,CAAC,EAAE;AAAA,MAC3C,CAAC;AACD,oBAAc,KAAK,MAAM;AAAA,IAC3B;AACA,UAAM,OAAO,gBAAgB,IAAI,IAAI;AACrC,QAAI,SAAS;AACb,QAAI,gBAAgB;AAClB,eAAS,SAAS;AAAA,QAChB,QAAQ,EAAE,GAAG,KAAK;AAAA,QAClB,SAAS;AAAA,QACT,OAAO,EAAE,OAAO,CAAC,UAAU,GAAG,SAAS,EAAE;AAAA,MAC3C,CAAC;AACD,oBAAc,KAAK,MAAM;AAAA,IAC3B;AACA,UAAM,UAAU,UAAU,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,OAAO,GAAG,SAAS,SAAS,CAAC;AACjF,UAAM,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,UAAU,KAAK,EAAE,CAAC;AACzF,kBAAc,KAAK,OAAO;AAAA,EAC5B,OAAO;AACL,UAAM,QAAQ,WAAW,EAAE,OAAO,EAAE,KAAK;AACzC,UAAM,UAAU,IAAI,oBAAoB,UAAU,UAAU,CAAC,UAAU,aAAa,WAAW,GAAG,YAAY,YAAY,SAAS,iBAAiB,2BAA2B,iBAAiB;AAChM,UAAM,SAAS,CAAC,KAAK,GAAG;AACxB,QAAI,QAAQ,MAAM;AAChB,aAAO,KAAK,IAAI;AAAA,IAClB;AACA,QAAI,2BAA2B;AAC7B,aAAO,KAAK,sBAAsB;AAAA,IACpC;AACA,QAAI,mBAAmB;AACrB,YAAM,kBAAkB,SAAS,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,gBAAgB,SAAS,CAAC;AACxH,aAAO,KAAK,eAAe;AAC3B,oBAAc,KAAK,eAAe;AAAA,IACpC;AACA,UAAM,SAAS,gBAAgB,SAAS,QAAQ,KAAK;AAAA,EACvD;AACA,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAClG,gBAAc,KAAK,GAAG;AACtB,aAAW,KAAK,eAAe;AAC7B,aAAS,8BAA8B,CAAC;AAAA,EAC1C;AACA,SAAO;AACT;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,MAAM,uBAAuB,IAAI;AAC/C,QAAM,EAAE,YAAY,YAAY,YAAY,aAAa,eAAe,IAAI;AAC5E,SAAO,gBAAgB;AAAA,IACrB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA,YAAY;AAAA,EACd,CAAC;AACH;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AACX,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,SAAS,mBAAmB,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,aAAa;AAC/D,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,YAAY,iBAAiB,MAAM,MAAM;AAC/C,WAAO,SAAS,eAAe,EAAE,OAAO,EAAE,OAAO,SAAS;AAAA,EAC5D;AACA,MAAI;AACJ,MAAI,IAAI,EAAE,QAAQ,6BAA6B,GAAG;AAChD,cAAU,IAAI,qBAAqB,EAAE,OAAO,IAAI;AAAA,EAClD,OAAO;AACL,cAAU,IAAI,eAAe,EAAE,OAAO,IAAI;AAAA,EAC5C;AACA,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAM/B,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,oBAAoB;AAAA;AAAA;AAGhC,IAAI,SAAS,iBAAiB,EAAE,WAAW,MAAM,CAAC;AAClD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AACV,IAAI,gBAAgB,kBAAkB;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EACjB,eAAe;AACjB,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,aAAa,QAAQ;AAC/B,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc;AACnB,SAAK,gBAAgB,OAAO,IAAI,CAAC,GAAG,MAAM,IAAI,GAAG;AACjD,UAAM,WAAW,CAAC;AAClB,SAAK,cAAc,QAAQ,CAAC,cAAc;AACxC,eAAS,KAAK,UAAU,kBAAkB,yBAAyB;AAAA,IACrE,CAAC;AACD,UAAM,YAAY,KAAK,cAAc,IAAI,CAAC,cAAc;AACtD,aAAO,IAAI;AAAA,IACb,CAAC,EAAE,KAAK,KAAK;AACb,SAAK,WAAW;AAAA;AAAA,UAEV,SAAS,KAAK,YAAY;AAAA;AAAA,yBAEX;AAAA;AAAA;AAAA;AAAA,EAIvB;AACF;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,aAAa,QAAQ;AAC/B,SAAK,cAAc,CAAC;AACpB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,gBAAgB,OAAO,IAAI,CAAC,GAAG,MAAM,IAAI,GAAG;AACjD,UAAM,WAAW,CAAC;AAClB,SAAK,cAAc,QAAQ,CAAC,cAAc;AACxC,eAAS,KAAK,SAAS,kBAAkB,yBAAyB;AAAA,IACpE,CAAC;AACD,UAAM,YAAY,KAAK,cAAc,IAAI,CAAC,cAAc;AACtD,aAAO,IAAI;AAAA,IACb,CAAC,EAAE,KAAK,KAAK;AACb,SAAK,WAAW;AAAA;AAAA,UAEV,SAAS,KAAK,YAAY;AAAA;AAAA,wBAEZ;AAAA;AAAA;AAAA;AAAA,EAItB;AACF;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,UAAU;AAChB,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,GAAG,SAAS,SAAS,CAAC;AAAA,EACnE;AACA,MAAI,QAAQ,SAAS,IAAI,EAAE,IAAI,8BAA8B,GAAG;AAC9D,UAAM,WAAW,KAAK,MAAM,QAAQ,SAAS,CAAC;AAC9C,UAAM,WAAW,MAAM,EAAE,QAAQ,QAAQ,MAAM,GAAG,QAAQ,GAAG,SAAS,SAAS,CAAC;AAChF,UAAM,YAAY,MAAM,EAAE,QAAQ,QAAQ,MAAM,QAAQ,GAAG,SAAS,SAAS,CAAC;AAC9E,WAAO,MAAM,EAAE,QAAQ,CAAC,UAAU,SAAS,GAAG,SAAS,SAAS,CAAC;AAAA,EACnE;AACA,QAAM,QAAQ,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,IAAI,OAAO,WAAW,IAAI,EAAE,CAAC;AAC/E,QAAM,SAAS,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK;AACzC,QAAM,cAAc,IAAI,EAAE,QAAQ,YAAY;AAC9C,QAAM,UAAU,cAAc,IAAI,kBAAkB,QAAQ,GAAG,OAAO,MAAM,IAAI,IAAI,YAAY,QAAQ,GAAG,OAAO,MAAM;AACxH,SAAO,SAAS,gBAAgB,SAAS,SAAS,KAAK;AACzD;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AAC1F,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAAA,EACjE;AACA,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACpG,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,QAAM,UAAU,OAAO,KAAK,IAAI,OAAO,OAAO,QAAQ;AACtD,MAAI;AACJ,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,UAAU,QAAQ;AAC7E,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F,OAAO;AACL,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F;AACA,WAAS,8BAA8B,GAAG;AAC1C,WAAS,8BAA8B,OAAO;AAC9C,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AAC1F,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAAA,EACjE;AACA,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACpG,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,QAAM,UAAU,OAAO,KAAK,IAAI,OAAO,OAAO,QAAQ;AACtD,MAAI;AACJ,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,UAAU,QAAQ;AAC7E,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F,OAAO;AACL,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F;AACA,WAAS,8BAA8B,GAAG;AAC1C,WAAS,8BAA8B,OAAO;AAC9C,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,YAAY,KAAK,WAAW;AACtC,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,EAAE,YAAY,WAAW,QAAQ,IAAI;AAC3C,QAAI,CAAC,WAAW;AACd,WAAK,cAAc,KAAK,cAAc;AAAA,IACxC;AACA,SAAK,cAAc,CAAC,WAAW,OAAO;AACtC,UAAM,SAAS,QAAQ,QAAQ,MAAM;AACrC,UAAM,eAAe,YAAY,kBAAkB;AACnD,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKc;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKJ;AAAA,wBACN;AAAA;AAAA,0BAEE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQxB;AACF;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,OAAO,YAAY,KAAK,WAAW;AAC7C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,iBAAa,OAAO,MAAM,SAAS,GAAG,MAAM,aAAa,IAAI,OAAO,CAAC,EAAE,YAAY,IAAI,IAAI,MAAM,CAAC,2CAA2C;AAC7I,UAAM,SAAS,MAAM,MAAM,SAAS;AACpC,UAAM,UAAU,KAAK,KAAK,SAAS,UAAU;AAC7C,SAAK,cAAc,MAAM,MAAM,GAAG,EAAE;AACpC,QAAI,UAAU,GAAG;AACf,WAAK,YAAY,KAAK,OAAO;AAAA,IAC/B;AACA,QAAI,CAAC,WAAW;AACd,WAAK,cAAc,KAAK,cAAc;AAAA,IACxC;AACA,UAAM,WAAW,KAAK;AACtB,UAAM,OAAO,SAAS;AACtB,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,UAAU,YAAY,UAAU,IAAI;AAC1C,QAAI;AACJ,QAAI;AACJ,QAAI,YAAY,GAAG;AACjB,mBAAa,OAAO;AACpB,YAAM,iBAAiB,kBAAkB,UAAU;AACnD,uBAAiB;AAAA,UACb,+BAA+B,kBAAkB,QAAQ,KAAK;AAAA,YAC5D,QAAQ,OAAO;AAAA,UACjB,+BAA+B,kBAAkB,QAAQ,KAAK;AAAA,YAC5D,QAAQ,OAAO;AAAA,UACjB,+BAA+B,kBAAkB,QAAQ,KAAK;AAAA,YAC5D,QAAQ,OAAO;AAAA,UACjB,+BAA+B,kBAAkB,QAAQ,KAAK;AAAA,YAC5D,QAAQ,OAAO;AAAA,IACvB,OAAO;AACL,mBAAa;AACb,uBAAiB;AAAA,UACb;AAAA,YACE,QAAQ,OAAO;AAAA,UACjB;AAAA,YACE,QAAQ,OAAO;AAAA,UACjB;AAAA,YACE,QAAQ,OAAO;AAAA,UACjB;AAAA,YACE,QAAQ,OAAO;AAAA,IACvB;AACA,UAAM,WAAW,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG,EAAE,MAAM,GAAG,UAAU;AACnE,UAAM,YAAY,MAAM,SAAS,aAAa;AAC9C,UAAM,cAAc,SAAS,IAAI,CAAC,MAAM,SAAS,CAAC;AAClD,UAAM,aAAa,YAAY,cAAc,aAAa,CAAC,EAAE,OAAO,SAAS;AAC7E,UAAM,aAAa,YAAY,cAAc,aAAa,CAAC,EAAE,OAAO,SAAS;AAC7E,UAAM,aAAa,YAAY,cAAc,aAAa,CAAC,EAAE,OAAO,SAAS;AAC7E,UAAM,aAAa,YAAY,cAAc,aAAa,CAAC,EAAE,OAAO,SAAS;AAC7E,UAAM,SAAS,QAAQ,QAAQ,gBAAgB;AAC/C,UAAM,oBAAoB,YAAY,KAAK;AAAA,sDACO,WAAW,KAAK;AAAA,sDAChB,WAAW,KAAK;AAAA,sDAChB,WAAW,KAAK;AAAA,sDAChB,WAAW,KAAK;AAClE,UAAM,aAAa;AAAA,0BACG,WAAW,KAAK;AAAA,uCACH,WAAW,KAAK;AAAA,uCAChB,WAAW,KAAK;AAAA,qDACF,WAAW,KAAK;AACjE,UAAM,gCAAgC,YAAY,KAAK;AAAA,qCACtB,YAAY,KAAK;AAAA,4CACV,SAAS,KAAK;AAAA,iDACT,SAAS,MAAM,EAAE,EAAE,KAAK;AAAA;AAErE,SAAK,WAAW;AAAA,0BACM,YAAY,KAAK;AAAA,iCACV,SAAS,KAAK;AAAA,sCACT,SAAS,MAAM,EAAE,EAAE,KAAK;AAAA;AAAA,QAEtD;AAAA;AAAA,UAEE;AAAA,4BACkB,QAAQ,OAAO,QAAQ,SAAS,OAAO,KAAK;AAAA,4BAC5C,QAAQ,OAAO,QAAQ,SAAS,OAAO,KAAK;AAAA,UAC9D;AAAA,yCAC+B,wBAAwB;AAAA,sBAC3C,wBAAwB,gBAAgB;AAAA;AAAA;AAAA,2BAGnC;AAAA;AAAA,8BAEG;AAAA;AAAA,YAElB;AAAA,6BACiB;AAAA;AAAA;AAAA,mBAGV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYjB;AACF;AAGA,SAAS,UAAU,UAAU,GAAG,YAAY,eAAe,MAAM;AAC/D,MAAI,YAAY,EAAE,MAAM;AACxB,MAAI,SAAS,EAAE,MAAM;AACrB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,aAAa,MAAM;AAC/B,aAAS,aAAa,MAAM;AAAA,EAC9B;AACA,QAAM,aAAa,qBAAqB,yBAAyB,MAAM;AACvE,QAAM,aAAa,EAAE,YAAY,QAAQ,WAAW,SAAS,KAAK,KAAK,SAAS,UAAU,EAAE;AAC5F,QAAM,UAAU,IAAI,iBAAiB,YAAY,YAAY,gBAAgB,IAAI;AACjF,QAAM,SAAS,CAAC,CAAC;AACjB,MAAI,gBAAgB,MAAM;AACxB,WAAO,KAAK,YAAY;AAAA,EAC1B;AACA,QAAM,SAAS,SAAS,gBAAgB,SAAS,QAAQ,OAAO;AAChE,MAAI,OAAO,MAAM,OAAO,GAAG;AACzB,WAAO;AAAA,EACT;AACA,QAAM,SAAS,UAAU,UAAU,GAAG,YAAY,MAAM;AACxD,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,SAAS,gBAAgB,UAAU,GAAG,YAAY,eAAe,MAAM;AACrE,QAAM,UAAU,gBAAgB,OAAO,aAAa,QAAQ,EAAE;AAC9D,QAAM,SAAS,QAAQ,QAAQ,SAAS;AACxC,QAAM,aAAa,qBAAqB,yBAAyB,MAAM;AACvE,QAAM,UAAU,IAAI,uBAAuB,SAAS,YAAY,YAAY,gBAAgB,IAAI;AAChG,QAAM,SAAS,gBAAgB,OAAO,CAAC,CAAC,IAAI,CAAC,GAAG,YAAY;AAC5D,QAAM,SAAS,SAAS,gBAAgB,SAAS,QAAQ,OAAO;AAChE,MAAI,OAAO,MAAM,WAAW,EAAE,MAAM,QAAQ;AAC1C,UAAM,SAAS,gBAAgB,UAAU,GAAG,YAAY,MAAM;AAC9D,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,SAAO;AACT;AACA,SAAS,gBAAgB,UAAU,GAAG,MAAM,YAAY;AACtD,QAAM,OAAO,CAAC,IAAI;AAClB,uBAAqB,2BAA2B,QAAQ,WAAW,OAAO,CAAC,EAAE,YAAY,IAAI,WAAW,MAAM,CAAC,GAAG,MAAM,EAAE,MAAM,MAAM;AACtI,MAAI,CAAC,IAAI,EAAE,QAAQ,mBAAmB,KAAK,EAAE,MAAM,UAAU,GAAG;AAC9D,UAAM,0BAA0B,CAAC;AACjC,UAAM,WAAW,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC9C,UAAM,YAAY,aAAa,QAAQ,SAAS;AAChD,QAAI,YAAY;AAChB,QAAI,WAAW;AACb,kBAAY,SAAS,aAAa,CAAC;AACnC,8BAAwB,KAAK,SAAS;AAAA,IACxC;AACA,UAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACpG,UAAM,SAAS,aAAa,cAAc,WAAW;AACrD,UAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,4BAAwB,KAAK,GAAG;AAChC,UAAM,UAAU,UAAU,UAAU,KAAK,UAAU;AACnD,4BAAwB,KAAK,OAAO;AACpC,UAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACnG,4BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,WAAO;AAAA,EACT;AACA,SAAO,gBAAgB,UAAU,GAAG,UAAU;AAChD;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACpD,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,QAAM,0BAA0B,CAAC;AACjC,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,4BAAwB,KAAK,EAAE;AAC/B,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,GAAG,MAAM,MAAM;AAAA,EAC3E;AACA,uBAAqB,2BAA2B,UAAU,CAAC,KAAK,EAAE,GAAG,GAAG,MAAM,MAAM;AACpF,QAAM,MAAM,gBAAgB,UAAU,IAAI,KAAK,IAAI,KAAK;AACxD,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACpD,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,EAAE,MAAM,MAAM;AACjF,MAAI,KAAK;AACT,QAAM,0BAA0B,CAAC;AACjC,MAAI,gBAAgB,MAAM;AACxB,SAAK,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AACnF,4BAAwB,KAAK,EAAE;AAC/B,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,GAAG,MAAM,MAAM;AAAA,EAC3E;AACA,uBAAqB,2BAA2B,UAAU,CAAC,KAAK,EAAE,GAAG,GAAG,MAAM,MAAM;AACpF,QAAM,MAAM,gBAAgB,UAAU,IAAI,KAAK,IAAI,KAAK;AACxD,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAM/B,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,oBAAoB;AAChC,IAAI,SAAS,iBAAiB,EAAE,WAAW,MAAM,CAAC;AAClD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO,oBAAoB;AAAA;AAAA;AAG/B,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,qBAAqB;AAAA;AAAA;AAGjC,IAAI,eAAe;AAAA;AAAA;AAAA;AAAA;AAAA,MAKb,2BAA2B;AAAA;AAAA;AAGjC,IAAI,SAAS,kBAAkB,EAAE,WAAW,OAAO,iBAAiB,aAAa,CAAC;AAClF,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,oBAAoB;AAAA;AAAA;AAGhC,IAAI,SAAS,iBAAiB,EAAE,WAAW,MAAM,CAAC;AAClD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,UAAU,UAAU,kBAAkB,mBAAmB,OAAO,sBAAsB,OAAO;AACvG,SAAK,gBAAgB,CAAC,GAAG;AACzB,QAAI,aAAa,SAAS,kBAAkB;AAC1C,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,SAAK,cAAc,SAAS;AAC5B,UAAM,YAAY,aAAa;AAC/B,UAAM,0BAA0B,cAAc,SAAS,oBAAoB,SAAS,mBAAmB,SAAS;AAChH,UAAM,qBAAqB,SAAS,SAAS,mBAAmB,SAAS;AACzE,QAAI,sBAAsB;AAC1B,QAAI,CAAC,WAAW;AACd,4BAAsB;AAAA,IACxB;AACA,QAAI,kBAAkB;AACpB,YAAM,aAAa;AACnB,WAAK,WAAW;AAAA,sCACgB,iBAAiB;AAAA,mCACpB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBZ;AAAA,sBACZ;AAAA;AAAA;AAAA,kCAGY,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP;AAAA,wBACZ;AAAA;AAAA;AAAA,oCAGY,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAUnB;AAAA;AAAA;AAAA,mCAGS,mBAAmB,sBAAsB,0BAA0B,qBAAqB,QAAQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAO7H;AAAA,IACF;AACA,UAAM,YAAY;AAClB,QAAI,cAAc,GAAG,YAAY,YAAY;AAC7C,QAAI,aAAa,OAAO;AACtB,oBAAc;AAAA,IAChB;AACA,UAAM,yBAAyB,KAAK,MAAM,cAAc,CAAC,IAAI;AAC7D,UAAM,2BAA2B,cAAc;AAC/C,UAAM,gBAAgB;AAAA,YACd;AAAA;AAAA;AAAA,wBAGY;AAAA;AAAA;AAGpB,SAAK,WAAW;AAAA,oCACgB,iBAAiB;AAAA,iCACpB,WAAW;AAAA,0CACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBL;AAAA;AAAA;AAAA;AAAA,gCAIF;AAAA,oBACZ;AAAA;AAAA;AAAA,gCAGY,SAAS;AAAA;AAAA;AAAA;AAAA,kCAIP;AAAA,uCACK;AAAA;AAAA;AAAA;AAAA,yCAIE;AAAA,6CACI;AAAA,6CACA;AAAA;AAAA;AAAA,cAG/B;AAAA;AAAA;AAAA,gCAGkB;AAAA,gBAChB,6BAA6B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,cAQ/B;AAAA,uBACS,6BAA6B;AAAA;AAAA;AAAA,yCAGX;AAAA;AAAA;AAAA;AAAA;AAAA,cAK3B;AAAA,uBACS,6BAA6B;AAAA;AAAA;AAAA,yCAGX;AAAA,6CACI;AAAA;AAAA;AAAA;AAAA,cAI/B;AAAA;AAAA;AAAA,oBAGM;AAAA;AAAA;AAAA,EAGlB;AACF;AACA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,UAAU,UAAU,kBAAkB,mBAAmB,OAAO,sBAAsB,OAAO;AACvG,SAAK,gBAAgB,CAAC,GAAG;AACzB,QAAI,aAAa,SAAS,kBAAkB;AAC1C,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,UAAM,cAAc,SAAS;AAC7B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,uBAAuB,SAAS;AACtC,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,WAAW,SAAS,QAAQ;AAClC,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,SAAK,cAAc,SAAS;AAC5B,UAAM,YAAY,aAAa;AAC/B,QAAI,sBAAsB;AAC1B,QAAI,CAAC,WAAW;AACd,4BAAsB;AAAA,IACxB;AACA,QAAI,kBAAkB;AACpB,YAAM,aAAa;AACnB,WAAK,WAAW;AAAA;AAAA,oBAEF,gBAAgB,iBAAiB;AAAA,mCAClB,aAAa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAkBzB;AAAA,sBACZ;AAAA;AAAA;AAAA,kCAGY,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP;AAAA,wBACZ;AAAA;AAAA;AAAA,oCAGY,SAAS;AAAA;AAAA;AAAA;AAAA,sCAIP;AAAA,0BACZ;AAAA;AAAA;AAAA,sCAGY,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,4BAUnB;AAAA;AAAA;AAAA,qCAGS,mBAAmB,sBAAsB,cAAc,SAAS,mBAAmB,SAAS,oBAAoB,SAAS,mBAAmB,SAAS,oBAAoB,UAAU,SAAS,oBAAoB,SAAS,mBAAmB,SAAS,oBAAoB,QAAQ,2BAA2B;AAAA,6BACpT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAQvB;AAAA,IACF;AACA,UAAM,YAAY;AAClB,QAAI,cAAc,GAAG,YAAY,YAAY;AAC7C,QAAI,aAAa,OAAO;AACtB,oBAAc;AAAA,IAChB;AACA,UAAM,yBAAyB,KAAK,MAAM,cAAc,CAAC,IAAI;AAC7D,UAAM,2BAA2B,cAAc;AAC/C,UAAM,gBAAgB;AAAA,YACd;AAAA;AAAA;AAAA,wBAGY;AAAA;AAAA;AAGpB,SAAK,WAAW;AAAA;AAAA,gBAEJ,gBAAgB,iBAAiB;AAAA,iCAChB,aAAa,WAAW;AAAA,0CACf;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAmBL;AAAA;AAAA;AAAA;AAAA,gCAIF;AAAA,oBACZ;AAAA;AAAA;AAAA,gCAGY,SAAS;AAAA;AAAA;AAAA;AAAA,kCAIP;AAAA,oBACd;AAAA;AAAA;AAAA,kCAGc,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP;AAAA,yCACK;AAAA;AAAA;AAAA;AAAA,+CAIM;AAAA,mDACI;AAAA,mDACA;AAAA;AAAA;AAAA,gBAGnC;AAAA;AAAA;AAAA,kCAGkB;AAAA,kBAChB,6BAA6B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gBAQ/B;AAAA,yBACS,6BAA6B;AAAA;AAAA;AAAA,+CAGP;AAAA;AAAA;AAAA;AAAA;AAAA,gBAK/B;AAAA,yBACS,6BAA6B;AAAA;AAAA;AAAA,+CAGP;AAAA,mDACI;AAAA;AAAA;AAAA;AAAA,gBAInC;AAAA;AAAA;AAAA,sBAGM;AAAA;AAAA;AAAA;AAAA,EAIpB;AACF;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,oBAAkB,GAAG,SAAS;AAC9B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY;AAClB,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AACjN,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,MAAI,SAAS,gBAAgB,KAAK,SAAS,iBAAiB,KAAK,aAAa,YAAY,SAAS,SAAS,SAAS,QAAQ,GAAG;AAC9H,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD;AACA,QAAM,iBAAiB,IAAI,cAAc,UAAU,OAAO,KAAK;AAC/D,SAAO,SAAS,gBAAgB,gBAAgB,CAAC,CAAC,GAAG,SAAS;AAChE;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACxE,QAAM,YAAY,CAAC,GAAG,GAAG,CAAC;AAC1B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,iBAAiB,UAAU;AAClI,QAAM,iBAAiB,IAAI,cAAc,UAAU,OAAO,KAAK;AAC/D,SAAO,SAAS,gBAAgB,gBAAgB,CAAC,CAAC,GAAG,SAAS;AAChE;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,MAAM;AAAA,EACnC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,IAAI;AAC1B,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,UAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,UAAM,gBAAgB,KAAK,eAAe;AAC1C,SAAK,WAAW;AAAA,iCACa,WAAW;AAAA,0CACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAcV;AAAA,oBACZ;AAAA,gDAC4B;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKX;AAAA,mBACf;AAAA,kDAC+B;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAc7C;AACF;AACA,IAAI,2BAA2B,MAAM;AAAA,EACnC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,IAAI;AAC1B,SAAK,cAAc,SAAS;AAC5B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,uBAAuB,SAAS;AACtC,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,WAAW,uBAAuB,IAAI,SAAS,QAAQ;AAC7D,UAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,UAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,UAAM,gBAAgB,KAAK,cAAc,eAAe;AACxD,SAAK,WAAW;AAAA,iCACa,aAAa,WAAW;AAAA,0CACf;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAiBV;AAAA,oBACZ;AAAA,gDAC4B;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKX;AAAA,sBACZ;AAAA,kDAC4B;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAMX;AAAA,wBACZ;AAAA,oDAC4B;AAAA;AAAA,wCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAe/C;AACF;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,IAAI;AACV,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY,CAAC,GAAG,GAAG,CAAC;AAC1B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,QAAM,yBAAyB,IAAI,yBAAyB,QAAQ;AACpE,SAAO,SAAS,gBAAgB,wBAAwB,CAAC,EAAE,GAAG,EAAE,KAAK;AACvE;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,IAAI;AACV,oBAAkB,CAAC,IAAI,MAAM,GAAG,aAAa;AAC7C,QAAM,EAAE,YAAY,SAAS,KAAK,KAAK,IAAI;AAC3C,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,IAAI;AAC7F,QAAM,yBAAyB,IAAI,yBAAyB,QAAQ;AACpE,SAAO,SAAS,gBAAgB,wBAAwB,CAAC,EAAE,GAAG,EAAE,KAAK;AACvE;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,EAAE,IAAI;AACjB,QAAM,EAAE,YAAY,WAAW,IAAI;AACnC,SAAO,gBAAgB,EAAE,GAAG,GAAG,YAAY,YAAY,SAAS,SAAS,CAAC;AAC5E;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,QAAQ,WAAW,eAAe,aAAa,YAAY,iBAAiB;AACtF,SAAK,cAAc,CAAC;AACpB,SAAK,gBAAgB,CAAC,KAAK,QAAQ,UAAU;AAC7C,yBAAqB,2BAA2B,QAAQ,SAAS;AACjE,yBAAqB,2BAA2B,QAAQ,aAAa;AACrE,QAAI,gBAAgB;AACpB,QAAI,eAAe,MAAM;AACvB,2BAAqB,2BAA2B,QAAQ,WAAW;AACnE,WAAK,cAAc,KAAK,QAAQ;AAChC,sBAAgB;AAAA,IAClB;AACA,QAAI,eAAe;AACnB,QAAI,cAAc,MAAM;AACtB,2BAAqB,2BAA2B,QAAQ,UAAU;AAClE,WAAK,cAAc,KAAK,OAAO;AAC/B,qBAAe;AAAA,IACjB;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKK;AAAA,wBACD;AAAA,2DACmC;AAAA;AAAA;AAAA;AAAA,EAIzD;AACF;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,QAAQ,WAAW,eAAe,aAAa,YAAY,iBAAiB;AACtF,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,gBAAgB,CAAC,KAAK,QAAQ,UAAU;AAC7C,yBAAqB,2BAA2B,QAAQ,SAAS;AACjE,yBAAqB,2BAA2B,QAAQ,aAAa;AACrE,QAAI,gBAAgB;AACpB,QAAI,eAAe,MAAM;AACvB,2BAAqB,2BAA2B,QAAQ,WAAW;AACnE,WAAK,cAAc,KAAK,QAAQ;AAChC,sBAAgB;AAAA,IAClB;AACA,QAAI,eAAe;AACnB,QAAI,cAAc,MAAM;AACtB,2BAAqB,2BAA2B,QAAQ,UAAU;AAClE,WAAK,cAAc,KAAK,OAAO;AAC/B,qBAAe;AAAA,IACjB;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA,wBAEI;AAAA,uBACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yDAMkC;AAAA;AAAA;AAAA;AAAA;AAAA,EAKvD;AACF;AAGA,IAAI,aAAa,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AACzD,QAAM,EAAE,GAAG,MAAM,OAAO,UAAU,QAAQ,OAAOI,QAAO,IAAI;AAC5D,eAAa,OAAO,MAAM,MAAM,WAAW,SAAS,MAAM,QAAQ,MAAM,8EAA8E;AACtJ,eAAa,OAAO,UAAU,QAAQ,MAAM,MAAM,WAAW,OAAO,MAAM,QAAQ,MAAM,4EAA4E;AACpK,eAAa,OAAOA,WAAU,QAAQ,MAAM,MAAM,WAAWA,QAAO,MAAM,QAAQ,MAAM,2EAA2E;AACnK,MAAI,EAAE,gBAAgB,IAAI;AAC1B,MAAI,mBAAmB,MAAM;AAC3B,sBAAkB;AAAA,EACpB;AACA,QAAM,cAAc,CAAC,GAAG,OAAO,QAAQ;AACvC,MAAI,cAAc;AAClB,MAAI,UAAU,MAAM;AAClB,kBAAc,OAAO;AACrB,gBAAY,KAAK,MAAM;AAAA,EACzB;AACA,MAAI,aAAa;AACjB,MAAIA,WAAU,MAAM;AAClB,iBAAaA,QAAO;AACpB,gBAAY,KAAKA,OAAM;AAAA,EACzB;AACA,QAAM,UAAU,IAAI,EAAE,QAAQ,0BAA0B,IAAI,IAAI,uBAAuB,EAAE,OAAO,MAAM,OAAO,SAAS,OAAO,aAAa,YAAY,eAAe,IAAI,IAAI,iBAAiB,EAAE,OAAO,MAAM,OAAO,SAAS,OAAO,aAAa,YAAY,eAAe;AAC5Q,QAAM,SAAS,SAAS,gBAAgB,SAAS,aAAa,YAAY,GAAG,KAAK;AAClF,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,QAAQ;AAC9B,SAAK,cAAc;AACnB,SAAK,OAAO,SAAS;AACrB,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,YAAY,KAAK,MAAM,MAAM,MAAM,CAAC;AAC5E,UAAM,eAAe,UAAU,KAAK,IAAI;AACxC,QAAI;AACJ,UAAM,WAAW,SAAS,IAAI,CAAC,GAAG,MAAM;AACtC,aAAO,aAAa,OAAO,cAAc,eAAe,OAAO;AAAA,IACjE,CAAC;AACD,WAAO;AAAA,UACD;AAAA,UACA;AAAA,UACA,SAAS,KAAK,IAAI;AAAA;AAExB,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,8BACoB;AAAA;AAAA;AAAA,EAG5B;AACF;AACA,IAAI,SAAS,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG;AAC1C,SAAS,UAAU,MAAM;AACvB,MAAI,SAAS,GAAG;AACd,WAAO;AAAA,EACT,WAAW,QAAQ,GAAG;AACpB,WAAO,OAAO,MAAM,GAAG,IAAI,EAAE,IAAI,CAAC,MAAM,eAAe,CAAC,EAAE,KAAK,GAAG;AAAA,EACpE,OAAO;AACL,UAAM,MAAM,oBAAoB,2BAA2B;AAAA,EAC7D;AACF;AAGA,IAAI,qBAAqB,MAAM;AAAA,EAC7B,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,QAAQ;AAC9B,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc;AACnB,SAAK,OAAO,SAAS;AACrB,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,YAAY,KAAK,MAAM,MAAM,MAAM,CAAC;AAC5E,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,UAAM,UAAU,YAAY,UAAU,KAAK,IAAI;AAC/C,UAAM,YAAY,YAAY,aAAa,KAAK,IAAI;AACpD,UAAM,YAAY,KAAK,SAAS,IAAI,cAAc,QAAQ,UAAU,MAAM,EAAE,EAAE,KAAK;AACnF,UAAM,aAAa,wBAAwB,UAAU,KAAK,OAAO;AACjE,UAAM,WAAW;AAAA,mBACF;AAAA,cACL,QAAQ,KAAK,OAAO,QAAQ,SAAS,KAAK,OAAO;AAAA,YACnD,UAAU,KAAK,OAAO;AAAA,qBACb;AAAA,YACT,UAAU,KAAK,OAAO;AAAA;AAAA;AAG9B,UAAM,WAAW,KAAK,SAAS,IAAI,KAAK;AAAA,UAClC,QAAQ,KAAK,OAAO;AAAA,cAChB,QAAQ,KAAK,OAAO,QAAQ,SAAS,KAAK,OAAO;AAAA,YACnD,UAAU,KAAK,OAAO;AAAA,qBACb;AAAA,gBACL,QAAQ,KAAK,OAAO,QAAQ,SAAS,KAAK,OAAO;AAAA,cACnD,UAAU,KAAK,OAAO;AAAA,uBACb;AAAA;AAAA;AAAA;AAInB,UAAM,iBAAiB,KAAK,QAAQ,IAAI;AAAA,cAC9B,SAAS,SAAS,IAAI,CAAC,GAAG,MAAM,SAAS,IAAI,EAAE,KAAK,QAAQ,SAAS,IAAI,CAAC,GAAG,MAAM,GAAG,UAAU,QAAQ,QAAQ,cAAc,KAAK,EAAE,KAAK,IAAI;AACxJ,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,UACA;AAAA,UACA;AAAA;AAAA,UAEA;AAAA,UACA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AAGA,SAAS,aAAa,GAAG,OAAO,MAAM,UAAU;AAC9C,QAAM,WAAW,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC9C,QAAM,IAAI,SAAS,eAAe,MAAM,EAAE,KAAK;AAC/C,QAAM,aAAa,SAAS,QAAQ,IAAI,EAAE,MAAM;AAChD,SAAO,OAAO,YAAY,QAAQ;AAClC,aAAW,WAAW;AACtB,aAAW,QAAQ;AACnB,aAAW,QAAQ,EAAE;AACrB,MAAI,aAAa,mBAAmB,kBAAkB,OAAO,aAAa,eAAe,EAAE,KAAK,CAAC;AACjG,MAAI,SAAS,OAAO;AAClB,kBAAc,SAAS,MAAM;AAAA,EAC/B;AACA,aAAW,QAAQ;AAAA,IACjB;AAAA,IACA,YAAY,SAAS,SAAS,SAAS,MAAM,cAAc,EAAE;AAAA,EAC/D;AACA,QAAM,WAAW,SAAS,aAAa,IAAI,WAAW,MAAM,UAAU,KAAK;AAC3E,WAAS,aAAa,IAAI,WAAW,MAAM,YAAY,WAAW,CAAC;AACnE,SAAO;AACT;AACA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,OAAO,KAAK,IAAI;AACxB,QAAM,CAAC,QAAQ,KAAK,IAAI,mBAAmB,iBAAiB,GAAG,OAAO,IAAI;AAC1E,qBAAmB,kBAAkB,GAAG,QAAQ,KAAK;AACrD,MAAI,aAAa,cAAc,KAAK,MAAM,GAAG;AAC3C,WAAO,SAAS,eAAe,OAAO,EAAE,OAAO,CAAC,CAAC;AAAA,EACnD;AACA,MAAI,SAAS,mBAAmB,CAAC,CAAC,CAAC,KAAK,EAAE,UAAU,UAAU;AAC5D,UAAM,WAAW,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC9C,UAAM,YAAY,aAAa,SAAS,QAAQ,QAAQ,OAAO,EAAE,OAAO,EAAE,KAAK;AAC/E,WAAO,SAAS,eAAe,OAAO,EAAE,OAAO,SAAS;AAAA,EAC1D;AACA,QAAM,EAAE,SAAS,IAAI,SAAS,QAAQ,IAAI,EAAE,MAAM;AAClD,QAAM,cAAc,mBAAmB,iBAAiB,EAAE,OAAO,QAAQ,KAAK;AAC9E,MAAI,YAAY,CAAC,aAAa;AAC5B,UAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,mBAAmB,KAAK,IAAI,IAAI,aAAa,KAAK;AACrH,UAAM,eAAe,CAAC,MAAM;AAC5B,WAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,OAAO,YAAY;AAAA,EACrE;AACA,WAAS,YAAY,EAAE,MAAM;AAC7B,SAAO,aAAa,GAAG,QAAQ,OAAO,QAAQ;AAChD;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,kBAAkB,CAAC,SAAS;AAC9B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,MAAM,IAAI;AAC9B,eAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,sEAAsE;AACrH,QAAM,QAAQ,WAAW,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAC/C,QAAM,WAAW,qBAAqB,YAAY,EAAE,OAAO,YAAY,KAAK;AAC5E,QAAM,WAAW,qBAAqB,YAAY,SAAS,QAAQ,WAAW,MAAM;AACpF,QAAM,mBAAmB,qBAAqB,oBAAoB,EAAE,OAAO,YAAY,KAAK;AAC5F,QAAM,mBAAmB,qBAAqB,oBAAoB,OAAO,WAAW,MAAM;AAC1F,QAAM,YAAY,qBAAqB,aAAa,kBAAkB,OAAO,WAAW,MAAM;AAC9F,QAAM,YAAY,CAAC;AACnB,QAAM,uBAAuB,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACtG,QAAM,yBAAyB,WAAW,EAAE,QAAQ,EAAE,GAAG,qBAAqB,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,SAAS,EAAE,CAAC;AAC/H,QAAM,wBAAwB,SAAS;AAAA,IACrC,QAAQ,EAAE,GAAG,uBAAuB;AAAA,IACpC,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,iBAAiB;AAAA,EACnC,CAAC;AACD,QAAM,SAAS,OAAO;AAAA,IACpB,QAAQ,EAAE,GAAG,sBAAsB;AAAA,IACnC,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,kBAAkB,MAAM,UAAU;AAAA,EACpD,CAAC;AACD,YAAU,KAAK,oBAAoB;AACnC,YAAU,KAAK,sBAAsB;AACrC,YAAU,KAAK,qBAAqB;AACpC,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,QAAQ,SAAS,SAAS,EAAE,MAAM;AACxC,QAAM,cAAc,SAAS,SAAS,QAAQ,MAAM;AACpD,QAAM,UAAU,gBAAgB,OAAO,aAAa,QAAQ,OAAO,QAAQ,OAAO,IAAI;AACtF,SAAO,SAAS,eAAe,CAAC,IAAI,GAAG,QAAQ,OAAO,OAAO;AAC/D;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,IAAI,GAAG,IAAI;AACnB,QAAM,SAAS,SAAS,SAAS,GAAG,MAAM;AAC1C,QAAM,SAAS,SAAS,SAAS,GAAG,MAAM;AAC1C,QAAM,iBAAiB,qBAAqB,2BAA2B,MAAM,KAAK,MAAM,GAAG,MAAM,KAAK,MAAM,CAAC;AAC7G,SAAO,SAAS,eAAe,CAAC,eAAe,MAAM,GAAG,SAAS,WAAW,KAAK,cAAc,CAAC;AAClG;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY;AAChB,IAAI,YAAY,kBAAkB,EAAE,WAAW,WAAW,eAAe,iBAAiB,OAAO,OAAO,CAAC;AACzG,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,YAAY,SAAS,QAAQ,IAAI,OAAO,MAAM;AACpD,SAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,UAAU,mBAAmB,KAAK,GAAG,SAAS,SAAS,CAAC;AAC1F;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS;AACb,SAAS,IAAI,QAAQ,UAAU;AAC7B,QAAM,UAAU,IAAI,eAAe,OAAO,OAAO,MAAM;AACvD,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO;AAClE,SAAO,EAAE,QAAQ,OAAO,QAAQ,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAC3E;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,MAAI,UAAU,aAAa;AACzB,QAAI,EAAE,UAAU,aAAa;AAC3B,aAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,IACvD;AACA,UAAM,cAAc,MAAM,EAAE,KAAK;AACjC,UAAM,SAAS,MAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AACtF,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,QAAQ,MAAM,YAAY,GAAG,SAAS,SAAS,CAAC;AAC1F,gBAAY,QAAQ;AACpB,aAAS,8BAA8B,MAAM;AAC7C,WAAO;AAAA,EACT;AACA,MAAI,EAAE,UAAU,aAAa;AAC3B,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,SAAS,MAAM,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AACrF,aAAS,8BAA8B,QAAQ;AAC/C,WAAO;AAAA,EACT;AACA,MAAI,CAAC,aAAa,gBAAgB,EAAE,OAAO,KAAK,GAAG;AACjD,UAAM,SAAS,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAC7D,WAAO,EAAE,QAAQ,OAAO,QAAQ,OAAO,OAAO,OAAO,MAAM;AAAA,EAC7D;AACA,MAAI,SAAS,mBAAmB,CAAC,CAAC,CAAC,GAAG;AACpC,UAAM,SAAS,SAAS,QAAQ,IAAI,EAAE,MAAM,EAAE;AAC9C,UAAM,CAAC,aAAa,YAAY,UAAU,IAAI,YAAY,QAAQ,EAAE,OAAO,EAAE,OAAO,KAAK;AACzF,WAAO,SAAS,eAAe,aAAa,YAAY,UAAU;AAAA,EACpE;AACA,MAAI,UAAU,SAAS;AACrB,WAAO,IAAI,GAAG,QAAQ;AAAA,EACxB;AACA,MAAI,UAAU,QAAQ;AACpB,UAAM,kBAAkB,SAAS,eAAe,CAAC,GAAG,QAAQ,aAAa,uBAAuB,QAAQ,CAAC,CAAC;AAC1G,UAAM,eAAe,EAAE,GAAG,GAAG,GAAG,gBAAgB;AAChD,UAAM,SAAS,UAAU,EAAE,QAAQ,cAAc,SAAS,SAAS,CAAC;AACpE,aAAS,8BAA8B,eAAe;AACtD,WAAO;AAAA,EACT;AACA,QAAM,IAAI,MAAM,iCAAiC,EAAE,YAAY,OAAO;AACxE;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AACX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,MAAM,iBAAiB,MAAM,eAAe,YAAY,CAAC;AACnG,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,QAAQ;AAClB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,MAChC,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,IAClC;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYlB;AACF;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,QAAQ;AAClB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,MAChC,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,IAClC;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYlB;AACF;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,cAAc,aAAa,IAAI;AACvC,MAAI;AACJ,MAAI,IAAI,EAAE,QAAQ,iBAAiB,GAAG;AACpC,cAAU,IAAI,kBAAkB,EAAE,KAAK;AAAA,EACzC,OAAO;AACL,cAAU,IAAI,YAAY,EAAE,KAAK;AAAA,EACnC;AACA,QAAM,eAAe,CAAC,CAAC,YAAY,GAAG,CAAC,YAAY,CAAC;AACpD,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,OAAO,YAAY;AACrE;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,OAAO;AACjB,SAAK,gBAAgB,CAAC,QAAQ,MAAM;AACpC,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAclB;AACF;AAGA,SAAS,+BAA+B,eAAe,aAAa;AAClE,SAAO;AAAA,IACL,QAAQ,YAAY;AAAA,IACpB,OAAO,YAAY;AAAA,IACnB,OAAO,cAAc;AAAA,EACvB;AACF;AACA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,QAAM,UAAU,IAAI,kBAAkB,EAAE,KAAK;AAC7C,QAAM,gBAAgB;AAAA,IACpB,+BAA+B,GAAG,MAAM,mBAAmB,IAAI;AAAA,IAC/D,+BAA+B,GAAG,MAAM,mBAAmB,IAAI;AAAA,EACjE;AACA,SAAO,SAAS,gBAAgB,SAAS,eAAe,cAAc,GAAG,KAAK;AAChF;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,QAAQ;AAClB,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc,qBAAqB,gBAAgB,QAAQ,CAAC;AACjE,SAAK,gBAAgB,OAAO,IAAI,CAAC,GAAG,MAAM,IAAI,GAAG;AACjD,UAAM,UAAU,IAAI,MAAM,OAAO,SAAS,CAAC;AAC3C,YAAQ,KAAK,OAAO,GAAG;AACvB,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,cAAQ,KAAK,QAAQ,IAAI,KAAK,OAAO,GAAG;AAAA,IAC1C;AACA,UAAM,WAAW,CAAC,YAAY,QAAQ,+BAA+B;AACrE,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,YAAM,QAAQ,QAAQ,IAAI;AAC1B,eAAS,KAAK,iBAAiB,QAAQ,qBAAqB,YAAY,UAAU;AAAA,IACpF;AACA,UAAM,YAAY,QAAQ;AAC1B,UAAM,YAAY,QAAQ,QAAQ,SAAS;AAC3C,aAAS,KAAK,sBAAsB,oBAAoB,cAAc;AACtE,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAMV,SAAS,KAAK,YAAY;AAAA;AAAA;AAAA,EAGlC;AACF;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,QAAQ,MAAM;AACxB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc,qBAAqB,gBAAgB,QAAQ,IAAI;AACpE,UAAM,QAAQ,KAAK;AACnB,UAAM,OAAO,MAAM;AACnB,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,UAAU,YAAY,UAAU,IAAI;AAC1C,UAAM,WAAW,CAAC,KAAK,KAAK,KAAK,KAAK,KAAK,GAAG,EAAE,MAAM,GAAG,IAAI;AAC7D,SAAK,gBAAgB,OAAO,IAAI,CAAC,GAAG,MAAM,IAAI,GAAG;AACjD,UAAM,UAAU,IAAI,MAAM,OAAO,SAAS,CAAC;AAC3C,YAAQ,KAAK,OAAO,GAAG;AACvB,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,cAAQ,KAAK,QAAQ,IAAI,KAAK,OAAO,GAAG;AAAA,IAC1C;AACA,UAAM,UAAU,SAAS;AACzB,UAAM,eAAe,SAAS,MAAM,EAAE;AACtC,UAAM,cAAc,SAAS,KAAK;AAClC,QAAI,kBAAkB,OAAO,aAAa,QAAQ;AAAA;AAAA,oBAElC,sBAAsB,aAAa,KAAK;AAAA;AAExD,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,YAAM,SAAS,QAAQ,IAAI;AAC3B,yBAAmB;AAAA,cACX,aAAa,QAAQ,UAAU,cAAc,QAAQ,IAAI;AAAA;AAAA,kBAErD,KAAK,gBAAgB,UAAU,SAAS,MAAM;AAAA,mBAC7C,gBAAgB,cAAc,SAAS,MAAM;AAAA;AAAA,IAE5D;AACA,UAAM,YAAY,QAAQ;AAC1B,UAAM,QAAQ,QAAQ,QAAQ,SAAS;AACvC,uBAAmB;AAAA;AAAA,gBAEP,aAAa,gBAAgB,UAAU,SAAS,KAAK;AAAA,iBACpD,gBAAgB,cAAc,SAAS,KAAK;AACzD,SAAK,WAAW;AAAA,uBACG,SAAS,IAAI,CAAC,MAAM,SAAS,CAAC;AAAA,UAC3C;AAAA;AAAA;AAAA;AAAA,UAIA;AAAA,sCAC4B;AAAA;AAAA,UAE5B,QAAQ,OAAO,QAAQ,QAAQ,OAAO;AAAA,cAClC,QAAQ,OAAO,QAAQ,MAAM,OAAO;AAAA,gCAClB;AAAA;AAAA;AAAA,UAGtB,QAAQ,OAAO,QAAQ,QAAQ,OAAO;AAAA,cAClC,QAAQ,OAAO,QAAQ,MAAM,OAAO;AAAA,gCAClB;AAAA;AAAA;AAAA,UAGtB,QAAQ,OAAO,QAAQ,QAAQ,OAAO;AAAA,cAClC,QAAQ,OAAO,QAAQ,MAAM,OAAO;AAAA,cACpC,QAAQ,OAAO,QAAQ,MAAM,OAAO;AAAA,gCAClB;AAAA;AAAA;AAAA;AAAA;AAAA,EAK9B;AACF;AACA,SAAS,gBAAgB,UAAU,SAAS,OAAO;AACjD,QAAM,aAAa,SAAS,QAAQ,OAAO;AAC3C,QAAM,MAAM,SAAS,IAAI,CAAC,GAAG,QAAQ;AACnC,QAAI,QAAQ,YAAY;AACtB,aAAO,GAAG,OAAO;AAAA,IACnB,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF,CAAC;AACD,SAAO,IAAI,KAAK;AAClB;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,YAAY,SAAS,QAAQ,IAAI,OAAO,MAAM;AACpD,SAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,UAAU,mBAAmB,KAAK,GAAG,SAAS,SAAS,CAAC;AAC1F;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,QAAQ,MAAM,UAAU;AAC3C,QAAM,QAAQ,OAAO,GAAG;AACxB,MAAI,UAAU,aAAa;AACzB,UAAM,QAAQ,OAAO,IAAI,CAAC,MAAM,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC,CAAC;AAClF,UAAM,QAAQ,OAAO,IAAI,CAAC,MAAM,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC,CAAC;AAClF,UAAM,eAAe,YAAY,OAAO,MAAM,QAAQ;AACtD,UAAM,eAAe,YAAY,OAAO,MAAM,QAAQ;AACtD,UAAM,UAAU,SAAS,EAAE,QAAQ,EAAE,MAAM,cAAc,MAAM,aAAa,GAAG,SAAS,SAAS,CAAC;AAClG,UAAM,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAC9D,UAAM,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAC9D,aAAS,8BAA8B,YAAY;AACnD,aAAS,8BAA8B,YAAY;AACnD,WAAO;AAAA,EACT;AACA,MAAI,WAAW,SAAS,mBAAmB,MAAM;AACjD,MAAI,UAAU,UAAU;AACtB,eAAW;AAAA,EACb;AACA,MAAI,UAAU;AACZ,UAAM,aAAa,OAAO,IAAI,CAAC,MAAM;AACnC,YAAM,YAAY,aAAa,cAAc,EAAE,MAAM,MAAM,IAAI,CAAC;AAChE,YAAM,QAAQ,CAAC,IAAI,SAAS;AAC5B,aAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AAAA,IAC3E,CAAC;AACD,UAAM,kBAAkB,WAAW,IAAI,CAAC,MAAM;AAC5C,aAAO,EAAE,MAAM,SAAS,SAAS,EAAE,MAAM,GAAG,OAAO,EAAE,MAAM;AAAA,IAC7D,CAAC;AACD,UAAM,YAAY,qBAAqB,gBAAgB,WAAW,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AACxF,UAAM,eAAe,WAAW,GAAG,MAAM,OAAO;AAChD,UAAM,UAAU,cAAc,iBAAiB,WAAW,OAAO,YAAY;AAC7E,UAAM,gBAAgB,qBAAqB,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,IAAI;AAC3F,UAAM,UAAU,SAAS,eAAe,eAAe,OAAO,OAAO;AACrE,eAAW,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AACnE,WAAO;AAAA,EACT;AACA,QAAM,sBAAsB,IAAI,EAAE,UAAU,8BAA8B;AAC1E,MAAI,OAAO,SAAS,qBAAqB;AACvC,UAAM,gBAAgB,CAAC;AACvB,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,qBAAqB;AAC3D,YAAM,WAAW,OAAO,MAAM,GAAG,IAAI,mBAAmB;AACxD,oBAAc,KAAK,YAAY,UAAU,MAAM,QAAQ,CAAC;AAAA,IAC1D;AACA,UAAM,UAAU,YAAY,eAAe,MAAM,QAAQ;AACzD,eAAW,KAAK,eAAe;AAC7B,eAAS,8BAA8B,CAAC;AAAA,IAC1C;AACA,WAAO;AAAA,EACT;AACA,MAAI,IAAI,EAAE,QAAQ,6BAA6B,KAAK,OAAO,GAAG,MAAM,SAAS,GAAG;AAC9E,UAAM,WAAW,IAAI,oBAAoB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,IAAI;AACzE,WAAO,SAAS,gBAAgB,UAAU,QAAQ,KAAK;AAAA,EACzD;AACA,QAAM,EAAE,WAAW,SAAS,IAAI,iBAAiB,QAAQ,MAAM,QAAQ;AACvE,QAAM,UAAU,IAAI,cAAc,UAAU,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC;AAC/D,QAAM,SAAS,SAAS,gBAAgB,SAAS,WAAW,KAAK;AACjE,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AACxG,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AAChD,QAAM,WAAW,qBAAqB,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,IAAI;AACtF,QAAM,YAAY,OAAO,IAAI,CAAC,MAAM,SAAS;AAAA,IAC3C,QAAQ,EAAE,EAAE;AAAA,IACZ,OAAO,EAAE,OAAO,CAAC,IAAI,aAAa,cAAc,EAAE,MAAM,MAAM,IAAI,CAAC,CAAC,EAAE;AAAA,IACtE,SAAS;AAAA,EACX,CAAC,CAAC;AACF,SAAO,EAAE,WAAW,SAAS;AAC/B;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,QAAQ,aAAa,eAAe,MAAM,OAAO,GAAG,KAAK,EAAE;AACjE,QAAM,SAAS,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AACxC,uBAAqB,uBAAuB,QAAQ,KAAK;AACzD,QAAM,WAAW,qBAAqB,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,KAAK;AACvF,MAAI,aAAa,cAAc,QAAQ,MAAM,GAAG;AAC9C,WAAO,SAAS,eAAe,UAAU,OAAO,GAAG,OAAO,CAAC,CAAC;AAAA,EAC9D;AACA,QAAM,UAAU,OAAO,OAAO,CAAC,MAAM,aAAa,cAAc,EAAE,KAAK,IAAI,CAAC;AAC5E,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,GAAG,SAAS,SAAS,CAAC;AAAA,EACnE;AACA,SAAO,YAAY,SAAS,OAAO,QAAQ;AAC7C;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,UAAU,UAAU,OAAO,cAAc,MAAM,4BAA4B,OAAO,oBAAoB,OAAO;AACvH,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,cAAc,SAAS;AAC5B,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,wBAAwB,KAAK,MAAM,SAAS,aAAa,CAAC,IAAI;AACpE,UAAM,0BAA0B,SAAS,aAAa;AACtD,UAAM,iBAAiB,SAAS,eAAe;AAC/C,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,aAAa,iBAAiB,IAAI;AACxC,QAAI,oBAAoB,IAAI,yBAAyB;AACrD,QAAI,aAAa;AACf,UAAI,2BAA2B;AAC7B,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,WAAW,mBAAmB;AAC5B,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,OAAO;AACL,4BAAoB;AAAA;AAAA,cAEd;AAAA;AAAA;AAAA,MAGR;AACA,+BAAyB;AAAA,IAC3B;AACA,UAAM,iBAAiB,UAAU,oCAAoC;AACrE,QAAI,SAAS;AACX,WAAK,cAAc,KAAK,MAAM;AAAA,IAChC;AACA,QAAI,2BAA2B;AAC7B,WAAK,cAAc,KAAK,wBAAwB;AAAA,IAClD;AACA,QAAI,mBAAmB;AACrB,WAAK,cAAc,KAAK,gBAAgB;AAAA,IAC1C;AACA,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA,oCAE4B,iBAAiB;AAAA,iCACpB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKlB;AAAA;AAAA;AAAA,2BAGC,mBAAmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAOd;AAAA,qCACK;AAAA;AAAA,gCAEL,SAAS;AAAA;AAAA;AAAA;AAAA,kCAIP;AAAA,uCACK;AAAA;AAAA,kCAEL,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oBAQhB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kBAmBF,4BAA4B;AAAA;AAAA,oBAE1B;AAAA;AAAA,0CAEsB;AAAA,mCACP;AAAA;AAAA;AAAA,kCAGD;AAAA,mCACC;AAAA;AAAA;AAAA,yBAGV,4BAA4B;AAAA;AAAA,+BAEtB;AAAA,+BACA;AAAA;AAAA;AAAA,oBAGX;AAAA;AAAA,wCAEoB;AAAA,wCACA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKR;AAAA,gCACA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKP,4BAA4B;AAAA;AAAA,+BAEtB;AAAA,+BACA;AAAA,+BACA;AAAA;AAAA;AAAA,oBAGX;AAAA;AAAA,wCAEoB;AAAA,wCACA;AAAA,wCACA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKR;AAAA,gCACA;AAAA,gCACA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAUtB;AAAA,UACA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AACA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,cAAc,SAAS;AAC5B,UAAM,WAAW,SAAS,QAAQ;AAClC,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,wBAAwB,KAAK,MAAM,SAAS,aAAa,CAAC,IAAI;AACpE,UAAM,0BAA0B,SAAS,aAAa;AACtD,SAAK,WAAW;AAAA,oCACgB,gBAAgB,iBAAiB;AAAA,iCACpC,aAAa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAgBzB;AAAA,qCACK;AAAA;AAAA,gCAEL,SAAS;AAAA;AAAA;AAAA;AAAA,kCAIP;AAAA,uCACK;AAAA;AAAA,kCAEL,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP;AAAA,yCACK;AAAA;AAAA,oCAEL,SAAS;AAAA;AAAA;AAAA;AAAA,sCAIP;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oBAiBlB,4BAA4B;AAAA;AAAA,4CAEJ;AAAA,qCACP;AAAA,2BACV,4BAA4B;AAAA;AAAA,4CAEX;AAAA,4CACA;AAAA;AAAA;AAAA,qCAGP;AAAA,qCACA;AAAA;AAAA;AAAA,2BAGV,4BAA4B;AAAA;AAAA,4CAEX;AAAA,4CACA;AAAA,4CACA;AAAA;AAAA;AAAA,qCAGP;AAAA,qCACA;AAAA,qCACA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUnC;AACF;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,UAAU,UAAU,OAAO,cAAc,MAAM,qBAAqB,OAAO,oBAAoB,OAAO;AAChH,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,QAAQ,MAAM,QAAQ;AAAA,MAC9B,EAAE,MAAM,WAAW,MAAM,QAAQ;AAAA,MACjC,EAAE,MAAM,aAAa,MAAM,QAAQ;AAAA,MACnC,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,IAClC;AACA,SAAK,cAAc,SAAS;AAC5B,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe;AACrB,QAAI,WAAW;AAAA;AAAA;AAGf,aAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAY;AAAA,yBACO,IAAI;AAAA,wBACL,IAAI;AAAA,yBACH,IAAI,IAAI;AAAA,wBACT,IAAI,IAAI;AAAA,oBACZ;AAAA,IAChB;AACA,gBAAY;AAAA,2BACW;AAAA,8BACG,SAAS;AAAA;AAEnC,aAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAY;AAAA,oBACE,IAAI;AAAA,oBACJ,IAAI;AAAA,oBACJ,IAAI,IAAI;AAAA,oBACR,IAAI,IAAI;AAAA,eACb;AAAA,IACX;AACA,gBAAY;AAAA;AAAA;AAAA;AAIZ,aAAS,SAAS,GAAG,UAAU,eAAe,KAAK,GAAG,UAAU;AAC9D,YAAM,WAAW,SAAS;AAC1B,kBAAY;AAAA,6BACW,WAAW;AAAA;AAElC,UAAI,gBAAgB,GAAG;AACrB,YAAI,WAAW,aAAa;AAC1B,cAAI,UAAU,MAAM,GAAG;AACrB,wBAAY;AAAA;AAAA,uEAE+C;AAAA,4BAC3C;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKE;AAAA;AAAA,4BAEF;AAAA;AAAA;AAGhB,gBAAI,kBAAkB,KAAK,WAAW,GAAG;AACvC,0BAAY;AAAA,qBACL,0BAA0B,WAAW,gBAAgB;AAAA;AAAA,YAE9D,OAAO;AACL,0BAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAYD,uCAAuC;AAAA;AAAA,yBAEvC,oCAAoC;AAAA;AAAA;AAAA,YAGjD;AAAA,UACF,OAAO;AACL,wBAAY;AAAA,2DACmC;AAAA,4BAC/B;AAAA;AAAA,8BAEE;AAAA;AAAA,4BAEF;AAAA;AAAA;AAAA,qBAGP,qBAAqB;AAAA;AAAA,UAEhC;AACA,cAAI,WAAW,IAAI,aAAa;AAC9B,kBAAM,kBAAkB,UAAU,MAAM,IAAI,aAAa,kBAAkB,aAAa,IAAI;AAC5F,gBAAI,gBAAgB,MAAM,KAAK,UAAU,MAAM,KAAK,gBAAgB,MAAM,KAAK,UAAU,MAAM,GAAG;AAChG,0BAAY;AAAA,wDAC8B;AAAA;AAAA,yEAEiB,WAAW;AAAA,8BACtD,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,gCAKT,WAAW;AAAA;AAAA,8BAEb,WAAW;AAAA;AAAA;AAG3B,kBAAI,gBAAgB,GAAG;AACrB,4BAAY;AAAA;AAAA;AAAA;AAAA,0BAIF,WAAW,gCAAgC,WAAW;AAAA;AAAA,0BAEtD,WAAW,6BAA6B,WAAW;AAAA;AAAA;AAAA,cAG/D,OAAO;AACL,4BAAY;AAAA,yBACH,WAAW,mBAAmB,uBAAuB,WAAW;AAAA;AAAA,cAE3E;AAAA,YACF,OAAO;AACL,kBAAI,oBAAoB,GAAG;AACzB,4BAAY;AAAA,yBACH,WAAW,cAAc;AAAA;AAAA,cAEpC,OAAO;AACL,4BAAY;AAAA,uCACW;AAAA;AAAA,2EAEoC,WAAW;AAAA,gCACtD,WAAW;AAAA;AAAA,kCAET,WAAW;AAAA;AAAA,gCAEb,WAAW;AAAA;AAAA;AAAA,yBAGlB,WAAW,cAAc,WAAW;AAAA;AAAA,cAE/C;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF,OAAO;AACL,YAAI,WAAW,aAAa;AAC1B,cAAI,UAAU,MAAM,GAAG;AACrB,wBAAY;AAAA;AAAA,sEAE8C;AAAA,4BAC1C;AAAA;AAAA;AAAA;AAAA,8BAIE;AAAA;AAAA,4BAEF;AAAA;AAAA;AAAA,kEAGsC,WAAW;AAAA,4BACjD,WAAW;AAAA;AAAA;AAAA;AAAA,8BAIT,WAAW;AAAA;AAAA,4BAEb,WAAW;AAAA;AAAA;AAAA,qBAGlB,0BAA0B,uBAAuB,WAAW;AAAA;AAErE,gBAAI,WAAW,IAAI,aAAa;AAC9B,0BAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uBAMH,WAAW,mBAAmB,WAAW;AAAA;AAAA,YAEpD;AAAA,UACF,OAAO;AACL,wBAAY;AAAA,0DACkC;AAAA,4BAC9B;AAAA;AAAA,8BAEE;AAAA;AAAA,4BAEF;AAAA;AAAA;AAAA;AAAA,sEAI0C,WAAW;AAAA,4BACrD,WAAW;AAAA;AAAA,8BAET,WAAW;AAAA;AAAA,4BAEb,WAAW;AAAA;AAAA;AAAA,qBAGlB;AAAA,4BACO,uBAAuB,WAAW;AAAA;AAElD,gBAAI,WAAW,IAAI,aAAa;AAC9B,0BAAY;AAAA,uBACH,WAAW,mBAAmB,uBAAuB,WAAW;AAAA;AAAA,YAE3E;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,WAAW,aAAa;AAC1B,oBAAY;AAAA,gCACY;AAAA,4BACJ;AAAA,2BACD,SAAS;AAAA,8BACN;AAAA;AAAA;AAGtB,YAAI,WAAW,IAAI,aAAa;AAC9B,sBAAY;AAAA,kCACY,WAAW;AAAA,8BACf,WAAW;AAAA,6BACZ,SAAS;AAAA,gCACN,WAAW;AAAA;AAAA;AAAA,QAGnC;AAAA,MACF;AAAA,IACF;AACA,gBAAY;AAAA;AAAA;AAGZ,gBAAY;AAAA;AAAA;AAGZ,gBAAY;AAAA;AAAA;AAGZ,QAAI,oBAAoB,IAAI,yBAAyB;AACrD,QAAI,aAAa;AACf,UAAI,oBAAoB;AACtB,4BAAoB;AAAA;AAAA,aAEf;AAAA;AAAA,MAEP,WAAW,mBAAmB;AAC5B,4BAAoB;AAAA;AAAA,aAEf;AAAA;AAAA,MAEP,OAAO;AACL,4BAAoB;AAAA,aACf;AAAA;AAAA,MAEP;AACA,+BAAyB;AAAA,IAC3B;AACA,UAAM,iBAAiB,UAAU,oCAAoC;AACrE,QAAI,SAAS;AACX,WAAK,cAAc,KAAK,MAAM;AAAA,IAChC;AACA,QAAI,oBAAoB;AACtB,WAAK,cAAc,KAAK,wBAAwB;AAAA,IAClD;AACA,QAAI,mBAAmB;AACrB,WAAK,cAAc,KAAK,gBAAgB;AAAA,IAC1C;AACA,SAAK,WAAW;AAAA,SACX;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,WAaE;AAAA;AAAA;AAAA,WAGA;AAAA,WACA;AAAA;AAAA;AAAA;AAAA,EAIT;AACF;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,aAAa,UAAU;AACjC,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,cAAc,MAAM,QAAQ;AAAA,MACpC,EAAE,MAAM,OAAO,MAAM,QAAQ;AAAA,MAC7B,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,MAChC,EAAE,MAAM,YAAY,MAAM,QAAQ;AAAA,MAClC,EAAE,MAAM,cAAc,MAAM,MAAM;AAAA,MAClC,EAAE,MAAM,oBAAoB,MAAM,MAAM;AAAA,MACxC,EAAE,MAAM,YAAY,MAAM,MAAM;AAAA,IAClC;AACA,SAAK,cAAc;AACnB,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,EAAE,WAAW,IAAI;AACvB,UAAM,OAAO,mBAAmB;AAChC,UAAM,iBAAiB,eAAe;AACtC,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,wBAAwB,KAAK,sBAAsB,wDAAwD,mBAAmB,YAAY,eAAe,YAAY;AAC3K,QAAI,WAAW;AACf,aAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,eAAS,MAAM,GAAG,OAAO,GAAG,OAAO;AACjC,oBAAY;AAAA,gCACY;AAAA,yBACP;AAAA;AAAA,YAEb;AAAA;AAAA;AAAA;AAAA,iCAIqB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAQE;AAAA;AAAA;AAAA;AAAA,sBAIb;AAAA;AAAA,2BAEK,MAAM,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA,2BAKV,MAAM,IAAI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAQ/B;AAAA,IACF;AACA,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UASV;AAAA;AAAA,UAEA,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,SAAS,uBAAuB,OAAO,gBAAgB;AACrD,QAAM,SAAS,MAAM;AACrB,MAAI,UAAU,GAAG;AACf,WAAO,iBAAiB;AAAA,MACtB,GAAG,MAAM,MAAM,GAAG,EAAE;AAAA,MACpB,MAAM,SAAS,KAAK,MAAM,SAAS;AAAA,MACnC,MAAM,SAAS;AAAA,IACjB,IAAI;AAAA,MACF,GAAG,MAAM,MAAM,GAAG,EAAE;AAAA,MACpB,MAAM,SAAS;AAAA,MACf,MAAM,SAAS,KAAK,MAAM,SAAS;AAAA,IACrC;AAAA,EACF,WAAW,CAAC,kBAAkB,WAAW,KAAK,MAAM,KAAK,GAAG;AAC1D,WAAO,CAAC,MAAM,IAAI,CAAC;AAAA,EACrB,OAAO;AACL,WAAO;AAAA,EACT;AACF;AACA,SAAS,eAAe,EAAE,GAAG,QAAQ,UAAU,SAAS,UAAU,OAAO,MAAM,yBAAyB,MAAM,iBAAiB,GAAG,YAAY,cAAc,KAAK,GAAG;AAClK,QAAM,SAAS,EAAE;AACjB,QAAM,WAAW,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC9C,QAAM,kBAAkB,SAAS;AACjC,QAAM,cAAc,OAAO,KAAK,OAAO,KAAK,OAAO;AACnD,QAAM,mBAAmB,SAAS;AAClC,QAAM,iBAAiB,SAAS,eAAe;AAC/C,QAAM,aAAa;AACnB,QAAM,aAAa;AACnB,MAAI;AACJ,QAAM,gBAAgB,CAAC;AACvB,MAAI,0BAA0B,MAAM;AAClC,UAAM,cAAc,uBAAuB,uBAAuB,OAAO,cAAc;AACvF,QAAI,eAAe,MAAM;AACvB,+BAAyB,SAAS;AAAA,QAChC,QAAQ,EAAE,GAAG,uBAAuB;AAAA,QACpC,SAAS;AAAA,QACT,OAAO,EAAE,OAAO,YAAY;AAAA,MAC9B,CAAC;AACD,oBAAc,KAAK,sBAAsB;AAAA,IAC3C;AAAA,EACF;AACA,MAAI,QAAQ,MAAM;AAChB,UAAM,cAAc,uBAAuB,KAAK,OAAO,cAAc;AACrE,QAAI,eAAe,MAAM;AACvB,aAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,KAAK,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AACzF,oBAAc,KAAK,IAAI;AAAA,IACzB;AAAA,EACF;AACA,QAAM,6BAA6B,gBAAgB,KAAK,qBAAqB,MAAM,kBAAkB;AACrG,QAAM,cAAc,CAAC,6BAA6B,SAAS,YAAY,kBAAkB,SAAS,WAAW,QAAQ,OAAO,KAAK,MAAM,KAAK,aAAa,YAAY,SAAS,MAAM,MAAM,EAAE,GAAG,OAAO,MAAM,EAAE,CAAC;AAC/M,MAAI,aAAa;AACf,UAAM,cAAc,OAAO,KAAK,OAAO,MAAM,OAAO,KAAK;AACzD,UAAM,YAAY;AAAA,MAChB,QAAQ,EAAE;AAAA,MACV,OAAO,CAAC,GAAG,aAAa,SAAS,UAAU;AAAA,MAC3C,OAAO,EAAE;AAAA,IACX;AACA,UAAM,wBAAwB,SAAS;AACvC,aAAS,QAAQ,SAAS,MAAM,MAAM;AACtC,aAAS,MAAM,SAAS,MAAM,SAAS;AACvC,iBAAa,OAAO,cAAc,SAAS,OAAO,UAAU,KAAK,GAAG,MAAM,kBAAkB,SAAS,YAAY,UAAU,kBAAkB;AAC7I,UAAM,iBAAiB,SAAS;AAAA,MAC9B,QAAQ,EAAE,GAAG,OAAO;AAAA,MACpB,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,CAAC,GAAG,SAAS,YAAY,SAAS,WAAW,EAAE;AAAA,IACjE,CAAC;AACD,kBAAc,KAAK,cAAc;AACjC,UAAM,gBAAgB,gBAAgB;AAAA,MACpC,GAAG;AAAA,MACH,GAAG;AAAA,MACH,SAAS;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,MACA,YAAY;AAAA,MACZ;AAAA,MACA;AAAA,IACF,CAAC;AACD,UAAM,uBAAuB,SAAS,QAAQ,IAAI,cAAc,MAAM;AACtE,iBAAa,OAAO,qBAAqB,UAAU,MAAM,6CAA6C;AACtG,aAAS,QAAQ;AACjB,yBAAqB,QAAQ,SAAS;AACtC,UAAM,UAAU,EAAE,QAAQ,EAAE,GAAG,cAAc,GAAG,SAAS,SAAS,CAAC;AACnE,QAAI,QAAQ,SAAS;AACrB,kBAAc,KAAK,aAAa;AAAA,EAClC,OAAO;AACL,UAAM,UAAU,SAAS,YAAY,SAAS;AAC9C,UAAM,YAAY,SAAS;AAAA,MACzB,QAAQ,EAAE,EAAE;AAAA,MACZ,SAAS;AAAA,MACT,OAAO;AAAA,QACL,OAAO,iBAAiB,CAAC,SAAS,WAAW,SAAS,SAAS,UAAU,IAAI,CAAC,SAAS,WAAW,SAAS,YAAY,OAAO;AAAA,MAChI;AAAA,IACF,CAAC;AACD,UAAM,iBAAiB,SAAS;AAAA,MAC9B,QAAQ,EAAE,GAAG,OAAO;AAAA,MACpB,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,CAAC,GAAG,SAAS,YAAY,SAAS,WAAW,EAAE;AAAA,IACjE,CAAC;AACD,UAAM,SAAS,gBAAgB;AAAA,MAC7B,GAAG,iBAAiB,YAAY;AAAA,MAChC,GAAG,iBAAiB,iBAAiB;AAAA,MACrC,YAAY,CAAC;AAAA,MACb;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,YAAY;AAAA,MACZ;AAAA,MACA;AAAA,IACF,CAAC;AACD,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,SAAS,EAAE,CAAC;AAChG,kBAAc,KAAK,SAAS;AAC5B,kBAAc,KAAK,cAAc;AACjC,kBAAc,KAAK,MAAM;AAAA,EAC3B;AACA,aAAW,KAAK,eAAe;AAC7B,aAAS,8BAA8B,CAAC;AAAA,EAC1C;AACA,SAAO;AACT;AACA,SAAS,iBAAiB,EAAE,GAAG,QAAQ,UAAU,SAAS,UAAU,OAAO,MAAM,yBAAyB,MAAM,iBAAiB,GAAG,YAAY,cAAc,KAAK,GAAG;AACpK,QAAM,EAAE,aAAa,cAAc,YAAY,UAAU,WAAW,WAAW,IAAI;AACnF,QAAM,iBAAiB,eAAe;AACtC,QAAM,YAAY,cAAc,eAAe;AAC/C,QAAM,UAAU,YAAY;AAC5B,QAAM,aAAa,CAAC,SAAS,WAAW,WAAW,OAAO;AAC1D,QAAM,aAAa;AACnB,QAAM,aAAa;AACnB,QAAM,gBAAgB,CAAC;AACvB,MAAI,0BAA0B,MAAM;AAClC,UAAM,cAAc,uBAAuB,uBAAuB,OAAO,cAAc;AACvF,QAAI,eAAe,MAAM;AACvB,+BAAyB,SAAS;AAAA,QAChC,QAAQ,EAAE,GAAG,uBAAuB;AAAA,QACpC,SAAS;AAAA,QACT,OAAO,EAAE,OAAO,YAAY;AAAA,MAC9B,CAAC;AACD,oBAAc,KAAK,sBAAsB;AAAA,IAC3C;AAAA,EACF;AACA,MAAI,QAAQ,MAAM;AAChB,UAAM,cAAc,uBAAuB,KAAK,OAAO,cAAc;AACrE,QAAI,eAAe,MAAM;AACvB,aAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,KAAK,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AACzF,oBAAc,KAAK,IAAI;AAAA,IACzB;AAAA,EACF;AACA,QAAM,QAAQ,SAAS;AAAA,IACrB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,GAAG,WAAW,aAAa,cAAc,OAAO,KAAK,IAAI,SAAS,EAAE;AAAA,EACvF,CAAC;AACD,gBAAc,KAAK,KAAK;AACxB,QAAM,gBAAgB,IAAI,oBAAoB,YAAY,QAAQ;AAClE,QAAM,eAAe;AAAA,IACnB,EAAE;AAAA,IACF,CAAC,SAAS,QAAQ,KAAK,SAAS,QAAQ,IAAI;AAAA,IAC5C,CAAC,SAAS,cAAc,SAAS,WAAW;AAAA,IAC5C,CAAC,SAAS,gBAAgB,SAAS,aAAa;AAAA,IAChD,CAAC,SAAS,UAAU;AAAA,IACpB,CAAC,SAAS,cAAc,SAAS,UAAU;AAAA,IAC3C,CAAC,SAAS,QAAQ;AAAA,EACpB;AACA,QAAM,SAAS,SAAS,gBAAgB,eAAe,CAAC,CAAC,GAAG,WAAW,YAAY;AACnF,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAC1G,gBAAc,KAAK,MAAM;AACzB,gBAAc,KAAK,cAAc;AACjC,QAAM,UAAU,QAAQ;AACxB,QAAM,4BAA4B,0BAA0B;AAC5D,QAAM,oBAAoB,gBAAgB;AAC1C,QAAM,kBAAkB,cAAc,6BAA6B,aAAa,IAAI,IAAI;AACxF,QAAM,gBAAgB,IAAI,oBAAoB,iBAAiB,eAAe,QAAQ,MAAM,OAAO,iBAAiB,MAAM,QAAQ,eAAe,OAAO,iBAAiB,CAAC,SAAS,WAAW,SAAS,SAAS,WAAW,IAAI,CAAC,SAAS,WAAW,SAAS,aAAa,OAAO,GAAG,YAAY,YAAY,SAAS,iBAAiB,2BAA2B,iBAAiB;AAClX,QAAM,SAAS,iBAAiB,CAAC,gBAAgB,KAAK,IAAI,CAAC,OAAO,cAAc;AAChF,MAAI,MAAM;AACR,WAAO,KAAK,IAAI;AAAA,EAClB;AACA,MAAI,2BAA2B;AAC7B,WAAO,KAAK,sBAAsB;AAAA,EACpC;AACA,MAAI,mBAAmB;AACrB,UAAM,kBAAkB,SAAS,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,gBAAgB,SAAS,CAAC;AACxH,WAAO,KAAK,eAAe;AAC3B,kBAAc,KAAK,eAAe;AAAA,EACpC;AACA,QAAM,UAAU,SAAS,gBAAgB,eAAe,QAAQ,SAAS;AACzE,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,SAAS,EAAE,CAAC;AACvG,gBAAc,KAAK,OAAO;AAC1B,aAAW,KAAK,eAAe;AAC7B,aAAS,8BAA8B,CAAC;AAAA,EAC1C;AACA,SAAO;AACT;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,gBAAgB,IAAI;AACvE,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,OAAO,WAAW;AAC5I,MAAI;AACJ,MAAI,SAAS,iBAAiB,KAAK,SAAS,gBAAgB,KAAK,SAAS,mBAAmB,KAAK,SAAS,kBAAkB,KAAK,SAAS,iBAAiB,KAAK,SAAS,gBAAgB,MAAM,SAAS,QAAQ,SAAS,UAAU,SAAS,QAAQ,SAAS,UAAU;AACtQ,UAAM,eAAe,EAAE,GAAG,QAAQ,UAAU,SAAS,SAAS,CAAC;AAAA,EACjE,WAAW,SAAS,eAAe,KAAK,gBAAgB,kBAAkB,IAAI,EAAE,QAAQ,gBAAgB,GAAG;AACzG,UAAM,UAAU,IAAI,oBAAoB,QAAQ;AAChD,UAAM,eAAe;AAAA,MACnB,CAAC,SAAS,QAAQ,KAAK,SAAS,QAAQ,IAAI;AAAA,MAC5C,CAAC,SAAS,cAAc,SAAS,WAAW;AAAA,MAC5C,CAAC,SAAS,gBAAgB,SAAS,aAAa;AAAA,MAChD,CAAC,SAAS,UAAU,SAAS,OAAO;AAAA,IACtC;AACA,UAAM,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,WAAW,YAAY;AAAA,EAC9E,WAAW,IAAI,EAAE,QAAQ,mBAAmB,GAAG;AAC7C,UAAM,iBAAiB,EAAE,GAAG,QAAQ,UAAU,SAAS,SAAS,CAAC;AAAA,EACnE,OAAO;AACL,UAAM,UAAU,IAAI,cAAc,QAAQ;AAC1C,UAAM,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,SAAS;AAAA,EAChE;AACA,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,SAAS,EAAE,CAAC;AAC3G,WAAS,8BAA8B,GAAG;AAC1C,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,KAAK,IAAI;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,iBAAiB,SAAS,eAAe;AAC/C,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAYU,SAAS;AAAA,kCACL,SAAS;AAAA,iCACV,kBAAkB;AAAA;AAAA,kCAEjB,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP,SAAS;AAAA,mCACV,iBAAiB;AAAA;AAAA,oCAEhB,SAAS;AAAA;AAAA;AAAA;AAAA,oBAIzB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBlB;AACF;AACA,IAAI,wBAAwB,MAAM;AAAA,EAChC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,MAAM,GAAG;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,iBAAiB,SAAS,eAAe;AAC/C,UAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,UAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,SAAS,iBAAiB,IAAI;AACpC,UAAM,aAAa,iBAAiB,IAAI;AACxC,SAAK,WAAW;AAAA,iCACa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKlB;AAAA;AAAA,wCAEc,mBAAmB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAO3B;AAAA,gDACgB;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKpB;AAAA;AAAA,kCAES;AAAA,kDACgB;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMpB;AAAA;AAAA,oCAES,SAAS;AAAA;AAAA,oBAEzB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBlB;AACF;AACA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,KAAK,IAAI;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,WAAW,SAAS,QAAQ;AAClC,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAWU,SAAS;AAAA,kCACL,SAAS;AAAA,iCACV,iBAAiB;AAAA;AAAA,kCAEhB,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP,SAAS;AAAA,mCACV,kBAAkB;AAAA;AAAA,oCAEjB,SAAS;AAAA;AAAA;AAAA;AAAA,sCAIP,SAAS;AAAA,qCACV,iBAAiB;AAAA;AAAA,sCAEhB,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAc7C;AACF;AACA,IAAI,wBAAwB,MAAM;AAAA,EAChC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,MAAM,GAAG;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,WAAW,cAAc,IAAI,SAAS,QAAQ;AACpD,UAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,UAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,SAAK,WAAW;AAAA,iCACa,aAAa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAczB;AAAA,gDACgB;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKpB;AAAA;AAAA,kCAES;AAAA,kDACgB;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMpB;AAAA;AAAA,oCAES;AAAA,oDACgB;AAAA;AAAA,wCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,6BAMpB;AAAA;AAAA,sCAES,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAW7C;AACF;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,iBAAiB,YAAY,IAAI;AACzE,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,GAAG,MAAM,iBAAiB,OAAO,WAAW;AACnI,QAAM,UAAU,IAAI,uBAAuB,QAAQ;AACnD,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,EAAE,GAAG,SAAS;AAC7D;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,YAAY,gBAAgB,IAAI;AACxE,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,GAAG,MAAM,iBAAiB,OAAO,WAAW;AACvI,QAAM,UAAU,IAAI,sBAAsB,QAAQ;AAClD,SAAO,SAAS,gBAAgB,SAAS,CAAC,IAAI,MAAM,GAAG,SAAS;AAClE;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,IAAI;AACvG,QAAM,UAAU,IAAI,cAAc,QAAQ;AAC1C,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,SAAS;AACjE;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,wBAAwB,MAAM;AACrC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,IAAI;AAC5C,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,GAAG,IAAI;AAC9F,QAAM,UAAU,IAAI,uBAAuB,QAAQ;AACnD,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,EAAE,GAAG,SAAS;AAC7D;AACA,IAAI,gCAAgC;AAAA,EAClC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,KAAK,MAAM,SAAS,WAAW,IAAI;AAC3C,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,GAAG,IAAI;AAClG,QAAM,UAAU,IAAI,sBAAsB,QAAQ;AAClD,SAAO,SAAS,gBAAgB,SAAS,CAAC,IAAI,MAAM,GAAG,SAAS;AAClE;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM,0BAA0B;AAAA;AAAA;AAGpC,IAAI,OAAO,iBAAiB,EAAE,WAAW,IAAI,CAAC;AAC9C,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AAAA;AAAA;AAAA;AAIX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,uBAAuB,MAAM;AAAA,EAC/B,YAAY,YAAY,UAAU,UAAU,QAAQ,oBAAoB;AACtE,SAAK,gBAAgB,CAAC,SAAS,SAAS,QAAQ;AAChD,SAAK,cAAc,CAAC;AACpB,UAAM,CAAC,OAAO,aAAa,YAAY,KAAK,IAAI;AAChD,UAAM,CAAC,QAAQ,IAAI;AACnB,UAAM,CAAC,YAAY,SAAS,IAAI;AAChC,SAAK,cAAc,CAAC,UAAU,YAAY,WAAW,KAAK;AAC1D,UAAM,WAAW,WAAW,aAAa,IAAI;AAC7C,UAAM,CAAC,kBAAkB,eAAe,IAAI,CAAC,GAAG,cAAc,OAAO,GAAG,aAAa,KAAK;AAC1F,UAAM,CAAC,aAAa,aAAa,GAAG,IAAI,aAAa,IAAI;AAAA,MACvD,IAAI,cAAc,MAAM,aAAa;AAAA,MACrC;AAAA,MACA,MAAM;AAAA,IACR,IAAI;AAAA,MACF;AAAA,MACA;AAAA,MACA,mBAAmB;AAAA,IACrB;AACA,UAAM,CAAC,YAAY,YAAY,GAAG,IAAI,YAAY,IAAI;AAAA,MACpD,IAAI,aAAa,MAAM,YAAY;AAAA,MACnC;AAAA,MACA,MAAM;AAAA,IACR,IAAI;AAAA,MACF;AAAA,MACA;AAAA,MACA,mBAAmB;AAAA,IACrB;AACA,SAAK,WAAW;AAAA,yCACqB;AAAA,wCACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,iCAgBP;AAAA;AAAA;AAAA;AAAA,+BAIF;AAAA,8BACD;AAAA;AAAA,uBAEP;AAAA,mCACY;AAAA,4BACP;AAAA;AAAA;AAAA,uBAGL;AAAA,mCACY;AAAA,4BACP;AAAA;AAAA;AAAA;AAAA;AAAA,aAKf;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAyBX;AACF;AAGA,IAAI,iBAAiB,CAAC,SAAS;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,OAAO,OAAO,IAAI;AACzC,QAAM,EAAE,UAAU,QAAQ,mBAAmB,IAAI;AACjD,QAAM,UAAU,IAAI,qBAAqB,OAAO,OAAO,MAAM,OAAO,UAAU,QAAQ,kBAAkB;AACxG,SAAO,SAAS,gBAAgB,SAAS,CAAC,QAAQ,OAAO,MAAM,GAAG,SAAS;AAC7E;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AAAA,CACH,SAAS,YAAY;AACpB,aAAW,UAAU;AACrB,aAAW,SAAS;AACtB,GAAG,cAAc,YAAY,CAAC,EAAE;AAChC,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,KAAK,aAAa,WAAW,UAAU;AACjD,SAAK,KAAK;AACV,SAAK,cAAc;AACnB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,MAAM,QAAQ,CAAC;AACvD,UAAM,OAAO,KAAK,YAAY;AAC9B,UAAM,UAAU,KAAK,OAAO,UAAU,OAAO,QAAQ;AACrD,UAAM,MAAM,YAAY,UAAU,QAAQ,WAAW,MAAM,UAAU,KAAK,EAAE;AAC5E,UAAM,SAAS,KAAK,YAAY,KAAK,YAAY,SAAS;AAC1D,QAAI,YAAY;AAChB,QAAI,YAAY;AAChB,QAAI,WAAW;AACb,kBAAY,WAAW,UAAU,SAAS,MAAM;AAChD,kBAAY,WAAW,YAAY;AAAA,IACrC,OAAO;AACL,kBAAY,WAAW,gBAAgB,WAAW;AAClD,kBAAY,WAAW,eAAe;AAAA,IACxC;AACA,SAAK,WAAW;AAAA;AAAA,UAEV,kBAAkB,IAAI;AAAA,oBACZ,cAAc,MAAM,UAAU,KAAK,EAAE;AAAA,sBACnC;AAAA;AAAA,cAER;AAAA,sBACQ;AAAA,YACV,cAAc,MAAM,UAAU,KAAK,EAAE;AAAA,gBACjC,KAAK,YAAY,WAAW,MAAM,UAAU,KAAK,EAAE;AAAA;AAAA;AAAA;AAAA;AAAA,EAKjE;AACF;AACA,SAAS,WAAW,MAAM,MAAM,KAAK;AACnC,MAAI,SAAS,GAAG;AACd,WAAO,GAAG;AAAA,EACZ,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG,WAAW;AAAA,EACvB,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG,WAAW,WAAW;AAAA,EAClC,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG,WAAW,WAAW,WAAW;AAAA,EAC7C,OAAO;AACL,UAAM,IAAI,MAAM,cAAc,gBAAgB,2BAA2B;AAAA,EAC3E;AACF;AACA,SAAS,cAAc,MAAM,MAAM,KAAK;AACtC,MAAI,SAAS,GAAG;AACd,WAAO,GAAG;AAAA,EACZ,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG;AAAA,EACZ,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG;AAAA,EACZ,WAAW,SAAS,GAAG;AACrB,WAAO,GAAG;AAAA,EACZ,OAAO;AACL,UAAM,IAAI,MAAM,cAAc,gBAAgB,2BAA2B;AAAA,EAC3E;AACF;AAGA,SAAS,QAAQ,KAAK,GAAG,UAAU,MAAM,WAAW,UAAU;AAC5D,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,KAAK;AACzE,MAAI,YAAY;AAChB,MAAI,eAAe,MAAM;AACvB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AAAA,EAC3F;AACA,QAAM,eAAe,qBAAqB,iBAAiB,GAAG,KAAK,EAAE;AACrE,MAAI,iBAAiB,QAAQ,GAAG;AAC9B,UAAM,IAAI,MAAM,mDAAmD,EAAE,MAAM,SAAS,kBAAkB,MAAM;AAAA,EAC9G;AACA,QAAM,OAAO,UAAU,MAAM;AAC7B,MAAI,SAAS,UAAU,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,SAAS,CAAC;AACtE,WAAS,IAAI,GAAG,KAAK,KAAK,KAAK,KAAK,KAAK,IAAI,CAAC,IAAI,GAAG,KAAK;AACxD,UAAM,UAAU,IAAI,WAAW,KAAK,UAAU,OAAO,OAAO,QAAQ;AACpE,UAAM,eAAe,CAAC,CAAC,CAAC,CAAC;AACzB,UAAM,aAAa;AACnB,aAAS,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,OAAO,YAAY;AAC/E,aAAS,8BAA8B,UAAU;AAAA,EACnD;AACA,MAAI,WAAW;AACb,UAAM,UAAU,IAAI,WAAW,KAAK,UAAU,OAAO,WAAW,QAAQ;AACxE,UAAM,aAAa;AACnB,aAAS,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,KAAK;AACjE,aAAS,8BAA8B,UAAU;AAAA,EACnD;AACA,MAAI,eAAe,MAAM;AACvB,UAAM,qBAAqB,qBAAqB,uBAAuB,WAAW;AAClF,UAAM,0BAA0B,WAAW,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,mBAAmB,EAAE,CAAC;AAC5H,aAAS,8BAA8B,MAAM;AAC7C,aAAS,8BAA8B,SAAS;AAChD,WAAO;AAAA,EACT;AACA,SAAO;AACT;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,SAAO,QAAQ,UAAU,MAAM,GAAG,UAAU,MAAM,WAAW,QAAQ;AACvE;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,SAAO,QAAQ,UAAU,KAAK,GAAG,UAAU,MAAM,WAAW,QAAQ;AACtE;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,MAAM,aAAa,IAAI;AAC/B,MAAI,EAAE,MAAM,WAAW,GAAG;AACxB,UAAM,QAAQ,SAAS,SAAS,EAAE,MAAM;AACxC,UAAM,cAAc,SAAS,SAAS,QAAQ,MAAM;AACpD,UAAM,UAAU,gBAAgB,OAAO,aAAa,QAAQ,OAAO,QAAQ,OAAO,IAAI;AACtF,WAAO,SAAS,eAAe,CAAC,IAAI,GAAG,QAAQ,OAAO,OAAO;AAAA,EAC/D,WAAW,EAAE,MAAM,WAAW,GAAG;AAC/B,UAAM,OAAO,SAAS,WAAW,CAAC;AAClC,UAAM,aAAa,SAAS,WAAW,OAAO;AAC9C,UAAM,SAAS,sBAAsB,MAAM,YAAY,MAAM,YAAY;AACzE,WAAO,SAAS,eAAe,OAAO,OAAO,QAAQ,OAAO,OAAO,MAAM;AAAA,EAC3E;AACA,QAAM,IAAI,MAAM,qEAAqE,EAAE,MAAM,SAAS;AACxG;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,aAAa,WAAW,YAAY;AAC9C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc;AACnB,SAAK,YAAY;AACjB,SAAK,aAAa;AAClB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA,gBAIJ,KAAK,qBAAqB;AAAA,gBAC1B,KAAK,oBAAoB;AAAA,gBACzB,KAAK,oBAAoB;AAAA;AAAA,uBAElB;AAAA,+BACQ;AAAA,uBACR;AAAA,+BACQ;AAAA,mCACI;AAAA,UACzB,KAAK,mBAAmB;AAAA;AAAA;AAAA,uBAGX,KAAK,uBAAuB;AAAA;AAAA;AAAA;AAAA,EAIjD;AAAA,EACA,uBAAuB;AACrB,QAAI,KAAK,eAAe,QAAQ;AAC9B,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,sBAAsB;AACpB,QAAI,KAAK,eAAe,QAAQ;AAC9B,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,sBAAsB;AACpB,QAAI,KAAK,eAAe,QAAQ;AAC9B,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA,qBAAqB;AACnB,QAAI,KAAK,eAAe,QAAQ;AAC9B,aAAO,KAAK,YAAY;AAAA,IAC1B,OAAO;AACL,aAAO,KAAK,YAAY;AAAA,IAC1B;AAAA,EACF;AAAA,EACA,yBAAyB;AACvB,QAAI,KAAK,eAAe,QAAQ;AAC9B,aAAO;AAAA,IACT,OAAO;AACL,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,WAAW,WAAW,IAAI;AAClC,QAAM,YAAY,EAAE,MAAM;AAC1B,QAAM,cAAc,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AACjE,QAAM,aAAa,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AAChE,QAAM,aAAa,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AAChE,QAAM,eAAe,cAAc;AACnC,QAAM,cAAc,aAAa;AACjC,QAAM,cAAc,cAAc,YAAY;AAC9C,QAAM,cAAc,eAAe,SAAS,CAAC,WAAW,cAAc,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,cAAc,WAAW;AACpJ,QAAM,UAAU,IAAI,oBAAoB,aAAa,WAAW,UAAU;AAC1E,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,UAAU,UAAU,OAAO,cAAc,MAAM,qBAAqB,OAAO,oBAAoB,OAAO;AAChH,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,QAAQ,MAAM,QAAQ;AAAA,MAC9B,EAAE,MAAM,WAAW,MAAM,QAAQ;AAAA,MACjC,EAAE,MAAM,aAAa,MAAM,QAAQ;AAAA,MACnC,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,IAClC;AACA,SAAK,cAAc,SAAS;AAC5B,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,aAAa,SAAS,cAAc,SAAS;AACnD,QAAI,oBAAoB,IAAI,yBAAyB;AACrD,QAAI,aAAa;AACf,UAAI,oBAAoB;AACtB,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,WAAW,mBAAmB;AAC5B,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,OAAO;AACL,4BAAoB;AAAA;AAAA,cAEd;AAAA;AAAA;AAAA,MAGR;AACA,+BAAyB;AAAA,IAC3B;AACA,UAAM,iBAAiB,UAAU,oCAAoC;AACrE,QAAI,SAAS;AACX,WAAK,cAAc,KAAK,MAAM;AAAA,IAChC;AACA,QAAI,oBAAoB;AACtB,WAAK,cAAc,KAAK,wBAAwB;AAAA,IAClD;AACA,QAAI,mBAAmB;AACrB,WAAK,cAAc,KAAK,gBAAgB;AAAA,IAC1C;AACA,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOgB;AAAA,4BACI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCASI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAOE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAcxB;AAAA,UACA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AAGA,IAAI,+BAA+B,MAAM;AAAA,EACvC,YAAY,UAAU,UAAU,OAAO,cAAc,MAAM,qBAAqB,OAAO,oBAAoB,OAAO;AAChH,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,QAAQ,MAAM,QAAQ;AAAA,MAC9B,EAAE,MAAM,WAAW,MAAM,QAAQ;AAAA,MACjC,EAAE,MAAM,aAAa,MAAM,QAAQ;AAAA,MACnC,EAAE,MAAM,UAAU,MAAM,QAAQ;AAAA,IAClC;AACA,SAAK,cAAc,SAAS;AAC5B,SAAK,sBAAsB,iBAAiB,KAAK,YAAY,MAAM;AACnE,UAAM,aAAa,SAAS,cAAc,SAAS;AACnD,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe;AACrB,QAAI,WAAW;AAAA;AAAA;AAGf,aAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAY;AAAA,wBACM,IAAI;AAAA,uBACL,IAAI;AAAA,wBACH,IAAI,IAAI;AAAA,uBACT,IAAI,IAAI;AAAA,mBACZ;AAAA,IACf;AACA,gBAAY;AAAA,0BACU;AAAA;AAEtB,aAAS,IAAI,GAAG,IAAI,aAAa,KAAK;AACpC,kBAAY;AAAA,mBACC,IAAI;AAAA,mBACJ,IAAI;AAAA,mBACJ,IAAI,IAAI;AAAA,mBACR,IAAI,IAAI;AAAA,cACb;AAAA,IACV;AACA,gBAAY;AAAA;AAAA;AAAA;AAIZ,aAAS,SAAS,GAAG,UAAU,eAAe,KAAK,GAAG,UAAU;AAC9D,YAAM,WAAW,SAAS;AAC1B,kBAAY;AAAA,4BACU,WAAW;AAAA;AAEjC,UAAI,gBAAgB,GAAG;AACrB,YAAI,WAAW,aAAa;AAC1B,cAAI,UAAU,MAAM,GAAG;AACrB,wBAAY;AAAA;AAAA,sEAE8C;AAAA,2BAC3C;AAAA;AAAA;AAAA;AAAA;AAAA,6BAKE;AAAA;AAAA,2BAEF;AAAA;AAAA;AAGf,gBAAI,kBAAkB,KAAK,WAAW,GAAG;AACvC,0BAAY;AAAA,oBACN,0BAA0B,WAAW,gBAAgB;AAAA;AAAA,YAE7D,OAAO;AACL,0BAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAYF,uCAAuC;AAAA;AAAA,wBAEvC,oCAAoC;AAAA;AAAA;AAAA,YAGhD;AAAA,UACF,OAAO;AACL,wBAAY;AAAA,0DACkC;AAAA,2BAC/B;AAAA;AAAA,6BAEE;AAAA;AAAA,2BAEF;AAAA;AAAA;AAAA,oBAGP,qBAAqB;AAAA;AAAA,UAE/B;AACA,cAAI,WAAW,IAAI,aAAa;AAC9B,kBAAM,kBAAkB,UAAU,MAAM,IAAI,aAAa,kBAAkB,aAAa,IAAI;AAC5F,gBAAI,gBAAgB,MAAM,KAAK,UAAU,MAAM,KAAK,gBAAgB,MAAM,KAAK,UAAU,MAAM,GAAG;AAChG,0BAAY;AAAA,uDAC6B;AAAA;AAAA,wEAEiB,WAAW;AAAA,6BACtD,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,+BAKT,WAAW;AAAA;AAAA,6BAEb,WAAW;AAAA;AAAA;AAG1B,kBAAI,gBAAgB,GAAG;AACrB,4BAAY;AAAA;AAAA;AAAA;AAAA,yBAIH,WAAW,gCAAgC,WAAW;AAAA;AAAA,yBAEtD,WAAW,6BAA6B,WAAW;AAAA;AAAA;AAAA,cAG9D,OAAO;AACL,4BAAY;AAAA,wBACJ,WAAW,mBAAmB,uBAAuB,WAAW;AAAA;AAAA,cAE1E;AAAA,YACF,OAAO;AACL,kBAAI,oBAAoB,GAAG;AACzB,4BAAY;AAAA,wBACJ,WAAW,cAAc;AAAA;AAAA,cAEnC,OAAO;AACL,4BAAY;AAAA,sCACU;AAAA;AAAA,0EAEoC,WAAW;AAAA,+BACtD,WAAW;AAAA;AAAA,iCAET,WAAW;AAAA;AAAA,+BAEb,WAAW;AAAA;AAAA;AAAA,wBAGlB,WAAW,cAAc,WAAW;AAAA;AAAA,cAE9C;AAAA,YACF;AAAA,UACF;AAAA,QACF;AAAA,MACF,OAAO;AACL,YAAI,WAAW,aAAa;AAC1B,cAAI,UAAU,MAAM,GAAG;AACrB,wBAAY;AAAA;AAAA,qEAE6C;AAAA,2BAC1C;AAAA;AAAA;AAAA;AAAA,6BAIE;AAAA;AAAA,2BAEF;AAAA;AAAA;AAAA,iEAGsC,WAAW;AAAA,2BACjD,WAAW;AAAA;AAAA;AAAA;AAAA,6BAIT,WAAW;AAAA;AAAA,2BAEb,WAAW;AAAA;AAAA;AAAA,oBAGlB,0BAA0B,uBAAuB,WAAW;AAAA;AAEpE,gBAAI,WAAW,IAAI,aAAa;AAC9B,0BAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sBAMJ,WAAW,mBAAmB,WAAW;AAAA;AAAA,YAEnD;AAAA,UACF,OAAO;AACL,wBAAY;AAAA,yDACiC;AAAA,2BAC9B;AAAA;AAAA,6BAEE;AAAA;AAAA,2BAEF;AAAA;AAAA;AAAA;AAAA,qEAI0C,WAAW;AAAA,2BACrD,WAAW;AAAA;AAAA,6BAET,WAAW;AAAA;AAAA,2BAEb,WAAW;AAAA;AAAA;AAAA,oBAGlB;AAAA,2BACO,uBAAuB,WAAW;AAAA;AAEjD,gBAAI,WAAW,IAAI,aAAa;AAC9B,0BAAY;AAAA,sBACJ,WAAW,mBAAmB,uBAAuB,WAAW;AAAA;AAAA,YAE1E;AAAA,UACF;AAAA,QACF;AAAA,MACF;AACA,UAAI,WAAW,aAAa;AAC1B,oBAAY;AAAA,+BACW;AAAA,2BACJ;AAAA;AAEnB,YAAI,WAAW,IAAI,aAAa;AAC9B,sBAAY;AAAA,iCACW,WAAW;AAAA,6BACf,WAAW;AAAA;AAAA,QAEhC;AAAA,MACF;AAAA,IACF;AACA,gBAAY;AAAA;AAAA;AAGZ,gBAAY;AAAA;AAAA;AAGZ,QAAI,oBAAoB,IAAI,yBAAyB;AACrD,QAAI,aAAa;AACf,UAAI,oBAAoB;AACtB,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,WAAW,mBAAmB;AAC5B,4BAAoB;AAAA;AAAA,YAEhB;AAAA;AAAA,MAEN,OAAO;AACL,4BAAoB;AAAA,YAChB;AAAA;AAAA,MAEN;AACA,+BAAyB;AAAA,IAC3B;AACA,UAAM,iBAAiB,UAAU,oCAAoC;AACrE,QAAI,SAAS;AACX,WAAK,cAAc,KAAK,MAAM;AAAA,IAChC;AACA,QAAI,oBAAoB;AACtB,WAAK,cAAc,KAAK,wBAAwB;AAAA,IAClD;AACA,QAAI,mBAAmB;AACrB,WAAK,cAAc,KAAK,gBAAgB;AAAA,IAC1C;AACA,SAAK,WAAW;AAAA,QACZ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOgB;AAAA,4BACI;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAOlB;AAAA;AAAA;AAAA,UAGA;AAAA,UACA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AAGA,SAAS,uBAAuB,MAAM;AACpC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,WAAW,gBAAgB,IAAI;AAC3D,MAAI,aAAa;AACjB,MAAI,cAAc,MAAM;AACtB,iBAAa,CAAC,GAAG,CAAC;AAAA,EACpB;AACA,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,UAAU,GAAG,MAAM,gFAAgF,0BAA0B,aAAa;AAC3N,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,YAAY,MAAM,iBAAiB,IAAI;AAC/H,MAAI;AACJ,MAAI,IAAI,EAAE,QAAQ,0BAA0B,KAAK,SAAS,eAAe,KAAK,SAAS,cAAc,SAAS,eAAe,GAAG;AAC9H,cAAU,IAAI,6BAA6B,QAAQ;AAAA,EACrD,OAAO;AACL,cAAU,IAAI,uBAAuB,QAAQ;AAAA,EAC/C;AACA,QAAM,eAAe;AAAA,IACnB,CAAC,SAAS,QAAQ,KAAK,SAAS,QAAQ,IAAI;AAAA,IAC5C,CAAC,SAAS,cAAc,SAAS,WAAW;AAAA,IAC5C,CAAC,SAAS,gBAAgB,SAAS,aAAa;AAAA,IAChD,CAAC,SAAS,UAAU,SAAS,OAAO;AAAA,EACtC;AACA,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,WAAW,YAAY;AAC/E;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,kCAAkC,MAAM;AAAA,EAC1C,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,KAAK,IAAI;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,SAAS,SAAS,QAAQ;AAChC,UAAM,UAAU,SAAS,QAAQ;AACjC,UAAM,aAAa,SAAS,cAAc,SAAS;AACnD,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBAOI;AAAA;AAAA;AAAA;AAAA;AAAA,8BAKM,SAAS;AAAA,kCACL,SAAS;AAAA,iCACV,kBAAkB;AAAA;AAAA,kCAEjB,SAAS;AAAA;AAAA;AAAA;AAAA,oCAIP,SAAS;AAAA,mCACV,iBAAiB;AAAA;AAAA,oCAEhB,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAa3C;AACF;AACA,IAAI,iCAAiC,MAAM;AAAA,EACzC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,MAAM,GAAG;AAC/B,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,UAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,UAAM,aAAa,SAAS,cAAc,SAAS;AACnD,SAAK,WAAW;AAAA,iCACa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAYZ;AAAA,gDACgB;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKpB;AAAA;AAAA,kCAES;AAAA,kDACgB;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAMpB;AAAA;AAAA;AAAA,oCAGS;AAAA,8BACN;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAU5B;AACF;AAGA,SAAS,qCAAqC,MAAM;AAClD,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,IAAI;AAClB,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,iBAAiB,YAAY,IAAI;AACxE,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,aAAa,SAAS,WAAW,MAAM,iBAAiB,IAAI;AAC7H,QAAM,UAAU,IAAI,gCAAgC,QAAQ;AAC5D,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,EAAE,GAAG,SAAS;AAC7D;AACA,IAAI,6CAA6C;AAAA,EAC/C,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,oCAAoC,MAAM;AACjD,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACvE,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,IAAI;AACjI,QAAM,UAAU,IAAI,+BAA+B,QAAQ;AAC3D,SAAO,SAAS,gBAAgB,SAAS,CAAC,IAAI,MAAM,GAAG,SAAS;AAClE;AACA,IAAI,4CAA4C;AAAA,EAC9C,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,MAAM;AAChB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC,MAAM,IAAI;AAC9B,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOlB;AACF;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,WAAW,CAAC,GAAG,EAAE,OAAO,GAAG,EAAE,KAAK;AACxC,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,OAAO,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,KAAK,EAAE,EAAE,CAAC;AACrF,QAAM,UAAU,IAAI,YAAY,KAAK;AACrC,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,IAAI,GAAG,KAAK,KAAK;AAChE,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC1F,WAAS,8BAA8B,IAAI;AAC3C,WAAS,8BAA8B,GAAG;AAC1C,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,KAAK,GAAG;AAC9B,SAAK,cAAc,SAAS;AAC5B,UAAM,EAAE,UAAU,SAAS,SAAS,cAAc,aAAa,cAAc,aAAa,gBAAgB,cAAc,IAAI;AAC5H,UAAM,EAAE,KAAK,QAAQ,MAAM,QAAQ,IAAI;AACvC,SAAK,WAAW;AAAA,oCACgB,iBAAiB;AAAA,iCACpB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAad;AAAA,iCACG;AAAA;AAAA,kCAEC;AAAA,kCACA;AAAA,qCACG;AAAA;AAAA,sCAEC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAiBpC;AACF;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,EAAE,SAAS,KAAK,MAAM,UAAU,IAAI;AAC1C,QAAM,WAAW,qBAAqB,sBAAsB,EAAE,OAAO,OAAO,OAAO,SAAS,MAAM,QAAQ,SAAS;AACnH,MAAI;AACJ,QAAM,UAAU,IAAI,kBAAkB,QAAQ;AAC9C,QAAM,SAAS,gBAAgB,SAAS,CAAC,GAAG,MAAM,GAAG,SAAS;AAC9D,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,SAAS,EAAE,CAAC;AAC3G,WAAS,8BAA8B,GAAG;AAC1C,SAAO;AACT;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,SAAS,IAAI;AACrB,QAAM,UAAU;AAChB,QAAM,EAAE,SAAS,YAAY,OAAO,IAAI,qBAAqB,qBAAqB,UAAU,QAAQ,MAAM;AAC1G,uBAAqB,oBAAoB,QAAQ,QAAQ,QAAQ,OAAO;AACxE,QAAM,EAAE,MAAM,MAAM,IAAI,qBAAqB,qBAAqB,YAAY,MAAM;AACpF,QAAM,SAAS,MAAM;AACrB,MAAI,MAAM;AACV,MAAI,mBAAmB,QAAQ;AAC/B,QAAM,mBAAmB,CAAC;AAC1B,WAAS,IAAI,GAAG,IAAI,QAAQ,EAAE,GAAG;AAC/B,eAAW,UAAU,MAAM,IAAI;AAC7B,YAAM,EAAE,oBAAoB,MAAM,YAAY,aAAa,IAAI,qBAAqB,qBAAqB,kBAAkB,OAAO,OAAO;AACzI,UAAI;AACJ,UAAI,qBAAqB,sBAAsB,IAAI,GAAG;AACpD,YAAI,QAAQ;AAAA,MACd,OAAO;AACL,YAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,QAAQ,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AACrF,yBAAiB,KAAK,CAAC;AAAA,MACzB;AACA,YAAM,cAAc,EAAE,MAAM,MAAM;AAClC,eAAS,IAAI,GAAG,IAAI,aAAa,QAAQ,EAAE,GAAG;AAC5C,oBAAY,OAAO,aAAa,IAAI,GAAG,CAAC;AAAA,MAC1C;AACA,UAAI,CAAC,aAAa,YAAY,EAAE,OAAO,WAAW,GAAG;AACnD,YAAI,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAChF,yBAAiB,KAAK,CAAC;AAAA,MACzB;AACA,UAAI,QAAQ,MAAM;AAChB,cAAM;AAAA,MACR,OAAO;AACL,cAAM,UAAU,EAAE,QAAQ,EAAE,GAAG,GAAG,GAAG,IAAI,GAAG,SAAS,SAAS,CAAC;AAC/D,yBAAiB,KAAK,GAAG;AAAA,MAC3B;AAAA,IACF;AACA,QAAI,IAAI,SAAS,GAAG;AAClB,UAAI,KAAK,MAAM,GAAG;AAChB,cAAM,KAAK;AAAA,UACT,QAAQ,EAAE,GAAG,IAAI;AAAA,UACjB,SAAS;AAAA,UACT,OAAO;AAAA,YACL,MAAM,KAAK,MAAM,QAAQ,SAAS;AAAA,YAClC,UAAU;AAAA,UACZ;AAAA,QACF,CAAC;AACD,yBAAiB,KAAK,GAAG;AAAA,MAC3B;AACA;AAAA,IACF;AAAA,EACF;AACA,aAAW,cAAc,kBAAkB;AACzC,QAAI,eAAe,KAAK;AACtB;AAAA,IACF;AACA,aAAS,8BAA8B,UAAU;AAAA,EACnD;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AACX,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUjB,IAAI,OAAO,iBAAiB,EAAE,WAAW,MAAM,iBAAiB,WAAW,CAAC;AAC5E,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU;AACd,IAAI,iBAAiB;AAAA;AAAA;AAAA;AAIrB,IAAI,WAAW,CAAC,SAAS;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,IAAI,EAAE,IAAI;AAClB,QAAM,UAAU,IAAI,EAAE,QAAQ,8BAA8B,IAAI,IAAI,sBAAsB,gBAAgB,GAAG,OAAO,EAAE,KAAK,IAAI,IAAI,gBAAgB,SAAS,GAAG,OAAO,EAAE,KAAK;AAC7K,SAAO,SAAS,gBAAgB,SAAS,CAAC,IAAI,CAAC,GAAG,GAAG,KAAK;AAC5D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,eAAe;AAAA;AAAA;AAGnB,IAAI,QAAQ;AACZ,IAAI,SAAS,kBAAkB;AAAA,EAC7B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,OAAO;AAAA,EACP,eAAe;AACjB,CAAC;AACD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AAAA;AAAA;AAAA;AAAA,cAII,qBAAqB;AAAA,eACpB,qBAAqB;AAAA,eACrB,qBAAqB;AAAA,eACrB,qBAAqB;AAAA,eACrB,qBAAqB;AAAA,eACrB,qBAAqB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAOpC,IAAI,OAAO,iBAAiB,EAAE,WAAW,IAAI,CAAC;AAC9C,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM,0BAA0B;AAAA;AAAA;AAGpC,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUjB,IAAI,OAAO,iBAAiB;AAAA,EAC1B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AAAA,EACf,OAAO;AACT,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,IAAI,IAAI;AAChB,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,YAAY,OAAO,MAAM;AAC/B,QAAM,WAAW,OAAO,MAAM,MAAM;AACpC,MAAI,OAAO;AACX,MAAI,MAAM,GAAG;AACX,iBAAa,OAAO,EAAE,YAAY,MAAM,KAAK,MAAM,iCAAiC,EAAE,YAAY,OAAO,YAAY;AACrH,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,WAAS,OAAO,MAAM,GAAG,CAAC;AAC1B,SAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC1F;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ;AACZ,IAAI,SAAS,iBAAiB,EAAE,WAAW,OAAO,iBAAiB,OAAO,eAAe,aAAa,CAAC;AACvG,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,WAAW,YAAY,SAAS;AAC1C,SAAK,gBAAgB,CAAC,QAAQ,MAAM;AACpC,UAAM,WAAW,WAAW;AAC5B,SAAK,cAAc;AACnB,UAAM,4BAA4B,UAAU,SAAS,KAAK,OAAO,UAAU,KAAK;AAChF,UAAM,oBAAoB,UAAU,GAAG,eAAe;AACtD,QAAI;AACJ,QAAI,cAAc,QAAQ;AACxB,iBAAW;AAAA,IACb,WAAW,cAAc,QAAQ;AAC/B,iBAAW;AAAA,IACb,OAAO;AACL,YAAM,IAAI,MAAM,sDAAsD,YAAY;AAAA,IACpF;AACA,SAAK,WAAW;AAAA,yCACqB;AAAA;AAAA;AAAA,UAG/B;AAAA;AAAA;AAAA;AAAA,kDAIwC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAMpB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yDAS2B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWvD;AACF;AAGA,SAAS,SAAS,GAAG,SAAS,UAAU;AACtC,QAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,QAAM,YAAY,aAAa,cAAc,EAAE,KAAK;AACpD,QAAM,qBAAqB,EAAE,MAAM,EAAE,MAAM,SAAS;AACpD,QAAM,QAAQ,YAAY;AAC1B,QAAM,UAAU,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,OAAO,kBAAkB,EAAE,EAAE,CAAC;AAC5G,QAAM,SAAS,QAAQ;AACvB,QAAM,cAAc,IAAI,WAAW,QAAQ,QAAQ,OAAO;AAC1D,QAAM,cAAc,IAAI,WAAW,QAAQ,QAAQ,OAAO;AAC1D,QAAM,SAAS;AAAA,IACb;AAAA,MACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,MACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,MACrC,OAAO;AAAA,IACT;AAAA,IACA;AAAA,MACE,QAAQ,MAAM,mBAAmB,KAAK;AAAA,MACtC,OAAO,MAAM,mBAAmB,KAAK;AAAA,MACrC,OAAO;AAAA,IACT;AAAA,EACF;AACA,QAAM,WAAW,SAAS,gBAAgB,aAAa,QAAQ,SAAS;AACxE,QAAM,WAAW,SAAS,gBAAgB,aAAa,QAAQ,SAAS;AACxE,QAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,MAAM,UAAU,MAAM,SAAS,GAAG,SAAS,SAAS,CAAC;AAChG,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,QAAQ;AAC/C,QAAM,wBAAwB,SAAS,EAAE,QAAQ,EAAE,GAAG,cAAc,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,EAAE,MAAM,EAAE,CAAC;AACrH,WAAS,8BAA8B,OAAO;AAC9C,WAAS,8BAA8B,aAAa;AACpD,SAAO;AACT;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,SAAO,SAAS,QAAQ,OAAO,QAAQ;AACzC;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,OAAO,OAAO;AACxB,SAAK,cAAc,CAAC;AACpB,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,MAAM,QAAQ,CAAC;AACvD,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMlB;AACF;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,IAAI;AACzB,MAAI,EAAE,MAAM,IAAI;AAChB,UAAQ,SAAS,aAAa,WAAW,KAAK;AAC9C,MAAI,UAAU,UAAU;AACtB,UAAM,SAAS,aAAa,kBAAkB,OAAO,aAAa,cAAc,KAAK,CAAC;AACtF,WAAO,KAAK,KAAK;AACjB,WAAO,SAAS,eAAe,OAAO,OAAO,MAAM;AAAA,EACrD,OAAO;AACL,UAAM,UAAU,IAAI,YAAY,OAAO,KAAK;AAC5C,UAAM,eAAe,CAAC,CAAC,KAAK,CAAC;AAC7B,WAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,OAAO,YAAY;AAAA,EAClE;AACF;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,uBAAuB,MAAM;AAAA,EAC/B,YAAY,YAAY;AACtB,SAAK,gBAAgB,CAAC,OAAO;AAC7B,SAAK,cAAc,CAAC;AACpB,UAAM,aAAa,WAAW;AAC9B,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,yBAKK;AAAA;AAAA,uCAEc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQrC;AACF;AAGA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,SAAS,SAAS,MAAM;AAC7C,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,eAAe;AACrB,UAAM,UAAU,IAAI,qBAAqB,OAAO,KAAK;AACrD,UAAM,SAAS,aAAa,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,KAAK;AAC3E,WAAO;AAAA,EACT;AACF;AAGA,IAAI,QAAQ;AACZ,IAAI,SAAS,iBAAiB,EAAE,WAAW,OAAO,iBAAiB,OAAO,eAAe,aAAa,CAAC;AACvG,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWd,IAAI,iBAAiB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAsBrB,IAAI,YAAY,kBAAkB,EAAE,WAAW,SAAS,iBAAiB,gBAAgB,OAAO,QAAQ,CAAC;AACzG,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,oBAAoB,MAAM;AAAA,EAC5B,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,OAAO,mBAAmB;AAChC,UAAM,CAAC,QAAQ,KAAK,IAAI;AACxB,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,uDAMmC,YAAY;AAAA;AAAA,wBAE3C,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAe3B;AACF;AAGA,IAAI,0BAA0B,MAAM;AAAA,EAClC,YAAY,aAAa;AACvB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,UAAM,OAAO,mBAAmB;AAChC,UAAM,CAAC,QAAQ,KAAK,IAAI;AACxB,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAeU,YAAY;AAAA,4BACd,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAgBvB,KAAK;AAAA;AAAA;AAAA,EAGb;AACF;AAGA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AACA,IAAI;AACJ,IAAI,qBAAqB,IAAI,EAAE,QAAQ,uCAAuC;AAC9E,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,MAAI,EAAE,OAAO,IAAI;AACjB,QAAM,EAAE,YAAY,IAAI;AACxB,QAAM,UAAU,OAAO,qBAAqB,eAAe,kBAAkB;AAC7E,QAAM,UAAU,OAAO,qBAAqB,eAAe,kBAAkB;AAC7E,QAAM,CAAC,OAAO,MAAM,IAAI,UAAU;AAAA,IAChC,OAAO;AAAA,IACP,OAAO;AAAA,EACT,IAAI,CAAC,OAAO,OAAO,OAAO,MAAM;AAChC,QAAM,WAAW,CAAC,QAAQ,KAAK;AAC/B,QAAM,WAAW,CAAC,QAAQ,OAAO,WAAW;AAC5C,MAAI,WAAW,SAAS;AACtB,UAAM,wBAAwB,IAAI,EAAE,QAAQ,uCAAuC;AACnF,QAAI,wBAAwB,QAAQ,0BAA0B,oBAAoB;AAChF,2BAAqB;AACrB,6BAAuB,SAAS,cAAc,QAAQ,EAAE,WAAW,MAAM,EAAE,mBAAmB,CAAC;AAAA,IACjG;AACA,yBAAqB,OAAO,QAAQ;AACpC,yBAAqB,OAAO,SAAS;AACrC,yBAAqB,UAAU,QAAQ,GAAG,GAAG,OAAO,MAAM;AAC1D,aAAS,qBAAqB;AAAA,EAChC;AACA,QAAM,kBAAkB,SAAS,eAAe,UAAU,OAAO;AACjE,WAAS,QAAQ,IAAI,gBAAgB,MAAM,EAAE,QAAQ,aAAa;AAClE,WAAS,MAAM,yBAAyB,SAAS,WAAW,gBAAgB,MAAM,GAAG,MAAM;AAC3F,QAAM,UAAU,IAAI,EAAE,QAAQ,YAAY,IAAI,IAAI,wBAAwB,QAAQ,IAAI,IAAI,kBAAkB,QAAQ;AACpH,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,eAAe,GAAG,OAAO;AACxE,WAAS,YAAY,gBAAgB,MAAM;AAC3C,SAAO;AACT;AAGA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,WAAW,iBAAiB,YAAY,aAAa,eAAe,IAAI;AAChH,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,OAAO,WAAW;AAC5I,MAAI;AACJ,QAAM,gBAAgB,CAAC;AACvB,QAAM,UAAU,QAAQ;AACxB,QAAM,4BAA4B,0BAA0B;AAC5D,QAAM,oBAAoB,gBAAgB;AAC1C,QAAM,gBAAgB,MAAM;AAC1B,UAAM,UAAU,CAAC,GAAG,MAAM;AAC1B,UAAM,2BAA2B,CAAC,QAAQ,gBAAgB;AACxD,UAAI,gBAAgB,UAAU,OAAO,MAAM,WAAW,KAAK,OAAO,MAAM,OAAO,GAAG;AAChF,cAAM,eAAe,SAAS;AAAA,UAC5B,QAAQ,EAAE,GAAG,OAAO;AAAA,UACpB,SAAS;AAAA,UACT,OAAO,EAAE,OAAO,CAAC,OAAO,MAAM,IAAI,GAAG,CAAC,EAAE;AAAA,QAC1C,CAAC;AACD,sBAAc,KAAK,YAAY;AAC/B,eAAO;AAAA,MACT;AACA,aAAO;AAAA,IACT;AACA,QAAI,SAAS;AACX,cAAQ,KAAK,yBAAyB,MAAM,UAAU,CAAC;AAAA,IACzD;AACA,QAAI,2BAA2B;AAC7B,cAAQ,KAAK,yBAAyB,wBAAwB,UAAU,CAAC;AAAA,IAC3E;AACA,QAAI,mBAAmB;AACrB,YAAM,kBAAkB,SAAS,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,gBAAgB,SAAS,CAAC;AACxH,cAAQ,KAAK,eAAe;AAC5B,oBAAc,KAAK,eAAe;AAAA,IACpC;AACA,WAAO;AAAA,EACT;AACA,MAAI,SAAS,iBAAiB,KAAK,SAAS,gBAAgB,KAAK,SAAS,mBAAmB,KAAK,SAAS,kBAAkB,KAAK,SAAS,iBAAiB,KAAK,SAAS,gBAAgB,MAAM,SAAS,QAAQ,SAAS,UAAU,SAAS,QAAQ,SAAS,UAAU;AACtQ,UAAM,eAAe;AAAA,MACnB;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,YAAY;AAAA,MACZ;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH,WAAW,SAAS,eAAe,KAAK,gBAAgB,kBAAkB,IAAI,EAAE,QAAQ,gBAAgB,GAAG;AACzG,UAAM,kBAAkB,cAAc,6BAA6B,aAAa,IAAI,IAAI;AACxF,UAAM,UAAU,IAAI,oBAAoB,UAAU,SAAS,iBAAiB,2BAA2B,iBAAiB;AACxH,UAAM,eAAe;AAAA,MACnB,CAAC,SAAS,QAAQ,KAAK,SAAS,QAAQ,IAAI;AAAA,MAC5C,CAAC,SAAS,cAAc,SAAS,WAAW;AAAA,MAC5C,CAAC,SAAS,gBAAgB,SAAS,aAAa;AAAA,MAChD,CAAC,SAAS,UAAU,SAAS,OAAO;AAAA,IACtC;AACA,UAAM,UAAU,cAAc;AAC9B,UAAM,SAAS,gBAAgB,SAAS,SAAS,WAAW,YAAY;AAAA,EAC1E,WAAW,IAAI,EAAE,QAAQ,mBAAmB,GAAG;AAC7C,UAAM,iBAAiB;AAAA,MACrB;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,MACT;AAAA,MACA,YAAY;AAAA,MACZ;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH,OAAO;AACL,UAAM,kBAAkB,cAAc,6BAA6B,aAAa,KAAK,IAAI;AACzF,UAAM,UAAU,IAAI,cAAc,UAAU,SAAS,iBAAiB,2BAA2B,iBAAiB;AAClH,UAAM,UAAU,cAAc;AAC9B,UAAM,SAAS,gBAAgB,SAAS,SAAS,SAAS;AAAA,EAC5D;AACA,QAAM,cAAc,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,SAAS,EAAE,CAAC;AAC3G,gBAAc,KAAK,GAAG;AACtB,gBAAc,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AACtE,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,WAAW,iBAAiB,YAAY,aAAa,eAAe,IAAI;AACpG,QAAM,gBAAgB,CAAC;AACvB,MAAI,aAAa;AACjB,MAAI,cAAc,MAAM;AACtB,iBAAa,CAAC,GAAG,CAAC;AAAA,EACpB;AACA,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,UAAU,GAAG,MAAM,gFAAgF,0BAA0B,aAAa;AAC3N,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,YAAY,MAAM,iBAAiB,IAAI;AAC/H,QAAM,0BAA0B,IAAI,EAAE,QAAQ,0BAA0B,KAAK,SAAS,eAAe,KAAK,SAAS,cAAc,SAAS,eAAe;AACzJ,QAAM,kBAAkB,cAAc,6BAA6B,aAAa,uBAAuB,IAAI;AAC3G,QAAM,gBAAgB,CAAC,GAAG,MAAM;AAChC,QAAM,UAAU,QAAQ;AACxB,QAAM,4BAA4B,0BAA0B;AAC5D,QAAM,oBAAoB,gBAAgB;AAC1C,MAAI,SAAS;AACX,kBAAc,KAAK,IAAI;AAAA,EACzB;AACA,MAAI,2BAA2B;AAC7B,kBAAc,KAAK,sBAAsB;AAAA,EAC3C;AACA,MAAI,mBAAmB;AACrB,UAAM,kBAAkB,SAAS,eAAe,CAAC,GAAG,WAAW,aAAa,kBAAkB,gBAAgB,SAAS,CAAC;AACxH,kBAAc,KAAK,eAAe;AAClC,kBAAc,KAAK,eAAe;AAAA,EACpC;AACA,MAAI;AACJ,MAAI,yBAAyB;AAC3B,cAAU,IAAI,6BAA6B,UAAU,SAAS,iBAAiB,2BAA2B,iBAAiB;AAAA,EAC7H,OAAO;AACL,cAAU,IAAI,uBAAuB,UAAU,SAAS,iBAAiB,2BAA2B,iBAAiB;AAAA,EACvH;AACA,QAAM,eAAe;AAAA,IACnB,CAAC,SAAS,QAAQ,KAAK,SAAS,QAAQ,IAAI;AAAA,IAC5C,CAAC,SAAS,cAAc,SAAS,WAAW;AAAA,IAC5C,CAAC,SAAS,gBAAgB,SAAS,aAAa;AAAA,IAChD,CAAC,SAAS,UAAU,SAAS,OAAO;AAAA,EACtC;AACA,QAAM,SAAS,SAAS,gBAAgB,SAAS,eAAe,WAAW,YAAY;AACvF,gBAAc,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AACtE,SAAO;AACT;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,kBAAkB,MAAM;AAAA,EAC1B,YAAY,UAAU,SAAS,OAAO,aAAa;AACjD,SAAK,WAAW;AAChB,SAAK,UAAU;AACf,SAAK,cAAc;AACnB,SAAK,gBAAgB,CAAC,KAAK,SAAS;AACpC,SAAK,cAAc;AACnB,UAAM,QAAQ,kBAAkB,MAAM,MAAM;AAC5C,QAAI,WAAW;AAAA;AAEf,aAAS,IAAI,GAAG,IAAI,KAAK,UAAU,KAAK;AACtC,kBAAY;AAAA,gDAC8B;AAAA;AAAA,sDAEM,KAAK,YAAY;AAAA,oCACnC,KAAK,QAAQ;AAAA,IAC7C;AACA,SAAK,WAAW;AAAA;AAAA,YAER;AAAA;AAAA;AAAA;AAAA,YAIA;AAAA;AAAA;AAAA;AAAA;AAAA,EAKV;AACF;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,QAAQ,QAAQ,IAAI;AAC5B,QAAM,eAAe,QAAQ;AAC7B,QAAM,YAAY,aAAa,aAAa,SAAS;AACrD,QAAM,aAAa,aAAa,cAAc,OAAO,KAAK;AAC1D,QAAM,CAAC,aAAa,WAAW,WAAW,OAAO,IAAI,qBAAqB,mBAAmB,QAAQ,OAAO;AAC5G,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,WAAW,SAAS,EAAE,EAAE,CAAC;AACvH,QAAM,WAAW,SAAS;AAAA,IACxB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,aAAa,cAAc,OAAO,KAAK,IAAI,WAAW,SAAS,EAAE;AAAA,EACpF,CAAC;AACD,MAAI,SAAS,mBAAmB,CAAC,QAAQ,OAAO,CAAC,KAAK,OAAO,UAAU,UAAU;AAC/E,UAAM,cAAc,SAAS,SAAS,QAAQ,MAAM;AACpD,UAAM,YAAY,SAAS,WAAW,MAAM;AAC5C,UAAM,WAAW,gBAAgB,aAAa,WAAW,OAAO,OAAO,WAAW,WAAW,WAAW,SAAS,OAAO,OAAO,UAAU;AACzI,WAAO,SAAS,eAAe,aAAa,OAAO,OAAO,SAAS,MAAM;AAAA,EAC3E;AACA,QAAM,UAAU,IAAI,gBAAgB,WAAW,SAAS,CAAC,WAAW,SAAS,GAAG,OAAO,KAAK;AAC5F,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,UAAU,cAAc,GAAG,SAAS,KAAK;AACxF,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAClG,WAAS,8BAA8B,cAAc;AACrD,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,GAAG;AAC1C,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,QAAQ,aAAa;AAC/B,SAAK,gBAAgB,CAAC,KAAK,SAAS;AACpC,SAAK,cAAc;AACnB,SAAK,OAAO,YAAY;AACxB,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,UAAM,eAAe,iBAAiB,QAAQ,CAAC;AAC/C,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA;AAAA,oDAE0C,OAAO;AAAA,oCACvB;AAAA;AAAA;AAAA,EAGlC;AACF;AACA,SAAS,iBAAiB,QAAQ,MAAM;AACtC,QAAM,gBAAgB,CAAC,WAAW,WAAW,WAAW,SAAS;AACjE,QAAM,eAAe,CAAC;AACtB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,QAAI,MAAM,GAAG;AACX,mBAAa,KAAK,OAAO;AAAA,IAC3B,OAAO;AACL,mBAAa,KAAK,GAAG,cAAc,IAAI;AAAA,IACzC;AAAA,EACF;AACA,SAAO,aAAa,KAAK;AAC3B;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,MAAM,UAAU,IAAI;AAC5B,QAAM,aAAa,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AAC9D,MAAI,IAAI,EAAE,IAAI,OAAO,GAAG;AACtB,UAAM,cAAc,SAAS,SAAS,QAAQ,MAAM;AACpD,UAAM,UAAU,EAAE,MAAM;AACxB,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,YAAM,QAAQ,YAAY;AAC1B,mBAAa,OAAO,SAAS,UAAU,KAAK,SAAS,GAAG,MAAM,6BAA6B,uBAAuB,UAAU,IAAI;AAAA,IAClI;AAAA,EACF;AACA,QAAM,YAAY,qBAAqB,aAAa,yBAAyB,GAAG,SAAS,YAAY,SAAS;AAC9G,QAAM,cAAc,aAAa,cAAc,QAAQ,KAAK;AAC5D,QAAM,YAAY,CAAC;AACnB,QAAM,WAAW,SAAS;AAAA,IACxB,QAAQ,EAAE,EAAE;AAAA,IACZ,SAAS;AAAA,IACT,OAAO;AAAA,MACL,OAAO;AAAA,QACL,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,MACZ;AAAA,IACF;AAAA,EACF,CAAC;AACD,QAAM,eAAe,SAAS;AAAA,IAC5B,QAAQ,EAAE,GAAG,QAAQ;AAAA,IACrB,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,CAAC,UAAU,WAAW,cAAc,UAAU,SAAS,EAAE;AAAA,EAC3E,CAAC;AACD,YAAU,KAAK,QAAQ;AACvB,YAAU,KAAK,YAAY;AAC3B,QAAM,qBAAqB;AAAA,IACzB,UAAU;AAAA,IACV,UAAU;AAAA,IACV,cAAc,UAAU;AAAA,IACxB,UAAU;AAAA,EACZ;AACA,MAAI,SAAS,mBAAmB,CAAC,GAAG,OAAO,CAAC,KAAK,EAAE,UAAU,UAAU;AACrE,UAAM,aAAa,SAAS,WAAW,YAAY;AACnD,UAAM,OAAO,SAAS,WAAW,QAAQ;AACzC,UAAM,SAAS,gBAAgB,MAAM,YAAY,kBAAkB;AACnE,cAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,WAAO,SAAS,eAAe,UAAU,aAAa,OAAO,OAAO,OAAO,MAAM;AAAA,EACnF;AACA,QAAM,UAAU,IAAI,cAAc,SAAS,OAAO,kBAAkB;AACpE,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,UAAU,YAAY,GAAG,SAAS,KAAK;AACtF,YAAU,KAAK,GAAG;AAClB,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,UAAU,YAAY,EAAE,CAAC;AAC5G,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU;AACd,IAAI,iBAAiB;AAAA;AAAA;AAGrB,IAAI,WAAW,kBAAkB;AAAA,EAC/B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AAAA,EACf,OAAO;AACT,CAAC;AACD,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB;AACpB,IAAI,uBAAuB;AAAA;AAAA;AAG3B,IAAI,gBAAgB,kBAAkB;AAAA,EACpC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,OAAO;AAAA,EACP,eAAe;AACjB,CAAC;AACD,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,SAAO,SAAS,QAAQ,MAAM,QAAQ;AACxC;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,YAAY;AAChB,IAAI,YAAY,iBAAiB,EAAE,WAAW,WAAW,OAAO,OAAO,CAAC;AACxE,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS;AACb,IAAI,SAAS,iBAAiB,EAAE,WAAW,QAAQ,OAAO,OAAO,CAAC;AAClE,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS;AACb,IAAI,SAAS,iBAAiB,EAAE,WAAW,QAAQ,OAAO,OAAO,CAAC;AAClE,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AACX,IAAI,cAAc;AAAA;AAAA;AAGlB,IAAI,QAAQ,kBAAkB;AAAA,EAC5B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AAAA,EACf,OAAO;AACT,CAAC;AACD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa;AACjB,IAAI,oBAAoB;AAAA;AAAA;AAGxB,IAAI,aAAa,kBAAkB;AAAA,EACjC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AAAA,EACf,OAAO;AACT,CAAC;AACD,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,IAAI,IAAI;AAC7B,QAAM,UAAU,gBAAgB,OAAO,MAAM,GAAG;AAChD,SAAO,SAAS,eAAe,CAAC,QAAQ,MAAM,GAAG,WAAW,OAAO;AACrE;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM,0BAA0B;AAAA;AAAA;AAGpC,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AASjB,IAAI,OAAO,iBAAiB,EAAE,WAAW,KAAK,iBAAiB,YAAY,eAAe,WAAW,CAAC;AACtG,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,0BAA0B;AAAA;AAAA;AAGtC,IAAI,SAAS,iBAAiB,EAAE,WAAW,MAAM,CAAC;AAClD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc;AAClB,IAAI,qBAAqB;AAAA;AAAA;AAAA;AAAA;AAKzB,IAAI,cAAc,kBAAkB;AAAA,EAClC,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,OAAO;AACT,CAAC;AACD,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc;AAClB,IAAI,cAAc,iBAAiB,EAAE,WAAW,YAAY,CAAC;AAC7D,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa;AACjB,IAAI,oBAAoB;AAAA;AAAA;AAAA;AAAA;AAAA;AAMxB,IAAI,aAAa,kBAAkB,EAAE,WAAW,YAAY,iBAAiB,mBAAmB,OAAO,OAAO,CAAC;AAC/G,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,QAAQ,QAAQ,MAAM,OAAO,MAAM;AAC7C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC;AACpB,UAAM,MAAM;AACZ,UAAM,OAAO,OAAO,KAAK;AACzB,SAAK,cAAc;AACnB,QAAI;AACJ,UAAM,QAAQ,SAAS,iBAAiB;AACxC,QAAI,SAAS,KAAK;AAChB,oBAAc,eAAe;AAAA,IAC/B,WAAW,SAAS,GAAG;AACrB,oBAAc,QAAQ;AAAA,IACxB,OAAO;AACL,oBAAc,WAAW,mBAAmB;AAAA,IAC9C;AACA,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,wBASI,aAAa;AAAA;AAAA,oCAED;AAAA;AAAA;AAAA;AAAA;AAAA,0BAKV;AAAA;AAAA;AAAA;AAAA,EAIxB;AACF;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,QAAQ,QAAQ,MAAM,OAAO,MAAM;AAC7C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC;AACpB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,UAAM,MAAM;AACZ,UAAM,OAAO,OAAO,KAAK;AACzB,SAAK,cAAc;AACnB,QAAI;AACJ,UAAM,QAAQ,SAAS,iBAAiB;AACxC,QAAI,SAAS,KAAK;AAChB,oBAAc,eAAe;AAAA,IAC/B,WAAW,SAAS,GAAG;AACrB,oBAAc,QAAQ;AAAA,IACxB,OAAO;AACL,oBAAc,WAAW,mBAAmB;AAAA,IAC9C;AACA,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAQY,KAAK,YAAY;AAAA,gCACjB,KAAK,YAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,iCAehB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAWR,aAAa;AAAA;AAAA;AAAA,6DAGuB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAqBnB;AAAA;AAAA;AAAA;AAAA,EAIxC;AACF;AAGA,IAAI,MAAM,CAAC,SAAS;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,aAAa,MAAM,OAAO,KAAK,IAAI;AAC3C,QAAM,UAAU,IAAI,EAAE,QAAQ,0BAA0B,IAAI,IAAI,iBAAiB,EAAE,OAAO,aAAa,MAAM,OAAO,IAAI,IAAI,IAAI,WAAW,EAAE,OAAO,aAAa,MAAM,OAAO,IAAI;AAClL,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,YAAY,aAAa,MAAM,OAAO,MAAM;AACtD,SAAK,gBAAgB,CAAC,cAAc,eAAe,IAAI;AACvD,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc;AACnB,SAAK,QAAQ,WAAW;AACxB,SAAK,cAAc;AACnB,SAAK,OAAO;AACZ,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAQU,KAAK;AAAA,oDACiB;AAAA,yCACX,KAAK;AAAA,0BACpB;AAAA;AAAA;AAAA,sCAGY,KAAK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAelB,yBAAyB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yCAOT;AAAA,0BACf;AAAA;AAAA;AAAA;AAAA,0CAIgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAexC;AACF;AAGA,IAAI,UAAU,CAAC,SAAS;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,GAAG,IAAI;AACrB,QAAM,EAAE,aAAa,MAAM,OAAO,KAAK,IAAI;AAC3C,QAAM,UAAU,IAAI,eAAe,EAAE,OAAO,aAAa,MAAM,OAAO,IAAI;AAC1E,SAAO,SAAS,gBAAgB,SAAS,CAAC,GAAG,GAAG,EAAE,GAAG,EAAE,KAAK;AAC9D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,GAAG,aAAa,UAAU,UAAU;AACpD,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,YAAY,QAAQ;AAC1B,QAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,OAAO,CAAC,WAAW,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC1G,QAAM,UAAU,OAAO,eAAe,EAAE,OAAO,OAAO,QAAQ;AAC9D,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AACzG,WAAS,8BAA8B,aAAa;AACpD,WAAS,8BAA8B,OAAO;AAC9C,SAAO;AACT;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,kBAAkB,SAAS,IAAI;AACvC,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,kBAAkB,EAAE,KAAK;AACtE,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,QAAM,uBAAuB,gBAAgB;AAC7C,QAAM,qBAAqB,SAAS,mBAAmB,CAAC,CAAC,CAAC;AAC1D,MAAI,WAAW;AACf,MAAI,sBAAsB;AACxB,QAAI,oBAAoB;AACtB,YAAM,WAAW,SAAS,QAAQ,IAAI,SAAS,MAAM;AACrD,YAAM,SAAS,SAAS;AACxB,YAAM,WAAW,IAAI,MAAM,KAAK;AAChC,eAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,iBAAS,KAAK,EAAE,MAAM,aAAa;AAAA,MACrC;AACA,YAAM,iBAAiB,iBAAiB,QAAQ,EAAE,OAAO,EAAE,OAAO,cAAc,QAAQ;AACxF,iBAAW,SAAS,eAAe,UAAU,EAAE,KAAK;AACpD,YAAM,eAAe,SAAS,QAAQ,IAAI,SAAS,MAAM;AACzD,mBAAa,SAAS;AAAA,IACxB,OAAO;AACL,iBAAW,eAAe,GAAG,cAAc,QAAQ;AAAA,IACrD;AACA,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAAA,EACjE;AACA,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,aAAa,WAAW,IAAI,qBAAqB,0BAA0B,SAAS,OAAO,IAAI;AACtG,MAAI,WAAW;AACf,MAAI,UAAU;AACZ,eAAW,qBAAqB,qBAAqB,aAAa,QAAQ;AAAA,EAC5E;AACA,MAAI;AACJ,MAAI,oBAAoB;AACtB,UAAM,WAAW,SAAS,QAAQ,IAAI,SAAS,MAAM;AACrD,UAAM,SAAS,SAAS;AACxB,UAAM,YAAY,WAAW,QAAQ,aAAa,cAAc,WAAW,GAAG,UAAU,EAAE,KAAK;AAC/F,UAAM,SAAS,eAAe,UAAU,EAAE,KAAK;AAC/C,UAAM,UAAU,SAAS,QAAQ,IAAI,IAAI,MAAM;AAC/C,YAAQ,SAAS;AAAA,EACnB,OAAO;AACL,UAAM,SAAS,UAAU,aAAa,UAAU,QAAQ;AAAA,EAC1D;AACA,MAAI,sBAAsB;AACxB,aAAS,8BAA8B,QAAQ;AAAA,EACjD;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU,qBAAqB;AAAA;AAAA;AAGnC,IAAI,iBAAiB;AAAA;AAAA;AAAA;AAAA;AAAA,MAKf,2BAA2B;AAAA;AAAA;AAGjC,IAAI,WAAW,kBAAkB;AAAA,EAC/B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AACjB,CAAC;AACD,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,oBAAkB,GAAG,SAAS;AAC9B,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY;AAClB,eAAa,OAAO,qBAAqB,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AACjN,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,MAAI,SAAS,gBAAgB,KAAK,SAAS,iBAAiB,KAAK,aAAa,YAAY,SAAS,SAAS,SAAS,QAAQ,GAAG;AAC9H,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD;AACA,QAAM,iBAAiB,IAAI,cAAc,UAAU,OAAO,KAAK;AAC/D,SAAO,SAAS,gBAAgB,gBAAgB,CAAC,CAAC,GAAG,EAAE,KAAK;AAC9D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,YAAY,gBAAgB,IAAI;AACxE,QAAM,YAAY,CAAC,GAAG,GAAG,CAAC;AAC1B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,iBAAiB,UAAU;AAClI,QAAM,iBAAiB,IAAI,cAAc,UAAU,OAAO,KAAK;AAC/D,SAAO,SAAS,gBAAgB,gBAAgB,CAAC,CAAC,GAAG,EAAE,KAAK;AAC9D;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,MAAM;AAAA,EACnC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,MAAM,QAAQ;AACpC,SAAK,cAAc,SAAS;AAC5B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,iBAAiB,SAAS;AAChC,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,UAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,UAAM,YAAY,wBAAwB,uBAAuB;AACjE,SAAK,WAAW;AAAA,iCACa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAcZ;AAAA,kBACd;AAAA,gDAC8B;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKX;AAAA,kDACgB;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAOf;AAAA;AAAA;AAAA;AAAA,qCAIK;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASnC;AACF;AACA,IAAI,2BAA2B,MAAM;AAAA,EACnC,YAAY,UAAU;AACpB,SAAK,gBAAgB,CAAC,MAAM,QAAQ;AACpC,SAAK,cAAc,SAAS;AAC5B,UAAM,cAAc,SAAS;AAC7B,UAAM,eAAe,SAAS;AAC9B,UAAM,cAAc,SAAS;AAC7B,UAAM,gBAAgB,SAAS;AAC/B,UAAM,iBAAiB,SAAS;AAChC,UAAM,gBAAgB,SAAS;AAC/B,UAAM,uBAAuB,SAAS;AACtC,UAAM,wBAAwB,SAAS;AACvC,UAAM,uBAAuB,SAAS;AACtC,UAAM,WAAW,uBAAuB,IAAI,SAAS,QAAQ;AAC7D,UAAM,SAAS,wBAAwB,IAAI,SAAS,QAAQ;AAC5D,UAAM,UAAU,uBAAuB,IAAI,SAAS,QAAQ;AAC5D,UAAM,YAAY,uBAAuB,wBAAwB,uBAAuB;AACxF,SAAK,WAAW;AAAA,iCACa,aAAa,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAiBzB;AAAA,mBACb;AAAA,gDAC6B;AAAA;AAAA,oCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA,kCAKX;AAAA,sBACZ;AAAA,kDAC4B;AAAA;AAAA,sCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAMX;AAAA,wBACZ;AAAA,oDAC4B;AAAA;AAAA,wCAEZ,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAOf;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,yBAMT,2BAA2B;AAAA,yBAC3B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUvB;AACF;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,OAAO,IAAI;AAC9B,QAAM,IAAI;AACV,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,YAAY,CAAC,GAAG,GAAG,CAAC;AAC1B,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,MAAM,eAAe;AACtH,QAAM,4BAA4B,IAAI,cAAc,UAAU,OAAO,IAAI;AACzE,QAAM,sBAAsB,SAAS,gBAAgB,2BAA2B,CAAC,CAAC,GAAG,EAAE,KAAK;AAC5F,QAAM,yBAAyB,IAAI,yBAAyB,QAAQ;AACpE,QAAM,SAAS,SAAS,gBAAgB,wBAAwB,CAAC,IAAI,mBAAmB,GAAG,EAAE,KAAK;AAClG,WAAS,8BAA8B,mBAAmB;AAC1D,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,QAAQ,OAAO,IAAI;AACtC,QAAM,IAAI;AACV,oBAAkB,CAAC,QAAQ,MAAM,GAAG,aAAa;AACjD,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AAC9G,QAAM,eAAe;AACrB,QAAM,0BAA0B,IAAI,cAAc,UAAU,OAAO,YAAY;AAC/E,QAAM,oBAAoB,SAAS,gBAAgB,yBAAyB,CAAC,CAAC,GAAG,EAAE,KAAK;AACxF,QAAM,yBAAyB,IAAI,yBAAyB,QAAQ;AACpE,QAAM,SAAS,SAAS,gBAAgB,wBAAwB,CAAC,IAAI,iBAAiB,GAAG,EAAE,KAAK;AAChG,WAAS,8BAA8B,iBAAiB;AACxD,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,uBAAuB,GAAG,qBAAqB,UAAU,UAAU;AAC1E,MAAI,UAAU,IAAI,cAAc,UAAU,OAAO,KAAK;AACtD,QAAM,aAAa,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,SAAS;AACnE,YAAU,IAAI,cAAc,UAAU,OAAO,MAAM,MAAM,mBAAmB;AAC5E,QAAM,cAAc,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,SAAS;AACpE,SAAO,CAAC,YAAY,WAAW;AACjC;AAGA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,EAAE,IAAI;AACd,UAAM,EAAE,YAAY,SAAS,KAAK,MAAM,oBAAoB,IAAI;AAChE,UAAM,eAAe;AACrB,iBAAa,OAAO,EAAE,MAAM,WAAW,GAAG,MAAM,uDAAuD,EAAE,MAAM,SAAS;AACxH,UAAM,YAAY,CAAC,GAAG,CAAC;AACvB,iBAAa,OAAO,qBAAqB,+BAA+B,SAAS,SAAS,GAAG,MAAM,wEAAwE,0BAA0B,YAAY;AACjN,UAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,WAAW,IAAI;AACrG,UAAM,CAAC,QAAQ,OAAO,IAAI,uBAAuB,GAAG,qBAAqB,UAAU,YAAY;AAC/F,WAAO,CAAC,QAAQ,OAAO;AAAA,EACzB;AACF;AAGA,SAAS,SAAS,GAAG,aAAa,UAAU,UAAU;AACpD,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,YAAY,QAAQ;AAC1B,QAAM,gBAAgB,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,OAAO,CAAC,WAAW,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC1G,QAAM,UAAU,OAAO,eAAe,WAAW,QAAQ,QAAQ;AACjE,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AACzG,WAAS,8BAA8B,aAAa;AACpD,WAAS,8BAA8B,OAAO;AAC9C,SAAO;AACT;AAGA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,EAAE,IAAI;AACd,UAAM,EAAE,UAAU,KAAK,IAAI;AAC3B,UAAM,eAAe;AACrB,UAAM,QAAQ,EAAE,MAAM;AACtB,UAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,QAAI,OAAO;AACX,UAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,UAAM,wBAAwB,gBAAgB;AAC9C,UAAM,qBAAqB,aAAa,mBAAmB,CAAC,CAAC,CAAC;AAC9D,UAAM,gBAAgB,CAAC;AACvB,QAAI,YAAY;AAChB,QAAI,uBAAuB;AACzB,UAAI,oBAAoB;AACtB,cAAM,WAAW,aAAa,QAAQ,IAAI,UAAU,MAAM;AAC1D,cAAM,SAAS,SAAS;AACxB,cAAM,WAAW,IAAI,MAAM,KAAK;AAChC,iBAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,mBAAS,KAAK,EAAE,MAAM,aAAa;AAAA,QACrC;AACA,cAAM,kBAAkB,iBAAiB,QAAQ,EAAE,OAAO,EAAE,OAAO,cAAc,QAAQ;AACzF,oBAAY,aAAa,eAAe,UAAU,EAAE,KAAK;AACzD,cAAM,gBAAgB,aAAa,QAAQ,IAAI,UAAU,MAAM;AAC/D,sBAAc,SAAS;AAAA,MACzB,OAAO;AACL,oBAAY,eAAe,GAAG,cAAc,YAAY;AAAA,MAC1D;AACA,oBAAc,KAAK,SAAS;AAC5B,aAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAAA,IACjE;AACA,yBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,UAAM,CAAC,cAAc,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACxG,QAAI,WAAW;AACf,QAAI,UAAU;AACZ,iBAAW,qBAAqB,qBAAqB,cAAc,QAAQ;AAAA,IAC7E;AACA,UAAM,MAAM,SAAS,WAAW,aAAa,UAAU,YAAY;AACnE,eAAW,KAAK,eAAe;AAC7B,mBAAa,8BAA8B,CAAC;AAAA,IAC9C;AACA,WAAO;AAAA,EACT;AACF;AAGA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AAC1F,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,EAC1E;AACA,uBAAqB,2BAA2B,OAAO,MAAM,KAAK;AAClE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACpG,QAAM,SAAS,aAAa,cAAc,WAAW;AACrD,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,QAAM,UAAU,OAAO,KAAK,IAAI,OAAO,OAAO,QAAQ;AACtD,MAAI;AACJ,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,UAAU,QAAQ;AAC7E,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F,OAAO;AACL,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EAC1F;AACA,WAAS,8BAA8B,GAAG;AAC1C,WAAS,8BAA8B,OAAO;AAC9C,MAAI,gBAAgB,MAAM;AACxB,aAAS,8BAA8B,SAAS;AAAA,EAClD;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,UAAU,qBAAqB;AAAA;AAAA;AAGnC,IAAI,iBAAiB;AAAA;AAAA;AAAA;AAAA;AAAA,MAKf,2BAA2B;AAAA;AAAA;AAGjC,IAAI,WAAW,kBAAkB;AAAA,EAC/B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AACjB,CAAC;AACD,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,QAAQ,UAAU,MAAM;AAClC,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,KAAK,GAAG,EAAE;AACpE,UAAM,OAAO,OAAO;AACpB,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE,EAAE,KAAK,GAAG;AAClD,UAAM,MAAM,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,EAAE,EAAE,KAAK,GAAG;AAC/D,UAAM,iBAAiB,CAAC,aAAa,aAAa,aAAa,WAAW,EAAE,MAAM,GAAG,IAAI;AACzF,UAAM,SAAS,SAAS,YAAY,IAAI;AACxC,QAAI,SAAS,GAAG;AACd,WAAK,WAAW;AAAA,sBACA;AAAA,oBACF;AAAA;AAAA;AAAA;AAAA;AAAA,wCAKoB;AAAA;AAAA,4CAEI;AAAA;AAAA;AAAA;AAAA;AAKtC;AAAA,IACF;AACA,SAAK,WAAW;AAAA,QACZ,iBAAiB,SAAS;AAAA,QAC1B,eAAe,SAAS;AAAA;AAAA;AAAA,UAGtB;AAAA,8BACoB;AAAA;AAAA,iDAEmB;AAAA;AAAA,qDAEI;AAAA;AAAA;AAAA,UAG3C;AAAA,yBACe;AAAA;AAAA;AAAA,EAGvB;AACF;AAGA,IAAI,yBAAyB,MAAM;AAAA,EACjC,YAAY,QAAQ,UAAU,MAAM;AAClC,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,KAAK,GAAG,EAAE;AACpE,UAAM,OAAO,OAAO;AACpB,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE,EAAE,KAAK,GAAG;AAClD,UAAM,MAAM,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,EAAE,EAAE,KAAK,GAAG;AAC/D,UAAM,UAAU,YAAY,MAAM,IAAI;AACtC,UAAM,SAAS,YAAY,UAAU,IAAI;AACzC,UAAM,SAAS,GAAG,QAAQ,OAAO,QAAQ,KAAK,YAAY,OAAO;AACjE,UAAM,YAAY,SAAS,IAAI,WAAW,QAAQ,OAAO,MAAM,EAAE,EAAE,KAAK;AACxE,UAAM,SAAS,SAAS,YAAY,IAAI;AACxC,QAAI,WAAW;AACf,QAAI,SAAS,GAAG;AACd,YAAM,WAAW;AAAA,UACb;AAAA;AAAA,0CAEgC;AAAA;AAAA,8CAEI;AAAA;AAAA;AAAA;AAIxC,iBAAW;AAAA,UACP;AAAA,UACA;AAAA,sCAC4B,OAAO,KAAK,OAAO;AAAA,UAC/C,QAAQ,OAAO;AAAA,aACZ;AAAA,YACD;AAAA,wCAC4B,OAAO,KAAK,OAAO;AAAA;AAAA;AAAA,IAGvD,OAAO;AACL,YAAM,WAAW;AAAA,UACb;AAAA,UACA,cAAc;AAAA,UACd,eAAe;AAAA,UACf;AAAA;AAAA,6CAEmC;AAAA,kDACK;AAAA;AAAA;AAG5C,iBAAW;AAAA,UACP;AAAA,UACA;AAAA,sCAC4B,OAAO,KAAK,OAAO;AAAA,UAC/C,QAAQ,OAAO;AAAA,aACZ;AAAA,YACD;AAAA,wCAC4B,OAAO,KAAK,OAAO;AAAA;AAAA;AAAA,UAGjD,QAAQ,OAAO;AAAA,aACZ,QAAQ,OAAO,QAAQ,KAAK,YAAY,OAAO;AAAA,YAChD;AAAA,wCAC4B,OAAO,KAAK,OAAO;AAAA,YAC/C,QAAQ,OAAO;AAAA,eACZ;AAAA,cACD;AAAA,0CAC4B,OAAO,KAAK,OAAO;AAAA;AAAA;AAAA;AAAA,IAIzD;AACA,SAAK,WAAW;AAAA,cACN,iBAAiB,SAAS;AAAA,cAC1B,eAAe,SAAS;AAAA;AAAA;AAAA,UAG5B;AAAA;AAAA,UAEA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AAGA,IAAI,sBAAsB,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AAClE,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,UAAU,KAAK,IAAI;AAC3B,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,uBAAuB,EAAE,OAAO,UAAU,IAAI,IAAI,IAAI,iBAAiB,EAAE,OAAO,UAAU,IAAI;AACjK,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AAC7D,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AAAA;AAEV,IAAI,aAAa;AAAA;AAAA;AAAA,MAGX,2BAA2B;AAAA;AAAA;AAGjC,IAAI,OAAO,kBAAkB;AAAA,EAC3B,WAAW;AAAA,EACX,iBAAiB;AACnB,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,qBAAqB,MAAM;AAAA,EAC7B,YAAY,WAAW,aAAa,YAAY;AAC9C,SAAK,gBAAgB,CAAC,OAAO;AAC7B,SAAK,iBAAiB,CAAC,EAAE,MAAM,QAAQ,MAAM,QAAQ,CAAC;AACtD,SAAK,cAAc,CAAC,WAAW,UAAU;AACzC,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,8BAQU,cAAc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0BAUlB,cAAc;AAAA;AAAA;AAAA,EAGtC;AACF;AAGA,IAAI,MAAM;AAAA;AAAA;AAAA;AAAA;AAKV,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAmBjB,IAAI,UAAU,kBAAkB,EAAE,WAAW,KAAK,iBAAiB,YAAY,kBAAkB,KAAK,CAAC;AACvG,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AACV,IAAI,OAAO,kBAAkB;AAAA,EAC3B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EACjB,eAAe;AACjB,CAAC;AACD,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,IAAI,IAAI;AAChB,QAAM,OAAO,aAAa,eAAe,CAAC,GAAG,GAAG,OAAO,KAAK;AAC5D,QAAM,WAAW,KAAK;AAAA,IACpB,QAAQ,EAAE,GAAG,OAAO;AAAA,IACpB,SAAS;AAAA,IACT,OAAO,EAAE,kBAAkB,MAAM,UAAU,MAAM;AAAA,EACnD,CAAC;AACD,QAAM,gBAAgB,qBAAqB,qBAAqB,SAAS,OAAO,IAAI;AACpF,QAAM,oBAAoB,SAAS,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAClH,QAAM,IAAI,KAAK,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,kBAAkB,GAAG,SAAS,SAAS,CAAC;AACjF,QAAM,IAAI,KAAK,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,SAAS,CAAC;AACtD,QAAM,SAAS,KAAK,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,MAAM,UAAU,MAAM,EAAE,CAAC;AACnG,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,cAAc,EAAE,CAAC;AAC7G,QAAM,MAAM,QAAQ,EAAE,QAAQ,EAAE,GAAG,GAAG,GAAG,eAAe,GAAG,SAAS,SAAS,CAAC;AAC9E,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,iBAAiB;AACxD,WAAS,8BAA8B,CAAC;AACxC,WAAS,8BAA8B,CAAC;AACxC,WAAS,8BAA8B,MAAM;AAC7C,WAAS,8BAA8B,cAAc;AACrD,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,YAAY,MAAM,WAAW,IAAI;AACzC,QAAM,QAAQ,aAAa,SAAS,SAAS,EAAE,QAAQ,EAAE,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,OAAO,MAAM,SAAS,EAAE,EAAE,CAAC;AAC/H,QAAM,YAAY,MAAM,MAAM;AAC9B,QAAM,cAAc,MAAM,MAAM;AAChC,QAAM,UAAU,IAAI,mBAAmB,WAAW,aAAa,UAAU;AACzE,QAAM,eAAe,CAAC,CAAC,IAAI,CAAC;AAC5B,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,KAAK,GAAG,SAAS,YAAY;AAC5E,MAAI,CAAC,YAAY;AACf,aAAS,8BAA8B,KAAK;AAAA,EAC9C;AACA,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM,oBAAoB;AAAA;AAAA;AAG9B,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWjB,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,SAAS,mBAAmB,CAAC,CAAC,CAAC,GAAG;AACpC,UAAM,QAAQ,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC3C,UAAM,CAAC,WAAW,QAAQ,IAAI,WAAW,MAAM,QAAQ,EAAE,OAAO,EAAE,KAAK;AACvE,WAAO,SAAS,eAAe,UAAU,EAAE,OAAO,SAAS;AAAA,EAC7D;AACA,MAAI;AACJ,MAAI,IAAI,EAAE,QAAQ,6BAA6B,GAAG;AAChD,cAAU,IAAI,qBAAqB,EAAE,OAAO,UAAU;AAAA,EACxD,OAAO;AACL,cAAU,IAAI,eAAe,EAAE,OAAO,GAAG;AAAA,EAC3C;AACA,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,qBAAqB,MAAM;AAClC,uBAAqB,KAAK,+FAA+F;AACzH,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,eAAe,IAAI;AACxD,QAAM,YAAY,SAAS,SAAS,MAAM,MAAM;AAChD,QAAM,aAAa,SAAS,SAAS,OAAO,MAAM;AAClD,QAAM,EAAE,gBAAgB,IAAI,yBAAyB,WAAW,YAAY,eAAe,cAAc,cAAc;AACvH,SAAO,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AACnG;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,qBAAqB,MAAM;AAClC,uBAAqB,KAAK,+FAA+F;AACzH,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,gBAAgB,mBAAmB,IAAI;AAC5E,QAAM,YAAY,SAAS,SAAS,MAAM,MAAM;AAChD,QAAM,aAAa,SAAS,SAAS,OAAO,MAAM;AAClD,QAAM,EAAE,iBAAiB,aAAa,IAAI,yBAAyB,WAAW,YAAY,eAAe,cAAc,gBAAgB,kBAAkB;AACzJ,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AAAA,IAC1F,SAAS,eAAe,CAAC,GAAG,SAAS,IAAI,WAAW,CAAC,YAAY,CAAC,CAAC;AAAA,EACrE;AACF;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,2BAA2B,qBAAqB;AACpD,SAAS,qBAAqB,MAAM;AAClC,uBAAqB,KAAK,+FAA+F;AACzH,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,eAAe,cAAc,gBAAgB,aAAa,IAAI;AACtE,QAAM,YAAY,SAAS,SAAS,MAAM,MAAM;AAChD,QAAM,aAAa,SAAS,SAAS,OAAO,MAAM;AAClD,QAAM,mBAAmB;AACzB,QAAM,kBAAkB;AACxB,QAAM,oBAAoB;AAC1B,QAAM,kBAAkB;AACxB,QAAM,EAAE,iBAAiB,eAAe,IAAI,yBAAyB,WAAW,YAAY,kBAAkB,iBAAiB,mBAAmB,eAAe;AACjK,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,SAAS,IAAI,WAAW,eAAe,CAAC;AAAA,IAC1F,SAAS,eAAe,CAAC,eAAe,MAAM,GAAG,WAAW,IAAI,aAAa,cAAc,CAAC;AAAA,EAC9F;AACF;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,YAAY,OAAO,SAAS,UAAU;AAChD,SAAK,gBAAgB,CAAC,SAAS;AAC/B,SAAK,cAAc,CAAC,YAAY,KAAK;AACrC,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA,8BAIU,oBAAoB;AAAA;AAAA;AAAA;AAAA,EAIhD;AACF;AAGA,IAAI,UAAU,CAAC,SAAS;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,IAAI;AACpB,QAAM,EAAE,OAAO,OAAO,SAAS,SAAS,IAAI;AAC5C,QAAM,cAAc,aAAa,cAAc,QAAQ,KAAK;AAC5D,QAAM,UAAU,IAAI,cAAc,aAAa,OAAO,SAAS,QAAQ;AACvE,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,WAAW,EAAE,EAAE,CAAC;AACxG,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,QAAQ,GAAG,KAAK;AAClE,WAAS,8BAA8B,QAAQ;AAC/C,QAAM,WAAW,CAAC,GAAG,QAAQ,OAAO,KAAK;AACzC,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC7F,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,EAAE,UAAU,aAAa;AAC3B,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC3E,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,WAAO;AAAA,EACT,OAAO;AACL,WAAO,MAAM;AAAA,MACX,OAAO;AAAA,QACL,OAAO,EAAE;AAAA,QACT,OAAO,EAAE;AAAA,QACT,OAAO,EAAE,UAAU,WAAW,KAAK;AAAA,MACrC;AAAA,MACA,SAAS;AAAA,IACX,CAAC;AAAA,EACH;AACF;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,EAAE,IAAI;AACd,MAAI,EAAE,UAAU,UAAU;AACxB,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE,WAAW,EAAE,UAAU,aAAa;AAClC,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,UAAU,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,WAAW,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAClE,UAAM,IAAI,WAAW,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,SAAS,CAAC;AACnE,UAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE,GAAG,SAAS,SAAS,CAAC;AAC3E,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,aAAS,8BAA8B,QAAQ;AAC/C,aAAS,8BAA8B,CAAC;AACxC,WAAO;AAAA,EACT,OAAO;AACL,WAAO,MAAM,EAAE,OAAO,EAAE,OAAO,EAAE,OAAO,OAAO,EAAE,OAAO,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACzF;AACF;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,YAAY,EAAE,QAAQ,EAAE,OAAO,OAAO,GAAG,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAAA,EAC9F;AACA,QAAM,QAAQ,OAAO,GAAG;AACxB,QAAM,QAAQ,OAAO,GAAG;AACxB,SAAO,QAAQ,CAAC,MAAM;AACpB,iBAAa,kBAAkB,OAAO,EAAE,OAAO,uDAAuD;AACtG,iBAAa,OAAO,UAAU,EAAE,OAAO,MAAM,uDAAuD;AAAA,EACtG,CAAC;AACD,QAAM,0BAA0B,CAAC;AACjC,QAAM,kBAAkB,OAAO,IAAI,CAAC,MAAM;AACxC,UAAM,YAAY,YAAY,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAC/F,4BAAwB,KAAK,SAAS;AACtC,WAAO;AAAA,EACT,CAAC;AACD,QAAM,SAAS,QAAQ,EAAE,QAAQ,iBAAiB,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AACtF,0BAAwB,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAChF,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa,MAAM;AAAA,EACrB,YAAY,QAAQ,UAAU,eAAe;AAC3C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,MAAM,QAAQ,CAAC;AACvD,SAAK,cAAc,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,KAAK,GAAG,EAAE;AACpE,UAAM,OAAO,OAAO;AACpB,UAAM,OAAO,kBAAkB,IAAI;AACnC,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE,EAAE,KAAK,GAAG;AAClD,UAAM,MAAM,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,EAAE,EAAE,KAAK,GAAG;AAC/D,UAAM,iBAAiB,CAAC,aAAa,aAAa,aAAa,WAAW,EAAE,MAAM,GAAG,IAAI;AACzF,QAAI,SAAS,GAAG;AACd,WAAK,WAAW;AAAA,sBACA;AAAA,oBACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWd;AAAA,IACF;AACA,SAAK,WAAW;AAAA,QACZ,gBAAgB,QAAQ;AAAA,QACxB,cAAc,QAAQ;AAAA;AAAA;AAAA,UAGpB;AAAA;AAAA;AAAA;AAAA,YAIE;AAAA,2BACe;AAAA;AAAA;AAAA;AAAA,EAIzB;AACF;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,QAAQ,UAAU,eAAe;AAC3C,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,iBAAiB,CAAC,EAAE,MAAM,SAAS,MAAM,QAAQ,CAAC;AACvD,SAAK,cAAc,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,KAAK,GAAG,EAAE;AACpE,UAAM,OAAO,OAAO;AACpB,UAAM,QAAQ,kBAAkB,IAAI;AACpC,UAAM,QAAQ,SAAS,IAAI,CAAC,OAAO,GAAG,EAAE,EAAE,KAAK,GAAG;AAClD,UAAM,MAAM,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,OAAO,EAAE,EAAE,KAAK,GAAG;AAC/D,UAAM,UAAU,YAAY,MAAM,IAAI;AACtC,UAAM,SAAS,YAAY,UAAU,IAAI;AACzC,UAAM,SAAS,GAAG,QAAQ,OAAO,QAAQ,KAAK,YAAY,OAAO;AACjE,UAAM,YAAY,SAAS,IAAI,WAAW,QAAQ,OAAO,MAAM,EAAE,EAAE,KAAK;AACxE,UAAM,iBAAiB;AAAA,MACrB,GAAG;AAAA,MACH,GAAG,QAAQ,OAAO;AAAA,YACZ;AAAA;AAAA,MAEN,SAAS,IAAI,KAAK;AAAA;AAAA,SAEf,QAAQ,OAAO;AAAA,YACZ,QAAQ,OAAO,QAAQ,KAAK,YAAY,OAAO;AAAA,MACrD,SAAS,IAAI,KAAK,KAAK,QAAQ,OAAO;AAAA,cAC9B;AAAA,IACV;AACA,UAAM,cAAc,SAAS,IAAI,4BAA4B;AAC7D,QAAI,WAAW;AACf,aAAS,IAAI,GAAG,IAAI,SAAS,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;AAClD,kBAAY;AAAA,UACR,eAAe;AAAA,cACX;AAAA,mBACK;AAAA;AAAA,YAEP;AAAA,mBACO,wBAAwB,OAAO,KAAK,OAAO;AAAA;AAAA;AAAA,IAG1D;AACA,gBAAY,SAAS,IAAI,OAAO;AAChC,SAAK,WAAW;AAAA,cACN,iBAAiB,SAAS;AAAA,cAC1B,eAAe,SAAS;AAAA;AAAA;AAAA,UAG5B;AAAA;AAAA,UAEA;AAAA;AAAA;AAAA;AAAA,EAIR;AACF;AAGA,IAAI,SAAS,CAAC,SAAS;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,UAAU,cAAc,IAAI;AACpC,MAAI,aAAa,cAAc,EAAE,KAAK,MAAM,GAAG;AAC7C,UAAM,cAAc,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,KAAK,GAAG,EAAE;AACtE,WAAO,MAAM;AAAA,MACX,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,aAAa,OAAO,eAAe,OAAO,EAAE,MAAM;AAAA,IACpE,CAAC;AAAA,EACH;AACA,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,iBAAiB,EAAE,OAAO,UAAU,aAAa,IAAI,IAAI,WAAW,EAAE,OAAO,UAAU,aAAa;AACvK,QAAM,eAAe,CAAC,CAAC,aAAa,CAAC;AACrC,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,OAAO,YAAY;AACrE;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUV,IAAI,aAAa;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,MAgBX,2BAA2B;AAAA;AAAA;AAGjC,IAAI,OAAO,kBAAkB,EAAE,WAAW,KAAK,iBAAiB,WAAW,CAAC;AAC5E,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,YAAY,CAAC;AACnB,QAAM,WAAW,aAAa,eAAe,MAAM,EAAE,KAAK;AAC1D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,aAAa,EAAE,CAAC;AAC1F,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAC/D,cAAU,KAAK,SAAS;AAAA,EAC1B;AACA,uBAAqB,2BAA2B,QAAQ,MAAM,KAAK;AACnE,MAAI;AACJ,MAAI,SAAS,mBAAmB,CAAC,SAAS,CAAC,GAAG;AAC5C,UAAM,QAAQ,SAAS,QAAQ,IAAI,UAAU,MAAM,EAAE;AACrD,UAAM,EAAE,SAAS,UAAU,SAAS,IAAI,YAAY,UAAU,OAAO,UAAU,OAAO,OAAO,IAAI;AACjG,UAAM,SAAS,eAAe,UAAU,UAAU,OAAO;AAAA,EAC3D,OAAO;AACL,UAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,UAAU,OAAO,IAAI;AACpG,UAAM,SAAS,aAAa,cAAc,WAAW;AACrD,UAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,UAAM,cAAc,WAAW,EAAE,KAAK;AACtC,UAAM,UAAU,OAAO,KAAK,aAAa,QAAQ,QAAQ;AACzD,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,cAAU,KAAK,GAAG;AAClB,cAAU,KAAK,OAAO;AAAA,EACxB;AACA,MAAI,UAAU;AACZ,cAAU,KAAK,GAAG;AAClB,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,QAAQ;AAC9E,UAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAAA,EACtF;AACA,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,oBAAoB,mBAAmB,QAAQ,IAAI;AAC3D,QAAM,EAAE,iBAAiB,IAAI;AAC7B,QAAM,sBAAsB,mBAAmB,IAAI,CAAC,MAAM,SAAS,SAAS,EAAE,MAAM,CAAC;AACrF,QAAM,4BAA4B,mBAAmB,IAAI,CAAC,MAAM,EAAE,KAAK;AACvE,QAAM,qBAAqB,SAAS,SAAS,kBAAkB,MAAM;AACrE,QAAM,WAAW,SAAS,SAAS,QAAQ,MAAM;AACjD,QAAM,CAAC,oBAAoB,mBAAmB,sBAAsB,IAAI,oBAAoB,qBAAqB,2BAA2B,oBAAoB,kBAAkB,OAAO,kBAAkB,OAAO,UAAU,QAAQ,OAAO,gBAAgB;AAC3P,QAAM,4BAA4B,mBAAmB,IAAI,CAAC,WAAW,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,SAAS,MAAM,CAAC;AAC9H,QAAM,0BAA0B,SAAS,eAAe,wBAAwB,kBAAkB,OAAO,iBAAiB;AAC1H,SAAO,0BAA0B,OAAO,CAAC,uBAAuB,CAAC;AACnE;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,QAAQ,QAAQ,OAAO,IAAI;AACnC,QAAM,UAAU,SAAS,SAAS,OAAO,MAAM;AAC/C,QAAM,UAAU,SAAS,SAAS,OAAO,MAAM;AAC/C,QAAM,UAAU,SAAS,SAAS,OAAO,MAAM;AAC/C,QAAM,CAAC,oBAAoB,iBAAiB,IAAI,mBAAmB,SAAS,OAAO,OAAO,OAAO,OAAO,SAAS,OAAO,OAAO,SAAS,OAAO,KAAK;AACpJ,QAAM,iBAAiB,SAAS,eAAe,CAAC,mBAAmB,MAAM,GAAG,SAAS,kBAAkB;AACvG,QAAM,gBAAgB,SAAS,eAAe,CAAC,kBAAkB,MAAM,GAAG,OAAO,OAAO,iBAAiB;AACzG,SAAO,CAAC,gBAAgB,aAAa;AACvC;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,sBAAsB,MAAM;AACnC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,cAAc,oBAAoB,IAAI;AAC7D,QAAM,EAAE,kBAAkB,IAAI;AAC9B,QAAM,SAAS,SAAS,SAAS,MAAM,MAAM;AAC7C,QAAM,UAAU,SAAS,SAAS,OAAO,MAAM;AAC/C,QAAM,gBAAgB,SAAS,SAAS,aAAa,MAAM;AAC3D,QAAM,sBAAsB,oBAAoB,IAAI,CAAC,MAAM,SAAS,SAAS,EAAE,MAAM,CAAC;AACtF,QAAM,2BAA2B,oBAAoB,IAAI,CAAC,MAAM,EAAE,KAAK;AACvE,QAAM,CAAC,aAAa,MAAM,IAAI,4BAA4B,QAAQ,MAAM,OAAO,SAAS,OAAO,OAAO,OAAO,OAAO,eAAe,aAAa,OAAO,qBAAqB,0BAA0B,iBAAiB;AACvN,SAAO,SAAS,eAAe,aAAa,OAAO,OAAO,MAAM;AAClE;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,CAAC,SAAS;AACrB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,MAAM,OAAO,MAAM,IAAI;AAC5C,QAAM,SAAS,aAAa,OAAO,MAAM,OAAO,KAAK;AACrD,SAAO,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,OAAO,MAAM;AAC/D;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,aAAa;AACjB,IAAI,cAAc,iBAAiB,EAAE,WAAW,WAAW,CAAC;AAC5D,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ,oBAAoB;AAAA;AAAA;AAGhC,IAAI,cAAc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWlB,IAAI,QAAQ,iBAAiB,EAAE,WAAW,OAAO,iBAAiB,YAAY,CAAC;AAC/E,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS,oBAAoB;AAAA;AAAA;AAGjC,IAAI,eAAe;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWnB,IAAI,SAAS,iBAAiB,EAAE,WAAW,QAAQ,iBAAiB,aAAa,CAAC;AAClF,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,wBAAwB,MAAM;AAAA,EAChC,YAAY,YAAY,WAAW,UAAU,cAAc,kBAAkB;AAC3E,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC;AACpB,UAAM,CAAC,OAAO,WAAW,UAAU,KAAK,IAAI;AAC5C,SAAK,cAAc,CAAC,OAAO,WAAW,UAAU,KAAK;AACrD,UAAM,kBAAkB;AAAA,MACtB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,mBAAmB;AAAA,MACvB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,QAAI;AACJ,QAAI,kBAAkB;AACpB,0BAAoB;AAAA,IACtB,OAAO;AACL,0BAAoB;AAAA,IACtB;AACA,SAAK,WAAW;AAAA;AAAA,YAER,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,uCACX,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCASpB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAqBjC;AACF;AAGA,IAAI,8BAA8B,MAAM;AAAA,EACtC,YAAY,YAAY,WAAW,UAAU,cAAc,kBAAkB;AAC3E,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,CAAC;AACpB,UAAM,CAAC,OAAO,WAAW,UAAU,KAAK,IAAI;AAC5C,SAAK,cAAc,CAAC,OAAO,WAAW,UAAU,KAAK;AACrD,UAAM,kBAAkB;AAAA,MACtB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,mBAAmB;AAAA,MACvB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,QAAI;AACJ,QAAI,kBAAkB;AACpB,0BAAoB;AAAA,IACtB,OAAO;AACL,0BAAoB;AAAA,IACtB;AACA,SAAK,WAAW;AAAA;AAAA,YAER,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,uCACX,gBAAgB;AAAA,uCAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAcJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,gCAQH,QAAQ;AAAA,uCACD,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAiDhD;AACF;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,4BAA4B,OAAO,OAAO,WAAW,UAAU,cAAc,gBAAgB,IAAI,IAAI,sBAAsB,OAAO,OAAO,WAAW,UAAU,cAAc,gBAAgB;AAC/P,SAAO,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,SAAS;AAC9D;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gCAAgC,MAAM;AAAA,EACxC,YAAY,SAAS,YAAY,cAAc;AAC7C,SAAK,gBAAgB,CAAC,IAAI;AAC1B,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc;AACnB,UAAM,CAAC,EAAE,SAAS,MAAM,IAAI;AAC5B,UAAM,CAAC,EAAE,SAAS,MAAM,IAAI;AAC5B,UAAM,iBAAiB;AAAA,MACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,MAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,IAC5C;AACA,UAAM,iBAAiB;AAAA,MACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,MAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,IAC5C;AACA,UAAM,cAAc,eAAe,KAAK,eAAe;AACvD,UAAM,aAAa,eAAe,KAAK,eAAe;AACtD,UAAM,iBAAiB,IAAI;AAC3B,UAAM,gBAAgB,IAAI;AAC1B,UAAM,YAAY,KAAK,KAAK,cAAc,IAAI,IAAI;AAClD,UAAM,WAAW,KAAK,KAAK,aAAa,IAAI,IAAI;AAChD,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAUsB;AAAA,yCACD;AAAA;AAAA,6CAEI;AAAA,4CACD;AAAA;AAAA,oCAER;AAAA,mCACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAcD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAQE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sDAMkB,UAAU;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,qDAMX,SAAS;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA+B5D;AACF;AAGA,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,GAAG,IAAI;AACvB,QAAM,EAAE,aAAa,IAAI;AACzB,QAAM,UAAU,IAAI,8BAA8B,GAAG,OAAO,OAAO,OAAO,YAAY;AACtF,SAAO,SAAS,gBAAgB,SAAS,CAAC,EAAE,GAAG,GAAG,KAAK;AACzD;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,+BAA+B,MAAM;AAAA,EACvC,YAAY,YAAY,WAAW,UAAU,cAAc,kBAAkB;AAC3E,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc,CAAC;AACpB,UAAM,CAAC,OAAO,WAAW,UAAU,KAAK,IAAI;AAC5C,SAAK,cAAc,CAAC,OAAO,WAAW,UAAU,KAAK;AACrD,UAAM,kBAAkB;AAAA,MACtB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,mBAAmB;AAAA,MACvB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,YAAY,eAAe,QAAQ;AACzC,QAAI;AACJ,QAAI,kBAAkB;AACpB,0BAAoB;AAAA,IACtB,OAAO;AACL,0BAAoB;AAAA,IACtB;AACA,SAAK,WAAW;AAAA;AAAA,YAER,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,uCACX,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCASpB;AAAA;AAAA;AAAA;AAAA,8DAI2B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAM5D;AACF;AAGA,IAAI,qCAAqC,MAAM;AAAA,EAC7C,YAAY,YAAY,WAAW,UAAU,cAAc,kBAAkB;AAC3E,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,SAAK,cAAc,CAAC;AACpB,UAAM,CAAC,OAAO,WAAW,UAAU,KAAK,IAAI;AAC5C,SAAK,cAAc,CAAC,OAAO,WAAW,UAAU,KAAK;AACrD,UAAM,kBAAkB;AAAA,MACtB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,mBAAmB;AAAA,MACvB,gBAAgB,YAAY,IAAI,YAAY,IAAI;AAAA,MAChD,gBAAgB,WAAW,IAAI,WAAW,IAAI;AAAA,IAChD;AACA,UAAM,YAAY,eAAe,QAAQ;AACzC,QAAI;AACJ,QAAI,kBAAkB;AACpB,0BAAoB;AAAA,IACtB,OAAO;AACL,0BAAoB;AAAA,IACtB;AACA,SAAK,WAAW;AAAA;AAAA,YAER,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,YACtC,gBAAgB,KAAK,iBAAiB;AAAA,uCACX,gBAAgB;AAAA,uCAChB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mCAcJ;AAAA;AAAA;AAAA;AAAA,8DAI2B;AAAA;AAAA;AAAA,gCAG9B,QAAQ;AAAA,uCACD,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAchD;AACF;AAGA,SAAS,uBAAuB,MAAM;AACpC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,mCAAmC,OAAO,OAAO,WAAW,UAAU,cAAc,gBAAgB,IAAI,IAAI,6BAA6B,OAAO,OAAO,WAAW,UAAU,cAAc,gBAAgB;AAC7Q,SAAO,SAAS,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,KAAK;AACjE;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,sCAAsC,MAAM;AAAA,EAC9C,YAAY,SAAS,YAAY,cAAc;AAC7C,SAAK,gBAAgB,CAAC,IAAI;AAC1B,SAAK,cAAc,CAAC;AACpB,SAAK,cAAc;AACnB,UAAM,CAAC,EAAE,SAAS,MAAM,IAAI;AAC5B,UAAM,CAAC,EAAE,SAAS,MAAM,IAAI;AAC5B,UAAM,iBAAiB;AAAA,MACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,MAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,IAC5C;AACA,UAAM,iBAAiB;AAAA,MACrB,gBAAgB,UAAU,IAAI,UAAU,IAAI;AAAA,MAC5C,gBAAgB,SAAS,IAAI,SAAS,IAAI;AAAA,IAC5C;AACA,UAAM,cAAc,eAAe,KAAK,eAAe;AACvD,UAAM,aAAa,eAAe,KAAK,eAAe;AACtD,UAAM,iBAAiB,IAAI;AAC3B,UAAM,gBAAgB,IAAI;AAC1B,UAAM,YAAY,KAAK,KAAK,cAAc,IAAI,IAAI;AAClD,UAAM,WAAW,KAAK,KAAK,aAAa,IAAI,IAAI;AAChD,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,0CAUsB;AAAA,yCACD;AAAA;AAAA,6CAEI;AAAA,4CACD;AAAA;AAAA,oCAER;AAAA,mCACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,kCAcD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,oCAQE;AAAA;AAAA;AAAA;AAAA;AAAA,sBAKd,eAAe;AAAA,sCACC,eAAe;AAAA;AAAA;AAAA,wBAG7B,eAAe;AAAA,wCACC,eAAe;AAAA;AAAA;AAAA,4BAG3B;AAAA,kBACV;AAAA;AAAA;AAAA;AAAA,4BAIU;AAAA,kBACV;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAahB;AACF;AAGA,SAAS,2BAA2B,MAAM;AACxC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,GAAG,IAAI;AACvB,QAAM,EAAE,aAAa,IAAI;AACzB,QAAM,UAAU,IAAI,oCAAoC,GAAG,OAAO,OAAO,OAAO,YAAY;AAC5F,SAAO,SAAS,gBAAgB,SAAS,CAAC,EAAE,GAAG,GAAG,KAAK;AACzD;AACA,IAAI,mCAAmC;AAAA,EACrC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,QAAQ,MAAM;AACxB,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,OAAO,OAAO;AACpB,QAAI,OAAO,GAAG;AACZ,YAAM,IAAI,MAAM,kCAAkC,kCAAkC;AAAA,IACtF;AACA,SAAK,cAAc;AACnB,QAAI,SAAS,GAAG;AACd,WAAK,WAAW;AAAA;AAAA;AAAA,2BAGK,OAAO;AAAA;AAAA;AAG5B;AAAA,IACF;AACA,UAAM,aAAa,CAAC,MAAM;AACxB,UAAI,KAAK,QAAQ,CAAC,MAAM,MAAM,OAAO,OAAO,GAAG;AAC7C,eAAO,GAAG,OAAO,eAAe;AAAA,MAClC;AACA,aAAO,UAAU;AAAA,IACnB;AACA,UAAM,WAAW,OAAO,IAAI,CAAC,GAAG,MAAM,WAAW,CAAC,CAAC,EAAE,KAAK,GAAG;AAC7D,UAAM,OAAO,kBAAkB,IAAI;AACnC,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,yBACe;AAAA;AAAA;AAAA,EAGvB;AACF;AAGA,IAAI,uBAAuB,MAAM;AAAA,EAC/B,YAAY,QAAQ,MAAM;AACxB,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,eAAe;AACpB,SAAK,eAAe;AACpB,UAAM,OAAO,OAAO;AACpB,QAAI,OAAO,GAAG;AACZ,YAAM,IAAI,MAAM,kCAAkC,kCAAkC;AAAA,IACtF;AACA,SAAK,cAAc;AACnB,UAAM,WAAW,YAAY,MAAM,IAAI;AACvC,UAAM,aAAa,GAAG,SAAS,OAAO,YAAY,KAAK,YAAY,OAAO;AAC1E,UAAM,UAAU,GAAG,SAAS,OAAO,YAAY,KAAK,YAAY,OAAO;AACvE,UAAM,OAAO,kBAAkB,IAAI;AACnC,QAAI,SAAS,GAAG;AACd,WAAK,WAAW;AAAA;AAAA;AAAA;AAAA,uCAIiB,OAAO;AAAA,cAChC,OAAO;AAAA,eACN;AAAA,2CAC4B,OAAO;AAAA,kBAChC,OAAO;AAAA;AAAA;AAAA;AAAA;AAAA,IAKrB,OAAO;AACL,WAAK,WAAW;AAAA;AAAA,YAEV;AAAA;AAAA,uBAEW,KAAK,SAAS,MAAM,CAAC;AAAA,eAC7B;AAAA,yBACU,KAAK,SAAS,MAAM,CAAC;AAAA;AAAA,eAE/B;AAAA,yBACU,KAAK,SAAS,MAAM,CAAC;AAAA,iBAC7B;AAAA,2BACU,KAAK,SAAS,MAAM,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAM5C;AACA,aAAS,KAAK,WAAW;AACvB,aAAO,WAAW,SAAS;AAAA,IAC7B;AACA,aAAS,KAAK,WAAW;AACvB,gBAAU,OAAO,KAAK,MAAM,UAAU,OAAO,KAAK;AAClD,aAAO,WAAW,SAAS;AAAA,IAC7B;AACA,aAAS,KAAK,WAAW;AACvB,gBAAU,OAAO,KAAK,MAAM,UAAU,OAAO,KAAK;AAClD,aAAO,WAAW,SAAS;AAAA,IAC7B;AACA,aAAS,KAAK,WAAW;AACvB,gBAAU,OAAO,KAAK,MAAM,UAAU,OAAO,KAAK;AAClD,gBAAU,OAAO,KAAK,MAAM,UAAU,OAAO,KAAK;AAClD,aAAO,WAAW,SAAS;AAAA,IAC7B;AACA,aAAS,WAAW,WAAW;AAC7B,YAAM,gBAAgB,OAAO,IAAI,CAAC,GAAG,MAAM,WAAW,GAAG,SAAS,CAAC;AACnE,YAAM,WAAW,cAAc,KAAK,GAAG;AACvC,YAAM,YAAY,cAAc,MAAM,EAAE,EAAE,KAAK,GAAG;AAClD,aAAO,mBAAmB,mBAAmB;AAAA,IAC/C;AACA,aAAS,WAAW,GAAG,WAAW;AAChC,UAAI,KAAK,QAAQ,CAAC,MAAM,MAAM,OAAO,OAAO,GAAG;AAC7C,eAAO,GAAG,OAAO,QAAQ,UAAU;AAAA,MACrC,OAAO;AACL,eAAO,GAAG,UAAU;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AACF;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,aAAa,eAAe,MAAM,EAAE,KAAK;AACvD,MAAI,UAAU,GAAG;AACf,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD;AACA,QAAM,UAAU,IAAI,EAAE,QAAQ,6BAA6B,IAAI,IAAI,qBAAqB,EAAE,OAAO,KAAK,IAAI,IAAI,eAAe,EAAE,OAAO,KAAK;AAC3I,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,YAAY,WAAW;AACjC,SAAK,gBAAgB,CAAC,OAAO;AAC7B,SAAK,cAAc,CAAC;AACpB,SAAK,iBAAiB,CAAC,EAAE,MAAM,UAAU,MAAM,OAAO,CAAC;AACvD,UAAM,cAAc,WAAW;AAC/B,UAAM,aAAa,WAAW;AAC9B,SAAK,cAAc;AACnB,QAAI,cAAc;AAClB,QAAI,OAAO,cAAc,UAAU;AACjC,oBAAc,uBAAuB,UAAU,QAAQ,CAAC;AAAA,IAC1D,OAAO;AACL,oBAAc;AAAA,2BACO,UAAU,KAAK,GAAG;AAAA;AAAA,IAEzC;AACA,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAWR;AAAA,uCAC2B,yCAAyC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAM9E;AACF;AAGA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY,CAAC,EAAE,QAAQ,OAAO,SAAS,SAAS,MAAM;AACpD,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,EAAE,SAAS,WAAW,OAAO,IAAI;AACvC,UAAM,eAAe;AACrB,UAAM,UAAU,IAAI,cAAc,OAAO,OAAO,SAAS;AACzD,UAAM,CAAC,SAAS,OAAO,IAAI,qBAAqB,eAAe,QAAQ,OAAO,MAAM,IAAI,OAAO,MAAM,EAAE;AACvG,UAAM,eAAe,CAAC,CAAC,SAAS,SAAS,KAAK,IAAI,OAAO,GAAG,KAAK,IAAI,OAAO,CAAC,CAAC;AAC9E,UAAM,SAAS,aAAa,gBAAgB,SAAS,CAAC,MAAM,GAAG,OAAO,OAAO,YAAY;AACzF,WAAO;AAAA,EACT;AACF;AAGA,IAAI,QAAQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAgBZ,IAAI,SAAS,iBAAiB,EAAE,WAAW,MAAM,CAAC;AAClD,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,QAAQ;AACZ,IAAI,SAAS,iBAAiB,EAAE,WAAW,OAAO,eAAe,aAAa,CAAC;AAC/E,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB,MAAM;AAAA,EACzB,YAAY,YAAY,UAAU,aAAa,aAAa,SAAS,OAAO,mBAAmB,MAAM;AACnG,SAAK,gBAAgB,CAAC,WAAW,WAAW,cAAc;AAC1D,SAAK,cAAc;AACnB,UAAM,cAAc,kBAAkB,QAAQ,MAAM;AACpD,UAAM,QAAQ,kBAAkB,MAAM,MAAM;AAC5C,QAAI,gBAAgB;AACpB,QAAI,gBAAgB,GAAG;AACrB,sBAAgB;AAAA,IAClB,WAAW,gBAAgB,GAAG;AAC5B,sBAAgB;AAAA,IAClB;AACA,UAAM,iBAAiB,cAAc;AACrC,QAAI,gBAAgB;AACpB,QAAI,gBAAgB,GAAG;AACrB,sBAAgB;AAAA,IAClB,WAAW,gBAAgB,GAAG;AAC5B,sBAAgB;AAAA,IAClB;AACA,UAAM,iBAAiB,cAAc;AACrC,UAAM,eAAe,WAAW,IAAI,eAAe;AACnD,SAAK,WAAW;AAAA,UACV,yBAAyB,eAAe;AAAA;AAAA;AAAA,YAGtC;AAAA;AAAA;AAAA,gCAGoB;AAAA;AAAA,kCAEE;AAAA,kCACA;AAAA,0CACQ;AAAA;AAAA;AAAA,uBAGnB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOrB;AACF;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,SAAS,QAAQ,IAAI;AAC7B,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW,IAAI,qBAAqB,gBAAgB,SAAS,SAAS,KAAK;AAC9H,QAAM,eAAe,CAAC,aAAa,WAAW,SAAS;AACvD,MAAI,eAAe,GAAG;AACpB,WAAO,SAAS,eAAe,OAAO,QAAQ,KAAK;AAAA,EACrD;AACA,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,YAAY,SAAS,EAAE,EAAE,CAAC;AACxH,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,YAAY,SAAS,EAAE,EAAE,CAAC;AAClH,QAAM,eAAe,SAAS,eAAe,CAAC,GAAG,WAAW,IAAI,aAAa,CAAC,CAAC,CAAC,CAAC;AACjF,QAAM,UAAU,IAAI,eAAe,YAAY,WAAW,eAAe,MAAM,QAAQ,SAAS,MAAM,QAAQ,SAAS,YAAY;AACnI,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,UAAU,gBAAgB,YAAY,GAAG,SAAS,KAAK;AACtG,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AACrF,WAAS,8BAA8B,cAAc;AACrD,WAAS,8BAA8B,QAAQ;AAC/C,WAAS,8BAA8B,GAAG;AAC1C,WAAS,8BAA8B,YAAY;AACnD,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,WAAW,WAAW,WAAW,MAAM;AACjD,SAAK,gBAAgB,CAAC,kBAAkB,QAAQ;AAChD,SAAK,iBAAiB,CAAC,EAAE,MAAM,aAAa,MAAM,MAAM,CAAC;AACzD,SAAK,cAAc,CAAC,WAAW,SAAS;AACxC,UAAM,iBAAiB;AACvB,UAAM,iBAAiB,uBAAuB,KAAK,KAAK,KAAK,KAAK,YAAY,CAAC,CAAC;AAChF,UAAM,WAAW,IAAI,EAAE,UAAU,eAAe,MAAM,IAAI,iBAAiB;AAC3E,UAAM,kBAAkB,SAAS,SAAS,MAAM;AAChD,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA,WAKT;AAAA;AAAA,+CAEoC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAmB7C;AACF;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,gBAAgB,OAAO,IAAI;AACnC,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,UAAU,IAAI,oBAAoB,eAAe,MAAM,IAAI,eAAe,MAAM,IAAI,OAAO,MAAM,IAAI,IAAI;AAC/G,QAAM,eAAe,CAAC,CAAC,eAAe,MAAM,EAAE,CAAC;AAC/C,SAAO,SAAS,gBAAgB,SAAS,CAAC,gBAAgB,MAAM,GAAG,SAAS,YAAY;AAC1F;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,gBAAgB,MAAM;AAAA,EACxB,YAAY,OAAO,OAAO,MAAM;AAC9B,SAAK,gBAAgB,CAAC,KAAK,KAAK,GAAG;AACnC,SAAK,cAAc;AACnB,QAAI;AACJ,QAAI;AACJ,QAAI,OAAO,GAAG;AACZ,YAAM,MAAM,kBAAkB,2BAA2B;AAAA,IAC3D;AACA,QAAI,SAAS,GAAG;AACd,iBAAW;AACX,gBAAU;AAAA,IACZ,OAAO;AACL,YAAM,gBAAgB,CAAC,WAAW,WAAW,WAAW,SAAS;AACjE,YAAM,aAAa,CAAC;AACpB,YAAM,cAAc,CAAC;AACrB,eAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,oBAAY,KAAK,GAAG,cAAc,IAAI;AACtC,YAAI,IAAI,OAAO;AACb,qBAAW,KAAK,GAAG,cAAc,IAAI;AAAA,QACvC;AAAA,MACF;AACA,gBAAU,WAAW,KAAK;AAC1B,iBAAW,YAAY,KAAK;AAAA,IAC9B;AACA,UAAM,QAAQ,kBAAkB,IAAI;AACpC,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,4BACkB;AAAA;AAAA,2BAED;AAAA;AAAA,2BAEA;AAAA;AAAA;AAAA;AAAA,EAIzB;AACF;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,WAAW,GAAG,EAAE,IAAI;AAC5B,QAAM,UAAU,IAAI,cAAc,UAAU,MAAM,QAAQ,EAAE,OAAO,EAAE,MAAM,MAAM;AACjF,SAAO,SAAS,gBAAgB,SAAS,CAAC,WAAW,GAAG,CAAC,GAAG,WAAW,EAAE,OAAO,EAAE,KAAK,CAAC;AAC1F;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AAAA;AAAA;AAAA,uBAGY,qBAAqB;AAAA,kBAC1B,qBAAqB;AAAA;AAAA;AAGvC,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,WAAW,0BAA0B;AAAA;AAAA;AAGzC,IAAI,iBAAiB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAWrB,IAAI,WAAW,iBAAiB;AAAA,EAC9B,WAAW;AAAA,EACX,iBAAiB;AAAA,EACjB,eAAe;AACjB,CAAC;AACD,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AAAA;AAAA;AAAA;AAIX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM,0BAA0B;AAAA;AAAA;AAGpC,IAAI,OAAO,iBAAiB,EAAE,WAAW,IAAI,CAAC;AAC9C,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AAAA;AAAA;AAAA;AAIX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAqBf,IAAI,YAAY,iBAAiB,EAAE,WAAW,SAAS,CAAC;AACxD,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,kBAAkB,CAAC,SAAS;AAC9B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,IAAI;AACjC,eAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,sEAAsE;AACrH,QAAM,QAAQ,WAAW,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAC/C,QAAM,mBAAmB,CAAC,CAAC,GAAG,CAAC,CAAC;AAChC,mBAAiB,KAAK,GAAG,QAAQ;AACjC,WAAS,IAAI,IAAI,WAAW,QAAQ,IAAI,EAAE,MAAM,QAAQ,EAAE,GAAG;AAC3D,qBAAiB,KAAK,CAAC,GAAG,CAAC,CAAC;AAAA,EAC9B;AACA,QAAM,YAAY,CAAC;AACnB,QAAM,UAAU,OAAO;AAAA,IACrB,QAAQ,EAAE,EAAE;AAAA,IACZ,SAAS;AAAA,IACT,OAAO,EAAE,UAAU,kBAAkB,eAAe,EAAE;AAAA,EACxD,CAAC;AACD,QAAM,sBAAsB,qBAAqB,YAAY,QAAQ,OAAO,YAAY,OAAO,KAAK;AACpG,QAAM,oCAAoC,qBAAqB,YAAY,oBAAoB,QAAQ,WAAW,QAAQ,KAAK;AAC/H,QAAM,eAAe,qBAAqB,oBAAoB,QAAQ,OAAO,YAAY,OAAO,KAAK;AACrG,QAAM,kBAAkB,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,oBAAoB,EAAE,CAAC;AACrH,QAAM,WAAW,WAAW;AAAA,IAC1B,QAAQ,EAAE,GAAG,gBAAgB;AAAA,IAC7B,SAAS;AAAA,IACT,OAAO,EAAE,MAAM,kCAAkC;AAAA,EACnD,CAAC;AACD,QAAM,SAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,SAAS,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,aAAa,EAAE,CAAC;AACtG,YAAU,KAAK,OAAO;AACtB,YAAU,KAAK,eAAe;AAC9B,YAAU,KAAK,QAAQ;AACvB,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,SAAS,QAAQ,YAAY,aAAa,IAAI;AACtD,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,WACT,WAAW,OAAO;AAAA,EAC3B;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,WACT,QAAQ,OAAO;AAAA,EACxB;AACA,MAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,UAAM,IAAI,MAAM;AAAA,WACT,OAAO,OAAO;AAAA,EACvB;AACA,MAAI,aAAa,MAAM,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM;AAAA,UACV,aAAa,OAAO;AAAA,EAC5B;AACA,QAAM,WAAW,SAAS,SAAS,QAAQ,MAAM;AACjD,QAAM,UAAU,SAAS,SAAS,OAAO,MAAM;AAC/C,QAAM,cAAc,SAAS,SAAS,WAAW,MAAM;AACvD,QAAM,gBAAgB,SAAS,SAAS,aAAa,MAAM,EAAE;AAC7D,QAAM,CAAC,eAAe,oBAAoB,cAAc,mBAAmB,eAAe,IAAI,2BAA2B,UAAU,QAAQ,OAAO,QAAQ,OAAO,SAAS,OAAO,OAAO,aAAa,aAAa;AAClN,SAAO;AAAA,IACL,SAAS,eAAe,oBAAoB,QAAQ,OAAO,aAAa;AAAA,IACxE,SAAS,eAAe,CAAC,mBAAmB,EAAE,GAAG,OAAO,OAAO,YAAY;AAAA,IAC3E,SAAS,eAAe,CAAC,kBAAkB,MAAM,GAAG,QAAQ,IAAI,WAAW,kBAAkB,IAAI,CAAC,UAAU,OAAO,KAAK,CAAC,CAAC,CAAC;AAAA,IAC3H,SAAS,eAAe,CAAC,gBAAgB,MAAM,GAAG,QAAQ,OAAO,IAAI,WAAW,eAAe,CAAC;AAAA,EAClG;AACF;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,cAAc,YAAY,SAAS,IAAI;AAC/C,MAAI,aAAa,MAAM,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM,uDAAuD,aAAa,OAAO;AAAA,EAC7F;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM,qDAAqD,WAAW,OAAO;AAAA,EACzF;AACA,MAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,UAAM,IAAI,MAAM,sDAAsD,SAAS,OAAO;AAAA,EACxF;AACA,QAAM,cAAc,MAAM,KAAK,SAAS,SAAS,WAAW,MAAM,CAAC;AACnE,QAAM,gBAAgB,SAAS,SAAS,aAAa,MAAM;AAC3D,QAAM,cAAc,MAAM,KAAK,SAAS,SAAS,SAAS,MAAM,CAAC;AACjE,QAAM,CAAC,YAAY,cAAc,WAAW,IAAI,qBAAqB,eAAe,aAAa,OAAO,aAAa,OAAO,aAAa,WAAW;AACpJ,SAAO;AAAA,IACL,SAAS,eAAe,cAAc,aAAa,OAAO,UAAU;AAAA,IACpE,SAAS,eAAe,CAAC,YAAY,MAAM,GAAG,SAAS,OAAO,IAAI,WAAW,WAAW,CAAC;AAAA,EAC3F;AACF;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,mBAAmB,MAAM;AAChC,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,SAAS,WAAW,IAAI;AACtC,MAAI,KAAK,MAAM,SAAS,GAAG;AACzB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,gBACJ,QAAQ,OAAO;AAAA,EAC7B;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,gBACJ,WAAW,OAAO;AAAA,EAChC;AACA,QAAM,QAAQ,SAAS,SAAS,KAAK,MAAM;AAC3C,QAAM,WAAW,SAAS,SAAS,QAAQ,MAAM;AACjD,QAAM,cAAc,SAAS,SAAS,WAAW,MAAM;AACvD,QAAM,CAAC,YAAY,eAAe,IAAI,8BAA8B,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,aAAa,IAAI;AAC9H,SAAO,SAAS,eAAe,iBAAiB,KAAK,OAAO,UAAU;AACxE;AACA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,kBAAkB,MAAM;AAC/B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,MAAM,SAAS,WAAW,IAAI;AACtC,MAAI,KAAK,MAAM,SAAS,GAAG;AACzB,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AACA,MAAI,QAAQ,MAAM,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM;AAAA,eACL,QAAQ,OAAO;AAAA,EAC5B;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,eACL,WAAW,OAAO;AAAA,EAC/B;AACA,QAAM,QAAQ,SAAS,SAAS,KAAK,MAAM;AAC3C,QAAM,WAAW,SAAS,SAAS,QAAQ,MAAM;AACjD,QAAM,cAAc,SAAS,SAAS,WAAW,MAAM;AACvD,QAAM,CAAC,YAAY,eAAe,IAAI,8BAA8B,OAAO,KAAK,OAAO,KAAK,OAAO,UAAU,WAAW;AACxH,SAAO,SAAS,eAAe,iBAAiB,KAAK,OAAO,UAAU;AACxE;AACA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,eAAe,cAAc,aAAa,IAAI;AACtD,QAAM,EAAE,YAAY,IAAI;AACxB,QAAM,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW,IAAI,qBAAqB,gBAAgB,cAAc,eAAe,WAAW;AAC/I,QAAM,iBAAiB;AACvB,MAAI,aAAa,UAAU,UAAU;AACnC,UAAM,aAAa,SAAS,WAAW,aAAa;AACpD,UAAM,aAAa,SAAS,WAAW,YAAY;AACnD,UAAM,gBAAgB,aAAa,aAAa,SAAS,SAAS,aAAa,MAAM,EAAE,EAAE;AACzF,UAAM,SAAS,eAAe,YAAY,YAAY,aAAa,YAAY,WAAW,YAAY,WAAW,SAAS,eAAe,cAAc;AACvJ,WAAO,SAAS,eAAe,aAAa,OAAO,OAAO,OAAO,MAAM;AAAA,EACzE;AACA,QAAM,UAAU,IAAI,eAAe,YAAY,WAAW,cAAc,MAAM,QAAQ,aAAa,MAAM,QAAQ,SAAS,CAAC,YAAY,CAAC,GAAG,cAAc;AACzJ,QAAM,MAAM,SAAS,gBAAgB,SAAS,CAAC,cAAc,eAAe,YAAY,GAAG,aAAa,KAAK;AAC7G,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,YAAY,EAAE,CAAC;AAClG,WAAS,8BAA8B,GAAG;AAC1C,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,iBAAiB,KAAK,IAAI;AAClC,QAAM,QAAQ,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AACzD,QAAM,aAAa,qBAAqB,iBAAiB,GAAG,iBAAiB,KAAK;AAClF,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,IAAI,MAAM,KAAK,EAAE,KAAK,CAAC;AACrC,QAAM,OAAO,EAAE,MAAM,MAAM;AAC3B,SAAO,WAAW,IAAI,CAAC,MAAM;AAC3B,UAAM,YAAY,CAAC,GAAG,IAAI;AAC1B,cAAU,SAAS;AACnB,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,MAAM,UAAU,EAAE,CAAC;AAC7F,UAAM,UAAU;AAChB,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AACX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,MAAM,iBAAiB,MAAM,eAAe,YAAY,CAAC;AACnG,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,SAAS;AACb,IAAI,UAAU,iBAAiB,EAAE,WAAW,OAAO,CAAC;AACpD,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,qBAAqB;AACzB,IAAI,qBAAqB,kBAAkB,EAAE,WAAW,oBAAoB,iBAAiB,mBAAmB,CAAC;AACjH,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,GAAG;AACnD,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,YAAY,oBAAoB;AAAA,mCACL,MAAM;AAAA;AAEvC,QAAM,UAAU,IAAI,eAAe,EAAE,OAAO,SAAS;AACrD,SAAO,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AACvD;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,sBAAsB,MAAM;AAAA,EAC9B,YAAY,OAAO,SAAS,MAAM;AAChC,SAAK,gBAAgB,CAAC,GAAG;AACzB,SAAK,cAAc;AACnB,UAAM,OAAO,KAAK;AAClB,UAAM,aAAa,kBAAkB,KAAK,MAAM;AAChD,UAAM,QAAQ,kBAAkB,KAAK,MAAM;AAC3C,QAAI,YAAY;AAChB,QAAI,SAAS,GAAG;AACd,kBAAY;AAAA,IACd,OAAO;AACL,UAAI,aAAa;AACjB,kBAAY,KAAK,IAAI,CAAC,GAAG,MAAM;AAC7B;AACA,eAAO,KAAK,WAAW,IAAI,oBAAoB,cAAc,OAAO,UAAU,aAAa,gBAAgB,cAAc;AAAA,MAC3H,CAAC,EAAE,KAAK,GAAG;AAAA,IACb;AACA,SAAK,WAAW;AAAA,QACZ,sBAAsB,cAAc;AAAA,QACpC,wBAAwB,cAAc;AAAA;AAAA;AAAA,UAGpC;AAAA,yBACe;AAAA;AAAA;AAAA,EAGvB;AACF;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,eAAe,IAAI;AAC/F,QAAM,EAAE,kBAAkB,YAAY,YAAY,WAAW,eAAe,OAAO,QAAQ,KAAK,MAAM,SAAS,SAAS,IAAI,mBAAmB,UAAU,EAAE,OAAO,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,cAAc;AACpP,MAAI;AACJ,MAAI,YAAY;AACd,aAAS,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAAA,EACtF,WAAW,aAAa,eAAe;AACrC,iBAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,yCAAyC,EAAE,MAAM,QAAQ;AACxG,UAAM,OAAO,mBAAmB,gBAAgB,QAAQ,MAAM,QAAQ;AACtE,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,QAAQ,KAAK,EAAE,CAAC;AAC1F,aAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAC5F,aAAS,8BAA8B,MAAM;AAAA,EAC/C,OAAO;AACL,UAAM,qBAAqB,SAAS,mBAAmB,CAAC,CAAC,CAAC;AAC1D,QAAI,oBAAoB;AACtB,YAAM,SAAS,SAAS,SAAS,EAAE,MAAM;AACzC,YAAM,OAAO,OAAO,EAAE,OAAO,EAAE,OAAO,MAAM;AAC5C,YAAM,eAAe,oBAAoB,kBAAkB,MAAM,UAAU,MAAM;AACjF,eAAS,SAAS,eAAe,YAAY,EAAE,OAAO,aAAa,MAAM;AAAA,IAC3E,OAAO;AACL,YAAM,UAAU,IAAI,oBAAoB,QAAQ,UAAU,gBAAgB;AAC1E,eAAS,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AAAA,IACzD;AAAA,EACF;AACA,QAAM,iBAAiB,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAC1G,WAAS,8BAA8B,MAAM;AAC7C,SAAO;AACT;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,WAAW,aAAa,SAAS,UAAU,WAAW,UAAU,uBAAuB,IAAI;AACnG,QAAM,EAAE,MAAM,WAAW,IAAI;AAC7B,QAAM,QAAQ,SAAS,SAAS,KAAK,MAAM;AAC3C,QAAM,cAAc,SAAS,SAAS,WAAW,MAAM;AACvD,QAAM,CAAC,QAAQ,YAAY,IAAI,oBAAoB,OAAO,aAAa,WAAW,aAAa,SAAS,WAAW,UAAU,sBAAsB;AACnJ,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,OAAO,MAAM,GAAG,UAAU,MAAM;AAAA,IACzD,SAAS,eAAe,WAAW,OAAO,SAAS,YAAY;AAAA,EACjE;AACF;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,UAAU,IAAI;AACtB,QAAM,EAAE,OAAO,QAAQ,UAAU,IAAI;AACrC,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AACA,MAAI,OAAO,MAAM,WAAW,GAAG;AAC7B,UAAM,IAAI,MAAM,sCAAsC,OAAO,OAAO;AAAA,EACtE;AACA,MAAI,UAAU,MAAM,WAAW,GAAG;AAChC,UAAM,IAAI,MAAM,0CAA0C,UAAU,OAAO;AAAA,EAC7E;AACA,QAAM,SAAS,SAAS,SAAS,OAAO,MAAM;AAC9C,QAAM,aAAa,SAAS,SAAS,UAAU,MAAM,EAAE;AACvD,QAAM,CAAC,SAAS,QAAQ,KAAK,IAAI,mBAAmB,QAAQ,YAAY,SAAS;AACjF,QAAM,aAAa,OAAO;AAC1B,SAAO;AAAA,IACL,SAAS,eAAe,CAAC,YAAY,CAAC,GAAG,SAAS,OAAO;AAAA,IACzD,SAAS,eAAe,CAAC,UAAU,GAAG,UAAU,MAAM;AAAA,IACtD,SAAS,eAAe,CAAC,CAAC,GAAG,SAAS,IAAI,WAAW,KAAK,CAAC;AAAA,EAC7D;AACF;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,wBAAwB,MAAM;AACrC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,WAAW,IAAI;AACvB,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,IAAI,MAAM,kCAAkC;AAAA,EACpD;AACA,MAAI,cAAc,GAAG;AACnB,UAAM,IAAI,MAAM,sCAAsC;AAAA,EACxD;AACA,QAAM,SAAS,SAAS,SAAS,OAAO,MAAM;AAC9C,QAAM,SAAS,8BAA8B,QAAQ,UAAU;AAC/D,SAAO,SAAS,eAAe,OAAO,OAAO,SAAS,MAAM;AAC9D;AACA,IAAI,gCAAgC;AAAA,EAClC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,MAAM;AACV,IAAI,OAAO,iBAAiB,EAAE,WAAW,IAAI,CAAC;AAC9C,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,OAAO;AAAA;AAAA;AAAA;AAIX,IAAI,QAAQ,iBAAiB,EAAE,WAAW,KAAK,CAAC;AAChD,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,QAAQ,MAAM;AACxB,SAAK,gBAAgB,CAAC,GAAG;AACzB,UAAM,cAAc,IAAI,MAAM,OAAO,MAAM;AAC3C,aAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,KAAK;AAC3C,kBAAY,KAAK,OAAO,KAAK,KAAK;AAAA,IACpC;AACA,SAAK,cAAc;AACnB,SAAK,OAAO,YAAY;AACxB,UAAM,QAAQ,kBAAkB,KAAK,IAAI;AACzC,UAAM,eAAe,iBAAiB,MAAM;AAC5C,SAAK,WAAW;AAAA;AAAA,UAEV;AAAA,yBACe;AAAA;AAAA;AAAA,EAGvB;AACF;AACA,SAAS,iBAAiB,QAAQ;AAChC,QAAM,OAAO,OAAO;AACpB,MAAI,OAAO,GAAG;AACZ,UAAM,MAAM,iBAAiB,2BAA2B;AAAA,EAC1D;AACA,MAAI,SAAS,GAAG;AACd,WAAO,eAAe,OAAO;AAAA,EAC/B;AACA,QAAM,gBAAgB,CAAC,WAAW,WAAW,WAAW,WAAW,SAAS;AAC5E,QAAM,eAAe,CAAC;AACtB,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,iBAAa,KAAK,QAAQ,cAAc,OAAO,OAAO,KAAK;AAAA,EAC7D;AACA,SAAO,aAAa,KAAK;AAC3B;AAGA,SAAS,MAAM,QAAQ;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,EAAE,UAAU,YAAY,EAAE,MAAM,SAAS,GAAG;AAC9C,UAAM,OAAO,SAAS,SAAS,EAAE,MAAM;AACvC,UAAM,QAAQ,EAAE,UAAU,WAAW,KAAK,IAAI,CAAC,MAAM,aAAa,aAAa,CAAC,CAAC,IAAI;AACrF,UAAM,MAAM,OAAO,EAAE,OAAO,EAAE,OAAO,KAAK;AAC1C,UAAM,SAAS,YAAY,KAAK,IAAI;AACpC,WAAO,SAAS,eAAe,OAAO,OAAO,OAAO,OAAO,OAAO,MAAM;AAAA,EAC1E;AACA,QAAM,UAAU,IAAI,YAAY,EAAE,OAAO,IAAI;AAC7C,QAAM,SAAS,SAAS,gBAAgB,SAAS,CAAC,CAAC,GAAG,EAAE,KAAK;AAC7D,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,MAAM;AAAA,EACtB,YAAY,OAAO;AACjB,SAAK,gBAAgB,CAAC,KAAK,SAAS;AACpC,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,KAAK,MAAM,MAAM;AAAA,MACzB,EAAE,MAAM,aAAa,MAAM,MAAM;AAAA,MACjC,EAAE,MAAM,eAAe,MAAM,QAAQ;AAAA,MACrC,EAAE,MAAM,OAAO,MAAM,MAAM;AAAA,MAC3B,EAAE,MAAM,OAAO,MAAM,MAAM;AAAA,IAC7B;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAyClB;AACF;AACA,IAAI,eAAe,MAAM;AAAA,EACvB,YAAY,OAAO;AACjB,SAAK,gBAAgB,CAAC,KAAK,SAAS;AACpC,SAAK,iBAAiB;AAAA,MACpB,EAAE,MAAM,KAAK,MAAM,MAAM;AAAA,MACzB,EAAE,MAAM,aAAa,MAAM,MAAM;AAAA,MACjC,EAAE,MAAM,KAAK,MAAM,MAAM;AAAA,IAC3B;AACA,SAAK,cAAc;AACnB,SAAK,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAmClB;AACF;AAGA,SAAS,oCAAoC,UAAU,YAAY;AACjE,MAAI,eAAe,MAAM;AACvB,aAAS,8BAA8B,UAAU;AAAA,EACnD;AACF;AACA,SAAS,cAAc,KAAK;AAC1B,MAAI,QAAQ;AACZ,SAAO,QAAQ,KAAK;AAClB,aAAS;AAAA,EACX;AACA,SAAO;AACT;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,2CAA2C,IAAI,EAAE,UAAU,0CAA0C;AAC3G,QAAM,+BAA+B,IAAI,EAAE,UAAU,8BAA8B;AACnF,QAAM,SAAS,EAAE;AACjB,QAAM,UAAU,OAAO,OAAO,SAAS;AACvC,MAAI,SAAS,mBAAmB,CAAC,CAAC,CAAC,KAAK,UAAU,4CAA4C,IAAI,8BAA8B;AAC9H,UAAM,QAAQ,SAAS,SAAS,EAAE,MAAM;AACxC,UAAM,CAAC,aAAa,cAAc,IAAI,YAAY,OAAO,QAAQ,EAAE,OAAO,GAAG,MAAM;AACnF,WAAO;AAAA,MACL,SAAS,eAAe,YAAY,OAAO,YAAY,OAAO,YAAY,MAAM;AAAA,MAChF,SAAS,eAAe,eAAe,OAAO,eAAe,OAAO,eAAe,MAAM;AAAA,IAC3F;AAAA,EACF;AACA,MAAI,MAAM,GAAG;AACX,WAAO,OAAO,SAAS,KAAK;AAC5B,WAAO;AAAA,MACL,SAAS,eAAe,QAAQ,EAAE,OAAO,CAAC,CAAC;AAAA,MAC3C,SAAS,eAAe,QAAQ,SAAS,CAAC,CAAC;AAAA,IAC7C;AAAA,EACF;AACA,MAAI,YAAY,GAAG;AACjB,WAAO;AAAA,MACL;AAAA,MACA,MAAM,EAAE,OAAO,EAAE,OAAO,QAAQ,OAAO,SAAS,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,IACjF;AAAA,EACF;AACA,QAAM,WAAW,SAAS,QAAQ,IAAI,EAAE,MAAM;AAC9C,QAAM,YAAY,aAAa,QAAQ,SAAS;AAChD,QAAM,YAAY,YAAY,SAAS,aAAa,CAAC,IAAI;AACzD,QAAM,QAAQ,aAAa,cAAc,MAAM;AAC/C,QAAM,QAAQ,QAAQ;AACtB,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,OAAO,EAAE,OAAO,CAAC,OAAO,OAAO,EAAE,GAAG,SAAS,SAAS,CAAC;AACxG,MAAI,WAAW;AACb,wCAAoC,UAAU,SAAS;AAAA,EACzD;AACA,QAAM,QAAQ,cAAc,CAAC;AAC7B,QAAM,cAAc,cAAc,OAAO;AACzC,MAAI,UAAU;AACd,QAAM,YAAY,MAAM,YAAY,OAAO,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK,OAAO;AACrE,QAAM,UAAU,CAAC,KAAK,KAAK,UAAU;AACnC,UAAM,UAAU,UAAU;AAC1B,UAAM,UAAU,IAAI,YAAY,KAAK;AACrC,UAAM,WAAW,YAAY,OAAO,IAAI;AACxC,UAAM,eAAe,CAAC,CAAC,OAAO,GAAG,CAAC,QAAQ,GAAG,CAAC,OAAO,iBAAiB,GAAG,CAAC,GAAG,GAAG,CAAC,GAAG,CAAC;AACrF,UAAM,eAAe;AACrB,cAAU,SAAS,gBAAgB,SAAS,SAAS,SAAS,YAAY;AAC1E,wCAAoC,UAAU,YAAY;AAAA,EAC5D;AACA,WAAS,MAAM,GAAG,MAAM,OAAO,OAAO,GAAG;AACvC,UAAM,MAAM,MAAM;AAClB,aAAS,MAAM,KAAK,OAAO,GAAG,OAAO,GAAG;AACtC,cAAQ,KAAK,KAAK,CAAC,OAAO,WAAW,CAAC;AAAA,IACxC;AAAA,EACF;AACA,WAAS,cAAc,aAAa,cAAc,OAAO,eAAe,GAAG;AACzE,UAAM,UAAU,UAAU;AAC1B,UAAM,eAAe,IAAI,aAAa,CAAC,OAAO,cAAc,CAAC,CAAC;AAC9D,UAAM,YAAY,YAAY,OAAO,IAAI;AACzC,UAAM,eAAe,CAAC,CAAC,OAAO,GAAG,CAAC,SAAS,GAAG,CAAC,KAAK,CAAC;AACrD,UAAM,eAAe;AACrB,cAAU,SAAS,gBAAgB,cAAc,SAAS,SAAS,YAAY;AAC/E,wCAAoC,UAAU,YAAY;AAC1D,UAAM,MAAM,QAAQ;AACpB,UAAM,MAAM,MAAM;AAClB,aAAS,MAAM,KAAK,OAAO,GAAG,OAAO,GAAG;AACtC,cAAQ,KAAK,KAAK,QAAQ,KAAK;AAAA,IACjC;AAAA,EACF;AACA,MAAI,cAAc;AAClB,YAAU,OAAO,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,GAAG,MAAM,CAAC,OAAO,CAAC,EAAE,EAAE,CAAC;AACrG,sCAAoC,UAAU,WAAW;AACzD,MAAI,SAAS,UAAU,EAAE,QAAQ,EAAE,GAAG,KAAK,QAAQ,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,GAAG,WAAW,EAAE,EAAE,CAAC;AAC3G,sCAAoC,UAAU,GAAG;AACjD,QAAM,WAAW,OAAO,MAAM,GAAG,EAAE;AACnC,WAAS,KAAK,CAAC;AACf,gBAAc;AACd,YAAU,SAAS,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AAC5F,sCAAoC,UAAU,WAAW;AACzD,QAAM,aAAa;AACnB,WAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,OAAO,EAAE,OAAO,SAAS,GAAG,SAAS,SAAS,CAAC;AAC1F,sCAAoC,UAAU,UAAU;AACxD,SAAO,CAAC,QAAQ,OAAO;AACzB;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,aAAa,YAAY,eAAe,UAAU,WAAW,UAAU;AACjF,SAAK,gBAAgB,CAAC,SAAS,YAAY;AAC3C,SAAK,cAAc;AACnB,UAAM,sBAAsB,kBAAkB,YAAY,IAAI;AAC9D,QAAI;AACJ,YAAQ,UAAU;AAAA,MAChB,KAAK;AACH,qBAAa;AACb;AAAA,MACF,KAAK;AACH,qBAAa;AACb;AAAA,MACF,KAAK;AACH,qBAAa;AACb;AAAA,MACF,KAAK;AACH,qBAAa;AACb;AAAA,MACF;AACE,qBAAa;AACb;AAAA,IACJ;AACA,SAAK,WAAW;AAAA;AAAA;AAAA,mBAGD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAwBQ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,2BAiBA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,4CAUiB,0CAA0C;AAAA;AAAA;AAAA,sCAGhD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,sCAwBA;AAAA;AAAA;AAAA;AAAA,mDAIa;AAAA,mDACA;AAAA;AAAA,sBAE7B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAyBpB;AACF;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,WAAW,IAAI;AACtC,QAAM,EAAE,eAAe,UAAU,WAAW,YAAY,IAAI;AAC5D,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,QAAM,CAAC,WAAW,QAAQ,IAAI,eAAe,OAAO,cAAc,CAAC,aAAa,UAAU;AAC1F,QAAM,WAAW;AAAA,IACf;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,QAAM,UAAU,IAAI,iBAAiB,aAAa,YAAY,eAAe,UAAU,WAAW,QAAQ;AAC1G,SAAO,SAAS,gBAAgB,SAAS,CAAC,QAAQ,UAAU,GAAG,SAAS;AAC1E;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,EAAE,EAAE,IAAI;AACd,oBAAkB,GAAG,QAAQ;AAC7B,UAAQ,KAAK,aAAa,4DAA4D;AACtF,QAAM,SAAS,SAAS,SAAS,EAAE,MAAM;AACzC,QAAM,EAAE,cAAc,aAAa,QAAQ,IAAI,cAAc,QAAQ,MAAM,EAAE,OAAO,EAAE,KAAK;AAC3F,SAAO;AAAA,IACL,SAAS,eAAe,aAAa,EAAE,OAAO,YAAY;AAAA,IAC1D,SAAS,eAAe,CAAC,QAAQ,MAAM,GAAG,SAAS,OAAO;AAAA,EAC5D;AACF;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,IAAI;AAClB,MAAI,EAAE,KAAK,IAAI;AACf,MAAI,OAAO,GAAG;AACZ,YAAQ,MAAM,MAAM;AAAA,EACtB;AACA,QAAM,IAAI;AACV,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,MAAM,MAAM,MAAM;AACxB,QAAM,WAAW,IAAI,MAAM,QAAQ,CAAC;AACpC,MAAI,WAAW;AACf,WAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,QAAI,MAAM,MAAM;AACd,eAAS,cAAc,EAAE,MAAM;AAAA,IACjC;AAAA,EACF;AACA,QAAM,YAAY,CAAC;AACnB,QAAM,QAAQ,IAAI,MAAM,KAAK,EAAE,KAAK,CAAC;AACrC,QAAM,OAAO,EAAE,MAAM,MAAM;AAC3B,OAAK,QAAQ;AACb,QAAM,MAAM,IAAI,MAAM,GAAG;AACzB,WAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,UAAM,QAAQ;AACd,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,KAAK,EAAE,CAAC;AAClF,UAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAClG,QAAI,KAAK;AACT,cAAU,KAAK,MAAM;AAAA,EACvB;AACA,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,mBAAmB,MAAM;AAAA,EAC3B,YAAY,WAAW,WAAW;AAChC,SAAK,gBAAgB,CAAC,KAAK,YAAY;AACvC,UAAM,aAAa,UAAU;AAC7B,UAAM,YAAY,UAAU;AAC5B,UAAM,SAAS,UAAU;AACzB,UAAM,cAAc,UAAU;AAC9B,UAAM,UAAU,cAAc,KAAK,KAAK,SAAS,UAAU;AAC3D,SAAK,cAAc,CAAC,WAAW,OAAO;AACtC,UAAM,sBAAsB;AAC5B,UAAM,cAAc;AACpB,UAAM,wBAAwB,KAAK,MAAM,aAAa,CAAC,IAAI;AAC3D,UAAM,0BAA0B,aAAa;AAC7C,UAAM,gBAAgB;AAAA;AAAA;AAGtB,QAAI,wBAAwB;AAC5B,QAAI,SAAS,aAAa,GAAG;AAC3B,8BAAwB;AAAA,oCACM;AAAA;AAAA;AAAA;AAAA,IAIhC;AACA,QAAI,4BAA4B;AAChC,QAAI,SAAS,aAAa,GAAG;AAC3B,kCAA4B;AAAA,oCACE;AAAA;AAAA;AAAA;AAAA,IAIhC;AACA,SAAK,WAAW;AAAA,0CACsB;AAAA;AAAA;AAAA,UAGhC;AAAA;AAAA;AAAA;AAAA;AAAA,UAKA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YASE,yBAAyB;AAAA,wDACmB;AAAA;AAAA;AAAA;AAAA,8BAI1B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAgBlB;AAAA;AAAA;AAAA,iCAGqB;AAAA,cACnB,4BAA4B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAiB9B;AAAA,qBACS,4BAA4B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAerC;AAAA,qBACS,4BAA4B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YAerC;AAAA;AAAA,oBAEQ;AAAA;AAAA;AAAA,EAGlB;AACF;AAGA,SAAS,oBAAoB,MAAM;AACjC,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,WAAW,IAAI;AAC1B,QAAM,EAAE,YAAY,IAAI;AACxB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,YAAY,CAAC;AACnB,MAAI,OAAO;AACX,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,KAAK;AACzE,MAAI,YAAY;AAChB,MAAI,eAAe,MAAM;AACvB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,YAAY,EAAE,CAAC;AACzF,cAAU,KAAK,SAAS;AACxB,WAAO,qBAAqB,iBAAiB,GAAG,KAAK,EAAE;AAAA,EACzD;AACA,QAAM,WAAW,qBAAqB,aAAa,gBAAgB,UAAU,OAAO,MAAM,WAAW;AACrG,QAAM,SAAS,aAAa,cAAc,CAAC,UAAU,MAAM,KAAK,CAAC;AACjE,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,CAAC,IAAI,MAAM,EAAE,EAAE,CAAC;AACpG,YAAU,KAAK,GAAG;AAClB,QAAM,cAAc,WAAW,EAAE,KAAK;AACtC,QAAM,eAAe,CAAC,IAAI,WAAW,aAAa,OAAO,iBAAiB;AACxE,UAAM,YAAY,GAAG,MAAM;AAC3B,UAAM,UAAU,GAAG,MAAM;AACzB,UAAM,aAAa,qBAAqB,aAAa,8BAA8B,SAAS,YAAY;AACxG,UAAM,YAAY,EAAE,YAAY,QAAQ,SAAS,WAAW,aAAa,aAAa;AACtF,UAAM,UAAU,IAAI,iBAAiB,WAAW,SAAS;AACzD,UAAM,SAAS,SAAS,cAAc,SAAS,CAAC,IAAI,WAAW,GAAG,KAAK;AACvE,cAAU,KAAK,MAAM;AACrB,QAAI,OAAO,MAAM,OAAO,cAAc;AACpC,aAAO;AAAA,IACT;AACA,UAAM,YAAY,OAAO;AAAA,MACvB,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,GAAG,MAAM,cAAc,MAAM,GAAG,OAAO,UAAU;AAAA,IACnE,CAAC;AACD,UAAM,WAAW,MAAM;AAAA,MACrB,QAAQ,EAAE,GAAG,UAAU;AAAA,MACvB,SAAS;AAAA,MACT,OAAO,EAAE,MAAM,CAAC,UAAU,UAAU,EAAE;AAAA,IACxC,CAAC;AACD,cAAU,KAAK,SAAS;AACxB,cAAU,KAAK,QAAQ;AACvB,UAAM,UAAU,aAAa,QAAQ,WAAW,UAAU,OAAO,YAAY;AAC7E,WAAO;AAAA,EACT;AACA,QAAM,cAAc,aAAa,KAAK,sBAAsB,YAAY,aAAa,WAAW;AAChG,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACvG,MAAI,SAAS;AACb,MAAI,eAAe,MAAM;AACvB,cAAU,KAAK,QAAQ;AACvB,UAAM,OAAO,qBAAqB,uBAAuB,WAAW;AACpE,aAAS,WAAW,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AAAA,EACnF;AACA,YAAU,QAAQ,CAAC,MAAM,SAAS,8BAA8B,CAAC,CAAC;AAClE,SAAO;AACT;AACA,IAAI,4BAA4B;AAAA,EAC9B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,WAAW,gBAAgB,gBAAgB;AACzC,iBAAe,YAAY;AAC7B;AAGA,IAAI;AAAA,CACH,SAAS,WAAW;AACnB,YAAU,UAAU,aAAa,KAAK;AACtC,YAAU,UAAU,WAAW,KAAK;AACpC,YAAU,UAAU,UAAU,KAAK;AACnC,YAAU,UAAU,YAAY,KAAK;AACrC,YAAU,UAAU,eAAe,KAAK;AAC1C,GAAG,aAAa,WAAW,CAAC,EAAE;AAC9B,IAAI;AAAA,CACH,SAAS,oBAAoB;AAC5B,qBAAmB,mBAAmB,YAAY,KAAK;AACvD,qBAAmB,mBAAmB,UAAU,KAAK;AACrD,qBAAmB,mBAAmB,WAAW,KAAK;AACtD,qBAAmB,mBAAmB,WAAW,KAAK;AACtD,qBAAmB,mBAAmB,eAAe,KAAK;AAC1D,qBAAmB,mBAAmB,aAAa,KAAK;AACxD,qBAAmB,mBAAmB,SAAS,KAAK;AACtD,GAAG,sBAAsB,oBAAoB,CAAC,EAAE;AAGhD,IAAI;AACJ,SAAS,MAAM,UAAU;AACvB,oBAAkB,SAAS,KAAK,MAAM,cAAc,MAAM;AAAA,IACxD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,iBAAiB,MAAM;AAC9B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,GAAG,MAAM,uBAAuB,IAAI;AAC/C,MAAI,EAAE,UAAU,aAAa,EAAE,UAAU,WAAW;AAClD,UAAM,IAAI,MAAM,6DAA6D;AAAA,EAC/E;AACA,QAAM,EAAE,YAAY,YAAY,YAAY,aAAa,eAAe,IAAI;AAC5E,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,SAAS;AACb,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,SAAS,UAAU,IAAI,KAAK,MAAM;AACnD,QAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,YAAM,IAAI,MAAM,uDAAuD,SAAS,MAAM,SAAS;AAAA,IACjG;AACA,aAAS,SAAS;AAAA,EACpB;AACA,QAAM,2BAA2B,0BAA0B,OAAO,IAAI,SAAS,UAAU,IAAI,uBAAuB,MAAM,EAAE;AAC5H,QAAM,kBAAkB,kBAAkB;AAC1C,MAAI,mBAAmB,MAAM;AAC3B,UAAM,IAAI,MAAM,GAAG,+EAA+E;AAAA,EACpG;AACA,QAAM,UAAU,aAAa,EAAE,MAAM,KAAK,EAAE,MAAM;AAClD,QAAM,WAAW,aAAa,EAAE,MAAM,KAAK,EAAE,MAAM;AACnD,QAAM,YAAY,uBAAuB,2BAA2B,EAAE,MAAM,MAAM,GAAG,EAAE,GAAG,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC;AAC9G,QAAM,MAAM,SAAS,WAAW,CAAC,GAAG,WAAW,SAAS,QAAQ,GAAG,EAAE,KAAK;AAC1E,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,kBAAgB,KAAK,aAAa,EAAE,MAAM,QAAQ,KAAK,aAAa,EAAE,MAAM,QAAQ,YAAY,YAAY,iBAAiB,QAAQ,0BAA0B,kBAAkB,GAAG,KAAK;AACzL,SAAO;AACT;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,wBAAwB,YAAY,SAAS;AACpD,MAAI;AACJ,WAAS,WAAW,UAAU;AAC5B,gBAAY,SAAS,KAAK,MAAM,YAAY,MAAM;AAAA,MAChD;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AACA,WAAS,YAAY,MAAM;AACzB,UAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,EAAE,EAAE,IAAI;AAC7C,UAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,MAAM,SAAS,WAAW,EAAE,OAAO,WAAW,EAAE,KAAK;AAC3D,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAI,aAAa,cAAc,IAAI,KAAK,MAAM,GAAG;AAC/C,aAAO;AAAA,IACT;AACA,cAAU,KAAK,SAAS,EAAE,QAAQ,KAAK;AACvC,WAAO;AAAA,EACT;AACA,SAAO,EAAE,YAAY,aAAa,QAAQ,WAAW,YAAY,YAAY,YAAY;AAC3F;AAGA,IAAI,aAAa,wBAAwB,GAAG;AAG5C,SAAS,yBAAyB,YAAY,yBAAyB,OAAO;AAC5E,MAAI;AACJ,WAAS,WAAW,UAAU;AAC5B,gBAAY,SAAS,KAAK,MAAM,YAAY,MAAM;AAAA,MAChD;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AACA,WAAS,YAAY,MAAM;AACzB,UAAM,EAAE,SAAS,UAAU,OAAO,IAAI;AACtC,UAAM,EAAE,GAAG,EAAE,IAAI;AACjB,UAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,aAAa,SAAS,OAAO,QAAQ,EAAE;AAC7C,UAAM,WAAW,qBAAqB,2BAA2B,EAAE,OAAO,EAAE,KAAK;AACjF,UAAM,MAAM,SAAS,WAAW,UAAU,UAAU;AACpD,QAAI,aAAa,cAAc,QAAQ,MAAM,GAAG;AAC9C,aAAO;AAAA,IACT;AACA,UAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,UAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,UAAM,cAAc,MAAM,UAAU,KAAK,aAAa,EAAE,MAAM,QAAQ,KAAK,aAAa,EAAE,MAAM,QAAQ,SAAS,EAAE,QAAQ,KAAK;AAChI,gBAAY;AACZ,WAAO;AAAA,EACT;AACA,SAAO,EAAE,YAAY,aAAa,QAAQ,WAAW,YAAY,YAAY,YAAY;AAC3F;AAGA,IAAI,wBAAwB;AAC5B,IAAI,aAAa,yBAAyB,KAAK,qBAAqB;AAGpE,IAAI;AACJ,SAAS,UAAU,UAAU;AAC3B,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM;AAAA,IACzC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,MAAM,SAAS,WAAW,OAAO,GAAG,OAAO,OAAO,GAAG,KAAK;AAChE,MAAI,aAAa,cAAc,IAAI,KAAK,MAAM,GAAG;AAC/C,WAAO;AAAA,EACT;AACA,QAAM,WAAW,OAAO,IAAI,CAAC,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE,EAAE;AACtE,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,QAAQ,EAAE,MAAM;AACpE,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,WAAS,eAAe,SAAS,QAAQ,SAAS,IAAI,QAAQ,KAAK;AACnE,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb;AAAA,EACA,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,IAAI;AAC7C,MAAI,EAAE,UAAU,UAAU;AACxB,WAAO,OAAO,SAAS,SAAS,EAAE,MAAM,GAAG,EAAE,OAAO,EAAE,KAAK;AAAA,EAC7D;AACA,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,SAAS,SAAS,mBAAmB,CAAC;AAC5C,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,IAAI,MAAM;AAClB,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,kBAAgB,SAAS,KAAK,MAAM,WAAW,MAAM;AAAA,IACnD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,CAAC,cAAc,IAAI,IAAI,kBAAkB,OAAO,EAAE,OAAO,MAAM,IAAI;AACzE,MAAI,aAAa;AACjB,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,QAAI,KAAK,OAAO,GAAG;AACjB,mBAAa;AAAA,IACf;AAAA,EACF;AACA,QAAM,WAAW,iBAAiB,OAAO,EAAE,OAAO,MAAM,IAAI;AAC5D,QAAM,IAAI;AAAA,IACR,QAAQ,OAAO,EAAE;AAAA,IACjB,OAAO;AAAA,IACP,OAAO,OAAO,EAAE;AAAA,EAClB;AACA,MAAI,YAAY;AACd,UAAM,SAAS,UAAU,EAAE,QAAQ,SAAS,SAAS,CAAC;AACtD,WAAO,QAAQ;AACf,WAAO;AAAA,EACT;AACA,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,YAAY,IAAI,WAAW,IAAI,WAAW,IAAI,EAAE,MAAM;AAC5D,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,gBAAc,KAAK,aAAa,EAAE,MAAM,QAAQ,SAAS,EAAE,QAAQ,OAAO,WAAW,KAAK,MAAM;AAChG,SAAO;AACT;AACA,SAAS,iBAAiB,SAAS,MAAM;AACvC,QAAM,WAAW,IAAI,MAAM,QAAQ,MAAM;AACzC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAS,KAAK,QAAQ,KAAK;AAAA,EAC7B;AACA,SAAO;AACT;AACA,SAAS,kBAAkB,OAAO,MAAM;AACtC,QAAM,WAAW,CAAC;AAClB,QAAM,UAAU,CAAC;AACjB,WAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,EAAE,GAAG;AACrC,QAAI,MAAM,OAAO,GAAG;AAClB,eAAS,KAAK,MAAM,EAAE;AAAA,IACxB;AACA,QAAI,MAAM,KAAK,QAAQ,GAAG;AACxB,cAAQ,KAAK,KAAK,EAAE;AAAA,IACtB;AAAA,EACF;AACA,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,QAAI,YAAY;AAChB,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,EAAE,GAAG;AACvC,UAAI,QAAQ,MAAM,MAAM,cAAc,MAAM,QAAQ,aAAa,QAAQ,KAAK;AAC5E,oBAAY;AAAA,MACd;AAAA,IACF;AACA,YAAQ,aAAa;AAAA,EACvB;AACA,SAAO,CAAC,UAAU,OAAO;AAC3B;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,SAAS,wBAAwB,GAAG,MAAM,UAAU;AAClD,QAAM,SAAS,EAAE;AACjB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,eAAe,aAAa,eAAe,MAAM,MAAM;AAC7D,MAAI,OAAO;AACX,QAAM,eAAe,qBAAqB,mBAAmB,MAAM,KAAK;AACxE,MAAI,cAAc;AAClB,MAAI,qBAAqB;AACzB,MAAI,gBAAgB,MAAM;AACxB,UAAM,WAAW,IAAI,MAAM,KAAK;AAChC,aAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,eAAS,KAAK,OAAO,aAAa;AAAA,IACpC;AACA,WAAO,qBAAqB,iBAAiB,KAAK,QAAQ,KAAK;AAC/D,kBAAc,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,MAAM,aAAa,GAAG,SAAS,SAAS,CAAC;AAC5F,UAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,eAAe,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AAChE,QAAI,iBAAiB,KAAK;AACxB,2BAAqB;AAAA,IACvB;AAAA,EACF;AACA,SAAO,EAAE,YAAY,aAAa,cAAc,MAAM,mBAAmB;AAC3E;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,YAAU,SAAS,KAAK,MAAM,KAAK,MAAM,CAAC,wBAAwB,CAAC;AACrE;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,aAAS;AACT,cAAU;AAAA,EACZ;AACA,QAAM,YAAY,OAAO,MAAM;AAC/B,uBAAqB,2BAA2B,OAAO,MAAM,SAAS;AACtE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,IAAI;AACjG,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAQ,SAAS,YAAY,KAAK;AAAA,EACpC;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,YAAU,SAAS,KAAK,MAAM,KAAK,MAAM,CAAC,wBAAwB,CAAC;AACrE;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,aAAS;AACT,cAAU;AAAA,EACZ;AACA,QAAM,YAAY,OAAO,MAAM;AAC/B,uBAAqB,2BAA2B,OAAO,MAAM,SAAS;AACtE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,IAAI;AACjG,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAQ,SAAS,YAAY,KAAK;AAAA,EACpC;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,cAAY,SAAS,KAAK,MAAM,QAAQ,MAAM;AAAA,IAC5C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AAC1F,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAI,iBAAiB,KAAK;AACxB,eAAS;AACT,gBAAU;AAAA,IACZ;AAAA,EACF;AACA,QAAM,WAAW,OAAO,MAAM,MAAM,GAAG,EAAE;AACzC,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO;AACjD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,YAAY,aAAa,cAAc,IAAI,KAAK;AACtD,QAAM,YAAY,OAAO,MAAM,KAAK;AACpC,YAAU,SAAS,SAAS,OAAO,QAAQ,WAAW,WAAW,KAAK;AACtE,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,gBAAc,SAAS,KAAK,MAAM,SAAS,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,IAAI,OAAO;AACjB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AAC9G,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,WAAW,SAAS;AAC1B,MAAI,SAAS,eAAe,gBAAgB;AAC1C,UAAM,IAAI,MAAM,6CAA6C,SAAS,yCAAyC;AAAA,EACjH;AACA,MAAI,SAAS,kBAAkB,KAAK,SAAS,mBAAmB,GAAG;AACjE,UAAM,IAAI,MAAM,0EAA0E,SAAS,mBAAmB,SAAS,iBAAiB;AAAA,EAClJ;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,cAAY,KAAK,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,cAAc,aAAa,QAAQ,UAAU,WAAW,SAAS,cAAc,aAAa,UAAU,KAAK;AAChK,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,MAAM,IAAI;AAC1B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,QAAQ,aAAa,cAAc,EAAE,KAAK;AAChD,QAAM,SAAS,aAAa,uBAAuB,OAAO,KAAK;AAC/D,eAAa,OAAO,UAAU,aAAa,cAAc,MAAM,GAAG,MAAM,cAAc,sBAAsB,EAAE,uEAAuE;AACrL,OAAK,QAAQ,OAAO,EAAE,MAAM;AAC5B,SAAO,EAAE,QAAQ,EAAE,QAAQ,OAAO,QAAQ,OAAO,EAAE,MAAM;AAC3D;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,oBAAkB,SAAS,KAAK,MAAM,aAAa,MAAM;AAAA,IACvD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,EAAE,IAAI;AACjB,QAAM,EAAE,YAAY,WAAW,IAAI;AACnC,MAAI,EAAE,UAAU,aAAa,EAAE,UAAU,WAAW;AAClD,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AACA,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,cAAc,aAAa,EAAE,MAAM,QAAQ,KAAK,EAAE,MAAM,QAAQ;AACtE,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,aAAa,EAAE,MAAM,MAAM,GAAG,EAAE;AACtC,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,YAAY,aAAa,cAAc,UAAU;AACvD,QAAM,oBAAoB,uBAAuB,2BAA2B,EAAE,MAAM,MAAM,GAAG,EAAE,GAAG,EAAE,MAAM,MAAM,GAAG,EAAE,CAAC;AACtH,QAAM,WAAW,kBAAkB,OAAO,CAAC,aAAa,WAAW,CAAC;AACpE,eAAa,OAAO,gBAAgB,aAAa,MAAM,kCAAkC,qBAAqB,uCAAuC,EAAE,aAAa,EAAE,wBAAwB,6BAA6B,wBAAwB;AACnP,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,WAAW,aAAa,CAAC,WAAW,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,WAAW;AAC1G,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,MAAM,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AACxF,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,UAAU,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACtD,QAAM,WAAW,aAAa,IAAI,MAAM,KAAK,IAAI,MAAM;AACvD,QAAM,WAAW,KAAK,IAAI,WAAW,SAAS;AAC9C,QAAM,MAAM,SAAS,WAAW,CAAC,UAAU,SAAS,QAAQ,GAAG,IAAI,KAAK;AACxE,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,IAAI,KAAK,EAAE,MAAM;AACnE,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,IAAI,KAAK,EAAE,MAAM;AACnE,kBAAgB,OAAO,aAAa,IAAI,MAAM,QAAQ,OAAO,aAAa,IAAI,MAAM,QAAQ,YAAY,YAAY,KAAK;AACzH,WAAS,YAAY,IAAI,MAAM;AAC/B,WAAS,YAAY,IAAI,MAAM;AAC/B,MAAI,QAAQ;AACZ,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,OAAO,KAAK,GAAG,SAAS,SAAS,IAAI;AACrE,QAAM,CAAC,QAAQ,KAAK,IAAI,mBAAmB,iBAAiB,GAAG,OAAO,IAAI;AAC1E,QAAM,cAAc,mBAAmB,iBAAiB,EAAE,OAAO,QAAQ,KAAK;AAC9E,QAAM,QAAQ,SAAS,SAAS,EAAE,MAAM;AACxC,QAAM,MAAM,SAAS,WAAW,OAAO,EAAE,KAAK;AAC9C,QAAM,WAAW,aAAa,eAAe,EAAE,KAAK;AACpD,QAAM,UAAU,SAAS,UAAU,IAAI,IAAI,MAAM;AACjD,MAAI,aAAa;AACf,UAAM,aAAa,mBAAmB,kBAAkB,QAAQ,QAAQ;AACxE,QAAI,EAAE,UAAU,UAAU;AACxB,cAAQ,cAAc,MAAM,MAAM,YAAY,aAAa,aAAa,cAAc,KAAK,CAAC;AAAA,IAC9F,OAAO;AACL,YAAM,WAAW,SAAS,mBAAmB,GAAG;AAChD,eAAS,IAAI,MAAM,SAAS,YAAY,aAAa,aAAa,cAAc,KAAK,CAAC,CAAC;AAAA,IACzF;AACA,WAAO;AAAA,EACT;AACA,MAAI,EAAE,UAAU,UAAU;AACxB,UAAM,MAAM,UAAU,OAAO,QAAQ,OAAO,EAAE,OAAO,EAAE,KAAK;AAC5D,YAAQ,cAAc;AACtB,WAAO;AAAA,EACT;AACA,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,QAAM,OAAO,EAAE,MAAM;AACrB,MAAI,SAAS,GAAG;AACd,aAAS,OAAO,SAAS,IAAI,SAAS,QAAQ,KAAK;AAAA,EACrD,WAAW,SAAS,GAAG;AACrB,aAAS,OAAO,SAAS,IAAI,SAAS,IAAI,SAAS,QAAQ,KAAK;AAAA,EAClE,WAAW,SAAS,GAAG;AACrB,aAAS,OAAO,SAAS,IAAI,SAAS,IAAI,SAAS,IAAI,SAAS,QAAQ,KAAK;AAAA,EAC/E,OAAO;AACL,UAAM,MAAM,UAAU,OAAO,QAAQ,OAAO,EAAE,OAAO,EAAE,KAAK;AAC5D,YAAQ,IAAI,GAAG;AAAA,EACjB;AACA,SAAO;AACT;AACA,SAAS,SAAS,OAAO,SAAS,SAAS,OAAO,MAAM;AACtD,MAAI,YAAY;AAChB,QAAM,SAAS,MAAM;AACrB,QAAM,SAAS,MAAM;AACrB,QAAM,OAAO,SAAS,KAAK;AAC3B,WAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,UAAM,UAAU,IAAI,UAAU;AAC9B,YAAQ,IAAI,MAAM,SAAS,SAAS,UAAU,KAAK,EAAE,GAAG,SAAS;AACjE,iBAAa,KAAK;AAAA,EACpB;AACF;AACA,SAAS,SAAS,OAAO,UAAU,UAAU,SAAS,OAAO,MAAM;AACjE,MAAI,YAAY;AAChB,QAAM,SAAS,MAAM;AACrB,QAAM,SAAS,MAAM;AACrB,QAAM,SAAS,MAAM;AACrB,QAAM,OAAO,SAAS,KAAK;AAC3B,QAAM,OAAO,SAAS,KAAK;AAC3B,WAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,aAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,YAAM,UAAU,IAAI,WAAW,IAAI,WAAW;AAC9C,cAAQ,IAAI,MAAM,SAAS,SAAS,UAAU,KAAK,EAAE,GAAG,SAAS;AACjE,mBAAa,KAAK;AAAA,IACpB;AAAA,EACF;AACF;AACA,SAAS,SAAS,OAAO,UAAU,UAAU,UAAU,SAAS,OAAO,MAAM;AAC3E,MAAI,YAAY;AAChB,QAAM,SAAS,MAAM;AACrB,QAAM,SAAS,MAAM;AACrB,QAAM,SAAS,MAAM;AACrB,QAAM,OAAO,SAAS,KAAK;AAC3B,QAAM,OAAO,SAAS,KAAK;AAC3B,QAAM,OAAO,SAAS,KAAK;AAC3B,QAAM,SAAS,MAAM;AACrB,WAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,aAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,eAAS,IAAI,QAAQ,IAAI,MAAM,KAAK;AAClC,cAAM,UAAU,IAAI,WAAW,IAAI,WAAW,IAAI,WAAW;AAC7D,gBAAQ,IAAI,MAAM,SAAS,SAAS,UAAU,KAAK,EAAE,GAAG,SAAS;AACjE,qBAAa,KAAK;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AACF;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,MAAM,IAAI;AAC9B,QAAM,QAAQ,WAAW,OAAO,CAAC,GAAG,MAAM,IAAI,CAAC;AAC/C,QAAM,WAAW,qBAAqB,YAAY,EAAE,OAAO,YAAY,KAAK;AAC5E,QAAM,WAAW,qBAAqB,YAAY,SAAS,QAAQ,WAAW,MAAM;AACpF,QAAM,mBAAmB,qBAAqB,oBAAoB,EAAE,OAAO,YAAY,KAAK;AAC5F,QAAM,mBAAmB,qBAAqB,oBAAoB,OAAO,WAAW,MAAM;AAC1F,QAAM,YAAY,qBAAqB,aAAa,kBAAkB,OAAO,WAAW,MAAM;AAC9F,QAAM,YAAY,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC3F,QAAM,cAAc,WAAW,EAAE,QAAQ,EAAE,GAAG,UAAU,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,SAAS,EAAE,CAAC;AACzG,QAAM,sBAAsB,SAAS,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,iBAAiB,EAAE,CAAC;AAC1H,QAAM,SAAS,OAAO;AAAA,IACpB,QAAQ,EAAE,GAAG,oBAAoB;AAAA,IACjC,SAAS;AAAA,IACT,OAAO,EAAE,OAAO,kBAAkB,MAAM,UAAU;AAAA,EACpD,CAAC;AACD,WAAS,YAAY,UAAU,MAAM;AACrC,WAAS,YAAY,YAAY,MAAM;AACvC,WAAS,YAAY,UAAU,MAAM;AACrC,SAAO;AACT;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,MAAM,GAAG,SAAS,SAAS,IAAI;AAC/D,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,KAAK;AAC9C,QAAM,SAAS,SAAS,mBAAmB,CAAC;AAC5C,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,IAAI,MAAM;AAClB,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,wBAAwB,IAAI;AAG9C,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,aAAW,SAAS,KAAK,MAAM,aAAa,MAAM;AAAA,IAChD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,cAAc,aAAa,IAAI;AACvC,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,WAAS,KAAK,cAAc,cAAc,KAAK;AAC/C,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,OAAO,aAAa,eAAe,KAAK,MAAM,MAAM,OAAO,GAAG,KAAK,EAAE;AAC3E,QAAM,SAAS,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK;AACxC,uBAAqB,uBAAuB,QAAQ,IAAI;AACxD,MAAI,WAAW,qBAAqB,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,IAAI;AACpF,QAAM,UAAU,OAAO,OAAO,CAAC,MAAM,aAAa,cAAc,EAAE,KAAK,IAAI,CAAC;AAC5E,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,UAAU,EAAE,QAAQ,EAAE,GAAG,QAAQ,GAAG,GAAG,SAAS,SAAS,CAAC;AAAA,EACnE;AACA,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO,GAAG,KAAK;AACzD,MAAI,aAAa,cAAc,QAAQ,MAAM,GAAG;AAC9C,WAAO;AAAA,EACT;AACA,MAAI,QAAQ,GAAG,UAAU,UAAU;AACjC,UAAM,WAAW,QAAQ,IAAI,CAAC,MAAM;AAClC,YAAM,YAAY,aAAa,cAAc,EAAE,MAAM,MAAM,IAAI,CAAC;AAChE,YAAM,QAAQ,CAAC,IAAI,SAAS;AAC5B,aAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,MAAM,EAAE,CAAC;AAAA,IAC3E,CAAC;AACD,UAAM,kBAAkB,SAAS,IAAI,CAAC,MAAM;AAC1C,aAAO,EAAE,MAAM,SAAS,SAAS,EAAE,MAAM,GAAG,OAAO,EAAE,MAAM;AAAA,IAC7D,CAAC;AACD,eAAW,qBAAqB,gBAAgB,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAC/E,UAAM,eAAe,SAAS,GAAG,MAAM,OAAO;AAC9C,UAAM,WAAW,WAAW,iBAAiB,UAAU,OAAO,GAAG,OAAO,YAAY;AACpF,UAAM,gBAAgB,qBAAqB,gBAAgB,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,IAAI;AAC5F,QAAI,QAAQ;AACZ,UAAM,UAAU,SAAS,UAAU,IAAI,IAAI,MAAM;AACjD,YAAQ,cAAc,qBAAqB,uBAAuB,QAAQ;AAC1E,aAAS,QAAQ,CAAC,MAAM,SAAS,YAAY,EAAE,MAAM,CAAC;AACtD,WAAO;AAAA,EACT;AACA,QAAM,WAAW,aAAa,cAAc,QAAQ,GAAG,MAAM,MAAM,GAAG,IAAI,CAAC;AAC3E,MAAI,eAAe;AACnB,QAAM,YAAY,QAAQ,IAAI,CAAC,WAAW;AACxC,UAAM,WAAW,aAAa,cAAc,OAAO,MAAM,MAAM,IAAI,CAAC;AACpE,oBAAgB;AAChB,WAAO;AAAA,EACT,CAAC;AACD,QAAM,SAAS,QAAQ,IAAI,CAAC,WAAW,SAAS,mBAAmB,MAAM,CAAC;AAC1E,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,WAAS,IAAI,GAAG,IAAI,UAAU,KAAK;AACjC,QAAI,YAAY,IAAI;AACpB,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,YAAM,WAAW,UAAU;AAC3B,YAAM,WAAW,IAAI;AACrB,YAAM,OAAO,OAAO,GAAG,SAAS,UAAU,WAAW,QAAQ;AAC7D,cAAQ,IAAI,MAAM,SAAS;AAC3B,mBAAa;AAAA,IACf;AAAA,EACF;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,OAAO,UAAU;AACxB,eAAa,SAAS,KAAK,MAAM,QAAQ,MAAM;AAAA,IAC7C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,iBAAiB,WAAW,IAAI;AACvE,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,OAAO,WAAW;AAC5I,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,YAAY,SAAS,QAAQ,SAAS,SAAS,IAAI;AACzD,MAAI,SAAS,eAAe,gBAAgB;AAC1C,UAAM,IAAI,MAAM,oDAAoD,SAAS,yCAAyC;AAAA,EACxH;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,aAAW,KAAK,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,UAAU,cAAc,aAAa,QAAQ,UAAU,WAAW,SAAS,WAAW,gBAAgB,eAAe,cAAc,aAAa,eAAe,gBAAgB,KAAK;AACxO,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,4BAA0B,SAAS,KAAK,MAAM,qBAAqB,MAAM;AAAA,IACvE;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,IAAI,OAAO,IAAI;AACvB,QAAM,EAAE,SAAS,KAAK,MAAM,YAAY,iBAAiB,WAAW,IAAI;AACxE,QAAM,YAAY;AAClB,QAAM,cAAc,qBAAqB,wBAAwB,UAAU;AAC3E,QAAM,WAAW,qBAAqB,kBAAkB,YAAY,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,OAAO,WAAW;AAC/I,QAAM,EAAE,WAAW,cAAc,aAAa,YAAY,UAAU,SAAS,aAAa,WAAW,UAAU,cAAc,YAAY,IAAI;AAC7I,QAAM,SAAS,eAAe,IAAI,SAAS,QAAQ;AACnD,QAAM,UAAU,cAAc,IAAI,SAAS,QAAQ;AACnD,QAAM,iBAAiB,SAAS,eAAe;AAC/C,QAAM,YAAY,aAAa,eAAe,SAAS,OAAO;AAC9D,QAAM,YAAY,aAAa,eAAe,GAAG,KAAK;AACtD,QAAM,CAAC,OAAO,OAAO,KAAK,IAAI,aAAa,eAAe,OAAO,KAAK;AACtE,QAAM,eAAe,UAAU;AAC/B,QAAM,aAAa,iBAAiB,UAAU,KAAK,UAAU;AAC7D,QAAM,aAAa,iBAAiB,UAAU,KAAK;AACnD,QAAM,iBAAiB,iBAAiB,IAAI,UAAU;AACtD,QAAM,eAAe,UAAU;AAC/B,QAAM,aAAa,iBAAiB,UAAU,KAAK,UAAU;AAC7D,QAAM,aAAa,iBAAiB,UAAU,KAAK;AACnD,QAAM,iBAAiB,iBAAiB,IAAI,UAAU;AACtD,QAAM,MAAM,SAAS,WAAW,SAAS,SAAS,SAAS;AAC3D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,OAAO,SAAS,UAAU,IAAI,GAAG,MAAM,EAAE;AAC/C,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,0BAAwB,MAAM,UAAU,WAAW,cAAc,aAAa,UAAU,SAAS,YAAY,WAAW,UAAU,aAAa,cAAc,aAAa,QAAQ,SAAS,OAAO,OAAO,OAAO,cAAc,YAAY,YAAY,gBAAgB,cAAc,YAAY,YAAY,gBAAgB,KAAK;AACjU,SAAO;AACT;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,aAAa,wBAAwB,GAAG;AAG5C,IAAI,cAAc,wBAAwB,IAAI;AAG9C,IAAI;AAAA,CACH,SAAS,sBAAsB;AAC9B,uBAAqB,qBAAqB,cAAc,KAAK;AAC7D,uBAAqB,qBAAqB,aAAa,KAAK;AAC9D,GAAG,wBAAwB,sBAAsB,CAAC,EAAE;AACpD,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,sBAAoB,SAAS,KAAK,MAAM,eAAe,MAAM;AAAA,IAC3D;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,oBAAoB,SAAS,IAAI;AACjD,QAAM,EAAE,OAAO,QAAQ,OAAO,OAAO,IAAI;AACzC,QAAM,WAAW,MAAM,MAAM;AAC7B,QAAM,CAAC,YAAY,SAAS,IAAI;AAChC,QAAM,WAAW,CAAC,UAAU,YAAY,WAAW,OAAO,MAAM,EAAE;AAClE,MAAI,aAAa,SAAS,UAAU,IAAI,OAAO,MAAM;AACrD,MAAI;AACJ,MAAI,OAAO,UAAU,WAAW;AAC9B,iBAAa,MAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,GAAG,OAAO,GAAG,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AAC5F,iBAAa,SAAS,UAAU,IAAI,WAAW,MAAM;AAAA,EACvD;AACA,QAAM,WAAW,WAAW;AAC5B,QAAM,UAAU,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACrD,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,MAAM,SAAS,WAAW,UAAU,SAAS;AACnD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,mBAAmB,IAAI,WAAW,IAAI,WAAW,OAAO,KAAK,EAAE,MAAM;AAC3E,oBAAkB,UAAU,SAAS,UAAU,UAAU,kBAAkB,YAAY,WAAW,oBAAoB,SAAS,oBAAoB,KAAK;AACxJ,MAAI,cAAc,MAAM;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,gBAAc,SAAS,KAAK,MAAM,SAAS,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,QAAM,QAAQ,EAAE,MAAM;AACtB,eAAa,OAAO,EAAE,UAAU,aAAa,EAAE,UAAU,SAAS,MAAM,4BAA4B,EAAE,mCAAmC;AACzI,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,KAAK;AACzE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,MAAM,YAAY,GAAG,SAAS,SAAS,CAAC;AAAA,EAC3F;AACA,QAAM,eAAe,qBAAqB,iBAAiB,GAAG,KAAK,EAAE;AACrE,uBAAqB,2BAA2B,WAAW,CAAC,YAAY,GAAG,KAAK;AAChF,QAAM,cAAc,SAAS,WAAW,UAAU,OAAO,UAAU,KAAK;AACxE,QAAM,WAAW,UAAU,MAAM;AACjC,QAAM,cAAc,SAAS,UAAU,IAAI,UAAU,MAAM,EAAE;AAC7D,QAAM,gBAAgB,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AACjE,cAAY,aAAa,YAAY,IAAI,GAAG,WAAW,IAAI,GAAG,UAAU,eAAe,SAAS,EAAE,MAAM;AACxG,MAAI,MAAM;AACV,MAAI,gBAAgB,MAAM;AACxB,UAAM,kBAAkB,qBAAqB,uBAAuB,WAAW;AAC/E,UAAM,WAAW,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,OAAO,EAAE,MAAM,gBAAgB,GAAG,SAAS,SAAS,CAAC;AACpG,aAAS,YAAY,UAAU,MAAM;AACrC,aAAS,YAAY,YAAY,MAAM;AAAA,EACzC;AACA,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,eAAa,SAAS,KAAK,MAAM,QAAQ,MAAM;AAAA,IAC7C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,MAAM,WAAW,SAAS,SAAS,IAAI;AAC/C,QAAM,QAAQ,EAAE,MAAM;AACtB,eAAa,OAAO,EAAE,UAAU,aAAa,EAAE,UAAU,SAAS,MAAM,2BAA2B,EAAE,mCAAmC;AACxI,QAAM,cAAc,qBAAqB,mBAAmB,CAAC,IAAI,GAAG,KAAK;AACzE,MAAI,YAAY;AAChB,MAAI,gBAAgB,MAAM;AACxB,gBAAY,WAAW,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,MAAM,YAAY,GAAG,SAAS,SAAS,CAAC;AAAA,EAC3F;AACA,QAAM,eAAe,qBAAqB,iBAAiB,GAAG,KAAK,EAAE;AACrE,uBAAqB,2BAA2B,UAAU,CAAC,YAAY,GAAG,KAAK;AAC/E,QAAM,cAAc,SAAS,WAAW,UAAU,OAAO,UAAU,KAAK;AACxE,QAAM,WAAW,UAAU,MAAM;AACjC,QAAM,cAAc,SAAS,UAAU,IAAI,UAAU,MAAM,EAAE;AAC7D,QAAM,gBAAgB,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AACjE,aAAW,aAAa,YAAY,IAAI,GAAG,WAAW,IAAI,GAAG,UAAU,eAAe,SAAS,EAAE,MAAM;AACvG,MAAI,MAAM;AACV,MAAI,gBAAgB,MAAM;AACxB,UAAM,kBAAkB,qBAAqB,uBAAuB,WAAW;AAC/E,UAAM,WAAW,EAAE,QAAQ,EAAE,GAAG,YAAY,GAAG,OAAO,EAAE,MAAM,gBAAgB,GAAG,SAAS,SAAS,CAAC;AACpG,aAAS,YAAY,UAAU,MAAM;AACrC,aAAS,YAAY,YAAY,MAAM;AAAA,EACzC;AACA,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,qBAAmB,SAAS,KAAK,MAAM,cAAc,MAAM;AAAA,IACzD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,WAAW,WAAW,IAAI;AAClC,QAAM,YAAY,EAAE,MAAM;AAC1B,QAAM,cAAc,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AACjE,QAAM,aAAa,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AAChE,QAAM,aAAa,eAAe,SAAS,EAAE,MAAM,KAAK,EAAE,MAAM;AAChE,QAAM,eAAe,cAAc;AACnC,QAAM,cAAc,aAAa;AACjC,QAAM,cAAc,cAAc,YAAY;AAC9C,QAAM,cAAc,eAAe,SAAS,CAAC,WAAW,cAAc,aAAa,WAAW,IAAI,CAAC,WAAW,aAAa,cAAc,WAAW;AACpJ,QAAM,MAAM,SAAS,WAAW,aAAa,SAAS;AACtD,QAAM,QAAQ,SAAS,UAAU,IAAI,EAAE,MAAM;AAC7C,QAAM,MAAM,MAAM;AAClB,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,EAAE,KAAK,CAAC,EAAE,MAAM;AAChG,QAAM,mBAAmB,IAAI,WAAW,IAAI,WAAW,WAAW,EAAE,MAAM;AAC1E,QAAM,kBAAkB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,WAAW,CAAC,EAAE,MAAM;AACtG,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,eAAe,eAAe,SAAS,IAAI;AACjD,mBAAiB,KAAK,WAAW,cAAc,eAAe,EAAE,MAAM,SAAS,GAAG,kBAAkB,iBAAiB,YAAY,QAAQ,KAAK;AAC9I,SAAO;AACT;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,wBAAsB,SAAS,KAAK,MAAM,uBAAuB,MAAM;AAAA,IACrE;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,iBAAiB,MAAM;AAC9B,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,EAAE,SAAS,WAAW,KAAK,MAAM,gBAAgB,IAAI;AAC3D,QAAM,aAAa,aAAa,OAAO,CAAC,GAAG,CAAC,IAAI;AAChD,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,YAAY,MAAM,iBAAiB,IAAI;AAC/H,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,QAAM,YAAY,SAAS,QAAQ,SAAS,SAAS,IAAI;AACzD,MAAI,SAAS,eAAe,gBAAgB;AAC1C,UAAM,IAAI,MAAM,mEAAmE,SAAS,yCAAyC;AAAA,EACvI;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,sBAAoB,KAAK,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,UAAU,cAAc,aAAa,QAAQ,UAAU,WAAW,SAAS,WAAW,gBAAgB,eAAe,cAAc,aAAa,eAAe,gBAAgB,KAAK;AACjP,SAAO;AACT;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,aAAa,wBAAwB,GAAG;AAG5C,IAAI,yBAAyB;AAC7B,IAAI,eAAe,yBAAyB,OAAO,wBAAwB,MAAM;AAGjF,IAAI,aAAa,wBAAwB,KAAK,SAAS;AAGvD,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,IAAI,IAAI;AAChB,QAAM,YAAY,OAAO,MAAM;AAC/B,QAAM,WAAW,OAAO,MAAM,MAAM;AACpC,MAAI,OAAO;AACX,MAAI,MAAM,GAAG;AACX,iBAAa,OAAO,EAAE,YAAY,MAAM,KAAK,MAAM,iCAAiC,EAAE,YAAY,OAAO,YAAY;AACrH,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,WAAS,OAAO,MAAM,GAAG,CAAC;AAC1B,SAAO,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,SAAS,EAAE,CAAC;AAC1F;AACA,IAAI,oBAAoB;AAAA,EACtB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,OAAO,EAAE,OAAO,OAAO,MAAM,GAAG,SAAS,SAAS,IAAI;AAC9D,QAAM,MAAM,SAAS,WAAW,OAAO,KAAK;AAC5C,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,KAAK,KAAK;AAClB,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,sBAAoB,SAAS,KAAK,MAAM,eAAe,MAAM;AAAA,IAC3D;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,MAAM,SAAS,WAAW,OAAO,OAAO,OAAO,KAAK;AAC1D,QAAM,UAAU,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACtD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,oBAAkB,SAAS,OAAO,aAAa,YAAY,aAAa,KAAK;AAC7E,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI,eAAe,wBAAwB,KAAK;AAGhD,IAAI,yBAAyB;AAC7B,IAAI,kBAAkB,yBAAyB,UAAU,sBAAsB;AAG/E,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,kBAAgB,SAAS,KAAK,MAAM,gBAAgB,MAAM,CAAC,UAAU,UAAU,UAAU,UAAU,UAAU,UAAU,QAAQ,CAAC;AAClI;AACA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,gBAAgB,IAAI;AAC5B,QAAM,EAAE,GAAG,MAAM,OAAO,UAAU,QAAQ,OAAOA,QAAO,IAAI;AAC5D,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,SAAS,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACpD,QAAM,aAAa,SAAS,UAAU,IAAI,SAAS,MAAM,EAAE;AAC3D,QAAM,WAAW,UAAU,OAAO,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE,KAAK;AAC7E,QAAM,UAAUA,WAAU,OAAO,SAAS,UAAU,IAAIA,QAAO,MAAM,EAAE,KAAK;AAC5E,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,MAAI,aAAa,cAAc,EAAE,KAAK,MAAM,GAAG;AAC7C,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,gBAAc,KAAK,QAAQ,YAAY,UAAU,SAAS,iBAAiB,KAAK;AAChF,SAAO;AACT;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,oBAAkB,SAAS,KAAK,MAAM,aAAa,MAAM;AAAA,IACvD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,WAAW,YAAY,iBAAiB,YAAY,aAAa,eAAe,IAAI;AAChH,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,eAAe;AACxH,QAAM,kBAAkB,kBAAkB;AAC1C,MAAI,mBAAmB,MAAM;AAC3B,UAAM,IAAI,MAAM,GAAG,+EAA+E;AAAA,EACpG;AACA,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,iBAAiB,SAAS;AAChC,MAAI,SAAS;AACb,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,SAAS,UAAU,IAAI,KAAK,MAAM;AACnD,QAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,YAAM,IAAI,MAAM,sDAAsD,SAAS,MAAM,SAAS;AAAA,IAChG;AACA,QAAI,SAAS,MAAM,OAAO,gBAAgB;AACxC,YAAM,IAAI,MAAM,2BAA2B,SAAS,wDAAwD,iBAAiB;AAAA,IAC/H;AACA,aAAS,SAAS;AAAA,EACpB;AACA,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,YAAY,SAAS,QAAQ,SAAS,SAAS,IAAI;AACzD,QAAM,YAAY,SAAS;AAC3B,QAAM,WAAW,SAAS;AAC1B,QAAM,UAAU,SAAS;AACzB,MAAI,eAAe,QAAQ;AACzB,UAAM,IAAI,MAAM,yDAAyD,iCAAiC;AAAA,EAC5G;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,2BAA2B,0BAA0B,OAAO,IAAI,SAAS,UAAU,IAAI,uBAAuB,MAAM,EAAE;AAC5H,kBAAgB,KAAK,WAAW,UAAU,SAAS,UAAU,cAAc,aAAa,QAAQ,QAAQ,UAAU,WAAW,SAAS,WAAW,gBAAgB,eAAe,cAAc,aAAa,eAAe,gBAAgB,iBAAiB,0BAA0B,kBAAkB,GAAG,KAAK;AAC/S,SAAO;AACT;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,6BAA2B,SAAS,KAAK,MAAM,sBAAsB,MAAM;AAAA,IACzE;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,MAAM,uBAAuB,IAAI;AACpD,QAAM,EAAE,SAAS,KAAK,MAAM,WAAW,YAAY,iBAAiB,YAAY,aAAa,eAAe,IAAI;AAChH,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,OAAO,OAAO,SAAS,WAAW,MAAM,iBAAiB,IAAI;AAC9H,QAAM,kBAAkB,kBAAkB;AAC1C,MAAI,mBAAmB,MAAM;AAC3B,UAAM,IAAI,MAAM,GAAG,wFAAwF;AAAA,EAC7G;AACA,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,iBAAiB,SAAS;AAChC,MAAI,SAAS;AACb,MAAI,QAAQ,MAAM;AAChB,UAAM,WAAW,SAAS,UAAU,IAAI,KAAK,MAAM;AACnD,QAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,YAAM,IAAI,MAAM,+DAA+D,SAAS,MAAM,SAAS;AAAA,IACzG;AACA,QAAI,SAAS,MAAM,OAAO,gBAAgB;AACxC,YAAM,IAAI,MAAM,oCAAoC,SAAS,wDAAwD,iBAAiB;AAAA,IACxI;AACA,aAAS,SAAS;AAAA,EACpB;AACA,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,YAAY,SAAS,QAAQ,SAAS,SAAS,IAAI;AACzD,QAAM,YAAY,SAAS;AAC3B,QAAM,WAAW,SAAS;AAC1B,QAAM,UAAU,SAAS;AACzB,MAAI,eAAe,QAAQ;AACzB,UAAM,IAAI,MAAM,kEAAkE,iCAAiC;AAAA,EACrH;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,2BAA2B,0BAA0B,OAAO,IAAI,SAAS,UAAU,IAAI,uBAAuB,MAAM,EAAE;AAC5H,2BAAyB,KAAK,WAAW,UAAU,SAAS,UAAU,cAAc,aAAa,QAAQ,QAAQ,UAAU,WAAW,SAAS,WAAW,gBAAgB,eAAe,cAAc,aAAa,eAAe,gBAAgB,iBAAiB,0BAA0B,kBAAkB,GAAG,KAAK;AACxT,SAAO;AACT;AACA,IAAI,8BAA8B;AAAA,EAChC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,iBAAe,SAAS,KAAK,MAAM,UAAU,MAAM;AAAA,IACjD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,SAAS,UAAU,OAAO,IAAI;AACtC,QAAM,EAAE,QAAQ,QAAQ,IAAI;AAC5B,QAAM,CAAC,aAAa,WAAW,WAAW,OAAO,IAAI,uBAAuB,mBAAmB,QAAQ,OAAO;AAC9G,QAAM,MAAM,SAAS,WAAW,aAAa,OAAO,KAAK;AACzD,MAAI,cAAc,GAAG;AACnB,WAAO;AAAA,EACT;AACA,QAAM,eAAe,QAAQ;AAC7B,QAAM,YAAY,aAAa,aAAa,SAAS;AACrD,QAAM,QAAQ,SAAS,UAAU,IAAI,OAAO,MAAM;AAClD,QAAM,MAAM,MAAM;AAClB,QAAM,cAAc,SAAS,UAAU,IAAI,QAAQ,MAAM;AACzD,QAAM,YAAY,YAAY;AAC9B,QAAM,eAAe,IAAI,WAAW,IAAI,WAAW,OAAO,EAAE,MAAM;AAClE,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,eAAa,KAAK,SAAS,OAAO,QAAQ,WAAW,WAAW,WAAW,WAAW,cAAc,KAAK;AACzG,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,eAAa,SAAS,KAAK,MAAM,UAAU,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,GAAG,QAAQ,IAAI;AACvB,QAAM,EAAE,MAAM,UAAU,IAAI;AAC5B,QAAM,aAAa,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AAC9D,QAAM,cAAc,SAAS,SAAS,QAAQ,MAAM;AACpD,QAAM,UAAU,EAAE,MAAM;AACxB,WAAS,IAAI,GAAG,IAAI,YAAY,QAAQ,EAAE,GAAG;AAC3C,UAAM,QAAQ,YAAY;AAC1B,iBAAa,OAAO,SAAS,UAAU,KAAK,SAAS,GAAG,MAAM,6BAA6B,uBAAuB,UAAU,IAAI;AAAA,EAClI;AACA,QAAM,YAAY,qBAAqB,aAAa,yBAAyB,GAAG,SAAS,YAAY,SAAS;AAC9G,QAAM,WAAW,SAAS;AAAA,IACxB,QAAQ,EAAE,EAAE;AAAA,IACZ,OAAO;AAAA,MACL,OAAO;AAAA,QACL,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,QACV,UAAU;AAAA,MACZ;AAAA,IACF;AAAA,IACA,SAAS;AAAA,EACX,CAAC;AACD,QAAM,cAAc,aAAa,cAAc,QAAQ,KAAK;AAC5D,QAAM,eAAe,SAAS;AAAA,IAC5B,QAAQ,EAAE,GAAG,QAAQ;AAAA,IACrB,OAAO,EAAE,OAAO,CAAC,UAAU,WAAW,cAAc,UAAU,SAAS,EAAE;AAAA,IACzE,SAAS;AAAA,EACX,CAAC;AACD,QAAM,qBAAqB;AAAA,IACzB,UAAU;AAAA,IACV,UAAU;AAAA,IACV,cAAc,UAAU;AAAA,IACxB,UAAU;AAAA,EACZ;AACA,QAAM,MAAM,SAAS,WAAW,oBAAoB,EAAE,KAAK;AAC3D,MAAI,aAAa,cAAc,EAAE,KAAK,MAAM,GAAG;AAC7C,WAAO;AAAA,EACT;AACA,QAAM,cAAc,SAAS,MAAM,SAAS;AAC5C,QAAM,QAAQ,SAAS,UAAU,IAAI,SAAS,MAAM;AACpD,QAAM,MAAM,MAAM;AAClB,QAAM,cAAc,SAAS,UAAU,IAAI,aAAa,MAAM;AAC9D,QAAM,YAAY,YAAY;AAC9B,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,SAAS,KAAK,CAAC,EAAE,MAAM;AACvG,QAAM,kBAAkB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,kBAAkB,CAAC,EAAE,MAAM;AAC7G,aAAW,KAAK,SAAS,EAAE,QAAQ,eAAe,aAAa,WAAW,UAAU,WAAW,iBAAiB,KAAK;AACrH,WAAS,YAAY,SAAS,MAAM;AACpC,WAAS,YAAY,aAAa,MAAM;AACxC,MAAI,QAAQ,UAAU;AACtB,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,yBAAyB;AAC7B,IAAI,iBAAiB,yBAAyB,SAAS,wBAAwB,MAAM;AAGrF,IAAI,yBAAyB;AAC7B,IAAI,sBAAsB,yBAAyB,cAAc,wBAAwB,MAAM;AAG/F,IAAI;AACJ,SAAS,WAAW,UAAU;AAC5B,cAAY,SAAS,KAAK,MAAM,WAAW,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,MAAM,GAAG,SAAS,SAAS,IAAI;AAC/D,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,SAAS;AAClD,MAAI,aAAa,cAAc,EAAE,KAAK,MAAM,GAAG;AAC7C,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,cAAU,KAAK,SAAS,EAAE,QAAQ,OAAO,KAAK;AAAA,EAChD;AACA,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,yBAAyB;AAC7B,IAAI,cAAc,yBAAyB,MAAM,wBAAwB,MAAM;AAG/E,IAAI,yBAAyB;AAC7B,IAAI,mBAAmB,yBAAyB,WAAW,wBAAwB,MAAM;AAGzF,IAAI,aAAa,wBAAwB,GAAG;AAG5C,IAAI,yBAAyB;AAC7B,IAAI,oBAAoB,yBAAyB,YAAY,wBAAwB,MAAM;AAG3F,IAAI,oBAAoB,wBAAwB,UAAU;AAG1D,IAAI,yBAAyB;AAC7B,IAAI,mBAAmB,yBAAyB,WAAW,wBAAwB,MAAM;AAGzF,IAAI,0BAA0B;AAC9B,IAAI,mBAAmB,yBAAyB,YAAY,yBAAyB,MAAM;AAG3F,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,YAAU,SAAS,KAAK,MAAM,KAAK,MAAM;AAAA,IACvC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,kBAAkB,MAAM,SAAS,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,aAAS;AACT,cAAU;AAAA,EACZ;AACA,QAAM,YAAY,OAAO,MAAM;AAC/B,uBAAqB,2BAA2B,OAAO,MAAM,SAAS;AACtE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,IAAI;AACjG,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAQ,SAAS,SAAS,EAAE,QAAQ,YAAY,KAAK;AAAA,EACvD;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,IAAI,iBAAiB,yBAAyB,SAAS,uBAAuB;AAG9E,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,gBAAc,SAAS,KAAK,MAAM,SAAS,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,IAAI,OAAO;AACjB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,eAAa,OAAO,EAAE,UAAU,WAAW,MAAM,0DAA0D,EAAE,QAAQ;AACrH,QAAM,EAAE,YAAY,SAAS,KAAK,MAAM,gBAAgB,IAAI;AAC5D,QAAM,WAAW,qBAAqB,kBAAkB,EAAE,OAAO,YAAY,SAAS,GAAG,MAAM,eAAe;AAC9G,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,SAAS,SAAS,QAAQ;AAChC,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,YAAY,SAAS,QAAQ;AACnC,QAAM,UAAU,SAAS,QAAQ;AACjC,QAAM,iBAAiB,SAAS;AAChC,QAAM,gBAAgB,SAAS;AAC/B,QAAM,eAAe,SAAS;AAC9B,QAAM,cAAc,SAAS;AAC7B,QAAM,gBAAgB,SAAS;AAC/B,QAAM,iBAAiB,SAAS;AAChC,MAAI,SAAS,eAAe,gBAAgB;AAC1C,UAAM,IAAI,MAAM,6CAA6C,SAAS,yCAAyC;AAAA,EACjH;AACA,QAAM,MAAM,SAAS,WAAW,SAAS,UAAU,SAAS;AAC5D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,cAAY,KAAK,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,EAAE,MAAM,IAAI,cAAc,aAAa,QAAQ,UAAU,WAAW,SAAS,gBAAgB,eAAe,cAAc,aAAa,eAAe,gBAAgB,KAAK;AACpN,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM,CAAC,wBAAwB,CAAC;AACvE;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,gBAAgB;AACpB,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAI,iBAAiB,KAAK;AACxB,eAAS;AACT,gBAAU;AACV,sBAAgB,qBAAqB,iBAAiB,cAAc,QAAQ,OAAO,MAAM,MAAM;AAAA,IACjG;AAAA,EACF;AACA,uBAAqB,2BAA2B,QAAQ,eAAe,OAAO,MAAM,MAAM;AAC1F,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,aAAa;AAC1G,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,MAAI,cAAc;AAClB,MAAI,OAAO,UAAU,WAAW;AAC9B,kBAAc,MAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,GAAG,OAAO,GAAG,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AAC7F,cAAU,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AAAA,EACvD;AACA,QAAM,MAAM,SAAS,WAAW,UAAU,SAAS;AACnD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,aAAS,SAAS,YAAY,KAAK;AAAA,EACrC;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,MAAI,OAAO,UAAU,WAAW;AAC9B,aAAS,YAAY,YAAY,MAAM;AAAA,EACzC;AACA,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,YAAU,SAAS,KAAK,MAAM,KAAK,MAAM;AAAA,IACvC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAI,iBAAiB,KAAK;AACxB,eAAS;AACT,gBAAU;AAAA,IACZ;AAAA,EACF;AACA,QAAM,YAAY,OAAO,MAAM;AAC/B,uBAAqB,2BAA2B,OAAO,MAAM,SAAS;AACtE,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,IAAI;AACjG,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO,KAAK;AACtD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAQ,SAAS,SAAS,EAAE,QAAQ,YAAY,KAAK;AAAA,EACvD;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,IAAI,iBAAiB,yBAAyB,SAAS,uBAAuB;AAG9E,IAAI;AAAA,CACH,SAAS,oBAAoB;AAC5B,qBAAmB,mBAAmB,aAAa,KAAK;AACxD,qBAAmB,mBAAmB,eAAe,KAAK;AAC5D,GAAG,sBAAsB,oBAAoB,CAAC,EAAE;AAChD,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,kBAAgB,SAAS,KAAK,MAAM,WAAW,MAAM;AAAA,IACnD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,UAAU,KAAK,EAAE,IAAI;AACxE,QAAM,WAAW,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,KAAK,GAAG,EAAE;AACnE,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,QAAM,kBAAkB,SAAS,IAAI,CAAC,aAAa,SAAS,EAAE;AAC9D,QAAM,mBAAmB,SAAS,IAAI,CAAC,aAAa,SAAS,EAAE;AAC/D,QAAM,mBAAmB,IAAI,WAAW,IAAI,WAAW,eAAe,EAAE,MAAM;AAC9E,QAAM,oBAAoB,IAAI,WAAW,IAAI,WAAW,gBAAgB,EAAE,MAAM;AAChF,gBAAc,KAAK,aAAa,EAAE,MAAM,QAAQ,SAAS,EAAE,QAAQ,kBAAkB,mBAAmB,kBAAkB,OAAO,KAAK;AACtI,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI,0BAA0B;AAC9B,IAAI,kBAAkB,yBAAyB,UAAU,uBAAuB;AAGhF,IAAI,aAAa,wBAAwB,GAAG;AAG5C,SAAS,kBAAkB,UAAU,WAAW;AAC9C,QAAM,SAAS,IAAI,WAAW,SAAS,KAAK,OAAO,QAAQ,WAAW,CAAC;AACvE,QAAM,mBAAmB,OAAO;AAChC,QAAM,eAAe,OAAO;AAC5B,QAAM,kBAAkB,OAAO;AAC/B,QAAM,gBAAgB,OAAO;AAC7B,WAAS,KAAK,MAAM,SAAS;AAC7B,SAAO,EAAE,kBAAkB,cAAc,iBAAiB,cAAc;AAC1E;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,cAAc,eAAe,eAAe,IAAI;AACxD,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,UAAU,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACrD,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,YAAY,UAAU,SAAS,UAAU,eAAe,cAAc,cAAc;AAC1F,QAAM,EAAE,kBAAkB,cAAc,iBAAiB,cAAc,IAAI,kBAAkB,UAAU,SAAS;AAChH,WAAS,KAAK,MAAM,eAAe;AACnC,WAAS,KAAK,MAAM,aAAa;AACjC,QAAM,wBAAwB,SAAS,WAAW,CAAC,YAAY,GAAG,SAAS,gBAAgB;AAC3F,SAAO;AACT;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX;AACF;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,cAAc,eAAe,gBAAgB,mBAAmB,IAAI;AAC5E,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,UAAU,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACrD,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,YAAY,UAAU,SAAS,UAAU,eAAe,cAAc,gBAAgB,kBAAkB;AAC9G,QAAM,EAAE,kBAAkB,cAAc,iBAAiB,cAAc,IAAI,kBAAkB,UAAU,SAAS;AAChH,WAAS,KAAK,MAAM,eAAe;AACnC,QAAM,wBAAwB,SAAS,WAAW,CAAC,YAAY,GAAG,SAAS,gBAAgB;AAC3F,QAAM,qBAAqB,SAAS,WAAW,CAAC,GAAG,SAAS,aAAa;AACzE,SAAO,CAAC,uBAAuB,kBAAkB;AACnD;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AACA,SAAS,YAAY,MAAM;AACzB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,cAAc,eAAe,gBAAgB,aAAa,IAAI;AACtE,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,UAAU,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACrD,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,YAAY,UAAU,SAAS,UAAU,eAAe,cAAc,gBAAgB,YAAY;AACxG,QAAM,EAAE,kBAAkB,cAAc,iBAAiB,cAAc,IAAI,kBAAkB,UAAU,SAAS;AAChH,WAAS,KAAK,MAAM,aAAa;AACjC,QAAM,wBAAwB,SAAS,WAAW,CAAC,YAAY,GAAG,SAAS,gBAAgB;AAC3F,QAAM,uBAAuB,SAAS,WAAW,CAAC,YAAY,GAAG,WAAW,eAAe;AAC3F,SAAO,CAAC,uBAAuB,oBAAoB;AACrD;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,IAAI,kBAAkB,yBAAyB,UAAU,yBAAyB,MAAM;AAGxF,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,eAAa,SAAS,KAAK,MAAM,QAAQ,MAAM;AAAA,IAC7C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,QAAQ,IAAI;AACpB,QAAM,EAAE,OAAO,OAAO,SAAS,SAAS,IAAI;AAC5C,QAAM,MAAM,SAAS,WAAW,CAAC,GAAG,QAAQ,OAAO,KAAK,GAAG,KAAK;AAChE,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,cAAc,SAAS,UAAU,IAAI,QAAQ,MAAM;AACzD,QAAM,YAAY,YAAY;AAC9B,aAAW,WAAW,OAAO,SAAS,UAAU,KAAK;AACrD,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,UAAU,MAAM;AACvB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,IAAI;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,KAAK,CAAC;AACd,SAAO;AACT;AACA,IAAI,kBAAkB;AAAA,EACpB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,MAAI,OAAO,WAAW,GAAG;AACvB,WAAO,YAAY,EAAE,QAAQ,EAAE,OAAO,OAAO,GAAG,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAAA,EAC9F;AACA,QAAM,QAAQ,OAAO,GAAG;AACxB,QAAM,QAAQ,OAAO,GAAG;AACxB,SAAO,QAAQ,CAAC,MAAM;AACpB,iBAAa,kBAAkB,OAAO,EAAE,OAAO,uDAAuD;AACtG,iBAAa,OAAO,UAAU,EAAE,OAAO,MAAM,uDAAuD;AAAA,EACtG,CAAC;AACD,QAAM,0BAA0B,CAAC;AACjC,QAAM,kBAAkB,OAAO,IAAI,CAAC,MAAM;AACxC,UAAM,YAAY,YAAY,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,KAAK,KAAK,EAAE,CAAC;AAC/F,4BAAwB,KAAK,SAAS;AACtC,WAAO;AAAA,EACT,CAAC;AACD,QAAM,SAAS,QAAQ,EAAE,QAAQ,iBAAiB,SAAS,UAAU,OAAO,EAAE,KAAK,EAAE,CAAC;AACtF,0BAAwB,QAAQ,CAAC,MAAM,SAAS,YAAY,EAAE,MAAM,CAAC;AACrE,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK,MAAM,OAAO,MAAM;AAAA,IAC3C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,UAAU,cAAc,EAAE,IAAI;AACjF,QAAM,WAAW,SAAS,IAAI,CAAC,IAAI,MAAM,GAAG,KAAK,EAAE,MAAM,KAAK,GAAG,EAAE;AACnE,MAAI,aAAa,cAAc,EAAE,KAAK,MAAM,GAAG;AAC7C,WAAO,MAAM;AAAA,MACX,SAAS;AAAA,MACT,OAAO,EAAE,OAAO,UAAU,OAAO,eAAe,OAAO,EAAE,MAAM;AAAA,IACjE,CAAC;AAAA,EACH;AACA,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,QAAM,gBAAgB,SAAS,UAAU,IAAI,IAAI,MAAM;AACvD,QAAM,QAAQ,cAAc;AAC5B,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,QAAM,kBAAkB,SAAS,IAAI,CAAC,aAAa,SAAS,EAAE;AAC9D,QAAM,mBAAmB,SAAS,IAAI,CAAC,aAAa,SAAS,EAAE;AAC/D,QAAM,mBAAmB,IAAI,WAAW,IAAI,WAAW,eAAe,EAAE,MAAM;AAC9E,QAAM,oBAAoB,IAAI,WAAW,IAAI,WAAW,gBAAgB,EAAE,MAAM;AAChF,YAAU,KAAK,aAAa,EAAE,MAAM,QAAQ,SAAS,EAAE,QAAQ,kBAAkB,mBAAmB,eAAe,KAAK;AACxH,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI,0BAA0B;AAC9B,IAAI,aAAa,yBAAyB,KAAK,uBAAuB;AAGtE,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK,MAAM,OAAO,MAAM;AAAA,IAC3C;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,OAAO,MAAM;AACpB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,GAAG,MAAM,IAAI;AACrB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,YAAY,SAAS,UAAU,IAAI,MAAM,MAAM,EAAE;AACvD,MAAI,UAAU;AACd,QAAM,SAAS;AACf,MAAI,cAAc;AAClB,MAAI,OAAO,UAAU,WAAW;AAC9B,kBAAc,MAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AACrF,cAAU,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AAAA,EACvD;AACA,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,SAAS;AAClD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAU,SAAS,WAAW,KAAK;AACnC,MAAI,OAAO,UAAU,WAAW;AAC9B,aAAS,YAAY,YAAY,MAAM;AAAA,EACzC;AACA,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM;AAAA,IACzC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,gBAAgB;AACpB,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAI,iBAAiB,KAAK;AACxB,eAAS;AACT,gBAAU;AACV,sBAAgB,qBAAqB,iBAAiB,cAAc,QAAQ,OAAO,MAAM,MAAM;AAAA,IACjG;AAAA,EACF;AACA,uBAAqB,2BAA2B,QAAQ,eAAe,OAAO,MAAM,MAAM;AAC1F,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,aAAa;AAC1G,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO,KAAK;AACtD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,aAAS,SAAS,YAAY,SAAS,IAAI,QAAQ,KAAK;AAAA,EAC1D;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,SAAS,CAAC,SAAS;AACrB,QAAM,EAAE,SAAS,UAAU,MAAM,IAAI;AACrC,QAAM,EAAE,OAAO,MAAM,MAAM,OAAO,MAAM,IAAI;AAC5C,QAAM,SAAS,UAAU,OAAO,MAAM,OAAO,KAAK;AAClD,QAAM,MAAM,SAAS,WAAW,CAAC,OAAO,MAAM,GAAG,KAAK;AACtD,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,IAAI,MAAM;AAClB,SAAO;AACT;AACA,IAAI,eAAe;AAAA,EACjB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,IAAI,iBAAiB,yBAAyB,SAAS,uBAAuB;AAG9E,IAAI,cAAc,wBAAwB,IAAI;AAG9C,IAAI,eAAe,wBAAwB,KAAK;AAGhD,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,uBAAqB,SAAS,KAAK,MAAM,gBAAgB,MAAM;AAAA,IAC7D;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,CAAC,OAAO,WAAW,UAAU,WAAW,IAAI,OAAO;AACzD,QAAM,WAAW,CAAC,OAAO,WAAW,UAAU,WAAW;AACzD,MAAI,QAAQ,SAAS,UAAU,IAAI,OAAO,MAAM;AAChD,MAAI;AACJ,MAAI,MAAM,UAAU,WAAW;AAC7B,iBAAa,MAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,GAAG,OAAO,GAAG,OAAO,EAAE,OAAO,UAAU,EAAE,CAAC;AAC5F,YAAQ,SAAS,UAAU,IAAI,WAAW,MAAM;AAAA,EAClD;AACA,QAAM,MAAM,MAAM;AAClB,QAAM,MAAM,SAAS,WAAW,UAAU,SAAS;AACnD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,WAAO;AAAA,EACT;AACA,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,qBAAmB,KAAK,OAAO,WAAW,UAAU,aAAa,WAAW,UAAU,eAAe,IAAI,GAAG,mBAAmB,IAAI,GAAG,KAAK;AAC3I,MAAI,cAAc,MAAM;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,SAAO;AACT;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,8BAA4B,SAAS,KAAK,MAAM,uBAAuB,MAAM;AAAA,IAC3E;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,uBAAuB,MAAM;AACpC,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,IAAI;AACnB,QAAM,EAAE,cAAc,kBAAkB,KAAK,IAAI;AACjD,QAAM,CAAC,WAAW,QAAQ,IAAI;AAC9B,QAAM,CAAC,OAAO,WAAW,UAAU,WAAW,IAAI,OAAO;AACzD,QAAM,WAAW,CAAC,OAAO,WAAW,UAAU,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,SAAS;AACnD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,WAAO;AAAA,EACT;AACA,MAAI,QAAQ,SAAS,UAAU,IAAI,OAAO,MAAM;AAChD,MAAI;AACJ,MAAI,MAAM,UAAU,WAAW;AAC7B,iBAAa,MAAM;AAAA,MACjB,SAAS;AAAA,MACT,QAAQ,EAAE,GAAG,OAAO;AAAA,MACpB,OAAO,EAAE,OAAO,UAAU;AAAA,IAC5B,CAAC;AACD,YAAQ,SAAS,UAAU,IAAI,WAAW,MAAM;AAAA,EAClD;AACA,QAAM,MAAM,MAAM;AAClB,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,4BAA0B,KAAK,OAAO,WAAW,UAAU,aAAa,WAAW,UAAU,eAAe,IAAI,GAAG,mBAAmB,IAAI,GAAG,KAAK;AAClJ,MAAI,cAAc,MAAM;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,SAAO;AACT;AACA,IAAI,+BAA+B;AAAA,EACjC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,gBAAc,SAAS,KAAK,MAAM,SAAS,MAAM;AAAA,IAC/C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,OAAO,aAAa,eAAe,MAAM,EAAE,KAAK;AACtD,MAAI,EAAE,MAAM,WAAW,GAAG;AACxB,WAAO,UAAU,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,CAAC;AAAA,EACvD;AACA,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,YAAY,IAAI,WAAW,IAAI,WAAW,IAAI,EAAE,MAAM;AAC5D,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACnE,cAAY,KAAK,WAAW,KAAK,QAAQ,eAAe,EAAE,MAAM,QAAQ,KAAK;AAC7E,QAAM,WAAW,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,OAAO,EAAE,OAAO,EAAE,MAAM,GAAG,SAAS,SAAS,CAAC;AAC9F,WAAS,YAAY,IAAI,MAAM;AAC/B,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,eAAa,SAAS,KAAK,MAAM,kBAAkB,MAAM;AAAA,IACvD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,kBAAkB,MAAM;AAC/B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,SAAS,WAAW,OAAO,IAAI;AACvC,QAAM,MAAM,SAAS,WAAW,OAAO,OAAO,OAAO,KAAK;AAC1D,QAAM,UAAU,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACtD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,QAAM,CAAC,SAAS,OAAO,IAAI,qBAAqB,eAAe,QAAQ,aAAa,UAAU;AAC9F,QAAM,cAAc,cAAc;AAClC,QAAM,mBAAmB;AACzB,QAAM,cAAc,OAAO,cAAc,WAAW,CAAC,WAAW,WAAW,WAAW,cAAc,IAAI,gBAAgB,IAAI,CAAC,GAAG,WAAW,gBAAgB;AAC3J,QAAM,YAAY,IAAI,WAAW,IAAI,WAAW,WAAW,EAAE,MAAM;AACnE,aAAW,SAAS,OAAO,aAAa,YAAY,aAAa,SAAS,SAAS,SAAS,WAAW,YAAY,QAAQ,KAAK;AAChI,SAAO;AACT;AACA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI,eAAe,wBAAwB,KAAK;AAGhD,IAAI,eAAe,wBAAwB,KAAK;AAGhD,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,kBAAgB,SAAS,KAAK,MAAM,WAAW,MAAM;AAAA,IACnD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,SAAS,QAAQ,IAAI;AAC7B,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,MAAM,SAAS,WAAW,OAAO,QAAQ,KAAK;AACpD,MAAI,aAAa,cAAc,KAAK,MAAM,GAAG;AAC3C,WAAO;AAAA,EACT;AACA,QAAM,EAAE,WAAW,YAAY,WAAW,SAAS,WAAW,IAAI,wBAAwB,gBAAgB,SAAS,SAAS,KAAK;AACjI,QAAM,cAAc,SAAS,UAAU,IAAI,QAAQ,MAAM;AACzD,QAAM,YAAY,YAAY;AAC9B,QAAM,cAAc,SAAS,UAAU,IAAI,QAAQ,MAAM;AACzD,QAAM,YAAY,YAAY;AAC9B,QAAM,eAAe,IAAI,WAAW,IAAI,WAAW,OAAO,EAAE,MAAM;AAClE,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,gBAAc,WAAW,WAAW,SAAS,QAAQ,QAAQ,WAAW,YAAY,WAAW,cAAc,YAAY,KAAK;AAC9H,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,eAAa,SAAS,KAAK,MAAM,YAAY,MAAM;AAAA,IACjD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,SAAS,IAAI;AACtC,QAAM,EAAE,WAAW,GAAG,EAAE,IAAI;AAC5B,QAAM,cAAc,SAAS,UAAU,IAAI,UAAU,MAAM,EAAE;AAC7D,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,QAAQ,UAAU,MAAM;AAC9B,QAAM,QAAQ,EAAE,MAAM;AACtB,QAAM,SAAS,UAAU,KAAK,QAAQ,KAAK,UAAU,IAAI,IAAI,aAAa,cAAc,EAAE,MAAM,MAAM,CAAC,CAAC;AACxG,aAAW,aAAa,KAAK,KAAK,QAAQ,KAAK;AAC/C,SAAO;AACT;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,WAAW;AACb;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK,MAAM,SAAS,MAAM,CAAC,UAAU,QAAQ,CAAC;AACrE;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,EAAE,EAAE,IAAI;AAC7C,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,MAAI,aAAa,cAAc,IAAI,KAAK,MAAM,GAAG;AAC/C,WAAO;AAAA,EACT;AACA,YAAU,KAAK,KAAK;AACpB,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,aAAa,wBAAwB,GAAG;AAG5C,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,cAAY,SAAS,KAAK,MAAM,SAAS,MAAM;AAAA,IAC7C;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,SAAS,MAAM;AACtB,QAAM,EAAE,SAAS,UAAU,QAAQ,EAAE,OAAO,GAAG,OAAO,EAAE,IAAI,EAAE,IAAI;AAClE,QAAM,MAAM,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AAClD,QAAM,MAAM,SAAS,WAAW,OAAO,OAAO,OAAO,KAAK;AAC1D,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,WAAW,OAAO,MAAM;AAC9B,QAAM,QAAQ,aAAa,cAAc,OAAO,KAAK,IAAI;AACzD,MAAI,aAAa,cAAc,IAAI,KAAK,MAAM,GAAG;AAC/C,WAAO;AAAA,EACT;AACA,YAAU,KAAK,OAAO,UAAU,KAAK;AACrC,SAAO;AACT;AACA,IAAI,iBAAiB;AAAA,EACnB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,gBAAgB,MAAM;AAC7B,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,YAAY,SAAS,IAAI;AACjC,QAAM,QAAQ,aAAa,cAAc,UAAU;AACnD,QAAM,mBAAmB,CAAC,CAAC,GAAG,CAAC,CAAC;AAChC,mBAAiB,KAAK,GAAG,QAAQ;AACjC,WAAS,IAAI,IAAI,WAAW,QAAQ,IAAI,EAAE,MAAM,QAAQ,EAAE,GAAG;AAC3D,qBAAiB,KAAK,CAAC,GAAG,CAAC,CAAC;AAAA,EAC9B;AACA,QAAM,UAAU,aAAa,WAAW;AAAA,IACtC,QAAQ,EAAE,EAAE;AAAA,IACZ,SAAS;AAAA,IACT,OAAO,EAAE,UAAU,kBAAkB,eAAe,EAAE;AAAA,EACxD,CAAC;AACD,QAAM,sBAAsB,qBAAqB,YAAY,QAAQ,OAAO,YAAY,OAAO,KAAK;AACpG,QAAM,oCAAoC,qBAAqB,YAAY,oBAAoB,QAAQ,WAAW,QAAQ,KAAK;AAC/H,QAAM,eAAe,qBAAqB,oBAAoB,QAAQ,OAAO,YAAY,OAAO,KAAK;AACrG,QAAM,gBAAgB,EAAE,GAAG,QAAQ;AACnC,QAAM,eAAe,EAAE,OAAO,oBAAoB;AAClD,QAAM,kBAAkB,SAAS,EAAE,QAAQ,eAAe,SAAS,UAAU,OAAO,aAAa,CAAC;AAClG,QAAM,kBAAkB,EAAE,GAAG,gBAAgB;AAC7C,QAAM,iBAAiB,EAAE,MAAM,kCAAkC;AACjE,QAAM,WAAW,WAAW,EAAE,QAAQ,iBAAiB,SAAS,UAAU,OAAO,eAAe,CAAC;AACjG,QAAM,sBAAsB,EAAE,GAAG,SAAS;AAC1C,QAAM,qBAAqB,EAAE,OAAO,aAAa;AACjD,QAAM,SAAS,SAAS,EAAE,QAAQ,qBAAqB,SAAS,UAAU,OAAO,mBAAmB,CAAC;AACrG,WAAS,YAAY,QAAQ,MAAM;AACnC,WAAS,YAAY,gBAAgB,MAAM;AAC3C,WAAS,YAAY,SAAS,MAAM;AACpC,SAAO;AACT;AACA,IAAI,wBAAwB;AAAA,EAC1B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,4BAA0B,SAAS,KAAK,MAAM,uBAAuB,UAAU;AAAA,IAC7E;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,qBAAqB,MAAM;AAClC,QAAM,EAAE,SAAS,UAAU,OAAO,IAAI;AACtC,QAAM,EAAE,SAAS,QAAQ,YAAY,aAAa,IAAI;AACtD,QAAM,eAAe,QAAQ,MAAM;AACnC,QAAM,OAAO,QAAQ,MAAM;AAC3B,QAAM,YAAY,SAAS,SAAS,WAAW,MAAM,EAAE;AACvD,QAAM,wBAAwB,CAAC,eAAe,WAAW,IAAI;AAC7D,QAAM,YAAY,SAAS,UAAU,IAAI,QAAQ,MAAM,EAAE;AACzD,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,iBAAiB,SAAS,UAAU,IAAI,aAAa,MAAM,EAAE;AACnE,QAAM,gBAAgB,SAAS,WAAW,uBAAuB,QAAQ,KAAK;AAC9E,QAAM,kBAAkB,SAAS,UAAU,IAAI,cAAc,MAAM,EAAE;AACrE,QAAM,eAAe,SAAS,WAAW,sBAAsB,MAAM,GAAG,CAAC,GAAG,OAAO,KAAK;AACxF,QAAM,iBAAiB,SAAS,UAAU,IAAI,aAAa,MAAM,EAAE;AACnE,QAAM,oBAAoB,SAAS,WAAW,CAAC,SAAS,GAAG,MAAM;AACjE,QAAM,sBAAsB,SAAS,UAAU,IAAI,kBAAkB,MAAM,EAAE;AAC7E,QAAM,kBAAkB,SAAS,WAAW,CAAC,YAAY,GAAG,QAAQ,KAAK;AACzE,QAAM,oBAAoB,SAAS,UAAU,IAAI,gBAAgB,MAAM,EAAE;AACzE,QAAM,kBAAkB,SAAS,WAAW,CAAC,CAAC,GAAG,OAAO;AACxD,QAAM,oBAAoB,SAAS,UAAU,IAAI,gBAAgB,MAAM,EAAE;AACzE,QAAM,aAAa,wBAAwB,WAAW,UAAU,SAAS,OAAO,QAAQ,cAAc,WAAW,MAAM,gBAAgB,iBAAiB,gBAAgB,qBAAqB,mBAAmB,iBAAiB;AACjO,QAAM,uBAAuB,SAAS,SAAS,gBAAgB,MAAM;AACrE,MAAI;AACJ,UAAQ,qBAAqB,IAAI;AAAA,IAC/B,KAAK,GAAG;AACN,yBAAmB,qBAAqB,gDAAgD,qBAAqB,EAAE;AAC/G;AAAA,IACF;AAAA,IACA,KAAK,GAAG;AACN,yBAAmB,qBAAqB,gDAAgD,qBAAqB,IAAI,qBAAqB,EAAE;AACxI;AAAA,IACF;AAAA,IACA,KAAK;AACH,yBAAmB,qBAAqB,kDAAkD,qBAAqB,IAAI,qBAAqB,IAAI,qBAAqB,EAAE;AACnK;AAAA,IACF;AACE,yBAAmB;AAAA,EACvB;AACA,WAAS,YAAY,gBAAgB,MAAM;AAC3C,MAAI,kBAAkB;AACpB,aAAS,YAAY,cAAc,MAAM;AACzC,aAAS,YAAY,aAAa,MAAM;AACxC,aAAS,YAAY,kBAAkB,MAAM;AAC7C,aAAS,YAAY,gBAAgB,MAAM;AAC3C,UAAM,IAAI,MAAM,gBAAgB;AAAA,EAClC;AACA,MAAI,iBAAiB;AACrB,MAAI,gBAAgB;AACpB,MAAI,eAAe,sBAAsB,IAAI;AAC3C,qBAAiB,OAAO;AAAA,MACtB,QAAQ,EAAE,GAAG,cAAc;AAAA,MAC3B,OAAO,EAAE,OAAO,GAAG,MAAM,CAAC,YAAY,IAAI,EAAE;AAAA,MAC5C,SAAS;AAAA,IACX,CAAC;AACD,oBAAgB,OAAO;AAAA,MACrB,QAAQ,EAAE,GAAG,aAAa;AAAA,MAC1B,OAAO,EAAE,OAAO,GAAG,MAAM,WAAW;AAAA,MACpC,SAAS;AAAA,IACX,CAAC;AACD,aAAS,YAAY,cAAc,MAAM;AACzC,aAAS,YAAY,aAAa,MAAM;AAAA,EAC1C;AACA,SAAO,CAAC,gBAAgB,eAAe,mBAAmB,eAAe;AAC3E;AACA,IAAI,6BAA6B;AAAA,EAC/B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,sBAAoB,SAAS,KAAK,MAAM,eAAe,MAAM;AAAA,IAC3D;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,eAAe,MAAM;AAC5B,QAAM,EAAE,SAAS,UAAU,OAAO,IAAI;AACtC,QAAM,EAAE,cAAc,YAAY,SAAS,IAAI;AAC/C,MAAI,aAAa,MAAM,WAAW,GAAG;AACnC,UAAM,IAAI,MAAM;AAAA,UACV,aAAa,OAAO;AAAA,EAC5B;AACA,MAAI,WAAW,MAAM,WAAW,GAAG;AACjC,UAAM,IAAI,MAAM;AAAA,UACV,WAAW,OAAO;AAAA,EAC1B;AACA,MAAI,SAAS,MAAM,WAAW,GAAG;AAC/B,UAAM,IAAI,MAAM,sDAAsD,SAAS,OAAO;AAAA,EACxF;AACA,QAAM,iBAAiB,SAAS,UAAU,IAAI,aAAa,MAAM,EAAE;AACnE,QAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAM,aAAa,SAAS,UAAU,IAAI,SAAS,MAAM,EAAE;AAC3D,QAAM,MAAM,aAAa,MAAM;AAC/B,QAAM,aAAa,aAAa,cAAc,SAAS,KAAK;AAC5D,QAAM,aAAa,SAAS,WAAW,CAAC,KAAK,UAAU,GAAG,aAAa,KAAK;AAC5E,QAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAM,cAAc,SAAS,WAAW,CAAC,UAAU,GAAG,SAAS,KAAK;AACpE,QAAM,gBAAgB,SAAS,UAAU,IAAI,YAAY,MAAM,EAAE;AACjE,QAAM,kBAAkB,SAAS,WAAW,CAAC,CAAC,GAAG,OAAO;AACxD,QAAM,oBAAoB,SAAS,UAAU,IAAI,gBAAgB,MAAM,EAAE;AACzE,oBAAkB,gBAAgB,cAAc,YAAY,KAAK,cAAc,eAAe,iBAAiB;AAC/G,QAAM,uBAAuB,SAAS,SAAS,gBAAgB,MAAM;AACrE,MAAI;AACJ,UAAQ,qBAAqB,IAAI;AAAA,IAC/B,KAAK,GAAG;AACN,yBAAmB,qBAAqB,yDAAyD,qBAAqB,IAAI,qBAAqB,EAAE;AACjJ;AAAA,IACF;AAAA,IACA,KAAK,GAAG;AACN,yBAAmB,qBAAqB,8CAA8C,qBAAqB,IAAI,qBAAqB,EAAE;AACtI;AAAA,IACF;AAAA,IACA,KAAK;AACH,yBAAmB,qBAAqB,qDAAqD;AAC7F;AAAA,IACF,KAAK,GAAG;AACN,YAAM,mBAAmB,MAAM,KAAK,SAAS,SAAS,WAAW,MAAM,CAAC,GAAG,oBAAoB,MAAM,KAAK,SAAS,SAAS,YAAY,MAAM,CAAC;AAC/I,yBAAmB,qBAAqB,gDAAgD,kBAAkB,iBAAiB;AAC3H;AAAA,IACF;AAAA,IACA,KAAK,GAAG;AACN,YAAM,mBAAmB,MAAM,KAAK,SAAS,SAAS,WAAW,MAAM,CAAC,GAAG,oBAAoB,MAAM,KAAK,SAAS,SAAS,YAAY,MAAM,CAAC;AAC/I,yBAAmB,qBAAqB,gDAAgD,kBAAkB,iBAAiB;AAC3H;AAAA,IACF;AAAA,IACA;AACE,yBAAmB;AAAA,EACvB;AACA,WAAS,YAAY,gBAAgB,MAAM;AAC3C,MAAI,kBAAkB;AACpB,aAAS,YAAY,WAAW,MAAM;AACtC,aAAS,YAAY,YAAY,MAAM;AACvC,UAAM,IAAI,MAAM,gBAAgB;AAAA,EAClC;AACA,SAAO,CAAC,YAAY,WAAW;AACjC;AACA,IAAI,uBAAuB;AAAA,EACzB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,+BAA6B,SAAS,KAAK,MAAM,0BAA0B,MAAM;AAAA,IAC/E;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,uBAAuB,MAAM,QAAQ;AAC5C,QAAM,EAAE,SAAS,UAAU,OAAO,IAAI;AACtC,QAAM,EAAE,MAAM,SAAS,WAAW,IAAI;AACtC,QAAM,aAAa,QAAQ,MAAM;AACjC,QAAM,iBAAiB,SAAS,SAAS,WAAW,QAAQ,aAAa,GAAG,UAAU,EAAE;AACxF,QAAM,uBAAuB,aAAa,IAAI,iBAAiB,IAAI;AACnE,QAAM,aAAa;AACnB,MAAI,aAAa,GAAG;AAClB,UAAM,IAAI,MAAM,qBAAqB,wDAAwD,CAAC;AAAA,EAChG;AACA,QAAM,cAAc,KAAK,MAAM,MAAM;AACrC,cAAY,KAAK;AACjB,QAAM,SAAS,SAAS,UAAU,IAAI,KAAK,MAAM,EAAE;AACnD,QAAM,YAAY,SAAS,UAAU,IAAI,QAAQ,MAAM,EAAE;AACzD,QAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAM,SAAS,SAAS,WAAW,aAAa,KAAK,KAAK;AAC1D,QAAM,WAAW,SAAS,UAAU,IAAI,OAAO,MAAM,EAAE;AACvD,QAAM,kBAAkB,SAAS,WAAW,CAAC,CAAC,GAAG,OAAO;AACxD,QAAM,oBAAoB,SAAS,UAAU,IAAI,gBAAgB,MAAM,EAAE;AACzE,6BAA2B,QAAQ,SAAS,KAAK,QAAQ,KAAK,MAAM,IAAI,WAAW,cAAc,UAAU,mBAAmB,QAAQ,CAAC;AACvI,QAAM,uBAAuB,SAAS,SAAS,gBAAgB,MAAM;AACrE,MAAI;AACJ,UAAQ,qBAAqB,IAAI;AAAA,IAC/B,KAAK,GAAG;AACN,yBAAmB,qBAAqB,wDAAwD;AAChG;AAAA,IACF;AAAA,IACA,KAAK,GAAG;AACN,yBAAmB,qBAAqB,6DAA6D;AACrG;AAAA,IACF;AAAA,IACA,KAAK;AACH,yBAAmB,qBAAqB,yDAAyD,qBAAqB,IAAI,qBAAqB,EAAE;AACjJ;AAAA,IACF,KAAK;AACH,yBAAmB,qBAAqB,uDAAuD,qBAAqB,IAAI,qBAAqB,IAAI,qBAAqB,EAAE;AACxK;AAAA,IACF;AACE,yBAAmB;AAAA,EACvB;AACA,WAAS,YAAY,gBAAgB,MAAM;AAC3C,MAAI,kBAAkB;AACpB,aAAS,YAAY,OAAO,MAAM;AAClC,UAAM,IAAI,MAAM,gBAAgB;AAAA,EAClC;AACA,SAAO;AACT;AAGA,SAAS,mBAAmB,MAAM;AAChC,SAAO,uBAAuB,MAAM,IAAI;AAC1C;AACA,IAAI,2BAA2B;AAAA,EAC7B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,kBAAkB,MAAM;AAC/B,SAAO,uBAAuB,MAAM,KAAK;AAC3C;AACA,IAAI,0BAA0B;AAAA,EAC5B,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,OAAO,SAAS,SAAS,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,iBAAiB,KAAK,IAAI;AAClC,QAAM,QAAQ,aAAa,eAAe,MAAM,EAAE,KAAK,EAAE;AACzD,QAAM,aAAa,qBAAqB,iBAAiB,GAAG,iBAAiB,KAAK;AAClF,QAAM,QAAQ,IAAI,MAAM,EAAE,MAAM,MAAM,EAAE,KAAK,CAAC;AAC9C,QAAM,OAAO,EAAE,MAAM,MAAM;AAC3B,SAAO,WAAW,IAAI,CAAC,MAAM;AAC3B,UAAM,aAAa,CAAC,GAAG,IAAI;AAC3B,eAAW,SAAS;AACpB,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,OAAO,EAAE,OAAO,MAAM,WAAW,GAAG,SAAS,SAAS,CAAC;AAC9F,UAAM,UAAU;AAChB,WAAO;AAAA,EACT,CAAC;AACH;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,cAAc,wBAAwB,IAAI;AAG9C,IAAI,gBAAgB,wBAAwB,MAAM;AAGlD,IAAI,0BAA0B;AAC9B,IAAI,2BAA2B,yBAAyB,mBAAmB,uBAAuB;AAGlG,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM;AAAA,IACzC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,IAAI;AAClB,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,WAAS,KAAK,OAAO,SAAS,EAAE,QAAQ,KAAK;AAC7C,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,qBAAmB,SAAS,KAAK,MAAM,cAAc,MAAM;AAAA,IACzD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,eAAe,IAAI;AAC/F,QAAM,EAAE,kBAAkB,YAAY,YAAY,WAAW,eAAe,OAAO,QAAQ,KAAK,MAAM,SAAS,SAAS,IAAI,mBAAmB,UAAU,EAAE,OAAO,OAAO,KAAK,SAAS,WAAW,SAAS,cAAc,aAAa,cAAc;AACpP,MAAI;AACJ,MAAI,YAAY;AACd,aAAS,SAAS,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAAA,EACtF,WAAW,aAAa,eAAe;AACrC,iBAAa,OAAO,EAAE,MAAM,UAAU,GAAG,MAAM,yCAAyC,EAAE,MAAM,QAAQ;AACxG,UAAM,OAAO,mBAAmB,gBAAgB,QAAQ,MAAM,QAAQ;AACtE,UAAM,SAAS,OAAO,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,QAAQ,KAAK,EAAE,CAAC;AAC1F,aAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,OAAO,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AAC5F,aAAS,YAAY,OAAO,MAAM;AAAA,EACpC,OAAO;AACL,UAAM,MAAM,SAAS,WAAW,kBAAkB,SAAS;AAC3D,UAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,UAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,EAAE,KAAK,CAAC,EAAE,MAAM;AAChG,UAAM,aAAa,IAAI,WAAW,IAAI,WAAW,MAAM,EAAE,MAAM;AAC/D,UAAM,WAAW,IAAI,WAAW,IAAI,WAAW,IAAI,EAAE,MAAM;AAC3D,UAAM,eAAe,IAAI,WAAW,IAAI,WAAW,QAAQ,EAAE,MAAM;AACnE,UAAM,mBAAmB,IAAI,WAAW,IAAI,WAAW,gBAAgB,EAAE,MAAM;AAC/E,UAAM,kBAAkB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,gBAAgB,CAAC,EAAE,MAAM;AAC3G,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,qBAAiB,KAAK,eAAe,EAAE,MAAM,QAAQ,YAAY,UAAU,cAAc,kBAAkB,iBAAiB,iBAAiB,QAAQ,KAAK;AAC1J,aAAS,SAAS,EAAE,QAAQ,EAAE,GAAG,IAAI,GAAG,SAAS,UAAU,OAAO,EAAE,OAAO,WAAW,EAAE,CAAC;AACzF,aAAS,YAAY,IAAI,MAAM;AAAA,EACjC;AACA,SAAO;AACT;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,cAAc,MAAM;AAC3B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,WAAW,IAAI;AAC7B,QAAM,EAAE,WAAW,aAAa,SAAS,UAAU,WAAW,UAAU,uBAAuB,IAAI;AACnG,QAAM,QAAQ,SAAS,SAAS,KAAK,MAAM;AAC3C,QAAM,cAAc,SAAS,SAAS,WAAW,MAAM;AACvD,QAAM,CAAC,QAAQ,YAAY,IAAI,iBAAiB,OAAO,aAAa,WAAW,aAAa,SAAS,WAAW,UAAU,sBAAsB;AAChJ,QAAM,YAAY,SAAS,WAAW,CAAC,OAAO,MAAM,GAAG,QAAQ;AAC/D,QAAM,gBAAgB,SAAS,UAAU,IAAI,UAAU,MAAM;AAC7D,gBAAc,cAAc;AAC5B,QAAM,kBAAkB,SAAS,WAAW,WAAW,OAAO,OAAO;AACrE,QAAM,sBAAsB,SAAS,mBAAmB,eAAe;AACvE,sBAAoB,IAAI,YAAY;AACpC,SAAO,CAAC,WAAW,eAAe;AACpC;AACA,IAAI,sBAAsB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,aAAa,MAAM;AAC1B,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,UAAU,IAAI;AACrC,QAAM,EAAE,UAAU,IAAI;AACtB,QAAM,YAAY,SAAS,SAAS,OAAO,MAAM;AACjD,QAAM,gBAAgB,SAAS,SAAS,UAAU,MAAM;AACxD,QAAM,CAAC,SAAS,QAAQ,KAAK,IAAI,gBAAgB,WAAW,cAAc,IAAI,SAAS;AACvF,QAAM,aAAa,OAAO;AAC1B,QAAM,aAAa,SAAS,WAAW,CAAC,YAAY,CAAC,GAAG,OAAO;AAC/D,QAAM,iBAAiB,SAAS,mBAAmB,UAAU;AAC7D,iBAAe,IAAI,OAAO;AAC1B,QAAM,YAAY,SAAS,WAAW,CAAC,UAAU,GAAG,QAAQ;AAC5D,QAAM,gBAAgB,SAAS,UAAU,IAAI,UAAU,MAAM;AAC7D,gBAAc,cAAc;AAC5B,QAAM,WAAW,SAAS,WAAW,CAAC,CAAC,GAAG,OAAO;AACjD,QAAM,eAAe,SAAS,mBAAmB,QAAQ;AACzD,eAAa,IAAI,KAAK;AACtB,SAAO,CAAC,YAAY,WAAW,QAAQ;AACzC;AACA,IAAI,qBAAqB;AAAA,EACvB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,wBAAwB,MAAM;AACrC,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,QAAM,EAAE,WAAW,IAAI;AACvB,QAAM,YAAY,SAAS,SAAS,OAAO,MAAM;AACjD,QAAM,SAAS,2BAA2B,WAAW,UAAU;AAC/D,QAAM,MAAM,SAAS,WAAW,OAAO,OAAO,OAAO;AACrD,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,IAAI,MAAM;AAClB,SAAO;AACT;AACA,IAAI,gCAAgC;AAAA,EAClC,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,0BAA0B;AAC9B,IAAI,aAAa,yBAAyB,KAAK,uBAAuB;AAGtE,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,YAAU,SAAS,KAAK,MAAM,KAAK,MAAM;AAAA,IACvC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,KAAK,MAAM;AAClB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,SAAS,IAAI;AAC3B,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,MAAI,UAAU;AACd,MAAI,SAAS;AACb,QAAM,EAAE,YAAY,MAAM,cAAc,mBAAmB,IAAI,wBAAwB,GAAG,MAAM,QAAQ;AACxG,MAAI,gBAAgB;AACpB,MAAI,oBAAoB;AACtB,UAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,QAAI,iBAAiB,KAAK;AACxB,eAAS;AACT,gBAAU;AACV,sBAAgB,qBAAqB,iBAAiB,cAAc,QAAQ,OAAO,MAAM,MAAM;AAAA,IACjG;AAAA,EACF;AACA,uBAAqB,2BAA2B,OAAO,eAAe,OAAO,MAAM,MAAM;AACzF,QAAM,CAAC,UAAU,WAAW,IAAI,qBAAqB,0BAA0B,OAAO,OAAO,aAAa;AAC1G,QAAM,aAAa,aAAa,cAAc,WAAW;AACzD,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO,KAAK;AACtD,MAAI,aAAa,cAAc,OAAO,KAAK,MAAM,GAAG;AAClD,UAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,YAAQ,SAAS,YAAY,SAAS,IAAI,QAAQ,KAAK;AAAA,EACzD;AACA,MAAI,oBAAoB;AACtB,aAAS,YAAY,WAAW,MAAM;AAAA,EACxC;AACA,MAAI,UAAU;AACZ,UAAM,WAAW,qBAAqB,qBAAqB,IAAI,OAAO,YAAY;AAClF,QAAI,QAAQ;AAAA,EACd;AACA,SAAO;AACT;AACA,IAAI,aAAa;AAAA,EACf,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI,aAAa,wBAAwB,GAAG;AAG5C,IAAI,cAAc,wBAAwB,IAAI;AAG9C,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM;AAAA,IACzC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,MAAM,MAAM;AACnB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,EAAE,KAAK,IAAI;AACjB,QAAM,WAAW,IAAI,MAAM,EAAE,MAAM,MAAM;AACzC,WAAS,IAAI,GAAG,IAAI,SAAS,QAAQ,KAAK;AACxC,aAAS,KAAK,EAAE,MAAM,KAAK,KAAK;AAAA,EAClC;AACA,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,QAAQ,EAAE,MAAM;AACpE,QAAM,MAAM,SAAS,WAAW,UAAU,EAAE,KAAK;AACjD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,WAAS,KAAK,aAAa,EAAE,MAAM,QAAQ,eAAe,SAAS,QAAQ,SAAS,IAAI,QAAQ,KAAK;AACrG,SAAO;AACT;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,aAAW,SAAS,KAAK,MAAM,MAAM,MAAM;AAAA,IACzC;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,IAAI,QAAQ,CAAC,EAAE,QAAQ,SAAS,UAAU,MAAM,MAAM;AACpD,QAAM,EAAE,EAAE,IAAI;AACd,QAAM,EAAE,GAAG,OAAO,IAAI;AACtB,QAAM,MAAM,SAAS,UAAU,IAAI,EAAE,MAAM,EAAE;AAC7C,QAAM,cAAc,IAAI,WAAW,IAAI,WAAW,EAAE,KAAK,EAAE,MAAM;AACjE,QAAM,cAAc,EAAE,MAAM,MAAM;AAClC,cAAY,YAAY,SAAS,KAAK;AACtC,QAAM,YAAY,SAAS,WAAW,aAAa,EAAE,KAAK;AAC1D,QAAM,cAAc,SAAS,UAAU,IAAI,UAAU,MAAM,EAAE;AAC7D,QAAM,aAAa,SAAS,WAAW,aAAa,OAAO;AAC3D,QAAM,eAAe,SAAS,UAAU,IAAI,WAAW,MAAM,EAAE;AAC/D,WAAS,KAAK,aAAa,EAAE,MAAM,QAAQ,SAAS,EAAE,QAAQ,GAAG,QAAQ,aAAa,YAAY;AAClG,SAAO,CAAC,WAAW,UAAU;AAC/B;AACA,IAAI,cAAc;AAAA,EAChB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,IAAI;AACJ,SAAS,QAAQ,UAAU;AACzB,kBAAgB,SAAS,KAAK,MAAM,WAAW,MAAM;AAAA,IACnD;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AACA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,SAAS,UAAU,QAAQ,MAAM,IAAI;AAC7C,QAAM,EAAE,OAAO,QAAQ,WAAW,IAAI;AACtC,QAAM,EAAE,eAAe,UAAU,WAAW,YAAY,IAAI;AAC5D,QAAM,CAAC,OAAO,aAAa,YAAY,WAAW,IAAI,OAAO;AAC7D,QAAM,CAAC,WAAW,QAAQ,IAAI,eAAe,OAAO,cAAc,CAAC,aAAa,UAAU;AAC1F,QAAM,WAAW;AAAA,IACf;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACA,QAAM,eAAe,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,OAAO,KAAK,CAAC,EAAE,MAAM;AACpG,QAAM,gBAAgB,IAAI,WAAW,IAAI,WAAW,aAAa,eAAe,QAAQ,CAAC,EAAE,MAAM;AACjG,QAAM,MAAM,SAAS,WAAW,UAAU,OAAO,KAAK;AACtD,QAAM,QAAQ,SAAS,UAAU,IAAI,IAAI,MAAM,EAAE;AACjD,QAAM,YAAY,SAAS,UAAU,IAAI,OAAO,MAAM;AACtD,QAAM,UAAU,UAAU;AAC1B,QAAM,iBAAiB,SAAS,UAAU,IAAI,WAAW,MAAM;AAC/D,QAAM,eAAe,eAAe;AACpC,QAAM,sBAAsB,kBAAkB,YAAY,IAAI;AAC9D,MAAI;AACJ,UAAQ,UAAU;AAAA,IAChB,KAAK;AACH,mBAAa;AACb;AAAA,IACF,KAAK;AACH,mBAAa;AACb;AAAA,IACF,KAAK;AACH,mBAAa;AACb;AAAA,IACF,KAAK;AACH,mBAAa;AACb;AAAA,IACF;AACE,mBAAa;AACb;AAAA,EACJ;AACA,gBAAc,SAAS,cAAc,WAAW,MAAM,KAAK,GAAG,OAAO,WAAW,UAAU,aAAa,YAAY,aAAa,cAAc,OAAO,MAAM,SAAS,GAAG,eAAe,SAAS,SAAS,GAAG,qBAAqB,YAAY,WAAW,KAAK;AAC5P,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,WAAW;AAAA,EACX,YAAY;AACd;AAGA,SAAS,QAAQ,MAAM;AACrB,QAAM,EAAE,QAAQ,SAAS,UAAU,MAAM,IAAI;AAC7C,QAAM,EAAE,MAAM,IAAI;AAClB,MAAI,EAAE,KAAK,IAAI;AACf,MAAI,OAAO,GAAG;AACZ,YAAQ,MAAM,MAAM;AAAA,EACtB;AACA,QAAM,aAAa,MAAM,MAAM;AAC/B,QAAM,OAAO,MAAM,MAAM;AACzB,QAAM,WAAW,IAAI,MAAM,OAAO,CAAC;AACnC,MAAI,WAAW;AACf,WAAS,IAAI,GAAG,IAAI,MAAM,KAAK;AAC7B,QAAI,MAAM,MAAM;AACd,eAAS,cAAc,MAAM,MAAM;AAAA,IACrC;AAAA,EACF;AACA,QAAM,OAAO,IAAI,MAAM,UAAU;AACjC,QAAM,QAAQ,IAAI,MAAM,IAAI,EAAE,KAAK,CAAC;AACpC,QAAM,OAAO,MAAM,MAAM,MAAM;AAC/B,OAAK,QAAQ;AACb,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,QAAQ;AACd,SAAK,KAAK,OAAO,EAAE,QAAQ,EAAE,GAAG,MAAM,GAAG,OAAO,EAAE,OAAO,KAAK,GAAG,SAAS,SAAS,CAAC;AAAA,EACtF;AACA,SAAO,KAAK,IAAI,CAAC,EAAE,QAAQ,MAAM,OAAO,EAAE,QAAQ,OAAO,OAAO,SAAS,EAAE;AAC7E;AACA,IAAI,gBAAgB;AAAA,EAClB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,SAAS,WAAW,MAAM;AACxB,QAAM,EAAE,QAAQ,EAAE,EAAE,GAAG,SAAS,SAAS,IAAI;AAC7C,QAAM,MAAM,SAAS,WAAW,EAAE,OAAO,EAAE,KAAK;AAChD,QAAM,UAAU,SAAS,mBAAmB,GAAG;AAC/C,UAAQ,KAAK,CAAC;AACd,SAAO;AACT;AACA,IAAI,mBAAmB;AAAA,EACrB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,YAAY;AACd;AAGA,IAAI,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AACA,WAAW,gBAAgB,gBAAgB;AACzC,iBAAe,YAAY;AAC7B;AAGA,IAAI,OAAO,IAAI;AACf,KAAK,aAAa,yBAAyB,YAAY;AACrD,MAAI;AACF,WAAO,YAAY,SAAS,IAAI,WAAW;AAAA,MACzC;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC,CAAC;AAAA,EACJ,SAAS,GAAP;AACA,WAAO;AAAA,EACT;AACF,CAAC;AACD,KAAK,aAAa,gCAAgC,YAAY;AAC5D,MAAI,KAAK,IAAI,SAAS,GAAG;AACvB,WAAO;AAAA,EACT;AACA,MAAI;AACF,QAAI,eAAe,EAAE,MAAM,YAAY,IAAI,kBAAkB,CAAC,CAAC;AAC/D,WAAO,YAAY,SAAS,IAAI,WAAW;AAAA,MACzC;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC,CAAC;AAAA,EACJ,SAAS,GAAP;AACA,WAAO;AAAA,EACT;AACF,CAAC;AAGD,IAAI,iCAAiC,QAAQ,wCAAwC,CAAC;AACtF,IAAI,gDAAgD,QAAQ,+CAA+C,CAAC;AAC5G,IAAI,qBAAqB,QAAQ,0BAA0B,CAAC;AAC5D,IAAI,0BAA0B,+BAA+B,WAAW;AACxE,IAAI,cAAc,mBAAmB,WAAW;AAChD,IAAI,cAAc,cAAc,cAAc;AAAA,EAC5C,YAAY,MAAM;AAChB,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,mBAAmB;AACxB,SAAK,KAAK,KAAK,qBAAqB,YAAY;AAChD,yBAAqB,KAAK,KAAK,KAAK,gBAAgB;AACpD,SAAK,YAAY,IAAI,YAAY,MAAM,OAAO,CAAC;AAAA,EACjD;AAAA,EACA,MAAM,QAAQ,OAAO,OAAO;AAC1B,UAAM,SAAS,EAAE,IAAI,KAAK,mBAAmB;AAC7C,SAAK,KAAK,QAAQ,QAAQ,OAAO,OAAO,CAAC;AACzC,WAAO;AAAA,EACT;AAAA,EACA,aAAa;AACX,WAAO,KAAK,UAAU,WAAW;AAAA,EACnC;AAAA,EACA,MAAM,KAAK,GAAG;AACZ,UAAM,QAAQ,aAAa,IAAI;AAC/B,MAAE;AACF,UAAM,WAAW,aAAa,IAAI,IAAI;AACtC,WAAO,EAAE,SAAS;AAAA,EACpB;AAAA,EACA,KAAK,QAAQ,QAAQ,OAAO,OAAO,UAAU;AAC3C,UAAM,KAAK,KAAK;AAChB,QAAI,UAAU,UAAU;AACtB,YAAM,cAAc;AACpB,WAAK,UAAU,IAAI,QAAQ,EAAE,IAAI,aAAa,OAAO,OAAO,cAAc,MAAM,SAAS,CAAC;AAC1F;AAAA,IACF;AACA,UAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,UAAM,WAAW,OAAO,aAAa,gBAAgB,KAAK;AAC1D,UAAM,eAAe,KAAK,KAAK,QAAQ,QAAQ;AAC/C,SAAK,UAAU,IAAI,QAAQ,EAAE,IAAI,cAAc,OAAO,OAAO,SAAS,CAAC;AACvE,SAAK,KAAK,KAAK,eAAe,IAAI,MAAM,YAAY;AACpD,QAAI,UAAU,MAAM;AAClB,WAAK,KAAK,OAAO,IAAI,IAAI,WAAW,OAAO,QAAQ,OAAO,YAAY,QAAQ,GAAG,YAAY;AAAA,IAC/F;AAAA,EACF;AAAA,EACA,MAAM,KAAK,QAAQ;AACjB,WAAO,KAAK,SAAS,MAAM;AAAA,EAC7B;AAAA,EACA,SAAS,QAAQ,OAAO,KAAK;AAC3B,UAAM,EAAE,cAAc,OAAO,OAAO,YAAY,IAAI,KAAK,UAAU,IAAI,MAAM;AAC7E,QAAI,UAAU,UAAU;AACtB,WAAK,SAAS,QAAQ,UAAU,OAAO,OAAO,QAAQ,OAAO,YAAY,SAAS;AAChF,eAAO;AAAA,MACT;AACA,aAAO,YAAY,MAAM,OAAO,GAAG;AAAA,IACrC;AACA,YAAQ,SAAS;AACjB,UAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,UAAM,mBAAmB,aAAa,gBAAgB,KAAK;AAC3D,UAAM,QAAQ,KAAK,KAAK,OAAO,MAAM,eAAe,QAAQ,kBAAkB,eAAe,MAAM,gBAAgB;AACnH,WAAO,qBAAqB,MAAM,QAAQ,KAAK;AAAA,EACjD;AAAA,EACA,YAAY,QAAQ,QAAQ,OAAO;AACjC,QAAI,KAAK,UAAU,IAAI,MAAM,GAAG;AAC9B,YAAM,OAAO,KAAK,UAAU,IAAI,MAAM;AACtC,WAAK;AACL,UAAI,CAAC,SAAS,KAAK,WAAW,GAAG;AAC/B,eAAO;AAAA,MACT;AACA,WAAK,KAAK,MAAM,KAAK,YAAY;AACjC,WAAK,KAAK,KAAK,YAAY,KAAK,EAAE;AAClC,WAAK,UAAU,OAAO,MAAM;AAAA,IAC9B;AACA,WAAO;AAAA,EACT;AAAA,EACA,SAAS,QAAQ;AACf,QAAI,KAAK,UAAU,IAAI,MAAM,GAAG;AAC9B,YAAM,aAAa,KAAK,UAAU,IAAI,MAAM;AAC5C,aAAO,WAAW;AAAA,IACpB;AACA,WAAO;AAAA,EACT;AAAA,EACA,OAAO,QAAQ;AACb,UAAM,OAAO,KAAK,UAAU,IAAI,MAAM;AACtC,QAAI,QAAQ,MAAM;AAChB,WAAK;AAAA,IACP;AAAA,EACF;AAAA,EACA,iBAAiB;AACf,WAAO;AAAA,EACT;AAAA,EACA,gBAAgB,QAAQ;AACtB,WAAO,KAAK,UAAU,IAAI,MAAM,EAAE;AAAA,EACpC;AAAA,EACA,UAAU;AACR,SAAK,KAAK,KAAK,QAAQ;AACvB,QAAI,aAAa,KAAK,MAAM;AAC1B,WAAK,KAAK,QAAQ,oBAAoB;AAAA,IACxC;AACA,SAAK,OAAO;AAAA,EACd;AAAA,EACA,SAAS;AACP,WAAO,EAAE,YAAY,MAAM;AAAA,EAC7B;AAAA,EACA,WAAW,OAAO,OAAO,cAAc;AACrC,QAAI;AACJ,QAAI,gBAAgB,MAAM;AACxB,eAAS,KAAK,MAAM,MAAM,OAAO,KAAK;AAAA,IACxC,OAAO;AACL,YAAM,KAAK,KAAK;AAChB,eAAS,EAAE,GAAG;AACd,WAAK,UAAU,IAAI,QAAQ,EAAE,IAAI,cAAc,OAAO,OAAO,UAAU,EAAE,CAAC;AAC1E,YAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,WAAK,KAAK,KAAK,eAAe,IAAI,MAAM,YAAY;AAAA,IACtD;AACA,WAAO,EAAE,QAAQ,OAAO,MAAM;AAAA,EAChC;AAAA,EACA,mBAAmB,EAAE,OAAO,OAAO,OAAO,GAAG;AAC3C,UAAM,UAAU,KAAK,KAAK,OAAO;AACjC,UAAM,EAAE,aAAa,IAAI,KAAK,UAAU,IAAI,MAAM;AAClD,UAAM,OAAO,aAAa,cAAc,KAAK;AAC7C,YAAQ,OAAO;AAAA,MACb,KAAK;AACH,eAAO,IAAI,aAAa,SAAS,cAAc,IAAI;AAAA,MACrD,KAAK;AACH,eAAO,IAAI,WAAW,SAAS,cAAc,IAAI;AAAA,MACnD,KAAK;AACH,eAAO,IAAI,WAAW,SAAS,cAAc,IAAI;AAAA,MACnD;AACE,cAAM,IAAI,MAAM,iBAAiB,OAAO;AAAA,IAC5C;AAAA,EACF;AACF;AACA,SAAS,0BAA0B,MAAM;AACvC,SAAO,CAAC,SAAS,aAAa;AAC5B,iBAAa,MAAM,MAAM,EAAE,aAAa,cAAc,CAAC,EAAE,KAAK,CAAC,aAAa;AAC1E,UAAI,CAAC,SAAS,OAAO;AACnB,gBAAQ,IAAI,EAAE,uCAAuC,OAAO;AAAA,MAC9D;AACA,eAAS,YAAY,EAAE,KAAK,CAAC,WAAW;AACtC,oBAAY,YAAY,QAAQ,OAAO,EAAE,KAAK,CAAC,WAAW;AACxD,mBAAS,OAAO,UAAU,OAAO,MAAM;AAAA,QACzC,CAAC;AAAA,MACH,CAAC;AAAA,IACH,CAAC;AACD,WAAO,CAAC;AAAA,EACV;AACF;AACA,SAAS,oBAAoB,eAAe,kBAAkB,kBAAkB;AAC9E,MAAI,YAAY,MAAM;AACpB,WAAO;AAAA,EACT;AACA,MAAI,OAAO;AACX,MAAI,iBAAiB,kBAAkB;AACrC,WAAO;AAAA,EACT,WAAW,eAAe;AACxB,WAAO;AAAA,EACT;AACA,MAAI,eAAe,MAAM;AACvB,QAAI,YAAY,SAAS,MAAM;AAC7B,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AACA,SAAO,mBAAmB;AAC5B;AACA,eAAe,OAAO;AACpB,QAAM,CAAC,eAAe,gBAAgB,IAAI,MAAM,QAAQ,IAAI;AAAA,IAC1D,IAAI,EAAE,SAAS,uBAAuB;AAAA,IACtC,IAAI,EAAE,SAAS,8BAA8B;AAAA,EAC/C,CAAC;AACD,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,UAAM,gBAAgB,CAAC;AACvB,kBAAc,aAAa,CAAC,MAAM,WAAW;AAC3C,UAAI,KAAK,SAAS,YAAY,GAAG;AAC/B,cAAM,WAAW,8CAA8C,mBAAmB,QAAQ,OAAO,KAAK;AACtG,cAAM,OAAO,IAAI,KAAK,CAAC,QAAQ,GAAG,EAAE,MAAM,yBAAyB,CAAC;AACpE,eAAO,IAAI,gBAAgB,IAAI;AAAA,MACjC;AACA,UAAI,KAAK,SAAS,OAAO,GAAG;AAC1B,eAAO,oBAAoB,eAAe,kBAAkB,kBAAkB,OAAO,iBAAiB,MAAM;AAAA,MAC9G;AACA,aAAO,SAAS;AAAA,IAClB;AACA,QAAI,aAAa;AACf,oBAAc,kBAAkB,0BAA0B,oBAAoB,eAAe,kBAAkB,kBAAkB,OAAO,iBAAiB,EAAE,CAAC;AAAA,IAC9J;AACA,QAAI,cAAc;AAClB,kBAAc,UAAU,MAAM;AAC5B,UAAI,aAAa;AACf;AAAA,MACF;AACA,UAAI,aAAa;AACf;AAAA,MACF;AACA,oBAAc;AACd,YAAM,YAAY;AAClB,aAAO,EAAE,SAAS,UAAU,CAAC;AAAA,IAC/B;AACA,QAAI;AACJ,QAAI,oBAAoB,iBAAiB,YAAY,MAAM;AACzD,oBAAc,sBAAsB,IAAI,KAAK,CAAC,yCAAyC,wBAAwB,SAAS,CAAC,GAAG,EAAE,MAAM,kBAAkB,CAAC;AACvJ,aAAO,wBAAwB,aAAa;AAAA,IAC9C,OAAO;AACL,aAAO,YAAY,aAAa;AAAA,IAClC;AACA,SAAK,KAAK,CAAC,WAAW;AACpB,oBAAc;AACd,oBAAc;AACd,YAAM,iBAAiB;AACvB,aAAO,OAAO;AAAA,QACZ,MAAM,OAAO,MAAM,QAAQ,MAAM,CAAC,CAAC;AAAA,QACnC,sBAAsB,OAAO,MAAM,2BAA2B,MAAM,CAAC,QAAQ,CAAC;AAAA,QAC9E,iBAAiB,OAAO,MAAM,qBAAqB,UAAU,CAAC,CAAC;AAAA,QAC/D,gBAAgB,OAAO,MAAM,mBAAmB,MAAM;AAAA,UACpD;AAAA,UACA;AAAA,UACA;AAAA,QACF,CAAC;AAAA,QACD,aAAa,OAAO,MAAM,gBAAgB,gBAAgB,CAAC,QAAQ,CAAC;AAAA,QACpE,SAAS,OAAO,MAAM,WAAW,gBAAgB,CAAC,CAAC;AAAA,MACrD;AACA,cAAQ,EAAE,MAAM,OAAO,CAAC;AAAA,IAC1B,CAAC,EAAE,MAAM,MAAM;AAAA,EACjB,CAAC;AACH;AACA,SAAS,qBAAqB,SAAS,OAAO;AAC5C,UAAQ,OAAO;AAAA,IACb,KAAK;AACH,aAAO,IAAI,aAAa,OAAO;AAAA,IACjC,KAAK;AACH,aAAO,IAAI,WAAW,OAAO;AAAA,IAC/B,KAAK;AACH,aAAO,IAAI,WAAW,OAAO;AAAA,IAC/B;AACE,YAAM,IAAI,MAAM,iBAAiB,OAAO;AAAA,EAC5C;AACF;AACA,IAAI,kBAAkB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AACF;AACA,IAAI,WAAW;AACf,IAAI,iBAAiB;AACrB,IAAI,cAAc,CAAC;AACnB,IAAI,cAAc;AAClB,IAAI,cAAc;AAClB,SAAS,YAAY,MAAM,mBAAmB,OAAO;AACnD,kBAAgB,mGAAmG;AACnH,MAAI,aAAa;AACf,UAAM,IAAI,MAAM,gIAAgI;AAAA,EAClJ;AACA,aAAW;AACX,gBAAc;AAChB;AACA,SAAS,aAAa,iBAAiB,mBAAmB,OAAO;AAC/D,MAAI,aAAa;AACf,UAAM,IAAI,MAAM,iIAAiI;AAAA,EACnJ;AACA,MAAI,OAAO,oBAAoB,UAAU;AACvC,qBAAiB;AAAA,EACnB,OAAO;AACL,kBAAc;AACd,UAAM,eAAe,gBAAgB,OAAO,CAAC,SAAS,YAAY,SAAS,IAAI;AAC/E,QAAI,aAAa,SAAS,GAAG;AAC3B,YAAM,IAAI,MAAM,2DAA2D,aAAa,KAAK,GAAG,gKAAgK;AAAA,IAClQ;AAAA,EACF;AACA,gBAAc;AAChB;AACA,IAAI,eAAe;AACnB,IAAI,qBAAqB;AACzB,SAAS,gBAAgB,YAAY;AACnC,iBAAe;AACjB;AACA,SAAS,kBAAkB;AACzB,MAAI,uBAAuB,IAAI;AAC7B,UAAM,IAAI,MAAM,+BAA+B;AAAA,EACjD;AACA,SAAO;AACT;AAGA,IAAI,WAAW;AAGf,IAAI,gBAAgB;AACpB,gBAAgB,QAAQ,YAAY;AAClC,QAAM,EAAE,KAAK,IAAI,MAAM,KAAK;AAC5B,SAAO,IAAI,YAAY,IAAI;AAC7B,GAAG,aAAa;AAGhB,IAAI,WAAW;AACf,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,IAAI,YAAY;AAChB,IAAI,YAAY;AAAA,EACd,MAAM;AAAA,EACN,aAAa;AAAA,EACb,kBAAkB;AAAA,EAClB,oBAAoB;AAAA,EACpB,sBAAsB;AAAA,EACtB,qBAAqB;AACvB;;;ACt/jEA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACEO,SAAS,YACd,KACA,QACA,WAAW,OACX;AACA,MAAI,UAAU;AAEd,SAAO,MAAM,CAAC,EAAE,QAAQ,CAAC,EAAE,GAAG,EAAE,GAAG,YAAY;AAC7C,UAAM,OAAO,OAAO;AACpB,QAAI,OAAO,KAAK,GAAG,KAAK,CAAC;AACzB,QAAI,OAAO,GAAG,CAAC;AAAA,EACjB,CAAC;AAED,MAAI,UAAU;AACZ,UAAM,OAAO,OAAO,OAAO,SAAS;AACpC,UAAM,KAAK,OAAO;AAClB,QAAI,CAAC,QAAQ,CAAC,IAAI;AAChB;AAAA,IACF;AAEA,QAAI,OAAO,KAAK,GAAG,KAAK,CAAC;AACzB,QAAI,OAAO,GAAG,GAAG,GAAG,CAAC;AAAA,EACvB;AAEA,MAAI,OAAO;AACb;;;AC3BA;AAAA;AAAA;AAAA;AAAA;AAAA,gBAAAG;AAAA,EAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,eAAAC;AAAA,EAAA,aAAAC;AAAA;;;ACOO,IAAM,aAAN,MAAwC;AAAA,EAK7C,YAAY,OAAe,QAAgB;AACzC,QAAI,CAAC,cAAc,KAAK,KAAK,CAAC,cAAc,MAAM,GAAG;AACnD,YAAM,IAAI,MAAM,wFAAwF,KAAK,UAAU,EAAE,OAAO,OAAO,CAAC,GAAG;AAAA,IAC7I;AAEA,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,SAAiB;AAAE,WAAO,KAAK;AAAA,EAAS;AAAA,EAE5C,UAAsB;AAC3B,WAAO,IAAI,WAAW,IAAI,KAAK,OAAO,IAAI,KAAK,MAAM;AAAA,EACvD;AACF;;;ADvBO,SAAS,SAASC,SAAa,KAAa;AACjD,SAAOA,mBAAqB,UAAUA,QAAO,MAAM,WAAW;AAChE;AAEO,SAAS,WAAWA,SAAoC;AAC7D,SAAO,SAASA,SAAQ,CAAC;AAC3B;AAEO,SAAS,WAAWA,SAAoC;AAC7D,SAAO,SAASA,SAAQ,CAAC;AAC3B;AAEO,SAAS,WAAWA,SAAoC;AAC7D,SAAO,SAASA,SAAQ,CAAC;AAC3B;AAEO,SAAS,WAAWA,SAAoC;AAC7D,SAAO,SAASA,SAAQ,CAAC;AAC3B;AAEO,SAAS,QAAQ,KAAa;AACnC,SAAO,MAAM,MAAM;AACrB;AAEO,SAASC,QAAO,KAAa;AAClC,SAAO,MAAM,MAAM;AACrB;AAEO,SAASC,OAAM,KAAa,OAAO,GAAG;AAC3C,QAAM,IAAI,MAAM;AAChB,SAAO,KAAK,MAAM,MAAM,CAAC,IAAI;AAC/B;AAEO,SAAS,aAAa,KAAmB;AAC9C,SAAO,OAAO,IAAI,SAAS,IAAI;AACjC;AAEO,SAAS,0BAA0B,EAAE,OAAO,OAAO,GAAgB,WAAmB;AAC3F,QAAMC,SAAQ,YAAY,KAAK,IAAI,QAAQ,KAAK;AAChD,SAAO,IAAI,WAAW,KAAK,MAAM,QAAQA,MAAK,GAAG,KAAK,MAAM,SAASA,MAAK,CAAC;AAC7E;AAEO,SAAS,eAAe,KAAqB;AAClD,SAAO,IAAI,OAAO,CAACC,MAAK,OAAOA,KAAI,IAAI,EAAE,GAAG,IAAI,MAAM,GAAG,CAAC,CAAC,EACxD,IAAI,IAAI,MAAM,IAAI,QAAQ,IAAI,MAAM,CAAC;AAC1C;AAEO,SAASC,OAAM,KAAa,OAAeC,OAAwB;AACxE,SAAO,MAAM,GAAG,EAAE,KAAK,CAAC,EAAE,IAAI,CAAC,GAAG,MAAM,QAAS,IAAIA,KAAK;AAC5D;AAEO,SAAS,cAAc,KAAU;AACtC,SAAO,CAAC,CAAC,OAAQ,QAAQ,YAAc,QAAQ,aAAc,CAAC,OAAO,MAAM,GAAG,KAAK,QAAQ;AAC7F;AAEO,SAAS,mBAAmB,KAAU;AAC3C,SAAO,cAAc,GAAG,KAAK,OAAO,KAAK,OAAO;AAClD;;;AEzDO,IAAM,QAAN,MAA8B;AAAA,EAKnC,YAAY,GAAW,GAAW;AAChC,SAAK,KAAK;AACV,SAAK,KAAK;AAAA,EACZ;AAAA,EAEA,IAAI,IAAY;AAAE,WAAO,KAAK;AAAA,EAAI;AAAA,EAElC,IAAI,IAAY;AAAE,WAAO,KAAK;AAAA,EAAI;AAAA,EAE3B,IAAI,IAAmB;AAC5B,WAAO,IAAI,MAAM,KAAK,IAAI,GAAG,GAAG,KAAK,IAAI,GAAG,CAAC;AAAA,EAC/C;AAAA,EAEO,IAAI,IAAmB;AAC5B,WAAO,IAAI,MAAM,KAAK,IAAI,GAAG,GAAG,KAAK,IAAI,GAAG,CAAC;AAAA,EAC/C;AAAA,EAEO,IAAI,IAAmB;AAC5B,WAAO,IAAI,MAAM,KAAK,IAAI,GAAG,GAAG,KAAK,IAAI,GAAG,CAAC;AAAA,EAC/C;AAAA,EAEO,IAAI,IAAmB;AAC5B,WAAO,IAAI,MAAM,KAAK,IAAI,GAAG,GAAG,KAAK,IAAI,GAAG,CAAC;AAAA,EAC/C;AAAA,EAEO,MAAa;AAClB,WAAO,IAAI,MAAM,KAAK,IAAI,KAAK,CAAC,GAAG,KAAK,IAAI,KAAK,CAAC,CAAC;AAAA,EACrD;AAAA,EAEO,YAAoB;AACzB,WAAO,KAAK,KAAM,KAAK,KAAK,IAAM,KAAK,KAAK,CAAE;AAAA,EAChD;AAAA,EAEO,QAAe;AACpB,WAAO,IAAI,MAAM,KAAK,MAAM,KAAK,CAAC,GAAG,KAAK,MAAM,KAAK,CAAC,CAAC;AAAA,EACzD;AACF;;;ACxCO,IAAM,MAAN,MAAwD;AAAA,EAC7D,OAAc,OAAO,MAAoB;AACvC,WAAO,CAAC,CAAC,QAAQ,CAAC,KAAK,GAAG,KAAK,GAAG,KAAK,OAAO,KAAK,MAAM,EAAE,MAAM,aAAa;AAAA,EAChF;AAAA,EAEA,OAAc,iBAAiB,KAAU,QAAgB,0BAA0B,OAAO;AACxF,QAAI,CAAC,IAAI,OAAO,GAAG,GAAG;AACpB,YAAM,IAAI,MAAM,GAAG,yBAAyB,KAAK,UAAU,GAAG,wDAAwD;AAAA,IACxH;AAEA,QAAI,CAAC,4BAA4B,IAAI,QAAQ,KAAK,IAAI,SAAS,IAAI;AACjE,YAAM,IAAI,MAAM,GAAG,mBAAmB,IAAI,sBAAsB,IAAI,kCAAkC;AAAA,IACxG;AAAA,EACF;AAAA,EAUA,YAAY,MAA4B,0BAA0B,MAAM;AACtE,UAAM,MAAO,QAAQ,CAAC;AAEtB,UAAM,SAAS,CAAC,IAAI,MAAM,IAAI,KAAK,IAAI,OAAO,IAAI,MAAM,EAAE,MAAM,aAAa;AAC7E,UAAM,SAAS,CAAC,IAAI,GAAG,IAAI,GAAG,IAAI,OAAO,IAAI,MAAM,EAAE,MAAM,aAAa;AAExE,QAAI,CAAC,UAAU,CAAC,QAAQ;AACtB,YAAM,IAAI,MAAM,2EAA2E,KAAK,UAAU,GAAG,GAAG;AAAA,IAClH;AAEA,UAAM,CAAC,GAAG,GAAG,OAAO,MAAM,IAAI,SAC1B,CAAC,IAAI,GAAG,IAAI,GAAG,IAAI,OAAO,IAAI,MAAM,IACpC,CAAC,IAAI,MAAM,IAAI,KAAK,IAAI,QAAQ,IAAI,MAAM,IAAI,SAAS,IAAI,GAAG;AAElE,QAAI,iBAAiB;AAAA,MACnB;AAAA,MAAG;AAAA,MAAG;AAAA,MAAO;AAAA,IACf,GAAG,mBAAmB,uBAAuB;AAE7C,SAAK,KAAK;AACV,SAAK,KAAK;AACV,SAAK,SAAS;AACd,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,IAAW,IAAY;AAAE,WAAO,KAAK;AAAA,EAAI;AAAA,EAEzC,IAAW,IAAY;AAAE,WAAO,KAAK;AAAA,EAAI;AAAA,EAEzC,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,SAAiB;AAAE,WAAO,KAAK;AAAA,EAAS;AAAA,EAEnD,IAAW,OAAe;AAAE,WAAO,KAAK;AAAA,EAAG;AAAA,EAE3C,IAAW,MAAc;AAAE,WAAO,KAAK;AAAA,EAAG;AAAA,EAE1C,IAAW,QAAgB;AAAE,WAAO,KAAK,IAAI,KAAK;AAAA,EAAO;AAAA,EAEzD,IAAW,SAAiB;AAAE,WAAO,KAAK,IAAI,KAAK;AAAA,EAAQ;AAAA,EAE3D,IAAW,OAAe;AAAE,WAAO,KAAK,QAAQ,KAAK;AAAA,EAAQ;AAAA,EAE7D,IAAW,UAAiB;AAAE,WAAO,IAAI,MAAM,KAAK,MAAM,KAAK,GAAG;AAAA,EAAG;AAAA,EAErE,IAAW,WAAkB;AAAE,WAAO,IAAI,MAAM,KAAK,OAAO,KAAK,GAAG;AAAA,EAAG;AAAA,EAEvE,IAAW,aAAoB;AAAE,WAAO,IAAI,MAAM,KAAK,MAAM,KAAK,MAAM;AAAA,EAAG;AAAA,EAE3E,IAAW,cAAqB;AAAE,WAAO,IAAI,MAAM,KAAK,OAAO,KAAK,MAAM;AAAA,EAAG;AAAA,EAEtE,QAAsB;AAC3B,UAAM,CAAC,GAAG,GAAG,OAAO,MAAM,IAAI,CAAC,KAAK,GAAG,KAAK,GAAG,KAAK,OAAO,KAAK,MAAM,EACnE,IAAI,CAAC,QAAQ,KAAK,MAAM,GAAG,CAAC;AAC/B,WAAO,IAAI,IAAI;AAAA,MACb;AAAA,MAAG;AAAA,MAAG;AAAA,MAAO;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEO,QAAsB;AAC3B,UAAM,CAAC,GAAG,GAAG,OAAO,MAAM,IAAI,CAAC,KAAK,GAAG,KAAK,GAAG,KAAK,OAAO,KAAK,MAAM,EACnE,IAAI,CAAC,QAAQ,KAAK,MAAM,GAAG,CAAC;AAC/B,WAAO,IAAI,IAAI;AAAA,MACb;AAAA,MAAG;AAAA,MAAG;AAAA,MAAO;AAAA,IACf,CAAC;AAAA,EACH;AAAA,EAEO,WAAyB;AAC9B,QAAI;AAAA,MACF;AAAA,MAAG;AAAA,MAAG;AAAA,MAAO;AAAA,IACf,IAAI;AACJ,UAAM,OAAO,KAAK,IAAI,QAAQ,MAAM;AACpC,QAAI,QAAQ,QAAQ;AAClB,WAAM,OAAO;AACb,eAAS;AAAA,IACX;AACA,QAAI,SAAS,OAAO;AAClB,WAAM,OAAO;AACb,gBAAU;AAAA,IACZ;AAEA,WAAO,IAAI,IAAI,EAAE,GAAG,GAAG,OAAO,OAAO,CAAC;AAAA,EACxC;AAAA,EAEO,QAAQ,GAAuC;AACpD,UAAM,SAAS,aAAa,CAAC,IAAK,EAAkB,QAAQ;AAC5D,UAAM,SAAS,aAAa,CAAC,IAAK,EAAkB,SAAS;AAC7D,WAAO,IAAI,IAAI;AAAA,MACb,GAAG,KAAK,IAAI;AAAA,MACZ,GAAG,KAAK,IAAI;AAAA,MACZ,OAAO,KAAK,QAAQ;AAAA,MACpB,QAAQ,KAAK,SAAS;AAAA,IACxB,CAAC;AAAA,EACH;AAAA,EAEO,IAAI,MAAc,MAA4B;AACnD,UAAM,CAAC,GAAG,GAAG,OAAO,MAAM,IAAI;AAAA,MAC5B,KAAK,IAAK,OAAO;AAAA,MACjB,KAAK,IAAK,OAAO;AAAA,MACjB,KAAK,QAAQ;AAAA,MACb,KAAK,SAAS;AAAA,IAChB;AACA,WAAO,IAAI,IAAI,EAAE,GAAG,GAAG,OAAO,OAAO,CAAC;AAAA,EACxC;AAAA,EAEO,mBAAmB,UAAkB,WAAiC;AAC3E,UAAM,EAAE,GAAG,GAAG,OAAO,OAAO,IAAI;AAChC,UAAM,WAAW,KAAK,IAAI,GAAG,CAAC;AAC9B,UAAM,WAAW,KAAK,IAAI,GAAG,CAAC;AAE9B,UAAM,WAAW,QAAQ;AACzB,UAAM,YAAY,SAAS;AAC3B,UAAM,eAAe,KAAK,IAAI,UAAU,WAAW,QAAQ;AAC3D,UAAM,gBAAgB,KAAK,IAAI,WAAW,YAAY,QAAQ;AAE9D,WAAQ,IAAI,IAAI,EAAE,GAAG,UAAU,GAAG,UAAU,OAAO,cAAc,QAAQ,cAAc,CAAC,EAAG,MAAM;AAAA,EACnG;AAAA,EAEO,MAAM,IAAY,IAA0B;AACjD,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,IAAI,KAAK,IAAI;AACnB,UAAM,IAAI,KAAK,IAAI;AAEnB,WAAO,IAAI,IAAI,EAAE,GAAG,GAAG,OAAO,OAAO,CAAC;AAAA,EACxC;AAAA,EAEO,aAAa,aAAqB,YAAoB;AAC3D,UAAM,IAAI,KAAK,QAAQ;AACvB,UAAM,IAAI,KAAK,SAAS;AAExB,UAAM,KAAK;AACX,UAAM,KAAK;AACX,QAAI,MAAM;AACV,QAAI,MAAM;AAEV,QAAI,IAAI,KAAK;AACb,QAAI,IAAI,KAAK;AACb,QAAI,KAAK,KAAK;AACd,QAAI,KAAK,KAAK;AAEd,QAAI,KAAK,YAAY;AACnB,YAAM,CAAC,KAAK,aAAa;AACzB,WAAK;AAAA,IACP;AACA,QAAI,KAAK,aAAa;AACpB,YAAM,CAAC,KAAK,cAAc;AAC1B,WAAK;AAAA,IACP;AACA,QAAI,IAAI,GAAG;AACT,YAAM,IAAI;AACV,UAAI;AAAA,IACN;AACA,QAAI,IAAI,GAAG;AACT,YAAM,IAAI;AACV,UAAI;AAAA,IACN;AAEA,WAAO,EAAE,IAAI,KAAK,IAAI,KAAK,GAAG,IAAI,GAAG,IAAI,GAAG,EAAE;AAAA,EAChD;AAAA,EAEO,UAAU,QAAa;AAC5B,WAAO,IAAI,IAAI;AAAA,MACb,MAAM,KAAK,OAAQ,OAAO,OAAO,KAAK;AAAA,MACtC,KAAK,KAAK,MAAO,OAAO,MAAM,KAAK;AAAA,MACnC,OAAO,KAAK,QAAS,OAAO,QAAQ,KAAK;AAAA,MACzC,QAAQ,KAAK,SAAU,OAAO,SAAS,KAAK;AAAA,IAC9C,CAAC,EAAE,SAAS,EAAE,MAAM;AAAA,EACtB;AACF;;;AC3LO,IAAM,cAAN,cAA0B,IAA4B;AAAA,EAC3D,YAAY,MAAc,KAAa,OAAe,QAAgB,0BAA0B,OAAO;AACrG,UAAM,EAAE,MAAM,KAAK,OAAO,OAAO,GAAG,uBAAuB;AAAA,EAC7D;AACF;;;ACTO,IAAM,kBAAN,MAAsB;AAAA,EAW3B,YACE,OACA,YACA,WACA,aACA,WACA;AACA,SAAK,aAAa,IAAI,WAAW,UAAU,OAAO,UAAU,MAAM;AAClE,SAAK,SAAS;AACd,SAAK,cAAc;AACnB,SAAK,aAAa;AAClB,SAAK,OAAO,IAAI,IAAI,WAAW,EAAE,QAAQ,KAAK,UAAU;AAAA,EAC1D;AAAA,EAEA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,aAAqB;AAAE,WAAO,KAAK;AAAA,EAAa;AAAA,EAE3D,IAAW,YAAoB;AAAE,WAAO,KAAK;AAAA,EAAY;AAAA,EAEzD,IAAW,MAAW;AAAE,WAAO,KAAK;AAAA,EAAM;AAAA,EAE1C,IAAW,YAAwB;AAAE,WAAO,KAAK;AAAA,EAAY;AAAA,EAE7D,IAAW,aAAqB;AAAE,WAAO,KAAK,UAAU;AAAA,EAAO;AAAA,EAE/D,IAAW,cAAsB;AAAE,WAAO,KAAK,UAAU;AAAA,EAAQ;AAAA,EAEjE,IAAW,cAAmB;AAAE,WAAO,IAAI,IAAI,KAAK,IAAI,EAAE,QAAQ,KAAK,UAAU,QAAQ,CAAC;AAAA,EAAG;AAAA,EAEtF,QAAQ,OAAe,QAAiC;AAC7D,WAAO,IAAI;AAAA,MACT,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK;AAAA,MACL,EAAE,OAAO,OAAO;AAAA,IAClB;AAAA,EACF;AACF;;;AC5CO,IAAM,gBAAN,cAA4B,gBAAyC;AAAA,EAC1E,YACE,OACA,aACA,WACA;AACA,UAAM,OAAO,OAAO,IAAI,aAAa,SAAS;AAAA,EAChD;AAAA,EAEgB,QAAQ,OAAe,QAA+B;AACpE,UAAM,EAAE,OAAO,aAAa,UAAU,IAAI,MAAM,QAAQ,OAAO,MAAM;AACrE,WAAO,IAAI,cAAc,OAAO,aAAa,SAAS;AAAA,EACxD;AACF;;;ACrBO,SAAS,IAAI,MAAW,MAAW,QAAQ,MAAM;AACtD,QAAM,QAAQ,KAAK,IAAI,GAAK,KAAK,IAAI,KAAK,OAAO,KAAK,KAAK,IAAI,KAAK,IAAI,KAAK,MAAM,KAAK,IAAI,CAAC;AAC7F,QAAM,SAAS,KAAK,IAAI,GAAK,KAAK,IAAI,KAAK,QAAQ,KAAK,MAAM,IAAI,KAAK,IAAI,KAAK,KAAK,KAAK,GAAG,CAAC;AAC9F,QAAM,eAAe,QAAQ;AAE7B,SAAO,QACH,gBAAgB,KAAK,OAAO,KAAK,OAAO,gBACxC,eAAe,KAAK,IAAI,KAAK,MAAM,KAAK,IAAI;AAClD;;;ACRO,SAAS,QAAQ,KAA4B;AAClD,QAAM,KAAK,IAAI,IAAI,CAAC,OAAO,GAAG,CAAC;AAC/B,QAAM,KAAK,IAAI,IAAI,CAAC,OAAO,GAAG,CAAC;AAC/B,QAAM,OAAO,GAAG,OAAO,CAACC,MAAK,MAAO,IAAIA,OAAM,IAAIA,MAAM,QAAQ;AAChE,QAAM,OAAO,GAAG,OAAO,CAACA,MAAK,MAAO,IAAIA,OAAM,IAAIA,MAAM,QAAQ;AAChE,QAAM,OAAO,GAAG,OAAO,CAACC,MAAK,MAAOA,OAAM,IAAI,IAAIA,MAAM,CAAC;AACzD,QAAM,OAAO,GAAG,OAAO,CAACA,MAAK,MAAOA,OAAM,IAAI,IAAIA,MAAM,CAAC;AAEzD,SAAO,IAAI,YAAY,MAAM,MAAM,MAAM,IAAI;AAC/C;;;ACRO,SAASC,mBACd,OACA,QACA,cACA,QAAQ,MACE;AACV,MAAI,uBAAuB,OACxB,IAAI,CAAC,OAAO,cAAc,EAAE,OAAO,SAAS,EAAE,EAC9C,KAAK,CAAC,IAAI,OAAO,GAAG,QAAQ,GAAG,KAAK,EACpC,IAAI,CAAC,MAAM,EAAE,QAAQ;AAExB,QAAM,OAAiB,CAAC;AAExB,SAAO,qBAAqB,SAAS,GAAG;AACtC,UAAM,OAAO,qBAAqB,IAAI;AACtC,SAAK,KAAK,IAAI;AAEd,UAAM,UAAU;AAEhB,UAAM,UAAoB,CAAC;AAC3B,aAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,YAAM,MAAM,QAAQ;AAEpB,YAAM,UAAU,MAAM;AACtB,YAAM,SAAS,MAAM;AAErB,cAAQ,KAAK,IAAI,SAAS,QAAQ,KAAK,CAAC;AAAA,IAC1C;AAEA,2BAAuB,qBAAqB;AAAA,MAC1C,CAAC,GAAG,MAAM,QAAQ,MAAM;AAAA,IAC1B;AAAA,EACF;AAEA,SAAO;AACT;;;ACpCO,SAAS,UAAU,GAAgB,SAAgC;AACxE,SAAU,KAAK,MAAM;AACnB,UAAM,CAAC,GAAG,GAAG,CAAC,IAAI;AAClB,UAAM,QAAW,KAAK,CAAC,GAAG,EAAE,MAAM,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,GAAG,SAAS;AAC/D,UAAM,QAAW,KAAK,CAAC,GAAG,EAAE,MAAM,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,GAAG,SAAS;AAC/D,UAAM,QAAW,KAAK,CAAC,GAAG,EAAE,MAAM,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,GAAG,SAAS;AAC/D,UAAM,UAAa,OAAO,CAAC,OAAO,OAAO,KAAK,GAAG,CAAC;AAElD,WAAU,IAAI,GAAG,OAAO;AAAA,EAC1B,CAAC;AACH;;;ACFO,SAAS,YAAY,WAAwB,gBAAgB,OAAoB;AACtF,SAAU,KAAK,MAAM;AACnB,UAAM,CAAC,QAAQ,KAAK,IAAI,UAAU,MAAM,MAAM,CAAC;AAC/C,QAAI,WAAW;AAAO,aAAO;AAC7B,UAAM,UAAU,KAAK,IAAI,SAAS,KAAK;AACvC,UAAM,gBAAgB,KAAK,MAAM,WAAW,gBAAgB,MAAM,EAAE;AACpE,UAAM,cAAc,SAAS,QAAQ,IAAI;AACzC,UAAM,sBAAsB,CAAC,uBAA0C;AACrE,YAAM,qBAAqB,UAAU,MAAM,MAAM;AACjD,yBAAmB,eAAe;AAClC,aAAU,KAAK,oBAAoB,GAAG,SAAS;AAAA,IACjD;AACA,UAAM,sBAAsB,oBAAoB,aAAa;AAC7D,UAAM,yBAAyB,UAAW,oBAAoB,MAAM;AACpE,UAAM,uBAAuB,iBAAiB,yBAAyB,oBAAoB,sBAAsB,IAAI;AACrH,UAAM,iBAAiB,CAAC,sBAAsB,WAAW,mBAAmB,EACzE,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,EACjB,IAAI,CAAC,MAAS,KAAK,GAAkB,SAAS,CAAC;AAClD,WAAU,OAAO,gBAAgB,WAAW;AAAA,EAC9C,CAAC;AACH;;;AC9BO,SAAS,aAAa,YAAmB;AAC9C,QAAMC,SAAQ,WAAW,MAAM;AAC/B,WAAS,IAAIA,OAAM,SAAS,GAAG,IAAI,GAAG,KAAK;AACzC,UAAM,IAAI,KAAK,MAAM,KAAK,OAAO,KAAK,IAAI,EAAE;AAC5C,UAAM,IAAIA,OAAM;AAChB,IAAAA,OAAM,KAAKA,OAAM;AACjB,IAAAA,OAAM,KAAK;AAAA,EACb;AACA,SAAOA;AACT;;;ACFO,SAASC,SAAQ,GAAW;AACjC,SAAO,KAAK,IAAI,KAAK,IAAI,CAAC,CAAC;AAC7B;AAEO,SAAS,eAAe,GAAW;AACxC,SAAO,KAAK,IAAI,KAAK,IAAI,EAAE;AAC7B;;;ACJO,IAAM,OAAN,cAAmB,IAAqB;AAAA,EAC7C,YAAY,GAAW,GAAW,OAAe,QAAgB,0BAA0B,OAAO;AAChG,UAAM,EAAE,GAAG,GAAG,OAAO,OAAO,GAAG,uBAAuB;AAAA,EACxD;AACF;;;ACHA,IAAM,OAAO;AACb,IAAM,OAAO;AACb,IAAM,WAAW;AAOV,IAAM,gBAAN,MAA8C;AAAA,EAOnD,YACE,+BACA,SACA,QAAe,IAAI,MAAM,GAAG,CAAC,GAC7B;AACA,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,SAAK,WAAW,IAAI,WAAW,OAAO,MAAM;AAC5C,SAAK,SAAS;AACd,SAAK,aAAa,8BAA8B;AAAA,MAC9C,CAAC,OAAO,GAAG,IAAI,IAAI,MAAM,OAAO,MAAM,CAAC,EAAE,IAAI,KAAK;AAAA,IACpD;AAAA,EACF;AAAA,EAEA,IAAW,QAAe;AAAE,WAAO,IAAI,MAAM,KAAK,OAAO,GAAG,KAAK,OAAO,CAAC;AAAA,EAAG;AAAA,EAE5E,IAAW,aAAqB;AAAE,WAAO,KAAK,SAAS;AAAA,EAAO;AAAA,EAE9D,IAAW,cAAsB;AAAE,WAAO,KAAK,SAAS;AAAA,EAAQ;AAAA,EAEhE,IAAW,YAAqB;AAAE,WAAO,KAAK;AAAA,EAAY;AAAA,EAE1D,IAAW,oBAA6B;AACtC,WAAO,KAAK,WAAW;AAAA,MACrB,CAAC,OAAO,GAAG,IAAI,KAAK,MAAM,EAAE,IAAI,IAAI,MAAM,KAAK,YAAY,KAAK,WAAW,CAAC;AAAA,IAC9E;AAAA,EACF;AAAA,EAEO,QAAiC,OAAe,QAAmB;AACxE,WAAO,IAAK,KAAK;AAAA,MACf,KAAK;AAAA,MACL,EAAE,OAAO,OAAO;AAAA,IAClB;AAAA,EACF;AAAA,EAEO,QAAiC,GAAW,GAAc;AAC/D,WAAO,IAAK,KAAK;AAAA,MACf,KAAK;AAAA,MACL,KAAK;AAAA,MACL,IAAI,MAAM,GAAG,CAAC;AAAA,IAChB;AAAA,EACF;AAAA,EAEO,aAAsC,IAAc;AACzD,WAAO,KAAK,QAAQ,GAAG,GAAG,GAAG,CAAC;AAAA,EAChC;AAAA,EAaO,MACL,WACA,UAAkE,CAAE,GAC/D;AACL,QAAI,WAAW;AACb,YAAM,MAAM,qBAAqB,gBAC7B,UAAU,IAAI,MAAM,IACpB,IAAI,IAAI,SAAS;AAErB,aAAO,KAAK,QAAQ,IAAI,GAAG,IAAI,CAAC,EAAE,MAAM,MAAM,OAAO;AAAA,IACvD;AAEA,UAAM,EAAE,kBAAkB,cAAc,IAAI,EAAE,kBAAkB,OAAO,eAAe,KAAK,GAAG,QAAQ;AAEtG,QAAI,kBAAkB;AACpB,aAAO,KAAK,UAAU;AAAA,IACxB;AAEA,WAAO,KAAK,aAAa,aAAa;AAAA,EACxC;AAAA,EAEQ,YAAiB;AACvB,UAAM,UAAU,KAAK,yBAAyB;AAE9C,UAAM,CAAC,eAAe,gBAAgB,WAAW,IAAI;AACrD,UAAM,cAAc,CAAC,OAAc,YAAY,IAAI,EAAE,EAAE,UAAU;AACjE,UAAM,kBAAkB,YAAY,aAAa,IAAI,YAAY,cAAc,KAAK;AAEpF,UAAM,OAAO,KAAK,MAAM,iBAAiB,QAAQ;AAEjD,UAAM,WAAW,eAAe,OAAO;AAEvC,UAAM,IAAI,KAAK,MAAM,KAAK,IAAI,GAAG,SAAS,IAAK,OAAO,IAAK,CAAC;AAC5D,UAAM,IAAI,KAAK,MAAM,KAAK,IAAI,GAAG,SAAS,IAAK,OAAO,IAAK,CAAC;AAE5D,WAAO,IAAI,KAAK,GAAG,GAAG,KAAK,IAAI,MAAM,KAAK,aAAa,CAAC,GAAG,KAAK,IAAI,MAAM,KAAK,cAAc,CAAC,CAAC;AAAA,EACjG;AAAA,EAEQ,aAAa,SAAsB;AACzC,UAAM,MAAM,QAAQ,KAAK,SAAS;AAClC,WAAO,IAAI,IAAI,IAAI,QAAQ,SAAS,IAAI,SAAS,OAAO;AAAA,EAC1D;AAAA,EAEU,2BAAoC;AAC5C,UAAM,IAAI,MAAM,wDAAwD;AAAA,EAC1E;AACF;;;AC7HO,IAAM,iBAAN,cAA6B,cAAc;AAAA,EAC7B,2BAAoC;AACrD,UAAM,MAAM,KAAK;AACjB,WAAO;AAAA,MACL,IAAI;AAAA,MACJ,IAAI;AAAA,MACJ,eAAe,CAAC,IAAI,IAAI,IAAI,EAAE,CAAC;AAAA,IACjC;AAAA,EACF;AACF;;;ACTO,IAAM,kBAAN,cAA8B,cAAc;AAAA,EAC1C,gBAAyB;AAC9B,WAAO,KAAK,UAAU,MAAM,GAAG,EAAE;AAAA,EACnC;AAAA,EAEO,iBAA0B;AAC/B,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEO,kBAA2B;AAChC,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEO,UAAmB;AACxB,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEO,aAAsB;AAC3B,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEO,cAAuB;AAC5B,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEO,WAAoB;AACzB,WAAO,KAAK,UAAU,MAAM,IAAI,EAAE;AAAA,EACpC;AAAA,EAEmB,2BAAoC;AACrD,WAAO;AAAA,MACL,KAAK,WAAW;AAAA,MAChB,KAAK,YAAY;AAAA,MACjB,KAAK,SAAS;AAAA,IAChB,EAAE,IAAI,cAAc;AAAA,EACtB;AACF;;;ACjCO,IAAM,YAAN,MAAsC;AAAA,EAI3C,YAAY,OAAe,UAAkB;AAC3C,SAAK,SAAS;AACd,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,WAAmB;AAAE,WAAO,KAAK;AAAA,EAAW;AAAA,EAEhD,SAAS,eAAe,MAAc;AAC3C,WAAO,GAAG,KAAK,QAAQ,eAAe,KAAKC,OAAM,KAAK,QAAQ,OAAO;AAAA,EACvE;AACF;;;AClBO,IAAM,aAAN,cAAyB,IAAI;AAAA,EAUlC,YAAY,KAAiC,OAAe;AAC1D,UAAM,GAAG;AACT,SAAK,SAAS;AAAA,EAChB;AAAA,EAZA,OAAc,wBAAwB,KAAU,QAAgB;AAC9D,QAAI,iBAAiB,KAAK,MAAM;AAChC,QAAI,CAAC,cAAc,IAAI,KAAK,GAAG;AAC7B,YAAM,IAAI,MAAM,GAAG,qCAAqC,IAAI,uBAAuB;AAAA,IACrF;AAAA,EACF;AAAA,EASA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AACnD;;;ACrBO,IAAM,yBAAN,MAA6B;AAAA,EAKlC,YAAY,OAAe,aAA6B;AACtD,QAAI,EAAE,OAAO,UAAU,WAAW;AAChC,YAAM,IAAI,MAAM,oEAAoE;AAAA,IACtF;AAEA,QAAI,CAAC,MAAM,QAAQ,WAAW,KAAK,YAAY,KAAK,CAAC,SAAS,EAAE,gBAAgB,aAAa,GAAG;AAC9F,YAAM,IAAI,MAAM,0FAA0F;AAAA,IAC5G;AAEA,SAAK,SAAS;AACd,SAAK,eAAe;AAAA,EACtB;AAAA,EAEA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,cAA8B;AAAE,WAAO,KAAK;AAAA,EAAc;AAAA,EAE9D,SAAc;AACnB,WAAO;AAAA,MACL,OAAO,KAAK;AAAA,MACZ,aAAa,KAAK,YAAY,IAAI,CAAC,MAAM,MAAM,KAAK,CAAC,CAAC;AAAA,IACxD;AAAA,EACF;AAAA,EAEA,OAAc,SAASC,QAAmC;AACxD,UAAM,cAAcA,OAAK,YAAY,IAAI,CAAC,MAAW,IAAI,aAAa,CAAC,CAAC;AACxE,WAAO,IAAI,uBAAuBA,OAAK,OAAO,WAAW;AAAA,EAC3D;AACF;;;AC5BO,IAAM,eAAN,cAA2B,WAAW;AAAA,EAgB3C,YAAY,KAAiC,OAAe,OAAe,YAAoB;AAC7F,UAAM,KAAK,KAAK;AAChB,SAAK,SAAS;AACd,SAAK,cAAc;AAAA,EACrB;AAAA,EAnBA,OAAc,0BAA0B,KAAU,QAAgB;AAChE,eAAW,wBAAwB,KAAK,MAAM;AAE9C,QACE,CAAC,mBAAmB,IAAI,KAAK,KAC1B,CAAC,mBAAmB,IAAI,UAAU,GACrC;AACA,YAAM,IAAI,MAAM,GAAG,uCAAuC,IAAI,eAAe,IAAI,2CAA2C;AAAA,IAC9H;AAAA,EACF;AAAA,EAYA,IAAW,QAAgB;AAAE,WAAO,KAAK;AAAA,EAAQ;AAAA,EAEjD,IAAW,aAAqB;AAAE,WAAO,KAAK;AAAA,EAAa;AAC7D;;;ACxBO,SAAS,oBAAoB,KAAwC;AAC1E,SAAO,IAAI,qBAAqB;AAClC;AAEO,SAAS,wBAAiC,WAAoB,WAAsD;AACzH,QAAM,YAAY,EAAE,UAAU;AAC9B,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACXO,SAAS,mBAAgC;AAC9C,QAAMC,SAAQ,OAAO;AACrB,MAAI,CAACA;AAAO,UAAM,IAAI,MAAM,8DAA8D;AAE1F,QAAM,WAAW,MAAM;AACrB,UAAM,IAAI,MAAM,6DAA6D;AAAA,EAC/E;AAEA,SAAO;AAAA,IACL,QAAQ;AAAA,IACR;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,OAAO;AAAA,IACP,qBAAqB,MAAM,SAAS,cAAc,QAAQ;AAAA,IAC1D,oBAAoB,MAAM,SAAS,cAAc,KAAK;AAAA,IACtD,oBAAoB,MAAM,SAAS,cAAc,OAAO;AAAA,IACxD,OAAAA;AAAA,IACA;AAAA,EACF;AACF;;;ACtBO,SAAS,WAAoB;AAClC,SAAO,OAAO,WAAW,YACpB,OAAO,YAAY,eACnB,QAAQ,YAAY,QACpB,QAAQ,SAAS,QAAQ;AAChC;;;ACFO,SAAS,iBAAiB,IAAsB;AACrD,MAAI,iBAAiB;AACrB,MAAI,CAAC,MAAM,SAAS,GAAG;AACrB,QAAI;AAEF,WAAK,UAAQ;AAAA,IACf,SAAS,KAAP;AACA,uBAAkB,IAAY,SAAS;AAAA,IACzC;AAAA,EACF;AAEA,QAAM,WAAW,KACb,CAAC,aAAqB,IAAI,QAAQ,CAAC,SAAS,WAAW;AAAE,OAAG,SAAS,UAAU,CAAC,KAAUC,YAAY,MAAM,OAAO,GAAG,IAAI,QAAQA,OAAM,CAAE;AAAA,EAAG,CAAC,IAC9I,MAAM;AAAE,UAAM,IAAI,MAAM,qEAAqE,gBAAgB;AAAA,EAAG;AACpH,SAAO,EAAE,SAAS;AACpB;;;ACdO,SAAS,kBAA+B;AAE7C,QAAM,SAAS,OAAO,aAAa,OAAO;AAC1C,QAAM,QAAQ,OAAO,SAAS,OAAO;AAErC,QAAM,QAAQ,OAAO,YAAY,OAAO;AAExC,QAAM,sBAAsB,MAAM;AAChC,QAAI;AAAQ,aAAO,IAAI,OAAO;AAC9B,UAAM,IAAI,MAAM,4EAA4E;AAAA,EAC9F;AAEA,QAAM,qBAAqB,MAAM;AAC/B,QAAI;AAAO,aAAO,IAAI,MAAM;AAC5B,UAAM,IAAI,MAAM,0EAA0E;AAAA,EAC5F;AAEA,QAAMC,sBAAqB,MAAM;AAC/B,QAAI;AAAO,aAAO,IAAI,MAAM;AAC5B,UAAM,IAAI,MAAM,0EAA0E;AAAA,EAC5F;AAEA,QAAMC,SAAQ,OAAO;AAGrB,QAAM,aAAa,iBAAiB;AAEpC,SAAO;AAAA,IACL,QAAQ,UAAU,MAAM;AAAA,IAAC;AAAA,IACzB,0BAA0B,OAAO,4BAA4B,MAAM;AAAA,IAAC;AAAA,IACpE,OAAO,SAAS,MAAM;AAAA,IAAC;AAAA,IACvB,WAAW,OAAO,aAAa,MAAM;AAAA,IAAC;AAAA,IACtC,OAAO,OAAO,oBAAoB,MAAM;AAAA,IAAC;AAAA,IACzC;AAAA,IACA;AAAA,IACA,oBAAAD;AAAA,IACA,OAAAC;AAAA,IACA,GAAG;AAAA,EACL;AACF;;;AC3CO,SAASC,aAAqB;AACnC,SAAO,OAAO,WAAW,YACpB,OAAO,aAAa,eACpB,OAAO,qBAAqB,eAC5B,OAAO,sBAAsB,eAC7B,OAAO,qBAAqB,eAC5B,OAAO,cAAc,eACrB,OAAO,6BAA6B;AAC3C;;;ACDA,IAAI;AAEJ,SAAS,SAAsB;AAC7B,MAAI,CAAC,aAAa;AAChB,UAAM,IAAI,MAAM,uEAAuE;AAAA,EACzF;AACA,SAAO;AACT;AAEA,SAAS,OAAOC,MAAkB;AAChC,gBAAcA;AAChB;AAEA,SAAS,aAAa;AAGpB,MAAIC,WAAU;AAAG,WAAO,OAAO,iBAAiB,CAAC;AACjD,MAAI,SAAS;AAAG,WAAO,OAAO,gBAAgB,CAAC;AAC/C,SAAO;AACT;AAEA,SAAS,YAAYD,MAA2B;AAC9C,MAAI,CAAC,aAAa;AAChB,eAAW;AAAA,EACb;AAEA,MAAI,CAAC,aAAa;AAChB,UAAM,IAAI,MAAM,4EAA4E;AAAA,EAC9F;AAEA,QAAM,EAAE,SAAS,YAAY,QAAQ,QAAQ,YAAY,MAAM,IAAIA;AACnE,cAAY,SAAS;AACrB,cAAY,QAAQ;AACpB,cAAY,sBAAsBA,KAAI,wBAAwB,MAAM,IAAI,OAAO;AAC/E,cAAY,qBAAqBA,KAAI,uBAAuB,MAAM,IAAI,MAAM;AAE5E,cAAY,YAAYA,KAAI,aAAa,YAAY;AACrD,cAAY,QAAQA,KAAI,SAAS,YAAY;AAC7C,cAAY,QAAQA,KAAI,SAAS,YAAY;AAC7C,cAAY,WAAWA,KAAI,YAAY,YAAY;AACrD;AAEO,IAAMA,OAAM;AAAA,EACjB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,WAAAC;AAAA,EACA;AACF;AAEA,WAAW;;;AC3DJ,SAAS,aAAa,KAAmB;AAC9C,MAAI,CAACC,KAAI,SAAS,KAAK,OAAO,QAAQ,UAAU;AAC9C,WAAO,SAAS,eAAe,GAAG;AAAA,EACpC;AACA,SAAO;AACT;;;ACJO,SAAS,oBAAoB,WAA4F;AAC9H,QAAM,EAAE,QAAQ,0BAAAC,0BAAyB,IAAIC,KAAI,OAAO;AAExD,MAAI,qBAAqBD,2BAA0B;AACjD,WAAO;AAAA,EACT;AAEA,QAAM,SAAS,aAAa,SAAS;AAErC,MAAI,EAAE,kBAAkB,SAAS;AAC/B,UAAM,IAAI,MAAM,gEAAgE;AAAA,EAClF;AAEA,QAAM,MAAM,OAAO,WAAW,IAAI;AAClC,MAAI,CAAC,KAAK;AACR,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AAEA,SAAO;AACT;;;AChBO,IAAK,iBAAL,kBAAKE,oBAAL;AAEL,EAAAA,gBAAA,cAAW;AAEX,EAAAA,gBAAA,eAAY;AAEZ,EAAAA,gBAAA,iBAAc;AAEd,EAAAA,gBAAA,kBAAe;AARL,SAAAA;AAAA,GAAA;AAoBL,IAAM,uBAAN,MAA4D;AAAA,EAajE,YAAY,UAAiC,CAAC,GAAG;AAC/C,UAAM;AAAA,MACJ;AAAA,MAAgB;AAAA,MAAiB;AAAA,MAAW;AAAA,MAAU;AAAA,MAAW;AAAA,IACnE,IAAI;AACJ,SAAK,iBAAiB,kBAAkB;AACxC,SAAK,kBAAkB,mBAAmB;AAC1C,SAAK,YAAY,aAAa;AAC9B,SAAK,WAAW,YAAY;AAC5B,SAAK,YAAY,aAAa;AAC9B,SAAK,UAAU,WAAW;AAAA,EAC5B;AACF;AAEO,IAAM,gBAAN,MAAoB;AAAA,EAOzB,YACE,MACA,QACA,UAAiC,CAAC,GAClC;AAEA,SAAK,OAAO,OAAO,SAAS,WACxB,CAAC,IAAI,IACJ,gBAAgB,gBAAgB,KAAK,OAAO;AACjD,SAAK,SAAS;AACd,SAAK,UAAU,IAAI,qBAAqB,OAAO;AAAA,EACjD;AAAA,EAEA,aAAa,KAAuC;AAClD,UAAM,EAAE,QAAQ,IAAI,KAAK;AACzB,WAAO,KAAK,KAAK,IAAI,CAAC,MAAM,IAAI,YAAY,CAAC,EAAE,KAAK,EAAE,OAAO,CAAC,IAAI,OAAQ,KAAK,KAAK,KAAK,IAAK,CAAC,IAAK,IAAI;AAAA,EAC1G;AAAA,EAEA,gBAAwB;AACtB,UAAM,EAAE,UAAU,QAAQ,IAAI,KAAK;AACnC,WAAO,KAAK,KAAK,SAAS,WAAY,IAAI;AAAA,EAC5C;AAAA,EAEA,aAAa,KAA+B,YAAkC;AAC5E,UAAM,EAAE,eAAe,IAAI,KAAK;AAChC,UAAM,cAAc,mBAAmB,qCAA+B,mBAAmB;AACzF,UAAM,aAAa,mBAAmB,mCAA8B,mBAAmB;AAEvF,UAAM,iBAAiB,KAAK,aAAa,GAAG;AAC5C,UAAM,kBAAkB,KAAK,cAAc;AAC3C,UAAM,IAAK,cAAc,KAAK,OAAO,IAAI,iBAAiB,KAAK,OAAO;AACtE,UAAM,IAAI,aAAa,KAAK,OAAO,IAAI,kBAAkB,KAAK,OAAO;AAGrE,QAAI,YAAY;AACd,YAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,YAAM,OAAO,KAAK,IAAI,KAAK,IAAI,GAAG,QAAQ,cAAc,GAAG,CAAC;AAC5D,YAAM,OAAO,KAAK,IAAI,KAAK,IAAI,GAAG,SAAS,eAAe,GAAG,CAAC;AAC9D,aAAO,EAAE,GAAG,MAAM,GAAG,KAAK;AAAA,IAC5B;AACA,WAAO,EAAE,GAAG,EAAE;AAAA,EAChB;AAAA,EAEA,KAAK,WAAkE;AACrE,UAAM,SAAS,aAAa,SAAS;AACrC,UAAM,MAAM,oBAAoB,MAAM;AAEtC,UAAM;AAAA,MACJ;AAAA,MAAiB;AAAA,MAAW;AAAA,MAAU;AAAA,MAAW;AAAA,IACnD,IAAI,KAAK;AAET,QAAI,OAAO,GAAG,cAAc;AAC5B,UAAM,eAAe,KAAK,aAAa,GAAG;AAC1C,UAAM,aAAa,KAAK,cAAc;AAEtC,QAAI,YAAY;AAChB,UAAM,YAAY,KAAK,aAAa,KAAK,MAAM;AAC/C,QAAI,SAAS,UAAU,GAAG,UAAU,GAAG,cAAc,UAAU;AAE/D,QAAI,YAAY;AAChB,SAAK,KAAK,QAAQ,CAAC,UAAU,MAAM;AACjC,YAAM,IAAI,UAAU,UAAU;AAC9B,YAAM,IAAI,UAAU,UAAU,KAAM,IAAI,KAAK;AAC7C,UAAI,SAAS,UAAU,GAAG,CAAC;AAAA,IAC7B,CAAC;AAAA,EACH;AACF;;;ACjHO,IAAM,iBAAN,MAAqB;AAAA,EAS1B,YAAY,UAA2B,CAAC,GAAG;AACzC,UAAM;AAAA,MACJ;AAAA,MAAU;AAAA,MAAW;AAAA,MAAO;AAAA,IAC9B,IAAI;AACJ,SAAK,WAAW,YAAY;AAC5B,SAAK,YAAY,aAAa;AAC9B,SAAK,QAAQ;AAEb,UAAM,0BAA0B;AAAA,MAC9B;AAAA,MACA,iBAAiB,KAAK;AAAA,IACxB;AACA,SAAK,mBAAmB,IAAI,qBAAqB,EAAE,GAAG,yBAAyB,GAAG,iBAAiB,CAAC;AAAA,EACtG;AACF;AAEO,IAAM,UAAN,MAAc;AAAA,EAKnB,YACE,KACA,UAA2B,CAAC,GAC5B;AACA,SAAK,MAAM,IAAI,IAAI,GAAG;AACtB,SAAK,UAAU,IAAI,eAAe,OAAO;AAAA,EAC3C;AAAA,EAEA,KAAK,WAAkE;AACrE,UAAM,MAAM,oBAAoB,SAAS;AAEzC,UAAM,EAAE,UAAU,UAAU,IAAI,KAAK;AAErC,UAAM;AAAA,MACJ;AAAA,MAAG;AAAA,MAAG;AAAA,MAAO;AAAA,IACf,IAAI,KAAK;AACT,QAAI,cAAc;AAClB,QAAI,YAAY;AAChB,QAAI,WAAW,GAAG,GAAG,OAAO,MAAM;AAElC,UAAM,EAAE,MAAM,IAAI,KAAK;AACvB,QAAI,OAAO;AACT,UAAI,cAAc,CAAC,KAAK,GAAG,EAAE,GAAG,IAAK,YAAY,GAAI,EAAE,GAAG,KAAK,QAAQ,gBAAgB,EAAE,KAAK,SAAS;AAAA,IACzG;AAAA,EACF;AACF;;;AC3DO,SAAS,eACd,WACA,YACA;AACA,QAAM,kBAAkB,MAAM,QAAQ,UAAU,IAAI,aAAa,CAAC,UAAU;AAE5E,kBAAgB,QAAQ,CAAC,QAAQ;AAE/B,UAAM,QAAQ,eAAe,gBACzB,IAAI,QACH,oBAAoB,GAAG,IAAI,IAAI,UAAU,QAAQ;AAGtD,UAAM,MAAM,eAAe,gBACvB,IAAI,MACH,oBAAoB,GAAG,IAAI,IAAI,UAAU,MAAM,IAAI,IAAI,GAAG;AAE/D,UAAM,QAAQ,QAAQ,GAAGC,OAAM,KAAK,MAAM;AAC1C,QAAI,QAAQ,KAAK,EAAE,MAAM,CAAC,EAAE,KAAK,SAAS;AAAA,EAC5C,CAAC;AACH;;;AC1BO,SAAS,cAAc,OAAsD;AAClF,QAAM,EAAE,OAAO,MAAM,IAAIC,KAAI,OAAO;AAEpC,SAAQ,iBAAiB,SAAS,MAAM,YAClC,iBAAiB,SAAS,MAAM,cAAc;AACtD;;;ACJO,SAAS,iBAAiB,OAAgE;AAE/F,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,QAAI,iBAAiBC,KAAI,OAAO,EAAE,UAAU,cAAc,KAAK;AAAG,cAAQ,IAAI;AAE9E,aAAS,QAAQ,GAAU;AACzB,UAAI,CAAC,EAAE;AAAe;AAEtB,QAAE,cAAc,oBAAoB,QAAQ,MAAM;AAClD,QAAE,cAAc,oBAAoB,SAAS,OAAO;AACpD,aAAO,CAAC;AAAA,IACV;AAEA,aAAS,OAAO,GAAU;AACxB,UAAI,CAAC,EAAE;AAAe;AACtB,QAAE,cAAc,oBAAoB,QAAQ,MAAM;AAClD,QAAE,cAAc,oBAAoB,SAAS,OAAO;AACpD,cAAQ,CAAC;AAAA,IACX;AAEA,UAAM,iBAAiB,QAAQ,MAAM;AACrC,UAAM,iBAAiB,SAAS,OAAO;AAAA,EACzC,CAAC;AACH;;;ACxBO,SAAS,cAAc,KAAsC;AAClE,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,QAAI,EAAE,eAAe;AAAO,aAAO,IAAI,MAAM,kDAAkD,CAAC;AAChG,UAAM,SAAS,IAAI,WAAW;AAC9B,WAAO,SAAS,MAAM;AACpB,UAAI,OAAO,OAAO,WAAW;AAAU,eAAO,IAAI,MAAM,kEAAkE,CAAC;AAC3H,YAAM,MAAMC,KAAI,OAAO,EAAE,mBAAmB;AAC5C,UAAI,SAAS,MAAM,QAAQ,GAAG;AAC9B,UAAI,UAAU;AACd,UAAI,MAAM,OAAO;AAAA,IACnB;AACA,WAAO,UAAU;AACjB,WAAO,cAAc,GAAG;AAAA,EAC1B,CAAC;AACH;;;ACbO,SAAS,mBAAmBC,QAA0F;AAC3H,QAAM,EAAE,OAAO,MAAM,IAAIC,KAAI,OAAO;AAEpC,MAAID,kBAAiB,OAAO;AAC1B,WAAO,IAAI,WAAWA,OAAM,cAAcA,OAAM,aAAa;AAAA,EAC/D;AACA,MAAIA,kBAAiB,OAAO;AAC1B,WAAO,IAAI,WAAWA,OAAM,YAAYA,OAAM,WAAW;AAAA,EAC3D;AACA,SAAO,IAAI,WAAWA,OAAM,OAAOA,OAAM,MAAM;AACjD;;;ACPO,SAASE,cAAa,EAAE,OAAO,OAAO,GAAmC;AAC9E,QAAM,EAAE,oBAAoB,IAAIC,KAAI,OAAO;AAC3C,QAAM,SAAS,oBAAoB;AACnC,SAAO,QAAQ;AACf,SAAO,SAAS;AAChB,SAAO;AACT;AAEO,SAAS,sBAAsB,OAAwD,MAAuC;AACnI,QAAM,EAAE,WAAAC,WAAU,IAAID,KAAI,OAAO;AAEjC,MAAI,EAAE,iBAAiBC,eAAc,CAAC,cAAc,KAAK,GAAG;AAC1D,UAAM,IAAI,MAAM,4DAA4D;AAAA,EAC9E;AAEA,QAAM,EAAE,OAAO,OAAO,IAAI,QAAQ,mBAAmB,KAAK;AAC1D,QAAM,SAASF,cAAa,EAAE,OAAO,OAAO,CAAC;AAE7C,MAAI,iBAAiBE,YAAW;AAC9B,wBAAoB,MAAM,EAAE,aAAa,OAAO,GAAG,CAAC;AAAA,EACtD,OAAO;AACL,wBAAoB,MAAM,EAAE,UAAU,OAAO,GAAG,GAAG,OAAO,MAAM;AAAA,EAClE;AACA,SAAO;AACT;;;ACzBA,eAAsB,oBACpB,WACA,QAC4B;AAC5B,QAAM,eAAe,UAAUC,KAAI,OAAO,EAAE,oBAAoB;AAEhE,QAAM,CAAC,QAAQ,OAAO,WAAW,IAAI,UAAU,MAAM,MAAM,WAAW,SAAS,IAAI,IAAI,CAAC;AACxF,QAAM,cAAiB,KAAK,MAAM,UAAU,KAAK,QAAQ,OAAO,WAAW,EAAE,MAAM,CAAC;AACpF,QAAS,gBAAW,SAAS,aAAa,YAAY;AAEtD,cAAY,QAAQ;AAEpB,SAAO;AACT;;;AChBO,SAAS,eAAeC,QAAY;AACzC,QAAM,EAAE,OAAO,QAAQ,MAAM,IAAIC,KAAI,OAAO;AAE5C,SAAOD,kBAAiB,SACnBA,kBAAiB,UACjBA,kBAAiB;AACxB;;;ACHO,SAAS,cAAcE,QAA6C,WAAmB,cAAc,OAAO;AACjH,QAAM,EAAE,OAAO,OAAO,IAAIC,KAAI,OAAO;AAErC,MAAI,EAAED,kBAAiB,SAASA,kBAAiB,SAAS;AACxD,UAAM,IAAI,MAAM,0EAA0E;AAAA,EAC5F;AAEA,MAAI,aAAa;AAAG,WAAOE,cAAa,EAAE,OAAO,GAAG,QAAQ,EAAE,CAAC;AAC/D,QAAM,OAAO,mBAAmBF,MAAK;AACrC,QAAMG,SAAQ,YAAY,KAAK,IAAI,KAAK,QAAQ,KAAK,KAAK;AAC1D,QAAM,QAAQA,SAAQ,KAAK;AAC3B,QAAM,SAASA,SAAQ,KAAK;AAE5B,QAAM,eAAeD,cAAa,EAAE,OAAO,WAAW,QAAQ,UAAU,CAAC;AACzE,QAAM,cAAcF,kBAAiB,SAASA,SAAQ,sBAAsBA,MAAK;AAEjF,QAAM,SAAS,KAAK,IAAI,QAAQ,MAAM,IAAI;AAC1C,QAAM,KAAK,eAAe,QAAQ,SAAS,SAAS;AACpD,QAAM,KAAK,eAAe,SAAS,QAAQ,SAAS;AACpD,MAAI,YAAY,QAAQ,KAAK,YAAY,SAAS;AAAG,wBAAoB,YAAY,EAAE,UAAU,aAAa,IAAI,IAAI,OAAO,MAAM;AAEnI,SAAO;AACT;;;ACjBO,IAAM,WAAN,MAAe;AAAA,EAapB,YAAY,QAAkC,oBAAoB,OAAO;AAZzE,SAAQ,gBAAkD,CAAC;AAE3D,SAAQ,YAAiC,CAAC;AAI1C,SAAQ,qBAAqB;AAE7B,SAAQ,mBAA+B,CAAC;AAExC,SAAQ,aAAa;AAGnB,QAAI,CAAC,MAAM,QAAQ,MAAM,GAAG;AAC1B,YAAM,IAAI,MAAM,4HAA4H,QAAQ;AAAA,IACtJ;AAEA,SAAK,qBAAqB;AAC1B,SAAK,aAAa,OAAO;AAEzB,WAAO,QAAQ,CAACI,QAAO,QAAQ;AAC7B,UAAI,WAAWA,MAAK,GAAG;AACrB,aAAK,cAAc,OAAOA;AAC1B,aAAK,iBAAiB,OAAOA,OAAM;AACnC;AAAA,MACF;AAEA,UAAI,WAAWA,MAAK,GAAG;AACrB,cAAM,YAAaA,OAAc,MAAM;AACvC,YAAI,cAAc,GAAG;AACnB,gBAAM,IAAI,MAAM,yCAAyC,oDAAoD;AAAA,QAC/G;AAEA,aAAK,cAAc,OAAOA;AAC1B,aAAK,iBAAiB,OAAQA,OAAc,MAAM,MAAM,CAAC;AACzD;AAAA,MACF;AAGA,YAAM,SAAUA,kBAAyBC,KAAI,OAAO,EAAE,SAASD,SAAQ,sBAAsBA,MAAK;AAClG,WAAK,UAAU,OAAO;AACtB,WAAK,iBAAiB,OAAO,CAAC,OAAO,QAAQ,OAAO,OAAO,CAAC;AAAA,IAC9D,CAAC;AAAA,EACH;AAAA,EAEA,IAAW,eAAiD;AAC1D,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,WAAgC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,eAAwB;AACjC,WAAO,KAAK,YAAY,KAAK,KAAK;AAAA,EACpC;AAAA,EAEA,IAAW,YAAoB;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,kBAA8B;AACvC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,YAAgC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,0BAAwC;AACjD,WAAOE,OAAM,KAAK,WAAW,GAAG,CAAC,EAAE;AAAA,MACjC,CAAC,GAAG,aAAa,KAAK,2BAA2B,QAAQ;AAAA,IAC3D;AAAA,EACF;AAAA,EAEO,SAAS,UAAiE;AAC/E,WAAO,KAAK,SAAS,aAAa,KAAK,aAAa;AAAA,EACtD;AAAA,EAEO,mBAAmB,UAA4B;AACpD,WAAO,KAAK,iBAAiB;AAAA,EAC/B;AAAA,EAEO,eAAe,UAA0B;AAC9C,WAAO,KAAK,iBAAiB,UAAU;AAAA,EACzC;AAAA,EAEO,cAAc,UAA0B;AAC7C,WAAO,KAAK,iBAAiB,UAAU;AAAA,EACzC;AAAA,EAEO,2BAA2B,UAA8B;AAC9D,QAAI,OAAO,KAAK,cAAc,UAAU;AACtC,YAAM,IAAI,MAAM,uFAAuF;AAAA,IACzG;AAEA,UAAM,QAAQ,KAAK,cAAc,QAAQ;AACzC,UAAM,SAAS,KAAK,eAAe,QAAQ;AAC3C,WAAO,0BAA0B,EAAE,OAAO,OAAO,GAAG,KAAK,SAAS;AAAA,EACpE;AAAA,EAWO,cAAc,WAAmB,iBAAiB,MAAmB;AAC1E,SAAK,aAAa;AAElB,WAAU,KAAK,MAAM;AACnB,YAAM,eAAeA,OAAM,KAAK,WAAW,GAAG,CAAC,EAAE,IAAI,CAAC,aAAa;AACjE,cAAMF,SAAQ,KAAK,SAAS,QAAQ;AAEpC,YAAIA,kBAAoB,QAAQ;AAC9B,cAAI,YAAY,WAAWA,MAAK,IAAIA,SAAW,WAAWA,MAAK;AAC/D,sBAAY,YAAY,WAA0B,cAAc;AAEhE,cAAI,UAAU,MAAM,OAAO,aAAa,UAAU,MAAM,OAAO,WAAW;AACxE,wBAAe,MAAS,eAAe,WAA0B,CAAC,WAAW,SAAS,GAAG,OAAO,KAAK;AAAA,UACvG;AAEA,iBAAO,UAAU,KAAK,WAAW,WAAW,CAAC;AAAA,QAC/C;AAEA,YAAIA,kBAAiBC,KAAI,OAAO,EAAE,QAAQ;AACxC,iBAAU,gBAAW,WAAW,cAAcD,QAAO,WAAW,cAAc,CAAC;AAAA,QACjF;AAEA,cAAM,IAAI,MAAM,+BAA+B,qGAAqGA,QAAO;AAAA,MAC7J,CAAC;AAED,YAAM,cAAiB,MAAM,aAAa,IAAI,CAAC,MAAS,KAAK,GAAG,SAAS,CAAC,CAAC,EAAE,KAAK,KAAK,WAAW,WAAW,WAAW,CAAC;AAGzH,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;;;AC1IA,eAAsB,WAAW,QAAsC;AACrE,MAAI,kBAAkB;AAAU,WAAO;AACvC,QAAM,gBAAgB,MAAM,QAAQ,MAAM,IAAI,SAAS,CAAC,MAAM;AAC9D,MAAI,CAAC,cAAc;AAAQ,UAAM,IAAI,MAAM,0CAA0C;AACrF,QAAM,aAAa,CAAC,QAAiB,MAAM,QAAQ,MAAM,IAAI,mBAAmB,SAAS;AACzF,QAAM,aAAa,cAAc,IAAI,YAAY;AACjD,aAAW,QAAQ,CAACG,QAAO,MAAM;AAC/B,QAAI,CAAC,eAAeA,MAAK,KAAK,CAAC,WAAWA,MAAK,KAAK,CAAC,WAAWA,MAAK,GAAG;AACtE,UAAI,OAAO,cAAc,OAAO;AAAU,cAAM,IAAI,MAAM,eAAe,WAAW,CAAC,qEAAqE,cAAc,IAAI;AAC5K,YAAM,IAAI,MAAM,eAAe,WAAW,CAAC,8HAA8H;AAAA,IAC3K;AACA,QAAI,WAAWA,MAAK,GAAG;AAErB,YAAM,YAAYA,OAAM,MAAM;AAC9B,UAAI,cAAc;AAAG,cAAM,IAAI,MAAM,eAAe,WAAW,CAAC,gCAAgC,oDAAoD;AAAA,IACtJ;AAAA,EACF,CAAC;AAED,QAAM,QAAQ,IAAI,WAAW,IAAI,CAACA,WAAU,eAAeA,MAAK,KAAK,iBAAiBA,MAAK,CAAC,CAAC;AAC7F,SAAO,IAAI,SAAS,YAAY,MAAM,QAAQ,MAAM,CAAC;AACvD;;;AClBA,eAAsB,aAAaC,QAAkB,YAAuE;AAC1H,QAAM,EAAE,OAAO,IAAIC,KAAI,OAAO;AAC9B,MAAI,SAASD;AACb,MAAI,EAAEA,kBAAiB,SAAS;AAC9B,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,QAAI,SAAS,YAAY;AAAG,YAAM,IAAI,MAAM,4CAA4C;AACxF,UAAM,iBAAiB,SAAS,SAAS,CAAC;AAC1C,aAAS,0BAA0B,SAAS,iBAAiB,MAAM,oBAAoB,cAAc;AAAA,EACvG;AACA,QAAM,MAAM,oBAAoB,MAAM;AACtC,QAAM,QAAQ,WACX,IAAI,CAAC,QAAS,eAAe,gBAAgB,IAAI,QAAQ,OAAO,OAAO,OAAO,MAAM,EAAE,IAAI,MAAM,IAAI,GAAI,EACxG,IAAI,CAAC,QAAQ,IAAI,mBAAmB,OAAO,OAAO,OAAO,MAAM,CAAC;AACnE,SAAO,MAAM,IAAI,CAAC,EAAE,GAAG,GAAG,OAAO,OAAO,MAAM;AAC5C,UAAM,UAAUE,cAAa,EAAE,OAAO,OAAO,CAAC;AAC9C,QAAI,QAAQ,KAAK,SAAS;AAAG,0BAAoB,OAAO,EAAE,aAAa,IAAI,aAAa,GAAG,GAAG,OAAO,MAAM,GAAG,GAAG,CAAC;AAClH,WAAO;AAAA,EACT,CAAC;AACH;;;AClBA,eAAsB,mBAAmB,aAAwC,YAAiE;AAChJ,MAAI,CAAC,WAAW,WAAW,KAAK,CAAC,WAAW,WAAW,GAAG;AACxD,UAAM,IAAI,MAAM,2DAA2D;AAAA,EAC7E;AAEA,MAAI,WAAW,WAAW,KAAK,YAAY,MAAM,KAAK,GAAG;AACvD,UAAM,IAAI,MAAM,kDAAkD;AAAA,EACpE;AAEA,SAAU,KAAK,MAAM;AACnB,UAAM,CAAC,WAAW,UAAU,WAAW,IAAI,YAAY,MAAM,MAAM,WAAW,WAAW,IAAI,IAAI,CAAC;AAClG,UAAM,QAAQ,WAAW,IAAI,CAAC,QAAS,eAAe,gBAAgB,IAAI,QAAQ,UAAU,SAAS,EAAE,MAAM,GAAI,EAC9G,IAAI,CAAC,QAAQ,IAAI,mBAAmB,UAAU,SAAS,CAAC;AAC3D,UAAM,cAAc,MACjB,OAAO,CAAC,QAAQ,IAAI,QAAQ,KAAK,IAAI,SAAS,CAAC,EAC/C,IAAI,CAAC,EAAE,GAAG,GAAG,OAAO,OAAO,MAAS,QAAQ,YAAY,KAAK,WAAW,UAAU,WAAW,GAAG,CAAC,GAAG,GAAG,CAAC,GAAG,CAAC,QAAQ,OAAO,WAAW,CAAC,CAAC;AAC3I,WAAO;AAAA,EACT,CAAC;AACH;;;AChCA,eAAsB,aACpB,KAEAC,OACmB;AACnB,QAAM,EAAE,OAAAC,OAAM,IAAIC,KAAI,OAAO;AAC7B,QAAM,MAAM,MAAMD,OAAM,KAAKD,KAAI;AACjC,MAAI,EAAE,IAAI,SAAS,MAAM;AACvB,UAAM,IAAI,MAAM,qBAAqB,IAAI,WAAW,IAAI,yBAAyB,IAAI,KAAK;AAAA,EAC5F;AACA,SAAO;AACT;;;ACVA,eAAsB,WAAW,KAAwC;AACvE,QAAM,MAAM,MAAM,aAAa,GAAG;AAClC,QAAM,OAAO,MAAO,IAAK,KAAK;AAE9B,MAAI,CAAC,KAAK,KAAK,WAAW,QAAQ,GAAG;AACnC,UAAM,IAAI,MAAM,wEAAwE,KAAK,kBAAkB,IAAI,KAAK;AAAA,EAC1H;AACA,SAAO,cAAc,IAAI;AAC3B;;;ACTA,eAAsB,UAAa,KAAyB;AAC1D,UAAQ,MAAM,aAAa,GAAG,GAAG,KAAK;AACxC;;;ACFA,eAAsB,gBAAgB,KAAoC;AACxE,SAAO,IAAI,aAAa,OAAO,MAAM,aAAa,GAAG,GAAG,YAAY,CAAC;AACvE;;;ACFO,SAAS,cAAc,KAAsC;AAClE,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,QAAI,EAAE,eAAe;AAAO,aAAO,IAAI,MAAM,kDAAkD,CAAC;AAEhG,UAAM,QAAQG,KAAI,OAAO,EAAE,mBAAmB;AAC9C,UAAM,YAAY,MAAM,QAAQ,KAAK;AACrC,UAAM,UAAU;AAChB,UAAM,cAAc;AACpB,UAAM,QAAQ;AACd,UAAM,MAAM,IAAI,gBAAgB,GAAG;AACnC,UAAM,KAAK;AAAA,EACb,CAAC;AACH;;;ACXA,eAAsB,WAAW,KAAwC;AACvE,QAAM,MAAM,MAAM,aAAa,GAAG;AAClC,QAAM,OAAO,MAAO,IAAK,KAAK;AAE9B,MAAI,CAAC,KAAK,KAAK,WAAW,QAAQ,GAAG;AACnC,UAAM,IAAI,MAAM,wEAAwE,KAAK,kBAAkB,IAAI,KAAK;AAAA,EAC1H;AACA,SAAO,cAAc,IAAI;AAC3B;;;ACXO,SAAS,aAAa,KAAyB,kBAA0B;AAC9E,QAAM,0BAA0B,GAAG;AAEnC,MAAI,CAAC,KAAK;AACR,WAAO;AAAA,MACL,cAAc;AAAA,MACd,aAAa;AAAA,IACf;AAAA,EACF;AAEA,MAAI,QAAQ,KAAK;AACf,WAAO;AAAA,MACL,cAAc;AAAA,MACd,aAAa,IAAI;AAAA,IACnB;AAAA,EACF;AAEA,QAAM,WAAW,IAAI,WAAW,SAAS,IAAI,YAAY,IAAI,WAAW,UAAU,IAAI,aAAa;AACnG,QAAM,IAAI,QAAQ,UAAU,EAAE;AAE9B,QAAM,QAAQ,IAAI,MAAM,GAAG,EAAE,OAAO,CAAC,MAAM,CAAC;AAE5C,QAAM,eAAe,IAAI,SAAS,OAAO,IACrC,MAAM,MAAM,SAAS,KACrB;AAEJ,MAAI,eAAe,YAAY,IAAI,SAAS,OAAO,IAAI,MAAM,MAAM,GAAG,MAAM,SAAS,CAAC,IAAI,OAAO,KAAK,GAAG;AACzG,iBAAe,IAAI,WAAW,GAAG,IAAI,IAAI,iBAAiB;AAE1D,SAAO;AAAA,IACL;AAAA,IACA,aAAa,iBAAiB,MAAM,IAAI,iBAAiB,GAAG,gBAAgB;AAAA,EAC9E;AACF;;;AC5BA,eAAsB,cACpB,KACA,kBAC4B;AAC5B,QAAM,EAAE,aAAa,aAAa,IAAI,aAAa,KAAK,gBAAgB;AAExE,QAAM,WAAW,MAAM,UAAuC,WAAW;AAEzE,SAAU,WAAM,YAAY,UAAU,YAAY;AACpD;;;ACXO,SAAS,gBAAgBC,QAAoB,WAAwB,qBAAqB,OAAO;AACtG,QAAM,EAAE,OAAO,OAAO,IAAI,qBACtB,mBAAmB,SAAS,IAC5B;AACJ,EAAAA,OAAM,QAAQ;AACd,EAAAA,OAAM,SAAS;AACf,SAAO,EAAE,OAAO,OAAO;AACzB;;;ACHO,IAAe,gBAAf,MAAyC;AAAA,EAC9C,YAAY,MAAc;AAI1B,SAAU,UAAkC;AAE5C,SAAU,iBAAiC,CAAC;AAL1C,SAAK,QAAQ;AAAA,EACf;AAAA,EAQA,IAAW,SAAiC;AAAE,WAAO,KAAK;AAAA,EAAS;AAAA,EAEnE,IAAW,gBAAgC;AAAE,WAAO,KAAK;AAAA,EAAgB;AAAA,EAEzE,IAAW,WAAoB;AAAE,WAAO,CAAC,CAAC,KAAK;AAAA,EAAQ;AAAA,EAEhD,iBAAiB,WAA8B;AACpD,UAAM,EAAE,KAAK,QAAQ,IAAI,KAAK,qBAAqB,SAAS;AAC5D,WAAO,IAAI;AAAA,EACb;AAAA,EAEO,sBAAsB,WAAmBC,SAAmB;AACjE,UAAM,EAAE,KAAK,QAAQ,IAAI,KAAK,qBAAqB,SAAS;AAC5D,QAAI,SAAS,QAAQ;AACrB,QAAI,WAAWA;AAAA,EACjB;AAAA,EAEO,eAAe;AACpB,WAAO,KAAK,eAAe,IAAI,CAAC,EAAE,UAAU,OAAO;AAAA,MACjD,MAAM;AAAA,MACN,QAAQ,KAAK,iBAAiB,SAAS;AAAA,IACzC,EAAE;AAAA,EACJ;AAAA,EAEO,qBAAqB;AAC1B,WAAO,KAAK,aAAa,EAAE,OAAO,CAAC,UAAU,MAAM,kBAAqB,QAAQ;AAAA,EAClF;AAAA,EAEO,kBAAkB;AACvB,WAAO,KAAK,aAAa,EAAE,OAAO,CAAC,UAAU,EAAE,MAAM,kBAAqB,SAAS;AAAA,EACrF;AAAA,EAEO,WAAW;AAChB,SAAK,gBAAgB,EAAE,QAAQ,CAAC,EAAE,MAAM,QAAAA,QAAO,MAAM;AACnD,WAAK,sBAAsB,MAAMA,QAAO,SAAS,CAAC;AAAA,IACpD,CAAC;AAAA,EACH;AAAA,EAEO,SAAS;AACd,SAAK,mBAAmB,EAAE,QAAQ,CAAC,EAAE,MAAM,QAAQC,UAAS,MAAM;AAChE,YAAMD,UAAY,OAAOC,UAAS,SAAS,CAAC;AAC5C,MAAAA,UAAS,QAAQ;AACjB,WAAK,sBAAsB,MAAMD,OAAM;AAAA,IACzC,CAAC;AAAA,EACH;AAAA,EAEO,QAAQ,mBAAmB,MAAM;AACtC,SAAK,aAAa,EAAE,QAAQ,CAAC,UAAU;AACrC,UAAI,oBAAoB,MAAM,OAAO,YAAY;AAC/C,cAAM,IAAI,MAAM,mDAAmD,MAAM,MAAM;AAAA,MACjF;AACA,YAAM,OAAO,QAAQ;AAAA,IACvB,CAAC;AACD,SAAK,UAAU;AAAA,EACjB;AAAA,EAEO,kBAAgC;AACrC,WAAO,IAAI;AAAA,MACT,KAAK,aAAa,EACf,IAAI,CAAC,EAAE,QAAAA,QAAO,MAAM,MAAM,KAAKA,QAAO,SAAS,CAAC,CAAa,EAC7D,OAAO,CAAC,MAAM,QAAQ,KAAK,OAAO,GAAG,CAAC;AAAA,IAC3C;AAAA,EACF;AAAA,EAEA,MAAa,KAAK,cAAgE;AAChF,QAAI,wBAAwB,cAAc;AACxC,WAAK,eAAe,YAAY;AAChC;AAAA,IACF;AACA,UAAM,KAAK,YAAY,YAAY;AAAA,EACrC;AAAA,EAEA,MAAa,YAAY,KAAyB;AAChD,QAAI,OAAO,OAAO,QAAQ,UAAU;AAClC,YAAM,IAAI,MAAM,GAAG,KAAK,wCAAwC;AAAA,IAClE;AACA,UAAM,YAAY,MAAM,cAAc,KAAK,KAAK,oBAAoB,CAAC;AACrE,SAAK,kBAAkB,SAAS;AAAA,EAClC;AAAA,EAEA,MAAa,aAAa,UAA8B;AACtD,QAAI,YAAY,OAAO,aAAa,UAAU;AAC5C,YAAM,IAAI,MAAM,GAAG,KAAK,+CAA+C;AAAA,IACzE;AACA,UAAM,EAAE,SAAS,IAAIE,KAAI,OAAO;AAChC,UAAM,EAAE,aAAa,aAAa,IAAI,aAAa,UAAU,KAAK,oBAAoB,CAAC;AACvF,UAAM,uBAAuB,CAAC,cAAwB,QAAQ,IAAI,UAAU,IAAI,CAAC,OAAO,SAAS,EAAE,EAAE,KAAK,CAAC,QAAQ,IAAI,MAAM,CAAC,CAAC;AAC/H,UAAMC,eAAiB,WAAM,qBAAqB,oBAAoB;AACtE,UAAM,WAAW,KAAK,OAAO,MAAM,SAAS,WAAW,GAAG,SAAS,CAAC;AACpE,UAAM,YAAY,MAAMA,aAAY,UAAU,YAAY;AAC1D,SAAK,kBAAkB,SAAS;AAAA,EAClC;AAAA,EAEO,kBAAkB,WAA8B;AACrD,UAAM,EAAE,eAAe,OAAO,IAAI,KAAK,2BAA2B,SAAS;AAC3E,SAAK,iBAAiB;AACtB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEO,eAAe,SAAuB;AAC3C,UAAM,EAAE,eAAe,OAAO,IAAI,KAAK,cAAc,OAAO;AAC5D,SAAK,iBAAiB;AACtB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEQ,qBAAqB,WAAmB;AAC9C,QAAI,CAAC,KAAK,QAAQ;AAChB,YAAM,IAAI,MAAM,mDAAmD;AAAA,IACrE;AAEA,UAAM,SAAS,UAAU,MAAM,GAAG,EAAE,OAAO,CAAC,KAAoDC,aAAY;AAE1G,UAAI,CAAC,IAAI,QAAQ,eAAeA,QAAO,GAAG;AACxC,cAAM,IAAI,MAAM,wDAAwDA,sBAAqB,WAAW;AAAA,MAC1G;AACA,aAAO,EAAE,KAAK,IAAI,SAAS,SAAAA,UAAS,SAAS,IAAI,QAAQA,UAAS;AAAA,IACpE,GAAG,EAAE,SAAS,KAAK,OAAO,CAAC;AAE3B,UAAM,EAAE,KAAK,QAAQ,IAAI;AACzB,QAAI,CAAC,OAAO,CAAC,WAAW,EAAE,IAAI,oBAAuB,SAAS;AAC5D,YAAM,IAAI,MAAM,8DAA8D,WAAW;AAAA,IAC3F;AAEA,WAAO,EAAE,KAAK,QAAQ;AAAA,EACxB;AASF;;;ACnJO,SAAS,uBACd,GACA,QACA,QACa;AACb,SAAU,KAAK,MAAM;AACnB,QAAI,MAAS,gBAAgB,GAAG,OAAO,kBAAkB,OAAO,kBAAkB,QAAQ,MAAM;AAChG,UAAS,KAAI,KAAK,OAAO,IAAI;AAC7B,WAAO;AAAA,EACT,CAAC;AACH;;;ACRO,SAAS,YACd,GACA,kBACA,eAAe,OACF;AACb,SAAU,KAAK,MAAM;AACnB,UAAM,OAAU;AAAA,MACd,eACO;AAAA,QACA,OAAO,GAAI,iBAAiB,MAAqB,SAAS,CAAC,GAAG,CAAC,GAAG,MAAM;AAAA,QAC3E,iBAAiB,MAAM;AAAA,MACzB,IACE,uBAAuB,GAAG,iBAAiB,OAA8B,CAAC,GAAG,CAAC,CAAC;AAAA,IACrF;AACA,UAAM,OAAO,uBAAuB,MAAM,iBAAiB,OAAO,CAAC,GAAG,CAAC,CAAC;AAExE,UAAM,MAAS,KAAQ,KAAI,MAAM,IAAI,CAAC;AACtC,UAAM,OAAO,uBAAuB,KAAK,iBAAiB,OAAO,CAAC,GAAG,CAAC,CAAC;AAEvE,WAAU,KAAQ,KAAI,MAAS,KAAI,MAAM,IAAI,CAAC,CAAC;AAAA,EACjD,CAAC;AACH;AAEO,SAAS,YACd,GACA,kBACA,eAAe,OACf,cAAc,MACD;AACb,SAAU,KAAK,MAAM;AACnB,UAAM,OAAU;AAAA,MACd,eACO;AAAA,QACA,OAAO,GAAI,iBAAiB,MAAqB,SAAS,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,MAAM;AAAA,QAClG,iBAAiB,MAAM;AAAA,MACzB,IACE,uBAAuB,GAAG,iBAAiB,OAA8B,cAAc,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;AAAA,IAC5G;AACA,UAAM,OAAO,uBAAuB,MAAM,iBAAiB,OAAO,CAAC,GAAG,CAAC,CAAC;AAExE,UAAM,MAAS,KAAQ,KAAI,MAAM,IAAI,CAAC;AACtC,UAAM,OAAO,uBAAuB,KAAK,iBAAiB,OAAO,CAAC,GAAG,CAAC,CAAC;AAEvE,UAAM,MAAS,KAAQ,KAAI,MAAS,KAAI,MAAM,IAAI,CAAC,CAAC;AACpD,UAAM,OAAO,uBAAuB,KAAK,iBAAiB,OAAO,CAAC,GAAG,CAAC,CAAC;AAEvE,WAAU,KAAQ,KAAI,MAAS,KAAI,MAAS,KAAI,MAAM,IAAI,CAAC,CAAC,CAAC;AAAA,EAC/D,CAAC;AACH;;;AClDO,SAAS,UACd,GACA,QACA,UAA4B,QAC5B,WAAW,OACE;AACb,SAAU,KAAK,MAAM;AACnB,UAAM,MAAS;AAAA,MACV,OAAO,GAAG,OAAO,SAAS,CAAC,GAAG,CAAC,GAAG,OAAO;AAAA,MAC5C,OAAO;AAAA,IACT;AAEA,WAAO,WAAc,KAAK,GAAG,IAAI;AAAA,EACnC,CAAC;AACH;;;AChBO,SAAS,2BAA2B,WAAgB,eAA+B;AACxF,SAAO,KAAK,SAAS,EAAE,QAAQ,CAAC,SAAS;AACvC,QAAI,CAAC,cAAc,KAAK,CAAC,OAAO,GAAG,iBAAiB,IAAI,GAAG;AACzD,gBAAU,MAAM,QAAQ;AAAA,IAC1B;AAAA,EACF,CAAC;AACH;;;ACJO,SAAS,yBACd,gBACA,eACA;AACA,SAAO,CACL,YACA,aACA,YACA,iBACe;AACf,UAAM,UAAa;AAAA,MACjB,eAAe,aAAa,cAAc,aAAa,UAAU;AAAA,MACjE,CAAC,YAAY,YAAY,YAAY,WAAW;AAAA,IAClD;AACA,UAAM,OAAU,SAAS,eAAe,WAAW,CAAC;AAEpD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,oBAAoB;AAAA,IACtC;AAEA,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AACF;;;ACvBO,SAAS,uBACd,gBACA,eACA;AACA,SAAO,CACL,YACA,aACA,iBACa;AACb,UAAM,aAAgB,SAAS,eAAe,aAAa,WAAW,GAAG,CAAC,YAAY,WAAW,CAAC;AAClG,UAAM,UAAa,SAAS,eAAe,WAAW,CAAC;AAEvD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,oBAAoB;AAAA,IACtC;AAEA,WAAO;AAAA,MACL,SAAS;AAAA,MACT,MAAM;AAAA,IACR;AAAA,EACF;AACF;;;ACNO,IAAM,sBAAN,MAA0B;AAAA,EAE/B,YAES,kBAEA,kBAEA,MAEP;AANO;AAEA;AAEA;AAAA,EAEN;AACL;;;AC3BO,SAAS,kCACd,gBACA,eACA;AACA,SAAO,CAAC,YAAoB,aAAqB,iBAA8C;AAC7F,UAAM,mBAAsB,SAAS,eAAe,IAAI,IAAI,UAAU,GAAG,CAAC,GAAG,GAAG,YAAY,CAAC,CAAC;AAC9F,UAAM,mBAAsB,SAAS,eAAe,aAAa,WAAW,GAAG,CAAC,GAAG,GAAG,YAAY,WAAW,CAAC;AAC9G,UAAM,OAAU,SAAS,eAAe,WAAW,CAAC;AAEpD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,gCAAgC;AAAA,MAChD,EAAE,WAAW,GAAG,gCAAgC;AAAA,MAChD,EAAE,WAAW,GAAG,oBAAoB;AAAA,IACtC;AAEA,WAAO,IAAI;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;AAEO,SAAS,+BAEd,oBACA;AACA,SAAO,CAAC,WAAwC;AAC9C,UAAM,mBAAmB,mBAAgC,GAAG,2BAA2B,CAAC;AACxF,UAAM,mBAAmB,mBAAgC,GAAG,2BAA2B,CAAC;AACxF,UAAM,OAAO,mBAAgC,GAAG,eAAe,CAAC;AAEhE,WAAO,IAAI;AAAA,MACT;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACF;;;ACvCO,SAAS,0BAA0B,WAAgB,eAA+B;AACvF,SAAO,CAAC,cAAsB,WAAmB,eAAwB;AACvE,UAAMC,UAAS,UAAU;AAEzB,QAAI,CAAC,SAASA,SAAQ,SAAS,GAAG;AAChC,YAAM,IAAI,MAAM,sBAAsB,+BAA+B,4BAA4BA,SAAQ;AAAA,IAC3G;AAEA,kBAAc;AAAA,MACZ,EAAE,cAAc,WAAW,cAAc,aAAa;AAAA,IACxD;AAEA,WAAOA;AAAA,EACT;AACF;;;ACjBO,SAAS,sBAAsB,SAAuB;AAC3D,MAAI,mBAAmB;AAEvB,WAAS,eAAe,YAAkC;AACxD,UAAM,MAAM,iBAAiB,MAAM,GAAG,UAAU;AAChD,uBAAmB,iBAAiB,MAAM,UAAU;AACpD,WAAO;AAAA,EACT;AAEA,WAAS,sBAAoC;AAC3C,WAAO;AAAA,EACT;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;;;ACdO,SAAS,kBAAkB,gBAAwC,eAA+B;AACvG,QAAM,oBAAoB,yBAAyB,gBAAgB,aAAa;AAChF,QAAM,6BAA6B,kCAAkC,gBAAgB,aAAa;AAElG,WAAS,yBAAyB,YAAoB,aAAqB,cAAsB,eAAe,OAA0B;AACxI,UAAM,QAAQ,eACV,kBAAkB,YAAY,aAAa,GAAG,GAAG,oBAAoB,IACrE,2BAA2B,YAAY,aAAa,GAAG,oBAAoB;AAC/E,UAAM,QAAQ,2BAA2B,aAAa,aAAa,GAAG,oBAAoB;AAC1F,UAAMC,SAAQ,2BAA2B,aAAa,aAAa,GAAG,oBAAoB;AAE1F,WAAO,EAAE,OAAO,OAAO,OAAAA,OAAM;AAAA,EAC/B;AAEA,WAAS,yBAAyB,YAAoB,aAAqB,cAAsB,eAAe,OAA0B;AACxI,UAAM,EAAE,OAAO,OAAO,OAAAA,OAAM,IAAI,yBAAyB,YAAY,aAAa,cAAc,YAAY;AAC5G,UAAM,QAAQ,2BAA2B,aAAa,aAAa,GAAG,oBAAoB;AAE1F,WAAO;AAAA,MACL;AAAA,MAAO;AAAA,MAAO,OAAAA;AAAA,MAAO;AAAA,IACvB;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;;;AC1BO,SAAS,cAAc,SAA8F;AAC1H,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM;AAAA,IACJ;AAAA,EACF,IAAI,kBAAkB,gBAAgB,aAAa;AAEnD,QAAM,SAAS,yBAAyB,GAAG,IAAI,UAAU,IAAI;AAC7D,QAAM,SAAS,yBAAyB,IAAI,IAAI,QAAQ;AACxD,QAAM,SAAS,yBAAyB,IAAI,KAAK,QAAQ;AACzD,QAAM,SAAS,yBAAyB,KAAK,KAAK,QAAQ;AAE1D,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,QAAQ;AAAA,MACN;AAAA,MAAQ;AAAA,MAAQ;AAAA,MAAQ;AAAA,IAC1B;AAAA,EACF;AACF;;;AC1BO,SAAS,sBAAsB,oBAAuE;AAC3G,SAAO,CAAC,WAA+B;AACrC,UAAM,UAAU,mBAAgC,GAAG,kBAAkB,CAAC;AACtE,UAAM,OAAO,mBAAgC,GAAG,eAAe,CAAC;AAEhE,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AACF;;;ACRO,SAAS,kBAAkB,WAAgB,eAA+B;AAC/E,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,QAAM,oBAAoB,sBAAsB,kBAAkB;AAClE,QAAM,6BAA6B,+BAA+B,kBAAkB;AAEpF,WAAS,yBAAyB,QAAgB,eAAe,OAA0B;AACzF,UAAM,QAAQ,eACV,kBAAkB,GAAG,cAAc,IACnC,2BAA2B,GAAG,cAAc;AAChD,UAAM,QAAQ,2BAA2B,GAAG,cAAc;AAC1D,UAAMC,SAAQ,2BAA2B,GAAG,cAAc;AAE1D,WAAO,EAAE,OAAO,OAAO,OAAAA,OAAM;AAAA,EAC/B;AAEA,WAAS,yBAAyB,QAAgB,eAAe,OAA0B;AACzF,UAAM,QAAQ,eACV,kBAAkB,GAAG,cAAc,IACnC,2BAA2B,GAAG,cAAc;AAChD,UAAM,QAAQ,2BAA2B,GAAG,cAAc;AAC1D,UAAMA,SAAQ,2BAA2B,GAAG,cAAc;AAC1D,UAAM,QAAQ,2BAA2B,GAAG,cAAc;AAE1D,WAAO;AAAA,MACL;AAAA,MAAO;AAAA,MAAO,OAAAA;AAAA,MAAO;AAAA,IACvB;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;;;AC/BO,SAAS,2BACd,WACuE;AACvE,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,EACF,IAAI,kBAAkB,WAAW,aAAa;AAE9C,QAAM,SAAS;AAAA,IACb,QAAQ,yBAAyB,UAAU,IAAI;AAAA,IAC/C,QAAQ,yBAAyB,QAAQ;AAAA,IACzC,QAAQ,yBAAyB,QAAQ;AAAA,IACzC,QAAQ,yBAAyB,QAAQ;AAAA,EAC3C;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACfO,IAAM,uBAAN,cAAmC,cAAuG;AAAA,EAC/I,cAAc;AACZ,UAAM,sBAAsB;AAAA,EAC9B;AAAA,EAEO,aAAaC,QAA8B;AAChD,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,oDAAoD;AAAA,IACtE;AAEA,WAAU,KAAK,MAAM;AACnB,YAAM,cAAiB,KAAKA,OAAM,cAAc,KAAK,IAAI,GAAG,SAAS;AACrE,YAAM,UAAU,CAAC,SAAS,SAAS,OAAO;AAC1C,YAAM,aAAa,UAAU,aAAa,OAAO,EAAE,IAAI,GAAG;AAE1D,UAAI,MAAM,YAAY,YAAY,OAAO,QAAQ,IAAI;AACrD,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,OAAO;AAE7C,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAO,2BAA2B,SAAS;AAAA,EAC7C;AAAA,EAEU,cAAc,SAAuB;AAC7C,WAAO,cAAc,OAAO;AAAA,EAC9B;AACF;;;AChDO,SAAS,oBACd,GACA,QACa;AACb,SAAU,KAAK,MAAS;AAAA,IACnB,OAAO,GAAG,OAAO,OAAO;AAAA,IAC3B,OAAO;AAAA,EACT,CAAC;AACH;;;ACTO,SAASC,eAAc,SAAuB,YAAoB,aAA2E;AAClJ,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM,kBAAkB,uBAAuB,gBAAgB,aAAa;AAE5E,QAAM,KAAK,gBAAgB,YAAY,aAAa,IAAI;AAExD,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,QAAQ,EAAE,GAAG;AAAA,EACf;AACF;;;AClBO,SAASC,4BACd,WACsD;AACtD,QAAM,gBAAgC,CAAC;AAEvC,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,WAAS,gBAAgB,QAA0B;AACjD,UAAM,UAAU,mBAAmB,GAAG,kBAAkB,CAAC;AACzD,UAAM,OAAO,mBAAmB,GAAG,eAAe,CAAC;AACnD,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,QAAM,SAAS;AAAA,IACb,IAAI,gBAAgB,IAAI;AAAA,EAC1B;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACvBO,SAAS,mBAAmB,WAA8B;AAC/D,QAAM,sBAAyC,CAAC;AAChD,QAAM,gBAAmC,CAAC;AAE1C,SAAO,KAAK,SAAS,EAAE,QAAQ,CAAC,QAAQ;AACtC,UAAM,MAAM,IAAI,WAAW,IAAI,IAAI,gBAAgB;AACnD,QAAI,OAAO,UAAU;AAAA,EACvB,CAAC;AAED,SAAO,EAAE,qBAAqB,cAAc;AAC9C;;;ACDO,IAAe,gBAAf,cAGG,cAAyB;AAAA,EAGjC,YAAY,OAAe,sBAA+D;AACxF,UAAM,KAAK;AACX,SAAK,wBAAwB;AAAA,EAC/B;AAAA,EAEA,IAAW,uBAAgE;AACzE,WAAO,KAAK;AAAA,EACd;AAAA,EAQO,OAAOC,QAA4C;AACxD,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,GAAG,KAAK,qCAAqC;AAAA,IAC/D;AAEA,WAAU,KAAK,MAAM;AACnB,YAAM,qBAAqBA,kBAAiB,WACxC,KAAK,qBAAqB,aAAaA,MAAK,IAC5CA;AACJ,aAAO,oBAAoB,mBAAmB,KAAK,mBAAmB,MAAM,IAAI,EAAE,GAAG,OAAO,EAAE;AAAA,IAChG,CAAC;AAAA,EACH;AAAA,EAEgB,QAAQ,mBAAmB,MAAM;AAC/C,SAAK,qBAAqB,QAAQ,gBAAgB;AAClD,UAAM,QAAQ,gBAAgB;AAAA,EAChC;AAAA,EAEO,qBAAqB,SAAuB;AACjD,UAAM,EAAE,QAAQ,cAAc,IAAI,KAAK,wBAAwB,OAAO;AACtE,SAAK,UAAU;AACf,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEO,wBAAwB,SAAuB;AACpD,WAAOC,eAAc,SAAS,KAAK,wBAAwB,GAAG,KAAK,yBAAyB,CAAC;AAAA,EAC/F;AAAA,EAEU,2BAA2B,WAA8B;AACjE,UAAM,EAAE,qBAAqB,cAAc,IAAI,mBAAmB,SAAS;AAE3E,SAAK,qBAAqB,kBAAkB,mBAAmB;AAE/D,WAAOC,4BAA2B,aAAa;AAAA,EACjD;AAAA,EAEU,cAAc,SAAuB;AAC7C,UAAM,MAAM,KAAK,wBAAwB;AACzC,UAAM,OAAO,KAAK,yBAAyB;AAC3C,UAAM,uBAAwB,OAAO,MAAO;AAE5C,UAAM,0BAA0B,QAAQ,MAAM,GAAG,QAAQ,SAAS,oBAAoB;AACtF,UAAM,oBAAoB,QAAQ,MAAM,QAAQ,SAAS,oBAAoB;AAE7E,SAAK,qBAAqB,eAAe,uBAAuB;AAChE,WAAO,KAAK,wBAAwB,iBAAiB;AAAA,EACvD;AACF;;;ACjFO,IAAM,yBAAyB,CAAC,WAAW,SAAS,OAAO,SAAS,WAAW,aAAa,WAAW;AAEvG,IAAM,kBAAN,MAAsB;AAAA,EAS3B,YAAY,eAAwC;AARpD,SAAO,UAAU;AACjB,SAAO,QAAQ;AACf,SAAO,MAAM;AACb,SAAO,QAAQ;AACf,SAAO,UAAU;AACjB,SAAO,YAAY;AACnB,SAAO,YAAY;AAGjB,QAAI,cAAc,WAAW,GAAG;AAC9B,YAAM,IAAI,MAAM,8EAA8E,cAAc,QAAQ;AAAA,IACtH;AAEA,2BAAuB,QAAQ,CAAC,YAAY,QAAQ;AAClD,WAAK,cAAc,cAAc;AAAA,IACnC,CAAC;AAAA,EACH;AAAA,EAEA,gBAAgB;AACd,WAAO,uBACJ,IAAI,CAAC,gBAAgB,EAAE,YAAY,aAAa,KAAK,YAAsB,EAAE,EAC7E,KAAK,CAAC,IAAI,OAAO,GAAG,cAAc,GAAG,WAAW;AAAA,EACrD;AACF;;;AClBO,IAAM,oBAAN,cAAgC,cAA0C;AAAA,EAC/E,YAAY,uBAA6C,IAAI,qBAAqB,GAAG;AACnF,UAAM,qBAAqB,oBAAoB;AAAA,EACjD;AAAA,EAEO,aAAaC,QAA4C;AAC9D,WAAU,KAAK,MAAS,QAAQ,KAAK,OAAOA,MAAK,CAAC,CAAC;AAAA,EACrD;AAAA,EAEA,MAAa,QAAQA,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEA,MAAa,mBAAmBA,QAAkB;AAChD,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,MAAM,MAAM,KAAK,aAAa,QAAQ;AAC5C,UAAM,sBAAsB,MAAM,QAAQ,IAAO,QAAQ,GAAG,EAAE,IAAI,OAAO,MAAM;AAC7E,YAAM,OAAO,EAAE,SAAS;AACxB,QAAE,QAAQ;AACV,aAAO;AAAA,IACT,CAAC,CAAC;AACF,QAAI,QAAQ;AAEZ,UAAM,qBAAqB,oBACxB,IAAI,CAAC,iBAAiB,IAAI,gBAAgB,YAA4B,CAAC;AAE1E,WAAO,SAAS,eACZ,qBACA,mBAAmB;AAAA,EACzB;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,0BAAkC;AAC1C,WAAO;AAAA,EACT;AAAA,EAEU,2BAAmC;AAC3C,WAAO;AAAA,EACT;AACF;;;AC9CO,SAAS,sBAAsB,KAA0C;AAC9E,SAAO,IAAI,uBAAuB;AACpC;AAEO,SAAS,0BAAmC,WAAoB,aAA4D;AACjI,QAAM,YAAY,EAAE,YAAY;AAChC,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACFO,SAAS,oBAAoB,WAAuC,iBAA6E,gBAAgB,KAAK,iBAA0B;AACrM,QAAM,uBAAuB,MAAM,QAAQ,eAAe,IAAI,kBAAkB,CAAC,eAAe;AAEhG,uBAAqB,QAAQ,CAAC,MAAM;AAElC,UAAM,OAAO,aAAa,kBACtB,IACC,sBAAsB,CAAC,IAAI,EAAE,cAAc;AAChD,QAAI,CAAC,MAAM;AACT,YAAM,IAAI,MAAM,iHAAiH;AAAA,IACnI;AAEA,UAAM,SAAS,KAAK,cAAc;AAClC,UAAM,mBAAmB,OAAO,OAAO,CAAC,cAAc,UAAU,cAAc,aAAa;AAE3F,UAAM,SAAS,oBAAoB,CAAC,IAChC,EAAE,UAAU,IAAI,aACf,mBAAmB,IAAI,MAAM,GAAG,CAAC;AAEtC,UAAM,gBAAgB,IAAI;AAAA,MACxB,iBAAiB,IAAI,CAAC,cAAc,GAAG,UAAU,eAAeC,OAAM,UAAU,WAAW,IAAI;AAAA,MAC/F;AAAA,IACF;AACA,kBAAc,KAAK,SAAS;AAAA,EAC9B,CAAC;AACH;;;ACfO,SAAS,oBACd,KACgE;AAChE,SACE,oBAAoB,GAAG,KAEpB,IAAI,wBAAwB,iBAE5B,IAAI,iCAAiC,iBAErC,IAAI,0BAA0B;AAErC;AAEA,SAAS,mBAAmB,MAAM;AAGhC,QAAM,UAAU,CAAC,YAAa,UAAU,MAAO,KAAK;AACpD,QAAM,6BAA6B,CAAC,GAAG,MAAM,KAAK,MAAM,EAAE,KAAK,EAAE,OAAO,KAAK,EAAE,KAAK,EAAE,OAAO,CAAC;AAE9F,QAAM,QAAQ;AAAA,IACZ,MAA0B;AAAA,IAC1B,OAA2B;AAAA,IAC3B,KAAyB;AAAA,EAC3B;AAEA,QAAM,UAAU,CAAC,WAAW,UAAU,eAAe;AAEnD,UAAM,iBAAiB,KAAK,MAAM,UAAU,KAAK,SAAS,EAAE;AAE5D,UAAM,kBAAkB,KAAK,MAAM,SAAS,KAAK,WAAW,EAAE;AAE9D,WAAO,iBAAiB;AAAA,EAC1B;AAEA,QAAM,WAAW,CAAC,OAAO,UAAU;AAIjC,UAAM,aAAa,KAAK,MAAM,MAAM,KAAK,MAAM,IAAI,MAAM,KAAK,MAAM,EAAE;AACtE,UAAM,WAAW,MAAM,KAAK,MAAM;AAClC,UAAM,iBAAiB,KAAK,KAAK,WAAW,UAAU;AACtD,UAAM,iBAAiB,QAAQ,cAAc;AAC7C,UAAM,iBAAiB,KAAK,MAAM,KAAK,cAAc;AAGrD,UAAM,gBAAgB,MAAM,KAAK,MAAM,KAAK,IAAI,KAAK;AACrD,UAAM,SAAS,iBAAiB;AAChC,WAAO;AAAA,EACT;AAEA,QAAM,YAAY,CAAC,WAAW,UAAU,eAAe;AAKrD,UAAM,OAAO,2BAA2B,WAAW,UAAU;AAE7D,UAAM,aAAa;AAAA,MACjB,KAAK,UAAU,KAAK,WAAW,MAAM;AAAA,MACrC,KAAK,UAAU,KAAK,WAAW,MAAM;AAAA,IACvC;AACA,UAAM,kBAAkB,2BAA2B,UAAU,UAAU;AACvE,UAAM,iBAAiB,KAAK,KAAK,kBAAkB,IAAI;AACvD,UAAM,iBAAiB,KAAK,MAAM,QAAQ,cAAc,CAAC;AAGzD,UAAM,YAAY,WAAW,KAAK,SAAS,KAAK,IAAI,KAAK;AACzD,UAAM,SAAS,iBAAiB;AAChC,WAAO;AAAA,EACT;AAEA,MAAI,CAAC,QAAQ,CAAC,KAAK,cAAc,KAAK,WAAW,WAAW;AAAI,WAAO;AACvE,QAAM,KAAK,KAAK;AAChB,QAAM,OAAO,SAAS,GAAG,KAAK,GAAG,GAAG;AACpC,QAAM,QAAQ,UAAU,GAAG,KAAK,GAAG,KAAK,GAAG,EAAE;AAC7C,QAAM,MAAM,QAAQ,GAAG,KAAK,GAAG,KAAK,GAAG,EAAE;AACzC,SAAO;AACT;AAEO,SAAS,wBACd,WACA,oBAC4C;AAC5C,QAAM,EAAE,KAAK,MAAM,IAAI,UAAU;AACjC,QAAM,YAAY,mBAAmB,QAAwB,MAAM,GAAG,MAAM,CAAC;AAC7E,QAAM,OAAO,UAAU,MAAM;AAC7B,QAAM,EAAE,UAAU,IAAI,UAAU;AAChC,QAAM,cAAc,IAAI;AAAA,IACtB,UAAU,UAAU;AAAA,IACpB,KAAK,QAAQ,UAAU,QAAQ,CAAC;AAAA,IAChC;AAAA,EACF;AACA,QAAM,QAAQ,mBAAmB,kBAAkB;AACnD,QAAM,YAAY,EAAE,WAAW,oBAAoB,aAAa,MAAM;AACtE,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACjGO,IAAM,2BAAN,MAA+B;AAAA,EAapC,YAAY,UAAqC,CAAC,GAAG;AACnD,UAAM;AAAA,MACJ,YAAY;AAAA,MAAM,aAAa;AAAA,MAAM;AAAA,MAAW;AAAA,MAAW;AAAA,MAAW;AAAA,IACxE,IAAI;AACJ,SAAK,YAAY;AACjB,SAAK,aAAa;AAClB,SAAK,YAAY,aAAa;AAC9B,SAAK,YAAY,aAAa;AAC9B,SAAK,YAAY,aAAa;AAC9B,SAAK,aAAa,cAAc;AAAA,EAClC;AACF;AAEO,IAAM,oBAAN,MAAwB;AAAA,EAK7B,YACE,eACA,UAAqC,CAAC,GACtC;AACA,SAAK,gBAAgB;AACrB,SAAK,UAAU,IAAI,yBAAyB,OAAO;AAAA,EACrD;AAAA,EAEA,KAAK,WAAkE;AACrE,UAAM,MAAM,oBAAoB,SAAS;AAEzC,UAAM;AAAA,MACJ;AAAA,MAAW;AAAA,MAAY;AAAA,MAAW;AAAA,MAAW;AAAA,MAAW;AAAA,IAC1D,IAAI,KAAK;AAET,QAAI,aAAa,KAAK,yBAAyB,iBAAiB;AAC9D,UAAI,cAAc;AAClB,UAAI,YAAY;AAChB,kBAAY,KAAK,KAAK,cAAc,cAAc,CAAC;AACnD,kBAAY,KAAK,KAAK,cAAc,eAAe,CAAC;AACpD,kBAAY,KAAK,KAAK,cAAc,gBAAgB,CAAC;AACrD,kBAAY,KAAK,KAAK,cAAc,QAAQ,CAAC;AAC7C,kBAAY,KAAK,KAAK,cAAc,WAAW,GAAG,IAAI;AACtD,kBAAY,KAAK,KAAK,cAAc,YAAY,GAAG,IAAI;AACvD,kBAAY,KAAK,KAAK,cAAc,SAAS,GAAG,IAAI;AAAA,IACtD;AAEA,QAAI,YAAY;AACd,UAAI,cAAc;AAClB,UAAI,YAAY;AAEhB,YAAM,YAAY,CAAC,OAAe;AAChC,YAAI,UAAU;AACd,YAAI,IAAI,GAAG,GAAG,GAAG,GAAG,WAAW,GAAG,IAAI,KAAK,EAAE;AAC7C,YAAI,KAAK;AAAA,MACX;AACA,WAAK,cAAc,UAAU,QAAQ,SAAS;AAAA,IAChD;AAAA,EACF;AACF;AAIO,SAAS,kBACd,WACA,eACA;AACA,QAAM,qBAAqB,MAAM,QAAQ,aAAa,IAAI,gBAAgB,CAAC,aAAa;AACxF,qBAAmB,QAAQ,CAAC,MAAM;AAEhC,UAAM,YAAY,aAAa,gBAC3B,IACC,oBAAoB,CAAC,IAAI,EAAE,YAAY;AAC5C,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,8HAA8H;AAAA,IAChJ;AAEA,QAAI,kBAAkB,SAAS,EAAE,KAAK,SAAS;AAAA,EACjD,CAAC;AACH;;;;;;ACvGA,SAASC,mBAAkB,gBAAwC,eAA+B;AAChG,QAAM,oBAAoB,yBAAyB,gBAAgB,aAAa;AAChF,QAAM,6BAA6B,kCAAkC,gBAAgB,aAAa;AAElG,WAAS,4BAA4B,YAAoB,aAAqB,cAA4C;AACxH,UAAM,kBAAkB,2BAA2B,YAAY,aAAa,GAAG,8BAA8B;AAC7G,UAAM,kBAAkB,2BAA2B,aAAa,aAAa,GAAG,8BAA8B;AAC9G,UAAM,iBAAiB,kBAAkB,YAAY,aAAa,GAAG,GAAG,6BAA6B;AAErG,WAAO,EAAE,iBAAiB,iBAAiB,eAAe;AAAA,EAC5D;AAEA,WAAS,uBAAuB,UAAkB,cAAuC;AACvF,UAAM,kBAAkB,2BAA2B,UAAU,UAAU,GAAG,8BAA8B;AACxG,UAAM,kBAAkB,2BAA2B,UAAU,UAAU,GAAG,8BAA8B;AACxG,UAAM,kBAAkB,2BAA2B,UAAU,UAAU,GAAG,8BAA8B;AAExG,WAAO,EAAE,iBAAiB,iBAAiB,gBAAgB;AAAA,EAC7D;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,eAAc,SAAuB,eAAsF;AACzI,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,IAAID,mBAAkB,gBAAgB,aAAa;AAEnD,QAAM,qBAAqB,kBAAkB,GAAG,IAAI,GAAG,oBAAoB;AAC3E,QAAM,+BAA+B,4BAA4B,IAAI,IAAI,8BAA8B;AACvG,QAAM,+BAA+B,4BAA4B,IAAI,KAAK,8BAA8B;AAExG,QAAM,aAAa;AAAA,IACjB,SAAS;AAAA,IACT,mBAAmB;AAAA,IACnB,mBAAmB;AAAA,EACrB;AAEA,QAAM,cAAc,CAAC;AACrB,EAAAE,OAAM,eAAe,GAAG,CAAC,EAAE,QAAQ,CAAC,QAAQ;AAC1C,gBAAY,cAAc,SAAS,uBAAuB,KAAK,0BAA0B,KAAK;AAAA,EAChG,CAAC;AAED,QAAM,4BAA4B,4BAA4B,KAAK,KAAK,2BAA2B;AACnG,QAAM,2BAA2B,2BAA2B,KAAK,KAAK,0BAA0B;AAEhG,QAAM,YAAY;AAAA,IAChB,iBAAiB;AAAA,IACjB,gBAAgB;AAAA,EAClB;AAEA,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,QAAQ,EAAE,YAAY,aAAa,UAAU;AAAA,EAC/C;AACF;;;ACxEA,SAASC,mBAAkB,WAAgB,eAA+B;AACxE,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,QAAM,oBAAoB,sBAAsB,kBAAkB;AAClE,QAAM,6BAA6B,+BAA+B,kBAAkB;AAEpF,WAAS,4BAA4B,cAA4C;AAC/E,UAAM,kBAAkB,2BAA2B,GAAG,8BAA8B;AACpF,UAAM,kBAAkB,2BAA2B,GAAG,8BAA8B;AACpF,UAAM,iBAAiB,kBAAkB,GAAG,6BAA6B;AAEzE,WAAO,EAAE,iBAAiB,iBAAiB,eAAe;AAAA,EAC5D;AAEA,WAAS,uBAAuB,cAAuC;AACrE,UAAM,kBAAkB,2BAA2B,GAAG,8BAA8B;AACpF,UAAM,kBAAkB,2BAA2B,GAAG,8BAA8B;AACpF,UAAM,kBAAkB,2BAA2B,GAAG,8BAA8B;AAEpF,WAAO,EAAE,iBAAiB,iBAAiB,gBAAgB;AAAA,EAC7D;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,4BACd,WACA,eAC+D;AAC/D,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,IAAID,mBAAkB,WAAW,aAAa;AAE9C,QAAM,qBAAqB,kBAAkB,oBAAoB;AACjE,QAAM,+BAA+B,4BAA4B,8BAA8B;AAC/F,QAAM,+BAA+B,4BAA4B,8BAA8B;AAE/F,QAAM,aAAa;AAAA,IACjB,SAAS;AAAA,IACT,mBAAmB;AAAA,IACnB,mBAAmB;AAAA,EACrB;AAEA,QAAM,cAAc,CAAC;AACrB,EAAAE,OAAM,eAAe,GAAG,CAAC,EAAE,QAAQ,CAAC,QAAQ;AAC1C,gBAAY,cAAc,SAAS,uBAAuB,0BAA0B,KAAK;AAAA,EAC3F,CAAC;AAED,QAAM,4BAA4B,4BAA4B,2BAA2B;AACzF,QAAM,2BAA2B,2BAA2B,0BAA0B;AAEtF,QAAM,YAAY;AAAA,IAChB,iBAAiB;AAAA,IACjB,gBAAgB;AAAA,EAClB;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,EAAE,YAAY,aAAa,UAAU,GAAG,cAAc;AACzE;;;ACjEA,SAAS,KAAK,GAAgB,QAAoB,QAAuC;AACvF,SAAU,KAAO,OAAO,GAAG,OAAO,SAAS,QAAQ,MAAM,GAAG,OAAO,IAAI;AACzE;AAEA,SAAS,eAAe,GAAgB,QAA8B,kBAAkB,MAAmB;AACzG,MAAI,MAAM,kBAAqB,KAAK,CAAC,IAAI;AACzC,QAAM,uBAAuB,KAAK,OAAO,iBAAiB,CAAC,GAAG,CAAC,CAAC;AAChE,QAAM,uBAA0B,KAAK,GAAG,GAAG,OAAO,iBAAiB,CAAC,GAAG,CAAC,CAAC;AACzE,QAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,QAAS,KAAI,KAAK,KAAK,GAAG,OAAO,gBAAgB,CAAC,GAAG,CAAC,CAAC,CAAC;AACxD,SAAO;AACT;AAEA,SAAS,UAAU,GAAgB,QAAsC;AACvE,MAAI,MAAM,uBAA0B,KAAK,CAAC,GAAG,OAAO,iBAAiB,CAAC,GAAG,CAAC,CAAC;AAC3E,QAAM,uBAA0B,KAAK,GAAG,GAAG,OAAO,iBAAiB,CAAC,GAAG,CAAC,CAAC;AACzE,QAAM,uBAA0B,KAAK,GAAG,GAAG,OAAO,iBAAiB,CAAC,GAAG,CAAC,CAAC;AACzE,QAAS,KAAI,KAAK,CAAC;AACnB,SAAO;AACT;AAEO,IAAM,eAAN,cAA2B,cAAkC;AAAA,EAGlE,YAAY,eAAuB;AACjC,UAAM,cAAc;AACpB,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEO,aAAaC,QAA8B;AAChD,UAAM,EAAE,OAAO,IAAI;AACnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,4CAA4C;AAAA,IAC9D;AACA,WAAU,KAAK,MAAM;AACnB,YAAM,cAAiB,KAAKA,OAAM,cAAc,KAAK,IAAI,GAAG,SAAS;AACrE,YAAM,UAAU,CAAC,SAAS,SAAS,OAAO;AAC1C,YAAM,aAAa,UAAU,aAAa,OAAO,EAAE,IAAI,GAAG;AAC1D,UAAI,MAAS,KAAK,KAAK,YAAY,OAAO,WAAW,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC;AACrE,YAAM,eAAe,KAAK,OAAO,WAAW,mBAAmB,KAAK;AACpE,YAAM,eAAe,KAAK,OAAO,WAAW,iBAAiB;AAC7D,MAAAC,OAAM,KAAK,gBAAgB,GAAG,CAAC,EAAE,QAAQ,CAAC,QAAQ;AAChD,cAAM,UAAU,KAAK,OAAO,YAAY,cAAc,MAAM;AAAA,MAC9D,CAAC;AACD,YAAM,eAAe,KAAK,OAAO,UAAU,eAAe;AAC1D,YAAS,KAAK,uBAAuB,KAAK,OAAO,UAAU,gBAAgB,CAAC,GAAG,CAAC,CAAC,CAAC;AAClF,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQD,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAOE,4BAA2B,WAAW,KAAK,cAAc;AAAA,EAClE;AAAA,EAEU,cAAc,SAAuB;AAC7C,WAAOC,eAAc,SAAS,KAAK,cAAc;AAAA,EACnD;AACF;;;ACzEO,SAASC,eAAc,SAA6E;AACzG,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM,kBAAkB,uBAAuB,gBAAgB,aAAa;AAE5E,QAAM,MAAM,gBAAgB,KAAK,GAAG,QAAQ;AAC5C,QAAM,SAAS,gBAAgB,KAAK,GAAG,WAAW;AAElD,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,QAAQ,EAAE,IAAI,EAAE,KAAK,OAAO,EAAE;AAAA,EAChC;AACF;;;ACnBO,SAASC,4BACd,WACsD;AACtD,QAAM,gBAAgC,CAAC;AAEvC,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,WAAS,gBAAgB,QAA0B;AACjD,UAAM,UAAU,mBAAmB,GAAG,kBAAkB,CAAC;AACzD,UAAM,OAAO,mBAAmB,GAAG,eAAe,CAAC;AACnD,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,QAAM,SAAS;AAAA,IACb,IAAI;AAAA,MACF,KAAK,gBAAgB,QAAQ;AAAA,MAC7B,QAAQ,gBAAgB,WAAW;AAAA,IACrC;AAAA,EACF;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACvBO,IAAK,SAAL,kBAAKC,YAAL;AAEL,EAAAA,QAAA,YAAS;AAET,EAAAA,QAAA,UAAO;AAJG,SAAAA;AAAA,GAAA;;;ACKL,IAAM,eAAN,cAA2B,cAAyB;AAAA,EAGzD,YAAY,uBAAqC,IAAI,aAAa,CAAC,GAAG;AACpE,UAAM,cAAc;AACpB,SAAK,wBAAwB;AAAA,EAC/B;AAAA,EAEA,IAAW,uBAAqC;AAC9C,WAAO,KAAK;AAAA,EACd;AAAA,EAEO,OAAOC,QAA0C;AACtD,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,GAAG,KAAK,qCAAqC;AAAA,IAC/D;AAEA,WAAU,KAAK,MAAM;AACnB,YAAM,qBAAqBA,kBAAiB,WACxC,KAAK,qBAAqB,aAAaA,MAAK,IAC5CA;AAEJ,YAAM,SAAY,QAAQ,oBAAoB,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,OAAO,EAAE,KAAK,mBAAmB,MAAM,IAAI,EAAE;AAC3G,YAAM,MAAM,oBAAoB,QAAQ,OAAO,GAAG,GAAG,EAAE,KAAK;AAC5D,YAAM,SAAS,oBAAoB,QAAQ,OAAO,GAAG,MAAM;AAC3D,aAAO,EAAE,KAAK,OAAO;AAAA,IACvB,CAAC;AAAA,EACH;AAAA,EAEO,aAAaA,QAA0C;AAC5D,WAAU,KAAK,MAAM;AACnB,YAAM,EAAE,KAAK,OAAO,IAAI,KAAK,OAAOA,MAAK;AACzC,aAAO,EAAE,KAAK,QAAW,QAAQ,MAAM,EAAE;AAAA,IAC3C,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAsC;AACzD,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEA,MAAa,oBAAoBA,QAA8E;AAC7G,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,MAAM,MAAM,KAAK,aAAa,QAAQ;AAE5C,UAAM,OAAU,QAAQ,IAAI,GAAG;AAC/B,UAAM,UAAa,QAAQ,IAAI,MAAM;AACrC,UAAM,sBAAsB,KAAK,IAAI,CAAC,WAAW,OAAO;AAAA,MACtD;AAAA,MACA,cAAc,QAAQ;AAAA,IACxB,EAAE;AAEF,UAAM,qBAAqB,MAAM,QAAQ;AAAA,MACvC,oBAAoB,IAAI,OAAO,EAAE,WAAW,aAAa,MAAM;AAC7D,cAAM,MAAO,UAAU,SAAS,EAAG;AACnC,cAAM,WAAY,aAAa,SAAS,EAAG;AAC3C,cAAM,SAAS,WAAW;AAC1B,cAAM,SAAS;AACf,cAAM,oBAAoB,SAAS,WAAY,IAAI;AAEnD,kBAAU,QAAQ;AAClB,qBAAa,QAAQ;AACrB,eAAO,EAAE,KAAK,QAAQ,kBAAkB;AAAA,MAC1C,CAAC;AAAA,IACH;AACA,QAAI,IAAI,QAAQ;AAChB,QAAI,OAAO,QAAQ;AAEnB,WAAO,SAAS,eAAe,qBAAiD,mBAAmB;AAAA,EACrG;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEgB,QAAQ,mBAAmB,MAAM;AAC/C,SAAK,qBAAqB,QAAQ,gBAAgB;AAClD,UAAM,QAAQ,gBAAgB;AAAA,EAChC;AAAA,EAEO,qBAAqB,SAAuB;AACjD,UAAM,EAAE,QAAQ,cAAc,IAAI,KAAK,wBAAwB,OAAO;AACtE,SAAK,UAAU;AACf,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEO,wBAAwB,SAAuB;AACpD,WAAOC,eAAc,OAAO;AAAA,EAC9B;AAAA,EAEU,2BAA2B,WAA8B;AACjE,UAAM,EAAE,qBAAqB,cAAc,IAAI,mBAAmB,SAAS;AAE3E,SAAK,qBAAqB,kBAAkB,mBAAmB;AAE/D,WAAOC,4BAA2B,aAAa;AAAA,EACjD;AAAA,EAEU,cAAc,SAAuB;AAC7C,UAAM,uBAAwB,MAAM,IAAI,KAAM,MAAM,IAAI;AAExD,UAAM,0BAA0B,QAAQ,MAAM,GAAG,QAAQ,SAAS,oBAAoB;AACtF,UAAM,oBAAoB,QAAQ,MAAM,QAAQ,SAAS,oBAAoB;AAE7E,SAAK,qBAAqB,eAAe,uBAAuB;AAChE,WAAO,KAAK,wBAAwB,iBAAiB;AAAA,EACvD;AACF;;;AC7GO,IAAe,wBAAf,cAGG,cAAgC;AAAA,EACjC,YAAY,QAAqB,WAAmB,oBAAgD;AACzG,UAAM,kBAAkB,mBAAmB,IAAI,CAAC,EAAE,OAAO,OAAO,MAAM;AACpE,YAAMC,SAAQ,YAAY,KAAK,IAAI,QAAQ,KAAK;AAChD,aAAO;AAAA,QACL,OAAO,QAAQA;AAAA,QACf,QAAQ,SAASA;AAAA,MACnB;AAAA,IACF,CAAC;AAED,UAAM,YAAY,gBAAgB;AAElC,WAAU,KAAK,MAAM;AACnB,YAAM,0BAA0B,CAAC,OAAe,UAAqB,MAAM,CAAI,KAAK,CAAC,EAAE,GAAG,OAAO,SAAS,GAAM,KAAK,CAAC,EAAE,GAAG,OAAO,SAAS,CAAC,GAAG,CAAC,EAAE,KAAK,GAAG,GAAG,EAAE,KAAK;AAGpK,YAAMC,cAAa,CAAC,UAAkB,SAAoD;AACxF,cAAM,EAAE,OAAO,OAAO,IAAI,gBAAgB;AAC1C,eAAO,KAAK,OAAO,MAAM,IAAI,KAAK,IAAI,QAAQ,MAAM,IAAI,IAAI;AAAA,MAC9D;AAEA,YAAM,cAAc,CAAC,aAAqBA,YAAW,UAAU,CAAC,GAAG,MAAM,IAAI,CAAC;AAC9E,YAAM,cAAc,CAAC,aAAqBA,YAAW,UAAU,CAAC,GAAG,MAAM,IAAI,CAAC;AAE9E,YAAM,kBAAkB,OACrB,IAAO,KAAK,CAAC,WAAW,GAAG,GAAG,WAAW,SAAS,CAAC,EACnD,IAAO,MAAM,MAAM,KAAK,MAAM,SAAS,GAAG,CAAC,GAAG,aAAa;AAAA,QAC1D,YAAY,QAAQ;AAAA,QACpB,YAAY,QAAQ;AAAA,MACtB,CAAC,CAAC,CAAC,EACF,IAAO,MAAM,MAAM,KAAK,MAAM,SAAS,GAAG,CAAC,GAAG,aAAa;AAAA,QAC1D,gBAAgB,UAAU;AAAA,QAC1B,gBAAgB,UAAU;AAAA,MAC5B,CAAC,CAAC,CAAC;AAEL,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EAEO,aAAaC,QAA8B;AAChD,WAAU,KAAK,MAAM;AACnB,YAAM,MAAM,KAAK,OAAOA,MAAK;AAC7B,aAAO,KAAK;AAAA,QACV;AAAA,QACAA,OAAM;AAAA,QACNA,OAAM,gBAAgB,IAAI,CAAC,CAAC,QAAQ,KAAK,OAAO,EAAE,QAAQ,MAAM,EAAE;AAAA,MACpE;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEA,MAAa,gBAAgBA,QAAgE;AAC3F,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,kBAAqB;AAAA,MACzB,MAAS,QAAQ,KAAK,aAAa,QAAQ,CAAC;AAAA,IAC9C;AAEA,UAAM,oBAAoB,MAAM,QAAQ,IAAI,gBAAgB;AAAA,MAC1D,OAAO,gBAAgB,aAAa;AAClC,cAAM,iBAAiB,MAAM,KAAK,eAAe,SAAS,CAAC;AAC3D,cAAM,UAAU,eAAe,OAAO,CAAC,GAAG,MAAMC,QAAO,CAAC,CAAC;AACzD,cAAM,UAAU,eAAe,OAAO,CAAC,GAAG,MAAM,CAACA,QAAO,CAAC,CAAC;AAE1D,eAAO,IAAI;AAAA,UACT,MAAM,EAAE,EAAE,KAAK,CAAC,EAAE,IAAI,CAAC,GAAG,MAAM,IAAI,MAAM,QAAQ,IAAc,QAAQ,EAAY,CAAC;AAAA,UACrF;AAAA,YACE,QAAQ,SAAS,eAAe,QAAQ;AAAA,YACxC,OAAO,SAAS,cAAc,QAAQ;AAAA,UACxC;AAAA,QACF;AAAA,MACF;AAAA,IACF,CAAC;AAED,oBAAgB,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC;AAE1C,WAAO,SAAS,eAAe,oBAAyC,kBAAkB;AAAA,EAC5F;AAAA,EAEU,2BAAmC;AAC3C,WAAO;AAAA,EACT;AACF;;;AC5FO,IAAM,oBAAN,cAAgC,sBAAkD;AAAA,EACvF,YAAY,uBAA6C,IAAI,qBAAqB,GAAG;AACnF,UAAM,qBAAqB,oBAAoB;AAAA,EACjD;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,0BAAkC;AAC1C,WAAO;AAAA,EACT;AACF;;;ACVO,SAAS,+BACd,WAC2E;AAC3E,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,EACF,IAAI,kBAAkB,WAAW,aAAa;AAE9C,QAAM,SAAS;AAAA,IACb,QAAQ,yBAAyB,UAAU,IAAI;AAAA,IAC/C,QAAQ,yBAAyB,QAAQ;AAAA,IACzC,QAAQ,yBAAyB,QAAQ;AAAA,EAC3C;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACpBO,SAAS,kBAAkB,SAAkG;AAClI,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM;AAAA,IACJ;AAAA,EACF,IAAI,kBAAkB,gBAAgB,aAAa;AAEnD,QAAM,SAAS,yBAAyB,GAAG,IAAI,UAAU,IAAI;AAC7D,QAAM,SAAS,yBAAyB,IAAI,IAAI,QAAQ;AACxD,QAAM,SAAS,yBAAyB,IAAI,KAAK,QAAQ;AAEzD,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,QAAQ,EAAE,QAAQ,QAAQ,OAAO;AAAA,EACnC;AACF;;;AClBO,IAAM,2BAAN,cAAuC,cAA+G;AAAA,EAC3J,cAAc;AACZ,UAAM,0BAA0B;AAAA,EAClC;AAAA,EAEO,aAAaC,QAA8B;AAChD,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAEA,WAAU,KAAK,MAAM;AACnB,YAAM,cAAiB,KAAKA,OAAM,cAAc,KAAK,IAAI,GAAG,SAAS;AACrE,YAAM,UAAU,CAAC,SAAS,SAAS,OAAO;AAC1C,YAAM,aAAa,UAAU,aAAa,OAAO,EAAE,IAAI,GAAG;AAE1D,UAAI,MAAM,YAAY,YAAY,OAAO,QAAQ,IAAI;AACrD,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAM,YAAY,KAAK,OAAO,MAAM;AACpC,YAAS,QAAQ,KAAK,CAAC,IAAI,EAAE,GAAG,CAAC,GAAG,CAAC,GAAG,OAAO;AAE/C,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAO,+BAA+B,SAAS;AAAA,EACjD;AAAA,EAEU,cAAc,SAAuB;AAC7C,WAAO,kBAAkB,OAAO;AAAA,EAClC;AACF;;;AC/CO,IAAM,wBAAN,cAAoC,sBAAsD;AAAA,EAC/F,YAAY,uBAAiD,IAAI,yBAAyB,GAAG;AAC3F,UAAM,yBAAyB,oBAAoB;AAAA,EACrD;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,0BAAkC;AAC1C,WAAO;AAAA,EACT;AACF;;;ACZO,IAAM,kBAAN,cAA8B,kBAAkB;AAAC;;;ACAjD,SAASC,OAAM,GAAgB,QAAuC;AAC3E,SAAU,KAAO,IAAI,GAAG,OAAO,OAAO,GAAG,OAAO,MAAM;AACxD;;;ACDA,SAASC,WACP,GACA,QACA,SACA,UACA,UAA4B,QACf;AACb,QAAM,EAAE,SAAS,KAAK,IAAI,OAAO;AAEjC,MAAI,MAAS,OAAO,GAAG,SAAS,SAAS,OAAO;AAChD,QAAS,KAAI,KAAK,IAAI;AACtB,QAAMC,OAAM,KAAK,OAAO,KAAK;AAC7B,SAAO,WAAc,KAAK,GAAG,IAAI;AACnC;AAEO,SAASC,MAAK,GAAgB,QAAyB;AAC5D,SAAOF,WAAU,GAAG,QAAQ,CAAC,GAAG,CAAC,GAAG,IAAI;AAC1C;AAEO,SAAS,WAAW,GAAgB,QAAyB;AAClE,SAAOA,WAAU,GAAG,QAAQ,CAAC,GAAG,CAAC,GAAG,KAAK;AAC3C;AAEO,SAAS,SAAS,GAAgB,QAAyB;AAChE,SAAOA,WAAU,GAAG,QAAQ,CAAC,GAAG,CAAC,GAAG,MAAM,OAAO;AACnD;;;ACxBA,SAASG,mBAAkB,gBAAwC,eAA+B;AAChG,WAAS,oBAAoB,iBAAyB,YAAoB,YAAiC;AACzG,UAAM,UAAU,eAAe,eAAe;AAC9C,UAAM,QAAQ,QAAQ,UAAU,aAAa,aAAa;AAE1D,QAAI,QAAQ,KAAK,GAAG;AAClB,YAAM,IAAI,MAAM,+BAA+B,0BAA0B,QAAQ,uBAAuB,2BAA2B,YAAY;AAAA,IACjJ;AAEA,WAAU;AAAA,MACR,MAAS;AAAA,QACJ,SAAS,SAAS,CAAC,YAAY,OAAO,YAAY,UAAU,CAAC;AAAA,QAChE,CAAC,GAAG,GAAG,GAAG,CAAC;AAAA,MACb;AAAA,IACF;AAAA,EACF;AAEA,WAAS,kBACP,iBACA,YACA,YACA,cACY;AACZ,UAAM,UAAU,oBAAoB,iBAAiB,YAAY,UAAU;AAC3E,UAAM,OAAU,SAAS,eAAe,UAAU,CAAC;AAEnD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,oBAAoB;AAAA,IACtC;AAEA,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,WAAS,wBAAwB,YAAoB,cAAwC;AAC3F,UAAM,UAAa,SAAS,eAAe,UAAU,CAAC;AACtD,UAAM,SAAY,SAAS,eAAe,UAAU,CAAC;AAErD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,sBAAsB;AAAA,IACxC;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,WAAS,uBACP,iBACA,YACA,YACA,cACiB;AACjB,UAAMC,QAAO,kBAAkB,iBAAiB,YAAY,YAAY,GAAG,mBAAmB;AAC9F,UAAMC,SAAQ,wBAAwB,YAAY,GAAG,oBAAoB;AAEzE,WAAO,EAAE,MAAAD,OAAM,OAAAC,OAAM;AAAA,EACvB;AAEA,WAAS,2BACP,iBACA,YACA,YACA,cACA,SAAS,OACY;AACrB,UAAM,QAAQ,wBAAwB,SAAS,MAAM,KAAK,iBAAiB,YAAY,YAAY,GAAG,oBAAoB;AAC1H,UAAMC,SAAQ,uBAAuB,iBAAiB,YAAY,YAAY,GAAG,oBAAoB;AAErG,WAAO,EAAE,OAAO,OAAAA,OAAM;AAAA,EACxB;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,eAAc,SAA6E;AACzG,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAIJ,mBAAkB,gBAAgB,aAAa;AAEnD,QAAM,cAAc,uBAAuB,MAAM,IAAI,GAAG,aAAa;AACrE,QAAM,WAAW,2BAA2B,MAAM,IAAI,GAAG,UAAU;AACnE,QAAM,WAAW,2BAA2B,MAAM,IAAI,GAAG,UAAU;AACnE,QAAM,WAAW,2BAA2B,MAAM,IAAI,GAAG,UAAU;AAEnE,QAAM,cAAc,2BAA2B,OAAO,IAAI,GAAG,eAAe,IAAI;AAChF,QAAM,WAAW,2BAA2B,OAAO,IAAI,GAAG,UAAU;AACpE,QAAM,WAAW,2BAA2B,OAAO,IAAI,GAAG,UAAU;AACpE,QAAM,WAAW,2BAA2B,OAAO,IAAI,GAAG,UAAU;AAEpE,QAAM,eAAe,2BAA2B,QAAQ,KAAK,GAAG,gBAAgB,IAAI;AACpF,QAAM,YAAY,2BAA2B,QAAQ,KAAK,GAAG,WAAW;AACxE,QAAM,YAAY,2BAA2B,QAAQ,KAAK,GAAG,WAAW;AAExE,QAAM,eAAe,2BAA2B,QAAQ,KAAK,GAAG,gBAAgB,IAAI;AACpF,QAAM,YAAY,2BAA2B,QAAQ,KAAK,GAAG,WAAW;AACxE,QAAM,YAAY,2BAA2B,QAAQ,KAAK,GAAG,WAAW;AACxE,QAAM,mBAAmB,2BAA2B,QAAQ,KAAK,GAAG,kBAAkB;AAEtF,QAAM,KAAQ;AAAA,IACZ,MAAS,UAAa,SAAS,eAAe,MAAM,GAAG,GAAG,CAAC,KAAK,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;AAAA,EAC/E;AACA,gBAAc,KAAK,EAAE,WAAW,KAAK,CAAC;AAEtC,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,QAAM,SAAS;AAAA,IACb;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,SAAO,EAAE,QAAQ,cAAc;AACjC;;;AC7IA,SAASK,mBAAkB,WAAgB,eAA+B;AACxE,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,WAAS,wBAAwB,QAAkC;AACjE,UAAM,UAAU,mBAAmB,GAAG,wBAAwB,CAAC;AAC/D,UAAM,SAAS,mBAAmB,GAAG,uBAAuB,CAAC;AAE7D,WAAO,EAAE,SAAS,OAAO;AAAA,EAC3B;AAEA,WAAS,uBAAuB,QAAiC;AAC/D,UAAM,UAAU,mBAAmB,GAAG,uBAAuB,CAAC;AAC9D,UAAM,OAAO,mBAAmB,GAAG,oBAAoB,CAAC;AACxD,UAAMC,SAAQ,wBAAwB,MAAM;AAE5C,WAAO,EAAE,MAAM,EAAE,SAAS,KAAK,GAAG,OAAAA,OAAM;AAAA,EAC1C;AAEA,WAAS,2BAA2B,QAAqC;AACvE,WAAO;AAAA,MACL,OAAO,uBAAuB,GAAG,cAAc;AAAA,MAC/C,OAAO,uBAAuB,GAAG,cAAc;AAAA,IACjD;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,4BACd,WACsD;AACtD,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAIF,mBAAkB,WAAW,aAAa;AAE9C,QAAM,cAAc,uBAAuB,aAAa;AACxD,QAAM,WAAW,2BAA2B,UAAU;AACtD,QAAM,WAAW,2BAA2B,UAAU;AACtD,QAAM,WAAW,2BAA2B,UAAU;AAEtD,QAAM,cAAc,2BAA2B,aAAa;AAC5D,QAAM,WAAW,2BAA2B,UAAU;AACtD,QAAM,WAAW,2BAA2B,UAAU;AACtD,QAAM,WAAW,2BAA2B,UAAU;AAEtD,QAAM,eAAe,2BAA2B,cAAc;AAC9D,QAAM,YAAY,2BAA2B,WAAW;AACxD,QAAM,YAAY,2BAA2B,WAAW;AAExD,QAAM,eAAe,2BAA2B,cAAc;AAC9D,QAAM,YAAY,2BAA2B,WAAW;AACxD,QAAM,YAAY,2BAA2B,WAAW;AACxD,QAAM,mBAAmB,2BAA2B,kBAAkB;AAEtE,QAAM,EAAE,GAAG,IAAI;AACf,gBAAc,KAAK,EAAE,cAAc,MAAM,WAAW,KAAK,CAAC;AAE1D,MAAI,CAAC,WAAW,EAAE,GAAG;AACnB,UAAM,IAAI,MAAM,yDAAyD,IAAI;AAAA,EAC/E;AAEA,QAAM,SAAS;AAAA,IACb;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,6BAA2B,WAAW,aAAa;AAEnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;AC1FO,SAAS,SAAS,GAAgB,QAA0C;AACjF,MAAI,MAAMG,MAAK,GAAG,OAAO,KAAK;AAC9B,QAAM,WAAW,KAAK,OAAO,KAAK;AAClC,QAAS,KAAI,KAAK,CAAC;AACnB,QAAS,KAAK,GAAG;AACjB,SAAO;AACT;AAEO,SAAS,aAAa,GAAgB,QAA0C;AACrF,MAAI,MAAM,SAAS,GAAG,OAAO,KAAK;AAClC,QAAM,WAAW,KAAK,OAAO,KAAK;AAElC,MAAI,SAAY,QAAQ,GAAG,GAAG,GAAG,OAAO;AACxC,QAAMC,SAAW,MAAkB,OAAO,KAAK;AAC/C,QAAM,QAAQ,OAAO,MAAM,OAAO,IAAI,MAAM;AAC5C,QAAM,gBAAgB,OAAO,MAAM,OAAO,IAAI,MAAM,MAAM,OAAO,MAAM,OAAO,IAAI,MAAM;AAExF,MAAI,eAAe;AACjB,UAAM,YAAY,CAAC,GAAG,IAAI,KAAK;AAC/B,cAAU,KAAK;AACf,UAAM,SAAY,MAAkB,SAAS;AAC7C,UAAS,OAAO,CAAC,KAAK,MAAM,GAAG,CAAC;AAEhC,UAAM,YAAY,CAAC,GAAG,IAAI,KAAK;AAC/B,cAAU,KAAK;AACf,UAAM,SAAY,MAAkB,SAAS;AAC7C,UAAS,OAAO,CAAC,KAAK,MAAM,GAAG,CAAC;AAAA,EAClC;AAEA,WAAS,QAAW,OAAO,CAAC,QAAQA,MAAK,GAAG,CAAC,IAAI;AACjD,QAAS,KAAI,QAAQ,GAAG;AAExB,QAAS,KAAK,GAAG;AACjB,SAAO;AACT;;;AC5BO,IAAM,qBAAN,cAAiC,cAAyB;AAAA,EAC/D,cAAc;AACZ,UAAM,oBAAoB;AAAA,EAC5B;AAAA,EAEO,aAAaC,QAA8B;AAChD,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,kDAAkD;AAAA,IACpE;AAEA,WAAU,KAAK,MAAM;AACnB,YAAM,cAAiB,KAAKA,OAAM,cAAc,KAAK,IAAI,GAAG,SAAS;AAErE,YAAM,UAAU,CAAC,SAAS,SAAS,OAAO;AAC1C,YAAM,aAAa,UAAU,aAAa,OAAO,EAAE,IAAI,GAAG;AAE1D,UAAI,MAAM,SAAS,YAAY,OAAO,WAAW;AACjD,YAAS,QAAQ,KAAK,GAAG,GAAG,OAAO;AAEnC,YAAM,SAAS,KAAK,OAAO,QAAQ;AACnC,YAAM,SAAS,KAAK,OAAO,QAAQ;AACnC,YAAM,SAAS,KAAK,OAAO,QAAQ;AAEnC,YAAM,aAAa,KAAK,OAAO,WAAW;AAC1C,YAAM,SAAS,KAAK,OAAO,QAAQ;AACnC,YAAM,SAAS,KAAK,OAAO,QAAQ;AACnC,YAAM,SAAS,KAAK,OAAO,QAAQ;AAEnC,YAAM,aAAa,KAAK,OAAO,YAAY;AAC3C,YAAM,SAAS,KAAK,OAAO,SAAS;AACpC,YAAM,SAAS,KAAK,OAAO,SAAS;AAEpC,YAAM,aAAa,KAAK,OAAO,YAAY;AAC3C,YAAM,SAAS,KAAK,OAAO,SAAS;AACpC,YAAM,SAAS,KAAK,OAAO,SAAS;AACpC,YAAM,aAAa,KAAK,OAAO,gBAAgB;AAE/C,YAAM,YAAY,IAAI,KAAK,CAAC,GAAG,CAAC,CAAC;AACjC,YAAM,iBAAoB,OAAO,WAAW,OAAO,EAAE;AAErD,aAAO;AAAA,IACT,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAwC;AAC3D,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEA,MAAa,sBAAsBA,QAAwD;AA7D7F;AA+DI,SAAI,KAAAA,UAAA,gBAAAA,OAAO,UAAP,mBAAc,KAAK,CAAC,QAAQ,OAAO;AAAI,aAAO,IAAI,aAAa,GAAG;AACtE,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,wBAA2B,KAAK,MAAS,QAAQ,KAAK,aAAa,QAAQ,CAAC,CAAC;AACnF,UAAM,0BAA0B,MAAM,QAAQ,IAAI,sBAAsB,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;AAC5F,0BAAsB,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC;AAChD,WAAO,SAAS,eAAe,0BAA0B,wBAAwB;AAAA,EACnF;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAOC,4BAA2B,SAAS;AAAA,EAC7C;AAAA,EAEU,cAAc,SAAuB;AAC7C,WAAOC,eAAc,OAAO;AAAA,EAC9B;AACF;;;AC9EO,SAAS,yBAAyB,SAAuB;AAC9D,QAAM,MAAM,IAAI,mBAAmB;AACnC,MAAI,eAAe,OAAO;AAC1B,SAAO;AACT;;;ACJO,SAAS,yBAGd,WACA,YAC6B;AAC7B,QAAM,YAAY,EAAE,WAAW;AAC/B,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACRO,SAAS,UAAU,KAA8B;AACtD,SAAO,OAAO,IAAI,QAAQ;AAC5B;AAEO,SAAS,cAGd,WACA,KACkB;AAClB,QAAM,YAAY,EAAE,IAAI;AACxB,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACRO,SAAS,aAAa,KAAiC;AAC5D,UAAQ,IAAI,gCAA0B,IAAI,qCACrC,mBAAmB,IAAI,iBAAiB;AAC/C;AAEO,SAAS,iBAGd,WACA,QACA,mBACqB;AACrB,QAAM,YAAY,EAAE,QAAQ,kBAAkB;AAC9C,SAAO,EAAE,GAAG,WAAW,GAAG,UAAU;AACtC;;;ACjBA,SAASC,mBAAkB,gBAAwC,eAA+B;AAChG,WAAS,2BAA2B,aAAqB,cAAuD;AAC9G,UAAM,UAAa,SAAS,eAAe,IAAI,IAAI,WAAW,GAAG,CAAC,GAAG,GAAG,aAAa,CAAC,CAAC;AACvF,UAAM,mBAAsB,SAAS,eAAe,WAAW,CAAC;AAChE,UAAM,oBAAuB,SAAS,eAAe,WAAW,CAAC;AACjE,UAAM,kBAAqB,SAAS,eAAe,WAAW,CAAC;AAC/D,UAAM,sBAAyB,SAAS,eAAe,WAAW,CAAC;AAEnE,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,gCAAgC;AAAA,MAChD,EAAE,WAAW,GAAG,iCAAiC;AAAA,MACjD,EAAE,WAAW,GAAG,+BAA+B;AAAA,MAC/C,EAAE,WAAW,GAAG,mCAAmC;AAAA,IACrD;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,WAAS,kBACP,YACA,aACA,YACA,cACA,iBACY;AACZ,UAAM,UAAa;AAAA,MACjB,eAAe,aAAa,cAAc,aAAa,UAAU;AAAA,MACjE,CAAC,YAAY,YAAY,YAAY,WAAW;AAAA,IAClD;AACA,UAAM,OAAU,SAAS,eAAe,WAAW,CAAC;AAEpD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,uBAAuB;AAAA,MACvC,EAAE,WAAW,GAAG,gBAAgB,kBAAkB,sBAAsB,SAAS;AAAA,IACnF;AAEA,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,WAAS,2BACP,YACA,aACA,YACA,cACqB;AACrB,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,IACF,IAAI,kBAAkB,YAAY,aAAa,YAAY,cAAc,IAAI;AAE7E,WAAO;AAAA,MACL;AAAA,MACA,mBAAmB;AAAA,IACrB;AAAA,EACF;AAEA,WAAS,sBACP,YACA,aACA,cAC4B;AAC5B,UAAM,iBAAiB,2BAA2B,YAAY,GAAG,6BAA6B;AAC9F,UAAM,iBAAiB,2BAA2B,YAAY,aAAa,GAAG,GAAG,6BAA6B;AAE9G,WAAO,EAAE,gBAAgB,eAAe;AAAA,EAC1C;AAEA,WAAS,2BAA+C;AACtD,UAAM,SAAS,2BAA2B,GAAG,IAAI,GAAG,oBAAoB;AACxE,UAAM,SAAS,sBAAsB,IAAI,IAAI,oBAAoB;AACjE,UAAM,SAAS,sBAAsB,IAAI,KAAK,oBAAoB;AAClE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,SAAS,sBAAsB,KAAK,KAAK,oBAAoB;AACnE,UAAM,UAAU,sBAAsB,KAAK,KAAK,qBAAqB;AACrE,UAAM,UAAU,sBAAsB,KAAK,KAAK,qBAAqB;AACrE,UAAM,UAAU,sBAAsB,KAAK,MAAM,qBAAqB;AACtE,UAAM,UAAU,sBAAsB,MAAM,MAAM,qBAAqB;AACvE,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,WAAS,+BAAsD;AAC7D,UAAM,SAAS,2BAA2B,MAAM,KAAK,GAAG,yBAAyB;AACjF,UAAM,SAAS,2BAA2B,KAAK,KAAK,GAAG,yBAAyB;AAChF,UAAM,SAAS,2BAA2B,KAAK,KAAK,GAAG,yBAAyB;AAChF,UAAM,SAAS,2BAA2B,KAAK,KAAK,GAAG,yBAAyB;AAChF,UAAM,SAAS,2BAA2B,KAAK,KAAK,GAAG,yBAAyB;AAChF,UAAM,SAAS,2BAA2B,KAAK,KAAK,GAAG,yBAAyB;AAChF,UAAM,SAAS,2BAA2B,KAAK,IAAI,GAAG,yBAAyB;AAC/E,UAAM,SAAS,2BAA2B,IAAI,KAAK,GAAG,yBAAyB;AAC/E,UAAM,2BAA2B,kBAAkB,KAAK,IAAI,GAAG,yDAAyD;AACxH,UAAM,oBAAoB,kBAAkB,KAAK,GAAG,GAAG,kDAAkD;AACzG,UAAM,2BAA2B,kBAAkB,MAAM,IAAI,GAAG,yDAAyD;AACzH,UAAM,oBAAoB,kBAAkB,MAAM,IAAI,GAAG,kDAAkD;AAC3G,UAAM,2BAA2B,kBAAkB,KAAK,IAAI,GAAG,yDAAyD;AACxH,UAAM,oBAAoB,kBAAkB,KAAK,IAAI,GAAG,kDAAkD;AAC1G,UAAM,2BAA2B,kBAAkB,KAAK,IAAI,GAAG,yDAAyD;AACxH,UAAM,oBAAoB,kBAAkB,KAAK,IAAI,GAAG,kDAAkD;AAC1G,UAAM,2BAA2B,kBAAkB,KAAK,IAAI,GAAG,yDAAyD;AACxH,UAAM,oBAAoB,kBAAkB,KAAK,IAAI,GAAG,kDAAkD;AAC1G,UAAM,2BAA2B,kBAAkB,KAAK,IAAI,GAAG,yDAAyD;AACxH,UAAM,oBAAoB,kBAAkB,KAAK,IAAI,GAAG,kDAAkD;AAE1G,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,UAAM,kBAAkB;AAAA,MACtB,wBAAwB;AAAA,MACxB,iBAAiB;AAAA,IACnB;AACA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,eAAc,SAA6E;AACzG,QAAM,gBAAgC,CAAC;AACvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AACjC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAID,mBAAkB,gBAAgB,aAAa;AACnD,QAAM,cAAc,yBAAyB;AAC7C,QAAM,mBAAmB,6BAA6B;AACtD,QAAM,YAAe;AAAA,IACnB,eAAe,OAAO,CAAC;AAAA,IACvB,CAAC,GAAG,MAAM,CAAC;AAAA,EACb;AACA,QAAM,eAAe;AAAA,IACnB;AAAA,EACF;AACA,gBAAc,KAAK,EAAE,WAAW,yBAAyB,CAAC;AAC1D,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AAEA,SAAO;AAAA,IACL,QAAQ;AAAA,MACN;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,IACA;AAAA,EACF;AACF;;;AChNA,SAASE,mBAAkB,WAAgB,eAA+B;AACxE,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,WAAS,2BAA2B,QAAgB,KAAa,cAA2C;AAC1G,UAAM,UAAU,mBAAmB,GAAG,iBAAiB,yBAAyB,GAAG,GAAG,sBAAsB;AAC5G,UAAM,oBAAoB,mBAAmB,GAAG,iBAAiB,uCAAuC,GAAG,GAAG,gCAAgC;AAC9I,WAAO,EAAE,SAAS,kBAAkB;AAAA,EACtC;AAEA,WAAS,sBAAsB,KAAyC;AACtE,UAAM,eAAe,oBAAoB;AACzC,UAAM,sBAAsB,sBAAsB;AAClD,UAAM,4BAA4B,GAAG;AACrC,UAAM,4BAA4B,GAAG;AAErC,UAAM,UAAU,mBAAmB,GAAG,yCAAyC,GAAG,GAAG,mCAAmC;AACxH,UAAM,mBAAmB,mBAAmB,GAAG,uCAAuC,GAAG,GAAG,4CAA4C;AACxI,UAAM,oBAAoB,mBAAmB,GAAG,sCAAsC,GAAG,GAAG,6CAA6C;AACzI,UAAM,kBAAkB,mBAAmB,GAAG,6CAA6C,GAAG,GAAG,2CAA2C;AAC5I,UAAM,sBAAsB,mBAAmB,GAAG,iDAAiD,GAAG,GAAG,+CAA+C;AAExJ,WAAO;AAAA,MACL,gBAAgB;AAAA,QACd;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAAA,MACA,gBAAgB,2BAA2B,eAAe,KAAK,yBAAyB;AAAA,IAC1F;AAAA,EACF;AAEA,WAAS,2BAA+C;AACtD,WAAO;AAAA,MACL,QAAQ,2BAA2B,eAAe,GAAG,oBAAoB;AAAA,MACzE,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,QAAQ,sBAAsB,CAAC;AAAA,MAC/B,SAAS,sBAAsB,EAAE;AAAA,MACjC,SAAS,sBAAsB,EAAE;AAAA,MACjC,SAAS,sBAAsB,EAAE;AAAA,MACjC,SAAS,sBAAsB,EAAE;AAAA,IACnC;AAAA,EACF;AAEA,WAAS,kBAAkB,QAAgB,cAAkC;AAC3E,UAAM,UAAU,mBAAmB,GAAG,kBAAkB,GAAG,GAAG,sBAAsB;AACpF,UAAM,OAAO,mBAAmB,GAAG,iBAAiB,GAAG,GAAG,mBAAmB;AAC7E,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,WAAS,0BAA0B,KAAkC;AACnE,UAAM,yBAAyB;AAAA,MAC7B,2BAA2B;AAAA,MAC3B,kCAAkC;AAAA,IACpC;AACA,UAAM,kBAAkB;AAAA,MACtB,2BAA2B;AAAA,MAC3B,kCAAkC;AAAA,IACpC;AACA,WAAO,EAAE,wBAAwB,gBAAgB;AAAA,EACnD;AAEA,WAAS,+BAAsD;AAC7D,WAAO;AAAA,MACL,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,QAAQ,2BAA2B,cAAc,GAAG,yBAAyB;AAAA,MAC7E,iBAAiB,0BAA0B,CAAC;AAAA,MAC5C,iBAAiB,0BAA0B,CAAC;AAAA,MAC5C,iBAAiB,0BAA0B,CAAC;AAAA,MAC5C,iBAAiB,0BAA0B,CAAC;AAAA,MAC5C,iBAAiB,0BAA0B,CAAC;AAAA,MAC5C,iBAAiB,0BAA0B,CAAC;AAAA,IAC9C;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,4BACd,WACsD;AACtD,QAAM,gBAAgC,CAAC;AACvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAID,mBAAkB,WAAW,aAAa;AAC9C,QAAM,YAAY,UAAU;AAC5B,gBAAc,KAAK,EAAE,cAAc,oBAAoB,WAAW,yBAAyB,CAAC;AAC5F,MAAI,CAAC,WAAW,SAAS,GAAG;AAC1B,UAAM,IAAI,MAAM,yEAAyE,WAAW;AAAA,EACtG;AAEA,QAAM,SAAS;AAAA,IACb,aAAa,yBAAyB;AAAA,IACtC,kBAAkB,6BAA6B;AAAA,IAC/C,cAAc;AAAA,MACZ;AAAA,IACF;AAAA,EACF;AAEA,6BAA2B,WAAW,aAAa;AACnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACzHO,SAAS,mBAAmB,GAAgB,QAA6B,SAA2B;AACzG,SAAU,KAAK,MAAM;AACnB,QAAI,MAAS,OAAO,GAAG,OAAO,SAAS,SAAS,MAAM;AACtD,UAAS,KAAI,KAAK,OAAO,iBAAiB;AAC1C,WAAU,YAAY,KAAK,GAAG,CAAC;AAAA,EACjC,CAAC;AACH;;;ACLA,IAAME,WAAU;AAEhB,SAAS,mBAAmB,GAAgB,QAAyC,SAA2B;AAC9G,SAAU,KAAK,MAAM;AACnB,QAAI,MAAS,gBAAgB,GAAG,OAAO,SAAS,SAAS,MAAM;AAC/D,UAAS;AAAA,MACP;AAAA,MACA,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACPA;AAAA,IACF;AACA,WAAU,YAAY,KAAK,GAAG,CAAC;AAAA,EACjC,CAAC;AACH;AAEA,SAAS,sBAAsB,UAAoC;AACjE,SAAO,CAAC,GAAG,GAAG,GAAG,EAAE,EAAE,KAAK,CAAC,QAAQ,QAAQ,QAAQ,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC;AACvE;AAEO,SAAS,YAAY,GAAgB,QAA4B;AACtE,SAAU,KAAK,MAAM;AACnB,QAAI;AACJ,QAAI,MAAM,mBAAmB,GAAG,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAErD,UAAM,iBAAiB;AAAA,MACrB,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,MACP,OAAO;AAAA,IACT;AAEA,mBAAe,QAAQ,CAAC,OAAO,MAAM;AACnC,YAAM,WAAW,IAAI;AACrB,YAAM,uBAAuB,sBAAsB,QAAQ;AAC3D,YAAM,mBAAmB,KAAK,MAAM,gBAAgB,oBAAoB;AACxE,YAAM,mBAAmB,KAAK,MAAM,gBAAgB,CAAC,GAAG,CAAC,CAAC;AAC1D,UAAI,aAAa;AAAI,iBAAS;AAAA,IAChC,CAAC;AAED,QAAI,WAAW,MAAM;AACnB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AAEA,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF,CAAC;AACH;;;AC9DA,SAAS,IAAI,OAAoB,GAAW,GAAW;AACrD,QAAM,YAAY,MAAM,UAAU;AAClC,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,IAAI,UAAU,GAAG,EAAE;AACvD,QAAM,SAAS,QAAQ,UAAU,QAAQ;AACzC,QAAM,SAAS,QAAQ,UAAU,QAAQ;AACzC,MAAI,SAAS,KAAK,SAAS;AAAG,WAAO;AACrC,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,OAAO,KAAK;AAC9C,QAAM,mBAAmB,KAAK,IAAI,mBAAmB,kBAAkB,CAAG,IAAI,KAAK,IAAI,mBAAmB,kBAAkB,CAAG;AAC/H,SAAO,oBAAoB,QAAQ,QAAQ;AAC7C;AAEO,SAASC,mBACd,OACA,QACA,eACA,cACA,gBACU;AACV,QAAM,WAAW,MAAM,MAAM;AAC7B,QAAM,aAAa,KAAK,IAAI,eAAe,QAAQ;AAEnD,QAAM,aAAa,OAChB,IAAI,CAAC,OAAO,cAAc,EAAE,OAAO,SAAS,EAAE,EAC9C,OAAO,CAAC,MAAM,EAAE,QAAQ,cAAc,EACtC,KAAK,CAAC,IAAI,OAAO,GAAG,QAAQ,GAAG,KAAK;AAEvC,QAAM,eAAe,CAAC,MAAe,KAAK,eAAe,IAAI;AAC7D,QAAM,WAAqB,CAAC;AAE5B,aAAW,QAAQ,CAAC,MAAM;AACxB,QAAI,SAAS,UAAU;AAAY;AACnC,UAAM,gBAAgB,EAAE;AACxB,aAAS,IAAI,SAAS,SAAS,GAAG,KAAK,GAAG,EAAE,GAAG;AAC7C,YAAMC,OAAM,IAAI,OAAO,EAAE,UAAU,SAAS,EAAE;AAC9C,UAAIA,SAAQ;AAAK;AACjB,QAAE,SAAS,aAAaA,IAAG;AAC3B,UAAI,EAAE,SAAS;AAAgB;AAAA,IACjC;AACA,QAAI,kBAAkB,EAAE,OAAO;AAC7B,eAAS,KAAK,EAAE,QAAQ;AAAA,IAC1B;AAAA,EACF,CAAC;AACD,SAAO;AACT;;;ACnDA,SAAS,kCAAkC,GAAgB;AACzD,QAAM,MAAS,QAAW,UAAU,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC;AAE9C,QAAM,QAAQ;AAAA,IACT,IAAI,IAAI,IAAI,IAAI,EAAE;AAAA,IAClB,IAAI,IAAI,IAAI,IAAI,EAAE;AAAA,EACvB;AACA,QAAM,UAAU;AAAA,IACX,KAAI,IAAI,IAAO,IAAI,MAAM,IAAI,CAAC,CAAC;AAAA,IAC/B,KAAI,IAAI,IAAO,IAAI,MAAM,IAAI,CAAC,CAAC;AAAA,EACpC;AACA,SAAO,EAAE,OAAO,QAAQ;AAC1B;AAEA,SAAS,iBAAiB,IAAiB,IAAiB;AAC1D,QAAM,EAAE,OAAO,QAAQ,IAAI,kCAAkC,EAAE;AAE/D,QAAM,MAAS,QAAW,UAAU,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC;AAC/C,QAAM,WAAc,IAAO,IAAO,IAAO,IAAI,IAAI,IAAI,CAAC,CAAC,GAAG,MAAM,EAAE,GAAG,CAAC;AACtE,QAAM,WAAc,KAAO,IAAO,IAAI,IAAI,IAAI,EAAE,GAAG,MAAM,EAAE,GAAG,QAAQ,EAAE;AACxE,QAAM,WAAc,IAAO,IAAO,IAAO,IAAI,IAAI,IAAI,CAAC,CAAC,GAAG,MAAM,EAAE,GAAG,CAAC;AACtE,QAAM,WAAc,KAAO,IAAO,IAAI,IAAI,IAAI,EAAE,GAAG,MAAM,EAAE,GAAG,QAAQ,EAAE;AAExE,SAAU;AAAA,IACL,MAAM;AAAA,MACJ,IAAI,UAAU,QAAQ;AAAA,MACtB,IAAI,UAAU,QAAQ;AAAA,MACtB,KAAI,UAAU,QAAQ;AAAA,MACtB,KAAI,UAAU,QAAQ;AAAA,IAC3B,CAAC;AAAA,IACD,CAAC,GAAG,CAAC;AAAA,EACP;AACF;AAEO,SAAS,YAAY,gBAA6B,kBAA+B,QAA2B;AACjH,SAAU,KAAK,MAAM;AACnB,UAAM,YAAY,eAAe,MAAM;AAEvC,QAAI,QAAQ;AAAA,MACP,QAAW,KAAK,OAAO,WAAW,CAAC,WAAW,GAAG,CAAC,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;AAAA,MAC7D,QAAQ,gBAAgB,CAAC,IAAI,CAAC,CAAC;AAAA,IACpC;AACA,YAAW,QAAQ,OAAO,CAAC,WAAY,MAAM,MAAM,KAAK,WAAY,CAAC,CAAC;AAEtE,UAAM,mBAAsB,QAAW,MAAM,kBAAkB,CAAC,GAAG,GAAG,CAAC,GAAG,CAAC,IAAI,IAAI,EAAE,CAAC,CAAC;AACvF,QAAI,SAAY,MAAM,kBAAkB,CAAC,GAAG,GAAG,CAAC,GAAG,CAAC,IAAI,IAAI,CAAC,CAAC;AAE9D,aAAY,QAAQ,QAAQ,CAAC,WAAW,OAAO,MAAM,EAAY,CAAC;AAElE,UAAM,eAAkB,QAAQ,KAAK;AACrC,UAAM,gBAAmB,QAAQ,MAAM;AAEvC,WAAO,EAAE,OAAO,cAAc,QAAQ,cAAc;AAAA,EACtD,CAAC;AACH;;;ACrDO,SAAS,mBACd,GACA,QACA;AACA,SAAU,KAAK,MAAM;AACnB,UAAM,YAAY,EAAE,MAAM;AAC1B,UAAM,wBAA2B;AAAA,MAC/B,UAAU,GAAG,OAAO,sBAAsB;AAAA,MAC1C,CAAC,WAAW,IAAI,GAAG,CAAC;AAAA,IACtB;AACA,UAAM,kBAAqB;AAAA,MACzB,UAAU,GAAG,OAAO,eAAe;AAAA,MACnC,CAAC,WAAW,IAAI,CAAC;AAAA,IACnB;AACA,WAAO,EAAE,uBAAuB,gBAAgB;AAAA,EAClD,CAAC;AACH;;;ACfO,SAAS,gBACd,GACA,QACA,QACA;AACA,SAAU,KAAK,MAAM;AACnB,UAAM,QAAQ,mBAAmB,GAAG,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AACzD,UAAM,QAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAMC,SAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAM,QAAQ,mBAAmBA,QAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAM,QAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAM,QAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAM,QAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAC7D,UAAM,QAAQ,mBAAmB,OAAO,OAAO,QAAQ,CAAC,GAAG,CAAC,CAAC;AAE7D,UAAM,iBAAiB,mBAAmB,QAAQ,OAAO,eAAe;AACxE,UAAM,iBAAiB,mBAAmB,GAAG,OAAO,eAAe;AACnE,UAAM,iBAAiB,mBAAmB,OAAO,OAAO,eAAe;AACvE,UAAM,iBAAiB,mBAAmB,OAAO,OAAO,eAAe;AACvE,UAAM,iBAAiB,mBAAmB,OAAO,OAAO,eAAe;AACvE,UAAM,iBAAiB,mBAAmB,OAAO,OAAO,eAAe;AAEvE,UAAM,iBAAoB,OAAO;AAAA,MAC/B,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,IACjB,GAAG,CAAC;AAEJ,UAAM,mBAAsB,OAAO;AAAA,MACjC,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,MACf,eAAe;AAAA,IACjB,GAAG,CAAC;AAEJ,WAAO;AAAA,MACL;AAAA,MACA;AAAA,IACF;AAAA,EACF,CAAC;AACH;;;AC9CO,IAAM,wBAAN,MAA4B;AAAA,EAOjC,YAAY,EAAE,eAAe,WAAW,IAA4B,CAAC,GAAG;AANxE,SAAU,QAAQ;AAOhB,SAAK,iBAAiB,iBAAiB;AACvC,SAAK,cAAc,cAAc;AAEjC,QAAI,OAAO,KAAK,mBAAmB,YAAY,KAAK,kBAAkB,KAAK,KAAK,kBAAkB,GAAG;AACnG,YAAM,IAAI,MAAM,GAAG,KAAK,+DAA+D;AAAA,IACzF;AAEA,QAAI,OAAO,KAAK,gBAAgB,UAAU;AACxC,YAAM,IAAI,MAAM,GAAG,KAAK,4CAA4C;AAAA,IACtE;AAAA,EACF;AAAA,EAEA,IAAI,gBAAwB;AAAE,WAAO,KAAK;AAAA,EAAgB;AAAA,EAE1D,IAAI,aAAqB;AAAE,WAAO,KAAK;AAAA,EAAa;AACtD;;;ACbO,IAAM,iBAAN,cAA6B,cAAyB;AAAA,EAC3D,cAAc;AACZ,UAAM,gBAAgB;AAAA,EACxB;AAAA,EAEO,aAAaC,QAAiB;AACnC,UAAM,EAAE,OAAO,IAAI;AACnB,QAAI,CAAC;AAAQ,YAAM,IAAI,MAAM,8CAA8C;AAC3E,WAAU,KAAK,MAAM;AACnB,YAAM,cAAiB,KAAKA,OAAM,cAAc,KAAK,KAAK,GAAG,SAAS;AACtE,YAAM,IAAO,IAAO,IAAI,aAAa,KAAK,GAAG,CAAC;AAC9C,YAAM,WAAW,YAAY,GAAG,OAAO,WAAW;AAClD,YAAM,EAAE,gBAAgB,iBAAiB,IAAI,gBAAgB,SAAS,KAAK,SAAS,QAAQ,OAAO,gBAAgB;AACnH,aAAO,YAAY,gBAAgB,kBAAkB,OAAO,YAAY;AAAA,IAC1E,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAkB;AACrC,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,CAAC;AAAA,EAClD;AAAA,EAEA,MAAa,YAAYA,QAAkB,UAAkC,CAAC,GAA6B;AACzG,UAAM,EAAE,YAAY,cAAc,IAAI,IAAI,sBAAsB,OAAO;AACvE,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,EAAE,OAAO,QAAQ,QAAQ,QAAQ,IAAI,KAAK,aAAa,QAAQ;AACrE,UAAM,QAAQ,OAAO;AACrB,UAAM,SAAS,QAAQ;AACvB,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,aAAO,GAAG,QAAQ;AAClB,cAAQ,GAAG,QAAQ;AAAA,IACrB;AACA,UAAM,aAAa,MAAM,KAAK,OAAO,SAAS,CAAC;AAC/C,UAAM,eAAe;AACrB,UAAM,UAAUC,mBAAkB,OAAO,YAAwB,YAAY,cAAc,aAAa;AACxG,UAAM,eAAe,SAAS,2BAA2B,CAAC;AAC1D,UAAM,YAAY,SAAS;AAC3B,UAAM,OAAO,YAAY,aAAa;AACtC,UAAM,OAAO,YAAY,aAAa;AACtC,UAAM,YAAY,MAAM,UAAU;AAClC,UAAM,UAAU,QACb,IAAI,CAAC,QAAQ;AACZ,YAAM,CAAC,KAAK,MAAM,IAAI;AAAA,QACpB,KAAK,IAAI,GAAG,UAAU,KAAK,EAAE;AAAA,QAC7B,KAAK,IAAI,GAAK,UAAU,KAAK,EAAE;AAAA,MACjC,EAAE,IAAI,CAAC,QAAQ,MAAM,IAAI;AACzB,YAAM,CAAC,MAAM,KAAK,IAAI;AAAA,QACpB,KAAK,IAAI,GAAG,UAAU,KAAK,EAAE;AAAA,QAC7B,KAAK,IAAI,GAAK,UAAU,KAAK,EAAE;AAAA,MACjC,EAAE,IAAI,CAAC,QAAQ,MAAM,IAAI;AACzB,aAAO,IAAI;AAAA,QACT,WAAW;AAAA,QACX,IAAI,KAAK,MAAM,KAAK,QAAQ,MAAM,SAAS,GAAG;AAAA,QAC9C,EAAE,QAAQ,SAAS,eAAe,CAAC,GAAG,OAAO,SAAS,cAAc,CAAC,EAAE;AAAA,MACzE;AAAA,IACF,CAAC;AACH,UAAM,QAAQ;AACd,WAAO,QAAQ;AACf,WAAO;AAAA,EACT;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAOC,4BAA2B,SAAS;AAAA,EAC7C;AAAA,EAEU,cAAc,SAAuB;AAC7C,WAAOC,eAAc,OAAO;AAAA,EAC9B;AACF;;;ACjFO,SAAS,qBAAqB,SAAuB;AAC1D,QAAM,MAAM,IAAI,eAAe;AAC/B,MAAI,eAAe,OAAO;AAC1B,SAAO;AACT;AAEO,SAAS,uBAAuB,SAAuB;AAC5D,SAAO,qBAAqB,OAAO;AACrC;AAGO,IAAM,mBAAN,cAA+B,eAAe;AAAC;;;ACd/C,IAAM,gBAAgB;AAEtB,IAAM,cAAc;AAAA,EACzB,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,SAAS,OAAO;AAAA,EAC1B,IAAI,MAAM,SAAS,OAAO;AAAA,EAC1B,IAAI,MAAM,QAAQ,OAAO;AAAA,EACzB,IAAI,MAAM,SAAS,OAAO;AAC5B;AAEO,IAAM,wBAAwB;AAAA,EACnC,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAC9B;AAEO,IAAM,qBAA+C,CAAC,SAAS,SAAS,MAAM;AAE9E,IAAMC,sBAAqB;AAC3B,IAAM,oCAAoC;;;ACVjD,IAAMC,YAAW,CAAC,QAAa,OAAO,QAAQ;AAEvC,SAAS,eAAe,QAAa;AAC1C,MAAI,CAAC,QAAQ;AACX,UAAM,IAAI,MAAM,mBAAmB,QAAQ;AAAA,EAC7C;AAEA,MAAI,OAAO,OAAO,uBAAuB,WAAW;AAClD,UAAM,IAAI,MAAM,wDAAwD,OAAO,oBAAoB;AAAA,EACrG;AAEA,MAAI,CAACA,UAAS,OAAO,YAAY,KAAK,OAAO,eAAe,KAAK,OAAO,eAAe,GAAK;AAC1F,UAAM,IAAI,MAAM,gEAAgE,OAAO,cAAc;AAAA,EACvG;AAEA,MACE,CAAC,MAAM,QAAQ,OAAO,OAAO,KAC1B,CAAC,OAAO,QAAQ,UAChB,CAAC,OAAO,QAAQ,MAAM,CAAC,MAAW,OAAO,MAAM,QAAQ,GAC1D;AACA,UAAM,IAAI,MAAM,kEAAkE,KAAK,UAAU,OAAO,OAAO,GAAG;AAAA,EACpH;AAEA,MACE,CAAC,MAAM,QAAQ,OAAO,OAAO,KAC1B,CAAC,OAAO,QAAQ,UAChB,CAAC,OAAO,QAAQ,IAAI,CAAC,MAAW,KAAK,CAAC,CAAC,EAAE,MAAM,CAAC,MAAWA,UAAS,EAAE,CAAC,KAAKA,UAAS,EAAE,CAAC,CAAC,GAC5F;AACA,UAAM,IAAI,MAAM,wEAAwE,KAAK,UAAU,OAAO,OAAO,GAAG;AAAA,EAC1H;AAEA,MAAI,OAAO,YACT,CAAC,MAAM,QAAQ,OAAO,OAAO,KAC1B,OAAO,QAAQ,WAAW,KAC1B,CAAC,OAAO,QAAQ,MAAMA,SAAQ,IAChC;AACD,UAAM,IAAI,MAAM,8EAA8E,KAAK,UAAU,OAAO,OAAO,GAAG;AAAA,EAChI;AACF;;;ACjDO,SAAS,MAAM,GAA6B;AACjD,SAAU,KAAK,MAAM;AACnB,UAAMC,OAAS,IAAI,GAAM,OAAO,mBAAmB,CAAC;AACpD,WAAU,KAAO,KAAQ,IAAI,GAAGA,IAAG,CAAC,GAAGA,IAAG;AAAA,EAC5C,CAAC;AACH;;;ACFO,SAAS,kBAAkB,GAAgB,QAAwC;AACxF,SAAU,KAAK,MAAM;AACnB,QAAI,MAAS,IAAI,GAAG,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC;AACpD,UAAS,OAAO,KAAK,OAAO,KAAK,SAAS,CAAC,GAAG,CAAC,GAAG,OAAO;AACzD,UAAS,IAAI,KAAK,OAAO,GAAG,GAAG;AAC/B,UAAS,IAAI,KAAK,OAAO,GAAG,OAAO;AACnC,UAAS,KAAI,KAAK,OAAO,KAAK,IAAI;AAClC,WAAO,MAAM,GAAG;AAAA,EAClB,CAAC;AACH;;;ACTO,SAASC,wBAAuB,GAAgB,QAA0C;AAC/F,SAAU,KAAK,MAAM;AACnB,QAAI,MAAS,IAAI,GAAG,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAC;AACpD,UAAS,gBAAgB,KAAK,OAAO,kBAAkB,OAAO,kBAAkB,CAAC,GAAG,CAAC,GAAG,OAAO;AAC/F,UAAS,KAAI,KAAK,OAAO,IAAI;AAC7B,WAAO,MAAM,GAAG;AAAA,EAClB,CAAC;AACH;;;ACHA,SAASC,mBAAkB,gBAAwC,eAA+B;AAChG,QAAM,oBAAoB,yBAAyB,gBAAgB,aAAa;AAEhF,WAAS,uBAAuB,MAAc,cAAiC;AAC7E,UAAMC,OAAS,SAAS,eAAe,IAAI,CAAC;AAC5C,UAAM,UAAa,SAAS,eAAe,IAAI,CAAC;AAEhD,kBAAc;AAAA,MACZ,EAAE,WAAW,GAAG,mBAAmB;AAAA,MACnC,EAAE,WAAW,GAAG,uBAAuB;AAAA,IACzC;AACA,WAAO,EAAE,KAAAA,MAAK,QAAQ;AAAA,EACxB;AAEA,WAAS,+BAA+B,YAAoB,aAAqB,cAAyC;AACxH,UAAMC,QAAO,kBAAkB,YAAY,aAAa,GAAG,GAAG,mBAAmB;AACjF,UAAM,KAAK,uBAAuB,aAAa,GAAG,iBAAiB;AACnE,WAAO,EAAE,MAAAA,OAAM,GAAG;AAAA,EACpB;AACA,QAAM,6BAA6B,kCAAkC,gBAAgB,aAAa;AAElG,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,eACd,SACA,QACA,iBACA,aACgE;AAChE,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,EACF,IAAI,sBAAsB,OAAO;AAEjC,QAAM,gBAAgC,CAAC;AACvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA;AAAA,EACF,IAAIH,mBAAkB,gBAAgB,aAAa;AACnD,MAAI;AAEJ,MAAI,OAAO,oBAAoB;AAC7B,UAAM,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,EAAE,IAAI;AAC7C,UAAM,QAAQ,OAAO,qBACjB,kBAAkB,IAAI,IAAI,GAAG,OAAO,IACpC,2BAA2B,IAAI,IAAI,OAAO;AAC9C,UAAM,QAAQ,2BAA2B,IAAI,IAAI,OAAO;AACxD,UAAMI,SAAQ,2BAA2B,IAAI,IAAI,OAAO;AACxD,UAAM,QAAQ,2BAA2B,IAAI,IAAI,OAAO;AACxD,UAAM,QAAQ,2BAA2B,IAAI,IAAI,OAAO;AACxD,UAAM,QAAQ,2BAA2B,IAAI,IAAI,OAAO;AACxD,UAAM,QAAQ,KAAK,2BAA2B,IAAI,IAAI,OAAO,IAAI;AACjE,UAAM,QAAQ,KAAK,2BAA2B,IAAI,IAAI,OAAO,IAAI;AACjE,UAAM,QAAQ,kBAAkB,MAAM,MAAM,IAAI,IAAI,iBAAiB,GAAG,OAAO;AAC/E,aAAS;AAAA,MACP;AAAA,MAAO;AAAA,MAAO,OAAAA;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,IAC1D;AAAA,EACF,OAAO;AACL,UAAM,CAAC,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,IAAI,EAAE,IAAI;AAC7C,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAMA,SAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,+BAA+B,IAAI,IAAI,OAAO;AAC5D,UAAM,QAAQ,kBAAkB,IAAI,IAAI,iBAAiB,GAAG,OAAO;AACnE,aAAS;AAAA,MACP;AAAA,MAAO;AAAA,MAAO,OAAAA;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,MAAO;AAAA,IAC1D;AAAA,EACF;AACA,MAAI,oBAAoB,EAAE,WAAW,GAAG;AACtC,UAAM,IAAI,MAAM,kCAAkC,oBAAoB,EAAE,QAAQ;AAAA,EAClF;AACA,SAAO,EAAE,QAAQ,cAAc;AACjC;;;ACjFA,SAASC,mBAAkB,WAAgB,eAA+B;AACxE,QAAM,qBAAqB,0BAA0B,WAAW,aAAa;AAE7E,WAAS,uBAAuB,QAA2B;AACzD,UAAMC,OAAM,mBAAmB,GAAG,cAAc,CAAC;AACjD,UAAM,UAAU,mBAAmB,GAAG,kBAAkB,CAAC;AACzD,WAAO,EAAE,KAAAA,MAAK,QAAQ;AAAA,EACxB;AAEA,WAAS,kBAAkB,QAA4B;AACrD,UAAM,UAAU,mBAAmB,GAAG,kBAAkB,CAAC;AACzD,UAAM,OAAO,mBAAmB,GAAG,eAAe,CAAC;AACnD,WAAO,EAAE,SAAS,KAAK;AAAA,EACzB;AAEA,WAAS,+BAA+B,QAAmC;AACzE,UAAMC,QAAO,kBAAkB,GAAG,aAAa;AAC/C,UAAM,KAAK,uBAAuB,GAAG,WAAW;AAChD,WAAO,EAAE,MAAAA,OAAM,GAAG;AAAA,EACpB;AAEA,QAAM,6BAA6B,+BAA+B,kBAAkB;AACpF,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAASC,4BACd,WACA,QACgE;AAChE,QAAM,gBAAgC,CAAC;AAEvC,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA;AAAA,EACF,IAAIH,mBAAkB,WAAW,aAAa;AAE9C,MAAI;AAEJ,MAAI,OAAO,oBAAoB;AAE7B,UAAM,aAAc,OAAO,eAAe,OAAO,YAAY,UAAU;AACvE,aAAS;AAAA,MACP,OAAO,OAAO,qBAAqB,kBAAkB,OAAO,IAAI,2BAA2B,OAAO;AAAA,MAClG,OAAO,2BAA2B,OAAO;AAAA,MACzC,OAAO,2BAA2B,OAAO;AAAA,MACzC,OAAO,2BAA2B,OAAO;AAAA,MACzC,OAAO,2BAA2B,OAAO;AAAA,MACzC,OAAO,2BAA2B,OAAO;AAAA,MACzC,OAAO,aAAa,IAAI,2BAA2B,OAAO,IAAI;AAAA,MAC9D,OAAO,aAAa,IAAI,2BAA2B,OAAO,IAAI;AAAA,MAC9D,OAAO,kBAAkB,OAAO;AAAA,IAClC;AAAA,EACF,OAAO;AACL,aAAS;AAAA,MACP,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,+BAA+B,OAAO;AAAA,MAC7C,OAAO,kBAAkB,OAAO;AAAA,IAClC;AAAA,EACF;AAEA,6BAA2B,WAAW,aAAa;AACnD,SAAO,EAAE,QAAQ,cAAc;AACjC;;;AC9EO,IAAM,oBAAN,MAAwB;AAAA,EAO7B,YAAY,EAAE,WAAW,eAAe,IAAwB,CAAC,GAAG;AANpE,SAAU,QAAQ;AAOhB,SAAK,aAAa,aAAa;AAC/B,SAAK,kBAAkB,kBAAkB;AAEzC,QAAI,OAAO,KAAK,eAAe,YAAY,KAAK,aAAa,OAAO,GAAG;AACrE,YAAM,IAAI,MAAM,GAAG,KAAK,2DAA2D;AAAA,IACrF;AAEA,QAAI,OAAO,KAAK,oBAAoB,YAAY,KAAK,mBAAmB,KAAK,KAAK,mBAAmB,GAAG;AACtG,YAAM,IAAI,MAAM,GAAG,KAAK,gEAAgE;AAAA,IAC1F;AAAA,EACF;AAAA,EAEA,IAAI,YAAoB;AAAE,WAAO,KAAK;AAAA,EAAY;AAAA,EAElD,IAAI,iBAAyB;AAAE,WAAO,KAAK;AAAA,EAAiB;AAC9D;;;ACLO,IAAM,kBAAN,cAA6B,cAAmC;AAAA,EAKrE,YAAY,QAA0B;AACpC,UAAM,YAAY;AAClB,mBAAe,MAAM;AACrB,SAAK,UAAU;AAAA,EACjB;AAAA,EAEA,IAAW,SAA2B;AACpC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAW,kBAA2B;AACpC,WAAO,KAAK,OAAO,mBAAmB,KAAK,OAAO,QAAQ,SAAS;AAAA,EACrE;AAAA,EAEA,IAAW,kBAA0B;AACnC,WAAO,KAAK,KAAK,kBAAkB,KAAK,OAAO,QAAQ,SAAS;AAAA,EAClE;AAAA,EAEO,cAAc,GAAgB,QAAiD;AACpF,QAAI,MAAM,kBAAkB,GAAG,OAAO,KAAK;AAC3C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,UAAM,kBAAkB,KAAK,OAAO,KAAK;AACzC,WAAO,UAAU,KAAK,OAAO,OAAO,SAAS,KAAK;AAAA,EACpD;AAAA,EAEO,aAAa,GAAgB,QAAsC;AACxE,QAAI,MAAM,KAAK,OAAO,qBAClB,MAAM,UAAU,GAAG,OAAO,OAAqB,SAAS,KAAK,CAAC,IAC9DI,wBAAuB,GAAG,OAAO,KAA4B;AACjE,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAMA,wBAAuB,KAAK,OAAO,KAAK;AAC9C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAMA,wBAAuB,KAAK,OAAO,KAAK;AAC9C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAMA,wBAAuB,KAAK,OAAO,KAAK;AAC9C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAMA,wBAAuB,KAAK,OAAO,KAAK;AAC9C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAMA,wBAAuB,KAAK,OAAO,KAAK;AAC9C,UAAS,QAAQ,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,GAAG,MAAM;AAC5C,UAAM,OAAO,QAAQA,wBAAuB,KAAK,OAAO,KAAK,IAAI;AACjE,UAAM,OAAO,QAAQA,wBAAuB,KAAK,OAAO,KAAK,IAAI;AACjE,WAAO,UAAU,KAAK,OAAO,OAAO,SAAS,KAAK;AAAA,EACpD;AAAA,EAEO,aAAaC,QAAiB,WAAgC;AACnE,UAAM,EAAE,OAAO,IAAI;AAEnB,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI,MAAM,0CAA0C;AAAA,IAC5D;AAEA,WAAU,KAAK,MAAM;AACnB,UAAI,cAAiB,KAAKA,OAAM,cAAc,WAAW,KAAK,GAAG,SAAS;AAC1E,oBAAc,KAAK,OAAO,UACtB,UAAU,aAAa,KAAK,OAAO,OAAO,IAC1C;AACJ,oBAAc,YAAY,IAAI,GAAG;AACjC,aAAO,KAAK,OAAO,qBACf,KAAK,aAAa,aAAa,MAAyB,IACxD,KAAK,cAAc,aAAa,MAAoC;AAAA,IAC1E,CAAC;AAAA,EACH;AAAA,EAEA,MAAa,QAAQA,QAAkB,WAAyC;AAC9E,WAAO,KAAK,aAAa,MAAM,WAAWA,MAAK,GAAG,SAAS;AAAA,EAC7D;AAAA,EAEA,MAAa,OAAOA,QAAkB,gBAAoC,CAAC,GAA+B;AACxG,UAAM,EAAE,WAAW,eAAe,IAAI,IAAI,kBAAkB,aAAa;AACzE,UAAM,WAAW,MAAM,WAAWA,MAAK;AACvC,UAAM,MAAM,MAAM,KAAK,aAAa,UAAU,SAAS;AACvD,UAAM,OAAU,KAAK,MAAS,QAAQ,GAAG,EAAE,GAAG,WAAW,CAAC;AAC1D,UAAM,kBAAkB;AAAA,MACtB,OAAO,SAAS,cAAc,CAAC;AAAA,MAC/B,QAAQ,SAAS,eAAe,CAAC;AAAA,IACnC;AAEA,UAAM,UAAU,MAAM,KAAK,aAAa,MAAM,SAAS,2BAA2B,CAAC,GAAG,cAAc;AACpG,QAAI,QAAQ;AACZ,SAAK,QAAQ;AAEb,UAAM,QAAQ,QAAQ,IAAI,CAAC,QAAQ,IAAI,GAAG;AAC1C,UAAM,SAAS,QAAQ,IAAI,CAAC,QAAQ,IAAI,KAAK;AAC7C,UAAM,cAAc,QAAQ,IAAI,CAAC,QAAQ,IAAI,UAAU;AACvD,UAAM,aAAa,QAAQ,IAAI,CAAC,QAAQ,KAAK,OAAO,QAAQ,IAAI,MAAM;AAEtE,UAAM,UAAUC;AAAA,MACd,MAAM,IAAI,CAAC,QAAQ,IAAI,QAAQ,SAAS,CAAC;AAAA,MACzC;AAAA,MACA,KAAK,OAAO;AAAA,MACZ;AAAA,IACF;AAEA,UAAM,aAAa,QAAQ,IAAI,CAAC,QAAQ,IAAI;AAAA,MAC1C,OAAO;AAAA,MACP,YAAY;AAAA,MACZ,WAAW;AAAA,MACX,MAAM;AAAA,MACN;AAAA,IACF,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAEU,sBAA8B;AACtC,WAAO;AAAA,EACT;AAAA,EAEU,2BAA2B,WAA8B;AACjE,WAAOC,4BAA2B,WAAW,KAAK,MAAM;AAAA,EAC1D;AAAA,EAEU,cAAc,SAAuB;AAC7C,UAAM,cAAc,KAAK,OAAO,eAAe,gBAAe;AAE9D,UAAM,aAAa,cAAc,YAAY,SAAS;AACtD,QAAI,eAAe,KAAK,eAAe,KAAK,eAAe,GAAG;AAC5D,YAAM,IAAI,MAAM,oEAAoE,kCAAkC;AAAA,IACxH;AACA,WAAOC,eAAc,SAAS,KAAK,QAAQ,KAAK,iBAAiB,WAAW;AAAA,EAC9E;AAAA,EAEA,MAAgB,aACd,cACA,qBACA,gBACA;AACA,UAAM,EAAE,OAAO,OAAO,IAAI;AAC1B,UAAM,YAAY,KAAK,IAAI,OAAO,MAAM;AACxC,UAAM,oBAAoB,YAAY;AACtC,UAAM,oBAAoB,YAAY;AAEtC,UAAM,WAAW,aAAa,MAAM;AACpC,UAAM,WAAW,KAAK,OAAO,QAAQ;AAErC,UAAM,CAAC,aAAa,cAAc,iBAAiB,IAAO,KAAK,MAAM;AACnE,YAAM,WAAW,aAAa,QAAQ,CAAC,UAAU,UAAU,UAAU,KAAK,eAAe,CAAC;AAE1F,YAAM,QAAQ,SAAS,MAAM,CAAC,GAAG,GAAG,GAAG,CAAC,GAAG,CAAC,UAAU,UAAU,UAAU,CAAC,CAAC;AAC5E,YAAM,SAAS,SAAS,MAAM,CAAC,GAAG,GAAG,GAAG,CAAC,GAAG,CAAC,UAAU,UAAU,UAAU,CAAC,CAAC;AAC7E,YAAM,cAAc,KAAK,kBAClB,QAAQ,SAAS,MAAM,CAAC,GAAG,GAAG,GAAG,CAAC,GAAG,CAAC,UAAU,UAAU,UAAU,KAAK,OAAO,QAAQ,MAAM,CAAC,GAAG,CAAC,IACnG,OAAO,CAAC;AACf,aAAO,CAAC,OAAO,QAAQ,WAAW;AAAA,IACpC,CAAC;AAED,UAAM,UAAU,CAAC;AACjB,UAAM,aAAa,MAAM,aAAa,MAAM;AAC5C,UAAM,YAAY,MAAM,YAAY,MAAM;AAC1C,aAAS,MAAM,GAAG,MAAM,UAAU,OAAO;AACvC,eAAS,MAAM,GAAG,MAAM,UAAU,OAAO;AACvC,iBAAS,SAAS,GAAG,SAAS,UAAU,UAAU;AAChD,gBAAM,QAAQC,SAAQ,WAAW,KAAK,KAAK,QAAQ,EAAE;AACrD,cAAI,CAAC,kBAAkB,QAAQ,gBAAgB;AAC7C,kBAAM,OAAQ,MAAMA,SAAQ,UAAU,KAAK,KAAK,QAAQ,EAAE,KAAK,WAAY;AAC3E,kBAAM,OAAQ,MAAMA,SAAQ,UAAU,KAAK,KAAK,QAAQ,EAAE,KAAK,WAAY;AAC3E,kBAAM,aAAe,KAAK,IAAI,UAAU,KAAK,KAAK,QAAQ,EAAE,IAAI,KAAK,OAAO,QAAQ,QAAQ,IAAK,WAAY;AAC7G,kBAAM,cAAgB,KAAK,IAAI,UAAU,KAAK,KAAK,QAAQ,EAAE,IAAI,KAAK,OAAO,QAAQ,QAAQ,IAAK,WAAY;AAC9G,kBAAM,IAAK,MAAO,aAAa;AAC/B,kBAAM,IAAK,MAAO,cAAc;AAChC,kBAAM,MAAM,EAAE,KAAK,KAAK,OAAO;AAC/B,kBAAM,EAAE,YAAY,MAAM,IAAI,KAAK,kBAC/B,MAAM,KAAK,sBAAsB,mBAAkC,GAAG,IACtE,EAAE,YAAY,GAAG,OAAO,EAAE;AAC9B,oBAAQ,KAAK;AAAA,cACX,KAAK,IAAI,YAAY,GAAG,GAAG,IAAI,YAAY,IAAI,WAAW;AAAA,cAC1D;AAAA,cACA,YAAY,QAAQ;AAAA,cACpB;AAAA,cACA,GAAG;AAAA,YACL,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,gBAAY,QAAQ;AACpB,iBAAa,QAAQ;AACrB,sBAAkB,QAAQ;AAC1B,WAAO;AAAA,EACT;AAAA,EAEA,MAAc,sBAAsB,eAA4B,KAAmD;AACjH,UAAM,EAAE,KAAK,KAAK,OAAO,IAAI;AAC7B,UAAM,cAAc,MAAM,cAAc,MAAM;AAC9C,WAAO,MAAM,KAAK,OAAO,QAAQ,MAAM,EAAE,KAAK,CAAC,EAC5C,IAAI,CAAC,GAAG,MAAM,YAAY,KAAK,KAAK,QAAQ,EAAE,EAC9C,IAAI,CAAC,YAAY,WAAW;AAAA,MAC3B;AAAA,MACA;AAAA,IACF,EAAE,EACD,OAAO,CAACC,MAAK,SAAUA,KAAI,aAAa,KAAK,aAAaA,OAAM,IAAK;AAAA,EAC1E;AACF;AAjNO,IAAM,iBAAN;AAAM,eACG,uBAAuB,CAAC,GAAG,IAAI,IAAI,IAAI,KAAK,KAAK,KAAK,MAAM,IAAI;;;ACPzE,IAAM,aAAN,cAAyB,eAAe;AAAA,EAC7C,YAAY,qBAAqB,MAAM;AACrC,UAAM,SAAS;AAAA,MACb;AAAA,MACA,cAAc;AAAA,MACd,SAAS,CAAC,MAAM;AAAA,MAChB,GAAI,qBACA;AAAA,QACA,SAAS;AAAA,QACT,SAAS;AAAA,MACX,IACE;AAAA,QACA,SAAS;AAAA,QACT,iBAAiB;AAAA,MACnB;AAAA,IACJ;AAEA,UAAM,MAAM;AAAA,EACd;AAAA,EAEA,IAAW,qBAA8B;AACvC,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,IAAW,UAAmB;AAC5B,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,MAAa,YAAYC,QAAkB,eAA6D;AACtG,UAAM,mBAAmB,MAAM,KAAK,OAAOA,QAAO,aAAa;AAC/D,WAAO,iBAAiB,IAAI,CAAC,QAAQ,IAAI,cAAc,IAAI,OAAO,IAAI,aAAa,EAAE,OAAO,IAAI,YAAY,QAAQ,IAAI,YAAY,CAAC,CAAC;AAAA,EACxI;AAAA,EAEmB,sBAA8B;AAC/C,WAAO,KAAK,qBAAqB,oCAAoCC;AAAA,EACvE;AAAA,EAEmB,2BAA2B,WAA8F;AAC1I,WAAO,MAAM,2BAA2B,SAAS;AAAA,EACnD;AACF;;;AClDO,SAAS,iBAAiB,SAAuB,qBAAqB,MAAM;AACjF,QAAM,MAAM,IAAI,WAAW,kBAAkB;AAC7C,MAAI,eAAe,OAAO;AAC1B,SAAO;AACT;;;ACPO,IAAM,0BAAN,cAAsC,kBAAkB;AAAA,EAAxD;AAAA;AACL,SAAmB,QAAQ;AAAA;AAC7B;;;ACNO,IAAM,iBAAN,MAAwB;AAAA,EAE7B,MAAa,KAAK,aAA2D;AAC3E,WAAO,YAAY,MAAM,KAAK,IAAI,CAAC;AAAA,EACrC;AAAA,EAEA,MAAa,MAAkB;AAC7B,UAAM,IAAI,MAAM,yCAAyC;AAAA,EAC3D;AACF;;;ACFA,eAAsB,iCACpB,eACAC,QAEA,gBACA,gBAEA,sBAAwF,CAAC,EAAE,YAAY,MAAM,aAC7G;AACA,QAAM,YAAY,cAAc,IAAI,CAAC,iBAAkB,oBAAoB,YAAY,IACnF,oBAAoB,YAAY,IAChC,aAAa,SAAU;AAC3B,QAAM,QAAgD,mBACpDA,kBAAoB,SAChB,MAAM,mBAAmBA,QAAO,SAAS,IACzC,MAAM,aAAaA,QAAO,SAAS;AAEzC,QAAM,UAAU,MAAM,eAAe,KAAK;AAC1C,QAAM,QAAQ,CAAC,MAAM,aAAgB,UAAU,EAAE,QAAQ,CAAC;AAC1D,SAAO;AACT;AAEA,eAAsB,kCACpB,cACAA,QAEA,eACA,gBAEA,qBACA;AACA,SAAO;AAAA,IACL,CAAC,YAAY;AAAA,IACbA;AAAA,IACA,OAAO,UAAU,cAAc,MAAM,EAAE;AAAA,IACvC;AAAA,IACA;AAAA,EACF;AACF;;;AC3CO,IAAMC,iBAAgB;AAEtB,IAAMC,eAAc;AAAA,EACzB,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAAA,EAC5B,IAAI,MAAM,UAAU,QAAQ;AAC9B;AAEO,IAAM,WAAqC,CAAC,SAAS,SAAS,MAAM;;;ACFpE,IAAM,mBAAN,cAA+B,eAAe;AAAA,EACnD,cAAc;AACZ,UAAM,SAAS;AAAA,MACb,oBAAoB;AAAA,MACpB,cAAcC;AAAA,MACd,SAAS,CAAC,MAAM;AAAA,MAChB,SAASC;AAAA,MACT,SAAS;AAAA,MACT,oBAAoB;AAAA,MACpB,aAAa,CAAC,GAAG,IAAI,IAAI,IAAI,KAAK,KAAK,GAAG;AAAA,IAC5C;AAEA,UAAM,MAAM;AAAA,EACd;AAAA,EAEA,IAAW,UAAmB;AAC5B,WAAO,KAAK,OAAO;AAAA,EACrB;AAAA,EAEA,MAAa,YAAYC,QAAkB,eAA6D;AACtG,UAAM,mBAAmB,MAAM,KAAK,OAAOA,QAAO,aAAa;AAC/D,WAAO,iBAAiB,IAAI,CAAC,QAAQ,IAAI,cAAc,IAAI,OAAO,IAAI,aAAa,EAAE,OAAO,IAAI,YAAY,QAAQ,IAAI,YAAY,CAAC,CAAC;AAAA,EACxI;AAAA,EAEmB,sBAA8B;AAC/C,WAAO;AAAA,EACT;AAAA,EAEmB,2BAA2B,WAA8F;AAC1I,WAAO,MAAM,2BAA2B,SAAS;AAAA,EACnD;AACF;;;ACzBO,IAAM,OAAO;AAAA,EAClB,gBAAgB,IAAI,eAAe;AAAA,EACnC,kBAAkB,IAAI,iBAAiB;AAAA,EACvC,YAAY,IAAI,WAAW;AAAA,EAC3B,mBAAmB,IAAI,kBAAkB;AAAA,EACzC,uBAAuB,IAAI,sBAAsB;AAAA,EACjD,oBAAoB,IAAI,mBAAmB;AAAA,EAC3C,mBAAmB,IAAI,kBAAkB;AAAA,EACzC,cAAc,IAAI,aAAa;AACjC;AASO,IAAM,iBAAiB,CAACC,QAAkB,YAA6D,KAAK,eAAe,YAAYA,QAAO,OAAO;AASrJ,IAAM,mBAAmB,CAACA,QAAkB,YAA+D,KAAK,iBAAiB,YAAYA,QAAO,OAAO;AAS3J,IAAM,aAAa,CAACA,QAAkB,YAA0D,KAAK,WAAW,YAAYA,QAAO,OAAO;AAS1I,IAAM,sBAAsB,CAACA,WAAmE,KAAK,kBAAkB,gBAAgBA,MAAK;AAW5I,IAAM,0BAA0B,CAACA,WAAmE,KAAK,sBAAsB,gBAAgBA,MAAK;AAYpJ,IAAM,wBAAwB,CAACA,WAA6D,KAAK,mBAAmB,sBAAsBA,MAAK;AAS/I,IAAM,2BAA2B,CAACA,WAAmE,KAAK,kBAAkB,mBAAmBA,MAAK;AASpJ,IAAM,sBAAsB,CAACA,WAAiF,KAAK,aAAa,oBAAoBA,MAAK;AAEzJ,IAAM,0BAA0B,CAAC,QAAgB,KAAK,eAAe,KAAK,GAAG;AAC7E,IAAM,4BAA4B,CAAC,QAAgB,KAAK,iBAAiB,KAAK,GAAG;AACjF,IAAM,sBAAsB,CAAC,QAAgB,KAAK,WAAW,KAAK,GAAG;AACrE,IAAM,wBAAwB,CAAC,QAAgB,KAAK,kBAAkB,KAAK,GAAG;AAC9E,IAAM,4BAA4B,CAAC,QAAgB,KAAK,sBAAsB,KAAK,GAAG;AACtF,IAAM,2BAA2B,CAAC,QAAgB,KAAK,mBAAmB,KAAK,GAAG;AAClF,IAAM,0BAA0B,CAAC,QAAgB,KAAK,kBAAkB,KAAK,GAAG;AAChF,IAAM,qBAAqB,CAAC,QAAgB,KAAK,aAAa,KAAK,GAAG;AAGtE,IAAM,yBAAyB;AAC/B,IAAM,cAAc;AACpB,IAAM,kBAAkB;;;ACtGxB,IAAM,iCAAN,cAAqE,eAAwB;AAAA,EAClG,YAEY,YAEAC,QAEA,gBACV;AACA,UAAM;AANI;AAEA,iBAAAA;AAEA;AAAA,EAGZ;AACF;AAEO,IAAM,gCAAN,cAAmF,+BAA0E;AAAA,EAClK,MAAsB,MAA+C;AACnE,UAAM,gBAAgB,MAAM,KAAK;AAEjC,UAAM,wBAAwB,MAAM;AAAA,MAClC;AAAA,MACA,KAAK;AAAA,MACL,OAAO,UAAU,QAAQ;AAAA,QACvB,MAAM,IAAI,CAAC,SAAS,KAAK,kBAAkB,mBAAmB,IAAI,CAA6B;AAAA,MACjG;AAAA,MACA,KAAK;AAAA,IACP;AAEA,WAAO,cAAc;AAAA,MACnB,CAAC,cAAc,MAAM,0BAAmC,cAAc,sBAAsB,EAAE;AAAA,IAChG;AAAA,EACF;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,2BAA2B,MAAM,KAAK,KAAK;AAAA,EACxD;AACF;AAEO,IAAM,mCAAN,cAAsF,+BAA8F;AAAA,EACzL,MAAsB,MAAyD;AAC7E,UAAM,eAAe,MAAM,KAAK;AAChC,QAAI,CAAC,cAAc;AACjB,aAAO;AAAA,IACT;AAEA,UAAM,kBAAkB,MAAM;AAAA,MAC5B;AAAA,MACA,KAAK;AAAA,MACL,CAAC,SAAS,KAAK,kBAAkB,mBAAmB,IAAI;AAAA,MACxD,KAAK;AAAA,IACP;AAEA,WAAO,0BAA0B,cAAc,eAAe;AAAA,EAChE;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,8BAA8B,MAAM,KAAK,KAAK;AAAA,EAC3D;AACF;AAEO,IAAM,iDAAN,cAAuH,8BAAuC;AAAA,EAC1J,mBAAmB;AAC1B,WAAO,IAAI,4CAA4C,MAAM,KAAK,KAAK;AAAA,EACzE;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,8BAA8B,MAAM,KAAK,KAAK;AAAA,EAC3D;AACF;AAEO,IAAM,oDAAN,cAA0H,iCAA0C;AAAA,EAChK,mBAAmB;AAC1B,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AAAA,EAEA,qBAAqB;AACnB,WAAO,IAAI,gCAAgC,MAAM,KAAK,KAAK;AAAA,EAC7D;AACF;;;AC3EO,IAAM,8BAAN,cAAkE,eAAwB;AAAA,EAC/F,YAEY,YAEAC,QAEA,gBACV;AACA,UAAM;AANI;AAEA,iBAAAA;AAEA;AAAA,EAGZ;AACF;AAEO,IAAM,6BAAN,cAAgF,4BAAuE;AAAA,EAC5J,MAAsB,MAA+C;AACnE,UAAM,gBAAgB,MAAM,KAAK;AACjC,UAAM,qBAAqB,MAAM;AAAA,MAC/B;AAAA,MACA,KAAK;AAAA,MACL,OAAO,UAAU,QAAQ,IAAI,MAAM,IAAI,CAAC,SAAS,KAAK,aAAa,oBAAoB,IAAI,CAAoC,CAAC;AAAA,MAChI,KAAK;AAAA,IACP;AACA,WAAO,cAAc,IAAI,CAAC,cAAc,MAAM;AAC5C,YAAM,EAAE,KAAK,QAAQ,kBAAkB,IAAI,mBAAmB;AAC9D,aAAO,cAAc,iBAAiB,cAAc,QAAQ,iBAAiB,GAAG,GAAG;AAAA,IACrF,CAAC;AAAA,EACH;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,8BAA8B,MAAM,KAAK,KAAK;AAAA,EAC3D;AACF;AAEO,IAAM,gCAAN,cAAmF,4BAA2F;AAAA,EACnL,MAAsB,MAAyD;AAC7E,UAAM,eAAe,MAAM,KAAK;AAChC,QAAI,CAAC;AAAc,aAAO;AAC1B,UAAM,EAAE,KAAK,QAAQ,kBAAkB,IAAI,MAAM;AAAA,MAC/C;AAAA,MACA,KAAK;AAAA,MACL,CAAC,SAAS,KAAK,aAAa,oBAAoB,IAAI;AAAA,MACpD,KAAK;AAAA,IACP;AACA,WAAO,cAAc,iBAAiB,cAAc,QAAQ,iBAAiB,GAAG,GAAG;AAAA,EACrF;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,iCAAiC,MAAM,KAAK,KAAK;AAAA,EAC9D;AACF;AAEO,IAAM,8CAAN,cAAoH,2BAAoC;AAAA,EACpJ,sBAAsB;AAC7B,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,8BAA8B,MAAM,KAAK,KAAK;AAAA,EAC3D;AACF;AAEO,IAAM,iDAAN,cAAuH,8BAAuC;AAAA,EAC1J,sBAAsB;AAC7B,WAAO,IAAI,kDAAkD,MAAM,KAAK,KAAK;AAAA,EAC/E;AAAA,EAEA,qBAAqB;AACnB,WAAO,IAAI,gCAAgC,MAAM,KAAK,KAAK;AAAA,EAC7D;AACF;;;ACzEO,IAAM,iCAAN,cAAqE,eAAwB;AAAA,EAClG,YAEY,YAEAC,QACV;AACA,UAAM;AAJI;AAEA,iBAAAA;AAAA,EAGZ;AACF;AAEO,IAAM,gCAAN,cAAsG,+BAAyE;AAAA,EACpL,MAAsB,MAA8C;AAClE,UAAM,gBAAgB,MAAM,KAAK;AACjC,UAAM,cAAc,MAAM;AAAA,MACxB;AAAA,MACA,KAAK;AAAA,MACL,CAAC,UAAU,QAAQ,IAAI,MAAM,IAAI,CAAC,SAAS,KAAK,mBAAmB,sBAAsB,IAAI,CAA0B,CAAC;AAAA,MACxH;AAAA,MACA,CAAC,iBAAiB,aAAa,UAAU,MAAM,MAAM,EAAE,kBAAkB,KAAK,CAAC;AAAA,IACjF;AACA,WAAO,YAAY,IAAI,CAAC,YAAY,MAAM,yBAAkC,cAAc,IAAI,UAAU,CAAC;AAAA,EAC3G;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,4CAA4C,MAAM,KAAK,KAAK;AAAA,EACzE;AACF;AAEO,IAAM,kCAAN,cAAwG,+BAA6F;AAAA,EAC1M,MAAsB,MAAwD;AAC5E,UAAM,eAAe,MAAM,KAAK;AAChC,QAAI,CAAC;AAAc,aAAO;AAC1B,UAAM,aAAa,MAAM;AAAA,MACvB;AAAA,MACA,KAAK;AAAA,MACL,CAAC,SAAS,KAAK,mBAAmB,sBAAsB,IAAI;AAAA,MAC5D;AAAA,MAEA,CAACC,kBAAiBA,cAAa,UAAU,MAAM,MAAM,EAAE,kBAAkB,KAAK,CAAC;AAAA,IACjF;AACA,WAAO,yBAAyB,cAAc,UAAU;AAAA,EAC1D;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,kDAAkD,MAAM,KAAK,KAAK;AAAA,EAC/E;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AACF;;;ACnDO,IAAM,8BAAN,cAAkE,eAAwB;AAAA,EAC/F,YAEY,YAEAC,QAEA,oBACV;AACA,UAAM;AANI;AAEA,iBAAAA;AAEA;AAAA,EAGZ;AAAA,EAEA,IAAc,cAAyD;AACrE,WAAO,KAAK,qBACR,KAAK,wBACL,KAAK;AAAA,EACX;AACF;AAEO,IAAM,6BAAN,cAAgF,4BAAqE;AAAA,EAC1J,MAAsB,MAA6C;AACjE,UAAM,gBAAgB,MAAM,KAAK;AACjC,UAAM,aAAa,cAAc,IAAI,CAAC,QAAQ,IAAI,SAAS;AAC3D,UAAM,QAAgD,KAAK,iBAAoB,SAC3E,MAAM,mBAAmB,KAAK,OAAO,UAAU,IAC/C,MAAM,aAAa,KAAK,OAAO,UAAU;AAC7C,UAAM,sBAAsB,MAAM,QAAQ,IAAI,MAAM,IAAI,CAAC,SAAS,KAAK,YAAY,gBAAgB,IAAI,CAAC,CAAC;AACzG,UAAM,QAAQ,CAAC,MAAM,aAAgB,UAAU,EAAE,QAAQ,CAAC;AAC1D,UAAM,SAAS,cACZ,OAAO,CAAC,eAAe,MAAM,oBAAoB,EAAE,EACnD,IAAI,CAAC,cAAc,MAAM,wBAAiC,cAAc,oBAAoB,EAAE,CAAC;AAClG,WAAO;AAAA,EACT;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,4CAA4C,MAAM,KAAK,KAAK;AAAA,EACzE;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,8BAA8B,MAAM,KAAK,KAAK;AAAA,EAC3D;AACF;AAEO,IAAM,gCAAN,cAAmF,4BAAyF;AAAA,EACjL,MAAsB,MAAuD;AAC3E,UAAM,eAAe,MAAM,KAAK;AAChC,QAAI,CAAC,cAAc;AACjB,aAAO;AAAA,IACT;AACA,UAAM,EAAE,UAAU,IAAI;AACtB,UAAM,QAAgD,KAAK,iBAAoB,SAC3E,MAAM,mBAAmB,KAAK,OAAO,CAAC,SAAS,CAAC,IAChD,MAAM,aAAa,KAAK,OAAO,CAAC,SAAS,CAAC;AAC9C,UAAM,YAAY,MAAM,KAAK,YAAY,gBAAgB,MAAM,EAAE;AACjE,UAAM,QAAQ,CAAC,MAAM,aAAgB,UAAU,EAAE,QAAQ,CAAC;AAC1D,WAAO,wBAAiC,cAAc,SAAS;AAAA,EACjE;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI,kDAAkD,MAAM,KAAK,KAAK;AAAA,EAC/E;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI,+CAA+C,MAAM,KAAK,KAAK;AAAA,EAC5E;AAAA,EAEA,qBAAqB;AACnB,WAAO,IAAI,gCAAgC,MAAM,KAAK,KAAK;AAAA,EAC7D;AACF;;;AC1EO,IAAM,sBAAN,cAA2C,eAAwB;AAAA,EAExE,YAAsBC,QAA4B,UAAgC,IAAI,sBAAsB,GAAG;AAC7G,UAAM;AADc,iBAAAA;AAA4B;AAAA,EAElD;AACF;AAEO,IAAM,qBAAN,cAAiC,oBAAqC;AAAA,EAC3E,MAAsB,MAAgC;AACpD,UAAM,EAAE,OAAAA,QAAO,QAAQ,IAAI;AAC3B,QAAI;AACJ,QAAI,mBAAmB;AAAyB,eAAS,KAAK,iBAAiB,YAAYA,QAAO,OAAO;AAAA,aAChG,mBAAmB;AAAuB,eAAS,KAAK,eAAe,YAAYA,QAAO,OAAO;AAAA,aACjG,mBAAmB;AAAmB,eAAS,KAAK,WAAW,YAAYA,QAAO,OAAO;AAAA;AAC7F,YAAM,IAAI,MAAM,sHAAsH;AAC3I,WAAO;AAAA,EACT;AAAA,EAEQ,iCAAmE;AACzE,WAAO,IAAI,QAAiC,CAAC,SAAS,WAAW;AAC/D,WAAK,IAAI,EACN,KAAK,CAAC,eAAe,QAAQ,WAAW,IAAI,CAAC,cAAc,wBAAwB,CAAC,GAAG,SAAS,CAAC,CAAC,CAAC,EACnG,MAAM,CAAC,QAAQ,OAAO,GAAG,CAAC;AAAA,IAC/B,CAAC;AAAA,EACH;AAAA,EAEA,kBAAkB,qBAAqB,OAAO;AAC5C,WAAO,IAAI;AAAA,MACT,KAAK,+BAA+B;AAAA,MACpC,KAAK;AAAA,MACL;AAAA,IACF;AAAA,EACF;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI;AAAA,MACT,KAAK,+BAA+B;AAAA,MACpC,KAAK;AAAA,IACP;AAAA,EACF;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI;AAAA,MACT,KAAK,+BAA+B;AAAA,MACpC,KAAK;AAAA,IACP;AAAA,EACF;AACF;AAEO,IAAM,uBAAN,cAAmC,oBAA+C;AAAA,EACvF,MAAsB,MAA0C;AAC9D,UAAM,iBAAiB,MAAM,IAAI,mBAAmB,KAAK,OAAO,KAAK,OAAO;AAC5E,QAAI,gCAAgC,eAAe;AACnD,mBAAe,QAAQ,CAAC,kBAAkB;AACxC,UAAI,cAAc,QAAQ,8BAA8B;AAAO,wCAAgC;AAAA,IACjG,CAAC;AACD,WAAO;AAAA,EACT;AAAA,EAEQ,gCAA4E;AAElF,WAAO,IAAI,QAA2C,OAAO,YAAY;AACvE,YAAM,YAAY,MAAM,KAAK,IAAI;AACjC,cAAQ,YAAY,wBAA4B,CAAC,GAAG,SAAS,IAAI,MAAS;AAAA,IAC5E,CAAC;AAAA,EACH;AAAA,EAEA,kBAAkB,qBAAqB,OAAO;AAC5C,WAAO,IAAI;AAAA,MACT,KAAK,8BAA8B;AAAA,MACnC,KAAK;AAAA,MACL;AAAA,IACF;AAAA,EACF;AAAA,EAEA,sBAAsB;AACpB,WAAO,IAAI;AAAA,MACT,KAAK,8BAA8B;AAAA,MACnC,KAAK;AAAA,IACP;AAAA,EACF;AAAA,EAEA,mBAAmB;AACjB,WAAO,IAAI;AAAA,MACT,KAAK,8BAA8B;AAAA,MACnC,KAAK;AAAA,IACP;AAAA,EACF;AACF;;;ACjGO,SAAS,iBAAiBC,QAAkB,UAAgC,IAAI,sBAAsB,GAAyB;AACpI,SAAO,IAAI,qBAAqBA,QAAO,OAAO;AAChD;AAEO,SAAS,eAAeA,QAAkB,UAAgC,IAAI,sBAAsB,GAAuB;AAChI,SAAO,IAAI,mBAAmBA,QAAO,OAAO;AAC9C;;;ACLA,eAAsB,uBAAuBC,QAAkB,eAAiG;AAC9J,SAAO,eAAeA,QAAO,IAAI,sBAAsB,gBAAgB,EAAE,cAAc,IAAI,CAAC,CAAC,CAAC,EAC3F,kBAAkB,EAClB,oBAAoB;AACzB;AAEA,eAAsB,mBAAmBA,QAAkB,gBAAoC,CAAC,GAA4E;AAC1K,SAAO,eAAeA,QAAO,IAAI,kBAAkB,aAAa,CAAC,EAC9D,kBAAkB,EAClB,oBAAoB;AACzB;AAEO,IAAM,WAAW;;;AClBjB,SAAS,kBAAkB,MAA+B,MAA+B;AAC9F,MAAI,KAAK,WAAW,KAAK;AAAQ,UAAM,IAAI,MAAM,gDAAgD;AACjG,QAAM,QAAQ,MAAM,KAAK,IAAI;AAC7B,QAAM,QAAQ,MAAM,KAAK,IAAI;AAC7B,SAAO,KAAK;AAAA,IACV,MACG,IAAI,CAAC,KAAK,MAAM,MAAM,MAAM,EAAE,EAC9B,OAAO,CAAC,KAAK,SAAS,MAAO,OAAO,MAAO,CAAC;AAAA,EACjD;AACF;;;ACJO,IAAM,cAAN,MAAkB;AAAA,EAIvB,YAAY,QAAkJ,oBAAoB,KAAK;AACrL,SAAK,qBAAqB;AAC1B,UAAM,aAAa,MAAM,QAAQ,MAAM,IAAI,SAAS,CAAC,MAAM;AAC3D,QAAI,CAAC,WAAW;AAAQ,YAAM,IAAI,MAAM,yDAAyD;AACjG,QAAIC,SAAQ;AACZ,UAAM,oBAAoB,MAAM,UAAUA;AAC1C,SAAK,sBAAsB,WAAW,IAAI,CAAC,SAAS;AAClD,UAAI,gBAAgB;AAAwB,eAAO;AACnD,UAAI,gBAAgB;AAAc,eAAO,IAAI,uBAAuB,kBAAkB,GAAG,CAAC,IAAI,CAAC;AAC/F,UAAI,KAAK,cAAc,KAAK,sBAAsB;AAAc,eAAO,IAAI,uBAAuB,kBAAkB,GAAG,CAAC,KAAK,UAAU,CAAC;AACxI,YAAM,IAAI,MAAM,qMAAqM;AAAA,IACvN,CAAC;AAAA,EACH;AAAA,EAEA,IAAW,qBAA+C;AAAE,WAAO,KAAK;AAAA,EAAqB;AAAA,EAE7F,IAAW,oBAA4B;AAAE,WAAO,KAAK;AAAA,EAAoB;AAAA,EAElE,oBAAoB,iBAA+B,aAAqC;AAC7F,WAAO,YACJ,IAAI,CAAC,MAAM,kBAAkB,GAAG,eAAe,CAAC,EAChD,OAAO,CAAC,IAAI,OAAO,KAAK,IAAI,CAAC,KAAK,YAAY,UAAU;AAAA,EAC7D;AAAA,EAEO,gBAAgB,iBAA0C;AAC/D,WAAO,KAAK,mBACT,IAAI,CAAC,EAAE,aAAa,MAAM,MAAM,IAAI,UAAU,OAAO,KAAK,oBAAoB,iBAAiB,WAAW,CAAC,CAAC,EAC5G,OAAO,CAAC,MAAM,SAAU,KAAK,WAAW,KAAK,WAAW,OAAO,IAAK;AAAA,EACzE;AAAA,EAEO,cAAc,iBAA0C;AAC7D,UAAM,YAAY,KAAK,gBAAgB,eAAe;AACtD,WAAQ,UAAU,WAAW,KAAK,qBAAsB,YAAY,IAAI,UAAU,WAAW,UAAU,QAAQ;AAAA,EACjH;AAAA,EAEO,SAAc;AACnB,WAAO;AAAA,MACL,mBAAmB,KAAK;AAAA,MACxB,oBAAoB,KAAK,oBAAoB,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;AAAA,IACtE;AAAA,EACF;AAAA,EAEA,OAAc,SAASC,QAAwB;AAC7C,UAAM,qBAAqBA,OAAK,mBAAmB,IAAI,CAAC,OAAY,uBAAuB,SAAS,EAAE,CAAC;AACvG,WAAO,IAAI,YAAY,oBAAoBA,OAAK,iBAAiB;AAAA,EACnE;AACF;;;AClDO,SAAS,uBAAuB,SAAuB;AAC5D,QAAM,MAAM,IAAI,iBAAiB;AACjC,MAAI,eAAe,OAAO;AAC1B,SAAO;AACT;;;ACHO,SAAS,cAAiB,SAAY,YAA4B;AACvE,QAAM,EAAE,OAAO,OAAO,IAAI,IAAI,WAAW,WAAW,OAAO,WAAW,MAAM;AAE5E,MAAI,SAAS,KAAK,UAAU,GAAG;AAC7B,UAAM,IAAI,MAAM,uCAAuC,KAAK,UAAU,EAAE,OAAO,OAAO,CAAC,GAAG;AAAA,EAC5F;AAEA,MAAI,MAAM,QAAQ,OAAO,GAAG;AAE1B,WAAQ,QAAuB,IAAI,CAAC,QAAQ,cAAc,KAAK,EAAE,OAAO,OAAO,CAAgB,CAAC;AAAA,EAClG;AAEA,MAAI,oBAAoB,OAAO,GAAG;AAChC,UAAM,mBAAmB,QAAQ,UAAU,QAAQ,OAAO,MAAM;AAChE,UAAM,mBAAmB,QAAQ,mBAAmB,QAAQ,iBAAiB,IAAI,OAAO,iBAAiB,IAAI,MAAM;AACnH,WAAO,wBAAwB,wBAAwB,SAAS,gBAAgB,GAAG,gBAAgB;AAAA,EACrG;AAEA,MAAI,oBAAoB,OAAO,GAAG;AAChC,WAAO,wBAAwB,SAAS,QAAQ,UAAU,QAAQ,OAAO,MAAM,CAAC;AAAA,EAClF;AAEA,MAAI,mBAAmB,iBAAiB,mBAAmB,eAAe;AACxE,WAAQ,QAAgB,QAAQ,OAAO,MAAM;AAAA,EAC/C;AAEA,SAAO;AACT;;;ACTO,IAAMC,WAAcA;",
+  "names": ["__defProp", "__require", "__require2", "__export", "isEven2", "environment", "epsilon3", "scale2", "iou", "isTensor", "isEven", "range", "round", "tensor", "isEven", "round", "scale", "sum", "range", "step", "min", "max", "nonMaxSuppression", "array", "sigmoid", "round", "json", "fetch", "buffer", "createVideoElement", "fetch", "isBrowser", "env", "isBrowser", "env", "CanvasRenderingContext2D", "env", "AnchorPosition", "round", "env", "env", "env", "input", "env", "createCanvas", "env", "ImageData", "env", "input", "env", "input", "env", "createCanvas", "scale", "input", "env", "range", "input", "input", "env", "createCanvas", "init", "fetch", "env", "env", "input", "tensor", "variable", "env", "loadWeights", "objProp", "tensor", "conv2", "conv2", "input", "extractParams", "extractParamsFromWeightMap", "input", "extractParams", "extractParamsFromWeightMap", "input", "round", "extractorsFactory", "extractParams", "range", "loadParamsFactory", "extractParamsFromWeightMap", "range", "input", "range", "extractParamsFromWeightMap", "extractParams", "extractParams", "extractParamsFromWeightMap", "Gender", "input", "extractParams", "extractParamsFromWeightMap", "scale", "getPadding", "input", "isEven", "input", "scale", "convLayer", "scale", "conv", "extractorsFactory", "conv", "scale", "conv2", "extractParams", "extractorsFactory", "scale", "extractParamsFromWeightMap", "conv", "zeros", "input", "extractParamsFromWeightMap", "extractParams", "extractorsFactory", "extractParams", "extractorsFactory", "extractParamsFromWeightMap", "epsilon", "nonMaxSuppression", "iou", "conv2", "input", "nonMaxSuppression", "extractParamsFromWeightMap", "extractParams", "DEFAULT_MODEL_NAME", "isNumber", "min", "depthwiseSeparableConv", "extractorsFactory", "sub", "conv", "extractParams", "conv2", "extractorsFactory", "sub", "conv", "extractParamsFromWeightMap", "depthwiseSeparableConv", "input", "nonMaxSuppression", "extractParamsFromWeightMap", "extractParams", "sigmoid", "max", "input", "DEFAULT_MODEL_NAME", "input", "IOU_THRESHOLD", "BOX_ANCHORS", "IOU_THRESHOLD", "BOX_ANCHORS", "input", "input", "input", "input", "input", "parentResult", "input", "input", "input", "input", "count", "json", "version"]
 }
diff --git a/dist/face-api.js b/dist/face-api.js
index f3f0724..44169ca 100644
--- a/dist/face-api.js
+++ b/dist/face-api.js
@@ -4,65 +4,65 @@
   author: <https://github.com/vladmandic>'
 */
 
-"use strict";var faceapi=(()=>{var cb=Object.defineProperty;var pF=Object.getOwnPropertyDescriptor;var cF=Object.getOwnPropertyNames;var dF=Object.prototype.hasOwnProperty;var Xr=(e=>typeof require!="undefined"?require:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof require!="undefined"?require:t)[n]}):e)(function(e){if(typeof require!="undefined")return require.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')});var rh=(e,t)=>{for(var n in t)cb(e,n,{get:t[n],enumerable:!0})},hF=(e,t,n,a)=>{if(t&&typeof t=="object"||typeof t=="function")for(let r of cF(t))!dF.call(e,r)&&r!==n&&cb(e,r,{get:()=>t[r],enumerable:!(a=pF(t,r))||a.enumerable});return e};var mF=e=>hF(cb({},"__esModule",{value:!0}),e);var Lce={};rh(Lce,{AgeGenderNet:()=>wd,BoundingBox:()=>Co,Box:()=>st,ComposableTask:()=>ka,ComputeAllFaceDescriptorsTask:()=>Gr,ComputeFaceDescriptorsTaskBase:()=>Td,ComputeSingleFaceDescriptorTask:()=>Hr,DetectAllFaceLandmarksTask:()=>Cd,DetectAllFacesTask:()=>xp,DetectFaceLandmarksTaskBase:()=>Nd,DetectFacesTaskBase:()=>Ed,DetectSingleFaceLandmarksTask:()=>_d,DetectSingleFaceTask:()=>Ad,Dimensions:()=>yn,FACE_EXPRESSION_LABELS:()=>D1,FaceDetection:()=>yt,FaceDetectionNet:()=>W1,FaceExpressionNet:()=>vd,FaceExpressions:()=>Br,FaceLandmark68Net:()=>Po,FaceLandmark68TinyNet:()=>kd,FaceLandmarkNet:()=>O1,FaceLandmarks:()=>ra,FaceLandmarks5:()=>x1,FaceLandmarks68:()=>Eo,FaceMatch:()=>sp,FaceMatcher:()=>$d,FaceRecognitionNet:()=>Oo,Gender:()=>gg,LabeledBox:()=>ip,LabeledFaceDescriptors:()=>gr,NetInput:()=>br,NeuralNetwork:()=>sn,ObjectDetection:()=>Lr,Point:()=>De,PredictedBox:()=>v1,Rect:()=>_o,SsdMobilenetv1:()=>As,SsdMobilenetv1Options:()=>wa,TinyFaceDetector:()=>Vo,TinyFaceDetectorOptions:()=>Sd,TinyYolov2:()=>Wo,TinyYolov2Options:()=>Qa,allFaces:()=>Mce,allFacesSsdMobilenetv1:()=>E$,allFacesTinyYolov2:()=>Rce,awaitMediaLoaded:()=>C1,bufferToImage:()=>_1,computeFaceDescriptor:()=>vce,createCanvas:()=>Do,createCanvasFromMedia:()=>gd,createFaceDetectionNet:()=>hce,createFaceRecognitionNet:()=>nce,createSsdMobilenetv1:()=>m$,createTinyFaceDetector:()=>Pce,createTinyYolov2:()=>gce,detectAllFaces:()=>Tg,detectFaceLandmarks:()=>C$,detectFaceLandmarksTiny:()=>xce,detectLandmarks:()=>Fce,detectSingleFace:()=>Dce,draw:()=>M1,env:()=>Je,euclideanDistance:()=>U1,extendWithAge:()=>vg,extendWithFaceDescriptor:()=>xg,extendWithFaceDetection:()=>Ao,extendWithFaceExpressions:()=>dg,extendWithFaceLandmarks:()=>fp,extendWithGender:()=>wg,extractFaceTensors:()=>up,extractFaces:()=>lp,fetchImage:()=>Vpe,fetchJson:()=>$1,fetchNetWeights:()=>Upe,fetchOrThrow:()=>Wr,fetchVideo:()=>Gpe,getContext2dOrThrow:()=>Gn,getMediaDimensions:()=>Fo,imageTensorToCanvas:()=>E1,imageToSquare:()=>A1,inverseSigmoid:()=>Mpe,iou:()=>f1,isMediaElement:()=>rg,isMediaLoaded:()=>fd,isWithAge:()=>ace,isWithFaceDetection:()=>yr,isWithFaceExpressions:()=>R1,isWithFaceLandmarks:()=>Mo,isWithGender:()=>rce,loadAgeGenderModel:()=>Ece,loadFaceDetectionModel:()=>Ace,loadFaceExpressionModel:()=>_ce,loadFaceLandmarkModel:()=>Tce,loadFaceLandmarkTinyModel:()=>Nce,loadFaceRecognitionModel:()=>Cce,loadSsdMobilenetv1Model:()=>_$,loadTinyFaceDetectorModel:()=>Ice,loadTinyYolov2Model:()=>Sce,loadWeightMap:()=>F1,locateFaces:()=>$ce,matchDimensions:()=>Hpe,minBbox:()=>g1,nets:()=>Qe,nonMaxSuppression:()=>y1,normalize:()=>Za,padToSquare:()=>b1,predictAgeAndGender:()=>kce,recognizeFaceExpressions:()=>wce,resizeResults:()=>A$,resolveInput:()=>$o,shuffleArray:()=>Rpe,sigmoid:()=>dd,ssdMobilenetv1:()=>N$,tf:()=>Pe,tinyFaceDetector:()=>yce,tinyYolov2:()=>bce,toNetInput:()=>bt,utils:()=>m1,validateConfig:()=>B1,version:()=>Oce});var Pe={};rh(Pe,{Abs:()=>Dl,Acos:()=>Rl,Acosh:()=>Ml,AdadeltaOptimizer:()=>kf,AdagradOptimizer:()=>If,AdamOptimizer:()=>Sf,AdamaxOptimizer:()=>Tf,Add:()=>gs,AddN:()=>fi,All:()=>Pl,Any:()=>Ol,ArgMax:()=>gi,ArgMin:()=>cc,Asin:()=>Ll,Asinh:()=>zl,Atan:()=>Wl,Atan2:()=>Vl,Atanh:()=>Bl,AvgPool:()=>yi,AvgPool3D:()=>dc,AvgPool3DGrad:()=>cm,AvgPoolGrad:()=>pm,BackendWasm:()=>UA,BatchMatMul:()=>bi,BatchToSpaceND:()=>Ul,Bincount:()=>dm,BroadcastArgs:()=>hm,BroadcastTo:()=>tS,Callback:()=>x2,CallbackList:()=>IN,Cast:()=>xi,Ceil:()=>vi,ClipByValue:()=>ys,Complex:()=>mm,ComplexAbs:()=>hc,Concat:()=>Gl,Conv2D:()=>wi,Conv2DBackpropFilter:()=>fm,Conv2DBackpropInput:()=>ki,Conv3D:()=>mc,Conv3DBackpropFilterV2:()=>gm,Conv3DBackpropInputV2:()=>ym,Cos:()=>Ii,Cosh:()=>Si,CropAndResize:()=>jl,Cumprod:()=>Hl,Cumsum:()=>Ti,CustomCallback:()=>TN,DataStorage:()=>om,DenseBincount:()=>bm,DepthToSpace:()=>ql,DepthwiseConv2dNative:()=>Ni,DepthwiseConv2dNativeBackpropFilter:()=>xm,DepthwiseConv2dNativeBackpropInput:()=>vm,Diag:()=>wm,Dilation2D:()=>fc,Dilation2DBackpropFilter:()=>$h,Dilation2DBackpropInput:()=>Ah,ENV:()=>$x,EarlyStopping:()=>v2,Einsum:()=>km,Elu:()=>_i,EluGrad:()=>Im,Environment:()=>QI,Equal:()=>Xl,Erf:()=>Kl,Exp:()=>Ei,ExpandDims:()=>Yl,Expm1:()=>Zl,FFT:()=>Sm,Fill:()=>gc,FlipLeftRight:()=>Jl,Floor:()=>Ai,FloorDiv:()=>$i,FromPixels:()=>Fh,FusedBatchNorm:()=>Fi,FusedConv2D:()=>ei,FusedDepthwiseConv2D:()=>ti,GPGPUContext:()=>Th,GatherNd:()=>eu,GatherV2:()=>Ql,GraphModel:()=>A0,Greater:()=>tu,GreaterEqual:()=>Di,History:()=>SN,IFFT:()=>Tm,Identity:()=>Ri,Imag:()=>Nm,InputSpec:()=>zt,IsFinite:()=>nu,IsInf:()=>au,IsNan:()=>ru,KernelBackend:()=>pc,LRN:()=>yc,LRNGrad:()=>_m,LayerVariable:()=>gN,LayersModel:()=>Er,LeakyRelu:()=>Mi,Less:()=>su,LessEqual:()=>iu,LinSpace:()=>Cm,Log:()=>Pi,Log1p:()=>ou,LogSoftmax:()=>aS,LogicalAnd:()=>lu,LogicalNot:()=>uu,LogicalOr:()=>pu,LogicalXor:()=>nS,LowerBound:()=>rD,MathBackendWebGL:()=>Xf,Max:()=>Oi,MaxPool:()=>zi,MaxPool3D:()=>bc,MaxPool3DGrad:()=>Am,MaxPoolGrad:()=>Em,MaxPoolWithArgmax:()=>$m,Maximum:()=>Li,Mean:()=>Wi,Min:()=>Bi,Minimum:()=>Vi,MirrorPad:()=>Ui,Mod:()=>cu,MomentumOptimizer:()=>Nf,Multinomial:()=>Fm,Multiply:()=>Gi,Neg:()=>du,NonMaxSuppressionV3:()=>mu,NonMaxSuppressionV4:()=>fu,NonMaxSuppressionV5:()=>gu,NotEqual:()=>hu,OP_SCOPE_SUFFIX:()=>Mx,OneHot:()=>Hi,OnesLike:()=>yu,Optimizer:()=>Rr,OptimizerConstructors:()=>Yr,Pack:()=>bu,PadV2:()=>ji,Pool:()=>sD,Pow:()=>qi,Prelu:()=>Ki,Prod:()=>Xi,RMSPropOptimizer:()=>Cf,RNN:()=>mr,RaggedGather:()=>Dm,RaggedRange:()=>Rm,RaggedTensorToTensor:()=>Mm,Range:()=>xc,Rank:()=>Eb,Real:()=>Pm,RealDiv:()=>Ci,Reciprocal:()=>xu,Reduction:()=>vn,Relu:()=>Yi,Relu6:()=>Qi,Reshape:()=>vu,ResizeBilinear:()=>Ji,ResizeBilinearGrad:()=>Lm,ResizeNearestNeighbor:()=>Zi,ResizeNearestNeighborGrad:()=>Om,Reverse:()=>eo,RotateWithOffset:()=>Pu,Round:()=>to,Rsqrt:()=>no,SGDOptimizer:()=>qc,ScatterNd:()=>wu,SearchSorted:()=>zm,Select:()=>ku,Selu:()=>Iu,Sequential:()=>Tl,Sigmoid:()=>ro,Sign:()=>Nu,Sin:()=>ao,Sinh:()=>Tu,Slice:()=>Su,Softmax:()=>oo,Softplus:()=>Cu,SpaceToBatchND:()=>_u,SparseFillEmptyRows:()=>vc,SparseReshape:()=>Au,SparseSegmentMean:()=>wc,SparseSegmentSum:()=>kc,SparseToDense:()=>Wm,SplitV:()=>Eu,Sqrt:()=>so,Square:()=>Ic,SquaredDifference:()=>lo,Step:()=>xs,StridedSlice:()=>$u,StringNGrams:()=>Sc,StringSplit:()=>Tc,StringToHashBucketFast:()=>Nc,Sub:()=>uo,Sum:()=>io,SymbolicTensor:()=>Ba,Tan:()=>po,Tanh:()=>co,Tensor:()=>Te,TensorBuffer:()=>Ht,Tile:()=>bs,TopK:()=>Fu,Transform:()=>Du,Transpose:()=>_r,Unique:()=>Bm,Unpack:()=>Ru,UnsortedSegmentSum:()=>Cc,UpperBound:()=>iD,Variable:()=>is,ZerosLike:()=>Mu,_FusedMatMul:()=>Qs,abs:()=>Lt,acos:()=>Xx,acosh:()=>Yx,add:()=>Y,addN:()=>KS,all:()=>jm,any:()=>Qp,argMax:()=>ri,argMin:()=>Zx,asin:()=>Jx,asinh:()=>Qx,atan:()=>ev,atan2:()=>tv,atanh:()=>nv,avgPool:()=>ya,avgPool3d:()=>rv,backend:()=>AS,backend_util:()=>N,basicLSTMCell:()=>JS,batchNorm:()=>ks,batchNorm2d:()=>sv,batchNorm3d:()=>iv,batchNorm4d:()=>ov,batchToSpaceND:()=>Pc,bincount:()=>lv,booleanMaskAsync:()=>RT,broadcastArgs:()=>QS,broadcastTo:()=>Xs,broadcast_util:()=>Ou,browser:()=>ho,buffer:()=>Oe,callbacks:()=>EH,cast:()=>oe,ceil:()=>uv,clipByValue:()=>en,clone:()=>ir,complex:()=>Ar,concat:()=>Ze,concat1d:()=>pv,concat2d:()=>cv,concat3d:()=>dv,concat4d:()=>hv,constraints:()=>xN,conv1d:()=>qm,conv2d:()=>$t,conv2dTranspose:()=>Km,conv3d:()=>fv,conv3dTranspose:()=>gv,copyRegisteredKernels:()=>pD,cos:()=>Oc,cosh:()=>Xm,cosineWindow:()=>yf,cumprod:()=>ec,cumsum:()=>Ym,customGrad:()=>ur,data:()=>G2,denseBincount:()=>zh,deprecationWarn:()=>Vx,depthToSpace:()=>yv,depthwiseConv2d:()=>Is,deregisterOp:()=>FH,device_util:()=>$c,diag:()=>tT,dilation2d:()=>bv,disableDeprecationWarnings:()=>LR,dispose:()=>_e,disposeVariables:()=>zR,div:()=>he,divNoNan:()=>xv,dot:()=>vv,dropout:()=>qv,einsum:()=>nT,elu:()=>Lu,enableDebugMode:()=>OR,enableProdMode:()=>PR,enclosingPowerOfTwo:()=>Kv,engine:()=>_a,env:()=>H,equal:()=>ea,erf:()=>wv,euclideanNorm:()=>Sv,exp:()=>fn,expandDims:()=>Zt,expm1:()=>Tv,eye:()=>Zm,fft:()=>Hc,fill:()=>gn,findBackend:()=>jR,findBackendFactory:()=>qR,floor:()=>Wu,floorDiv:()=>Hm,forceHalfFloat:()=>uE,fused:()=>kl,gather:()=>Bu,gatherND:()=>LT,gather_util:()=>Ux,getBackend:()=>GR,getGradient:()=>Cb,getKernel:()=>Dh,getKernelsForBackend:()=>Rh,getThreadsCount:()=>Tpe,gpgpu_util:()=>B_,grad:()=>yO,grads:()=>bO,greater:()=>Cn,greaterEqual:()=>Fr,ifft:()=>wl,imag:()=>Dc,image:()=>za,inTopKAsync:()=>zT,initializers:()=>vN,input:()=>zN,io:()=>Ut,irfft:()=>hf,isFinite:()=>Nv,isInf:()=>Cv,isNaN:()=>_v,keep:()=>Jt,kernel_impls:()=>hr,layers:()=>wN,leakyRelu:()=>Lc,less:()=>Jm,lessEqual:()=>Ss,linalg:()=>Zv,linspace:()=>oT,loadGraphModel:()=>M6,loadGraphModelSync:()=>P6,loadLayersModel:()=>AU,localResponseNormalization:()=>Ev,log:()=>ta,log1p:()=>zc,logSigmoid:()=>Av,logSoftmax:()=>ef,logSumExp:()=>tf,logicalAnd:()=>$a,logicalNot:()=>Wc,logicalOr:()=>nf,logicalXor:()=>$v,losses:()=>ZT,lowerBound:()=>uT,matMul:()=>Fe,math:()=>ES,max:()=>ma,maxPool:()=>Dt,maxPool3d:()=>Fv,maxPoolWithArgmax:()=>pT,maximum:()=>dr,mean:()=>Nt,memory:()=>Oh,meshgrid:()=>cT,metrics:()=>g2,min:()=>bl,minimum:()=>Vu,mirrorPad:()=>Dv,mod:()=>Rv,model:()=>DU,models:()=>y2,moments:()=>Bc,movingAverage:()=>MT,mul:()=>z,multiRNNCell:()=>dT,multinomial:()=>hT,neg:()=>vt,nextFrame:()=>Jv,norm:()=>zu,notEqual:()=>oi,oneHot:()=>gl,ones:()=>Jn,onesLike:()=>na,op:()=>L,outerProduct:()=>mT,pad:()=>ba,pad1d:()=>fT,pad2d:()=>gT,pad3d:()=>yT,pad4d:()=>bT,pool:()=>Mv,pow:()=>$r,prelu:()=>Uc,print:()=>zx,prod:()=>Pv,profile:()=>WR,raggedGather:()=>xT,raggedRange:()=>vT,raggedTensorToTensor:()=>wT,rand:()=>kT,randomGamma:()=>IT,randomNormal:()=>rf,randomStandardNormal:()=>ST,randomUniform:()=>Uu,range:()=>xl,ready:()=>UR,real:()=>yl,reciprocal:()=>zv,registerBackend:()=>Gm,registerCallbackConstructor:()=>MU,registerGradient:()=>rS,registerKernel:()=>_c,registerOp:()=>$H,regularizers:()=>b2,relu:()=>Xe,relu6:()=>sf,removeBackend:()=>HR,reshape:()=>W,reverse:()=>ga,reverse1d:()=>TT,reverse2d:()=>NT,reverse3d:()=>CT,reverse4d:()=>_T,rfft:()=>jc,round:()=>of,rsqrt:()=>lf,scalar:()=>be,scatterND:()=>PT,scatter_util:()=>Gx,searchSorted:()=>af,selu:()=>uf,separableConv2d:()=>Ts,sequential:()=>RU,serialization:()=>ne,setBackend:()=>VR,setPlatform:()=>KR,setThreadsCount:()=>Spe,setWasmPath:()=>kpe,setWasmPaths:()=>Ipe,setWebGLContext:()=>c_,setdiff1dAsync:()=>ET,sigmoid:()=>ha,sign:()=>Wv,signal:()=>YT,sin:()=>pf,sinh:()=>cf,slice:()=>Be,slice1d:()=>Gc,slice2d:()=>df,slice3d:()=>fo,slice4d:()=>vl,slice_util:()=>jt,softmax:()=>Ka,softplus:()=>mo,spaceToBatchND:()=>Vc,sparse:()=>JT,sparseToDense:()=>OT,spectral:()=>XT,split:()=>zn,sqrt:()=>un,square:()=>lt,squaredDifference:()=>mf,squeeze:()=>Ns,stack:()=>Ft,step:()=>go,stridedSlice:()=>Bv,string:()=>QT,sub:()=>pe,sum:()=>fe,sumOutType:()=>Um,tan:()=>Vv,tanh:()=>si,tensor:()=>In,tensor1d:()=>Ke,tensor2d:()=>Aa,tensor3d:()=>Rc,tensor4d:()=>Da,tensor5d:()=>AT,tensor6d:()=>$T,tensor_util:()=>Va,test_util:()=>HS,tidy:()=>P,tile:()=>Ln,time:()=>BR,topk:()=>Uv,train:()=>Ws,transpose:()=>Ee,truncatedNormal:()=>ff,unique:()=>Gv,unregisterGradient:()=>uD,unregisterKernel:()=>lD,unsortedSegmentSum:()=>gf,unstack:()=>ct,upcastType:()=>fa,upperBound:()=>FT,util:()=>v,valueAndGrad:()=>xO,valueAndGrads:()=>vO,variable:()=>Hv,variableGrads:()=>lT,version:()=>Fpe,version_converter:()=>L6,version_core:()=>NM,version_layers:()=>vw,version_wasm:()=>Npe,version_webgl:()=>fJ,webgl:()=>gJ,webgl_util:()=>p_,where:()=>mn,whereAsync:()=>jv,zeros:()=>It,zerosLike:()=>qe});var fF=Object.create,Nx=Object.defineProperty,gF=Object.getOwnPropertyDescriptor,yF=Object.getOwnPropertyNames,bF=Object.getPrototypeOf,xF=Object.prototype.hasOwnProperty,BI=(e=>typeof Xr!="undefined"?Xr:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof Xr!="undefined"?Xr:t)[n]}):e)(function(e){if(typeof Xr!="undefined")return Xr.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')}),Bt=(e,t)=>()=>(t||e((t={exports:{}}).exports,t),t.exports),Ae=(e,t)=>{for(var n in t)Nx(e,n,{get:t[n],enumerable:!0})},vF=(e,t,n,a)=>{if(t&&typeof t=="object"||typeof t=="function")for(let r of yF(t))!xF.call(e,r)&&r!==n&&Nx(e,r,{get:()=>t[r],enumerable:!(a=gF(t,r))||a.enumerable});return e},ms=(e,t,n)=>(n=e!=null?fF(bF(e)):{},vF(t||!e||!e.__esModule?Nx(n,"default",{value:e,enumerable:!0}):n,e)),wF=Bt((e,t)=>{t.exports=a;var n=null;try{n=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(S){}function a(S,M,B){this.low=S|0,this.high=M|0,this.unsigned=!!B}a.prototype.__isLong__,Object.defineProperty(a.prototype,"__isLong__",{value:!0});function r(S){return(S&&S.__isLong__)===!0}a.isLong=r;var s={},i={};function o(S,M){var B,U,G;return M?(S>>>=0,(G=0<=S&&S<256)&&(U=i[S],U)?U:(B=u(S,(S|0)<0?-1:0,!0),G&&(i[S]=B),B)):(S|=0,(G=-128<=S&&S<128)&&(U=s[S],U)?U:(B=u(S,S<0?-1:0,!1),G&&(s[S]=B),B))}a.fromInt=o;function l(S,M){if(isNaN(S))return M?w:x;if(M){if(S<0)return w;if(S>=g)return A}else{if(S<=-y)return R;if(S+1>=y)return E}return S<0?l(-S,M).neg():u(S%f|0,S/f|0,M)}a.fromNumber=l;function u(S,M,B){return new a(S,M,B)}a.fromBits=u;var p=Math.pow;function d(S,M,B){if(S.length===0)throw Error("empty string");if(S==="NaN"||S==="Infinity"||S==="+Infinity"||S==="-Infinity")return x;if(typeof M=="number"?(B=M,M=!1):M=!!M,B=B||10,B<2||36<B)throw RangeError("radix");var U;if((U=S.indexOf("-"))>0)throw Error("interior hyphen");if(U===0)return d(S.substring(1),M,B).neg();for(var G=l(p(B,8)),q=x,K=0;K<S.length;K+=8){var Z=Math.min(8,S.length-K),Q=parseInt(S.substring(K,K+Z),B);if(Z<8){var ee=l(p(B,Z));q=q.mul(ee).add(l(Q))}else q=q.mul(G),q=q.add(l(Q))}return q.unsigned=M,q}a.fromString=d;function c(S,M){return typeof S=="number"?l(S,M):typeof S=="string"?d(S,M):u(S.low,S.high,typeof M=="boolean"?M:S.unsigned)}a.fromValue=c;var h=1<<16,m=1<<24,f=h*h,g=f*f,y=g/2,b=o(m),x=o(0);a.ZERO=x;var w=o(0,!0);a.UZERO=w;var I=o(1);a.ONE=I;var T=o(1,!0);a.UONE=T;var C=o(-1);a.NEG_ONE=C;var E=u(-1,2147483647,!1);a.MAX_VALUE=E;var A=u(-1,-1,!0);a.MAX_UNSIGNED_VALUE=A;var R=u(0,-2147483648,!1);a.MIN_VALUE=R;var F=a.prototype;F.toInt=function(){return this.unsigned?this.low>>>0:this.low},F.toNumber=function(){return this.unsigned?(this.high>>>0)*f+(this.low>>>0):this.high*f+(this.low>>>0)},F.toString=function(S){if(S=S||10,S<2||36<S)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative())if(this.eq(R)){var M=l(S),B=this.div(M),U=B.mul(M).sub(this);return B.toString(S)+U.toInt().toString(S)}else return"-"+this.neg().toString(S);for(var G=l(p(S,6),this.unsigned),q=this,K="";;){var Z=q.div(G),Q=q.sub(Z.mul(G)).toInt()>>>0,ee=Q.toString(S);if(q=Z,q.isZero())return ee+K;for(;ee.length<6;)ee="0"+ee;K=""+ee+K}},F.getHighBits=function(){return this.high},F.getHighBitsUnsigned=function(){return this.high>>>0},F.getLowBits=function(){return this.low},F.getLowBitsUnsigned=function(){return this.low>>>0},F.getNumBitsAbs=function(){if(this.isNegative())return this.eq(R)?64:this.neg().getNumBitsAbs();for(var S=this.high!=0?this.high:this.low,M=31;M>0&&(S&1<<M)==0;M--);return this.high!=0?M+33:M+1},F.isZero=function(){return this.high===0&&this.low===0},F.eqz=F.isZero,F.isNegative=function(){return!this.unsigned&&this.high<0},F.isPositive=function(){return this.unsigned||this.high>=0},F.isOdd=function(){return(this.low&1)===1},F.isEven=function(){return(this.low&1)===0},F.equals=function(S){return r(S)||(S=c(S)),this.unsigned!==S.unsigned&&this.high>>>31===1&&S.high>>>31===1?!1:this.high===S.high&&this.low===S.low},F.eq=F.equals,F.notEquals=function(S){return!this.eq(S)},F.neq=F.notEquals,F.ne=F.notEquals,F.lessThan=function(S){return this.comp(S)<0},F.lt=F.lessThan,F.lessThanOrEqual=function(S){return this.comp(S)<=0},F.lte=F.lessThanOrEqual,F.le=F.lessThanOrEqual,F.greaterThan=function(S){return this.comp(S)>0},F.gt=F.greaterThan,F.greaterThanOrEqual=function(S){return this.comp(S)>=0},F.gte=F.greaterThanOrEqual,F.ge=F.greaterThanOrEqual,F.compare=function(S){if(r(S)||(S=c(S)),this.eq(S))return 0;var M=this.isNegative(),B=S.isNegative();return M&&!B?-1:!M&&B?1:this.unsigned?S.high>>>0>this.high>>>0||S.high===this.high&&S.low>>>0>this.low>>>0?-1:1:this.sub(S).isNegative()?-1:1},F.comp=F.compare,F.negate=function(){return!this.unsigned&&this.eq(R)?R:this.not().add(I)},F.neg=F.negate,F.add=function(S){r(S)||(S=c(S));var M=this.high>>>16,B=this.high&65535,U=this.low>>>16,G=this.low&65535,q=S.high>>>16,K=S.high&65535,Z=S.low>>>16,Q=S.low&65535,ee=0,ae=0,te=0,le=0;return le+=G+Q,te+=le>>>16,le&=65535,te+=U+Z,ae+=te>>>16,te&=65535,ae+=B+K,ee+=ae>>>16,ae&=65535,ee+=M+q,ee&=65535,u(te<<16|le,ee<<16|ae,this.unsigned)},F.subtract=function(S){return r(S)||(S=c(S)),this.add(S.neg())},F.sub=F.subtract,F.multiply=function(S){if(this.isZero())return x;if(r(S)||(S=c(S)),n){var M=n.mul(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}if(S.isZero())return x;if(this.eq(R))return S.isOdd()?R:x;if(S.eq(R))return this.isOdd()?R:x;if(this.isNegative())return S.isNegative()?this.neg().mul(S.neg()):this.neg().mul(S).neg();if(S.isNegative())return this.mul(S.neg()).neg();if(this.lt(b)&&S.lt(b))return l(this.toNumber()*S.toNumber(),this.unsigned);var B=this.high>>>16,U=this.high&65535,G=this.low>>>16,q=this.low&65535,K=S.high>>>16,Z=S.high&65535,Q=S.low>>>16,ee=S.low&65535,ae=0,te=0,le=0,ie=0;return ie+=q*ee,le+=ie>>>16,ie&=65535,le+=G*ee,te+=le>>>16,le&=65535,le+=q*Q,te+=le>>>16,le&=65535,te+=U*ee,ae+=te>>>16,te&=65535,te+=G*Q,ae+=te>>>16,te&=65535,te+=q*Z,ae+=te>>>16,te&=65535,ae+=B*ee+U*Q+G*Z+q*K,ae&=65535,u(le<<16|ie,ae<<16|te,this.unsigned)},F.mul=F.multiply,F.divide=function(S){if(r(S)||(S=c(S)),S.isZero())throw Error("division by zero");if(n){if(!this.unsigned&&this.high===-2147483648&&S.low===-1&&S.high===-1)return this;var M=(this.unsigned?n.div_u:n.div_s)(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?w:x;var B,U,G;if(this.unsigned){if(S.unsigned||(S=S.toUnsigned()),S.gt(this))return w;if(S.gt(this.shru(1)))return T;G=w}else{if(this.eq(R)){if(S.eq(I)||S.eq(C))return R;if(S.eq(R))return I;var q=this.shr(1);return B=q.div(S).shl(1),B.eq(x)?S.isNegative()?I:C:(U=this.sub(S.mul(B)),G=B.add(U.div(S)),G)}else if(S.eq(R))return this.unsigned?w:x;if(this.isNegative())return S.isNegative()?this.neg().div(S.neg()):this.neg().div(S).neg();if(S.isNegative())return this.div(S.neg()).neg();G=x}for(U=this;U.gte(S);){B=Math.max(1,Math.floor(U.toNumber()/S.toNumber()));for(var K=Math.ceil(Math.log(B)/Math.LN2),Z=K<=48?1:p(2,K-48),Q=l(B),ee=Q.mul(S);ee.isNegative()||ee.gt(U);)B-=Z,Q=l(B,this.unsigned),ee=Q.mul(S);Q.isZero()&&(Q=I),G=G.add(Q),U=U.sub(ee)}return G},F.div=F.divide,F.modulo=function(S){if(r(S)||(S=c(S)),n){var M=(this.unsigned?n.rem_u:n.rem_s)(this.low,this.high,S.low,S.high);return u(M,n.get_high(),this.unsigned)}return this.sub(this.div(S).mul(S))},F.mod=F.modulo,F.rem=F.modulo,F.not=function(){return u(~this.low,~this.high,this.unsigned)},F.and=function(S){return r(S)||(S=c(S)),u(this.low&S.low,this.high&S.high,this.unsigned)},F.or=function(S){return r(S)||(S=c(S)),u(this.low|S.low,this.high|S.high,this.unsigned)},F.xor=function(S){return r(S)||(S=c(S)),u(this.low^S.low,this.high^S.high,this.unsigned)},F.shiftLeft=function(S){return r(S)&&(S=S.toInt()),(S&=63)===0?this:S<32?u(this.low<<S,this.high<<S|this.low>>>32-S,this.unsigned):u(0,this.low<<S-32,this.unsigned)},F.shl=F.shiftLeft,F.shiftRight=function(S){return r(S)&&(S=S.toInt()),(S&=63)===0?this:S<32?u(this.low>>>S|this.high<<32-S,this.high>>S,this.unsigned):u(this.high>>S-32,this.high>=0?0:-1,this.unsigned)},F.shr=F.shiftRight,F.shiftRightUnsigned=function(S){if(r(S)&&(S=S.toInt()),S&=63,S===0)return this;var M=this.high;if(S<32){var B=this.low;return u(B>>>S|M<<32-S,M>>>S,this.unsigned)}else return S===32?u(M,0,this.unsigned):u(M>>>S-32,0,this.unsigned)},F.shru=F.shiftRightUnsigned,F.shr_u=F.shiftRightUnsigned,F.toSigned=function(){return this.unsigned?u(this.low,this.high,!1):this},F.toUnsigned=function(){return this.unsigned?this:u(this.low,this.high,!0)},F.toBytes=function(S){return S?this.toBytesLE():this.toBytesBE()},F.toBytesLE=function(){var S=this.high,M=this.low;return[M&255,M>>>8&255,M>>>16&255,M>>>24,S&255,S>>>8&255,S>>>16&255,S>>>24]},F.toBytesBE=function(){var S=this.high,M=this.low;return[S>>>24,S>>>16&255,S>>>8&255,S&255,M>>>24,M>>>16&255,M>>>8&255,M&255]},a.fromBytes=function(S,M,B){return B?a.fromBytesLE(S,M):a.fromBytesBE(S,M)},a.fromBytesLE=function(S,M){return new a(S[0]|S[1]<<8|S[2]<<16|S[3]<<24,S[4]|S[5]<<8|S[6]<<16|S[7]<<24,M)},a.fromBytesBE=function(S,M){return new a(S[4]<<24|S[5]<<16|S[6]<<8|S[7],S[0]<<24|S[1]<<16|S[2]<<8|S[3],M)}}),kF=Bt(()=>{}),IF=Bt(()=>{}),SF=Bt((e,t)=>{(function(n,a,r){function s(u){var p=this,d=l();p.next=function(){var c=2091639*p.s0+p.c*23283064365386963e-26;return p.s0=p.s1,p.s1=p.s2,p.s2=c-(p.c=c|0)},p.c=1,p.s0=d(" "),p.s1=d(" "),p.s2=d(" "),p.s0-=d(u),p.s0<0&&(p.s0+=1),p.s1-=d(u),p.s1<0&&(p.s1+=1),p.s2-=d(u),p.s2<0&&(p.s2+=1),d=null}function i(u,p){return p.c=u.c,p.s0=u.s0,p.s1=u.s1,p.s2=u.s2,p}function o(u,p){var d=new s(u),c=p&&p.state,h=d.next;return h.int32=function(){return d.next()*4294967296|0},h.double=function(){return h()+(h()*2097152|0)*11102230246251565e-32},h.quick=h,c&&(typeof c=="object"&&i(c,d),h.state=function(){return i(d,{})}),h}function l(){var u=4022871197,p=function(d){d=String(d);for(var c=0;c<d.length;c++){u+=d.charCodeAt(c);var h=.02519603282416938*u;u=h>>>0,h-=u,h*=u,u=h>>>0,h-=u,u+=h*4294967296}return(u>>>0)*23283064365386963e-26};return p}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.alea=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),TF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.x=0,u.y=0,u.z=0,u.w=0,u.next=function(){var c=u.x^u.x<<11;return u.x=u.y,u.y=u.z,u.z=u.w,u.w^=u.w>>>19^c^c>>>8},l===(l|0)?u.x=l:p+=l;for(var d=0;d<p.length+64;d++)u.x^=p.charCodeAt(d)|0,u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xor128=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),NF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.next=function(){var c=u.x^u.x>>>2;return u.x=u.y,u.y=u.z,u.z=u.w,u.w=u.v,(u.d=u.d+362437|0)+(u.v=u.v^u.v<<4^(c^c<<1))|0},u.x=0,u.y=0,u.z=0,u.w=0,u.v=0,l===(l|0)?u.x=l:p+=l;for(var d=0;d<p.length+64;d++)u.x^=p.charCodeAt(d)|0,d==p.length&&(u.d=u.x<<10^u.x>>>4),u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u.v=l.v,u.d=l.d,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xorwow=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),CF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this;u.next=function(){var d=u.x,c=u.i,h,m,f;return h=d[c],h^=h>>>7,m=h^h<<24,h=d[c+1&7],m^=h^h>>>10,h=d[c+3&7],m^=h^h>>>3,h=d[c+4&7],m^=h^h<<7,h=d[c+7&7],h=h^h<<13,m^=h^h<<9,d[c]=m,u.i=c+1&7,m};function p(d,c){var h,m,f=[];if(c===(c|0))m=f[0]=c;else for(c=""+c,h=0;h<c.length;++h)f[h&7]=f[h&7]<<15^c.charCodeAt(h)+f[h+1&7]<<13;for(;f.length<8;)f.push(0);for(h=0;h<8&&f[h]===0;++h);for(h==8?m=f[7]=-1:m=f[h],d.x=f,d.i=0,h=256;h>0;--h)d.next()}p(u,l)}function i(l,u){return u.x=l.x.slice(),u.i=l.i,u}function o(l,u){l==null&&(l=+new Date);var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(d.x&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xorshift7=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),_F=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this;u.next=function(){var d=u.w,c=u.X,h=u.i,m,f;return u.w=d=d+1640531527|0,f=c[h+34&127],m=c[h=h+1&127],f^=f<<13,m^=m<<17,f^=f>>>15,m^=m>>>12,f=c[h]=f^m,u.i=h,f+(d^d>>>16)|0};function p(d,c){var h,m,f,g,y,b=[],x=128;for(c===(c|0)?(m=c,c=null):(c=c+"\0",m=0,x=Math.max(x,c.length)),f=0,g=-32;g<x;++g)c&&(m^=c.charCodeAt((g+32)%c.length)),g===0&&(y=m),m^=m<<10,m^=m>>>15,m^=m<<4,m^=m>>>13,g>=0&&(y=y+1640531527|0,h=b[g&127]^=m+y,f=h==0?f+1:0);for(f>=128&&(b[(c&&c.length||0)&127]=-1),f=127,g=4*128;g>0;--g)m=b[f+34&127],h=b[f=f+1&127],m^=m<<13,h^=h<<17,m^=m>>>15,h^=h>>>12,b[f]=m^h;d.w=y,d.X=b,d.i=f}p(u,l)}function i(l,u){return u.i=l.i,u.w=l.w,u.X=l.X.slice(),u}function o(l,u){l==null&&(l=+new Date);var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(d.X&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.xor4096=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),EF=Bt((e,t)=>{(function(n,a,r){function s(l){var u=this,p="";u.next=function(){var c=u.b,h=u.c,m=u.d,f=u.a;return c=c<<25^c>>>7^h,h=h-m|0,m=m<<24^m>>>8^f,f=f-c|0,u.b=c=c<<20^c>>>12^h,u.c=h=h-m|0,u.d=m<<16^h>>>16^f,u.a=f-c|0},u.a=0,u.b=0,u.c=-1640531527,u.d=1367130551,l===Math.floor(l)?(u.a=l/4294967296|0,u.b=l|0):p+=l;for(var d=0;d<p.length+20;d++)u.b^=p.charCodeAt(d)|0,u.next()}function i(l,u){return u.a=l.a,u.b=l.b,u.c=l.c,u.d=l.d,u}function o(l,u){var p=new s(l),d=u&&u.state,c=function(){return(p.next()>>>0)/4294967296};return c.double=function(){do var h=p.next()>>>11,m=(p.next()>>>0)/4294967296,f=(h+m)/(1<<21);while(f===0);return f},c.int32=p.next,c.quick=c,d&&(typeof d=="object"&&i(d,p),c.state=function(){return i(p,{})}),c}a&&a.exports?a.exports=o:r&&r.amd?r(function(){return o}):this.tychei=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}),AF=Bt(()=>{}),$F=Bt((e,t)=>{(function(n,a,r){var s=256,i=6,o=52,l="random",u=r.pow(s,i),p=r.pow(2,o),d=p*2,c=s-1,h;function m(I,T,C){var E=[];T=T==!0?{entropy:!0}:T||{};var A=b(y(T.entropy?[I,w(a)]:I==null?x():I,3),E),R=new f(E),F=function(){for(var S=R.g(i),M=u,B=0;S<p;)S=(S+B)*s,M*=s,B=R.g(1);for(;S>=d;)S/=2,M/=2,B>>>=1;return(S+B)/M};return F.int32=function(){return R.g(4)|0},F.quick=function(){return R.g(4)/4294967296},F.double=F,b(w(R.S),a),(T.pass||C||function(S,M,B,U){return U&&(U.S&&g(U,R),S.state=function(){return g(R,{})}),B?(r[l]=S,M):S})(F,A,"global"in T?T.global:this==r,T.state)}function f(I){var T,C=I.length,E=this,A=0,R=E.i=E.j=0,F=E.S=[];for(C||(I=[C++]);A<s;)F[A]=A++;for(A=0;A<s;A++)F[A]=F[R=c&R+I[A%C]+(T=F[A])],F[R]=T;(E.g=function(S){for(var M,B=0,U=E.i,G=E.j,q=E.S;S--;)M=q[U=c&U+1],B=B*s+q[c&(q[U]=q[G=c&G+M])+(q[G]=M)];return E.i=U,E.j=G,B})(s)}function g(I,T){return T.i=I.i,T.j=I.j,T.S=I.S.slice(),T}function y(I,T){var C=[],E=typeof I,A;if(T&&E=="object")for(A in I)try{C.push(y(I[A],T-1))}catch(R){}return C.length?C:E=="string"?I:I+"\0"}function b(I,T){for(var C=I+"",E,A=0;A<C.length;)T[c&A]=c&(E^=T[c&A]*19)+C.charCodeAt(A++);return w(T)}function x(){try{var I;return h&&(I=h.randomBytes)?I=I(s):(I=new Uint8Array(s),(n.crypto||n.msCrypto).getRandomValues(I)),w(I)}catch(E){var T=n.navigator,C=T&&T.plugins;return[+new Date,n,C,n.screen,w(a)]}}function w(I){return String.fromCharCode.apply(0,I)}if(b(r.random(),a),typeof t=="object"&&t.exports){t.exports=m;try{h=AF()}catch(I){}}else typeof define=="function"&&define.amd?define(function(){return m}):r["seed"+l]=m})(typeof self!="undefined"?self:e,[],Math)}),im=Bt((e,t)=>{var n=SF(),a=TF(),r=NF(),s=CF(),i=_F(),o=EF(),l=$F();l.alea=n,l.xor128=a,l.xorwow=r,l.xorshift7=s,l.xor4096=i,l.tychei=o,t.exports=l}),VI=Bt(()=>{}),Cx=Bt(()=>{}),Ch=Bt(()=>{}),FF=Bt(()=>{}),DF=Bt(()=>{}),RF=Bt(()=>{}),MF=Bt((e,t)=>{var n=(()=>{var a=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(a=a||__filename),function(r){r=r||{};function s(){return ue.buffer!=ze&&ot(ue.buffer),dt}function i(){return ue.buffer!=ze&&ot(ue.buffer),jn}function o(){return ue.buffer!=ze&&ot(ue.buffer),Mt}function l(){return ue.buffer!=ze&&ot(ue.buffer),on}function u(){return ue.buffer!=ze&&ot(ue.buffer),Fn}function p(){return ue.buffer!=ze&&ot(ue.buffer),oa}function d(){return ue.buffer!=ze&&ot(ue.buffer),Dn}var c=typeof r!="undefined"?r:{},h,m;c.ready=new Promise(function(D,j){h=D,m=j});var f;typeof process!="undefined"&&process.listeners&&(f={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var g=Object.assign({},c),y=[],b="./this.program",x=(D,j)=>{throw j},w=typeof window=="object",I=typeof importScripts=="function",T=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",C=c.ENVIRONMENT_IS_PTHREAD||!1,E="";function A(D){return c.locateFile?c.locateFile(D,E):E+D}var R,F,S,M;function B(D){D instanceof Ms||Q("exiting due to exception: "+D)}if(T){I?E=Ch().dirname(E)+"/":E=__dirname+"/";var U,G;typeof BI=="function"&&(U=Cx(),G=Ch()),R=(j,re)=>(j=G.normalize(j),U.readFileSync(j,re?void 0:"utf8")),S=j=>{var re=R(j,!0);return re.buffer||(re=new Uint8Array(re)),re},F=(j,re,ce)=>{j=G.normalize(j),U.readFile(j,function(ke,je){ke?ce(ke):re(je.buffer)})},process.argv.length>1&&(b=process.argv[1].replace(/\\/g,"/")),y=process.argv.slice(2),process.on("uncaughtException",function(j){if(!(j instanceof Ms))throw j}),process.on("unhandledRejection",function(j){throw j}),x=(j,re)=>{if(Ia())throw process.exitCode=j,re;B(re),process.exit(j)},c.inspect=function(){return"[Emscripten Module object]"};let D;try{D=FF()}catch(j){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),j}global.Worker=D.Worker}else(w||I)&&(I?E=self.location.href:typeof document!="undefined"&&document.currentScript&&(E=document.currentScript.src),typeof a!="undefined"&&a&&(E=a),E.indexOf("blob:")!==0?E=E.substr(0,E.replace(/[?#].*/,"").lastIndexOf("/")+1):E="",T||(R=D=>{var j=new XMLHttpRequest;return j.open("GET",D,!1),j.send(null),j.responseText},I&&(S=D=>{var j=new XMLHttpRequest;return j.open("GET",D,!1),j.responseType="arraybuffer",j.send(null),new Uint8Array(j.response)}),F=(D,j,re)=>{var ce=new XMLHttpRequest;ce.open("GET",D,!0),ce.responseType="arraybuffer",ce.onload=()=>{if(ce.status==200||ce.status==0&&ce.response){j(ce.response);return}re()},ce.onerror=re,ce.send(null)}),M=D=>document.title=D);T&&typeof performance=="undefined"&&(global.performance=DF().performance);var q=console.log.bind(console),K=console.warn.bind(console);T&&(q=D=>U.writeSync(1,D+`
-`),K=D=>U.writeSync(2,D+`
-`));var Z=c.print||q,Q=c.printErr||K;Object.assign(c,g),g=null,c.arguments&&(y=c.arguments),c.thisProgram&&(b=c.thisProgram),c.quit&&(x=c.quit);var ee=4,ae=Atomics.load,te=Atomics.store,le=Atomics.compareExchange,ie;c.wasmBinary&&(ie=c.wasmBinary);var ye=c.noExitRuntime||!0;typeof WebAssembly!="object"&&Yo("no native wasm support detected");var ue,xe,Ie=!1,Se;function Le(D,j){D||Yo(j)}var Ve=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function nt(D,j,re){for(var ce=j+re,ke=j;D[ke]&&!(ke>=ce);)++ke;if(ke-j>16&&D.buffer&&Ve)return Ve.decode(D.buffer instanceof SharedArrayBuffer?D.slice(j,ke):D.subarray(j,ke));for(var je="";j<ke;){var Ce=D[j++];if(!(Ce&128)){je+=String.fromCharCode(Ce);continue}var Re=D[j++]&63;if((Ce&224)==192){je+=String.fromCharCode((Ce&31)<<6|Re);continue}var Ot=D[j++]&63;if((Ce&240)==224?Ce=(Ce&15)<<12|Re<<6|Ot:Ce=(Ce&7)<<18|Re<<12|Ot<<6|D[j++]&63,Ce<65536)je+=String.fromCharCode(Ce);else{var ua=Ce-65536;je+=String.fromCharCode(55296|ua>>10,56320|ua&1023)}}return je}function it(D,j){return D?nt(i(),D,j):""}function et(D,j,re,ce){if(!(ce>0))return 0;for(var ke=re,je=re+ce-1,Ce=0;Ce<D.length;++Ce){var Re=D.charCodeAt(Ce);if(Re>=55296&&Re<=57343){var Ot=D.charCodeAt(++Ce);Re=65536+((Re&1023)<<10)|Ot&1023}if(Re<=127){if(re>=je)break;j[re++]=Re}else if(Re<=2047){if(re+1>=je)break;j[re++]=192|Re>>6,j[re++]=128|Re&63}else if(Re<=65535){if(re+2>=je)break;j[re++]=224|Re>>12,j[re++]=128|Re>>6&63,j[re++]=128|Re&63}else{if(re+3>=je)break;j[re++]=240|Re>>18,j[re++]=128|Re>>12&63,j[re++]=128|Re>>6&63,j[re++]=128|Re&63}}return j[re]=0,re-ke}function at(D,j,re){return et(D,i(),j,re)}var ze,dt,jn,Mt,ia,on,Fn,oa,Dn;C&&(ze=c.buffer);function ot(D){ze=D,c.HEAP8=dt=new Int8Array(D),c.HEAP16=Mt=new Int16Array(D),c.HEAP32=on=new Int32Array(D),c.HEAPU8=jn=new Uint8Array(D),c.HEAPU16=ia=new Uint16Array(D),c.HEAPU32=Fn=new Uint32Array(D),c.HEAPF32=oa=new Float32Array(D),c.HEAPF64=Dn=new Float64Array(D)}var Rn=c.INITIAL_MEMORY||16777216;if(C)ue=c.wasmMemory,ze=c.buffer;else if(c.wasmMemory)ue=c.wasmMemory;else if(ue=new WebAssembly.Memory({initial:Rn/65536,maximum:32768,shared:!0}),!(ue.buffer instanceof SharedArrayBuffer))throw Q("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),T&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");ue&&(ze=ue.buffer),Rn=ze.byteLength,ot(ze);var qn,xr=[],qo=[],er=[],vp=!1;function Ia(){return ye}function Ko(){if(c.preRun)for(typeof c.preRun=="function"&&(c.preRun=[c.preRun]);c.preRun.length;)Ng(c.preRun.shift());Sp(xr)}function Xt(){vp=!0,!C&&Sp(qo)}function Fd(){if(!C){if(c.postRun)for(typeof c.postRun=="function"&&(c.postRun=[c.postRun]);c.postRun.length;)G1(c.postRun.shift());Sp(er)}}function Ng(D){xr.unshift(D)}function Cg(D){qo.unshift(D)}function G1(D){er.unshift(D)}var jr=0,Xo=null,vr=null;function H1(D){jr++,c.monitorRunDependencies&&c.monitorRunDependencies(jr)}function j1(D){if(jr--,c.monitorRunDependencies&&c.monitorRunDependencies(jr),jr==0&&(Xo!==null&&(clearInterval(Xo),Xo=null),vr)){var j=vr;vr=null,j()}}function Yo(D){C?postMessage({cmd:"onAbort",arg:D}):c.onAbort&&c.onAbort(D),D="Aborted("+D+")",Q(D),Ie=!0,Se=1,D+=". Build with -sASSERTIONS for more info.";var j=new WebAssembly.RuntimeError(D);throw m(j),j}var _g="data:application/octet-stream;base64,";function Dd(D){return D.startsWith(_g)}function wp(D){return D.startsWith("file://")}var bn;bn="tfjs-backend-wasm-threaded-simd.wasm",Dd(bn)||(bn=A(bn));function Rd(D){try{if(D==bn&&ie)return new Uint8Array(ie);if(S)return S(D);throw"both async and sync fetching of the wasm failed"}catch(j){Yo(j)}}function Eg(){if(!ie&&(w||I)){if(typeof fetch=="function"&&!wp(bn))return fetch(bn,{credentials:"same-origin"}).then(function(D){if(!D.ok)throw"failed to load wasm binary file at '"+bn+"'";return D.arrayBuffer()}).catch(function(){return Rd(bn)});if(F)return new Promise(function(D,j){F(bn,function(re){D(new Uint8Array(re))},j)})}return Promise.resolve().then(function(){return Rd(bn)})}function Ag(){var D={env:jd,wasi_snapshot_preview1:jd};function j(Ce,Re){var Ot=Ce.exports;if(c.asm=Ot,Wg(c.asm._emscripten_tls_init),qn=c.asm.__indirect_function_table,Cg(c.asm.__wasm_call_ctors),xe=Re,!C){var ua=$e.unusedWorkers.length;$e.unusedWorkers.forEach(function(kr){$e.loadWasmModuleToWorker(kr,function(){--ua||j1("wasm-instantiate")})})}}C||H1("wasm-instantiate");function re(Ce){j(Ce.instance,Ce.module)}function ce(Ce){return Eg().then(function(Re){return WebAssembly.instantiate(Re,D)}).then(function(Re){return Re}).then(Ce,function(Re){Q("failed to asynchronously prepare wasm: "+Re),Yo(Re)})}function ke(){return!ie&&typeof WebAssembly.instantiateStreaming=="function"&&!Dd(bn)&&!wp(bn)&&!T&&typeof fetch=="function"?fetch(bn,{credentials:"same-origin"}).then(function(Ce){var Re=WebAssembly.instantiateStreaming(Ce,D);return Re.then(re,function(Ot){return Q("wasm streaming compile failed: "+Ot),Q("falling back to ArrayBuffer instantiation"),ce(re)})}):ce(re)}if(c.instantiateWasm)try{var je=c.instantiateWasm(D,j);return je}catch(Ce){Q("Module.instantiateWasm callback failed with error: "+Ce),m(Ce)}return ke().catch(m),{}}var $g,q1,Fg={};function Ms(D){this.name="ExitStatus",this.message="Program terminated with exit("+D+")",this.status=D}function Dg(D){var j=$e.pthreads[D];delete $e.pthreads[D],j.terminate(),lb(D),$e.runningWorkers.splice($e.runningWorkers.indexOf(j),1),j.pthread_ptr=0}function Rg(D){var j=$e.pthreads[D];j.postMessage({cmd:"cancel"})}function kp(D){var j=$e.pthreads[D];Le(j),$e.returnWorkerToPool(j)}function Md(D){var j=$e.getNewWorker();if(!j)return 6;$e.runningWorkers.push(j),$e.pthreads[D.pthread_ptr]=j,j.pthread_ptr=D.pthread_ptr;var re={cmd:"run",start_routine:D.startRoutine,arg:D.arg,pthread_ptr:D.pthread_ptr};return j.runPthread=()=>{re.time=performance.now(),j.postMessage(re,D.transferList)},j.loaded&&(j.runPthread(),delete j.runPthread),0}var Pd={varargs:void 0,get:function(){Pd.varargs+=4;var D=l()[Pd.varargs-4>>2];return D},getStr:function(D){var j=it(D);return j}};function Ip(D){if(C)return qr(1,1,D);Se=D,Ia()||($e.terminateAllThreads(),c.onExit&&c.onExit(D),Ie=!0),x(D,new Ms(D))}function K1(D,j){if(Se=D,!j&&C)throw Ld(D),"unwind";Ip(D)}var Od=K1;function Mg(D){if(D instanceof Ms||D=="unwind")return Se;x(1,D)}var $e={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){C?$e.initWorker():$e.initMainThread()},initMainThread:function(){for(var D=8;D--;)$e.allocateUnusedWorker()},initWorker:function(){ye=!1},setExitStatus:function(D){Se=D},terminateAllThreads:function(){for(var D of Object.values($e.pthreads))$e.returnWorkerToPool(D);for(var D of $e.unusedWorkers)D.terminate();$e.unusedWorkers=[]},returnWorkerToPool:function(D){var j=D.pthread_ptr;delete $e.pthreads[j],$e.unusedWorkers.push(D),$e.runningWorkers.splice($e.runningWorkers.indexOf(D),1),D.pthread_ptr=0,lb(j)},receiveObjectTransfer:function(D){},threadInitTLS:function(){$e.tlsInitFunctions.forEach(D=>D())},loadWasmModuleToWorker:function(D,j){D.onmessage=re=>{var ce=re.data,ke=ce.cmd;if(D.pthread_ptr&&($e.currentProxiedOperationCallerThread=D.pthread_ptr),ce.targetThread&&ce.targetThread!=Jd()){var je=$e.pthreads[ce.targetThread];je?je.postMessage(ce,ce.transferList):Q('Internal error! Worker sent a message "'+ke+'" to target pthread '+ce.targetThread+", but that thread no longer exists!"),$e.currentProxiedOperationCallerThread=void 0;return}ke==="processProxyingQueue"?Tp(ce.queue):ke==="spawnThread"?Md(ce):ke==="cleanupThread"?kp(ce.thread):ke==="killThread"?Dg(ce.thread):ke==="cancelThread"?Rg(ce.thread):ke==="loaded"?(D.loaded=!0,j&&j(D),D.runPthread&&(D.runPthread(),delete D.runPthread)):ke==="print"?Z("Thread "+ce.threadId+": "+ce.text):ke==="printErr"?Q("Thread "+ce.threadId+": "+ce.text):ke==="alert"?alert("Thread "+ce.threadId+": "+ce.text):ce.target==="setimmediate"?D.postMessage(ce):ke==="onAbort"?c.onAbort&&c.onAbort(ce.arg):ke&&Q("worker sent an unknown command "+ke),$e.currentProxiedOperationCallerThread=void 0},D.onerror=re=>{var ce="worker sent an error!";throw Q(ce+" "+re.filename+":"+re.lineno+": "+re.message),re},T&&(D.on("message",function(re){D.onmessage({data:re})}),D.on("error",function(re){D.onerror(re)}),D.on("detachedExit",function(){})),D.postMessage({cmd:"load",urlOrBlob:c.mainScriptUrlOrBlob||a,wasmMemory:ue,wasmModule:xe})},allocateUnusedWorker:function(){var D=A("tfjs-backend-wasm-threaded-simd.worker.js");$e.unusedWorkers.push(new Worker(D))},getNewWorker:function(){return $e.unusedWorkers.length==0&&($e.allocateUnusedWorker(),$e.loadWasmModuleToWorker($e.unusedWorkers[0])),$e.unusedWorkers.pop()}};c.PThread=$e;function Sp(D){for(;D.length>0;)D.shift()(c)}function Pg(D){var j=ub(),re=D();return Qd(j),re}function X1(D){return D}function Y1(D){var j=/\b_Z[\w\d_]+/g;return D.replace(j,function(re){var ce=re;return re===ce?re:ce+" ["+re+"]"})}function Og(){var D=Jd(),j=l()[D+44>>2],re=l()[D+48>>2],ce=j-re;ak(j,ce),Qd(j)}c.establishStackSpace=Og;function Ld(D){if(C)return qr(2,0,D);try{Od(D)}catch(j){Mg(j)}}var Zo=[];function Lg(D){var j=Zo[D];return j||(D>=Zo.length&&(Zo.length=D+1),Zo[D]=j=qn.get(D)),j}function zg(D,j){var re=Lg(D)(j);Ia()?$e.setExitStatus(re):nk(re)}c.invokeEntryPoint=zg;function Z1(){var D=new Error;if(!D.stack){try{throw new Error}catch(j){D=j}if(!D.stack)return"(no stack trace available)"}return D.stack.toString()}function Wg(D){$e.tlsInitFunctions.push(D)}function Bg(D,j){s().set(D,j)}function Vg(D){Q1(D,!I,1,!w),$e.threadInitTLS()}function Ug(D){C?postMessage({cmd:"cleanupThread",thread:D}):kp(D)}function zd(D,j,re,ce){return C?qr(3,1,D,j,re,ce):Wd(D,j,re,ce)}function Wd(D,j,re,ce){if(typeof SharedArrayBuffer=="undefined")return Q("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var ke=[],je=0;if(C&&(ke.length===0||je))return zd(D,j,re,ce);if(je)return je;var Ce={startRoutine:re,pthread_ptr:D,arg:ce,transferList:ke};return C?(Ce.cmd="spawnThread",postMessage(Ce,ke),0):Md(Ce)}function Gg(){return 2097152}var Hg=!0;function jg(){return Hg}function Tp(D){Atomics.store(l(),D>>2,1),Jd()&&tk(D),Atomics.compareExchange(l(),D>>2,1,0)}c.executeNotifiedProxyingQueue=Tp;function qg(D,j,re,ce){if(D==j)setTimeout(()=>Tp(ce));else if(C)postMessage({targetThread:D,cmd:"processProxyingQueue",queue:ce});else{var ke=$e.pthreads[D];if(!ke)return;ke.postMessage({cmd:"processProxyingQueue",queue:ce})}return 1}function Kg(D,j,re){return-1}function Xg(){Yo("")}function Ps(D){Ps.shown||(Ps.shown={}),Ps.shown[D]||(Ps.shown[D]=1,T&&(D="warning: "+D),Q(D))}function Yg(){T||I||Ps("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function Zg(){return Date.now()}function Bd(){return 2147483648}function Jg(){return Bd()}var Jo;T?Jo=()=>{var D=process.hrtime();return D[0]*1e3+D[1]/1e6}:C?Jo=()=>performance.now()-c.__performance_now_clock_drift:Jo=()=>performance.now();function Qg(D,j,re){i().copyWithin(D,j,j+re)}function ey(){return T?RF().cpus().length:navigator.hardwareConcurrency}function qr(D,j){var re=arguments.length-2,ce=arguments;return Pg(()=>{for(var ke=re,je=eh(ke*8),Ce=je>>3,Re=0;Re<re;Re++){var Ot=ce[2+Re];d()[Ce+Re]=Ot}return ek(D,ke,je,j)})}var Np=[];function ty(D,j,re){Np.length=j;for(var ce=re>>3,ke=0;ke<j;ke++)Np[ke]=d()[ce+ke];var je=D<0,Ce=je?Fg[-D-1]:uy[D];return Ce.apply(null,Np)}function ny(D){try{return ue.grow(D-ze.byteLength+65535>>>16),ot(ue.buffer),1}catch(j){}}function ay(D){var j=i().length;if(D=D>>>0,D<=j)return!1;var re=Bd();if(D>re)return!1;let ce=(Ot,ua)=>Ot+(ua-Ot%ua)%ua;for(var ke=1;ke<=4;ke*=2){var je=j*(1+.2/ke);je=Math.min(je,D+100663296);var Ce=Math.min(re,ce(Math.max(D,je),65536)),Re=ny(Ce);if(Re)return!0}return!1}function ry(){throw"unwind"}function Vd(D){return C?qr(4,1,D):52}function Ud(D,j,re,ce,ke){return C?qr(5,1,D,j,re,ce,ke):70}var sy=[null,[],[]];function iy(D,j){var re=sy[D];j===0||j===10?((D===1?Z:Q)(nt(re,0)),re.length=0):re.push(j)}function Gd(D,j,re,ce){if(C)return qr(6,1,D,j,re,ce);for(var ke=0,je=0;je<re;je++){var Ce=u()[j>>2],Re=u()[j+4>>2];j+=8;for(var Ot=0;Ot<Re;Ot++)iy(D,i()[Ce+Ot]);ke+=Re}return u()[ce>>2]=ke,0}function Hd(D){var j=c["_"+D];return j}function oy(D,j,re,ce,ke){var je={string:pa=>{var nl=0;if(pa!=null&&pa!==0){var ik=(pa.length<<2)+1;nl=eh(ik),at(pa,nl,ik)}return nl},array:pa=>{var nl=eh(pa.length);return Bg(pa,nl),nl}};function Ce(pa){return j==="string"?it(pa):j==="boolean"?Boolean(pa):pa}var Re=Hd(D),Ot=[],ua=0;if(ce)for(var kr=0;kr<ce.length;kr++){var sk=je[re[kr]];sk?(ua===0&&(ua=ub()),Ot[kr]=sk(ce[kr])):Ot[kr]=ce[kr]}var pb=Re.apply(null,Ot);function uF(pa){return ua!==0&&Qd(ua),Ce(pa)}return pb=uF(pb),pb}function ly(D,j,re,ce){re=re||[];var ke=re.every(Ce=>Ce==="number"||Ce==="boolean"),je=j!=="string";return je&&ke&&!ce?Hd(D):function(){return oy(D,j,re,arguments,ce)}}$e.init();var uy=[null,Ip,Ld,zd,Vd,Ud,Gd],jd={__emscripten_init_main_thread_js:Vg,__emscripten_thread_cleanup:Ug,__pthread_create_js:Wd,_emscripten_default_pthread_stack_size:Gg,_emscripten_get_now_is_monotonic:jg,_emscripten_notify_task_queue:qg,_emscripten_set_offscreencanvas_size:Kg,abort:Xg,emscripten_check_blocking_allowed:Yg,emscripten_date_now:Zg,emscripten_get_heap_max:Jg,emscripten_get_now:Jo,emscripten_memcpy_big:Qg,emscripten_num_logical_cores:ey,emscripten_receive_on_main_thread_js:ty,emscripten_resize_heap:ay,emscripten_unwind_to_js_event_loop:ry,exit:Od,fd_close:Vd,fd_seek:Ud,fd_write:Gd,memory:ue||c.wasmMemory},J1=Ag(),py=c.___wasm_call_ctors=function(){return(py=c.___wasm_call_ctors=c.asm.__wasm_call_ctors).apply(null,arguments)},cy=c._init=function(){return(cy=c._init=c.asm.init).apply(null,arguments)},dy=c._init_with_threads_count=function(){return(dy=c._init_with_threads_count=c.asm.init_with_threads_count).apply(null,arguments)},hy=c._get_threads_count=function(){return(hy=c._get_threads_count=c.asm.get_threads_count).apply(null,arguments)},my=c._register_tensor=function(){return(my=c._register_tensor=c.asm.register_tensor).apply(null,arguments)},fy=c._dispose_data=function(){return(fy=c._dispose_data=c.asm.dispose_data).apply(null,arguments)},gy=c._dispose=function(){return(gy=c._dispose=c.asm.dispose).apply(null,arguments)},yy=c._Abs=function(){return(yy=c._Abs=c.asm.Abs).apply(null,arguments)},by=c._Add=function(){return(by=c._Add=c.asm.Add).apply(null,arguments)},xy=c._AddN=function(){return(xy=c._AddN=c.asm.AddN).apply(null,arguments)},vy=c._All=function(){return(vy=c._All=c.asm.All).apply(null,arguments)},wy=c._Any=function(){return(wy=c._Any=c.asm.Any).apply(null,arguments)},ky=c._ArgMax=function(){return(ky=c._ArgMax=c.asm.ArgMax).apply(null,arguments)},Iy=c._AvgPool=function(){return(Iy=c._AvgPool=c.asm.AvgPool).apply(null,arguments)},Sy=c._BatchMatMul=function(){return(Sy=c._BatchMatMul=c.asm.BatchMatMul).apply(null,arguments)},Ty=c._Ceil=function(){return(Ty=c._Ceil=c.asm.Ceil).apply(null,arguments)},Ny=c._ClipByValue=function(){return(Ny=c._ClipByValue=c.asm.ClipByValue).apply(null,arguments)},Cy=c._Conv2D=function(){return(Cy=c._Conv2D=c.asm.Conv2D).apply(null,arguments)},_y=c._Conv2DBackpropInput=function(){return(_y=c._Conv2DBackpropInput=c.asm.Conv2DBackpropInput).apply(null,arguments)},Ey=c._Cos=function(){return(Ey=c._Cos=c.asm.Cos).apply(null,arguments)},Ay=c._Cosh=function(){return(Ay=c._Cosh=c.asm.Cosh).apply(null,arguments)},$y=c._CropAndResize=function(){return($y=c._CropAndResize=c.asm.CropAndResize).apply(null,arguments)},Fy=c._Cumprod=function(){return(Fy=c._Cumprod=c.asm.Cumprod).apply(null,arguments)},Dy=c._Cumsum=function(){return(Dy=c._Cumsum=c.asm.Cumsum).apply(null,arguments)},Ry=c._DepthToSpace=function(){return(Ry=c._DepthToSpace=c.asm.DepthToSpace).apply(null,arguments)},My=c._DepthwiseConv2dNative=function(){return(My=c._DepthwiseConv2dNative=c.asm.DepthwiseConv2dNative).apply(null,arguments)},Py=c._Elu=function(){return(Py=c._Elu=c.asm.Elu).apply(null,arguments)},Oy=c._Equal=function(){return(Oy=c._Equal=c.asm.Equal).apply(null,arguments)},Ly=c._Exp=function(){return(Ly=c._Exp=c.asm.Exp).apply(null,arguments)},zy=c._FlipLeftRight=function(){return(zy=c._FlipLeftRight=c.asm.FlipLeftRight).apply(null,arguments)},Wy=c._Floor=function(){return(Wy=c._Floor=c.asm.Floor).apply(null,arguments)},By=c._FloorDiv=function(){return(By=c._FloorDiv=c.asm.FloorDiv).apply(null,arguments)},Vy=c._FusedBatchNorm=function(){return(Vy=c._FusedBatchNorm=c.asm.FusedBatchNorm).apply(null,arguments)},Uy=c._FusedConv2D=function(){return(Uy=c._FusedConv2D=c.asm.FusedConv2D).apply(null,arguments)},Gy=c._FusedDepthwiseConv2D=function(){return(Gy=c._FusedDepthwiseConv2D=c.asm.FusedDepthwiseConv2D).apply(null,arguments)},Hy=c._Gather=function(){return(Hy=c._Gather=c.asm.Gather).apply(null,arguments)},jy=c._GatherNd=function(){return(jy=c._GatherNd=c.asm.GatherNd).apply(null,arguments)},qy=c._Greater=function(){return(qy=c._Greater=c.asm.Greater).apply(null,arguments)},Ky=c._GreaterEqual=function(){return(Ky=c._GreaterEqual=c.asm.GreaterEqual).apply(null,arguments)},Xy=c._LeakyRelu=function(){return(Xy=c._LeakyRelu=c.asm.LeakyRelu).apply(null,arguments)},Yy=c._Less=function(){return(Yy=c._Less=c.asm.Less).apply(null,arguments)},Zy=c._LessEqual=function(){return(Zy=c._LessEqual=c.asm.LessEqual).apply(null,arguments)},Jy=c._Log=function(){return(Jy=c._Log=c.asm.Log).apply(null,arguments)},Qy=c._LogicalAnd=function(){return(Qy=c._LogicalAnd=c.asm.LogicalAnd).apply(null,arguments)},eb=c._LogicalNot=function(){return(eb=c._LogicalNot=c.asm.LogicalNot).apply(null,arguments)},tb=c._LogicalOr=function(){return(tb=c._LogicalOr=c.asm.LogicalOr).apply(null,arguments)},nb=c._LogicalXor=function(){return(nb=c._LogicalXor=c.asm.LogicalXor).apply(null,arguments)},ab=c._Max=function(){return(ab=c._Max=c.asm.Max).apply(null,arguments)},qd=c._MaxPool=function(){return(qd=c._MaxPool=c.asm.MaxPool).apply(null,arguments)},Kd=c._Maximum=function(){return(Kd=c._Maximum=c.asm.Maximum).apply(null,arguments)},Cp=c._Mean=function(){return(Cp=c._Mean=c.asm.Mean).apply(null,arguments)},rb=c._Min=function(){return(rb=c._Min=c.asm.Min).apply(null,arguments)},sb=c._Minimum=function(){return(sb=c._Minimum=c.asm.Minimum).apply(null,arguments)},Qo=c._MirrorPad=function(){return(Qo=c._MirrorPad=c.asm.MirrorPad).apply(null,arguments)},Xd=c._Multiply=function(){return(Xd=c._Multiply=c.asm.Multiply).apply(null,arguments)},el=c._Neg=function(){return(el=c._Neg=c.asm.Neg).apply(null,arguments)},tl=c._NonMaxSuppressionV3=function(){return(tl=c._NonMaxSuppressionV3=c.asm.NonMaxSuppressionV3).apply(null,arguments)},ib=c._NonMaxSuppressionV4=function(){return(ib=c._NonMaxSuppressionV4=c.asm.NonMaxSuppressionV4).apply(null,arguments)},X=c._NonMaxSuppressionV5=function(){return(X=c._NonMaxSuppressionV5=c.asm.NonMaxSuppressionV5).apply(null,arguments)},se=c._NotEqual=function(){return(se=c._NotEqual=c.asm.NotEqual).apply(null,arguments)},we=c._OneHot=function(){return(we=c._OneHot=c.asm.OneHot).apply(null,arguments)},He=c._PadV2=function(){return(He=c._PadV2=c.asm.PadV2).apply(null,arguments)},wt=c._Pow=function(){return(wt=c._Pow=c.asm.Pow).apply(null,arguments)},kt=c._Prelu=function(){return(kt=c._Prelu=c.asm.Prelu).apply(null,arguments)},Ue=c._Prod=function(){return(Ue=c._Prod=c.asm.Prod).apply(null,arguments)},We=c._RealDiv=function(){return(We=c._RealDiv=c.asm.RealDiv).apply(null,arguments)},Pt=c._Relu=function(){return(Pt=c._Relu=c.asm.Relu).apply(null,arguments)},la=c._Relu6=function(){return(la=c._Relu6=c.asm.Relu6).apply(null,arguments)},wr=c._ResizeBilinear=function(){return(wr=c._ResizeBilinear=c.asm.ResizeBilinear).apply(null,arguments)},Yd=c._ResizeNearestNeighbor=function(){return(Yd=c._ResizeNearestNeighbor=c.asm.ResizeNearestNeighbor).apply(null,arguments)},_p=c._Reverse=function(){return(_p=c._Reverse=c.asm.Reverse).apply(null,arguments)},ob=c._RotateWithOffset=function(){return(ob=c._RotateWithOffset=c.asm.RotateWithOffset).apply(null,arguments)},Mn=c._Round=function(){return(Mn=c._Round=c.asm.Round).apply(null,arguments)},Kr=c._Rsqrt=function(){return(Kr=c._Rsqrt=c.asm.Rsqrt).apply(null,arguments)},Zd=c._ScatterNd=function(){return(Zd=c._ScatterNd=c.asm.ScatterNd).apply(null,arguments)},$$=c._SelectV2=function(){return($$=c._SelectV2=c.asm.SelectV2).apply(null,arguments)},F$=c._Sigmoid=function(){return(F$=c._Sigmoid=c.asm.Sigmoid).apply(null,arguments)},D$=c._Sin=function(){return(D$=c._Sin=c.asm.Sin).apply(null,arguments)},R$=c._Softmax=function(){return(R$=c._Softmax=c.asm.Softmax).apply(null,arguments)},M$=c._SparseFillEmptyRows=function(){return(M$=c._SparseFillEmptyRows=c.asm.SparseFillEmptyRows).apply(null,arguments)},P$=c._SparseReshape=function(){return(P$=c._SparseReshape=c.asm.SparseReshape).apply(null,arguments)},O$=c._SparseSegmentReduction=function(){return(O$=c._SparseSegmentReduction=c.asm.SparseSegmentReduction).apply(null,arguments)},L$=c._Sqrt=function(){return(L$=c._Sqrt=c.asm.Sqrt).apply(null,arguments)},z$=c._Square=function(){return(z$=c._Square=c.asm.Square).apply(null,arguments)},W$=c._SquaredDifference=function(){return(W$=c._SquaredDifference=c.asm.SquaredDifference).apply(null,arguments)},B$=c._Step=function(){return(B$=c._Step=c.asm.Step).apply(null,arguments)},V$=c._StridedSlice=function(){return(V$=c._StridedSlice=c.asm.StridedSlice).apply(null,arguments)},U$=c._Sub=function(){return(U$=c._Sub=c.asm.Sub).apply(null,arguments)},G$=c._Sum=function(){return(G$=c._Sum=c.asm.Sum).apply(null,arguments)},H$=c._Tan=function(){return(H$=c._Tan=c.asm.Tan).apply(null,arguments)},j$=c._Tanh=function(){return(j$=c._Tanh=c.asm.Tanh).apply(null,arguments)},q$=c._Tile=function(){return(q$=c._Tile=c.asm.Tile).apply(null,arguments)},K$=c._TopK=function(){return(K$=c._TopK=c.asm.TopK).apply(null,arguments)},X$=c._Transform=function(){return(X$=c._Transform=c.asm.Transform).apply(null,arguments)},Y$=c._Transpose=function(){return(Y$=c._Transpose=c.asm.Transpose).apply(null,arguments)},Z$=c.__FusedMatMul=function(){return(Z$=c.__FusedMatMul=c.asm._FusedMatMul).apply(null,arguments)},J$=c._malloc=function(){return(J$=c._malloc=c.asm.malloc).apply(null,arguments)},Q$=c._free=function(){return(Q$=c._free=c.asm.free).apply(null,arguments)},eF=c.__emscripten_tls_init=function(){return(eF=c.__emscripten_tls_init=c.asm._emscripten_tls_init).apply(null,arguments)},Jd=c._pthread_self=function(){return(Jd=c._pthread_self=c.asm.pthread_self).apply(null,arguments)},tF=c.___errno_location=function(){return(tF=c.___errno_location=c.asm.__errno_location).apply(null,arguments)},Q1=c.__emscripten_thread_init=function(){return(Q1=c.__emscripten_thread_init=c.asm._emscripten_thread_init).apply(null,arguments)},nF=c.__emscripten_thread_crashed=function(){return(nF=c.__emscripten_thread_crashed=c.asm._emscripten_thread_crashed).apply(null,arguments)},aF=c._emscripten_main_thread_process_queued_calls=function(){return(aF=c._emscripten_main_thread_process_queued_calls=c.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},rF=c._emscripten_main_browser_thread_id=function(){return(rF=c._emscripten_main_browser_thread_id=c.asm.emscripten_main_browser_thread_id).apply(null,arguments)},ek=c._emscripten_run_in_main_runtime_thread_js=function(){return(ek=c._emscripten_run_in_main_runtime_thread_js=c.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},sF=c._emscripten_dispatch_to_thread_=function(){return(sF=c._emscripten_dispatch_to_thread_=c.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},tk=c.__emscripten_proxy_execute_task_queue=function(){return(tk=c.__emscripten_proxy_execute_task_queue=c.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)},lb=c.__emscripten_thread_free_data=function(){return(lb=c.__emscripten_thread_free_data=c.asm._emscripten_thread_free_data).apply(null,arguments)},nk=c.__emscripten_thread_exit=function(){return(nk=c.__emscripten_thread_exit=c.asm._emscripten_thread_exit).apply(null,arguments)},ak=c._emscripten_stack_set_limits=function(){return(ak=c._emscripten_stack_set_limits=c.asm.emscripten_stack_set_limits).apply(null,arguments)},ub=c.stackSave=function(){return(ub=c.stackSave=c.asm.stackSave).apply(null,arguments)},Qd=c.stackRestore=function(){return(Qd=c.stackRestore=c.asm.stackRestore).apply(null,arguments)},eh=c.stackAlloc=function(){return(eh=c.stackAlloc=c.asm.stackAlloc).apply(null,arguments)},iF=c.dynCall_iijjiiii=function(){return(iF=c.dynCall_iijjiiii=c.asm.dynCall_iijjiiii).apply(null,arguments)},oF=c.dynCall_jiji=function(){return(oF=c.dynCall_jiji=c.asm.dynCall_jiji).apply(null,arguments)};c.keepRuntimeAlive=Ia,c.wasmMemory=ue,c.cwrap=ly,c.ExitStatus=Ms,c.PThread=$e;var th;vr=function D(){th||rk(),th||(vr=D)};function rk(D){if(D=D||y,jr>0)return;if(C){h(c),Xt(),postMessage({cmd:"loaded"});return}if(Ko(),jr>0)return;function j(){th||(th=!0,c.calledRun=!0,!Ie&&(Xt(),h(c),c.onRuntimeInitialized&&c.onRuntimeInitialized(),Fd()))}c.setStatus?(c.setStatus("Running..."),setTimeout(function(){setTimeout(function(){c.setStatus("")},1),j()},1)):j()}if(c.preInit)for(typeof c.preInit=="function"&&(c.preInit=[c.preInit]);c.preInit.length>0;)c.preInit.pop()();rk();var nh;f&&(nh={uncaughtException:process.listeners("uncaughtException").filter(function(D){return!f.uncaughtException.indexOf(D)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(D){return!f.unhandledRejection.indexOf(D)>-1})});var ah;if(typeof WasmBackendModule!="undefined")ah=WasmBackendModule;else if(typeof r!="undefined")ah=r;else throw new Error("Could not find wasm module in post.js");if(nh){var lF=ah._dispose;ah._dispose=function(){lF(),nh.uncaughtException.forEach(function(D){process.removeListener("uncaughtException",D)}),nh.unhandledRejection.forEach(function(D){process.removeListener("unhandledRejection",D)})}}return r.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModuleThreadedSimd=n)}),PF=Bt((e,t)=>{t.exports.wasmWorkerContents=`"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
-");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`}),OF=Bt((e,t)=>{var n=(()=>{var a=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(a=a||__filename),function(r){r=r||{};var s=typeof r!="undefined"?r:{},i,o;s.ready=new Promise(function(X,se){i=X,o=se});var l;typeof process!="undefined"&&process.listeners&&(l={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var u=Object.assign({},s),p=[],d="./this.program",c=(X,se)=>{throw se},h=typeof window=="object",m=typeof importScripts=="function",f=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",g="";function y(X){return s.locateFile?s.locateFile(X,g):g+X}var b,x,w,I;function T(X){X instanceof Xo||R("exiting due to exception: "+X)}if(f){m?g=Ch().dirname(g)+"/":g=__dirname+"/";var C,E;typeof BI=="function"&&(C=Cx(),E=Ch()),b=(X,se)=>(X=E.normalize(X),C.readFileSync(X,se?void 0:"utf8")),w=X=>{var se=b(X,!0);return se.buffer||(se=new Uint8Array(se)),se},x=(X,se,we)=>{X=E.normalize(X),C.readFile(X,function(He,wt){He?we(He):se(wt.buffer)})},process.argv.length>1&&(d=process.argv[1].replace(/\\/g,"/")),p=process.argv.slice(2),process.on("uncaughtException",function(X){if(!(X instanceof Xo))throw X}),process.on("unhandledRejection",function(X){throw X}),c=(X,se)=>{if(jn())throw process.exitCode=X,se;T(se),process.exit(X)},s.inspect=function(){return"[Emscripten Module object]"}}else(h||m)&&(m?g=self.location.href:typeof document!="undefined"&&document.currentScript&&(g=document.currentScript.src),a&&(g=a),g.indexOf("blob:")!==0?g=g.substr(0,g.replace(/[?#].*/,"").lastIndexOf("/")+1):g="",b=X=>{var se=new XMLHttpRequest;return se.open("GET",X,!1),se.send(null),se.responseText},m&&(w=X=>{var se=new XMLHttpRequest;return se.open("GET",X,!1),se.responseType="arraybuffer",se.send(null),new Uint8Array(se.response)}),x=(X,se,we)=>{var He=new XMLHttpRequest;He.open("GET",X,!0),He.responseType="arraybuffer",He.onload=()=>{if(He.status==200||He.status==0&&He.response){se(He.response);return}we()},He.onerror=we,He.send(null)},I=X=>document.title=X);var A=s.print||console.log.bind(console),R=s.printErr||console.warn.bind(console);Object.assign(s,u),u=null,s.arguments&&(p=s.arguments),s.thisProgram&&(d=s.thisProgram),s.quit&&(c=s.quit);var F=4,S;s.wasmBinary&&(S=s.wasmBinary);var M=s.noExitRuntime||!0;typeof WebAssembly!="object"&&er("no native wasm support detected");var B,U=!1,G;function q(X,se){X||er(se)}var K=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function Z(X,se,we){for(var He=se+we,wt=se;X[wt]&&!(wt>=He);)++wt;if(wt-se>16&&X.buffer&&K)return K.decode(X.subarray(se,wt));for(var kt="";se<wt;){var Ue=X[se++];if(!(Ue&128)){kt+=String.fromCharCode(Ue);continue}var We=X[se++]&63;if((Ue&224)==192){kt+=String.fromCharCode((Ue&31)<<6|We);continue}var Pt=X[se++]&63;if((Ue&240)==224?Ue=(Ue&15)<<12|We<<6|Pt:Ue=(Ue&7)<<18|We<<12|Pt<<6|X[se++]&63,Ue<65536)kt+=String.fromCharCode(Ue);else{var la=Ue-65536;kt+=String.fromCharCode(55296|la>>10,56320|la&1023)}}return kt}function Q(X,se){return X?Z(ie,X,se):""}function ee(X,se,we,He){if(!(He>0))return 0;for(var wt=we,kt=we+He-1,Ue=0;Ue<X.length;++Ue){var We=X.charCodeAt(Ue);if(We>=55296&&We<=57343){var Pt=X.charCodeAt(++Ue);We=65536+((We&1023)<<10)|Pt&1023}if(We<=127){if(we>=kt)break;se[we++]=We}else if(We<=2047){if(we+1>=kt)break;se[we++]=192|We>>6,se[we++]=128|We&63}else if(We<=65535){if(we+2>=kt)break;se[we++]=224|We>>12,se[we++]=128|We>>6&63,se[we++]=128|We&63}else{if(we+3>=kt)break;se[we++]=240|We>>18,se[we++]=128|We>>12&63,se[we++]=128|We>>6&63,se[we++]=128|We&63}}return se[we]=0,we-wt}function ae(X,se,we){return ee(X,ie,se,we)}var te,le,ie,ye,ue,xe,Ie,Se,Le;function Ve(X){te=X,s.HEAP8=le=new Int8Array(X),s.HEAP16=ye=new Int16Array(X),s.HEAP32=xe=new Int32Array(X),s.HEAPU8=ie=new Uint8Array(X),s.HEAPU16=ue=new Uint16Array(X),s.HEAPU32=Ie=new Uint32Array(X),s.HEAPF32=Se=new Float32Array(X),s.HEAPF64=Le=new Float64Array(X)}var nt=s.INITIAL_MEMORY||16777216,it,et=[],at=[],ze=[],dt=!1;function jn(){return M}function Mt(){if(s.preRun)for(typeof s.preRun=="function"&&(s.preRun=[s.preRun]);s.preRun.length;)Fn(s.preRun.shift());vr(et)}function ia(){dt=!0,vr(at)}function on(){if(s.postRun)for(typeof s.postRun=="function"&&(s.postRun=[s.postRun]);s.postRun.length;)Dn(s.postRun.shift());vr(ze)}function Fn(X){et.unshift(X)}function oa(X){at.unshift(X)}function Dn(X){ze.unshift(X)}var ot=0,Rn=null,qn=null;function xr(X){ot++,s.monitorRunDependencies&&s.monitorRunDependencies(ot)}function qo(X){if(ot--,s.monitorRunDependencies&&s.monitorRunDependencies(ot),ot==0&&(Rn!==null&&(clearInterval(Rn),Rn=null),qn)){var se=qn;qn=null,se()}}function er(X){s.onAbort&&s.onAbort(X),X="Aborted("+X+")",R(X),U=!0,G=1,X+=". Build with -sASSERTIONS for more info.";var se=new WebAssembly.RuntimeError(X);throw o(se),se}var vp="data:application/octet-stream;base64,";function Ia(X){return X.startsWith(vp)}function Ko(X){return X.startsWith("file://")}var Xt;Xt="tfjs-backend-wasm.wasm",Ia(Xt)||(Xt=y(Xt));function Fd(X){try{if(X==Xt&&S)return new Uint8Array(S);if(w)return w(X);throw"both async and sync fetching of the wasm failed"}catch(se){er(se)}}function Ng(){if(!S&&(h||m)){if(typeof fetch=="function"&&!Ko(Xt))return fetch(Xt,{credentials:"same-origin"}).then(function(X){if(!X.ok)throw"failed to load wasm binary file at '"+Xt+"'";return X.arrayBuffer()}).catch(function(){return Fd(Xt)});if(x)return new Promise(function(X,se){x(Xt,function(we){X(new Uint8Array(we))},se)})}return Promise.resolve().then(function(){return Fd(Xt)})}function Cg(){var X={env:Ip,wasi_snapshot_preview1:Ip};function se(Ue,We){var Pt=Ue.exports;s.asm=Pt,B=s.asm.memory,Ve(B.buffer),it=s.asm.__indirect_function_table,oa(s.asm.__wasm_call_ctors),qo("wasm-instantiate")}xr("wasm-instantiate");function we(Ue){se(Ue.instance)}function He(Ue){return Ng().then(function(We){return WebAssembly.instantiate(We,X)}).then(function(We){return We}).then(Ue,function(We){R("failed to asynchronously prepare wasm: "+We),er(We)})}function wt(){return!S&&typeof WebAssembly.instantiateStreaming=="function"&&!Ia(Xt)&&!Ko(Xt)&&!f&&typeof fetch=="function"?fetch(Xt,{credentials:"same-origin"}).then(function(Ue){var We=WebAssembly.instantiateStreaming(Ue,X);return We.then(we,function(Pt){return R("wasm streaming compile failed: "+Pt),R("falling back to ArrayBuffer instantiation"),He(we)})}):He(we)}if(s.instantiateWasm)try{var kt=s.instantiateWasm(X,se);return kt}catch(Ue){R("Module.instantiateWasm callback failed with error: "+Ue),o(Ue)}return wt().catch(o),{}}var G1,jr;function Xo(X){this.name="ExitStatus",this.message="Program terminated with exit("+X+")",this.status=X}function vr(X){for(;X.length>0;)X.shift()(s)}function H1(X){return X}function j1(X){var se=/\b_Z[\w\d_]+/g;return X.replace(se,function(we){var He=we;return we===He?we:He+" ["+we+"]"})}function Yo(){var X=new Error;if(!X.stack){try{throw new Error}catch(se){X=se}if(!X.stack)return"(no stack trace available)"}return X.stack.toString()}function _g(X,se){le.set(X,se)}function Dd(){er("")}function wp(){return 2147483648}function bn(){return wp()}function Rd(X,se,we){ie.copyWithin(X,se,se+we)}function Eg(X){try{return B.grow(X-te.byteLength+65535>>>16),Ve(B.buffer),1}catch(se){}}function Ag(X){var se=ie.length;X=X>>>0;var we=wp();if(X>we)return!1;let He=(Pt,la)=>Pt+(la-Pt%la)%la;for(var wt=1;wt<=4;wt*=2){var kt=se*(1+.2/wt);kt=Math.min(kt,X+100663296);var Ue=Math.min(we,He(Math.max(X,kt),65536)),We=Eg(Ue);if(We)return!0}return!1}var $g={varargs:void 0,get:function(){$g.varargs+=4;var X=xe[$g.varargs-4>>2];return X},getStr:function(X){var se=Q(X);return se}};function q1(X){return 52}function Fg(X,se,we,He,wt){return 70}var Ms=[null,[],[]];function Dg(X,se){var we=Ms[X];se===0||se===10?((X===1?A:R)(Z(we,0)),we.length=0):we.push(se)}function Rg(X,se,we,He){for(var wt=0,kt=0;kt<we;kt++){var Ue=Ie[se>>2],We=Ie[se+4>>2];se+=8;for(var Pt=0;Pt<We;Pt++)Dg(X,ie[Ue+Pt]);wt+=We}return Ie[He>>2]=wt,0}function kp(X){var se=s["_"+X];return se}function Md(X,se,we,He,wt){var kt={string:Mn=>{var Kr=0;if(Mn!=null&&Mn!==0){var Zd=(Mn.length<<2)+1;Kr=Cp(Zd),ae(Mn,Kr,Zd)}return Kr},array:Mn=>{var Kr=Cp(Mn.length);return _g(Mn,Kr),Kr}};function Ue(Mn){return se==="string"?Q(Mn):se==="boolean"?Boolean(Mn):Mn}var We=kp(X),Pt=[],la=0;if(He)for(var wr=0;wr<He.length;wr++){var Yd=kt[we[wr]];Yd?(la===0&&(la=qd()),Pt[wr]=Yd(He[wr])):Pt[wr]=He[wr]}var _p=We.apply(null,Pt);function ob(Mn){return la!==0&&Kd(la),Ue(Mn)}return _p=ob(_p),_p}function Pd(X,se,we,He){we=we||[];var wt=we.every(Ue=>Ue==="number"||Ue==="boolean"),kt=se!=="string";return kt&&wt&&!He?kp(X):function(){return Md(X,se,we,arguments,He)}}var Ip={abort:Dd,emscripten_get_heap_max:bn,emscripten_memcpy_big:Rd,emscripten_resize_heap:Ag,fd_close:q1,fd_seek:Fg,fd_write:Rg},K1=Cg(),Od=s.___wasm_call_ctors=function(){return(Od=s.___wasm_call_ctors=s.asm.__wasm_call_ctors).apply(null,arguments)},Mg=s._init=function(){return(Mg=s._init=s.asm.init).apply(null,arguments)},$e=s._init_with_threads_count=function(){return($e=s._init_with_threads_count=s.asm.init_with_threads_count).apply(null,arguments)},Sp=s._get_threads_count=function(){return(Sp=s._get_threads_count=s.asm.get_threads_count).apply(null,arguments)},Pg=s._register_tensor=function(){return(Pg=s._register_tensor=s.asm.register_tensor).apply(null,arguments)},X1=s._dispose_data=function(){return(X1=s._dispose_data=s.asm.dispose_data).apply(null,arguments)},Y1=s._dispose=function(){return(Y1=s._dispose=s.asm.dispose).apply(null,arguments)},Og=s._Abs=function(){return(Og=s._Abs=s.asm.Abs).apply(null,arguments)},Ld=s._Add=function(){return(Ld=s._Add=s.asm.Add).apply(null,arguments)},Zo=s._AddN=function(){return(Zo=s._AddN=s.asm.AddN).apply(null,arguments)},Lg=s._All=function(){return(Lg=s._All=s.asm.All).apply(null,arguments)},zg=s._Any=function(){return(zg=s._Any=s.asm.Any).apply(null,arguments)},Z1=s._ArgMax=function(){return(Z1=s._ArgMax=s.asm.ArgMax).apply(null,arguments)},Wg=s._AvgPool=function(){return(Wg=s._AvgPool=s.asm.AvgPool).apply(null,arguments)},Bg=s._BatchMatMul=function(){return(Bg=s._BatchMatMul=s.asm.BatchMatMul).apply(null,arguments)},Vg=s._Ceil=function(){return(Vg=s._Ceil=s.asm.Ceil).apply(null,arguments)},Ug=s._ClipByValue=function(){return(Ug=s._ClipByValue=s.asm.ClipByValue).apply(null,arguments)},zd=s._Conv2D=function(){return(zd=s._Conv2D=s.asm.Conv2D).apply(null,arguments)},Wd=s._Conv2DBackpropInput=function(){return(Wd=s._Conv2DBackpropInput=s.asm.Conv2DBackpropInput).apply(null,arguments)},Gg=s._Cos=function(){return(Gg=s._Cos=s.asm.Cos).apply(null,arguments)},Hg=s._Cosh=function(){return(Hg=s._Cosh=s.asm.Cosh).apply(null,arguments)},jg=s._CropAndResize=function(){return(jg=s._CropAndResize=s.asm.CropAndResize).apply(null,arguments)},Tp=s._Cumprod=function(){return(Tp=s._Cumprod=s.asm.Cumprod).apply(null,arguments)},qg=s._Cumsum=function(){return(qg=s._Cumsum=s.asm.Cumsum).apply(null,arguments)},Kg=s._DepthToSpace=function(){return(Kg=s._DepthToSpace=s.asm.DepthToSpace).apply(null,arguments)},Xg=s._DepthwiseConv2dNative=function(){return(Xg=s._DepthwiseConv2dNative=s.asm.DepthwiseConv2dNative).apply(null,arguments)},Ps=s._Elu=function(){return(Ps=s._Elu=s.asm.Elu).apply(null,arguments)},Yg=s._Equal=function(){return(Yg=s._Equal=s.asm.Equal).apply(null,arguments)},Zg=s._Exp=function(){return(Zg=s._Exp=s.asm.Exp).apply(null,arguments)},Bd=s._FlipLeftRight=function(){return(Bd=s._FlipLeftRight=s.asm.FlipLeftRight).apply(null,arguments)},Jg=s._Floor=function(){return(Jg=s._Floor=s.asm.Floor).apply(null,arguments)},Jo=s._FloorDiv=function(){return(Jo=s._FloorDiv=s.asm.FloorDiv).apply(null,arguments)},Qg=s._FusedBatchNorm=function(){return(Qg=s._FusedBatchNorm=s.asm.FusedBatchNorm).apply(null,arguments)},ey=s._FusedConv2D=function(){return(ey=s._FusedConv2D=s.asm.FusedConv2D).apply(null,arguments)},qr=s._FusedDepthwiseConv2D=function(){return(qr=s._FusedDepthwiseConv2D=s.asm.FusedDepthwiseConv2D).apply(null,arguments)},Np=s._Gather=function(){return(Np=s._Gather=s.asm.Gather).apply(null,arguments)},ty=s._GatherNd=function(){return(ty=s._GatherNd=s.asm.GatherNd).apply(null,arguments)},ny=s._Greater=function(){return(ny=s._Greater=s.asm.Greater).apply(null,arguments)},ay=s._GreaterEqual=function(){return(ay=s._GreaterEqual=s.asm.GreaterEqual).apply(null,arguments)},ry=s._LeakyRelu=function(){return(ry=s._LeakyRelu=s.asm.LeakyRelu).apply(null,arguments)},Vd=s._Less=function(){return(Vd=s._Less=s.asm.Less).apply(null,arguments)},Ud=s._LessEqual=function(){return(Ud=s._LessEqual=s.asm.LessEqual).apply(null,arguments)},sy=s._Log=function(){return(sy=s._Log=s.asm.Log).apply(null,arguments)},iy=s._LogicalAnd=function(){return(iy=s._LogicalAnd=s.asm.LogicalAnd).apply(null,arguments)},Gd=s._LogicalNot=function(){return(Gd=s._LogicalNot=s.asm.LogicalNot).apply(null,arguments)},Hd=s._LogicalOr=function(){return(Hd=s._LogicalOr=s.asm.LogicalOr).apply(null,arguments)},oy=s._LogicalXor=function(){return(oy=s._LogicalXor=s.asm.LogicalXor).apply(null,arguments)},ly=s._Max=function(){return(ly=s._Max=s.asm.Max).apply(null,arguments)},uy=s._MaxPool=function(){return(uy=s._MaxPool=s.asm.MaxPool).apply(null,arguments)},jd=s._Maximum=function(){return(jd=s._Maximum=s.asm.Maximum).apply(null,arguments)},J1=s._Mean=function(){return(J1=s._Mean=s.asm.Mean).apply(null,arguments)},py=s._Min=function(){return(py=s._Min=s.asm.Min).apply(null,arguments)},cy=s._Minimum=function(){return(cy=s._Minimum=s.asm.Minimum).apply(null,arguments)},dy=s._MirrorPad=function(){return(dy=s._MirrorPad=s.asm.MirrorPad).apply(null,arguments)},hy=s._Multiply=function(){return(hy=s._Multiply=s.asm.Multiply).apply(null,arguments)},my=s._Neg=function(){return(my=s._Neg=s.asm.Neg).apply(null,arguments)},fy=s._NonMaxSuppressionV3=function(){return(fy=s._NonMaxSuppressionV3=s.asm.NonMaxSuppressionV3).apply(null,arguments)},gy=s._NonMaxSuppressionV4=function(){return(gy=s._NonMaxSuppressionV4=s.asm.NonMaxSuppressionV4).apply(null,arguments)},yy=s._NonMaxSuppressionV5=function(){return(yy=s._NonMaxSuppressionV5=s.asm.NonMaxSuppressionV5).apply(null,arguments)},by=s._NotEqual=function(){return(by=s._NotEqual=s.asm.NotEqual).apply(null,arguments)},xy=s._OneHot=function(){return(xy=s._OneHot=s.asm.OneHot).apply(null,arguments)},vy=s._PadV2=function(){return(vy=s._PadV2=s.asm.PadV2).apply(null,arguments)},wy=s._Pow=function(){return(wy=s._Pow=s.asm.Pow).apply(null,arguments)},ky=s._Prelu=function(){return(ky=s._Prelu=s.asm.Prelu).apply(null,arguments)},Iy=s._Prod=function(){return(Iy=s._Prod=s.asm.Prod).apply(null,arguments)},Sy=s._RealDiv=function(){return(Sy=s._RealDiv=s.asm.RealDiv).apply(null,arguments)},Ty=s._Relu=function(){return(Ty=s._Relu=s.asm.Relu).apply(null,arguments)},Ny=s._Relu6=function(){return(Ny=s._Relu6=s.asm.Relu6).apply(null,arguments)},Cy=s._ResizeBilinear=function(){return(Cy=s._ResizeBilinear=s.asm.ResizeBilinear).apply(null,arguments)},_y=s._ResizeNearestNeighbor=function(){return(_y=s._ResizeNearestNeighbor=s.asm.ResizeNearestNeighbor).apply(null,arguments)},Ey=s._Reverse=function(){return(Ey=s._Reverse=s.asm.Reverse).apply(null,arguments)},Ay=s._RotateWithOffset=function(){return(Ay=s._RotateWithOffset=s.asm.RotateWithOffset).apply(null,arguments)},$y=s._Round=function(){return($y=s._Round=s.asm.Round).apply(null,arguments)},Fy=s._Rsqrt=function(){return(Fy=s._Rsqrt=s.asm.Rsqrt).apply(null,arguments)},Dy=s._ScatterNd=function(){return(Dy=s._ScatterNd=s.asm.ScatterNd).apply(null,arguments)},Ry=s._SelectV2=function(){return(Ry=s._SelectV2=s.asm.SelectV2).apply(null,arguments)},My=s._Sigmoid=function(){return(My=s._Sigmoid=s.asm.Sigmoid).apply(null,arguments)},Py=s._Sin=function(){return(Py=s._Sin=s.asm.Sin).apply(null,arguments)},Oy=s._Softmax=function(){return(Oy=s._Softmax=s.asm.Softmax).apply(null,arguments)},Ly=s._SparseFillEmptyRows=function(){return(Ly=s._SparseFillEmptyRows=s.asm.SparseFillEmptyRows).apply(null,arguments)},zy=s._SparseReshape=function(){return(zy=s._SparseReshape=s.asm.SparseReshape).apply(null,arguments)},Wy=s._SparseSegmentReduction=function(){return(Wy=s._SparseSegmentReduction=s.asm.SparseSegmentReduction).apply(null,arguments)},By=s._Sqrt=function(){return(By=s._Sqrt=s.asm.Sqrt).apply(null,arguments)},Vy=s._Square=function(){return(Vy=s._Square=s.asm.Square).apply(null,arguments)},Uy=s._SquaredDifference=function(){return(Uy=s._SquaredDifference=s.asm.SquaredDifference).apply(null,arguments)},Gy=s._Step=function(){return(Gy=s._Step=s.asm.Step).apply(null,arguments)},Hy=s._StridedSlice=function(){return(Hy=s._StridedSlice=s.asm.StridedSlice).apply(null,arguments)},jy=s._Sub=function(){return(jy=s._Sub=s.asm.Sub).apply(null,arguments)},qy=s._Sum=function(){return(qy=s._Sum=s.asm.Sum).apply(null,arguments)},Ky=s._Tan=function(){return(Ky=s._Tan=s.asm.Tan).apply(null,arguments)},Xy=s._Tanh=function(){return(Xy=s._Tanh=s.asm.Tanh).apply(null,arguments)},Yy=s._Tile=function(){return(Yy=s._Tile=s.asm.Tile).apply(null,arguments)},Zy=s._TopK=function(){return(Zy=s._TopK=s.asm.TopK).apply(null,arguments)},Jy=s._Transform=function(){return(Jy=s._Transform=s.asm.Transform).apply(null,arguments)},Qy=s._Transpose=function(){return(Qy=s._Transpose=s.asm.Transpose).apply(null,arguments)},eb=s.__FusedMatMul=function(){return(eb=s.__FusedMatMul=s.asm._FusedMatMul).apply(null,arguments)},tb=s._malloc=function(){return(tb=s._malloc=s.asm.malloc).apply(null,arguments)},nb=s._free=function(){return(nb=s._free=s.asm.free).apply(null,arguments)},ab=s.___errno_location=function(){return(ab=s.___errno_location=s.asm.__errno_location).apply(null,arguments)},qd=s.stackSave=function(){return(qd=s.stackSave=s.asm.stackSave).apply(null,arguments)},Kd=s.stackRestore=function(){return(Kd=s.stackRestore=s.asm.stackRestore).apply(null,arguments)},Cp=s.stackAlloc=function(){return(Cp=s.stackAlloc=s.asm.stackAlloc).apply(null,arguments)},rb=s.dynCall_iijjiiii=function(){return(rb=s.dynCall_iijjiiii=s.asm.dynCall_iijjiiii).apply(null,arguments)},sb=s.dynCall_jiji=function(){return(sb=s.dynCall_jiji=s.asm.dynCall_jiji).apply(null,arguments)};s.cwrap=Pd;var Qo;qn=function X(){Qo||Xd(),Qo||(qn=X)};function Xd(X){if(X=X||p,ot>0||(Mt(),ot>0))return;function se(){Qo||(Qo=!0,s.calledRun=!0,!U&&(ia(),i(s),s.onRuntimeInitialized&&s.onRuntimeInitialized(),on()))}s.setStatus?(s.setStatus("Running..."),setTimeout(function(){setTimeout(function(){s.setStatus("")},1),se()},1)):se()}if(s.preInit)for(typeof s.preInit=="function"&&(s.preInit=[s.preInit]);s.preInit.length>0;)s.preInit.pop()();Xd();var el;l&&(el={uncaughtException:process.listeners("uncaughtException").filter(function(X){return!l.uncaughtException.indexOf(X)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(X){return!l.unhandledRejection.indexOf(X)>-1})});var tl;if(typeof r!="undefined")tl=r;else if(typeof WasmBackendModuleThreadedSimd!="undefined")tl=WasmBackendModuleThreadedSimd;else throw new Error("Could not find wasm module in post.js");if(el){var ib=tl._dispose;tl._dispose=function(){ib(),el.uncaughtException.forEach(function(X){process.removeListener("uncaughtException",X)}),el.unhandledRejection.forEach(function(X){process.removeListener("unhandledRejection",X)})}}return r.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModule=n)}),om=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},pc=class{refCount(e){return Kn("refCount")}incRef(e){return Kn("incRef")}timerAvailable(){return!0}time(e){return Kn("time")}read(e){return Kn("read")}readSync(e){return Kn("readSync")}readToGPU(e,t){return Kn("readToGPU")}numDataIds(){return Kn("numDataIds")}disposeData(e,t){return Kn("disposeData")}write(e,t,n){return Kn("write")}move(e,t,n,a,r){return Kn("move")}createTensorFromTexture(e,t,n){return Kn("createTensorFromTexture")}memory(){return Kn("memory")}floatPrecision(){return Kn("floatPrecision")}epsilon(){return this.floatPrecision()===32?1e-7:1e-4}dispose(){return Kn("dispose")}};function Kn(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function UI(e){let t=e.length,n=0;for(;t>0;)n=Math.random()*t|0,t--,_h(e,t,n)}function LF(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let n=e.length,a=0;for(;n>0;)a=Math.random()*n|0,n--,_h(e,n,a),_h(t,n,a)}function Kp(e,t,n){return Math.max(e,Math.min(t,n))}function zF(e){return e%2===0?e:e+1}function _h(e,t,n){let a=e[t];e[t]=e[n],e[n]=a}function WF(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function BF(e,t){let n=Math.random();return t*n+(1-n)*e}function VF(e,t){let n=0;for(let a=0;a<e.length;a++){let r=Number(e[a])-Number(t[a]);n+=r*r}return n}function $(e,t){if(!e)throw new Error(typeof t=="string"?t:t())}function Tn(e,t,n=""){$(fs(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function mi(e){$(e!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Js(e,t=[],n=!1){if(t==null&&(t=[]),Array.isArray(e)||hn(e)&&!n)for(let a=0;a<e.length;++a)Js(e[a],t,n);else t.push(e);return t}function mt(e){if(e.length===0)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function UF(e){return e.length===0}function fs(e,t){if(e===t)return!0;if(e==null||t==null||e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function hl(e){return e%1===0}function GF(e){if(Math.tanh!=null)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{let t=Math.exp(2*e);return(t-1)/(t+1)}}function HF(e){let t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function jF(e){let t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return UI(t),t}function Gp(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function qF(e,t=r=>0,n,a){return new Promise((r,s)=>{let i=0,o=()=>{if(e()){r();return}i++;let l=t(i);if(n!=null&&i>=n){s();return}a!=null?a(o,l):setTimeout(o,l)};o()})}function KF(e,t){let n=1,a=-1;for(let s=0;s<e.length;++s)if(e[s]>=0)n*=e[s];else if(e[s]===-1){if(a!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${a} and dim ${s}`);a=s}else if(e[s]<0)throw Error(`Shapes can not be < 0. Found ${e[s]} at dim ${s}`);if(a===-1){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(n===0)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);let r=e.slice();return r[a]=t/n,r}function Fa(e,t){let n=t.length;return e=e==null?t.map((a,r)=>r):[].concat(e),$(e.every(a=>a>=-n&&a<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),$(e.every(a=>hl(a)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(a=>a<0?n+a:a)}function GI(e,t){let n=[],a=[],r=t!=null&&Array.isArray(t)&&t.length===0,s=t==null||r?null:Fa(t,e).sort(),i=0;for(let o=0;o<e.length;++o){if(s!=null){if(s[i]===o&&e[o]!==1)throw new Error(`Can't squeeze axis ${o} since its dim '${e[o]}' is not 1`);(s[i]==null||s[i]>o)&&e[o]===1&&(n.push(e[o]),a.push(o)),s[i]<=o&&i++}e[o]!==1&&(n.push(e[o]),a.push(o))}return{newShape:n,keptDims:a}}function HI(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else throw new Error(`Unknown data type ${e}`);return n}function jI(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else if(e==="string")n=new Array(t);else throw new Error(`Unknown data type ${e}`);return n}function qI(e,t){for(let n=0;n<e.length;n++){let a=e[n];if(isNaN(a)||!isFinite(a))throw Error(`A tensor of type ${t} being uploaded contains ${a}.`)}}function KI(e){return e==="bool"||e==="complex64"||e==="float32"||e==="int32"||e==="string"}function XF(e,t){return!(t==="complex64"||t==="float32"&&e!=="complex64"||t==="int32"&&e!=="float32"&&e!=="complex64"||t==="bool"&&e==="bool")}function hn(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}function Nb(e){if(e==="float32"||e==="int32")return 4;if(e==="complex64")return 8;if(e==="bool")return 1;throw new Error(`Unknown dtype ${e}`)}function XI(e){if(e==null)return 0;let t=0;return e.forEach(n=>t+=n.length),t}function Qr(e){return typeof e=="string"||e instanceof String}function YI(e){return typeof e=="boolean"}function ZI(e){return typeof e=="number"}function lm(e){return Array.isArray(e)?lm(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":ZI(e)?"float32":Qr(e)?"string":YI(e)?"bool":"float32"}function ss(e){return!!(e&&e.constructor&&e.call&&e.apply)}function Eh(e,t){for(let n=t;n<e;++n)if(e%n===0)return n;return e}function Fl(e){let t=e.length;if(t<2)return[];let n=new Array(t-1);n[t-2]=e[t-1];for(let a=t-3;a>=0;--a)n[a]=n[a+1]*e[a+1];return n}function JI(e,t,n,a=!1){let r=new Array;if(t.length===1){let s=t[0]*(a?2:1);for(let i=0;i<s;i++)r[i]=n[e+i]}else{let s=t[0],i=t.slice(1),o=i.reduce((l,u)=>l*u)*(a?2:1);for(let l=0;l<s;l++)r[l]=JI(e+l*o,i,n,a)}return r}function ul(e,t,n=!1){if(e.length===0)return t[0];let a=e.reduce((r,s)=>r*s)*(n?2:1);if(a===0)return[];if(a!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);return JI(0,e,t,n)}function _x(e,t){let n=um(e,t);for(let a=0;a<n.length;a++)n[a]=1;return n}function um(e,t){if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool")return new Uint8Array(e);throw new Error(`Unknown data type ${t}`)}function YF(e,t){let n=e.reduce((a,r)=>a*r,1);if(t==null||t==="float32")return ul(e,new Float32Array(n));if(t==="int32")return ul(e,new Int32Array(n));if(t==="bool")return ul(e,new Uint8Array(n));throw new Error(`Unknown data type ${t}`)}function Ex(e){e.forEach(t=>{$(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function ZF(e,t,n){if(t===0)return 0;if(t===1)return e[0];let a=e[e.length-1];for(let r=0;r<e.length-1;++r)a+=n[r]*e[r];return a}function JF(e,t,n){if(t===0)return[];if(t===1)return[e];let a=new Array(t);for(let r=0;r<a.length-1;++r)a[r]=Math.floor(e/n[r]),e-=a[r]*n[r];return a[a.length-1]=e,a}function Ax(e){return e&&e.then&&typeof e.then=="function"}var ok="tfjsflags",QI=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=QF,this.populateURLFlags()}setPlatform(e,t){this.platform!=null&&(H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},this.urlFlags[e]!=null){let a=this.urlFlags[e];H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${a}.`),this.set(e,a)}}async getAsync(e){return e in this.flags?this.flags[e]:(this.flags[e]=await this.evaluateFlag(e),this.flags[e])}get(e){if(e in this.flags)return this.flags[e];let t=this.evaluateFlag(e);if(Ax(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(this.flagRegistry[e]==null)throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,this.flagRegistry[e].setHook!=null&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(this.flagRegistry[e]==null)throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);return this.flagRegistry[e].evaluationFn()}setFlags(e){this.flags=Object.assign({},e)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global=="undefined"||typeof this.global.location=="undefined"||typeof this.global.location.search=="undefined")return;let e=this.getQueryParams(this.global.location.search);ok in e&&e[ok].split(",").forEach(t=>{let[n,a]=t.split(":");this.urlFlags[n]=tD(n,a)})}};function QF(e){let t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(n,...a)=>(eD(t,a[0],a[1]),a.join("="))),t}function eD(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}function tD(e,t){if(t=t.toLowerCase(),t==="true"||t==="false")return t==="true";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)}function H(){return $x}var $x=null;function nD(e){$x=e}var db;function eS(){if(db==null){let e;if(typeof window!="undefined")e=window;else if(typeof global!="undefined")e=global;else if(typeof process!="undefined")e=process;else if(typeof self!="undefined")e=self;else throw new Error("Could not find a global object");db=e}return db}function aD(){let e=eS();return e._tfGlobals==null&&(e._tfGlobals=new Map),e._tfGlobals}function Fx(e,t){let n=aD();if(n.has(e))return n.get(e);{let a=t();return n.set(e,a),n.get(e)}}var Dl="Abs",Rl="Acos",Ml="Acosh",gs="Add",fi="AddN",Pl="All",Ol="Any",gi="ArgMax",cc="ArgMin",Ll="Asin",zl="Asinh",Wl="Atan",Bl="Atanh",Vl="Atan2",yi="AvgPool",pm="AvgPoolGrad",dc="AvgPool3D",cm="AvgPool3DGrad",bi="BatchMatMul",Ul="BatchToSpaceND",dm="Bincount",tS="BroadcastTo",hm="BroadcastArgs",xi="Cast",vi="Ceil",ys="ClipByValue",mm="Complex",hc="ComplexAbs",Gl="Concat",wi="Conv2D",fm="Conv2DBackpropFilter",ki="Conv2DBackpropInput",mc="Conv3D",gm="Conv3DBackpropFilterV2",ym="Conv3DBackpropInputV2",Ii="Cos",Si="Cosh",Hl="Cumprod",Ti="Cumsum",jl="CropAndResize",bm="DenseBincount",ql="DepthToSpace",Ni="DepthwiseConv2dNative",xm="DepthwiseConv2dNativeBackpropFilter",vm="DepthwiseConv2dNativeBackpropInput",wm="Diag",fc="Dilation2D",Ah="Dilation2DBackpropInput",$h="Dilation2DBackpropFilter",Ci="RealDiv",km="Einsum",_i="Elu",Im="EluGrad",Kl="Erf",Xl="Equal",Ei="Exp",Yl="ExpandDims",Zl="Expm1",Sm="FFT",gc="Fill",Jl="FlipLeftRight",Ai="Floor",$i="FloorDiv",Fi="FusedBatchNorm",Ql="GatherV2",eu="GatherNd",tu="Greater",Di="GreaterEqual",Ri="Identity",Tm="IFFT",Nm="Imag",nu="IsFinite",au="IsInf",ru="IsNan",Mi="LeakyRelu",su="Less",iu="LessEqual",Cm="LinSpace",Pi="Log",ou="Log1p",lu="LogicalAnd",uu="LogicalNot",pu="LogicalOr",nS="LogicalXor",aS="LogSoftmax",rD="LowerBound",yc="LRN",_m="LRNGrad",Oi="Max",Li="Maximum",zi="MaxPool",Em="MaxPoolGrad",bc="MaxPool3D",Am="MaxPool3DGrad",$m="MaxPoolWithArgmax",Wi="Mean",Bi="Min",Vi="Minimum",Ui="MirrorPad",cu="Mod",Fm="Multinomial",Gi="Multiply",du="Neg",hu="NotEqual",mu="NonMaxSuppressionV3",fu="NonMaxSuppressionV4",gu="NonMaxSuppressionV5",yu="OnesLike",Hi="OneHot",bu="Pack",ji="PadV2",sD="Pool",qi="Pow",Ki="Prelu",Xi="Prod",Dm="RaggedGather",Rm="RaggedRange",Mm="RaggedTensorToTensor",xc="Range",Pm="Real",xu="Reciprocal",Yi="Relu",vu="Reshape",Zi="ResizeNearestNeighbor",Om="ResizeNearestNeighborGrad",Ji="ResizeBilinear",Lm="ResizeBilinearGrad",Qi="Relu6",eo="Reverse",to="Round",no="Rsqrt",wu="ScatterNd",zm="SearchSorted",ku="Select",Iu="Selu",Su="Slice",ao="Sin",Tu="Sinh",Nu="Sign",ro="Sigmoid",Cu="Softplus",so="Sqrt",io="Sum",_u="SpaceToBatchND",Eu="SplitV",oo="Softmax",vc="SparseFillEmptyRows",Au="SparseReshape",wc="SparseSegmentMean",kc="SparseSegmentSum",Wm="SparseToDense",lo="SquaredDifference",Ic="Square",$u="StridedSlice",Sc="StringNGrams",Tc="StringSplit",Nc="StringToHashBucketFast",uo="Sub",po="Tan",co="Tanh",bs="Tile",Fu="TopK",Du="Transform",_r="Transpose",Bm="Unique",Ru="Unpack",Cc="UnsortedSegmentSum",iD="UpperBound",Mu="ZerosLike",xs="Step",Fh="FromPixels",Pu="RotateWithOffset",Qs="_FusedMatMul",ei="FusedConv2D",ti="FusedDepthwiseConv2D";function Jr(...e){H().getBool("IS_TEST")||H().getBool("PROD")||console.warn(...e)}function oD(...e){H().getBool("IS_TEST")||H().getBool("PROD")||console.log(...e)}var ml=Fx("kernelRegistry",()=>new Map),Xp=Fx("gradRegistry",()=>new Map);function Dh(e,t){let n=Dx(e,t);return ml.get(n)}function Cb(e){return Xp.get(e)}function Rh(e){let t=ml.entries(),n=[];for(;;){let{done:a,value:r}=t.next();if(a)break;let[s,i]=r,[o]=s.split("_");o===e&&n.push(i)}return n}function _c(e){let{kernelName:t,backendName:n}=e,a=Dx(t,n);ml.has(a)&&Jr(`The kernel '${t}' for backend '${n}' is already registered`),ml.set(a,e)}function rS(e){let{kernelName:t}=e;Xp.has(t)&&H().getBool("DEBUG")&&Jr(`Overriding the gradient for '${t}'`),Xp.set(t,e)}function lD(e,t){let n=Dx(e,t);if(!ml.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);ml.delete(n)}function uD(e){if(!Xp.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);Xp.delete(e)}function pD(e,t){Rh(e).forEach(n=>{let a=Object.assign({},n,{backendName:t});_c(a)})}function Dx(e,t){return`${t}_${e}`}var v={};Ae(v,{arraysEqual:()=>fs,assert:()=>$,assertNonNegativeIntegerDimensions:()=>Ex,assertNonNull:()=>mi,assertShapesMatch:()=>Tn,bytesFromStringArray:()=>XI,bytesPerElement:()=>Nb,checkConversionForErrors:()=>qI,clamp:()=>Kp,computeStrides:()=>Fl,createScalarValue:()=>gD,createShuffledIndices:()=>jF,decodeString:()=>Mh,distSquared:()=>VF,encodeString:()=>Ac,fetch:()=>bD,fingerPrint64:()=>fD,flatten:()=>Js,getArrayFromDType:()=>jI,getTypedArrayFromDType:()=>HI,hasEncodingLoss:()=>XF,hexToLong:()=>Ec,indexToLoc:()=>JF,inferDtype:()=>lm,inferFromImplicitShape:()=>KF,isBoolean:()=>YI,isFunction:()=>ss,isInt:()=>hl,isNumber:()=>ZI,isPromise:()=>Ax,isScalarShape:()=>UF,isString:()=>Qr,isTypedArray:()=>hn,isValidDtype:()=>KI,locToIndex:()=>ZF,makeOnesTypedArray:()=>_x,makeZerosNestedTypedArray:()=>YF,makeZerosTypedArray:()=>um,nearestDivisor:()=>Eh,nearestLargerEven:()=>zF,now:()=>Yp,parseAxisParam:()=>Fa,randUniform:()=>BF,repeatedTry:()=>qF,rightPad:()=>Gp,shuffle:()=>UI,shuffleCombo:()=>LF,sizeFromShape:()=>mt,sizeToSquarishShape:()=>HF,squeezeShape:()=>GI,sum:()=>WF,swap:()=>_h,tanh:()=>GF,toNestedArray:()=>ul,toTypedArray:()=>Vm});var lk=ms(wF()),Bs=lk.default||lk;function Ec(e){return Bs.fromString(e,!0,16)}var sS=Ec("c3a5c85c97cb3127"),zs=Ec("b492b66fbe98f273"),xn=Ec("9ae16a3b2f90404f");function _b(e){return e.xor(e.shru(47))}function iS(e,t,n){let a=e.slice(t,t+n);return Bs.fromBytes(Array.from(a),!0,!0)}function ht(e,t){return iS(e,t,8)}function uk(e,t){return iS(e,t,4)}function Yt(e,t){return t===0?e:e.shru(t).or(e.shl(64-t))}function ns(e,t,n=Ec("9ddfea08eb382d69")){let a=e.xor(t).mul(n);a=a.xor(a.shru(47));let r=t.xor(a).mul(n);return r=r.xor(r.shru(47)),r=r.mul(n),r}function cD(e,t,n,a,r,s){r=r.add(e),s=Yt(s.add(r).add(a),21);let i=r;return r=r.add(t),r=r.add(n),s=s.add(Yt(r,44)),[r.add(a),s.add(i)]}function sh(e,t,n,a){return cD(ht(e,t),ht(e,t+8),ht(e,t+16),ht(e,t+24),n,a)}function dD(e,t=e.length){if(t>=8){let n=xn.add(t*2),a=ht(e,0).add(xn),r=ht(e,t-8),s=Yt(r,37).mul(n).add(a),i=Yt(a,25).add(r).mul(n);return ns(s,i,n)}if(t>=4){let n=xn.add(t*2),a=uk(e,0);return ns(a.shl(3).add(t),uk(e,t-4),n)}if(t>0){let n=e[0],a=e[t>>1],r=e[t-1],s=n+(a<<8),i=t+(r<<2);return _b(xn.mul(s).xor(sS.mul(i))).mul(xn)}return xn}function hD(e,t=e.length){let n=xn.add(t*2),a=ht(e,0).mul(zs),r=ht(e,8),s=ht(e,t-8).mul(n),i=ht(e,t-16).mul(xn);return ns(Yt(a.add(r),43).add(Yt(s,30)).add(i),a.add(Yt(r.add(xn),18)).add(s),n)}function mD(e,t=e.length){let n=xn.add(t*2),a=ht(e,0).mul(xn),r=ht(e,8),s=ht(e,t-8).mul(n),i=ht(e,t-16).mul(xn),o=Yt(a.add(r),43).add(Yt(s,30)).add(i),l=ns(o,a.add(Yt(r.add(xn),18)).add(s),n),u=ht(e,16).mul(n),p=ht(e,24),d=o.add(ht(e,t-32)).mul(n),c=l.add(ht(e,t-24)).mul(n);return ns(Yt(u.add(p),43).add(Yt(d,30)).add(c),u.add(Yt(p.add(a),18)).add(d),n)}function fD(e,t=e.length){let n=Bs.fromNumber(81,!0);if(t<=32)return t<=16?dD(e,t):hD(e,t);if(t<=64)return mD(e,t);let a=n,r=n.mul(zs).add(113),s=_b(r.mul(xn).add(113)).mul(xn),i=[Bs.UZERO,Bs.UZERO],o=[Bs.UZERO,Bs.UZERO];a=a.mul(xn).add(ht(e,0));let l=0,u=(t-1>>6)*64,p=u+(t-1&63)-63;do a=Yt(a.add(r).add(i[0]).add(ht(e,l+8)),37).mul(zs),r=Yt(r.add(i[1]).add(ht(e,l+48)),42).mul(zs),a=a.xor(o[1]),r=r.add(i[0]).add(ht(e,l+40)),s=Yt(s.add(o[0]),33).mul(zs),i=sh(e,l,i[1].mul(zs),a.add(o[0])),o=sh(e,l+32,s.add(o[1]),r.add(ht(e,l+16))),[s,a]=[a,s],l+=64;while(l!==u);let d=zs.add(s.and(255).shl(1));return l=p,o[0]=o[0].add(t-1&63),i[0]=i[0].add(o[0]),o[0]=o[0].add(i[0]),a=Yt(a.add(r).add(i[0]).add(ht(e,l+8)),37).mul(d),r=Yt(r.add(i[1]).add(ht(e,l+48)),42).mul(d),a=a.xor(o[1].mul(9)),r=r.add(i[0].mul(9).add(ht(e,l+40))),s=Yt(s.add(o[0]),33).mul(d),i=sh(e,l,i[1].mul(d),a.add(o[0])),o=sh(e,l+32,s.add(o[1]),r.add(ht(e,l+16))),[s,a]=[a,s],ns(ns(i[0],o[0],d).add(_b(r).mul(sS)).add(s),ns(i[1],o[1],d).add(a),d)}function gD(e,t){return t==="string"?Ac(e):Vm([e],t)}function yD(e,t){return e instanceof Float32Array&&t==="float32"||e instanceof Int32Array&&t==="int32"||e instanceof Uint8Array&&t==="bool"}function Vm(e,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Js(e)),H().getBool("DEBUG")&&qI(e,t),yD(e,t))return e;if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool"){let n=new Uint8Array(e.length);for(let a=0;a<n.length;++a)Math.round(e[a])!==0&&(n[a]=1);return n}else throw new Error(`Unknown data type ${t}`)}function Yp(){return H().platform.now()}function bD(e,t){return H().platform.fetch(e,t)}function Ac(e,t="utf-8"){return t=t||"utf-8",H().platform.encode(e,t)}function Mh(e,t="utf-8"){return t=t||"utf-8",H().platform.decode(e,t)}var xD=class{constructor(e,t){this.backendTimer=e,this.logger=t,t==null&&(this.logger=new wD)}profileKernel(e,t,n){let a,r=()=>{a=n()},s,i=Yp();if(this.backendTimer.timerAvailable())s=this.backendTimer.time(r);else{r();for(let o of a)o.dataSync();s=Promise.resolve({kernelMs:Yp()-i})}if(H().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let o=0;o<a.length;o++){let l=a[o];l.data().then(u=>{vD(u,l.dtype,e)})}return{kernelName:e,outputs:a,inputs:t,timeMs:s.then(o=>o.kernelMs),extraInfo:s.then(o=>o.getExtraProfileInfo!=null?o.getExtraProfileInfo():"")}}logKernelProfile(e){let{kernelName:t,outputs:n,timeMs:a,inputs:r,extraInfo:s}=e;n.forEach(i=>{Promise.all([i.data(),a,s]).then(o=>{this.logger.logKernelProfile(t,i,o[0],o[1],r,o[2])})})}};function vD(e,t,n){if(t!=="float32")return!1;for(let a=0;a<e.length;a++){let r=e[a];if(isNaN(r)||!isFinite(r))return console.warn(`Found ${r} in the result of '${n}'`),!0}return!1}var wD=class{logKernelProfile(e,t,n,a,r,s){let i=typeof a=="number"?Gp(`${a}ms`,9):a.error,o=Gp(e,25),l=t.rank,u=t.size,p=Gp(t.shape.toString(),14),d="";for(let c in r){let h=r[c];if(h!=null){let m=h.shape||t.shape,f=m.length;d+=`${c}: ${f}D ${f>0?m:""} `}}console.log(`%c${o}	%c${i}	%c${l}D ${p}	%c${u}	%c${d}	%c${s}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function kD(e,t,n){let a={},r={};for(let l=0;l<t.length;l++)a[t[l].id]=!0;for(let l=0;l<e.length;l++){let u=e[l],p=u.inputs;for(let d in p){let c=p[d],h=!1;for(let m=0;m<t.length;m++)if(a[c.id]){u.outputs.forEach(f=>a[f.id]=!0),h=!0,r[u.id]=!0;break}if(h)break}}let s={};s[n.id]=!0;let i={};for(let l=e.length-1;l>=0;l--){let u=e[l],p=u.inputs;for(let d=0;d<u.outputs.length;d++)if(s[u.outputs[d].id]){for(let c in p)s[p[c].id]=!0,i[u.id]=!0;break}}let o=[];for(let l=0;l<e.length;l++){let u=e[l];if(r[u.id]&&i[u.id]){let p={};for(let c in u.inputs){let h=u.inputs[c];a[h.id]&&(p[c]=h)}let d=Object.assign({},u);d.inputs=p,d.outputs=u.outputs,o.push(d)}}return o}function ID(e,t,n,a){for(let r=t.length-1;r>=0;r--){let s=t[r],i=[];if(s.outputs.forEach(l=>{let u=e[l.id];u!=null?i.push(u):i.push(null)}),s.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let o=s.gradient(i);for(let l in s.inputs){if(!(l in o))throw new Error(`Cannot backprop through input ${l}. Available gradients found: ${Object.keys(o)}.`);let u=n(()=>o[l]());if(u.dtype!=="float32")throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${l} must have 'float32' dtype, but has '${u.dtype}'`);let p=s.inputs[l];if(!fs(u.shape,p.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${l}' has shape '${u.shape}', which does not match the shape of the input '${p.shape}'`);if(e[p.id]==null)e[p.id]=u;else{let d=e[p.id];e[p.id]=a(d,u),d.dispose()}}}}var pk=20,Ep=3,hb=7;function SD(e,t,n,a){let r=Fl(t),s=TD(e,t,n,r),i=t.length,o=bh(e,t,n,r,s),l=["Tensor"];return a&&(l.push(`  dtype: ${n}`),l.push(`  rank: ${i}`),l.push(`  shape: [${t}]`),l.push("  values:")),l.push(o.map(u=>"    "+u).join(`
-`)),l.join(`
-`)}function TD(e,t,n,a){let r=mt(t),s=a[a.length-1],i=new Array(s).fill(0),o=t.length,l=n==="complex64"?Rp(e):e;if(o>1)for(let u=0;u<r/s;u++){let p=u*s;for(let d=0;d<s;d++)i[d]=Math.max(i[d],Dp(l[p+d],0,n).length)}return i}function Dp(e,t,n){let a;return Array.isArray(e)?a=`${parseFloat(e[0].toFixed(hb))} + ${parseFloat(e[1].toFixed(hb))}j`:Qr(e)?a=`'${e}'`:n==="bool"?a=oS(e):a=parseFloat(e.toFixed(hb)).toString(),Gp(a,t)}function oS(e){return e===0?"false":"true"}function bh(e,t,n,a,r,s=!0){let i=n==="complex64"?2:1,o=t[0],l=t.length;if(l===0){if(n==="complex64"){let f=Rp(e);return[Dp(f[0],0,n)]}return n==="bool"?[oS(e[0])]:[e[0].toString()]}if(l===1){if(o>pk){let g=Ep*i,y=Array.from(e.slice(0,g)),b=Array.from(e.slice((o-Ep)*i,o*i));return n==="complex64"&&(y=Rp(y),b=Rp(b)),["["+y.map((x,w)=>Dp(x,r[w],n)).join(", ")+", ..., "+b.map((x,w)=>Dp(x,r[o-Ep+w],n)).join(", ")+"]"]}let f=n==="complex64"?Rp(e):Array.from(e);return["["+f.map((g,y)=>Dp(g,r[y],n)).join(", ")+"]"]}let u=t.slice(1),p=a.slice(1),d=a[0]*i,c=[];if(o>pk){for(let f=0;f<Ep;f++){let g=f*d,y=g+d;c.push(...bh(e.slice(g,y),u,n,p,r,!1))}c.push("...");for(let f=o-Ep;f<o;f++){let g=f*d,y=g+d;c.push(...bh(e.slice(g,y),u,n,p,r,f===o-1))}}else for(let f=0;f<o;f++){let g=f*d,y=g+d;c.push(...bh(e.slice(g,y),u,n,p,r,f===o-1))}let h=l===2?",":"";c[0]="["+c[0]+h;for(let f=1;f<c.length-1;f++)c[f]=" "+c[f]+h;let m=`,
-`;for(let f=2;f<l;f++)m+=`
-`;return c[c.length-1]=" "+c[c.length-1]+"]"+(s?"":m),c}function Rp(e){let t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}var Ht=class{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=mt(e),n!=null){let a=n.length;$(a===this.size,()=>`Length of values '${a}' does not match the size inferred by the shape '${this.size}'.`)}if(t==="complex64")throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=n||jI(t,this.size),this.strides=Fl(e)}set(e,...t){t.length===0&&(t=[0]),$(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);let n=this.locToIndex(t);this.values[n]=e}get(...e){e.length===0&&(e=[0]);let t=0;for(let a of e){if(a<0||a>=this.shape[t]){let r=`Requested out of range element at ${e}.   Buffer shape=${this.shape}`;throw new Error(r)}t++}let n=e[e.length-1];for(let a=0;a<e.length-1;++a)n+=this.strides[a]*e[a];return this.values[n]}locToIndex(e){if(this.rank===0)return 0;if(this.rank===1)return e[0];let t=e[e.length-1];for(let n=0;n<e.length-1;++n)t+=this.strides[n]*e[n];return t}indexToLoc(e){if(this.rank===0)return[];if(this.rank===1)return[e];let t=new Array(this.shape.length);for(let n=0;n<t.length-1;++n)t[n]=Math.floor(e/this.strides[n]),e-=t[n]*this.strides[n];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return Oa().makeTensor(this.values,this.shape,this.dtype)}},Oa=null,il=null,ND=null;function CD(e){Oa=e}function _D(e){il=e}function ED(e){ND=e}var Te=class{constructor(e,t,n,a){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=mt(e),this.strides=Fl(e),this.dataId=n,this.id=a,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let e=await this.data();return il.buffer(this.shape,this.dtype,e)}bufferSync(){return il.buffer(this.shape,this.dtype,this.dataSync())}async array(){let e=await this.data();return ul(this.shape,e,this.dtype==="complex64")}arraySync(){return ul(this.shape,this.dataSync(),this.dtype==="complex64")}async data(){this.throwIfDisposed();let e=Oa().read(this.dataId);if(this.dtype==="string"){let t=await e;try{return t.map(n=>Mh(n))}catch(n){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),Oa().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();let e=Oa().readSync(this.dataId);if(this.dtype==="string")try{return e.map(t=>Mh(t))}catch(t){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();let e=await Oa().read(this.dataId);return this.dtype==="string"?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(Oa().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return il.print(this,e)}clone(){return this.throwIfDisposed(),il.clone(this)}toString(e=!1){let t=this.dataSync();return SD(t,this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),il.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),Oa().makeVariable(this,e,t,n)}};Object.defineProperty(Te,Symbol.hasInstance,{value:e=>!!e&&e.data!=null&&e.dataSync!=null&&e.throwIfDisposed!=null});function J(){return Fx("Tensor",()=>Te)}J();var is=class extends Te{constructor(e,t,n,a){super(e.shape,e.dtype,e.dataId,a),this.trainable=t,this.name=n}assign(e){if(e.dtype!==this.dtype)throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);if(!fs(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);Oa().disposeTensor(this),this.dataId=e.dataId,Oa().incRef(this,null)}dispose(){Oa().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(is,Symbol.hasInstance,{value:e=>e instanceof Te&&e.assign!=null&&e.assign instanceof Function});var Va={};Ae(Va,{assertTypesMatch:()=>lS,getTensorsInContainer:()=>Rx,isTensorInList:()=>$D,makeTypesMatch:()=>_t});var Eb;(function(e){e.R0="R0",e.R1="R1",e.R2="R2",e.R3="R3",e.R4="R4",e.R5="R5",e.R6="R6"})(Eb||(Eb={}));var Ab;(function(e){e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64"})(Ab||(Ab={}));var $b;(function(e){e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64"})($b||($b={}));var Fb;(function(e){e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64"})(Fb||(Fb={}));var Db;(function(e){e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64"})(Db||(Db={}));var AD={float32:Fb,int32:Ab,bool:$b,complex64:Db};function fa(e,t){if(e==="string"||t==="string"){if(e==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return AD[e][t]}function Um(e){return fa(e,"int32")}function _t(e,t){if(e.dtype===t.dtype)return[e,t];let n=fa(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function lS(e,t){$(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function $D(e,t){return t.some(n=>n.id===e.id)}function Rx(e){let t=[];return uS(e,t,new Set),t}function uS(e,t,n){if(e==null)return;if(e instanceof Te){t.push(e);return}if(!FD(e))return;let a=e;for(let r in a){let s=a[r];n.has(s)||(n.add(s),uS(s,t,n))}}function FD(e){return Array.isArray(e)||typeof e=="object"}function mb(e){return e.kernelName!=null}var ck=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(let e in this.registeredVariables)this.registeredVariables[e].dispose()}},Zp=class{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new ck}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t];if(await this.initializeBackend(n).success){await this.setBackend(n);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry))if(e in this.registryFactory){let{asyncInit:t}=this.initializeBackend(e);if(t)return null}else return null;return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(Jr(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:n},!0)}async setBackend(e){if(this.registryFactory[e]==null)throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,this.registry[e]==null){this.backendInstance=null;let{success:t,asyncInit:n}=this.initializeBackend(e);if(!(n?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new xD(this.backendInstance),!0}setupRegisteredKernels(){Rh(this.backendName).forEach(e=>{e.setupFunc!=null&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){Rh(e).forEach(t=>{t.disposeFunc!=null&&t.disposeFunc(this.registry[e])})}initializeBackend(e){let t=this.registryFactory[e];if(t==null)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{let n=t.factory();if(n&&!(n instanceof pc)&&typeof n.then=="function"){let a=++this.pendingBackendInitId,r=n.then(s=>a<this.pendingBackendInitId?!1:(this.registry[e]=s,this.pendingBackendInit=null,!0)).catch(s=>(a<this.pendingBackendInitId||(this.pendingBackendInit=null,Jr(`Initialization of backend ${e} failed`),Jr(s.stack||s.message)),!1));return this.pendingBackendInit=r,{success:r,asyncInit:!0}}else return this.registry[e]=n,{success:!0,asyncInit:!1}}catch(n){return Jr(`Initialization of backend ${e} failed`),Jr(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${e} backend not found in registry`);this.backendName===e&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,e in this.registry&&(this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e]),delete this.registryFactory[e],this.backendName===e&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t],{success:a,asyncInit:r}=this.initializeBackend(n);if(r||a)return{name:n,asyncInit:r}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){let n=this.state.tensorInfo.get(t),a=n.backend,r=this.readSync(t),s=a.refCount(t);a.disposeData(t,!0),n.backend=e,e.move(t,r,n.shape,n.dtype,s),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n=null;if(t==null){if(typeof e!="function")throw new Error("Please provide a function to tidy()");t=e}else{if(typeof e!="string"&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if(typeof t!="function")throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");n=e}let a;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(a),()=>(a=t(),a instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),a))}scopedRun(e,t,n){e();try{let a=n();return t(),a}catch(a){throw t(),a}}nextTensorId(){return Zp.nextTensorId++}nextVariableId(){return Zp.nextVariableId++}clone(e){let t=O.runKernel(Ri,{x:e}),n={x:e},a=s=>({x:()=>{let i="float32",o={x:s},l={dtype:i};return O.runKernel(xi,o,l)}}),r=[];return this.addTapeNode(this.state.activeScope.name,n,[t],a,r,{}),t}runKernel(e,t,n){if(this.backendName==null&&this.backend,Dh(e,this.backendName)==null)throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){let a=this.backend.numDataIds(),r=0;n.forEach(o=>{r+=o.dtype==="complex64"?3:1});let s=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],i=a-t-r-s;if(i>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)}runKernelFunc(e){let t,n=[],a=this.isTapeOn(),r=this.state.numBytes,s=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let i;this.backendName==null&&this.backend;let o,l=mb(e)?e.kernelName:this.state.activeScope!=null?this.state.activeScope.name:"";if(mb(e)){let{kernelName:h,inputs:m,attrs:f}=e;this.backendName==null&&this.backend;let g=Dh(h,this.backendName);$(g!=null,()=>`Cannot find registered kernel '${h}' for backend '${this.backendName}'`),i=()=>{let y=this.backend.numDataIds();o=g.kernelFunc({inputs:m,attrs:f,backend:this.backend});let b=Array.isArray(o)?o:[o];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(h,y,b);let x=b.map(w=>w.rank!=null?w:this.makeTensorFromTensorInfo(w));if(a){let w=this.getTensorsForGradient(h,m,x);n=this.saveTensorsForBackwardMode(w)}return x}}else{let{forwardFunc:h}=e,m=f=>{!a||(n=f.map(g=>this.keep(this.clone(g))))};i=()=>{let f=this.backend.numDataIds();o=this.tidy(()=>h(this.backend,m));let g=Array.isArray(o)?o:[o];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,f,g),g}}let{inputs:u,attrs:p}=e,d=mb(e)?null:e.backwardsFunc,c;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?t=i():(c=this.profiler.profileKernel(l,u,()=>i()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(c),t=c.outputs)}),a&&this.addTapeNode(l,u,t,d,n,p),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-r,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-s,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(u).map(h=>u[h]!=null?u[h].shape:null),outputShapes:t.map(h=>h.shape),kernelTimeMs:c.timeMs,extraInfo:c.extraInfo}),Array.isArray(o)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(t=>this.keep(this.clone(t)))}getTensorsForGradient(e,t,n){let a=Cb(e);if(a!=null){let r=a.inputsToSave||[],s=a.outputsToSave||[],i;a.saveAllInputs?($(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),i=Object.keys(t).map(l=>t[l])):i=r.map(l=>t[l]);let o=n.filter((l,u)=>s[u]);return i.concat(o)}return[]}makeTensor(e,t,n,a){if(e==null)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",a=a||this.backend;let r=e;n==="string"&&Qr(e[0])&&(r=e.map(o=>Ac(o)));let s=a.write(r,t,n),i=new Te(t,n,s,this.nextTensorId());if(this.trackTensor(i,a),n==="string"){let o=this.state.tensorInfo.get(s),l=XI(r);this.state.numBytes+=l-o.bytes,o.bytes=l}return i}makeTensorFromDataId(e,t,n,a){n=n||"float32";let r={dataId:e,shape:t,dtype:n};return this.makeTensorFromTensorInfo(r,a)}makeTensorFromTensorInfo(e,t){let{dataId:n,shape:a,dtype:r}=e,s=new Te(a,r,n,this.nextTensorId());return this.trackTensor(s,t),s}makeVariable(e,t=!0,n,a){n=n||this.nextVariableId().toString(),a!=null&&a!==e.dtype&&(e=e.cast(a));let r=new is(e,t,n,this.nextTensorId());if(this.state.registeredVariables[r.name]!=null)throw new Error(`Variable with name ${r.name} was already registered`);return this.state.registeredVariables[r.name]=r,this.incRef(r,this.backend),r}trackTensor(e,t){this.state.numTensors++,e.dtype==="string"&&this.state.numStringTensors++;let n=0;e.dtype!=="complex64"&&e.dtype!=="string"&&(n=e.size*Nb(e.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:n})),e instanceof is||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;let t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,e.dtype==="string"&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),e.dtype!=="complex64"&&e.dtype!=="string"){let n=e.size*Nb(e.dtype);this.state.numBytes-=n}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(let e in this.state.registeredVariables){let t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),this.state.registeredVariables[e.name]!=null&&delete this.state.registeredVariables[e.name]}memory(){let e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,e.reasons==null&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;let t=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await e(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(a=>a.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let a of this.state.activeProfile.kernels)a.kernelTimeMs=await a.kernelTimeMs,a.extraInfo=await a.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(e,t,n,a,r,s){let i={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:r},o=Cb(e);o!=null&&(a=o.gradFunc),a!=null&&(i.gradient=l=>(l=l.map((u,p)=>{if(u==null){let d=n[p],c=um(d.size,d.dtype);return this.makeTensor(c,d.shape,d.dtype)}return u}),a(l.length>1?l:l[0],r,s))),this.state.activeTape.push(i)}keep(e){return e.kept=!0,e}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){let t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){let t=Rx(e),n=new Set(t.map(r=>r.id));for(let r=0;r<this.state.activeScope.track.length;r++){let s=this.state.activeScope.track[r];!s.kept&&!n.has(s.id)&&s.dispose()}let a=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(r=>{!r.kept&&r.scopeId===a.id&&this.track(r)})}gradients(e,t,n,a=!1){if($(t.length>0,()=>"gradients() received an empty list of xs."),n!=null&&n.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let r=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));$(r instanceof Te,()=>"The result y returned by f() must be a tensor.");let s=kD(this.state.activeTape,t,r);if(!a&&s.length===0&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let i={};i[r.id]=n==null?DD(r.shape):n,ID(i,s,l=>this.tidy(l),RD);let o=t.map(l=>i[l.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(l=>{for(let u of l.saved)u.dispose()}),this.state.activeTape=null),{value:r,grads:o}})}customGrad(e){return $(ss(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{$(t.every(i=>i instanceof Te),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n,a={};t.forEach((i,o)=>{a[o]=i});let r=(i,o)=>(n=e(...t,o),$(n.value instanceof Te,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),$(ss(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),s=(i,o)=>{let l=n.gradFunc(i,o),u=Array.isArray(l)?l:[l];$(u.length===t.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),$(u.every(d=>d instanceof Te),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let p={};return u.forEach((d,c)=>{p[c]=()=>d}),p};return this.runKernelFunc({forwardFunc:r,backwardsFunc:s,inputs:a})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){let t=Yp(),n=await this.backend.time(e);return n.wallMs=Yp()-t,n}track(e){return this.state.activeScope!=null&&(e.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(e)),e}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new ck;for(let e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};Zp.nextTensorId=0;Zp.nextVariableId=0;function DD(e){let t=_x(mt(e),"float32");return O.makeTensor(t,e,"float32")}function pS(){let e=eS();if(e._tfengine==null){let t=new QI(e);e._tfengine=new Zp(t)}return nD(e._tfengine.ENV),CD(()=>e._tfengine),e._tfengine}var O=pS();function RD(e,t){let n={a:e,b:t};return O.runKernel(gs,n)}var $c={};Ae($c,{isBrowser:()=>cS,isMobile:()=>OD,mockIsMobile:()=>PD});function MD(){return typeof navigator!="undefined"&&navigator!=null}var Rb;function PD(e){Rb=e}function OD(e){if(Rb!==void 0)return Rb;if(e||MD()){if(e||(e=navigator),e.product==="ReactNative")return!0;let t=e.userAgent||e.vendor||(typeof window!="undefined"?window.opera:"");if(!t){let n=e;return n.userAgentData&&n.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function cS(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}var Qn=H();Qn.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")});Qn.registerFlag("IS_BROWSER",()=>cS());Qn.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");Qn.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));Qn.registerFlag("PROD",()=>!1);Qn.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>Qn.getBool("DEBUG"));Qn.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);Qn.registerFlag("IS_TEST",()=>!1);Qn.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0);Qn.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);Qn.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);Qn.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU",()=>!1);Qn.registerFlag("USE_SETTIMEOUTCUSTOM",()=>!1);function lr(e,t){let n=e;if(hn(e))return t==="string"?[]:[e.length];if(typeof e=="object"&&"texture"in e){let r=e.channels||"RGBA";return[e.height,e.width*r.length]}if(!Array.isArray(e))return[];let a=[];for(;Array.isArray(n)||hn(n)&&t!=="string";)a.push(n.length),n=n[0];return Array.isArray(e)&&H().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&dS(e,a,[]),a}function dS(e,t,n){if(n=n||[],!Array.isArray(e)&&!hn(e)){$(t.length===0,()=>`Element arr[${n.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}$(t.length>0,()=>`Element arr[${n.join("][")}] should be a primitive, but is an array of ${e.length} elements`),$(e.length===t[0],()=>`Element arr[${n.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);let a=t.slice(1);for(let r=0;r<e.length;++r)dS(e[r],a,n.concat(r))}function dk(e,t,n,a){if(e!=="string_or_numeric"){if(e==null)throw new Error("Expected dtype cannot be null.");if(e!=="numeric"&&e!==t||e==="numeric"&&t==="string")throw new Error(`Argument '${n}' passed to '${a}' must be ${e} tensor, but got ${t} tensor`)}}function _(e,t,n,a="numeric"){if(e instanceof Te)return dk(a,e.dtype,t,n),e;let r=lm(e);if(r!=="string"&&["bool","int32","float32"].indexOf(a)>=0&&(r=a),dk(a,r,t,n),e==null||!hn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string"){let o=e==null?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${o}'`)}let s=lr(e,r);!hn(e)&&!Array.isArray(e)&&(e=[e]);let i=r!=="string"?Vm(e,r):Js(e,[],!0);return O.makeTensor(i,s,r)}function Jp(e,t,n,a="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((r,s)=>_(r,`${t}[${s}]`,n,a))}var Mx="__op";function L(e){let t=Object.keys(e);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let n=t[0],a=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1)),n=n+Mx;let r=(...s)=>{O.startScope(n);try{let i=a(...s);return Ax(i)&&console.error("Cannot return a Promise inside of tidy."),O.endScope(i),i}catch(i){throw O.endScope(null),i}};return Object.defineProperty(r,"name",{value:n,configurable:!0}),r}function LD(e,t){let n=_(e,"real","complex"),a=_(t,"imag","complex");Tn(n.shape,a.shape,`real and imag shapes, ${n.shape} and ${a.shape}, must match in call to tf.complex().`);let r={real:n,imag:a};return O.runKernel(mm,r)}var Ar=L({complex_:LD});function vs(e,t,n,a){if(a==null&&(a=lm(e)),a==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(typeof e=="object"&&"texture"in e){if(a!=="float32"&&a!=="int32")throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${a}.`);return e.channels=e.channels||"RGBA",O.backend.createTensorFromTexture(e,t||n,a)}if(!hn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){Ex(t);let r=mt(t),s=mt(n);$(r===s,()=>`Based on the provided shape, [${t}], the tensor should have ${r} values but has ${s}`);for(let i=0;i<n.length;++i){let o=n[i],l=i===n.length-1?o!==mt(t.slice(i)):!0;$(n[i]===t[i]||!l,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return!hn(e)&&!Array.isArray(e)&&(e=[e]),t=t||n,e=a!=="string"?Vm(e,a):Js(e,[],!0),O.makeTensor(e,t,a)}function In(e,t,n){let a=lr(e,n);return vs(e,t,a,n)}var Mb={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8},Ph=4;async function zD(e,t){let n=[],a=[],r=Array.isArray(e)?e.map(i=>i.name):Object.keys(e);for(let i=0;i<r.length;++i){let o=r[i],l=Array.isArray(e)?e[i].tensor:e[o];if(l.dtype!=="float32"&&l.dtype!=="int32"&&l.dtype!=="bool"&&l.dtype!=="string"&&l.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${o}': ${l.dtype}`);let u={name:o,shape:l.shape,dtype:l.dtype};if(l.dtype==="string"){let p=new Promise(async d=>{let c=await l.bytes(),h=c.reduce((g,y)=>g+y.length,0)+Ph*c.length,m=new Uint8Array(h),f=0;for(let g=0;g<c.length;g++){let y=c[g],b=new Uint8Array(new Uint32Array([y.length]).buffer);m.set(b,f),f+=Ph,m.set(y,f),f+=y.length}d(m)});a.push(p)}else a.push(l.data());t!=null&&(u.group=t),n.push(u)}let s=await Promise.all(a);return{data:WD(s),specs:n}}function hS(e,t){let n={},a,r=0;for(let s of t){let i=s.name,o=s.dtype,l=s.shape,u=mt(l),p;if("quantization"in s){let d=s.quantization;if(d.dtype==="uint8"||d.dtype==="uint16"){if(!("min"in d&&"scale"in d))throw new Error(`Weight ${s.name} with quantization ${d.dtype} doesn't have corresponding metadata min and scale.`)}else if(d.dtype==="float16"){if(o!=="float32")throw new Error(`Weight ${s.name} is quantized with ${d.dtype} which only supports weights of type float32 not ${o}.`)}else throw new Error(`Weight ${s.name} has unknown quantization dtype ${d.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let c=Mb[d.dtype],h=e.slice(r,r+u*c),m=d.dtype==="uint8"?new Uint8Array(h):new Uint16Array(h);if(o==="float32")if(d.dtype==="uint8"||d.dtype==="uint16"){p=new Float32Array(m.length);for(let f=0;f<m.length;f++){let g=m[f];p[f]=g*d.scale+d.min}}else if(d.dtype==="float16")a===void 0&&(a=jD()),p=a(m);else throw new Error(`Unsupported quantization type ${d.dtype} for weight type float32.`);else if(o==="int32"){if(d.dtype!=="uint8"&&d.dtype!=="uint16")throw new Error(`Unsupported quantization type ${d.dtype} for weight type int32.`);p=new Int32Array(m.length);for(let f=0;f<m.length;f++){let g=m[f];p[f]=Math.round(g*d.scale+d.min)}}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);r+=u*c}else if(o==="string"){let d=mt(s.shape);p=[];for(let c=0;c<d;c++){let h=new Uint32Array(e.slice(r,r+Ph))[0];r+=Ph;let m=new Uint8Array(e.slice(r,r+h));p.push(m),r+=h}}else{let d=Mb[o],c=e.slice(r,r+u*d);if(o==="float32")p=new Float32Array(c);else if(o==="int32")p=new Int32Array(c);else if(o==="bool")p=new Uint8Array(c);else if(o==="complex64"){p=new Float32Array(c);let h=new Float32Array(p.length/2),m=new Float32Array(p.length/2);for(let y=0;y<h.length;y++)h[y]=p[y*2],m[y]=p[y*2+1];let f=In(h,l,"float32"),g=In(m,l,"float32");n[i]=Ar(f,g),f.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);r+=u*d}o!=="complex64"&&(n[i]=In(p,l,o))}return n}function WD(e){if(e===null)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0,n=[];e.forEach(s=>{if(t+=s.byteLength,n.push(s.byteLength===s.buffer.byteLength?s:new s.constructor(s)),!(s instanceof Float32Array||s instanceof Int32Array||s instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${s.constructor.name}`)});let a=new Uint8Array(t),r=0;return n.forEach(s=>{a.set(new Uint8Array(s.buffer),r),r+=s.byteLength}),a.buffer}var Px=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function hk(e){return Px?Buffer.byteLength(e):new Blob([e]).size}function BD(e){if(Px)return Buffer.from(e).toString("base64");let t=new Uint8Array(e),n="";for(let a=0,r=t.length;a<r;a++)n+=String.fromCharCode(t[a]);return btoa(n)}function VD(e){if(Px){let a=Buffer.from(e,"base64");return a.buffer.slice(a.byteOffset,a.byteOffset+a.byteLength)}let t=atob(e),n=new Uint8Array(t.length);for(let a=0;a<t.length;++a)n.set([t.charCodeAt(a)],a);return n.buffer}function Ox(e){if(e.length===1)return e[0];let t=0;e.forEach(r=>{t+=r.byteLength});let n=new Uint8Array(t),a=0;return e.forEach(r=>{n.set(new Uint8Array(r),a),a+=r.byteLength}),n.buffer}function mk(e){let t="/";for(e=e.trim();e.endsWith(t);)e=e.slice(0,e.length-1);let n=e.split(t);return n[n.length-1]}function mS(e,t){let n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return e.signature!=null&&(n.signature=e.signature),e.userDefinedMetadata!=null&&(n.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(n.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(n.initializerSignature=e.initializerSignature),e.trainingConfig!=null&&(n.trainingConfig=e.trainingConfig),n}function fS(e,t,n){let a={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(e.trainingConfig!=null&&(a.trainingConfig=e.trainingConfig),e.weightsManifest!=null){if(!t)throw new Error("modelJSON has weightsManifest but weightSpecs is null");if(!n)throw new Error("modelJSON has weightsManifest but weightData is null");a.weightSpecs=t,a.weightData=n}return e.signature!=null&&(a.signature=e.signature),e.userDefinedMetadata!=null&&(a.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(a.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(a.initializerSignature=e.initializerSignature),a}async function Lx(e,t){let n,a;return e.weightsManifest!=null&&([n,a]=await t(e.weightsManifest)),fS(e,n,a)}function Fc(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:e.modelTopology==null?0:hk(JSON.stringify(e.modelTopology)),weightSpecsBytes:e.weightSpecs==null?0:hk(JSON.stringify(e.weightSpecs)),weightDataBytes:e.weightData==null?0:e.weightData.byteLength}}function gS(e){let t=[];for(let n of e)t.push(...n.weights);return t}function UD(){let e=n=>{let a=n<<13,r=0;for(;(a&8388608)===0;)r-=8388608,a<<=1;return a&=-8388609,r+=947912704,a|r},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let n=1024;n<2048;n++)t[n]=939524096+(n-1024<<13);return t}function GD(){let e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}function HD(){let e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}function jD(){let e=UD(),t=GD(),n=HD();return a=>{let r=new ArrayBuffer(4*a.length),s=new Uint32Array(r);for(let i=0;i<a.length;i++){let o=a[i],l=e[n[o>>10]+(o&1023)]+t[o>>10];s[i]=l}return new Float32Array(r)}}var At=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return At.instance==null&&(At.instance=new At),At.instance}static registerSaveRouter(e){At.getInstance().saveRouters.push(e)}static registerLoadRouter(e){At.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return At.getHandlers(e,"save")}static getLoadHandlers(e,t){return At.getHandlers(e,"load",t)}static getHandlers(e,t,n){let a=[];return(t==="load"?At.getInstance().loadRouters:At.getInstance().saveRouters).forEach(r=>{let s=r(e,n);s!==null&&a.push(s)}),a}},qD=e=>At.registerSaveRouter(e),KD=e=>At.registerLoadRouter(e),XD=e=>At.getSaveHandlers(e),YD=(e,t)=>At.getLoadHandlers(e,t),Pb="tensorflowjs",Ob=1,Hs="models_store",es="model_info_store";function yS(){if(!H().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let e=typeof window=="undefined"?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(t==null)throw new Error("The current browser does not appear to support IndexedDB.");return t}function Lb(e){let t=e.result;t.createObjectStore(Hs,{keyPath:"modelPath"}),t.createObjectStore(es,{keyPath:"modelPath"})}var ni=class{constructor(e){if(this.indexedDB=yS(),e==null||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((n,a)=>{let r=this.indexedDB.open(Pb,Ob);r.onupgradeneeded=()=>Lb(r),r.onsuccess=()=>{let s=r.result;if(t==null){let i=s.transaction(Hs,"readonly"),o=i.objectStore(Hs).get(this.modelPath);o.onsuccess=()=>{if(o.result==null)return s.close(),a(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));n(o.result.modelArtifacts)},o.onerror=l=>(s.close(),a(o.error)),i.oncomplete=()=>s.close()}else{let i=Fc(t),o=s.transaction(es,"readwrite"),l=o.objectStore(es),u=l.put({modelPath:this.modelPath,modelArtifactsInfo:i}),p;u.onsuccess=()=>{p=s.transaction(Hs,"readwrite");let d=p.objectStore(Hs).put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:i});d.onsuccess=()=>n({modelArtifactsInfo:i}),d.onerror=c=>{l=o.objectStore(es);let h=l.delete(this.modelPath);h.onsuccess=()=>(s.close(),a(d.error)),h.onerror=m=>(s.close(),a(d.error))}},u.onerror=d=>(s.close(),a(u.error)),o.oncomplete=()=>{p==null?s.close():p.oncomplete=()=>s.close()}}},r.onerror=s=>a(r.error)})}};ni.URL_SCHEME="indexeddb://";var bS=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ni.URL_SCHEME)?ZD(e.slice(ni.URL_SCHEME.length)):null;At.registerSaveRouter(bS);At.registerLoadRouter(bS);function ZD(e){return new ni(e)}function JD(e){return e.startsWith(ni.URL_SCHEME)?e.slice(ni.URL_SCHEME.length):e}var QD=class{constructor(){this.indexedDB=yS()}async listModels(){return new Promise((e,t)=>{let n=this.indexedDB.open(Pb,Ob);n.onupgradeneeded=()=>Lb(n),n.onsuccess=()=>{let a=n.result,r=a.transaction(es,"readonly"),s=r.objectStore(es).getAll();s.onsuccess=()=>{let i={};for(let o of s.result)i[o.modelPath]=o.modelArtifactsInfo;e(i)},s.onerror=i=>(a.close(),t(s.error)),r.oncomplete=()=>a.close()},n.onerror=a=>t(n.error)})}async removeModel(e){return e=JD(e),new Promise((t,n)=>{let a=this.indexedDB.open(Pb,Ob);a.onupgradeneeded=()=>Lb(a),a.onsuccess=()=>{let r=a.result,s=r.transaction(es,"readwrite"),i=s.objectStore(es),o=i.get(e),l;o.onsuccess=()=>{if(o.result==null)return r.close(),n(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{let u=i.delete(e),p=()=>{l=r.transaction(Hs,"readwrite");let d=l.objectStore(Hs).delete(e);d.onsuccess=()=>t(o.result.modelArtifactsInfo),d.onerror=c=>n(o.error)};u.onsuccess=p,u.onerror=d=>(p(),r.close(),n(o.error))}},o.onerror=u=>(r.close(),n(o.error)),s.oncomplete=()=>{l==null?r.close():l.oncomplete=()=>r.close()}},a.onerror=r=>n(a.error)})}},Nr="/",ol="tensorflowjs_models",xS="info",eR="model_topology",tR="weight_specs",nR="weight_data",aR="model_metadata";function vS(e){return{info:[ol,e,xS].join(Nr),topology:[ol,e,eR].join(Nr),weightSpecs:[ol,e,tR].join(Nr),weightData:[ol,e,nR].join(Nr),modelMetadata:[ol,e,aR].join(Nr)}}function wS(e){for(let t of Object.values(e))window.localStorage.removeItem(t)}function rR(e){let t=e.split(Nr);if(t.length<3)throw new Error(`Invalid key format: ${e}`);return t.slice(1,t.length-1).join(Nr)}function sR(e){return e.startsWith(ai.URL_SCHEME)?e.slice(ai.URL_SCHEME.length):e}var ai=class{constructor(e){if(!H().getBool("IS_BROWSER")||typeof window=="undefined"||typeof window.localStorage=="undefined")throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,e==null||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=vS(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let t=JSON.stringify(e.modelTopology),n=JSON.stringify(e.weightSpecs),a=Fc(e);try{this.LS.setItem(this.keys.info,JSON.stringify(a)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,BD(e.weightData));let r={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:e.signature!=null?e.signature:void 0,userDefinedMetadata:e.userDefinedMetadata!=null?e.userDefinedMetadata:void 0,modelInitializer:e.modelInitializer!=null?e.modelInitializer:void 0,initializerSignature:e.initializerSignature!=null?e.initializerSignature:void 0,trainingConfig:e.trainingConfig!=null?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(r)),{modelArtifactsInfo:a}}catch(r){throw wS(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${a.modelTopologyBytes}, weightSpecsBytes=${a.weightSpecsBytes}, weightDataBytes=${a.weightDataBytes}.`)}}}async load(){let e=JSON.parse(this.LS.getItem(this.keys.info));if(e==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(e.modelTopologyType!=="JSON")throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let t={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(n==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=n;let a=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(a==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=a;let r=this.LS.getItem(this.keys.modelMetadata);if(r!=null){let i=JSON.parse(r);t.format=i.format,t.generatedBy=i.generatedBy,t.convertedBy=i.convertedBy,i.signature!=null&&(t.signature=i.signature),i.userDefinedMetadata!=null&&(t.userDefinedMetadata=i.userDefinedMetadata),i.modelInitializer!=null&&(t.modelInitializer=i.modelInitializer),i.initializerSignature!=null&&(t.initializerSignature=i.initializerSignature),i.trainingConfig!=null&&(t.trainingConfig=i.trainingConfig)}let s=this.LS.getItem(this.keys.weightData);if(s==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=VD(s),t}};ai.URL_SCHEME="localstorage://";var kS=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ai.URL_SCHEME)?iR(e.slice(ai.URL_SCHEME.length)):null;At.registerSaveRouter(kS);At.registerLoadRouter(kS);function iR(e){return new ai(e)}var oR=class{constructor(){$(H().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),$(typeof window=="undefined"||typeof window.localStorage!="undefined",()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let e={},t=ol+Nr,n=Nr+xS;for(let a=0;a<this.LS.length;++a){let r=this.LS.key(a);if(r.startsWith(t)&&r.endsWith(n)){let s=rR(r);e[s]=JSON.parse(this.LS.getItem(r))}}return e}async removeModel(e){e=sR(e);let t=vS(e);if(this.LS.getItem(t.info)==null)throw new Error(`Cannot find model at path '${e}'`);let n=JSON.parse(this.LS.getItem(t.info));return wS(t),n}},pl="://",Pn=class{constructor(){this.managers={}}static getInstance(){return Pn.instance==null&&(Pn.instance=new Pn),Pn.instance}static registerManager(e,t){$(e!=null,()=>"scheme must not be undefined or null."),e.endsWith(pl)&&(e=e.slice(0,e.indexOf(pl))),$(e.length>0,()=>"scheme must not be an empty string.");let n=Pn.getInstance();$(n.managers[e]==null,()=>`A model store manager is already registered for scheme '${e}'.`),n.managers[e]=t}static getManager(e){let t=Pn.getInstance().managers[e];if(t==null)throw new Error(`Cannot find model manager for scheme '${e}'`);return t}static getSchemes(){return Object.keys(Pn.getInstance().managers)}};function xh(e){if(e.indexOf(pl)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Pn.getSchemes().join(",")}`);return{scheme:e.split(pl)[0],path:e.split(pl)[1]}}async function IS(e,t,n=!1){$(e!==t,()=>`Old path and new path are the same: '${e}'`);let a=At.getLoadHandlers(e);$(a.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),$(a.length<2,()=>`Copying failed because more than one (${a.length}) load handlers for source URL ${e}.`);let r=a[0],s=At.getSaveHandlers(t);$(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),$(s.length<2,()=>`Copying failed because more than one (${a.length}) save handlers for destination URL ${t}.`);let i=s[0],o=xh(e).scheme,l=xh(e).path,u=o===xh(e).scheme,p=await r.load();n&&u&&await Pn.getManager(o).removeModel(l);let d=await i.save(p);return n&&!u&&await Pn.getManager(o).removeModel(l),d.modelArtifactsInfo}async function lR(){let e=Pn.getSchemes(),t={};for(let n of e){let a=await Pn.getManager(n).listModels();for(let r in a){let s=n+pl+r;t[s]=a[r]}}return t}async function uR(e){let t=xh(e);return Pn.getManager(t.scheme).removeModel(t.path)}async function pR(e,t){return IS(e,t,!1)}async function cR(e,t){return IS(e,t,!0)}var dR=class{constructor(){this.messageName="setTimeoutCustom",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Browser's encoder only supports utf-8, but got ${t}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}setTimeoutCustom(e,t){if(typeof window=="undefined"||!H().getBool("USE_SETTIMEOUTCUSTOM")){setTimeout(e,t);return}this.functionRefs.push(e),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},"*")},t),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener("message",n=>{if(n.source===window&&n.data.name===this.messageName){n.stopPropagation();let a=this.functionRefs[n.data.index];a(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0)}},!0))}};if(H().get("IS_BROWSER")){H().setPlatform("browser",new dR);try{Pn.registerManager(ai.URL_SCHEME,new oR)}catch(e){}try{Pn.registerManager(ni.URL_SCHEME,new QD)}catch(e){}}var hR={importFetch:()=>kF()},fb,mR=class{constructor(){this.util=IF(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return H().global.fetch!=null?H().global.fetch(e,t):(fb==null&&(fb=hR.importFetch()),fb(e,t))}now(){let e=process.hrtime();return e[0]*1e3+e[1]/1e6}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Node built-in encoder only supports utf-8, but got ${t}`);return this.textEncoder.encode(e)}decode(e,t){return e.length===0?"":new this.util.TextDecoder(t).decode(e)}};H().get("IS_NODE")&&!H().get("IS_BROWSER")&&H().setPlatform("node",new mR);function Oe(e,t="float32",n){return t=t||"float32",Ex(e),new Ht(e,t,n)}function fR(e,t){let n=_(e,"x","cast");if(!KI(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&n.dtype!=="string"||t!=="string"&&n.dtype==="string")throw new Error("Only strings can be casted to strings");let a={x:n},r={dtype:t};return O.runKernel(xi,a,r)}var oe=L({cast_:fR});function gR(e){let t={x:_(e,"x","clone","string_or_numeric")};return O.runKernel(Ri,t)}var ir=L({clone_:gR});function zx(e,t=!1){console.log(e.toString(t))}pS();var yR={buffer:Oe,cast:oe,clone:ir,print:zx};_D(yR);var Ut={};Ae(Ut,{browserFiles:()=>SR,browserHTTPRequest:()=>ER,concatenateArrayBuffers:()=>Ox,copyModel:()=>pR,decodeWeights:()=>hS,encodeWeights:()=>zD,fromMemory:()=>$R,fromMemorySync:()=>_S,getLoadHandlers:()=>YD,getModelArtifactsForJSON:()=>Lx,getModelArtifactsForJSONSync:()=>fS,getModelArtifactsInfoForJSON:()=>Fc,getSaveHandlers:()=>XD,getWeightSpecs:()=>gS,http:()=>Bx,isHTTPScheme:()=>zb,listModels:()=>lR,loadWeights:()=>TR,moveModel:()=>cR,registerLoadRouter:()=>KD,registerSaveRouter:()=>qD,removeModel:()=>uR,weightsLoaderFactory:()=>TS,withSaveHandler:()=>FR,withSaveHandlerSync:()=>DR});var bR="model",xR=".json",vR=".weights.bin";function fk(e){return new Promise(t=>setTimeout(t)).then(e)}var fl=class{constructor(e){if(!H().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(fl.URL_SCHEME)&&(e=e.slice(fl.URL_SCHEME.length)),(e==null||e.length===0)&&(e=bR),this.modelJsonFileName=e+xR,this.weightDataFileName=e+vR}async save(e){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");let t=window.URL.createObjectURL(new Blob([e.weightData],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let n=[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}],a=mS(e,n),r=window.URL.createObjectURL(new Blob([JSON.stringify(a)],{type:"application/json"})),s=this.modelJsonAnchor==null?document.createElement("a"):this.modelJsonAnchor;if(s.download=this.modelJsonFileName,s.href=r,await fk(()=>s.dispatchEvent(new MouseEvent("click"))),e.weightData!=null){let i=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;i.download=this.weightDataFileName,i.href=t,await fk(()=>i.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Fc(e)}}}};fl.URL_SCHEME="downloads://";var wR=class{constructor(e){if(e==null||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{let n=new FileReader;n.onload=a=>{let r=JSON.parse(a.target.result),s=r.modelTopology;if(s==null){t(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(r.weightsManifest==null){t(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){e({modelTopology:s});return}let i=Lx(r,o=>this.loadWeights(o));e(i)},n.onerror=a=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(e){let t=[],n=[];for(let s of e)t.push(...s.weights),n.push(...s.paths);let a=this.checkManifestAndWeightFiles(e),r=n.map(s=>this.loadWeightsFile(s,a[s]));return Promise.all(r).then(s=>[t,Ox(s)])}loadWeightsFile(e,t){return new Promise((n,a)=>{let r=new FileReader;r.onload=s=>{let i=s.target.result;n(i)},r.onerror=s=>a(`Failed to weights data from file of path '${e}'.`),r.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){let t=[],n=this.weightsFiles.map(r=>mk(r.name)),a={};for(let r of e)r.paths.forEach(s=>{let i=mk(s);if(t.indexOf(i)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${i}'`);if(t.push(i),n.indexOf(i)===-1)throw new Error(`Weight file with basename '${i}' is not provided.`);a[s]=this.weightsFiles[n.indexOf(i)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return a}},kR=e=>H().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(fl.URL_SCHEME)?IR(e.slice(fl.URL_SCHEME.length)):null;At.registerSaveRouter(kR);function IR(e="model"){return new fl(e)}function SR(e){return new wR(e)}function gk(e,t,n,a){i(e),n=n==null?0:n,a=a==null?1:a,o(n,a);let r=0,s=l=>(l.then(u=>{let p=n+ ++r/e.length*(a-n);return t(p),u}),l);function i(l){$(l!=null&&Array.isArray(l)&&l.length>0,()=>"promises must be a none empty array")}function o(l,u){$(l>=0&&l<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${l}`),$(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${u}`),$(u>=l,()=>`startFraction must be no more than endFraction, but got startFraction ${l} and endFraction ${u}`)}return Promise.all(e.map(s))}async function SS(e,t){t==null&&(t={});let n=t.fetchFunc==null?H().platform.fetch:t.fetchFunc,a=e.map(u=>n(u,t.requestInit,{isBinary:!0})),r=0,s=.5,i=(t.onProgress==null?await Promise.all(a):await gk(a,t.onProgress,r,s)).map(u=>u.arrayBuffer()),o=.5,l=1;return t.onProgress==null?await Promise.all(i):await gk(i,t.onProgress,o,l)}async function TR(e,t="",n,a){return TS(r=>SS(r,{requestInit:a}))(e,t,n)}function TS(e){return async(t,n="",a)=>{let r=t.map(()=>!1),s={},i=a!=null?a.map(()=>!1):[],o=[];if(t.forEach((h,m)=>{let f=0;h.weights.forEach(g=>{let y="quantization"in g?g.quantization.dtype:g.dtype,b=Mb[y]*mt(g.shape),x=()=>{r[m]=!0,s[m]==null&&(s[m]=[]),s[m].push({manifestEntry:g,groupOffset:f,sizeBytes:b})};a!=null?a.forEach((w,I)=>{w===g.name&&(x(),i[I]=!0)}):x(),o.push(g.name),f+=b})}),!i.every(h=>h)){let h=a.filter((m,f)=>!i[f]);throw new Error(`Could not find weights in manifest with names: ${h.join(", ")}. 
-Manifest JSON has weights with names: ${o.join(", ")}.`)}let l=r.reduce((h,m,f)=>(m&&h.push(f),h),[]),u=[];l.forEach(h=>{t[h].paths.forEach(m=>{let f=n+(n.endsWith("/")?"":"/")+m;u.push(f)})});let p=await e(u),d={},c=0;return l.forEach(h=>{let m=t[h].paths.length,f=0;for(let x=0;x<m;x++)f+=p[c+x].byteLength;let g=new ArrayBuffer(f),y=new Uint8Array(g),b=0;for(let x=0;x<m;x++){let w=new Uint8Array(p[c+x]);y.set(w,b),b+=w.byteLength}s[h].forEach(x=>{let w=g.slice(x.groupOffset,x.groupOffset+x.sizeBytes),I=hS(w,[x.manifestEntry]);for(let T in I)d[T]=I[T]}),c+=m}),d}}var NR="application/octet-stream",CR="application/json",Wx=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",t==null&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,this.weightUrlConverter=t.weightUrlConverter,t.fetchFunc!=null?($(typeof t.fetchFunc=="function",()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=t.fetchFunc):this.fetch=H().platform.fetch,$(e!=null&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&$(e.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,t.requestInit!=null&&t.requestInit.body!=null)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{}}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;let n=[{paths:["./model.weights.bin"],weights:e.weightSpecs}],a=mS(e,n);t.body.append("model.json",new Blob([JSON.stringify(a)],{type:CR}),"model.json"),e.weightData!=null&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:NR}),"model.weights.bin");let r=await this.fetch(this.path,t);if(r.ok)return{modelArtifactsInfo:Fc(e),responses:[r]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${r.status}.`)}async load(){let e=await this.fetch(this.path,this.requestInit);if(!e.ok)throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);let t;try{t=await e.json()}catch(r){let s=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?s+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":s+=" Please make sure the server is serving valid JSON for this request.",new Error(s)}let n=t.modelTopology,a=t.weightsManifest;if(n==null&&a==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return Lx(t,r=>this.loadWeights(r))}async loadWeights(e){let t=Array.isArray(this.path)?this.path[1]:this.path,[n,a]=_R(t),r=this.weightPathPrefix||n,s=gS(e),i=[],o=[];for(let u of e)for(let p of u.paths)this.weightUrlConverter!=null?o.push(this.weightUrlConverter(p)):i.push(r+p+a);this.weightUrlConverter&&i.push(...await Promise.all(o));let l=await SS(i,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[s,Ox(l)]}};Wx.URL_SCHEME_REGEX=/^https?:\/\//;function _R(e){let t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),a=e.substring(0,t),r=n>t?e.substring(n):"";return[a+"/",r]}function zb(e){return e.match(Wx.URL_SCHEME_REGEX)!=null}var NS=(e,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let n=!0;if(Array.isArray(e)?n=e.every(a=>zb(a)):n=zb(e),n)return Bx(e,t)}return null};At.registerSaveRouter(NS);At.registerLoadRouter(NS);function Bx(e,t){return new Wx(e,t)}function ER(e,t){return Bx(e,t)}var gb=class{constructor(e){this.modelArtifacts=e}load(){return this.modelArtifacts}},CS=class{constructor(e){this.saveHandler=e}save(e){return this.saveHandler(e)}},AR=class{constructor(e){e.load&&(this.load=()=>Promise.resolve(e.load())),e.save&&(this.save=t=>Promise.resolve(e.save(t)))}};function $R(e,t,n,a){let r=arguments;return new AR(_S(...r))}function _S(e,t,n,a){return arguments.length===1?e.modelTopology!=null||e.weightSpecs!=null?new gb(e):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new gb({modelTopology:e})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new gb({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:a}))}function FR(e){return new CS(e)}function DR(e){return new CS(e)}var ES={};Ae(ES,{confusionMatrix:()=>eM});function RR(e,t,n=!1,a=!1){let r=_(e,"a","matMul"),s=_(t,"b","matMul");[r,s]=_t(r,s);let i={a:r,b:s},o={transposeA:n,transposeB:a};return O.runKernel(bi,i,o)}var Fe=L({matMul_:RR});function MR(e,t,n=1,a=0,r="int32"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let s={indices:_(e,"indices","oneHot","int32")},i={dtype:r,depth:t,onValue:n,offValue:a};return O.runKernel(Hi,s,i)}var gl=L({oneHot_:MR});function PR(){H().set("PROD",!0)}function OR(){H().set("DEBUG",!0)}function LR(){H().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function Vx(e){H().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}ED(Vx);function zR(){O.disposeVariables()}function _a(){return O}function Oh(){return O.memory()}function WR(e){return O.profile(e)}function P(e,t){return O.tidy(e,t)}function _e(e){Rx(e).forEach(t=>t.dispose())}function Jt(e){return O.keep(e)}function BR(e){return O.time(e)}function VR(e){return O.setBackend(e)}function UR(){return O.ready()}function GR(){return O.backendName}function HR(e){O.removeBackend(e)}function jR(e){return O.findBackend(e)}function qR(e){return O.findBackendFactory(e)}function Gm(e,t,n=1){return O.registerBackend(e,t,n)}function AS(){return O.backend}function KR(e,t){H().setPlatform(e,t)}function XR(e){let t={input:_(e,"input","imag")};return O.runKernel(Nm,t)}var Dc=L({imag_:XR});function YR(e){let t={x:_(e,"x","neg")};return O.runKernel(du,t)}var vt=L({neg_:YR});function ZR(e){let t={input:_(e,"input","real")};return O.runKernel(Pm,t)}var yl=L({real_:ZR});function JR(e,t,n){let a=_(e,"x","transpose");if(t==null&&(t=a.shape.map((i,o)=>o).reverse()),$(a.rank===t.length,()=>`Error in transpose: rank of input ${a.rank} must match length of perm ${t}.`),t.forEach(i=>{$(i>=0&&i<a.rank,()=>`All entries in 'perm' must be between 0 and ${a.rank-1} but got ${t}`)}),a.rank<=1)return a.clone();let r={x:a},s={perm:t};return a.dtype==="complex64"?P(()=>{let i=yl(a),o=Dc(a);return i=O.runKernel(_r,{x:i},s),o=O.runKernel(_r,{x:o},s),n&&(o=vt(o)),Ar(i,o)}):O.runKernel(_r,r,s)}var Ee=L({transpose_:JR});function QR(e,t,n){let a=_(e,"labels","confusionMatrix"),r=_(t,"predictions","confusionMatrix");$(n==null||n>0&&Number.isInteger(n),()=>`If provided, numClasses must be a positive integer, but got ${n}`),$(a.rank===1,()=>`Expected the rank of labels to be 1, but got ${a.rank}`),$(r.rank===1,()=>`Expected the rank of predictions to be 1, but got ${r.rank}`),$(a.shape[0]===r.shape[0],()=>`Mismatch in the number of examples: ${a.shape[0]} vs. ${r.shape[0]}. Labels and predictions should have the same number of elements.`),$(n>0&&Number.isInteger(n),()=>`numClasses is required to be a positive integer, but got ${n}`);let s=gl(oe(a,"int32"),n),i=gl(oe(r,"int32"),n),o=Ee(s),l=Fe(o,i);return oe(l,"int32")}var eM=L({confusionMatrix_:QR}),Ou={};Ae(Ou,{assertAndGetBroadcastShape:()=>ut,getBroadcastDims:()=>$S,getReductionAxes:()=>Wt});function $S(e,t){let n=e.length,a=[];for(let r=0;r<n;r++){let s=n-1-r,i=e[s]||1;(t[t.length-1-r]||1)>1&&i===1&&a.unshift(s)}return a}function Wt(e,t){let n=[];for(let a=0;a<t.length;a++){let r=e[e.length-a-1],s=t.length-a-1,i=t[s];(r==null||r===1&&i>1)&&n.unshift(s)}return n}function ut(e,t){let n=[],a=Math.max(e.length,t.length);for(let r=0;r<a;r++){let s=e[e.length-r-1];s==null&&(s=1);let i=t[t.length-r-1];if(i==null&&(i=1),s===1)n.unshift(i);else if(i===1)n.unshift(s);else if(s!==i){let o=`Operands could not be broadcast together with shapes ${e} and ${t}.`;throw Error(o)}else n.unshift(s)}return n}var ho={};Ae(ho,{fromPixels:()=>oM,fromPixelsAsync:()=>sM,toPixels:()=>iM});function Rc(e,t,n){if(mi(e),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");let a=lr(e,n);if(a.length!==3&&a.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return vs(e,t,a,n)}var Os;function FS(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(e==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let n=!1,a=!1,r=!1,s=!1,i=!1,o=!1;if(e.data instanceof Uint8Array)n=!0;else if(typeof ImageData!="undefined"&&e instanceof ImageData)a=!0;else if(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)r=!0;else if(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)s=!0;else if(e.getContext!=null)i=!0;else if(typeof ImageBitmap!="undefined"&&e instanceof ImageBitmap)o=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${e.constructor.name}`);if(Dh(Fh,O.backendName)!=null){let c={pixels:e},h={numChannels:t};return O.runKernel(Fh,c,h)}let[l,u]=r?[e.videoWidth,e.videoHeight]:[e.width,e.height],p;if(i)p=e.getContext("2d").getImageData(0,0,l,u).data;else if(a||n)p=e.data;else if(s||r||o){if(Os==null)if(typeof document=="undefined")if(typeof OffscreenCanvas!="undefined"&&typeof OffscreenCanvasRenderingContext2D!="undefined")Os=new OffscreenCanvas(1,1).getContext("2d");else throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");else Os=document.createElement("canvas").getContext("2d",{willReadFrequently:!0});Os.canvas.width=l,Os.canvas.height=u,Os.drawImage(e,0,0,l,u),p=Os.getImageData(0,0,l,u).data}let d;if(t===4)d=new Int32Array(p);else{let c=l*u;d=new Int32Array(c*t);for(let h=0;h<c;h++)for(let m=0;m<t;++m)d[h*t+m]=p[h*4+m]}return Rc(d,[u,l,t],"int32")}function tM(e){return e!=null&&e.data instanceof Uint8Array}function nM(){return typeof window!="undefined"&&typeof ImageBitmap!="undefined"&&window.hasOwnProperty("createImageBitmap")}function aM(e){return e!=null&&e.width!==0&&e.height!==0}function rM(e){return nM()&&!(e instanceof ImageBitmap)&&aM(e)&&!tM(e)}async function sM(e,t=3){let n=null;if(H().getBool("WRAP_TO_IMAGEBITMAP")&&rM(e)){let a;try{a=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(r){a=null}a!=null&&a.width===e.width&&a.height===e.height?n=a:n=e}else n=e;return FS(n,t)}async function iM(e,t){let n=_(e,"img","toPixels");if(!(e instanceof Te)){let u=n;n=oe(u,"int32"),u.dispose()}if(n.rank!==2&&n.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);let[a,r]=n.shape.slice(0,2),s=n.rank===2?1:n.shape[2];if(s>4||s===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${s}`);if(n.dtype!=="float32"&&n.dtype!=="int32")throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);let i=await n.data(),o=n.dtype==="float32"?255:1,l=new Uint8ClampedArray(r*a*4);for(let u=0;u<a*r;++u){let p=[0,0,0,255];for(let c=0;c<s;c++){let h=i[u*s+c];if(n.dtype==="float32"){if(h<0||h>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${h}.`)}else if(n.dtype==="int32"&&(h<0||h>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${h}.`);s===1?(p[0]=h*o,p[1]=h*o,p[2]=h*o):p[c]=h*o}let d=u*4;l[d+0]=Math.round(p[0]),l[d+1]=Math.round(p[1]),l[d+2]=Math.round(p[2]),l[d+3]=Math.round(p[3])}if(t!=null){t.width=r,t.height=a;let u=t.getContext("2d"),p=new ImageData(l,r,a);u.putImageData(p,0,0)}return n!==e&&n.dispose(),l}var oM=L({fromPixels_:FS}),Ux={};Ae(Ux,{prepareAndValidate:()=>DS});function DS(e,t){let n=e.shape.length,a=t.shape.length;if(n<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);if(a<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${a}.`);if(t.dtype!=="int32")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[a-1]>n)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[a-1]} vs. ${n}`);if(mt(e.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);let r=t.shape,s=r[r.length-1],i=1;for(let d=0;d<r.length-1;++d)i*=r[d];let o=e.shape,l=r.slice();l.pop();let u=1;for(let d=s;d<n;++d)u*=o[d],l.push(o[d]);let p=[...Fl(e.shape).map(d=>d/u),1].slice(0,s);return[l,i,u,p]}var Gx={};Ae(Gx,{calculateShapes:()=>RS,validateInput:()=>jx,validateUpdateShape:()=>Hx});function Hx(e,t,n){let a=t.rank>1?t.shape[t.rank-1]:1,r=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${a}, and batchDim: ${r}.`;if(n.rank<r)throw new Error(s+` update.rank < ${r}. `);if(e.length<a+(n.rank-r))throw new Error(s+` Output shape length < ${a+(n.rank-r)}`);if(n.rank!==r+e.length-a)throw new Error(s+` update.rank != ${r+e.length-a}`);for(let i=0;i<r;++i)if(n.shape[i]!==t.shape[i])throw new Error(s+` updates.shape[${i}] (${n.shape[i]}) != indices.shape[${i}] (${t.shape[i]}).`);for(let i=0;i<n.rank-r;++i)if(n.shape[i+r]!==e[i+a])throw new Error(s+` updates.shape[${i+r}] (${n.shape[i+r]}) != shape[${i+r}] (${e[i+r]})`)}function jx(e,t,n){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(e.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);if(t.dtype!=="int32")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(n.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${n}`);if(n.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(e.size===0)throw new Error(`Updates specified for empty output. updates shape: ${e.shape}`)}Hx(n,t,e)}function RS(e,t,n){let a=t.shape.length,r=a>1?t.shape[a-1]:1,s=n.length,i=1;for(let d=r;d<s;++d)i*=n[d];let o=r<1?1:r,l=mt(t.shape)/o,u=[...Fl(n.slice(0,r)),1],p=mt(n);return{sliceRank:r,numUpdates:l,sliceSize:i,strides:u,outputSize:p}}var jt={};Ae(jt,{assertParamsValid:()=>uM,computeFlatOffset:()=>mM,computeOutShape:()=>cM,getNormalizedAxes:()=>dM,isSliceContinous:()=>hM,maskToAxes:()=>pM,parseSliceParams:()=>US,sliceInfo:()=>fM,startForAxis:()=>BS,startIndicesWithElidedDims:()=>LS,stopForAxis:()=>VS,stopIndicesWithElidedDims:()=>zS,stridesForAxis:()=>WS,stridesWithElidedDims:()=>MS});var Wb=-2,lM=-1;function uM(e,t,n){let a=e.shape.length;$(a===t.length,()=>`Error in slice${a}D: Length of begin ${t} must match the rank of the array (${a}).`),$(a===n.length,()=>`Error in slice${a}D: Length of size ${n} must match the rank of the array (${a}).`);for(let r=0;r<a;++r)$(t[r]+n[r]<=e.shape[r],()=>`Error in slice${a}D: begin[${r}] + size[${r}] (${t[r]+n[r]}) would overflow input.shape[${r}] (${e.shape[r]})`)}function pM(e){let t=[],n=0;for(;e>0;)e&1&&t.push(n),e/=2,n++;return t}function cM(e,t,n){let a=[];for(let r=0;r<e.length;r++)a[r]=Math.ceil((t[r]-e[r])/n[r]);return a}function MS(e,t,n,a){let r=[...e];for(let s=r.length;s<a.length;s++)r.push(1);for(let s=0;s<n;s++)s===0?r[t]=1:(r.splice(t,0,1),r.pop());return r}function PS(e,t,n){return n<=e?n:n-(t-1)}function OS(e,t){let n=[];for(let a=0;a<e;a++)n.push(t+a);return n}function dM(e,t,n,a,r,s,i,o,l){let u=e.length,p=new Array(u),d=new Array(u),c=new Array(u);if(t.length&&n>0){let h=t[0],m=n+1;p=LS(i,h,m,a,e),d=zS(o,h,m,r,e),c=MS(s,h,m,e)}else for(let h=0;h<u;h++)p[h]=BS(i,a,s,e,h,l),d[h]=VS(o,r,s,e,h,l),c[h]=WS(s,h,l);return{begin:p,end:d,strides:c}}function LS(e,t,n,a,r){let s=[...r],i=OS(n,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=0;else{let l=PS(t,n,o),u=a[l];e&1<<l&&(u=0),s[o]=u}return s}function zS(e,t,n,a,r){let s=[...r],i=OS(n,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=Number.MAX_SAFE_INTEGER;else{let l=PS(t,n,o),u=a[l];e&1<<l&&(u=Number.MAX_SAFE_INTEGER),s[o]=u}for(let o=0;o<s.length;o++){let l=r[o];s[o]<0&&(s[o]+=l),s[o]=Kp(0,s[o],r[o])}return s}function WS(e,t,n){let a=e[t];return(n&1<<t||a==null)&&(a=1),a}function BS(e,t,n,a,r,s){let i=t[r],o=n[r]||1;(e&1<<r||s&1<<r||i==null)&&(o>0?i=Number.MIN_SAFE_INTEGER:i=Number.MAX_SAFE_INTEGER);let l=a[r];return i<0&&(i+=l),i=Kp(0,i,l-1),i}function VS(e,t,n,a,r,s){let i=t[r],o=n[r]||1;(e&1<<r||s&1<<r||i==null)&&(o>0?i=Number.MAX_SAFE_INTEGER:i=Number.MIN_SAFE_INTEGER);let l=a[r];return i<0&&(i+=l),o>0?i=Kp(0,i,l):i=Kp(-1,i,l-1),i}function hM(e,t,n){let a=n.length;for(let r=0;r<n.length;r++)if(n[r]>1){a=r;break}for(let r=a+1;r<n.length;r++)if(t[r]>0||n[r]!==e[r])return!1;return!0}function mM(e,t){let n=e.length>0?e[e.length-1]:1;for(let a=0;a<e.length-1;a++)n+=e[a]*t[a];return n}function US(e,t,n){let a,r=e.shape.length;typeof t=="number"?a=[t,...new Array(r-1).fill(0)]:t.length<r?a=t.concat(new Array(r-t.length).fill(0)):a=t.slice(),a.forEach(i=>{$(i!==-1,()=>"slice() does not support negative begin indexing.")});let s;return n==null?s=new Array(r).fill(-1):typeof n=="number"?s=[n,...new Array(r-1).fill(-1)]:n.length<r?s=n.concat(new Array(r-n.length).fill(-1)):s=n,s=s.map((i,o)=>i>=0?i:($(i===-1,()=>`Negative size values should be exactly -1 but got ${i} for the slice() size at index ${o}.`),e.shape[o]-a[o])),[a,s]}function fM(e,t,n,a,r,s,i,o,l){let u;if(a==null?(u=new Array(t.length),u.fill(1)):u=a,i!=null&&(i&i-1)!==0)throw new Error("Multiple ellipses in slice is not allowed.");let p=!1,d={dims:u.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:n.slice(),strides:u.slice(),beginMask:r,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};for(let b=0;b<d.dims;b++)p&&(1<<b&o)!==0&&d.numAddAxisAfterEllipsis++,1<<b&i&&(p=!0);p||(d.ellipsisMask|=1<<d.dims,d.dims++);let c={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};gM(d,c);let h=!0,m=!0,f=!0,g=[],y=[];for(let b=0;b<e.length;++b){if(c.strides[b]===0)throw Error(`strides[${b}] must be non-zero`);let x=!!(c.shrinkAxisMask&1<<b),w=e[b];if(w===-1){g.push(x?1:-1);continue}let I=[c.beginMask&1<<b,c.endMask&1<<b],T=[c.strides[b]>0?0:-1,c.strides[b]>0?w:w-1];if(x&&c.strides[b]<=0)throw Error("only stride 1 allowed on non-range indexing.");f=f&&c.strides[b]===1;let C=!!(c.beginMask&1<<b&&c.endMask&1<<b);if(c.beginValid&&c.endValid){if(x){let F=c.begin[b]<0?w+c.begin[b]:c.begin[b];if(c.begin[b]=F,c.end[b]=c.begin[b]+1,F<0||F>=w)throw Error(`slice index ${c.begin[b]} of dimension ${b} out of bounds.`)}else c.begin[b]=yk(c.begin[b],0,c.strides[b],w,I,T),c.end[b]=yk(c.end[b],1,c.strides[b],w,I,T);let R=c.strides[b]===1&&c.begin[b]===0&&c.end[b]===w;h=h&&R,m=m&&(b===0&&c.strides[b]===1||R)}else h=h&&c.strides[b]===1&&C,m=m&&(b===0&&c.strides[b]===1||C);let E,A=!1;if(c.beginValid&&c.endValid?(E=c.end[b]-c.begin[b],A=!0):x?(E=1,A=!0):C&&w>=0&&(c.strides[b]<0?E=-w:E=w,A=!0),A){let R;E===0||E<0!=c.strides[b]<0?R=0:R=Math.trunc(E/c.strides[b])+(E%c.strides[b]!==0?1:0),g.push(R)}else g.push(-1)}for(let b=0;b<c.finalShapeGatherIndices.length;++b){let x=c.finalShapeGatherIndices[b];x>=0?y.push(g[x]):x===Wb&&y.push(1)}return{finalShapeSparse:y.filter((b,x)=>c.finalShapeGatherIndices[x]!==Wb),finalShape:y,isIdentity:h,sliceDim0:m,isSimpleSlice:f,begin:c.begin,end:c.end,strides:c.strides}}function gM(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let n=0;t.beginValid=e.begin!=null,t.endValid=e.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let a=0;a<e.dims;a++)if(1<<a&e.ellipsisMask){let r=Math.min(t.dims-(e.dims-a)+1+e.numAddAxisAfterEllipsis,t.dims);for(;n<r;n++)t.begin[n]=0,t.end[n]=0,t.strides[n]=1,t.beginMask|=1<<n,t.endMask|=1<<n,t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[n]=a}else if(1<<a&e.newAxisMask)t.finalShapeGatherIndices.push(Wb),t.finalShapeGatherIndicesSparse.push(-1);else{if(n===t.begin.length)throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);e.begin!=null&&(t.begin[n]=e.begin[a]),e.end!=null&&(t.end[n]=e.end[a]),t.strides[n]=e.strides[a],e.beginMask&1<<a&&(t.beginMask|=1<<n),e.endMask&1<<a&&(t.endMask|=1<<n),e.shrinkAxisMask&1<<a?(t.finalShapeGatherIndices.push(lM),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<n):(t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(a)),t.inputShapeGatherIndicesSparse[n]=a,n++}}function yk(e,t,n,a,r,s){if(r[t])return n>0?s[t]:s[t+1&1];{let i=e<0?a+e:e;return i<s[0]?s[0]:i>s[1]?s[1]:i}}var ne={};Ae(ne,{Serializable:()=>GS,SerializationMap:()=>Vs,registerClass:()=>ws});var GS=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},Vs=class{constructor(){this.classNameMap={}}static getMap(){return Vs.instance==null&&(Vs.instance=new Vs),Vs.instance}static register(e){Vs.getMap().classNameMap[e.className]=[e,e.fromConfig]}};function ws(e){$(e.className!=null,()=>"Class being registered does not have the static className property defined."),$(typeof e.className=="string",()=>"className is required to be a string, but got type "+typeof e.className),$(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),Vs.register(e)}var HS={};Ae(HS,{TEST_EPSILON_FLOAT16:()=>jS,createVideoElement:()=>SM,encodeStrings:()=>qS,expectArrayBuffersEqual:()=>IM,expectArraysClose:()=>bM,expectArraysEqual:()=>vM,expectNumbersClose:()=>wM,expectPromiseToFail:()=>xM,expectValuesInRange:()=>kM,play:()=>TM,testEpsilon:()=>qx});var yM=.001,jS=.1;function bM(e,t,n){return n==null&&(n=qx()),Bb(e,t,(a,r)=>Kx(a,r,n))}function qx(){return O.backend.floatPrecision()===32?yM:jS}function Bb(e,t,n){let a=!0;if((hn(e)||hn(t))&&(a=!1),hn(e)&&hn(t)&&(a=!0),a){let i=e.constructor.name,o=t.constructor.name;if(i!==o)throw new Error(`Arrays are of different type. Actual: ${i}. Expected: ${o}`)}if(Array.isArray(e)&&Array.isArray(t)){let i=lr(e),o=lr(t);if(!fs(i,o))throw new Error(`Arrays have different shapes. Actual: [${i}]. Expected: [${o}]`)}let r=hn(e)?e:Js(e),s=hn(t)?t:Js(t);if(r.length!==s.length)throw new Error(`Arrays have different lengths actual: ${r.length} vs expected: ${s.length}.
-Actual:   ${r}.
-Expected: ${s}.`);for(let i=0;i<s.length;++i){let o=r[i],l=s[i];if(!n(o,l))throw new Error(`Arrays differ: actual[${i}] = ${o}, expected[${i}] = ${l}.
-Actual:   ${r}.
-Expected: ${s}.`)}typeof expect!="undefined"&&expect().nothing()}function xM(e,t){e().then(()=>t.fail(),()=>t()),typeof expect!="undefined"&&expect().nothing()}function vM(e,t){let n=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return Qr(e)||Qr(e[0])||Qr(t)||Qr(t[0])?Bb(e,n,(a,r)=>a==r):Bb(e,t,(a,r)=>Kx(a,r,0))}function wM(e,t,n){if(n==null&&(n=qx()),!Kx(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`);typeof expect!="undefined"&&expect().nothing()}function Kx(e,t,n){return!isFinite(e)&&!isFinite(t)?!0:!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function kM(e,t,n){for(let a=0;a<e.length;a++)if(e[a]<t||e[a]>n)throw new Error(`Value out of range:${e[a]} low: ${t}, high: ${n}`)}function IM(e,t){let n=new Float32Array(e),a=new Float32Array(t);if(n.length!==a.length)throw new Error(`Expected ArrayBuffer to be of length ${a.length}, but it was ${n.length}`);for(let r=0;r<a.length;r++)if(n[r]!==a[r])throw new Error(`Expected ArrayBuffer value at ${r} to be ${a[r]} but got ${n[r]} instead`)}function qS(e){for(let t=0;t<e.length;t++){let n=e[t];Array.isArray(n)?qS(n):e[t]=Ac(n)}return e}function SM(e){let t=document.createElement("video");return"playsInline"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position="fixed",t.style.left="0px",t.style.top="0px",t.preload="auto",t.appendChild(e),new Promise(n=>{t.addEventListener("loadeddata",a=>n(t)),t.load()})}async function TM(e){await e.play(),"requestVideoFrameCallback"in e&&await new Promise(t=>{e.requestVideoFrameCallback(t)})}var NM="4.0.0";function CM(e,t){let n=_(e,"a","add"),a=_(t,"b","add");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(gs,r)}var Y=L({add_:CM});function _M(e,t){let n=_(e,"a","floorDiv"),a=_(t,"b","floorDiv");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel($i,r)}var Hm=L({floorDiv_:_M});function EM(e,t){let n=_(e,"a","div"),a=_(t,"b","div");if([n,a]=_t(n,a),n.dtype==="int32"&&a.dtype==="int32")return Hm(n,a);let r={a:n,b:a},s={};return O.runKernel(Ci,r,s)}var he=L({div_:EM});function AM(e,t){let n=_(e,"a","mul"),a=_(t,"b","mul");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Gi,r)}var z=L({mul_:AM});function $M(e){let t=_(e,"x","abs");if(t.dtype==="complex64"){let n={x:t};return O.runKernel(hc,n)}else{let n={x:t};return O.runKernel(Dl,n)}}var Lt=L({abs_:$M});function FM(e){let t={x:_(e,"x","acos")};return O.runKernel(Rl,t)}var Xx=L({acos_:FM});function DM(e){let t={x:_(e,"x","acosh")};return O.runKernel(Ml,t)}var Yx=L({acosh_:DM});function RM(e){$(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),$(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);let t=e.map((r,s)=>_(r,`tensors${s}`,"addN")),n=t[0];t.forEach(r=>{if(r.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(r=>{if(!fs(r.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let a=t;return O.runKernel(fi,a)}var KS=L({addN_:RM});function MM(e,t=null,n=!1){let a={x:_(e,"x","all","bool")},r={axis:t,keepDims:n};return O.runKernel(Pl,a,r)}var jm=L({all_:MM});function PM(e,t=null,n=!1){let a={x:_(e,"x","any","bool")},r={axis:t,keepDims:n};return O.runKernel(Ol,a,r)}var Qp=L({any_:PM});function OM(e,t=0){let n={x:_(e,"x","argMax")},a={axis:t};return O.runKernel(gi,n,a)}var ri=L({argMax_:OM});function LM(e,t=0){let n={x:_(e,"x","argMin")},a={axis:t};return O.runKernel(cc,n,a)}var Zx=L({argMin_:LM});function zM(e){let t={x:_(e,"x","asin")};return O.runKernel(Ll,t)}var Jx=L({asin_:zM});function WM(e){let t={x:_(e,"x","asinh")};return O.runKernel(zl,t)}var Qx=L({asinh_:WM});function BM(e){let t={x:_(e,"x","atan")};return O.runKernel(Wl,t)}var ev=L({atan_:BM});function VM(e,t){let n=_(e,"a","atan2"),a=_(t,"b","atan2");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(Vl,r)}var tv=L({atan2_:VM});function UM(e){let t={x:_(e,"x","atanh")};return O.runKernel(Bl,t)}var nv=L({atanh_:UM});function GM(e,t,n,a,r="NHWC",s){let i=e[3],o=[...t,i],l=ZS(r);return Mc(e,o,n,s,a,null,null,l)}function XS(e,t,n,a,r,s,i="channelsLast"){let[o,l]=Lh(t),u;if(i==="channelsLast")u=[o,l,e[3],e[3]];else if(i==="channelsFirst")u=[o,l,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return Mc(e,u,n,a,r,s,!1,i)}function HM(e,t,n,a,r,s,i="NDHWC"){let[o,l,u]=Vb(t),p,d;if(i==="NDHWC")d="channelsLast",p=[o,l,u,e[4],e[4]];else if(i==="NCDHW")d="channelsFirst",p=[o,l,u,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return YS(e,p,n,a,r,!1,d,s)}function Mc(e,t,n,a,r,s,i=!1,o="channelsLast"){let[l,u,p,d]=[-1,-1,-1,-1];if(o==="channelsLast")[l,u,p,d]=e;else if(o==="channelsFirst")[l,d,u,p]=e;else throw new Error(`Unknown dataFormat ${o}`);let[c,h,,m]=t,[f,g]=Lh(n),[y,b]=Lh(a),x=cl(c,y),w=cl(h,b),{padInfo:I,outHeight:T,outWidth:C}=KM(r,u,p,f,g,x,w,s,o),E=i?m*d:m,A;return o==="channelsFirst"?A=[l,E,T,C]:o==="channelsLast"&&(A=[l,T,C,E]),{batchSize:l,dataFormat:o,inHeight:u,inWidth:p,inChannels:d,outHeight:T,outWidth:C,outChannels:E,padInfo:I,strideHeight:f,strideWidth:g,filterHeight:c,filterWidth:h,effectiveFilterHeight:x,effectiveFilterWidth:w,dilationHeight:y,dilationWidth:b,inShape:e,outShape:A,filterShape:t}}function YS(e,t,n,a,r,s=!1,i="channelsLast",o){let[l,u,p,d,c]=[-1,-1,-1,-1,-1];if(i==="channelsLast")[l,u,p,d,c]=e;else if(i==="channelsFirst")[l,c,u,p,d]=e;else throw new Error(`Unknown dataFormat ${i}`);let[h,m,f,,g]=t,[y,b,x]=Vb(n),[w,I,T]=Vb(a),C=cl(h,w),E=cl(m,I),A=cl(f,T),{padInfo:R,outDepth:F,outHeight:S,outWidth:M}=XM(r,u,p,d,y,b,x,C,E,A,o),B=s?g*c:g,U;return i==="channelsFirst"?U=[l,B,F,S,M]:i==="channelsLast"&&(U=[l,F,S,M,B]),{batchSize:l,dataFormat:i,inDepth:u,inHeight:p,inWidth:d,inChannels:c,outDepth:F,outHeight:S,outWidth:M,outChannels:B,padInfo:R,strideDepth:y,strideHeight:b,strideWidth:x,filterDepth:h,filterHeight:m,filterWidth:f,effectiveFilterDepth:C,effectiveFilterHeight:E,effectiveFilterWidth:A,dilationDepth:w,dilationHeight:I,dilationWidth:T,inShape:e,outShape:U,filterShape:t}}function jM(e,t,n,a,r){a==null&&(a=av(e,t,n));let s=e[0],i=e[1],o=Ks((s-t+2*a)/n+1,r),l=Ks((i-t+2*a)/n+1,r);return[o,l]}function qM(e,t,n,a,r,s){r==null&&(r=av(e,t,a));let i=e[0],o=e[1],l=e[2],u=Ks((i-t+2*r)/a+1,s),p=Ks((o-t+2*r)/a+1,s),d=Ks((l-t+2*r)/a+1,s);return[u,p,d,n]}function av(e,t,n,a=1){let r=cl(t,a);return Math.floor((e[0]*(n-1)-n+r)/2)}function Lh(e){return typeof e=="number"?[e,e,e]:e.length===2?[e[0],e[1],1]:e}function Vb(e){return typeof e=="number"?[e,e,e]:e}function cl(e,t){return t<=1?e:e+(e-1)*(t-1)}function KM(e,t,n,a,r,s,i,o,l){let u,p,d;if(typeof e=="number"){u={top:e,bottom:e,left:e,right:e,type:e===0?"VALID":"NUMBER"};let c=jM([t,n],s,a,e,o);p=c[0],d=c[1]}else if(e==="same"){p=Math.ceil(t/a),d=Math.ceil(n/r);let c=Math.max(0,(p-1)*a+s-t),h=Math.max(0,(d-1)*r+i-n),m=Math.floor(c/2),f=c-m,g=Math.floor(h/2),y=h-g;u={top:m,bottom:f,left:g,right:y,type:"SAME"}}else if(e==="valid")u={top:0,bottom:0,left:0,right:0,type:"VALID"},p=Math.ceil((t-s+1)/a),d=Math.ceil((n-i+1)/r);else if(typeof e=="object"){let c=l==="channelsLast"?e[1][0]:e[2][0],h=l==="channelsLast"?e[1][1]:e[2][1],m=l==="channelsLast"?e[2][0]:e[3][0],f=l==="channelsLast"?e[2][1]:e[3][1];u={top:c,bottom:h,left:m,right:f,type:c===0&&h===0&&m===0&&f===0?"VALID":"EXPLICIT"},p=Ks((t-s+c+h)/a+1,o),d=Ks((n-i+m+f)/r+1,o)}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:u,outHeight:p,outWidth:d}}function XM(e,t,n,a,r,s,i,o,l,u,p){let d,c,h,m;if(typeof e=="number"){d={top:e,bottom:e,left:e,right:e,front:e,back:e,type:e===0?"VALID":"NUMBER"};let f=qM([t,n,a,1],o,1,r,e,p);c=f[0],h=f[1],m=f[2]}else if(e==="same"){c=Math.ceil(t/r),h=Math.ceil(n/s),m=Math.ceil(a/i);let f=(c-1)*r+o-t,g=(h-1)*s+l-n,y=(m-1)*i+u-a,b=Math.floor(f/2),x=f-b,w=Math.floor(g/2),I=g-w,T=Math.floor(y/2),C=y-T;d={top:w,bottom:I,left:T,right:C,front:b,back:x,type:"SAME"}}else if(e==="valid")d={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},c=Math.ceil((t-o+1)/r),h=Math.ceil((n-l+1)/s),m=Math.ceil((a-u+1)/i);else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:d,outDepth:c,outHeight:h,outWidth:m}}function Ks(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error(`Unknown roundingMode ${t}`)}}function os(e){let[t,n,a]=Lh(e);return t===1&&n===1&&a===1}function cr(e,t){return os(e)||os(t)}function ZS(e){if(e==="NHWC")return"channelsLast";if(e==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function Nn(e,t,n){if(n!=null){if(typeof t=="string")throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);if(typeof t=="number")$(hl(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);else if(typeof t=="object")t.forEach(a=>{a.forEach(r=>{$(hl(r),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${r}.`)})});else throw Error(`Error in ${e}: Unknown padding parameter: ${t}`)}}function YM(e,t){let n={x:_(e,"x","reshape","string_or_numeric")},a={shape:t};return O.runKernel(vu,n,a)}var W=L({reshape_:YM});function ZM(e,t,n,a,r){let s=_(e,"x","avgPool","float32"),i=1;$(cr(n,i),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${i}'`);let o=s,l=!1;s.rank===3&&(l=!0,o=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(o.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${o.rank}.`),Nn("avgPool",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r},d=O.runKernel(yi,u,p);return d=oe(d,s.dtype),l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var ya=L({avgPool_:ZM});function JM(e,t,n,a,r,s="NDHWC"){let i=_(e,"x","avgPool3d","float32"),o=i,l=!1;i.rank===4&&(l=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(o.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`),$(s==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Nn("avgPool3d",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r,dataFormat:s},d=O.runKernel(dc,u,p);return d=oe(d,o.dtype),l?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var rv=L({avgPool3d_:JM});function QM(e,t=0){$(e.length>=1,()=>"Pass at least one tensor to concat");let n=Jp(e,"tensors","concat","string_or_numeric");if(n[0].dtype==="complex64"&&n.forEach(s=>{if(s.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
-          with dtype ${s.dtype}. `)}),n.length===1)return ir(n[0]);let a=n,r={axis:t};return O.runKernel(Gl,a,r)}var Ze=L({concat_:QM});function eP(e){let t={x:_(e,"x","sigmoid","float32")};return O.runKernel(ro,t)}var ha=L({sigmoid_:eP});function tP(e,t,n){let a=_(e,"x","slice","string_or_numeric");if(a.rank===0)throw new Error("Slicing scalar is not possible");let r={x:a},s={begin:t,size:n};return O.runKernel(Su,r,s)}var Be=L({slice_:tP});function nP(e){let t={x:_(e,"x","tanh","float32")};return O.runKernel(co,t)}var si=L({tanh_:nP});function aP(e,t,n,a,r,s){let i=_(e,"forgetBias","basicLSTMCell"),o=_(t,"lstmKernel","basicLSTMCell"),l=_(n,"lstmBias","basicLSTMCell"),u=_(a,"data","basicLSTMCell"),p=_(r,"c","basicLSTMCell"),d=_(s,"h","basicLSTMCell"),c=Ze([u,d],1),h=Fe(c,o),m=Y(h,l),f=m.shape[0],g=m.shape[1]/4,y=[f,g],b=Be(m,[0,0],y),x=Be(m,[0,g],y),w=Be(m,[0,g*2],y),I=Be(m,[0,g*3],y),T=Y(z(ha(b),si(x)),z(p,ha(Y(i,w)))),C=z(si(T),ha(I));return[T,C]}var JS=L({basicLSTMCell_:aP});function rP(e,t,n){let a=_(e,"x","batchToSpaceND"),r=t.reduce((o,l)=>o*l);$(a.rank>=1+t.length,()=>`input rank is ${a.rank} but should be > than blockShape.length ${t.length}`),$(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`),$(a.shape[0]%r===0,()=>`input tensor batch is ${a.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${r}`);let s={x:a},i={blockShape:t,crops:n};return O.runKernel(Ul,s,i)}var Pc=L({batchToSpaceND_:rP});function sP(e){let t;return e.rank===0||e.rank===1?t=W(e,[1,1,1,e.size]):e.rank===2?t=W(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?t=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]):t=e,t}function iP(e,t,n,a,r,s){s==null&&(s=.001);let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;a!=null&&(p=_(a,"offset","batchNorm")),$(o.rank===l.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),$(p==null||o.rank===p.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),$(u==null||o.rank===u.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let d={x:sP(i),scale:u,offset:p,mean:o,variance:l},c={varianceEpsilon:s},h=O.runKernel(Fi,d,c);return W(h,i.shape)}var ks=L({batchNorm_:iP});function oP(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),$(o.rank===2||o.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${o.rank}.`),$(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===2||p.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${p.rank}.`),ks(i,o,l,p,u,s)}var sv=L({batchNorm2d_:oP});function lP(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),$(o.rank===3||o.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${o.rank}.`),$(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===3||p.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${p.rank}.`),ks(i,o,l,p,u,s)}var iv=L({batchNorm3d_:lP});function uP(e,t,n,a,r,s){let i=_(e,"x","batchNorm"),o=_(t,"mean","batchNorm"),l=_(n,"variance","batchNorm"),u;r!=null&&(u=_(r,"scale","batchNorm"));let p;return a!=null&&(p=_(a,"offset","batchNorm")),$(i.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),$(o.rank===4||o.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${o.rank}.`),$(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${l.rank}.`),u!=null&&$(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${u.rank}.`),p!=null&&$(p.rank===4||p.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${p.rank}.`),ks(i,o,l,p,u,s)}var ov=L({batchNorm4d_:uP});function pP(e,t,n){let a=_(e,"x","bincount"),r=_(t,"weights","bincount");$(a.dtype==="int32",()=>`Error in bincount: input dtype must be int32, but got ${a.dtype}`),$(n>=0,()=>`size must be non-negative, but got ${n}.`),$(r.size===a.size||r.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${a.shape}, weights shape: ${r.shape}.`);let s={x:a,weights:r},i={size:n};return O.runKernel(dm,s,i)}var lv=L({bincount_:pP});function cP(e,t){let n=_(e,"s0","broadcastArgs","int32"),a=_(t,"s1","broadcastArgs","int32");if(n.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${n.rank}`);if(a.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${a.rank}`);let r={s0:n,s1:a};return O.runKernel(hm,r)}var QS=L({broadcastArgs_:cP});function dP(e,t){let n=_(e,"broadcastTo","x"),a=n.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){let l=n.shape.slice();for(;l.length<t.length;)l.unshift(1);n=W(n,l)}let r=n.shape,s=Array.from(t);for(let l=t.length-1;l>=0;l--)if(r[l]===t[l])s[l]=1;else if(n.shape[l]!==1)throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${t}].`);if(s.map((l,u)=>l>1?u:-1).filter(l=>l>=0).length===0)return ir(n);let i={x:n},o={reps:s};return O.runKernel(bs,i,o)}var Xs=L({broadcastTo_:dP});function hP(e){let t={x:_(e,"x","ceil","float32")};return O.runKernel(vi,t)}var uv=L({ceil_:hP});function gn(e,t,n){let a={shape:e,value:t,dtype:n};return O.runKernel(gc,{},a)}function mP(e,t,n){let a=_(e,"x","clipByValue");if($(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`),t===n)return gn(a.shape,t,a.dtype);let r={x:a},s={clipValueMin:t,clipValueMax:n};return O.runKernel(ys,r,s)}var en=L({clipByValue_:mP});function fP(e){return Ze(e,0)}var pv=L({concat1d_:fP});function gP(e,t){return Ze(e,t)}var cv=L({concat2d_:gP});function yP(e,t){return Ze(e,t)}var dv=L({concat3d_:yP});function bP(e,t){return Ze(e,t)}var hv=L({concat4d_:bP});function xP(e,t,n,a,r="NHWC",s=[1,1],i){let o=_(e,"x","conv2d","float32"),l=_(t,"filter","conv2d","float32"),u=o,p=!1;o.rank===3&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(u.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${u.rank}.`),$(l.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${l.rank}.`),Nn("conv2d",a,i);let d=r==="NHWC"?u.shape[3]:u.shape[1];$(d===l.shape[2],()=>`Error in conv2d: depth of input (${d}) must match input depth for filter ${l.shape[2]}.`),$(cr(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);let c={x:u,filter:l},h={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i},m=O.runKernel(wi,c,h);return p?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var $t=L({conv2d_:xP});function vP(e,t,n,a,r="NWC",s=1,i){let o=_(e,"x","conv1d"),l=_(t,"filter","conv1d"),u=o,p=!1;o.rank===2&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1]])),$(u.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${u.rank}.`),$(l.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${l.rank}.`),Nn("conv1d",a,i),$(u.shape[2]===l.shape[1],()=>`Error in conv1d: depth of input (${u.shape[2]}) must match input depth for filter ${l.shape[1]}.`),$(cr(n,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${s}'`),$(r==="NWC",()=>`Error in conv1d: got dataFormat of ${r} but only NWC is currently supported.`);let d=W(l,[1,l.shape[0],l.shape[1],l.shape[2]]),c=W(u,[u.shape[0],1,u.shape[1],u.shape[2]]),h=$t(c,d,[1,n],a,"NHWC",[1,s],i);return p?W(h,[h.shape[2],h.shape[3]]):W(h,[h.shape[0],h.shape[2],h.shape[3]])}var qm=L({conv1d_:vP});function wP(e,t,n,a,r,s="NHWC",i){$(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,l=t,u=!1;t.rank===3&&(u=!0,l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]),o=[1,e[0],e[1],e[2]]),$(o.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`),$(l.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${l.rank}`),$(n.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${n.rank}`);let p=s==="NHWC"?o[3]:o[1],d=s==="NHWC"?l.shape[3]:l.shape[1];$(p===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${p}) must match input depth for filter ${n.shape[2]}.`),$(d===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${d}) must match output depth for filter ${n.shape[3]}.`),Nn("conv2dDerInput",r,i);let c={dy:l,filter:n},h={strides:a,pad:r,dataFormat:s,dimRoundingMode:i,inputShape:o},m=O.runKernel(ki,c,h);return u?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var mv=L({conv2DBackpropInput_:wP});function kP(e,t,n,a,r,s){let i=_(e,"x","conv2dTranspose"),o=_(t,"filter","conv2dTranspose");return mv(n,i,o,a,r,"NHWC",s)}var Km=L({conv2dTranspose_:kP});function IP(e,t,n,a,r="NDHWC",s=[1,1,1]){let i=_(e,"x","conv3d"),o=_(t,"filter","conv3d"),l=i,u=!1;i.rank===4&&(u=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(l.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${l.rank}.`),$(o.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`),$(l.shape[4]===o.shape[3],()=>`Error in conv3d: depth of input (${l.shape[4]}) must match input depth for filter ${o.shape[3]}.`),$(cr(n,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),$(r==="NDHWC",()=>`Error in conv3d: got dataFormat of ${r} but only NDHWC is currently supported.`);let p={x:l,filter:o},d={strides:n,pad:a,dataFormat:r,dilations:s},c=O.runKernel(mc,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}var fv=L({conv3d_:IP});function SP(e,t,n,a,r){$(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,i=t,o=!1;t.rank===4&&(o=!0,i=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,e[0],e[1],e[2],e[3]]);let l=s[4],u=i.shape[4];$(s.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),$(i.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),$(n.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${n.rank}`),$(l===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${l}) must match input depth for filter ${n.shape[3]}.`),$(u===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${u}) must match output depth for filter ${n.shape[4]}.`);let p={dy:i,filter:n},d={pad:r,strides:a,inputShape:s},c=O.runKernel(ym,p,d);return o?W(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}var eT=L({conv3DBackpropInput_:SP});function TP(e,t,n,a,r){let s=_(e,"x","conv3dTranspose"),i=_(t,"filter","conv3dTranspose");return eT(n,s,i,a,r)}var gv=L({conv3dTranspose_:TP});function NP(e){let t={x:_(e,"x","cos","float32")};return O.runKernel(Ii,t)}var Oc=L({cos_:NP});function CP(e){let t={x:_(e,"x","cosh","float32")};return O.runKernel(Si,t)}var Xm=L({cosh_:CP});function _P(e,t=0,n=!1,a=!1){let r={x:_(e,"x","cumprod")},s={axis:t,exclusive:n,reverse:a};return O.runKernel(Hl,r,s)}var ec=L({cumprod_:_P});function EP(e,t=0,n=!1,a=!1){let r={x:_(e,"x","cumsum")},s={axis:t,exclusive:n,reverse:a};return O.runKernel(Ti,r,s)}var Ym=L({cumsum_:EP});function AP(e,t,n,a=!1){let r=_(e,"x","denseBincount"),s=_(t,"weights","denseBincount");$(r.dtype==="int32",()=>`Error in denseBincount: input dtype must be int32, but got ${r.dtype}`),$(r.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${r.rank}.`),$(n>=0,()=>`size must be non-negative, but got ${n}.`),$(s.size===r.size||s.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${r.shape}, weights shape: ${s.shape}.`);let i={x:r,weights:s},o={size:n,binaryOutput:a};return O.runKernel(bm,i,o)}var zh=L({denseBincount_:AP});function $P(e,t,n="NHWC"){let a=_(e,"x","depthToSpace","float32"),r=n==="NHWC"?a.shape[1]:a.shape[2],s=n==="NHWC"?a.shape[2]:a.shape[3],i=n==="NHWC"?a.shape[3]:a.shape[1];$(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),$(r*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${r} and ${t}  for depthToSpace with input shape
-    ${a.shape}`),$(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${s} and ${t} for depthToSpace with input shape
-        ${a.shape}`),$(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${a.shape}`);let o={x:a},l={blockSize:t,dataFormat:n};return O.runKernel(ql,o,l)}var yv=L({depthToSpace_:$P});function FP(e,t,n,a,r="NHWC",s=[1,1],i){let o=_(e,"x","depthwiseConv2d","float32"),l=_(t,"filter","depthwiseConv2d","float32"),u=o,p=!1;o.rank===3&&(p=!0,u=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(u.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${u.rank}.`),$(l.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${l.rank}.`);let d=r==="NHWC"?u.shape[3]:u.shape[1];$(d===l.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${d}) must match the inChannels dimension in filter ${l.shape[2]}.`),Nn("depthwiseConv2d",a,i);let c={x:u,filter:l},h={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i},m=O.runKernel(Ni,c,h);return p?W(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var Is=L({depthwiseConv2d_:FP});function DP(e){let t={x:_(e,"x","diag")};return O.runKernel(wm,t)}var tT=L({diag_:DP});function RP(e,t,n,a,r=[1,1],s="NHWC"){let i=_(e,"x","dilation2d"),o=_(t,"filter","dilation2d");$(i.rank===3||i.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),$(o.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`),$(s==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let l=i,u=!1;i.rank===3&&(l=W(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=!0);let p={x:l,filter:o},d={strides:n,pad:a,dilations:r},c=O.runKernel(fc,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var bv=L({dilation2d_:RP});function MP(e,t){let n=_(e,"a","equal","string_or_numeric"),a=_(t,"b","equal","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Xl,r)}var ea=L({equal_:MP});function PP(e,t,n){let a=_(t,"a","where"),r=_(n,"b","where"),s=_(e,"condition","where","bool"),i=ut(ut(s.shape,a.shape),r.shape),o=Xs(s,i),l=Xs(a,i),u=Xs(r,i),p={condition:o,t:l,e:u};return O.runKernel(ku,p)}var mn=L({where_:PP});function OP(e){let t={x:_(e,"x","zerosLike")};return O.runKernel(Mu,t)}var qe=L({zerosLike_:OP});function LP(e,t){let n=_(e,"a","div"),a=_(t,"b","div");[n,a]=_t(n,a);let r=he(n,a),s=qe(r),i=ea(a,s);return mn(i,s,r)}var xv=L({divNoNan_:LP});function zP(e,t){let n=_(e,"t1","dot"),a=_(t,"t2","dot");$((n.rank===1||n.rank===2)&&(a.rank===1||a.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${a.rank}.`);let r=n.rank===1?n.size:n.shape[1],s=a.rank===1?a.size:a.shape[0];if($(r===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${r} and ${s}.`),n.rank===1&&a.rank===1){let i=W(n,[1,-1]),o=W(a,[-1,1]),l=Fe(i,o);return W(l,[])}else if(n.rank===1&&a.rank===2){let i=W(n,[1,-1]),o=W(a,[a.shape[0],a.shape[1]]),l=Fe(i,o);return W(l,[l.size])}else if(n.rank===2&&a.rank===1){let i=W(a,[-1,1]),o=Fe(n,i);return W(o,[o.size])}else{let i=W(a,[a.shape[0],a.shape[1]]);return Fe(n,i)}}var vv=L({dot_:zP});function WP(e,...t){let n=t.map((r,s)=>_(r,`tensors${s}`,"einsum")),a={equation:e};return O.runKernel(km,n,a)}var nT=L({einsum_:WP});function BP(e){let t={x:_(e,"x","elu","float32")};return O.runKernel(_i,t)}var Lu=L({elu_:BP});function VP(e){let t=_(e,"x","erf");$(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=oe(t,"float32"));let n={x:t};return O.runKernel(Kl,n)}var wv=L({erf_:VP});function kv(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function aT(e,t,n){let a=e.length+t.length,r=[],s=0,i=0;for(let o=0;o<a;o++)n.indexOf(o)===-1?r.push(e[s++]):r.push(t[i++]);return r}function rT(e,t){let n=[],a=e.length;for(let s=0;s<a;s++)t.indexOf(s)===-1&&n.push(e[s]);let r=t.map(s=>e[s]);return[n,r]}function ii(e,t){let n=t.map(a=>1);return aT(e,n,t)}function UP(e,t,n){$(kv(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function sT(e,t){if(kv(e,t))return null;let n=[];for(let a=0;a<t;++a)e.indexOf(a)===-1&&n.push(a);return e.forEach(a=>n.push(a)),n}function Iv(e){return e.map((t,n)=>[n,t]).sort((t,n)=>t[1]-n[1]).map(t=>t[0])}function GP(e,t){let n=[];for(let a=t-e;a<t;++a)n.push(a);return n}function HP(e,t=null,n=!1){let a={x:_(e,"x","max")},r={reductionIndices:t,keepDims:n};return O.runKernel(Oi,a,r)}var ma=L({max_:HP});function jP(e,t=null,n=!1){let a={x:_(e,"x","min")},r={axis:t,keepDims:n};return O.runKernel(Bi,a,r)}var bl=L({min_:jP});function qP(e,t){let n=_(e,"base","pow"),a=_(t,"exp","pow");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(qi,r)}var $r=L({pow_:qP});function be(e,t){if((hn(e)&&t!=="string"||Array.isArray(e))&&t!=="complex64")throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if(t==="string"&&hn(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return vs(e,[],[],t)}function KP(e){let t={x:_(e,"x","sqrt","float32")};return O.runKernel(so,t)}var un=L({sqrt_:KP});function XP(e){let t=_(e,"x","square"),n={};return O.runKernel("Square",{x:t},n)}var lt=L({square_:XP});function YP(e,t=null,n=!1){let a=_(e,"x","sum");a.dtype==="bool"&&(a=oe(a,"int32"));let r={x:a},s={axis:t,keepDims:n};return O.runKernel(io,r,s)}var fe=L({sum_:YP});function ZP(e,t="euclidean",n=null,a=!1){e=_(e,"x","norm");let r=iT(e,t,n),s=r.shape;if(a){let i=Fa(n,e.shape);s=ii(r.shape,i)}return W(r,s)}function iT(e,t,n=null){if(e.rank===0)return Lt(e);if(e.rank!==1&&n===null)return iT(W(e,[-1]),t,n);if(e.rank===1||typeof n=="number"||Array.isArray(n)&&n.length===1){if(t===1)return fe(Lt(e),n);if(t===1/0)return ma(Lt(e),n);if(t===-1/0)return bl(Lt(e),n);if(t==="euclidean"||t===2)return un(fe($r(Lt(e),be(2,"int32")),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(n)&&n.length===2){if(t===1)return ma(fe(Lt(e),n[0]),n[1]-1);if(t===1/0)return ma(fe(Lt(e),n[1]),n[0]);if(t===-1/0)return bl(fe(Lt(e),n[1]),n[0]);if(t==="fro"||t==="euclidean")return un(fe(lt(e),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${n}`)}var zu=L({norm_:ZP});function JP(e,t=null,n=!1){return zu(e,"euclidean",t,n)}var Sv=L({euclideanNorm_:JP});function QP(e){let t={x:_(e,"x","exp")};return O.runKernel(Ei,t)}var fn=L({exp_:QP});function eO(e,t=0){let n=_(e,"x","expandDims","string_or_numeric");$(t<=n.rank,()=>"Axis must be <= rank of the tensor");let a={input:n},r={dim:t};return O.runKernel(Yl,a,r)}var Zt=L({expandDims_:eO});function tO(e){let t={x:_(e,"x","expm1")};return O.runKernel(Zl,t)}var Tv=L({expm1_:tO});function nO(e,t){let n=_(e,"x","tile","string_or_numeric");$(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);let a={x:n},r={reps:t};return O.runKernel(bs,a,r)}var Ln=L({tile_:nO});function aO(e,t,n,a="float32"){t==null&&(t=e);let r=Oe([e,t],a),s=e<=t?e:t;for(let o=0;o<s;++o)r.set(1,o,o);let i=W(r.toTensor(),[e,t]);if(n==null)return i;if(n.length===1)return Ln(Zt(i,0),[n[0],1,1]);if(n.length===2)return Ln(Zt(Zt(i,0),0),[n[0],n[1],1,1]);if(n.length===3)return Ln(Zt(Zt(Zt(i,0),0),0),[n[0],n[1],n[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)}var Zm=L({eye_:aO});function rO(e){let t={x:_(e,"x","floor","float32")};return O.runKernel(Ai,t)}var Wu=L({floor_:rO});function sO(e,t,n=0,a=0){let r=_(e,"x","gather"),s=_(t,"indices","gather","int32"),i={x:r,indices:s},o={axis:n,batchDims:a};return O.runKernel(Ql,i,o)}var Bu=L({gather_:sO});function iO(e,t){let n=_(e,"a","greater","string_or_numeric"),a=_(t,"b","greater","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(tu,r)}var Cn=L({greater_:iO});function oO(e,t){let n=_(e,"a","greaterEqual","string_or_numeric"),a=_(t,"b","greaterEqual","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Di,r)}var Fr=L({greaterEqual_:oO});function lO(e){let t={x:_(e,"x","isFinite")};return O.runKernel(nu,t)}var Nv=L({isFinite_:lO});function uO(e){let t={x:_(e,"x","isInf")};return O.runKernel(au,t)}var Cv=L({isInf_:uO});function pO(e){let t={x:_(e,"x","isNaN")};return O.runKernel(ru,t)}var _v=L({isNaN_:pO});function cO(e,t=.2){let n={x:_(e,"x","leakyRelu")},a={alpha:t};return O.runKernel(Mi,n,a)}var Lc=L({leakyRelu_:cO});function dO(e,t){let n=_(e,"a","less","string_or_numeric"),a=_(t,"b","less","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(su,r)}var Jm=L({less_:dO});function hO(e,t){let n=_(e,"a","lessEqual","string_or_numeric"),a=_(t,"b","lessEqual","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(iu,r)}var Ss=L({lessEqual_:hO});function oT(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");let a={start:e,stop:t,num:n};return O.runKernel(Cm,{},a)}function mO(e,t=5,n=1,a=1,r=.5){let s=_(e,"x","localResponseNormalization");$(s.rank===4||s.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
-               rank ${s.rank}.`),$(hl(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,o=!1;s.rank===3&&(o=!0,i=W(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let l={x:i},u={depthRadius:t,bias:n,alpha:a,beta:r},p=O.runKernel(yc,l,u);return o?W(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var Ev=L({localResponseNormalization_:mO});function fO(e){let t={x:_(e,"x","log","float32")};return O.runKernel(Pi,t)}var ta=L({log_:fO});function gO(e){let t={x:_(e,"x","log1p")};return O.runKernel(ou,t)}var zc=L({log1p_:gO});function yO(e){return $(ss(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{let a=_(t,"x","tf.grad","string_or_numeric"),r=n!=null?_(n,"dy","tf.grad"):null;return O.tidy(()=>{let{value:s,grads:i}=O.gradients(()=>e(a),[a],r);return r!=null&&Tn(s.shape,r.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Qm(i),i[0]})}}function bO(e){return $(ss(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{$(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let a=Jp(t,"args","tf.grads","string_or_numeric"),r=n!=null?_(n,"dy","tf.grads"):null;return O.tidy(()=>{let{value:s,grads:i}=O.gradients(()=>e(...a),a,r);return r!=null&&Tn(s.shape,r.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Qm(i),i})}}function xO(e){return $(ss(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{$(t instanceof Te,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),$(n==null||n instanceof Te,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:a,value:r}=O.gradients(()=>e(t),[t],n);return Qm(a),{grad:a[0],value:r}}}function vO(e){return $(ss(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{$(Array.isArray(t)&&t.every(r=>r instanceof Te),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),$(n==null||n instanceof Te,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let a=O.gradients(()=>e(...t),t,n);return n!=null&&Tn(a.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Qm(a.grads),a}}function lT(e,t){$(ss(e),()=>"The f passed in variableGrads(f) must be a function"),$(t==null||Array.isArray(t)&&t.every(u=>u instanceof is),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let n=t!=null;if(!n){t=[];for(let u in O.registeredVariables)t.push(O.registeredVariables[u])}let a=n?t.filter(u=>!u.trainable):null,r=t.length;t=t.filter(u=>u.trainable),$(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${r} variables is trainable.`);let s=!0,{value:i,grads:o}=O.gradients(e,t,null,s);$(o.some(u=>u!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),$(i.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${i.rank} tensor`);let l={};return t.forEach((u,p)=>{o[p]!=null&&(l[u.name]=o[p])}),a!=null&&a.forEach(u=>l[u.name]=null),{value:i,grads:l}}function ur(e){return O.customGrad(e)}function Qm(e){if(e.filter(t=>t==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
-    the f you passed encloses all operations that lead from x to y.`)}function wO(e){let t={x:_(e,"x","softplus")};return O.runKernel(Cu,t)}var mo=L({softplus_:wO});function kO(e){let t=_(e,"x","logSigmoid");return ur(n=>({value:vt(mo(vt(n))),gradFunc:a=>z(a,ha(vt(n)))}))(t)}var Av=L({logSigmoid_:kO});function IO(e,t){let n=_(e,"a","sub"),a=_(t,"b","sub");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(uo,r)}var pe=L({sub_:IO});function SO(e,t=-1){let n=_(e,"logits","logSoftmax");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);return ur((a,r)=>{let s=ma(a,t,!0),i=pe(a,s),o=pe(oe(i,"float32"),ta(fe(fn(i),t,!0)));return r([o]),{value:o,gradFunc:(l,u)=>{let[p]=u,d=!0,c=fn(p);return pe(l,z(fe(l,t,d),c))}}})(n)}var ef=L({logSoftmax_:SO});function TO(e,t=null,n=!1){let a=_(e,"x","logSumExp"),r=Fa(t,a.shape),s=ma(a,r,!0),i=pe(a,s),o=fn(i),l=fe(o,r),u=ta(l),p=Y(W(s,u.shape),u);if(n){let d=ii(p.shape,r);return W(p,d)}return p}var tf=L({logSumExp_:TO});function NO(e,t){let n=_(e,"a","logicalAnd","bool"),a=_(t,"b","logicalAnd","bool");ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(lu,r)}var $a=L({logicalAnd_:NO});function CO(e){let t={x:_(e,"x","logicalNot","bool")};return O.runKernel(uu,t)}var Wc=L({logicalNot_:CO});function _O(e,t){let n=_(e,"a","logicalOr","bool"),a=_(t,"b","logicalOr","bool");ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(pu,r)}var nf=L({logicalOr_:_O});function EO(e,t){let n=_(e,"a","logicalXor","bool"),a=_(t,"b","logicalXor","bool");return ut(n.shape,a.shape),$a(nf(e,t),Wc($a(e,t)))}var $v=L({logicalXor_:EO}),ih=2147483648;function AO(e,t,n="left"){let a=_(e,"sortedSequence","searchSorted"),r=_(t,"values","searchSorted"),s=a.shape[a.shape.length-1],i=r.shape[r.shape.length-1],o=W(a,[-1,s]),l=W(r,[-1,i]);if(o.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(o.shape[0]!==l.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(mt(l.shape)>=ih)throw new Error(`values tensor size must less than ${ih}`);if(o.shape[1]>=ih)throw new Error(`trailing dim_size must less than ${ih} for int32 output type, was ${o.shape[1]}`);let u={sortedSequence:o,values:l},p={side:n};return O.runKernel(zm,u,p)}var af=L({searchSorted_:AO});function uT(e,t){return af(e,t,"left")}function $O(e,t,n,a,r){let s=_(e,"x","maxPool"),i=1,o=s,l=!1;s.rank===3&&(l=!0,o=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(o.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${o.rank}.`),$(cr(n,i),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${i}'`),Nn("maxPool",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r},d=O.runKernel(zi,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var Dt=L({maxPool_:$O});function FO(e,t=[1,1,1],n,a,r,s="NDHWC"){let i=_(e,"x","maxPool3d"),o=i,l=!1;i.rank===4&&(l=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),$(o.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`),$(s==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Nn("maxPool3d",a,r);let u={x:o},p={filterSize:t,strides:n,pad:a,dimRoundingMode:r,dataFormat:s},d=O.runKernel(bc,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var Fv=L({maxPool3d_:FO});function DO(e,t,n,a,r=!1){let s={x:_(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:n,pad:a,includeBatchInIndex:r},o=O.runKernel($m,s,i);return{result:o[0],indexes:o[1]}}var pT=L({maxPoolWithArgmax_:DO});function RO(e,t){let n=_(e,"a","maximum"),a=_(t,"b","maximum");[n,a]=_t(n,a),n.dtype==="bool"&&(n=oe(n,"int32"),a=oe(a,"int32")),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Li,r)}var dr=L({maximum_:RO});function MO(e,t=null,n=!1){let a={x:_(e,"x","mean")},r={axis:t,keepDims:n};return O.runKernel(Wi,a,r)}var Nt=L({mean_:MO});function It(e,t="float32"){if(t==="complex64"){let a=It(e,"float32"),r=It(e,"float32");return Ar(a,r)}let n=um(mt(e),t);return O.makeTensor(n,e,t)}function Jn(e,t="float32"){if(t==="complex64"){let a=Jn(e,"float32"),r=It(e,"float32");return Ar(a,r)}let n=_x(mt(e),t);return O.makeTensor(n,e,t)}function cT(e,t,{indexing:n="xy"}={}){if(n!=="xy"&&n!=="ij")throw new TypeError(`${n} is not a valid third argument to meshgrid`);if(e===void 0)return[];let a=_(e,"x","meshgrid",e instanceof Te?e.dtype:"float32");if(t===void 0)return[a];let r=_(t,"y","meshgrid",t instanceof Te?t.dtype:"float32"),s=mt(a.shape),i=mt(r.shape);return n==="xy"?(a=W(a,[1,-1]),r=W(r,[-1,1]),[Fe(Jn([i,1],a.dtype),a),Fe(r,Jn([1,s],r.dtype))]):(a=W(a,[-1,1]),r=W(r,[1,-1]),[Fe(a,Jn([1,i],a.dtype)),Fe(Jn([s,1],r.dtype),r)])}function PO(e,t){let n=_(e,"a","minimum"),a=_(t,"b","minimum");[n,a]=_t(n,a),n.dtype==="bool"&&(n=oe(n,"int32"),a=oe(a,"int32")),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(Vi,r)}var Vu=L({minimum_:PO});function OO(e,t,n){$(n==="reflect"||n==="symmetric",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);let a=_(e,"x","mirrorPad");if(a.rank===0)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");$(t.length===a.rank,()=>`Padding doesn't match input. Must be ${a.rank}. Got ${t.length}.`);let r=n==="reflect"?1:0;for(let o=0;o<a.rank;o++)$(t[o].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),$(t[o][0]>=0&&t[o][0]<=a.shape[o]-r&&t[o][1]>=0&&t[o][1]<=a.shape[o]-r,()=>`Padding in dimension ${o} cannot be greater than or equal to ${a.shape[o]-r} or less than 0 for input of shape ${a.shape}`);let s={paddings:t,mode:n},i={x:a};return O.runKernel(Ui,i,s)}var Dv=L({mirrorPad_:OO});function LO(e,t){let n=_(e,"a","mod"),a=_(t,"b","mod");[n,a]=_t(n,a);let r={a:n,b:a};return O.runKernel(cu,r)}var Rv=L({mod_:LO});function zO(e,t=null,n=!1){e=_(e,"x","moments");let a=Fa(t,e.shape),r=Nt(e,a,n),s=r.shape;n||(s=ii(r.shape,a));let i=lt(pe(oe(e,"float32"),W(r,s))),o=Nt(i,a,n);return{mean:r,variance:o}}var Bc=L({moments_:zO});function WO(e,t,n,a){let r=_(t,"data","multiRNNCell"),s=Jp(n,"c","multiRNNCell"),i=Jp(a,"h","multiRNNCell"),o=r,l=[];for(let d=0;d<e.length;d++){let c=e[d](o,s[d],i[d]);l.push(c[0]),l.push(c[1]),o=c[1]}let u=[],p=[];for(let d=0;d<l.length;d+=2)u.push(l[d]),p.push(l[d+1]);return[u,p]}var dT=L({multiRNNCell_:WO});function BO(e,t,n,a=!1){let r=_(e,"logits","multinomial"),s=r.size,i=r.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);n=n||Math.random();let o={logits:i===1?W(r,[1,-1]):r},l={numSamples:t,seed:n,normalized:a},u=O.runKernel(Fm,o,l);return i===1?W(u,[u.size]):u}var hT=L({multinomial_:BO});function VO(e,t){let n=_(e,"a","notEqual","string_or_numeric"),a=_(t,"b","notEqual","string_or_numeric");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a};return O.runKernel(hu,r)}var oi=L({notEqual_:VO});function UO(e){let t={x:_(e,"x","onesLike")};return O.runKernel(yu,t)}var na=L({onesLike_:UO});function GO(e,t){let n=_(e,"v1","outerProduct"),a=_(t,"v2","outerProduct");$(n.rank===1&&a.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${a.rank}.`);let r=W(n,[-1,1]),s=W(a,[1,-1]);return Fe(r,s)}var mT=L({outerProduct_:GO});function HO(e,t,n=0){let a=_(e,"x","pad");if(a.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let r={paddings:t,constantValue:n},s={x:a};return O.runKernel(ji,s,r)}var ba=L({pad_:HO});function jO(e,t,n=0){return $(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),ba(e,[t],n)}var fT=L({pad1d_:jO});function qO(e,t,n=0){return $(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ba(e,t,n)}var gT=L({pad2d_:qO});function KO(e,t,n=0){return $(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ba(e,t,n)}var yT=L({pad3d_:KO});function XO(e,t,n=0){return $(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ba(e,t,n)}var bT=L({pad4d_:XO});function YO(e,t,n){let a=_(e,"x","spaceToBatchND");$(a.rank>=1+t.length,()=>`input rank ${a.rank} should be > than [blockShape] ${t.length}`),$(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),$(a.shape.reduce((i,o,l)=>l>0&&l<=t.length?i&&(o+n[l-1][0]+n[l-1][1])%t[l-1]===0:i,!0),()=>`input spatial dimensions ${a.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);let r={x:a},s={blockShape:t,paddings:n};return O.runKernel(_u,r,s)}var Vc=L({spaceToBatchND_:YO});function ZO(e,t,n,a,r,s,i){r==null&&(r=[1,1]),s==null&&(s=1),a===0&&(a="valid");let o=_(e,"x","maxPool"),l=o,u=!1;o.rank===3&&(u=!0,l=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),$(cr(s,r),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${r}'`);let p=XS(l.shape,t,s,r,a),d=[p.dilationHeight,p.dilationWidth],c;a==="same"?c=QO([p.filterHeight,p.filterWidth],d):c=[[0,0],[0,0]];let h=d[0]===1&&d[1]===1,[m,f]=JO([p.inHeight,p.inWidth],d,c),g=h?a:"valid",y=h?l:Vc(l,d,m),b=(n==="avg"?()=>ya(y,t,s,g,i):()=>Dt(y,t,s,g,i))(),x=h?b:Pc(b,d,f);return u?W(x,[x.shape[1],x.shape[2],x.shape[3]]):x}function JO(e,t,n){let a=n.map(p=>p[0]),r=n.map(p=>p[1]),s=e.concat(a,r),i=t.map((p,d)=>(p-s[d]%p)%p),o=r.map((p,d)=>p+i[d]),l=t.map((p,d)=>[a[d],o[d]]),u=t.map((p,d)=>[0,i[d]]);return[l,u]}function QO(e,t){let n=e.map((s,i)=>s+(s-1)*(t[i]-1)).map(s=>s-1),a=n.map(s=>Math.floor(s/2)),r=n.map((s,i)=>s-a[i]);return n.map((s,i)=>[a[i],r[i]])}var Mv=L({pool_:ZO});function e3(e,t){let n=_(e,"x","prelu"),a=_(t,"alpha","prelu"),r={x:n,alpha:a};return O.runKernel(Ki,r)}var Uc=L({prelu_:e3});function t3(e,t=null,n=!1){let a=_(e,"x","prod");a.dtype==="bool"&&(a=oe(a,"int32"));let r={x:a},s={axis:t,keepDims:n};return O.runKernel(Xi,r,s)}var Pv=L({prod_:t3});function n3(e,t,n,a){let r=e.map((p,d)=>_(p,`tensors${d}`,"raggedGather","int32")),s=_(t,"paramsDenseValues","raggedGather"),i=_(n,"indices","raggedGather","int32"),o={paramsNestedSplits:r,paramsDenseValues:s,indices:i},l={outputRaggedRank:a},u=O.runKernel(Dm,o,l);return{outputNestedSplits:u.slice(0,u.length-1),outputDenseValues:u[u.length-1]}}var xT=L({raggedGather_:n3});function a3(e,t,n){let a=_(e,"starts","raggedRange"),r=_(t,"limits","raggedRange",a.dtype),s=_(n,"deltas","raggedRange",a.dtype),i={starts:a,limits:r,deltas:s},o=O.runKernel(Rm,i);return{rtNestedSplits:o[0],rtDenseValues:o[1]}}var vT=L({raggedRange_:a3});function r3(e,t,n,a,r){let s=_(e,"shape","raggedTensorToTensor","int32"),i=_(t,"values","raggedTensorToTensor"),o=_(n,"defaultValue","raggedTensorToTensor",i.dtype),l=a.map((d,c)=>_(d,`tensors${c}`,"raggedTensorToTensor","int32")),u={shape:s,values:i,defaultValue:o,rowPartitionTensors:l},p={rowPartitionTypes:r};return O.runKernel(Mm,u,p)}var wT=L({raggedTensorToTensor_:r3});function s3(e,t,n){let a=mt(e),r=null;if(n==null||n==="float32")r=new Float32Array(a);else if(n==="int32")r=new Int32Array(a);else if(n==="bool")r=new Uint8Array(a);else throw new Error(`Unknown data type ${n}`);for(let s=0;s<a;s++)r[s]=t();return O.makeTensor(r,e,n)}var kT=L({rand_:s3}),Ov=ms(im()),Lv=class{constructor(e,t,n,a,r){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=a,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let s=r||Math.random();this.random=Ov.alea(s.toString())}nextValue(){if(!isNaN(this.nextVal)){let a=this.nextVal;return this.nextVal=NaN,a}let e,t,n=!1;for(;!n;){let a,r,s;do a=2*this.random()-1,r=2*this.random()-1,s=a*a+r*r;while(s>=1||s===0);let i=Math.sqrt(-2*Math.log(s)/s);e=this.mean+this.stdDev*a*i,t=this.mean+this.stdDev*r*i,(!this.truncated||this.isValidTruncated(e))&&(n=!0)}return(!this.truncated||this.isValidTruncated(t))&&(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return this.dtype==null||this.dtype==="float32"?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}},i3=class{constructor(e,t,n,a){this.alpha=e,this.beta=1/t,this.dtype=n;let r=a||Math.random();this.randu=Ov.alea(r.toString()),this.randn=new Lv(0,1,n,!1,this.randu()),e<1?this.d=e+2/3:this.d=e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,n,a,r,s;for(;;){do a=this.randn.nextValue(),s=1+this.c*a;while(s<=0);if(s*=s*s,e=a*a,t=1-.331*e*e,n=.5*e+this.d*(1-s+Math.log(s)),r=this.randu(),r<t||Math.log(r)<n)break}return s=1/this.beta*this.d*s,this.alpha<1&&(s*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(s)}convertValue(e){return this.dtype==="float32"?e:Math.round(e)}},o3=class{constructor(e=0,t=1,n,a){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=e,this.range=t-e,this.dtype=n,a==null&&(a=Math.random()),typeof a=="number"&&(a=a.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=Ov.alea(a)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function l3(e,t,n=1,a="float32",r){if(n==null&&(n=1),a==null&&(a="float32"),a!=="float32"&&a!=="int32")throw new Error(`Unsupported data type ${a}`);let s=new i3(t,n,a,r),i=Oe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var IT=L({randomGamma_:l3});function u3(e,t=0,n=1,a,r){if(a!=null&&a==="bool")throw new Error(`Unsupported data type ${a}`);let s=new Lv(t,n,a,!1,r),i=Oe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var rf=L({randomNormal_:u3});function p3(e,t,n){if(t!=null&&t==="bool")throw new Error(`Unsupported data type ${t}`);return rf(e,0,1,t,n)}var ST=L({randomStandardNormal_:p3});function c3(e,t=0,n=1,a="float32",r){let s=Oe(e,a),i=new o3(t,n,null,r);for(let o=0;o<s.values.length;o++)s.values[o]=i.nextValue();return s.toTensor()}var Uu=L({randomUniform_:c3});function xl(e,t,n=1,a="float32"){if(n===0)throw new Error("Cannot have a step of zero");let r={start:e,stop:t,step:n,dtype:a};return O.runKernel(xc,{},r)}function d3(e){let t={x:_(e,"x","reciprocal")};return O.runKernel(xu,t)}var zv=L({reciprocal_:d3});function h3(e){let t={x:_(e,"x","relu")};return O.runKernel(Yi,t)}var Xe=L({relu_:h3});function m3(e){let t={x:_(e,"x","relu6")};return O.runKernel(Qi,t)}var sf=L({relu6_:m3});function f3(e,t){let n={x:_(e,"x","reverse")},a={dims:t};return O.runKernel(eo,n,a)}var ga=L({reverse_:f3});function g3(e){let t=_(e,"x","reverse");return $(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),ga(t,0)}var TT=L({reverse1d_:g3});function y3(e,t){let n=_(e,"x","reverse");return $(n.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),ga(n,t)}var NT=L({reverse2d_:y3});function b3(e,t){let n=_(e,"x","reverse");return $(n.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),ga(n,t)}var CT=L({reverse3d_:b3});function x3(e,t){let n=_(e,"x","reverse");return $(n.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),ga(n,t)}var _T=L({reverse4d_:x3});function v3(e){let t={x:_(e,"x","round")};return O.runKernel(to,t)}var of=L({round_:v3});function w3(e){let t={x:_(e,"x","rsqrt","float32")};return O.runKernel(no,t)}var lf=L({rsqrt_:w3});function k3(e){let t={x:_(e,"x","selu")};return O.runKernel(Iu,t)}var uf=L({selu_:k3});function I3(e,t,n,a,r,s=[1,1],i="NHWC"){let o=_(e,"x","separableConv2d"),l=_(t,"depthwiseFilter","separableConv2d"),u=_(n,"pointwiseFilter","separableConv2d"),p=o,d=!1;if(o.rank===3&&(d=!0,p=W(o,[1,o.shape[0],o.shape[1],o.shape[2]])),i==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");$(p.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${p.rank}.`),$(l.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${l.rank}.`),$(u.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${l.rank}.`),$(u.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${u.shape[0]}.`),$(u.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${u.shape[1]}.`);let c=l.shape[2],h=l.shape[3];$(u.shape[2]===c*h,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${c*h}, but got ${u.shape[2]}.`);let m=Is(p,l,a,r,i,s),f=$t(m,u,1,"valid",i);return d?W(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var Ts=L({separableConv2d_:I3});async function S3(e,t){let n=_(e,"x","setdiff1d"),a=_(t,"y","setdiff1d");$(n.dtype===a.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${a.dtype}).`),$(n.rank===1,()=>`x should be 1D tensor, but got x (${n.shape}).`),$(a.rank===1,()=>`y should be 1D tensor, but got y (${a.shape}).`);let r=await n.data(),s=await a.data(),i=new Set(s),o=0;for(let p=0;p<r.length;p++)i.has(r[p])||o++;let l=new Ht([o],n.dtype),u=new Ht([o],"int32");for(let p=0,d=0;p<r.length;p++)i.has(r[p])||(l.values[d]=r[p],u.values[d]=p,d++);return[l.toTensor(),u.toTensor()]}var ET=S3;function T3(e){let t={x:_(e,"x","sign")};return O.runKernel(Nu,t)}var Wv=L({sign_:T3});function N3(e){let t={x:_(e,"x","sin","float32")};return O.runKernel(ao,t)}var pf=L({sin_:N3});function C3(e){let t={x:_(e,"x","sinh")};return O.runKernel(Tu,t)}var cf=L({sinh_:C3});function _3(e,t,n){let a=_(e,"x","slice1d");return $(a.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${a.rank} tensor`),Be(a,[t],[n])}var Gc=L({slice1d_:_3});function E3(e,t,n){let a=_(e,"x","slice2d");return $(a.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var df=L({slice2d_:E3});function A3(e,t,n){let a=_(e,"x","slice3d");return $(a.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var fo=L({slice3d_:A3});function $3(e,t,n){let a=_(e,"x","slice4d");return $(a.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${a.rank} tensor`),Be(a,t,n)}var vl=L({slice4d_:$3});function F3(e,t=-1){let n=_(e,"logits","softmax","float32");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);let a={logits:n},r={dim:t};return O.runKernel(oo,a,r)}var Ka=L({softmax_:F3});function D3(e){$(e.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);let t={input:e};return O.runKernel(Sm,t)}var Hc=L({fft_:D3});function R3(e){$(e.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);let t={input:e};return O.runKernel(Tm,t)}var wl=L({ifft_:R3});function M3(e){let t=e.shape[e.shape.length-1],n=e.size/t,a;if(t<=2){let r=W(e,[n,t]);a=wl(r)}else{let r=[n,2*(t-1)],s=W(yl(e),[n,t]),i=W(Dc(e),[n,t]),o=ga(Be(s,[0,1],[n,t-2]),1),l=z(ga(Be(i,[0,1],[n,t-2]),1),be(-1)),u=Ze([s,o],1),p=Ze([i,l],1),d=W(Ar(u,p),[r[0],r[1]]);a=wl(d)}if(a=yl(a),e.rank===3&&e.shape[0]!==0){let r=a,s=e.shape[0];a=W(a,[s,a.shape[0]/s,a.shape[1]]),r.dispose()}return a}var hf=L({irfft_:M3});function P3(e,t,n=0){let a={x:_(e,"x","split")},r={numOrSizeSplits:t,axis:n};return O.runKernel(Eu,a,r)}var zn=L({split_:P3});function O3(e,t){$(e.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let n=e.shape[e.shape.length-1],a=e.size/n,r;if(t!=null&&t<n){let m=e.shape.map(g=>0),f=e.shape.map(g=>g);f[e.shape.length-1]=t,r=Be(e,m,f),n=t}else if(t!=null&&t>n){let m=e.shape.map(f=>f);m[e.shape.length-1]=t-n,r=Ze([e,It(m)],e.shape.length-1),n=t}else r=e;let s=qe(r),i=W(Ar(r,s),[a,n]),o=Hc(i),l=Math.floor(n/2)+1,u=yl(o),p=Dc(o),d=zn(u,[l,n-l],u.shape.length-1),c=zn(p,[l,n-l],p.shape.length-1),h=r.shape.slice();return h[r.shape.length-1]=l,W(Ar(d[0],c[0]),h)}var jc=L({rfft_:O3});function L3(e,t){let n=_(e,"a","squaredDifference"),a=_(t,"b","squaredDifference");[n,a]=_t(n,a),ut(n.shape,a.shape);let r={a:n,b:a},s={};return O.runKernel(lo,r,s)}var mf=L({squaredDifference_:L3});function z3(e,t){let n=_(e,"x","squeeze","string_or_numeric");return W(n,GI(n.shape,t).newShape)}var Ns=L({squeeze_:z3});function W3(e,t=0){let n=Jp(e,"tensors","stack","string_or_numeric");$(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length>0&&$(t<=n[0].rank,()=>"Axis must be <= rank of the tensor");let a=n,r={axis:t};return O.runKernel(bu,a,r)}var Ft=L({stack_:W3});function B3(e,t=0){let n={x:_(e,"x","step")},a={alpha:t};return O.runKernel(xs,n,a)}var go=L({step_:B3});function V3(e,t,n,a,r=0,s=0,i=0,o=0,l=0){let u={x:_(e,"x","stridedSlice","string_or_numeric")},p={begin:t,end:n,strides:a,beginMask:r,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};return O.runKernel($u,u,p)}var Bv=L({stridedSlice_:V3});function U3(e){let t={x:_(e,"x","tan","float32")};return O.runKernel(po,t)}var Vv=L({tan_:U3});function Ke(e,t){mi(e);let n=lr(e,t);if(n.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");return vs(e,null,n,t)}function Aa(e,t,n){if(mi(e),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");let a=lr(e,n);if(a.length!==2&&a.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return vs(e,t,a,n)}function Da(e,t,n){if(mi(e),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");let a=lr(e,n);if(a.length!==4&&a.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return vs(e,t,a,n)}function AT(e,t,n){if(mi(e),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");let a=lr(e,n);if(a.length!==5&&a.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return vs(e,t,a,n)}function $T(e,t,n){if(mi(e),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");let a=lr(e,n);if(a.length!==6&&a.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(a.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||a,vs(e,t,a,n)}function G3(e,t=1,n=!0){let a=_(e,"x","topk");if(a.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");let r=a.shape[a.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>r)throw new Error(`'k' passed to topk() must be <= the last dimension (${r}) but got ${t}`);let s={x:a},i={k:t,sorted:n},[o,l]=O.runKernel(Fu,s,i);return{values:o,indices:l}}var Uv=L({topk_:G3});function H3(e,t=0,n=1,a,r){if(a!=null&&a==="bool")throw new Error("Unsupported data type $ { dtype }");let s=new Lv(t,n,a,!0,r),i=Oe(e,a);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var ff=L({truncatedNormal_:H3});function j3(e,t=0){let n=_(e,"x","unique","string_or_numeric");$(n.rank>0,()=>"The input tensor must be at least 1D");let a={x:n},r={axis:t},[s,i]=O.runKernel(Bm,a,r);return{values:s,indices:i}}var Gv=L({unique_:j3});function q3(e,t,n){let a=_(e,"x","unsortedSegmentSum"),r=_(t,"segmentIds","unsortedSegmentSum","int32");$(hl(n),()=>"numSegments must be of dtype int");let s={x:a,segmentIds:r},i={numSegments:n};return O.runKernel(Cc,s,i)}var gf=L({unsortedSegmentSum_:q3});function K3(e,t=0){let n=_(e,"x","unstack","string_or_numeric");$(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);let a={value:n},r={axis:t};return O.runKernel(Ru,a,r)}var ct=L({unstack_:K3});function FT(e,t){return af(e,t,"right")}function Hv(e,t=!0,n,a){return O.makeVariable(e,t,n,a)}function DT(e,t){let n=[];for(let s=0;s<t.length;s++)t[s]&&n.push(s);let a=Oe(e,"int32"),r=Oe([n.length,e.length],"int32");for(let s=0;s<n.length;s++){let i=a.indexToLoc(n[s]),o=s*e.length;r.values.set(i,o)}return r.toTensor()}async function X3(e){let t=_(e,"condition","whereAsync","bool"),n=await t.data(),a=DT(t.shape,n);return e!==t&&t.dispose(),a}var jv=X3;async function Y3(e,t,n){let a=_(e,"tensor","boolMask"),r=_(t,"mask","boolMask","bool"),s=n==null?0:n,i=r.rank,o=a.shape;$(i>0,()=>"mask cannot be scalar"),Tn(o.slice(s,s+i),r.shape,"mask's shape must match the first K dimensions of tensor's shape,");let l=1;for(let f=s;f<s+i;f++)l*=o[f];let u=o.slice(0,s).concat([l],o.slice(s+i)),p=W(a,u),d=W(r,[-1]),c=await jv(d),h=Ns(c,[1]),m=Bu(p,h,s);return e!==a&&a.dispose(),t!==r&&r.dispose(),h.dispose(),p.dispose(),d.dispose(),c.dispose(),m}var RT=Y3;function Z3(e,t,n,a,r=!0){let s=_(e,"v","movingAverage"),i=_(t,"x","movingAverage"),o=_(n,"decay","movingAverage");lS(s,i),$(fs(s.shape,i.shape),()=>"Shape mismatch in v and x");let l=be(1),u=pe(l,o),p=z(pe(i,s),u);if(r){$(a!=null,()=>"When using zeroDebias: true, step is required.");let d=_(a,"step","movingAverage");p=he(p,pe(l,$r(o,d)))}return Y(s,p)}var MT=L({movingAverage_:Z3});function J3(e,t,n){let a=_(e,"indices","scatterND","int32"),r=_(t,"updates","scatterND");jx(r,a,n);let s={indices:a,updates:r},i={shape:n};return O.runKernel(wu,s,i)}var PT=L({scatterND_:J3});function Q3(e,t,n,a){if(e.dtype!=="int32")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);let r=e.rank>0?e.shape[0]:1,s=e.rank>1?e.shape[1]:1;if(n.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${s}.`);let i=t.size;if(!(t.rank===0||t.rank===1&&i===r))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${r}]`);if(t.dtype!==a.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function eL(e,t,n,a=0){let r=_(e,"sparseIndices","sparseToDense","int32"),s=_(t,"sparseValues","sparseToDense","string_or_numeric"),i=_(a,"defaultValue","sparseToDense",s.dtype);Q3(r,s,n,i);let o={sparseIndices:r,sparseValues:s,defaultValue:i},l={outputShape:n};return O.runKernel(Wm,o,l)}var OT=L({sparseToDense_:eL});function tL(e,t){let n=_(t,"indices","gatherND","int32"),a={params:_(e,"x","gatherND","string_or_numeric"),indices:n};return O.runKernel(eu,a)}var LT=L({gatherND_:tL});function nL(e,t){if(t==null)return e.shape.slice();if(fs(e.shape,t))return t;if(e.shape.length===t.length){let n=[];for(let a=0;a<e.shape.length;a++)t[a]==null&&e.shape[a]!=null?n.push(e.shape[a]):n.push(t[a]);return n}return t}function aL(e,t,n,a){let r=_(e,"x","dropout");if($(r.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${r.dtype} tensor instead.`),$(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return e instanceof Te?r.clone():r;let s=nL(r,n),i=1-t,o=he(Wu(Y(Uu(s,0,1,"float32",a),i)),i);return z(r,o)}var qv=L({dropout_:aL});function Kv(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function yf(e,t,n){let a=1-e%2,r=new Float32Array(e);for(let s=0;s<e;++s){let i=2*Math.PI*s/(e+a-1);r[s]=t-n*Math.cos(i)}return Ke(r,"float32")}async function rL(e,t,n=1){let a=_(e,"predictions","inTopK"),r=_(t,"targets","inTopK");$(a.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${a.rank}`),$(a.rank-1===r.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${a.rank} and targets rank ${r.rank}`),Tn(a.shape.slice(0,a.shape.length-1),r.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let s=a.shape[a.shape.length-1];$(n>0&&n<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${n}`);let i=await a.data(),o=await r.data(),[l,u]=[i.length/s,s],p=HI("bool",l);for(let d=0;d<l;d++){let c=d*u,h=i.subarray(c,c+u),m=[];for(let f=0;f<h.length;f++)m.push({value:h[f],index:f});m.sort((f,g)=>g.value-f.value),p[d]=0;for(let f=0;f<n;f++)if(m[f].index===o[d]){p[d]=1;break}}return e!==a&&a.dispose(),t!==r&&r.dispose(),In(p,r.shape,"bool")}var zT=rL,kl={};Ae(kl,{conv2d:()=>oL,depthwiseConv2d:()=>cL,matMul:()=>hL});function sL(e,t,n,a,r,s="NHWC",i){let o=e;e.rank===3&&(o=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]])),$(o.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`),$(l.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${l.shape}.`),$(n.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);let u=s==="NHWC"?o.shape[3]:o.shape[1],p=s==="NHWC"?l.shape[3]:l.shape[1];$(u===n[2],()=>`Error in conv2dDerFilter: depth of input ${u}) must match input depth in filter (${n[2]}.`),$(p===n[3],()=>`Error in conv2dDerFilter: depth of dy (${p}) must match output depth for filter (${n[3]}).`),Nn("conv2dDerFilter",r,i);let d={x:o,dy:l},c={strides:a,pad:r,dataFormat:s,dimRoundingMode:i,filterShape:n};return O.runKernel(fm,d,c)}var Xv=L({conv2DBackpropFilter_:sL});function bf(e,t,n){if(n==null||n==="linear")return e;if(n==="relu")return z(e,go(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function xf(e,t){let n=t,a=Wt(e.shape,t.shape);return a.length>0&&(n=fe(n,a)),W(n,e.shape)}function vf(e,t,n,a){if(t==="linear")return e;if(t==="relu")return Xe(e);if(t==="elu")return Lu(e);if(t==="relu6")return sf(e);if(t==="prelu")return Uc(e,n);if(t==="leakyrelu")return Lc(e,a);if(t==="sigmoid")return ha(e);throw new Error(`Unknown fused activation ${t}.`)}var wf=(e,t)=>!(e>0)||t==="linear";function iL({x:e,filter:t,strides:n,pad:a,dataFormat:r="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:p}){if(l=l||"linear",wf(O.state.gradientDepth,l)===!1){$(r==="NHWC",()=>`Error in fused conv2d: got dataFormat of ${r} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let T=$t(e,t,n,a,r,s,i);return o!=null&&(T=Y(T,o)),vf(T,l,u,p)}let d=_(e,"x","conv2d","float32"),c=_(t,"filter","conv2d","float32"),h=d,m=!1;d.rank===3&&(m=!0,h=W(d,[1,d.shape[0],d.shape[1],d.shape[2]])),$(h.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${h.rank}.`),$(c.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${c.rank}.`),Nn("fused conv2d",a,i);let f=r==="NHWC"?h.shape[3]:h.shape[1];$(c.shape[2]===f,()=>`Error in conv2d: depth of input (${f}) must match input depth for filter ${c.shape[2]}.`),$(cr(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);let g=Mc(h.shape,c.shape,n,s,a,i),y;o!=null&&(y=_(o,"bias","fused conv2d"),[y]=_t(y,d),r==="NHWC"?ut(g.outShape,y.shape):($(y.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${y.shape.length}.`),$(y.shape.length===0||y.shape[0]===g.outChannels||y.shape[0]===1,()=>`Error in fused conv2d: bias shape (${y.shape}) is not compatible with the number of output channels (${g.outChannels})`)));let b;if(u!=null){let T=u.shape;if($(T.length<=1||T.length===3,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${T.length}.`),T.length===1)$(T[0]===1||T[0]===g.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${T}) is not compatible with the number of output channels (${g.outChannels}).`);else if(T.length===3)try{ut(T,g.outShape)}catch(C){let E=`Error in fused conv2d: PReLU activation weights (${T}) is not compatible with the output shape of the conv2d (${g.outShape}).`;throw Error(E)}b=_(u,"prelu weights","fused conv2d")}let x=(T,C)=>{$(r==="NHWC",()=>`Error in gradient of fused conv2D: got dataFormat of ${r} but only NHWC is currently supported.`);let[E,A,R,F]=C,S=bf(T,R,l);$(os(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let M=mv(A.shape,S,E,n,a),B=Xv(A,S,E.shape,n,a),U=[M,B];if(F!=null){let G=xf(F,S);U.push(G)}return U},w={x:h,filter:c,bias:y,preluActivationWeights:b},I={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:p};return o==null?ur((T,C,E)=>{let A=O.runKernel(ei,w,I);return E([C,T,A]),m&&(A=W(A,[A.shape[1],A.shape[2],A.shape[3]])),{value:A,gradFunc:x}})(h,c):ur((T,C,E,A)=>{let R=O.runKernel(ei,w,I);return A([C,T,R,E]),m&&(R=W(R,[R.shape[1],R.shape[2],R.shape[3]])),{value:R,gradFunc:x}})(h,c,y)}var oL=L({fusedConv2d_:iL});function lL(e,t,n,a,r,s=[1,1],i){let o=e;e.rank===3&&(o=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={x:o,dy:l},p={strides:a,pad:r,dimRoundingMode:i,dilations:s,filterShape:n};return O.runKernel(xm,u,p)}var WT=L({depthwiseConv2dNativeBackpropFilter_:lL});function uL(e,t,n,a,r,s=[1,1],i){let o=t,l=!1;t.rank===3&&(l=!0,o=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={dy:o,filter:n},p={strides:a,pad:r,dimRoundingMode:i,dilations:s,inputShape:e},d=O.runKernel(vm,u,p);return l?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var BT=L({depthwiseConv2dNativeBackpropInput_:uL});function pL({x:e,filter:t,strides:n,pad:a,dataFormat:r="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:p}){if(wf(O.state.gradientDepth,l)===!1){let I=Is(e,t,n,a,r,s,i);return o!=null&&(I=Y(I,o)),vf(I,l,u,p)}let d=_(e,"x","depthwiseConv2d","float32"),c=_(t,"filter","depthwiseConv2d","float32"),h=d,m=!1;d.rank===3&&(m=!0,h=W(d,[1,d.shape[0],d.shape[1],d.shape[2]])),$(h.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${h.rank}.`),$(c.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${c.rank}.`),$(h.shape[3]===c.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${h.shape[3]}) must match the inChannels dimension in filter ${c.shape[2]}.`),s==null&&(s=[1,1]),$(cr(n,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),Nn("fused depthwiseConv2d",a,i);let f=Mc(h.shape,c.shape,n,s,a,i,!0),g;o!=null&&(g=_(o,"bias","fused conv2d"),[g]=_t(g,d),ut(f.outShape,g.shape));let y;u!=null&&(y=_(u,"prelu weights","fused depthwiseConv2d"));let b=(I,T)=>{$(os(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[C,E,A,R]=T,F=bf(I,A,l),S=BT(E.shape,F,C,n,a,s,i),M=WT(E,F,C.shape,n,a,s,i);if(R!=null){let B=xf(g,F);return[S,M,B]}return[S,M]},x={x:h,filter:c,bias:g,preluActivationWeights:y},w={strides:n,pad:a,dataFormat:r,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:p};return o==null?ur((I,T,C)=>{let E=O.runKernel(ti,x,w);return C([T,I,E]),m&&(E=W(E,[E.shape[1],E.shape[2],E.shape[3]])),{value:E,gradFunc:b}})(h,c):ur((I,T,C,E)=>{let A=O.runKernel(ti,x,w);return E([T,I,A,C]),m&&(A=W(A,[A.shape[1],A.shape[2],A.shape[3]])),{value:A,gradFunc:b}})(h,c,g)}var cL=L({fusedDepthwiseConv2d_:pL});function dL({a:e,b:t,transposeA:n=!1,transposeB:a=!1,bias:r,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:o=.2}){if(wf(O.state.gradientDepth,s)===!1){let R=Fe(e,t,n,a);return r!=null&&(R=Y(R,r)),vf(R,s,i,o)}let l=_(e,"a","fused matMul"),u=_(t,"b","fused matMul");[l,u]=_t(l,u);let p=n?l.shape[l.rank-2]:l.shape[l.rank-1],d=a?u.shape[u.rank-1]:u.shape[u.rank-2],c=n?l.shape[l.rank-1]:l.shape[l.rank-2],h=a?u.shape[u.rank-2]:u.shape[u.rank-1],m=l.shape.slice(0,-2),f=u.shape.slice(0,-2),g=mt(m),y=mt(f);$(p===d,()=>`Error in fused matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${l.shape} and ${u.shape} and transposeA=${n} and transposeB=${a} must match.`);let b=ut(l.shape.slice(0,-2),u.shape.slice(0,-2)).concat([c,h]),x=n?W(l,[g,p,c]):W(l,[g,c,p]),w=a?W(u,[y,h,d]):W(u,[y,d,h]),I;r!=null&&(I=_(r,"bias","fused matMul"),[I]=_t(I,l),ut(b,I.shape));let T;i!=null&&(T=_(i,"prelu weights","fused matMul"));let C=(R,F)=>{let[S,M,B,U]=F,G=bf(W(R,B.shape),B,s),q,K;if(!n&&!a?(q=Fe(G,M,!1,!0),K=Fe(S,G,!0,!1)):!n&&a?(q=Fe(G,M,!1,!1),K=Fe(G,S,!0,!1)):n&&!a?(q=Fe(M,G,!1,!0),K=Fe(S,G,!1,!1)):(q=Fe(M,G,!0,!0),K=Fe(G,S,!0,!0)),r!=null){let Z=xf(U,G);return[q,K,Z]}else return[q,K]},E={a:x,b:w,bias:I,preluActivationWeights:T},A={transposeA:n,transposeB:a,activation:s,leakyreluAlpha:o};return r==null?ur((R,F,S)=>{let M=O.runKernel(Qs,E,A);return S([R,F,M]),{value:W(M,b),gradFunc:C}})(x,w):ur((R,F,S,M)=>{let B=O.runKernel(Qs,E,A);return M([R,F,B,S]),{value:W(B,b),gradFunc:C}})(x,w,I)}var hL=L({fusedMatMul_:dL});function mL(e){return yf(e,.54,.46)}var fL=L({hammingWindow_:mL});function gL(e){return yf(e,.5,.5)}var VT=L({hannWindow_:gL});function yL(e,t,n,a=!1,r=0){let s=0,i=[];for(;s+t<=e.size;)i.push(Be(e,s,t)),s+=n;if(a)for(;s<e.size;){let o=s+t-e.size,l=Ze([Be(e,s,t-o),gn([o],r)]);i.push(l),s+=n}return i.length===0?Aa([],[0,t]):W(Ze(i),[i.length,t])}var UT=L({frame_:yL});function bL(e,t,n,a,r=VT){a==null&&(a=Kv(t));let s=UT(e,t,n),i=z(s,r(t));return jc(i,a)}var xL=L({stft_:bL});function vL(e,t,n,a,r="bilinear",s=0){let i=_(e,"image","cropAndResize"),o=_(t,"boxes","cropAndResize","float32"),l=_(n,"boxInd","cropAndResize","int32"),u=o.shape[0];$(i.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),$(o.rank===2&&o.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${u},4] but had shape ${o.shape}.`),$(l.rank===1&&l.shape[0]===u,()=>`Error in cropAndResize: boxInd must be have size [${u}] but had shape ${o.shape}.`),$(a.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${a.length}.`),$(a[0]>=1&&a[1]>=1,()=>`cropSize must be atleast [1,1], but was ${a}`),$(r==="bilinear"||r==="nearest",()=>`method must be bilinear or nearest, but was ${r}`);let p={image:i,boxes:o,boxInd:l},d={method:r,extrapolationValue:s,cropSize:a};return O.runKernel(jl,p,d)}var wL=L({cropAndResize_:vL});function kL(e){let t=_(e,"image","flipLeftRight","float32");$(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let n={image:t};return O.runKernel(Jl,n,{})}var IL=L({flipLeftRight_:kL});function SL(e){let t=_(e,"image","grayscaleToRGB"),n=t.rank-1,a=t.shape[n];$(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),$(a===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${a}.`);let r=new Array(t.rank);return r.fill(1,0,n),r[n]=3,Ln(t,r)}var TL=L({grayscaleToRGB_:SL});function NL(e,t,n=0,a=.5){let r=_(e,"image","rotateWithOffset","float32");$(r.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${r.rank}.`);let s={image:r},i={radians:t,fillValue:n,center:a};return O.runKernel(Pu,s,i)}var CL=L({rotateWithOffset_:NL});function Gu(e,t,n,a,r,s){a==null&&(a=.5),r==null&&(r=Number.NEGATIVE_INFINITY),s==null&&(s=0);let i=e.shape[0];return n=Math.min(n,i),$(0<=a&&a<=1,()=>`iouThreshold must be in [0, 1], but was '${a}'`),$(e.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),$(e.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),$(t.rank===1,()=>"scores must be a 1D tensor"),$(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),$(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:n,iouThreshold:a,scoreThreshold:r,softNmsSigma:s}}function _L(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY){let s=_(e,"boxes","nonMaxSuppression","float32"),i=_(t,"scores","nonMaxSuppression","float32"),o=Gu(s,i,n,a,r);n=o.maxOutputSize,a=o.iouThreshold,r=o.scoreThreshold;let l={maxOutputSize:n,iouThreshold:a,scoreThreshold:r};return O.runKernel(mu,{boxes:s,scores:i},l)}var EL=L({nonMaxSuppression_:_L});function AL(e,t,n){let a=$L(e,t,n),r=a<0?-(a+1):a;e.splice(r,0,t)}function $L(e,t,n){return DL(e,t,n||FL)}function FL(e,t){return e>t?1:e<t?-1:0}function DL(e,t,n){let a=0,r=e.length,s=0,i=!1;for(;a<r;){s=a+(r-a>>>1);let o=n(t,e[s]);o>0?a=s+1:(r=s,i=!o)}return i?a:-a-1}function GT(e,t,n,a,r){return Yv(e,t,n,a,r,0)}function HT(e,t,n,a,r,s){return Yv(e,t,n,a,r,0,!1,s,!0)}function jT(e,t,n,a,r,s){return Yv(e,t,n,a,r,s,!0)}function Yv(e,t,n,a,r,s,i=!1,o=!1,l=!1){let u=[];for(let g=0;g<t.length;g++)t[g]>r&&u.push({score:t[g],boxIndex:g,suppressBeginIndex:0});u.sort(bk);let p=s>0?-.5/s:0,d=[],c=[];for(;d.length<n&&u.length>0;){let g=u.pop(),{score:y,boxIndex:b,suppressBeginIndex:x}=g;if(y<r)break;let w=!1;for(let I=d.length-1;I>=x;--I){let T=RL(e,b,d[I]);if(T>=a){w=!0;break}if(g.score=g.score*ML(a,p,T),g.score<=r)break}g.suppressBeginIndex=d.length,w||(g.score===y?(d.push(b),c.push(g.score)):g.score>r&&AL(u,g,bk))}let h=d.length,m=n-h;o&&m>0&&(d.push(...new Array(m).fill(0)),c.push(...new Array(m).fill(0)));let f={selectedIndices:d};return i&&(f.selectedScores=c),l&&(f.validOutputs=h),f}function RL(e,t,n){let a=e.subarray(t*4,t*4+4),r=e.subarray(n*4,n*4+4),s=Math.min(a[0],a[2]),i=Math.min(a[1],a[3]),o=Math.max(a[0],a[2]),l=Math.max(a[1],a[3]),u=Math.min(r[0],r[2]),p=Math.min(r[1],r[3]),d=Math.max(r[0],r[2]),c=Math.max(r[1],r[3]),h=(o-s)*(l-i),m=(d-u)*(c-p);if(h<=0||m<=0)return 0;let f=Math.max(s,u),g=Math.max(i,p),y=Math.min(o,d),b=Math.min(l,c),x=Math.max(y-f,0)*Math.max(b-g,0);return x/(h+m-x)}function ML(e,t,n){let a=Math.exp(t*n*n);return n<=e?a:0}function bk(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}async function PL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY){let s=_(e,"boxes","nonMaxSuppressionAsync"),i=_(t,"scores","nonMaxSuppressionAsync"),o=Gu(s,i,n,a,r);n=o.maxOutputSize,a=o.iouThreshold,r=o.scoreThreshold;let l=await Promise.all([s.data(),i.data()]),u=l[0],p=l[1],{selectedIndices:d}=GT(u,p,n,a,r);return s!==e&&s.dispose(),i!==t&&i.dispose(),Ke(d,"int32")}var OL=PL;function LL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=0){let i=_(e,"boxes","nonMaxSuppression"),o=_(t,"scores","nonMaxSuppression"),l=Gu(i,o,n,a,r,s);n=l.maxOutputSize,a=l.iouThreshold,r=l.scoreThreshold,s=l.softNmsSigma;let u={boxes:i,scores:o},p={maxOutputSize:n,iouThreshold:a,scoreThreshold:r,softNmsSigma:s},d=O.runKernel(gu,u,p);return{selectedIndices:d[0],selectedScores:d[1]}}var zL=L({nonMaxSuppressionWithScore_:LL});async function WL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=0){let i=_(e,"boxes","nonMaxSuppressionAsync"),o=_(t,"scores","nonMaxSuppressionAsync"),l=Gu(i,o,n,a,r,s);n=l.maxOutputSize,a=l.iouThreshold,r=l.scoreThreshold,s=l.softNmsSigma;let u=await Promise.all([i.data(),o.data()]),p=u[0],d=u[1],{selectedIndices:c,selectedScores:h}=jT(p,d,n,a,r,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Ke(c,"int32"),selectedScores:Ke(h)}}var BL=WL;function VL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=!1){let i=_(e,"boxes","nonMaxSuppression"),o=_(t,"scores","nonMaxSuppression"),l=Gu(i,o,n,a,r,null),u=l.maxOutputSize,p=l.iouThreshold,d=l.scoreThreshold,c={boxes:i,scores:o},h={maxOutputSize:u,iouThreshold:p,scoreThreshold:d,padToMaxOutputSize:s},m=O.runKernel(fu,c,h);return{selectedIndices:m[0],validOutputs:m[1]}}var UL=L({nonMaxSuppressionPadded_:VL});async function GL(e,t,n,a=.5,r=Number.NEGATIVE_INFINITY,s=!1){let i=_(e,"boxes","nonMaxSuppressionAsync"),o=_(t,"scores","nonMaxSuppressionAsync"),l=Gu(i,o,n,a,r,null),u=l.maxOutputSize,p=l.iouThreshold,d=l.scoreThreshold,[c,h]=await Promise.all([i.data(),o.data()]),{selectedIndices:m,validOutputs:f}=HT(c,h,u,p,d,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Ke(m,"int32"),validOutputs:be(f,"int32")}}var HL=GL;function jL(e,t,n=!1,a=!1){let r=_(e,"images","resizeBilinear");$(r.rank===3||r.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${r.rank}.`),$(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),$(a===!1||n===!1,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=r,i=!1;r.rank===3&&(i=!0,s=W(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:a,size:t},u=O.runKernel(Ji,o,l);return i?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var qT=L({resizeBilinear_:jL});function qL(e,t,n=!1,a=!1){let r=_(e,"images","resizeNearestNeighbor");$(r.rank===3||r.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${r.rank}.`),$(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),$(r.dtype==="float32"||r.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype"),$(a===!1||n===!1,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=r,i=!1;r.rank===3&&(i=!0,s=W(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:a,size:t},u=O.runKernel(Zi,o,l);return i?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var KT=L({resizeNearestNeighbor_:qL});function KL(e,t="binary",n=!1,a=.5){let r=_(e,"image","threshold"),s=.2989,i=.587,o=.114,l=r.shape[0]*r.shape[1],u=z(Ke([a]),255),p,d,c,h;if($(r.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${r.rank}.`),$(r.shape[2]===3||r.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${r.shape[2]}.`),$(r.dtype==="int32"||r.dtype==="float32",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${r.dtype}.`),$(t==="otsu"||t==="binary",()=>`Method must be binary or otsu, but was ${t}`),r.shape[2]===3){[p,d,c]=zn(r,[1,1,1],-1);let f=z(p,s),g=z(d,i),y=z(c,o);h=Y(Y(f,g),y)}else h=e;if(t==="otsu"){let f=lv(oe(of(h),"int32"),In([]),256);u=XL(f,l)}let m=n?Ss(h,u):Cn(h,u);return oe(z(m,255),"int32")}function XL(e,t){let n=Ke([-1]),a=Ke([0]),r=Ke([0]),s,i,o,l,u,p;for(let d=0;d<e.size-1;d++){s=Be(e,0,d+1),i=Be(e,d+1),u=he(fe(s),t),p=he(fe(i),t);let c=fe(z(s,xl(0,s.size)));o=he(c,fe(s));let h=gn(i.shape,s.size),m=Y(xl(0,i.size),h),f=z(i,m);l=he(fe(f),fe(i));let g=pe(o,l),y=pe(o,l),b=z(u,p);r=z(z(b,g),y);let x=Cn(r,a);a=mn(x,r,a),n=mn(x,Ke([d]),n)}return n}var YL=L({threshold_:KL});function ZL(e,t,n="nearest",a="constant",r=0,s){let i=_(e,"image","transform","float32"),o=_(t,"transforms","transform","float32");$(i.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),$(o.rank===2&&(o.shape[0]===i.shape[0]||o.shape[0]===1)&&o.shape[1]===8,()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),$(s==null||s.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let l={image:i,transforms:o},u={interpolation:n,fillMode:a,fillValue:r,outputShape:s};return O.runKernel(Du,l,u)}var JL=L({transform_:ZL});function QL(e,t,n){$(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),$(n%1===0,()=>`bandPart(): numUpper must be an integer, got ${n}.`);let a=_(e,"a","bandPart");$(a.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${a.rank}.`);let r=a.shape,[s,i]=a.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(n<=i))throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),n<0&&(n=i);let o=W(xl(0,s,1,"int32"),[-1,1]),l=xl(0,i,1,"int32"),u=pe(o,l),p=$a(Ss(u,be(+t,"int32")),Fr(u,be(-n,"int32"))),d=It([s,i],a.dtype);return W(Ft(ct(W(a,[-1,s,i])).map(c=>mn(p,c,d))),r)}var ez=L({bandPart_:QL});function tz(e){let t;if(Array.isArray(e)){t=!1,$(e!=null&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let r=e[0].shape[0];for(let s=1;s<e.length;++s)$(e[s].shape[0]===r,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[s].shape[0]} vs. ${r})`)}else t=!0,e=zn(e,e.shape[0],0).map(r=>Ns(r,[0]));$(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);let n=[],a=e;for(let r=0;r<e.length;++r)n.push(O.tidy(()=>{let s=a[r];if(r>0)for(let i=0;i<r;++i){let o=z(fe(z(n[i],s)),n[i]);s=pe(s,o)}return he(s,zu(s,"euclidean"))}));return t?Ft(n,0):n}var nz=L({gramSchmidt_:tz});function az(e,t=!1){if($(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),e.rank===2)return xk(e,t);{let n=e.shape.slice(0,e.shape.length-2).reduce((l,u)=>l*u),a=ct(W(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),r=[],s=[];a.forEach(l=>{let[u,p]=xk(l,t);r.push(u),s.push(p)});let i=W(Ft(r,0),e.shape),o=W(Ft(s,0),e.shape);return[i,o]}}function xk(e,t=!1){return O.tidy(()=>{$(e.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);let n=e.shape[0],a=e.shape[1],r=Zm(n),s=ir(e),i=Aa([[1]],[1,1]),o=ir(i),l=n>=a?a:n;for(let u=0;u<l;++u){let p=s,d=o,c=r;[o,s,r]=O.tidy(()=>{let h=Be(s,[u,u],[n-u,1]),m=zu(h),f=Be(s,[u,u],[1,1]),g=mn(Cn(f,0),Aa([[-1]]),Aa([[1]])),y=pe(f,z(g,m)),b=he(h,y);b.shape[0]===1?o=ir(i):o=Ze([i,Be(b,[1,0],[b.shape[0]-1,b.shape[1]])],0);let x=vt(he(Fe(g,y),m)),w=Be(s,[u,0],[n-u,a]),I=z(x,o),T=Ee(o);if(u===0)s=pe(w,Fe(I,Fe(T,w)));else{let A=pe(w,Fe(I,Fe(T,w)));s=Ze([Be(s,[0,0],[u,a]),A],0)}let C=Ee(I),E=Be(r,[0,u],[n,r.shape[1]-u]);if(u===0)r=pe(E,Fe(Fe(E,o),C));else{let A=pe(E,Fe(Fe(E,o),C));r=Ze([Be(r,[0,0],[n,u]),A],1)}return[o,s,r]}),_e([p,d,c])}return!t&&n>a&&(r=Be(r,[0,0],[n,a]),s=Be(s,[0,0],[a,a])),[r,s]})}var rz=L({qr_:az}),vn;(function(e){e[e.NONE=0]="NONE",e[e.MEAN=1]="MEAN",e[e.SUM=2]="SUM",e[e.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS"})(vn||(vn={}));function sz(e,t,n=vn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"losses","computeWeightedLoss"),r=null;t!=null&&(r=_(t,"weights","computeWeightedLoss"));let s=r==null?a:z(a,r);if(n===vn.NONE)return s;if(n===vn.SUM)return fe(s);if(n===vn.MEAN){if(r==null)return Nt(s);{let i=a.size/r.size,o=he(fe(s),fe(r));return i>1?he(o,be(i)):o}}if(n===vn.SUM_BY_NONZERO_WEIGHTS){if(r==null)return he(fe(s),be(a.size));{let i=z(r,Jn(a.shape)),o=oe(fe(oi(i,be(0))),"float32");return he(fe(s),o)}}throw Error(`Unknown reduction: ${n}`)}var Dr=L({computeWeightedLoss_:sz});function iz(e,t,n,a=vn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","absoluteDifference"),s=_(t,"predictions","absoluteDifference"),i=null;n!=null&&(i=_(n,"weights","absoluteDifference")),Tn(r.shape,s.shape,"Error in absoluteDifference: ");let o=Lt(pe(r,s));return Dr(o,i,a)}var oz=L({absoluteDifference_:iz});function lz(e,t,n,a,r=vn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","cosineDistance"),i=_(t,"predictions","cosineDistance"),o=null;a!=null&&(o=_(a,"weights","cosineDistance")),Tn(s.shape,i.shape,"Error in cosineDistance: ");let l=be(1),u=pe(l,fe(z(s,i),n,!0));return Dr(u,o,r)}var uz=L({cosineDistance_:lz});function pz(e,t,n,a=vn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","hingeLoss"),s=_(t,"predictions","hingeLoss"),i=null;n!=null&&(i=_(n,"weights","hingeLoss")),Tn(r.shape,s.shape,"Error in hingeLoss: ");let o=be(1);r=pe(z(be(2),r),o);let l=Xe(pe(o,z(r,s)));return Dr(l,i,a)}var cz=L({hingeLoss_:pz});function dz(e,t,n,a=1,r=vn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","huberLoss"),i=_(t,"predictions","huberLoss"),o=null;n!=null&&(o=_(n,"weights","huberLoss")),Tn(s.shape,i.shape,"Error in huberLoss: ");let l=be(a),u=Lt(pe(i,s)),p=Vu(u,l),d=pe(u,p),c=Y(z(be(.5),lt(p)),z(l,d));return Dr(c,o,r)}var hz=L({huberLoss_:dz});function mz(e,t,n,a=1e-7,r=vn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","logLoss"),i=_(t,"predictions","logLoss"),o=null;n!=null&&(o=_(n,"weights","logLoss")),Tn(s.shape,i.shape,"Error in logLoss: ");let l=be(1),u=be(a),p=vt(z(s,ta(Y(i,u)))),d=z(pe(l,s),ta(Y(pe(l,i),u))),c=pe(p,d);return Dr(c,o,r)}var fz=L({logLoss_:mz});function gz(e,t,n,a=vn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"labels","meanSquaredError"),s=_(t,"predictions","meanSquaredError"),i=null;n!=null&&(i=_(n,"weights","meanSquaredError")),Tn(r.shape,s.shape,"Error in meanSquaredError: ");let o=mf(r,s);return Dr(o,i,a)}var yz=L({meanSquaredError_:gz});function bz(e,t){let n=_(e,"labels","sigmoidCrossEntropyWithLogits"),a=_(t,"logits","sigmoidCrossEntropyWithLogits");Tn(n.shape,a.shape,"Error in sigmoidCrossEntropyWithLogits: ");let r=Xe(a),s=z(a,n),i=zc(fn(vt(Lt(a))));return Y(pe(r,s),i)}function xz(e,t,n,a=0,r=vn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"multiClassLabels","sigmoidCrossEntropy"),i=_(t,"logits","sigmoidCrossEntropy"),o=null;if(n!=null&&(o=_(n,"weights","sigmoidCrossEntropy")),Tn(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),a>0){let u=be(a),p=be(1),d=be(.5);s=Y(z(s,pe(p,u)),z(d,u))}let l=bz(s,i);return Dr(l,o,r)}var vz=L({sigmoidCrossEntropy_:xz});function wz(e,t,n=-1){if(n===-1&&(n=t.rank-1),n!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${n}`);return ur((a,r,s)=>{let i=tf(r,[n],!0),o=pe(oe(r,"float32"),i);s([a,o]);let l=vt(z(o,a));return{value:fe(l,[n]),gradFunc:(u,p)=>{let[d,c]=p,h=ii(u.shape,[n]);return[z(W(u,h),pe(oe(d,"float32"),fn(c))),z(W(u,h),pe(fn(c),oe(d,"float32")))]}}})(e,t)}function kz(e,t,n,a=0,r=vn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"onehotLabels","softmaxCrossEntropy"),i=_(t,"logits","softmaxCrossEntropy"),o=null;if(n!=null&&(o=_(n,"weights","softmaxCrossEntropy")),Tn(s.shape,i.shape,"Error in softmaxCrossEntropy: "),a>0){let u=be(a),p=be(1),d=be(s.shape[1]);s=Y(z(s,pe(p,u)),he(u,d))}let l=wz(s,i);return Dr(l,o,r)}var Iz=L({softmaxCrossEntropy_:kz});function Sz(e,t,n,a){let r=_(e,"indices","sparseFillEmptyRows","int32"),s=_(t,"values","sparseFillEmptyRows"),i=_(n,"denseShape","sparseFillEmptyRows","int32"),o=_(a,"defaultValue","sparseFillEmptyRows",s.dtype);if(r.rank!==2)throw new Error(`Indices should be Tensor2D but received shape
-        ${r.shape}`);if(s.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(i.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(o.rank!==0)throw new Error(`Default value should be a scalar but received shape ${o.shape}`);let l={indices:r,values:s,denseShape:i,defaultValue:o},u=O.runKernel(vc,l);return{outputIndices:u[0],outputValues:u[1],emptyRowIndicator:u[2],reverseIndexMap:u[3]}}var Tz=L({sparseFillEmptyRows_:Sz});function Nz(e,t,n){let a=_(e,"inputIndices","sparseReshape","int32"),r=_(t,"inputShape","sparseReshape","int32"),s=_(n,"newShape","sparseReshape","int32");if(a.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape
-        ${a.shape}`);if(r.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${r.shape}`);if(s.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:a,inputShape:r,newShape:s},o=O.runKernel(Au,i);return{outputIndices:o[0],outputShape:o[1]}}var Cz=L({sparseReshape_:Nz});function _z(e,t,n){let a=_(e,"data","sparseSegmentMean"),r=_(t,"indices","sparseSegmentMean","int32"),s=_(n,"segmentIds","sparseSegmentMean","int32");if(a.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-          ${r.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-          ${s.shape}`);let i={data:a,indices:r,segmentIds:s};return O.runKernel(wc,i)}var Ez=L({sparseSegmentMean_:_z});function Az(e,t,n){let a=_(e,"data","sparseSegmentSum"),r=_(t,"indices","sparseSegmentSum","int32"),s=_(n,"segmentIds","sparseSegmentSum","int32");if(a.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-         ${r.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-         ${s.shape}`);let i={data:a,indices:r,segmentIds:s};return O.runKernel(kc,i)}var $z=L({sparseSegmentSum_:Az});function Fz(e,t,n,a,r,s,i,o){let l=_(e,"data","stringNGrams","string");if(l.dtype!=="string")throw new Error("Data must be of datatype string");if(l.shape.length!==1)throw new Error(`Data must be a vector, saw: ${l.shape}`);let u=_(t,"dataSplits","stringNGrams");if(u.dtype!=="int32")throw new Error("Data splits must be of datatype int32");let p={separator:n,nGramWidths:a,leftPad:r,rightPad:s,padWidth:i,preserveShortSequences:o},d={data:l,dataSplits:u},c=O.runKernel(Sc,d,p);return{nGrams:c[0],nGramsSplits:c[1]}}var Dz=L({stringNGrams_:Fz});function Rz(e,t,n=!0){let a=_(e,"input","stringSplit","string"),r=_(t,"delimiter","stringSplit","string");if(a.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${a.shape}`);if(r.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${r.shape}`);let s={skipEmpty:n},i={input:a,delimiter:r},o=O.runKernel(Tc,i,s);return{indices:o[0],values:o[1],shape:o[2]}}var Mz=L({stringSplit_:Rz});function Pz(e,t){let n=_(e,"input","stringToHashBucketFast","string"),a={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let r={input:n};return O.runKernel(Nc,r,a)}var Oz=L({stringToHashBucketFast_:Pz}),XT={fft:Hc,ifft:wl,rfft:jc,irfft:hf},YT={hammingWindow:fL,hannWindow:VT,frame:UT,stft:xL},za={flipLeftRight:IL,grayscaleToRGB:TL,resizeNearestNeighbor:KT,resizeBilinear:qT,rotateWithOffset:CL,cropAndResize:wL,nonMaxSuppression:EL,nonMaxSuppressionAsync:OL,nonMaxSuppressionWithScore:zL,nonMaxSuppressionWithScoreAsync:BL,nonMaxSuppressionPadded:UL,nonMaxSuppressionPaddedAsync:HL,threshold:YL,transform:JL},Zv={bandPart:ez,gramSchmidt:nz,qr:rz},ZT={absoluteDifference:oz,computeWeightedLoss:Dr,cosineDistance:uz,hingeLoss:cz,huberLoss:hz,logLoss:fz,meanSquaredError:yz,sigmoidCrossEntropy:vz,softmaxCrossEntropy:Iz},JT={sparseFillEmptyRows:Tz,sparseReshape:Cz,sparseSegmentMean:Ez,sparseSegmentSum:$z},QT={stringNGrams:Dz,stringSplit:Mz,stringToHashBucketFast:Oz},Rr=class extends GS{minimize(e,t=!1,n){let{value:a,grads:r}=this.computeGradients(e,n);if(n!=null){let s=n.map(i=>({name:i.name,tensor:r[i.name]}));this.applyGradients(s)}else this.applyGradients(r);return _e(r),t?a:(a.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return lT(e,t)}dispose(){this.iterations_!=null&&_e(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:be(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(e){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty(Rr,Symbol.hasInstance,{value:e=>e.minimize!=null&&e.computeGradients!=null&&e.applyGradients!=null});var kf=class extends Rr{constructor(e,t,n=null){super(),this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t],r=!1;this.accumulatedGrads[n]==null&&(this.accumulatedGrads[n]={originalName:`${t}/accum_grad`,variable:P(()=>qe(a).variable(r))}),this.accumulatedUpdates[n]==null&&(this.accumulatedUpdates[n]={originalName:`${t}/accum_var`,variable:P(()=>qe(a).variable(r))});let s=Array.isArray(e)?e[n].tensor:e[t];if(s==null)return;let i=this.accumulatedGrads[n].variable,o=this.accumulatedUpdates[n].variable;P(()=>{let l=Y(z(i,this.rho),z(lt(s),1-this.rho)),u=z(he(un(Y(o,this.epsilon)),un(Y(i,this.epsilon))),s),p=Y(z(o,this.rho),z(lt(u),1-this.rho));i.assign(l),o.assign(p);let d=Y(z(u,-this.learningRate),a);a.assign(d)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(_e(this.accumulatedGrads.map(e=>e.variable)),_e(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){let e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=e.length/2,n=!1;this.accumulatedGrads=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedUpdates=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}};kf.className="Adadelta";ws(kf);var If=class extends Rr{constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t];this.accumulatedGrads[n]==null&&(this.accumulatedGrads[n]={originalName:`${t}/accumulator`,variable:P(()=>gn(a.shape,this.initialAccumulatorValue).variable(!1))});let r=Array.isArray(e)?e[n].tensor:e[t];if(r==null)return;let s=this.accumulatedGrads[n].variable;P(()=>{let i=Y(s,lt(r));s.assign(i);let o=Y(z(he(r,un(Y(i,O.backend.epsilon()))),-this.learningRate),a);a.assign(o)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&_e(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulatedGrads=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}};If.className="Adagrad";ws(If);var Sf=class extends Rr{constructor(e,t,n,a=null){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=a,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],P(()=>{this.accBeta1=be(t).variable(),this.accBeta2=be(n).variable()}),a==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);P(()=>{let n=pe(1,this.accBeta1),a=pe(1,this.accBeta2);t.forEach((r,s)=>{let i=O.registeredVariables[r],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${r}/m`,variable:P(()=>qe(i).variable(o))}),this.accumulatedSecondMoment[s]==null&&(this.accumulatedSecondMoment[s]={originalName:`${r}/v`,variable:P(()=>qe(i).variable(o))});let l=Array.isArray(e)?e[s].tensor:e[r];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,p=this.accumulatedSecondMoment[s].variable,d=Y(z(u,this.beta1),z(l,1-this.beta1)),c=Y(z(p,this.beta2),z(lt(l),1-this.beta2)),h=he(d,n),m=he(c,a);u.assign(d),p.assign(c);let f=Y(z(he(h,Y(un(m),this.epsilon)),-this.learningRate),i);i.assign(f)}),this.accBeta1.assign(z(this.accBeta1,this.beta1)),this.accBeta2.assign(z(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedSecondMoment!=null&&_e(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e),P(()=>{this.accBeta1.assign($r(this.beta1,this.iterations_+1)),this.accBeta2.assign($r(this.beta2,this.iterations_+1))});let t=e.length/2,n=!1;this.accumulatedFirstMoment=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedSecondMoment=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon)}};Sf.className="Adam";ws(Sf);var Tf=class extends Rr{constructor(e,t,n,a=null,r=0){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=a,this.decay=r,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],P(()=>{this.iteration=be(0).variable(),this.accBeta1=be(t).variable()}),a==null&&(this.epsilon=O.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);P(()=>{let n=pe(1,this.accBeta1),a=he(-this.learningRate,Y(z(this.iteration,this.decay),1));t.forEach((r,s)=>{let i=O.registeredVariables[r],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${r}/m`,variable:qe(i).variable(o)}),this.accumulatedWeightedInfNorm[s]==null&&(this.accumulatedWeightedInfNorm[s]={originalName:`${r}/v`,variable:qe(i).variable(o)});let l=Array.isArray(e)?e[s].tensor:e[r];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,p=this.accumulatedWeightedInfNorm[s].variable,d=Y(z(u,this.beta1),z(l,1-this.beta1)),c=z(p,this.beta2),h=Lt(l),m=dr(c,h);u.assign(d),p.assign(m);let f=Y(z(he(a,n),he(d,Y(m,this.epsilon))),i);i.assign(f)}),this.iteration.assign(Y(this.iteration,1)),this.accBeta1.assign(z(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedWeightedInfNorm!=null&&_e(this.accumulatedWeightedInfNorm.map(e=>e.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon,t.decay)}};Tf.className="Adamax";ws(Tf);var qc=class extends Rr{constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=Array.isArray(e)?e[n].tensor:e[t];if(a==null)return;let r=O.registeredVariables[t];P(()=>{let s=Y(z(this.c,a),r);r.assign(s)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,this.c!=null&&this.c.dispose(),this.c=Jt(be(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(e=await this.extractIterations(e),e.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}};qc.className="SGD";ws(qc);var Nf=class extends qc{constructor(e,t,n=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=be(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t];this.accumulations[n]==null&&(this.accumulations[n]={originalName:`${t}/momentum`,variable:P(()=>qe(a).variable(!1))});let r=this.accumulations[n].variable,s=Array.isArray(e)?e[n].tensor:e[t];s!=null&&P(()=>{let i,o=Y(z(this.m,r),s);this.useNesterov?i=Y(z(this.c,Y(s,z(o,this.m))),a):i=Y(z(this.c,o),a),r.assign(o),a.assign(i)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&_e(this.accumulations.map(e=>e.variable))}setMomentum(e){this.momentum=e}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulations=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}};Nf.className="Momentum";ws(Nf);var Cf=class extends Rr{constructor(e,t=.9,n=0,a=null,r=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=a,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=r,a==null&&(this.epsilon=O.backend.epsilon()),e==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,n)=>{let a=O.registeredVariables[t],r=!1;this.accumulatedMeanSquares[n]==null&&(this.accumulatedMeanSquares[n]={originalName:`${t}/rms`,variable:P(()=>qe(a).variable(r))}),this.accumulatedMoments[n]==null&&(this.accumulatedMoments[n]={originalName:`${t}/momentum`,variable:P(()=>qe(a).variable(r))}),this.accumulatedMeanGrads[n]==null&&this.centered&&(this.accumulatedMeanGrads[n]={originalName:`${t}/mg`,variable:P(()=>qe(a).variable(r))});let s=Array.isArray(e)?e[n].tensor:e[t];if(s==null)return;let i=this.accumulatedMeanSquares[n].variable,o=this.accumulatedMoments[n].variable;P(()=>{let l=Y(z(i,this.decay),z(lt(s),1-this.decay));if(this.centered){let u=this.accumulatedMeanGrads[n].variable,p=Y(z(u,this.decay),z(s,1-this.decay)),d=he(z(s,this.learningRate),un(pe(l,Y(lt(p),this.epsilon)))),c=Y(z(o,this.momentum),d);i.assign(l),u.assign(p),o.assign(c);let h=pe(a,c);a.assign(h)}else{let u=Y(z(i,this.decay),z(lt(s),1-this.decay)),p=Y(z(o,this.momentum),he(z(s,this.learningRate),un(Y(u,this.epsilon))));i.assign(u),o.assign(p);let d=pe(a,p);a.assign(d)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&_e(this.accumulatedMeanSquares.map(e=>e.variable)),this.accumulatedMeanGrads!=null&&this.centered&&_e(this.accumulatedMeanGrads.map(e=>e.variable)),this.accumulatedMoments!=null&&_e(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=this.centered?e.length/3:e.length/2,n=!1;this.accumulatedMeanSquares=e.slice(0,t).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.accumulatedMoments=e.slice(t,t*2).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=e.slice(t*2,t*3).map(a=>({originalName:a.name,variable:a.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}};Cf.className="RMSProp";ws(Cf);var Yr=class{static sgd(e){return new qc(e)}static momentum(e,t,n=!1){return new Nf(e,t,n)}static rmsprop(e,t=.9,n=0,a=null,r=!1){return new Cf(e,t,n,a,r)}static adam(e=.001,t=.9,n=.999,a=null){return new Sf(e,t,n,a)}static adadelta(e=.001,t=.95,n=null){return new kf(e,t,n)}static adamax(e=.002,t=.9,n=.999,a=null,r=0){return new Tf(e,t,n,a,r)}static adagrad(e,t=.1){return new If(e,t)}},Ws={sgd:Yr.sgd,momentum:Yr.momentum,adadelta:Yr.adadelta,adagrad:Yr.adagrad,rmsprop:Yr.rmsprop,adamax:Yr.adamax,adam:Yr.adam},Lz=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:e=>e())();function Jv(){return new Promise(e=>Lz(()=>e()))}var N={};Ae(N,{ERF_A1:()=>Qz,ERF_A2:()=>eW,ERF_A3:()=>tW,ERF_A4:()=>nW,ERF_A5:()=>aW,ERF_P:()=>Jz,PARALLELIZE_THRESHOLD:()=>Qv,RowPartitionType:()=>nr,SELU_SCALE:()=>tN,SELU_SCALEALPHA:()=>eN,applyActivation:()=>vf,assertAndGetBroadcastShape:()=>ut,assertAxesAreInnerMostDims:()=>UP,assertParamsConsistent:()=>zz,assignToTypedArray:()=>uW,axesAreInnerMostDims:()=>kv,calculateShapes:()=>RS,checkEinsumDimSizes:()=>fW,checkPadOnDimRoundingMode:()=>Nn,combineLocations:()=>aT,combineRaggedTensorToTensorShapes:()=>Bz,complexWithEvenIndex:()=>iW,complexWithOddIndex:()=>oW,computeConv2DInfo:()=>Mc,computeConv3DInfo:()=>YS,computeDefaultPad:()=>av,computeDilation2DInfo:()=>GM,computeOptimalWindowSize:()=>Hz,computeOutAndReduceShapes:()=>rT,computeOutShape:()=>Wz,computePool2DInfo:()=>XS,computePool3DInfo:()=>HM,convertConv2DDataFormat:()=>ZS,decodeEinsumEquation:()=>hW,eitherStridesOrDilationsAreOne:()=>cr,expandShapeToKeepDim:()=>ii,exponent:()=>cW,exponents:()=>pW,fromStringArrayToUint8:()=>PW,fromUint8ToStringArray:()=>MW,getAxesPermutation:()=>sT,getBroadcastDims:()=>$S,getComplexWithIndex:()=>lW,getEinsumComputePath:()=>gW,getEinsumPermutation:()=>mW,getFusedBiasGradient:()=>xf,getFusedDyActivation:()=>bf,getImageCenter:()=>jz,getInnerMostAxes:()=>GP,getPermuted:()=>Kz,getRaggedRank:()=>Uz,getReductionAxes:()=>Wt,getReshaped:()=>qz,getReshapedPermuted:()=>Xz,getRowPartitionTypesHelper:()=>Vz,getSliceBeginCoords:()=>Yz,getSliceSize:()=>Zz,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>vW,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>wW,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>kW,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>TW,getSparseReshapeInputOutputMismatchErrorMessage:()=>CW,getSparseReshapeInputOutputMultipleErrorMessage:()=>NW,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>IW,getSparseReshapeNegativeOutputDimErrorMessage:()=>SW,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>$W,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>_W,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>EW,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>AW,getUndoAxesPermutation:()=>Iv,isIdentityPermutation:()=>yW,log:()=>oD,mergeRealAndImagArrays:()=>rW,prepareAndValidate:()=>DS,prepareSplitSize:()=>xW,segment_util:()=>nN,shouldFuse:()=>wf,slice_util:()=>jt,splitRealAndImagArrays:()=>sW,tupleValuesAreOne:()=>os,upcastType:()=>fa,validateDefaultValueShape:()=>Gz,validateInput:()=>jx,validateUpdateShape:()=>Hx,warn:()=>Jr});function zz(e,t){let n=e[0].length;e.forEach((r,s)=>{$(r.length===n,()=>`Error in concat${n}D: rank of tensors[${s}] must be the same as the rank of the rest (${n})`)}),$(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);let a=e[0];e.forEach((r,s)=>{for(let i=0;i<n;i++)$(i===t||r[i]===a[i],()=>`Error in concat${n}D: Shape of tensors[${s}] (${r}) does not match the shape of the rest (${a}) along the non-concatenated axis ${s}.`)})}function Wz(e,t){let n=e[0].slice();for(let a=1;a<e.length;a++)n[t]+=e[a][t];return n}var nr;(function(e){e[e.FIRST_DIM_SIZE=0]="FIRST_DIM_SIZE",e[e.VALUE_ROWIDS=1]="VALUE_ROWIDS",e[e.ROW_LENGTHS=2]="ROW_LENGTHS",e[e.ROW_SPLITS=3]="ROW_SPLITS",e[e.ROW_LIMITS=4]="ROW_LIMITS",e[e.ROW_STARTS=5]="ROW_STARTS"})(nr||(nr={}));function Bz(e,t,n){let a=new Array;if(n==null&&t==null)return a;if(t==null)for(;a.length<e+n.length;)a.push(-1);else a=t.slice();if(n==null)return a;if(e+n.length!==a.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${e+n.length}, but shape.rank = ${a.length}`);for(let r=1;r<n.length;++r){let s=n[r],i=a[a.length-n.length+r],o=a[i];if(s>=0)if(o>=0){if(o!==s)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${r+e}] = ${s} but shape[${r+e}] = ${o}`)}else a[i]=s}return a}function Vz(e){let t={FIRST_DIM_SIZE:nr.FIRST_DIM_SIZE,VALUE_ROWIDS:nr.VALUE_ROWIDS,ROW_LENGTHS:nr.ROW_LENGTHS,ROW_SPLITS:nr.ROW_SPLITS,ROW_LIMITS:nr.ROW_LIMITS,ROW_STARTS:nr.ROW_STARTS},n=[];for(let a of e)if(a in t)n.push(t[a]);else break;return n}function Uz(e){return e.length===0?0:e[0]===nr.FIRST_DIM_SIZE?e.length-1:e.length}function Gz(e,t){if(e==null||t==null)return;let n=e.length,a=t.length;if(n>=a)throw new Error(`defaultValue.shape=${e} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${n} must be less than ragged tensor input flatValues.rank = ${a})`);for(let r=0;r<Math.min(n,a-1);++r){let s=e[r],i=t[r+1];if(s>=0&&i>=0&&s!==1&&s!==i)throw new Error(`defaultValue.shape=${e}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${r-e.length}] = ${s} but ragged tensor input.flatValues.shape[${r-e.length}] = ${i}`)}}var Qv=30;function Hz(e){return e<=Qv?e:Eh(e,Math.floor(Math.sqrt(e)))}function jz(e,t,n){let a=n*(typeof e=="number"?e:e[0]),r=t*(typeof e=="number"?e:e[1]);return[a,r]}function qz(e,t,n,a=!0){let r=[];if(a)r=r.concat(t.slice(0)),r.push(e[0]/n),r=r.concat(e.slice(1));else{r=r.concat(e[0]);let s=t.length;for(let i=0;i<s;++i)r=r.concat([e[i+1]/t[i],t[i]]);r=r.concat(e.slice(s+1))}return r}function Kz(e,t,n=!0){let a=[];if(n){a.push(t);for(let r=t+1;r<e;++r)r<=2*t?(a.push(r),a.push(r-(t+1))):a.push(r)}else{let r=[],s=[];for(let i=1;i<e;++i)i>=t*2+1||i%2===1?s.push(i):r.push(i);a.push(...r),a.push(0),a.push(...s)}return a}function Xz(e,t,n,a=!0){let r=[];a?r.push(e[0]/n):r.push(e[0]*n);for(let s=1;s<e.length;++s)s<=t.length?a?r.push(t[s-1]*e[s]):r.push(e[s]/t[s-1]):r.push(e[s]);return r}function Yz(e,t){let n=[0];for(let a=0;a<t;++a)n.push(e[a][0]);return n}function Zz(e,t,n){let a=e.slice(0,1);for(let r=0;r<n;++r)a.push(e[r+1]-t[r][0]-t[r][1]);return a}var eN=1.7580993408473768,tN=1.0507009873554805,Jz=.3275911,Qz=.254829592,eW=-.284496736,tW=1.421413741,nW=-1.453152027,aW=1.061405429;function rW(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);let n=new Float32Array(e.length*2);for(let a=0;a<n.length;a+=2)n[a]=e[a/2],n[a+1]=t[a/2];return n}function sW(e){let t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let a=0;a<e.length;a+=2)t[a/2]=e[a],n[a/2]=e[a+1];return{real:t,imag:n}}function iW(e){let t=Math.ceil(e.length/4),n=new Float32Array(t),a=new Float32Array(t);for(let r=0;r<e.length;r+=4)n[Math.floor(r/4)]=e[r],a[Math.floor(r/4)]=e[r+1];return{real:n,imag:a}}function oW(e){let t=Math.floor(e.length/4),n=new Float32Array(t),a=new Float32Array(t);for(let r=2;r<e.length;r+=4)n[Math.floor(r/4)]=e[r],a[Math.floor(r/4)]=e[r+1];return{real:n,imag:a}}function lW(e,t){let n=e[t*2],a=e[t*2+1];return{real:n,imag:a}}function uW(e,t,n,a){e[a*2]=t,e[a*2+1]=n}function pW(e,t){let n=new Float32Array(e/2),a=new Float32Array(e/2);for(let r=0;r<Math.ceil(e/2);r++){let s=(t?2:-2)*Math.PI*(r/e);n[r]=Math.cos(s),a[r]=Math.sin(s)}return{real:n,imag:a}}function cW(e,t,n){let a=(n?2:-2)*Math.PI*(e/t),r=Math.cos(a),s=Math.sin(a);return{real:r,imag:s}}var yb="->",dW=/->/g,vk=",",wk="...";function hW(e,t){e=e.replace(/\s/g,"");let n=(e.length-e.replace(dW,"").length)/yb.length;if(n<1)throw new Error("Equations without an arrow are not supported.");if(n>1)throw new Error(`Equation must contain exactly one arrow ("${yb}").`);let[a,r]=e.split(yb);$(a.indexOf(wk)===-1,()=>`The ellipsis notation ("${wk}") is not supported yet.`);let s=a.split(vk),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let o=[];for(let c=0;c<r.length;++c){let h=r[c];if(!s.some(m=>m.indexOf(h)!==-1))throw new Error(`Output subscripts contain the label ${h} not present in the input subscripts.`);o.indexOf(h)===-1&&o.push(h)}for(let c=0;c<a.length;++c){let h=a[c];o.indexOf(h)===-1&&h!==vk&&o.push(h)}let l=new Array(s.length);for(let c=0;c<i;++c){if(new Set(s[c].split("")).size!==s[c].length)throw new Error(`Found duplicate axes in input component ${s[c]}. Support for duplicate axes in input is not implemented yet.`);l[c]=[];for(let h=0;h<s[c].length;++h)l[c].push(o.indexOf(s[c][h]))}let u=o.length,p=r.length,d=[];for(let c=p;c<u;++c)d.push(c);return{allDims:o,summedDims:d,idDims:l}}function mW(e,t){let n=new Array(e);n.fill(-1);for(let r=0;r<t.length;++r)n[t[r]]=r;let a=[];for(let r=0;r<e;++r)n[r]===-1&&a.push(r);return n=n.filter(r=>r!==-1),{permutationIndices:n,expandDims:a}}function fW(e,t,n){let a=new Array(e);for(let r=0;r<n.length;++r){let s=n[r].shape;for(let i=0;i<t[r].length;++i)a[t[r][i]]===void 0?a[t[r][i]]=s[i]:$(a[t[r][i]]===s[i],()=>`Expected dimension ${a[t[r][i]]} at axis ${i} of input shaped ${JSON.stringify(s)}, but got dimension ${s[i]}`)}}function gW(e,t){let n=e,a=[],r=0;e.length===0&&n.push(-1),r=e.length+1;for(let i=0;i<r;++i)a.push([]);let s=[];for(let i=0;i<n.length;++i){let o=n[i],l=bW(t,o);for(let u of l)s.indexOf(u)===-1&&(a[i].push(u),s.push(u))}return{path:n,steps:a}}function yW(e){return e.every((t,n)=>t===n)}function bW(e,t){let n=[];for(let a=0;a<e.length;++a)(e[a].length===0||e[a].indexOf(t)!==-1||t===-1)&&n.push(a);return n}function xW(e,t,n=0){let a=[];if(typeof t=="number")$(e.shape[n]%t===0,()=>"Number of splits must evenly divide the axis."),a=new Array(t).fill(e.shape[n]/t);else{let r=t.reduce((i,o)=>(o===-1&&(i+=1),i),0);$(r<=1,()=>"There should be only one negative value in split array.");let s=t.indexOf(-1);if(s!==-1){let i=t.reduce((o,l)=>l>0?o+l:o);t[s]=e.shape[n]-i}$(e.shape[n]===t.reduce((i,o)=>i+o),()=>"The sum of sizes must match the size of the axis dimension."),a=t}return a}function vW(e){return`Received SparseTensor with denseShape[0] = 0 but
-  indices.shape[0] = ${e}`}function wW(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function kW(e,t,n){return`indices(${e}, 0) is invalid: ${t} >= ${n}`}function IW(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function SW(e,t){return`size ${e} must be non-negative, not ${t}`}function TW(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function NW(e,t){let n=mt(e),a=mt(t);return`Input to reshape is a SparseTensor with ${n}
-  dense values, but the requested shape requires a multiple of ${a}. inputShape=${e} outputShape= ${t}`}function CW(e,t){let n=mt(e),a=mt(t);return`Input to reshape is a tensor with ${n} dense values, but the requested shape has ${a}. inputShape=${e} outputShape=${t}`}function _W(){return"segment ids must be >= 0"}function EW(){return"segment ids are not increasing"}function AW(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function $W(e,t,n){return`Bad: indices[${e}] == ${t} out of range [0, ${n})`}var nN={};Ae(nN,{collectGatherOpShapeInfo:()=>RW,computeOutShape:()=>DW,segOpComputeOptimalWindowSize:()=>FW});function FW(e,t){let n=!1,a;for(e<=Qv?(a=e,n=!0):a=Eh(e,Math.floor(Math.sqrt(e)));!n;)a>t||a===e?n=!0:a=Eh(e,a+1);return a}function DW(e,t,n){let a=[],r=e.length;for(let s=0;s<r;s++)s!==t?a.push(e[s]):a.push(n);return a}function RW(e,t,n,a){let r=t.shape.length,s=e.shape.length;if(a!==0&&(a<-r||a>r))throw new Error(`Expect batchDims in the range of [-${r}, ${r}], but got ${a}`);if(a<0&&(a+=r),a>s)throw new Error(`batchDims (${a}) must be less than rank(x) (
-    ${s}).`);if(n<a)throw new Error(`batchDims (${a}) must be less than or equal to axis (${n}).`);for(let d=0;d<a;++d)if(e.shape[d]!==t.shape[d])throw new Error(`x.shape[${d}]: ${e.shape[d]} should be equal to indices.shape[${d}]: ${t.shape[d]}.`);let i=e.shape[n],o=[],l=1,u=1,p=1;for(let d=0;d<a;++d)o.push(e.shape[d]),l*=e.shape[d];for(let d=a;d<n;d++)o.push(e.shape[d]),u*=e.shape[d];for(let d=a;d<r;d++)o.push(t.shape[d]);for(let d=n+1;d<s;d++)o.push(e.shape[d]),p*=e.shape[d];return{batchSize:l,sliceSize:p,outerSize:u,dimSize:i,outputShape:o}}function MW(e){try{return e.map(t=>Mh(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function PW(e){return e.map(t=>Ac(t))}var hr={};Ae(hr,{nonMaxSuppressionV3Impl:()=>GT,nonMaxSuppressionV4Impl:()=>HT,nonMaxSuppressionV5Impl:()=>jT,whereImpl:()=>DT});var aN={kernelName:Dl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,go(oe(n,"float32"),-1))}}},OW={kernelName:Rl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=lt(oe(n,"float32")),r=un(pe(be(1),a));return vt(he(e,r))}}}},LW={kernelName:Ml,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=un(pe(lt(oe(n,"float32")),1));return he(e,a)}}}},zW={kernelName:gs,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=e,i=Wt(n.shape,r);return i.length>0&&(s=fe(s,i)),W(s,n.shape)},b:()=>{let s=e,i=Wt(a.shape,r);return i.length>0&&(s=fe(s,i)),W(s,a.shape)}}}},WW={kernelName:fi,saveAllInputs:!0,gradFunc:(e,t)=>{let n={};return t.forEach((a,r)=>{n[r]=()=>e.clone()}),n}},BW={kernelName:gi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},VW={kernelName:cc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},UW={kernelName:Ll,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,un(pe(be(1),lt(oe(n,"float32")))))}}},GW={kernelName:zl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=un(Y(be(1),lt(oe(n,"float32"))));return he(e,a)}}}},HW={kernelName:Vl,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=Y(lt(n),lt(a)),i=z(e,he(a,s)),o=Wt(n.shape,r);return o.length>0&&(i=fe(i,o)),W(i,n.shape)},b:()=>{let s=Y(lt(n),lt(a)),i=vt(z(e,he(n,s))),o=Wt(a.shape,r);return o.length>0&&(i=fe(i,o)),W(i,a.shape)}}}},jW={kernelName:Wl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,Y(lt(oe(n,"float32")),1))}}},qW={kernelName:Bl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,pe(be(1),lt(oe(n,"float32"))))}}};function KW(e,t,n,a,r,s){let i=_(e,"dy","avgPool3dGrad"),o=_(t,"input","avgPool3dGrad"),l=i,u=o,p=!1;o.rank===4&&(p=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),u=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),$(l.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${l.rank}.`),$(u.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${u.rank}.`),Nn("avgPool3dGrad",r,s);let d={dy:l,input:u},c={filterSize:n,strides:a,pad:r,dimRoundingMode:s},h=O.runKernel(cm,d,c);return p?W(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}var XW=L({avgPool3dGrad_:KW}),YW={kernelName:dc,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{filterSize:r,strides:s,pad:i,dimRoundingMode:o}=n;return{x:()=>XW(e,a,r,s,i,o)}}};function ZW(e,t,n,a,r){let s=_(e,"dy","avgPoolGrad"),i=_(t,"input","avgPoolGrad");$(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let o=i,l=s,u=!1;i.rank===3&&(u=!0,o=W(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=W(s,[1,s.shape[0],s.shape[1],s.shape[2]])),$(l.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${l.rank}.`),$(o.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);let p={dy:l,input:o},d={filterSize:n,strides:a,pad:r},c=O.runKernel(pm,p,d);return u?W(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var JW=L({avgPoolGrad_:ZW}),QW={kernelName:yi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{filterSize:r,strides:s,pad:i}=n;return{x:()=>JW(e,a,r,s,i)}}},eB={kernelName:bi,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{let[a,r]=t,{transposeA:s,transposeB:i}=n;return!s&&!i?{a:()=>Fe(e,r,!1,!0),b:()=>Fe(a,e,!0,!1)}:!s&&i?{a:()=>Fe(e,r,!1,!1),b:()=>Fe(e,a,!0,!1)}:s&&!i?{a:()=>Fe(r,e,!1,!0),b:()=>Fe(a,e,!1,!1)}:{a:()=>Fe(r,e,!0,!0),b:()=>Fe(e,a,!0,!0)}}},tB={kernelName:Ul,gradFunc:(e,t,n)=>{let{blockShape:a,crops:r}=n;return{x:()=>Vc(e,a,r)}}},nB={kernelName:tS,gradFunc:(e,t,n)=>{let a=n,r=a.inputShape,s=a.shape,i=Array.from(s);for(let l=r.length-1;l>=0;l--)if(r[l]===s[l])i[l]=1;else if(r[l]!==1)throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${s}].`);let o=[];for(let l=0;l<i.length;l++)i[l]>1&&o.push(l);return{x:()=>fe(e,o,!0)}}},aB={kernelName:xi,gradFunc:e=>({x:()=>e.clone()})},rB={kernelName:vi,gradFunc:e=>({x:()=>qe(e)})},sB={kernelName:ys,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{clipValueMin:r,clipValueMax:s}=n;return{x:()=>mn($a(Fr(a,r),Ss(a,s)),e,qe(e))}}},iB={kernelName:hc,inputsToSave:["x"],gradFunc:aN.gradFunc},oB={kernelName:Gl,saveAllInputs:!0,gradFunc:(e,t,n)=>{let a=t.map(o=>o.shape),{axis:r}=n,s=Fa(r,t[0].shape)[0],i=a.map(o=>o[s]);return zn(e,i,s).map(o=>()=>o)}},lB={kernelName:wi,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,{dilations:s,strides:i,pad:o,dataFormat:l}=n;return $(os(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>mv(a.shape,e,r,i,o,l),filter:()=>Xv(a,e,r.shape,i,o,l)}}},uB={kernelName:ki,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,{strides:s,pad:i,dataFormat:o,dimRoundingMode:l}=n;return{dy:()=>$t(e,r,s,i,o,1,l),filter:()=>Xv(e,a,r.shape,s,i,o,l)}}};function pB(e,t,n,a,r){let s=e;e.rank===4&&(s=W(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let i=t;i.rank===4&&(i=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),$(s.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),$(i.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),$(n.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),$(s.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${n[3]}.`),$(i.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${n[4]}).`);let o={x:s,dy:i},l={strides:a,pad:r,filterShape:n};return O.runKernel(gm,o,l)}var cB=L({conv3DBackpropFilter_:pB}),dB={kernelName:mc,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:a,strides:r,pad:s}=n;$(os(a),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${a}'`);let[i,o]=t;return{x:()=>eT(i.shape,e,o,r,s),filter:()=>cB(i,e,o.shape,r,s)}}},hB={kernelName:Ii,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(vt(pf(oe(n,"float32"))),e)}}},mB={kernelName:Si,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(cf(oe(n,"float32")),e)}}},fB={kernelName:Ti,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r,exclusive:s,reverse:i}=n;return{x:()=>{let o=sT([r],a.rank),l=Ym(e,r,s,!i);return o!=null&&(l=Ee(l,o)),l}}}},gB={kernelName:Ni,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:a,strides:r,pad:s,dimRoundingMode:i}=n,o=a==null?[1,1]:a;$(os(o),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);let[l,u]=t;return $(l.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${l.rank}.`),$(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${u.rank}.`),$(l.shape[3]===u.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),$(cr(r,o),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${r} and dilations '${o}'.`),Nn("depthwiseConv2d",s,i),{x:()=>BT(l.shape,e,u,r,s,o,i),filter:()=>WT(l,e,u.shape,r,s,o,i)}}},yB={kernelName:fc,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[a,r]=t,s={x:a,filter:r,dy:e},i={x:a,filter:r,dy:e};return{x:()=>O.runKernel(Ah,s,n),filter:()=>O.runKernel($h,i,n)}}},bB={kernelName:_i,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t,a={dy:e,y:n};return{x:()=>O.runKernel(Im,a)}}},xB={kernelName:Kl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,a=z(fn(vt(lt(n))),2/Math.sqrt(Math.PI));return{x:()=>z(e,a)}}},vB={kernelName:Ei,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,n)}}},wB={kernelName:Yl,inputsToSave:["input"],gradFunc:(e,t)=>{let[n]=t;return{input:()=>W(e,n.shape)}}},kB={kernelName:Zl,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,fn(n))}}},IB={kernelName:Ai,gradFunc:e=>({x:()=>qe(e)})},SB={kernelName:$i,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=he(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);i.length>0&&(s=W(fe(s,i),a.shape));let o=lt(a);return vt(he(s,oe(o,"float32")))}}}},TB={kernelName:Fi,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{let{varianceEpsilon:a}=n,[r,s,i,o]=t,l=o==null?be(1):o,u=Wt(s.shape,r.shape),p=[];if(s.rank===1){for(let f=0;f<r.shape.length-1;++f)p.push(r.shape[f]);p.push(1)}let d=pe(r,s),c=z(e,l),h=lf(Y(i,be(a))),m=z(z(z(h,h),h),be(-.5));return{x:()=>s.rank===1?W(z(z(e,Ln(W(h,[1,1,1,s.shape[0]]),p)),l),r.shape):W(z(z(e,h),l),r.shape),mean:()=>{let f=z(z(h,be(-1)),c);return s.rank===1&&(f=fe(f,u)),W(f,s.shape)},variance:()=>{let f=z(z(m,d),c);return s.rank===1&&(f=fe(f,u)),W(f,s.shape)},scale:()=>{let f=z(d,h),g=z(e,f);return s.rank===1&&(g=fe(g,u)),W(g,s.shape)},offset:()=>{let f=e;return s.rank===1&&(f=fe(f,u)),W(f,s.shape)}}}},NB={kernelName:Ql,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{let[a,r]=t,{axis:s}=n,i=Fa(s,a.shape)[0];return{x:()=>{let o=a.shape,l=r.size,u=o.slice(0,i),p=u.length,d=o.slice(s,o.length).slice(1),c=d.length,h=kk(0,p),m=kk(p+1,p+1+c),f=Ik([u,[l],d]),g=W(e,f),y=W(r,[l]),b=Ik([[p],h,m]),x=Ee(g,b),w=gf(x,y,a.shape[i]),I=Iv(b);return w=Ee(w,I),w},indices:()=>r}}};function kk(e,t){let n=[];for(let a=e;a<t;++a)n.push(a);return n}function Ik(e){let t=[];for(let n=0;n<e.length;++n)for(let a=0;a<e[n].length;++a)t.push(e[n][a]);return t}var CB={kernelName:Di,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>qe(n),b:()=>qe(a)}}},_B={kernelName:Ri,gradFunc:e=>({x:()=>oe(e,"float32")})},EB={kernelName:nu,gradFunc:e=>({x:()=>qe(e)})},AB={kernelName:au,gradFunc:e=>({x:()=>qe(e)})},$B={kernelName:ru,gradFunc:e=>({x:()=>qe(e)})},FB={kernelName:Mi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{alpha:r}=n,s=Cn(a,0);return{x:()=>mn(s,e,z(e,r))}}},DB={kernelName:ou,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,Y(n,1))}}},RB={kernelName:Pi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,oe(n,"float32"))}}},MB={kernelName:aS,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n;return{logits:()=>{let s=fn(a);return pe(e,z(fe(e,r,!0),s))}}}};function PB(e,t,n,a=5,r=1,s=1,i=.5){let o={x:e,y:t,dy:n},l={depthRadius:a,bias:r,alpha:s,beta:i};return O.runKernel(_m,o,l)}var OB=L({localResponseNormalizationBackprop_:PB}),LB={kernelName:yc,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{depthRadius:s,bias:i,alpha:o,beta:l}=n;return{x:()=>OB(a,r,e,s,i,o,l)}}};function rN(e,t,n,a){return t.rank<n.rank&&(t=W(t,ii(t.shape,a))),e.rank<n.rank&&(e=W(e,ii(e.shape,a))),{x:()=>z(e,oe(ea(n,t),e.dtype))}}var Sk={kernelName:Oi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let a=n,{reductionIndices:r}=a,s=t[0],i=t[1],o=Fa(r,s.shape),l=rN(e,i,s,o);return{x:()=>l.x()}}},zB={kernelName:Li,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>z(e,oe(Fr(n,a),"float32")),b:()=>z(e,oe(Jm(n,a),"float32"))}}};function WB(e,t,n,a,r,s,i){let o=_(e,"dy","maxPool3dGrad"),l=_(t,"input","maxPool3dGrad"),u=_(n,"output","maxPool3dGrad"),p=o,d=l,c=u,h=!1;l.rank===4&&(h=!0,p=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),d=W(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]]),c=W(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]])),$(p.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${p.rank}.`),$(d.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${d.rank}.`),$(c.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${c.rank}.`),Nn("maxPool3dGrad",s,i);let m={dy:p,input:d,output:c},f={filterSize:a,strides:r,pad:s,dimRoundingMode:i},g=O.runKernel(Am,m,f);return h?W(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var BB=L({maxPool3dGrad_:WB}),VB={kernelName:bc,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n;return{x:()=>BB(e,a,r,s,i,o,l)}}};function UB(e,t,n,a,r,s,i){let o=_(e,"dy","maxPoolGrad"),l=_(t,"input","maxPoolGrad"),u=_(n,"output","maxPoolGrad");$(l.rank===o.rank,()=>`Rank of input (${l.rank}) does not match rank of dy (${o.rank})`),$(o.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`),$(l.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${l.rank}.`),Nn("maxPoolGrad",s,i);let p={dy:o,input:l,output:u},d={filterSize:a,strides:r,pad:s,dimRoundingMode:i};return O.runKernel(Em,p,d)}var GB=L({maxPoolGrad_:UB}),HB={kernelName:zi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a,r]=t,{filterSize:s,strides:i,pad:o}=n;return{x:()=>GB(e,a,r,s,i,o)}}},jB={kernelName:Wi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n,s=Fa(r,a.shape),i=rT(a.shape,s)[1],o=mt(i);return{x:()=>{let l=a.shape.slice();s.forEach(p=>{l[p]=1});let u=W(e,l);return he(z(u,Jn(a.shape,"float32")),o)}}}},qB={kernelName:Bi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let a=n,{axis:r}=a,[s,i]=t,o=Fa(r,s.shape),l=rN(e,i,s,o);return{x:()=>l.x()}}},KB={kernelName:Vi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t;return{a:()=>z(e,oe(Ss(n,a),"float32")),b:()=>z(e,oe(Cn(n,a),"float32"))}}},XB={kernelName:Ui,inputsToSave:["x"],gradFunc:(e,t,n)=>{let a=t[0],{paddings:r}=n,s=r.map(i=>i[0]);return{x:()=>Be(e,s,a.shape)}}},YB={kernelName:cu,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=Wt(n.shape,r);return s.length>0?W(fe(e,s),n.shape):e},b:()=>{let s=z(e,vt(Wu(he(n,a)))),i=Wt(a.shape,r);return i.length>0?W(fe(s,i),a.shape):s}}}},ZB={kernelName:Gi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=z(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);return i.length>0?W(fe(s,i),a.shape):s}}}},JB={kernelName:du,gradFunc:e=>({x:()=>vt(e)})},QB={kernelName:Hi,inputsToSave:["indices"],gradFunc:(e,t)=>{let n=t[0];return{indices:()=>It(n.shape,"float32")}}},e4={kernelName:yu,gradFunc:e=>({x:()=>qe(e)})},t4={kernelName:bu,saveAllInputs:!0,gradFunc:(e,t,n)=>{let{axis:a}=n;return ct(e,a).map(r=>()=>r)}},Tk={kernelName:ji,inputsToSave:["x"],gradFunc:(e,t,n)=>{let a=t[0],{paddings:r}=n,s=r.map(i=>i[0]);return{x:()=>Be(e,s,a.shape)}}},n4={kernelName:qi,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{let[n,a,r]=t,s=n,i=a,o=ut(s.shape,i.shape);return{a:()=>{let l=oe(i,"float32"),u=z(e,z(l,$r(s,pe(l,be(1))))),p=Wt(s.shape,o);return p.length>0&&(u=fe(u,p)),W(u,s.shape)},b:()=>{let l=Cn(s,0),u=mn(l,ta(s),qe(s)),p=z(e,z(r,u)),d=Wt(i.shape,o);return d.length>0&&(p=fe(p,d)),W(p,i.shape)}}}},a4={kernelName:Ki,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{let[n,a]=t,r=Cn(n,0);return{x:()=>mn(r,e,z(e,a)),alpha:()=>{let s=mn(r,qe(e),z(e,n)),i=Wt(a.shape,e.shape);return i.length>0&&(s=fe(s,i)),W(s,a.shape)}}}};function r4(e,t,n){let a=e.shape.slice();a[n]=1;let r=W(t,a),s=ec(e,n,!0,!1),i=ec(e,n,!0,!0),o=z(s,i);return z(r,o)}function s4(e,t,n){let a=e.shape.length,r=a-n.length,s=N.getAxesPermutation(n,a),i=e;s!=null&&(i=Ee(e,s));let o=i.shape.slice(),l=o.splice(a-n.length,n.length).reduce((d,c)=>d*c,1);o.push(l);let u=i.reshape(o),p=r4(u,t,r);if(p=p.reshape(i.shape),s!=null){let d=N.getUndoAxesPermutation(s);p=Ee(p,d)}return p}var i4={kernelName:Xi,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{axis:r}=n,s=[];return r==null?s=a.shape.map((i,o)=>o):typeof r=="number"?s=[r]:s=r,{x:()=>s4(a,e,s)}}},o4={kernelName:Ci,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=he(e,oe(a,"float32")),i=Wt(n.shape,r);return i.length>0?W(fe(s,i),n.shape):s},b:()=>{let s=z(e,oe(n,"float32")),i=Wt(a.shape,r);i.length>0&&(s=W(fe(s,i),a.shape));let o=lt(a);return vt(he(s,oe(o,"float32")))}}}},l4={kernelName:xu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,vt(lt(n)))}}},u4={kernelName:Qi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,a=z(Ss(n,6),go(n));return{x:()=>z(e,oe(a,"float32"))}}},p4={kernelName:Yi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,oe(go(n),"float32"))}}},c4={kernelName:vu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>W(e,n.shape)}}},d4={kernelName:Ji,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[a]=t,r={dy:e,images:a};return{images:()=>O.runKernel(Lm,r,n)}}},h4={kernelName:Zi,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[a]=t,r={dy:e,images:a};return{images:()=>O.runKernel(Om,r,n)}}},m4={kernelName:eo,gradFunc:(e,t,n)=>{let{dims:a}=n,r=Fa(a,e.shape);return{x:()=>ga(e,r)}}},f4={kernelName:to,gradFunc:e=>({x:()=>qe(e)})},g4={kernelName:no,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>vt(he(e,z($r(n,1.5),2)))}}},y4={kernelName:ku,inputsToSave:["condition"],gradFunc:(e,t)=>{let[n]=t;return{condition:()=>oe(qe(n),"float32"),t:()=>z(e,oe(n,e.dtype)),e:()=>z(e,oe(Wc(n),e.dtype))}}},b4={kernelName:Iu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let a=Cn(n,be(0)),r=be(eN),s=be(tN),i=z(e,s),o=z(z(e,r),fn(oe(n,"float32")));return mn(a,i,o)}}}},x4={kernelName:ro,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,z(n,pe(be(1),n)))}}},v4={kernelName:Nu,gradFunc:e=>({x:()=>qe(e)})},w4={kernelName:ao,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(Oc(oe(n,"float32")),e)}}},k4={kernelName:Tu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(Xm(oe(n,"float32")),e)}}},I4={kernelName:Su,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{begin:r,size:s}=n,i=a.shape,[o,l]=US(a,r,s),u=[];for(let p=0;p<e.rank;p++)u.push([o[p],i[p]-o[p]-l[p]]);return{x:()=>ba(e,u)}}},S4={kernelName:oo,outputsToSave:[!0],gradFunc:(e,t,n)=>{let[a]=t,{dim:r}=n,s=!0,i=z(e,a);return{logits:()=>pe(i,z(fe(i,[r],s),a))}}},T4={kernelName:Cu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,ha(n))}}},Nk={kernelName:_u,gradFunc:(e,t,n)=>{let{blockShape:a,paddings:r}=n;return{x:()=>Pc(e,a,r)}}},Ck={kernelName:Eu,gradFunc:(e,t,n)=>{let{axis:a}=n;return{x:()=>Ze(e,a)}}},N4={kernelName:so,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,z(un(oe(n,"float32")),2))}}},C4={kernelName:Ic,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(e,z(oe(n,"float32"),2))}}},_4={kernelName:lo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=be(2);return{a:()=>z(e,z(r,pe(n,a))),b:()=>z(e,z(r,pe(a,n)))}}},E4={kernelName:xs,gradFunc:e=>({x:()=>qe(e)})},A4={kernelName:uo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,a]=t,r=ut(n.shape,a.shape);return{a:()=>{let s=e,i=Wt(n.shape,r);return i.length>0&&(s=fe(s,i)),W(s,n.shape)},b:()=>{let s=e,i=Wt(a.shape,r);return i.length>0&&(s=fe(s,i)),W(vt(s),a.shape)}}}},$4={kernelName:io,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,r=a.shape.slice(),{axis:s}=n;Fa(s,a.shape).forEach(l=>{r[l]=1});let i=W(e,r),o=z(i,Jn(a.shape,"float32"));return{x:()=>o}}},F4={kernelName:po,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>he(e,lt(Oc(n)))}}},D4={kernelName:co,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>z(pe(be(1),lt(n)),e)}}},R4={kernelName:bs,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[a]=t,{reps:r}=n;return{x:()=>{let s=qe(a);if(a.rank===1)for(let i=0;i<r[0];++i)s=Y(s,Be(e,[i*a.shape[0]],[a.shape[0]]));else if(a.rank===2)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1]],[a.shape[0],a.shape[1]]));else if(a.rank===3)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)for(let l=0;l<r[2];++l)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1],l*a.shape[2]],[a.shape[0],a.shape[1],a.shape[2]]));else if(a.rank===4)for(let i=0;i<r[0];++i)for(let o=0;o<r[1];++o)for(let l=0;l<r[2];++l)for(let u=0;u<r[3];++u)s=Y(s,Be(e,[i*a.shape[0],o*a.shape[1],l*a.shape[2],u*a.shape[3]],[a.shape[0],a.shape[1],a.shape[2],a.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${a.rank} tensors yet.`);return s}}}},M4={kernelName:_r,gradFunc:(e,t,n)=>{let a=n,{perm:r}=a,s=Iv(r);return{x:()=>Ee(e,s)}}},P4={kernelName:Ru,gradFunc:(e,t,n)=>{let a=n,{axis:r}=a;return{value:()=>Ft(e,r)}}},O4={kernelName:Cc,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>L4(e,n)}}};function L4(e,t){let n=dr(t,qe(t)),a=Bu(e,n),r=Fr(t,be(0,"int32")),s=a.rank-r.rank;for(let o=0;o<s;++o)r=Zt(r,o+1);r=$a(r,Jn(a.shape,"bool"));let i=qe(a);return mn(r,a,i)}var z4={kernelName:Mu,gradFunc:e=>({x:()=>qe(e)})},W4=[aN,OW,LW,zW,WW,BW,VW,UW,GW,HW,jW,qW,YW,QW,eB,tB,nB,aB,rB,sB,iB,oB,uB,lB,dB,hB,mB,fB,gB,yB,o4,bB,xB,vB,wB,kB,SB,IB,TB,NB,CB,_B,EB,AB,$B,FB,DB,RB,MB,LB,Sk,Sk,zB,VB,HB,jB,qB,KB,XB,YB,ZB,JB,QB,e4,t4,Tk,Tk,n4,a4,i4,l4,u4,p4,c4,d4,h4,m4,f4,g4,y4,b4,x4,v4,w4,k4,I4,S4,T4,Nk,Nk,Ck,Ck,N4,_4,C4,E4,A4,$4,F4,D4,R4,M4,P4,O4,z4];for(let e of W4)rS(e);J().prototype.abs=function(){return this.throwIfDisposed(),Lt(this)};J().prototype.acos=function(){return this.throwIfDisposed(),Xx(this)};J().prototype.acosh=function(){return this.throwIfDisposed(),Yx(this)};J().prototype.add=function(e){return this.throwIfDisposed(),Y(this,e)};J().prototype.all=function(e,t){return this.throwIfDisposed(),jm(this,e,t)};J().prototype.any=function(e,t){return this.throwIfDisposed(),Qp(this,e,t)};J().prototype.argMax=function(e){return this.throwIfDisposed(),ri(this,e)};J().prototype.argMin=function(e){return this.throwIfDisposed(),Zx(this,e)};J().prototype.asScalar=function(){return this.throwIfDisposed(),$(this.size===1,()=>"The array must have only 1 element."),W(this,[])};J().prototype.asType=function(e){return this.throwIfDisposed(),oe(this,e)};J().prototype.as1D=function(){return this.throwIfDisposed(),W(this,[this.size])};J().prototype.as2D=function(e,t){return this.throwIfDisposed(),W(this,[e,t])};J().prototype.as3D=function(e,t,n){return this.throwIfDisposed(),W(this,[e,t,n])};J().prototype.as4D=function(e,t,n,a){return this.throwIfDisposed(),W(this,[e,t,n,a])};J().prototype.as5D=function(e,t,n,a,r){return this.throwIfDisposed(),W(this,[e,t,n,a,r])};J().prototype.asin=function(){return this.throwIfDisposed(),Jx(this)};J().prototype.asinh=function(){return this.throwIfDisposed(),Qx(this)};J().prototype.atan=function(){return this.throwIfDisposed(),ev(this)};J().prototype.atan2=function(e){return this.throwIfDisposed(),tv(this,e)};J().prototype.atanh=function(){return this.throwIfDisposed(),nv(this)};J().prototype.avgPool=function(e,t,n,a){return this.throwIfDisposed(),ya(this,e,t,n,a)};J().prototype.batchToSpaceND=function(e,t){return this.throwIfDisposed(),Pc(this,e,t)};J().prototype.batchNorm=function(e,t,n,a,r){return this.throwIfDisposed(),ks(this,e,t,n,a,r)};J().prototype.broadcastTo=function(e){return this.throwIfDisposed(),Xs(this,e)};J().prototype.cast=function(e){return this.throwIfDisposed(),oe(this,e)};J().prototype.ceil=function(){return this.throwIfDisposed(),uv(this)};J().prototype.clipByValue=function(e,t){return this.throwIfDisposed(),en(this,e,t)};J().prototype.concat=function(e,t){return this.throwIfDisposed(),e instanceof Te&&(e=[e]),Ze([this,...e],t)};J().prototype.conv1d=function(e,t,n,a,r,s){return this.throwIfDisposed(),qm(this,e,t,n,a,r,s)};J().prototype.conv2dTranspose=function(e,t,n,a,r){return this.throwIfDisposed(),Km(this,e,t,n,a,r)};J().prototype.conv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),$t(this,e,t,n,a,r,s)};J().prototype.cos=function(){return this.throwIfDisposed(),Oc(this)};J().prototype.cosh=function(){return this.throwIfDisposed(),Xm(this)};J().prototype.cumprod=function(e,t,n){return this.throwIfDisposed(),ec(this,e,t,n)};J().prototype.cumsum=function(e,t,n){return this.throwIfDisposed(),Ym(this,e,t,n)};J().prototype.depthToSpace=function(e,t){return this.throwIfDisposed(),yv(this,e,t)};J().prototype.depthwiseConv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),Is(this,e,t,n,a,r,s)};J().prototype.dilation2d=function(e,t,n,a,r){return this.throwIfDisposed(),bv(this,e,t,n,a,r)};J().prototype.divNoNan=function(e){return this.throwIfDisposed(),xv(this,e)};J().prototype.div=function(e){return this.throwIfDisposed(),he(this,e)};J().prototype.dot=function(e){return this.throwIfDisposed(),vv(this,e)};J().prototype.elu=function(){return this.throwIfDisposed(),Lu(this)};J().prototype.equal=function(e){return this.throwIfDisposed(),ea(this,e)};J().prototype.erf=function(){return this.throwIfDisposed(),wv(this)};J().prototype.euclideanNorm=function(e,t){return this.throwIfDisposed(),Sv(this,e,t)};J().prototype.exp=function(){return this.throwIfDisposed(),fn(this)};J().prototype.expandDims=function(e){return this.throwIfDisposed(),Zt(this,e)};J().prototype.expm1=function(){return this.throwIfDisposed(),Tv(this)};J().prototype.fft=function(){return this.throwIfDisposed(),Hc(this)};J().prototype.flatten=function(){return this.throwIfDisposed(),W(this,[this.size])};J().prototype.floor=function(){return this.throwIfDisposed(),Wu(this)};J().prototype.floorDiv=function(e){return this.throwIfDisposed(),Hm(this,e)};J().prototype.gather=function(e,t){return this.throwIfDisposed(),Bu(this,e,t)};J().prototype.greaterEqual=function(e){return this.throwIfDisposed(),Fr(this,e)};J().prototype.greater=function(e){return this.throwIfDisposed(),Cn(this,e)};J().prototype.ifft=function(){return this.throwIfDisposed(),wl(this)};J().prototype.irfft=function(){return this.throwIfDisposed(),hf(this)};J().prototype.isFinite=function(){return this.throwIfDisposed(),Nv(this)};J().prototype.isInf=function(){return this.throwIfDisposed(),Cv(this)};J().prototype.isNaN=function(){return this.throwIfDisposed(),_v(this)};J().prototype.leakyRelu=function(e){return this.throwIfDisposed(),Lc(this,e)};J().prototype.lessEqual=function(e){return this.throwIfDisposed(),Ss(this,e)};J().prototype.less=function(e){return this.throwIfDisposed(),Jm(this,e)};J().prototype.localResponseNormalization=function(e,t,n,a){return this.throwIfDisposed(),Ev(this,e,t,n,a)};J().prototype.logSigmoid=function(){return this.throwIfDisposed(),Av(this)};J().prototype.logSoftmax=function(e){return this.throwIfDisposed(),ef(this,e)};J().prototype.logSumExp=function(e,t){return this.throwIfDisposed(),tf(this,e,t)};J().prototype.log=function(){return this.throwIfDisposed(),ta(this)};J().prototype.log1p=function(){return this.throwIfDisposed(),zc(this)};J().prototype.logicalAnd=function(e){return this.throwIfDisposed(),$a(this,e)};J().prototype.logicalNot=function(){return this.throwIfDisposed(),Wc(this)};J().prototype.logicalOr=function(e){return this.throwIfDisposed(),nf(this,e)};J().prototype.logicalXor=function(e){return this.throwIfDisposed(),$v(this,e)};J().prototype.matMul=function(e,t,n){return this.throwIfDisposed(),Fe(this,e,t,n)};J().prototype.maxPool=function(e,t,n,a){return this.throwIfDisposed(),Dt(this,e,t,n,a)};J().prototype.max=function(e,t){return this.throwIfDisposed(),ma(this,e,t)};J().prototype.maximum=function(e){return this.throwIfDisposed(),dr(this,e)};J().prototype.mean=function(e,t){return this.throwIfDisposed(),Nt(this,e,t)};J().prototype.min=function(e,t){return this.throwIfDisposed(),bl(this,e,t)};J().prototype.minimum=function(e){return this.throwIfDisposed(),Vu(this,e)};J().prototype.mirrorPad=function(e,t){return this.throwIfDisposed(),Dv(this,e,t)};J().prototype.mod=function(e){return this.throwIfDisposed(),Rv(this,e)};J().prototype.mul=function(e){return this.throwIfDisposed(),z(this,e)};J().prototype.neg=function(){return this.throwIfDisposed(),vt(this)};J().prototype.norm=function(e,t,n){return this.throwIfDisposed(),zu(this,e,t,n)};J().prototype.notEqual=function(e){return this.throwIfDisposed(),oi(this,e)};J().prototype.oneHot=function(e,t=1,n=0){return this.throwIfDisposed(),gl(this,e,t,n)};J().prototype.onesLike=function(){return this.throwIfDisposed(),na(this)};J().prototype.pad=function(e,t){return this.throwIfDisposed(),ba(this,e,t)};J().prototype.pool=function(e,t,n,a,r,s){return this.throwIfDisposed(),Mv(this,e,t,n,a,r,s)};J().prototype.pow=function(e){return this.throwIfDisposed(),$r(this,e)};J().prototype.prelu=function(e){return this.throwIfDisposed(),Uc(this,e)};J().prototype.prod=function(e,t){return this.throwIfDisposed(),Pv(this,e,t)};J().prototype.reciprocal=function(){return this.throwIfDisposed(),zv(this)};J().prototype.relu=function(){return this.throwIfDisposed(),Xe(this)};J().prototype.relu6=function(){return this.throwIfDisposed(),sf(this)};J().prototype.reshapeAs=function(e){return this.throwIfDisposed(),W(this,e.shape)};J().prototype.reshape=function(e){return this.throwIfDisposed(),W(this,e)};J().prototype.resizeBilinear=function(e,t,n){return this.throwIfDisposed(),qT(this,e,t,n)};J().prototype.resizeNearestNeighbor=function(e,t,n){return this.throwIfDisposed(),KT(this,e,t,n)};J().prototype.reverse=function(e){return this.throwIfDisposed(),ga(this,e)};J().prototype.rfft=function(){return this.throwIfDisposed(),jc(this)};J().prototype.round=function(){return this.throwIfDisposed(),of(this)};J().prototype.rsqrt=function(){return this.throwIfDisposed(),lf(this)};J().prototype.selu=function(){return this.throwIfDisposed(),uf(this)};J().prototype.separableConv2d=function(e,t,n,a,r,s){return this.throwIfDisposed(),Ts(this,e,t,n,a,r,s)};J().prototype.sigmoid=function(){return this.throwIfDisposed(),ha(this)};J().prototype.sign=function(){return this.throwIfDisposed(),Wv(this)};J().prototype.sin=function(){return this.throwIfDisposed(),pf(this)};J().prototype.sinh=function(){return this.throwIfDisposed(),cf(this)};J().prototype.slice=function(e,t){return this.throwIfDisposed(),Be(this,e,t)};J().prototype.softmax=function(e){return this.throwIfDisposed(),Ka(this,e)};J().prototype.softplus=function(){return this.throwIfDisposed(),mo(this)};J().prototype.spaceToBatchND=function(e,t){return this.throwIfDisposed(),Vc(this,e,t)};J().prototype.split=function(e,t){return this.throwIfDisposed(),zn(this,e,t)};J().prototype.sqrt=function(){return this.throwIfDisposed(),un(this)};J().prototype.square=function(){return this.throwIfDisposed(),lt(this)};J().prototype.squaredDifference=function(e){return this.throwIfDisposed(),mf(this,e)};J().prototype.squeeze=function(e){return this.throwIfDisposed(),Ns(this,e)};J().prototype.stack=function(e,t){this.throwIfDisposed();let n=e instanceof Te?[this,e]:[this,...e];return Ft(n,t)};J().prototype.step=function(e){return this.throwIfDisposed(),go(this,e)};J().prototype.stridedSlice=function(e,t,n,a,r,s,i,o){return this.throwIfDisposed(),Bv(this,e,t,n,a,r,s,i,o)};J().prototype.sub=function(e){return this.throwIfDisposed(),pe(this,e)};J().prototype.sum=function(e,t){return this.throwIfDisposed(),fe(this,e,t)};J().prototype.tan=function(){return this.throwIfDisposed(),Vv(this)};J().prototype.tanh=function(){return this.throwIfDisposed(),si(this)};J().prototype.tile=function(e){return this.throwIfDisposed(),Ln(this,e)};J().prototype.toBool=function(){return this.throwIfDisposed(),oe(this,"bool")};J().prototype.toFloat=function(){return this.throwIfDisposed(),oe(this,"float32")};J().prototype.toInt=function(){return this.throwIfDisposed(),oe(this,"int32")};J().prototype.topk=function(e,t){return this.throwIfDisposed(),Uv(this,e,t)};J().prototype.transpose=function(e){return this.throwIfDisposed(),Ee(this,e)};J().prototype.unique=function(e){return this.throwIfDisposed(),Gv(this,e)};J().prototype.unsortedSegmentSum=function(e,t){return this.throwIfDisposed(),gf(this,e,t)};J().prototype.unstack=function(e){return this.throwIfDisposed(),ct(this,e)};J().prototype.where=function(e,t){return this.throwIfDisposed(),mn(e,this,t)};J().prototype.zerosLike=function(){return this.throwIfDisposed(),qe(this)};var Ir=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Ir.prototype)}},Wa=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Wa.prototype)}},V=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,V.prototype)}},Me=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Me.prototype)}},sN=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,sN.prototype)}},iN=class{constructor(e){this.maxEntries=e||100,this.cache=new Map}get(e){let t;return this.cache.has(e)&&(t=this.cache.get(e),this.cache.delete(e),this.cache.set(e,t)),t}put(e,t){if(this.cache.has(e))this.cache.delete(e);else if(this.cache.size>=this.maxEntries){let n=this.cache.keys().next().value;this.cache.delete(n)}this.cache.set(e,t)}getMaxEntries(){return this.maxEntries}setMaxEntries(e){if(e<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${e}.`);if(this.maxEntries>e)for(let t=0;t<this.maxEntries-e;t++){let n=this.cache.keys().next().value;this.cache.delete(n)}this.maxEntries=e}};function li(e,t){if(Array.isArray(e)){let n=[];for(let a=0;a<t;a++)n=n.concat(e);return n}else{let n=new Array(t);return n.fill(e),n}}function ar(e,t){if(!e)throw new sN(t)}function _k(e,t){let n=0;for(let a of e)a===t&&n++;return n}function On(e){return e.length===1?e[0]:e}function xt(e){return Array.isArray(e)?e:[e]}function Sr(e){let t=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return t[0]!=="_"?t:"private"+t}function Us(e){return e.length<=1||e.indexOf("_")===-1?e:e.replace(/[_]+(\w|$)/g,(t,n)=>n.toUpperCase())}var Sa={};function ew(e){if(e==null)return null;let t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function Ub(e){if(!(e==null||typeof e!="object"))if(Array.isArray(e))e.forEach(t=>Ub(t));else{let t=Object.keys(e);for(let n of t){let a=e[n];a!=null&&typeof a=="object"&&(!Array.isArray(a)&&a.type==="ndarray"&&typeof a.value=="number"?e[n]=a.value:Ub(a))}}}function Kc(e,t={},n={},a="object",r=!1){if(typeof e=="string"){let s=e,i;if(s in n)i=n[s];else if(s in Sa)i=Sa[s];else if(i=t[s],i==null)throw new V(`Unknown ${a}: ${e}. This may be due to one of the following reasons:
-1. The ${a} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${a} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return i}else{let s=e;if(s.className==null||s.config==null)throw new V(`${a}: Improper config format: ${JSON.stringify(s)}.
-'className' and 'config' must set.`);let i=s.className,o,l;if(i in n?[o,l]=n[i]:i in Sa?[o,l]=Sa.className:i in t&&([o,l]=t[i]),o==null)throw new V(`Unknown ${a}: ${i}. This may be due to one of the following reasons:
-1. The ${a} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${a} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(l!=null){let u={};for(let h of Object.keys(Sa))u[h]=Sa[h];for(let h of Object.keys(n))u[h]=n[h];let p=s.config;p.customObjects=u;let d=Object.assign({},Sa);for(let h of Object.keys(n))Sa[h]=n[h];Ub(s.config);let c=l(o,s.config,n,r);return Sa=Object.assign({},d),c}else{let u=Object.assign({},Sa);for(let d of Object.keys(n))Sa[d]=n[d];let p=new o(s.config);return Sa=Object.assign({},u),p}}}function B4(e,t){return e<t?-1:e>t?1:0}function oh(e,t){return-1*B4(e,t)}function as(e){if(e==null)return e;let t=[];for(let n of e)t.indexOf(n)===-1&&t.push(n);return t}function V4(e){if(e==null)throw new V(`Invalid value in obj: ${JSON.stringify(e)}`);for(let t in e)if(e.hasOwnProperty(t))return!1;return!0}function yo(e,t,n){if(n!=null&&e.indexOf(n)<0)throw new V(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)}function tw(e,t,n=0,a=1/0){return ar(n>=0),ar(a>=n),Array.isArray(e)&&e.length>=n&&e.length<=a&&e.every(r=>typeof r===t)}function Qt(e,t){Array.isArray(e)?(v.assert(e.length>0,()=>`${t} is unexpectedly an empty array.`),e.forEach((n,a)=>Qt(n,`element ${a+1} of ${t}`))):v.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${oN(e)}.`)}function oN(e){return e===null?"null":Array.isArray(e)?"["+e.map(t=>oN(t)).join(",")+"]":typeof e=="string"?`"${e}"`:`${e}`}function U4(e,t,n){let a=n!=null?n():v.now(),r;return(...s)=>{let i=n!=null?n():v.now();return i-a<t||(a=i,r=e(...s)),r}}function lN(e){return e==="relu"?"relu":e==="linear"?"linear":e==="elu"?"elu":null}var G4=0;function uN(){return G4++}var lh={};function _f(e=""){return e in lh||(lh[e]=0),lh[e]+=1,e+lh[e].toString()}var H4=["channelsFirst","channelsLast"],j4=["nearest","bilinear"],q4=["valid","same","causal"],K4=["max","avg"],X4=["sum","mul","concat","ave"],al=new Map;function Rt(e){yo(H4,"DataFormat",e)}function Y4(e){yo(j4,"InterpolationFormat",e)}function xa(e){yo(q4,"PaddingMode",e)}function pN(e){yo(K4,"PoolMode",e)}var Hp=[],Ek="/";function Ys(e,t){Hp.push(e);try{let n=t();return Hp.pop(),n}catch(n){throw Hp.pop(),n}}function Z4(){return Hp.length===0?"":Hp.join(Ek)+Ek}function cN(e){if(!hN(e))throw new Error("Not a valid tensor name: '"+e+"'");return Z4()+e}function dN(e){if(!hN(e))throw new Error("Not a valid tensor name: '"+e+"'");al.has(e)||al.set(e,0);let t=al.get(e);if(al.set(e,al.get(e)+1),t>0){let n=`${e}_${t}`;return al.set(n,1),n}else return e}var J4=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function hN(e){return!!e.match(J4)}function Q4(e){return e===parseInt(e.toString(),10)}function rs(e,t,n){t==null&&(t=0),n==null&&(n=e.length);let a=1;for(let r=t;r<n;++r)a*=e[r];return a}function Il(e){if(e.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let n=0;n<e.length;n++){let a=e[n];a<t&&(t=a)}return t}function ls(e){if(e.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let n=0;n<e.length;n++){let a=e[n];a>t&&(t=a)}return t}function Ha(e,t){if(t<e)throw new V(`end (${t}) < begin (${e}) is forbidden.`);let n=[];for(let a=e;a<t;++a)n.push(a);return n}var bb;function Gt(){return bb==null&&(bb=AS().epsilon()),bb}function ja(){return"channelsLast"}function bo(e,t){return oe(e,t)}function Xc(e,t=-1){let n=e.shape.slice();return t<0&&(t=n.length+t+1),n.splice(t,0,1),W(e,n)}function eV(e,t){return P(()=>{if(e.shape.length!==2)throw new V(`repeat() expects a rank-2 tensor, but received a rank-${e.shape.length} tensor.`);let n=Xc(e,1);return Gb(n,[1,t,1])})}function tV(e){let t=[rs(e.shape)];return W(e,t)}function nV(e){if(e.rank<=1)throw new V(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);let t=[e.shape[0],rs(e.shape,1)];return W(e,t)}function Zs(e,t,n){return P(()=>{switch(e.rank){case 1:return Gc(e,t,n);case 2:return df(e,[t,0],[n,e.shape[1]]);case 3:return fo(e,[t,0,0],[n,e.shape[1],e.shape[2]]);case 4:return vl(e,[t,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3]]);case 5:return Be(e,[t,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return Be(e,[t,0,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new V(`sliceAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}})}function xb(e,t,n){return P(()=>{switch(e.rank){case 1:return Gc(e,t,n);case 2:return df(e,[0,t],[e.shape[0],n]);case 3:return fo(e,[0,0,t],[e.shape[0],e.shape[1],n]);case 4:return vl(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],n]);default:throw new V(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function uh(e,t,n,a){return P(()=>{switch(e.rank){case 1:return Gc(e,t,n);case 2:switch(a){case 1:return Zs(e,t,n);case 2:return xb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}case 3:switch(a){case 1:return Zs(e,t,n);case 2:return fo(e,[0,t,0],[e.shape[0],n,e.shape[2]]);case 3:return xb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}case 4:switch(a){case 1:return Zs(e,t,n);case 2:return vl(e,[0,t,0,0],[e.shape[0],n,e.shape[2],e.shape[3]]);case 3:return vl(e,[0,0,t,0],[e.shape[0],e.shape[1],n,e.shape[3]]);case 4:return xb(e,t,n);default:throw new V(`The axis is not within the rank of the tensor ${a}`)}default:throw new V(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function nw(e,t=-1){let n;return t<0&&(n=e[0].rank,n!==0?t=n:t=0),t===e[0].rank&&(t=-1),Ze(e,t)}function Ak(e,t){switch(e.rank){case 1:return pv([e,t]);case 2:return cv([e,t],0);case 3:return dv([e,t],0);case 4:return hv([e,t],0);default:throw new V(`concatAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}}function Gb(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new V(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return Ln(e,t)}function Ef(e,t=0,n=1,a,r){return rf(e,t,n,a,r)}function or(e,t,n,a){if(e.rank<2||t.rank<2)throw new Me(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){let r=e.shape.slice(-1)[0],s=t.shape.slice(-2)[0];if(r!==s)throw new Me(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)}if(e.rank===2&&t.rank===2)return kl.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:a?Hb(e.rank,a,ja()):null,activation:n});{let r=e.shape.slice(),s=r.pop();e=W(e,[-1,s]);let i=t.shape.slice(),o=i.pop(),l=i.pop(),u=[...i,o],p=Array.from({length:t.rank},(m,f)=>f===0?t.rank-2:f<=t.rank-2?f-1:f);t=W(Ee(t,p),[l,-1]);let d=[...r,...u],c=!1,h=!1;return W(kl.matMul({a:e,b:t,transposeA:c,transposeB:h,bias:a?Hb(e.rank,a,ja()):null,activation:n}),d)}}function mN(e,t,n){return P(()=>(Array.isArray(t)?t=Ke(t,"int32"):t=oe(t,"int32"),Bu(e,t,n)))}function Yc(e){return z(e,e)}function Hb(e,t,n){let a=t.shape;if(t.rank!==1&&t.rank!==e)throw new V(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(e===5){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1,1,1]):W(t,[1,a[3],a[0],a[1],a[2]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,1,1,a[0]]):W(t,[1].concat(a))}else if(e===4){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1,1]):W(t,[1,a[2],a[0],a[1]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,1,a[0]]):W(t,[1].concat(a))}else if(e===3){if(n==="channelsFirst")return a.length===1?W(t,[1,a[0],1]):W(t,[1,a[1],a[0]]);if(n==="channelsLast")return a.length===1?W(t,[1,1,a[0]]):W(t,[1].concat(a))}else if(e<3)return t;throw new V(`Unsupported input rank by biasAdd: ${t.rank}`)}function Xa(e,t,n){return P(()=>(n==null&&(n=ja()),Rt(n),Y(e,Hb(e.rank,t,n))))}function aV(e,t=1){if(t!==1)throw new Me(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Lu(e)}function rV(e){return P(()=>he(e,Y(Lt(e),1)))}function fN(e,t,n,a){return P(()=>qv(e,t,n,a))}function sV(e){return P(()=>{let t=Y(.5,z(.2,e));return en(t,0,1)})}function Zc(e,t,n=!1){return n?e():t()}var iV=["fanIn","fanOut","fanAvg"],oV=["normal","uniform","truncatedNormal"];function lV(e){yo(iV,"FanMode",e)}function uV(e){yo(oV,"Distribution",e)}var Ra=class extends ne.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},aw=class extends Ra{apply(e,t){return It(e,t)}};aw.className="Zeros";ne.registerClass(aw);var Af=class extends Ra{apply(e,t){return Jn(e,t)}};Af.className="Ones";ne.registerClass(Af);var rw=class extends Ra{constructor(e){if(super(),typeof e!="object")throw new V(`Expected argument of type ConstantConfig but got ${e}`);if(e.value===void 0)throw new V(`config must have value set but got ${e}`);this.value=e.value}apply(e,t){return P(()=>z(be(this.value),Jn(e,t)))}getConfig(){return{value:this.value}}};rw.className="Constant";ne.registerClass(rw);var sw=class extends Ra{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return Uu(e,this.minval,this.maxval,t)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};sw.className="RandomUniform";ne.registerClass(sw);var iw=class extends Ra{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Me(`randomNormal does not support dType ${t}.`);return Ef(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};iw.className="RandomNormal";ne.registerClass(iw);var ow=class extends Ra{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Me(`truncatedNormal does not support dType ${t}.`);return ff(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};ow.className="TruncatedNormal";ne.registerClass(ow);var lw=class extends Ra{constructor(e){super(),this.gain=e.gain!=null?e.gain:1}apply(e,t){return P(()=>{if(e.length!==2||e[0]!==e[1])throw new V("Identity matrix initializer can only be used for 2D square matrices.");return z(this.gain,Zm(e[0]))})}getConfig(){return{gain:this.gain}}};lw.className="Identity";ne.registerClass(lw);function pV(e,t="channelsLast"){let n,a;if(Rt(t),e.length===2)n=e[0],a=e[1];else if([3,4,5].indexOf(e.length)!==-1){if(t==="channelsFirst"){let r=rs(e,2);n=e[1]*r,a=e[0]*r}else if(t==="channelsLast"){let r=rs(e,0,e.length-2);n=e[e.length-2]*r,a=e[e.length-1]*r}}else{let r=rs(e);n=Math.sqrt(r),a=Math.sqrt(r)}return[n,a]}var Bn=class extends Ra{constructor(e){if(super(),e.scale<0)throw new V(`scale must be a positive float. Got: ${e.scale}`);this.scale=e.scale==null?1:e.scale,this.mode=e.mode==null?"fanIn":e.mode,lV(this.mode),this.distribution=e.distribution==null?"normal":e.distribution,uV(this.distribution),this.seed=e.seed}apply(e,t){let n=pV(e),a=n[0],r=n[1],s=this.scale;if(this.mode==="fanIn"?s/=Math.max(1,a):this.mode==="fanOut"?s/=Math.max(1,r):s/=Math.max(1,(a+r)/2),this.distribution==="normal"){let i=Math.sqrt(s);if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Me(`${this.getClassName()} does not support dType ${t}.`);return ff(e,0,i,t,this.seed)}else{let i=Math.sqrt(3*s);return Uu(e,-i,i,t)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};Bn.className="VarianceScaling";ne.registerClass(Bn);var $f=class extends Bn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};$f.className="GlorotUniform";ne.registerClass($f);var Ff=class extends Bn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};Ff.className="GlorotNormal";ne.registerClass(Ff);var Df=class extends Bn{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};Df.className="HeNormal";ne.registerClass(Df);var Rf=class extends Bn{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};Rf.className="HeUniform";ne.registerClass(Rf);var Mf=class extends Bn{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};Mf.className="LeCunNormal";ne.registerClass(Mf);var Pf=class extends Bn{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Bn.className}};Pf.className="LeCunNormal";ne.registerClass(Pf);var uw=class extends Ra{constructor(e){if(super(),this.DEFAULT_GAIN=1,this.gain=e.gain==null?this.DEFAULT_GAIN:e.gain,this.seed=e.seed,this.seed!=null)throw new Me("Random seed is not implemented for Orthogonal Initializer yet.")}apply(e,t){return P(()=>{if(e.length<2)throw new Me("Shape must be at least 2D.");e[0]*e[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);let n=e[0]>e[1]?[e[1],e[0]]:e,a=Ef(n,0,1,"float32"),r=Zv.gramSchmidt(a);return e[0]>e[1]&&(r=Ee(r)),z(this.gain,r)})}getConfig(){return{gain:this.gain,seed:this.seed}}};uw.className="Orthogonal";ne.registerClass(uw);var $k={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function Fk(e,t={}){return Kc(e,ne.SerializationMap.getMap().classNameMap,t,"initializer")}function Ct(e){return ew(e)}function St(e){if(typeof e=="string"){let t=e in $k?$k[e]:e;if(t==="GlorotNormal")return new Ff;if(t==="GlorotUniform")return new $f;if(t==="HeNormal")return new Df;if(t==="HeUniform")return new Rf;if(t==="LeCunNormal")return new Mf;if(t==="LeCunUniform")return new Pf;{let n={};return n.className=t,n.config={},Fk(n)}}else return e instanceof Ra?e:Fk(e)}function jb(e){return Array.isArray(e)&&Array.isArray(e[0])}function Wh(e){return e.length===0?[]:Array.isArray(e[0])?e:[e]}function Ne(e){let t;if(Array.isArray(e)){if(e.length!==1)throw new V(`Expected Tensor length to be 1; got ${e.length}`);t=e[0]}else t=e;return t}function tt(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(e.length===1)return e=e,e[0];throw new V(`Expected exactly 1 Shape; got ${e.length}`)}else return e}function Bh(e){let t=0;for(let n of e)n.shape.length===0?t+=1:t+=n.shape.reduce((a,r)=>a*r);return t}var Dk="Variable",gN=class{constructor(e,t="float32",n=Dk,a=!0,r=null){this.dtype=t==null?"float32":t,this.shape=e.shape,this.id=uN(),n=n==null?Dk:n,this.originalName=cN(n),this.name=dN(this.originalName),this.trainable_=a,this.constraint=r,this.val=Hv(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),cV(this.val,e),this.val.id!==e.id&&(this.val.assign(e),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function cV(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}function qb(e){return e.map(t=>t.read())}function pw(e){e.forEach(t=>{t[0].write(t[1])})}var zt=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,e.shape!=null?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},Ba=class{constructor(e,t,n,a,r,s,i){this.dtype=e,this.shape=t,this.sourceLayer=n,this.inputs=a,this.callArgs=r,this.outputTensorIndex=i,this.id=uN(),s!=null&&(this.originalName=cN(s),this.name=dN(this.originalName)),this.rank=t.length}},dV=0,Of=class{constructor(e,t){this.callArgs=t,this.id=dV++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(let n of e.inboundLayers)n!=null&&n.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){let e=[];for(let t of this.inboundLayers)t!=null?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},hV=0,Ge=class extends ne.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=hV++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){let n=this.getClassName();t=Sr(n)+"_"+_f(n)}if(this.name=t,this.trainable_=e.trainable==null?!0:e.trainable,e.inputShape!=null||e.batchInputShape!=null){let n;if(e.batchInputShape!=null)n=e.batchInputShape;else if(e.inputShape!=null){let r=null;e.batchSize!=null&&(r=e.batchSize),n=[r].concat(e.inputShape)}this.batchInputShape=n;let a=e.dtype;a==null&&(a=e.inputDType),a==null&&(a="float32"),this.dtype=a}e.weights!=null?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(this.inboundNodes.length===0)throw new Wa(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new V(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return On(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return On(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new Ir(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(this.inboundNodes.length===0)throw new Ir(`Layer ${this.name} is not connected, no input to return.`);return On(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new Ir(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new Ir(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return On(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){if(e=xt(e),this.inputSpec==null||this.inputSpec.length===0)return;let t=xt(this.inputSpec);if(e.length!==t.length)throw new V(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);for(let n=0;n<e.length;n++){let a=e[n],r=t[n];if(r==null)continue;let s=a.rank;if(r.ndim!=null&&s!==r.ndim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${r.ndim}, found ndim=${s}`);if(r.maxNDim!=null&&s>r.maxNDim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${r.maxNDim}, found ndim=${s}`);if(r.minNDim!=null&&s<r.minNDim)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${r.minNDim}, found ndim=${s}.`);if(r.dtype!=null&&a.dtype!==r.dtype)throw new V(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${r.dtype}, found dtype=${a.dtype}.`);if(r.axes){let i=a.shape;for(let o in r.axes){let l=Number(o),u=r.axes[o],p=l>=0?i[l]:i[i.length+l];if(u!=null&&[u,null].indexOf(p)===-1)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected axis ${l} of input shape to have value ${u} but got shape ${i}.`)}}if(r.shape!=null)for(let i=0;i<r.shape.length;++i){let o=r.shape[i],l=a.shape[i];if(o!=null&&l!=null&&o!==l)throw new V(`Input ${n} is incompatible with layer ${this.name}: expected shape=${r.shape}, found shape=${a.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){this._callHook!=null&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();let n=xt(e),a=!0;for(let s of n)if(!(s instanceof Ba)){a=!1;break}let r=!0;for(let s of n)if(s instanceof Ba){r=!1;break}if(a===r)throw new V("Arguments to apply() must be all SymbolicTensors or all Tensors");return Ys(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);let s=[];for(let i of xt(e))s.push(i.shape);this.build(On(s)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&r&&(this._refCount=1)}if(this.assertInputCompatibility(e),r){let s=this.call(e,t),i=xt(s),o=[];for(let l of i)n.indexOf(l)!==-1&&(l=l.clone()),o.push(l);if(s=On(o),this.activityRegularizer!=null)throw new Me("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return s}else{let s=mV(e),i=this.computeOutputShape(s),o,l=fV(e);if(this.warnOnIncompatibleInputShape(Array.isArray(e)?s[0]:s),i!=null&&i.length>0&&Array.isArray(i[0])?o=i.map((u,p)=>new Ba(l,u,this,xt(e),t,this.name,p)):o=new Ba(l,i,this,xt(e),t,this.name),this.addInboundNode(e,o,null,null,s,i,t),this._refCount++,this.activityRegularizer!=null)throw new Me("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return o}})}warnOnIncompatibleInputShape(e){if(this.batchInputShape!=null)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((n,a)=>{n!=null&&e[a]!=null&&e[a]!==n&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new Ir(`The layer ${this.name} has never been called and thus has no defined output shape.`);let e=[];for(let t of this.inboundNodes){let n=JSON.stringify(t.outputShapes);e.indexOf(n)===-1&&e.push(n)}if(e.length===1){let t=this.inboundNodes[0].outputShapes;return Array.isArray(t)&&Array.isArray(t[0])&&t.length===1?t[0]:t}else throw new Ir(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new Wa(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Bh(this.weights)}build(e){this.built=!0}getWeights(e=!1){return qb(e?this.trainableWeights:this.weights)}setWeights(e){P(()=>{let t=this.weights;if(t.length!==e.length)throw new V(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(t.length===0)return;let n=[],a=qb(t);for(let r=0;r<a.length;++r){let s=a[r],i=t[r],o=e[r];if(!v.arraysEqual(s.shape,o.shape))throw new V(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);n.push([i,o])}pw(n)})}addWeight(e,t,n,a,r,s,i,o){if(this._addedWeightNames.indexOf(e)!==-1)throw new V(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),n==null&&(n="float32"),this.fastWeightInitDuringBuild&&(a=o!=null?o():St("zeros"));let l=a.apply(t,n),u=new gN(l,n,e,s,i);return l.dispose(),r!=null&&this.addLoss(()=>r.apply(u.read())),s==null&&(s=!0),s?this._trainableWeights.push(u):this._nonTrainableWeights.push(u),u}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){e==null||Array.isArray(e)&&e.length===0||(e=xt(e),this._losses!==void 0&&this._losses!==null&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(t!=null)if(Array.isArray(t))t.forEach(n=>{if(n!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return t}addInboundNode(e,t,n,a,r,s,i=null){let o=xt(e);t=xt(t),n=xt(n),a=xt(a),r=Wh(r),s=Wh(s);let l=[],u=[],p=[];for(let d of o)l.push(d.sourceLayer),u.push(d.nodeIndex),p.push(d.tensorIndex);new Of({outboundLayer:this,inboundLayers:l,nodeIndices:u,tensorIndices:p,inputTensors:o,outputTensors:t,inputMasks:n,outputMasks:a,inputShapes:r,outputShapes:s},i);for(let d=0;d<t.length;d++)t[d].sourceLayer=this,t[d].nodeIndex=this.inboundNodes.length-1,t[d].tensorIndex=d}getConfig(){let e={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(e.batchInputShape=this.batchInputShape),this.dtype!=null&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return--this._refCount===0&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};function mV(e){e=xt(e);let t=[];for(let n of e)t.push(n.shape);return On(t)}function fV(e){return"float32"}function yN(e,t,n){if((t==null||n!=null&&n>0)&&(t=e.sourceLayer,n=e.nodeIndex),t.inboundNodes.length===0)return[e];{let a=t.inboundNodes[n];if(a.inboundLayers.length===0)return a.inputTensors;{let r=[];for(let s=0;s<a.inboundLayers.length;s++){let i=a.inputTensors[s],o=a.inboundLayers[s],l=a.nodeIndices[s],u=yN(i,o,l);for(let p of u)r.indexOf(p)===-1&&r.push(p)}return r}}}var Hu=class extends Ge{constructor(e){if(super({dtype:e.dtype,name:e.name!=null?e.name:_f("input").toString()}),e.batchSize==null&&(e.batchSize=null),e.sparse==null&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,e.inputShape!=null&&e.batchInputShape!=null)throw new V("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(t==null){if(e.inputShape==null)throw new V("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(e.batchSize!=null)throw new V("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let n=e.dtype||"float32";this.batchInputShape=t,this.dtype=n,this.inputSpec=[{shape:t}];let a=new Ba(this.dtype,this.batchInputShape,this,[],{},this.name);a.nodeIndex=0,a.tensorIndex=0,new Of({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[a],outputTensors:[a],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new V(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};Hu.className="InputLayer";ne.registerClass(Hu);function bN(e){if(e.batchShape==null&&e.shape==null)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(e.batchShape!=null&&e.shape!=null)throw new V("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;e.shape!=null&&t==null&&(t=[null].concat(e.shape));let n=e.dtype;return n==null&&(n="float32"),new Hu({batchInputShape:t,name:e.name,dtype:n,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}function gV(e,t){if(e.dtype==null||e.dtype===t.dtype)return t;try{return oe(t,e.dtype)}catch(n){throw new V(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}var js=class{constructor(e){if(this.id2Value={},this.id2Mask={},this.name2Id={},e instanceof js)for(let t in e.id2Value)this.id2Value[t]=e.id2Value[t],t in e.id2Mask&&(this.id2Mask[t]=e.id2Mask[t]);else{if(e==null)return;for(let t of e)this.add(t.key,t.value)}}add(e,t,n){if(this.id2Value[e.id]==null)this.id2Value[e.id]=gV(e,t),this.name2Id[e.name]=e.id,n!=null&&(this.id2Mask[e.id]=n);else throw new V(`Duplicate key: name=${e.name}, id=${e.id}`);return this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return this.id2Value[e.id]!=null}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof Ba){if(this.id2Value[e.id]==null)throw new V(`Nonexistent key: ${e.name}`);return this.id2Value[e.id]}else{let t=this.name2Id[e];if(t==null)throw new V(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Value[t]}}getMask(e){if(e instanceof Ba){if(this.id2Value[e.id]==null)throw new V(`Nonexistent key: ${e.name}`);return this.id2Mask[e.id]}else{let t=this.name2Id[e];if(t==null)throw new V(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Mask[t]}}disposeMasks(){this.id2Mask!=null&&_e(this.id2Mask)}},Vh=new iN,Uh=new iN;function yV(e){Vh!=null&&Vh.setMaxEntries(e),Uh!=null&&Uh.setMaxEntries(e)}function Mp(e,t,n,a){let r=n==null?!1:n.training,s=Array.isArray(e),i=s?e:[e],o=i.map(m=>m.name),l=[],u=t.names();for(let m of o)u.indexOf(m)!==-1?l.push(t.getValue(m)):l.push(null);a!=null&&(a.maxNumTensors=-1/0,a.minNumTensors=1/0);let p=o.join(",")+"|"+t.names().sort().join(","),d=Vh.get(p),c;if(d==null){let m=bV(i,t);d=m.sorted,c=m.recipientCounts,Vh.put(p,d),Uh.put(p,c)}c={},r||Object.assign(c,Uh.get(p));let h=new js(t);for(let m=0;m<d.length;++m){if(a!=null){let A=Oh().numTensors;A>a.maxNumTensors&&(a.maxNumTensors=A),A<a.minNumTensors&&(a.minNumTensors=A)}let f=d[m],g=f.sourceLayer;if(g instanceof Hu)continue;let y=[],b=[],x=[],w=!1;for(let A of f.inputs){let R=h.getValue(A),F=h.getMask(A);y.push(R),b.push(F),F!=null&&(w=!0),r||(c[A.name]--,c[A.name]===0&&!t.hasKey(A)&&o.indexOf(A.name)===-1&&!R.isDisposed&&A.sourceLayer.stateful!==!0&&x.push(R))}w&&(n=n||{},n.mask=b[0]);let I=xt(g.apply(y,n)),T=null;g.supportsMasking&&(T=g.computeMask(y,b));let C=vV(f),E=Array.isArray(C)?C:[C];for(let A=0;A<E.length;++A){h.hasKey(E[A])||h.add(E[A],I[A],Array.isArray(T)?T[0]:T);let R=o.indexOf(E[A].name);R!==-1&&(l[R]=I[A])}r||_e(x)}return h.disposeMasks(),s?l:l[0]}function bV(e,t){v.assert(e!=null&&e.length>0,()=>"Expected at least one fetch, got none");let n=[],a={};if(e.length===1){let r=Rk(e[0],t);n=r.sorted,a=r.recipientMap}else{let r=new Set;for(let s of e){let{sorted:i,recipientMap:o}=Rk(s,t);for(let l of i)r.has(l.name)||(n.push(l),r.add(l.name));for(let l in o)a[l]==null&&(a[l]=new Set),o[l].forEach(u=>a[l].add(u))}}return{sorted:n,recipientCounts:xV(a)}}function xV(e){let t={};for(let n in e)t[n]=e[n].size;return t}function Rk(e,t){let n=new Set,a=[],r={};for(let o of t.names())n.add(o);let s=[],i=[];for(s.push(e);s.length>0;){let o=s[s.length-1];if(n.has(o.name)){s.pop();continue}let l=i[i.length-1]===s.length-1;if(o.inputs.length===0||l)s.pop(),a.push(o),n.add(o.name),l&&i.pop();else{i.push(s.length-1);for(let u of o.inputs)r[u.name]==null&&(r[u.name]=new Set),r[u.name].add(o.name),!n.has(u.name)&&s.push(u)}}return{sorted:a,recipientMap:r}}function vV(e){let t;if(e.sourceLayer.inboundNodes.length===1)t=e.sourceLayer.output;else{let n=null;for(let a=0;a<e.sourceLayer.inboundNodes.length;++a)for(let r of e.sourceLayer.inboundNodes[a].outputTensors)if(r.id===e.id){n=a;break}t=e.sourceLayer.getOutputAt(n)}return t}var wV=H();wV.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,yV);var xN={};Ae(xN,{maxNorm:()=>kV,minMaxNorm:()=>TV,nonNeg:()=>SV,unitNorm:()=>IV});function cw(e,t){return P(()=>un(fe(z(e,e),t,!0)))}var Jc=class extends ne.Serializable{getConfig(){return{}}},dw=class extends Jc{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>{let t=cw(e,this.axis),n=en(t,0,this.maxValue);return z(e,he(n,Y(Gt(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};dw.className="MaxNorm";ne.registerClass(dw);var hw=class extends Jc{constructor(e){super(),this.defaultAxis=0,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>he(e,Y(Gt(),cw(e,this.axis))))}getConfig(){return{axis:this.axis}}};hw.className="UnitNorm";ne.registerClass(hw);var mw=class extends Jc{apply(e){return Xe(e)}};mw.className="NonNeg";ne.registerClass(mw);var fw=class extends Jc{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=e.minValue!=null?e.minValue:this.defaultMinValue,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.rate=e.rate!=null?e.rate:this.defaultRate,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return P(()=>{let t=cw(e,this.axis),n=Y(z(this.rate,en(t,this.minValue,this.maxValue)),z(1-this.rate,t));return z(e,he(n,Y(Gt(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};fw.className="MinMaxNorm";ne.registerClass(fw);var Mk={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function qt(e){return ew(e)}function Pk(e,t={}){return Kc(e,ne.SerializationMap.getMap().classNameMap,t,"constraint")}function Kt(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in Mk?Mk[e]:e,config:{}};return Pk(t)}else return e instanceof Jc?e:Pk(e)}function kV(e){return new dw(e)}function IV(e){return new hw(e)}function SV(){return new mw}function TV(e){return new fw(e)}var vN={};Ae(vN,{constant:()=>_V,glorotNormal:()=>MV,glorotUniform:()=>RV,heNormal:()=>PV,heUniform:()=>OV,identity:()=>FV,leCunNormal:()=>LV,leCunUniform:()=>zV,ones:()=>CV,orthogonal:()=>WV,randomNormal:()=>AV,randomUniform:()=>EV,truncatedNormal:()=>$V,varianceScaling:()=>DV,zeros:()=>NV});function NV(){return new aw}function CV(){return new Af}function _V(e){return new rw(e)}function EV(e){return new sw(e)}function AV(e){return new iw(e)}function $V(e){return new ow(e)}function FV(e){return new lw(e)}function DV(e){return new Bn(e)}function RV(e){return new $f(e)}function MV(e){return new Ff(e)}function PV(e){return new Df(e)}function OV(e){return new Rf(e)}function LV(e){return new Mf(e)}function zV(e){return new Pf(e)}function WV(e){return new uw(e)}var wN={};Ae(wN,{Layer:()=>Ge,RNN:()=>mr,RNNCell:()=>ad,activation:()=>hG,add:()=>kG,alphaDropout:()=>sH,average:()=>IG,averagePooling1d:()=>T0,averagePooling2d:()=>N0,averagePooling3d:()=>C0,avgPool1d:()=>FG,avgPool2d:()=>RG,avgPool3d:()=>PG,avgPooling1d:()=>DG,avgPooling2d:()=>MG,avgPooling3d:()=>OG,batchNormalization:()=>EG,bidirectional:()=>ZG,categoryEncoding:()=>uH,concatenate:()=>SG,conv1d:()=>rG,conv2d:()=>sG,conv2dTranspose:()=>iG,conv3d:()=>oG,conv3dTranspose:()=>lG,convLstm2d:()=>qG,convLstm2dCell:()=>KG,cropping2D:()=>pG,dense:()=>mG,depthwiseConv2d:()=>dG,dot:()=>_G,dropout:()=>fG,elu:()=>JU,embedding:()=>wG,flatten:()=>yG,gaussianDropout:()=>rH,gaussianNoise:()=>aH,globalAveragePooling1d:()=>LG,globalAveragePooling2d:()=>zG,globalMaxPool1d:()=>QG,globalMaxPool2d:()=>eH,globalMaxPooling1d:()=>d2,globalMaxPooling2d:()=>h2,gru:()=>BG,gruCell:()=>VG,input:()=>zN,inputLayer:()=>ZU,layerNormalization:()=>AG,leakyReLU:()=>eG,lstm:()=>UG,lstmCell:()=>GG,masking:()=>iH,maxPool1d:()=>tH,maxPool2d:()=>nH,maxPooling1d:()=>m2,maxPooling2d:()=>f2,maxPooling3d:()=>WG,maximum:()=>TG,minimum:()=>NG,multiply:()=>CG,permute:()=>vG,prelu:()=>tG,reLU:()=>QU,repeatVector:()=>bG,rescaling:()=>oH,reshape:()=>xG,resizing:()=>lH,rnn:()=>XG,separableConv2d:()=>uG,simpleRNN:()=>HG,simpleRNNCell:()=>jG,softmax:()=>nG,spatialDropout1d:()=>gG,stackedRNNCells:()=>YG,thresholdedReLU:()=>aG,timeDistributed:()=>JG,upSampling2d:()=>cG,zeroPadding2d:()=>$G});async function Zr(e){if(e==null)return;let t=[],n=[],a=[];for(let r in e){let s=e[r];if(typeof s!="number"){let i=s;t.push(i.data()),n.push(r),a.push(i)}}if(t.length>0){let r=await Promise.all(t);for(let s=0;s<r.length;++s)e[n[s]]=r[s][0];_e(a)}}function kN(e){if(e!=null)for(let t in e){let n=e[t];typeof n!="number"&&n.dispose()}}var Ok;(function(e){e[e.SILENT=0]="SILENT",e[e.VERBOSE=1]="VERBOSE"})(Ok||(Ok={}));var BV=125,Sl=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},IN=class{constructor(e,t=10){e==null&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(let t of this.callbacks)t.setParams(e)}setModel(e){for(let t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochBegin(e,t)}async onEpochEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochEnd(e,t)}async onBatchBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchBegin(e,t)}async onBatchEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchEnd(e,t)}async onTrainBegin(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainEnd(e)}},VV=class extends Sl{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){t==null&&(t={});let n=t.size==null?0:t.size;this.seen+=n;for(let a in t){let r=t[a];if(typeof r=="number")this.totals.hasOwnProperty(a)||(this.totals[a]=0),this.totals[a]=this.totals[a]+r*n;else{let s;a in this.totals?s=this.totals[a]:this.totals[a]=0;let i=P(()=>Y(this.totals[a],z(r,n)));this.totals[a]=i,s!=null&&s.dispose()}}}async onEpochEnd(e,t){if(t!=null)for(let n of this.params.metrics)this.totals[n]!=null&&(typeof this.totals[n]=="number"?t[n]=this.totals[n]/this.seen:P(()=>{let a=z(he(1,this.seen),this.totals[n]);t[n]=a,this.totals[n].dispose(),Jt(t[n])}))}},SN=class extends Sl{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){t==null&&(t={}),this.epoch.push(e);for(let n in t)this.history[n]==null&&(this.history[n]=[]),this.history[n].push(t[n])}async syncData(){let e=[],t=[],n=[];for(let r in this.history){let s=this.history[r];for(let i=0;i<s.length;++i)if(typeof s[i]!="number"){let o=s[i];e.push(o.data()),t.push(r),n.push(i)}}let a=await Promise.all(e);for(let r=0;r<a.length;++r)this.history[t[r]][n[r]].dispose(),this.history[t[r]][n[r]]=a[r][0]}},TN=class extends Sl{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||Jv,this.yieldEvery=t||"auto",this.yieldEvery==="auto"&&(this.yieldEvery=BV),this.yieldEvery==="never"&&e.onYield!=null)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");v.isNumber(this.yieldEvery)&&(this.maybeWait=U4(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,n){let a=[];this.yield!=null&&(await Zr(n),a.push(this.yield(e,t,n))),a.push(this.nextFrameFunc()),await Promise.all(a)}async onEpochBegin(e,t){this.currentEpoch=e,this.epochBegin!=null&&(await Zr(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){let n=[];this.epochEnd!=null&&(await Zr(t),n.push(this.epochEnd(e,t))),this.yieldEvery==="epoch"&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(e,t){this.batchBegin!=null&&(await Zr(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){let n=[];this.batchEnd!=null&&(await Zr(t),n.push(this.batchEnd(e,t))),this.yieldEvery==="batch"?n.push(this.nextFrameFunc()):v.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(n)}async onTrainBegin(e){this.trainBegin!=null&&(await Zr(e),await this.trainBegin(e))}async onTrainEnd(e){this.trainEnd!=null&&(await Zr(e),await this.trainEnd(e))}};function NN(e,t){return e==null&&(e={}),e instanceof Sl?[e]:Array.isArray(e)&&e[0]instanceof Sl?e:xt(e).map(n=>new TN(n,t))}var Ca=class{constructor(){}static registerCallbackConstructor(e,t){v.assert(e>=0&&Number.isInteger(e),()=>`Verbosity level is expected to be an integer >= 0, but got ${e}`),Ca.checkForDuplicate(t),Ca.constructors[e]==null&&(Ca.constructors[e]=[]),Ca.constructors[e].push(t)}static checkForDuplicate(e){for(let t in Ca.constructors)Ca.constructors[+t].forEach(n=>{if(n===e)throw new V("Duplicate callback constructor.")})}static clear(){Ca.constructors={}}static createCallbacks(e){let t=[];for(let n in Ca.constructors){let a=+n;e>=a&&t.push(...Ca.constructors[a])}return t.map(n=>new n)}};Ca.constructors={};function CN(e,t,n,a,r,s,i,o,l){let u=new SN,p=[new VV,...Ca.createCallbacks(t)];e!=null&&p.push(...e),p.push(u);let d=new IN(p);return d.setParams({epochs:n,initialEpoch:a,samples:r,steps:s,batchSize:i,verbose:t,doValidation:o,metrics:l}),{callbackList:d,history:u}}function Ua(e,t={},n=!1){return Kc(e,ne.SerializationMap.getMap().classNameMap,t,"layer",n)}function Gh(e,t){return P(()=>{e.dtype!=="float32"&&(e=oe(e,"float32"));let n=fe(Yc(e),t,!0),a=gn(n.shape,Gt()),r=un(dr(n,a));return he(e,r)})}function xo(e,t){return P(()=>Nt(Yc(pe(t,e)),-1))}function Lf(e,t){return P(()=>Nt(Lt(pe(t,e)),-1))}function ju(e,t){return P(()=>{let n=pe(e,t),a=en(Lt(e),Gt(),Number.MAX_VALUE),r=Lt(he(n,a));return z(100,Nt(r,-1))})}function UV(e,t){return P(()=>{let n=en(t,Gt(),Number.MAX_VALUE),a=ta(Y(1,n)),r=en(e,Gt(),Number.MAX_VALUE),s=ta(Y(1,r));return Nt(Yc(pe(a,s)),-1)})}function GV(e,t){return P(()=>{let n=dr(0,pe(1,z(e,t)));return Nt(Yc(n),-1)})}function HV(e,t){return P(()=>{let n=dr(0,pe(1,z(e,t)));return Nt(n,-1)})}function jV(e,t){return P(()=>{let n=fe(z(e,t),-1),a=ma(z(pe(1,e),t),-1);return dr(0,Y(1,pe(a,n)))})}function qV(e,t){return P(()=>{let n=Math.log(2),a=pe(t,e),r=pe(Y(a,mo(z(-2,a))),n);return Nt(r,-1)})}function tc(e,t,n=!1){return P(()=>{if(n)t=Ka(t);else{let a=fe(t,t.shape.length-1,!0);t=he(t,a)}return t=en(t,Gt(),1-Gt()),vt(fe(z(oe(e,"float32"),ta(t)),t.shape.length-1))})}function Hh(e,t,n=!1){return P(()=>{let a=oe(Wu(tV(e)),"int32");t=en(t,Gt(),1-Gt());let r=t.shape,s=W(gl(a,r[r.length-1]),r);return tc(s,t,n)})}function KV(e,t){if(!v.arraysEqual(e.shape,t.shape))throw new V(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return P(()=>{let n=Xe(t),a=vt(Lt(t));return Y(pe(n,z(t,e)),zc(fn(a)))})}function zf(e,t){return P(()=>{let n;return n=en(t,Gt(),1-Gt()),n=ta(he(n,pe(1,n))),Nt(KV(e,n),-1)})}function XV(e,t){return P(()=>{let n=en(e,Gt(),1),a=en(t,Gt(),1);return fe(z(e,ta(he(n,a))),-1)})}function YV(e,t){return P(()=>{let n=ta(Y(Gt(),t));return Nt(pe(t,z(e,n)),-1)})}function gw(e,t){return P(()=>{let n=Gh(e,-1),a=Gh(t,-1),r=z(n,a);return vt(fe(r,-1))})}var jh={meanSquaredError:xo,meanAbsoluteError:Lf,meanAbsolutePercentageError:ju,meanSquaredLogarithmicError:UV,squaredHinge:GV,hinge:HV,categoricalHinge:jV,logcosh:qV,categoricalCrossentropy:tc,sparseCategoricalCrossentropy:Hh,binaryCrossentropy:zf,kullbackLeiblerDivergence:XV,poisson:YV,cosineProximity:gw};function vb(e){if(typeof e=="string"){if(e in jh)return jh[e];let t=`Unknown loss ${e}`;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new V(t)}else return e}function yw(e,t){return P(()=>{let n=z(.5,na(t)),a=bo(Cn(t,n),e.dtype);return Nt(ea(e,a),-1)})}function bw(e,t){return P(()=>bo(ea(ri(e,-1),ri(t,-1)),"float32"))}function _N(e,t){return P(()=>oe(fe($a(ea(e,1),ea(t,1))),"float32"))}function ZV(e,t){return P(()=>oe(fe($a(ea(e,1),ea(t,0))),"float32"))}function JV(e,t){return P(()=>oe(fe($a(ea(e,0),ea(t,1))),"float32"))}function EN(e,t){return P(()=>{let n=_N(e,t),a=JV(e,t),r=Y(n,a);return oe(mn(Cn(r,0),he(n,r),0),"float32")})}function QV(e,t){return P(()=>{let n=_N(e,t),a=ZV(e,t),r=Y(n,a);return oe(mn(Cn(r,0),he(n,r),0),"float32")})}function AN(e,t){return zf(e,t)}function $N(e,t){return e.rank===t.rank&&(e=Ns(e,[e.rank-1])),t=ri(t,-1),t.dtype!==e.dtype&&(t=oe(t,e.dtype)),oe(ea(e,t),"float32")}var eU=xo,tU=xo,nU=Lf,aU=Lf,rU=ju,sU=ju,xw=tc,iU=gw,FN=Hh,qh={binaryAccuracy:yw,categoricalAccuracy:bw,precision:EN,categoricalCrossentropy:xw,sparseCategoricalCrossentropy:FN,mse:eU,MSE:tU,mae:nU,MAE:aU,mape:rU,MAPE:sU,cosine:iU};function oU(e){if(typeof e=="string"&&e in qh)return qh[e];if(typeof e!="string"&&e!=null)return e;throw new V(`Unknown metric ${e}`)}function ph(e){if(ar(e!==null,`Unknown LossOrMetricFn ${e}`),typeof e=="string")return e;{let t;for(let n of Object.keys(jh))if(jh[n]===e){t=n;break}if(t!==void 0)return t;for(let n of Object.keys(qh))if(qh[n]===e){t=n;break}return t!==void 0?t:e.name}}function lU(e){let t={Adagrad:()=>Ws.adagrad(.01),Adadelta:()=>Ws.adadelta(1,.95,Gt()),Adam:()=>Ws.adam(.001,.9,.999,Gt()),Adamax:()=>Ws.adamax(.002,.9,.999,Gt(),0),RMSProp:()=>Ws.rmsprop(.001,.9,0,Gt()),SGD:()=>Ws.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new V(`Unknown Optimizer ${e}`)}function Lk(e,t,n=!1){if(e==null||typeof e!="object"||Object.getPrototypeOf(e)!==Object.prototype||!Kb(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(n){let a=JSON.stringify(e);a.length>1048576&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${a.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${1048576}.`)}}function Kb(e){if(e===null)return!0;if(typeof e=="object")if(Object.getPrototypeOf(e)===Object.prototype){let t=Object.keys(e);for(let n of t)if(typeof n!="string"||!Kb(e[n]))return!1;return!0}else if(Array.isArray(e)){for(let t of e)if(!Kb(t))return!1;return!0}else return!1;else{let t=typeof e;return t==="string"||t==="number"||t==="boolean"}}function uU(e,t,n,a=console.log){let r=cU(e),s=["Layer (type)","Input Shape","Output shape","Param #"];r?(t=t||90,n=n||[.32,.61,.89,1]):(t=t||115,n=n||[.24,.48,.7,.8,1]),n[n.length-1]<=1&&(n=n.map(p=>Math.floor(t*p)));let i;if(!r){s.push("Receives inputs"),i=[];for(let p in e.nodesByDepth)i.push(...e.nodesByDepth[p])}a("_".repeat(t)),Kh(s,n,a),a("=".repeat(t));let o=e.layers;for(let p=0;p<o.length;++p)r?dU(o[p],n,a):hU(o[p],n,i,a),a((p===o.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();let l=pU(e),u=Bh(e.nonTrainableWeights);a(`Total params: ${l+u}`),a(`Trainable params: ${l}`),a(`Non-trainable params: ${u}`),a("_".repeat(t))}function pU(e){let t;return e.collectedTrainableWeights!=null?t=Bh(e.collectedTrainableWeights):t=Bh(e.trainableWeights),t}function cU(e){let t=!0,n=[],a=[];for(let r in e.nodesByDepth)n.push(e.nodesByDepth[r]);for(let r of n){if(r.length>1||r.length===1&&r[0].inboundLayers.length>1){t=!1;break}a.push(...r)}if(t)for(let r of e.layers){let s=!1;for(let i of r.inboundNodes)if(a.indexOf(i)!==-1)if(s){t=!1;break}else s=!0;if(!t)break}return t}function Kh(e,t,n=console.log){let a="";for(let r=0;r<e.length;++r)r>0&&(a=a.slice(0,a.length-1)+" "),a+=e[r],a=a.slice(0,t[r]),a+=" ".repeat(t[r]-a.length);n(a)}function dU(e,t,n){let a,r;try{r=e.inboundNodes.map(l=>JSON.stringify(l.inputShapes)).join(",")}catch(l){r="multiple"}try{a=JSON.stringify(e.outputShape)}catch(l){a="multiple"}let s=e.name,i=e.getClassName(),o=[`${s} (${i})`,r,a,e.countParams().toString()];Kh(o,t,n)}function hU(e,t,n,a){let r,s;try{s=e.inboundNodes.map(d=>JSON.stringify(d.inputShapes)).join(",")}catch(d){s="multiple"}try{r=JSON.stringify(e.outputShape)}catch(d){r="multiple"}let i=[];for(let d of e.inboundNodes)if(!(n!=null&&n.length>0&&n.indexOf(d)===-1))for(let c=0;c<d.inboundLayers.length;++c){let h=d.inboundLayers[c].name,m=d.nodeIndices[c],f=d.tensorIndices[c];i.push(`${h}[${m}][${f}]`)}let o=e.name,l=e.getClassName(),u=i.length===0?"":i[0],p=[`${o} (${l})`,s,r,e.countParams().toString(),u];Kh(p,t,a);for(let d=1;d<i.length;++d)Kh(["","","","",i[d]],t,a)}function DN(e,t,n){return(e==="inboundNodes"||e==="outputLayers"||e==="inputLayers")&&t===0&&typeof n=="string"}function nc(e,t){if(e===null)return null;if(typeof e=="string")return Us(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],a=e.length;for(let r=0;r<a;++r){let s=e[r];DN(t,r,s)?n.push(s):n.push(nc(s,t))}return n}else{let n={};for(let a of Object.keys(e)){let r=e[a];if(a==="name"&&typeof r=="string")n[a]=r;else{let s=Us(a);n[s]=nc(r,s)}}return n}}function Xb(e,t){if(e==null)return null;if(typeof e=="string")return Sr(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],a=e.length;for(let r=0;r<a;++r){let s=e[r];DN(t,r,s)?n.push(s):n.push(Xb(s,t))}return n}else{let n={};for(let a of Object.keys(e)){let r=e[a],s=Sr(a);(a==="name"||a==="className")&&typeof r=="string"?n[s]=r:n[s]=Xb(r,a)}return n}}var vw="4.0.0",tr=class extends Ge{constructor(e){if(super({}),this.containerNodes=new Set,this.name=e.name,this.name==null){let y=this.getClassName().toLowerCase();this.name=_f(y)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(e.inputs)?this.inputs=e.inputs.slice():this.inputs=[e.inputs],Array.isArray(e.outputs)?this.outputs=e.outputs.slice():this.outputs=[e.outputs],as(this.inputs).length!==this.inputs.length)throw new V(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(y=>y.name)}`);as(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(y=>y.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let y of this.outputs){let b=y.sourceLayer,x=y.nodeIndex,w=y.tensorIndex;this.outputLayers.push(b),this.outputLayersNodeIndices.push(x),this.outputLayersTensorIndices.push(w)}for(let y of this.inputs){let b=y.sourceLayer,x=y.nodeIndex,w=y.tensorIndex;ar(x===0,"input layer has >1 nodes"),ar(w===0,"input layer has >1 tensors"),this.inputLayers.push(b),this.inputLayersNodeIndices.push(x),this.inputLayersTensorIndices.push(w)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let y=0;y<this.inputLayers.length;y++){let b=this.inputLayers[y];if(!(b instanceof Hu))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${y} (0-based) originates from layer type ${b.getClassName()}.`);this.inputNames.push(b.name),this.feedInputShapes.push(b.batchInputShape),this.feedInputNames.push(b.name)}for(let y of this.outputLayers)this.outputNames.push(y.name);this.internalInputShapes=this.inputs.map(y=>y.shape),this.internalOutputShapes=this.outputs.map(y=>y.shape);let t={},n={},a={},r={},s={},i=[],o=(y,b,x,w,I,T)=>{(w==null||I==null||T==null)&&(w=y.sourceLayer,I=y.nodeIndex,T=y.tensorIndex);let C=w.inboundNodes[I];if(x.indexOf(C)!==-1)throw new Wa(`The tensor ${y.name} at layer "${w.name}" is part of a cycle.`);if(b.indexOf(C)!==-1)return;this.containerNodes.add(tr.nodeKey(w,I)),w.id in s||(s[w.id]=Object.keys(s).length),x.indexOf(C)===-1&&x.push(C);let E=C.inboundLayers.length;for(let A=0;A<E;A++){let R=C.inputTensors[A],F=C.inboundLayers[A],S=C.nodeIndices[A],M=C.tensorIndices[A];o(R,b,x,F,S,M)}for(b.push(C);x.indexOf(C)>=0;)x.splice(x.indexOf(C),1);i.push(C)},l=[],u=[];for(let y of this.outputs)o(y,l,u);let p=i.slice().reverse();for(let y of p){n[y.id]=y,y.id in t||(t[y.id]=0);let b=t[y.id],x=a[y.outboundLayer.id]==null?0:a[y.outboundLayer.id];b=Math.max(b,x),a[y.outboundLayer.id]=b,r[y.outboundLayer.id]=y.outboundLayer,t[y.id]=b;for(let w=0;w<y.inboundLayers.length;w++){let I=y.inboundLayers[w],T=y.nodeIndices[w],C=I.inboundNodes[T],E=t[C.id]==null?0:t[C.id];t[C.id]=Math.max(b+1,E),n[C.id]=C}}let d={};for(let y in t){let b=t[y];b in d||(d[b]=[]),d[b].push(n[y])}let c={};for(let y in a){let b=a[y];b in c||(c[b]=[]),c[b].push(r[y])}let h=Object.keys(c).map(y=>parseInt(y,10)).sort(oh);this.layers=[];for(let y of h){let b=c[y];b.sort((x,w)=>{let I=s[x.id],T=s[w.id];return I<T?-1:I>T?1:0});for(let x of b)x instanceof tr&&this.internalContainerRefs.push(x),this.layers.push(x)}this.layersByDepth=c,h=Object.keys(d).map(y=>parseInt(y,10)).sort(oh);let m=this.inputs.slice(),f=[];for(let y of h)for(let b of d[y]){let x=b.outboundLayer;if(x!=null){for(let w of b.inputTensors)if(m.indexOf(w)===-1)throw new Wa(`Graph disconnected: cannot obtain value for tensor ${w} at layer "${x.name}". The following previous layers were accessed without issue: ${f}`);for(let w of b.outputTensors)m.push(w);f.push(x.name)}}this.nodesByDepth=d;let g=this.layers.map(y=>y.name);for(let y of g){let b=g.filter(x=>x===y).length;if(b!==1)throw new Wa(`The name "${y}" is used ${b} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(g))}this.outboundNodes=[],this.inboundNodes=[],new Of({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(y=>null),outputMasks:this.outputs.map(y=>null),inputShapes:this.inputs.map(y=>y.shape),outputShapes:this.outputs.map(y=>y.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let e={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(let t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(n=>n.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new V("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(let t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.layers)t.push(...n.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){let n={},a=0;for(let s of this.layers)for(let i of s.weights){if(n[i.originalName]!=null)throw new V(`Duplicate weight name: ${i.originalName}`);n[i.originalName]=i,a++}let r=[];for(let s in e){let i=s;if(n[s]==null){let o=s.split("/");i=o.slice(0,-2).concat([o[o.length-1]]).join("/")}if(n[i]!=null)r.push([n[i],e[s]]);else if(t)throw new V(`Provided weight data has no target variable: ${s}`);delete n[i]}if(t){let s=[];for(let i in n)s.push(i);if(s.length>0)throw new V(`${s.length} of ${a} weights are not set: ${s}`)}pw(r)}updatedConfig(){let e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion=`tfjs-layers ${vw}`,t.backend="TensorFlow.js",t}toJSON(e,t=!0){let n=Xb(this.updatedConfig());return t?JSON.stringify(n):n}call(e,t){return P(()=>{e=xt(e);let n=new js;for(let a=0;a<this.inputs.length;++a)n.add(this.inputs[a],e[a]);return Mp(this.outputs,n,t)})}computeMask(e,t){return P(()=>{e=xt(e);let n;return t==null?n=li(null,e.length):n=xt(t),this.runInternalGraph(e,n)[1]})}computeOutputShape(e){let t=Wh(e);if(t.length!==this.inputLayers.length)throw new V(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let i=0;i<t.length;i++){let o=this.inputLayers[i],l=t[i],u=o.name+"_0_0";n[u]=l}let a=Object.keys(this.nodesByDepth).map(i=>parseInt(i,10)).sort(oh);if(a.length>1)for(let i of a){let o=this.nodesByDepth[i];for(let l of o){let u=l.outboundLayer;if(this.inputLayers.map(m=>m.id).indexOf(u.id)!==-1)continue;let p=[];for(let m=0;m<l.inboundLayers.length;m++){let f=l.inboundLayers[m],g=l.nodeIndices[m],y=l.tensorIndices[m],b=`${f.name}_${g}_${y}`,x=n[b];p.push(x)}let d=u.computeOutputShape(On(p)),c=Wh(d),h=u.inboundNodes.indexOf(l);for(let m=0;m<c.length;m++){let f=`${u.name}_${h}_${m}`;n[f]=c[m]}}}let r=[],s=[];for(let i=0;i<this.outputLayers.length;i++){let o=this.outputLayers[i],l=this.outputLayersNodeIndices[i],u=this.outputLayersTensorIndices[i],p=`${o.name}_${l}_${u}`;s.push(p)}for(let i=0;i<s.length;i++){let o=s[i];ar(o in n),r.push(n[o])}return On(r)}runInternalGraph(e,t){t==null&&(t=li(null,e.length));let n={};for(let o=0;o<this.inputs.length;++o){let l=this.inputs[o],u=e[o],p=t[o];n[l.id]=[u,p]}let a=Object.keys(this.nodesByDepth).map(o=>parseInt(o,10)).sort(oh);for(let o of a){let l=this.nodesByDepth[o];for(let u of l){let p=u.outboundLayer,d=u.inputTensors,c=u.outputTensors,h=new Array;for(let m of d)m.id in n&&h.push(n[m.id]);if(h.length===d.length){let m={},f,g,y,b;if(u.callArgs!=null&&(m=u.callArgs),h.length===1){let[x,w]=h[0];m.mask==null&&(m.mask=w),y=xt(p.call(x,m)),b=xt(p.computeMask(x,w)),f=[x],g=[w]}else f=h.map(x=>x[0]),g=h.map(x=>x[1]),m.mask==null&&(m.mask=g),y=xt(p.call(f,m)),b=xt(p.computeMask(f,g));if(p.activityRegularizer)throw new Me("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let x=0;x<c.length;++x){let w=c[x],I=y[x],T=b[x];n[w.id]=[I,T]}}}}let r=[],s=[],i=[];for(let o of this.outputs){ar(o.id in n,`Could not compute output ${o.name} : ${o.id}`);let[l,u]=n[o.id];i.push(l.shape),r.push(l),s.push(u)}return[r,s,i]}buildNodeConversionMap(e){let t={},n;for(let a of this.layers){n=a instanceof tr?1:0;for(let r=0;r<a.inboundNodes.length;r++){let s=tr.nodeKey(a,r);this.containerNodes.has(s)&&(t[s]=n,n+=1)}}return t}getLayer(e,t){if(t!=null){if(this.layers.length<=t)throw new V(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);return this.layers[t]}else if(e==null)throw new V("Provide either a layer name or layer index");for(let n of this.layers)if(n.name===e)return n;throw new V(`No such layer: ${e}`)}calculateLosses(){return P(()=>{let e=[];for(let t of this.layers)for(let n=0;n<t.inboundNodes.length;++n){let a=tr.nodeKey(t,n);this.containerNodes.has(a)&&e.push(...t.calculateLosses())}return e})}getConfig(){let e={name:this.name},t=this.buildNodeConversionMap(this.layers),n=[];for(let s of this.layers){let i=s.getClassName(),o=s.getConfig(),l=[];for(let p=0;p<s.inboundNodes.length;p++){let d=s.inboundNodes[p],c=tr.nodeKey(s,p),h={};if(this.containerNodes.has(c)){if(d.callArgs)try{JSON.stringify(d.callArgs),h=d.callArgs}catch(m){console.warn(`Layer ${s.name} was passed non-serializable keyword arguments: ${d.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),h={}}if(d.inboundLayers.length>0){let m=[];for(let f=0;f<d.inboundLayers.length;f++){let g=d.inboundLayers[f],y=d.nodeIndices[f],b=d.tensorIndices[f],x=tr.nodeKey(g,y),w=t[x];w==null&&(w=0),m.push([g.name,w,b,h])}l.push(m)}}}let u={};u.name=s.name,u.className=i,u.config=o,u.inboundNodes=l,n.push(u)}e.layers=n;let a=[];for(let s=0;s<this.inputLayers.length;s++){let i=this.inputLayers[s],o=this.inputLayersNodeIndices[s],l=tr.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let p=this.inputLayersTensorIndices[s];a.push([i.name,u,p])}e.inputLayers=a;let r=[];for(let s=0;s<this.outputLayers.length;s++){let i=this.outputLayers[s],o=this.outputLayersNodeIndices[s],l=tr.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let p=this.outputLayersTensorIndices[s];r.push([i.name,u,p])}return e.outputLayers=r,e}static fromConfig(e,t,n={},a=!1){let r={},s={};function i(f,g){f.name in s?s[f.name].push(g):s[f.name]=[g]}function o(f,g){let y=[],b;for(let x of g){let w=x[0],I=x[1],T=x[2];if(b=x[3]==null?{}:x[3],!(w in r)){i(f,g);return}let C=r[w];if(C.inboundNodes.length<=I){i(f,g);return}let E=C.inboundNodes[I];y.push(E.outputTensors[T])}y.length>0&&f.apply(On(y),b)}function l(f){let g=f.name,y=Ua(f,t.customObjects!=null?t.customObjects:{});y.setFastWeightInitDuringBuild(a),r[g]=y,f.inboundNodes.forEach(b=>{if(!(b instanceof Array))throw new V(`Corrupted configuration, expected array for nodeData: ${b}`);i(y,b)})}let u=t.name,p=t.layers;for(let f of p)l(f);for(;!V4(s);)for(let f of p){let g=r[f.name];if(g.name in s){let y=s[g.name];delete s[g.name];for(let b of y)o(g,b)}}let d=[],c=[],h=t.inputLayers;for(let f of h){let g=f[0],y=f[1],b=f[2];ar(g in r);let x=r[g].inboundNodes[y].outputTensors;d.push(x[b])}let m=t.outputLayers;for(let f of m){let g=f[0],y=f[1],b=f[2];ar(g in r);let x=r[g].inboundNodes[y].outputTensors;c.push(x[b])}return new e({inputs:d,outputs:c,name:u})}get stateful(){if(this._stateful)throw new V("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let e of this.layers)if(e.stateful)return!0;return!1}resetStates(){P(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function mU(e,t,n){let a=t.length;if(e==null||Array.isArray(e)&&e.length===0)return t.map(r=>null);if(a===1)return Array.isArray(e)&&e.length===1?e:typeof e=="object"&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==a)throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${a} outputs. Make sure a set of weights is provided for each model output.`);return e}else if(typeof e=="object"&&Object.keys(e).length>0&&typeof e[Object.keys(e)[0]]=="object"){let r=[];return t.forEach(s=>{s in e?r.push(e[s]):r.push(null)}),r}else throw new Error(`The model has multiple (${a}) outputs, so ${n} must be either an array with ${a} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)}function RN(e,t){return mU(e,t,"classWeight")}async function MN(e,t,n,a){if(t!=null||a!=null)throw new Error("Support sampleWeight is not implemented yet");if(n!=null){let r=P(()=>{if(e.shape.length===1)return ir(e);if(e.shape.length===2){if(e.shape[1]>1)return ri(e,1);if(e.shape[1]===1)return W(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),s=Array.from(await r.data());_e(r);let i=[];return s.forEach(o=>{if(n[o]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${o} exists in the data but not in classWeight`);i.push(n[o])}),Ke(i,"float32")}else return null}function fU(e,t){return z(e,t)}var gU=32;function PN(e,t){let n,a,r=t;n=r.xs,a=r.ys,v.assert(n!=null&&a!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let s=zk("input",e.inputNames,n),i=zk("output",e.outputNames,a),o=s[0].shape[0];v.assert(s.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`),v.assert(i.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let l=0;l<s.length;l++)v.assert(s[l].shape[0]===o,()=>`Batch size mismatch: input ${e.inputNames[l]} has ${s[l].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);for(let l=0;l<i.length;l++)v.assert(i[l].shape[0]===o,()=>`Batch size mismatch: output ${e.outputNames[l]} has ${i[l].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);return{xs:s,ys:i}}function zk(e,t,n){if(n instanceof Te)return[n];if(Array.isArray(n))return v.assert(n.length===t.length,()=>`Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),n;{let a=[];for(let r of t){if(n[r]==null)throw new V(`The feature data generated by the dataset lacks the required ${e} key '${r}'.`);a.push(n[r])}return a}}function yU(e){if(e.length===3)throw new Me("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}async function bU(e,t,n){let a=n.batchesPerEpoch!=null;if(v.assert(e.optimizer!=null,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),v.assert(n!=null,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),v.assert(n.epochs!=null&&n.epochs>0&&Number.isInteger(n.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${n.epochs}`),v.assert(!a||n.batchesPerEpoch>0&&Number.isInteger(n.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${n.batchesPerEpoch}`),v.assert(n.validationSplit==null,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{let r=n.validationData!=null,s,i;if(r)if(Wk(n.validationData))v.assert(n.validationBatches==null||n.validationBatches>0&&Number.isInteger(n.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${n.validationBatches}`);else{let g=yU(n.validationData);s=g.xs,i=g.ys}let o=e.makeTrainFunction(),l=e.getDedupedMetricsNames(),u;r?u=l.slice().concat(l.map(g=>"val_"+g)):u=l.slice();let p=NN(n.callbacks,n.yieldEvery),d=n.verbose==null?1:n.verbose,{callbackList:c,history:h}=CN(p,d,n.epochs,null,null,xU(t,n),null,r,u);c.setModel(e),e.history=h,await c.onTrainBegin(),e.stopTraining_=!1;let m=n.initialEpoch==null?0:n.initialEpoch,f=await t.iterator();for(;m<n.epochs;){let g={};await c.onEpochBegin(m);let y=0,b=0;for(a||(f=await t.iterator());!a||y<n.batchesPerEpoch;){let x=await f.next();if(a&&x.done){console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${y} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(x.value!=null){let{xs:w,ys:I}=PN(e,x.value),T={};T.batch=b,T.size=w[0].shape[0],await c.onBatchBegin(b,T);let C=[];if(n.classWeight!=null){let R=RN(n.classWeight,e.outputNames);for(let F=0;F<R.length;++F)C.push(await MN(I[F],null,R[F]))}let E=w.concat(I).concat(C),A=o(E);_e(E);for(let R=0;R<l.length;++R){let F=l[R],S=A[R];T[F]=S,Jt(S)}await c.onBatchEnd(b,T),kN(T),b++,y++}if(a?y>=n.batchesPerEpoch:x.done){if(r){let w;Wk(n.validationData)?w=xt(await e.evaluateDataset(n.validationData,{batches:n.validationBatches})):w=xt(e.evaluate(s,i,{batchSize:n.validationBatchSize==null?gU:n.validationBatchSize,verbose:0}));for(let I=0;I<e.metricsNames.length;++I)g[`val_${e.metricsNames[I]}`]=w[I]}break}if(e.stopTraining_)break}if(await c.onEpochEnd(m,g),m++,e.stopTraining_)break}return await c.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function xU(e,t){let n=null;return t.batchesPerEpoch!=null?n=t.batchesPerEpoch:Number.isFinite(e.size)&&(n=e.size),n}function Wk(e){return typeof e.iterator=="function"}function vU(e){return typeof e.next=="function"}async function wU(e,t,n){n=n||{};let a=n.batches!=null,r=e.testFunction,s=[];if(n.verbose>0)throw new Me("Verbose mode is not implemented yet.");v.assert(!a||n.batches>0&&Number.isInteger(n.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(n.batches)}`);let i=vU(t)?t:await t.iterator(),o=0,l=0;for(;!a||l<n.batches;){let u=await i.next();if(s=P(()=>{if(u.value){let{xs:p,ys:d}=PN(e,u.value),c=p.concat(d),h=P(()=>r(c));if(_e(c),l===0)for(let f=0;f<h.length;++f)s.push(be(0));let m=c[0].shape[0];for(let f=0;f<h.length;++f){let g=h[f],y=s[f];s[f]=P(()=>Y(s[f],z(m,g))),l>0&&_e(y)}_e(h),o+=m,++l}return s}),u.done){a&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let u=0;u<s.length;++u){let p=s[u];s[u]=he(s[u],o),_e(p)}return On(s)}function Yb(e){v.assert(e>0&&Number.isInteger(e),()=>`batchSize is required to be a positive integer, but got ${e}`)}function Pp(e,t,n){return e==null?[null]:Array.isArray(e)?e.map(a=>Zs(a,t,n-t)):Zs(e,t,n-t)}function ww(e,t){return P(()=>e==null?null:Array.isArray(e)?e.map(n=>ww(n,t)):mN(e,t.dtype==="int32"?t:oe(t,"int32")))}function Zb(e,t){let n=[],a=0,r=null;for(;a<e;)r=a+t,r>=e&&(r=e),n.push([a,r]),a=r;return n}async function kU(e,t,n,a,r,s,i,o,l,u,p,d,c,h,m){r==null&&(r=32),s==null&&(s=1),p==null&&(p=!0),c==null&&(c=0);let f=!1;if(l!=null&&u!=null&&(f=!0),m!=null&&(f=!0,h==null))throw new V("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let g=e.checkNumSamples(n,r,h,"steps_per_epoch"),y;g!=null&&(y=Ha(0,g)),i==null&&(i=1);let{callbackList:b,history:x}=CN(o,i,s,c,g,h,r,f,d);b.setModel(e),e.history=x,await b.onTrainBegin(),e.stopTraining_=!1;for(let w=c;w<s;++w){await b.onEpochBegin(w);let I={};if(h!=null)throw new Me("stepsPerEpoch mode is not implemented yet.");{if(p==="batch")throw new Me("batch shuffling is not implemneted yet");p&&v.shuffle(y);let T=Ke(y),C=Zb(g,r);for(let E=0;E<C.length;++E){let A={};if(await b.onBatchBegin(E,A),P(()=>{let R=C[E][0],F=C[E][1],S=Zs(T,R,F-R);A.batch=E,A.size=F-R;let M=ww(n,S),B=t(M);for(let U=0;U<a.length;++U){let G=a[U],q=B[U];A[G]=q,Jt(q)}if(E===C.length-1&&f){let U=e.testLoop(l,u,r);for(let G=0;G<a.length;++G){let q=a[G],K=U[G];Jt(K),I["val_"+q]=K}}}),await b.onBatchEnd(E,A),kN(A),e.stopTraining_)break}T.dispose()}if(await b.onEpochEnd(w,I),e.stopTraining_)break}return await b.onTrainEnd(),await e.history.syncData(),e.history}async function IU(e,t,n,a={}){if(e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;let r,s,i,o,l,u,p,d,c;try{let h=a.batchSize==null?32:a.batchSize;Yb(h);let m=!1,f=await e.standardizeUserData(t,n,a.sampleWeight,a.classWeight,m,h);r=f[0],s=f[1],c=f[2];let g=!1,y;if(a.validationData!=null&&a.validationData.length>0){if(g=!0,a.validationData.length===2)l=a.validationData[0],u=a.validationData[1];else throw a.validationData.length===3?new Me("validationData including sample weights is not supported yet."):new V(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${a.validationData} is invalid.`);let E=!0,A=await e.standardizeUserData(l,u,null,null,E,h);p=A[0],d=A[1],y=p.concat(d)}else if(a.validationSplit!=null&&a.validationSplit>0&&a.validationSplit<1){g=!0;let E=Math.floor(r[0].shape[0]*(1-a.validationSplit)),A=r[0].shape[0];p=Pp(r,E,A),i=r,r=Pp(r,0,E),d=Pp(s,E,A),o=s,s=Pp(s,0,E),y=p.concat(d)}else a.validationSteps!=null&&(g=!0);let b=r.concat(s).concat(c);e.checkTrainableWeightsConsistency();let x=e.makeTrainFunction(),w=e.getDedupedMetricsNames(),I,T;g?(e.makeTestFunction(),I=e.testFunction,T=w.slice().concat(w.map(E=>"val_"+E))):(I=null,y=[],T=w.slice());let C=NN(a.callbacks,a.yieldEvery);return await kU(e,x,b,w,h,a.epochs,a.verbose,C,I,y,a.shuffle,T,a.initialEpoch,null,null)}finally{e.isTraining=!1,La(r,t),La(s,n),La(i,t),La(o,n),La(p,l),La(d,u),c!=null&&_e(c)}}function ON(e){let t=[];e instanceof Te&&(e=[e]);for(let n=0;n<e.length;++n){let a=e[n];if(a.rank===1)t.push(Xc(a,1));else{if(a.rank===0)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(a)}}return t}function La(e,t){if(e==null)return;let n=[];if(t instanceof Te)n.push(t.id);else if(Array.isArray(t))t.forEach(r=>n.push(r.id));else if(t!=null)for(let r in t){let s=t[r];n.push(s.id)}let a=[];if(e instanceof Te)n.indexOf(e.id)===-1&&a.push(e);else if(Array.isArray(e))e.forEach(r=>{n.indexOf(r.id)===-1&&a.push(r)});else if(e!=null)for(let r in e){let s=e[r];n.indexOf(s.id)===-1&&a.push(s)}a.forEach(r=>{r.isDisposed||r.dispose()})}function SU(e){return e instanceof Te}function Jb(e){return Array.isArray(e)}function Bk(e){return!SU(e)&&!Jb(e)}function Vk(e,t,n,a=!0,r=""){if(t==null||t.length===0){if(e!=null){let i=!1;if(Jb(e)&&e.length>0)i=!0;else if(Bk(e)){for(let o in e)if(e.hasOwnProperty(o)){i=!0;break}}else i=!0;if(i)throw new V(`Error when checking model ${r} expected no data, but got ${e}`)}return[]}if(e==null)return t.map(i=>null);let s;if(Bk(e)){e=e,s=[];for(let i of t){if(e[i]==null)throw new V(`No data provided for "${i}". Need data for each key in: ${t}`);s.push(e[i])}}else if(Jb(e)){if(e=e,e.length!==t.length)throw new V(`Error when checking model ${r}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);s=e}else{if(e=e,t.length>1)throw new V(`The model ${r} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);s=[e]}if(s=ON(s),n!=null)for(let i=0;i<t.length;++i){if(n[i]==null)continue;let o=s[i];if(o.shape.length!==n[i].length)throw new V(`Error when checking ${r}: expected ${t[i]} to have ${n[i].length} dimension(s). but got array with shape ${o.shape}`);for(let l=0;l<n[i].length;++l){if(l===0&&!a)continue;let u=o.shape[l],p=n[i][l];if(p!=null&&p>=0&&u!==p)throw new V(`${r} expected a batch of elements where each example has shape [${n[i].slice(1,n[i].length)}] (i.e.,tensor shape [*,${n[i].slice(1,n[i].length)}]) but the ${r} received an input with ${o.shape[0]} examples, each with shape [${o.shape.slice(1,o.shape.length)}] (tensor shape [${o.shape}])`)}}return s}function TU(e,t,n){let a=as(e.map(s=>s.shape[0]));a.sort();let r=as(t.map(s=>s.shape[0]));if(r.sort(),a.length>1)throw new V(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(e.map(s=>s.shape))}`);if(r.length>1)throw new V(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(s=>s.shape))}`);if(a.length>0&&r.length>0&&!v.arraysEqual(a,r))throw new V(`Input Tensors should have the same number of samples as target Tensors. Found ${a[0]} input sample(s) and ${r[0]} target sample(s).`)}function NU(e,t,n){let a=[xo,zf,tc];for(let r=0;r<e.length;++r){let s=e[r],i=t[r],o=n[r];if(i!=null){if(i===tc&&s.shape[s.shape.length-1]===1)throw new V(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(a.indexOf(i)!==-1){let l=s.shape.slice(1),u=o.slice(1);for(let p=0;p<l.length;++p){let d=l[p],c=u[p];if(c!=null&&d!==c)throw new V(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function Uk(e,t,n,a=!0,r=""){let s;if(Array.isArray(e)){if(e.length!==t.length)throw new V(`Error when checking model ${r}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);s=e}else{if(t.length>1)throw new V(`The model expects ${t.length} ${r} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);s=[e]}if(n!=null)for(let i=0;i<t.length;++i){if(n[i]==null)continue;let o=s[i];if(o.shape.length!==n[i].length)throw new V(`Error when checking ${r}: expected ${t[i]} to have ${n[i].length} dimension(s), but got array with shape ${JSON.stringify(o.shape)}`);for(let l=0;l<n[i].length;++l){if(l===0&&!a)continue;let u=o.shape[l],p=n[i][l];if(p!=null&&p!==u)throw new V(`Error when checking ${r}: expected ${t[i]} to have shape ${JSON.stringify(n[i])} but got array with shape ${JSON.stringify(o.shape)}.`)}}}function CU(e,t){if(e==null||Array.isArray(e)&&e.length===0)return t.map(a=>[]);let n;if(typeof e=="string"||typeof e=="function")n=[e];else if(Array.isArray(e)||typeof e=="object")n=e;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${e}`);if(Array.isArray(n))return t.map(a=>n);{let a=[];for(let r of t){let s=n.hasOwnProperty(r)?n[r]:[];Array.isArray(s)||(s=[s]),a.push(s)}return a}}var _U="layers-model",Er=class extends tr{constructor(e){super(e),this.isTraining=!1}summary(e,t,n=console.log){if(!this.built)throw new V("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");uU(this,e,t,n)}compile(e){if(e.loss==null&&(e.loss=[]),this.loss=e.loss,typeof e.optimizer=="string")this.optimizer_=lU(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof Rr))throw new V("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(!Array.isArray(e.loss)&&typeof e.loss!="string"&&typeof e.loss!="function"){e.loss=e.loss;for(let s in e.loss)if(this.outputNames.indexOf(s)===-1)throw new V(`Unknown entry in loss dictionary: "${s}". Only expected the following keys: ${this.outputNames}`);for(let s of this.outputNames)e.loss[s]==null&&console.warn(`Output "${s}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${s} during training`),t.push(vb(e.loss[s]))}else if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new V(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);t=e.loss.map(s=>vb(s))}else{let s=vb(e.loss);this.outputs.forEach(i=>{t.push(s)})}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let s=0;s<this.outputs.length;++s){let i=this.internalOutputShapes[s],o=this.outputNames[s];this.feedOutputNames.push(o),this.feedOutputShapes.push(i),this.feedLossFns.push(this.lossFunctions[s])}let n=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Ys("loss",()=>{for(let s=0;s<this.outputs.length;++s){if(n.indexOf(s)!==-1)continue;let i=this.lossFunctions[s];this.outputs.length>1&&(this.metricsTensors.push([i,s]),this.metricsNames.push(this.outputNames[s]+"_loss"))}});let a=CU(e.metrics,this.outputNames),r=(s,i,o)=>{this.outputNames.length>1&&(i=this.outputNames[s]+"_"+i),this.metricsNames.push(i),this.metricsTensors.push([o,s])};Ys("metric",()=>{for(let s=0;s<this.outputs.length;++s){if(n.indexOf(s)!==-1)continue;let i=a[s];(o=>{let l="",u,p,d;for(let c of o){if(typeof c=="string"&&["accuracy","acc","crossentropy","ce"].indexOf(c)!==-1){let m=this.internalOutputShapes[s];m[m.length-1]===1||this.lossFunctions[s]===zf?["accuracy","acc"].indexOf(c)!==-1?p=yw:["crossentropy","ce"].indexOf(c)!==-1&&(p=AN):this.lossFunctions[s]===Hh?["accuracy","acc"].indexOf(c)!==-1?p=$N:["crossentropy","ce"].indexOf(c)!==-1&&(p=FN):["accuracy","acc"].indexOf(c)!==-1?p=bw:["crossentropy","ce"].indexOf(c)!==-1&&(p=xw);let f;["accuracy","acc"].indexOf(c)!==-1?f="acc":["crossentropy","ce"].indexOf(c)!==-1&&(f="ce"),d=p,u=l+f}else d=oU(c),u=l+ph(c);let h;Ys(u,()=>{h=d}),r(s,u,h)}})(i)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,n={}){let a=n.batchSize==null?32:n.batchSize;Yb(a);let r=!0,s=this.standardizeUserDataXY(e,t,r,a);try{let i=s[0].concat(s[1]);this.makeTestFunction();let o=this.testFunction,l=this.testLoop(o,i,a,n.verbose,n.steps);return On(l)}finally{La(s[0],e),La(s[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),wU(this,e,t)}checkNumSamples(e,t,n,a="steps"){let r;if(n!=null){if(r=null,t!=null)throw new V(`If ${a} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else if(e!=null)Array.isArray(e)?r=e[0].shape[0]:r=e.shape[0];else throw new V(`Either the input data should have a defined shape, or ${a} shoud be specified.`);return r}execute(e,t){if(Array.isArray(t)&&t.length===0)throw new V("`outputs` is an empty Array, which is not allowed.");let n=Array.isArray(t),a=n?t:[t],r=this.retrieveSymbolicTensors(a),s=new js;if(e instanceof Te&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new V(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let o=0;o<this.inputs.length;++o)s.add(this.inputs[o],e[o])}else for(let o of this.inputs){let l=e[o.name];if(l==null)throw new V(`No value is provided for the model's input ${o.name}`);s.add(o,l)}let i=Mp(r,s);return n?i:i[0]}retrieveSymbolicTensors(e){let t=li(null,e.length),n=e.length;for(let a of this.layers){let r=Array.isArray(a.output)?a.output:[a.output],s=r.map(i=>i.name);for(let i=0;i<e.length;++i){let o=s.indexOf(e[i]);if(o!==-1&&(t[i]=r[o],n--),n===0)break}if(n===0)break}if(n>0){let a=[];throw t.forEach((r,s)=>{r==null&&a.push(e[s])}),new V(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(a)}`)}return t}predictLoop(e,t=32,n=!1){return P(()=>{let a=this.checkNumSamples(e);if(n)throw new Me("Verbose predictLoop() is not implemented yet.");let r=Zb(a,t),s=this.outputs.map(i=>[]);for(let i=0;i<r.length;++i)P(()=>{let o=r[i][0],l=r[i][1],u=Pp(e,o,l),p=[];if(Array.isArray(u))for(let c=0;c<u.length;++c)p.push({key:this.inputs[c],value:u[c]});else p.push({key:this.inputs[0],value:u});let d=new js(p);return Mp(this.outputs,d)}).forEach((o,l)=>s[l].push(o));return On(s.map(i=>Ze(i,0)))})}predict(e,t={}){let n=ON(e);Uk(n,this.inputNames,this.feedInputShapes,!1);try{let a=t.batchSize==null?32:t.batchSize;return Yb(a),this.predictLoop(n,a)}finally{La(n,e)}}predictOnBatch(e){Uk(e,this.inputNames,this.feedInputShapes,!0);let t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,n=!0,a){if(this.optimizer_==null)throw new Wa("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let r=[];for(let s=0;s<this.feedOutputShapes.length;++s){let i=this.feedOutputShapes[s];this.feedLossFns[s]===Hh?r.push(i.slice(0,i.length-1).concat([1])):r.push(i)}if(e=Vk(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=Vk(t,this.feedOutputNames,r,!1,"target"),TU(e,t,null),NU(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&a!=null&&a>0&&e[0].shape[0]%a!==0)throw new V(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${a}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,n,a,r=!0,s){let[i,o]=this.standardizeUserDataXY(e,t,r,s);if(n!=null)throw new Error("sample weight is not supported yet.");let l=null;if(a!=null){let u=RN(a,this.outputNames);l=[];for(let p=0;p<u.length;++p)l.push(await MN(o[p],null,u[p]))}return[i,o,l]}testLoop(e,t,n,a=0,r){return P(()=>{let s=this.checkNumSamples(t,n,r,"steps"),i=[];if(a>0)throw new Me("Verbose mode is not implemented yet.");if(r!=null)throw new Me("steps mode in testLoop() is not implemented yet");{let o=Zb(s,n),l=Ke(Ha(0,s));for(let u=0;u<o.length;++u){let p=o[u][0],d=o[u][1],c=Zs(l,p,d-p),h=ww(t,c),m=e(h);if(u===0)for(let f=0;f<m.length;++f)i.push(be(0));for(let f=0;f<m.length;++f){let g=m[f];i[f]=Y(i[f],z(d-p,g))}}for(let u=0;u<i.length;++u)i[u]=he(i[u],s)}return i})}getDedupedMetricsNames(){let e=this.metricsNames,t=[];for(let n=0;n<e.length;++n){let a=e[n],r=a;_k(e,a)>1&&(r+=`_${_k(e.slice(0,n),a)}`),t.push(r)}return t}makeTrainFunction(){return e=>{let t=[],n=e.slice(0,this.inputs.length),a=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),r=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),s=[],i=()=>{let u=[];for(let h=0;h<this.inputs.length;++h)u.push({key:this.inputs[h],value:n[h]});let p=new js(u),d=Mp(this.outputs,p,{training:!0}),c;for(let h=0;h<this.lossFunctions.length;++h){let m=this.lossFunctions[h],f=m(a[h],d[h]);r[h]!=null&&(f=fU(f,r[h]));let g=Nt(f);t.push(g),h===0?c=f:c=Y(c,f)}for(let h=0;h<this.metricsTensors.length;++h){let m;if(this.outputs.length>1&&h<this.outputs.length)m=t[h];else{let f=this.metricsTensors[h][0],g=this.metricsTensors[h][1];m=Nt(f(a[g],d[g]))}Jt(m),s.push(m)}return c=Nt(c),this.calculateLosses().forEach(h=>{c=Y(c,h)}),c},o=this.collectedTrainableWeights.map(u=>u.read()),l=!0;return[this.optimizer_.minimize(i,l,o)].concat(s)}}makeTestFunction(){this.testFunction=e=>P(()=>{let t=[],n,a=e.slice(0,this.inputs.length),r=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=[];for(let l=0;l<this.inputs.length;++l)s.push({key:this.inputs[l],value:a[l]});let i=new js(s),o=Mp(this.outputs,i);for(let l=0;l<this.lossFunctions.length;++l){let u=this.lossFunctions[l],p=Nt(u(r[l],o[l]));l===0?n=p:n=Y(n,p),t.push(n)}for(let l=0;l<this.metricsTensors.length;++l){let u=this.metricsTensors[l][0],p=this.metricsTensors[l][1],d=Nt(u(r[p],o[p]));t.push(d)}return t})}async fit(e,t,n={}){return IU(this,e,t,n)}async fitDataset(e,t){return bU(this,e,t)}async trainOnBatch(e,t){let n=await this.standardizeUserData(e,t),a=n[0],r=n[1],s=this.makeTrainFunction()(a.concat(r)),i=[];for(let o of s){let l=await o.data();i.push(l[0])}return _e(s),La(n[0],e),La(n[1],t),On(i)}getNamedWeights(e){let t=[],n=e!=null&&e.trainableOnly,a=n?this.trainableWeights:this.weights,r=this.getWeights(n);for(let s=0;s<a.length;++s)n&&!a[s].trainable||t.push({name:a[s].originalName,tensor:r[s]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){let e=super.dispose();if(e.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let t=Oh().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-Oh().numTensors}return e}getLossIdentifiers(){let e;if(typeof this.loss=="string")e=Sr(this.loss);else if(Array.isArray(this.loss)){for(let t of this.loss)if(typeof t!="string")throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(t=>Sr(t))}else{let t=Object.keys(this.loss);e={};let n=this.loss;for(let a of t)if(typeof n[a]=="string")e[a]=Sr(n[a]);else throw new Error("Serialization of non-string loss is not supported.")}return e}getMetricIdentifiers(){if(typeof this.metrics=="string"||typeof this.metrics=="function")return[Sr(ph(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>Sr(ph(e)));{let e={};for(let t in this.metrics)e[t]=Sr(ph(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(e.weighted_metrics!=null)throw new Error("Loading weight_metrics is not supported yet.");if(e.loss_weights!=null)throw new Error("Loading loss_weights is not supported yet.");if(e.sample_weight_mode!=null)throw new Error("Loading sample_weight_mode is not supported yet.");let t=nc(e.optimizer_config),n=Ua(t),a;if(typeof e.loss=="string")a=Us(e.loss);else if(Array.isArray(e.loss))a=e.loss.map(s=>Us(s));else if(e.loss!=null){a={};for(let s in e.loss)a[s]=Us(e.loss[s])}let r;if(Array.isArray(e.metrics))r=e.metrics.map(s=>Us(s));else if(e.metrics!=null){r={};for(let s in e.metrics)r[s]=Us(e.metrics[s])}this.compile({loss:a,metrics:r,optimizer:n})}async save(e,t){if(typeof e=="string"){let i=Ut.getSaveHandlers(e);if(i.length===0)throw new V(`Cannot find any save handlers for URL '${e}'`);if(i.length>1)throw new V(`Found more than one (${i.length}) save handlers for URL '${e}'`);e=i[0]}if(e.save==null)throw new V("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let n=await Ut.encodeWeights(this.getNamedWeights(t)),a=!1,r=null,s={modelTopology:this.toJSON(r,a),format:_U,generatedBy:`TensorFlow.js tfjs-layers v${vw}`,convertedBy:null};if((t==null?!1:t.includeOptimizer)&&this.optimizer!=null){s.trainingConfig=this.getTrainingConfig();let i="optimizer",{data:o,specs:l}=await Ut.encodeWeights(await this.optimizer.getWeights(),i);n.specs.push(...l),n.data=Ut.concatenateArrayBuffers([n.data,o])}return this.userDefinedMetadata!=null&&(Lk(this.userDefinedMetadata,this.name,!0),s.userDefinedMetadata=this.userDefinedMetadata),s.weightData=n.data,s.weightSpecs=n.specs,e.save(s)}setUserDefinedMetadata(e){Lk(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Er.className="Model";ne.registerClass(Er);var LN=class extends Er{};LN.className="Functional";ne.registerClass(LN);async function EU(e,t){"modelTopology"in e||(e={modelTopology:e}),e=e;let n=e.modelTopology;n.model_config!=null&&(n=n.model_config);let a=nc(n),r=Ua(a,t);if(e.weightsManifest!=null){let s=await Ut.loadWeights(e.weightsManifest,e.pathPrefix,r.weights.map(o=>o.originalName)),i={};for(let o of r.weights)i[o.originalName]=s[o.originalName];r.loadWeights(i),_e(s)}return r}async function AU(e,t){if(t==null&&(t={}),typeof e=="string"){let n=Ut.getLoadHandlers(e,t);if(n.length===0)n.push(Ut.browserHTTPRequest(e,t));else if(n.length>1)throw new V(`Found more than one (${n.length}) load handlers for URL '${e}'`);e=n[0]}return $U(e,void 0,t)}async function $U(e,t,n){if(n==null&&(n={}),e.load==null)throw new V("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let a=await e.load(),r=a.modelTopology;r.model_config!=null&&(r=r.model_config);let s=n.strict==null?!0:n.strict,i=a.weightData!=null&&a.weightSpecs!=null&&s,o=Ua(nc(r),t,i),l=a.trainingConfig;if(l!=null&&o.loadTrainingConfig(l),a.userDefinedMetadata!=null&&o.setUserDefinedMetadata(a.userDefinedMetadata),a.weightData!=null){if(a.weightSpecs==null)throw new V("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:u,optimizerWeights:p}=FU(a.weightData,a.weightSpecs);o.loadWeights(u,s),o.optimizer!=null&&p.length>0&&await o.optimizer.setWeights(p),_e(u),_e(p.map(d=>d.tensor))}return o}function FU(e,t){let n=Ut.decodeWeights(e,t),a={},r=[];return t.forEach(s=>{s.group==="optimizer"?r.push({name:s.name,tensor:n[s.name]}):a[s.name]=n[s.name]}),{modelWeights:a,optimizerWeights:r}}var Tl=class extends Er{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=e.name!=null?e.name:_f("sequential_"),e.layers!=null)for(let t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(t=>t<0))throw new V(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(e){let t=e instanceof Tl||e instanceof Er,n;if(t){if(n=e,n.outputs.length!==1)throw new V("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(n.inputs.length!==1)throw new V("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(this.outputs.length===0){if(e.inboundNodes.length===0){if(e.batchInputShape==null)throw new V("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let a=bN({batchShape:e.batchInputShape,dtype:e.dtype,name:e.name+"_input"});e.apply(a)}if(t)this.outputs=n.outputs,this.inputs=n.inputs;else{if(e.inboundNodes.length!==1)throw new V(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);if(e.inboundNodes[0].outputTensors.length!==1)throw new V("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[e.inboundNodes[0].outputTensors[0]],this.inputs=yN(this.outputs[0])}this.inboundNodes=[],new Of({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:li(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(a=>a.shape),outputShapes:this.outputs[0].shape})}else{let a=e.apply(this.outputs[0]);if(Array.isArray(a))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[a],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(e),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return this.model==null&&this.build(),this.model.call(e,t)}build(e){if(tt(e),this.inputs.length===0||this.outputs.length===0)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Er({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,n=console.log){this.built||this.build(),super.summary(e,t,n)}setWeights(e){this.model==null&&this.build(),this.model.setWeights(e)}evaluate(e,t,n={}){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.evaluate(e,t,n)}async evaluateDataset(e,t){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return this.model==null&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return this.model==null&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,n={}){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.fit(e,t,n)}async fitDataset(e,t){if(!this.built)throw new Wa("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(e,t,n={},a=!1){let r,s={};if(t instanceof Array){if(t[0].className==null||t[0].className==="Merge")throw new V("Legacy serialization format not supported yet.");r=t}else v.assert(t.layers!=null,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),r=t.layers,delete t.layers,s=t;let i=new e(s);if(!(i instanceof Tl))throw new Me(`Sequential.fromConfig called on non-Sequential input: ${i}`);for(let o of r){let l=Ua(o,void 0,a);a&&l.setFastWeightInitDuringBuild(!0),i.add(l)}return i}set stopTraining(e){if(this.model==null)throw new V("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(this.model==null)throw new V("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let e=[];for(let t of this.layers){let n={};n.className=t.getClassName(),n.config=t.getConfig(),e.push(n)}return{name:this.name,layers:e}}};Tl.className="Sequential";ne.registerClass(Tl);function DU(e){return new Er(e)}function RU(e){return new Tl(e)}function zN(e){return bN(e)}function MU(e,t){Ca.registerCallbackConstructor(e,t)}var Un=class extends ne.Serializable{getConfig(){return{}}},WN=class extends Un{apply(e,t=1){return aV(e,t)}};WN.className="elu";ne.registerClass(WN);var BN=class extends Un{apply(e){return uf(e)}};BN.className="selu";ne.registerClass(BN);var VN=class extends Un{apply(e){return Xe(e)}};VN.className="relu";ne.registerClass(VN);var UN=class extends Un{apply(e){return P(()=>Vu(6,Xe(e)))}};UN.className="relu6";ne.registerClass(UN);var GN=class extends Un{apply(e){return e}};GN.className="linear";ne.registerClass(GN);var HN=class extends Un{apply(e){return ha(e)}};HN.className="sigmoid";ne.registerClass(HN);var jN=class extends Un{apply(e){return sV(e)}};jN.className="hardSigmoid";ne.registerClass(jN);var qN=class extends Un{apply(e){return mo(e)}};qN.className="softplus";ne.registerClass(qN);var KN=class extends Un{apply(e){return rV(e)}};KN.className="softsign";ne.registerClass(KN);var XN=class extends Un{apply(e){return si(e)}};XN.className="tanh";ne.registerClass(XN);var kw=class extends Un{apply(e,t=-1){return Ka(e,t)}};kw.className="softmax";ne.registerClass(kw);var YN=class extends Un{apply(e,t=-1){return ef(e,t)}};YN.className="logSoftmax";ne.registerClass(YN);var ZN=class extends Un{apply(e,t=1){return P(()=>z(ha(z(e,t)),e))}};ZN.className="swish";ne.registerClass(ZN);var JN=class extends Un{apply(e){return P(()=>z(e,si(mo(e))))}};JN.className="mish";ne.registerClass(JN);function us(e){return e.getClassName()}function wb(e,t={}){return Kc(e,ne.SerializationMap.getMap().classNameMap,t,"activation")}function ps(e){if(e==null){let t={};return t.className="linear",t.config={},wb(t)}if(typeof e=="string"){let t={};return t.className=e,t.config={},wb(t)}else return e instanceof Un?e:wb(e)}function Iw(e){if(e!=null&&typeof e!="object")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${e}`)}var QN=class extends ne.Serializable{},Qc=class extends QN{constructor(e){super(),Iw(e),this.l1=e==null||e.l1==null?.01:e.l1,this.l2=e==null||e.l2==null?.01:e.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(e){return P(()=>{let t=It([1]);return this.hasL1&&(t=Y(t,fe(z(this.l1,Lt(e))))),this.hasL2&&(t=Y(t,fe(z(this.l2,Yc(e))))),W(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};Qc.className="L1L2";ne.registerClass(Qc);function PU(e){return Iw(e),new Qc({l1:e!=null?e.l1:null,l2:0})}function OU(e){return Iw(e),new Qc({l2:e!=null?e.l2:null,l1:0})}var Gk={l1l2:"L1L2"};function pt(e){return ew(e)}function Hk(e,t={}){return Kc(e,ne.SerializationMap.getMap().classNameMap,t,"regularizer")}function Tt(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in Gk?Gk[e]:e,config:{}};return Hk(t)}else return e instanceof QN?e:Hk(e)}var Sw=class extends Ge{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null&&(this.maxValue=e.maxValue)}call(e,t){e=Ne(e);let n=Xe(e);return this.maxValue!=null&&(n=en(n,0,this.maxValue)),n}computeOutputShape(e){return e}getConfig(){let e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};Sw.className="ReLU";ne.registerClass(Sw);var Tw=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_ALPHA=.3,e==null&&(e={}),this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ne(e);return Lc(n,this.alpha)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Tw.className="LeakyReLU";ne.registerClass(Tw);var Nw=class extends Ge{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA_INITIALIZER="zeros",e==null&&(e={}),this.supportsMasking=!0,this.alphaInitializer=St(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=Tt(e.alphaRegularizer),this.alphaConstraint=Kt(e.alphaConstraint),e.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else if(typeof e.sharedAxes=="number")this.sharedAxes=[e.sharedAxes];else throw new V(`Expected sharedAxes to be a number or an array of numbers, but got ${e.sharedAxes}`)}build(e){e=tt(e);let t=e.slice(1);if(this.sharedAxes!=null)for(let a of this.sharedAxes)t[a-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(this.sharedAxes!=null)for(let a=1;a<e.length;++a)n[a]=e[a];this.inputSpec=[new zt({ndim:e.length,axes:n})],this.built=!0}call(e,t){return e=Ne(e),Uc(e,this.alpha.read())}getConfig(){let e={alphaInitializer:Ct(this.alphaInitializer),alphaRegularizer:pt(this.alphaRegularizer),alphaConstraint:qt(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};Nw.className="PReLU";ne.registerClass(Nw);var Cw=class extends Ge{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA=1,e==null&&(e={}),e.alpha!=null&&e.alpha!==this.DEFAULT_ALPHA)throw new Me(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ne(e);return Lu(n)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Cw.className="ELU";ne.registerClass(Cw);var _w=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_THETA=1,e==null&&(e={}),this.theta=e.theta==null?this.DEFAULT_THETA:e.theta}call(e,t){let n=Ne(e);return z(n,oe(Cn(n,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){let e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};_w.className="ThresholdedReLU";ne.registerClass(_w);var Ew=class extends Ge{constructor(e){super(e==null?{}:e),this.DEFAULT_AXIS=1,e==null&&(e={}),this.softmax=new kw().apply,this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis}call(e,t){let n=Ne(e);return this.softmax(n,this.axis)}computeOutputShape(e){return e}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};Ew.className="Softmax";ne.registerClass(Ew);function dl(e,t,n){if(typeof e=="number")return li(e,t);if(e.length!==t)throw new V(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let a=0;a<t;++a){let r=e[a];if(!Q4(r))throw new V(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${r}`)}return e}function Ga(e,t,n,a,r=1){if(e==null)return e;let s=t+(t-1)*(r-1),i;return n==="same"?i=e:i=e-s+1,Math.floor((i+a-1)/a)}function rr(e,t,n,a){if(e==null)return null;if(a==="valid")e=e*t+ls([n-t,0]);else if(a==="same")e=e*t;else throw new V(`Unsupport padding mode: ${a}.`);return e}function Aw(e,t){return P(()=>(Rt(t),t==="channelsFirst"?Ee(e,[0,2,3,1]):e))}function e2(e,t){return P(()=>(Rt(t),t==="channelsFirst"?Ee(e,[0,2,3,4,1]):e))}function LU(e,t,n,a=1,r="valid",s,i=1){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.shape.length!==3)throw new V(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(t.shape.length!==3)throw new V(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(n!=null&&n.shape.length!==1)throw new V(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(s==="channelsFirst"&&(e=Ee(e,[0,2,1])),r==="causal")throw new Me("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let o=qm(e,t,a,r==="same"?"same":"valid","NWC",i);return n!=null&&(o=Xa(o,n)),o})}function jk(e,t,n,a=[1,1],r="valid",s,i,o=null){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.rank!==3&&e.rank!==4)throw new V(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(t.rank!==3&&t.rank!==4)throw new V(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let l=Aw(e,s);if(r==="causal")throw new Me("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return l=kl.conv2d({x:l,filter:t,strides:a,pad:r==="same"?"same":"valid",dilations:i,dataFormat:"NHWC",bias:n,activation:o}),s==="channelsFirst"&&(l=Ee(l,[0,3,1,2])),l})}function zU(e,t,n,a=[1,1,1],r="valid",s,i){return P(()=>{if(s==null&&(s=ja()),Rt(s),e.rank!==4&&e.rank!==5)throw new V(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(t.rank!==4&&t.rank!==5)throw new V(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let o=e2(e,s);if(r==="causal")throw new Me("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return o=fv(o,t,a,r==="same"?"same":"valid","NDHWC",i),n!=null&&(o=Xa(o,n)),s==="channelsFirst"&&(o=Ee(o,[0,4,1,2,3])),o})}var $w=class extends Ge{constructor(e,t){if(super(t),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",$w.verifyArgs(t),this.rank=e,Qt(this.rank,"rank"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new Me(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=dl(t.kernelSize,e,"kernelSize"),this.strides=dl(t.strides==null?1:t.strides,e,"strides"),this.padding=t.padding==null?"valid":t.padding,xa(this.padding),this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Rt(this.dataFormat),this.activation=ps(t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.biasInitializer=St(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Kt(t.biasConstraint),this.biasRegularizer=Tt(t.biasRegularizer),this.activityRegularizer=Tt(t.activityRegularizer),this.dilationRate=dl(t.dilationRate==null?1:t.dilationRate,e,"dilationRate"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new V(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new V(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new V(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(e){if(ar("kernelSize"in e,"required key 'kernelSize' not in config"),typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,3))throw new V(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){let e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:us(this.activation),useBias:this.useBias,biasInitializer:Ct(this.biasInitializer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),biasConstraint:qt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},ed=class extends $w{constructor(e,t){super(e,t),this.kernel=null,ed.verifyArgs(t),this.filters=t.filters,Qt(this.filters,"filters"),this.kernelInitializer=St(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Kt(t.kernelConstraint),this.kernelRegularizer=Tt(t.kernelRegularizer)}build(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V(`The channel dimension of the input should be defined. Found ${e[t]}`);let n=e[t],a=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",a,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:n}}],this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n,a=this.bias==null?null:this.bias.read(),r=lN(this.activation.getClassName());if(r!=null&&this.rank===2)n=jk(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate,r);else{if(this.rank===1)n=LU(e,this.kernel.read(),a,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)n=jk(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)n=zU(e,this.kernel.read(),a,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new Me("convolutions greater than 3D are not implemented yet.");this.activation!=null&&(n=this.activation.apply(n))}return n})}computeOutputShape(e){e=tt(e);let t=[],n=this.dataFormat==="channelsLast"?e.slice(1,e.length-1):e.slice(2);for(let r=0;r<n.length;++r){let s=Ga(n[r],this.kernelSize[r],this.padding,this.strides[r],typeof this.dilationRate=="number"?this.dilationRate:this.dilationRate[r]);t.push(s)}let a=[e[0]];return this.dataFormat==="channelsLast"?(a=a.concat(t),a.push(this.filters)):(a.push(this.filters),a=a.concat(t)),a}getConfig(){let e={filters:this.filters,kernelInitializer:Ct(this.kernelInitializer),kernelRegularizer:pt(this.kernelRegularizer),kernelConstraint:qt(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||typeof e.filters!="number"||e.filters<1)throw new V(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(e.filters)}`)}},td=class extends ed{constructor(e){super(2,e),td.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,2))throw new V(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}};td.className="Conv2D";ne.registerClass(td);var nd=class extends ed{constructor(e){super(3,e),nd.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!(Array.isArray(e.kernelSize)&&(e.kernelSize.length===1||e.kernelSize.length===3)))throw new V(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}};nd.className="Conv3D";ne.registerClass(nd);var Fw=class extends td{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:4})],this.padding!=="same"&&this.padding!=="valid")throw new V(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=tt(e),e.length!==4)throw new V("Input should have rank 4; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],a=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",a,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:4,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{let n=Ne(e);if(n.shape.length!==4)throw new V(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let a=n.shape,r=a[0],s,i;this.dataFormat==="channelsFirst"?(s=2,i=3):(s=1,i=2);let o=a[s],l=a[i],u=this.kernelSize[0],p=this.kernelSize[1],d=this.strides[0],c=this.strides[1],h=rr(o,d,u,this.padding),m=rr(l,c,p,this.padding),f=[r,h,m,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,1]));let g=Km(n,this.kernel.read(),f,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(g=Ee(g,[0,3,1,2])),this.bias!=null&&(g=Xa(g,this.bias.read(),this.dataFormat)),this.activation!=null&&(g=this.activation.apply(g)),g})}computeOutputShape(e){e=tt(e);let t=e.slice(),n,a,r;this.dataFormat==="channelsFirst"?(n=1,a=2,r=3):(n=3,a=1,r=2);let s=this.kernelSize[0],i=this.kernelSize[1],o=this.strides[0],l=this.strides[1];return t[n]=this.filters,t[a]=rr(t[a],o,s,this.padding),t[r]=rr(t[r],l,i,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Fw.className="Conv2DTranspose";ne.registerClass(Fw);var Dw=class extends nd{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:5})],this.padding!=="same"&&this.padding!=="valid")throw new V(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=tt(e),e.length!==5)throw new V("Input should have rank 5; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new V("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],a=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",a,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:5,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{let n=Ne(e);if(n.shape.length!==5)throw new V(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let a=n.shape,r=a[0],s,i,o;this.dataFormat==="channelsFirst"?(o=2,s=3,i=4):(o=1,s=2,i=3);let l=a[o],u=a[s],p=a[i],d=this.kernelSize[0],c=this.kernelSize[1],h=this.kernelSize[2],m=this.strides[0],f=this.strides[1],g=this.strides[2],y=rr(l,m,d,this.padding),b=rr(u,f,c,this.padding),x=rr(p,g,h,this.padding),w=[r,y,b,x,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,4,1]));let I=gv(n,this.kernel.read(),w,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(I=Ee(I,[0,4,1,2,3])),this.bias!==null&&(I=Xa(I,this.bias.read(),this.dataFormat)),this.activation!==null&&(I=this.activation.apply(I)),I})}computeOutputShape(e){e=tt(e);let t=e.slice(),n,a,r,s;this.dataFormat==="channelsFirst"?(n=1,a=2,r=3,s=4):(n=4,a=1,r=2,s=3);let i=this.kernelSize[0],o=this.kernelSize[1],l=this.kernelSize[2],u=this.strides[0],p=this.strides[1],d=this.strides[2];return t[n]=this.filters,t[a]=rr(t[a],u,i,this.padding),t[r]=rr(t[r],p,o,this.padding),t[s]=rr(t[s],d,l,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Dw.className="Conv3DTranspose";ne.registerClass(Dw);var t2=class extends ed{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,t.filters==null)throw new V("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(t.kernelInitializer!=null||t.kernelRegularizer!=null||t.kernelConstraint!=null)throw new V("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(t.padding!=null&&t.padding!=="same"&&t.padding!=="valid")throw new V(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=Tt(t.depthwiseRegularizer),this.depthwiseConstraint=Kt(t.depthwiseConstraint),this.pointwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=Tt(t.pointwiseRegularizer),this.pointwiseConstraint=Kt(t.pointwiseConstraint)}build(e){if(e=tt(e),e.length<this.rank+2)throw new V(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null||e[t]<0)throw new V(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(e[t])}`);let n=e[t],a=this.kernelSize.concat([n,this.depthMultiplier]),r=[];for(let i=0;i<this.rank;++i)r.push(1);r.push(n*this.depthMultiplier,this.filters);let s=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",a,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,s,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",r,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,s,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,s,this.biasConstraint):this.bias=null,this.inputSpec=[new zt({ndim:this.rank+2,axes:{[t]:n}})],this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n;if(this.rank===1)throw new Me("1D separable convolution is not implemented yet.");return this.rank===2&&(this.dataFormat==="channelsFirst"&&(e=Ee(e,[0,2,3,1])),n=Ts(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(n=Xa(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),this.dataFormat==="channelsFirst"&&(n=Ee(n,[0,3,1,2])),n})}getConfig(){let e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=Ct(this.depthwiseInitializer),e.pointwiseInitializer=Ct(this.pointwiseInitializer),e.depthwiseRegularizer=pt(this.depthwiseRegularizer),e.pointwiseRegularizer=pt(this.pointwiseRegularizer),e.depthwiseConstraint=qt(this.depthwiseConstraint),e.pointwiseConstraint=qt(this.pointwiseConstraint),e}};t2.className="SeparableConv";var Rw=class extends t2{constructor(e){super(2,e)}};Rw.className="SeparableConv2D";ne.registerClass(Rw);var Wf=class extends ed{constructor(e){super(1,e),Wf.verifyArgs(e),this.inputSpec=[{ndim:3}]}getConfig(){let e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,1))throw new V(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}};Wf.className="Conv1D";ne.registerClass(Wf);var Mw=class extends Ge{constructor(e){super(e),typeof e.cropping=="number"?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:typeof e.cropping[0]=="number"?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=e.dataFormat===void 0?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return this.dataFormat==="channelsFirst"?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return P(()=>{if(e=Ne(e),this.dataFormat==="channelsLast"){let n=uh(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return uh(n,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let n=uh(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return uh(n,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Mw.className="Cropping2D";ne.registerClass(Mw);var Pw=class extends Ge{constructor(e){super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=e.size==null?this.DEFAULT_SIZE:e.size,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),this.interpolation=e.interpolation==null?"nearest":e.interpolation,Y4(this.interpolation)}computeOutputShape(e){if(this.dataFormat==="channelsFirst"){let t=e[2]==null?null:this.size[0]*e[2],n=e[3]==null?null:this.size[1]*e[3];return[e[0],e[1],t,n]}else{let t=e[1]==null?null:this.size[0]*e[1],n=e[2]==null?null:this.size[1]*e[2];return[e[0],t,n,e[3]]}}call(e,t){return P(()=>{let n=Ne(e),a=n.shape;if(this.dataFormat==="channelsFirst"){n=Ee(n,[0,2,3,1]);let r=this.size[0]*a[2],s=this.size[1]*a[3],i=this.interpolation==="nearest"?za.resizeNearestNeighbor(n,[r,s]):za.resizeBilinear(n,[r,s]);return Ee(i,[0,3,1,2])}else{let r=this.size[0]*a[1],s=this.size[1]*a[2];return this.interpolation==="nearest"?za.resizeNearestNeighbor(n,[r,s]):za.resizeBilinear(n,[r,s])}})}getConfig(){let e={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}};Pw.className="UpSampling2D";ne.registerClass(Pw);function WU(e,t,n=[1,1],a="valid",r,s){return P(()=>{r==null&&(r=ja()),Rt(r);let i=Aw(e,r);if(e.rank!==4)throw new V(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(t.rank!==4)throw new V(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=Is(i,t,n,a==="same"?"same":"valid","NHWC",s),r==="channelsFirst"&&(i=Ee(i,[0,3,1,2])),i})}var Ow=class extends $w{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=St(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Kt(e.depthwiseConstraint),this.depthwiseRegularizer=Tt(e.depthwiseRegularizer)}build(e){if(e=tt(e),e.length<4)throw new V(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);let t=this.dataFormat==="channelsFirst"?1:3;if(e[t]==null||e[t]<0)throw new V(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);let n=e[t],a=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",a,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{e=Ne(e);let n=WU(e,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(n=Xa(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),n})}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],a=this.dataFormat==="channelsFirst"?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,r=Ga(t,this.kernelSize[0],this.padding,this.strides[0]),s=Ga(n,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat==="channelsFirst"?[e[0],a,r,s]:[e[0],r,s,a]}getConfig(){let e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=Ct(this.depthwiseInitializer),e.depthwiseRegularizer=pt(this.depthwiseRegularizer),e.depthwiseConstraint=qt(this.depthwiseRegularizer),e}};Ow.className="DepthwiseConv2D";ne.registerClass(Ow);function n2(e,t,n,a){if(Array.isArray(e)){if(t!=null||n!=null)throw new V("When inputs is an array, neither initialState or constants should be provided");a!=null&&(n=e.slice(e.length-a,e.length),e=e.slice(0,e.length-a)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function r(s){return s==null||Array.isArray(s)?s:[s]}return t=r(t),n=r(n),{inputs:e,initialState:t,constants:n}}function a2(e,t,n,a=!1,r,s,i=!1,o=!1){return P(()=>{let l=t.shape.length;if(l<3)throw new V(`Input should be at least 3D, but is ${l}D.`);let u=[1,0].concat(Ha(2,l));if(t=Ee(t,u),s!=null)throw new Me("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),r!=null&&(r=oe(oe(r,"bool"),"float32"),r.rank===l-1&&(r=Zt(r,-1)),r=Ee(r,u)),a&&(t=ga(t,0),r!=null&&(r=ga(r,0)));let p=[],d,c=n,h=t.shape[0],m=ct(t),f;r!=null&&(f=ct(r));for(let y=0;y<h;++y){let b=m[y],x=P(()=>e(b,c));if(r==null)d=x[0],c=x[1];else{let w=P(()=>{let I=f[y],T=pe(na(I),I),C=Y(z(x[0],I),z(c[0],T)),E=c.map((A,R)=>Y(z(x[1][R],I),z(A,T)));return{output:C,newStates:E}});d=w.output,c=w.newStates}o&&p.push(d)}let g;return o&&(g=Ft(p,1)),[d,g,c]})}var mr=class extends Ge{constructor(e){super(e);let t;if(e.cell==null)throw new V("cell property is missing for the constructor of RNN.");if(Array.isArray(e.cell)?t=new Uf({cells:e.cell}):t=e.cell,t.stateSize==null)throw new V("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=e.returnSequences==null?!1:e.returnSequences,this.returnState=e.returnState==null?!1:e.returnState,this.goBackwards=e.goBackwards==null?!1:e.goBackwards,this._stateful=e.stateful==null?!1:e.stateful,this.unroll=e.unroll==null?!1:e.unroll,this.supportsMasking=!0,this.inputSpec=[new zt({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return Ha(0,e).map(t=>null)}else return this.states_}setStates(e){this.states_=e}computeOutputShape(e){jb(e)&&(e=e[0]),e=e;let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);let n=t[0],a;if(this.returnSequences?a=[e[0],e[1],n]:a=[e[0],n],this.returnState){let r=[];for(let s of t)r.push([e[0],s]);return[a].concat(r)}else return a}computeMask(e,t){return P(()=>{Array.isArray(t)&&(t=t[0]);let n=this.returnSequences?t:null;if(this.returnState){let a=this.states.map(r=>null);return[n].concat(a)}else return n})}get states(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let n=0;n<e;++n)t.push(null);return t}else return this.states_}set states(e){this.states_=e}build(e){if(this.numConstants!=null)throw new Me("Constants support is not implemented in RNN yet.");jb(e)&&(e=e[0]),e=e;let t=this.stateful?e[0]:null,n=e.slice(2);this.inputSpec[0]=new zt({shape:[t,null,...n]});let a=[e[0]].concat(e.slice(2));this.cell.build(a);let r;if(Array.isArray(this.cell.stateSize)?r=this.cell.stateSize:r=[this.cell.stateSize],this.stateSpec!=null){if(!v.arraysEqual(this.stateSpec.map(s=>s.shape[s.shape.length-1]),r))throw new V(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=r.map(s=>new zt({shape:[null,s]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){P(()=>{if(!this.stateful)throw new Ir("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape[0];if(n==null)throw new V("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(a=>It([n,a])):this.states_=[It([n,this.cell.stateSize])];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(a=>It([n,a])):this.states_[0]=It([n,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new V(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t===!0?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let a=0;a<this.states_.length;++a){let r=e[a],s=Array.isArray(this.cell.stateSize)?this.cell.stateSize[a]:this.cell.stateSize,i=[n,s];if(!v.arraysEqual(r.shape,i))throw new V(`State ${a} is incompatible with layer ${this.name}: expected shape=${i}, received shape=${r.shape}`);this.states_[a]=r}}this.states_=this.states_.map(a=>Jt(a.clone()))})}apply(e,t){let n=t==null?null:t.initialState,a=t==null?null:t.constants;t==null&&(t={});let r=n2(e,n,a,this.numConstants);e=r.inputs,n=r.initialState,a=r.constants;let s=[],i=[];if(n!=null){t.initialState=n,s=s.concat(n),this.stateSpec=[];for(let o of n)this.stateSpec.push(new zt({shape:o.shape}));i=i.concat(this.stateSpec)}if(a!=null&&(t.constants=a,s=s.concat(a),this.numConstants=a.length),s[0]instanceof Ba){let o=[e].concat(s),l=this.inputSpec.concat(i),u=this.inputSpec;this.inputSpec=l;let p=super.apply(o,t);return this.inputSpec=u,p}else return super.apply(e,t)}call(e,t){return P(()=>{let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;e=Ne(e),r==null&&(this.stateful?r=this.states_:r=this.getInitialState(e));let s=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(r.length!==s)throw new V(`RNN Layer has ${s} state(s) but was passed ${r.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let i={training:a},o=a2((c,h)=>{let m=this.cell.call([c].concat(h),i);return[m[0],m.slice(1)]},e,r,this.goBackwards,n,null,this.unroll,this.returnSequences),l=o[0],u=o[1],p=o[2];this.stateful&&this.resetStates(p,a);let d=this.returnSequences?u:l;return this.returnState?[d].concat(p):d})}getInitialState(e){return P(()=>{let t=It(e.shape);return t=fe(t,[1,2]),t=Xc(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(n=>n>1?Gb(t,[1,n]):t):this.cell.stateSize>1?[Gb(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){let e=super.getConfig(),t={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(t.numConstants=this.numConstants);let n=this.cell.getConfig();return this.getClassName()===mr.className&&(t.cell={className:this.cell.getClassName(),config:n}),Object.assign(Object.assign(Object.assign({},n),e),t)}static fromConfig(e,t,n={}){let a=t.cell,r=Ua(a,n);return new e(Object.assign(t,{cell:r}))}};mr.className="RNN";ne.registerClass(mr);var ad=class extends Ge{},Bf=class extends ad{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=ps(e.activation==null?this.DEFAULT_ACTIVATION:e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=Il([1,ls([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Il([1,ls([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=tt(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{if(e=e,e.length!==2)throw new V(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let n=e[1];e=e[0];let a=t.training==null?!1:t.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cs({ones:()=>na(e),rate:this.dropout,training:a,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cs({ones:()=>na(n),rate:this.recurrentDropout,training:a,dropoutFunc:this.dropoutFunc}));let r,s=this.dropoutMask,i=this.recurrentDropoutMask;s!=null?r=or(z(e,s),this.kernel.read()):r=or(e,this.kernel.read()),this.bias!=null&&(r=Xa(r,this.bias.read())),i!=null&&(n=z(n,i));let o=Y(r,or(n,this.recurrentKernel.read()));return this.activation!=null&&(o=this.activation.apply(o)),[o,o]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:us(this.activation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},e),t)}};Bf.className="SimpleRNNCell";ne.registerClass(Bf);var Lw=class extends mr{constructor(e){e.cell=new Bf(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return new e(t)}};Lw.className="SimpleRNN";ne.registerClass(Lw);var Vf=class extends ad{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new V("GRUCell does not support reset_after parameter set to true.");this.units=e.units,Qt(this.units,"units"),this.activation=ps(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=ps(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=Il([1,ls([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Il([1,ls([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=tt(e);let t=e[e.length-1];this.kernel=this.addWeight("kernel",[t,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return P(()=>{if(e=e,e.length!==2)throw new V(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training==null?!1:t.training,a=e[1];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cs({ones:()=>na(e),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cs({ones:()=>na(a),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let r=this.dropoutMask,s=this.recurrentDropoutMask,i,o,l;0<this.dropout&&this.dropout<1&&(e=z(e,r[0]));let u=or(e,this.kernel.read());this.useBias&&(u=Xa(u,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(a=z(a,s[0]));let p=this.recurrentKernel.read(),[d,c]=zn(p,[2*this.units,this.units],p.rank-1),h=or(a,d),[m,f,g]=zn(u,3,u.rank-1),[y,b]=zn(h,2,h.rank-1);i=this.recurrentActivation.apply(Y(m,y)),o=this.recurrentActivation.apply(Y(f,b));let x=or(z(o,a),c);l=this.activation.apply(Y(g,x));let w=Y(z(i,a),z(Y(1,vt(i)),l));return[w,w]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:us(this.activation),recurrentActivation:us(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},e),t)}};Vf.className="GRUCell";ne.registerClass(Vf);var zw=class extends mr{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new Vf(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};zw.className="GRU";ne.registerClass(zw);var rd=class extends ad{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=ps(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=ps(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=Il([1,ls([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Il([1,ls([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;e=tt(e);let n=e[e.length-1];this.kernel=this.addWeight("kernel",[n,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let a;if(this.useBias){if(this.unitForgetBias){let r=this.biasInitializer,s=this.units;a=new(t=class extends Ra{apply(i,o){let l=r.apply([s]),u=new Af().apply([s]),p=r.apply([s*2]);return Ak(Ak(l,u),p)}},t.className="CustomInit",t)}else a=this.biasInitializer;this.bias=this.addWeight("bias",[this.units*4],null,a,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return P(()=>{let n=t.training==null?!1:t.training;if(e=e,e.length!==3)throw new V(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let a=e[1],r=e[2];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cs({ones:()=>na(e),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cs({ones:()=>na(a),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,i=this.recurrentDropoutMask,o,l,u,p;0<this.dropout&&this.dropout<1&&(e=z(e,s[0]));let d=or(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(a=z(a,i[0])),d=Y(d,or(a,this.recurrentKernel.read())),this.useBias&&(d=Xa(d,this.bias.read()));let[c,h,m,f]=zn(d,4,d.rank-1);o=this.recurrentActivation.apply(c),l=this.recurrentActivation.apply(h),u=Y(z(l,r),z(o,this.activation.apply(m))),p=this.recurrentActivation.apply(f);let g=z(p,this.activation.apply(u));return[g,g,u]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:us(this.activation),recurrentActivation:us(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),recurrentInitializer:Ct(this.recurrentInitializer),biasInitializer:Ct(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:pt(this.kernelRegularizer),recurrentRegularizer:pt(this.recurrentRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),recurrentConstraint:qt(this.recurrentConstraint),biasConstraint:qt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},e),t)}};rd.className="LSTMCell";ne.registerClass(rd);var Ww=class extends mr{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new rd(e),super(e)}call(e,t){return P(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};Ww.className="LSTM";ne.registerClass(Ww);var Uf=class extends ad{constructor(e){super(e),this.cells=e.cells}get stateSize(){let e=[];for(let t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return P(()=>{e=e;let n=e.slice(1),a=[];for(let i of this.cells.slice().reverse())Array.isArray(i.stateSize)?a.push(n.splice(0,i.stateSize.length)):a.push(n.splice(0,1));a.reverse();let r=[],s;for(let i=0;i<this.cells.length;++i){let o=this.cells[i];n=a[i],i===0?s=[e[0]].concat(n):s=[s[0]].concat(n),s=o.call(s,t),r.push(s.slice(1))}n=[];for(let i of r.slice().reverse())n.push(...i);return[s[0]].concat(n)})}build(e){jb(e)&&(e=e[0]),e=e;let t;this.cells.forEach((n,a)=>{Ys(`RNNCell_${a}`,()=>{n.build(e),Array.isArray(n.stateSize)?t=n.stateSize[0]:t=n.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){let e=super.getConfig(),t=a=>({className:a.getClassName(),config:a.getConfig()}),n={cells:this.cells.map(t)};return Object.assign(Object.assign({},e),n)}static fromConfig(e,t,n={}){let a=[];for(let r of t.cells)a.push(Ua(r,n));return new e({cells:a})}get trainableWeights(){if(!this.trainable)return[];let e=[];for(let t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.cells)t.push(...n.trainableWeights);return t.concat(e)}return e}getWeights(){let e=[];for(let t of this.cells)e.push(...t.weights);return qb(e)}setWeights(e){let t=[];for(let n of this.cells){let a=n.weights.length,r=e.splice(a);for(let s=0;s<n.weights.length;++s)t.push([n.weights[s],r[s]])}pw(t)}};Uf.className="StackedRNNCells";ne.registerClass(Uf);function cs(e){let{ones:t,rate:n,training:a=!1,count:r=1,dropoutFunc:s}=e,i=()=>s!=null?s(t(),n):fN(t(),n),o=()=>Zc(i,t,a);return!r||r<=1?Jt(o().clone()):Array(r).fill(void 0).map(o).map(l=>Jt(l.clone()))}var BU=function(e,t){var n={};for(var a in e)Object.prototype.hasOwnProperty.call(e,a)&&t.indexOf(a)<0&&(n[a]=e[a]);if(e!=null&&typeof Object.getOwnPropertySymbols=="function")for(var r=0,a=Object.getOwnPropertySymbols(e);r<a.length;r++)t.indexOf(a[r])<0&&Object.prototype.propertyIsEnumerable.call(e,a[r])&&(n[a[r]]=e[a[r]]);return n},r2=class extends mr{constructor(e){if(e.unroll)throw new Me("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new Me("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new zt({ndim:5})]}call(e,t){return P(()=>{if(this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new V("ConvRNN2D cell does not support constants");let n=t==null?null:t.mask,a=t==null?null:t.training,r=t==null?null:t.initialState;return super.call(e,{mask:n,training:a,initialState:r})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return P(()=>{let{stateSize:t}=this.cell,n=e.shape,a=this.computeSingleOutputShape(n),r=[a[0],...a.slice(2)],s=It(r);return Array.isArray(t)?Array(t.length).fill(s):[s]})}resetStates(e,t=!1){P(()=>{if(!this.stateful)throw new Ir("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape,a=this.computeSingleOutputShape(n),r=[a[0],...a.slice(2)];if(n[0]==null)throw new V("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>It(r)):this.states_=[It(r)];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>It(r)):this.states_[0]=It(r);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new V(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let s=0;s<this.states_.length;++s){let i=e[s],o=r;if(!v.arraysEqual(i.shape,o))throw new V(`State ${s} is incompatible with layer ${this.name}: expected shape=${o}, received shape=${i.shape}`);this.states_[s]=i}}this.states_=this.states_.map(s=>Jt(s.clone()))})}computeSingleOutputShape(e){let{dataFormat:t,filters:n,kernelSize:a,padding:r,strides:s,dilationRate:i}=this.cell,o=t==="channelsFirst",l=e[o?3:2],u=e[o?4:3],p=Ga(l,a[0],r,s[0],i[0]),d=Ga(u,a[1],r,s[1],i[1]);return[...e.slice(0,2),...o?[n,p,d]:[p,d,n]]}};r2.className="ConvRNN2D";var Gf=class extends rd{constructor(e){let{filters:t,kernelSize:n,strides:a,padding:r,dataFormat:s,dilationRate:i}=e;super(Object.assign(Object.assign({},e),{units:t})),this.filters=t,Qt(this.filters,"filters"),this.kernelSize=dl(n,2,"kernelSize"),this.kernelSize.forEach(o=>Qt(o,"kernelSize")),this.strides=dl(a||1,2,"strides"),this.strides.forEach(o=>Qt(o,"strides")),this.padding=r||"valid",xa(this.padding),this.dataFormat=s||"channelsLast",Rt(this.dataFormat),this.dilationRate=dl(i||1,2,"dilationRate"),this.dilationRate.forEach(o=>Qt(o,"dilationRate"))}build(e){var t;e=tt(e);let n=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[n]==null)throw new V(`The channel dimension of the input should be defined. Found ${e[n]}`);let a=e[n],r=4,s=this.kernelSize.concat([a,this.filters*r]);this.kernel=this.addWeight("kernel",s,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let i=this.kernelSize.concat([this.filters,this.filters*r]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",i,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let o;if(this.unitForgetBias){let l=this.biasInitializer,u=this.filters;o=new(t=class extends Ra{apply(p,d){let c=l.apply([u]),h=Jn([u]),m=l.apply([u*2]);return nw([c,h,m])}},t.className="CustomInit",t)}else o=this.biasInitializer;this.bias=this.addWeight("bias",[this.filters*r],null,o,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return P(()=>{if(e.length!==3)throw new V(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training||!1,a=e[0],r=e[1],s=e[2],i=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cs({ones:()=>na(a),rate:this.dropout,training:n,count:i,dropoutFunc:this.dropoutFunc}));let o=this.dropoutMask,l=(Z,Q,ee)=>!Q||!Q[ee]?Z:z(Q[ee],Z),u=l(a,o,0),p=l(a,o,1),d=l(a,o,2),c=l(a,o,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cs({ones:()=>na(r),rate:this.recurrentDropout,training:n,count:i,dropoutFunc:this.dropoutFunc}));let h=this.recurrentDropoutMask,m=l(r,h,0),f=l(r,h,1),g=l(r,h,2),y=l(r,h,3),b=3,[x,w,I,T]=zn(this.kernel.read(),i,b),[C,E,A,R]=this.useBias?zn(this.bias.read(),i):[null,null,null,null];u=this.inputConv(u,x,C,this.padding),p=this.inputConv(p,w,E,this.padding),d=this.inputConv(d,I,A,this.padding),c=this.inputConv(c,T,R,this.padding);let[F,S,M,B]=zn(this.recurrentKernel.read(),i,b);m=this.recurrentConv(m,F),f=this.recurrentConv(f,S),g=this.recurrentConv(g,M),y=this.recurrentConv(y,B);let U=this.recurrentActivation.apply(Y(u,m)),G=this.recurrentActivation.apply(Y(p,f)),q=Y(z(G,s),z(U,this.activation.apply(Y(d,g)))),K=z(this.recurrentActivation.apply(Y(c,y)),this.activation.apply(q));return[K,K,q]})}getConfig(){let e=super.getConfig(),{units:t}=e,n=BU(e,["units"]),a={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),a)}inputConv(e,t,n,a){let r=$t(e,t,this.strides,a||"valid",this.dataFormat==="channelsFirst"?"NCHW":"NHWC",this.dilationRate);return n?Xa(r,n,this.dataFormat):r}recurrentConv(e,t){return $t(e,t,1,"same",this.dataFormat==="channelsFirst"?"NCHW":"NHWC")}};Gf.className="ConvLSTM2DCell";ne.registerClass(Gf);var Bw=class extends r2{constructor(e){let t=new Gf(e);super(Object.assign(Object.assign({},e),{cell:t}))}static fromConfig(e,t){return new e(t)}};Bw.className="ConvLSTM2D";ne.registerClass(Bw);var Hf=class extends Ge{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(this.noiseShape==null)return this.noiseShape;let t=e.shape,n=[];for(let a=0;a<this.noiseShape.length;++a)n.push(this.noiseShape[a]==null?t[a]:this.noiseShape[a]);return n}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);if(0<this.rate&&this.rate<1){let a=t.training==null?!1:t.training,r=this.getNoiseShape(n);return Zc(()=>fN(n,this.rate,r,this.seed),()=>n,a)}return e})}getConfig(){let e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};Hf.className="Dropout";ne.registerClass(Hf);var Vw=class extends Hf{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){let t=e.shape;return[t[0],1,t[2]]}};Vw.className="SpatialDropout1D";ne.registerClass(Vw);var Uw=class extends Ge{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.batchInputShape==null&&e.inputShape==null&&e.inputDim!=null){let t=null;e.batchSize!=null&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,Qt(this.units,"units"),this.activation=ps(e.activation),e.useBias!=null&&(this.useBias=e.useBias),this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Kt(e.kernelConstraint),this.biasConstraint=Kt(e.biasConstraint),this.kernelRegularizer=Tt(e.kernelRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){e=tt(e);let t=e[e.length-1];this.kernel==null&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){e=tt(e);let t=e.slice();return t[t.length-1]=this.units,t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=lN(this.activation.getClassName()),r;return a!=null?r=or(n,this.kernel.read(),a,this.bias?this.bias.read():null):(r=or(n,this.kernel.read()),this.bias!=null&&(r=Xa(r,this.bias.read())),this.activation!=null&&(r=this.activation.apply(r))),r})}getConfig(){let e={units:this.units,activation:us(this.activation),useBias:this.useBias,kernelInitializer:Ct(this.kernelInitializer),biasInitializer:Ct(this.biasInitializer),kernelRegularizer:pt(this.kernelRegularizer),biasRegularizer:pt(this.biasRegularizer),activityRegularizer:pt(this.activityRegularizer),kernelConstraint:qt(this.kernelConstraint),biasConstraint:qt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};Uw.className="Dense";ne.registerClass(Uw);var Gw=class extends Ge{constructor(e){e=e||{},super(e),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=tt(e);for(let t of e.slice(1))if(t==null)throw new V(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],rs(e,1)]}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);if(this.dataFormat==="channelsFirst"&&n.rank>1){let a=[0];for(let r=2;r<n.rank;++r)a.push(r);a.push(1),n=Ee(n,a)}return nV(n)})}getConfig(){let e={};this.dataFormat!=null&&(e.dataFormat=this.dataFormat);let t=super.getConfig();return Object.assign(e,t),e}};Gw.className="Flatten";ne.registerClass(Gw);var Hw=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.activation=ps(e.activation)}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return this.activation.apply(n)})}getConfig(){let e={activation:us(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};Hw.className="Activation";ne.registerClass(Hw);var jw=class extends Ge{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return P(()=>(e=Ne(e),eV(e,this.n)))}getConfig(){let e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};jw.className="RepeatVector";ne.registerClass(jw);var qw=class extends Ge{constructor(e){super(e),this.targetShape=e.targetShape;for(let t=0;t<this.targetShape.length;++t)this.isUnknown(this.targetShape[t])&&(this.targetShape[t]=null)}isUnknown(e){return e<0||e==null}fixUnknownDimension(e,t){let n="Total size of new array must be unchanged.",a=t.slice(),r=1,s=null;for(let o=0;o<a.length;++o){let l=a[o];if(this.isUnknown(l))if(s===null)s=o;else throw new V("Can only specifiy one unknown dimension.");else r*=l}let i=rs(e);if(s!==null){if(r===0||i%r!==0)throw new V(n);a[s]=i/r}else if(i!==r)throw new V(n);return a}computeOutputShape(e){let t=!1;for(let n=0;n<e.length;++n)if(this.isUnknown(e[n])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=n.shape,r=a.slice(0,1).concat(this.fixUnknownDimension(a.slice(1),this.targetShape));return W(n,r)})}getConfig(){let e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};qw.className="Reshape";ne.registerClass(qw);var Kw=class extends Ge{constructor(e){if(super(e),e.dims==null)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);let t=Ha(1,e.dims.length+1);if(!v.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new zt({ndim:this.dims.length+1})]}computeOutputShape(e){e=tt(e);let t=e.slice();return this.dims.forEach((n,a)=>{t[a+1]=e[n]}),t}call(e,t){return Ee(Ne(e),this.dimsIncludingBatch)}getConfig(){let e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};Kw.className="Permute";ne.registerClass(Kw);var Xw=class extends Ge{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null?this.maskValue=e.maskValue==null?0:e.maskValue:this.maskValue=0}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){let n=Ne(e),a=-1;return Qp(oi(n,this.maskValue),a)}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e),a=-1,r=!0,s=Qp(oi(n,this.maskValue),a,r);return z(n,oe(s,n.dtype))})}};Xw.className="Masking";ne.registerClass(Xw);var Yw=class extends Ge{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",e.batchInputShape==null&&e.inputShape==null){let t=null;e.batchSize!=null&&(t=e.batchSize),e.inputLength==null?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(xt(e.inputLength))}this.inputDim=e.inputDim,Qt(this.inputDim,"inputDim"),this.outputDim=e.outputDim,Qt(this.outputDim,"outputDim"),this.embeddingsInitializer=St(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=Tt(e.embeddingsRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.embeddingsConstraint=Kt(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return P(()=>this.maskZero?(e=Ne(e),oi(e,qe(e))):null)}computeOutputShape(e){if(e=tt(e),this.inputLength==null)return[...e,this.outputDim];let t=xt(this.inputLength);if(t.length!==e.length-1)throw new V(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let n=0;for(let a=0;a<t.length;++a){let r=t[a],s=e[a+1];if(r!=null&&s!=null&&r!==s)throw new V(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);r==null&&(t[n]=s),n++}}return[e[0],...t,this.outputDim]}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);n.dtype!=="int32"&&(n=bo(n,"int32"));let a=mN(this.embeddings.read(),W(n,[n.size]));return W(a,tt(this.computeOutputShape(n.shape)))})}getConfig(){let e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Ct(this.embeddingsInitializer),embeddingsRegularizer:pt(this.embeddingsRegularizer),activityRegularizer:pt(this.activityRegularizer),embeddingsConstraint:qt(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};Yw.className="Embedding";ne.registerClass(Yw);var vo=class extends Ge{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new Me}computeElementwiseOpOutputShape(e,t){if(e==null||t==null)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(t.length===0)return e;let n=e.slice(0,e.length-t.length);for(let a=0;a<t.length;++a){let r=e[e.length-t.length+a],s=t[a];if(r==null||s==null||r<0||s<0)n.push(null);else if(r===1)n.push(s);else if(s===1)n.push(r);else{if(r!==s)throw new V("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));n.push(r)}}return n}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[tt(e)]),e=e,e.length<2)throw new V(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(let r of e)r!=null&&r[0]!==null&&t.push(r[0]);if(t=as(t),t.length>1)throw new V(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let n=e[0]==null?null:e[0].slice(1);for(let r=1;r<e.length;++r){let s=e[r]==null?null:e[r].slice(1);n=this.computeElementwiseOpOutputShape(n,s)}let a=e.map(r=>r.length);e.indexOf(null)===-1&&as(a).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return P(()=>{if(e=e,this.reshapeRequired){let n=[],a=e.map(r=>r.rank);if(a.indexOf(null)===-1){let r=ls(a);for(let s of e){let i=s.rank;for(let o=0;o<r-i;++o)s=Xc(s,1);n.push(s)}return this.mergeFunction(n)}else{let r=!1;for(let o of e){let l=o.rank;if(l==null){let u=o.shape,p=u[0],d=u.slice(1).concat([p]),c=W(o,[p].concat(rs(u.slice(1))));c=Ee(c,[1,0]),c=W(c,d),n.push(c),r=!0}else if(l>1){let u=Ha(1,l).concat([0]);n.push(Ee(o,u)),r=!0}else n.push(o)}let s=this.mergeFunction(n),i=s.rank;if(r){if(i==null){let o=s.shape,l=o.length,u=o[l-1],p=[u].concat(o.slice(0,o.length-1));s=W(Ee(W(s,[-1,u]),[1,0]),p)}else if(i>1){let o=[i-1].concat(Ha(0,i-1));s=Ee(s,o)}}return s}}else return this.mergeFunction(e)})}computeOutputShape(e){e=e;let t;e[0]==null?t=null:t=e[0].slice(1);for(let a=1;a<e.length;++a){let r=e[a]==null?null:e[a].slice(1);t=this.computeElementwiseOpOutputShape(t,r)}let n=[];for(let a of e)a!=null&&a[0]!==null&&n.push(a[0]);return n=as(n),n.length===1?t=n.concat(t):t=[null].concat(t),t}computeMask(e,t){return P(()=>{if(t==null)return null;if(!Array.isArray(t))throw new V("`mask` should be an Array");if(!Array.isArray(e))throw new V("`inputs` should be an Array");if(t.length!==e.length)throw new V(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(a=>a==null))return null;t=t.map(a=>a==null?a:Zt(a,0));let n=t[0];for(let a=1;a<t.length-1;++a)n=$a(n,t[a]);return n})}},Zw=class extends vo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return t})}};Zw.className="Add";ne.registerClass(Zw);var Jw=class extends vo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=z(t,e[n]);return t})}};Jw.className="Multiply";ne.registerClass(Jw);var Qw=class extends vo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return z(1/e.length,t)})}};Qw.className="Average";ne.registerClass(Qw);var e0=class extends vo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=dr(t,e[n]);return t})}};e0.className="Maximum";ne.registerClass(e0);var t0=class extends vo{constructor(e){super(e)}mergeFunction(e){return P(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=Vu(t,e[n]);return t})}};t0.className="Minimum";ne.registerClass(t0);var n0=class extends vo{constructor(e){super(e),this.DEFAULT_AXIS=-1,e==null&&(e={}),this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!(Array.isArray(e)&&Array.isArray(e[0]))||e.length===1)throw new V("A `Concatenate` layer should be called on a list of at least 2 inputs");e=e;let t=!0;for(let a of e)if(a!=null){t=!1;break}if(t)return;let n=[];for(let a=0;a<e.length;++a){let r=e[a].slice();r.splice(this.axis,1);let s=!1;for(let i of n)if(v.arraysEqual(i,r)){s=!0;break}s||n.push(r)}if(n.length>1)throw new V("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return P(()=>nw(e,this.axis))}computeOutputShape(e){if(!(Array.isArray(e)&&Array.isArray(e[0])))throw new V("A `Concatenate` layer should be called on a list of inputs.");let t=e,n=t[0].slice(),a=this.axis<0?n.length+this.axis:this.axis;for(let r of t.slice(1)){if(n[a]==null||r[a]==null){n[a]=null;break}n[a]+=r[a]}return n}computeMask(e,t){if(t==null)return null;if(!Array.isArray(t))throw new V("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new V("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new V(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return P(()=>{let n=!0;if(t.forEach(s=>{if(s!=null){n=!1;return}}),n)return null;let a=[];for(let s=0;s<e.length;++s)t[s]==null?a.push(oe(na(e[s]),"bool")):t[s].rank<e[s].rank?a.push(Zt(t[s],-1)):a.push(t[s]);let r=Ze(a,this.axis);return jm(r,-1,!1)})}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};n0.className="Concatenate";ne.registerClass(n0);function Ap(e,t){for(;e<0;)e+=t;return e}function VU(e,t,n){if(e.shape.length>3||t.shape.length>3)throw new Me("batchDot is not implemented for tensors of 4D or higher rank yet");if(v.assert(e.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${e.shape.length}`),v.assert(e.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof n=="number"&&(n=[n,n]),e.dtype==="complex64"||t.dtype==="complex64")throw new Me("batchDot is not implemented for complex64-type Tensors yet.");let a=e.shape.length,r=t.shape.length;n==null&&(n=[a-1,r-2]);let s=n;return P(()=>{let i;if(a>r){i=a-r;let l=[];for(let u=0;u<i;++u)l.push(1);t=W(t,t.shape.concat(l))}else if(r>a){i=r-a;let l=[];for(let u=0;u<i;++u)l.push(1);e=W(e,e.shape.concat(l))}else i=0;let o;if(e.shape.length===2&&t.shape.length===2)s[0]===s[1]?o=fe(z(e,t),s[0]):o=fe(z(Ee(e,[1,0]),t),s[1]);else{let l=s[0]!==e.shape.length-1,u=s[1]===t.shape.length-1;o=Fe(e,t,l,u)}if(i>0){let l;a>r?l=a+r-3:l=a-1;let u=[];for(let p=l;p<l+i;++p)u.push(p);o=Ns(o,u)}return o.shape.length===1&&(o=Zt(o,1)),o})}var a0=class extends vo{constructor(e){super(e),this.axes=e.axes,this.normalize=e.normalize==null?!1:e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0],n=e[1];if(t.length>3||n.length>3)throw new Me("Dot layer does not support tensors of 4D or higher rank yet.");let a=this.interpretAxes(t,n);if(t[a[0]]!==n[a[1]])throw new V(`Dimension incompatibility: ${t[a[0]]} !== ${n[a[1]]}`)}mergeFunction(e){if(e.length!==2)throw new V(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t=e[0],n=e[1],a;return Array.isArray(this.axes)?a=this.axes.map((r,s)=>Ap(r,e[s].shape.length)):a=[Ap(this.axes,t.shape.length),Ap(this.axes,n.shape.length)],this.normalize&&(t=Gh(t,a[0]),n=Gh(n,a[1])),VU(t,n,a)}interpretAxes(e,t){let n;return Array.isArray(this.axes)?n=this.axes:n=[Ap(this.axes,e.length),Ap(this.axes,t.length)],n}computeOutputShape(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0].slice(),n=e[1].slice();if(t.length>3||n.length>3)throw new Me("Dot layer does not support tensors of 4D or higher rank yet.");let a=this.interpretAxes(t,n);t.splice(a[0],1),n.splice(a[1],1),n.splice(0,1);let r=t.concat(n);return r.length===1&&r.push(1),r}computeMask(e,t){return null}getConfig(){let e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};a0.className="Dot";ne.registerClass(a0);var r0=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return Zc(()=>Y(Ef(n.shape,0,this.stddev),n),()=>n,t.training||!1)})}};r0.className="GaussianNoise";ne.registerClass(r0);var s0=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return P(()=>{this.invokeCallHook(e,t);let n=Ne(e);return this.rate>0&&this.rate<1?Zc(()=>{let a=Math.sqrt(this.rate/(1-this.rate));return z(n,Ef(n.shape,1,a))},()=>n,t.training||!1):n})}};s0.className="GaussianDropout";ne.registerClass(s0);var i0=class extends Ge{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||Ne(e).shape}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return P(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(e);return Zc(()=>{let a=Ne(e),r=1.6732632423543772,s=1.0507009873554805,i=-r*s,o=Fr(Uu(n),this.rate);o=bo(o,"float32");let l=((1-this.rate)*(1+this.rate*i**2))**-.5,u=-l*i*this.rate,p=Y(z(a,o),z(Y(o,-1),i));return Y(z(p,l),u)},()=>Ne(e),t.training||!1)}return e})}};i0.className="AlphaDropout";ne.registerClass(i0);function ac(e,t,n,a,r,s=.001){let i;if(e.rank===2)i=sv(e,t,n,a,r,s);else if(e.rank===3)i=iv(e,t,n,a,r,s);else if(e.rank===4)i=ov(e,t,n,a,r,s);else throw new Me(`batchNormalization is not implemented for array of rank ${e.rank} yet`);return i}function UU(e,t,n,a,r=.001){return P(()=>{let s=Bc(e,a),i=s.mean,o=s.variance;return[ac(e,i,o,n,t,r),i,o]})}function GU(e,t,n,a,r=.001){return P(()=>{let s=Bc(e,a),i=s.mean,o=s.variance,l=[];for(let h of Ha(0,e.rank))a.indexOf(h)!==-1?l.push(1):l.push(e.shape[h]);let u=W(i,l),p=W(o,l),d=t==null?null:W(t,l),c=n==null?null:W(n,l);return[ac(e,u,p,c,d,r),i,o]})}function HU(e,t,n,a,r=.001){return v.arraysEqual(a.slice().sort(),Ha(0,e.rank-1))?UU(e,t,n,a,r):GU(e,t,n,a,r)}var o0=class extends Ge{constructor(e){e==null&&(e={}),super(e),this.supportsMasking=!0,this.axis=e.axis==null?-1:e.axis,this.momentum=e.momentum==null?.99:e.momentum,this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.movingMeanInitializer=St(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=St(e.movingVarianceInitializer||"ones"),this.betaConstraint=Kt(e.betaConstraint),this.gammaConstraint=Kt(e.gammaConstraint),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer)}build(e){e=tt(e);let t=this.axis>=0?this.axis:this.axis+e.length,n=e[t];if(n==null)throw new V(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new zt({ndim:e.length,axes:{[t]:n}})];let a=[n];this.scale&&(this.gamma=this.addWeight("gamma",a,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",a,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",a,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",a,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return P(()=>{let n=t.training==null?!1:t.training,a=Ne(e),r=a.shape,s=r.length,i=Ha(0,s),o=this.axis>=0?this.axis:this.axis+s;i.splice(o,1);let l=li(1,s);l[o]=r[o];let u=i.slice();u.sort();let p=!v.arraysEqual(u,Ha(0,s).slice(0,s-1)),d=()=>{if(p){let g=W(this.movingMean.read(),l),y=W(this.movingVariance.read(),l),b=this.center?W(this.beta.read(),l):null,x=this.scale?W(this.gamma.read(),l):null;return ac(a,g,y,b,x,this.epsilon)}else return ac(a,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!n)return d();let[c,h,m]=HU(a,this.gamma.read(),this.beta.read(),i,this.epsilon),f=(g,y,b)=>{P(()=>{let x=1-b,w=g.read(),I=z(pe(w,y),x);g.write(pe(w,I))})};return f(this.movingMean,h,this.momentum),f(this.movingVariance,m,this.momentum),c})}getConfig(){let e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Ct(this.betaInitializer),gammaInitializer:Ct(this.gammaInitializer),movingMeanInitializer:Ct(this.movingMeanInitializer),movingVarianceInitializer:Ct(this.movingVarianceInitializer),betaRegularizer:pt(this.betaRegularizer),gammaRegularizer:pt(this.gammaRegularizer),betaConstraint:qt(this.betaConstraint),gammaConstraint:qt(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};o0.className="BatchNormalization";ne.registerClass(o0);var l0=class extends Ge{constructor(e){if(e==null&&(e={}),super(e),this.axis=e.axis==null?-1:e.axis,typeof this.axis=="number"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let t of this.axis)if(!Number.isInteger(t))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer),this.supportsMasking=!0}build(e){e=tt(e);let t=e.length;typeof this.axis=="number"&&(this.axis=[this.axis]);for(let r=0;r<this.axis.length;++r)this.axis[r]<0&&(this.axis[r]+=t);for(let r of this.axis)if(r<0||r>=t)throw new Error(`Invalid axis: ${r}`);if(this.axis.length!==as(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(r=>e[r]),a=!0;this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,a):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,a):this.beta=null,this.built=!0}call(e,t){let n=Ne(e),a=n.shape,r=a.length;return P(()=>{let{mean:s,variance:i}=Bc(n,this.axis,!0),o=li(1,r);for(let h of this.axis)o[h]=a[h];let l=h=>h!=null&&h.shape.length!==r?W(h,o):h,u=this.scale?l(this.gamma.read()):null,p=this.center?l(this.beta.read()):null,d=[],c=[];for(let h=0;h<r;++h)this.axis.indexOf(h)!==-1?(d.push(a[h]),c.push(1)):(d.push(1),c.push(a[h]));return s=Ln(s,d),i=Ln(i,d),u!=null&&(u=Ln(u,c)),p!=null&&(p=Ln(p,c)),ac(n,s,i,p,u,this.epsilon)})}getConfig(){let e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Ct(this.betaInitializer),gammaInitializer:Ct(this.gammaInitializer),betaRegularizer:pt(this.betaRegularizer),gammaRegularizer:pt(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};l0.className="LayerNormalization";ne.registerClass(l0);function jU(e,t,n){return P(()=>{if(e.rank!==4)throw new V(`temporalPadding expects input tensor to be 4-D, but received a ${e.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new V("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(n==null&&(n=ja()),n!=="channelsLast"&&n!=="channelsFirst")throw new V(`Unknown data format: ${n}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let a;return n==="channelsFirst"?a=[[0,0],[0,0],t[0],t[1]]:a=[[0,0],t[0],t[1],[0,0]],ba(e,a)})}var u0=class extends Ge{constructor(e){if(e==null&&(e={}),super(e),this.dataFormat=e.dataFormat==null?ja():e.dataFormat,e.padding==null)this.padding=[[1,1],[1,1]];else if(typeof e.padding=="number")this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,e.padding.length!==2)throw new V(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,n;if(typeof e.padding[0]=="number")t=[e.padding[0],e.padding[0]],n=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,e.padding[0].length!==2)throw new V(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],e.padding[1].length!==2)throw new V(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);n=e.padding[1]}this.padding=[t,n]}this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=tt(e);let t,n;return this.dataFormat==="channelsFirst"?(e[2]!=null&&e[2]>=0?t=e[2]+this.padding[0][0]+this.padding[0][1]:t=null,e[3]!=null&&e[3]>=0?n=e[3]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],e[1],t,n]):(e[1]!=null&&e[1]>=0?t=e[1]+this.padding[0][0]+this.padding[0][1]:t=null,e[2]!=null&&e[2]>=0?n=e[2]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],t,n,e[3]])}call(e,t){return P(()=>jU(Ne(e),this.padding,this.dataFormat))}getConfig(){let e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};u0.className="ZeroPadding2D";ne.registerClass(u0);function jf(e,t,n,a,r,s){return P(()=>{Rt(r),pN(s),xa(a),n==null&&(n=[1,1]),a==null&&(a="valid"),r==null&&(r=ja()),s==null&&(s="max"),e=Aw(e,r);let i,o=a==="same"?"same":"valid";return s==="max"?i=Dt(e,t,n,o):i=ya(e,t,n,o),r==="channelsFirst"&&(i=Ee(i,[0,3,1,2])),i})}function s2(e,t,n,a,r,s){return P(()=>{Rt(r),pN(s),xa(a),n==null&&(n=[1,1,1]),a==null&&(a="valid"),r==null&&(r=ja()),s==null&&(s="max"),e=e2(e,r);let i,o=a==="same"?"same":"valid";return s==="max"?i=Fv(e,t,n,o):i=rv(e,t,n,o),r==="channelsFirst"&&(i=Ee(i,[0,4,1,2,3])),i})}var i2=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=2),super(e),typeof e.poolSize=="number")this.poolSize=[e.poolSize];else if(Array.isArray(e.poolSize)&&e.poolSize.length===1&&typeof e.poolSize[0]=="number")this.poolSize=e.poolSize;else throw new V(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.poolSize)}`);if(Qt(this.poolSize,"poolSize"),e.strides==null)this.strides=this.poolSize;else if(typeof e.strides=="number")this.strides=[e.strides];else if(Array.isArray(e.strides)&&e.strides.length===1&&typeof e.strides[0]=="number")this.strides=e.strides;else throw new V(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.strides)}`);Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,xa(this.padding),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){e=tt(e);let t=Ga(e[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return P(()=>{this.invokeCallHook(e,t),e=Xc(Ne(e),2);let n=this.poolingFunction(Ne(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return Ns(n,[2])})}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},p0=class extends i2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),jf(e,t,n,a,r,"max")}};p0.className="MaxPooling1D";ne.registerClass(p0);var c0=class extends i2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),jf(e,t,n,a,r,"avg")}};c0.className="AveragePooling1D";ne.registerClass(c0);var o2=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==2)throw new V(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),xa(this.padding),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2];return t=Ga(t,this.poolSize[0],this.padding,this.strides[0]),n=Ga(n,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n]:[e[0],t,n,e[3]]}call(e,t){return P(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ne(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},d0=class extends o2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),jf(e,t,n,a,r,"max")}};d0.className="MaxPooling2D";ne.registerClass(d0);var h0=class extends o2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),jf(e,t,n,a,r,"avg")}};h0.className="AveragePooling2D";ne.registerClass(h0);var l2=class extends Ge{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==3)throw new V(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),xa(this.padding),this.inputSpec=[new zt({ndim:5})]}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],a=this.dataFormat==="channelsFirst"?e[4]:e[3];return t=Ga(t,this.poolSize[0],this.padding,this.strides[0]),n=Ga(n,this.poolSize[1],this.padding,this.strides[1]),a=Ga(a,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n,a]:[e[0],t,n,a,e[4]]}call(e,t){return P(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ne(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},m0=class extends l2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),s2(e,t,n,a,r,"max")}};m0.className="MaxPooling3D";ne.registerClass(m0);var f0=class extends l2{constructor(e){super(e)}poolingFunction(e,t,n,a,r){return Rt(r),xa(a),s2(e,t,n,a,r,"avg")}};f0.className="AveragePooling3D";ne.registerClass(f0);var u2=class extends Ge{constructor(e){super(e),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new Me}},g0=class extends u2{constructor(e){super(e||{})}call(e,t){return P(()=>{let n=Ne(e);return Nt(n,1)})}};g0.className="GlobalAveragePooling1D";ne.registerClass(g0);var y0=class extends u2{constructor(e){super(e||{})}call(e,t){return P(()=>{let n=Ne(e);return ma(n,1)})}};y0.className="GlobalMaxPooling1D";ne.registerClass(y0);var p2=class extends Ge{constructor(e){super(e),this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Rt(this.dataFormat),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){return e=e,this.dataFormat==="channelsLast"?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new Me}getConfig(){let e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},b0=class extends p2{call(e,t){return P(()=>{let n=Ne(e);return this.dataFormat==="channelsLast"?Nt(n,[1,2]):Nt(n,[2,3])})}};b0.className="GlobalAveragePooling2D";ne.registerClass(b0);var x0=class extends p2{call(e,t){return P(()=>{let n=Ne(e);return this.dataFormat==="channelsLast"?ma(n,[1,2]):ma(n,[2,3])})}};x0.className="GlobalMaxPooling2D";ne.registerClass(x0);var c2=class extends Ge{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(e){this.layer!=null&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){let e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,n={}){let a=t.layer,r=Ua(a,n);delete t.layer;let s={layer:r};return Object.assign(s,t),new e(s)}},v0=class extends c2{constructor(e){super(e),this.supportsMasking=!0}build(e){if(e=tt(e),e.length<3)throw new V(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(e)}`);this.inputSpec=[{shape:e}];let t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){e=tt(e);let t=[e[0]].concat(e.slice(2)),n=this.layer.computeOutputShape(t),a=e[1];return[n[0],a].concat(n.slice(1))}call(e,t){return P(()=>(e=Ne(e),a2((n,a)=>[Ne(this.layer.call(n,t)),[]],e,[],!1,null,null,!1,!0)[1]))}};v0.className="TimeDistributed";ne.registerClass(v0);function qU(e){yo(X4,"BidirectionalMergeMode",e)}var KU="concat",w0=class extends c2{constructor(e){super(e);let t=e.layer.getConfig(),n={};n.className=e.layer.getClassName(),n.config=t,this.forwardLayer=Ua(n),t.goBackwards=t.goBackwards!==!0;let a={};if(a.className=e.layer.getClassName(),a.config=t,this.backwardLayer=Ua(a),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=e.mergeMode===void 0?KU:e.mergeMode,qU(this.mergeMode),e.weights)throw new Me("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,this.forwardLayer!=null&&(this.forwardLayer.trainable=e),this.backwardLayer!=null&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){let t=e.length,n=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,n)),this.backwardLayer.setWeights(e.slice(n))}computeOutputShape(e){let t=this.forwardLayer.computeOutputShape(e);Array.isArray(t)&&Array.isArray(t[0])||(t=[t]),t=t;let n,a,r;return this.returnState&&(r=t.slice(1)),n=t[0],n=n,this.mergeMode==="concat"?(n[n.length-1]*=2,a=[n]):this.mergeMode==null?a=[n,n.slice()]:a=[n],this.returnState?this.mergeMode==null?a.concat(r).concat(r.slice()):[n].concat(r).concat(r.slice()):On(a)}apply(e,t){let n=t==null?null:t.initialState,a=t==null?null:t.constants;t==null&&(t={});let r=n2(e,n,a,this.numConstants);if(e=r.inputs,n=r.initialState,a=r.constants,Array.isArray(e)&&(n=e.slice(1),e=e[0]),(n==null||n.length===0)&&a==null)return super.apply(e,t);let s=[],i=[];if(n!=null){let l=n.length;if(l%2>0)throw new V("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=n,s.push(...n);let u=n.map(p=>new zt({shape:p.shape}));this.forwardLayer.stateSpec=u.slice(0,l/2),this.backwardLayer.stateSpec=u.slice(l/2),i.push(...u)}if(a!=null)throw new Me("Support for constants in Bidirectional layers is not implemented yet.");let o=s[0]instanceof Ba;for(let l of s)if(l instanceof Ba!==o)throw new V("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(o){let l=[e].concat(s),u=this.inputSpec.concat(i),p=this.inputSpec;this.inputSpec=u;let d=super.apply(l,t);return this.inputSpec=p,d}else return super.apply(e,t)}call(e,t){return P(()=>{let n=t.initialState,a,r;if(n==null)a=this.forwardLayer.call(e,t),r=this.backwardLayer.call(e,t);else{let o=n.slice(0,n.length/2),l=n.slice(n.length/2);a=this.forwardLayer.call(e,Object.assign(t,{initialState:o})),r=this.backwardLayer.call(e,Object.assign(t,{initialState:l}))}let s;this.returnState&&(Array.isArray(a)&&(s=a.slice(1).concat(r.slice(1))),a=a[0],r=r[0]),this.returnSequences&&(r=ga(r,1));let i;return this.mergeMode==="concat"?i=nw([a,r]):this.mergeMode==="sum"?i=Y(a,r):this.mergeMode==="ave"?i=z(.5,Y(a,r)):this.mergeMode==="mul"?i=z(a,r):this.mergeMode==null&&(i=[a,r]),this.returnState?this.mergeMode==null?i.concat(s):[i].concat(s):i})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Ys(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Ys(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){Array.isArray(t)&&(t=t[0]);let n;if(this.returnSequences?this.mergeMode==null?n=[t,t]:n=t:this.mergeMode==null?n=[null,null]:n=null,this.returnState){let a=this.forwardLayer.states.map(r=>null);return Array.isArray(n)?n.concat(a).concat(a):[n].concat(a).concat(a)}else return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(e),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){let e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){let n=Ua(t.layer);if(delete t.layer,t.numConstants!=null)throw new Me("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let a=t;return a.layer=n,new e(a)}};w0.className="Bidirectional";ne.registerClass(w0);var k0=class extends Ge{constructor(e){super(e),this.scale=e.scale,e.offset?this.offset=e.offset:this.offset=0}getConfig(){let e={scale:this.scale,offset:this.offset},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return P(()=>(e=Ne(e),e.dtype!=="float32"&&(e=bo(e,"float32")),Y(z(e,this.scale),this.offset)))}};k0.className="Rescaling";ne.registerClass(k0);var XU=["bilinear","nearest"],qk=new Set(XU),I0=class extends Ge{constructor(e){if(super(e),this.height=e.height,this.width=e.width,e.interpolation)if(qk.has(e.interpolation))this.interpolation=e.interpolation;else throw new V(`Invalid interpolation parameter: ${e.interpolation} is not implemented`);else this.interpolation="bilinear";this.cropToAspectRatio=Boolean(e.cropToAspectRatio)}computeOutputShape(e){e=tt(e);let t=e[2];return[this.height,this.width,t]}getConfig(){let e={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return P(()=>{let n=[this.height,this.width];if(this.interpolation==="bilinear")return za.resizeBilinear(e,n,!this.cropToAspectRatio);if(this.interpolation==="nearest")return za.resizeNearestNeighbor(e,n,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...qk]} are supported`)})}};I0.className="Resizing";ne.registerClass(I0);function YU(e,t,n,a){let r=Ne(e);if(r.dtype!=="int32"&&(r=bo(r,"int32")),t==="int")return r;let s=r.shape;if(r.rank===0&&(r=Zt(r,-1)),t==="oneHot"&&r.shape[r.shape.length-1]!==1&&(r=Zt(r,-1)),r.rank>2)throw new V(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${s} which would result in output rank ${r.rank}.`);let i=["multiHot","oneHot"].includes(t),o=r,l;if(typeof a!="undefined"&&t==="count"?l=zh(o,a,n,i):l=zh(o,[],n,i),t!=="tfIdf")return l;if(a)return z(l,a);throw new V("When outputMode is 'tfIdf', weights must be provided.")}var S0=class extends Ge{constructor(e){super(e),this.numTokens=e.numTokens,e.outputMode?this.outputMode=e.outputMode:this.outputMode="multiHot"}getConfig(){let e={numTokens:this.numTokens,outputMode:this.outputMode},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){return e=tt(e),e==null?[this.numTokens]:this.outputMode==="oneHot"&&e[e.length-1]!==1?(e.push(this.numTokens),e):(e[e.length-1]=this.numTokens,e)}call(e,t){return P(()=>{e=Ne(e),e.dtype!=="int32"&&(e=bo(e,"int32"));let n;if(typeof t.countWeights!="undefined"){if(this.outputMode!=="count")throw new V(`countWeights is not used when outputMode !== count.
-              Received countWeights=${t.countWeights}`);n=Ne(t.countWeights)}let a=ma(e),r=bl(e),s=Cn(this.numTokens,a).bufferSync().get(0),i=Fr(r,0).bufferSync().get(0);if(!(s&&i))throw new V(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return YU(e,this.outputMode,this.numTokens,n)})}};S0.className="CategoryEncoding";ne.registerClass(S0);function ZU(e){return new Hu(e)}function JU(e){return new Cw(e)}function QU(e){return new Sw(e)}function eG(e){return new Tw(e)}function tG(e){return new Nw(e)}function nG(e){return new Ew(e)}function aG(e){return new _w(e)}function rG(e){return new Wf(e)}function sG(e){return new td(e)}function iG(e){return new Fw(e)}function oG(e){return new nd(e)}function lG(e){return new Dw(e)}function uG(e){return new Rw(e)}function pG(e){return new Mw(e)}function cG(e){return new Pw(e)}function dG(e){return new Ow(e)}function hG(e){return new Hw(e)}function mG(e){return new Uw(e)}function fG(e){return new Hf(e)}function gG(e){return new Vw(e)}function yG(e){return new Gw(e)}function bG(e){return new jw(e)}function xG(e){return new qw(e)}function vG(e){return new Kw(e)}function wG(e){return new Yw(e)}function kG(e){return new Zw(e)}function IG(e){return new Qw(e)}function SG(e){return new n0(e)}function TG(e){return new e0(e)}function NG(e){return new t0(e)}function CG(e){return new Jw(e)}function _G(e){return new a0(e)}function EG(e){return new o0(e)}function AG(e){return new l0(e)}function $G(e){return new u0(e)}function T0(e){return new c0(e)}function FG(e){return T0(e)}function DG(e){return T0(e)}function N0(e){return new h0(e)}function RG(e){return N0(e)}function MG(e){return N0(e)}function C0(e){return new f0(e)}function PG(e){return C0(e)}function OG(e){return C0(e)}function LG(e){return new g0(e)}function zG(e){return new b0(e)}function d2(e){return new y0(e)}function h2(e){return new x0(e)}function m2(e){return new p0(e)}function f2(e){return new d0(e)}function WG(e){return new m0(e)}function BG(e){return new zw(e)}function VG(e){return new Vf(e)}function UG(e){return new Ww(e)}function GG(e){return new rd(e)}function HG(e){return new Lw(e)}function jG(e){return new Bf(e)}function qG(e){return new Bw(e)}function KG(e){return new Gf(e)}function XG(e){return new mr(e)}function YG(e){return new Uf(e)}function ZG(e){return new w0(e)}function JG(e){return new v0(e)}var QG=d2,eH=h2,tH=m2,nH=f2;function aH(e){return new r0(e)}function rH(e){return new s0(e)}function sH(e){return new i0(e)}function iH(e){return new Xw(e)}function oH(e){return new k0(e)}function lH(e){return new I0(e)}function uH(e){return new S0(e)}var g2={};Ae(g2,{MAPE:()=>vH,MSE:()=>IH,binaryAccuracy:()=>pH,binaryCrossentropy:()=>cH,categoricalAccuracy:()=>hH,categoricalCrossentropy:()=>mH,cosineProximity:()=>yH,mape:()=>wH,meanAbsoluteError:()=>bH,meanAbsolutePercentageError:()=>xH,meanSquaredError:()=>kH,mse:()=>SH,precision:()=>fH,recall:()=>gH,sparseCategoricalAccuracy:()=>dH});function pH(e,t){return yw(e,t)}function cH(e,t){return AN(e,t)}function dH(e,t){return $N(e,t)}function hH(e,t){return bw(e,t)}function mH(e,t){return xw(e,t)}function fH(e,t){return EN(e,t)}function gH(e,t){return QV(e,t)}function yH(e,t){return gw(e,t)}function bH(e,t){return Lf(e,t)}function xH(e,t){return ju(e,t)}function vH(e,t){return ju(e,t)}function wH(e,t){return ju(e,t)}function kH(e,t){return xo(e,t)}function IH(e,t){return xo(e,t)}function SH(e,t){return xo(e,t)}var y2={};Ae(y2,{modelFromJSON:()=>EU});var b2={};Ae(b2,{l1:()=>NH,l1l2:()=>TH,l2:()=>CH});function TH(e){return new Qc(e)}function NH(e){return PU(e)}function CH(e){return OU(e)}var x2=class extends Sl{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Er))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function ch(e,t){return e<t}function Kk(e,t){return e>t}var v2=class extends x2{constructor(e){if(super(),e==null&&(e={}),e.restoreBestWeights)throw new Me("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,["auto","min","max"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),this.mode==="min"?this.monitorFunc=ch:this.mode==="max"?this.monitorFunc=Kk:this.monitor.indexOf("acc")!==-1?this.monitorFunc=Kk:this.monitorFunc=ch,this.monitorFunc===ch&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===ch?1/0:-1/0}async onEpochEnd(e,t){await Zr(t);let n=this.getMonitorValue(t);n!=null&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){e==null&&(e={});let t=e[this.monitor];return t==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};function _H(e){return new v2(e)}var EH={earlyStopping:_H},AH=H();AH.registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")});var Na;(function(e){e[e.DT_INVALID=0]="DT_INVALID",e[e.DT_FLOAT=1]="DT_FLOAT",e[e.DT_DOUBLE=2]="DT_DOUBLE",e[e.DT_INT32=3]="DT_INT32",e[e.DT_UINT8=4]="DT_UINT8",e[e.DT_INT16=5]="DT_INT16",e[e.DT_INT8=6]="DT_INT8",e[e.DT_STRING=7]="DT_STRING",e[e.DT_COMPLEX64=8]="DT_COMPLEX64",e[e.DT_INT64=9]="DT_INT64",e[e.DT_BOOL=10]="DT_BOOL",e[e.DT_QINT8=11]="DT_QINT8",e[e.DT_QUINT8=12]="DT_QUINT8",e[e.DT_QINT32=13]="DT_QINT32",e[e.DT_BFLOAT16=14]="DT_BFLOAT16",e[e.DT_QINT16=15]="DT_QINT16",e[e.DT_QUINT16=16]="DT_QUINT16",e[e.DT_UINT16=17]="DT_UINT16",e[e.DT_COMPLEX128=18]="DT_COMPLEX128",e[e.DT_HALF=19]="DT_HALF",e[e.DT_RESOURCE=20]="DT_RESOURCE",e[e.DT_VARIANT=21]="DT_VARIANT",e[e.DT_UINT32=22]="DT_UINT32",e[e.DT_UINT64=23]="DT_UINT64",e[e.DT_FLOAT_REF=101]="DT_FLOAT_REF",e[e.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",e[e.DT_INT32_REF=103]="DT_INT32_REF",e[e.DT_UINT8_REF=104]="DT_UINT8_REF",e[e.DT_INT16_REF=105]="DT_INT16_REF",e[e.DT_INT8_REF=106]="DT_INT8_REF",e[e.DT_STRING_REF=107]="DT_STRING_REF",e[e.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",e[e.DT_INT64_REF=109]="DT_INT64_REF",e[e.DT_BOOL_REF=110]="DT_BOOL_REF",e[e.DT_QINT8_REF=111]="DT_QINT8_REF",e[e.DT_QUINT8_REF=112]="DT_QUINT8_REF",e[e.DT_QINT32_REF=113]="DT_QINT32_REF",e[e.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",e[e.DT_QINT16_REF=115]="DT_QINT16_REF",e[e.DT_QUINT16_REF=116]="DT_QUINT16_REF",e[e.DT_UINT16_REF=117]="DT_UINT16_REF",e[e.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",e[e.DT_HALF_REF=119]="DT_HALF_REF",e[e.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",e[e.DT_VARIANT_REF=121]="DT_VARIANT_REF",e[e.DT_UINT32_REF=122]="DT_UINT32_REF",e[e.DT_UINT64_REF=123]="DT_UINT64_REF"})(Na||(Na={}));var Xk;(function(e){let t;(function(n){n[n.LEGACY=0]="LEGACY",n[n.V1=1]="V1",n[n.V2=2]="V2"})(t=e.CheckpointFormatVersion||(e.CheckpointFormatVersion={}))})(Xk||(Xk={}));var _0={};function $H(e,t){let n={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};_0[e]=n}function w2(e){return _0[e]}function FH(e){delete _0[e]}function k(e,t,n,a,r){let s=t.inputParams[e];if(s&&s.inputIndexStart!==void 0){let o=s.inputIndexStart,l=s.inputIndexEnd===0?void 0:s.inputIndexEnd===void 0?o+1:s.inputIndexEnd;if(s.type==="tensor")return wn(t.inputNames[s.inputIndexStart],n,a,r);if(s.type==="tensors")return t.inputNames.slice(o,l).map(d=>wn(d,n,a,r));let u=wn(t.inputNames.slice(o)[0],n,a,r),p=u.dataSync();return s.type==="number"?p[0]:v.toNestedArray(u.shape,p)}let i=t.attrParams[e];return i&&i.value}function wn(e,t,n,a){let[r,s]=Yn(e);if(a!=null){let o=a.getHashTableHandleByName(r);if(o!=null)return o}let i=n.currentContextIds.find(o=>!!t[Xh(r,o)]);return i!==void 0?t[Xh(r,i)][s]:void 0}function DH(e,t,n){return t[Xh(e,n.currentContextId)]}function sr(e,t){let[n,a,r]=Yn(e);return[Xh(n,t&&t.currentContextId),a,r]}function Xh(e,t){return t?`${e}-${t}`:e}function Yn(e){let t=e.split(":");if(t.length===1)return[e,0,void 0];let n=t[0],a=t.length===3?t[1]:void 0,r=Number(t[t.length-1]);return[n,r,a]}function vh(e,t,n){let a=k("pad",e,t,n);if(a==="explicit"){a=k("explicitPaddings",e,t,n);let r=[[0,0],[0,0],[0,0],[0,0]];for(let s=0;s<4;s++)r[s][0]=a[s*2],r[s][1]=a[s*2+1];return r}return a}function Tr(e){return e.kept?e:ir(e)}var k2={};Ae(k2,{json:()=>RH});var RH=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],I2={};Ae(I2,{json:()=>MH});var MH=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],S2={};Ae(S2,{json:()=>PH});var PH=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}],T2={};Ae(T2,{json:()=>OH});var OH=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],N2={};Ae(N2,{json:()=>LH});var LH=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomStandardNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],C2={};Ae(C2,{json:()=>zH});var zH=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],_2={};Ae(_2,{json:()=>WH});var WH=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],E2={};Ae(E2,{json:()=>BH});var BH=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],A2={};Ae(A2,{json:()=>VH});var VH=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}],$2={};Ae($2,{json:()=>UH});var UH=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}],F2={};Ae(F2,{json:()=>GH});var GH=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],D2={};Ae(D2,{json:()=>HH});var HH=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}],R2={};Ae(R2,{json:()=>jH});var jH=[{tfOpName:"EuclideanNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",defaultValue:!1}]},{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}],M2={};Ae(M2,{json:()=>qH});var qH=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],P2={};Ae(P2,{json:()=>KH});var KH=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}],O2={};Ae(O2,{json:()=>XH});var XH=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],L2={};Ae(L2,{json:()=>YH});var YH=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],z2={};Ae(z2,{json:()=>ZH});var ZH=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],W2={};Ae(W2,{json:()=>JH});var JH=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],Yk=class{constructor(){let e=[k2,I2,S2,T2,N2,C2,_2,E2,A2,$2,F2,D2,R2,M2,P2,O2,L2,z2,W2],t=[].concat(...e.map(n=>n.json));this.opMappers=t.reduce((n,a)=>(n[a.tfOpName]=a,n),{})}static get Instance(){return this._instance||(this._instance=new this)}transformGraph(e,t={}){let n=e.node,a=[],r=[],s=[],i=n.reduce((m,f)=>(m[f.name]=this.mapNode(f),f.op.startsWith("Placeholder")?a.push(m[f.name]):f.op==="Const"?r.push(m[f.name]):(f.input==null||f.input.length===0)&&s.push(m[f.name]),m),{}),o=[],l=[],u={},p={};t!=null&&(u=this.mapSignatureEntries(t.inputs),p=this.mapSignatureEntries(t.outputs));let d=Object.keys(i);d.forEach(m=>{let f=i[m];f.inputNames.forEach((g,y)=>{let[b,,x]=sr(g),w=i[b];if(w.outputs!=null){let I=w.outputs.indexOf(x);if(I!==-1){let T=`${b}:${I}`;f.inputNames[y]=T}}f.inputs.push(w),w.children.push(f)})}),Object.keys(p).length===0?d.forEach(m=>{let f=i[m];f.children.length===0&&l.push(f)}):Object.keys(p).forEach(m=>{let[f]=sr(m),g=i[f];g!=null&&(g.signatureKey=p[m],l.push(g))}),Object.keys(u).length>0?Object.keys(u).forEach(m=>{let[f]=sr(m),g=i[f];g&&(g.signatureKey=u[m],o.push(g))}):o=a;let c={};e.library!=null&&e.library.function!=null&&(c=e.library.function.reduce((m,f)=>(m[f.signature.name]=this.mapFunction(f),m),{}));let h={nodes:i,inputs:o,outputs:l,weights:r,placeholders:a,signature:t,functions:c};return s.length>0&&(h.initNodes=s),h}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,n)=>(t[e[n].name]=n,t),{})}mapNode(e){let t=w2(e.op)||this.opMappers[e.op]||{};e.attr==null&&(e.attr={});let n={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(a=>a.startsWith("^")?a.slice(1):a),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return t.inputs!=null&&(n.inputParams=t.inputs.reduce((a,r)=>(a[r.name]={type:r.type,inputIndexStart:r.start,inputIndexEnd:r.end},a),{})),t.attrs!=null&&(n.attrParams=t.attrs.reduce((a,r)=>{let s=r.type,i;switch(r.type){case"string":i=Qb(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Qb(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"string[]":i=ix(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=ix(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"number":i=tx(e.attr,r.tfName,r.defaultValue||0),i===void 0&&!!r.tfDeprecatedName&&(i=tx(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"number[]":i=sx(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=sx(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"bool":i=ex(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=ex(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"bool[]":i=lx(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=lx(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"shape":i=rx(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=rx(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"shape[]":i=ox(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=ox(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"dtype":i=nx(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=nx(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"dtype[]":i=ax(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=ax(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"func":i=Zk(e.attr,r.tfName,r.defaultValue),i===void 0&&!!r.tfDeprecatedName&&(i=Zk(e.attr,r.tfDeprecatedName,r.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${r.type} for op: ${e.op}`)}return a[r.name]={value:i,type:s},a},{})),n}mapFunction(e){let t=e.nodeDef,n=[],a=[],r={};t!=null&&(r=t.reduce((u,p)=>(u[p.name]=this.mapNode(p),p.op==="Const"&&a.push(u[p.name]),u),{}));let s=[],i=[];e.signature.inputArg.forEach(u=>{let[p]=sr(u.name),d={name:p,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:E0(u.type),type:"dtype"}},children:[]};d.signatureKey=u.name,s.push(d),r[p]=d}),Object.keys(r).forEach(u=>{let p=r[u];p.inputNames.forEach((d,c)=>{let[h,,m]=sr(d),f=r[h];if(f.outputs!=null){let g=f.outputs.indexOf(m);if(g!==-1){let y=`${h}:${g}`;p.inputNames[c]=y}}p.inputs.push(f),f.children.push(p)})});let o=e.ret;e.signature.outputArg.forEach(u=>{let[p,d]=sr(o[u.name]),c=r[p];c!=null&&(c.defaultOutput=d,i.push(c))});let l=this.mapArgsToSignature(e);return{nodes:r,inputs:s,outputs:i,weights:a,placeholders:n,signature:l}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n),t),{}),outputs:e.signature.outputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n,e.ret),t),{})}}mapArgToTensorInfo(e,t){let n=e.name;return t!=null&&(n=t[n]),{name:n,dtype:e.type}}};function QH(e){let t=H().global;if(typeof t.atob!="undefined")return t.atob(e);if(typeof Buffer!="undefined")return new Buffer(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}function B2(e,t){let n=Array.isArray(e)?String.fromCharCode.apply(null,e):QH(e);return t?n:n.toLowerCase()}function Qb(e,t,n,a=!1){let r=e[t];return r!=null?B2(r.s,a):n}function ex(e,t,n){let a=e[t];return a?a.b:n}function tx(e,t,n){let a=e[t]||{},r=a.i!=null?a.i:a.f!=null?a.f:n;return typeof r=="number"?r:parseInt(r,10)}function E0(e){switch(typeof e=="string"&&(e=Na[e]),e){case Na.DT_FLOAT:case Na.DT_HALF:return"float32";case Na.DT_INT32:case Na.DT_INT64:case Na.DT_INT8:case Na.DT_UINT8:return"int32";case Na.DT_BOOL:return"bool";case Na.DT_DOUBLE:return"float32";case Na.DT_STRING:return"string";default:return null}}function Zk(e,t,n){let a=e[t];return a&&a.func?a.func.name:n}function nx(e,t,n){let a=e[t];return a&&a.type?E0(a.type):n}function ax(e,t,n){let a=e[t];return a&&a.list&&a.list.type?a.list.type.map(r=>E0(r)):n}function V2(e){if(!e.unknownRank)return e.dim!=null?e.dim.map(t=>typeof t.size=="number"?t.size:parseInt(t.size,10)):[]}function rx(e,t,n){let a=e[t];return a&&a.shape?V2(a.shape):n}function sx(e,t,n){let a=e[t];return a?((a.list.f&&a.list.f.length?a.list.f:a.list.i)||[]).map(r=>typeof r=="number"?r:parseInt(r,10)):n}function ix(e,t,n,a=!1){let r=e[t];return r&&r.list&&r.list.s?r.list.s.map(s=>B2(s,a)):n}function ox(e,t,n){let a=e[t];return a&&a.list&&a.list.shape?a.list.shape.map(r=>V2(r)):n}function lx(e,t,n){let a=e[t];return a&&a.list&&a.list.b?a.list.b:n}var e6=class{constructor(e,t,n){this.node=e,this.tensorMap=t,this.context=n,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(a=>this.getInput(a)),e.rawAttrs!=null&&(this.attrs=Object.keys(e.rawAttrs).reduce((a,r)=>(a[r]=this.getAttr(r),a),{}))}getInput(e){return wn(e,this.tensorMap,this.context)}getAttr(e,t){let n=this.node.rawAttrs[e];if(n.tensor!=null)return wn(e,this.tensorMap,this.context);if(n.i!=null||n.f!=null)return tx(this.node.rawAttrs,e,t);if(n.s!=null)return Qb(this.node.rawAttrs,e,t);if(n.b!=null)return ex(this.node.rawAttrs,e,t);if(n.shape!=null)return rx(this.node.rawAttrs,e,t);if(n.type!=null)return nx(this.node.rawAttrs,e,t);if(n.list!=null){if(n.list.i!=null||n.list.f!=null)return sx(this.node.rawAttrs,e,t);if(n.list.s!=null)return ix(this.node.rawAttrs,e,t);if(n.list.shape!=null)return ox(this.node.rawAttrs,e,t);if(n.list.b!=null)return lx(this.node.rawAttrs,e,t);if(n.list.type!=null)return ax(this.node.rawAttrs,e,t)}return t}},pn={};Ae(pn,{OP_SCOPE_SUFFIX:()=>Mx,abs:()=>Lt,acos:()=>Xx,acosh:()=>Yx,add:()=>Y,addN:()=>KS,all:()=>jm,any:()=>Qp,argMax:()=>ri,argMin:()=>Zx,asin:()=>Jx,asinh:()=>Qx,atan:()=>ev,atan2:()=>tv,atanh:()=>nv,avgPool:()=>ya,avgPool3d:()=>rv,basicLSTMCell:()=>JS,batchNorm:()=>ks,batchNorm2d:()=>sv,batchNorm3d:()=>iv,batchNorm4d:()=>ov,batchToSpaceND:()=>Pc,bincount:()=>lv,booleanMaskAsync:()=>RT,broadcastArgs:()=>QS,broadcastTo:()=>Xs,buffer:()=>Oe,cast:()=>oe,ceil:()=>uv,clipByValue:()=>en,clone:()=>ir,complex:()=>Ar,concat:()=>Ze,concat1d:()=>pv,concat2d:()=>cv,concat3d:()=>dv,concat4d:()=>hv,conv1d:()=>qm,conv2d:()=>$t,conv2dTranspose:()=>Km,conv3d:()=>fv,conv3dTranspose:()=>gv,cos:()=>Oc,cosh:()=>Xm,cosineWindow:()=>yf,cumprod:()=>ec,cumsum:()=>Ym,denseBincount:()=>zh,depthToSpace:()=>yv,depthwiseConv2d:()=>Is,diag:()=>tT,dilation2d:()=>bv,div:()=>he,divNoNan:()=>xv,dot:()=>vv,dropout:()=>qv,einsum:()=>nT,elu:()=>Lu,enclosingPowerOfTwo:()=>Kv,equal:()=>ea,erf:()=>wv,euclideanNorm:()=>Sv,exp:()=>fn,expandDims:()=>Zt,expm1:()=>Tv,eye:()=>Zm,fft:()=>Hc,fill:()=>gn,floor:()=>Wu,floorDiv:()=>Hm,fused:()=>kl,gather:()=>Bu,gatherND:()=>LT,greater:()=>Cn,greaterEqual:()=>Fr,ifft:()=>wl,imag:()=>Dc,image:()=>za,inTopKAsync:()=>zT,irfft:()=>hf,isFinite:()=>Nv,isInf:()=>Cv,isNaN:()=>_v,leakyRelu:()=>Lc,less:()=>Jm,lessEqual:()=>Ss,linalg:()=>Zv,linspace:()=>oT,localResponseNormalization:()=>Ev,log:()=>ta,log1p:()=>zc,logSigmoid:()=>Av,logSoftmax:()=>ef,logSumExp:()=>tf,logicalAnd:()=>$a,logicalNot:()=>Wc,logicalOr:()=>nf,logicalXor:()=>$v,losses:()=>ZT,lowerBound:()=>uT,matMul:()=>Fe,max:()=>ma,maxPool:()=>Dt,maxPool3d:()=>Fv,maxPoolWithArgmax:()=>pT,maximum:()=>dr,mean:()=>Nt,meshgrid:()=>cT,min:()=>bl,minimum:()=>Vu,mirrorPad:()=>Dv,mod:()=>Rv,moments:()=>Bc,movingAverage:()=>MT,mul:()=>z,multiRNNCell:()=>dT,multinomial:()=>hT,neg:()=>vt,norm:()=>zu,notEqual:()=>oi,oneHot:()=>gl,ones:()=>Jn,onesLike:()=>na,op:()=>L,outerProduct:()=>mT,pad:()=>ba,pad1d:()=>fT,pad2d:()=>gT,pad3d:()=>yT,pad4d:()=>bT,pool:()=>Mv,pow:()=>$r,prelu:()=>Uc,print:()=>zx,prod:()=>Pv,raggedGather:()=>xT,raggedRange:()=>vT,raggedTensorToTensor:()=>wT,rand:()=>kT,randomGamma:()=>IT,randomNormal:()=>rf,randomStandardNormal:()=>ST,randomUniform:()=>Uu,range:()=>xl,real:()=>yl,reciprocal:()=>zv,relu:()=>Xe,relu6:()=>sf,reshape:()=>W,reverse:()=>ga,reverse1d:()=>TT,reverse2d:()=>NT,reverse3d:()=>CT,reverse4d:()=>_T,rfft:()=>jc,round:()=>of,rsqrt:()=>lf,scalar:()=>be,scatterND:()=>PT,searchSorted:()=>af,selu:()=>uf,separableConv2d:()=>Ts,setdiff1dAsync:()=>ET,sigmoid:()=>ha,sign:()=>Wv,signal:()=>YT,sin:()=>pf,sinh:()=>cf,slice:()=>Be,slice1d:()=>Gc,slice2d:()=>df,slice3d:()=>fo,slice4d:()=>vl,softmax:()=>Ka,softplus:()=>mo,spaceToBatchND:()=>Vc,sparse:()=>JT,sparseToDense:()=>OT,spectral:()=>XT,split:()=>zn,sqrt:()=>un,square:()=>lt,squaredDifference:()=>mf,squeeze:()=>Ns,stack:()=>Ft,step:()=>go,stridedSlice:()=>Bv,string:()=>QT,sub:()=>pe,sum:()=>fe,tan:()=>Vv,tanh:()=>si,tensor:()=>In,tensor1d:()=>Ke,tensor2d:()=>Aa,tensor3d:()=>Rc,tensor4d:()=>Da,tensor5d:()=>AT,tensor6d:()=>$T,tile:()=>Ln,topk:()=>Uv,transpose:()=>Ee,truncatedNormal:()=>ff,unique:()=>Gv,unsortedSegmentSum:()=>gf,unstack:()=>ct,upperBound:()=>FT,variable:()=>Hv,where:()=>mn,whereAsync:()=>jv,zeros:()=>It,zerosLike:()=>qe});var t6=(e,t,n,a=pn)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[a.add(k("a",e,t,n),k("b",e,t,n))];case"AddN":return[a.addN(k("tensors",e,t,n))];case"FloorMod":case"Mod":return[a.mod(k("a",e,t,n),k("b",e,t,n))];case"Mul":return[a.mul(k("a",e,t,n),k("b",e,t,n))];case"RealDiv":case"Div":return[a.div(k("a",e,t,n),k("b",e,t,n))];case"DivNoNan":return[a.divNoNan(k("a",e,t,n),k("b",e,t,n))];case"FloorDiv":return[a.floorDiv(k("a",e,t,n),k("b",e,t,n))];case"Sub":return[a.sub(k("a",e,t,n),k("b",e,t,n))];case"Minimum":return[a.minimum(k("a",e,t,n),k("b",e,t,n))];case"Maximum":return[a.maximum(k("a",e,t,n),k("b",e,t,n))];case"Pow":return[a.pow(k("a",e,t,n),k("b",e,t,n))];case"SquaredDifference":return[a.squaredDifference(k("a",e,t,n),k("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},n6=(e,t,n,a=pn)=>{switch(e.op){case"Abs":case"ComplexAbs":return[a.abs(k("x",e,t,n))];case"Acos":return[a.acos(k("x",e,t,n))];case"Acosh":return[a.acosh(k("x",e,t,n))];case"Asin":return[a.asin(k("x",e,t,n))];case"Asinh":return[a.asinh(k("x",e,t,n))];case"Atan":return[a.atan(k("x",e,t,n))];case"Atan2":return[a.atan2(k("x",e,t,n),k("y",e,t,n))];case"Atanh":return[a.atanh(k("x",e,t,n))];case"Ceil":return[a.ceil(k("x",e,t,n))];case"Complex":return[a.complex(k("real",e,t,n),k("imag",e,t,n))];case"Cos":return[a.cos(k("x",e,t,n))];case"Cosh":return[a.cosh(k("x",e,t,n))];case"Elu":return[a.elu(k("x",e,t,n))];case"Erf":return[a.erf(k("x",e,t,n))];case"Exp":return[a.exp(k("x",e,t,n))];case"Expm1":return[a.expm1(k("x",e,t,n))];case"Floor":return[a.floor(k("x",e,t,n))];case"Log":return[a.log(k("x",e,t,n))];case"Log1p":return[a.log1p(k("x",e,t,n))];case"Imag":return[a.imag(k("x",e,t,n))];case"Neg":return[a.neg(k("x",e,t,n))];case"Reciprocal":return[a.reciprocal(k("x",e,t,n))];case"Real":return[a.real(k("x",e,t,n))];case"Relu":return[a.relu(k("x",e,t,n))];case"Round":return[a.round(k("x",e,t,n))];case"Selu":return[a.selu(k("x",e,t,n))];case"Sigmoid":return[a.sigmoid(k("x",e,t,n))];case"Sin":return[a.sin(k("x",e,t,n))];case"Sign":return[a.sign(k("x",e,t,n))];case"Sinh":return[a.sinh(k("x",e,t,n))];case"Softplus":return[a.softplus(k("x",e,t,n))];case"Sqrt":return[a.sqrt(k("x",e,t,n))];case"Square":return[a.square(k("x",e,t,n))];case"Tanh":return[a.tanh(k("x",e,t,n))];case"Tan":return[a.tan(k("x",e,t,n))];case"ClipByValue":return[a.clipByValue(k("x",e,t,n),k("clipValueMin",e,t,n),k("clipValueMax",e,t,n))];case"Relu6":return[a.relu6(k("x",e,t,n))];case"Rsqrt":return[a.rsqrt(wn(e.inputNames[0],t,n))];case"Prod":return[a.prod(k("x",e,t,n),k("axes",e,t,n))];case"LeakyRelu":return[a.leakyRelu(k("x",e,t,n),k("alpha",e,t,n))];case"Prelu":return[a.prelu(k("x",e,t,n),k("alpha",e,t,n))];case"IsNan":return[a.isNaN(wn(e.inputNames[0],t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Ea(e,t,n=""){if(!(typeof e=="number"||typeof t=="number")){v.assert(e.length===t.length,()=>n+` Shapes ${e} and ${t} must match`);for(let a=0;a<e.length;a++){let r=e[a],s=t[a];v.assert(r<0||s<0||r===s,()=>n+` Shapes ${e} and ${t} must match`)}}}function Jk(e){return!(typeof e=="number"||e.some(t=>t<0))}function $p(e,t,n){let a=ux(e,n),r=!Jk(a);if(r&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${a}`);if(r&&t.forEach(s=>{a=ux(s.shape,a)}),!Jk(a))throw new Error(`Non-fully-defined elementShape: ${a}`);return a}function ux(e,t){if(typeof e=="number")return t;if(typeof t=="number")return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);let n=[];for(let a=0;a<e.length;++a){let r=e[a],s=t[a];if(r>=0&&s>=0&&r!==s)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);n[a]=r>=0?r:s}return n}var a6=class{constructor(e,t,n,a,r,s,i){this.name=e,this.dtype=t,this.maxSize=n,this.elementShape=a,this.identicalElementShapes=r,this.dynamicSize=s,this.clearAfterRead=i,this.tensors=[],this.closed_=!1,this.idTensor=be(0),Jt(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.tensor.id))&&t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);let t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(t=>this.read(t))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},
-          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=t.shape),Ea(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);n.tensor=t,Jt(t),n.written=!0,this.tensors[e]=n}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((n,a)=>this.write(n,t[a]))}gather(e,t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let a=0;a<this.size();a++)e.push(a)}if(e.length===0)return In([],[0].concat(this.elementShape));let n=this.readMany(e);return Ea(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),Ft(n,0)}concat(e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(this.size()===0)return In([],[0].concat(this.elementShape));let t=[];for(let a=0;a<this.size();a++)t.push(a);let n=this.readMany(t);return Ea(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),Ze(n,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);let n=Math.max(...e);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(e,ct(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let n=0,a=e.map(o=>(n+=o,n));if(n!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to
+"use strict";var faceapi=(()=>{var py=Object.defineProperty;var pD=Object.getOwnPropertyDescriptor;var hD=Object.getOwnPropertyNames;var fD=Object.prototype.hasOwnProperty;var Ys=(e=>typeof require!="undefined"?require:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof require!="undefined"?require:t)[n]}):e)(function(e){if(typeof require!="undefined")return require.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')});var sh=(e,t)=>{for(var n in t)py(e,n,{get:t[n],enumerable:!0})},mD=(e,t,n,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let s of hD(t))!fD.call(e,s)&&s!==n&&py(e,s,{get:()=>t[s],enumerable:!(r=pD(t,s))||r.enumerable});return e};var gD=e=>mD(py({},"__esModule",{value:!0}),e);var Wde={};sh(Wde,{AgeGenderNet:()=>wp,BoundingBox:()=>_i,Box:()=>ot,ComposableTask:()=>xr,ComputeAllFaceDescriptorsTask:()=>Gs,ComputeFaceDescriptorsTaskBase:()=>Tp,ComputeSingleFaceDescriptorTask:()=>Hs,DetectAllFaceLandmarksTask:()=>Np,DetectAllFacesTask:()=>vl,DetectFaceLandmarksTaskBase:()=>Cp,DetectFacesTaskBase:()=>Ep,DetectSingleFaceLandmarksTask:()=>_p,DetectSingleFaceTask:()=>Ap,Dimensions:()=>yn,FACE_EXPRESSION_LABELS:()=>FI,FaceDetection:()=>gt,FaceDetectionNet:()=>BI,FaceExpressionNet:()=>xp,FaceExpressions:()=>Ws,FaceLandmark68Net:()=>Mi,FaceLandmark68TinyNet:()=>Ip,FaceLandmarkNet:()=>MI,FaceLandmarks:()=>ar,FaceLandmarks5:()=>vI,FaceLandmarks68:()=>Ai,FaceMatch:()=>al,FaceMatcher:()=>$p,FaceRecognitionNet:()=>Li,Gender:()=>bg,LabeledBox:()=>ol,LabeledFaceDescriptors:()=>bs,NetInput:()=>vs,NeuralNetwork:()=>an,ObjectDetection:()=>Ls,Point:()=>Fe,PredictedBox:()=>xI,Rect:()=>Ei,SsdMobilenetv1:()=>$a,SsdMobilenetv1Options:()=>vr,TinyFaceDetector:()=>Ui,TinyFaceDetectorOptions:()=>Sp,TinyYolov2:()=>Wi,TinyYolov2Options:()=>es,allFaces:()=>Lde,allFacesSsdMobilenetv1:()=>$$,allFacesTinyYolov2:()=>Mde,awaitMediaLoaded:()=>NI,bufferToImage:()=>_I,computeFaceDescriptor:()=>kde,createCanvas:()=>Ri,createCanvasFromMedia:()=>gp,createFaceDetectionNet:()=>gde,createFaceRecognitionNet:()=>ade,createSsdMobilenetv1:()=>g$,createTinyFaceDetector:()=>zde,createTinyYolov2:()=>vde,detectAllFaces:()=>Cg,detectFaceLandmarks:()=>E$,detectFaceLandmarksTiny:()=>Ide,detectLandmarks:()=>Pde,detectSingleFace:()=>Ode,draw:()=>PI,env:()=>Qe,euclideanDistance:()=>UI,extendWithAge:()=>wg,extendWithFaceDescriptor:()=>xg,extendWithFaceDetection:()=>$i,extendWithFaceExpressions:()=>hg,extendWithFaceLandmarks:()=>ml,extendWithGender:()=>Ig,extractFaceTensors:()=>ul,extractFaces:()=>cl,fetchImage:()=>Hle,fetchJson:()=>$I,fetchNetWeights:()=>qle,fetchOrThrow:()=>Bs,fetchVideo:()=>jle,getContext2dOrThrow:()=>Gn,getMediaDimensions:()=>Fi,imageTensorToCanvas:()=>EI,imageToSquare:()=>AI,inverseSigmoid:()=>Lle,iou:()=>mI,isMediaElement:()=>ag,isMediaLoaded:()=>mp,isWithAge:()=>ode,isWithFaceDetection:()=>ys,isWithFaceExpressions:()=>RI,isWithFaceLandmarks:()=>Oi,isWithGender:()=>ide,loadAgeGenderModel:()=>Dde,loadFaceDetectionModel:()=>Fde,loadFaceExpressionModel:()=>$de,loadFaceLandmarkModel:()=>_de,loadFaceLandmarkTinyModel:()=>Ede,loadFaceRecognitionModel:()=>Ade,loadSsdMobilenetv1Model:()=>A$,loadTinyFaceDetectorModel:()=>Cde,loadTinyYolov2Model:()=>Nde,loadWeightMap:()=>DI,locateFaces:()=>Rde,matchDimensions:()=>Kle,minBbox:()=>gI,nets:()=>et,nonMaxSuppression:()=>bI,normalize:()=>Jr,padToSquare:()=>yI,predictAgeAndGender:()=>Tde,recognizeFaceExpressions:()=>Sde,resizeResults:()=>D$,resolveInput:()=>Di,shuffleArray:()=>Mle,sigmoid:()=>pp,ssdMobilenetv1:()=>_$,tf:()=>Oe,tinyFaceDetector:()=>xde,tinyYolov2:()=>wde,toNetInput:()=>bt,utils:()=>fI,validateConfig:()=>WI,version:()=>Bde});var Oe={};sh(Oe,{Abs:()=>Fc,Acos:()=>Rc,Acosh:()=>Pc,AdadeltaOptimizer:()=>km,AdagradOptimizer:()=>Sm,AdamOptimizer:()=>Tm,AdamaxOptimizer:()=>Cm,Add:()=>ba,AddN:()=>bo,All:()=>Oc,Any:()=>Mc,ArgMax:()=>yo,ArgMin:()=>dd,Asin:()=>Lc,Asinh:()=>zc,Atan:()=>Bc,Atan2:()=>Vc,Atanh:()=>Wc,AvgPool:()=>vo,AvgPool3D:()=>pd,AvgPool3DGrad:()=>hf,AvgPoolGrad:()=>pf,BackendWasm:()=>HA,BatchMatMul:()=>xo,BatchToSpaceND:()=>Uc,Bincount:()=>ff,BroadcastArgs:()=>mf,BroadcastTo:()=>rS,Callback:()=>wN,CallbackList:()=>TC,Cast:()=>wo,Ceil:()=>Io,ClipByValue:()=>ya,Complex:()=>gf,ComplexAbs:()=>hd,Concat:()=>Gc,Conv2D:()=>ko,Conv2DBackpropFilter:()=>bf,Conv2DBackpropInput:()=>So,Conv3D:()=>fd,Conv3DBackpropFilterV2:()=>yf,Conv3DBackpropInputV2:()=>vf,Cos:()=>To,Cosh:()=>Co,CropAndResize:()=>qc,Cumprod:()=>Hc,Cumsum:()=>No,CustomCallback:()=>NC,DataStorage:()=>uf,DenseBincount:()=>xf,DepthToSpace:()=>jc,DepthwiseConv2dNative:()=>_o,DepthwiseConv2dNativeBackpropFilter:()=>wf,DepthwiseConv2dNativeBackpropInput:()=>If,Diag:()=>kf,Dilation2D:()=>md,Dilation2DBackpropFilter:()=>Dh,Dilation2DBackpropInput:()=>$h,ENV:()=>$v,EarlyStopping:()=>IN,Einsum:()=>Sf,Elu:()=>Ao,EluGrad:()=>Tf,Environment:()=>tS,Equal:()=>Xc,Erf:()=>Kc,Exp:()=>$o,ExpandDims:()=>Yc,Expm1:()=>Zc,FFT:()=>Cf,Fill:()=>gd,FlipLeftRight:()=>Jc,Floor:()=>Do,FloorDiv:()=>Fo,FromPixels:()=>Fh,FusedBatchNorm:()=>Ro,FusedConv2D:()=>to,FusedDepthwiseConv2D:()=>no,GPGPUContext:()=>Ch,GatherNd:()=>eu,GatherV2:()=>Qc,GraphModel:()=>A0,Greater:()=>tu,GreaterEqual:()=>Po,History:()=>CC,IFFT:()=>Nf,Identity:()=>Oo,Imag:()=>_f,InputSpec:()=>zt,IsFinite:()=>nu,IsInf:()=>ru,IsNan:()=>su,KernelBackend:()=>ld,LRN:()=>bd,LRNGrad:()=>Af,LayerVariable:()=>yC,LayersModel:()=>Es,LeakyRelu:()=>Mo,Less:()=>au,LessEqual:()=>ou,LinSpace:()=>Ef,Log:()=>Lo,Log1p:()=>iu,LogSoftmax:()=>aS,LogicalAnd:()=>cu,LogicalNot:()=>uu,LogicalOr:()=>lu,LogicalXor:()=>sS,LowerBound:()=>iF,MathBackendWebGL:()=>Ym,Max:()=>zo,MaxPool:()=>Wo,MaxPool3D:()=>yd,MaxPool3DGrad:()=>Df,MaxPoolGrad:()=>$f,MaxPoolWithArgmax:()=>Ff,Maximum:()=>Bo,Mean:()=>Vo,Min:()=>Uo,Minimum:()=>Go,MirrorPad:()=>Ho,Mod:()=>du,MomentumOptimizer:()=>Nm,Multinomial:()=>Rf,Multiply:()=>qo,Neg:()=>pu,NonMaxSuppressionV3:()=>fu,NonMaxSuppressionV4:()=>mu,NonMaxSuppressionV5:()=>gu,NotEqual:()=>hu,OP_SCOPE_SUFFIX:()=>Pv,OneHot:()=>jo,OnesLike:()=>bu,Optimizer:()=>Rs,OptimizerConstructors:()=>Zs,Pack:()=>yu,PadV2:()=>Ko,Pool:()=>cF,Pow:()=>Xo,Prelu:()=>Yo,Prod:()=>Zo,RMSPropOptimizer:()=>_m,RNN:()=>ms,RaggedGather:()=>Pf,RaggedRange:()=>Of,RaggedTensorToTensor:()=>Mf,Range:()=>vd,Rank:()=>Ey,Real:()=>Lf,RealDiv:()=>Eo,Reciprocal:()=>vu,Reduction:()=>kn,Relu:()=>Jo,Relu6:()=>ti,Reshape:()=>xu,ResizeBilinear:()=>ei,ResizeBilinearGrad:()=>Bf,ResizeNearestNeighbor:()=>Qo,ResizeNearestNeighborGrad:()=>zf,Reverse:()=>ni,RotateWithOffset:()=>Ou,Round:()=>ri,Rsqrt:()=>si,SGDOptimizer:()=>jd,ScatterNd:()=>wu,SearchSorted:()=>Wf,Select:()=>Iu,Selu:()=>ku,Sequential:()=>Tc,Sigmoid:()=>oi,Sign:()=>Cu,Sin:()=>ai,Sinh:()=>Tu,Slice:()=>Su,Softmax:()=>ui,Softplus:()=>Nu,SpaceToBatchND:()=>_u,SparseFillEmptyRows:()=>xd,SparseReshape:()=>Au,SparseSegmentMean:()=>wd,SparseSegmentSum:()=>Id,SparseToDense:()=>Vf,SplitV:()=>Eu,Sqrt:()=>ii,Square:()=>kd,SquaredDifference:()=>li,Step:()=>xa,StridedSlice:()=>$u,StringNGrams:()=>Sd,StringSplit:()=>Td,StringToHashBucketFast:()=>Cd,Sub:()=>di,Sum:()=>ci,SymbolicTensor:()=>Vr,Tan:()=>pi,Tanh:()=>hi,Tensor:()=>Te,TensorBuffer:()=>Ht,Tile:()=>va,TopK:()=>Du,Transform:()=>Fu,Transpose:()=>_s,Unique:()=>Uf,Unpack:()=>Ru,UnsortedSegmentSum:()=>Nd,UpperBound:()=>uF,Variable:()=>ia,ZerosLike:()=>Pu,_FusedMatMul:()=>eo,abs:()=>Lt,acos:()=>Xv,acosh:()=>Yv,add:()=>Y,addN:()=>YS,all:()=>Kf,any:()=>Ql,argMax:()=>ao,argMin:()=>Zv,asin:()=>Jv,asinh:()=>Qv,atan:()=>ex,atan2:()=>tx,atanh:()=>nx,avgPool:()=>mr,avgPool3d:()=>sx,backend:()=>DS,backend_util:()=>N,basicLSTMCell:()=>eT,batchNorm:()=>ka,batchNorm2d:()=>ax,batchNorm3d:()=>ox,batchNorm4d:()=>ix,batchToSpaceND:()=>Od,bincount:()=>cx,booleanMaskAsync:()=>OT,broadcastArgs:()=>tT,broadcastTo:()=>Ya,broadcast_util:()=>Mu,browser:()=>fi,buffer:()=>Me,callbacks:()=>DH,cast:()=>ce,ceil:()=>ux,clipByValue:()=>en,clone:()=>is,complex:()=>As,concat:()=>Je,concat1d:()=>lx,concat2d:()=>dx,concat3d:()=>px,concat4d:()=>hx,constraints:()=>wC,conv1d:()=>Xf,conv2d:()=>Dt,conv2dTranspose:()=>Yf,conv3d:()=>mx,conv3dTranspose:()=>gx,copyRegisteredKernels:()=>hF,cos:()=>Md,cosh:()=>Zf,cosineWindow:()=>ym,cumprod:()=>ed,cumsum:()=>Jf,customGrad:()=>ls,data:()=>qN,denseBincount:()=>Bh,deprecationWarn:()=>Vv,depthToSpace:()=>bx,depthwiseConv2d:()=>Sa,deregisterOp:()=>PH,device_util:()=>$d,diag:()=>rT,dilation2d:()=>yx,disableDeprecationWarnings:()=>WR,dispose:()=>_e,disposeVariables:()=>VR,div:()=>fe,divNoNan:()=>vx,dot:()=>xx,dropout:()=>jx,einsum:()=>sT,elu:()=>Lu,enableDebugMode:()=>BR,enableProdMode:()=>zR,enclosingPowerOfTwo:()=>Kx,engine:()=>Er,env:()=>q,equal:()=>tr,erf:()=>wx,euclideanNorm:()=>Sx,exp:()=>gn,expandDims:()=>Zt,expm1:()=>Tx,eye:()=>Qf,fft:()=>Hd,fill:()=>bn,findBackend:()=>XR,findBackendFactory:()=>YR,floor:()=>Bu,floorDiv:()=>jf,forceHalfFloat:()=>dE,fused:()=>Ic,gather:()=>Wu,gatherND:()=>BT,gather_util:()=>Uv,getBackend:()=>jR,getGradient:()=>Ny,getKernel:()=>Rh,getKernelsForBackend:()=>Ph,getThreadsCount:()=>_le,gpgpu_util:()=>U2,grad:()=>xM,grads:()=>wM,greater:()=>An,greaterEqual:()=>Ds,ifft:()=>wc,imag:()=>Fd,image:()=>Br,inTopKAsync:()=>WT,initializers:()=>IC,input:()=>WC,io:()=>Ut,irfft:()=>fm,isFinite:()=>Cx,isInf:()=>Nx,isNaN:()=>_x,keep:()=>Jt,kernel_impls:()=>fs,layers:()=>kC,leakyRelu:()=>Ld,less:()=>em,lessEqual:()=>Ta,linalg:()=>Zx,linspace:()=>uT,loadGraphModel:()=>L6,loadGraphModelSync:()=>z6,loadLayersModel:()=>FU,localResponseNormalization:()=>Ex,log:()=>nr,log1p:()=>zd,logSigmoid:()=>Ax,logSoftmax:()=>nm,logSumExp:()=>rm,logicalAnd:()=>Dr,logicalNot:()=>Bd,logicalOr:()=>sm,logicalXor:()=>$x,losses:()=>QT,lowerBound:()=>dT,matMul:()=>De,math:()=>$S,max:()=>pr,maxPool:()=>Rt,maxPool3d:()=>Dx,maxPoolWithArgmax:()=>pT,maximum:()=>hs,mean:()=>Ct,memory:()=>Lh,meshgrid:()=>hT,metrics:()=>yN,min:()=>yc,minimum:()=>Vu,mirrorPad:()=>Fx,mod:()=>Rx,model:()=>OU,models:()=>vN,moments:()=>Wd,movingAverage:()=>MT,mul:()=>B,multiRNNCell:()=>fT,multinomial:()=>mT,neg:()=>xt,nextFrame:()=>Jx,norm:()=>zu,notEqual:()=>co,oneHot:()=>gc,ones:()=>Qn,onesLike:()=>rr,op:()=>z,outerProduct:()=>gT,pad:()=>gr,pad1d:()=>bT,pad2d:()=>yT,pad3d:()=>vT,pad4d:()=>xT,pool:()=>Px,pow:()=>$s,prelu:()=>Ud,print:()=>zv,prod:()=>Ox,profile:()=>UR,raggedGather:()=>wT,raggedRange:()=>IT,raggedTensorToTensor:()=>kT,rand:()=>ST,randomGamma:()=>TT,randomNormal:()=>om,randomStandardNormal:()=>CT,randomUniform:()=>Uu,range:()=>vc,ready:()=>qR,real:()=>bc,reciprocal:()=>zx,registerBackend:()=>qf,registerCallbackConstructor:()=>LU,registerGradient:()=>oS,registerKernel:()=>_d,registerOp:()=>RH,regularizers:()=>xN,relu:()=>Xe,relu6:()=>im,removeBackend:()=>KR,reshape:()=>W,reverse:()=>fr,reverse1d:()=>NT,reverse2d:()=>_T,reverse3d:()=>ET,reverse4d:()=>AT,rfft:()=>qd,round:()=>cm,rsqrt:()=>um,scalar:()=>xe,scatterND:()=>LT,scatter_util:()=>Gv,searchSorted:()=>am,selu:()=>lm,separableConv2d:()=>Ca,sequential:()=>MU,serialization:()=>se,setBackend:()=>HR,setPlatform:()=>ZR,setThreadsCount:()=>Nle,setWasmPath:()=>Tle,setWasmPaths:()=>Cle,setWebGLContext:()=>h2,setdiff1dAsync:()=>$T,sigmoid:()=>dr,sign:()=>Bx,signal:()=>JT,sin:()=>dm,sinh:()=>pm,slice:()=>We,slice1d:()=>Gd,slice2d:()=>hm,slice3d:()=>gi,slice4d:()=>xc,slice_util:()=>qt,softmax:()=>Xr,softplus:()=>mi,spaceToBatchND:()=>Vd,sparse:()=>eC,sparseToDense:()=>zT,spectral:()=>ZT,split:()=>zn,sqrt:()=>un,square:()=>it,squaredDifference:()=>mm,squeeze:()=>Na,stack:()=>Ft,step:()=>bi,stridedSlice:()=>Wx,string:()=>tC,sub:()=>de,sum:()=>ge,sumOutType:()=>Hf,tan:()=>Vx,tanh:()=>oo,tensor:()=>Cn,tensor1d:()=>Ke,tensor2d:()=>$r,tensor3d:()=>Rd,tensor4d:()=>Rr,tensor5d:()=>DT,tensor6d:()=>FT,tensor_util:()=>Ur,test_util:()=>jS,tidy:()=>O,tile:()=>Ln,time:()=>GR,topk:()=>Ux,train:()=>Wa,transpose:()=>Ee,truncatedNormal:()=>gm,unique:()=>Gx,unregisterGradient:()=>pF,unregisterKernel:()=>dF,unsortedSegmentSum:()=>bm,unstack:()=>lt,upcastType:()=>hr,upperBound:()=>RT,util:()=>w,valueAndGrad:()=>IM,valueAndGrads:()=>kM,variable:()=>Hx,variableGrads:()=>lT,version:()=>Ple,version_converter:()=>W6,version_core:()=>EP,version_layers:()=>xw,version_wasm:()=>Ele,version_webgl:()=>yZ,webgl:()=>vZ,webgl_util:()=>p2,where:()=>mn,whereAsync:()=>qx,zeros:()=>kt,zerosLike:()=>qe});var bD=Object.create,Cv=Object.defineProperty,yD=Object.getOwnPropertyDescriptor,V1=Object.getOwnPropertyNames,vD=Object.getPrototypeOf,xD=Object.prototype.hasOwnProperty,U1=(e=>typeof Ys!="undefined"?Ys:typeof Proxy!="undefined"?new Proxy(e,{get:(t,n)=>(typeof Ys!="undefined"?Ys:t)[n]}):e)(function(e){if(typeof Ys!="undefined")return Ys.apply(this,arguments);throw new Error('Dynamic require of "'+e+'" is not supported')}),Wt=(e,t)=>function(){return t||(0,e[V1(e)[0]])((t={exports:{}}).exports,t),t.exports},Ae=(e,t)=>{for(var n in t)Cv(e,n,{get:t[n],enumerable:!0})},wD=(e,t,n,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let s of V1(t))!xD.call(e,s)&&s!==n&&Cv(e,s,{get:()=>t[s],enumerable:!(r=yD(t,s))||r.enumerable});return e},ma=(e,t,n)=>(n=e!=null?bD(vD(e)):{},wD(t||!e||!e.__esModule?Cv(n,"default",{value:e,enumerable:!0}):n,e)),ID=Wt({"node_modules/long/src/long.js"(e,t){t.exports=r;var n=null;try{n=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(R){}function r(R,T,L){this.low=R|0,this.high=T|0,this.unsigned=!!L}r.prototype.__isLong__,Object.defineProperty(r.prototype,"__isLong__",{value:!0});function s(R){return(R&&R.__isLong__)===!0}r.isLong=s;var a={},o={};function i(R,T){var L,V,G;return T?(R>>>=0,(G=0<=R&&R<256)&&(V=o[R],V)?V:(L=u(R,(R|0)<0?-1:0,!0),G&&(o[R]=L),L)):(R|=0,(G=-128<=R&&R<128)&&(V=a[R],V)?V:(L=u(R,R<0?-1:0,!1),G&&(a[R]=L),L))}r.fromInt=i;function c(R,T){if(isNaN(R))return T?x:v;if(T){if(R<0)return x;if(R>=g)return $}else{if(R<=-b)return F;if(R+1>=b)return E}return R<0?c(-R,T).neg():u(R%m|0,R/m|0,T)}r.fromNumber=c;function u(R,T,L){return new r(R,T,L)}r.fromBits=u;var l=Math.pow;function p(R,T,L){if(R.length===0)throw Error("empty string");if(R==="NaN"||R==="Infinity"||R==="+Infinity"||R==="-Infinity")return v;if(typeof T=="number"?(L=T,T=!1):T=!!T,L=L||10,L<2||36<L)throw RangeError("radix");var V;if((V=R.indexOf("-"))>0)throw Error("interior hyphen");if(V===0)return p(R.substring(1),T,L).neg();for(var G=c(l(L,8)),j=v,H=0;H<R.length;H+=8){var Z=Math.min(8,R.length-H),J=parseInt(R.substring(H,H+Z),L);if(Z<8){var ee=c(l(L,Z));j=j.mul(ee).add(c(J))}else j=j.mul(G),j=j.add(c(J))}return j.unsigned=T,j}r.fromString=p;function d(R,T){return typeof R=="number"?c(R,T):typeof R=="string"?p(R,T):u(R.low,R.high,typeof T=="boolean"?T:R.unsigned)}r.fromValue=d;var h=1<<16,f=1<<24,m=h*h,g=m*m,b=g/2,y=i(f),v=i(0);r.ZERO=v;var x=i(0,!0);r.UZERO=x;var k=i(1);r.ONE=k;var S=i(1,!0);r.UONE=S;var C=i(-1);r.NEG_ONE=C;var E=u(-1,2147483647,!1);r.MAX_VALUE=E;var $=u(-1,-1,!0);r.MAX_UNSIGNED_VALUE=$;var F=u(0,-2147483648,!1);r.MIN_VALUE=F;var A=r.prototype;A.toInt=function(){return this.unsigned?this.low>>>0:this.low},A.toNumber=function(){return this.unsigned?(this.high>>>0)*m+(this.low>>>0):this.high*m+(this.low>>>0)},A.toString=function(T){if(T=T||10,T<2||36<T)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative())if(this.eq(F)){var L=c(T),V=this.div(L),G=V.mul(L).sub(this);return V.toString(T)+G.toInt().toString(T)}else return"-"+this.neg().toString(T);for(var j=c(l(T,6),this.unsigned),H=this,Z="";;){var J=H.div(j),ee=H.sub(J.mul(j)).toInt()>>>0,re=ee.toString(T);if(H=J,H.isZero())return re+Z;for(;re.length<6;)re="0"+re;Z=""+re+Z}},A.getHighBits=function(){return this.high},A.getHighBitsUnsigned=function(){return this.high>>>0},A.getLowBits=function(){return this.low},A.getLowBitsUnsigned=function(){return this.low>>>0},A.getNumBitsAbs=function(){if(this.isNegative())return this.eq(F)?64:this.neg().getNumBitsAbs();for(var T=this.high!=0?this.high:this.low,L=31;L>0&&(T&1<<L)==0;L--);return this.high!=0?L+33:L+1},A.isZero=function(){return this.high===0&&this.low===0},A.eqz=A.isZero,A.isNegative=function(){return!this.unsigned&&this.high<0},A.isPositive=function(){return this.unsigned||this.high>=0},A.isOdd=function(){return(this.low&1)===1},A.isEven=function(){return(this.low&1)===0},A.equals=function(T){return s(T)||(T=d(T)),this.unsigned!==T.unsigned&&this.high>>>31===1&&T.high>>>31===1?!1:this.high===T.high&&this.low===T.low},A.eq=A.equals,A.notEquals=function(T){return!this.eq(T)},A.neq=A.notEquals,A.ne=A.notEquals,A.lessThan=function(T){return this.comp(T)<0},A.lt=A.lessThan,A.lessThanOrEqual=function(T){return this.comp(T)<=0},A.lte=A.lessThanOrEqual,A.le=A.lessThanOrEqual,A.greaterThan=function(T){return this.comp(T)>0},A.gt=A.greaterThan,A.greaterThanOrEqual=function(T){return this.comp(T)>=0},A.gte=A.greaterThanOrEqual,A.ge=A.greaterThanOrEqual,A.compare=function(T){if(s(T)||(T=d(T)),this.eq(T))return 0;var L=this.isNegative(),V=T.isNegative();return L&&!V?-1:!L&&V?1:this.unsigned?T.high>>>0>this.high>>>0||T.high===this.high&&T.low>>>0>this.low>>>0?-1:1:this.sub(T).isNegative()?-1:1},A.comp=A.compare,A.negate=function(){return!this.unsigned&&this.eq(F)?F:this.not().add(k)},A.neg=A.negate,A.add=function(T){s(T)||(T=d(T));var L=this.high>>>16,V=this.high&65535,G=this.low>>>16,j=this.low&65535,H=T.high>>>16,Z=T.high&65535,J=T.low>>>16,ee=T.low&65535,re=0,te=0,ie=0,ne=0;return ne+=j+ee,ie+=ne>>>16,ne&=65535,ie+=G+J,te+=ie>>>16,ie&=65535,te+=V+Z,re+=te>>>16,te&=65535,re+=L+H,re&=65535,u(ie<<16|ne,re<<16|te,this.unsigned)},A.subtract=function(T){return s(T)||(T=d(T)),this.add(T.neg())},A.sub=A.subtract,A.multiply=function(T){if(this.isZero())return v;if(s(T)||(T=d(T)),n){var L=n.mul(this.low,this.high,T.low,T.high);return u(L,n.get_high(),this.unsigned)}if(T.isZero())return v;if(this.eq(F))return T.isOdd()?F:v;if(T.eq(F))return this.isOdd()?F:v;if(this.isNegative())return T.isNegative()?this.neg().mul(T.neg()):this.neg().mul(T).neg();if(T.isNegative())return this.mul(T.neg()).neg();if(this.lt(y)&&T.lt(y))return c(this.toNumber()*T.toNumber(),this.unsigned);var V=this.high>>>16,G=this.high&65535,j=this.low>>>16,H=this.low&65535,Z=T.high>>>16,J=T.high&65535,ee=T.low>>>16,re=T.low&65535,te=0,ie=0,ne=0,le=0;return le+=H*re,ne+=le>>>16,le&=65535,ne+=j*re,ie+=ne>>>16,ne&=65535,ne+=H*ee,ie+=ne>>>16,ne&=65535,ie+=G*re,te+=ie>>>16,ie&=65535,ie+=j*ee,te+=ie>>>16,ie&=65535,ie+=H*J,te+=ie>>>16,ie&=65535,te+=V*re+G*ee+j*J+H*Z,te&=65535,u(ne<<16|le,te<<16|ie,this.unsigned)},A.mul=A.multiply,A.divide=function(T){if(s(T)||(T=d(T)),T.isZero())throw Error("division by zero");if(n){if(!this.unsigned&&this.high===-2147483648&&T.low===-1&&T.high===-1)return this;var L=(this.unsigned?n.div_u:n.div_s)(this.low,this.high,T.low,T.high);return u(L,n.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?x:v;var V,G,j;if(this.unsigned){if(T.unsigned||(T=T.toUnsigned()),T.gt(this))return x;if(T.gt(this.shru(1)))return S;j=x}else{if(this.eq(F)){if(T.eq(k)||T.eq(C))return F;if(T.eq(F))return k;var H=this.shr(1);return V=H.div(T).shl(1),V.eq(v)?T.isNegative()?k:C:(G=this.sub(T.mul(V)),j=V.add(G.div(T)),j)}else if(T.eq(F))return this.unsigned?x:v;if(this.isNegative())return T.isNegative()?this.neg().div(T.neg()):this.neg().div(T).neg();if(T.isNegative())return this.div(T.neg()).neg();j=v}for(G=this;G.gte(T);){V=Math.max(1,Math.floor(G.toNumber()/T.toNumber()));for(var Z=Math.ceil(Math.log(V)/Math.LN2),J=Z<=48?1:l(2,Z-48),ee=c(V),re=ee.mul(T);re.isNegative()||re.gt(G);)V-=J,ee=c(V,this.unsigned),re=ee.mul(T);ee.isZero()&&(ee=k),j=j.add(ee),G=G.sub(re)}return j},A.div=A.divide,A.modulo=function(T){if(s(T)||(T=d(T)),n){var L=(this.unsigned?n.rem_u:n.rem_s)(this.low,this.high,T.low,T.high);return u(L,n.get_high(),this.unsigned)}return this.sub(this.div(T).mul(T))},A.mod=A.modulo,A.rem=A.modulo,A.not=function(){return u(~this.low,~this.high,this.unsigned)},A.and=function(T){return s(T)||(T=d(T)),u(this.low&T.low,this.high&T.high,this.unsigned)},A.or=function(T){return s(T)||(T=d(T)),u(this.low|T.low,this.high|T.high,this.unsigned)},A.xor=function(T){return s(T)||(T=d(T)),u(this.low^T.low,this.high^T.high,this.unsigned)},A.shiftLeft=function(T){return s(T)&&(T=T.toInt()),(T&=63)===0?this:T<32?u(this.low<<T,this.high<<T|this.low>>>32-T,this.unsigned):u(0,this.low<<T-32,this.unsigned)},A.shl=A.shiftLeft,A.shiftRight=function(T){return s(T)&&(T=T.toInt()),(T&=63)===0?this:T<32?u(this.low>>>T|this.high<<32-T,this.high>>T,this.unsigned):u(this.high>>T-32,this.high>=0?0:-1,this.unsigned)},A.shr=A.shiftRight,A.shiftRightUnsigned=function(T){if(s(T)&&(T=T.toInt()),T&=63,T===0)return this;var L=this.high;if(T<32){var V=this.low;return u(V>>>T|L<<32-T,L>>>T,this.unsigned)}else return T===32?u(L,0,this.unsigned):u(L>>>T-32,0,this.unsigned)},A.shru=A.shiftRightUnsigned,A.shr_u=A.shiftRightUnsigned,A.toSigned=function(){return this.unsigned?u(this.low,this.high,!1):this},A.toUnsigned=function(){return this.unsigned?this:u(this.low,this.high,!0)},A.toBytes=function(T){return T?this.toBytesLE():this.toBytesBE()},A.toBytesLE=function(){var T=this.high,L=this.low;return[L&255,L>>>8&255,L>>>16&255,L>>>24,T&255,T>>>8&255,T>>>16&255,T>>>24]},A.toBytesBE=function(){var T=this.high,L=this.low;return[T>>>24,T>>>16&255,T>>>8&255,T&255,L>>>24,L>>>16&255,L>>>8&255,L&255]},r.fromBytes=function(T,L,V){return V?r.fromBytesLE(T,L):r.fromBytesBE(T,L)},r.fromBytesLE=function(T,L){return new r(T[0]|T[1]<<8|T[2]<<16|T[3]<<24,T[4]|T[5]<<8|T[6]<<16|T[7]<<24,L)},r.fromBytesBE=function(T,L){return new r(T[4]<<24|T[5]<<16|T[6]<<8|T[7],T[0]<<24|T[1]<<16|T[2]<<8|T[3],L)}}}),kD=Wt({"(disabled):node_modules/node-fetch/browser.js"(){}}),SD=Wt({"(disabled):util"(){}}),TD=Wt({"node_modules/seedrandom/lib/alea.js"(e,t){(function(n,r,s){function a(u){var l=this,p=c();l.next=function(){var d=2091639*l.s0+l.c*23283064365386963e-26;return l.s0=l.s1,l.s1=l.s2,l.s2=d-(l.c=d|0)},l.c=1,l.s0=p(" "),l.s1=p(" "),l.s2=p(" "),l.s0-=p(u),l.s0<0&&(l.s0+=1),l.s1-=p(u),l.s1<0&&(l.s1+=1),l.s2-=p(u),l.s2<0&&(l.s2+=1),p=null}function o(u,l){return l.c=u.c,l.s0=u.s0,l.s1=u.s1,l.s2=u.s2,l}function i(u,l){var p=new a(u),d=l&&l.state,h=p.next;return h.int32=function(){return p.next()*4294967296|0},h.double=function(){return h()+(h()*2097152|0)*11102230246251565e-32},h.quick=h,d&&(typeof d=="object"&&o(d,p),h.state=function(){return o(p,{})}),h}function c(){var u=4022871197,l=function(p){p=String(p);for(var d=0;d<p.length;d++){u+=p.charCodeAt(d);var h=.02519603282416938*u;u=h>>>0,h-=u,h*=u,u=h>>>0,h-=u,u+=h*4294967296}return(u>>>0)*23283064365386963e-26};return l}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.alea=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),CD=Wt({"node_modules/seedrandom/lib/xor128.js"(e,t){(function(n,r,s){function a(c){var u=this,l="";u.x=0,u.y=0,u.z=0,u.w=0,u.next=function(){var d=u.x^u.x<<11;return u.x=u.y,u.y=u.z,u.z=u.w,u.w^=u.w>>>19^d^d>>>8},c===(c|0)?u.x=c:l+=c;for(var p=0;p<l.length+64;p++)u.x^=l.charCodeAt(p)|0,u.next()}function o(c,u){return u.x=c.x,u.y=c.y,u.z=c.z,u.w=c.w,u}function i(c,u){var l=new a(c),p=u&&u.state,d=function(){return(l.next()>>>0)/4294967296};return d.double=function(){do var h=l.next()>>>11,f=(l.next()>>>0)/4294967296,m=(h+f)/(1<<21);while(m===0);return m},d.int32=l.next,d.quick=d,p&&(typeof p=="object"&&o(p,l),d.state=function(){return o(l,{})}),d}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.xor128=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),ND=Wt({"node_modules/seedrandom/lib/xorwow.js"(e,t){(function(n,r,s){function a(c){var u=this,l="";u.next=function(){var d=u.x^u.x>>>2;return u.x=u.y,u.y=u.z,u.z=u.w,u.w=u.v,(u.d=u.d+362437|0)+(u.v=u.v^u.v<<4^(d^d<<1))|0},u.x=0,u.y=0,u.z=0,u.w=0,u.v=0,c===(c|0)?u.x=c:l+=c;for(var p=0;p<l.length+64;p++)u.x^=l.charCodeAt(p)|0,p==l.length&&(u.d=u.x<<10^u.x>>>4),u.next()}function o(c,u){return u.x=c.x,u.y=c.y,u.z=c.z,u.w=c.w,u.v=c.v,u.d=c.d,u}function i(c,u){var l=new a(c),p=u&&u.state,d=function(){return(l.next()>>>0)/4294967296};return d.double=function(){do var h=l.next()>>>11,f=(l.next()>>>0)/4294967296,m=(h+f)/(1<<21);while(m===0);return m},d.int32=l.next,d.quick=d,p&&(typeof p=="object"&&o(p,l),d.state=function(){return o(l,{})}),d}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.xorwow=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),_D=Wt({"node_modules/seedrandom/lib/xorshift7.js"(e,t){(function(n,r,s){function a(c){var u=this;u.next=function(){var p=u.x,d=u.i,h,f,m;return h=p[d],h^=h>>>7,f=h^h<<24,h=p[d+1&7],f^=h^h>>>10,h=p[d+3&7],f^=h^h>>>3,h=p[d+4&7],f^=h^h<<7,h=p[d+7&7],h=h^h<<13,f^=h^h<<9,p[d]=f,u.i=d+1&7,f};function l(p,d){var h,f,m=[];if(d===(d|0))f=m[0]=d;else for(d=""+d,h=0;h<d.length;++h)m[h&7]=m[h&7]<<15^d.charCodeAt(h)+m[h+1&7]<<13;for(;m.length<8;)m.push(0);for(h=0;h<8&&m[h]===0;++h);for(h==8?f=m[7]=-1:f=m[h],p.x=m,p.i=0,h=256;h>0;--h)p.next()}l(u,c)}function o(c,u){return u.x=c.x.slice(),u.i=c.i,u}function i(c,u){c==null&&(c=+new Date);var l=new a(c),p=u&&u.state,d=function(){return(l.next()>>>0)/4294967296};return d.double=function(){do var h=l.next()>>>11,f=(l.next()>>>0)/4294967296,m=(h+f)/(1<<21);while(m===0);return m},d.int32=l.next,d.quick=d,p&&(p.x&&o(p,l),d.state=function(){return o(l,{})}),d}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.xorshift7=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),ED=Wt({"node_modules/seedrandom/lib/xor4096.js"(e,t){(function(n,r,s){function a(c){var u=this;u.next=function(){var p=u.w,d=u.X,h=u.i,f,m;return u.w=p=p+1640531527|0,m=d[h+34&127],f=d[h=h+1&127],m^=m<<13,f^=f<<17,m^=m>>>15,f^=f>>>12,m=d[h]=m^f,u.i=h,m+(p^p>>>16)|0};function l(p,d){var h,f,m,g,b,y=[],v=128;for(d===(d|0)?(f=d,d=null):(d=d+"\0",f=0,v=Math.max(v,d.length)),m=0,g=-32;g<v;++g)d&&(f^=d.charCodeAt((g+32)%d.length)),g===0&&(b=f),f^=f<<10,f^=f>>>15,f^=f<<4,f^=f>>>13,g>=0&&(b=b+1640531527|0,h=y[g&127]^=f+b,m=h==0?m+1:0);for(m>=128&&(y[(d&&d.length||0)&127]=-1),m=127,g=4*128;g>0;--g)f=y[m+34&127],h=y[m=m+1&127],f^=f<<13,h^=h<<17,f^=f>>>15,h^=h>>>12,y[m]=f^h;p.w=b,p.X=y,p.i=m}l(u,c)}function o(c,u){return u.i=c.i,u.w=c.w,u.X=c.X.slice(),u}function i(c,u){c==null&&(c=+new Date);var l=new a(c),p=u&&u.state,d=function(){return(l.next()>>>0)/4294967296};return d.double=function(){do var h=l.next()>>>11,f=(l.next()>>>0)/4294967296,m=(h+f)/(1<<21);while(m===0);return m},d.int32=l.next,d.quick=d,p&&(p.X&&o(p,l),d.state=function(){return o(l,{})}),d}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.xor4096=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),AD=Wt({"node_modules/seedrandom/lib/tychei.js"(e,t){(function(n,r,s){function a(c){var u=this,l="";u.next=function(){var d=u.b,h=u.c,f=u.d,m=u.a;return d=d<<25^d>>>7^h,h=h-f|0,f=f<<24^f>>>8^m,m=m-d|0,u.b=d=d<<20^d>>>12^h,u.c=h=h-f|0,u.d=f<<16^h>>>16^m,u.a=m-d|0},u.a=0,u.b=0,u.c=-1640531527,u.d=1367130551,c===Math.floor(c)?(u.a=c/4294967296|0,u.b=c|0):l+=c;for(var p=0;p<l.length+20;p++)u.b^=l.charCodeAt(p)|0,u.next()}function o(c,u){return u.a=c.a,u.b=c.b,u.c=c.c,u.d=c.d,u}function i(c,u){var l=new a(c),p=u&&u.state,d=function(){return(l.next()>>>0)/4294967296};return d.double=function(){do var h=l.next()>>>11,f=(l.next()>>>0)/4294967296,m=(h+f)/(1<<21);while(m===0);return m},d.int32=l.next,d.quick=d,p&&(typeof p=="object"&&o(p,l),d.state=function(){return o(l,{})}),d}r&&r.exports?r.exports=i:s&&s.amd?s(function(){return i}):this.tychei=i})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),$D=Wt({"(disabled):crypto"(){}}),DD=Wt({"node_modules/seedrandom/seedrandom.js"(e,t){(function(n,r,s){var a=256,o=6,i=52,c="random",u=s.pow(a,o),l=s.pow(2,i),p=l*2,d=a-1,h;function f(k,S,C){var E=[];S=S==!0?{entropy:!0}:S||{};var $=y(b(S.entropy?[k,x(r)]:k==null?v():k,3),E),F=new m(E),A=function(){for(var R=F.g(o),T=u,L=0;R<l;)R=(R+L)*a,T*=a,L=F.g(1);for(;R>=p;)R/=2,T/=2,L>>>=1;return(R+L)/T};return A.int32=function(){return F.g(4)|0},A.quick=function(){return F.g(4)/4294967296},A.double=A,y(x(F.S),r),(S.pass||C||function(R,T,L,V){return V&&(V.S&&g(V,F),R.state=function(){return g(F,{})}),L?(s[c]=R,T):R})(A,$,"global"in S?S.global:this==s,S.state)}function m(k){var S,C=k.length,E=this,$=0,F=E.i=E.j=0,A=E.S=[];for(C||(k=[C++]);$<a;)A[$]=$++;for($=0;$<a;$++)A[$]=A[F=d&F+k[$%C]+(S=A[$])],A[F]=S;(E.g=function(R){for(var T,L=0,V=E.i,G=E.j,j=E.S;R--;)T=j[V=d&V+1],L=L*a+j[d&(j[V]=j[G=d&G+T])+(j[G]=T)];return E.i=V,E.j=G,L})(a)}function g(k,S){return S.i=k.i,S.j=k.j,S.S=k.S.slice(),S}function b(k,S){var C=[],E=typeof k,$;if(S&&E=="object")for($ in k)try{C.push(b(k[$],S-1))}catch(F){}return C.length?C:E=="string"?k:k+"\0"}function y(k,S){for(var C=k+"",E,$=0;$<C.length;)S[d&$]=d&(E^=S[d&$]*19)+C.charCodeAt($++);return x(S)}function v(){try{var k;return h&&(k=h.randomBytes)?k=k(a):(k=new Uint8Array(a),(n.crypto||n.msCrypto).getRandomValues(k)),x(k)}catch(E){var S=n.navigator,C=S&&S.plugins;return[+new Date,n,C,n.screen,x(r)]}}function x(k){return String.fromCharCode.apply(0,k)}if(y(s.random(),r),typeof t=="object"&&t.exports){t.exports=f;try{h=$D()}catch(k){}}else typeof define=="function"&&define.amd?define(function(){return f}):s["seed"+c]=f})(typeof self!="undefined"?self:e,[],Math)}}),cf=Wt({"node_modules/seedrandom/index.js"(e,t){var n=TD(),r=CD(),s=ND(),a=_D(),o=ED(),i=AD(),c=DD();c.alea=n,c.xor128=r,c.xorwow=s,c.xorshift7=a,c.xor4096=o,c.tychei=i,t.exports=c}}),G1=Wt({"(disabled):node_modules/string_decoder/lib/string_decoder.js"(){}}),Nv=Wt({"(disabled):fs"(){}}),_h=Wt({"(disabled):path"(){}}),FD=Wt({"(disabled):worker_threads"(){}}),RD=Wt({"(disabled):perf_hooks"(){}}),PD=Wt({"(disabled):os"(){}}),OD=Wt({"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(e,t){var n=(()=>{var r=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(r=r||__filename),function(s){s=s||{};function a(){return ue.buffer!=ze&&yt(ue.buffer),dt}function o(){return ue.buffer!=ze&&yt(ue.buffer),qn}function i(){return ue.buffer!=ze&&yt(ue.buffer),At}function c(){return ue.buffer!=ze&&yt(ue.buffer),on}function u(){return ue.buffer!=ze&&yt(ue.buffer),Ir}function l(){return ue.buffer!=ze&&yt(ue.buffer),vn}function p(){return ue.buffer!=ze&&yt(ue.buffer),jn}var d=typeof s!="undefined"?s:{},h,f;d.ready=new Promise(function(P,K){h=P,f=K});var m;typeof process!="undefined"&&process.listeners&&(m={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var g=Object.assign({},d),b=[],y="./this.program",v=(P,K)=>{throw K},x=typeof window=="object",k=typeof importScripts=="function",S=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",C=d.ENVIRONMENT_IS_PTHREAD||!1,E="";function $(P){return d.locateFile?d.locateFile(P,E):E+P}var F,A,R,T;function L(P){if(P instanceof Oa)return;J("exiting due to exception: "+P)}if(S){k?E=_h().dirname(E)+"/":E=__dirname+"/";var V,G;typeof U1=="function"&&(V=Nv(),G=_h()),F=(K,oe)=>(K=G.normalize(K),V.readFileSync(K,oe?void 0:"utf8")),R=K=>{var oe=F(K,!0);return oe.buffer||(oe=new Uint8Array(oe)),oe},A=(K,oe,pe)=>{K=G.normalize(K),V.readFile(K,function(ke,He){ke?pe(ke):oe(He.buffer)})},process.argv.length>1&&(y=process.argv[1].replace(/\\/g,"/")),b=process.argv.slice(2),process.on("uncaughtException",function(K){if(!(K instanceof Oa))throw K}),process.on("unhandledRejection",function(K){throw K}),v=(K,oe)=>{if(Sr())throw process.exitCode=K,oe;L(oe),process.exit(K)},d.inspect=function(){return"[Emscripten Module object]"};let P;try{P=FD()}catch(K){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),K}global.Worker=P.Worker}else(x||k)&&(k?E=self.location.href:typeof document!="undefined"&&document.currentScript&&(E=document.currentScript.src),typeof r!="undefined"&&r&&(E=r),E.indexOf("blob:")!==0?E=E.substr(0,E.replace(/[?#].*/,"").lastIndexOf("/")+1):E="",S||(F=P=>{var K=new XMLHttpRequest;return K.open("GET",P,!1),K.send(null),K.responseText},k&&(R=P=>{var K=new XMLHttpRequest;return K.open("GET",P,!1),K.responseType="arraybuffer",K.send(null),new Uint8Array(K.response)}),A=(P,K,oe)=>{var pe=new XMLHttpRequest;pe.open("GET",P,!0),pe.responseType="arraybuffer",pe.onload=()=>{if(pe.status==200||pe.status==0&&pe.response){K(pe.response);return}oe()},pe.onerror=oe,pe.send(null)}),T=P=>document.title=P);S&&typeof performance=="undefined"&&(global.performance=RD().performance);var j=console.log.bind(console),H=console.warn.bind(console);S&&(j=P=>V.writeSync(1,P+`
+`),H=P=>V.writeSync(2,P+`
+`));var Z=d.print||j,J=d.printErr||H;Object.assign(d,g),g=null,d.arguments&&(b=d.arguments),d.thisProgram&&(y=d.thisProgram),d.quit&&(v=d.quit);var ee=4,re=Atomics.load,te=Atomics.store,ie=Atomics.compareExchange,ne;d.wasmBinary&&(ne=d.wasmBinary);var le=d.noExitRuntime||!0;typeof WebAssembly!="object"&&Yi("no native wasm support detected");var ue,ye,ve=!1,Se;function Le(P,K){P||Yi(K)}var je=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function nt(P,K,oe){for(var pe=K+oe,ke=K;P[ke]&&!(ke>=pe);)++ke;if(ke-K>16&&P.buffer&&je)return je.decode(P.buffer instanceof SharedArrayBuffer?P.slice(K,ke):P.subarray(K,ke));for(var He="";K<ke;){var Ne=P[K++];if(!(Ne&128)){He+=String.fromCharCode(Ne);continue}var Re=P[K++]&63;if((Ne&224)==192){He+=String.fromCharCode((Ne&31)<<6|Re);continue}var Mt=P[K++]&63;if((Ne&240)==224?Ne=(Ne&15)<<12|Re<<6|Mt:Ne=(Ne&7)<<18|Re<<12|Mt<<6|P[K++]&63,Ne<65536)He+=String.fromCharCode(Ne);else{var cr=Ne-65536;He+=String.fromCharCode(55296|cr>>10,56320|cr&1023)}}return He}function rt(P,K){return P?nt(o(),P,K):""}function st(P,K,oe,pe){if(!(pe>0))return 0;for(var ke=oe,He=oe+pe-1,Ne=0;Ne<P.length;++Ne){var Re=P.charCodeAt(Ne);if(Re>=55296&&Re<=57343){var Mt=P.charCodeAt(++Ne);Re=65536+((Re&1023)<<10)|Mt&1023}if(Re<=127){if(oe>=He)break;K[oe++]=Re}else if(Re<=2047){if(oe+1>=He)break;K[oe++]=192|Re>>6,K[oe++]=128|Re&63}else if(Re<=65535){if(oe+2>=He)break;K[oe++]=224|Re>>12,K[oe++]=128|Re>>6&63,K[oe++]=128|Re&63}else{if(oe+3>=He)break;K[oe++]=240|Re>>18,K[oe++]=128|Re>>12&63,K[oe++]=128|Re>>6&63,K[oe++]=128|Re&63}}return K[oe]=0,oe-ke}function Ze(P,K,oe){return st(P,o(),K,oe)}var ze,dt,qn,At,wr,on,Ir,vn,jn;C&&(ze=d.buffer);function yt(P){ze=P,d.HEAP8=dt=new Int8Array(P),d.HEAP16=At=new Int16Array(P),d.HEAP32=on=new Int32Array(P),d.HEAPU8=qn=new Uint8Array(P),d.HEAPU16=wr=new Uint16Array(P),d.HEAPU32=Ir=new Uint32Array(P),d.HEAPF32=vn=new Float32Array(P),d.HEAPF64=jn=new Float64Array(P)}var Pn=d.INITIAL_MEMORY||16777216;if(C)ue=d.wasmMemory,ze=d.buffer;else if(d.wasmMemory)ue=d.wasmMemory;else if(ue=new WebAssembly.Memory({initial:Pn/65536,maximum:32768,shared:!0}),!(ue.buffer instanceof SharedArrayBuffer))throw J("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),S&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");ue&&(ze=ue.buffer),Pn=ze.byteLength,yt(ze);var hn,ts=[],qs=[],kr=[],xl=!1;function Sr(){return le}function Ki(){if(d.preRun)for(typeof d.preRun=="function"&&(d.preRun=[d.preRun]);d.preRun.length;)Ng(d.preRun.shift());Sl(ts)}function Xt(){xl=!0,!C&&Sl(qs)}function Dp(){if(!C){if(d.postRun)for(typeof d.postRun=="function"&&(d.postRun=[d.postRun]);d.postRun.length;)GI(d.postRun.shift());Sl(kr)}}function Ng(P){ts.unshift(P)}function _g(P){qs.unshift(P)}function GI(P){kr.unshift(P)}var js=0,Xi=null,xs=null;function HI(P){js++,d.monitorRunDependencies&&d.monitorRunDependencies(js)}function qI(P){if(js--,d.monitorRunDependencies&&d.monitorRunDependencies(js),js==0&&(Xi!==null&&(clearInterval(Xi),Xi=null),xs)){var K=xs;xs=null,K()}}function Yi(P){C?postMessage({cmd:"onAbort",arg:P}):d.onAbort&&d.onAbort(P),P="Aborted("+P+")",J(P),ve=!0,Se=1,P+=". Build with -sASSERTIONS for more info.";var K=new WebAssembly.RuntimeError(P);throw f(K),K}var Eg="data:application/octet-stream;base64,";function Fp(P){return P.startsWith(Eg)}function wl(P){return P.startsWith("file://")}var xn;xn="tfjs-backend-wasm-threaded-simd.wasm",Fp(xn)||(xn=$(xn));function Rp(P){try{if(P==xn&&ne)return new Uint8Array(ne);if(R)return R(P);throw"both async and sync fetching of the wasm failed"}catch(K){Yi(K)}}function Ag(){if(!ne&&(x||k)){if(typeof fetch=="function"&&!wl(xn))return fetch(xn,{credentials:"same-origin"}).then(function(P){if(!P.ok)throw"failed to load wasm binary file at '"+xn+"'";return P.arrayBuffer()}).catch(function(){return Rp(xn)});if(A)return new Promise(function(P,K){A(xn,function(oe){P(new Uint8Array(oe))},K)})}return Promise.resolve().then(function(){return Rp(xn)})}function $g(){var P={env:qp,wasi_snapshot_preview1:qp};function K(Ne,Re){var Mt=Ne.exports;if(d.asm=Mt,Wg(d.asm._emscripten_tls_init),hn=d.asm.__indirect_function_table,_g(d.asm.__wasm_call_ctors),ye=Re,!C){var cr=$e.unusedWorkers.length;$e.unusedWorkers.forEach(function(Is){$e.loadWasmModuleToWorker(Is,function(){--cr||qI("wasm-instantiate")})})}}C||HI("wasm-instantiate");function oe(Ne){K(Ne.instance,Ne.module)}function pe(Ne){return Ag().then(function(Re){return WebAssembly.instantiate(Re,P)}).then(function(Re){return Re}).then(Ne,function(Re){J("failed to asynchronously prepare wasm: "+Re),Yi(Re)})}function ke(){return!ne&&typeof WebAssembly.instantiateStreaming=="function"&&!Fp(xn)&&!wl(xn)&&!S&&typeof fetch=="function"?fetch(xn,{credentials:"same-origin"}).then(function(Ne){var Re=WebAssembly.instantiateStreaming(Ne,P);return Re.then(oe,function(Mt){return J("wasm streaming compile failed: "+Mt),J("falling back to ArrayBuffer instantiation"),pe(oe)})}):pe(oe)}if(d.instantiateWasm)try{var He=d.instantiateWasm(P,K);return He}catch(Ne){J("Module.instantiateWasm callback failed with error: "+Ne),f(Ne)}return ke().catch(f),{}}var Dg,jI,Fg={};function Oa(P){this.name="ExitStatus",this.message="Program terminated with exit("+P+")",this.status=P}function Rg(P){var K=$e.pthreads[P];delete $e.pthreads[P],K.terminate(),uy(P),$e.runningWorkers.splice($e.runningWorkers.indexOf(K),1),K.pthread_ptr=0}function Pg(P){var K=$e.pthreads[P];K.postMessage({cmd:"cancel"})}function Il(P){var K=$e.pthreads[P];Le(K),$e.returnWorkerToPool(K)}function Pp(P){var K=$e.getNewWorker();if(!K)return 6;$e.runningWorkers.push(K),$e.pthreads[P.pthread_ptr]=K,K.pthread_ptr=P.pthread_ptr;var oe={cmd:"run",start_routine:P.startRoutine,arg:P.arg,pthread_ptr:P.pthread_ptr};return K.runPthread=()=>{oe.time=performance.now(),K.postMessage(oe,P.transferList)},K.loaded&&(K.runPthread(),delete K.runPthread),0}var Op={varargs:void 0,get:function(){Op.varargs+=4;var P=c()[Op.varargs-4>>2];return P},getStr:function(P){var K=rt(P);return K}};function kl(P){if(C)return Ks(1,1,P);Se=P,Sr()||($e.terminateAllThreads(),d.onExit&&d.onExit(P),ve=!0),v(P,new Oa(P))}function KI(P,K){if(Se=P,!K&&C)throw Lp(P),"unwind";kl(P)}var Mp=KI;function Og(P){if(P instanceof Oa||P=="unwind")return Se;v(1,P)}var $e={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){C?$e.initWorker():$e.initMainThread()},initMainThread:function(){for(var P=8;P--;)$e.allocateUnusedWorker()},initWorker:function(){le=!1},setExitStatus:function(P){Se=P},terminateAllThreads:function(){for(var P of Object.values($e.pthreads))$e.returnWorkerToPool(P);for(var P of $e.unusedWorkers)P.terminate();$e.unusedWorkers=[]},returnWorkerToPool:function(P){var K=P.pthread_ptr;delete $e.pthreads[K],$e.unusedWorkers.push(P),$e.runningWorkers.splice($e.runningWorkers.indexOf(P),1),P.pthread_ptr=0,uy(K)},receiveObjectTransfer:function(P){},threadInitTLS:function(){$e.tlsInitFunctions.forEach(P=>P())},loadWasmModuleToWorker:function(P,K){P.onmessage=oe=>{var pe=oe.data,ke=pe.cmd;if(P.pthread_ptr&&($e.currentProxiedOperationCallerThread=P.pthread_ptr),pe.targetThread&&pe.targetThread!=Jp()){var He=$e.pthreads[pe.targetThread];He?He.postMessage(pe,pe.transferList):J('Internal error! Worker sent a message "'+ke+'" to target pthread '+pe.targetThread+", but that thread no longer exists!"),$e.currentProxiedOperationCallerThread=void 0;return}ke==="processProxyingQueue"?Tl(pe.queue):ke==="spawnThread"?Pp(pe):ke==="cleanupThread"?Il(pe.thread):ke==="killThread"?Rg(pe.thread):ke==="cancelThread"?Pg(pe.thread):ke==="loaded"?(P.loaded=!0,K&&K(P),P.runPthread&&(P.runPthread(),delete P.runPthread)):ke==="print"?Z("Thread "+pe.threadId+": "+pe.text):ke==="printErr"?J("Thread "+pe.threadId+": "+pe.text):ke==="alert"?alert("Thread "+pe.threadId+": "+pe.text):pe.target==="setimmediate"?P.postMessage(pe):ke==="onAbort"?d.onAbort&&d.onAbort(pe.arg):ke&&J("worker sent an unknown command "+ke),$e.currentProxiedOperationCallerThread=void 0},P.onerror=oe=>{var pe="worker sent an error!";throw J(pe+" "+oe.filename+":"+oe.lineno+": "+oe.message),oe},S&&(P.on("message",function(oe){P.onmessage({data:oe})}),P.on("error",function(oe){P.onerror(oe)}),P.on("detachedExit",function(){})),P.postMessage({cmd:"load",urlOrBlob:d.mainScriptUrlOrBlob||r,wasmMemory:ue,wasmModule:ye})},allocateUnusedWorker:function(){var P=$("tfjs-backend-wasm-threaded-simd.worker.js");$e.unusedWorkers.push(new Worker(P))},getNewWorker:function(){return $e.unusedWorkers.length==0&&($e.allocateUnusedWorker(),$e.loadWasmModuleToWorker($e.unusedWorkers[0])),$e.unusedWorkers.pop()}};d.PThread=$e;function Sl(P){for(;P.length>0;)P.shift()(d)}function Mg(P){var K=ly(),oe=P();return Qp(K),oe}function XI(P){return P}function YI(P){var K=/\b_Z[\w\d_]+/g;return P.replace(K,function(oe){var pe=oe;return oe===pe?oe:pe+" ["+oe+"]"})}function Lg(){var P=Jp(),K=c()[P+44>>2],oe=c()[P+48>>2],pe=K-oe;rk(K,pe),Qp(K)}d.establishStackSpace=Lg;function Lp(P){if(C)return Ks(2,0,P);try{Mp(P)}catch(K){Og(K)}}var Zi=[];function zg(P){var K=Zi[P];return K||(P>=Zi.length&&(Zi.length=P+1),Zi[P]=K=hn.get(P)),K}function Bg(P,K){var oe=zg(P)(K);Sr()?$e.setExitStatus(oe):nk(oe)}d.invokeEntryPoint=Bg;function ZI(){var P=new Error;if(!P.stack){try{throw new Error}catch(K){P=K}if(!P.stack)return"(no stack trace available)"}return P.stack.toString()}function Wg(P){$e.tlsInitFunctions.push(P)}function Vg(P,K){a().set(P,K)}function Ug(P){QI(P,!k,1,!x),$e.threadInitTLS()}function Gg(P){C?postMessage({cmd:"cleanupThread",thread:P}):Il(P)}function zp(P,K,oe,pe){return C?Ks(3,1,P,K,oe,pe):Bp(P,K,oe,pe)}function Bp(P,K,oe,pe){if(typeof SharedArrayBuffer=="undefined")return J("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var ke=[],He=0;if(C&&(ke.length===0||He))return zp(P,K,oe,pe);if(He)return He;var Ne={startRoutine:oe,pthread_ptr:P,arg:pe,transferList:ke};return C?(Ne.cmd="spawnThread",postMessage(Ne,ke),0):Pp(Ne)}function Hg(){return 2097152}var qg=!0;function jg(){return qg}function Tl(P){Atomics.store(c(),P>>2,1),Jp()&&tk(P),Atomics.compareExchange(c(),P>>2,1,0)}d.executeNotifiedProxyingQueue=Tl;function Kg(P,K,oe,pe){if(P==K)setTimeout(()=>Tl(pe));else if(C)postMessage({targetThread:P,cmd:"processProxyingQueue",queue:pe});else{var ke=$e.pthreads[P];if(!ke)return;ke.postMessage({cmd:"processProxyingQueue",queue:pe})}return 1}function Xg(P,K,oe){return-1}function Yg(){Yi("")}function Ma(P){Ma.shown||(Ma.shown={}),Ma.shown[P]||(Ma.shown[P]=1,S&&(P="warning: "+P),J(P))}function Zg(){S||k||Ma("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function Jg(){return Date.now()}function Wp(){return 2147483648}function Qg(){return Wp()}var Ji;S?Ji=()=>{var P=process.hrtime();return P[0]*1e3+P[1]/1e6}:C?Ji=()=>performance.now()-d.__performance_now_clock_drift:Ji=()=>performance.now();function eb(P,K,oe){o().copyWithin(P,K,K+oe)}function tb(){return S?PD().cpus().length:navigator.hardwareConcurrency}function Ks(P,K){var oe=arguments.length-2,pe=arguments;return Mg(()=>{for(var ke=oe,He=eh(ke*8),Ne=He>>3,Re=0;Re<oe;Re++){var Mt=pe[2+Re];p()[Ne+Re]=Mt}return ek(P,ke,He,K)})}var Cl=[];function nb(P,K,oe){Cl.length=K;for(var pe=oe>>3,ke=0;ke<K;ke++)Cl[ke]=p()[pe+ke];var He=P<0,Ne=He?Fg[-P-1]:lb[P];return Ne.apply(null,Cl)}function rb(P){try{return ue.grow(P-ze.byteLength+65535>>>16),yt(ue.buffer),1}catch(K){}}function sb(P){var K=o().length;if(P=P>>>0,P<=K)return!1;var oe=Wp();if(P>oe)return!1;let pe=(Mt,cr)=>Mt+(cr-Mt%cr)%cr;for(var ke=1;ke<=4;ke*=2){var He=K*(1+.2/ke);He=Math.min(He,P+100663296);var Ne=Math.min(oe,pe(Math.max(P,He),65536)),Re=rb(Ne);if(Re)return!0}return!1}function ab(){throw"unwind"}function Vp(P){return C?Ks(4,1,P):52}function Up(P,K,oe,pe,ke){return C?Ks(5,1,P,K,oe,pe,ke):70}var ob=[null,[],[]];function ib(P,K){var oe=ob[P];K===0||K===10?((P===1?Z:J)(nt(oe,0)),oe.length=0):oe.push(K)}function Gp(P,K,oe,pe){if(C)return Ks(6,1,P,K,oe,pe);for(var ke=0,He=0;He<oe;He++){var Ne=u()[K>>2],Re=u()[K+4>>2];K+=8;for(var Mt=0;Mt<Re;Mt++)ib(P,o()[Ne+Mt]);ke+=Re}return u()[pe>>2]=ke,0}function Hp(P){var K=d["_"+P];return K}function cb(P,K,oe,pe,ke){var He={string:Kn=>{var nc=0;if(Kn!=null&&Kn!==0){var ok=(Kn.length<<2)+1;nc=eh(ok),Ze(Kn,nc,ok)}return nc},array:Kn=>{var nc=eh(Kn.length);return Vg(Kn,nc),nc}};function Ne(Kn){return K==="string"?rt(Kn):K==="boolean"?Boolean(Kn):Kn}var Re=Hp(P),Mt=[],cr=0;if(pe)for(var Is=0;Is<pe.length;Is++){var ak=He[oe[Is]];ak?(cr===0&&(cr=ly()),Mt[Is]=ak(pe[Is])):Mt[Is]=pe[Is]}var dy=Re.apply(null,Mt);function dD(Kn){return cr!==0&&Qp(cr),Ne(Kn)}return dy=dD(dy),dy}function ub(P,K,oe,pe){oe=oe||[];var ke=oe.every(Ne=>Ne==="number"||Ne==="boolean"),He=K!=="string";return He&&ke&&!pe?Hp(P):function(){return cb(P,K,oe,arguments,pe)}}$e.init();var lb=[null,kl,Lp,zp,Vp,Up,Gp],qp={__emscripten_init_main_thread_js:Ug,__emscripten_thread_cleanup:Gg,__pthread_create_js:Bp,_emscripten_default_pthread_stack_size:Hg,_emscripten_get_now_is_monotonic:jg,_emscripten_notify_task_queue:Kg,_emscripten_set_offscreencanvas_size:Xg,abort:Yg,emscripten_check_blocking_allowed:Zg,emscripten_date_now:Jg,emscripten_get_heap_max:Qg,emscripten_get_now:Ji,emscripten_memcpy_big:eb,emscripten_num_logical_cores:tb,emscripten_receive_on_main_thread_js:nb,emscripten_resize_heap:sb,emscripten_unwind_to_js_event_loop:ab,exit:Mp,fd_close:Vp,fd_seek:Up,fd_write:Gp,memory:ue||d.wasmMemory},JI=$g(),db=d.___wasm_call_ctors=function(){return(db=d.___wasm_call_ctors=d.asm.__wasm_call_ctors).apply(null,arguments)},pb=d._init=function(){return(pb=d._init=d.asm.init).apply(null,arguments)},hb=d._init_with_threads_count=function(){return(hb=d._init_with_threads_count=d.asm.init_with_threads_count).apply(null,arguments)},fb=d._get_threads_count=function(){return(fb=d._get_threads_count=d.asm.get_threads_count).apply(null,arguments)},mb=d._register_tensor=function(){return(mb=d._register_tensor=d.asm.register_tensor).apply(null,arguments)},gb=d._dispose_data=function(){return(gb=d._dispose_data=d.asm.dispose_data).apply(null,arguments)},bb=d._dispose=function(){return(bb=d._dispose=d.asm.dispose).apply(null,arguments)},yb=d._Abs=function(){return(yb=d._Abs=d.asm.Abs).apply(null,arguments)},vb=d._Add=function(){return(vb=d._Add=d.asm.Add).apply(null,arguments)},xb=d._AddN=function(){return(xb=d._AddN=d.asm.AddN).apply(null,arguments)},wb=d._All=function(){return(wb=d._All=d.asm.All).apply(null,arguments)},Ib=d._Any=function(){return(Ib=d._Any=d.asm.Any).apply(null,arguments)},kb=d._ArgMax=function(){return(kb=d._ArgMax=d.asm.ArgMax).apply(null,arguments)},Sb=d._AvgPool=function(){return(Sb=d._AvgPool=d.asm.AvgPool).apply(null,arguments)},Tb=d._BatchMatMul=function(){return(Tb=d._BatchMatMul=d.asm.BatchMatMul).apply(null,arguments)},Cb=d._Ceil=function(){return(Cb=d._Ceil=d.asm.Ceil).apply(null,arguments)},Nb=d._ClipByValue=function(){return(Nb=d._ClipByValue=d.asm.ClipByValue).apply(null,arguments)},_b=d._Conv2D=function(){return(_b=d._Conv2D=d.asm.Conv2D).apply(null,arguments)},Eb=d._Conv2DBackpropInput=function(){return(Eb=d._Conv2DBackpropInput=d.asm.Conv2DBackpropInput).apply(null,arguments)},Ab=d._Cos=function(){return(Ab=d._Cos=d.asm.Cos).apply(null,arguments)},$b=d._Cosh=function(){return($b=d._Cosh=d.asm.Cosh).apply(null,arguments)},Db=d._CropAndResize=function(){return(Db=d._CropAndResize=d.asm.CropAndResize).apply(null,arguments)},Fb=d._Cumprod=function(){return(Fb=d._Cumprod=d.asm.Cumprod).apply(null,arguments)},Rb=d._Cumsum=function(){return(Rb=d._Cumsum=d.asm.Cumsum).apply(null,arguments)},Pb=d._DepthToSpace=function(){return(Pb=d._DepthToSpace=d.asm.DepthToSpace).apply(null,arguments)},Ob=d._DepthwiseConv2dNative=function(){return(Ob=d._DepthwiseConv2dNative=d.asm.DepthwiseConv2dNative).apply(null,arguments)},Mb=d._Elu=function(){return(Mb=d._Elu=d.asm.Elu).apply(null,arguments)},Lb=d._Equal=function(){return(Lb=d._Equal=d.asm.Equal).apply(null,arguments)},zb=d._Exp=function(){return(zb=d._Exp=d.asm.Exp).apply(null,arguments)},Bb=d._FlipLeftRight=function(){return(Bb=d._FlipLeftRight=d.asm.FlipLeftRight).apply(null,arguments)},Wb=d._Floor=function(){return(Wb=d._Floor=d.asm.Floor).apply(null,arguments)},Vb=d._FloorDiv=function(){return(Vb=d._FloorDiv=d.asm.FloorDiv).apply(null,arguments)},Ub=d._FusedBatchNorm=function(){return(Ub=d._FusedBatchNorm=d.asm.FusedBatchNorm).apply(null,arguments)},Gb=d._FusedConv2D=function(){return(Gb=d._FusedConv2D=d.asm.FusedConv2D).apply(null,arguments)},Hb=d._FusedDepthwiseConv2D=function(){return(Hb=d._FusedDepthwiseConv2D=d.asm.FusedDepthwiseConv2D).apply(null,arguments)},qb=d._Gather=function(){return(qb=d._Gather=d.asm.Gather).apply(null,arguments)},jb=d._GatherNd=function(){return(jb=d._GatherNd=d.asm.GatherNd).apply(null,arguments)},Kb=d._Greater=function(){return(Kb=d._Greater=d.asm.Greater).apply(null,arguments)},Xb=d._GreaterEqual=function(){return(Xb=d._GreaterEqual=d.asm.GreaterEqual).apply(null,arguments)},Yb=d._LeakyRelu=function(){return(Yb=d._LeakyRelu=d.asm.LeakyRelu).apply(null,arguments)},Zb=d._Less=function(){return(Zb=d._Less=d.asm.Less).apply(null,arguments)},Jb=d._LessEqual=function(){return(Jb=d._LessEqual=d.asm.LessEqual).apply(null,arguments)},Qb=d._Log=function(){return(Qb=d._Log=d.asm.Log).apply(null,arguments)},ey=d._LogicalAnd=function(){return(ey=d._LogicalAnd=d.asm.LogicalAnd).apply(null,arguments)},ty=d._LogicalNot=function(){return(ty=d._LogicalNot=d.asm.LogicalNot).apply(null,arguments)},ny=d._LogicalOr=function(){return(ny=d._LogicalOr=d.asm.LogicalOr).apply(null,arguments)},ry=d._LogicalXor=function(){return(ry=d._LogicalXor=d.asm.LogicalXor).apply(null,arguments)},sy=d._Max=function(){return(sy=d._Max=d.asm.Max).apply(null,arguments)},jp=d._MaxPool=function(){return(jp=d._MaxPool=d.asm.MaxPool).apply(null,arguments)},Kp=d._Maximum=function(){return(Kp=d._Maximum=d.asm.Maximum).apply(null,arguments)},Nl=d._Mean=function(){return(Nl=d._Mean=d.asm.Mean).apply(null,arguments)},ay=d._Min=function(){return(ay=d._Min=d.asm.Min).apply(null,arguments)},oy=d._Minimum=function(){return(oy=d._Minimum=d.asm.Minimum).apply(null,arguments)},Qi=d._MirrorPad=function(){return(Qi=d._MirrorPad=d.asm.MirrorPad).apply(null,arguments)},Xp=d._Multiply=function(){return(Xp=d._Multiply=d.asm.Multiply).apply(null,arguments)},ec=d._Neg=function(){return(ec=d._Neg=d.asm.Neg).apply(null,arguments)},tc=d._NonMaxSuppressionV3=function(){return(tc=d._NonMaxSuppressionV3=d.asm.NonMaxSuppressionV3).apply(null,arguments)},iy=d._NonMaxSuppressionV4=function(){return(iy=d._NonMaxSuppressionV4=d.asm.NonMaxSuppressionV4).apply(null,arguments)},X=d._NonMaxSuppressionV5=function(){return(X=d._NonMaxSuppressionV5=d.asm.NonMaxSuppressionV5).apply(null,arguments)},ae=d._NotEqual=function(){return(ae=d._NotEqual=d.asm.NotEqual).apply(null,arguments)},Ie=d._OneHot=function(){return(Ie=d._OneHot=d.asm.OneHot).apply(null,arguments)},Ge=d._PadV2=function(){return(Ge=d._PadV2=d.asm.PadV2).apply(null,arguments)},wt=d._Pow=function(){return(wt=d._Pow=d.asm.Pow).apply(null,arguments)},It=d._Prelu=function(){return(It=d._Prelu=d.asm.Prelu).apply(null,arguments)},Ve=d._Prod=function(){return(Ve=d._Prod=d.asm.Prod).apply(null,arguments)},Be=d._RealDiv=function(){return(Be=d._RealDiv=d.asm.RealDiv).apply(null,arguments)},Ot=d._Relu=function(){return(Ot=d._Relu=d.asm.Relu).apply(null,arguments)},ir=d._Relu6=function(){return(ir=d._Relu6=d.asm.Relu6).apply(null,arguments)},ws=d._ResizeBilinear=function(){return(ws=d._ResizeBilinear=d.asm.ResizeBilinear).apply(null,arguments)},Yp=d._ResizeNearestNeighbor=function(){return(Yp=d._ResizeNearestNeighbor=d.asm.ResizeNearestNeighbor).apply(null,arguments)},_l=d._Reverse=function(){return(_l=d._Reverse=d.asm.Reverse).apply(null,arguments)},cy=d._RotateWithOffset=function(){return(cy=d._RotateWithOffset=d.asm.RotateWithOffset).apply(null,arguments)},wn=d._Round=function(){return(wn=d._Round=d.asm.Round).apply(null,arguments)},Xs=d._Rsqrt=function(){return(Xs=d._Rsqrt=d.asm.Rsqrt).apply(null,arguments)},Zp=d._ScatterNd=function(){return(Zp=d._ScatterNd=d.asm.ScatterNd).apply(null,arguments)},F$=d._SelectV2=function(){return(F$=d._SelectV2=d.asm.SelectV2).apply(null,arguments)},R$=d._Sigmoid=function(){return(R$=d._Sigmoid=d.asm.Sigmoid).apply(null,arguments)},P$=d._Sin=function(){return(P$=d._Sin=d.asm.Sin).apply(null,arguments)},O$=d._Softmax=function(){return(O$=d._Softmax=d.asm.Softmax).apply(null,arguments)},M$=d._SparseFillEmptyRows=function(){return(M$=d._SparseFillEmptyRows=d.asm.SparseFillEmptyRows).apply(null,arguments)},L$=d._SparseReshape=function(){return(L$=d._SparseReshape=d.asm.SparseReshape).apply(null,arguments)},z$=d._SparseSegmentReduction=function(){return(z$=d._SparseSegmentReduction=d.asm.SparseSegmentReduction).apply(null,arguments)},B$=d._Sqrt=function(){return(B$=d._Sqrt=d.asm.Sqrt).apply(null,arguments)},W$=d._Square=function(){return(W$=d._Square=d.asm.Square).apply(null,arguments)},V$=d._SquaredDifference=function(){return(V$=d._SquaredDifference=d.asm.SquaredDifference).apply(null,arguments)},U$=d._Step=function(){return(U$=d._Step=d.asm.Step).apply(null,arguments)},G$=d._StridedSlice=function(){return(G$=d._StridedSlice=d.asm.StridedSlice).apply(null,arguments)},H$=d._Sub=function(){return(H$=d._Sub=d.asm.Sub).apply(null,arguments)},q$=d._Sum=function(){return(q$=d._Sum=d.asm.Sum).apply(null,arguments)},j$=d._Tan=function(){return(j$=d._Tan=d.asm.Tan).apply(null,arguments)},K$=d._Tanh=function(){return(K$=d._Tanh=d.asm.Tanh).apply(null,arguments)},X$=d._Tile=function(){return(X$=d._Tile=d.asm.Tile).apply(null,arguments)},Y$=d._TopK=function(){return(Y$=d._TopK=d.asm.TopK).apply(null,arguments)},Z$=d._Transform=function(){return(Z$=d._Transform=d.asm.Transform).apply(null,arguments)},J$=d._Transpose=function(){return(J$=d._Transpose=d.asm.Transpose).apply(null,arguments)},Q$=d.__FusedMatMul=function(){return(Q$=d.__FusedMatMul=d.asm._FusedMatMul).apply(null,arguments)},eD=d._malloc=function(){return(eD=d._malloc=d.asm.malloc).apply(null,arguments)},tD=d._free=function(){return(tD=d._free=d.asm.free).apply(null,arguments)},nD=d.__emscripten_tls_init=function(){return(nD=d.__emscripten_tls_init=d.asm._emscripten_tls_init).apply(null,arguments)},Jp=d._pthread_self=function(){return(Jp=d._pthread_self=d.asm.pthread_self).apply(null,arguments)},rD=d.___errno_location=function(){return(rD=d.___errno_location=d.asm.__errno_location).apply(null,arguments)},QI=d.__emscripten_thread_init=function(){return(QI=d.__emscripten_thread_init=d.asm._emscripten_thread_init).apply(null,arguments)},sD=d.__emscripten_thread_crashed=function(){return(sD=d.__emscripten_thread_crashed=d.asm._emscripten_thread_crashed).apply(null,arguments)},aD=d._emscripten_main_thread_process_queued_calls=function(){return(aD=d._emscripten_main_thread_process_queued_calls=d.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},oD=d._emscripten_main_browser_thread_id=function(){return(oD=d._emscripten_main_browser_thread_id=d.asm.emscripten_main_browser_thread_id).apply(null,arguments)},ek=d._emscripten_run_in_main_runtime_thread_js=function(){return(ek=d._emscripten_run_in_main_runtime_thread_js=d.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},iD=d._emscripten_dispatch_to_thread_=function(){return(iD=d._emscripten_dispatch_to_thread_=d.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},tk=d.__emscripten_proxy_execute_task_queue=function(){return(tk=d.__emscripten_proxy_execute_task_queue=d.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)},uy=d.__emscripten_thread_free_data=function(){return(uy=d.__emscripten_thread_free_data=d.asm._emscripten_thread_free_data).apply(null,arguments)},nk=d.__emscripten_thread_exit=function(){return(nk=d.__emscripten_thread_exit=d.asm._emscripten_thread_exit).apply(null,arguments)},rk=d._emscripten_stack_set_limits=function(){return(rk=d._emscripten_stack_set_limits=d.asm.emscripten_stack_set_limits).apply(null,arguments)},ly=d.stackSave=function(){return(ly=d.stackSave=d.asm.stackSave).apply(null,arguments)},Qp=d.stackRestore=function(){return(Qp=d.stackRestore=d.asm.stackRestore).apply(null,arguments)},eh=d.stackAlloc=function(){return(eh=d.stackAlloc=d.asm.stackAlloc).apply(null,arguments)},cD=d.dynCall_iijjiiii=function(){return(cD=d.dynCall_iijjiiii=d.asm.dynCall_iijjiiii).apply(null,arguments)},uD=d.dynCall_jiji=function(){return(uD=d.dynCall_jiji=d.asm.dynCall_jiji).apply(null,arguments)};d.keepRuntimeAlive=Sr,d.wasmMemory=ue,d.cwrap=ub,d.ExitStatus=Oa,d.PThread=$e;var th;xs=function P(){th||sk(),th||(xs=P)};function sk(P){if(P=P||b,js>0)return;if(C){h(d),Xt(),postMessage({cmd:"loaded"});return}if(Ki(),js>0)return;function K(){th||(th=!0,d.calledRun=!0,!ve&&(Xt(),h(d),d.onRuntimeInitialized&&d.onRuntimeInitialized(),Dp()))}d.setStatus?(d.setStatus("Running..."),setTimeout(function(){setTimeout(function(){d.setStatus("")},1),K()},1)):K()}if(d.preInit)for(typeof d.preInit=="function"&&(d.preInit=[d.preInit]);d.preInit.length>0;)d.preInit.pop()();sk();var nh;m&&(nh={uncaughtException:process.listeners("uncaughtException").filter(function(P){return!m.uncaughtException.indexOf(P)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(P){return!m.unhandledRejection.indexOf(P)>-1})});var rh;if(typeof WasmBackendModule!="undefined")rh=WasmBackendModule;else if(typeof s!="undefined")rh=s;else throw new Error("Could not find wasm module in post.js");if(nh){var lD=rh._dispose;rh._dispose=function(){lD(),nh.uncaughtException.forEach(function(P){process.removeListener("uncaughtException",P)}),nh.unhandledRejection.forEach(function(P){process.removeListener("unhandledRejection",P)})}}return s.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModuleThreadedSimd=n)}}),MD=Wt({"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js"(e,t){t.exports.wasmWorkerContents=`"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
+");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`}}),LD=Wt({"node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(e,t){var n=(()=>{var r=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(r=r||__filename),function(s){s=s||{};var a=typeof s!="undefined"?s:{},o,i;a.ready=new Promise(function(X,ae){o=X,i=ae});var c;typeof process!="undefined"&&process.listeners&&(c={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var u=Object.assign({},a),l=[],p="./this.program",d=(X,ae)=>{throw ae},h=typeof window=="object",f=typeof importScripts=="function",m=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",g="";function b(X){return a.locateFile?a.locateFile(X,g):g+X}var y,v,x,k;function S(X){if(X instanceof Xi)return;F("exiting due to exception: "+X)}if(m){f?g=_h().dirname(g)+"/":g=__dirname+"/";var C,E;typeof U1=="function"&&(C=Nv(),E=_h()),y=(X,ae)=>(X=E.normalize(X),C.readFileSync(X,ae?void 0:"utf8")),x=X=>{var ae=y(X,!0);return ae.buffer||(ae=new Uint8Array(ae)),ae},v=(X,ae,Ie)=>{X=E.normalize(X),C.readFile(X,function(Ge,wt){Ge?Ie(Ge):ae(wt.buffer)})},process.argv.length>1&&(p=process.argv[1].replace(/\\/g,"/")),l=process.argv.slice(2),process.on("uncaughtException",function(X){if(!(X instanceof Xi))throw X}),process.on("unhandledRejection",function(X){throw X}),d=(X,ae)=>{if(qn())throw process.exitCode=X,ae;S(ae),process.exit(X)},a.inspect=function(){return"[Emscripten Module object]"}}else(h||f)&&(f?g=self.location.href:typeof document!="undefined"&&document.currentScript&&(g=document.currentScript.src),r&&(g=r),g.indexOf("blob:")!==0?g=g.substr(0,g.replace(/[?#].*/,"").lastIndexOf("/")+1):g="",y=X=>{var ae=new XMLHttpRequest;return ae.open("GET",X,!1),ae.send(null),ae.responseText},f&&(x=X=>{var ae=new XMLHttpRequest;return ae.open("GET",X,!1),ae.responseType="arraybuffer",ae.send(null),new Uint8Array(ae.response)}),v=(X,ae,Ie)=>{var Ge=new XMLHttpRequest;Ge.open("GET",X,!0),Ge.responseType="arraybuffer",Ge.onload=()=>{if(Ge.status==200||Ge.status==0&&Ge.response){ae(Ge.response);return}Ie()},Ge.onerror=Ie,Ge.send(null)},k=X=>document.title=X);var $=a.print||console.log.bind(console),F=a.printErr||console.warn.bind(console);Object.assign(a,u),u=null,a.arguments&&(l=a.arguments),a.thisProgram&&(p=a.thisProgram),a.quit&&(d=a.quit);var A=4,R;a.wasmBinary&&(R=a.wasmBinary);var T=a.noExitRuntime||!0;typeof WebAssembly!="object"&&kr("no native wasm support detected");var L,V=!1,G;function j(X,ae){X||kr(ae)}var H=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function Z(X,ae,Ie){for(var Ge=ae+Ie,wt=ae;X[wt]&&!(wt>=Ge);)++wt;if(wt-ae>16&&X.buffer&&H)return H.decode(X.subarray(ae,wt));for(var It="";ae<wt;){var Ve=X[ae++];if(!(Ve&128)){It+=String.fromCharCode(Ve);continue}var Be=X[ae++]&63;if((Ve&224)==192){It+=String.fromCharCode((Ve&31)<<6|Be);continue}var Ot=X[ae++]&63;if((Ve&240)==224?Ve=(Ve&15)<<12|Be<<6|Ot:Ve=(Ve&7)<<18|Be<<12|Ot<<6|X[ae++]&63,Ve<65536)It+=String.fromCharCode(Ve);else{var ir=Ve-65536;It+=String.fromCharCode(55296|ir>>10,56320|ir&1023)}}return It}function J(X,ae){return X?Z(ne,X,ae):""}function ee(X,ae,Ie,Ge){if(!(Ge>0))return 0;for(var wt=Ie,It=Ie+Ge-1,Ve=0;Ve<X.length;++Ve){var Be=X.charCodeAt(Ve);if(Be>=55296&&Be<=57343){var Ot=X.charCodeAt(++Ve);Be=65536+((Be&1023)<<10)|Ot&1023}if(Be<=127){if(Ie>=It)break;ae[Ie++]=Be}else if(Be<=2047){if(Ie+1>=It)break;ae[Ie++]=192|Be>>6,ae[Ie++]=128|Be&63}else if(Be<=65535){if(Ie+2>=It)break;ae[Ie++]=224|Be>>12,ae[Ie++]=128|Be>>6&63,ae[Ie++]=128|Be&63}else{if(Ie+3>=It)break;ae[Ie++]=240|Be>>18,ae[Ie++]=128|Be>>12&63,ae[Ie++]=128|Be>>6&63,ae[Ie++]=128|Be&63}}return ae[Ie]=0,Ie-wt}function re(X,ae,Ie){return ee(X,ne,ae,Ie)}var te,ie,ne,le,ue,ye,ve,Se,Le;function je(X){te=X,a.HEAP8=ie=new Int8Array(X),a.HEAP16=le=new Int16Array(X),a.HEAP32=ye=new Int32Array(X),a.HEAPU8=ne=new Uint8Array(X),a.HEAPU16=ue=new Uint16Array(X),a.HEAPU32=ve=new Uint32Array(X),a.HEAPF32=Se=new Float32Array(X),a.HEAPF64=Le=new Float64Array(X)}var nt=a.INITIAL_MEMORY||16777216,rt,st=[],Ze=[],ze=[],dt=!1;function qn(){return T}function At(){if(a.preRun)for(typeof a.preRun=="function"&&(a.preRun=[a.preRun]);a.preRun.length;)Ir(a.preRun.shift());xs(st)}function wr(){dt=!0,xs(Ze)}function on(){if(a.postRun)for(typeof a.postRun=="function"&&(a.postRun=[a.postRun]);a.postRun.length;)jn(a.postRun.shift());xs(ze)}function Ir(X){st.unshift(X)}function vn(X){Ze.unshift(X)}function jn(X){ze.unshift(X)}var yt=0,Pn=null,hn=null;function ts(X){yt++,a.monitorRunDependencies&&a.monitorRunDependencies(yt)}function qs(X){if(yt--,a.monitorRunDependencies&&a.monitorRunDependencies(yt),yt==0&&(Pn!==null&&(clearInterval(Pn),Pn=null),hn)){var ae=hn;hn=null,ae()}}function kr(X){a.onAbort&&a.onAbort(X),X="Aborted("+X+")",F(X),V=!0,G=1,X+=". Build with -sASSERTIONS for more info.";var ae=new WebAssembly.RuntimeError(X);throw i(ae),ae}var xl="data:application/octet-stream;base64,";function Sr(X){return X.startsWith(xl)}function Ki(X){return X.startsWith("file://")}var Xt;Xt="tfjs-backend-wasm.wasm",Sr(Xt)||(Xt=b(Xt));function Dp(X){try{if(X==Xt&&R)return new Uint8Array(R);if(x)return x(X);throw"both async and sync fetching of the wasm failed"}catch(ae){kr(ae)}}function Ng(){if(!R&&(h||f)){if(typeof fetch=="function"&&!Ki(Xt))return fetch(Xt,{credentials:"same-origin"}).then(function(X){if(!X.ok)throw"failed to load wasm binary file at '"+Xt+"'";return X.arrayBuffer()}).catch(function(){return Dp(Xt)});if(v)return new Promise(function(X,ae){v(Xt,function(Ie){X(new Uint8Array(Ie))},ae)})}return Promise.resolve().then(function(){return Dp(Xt)})}function _g(){var X={env:kl,wasi_snapshot_preview1:kl};function ae(Ve,Be){var Ot=Ve.exports;a.asm=Ot,L=a.asm.memory,je(L.buffer),rt=a.asm.__indirect_function_table,vn(a.asm.__wasm_call_ctors),qs("wasm-instantiate")}ts("wasm-instantiate");function Ie(Ve){ae(Ve.instance)}function Ge(Ve){return Ng().then(function(Be){return WebAssembly.instantiate(Be,X)}).then(function(Be){return Be}).then(Ve,function(Be){F("failed to asynchronously prepare wasm: "+Be),kr(Be)})}function wt(){return!R&&typeof WebAssembly.instantiateStreaming=="function"&&!Sr(Xt)&&!Ki(Xt)&&!m&&typeof fetch=="function"?fetch(Xt,{credentials:"same-origin"}).then(function(Ve){var Be=WebAssembly.instantiateStreaming(Ve,X);return Be.then(Ie,function(Ot){return F("wasm streaming compile failed: "+Ot),F("falling back to ArrayBuffer instantiation"),Ge(Ie)})}):Ge(Ie)}if(a.instantiateWasm)try{var It=a.instantiateWasm(X,ae);return It}catch(Ve){F("Module.instantiateWasm callback failed with error: "+Ve),i(Ve)}return wt().catch(i),{}}var GI,js;function Xi(X){this.name="ExitStatus",this.message="Program terminated with exit("+X+")",this.status=X}function xs(X){for(;X.length>0;)X.shift()(a)}function HI(X){return X}function qI(X){var ae=/\b_Z[\w\d_]+/g;return X.replace(ae,function(Ie){var Ge=Ie;return Ie===Ge?Ie:Ge+" ["+Ie+"]"})}function Yi(){var X=new Error;if(!X.stack){try{throw new Error}catch(ae){X=ae}if(!X.stack)return"(no stack trace available)"}return X.stack.toString()}function Eg(X,ae){ie.set(X,ae)}function Fp(){kr("")}function wl(){return 2147483648}function xn(){return wl()}function Rp(X,ae,Ie){ne.copyWithin(X,ae,ae+Ie)}function Ag(X){try{return L.grow(X-te.byteLength+65535>>>16),je(L.buffer),1}catch(ae){}}function $g(X){var ae=ne.length;X=X>>>0;var Ie=wl();if(X>Ie)return!1;let Ge=(Ot,ir)=>Ot+(ir-Ot%ir)%ir;for(var wt=1;wt<=4;wt*=2){var It=ae*(1+.2/wt);It=Math.min(It,X+100663296);var Ve=Math.min(Ie,Ge(Math.max(X,It),65536)),Be=Ag(Ve);if(Be)return!0}return!1}var Dg={varargs:void 0,get:function(){Dg.varargs+=4;var X=ye[Dg.varargs-4>>2];return X},getStr:function(X){var ae=J(X);return ae}};function jI(X){return 52}function Fg(X,ae,Ie,Ge,wt){return 70}var Oa=[null,[],[]];function Rg(X,ae){var Ie=Oa[X];ae===0||ae===10?((X===1?$:F)(Z(Ie,0)),Ie.length=0):Ie.push(ae)}function Pg(X,ae,Ie,Ge){for(var wt=0,It=0;It<Ie;It++){var Ve=ve[ae>>2],Be=ve[ae+4>>2];ae+=8;for(var Ot=0;Ot<Be;Ot++)Rg(X,ne[Ve+Ot]);wt+=Be}return ve[Ge>>2]=wt,0}function Il(X){var ae=a["_"+X];return ae}function Pp(X,ae,Ie,Ge,wt){var It={string:wn=>{var Xs=0;if(wn!=null&&wn!==0){var Zp=(wn.length<<2)+1;Xs=Nl(Zp),re(wn,Xs,Zp)}return Xs},array:wn=>{var Xs=Nl(wn.length);return Eg(wn,Xs),Xs}};function Ve(wn){return ae==="string"?J(wn):ae==="boolean"?Boolean(wn):wn}var Be=Il(X),Ot=[],ir=0;if(Ge)for(var ws=0;ws<Ge.length;ws++){var Yp=It[Ie[ws]];Yp?(ir===0&&(ir=jp()),Ot[ws]=Yp(Ge[ws])):Ot[ws]=Ge[ws]}var _l=Be.apply(null,Ot);function cy(wn){return ir!==0&&Kp(ir),Ve(wn)}return _l=cy(_l),_l}function Op(X,ae,Ie,Ge){Ie=Ie||[];var wt=Ie.every(Ve=>Ve==="number"||Ve==="boolean"),It=ae!=="string";return It&&wt&&!Ge?Il(X):function(){return Pp(X,ae,Ie,arguments,Ge)}}var kl={abort:Fp,emscripten_get_heap_max:xn,emscripten_memcpy_big:Rp,emscripten_resize_heap:$g,fd_close:jI,fd_seek:Fg,fd_write:Pg},KI=_g(),Mp=a.___wasm_call_ctors=function(){return(Mp=a.___wasm_call_ctors=a.asm.__wasm_call_ctors).apply(null,arguments)},Og=a._init=function(){return(Og=a._init=a.asm.init).apply(null,arguments)},$e=a._init_with_threads_count=function(){return($e=a._init_with_threads_count=a.asm.init_with_threads_count).apply(null,arguments)},Sl=a._get_threads_count=function(){return(Sl=a._get_threads_count=a.asm.get_threads_count).apply(null,arguments)},Mg=a._register_tensor=function(){return(Mg=a._register_tensor=a.asm.register_tensor).apply(null,arguments)},XI=a._dispose_data=function(){return(XI=a._dispose_data=a.asm.dispose_data).apply(null,arguments)},YI=a._dispose=function(){return(YI=a._dispose=a.asm.dispose).apply(null,arguments)},Lg=a._Abs=function(){return(Lg=a._Abs=a.asm.Abs).apply(null,arguments)},Lp=a._Add=function(){return(Lp=a._Add=a.asm.Add).apply(null,arguments)},Zi=a._AddN=function(){return(Zi=a._AddN=a.asm.AddN).apply(null,arguments)},zg=a._All=function(){return(zg=a._All=a.asm.All).apply(null,arguments)},Bg=a._Any=function(){return(Bg=a._Any=a.asm.Any).apply(null,arguments)},ZI=a._ArgMax=function(){return(ZI=a._ArgMax=a.asm.ArgMax).apply(null,arguments)},Wg=a._AvgPool=function(){return(Wg=a._AvgPool=a.asm.AvgPool).apply(null,arguments)},Vg=a._BatchMatMul=function(){return(Vg=a._BatchMatMul=a.asm.BatchMatMul).apply(null,arguments)},Ug=a._Ceil=function(){return(Ug=a._Ceil=a.asm.Ceil).apply(null,arguments)},Gg=a._ClipByValue=function(){return(Gg=a._ClipByValue=a.asm.ClipByValue).apply(null,arguments)},zp=a._Conv2D=function(){return(zp=a._Conv2D=a.asm.Conv2D).apply(null,arguments)},Bp=a._Conv2DBackpropInput=function(){return(Bp=a._Conv2DBackpropInput=a.asm.Conv2DBackpropInput).apply(null,arguments)},Hg=a._Cos=function(){return(Hg=a._Cos=a.asm.Cos).apply(null,arguments)},qg=a._Cosh=function(){return(qg=a._Cosh=a.asm.Cosh).apply(null,arguments)},jg=a._CropAndResize=function(){return(jg=a._CropAndResize=a.asm.CropAndResize).apply(null,arguments)},Tl=a._Cumprod=function(){return(Tl=a._Cumprod=a.asm.Cumprod).apply(null,arguments)},Kg=a._Cumsum=function(){return(Kg=a._Cumsum=a.asm.Cumsum).apply(null,arguments)},Xg=a._DepthToSpace=function(){return(Xg=a._DepthToSpace=a.asm.DepthToSpace).apply(null,arguments)},Yg=a._DepthwiseConv2dNative=function(){return(Yg=a._DepthwiseConv2dNative=a.asm.DepthwiseConv2dNative).apply(null,arguments)},Ma=a._Elu=function(){return(Ma=a._Elu=a.asm.Elu).apply(null,arguments)},Zg=a._Equal=function(){return(Zg=a._Equal=a.asm.Equal).apply(null,arguments)},Jg=a._Exp=function(){return(Jg=a._Exp=a.asm.Exp).apply(null,arguments)},Wp=a._FlipLeftRight=function(){return(Wp=a._FlipLeftRight=a.asm.FlipLeftRight).apply(null,arguments)},Qg=a._Floor=function(){return(Qg=a._Floor=a.asm.Floor).apply(null,arguments)},Ji=a._FloorDiv=function(){return(Ji=a._FloorDiv=a.asm.FloorDiv).apply(null,arguments)},eb=a._FusedBatchNorm=function(){return(eb=a._FusedBatchNorm=a.asm.FusedBatchNorm).apply(null,arguments)},tb=a._FusedConv2D=function(){return(tb=a._FusedConv2D=a.asm.FusedConv2D).apply(null,arguments)},Ks=a._FusedDepthwiseConv2D=function(){return(Ks=a._FusedDepthwiseConv2D=a.asm.FusedDepthwiseConv2D).apply(null,arguments)},Cl=a._Gather=function(){return(Cl=a._Gather=a.asm.Gather).apply(null,arguments)},nb=a._GatherNd=function(){return(nb=a._GatherNd=a.asm.GatherNd).apply(null,arguments)},rb=a._Greater=function(){return(rb=a._Greater=a.asm.Greater).apply(null,arguments)},sb=a._GreaterEqual=function(){return(sb=a._GreaterEqual=a.asm.GreaterEqual).apply(null,arguments)},ab=a._LeakyRelu=function(){return(ab=a._LeakyRelu=a.asm.LeakyRelu).apply(null,arguments)},Vp=a._Less=function(){return(Vp=a._Less=a.asm.Less).apply(null,arguments)},Up=a._LessEqual=function(){return(Up=a._LessEqual=a.asm.LessEqual).apply(null,arguments)},ob=a._Log=function(){return(ob=a._Log=a.asm.Log).apply(null,arguments)},ib=a._LogicalAnd=function(){return(ib=a._LogicalAnd=a.asm.LogicalAnd).apply(null,arguments)},Gp=a._LogicalNot=function(){return(Gp=a._LogicalNot=a.asm.LogicalNot).apply(null,arguments)},Hp=a._LogicalOr=function(){return(Hp=a._LogicalOr=a.asm.LogicalOr).apply(null,arguments)},cb=a._LogicalXor=function(){return(cb=a._LogicalXor=a.asm.LogicalXor).apply(null,arguments)},ub=a._Max=function(){return(ub=a._Max=a.asm.Max).apply(null,arguments)},lb=a._MaxPool=function(){return(lb=a._MaxPool=a.asm.MaxPool).apply(null,arguments)},qp=a._Maximum=function(){return(qp=a._Maximum=a.asm.Maximum).apply(null,arguments)},JI=a._Mean=function(){return(JI=a._Mean=a.asm.Mean).apply(null,arguments)},db=a._Min=function(){return(db=a._Min=a.asm.Min).apply(null,arguments)},pb=a._Minimum=function(){return(pb=a._Minimum=a.asm.Minimum).apply(null,arguments)},hb=a._MirrorPad=function(){return(hb=a._MirrorPad=a.asm.MirrorPad).apply(null,arguments)},fb=a._Multiply=function(){return(fb=a._Multiply=a.asm.Multiply).apply(null,arguments)},mb=a._Neg=function(){return(mb=a._Neg=a.asm.Neg).apply(null,arguments)},gb=a._NonMaxSuppressionV3=function(){return(gb=a._NonMaxSuppressionV3=a.asm.NonMaxSuppressionV3).apply(null,arguments)},bb=a._NonMaxSuppressionV4=function(){return(bb=a._NonMaxSuppressionV4=a.asm.NonMaxSuppressionV4).apply(null,arguments)},yb=a._NonMaxSuppressionV5=function(){return(yb=a._NonMaxSuppressionV5=a.asm.NonMaxSuppressionV5).apply(null,arguments)},vb=a._NotEqual=function(){return(vb=a._NotEqual=a.asm.NotEqual).apply(null,arguments)},xb=a._OneHot=function(){return(xb=a._OneHot=a.asm.OneHot).apply(null,arguments)},wb=a._PadV2=function(){return(wb=a._PadV2=a.asm.PadV2).apply(null,arguments)},Ib=a._Pow=function(){return(Ib=a._Pow=a.asm.Pow).apply(null,arguments)},kb=a._Prelu=function(){return(kb=a._Prelu=a.asm.Prelu).apply(null,arguments)},Sb=a._Prod=function(){return(Sb=a._Prod=a.asm.Prod).apply(null,arguments)},Tb=a._RealDiv=function(){return(Tb=a._RealDiv=a.asm.RealDiv).apply(null,arguments)},Cb=a._Relu=function(){return(Cb=a._Relu=a.asm.Relu).apply(null,arguments)},Nb=a._Relu6=function(){return(Nb=a._Relu6=a.asm.Relu6).apply(null,arguments)},_b=a._ResizeBilinear=function(){return(_b=a._ResizeBilinear=a.asm.ResizeBilinear).apply(null,arguments)},Eb=a._ResizeNearestNeighbor=function(){return(Eb=a._ResizeNearestNeighbor=a.asm.ResizeNearestNeighbor).apply(null,arguments)},Ab=a._Reverse=function(){return(Ab=a._Reverse=a.asm.Reverse).apply(null,arguments)},$b=a._RotateWithOffset=function(){return($b=a._RotateWithOffset=a.asm.RotateWithOffset).apply(null,arguments)},Db=a._Round=function(){return(Db=a._Round=a.asm.Round).apply(null,arguments)},Fb=a._Rsqrt=function(){return(Fb=a._Rsqrt=a.asm.Rsqrt).apply(null,arguments)},Rb=a._ScatterNd=function(){return(Rb=a._ScatterNd=a.asm.ScatterNd).apply(null,arguments)},Pb=a._SelectV2=function(){return(Pb=a._SelectV2=a.asm.SelectV2).apply(null,arguments)},Ob=a._Sigmoid=function(){return(Ob=a._Sigmoid=a.asm.Sigmoid).apply(null,arguments)},Mb=a._Sin=function(){return(Mb=a._Sin=a.asm.Sin).apply(null,arguments)},Lb=a._Softmax=function(){return(Lb=a._Softmax=a.asm.Softmax).apply(null,arguments)},zb=a._SparseFillEmptyRows=function(){return(zb=a._SparseFillEmptyRows=a.asm.SparseFillEmptyRows).apply(null,arguments)},Bb=a._SparseReshape=function(){return(Bb=a._SparseReshape=a.asm.SparseReshape).apply(null,arguments)},Wb=a._SparseSegmentReduction=function(){return(Wb=a._SparseSegmentReduction=a.asm.SparseSegmentReduction).apply(null,arguments)},Vb=a._Sqrt=function(){return(Vb=a._Sqrt=a.asm.Sqrt).apply(null,arguments)},Ub=a._Square=function(){return(Ub=a._Square=a.asm.Square).apply(null,arguments)},Gb=a._SquaredDifference=function(){return(Gb=a._SquaredDifference=a.asm.SquaredDifference).apply(null,arguments)},Hb=a._Step=function(){return(Hb=a._Step=a.asm.Step).apply(null,arguments)},qb=a._StridedSlice=function(){return(qb=a._StridedSlice=a.asm.StridedSlice).apply(null,arguments)},jb=a._Sub=function(){return(jb=a._Sub=a.asm.Sub).apply(null,arguments)},Kb=a._Sum=function(){return(Kb=a._Sum=a.asm.Sum).apply(null,arguments)},Xb=a._Tan=function(){return(Xb=a._Tan=a.asm.Tan).apply(null,arguments)},Yb=a._Tanh=function(){return(Yb=a._Tanh=a.asm.Tanh).apply(null,arguments)},Zb=a._Tile=function(){return(Zb=a._Tile=a.asm.Tile).apply(null,arguments)},Jb=a._TopK=function(){return(Jb=a._TopK=a.asm.TopK).apply(null,arguments)},Qb=a._Transform=function(){return(Qb=a._Transform=a.asm.Transform).apply(null,arguments)},ey=a._Transpose=function(){return(ey=a._Transpose=a.asm.Transpose).apply(null,arguments)},ty=a.__FusedMatMul=function(){return(ty=a.__FusedMatMul=a.asm._FusedMatMul).apply(null,arguments)},ny=a._malloc=function(){return(ny=a._malloc=a.asm.malloc).apply(null,arguments)},ry=a._free=function(){return(ry=a._free=a.asm.free).apply(null,arguments)},sy=a.___errno_location=function(){return(sy=a.___errno_location=a.asm.__errno_location).apply(null,arguments)},jp=a.stackSave=function(){return(jp=a.stackSave=a.asm.stackSave).apply(null,arguments)},Kp=a.stackRestore=function(){return(Kp=a.stackRestore=a.asm.stackRestore).apply(null,arguments)},Nl=a.stackAlloc=function(){return(Nl=a.stackAlloc=a.asm.stackAlloc).apply(null,arguments)},ay=a.dynCall_iijjiiii=function(){return(ay=a.dynCall_iijjiiii=a.asm.dynCall_iijjiiii).apply(null,arguments)},oy=a.dynCall_jiji=function(){return(oy=a.dynCall_jiji=a.asm.dynCall_jiji).apply(null,arguments)};a.cwrap=Op;var Qi;hn=function X(){Qi||Xp(),Qi||(hn=X)};function Xp(X){if(X=X||l,yt>0||(At(),yt>0))return;function ae(){Qi||(Qi=!0,a.calledRun=!0,!V&&(wr(),o(a),a.onRuntimeInitialized&&a.onRuntimeInitialized(),on()))}a.setStatus?(a.setStatus("Running..."),setTimeout(function(){setTimeout(function(){a.setStatus("")},1),ae()},1)):ae()}if(a.preInit)for(typeof a.preInit=="function"&&(a.preInit=[a.preInit]);a.preInit.length>0;)a.preInit.pop()();Xp();var ec;c&&(ec={uncaughtException:process.listeners("uncaughtException").filter(function(X){return!c.uncaughtException.indexOf(X)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(X){return!c.unhandledRejection.indexOf(X)>-1})});var tc;if(typeof s!="undefined")tc=s;else if(typeof WasmBackendModuleThreadedSimd!="undefined")tc=WasmBackendModuleThreadedSimd;else throw new Error("Could not find wasm module in post.js");if(ec){var iy=tc._dispose;tc._dispose=function(){iy(),ec.uncaughtException.forEach(function(X){process.removeListener("uncaughtException",X)}),ec.unhandledRejection.forEach(function(X){process.removeListener("unhandledRejection",X)})}}return s.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=n:typeof define=="function"&&define.amd?define([],function(){return n}):typeof e=="object"&&(e.WasmBackendModule=n)}}),zD=1e-7,BD=1e-4,uf=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},ld=class{refCount(e){return Xn("refCount")}incRef(e){return Xn("incRef")}timerAvailable(){return!0}time(e){return Xn("time")}read(e){return Xn("read")}readSync(e){return Xn("readSync")}readToGPU(e,t){return Xn("readToGPU")}numDataIds(){return Xn("numDataIds")}disposeData(e,t){return Xn("disposeData")}write(e,t,n){return Xn("write")}move(e,t,n,r,s){return Xn("move")}createTensorFromTexture(e,t,n){return Xn("createTensorFromTexture")}memory(){return Xn("memory")}floatPrecision(){return Xn("floatPrecision")}epsilon(){return this.floatPrecision()===32?zD:BD}dispose(){return Xn("dispose")}};function Xn(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function H1(e){let t=e.length,n=0;for(;t>0;)n=Math.random()*t|0,t--,Eh(e,t,n)}function WD(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let n=e.length,r=0;for(;n>0;)r=Math.random()*n|0,n--,Eh(e,n,r),Eh(t,n,r)}function Kl(e,t,n){return Math.max(e,Math.min(t,n))}function VD(e){return e%2===0?e:e+1}function Eh(e,t,n){let r=e[t];e[t]=e[n],e[n]=r}function UD(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function GD(e,t){let n=Math.random();return t*n+(1-n)*e}function HD(e,t){let n=0;for(let r=0;r<e.length;r++){let s=Number(e[r])-Number(t[r]);n+=s*s}return n}function D(e,t){if(!e)throw new Error(typeof t=="string"?t:t())}function _n(e,t,n=""){D(ga(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function go(e){D(e!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Qa(e,t=[],n=!1){if(t==null&&(t=[]),Array.isArray(e)||fn(e)&&!n)for(let r=0;r<e.length;++r)Qa(e[r],t,n);else t.push(e);return t}function ht(e){if(e.length===0)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function qD(e){return e.length===0}function ga(e,t){if(e===t)return!0;if(e==null||t==null||e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function hc(e){return e%1===0}function jD(e){if(Math.tanh!=null)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{let t=Math.exp(2*e);return(t-1)/(t+1)}}function KD(e){let t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function XD(e){let t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return H1(t),t}function Gl(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function YD(e,t=s=>0,n,r){return new Promise((s,a)=>{let o=0,i=()=>{if(e()){s();return}o++;let c=t(o);if(n!=null&&o>=n){a();return}r!=null?r(i,c):setTimeout(i,c)};i()})}function ZD(e,t){let n=1,r=-1;for(let a=0;a<e.length;++a)if(e[a]>=0)n*=e[a];else if(e[a]===-1){if(r!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${r} and dim ${a}`);r=a}else if(e[a]<0)throw Error(`Shapes can not be < 0. Found ${e[a]} at dim ${a}`);if(r===-1){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(n===0)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);let s=e.slice();return s[r]=t/n,s}function Fr(e,t){let n=t.length;return e=e==null?t.map((r,s)=>s):[].concat(e),D(e.every(r=>r>=-n&&r<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),D(e.every(r=>hc(r)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(r=>r<0?n+r:r)}function q1(e,t){let n=[],r=[],s=t!=null&&Array.isArray(t)&&t.length===0,a=t==null||s?null:Fr(t,e).sort(),o=0;for(let i=0;i<e.length;++i){if(a!=null){if(a[o]===i&&e[i]!==1)throw new Error(`Can't squeeze axis ${i} since its dim '${e[i]}' is not 1`);(a[o]==null||a[o]>i)&&e[i]===1&&(n.push(e[i]),r.push(i)),a[o]<=i&&o++}e[i]!==1&&(n.push(e[i]),r.push(i))}return{newShape:n,keptDims:r}}function j1(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else throw new Error(`Unknown data type ${e}`);return n}function K1(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else if(e==="string")n=new Array(t);else throw new Error(`Unknown data type ${e}`);return n}function X1(e,t){for(let n=0;n<e.length;n++){let r=e[n];if(isNaN(r)||!isFinite(r))throw Error(`A tensor of type ${t} being uploaded contains ${r}.`)}}function Y1(e){return e==="bool"||e==="complex64"||e==="float32"||e==="int32"||e==="string"}function JD(e,t){return!(t==="complex64"||t==="float32"&&e!=="complex64"||t==="int32"&&e!=="float32"&&e!=="complex64"||t==="bool"&&e==="bool")}function fn(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}function Cy(e){if(e==="float32"||e==="int32")return 4;if(e==="complex64")return 8;if(e==="bool")return 1;throw new Error(`Unknown dtype ${e}`)}function Z1(e){if(e==null)return 0;let t=0;return e.forEach(n=>t+=n.length),t}function ea(e){return typeof e=="string"||e instanceof String}function J1(e){return typeof e=="boolean"}function Q1(e){return typeof e=="number"}function lf(e){return Array.isArray(e)?lf(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":Q1(e)?"float32":ea(e)?"string":J1(e)?"bool":"float32"}function oa(e){return!!(e&&e.constructor&&e.call&&e.apply)}function Ah(e,t){for(let n=t;n<e;++n)if(e%n===0)return n;return e}function Dc(e){let t=e.length;if(t<2)return[];let n=new Array(t-1);n[t-2]=e[t-1];for(let r=t-3;r>=0;--r)n[r]=n[r+1]*e[r+1];return n}function eS(e,t,n,r=!1){let s=new Array;if(t.length===1){let a=t[0]*(r?2:1);for(let o=0;o<a;o++)s[o]=n[e+o]}else{let a=t[0],o=t.slice(1),i=o.reduce((c,u)=>c*u)*(r?2:1);for(let c=0;c<a;c++)s[c]=eS(e+c*i,o,n,r)}return s}function uc(e,t,n=!1){if(e.length===0)return t[0];let r=e.reduce((s,a)=>s*a)*(n?2:1);if(r===0)return[];if(r!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);return eS(0,e,t,n)}function _v(e,t){let n=df(e,t);for(let r=0;r<n.length;r++)n[r]=1;return n}function df(e,t){if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool")return new Uint8Array(e);throw new Error(`Unknown data type ${t}`)}function QD(e,t){let n=e.reduce((r,s)=>r*s,1);if(t==null||t==="float32")return uc(e,new Float32Array(n));if(t==="int32")return uc(e,new Int32Array(n));if(t==="bool")return uc(e,new Uint8Array(n));throw new Error(`Unknown data type ${t}`)}function Ev(e){e.forEach(t=>{D(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function eF(e,t,n){if(t===0)return 0;if(t===1)return e[0];let r=e[e.length-1];for(let s=0;s<e.length-1;++s)r+=n[s]*e[s];return r}function tF(e,t,n){if(t===0)return[];if(t===1)return[e];let r=new Array(t);for(let s=0;s<r.length-1;++s)r[s]=Math.floor(e/n[s]),e-=r[s]*n[s];return r[r.length-1]=e,r}function Av(e){return e&&e.then&&typeof e.then=="function"}var ik="tfjsflags",tS=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=nF,this.populateURLFlags()}setPlatform(e,t){this.platform!=null&&(q().getBool("IS_TEST")||q().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},this.urlFlags[e]!=null){let r=this.urlFlags[e];q().getBool("IS_TEST")||q().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${r}.`),this.set(e,r)}}async getAsync(e){return e in this.flags?this.flags[e]:(this.flags[e]=await this.evaluateFlag(e),this.flags[e])}get(e){if(e in this.flags)return this.flags[e];let t=this.evaluateFlag(e);if(Av(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(this.flagRegistry[e]==null)throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,this.flagRegistry[e].setHook!=null&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(this.flagRegistry[e]==null)throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);return this.flagRegistry[e].evaluationFn()}setFlags(e){this.flags=Object.assign({},e)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global=="undefined"||typeof this.global.location=="undefined"||typeof this.global.location.search=="undefined")return;let e=this.getQueryParams(this.global.location.search);ik in e&&e[ik].split(",").forEach(n=>{let[r,s]=n.split(":");this.urlFlags[r]=sF(r,s)})}};function nF(e){let t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(n,...r)=>(rF(t,r[0],r[1]),r.join("="))),t}function rF(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}function sF(e,t){if(t=t.toLowerCase(),t==="true"||t==="false")return t==="true";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)}function q(){return $v}var $v=null;function aF(e){$v=e}var hy;function nS(){if(hy==null){let e;if(typeof window!="undefined")e=window;else if(typeof global!="undefined")e=global;else if(typeof process!="undefined")e=process;else if(typeof self!="undefined")e=self;else throw new Error("Could not find a global object");hy=e}return hy}function oF(){let e=nS();return e._tfGlobals==null&&(e._tfGlobals=new Map),e._tfGlobals}function Dv(e,t){let n=oF();if(n.has(e))return n.get(e);{let r=t();return n.set(e,r),n.get(e)}}var Fc="Abs",Rc="Acos",Pc="Acosh",ba="Add",bo="AddN",Oc="All",Mc="Any",yo="ArgMax",dd="ArgMin",Lc="Asin",zc="Asinh",Bc="Atan",Wc="Atanh",Vc="Atan2",vo="AvgPool",pf="AvgPoolGrad",pd="AvgPool3D",hf="AvgPool3DGrad",xo="BatchMatMul",Uc="BatchToSpaceND",ff="Bincount",rS="BroadcastTo",mf="BroadcastArgs",wo="Cast",Io="Ceil",ya="ClipByValue",gf="Complex",hd="ComplexAbs",Gc="Concat",ko="Conv2D",bf="Conv2DBackpropFilter",So="Conv2DBackpropInput",fd="Conv3D",yf="Conv3DBackpropFilterV2",vf="Conv3DBackpropInputV2",To="Cos",Co="Cosh",Hc="Cumprod",No="Cumsum",qc="CropAndResize",xf="DenseBincount",jc="DepthToSpace",_o="DepthwiseConv2dNative",wf="DepthwiseConv2dNativeBackpropFilter",If="DepthwiseConv2dNativeBackpropInput",kf="Diag",md="Dilation2D",$h="Dilation2DBackpropInput",Dh="Dilation2DBackpropFilter",Eo="RealDiv",Sf="Einsum",Ao="Elu",Tf="EluGrad",Kc="Erf",Xc="Equal",$o="Exp",Yc="ExpandDims",Zc="Expm1",Cf="FFT",gd="Fill",Jc="FlipLeftRight",Do="Floor",Fo="FloorDiv",Ro="FusedBatchNorm",Qc="GatherV2",eu="GatherNd",tu="Greater",Po="GreaterEqual",Oo="Identity",Nf="IFFT",_f="Imag",nu="IsFinite",ru="IsInf",su="IsNan",Mo="LeakyRelu",au="Less",ou="LessEqual",Ef="LinSpace",Lo="Log",iu="Log1p",cu="LogicalAnd",uu="LogicalNot",lu="LogicalOr",sS="LogicalXor",aS="LogSoftmax",iF="LowerBound",bd="LRN",Af="LRNGrad",zo="Max",Bo="Maximum",Wo="MaxPool",$f="MaxPoolGrad",yd="MaxPool3D",Df="MaxPool3DGrad",Ff="MaxPoolWithArgmax",Vo="Mean",Uo="Min",Go="Minimum",Ho="MirrorPad",du="Mod",Rf="Multinomial",qo="Multiply",pu="Neg",hu="NotEqual",fu="NonMaxSuppressionV3",mu="NonMaxSuppressionV4",gu="NonMaxSuppressionV5",bu="OnesLike",jo="OneHot",yu="Pack",Ko="PadV2",cF="Pool",Xo="Pow",Yo="Prelu",Zo="Prod",Pf="RaggedGather",Of="RaggedRange",Mf="RaggedTensorToTensor",vd="Range",Lf="Real",vu="Reciprocal",Jo="Relu",xu="Reshape",Qo="ResizeNearestNeighbor",zf="ResizeNearestNeighborGrad",ei="ResizeBilinear",Bf="ResizeBilinearGrad",ti="Relu6",ni="Reverse",ri="Round",si="Rsqrt",wu="ScatterNd",Wf="SearchSorted",Iu="Select",ku="Selu",Su="Slice",ai="Sin",Tu="Sinh",Cu="Sign",oi="Sigmoid",Nu="Softplus",ii="Sqrt",ci="Sum",_u="SpaceToBatchND",Eu="SplitV",ui="Softmax",xd="SparseFillEmptyRows",Au="SparseReshape",wd="SparseSegmentMean",Id="SparseSegmentSum",Vf="SparseToDense",li="SquaredDifference",kd="Square",$u="StridedSlice",Sd="StringNGrams",Td="StringSplit",Cd="StringToHashBucketFast",di="Sub",pi="Tan",hi="Tanh",va="Tile",Du="TopK",Fu="Transform",_s="Transpose",Uf="Unique",Ru="Unpack",Nd="UnsortedSegmentSum",uF="UpperBound",Pu="ZerosLike",xa="Step",Fh="FromPixels",Ou="RotateWithOffset",eo="_FusedMatMul",to="FusedConv2D",no="FusedDepthwiseConv2D";function Qs(...e){q().getBool("IS_TEST")||q().getBool("PROD")||console.warn(...e)}function lF(...e){q().getBool("IS_TEST")||q().getBool("PROD")||console.log(...e)}var fc=Dv("kernelRegistry",()=>new Map),Xl=Dv("gradRegistry",()=>new Map);function Rh(e,t){let n=Fv(e,t);return fc.get(n)}function Ny(e){return Xl.get(e)}function Ph(e){let t=fc.entries(),n=[];for(;;){let{done:r,value:s}=t.next();if(r)break;let[a,o]=s,[i]=a.split("_");i===e&&n.push(o)}return n}function _d(e){let{kernelName:t,backendName:n}=e,r=Fv(t,n);fc.has(r)&&Qs(`The kernel '${t}' for backend '${n}' is already registered`),fc.set(r,e)}function oS(e){let{kernelName:t}=e;Xl.has(t)&&q().getBool("DEBUG")&&Qs(`Overriding the gradient for '${t}'`),Xl.set(t,e)}function dF(e,t){let n=Fv(e,t);if(!fc.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);fc.delete(n)}function pF(e){if(!Xl.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);Xl.delete(e)}function hF(e,t){Ph(e).forEach(r=>{let s=Object.assign({},r,{backendName:t});_d(s)})}function Fv(e,t){return`${t}_${e}`}var w={};Ae(w,{arraysEqual:()=>ga,assert:()=>D,assertNonNegativeIntegerDimensions:()=>Ev,assertNonNull:()=>go,assertShapesMatch:()=>_n,bytesFromStringArray:()=>Z1,bytesPerElement:()=>Cy,checkConversionForErrors:()=>X1,clamp:()=>Kl,computeStrides:()=>Dc,createScalarValue:()=>vF,createShuffledIndices:()=>XD,decodeString:()=>Oh,distSquared:()=>HD,encodeString:()=>Ad,fetch:()=>wF,fingerPrint64:()=>yF,flatten:()=>Qa,getArrayFromDType:()=>K1,getTypedArrayFromDType:()=>j1,hasEncodingLoss:()=>JD,hexToLong:()=>Ed,indexToLoc:()=>tF,inferDtype:()=>lf,inferFromImplicitShape:()=>ZD,isBoolean:()=>J1,isFunction:()=>oa,isInt:()=>hc,isNumber:()=>Q1,isPromise:()=>Av,isScalarShape:()=>qD,isString:()=>ea,isTypedArray:()=>fn,isValidDtype:()=>Y1,locToIndex:()=>eF,makeOnesTypedArray:()=>_v,makeZerosNestedTypedArray:()=>QD,makeZerosTypedArray:()=>df,nearestDivisor:()=>Ah,nearestLargerEven:()=>VD,now:()=>Yl,parseAxisParam:()=>Fr,randUniform:()=>GD,repeatedTry:()=>YD,rightPad:()=>Gl,shuffle:()=>H1,shuffleCombo:()=>WD,sizeFromShape:()=>ht,sizeToSquarishShape:()=>KD,squeezeShape:()=>q1,sum:()=>UD,swap:()=>Eh,tanh:()=>jD,toNestedArray:()=>uc,toTypedArray:()=>Gf});var ck=ma(ID()),Va=ck.default||ck;function Ed(e){return Va.fromString(e,!0,16)}var iS=Ed("c3a5c85c97cb3127"),Ba=Ed("b492b66fbe98f273"),In=Ed("9ae16a3b2f90404f");function _y(e){return e.xor(e.shru(47))}function cS(e,t,n){let r=e.slice(t,t+n);return Va.fromBytes(Array.from(r),!0,!0)}function pt(e,t){return cS(e,t,8)}function uk(e,t){return cS(e,t,4)}function Yt(e,t){return t===0?e:e.shru(t).or(e.shl(64-t))}function ra(e,t,n=Ed("9ddfea08eb382d69")){let r=e.xor(t).mul(n);r=r.xor(r.shru(47));let s=t.xor(r).mul(n);return s=s.xor(s.shru(47)),s=s.mul(n),s}function fF(e,t,n,r,s,a){s=s.add(e),a=Yt(a.add(s).add(r),21);let o=s;return s=s.add(t),s=s.add(n),a=a.add(Yt(s,44)),[s.add(r),a.add(o)]}function ah(e,t,n,r){return fF(pt(e,t),pt(e,t+8),pt(e,t+16),pt(e,t+24),n,r)}function mF(e,t=e.length){if(t>=8){let n=In.add(t*2),r=pt(e,0).add(In),s=pt(e,t-8),a=Yt(s,37).mul(n).add(r),o=Yt(r,25).add(s).mul(n);return ra(a,o,n)}if(t>=4){let n=In.add(t*2),r=uk(e,0);return ra(r.shl(3).add(t),uk(e,t-4),n)}if(t>0){let n=e[0],r=e[t>>1],s=e[t-1],a=n+(r<<8),o=t+(s<<2);return _y(In.mul(a).xor(iS.mul(o))).mul(In)}return In}function gF(e,t=e.length){let n=In.add(t*2),r=pt(e,0).mul(Ba),s=pt(e,8),a=pt(e,t-8).mul(n),o=pt(e,t-16).mul(In);return ra(Yt(r.add(s),43).add(Yt(a,30)).add(o),r.add(Yt(s.add(In),18)).add(a),n)}function bF(e,t=e.length){let n=In.add(t*2),r=pt(e,0).mul(In),s=pt(e,8),a=pt(e,t-8).mul(n),o=pt(e,t-16).mul(In),i=Yt(r.add(s),43).add(Yt(a,30)).add(o),c=ra(i,r.add(Yt(s.add(In),18)).add(a),n),u=pt(e,16).mul(n),l=pt(e,24),p=i.add(pt(e,t-32)).mul(n),d=c.add(pt(e,t-24)).mul(n);return ra(Yt(u.add(l),43).add(Yt(p,30)).add(d),u.add(Yt(l.add(r),18)).add(p),n)}function yF(e,t=e.length){let n=Va.fromNumber(81,!0);if(t<=32)return t<=16?mF(e,t):gF(e,t);if(t<=64)return bF(e,t);let r=n,s=n.mul(Ba).add(113),a=_y(s.mul(In).add(113)).mul(In),o=[Va.UZERO,Va.UZERO],i=[Va.UZERO,Va.UZERO];r=r.mul(In).add(pt(e,0));let c=0,u=(t-1>>6)*64,l=u+(t-1&63)-63;do r=Yt(r.add(s).add(o[0]).add(pt(e,c+8)),37).mul(Ba),s=Yt(s.add(o[1]).add(pt(e,c+48)),42).mul(Ba),r=r.xor(i[1]),s=s.add(o[0]).add(pt(e,c+40)),a=Yt(a.add(i[0]),33).mul(Ba),o=ah(e,c,o[1].mul(Ba),r.add(i[0])),i=ah(e,c+32,a.add(i[1]),s.add(pt(e,c+16))),[a,r]=[r,a],c+=64;while(c!==u);let p=Ba.add(a.and(255).shl(1));return c=l,i[0]=i[0].add(t-1&63),o[0]=o[0].add(i[0]),i[0]=i[0].add(o[0]),r=Yt(r.add(s).add(o[0]).add(pt(e,c+8)),37).mul(p),s=Yt(s.add(o[1]).add(pt(e,c+48)),42).mul(p),r=r.xor(i[1].mul(9)),s=s.add(o[0].mul(9).add(pt(e,c+40))),a=Yt(a.add(i[0]),33).mul(p),o=ah(e,c,o[1].mul(p),r.add(i[0])),i=ah(e,c+32,a.add(i[1]),s.add(pt(e,c+16))),[a,r]=[r,a],ra(ra(o[0],i[0],p).add(_y(s).mul(iS)).add(a),ra(o[1],i[1],p).add(r),p)}function vF(e,t){return t==="string"?Ad(e):Gf([e],t)}function xF(e,t){return e instanceof Float32Array&&t==="float32"||e instanceof Int32Array&&t==="int32"||e instanceof Uint8Array&&t==="bool"}function Gf(e,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Qa(e)),q().getBool("DEBUG")&&X1(e,t),xF(e,t))return e;if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool"){let n=new Uint8Array(e.length);for(let r=0;r<n.length;++r)Math.round(e[r])!==0&&(n[r]=1);return n}else throw new Error(`Unknown data type ${t}`)}function Yl(){return q().platform.now()}function wF(e,t){return q().platform.fetch(e,t)}function Ad(e,t="utf-8"){return t=t||"utf-8",q().platform.encode(e,t)}function Oh(e,t="utf-8"){return t=t||"utf-8",q().platform.decode(e,t)}var IF=class{constructor(e,t){this.backendTimer=e,this.logger=t,t==null&&(this.logger=new SF)}profileKernel(e,t,n){let r,s=()=>{r=n()},a,o=Yl();if(this.backendTimer.timerAvailable())a=this.backendTimer.time(s);else{s();for(let c of r)c.dataSync();a=Promise.resolve({kernelMs:Yl()-o})}if(q().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let c=0;c<r.length;c++){let u=r[c];u.data().then(l=>{kF(l,u.dtype,e)})}return{kernelName:e,outputs:r,inputs:t,timeMs:a.then(c=>c.kernelMs),extraInfo:a.then(c=>c.getExtraProfileInfo!=null?c.getExtraProfileInfo():"")}}logKernelProfile(e){let{kernelName:t,outputs:n,timeMs:r,inputs:s,extraInfo:a}=e;n.forEach(o=>{Promise.all([o.data(),r,a]).then(i=>{this.logger.logKernelProfile(t,o,i[0],i[1],s,i[2])})})}};function kF(e,t,n){if(t!=="float32")return!1;for(let r=0;r<e.length;r++){let s=e[r];if(isNaN(s)||!isFinite(s))return console.warn(`Found ${s} in the result of '${n}'`),!0}return!1}var SF=class{logKernelProfile(e,t,n,r,s,a){let o=typeof r=="number"?Gl(`${r}ms`,9):r.error,i=Gl(e,25),c=t.rank,u=t.size,l=Gl(t.shape.toString(),14),p="";for(let d in s){let h=s[d];if(h!=null){let f=h.shape||t.shape,m=f.length;p+=`${d}: ${m}D ${m>0?f:""} `}}console.log(`%c${i}	%c${o}	%c${c}D ${l}	%c${u}	%c${p}	%c${a}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function TF(e,t,n){let r={},s={};for(let c=0;c<t.length;c++)r[t[c].id]=!0;for(let c=0;c<e.length;c++){let u=e[c],l=u.inputs;for(let p in l){let d=l[p],h=!1;for(let f=0;f<t.length;f++)if(r[d.id]){u.outputs.forEach(m=>r[m.id]=!0),h=!0,s[u.id]=!0;break}if(h)break}}let a={};a[n.id]=!0;let o={};for(let c=e.length-1;c>=0;c--){let u=e[c],l=u.inputs;for(let p=0;p<u.outputs.length;p++)if(a[u.outputs[p].id]){for(let d in l)a[l[d].id]=!0,o[u.id]=!0;break}}let i=[];for(let c=0;c<e.length;c++){let u=e[c];if(s[u.id]&&o[u.id]){let l={};for(let d in u.inputs){let h=u.inputs[d];r[h.id]&&(l[d]=h)}let p=Object.assign({},u);p.inputs=l,p.outputs=u.outputs,i.push(p)}}return i}function CF(e,t,n,r){for(let s=t.length-1;s>=0;s--){let a=t[s],o=[];if(a.outputs.forEach(c=>{let u=e[c.id];u!=null?o.push(u):o.push(null)}),a.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${a.kernelName}.`);let i=a.gradient(o);for(let c in a.inputs){if(!(c in i))throw new Error(`Cannot backprop through input ${c}. Available gradients found: ${Object.keys(i)}.`);let u=n(()=>i[c]());if(u.dtype!=="float32")throw new Error(`Error in gradient for op ${a.kernelName}. The gradient of input ${c} must have 'float32' dtype, but has '${u.dtype}'`);let l=a.inputs[c];if(!ga(u.shape,l.shape))throw new Error(`Error in gradient for op ${a.kernelName}. The gradient of input '${c}' has shape '${u.shape}', which does not match the shape of the input '${l.shape}'`);if(e[l.id]==null)e[l.id]=u;else{let p=e[l.id];e[l.id]=r(p,u),p.dispose()}}}}var lk=20,El=3,fy=7;function NF(e,t,n,r){let s=Dc(t),a=_F(e,t,n,s),o=t.length,i=vh(e,t,n,s,a),c=["Tensor"];return r&&(c.push(`  dtype: ${n}`),c.push(`  rank: ${o}`),c.push(`  shape: [${t}]`),c.push("  values:")),c.push(i.map(u=>"    "+u).join(`
+`)),c.join(`
+`)}function _F(e,t,n,r){let s=ht(t),a=r[r.length-1],o=new Array(a).fill(0),i=t.length,c=n==="complex64"?Rl(e):e;if(i>1)for(let u=0;u<s/a;u++){let l=u*a;for(let p=0;p<a;p++)o[p]=Math.max(o[p],Fl(c[l+p],0,n).length)}return o}function Fl(e,t,n){let r;return Array.isArray(e)?r=`${parseFloat(e[0].toFixed(fy))} + ${parseFloat(e[1].toFixed(fy))}j`:ea(e)?r=`'${e}'`:n==="bool"?r=uS(e):r=parseFloat(e.toFixed(fy)).toString(),Gl(r,t)}function uS(e){return e===0?"false":"true"}function vh(e,t,n,r,s,a=!0){let o=n==="complex64"?2:1,i=t[0],c=t.length;if(c===0){if(n==="complex64"){let m=Rl(e);return[Fl(m[0],0,n)]}return n==="bool"?[uS(e[0])]:[e[0].toString()]}if(c===1){if(i>lk){let g=El*o,b=Array.from(e.slice(0,g)),y=Array.from(e.slice((i-El)*o,i*o));return n==="complex64"&&(b=Rl(b),y=Rl(y)),["["+b.map((v,x)=>Fl(v,s[x],n)).join(", ")+", ..., "+y.map((v,x)=>Fl(v,s[i-El+x],n)).join(", ")+"]"]}return["["+(n==="complex64"?Rl(e):Array.from(e)).map((g,b)=>Fl(g,s[b],n)).join(", ")+"]"]}let u=t.slice(1),l=r.slice(1),p=r[0]*o,d=[];if(i>lk){for(let m=0;m<El;m++){let g=m*p,b=g+p;d.push(...vh(e.slice(g,b),u,n,l,s,!1))}d.push("...");for(let m=i-El;m<i;m++){let g=m*p,b=g+p;d.push(...vh(e.slice(g,b),u,n,l,s,m===i-1))}}else for(let m=0;m<i;m++){let g=m*p,b=g+p;d.push(...vh(e.slice(g,b),u,n,l,s,m===i-1))}let h=c===2?",":"";d[0]="["+d[0]+h;for(let m=1;m<d.length-1;m++)d[m]=" "+d[m]+h;let f=`,
+`;for(let m=2;m<c;m++)f+=`
+`;return d[d.length-1]=" "+d[d.length-1]+"]"+(a?"":f),d}function Rl(e){let t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}var Ht=class{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=ht(e),n!=null){let r=n.length;D(r===this.size,()=>`Length of values '${r}' does not match the size inferred by the shape '${this.size}'.`)}if(t==="complex64")throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=n||K1(t,this.size),this.strides=Dc(e)}set(e,...t){t.length===0&&(t=[0]),D(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);let n=this.locToIndex(t);this.values[n]=e}get(...e){e.length===0&&(e=[0]);let t=0;for(let r of e){if(r<0||r>=this.shape[t]){let s=`Requested out of range element at ${e}.   Buffer shape=${this.shape}`;throw new Error(s)}t++}let n=e[e.length-1];for(let r=0;r<e.length-1;++r)n+=this.strides[r]*e[r];return this.values[n]}locToIndex(e){if(this.rank===0)return 0;if(this.rank===1)return e[0];let t=e[e.length-1];for(let n=0;n<e.length-1;++n)t+=this.strides[n]*e[n];return t}indexToLoc(e){if(this.rank===0)return[];if(this.rank===1)return[e];let t=new Array(this.shape.length);for(let n=0;n<t.length-1;++n)t[n]=Math.floor(e/this.strides[n]),e-=t[n]*this.strides[n];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return Lr().makeTensor(this.values,this.shape,this.dtype)}},Lr=null,oc=null,EF=null;function AF(e){Lr=e}function $F(e){oc=e}function DF(e){EF=e}var Te=class{constructor(e,t,n,r){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=ht(e),this.strides=Dc(e),this.dataId=n,this.id=r,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let e=await this.data();return oc.buffer(this.shape,this.dtype,e)}bufferSync(){return oc.buffer(this.shape,this.dtype,this.dataSync())}async array(){let e=await this.data();return uc(this.shape,e,this.dtype==="complex64")}arraySync(){return uc(this.shape,this.dataSync(),this.dtype==="complex64")}async data(){this.throwIfDisposed();let e=Lr().read(this.dataId);if(this.dtype==="string"){let t=await e;try{return t.map(n=>Oh(n))}catch(n){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),Lr().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();let e=Lr().readSync(this.dataId);if(this.dtype==="string")try{return e.map(t=>Oh(t))}catch(t){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();let e=await Lr().read(this.dataId);return this.dtype==="string"?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(Lr().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return oc.print(this,e)}clone(){return this.throwIfDisposed(),oc.clone(this)}toString(e=!1){let t=this.dataSync();return NF(t,this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),oc.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),Lr().makeVariable(this,e,t,n)}};Object.defineProperty(Te,Symbol.hasInstance,{value:e=>!!e&&e.data!=null&&e.dataSync!=null&&e.throwIfDisposed!=null});function Q(){return Dv("Tensor",()=>Te)}Q();var ia=class extends Te{constructor(e,t,n,r){super(e.shape,e.dtype,e.dataId,r),this.trainable=t,this.name=n}assign(e){if(e.dtype!==this.dtype)throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);if(!ga(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);Lr().disposeTensor(this),this.dataId=e.dataId,Lr().incRef(this,null)}dispose(){Lr().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(ia,Symbol.hasInstance,{value:e=>e instanceof Te&&e.assign!=null&&e.assign instanceof Function});var Ur={};Ae(Ur,{assertTypesMatch:()=>lS,getTensorsInContainer:()=>Rv,isTensorInList:()=>RF,makeTypesMatch:()=>_t});var Ey;(function(e){e.R0="R0",e.R1="R1",e.R2="R2",e.R3="R3",e.R4="R4",e.R5="R5",e.R6="R6"})(Ey||(Ey={}));var Ay;(function(e){e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64"})(Ay||(Ay={}));var $y;(function(e){e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64"})($y||($y={}));var Dy;(function(e){e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64"})(Dy||(Dy={}));var Fy;(function(e){e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64"})(Fy||(Fy={}));var FF={float32:Dy,int32:Ay,bool:$y,complex64:Fy};function hr(e,t){if(e==="string"||t==="string"){if(e==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return FF[e][t]}function Hf(e){return hr(e,"int32")}function _t(e,t){if(e.dtype===t.dtype)return[e,t];let n=hr(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function lS(e,t){D(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function RF(e,t){return t.some(n=>n.id===e.id)}function Rv(e){let t=[];return dS(e,t,new Set),t}function dS(e,t,n){if(e==null)return;if(e instanceof Te){t.push(e);return}if(!PF(e))return;let r=e;for(let s in r){let a=r[s];n.has(a)||(n.add(a),dS(a,t,n))}}function PF(e){return Array.isArray(e)||typeof e=="object"}function my(e){return e.kernelName!=null}var dk=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(let e in this.registeredVariables)this.registeredVariables[e].dispose()}},Zl=class{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new dk}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t];if(await this.initializeBackend(n).success){await this.setBackend(n);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry))if(e in this.registryFactory){let{asyncInit:t}=this.initializeBackend(e);if(t)return null}else return null;return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(Qs(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:n},!0)}async setBackend(e){if(this.registryFactory[e]==null)throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,this.registry[e]==null){this.backendInstance=null;let{success:t,asyncInit:n}=this.initializeBackend(e);if(!(n?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new IF(this.backendInstance),!0}setupRegisteredKernels(){Ph(this.backendName).forEach(t=>{t.setupFunc!=null&&t.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){Ph(e).forEach(n=>{n.disposeFunc!=null&&n.disposeFunc(this.registry[e])})}initializeBackend(e){let t=this.registryFactory[e];if(t==null)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{let n=t.factory();if(n&&!(n instanceof ld)&&typeof n.then=="function"){let r=++this.pendingBackendInitId,s=n.then(a=>r<this.pendingBackendInitId?!1:(this.registry[e]=a,this.pendingBackendInit=null,!0)).catch(a=>(r<this.pendingBackendInitId||(this.pendingBackendInit=null,Qs(`Initialization of backend ${e} failed`),Qs(a.stack||a.message)),!1));return this.pendingBackendInit=s,{success:s,asyncInit:!0}}else return this.registry[e]=n,{success:!0,asyncInit:!1}}catch(n){return Qs(`Initialization of backend ${e} failed`),Qs(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${e} backend not found in registry`);this.backendName===e&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,e in this.registry&&(this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e]),delete this.registryFactory[e],this.backendName===e&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t],{success:r,asyncInit:s}=this.initializeBackend(n);if(s||r)return{name:n,asyncInit:s}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){let n=this.state.tensorInfo.get(t),r=n.backend,s=this.readSync(t),a=r.refCount(t);r.disposeData(t,!0),n.backend=e,e.move(t,s,n.shape,n.dtype,a),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n=null;if(t==null){if(typeof e!="function")throw new Error("Please provide a function to tidy()");t=e}else{if(typeof e!="string"&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if(typeof t!="function")throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");n=e}let r;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(r),()=>(r=t(),r instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),r))}scopedRun(e,t,n){e();try{let r=n();return t(),r}catch(r){throw t(),r}}nextTensorId(){return Zl.nextTensorId++}nextVariableId(){return Zl.nextVariableId++}clone(e){let t=M.runKernel(Oo,{x:e}),n={x:e},r=a=>({x:()=>{let o="float32",i={x:a},c={dtype:o};return M.runKernel(wo,i,c)}}),s=[];return this.addTapeNode(this.state.activeScope.name,n,[t],r,s,{}),t}runKernel(e,t,n){if(this.backendName==null&&this.backend,!(Rh(e,this.backendName)!=null))throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){let r=this.backend.numDataIds(),s=0;n.forEach(i=>{s+=i.dtype==="complex64"?3:1});let a=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],o=r-t-s-a;if(o>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${o} data ids) after running '${e}'`)}runKernelFunc(e){let t,n=[],r=this.isTapeOn(),s=this.state.numBytes,a=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let o;this.backendName==null&&this.backend;let i,c=my(e)?e.kernelName:this.state.activeScope!=null?this.state.activeScope.name:"";if(my(e)){let{kernelName:h,inputs:f,attrs:m}=e;this.backendName==null&&this.backend;let g=Rh(h,this.backendName);D(g!=null,()=>`Cannot find registered kernel '${h}' for backend '${this.backendName}'`),o=()=>{let b=this.backend.numDataIds();i=g.kernelFunc({inputs:f,attrs:m,backend:this.backend});let y=Array.isArray(i)?i:[i];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(h,b,y);let v=y.map(x=>x.rank!=null?x:this.makeTensorFromTensorInfo(x));if(r){let x=this.getTensorsForGradient(h,f,v);n=this.saveTensorsForBackwardMode(x)}return v}}else{let{forwardFunc:h}=e,f=m=>{!r||(n=m.map(g=>this.keep(this.clone(g))))};o=()=>{let m=this.backend.numDataIds();i=this.tidy(()=>h(this.backend,f));let g=Array.isArray(i)?i:[i];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(c,m,g),g}}let{inputs:u,attrs:l}=e,p=my(e)?null:e.backwardsFunc,d;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?t=o():(d=this.profiler.profileKernel(c,u,()=>o()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(d),t=d.outputs)}),r&&this.addTapeNode(c,u,t,p,n,l),this.state.profiling&&this.state.activeProfile.kernels.push({name:c,bytesAdded:this.state.numBytes-s,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-a,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(u).map(h=>u[h]!=null?u[h].shape:null),outputShapes:t.map(h=>h.shape),kernelTimeMs:d.timeMs,extraInfo:d.extraInfo}),Array.isArray(i)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(n=>this.keep(this.clone(n)))}getTensorsForGradient(e,t,n){let r=Ny(e);if(r!=null){let s=r.inputsToSave||[],a=r.outputsToSave||[],o;r.saveAllInputs?(D(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),o=Object.keys(t).map(c=>t[c])):o=s.map(c=>t[c]);let i=n.filter((c,u)=>a[u]);return o.concat(i)}return[]}makeTensor(e,t,n,r){if(e==null)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",r=r||this.backend;let s=e;n==="string"&&ea(e[0])&&(s=e.map(i=>Ad(i)));let a=r.write(s,t,n),o=new Te(t,n,a,this.nextTensorId());if(this.trackTensor(o,r),n==="string"){let i=this.state.tensorInfo.get(a),c=Z1(s);this.state.numBytes+=c-i.bytes,i.bytes=c}return o}makeTensorFromDataId(e,t,n,r){n=n||"float32";let s={dataId:e,shape:t,dtype:n};return this.makeTensorFromTensorInfo(s,r)}makeTensorFromTensorInfo(e,t){let{dataId:n,shape:r,dtype:s}=e,a=new Te(r,s,n,this.nextTensorId());return this.trackTensor(a,t),a}makeVariable(e,t=!0,n,r){n=n||this.nextVariableId().toString(),r!=null&&r!==e.dtype&&(e=e.cast(r));let s=new ia(e,t,n,this.nextTensorId());if(this.state.registeredVariables[s.name]!=null)throw new Error(`Variable with name ${s.name} was already registered`);return this.state.registeredVariables[s.name]=s,this.incRef(s,this.backend),s}trackTensor(e,t){this.state.numTensors++,e.dtype==="string"&&this.state.numStringTensors++;let n=0;e.dtype!=="complex64"&&e.dtype!=="string"&&(n=e.size*Cy(e.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:n})),e instanceof ia||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;let t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,e.dtype==="string"&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),e.dtype!=="complex64"&&e.dtype!=="string"){let n=e.size*Cy(e.dtype);this.state.numBytes-=n}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(let e in this.state.registeredVariables){let t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),this.state.registeredVariables[e.name]!=null&&delete this.state.registeredVariables[e.name]}memory(){let e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,e.reasons==null&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;let t=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await e(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(r=>r.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let r of this.state.activeProfile.kernels)r.kernelTimeMs=await r.kernelTimeMs,r.extraInfo=await r.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(e,t,n,r,s,a){let o={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:s},i=Ny(e);i!=null&&(r=i.gradFunc),r!=null&&(o.gradient=c=>(c=c.map((u,l)=>{if(u==null){let p=n[l],d=df(p.size,p.dtype);return this.makeTensor(d,p.shape,p.dtype)}return u}),r(c.length>1?c:c[0],s,a))),this.state.activeTape.push(o)}keep(e){return e.kept=!0,e}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){let t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){let t=Rv(e),n=new Set(t.map(s=>s.id));for(let s=0;s<this.state.activeScope.track.length;s++){let a=this.state.activeScope.track[s];!a.kept&&!n.has(a.id)&&a.dispose()}let r=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(s=>{!s.kept&&s.scopeId===r.id&&this.track(s)})}gradients(e,t,n,r=!1){if(D(t.length>0,()=>"gradients() received an empty list of xs."),n!=null&&n.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let s=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));D(s instanceof Te,()=>"The result y returned by f() must be a tensor.");let a=TF(this.state.activeTape,t,s);if(!r&&a.length===0&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let o={};o[s.id]=n==null?OF(s.shape):n,CF(o,a,c=>this.tidy(c),MF);let i=t.map(c=>o[c.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(c=>{for(let u of c.saved)u.dispose()}),this.state.activeTape=null),{value:s,grads:i}})}customGrad(e){return D(oa(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{D(t.every(o=>o instanceof Te),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n,r={};t.forEach((o,i)=>{r[i]=o});let s=(o,i)=>(n=e(...t,i),D(n.value instanceof Te,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),D(oa(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),a=(o,i)=>{let c=n.gradFunc(o,i),u=Array.isArray(c)?c:[c];D(u.length===t.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),D(u.every(p=>p instanceof Te),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let l={};return u.forEach((p,d)=>{l[d]=()=>p}),l};return this.runKernelFunc({forwardFunc:s,backwardsFunc:a,inputs:r})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){let t=Yl(),n=await this.backend.time(e);return n.wallMs=Yl()-t,n}track(e){return this.state.activeScope!=null&&(e.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(e)),e}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new dk;for(let e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};Zl.nextTensorId=0;Zl.nextVariableId=0;function OF(e){let t=_v(ht(e),"float32");return M.makeTensor(t,e,"float32")}function pS(){let e=nS();if(e._tfengine==null){let t=new tS(e);e._tfengine=new Zl(t)}return aF(e._tfengine.ENV),AF(()=>e._tfengine),e._tfengine}var M=pS();function MF(e,t){let n={a:e,b:t};return M.runKernel(ba,n)}var $d={};Ae($d,{isBrowser:()=>hS,isMobile:()=>BF,mockIsMobile:()=>zF});function LF(){return typeof navigator!="undefined"&&navigator!=null}var Ry;function zF(e){Ry=e}function BF(e){if(Ry!==void 0)return Ry;if(e||LF()){if(e||(e=navigator),e.product==="ReactNative")return!0;let t=e.userAgent||e.vendor||(typeof window!="undefined"?window.opera:"");if(!t){let n=e;return n.userAgentData&&n.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function hS(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}var er=q();er.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")});er.registerFlag("IS_BROWSER",()=>hS());er.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");er.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));er.registerFlag("PROD",()=>!1);er.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>er.getBool("DEBUG"));er.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);er.registerFlag("IS_TEST",()=>!1);er.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0);er.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);er.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);er.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU",()=>!1);er.registerFlag("USE_SETTIMEOUTCUSTOM",()=>!1);function us(e,t){let n=e;if(fn(e))return t==="string"?[]:[e.length];if(typeof e=="object"&&"texture"in e){let s=e.channels||"RGBA";return[e.height,e.width*s.length]}if(!Array.isArray(e))return[];let r=[];for(;Array.isArray(n)||fn(n)&&t!=="string";)r.push(n.length),n=n[0];return Array.isArray(e)&&q().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&fS(e,r,[]),r}function fS(e,t,n){if(n=n||[],!Array.isArray(e)&&!fn(e)){D(t.length===0,()=>`Element arr[${n.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}D(t.length>0,()=>`Element arr[${n.join("][")}] should be a primitive, but is an array of ${e.length} elements`),D(e.length===t[0],()=>`Element arr[${n.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);let r=t.slice(1);for(let s=0;s<e.length;++s)fS(e[s],r,n.concat(s))}function pk(e,t,n,r){if(e!=="string_or_numeric"){if(e==null)throw new Error("Expected dtype cannot be null.");if(e!=="numeric"&&e!==t||e==="numeric"&&t==="string")throw new Error(`Argument '${n}' passed to '${r}' must be ${e} tensor, but got ${t} tensor`)}}function _(e,t,n,r="numeric"){if(e instanceof Te)return pk(r,e.dtype,t,n),e;let s=lf(e);if(s!=="string"&&["bool","int32","float32"].indexOf(r)>=0&&(s=r),pk(r,s,t,n),e==null||!fn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string"){let c=e==null?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${c}'`)}let a=us(e,s);!fn(e)&&!Array.isArray(e)&&(e=[e]);let i=s!=="string"?Gf(e,s):Qa(e,[],!0);return M.makeTensor(i,a,s)}function Jl(e,t,n,r="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((a,o)=>_(a,`${t}[${o}]`,n,r))}var Pv="__op";function z(e){let t=Object.keys(e);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let n=t[0],r=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1)),n=n+Pv;let s=(...a)=>{M.startScope(n);try{let o=r(...a);return Av(o)&&console.error("Cannot return a Promise inside of tidy."),M.endScope(o),o}catch(o){throw M.endScope(null),o}};return Object.defineProperty(s,"name",{value:n,configurable:!0}),s}function WF(e,t){let n=_(e,"real","complex"),r=_(t,"imag","complex");_n(n.shape,r.shape,`real and imag shapes, ${n.shape} and ${r.shape}, must match in call to tf.complex().`);let s={real:n,imag:r};return M.runKernel(gf,s)}var As=z({complex_:WF});function wa(e,t,n,r){if(r==null&&(r=lf(e)),r==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(typeof e=="object"&&"texture"in e){if(r!=="float32"&&r!=="int32")throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${r}.`);return e.channels=e.channels||"RGBA",M.backend.createTensorFromTexture(e,t||n,r)}if(!fn(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){Ev(t);let s=ht(t),a=ht(n);D(s===a,()=>`Based on the provided shape, [${t}], the tensor should have ${s} values but has ${a}`);for(let o=0;o<n.length;++o){let i=n[o],c=o===n.length-1?i!==ht(t.slice(o)):!0;D(n[o]===t[o]||!c,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return!fn(e)&&!Array.isArray(e)&&(e=[e]),t=t||n,e=r!=="string"?Gf(e,r):Qa(e,[],!0),M.makeTensor(e,t,r)}function Cn(e,t,n){let r=us(e,n);return wa(e,t,r,n)}var Py={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8},Mh=4;async function VF(e,t){let n=[],r=[],s=Array.isArray(e)?e.map(o=>o.name):Object.keys(e);for(let o=0;o<s.length;++o){let i=s[o],c=Array.isArray(e)?e[o].tensor:e[i];if(c.dtype!=="float32"&&c.dtype!=="int32"&&c.dtype!=="bool"&&c.dtype!=="string"&&c.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${i}': ${c.dtype}`);let u={name:i,shape:c.shape,dtype:c.dtype};if(c.dtype==="string"){let l=new Promise(async p=>{let d=await c.bytes(),h=d.reduce((g,b)=>g+b.length,0)+Mh*d.length,f=new Uint8Array(h),m=0;for(let g=0;g<d.length;g++){let b=d[g],y=new Uint8Array(new Uint32Array([b.length]).buffer);f.set(y,m),m+=Mh,f.set(b,m),m+=b.length}p(f)});r.push(l)}else r.push(c.data());t!=null&&(u.group=t),n.push(u)}let a=await Promise.all(r);return{data:UF(a),specs:n}}function mS(e,t){let n={},r,s=0;for(let a of t){let o=a.name,i=a.dtype,c=a.shape,u=ht(c),l;if("quantization"in a){let p=a.quantization;if(p.dtype==="uint8"||p.dtype==="uint16"){if(!("min"in p&&"scale"in p))throw new Error(`Weight ${a.name} with quantization ${p.dtype} doesn't have corresponding metadata min and scale.`)}else if(p.dtype==="float16"){if(i!=="float32")throw new Error(`Weight ${a.name} is quantized with ${p.dtype} which only supports weights of type float32 not ${i}.`)}else throw new Error(`Weight ${a.name} has unknown quantization dtype ${p.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let d=Py[p.dtype],h=e.slice(s,s+u*d),f=p.dtype==="uint8"?new Uint8Array(h):new Uint16Array(h);if(i==="float32")if(p.dtype==="uint8"||p.dtype==="uint16"){l=new Float32Array(f.length);for(let m=0;m<f.length;m++){let g=f[m];l[m]=g*p.scale+p.min}}else if(p.dtype==="float16")r===void 0&&(r=XF()),l=r(f);else throw new Error(`Unsupported quantization type ${p.dtype} for weight type float32.`);else if(i==="int32"){if(p.dtype!=="uint8"&&p.dtype!=="uint16")throw new Error(`Unsupported quantization type ${p.dtype} for weight type int32.`);l=new Int32Array(f.length);for(let m=0;m<f.length;m++){let g=f[m];l[m]=Math.round(g*p.scale+p.min)}}else throw new Error(`Unsupported dtype in weight '${o}': ${i}`);s+=u*d}else if(i==="string"){let p=ht(a.shape);l=[];for(let d=0;d<p;d++){let h=new Uint32Array(e.slice(s,s+Mh))[0];s+=Mh;let f=new Uint8Array(e.slice(s,s+h));l.push(f),s+=h}}else{let p=Py[i],d=e.slice(s,s+u*p);if(i==="float32")l=new Float32Array(d);else if(i==="int32")l=new Int32Array(d);else if(i==="bool")l=new Uint8Array(d);else if(i==="complex64"){l=new Float32Array(d);let h=new Float32Array(l.length/2),f=new Float32Array(l.length/2);for(let b=0;b<h.length;b++)h[b]=l[b*2],f[b]=l[b*2+1];let m=Cn(h,c,"float32"),g=Cn(f,c,"float32");n[o]=As(m,g),m.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${o}': ${i}`);s+=u*p}i!=="complex64"&&(n[o]=Cn(l,c,i))}return n}function UF(e){if(e===null)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0,n=[];e.forEach(a=>{if(t+=a.byteLength,n.push(a.byteLength===a.buffer.byteLength?a:new a.constructor(a)),!(a instanceof Float32Array||a instanceof Int32Array||a instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${a.constructor.name}`)});let r=new Uint8Array(t),s=0;return n.forEach(a=>{r.set(new Uint8Array(a.buffer),s),s+=a.byteLength}),r.buffer}var Ov=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function hk(e){return Ov?Buffer.byteLength(e):new Blob([e]).size}function GF(e){if(Ov)return Buffer.from(e).toString("base64");let t=new Uint8Array(e),n="";for(let r=0,s=t.length;r<s;r++)n+=String.fromCharCode(t[r]);return btoa(n)}function HF(e){if(Ov){let r=Buffer.from(e,"base64");return r.buffer.slice(r.byteOffset,r.byteOffset+r.byteLength)}let t=atob(e),n=new Uint8Array(t.length);for(let r=0;r<t.length;++r)n.set([t.charCodeAt(r)],r);return n.buffer}function Mv(e){if(e.length===1)return e[0];let t=0;e.forEach(s=>{t+=s.byteLength});let n=new Uint8Array(t),r=0;return e.forEach(s=>{n.set(new Uint8Array(s),r),r+=s.byteLength}),n.buffer}function fk(e){let t="/";for(e=e.trim();e.endsWith(t);)e=e.slice(0,e.length-1);let n=e.split(t);return n[n.length-1]}function gS(e,t){let n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return e.signature!=null&&(n.signature=e.signature),e.userDefinedMetadata!=null&&(n.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(n.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(n.initializerSignature=e.initializerSignature),e.trainingConfig!=null&&(n.trainingConfig=e.trainingConfig),n}function bS(e,t,n){let r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(e.trainingConfig!=null&&(r.trainingConfig=e.trainingConfig),e.weightsManifest!=null){if(!t)throw new Error("modelJSON has weightsManifest but weightSpecs is null");if(!n)throw new Error("modelJSON has weightsManifest but weightData is null");r.weightSpecs=t,r.weightData=n}return e.signature!=null&&(r.signature=e.signature),e.userDefinedMetadata!=null&&(r.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(r.modelInitializer=e.modelInitializer),e.initializerSignature!=null&&(r.initializerSignature=e.initializerSignature),r}async function Lv(e,t){let n,r;return e.weightsManifest!=null&&([n,r]=await t(e.weightsManifest)),bS(e,n,r)}function Dd(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:e.modelTopology==null?0:hk(JSON.stringify(e.modelTopology)),weightSpecsBytes:e.weightSpecs==null?0:hk(JSON.stringify(e.weightSpecs)),weightDataBytes:e.weightData==null?0:e.weightData.byteLength}}function yS(e){let t=[];for(let n of e)t.push(...n.weights);return t}function qF(){let e=n=>{let r=n<<13,s=0;for(;(r&8388608)===0;)s-=8388608,r<<=1;return r&=-8388609,s+=947912704,r|s},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let n=1024;n<2048;n++)t[n]=939524096+(n-1024<<13);return t}function jF(){let e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}function KF(){let e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}function XF(){let e=qF(),t=jF(),n=KF();return r=>{let s=new ArrayBuffer(4*r.length),a=new Uint32Array(s);for(let o=0;o<r.length;o++){let i=r[o],c=e[n[i>>10]+(i&1023)]+t[i>>10];a[o]=c}return new Float32Array(s)}}var $t=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return $t.instance==null&&($t.instance=new $t),$t.instance}static registerSaveRouter(e){$t.getInstance().saveRouters.push(e)}static registerLoadRouter(e){$t.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return $t.getHandlers(e,"save")}static getLoadHandlers(e,t){return $t.getHandlers(e,"load",t)}static getHandlers(e,t,n){let r=[];return(t==="load"?$t.getInstance().loadRouters:$t.getInstance().saveRouters).forEach(a=>{let o=a(e,n);o!==null&&r.push(o)}),r}},YF=e=>$t.registerSaveRouter(e),ZF=e=>$t.registerLoadRouter(e),JF=e=>$t.getSaveHandlers(e),QF=(e,t)=>$t.getLoadHandlers(e,t),Oy="tensorflowjs",My=1,qa="models_store",ta="model_info_store";function vS(){if(!q().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let e=typeof window=="undefined"?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(t==null)throw new Error("The current browser does not appear to support IndexedDB.");return t}function Ly(e){let t=e.result;t.createObjectStore(qa,{keyPath:"modelPath"}),t.createObjectStore(ta,{keyPath:"modelPath"})}var ro=class{constructor(e){if(this.indexedDB=vS(),e==null||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((n,r)=>{let s=this.indexedDB.open(Oy,My);s.onupgradeneeded=()=>Ly(s),s.onsuccess=()=>{let a=s.result;if(t==null){let o=a.transaction(qa,"readonly"),c=o.objectStore(qa).get(this.modelPath);c.onsuccess=()=>{if(c.result==null)return a.close(),r(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));n(c.result.modelArtifacts)},c.onerror=u=>(a.close(),r(c.error)),o.oncomplete=()=>a.close()}else{let o=Dd(t),i=a.transaction(ta,"readwrite"),c=i.objectStore(ta),u=c.put({modelPath:this.modelPath,modelArtifactsInfo:o}),l;u.onsuccess=()=>{l=a.transaction(qa,"readwrite");let d=l.objectStore(qa).put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:o});d.onsuccess=()=>n({modelArtifactsInfo:o}),d.onerror=h=>{c=i.objectStore(ta);let f=c.delete(this.modelPath);f.onsuccess=()=>(a.close(),r(d.error)),f.onerror=m=>(a.close(),r(d.error))}},u.onerror=p=>(a.close(),r(u.error)),i.oncomplete=()=>{l==null?a.close():l.oncomplete=()=>a.close()}}},s.onerror=a=>r(s.error)})}};ro.URL_SCHEME="indexeddb://";var xS=e=>q().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(ro.URL_SCHEME)?eR(e.slice(ro.URL_SCHEME.length)):null;$t.registerSaveRouter(xS);$t.registerLoadRouter(xS);function eR(e){return new ro(e)}function tR(e){return e.startsWith(ro.URL_SCHEME)?e.slice(ro.URL_SCHEME.length):e}var nR=class{constructor(){this.indexedDB=vS()}async listModels(){return new Promise((e,t)=>{let n=this.indexedDB.open(Oy,My);n.onupgradeneeded=()=>Ly(n),n.onsuccess=()=>{let r=n.result,s=r.transaction(ta,"readonly"),o=s.objectStore(ta).getAll();o.onsuccess=()=>{let i={};for(let c of o.result)i[c.modelPath]=c.modelArtifactsInfo;e(i)},o.onerror=i=>(r.close(),t(o.error)),s.oncomplete=()=>r.close()},n.onerror=r=>t(n.error)})}async removeModel(e){return e=tR(e),new Promise((t,n)=>{let r=this.indexedDB.open(Oy,My);r.onupgradeneeded=()=>Ly(r),r.onsuccess=()=>{let s=r.result,a=s.transaction(ta,"readwrite"),o=a.objectStore(ta),i=o.get(e),c;i.onsuccess=()=>{if(i.result==null)return s.close(),n(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{let u=o.delete(e),l=()=>{c=s.transaction(qa,"readwrite");let d=c.objectStore(qa).delete(e);d.onsuccess=()=>t(i.result.modelArtifactsInfo),d.onerror=h=>n(i.error)};u.onsuccess=l,u.onerror=p=>(l(),s.close(),n(i.error))}},i.onerror=u=>(s.close(),n(i.error)),a.oncomplete=()=>{c==null?s.close():c.oncomplete=()=>s.close()}},r.onerror=s=>n(r.error)})}},Cs="/",ic="tensorflowjs_models",wS="info",rR="model_topology",sR="weight_specs",aR="weight_data",oR="model_metadata";function IS(e){return{info:[ic,e,wS].join(Cs),topology:[ic,e,rR].join(Cs),weightSpecs:[ic,e,sR].join(Cs),weightData:[ic,e,aR].join(Cs),modelMetadata:[ic,e,oR].join(Cs)}}function kS(e){for(let t of Object.values(e))window.localStorage.removeItem(t)}function iR(e){let t=e.split(Cs);if(t.length<3)throw new Error(`Invalid key format: ${e}`);return t.slice(1,t.length-1).join(Cs)}function cR(e){return e.startsWith(so.URL_SCHEME)?e.slice(so.URL_SCHEME.length):e}var so=class{constructor(e){if(!q().getBool("IS_BROWSER")||typeof window=="undefined"||typeof window.localStorage=="undefined")throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,e==null||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=IS(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let t=JSON.stringify(e.modelTopology),n=JSON.stringify(e.weightSpecs),r=Dd(e);try{this.LS.setItem(this.keys.info,JSON.stringify(r)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,GF(e.weightData));let s={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:e.signature!=null?e.signature:void 0,userDefinedMetadata:e.userDefinedMetadata!=null?e.userDefinedMetadata:void 0,modelInitializer:e.modelInitializer!=null?e.modelInitializer:void 0,initializerSignature:e.initializerSignature!=null?e.initializerSignature:void 0,trainingConfig:e.trainingConfig!=null?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(s)),{modelArtifactsInfo:r}}catch(s){throw kS(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${r.modelTopologyBytes}, weightSpecsBytes=${r.weightSpecsBytes}, weightDataBytes=${r.weightDataBytes}.`)}}}async load(){let e=JSON.parse(this.LS.getItem(this.keys.info));if(e==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(e.modelTopologyType!=="JSON")throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let t={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(n==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=n;let r=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(r==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=r;let s=this.LS.getItem(this.keys.modelMetadata);if(s!=null){let o=JSON.parse(s);t.format=o.format,t.generatedBy=o.generatedBy,t.convertedBy=o.convertedBy,o.signature!=null&&(t.signature=o.signature),o.userDefinedMetadata!=null&&(t.userDefinedMetadata=o.userDefinedMetadata),o.modelInitializer!=null&&(t.modelInitializer=o.modelInitializer),o.initializerSignature!=null&&(t.initializerSignature=o.initializerSignature),o.trainingConfig!=null&&(t.trainingConfig=o.trainingConfig)}let a=this.LS.getItem(this.keys.weightData);if(a==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=HF(a),t}};so.URL_SCHEME="localstorage://";var SS=e=>q().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(so.URL_SCHEME)?uR(e.slice(so.URL_SCHEME.length)):null;$t.registerSaveRouter(SS);$t.registerLoadRouter(SS);function uR(e){return new so(e)}var lR=class{constructor(){D(q().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),D(typeof window=="undefined"||typeof window.localStorage!="undefined",()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let e={},t=ic+Cs,n=Cs+wS;for(let r=0;r<this.LS.length;++r){let s=this.LS.key(r);if(s.startsWith(t)&&s.endsWith(n)){let a=iR(s);e[a]=JSON.parse(this.LS.getItem(s))}}return e}async removeModel(e){e=cR(e);let t=IS(e);if(this.LS.getItem(t.info)==null)throw new Error(`Cannot find model at path '${e}'`);let n=JSON.parse(this.LS.getItem(t.info));return kS(t),n}},lc="://",On=class{constructor(){this.managers={}}static getInstance(){return On.instance==null&&(On.instance=new On),On.instance}static registerManager(e,t){D(e!=null,()=>"scheme must not be undefined or null."),e.endsWith(lc)&&(e=e.slice(0,e.indexOf(lc))),D(e.length>0,()=>"scheme must not be an empty string.");let n=On.getInstance();D(n.managers[e]==null,()=>`A model store manager is already registered for scheme '${e}'.`),n.managers[e]=t}static getManager(e){let t=On.getInstance().managers[e];if(t==null)throw new Error(`Cannot find model manager for scheme '${e}'`);return t}static getSchemes(){return Object.keys(On.getInstance().managers)}};function xh(e){if(e.indexOf(lc)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${On.getSchemes().join(",")}`);return{scheme:e.split(lc)[0],path:e.split(lc)[1]}}async function TS(e,t,n=!1){D(e!==t,()=>`Old path and new path are the same: '${e}'`);let r=$t.getLoadHandlers(e);D(r.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),D(r.length<2,()=>`Copying failed because more than one (${r.length}) load handlers for source URL ${e}.`);let s=r[0],a=$t.getSaveHandlers(t);D(a.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),D(a.length<2,()=>`Copying failed because more than one (${r.length}) save handlers for destination URL ${t}.`);let o=a[0],i=xh(e).scheme,c=xh(e).path,u=i===xh(e).scheme,l=await s.load();n&&u&&await On.getManager(i).removeModel(c);let p=await o.save(l);return n&&!u&&await On.getManager(i).removeModel(c),p.modelArtifactsInfo}async function dR(){let e=On.getSchemes(),t={};for(let n of e){let r=await On.getManager(n).listModels();for(let s in r){let a=n+lc+s;t[a]=r[s]}}return t}async function pR(e){let t=xh(e);return On.getManager(t.scheme).removeModel(t.path)}async function hR(e,t){return TS(e,t,!1)}async function fR(e,t){return TS(e,t,!0)}var mR=class{constructor(){this.messageName="setTimeoutCustom",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Browser's encoder only supports utf-8, but got ${t}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}setTimeoutCustom(e,t){if(typeof window=="undefined"||!q().getBool("USE_SETTIMEOUTCUSTOM")){setTimeout(e,t);return}this.functionRefs.push(e),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},"*")},t),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener("message",n=>{if(n.source===window&&n.data.name===this.messageName){n.stopPropagation();let r=this.functionRefs[n.data.index];r(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0)}},!0))}};if(q().get("IS_BROWSER")){q().setPlatform("browser",new mR);try{On.registerManager(so.URL_SCHEME,new lR)}catch(e){}try{On.registerManager(ro.URL_SCHEME,new nR)}catch(e){}}var gR={importFetch:()=>kD()},gy,bR=class{constructor(){this.util=SD(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return q().global.fetch!=null?q().global.fetch(e,t):(gy==null&&(gy=gR.importFetch()),gy(e,t))}now(){let e=process.hrtime();return e[0]*1e3+e[1]/1e6}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Node built-in encoder only supports utf-8, but got ${t}`);return this.textEncoder.encode(e)}decode(e,t){return e.length===0?"":new this.util.TextDecoder(t).decode(e)}};q().get("IS_NODE")&&!q().get("IS_BROWSER")&&q().setPlatform("node",new bR);function Me(e,t="float32",n){return t=t||"float32",Ev(e),new Ht(e,t,n)}function yR(e,t){let n=_(e,"x","cast");if(!Y1(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&n.dtype!=="string"||t!=="string"&&n.dtype==="string")throw new Error("Only strings can be casted to strings");let r={x:n},s={dtype:t};return M.runKernel(wo,r,s)}var ce=z({cast_:yR});function vR(e){let n={x:_(e,"x","clone","string_or_numeric")};return M.runKernel(Oo,n)}var is=z({clone_:vR});function zv(e,t=!1){console.log(e.toString(t))}pS();var xR={buffer:Me,cast:ce,clone:is,print:zv};$F(xR);var Ut={};Ae(Ut,{browserFiles:()=>NR,browserHTTPRequest:()=>DR,concatenateArrayBuffers:()=>Mv,copyModel:()=>hR,decodeWeights:()=>mS,encodeWeights:()=>VF,fromMemory:()=>RR,fromMemorySync:()=>AS,getLoadHandlers:()=>QF,getModelArtifactsForJSON:()=>Lv,getModelArtifactsForJSONSync:()=>bS,getModelArtifactsInfoForJSON:()=>Dd,getSaveHandlers:()=>JF,getWeightSpecs:()=>yS,http:()=>Wv,isHTTPScheme:()=>zy,listModels:()=>dR,loadWeights:()=>_R,moveModel:()=>fR,registerLoadRouter:()=>ZF,registerSaveRouter:()=>YF,removeModel:()=>pR,weightsLoaderFactory:()=>NS,withSaveHandler:()=>PR,withSaveHandlerSync:()=>OR});var wR="model",IR=".json",kR=".weights.bin";function mk(e){return new Promise(t=>setTimeout(t)).then(e)}var mc=class{constructor(e){if(!q().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(mc.URL_SCHEME)&&(e=e.slice(mc.URL_SCHEME.length)),(e==null||e.length===0)&&(e=wR),this.modelJsonFileName=e+IR,this.weightDataFileName=e+kR}async save(e){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");let t=window.URL.createObjectURL(new Blob([e.weightData],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let n=[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}],r=gS(e,n),s=window.URL.createObjectURL(new Blob([JSON.stringify(r)],{type:"application/json"})),a=this.modelJsonAnchor==null?document.createElement("a"):this.modelJsonAnchor;if(a.download=this.modelJsonFileName,a.href=s,await mk(()=>a.dispatchEvent(new MouseEvent("click"))),e.weightData!=null){let o=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;o.download=this.weightDataFileName,o.href=t,await mk(()=>o.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Dd(e)}}}};mc.URL_SCHEME="downloads://";var SR=class{constructor(e){if(e==null||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{let n=new FileReader;n.onload=r=>{let s=JSON.parse(r.target.result),a=s.modelTopology;if(a==null){t(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(s.weightsManifest==null){t(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){e({modelTopology:a});return}let i=Lv(s,c=>this.loadWeights(c));e(i)},n.onerror=r=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(e){let t=[],n=[];for(let a of e)t.push(...a.weights),n.push(...a.paths);let r=this.checkManifestAndWeightFiles(e),s=n.map(a=>this.loadWeightsFile(a,r[a]));return Promise.all(s).then(a=>[t,Mv(a)])}loadWeightsFile(e,t){return new Promise((n,r)=>{let s=new FileReader;s.onload=a=>{let o=a.target.result;n(o)},s.onerror=a=>r(`Failed to weights data from file of path '${e}'.`),s.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){let t=[],n=this.weightsFiles.map(s=>fk(s.name)),r={};for(let s of e)s.paths.forEach(a=>{let o=fk(a);if(t.indexOf(o)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${o}'`);if(t.push(o),n.indexOf(o)===-1)throw new Error(`Weight file with basename '${o}' is not provided.`);r[a]=this.weightsFiles[n.indexOf(o)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return r}},TR=e=>q().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(mc.URL_SCHEME)?CR(e.slice(mc.URL_SCHEME.length)):null;$t.registerSaveRouter(TR);function CR(e="model"){return new mc(e)}function NR(e){return new SR(e)}function gk(e,t,n,r){o(e),n=n==null?0:n,r=r==null?1:r,i(n,r);let s=0,a=c=>(c.then(u=>{let l=n+ ++s/e.length*(r-n);return t(l),u}),c);function o(c){D(c!=null&&Array.isArray(c)&&c.length>0,()=>"promises must be a none empty array")}function i(c,u){D(c>=0&&c<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${c}`),D(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${u}`),D(u>=c,()=>`startFraction must be no more than endFraction, but got startFraction ${c} and endFraction ${u}`)}return Promise.all(e.map(a))}async function CS(e,t){t==null&&(t={});let n=t.fetchFunc==null?q().platform.fetch:t.fetchFunc,r=e.map(p=>n(p,t.requestInit,{isBinary:!0})),s=0,a=.5,i=(t.onProgress==null?await Promise.all(r):await gk(r,t.onProgress,s,a)).map(p=>p.arrayBuffer()),c=.5,u=1;return t.onProgress==null?await Promise.all(i):await gk(i,t.onProgress,c,u)}async function _R(e,t="",n,r){return NS(o=>CS(o,{requestInit:r}))(e,t,n)}function NS(e){return async(t,n="",r)=>{let s=t.map(()=>!1),a={},o=r!=null?r.map(()=>!1):[],i=[];if(t.forEach((h,f)=>{let m=0;h.weights.forEach(g=>{let b="quantization"in g?g.quantization.dtype:g.dtype,y=Py[b]*ht(g.shape),v=()=>{s[f]=!0,a[f]==null&&(a[f]=[]),a[f].push({manifestEntry:g,groupOffset:m,sizeBytes:y})};r!=null?r.forEach((x,k)=>{x===g.name&&(v(),o[k]=!0)}):v(),i.push(g.name),m+=y})}),!o.every(h=>h)){let h=r.filter((f,m)=>!o[m]);throw new Error(`Could not find weights in manifest with names: ${h.join(", ")}. 
+Manifest JSON has weights with names: ${i.join(", ")}.`)}let c=s.reduce((h,f,m)=>(f&&h.push(m),h),[]),u=[];c.forEach(h=>{t[h].paths.forEach(f=>{let m=n+(n.endsWith("/")?"":"/")+f;u.push(m)})});let l=await e(u),p={},d=0;return c.forEach(h=>{let f=t[h].paths.length,m=0;for(let x=0;x<f;x++)m+=l[d+x].byteLength;let g=new ArrayBuffer(m),b=new Uint8Array(g),y=0;for(let x=0;x<f;x++){let k=new Uint8Array(l[d+x]);b.set(k,y),y+=k.byteLength}a[h].forEach(x=>{let k=g.slice(x.groupOffset,x.groupOffset+x.sizeBytes),S=mS(k,[x.manifestEntry]);for(let C in S)p[C]=S[C]}),d+=f}),p}}var ER="application/octet-stream",AR="application/json",Bv=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",t==null&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,this.weightUrlConverter=t.weightUrlConverter,t.fetchFunc!=null?(D(typeof t.fetchFunc=="function",()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=t.fetchFunc):this.fetch=q().platform.fetch,D(e!=null&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&D(e.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,t.requestInit!=null&&t.requestInit.body!=null)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{}}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;let n=[{paths:["./model.weights.bin"],weights:e.weightSpecs}],r=gS(e,n);t.body.append("model.json",new Blob([JSON.stringify(r)],{type:AR}),"model.json"),e.weightData!=null&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:ER}),"model.weights.bin");let s=await this.fetch(this.path,t);if(s.ok)return{modelArtifactsInfo:Dd(e),responses:[s]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${s.status}.`)}async load(){let e=await this.fetch(this.path,this.requestInit);if(!e.ok)throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);let t;try{t=await e.json()}catch(s){let a=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?a+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":a+=" Please make sure the server is serving valid JSON for this request.",new Error(a)}let n=t.modelTopology,r=t.weightsManifest;if(n==null&&r==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return Lv(t,s=>this.loadWeights(s))}async loadWeights(e){let t=Array.isArray(this.path)?this.path[1]:this.path,[n,r]=$R(t),s=this.weightPathPrefix||n,a=yS(e),o=[],i=[];for(let u of e)for(let l of u.paths)this.weightUrlConverter!=null?i.push(this.weightUrlConverter(l)):o.push(s+l+r);this.weightUrlConverter&&o.push(...await Promise.all(i));let c=await CS(o,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[a,Mv(c)]}};Bv.URL_SCHEME_REGEX=/^https?:\/\//;function $R(e){let t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),r=e.substring(0,t),s=n>t?e.substring(n):"";return[r+"/",s]}function zy(e){return e.match(Bv.URL_SCHEME_REGEX)!=null}var _S=(e,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let n=!0;if(Array.isArray(e)?n=e.every(r=>zy(r)):n=zy(e),n)return Wv(e,t)}return null};$t.registerSaveRouter(_S);$t.registerLoadRouter(_S);function Wv(e,t){return new Bv(e,t)}function DR(e,t){return Wv(e,t)}var by=class{constructor(e){this.modelArtifacts=e}load(){return this.modelArtifacts}},ES=class{constructor(e){this.saveHandler=e}save(e){return this.saveHandler(e)}},FR=class{constructor(e){e.load&&(this.load=()=>Promise.resolve(e.load())),e.save&&(this.save=t=>Promise.resolve(e.save(t)))}};function RR(e,t,n,r){let s=arguments;return new FR(AS(...s))}function AS(e,t,n,r){return arguments.length===1?e.modelTopology!=null||e.weightSpecs!=null?new by(e):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new by({modelTopology:e})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new by({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:r}))}function PR(e){return new ES(e)}function OR(e){return new ES(e)}var $S={};Ae($S,{confusionMatrix:()=>rP});function MR(e,t,n=!1,r=!1){let s=_(e,"a","matMul"),a=_(t,"b","matMul");[s,a]=_t(s,a);let o={a:s,b:a},i={transposeA:n,transposeB:r};return M.runKernel(xo,o,i)}var De=z({matMul_:MR});function LR(e,t,n=1,r=0,s="int32"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let o={indices:_(e,"indices","oneHot","int32")},i={dtype:s,depth:t,onValue:n,offValue:r};return M.runKernel(jo,o,i)}var gc=z({oneHot_:LR});function zR(){q().set("PROD",!0)}function BR(){q().set("DEBUG",!0)}function WR(){q().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function Vv(e){q().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}DF(Vv);function VR(){M.disposeVariables()}function Er(){return M}function Lh(){return M.memory()}function UR(e){return M.profile(e)}function O(e,t){return M.tidy(e,t)}function _e(e){Rv(e).forEach(n=>n.dispose())}function Jt(e){return M.keep(e)}function GR(e){return M.time(e)}function HR(e){return M.setBackend(e)}function qR(){return M.ready()}function jR(){return M.backendName}function KR(e){M.removeBackend(e)}function XR(e){return M.findBackend(e)}function YR(e){return M.findBackendFactory(e)}function qf(e,t,n=1){return M.registerBackend(e,t,n)}function DS(){return M.backend}function ZR(e,t){q().setPlatform(e,t)}function JR(e){let n={input:_(e,"input","imag")};return M.runKernel(_f,n)}var Fd=z({imag_:JR});function QR(e){let n={x:_(e,"x","neg")};return M.runKernel(pu,n)}var xt=z({neg_:QR});function eP(e){let n={input:_(e,"input","real")};return M.runKernel(Lf,n)}var bc=z({real_:eP});function tP(e,t,n){let r=_(e,"x","transpose");if(t==null&&(t=r.shape.map((o,i)=>i).reverse()),D(r.rank===t.length,()=>`Error in transpose: rank of input ${r.rank} must match length of perm ${t}.`),t.forEach(o=>{D(o>=0&&o<r.rank,()=>`All entries in 'perm' must be between 0 and ${r.rank-1} but got ${t}`)}),r.rank<=1)return r.clone();let s={x:r},a={perm:t};return r.dtype==="complex64"?O(()=>{let o=bc(r),i=Fd(r);return o=M.runKernel(_s,{x:o},a),i=M.runKernel(_s,{x:i},a),n&&(i=xt(i)),As(o,i)}):M.runKernel(_s,s,a)}var Ee=z({transpose_:tP});function nP(e,t,n){let r=_(e,"labels","confusionMatrix"),s=_(t,"predictions","confusionMatrix");D(n==null||n>0&&Number.isInteger(n),()=>`If provided, numClasses must be a positive integer, but got ${n}`),D(r.rank===1,()=>`Expected the rank of labels to be 1, but got ${r.rank}`),D(s.rank===1,()=>`Expected the rank of predictions to be 1, but got ${s.rank}`),D(r.shape[0]===s.shape[0],()=>`Mismatch in the number of examples: ${r.shape[0]} vs. ${s.shape[0]}. Labels and predictions should have the same number of elements.`),D(n>0&&Number.isInteger(n),()=>`numClasses is required to be a positive integer, but got ${n}`);let a=gc(ce(r,"int32"),n),o=gc(ce(s,"int32"),n),i=Ee(a),c=De(i,o);return ce(c,"int32")}var rP=z({confusionMatrix_:nP}),Mu={};Ae(Mu,{assertAndGetBroadcastShape:()=>ct,getBroadcastDims:()=>FS,getReductionAxes:()=>Bt});function FS(e,t){let n=e.length,r=[];for(let s=0;s<n;s++){let a=n-1-s,o=e[a]||1;(t[t.length-1-s]||1)>1&&o===1&&r.unshift(a)}return r}function Bt(e,t){let n=[];for(let r=0;r<t.length;r++){let s=e[e.length-r-1],a=t.length-r-1,o=t[a];(s==null||s===1&&o>1)&&n.unshift(a)}return n}function ct(e,t){let n=[],r=Math.max(e.length,t.length);for(let s=0;s<r;s++){let a=e[e.length-s-1];a==null&&(a=1);let o=t[t.length-s-1];if(o==null&&(o=1),a===1)n.unshift(o);else if(o===1)n.unshift(a);else if(a!==o){let i=`Operands could not be broadcast together with shapes ${e} and ${t}.`;throw Error(i)}else n.unshift(a)}return n}var fi={};Ae(fi,{fromPixels:()=>lP,fromPixelsAsync:()=>cP,toPixels:()=>uP});function Rd(e,t,n){if(go(e),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");let r=us(e,n);if(r.length!==3&&r.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return wa(e,t,r,n)}var La;function RS(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(e==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let n=!1,r=!1,s=!1,a=!1,o=!1,i=!1;if(e.data instanceof Uint8Array)n=!0;else if(typeof ImageData!="undefined"&&e instanceof ImageData)r=!0;else if(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)s=!0;else if(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)a=!0;else if(e.getContext!=null)o=!0;else if(typeof ImageBitmap!="undefined"&&e instanceof ImageBitmap)i=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${e.constructor.name}`);if(Rh(Fh,M.backendName)!=null){let f={pixels:e},m={numChannels:t};return M.runKernel(Fh,f,m)}let[u,l]=s?[e.videoWidth,e.videoHeight]:[e.width,e.height],p;if(o)p=e.getContext("2d").getImageData(0,0,u,l).data;else if(r||n)p=e.data;else if(a||s||i){if(La==null)if(typeof document=="undefined")if(typeof OffscreenCanvas!="undefined"&&typeof OffscreenCanvasRenderingContext2D!="undefined")La=new OffscreenCanvas(1,1).getContext("2d");else throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");else La=document.createElement("canvas").getContext("2d",{willReadFrequently:!0});La.canvas.width=u,La.canvas.height=l,La.drawImage(e,0,0,u,l),p=La.getImageData(0,0,u,l).data}let d;if(t===4)d=new Int32Array(p);else{let f=u*l;d=new Int32Array(f*t);for(let m=0;m<f;m++)for(let g=0;g<t;++g)d[m*t+g]=p[m*4+g]}return Rd(d,[l,u,t],"int32")}function sP(e){return e!=null&&e.data instanceof Uint8Array}function aP(){return typeof window!="undefined"&&typeof ImageBitmap!="undefined"&&window.hasOwnProperty("createImageBitmap")}function oP(e){return e!=null&&e.width!==0&&e.height!==0}function iP(e){return aP()&&!(e instanceof ImageBitmap)&&oP(e)&&!sP(e)}async function cP(e,t=3){let n=null;if(q().getBool("WRAP_TO_IMAGEBITMAP")&&iP(e)){let r;try{r=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(s){r=null}r!=null&&r.width===e.width&&r.height===e.height?n=r:n=e}else n=e;return RS(n,t)}async function uP(e,t){let n=_(e,"img","toPixels");if(!(e instanceof Te)){let u=n;n=ce(u,"int32"),u.dispose()}if(n.rank!==2&&n.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);let[r,s]=n.shape.slice(0,2),a=n.rank===2?1:n.shape[2];if(a>4||a===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${a}`);if(n.dtype!=="float32"&&n.dtype!=="int32")throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);let o=await n.data(),i=n.dtype==="float32"?255:1,c=new Uint8ClampedArray(s*r*4);for(let u=0;u<r*s;++u){let l=[0,0,0,255];for(let d=0;d<a;d++){let h=o[u*a+d];if(n.dtype==="float32"){if(h<0||h>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${h}.`)}else if(n.dtype==="int32"&&(h<0||h>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${h}.`);a===1?(l[0]=h*i,l[1]=h*i,l[2]=h*i):l[d]=h*i}let p=u*4;c[p+0]=Math.round(l[0]),c[p+1]=Math.round(l[1]),c[p+2]=Math.round(l[2]),c[p+3]=Math.round(l[3])}if(t!=null){t.width=s,t.height=r;let u=t.getContext("2d"),l=new ImageData(c,s,r);u.putImageData(l,0,0)}return n!==e&&n.dispose(),c}var lP=z({fromPixels_:RS}),Uv={};Ae(Uv,{prepareAndValidate:()=>PS});function PS(e,t){let n=e.shape.length,r=t.shape.length;if(n<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);if(r<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${r}.`);if(t.dtype!=="int32")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[r-1]>n)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[r-1]} vs. ${n}`);if(ht(e.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);let s=t.shape,a=s[s.length-1],o=1;for(let p=0;p<s.length-1;++p)o*=s[p];let i=e.shape,c=s.slice();c.pop();let u=1;for(let p=a;p<n;++p)u*=i[p],c.push(i[p]);let l=[...Dc(e.shape).map(p=>p/u),1].slice(0,a);return[c,o,u,l]}var Gv={};Ae(Gv,{calculateShapes:()=>OS,validateInput:()=>qv,validateUpdateShape:()=>Hv});function Hv(e,t,n){let r=t.rank>1?t.shape[t.rank-1]:1,s=t.rank>1?t.rank-1:1,a=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${r}, and batchDim: ${s}.`;if(n.rank<s)throw new Error(a+` update.rank < ${s}. `);if(e.length<r+(n.rank-s))throw new Error(a+` Output shape length < ${r+(n.rank-s)}`);if(n.rank!==s+e.length-r)throw new Error(a+` update.rank != ${s+e.length-r}`);for(let o=0;o<s;++o)if(n.shape[o]!==t.shape[o])throw new Error(a+` updates.shape[${o}] (${n.shape[o]}) != indices.shape[${o}] (${t.shape[o]}).`);for(let o=0;o<n.rank-s;++o)if(n.shape[o+s]!==e[o+r])throw new Error(a+` updates.shape[${o+s}] (${n.shape[o+s]}) != shape[${o+s}] (${e[o+s]})`)}function qv(e,t,n){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(e.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);if(t.dtype!=="int32")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(n.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${n}`);if(n.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(e.size===0)throw new Error(`Updates specified for empty output. updates shape: ${e.shape}`)}Hv(n,t,e)}function OS(e,t,n){let r=t.shape.length,s=r>1?t.shape[r-1]:1,a=n.length,o=1;for(let p=s;p<a;++p)o*=n[p];let i=s<1?1:s,c=ht(t.shape)/i,u=[...Dc(n.slice(0,s)),1],l=ht(n);return{sliceRank:s,numUpdates:c,sliceSize:o,strides:u,outputSize:l}}var qt={};Ae(qt,{assertParamsValid:()=>pP,computeFlatOffset:()=>bP,computeOutShape:()=>fP,getNormalizedAxes:()=>mP,isSliceContinous:()=>gP,maskToAxes:()=>hP,parseSliceParams:()=>HS,sliceInfo:()=>yP,startForAxis:()=>US,startIndicesWithElidedDims:()=>BS,stopForAxis:()=>GS,stopIndicesWithElidedDims:()=>WS,stridesForAxis:()=>VS,stridesWithElidedDims:()=>MS});var By=-2,dP=-1;function pP(e,t,n){let r=e.shape.length;D(r===t.length,()=>`Error in slice${r}D: Length of begin ${t} must match the rank of the array (${r}).`),D(r===n.length,()=>`Error in slice${r}D: Length of size ${n} must match the rank of the array (${r}).`);for(let s=0;s<r;++s)D(t[s]+n[s]<=e.shape[s],()=>`Error in slice${r}D: begin[${s}] + size[${s}] (${t[s]+n[s]}) would overflow input.shape[${s}] (${e.shape[s]})`)}function hP(e){let t=[],n=0;for(;e>0;)e&1&&t.push(n),e/=2,n++;return t}function fP(e,t,n){let r=[];for(let s=0;s<e.length;s++)r[s]=Math.ceil((t[s]-e[s])/n[s]);return r}function MS(e,t,n,r){let s=[...e];for(let a=s.length;a<r.length;a++)s.push(1);for(let a=0;a<n;a++)a===0?s[t]=1:(s.splice(t,0,1),s.pop());return s}function LS(e,t,n){return n<=e?n:n-(t-1)}function zS(e,t){let n=[];for(let r=0;r<e;r++)n.push(t+r);return n}function mP(e,t,n,r,s,a,o,i,c){let u=e.length,l=new Array(u),p=new Array(u),d=new Array(u);if(t.length&&n>0){let h=t[0],f=n+1;l=BS(o,h,f,r,e),p=WS(i,h,f,s,e),d=MS(a,h,f,e)}else for(let h=0;h<u;h++)l[h]=US(o,r,a,e,h,c),p[h]=GS(i,s,a,e,h,c),d[h]=VS(a,h,c);return{begin:l,end:p,strides:d}}function BS(e,t,n,r,s){let a=[...s],o=zS(n,t);for(let i=0;i<a.length;i++)if(o.indexOf(i)>-1)a[i]=0;else{let c=LS(t,n,i),u=r[c];e&1<<c&&(u=0),a[i]=u}return a}function WS(e,t,n,r,s){let a=[...s],o=zS(n,t);for(let i=0;i<a.length;i++)if(o.indexOf(i)>-1)a[i]=Number.MAX_SAFE_INTEGER;else{let c=LS(t,n,i),u=r[c];e&1<<c&&(u=Number.MAX_SAFE_INTEGER),a[i]=u}for(let i=0;i<a.length;i++){let c=s[i];a[i]<0&&(a[i]+=c),a[i]=Kl(0,a[i],s[i])}return a}function VS(e,t,n){let r=e[t];return(n&1<<t||r==null)&&(r=1),r}function US(e,t,n,r,s,a){let o=t[s],i=n[s]||1;(e&1<<s||a&1<<s||o==null)&&(i>0?o=Number.MIN_SAFE_INTEGER:o=Number.MAX_SAFE_INTEGER);let c=r[s];return o<0&&(o+=c),o=Kl(0,o,c-1),o}function GS(e,t,n,r,s,a){let o=t[s],i=n[s]||1;(e&1<<s||a&1<<s||o==null)&&(i>0?o=Number.MAX_SAFE_INTEGER:o=Number.MIN_SAFE_INTEGER);let c=r[s];return o<0&&(o+=c),i>0?o=Kl(0,o,c):o=Kl(-1,o,c-1),o}function gP(e,t,n){let r=n.length;for(let s=0;s<n.length;s++)if(n[s]>1){r=s;break}for(let s=r+1;s<n.length;s++)if(t[s]>0||n[s]!==e[s])return!1;return!0}function bP(e,t){let n=e.length>0?e[e.length-1]:1;for(let r=0;r<e.length-1;r++)n+=e[r]*t[r];return n}function HS(e,t,n){let r,s=e.shape.length;typeof t=="number"?r=[t,...new Array(s-1).fill(0)]:t.length<s?r=t.concat(new Array(s-t.length).fill(0)):r=t.slice(),r.forEach(o=>{D(o!==-1,()=>"slice() does not support negative begin indexing.")});let a;return n==null?a=new Array(s).fill(-1):typeof n=="number"?a=[n,...new Array(s-1).fill(-1)]:n.length<s?a=n.concat(new Array(s-n.length).fill(-1)):a=n,a=a.map((o,i)=>o>=0?o:(D(o===-1,()=>`Negative size values should be exactly -1 but got ${o} for the slice() size at index ${i}.`),e.shape[i]-r[i])),[r,a]}function yP(e,t,n,r,s,a,o,i,c){let u;if(r==null?(u=new Array(t.length),u.fill(1)):u=r,o!=null&&(o&o-1)!==0)throw new Error("Multiple ellipses in slice is not allowed.");let l=!1,p={dims:u.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:n.slice(),strides:u.slice(),beginMask:s,endMask:a,ellipsisMask:o,newAxisMask:i,shrinkAxisMask:c};for(let v=0;v<p.dims;v++)l&&(1<<v&i)!==0&&p.numAddAxisAfterEllipsis++,1<<v&o&&(l=!0);l||(p.ellipsisMask|=1<<p.dims,p.dims++);let d={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};vP(p,d);let h=!0,f=!0,m=!0,g=[],b=[];for(let v=0;v<e.length;++v){if(d.strides[v]===0)throw Error(`strides[${v}] must be non-zero`);let x=!!(d.shrinkAxisMask&1<<v),k=e[v];if(k===-1){g.push(x?1:-1);continue}let S=[d.beginMask&1<<v,d.endMask&1<<v],C=[d.strides[v]>0?0:-1,d.strides[v]>0?k:k-1];if(x&&d.strides[v]<=0)throw Error("only stride 1 allowed on non-range indexing.");m=m&&d.strides[v]===1;let E=!!(d.beginMask&1<<v&&d.endMask&1<<v);if(d.beginValid&&d.endValid){if(x){let R=d.begin[v]<0?k+d.begin[v]:d.begin[v];if(d.begin[v]=R,d.end[v]=d.begin[v]+1,R<0||R>=k)throw Error(`slice index ${d.begin[v]} of dimension ${v} out of bounds.`)}else d.begin[v]=bk(d.begin[v],0,d.strides[v],k,S,C),d.end[v]=bk(d.end[v],1,d.strides[v],k,S,C);let A=d.strides[v]===1&&d.begin[v]===0&&d.end[v]===k;h=h&&A,f=f&&(v===0&&d.strides[v]===1||A)}else h=h&&d.strides[v]===1&&E,f=f&&(v===0&&d.strides[v]===1||E);let $,F=!1;if(d.beginValid&&d.endValid?($=d.end[v]-d.begin[v],F=!0):x?($=1,F=!0):E&&k>=0&&(d.strides[v]<0?$=-k:$=k,F=!0),F){let A;$===0||$<0!=d.strides[v]<0?A=0:A=Math.trunc($/d.strides[v])+($%d.strides[v]!==0?1:0),g.push(A)}else g.push(-1)}for(let v=0;v<d.finalShapeGatherIndices.length;++v){let x=d.finalShapeGatherIndices[v];x>=0?b.push(g[x]):x===By&&b.push(1)}return{finalShapeSparse:b.filter((v,x)=>d.finalShapeGatherIndices[x]!==By),finalShape:b,isIdentity:h,sliceDim0:f,isSimpleSlice:m,begin:d.begin,end:d.end,strides:d.strides}}function vP(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let n=0;t.beginValid=e.begin!=null,t.endValid=e.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let r=0;r<e.dims;r++)if(1<<r&e.ellipsisMask){let s=Math.min(t.dims-(e.dims-r)+1+e.numAddAxisAfterEllipsis,t.dims);for(;n<s;n++)t.begin[n]=0,t.end[n]=0,t.strides[n]=1,t.beginMask|=1<<n,t.endMask|=1<<n,t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[n]=r}else if(1<<r&e.newAxisMask)t.finalShapeGatherIndices.push(By),t.finalShapeGatherIndicesSparse.push(-1);else{if(n===t.begin.length)throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);e.begin!=null&&(t.begin[n]=e.begin[r]),e.end!=null&&(t.end[n]=e.end[r]),t.strides[n]=e.strides[r],e.beginMask&1<<r&&(t.beginMask|=1<<n),e.endMask&1<<r&&(t.endMask|=1<<n),e.shrinkAxisMask&1<<r?(t.finalShapeGatherIndices.push(dP),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<n):(t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(r)),t.inputShapeGatherIndicesSparse[n]=r,n++}}function bk(e,t,n,r,s,a){if(s[t])return n>0?a[t]:a[t+1&1];{let o=e<0?r+e:e;return o<a[0]?a[0]:o>a[1]?a[1]:o}}var se={};Ae(se,{Serializable:()=>qS,SerializationMap:()=>Ua,registerClass:()=>Ia});var qS=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},Ua=class{constructor(){this.classNameMap={}}static getMap(){return Ua.instance==null&&(Ua.instance=new Ua),Ua.instance}static register(e){Ua.getMap().classNameMap[e.className]=[e,e.fromConfig]}};function Ia(e){D(e.className!=null,()=>"Class being registered does not have the static className property defined."),D(typeof e.className=="string",()=>"className is required to be a string, but got type "+typeof e.className),D(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),Ua.register(e)}var jS={};Ae(jS,{TEST_EPSILON_FLOAT16:()=>KS,createVideoElement:()=>NP,encodeStrings:()=>XS,expectArrayBuffersEqual:()=>CP,expectArraysClose:()=>wP,expectArraysEqual:()=>kP,expectNumbersClose:()=>SP,expectPromiseToFail:()=>IP,expectValuesInRange:()=>TP,play:()=>_P,testEpsilon:()=>jv});var xP=.001,KS=.1;function wP(e,t,n){return n==null&&(n=jv()),Wy(e,t,(r,s)=>Kv(r,s,n))}function jv(){return M.backend.floatPrecision()===32?xP:KS}function Wy(e,t,n){let r=!0;if((fn(e)||fn(t))&&(r=!1),fn(e)&&fn(t)&&(r=!0),r){let o=e.constructor.name,i=t.constructor.name;if(o!==i)throw new Error(`Arrays are of different type. Actual: ${o}. Expected: ${i}`)}if(Array.isArray(e)&&Array.isArray(t)){let o=us(e),i=us(t);if(!ga(o,i))throw new Error(`Arrays have different shapes. Actual: [${o}]. Expected: [${i}]`)}let s=fn(e)?e:Qa(e),a=fn(t)?t:Qa(t);if(s.length!==a.length)throw new Error(`Arrays have different lengths actual: ${s.length} vs expected: ${a.length}.
+Actual:   ${s}.
+Expected: ${a}.`);for(let o=0;o<a.length;++o){let i=s[o],c=a[o];if(!n(i,c))throw new Error(`Arrays differ: actual[${o}] = ${i}, expected[${o}] = ${c}.
+Actual:   ${s}.
+Expected: ${a}.`)}typeof expect!="undefined"&&expect().nothing()}function IP(e,t){e().then(()=>t.fail(),()=>t()),typeof expect!="undefined"&&expect().nothing()}function kP(e,t){let n=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return ea(e)||ea(e[0])||ea(t)||ea(t[0])?Wy(e,n,(r,s)=>r==s):Wy(e,t,(r,s)=>Kv(r,s,0))}function SP(e,t,n){if(n==null&&(n=jv()),!Kv(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`);typeof expect!="undefined"&&expect().nothing()}function Kv(e,t,n){return!isFinite(e)&&!isFinite(t)?!0:!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function TP(e,t,n){for(let r=0;r<e.length;r++)if(e[r]<t||e[r]>n)throw new Error(`Value out of range:${e[r]} low: ${t}, high: ${n}`)}function CP(e,t){let n=new Float32Array(e),r=new Float32Array(t);if(n.length!==r.length)throw new Error(`Expected ArrayBuffer to be of length ${r.length}, but it was ${n.length}`);for(let s=0;s<r.length;s++)if(n[s]!==r[s])throw new Error(`Expected ArrayBuffer value at ${s} to be ${r[s]} but got ${n[s]} instead`)}function XS(e){for(let t=0;t<e.length;t++){let n=e[t];Array.isArray(n)?XS(n):e[t]=Ad(n)}return e}function NP(e){let t=document.createElement("video");return"playsInline"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position="fixed",t.style.left="0px",t.style.top="0px",t.preload="auto",t.appendChild(e),new Promise(n=>{t.addEventListener("loadeddata",r=>n(t)),t.load()})}async function _P(e){await e.play(),"requestVideoFrameCallback"in e&&await new Promise(t=>{e.requestVideoFrameCallback(t)})}var EP="4.0.0";function AP(e,t){let n=_(e,"a","add"),r=_(t,"b","add");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(ba,s)}var Y=z({add_:AP});function $P(e,t){let n=_(e,"a","floorDiv"),r=_(t,"b","floorDiv");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(Fo,s)}var jf=z({floorDiv_:$P});function DP(e,t){let n=_(e,"a","div"),r=_(t,"b","div");if([n,r]=_t(n,r),n.dtype==="int32"&&r.dtype==="int32")return jf(n,r);let s={a:n,b:r},a={};return M.runKernel(Eo,s,a)}var fe=z({div_:DP});function FP(e,t){let n=_(e,"a","mul"),r=_(t,"b","mul");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(qo,s)}var B=z({mul_:FP});function RP(e){let t=_(e,"x","abs");if(t.dtype==="complex64"){let n={x:t};return M.runKernel(hd,n)}else{let n={x:t};return M.runKernel(Fc,n)}}var Lt=z({abs_:RP});function PP(e){let n={x:_(e,"x","acos")};return M.runKernel(Rc,n)}var Xv=z({acos_:PP});function OP(e){let n={x:_(e,"x","acosh")};return M.runKernel(Pc,n)}var Yv=z({acosh_:OP});function MP(e){D(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),D(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);let t=e.map((s,a)=>_(s,`tensors${a}`,"addN")),n=t[0];t.forEach(s=>{if(s.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(s=>{if(!ga(s.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let r=t;return M.runKernel(bo,r)}var YS=z({addN_:MP});function LP(e,t=null,n=!1){let s={x:_(e,"x","all","bool")},a={axis:t,keepDims:n};return M.runKernel(Oc,s,a)}var Kf=z({all_:LP});function zP(e,t=null,n=!1){let s={x:_(e,"x","any","bool")},a={axis:t,keepDims:n};return M.runKernel(Mc,s,a)}var Ql=z({any_:zP});function BP(e,t=0){let r={x:_(e,"x","argMax")},s={axis:t};return M.runKernel(yo,r,s)}var ao=z({argMax_:BP});function WP(e,t=0){let r={x:_(e,"x","argMin")},s={axis:t};return M.runKernel(dd,r,s)}var Zv=z({argMin_:WP});function VP(e){let n={x:_(e,"x","asin")};return M.runKernel(Lc,n)}var Jv=z({asin_:VP});function UP(e){let n={x:_(e,"x","asinh")};return M.runKernel(zc,n)}var Qv=z({asinh_:UP});function GP(e){let n={x:_(e,"x","atan")};return M.runKernel(Bc,n)}var ex=z({atan_:GP});function HP(e,t){let n=_(e,"a","atan2"),r=_(t,"b","atan2");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(Vc,s)}var tx=z({atan2_:HP});function qP(e){let n={x:_(e,"x","atanh")};return M.runKernel(Wc,n)}var nx=z({atanh_:qP});function jP(e,t,n,r,s="NHWC",a){let o=e[3],i=[...t,o],c=QS(s);return Pd(e,i,n,a,r,null,null,c)}function ZS(e,t,n,r,s,a,o="channelsLast"){let[i,c]=zh(t),u;if(o==="channelsLast")u=[i,c,e[3],e[3]];else if(o==="channelsFirst")u=[i,c,e[1],e[1]];else throw new Error(`Unknown dataFormat ${o}`);return Pd(e,u,n,r,s,a,!1,o)}function KP(e,t,n,r,s,a,o="NDHWC"){let[i,c,u]=Vy(t),l,p;if(o==="NDHWC")p="channelsLast",l=[i,c,u,e[4],e[4]];else if(o==="NCDHW")p="channelsFirst",l=[i,c,u,e[1],e[1]];else throw new Error(`Unknown dataFormat ${o}`);return JS(e,l,n,r,s,!1,p,a)}function Pd(e,t,n,r,s,a,o=!1,i="channelsLast"){let[c,u,l,p]=[-1,-1,-1,-1];if(i==="channelsLast")[c,u,l,p]=e;else if(i==="channelsFirst")[c,p,u,l]=e;else throw new Error(`Unknown dataFormat ${i}`);let[d,h,,f]=t,[m,g]=zh(n),[b,y]=zh(r),v=dc(d,b),x=dc(h,y),{padInfo:k,outHeight:S,outWidth:C}=ZP(s,u,l,m,g,v,x,a,i),E=o?f*p:f,$;return i==="channelsFirst"?$=[c,E,S,C]:i==="channelsLast"&&($=[c,S,C,E]),{batchSize:c,dataFormat:i,inHeight:u,inWidth:l,inChannels:p,outHeight:S,outWidth:C,outChannels:E,padInfo:k,strideHeight:m,strideWidth:g,filterHeight:d,filterWidth:h,effectiveFilterHeight:v,effectiveFilterWidth:x,dilationHeight:b,dilationWidth:y,inShape:e,outShape:$,filterShape:t}}function JS(e,t,n,r,s,a=!1,o="channelsLast",i){let[c,u,l,p,d]=[-1,-1,-1,-1,-1];if(o==="channelsLast")[c,u,l,p,d]=e;else if(o==="channelsFirst")[c,d,u,l,p]=e;else throw new Error(`Unknown dataFormat ${o}`);let[h,f,m,,g]=t,[b,y,v]=Vy(n),[x,k,S]=Vy(r),C=dc(h,x),E=dc(f,k),$=dc(m,S),{padInfo:F,outDepth:A,outHeight:R,outWidth:T}=JP(s,u,l,p,b,y,v,C,E,$,i),L=a?g*d:g,V;return o==="channelsFirst"?V=[c,L,A,R,T]:o==="channelsLast"&&(V=[c,A,R,T,L]),{batchSize:c,dataFormat:o,inDepth:u,inHeight:l,inWidth:p,inChannels:d,outDepth:A,outHeight:R,outWidth:T,outChannels:L,padInfo:F,strideDepth:b,strideHeight:y,strideWidth:v,filterDepth:h,filterHeight:f,filterWidth:m,effectiveFilterDepth:C,effectiveFilterHeight:E,effectiveFilterWidth:$,dilationDepth:x,dilationHeight:k,dilationWidth:S,inShape:e,outShape:V,filterShape:t}}function XP(e,t,n,r,s){r==null&&(r=rx(e,t,n));let a=e[0],o=e[1],i=Xa((a-t+2*r)/n+1,s),c=Xa((o-t+2*r)/n+1,s);return[i,c]}function YP(e,t,n,r,s,a){s==null&&(s=rx(e,t,r));let o=e[0],i=e[1],c=e[2],u=Xa((o-t+2*s)/r+1,a),l=Xa((i-t+2*s)/r+1,a),p=Xa((c-t+2*s)/r+1,a);return[u,l,p,n]}function rx(e,t,n,r=1){let s=dc(t,r);return Math.floor((e[0]*(n-1)-n+s)/2)}function zh(e){return typeof e=="number"?[e,e,e]:e.length===2?[e[0],e[1],1]:e}function Vy(e){return typeof e=="number"?[e,e,e]:e}function dc(e,t){return t<=1?e:e+(e-1)*(t-1)}function ZP(e,t,n,r,s,a,o,i,c){let u,l,p;if(typeof e=="number"){u={top:e,bottom:e,left:e,right:e,type:e===0?"VALID":"NUMBER"};let h=XP([t,n],a,r,e,i);l=h[0],p=h[1]}else if(e==="same"){l=Math.ceil(t/r),p=Math.ceil(n/s);let d=Math.max(0,(l-1)*r+a-t),h=Math.max(0,(p-1)*s+o-n),f=Math.floor(d/2),m=d-f,g=Math.floor(h/2),b=h-g;u={top:f,bottom:m,left:g,right:b,type:"SAME"}}else if(e==="valid")u={top:0,bottom:0,left:0,right:0,type:"VALID"},l=Math.ceil((t-a+1)/r),p=Math.ceil((n-o+1)/s);else if(typeof e=="object"){let d=c==="channelsLast"?e[1][0]:e[2][0],h=c==="channelsLast"?e[1][1]:e[2][1],f=c==="channelsLast"?e[2][0]:e[3][0],m=c==="channelsLast"?e[2][1]:e[3][1];u={top:d,bottom:h,left:f,right:m,type:d===0&&h===0&&f===0&&m===0?"VALID":"EXPLICIT"},l=Xa((t-a+d+h)/r+1,i),p=Xa((n-o+f+m)/s+1,i)}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:u,outHeight:l,outWidth:p}}function JP(e,t,n,r,s,a,o,i,c,u,l){let p,d,h,f;if(typeof e=="number"){p={top:e,bottom:e,left:e,right:e,front:e,back:e,type:e===0?"VALID":"NUMBER"};let g=YP([t,n,r,1],i,1,s,e,l);d=g[0],h=g[1],f=g[2]}else if(e==="same"){d=Math.ceil(t/s),h=Math.ceil(n/a),f=Math.ceil(r/o);let m=(d-1)*s+i-t,g=(h-1)*a+c-n,b=(f-1)*o+u-r,y=Math.floor(m/2),v=m-y,x=Math.floor(g/2),k=g-x,S=Math.floor(b/2),C=b-S;p={top:x,bottom:k,left:S,right:C,front:y,back:v,type:"SAME"}}else if(e==="valid")p={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},d=Math.ceil((t-i+1)/s),h=Math.ceil((n-c+1)/a),f=Math.ceil((r-u+1)/o);else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:p,outDepth:d,outHeight:h,outWidth:f}}function Xa(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error(`Unknown roundingMode ${t}`)}}function ca(e){let[t,n,r]=zh(e);return t===1&&n===1&&r===1}function ps(e,t){return ca(e)||ca(t)}function QS(e){if(e==="NHWC")return"channelsLast";if(e==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function En(e,t,n){if(n!=null){if(typeof t=="string")throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);if(typeof t=="number")D(hc(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);else if(typeof t=="object")t.forEach(r=>{r.forEach(s=>{D(hc(s),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${s}.`)})});else throw Error(`Error in ${e}: Unknown padding parameter: ${t}`)}}function QP(e,t){let r={x:_(e,"x","reshape","string_or_numeric")},s={shape:t};return M.runKernel(xu,r,s)}var W=z({reshape_:QP});function eO(e,t,n,r,s){let a=_(e,"x","avgPool","float32"),o=1;D(ps(n,o),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`);let i=a,c=!1;a.rank===3&&(c=!0,i=W(a,[1,a.shape[0],a.shape[1],a.shape[2]])),D(i.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${i.rank}.`),En("avgPool",r,s);let u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:s},p=M.runKernel(vo,u,l);return p=ce(p,a.dtype),c?W(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var mr=z({avgPool_:eO});function tO(e,t,n,r,s,a="NDHWC"){let o=_(e,"x","avgPool3d","float32"),i=o,c=!1;o.rank===4&&(c=!0,i=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),D(i.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${i.rank}.`),D(a==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${a}`),En("avgPool3d",r,s);let u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:s,dataFormat:a},p=M.runKernel(pd,u,l);return p=ce(p,i.dtype),c?W(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var sx=z({avgPool3d_:tO});function nO(e,t=0){D(e.length>=1,()=>"Pass at least one tensor to concat");let n=Jl(e,"tensors","concat","string_or_numeric");if(n[0].dtype==="complex64"&&n.forEach(a=>{if(a.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
+          with dtype ${a.dtype}. `)}),n.length===1)return is(n[0]);let r=n,s={axis:t};return M.runKernel(Gc,r,s)}var Je=z({concat_:nO});function rO(e){let n={x:_(e,"x","sigmoid","float32")};return M.runKernel(oi,n)}var dr=z({sigmoid_:rO});function sO(e,t,n){let r=_(e,"x","slice","string_or_numeric");if(r.rank===0)throw new Error("Slicing scalar is not possible");let s={x:r},a={begin:t,size:n};return M.runKernel(Su,s,a)}var We=z({slice_:sO});function aO(e){let n={x:_(e,"x","tanh","float32")};return M.runKernel(hi,n)}var oo=z({tanh_:aO});function oO(e,t,n,r,s,a){let o=_(e,"forgetBias","basicLSTMCell"),i=_(t,"lstmKernel","basicLSTMCell"),c=_(n,"lstmBias","basicLSTMCell"),u=_(r,"data","basicLSTMCell"),l=_(s,"c","basicLSTMCell"),p=_(a,"h","basicLSTMCell"),d=Je([u,p],1),h=De(d,i),f=Y(h,c),m=f.shape[0],g=f.shape[1]/4,b=[m,g],y=We(f,[0,0],b),v=We(f,[0,g],b),x=We(f,[0,g*2],b),k=We(f,[0,g*3],b),S=Y(B(dr(y),oo(v)),B(l,dr(Y(o,x)))),C=B(oo(S),dr(k));return[S,C]}var eT=z({basicLSTMCell_:oO});function iO(e,t,n){let r=_(e,"x","batchToSpaceND"),s=t.reduce((i,c)=>i*c);D(r.rank>=1+t.length,()=>`input rank is ${r.rank} but should be > than blockShape.length ${t.length}`),D(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`),D(r.shape[0]%s===0,()=>`input tensor batch is ${r.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${s}`);let a={x:r},o={blockShape:t,crops:n};return M.runKernel(Uc,a,o)}var Od=z({batchToSpaceND_:iO});function cO(e){let t;return e.rank===0||e.rank===1?t=W(e,[1,1,1,e.size]):e.rank===2?t=W(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?t=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]):t=e,t}function uO(e,t,n,r,s,a){a==null&&(a=.001);let o=_(e,"x","batchNorm"),i=_(t,"mean","batchNorm"),c=_(n,"variance","batchNorm"),u;s!=null&&(u=_(s,"scale","batchNorm"));let l;r!=null&&(l=_(r,"offset","batchNorm")),D(i.rank===c.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),D(l==null||i.rank===l.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),D(u==null||i.rank===u.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let d={x:cO(o),scale:u,offset:l,mean:i,variance:c},h={varianceEpsilon:a},f=M.runKernel(Ro,d,h);return W(f,o.shape)}var ka=z({batchNorm_:uO});function lO(e,t,n,r,s,a){let o=_(e,"x","batchNorm"),i=_(t,"mean","batchNorm"),c=_(n,"variance","batchNorm"),u;s!=null&&(u=_(s,"scale","batchNorm"));let l;return r!=null&&(l=_(r,"offset","batchNorm")),D(o.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${o.rank}.`),D(i.rank===2||i.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${i.rank}.`),D(c.rank===2||c.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${c.rank}.`),u!=null&&D(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${u.rank}.`),l!=null&&D(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${l.rank}.`),ka(o,i,c,l,u,a)}var ax=z({batchNorm2d_:lO});function dO(e,t,n,r,s,a){let o=_(e,"x","batchNorm"),i=_(t,"mean","batchNorm"),c=_(n,"variance","batchNorm"),u;s!=null&&(u=_(s,"scale","batchNorm"));let l;return r!=null&&(l=_(r,"offset","batchNorm")),D(o.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${o.rank}.`),D(i.rank===3||i.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${i.rank}.`),D(c.rank===3||c.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${c.rank}.`),u!=null&&D(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${u.rank}.`),l!=null&&D(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${l.rank}.`),ka(o,i,c,l,u,a)}var ox=z({batchNorm3d_:dO});function pO(e,t,n,r,s,a){let o=_(e,"x","batchNorm"),i=_(t,"mean","batchNorm"),c=_(n,"variance","batchNorm"),u;s!=null&&(u=_(s,"scale","batchNorm"));let l;return r!=null&&(l=_(r,"offset","batchNorm")),D(o.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${o.rank}.`),D(i.rank===4||i.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${i.rank}.`),D(c.rank===4||c.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${c.rank}.`),u!=null&&D(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${u.rank}.`),l!=null&&D(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${l.rank}.`),ka(o,i,c,l,u,a)}var ix=z({batchNorm4d_:pO});function hO(e,t,n){let r=_(e,"x","bincount"),s=_(t,"weights","bincount");D(r.dtype==="int32",()=>`Error in bincount: input dtype must be int32, but got ${r.dtype}`),D(n>=0,()=>`size must be non-negative, but got ${n}.`),D(s.size===r.size||s.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${r.shape}, weights shape: ${s.shape}.`);let a={x:r,weights:s},o={size:n};return M.runKernel(ff,a,o)}var cx=z({bincount_:hO});function fO(e,t){let n=_(e,"s0","broadcastArgs","int32"),r=_(t,"s1","broadcastArgs","int32");if(n.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${n.rank}`);if(r.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${r.rank}`);let s={s0:n,s1:r};return M.runKernel(mf,s)}var tT=z({broadcastArgs_:fO});function mO(e,t){let n=_(e,"broadcastTo","x"),r=n.shape;if(t.some(u=>!(u>0)||u%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){let u=n.shape.slice();for(;u.length<t.length;)u.unshift(1);n=W(n,u)}let s=n.shape,a=Array.from(t);for(let u=t.length-1;u>=0;u--)if(s[u]===t[u])a[u]=1;else if(n.shape[u]!==1)throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${t}].`);if(a.map((u,l)=>u>1?l:-1).filter(u=>u>=0).length===0)return is(n);let i={x:n},c={reps:a};return M.runKernel(va,i,c)}var Ya=z({broadcastTo_:mO});function gO(e){let n={x:_(e,"x","ceil","float32")};return M.runKernel(Io,n)}var ux=z({ceil_:gO});function bn(e,t,n){let r={shape:e,value:t,dtype:n};return M.runKernel(gd,{},r)}function bO(e,t,n){let r=_(e,"x","clipByValue");if(D(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`),t===n)return bn(r.shape,t,r.dtype);let s={x:r},a={clipValueMin:t,clipValueMax:n};return M.runKernel(ya,s,a)}var en=z({clipByValue_:bO});function yO(e){return Je(e,0)}var lx=z({concat1d_:yO});function vO(e,t){return Je(e,t)}var dx=z({concat2d_:vO});function xO(e,t){return Je(e,t)}var px=z({concat3d_:xO});function wO(e,t){return Je(e,t)}var hx=z({concat4d_:wO});function IO(e,t,n,r,s="NHWC",a=[1,1],o){let i=_(e,"x","conv2d","float32"),c=_(t,"filter","conv2d","float32"),u=i,l=!1;i.rank===3&&(l=!0,u=W(i,[1,i.shape[0],i.shape[1],i.shape[2]])),D(u.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${u.rank}.`),D(c.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${c.rank}.`),En("conv2d",r,o);let p=s==="NHWC"?u.shape[3]:u.shape[1];D(p===c.shape[2],()=>`Error in conv2d: depth of input (${p}) must match input depth for filter ${c.shape[2]}.`),D(ps(n,a),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`);let d={x:u,filter:c},h={strides:n,pad:r,dataFormat:s,dilations:a,dimRoundingMode:o},f=M.runKernel(ko,d,h);return l?W(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var Dt=z({conv2d_:IO});function kO(e,t,n,r,s="NWC",a=1,o){let i=_(e,"x","conv1d"),c=_(t,"filter","conv1d"),u=i,l=!1;i.rank===2&&(l=!0,u=W(i,[1,i.shape[0],i.shape[1]])),D(u.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${u.rank}.`),D(c.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${c.rank}.`),En("conv1d",r,o),D(u.shape[2]===c.shape[1],()=>`Error in conv1d: depth of input (${u.shape[2]}) must match input depth for filter ${c.shape[1]}.`),D(ps(n,a),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${a}'`),D(s==="NWC",()=>`Error in conv1d: got dataFormat of ${s} but only NWC is currently supported.`);let p=W(c,[1,c.shape[0],c.shape[1],c.shape[2]]),d=W(u,[u.shape[0],1,u.shape[1],u.shape[2]]),g=Dt(d,p,[1,n],r,"NHWC",[1,a],o);return l?W(g,[g.shape[2],g.shape[3]]):W(g,[g.shape[0],g.shape[2],g.shape[3]])}var Xf=z({conv1d_:kO});function SO(e,t,n,r,s,a="NHWC",o){D(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let i=e,c=t,u=!1;t.rank===3&&(u=!0,c=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]),i=[1,e[0],e[1],e[2]]),D(i.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${i.length}.`),D(c.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${c.rank}`),D(n.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${n.rank}`);let l=a==="NHWC"?i[3]:i[1],p=a==="NHWC"?c.shape[3]:c.shape[1];D(l===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${l}) must match input depth for filter ${n.shape[2]}.`),D(p===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${p}) must match output depth for filter ${n.shape[3]}.`),En("conv2dDerInput",s,o);let d={dy:c,filter:n},h={strides:r,pad:s,dataFormat:a,dimRoundingMode:o,inputShape:i},f=M.runKernel(So,d,h);return u?W(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var fx=z({conv2DBackpropInput_:SO});function TO(e,t,n,r,s,a){let o=_(e,"x","conv2dTranspose"),i=_(t,"filter","conv2dTranspose");return fx(n,o,i,r,s,"NHWC",a)}var Yf=z({conv2dTranspose_:TO});function CO(e,t,n,r,s="NDHWC",a=[1,1,1]){let o=_(e,"x","conv3d"),i=_(t,"filter","conv3d"),c=o,u=!1;o.rank===4&&(u=!0,c=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),D(c.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${c.rank}.`),D(i.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${i.rank}.`),D(c.shape[4]===i.shape[3],()=>`Error in conv3d: depth of input (${c.shape[4]}) must match input depth for filter ${i.shape[3]}.`),D(ps(n,a),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`),D(s==="NDHWC",()=>`Error in conv3d: got dataFormat of ${s} but only NDHWC is currently supported.`);let l={x:c,filter:i},p={strides:n,pad:r,dataFormat:s,dilations:a},d=M.runKernel(fd,l,p);return u?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var mx=z({conv3d_:CO});function NO(e,t,n,r,s){D(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let a=e,o=t,i=!1;t.rank===4&&(i=!0,o=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),a=[1,e[0],e[1],e[2],e[3]]);let c=a[4],u=o.shape[4];D(a.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${a.length}.`),D(o.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${o.rank}`),D(n.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${n.rank}`),D(c===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${c}) must match input depth for filter ${n.shape[3]}.`),D(u===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${u}) must match output depth for filter ${n.shape[4]}.`);let l={dy:o,filter:n},p={pad:s,strides:r,inputShape:a},d=M.runKernel(vf,l,p);return i?W(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}var nT=z({conv3DBackpropInput_:NO});function _O(e,t,n,r,s){let a=_(e,"x","conv3dTranspose"),o=_(t,"filter","conv3dTranspose");return nT(n,a,o,r,s)}var gx=z({conv3dTranspose_:_O});function EO(e){let n={x:_(e,"x","cos","float32")};return M.runKernel(To,n)}var Md=z({cos_:EO});function AO(e){let n={x:_(e,"x","cosh","float32")};return M.runKernel(Co,n)}var Zf=z({cosh_:AO});function $O(e,t=0,n=!1,r=!1){let a={x:_(e,"x","cumprod")},o={axis:t,exclusive:n,reverse:r};return M.runKernel(Hc,a,o)}var ed=z({cumprod_:$O});function DO(e,t=0,n=!1,r=!1){let a={x:_(e,"x","cumsum")},o={axis:t,exclusive:n,reverse:r};return M.runKernel(No,a,o)}var Jf=z({cumsum_:DO});function FO(e,t,n,r=!1){let s=_(e,"x","denseBincount"),a=_(t,"weights","denseBincount");D(s.dtype==="int32",()=>`Error in denseBincount: input dtype must be int32, but got ${s.dtype}`),D(s.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${s.rank}.`),D(n>=0,()=>`size must be non-negative, but got ${n}.`),D(a.size===s.size||a.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${s.shape}, weights shape: ${a.shape}.`);let o={x:s,weights:a},i={size:n,binaryOutput:r};return M.runKernel(xf,o,i)}var Bh=z({denseBincount_:FO});function RO(e,t,n="NHWC"){let r=_(e,"x","depthToSpace","float32"),s=n==="NHWC"?r.shape[1]:r.shape[2],a=n==="NHWC"?r.shape[2]:r.shape[3],o=n==="NHWC"?r.shape[3]:r.shape[1];D(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),D(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying
+    ${s} and ${t}  for depthToSpace with input shape
+    ${r.shape}`),D(a*t>=0,()=>`Negative dimension size caused by overflow when multiplying
+    ${a} and ${t} for depthToSpace with input shape
+        ${r.shape}`),D(o%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${o} for depthToSpace with input shape ${r.shape}`);let i={x:r},c={blockSize:t,dataFormat:n};return M.runKernel(jc,i,c)}var bx=z({depthToSpace_:RO});function PO(e,t,n,r,s="NHWC",a=[1,1],o){let i=_(e,"x","depthwiseConv2d","float32"),c=_(t,"filter","depthwiseConv2d","float32"),u=i,l=!1;i.rank===3&&(l=!0,u=W(i,[1,i.shape[0],i.shape[1],i.shape[2]])),D(u.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${u.rank}.`),D(c.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${c.rank}.`);let p=s==="NHWC"?u.shape[3]:u.shape[1];D(p===c.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${p}) must match the inChannels dimension in filter ${c.shape[2]}.`),En("depthwiseConv2d",r,o);let d={x:u,filter:c},h={strides:n,pad:r,dataFormat:s,dilations:a,dimRoundingMode:o},f=M.runKernel(_o,d,h);return l?W(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var Sa=z({depthwiseConv2d_:PO});function OO(e){let n={x:_(e,"x","diag")};return M.runKernel(kf,n)}var rT=z({diag_:OO});function MO(e,t,n,r,s=[1,1],a="NHWC"){let o=_(e,"x","dilation2d"),i=_(t,"filter","dilation2d");D(o.rank===3||o.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${o.rank}.`),D(i.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${i.rank}.`),D(a==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${a}`);let c=o,u=!1;o.rank===3&&(c=W(o,[1,o.shape[0],o.shape[1],o.shape[2]]),u=!0);let l={x:c,filter:i},p={strides:n,pad:r,dilations:s},d=M.runKernel(md,l,p);return u?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var yx=z({dilation2d_:MO});function LO(e,t){let n=_(e,"a","equal","string_or_numeric"),r=_(t,"b","equal","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(Xc,s)}var tr=z({equal_:LO});function zO(e,t,n){let r=_(t,"a","where"),s=_(n,"b","where"),a=_(e,"condition","where","bool"),o=ct(ct(a.shape,r.shape),s.shape),i=Ya(a,o),c=Ya(r,o),u=Ya(s,o),l={condition:i,t:c,e:u};return M.runKernel(Iu,l)}var mn=z({where_:zO});function BO(e){let n={x:_(e,"x","zerosLike")};return M.runKernel(Pu,n)}var qe=z({zerosLike_:BO});function WO(e,t){let n=_(e,"a","div"),r=_(t,"b","div");[n,r]=_t(n,r);let s=fe(n,r),a=qe(s),o=tr(r,a);return mn(o,a,s)}var vx=z({divNoNan_:WO});function VO(e,t){let n=_(e,"t1","dot"),r=_(t,"t2","dot");D((n.rank===1||n.rank===2)&&(r.rank===1||r.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${r.rank}.`);let s=n.rank===1?n.size:n.shape[1],a=r.rank===1?r.size:r.shape[0];if(D(s===a,()=>`Error in dot: inner dimensions of inputs must match, but got ${s} and ${a}.`),n.rank===1&&r.rank===1){let o=W(n,[1,-1]),i=W(r,[-1,1]),c=De(o,i);return W(c,[])}else if(n.rank===1&&r.rank===2){let o=W(n,[1,-1]),i=W(r,[r.shape[0],r.shape[1]]),c=De(o,i);return W(c,[c.size])}else if(n.rank===2&&r.rank===1){let o=W(r,[-1,1]),i=De(n,o);return W(i,[i.size])}else{let o=W(r,[r.shape[0],r.shape[1]]);return De(n,o)}}var xx=z({dot_:VO});function UO(e,...t){let n=t.map((s,a)=>_(s,`tensors${a}`,"einsum")),r={equation:e};return M.runKernel(Sf,n,r)}var sT=z({einsum_:UO});function GO(e){let n={x:_(e,"x","elu","float32")};return M.runKernel(Ao,n)}var Lu=z({elu_:GO});function HO(e){let t=_(e,"x","erf");D(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=ce(t,"float32"));let n={x:t};return M.runKernel(Kc,n)}var wx=z({erf_:HO});function Ix(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function aT(e,t,n){let r=e.length+t.length,s=[],a=0,o=0;for(let i=0;i<r;i++)n.indexOf(i)===-1?s.push(e[a++]):s.push(t[o++]);return s}function oT(e,t){let n=[],r=e.length;for(let a=0;a<r;a++)t.indexOf(a)===-1&&n.push(e[a]);let s=t.map(a=>e[a]);return[n,s]}function io(e,t){let n=t.map(r=>1);return aT(e,n,t)}function qO(e,t,n){D(Ix(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function iT(e,t){if(Ix(e,t))return null;let n=[];for(let r=0;r<t;++r)e.indexOf(r)===-1&&n.push(r);return e.forEach(r=>n.push(r)),n}function kx(e){return e.map((t,n)=>[n,t]).sort((t,n)=>t[1]-n[1]).map(t=>t[0])}function jO(e,t){let n=[];for(let r=t-e;r<t;++r)n.push(r);return n}function KO(e,t=null,n=!1){let s={x:_(e,"x","max")},a={reductionIndices:t,keepDims:n};return M.runKernel(zo,s,a)}var pr=z({max_:KO});function XO(e,t=null,n=!1){let s={x:_(e,"x","min")},a={axis:t,keepDims:n};return M.runKernel(Uo,s,a)}var yc=z({min_:XO});function YO(e,t){let n=_(e,"base","pow"),r=_(t,"exp","pow");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(Xo,s)}var $s=z({pow_:YO});function xe(e,t){if((fn(e)&&t!=="string"||Array.isArray(e))&&t!=="complex64")throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if(t==="string"&&fn(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return wa(e,[],[],t)}function ZO(e){let n={x:_(e,"x","sqrt","float32")};return M.runKernel(ii,n)}var un=z({sqrt_:ZO});function JO(e){let t=_(e,"x","square"),n={};return M.runKernel("Square",{x:t},n)}var it=z({square_:JO});function QO(e,t=null,n=!1){let r=_(e,"x","sum");r.dtype==="bool"&&(r=ce(r,"int32"));let s={x:r},a={axis:t,keepDims:n};return M.runKernel(ci,s,a)}var ge=z({sum_:QO});function eM(e,t="euclidean",n=null,r=!1){e=_(e,"x","norm");let s=cT(e,t,n),a=s.shape;if(r){let o=Fr(n,e.shape);a=io(s.shape,o)}return W(s,a)}function cT(e,t,n=null){if(e.rank===0)return Lt(e);if(e.rank!==1&&n===null)return cT(W(e,[-1]),t,n);if(e.rank===1||typeof n=="number"||Array.isArray(n)&&n.length===1){if(t===1)return ge(Lt(e),n);if(t===1/0)return pr(Lt(e),n);if(t===-1/0)return yc(Lt(e),n);if(t==="euclidean"||t===2)return un(ge($s(Lt(e),xe(2,"int32")),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(n)&&n.length===2){if(t===1)return pr(ge(Lt(e),n[0]),n[1]-1);if(t===1/0)return pr(ge(Lt(e),n[1]),n[0]);if(t===-1/0)return yc(ge(Lt(e),n[1]),n[0]);if(t==="fro"||t==="euclidean")return un(ge(it(e),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${n}`)}var zu=z({norm_:eM});function tM(e,t=null,n=!1){return zu(e,"euclidean",t,n)}var Sx=z({euclideanNorm_:tM});function nM(e){let n={x:_(e,"x","exp")};return M.runKernel($o,n)}var gn=z({exp_:nM});function rM(e,t=0){let n=_(e,"x","expandDims","string_or_numeric");D(t<=n.rank,()=>"Axis must be <= rank of the tensor");let r={input:n},s={dim:t};return M.runKernel(Yc,r,s)}var Zt=z({expandDims_:rM});function sM(e){let n={x:_(e,"x","expm1")};return M.runKernel(Zc,n)}var Tx=z({expm1_:sM});function aM(e,t){let n=_(e,"x","tile","string_or_numeric");D(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);let r={x:n},s={reps:t};return M.runKernel(va,r,s)}var Ln=z({tile_:aM});function oM(e,t,n,r="float32"){t==null&&(t=e);let s=Me([e,t],r),a=e<=t?e:t;for(let i=0;i<a;++i)s.set(1,i,i);let o=W(s.toTensor(),[e,t]);if(n==null)return o;if(n.length===1)return Ln(Zt(o,0),[n[0],1,1]);if(n.length===2)return Ln(Zt(Zt(o,0),0),[n[0],n[1],1,1]);if(n.length===3)return Ln(Zt(Zt(Zt(o,0),0),0),[n[0],n[1],n[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)}var Qf=z({eye_:oM});function iM(e){let n={x:_(e,"x","floor","float32")};return M.runKernel(Do,n)}var Bu=z({floor_:iM});function cM(e,t,n=0,r=0){let s=_(e,"x","gather"),a=_(t,"indices","gather","int32"),o={x:s,indices:a},i={axis:n,batchDims:r};return M.runKernel(Qc,o,i)}var Wu=z({gather_:cM});function uM(e,t){let n=_(e,"a","greater","string_or_numeric"),r=_(t,"b","greater","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(tu,s)}var An=z({greater_:uM});function lM(e,t){let n=_(e,"a","greaterEqual","string_or_numeric"),r=_(t,"b","greaterEqual","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(Po,s)}var Ds=z({greaterEqual_:lM});function dM(e){let n={x:_(e,"x","isFinite")};return M.runKernel(nu,n)}var Cx=z({isFinite_:dM});function pM(e){let n={x:_(e,"x","isInf")};return M.runKernel(ru,n)}var Nx=z({isInf_:pM});function hM(e){let n={x:_(e,"x","isNaN")};return M.runKernel(su,n)}var _x=z({isNaN_:hM});function fM(e,t=.2){let r={x:_(e,"x","leakyRelu")},s={alpha:t};return M.runKernel(Mo,r,s)}var Ld=z({leakyRelu_:fM});function mM(e,t){let n=_(e,"a","less","string_or_numeric"),r=_(t,"b","less","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(au,s)}var em=z({less_:mM});function gM(e,t){let n=_(e,"a","lessEqual","string_or_numeric"),r=_(t,"b","lessEqual","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(ou,s)}var Ta=z({lessEqual_:gM});function uT(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");let r={start:e,stop:t,num:n};return M.runKernel(Ef,{},r)}function bM(e,t=5,n=1,r=1,s=.5){let a=_(e,"x","localResponseNormalization");D(a.rank===4||a.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
+               rank ${a.rank}.`),D(hc(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let o=a,i=!1;a.rank===3&&(i=!0,o=W(a,[1,a.shape[0],a.shape[1],a.shape[2]]));let c={x:o},u={depthRadius:t,bias:n,alpha:r,beta:s},l=M.runKernel(bd,c,u);return i?W(l,[l.shape[1],l.shape[2],l.shape[3]]):l}var Ex=z({localResponseNormalization_:bM});function yM(e){let n={x:_(e,"x","log","float32")};return M.runKernel(Lo,n)}var nr=z({log_:yM});function vM(e){let n={x:_(e,"x","log1p")};return M.runKernel(iu,n)}var zd=z({log1p_:vM});function xM(e){return D(oa(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{let r=_(t,"x","tf.grad","string_or_numeric"),s=n!=null?_(n,"dy","tf.grad"):null;return M.tidy(()=>{let{value:a,grads:o}=M.gradients(()=>e(r),[r],s);return s!=null&&_n(a.shape,s.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),tm(o),o[0]})}}function wM(e){return D(oa(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{D(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let r=Jl(t,"args","tf.grads","string_or_numeric"),s=n!=null?_(n,"dy","tf.grads"):null;return M.tidy(()=>{let{value:a,grads:o}=M.gradients(()=>e(...r),r,s);return s!=null&&_n(a.shape,s.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),tm(o),o})}}function IM(e){return D(oa(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{D(t instanceof Te,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),D(n==null||n instanceof Te,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:r,value:s}=M.gradients(()=>e(t),[t],n);return tm(r),{grad:r[0],value:s}}}function kM(e){return D(oa(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{D(Array.isArray(t)&&t.every(s=>s instanceof Te),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),D(n==null||n instanceof Te,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let r=M.gradients(()=>e(...t),t,n);return n!=null&&_n(r.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),tm(r.grads),r}}function lT(e,t){D(oa(e),()=>"The f passed in variableGrads(f) must be a function"),D(t==null||Array.isArray(t)&&t.every(u=>u instanceof ia),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let n=t!=null;if(!n){t=[];for(let u in M.registeredVariables)t.push(M.registeredVariables[u])}let r=n?t.filter(u=>!u.trainable):null,s=t.length;t=t.filter(u=>u.trainable),D(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${s} variables is trainable.`);let a=!0,{value:o,grads:i}=M.gradients(e,t,null,a);D(i.some(u=>u!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),D(o.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${o.rank} tensor`);let c={};return t.forEach((u,l)=>{i[l]!=null&&(c[u.name]=i[l])}),r!=null&&r.forEach(u=>c[u.name]=null),{value:o,grads:c}}function ls(e){return M.customGrad(e)}function tm(e){if(e.filter(n=>n==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
+    the f you passed encloses all operations that lead from x to y.`)}function SM(e){let n={x:_(e,"x","softplus")};return M.runKernel(Nu,n)}var mi=z({softplus_:SM});function TM(e){let t=_(e,"x","logSigmoid");return ls(r=>({value:xt(mi(xt(r))),gradFunc:o=>B(o,dr(xt(r)))}))(t)}var Ax=z({logSigmoid_:TM});function CM(e,t){let n=_(e,"a","sub"),r=_(t,"b","sub");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(di,s)}var de=z({sub_:CM});function NM(e,t=-1){let n=_(e,"logits","logSoftmax");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);return ls((s,a)=>{let i=pr(s,t,!0),c=de(s,i),u=de(ce(c,"float32"),nr(ge(gn(c),t,!0)));return a([u]),{value:u,gradFunc:(p,d)=>{let[h]=d,f=!0,m=gn(h);return de(p,B(ge(p,t,f),m))}}})(n)}var nm=z({logSoftmax_:NM});function _M(e,t=null,n=!1){let r=_(e,"x","logSumExp"),s=Fr(t,r.shape),a=pr(r,s,!0),o=de(r,a),i=gn(o),c=ge(i,s),u=nr(c),l=Y(W(a,u.shape),u);if(n){let p=io(l.shape,s);return W(l,p)}return l}var rm=z({logSumExp_:_M});function EM(e,t){let n=_(e,"a","logicalAnd","bool"),r=_(t,"b","logicalAnd","bool");ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(cu,s)}var Dr=z({logicalAnd_:EM});function AM(e){let n={x:_(e,"x","logicalNot","bool")};return M.runKernel(uu,n)}var Bd=z({logicalNot_:AM});function $M(e,t){let n=_(e,"a","logicalOr","bool"),r=_(t,"b","logicalOr","bool");ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(lu,s)}var sm=z({logicalOr_:$M});function DM(e,t){let n=_(e,"a","logicalXor","bool"),r=_(t,"b","logicalXor","bool");return ct(n.shape,r.shape),Dr(sm(e,t),Bd(Dr(e,t)))}var $x=z({logicalXor_:DM}),oh=2147483648;function FM(e,t,n="left"){let r=_(e,"sortedSequence","searchSorted"),s=_(t,"values","searchSorted"),a=r.shape[r.shape.length-1],o=s.shape[s.shape.length-1],i=W(r,[-1,a]),c=W(s,[-1,o]);if(i.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(i.shape[0]!==c.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(ht(c.shape)>=oh)throw new Error(`values tensor size must less than ${oh}`);if(i.shape[1]>=oh)throw new Error(`trailing dim_size must less than ${oh} for int32 output type, was ${i.shape[1]}`);let u={sortedSequence:i,values:c},l={side:n};return M.runKernel(Wf,u,l)}var am=z({searchSorted_:FM});function dT(e,t){return am(e,t,"left")}function RM(e,t,n,r,s){let a=_(e,"x","maxPool"),o=1,i=a,c=!1;a.rank===3&&(c=!0,i=W(a,[1,a.shape[0],a.shape[1],a.shape[2]])),D(i.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${i.rank}.`),D(ps(n,o),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`),En("maxPool",r,s);let u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:s},p=M.runKernel(Wo,u,l);return c?W(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var Rt=z({maxPool_:RM});function PM(e,t=[1,1,1],n,r,s,a="NDHWC"){let o=_(e,"x","maxPool3d"),i=o,c=!1;o.rank===4&&(c=!0,i=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),D(i.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${i.rank}.`),D(a==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${a}`),En("maxPool3d",r,s);let u={x:i},l={filterSize:t,strides:n,pad:r,dimRoundingMode:s,dataFormat:a},p=M.runKernel(yd,u,l);return c?W(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var Dx=z({maxPool3d_:PM});function OM(e,t,n,r,s=!1){let o={x:_(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:n,pad:r,includeBatchInIndex:s},c=M.runKernel(Ff,o,i);return{result:c[0],indexes:c[1]}}var pT=z({maxPoolWithArgmax_:OM});function MM(e,t){let n=_(e,"a","maximum"),r=_(t,"b","maximum");[n,r]=_t(n,r),n.dtype==="bool"&&(n=ce(n,"int32"),r=ce(r,"int32")),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(Bo,s)}var hs=z({maximum_:MM});function LM(e,t=null,n=!1){let s={x:_(e,"x","mean")},a={axis:t,keepDims:n};return M.runKernel(Vo,s,a)}var Ct=z({mean_:LM});function kt(e,t="float32"){if(t==="complex64"){let r=kt(e,"float32"),s=kt(e,"float32");return As(r,s)}let n=df(ht(e),t);return M.makeTensor(n,e,t)}function Qn(e,t="float32"){if(t==="complex64"){let r=Qn(e,"float32"),s=kt(e,"float32");return As(r,s)}let n=_v(ht(e),t);return M.makeTensor(n,e,t)}function hT(e,t,{indexing:n="xy"}={}){if(n!=="xy"&&n!=="ij")throw new TypeError(`${n} is not a valid third argument to meshgrid`);if(e===void 0)return[];let r=_(e,"x","meshgrid",e instanceof Te?e.dtype:"float32");if(t===void 0)return[r];let s=_(t,"y","meshgrid",t instanceof Te?t.dtype:"float32"),a=ht(r.shape),o=ht(s.shape);return n==="xy"?(r=W(r,[1,-1]),s=W(s,[-1,1]),[De(Qn([o,1],r.dtype),r),De(s,Qn([1,a],s.dtype))]):(r=W(r,[-1,1]),s=W(s,[1,-1]),[De(r,Qn([1,o],r.dtype)),De(Qn([a,1],s.dtype),s)])}function zM(e,t){let n=_(e,"a","minimum"),r=_(t,"b","minimum");[n,r]=_t(n,r),n.dtype==="bool"&&(n=ce(n,"int32"),r=ce(r,"int32")),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(Go,s)}var Vu=z({minimum_:zM});function BM(e,t,n){D(n==="reflect"||n==="symmetric",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);let r=_(e,"x","mirrorPad");if(r.rank===0)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");D(t.length===r.rank,()=>`Padding doesn't match input. Must be ${r.rank}. Got ${t.length}.`);let s=n==="reflect"?1:0;for(let i=0;i<r.rank;i++)D(t[i].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),D(t[i][0]>=0&&t[i][0]<=r.shape[i]-s&&t[i][1]>=0&&t[i][1]<=r.shape[i]-s,()=>`Padding in dimension ${i} cannot be greater than or equal to ${r.shape[i]-s} or less than 0 for input of shape ${r.shape}`);let a={paddings:t,mode:n},o={x:r};return M.runKernel(Ho,o,a)}var Fx=z({mirrorPad_:BM});function WM(e,t){let n=_(e,"a","mod"),r=_(t,"b","mod");[n,r]=_t(n,r);let s={a:n,b:r};return M.runKernel(du,s)}var Rx=z({mod_:WM});function VM(e,t=null,n=!1){e=_(e,"x","moments");let r=Fr(t,e.shape),s=Ct(e,r,n),a=s.shape;n||(a=io(s.shape,r));let o=it(de(ce(e,"float32"),W(s,a))),i=Ct(o,r,n);return{mean:s,variance:i}}var Wd=z({moments_:VM});function UM(e,t,n,r){let s=_(t,"data","multiRNNCell"),a=Jl(n,"c","multiRNNCell"),o=Jl(r,"h","multiRNNCell"),i=s,c=[];for(let p=0;p<e.length;p++){let d=e[p](i,a[p],o[p]);c.push(d[0]),c.push(d[1]),i=d[1]}let u=[],l=[];for(let p=0;p<c.length;p+=2)u.push(c[p]),l.push(c[p+1]);return[u,l]}var fT=z({multiRNNCell_:UM});function GM(e,t,n,r=!1){let s=_(e,"logits","multinomial"),a=s.size,o=s.rank;if(a<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${a}.`);if(o>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${o}`);n=n||Math.random();let c={logits:o===1?W(s,[1,-1]):s},u={numSamples:t,seed:n,normalized:r},l=M.runKernel(Rf,c,u);return o===1?W(l,[l.size]):l}var mT=z({multinomial_:GM});function HM(e,t){let n=_(e,"a","notEqual","string_or_numeric"),r=_(t,"b","notEqual","string_or_numeric");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r};return M.runKernel(hu,s)}var co=z({notEqual_:HM});function qM(e){let n={x:_(e,"x","onesLike")};return M.runKernel(bu,n)}var rr=z({onesLike_:qM});function jM(e,t){let n=_(e,"v1","outerProduct"),r=_(t,"v2","outerProduct");D(n.rank===1&&r.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${r.rank}.`);let s=W(n,[-1,1]),a=W(r,[1,-1]);return De(s,a)}var gT=z({outerProduct_:jM});function KM(e,t,n=0){let r=_(e,"x","pad");if(r.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let s={paddings:t,constantValue:n},a={x:r};return M.runKernel(Ko,a,s)}var gr=z({pad_:KM});function XM(e,t,n=0){return D(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),gr(e,[t],n)}var bT=z({pad1d_:XM});function YM(e,t,n=0){return D(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),gr(e,t,n)}var yT=z({pad2d_:YM});function ZM(e,t,n=0){return D(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),gr(e,t,n)}var vT=z({pad3d_:ZM});function JM(e,t,n=0){return D(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),gr(e,t,n)}var xT=z({pad4d_:JM});function QM(e,t,n){let r=_(e,"x","spaceToBatchND");D(r.rank>=1+t.length,()=>`input rank ${r.rank} should be > than [blockShape] ${t.length}`),D(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),D(r.shape.reduce((o,i,c)=>c>0&&c<=t.length?o&&(i+n[c-1][0]+n[c-1][1])%t[c-1]===0:o,!0),()=>`input spatial dimensions ${r.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);let s={x:r},a={blockShape:t,paddings:n};return M.runKernel(_u,s,a)}var Vd=z({spaceToBatchND_:QM});function e3(e,t,n,r,s,a,o){s==null&&(s=[1,1]),a==null&&(a=1),r===0&&(r="valid");let i=_(e,"x","maxPool"),c=i,u=!1;i.rank===3&&(u=!0,c=W(i,[1,i.shape[0],i.shape[1],i.shape[2]])),D(ps(a,s),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${a} and dilations '${s}'`);let l=ZS(c.shape,t,a,s,r),p=[l.dilationHeight,l.dilationWidth],d;r==="same"?d=n3([l.filterHeight,l.filterWidth],p):d=[[0,0],[0,0]];let h=p[0]===1&&p[1]===1,[f,m]=t3([l.inHeight,l.inWidth],p,d),g=h?r:"valid",b=h?c:Vd(c,p,f),v=(n==="avg"?()=>mr(b,t,a,g,o):()=>Rt(b,t,a,g,o))(),x=h?v:Od(v,p,m);return u?W(x,[x.shape[1],x.shape[2],x.shape[3]]):x}function t3(e,t,n){let r=n.map(l=>l[0]),s=n.map(l=>l[1]),a=e.concat(r,s),o=t.map((l,p)=>(l-a[p]%l)%l),i=s.map((l,p)=>l+o[p]),c=t.map((l,p)=>[r[p],i[p]]),u=t.map((l,p)=>[0,o[p]]);return[c,u]}function n3(e,t){let r=e.map((o,i)=>o+(o-1)*(t[i]-1)).map(o=>o-1),s=r.map(o=>Math.floor(o/2)),a=r.map((o,i)=>o-s[i]);return r.map((o,i)=>[s[i],a[i]])}var Px=z({pool_:e3});function r3(e,t){let n=_(e,"x","prelu"),r=_(t,"alpha","prelu"),s={x:n,alpha:r};return M.runKernel(Yo,s)}var Ud=z({prelu_:r3});function s3(e,t=null,n=!1){let r=_(e,"x","prod");r.dtype==="bool"&&(r=ce(r,"int32"));let s={x:r},a={axis:t,keepDims:n};return M.runKernel(Zo,s,a)}var Ox=z({prod_:s3});function a3(e,t,n,r){let s=e.map((l,p)=>_(l,`tensors${p}`,"raggedGather","int32")),a=_(t,"paramsDenseValues","raggedGather"),o=_(n,"indices","raggedGather","int32"),i={paramsNestedSplits:s,paramsDenseValues:a,indices:o},c={outputRaggedRank:r},u=M.runKernel(Pf,i,c);return{outputNestedSplits:u.slice(0,u.length-1),outputDenseValues:u[u.length-1]}}var wT=z({raggedGather_:a3});function o3(e,t,n){let r=_(e,"starts","raggedRange"),s=_(t,"limits","raggedRange",r.dtype),a=_(n,"deltas","raggedRange",r.dtype),o={starts:r,limits:s,deltas:a},i=M.runKernel(Of,o);return{rtNestedSplits:i[0],rtDenseValues:i[1]}}var IT=z({raggedRange_:o3});function i3(e,t,n,r,s){let a=_(e,"shape","raggedTensorToTensor","int32"),o=_(t,"values","raggedTensorToTensor"),i=_(n,"defaultValue","raggedTensorToTensor",o.dtype),c=r.map((p,d)=>_(p,`tensors${d}`,"raggedTensorToTensor","int32")),u={shape:a,values:o,defaultValue:i,rowPartitionTensors:c},l={rowPartitionTypes:s};return M.runKernel(Mf,u,l)}var kT=z({raggedTensorToTensor_:i3});function c3(e,t,n){let r=ht(e),s=null;if(n==null||n==="float32")s=new Float32Array(r);else if(n==="int32")s=new Int32Array(r);else if(n==="bool")s=new Uint8Array(r);else throw new Error(`Unknown data type ${n}`);for(let a=0;a<r;a++)s[a]=t();return M.makeTensor(s,e,n)}var ST=z({rand_:c3}),Mx=ma(cf()),Lx=class{constructor(e,t,n,r,s){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=r,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let a=s||Math.random();this.random=Mx.alea(a.toString())}nextValue(){if(!isNaN(this.nextVal)){let r=this.nextVal;return this.nextVal=NaN,r}let e,t,n=!1;for(;!n;){let r,s,a;do r=2*this.random()-1,s=2*this.random()-1,a=r*r+s*s;while(a>=1||a===0);let o=Math.sqrt(-2*Math.log(a)/a);e=this.mean+this.stdDev*r*o,t=this.mean+this.stdDev*s*o,(!this.truncated||this.isValidTruncated(e))&&(n=!0)}return(!this.truncated||this.isValidTruncated(t))&&(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return this.dtype==null||this.dtype==="float32"?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}},u3=class{constructor(e,t,n,r){this.alpha=e,this.beta=1/t,this.dtype=n;let s=r||Math.random();this.randu=Mx.alea(s.toString()),this.randn=new Lx(0,1,n,!1,this.randu()),e<1?this.d=e+2/3:this.d=e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,n,r,s,a;for(;;){do r=this.randn.nextValue(),a=1+this.c*r;while(a<=0);if(a*=a*a,e=r*r,t=1-.331*e*e,n=.5*e+this.d*(1-a+Math.log(a)),s=this.randu(),s<t||Math.log(s)<n)break}return a=1/this.beta*this.d*a,this.alpha<1&&(a*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(a)}convertValue(e){return this.dtype==="float32"?e:Math.round(e)}},l3=class{constructor(e=0,t=1,n,r){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=e,this.range=t-e,this.dtype=n,r==null&&(r=Math.random()),typeof r=="number"&&(r=r.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=Mx.alea(r)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function d3(e,t,n=1,r="float32",s){if(n==null&&(n=1),r==null&&(r="float32"),r!=="float32"&&r!=="int32")throw new Error(`Unsupported data type ${r}`);let a=new u3(t,n,r,s),o=Me(e,r);for(let i=0;i<o.values.length;i++)o.values[i]=a.nextValue();return o.toTensor()}var TT=z({randomGamma_:d3});function p3(e,t=0,n=1,r,s){if(r!=null&&r==="bool")throw new Error(`Unsupported data type ${r}`);let a=new Lx(t,n,r,!1,s),o=Me(e,r);for(let i=0;i<o.values.length;i++)o.values[i]=a.nextValue();return o.toTensor()}var om=z({randomNormal_:p3});function h3(e,t,n){if(t!=null&&t==="bool")throw new Error(`Unsupported data type ${t}`);return om(e,0,1,t,n)}var CT=z({randomStandardNormal_:h3});function f3(e,t=0,n=1,r="float32",s){let a=Me(e,r),o=new l3(t,n,null,s);for(let i=0;i<a.values.length;i++)a.values[i]=o.nextValue();return a.toTensor()}var Uu=z({randomUniform_:f3});function vc(e,t,n=1,r="float32"){if(n===0)throw new Error("Cannot have a step of zero");let s={start:e,stop:t,step:n,dtype:r};return M.runKernel(vd,{},s)}function m3(e){let n={x:_(e,"x","reciprocal")};return M.runKernel(vu,n)}var zx=z({reciprocal_:m3});function g3(e){let n={x:_(e,"x","relu")};return M.runKernel(Jo,n)}var Xe=z({relu_:g3});function b3(e){let n={x:_(e,"x","relu6")};return M.runKernel(ti,n)}var im=z({relu6_:b3});function y3(e,t){let r={x:_(e,"x","reverse")},s={dims:t};return M.runKernel(ni,r,s)}var fr=z({reverse_:y3});function v3(e){let t=_(e,"x","reverse");return D(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),fr(t,0)}var NT=z({reverse1d_:v3});function x3(e,t){let n=_(e,"x","reverse");return D(n.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),fr(n,t)}var _T=z({reverse2d_:x3});function w3(e,t){let n=_(e,"x","reverse");return D(n.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),fr(n,t)}var ET=z({reverse3d_:w3});function I3(e,t){let n=_(e,"x","reverse");return D(n.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),fr(n,t)}var AT=z({reverse4d_:I3});function k3(e){let n={x:_(e,"x","round")};return M.runKernel(ri,n)}var cm=z({round_:k3});function S3(e){let n={x:_(e,"x","rsqrt","float32")};return M.runKernel(si,n)}var um=z({rsqrt_:S3});function T3(e){let n={x:_(e,"x","selu")};return M.runKernel(ku,n)}var lm=z({selu_:T3});function C3(e,t,n,r,s,a=[1,1],o="NHWC"){let i=_(e,"x","separableConv2d"),c=_(t,"depthwiseFilter","separableConv2d"),u=_(n,"pointwiseFilter","separableConv2d"),l=i,p=!1;if(i.rank===3&&(p=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2]])),o==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");D(l.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${l.rank}.`),D(c.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${c.rank}.`),D(u.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${c.rank}.`),D(u.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${u.shape[0]}.`),D(u.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${u.shape[1]}.`);let d=c.shape[2],h=c.shape[3];D(u.shape[2]===d*h,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${d*h}, but got ${u.shape[2]}.`);let f=Sa(l,c,r,s,o,a),g=Dt(f,u,1,"valid",o);return p?W(g,[g.shape[1],g.shape[2],g.shape[3]]):g}var Ca=z({separableConv2d_:C3});async function N3(e,t){let n=_(e,"x","setdiff1d"),r=_(t,"y","setdiff1d");D(n.dtype===r.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${r.dtype}).`),D(n.rank===1,()=>`x should be 1D tensor, but got x (${n.shape}).`),D(r.rank===1,()=>`y should be 1D tensor, but got y (${r.shape}).`);let s=await n.data(),a=await r.data(),o=new Set(a),i=0;for(let l=0;l<s.length;l++)o.has(s[l])||i++;let c=new Ht([i],n.dtype),u=new Ht([i],"int32");for(let l=0,p=0;l<s.length;l++)o.has(s[l])||(c.values[p]=s[l],u.values[p]=l,p++);return[c.toTensor(),u.toTensor()]}var $T=N3;function _3(e){let n={x:_(e,"x","sign")};return M.runKernel(Cu,n)}var Bx=z({sign_:_3});function E3(e){let n={x:_(e,"x","sin","float32")};return M.runKernel(ai,n)}var dm=z({sin_:E3});function A3(e){let n={x:_(e,"x","sinh")};return M.runKernel(Tu,n)}var pm=z({sinh_:A3});function $3(e,t,n){let r=_(e,"x","slice1d");return D(r.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${r.rank} tensor`),We(r,[t],[n])}var Gd=z({slice1d_:$3});function D3(e,t,n){let r=_(e,"x","slice2d");return D(r.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${r.rank} tensor`),We(r,t,n)}var hm=z({slice2d_:D3});function F3(e,t,n){let r=_(e,"x","slice3d");return D(r.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${r.rank} tensor`),We(r,t,n)}var gi=z({slice3d_:F3});function R3(e,t,n){let r=_(e,"x","slice4d");return D(r.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${r.rank} tensor`),We(r,t,n)}var xc=z({slice4d_:R3});function P3(e,t=-1){let n=_(e,"logits","softmax","float32");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);let r={logits:n},s={dim:t};return M.runKernel(ui,r,s)}var Xr=z({softmax_:P3});function O3(e){D(e.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);let t={input:e};return M.runKernel(Cf,t)}var Hd=z({fft_:O3});function M3(e){D(e.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);let t={input:e};return M.runKernel(Nf,t)}var wc=z({ifft_:M3});function L3(e){let t=e.shape[e.shape.length-1],n=e.size/t,r;if(t<=2){let s=W(e,[n,t]);r=wc(s)}else{let s=[n,2*(t-1)],a=W(bc(e),[n,t]),o=W(Fd(e),[n,t]),i=fr(We(a,[0,1],[n,t-2]),1),c=B(fr(We(o,[0,1],[n,t-2]),1),xe(-1)),u=Je([a,i],1),l=Je([o,c],1),p=W(As(u,l),[s[0],s[1]]);r=wc(p)}if(r=bc(r),e.rank===3&&e.shape[0]!==0){let s=r,a=e.shape[0];r=W(r,[a,r.shape[0]/a,r.shape[1]]),s.dispose()}return r}var fm=z({irfft_:L3});function z3(e,t,n=0){let s={x:_(e,"x","split")},a={numOrSizeSplits:t,axis:n};return M.runKernel(Eu,s,a)}var zn=z({split_:z3});function B3(e,t){D(e.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let n=e.shape[e.shape.length-1],r=e.size/n,s;if(t!=null&&t<n){let f=e.shape.map(g=>0),m=e.shape.map(g=>g);m[e.shape.length-1]=t,s=We(e,f,m),n=t}else if(t!=null&&t>n){let f=e.shape.map(m=>m);f[e.shape.length-1]=t-n,s=Je([e,kt(f)],e.shape.length-1),n=t}else s=e;let a=qe(s),o=W(As(s,a),[r,n]),i=Hd(o),c=Math.floor(n/2)+1,u=bc(i),l=Fd(i),p=zn(u,[c,n-c],u.shape.length-1),d=zn(l,[c,n-c],l.shape.length-1),h=s.shape.slice();return h[s.shape.length-1]=c,W(As(p[0],d[0]),h)}var qd=z({rfft_:B3});function W3(e,t){let n=_(e,"a","squaredDifference"),r=_(t,"b","squaredDifference");[n,r]=_t(n,r),ct(n.shape,r.shape);let s={a:n,b:r},a={};return M.runKernel(li,s,a)}var mm=z({squaredDifference_:W3});function V3(e,t){let n=_(e,"x","squeeze","string_or_numeric");return W(n,q1(n.shape,t).newShape)}var Na=z({squeeze_:V3});function U3(e,t=0){let n=Jl(e,"tensors","stack","string_or_numeric");D(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length>0&&D(t<=n[0].rank,()=>"Axis must be <= rank of the tensor");let r=n,s={axis:t};return M.runKernel(yu,r,s)}var Ft=z({stack_:U3});function G3(e,t=0){let r={x:_(e,"x","step")},s={alpha:t};return M.runKernel(xa,r,s)}var bi=z({step_:G3});function H3(e,t,n,r,s=0,a=0,o=0,i=0,c=0){let l={x:_(e,"x","stridedSlice","string_or_numeric")},p={begin:t,end:n,strides:r,beginMask:s,endMask:a,ellipsisMask:o,newAxisMask:i,shrinkAxisMask:c};return M.runKernel($u,l,p)}var Wx=z({stridedSlice_:H3});function q3(e){let n={x:_(e,"x","tan","float32")};return M.runKernel(pi,n)}var Vx=z({tan_:q3});function Ke(e,t){go(e);let n=us(e,t);if(n.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");return wa(e,null,n,t)}function $r(e,t,n){if(go(e),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");let r=us(e,n);if(r.length!==2&&r.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return wa(e,t,r,n)}function Rr(e,t,n){if(go(e),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");let r=us(e,n);if(r.length!==4&&r.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return wa(e,t,r,n)}function DT(e,t,n){if(go(e),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");let r=us(e,n);if(r.length!==5&&r.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return wa(e,t,r,n)}function FT(e,t,n){if(go(e),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");let r=us(e,n);if(r.length!==6&&r.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||r,wa(e,t,r,n)}function j3(e,t=1,n=!0){let r=_(e,"x","topk");if(r.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");let s=r.shape[r.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>s)throw new Error(`'k' passed to topk() must be <= the last dimension (${s}) but got ${t}`);let a={x:r},o={k:t,sorted:n},[i,c]=M.runKernel(Du,a,o);return{values:i,indices:c}}var Ux=z({topk_:j3});function K3(e,t=0,n=1,r,s){if(r!=null&&r==="bool")throw new Error("Unsupported data type $ { dtype }");let a=new Lx(t,n,r,!0,s),o=Me(e,r);for(let i=0;i<o.values.length;i++)o.values[i]=a.nextValue();return o.toTensor()}var gm=z({truncatedNormal_:K3});function X3(e,t=0){let n=_(e,"x","unique","string_or_numeric");D(n.rank>0,()=>"The input tensor must be at least 1D");let r={x:n},s={axis:t},[a,o]=M.runKernel(Uf,r,s);return{values:a,indices:o}}var Gx=z({unique_:X3});function Y3(e,t,n){let r=_(e,"x","unsortedSegmentSum"),s=_(t,"segmentIds","unsortedSegmentSum","int32");D(hc(n),()=>"numSegments must be of dtype int");let a={x:r,segmentIds:s},o={numSegments:n};return M.runKernel(Nd,a,o)}var bm=z({unsortedSegmentSum_:Y3});function Z3(e,t=0){let n=_(e,"x","unstack","string_or_numeric");D(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);let r={value:n},s={axis:t};return M.runKernel(Ru,r,s)}var lt=z({unstack_:Z3});function RT(e,t){return am(e,t,"right")}function Hx(e,t=!0,n,r){return M.makeVariable(e,t,n,r)}function PT(e,t){let n=[];for(let a=0;a<t.length;a++)t[a]&&n.push(a);let r=Me(e,"int32"),s=Me([n.length,e.length],"int32");for(let a=0;a<n.length;a++){let o=r.indexToLoc(n[a]),i=a*e.length;s.values.set(o,i)}return s.toTensor()}async function J3(e){let t=_(e,"condition","whereAsync","bool"),n=await t.data(),r=PT(t.shape,n);return e!==t&&t.dispose(),r}var qx=J3;async function Q3(e,t,n){let r=_(e,"tensor","boolMask"),s=_(t,"mask","boolMask","bool"),a=n==null?0:n,o=s.rank,i=r.shape;D(o>0,()=>"mask cannot be scalar"),_n(i.slice(a,a+o),s.shape,"mask's shape must match the first K dimensions of tensor's shape,");let c=1;for(let m=a;m<a+o;m++)c*=i[m];let u=i.slice(0,a).concat([c],i.slice(a+o)),l=W(r,u),p=W(s,[-1]),d=await qx(p),h=Na(d,[1]),f=Wu(l,h,a);return e!==r&&r.dispose(),t!==s&&s.dispose(),h.dispose(),l.dispose(),p.dispose(),d.dispose(),f}var OT=Q3;function eL(e,t,n,r,s=!0){let a=_(e,"v","movingAverage"),o=_(t,"x","movingAverage"),i=_(n,"decay","movingAverage");lS(a,o),D(ga(a.shape,o.shape),()=>"Shape mismatch in v and x");let c=xe(1),u=de(c,i),l=B(de(o,a),u);if(s){D(r!=null,()=>"When using zeroDebias: true, step is required.");let p=_(r,"step","movingAverage");l=fe(l,de(c,$s(i,p)))}return Y(a,l)}var MT=z({movingAverage_:eL});function tL(e,t,n){let r=_(e,"indices","scatterND","int32"),s=_(t,"updates","scatterND");qv(s,r,n);let a={indices:r,updates:s},o={shape:n};return M.runKernel(wu,a,o)}var LT=z({scatterND_:tL});function nL(e,t,n,r){if(e.dtype!=="int32")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);let s=e.rank>0?e.shape[0]:1,a=e.rank>1?e.shape[1]:1;if(n.length!==a)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${a}.`);let o=t.size;if(!(t.rank===0||t.rank===1&&o===s))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${s}]`);if(t.dtype!==r.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function rL(e,t,n,r=0){let s=_(e,"sparseIndices","sparseToDense","int32"),a=_(t,"sparseValues","sparseToDense","string_or_numeric"),o=_(r,"defaultValue","sparseToDense",a.dtype);nL(s,a,n,o);let i={sparseIndices:s,sparseValues:a,defaultValue:o},c={outputShape:n};return M.runKernel(Vf,i,c)}var zT=z({sparseToDense_:rL});function sL(e,t){let n=_(t,"indices","gatherND","int32"),s={params:_(e,"x","gatherND","string_or_numeric"),indices:n};return M.runKernel(eu,s)}var BT=z({gatherND_:sL});function aL(e,t){if(t==null)return e.shape.slice();if(ga(e.shape,t))return t;if(e.shape.length===t.length){let n=[];for(let r=0;r<e.shape.length;r++)t[r]==null&&e.shape[r]!=null?n.push(e.shape[r]):n.push(t[r]);return n}return t}function oL(e,t,n,r){let s=_(e,"x","dropout");if(D(s.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${s.dtype} tensor instead.`),D(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return e instanceof Te?s.clone():s;let a=aL(s,n),o=1-t,i=fe(Bu(Y(Uu(a,0,1,"float32",r),o)),o);return B(s,i)}var jx=z({dropout_:oL});function Kx(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function ym(e,t,n){let r=1-e%2,s=new Float32Array(e);for(let a=0;a<e;++a){let o=2*Math.PI*a/(e+r-1);s[a]=t-n*Math.cos(o)}return Ke(s,"float32")}async function iL(e,t,n=1){let r=_(e,"predictions","inTopK"),s=_(t,"targets","inTopK");D(r.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${r.rank}`),D(r.rank-1===s.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${r.rank} and targets rank ${s.rank}`),_n(r.shape.slice(0,r.shape.length-1),s.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let a=r.shape[r.shape.length-1];D(n>0&&n<=a,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${a}), but got ${n}`);let o=await r.data(),i=await s.data(),[c,u]=[o.length/a,a],l=j1("bool",c);for(let p=0;p<c;p++){let d=p*u,h=o.subarray(d,d+u),f=[];for(let m=0;m<h.length;m++)f.push({value:h[m],index:m});f.sort((m,g)=>g.value-m.value),l[p]=0;for(let m=0;m<n;m++)if(f[m].index===i[p]){l[p]=1;break}}return e!==r&&r.dispose(),t!==s&&s.dispose(),Cn(l,s.shape,"bool")}var WT=iL,Ic={};Ae(Ic,{conv2d:()=>lL,depthwiseConv2d:()=>fL,matMul:()=>gL});function cL(e,t,n,r,s,a="NHWC",o){let i=e;e.rank===3&&(i=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let c=t;c.rank===3&&(c=W(t,[1,t.shape[0],t.shape[1],t.shape[2]])),D(i.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${i.shape}.`),D(c.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${c.shape}.`),D(n.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);let u=a==="NHWC"?i.shape[3]:i.shape[1],l=a==="NHWC"?c.shape[3]:c.shape[1];D(u===n[2],()=>`Error in conv2dDerFilter: depth of input ${u}) must match input depth in filter (${n[2]}.`),D(l===n[3],()=>`Error in conv2dDerFilter: depth of dy (${l}) must match output depth for filter (${n[3]}).`),En("conv2dDerFilter",s,o);let p={x:i,dy:c},d={strides:r,pad:s,dataFormat:a,dimRoundingMode:o,filterShape:n};return M.runKernel(bf,p,d)}var Xx=z({conv2DBackpropFilter_:cL});function vm(e,t,n){if(n==null||n==="linear")return e;if(n==="relu")return B(e,bi(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function xm(e,t){let n=t,r=Bt(e.shape,t.shape);return r.length>0&&(n=ge(n,r)),W(n,e.shape)}function wm(e,t,n,r){if(t==="linear")return e;if(t==="relu")return Xe(e);if(t==="elu")return Lu(e);if(t==="relu6")return im(e);if(t==="prelu")return Ud(e,n);if(t==="leakyrelu")return Ld(e,r);if(t==="sigmoid")return dr(e);throw new Error(`Unknown fused activation ${t}.`)}var Im=(e,t)=>!(e>0)||t==="linear";function uL({x:e,filter:t,strides:n,pad:r,dataFormat:s="NHWC",dilations:a=[1,1],dimRoundingMode:o,bias:i,activation:c="linear",preluActivationWeights:u,leakyreluAlpha:l}){if(c=c||"linear",Im(M.state.gradientDepth,c)===!1){D(s==="NHWC",()=>`Error in fused conv2d: got dataFormat of ${s} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let S=Dt(e,t,n,r,s,a,o);return i!=null&&(S=Y(S,i)),wm(S,c,u,l)}let p=_(e,"x","conv2d","float32"),d=_(t,"filter","conv2d","float32"),h=p,f=!1;p.rank===3&&(f=!0,h=W(p,[1,p.shape[0],p.shape[1],p.shape[2]])),D(h.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${h.rank}.`),D(d.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${d.rank}.`),En("fused conv2d",r,o);let m=s==="NHWC"?h.shape[3]:h.shape[1];D(d.shape[2]===m,()=>`Error in conv2d: depth of input (${m}) must match input depth for filter ${d.shape[2]}.`),D(ps(n,a),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`);let g=Pd(h.shape,d.shape,n,a,r,o),b;i!=null&&(b=_(i,"bias","fused conv2d"),[b]=_t(b,p),s==="NHWC"?ct(g.outShape,b.shape):(D(b.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${b.shape.length}.`),D(b.shape.length===0||b.shape[0]===g.outChannels||b.shape[0]===1,()=>`Error in fused conv2d: bias shape (${b.shape}) is not compatible with the number of output channels (${g.outChannels})`)));let y;if(u!=null){let S=u.shape;if(D(S.length<=1||S.length===3,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${S.length}.`),S.length===1)D(S[0]===1||S[0]===g.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${S}) is not compatible with the number of output channels (${g.outChannels}).`);else if(S.length===3)try{ct(S,g.outShape)}catch(C){let E=`Error in fused conv2d: PReLU activation weights (${S}) is not compatible with the output shape of the conv2d (${g.outShape}).`;throw Error(E)}y=_(u,"prelu weights","fused conv2d")}let v=(S,C)=>{D(s==="NHWC",()=>`Error in gradient of fused conv2D: got dataFormat of ${s} but only NHWC is currently supported.`);let[E,$,F,A]=C,R=vm(S,F,c);D(ca(a),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${a}'`);let T=fx($.shape,R,E,n,r),L=Xx($,R,E.shape,n,r),V=[T,L];if(A!=null){let G=xm(A,R);V.push(G)}return V},x={x:h,filter:d,bias:b,preluActivationWeights:y},k={strides:n,pad:r,dataFormat:s,dilations:a,dimRoundingMode:o,activation:c,leakyreluAlpha:l};return i==null?ls((C,E,$)=>{let F=M.runKernel(to,x,k);return $([E,C,F]),f&&(F=W(F,[F.shape[1],F.shape[2],F.shape[3]])),{value:F,gradFunc:v}})(h,d):ls((C,E,$,F)=>{let A=M.runKernel(to,x,k);return F([E,C,A,$]),f&&(A=W(A,[A.shape[1],A.shape[2],A.shape[3]])),{value:A,gradFunc:v}})(h,d,b)}var lL=z({fusedConv2d_:uL});function dL(e,t,n,r,s,a=[1,1],o){let i=e;e.rank===3&&(i=W(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let c=t;c.rank===3&&(c=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={x:i,dy:c},l={strides:r,pad:s,dimRoundingMode:o,dilations:a,filterShape:n};return M.runKernel(wf,u,l)}var VT=z({depthwiseConv2dNativeBackpropFilter_:dL});function pL(e,t,n,r,s,a=[1,1],o){let i=t,c=!1;t.rank===3&&(c=!0,i=W(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={dy:i,filter:n},l={strides:r,pad:s,dimRoundingMode:o,dilations:a,inputShape:e},p=M.runKernel(If,u,l);return c?W(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var UT=z({depthwiseConv2dNativeBackpropInput_:pL});function hL({x:e,filter:t,strides:n,pad:r,dataFormat:s="NHWC",dilations:a=[1,1],dimRoundingMode:o,bias:i,activation:c="linear",preluActivationWeights:u,leakyreluAlpha:l}){if(Im(M.state.gradientDepth,c)===!1){let k=Sa(e,t,n,r,s,a,o);return i!=null&&(k=Y(k,i)),wm(k,c,u,l)}let p=_(e,"x","depthwiseConv2d","float32"),d=_(t,"filter","depthwiseConv2d","float32"),h=p,f=!1;p.rank===3&&(f=!0,h=W(p,[1,p.shape[0],p.shape[1],p.shape[2]])),D(h.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${h.rank}.`),D(d.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${d.rank}.`),D(h.shape[3]===d.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${h.shape[3]}) must match the inChannels dimension in filter ${d.shape[2]}.`),a==null&&(a=[1,1]),D(ps(n,a),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`),En("fused depthwiseConv2d",r,o);let m=Pd(h.shape,d.shape,n,a,r,o,!0),g;i!=null&&(g=_(i,"bias","fused conv2d"),[g]=_t(g,p),ct(m.outShape,g.shape));let b;u!=null&&(b=_(u,"prelu weights","fused depthwiseConv2d"));let y=(k,S)=>{D(ca(a),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${a}'`);let[C,E,$,F]=S,A=vm(k,$,c),R=UT(E.shape,A,C,n,r,a,o),T=VT(E,A,C.shape,n,r,a,o);if(F!=null){let L=xm(g,A);return[R,T,L]}return[R,T]},v={x:h,filter:d,bias:g,preluActivationWeights:b},x={strides:n,pad:r,dataFormat:s,dilations:a,dimRoundingMode:o,activation:c,leakyreluAlpha:l};return i==null?ls((S,C,E)=>{let $=M.runKernel(no,v,x);return E([C,S,$]),f&&($=W($,[$.shape[1],$.shape[2],$.shape[3]])),{value:$,gradFunc:y}})(h,d):ls((S,C,E,$)=>{let F=M.runKernel(no,v,x);return $([C,S,F,E]),f&&(F=W(F,[F.shape[1],F.shape[2],F.shape[3]])),{value:F,gradFunc:y}})(h,d,g)}var fL=z({fusedDepthwiseConv2d_:hL});function mL({a:e,b:t,transposeA:n=!1,transposeB:r=!1,bias:s,activation:a="linear",preluActivationWeights:o,leakyreluAlpha:i=.2}){if(Im(M.state.gradientDepth,a)===!1){let A=De(e,t,n,r);return s!=null&&(A=Y(A,s)),wm(A,a,o,i)}let c=_(e,"a","fused matMul"),u=_(t,"b","fused matMul");[c,u]=_t(c,u);let l=n?c.shape[c.rank-2]:c.shape[c.rank-1],p=r?u.shape[u.rank-1]:u.shape[u.rank-2],d=n?c.shape[c.rank-1]:c.shape[c.rank-2],h=r?u.shape[u.rank-2]:u.shape[u.rank-1],f=c.shape.slice(0,-2),m=u.shape.slice(0,-2),g=ht(f),b=ht(m);D(l===p,()=>`Error in fused matMul: inner shapes (${l}) and (${p}) of Tensors with shapes ${c.shape} and ${u.shape} and transposeA=${n} and transposeB=${r} must match.`);let v=ct(c.shape.slice(0,-2),u.shape.slice(0,-2)).concat([d,h]),x=n?W(c,[g,l,d]):W(c,[g,d,l]),k=r?W(u,[b,h,p]):W(u,[b,p,h]),S;s!=null&&(S=_(s,"bias","fused matMul"),[S]=_t(S,c),ct(v,S.shape));let C;o!=null&&(C=_(o,"prelu weights","fused matMul"));let E=(A,R)=>{let[T,L,V,G]=R,j=vm(W(A,V.shape),V,a),H,Z;if(!n&&!r?(H=De(j,L,!1,!0),Z=De(T,j,!0,!1)):!n&&r?(H=De(j,L,!1,!1),Z=De(j,T,!0,!1)):n&&!r?(H=De(L,j,!1,!0),Z=De(T,j,!1,!1)):(H=De(L,j,!0,!0),Z=De(j,T,!0,!0)),s!=null){let J=xm(G,j);return[H,Z,J]}else return[H,Z]},$={a:x,b:k,bias:S,preluActivationWeights:C},F={transposeA:n,transposeB:r,activation:a,leakyreluAlpha:i};return s==null?ls((R,T,L)=>{let V=M.runKernel(eo,$,F);return L([R,T,V]),{value:W(V,v),gradFunc:E}})(x,k):ls((R,T,L,V)=>{let G=M.runKernel(eo,$,F);return V([R,T,G,L]),{value:W(G,v),gradFunc:E}})(x,k,S)}var gL=z({fusedMatMul_:mL});function bL(e){return ym(e,.54,.46)}var yL=z({hammingWindow_:bL});function vL(e){return ym(e,.5,.5)}var GT=z({hannWindow_:vL});function xL(e,t,n,r=!1,s=0){let a=0,o=[];for(;a+t<=e.size;)o.push(We(e,a,t)),a+=n;if(r)for(;a<e.size;){let i=a+t-e.size,c=Je([We(e,a,t-i),bn([i],s)]);o.push(c),a+=n}return o.length===0?$r([],[0,t]):W(Je(o),[o.length,t])}var HT=z({frame_:xL});function wL(e,t,n,r,s=GT){r==null&&(r=Kx(t));let a=HT(e,t,n),o=B(a,s(t));return qd(o,r)}var IL=z({stft_:wL});function kL(e,t,n,r,s="bilinear",a=0){let o=_(e,"image","cropAndResize"),i=_(t,"boxes","cropAndResize","float32"),c=_(n,"boxInd","cropAndResize","int32"),u=i.shape[0];D(o.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${o.rank}.`),D(i.rank===2&&i.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${u},4] but had shape ${i.shape}.`),D(c.rank===1&&c.shape[0]===u,()=>`Error in cropAndResize: boxInd must be have size [${u}] but had shape ${i.shape}.`),D(r.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${r.length}.`),D(r[0]>=1&&r[1]>=1,()=>`cropSize must be atleast [1,1], but was ${r}`),D(s==="bilinear"||s==="nearest",()=>`method must be bilinear or nearest, but was ${s}`);let l={image:o,boxes:i,boxInd:c},p={method:s,extrapolationValue:a,cropSize:r};return M.runKernel(qc,l,p)}var SL=z({cropAndResize_:kL});function TL(e){let t=_(e,"image","flipLeftRight","float32");D(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let n={image:t};return M.runKernel(Jc,n,{})}var CL=z({flipLeftRight_:TL});function NL(e){let t=_(e,"image","grayscaleToRGB"),n=t.rank-1,r=t.shape[n];D(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),D(r===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${r}.`);let s=new Array(t.rank);return s.fill(1,0,n),s[n]=3,Ln(t,s)}var _L=z({grayscaleToRGB_:NL});function EL(e,t,n=0,r=.5){let s=_(e,"image","rotateWithOffset","float32");D(s.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${s.rank}.`);let a={image:s},o={radians:t,fillValue:n,center:r};return M.runKernel(Ou,a,o)}var AL=z({rotateWithOffset_:EL});function Gu(e,t,n,r,s,a){r==null&&(r=.5),s==null&&(s=Number.NEGATIVE_INFINITY),a==null&&(a=0);let o=e.shape[0];return n=Math.min(n,o),D(0<=r&&r<=1,()=>`iouThreshold must be in [0, 1], but was '${r}'`),D(e.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),D(e.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),D(t.rank===1,()=>"scores must be a 1D tensor"),D(t.shape[0]===o,()=>`scores has incompatible shape with boxes. Expected ${o}, but was ${t.shape[0]}`),D(0<=a&&a<=1,()=>`softNmsSigma must be in [0, 1], but was '${a}'`),{maxOutputSize:n,iouThreshold:r,scoreThreshold:s,softNmsSigma:a}}function $L(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY){let a=_(e,"boxes","nonMaxSuppression","float32"),o=_(t,"scores","nonMaxSuppression","float32"),i=Gu(a,o,n,r,s);n=i.maxOutputSize,r=i.iouThreshold,s=i.scoreThreshold;let c={maxOutputSize:n,iouThreshold:r,scoreThreshold:s};return M.runKernel(fu,{boxes:a,scores:o},c)}var DL=z({nonMaxSuppression_:$L});function FL(e,t,n){let r=RL(e,t,n),s=r<0?-(r+1):r;e.splice(s,0,t)}function RL(e,t,n){return OL(e,t,n||PL)}function PL(e,t){return e>t?1:e<t?-1:0}function OL(e,t,n){let r=0,s=e.length,a=0,o=!1;for(;r<s;){a=r+(s-r>>>1);let i=n(t,e[a]);i>0?r=a+1:(s=a,o=!i)}return o?r:-r-1}function qT(e,t,n,r,s){return Yx(e,t,n,r,s,0)}function jT(e,t,n,r,s,a){return Yx(e,t,n,r,s,0,!1,a,!0)}function KT(e,t,n,r,s,a){return Yx(e,t,n,r,s,a,!0)}function Yx(e,t,n,r,s,a,o=!1,i=!1,c=!1){let u=[];for(let g=0;g<t.length;g++)t[g]>s&&u.push({score:t[g],boxIndex:g,suppressBeginIndex:0});u.sort(yk);let l=a>0?-.5/a:0,p=[],d=[];for(;p.length<n&&u.length>0;){let g=u.pop(),{score:b,boxIndex:y,suppressBeginIndex:v}=g;if(b<s)break;let x=!1;for(let k=p.length-1;k>=v;--k){let S=ML(e,y,p[k]);if(S>=r){x=!0;break}if(g.score=g.score*LL(r,l,S),g.score<=s)break}g.suppressBeginIndex=p.length,x||(g.score===b?(p.push(y),d.push(g.score)):g.score>s&&FL(u,g,yk))}let h=p.length,f=n-h;i&&f>0&&(p.push(...new Array(f).fill(0)),d.push(...new Array(f).fill(0)));let m={selectedIndices:p};return o&&(m.selectedScores=d),c&&(m.validOutputs=h),m}function ML(e,t,n){let r=e.subarray(t*4,t*4+4),s=e.subarray(n*4,n*4+4),a=Math.min(r[0],r[2]),o=Math.min(r[1],r[3]),i=Math.max(r[0],r[2]),c=Math.max(r[1],r[3]),u=Math.min(s[0],s[2]),l=Math.min(s[1],s[3]),p=Math.max(s[0],s[2]),d=Math.max(s[1],s[3]),h=(i-a)*(c-o),f=(p-u)*(d-l);if(h<=0||f<=0)return 0;let m=Math.max(a,u),g=Math.max(o,l),b=Math.min(i,p),y=Math.min(c,d),v=Math.max(b-m,0)*Math.max(y-g,0);return v/(h+f-v)}function LL(e,t,n){let r=Math.exp(t*n*n);return n<=e?r:0}function yk(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}async function zL(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY){let a=_(e,"boxes","nonMaxSuppressionAsync"),o=_(t,"scores","nonMaxSuppressionAsync"),i=Gu(a,o,n,r,s);n=i.maxOutputSize,r=i.iouThreshold,s=i.scoreThreshold;let c=await Promise.all([a.data(),o.data()]),u=c[0],l=c[1],{selectedIndices:p}=qT(u,l,n,r,s);return a!==e&&a.dispose(),o!==t&&o.dispose(),Ke(p,"int32")}var BL=zL;function WL(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY,a=0){let o=_(e,"boxes","nonMaxSuppression"),i=_(t,"scores","nonMaxSuppression"),c=Gu(o,i,n,r,s,a);n=c.maxOutputSize,r=c.iouThreshold,s=c.scoreThreshold,a=c.softNmsSigma;let u={boxes:o,scores:i},l={maxOutputSize:n,iouThreshold:r,scoreThreshold:s,softNmsSigma:a},p=M.runKernel(gu,u,l);return{selectedIndices:p[0],selectedScores:p[1]}}var VL=z({nonMaxSuppressionWithScore_:WL});async function UL(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY,a=0){let o=_(e,"boxes","nonMaxSuppressionAsync"),i=_(t,"scores","nonMaxSuppressionAsync"),c=Gu(o,i,n,r,s,a);n=c.maxOutputSize,r=c.iouThreshold,s=c.scoreThreshold,a=c.softNmsSigma;let u=await Promise.all([o.data(),i.data()]),l=u[0],p=u[1],{selectedIndices:d,selectedScores:h}=KT(l,p,n,r,s,a);return o!==e&&o.dispose(),i!==t&&i.dispose(),{selectedIndices:Ke(d,"int32"),selectedScores:Ke(h)}}var GL=UL;function HL(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY,a=!1){let o=_(e,"boxes","nonMaxSuppression"),i=_(t,"scores","nonMaxSuppression"),c=Gu(o,i,n,r,s,null),u=c.maxOutputSize,l=c.iouThreshold,p=c.scoreThreshold,d={boxes:o,scores:i},h={maxOutputSize:u,iouThreshold:l,scoreThreshold:p,padToMaxOutputSize:a},f=M.runKernel(mu,d,h);return{selectedIndices:f[0],validOutputs:f[1]}}var qL=z({nonMaxSuppressionPadded_:HL});async function jL(e,t,n,r=.5,s=Number.NEGATIVE_INFINITY,a=!1){let o=_(e,"boxes","nonMaxSuppressionAsync"),i=_(t,"scores","nonMaxSuppressionAsync"),c=Gu(o,i,n,r,s,null),u=c.maxOutputSize,l=c.iouThreshold,p=c.scoreThreshold,[d,h]=await Promise.all([o.data(),i.data()]),{selectedIndices:f,validOutputs:m}=jT(d,h,u,l,p,a);return o!==e&&o.dispose(),i!==t&&i.dispose(),{selectedIndices:Ke(f,"int32"),validOutputs:xe(m,"int32")}}var KL=jL;function XL(e,t,n=!1,r=!1){let s=_(e,"images","resizeBilinear");D(s.rank===3||s.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${s.rank}.`),D(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),D(r===!1||n===!1,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let a=s,o=!1;s.rank===3&&(o=!0,a=W(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let[]=t,i={images:a},c={alignCorners:n,halfPixelCenters:r,size:t},u=M.runKernel(ei,i,c);return o?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var XT=z({resizeBilinear_:XL});function YL(e,t,n=!1,r=!1){let s=_(e,"images","resizeNearestNeighbor");D(s.rank===3||s.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${s.rank}.`),D(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),D(s.dtype==="float32"||s.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype"),D(r===!1||n===!1,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let a=s,o=!1;s.rank===3&&(o=!0,a=W(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let[]=t,i={images:a},c={alignCorners:n,halfPixelCenters:r,size:t},u=M.runKernel(Qo,i,c);return o?W(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var YT=z({resizeNearestNeighbor_:YL});function ZL(e,t="binary",n=!1,r=.5){let s=_(e,"image","threshold"),a=.2989,o=.587,i=.114,c=s.shape[0]*s.shape[1],u=B(Ke([r]),255),l,p,d,h;if(D(s.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${s.rank}.`),D(s.shape[2]===3||s.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${s.shape[2]}.`),D(s.dtype==="int32"||s.dtype==="float32",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${s.dtype}.`),D(t==="otsu"||t==="binary",()=>`Method must be binary or otsu, but was ${t}`),s.shape[2]===3){[l,p,d]=zn(s,[1,1,1],-1);let g=B(l,a),b=B(p,o),y=B(d,i);h=Y(Y(g,b),y)}else h=e;if(t==="otsu"){let g=cx(ce(cm(h),"int32"),Cn([]),256);u=JL(g,c)}let f=n?Ta(h,u):An(h,u);return ce(B(f,255),"int32")}function JL(e,t){let n=Ke([-1]),r=Ke([0]),s=Ke([0]),a,o,i,c,u,l;for(let p=0;p<e.size-1;p++){a=We(e,0,p+1),o=We(e,p+1),u=fe(ge(a),t),l=fe(ge(o),t);let d=ge(B(a,vc(0,a.size)));i=fe(d,ge(a));let h=bn(o.shape,a.size),f=Y(vc(0,o.size),h),m=B(o,f);c=fe(ge(m),ge(o));let g=de(i,c),b=de(i,c),y=B(u,l);s=B(B(y,g),b);let v=An(s,r);r=mn(v,s,r),n=mn(v,Ke([p]),n)}return n}var QL=z({threshold_:ZL});function ez(e,t,n="nearest",r="constant",s=0,a){let o=_(e,"image","transform","float32"),i=_(t,"transforms","transform","float32");D(o.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${o.rank}.`),D(i.rank===2&&(i.shape[0]===o.shape[0]||i.shape[0]===1)&&i.shape[1]===8,()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),D(a==null||a.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${a}.`);let c={image:o,transforms:i},u={interpolation:n,fillMode:r,fillValue:s,outputShape:a};return M.runKernel(Fu,c,u)}var tz=z({transform_:ez});function nz(e,t,n){D(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),D(n%1===0,()=>`bandPart(): numUpper must be an integer, got ${n}.`);let r=_(e,"a","bandPart");D(r.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${r.rank}.`);let s=r.shape,[a,o]=r.shape.slice(-2);if(!(t<=a))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${a}).`);if(!(n<=o))throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${o}).`);t<0&&(t=a),n<0&&(n=o);let i=W(vc(0,a,1,"int32"),[-1,1]),c=vc(0,o,1,"int32"),u=de(i,c),l=Dr(Ta(u,xe(+t,"int32")),Ds(u,xe(-n,"int32"))),p=kt([a,o],r.dtype);return W(Ft(lt(W(r,[-1,a,o])).map(d=>mn(l,d,p))),s)}var rz=z({bandPart_:nz});function sz(e){let t;if(Array.isArray(e)){t=!1,D(e!=null&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let s=e[0].shape[0];for(let a=1;a<e.length;++a)D(e[a].shape[0]===s,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[a].shape[0]} vs. ${s})`)}else t=!0,e=zn(e,e.shape[0],0).map(s=>Na(s,[0]));D(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);let n=[],r=e;for(let s=0;s<e.length;++s)n.push(M.tidy(()=>{let a=r[s];if(s>0)for(let o=0;o<s;++o){let i=B(ge(B(n[o],a)),n[o]);a=de(a,i)}return fe(a,zu(a,"euclidean"))}));return t?Ft(n,0):n}var az=z({gramSchmidt_:sz});function oz(e,t=!1){if(D(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),e.rank===2)return vk(e,t);{let n=e.shape.slice(0,e.shape.length-2).reduce((c,u)=>c*u),r=lt(W(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),s=[],a=[];r.forEach(c=>{let[u,l]=vk(c,t);s.push(u),a.push(l)});let o=W(Ft(s,0),e.shape),i=W(Ft(a,0),e.shape);return[o,i]}}function vk(e,t=!1){return M.tidy(()=>{D(e.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);let n=e.shape[0],r=e.shape[1],s=Qf(n),a=is(e),o=$r([[1]],[1,1]),i=is(o),c=n>=r?r:n;for(let u=0;u<c;++u){let l=a,p=i,d=s;[i,a,s]=M.tidy(()=>{let h=We(a,[u,u],[n-u,1]),f=zu(h),m=We(a,[u,u],[1,1]),g=mn(An(m,0),$r([[-1]]),$r([[1]])),b=de(m,B(g,f)),y=fe(h,b);y.shape[0]===1?i=is(o):i=Je([o,We(y,[1,0],[y.shape[0]-1,y.shape[1]])],0);let v=xt(fe(De(g,b),f)),x=We(a,[u,0],[n-u,r]),k=B(v,i),S=Ee(i);if(u===0)a=de(x,De(k,De(S,x)));else{let $=de(x,De(k,De(S,x)));a=Je([We(a,[0,0],[u,r]),$],0)}let C=Ee(k),E=We(s,[0,u],[n,s.shape[1]-u]);if(u===0)s=de(E,De(De(E,i),C));else{let $=de(E,De(De(E,i),C));s=Je([We(s,[0,0],[n,u]),$],1)}return[i,a,s]}),_e([l,p,d])}return!t&&n>r&&(s=We(s,[0,0],[n,r]),a=We(a,[0,0],[r,r])),[s,a]})}var iz=z({qr_:oz}),kn;(function(e){e[e.NONE=0]="NONE",e[e.MEAN=1]="MEAN",e[e.SUM=2]="SUM",e[e.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS"})(kn||(kn={}));function cz(e,t,n=kn.SUM_BY_NONZERO_WEIGHTS){let r=_(e,"losses","computeWeightedLoss"),s=null;t!=null&&(s=_(t,"weights","computeWeightedLoss"));let a=s==null?r:B(r,s);if(n===kn.NONE)return a;if(n===kn.SUM)return ge(a);if(n===kn.MEAN){if(s==null)return Ct(a);{let o=r.size/s.size,i=fe(ge(a),ge(s));return o>1?fe(i,xe(o)):i}}if(n===kn.SUM_BY_NONZERO_WEIGHTS){if(s==null)return fe(ge(a),xe(r.size));{let o=B(s,Qn(r.shape)),i=ce(ge(co(o,xe(0))),"float32");return fe(ge(a),i)}}throw Error(`Unknown reduction: ${n}`)}var Fs=z({computeWeightedLoss_:cz});function uz(e,t,n,r=kn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","absoluteDifference"),a=_(t,"predictions","absoluteDifference"),o=null;n!=null&&(o=_(n,"weights","absoluteDifference")),_n(s.shape,a.shape,"Error in absoluteDifference: ");let i=Lt(de(s,a));return Fs(i,o,r)}var lz=z({absoluteDifference_:uz});function dz(e,t,n,r,s=kn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"labels","cosineDistance"),o=_(t,"predictions","cosineDistance"),i=null;r!=null&&(i=_(r,"weights","cosineDistance")),_n(a.shape,o.shape,"Error in cosineDistance: ");let c=xe(1),u=de(c,ge(B(a,o),n,!0));return Fs(u,i,s)}var pz=z({cosineDistance_:dz});function hz(e,t,n,r=kn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","hingeLoss"),a=_(t,"predictions","hingeLoss"),o=null;n!=null&&(o=_(n,"weights","hingeLoss")),_n(s.shape,a.shape,"Error in hingeLoss: ");let i=xe(1);s=de(B(xe(2),s),i);let c=Xe(de(i,B(s,a)));return Fs(c,o,r)}var fz=z({hingeLoss_:hz});function mz(e,t,n,r=1,s=kn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"labels","huberLoss"),o=_(t,"predictions","huberLoss"),i=null;n!=null&&(i=_(n,"weights","huberLoss")),_n(a.shape,o.shape,"Error in huberLoss: ");let c=xe(r),u=Lt(de(o,a)),l=Vu(u,c),p=de(u,l),d=Y(B(xe(.5),it(l)),B(c,p));return Fs(d,i,s)}var gz=z({huberLoss_:mz});function bz(e,t,n,r=1e-7,s=kn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"labels","logLoss"),o=_(t,"predictions","logLoss"),i=null;n!=null&&(i=_(n,"weights","logLoss")),_n(a.shape,o.shape,"Error in logLoss: ");let c=xe(1),u=xe(r),l=xt(B(a,nr(Y(o,u)))),p=B(de(c,a),nr(Y(de(c,o),u))),d=de(l,p);return Fs(d,i,s)}var yz=z({logLoss_:bz});function vz(e,t,n,r=kn.SUM_BY_NONZERO_WEIGHTS){let s=_(e,"labels","meanSquaredError"),a=_(t,"predictions","meanSquaredError"),o=null;n!=null&&(o=_(n,"weights","meanSquaredError")),_n(s.shape,a.shape,"Error in meanSquaredError: ");let i=mm(s,a);return Fs(i,o,r)}var xz=z({meanSquaredError_:vz});function wz(e,t){let n=_(e,"labels","sigmoidCrossEntropyWithLogits"),r=_(t,"logits","sigmoidCrossEntropyWithLogits");_n(n.shape,r.shape,"Error in sigmoidCrossEntropyWithLogits: ");let s=Xe(r),a=B(r,n),o=zd(gn(xt(Lt(r))));return Y(de(s,a),o)}function Iz(e,t,n,r=0,s=kn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"multiClassLabels","sigmoidCrossEntropy"),o=_(t,"logits","sigmoidCrossEntropy"),i=null;if(n!=null&&(i=_(n,"weights","sigmoidCrossEntropy")),_n(a.shape,o.shape,"Error in sigmoidCrossEntropy: "),r>0){let u=xe(r),l=xe(1),p=xe(.5);a=Y(B(a,de(l,u)),B(p,u))}let c=wz(a,o);return Fs(c,i,s)}var kz=z({sigmoidCrossEntropy_:Iz});function Sz(e,t,n=-1){if(n===-1&&(n=t.rank-1),n!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${n}`);return ls((s,a,o)=>{let c=rm(a,[n],!0),u=de(ce(a,"float32"),c);o([s,u]);let l=xt(B(u,s));return{value:ge(l,[n]),gradFunc:(h,f)=>{let[m,g]=f,b=io(h.shape,[n]);return[B(W(h,b),de(ce(m,"float32"),gn(g))),B(W(h,b),de(gn(g),ce(m,"float32")))]}}})(e,t)}function Tz(e,t,n,r=0,s=kn.SUM_BY_NONZERO_WEIGHTS){let a=_(e,"onehotLabels","softmaxCrossEntropy"),o=_(t,"logits","softmaxCrossEntropy"),i=null;if(n!=null&&(i=_(n,"weights","softmaxCrossEntropy")),_n(a.shape,o.shape,"Error in softmaxCrossEntropy: "),r>0){let u=xe(r),l=xe(1),p=xe(a.shape[1]);a=Y(B(a,de(l,u)),fe(u,p))}let c=Sz(a,o);return Fs(c,i,s)}var Cz=z({softmaxCrossEntropy_:Tz});function Nz(e,t,n,r){let s=_(e,"indices","sparseFillEmptyRows","int32"),a=_(t,"values","sparseFillEmptyRows"),o=_(n,"denseShape","sparseFillEmptyRows","int32"),i=_(r,"defaultValue","sparseFillEmptyRows",a.dtype);if(s.rank!==2)throw new Error(`Indices should be Tensor2D but received shape
+        ${s.shape}`);if(a.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${a.shape}`);if(o.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${o.shape}`);if(i.rank!==0)throw new Error(`Default value should be a scalar but received shape ${i.shape}`);let c={indices:s,values:a,denseShape:o,defaultValue:i},u=M.runKernel(xd,c);return{outputIndices:u[0],outputValues:u[1],emptyRowIndicator:u[2],reverseIndexMap:u[3]}}var _z=z({sparseFillEmptyRows_:Nz});function Ez(e,t,n){let r=_(e,"inputIndices","sparseReshape","int32"),s=_(t,"inputShape","sparseReshape","int32"),a=_(n,"newShape","sparseReshape","int32");if(r.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape
+        ${r.shape}`);if(s.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${s.shape}`);if(a.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${a.shape}`);let o={inputIndices:r,inputShape:s,newShape:a},i=M.runKernel(Au,o);return{outputIndices:i[0],outputShape:i[1]}}var Az=z({sparseReshape_:Ez});function $z(e,t,n){let r=_(e,"data","sparseSegmentMean"),s=_(t,"indices","sparseSegmentMean","int32"),a=_(n,"segmentIds","sparseSegmentMean","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
+          ${s.shape}`);if(a.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
+          ${a.shape}`);let o={data:r,indices:s,segmentIds:a};return M.runKernel(wd,o)}var Dz=z({sparseSegmentMean_:$z});function Fz(e,t,n){let r=_(e,"data","sparseSegmentSum"),s=_(t,"indices","sparseSegmentSum","int32"),a=_(n,"segmentIds","sparseSegmentSum","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
+         ${s.shape}`);if(a.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
+         ${a.shape}`);let o={data:r,indices:s,segmentIds:a};return M.runKernel(Id,o)}var Rz=z({sparseSegmentSum_:Fz});function Pz(e,t,n,r,s,a,o,i){let c=_(e,"data","stringNGrams","string");if(c.dtype!=="string")throw new Error("Data must be of datatype string");if(c.shape.length!==1)throw new Error(`Data must be a vector, saw: ${c.shape}`);let u=_(t,"dataSplits","stringNGrams");if(u.dtype!=="int32")throw new Error("Data splits must be of datatype int32");let l={separator:n,nGramWidths:r,leftPad:s,rightPad:a,padWidth:o,preserveShortSequences:i},p={data:c,dataSplits:u},d=M.runKernel(Sd,p,l);return{nGrams:d[0],nGramsSplits:d[1]}}var Oz=z({stringNGrams_:Pz});function Mz(e,t,n=!0){let r=_(e,"input","stringSplit","string"),s=_(t,"delimiter","stringSplit","string");if(r.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${r.shape}`);if(s.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${s.shape}`);let a={skipEmpty:n},o={input:r,delimiter:s},i=M.runKernel(Td,o,a);return{indices:i[0],values:i[1],shape:i[2]}}var Lz=z({stringSplit_:Mz});function zz(e,t){let n=_(e,"input","stringToHashBucketFast","string"),r={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let s={input:n};return M.runKernel(Cd,s,r)}var Bz=z({stringToHashBucketFast_:zz}),ZT={fft:Hd,ifft:wc,rfft:qd,irfft:fm},JT={hammingWindow:yL,hannWindow:GT,frame:HT,stft:IL},Br={flipLeftRight:CL,grayscaleToRGB:_L,resizeNearestNeighbor:YT,resizeBilinear:XT,rotateWithOffset:AL,cropAndResize:SL,nonMaxSuppression:DL,nonMaxSuppressionAsync:BL,nonMaxSuppressionWithScore:VL,nonMaxSuppressionWithScoreAsync:GL,nonMaxSuppressionPadded:qL,nonMaxSuppressionPaddedAsync:KL,threshold:QL,transform:tz},Zx={bandPart:rz,gramSchmidt:az,qr:iz},QT={absoluteDifference:lz,computeWeightedLoss:Fs,cosineDistance:pz,hingeLoss:fz,huberLoss:gz,logLoss:yz,meanSquaredError:xz,sigmoidCrossEntropy:kz,softmaxCrossEntropy:Cz},eC={sparseFillEmptyRows:_z,sparseReshape:Az,sparseSegmentMean:Dz,sparseSegmentSum:Rz},tC={stringNGrams:Oz,stringSplit:Lz,stringToHashBucketFast:Bz},Rs=class extends qS{minimize(e,t=!1,n){let{value:r,grads:s}=this.computeGradients(e,n);if(n!=null){let a=n.map(o=>({name:o.name,tensor:s[o.name]}));this.applyGradients(a)}else this.applyGradients(s);return _e(s),t?r:(r.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return lT(e,t)}dispose(){this.iterations_!=null&&_e(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:xe(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(e){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty(Rs,Symbol.hasInstance,{value:e=>e.minimize!=null&&e.computeGradients!=null&&e.applyGradients!=null});var km=class extends Rs{constructor(e,t,n=null){super(),this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=M.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(n=>n.name):Object.keys(e)).forEach((n,r)=>{let s=M.registeredVariables[n],a=!1;this.accumulatedGrads[r]==null&&(this.accumulatedGrads[r]={originalName:`${n}/accum_grad`,variable:O(()=>qe(s).variable(a))}),this.accumulatedUpdates[r]==null&&(this.accumulatedUpdates[r]={originalName:`${n}/accum_var`,variable:O(()=>qe(s).variable(a))});let o=Array.isArray(e)?e[r].tensor:e[n];if(o==null)return;let i=this.accumulatedGrads[r].variable,c=this.accumulatedUpdates[r].variable;O(()=>{let u=Y(B(i,this.rho),B(it(o),1-this.rho)),l=B(fe(un(Y(c,this.epsilon)),un(Y(i,this.epsilon))),o),p=Y(B(c,this.rho),B(it(l),1-this.rho));i.assign(u),c.assign(p);let d=Y(B(l,-this.learningRate),s);s.assign(d)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(_e(this.accumulatedGrads.map(e=>e.variable)),_e(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){let e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=e.length/2,n=!1;this.accumulatedGrads=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedUpdates=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}};km.className="Adadelta";Ia(km);var Sm=class extends Rs{constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(n=>n.name):Object.keys(e)).forEach((n,r)=>{let s=M.registeredVariables[n];this.accumulatedGrads[r]==null&&(this.accumulatedGrads[r]={originalName:`${n}/accumulator`,variable:O(()=>bn(s.shape,this.initialAccumulatorValue).variable(!1))});let a=Array.isArray(e)?e[r].tensor:e[n];if(a==null)return;let o=this.accumulatedGrads[r].variable;O(()=>{let i=Y(o,it(a));o.assign(i);let c=Y(B(fe(a,un(Y(i,M.backend.epsilon()))),-this.learningRate),s);s.assign(c)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&_e(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulatedGrads=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}};Sm.className="Adagrad";Ia(Sm);var Tm=class extends Rs{constructor(e,t,n,r=null){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],O(()=>{this.accBeta1=xe(t).variable(),this.accBeta2=xe(n).variable()}),r==null&&(this.epsilon=M.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);O(()=>{let n=de(1,this.accBeta1),r=de(1,this.accBeta2);t.forEach((s,a)=>{let o=M.registeredVariables[s],i=!1;this.accumulatedFirstMoment[a]==null&&(this.accumulatedFirstMoment[a]={originalName:`${s}/m`,variable:O(()=>qe(o).variable(i))}),this.accumulatedSecondMoment[a]==null&&(this.accumulatedSecondMoment[a]={originalName:`${s}/v`,variable:O(()=>qe(o).variable(i))});let c=Array.isArray(e)?e[a].tensor:e[s];if(c==null)return;let u=this.accumulatedFirstMoment[a].variable,l=this.accumulatedSecondMoment[a].variable,p=Y(B(u,this.beta1),B(c,1-this.beta1)),d=Y(B(l,this.beta2),B(it(c),1-this.beta2)),h=fe(p,n),f=fe(d,r);u.assign(p),l.assign(d);let m=Y(B(fe(h,Y(un(f),this.epsilon)),-this.learningRate),o);o.assign(m)}),this.accBeta1.assign(B(this.accBeta1,this.beta1)),this.accBeta2.assign(B(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedSecondMoment!=null&&_e(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e),O(()=>{this.accBeta1.assign($s(this.beta1,this.iterations_+1)),this.accBeta2.assign($s(this.beta2,this.iterations_+1))});let t=e.length/2,n=!1;this.accumulatedFirstMoment=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedSecondMoment=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon)}};Tm.className="Adam";Ia(Tm);var Cm=class extends Rs{constructor(e,t,n,r=null,s=0){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.decay=s,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],O(()=>{this.iteration=xe(0).variable(),this.accBeta1=xe(t).variable()}),r==null&&(this.epsilon=M.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);O(()=>{let n=de(1,this.accBeta1),r=fe(-this.learningRate,Y(B(this.iteration,this.decay),1));t.forEach((s,a)=>{let o=M.registeredVariables[s],i=!1;this.accumulatedFirstMoment[a]==null&&(this.accumulatedFirstMoment[a]={originalName:`${s}/m`,variable:qe(o).variable(i)}),this.accumulatedWeightedInfNorm[a]==null&&(this.accumulatedWeightedInfNorm[a]={originalName:`${s}/v`,variable:qe(o).variable(i)});let c=Array.isArray(e)?e[a].tensor:e[s];if(c==null)return;let u=this.accumulatedFirstMoment[a].variable,l=this.accumulatedWeightedInfNorm[a].variable,p=Y(B(u,this.beta1),B(c,1-this.beta1)),d=B(l,this.beta2),h=Lt(c),f=hs(d,h);u.assign(p),l.assign(f);let m=Y(B(fe(r,n),fe(p,Y(f,this.epsilon))),o);o.assign(m)}),this.iteration.assign(Y(this.iteration,1)),this.accBeta1.assign(B(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&_e(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedWeightedInfNorm!=null&&_e(this.accumulatedWeightedInfNorm.map(e=>e.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon,t.decay)}};Cm.className="Adamax";Ia(Cm);var jd=class extends Rs{constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(n=>n.name):Object.keys(e)).forEach((n,r)=>{let s=Array.isArray(e)?e[r].tensor:e[n];if(s==null)return;let a=M.registeredVariables[n];O(()=>{let o=Y(B(this.c,s),a);a.assign(o)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,this.c!=null&&this.c.dispose(),this.c=Jt(xe(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(e=await this.extractIterations(e),e.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}};jd.className="SGD";Ia(jd);var Nm=class extends jd{constructor(e,t,n=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=xe(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(n=>n.name):Object.keys(e)).forEach((n,r)=>{let s=M.registeredVariables[n];this.accumulations[r]==null&&(this.accumulations[r]={originalName:`${n}/momentum`,variable:O(()=>qe(s).variable(!1))});let a=this.accumulations[r].variable,o=Array.isArray(e)?e[r].tensor:e[n];o!=null&&O(()=>{let i,c=Y(B(this.m,a),o);this.useNesterov?i=Y(B(this.c,Y(o,B(c,this.m))),s):i=Y(B(this.c,c),s),a.assign(c),s.assign(i)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&_e(this.accumulations.map(e=>e.variable))}setMomentum(e){this.momentum=e}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulations=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}};Nm.className="Momentum";Ia(Nm);var _m=class extends Rs{constructor(e,t=.9,n=0,r=null,s=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=r,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=s,r==null&&(this.epsilon=M.backend.epsilon()),e==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(n=>n.name):Object.keys(e)).forEach((n,r)=>{let s=M.registeredVariables[n],a=!1;this.accumulatedMeanSquares[r]==null&&(this.accumulatedMeanSquares[r]={originalName:`${n}/rms`,variable:O(()=>qe(s).variable(a))}),this.accumulatedMoments[r]==null&&(this.accumulatedMoments[r]={originalName:`${n}/momentum`,variable:O(()=>qe(s).variable(a))}),this.accumulatedMeanGrads[r]==null&&this.centered&&(this.accumulatedMeanGrads[r]={originalName:`${n}/mg`,variable:O(()=>qe(s).variable(a))});let o=Array.isArray(e)?e[r].tensor:e[n];if(o==null)return;let i=this.accumulatedMeanSquares[r].variable,c=this.accumulatedMoments[r].variable;O(()=>{let u=Y(B(i,this.decay),B(it(o),1-this.decay));if(this.centered){let l=this.accumulatedMeanGrads[r].variable,p=Y(B(l,this.decay),B(o,1-this.decay)),d=fe(B(o,this.learningRate),un(de(u,Y(it(p),this.epsilon)))),h=Y(B(c,this.momentum),d);i.assign(u),l.assign(p),c.assign(h);let f=de(s,h);s.assign(f)}else{let l=Y(B(i,this.decay),B(it(o),1-this.decay)),p=Y(B(c,this.momentum),fe(B(o,this.learningRate),un(Y(l,this.epsilon))));i.assign(l),c.assign(p);let d=de(s,p);s.assign(d)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&_e(this.accumulatedMeanSquares.map(e=>e.variable)),this.accumulatedMeanGrads!=null&&this.centered&&_e(this.accumulatedMeanGrads.map(e=>e.variable)),this.accumulatedMoments!=null&&_e(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=this.centered?e.length/3:e.length/2,n=!1;this.accumulatedMeanSquares=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedMoments=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=e.slice(t*2,t*3).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}};_m.className="RMSProp";Ia(_m);var Zs=class{static sgd(e){return new jd(e)}static momentum(e,t,n=!1){return new Nm(e,t,n)}static rmsprop(e,t=.9,n=0,r=null,s=!1){return new _m(e,t,n,r,s)}static adam(e=.001,t=.9,n=.999,r=null){return new Tm(e,t,n,r)}static adadelta(e=.001,t=.95,n=null){return new km(e,t,n)}static adamax(e=.002,t=.9,n=.999,r=null,s=0){return new Cm(e,t,n,r,s)}static adagrad(e,t=.1){return new Sm(e,t)}},Wa={sgd:Zs.sgd,momentum:Zs.momentum,adadelta:Zs.adadelta,adagrad:Zs.adagrad,rmsprop:Zs.rmsprop,adamax:Zs.adamax,adam:Zs.adam},Wz=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:e=>e())();function Jx(){return new Promise(e=>Wz(()=>e()))}var N={};Ae(N,{ERF_A1:()=>nB,ERF_A2:()=>rB,ERF_A3:()=>sB,ERF_A4:()=>aB,ERF_A5:()=>oB,ERF_P:()=>tB,PARALLELIZE_THRESHOLD:()=>Qx,RowPartitionType:()=>rs,SELU_SCALE:()=>rC,SELU_SCALEALPHA:()=>nC,applyActivation:()=>wm,assertAndGetBroadcastShape:()=>ct,assertAxesAreInnerMostDims:()=>qO,assertParamsConsistent:()=>Vz,assignToTypedArray:()=>pB,axesAreInnerMostDims:()=>Ix,calculateShapes:()=>OS,checkEinsumDimSizes:()=>yB,checkPadOnDimRoundingMode:()=>En,combineLocations:()=>aT,combineRaggedTensorToTensorShapes:()=>Gz,complexWithEvenIndex:()=>uB,complexWithOddIndex:()=>lB,computeConv2DInfo:()=>Pd,computeConv3DInfo:()=>JS,computeDefaultPad:()=>rx,computeDilation2DInfo:()=>jP,computeOptimalWindowSize:()=>Kz,computeOutAndReduceShapes:()=>oT,computeOutShape:()=>Uz,computePool2DInfo:()=>ZS,computePool3DInfo:()=>KP,convertConv2DDataFormat:()=>QS,decodeEinsumEquation:()=>gB,eitherStridesOrDilationsAreOne:()=>ps,expandShapeToKeepDim:()=>io,exponent:()=>fB,exponents:()=>hB,fromStringArrayToUint8:()=>zB,fromUint8ToStringArray:()=>LB,getAxesPermutation:()=>iT,getBroadcastDims:()=>FS,getComplexWithIndex:()=>dB,getEinsumComputePath:()=>vB,getEinsumPermutation:()=>bB,getFusedBiasGradient:()=>xm,getFusedDyActivation:()=>vm,getImageCenter:()=>Xz,getInnerMostAxes:()=>jO,getPermuted:()=>Zz,getRaggedRank:()=>qz,getReductionAxes:()=>Bt,getReshaped:()=>Yz,getReshapedPermuted:()=>Jz,getRowPartitionTypesHelper:()=>Hz,getSliceBeginCoords:()=>Qz,getSliceSize:()=>eB,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>kB,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>SB,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>TB,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>_B,getSparseReshapeInputOutputMismatchErrorMessage:()=>AB,getSparseReshapeInputOutputMultipleErrorMessage:()=>EB,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>CB,getSparseReshapeNegativeOutputDimErrorMessage:()=>NB,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>RB,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>$B,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>DB,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>FB,getUndoAxesPermutation:()=>kx,isIdentityPermutation:()=>xB,log:()=>lF,mergeRealAndImagArrays:()=>iB,prepareAndValidate:()=>PS,prepareSplitSize:()=>IB,segment_util:()=>sC,shouldFuse:()=>Im,slice_util:()=>qt,splitRealAndImagArrays:()=>cB,tupleValuesAreOne:()=>ca,upcastType:()=>hr,validateDefaultValueShape:()=>jz,validateInput:()=>qv,validateUpdateShape:()=>Hv,warn:()=>Qs});function Vz(e,t){let n=e[0].length;e.forEach((s,a)=>{D(s.length===n,()=>`Error in concat${n}D: rank of tensors[${a}] must be the same as the rank of the rest (${n})`)}),D(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);let r=e[0];e.forEach((s,a)=>{for(let o=0;o<n;o++)D(o===t||s[o]===r[o],()=>`Error in concat${n}D: Shape of tensors[${a}] (${s}) does not match the shape of the rest (${r}) along the non-concatenated axis ${a}.`)})}function Uz(e,t){let n=e[0].slice();for(let r=1;r<e.length;r++)n[t]+=e[r][t];return n}var rs;(function(e){e[e.FIRST_DIM_SIZE=0]="FIRST_DIM_SIZE",e[e.VALUE_ROWIDS=1]="VALUE_ROWIDS",e[e.ROW_LENGTHS=2]="ROW_LENGTHS",e[e.ROW_SPLITS=3]="ROW_SPLITS",e[e.ROW_LIMITS=4]="ROW_LIMITS",e[e.ROW_STARTS=5]="ROW_STARTS"})(rs||(rs={}));function Gz(e,t,n){let r=new Array;if(n==null&&t==null)return r;if(t==null)for(;r.length<e+n.length;)r.push(-1);else r=t.slice();if(n==null)return r;if(e+n.length!==r.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${e+n.length}, but shape.rank = ${r.length}`);for(let s=1;s<n.length;++s){let a=n[s],o=r[r.length-n.length+s],i=r[o];if(a>=0)if(i>=0){if(i!==a)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${s+e}] = ${a} but shape[${s+e}] = ${i}`)}else r[o]=a}return r}function Hz(e){let t={FIRST_DIM_SIZE:rs.FIRST_DIM_SIZE,VALUE_ROWIDS:rs.VALUE_ROWIDS,ROW_LENGTHS:rs.ROW_LENGTHS,ROW_SPLITS:rs.ROW_SPLITS,ROW_LIMITS:rs.ROW_LIMITS,ROW_STARTS:rs.ROW_STARTS},n=[];for(let r of e)if(r in t)n.push(t[r]);else break;return n}function qz(e){return e.length===0?0:e[0]===rs.FIRST_DIM_SIZE?e.length-1:e.length}function jz(e,t){if(e==null||t==null)return;let n=e.length,r=t.length;if(n>=r)throw new Error(`defaultValue.shape=${e} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${n} must be less than ragged tensor input flatValues.rank = ${r})`);for(let s=0;s<Math.min(n,r-1);++s){let a=e[s],o=t[s+1];if(a>=0&&o>=0&&a!==1&&a!==o)throw new Error(`defaultValue.shape=${e}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${s-e.length}] = ${a} but ragged tensor input.flatValues.shape[${s-e.length}] = ${o}`)}}var Qx=30;function Kz(e){return e<=Qx?e:Ah(e,Math.floor(Math.sqrt(e)))}function Xz(e,t,n){let r=n*(typeof e=="number"?e:e[0]),s=t*(typeof e=="number"?e:e[1]);return[r,s]}function Yz(e,t,n,r=!0){let s=[];if(r)s=s.concat(t.slice(0)),s.push(e[0]/n),s=s.concat(e.slice(1));else{s=s.concat(e[0]);let a=t.length;for(let o=0;o<a;++o)s=s.concat([e[o+1]/t[o],t[o]]);s=s.concat(e.slice(a+1))}return s}function Zz(e,t,n=!0){let r=[];if(n){r.push(t);for(let s=t+1;s<e;++s)s<=2*t?(r.push(s),r.push(s-(t+1))):r.push(s)}else{let s=[],a=[];for(let o=1;o<e;++o)o>=t*2+1||o%2===1?a.push(o):s.push(o);r.push(...s),r.push(0),r.push(...a)}return r}function Jz(e,t,n,r=!0){let s=[];r?s.push(e[0]/n):s.push(e[0]*n);for(let a=1;a<e.length;++a)a<=t.length?r?s.push(t[a-1]*e[a]):s.push(e[a]/t[a-1]):s.push(e[a]);return s}function Qz(e,t){let n=[0];for(let r=0;r<t;++r)n.push(e[r][0]);return n}function eB(e,t,n){let r=e.slice(0,1);for(let s=0;s<n;++s)r.push(e[s+1]-t[s][0]-t[s][1]);return r}var nC=1.7580993408473768,rC=1.0507009873554805,tB=.3275911,nB=.254829592,rB=-.284496736,sB=1.421413741,aB=-1.453152027,oB=1.061405429;function iB(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);let n=new Float32Array(e.length*2);for(let r=0;r<n.length;r+=2)n[r]=e[r/2],n[r+1]=t[r/2];return n}function cB(e){let t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let r=0;r<e.length;r+=2)t[r/2]=e[r],n[r/2]=e[r+1];return{real:t,imag:n}}function uB(e){let t=Math.ceil(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let s=0;s<e.length;s+=4)n[Math.floor(s/4)]=e[s],r[Math.floor(s/4)]=e[s+1];return{real:n,imag:r}}function lB(e){let t=Math.floor(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let s=2;s<e.length;s+=4)n[Math.floor(s/4)]=e[s],r[Math.floor(s/4)]=e[s+1];return{real:n,imag:r}}function dB(e,t){let n=e[t*2],r=e[t*2+1];return{real:n,imag:r}}function pB(e,t,n,r){e[r*2]=t,e[r*2+1]=n}function hB(e,t){let n=new Float32Array(e/2),r=new Float32Array(e/2);for(let s=0;s<Math.ceil(e/2);s++){let a=(t?2:-2)*Math.PI*(s/e);n[s]=Math.cos(a),r[s]=Math.sin(a)}return{real:n,imag:r}}function fB(e,t,n){let r=(n?2:-2)*Math.PI*(e/t),s=Math.cos(r),a=Math.sin(r);return{real:s,imag:a}}var yy="->",mB=/->/g,xk=",",wk="...";function gB(e,t){e=e.replace(/\s/g,"");let n=(e.length-e.replace(mB,"").length)/yy.length;if(n<1)throw new Error("Equations without an arrow are not supported.");if(n>1)throw new Error(`Equation must contain exactly one arrow ("${yy}").`);let[r,s]=e.split(yy);D(r.indexOf(wk)===-1,()=>`The ellipsis notation ("${wk}") is not supported yet.`);let a=r.split(xk),o=a.length;if(t!==o)throw new Error(`Expected ${o} input tensors, received ${t}`);if(o>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let i=[];for(let d=0;d<s.length;++d){let h=s[d];if(!a.some(f=>f.indexOf(h)!==-1))throw new Error(`Output subscripts contain the label ${h} not present in the input subscripts.`);i.indexOf(h)===-1&&i.push(h)}for(let d=0;d<r.length;++d){let h=r[d];i.indexOf(h)===-1&&h!==xk&&i.push(h)}let c=new Array(a.length);for(let d=0;d<o;++d){if(new Set(a[d].split("")).size!==a[d].length)throw new Error(`Found duplicate axes in input component ${a[d]}. Support for duplicate axes in input is not implemented yet.`);c[d]=[];for(let h=0;h<a[d].length;++h)c[d].push(i.indexOf(a[d][h]))}let u=i.length,l=s.length,p=[];for(let d=l;d<u;++d)p.push(d);return{allDims:i,summedDims:p,idDims:c}}function bB(e,t){let n=new Array(e);n.fill(-1);for(let s=0;s<t.length;++s)n[t[s]]=s;let r=[];for(let s=0;s<e;++s)n[s]===-1&&r.push(s);return n=n.filter(s=>s!==-1),{permutationIndices:n,expandDims:r}}function yB(e,t,n){let r=new Array(e);for(let s=0;s<n.length;++s){let a=n[s].shape;for(let o=0;o<t[s].length;++o)r[t[s][o]]===void 0?r[t[s][o]]=a[o]:D(r[t[s][o]]===a[o],()=>`Expected dimension ${r[t[s][o]]} at axis ${o} of input shaped ${JSON.stringify(a)}, but got dimension ${a[o]}`)}}function vB(e,t){let n=e,r=[],s=0;e.length===0&&n.push(-1),s=e.length+1;for(let o=0;o<s;++o)r.push([]);let a=[];for(let o=0;o<n.length;++o){let i=n[o],c=wB(t,i);for(let u of c)a.indexOf(u)===-1&&(r[o].push(u),a.push(u))}return{path:n,steps:r}}function xB(e){return e.every((t,n)=>t===n)}function wB(e,t){let n=[];for(let r=0;r<e.length;++r)(e[r].length===0||e[r].indexOf(t)!==-1||t===-1)&&n.push(r);return n}function IB(e,t,n=0){let r=[];if(typeof t=="number")D(e.shape[n]%t===0,()=>"Number of splits must evenly divide the axis."),r=new Array(t).fill(e.shape[n]/t);else{let s=t.reduce((o,i)=>(i===-1&&(o+=1),o),0);D(s<=1,()=>"There should be only one negative value in split array.");let a=t.indexOf(-1);if(a!==-1){let o=t.reduce((i,c)=>c>0?i+c:i);t[a]=e.shape[n]-o}D(e.shape[n]===t.reduce((o,i)=>o+i),()=>"The sum of sizes must match the size of the axis dimension."),r=t}return r}function kB(e){return`Received SparseTensor with denseShape[0] = 0 but
+  indices.shape[0] = ${e}`}function SB(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function TB(e,t,n){return`indices(${e}, 0) is invalid: ${t} >= ${n}`}function CB(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function NB(e,t){return`size ${e} must be non-negative, not ${t}`}function _B(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function EB(e,t){let n=ht(e),r=ht(t);return`Input to reshape is a SparseTensor with ${n}
+  dense values, but the requested shape requires a multiple of ${r}. inputShape=${e} outputShape= ${t}`}function AB(e,t){let n=ht(e),r=ht(t);return`Input to reshape is a tensor with ${n} dense values, but the requested shape has ${r}. inputShape=${e} outputShape=${t}`}function $B(){return"segment ids must be >= 0"}function DB(){return"segment ids are not increasing"}function FB(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function RB(e,t,n){return`Bad: indices[${e}] == ${t} out of range [0, ${n})`}var sC={};Ae(sC,{collectGatherOpShapeInfo:()=>MB,computeOutShape:()=>OB,segOpComputeOptimalWindowSize:()=>PB});function PB(e,t){let n=!1,r;for(e<=Qx?(r=e,n=!0):r=Ah(e,Math.floor(Math.sqrt(e)));!n;)r>t||r===e?n=!0:r=Ah(e,r+1);return r}function OB(e,t,n){let r=[],s=e.length;for(let a=0;a<s;a++)a!==t?r.push(e[a]):r.push(n);return r}function MB(e,t,n,r){let s=t.shape.length,a=e.shape.length;if(r!==0&&(r<-s||r>s))throw new Error(`Expect batchDims in the range of [-${s}, ${s}], but got ${r}`);if(r<0&&(r+=s),r>a)throw new Error(`batchDims (${r}) must be less than rank(x) (
+    ${a}).`);if(n<r)throw new Error(`batchDims (${r}) must be less than or equal to axis (${n}).`);for(let p=0;p<r;++p)if(e.shape[p]!==t.shape[p])throw new Error(`x.shape[${p}]: ${e.shape[p]} should be equal to indices.shape[${p}]: ${t.shape[p]}.`);let o=e.shape[n],i=[],c=1,u=1,l=1;for(let p=0;p<r;++p)i.push(e.shape[p]),c*=e.shape[p];for(let p=r;p<n;p++)i.push(e.shape[p]),u*=e.shape[p];for(let p=r;p<s;p++)i.push(t.shape[p]);for(let p=n+1;p<a;p++)i.push(e.shape[p]),l*=e.shape[p];return{batchSize:c,sliceSize:l,outerSize:u,dimSize:o,outputShape:i}}function LB(e){try{return e.map(t=>Oh(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function zB(e){return e.map(t=>Ad(t))}var fs={};Ae(fs,{nonMaxSuppressionV3Impl:()=>qT,nonMaxSuppressionV4Impl:()=>jT,nonMaxSuppressionV5Impl:()=>KT,whereImpl:()=>PT});var aC={kernelName:Fc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,bi(ce(n,"float32"),-1))}}},BB={kernelName:Rc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let r=it(ce(n,"float32")),s=un(de(xe(1),r));return xt(fe(e,s))}}}},WB={kernelName:Pc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let r=un(de(it(ce(n,"float32")),1));return fe(e,r)}}}},VB={kernelName:ba,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=e,c=Bt(n.shape,s);return c.length>0&&(i=ge(i,c)),W(i,n.shape)},b:()=>{let i=e,c=Bt(r.shape,s);return c.length>0&&(i=ge(i,c)),W(i,r.shape)}}}},UB={kernelName:bo,saveAllInputs:!0,gradFunc:(e,t)=>{let n={};return t.forEach((r,s)=>{n[s]=()=>e.clone()}),n}},GB={kernelName:yo,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},HB={kernelName:dd,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>qe(n)}}},qB={kernelName:Lc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,un(de(xe(1),it(ce(n,"float32")))))}}},jB={kernelName:zc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let r=un(Y(xe(1),it(ce(n,"float32"))));return fe(e,r)}}}},KB={kernelName:Vc,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=Y(it(n),it(r)),c=B(e,fe(r,i)),u=Bt(n.shape,s);return u.length>0&&(c=ge(c,u)),W(c,n.shape)},b:()=>{let i=Y(it(n),it(r)),c=xt(B(e,fe(n,i))),u=Bt(r.shape,s);return u.length>0&&(c=ge(c,u)),W(c,r.shape)}}}},XB={kernelName:Bc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,Y(it(ce(n,"float32")),1))}}},YB={kernelName:Wc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,de(xe(1),it(ce(n,"float32"))))}}};function ZB(e,t,n,r,s,a){let o=_(e,"dy","avgPool3dGrad"),i=_(t,"input","avgPool3dGrad"),c=o,u=i,l=!1;i.rank===4&&(l=!0,c=W(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),u=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),D(c.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${c.rank}.`),D(u.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${u.rank}.`),En("avgPool3dGrad",s,a);let p={dy:c,input:u},d={filterSize:n,strides:r,pad:s,dimRoundingMode:a},h=M.runKernel(hf,p,d);return l?W(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}var JB=z({avgPool3dGrad_:ZB}),QB={kernelName:pd,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{filterSize:s,strides:a,pad:o,dimRoundingMode:i}=n;return{x:()=>JB(e,r,s,a,o,i)}}};function eW(e,t,n,r,s){let a=_(e,"dy","avgPoolGrad"),o=_(t,"input","avgPoolGrad");D(o.rank===a.rank,()=>`Rank of input (${o.rank}) does not match rank of dy (${a.rank})`);let i=o,c=a,u=!1;o.rank===3&&(u=!0,i=W(o,[1,o.shape[0],o.shape[1],o.shape[2]]),c=W(a,[1,a.shape[0],a.shape[1],a.shape[2]])),D(c.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${c.rank}.`),D(i.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${i.rank}.`);let l={dy:c,input:i},p={filterSize:n,strides:r,pad:s},d=M.runKernel(pf,l,p);return u?W(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var tW=z({avgPoolGrad_:eW}),nW={kernelName:vo,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{filterSize:s,strides:a,pad:o}=n;return{x:()=>tW(e,r,s,a,o)}}},rW={kernelName:xo,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{let[r,s]=t,{transposeA:a,transposeB:o}=n;return!a&&!o?{a:()=>De(e,s,!1,!0),b:()=>De(r,e,!0,!1)}:!a&&o?{a:()=>De(e,s,!1,!1),b:()=>De(e,r,!0,!1)}:a&&!o?{a:()=>De(s,e,!1,!0),b:()=>De(r,e,!1,!1)}:{a:()=>De(s,e,!0,!0),b:()=>De(e,r,!0,!0)}}},sW={kernelName:Uc,gradFunc:(e,t,n)=>{let{blockShape:r,crops:s}=n;return{x:()=>Vd(e,r,s)}}},aW={kernelName:rS,gradFunc:(e,t,n)=>{let r=n,s=r.inputShape,a=r.shape,o=Array.from(a);for(let c=s.length-1;c>=0;c--)if(s[c]===a[c])o[c]=1;else if(s[c]!==1)throw new Error(`broadcastTo(): [${s}] cannot be broadcast to [${a}].`);let i=[];for(let c=0;c<o.length;c++)o[c]>1&&i.push(c);return{x:()=>ge(e,i,!0)}}},oW={kernelName:wo,gradFunc:e=>({x:()=>e.clone()})},iW={kernelName:Io,gradFunc:e=>({x:()=>qe(e)})},cW={kernelName:ya,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{clipValueMin:s,clipValueMax:a}=n;return{x:()=>mn(Dr(Ds(r,s),Ta(r,a)),e,qe(e))}}},uW={kernelName:hd,inputsToSave:["x"],gradFunc:aC.gradFunc},lW={kernelName:Gc,saveAllInputs:!0,gradFunc:(e,t,n)=>{let r=t.map(c=>c.shape),{axis:s}=n,a=Fr(s,t[0].shape)[0],o=r.map(c=>c[a]);return zn(e,o,a).map(c=>()=>c)}},dW={kernelName:ko,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[r,s]=t,{dilations:a,strides:o,pad:i,dataFormat:c}=n;return D(ca(a),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${a}'`),{x:()=>fx(r.shape,e,s,o,i,c),filter:()=>Xx(r,e,s.shape,o,i,c)}}},pW={kernelName:So,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{let[r,s]=t,{strides:a,pad:o,dataFormat:i,dimRoundingMode:c}=n;return{dy:()=>Dt(e,s,a,o,i,1,c),filter:()=>Xx(e,r,s.shape,a,o,i,c)}}};function hW(e,t,n,r,s){let a=e;e.rank===4&&(a=W(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let o=t;o.rank===4&&(o=W(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),D(a.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${a.shape}.`),D(o.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${o.shape}.`),D(n.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),D(a.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${a.shape[4]}) must match input depth in filter (${n[3]}.`),D(o.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${o.shape[4]}) must match output depth for filter (${n[4]}).`);let i={x:a,dy:o},c={strides:r,pad:s,filterShape:n};return M.runKernel(yf,i,c)}var fW=z({conv3DBackpropFilter_:hW}),mW={kernelName:fd,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:r,strides:s,pad:a}=n;D(ca(r),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${r}'`);let[o,i]=t;return{x:()=>nT(o.shape,e,i,s,a),filter:()=>fW(o,e,i.shape,s,a)}}},gW={kernelName:To,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(xt(dm(ce(n,"float32"))),e)}}},bW={kernelName:Co,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(pm(ce(n,"float32")),e)}}},yW={kernelName:No,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:s,exclusive:a,reverse:o}=n;return{x:()=>{let i=iT([s],r.rank),c=Jf(e,s,a,!o);return i!=null&&(c=Ee(c,i)),c}}}},vW={kernelName:_o,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:r,strides:s,pad:a,dimRoundingMode:o}=n,i=r==null?[1,1]:r;D(ca(i),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${i}'`);let[c,u]=t;return D(c.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${c.rank}.`),D(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${u.rank}.`),D(c.shape[3]===u.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${c.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),D(ps(s,i),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${s} and dilations '${i}'.`),En("depthwiseConv2d",a,o),{x:()=>UT(c.shape,e,u,s,a,i,o),filter:()=>VT(c,e,u.shape,s,a,i,o)}}},xW={kernelName:md,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[r,s]=t,a={x:r,filter:s,dy:e},o={x:r,filter:s,dy:e};return{x:()=>M.runKernel($h,a,n),filter:()=>M.runKernel(Dh,o,n)}}},wW={kernelName:Ao,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t,r={dy:e,y:n};return{x:()=>M.runKernel(Tf,r)}}},IW={kernelName:Kc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,r=B(gn(xt(it(n))),2/Math.sqrt(Math.PI));return{x:()=>B(e,r)}}},kW={kernelName:$o,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,n)}}},SW={kernelName:Yc,inputsToSave:["input"],gradFunc:(e,t)=>{let[n]=t;return{input:()=>W(e,n.shape)}}},TW={kernelName:Zc,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,gn(n))}}},CW={kernelName:Do,gradFunc:e=>({x:()=>qe(e)})},NW={kernelName:Fo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=fe(e,ce(r,"float32")),c=Bt(n.shape,s);return c.length>0?W(ge(i,c),n.shape):i},b:()=>{let i=B(e,ce(n,"float32")),c=Bt(r.shape,s);c.length>0&&(i=W(ge(i,c),r.shape));let u=it(r);return xt(fe(i,ce(u,"float32")))}}}},_W={kernelName:Ro,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{let{varianceEpsilon:r}=n,[s,a,o,i]=t,c=i==null?xe(1):i,u=Bt(a.shape,s.shape),l=[];if(a.rank===1){for(let x=0;x<s.shape.length-1;++x)l.push(s.shape[x]);l.push(1)}let p=de(s,a),d=B(e,c),h=um(Y(o,xe(r))),f=B(B(B(h,h),h),xe(-.5));return{x:()=>a.rank===1?W(B(B(e,Ln(W(h,[1,1,1,a.shape[0]]),l)),c),s.shape):W(B(B(e,h),c),s.shape),mean:()=>{let x=B(B(h,xe(-1)),d);return a.rank===1&&(x=ge(x,u)),W(x,a.shape)},variance:()=>{let x=B(B(f,p),d);return a.rank===1&&(x=ge(x,u)),W(x,a.shape)},scale:()=>{let x=B(p,h),k=B(e,x);return a.rank===1&&(k=ge(k,u)),W(k,a.shape)},offset:()=>{let x=e;return a.rank===1&&(x=ge(x,u)),W(x,a.shape)}}}},EW={kernelName:Qc,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{let[r,s]=t,{axis:a}=n,o=Fr(a,r.shape)[0];return{x:()=>{let c=r.shape,u=s.size,l=c.slice(0,o),p=l.length,d=c.slice(a,c.length).slice(1),h=d.length,f=Ik(0,p),m=Ik(p+1,p+1+h),g=kk([l,[u],d]),b=W(e,g),y=W(s,[u]),v=kk([[p],f,m]),x=Ee(b,v),k=bm(x,y,r.shape[o]),S=kx(v);return k=Ee(k,S),k},indices:()=>s}}};function Ik(e,t){let n=[];for(let r=e;r<t;++r)n.push(r);return n}function kk(e){let t=[];for(let n=0;n<e.length;++n)for(let r=0;r<e[n].length;++r)t.push(e[n][r]);return t}var AW={kernelName:Po,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>qe(n),b:()=>qe(r)}}},$W={kernelName:Oo,gradFunc:e=>({x:()=>ce(e,"float32")})},DW={kernelName:nu,gradFunc:e=>({x:()=>qe(e)})},FW={kernelName:ru,gradFunc:e=>({x:()=>qe(e)})},RW={kernelName:su,gradFunc:e=>({x:()=>qe(e)})},PW={kernelName:Mo,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{alpha:s}=n,a=An(r,0);return{x:()=>mn(a,e,B(e,s))}}},OW={kernelName:iu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,Y(n,1))}}},MW={kernelName:Lo,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,ce(n,"float32"))}}},LW={kernelName:aS,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r]=t,{axis:s}=n;return{logits:()=>{let o=gn(r);return de(e,B(ge(e,s,!0),o))}}}};function zW(e,t,n,r=5,s=1,a=1,o=.5){let i={x:e,y:t,dy:n},c={depthRadius:r,bias:s,alpha:a,beta:o};return M.runKernel(Af,i,c)}var BW=z({localResponseNormalizationBackprop_:zW}),WW={kernelName:bd,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,s]=t,{depthRadius:a,bias:o,alpha:i,beta:c}=n;return{x:()=>BW(r,s,e,a,o,i,c)}}};function oC(e,t,n,r){return t.rank<n.rank&&(t=W(t,io(t.shape,r))),e.rank<n.rank&&(e=W(e,io(e.shape,r))),{x:()=>B(e,ce(tr(n,t),e.dtype))}}var Sk={kernelName:zo,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let r=n,{reductionIndices:s}=r,a=t[0],o=t[1],i=Fr(s,a.shape),c=oC(e,o,a,i);return{x:()=>c.x()}}},VW={kernelName:Bo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>B(e,ce(Ds(n,r),"float32")),b:()=>B(e,ce(em(n,r),"float32"))}}};function UW(e,t,n,r,s,a,o){let i=_(e,"dy","maxPool3dGrad"),c=_(t,"input","maxPool3dGrad"),u=_(n,"output","maxPool3dGrad"),l=i,p=c,d=u,h=!1;c.rank===4&&(h=!0,l=W(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),p=W(c,[1,c.shape[0],c.shape[1],c.shape[2],c.shape[3]]),d=W(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]])),D(l.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${l.rank}.`),D(p.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${p.rank}.`),D(d.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${d.rank}.`),En("maxPool3dGrad",a,o);let f={dy:l,input:p,output:d},m={filterSize:r,strides:s,pad:a,dimRoundingMode:o},g=M.runKernel(Df,f,m);return h?W(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var GW=z({maxPool3dGrad_:UW}),HW={kernelName:yd,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,s]=t,{filterSize:a,strides:o,pad:i,dimRoundingMode:c}=n;return{x:()=>GW(e,r,s,a,o,i,c)}}};function qW(e,t,n,r,s,a,o){let i=_(e,"dy","maxPoolGrad"),c=_(t,"input","maxPoolGrad"),u=_(n,"output","maxPoolGrad");D(c.rank===i.rank,()=>`Rank of input (${c.rank}) does not match rank of dy (${i.rank})`),D(i.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${i.rank}.`),D(c.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${c.rank}.`),En("maxPoolGrad",a,o);let l={dy:i,input:c,output:u},p={filterSize:r,strides:s,pad:a,dimRoundingMode:o};return M.runKernel($f,l,p)}var jW=z({maxPoolGrad_:qW}),KW={kernelName:Wo,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,s]=t,{filterSize:a,strides:o,pad:i}=n;return{x:()=>jW(e,r,s,a,o,i)}}},XW={kernelName:Vo,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:s}=n,a=Fr(s,r.shape),i=oT(r.shape,a)[1],c=ht(i);return{x:()=>{let l=r.shape.slice();a.forEach(h=>{l[h]=1});let p=W(e,l);return fe(B(p,Qn(r.shape,"float32")),c)}}}},YW={kernelName:Uo,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let r=n,{axis:s}=r,[a,o]=t,i=Fr(s,a.shape),c=oC(e,o,a,i);return{x:()=>c.x()}}},ZW={kernelName:Go,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>B(e,ce(Ta(n,r),"float32")),b:()=>B(e,ce(An(n,r),"float32"))}}},JW={kernelName:Ho,inputsToSave:["x"],gradFunc:(e,t,n)=>{let r=t[0],{paddings:s}=n,a=s.map(o=>o[0]);return{x:()=>We(e,a,r.shape)}}},QW={kernelName:du,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=Bt(n.shape,s);return i.length>0?W(ge(e,i),n.shape):e},b:()=>{let i=B(e,xt(Bu(fe(n,r)))),c=Bt(r.shape,s);return c.length>0?W(ge(i,c),r.shape):i}}}},e4={kernelName:qo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=B(e,ce(r,"float32")),c=Bt(n.shape,s);return c.length>0?W(ge(i,c),n.shape):i},b:()=>{let i=B(e,ce(n,"float32")),c=Bt(r.shape,s);return c.length>0?W(ge(i,c),r.shape):i}}}},t4={kernelName:pu,gradFunc:e=>({x:()=>xt(e)})},n4={kernelName:jo,inputsToSave:["indices"],gradFunc:(e,t)=>{let n=t[0];return{indices:()=>kt(n.shape,"float32")}}},r4={kernelName:bu,gradFunc:e=>({x:()=>qe(e)})},s4={kernelName:yu,saveAllInputs:!0,gradFunc:(e,t,n)=>{let{axis:r}=n;return lt(e,r).map(a=>()=>a)}},Tk={kernelName:Ko,inputsToSave:["x"],gradFunc:(e,t,n)=>{let r=t[0],{paddings:s}=n,a=s.map(o=>o[0]);return{x:()=>We(e,a,r.shape)}}},a4={kernelName:Xo,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{let[n,r,s]=t,a=n,o=r,i=ct(a.shape,o.shape);return{a:()=>{let l=ce(o,"float32"),p=B(e,B(l,$s(a,de(l,xe(1))))),d=Bt(a.shape,i);return d.length>0&&(p=ge(p,d)),W(p,a.shape)},b:()=>{let l=An(a,0),p=mn(l,nr(a),qe(a)),d=B(e,B(s,p)),h=Bt(o.shape,i);return h.length>0&&(d=ge(d,h)),W(d,o.shape)}}}},o4={kernelName:Yo,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{let[n,r]=t,s=An(n,0);return{x:()=>mn(s,e,B(e,r)),alpha:()=>{let a=mn(s,qe(e),B(e,n)),o=Bt(r.shape,e.shape);return o.length>0&&(a=ge(a,o)),W(a,r.shape)}}}};function i4(e,t,n){let r=e.shape.slice();r[n]=1;let s=W(t,r),a=ed(e,n,!0,!1),o=ed(e,n,!0,!0),i=B(a,o);return B(s,i)}function c4(e,t,n){let r=e.shape.length,s=r-n.length,a=N.getAxesPermutation(n,r),o=e;a!=null&&(o=Ee(e,a));let i=o.shape.slice(),u=i.splice(r-n.length,n.length).reduce((d,h)=>d*h,1);i.push(u);let l=o.reshape(i),p=i4(l,t,s);if(p=p.reshape(o.shape),a!=null){let d=N.getUndoAxesPermutation(a);p=Ee(p,d)}return p}var u4={kernelName:Zo,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:s}=n,a=[];return s==null?a=r.shape.map((o,i)=>i):typeof s=="number"?a=[s]:a=s,{x:()=>c4(r,e,a)}}},l4={kernelName:Eo,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=fe(e,ce(r,"float32")),c=Bt(n.shape,s);return c.length>0?W(ge(i,c),n.shape):i},b:()=>{let i=B(e,ce(n,"float32")),c=Bt(r.shape,s);c.length>0&&(i=W(ge(i,c),r.shape));let u=it(r);return xt(fe(i,ce(u,"float32")))}}}},d4={kernelName:vu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,xt(it(n)))}}},p4={kernelName:ti,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,r=B(Ta(n,6),bi(n));return{x:()=>B(e,ce(r,"float32"))}}},h4={kernelName:Jo,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,ce(bi(n),"float32"))}}},f4={kernelName:xu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>W(e,n.shape)}}},m4={kernelName:ei,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[r]=t,s={dy:e,images:r};return{images:()=>M.runKernel(Bf,s,n)}}},g4={kernelName:Qo,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[r]=t,s={dy:e,images:r};return{images:()=>M.runKernel(zf,s,n)}}},b4={kernelName:ni,gradFunc:(e,t,n)=>{let{dims:r}=n,s=Fr(r,e.shape);return{x:()=>fr(e,s)}}},y4={kernelName:ri,gradFunc:e=>({x:()=>qe(e)})},v4={kernelName:si,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>xt(fe(e,B($s(n,1.5),2)))}}},x4={kernelName:Iu,inputsToSave:["condition"],gradFunc:(e,t)=>{let[n]=t;return{condition:()=>ce(qe(n),"float32"),t:()=>B(e,ce(n,e.dtype)),e:()=>B(e,ce(Bd(n),e.dtype))}}},w4={kernelName:ku,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let r=An(n,xe(0)),s=xe(nC),a=xe(rC),o=B(e,a),i=B(B(e,s),gn(ce(n,"float32")));return mn(r,o,i)}}}},I4={kernelName:oi,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,B(n,de(xe(1),n)))}}},k4={kernelName:Cu,gradFunc:e=>({x:()=>qe(e)})},S4={kernelName:ai,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(Md(ce(n,"float32")),e)}}},T4={kernelName:Tu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(Zf(ce(n,"float32")),e)}}},C4={kernelName:Su,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{begin:s,size:a}=n,o=r.shape,[i,c]=HS(r,s,a),u=[];for(let l=0;l<e.rank;l++)u.push([i[l],o[l]-i[l]-c[l]]);return{x:()=>gr(e,u)}}},N4={kernelName:ui,outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r]=t,{dim:s}=n,a=!0,o=B(e,r);return{logits:()=>de(o,B(ge(o,[s],a),r))}}},_4={kernelName:Nu,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,dr(n))}}},Ck={kernelName:_u,gradFunc:(e,t,n)=>{let{blockShape:r,paddings:s}=n;return{x:()=>Od(e,r,s)}}},Nk={kernelName:Eu,gradFunc:(e,t,n)=>{let{axis:r}=n;return{x:()=>Je(e,r)}}},E4={kernelName:ii,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,B(un(ce(n,"float32")),2))}}},A4={kernelName:kd,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(e,B(ce(n,"float32"),2))}}},$4={kernelName:li,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=xe(2);return{a:()=>B(e,B(s,de(n,r))),b:()=>B(e,B(s,de(r,n)))}}},D4={kernelName:xa,gradFunc:e=>({x:()=>qe(e)})},F4={kernelName:di,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,s=ct(n.shape,r.shape);return{a:()=>{let i=e,c=Bt(n.shape,s);return c.length>0&&(i=ge(i,c)),W(i,n.shape)},b:()=>{let i=e,c=Bt(r.shape,s);return c.length>0&&(i=ge(i,c)),W(xt(i),r.shape)}}}},R4={kernelName:ci,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,s=r.shape.slice(),{axis:a}=n;Fr(a,r.shape).forEach(u=>{s[u]=1});let i=W(e,s),c=B(i,Qn(r.shape,"float32"));return{x:()=>c}}},P4={kernelName:pi,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>fe(e,it(Md(n)))}}},O4={kernelName:hi,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>B(de(xe(1),it(n)),e)}}},M4={kernelName:va,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{reps:s}=n;return{x:()=>{let o=qe(r);if(r.rank===1)for(let i=0;i<s[0];++i)o=Y(o,We(e,[i*r.shape[0]],[r.shape[0]]));else if(r.rank===2)for(let i=0;i<s[0];++i)for(let c=0;c<s[1];++c)o=Y(o,We(e,[i*r.shape[0],c*r.shape[1]],[r.shape[0],r.shape[1]]));else if(r.rank===3)for(let i=0;i<s[0];++i)for(let c=0;c<s[1];++c)for(let u=0;u<s[2];++u)o=Y(o,We(e,[i*r.shape[0],c*r.shape[1],u*r.shape[2]],[r.shape[0],r.shape[1],r.shape[2]]));else if(r.rank===4)for(let i=0;i<s[0];++i)for(let c=0;c<s[1];++c)for(let u=0;u<s[2];++u)for(let l=0;l<s[3];++l)o=Y(o,We(e,[i*r.shape[0],c*r.shape[1],u*r.shape[2],l*r.shape[3]],[r.shape[0],r.shape[1],r.shape[2],r.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${r.rank} tensors yet.`);return o}}}},L4={kernelName:_s,gradFunc:(e,t,n)=>{let r=n,{perm:s}=r,a=kx(s);return{x:()=>Ee(e,a)}}},z4={kernelName:Ru,gradFunc:(e,t,n)=>{let r=n,{axis:s}=r;return{value:()=>Ft(e,s)}}},B4={kernelName:Nd,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>W4(e,n)}}};function W4(e,t){let n=hs(t,qe(t)),r=Wu(e,n),s=Ds(t,xe(0,"int32")),a=r.rank-s.rank;for(let i=0;i<a;++i)s=Zt(s,i+1);s=Dr(s,Qn(r.shape,"bool"));let o=qe(r);return mn(s,r,o)}var V4={kernelName:Pu,gradFunc:e=>({x:()=>qe(e)})},U4=[aC,BB,WB,VB,UB,GB,HB,qB,jB,KB,XB,YB,QB,nW,rW,sW,aW,oW,iW,cW,uW,lW,pW,dW,mW,gW,bW,yW,vW,xW,l4,wW,IW,kW,SW,TW,NW,CW,_W,EW,AW,$W,DW,FW,RW,PW,OW,MW,LW,WW,Sk,Sk,VW,HW,KW,XW,YW,ZW,JW,QW,e4,t4,n4,r4,s4,Tk,Tk,a4,o4,u4,d4,p4,h4,f4,m4,g4,b4,y4,v4,x4,w4,I4,k4,S4,T4,C4,N4,_4,Ck,Ck,Nk,Nk,E4,$4,A4,D4,F4,R4,P4,O4,M4,L4,z4,B4,V4];for(let e of U4)oS(e);Q().prototype.abs=function(){return this.throwIfDisposed(),Lt(this)};Q().prototype.acos=function(){return this.throwIfDisposed(),Xv(this)};Q().prototype.acosh=function(){return this.throwIfDisposed(),Yv(this)};Q().prototype.add=function(e){return this.throwIfDisposed(),Y(this,e)};Q().prototype.all=function(e,t){return this.throwIfDisposed(),Kf(this,e,t)};Q().prototype.any=function(e,t){return this.throwIfDisposed(),Ql(this,e,t)};Q().prototype.argMax=function(e){return this.throwIfDisposed(),ao(this,e)};Q().prototype.argMin=function(e){return this.throwIfDisposed(),Zv(this,e)};Q().prototype.asScalar=function(){return this.throwIfDisposed(),D(this.size===1,()=>"The array must have only 1 element."),W(this,[])};Q().prototype.asType=function(e){return this.throwIfDisposed(),ce(this,e)};Q().prototype.as1D=function(){return this.throwIfDisposed(),W(this,[this.size])};Q().prototype.as2D=function(e,t){return this.throwIfDisposed(),W(this,[e,t])};Q().prototype.as3D=function(e,t,n){return this.throwIfDisposed(),W(this,[e,t,n])};Q().prototype.as4D=function(e,t,n,r){return this.throwIfDisposed(),W(this,[e,t,n,r])};Q().prototype.as5D=function(e,t,n,r,s){return this.throwIfDisposed(),W(this,[e,t,n,r,s])};Q().prototype.asin=function(){return this.throwIfDisposed(),Jv(this)};Q().prototype.asinh=function(){return this.throwIfDisposed(),Qv(this)};Q().prototype.atan=function(){return this.throwIfDisposed(),ex(this)};Q().prototype.atan2=function(e){return this.throwIfDisposed(),tx(this,e)};Q().prototype.atanh=function(){return this.throwIfDisposed(),nx(this)};Q().prototype.avgPool=function(e,t,n,r){return this.throwIfDisposed(),mr(this,e,t,n,r)};Q().prototype.batchToSpaceND=function(e,t){return this.throwIfDisposed(),Od(this,e,t)};Q().prototype.batchNorm=function(e,t,n,r,s){return this.throwIfDisposed(),ka(this,e,t,n,r,s)};Q().prototype.broadcastTo=function(e){return this.throwIfDisposed(),Ya(this,e)};Q().prototype.cast=function(e){return this.throwIfDisposed(),ce(this,e)};Q().prototype.ceil=function(){return this.throwIfDisposed(),ux(this)};Q().prototype.clipByValue=function(e,t){return this.throwIfDisposed(),en(this,e,t)};Q().prototype.concat=function(e,t){return this.throwIfDisposed(),e instanceof Te&&(e=[e]),Je([this,...e],t)};Q().prototype.conv1d=function(e,t,n,r,s,a){return this.throwIfDisposed(),Xf(this,e,t,n,r,s,a)};Q().prototype.conv2dTranspose=function(e,t,n,r,s){return this.throwIfDisposed(),Yf(this,e,t,n,r,s)};Q().prototype.conv2d=function(e,t,n,r,s,a){return this.throwIfDisposed(),Dt(this,e,t,n,r,s,a)};Q().prototype.cos=function(){return this.throwIfDisposed(),Md(this)};Q().prototype.cosh=function(){return this.throwIfDisposed(),Zf(this)};Q().prototype.cumprod=function(e,t,n){return this.throwIfDisposed(),ed(this,e,t,n)};Q().prototype.cumsum=function(e,t,n){return this.throwIfDisposed(),Jf(this,e,t,n)};Q().prototype.depthToSpace=function(e,t){return this.throwIfDisposed(),bx(this,e,t)};Q().prototype.depthwiseConv2d=function(e,t,n,r,s,a){return this.throwIfDisposed(),Sa(this,e,t,n,r,s,a)};Q().prototype.dilation2d=function(e,t,n,r,s){return this.throwIfDisposed(),yx(this,e,t,n,r,s)};Q().prototype.divNoNan=function(e){return this.throwIfDisposed(),vx(this,e)};Q().prototype.div=function(e){return this.throwIfDisposed(),fe(this,e)};Q().prototype.dot=function(e){return this.throwIfDisposed(),xx(this,e)};Q().prototype.elu=function(){return this.throwIfDisposed(),Lu(this)};Q().prototype.equal=function(e){return this.throwIfDisposed(),tr(this,e)};Q().prototype.erf=function(){return this.throwIfDisposed(),wx(this)};Q().prototype.euclideanNorm=function(e,t){return this.throwIfDisposed(),Sx(this,e,t)};Q().prototype.exp=function(){return this.throwIfDisposed(),gn(this)};Q().prototype.expandDims=function(e){return this.throwIfDisposed(),Zt(this,e)};Q().prototype.expm1=function(){return this.throwIfDisposed(),Tx(this)};Q().prototype.fft=function(){return this.throwIfDisposed(),Hd(this)};Q().prototype.flatten=function(){return this.throwIfDisposed(),W(this,[this.size])};Q().prototype.floor=function(){return this.throwIfDisposed(),Bu(this)};Q().prototype.floorDiv=function(e){return this.throwIfDisposed(),jf(this,e)};Q().prototype.gather=function(e,t){return this.throwIfDisposed(),Wu(this,e,t)};Q().prototype.greaterEqual=function(e){return this.throwIfDisposed(),Ds(this,e)};Q().prototype.greater=function(e){return this.throwIfDisposed(),An(this,e)};Q().prototype.ifft=function(){return this.throwIfDisposed(),wc(this)};Q().prototype.irfft=function(){return this.throwIfDisposed(),fm(this)};Q().prototype.isFinite=function(){return this.throwIfDisposed(),Cx(this)};Q().prototype.isInf=function(){return this.throwIfDisposed(),Nx(this)};Q().prototype.isNaN=function(){return this.throwIfDisposed(),_x(this)};Q().prototype.leakyRelu=function(e){return this.throwIfDisposed(),Ld(this,e)};Q().prototype.lessEqual=function(e){return this.throwIfDisposed(),Ta(this,e)};Q().prototype.less=function(e){return this.throwIfDisposed(),em(this,e)};Q().prototype.localResponseNormalization=function(e,t,n,r){return this.throwIfDisposed(),Ex(this,e,t,n,r)};Q().prototype.logSigmoid=function(){return this.throwIfDisposed(),Ax(this)};Q().prototype.logSoftmax=function(e){return this.throwIfDisposed(),nm(this,e)};Q().prototype.logSumExp=function(e,t){return this.throwIfDisposed(),rm(this,e,t)};Q().prototype.log=function(){return this.throwIfDisposed(),nr(this)};Q().prototype.log1p=function(){return this.throwIfDisposed(),zd(this)};Q().prototype.logicalAnd=function(e){return this.throwIfDisposed(),Dr(this,e)};Q().prototype.logicalNot=function(){return this.throwIfDisposed(),Bd(this)};Q().prototype.logicalOr=function(e){return this.throwIfDisposed(),sm(this,e)};Q().prototype.logicalXor=function(e){return this.throwIfDisposed(),$x(this,e)};Q().prototype.matMul=function(e,t,n){return this.throwIfDisposed(),De(this,e,t,n)};Q().prototype.maxPool=function(e,t,n,r){return this.throwIfDisposed(),Rt(this,e,t,n,r)};Q().prototype.max=function(e,t){return this.throwIfDisposed(),pr(this,e,t)};Q().prototype.maximum=function(e){return this.throwIfDisposed(),hs(this,e)};Q().prototype.mean=function(e,t){return this.throwIfDisposed(),Ct(this,e,t)};Q().prototype.min=function(e,t){return this.throwIfDisposed(),yc(this,e,t)};Q().prototype.minimum=function(e){return this.throwIfDisposed(),Vu(this,e)};Q().prototype.mirrorPad=function(e,t){return this.throwIfDisposed(),Fx(this,e,t)};Q().prototype.mod=function(e){return this.throwIfDisposed(),Rx(this,e)};Q().prototype.mul=function(e){return this.throwIfDisposed(),B(this,e)};Q().prototype.neg=function(){return this.throwIfDisposed(),xt(this)};Q().prototype.norm=function(e,t,n){return this.throwIfDisposed(),zu(this,e,t,n)};Q().prototype.notEqual=function(e){return this.throwIfDisposed(),co(this,e)};Q().prototype.oneHot=function(e,t=1,n=0){return this.throwIfDisposed(),gc(this,e,t,n)};Q().prototype.onesLike=function(){return this.throwIfDisposed(),rr(this)};Q().prototype.pad=function(e,t){return this.throwIfDisposed(),gr(this,e,t)};Q().prototype.pool=function(e,t,n,r,s,a){return this.throwIfDisposed(),Px(this,e,t,n,r,s,a)};Q().prototype.pow=function(e){return this.throwIfDisposed(),$s(this,e)};Q().prototype.prelu=function(e){return this.throwIfDisposed(),Ud(this,e)};Q().prototype.prod=function(e,t){return this.throwIfDisposed(),Ox(this,e,t)};Q().prototype.reciprocal=function(){return this.throwIfDisposed(),zx(this)};Q().prototype.relu=function(){return this.throwIfDisposed(),Xe(this)};Q().prototype.relu6=function(){return this.throwIfDisposed(),im(this)};Q().prototype.reshapeAs=function(e){return this.throwIfDisposed(),W(this,e.shape)};Q().prototype.reshape=function(e){return this.throwIfDisposed(),W(this,e)};Q().prototype.resizeBilinear=function(e,t,n){return this.throwIfDisposed(),XT(this,e,t,n)};Q().prototype.resizeNearestNeighbor=function(e,t,n){return this.throwIfDisposed(),YT(this,e,t,n)};Q().prototype.reverse=function(e){return this.throwIfDisposed(),fr(this,e)};Q().prototype.rfft=function(){return this.throwIfDisposed(),qd(this)};Q().prototype.round=function(){return this.throwIfDisposed(),cm(this)};Q().prototype.rsqrt=function(){return this.throwIfDisposed(),um(this)};Q().prototype.selu=function(){return this.throwIfDisposed(),lm(this)};Q().prototype.separableConv2d=function(e,t,n,r,s,a){return this.throwIfDisposed(),Ca(this,e,t,n,r,s,a)};Q().prototype.sigmoid=function(){return this.throwIfDisposed(),dr(this)};Q().prototype.sign=function(){return this.throwIfDisposed(),Bx(this)};Q().prototype.sin=function(){return this.throwIfDisposed(),dm(this)};Q().prototype.sinh=function(){return this.throwIfDisposed(),pm(this)};Q().prototype.slice=function(e,t){return this.throwIfDisposed(),We(this,e,t)};Q().prototype.softmax=function(e){return this.throwIfDisposed(),Xr(this,e)};Q().prototype.softplus=function(){return this.throwIfDisposed(),mi(this)};Q().prototype.spaceToBatchND=function(e,t){return this.throwIfDisposed(),Vd(this,e,t)};Q().prototype.split=function(e,t){return this.throwIfDisposed(),zn(this,e,t)};Q().prototype.sqrt=function(){return this.throwIfDisposed(),un(this)};Q().prototype.square=function(){return this.throwIfDisposed(),it(this)};Q().prototype.squaredDifference=function(e){return this.throwIfDisposed(),mm(this,e)};Q().prototype.squeeze=function(e){return this.throwIfDisposed(),Na(this,e)};Q().prototype.stack=function(e,t){this.throwIfDisposed();let n=e instanceof Te?[this,e]:[this,...e];return Ft(n,t)};Q().prototype.step=function(e){return this.throwIfDisposed(),bi(this,e)};Q().prototype.stridedSlice=function(e,t,n,r,s,a,o,i){return this.throwIfDisposed(),Wx(this,e,t,n,r,s,a,o,i)};Q().prototype.sub=function(e){return this.throwIfDisposed(),de(this,e)};Q().prototype.sum=function(e,t){return this.throwIfDisposed(),ge(this,e,t)};Q().prototype.tan=function(){return this.throwIfDisposed(),Vx(this)};Q().prototype.tanh=function(){return this.throwIfDisposed(),oo(this)};Q().prototype.tile=function(e){return this.throwIfDisposed(),Ln(this,e)};Q().prototype.toBool=function(){return this.throwIfDisposed(),ce(this,"bool")};Q().prototype.toFloat=function(){return this.throwIfDisposed(),ce(this,"float32")};Q().prototype.toInt=function(){return this.throwIfDisposed(),ce(this,"int32")};Q().prototype.topk=function(e,t){return this.throwIfDisposed(),Ux(this,e,t)};Q().prototype.transpose=function(e){return this.throwIfDisposed(),Ee(this,e)};Q().prototype.unique=function(e){return this.throwIfDisposed(),Gx(this,e)};Q().prototype.unsortedSegmentSum=function(e,t){return this.throwIfDisposed(),bm(this,e,t)};Q().prototype.unstack=function(e){return this.throwIfDisposed(),lt(this,e)};Q().prototype.where=function(e,t){return this.throwIfDisposed(),mn(e,this,t)};Q().prototype.zerosLike=function(){return this.throwIfDisposed(),qe(this)};var ks=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,ks.prototype)}},Wr=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Wr.prototype)}},U=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,U.prototype)}},Pe=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Pe.prototype)}},iC=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,iC.prototype)}},cC=class{constructor(e){this.maxEntries=e||100,this.cache=new Map}get(e){let t;return this.cache.has(e)&&(t=this.cache.get(e),this.cache.delete(e),this.cache.set(e,t)),t}put(e,t){if(this.cache.has(e))this.cache.delete(e);else if(this.cache.size>=this.maxEntries){let n=this.cache.keys().next().value;this.cache.delete(n)}this.cache.set(e,t)}getMaxEntries(){return this.maxEntries}setMaxEntries(e){if(e<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${e}.`);if(this.maxEntries>e)for(let t=0;t<this.maxEntries-e;t++){let n=this.cache.keys().next().value;this.cache.delete(n)}this.maxEntries=e}};function uo(e,t){if(Array.isArray(e)){let n=[];for(let r=0;r<t;r++)n=n.concat(e);return n}else{let n=new Array(t);return n.fill(e),n}}function ss(e,t){if(!e)throw new iC(t)}function _k(e,t){let n=0;for(let r of e)r===t&&n++;return n}function Mn(e){return e.length===1?e[0]:e}function vt(e){return Array.isArray(e)?e:[e]}function Ss(e){let n=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return n[0]!=="_"?n:"private"+n}function Ga(e){return e.length<=1||e.indexOf("_")===-1?e:e.replace(/[_]+(\w|$)/g,(t,n)=>n.toUpperCase())}var Tr={};function ew(e){if(e==null)return null;let t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function Uy(e){if(!(e==null||typeof e!="object"))if(Array.isArray(e))e.forEach(t=>Uy(t));else{let t=Object.keys(e);for(let n of t){let r=e[n];r!=null&&typeof r=="object"&&(!Array.isArray(r)&&r.type==="ndarray"&&typeof r.value=="number"?e[n]=r.value:Uy(r))}}}function Kd(e,t={},n={},r="object",s=!1){if(typeof e=="string"){let a=e,o;if(a in n)o=n[a];else if(a in Tr)o=Tr[a];else if(o=t[a],o==null)throw new U(`Unknown ${r}: ${e}. This may be due to one of the following reasons:
+1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return o}else{let a=e;if(a.className==null||a.config==null)throw new U(`${r}: Improper config format: ${JSON.stringify(a)}.
+'className' and 'config' must set.`);let o=a.className,i,c;if(o in n?[i,c]=n[o]:o in Tr?[i,c]=Tr.className:o in t&&([i,c]=t[o]),i==null)throw new U(`Unknown ${r}: ${o}. This may be due to one of the following reasons:
+1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(c!=null){let u={};for(let h of Object.keys(Tr))u[h]=Tr[h];for(let h of Object.keys(n))u[h]=n[h];let l=a.config;l.customObjects=u;let p=Object.assign({},Tr);for(let h of Object.keys(n))Tr[h]=n[h];Uy(a.config);let d=c(i,a.config,n,s);return Tr=Object.assign({},p),d}else{let u=Object.assign({},Tr);for(let p of Object.keys(n))Tr[p]=n[p];let l=new i(a.config);return Tr=Object.assign({},u),l}}}function G4(e,t){return e<t?-1:e>t?1:0}function ih(e,t){return-1*G4(e,t)}function sa(e){if(e==null)return e;let t=[];for(let n of e)t.indexOf(n)===-1&&t.push(n);return t}function H4(e){if(e==null)throw new U(`Invalid value in obj: ${JSON.stringify(e)}`);for(let t in e)if(e.hasOwnProperty(t))return!1;return!0}function yi(e,t,n){if(n!=null&&e.indexOf(n)<0)throw new U(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)}function tw(e,t,n=0,r=1/0){return ss(n>=0),ss(r>=n),Array.isArray(e)&&e.length>=n&&e.length<=r&&e.every(s=>typeof s===t)}function Qt(e,t){Array.isArray(e)?(w.assert(e.length>0,()=>`${t} is unexpectedly an empty array.`),e.forEach((n,r)=>Qt(n,`element ${r+1} of ${t}`))):w.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${uC(e)}.`)}function uC(e){return e===null?"null":Array.isArray(e)?"["+e.map(t=>uC(t)).join(",")+"]":typeof e=="string"?`"${e}"`:`${e}`}function q4(e,t,n){let r=n!=null?n():w.now(),s;return(...o)=>{let i=n!=null?n():w.now();return i-r<t||(r=i,s=e(...o)),s}}function lC(e){return e==="relu"?"relu":e==="linear"?"linear":e==="elu"?"elu":null}var j4=0;function dC(){return j4++}var ch={};function Em(e=""){return e in ch||(ch[e]=0),ch[e]+=1,e+ch[e].toString()}var K4=["channelsFirst","channelsLast"],X4=["nearest","bilinear"],Y4=["valid","same","causal"],Z4=["max","avg"],J4=["sum","mul","concat","ave"],rc=new Map;function Pt(e){yi(K4,"DataFormat",e)}function Q4(e){yi(X4,"InterpolationFormat",e)}function br(e){yi(Y4,"PaddingMode",e)}function pC(e){yi(Z4,"PoolMode",e)}var Hl=[],Ek="/";function Za(e,t){Hl.push(e);try{let n=t();return Hl.pop(),n}catch(n){throw Hl.pop(),n}}function eV(){return Hl.length===0?"":Hl.join(Ek)+Ek}function hC(e){if(!mC(e))throw new Error("Not a valid tensor name: '"+e+"'");return eV()+e}function fC(e){if(!mC(e))throw new Error("Not a valid tensor name: '"+e+"'");rc.has(e)||rc.set(e,0);let t=rc.get(e);if(rc.set(e,rc.get(e)+1),t>0){let n=`${e}_${t}`;return rc.set(n,1),n}else return e}var tV=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function mC(e){return!!e.match(tV)}function nV(e){return e===parseInt(e.toString(),10)}function aa(e,t,n){t==null&&(t=0),n==null&&(n=e.length);let r=1;for(let s=t;s<n;++s)r*=e[s];return r}function kc(e){if(e.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let n=0;n<e.length;n++){let r=e[n];r<t&&(t=r)}return t}function ua(e){if(e.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let n=0;n<e.length;n++){let r=e[n];r>t&&(t=r)}return t}function qr(e,t){if(t<e)throw new U(`end (${t}) < begin (${e}) is forbidden.`);let n=[];for(let r=e;r<t;++r)n.push(r);return n}var vy;function Gt(){return vy==null&&(vy=DS().epsilon()),vy}function jr(){return"channelsLast"}function vi(e,t){return ce(e,t)}function Xd(e,t=-1){let n=e.shape.slice();return t<0&&(t=n.length+t+1),n.splice(t,0,1),W(e,n)}function rV(e,t){return O(()=>{if(e.shape.length!==2)throw new U(`repeat() expects a rank-2 tensor, but received a rank-${e.shape.length} tensor.`);let n=Xd(e,1);return Gy(n,[1,t,1])})}function sV(e){let t=[aa(e.shape)];return W(e,t)}function aV(e){if(e.rank<=1)throw new U(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);let t=[e.shape[0],aa(e.shape,1)];return W(e,t)}function Ja(e,t,n){return O(()=>{switch(e.rank){case 1:return Gd(e,t,n);case 2:return hm(e,[t,0],[n,e.shape[1]]);case 3:return gi(e,[t,0,0],[n,e.shape[1],e.shape[2]]);case 4:return xc(e,[t,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3]]);case 5:return We(e,[t,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return We(e,[t,0,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new U(`sliceAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}})}function xy(e,t,n){return O(()=>{switch(e.rank){case 1:return Gd(e,t,n);case 2:return hm(e,[0,t],[e.shape[0],n]);case 3:return gi(e,[0,0,t],[e.shape[0],e.shape[1],n]);case 4:return xc(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],n]);default:throw new U(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function uh(e,t,n,r){return O(()=>{switch(e.rank){case 1:return Gd(e,t,n);case 2:switch(r){case 1:return Ja(e,t,n);case 2:return xy(e,t,n);default:throw new U(`The axis is not within the rank of the tensor ${r}`)}case 3:switch(r){case 1:return Ja(e,t,n);case 2:return gi(e,[0,t,0],[e.shape[0],n,e.shape[2]]);case 3:return xy(e,t,n);default:throw new U(`The axis is not within the rank of the tensor ${r}`)}case 4:switch(r){case 1:return Ja(e,t,n);case 2:return xc(e,[0,t,0,0],[e.shape[0],n,e.shape[2],e.shape[3]]);case 3:return xc(e,[0,0,t,0],[e.shape[0],e.shape[1],n,e.shape[3]]);case 4:return xy(e,t,n);default:throw new U(`The axis is not within the rank of the tensor ${r}`)}default:throw new U(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function nw(e,t=-1){let n;return t<0&&(n=e[0].rank,n!==0?t=n:t=0),t===e[0].rank&&(t=-1),Je(e,t)}function Ak(e,t){switch(e.rank){case 1:return lx([e,t]);case 2:return dx([e,t],0);case 3:return px([e,t],0);case 4:return hx([e,t],0);default:throw new U(`concatAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}}function Gy(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new U(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return Ln(e,t)}function Am(e,t=0,n=1,r,s){return om(e,t,n,r,s)}function cs(e,t,n,r){if(e.rank<2||t.rank<2)throw new Pe(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){let s=e.shape.slice(-1)[0],a=t.shape.slice(-2)[0];if(s!==a)throw new Pe(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)}if(e.rank===2&&t.rank===2)return Ic.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:r?Hy(e.rank,r,jr()):null,activation:n});{let s=e.shape.slice(),a=s.pop();e=W(e,[-1,a]);let o=t.shape.slice(),i=o.pop(),c=o.pop(),u=[...o,i],l=Array.from({length:t.rank},(f,m)=>m===0?t.rank-2:m<=t.rank-2?m-1:m);t=W(Ee(t,l),[c,-1]);let p=[...s,...u],d=!1,h=!1;return W(Ic.matMul({a:e,b:t,transposeA:d,transposeB:h,bias:r?Hy(e.rank,r,jr()):null,activation:n}),p)}}function gC(e,t,n){return O(()=>(Array.isArray(t)?t=Ke(t,"int32"):t=ce(t,"int32"),Wu(e,t,n)))}function Yd(e){return B(e,e)}function Hy(e,t,n){let r=t.shape;if(t.rank!==1&&t.rank!==e)throw new U(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(e===5){if(n==="channelsFirst")return r.length===1?W(t,[1,r[0],1,1,1]):W(t,[1,r[3],r[0],r[1],r[2]]);if(n==="channelsLast")return r.length===1?W(t,[1,1,1,1,r[0]]):W(t,[1].concat(r))}else if(e===4){if(n==="channelsFirst")return r.length===1?W(t,[1,r[0],1,1]):W(t,[1,r[2],r[0],r[1]]);if(n==="channelsLast")return r.length===1?W(t,[1,1,1,r[0]]):W(t,[1].concat(r))}else if(e===3){if(n==="channelsFirst")return r.length===1?W(t,[1,r[0],1]):W(t,[1,r[1],r[0]]);if(n==="channelsLast")return r.length===1?W(t,[1,1,r[0]]):W(t,[1].concat(r))}else if(e<3)return t;throw new U(`Unsupported input rank by biasAdd: ${t.rank}`)}function Yr(e,t,n){return O(()=>(n==null&&(n=jr()),Pt(n),Y(e,Hy(e.rank,t,n))))}function oV(e,t=1){if(t!==1)throw new Pe(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Lu(e)}function iV(e){return O(()=>fe(e,Y(Lt(e),1)))}function bC(e,t,n,r){return O(()=>jx(e,t,n,r))}function cV(e){return O(()=>{let t=Y(.5,B(.2,e));return en(t,0,1)})}function Zd(e,t,n=!1){return n?e():t()}var uV=["fanIn","fanOut","fanAvg"],lV=["normal","uniform","truncatedNormal"];function dV(e){yi(uV,"FanMode",e)}function pV(e){yi(lV,"Distribution",e)}var Pr=class extends se.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},rw=class extends Pr{apply(e,t){return kt(e,t)}};rw.className="Zeros";se.registerClass(rw);var $m=class extends Pr{apply(e,t){return Qn(e,t)}};$m.className="Ones";se.registerClass($m);var sw=class extends Pr{constructor(e){if(super(),typeof e!="object")throw new U(`Expected argument of type ConstantConfig but got ${e}`);if(e.value===void 0)throw new U(`config must have value set but got ${e}`);this.value=e.value}apply(e,t){return O(()=>B(xe(this.value),Qn(e,t)))}getConfig(){return{value:this.value}}};sw.className="Constant";se.registerClass(sw);var aw=class extends Pr{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return Uu(e,this.minval,this.maxval,t)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};aw.className="RandomUniform";se.registerClass(aw);var ow=class extends Pr{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Pe(`randomNormal does not support dType ${t}.`);return Am(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};ow.className="RandomNormal";se.registerClass(ow);var iw=class extends Pr{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Pe(`truncatedNormal does not support dType ${t}.`);return gm(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};iw.className="TruncatedNormal";se.registerClass(iw);var cw=class extends Pr{constructor(e){super(),this.gain=e.gain!=null?e.gain:1}apply(e,t){return O(()=>{if(e.length!==2||e[0]!==e[1])throw new U("Identity matrix initializer can only be used for 2D square matrices.");return B(this.gain,Qf(e[0]))})}getConfig(){return{gain:this.gain}}};cw.className="Identity";se.registerClass(cw);function hV(e,t="channelsLast"){let n,r;if(Pt(t),e.length===2)n=e[0],r=e[1];else if([3,4,5].indexOf(e.length)!==-1){if(t==="channelsFirst"){let s=aa(e,2);n=e[1]*s,r=e[0]*s}else if(t==="channelsLast"){let s=aa(e,0,e.length-2);n=e[e.length-2]*s,r=e[e.length-1]*s}}else{let s=aa(e);n=Math.sqrt(s),r=Math.sqrt(s)}return[n,r]}var Wn=class extends Pr{constructor(e){if(super(),e.scale<0)throw new U(`scale must be a positive float. Got: ${e.scale}`);this.scale=e.scale==null?1:e.scale,this.mode=e.mode==null?"fanIn":e.mode,dV(this.mode),this.distribution=e.distribution==null?"normal":e.distribution,pV(this.distribution),this.seed=e.seed}apply(e,t){let n=hV(e),r=n[0],s=n[1],a=this.scale;if(this.mode==="fanIn"?a/=Math.max(1,r):this.mode==="fanOut"?a/=Math.max(1,s):a/=Math.max(1,(r+s)/2),this.distribution==="normal"){let o=Math.sqrt(a);if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Pe(`${this.getClassName()} does not support dType ${t}.`);return gm(e,0,o,t,this.seed)}else{let o=Math.sqrt(3*a);return Uu(e,-o,o,t)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};Wn.className="VarianceScaling";se.registerClass(Wn);var Dm=class extends Wn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Dm.className="GlorotUniform";se.registerClass(Dm);var Fm=class extends Wn{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Fm.className="GlorotNormal";se.registerClass(Fm);var Rm=class extends Wn{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Rm.className="HeNormal";se.registerClass(Rm);var Pm=class extends Wn{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Pm.className="HeUniform";se.registerClass(Pm);var Om=class extends Wn{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Om.className="LeCunNormal";se.registerClass(Om);var Mm=class extends Wn{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return Wn.className}};Mm.className="LeCunNormal";se.registerClass(Mm);var uw=class extends Pr{constructor(e){if(super(),this.DEFAULT_GAIN=1,this.gain=e.gain==null?this.DEFAULT_GAIN:e.gain,this.seed=e.seed,this.seed!=null)throw new Pe("Random seed is not implemented for Orthogonal Initializer yet.")}apply(e,t){return O(()=>{if(e.length<2)throw new Pe("Shape must be at least 2D.");e[0]*e[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);let n=e[0]>e[1]?[e[1],e[0]]:e,r=Am(n,0,1,"float32"),s=Zx.gramSchmidt(r);return e[0]>e[1]&&(s=Ee(s)),B(this.gain,s)})}getConfig(){return{gain:this.gain,seed:this.seed}}};uw.className="Orthogonal";se.registerClass(uw);var $k={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function Dk(e,t={}){return Kd(e,se.SerializationMap.getMap().classNameMap,t,"initializer")}function Nt(e){return ew(e)}function St(e){if(typeof e=="string"){let t=e in $k?$k[e]:e;if(t==="GlorotNormal")return new Fm;if(t==="GlorotUniform")return new Dm;if(t==="HeNormal")return new Rm;if(t==="HeUniform")return new Pm;if(t==="LeCunNormal")return new Om;if(t==="LeCunUniform")return new Mm;{let n={};return n.className=t,n.config={},Dk(n)}}else return e instanceof Pr?e:Dk(e)}function qy(e){return Array.isArray(e)&&Array.isArray(e[0])}function Wh(e){return e.length===0?[]:Array.isArray(e[0])?e:[e]}function Ce(e){let t;if(Array.isArray(e)){if(e.length!==1)throw new U(`Expected Tensor length to be 1; got ${e.length}`);t=e[0]}else t=e;return t}function tt(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(e.length===1)return e=e,e[0];throw new U(`Expected exactly 1 Shape; got ${e.length}`)}else return e}function Vh(e){let t=0;for(let n of e)n.shape.length===0?t+=1:t+=n.shape.reduce((r,s)=>r*s);return t}var Fk="Variable",yC=class{constructor(e,t="float32",n=Fk,r=!0,s=null){this.dtype=t==null?"float32":t,this.shape=e.shape,this.id=dC(),n=n==null?Fk:n,this.originalName=hC(n),this.name=fC(this.originalName),this.trainable_=r,this.constraint=s,this.val=Hx(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),fV(this.val,e),this.val.id!==e.id&&(this.val.assign(e),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function fV(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}function jy(e){return e.map(t=>t.read())}function lw(e){e.forEach(t=>{t[0].write(t[1])})}var zt=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,e.shape!=null?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},Vr=class{constructor(e,t,n,r,s,a,o){this.dtype=e,this.shape=t,this.sourceLayer=n,this.inputs=r,this.callArgs=s,this.outputTensorIndex=o,this.id=dC(),a!=null&&(this.originalName=hC(a),this.name=fC(this.originalName)),this.rank=t.length}},mV=0,Lm=class{constructor(e,t){this.callArgs=t,this.id=mV++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(let n of e.inboundLayers)n!=null&&n.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){let e=[];for(let t of this.inboundLayers)t!=null?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},gV=0,Ue=class extends se.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=gV++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){let n=this.getClassName();t=Ss(n)+"_"+Em(n)}if(this.name=t,this.trainable_=e.trainable==null?!0:e.trainable,e.inputShape!=null||e.batchInputShape!=null){let n;if(e.batchInputShape!=null)n=e.batchInputShape;else if(e.inputShape!=null){let s=null;e.batchSize!=null&&(s=e.batchSize),n=[s].concat(e.inputShape)}this.batchInputShape=n;let r=e.dtype;r==null&&(r=e.inputDType),r==null&&(r="float32"),this.dtype=r}e.weights!=null?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(this.inboundNodes.length===0)throw new Wr(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new U(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return Mn(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return Mn(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new ks(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(this.inboundNodes.length===0)throw new ks(`Layer ${this.name} is not connected, no input to return.`);return Mn(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new ks(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new ks(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return Mn(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){if(e=vt(e),this.inputSpec==null||this.inputSpec.length===0)return;let t=vt(this.inputSpec);if(e.length!==t.length)throw new U(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);for(let n=0;n<e.length;n++){let r=e[n],s=t[n];if(s==null)continue;let a=r.rank;if(s.ndim!=null&&a!==s.ndim)throw new U(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${s.ndim}, found ndim=${a}`);if(s.maxNDim!=null&&a>s.maxNDim)throw new U(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${s.maxNDim}, found ndim=${a}`);if(s.minNDim!=null&&a<s.minNDim)throw new U(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${s.minNDim}, found ndim=${a}.`);if(s.dtype!=null&&r.dtype!==s.dtype)throw new U(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${s.dtype}, found dtype=${r.dtype}.`);if(s.axes){let o=r.shape;for(let i in s.axes){let c=Number(i),u=s.axes[i],l=c>=0?o[c]:o[o.length+c];if(u!=null&&[u,null].indexOf(l)===-1)throw new U(`Input ${n} is incompatible with layer ${this.name}: expected axis ${c} of input shape to have value ${u} but got shape ${o}.`)}}if(s.shape!=null)for(let o=0;o<s.shape.length;++o){let i=s.shape[o],c=r.shape[o];if(i!=null&&c!=null&&i!==c)throw new U(`Input ${n} is incompatible with layer ${this.name}: expected shape=${s.shape}, found shape=${r.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){this._callHook!=null&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();let n=vt(e),r=!0;for(let a of n)if(!(a instanceof Vr)){r=!1;break}let s=!0;for(let a of n)if(a instanceof Vr){s=!1;break}if(r===s)throw new U("Arguments to apply() must be all SymbolicTensors or all Tensors");return Za(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);let a=[];for(let o of vt(e))a.push(o.shape);this.build(Mn(a)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&s&&(this._refCount=1)}if(this.assertInputCompatibility(e),s){let a=this.call(e,t),o=vt(a),i=[];for(let c of o)n.indexOf(c)!==-1&&(c=c.clone()),i.push(c);if(a=Mn(i),this.activityRegularizer!=null)throw new Pe("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return a}else{let a=bV(e),o=this.computeOutputShape(a),i,c=yV(e);if(this.warnOnIncompatibleInputShape(Array.isArray(e)?a[0]:a),o!=null&&o.length>0&&Array.isArray(o[0])?i=o.map((u,l)=>new Vr(c,u,this,vt(e),t,this.name,l)):i=new Vr(c,o,this,vt(e),t,this.name),this.addInboundNode(e,i,null,null,a,o,t),this._refCount++,this.activityRegularizer!=null)throw new Pe("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return i}})}warnOnIncompatibleInputShape(e){if(this.batchInputShape!=null)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((n,r)=>{n!=null&&e[r]!=null&&e[r]!==n&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new ks(`The layer ${this.name} has never been called and thus has no defined output shape.`);let e=[];for(let t of this.inboundNodes){let n=JSON.stringify(t.outputShapes);e.indexOf(n)===-1&&e.push(n)}if(e.length===1){let t=this.inboundNodes[0].outputShapes;return Array.isArray(t)&&Array.isArray(t[0])&&t.length===1?t[0]:t}else throw new ks(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new Wr(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Vh(this.weights)}build(e){this.built=!0}getWeights(e=!1){return jy(e?this.trainableWeights:this.weights)}setWeights(e){O(()=>{let t=this.weights;if(t.length!==e.length)throw new U(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(t.length===0)return;let n=[],r=jy(t);for(let s=0;s<r.length;++s){let a=r[s],o=t[s],i=e[s];if(!w.arraysEqual(a.shape,i.shape))throw new U(`Layer weight shape ${a.shape} not compatible with provided weight shape ${i.shape}`);n.push([o,i])}lw(n)})}addWeight(e,t,n,r,s,a,o,i){if(this._addedWeightNames.indexOf(e)!==-1)throw new U(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),n==null&&(n="float32"),this.fastWeightInitDuringBuild&&(r=i!=null?i():St("zeros"));let c=r.apply(t,n),u=new yC(c,n,e,a,o);return c.dispose(),s!=null&&this.addLoss(()=>s.apply(u.read())),a==null&&(a=!0),a?this._trainableWeights.push(u):this._nonTrainableWeights.push(u),u}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){e==null||Array.isArray(e)&&e.length===0||(e=vt(e),this._losses!==void 0&&this._losses!==null&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(t!=null)if(Array.isArray(t))t.forEach(n=>{if(n!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return t}addInboundNode(e,t,n,r,s,a,o=null){let i=vt(e);t=vt(t),n=vt(n),r=vt(r),s=Wh(s),a=Wh(a);let c=[],u=[],l=[];for(let p of i)c.push(p.sourceLayer),u.push(p.nodeIndex),l.push(p.tensorIndex);new Lm({outboundLayer:this,inboundLayers:c,nodeIndices:u,tensorIndices:l,inputTensors:i,outputTensors:t,inputMasks:n,outputMasks:r,inputShapes:s,outputShapes:a},o);for(let p=0;p<t.length;p++)t[p].sourceLayer=this,t[p].nodeIndex=this.inboundNodes.length-1,t[p].tensorIndex=p}getConfig(){let e={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(e.batchInputShape=this.batchInputShape),this.dtype!=null&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return--this._refCount===0&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};function bV(e){e=vt(e);let t=[];for(let n of e)t.push(n.shape);return Mn(t)}function yV(e){return"float32"}function vC(e,t,n){if((t==null||n!=null&&n>0)&&(t=e.sourceLayer,n=e.nodeIndex),t.inboundNodes.length===0)return[e];{let r=t.inboundNodes[n];if(r.inboundLayers.length===0)return r.inputTensors;{let s=[];for(let a=0;a<r.inboundLayers.length;a++){let o=r.inputTensors[a],i=r.inboundLayers[a],c=r.nodeIndices[a],u=vC(o,i,c);for(let l of u)s.indexOf(l)===-1&&s.push(l)}return s}}}var Hu=class extends Ue{constructor(e){if(super({dtype:e.dtype,name:e.name!=null?e.name:Em("input").toString()}),e.batchSize==null&&(e.batchSize=null),e.sparse==null&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,e.inputShape!=null&&e.batchInputShape!=null)throw new U("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(t==null){if(e.inputShape==null)throw new U("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(e.batchSize!=null)throw new U("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let n=e.dtype||"float32";this.batchInputShape=t,this.dtype=n,this.inputSpec=[{shape:t}];let r=new Vr(this.dtype,this.batchInputShape,this,[],{},this.name);r.nodeIndex=0,r.tensorIndex=0,new Lm({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[r],outputTensors:[r],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new U(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};Hu.className="InputLayer";se.registerClass(Hu);function xC(e){if(e.batchShape==null&&e.shape==null)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(e.batchShape!=null&&e.shape!=null)throw new U("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;e.shape!=null&&t==null&&(t=[null].concat(e.shape));let n=e.dtype;return n==null&&(n="float32"),new Hu({batchInputShape:t,name:e.name,dtype:n,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}function vV(e,t){if(e.dtype==null||e.dtype===t.dtype)return t;try{return ce(t,e.dtype)}catch(n){throw new U(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}var ja=class{constructor(e){if(this.id2Value={},this.id2Mask={},this.name2Id={},e instanceof ja)for(let t in e.id2Value)this.id2Value[t]=e.id2Value[t],t in e.id2Mask&&(this.id2Mask[t]=e.id2Mask[t]);else{if(e==null)return;for(let t of e)this.add(t.key,t.value)}}add(e,t,n){if(this.id2Value[e.id]==null)this.id2Value[e.id]=vV(e,t),this.name2Id[e.name]=e.id,n!=null&&(this.id2Mask[e.id]=n);else throw new U(`Duplicate key: name=${e.name}, id=${e.id}`);return this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return this.id2Value[e.id]!=null}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof Vr){if(this.id2Value[e.id]==null)throw new U(`Nonexistent key: ${e.name}`);return this.id2Value[e.id]}else{let t=this.name2Id[e];if(t==null)throw new U(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Value[t]}}getMask(e){if(e instanceof Vr){if(this.id2Value[e.id]==null)throw new U(`Nonexistent key: ${e.name}`);return this.id2Mask[e.id]}else{let t=this.name2Id[e];if(t==null)throw new U(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Mask[t]}}disposeMasks(){this.id2Mask!=null&&_e(this.id2Mask)}},Uh=new cC,Gh=new cC;function xV(e){Uh!=null&&Uh.setMaxEntries(e),Gh!=null&&Gh.setMaxEntries(e)}function Pl(e,t,n,r){let s=n==null?!1:n.training,a=Array.isArray(e),o=a?e:[e],i=o.map(f=>f.name),c=[],u=t.names();for(let f of i)u.indexOf(f)!==-1?c.push(t.getValue(f)):c.push(null);r!=null&&(r.maxNumTensors=-1/0,r.minNumTensors=1/0);let l=i.join(",")+"|"+t.names().sort().join(","),p=Uh.get(l),d;if(p==null){let f=wV(o,t);p=f.sorted,d=f.recipientCounts,Uh.put(l,p),Gh.put(l,d)}d={},s||Object.assign(d,Gh.get(l));let h=new ja(t);for(let f=0;f<p.length;++f){if(r!=null){let $=Lh().numTensors;$>r.maxNumTensors&&(r.maxNumTensors=$),$<r.minNumTensors&&(r.minNumTensors=$)}let m=p[f],g=m.sourceLayer;if(g instanceof Hu)continue;let b=[],y=[],v=[],x=!1;for(let $ of m.inputs){let F=h.getValue($),A=h.getMask($);b.push(F),y.push(A),A!=null&&(x=!0),s||(d[$.name]--,d[$.name]===0&&!t.hasKey($)&&i.indexOf($.name)===-1&&!F.isDisposed&&$.sourceLayer.stateful!==!0&&v.push(F))}x&&(n=n||{},n.mask=y[0]);let k=vt(g.apply(b,n)),S=null;g.supportsMasking&&(S=g.computeMask(b,y));let C=kV(m),E=Array.isArray(C)?C:[C];for(let $=0;$<E.length;++$){h.hasKey(E[$])||h.add(E[$],k[$],Array.isArray(S)?S[0]:S);let F=i.indexOf(E[$].name);F!==-1&&(c[F]=k[$])}s||_e(v)}return h.disposeMasks(),a?c:c[0]}function wV(e,t){w.assert(e!=null&&e.length>0,()=>"Expected at least one fetch, got none");let n=[],r={};if(e.length===1){let s=Rk(e[0],t);n=s.sorted,r=s.recipientMap}else{let s=new Set;for(let a of e){let{sorted:o,recipientMap:i}=Rk(a,t);for(let c of o)s.has(c.name)||(n.push(c),s.add(c.name));for(let c in i)r[c]==null&&(r[c]=new Set),i[c].forEach(u=>r[c].add(u))}}return{sorted:n,recipientCounts:IV(r)}}function IV(e){let t={};for(let n in e)t[n]=e[n].size;return t}function Rk(e,t){let n=new Set,r=[],s={};for(let i of t.names())n.add(i);let a=[],o=[];for(a.push(e);a.length>0;){let i=a[a.length-1];if(n.has(i.name)){a.pop();continue}let c=o[o.length-1]===a.length-1;if(i.inputs.length===0||c)a.pop(),r.push(i),n.add(i.name),c&&o.pop();else{o.push(a.length-1);for(let u of i.inputs)s[u.name]==null&&(s[u.name]=new Set),s[u.name].add(i.name),!n.has(u.name)&&a.push(u)}}return{sorted:r,recipientMap:s}}function kV(e){let t;if(e.sourceLayer.inboundNodes.length===1)t=e.sourceLayer.output;else{let n=null;for(let r=0;r<e.sourceLayer.inboundNodes.length;++r)for(let s of e.sourceLayer.inboundNodes[r].outputTensors)if(s.id===e.id){n=r;break}t=e.sourceLayer.getOutputAt(n)}return t}var SV=q();SV.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,xV);var wC={};Ae(wC,{maxNorm:()=>TV,minMaxNorm:()=>_V,nonNeg:()=>NV,unitNorm:()=>CV});function dw(e,t){return O(()=>un(ge(B(e,e),t,!0)))}var Jd=class extends se.Serializable{getConfig(){return{}}},pw=class extends Jd{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return O(()=>{let t=dw(e,this.axis),n=en(t,0,this.maxValue);return B(e,fe(n,Y(Gt(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};pw.className="MaxNorm";se.registerClass(pw);var hw=class extends Jd{constructor(e){super(),this.defaultAxis=0,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return O(()=>fe(e,Y(Gt(),dw(e,this.axis))))}getConfig(){return{axis:this.axis}}};hw.className="UnitNorm";se.registerClass(hw);var fw=class extends Jd{apply(e){return Xe(e)}};fw.className="NonNeg";se.registerClass(fw);var mw=class extends Jd{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=e.minValue!=null?e.minValue:this.defaultMinValue,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.rate=e.rate!=null?e.rate:this.defaultRate,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return O(()=>{let t=dw(e,this.axis),n=Y(B(this.rate,en(t,this.minValue,this.maxValue)),B(1-this.rate,t));return B(e,fe(n,Y(Gt(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};mw.className="MinMaxNorm";se.registerClass(mw);var Pk={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function jt(e){return ew(e)}function Ok(e,t={}){return Kd(e,se.SerializationMap.getMap().classNameMap,t,"constraint")}function Kt(e){if(e==null)return null;if(typeof e=="string"){let n={className:e in Pk?Pk[e]:e,config:{}};return Ok(n)}else return e instanceof Jd?e:Ok(e)}function TV(e){return new pw(e)}function CV(e){return new hw(e)}function NV(){return new fw}function _V(e){return new mw(e)}var IC={};Ae(IC,{constant:()=>$V,glorotNormal:()=>LV,glorotUniform:()=>MV,heNormal:()=>zV,heUniform:()=>BV,identity:()=>PV,leCunNormal:()=>WV,leCunUniform:()=>VV,ones:()=>AV,orthogonal:()=>UV,randomNormal:()=>FV,randomUniform:()=>DV,truncatedNormal:()=>RV,varianceScaling:()=>OV,zeros:()=>EV});function EV(){return new rw}function AV(){return new $m}function $V(e){return new sw(e)}function DV(e){return new aw(e)}function FV(e){return new ow(e)}function RV(e){return new iw(e)}function PV(e){return new cw(e)}function OV(e){return new Wn(e)}function MV(e){return new Dm(e)}function LV(e){return new Fm(e)}function zV(e){return new Rm(e)}function BV(e){return new Pm(e)}function WV(e){return new Om(e)}function VV(e){return new Mm(e)}function UV(e){return new uw(e)}var kC={};Ae(kC,{Layer:()=>Ue,RNN:()=>ms,RNNCell:()=>rp,activation:()=>gG,add:()=>TG,alphaDropout:()=>cH,average:()=>CG,averagePooling1d:()=>T0,averagePooling2d:()=>C0,averagePooling3d:()=>N0,avgPool1d:()=>PG,avgPool2d:()=>MG,avgPool3d:()=>zG,avgPooling1d:()=>OG,avgPooling2d:()=>LG,avgPooling3d:()=>BG,batchNormalization:()=>DG,bidirectional:()=>eH,categoryEncoding:()=>pH,concatenate:()=>NG,conv1d:()=>iG,conv2d:()=>cG,conv2dTranspose:()=>uG,conv3d:()=>lG,conv3dTranspose:()=>dG,convLstm2d:()=>YG,convLstm2dCell:()=>ZG,cropping2D:()=>hG,dense:()=>bG,depthwiseConv2d:()=>mG,dot:()=>$G,dropout:()=>yG,elu:()=>tG,embedding:()=>SG,flatten:()=>xG,gaussianDropout:()=>iH,gaussianNoise:()=>oH,globalAveragePooling1d:()=>WG,globalAveragePooling2d:()=>VG,globalMaxPool1d:()=>nH,globalMaxPool2d:()=>rH,globalMaxPooling1d:()=>fN,globalMaxPooling2d:()=>mN,gru:()=>GG,gruCell:()=>HG,input:()=>WC,inputLayer:()=>eG,layerNormalization:()=>FG,leakyReLU:()=>rG,lstm:()=>qG,lstmCell:()=>jG,masking:()=>uH,maxPool1d:()=>sH,maxPool2d:()=>aH,maxPooling1d:()=>gN,maxPooling2d:()=>bN,maxPooling3d:()=>UG,maximum:()=>_G,minimum:()=>EG,multiply:()=>AG,permute:()=>kG,prelu:()=>sG,reLU:()=>nG,repeatVector:()=>wG,rescaling:()=>lH,reshape:()=>IG,resizing:()=>dH,rnn:()=>JG,separableConv2d:()=>pG,simpleRNN:()=>KG,simpleRNNCell:()=>XG,softmax:()=>aG,spatialDropout1d:()=>vG,stackedRNNCells:()=>QG,thresholdedReLU:()=>oG,timeDistributed:()=>tH,upSampling2d:()=>fG,zeroPadding2d:()=>RG});async function Js(e){if(e==null)return;let t=[],n=[],r=[];for(let s in e){let a=e[s];if(typeof a!="number"){let o=a;t.push(o.data()),n.push(s),r.push(o)}}if(t.length>0){let s=await Promise.all(t);for(let a=0;a<s.length;++a)e[n[a]]=s[a][0];_e(r)}}function SC(e){if(e!=null)for(let t in e){let n=e[t];typeof n!="number"&&n.dispose()}}var Mk;(function(e){e[e.SILENT=0]="SILENT",e[e.VERBOSE=1]="VERBOSE"})(Mk||(Mk={}));var GV=125,Sc=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},TC=class{constructor(e,t=10){e==null&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(let t of this.callbacks)t.setParams(e)}setModel(e){for(let t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochBegin(e,t)}async onEpochEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onEpochEnd(e,t)}async onBatchBegin(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchBegin(e,t)}async onBatchEnd(e,t){t==null&&(t={});for(let n of this.callbacks)await n.onBatchEnd(e,t)}async onTrainBegin(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainEnd(e)}},HV=class extends Sc{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){t==null&&(t={});let n=t.size==null?0:t.size;this.seen+=n;for(let r in t){let s=t[r];if(typeof s=="number")this.totals.hasOwnProperty(r)||(this.totals[r]=0),this.totals[r]=this.totals[r]+s*n;else{let a;r in this.totals?a=this.totals[r]:this.totals[r]=0;let o=O(()=>Y(this.totals[r],B(s,n)));this.totals[r]=o,a!=null&&a.dispose()}}}async onEpochEnd(e,t){if(t!=null)for(let n of this.params.metrics)this.totals[n]!=null&&(typeof this.totals[n]=="number"?t[n]=this.totals[n]/this.seen:O(()=>{let r=B(fe(1,this.seen),this.totals[n]);t[n]=r,this.totals[n].dispose(),Jt(t[n])}))}},CC=class extends Sc{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){t==null&&(t={}),this.epoch.push(e);for(let n in t)this.history[n]==null&&(this.history[n]=[]),this.history[n].push(t[n])}async syncData(){let e=[],t=[],n=[];for(let s in this.history){let a=this.history[s];for(let o=0;o<a.length;++o)if(typeof a[o]!="number"){let i=a[o];e.push(i.data()),t.push(s),n.push(o)}}let r=await Promise.all(e);for(let s=0;s<r.length;++s)this.history[t[s]][n[s]].dispose(),this.history[t[s]][n[s]]=r[s][0]}},NC=class extends Sc{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||Jx,this.yieldEvery=t||"auto",this.yieldEvery==="auto"&&(this.yieldEvery=GV),this.yieldEvery==="never"&&e.onYield!=null)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");w.isNumber(this.yieldEvery)&&(this.maybeWait=q4(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,n){let r=[];this.yield!=null&&(await Js(n),r.push(this.yield(e,t,n))),r.push(this.nextFrameFunc()),await Promise.all(r)}async onEpochBegin(e,t){this.currentEpoch=e,this.epochBegin!=null&&(await Js(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){let n=[];this.epochEnd!=null&&(await Js(t),n.push(this.epochEnd(e,t))),this.yieldEvery==="epoch"&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(e,t){this.batchBegin!=null&&(await Js(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){let n=[];this.batchEnd!=null&&(await Js(t),n.push(this.batchEnd(e,t))),this.yieldEvery==="batch"?n.push(this.nextFrameFunc()):w.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(n)}async onTrainBegin(e){this.trainBegin!=null&&(await Js(e),await this.trainBegin(e))}async onTrainEnd(e){this.trainEnd!=null&&(await Js(e),await this.trainEnd(e))}};function _C(e,t){return e==null&&(e={}),e instanceof Sc?[e]:Array.isArray(e)&&e[0]instanceof Sc?e:vt(e).map(r=>new NC(r,t))}var _r=class{constructor(){}static registerCallbackConstructor(e,t){w.assert(e>=0&&Number.isInteger(e),()=>`Verbosity level is expected to be an integer >= 0, but got ${e}`),_r.checkForDuplicate(t),_r.constructors[e]==null&&(_r.constructors[e]=[]),_r.constructors[e].push(t)}static checkForDuplicate(e){for(let t in _r.constructors)_r.constructors[+t].forEach(r=>{if(r===e)throw new U("Duplicate callback constructor.")})}static clear(){_r.constructors={}}static createCallbacks(e){let t=[];for(let n in _r.constructors){let r=+n;e>=r&&t.push(..._r.constructors[r])}return t.map(n=>new n)}};_r.constructors={};function EC(e,t,n,r,s,a,o,i,c){let u=new CC,l=[new HV,..._r.createCallbacks(t)];e!=null&&l.push(...e),l.push(u);let p=new TC(l);return p.setParams({epochs:n,initialEpoch:r,samples:s,steps:a,batchSize:o,verbose:t,doValidation:i,metrics:c}),{callbackList:p,history:u}}function Gr(e,t={},n=!1){return Kd(e,se.SerializationMap.getMap().classNameMap,t,"layer",n)}function Hh(e,t){return O(()=>{e.dtype!=="float32"&&(e=ce(e,"float32"));let n=ge(Yd(e),t,!0),r=bn(n.shape,Gt()),s=un(hs(n,r));return fe(e,s)})}function xi(e,t){return O(()=>Ct(Yd(de(t,e)),-1))}function zm(e,t){return O(()=>Ct(Lt(de(t,e)),-1))}function qu(e,t){return O(()=>{let n=de(e,t),r=en(Lt(e),Gt(),Number.MAX_VALUE),s=Lt(fe(n,r));return B(100,Ct(s,-1))})}function qV(e,t){return O(()=>{let n=en(t,Gt(),Number.MAX_VALUE),r=nr(Y(1,n)),s=en(e,Gt(),Number.MAX_VALUE),a=nr(Y(1,s));return Ct(Yd(de(r,a)),-1)})}function jV(e,t){return O(()=>{let n=hs(0,de(1,B(e,t)));return Ct(Yd(n),-1)})}function KV(e,t){return O(()=>{let n=hs(0,de(1,B(e,t)));return Ct(n,-1)})}function XV(e,t){return O(()=>{let n=ge(B(e,t),-1),r=pr(B(de(1,e),t),-1);return hs(0,Y(1,de(r,n)))})}function YV(e,t){return O(()=>{let n=Math.log(2),r=de(t,e),s=de(Y(r,mi(B(-2,r))),n);return Ct(s,-1)})}function td(e,t,n=!1){return O(()=>{if(n)t=Xr(t);else{let r=ge(t,t.shape.length-1,!0);t=fe(t,r)}return t=en(t,Gt(),1-Gt()),xt(ge(B(ce(e,"float32"),nr(t)),t.shape.length-1))})}function qh(e,t,n=!1){return O(()=>{let r=ce(Bu(sV(e)),"int32");t=en(t,Gt(),1-Gt());let s=t.shape,a=W(gc(r,s[s.length-1]),s);return td(a,t,n)})}function ZV(e,t){if(!w.arraysEqual(e.shape,t.shape))throw new U(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return O(()=>{let n=Xe(t),r=xt(Lt(t));return Y(de(n,B(t,e)),zd(gn(r)))})}function Bm(e,t){return O(()=>{let n;return n=en(t,Gt(),1-Gt()),n=nr(fe(n,de(1,n))),Ct(ZV(e,n),-1)})}function JV(e,t){return O(()=>{let n=en(e,Gt(),1),r=en(t,Gt(),1);return ge(B(e,nr(fe(n,r))),-1)})}function QV(e,t){return O(()=>{let n=nr(Y(Gt(),t));return Ct(de(t,B(e,n)),-1)})}function gw(e,t){return O(()=>{let n=Hh(e,-1),r=Hh(t,-1),s=B(n,r);return xt(ge(s,-1))})}var jh={meanSquaredError:xi,meanAbsoluteError:zm,meanAbsolutePercentageError:qu,meanSquaredLogarithmicError:qV,squaredHinge:jV,hinge:KV,categoricalHinge:XV,logcosh:YV,categoricalCrossentropy:td,sparseCategoricalCrossentropy:qh,binaryCrossentropy:Bm,kullbackLeiblerDivergence:JV,poisson:QV,cosineProximity:gw};function wy(e){if(typeof e=="string"){if(e in jh)return jh[e];let t=`Unknown loss ${e}`;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new U(t)}else return e}function bw(e,t){return O(()=>{let n=B(.5,rr(t)),r=vi(An(t,n),e.dtype);return Ct(tr(e,r),-1)})}function yw(e,t){return O(()=>vi(tr(ao(e,-1),ao(t,-1)),"float32"))}function AC(e,t){return O(()=>ce(ge(Dr(tr(e,1),tr(t,1))),"float32"))}function eU(e,t){return O(()=>ce(ge(Dr(tr(e,1),tr(t,0))),"float32"))}function tU(e,t){return O(()=>ce(ge(Dr(tr(e,0),tr(t,1))),"float32"))}function $C(e,t){return O(()=>{let n=AC(e,t),r=tU(e,t),s=Y(n,r);return ce(mn(An(s,0),fe(n,s),0),"float32")})}function nU(e,t){return O(()=>{let n=AC(e,t),r=eU(e,t),s=Y(n,r);return ce(mn(An(s,0),fe(n,s),0),"float32")})}function DC(e,t){return Bm(e,t)}function FC(e,t){return e.rank===t.rank&&(e=Na(e,[e.rank-1])),t=ao(t,-1),t.dtype!==e.dtype&&(t=ce(t,e.dtype)),ce(tr(e,t),"float32")}var rU=xi,sU=xi,aU=zm,oU=zm,iU=qu,cU=qu,vw=td,uU=gw,RC=qh,Kh={binaryAccuracy:bw,categoricalAccuracy:yw,precision:$C,categoricalCrossentropy:vw,sparseCategoricalCrossentropy:RC,mse:rU,MSE:sU,mae:aU,MAE:oU,mape:iU,MAPE:cU,cosine:uU};function lU(e){if(typeof e=="string"&&e in Kh)return Kh[e];if(typeof e!="string"&&e!=null)return e;throw new U(`Unknown metric ${e}`)}function lh(e){if(ss(e!==null,`Unknown LossOrMetricFn ${e}`),typeof e=="string")return e;{let t;for(let n of Object.keys(jh))if(jh[n]===e){t=n;break}if(t!==void 0)return t;for(let n of Object.keys(Kh))if(Kh[n]===e){t=n;break}return t!==void 0?t:e.name}}function dU(e){let t={Adagrad:()=>Wa.adagrad(.01),Adadelta:()=>Wa.adadelta(1,.95,Gt()),Adam:()=>Wa.adam(.001,.9,.999,Gt()),Adamax:()=>Wa.adamax(.002,.9,.999,Gt(),0),RMSProp:()=>Wa.rmsprop(.001,.9,0,Gt()),SGD:()=>Wa.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new U(`Unknown Optimizer ${e}`)}var Lk=1*1024*1024;function zk(e,t,n=!1){if(e==null||typeof e!="object"||Object.getPrototypeOf(e)!==Object.prototype||!Ky(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(n){let r=JSON.stringify(e);r.length>Lk&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${r.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${Lk}.`)}}function Ky(e){if(e===null)return!0;if(typeof e=="object")if(Object.getPrototypeOf(e)===Object.prototype){let t=Object.keys(e);for(let n of t)if(typeof n!="string"||!Ky(e[n]))return!1;return!0}else if(Array.isArray(e)){for(let t of e)if(!Ky(t))return!1;return!0}else return!1;else{let t=typeof e;return t==="string"||t==="number"||t==="boolean"}}function pU(e,t,n,r=console.log){let s=fU(e),a=["Layer (type)","Input Shape","Output shape","Param #"];s?(t=t||90,n=n||[.32,.61,.89,1]):(t=t||115,n=n||[.24,.48,.7,.8,1]),n[n.length-1]<=1&&(n=n.map(l=>Math.floor(t*l)));let o;if(!s){a.push("Receives inputs"),o=[];for(let l in e.nodesByDepth)o.push(...e.nodesByDepth[l])}r("_".repeat(t)),Xh(a,n,r),r("=".repeat(t));let i=e.layers;for(let l=0;l<i.length;++l)s?mU(i[l],n,r):gU(i[l],n,o,r),r((l===i.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();let c=hU(e),u=Vh(e.nonTrainableWeights);r(`Total params: ${c+u}`),r(`Trainable params: ${c}`),r(`Non-trainable params: ${u}`),r("_".repeat(t))}function hU(e){let t;return e.collectedTrainableWeights!=null?t=Vh(e.collectedTrainableWeights):t=Vh(e.trainableWeights),t}function fU(e){let t=!0,n=[],r=[];for(let s in e.nodesByDepth)n.push(e.nodesByDepth[s]);for(let s of n){if(s.length>1||s.length===1&&s[0].inboundLayers.length>1){t=!1;break}r.push(...s)}if(t)for(let s of e.layers){let a=!1;for(let o of s.inboundNodes)if(r.indexOf(o)!==-1)if(a){t=!1;break}else a=!0;if(!t)break}return t}function Xh(e,t,n=console.log){let r="";for(let s=0;s<e.length;++s)s>0&&(r=r.slice(0,r.length-1)+" "),r+=e[s],r=r.slice(0,t[s]),r+=" ".repeat(t[s]-r.length);n(r)}function mU(e,t,n){let r,s;try{s=e.inboundNodes.map(c=>JSON.stringify(c.inputShapes)).join(",")}catch(c){s="multiple"}try{r=JSON.stringify(e.outputShape)}catch(c){r="multiple"}let a=e.name,o=e.getClassName(),i=[`${a} (${o})`,s,r,e.countParams().toString()];Xh(i,t,n)}function gU(e,t,n,r){let s,a;try{a=e.inboundNodes.map(p=>JSON.stringify(p.inputShapes)).join(",")}catch(p){a="multiple"}try{s=JSON.stringify(e.outputShape)}catch(p){s="multiple"}let o=[];for(let p of e.inboundNodes)if(!(n!=null&&n.length>0&&n.indexOf(p)===-1))for(let d=0;d<p.inboundLayers.length;++d){let h=p.inboundLayers[d].name,f=p.nodeIndices[d],m=p.tensorIndices[d];o.push(`${h}[${f}][${m}]`)}let i=e.name,c=e.getClassName(),u=o.length===0?"":o[0],l=[`${i} (${c})`,a,s,e.countParams().toString(),u];Xh(l,t,r);for(let p=1;p<o.length;++p)Xh(["","","","",o[p]],t,r)}function PC(e,t,n){return(e==="inboundNodes"||e==="outputLayers"||e==="inputLayers")&&t===0&&typeof n=="string"}function nd(e,t){if(e===null)return null;if(typeof e=="string")return Ga(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],r=e.length;for(let s=0;s<r;++s){let a=e[s];PC(t,s,a)?n.push(a):n.push(nd(a,t))}return n}else{let n={};for(let r of Object.keys(e)){let s=e[r];if(r==="name"&&typeof s=="string")n[r]=s;else{let a=Ga(r);n[a]=nd(s,a)}}return n}}function Xy(e,t){if(e==null)return null;if(typeof e=="string")return Ss(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let n=[],r=e.length;for(let s=0;s<r;++s){let a=e[s];PC(t,s,a)?n.push(a):n.push(Xy(a,t))}return n}else{let n={};for(let r of Object.keys(e)){let s=e[r],a=Ss(r);(r==="name"||r==="className")&&typeof s=="string"?n[a]=s:n[a]=Xy(s,r)}return n}}var xw="4.0.0",ns=class extends Ue{constructor(e){if(super({}),this.containerNodes=new Set,this.name=e.name,this.name==null){let b=this.getClassName().toLowerCase();this.name=Em(b)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(e.inputs)?this.inputs=e.inputs.slice():this.inputs=[e.inputs],Array.isArray(e.outputs)?this.outputs=e.outputs.slice():this.outputs=[e.outputs],sa(this.inputs).length!==this.inputs.length)throw new U(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(b=>b.name)}`);sa(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(b=>b.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let b of this.outputs){let y=b.sourceLayer,v=b.nodeIndex,x=b.tensorIndex;this.outputLayers.push(y),this.outputLayersNodeIndices.push(v),this.outputLayersTensorIndices.push(x)}for(let b of this.inputs){let y=b.sourceLayer,v=b.nodeIndex,x=b.tensorIndex;ss(v===0,"input layer has >1 nodes"),ss(x===0,"input layer has >1 tensors"),this.inputLayers.push(y),this.inputLayersNodeIndices.push(v),this.inputLayersTensorIndices.push(x)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let b=0;b<this.inputLayers.length;b++){let y=this.inputLayers[b];if(!(y instanceof Hu))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${b} (0-based) originates from layer type ${y.getClassName()}.`);this.inputNames.push(y.name),this.feedInputShapes.push(y.batchInputShape),this.feedInputNames.push(y.name)}for(let b of this.outputLayers)this.outputNames.push(b.name);this.internalInputShapes=this.inputs.map(b=>b.shape),this.internalOutputShapes=this.outputs.map(b=>b.shape);let t={},n={},r={},s={},a={},o=[],i=(b,y,v,x,k,S)=>{(x==null||k==null||S==null)&&(x=b.sourceLayer,k=b.nodeIndex,S=b.tensorIndex);let C=x.inboundNodes[k];if(v.indexOf(C)!==-1)throw new Wr(`The tensor ${b.name} at layer "${x.name}" is part of a cycle.`);if(y.indexOf(C)!==-1)return;this.containerNodes.add(ns.nodeKey(x,k)),x.id in a||(a[x.id]=Object.keys(a).length),v.indexOf(C)===-1&&v.push(C);let E=C.inboundLayers.length;for(let $=0;$<E;$++){let F=C.inputTensors[$],A=C.inboundLayers[$],R=C.nodeIndices[$],T=C.tensorIndices[$];i(F,y,v,A,R,T)}for(y.push(C);v.indexOf(C)>=0;)v.splice(v.indexOf(C),1);o.push(C)},c=[],u=[];for(let b of this.outputs)i(b,c,u);let l=o.slice().reverse();for(let b of l){n[b.id]=b,b.id in t||(t[b.id]=0);let y=t[b.id],v=r[b.outboundLayer.id]==null?0:r[b.outboundLayer.id];y=Math.max(y,v),r[b.outboundLayer.id]=y,s[b.outboundLayer.id]=b.outboundLayer,t[b.id]=y;for(let x=0;x<b.inboundLayers.length;x++){let k=b.inboundLayers[x],S=b.nodeIndices[x],C=k.inboundNodes[S],E=t[C.id]==null?0:t[C.id];t[C.id]=Math.max(y+1,E),n[C.id]=C}}let p={};for(let b in t){let y=t[b];y in p||(p[y]=[]),p[y].push(n[b])}let d={};for(let b in r){let y=r[b];y in d||(d[y]=[]),d[y].push(s[b])}let h=Object.keys(d).map(b=>parseInt(b,10)).sort(ih);this.layers=[];for(let b of h){let y=d[b];y.sort((v,x)=>{let k=a[v.id],S=a[x.id];return k<S?-1:k>S?1:0});for(let v of y)v instanceof ns&&this.internalContainerRefs.push(v),this.layers.push(v)}this.layersByDepth=d,h=Object.keys(p).map(b=>parseInt(b,10)).sort(ih);let f=this.inputs.slice(),m=[];for(let b of h)for(let y of p[b]){let v=y.outboundLayer;if(v!=null){for(let x of y.inputTensors)if(f.indexOf(x)===-1)throw new Wr(`Graph disconnected: cannot obtain value for tensor ${x} at layer "${v.name}". The following previous layers were accessed without issue: ${m}`);for(let x of y.outputTensors)f.push(x);m.push(v.name)}}this.nodesByDepth=p;let g=this.layers.map(b=>b.name);for(let b of g){let y=g.filter(v=>v===b).length;if(y!==1)throw new Wr(`The name "${b}" is used ${y} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(g))}this.outboundNodes=[],this.inboundNodes=[],new Lm({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(b=>null),outputMasks:this.outputs.map(b=>null),inputShapes:this.inputs.map(b=>b.shape),outputShapes:this.outputs.map(b=>b.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let e={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(let t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(n=>n.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new U("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(let t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.layers)t.push(...n.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){let n={},r=0;for(let a of this.layers)for(let o of a.weights){if(n[o.originalName]!=null)throw new U(`Duplicate weight name: ${o.originalName}`);n[o.originalName]=o,r++}let s=[];for(let a in e){let o=a;if(n[a]==null){let i=a.split("/");o=i.slice(0,-2).concat([i[i.length-1]]).join("/")}if(n[o]!=null)s.push([n[o],e[a]]);else if(t)throw new U(`Provided weight data has no target variable: ${a}`);delete n[o]}if(t){let a=[];for(let o in n)a.push(o);if(a.length>0)throw new U(`${a.length} of ${r} weights are not set: ${a}`)}lw(s)}updatedConfig(){let e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion=`tfjs-layers ${xw}`,t.backend="TensorFlow.js",t}toJSON(e,t=!0){let n=Xy(this.updatedConfig());return t?JSON.stringify(n):n}call(e,t){return O(()=>{e=vt(e);let n=new ja;for(let r=0;r<this.inputs.length;++r)n.add(this.inputs[r],e[r]);return Pl(this.outputs,n,t)})}computeMask(e,t){return O(()=>{e=vt(e);let n;return t==null?n=uo(null,e.length):n=vt(t),this.runInternalGraph(e,n)[1]})}computeOutputShape(e){let t=Wh(e);if(t.length!==this.inputLayers.length)throw new U(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let o=0;o<t.length;o++){let i=this.inputLayers[o],c=t[o],u=i.name+"_0_0";n[u]=c}let r=Object.keys(this.nodesByDepth).map(o=>parseInt(o,10)).sort(ih);if(r.length>1)for(let o of r){let i=this.nodesByDepth[o];for(let c of i){let u=c.outboundLayer;if(this.inputLayers.map(f=>f.id).indexOf(u.id)!==-1)continue;let l=[];for(let f=0;f<c.inboundLayers.length;f++){let m=c.inboundLayers[f],g=c.nodeIndices[f],b=c.tensorIndices[f],y=`${m.name}_${g}_${b}`,v=n[y];l.push(v)}let p=u.computeOutputShape(Mn(l)),d=Wh(p),h=u.inboundNodes.indexOf(c);for(let f=0;f<d.length;f++){let m=`${u.name}_${h}_${f}`;n[m]=d[f]}}}let s=[],a=[];for(let o=0;o<this.outputLayers.length;o++){let i=this.outputLayers[o],c=this.outputLayersNodeIndices[o],u=this.outputLayersTensorIndices[o],l=`${i.name}_${c}_${u}`;a.push(l)}for(let o=0;o<a.length;o++){let i=a[o];ss(i in n),s.push(n[i])}return Mn(s)}runInternalGraph(e,t){t==null&&(t=uo(null,e.length));let n={};for(let i=0;i<this.inputs.length;++i){let c=this.inputs[i],u=e[i],l=t[i];n[c.id]=[u,l]}let r=Object.keys(this.nodesByDepth).map(i=>parseInt(i,10)).sort(ih);for(let i of r){let c=this.nodesByDepth[i];for(let u of c){let l=u.outboundLayer,p=u.inputTensors,d=u.outputTensors,h=new Array;for(let f of p)f.id in n&&h.push(n[f.id]);if(h.length===p.length){let f={},m,g,b,y;if(u.callArgs!=null&&(f=u.callArgs),h.length===1){let[v,x]=h[0];f.mask==null&&(f.mask=x),b=vt(l.call(v,f)),y=vt(l.computeMask(v,x)),m=[v],g=[x]}else m=h.map(v=>v[0]),g=h.map(v=>v[1]),f.mask==null&&(f.mask=g),b=vt(l.call(m,f)),y=vt(l.computeMask(m,g));if(l.activityRegularizer)throw new Pe("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let v=0;v<d.length;++v){let x=d[v],k=b[v],S=y[v];n[x.id]=[k,S]}}}}let s=[],a=[],o=[];for(let i of this.outputs){ss(i.id in n,`Could not compute output ${i.name} : ${i.id}`);let[c,u]=n[i.id];o.push(c.shape),s.push(c),a.push(u)}return[s,a,o]}buildNodeConversionMap(e){let t={},n;for(let r of this.layers){n=r instanceof ns?1:0;for(let s=0;s<r.inboundNodes.length;s++){let a=ns.nodeKey(r,s);this.containerNodes.has(a)&&(t[a]=n,n+=1)}}return t}getLayer(e,t){if(t!=null){if(this.layers.length<=t)throw new U(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);return this.layers[t]}else if(e==null)throw new U("Provide either a layer name or layer index");for(let n of this.layers)if(n.name===e)return n;throw new U(`No such layer: ${e}`)}calculateLosses(){return O(()=>{let e=[];for(let t of this.layers)for(let n=0;n<t.inboundNodes.length;++n){let r=ns.nodeKey(t,n);this.containerNodes.has(r)&&e.push(...t.calculateLosses())}return e})}getConfig(){let e={name:this.name},t=this.buildNodeConversionMap(this.layers),n=[];for(let a of this.layers){let o=a.getClassName(),i=a.getConfig(),c=[];for(let l=0;l<a.inboundNodes.length;l++){let p=a.inboundNodes[l],d=ns.nodeKey(a,l),h={};if(this.containerNodes.has(d)){if(p.callArgs)try{JSON.stringify(p.callArgs),h=p.callArgs}catch(f){console.warn(`Layer ${a.name} was passed non-serializable keyword arguments: ${p.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),h={}}if(p.inboundLayers.length>0){let f=[];for(let m=0;m<p.inboundLayers.length;m++){let g=p.inboundLayers[m],b=p.nodeIndices[m],y=p.tensorIndices[m],v=ns.nodeKey(g,b),x=t[v];x==null&&(x=0),f.push([g.name,x,y,h])}c.push(f)}}}let u={};u.name=a.name,u.className=o,u.config=i,u.inboundNodes=c,n.push(u)}e.layers=n;let r=[];for(let a=0;a<this.inputLayers.length;a++){let o=this.inputLayers[a],i=this.inputLayersNodeIndices[a],c=ns.nodeKey(o,i);if(!this.containerNodes.has(c))continue;let u=t[c];u==null&&(u=0);let l=this.inputLayersTensorIndices[a];r.push([o.name,u,l])}e.inputLayers=r;let s=[];for(let a=0;a<this.outputLayers.length;a++){let o=this.outputLayers[a],i=this.outputLayersNodeIndices[a],c=ns.nodeKey(o,i);if(!this.containerNodes.has(c))continue;let u=t[c];u==null&&(u=0);let l=this.outputLayersTensorIndices[a];s.push([o.name,u,l])}return e.outputLayers=s,e}static fromConfig(e,t,n={},r=!1){let s={},a={};function o(m,g){m.name in a?a[m.name].push(g):a[m.name]=[g]}function i(m,g){let b=[],y;for(let v of g){let x=v[0],k=v[1],S=v[2];if(y=v[3]==null?{}:v[3],!(x in s)){o(m,g);return}let C=s[x];if(C.inboundNodes.length<=k){o(m,g);return}let E=C.inboundNodes[k];b.push(E.outputTensors[S])}b.length>0&&m.apply(Mn(b),y)}function c(m){let g=m.name,b=Gr(m,t.customObjects!=null?t.customObjects:{});b.setFastWeightInitDuringBuild(r),s[g]=b,m.inboundNodes.forEach(v=>{if(!(v instanceof Array))throw new U(`Corrupted configuration, expected array for nodeData: ${v}`);o(b,v)})}let u=t.name,l=t.layers;for(let m of l)c(m);for(;!H4(a);)for(let m of l){let g=s[m.name];if(g.name in a){let b=a[g.name];delete a[g.name];for(let y of b)i(g,y)}}let p=[],d=[],h=t.inputLayers;for(let m of h){let g=m[0],b=m[1],y=m[2];ss(g in s);let x=s[g].inboundNodes[b].outputTensors;p.push(x[y])}let f=t.outputLayers;for(let m of f){let g=m[0],b=m[1],y=m[2];ss(g in s);let x=s[g].inboundNodes[b].outputTensors;d.push(x[y])}return new e({inputs:p,outputs:d,name:u})}get stateful(){if(this._stateful)throw new U("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let e of this.layers)if(e.stateful)return!0;return!1}resetStates(){O(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function bU(e,t,n){let r=t.length;if(e==null||Array.isArray(e)&&e.length===0)return t.map(s=>null);if(r===1)return Array.isArray(e)&&e.length===1?e:typeof e=="object"&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==r)throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${r} outputs. Make sure a set of weights is provided for each model output.`);return e}else if(typeof e=="object"&&Object.keys(e).length>0&&typeof e[Object.keys(e)[0]]=="object"){let s=[];return t.forEach(a=>{a in e?s.push(e[a]):s.push(null)}),s}else throw new Error(`The model has multiple (${r}) outputs, so ${n} must be either an array with ${r} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)}function OC(e,t){return bU(e,t,"classWeight")}async function MC(e,t,n,r){if(t!=null||r!=null)throw new Error("Support sampleWeight is not implemented yet");if(n!=null){let s=O(()=>{if(e.shape.length===1)return is(e);if(e.shape.length===2){if(e.shape[1]>1)return ao(e,1);if(e.shape[1]===1)return W(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),a=Array.from(await s.data());_e(s);let o=[];return a.forEach(i=>{if(n[i]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${i} exists in the data but not in classWeight`);o.push(n[i])}),Ke(o,"float32")}else return null}function yU(e,t){return B(e,t)}var vU=32;function LC(e,t){let n,r,s=t;n=s.xs,r=s.ys,w.assert(n!=null&&r!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let a=Bk("input",e.inputNames,n),o=Bk("output",e.outputNames,r),i=a[0].shape[0];w.assert(a.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${a.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`),w.assert(o.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${o.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let c=0;c<a.length;c++)w.assert(a[c].shape[0]===i,()=>`Batch size mismatch: input ${e.inputNames[c]} has ${a[c].shape[0]}; expected  ${i} based on input ${e.inputNames[0]}.`);for(let c=0;c<o.length;c++)w.assert(o[c].shape[0]===i,()=>`Batch size mismatch: output ${e.outputNames[c]} has ${o[c].shape[0]}; expected  ${i} based on input ${e.inputNames[0]}.`);return{xs:a,ys:o}}function Bk(e,t,n){if(n instanceof Te)return[n];if(Array.isArray(n))return w.assert(n.length===t.length,()=>`Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),n;{let r=[];for(let s of t){if(n[s]==null)throw new U(`The feature data generated by the dataset lacks the required ${e} key '${s}'.`);r.push(n[s])}return r}}function xU(e){if(e.length===3)throw new Pe("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}async function wU(e,t,n){let r=n.batchesPerEpoch!=null;if(w.assert(e.optimizer!=null,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),w.assert(n!=null,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),w.assert(n.epochs!=null&&n.epochs>0&&Number.isInteger(n.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${n.epochs}`),w.assert(!r||n.batchesPerEpoch>0&&Number.isInteger(n.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${n.batchesPerEpoch}`),w.assert(n.validationSplit==null,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{let s=n.validationData!=null,a,o;if(s)if(Wk(n.validationData))w.assert(n.validationBatches==null||n.validationBatches>0&&Number.isInteger(n.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${n.validationBatches}`);else{let g=xU(n.validationData);a=g.xs,o=g.ys}let i=e.makeTrainFunction(),c=e.getDedupedMetricsNames(),u;s?u=c.slice().concat(c.map(g=>"val_"+g)):u=c.slice();let l=_C(n.callbacks,n.yieldEvery),p=n.verbose==null?1:n.verbose,{callbackList:d,history:h}=EC(l,p,n.epochs,null,null,IU(t,n),null,s,u);d.setModel(e),e.history=h,await d.onTrainBegin(),e.stopTraining_=!1;let f=n.initialEpoch==null?0:n.initialEpoch,m=await t.iterator();for(;f<n.epochs;){let g={};await d.onEpochBegin(f);let b=0,y=0;for(r||(m=await t.iterator());!r||b<n.batchesPerEpoch;){let v=await m.next();if(r&&v.done){console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${b} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(v.value!=null){let{xs:x,ys:k}=LC(e,v.value),S={};S.batch=y,S.size=x[0].shape[0],await d.onBatchBegin(y,S);let C=[];if(n.classWeight!=null){let F=OC(n.classWeight,e.outputNames);for(let A=0;A<F.length;++A)C.push(await MC(k[A],null,F[A]))}let E=x.concat(k).concat(C),$=i(E);_e(E);for(let F=0;F<c.length;++F){let A=c[F],R=$[F];S[A]=R,Jt(R)}await d.onBatchEnd(y,S),SC(S),y++,b++}if(r?b>=n.batchesPerEpoch:v.done){if(s){let x;Wk(n.validationData)?x=vt(await e.evaluateDataset(n.validationData,{batches:n.validationBatches})):x=vt(e.evaluate(a,o,{batchSize:n.validationBatchSize==null?vU:n.validationBatchSize,verbose:0}));for(let k=0;k<e.metricsNames.length;++k)g[`val_${e.metricsNames[k]}`]=x[k]}break}if(e.stopTraining_)break}if(await d.onEpochEnd(f,g),f++,e.stopTraining_)break}return await d.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function IU(e,t){let n=null;return t.batchesPerEpoch!=null?n=t.batchesPerEpoch:Number.isFinite(e.size)&&(n=e.size),n}function Wk(e){return typeof e.iterator=="function"}function kU(e){return typeof e.next=="function"}async function SU(e,t,n){n=n||{};let r=n.batches!=null,s=e.testFunction,a=[];if(n.verbose>0)throw new Pe("Verbose mode is not implemented yet.");w.assert(!r||n.batches>0&&Number.isInteger(n.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(n.batches)}`);let o=kU(t)?t:await t.iterator(),i=0,c=0;for(;!r||c<n.batches;){let u=await o.next();if(a=O(()=>{if(u.value){let{xs:l,ys:p}=LC(e,u.value),d=l.concat(p),h=O(()=>s(d));if(_e(d),c===0)for(let m=0;m<h.length;++m)a.push(xe(0));let f=d[0].shape[0];for(let m=0;m<h.length;++m){let g=h[m],b=a[m];a[m]=O(()=>Y(a[m],B(f,g))),c>0&&_e(b)}_e(h),i+=f,++c}return a}),u.done){r&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let u=0;u<a.length;++u){let l=a[u];a[u]=fe(a[u],i),_e(l)}return Mn(a)}function Yy(e){w.assert(e>0&&Number.isInteger(e),()=>`batchSize is required to be a positive integer, but got ${e}`)}function Ol(e,t,n){return e==null?[null]:Array.isArray(e)?e.map(r=>Ja(r,t,n-t)):Ja(e,t,n-t)}function ww(e,t){return O(()=>e==null?null:Array.isArray(e)?e.map(n=>ww(n,t)):gC(e,t.dtype==="int32"?t:ce(t,"int32")))}function Zy(e,t){let n=[],r=0,s=null;for(;r<e;)s=r+t,s>=e&&(s=e),n.push([r,s]),r=s;return n}async function TU(e,t,n,r,s,a,o,i,c,u,l,p,d,h,f){s==null&&(s=32),a==null&&(a=1),l==null&&(l=!0),d==null&&(d=0);let m=!1;if(c!=null&&u!=null&&(m=!0),f!=null&&(m=!0,h==null))throw new U("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let g=e.checkNumSamples(n,s,h,"steps_per_epoch"),b;g!=null&&(b=qr(0,g)),o==null&&(o=1);let{callbackList:y,history:v}=EC(i,o,a,d,g,h,s,m,p);y.setModel(e),e.history=v,await y.onTrainBegin(),e.stopTraining_=!1;for(let x=d;x<a;++x){await y.onEpochBegin(x);let k={};if(h!=null)throw new Pe("stepsPerEpoch mode is not implemented yet.");{if(l==="batch")throw new Pe("batch shuffling is not implemneted yet");l&&w.shuffle(b);let S=Ke(b),C=Zy(g,s);for(let E=0;E<C.length;++E){let $={};if(await y.onBatchBegin(E,$),O(()=>{let F=C[E][0],A=C[E][1],R=Ja(S,F,A-F);$.batch=E,$.size=A-F;let T=ww(n,R),L=t(T);for(let V=0;V<r.length;++V){let G=r[V],j=L[V];$[G]=j,Jt(j)}if(E===C.length-1&&m){let V=e.testLoop(c,u,s);for(let G=0;G<r.length;++G){let j=r[G],H=V[G];Jt(H),k["val_"+j]=H}}}),await y.onBatchEnd(E,$),SC($),e.stopTraining_)break}S.dispose()}if(await y.onEpochEnd(x,k),e.stopTraining_)break}return await y.onTrainEnd(),await e.history.syncData(),e.history}async function CU(e,t,n,r={}){if(e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;let s,a,o,i,c,u,l,p,d;try{let h=r.batchSize==null?32:r.batchSize;Yy(h);let f=!1,m=await e.standardizeUserData(t,n,r.sampleWeight,r.classWeight,f,h);s=m[0],a=m[1],d=m[2];let g=!1,b;if(r.validationData!=null&&r.validationData.length>0){if(g=!0,r.validationData.length===2)c=r.validationData[0],u=r.validationData[1];else throw r.validationData.length===3?new Pe("validationData including sample weights is not supported yet."):new U(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${r.validationData} is invalid.`);let $=!0,F=await e.standardizeUserData(c,u,null,null,$,h);l=F[0],p=F[1],b=l.concat(p)}else if(r.validationSplit!=null&&r.validationSplit>0&&r.validationSplit<1){g=!0;let $=Math.floor(s[0].shape[0]*(1-r.validationSplit)),F=s[0].shape[0];l=Ol(s,$,F),o=s,s=Ol(s,0,$),p=Ol(a,$,F),i=a,a=Ol(a,0,$),b=l.concat(p)}else r.validationSteps!=null&&(g=!0);let y=s.concat(a).concat(d);e.checkTrainableWeightsConsistency();let v=e.makeTrainFunction(),x=e.getDedupedMetricsNames(),k,S;g?(e.makeTestFunction(),k=e.testFunction,S=x.slice().concat(x.map($=>"val_"+$))):(k=null,b=[],S=x.slice());let C=_C(r.callbacks,r.yieldEvery);return await TU(e,v,y,x,h,r.epochs,r.verbose,C,k,b,r.shuffle,S,r.initialEpoch,null,null)}finally{e.isTraining=!1,zr(s,t),zr(a,n),zr(o,t),zr(i,n),zr(l,c),zr(p,u),d!=null&&_e(d)}}function zC(e){let t=[];e instanceof Te&&(e=[e]);for(let n=0;n<e.length;++n){let r=e[n];if(r.rank===1)t.push(Xd(r,1));else{if(r.rank===0)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(r)}}return t}function zr(e,t){if(e==null)return;let n=[];if(t instanceof Te)n.push(t.id);else if(Array.isArray(t))t.forEach(s=>n.push(s.id));else if(t!=null)for(let s in t){let a=t[s];n.push(a.id)}let r=[];if(e instanceof Te)n.indexOf(e.id)===-1&&r.push(e);else if(Array.isArray(e))e.forEach(s=>{n.indexOf(s.id)===-1&&r.push(s)});else if(e!=null)for(let s in e){let a=e[s];n.indexOf(a.id)===-1&&r.push(a)}r.forEach(s=>{s.isDisposed||s.dispose()})}function NU(e){return e instanceof Te}function Jy(e){return Array.isArray(e)}function Vk(e){return!NU(e)&&!Jy(e)}function Uk(e,t,n,r=!0,s=""){if(t==null||t.length===0){if(e!=null){let o=!1;if(Jy(e)&&e.length>0)o=!0;else if(Vk(e)){for(let i in e)if(e.hasOwnProperty(i)){o=!0;break}}else o=!0;if(o)throw new U(`Error when checking model ${s} expected no data, but got ${e}`)}return[]}if(e==null)return t.map(o=>null);let a;if(Vk(e)){e=e,a=[];for(let o of t){if(e[o]==null)throw new U(`No data provided for "${o}". Need data for each key in: ${t}`);a.push(e[o])}}else if(Jy(e)){if(e=e,e.length!==t.length)throw new U(`Error when checking model ${s}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);a=e}else{if(e=e,t.length>1)throw new U(`The model ${s} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);a=[e]}if(a=zC(a),n!=null)for(let o=0;o<t.length;++o){if(n[o]==null)continue;let i=a[o];if(i.shape.length!==n[o].length)throw new U(`Error when checking ${s}: expected ${t[o]} to have ${n[o].length} dimension(s). but got array with shape ${i.shape}`);for(let c=0;c<n[o].length;++c){if(c===0&&!r)continue;let u=i.shape[c],l=n[o][c];if(l!=null&&l>=0&&u!==l)throw new U(`${s} expected a batch of elements where each example has shape [${n[o].slice(1,n[o].length)}] (i.e.,tensor shape [*,${n[o].slice(1,n[o].length)}]) but the ${s} received an input with ${i.shape[0]} examples, each with shape [${i.shape.slice(1,i.shape.length)}] (tensor shape [${i.shape}])`)}}return a}function _U(e,t,n){let r=sa(e.map(a=>a.shape[0]));r.sort();let s=sa(t.map(a=>a.shape[0]));if(s.sort(),r.length>1)throw new U(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(e.map(a=>a.shape))}`);if(s.length>1)throw new U(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(a=>a.shape))}`);if(r.length>0&&s.length>0&&!w.arraysEqual(r,s))throw new U(`Input Tensors should have the same number of samples as target Tensors. Found ${r[0]} input sample(s) and ${s[0]} target sample(s).`)}function EU(e,t,n){let r=[xi,Bm,td];for(let s=0;s<e.length;++s){let a=e[s],o=t[s],i=n[s];if(o!=null){if(o===td&&a.shape[a.shape.length-1]===1)throw new U(`You are passing a target array of shape ${a.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(r.indexOf(o)!==-1){let c=a.shape.slice(1),u=i.slice(1);for(let l=0;l<c.length;++l){let p=c[l],d=u[l];if(d!=null&&p!==d)throw new U(`A target Tensor with shape ${a.shape} was passed for an output of shape ${i}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function Gk(e,t,n,r=!0,s=""){let a;if(Array.isArray(e)){if(e.length!==t.length)throw new U(`Error when checking model ${s}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);a=e}else{if(t.length>1)throw new U(`The model expects ${t.length} ${s} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);a=[e]}if(n!=null)for(let o=0;o<t.length;++o){if(n[o]==null)continue;let i=a[o];if(i.shape.length!==n[o].length)throw new U(`Error when checking ${s}: expected ${t[o]} to have ${n[o].length} dimension(s), but got array with shape ${JSON.stringify(i.shape)}`);for(let c=0;c<n[o].length;++c){if(c===0&&!r)continue;let u=i.shape[c],l=n[o][c];if(l!=null&&l!==u)throw new U(`Error when checking ${s}: expected ${t[o]} to have shape ${JSON.stringify(n[o])} but got array with shape ${JSON.stringify(i.shape)}.`)}}}function AU(e,t){if(e==null||Array.isArray(e)&&e.length===0)return t.map(r=>[]);let n;if(typeof e=="string"||typeof e=="function")n=[e];else if(Array.isArray(e)||typeof e=="object")n=e;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${e}`);if(Array.isArray(n))return t.map(r=>n);{let r=[];for(let s of t){let a=n.hasOwnProperty(s)?n[s]:[];Array.isArray(a)||(a=[a]),r.push(a)}return r}}var $U="layers-model",Es=class extends ns{constructor(e){super(e),this.isTraining=!1}summary(e,t,n=console.log){if(!this.built)throw new U("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");pU(this,e,t,n)}compile(e){if(e.loss==null&&(e.loss=[]),this.loss=e.loss,typeof e.optimizer=="string")this.optimizer_=dU(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof Rs))throw new U("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(!Array.isArray(e.loss)&&typeof e.loss!="string"&&typeof e.loss!="function"){e.loss=e.loss;for(let a in e.loss)if(this.outputNames.indexOf(a)===-1)throw new U(`Unknown entry in loss dictionary: "${a}". Only expected the following keys: ${this.outputNames}`);for(let a of this.outputNames)e.loss[a]==null&&console.warn(`Output "${a}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${a} during training`),t.push(wy(e.loss[a]))}else if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new U(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);t=e.loss.map(o=>wy(o))}else{let a=wy(e.loss);this.outputs.forEach(o=>{t.push(a)})}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let a=0;a<this.outputs.length;++a){let o=this.internalOutputShapes[a],i=this.outputNames[a];this.feedOutputNames.push(i),this.feedOutputShapes.push(o),this.feedLossFns.push(this.lossFunctions[a])}let n=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Za("loss",()=>{for(let a=0;a<this.outputs.length;++a){if(n.indexOf(a)!==-1)continue;let o=this.lossFunctions[a];this.outputs.length>1&&(this.metricsTensors.push([o,a]),this.metricsNames.push(this.outputNames[a]+"_loss"))}});let r=AU(e.metrics,this.outputNames),s=(a,o,i)=>{this.outputNames.length>1&&(o=this.outputNames[a]+"_"+o),this.metricsNames.push(o),this.metricsTensors.push([i,a])};Za("metric",()=>{for(let a=0;a<this.outputs.length;++a){if(n.indexOf(a)!==-1)continue;let o=r[a];(c=>{let u="",l,p,d;for(let h of c){if(typeof h=="string"&&["accuracy","acc","crossentropy","ce"].indexOf(h)!==-1){let m=this.internalOutputShapes[a];m[m.length-1]===1||this.lossFunctions[a]===Bm?["accuracy","acc"].indexOf(h)!==-1?p=bw:["crossentropy","ce"].indexOf(h)!==-1&&(p=DC):this.lossFunctions[a]===qh?["accuracy","acc"].indexOf(h)!==-1?p=FC:["crossentropy","ce"].indexOf(h)!==-1&&(p=RC):["accuracy","acc"].indexOf(h)!==-1?p=yw:["crossentropy","ce"].indexOf(h)!==-1&&(p=vw);let g;["accuracy","acc"].indexOf(h)!==-1?g="acc":["crossentropy","ce"].indexOf(h)!==-1&&(g="ce"),d=p,l=u+g}else d=lU(h),l=u+lh(h);let f;Za(l,()=>{f=d}),s(a,l,f)}})(o)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,n={}){let r=n.batchSize==null?32:n.batchSize;Yy(r);let s=!0,a=this.standardizeUserDataXY(e,t,s,r);try{let o=a[0].concat(a[1]);this.makeTestFunction();let i=this.testFunction,c=this.testLoop(i,o,r,n.verbose,n.steps);return Mn(c)}finally{zr(a[0],e),zr(a[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),SU(this,e,t)}checkNumSamples(e,t,n,r="steps"){let s;if(n!=null){if(s=null,t!=null)throw new U(`If ${r} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else if(e!=null)Array.isArray(e)?s=e[0].shape[0]:s=e.shape[0];else throw new U(`Either the input data should have a defined shape, or ${r} shoud be specified.`);return s}execute(e,t){if(Array.isArray(t)&&t.length===0)throw new U("`outputs` is an empty Array, which is not allowed.");let n=Array.isArray(t),r=n?t:[t],s=this.retrieveSymbolicTensors(r),a=new ja;if(e instanceof Te&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new U(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let i=0;i<this.inputs.length;++i)a.add(this.inputs[i],e[i])}else for(let i of this.inputs){let c=e[i.name];if(c==null)throw new U(`No value is provided for the model's input ${i.name}`);a.add(i,c)}let o=Pl(s,a);return n?o:o[0]}retrieveSymbolicTensors(e){let t=uo(null,e.length),n=e.length;for(let r of this.layers){let s=Array.isArray(r.output)?r.output:[r.output],a=s.map(o=>o.name);for(let o=0;o<e.length;++o){let i=a.indexOf(e[o]);if(i!==-1&&(t[o]=s[i],n--),n===0)break}if(n===0)break}if(n>0){let r=[];throw t.forEach((s,a)=>{s==null&&r.push(e[a])}),new U(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(r)}`)}return t}predictLoop(e,t=32,n=!1){return O(()=>{let r=this.checkNumSamples(e);if(n)throw new Pe("Verbose predictLoop() is not implemented yet.");let s=Zy(r,t),a=this.outputs.map(o=>[]);for(let o=0;o<s.length;++o)O(()=>{let c=s[o][0],u=s[o][1],l=Ol(e,c,u),p=[];if(Array.isArray(l))for(let h=0;h<l.length;++h)p.push({key:this.inputs[h],value:l[h]});else p.push({key:this.inputs[0],value:l});let d=new ja(p);return Pl(this.outputs,d)}).forEach((c,u)=>a[u].push(c));return Mn(a.map(o=>Je(o,0)))})}predict(e,t={}){let n=zC(e);Gk(n,this.inputNames,this.feedInputShapes,!1);try{let r=t.batchSize==null?32:t.batchSize;return Yy(r),this.predictLoop(n,r)}finally{zr(n,e)}}predictOnBatch(e){Gk(e,this.inputNames,this.feedInputShapes,!0);let t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,n=!0,r){if(this.optimizer_==null)throw new Wr("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let s=[];for(let a=0;a<this.feedOutputShapes.length;++a){let o=this.feedOutputShapes[a];this.feedLossFns[a]===qh?s.push(o.slice(0,o.length-1).concat([1])):s.push(o)}if(e=Uk(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=Uk(t,this.feedOutputNames,s,!1,"target"),_U(e,t,null),EU(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&r!=null&&r>0&&e[0].shape[0]%r!==0)throw new U(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${r}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,n,r,s=!0,a){let[o,i]=this.standardizeUserDataXY(e,t,s,a);if(n!=null)throw new Error("sample weight is not supported yet.");let c=null;if(r!=null){let u=OC(r,this.outputNames);c=[];for(let l=0;l<u.length;++l)c.push(await MC(i[l],null,u[l]))}return[o,i,c]}testLoop(e,t,n,r=0,s){return O(()=>{let a=this.checkNumSamples(t,n,s,"steps"),o=[];if(r>0)throw new Pe("Verbose mode is not implemented yet.");if(s!=null)throw new Pe("steps mode in testLoop() is not implemented yet");{let i=Zy(a,n),c=Ke(qr(0,a));for(let u=0;u<i.length;++u){let l=i[u][0],p=i[u][1],d=Ja(c,l,p-l),h=ww(t,d),f=e(h);if(u===0)for(let m=0;m<f.length;++m)o.push(xe(0));for(let m=0;m<f.length;++m){let g=f[m];o[m]=Y(o[m],B(p-l,g))}}for(let u=0;u<o.length;++u)o[u]=fe(o[u],a)}return o})}getDedupedMetricsNames(){let e=this.metricsNames,t=[];for(let n=0;n<e.length;++n){let r=e[n],s=r;if(_k(e,r)>1){let a=_k(e.slice(0,n),r);s+=`_${a}`}t.push(s)}return t}makeTrainFunction(){return e=>{let t=[],n=e.slice(0,this.inputs.length),r=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),a=[],o=()=>{let l=[];for(let f=0;f<this.inputs.length;++f)l.push({key:this.inputs[f],value:n[f]});let p=new ja(l),d=Pl(this.outputs,p,{training:!0}),h;for(let f=0;f<this.lossFunctions.length;++f){let m=this.lossFunctions[f],g=m(r[f],d[f]);s[f]!=null&&(g=yU(g,s[f]));let b=Ct(g);t.push(b),f===0?h=g:h=Y(h,g)}for(let f=0;f<this.metricsTensors.length;++f){let m;if(this.outputs.length>1&&f<this.outputs.length)m=t[f];else{let g=this.metricsTensors[f][0],b=this.metricsTensors[f][1];m=Ct(g(r[b],d[b]))}Jt(m),a.push(m)}return h=Ct(h),this.calculateLosses().forEach(f=>{h=Y(h,f)}),h},i=this.collectedTrainableWeights.map(l=>l.read()),c=!0;return[this.optimizer_.minimize(o,c,i)].concat(a)}}makeTestFunction(){this.testFunction=e=>O(()=>{let t=[],n,r=e.slice(0,this.inputs.length),s=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),a=[];for(let c=0;c<this.inputs.length;++c)a.push({key:this.inputs[c],value:r[c]});let o=new ja(a),i=Pl(this.outputs,o);for(let c=0;c<this.lossFunctions.length;++c){let u=this.lossFunctions[c],l=Ct(u(s[c],i[c]));c===0?n=l:n=Y(n,l),t.push(n)}for(let c=0;c<this.metricsTensors.length;++c){let u=this.metricsTensors[c][0],l=this.metricsTensors[c][1],p=Ct(u(s[l],i[l]));t.push(p)}return t})}async fit(e,t,n={}){return CU(this,e,t,n)}async fitDataset(e,t){return wU(this,e,t)}async trainOnBatch(e,t){let n=await this.standardizeUserData(e,t),r=n[0],s=n[1],o=this.makeTrainFunction()(r.concat(s)),i=[];for(let c of o){let u=await c.data();i.push(u[0])}return _e(o),zr(n[0],e),zr(n[1],t),Mn(i)}getNamedWeights(e){let t=[],n=e!=null&&e.trainableOnly,r=n?this.trainableWeights:this.weights,s=this.getWeights(n);for(let a=0;a<r.length;++a)n&&!r[a].trainable||t.push({name:r[a].originalName,tensor:s[a]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){let e=super.dispose();if(e.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let t=Lh().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-Lh().numTensors}return e}getLossIdentifiers(){let e;if(typeof this.loss=="string")e=Ss(this.loss);else if(Array.isArray(this.loss)){for(let t of this.loss)if(typeof t!="string")throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(t=>Ss(t))}else{let t=Object.keys(this.loss);e={};let n=this.loss;for(let r of t)if(typeof n[r]=="string")e[r]=Ss(n[r]);else throw new Error("Serialization of non-string loss is not supported.")}return e}getMetricIdentifiers(){if(typeof this.metrics=="string"||typeof this.metrics=="function")return[Ss(lh(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>Ss(lh(e)));{let e={};for(let t in this.metrics)e[t]=Ss(lh(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(e.weighted_metrics!=null)throw new Error("Loading weight_metrics is not supported yet.");if(e.loss_weights!=null)throw new Error("Loading loss_weights is not supported yet.");if(e.sample_weight_mode!=null)throw new Error("Loading sample_weight_mode is not supported yet.");let t=nd(e.optimizer_config),n=Gr(t),r;if(typeof e.loss=="string")r=Ga(e.loss);else if(Array.isArray(e.loss))r=e.loss.map(a=>Ga(a));else if(e.loss!=null){r={};for(let a in e.loss)r[a]=Ga(e.loss[a])}let s;if(Array.isArray(e.metrics))s=e.metrics.map(a=>Ga(a));else if(e.metrics!=null){s={};for(let a in e.metrics)s[a]=Ga(e.metrics[a])}this.compile({loss:r,metrics:s,optimizer:n})}async save(e,t){if(typeof e=="string"){let c=Ut.getSaveHandlers(e);if(c.length===0)throw new U(`Cannot find any save handlers for URL '${e}'`);if(c.length>1)throw new U(`Found more than one (${c.length}) save handlers for URL '${e}'`);e=c[0]}if(e.save==null)throw new U("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let n=await Ut.encodeWeights(this.getNamedWeights(t)),r=!1,s=null,o={modelTopology:this.toJSON(s,r),format:$U,generatedBy:`TensorFlow.js tfjs-layers v${xw}`,convertedBy:null};if((t==null?!1:t.includeOptimizer)&&this.optimizer!=null){o.trainingConfig=this.getTrainingConfig();let c="optimizer",{data:u,specs:l}=await Ut.encodeWeights(await this.optimizer.getWeights(),c);n.specs.push(...l),n.data=Ut.concatenateArrayBuffers([n.data,u])}return this.userDefinedMetadata!=null&&(zk(this.userDefinedMetadata,this.name,!0),o.userDefinedMetadata=this.userDefinedMetadata),o.weightData=n.data,o.weightSpecs=n.specs,e.save(o)}setUserDefinedMetadata(e){zk(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Es.className="Model";se.registerClass(Es);var BC=class extends Es{};BC.className="Functional";se.registerClass(BC);async function DU(e,t){"modelTopology"in e||(e={modelTopology:e}),e=e;let n=e.modelTopology;n.model_config!=null&&(n=n.model_config);let r=nd(n),s=Gr(r,t);if(e.weightsManifest!=null){let a=await Ut.loadWeights(e.weightsManifest,e.pathPrefix,s.weights.map(i=>i.originalName)),o={};for(let i of s.weights)o[i.originalName]=a[i.originalName];s.loadWeights(o),_e(a)}return s}async function FU(e,t){if(t==null&&(t={}),typeof e=="string"){let n=Ut.getLoadHandlers(e,t);if(n.length===0)n.push(Ut.browserHTTPRequest(e,t));else if(n.length>1)throw new U(`Found more than one (${n.length}) load handlers for URL '${e}'`);e=n[0]}return RU(e,void 0,t)}async function RU(e,t,n){if(n==null&&(n={}),e.load==null)throw new U("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let r=await e.load(),s=r.modelTopology;s.model_config!=null&&(s=s.model_config);let a=n.strict==null?!0:n.strict,o=r.weightData!=null&&r.weightSpecs!=null&&a,i=Gr(nd(s),t,o),c=r.trainingConfig;if(c!=null&&i.loadTrainingConfig(c),r.userDefinedMetadata!=null&&i.setUserDefinedMetadata(r.userDefinedMetadata),r.weightData!=null){if(r.weightSpecs==null)throw new U("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:u,optimizerWeights:l}=PU(r.weightData,r.weightSpecs);i.loadWeights(u,a),i.optimizer!=null&&l.length>0&&await i.optimizer.setWeights(l),_e(u),_e(l.map(p=>p.tensor))}return i}function PU(e,t){let n=Ut.decodeWeights(e,t),r={},s=[];return t.forEach(a=>{a.group==="optimizer"?s.push({name:a.name,tensor:n[a.name]}):r[a.name]=n[a.name]}),{modelWeights:r,optimizerWeights:s}}var Tc=class extends Es{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=e.name!=null?e.name:Em("sequential_"),e.layers!=null)for(let t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(n=>n<0))throw new U(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(e){let t=e instanceof Tc||e instanceof Es,n;if(t){if(n=e,n.outputs.length!==1)throw new U("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(n.inputs.length!==1)throw new U("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(this.outputs.length===0){if(e.inboundNodes.length===0){if(e.batchInputShape==null)throw new U("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let r=xC({batchShape:e.batchInputShape,dtype:e.dtype,name:e.name+"_input"});e.apply(r)}if(t)this.outputs=n.outputs,this.inputs=n.inputs;else{if(e.inboundNodes.length!==1)throw new U(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);if(e.inboundNodes[0].outputTensors.length!==1)throw new U("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[e.inboundNodes[0].outputTensors[0]],this.inputs=vC(this.outputs[0])}this.inboundNodes=[],new Lm({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:uo(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(r=>r.shape),outputShapes:this.outputs[0].shape})}else{let r=e.apply(this.outputs[0]);if(Array.isArray(r))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[r],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(e),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return this.model==null&&this.build(),this.model.call(e,t)}build(e){if(tt(e),this.inputs.length===0||this.outputs.length===0)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Es({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,n=console.log){this.built||this.build(),super.summary(e,t,n)}setWeights(e){this.model==null&&this.build(),this.model.setWeights(e)}evaluate(e,t,n={}){if(!this.built)throw new Wr("The model needs to be compiled before being used.");return this.model.evaluate(e,t,n)}async evaluateDataset(e,t){if(!this.built)throw new Wr("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return this.model==null&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return this.model==null&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,n={}){if(!this.built)throw new Wr("The model needs to be compiled before being used.");return this.model.fit(e,t,n)}async fitDataset(e,t){if(!this.built)throw new Wr("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(e,t,n={},r=!1){let s,a={};if(t instanceof Array){if(t[0].className==null||t[0].className==="Merge")throw new U("Legacy serialization format not supported yet.");s=t}else w.assert(t.layers!=null,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),s=t.layers,delete t.layers,a=t;let o=new e(a);if(!(o instanceof Tc))throw new Pe(`Sequential.fromConfig called on non-Sequential input: ${o}`);for(let i of s){let u=Gr(i,void 0,r);r&&u.setFastWeightInitDuringBuild(!0),o.add(u)}return o}set stopTraining(e){if(this.model==null)throw new U("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(this.model==null)throw new U("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let e=[];for(let t of this.layers){let n={};n.className=t.getClassName(),n.config=t.getConfig(),e.push(n)}return{name:this.name,layers:e}}};Tc.className="Sequential";se.registerClass(Tc);function OU(e){return new Es(e)}function MU(e){return new Tc(e)}function WC(e){return xC(e)}function LU(e,t){_r.registerCallbackConstructor(e,t)}var Un=class extends se.Serializable{getConfig(){return{}}},VC=class extends Un{apply(e,t=1){return oV(e,t)}};VC.className="elu";se.registerClass(VC);var UC=class extends Un{apply(e){return lm(e)}};UC.className="selu";se.registerClass(UC);var GC=class extends Un{apply(e){return Xe(e)}};GC.className="relu";se.registerClass(GC);var HC=class extends Un{apply(e){return O(()=>Vu(6,Xe(e)))}};HC.className="relu6";se.registerClass(HC);var qC=class extends Un{apply(e){return e}};qC.className="linear";se.registerClass(qC);var jC=class extends Un{apply(e){return dr(e)}};jC.className="sigmoid";se.registerClass(jC);var KC=class extends Un{apply(e){return cV(e)}};KC.className="hardSigmoid";se.registerClass(KC);var XC=class extends Un{apply(e){return mi(e)}};XC.className="softplus";se.registerClass(XC);var YC=class extends Un{apply(e){return iV(e)}};YC.className="softsign";se.registerClass(YC);var ZC=class extends Un{apply(e){return oo(e)}};ZC.className="tanh";se.registerClass(ZC);var Iw=class extends Un{apply(e,t=-1){return Xr(e,t)}};Iw.className="softmax";se.registerClass(Iw);var JC=class extends Un{apply(e,t=-1){return nm(e,t)}};JC.className="logSoftmax";se.registerClass(JC);var QC=class extends Un{apply(e,t=1){return O(()=>B(dr(B(e,t)),e))}};QC.className="swish";se.registerClass(QC);var eN=class extends Un{apply(e){return O(()=>B(e,oo(mi(e))))}};eN.className="mish";se.registerClass(eN);function la(e){return e.getClassName()}function Iy(e,t={}){return Kd(e,se.SerializationMap.getMap().classNameMap,t,"activation")}function da(e){if(e==null){let t={};return t.className="linear",t.config={},Iy(t)}if(typeof e=="string"){let t={};return t.className=e,t.config={},Iy(t)}else return e instanceof Un?e:Iy(e)}function kw(e){if(e!=null&&typeof e!="object")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${e}`)}var tN=class extends se.Serializable{},Qd=class extends tN{constructor(e){super(),kw(e),this.l1=e==null||e.l1==null?.01:e.l1,this.l2=e==null||e.l2==null?.01:e.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(e){return O(()=>{let t=kt([1]);return this.hasL1&&(t=Y(t,ge(B(this.l1,Lt(e))))),this.hasL2&&(t=Y(t,ge(B(this.l2,Yd(e))))),W(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};Qd.className="L1L2";se.registerClass(Qd);function zU(e){return kw(e),new Qd({l1:e!=null?e.l1:null,l2:0})}function BU(e){return kw(e),new Qd({l2:e!=null?e.l2:null,l1:0})}var Hk={l1l2:"L1L2"};function ut(e){return ew(e)}function qk(e,t={}){return Kd(e,se.SerializationMap.getMap().classNameMap,t,"regularizer")}function Tt(e){if(e==null)return null;if(typeof e=="string"){let n={className:e in Hk?Hk[e]:e,config:{}};return qk(n)}else return e instanceof tN?e:qk(e)}var Sw=class extends Ue{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null&&(this.maxValue=e.maxValue)}call(e,t){e=Ce(e);let n=Xe(e);return this.maxValue!=null&&(n=en(n,0,this.maxValue)),n}computeOutputShape(e){return e}getConfig(){let e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};Sw.className="ReLU";se.registerClass(Sw);var Tw=class extends Ue{constructor(e){super(e==null?{}:e),this.DEFAULT_ALPHA=.3,e==null&&(e={}),this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ce(e);return Ld(n,this.alpha)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Tw.className="LeakyReLU";se.registerClass(Tw);var Cw=class extends Ue{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA_INITIALIZER="zeros",e==null&&(e={}),this.supportsMasking=!0,this.alphaInitializer=St(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=Tt(e.alphaRegularizer),this.alphaConstraint=Kt(e.alphaConstraint),e.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else if(typeof e.sharedAxes=="number")this.sharedAxes=[e.sharedAxes];else throw new U(`Expected sharedAxes to be a number or an array of numbers, but got ${e.sharedAxes}`)}build(e){e=tt(e);let t=e.slice(1);if(this.sharedAxes!=null)for(let r of this.sharedAxes)t[r-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(this.sharedAxes!=null)for(let r=1;r<e.length;++r)n[r]=e[r];this.inputSpec=[new zt({ndim:e.length,axes:n})],this.built=!0}call(e,t){return e=Ce(e),Ud(e,this.alpha.read())}getConfig(){let e={alphaInitializer:Nt(this.alphaInitializer),alphaRegularizer:ut(this.alphaRegularizer),alphaConstraint:jt(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};Cw.className="PReLU";se.registerClass(Cw);var Nw=class extends Ue{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA=1,e==null&&(e={}),e.alpha!=null&&e.alpha!==this.DEFAULT_ALPHA)throw new Pe(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ce(e);return Lu(n)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Nw.className="ELU";se.registerClass(Nw);var _w=class extends Ue{constructor(e){super(e==null?{}:e),this.DEFAULT_THETA=1,e==null&&(e={}),this.theta=e.theta==null?this.DEFAULT_THETA:e.theta}call(e,t){let n=Ce(e);return B(n,ce(An(n,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){let e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};_w.className="ThresholdedReLU";se.registerClass(_w);var Ew=class extends Ue{constructor(e){super(e==null?{}:e),this.DEFAULT_AXIS=1,e==null&&(e={}),this.softmax=new Iw().apply,this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis}call(e,t){let n=Ce(e);return this.softmax(n,this.axis)}computeOutputShape(e){return e}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};Ew.className="Softmax";se.registerClass(Ew);function pc(e,t,n){if(typeof e=="number")return uo(e,t);if(e.length!==t)throw new U(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let r=0;r<t;++r){let s=e[r];if(!nV(s))throw new U(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${s}`)}return e}function Hr(e,t,n,r,s=1){if(e==null)return e;let a=t+(t-1)*(s-1),o;return n==="same"?o=e:o=e-a+1,Math.floor((o+r-1)/r)}function as(e,t,n,r){if(e==null)return null;if(r==="valid")e=e*t+ua([n-t,0]);else if(r==="same")e=e*t;else throw new U(`Unsupport padding mode: ${r}.`);return e}function Aw(e,t){return O(()=>(Pt(t),t==="channelsFirst"?Ee(e,[0,2,3,1]):e))}function nN(e,t){return O(()=>(Pt(t),t==="channelsFirst"?Ee(e,[0,2,3,4,1]):e))}function WU(e,t,n,r=1,s="valid",a,o=1){return O(()=>{if(a==null&&(a=jr()),Pt(a),e.shape.length!==3)throw new U(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(t.shape.length!==3)throw new U(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(n!=null&&n.shape.length!==1)throw new U(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(a==="channelsFirst"&&(e=Ee(e,[0,2,1])),s==="causal")throw new Pe("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let i=Xf(e,t,r,s==="same"?"same":"valid","NWC",o);return n!=null&&(i=Yr(i,n)),i})}function jk(e,t,n,r=[1,1],s="valid",a,o,i=null){return O(()=>{if(a==null&&(a=jr()),Pt(a),e.rank!==3&&e.rank!==4)throw new U(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(t.rank!==3&&t.rank!==4)throw new U(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let c=Aw(e,a);if(s==="causal")throw new Pe("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return c=Ic.conv2d({x:c,filter:t,strides:r,pad:s==="same"?"same":"valid",dilations:o,dataFormat:"NHWC",bias:n,activation:i}),a==="channelsFirst"&&(c=Ee(c,[0,3,1,2])),c})}function VU(e,t,n,r=[1,1,1],s="valid",a,o){return O(()=>{if(a==null&&(a=jr()),Pt(a),e.rank!==4&&e.rank!==5)throw new U(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(t.rank!==4&&t.rank!==5)throw new U(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let i=nN(e,a);if(s==="causal")throw new Pe("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return i=mx(i,t,r,s==="same"?"same":"valid","NDHWC",o),n!=null&&(i=Yr(i,n)),a==="channelsFirst"&&(i=Ee(i,[0,4,1,2,3])),i})}var $w=class extends Ue{constructor(e,t){if(super(t),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",$w.verifyArgs(t),this.rank=e,Qt(this.rank,"rank"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new Pe(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=pc(t.kernelSize,e,"kernelSize"),this.strides=pc(t.strides==null?1:t.strides,e,"strides"),this.padding=t.padding==null?"valid":t.padding,br(this.padding),this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Pt(this.dataFormat),this.activation=da(t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.biasInitializer=St(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Kt(t.biasConstraint),this.biasRegularizer=Tt(t.biasRegularizer),this.activityRegularizer=Tt(t.activityRegularizer),this.dilationRate=pc(t.dilationRate==null?1:t.dilationRate,e,"dilationRate"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new U(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new U(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new U(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(e){if(ss("kernelSize"in e,"required key 'kernelSize' not in config"),typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,3))throw new U(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){let e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:la(this.activation),useBias:this.useBias,biasInitializer:Nt(this.biasInitializer),biasRegularizer:ut(this.biasRegularizer),activityRegularizer:ut(this.activityRegularizer),biasConstraint:jt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},ep=class extends $w{constructor(e,t){super(e,t),this.kernel=null,ep.verifyArgs(t),this.filters=t.filters,Qt(this.filters,"filters"),this.kernelInitializer=St(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Kt(t.kernelConstraint),this.kernelRegularizer=Tt(t.kernelRegularizer)}build(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new U(`The channel dimension of the input should be defined. Found ${e[t]}`);let n=e[t],r=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",r,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:n}}],this.built=!0}call(e,t){return O(()=>{e=Ce(e);let n,r=this.bias==null?null:this.bias.read(),s=lC(this.activation.getClassName());if(s!=null&&this.rank===2)n=jk(e,this.kernel.read(),r,this.strides,this.padding,this.dataFormat,this.dilationRate,s);else{if(this.rank===1)n=WU(e,this.kernel.read(),r,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)n=jk(e,this.kernel.read(),r,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)n=VU(e,this.kernel.read(),r,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new Pe("convolutions greater than 3D are not implemented yet.");this.activation!=null&&(n=this.activation.apply(n))}return n})}computeOutputShape(e){e=tt(e);let t=[],n=this.dataFormat==="channelsLast"?e.slice(1,e.length-1):e.slice(2);for(let s=0;s<n.length;++s){let a=Hr(n[s],this.kernelSize[s],this.padding,this.strides[s],typeof this.dilationRate=="number"?this.dilationRate:this.dilationRate[s]);t.push(a)}let r=[e[0]];return this.dataFormat==="channelsLast"?(r=r.concat(t),r.push(this.filters)):(r.push(this.filters),r=r.concat(t)),r}getConfig(){let e={filters:this.filters,kernelInitializer:Nt(this.kernelInitializer),kernelRegularizer:ut(this.kernelRegularizer),kernelConstraint:jt(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||typeof e.filters!="number"||e.filters<1)throw new U(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(e.filters)}`)}},tp=class extends ep{constructor(e){super(2,e),tp.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,2))throw new U(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}};tp.className="Conv2D";se.registerClass(tp);var np=class extends ep{constructor(e){super(3,e),np.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!(Array.isArray(e.kernelSize)&&(e.kernelSize.length===1||e.kernelSize.length===3)))throw new U(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}};np.className="Conv3D";se.registerClass(np);var Dw=class extends tp{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:4})],this.padding!=="same"&&this.padding!=="valid")throw new U(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=tt(e),e.length!==4)throw new U("Input should have rank 4; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new U("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:4,axes:{[t]:n}})],this.built=!0}call(e,t){return O(()=>{let n=Ce(e);if(n.shape.length!==4)throw new U(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let r=n.shape,s=r[0],a,o;this.dataFormat==="channelsFirst"?(a=2,o=3):(a=1,o=2);let i=r[a],c=r[o],u=this.kernelSize[0],l=this.kernelSize[1],p=this.strides[0],d=this.strides[1],h=as(i,p,u,this.padding),f=as(c,d,l,this.padding),m=[s,h,f,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,1]));let g=Yf(n,this.kernel.read(),m,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(g=Ee(g,[0,3,1,2])),this.bias!=null&&(g=Yr(g,this.bias.read(),this.dataFormat)),this.activation!=null&&(g=this.activation.apply(g)),g})}computeOutputShape(e){e=tt(e);let t=e.slice(),n,r,s;this.dataFormat==="channelsFirst"?(n=1,r=2,s=3):(n=3,r=1,s=2);let a=this.kernelSize[0],o=this.kernelSize[1],i=this.strides[0],c=this.strides[1];return t[n]=this.filters,t[r]=as(t[r],i,a,this.padding),t[s]=as(t[s],c,o,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Dw.className="Conv2DTranspose";se.registerClass(Dw);var Fw=class extends np{constructor(e){if(super(e),this.inputSpec=[new zt({ndim:5})],this.padding!=="same"&&this.padding!=="valid")throw new U(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=tt(e),e.length!==5)throw new U("Input should have rank 5; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new U("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new zt({ndim:5,axes:{[t]:n}})],this.built=!0}call(e,t){return O(()=>{let n=Ce(e);if(n.shape.length!==5)throw new U(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let r=n.shape,s=r[0],a,o,i;this.dataFormat==="channelsFirst"?(i=2,a=3,o=4):(i=1,a=2,o=3);let c=r[i],u=r[a],l=r[o],p=this.kernelSize[0],d=this.kernelSize[1],h=this.kernelSize[2],f=this.strides[0],m=this.strides[1],g=this.strides[2],b=as(c,f,p,this.padding),y=as(u,m,d,this.padding),v=as(l,g,h,this.padding),x=[s,b,y,v,this.filters];this.dataFormat!=="channelsLast"&&(n=Ee(n,[0,2,3,4,1]));let k=gx(n,this.kernel.read(),x,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(k=Ee(k,[0,4,1,2,3])),this.bias!==null&&(k=Yr(k,this.bias.read(),this.dataFormat)),this.activation!==null&&(k=this.activation.apply(k)),k})}computeOutputShape(e){e=tt(e);let t=e.slice(),n,r,s,a;this.dataFormat==="channelsFirst"?(n=1,r=2,s=3,a=4):(n=4,r=1,s=2,a=3);let o=this.kernelSize[0],i=this.kernelSize[1],c=this.kernelSize[2],u=this.strides[0],l=this.strides[1],p=this.strides[2];return t[n]=this.filters,t[r]=as(t[r],u,o,this.padding),t[s]=as(t[s],l,i,this.padding),t[a]=as(t[a],p,c,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Fw.className="Conv3DTranspose";se.registerClass(Fw);var rN=class extends ep{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,t.filters==null)throw new U("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(t.kernelInitializer!=null||t.kernelRegularizer!=null||t.kernelConstraint!=null)throw new U("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(t.padding!=null&&t.padding!=="same"&&t.padding!=="valid")throw new U(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=Tt(t.depthwiseRegularizer),this.depthwiseConstraint=Kt(t.depthwiseConstraint),this.pointwiseInitializer=St(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=Tt(t.pointwiseRegularizer),this.pointwiseConstraint=Kt(t.pointwiseConstraint)}build(e){if(e=tt(e),e.length<this.rank+2)throw new U(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null||e[t]<0)throw new U(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(e[t])}`);let n=e[t],r=this.kernelSize.concat([n,this.depthMultiplier]),s=[];for(let o=0;o<this.rank;++o)s.push(1);s.push(n*this.depthMultiplier,this.filters);let a=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",r,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,a,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",s,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,a,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,a,this.biasConstraint):this.bias=null,this.inputSpec=[new zt({ndim:this.rank+2,axes:{[t]:n}})],this.built=!0}call(e,t){return O(()=>{e=Ce(e);let n;if(this.rank===1)throw new Pe("1D separable convolution is not implemented yet.");return this.rank===2&&(this.dataFormat==="channelsFirst"&&(e=Ee(e,[0,2,3,1])),n=Ca(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(n=Yr(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),this.dataFormat==="channelsFirst"&&(n=Ee(n,[0,3,1,2])),n})}getConfig(){let e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=Nt(this.depthwiseInitializer),e.pointwiseInitializer=Nt(this.pointwiseInitializer),e.depthwiseRegularizer=ut(this.depthwiseRegularizer),e.pointwiseRegularizer=ut(this.pointwiseRegularizer),e.depthwiseConstraint=jt(this.depthwiseConstraint),e.pointwiseConstraint=jt(this.pointwiseConstraint),e}};rN.className="SeparableConv";var Rw=class extends rN{constructor(e){super(2,e)}};Rw.className="SeparableConv2D";se.registerClass(Rw);var Wm=class extends ep{constructor(e){super(1,e),Wm.verifyArgs(e),this.inputSpec=[{ndim:3}]}getConfig(){let e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!tw(e.kernelSize,"number",1,1))throw new U(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}};Wm.className="Conv1D";se.registerClass(Wm);var Pw=class extends Ue{constructor(e){super(e),typeof e.cropping=="number"?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:typeof e.cropping[0]=="number"?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=e.dataFormat===void 0?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return this.dataFormat==="channelsFirst"?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return O(()=>{if(e=Ce(e),this.dataFormat==="channelsLast"){let n=uh(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return uh(n,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let n=uh(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return uh(n,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Pw.className="Cropping2D";se.registerClass(Pw);var Ow=class extends Ue{constructor(e){super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=e.size==null?this.DEFAULT_SIZE:e.size,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Pt(this.dataFormat),this.interpolation=e.interpolation==null?"nearest":e.interpolation,Q4(this.interpolation)}computeOutputShape(e){if(this.dataFormat==="channelsFirst"){let t=e[2]==null?null:this.size[0]*e[2],n=e[3]==null?null:this.size[1]*e[3];return[e[0],e[1],t,n]}else{let t=e[1]==null?null:this.size[0]*e[1],n=e[2]==null?null:this.size[1]*e[2];return[e[0],t,n,e[3]]}}call(e,t){return O(()=>{let n=Ce(e),r=n.shape;if(this.dataFormat==="channelsFirst"){n=Ee(n,[0,2,3,1]);let s=this.size[0]*r[2],a=this.size[1]*r[3],o=this.interpolation==="nearest"?Br.resizeNearestNeighbor(n,[s,a]):Br.resizeBilinear(n,[s,a]);return Ee(o,[0,3,1,2])}else{let s=this.size[0]*r[1],a=this.size[1]*r[2];return this.interpolation==="nearest"?Br.resizeNearestNeighbor(n,[s,a]):Br.resizeBilinear(n,[s,a])}})}getConfig(){let e={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}};Ow.className="UpSampling2D";se.registerClass(Ow);function UU(e,t,n=[1,1],r="valid",s,a){return O(()=>{s==null&&(s=jr()),Pt(s);let o=Aw(e,s);if(e.rank!==4)throw new U(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(t.rank!==4)throw new U(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return o=Sa(o,t,n,r==="same"?"same":"valid","NHWC",a),s==="channelsFirst"&&(o=Ee(o,[0,3,1,2])),o})}var Mw=class extends $w{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=St(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Kt(e.depthwiseConstraint),this.depthwiseRegularizer=Tt(e.depthwiseRegularizer)}build(e){if(e=tt(e),e.length<4)throw new U(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);let t=this.dataFormat==="channelsFirst"?1:3;if(e[t]==null||e[t]<0)throw new U(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);let n=e[t],r=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",r,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return O(()=>{e=Ce(e);let n=UU(e,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(n=Yr(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),n})}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],r=this.dataFormat==="channelsFirst"?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,s=Hr(t,this.kernelSize[0],this.padding,this.strides[0]),a=Hr(n,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat==="channelsFirst"?[e[0],r,s,a]:[e[0],s,a,r]}getConfig(){let e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=Nt(this.depthwiseInitializer),e.depthwiseRegularizer=ut(this.depthwiseRegularizer),e.depthwiseConstraint=jt(this.depthwiseRegularizer),e}};Mw.className="DepthwiseConv2D";se.registerClass(Mw);function sN(e,t,n,r){if(Array.isArray(e)){if(t!=null||n!=null)throw new U("When inputs is an array, neither initialState or constants should be provided");r!=null&&(n=e.slice(e.length-r,e.length),e=e.slice(0,e.length-r)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function s(a){return a==null||Array.isArray(a)?a:[a]}return t=s(t),n=s(n),{inputs:e,initialState:t,constants:n}}function aN(e,t,n,r=!1,s,a,o=!1,i=!1){return O(()=>{let c=t.shape.length;if(c<3)throw new U(`Input should be at least 3D, but is ${c}D.`);let u=[1,0].concat(qr(2,c));if(t=Ee(t,u),a!=null)throw new Pe("The rnn() functoin of the deeplearn.js backend does not support constants yet.");o&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),s!=null&&(s=ce(ce(s,"bool"),"float32"),s.rank===c-1&&(s=Zt(s,-1)),s=Ee(s,u)),r&&(t=fr(t,0),s!=null&&(s=fr(s,0)));let l=[],p,d=n,h=t.shape[0],f=lt(t),m;s!=null&&(m=lt(s));for(let b=0;b<h;++b){let y=f[b],v=O(()=>e(y,d));if(s==null)p=v[0],d=v[1];else{let x=O(()=>{let k=m[b],S=de(rr(k),k),C=Y(B(v[0],k),B(d[0],S)),E=d.map(($,F)=>Y(B(v[1][F],k),B($,S)));return{output:C,newStates:E}});p=x.output,d=x.newStates}i&&l.push(p)}let g;return i&&(g=Ft(l,1)),[p,g,d]})}var ms=class extends Ue{constructor(e){super(e);let t;if(e.cell==null)throw new U("cell property is missing for the constructor of RNN.");if(Array.isArray(e.cell)?t=new Gm({cells:e.cell}):t=e.cell,t.stateSize==null)throw new U("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=e.returnSequences==null?!1:e.returnSequences,this.returnState=e.returnState==null?!1:e.returnState,this.goBackwards=e.goBackwards==null?!1:e.goBackwards,this._stateful=e.stateful==null?!1:e.stateful,this.unroll=e.unroll==null?!1:e.unroll,this.supportsMasking=!0,this.inputSpec=[new zt({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return qr(0,e).map(t=>null)}else return this.states_}setStates(e){this.states_=e}computeOutputShape(e){qy(e)&&(e=e[0]),e=e;let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);let n=t[0],r;if(this.returnSequences?r=[e[0],e[1],n]:r=[e[0],n],this.returnState){let s=[];for(let a of t)s.push([e[0],a]);return[r].concat(s)}else return r}computeMask(e,t){return O(()=>{Array.isArray(t)&&(t=t[0]);let n=this.returnSequences?t:null;if(this.returnState){let r=this.states.map(s=>null);return[n].concat(r)}else return n})}get states(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let n=0;n<e;++n)t.push(null);return t}else return this.states_}set states(e){this.states_=e}build(e){if(this.numConstants!=null)throw new Pe("Constants support is not implemented in RNN yet.");qy(e)&&(e=e[0]),e=e;let n=this.stateful?e[0]:null,r=e.slice(2);this.inputSpec[0]=new zt({shape:[n,null,...r]});let s=[e[0]].concat(e.slice(2));this.cell.build(s);let a;if(Array.isArray(this.cell.stateSize)?a=this.cell.stateSize:a=[this.cell.stateSize],this.stateSpec!=null){if(!w.arraysEqual(this.stateSpec.map(o=>o.shape[o.shape.length-1]),a))throw new U(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=a.map(o=>new zt({shape:[null,o]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){O(()=>{if(!this.stateful)throw new ks("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape[0];if(n==null)throw new U("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(r=>kt([n,r])):this.states_=[kt([n,this.cell.stateSize])];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(r=>kt([n,r])):this.states_[0]=kt([n,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new U(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t===!0?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let r=0;r<this.states_.length;++r){let s=e[r],a=Array.isArray(this.cell.stateSize)?this.cell.stateSize[r]:this.cell.stateSize,o=[n,a];if(!w.arraysEqual(s.shape,o))throw new U(`State ${r} is incompatible with layer ${this.name}: expected shape=${o}, received shape=${s.shape}`);this.states_[r]=s}}this.states_=this.states_.map(r=>Jt(r.clone()))})}apply(e,t){let n=t==null?null:t.initialState,r=t==null?null:t.constants;t==null&&(t={});let s=sN(e,n,r,this.numConstants);e=s.inputs,n=s.initialState,r=s.constants;let a=[],o=[];if(n!=null){t.initialState=n,a=a.concat(n),this.stateSpec=[];for(let c of n)this.stateSpec.push(new zt({shape:c.shape}));o=o.concat(this.stateSpec)}if(r!=null&&(t.constants=r,a=a.concat(r),this.numConstants=r.length),a[0]instanceof Vr){let c=[e].concat(a),u=this.inputSpec.concat(o),l=this.inputSpec;this.inputSpec=u;let p=super.apply(c,t);return this.inputSpec=l,p}else return super.apply(e,t)}call(e,t){return O(()=>{let n=t==null?null:t.mask,r=t==null?null:t.training,s=t==null?null:t.initialState;e=Ce(e),s==null&&(this.stateful?s=this.states_:s=this.getInitialState(e));let a=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(s.length!==a)throw new U(`RNN Layer has ${a} state(s) but was passed ${s.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let o={training:r},c=aN((h,f)=>{let m=this.cell.call([h].concat(f),o);return[m[0],m.slice(1)]},e,s,this.goBackwards,n,null,this.unroll,this.returnSequences),u=c[0],l=c[1],p=c[2];this.stateful&&this.resetStates(p,r);let d=this.returnSequences?l:u;return this.returnState?[d].concat(p):d})}getInitialState(e){return O(()=>{let t=kt(e.shape);return t=ge(t,[1,2]),t=Xd(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(n=>n>1?Gy(t,[1,n]):t):this.cell.stateSize>1?[Gy(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){let e=super.getConfig(),t={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(t.numConstants=this.numConstants);let n=this.cell.getConfig();return this.getClassName()===ms.className&&(t.cell={className:this.cell.getClassName(),config:n}),Object.assign(Object.assign(Object.assign({},n),e),t)}static fromConfig(e,t,n={}){let r=t.cell,s=Gr(r,n);return new e(Object.assign(t,{cell:s}))}};ms.className="RNN";se.registerClass(ms);var rp=class extends Ue{},Vm=class extends rp{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=da(e.activation==null?this.DEFAULT_ACTIVATION:e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=kc([1,ua([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=kc([1,ua([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=tt(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return O(()=>{if(e=e,e.length!==2)throw new U(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let n=e[1];e=e[0];let r=t.training==null?!1:t.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=pa({ones:()=>rr(e),rate:this.dropout,training:r,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=pa({ones:()=>rr(n),rate:this.recurrentDropout,training:r,dropoutFunc:this.dropoutFunc}));let s,a=this.dropoutMask,o=this.recurrentDropoutMask;a!=null?s=cs(B(e,a),this.kernel.read()):s=cs(e,this.kernel.read()),this.bias!=null&&(s=Yr(s,this.bias.read())),o!=null&&(n=B(n,o));let i=Y(s,cs(n,this.recurrentKernel.read()));return this.activation!=null&&(i=this.activation.apply(i)),[i,i]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:la(this.activation),useBias:this.useBias,kernelInitializer:Nt(this.kernelInitializer),recurrentInitializer:Nt(this.recurrentInitializer),biasInitializer:Nt(this.biasInitializer),kernelRegularizer:ut(this.kernelRegularizer),recurrentRegularizer:ut(this.recurrentRegularizer),biasRegularizer:ut(this.biasRegularizer),activityRegularizer:ut(this.activityRegularizer),kernelConstraint:jt(this.kernelConstraint),recurrentConstraint:jt(this.recurrentConstraint),biasConstraint:jt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},e),t)}};Vm.className="SimpleRNNCell";se.registerClass(Vm);var Lw=class extends ms{constructor(e){e.cell=new Vm(e),super(e)}call(e,t){return O(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,r=t==null?null:t.training,s=t==null?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:s})})}static fromConfig(e,t){return new e(t)}};Lw.className="SimpleRNN";se.registerClass(Lw);var Um=class extends rp{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new U("GRUCell does not support reset_after parameter set to true.");this.units=e.units,Qt(this.units,"units"),this.activation=da(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=da(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=kc([1,ua([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=kc([1,ua([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=tt(e);let t=e[e.length-1];this.kernel=this.addWeight("kernel",[t,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return O(()=>{if(e=e,e.length!==2)throw new U(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training==null?!1:t.training,r=e[1];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=pa({ones:()=>rr(e),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=pa({ones:()=>rr(r),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,a=this.recurrentDropoutMask,o,i,c;0<this.dropout&&this.dropout<1&&(e=B(e,s[0]));let u=cs(e,this.kernel.read());this.useBias&&(u=Yr(u,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(r=B(r,a[0]));let l=this.recurrentKernel.read(),[p,d]=zn(l,[2*this.units,this.units],l.rank-1),h=cs(r,p),[f,m,g]=zn(u,3,u.rank-1),[b,y]=zn(h,2,h.rank-1);o=this.recurrentActivation.apply(Y(f,b)),i=this.recurrentActivation.apply(Y(m,y));let v=cs(B(i,r),d);c=this.activation.apply(Y(g,v));let x=Y(B(o,r),B(Y(1,xt(o)),c));return[x,x]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:la(this.activation),recurrentActivation:la(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Nt(this.kernelInitializer),recurrentInitializer:Nt(this.recurrentInitializer),biasInitializer:Nt(this.biasInitializer),kernelRegularizer:ut(this.kernelRegularizer),recurrentRegularizer:ut(this.recurrentRegularizer),biasRegularizer:ut(this.biasRegularizer),activityRegularizer:ut(this.activityRegularizer),kernelConstraint:jt(this.kernelConstraint),recurrentConstraint:jt(this.recurrentConstraint),biasConstraint:jt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},e),t)}};Um.className="GRUCell";se.registerClass(Um);var zw=class extends ms{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new Um(e),super(e)}call(e,t){return O(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,r=t==null?null:t.training,s=t==null?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:s})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};zw.className="GRU";se.registerClass(zw);var sp=class extends rp{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,Qt(this.units,"units"),this.activation=da(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=da(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=St(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=Tt(e.kernelRegularizer),this.recurrentRegularizer=Tt(e.recurrentRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.kernelConstraint=Kt(e.kernelConstraint),this.recurrentConstraint=Kt(e.recurrentConstraint),this.biasConstraint=Kt(e.biasConstraint),this.dropout=kc([1,ua([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=kc([1,ua([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;e=tt(e);let n=e[e.length-1];this.kernel=this.addWeight("kernel",[n,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let r;if(this.useBias){if(this.unitForgetBias){let s=this.biasInitializer,a=this.units;r=new(t=class extends Pr{apply(i,c){let u=s.apply([a]),l=new $m().apply([a]),p=s.apply([a*2]);return Ak(Ak(u,l),p)}},t.className="CustomInit",t)}else r=this.biasInitializer;this.bias=this.addWeight("bias",[this.units*4],null,r,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return O(()=>{let n=t.training==null?!1:t.training;if(e=e,e.length!==3)throw new U(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let r=e[1],s=e[2];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=pa({ones:()=>rr(e),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=pa({ones:()=>rr(r),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let a=this.dropoutMask,o=this.recurrentDropoutMask,i,c,u,l;0<this.dropout&&this.dropout<1&&(e=B(e,a[0]));let p=cs(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(r=B(r,o[0])),p=Y(p,cs(r,this.recurrentKernel.read())),this.useBias&&(p=Yr(p,this.bias.read()));let[d,h,f,m]=zn(p,4,p.rank-1);i=this.recurrentActivation.apply(d),c=this.recurrentActivation.apply(h),u=Y(B(c,s),B(i,this.activation.apply(f))),l=this.recurrentActivation.apply(m);let g=B(l,this.activation.apply(u));return[g,g,u]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:la(this.activation),recurrentActivation:la(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Nt(this.kernelInitializer),recurrentInitializer:Nt(this.recurrentInitializer),biasInitializer:Nt(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:ut(this.kernelRegularizer),recurrentRegularizer:ut(this.recurrentRegularizer),biasRegularizer:ut(this.biasRegularizer),activityRegularizer:ut(this.activityRegularizer),kernelConstraint:jt(this.kernelConstraint),recurrentConstraint:jt(this.recurrentConstraint),biasConstraint:jt(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},e),t)}};sp.className="LSTMCell";se.registerClass(sp);var Bw=class extends ms{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new sp(e),super(e)}call(e,t){return O(()=>{this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=t==null?null:t.mask,r=t==null?null:t.training,s=t==null?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:s})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};Bw.className="LSTM";se.registerClass(Bw);var Gm=class extends rp{constructor(e){super(e),this.cells=e.cells}get stateSize(){let e=[];for(let t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return O(()=>{e=e;let n=e.slice(1),r=[];for(let o of this.cells.slice().reverse())Array.isArray(o.stateSize)?r.push(n.splice(0,o.stateSize.length)):r.push(n.splice(0,1));r.reverse();let s=[],a;for(let o=0;o<this.cells.length;++o){let i=this.cells[o];n=r[o],o===0?a=[e[0]].concat(n):a=[a[0]].concat(n),a=i.call(a,t),s.push(a.slice(1))}n=[];for(let o of s.slice().reverse())n.push(...o);return[a[0]].concat(n)})}build(e){qy(e)&&(e=e[0]),e=e;let t;this.cells.forEach((n,r)=>{Za(`RNNCell_${r}`,()=>{n.build(e),Array.isArray(n.stateSize)?t=n.stateSize[0]:t=n.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){let e=super.getConfig(),t=s=>({className:s.getClassName(),config:s.getConfig()}),r={cells:this.cells.map(t)};return Object.assign(Object.assign({},e),r)}static fromConfig(e,t,n={}){let r=[];for(let s of t.cells)r.push(Gr(s,n));return new e({cells:r})}get trainableWeights(){if(!this.trainable)return[];let e=[];for(let t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let n of this.cells)t.push(...n.trainableWeights);return t.concat(e)}return e}getWeights(){let e=[];for(let t of this.cells)e.push(...t.weights);return jy(e)}setWeights(e){let t=[];for(let n of this.cells){let r=n.weights.length,s=e.splice(r);for(let a=0;a<n.weights.length;++a)t.push([n.weights[a],s[a]])}lw(t)}};Gm.className="StackedRNNCells";se.registerClass(Gm);function pa(e){let{ones:t,rate:n,training:r=!1,count:s=1,dropoutFunc:a}=e,o=()=>a!=null?a(t(),n):bC(t(),n),i=()=>Zd(o,t,r);return!s||s<=1?Jt(i().clone()):Array(s).fill(void 0).map(i).map(u=>Jt(u.clone()))}var GU=function(e,t){var n={};for(var r in e)Object.prototype.hasOwnProperty.call(e,r)&&t.indexOf(r)<0&&(n[r]=e[r]);if(e!=null&&typeof Object.getOwnPropertySymbols=="function")for(var s=0,r=Object.getOwnPropertySymbols(e);s<r.length;s++)t.indexOf(r[s])<0&&Object.prototype.propertyIsEnumerable.call(e,r[s])&&(n[r[s]]=e[r[s]]);return n},oN=class extends ms{constructor(e){if(e.unroll)throw new Pe("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new Pe("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new zt({ndim:5})]}call(e,t){return O(()=>{if(this.cell.dropoutMask!=null&&(_e(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(_e(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new U("ConvRNN2D cell does not support constants");let n=t==null?null:t.mask,r=t==null?null:t.training,s=t==null?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:s})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return O(()=>{let{stateSize:t}=this.cell,n=e.shape,r=this.computeSingleOutputShape(n),s=[r[0],...r.slice(2)],a=kt(s);return Array.isArray(t)?Array(t.length).fill(a):[a]})}resetStates(e,t=!1){O(()=>{if(!this.stateful)throw new ks("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape,r=this.computeSingleOutputShape(n),s=[r[0],...r.slice(2)];if(n[0]==null)throw new U("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>kt(s)):this.states_=[kt(s)];else if(e==null)_e(this.states_),this.keptStates!=null&&(_e(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>kt(s)):this.states_[0]=kt(s);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new U(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):_e(this.states_);for(let o=0;o<this.states_.length;++o){let i=e[o],c=s;if(!w.arraysEqual(i.shape,c))throw new U(`State ${o} is incompatible with layer ${this.name}: expected shape=${c}, received shape=${i.shape}`);this.states_[o]=i}}this.states_=this.states_.map(o=>Jt(o.clone()))})}computeSingleOutputShape(e){let{dataFormat:t,filters:n,kernelSize:r,padding:s,strides:a,dilationRate:o}=this.cell,i=t==="channelsFirst",c=e[i?3:2],u=e[i?4:3],l=Hr(c,r[0],s,a[0],o[0]),p=Hr(u,r[1],s,a[1],o[1]);return[...e.slice(0,2),...i?[n,l,p]:[l,p,n]]}};oN.className="ConvRNN2D";var Hm=class extends sp{constructor(e){let{filters:t,kernelSize:n,strides:r,padding:s,dataFormat:a,dilationRate:o}=e;super(Object.assign(Object.assign({},e),{units:t})),this.filters=t,Qt(this.filters,"filters"),this.kernelSize=pc(n,2,"kernelSize"),this.kernelSize.forEach(i=>Qt(i,"kernelSize")),this.strides=pc(r||1,2,"strides"),this.strides.forEach(i=>Qt(i,"strides")),this.padding=s||"valid",br(this.padding),this.dataFormat=a||"channelsLast",Pt(this.dataFormat),this.dilationRate=pc(o||1,2,"dilationRate"),this.dilationRate.forEach(i=>Qt(i,"dilationRate"))}build(e){var t;e=tt(e);let n=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[n]==null)throw new U(`The channel dimension of the input should be defined. Found ${e[n]}`);let r=e[n],s=4,a=this.kernelSize.concat([r,this.filters*s]);this.kernel=this.addWeight("kernel",a,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let o=this.kernelSize.concat([this.filters,this.filters*s]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",o,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let i;if(this.unitForgetBias){let c=this.biasInitializer,u=this.filters;i=new(t=class extends Pr{apply(p,d){let h=c.apply([u]),f=Qn([u]),m=c.apply([u*2]);return nw([h,f,m])}},t.className="CustomInit",t)}else i=this.biasInitializer;this.bias=this.addWeight("bias",[this.filters*s],null,i,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return O(()=>{if(e.length!==3)throw new U(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training||!1,r=e[0],s=e[1],a=e[2],o=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=pa({ones:()=>rr(r),rate:this.dropout,training:n,count:o,dropoutFunc:this.dropoutFunc}));let i=this.dropoutMask,c=(Z,J,ee)=>!J||!J[ee]?Z:B(J[ee],Z),u=c(r,i,0),l=c(r,i,1),p=c(r,i,2),d=c(r,i,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=pa({ones:()=>rr(s),rate:this.recurrentDropout,training:n,count:o,dropoutFunc:this.dropoutFunc}));let h=this.recurrentDropoutMask,f=c(s,h,0),m=c(s,h,1),g=c(s,h,2),b=c(s,h,3),y=3,[v,x,k,S]=zn(this.kernel.read(),o,y),[C,E,$,F]=this.useBias?zn(this.bias.read(),o):[null,null,null,null];u=this.inputConv(u,v,C,this.padding),l=this.inputConv(l,x,E,this.padding),p=this.inputConv(p,k,$,this.padding),d=this.inputConv(d,S,F,this.padding);let[A,R,T,L]=zn(this.recurrentKernel.read(),o,y);f=this.recurrentConv(f,A),m=this.recurrentConv(m,R),g=this.recurrentConv(g,T),b=this.recurrentConv(b,L);let V=this.recurrentActivation.apply(Y(u,f)),G=this.recurrentActivation.apply(Y(l,m)),j=Y(B(G,a),B(V,this.activation.apply(Y(p,g)))),H=B(this.recurrentActivation.apply(Y(d,b)),this.activation.apply(j));return[H,H,j]})}getConfig(){let e=super.getConfig(),{units:t}=e,n=GU(e,["units"]),r={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),r)}inputConv(e,t,n,r){let s=Dt(e,t,this.strides,r||"valid",this.dataFormat==="channelsFirst"?"NCHW":"NHWC",this.dilationRate);return n?Yr(s,n,this.dataFormat):s}recurrentConv(e,t){return Dt(e,t,1,"same",this.dataFormat==="channelsFirst"?"NCHW":"NHWC")}};Hm.className="ConvLSTM2DCell";se.registerClass(Hm);var Ww=class extends oN{constructor(e){let t=new Hm(e);super(Object.assign(Object.assign({},e),{cell:t}))}static fromConfig(e,t){return new e(t)}};Ww.className="ConvLSTM2D";se.registerClass(Ww);var qm=class extends Ue{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(this.noiseShape==null)return this.noiseShape;let t=e.shape,n=[];for(let r=0;r<this.noiseShape.length;++r)n.push(this.noiseShape[r]==null?t[r]:this.noiseShape[r]);return n}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);if(0<this.rate&&this.rate<1){let r=t.training==null?!1:t.training,s=this.getNoiseShape(n);return Zd(()=>bC(n,this.rate,s,this.seed),()=>n,r)}return e})}getConfig(){let e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};qm.className="Dropout";se.registerClass(qm);var Vw=class extends qm{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){let t=e.shape;return[t[0],1,t[2]]}};Vw.className="SpatialDropout1D";se.registerClass(Vw);var Uw=class extends Ue{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.batchInputShape==null&&e.inputShape==null&&e.inputDim!=null){let t=null;e.batchSize!=null&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,Qt(this.units,"units"),this.activation=da(e.activation),e.useBias!=null&&(this.useBias=e.useBias),this.kernelInitializer=St(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=St(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Kt(e.kernelConstraint),this.biasConstraint=Kt(e.biasConstraint),this.kernelRegularizer=Tt(e.kernelRegularizer),this.biasRegularizer=Tt(e.biasRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){e=tt(e);let t=e[e.length-1];this.kernel==null&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){e=tt(e);let t=e.slice();return t[t.length-1]=this.units,t}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e),r=lC(this.activation.getClassName()),s;return r!=null?s=cs(n,this.kernel.read(),r,this.bias?this.bias.read():null):(s=cs(n,this.kernel.read()),this.bias!=null&&(s=Yr(s,this.bias.read())),this.activation!=null&&(s=this.activation.apply(s))),s})}getConfig(){let e={units:this.units,activation:la(this.activation),useBias:this.useBias,kernelInitializer:Nt(this.kernelInitializer),biasInitializer:Nt(this.biasInitializer),kernelRegularizer:ut(this.kernelRegularizer),biasRegularizer:ut(this.biasRegularizer),activityRegularizer:ut(this.activityRegularizer),kernelConstraint:jt(this.kernelConstraint),biasConstraint:jt(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};Uw.className="Dense";se.registerClass(Uw);var Gw=class extends Ue{constructor(e){e=e||{},super(e),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=tt(e);for(let t of e.slice(1))if(t==null)throw new U(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],aa(e,1)]}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);if(this.dataFormat==="channelsFirst"&&n.rank>1){let r=[0];for(let s=2;s<n.rank;++s)r.push(s);r.push(1),n=Ee(n,r)}return aV(n)})}getConfig(){let e={};this.dataFormat!=null&&(e.dataFormat=this.dataFormat);let t=super.getConfig();return Object.assign(e,t),e}};Gw.className="Flatten";se.registerClass(Gw);var Hw=class extends Ue{constructor(e){super(e),this.supportsMasking=!0,this.activation=da(e.activation)}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);return this.activation.apply(n)})}getConfig(){let e={activation:la(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};Hw.className="Activation";se.registerClass(Hw);var qw=class extends Ue{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return O(()=>(e=Ce(e),rV(e,this.n)))}getConfig(){let e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};qw.className="RepeatVector";se.registerClass(qw);var jw=class extends Ue{constructor(e){super(e),this.targetShape=e.targetShape;for(let t=0;t<this.targetShape.length;++t)this.isUnknown(this.targetShape[t])&&(this.targetShape[t]=null)}isUnknown(e){return e<0||e==null}fixUnknownDimension(e,t){let n="Total size of new array must be unchanged.",r=t.slice(),s=1,a=null;for(let i=0;i<r.length;++i){let c=r[i];if(this.isUnknown(c))if(a===null)a=i;else throw new U("Can only specifiy one unknown dimension.");else s*=c}let o=aa(e);if(a!==null){if(s===0||o%s!==0)throw new U(n);r[a]=o/s}else if(o!==s)throw new U(n);return r}computeOutputShape(e){let t=!1;for(let n=0;n<e.length;++n)if(this.isUnknown(e[n])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e),r=n.shape,s=r.slice(0,1).concat(this.fixUnknownDimension(r.slice(1),this.targetShape));return W(n,s)})}getConfig(){let e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};jw.className="Reshape";se.registerClass(jw);var Kw=class extends Ue{constructor(e){if(super(e),e.dims==null)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);let t=qr(1,e.dims.length+1);if(!w.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new zt({ndim:this.dims.length+1})]}computeOutputShape(e){e=tt(e);let t=e.slice();return this.dims.forEach((n,r)=>{t[r+1]=e[n]}),t}call(e,t){return Ee(Ce(e),this.dimsIncludingBatch)}getConfig(){let e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};Kw.className="Permute";se.registerClass(Kw);var Xw=class extends Ue{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null?this.maskValue=e.maskValue==null?0:e.maskValue:this.maskValue=0}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){let n=Ce(e),r=-1;return Ql(co(n,this.maskValue),r)}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e),r=-1,s=!0,a=Ql(co(n,this.maskValue),r,s);return B(n,ce(a,n.dtype))})}};Xw.className="Masking";se.registerClass(Xw);var Yw=class extends Ue{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",e.batchInputShape==null&&e.inputShape==null){let t=null;e.batchSize!=null&&(t=e.batchSize),e.inputLength==null?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(vt(e.inputLength))}this.inputDim=e.inputDim,Qt(this.inputDim,"inputDim"),this.outputDim=e.outputDim,Qt(this.outputDim,"outputDim"),this.embeddingsInitializer=St(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=Tt(e.embeddingsRegularizer),this.activityRegularizer=Tt(e.activityRegularizer),this.embeddingsConstraint=Kt(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return O(()=>this.maskZero?(e=Ce(e),co(e,qe(e))):null)}computeOutputShape(e){if(e=tt(e),this.inputLength==null)return[...e,this.outputDim];let t=vt(this.inputLength);if(t.length!==e.length-1)throw new U(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let n=0;for(let r=0;r<t.length;++r){let s=t[r],a=e[r+1];if(s!=null&&a!=null&&s!==a)throw new U(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);s==null&&(t[n]=a),n++}}return[e[0],...t,this.outputDim]}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);n.dtype!=="int32"&&(n=vi(n,"int32"));let r=gC(this.embeddings.read(),W(n,[n.size]));return W(r,tt(this.computeOutputShape(n.shape)))})}getConfig(){let e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Nt(this.embeddingsInitializer),embeddingsRegularizer:ut(this.embeddingsRegularizer),activityRegularizer:ut(this.activityRegularizer),embeddingsConstraint:jt(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};Yw.className="Embedding";se.registerClass(Yw);var wi=class extends Ue{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new Pe}computeElementwiseOpOutputShape(e,t){if(e==null||t==null)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(t.length===0)return e;let n=e.slice(0,e.length-t.length);for(let r=0;r<t.length;++r){let s=e[e.length-t.length+r],a=t[r];if(s==null||a==null||s<0||a<0)n.push(null);else if(s===1)n.push(a);else if(a===1)n.push(s);else{if(s!==a)throw new U("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));n.push(s)}}return n}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[tt(e)]),e=e,e.length<2)throw new U(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(let s of e)s!=null&&s[0]!==null&&t.push(s[0]);if(t=sa(t),t.length>1)throw new U(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let n=e[0]==null?null:e[0].slice(1);for(let s=1;s<e.length;++s){let a=e[s]==null?null:e[s].slice(1);n=this.computeElementwiseOpOutputShape(n,a)}let r=e.map(s=>s.length);e.indexOf(null)===-1&&sa(r).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return O(()=>{if(e=e,this.reshapeRequired){let n=[],r=e.map(s=>s.rank);if(r.indexOf(null)===-1){let s=ua(r);for(let a of e){let o=a.rank;for(let i=0;i<s-o;++i)a=Xd(a,1);n.push(a)}return this.mergeFunction(n)}else{let s=!1;for(let i of e){let c=i.rank;if(c==null){let u=i.shape,l=u[0],p=u.slice(1).concat([l]),d=W(i,[l].concat(aa(u.slice(1))));d=Ee(d,[1,0]),d=W(d,p),n.push(d),s=!0}else if(c>1){let u=qr(1,c).concat([0]);n.push(Ee(i,u)),s=!0}else n.push(i)}let a=this.mergeFunction(n),o=a.rank;if(s){if(o==null){let i=a.shape,c=i.length,u=i[c-1],l=[u].concat(i.slice(0,i.length-1));a=W(Ee(W(a,[-1,u]),[1,0]),l)}else if(o>1){let i=[o-1].concat(qr(0,o-1));a=Ee(a,i)}}return a}}else return this.mergeFunction(e)})}computeOutputShape(e){e=e;let t;e[0]==null?t=null:t=e[0].slice(1);for(let r=1;r<e.length;++r){let s=e[r]==null?null:e[r].slice(1);t=this.computeElementwiseOpOutputShape(t,s)}let n=[];for(let r of e)r!=null&&r[0]!==null&&n.push(r[0]);return n=sa(n),n.length===1?t=n.concat(t):t=[null].concat(t),t}computeMask(e,t){return O(()=>{if(t==null)return null;if(!Array.isArray(t))throw new U("`mask` should be an Array");if(!Array.isArray(e))throw new U("`inputs` should be an Array");if(t.length!==e.length)throw new U(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(r=>r==null))return null;t=t.map(r=>r==null?r:Zt(r,0));let n=t[0];for(let r=1;r<t.length-1;++r)n=Dr(n,t[r]);return n})}},Zw=class extends wi{constructor(e){super(e)}mergeFunction(e){return O(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return t})}};Zw.className="Add";se.registerClass(Zw);var Jw=class extends wi{constructor(e){super(e)}mergeFunction(e){return O(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=B(t,e[n]);return t})}};Jw.className="Multiply";se.registerClass(Jw);var Qw=class extends wi{constructor(e){super(e)}mergeFunction(e){return O(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=Y(t,e[n]);return B(1/e.length,t)})}};Qw.className="Average";se.registerClass(Qw);var e0=class extends wi{constructor(e){super(e)}mergeFunction(e){return O(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=hs(t,e[n]);return t})}};e0.className="Maximum";se.registerClass(e0);var t0=class extends wi{constructor(e){super(e)}mergeFunction(e){return O(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=Vu(t,e[n]);return t})}};t0.className="Minimum";se.registerClass(t0);var n0=class extends wi{constructor(e){super(e),this.DEFAULT_AXIS=-1,e==null&&(e={}),this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!(Array.isArray(e)&&Array.isArray(e[0]))||e.length===1)throw new U("A `Concatenate` layer should be called on a list of at least 2 inputs");e=e;let t=!0;for(let r of e)if(r!=null){t=!1;break}if(t)return;let n=[];for(let r=0;r<e.length;++r){let s=e[r].slice();s.splice(this.axis,1);let a=!1;for(let o of n)if(w.arraysEqual(o,s)){a=!0;break}a||n.push(s)}if(n.length>1)throw new U("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return O(()=>nw(e,this.axis))}computeOutputShape(e){if(!(Array.isArray(e)&&Array.isArray(e[0])))throw new U("A `Concatenate` layer should be called on a list of inputs.");let t=e,n=t[0].slice(),r=this.axis<0?n.length+this.axis:this.axis;for(let s of t.slice(1)){if(n[r]==null||s[r]==null){n[r]=null;break}n[r]+=s[r]}return n}computeMask(e,t){if(t==null)return null;if(!Array.isArray(t))throw new U("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new U("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new U(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return O(()=>{let n=!0;if(t.forEach(a=>{if(a!=null){n=!1;return}}),n)return null;let r=[];for(let a=0;a<e.length;++a)t[a]==null?r.push(ce(rr(e[a]),"bool")):t[a].rank<e[a].rank?r.push(Zt(t[a],-1)):r.push(t[a]);let s=Je(r,this.axis);return Kf(s,-1,!1)})}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};n0.className="Concatenate";se.registerClass(n0);function Al(e,t){for(;e<0;)e+=t;return e}function HU(e,t,n){if(e.shape.length>3||t.shape.length>3)throw new Pe("batchDot is not implemented for tensors of 4D or higher rank yet");if(w.assert(e.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${e.shape.length}`),w.assert(e.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof n=="number"&&(n=[n,n]),e.dtype==="complex64"||t.dtype==="complex64")throw new Pe("batchDot is not implemented for complex64-type Tensors yet.");let r=e.shape.length,s=t.shape.length;n==null&&(n=[r-1,s-2]);let a=n;return O(()=>{let o;if(r>s){o=r-s;let c=[];for(let u=0;u<o;++u)c.push(1);t=W(t,t.shape.concat(c))}else if(s>r){o=s-r;let c=[];for(let u=0;u<o;++u)c.push(1);e=W(e,e.shape.concat(c))}else o=0;let i;if(e.shape.length===2&&t.shape.length===2)a[0]===a[1]?i=ge(B(e,t),a[0]):i=ge(B(Ee(e,[1,0]),t),a[1]);else{let c=a[0]!==e.shape.length-1,u=a[1]===t.shape.length-1;i=De(e,t,c,u)}if(o>0){let c;r>s?c=r+s-3:c=r-1;let u=[];for(let l=c;l<c+o;++l)u.push(l);i=Na(i,u)}return i.shape.length===1&&(i=Zt(i,1)),i})}var r0=class extends wi{constructor(e){super(e),this.axes=e.axes,this.normalize=e.normalize==null?!1:e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){w.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0],n=e[1];if(t.length>3||n.length>3)throw new Pe("Dot layer does not support tensors of 4D or higher rank yet.");let r=this.interpretAxes(t,n);if(t[r[0]]!==n[r[1]])throw new U(`Dimension incompatibility: ${t[r[0]]} !== ${n[r[1]]}`)}mergeFunction(e){if(e.length!==2)throw new U(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t=e[0],n=e[1],r;return Array.isArray(this.axes)?r=this.axes.map((s,a)=>Al(s,e[a].shape.length)):r=[Al(this.axes,t.shape.length),Al(this.axes,n.shape.length)],this.normalize&&(t=Hh(t,r[0]),n=Hh(n,r[1])),HU(t,n,r)}interpretAxes(e,t){let n;return Array.isArray(this.axes)?n=this.axes:n=[Al(this.axes,e.length),Al(this.axes,t.length)],n}computeOutputShape(e){w.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0].slice(),n=e[1].slice();if(t.length>3||n.length>3)throw new Pe("Dot layer does not support tensors of 4D or higher rank yet.");let r=this.interpretAxes(t,n);t.splice(r[0],1),n.splice(r[1],1),n.splice(0,1);let s=t.concat(n);return s.length===1&&s.push(1),s}computeMask(e,t){return null}getConfig(){let e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};r0.className="Dot";se.registerClass(r0);var s0=class extends Ue{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);return Zd(()=>Y(Am(n.shape,0,this.stddev),n),()=>n,t.training||!1)})}};s0.className="GaussianNoise";se.registerClass(s0);var a0=class extends Ue{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return O(()=>{this.invokeCallHook(e,t);let n=Ce(e);return this.rate>0&&this.rate<1?Zd(()=>{let s=Math.sqrt(this.rate/(1-this.rate));return B(n,Am(n.shape,1,s))},()=>n,t.training||!1):n})}};a0.className="GaussianDropout";se.registerClass(a0);var o0=class extends Ue{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||Ce(e).shape}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return O(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(e);return Zd(()=>{let s=Ce(e),a=1.6732632423543772,o=1.0507009873554805,i=-a*o,c=Ds(Uu(n),this.rate);c=vi(c,"float32");let u=((1-this.rate)*(1+this.rate*i**2))**-.5,l=-u*i*this.rate,p=Y(B(s,c),B(Y(c,-1),i));return Y(B(p,u),l)},()=>Ce(e),t.training||!1)}return e})}};o0.className="AlphaDropout";se.registerClass(o0);function rd(e,t,n,r,s,a=.001){let o;if(e.rank===2)o=ax(e,t,n,r,s,a);else if(e.rank===3)o=ox(e,t,n,r,s,a);else if(e.rank===4)o=ix(e,t,n,r,s,a);else throw new Pe(`batchNormalization is not implemented for array of rank ${e.rank} yet`);return o}function qU(e,t,n,r,s=.001){return O(()=>{let a=Wd(e,r),o=a.mean,i=a.variance;return[rd(e,o,i,n,t,s),o,i]})}function jU(e,t,n,r,s=.001){return O(()=>{let a=Wd(e,r),o=a.mean,i=a.variance,c=[];for(let f of qr(0,e.rank))r.indexOf(f)!==-1?c.push(1):c.push(e.shape[f]);let u=W(o,c),l=W(i,c),p=t==null?null:W(t,c),d=n==null?null:W(n,c);return[rd(e,u,l,d,p,s),o,i]})}function KU(e,t,n,r,s=.001){return w.arraysEqual(r.slice().sort(),qr(0,e.rank-1))?qU(e,t,n,r,s):jU(e,t,n,r,s)}var i0=class extends Ue{constructor(e){e==null&&(e={}),super(e),this.supportsMasking=!0,this.axis=e.axis==null?-1:e.axis,this.momentum=e.momentum==null?.99:e.momentum,this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.movingMeanInitializer=St(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=St(e.movingVarianceInitializer||"ones"),this.betaConstraint=Kt(e.betaConstraint),this.gammaConstraint=Kt(e.gammaConstraint),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer)}build(e){e=tt(e);let t=this.axis>=0?this.axis:this.axis+e.length,n=e[t];if(n==null)throw new U(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new zt({ndim:e.length,axes:{[t]:n}})];let r=[n];this.scale&&(this.gamma=this.addWeight("gamma",r,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",r,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",r,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",r,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return O(()=>{let n=t.training==null?!1:t.training,r=Ce(e),s=r.shape,a=s.length,o=qr(0,a),i=this.axis>=0?this.axis:this.axis+a;o.splice(i,1);let c=uo(1,a);c[i]=s[i];let u=o.slice();u.sort();let l=!w.arraysEqual(u,qr(0,a).slice(0,a-1)),p=()=>{if(l){let b=W(this.movingMean.read(),c),y=W(this.movingVariance.read(),c),v=this.center?W(this.beta.read(),c):null,x=this.scale?W(this.gamma.read(),c):null;return rd(r,b,y,v,x,this.epsilon)}else return rd(r,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!n)return p();let[d,h,f]=KU(r,this.gamma.read(),this.beta.read(),o,this.epsilon),m=(b,y,v)=>{O(()=>{let x=1-v,k=b.read(),S=B(de(k,y),x);b.write(de(k,S))})};return(()=>{m(this.movingMean,h,this.momentum),m(this.movingVariance,f,this.momentum)})(),d})}getConfig(){let e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Nt(this.betaInitializer),gammaInitializer:Nt(this.gammaInitializer),movingMeanInitializer:Nt(this.movingMeanInitializer),movingVarianceInitializer:Nt(this.movingVarianceInitializer),betaRegularizer:ut(this.betaRegularizer),gammaRegularizer:ut(this.gammaRegularizer),betaConstraint:jt(this.betaConstraint),gammaConstraint:jt(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};i0.className="BatchNormalization";se.registerClass(i0);var c0=class extends Ue{constructor(e){if(e==null&&(e={}),super(e),this.axis=e.axis==null?-1:e.axis,typeof this.axis=="number"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let t of this.axis)if(!Number.isInteger(t))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=St(e.betaInitializer||"zeros"),this.gammaInitializer=St(e.gammaInitializer||"ones"),this.betaRegularizer=Tt(e.betaRegularizer),this.gammaRegularizer=Tt(e.gammaRegularizer),this.supportsMasking=!0}build(e){e=tt(e);let t=e.length;typeof this.axis=="number"&&(this.axis=[this.axis]);for(let s=0;s<this.axis.length;++s)this.axis[s]<0&&(this.axis[s]+=t);for(let s of this.axis)if(s<0||s>=t)throw new Error(`Invalid axis: ${s}`);if(this.axis.length!==sa(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(s=>e[s]),r=!0;this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,r):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,r):this.beta=null,this.built=!0}call(e,t){let n=Ce(e),r=n.shape,s=r.length;return O(()=>{let{mean:o,variance:i}=Wd(n,this.axis,!0),c=uo(1,s);for(let f of this.axis)c[f]=r[f];let u=f=>f!=null&&f.shape.length!==s?W(f,c):f,l=this.scale?u(this.gamma.read()):null,p=this.center?u(this.beta.read()):null,d=[],h=[];for(let f=0;f<s;++f)this.axis.indexOf(f)!==-1?(d.push(r[f]),h.push(1)):(d.push(1),h.push(r[f]));return o=Ln(o,d),i=Ln(i,d),l!=null&&(l=Ln(l,h)),p!=null&&(p=Ln(p,h)),rd(n,o,i,p,l,this.epsilon)})}getConfig(){let e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Nt(this.betaInitializer),gammaInitializer:Nt(this.gammaInitializer),betaRegularizer:ut(this.betaRegularizer),gammaRegularizer:ut(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};c0.className="LayerNormalization";se.registerClass(c0);function XU(e,t,n){return O(()=>{if(e.rank!==4)throw new U(`temporalPadding expects input tensor to be 4-D, but received a ${e.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new U("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(n==null&&(n=jr()),n!=="channelsLast"&&n!=="channelsFirst")throw new U(`Unknown data format: ${n}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let r;return n==="channelsFirst"?r=[[0,0],[0,0],t[0],t[1]]:r=[[0,0],t[0],t[1],[0,0]],gr(e,r)})}var u0=class extends Ue{constructor(e){if(e==null&&(e={}),super(e),this.dataFormat=e.dataFormat==null?jr():e.dataFormat,e.padding==null)this.padding=[[1,1],[1,1]];else if(typeof e.padding=="number")this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,e.padding.length!==2)throw new U(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,n;if(typeof e.padding[0]=="number")t=[e.padding[0],e.padding[0]],n=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,e.padding[0].length!==2)throw new U(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],e.padding[1].length!==2)throw new U(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);n=e.padding[1]}this.padding=[t,n]}this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=tt(e);let t,n;return this.dataFormat==="channelsFirst"?(e[2]!=null&&e[2]>=0?t=e[2]+this.padding[0][0]+this.padding[0][1]:t=null,e[3]!=null&&e[3]>=0?n=e[3]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],e[1],t,n]):(e[1]!=null&&e[1]>=0?t=e[1]+this.padding[0][0]+this.padding[0][1]:t=null,e[2]!=null&&e[2]>=0?n=e[2]+this.padding[1][0]+this.padding[1][1]:n=null,[e[0],t,n,e[3]])}call(e,t){return O(()=>XU(Ce(e),this.padding,this.dataFormat))}getConfig(){let e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};u0.className="ZeroPadding2D";se.registerClass(u0);function jm(e,t,n,r,s,a){return O(()=>{Pt(s),pC(a),br(r),n==null&&(n=[1,1]),r==null&&(r="valid"),s==null&&(s=jr()),a==null&&(a="max"),e=Aw(e,s);let o,i=r==="same"?"same":"valid";return a==="max"?o=Rt(e,t,n,i):o=mr(e,t,n,i),s==="channelsFirst"&&(o=Ee(o,[0,3,1,2])),o})}function iN(e,t,n,r,s,a){return O(()=>{Pt(s),pC(a),br(r),n==null&&(n=[1,1,1]),r==null&&(r="valid"),s==null&&(s=jr()),a==null&&(a="max"),e=nN(e,s);let o,i=r==="same"?"same":"valid";return a==="max"?o=Dx(e,t,n,i):o=sx(e,t,n,i),s==="channelsFirst"&&(o=Ee(o,[0,4,1,2,3])),o})}var cN=class extends Ue{constructor(e){if(e.poolSize==null&&(e.poolSize=2),super(e),typeof e.poolSize=="number")this.poolSize=[e.poolSize];else if(Array.isArray(e.poolSize)&&e.poolSize.length===1&&typeof e.poolSize[0]=="number")this.poolSize=e.poolSize;else throw new U(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.poolSize)}`);if(Qt(this.poolSize,"poolSize"),e.strides==null)this.strides=this.poolSize;else if(typeof e.strides=="number")this.strides=[e.strides];else if(Array.isArray(e.strides)&&e.strides.length===1&&typeof e.strides[0]=="number")this.strides=e.strides;else throw new U(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.strides)}`);Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,br(this.padding),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){e=tt(e);let t=Hr(e[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return O(()=>{this.invokeCallHook(e,t),e=Xd(Ce(e),2);let n=this.poolingFunction(Ce(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return Na(n,[2])})}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},l0=class extends cN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),jm(e,t,n,r,s,"max")}};l0.className="MaxPooling1D";se.registerClass(l0);var d0=class extends cN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),jm(e,t,n,r,s,"avg")}};d0.className="AveragePooling1D";se.registerClass(d0);var uN=class extends Ue{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==2)throw new U(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Pt(this.dataFormat),br(this.padding),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2];return t=Hr(t,this.poolSize[0],this.padding,this.strides[0]),n=Hr(n,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n]:[e[0],t,n,e[3]]}call(e,t){return O(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ce(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},p0=class extends uN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),jm(e,t,n,r,s,"max")}};p0.className="MaxPooling2D";se.registerClass(p0);var h0=class extends uN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),jm(e,t,n,r,s,"avg")}};h0.className="AveragePooling2D";se.registerClass(h0);var lN=class extends Ue{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==3)throw new U(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];Qt(this.poolSize,"poolSize"),Qt(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Pt(this.dataFormat),br(this.padding),this.inputSpec=[new zt({ndim:5})]}computeOutputShape(e){e=tt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],n=this.dataFormat==="channelsFirst"?e[3]:e[2],r=this.dataFormat==="channelsFirst"?e[4]:e[3];return t=Hr(t,this.poolSize[0],this.padding,this.strides[0]),n=Hr(n,this.poolSize[1],this.padding,this.strides[1]),r=Hr(r,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,n,r]:[e[0],t,n,r,e[4]]}call(e,t){return O(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ce(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},f0=class extends lN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),iN(e,t,n,r,s,"max")}};f0.className="MaxPooling3D";se.registerClass(f0);var m0=class extends lN{constructor(e){super(e)}poolingFunction(e,t,n,r,s){return Pt(s),br(r),iN(e,t,n,r,s,"avg")}};m0.className="AveragePooling3D";se.registerClass(m0);var dN=class extends Ue{constructor(e){super(e),this.inputSpec=[new zt({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new Pe}},g0=class extends dN{constructor(e){super(e||{})}call(e,t){return O(()=>{let n=Ce(e);return Ct(n,1)})}};g0.className="GlobalAveragePooling1D";se.registerClass(g0);var b0=class extends dN{constructor(e){super(e||{})}call(e,t){return O(()=>{let n=Ce(e);return pr(n,1)})}};b0.className="GlobalMaxPooling1D";se.registerClass(b0);var pN=class extends Ue{constructor(e){super(e),this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Pt(this.dataFormat),this.inputSpec=[new zt({ndim:4})]}computeOutputShape(e){return e=e,this.dataFormat==="channelsLast"?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new Pe}getConfig(){let e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},y0=class extends pN{call(e,t){return O(()=>{let n=Ce(e);return this.dataFormat==="channelsLast"?Ct(n,[1,2]):Ct(n,[2,3])})}};y0.className="GlobalAveragePooling2D";se.registerClass(y0);var v0=class extends pN{call(e,t){return O(()=>{let n=Ce(e);return this.dataFormat==="channelsLast"?pr(n,[1,2]):pr(n,[2,3])})}};v0.className="GlobalMaxPooling2D";se.registerClass(v0);var hN=class extends Ue{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(e){this.layer!=null&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){let e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,n={}){let r=t.layer,s=Gr(r,n);delete t.layer;let a={layer:s};return Object.assign(a,t),new e(a)}},x0=class extends hN{constructor(e){super(e),this.supportsMasking=!0}build(e){if(e=tt(e),e.length<3)throw new U(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(e)}`);this.inputSpec=[{shape:e}];let t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){e=tt(e);let t=[e[0]].concat(e.slice(2)),n=this.layer.computeOutputShape(t),r=e[1];return[n[0],r].concat(n.slice(1))}call(e,t){return O(()=>(e=Ce(e),aN((a,o)=>[Ce(this.layer.call(a,t)),[]],e,[],!1,null,null,!1,!0)[1]))}};x0.className="TimeDistributed";se.registerClass(x0);function YU(e){yi(J4,"BidirectionalMergeMode",e)}var ZU="concat",w0=class extends hN{constructor(e){super(e);let t=e.layer.getConfig(),n={};n.className=e.layer.getClassName(),n.config=t,this.forwardLayer=Gr(n),t.goBackwards=t.goBackwards!==!0;let r={};if(r.className=e.layer.getClassName(),r.config=t,this.backwardLayer=Gr(r),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=e.mergeMode===void 0?ZU:e.mergeMode,YU(this.mergeMode),e.weights)throw new Pe("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,this.forwardLayer!=null&&(this.forwardLayer.trainable=e),this.backwardLayer!=null&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){let t=e.length,n=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,n)),this.backwardLayer.setWeights(e.slice(n))}computeOutputShape(e){let t=this.forwardLayer.computeOutputShape(e);Array.isArray(t)&&Array.isArray(t[0])||(t=[t]),t=t;let n,r,s;return this.returnState&&(s=t.slice(1)),n=t[0],n=n,this.mergeMode==="concat"?(n[n.length-1]*=2,r=[n]):this.mergeMode==null?r=[n,n.slice()]:r=[n],this.returnState?this.mergeMode==null?r.concat(s).concat(s.slice()):[n].concat(s).concat(s.slice()):Mn(r)}apply(e,t){let n=t==null?null:t.initialState,r=t==null?null:t.constants;t==null&&(t={});let s=sN(e,n,r,this.numConstants);if(e=s.inputs,n=s.initialState,r=s.constants,Array.isArray(e)&&(n=e.slice(1),e=e[0]),(n==null||n.length===0)&&r==null)return super.apply(e,t);let a=[],o=[];if(n!=null){let c=n.length;if(c%2>0)throw new U("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=n,a.push(...n);let u=n.map(l=>new zt({shape:l.shape}));this.forwardLayer.stateSpec=u.slice(0,c/2),this.backwardLayer.stateSpec=u.slice(c/2),o.push(...u)}if(r!=null)throw new Pe("Support for constants in Bidirectional layers is not implemented yet.");let i=a[0]instanceof Vr;for(let c of a)if(c instanceof Vr!==i)throw new U("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(i){let c=[e].concat(a),u=this.inputSpec.concat(o),l=this.inputSpec;this.inputSpec=u;let p=super.apply(c,t);return this.inputSpec=l,p}else return super.apply(e,t)}call(e,t){return O(()=>{let n=t.initialState,r,s;if(n==null)r=this.forwardLayer.call(e,t),s=this.backwardLayer.call(e,t);else{let i=n.slice(0,n.length/2),c=n.slice(n.length/2);r=this.forwardLayer.call(e,Object.assign(t,{initialState:i})),s=this.backwardLayer.call(e,Object.assign(t,{initialState:c}))}let a;this.returnState&&(Array.isArray(r)&&(a=r.slice(1).concat(s.slice(1))),r=r[0],s=s[0]),this.returnSequences&&(s=fr(s,1));let o;return this.mergeMode==="concat"?o=nw([r,s]):this.mergeMode==="sum"?o=Y(r,s):this.mergeMode==="ave"?o=B(.5,Y(r,s)):this.mergeMode==="mul"?o=B(r,s):this.mergeMode==null&&(o=[r,s]),this.returnState?this.mergeMode==null?o.concat(a):[o].concat(a):o})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Za(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Za(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){Array.isArray(t)&&(t=t[0]);let n;if(this.returnSequences?this.mergeMode==null?n=[t,t]:n=t:this.mergeMode==null?n=[null,null]:n=null,this.returnState){let s=this.forwardLayer.states.map(a=>null);return Array.isArray(n)?n.concat(s).concat(s):[n].concat(s).concat(s)}else return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(e),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){let e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){let n=Gr(t.layer);if(delete t.layer,t.numConstants!=null)throw new Pe("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let r=t;return r.layer=n,new e(r)}};w0.className="Bidirectional";se.registerClass(w0);var I0=class extends Ue{constructor(e){super(e),this.scale=e.scale,e.offset?this.offset=e.offset:this.offset=0}getConfig(){let e={scale:this.scale,offset:this.offset},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return O(()=>(e=Ce(e),e.dtype!=="float32"&&(e=vi(e,"float32")),Y(B(e,this.scale),this.offset)))}};I0.className="Rescaling";se.registerClass(I0);var JU=["bilinear","nearest"],Kk=new Set(JU),k0=class extends Ue{constructor(e){if(super(e),this.height=e.height,this.width=e.width,e.interpolation)if(Kk.has(e.interpolation))this.interpolation=e.interpolation;else throw new U(`Invalid interpolation parameter: ${e.interpolation} is not implemented`);else this.interpolation="bilinear";this.cropToAspectRatio=Boolean(e.cropToAspectRatio)}computeOutputShape(e){e=tt(e);let t=e[2];return[this.height,this.width,t]}getConfig(){let e={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return O(()=>{let n=[this.height,this.width];if(this.interpolation==="bilinear")return Br.resizeBilinear(e,n,!this.cropToAspectRatio);if(this.interpolation==="nearest")return Br.resizeNearestNeighbor(e,n,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...Kk]} are supported`)})}};k0.className="Resizing";se.registerClass(k0);function QU(e,t,n,r){let s=Ce(e);if(s.dtype!=="int32"&&(s=vi(s,"int32")),t==="int")return s;let a=s.shape;if(s.rank===0&&(s=Zt(s,-1)),t==="oneHot"&&s.shape[s.shape.length-1]!==1&&(s=Zt(s,-1)),s.rank>2)throw new U(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${a} which would result in output rank ${s.rank}.`);let o=["multiHot","oneHot"].includes(t),i=s,c;if(typeof r!="undefined"&&t==="count"?c=Bh(i,r,n,o):c=Bh(i,[],n,o),t!=="tfIdf")return c;if(r)return B(c,r);throw new U("When outputMode is 'tfIdf', weights must be provided.")}var S0=class extends Ue{constructor(e){super(e),this.numTokens=e.numTokens,e.outputMode?this.outputMode=e.outputMode:this.outputMode="multiHot"}getConfig(){let e={numTokens:this.numTokens,outputMode:this.outputMode},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){return e=tt(e),e==null?[this.numTokens]:this.outputMode==="oneHot"&&e[e.length-1]!==1?(e.push(this.numTokens),e):(e[e.length-1]=this.numTokens,e)}call(e,t){return O(()=>{e=Ce(e),e.dtype!=="int32"&&(e=vi(e,"int32"));let n;if(typeof t.countWeights!="undefined"){if(this.outputMode!=="count")throw new U(`countWeights is not used when outputMode !== count.
+              Received countWeights=${t.countWeights}`);n=Ce(t.countWeights)}let r=pr(e),s=yc(e),a=An(this.numTokens,r).bufferSync().get(0),o=Ds(s,0).bufferSync().get(0);if(!(a&&o))throw new U(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return QU(e,this.outputMode,this.numTokens,n)})}};S0.className="CategoryEncoding";se.registerClass(S0);function eG(e){return new Hu(e)}function tG(e){return new Nw(e)}function nG(e){return new Sw(e)}function rG(e){return new Tw(e)}function sG(e){return new Cw(e)}function aG(e){return new Ew(e)}function oG(e){return new _w(e)}function iG(e){return new Wm(e)}function cG(e){return new tp(e)}function uG(e){return new Dw(e)}function lG(e){return new np(e)}function dG(e){return new Fw(e)}function pG(e){return new Rw(e)}function hG(e){return new Pw(e)}function fG(e){return new Ow(e)}function mG(e){return new Mw(e)}function gG(e){return new Hw(e)}function bG(e){return new Uw(e)}function yG(e){return new qm(e)}function vG(e){return new Vw(e)}function xG(e){return new Gw(e)}function wG(e){return new qw(e)}function IG(e){return new jw(e)}function kG(e){return new Kw(e)}function SG(e){return new Yw(e)}function TG(e){return new Zw(e)}function CG(e){return new Qw(e)}function NG(e){return new n0(e)}function _G(e){return new e0(e)}function EG(e){return new t0(e)}function AG(e){return new Jw(e)}function $G(e){return new r0(e)}function DG(e){return new i0(e)}function FG(e){return new c0(e)}function RG(e){return new u0(e)}function T0(e){return new d0(e)}function PG(e){return T0(e)}function OG(e){return T0(e)}function C0(e){return new h0(e)}function MG(e){return C0(e)}function LG(e){return C0(e)}function N0(e){return new m0(e)}function zG(e){return N0(e)}function BG(e){return N0(e)}function WG(e){return new g0(e)}function VG(e){return new y0(e)}function fN(e){return new b0(e)}function mN(e){return new v0(e)}function gN(e){return new l0(e)}function bN(e){return new p0(e)}function UG(e){return new f0(e)}function GG(e){return new zw(e)}function HG(e){return new Um(e)}function qG(e){return new Bw(e)}function jG(e){return new sp(e)}function KG(e){return new Lw(e)}function XG(e){return new Vm(e)}function YG(e){return new Ww(e)}function ZG(e){return new Hm(e)}function JG(e){return new ms(e)}function QG(e){return new Gm(e)}function eH(e){return new w0(e)}function tH(e){return new x0(e)}var nH=fN,rH=mN,sH=gN,aH=bN;function oH(e){return new s0(e)}function iH(e){return new a0(e)}function cH(e){return new o0(e)}function uH(e){return new Xw(e)}function lH(e){return new I0(e)}function dH(e){return new k0(e)}function pH(e){return new S0(e)}var yN={};Ae(yN,{MAPE:()=>kH,MSE:()=>CH,binaryAccuracy:()=>hH,binaryCrossentropy:()=>fH,categoricalAccuracy:()=>gH,categoricalCrossentropy:()=>bH,cosineProximity:()=>xH,mape:()=>SH,meanAbsoluteError:()=>wH,meanAbsolutePercentageError:()=>IH,meanSquaredError:()=>TH,mse:()=>NH,precision:()=>yH,recall:()=>vH,sparseCategoricalAccuracy:()=>mH});function hH(e,t){return bw(e,t)}function fH(e,t){return DC(e,t)}function mH(e,t){return FC(e,t)}function gH(e,t){return yw(e,t)}function bH(e,t){return vw(e,t)}function yH(e,t){return $C(e,t)}function vH(e,t){return nU(e,t)}function xH(e,t){return gw(e,t)}function wH(e,t){return zm(e,t)}function IH(e,t){return qu(e,t)}function kH(e,t){return qu(e,t)}function SH(e,t){return qu(e,t)}function TH(e,t){return xi(e,t)}function CH(e,t){return xi(e,t)}function NH(e,t){return xi(e,t)}var vN={};Ae(vN,{modelFromJSON:()=>DU});var xN={};Ae(xN,{l1:()=>EH,l1l2:()=>_H,l2:()=>AH});function _H(e){return new Qd(e)}function EH(e){return zU(e)}function AH(e){return BU(e)}var wN=class extends Sc{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Es))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function dh(e,t){return e<t}function Xk(e,t){return e>t}var IN=class extends wN{constructor(e){if(super(),e==null&&(e={}),e.restoreBestWeights)throw new Pe("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,["auto","min","max"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),this.mode==="min"?this.monitorFunc=dh:this.mode==="max"?this.monitorFunc=Xk:this.monitor.indexOf("acc")!==-1?this.monitorFunc=Xk:this.monitorFunc=dh,this.monitorFunc===dh&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===dh?1/0:-1/0}async onEpochEnd(e,t){await Js(t);let n=this.getMonitorValue(t);n!=null&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){e==null&&(e={});let t=e[this.monitor];return t==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};function $H(e){return new IN(e)}var DH={earlyStopping:$H},FH=q();FH.registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")});var Nr;(function(e){e[e.DT_INVALID=0]="DT_INVALID",e[e.DT_FLOAT=1]="DT_FLOAT",e[e.DT_DOUBLE=2]="DT_DOUBLE",e[e.DT_INT32=3]="DT_INT32",e[e.DT_UINT8=4]="DT_UINT8",e[e.DT_INT16=5]="DT_INT16",e[e.DT_INT8=6]="DT_INT8",e[e.DT_STRING=7]="DT_STRING",e[e.DT_COMPLEX64=8]="DT_COMPLEX64",e[e.DT_INT64=9]="DT_INT64",e[e.DT_BOOL=10]="DT_BOOL",e[e.DT_QINT8=11]="DT_QINT8",e[e.DT_QUINT8=12]="DT_QUINT8",e[e.DT_QINT32=13]="DT_QINT32",e[e.DT_BFLOAT16=14]="DT_BFLOAT16",e[e.DT_QINT16=15]="DT_QINT16",e[e.DT_QUINT16=16]="DT_QUINT16",e[e.DT_UINT16=17]="DT_UINT16",e[e.DT_COMPLEX128=18]="DT_COMPLEX128",e[e.DT_HALF=19]="DT_HALF",e[e.DT_RESOURCE=20]="DT_RESOURCE",e[e.DT_VARIANT=21]="DT_VARIANT",e[e.DT_UINT32=22]="DT_UINT32",e[e.DT_UINT64=23]="DT_UINT64",e[e.DT_FLOAT_REF=101]="DT_FLOAT_REF",e[e.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",e[e.DT_INT32_REF=103]="DT_INT32_REF",e[e.DT_UINT8_REF=104]="DT_UINT8_REF",e[e.DT_INT16_REF=105]="DT_INT16_REF",e[e.DT_INT8_REF=106]="DT_INT8_REF",e[e.DT_STRING_REF=107]="DT_STRING_REF",e[e.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",e[e.DT_INT64_REF=109]="DT_INT64_REF",e[e.DT_BOOL_REF=110]="DT_BOOL_REF",e[e.DT_QINT8_REF=111]="DT_QINT8_REF",e[e.DT_QUINT8_REF=112]="DT_QUINT8_REF",e[e.DT_QINT32_REF=113]="DT_QINT32_REF",e[e.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",e[e.DT_QINT16_REF=115]="DT_QINT16_REF",e[e.DT_QUINT16_REF=116]="DT_QUINT16_REF",e[e.DT_UINT16_REF=117]="DT_UINT16_REF",e[e.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",e[e.DT_HALF_REF=119]="DT_HALF_REF",e[e.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",e[e.DT_VARIANT_REF=121]="DT_VARIANT_REF",e[e.DT_UINT32_REF=122]="DT_UINT32_REF",e[e.DT_UINT64_REF=123]="DT_UINT64_REF"})(Nr||(Nr={}));var Yk;(function(e){let t;(function(n){n[n.LEGACY=0]="LEGACY",n[n.V1=1]="V1",n[n.V2=2]="V2"})(t=e.CheckpointFormatVersion||(e.CheckpointFormatVersion={}))})(Yk||(Yk={}));var _0={};function RH(e,t){let n={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};_0[e]=n}function kN(e){return _0[e]}function PH(e){delete _0[e]}function I(e,t,n,r,s){let a=t.inputParams[e];if(a&&a.inputIndexStart!==void 0){let i=a.inputIndexStart,c=a.inputIndexEnd===0?void 0:a.inputIndexEnd===void 0?i+1:a.inputIndexEnd;if(a.type==="tensor")return Sn(t.inputNames[a.inputIndexStart],n,r,s);if(a.type==="tensors")return t.inputNames.slice(i,c).map(d=>Sn(d,n,r,s));let u=Sn(t.inputNames.slice(i)[0],n,r,s),l=u.dataSync();return a.type==="number"?l[0]:w.toNestedArray(u.shape,l)}let o=t.attrParams[e];return o&&o.value}function Sn(e,t,n,r){let[s,a]=Zn(e);if(r!=null){let i=r.getHashTableHandleByName(s);if(i!=null)return i}let o=n.currentContextIds.find(i=>!!t[Yh(s,i)]);return o!==void 0?t[Yh(s,o)][a]:void 0}function OH(e,t,n){return t[Yh(e,n.currentContextId)]}function os(e,t){let[n,r,s]=Zn(e);return[Yh(n,t&&t.currentContextId),r,s]}function Yh(e,t){return t?`${e}-${t}`:e}function Zn(e){let t=e.split(":");if(t.length===1)return[e,0,void 0];let n=t[0],r=t.length===3?t[1]:void 0,s=Number(t[t.length-1]);return[n,s,r]}function wh(e,t,n){let r=I("pad",e,t,n);if(r==="explicit"){r=I("explicitPaddings",e,t,n);let s=[[0,0],[0,0],[0,0],[0,0]];for(let a=0;a<4;a++)s[a][0]=r[a*2],s[a][1]=r[a*2+1];return s}return r}function Ts(e){return e.kept?e:is(e)}var SN={};Ae(SN,{json:()=>MH});var MH=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],TN={};Ae(TN,{json:()=>LH});var LH=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],CN={};Ae(CN,{json:()=>zH});var zH=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}],NN={};Ae(NN,{json:()=>BH});var BH=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],_N={};Ae(_N,{json:()=>WH});var WH=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomStandardNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],EN={};Ae(EN,{json:()=>VH});var VH=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],AN={};Ae(AN,{json:()=>UH});var UH=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],$N={};Ae($N,{json:()=>GH});var GH=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],DN={};Ae(DN,{json:()=>HH});var HH=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}],FN={};Ae(FN,{json:()=>qH});var qH=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}],RN={};Ae(RN,{json:()=>jH});var jH=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],PN={};Ae(PN,{json:()=>KH});var KH=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}],ON={};Ae(ON,{json:()=>XH});var XH=[{tfOpName:"EuclideanNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",defaultValue:!1}]},{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}],MN={};Ae(MN,{json:()=>YH});var YH=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],LN={};Ae(LN,{json:()=>ZH});var ZH=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}],zN={};Ae(zN,{json:()=>JH});var JH=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],BN={};Ae(BN,{json:()=>QH});var QH=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],WN={};Ae(WN,{json:()=>e6});var e6=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],VN={};Ae(VN,{json:()=>t6});var t6=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],Zk=class{constructor(){let e=[SN,TN,CN,NN,_N,EN,AN,$N,DN,FN,RN,PN,ON,MN,LN,zN,BN,WN,VN],t=[].concat(...e.map(n=>n.json));this.opMappers=t.reduce((n,r)=>(n[r.tfOpName]=r,n),{})}static get Instance(){return this._instance||(this._instance=new this)}transformGraph(e,t={}){let n=e.node,r=[],s=[],a=[],o=n.reduce((f,m)=>(f[m.name]=this.mapNode(m),m.op.startsWith("Placeholder")?r.push(f[m.name]):m.op==="Const"?s.push(f[m.name]):(m.input==null||m.input.length===0)&&a.push(f[m.name]),f),{}),i=[],c=[],u={},l={};t!=null&&(u=this.mapSignatureEntries(t.inputs),l=this.mapSignatureEntries(t.outputs));let p=Object.keys(o);p.forEach(f=>{let m=o[f];m.inputNames.forEach((g,b)=>{let[y,,v]=os(g),x=o[y];if(x.outputs!=null){let k=x.outputs.indexOf(v);if(k!==-1){let S=`${y}:${k}`;m.inputNames[b]=S}}m.inputs.push(x),x.children.push(m)})}),Object.keys(l).length===0?p.forEach(f=>{let m=o[f];m.children.length===0&&c.push(m)}):Object.keys(l).forEach(f=>{let[m]=os(f),g=o[m];g!=null&&(g.signatureKey=l[f],c.push(g))}),Object.keys(u).length>0?Object.keys(u).forEach(f=>{let[m]=os(f),g=o[m];g&&(g.signatureKey=u[f],i.push(g))}):i=r;let d={};e.library!=null&&e.library.function!=null&&(d=e.library.function.reduce((f,m)=>(f[m.signature.name]=this.mapFunction(m),f),{}));let h={nodes:o,inputs:i,outputs:c,weights:s,placeholders:r,signature:t,functions:d};return a.length>0&&(h.initNodes=a),h}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,n)=>(t[e[n].name]=n,t),{})}mapNode(e){let t=kN(e.op)||this.opMappers[e.op]||{};e.attr==null&&(e.attr={});let n={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(r=>r.startsWith("^")?r.slice(1):r),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return t.inputs!=null&&(n.inputParams=t.inputs.reduce((r,s)=>(r[s.name]={type:s.type,inputIndexStart:s.start,inputIndexEnd:s.end},r),{})),t.attrs!=null&&(n.attrParams=t.attrs.reduce((r,s)=>{let a=s.type,o;switch(s.type){case"string":o=Qy(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=Qy(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"string[]":o=ov(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=ov(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"number":o=tv(e.attr,s.tfName,s.defaultValue||0),o===void 0&&!!s.tfDeprecatedName&&(o=tv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"number[]":o=av(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=av(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"bool":o=ev(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=ev(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"bool[]":o=cv(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=cv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"shape":o=sv(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=sv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"shape[]":o=iv(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=iv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"dtype":o=nv(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=nv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"dtype[]":o=rv(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=rv(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"func":o=Jk(e.attr,s.tfName,s.defaultValue),o===void 0&&!!s.tfDeprecatedName&&(o=Jk(e.attr,s.tfDeprecatedName,s.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${s.type} for op: ${e.op}`)}return r[s.name]={value:o,type:a},r},{})),n}mapFunction(e){let t=e.nodeDef,n=[],r=[],s={};t!=null&&(s=t.reduce((l,p)=>(l[p.name]=this.mapNode(p),p.op==="Const"&&r.push(l[p.name]),l),{}));let a=[],o=[];e.signature.inputArg.forEach(l=>{let[p]=os(l.name),d={name:p,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:E0(l.type),type:"dtype"}},children:[]};d.signatureKey=l.name,a.push(d),s[p]=d}),Object.keys(s).forEach(l=>{let p=s[l];p.inputNames.forEach((d,h)=>{let[f,,m]=os(d),g=s[f];if(g.outputs!=null){let b=g.outputs.indexOf(m);if(b!==-1){let y=`${f}:${b}`;p.inputNames[h]=y}}p.inputs.push(g),g.children.push(p)})});let c=e.ret;e.signature.outputArg.forEach(l=>{let[p,d]=os(c[l.name]),h=s[p];h!=null&&(h.defaultOutput=d,o.push(h))});let u=this.mapArgsToSignature(e);return{nodes:s,inputs:a,outputs:o,weights:r,placeholders:n,signature:u}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n),t),{}),outputs:e.signature.outputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n,e.ret),t),{})}}mapArgToTensorInfo(e,t){let n=e.name;return t!=null&&(n=t[n]),{name:n,dtype:e.type}}};function n6(e){let t=q().global;if(typeof t.atob!="undefined")return t.atob(e);if(typeof Buffer!="undefined")return new Buffer(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}function UN(e,t){let n=Array.isArray(e)?String.fromCharCode.apply(null,e):n6(e);return t?n:n.toLowerCase()}function Qy(e,t,n,r=!1){let s=e[t];return s!=null?UN(s.s,r):n}function ev(e,t,n){let r=e[t];return r?r.b:n}function tv(e,t,n){let r=e[t]||{},s=r.i!=null?r.i:r.f!=null?r.f:n;return typeof s=="number"?s:parseInt(s,10)}function E0(e){switch(typeof e=="string"&&(e=Nr[e]),e){case Nr.DT_FLOAT:case Nr.DT_HALF:return"float32";case Nr.DT_INT32:case Nr.DT_INT64:case Nr.DT_INT8:case Nr.DT_UINT8:return"int32";case Nr.DT_BOOL:return"bool";case Nr.DT_DOUBLE:return"float32";case Nr.DT_STRING:return"string";default:return null}}function Jk(e,t,n){let r=e[t];return r&&r.func?r.func.name:n}function nv(e,t,n){let r=e[t];return r&&r.type?E0(r.type):n}function rv(e,t,n){let r=e[t];return r&&r.list&&r.list.type?r.list.type.map(s=>E0(s)):n}function GN(e){if(!e.unknownRank)return e.dim!=null?e.dim.map(t=>typeof t.size=="number"?t.size:parseInt(t.size,10)):[]}function sv(e,t,n){let r=e[t];return r&&r.shape?GN(r.shape):n}function av(e,t,n){let r=e[t];return r?((r.list.f&&r.list.f.length?r.list.f:r.list.i)||[]).map(s=>typeof s=="number"?s:parseInt(s,10)):n}function ov(e,t,n,r=!1){let s=e[t];return s&&s.list&&s.list.s?s.list.s.map(a=>UN(a,r)):n}function iv(e,t,n){let r=e[t];return r&&r.list&&r.list.shape?r.list.shape.map(s=>GN(s)):n}function cv(e,t,n){let r=e[t];return r&&r.list&&r.list.b?r.list.b:n}var r6=class{constructor(e,t,n){this.node=e,this.tensorMap=t,this.context=n,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(r=>this.getInput(r)),e.rawAttrs!=null&&(this.attrs=Object.keys(e.rawAttrs).reduce((r,s)=>(r[s]=this.getAttr(s),r),{}))}getInput(e){return Sn(e,this.tensorMap,this.context)}getAttr(e,t){let n=this.node.rawAttrs[e];if(n.tensor!=null)return Sn(e,this.tensorMap,this.context);if(n.i!=null||n.f!=null)return tv(this.node.rawAttrs,e,t);if(n.s!=null)return Qy(this.node.rawAttrs,e,t);if(n.b!=null)return ev(this.node.rawAttrs,e,t);if(n.shape!=null)return sv(this.node.rawAttrs,e,t);if(n.type!=null)return nv(this.node.rawAttrs,e,t);if(n.list!=null){if(n.list.i!=null||n.list.f!=null)return av(this.node.rawAttrs,e,t);if(n.list.s!=null)return ov(this.node.rawAttrs,e,t);if(n.list.shape!=null)return iv(this.node.rawAttrs,e,t);if(n.list.b!=null)return cv(this.node.rawAttrs,e,t);if(n.list.type!=null)return rv(this.node.rawAttrs,e,t)}return t}},ln={};Ae(ln,{OP_SCOPE_SUFFIX:()=>Pv,abs:()=>Lt,acos:()=>Xv,acosh:()=>Yv,add:()=>Y,addN:()=>YS,all:()=>Kf,any:()=>Ql,argMax:()=>ao,argMin:()=>Zv,asin:()=>Jv,asinh:()=>Qv,atan:()=>ex,atan2:()=>tx,atanh:()=>nx,avgPool:()=>mr,avgPool3d:()=>sx,basicLSTMCell:()=>eT,batchNorm:()=>ka,batchNorm2d:()=>ax,batchNorm3d:()=>ox,batchNorm4d:()=>ix,batchToSpaceND:()=>Od,bincount:()=>cx,booleanMaskAsync:()=>OT,broadcastArgs:()=>tT,broadcastTo:()=>Ya,buffer:()=>Me,cast:()=>ce,ceil:()=>ux,clipByValue:()=>en,clone:()=>is,complex:()=>As,concat:()=>Je,concat1d:()=>lx,concat2d:()=>dx,concat3d:()=>px,concat4d:()=>hx,conv1d:()=>Xf,conv2d:()=>Dt,conv2dTranspose:()=>Yf,conv3d:()=>mx,conv3dTranspose:()=>gx,cos:()=>Md,cosh:()=>Zf,cosineWindow:()=>ym,cumprod:()=>ed,cumsum:()=>Jf,denseBincount:()=>Bh,depthToSpace:()=>bx,depthwiseConv2d:()=>Sa,diag:()=>rT,dilation2d:()=>yx,div:()=>fe,divNoNan:()=>vx,dot:()=>xx,dropout:()=>jx,einsum:()=>sT,elu:()=>Lu,enclosingPowerOfTwo:()=>Kx,equal:()=>tr,erf:()=>wx,euclideanNorm:()=>Sx,exp:()=>gn,expandDims:()=>Zt,expm1:()=>Tx,eye:()=>Qf,fft:()=>Hd,fill:()=>bn,floor:()=>Bu,floorDiv:()=>jf,fused:()=>Ic,gather:()=>Wu,gatherND:()=>BT,greater:()=>An,greaterEqual:()=>Ds,ifft:()=>wc,imag:()=>Fd,image:()=>Br,inTopKAsync:()=>WT,irfft:()=>fm,isFinite:()=>Cx,isInf:()=>Nx,isNaN:()=>_x,leakyRelu:()=>Ld,less:()=>em,lessEqual:()=>Ta,linalg:()=>Zx,linspace:()=>uT,localResponseNormalization:()=>Ex,log:()=>nr,log1p:()=>zd,logSigmoid:()=>Ax,logSoftmax:()=>nm,logSumExp:()=>rm,logicalAnd:()=>Dr,logicalNot:()=>Bd,logicalOr:()=>sm,logicalXor:()=>$x,losses:()=>QT,lowerBound:()=>dT,matMul:()=>De,max:()=>pr,maxPool:()=>Rt,maxPool3d:()=>Dx,maxPoolWithArgmax:()=>pT,maximum:()=>hs,mean:()=>Ct,meshgrid:()=>hT,min:()=>yc,minimum:()=>Vu,mirrorPad:()=>Fx,mod:()=>Rx,moments:()=>Wd,movingAverage:()=>MT,mul:()=>B,multiRNNCell:()=>fT,multinomial:()=>mT,neg:()=>xt,norm:()=>zu,notEqual:()=>co,oneHot:()=>gc,ones:()=>Qn,onesLike:()=>rr,op:()=>z,outerProduct:()=>gT,pad:()=>gr,pad1d:()=>bT,pad2d:()=>yT,pad3d:()=>vT,pad4d:()=>xT,pool:()=>Px,pow:()=>$s,prelu:()=>Ud,print:()=>zv,prod:()=>Ox,raggedGather:()=>wT,raggedRange:()=>IT,raggedTensorToTensor:()=>kT,rand:()=>ST,randomGamma:()=>TT,randomNormal:()=>om,randomStandardNormal:()=>CT,randomUniform:()=>Uu,range:()=>vc,real:()=>bc,reciprocal:()=>zx,relu:()=>Xe,relu6:()=>im,reshape:()=>W,reverse:()=>fr,reverse1d:()=>NT,reverse2d:()=>_T,reverse3d:()=>ET,reverse4d:()=>AT,rfft:()=>qd,round:()=>cm,rsqrt:()=>um,scalar:()=>xe,scatterND:()=>LT,searchSorted:()=>am,selu:()=>lm,separableConv2d:()=>Ca,setdiff1dAsync:()=>$T,sigmoid:()=>dr,sign:()=>Bx,signal:()=>JT,sin:()=>dm,sinh:()=>pm,slice:()=>We,slice1d:()=>Gd,slice2d:()=>hm,slice3d:()=>gi,slice4d:()=>xc,softmax:()=>Xr,softplus:()=>mi,spaceToBatchND:()=>Vd,sparse:()=>eC,sparseToDense:()=>zT,spectral:()=>ZT,split:()=>zn,sqrt:()=>un,square:()=>it,squaredDifference:()=>mm,squeeze:()=>Na,stack:()=>Ft,step:()=>bi,stridedSlice:()=>Wx,string:()=>tC,sub:()=>de,sum:()=>ge,tan:()=>Vx,tanh:()=>oo,tensor:()=>Cn,tensor1d:()=>Ke,tensor2d:()=>$r,tensor3d:()=>Rd,tensor4d:()=>Rr,tensor5d:()=>DT,tensor6d:()=>FT,tile:()=>Ln,topk:()=>Ux,transpose:()=>Ee,truncatedNormal:()=>gm,unique:()=>Gx,unsortedSegmentSum:()=>bm,unstack:()=>lt,upperBound:()=>RT,variable:()=>Hx,where:()=>mn,whereAsync:()=>qx,zeros:()=>kt,zerosLike:()=>qe});var s6=(e,t,n,r=ln)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[r.add(I("a",e,t,n),I("b",e,t,n))];case"AddN":return[r.addN(I("tensors",e,t,n))];case"FloorMod":case"Mod":return[r.mod(I("a",e,t,n),I("b",e,t,n))];case"Mul":return[r.mul(I("a",e,t,n),I("b",e,t,n))];case"RealDiv":case"Div":return[r.div(I("a",e,t,n),I("b",e,t,n))];case"DivNoNan":return[r.divNoNan(I("a",e,t,n),I("b",e,t,n))];case"FloorDiv":return[r.floorDiv(I("a",e,t,n),I("b",e,t,n))];case"Sub":return[r.sub(I("a",e,t,n),I("b",e,t,n))];case"Minimum":return[r.minimum(I("a",e,t,n),I("b",e,t,n))];case"Maximum":return[r.maximum(I("a",e,t,n),I("b",e,t,n))];case"Pow":return[r.pow(I("a",e,t,n),I("b",e,t,n))];case"SquaredDifference":return[r.squaredDifference(I("a",e,t,n),I("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},a6=(e,t,n,r=ln)=>{switch(e.op){case"Abs":case"ComplexAbs":return[r.abs(I("x",e,t,n))];case"Acos":return[r.acos(I("x",e,t,n))];case"Acosh":return[r.acosh(I("x",e,t,n))];case"Asin":return[r.asin(I("x",e,t,n))];case"Asinh":return[r.asinh(I("x",e,t,n))];case"Atan":return[r.atan(I("x",e,t,n))];case"Atan2":return[r.atan2(I("x",e,t,n),I("y",e,t,n))];case"Atanh":return[r.atanh(I("x",e,t,n))];case"Ceil":return[r.ceil(I("x",e,t,n))];case"Complex":return[r.complex(I("real",e,t,n),I("imag",e,t,n))];case"Cos":return[r.cos(I("x",e,t,n))];case"Cosh":return[r.cosh(I("x",e,t,n))];case"Elu":return[r.elu(I("x",e,t,n))];case"Erf":return[r.erf(I("x",e,t,n))];case"Exp":return[r.exp(I("x",e,t,n))];case"Expm1":return[r.expm1(I("x",e,t,n))];case"Floor":return[r.floor(I("x",e,t,n))];case"Log":return[r.log(I("x",e,t,n))];case"Log1p":return[r.log1p(I("x",e,t,n))];case"Imag":return[r.imag(I("x",e,t,n))];case"Neg":return[r.neg(I("x",e,t,n))];case"Reciprocal":return[r.reciprocal(I("x",e,t,n))];case"Real":return[r.real(I("x",e,t,n))];case"Relu":return[r.relu(I("x",e,t,n))];case"Round":return[r.round(I("x",e,t,n))];case"Selu":return[r.selu(I("x",e,t,n))];case"Sigmoid":return[r.sigmoid(I("x",e,t,n))];case"Sin":return[r.sin(I("x",e,t,n))];case"Sign":return[r.sign(I("x",e,t,n))];case"Sinh":return[r.sinh(I("x",e,t,n))];case"Softplus":return[r.softplus(I("x",e,t,n))];case"Sqrt":return[r.sqrt(I("x",e,t,n))];case"Square":return[r.square(I("x",e,t,n))];case"Tanh":return[r.tanh(I("x",e,t,n))];case"Tan":return[r.tan(I("x",e,t,n))];case"ClipByValue":return[r.clipByValue(I("x",e,t,n),I("clipValueMin",e,t,n),I("clipValueMax",e,t,n))];case"Relu6":return[r.relu6(I("x",e,t,n))];case"Rsqrt":return[r.rsqrt(Sn(e.inputNames[0],t,n))];case"Prod":return[r.prod(I("x",e,t,n),I("axes",e,t,n))];case"LeakyRelu":return[r.leakyRelu(I("x",e,t,n),I("alpha",e,t,n))];case"Prelu":return[r.prelu(I("x",e,t,n),I("alpha",e,t,n))];case"IsNan":return[r.isNaN(Sn(e.inputNames[0],t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Ar(e,t,n=""){if(!(typeof e=="number"||typeof t=="number")){w.assert(e.length===t.length,()=>n+` Shapes ${e} and ${t} must match`);for(let r=0;r<e.length;r++){let s=e[r],a=t[r];w.assert(s<0||a<0||s===a,()=>n+` Shapes ${e} and ${t} must match`)}}}function Qk(e){return!(typeof e=="number"||e.some(t=>t<0))}function $l(e,t,n){let r=uv(e,n),s=!Qk(r);if(s&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${r}`);if(s&&t.forEach(a=>{r=uv(a.shape,r)}),!Qk(r))throw new Error(`Non-fully-defined elementShape: ${r}`);return r}function uv(e,t){if(typeof e=="number")return t;if(typeof t=="number")return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);let n=[];for(let r=0;r<e.length;++r){let s=e[r],a=t[r];if(s>=0&&a>=0&&s!==a)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);n[r]=s>=0?s:a}return n}var o6=class{constructor(e,t,n,r,s,a,o){this.name=e,this.dtype=t,this.maxSize=n,this.elementShape=r,this.identicalElementShapes=s,this.dynamicSize=a,this.clearAfterRead=o,this.tensors=[],this.closed_=!1,this.idTensor=xe(0),Jt(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.tensor.id))&&t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);let t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(t=>this.read(t))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},
+          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=t.shape),Ar(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);n.tensor=t,Jt(t),n.written=!0,this.tensors[e]=n}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((n,r)=>this.write(n,t[r]))}gather(e,t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let r=0;r<this.size();r++)e.push(r)}if(e.length===0)return Cn([],[0].concat(this.elementShape));let n=this.readMany(e);return Ar(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),Ft(n,0)}concat(e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(this.size()===0)return Cn([],[0].concat(this.elementShape));let t=[];for(let r=0;r<this.size();r++)t.push(r);let n=this.readMany(t);return Ar(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),Je(n,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);let n=Math.max(...e);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(e,lt(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let n=0,r=e.map(i=>(n+=i,n));if(n!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${n}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);let r=n===0?0:t.size/n,s=[];P(()=>{t=W(t,[1,n,r]);for(let o=0;o<e.length;++o){let l=o===0?0:a[o-1],u=[0,l,0],p=[1,e[o],r];s[o]=W(Be(t,u,p),this.elementShape)}return s});let i=[];for(let o=0;o<e.length;o++)i[o]=o;this.writeMany(i,s)}},Nl=class{constructor(e,t,n,a=-1){this.tensors=e,this.elementShape=t,this.elementDtype=n,e!=null&&e.forEach(r=>{if(n!==r.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${r.dtype}`);Ea(t,r.shape,"TensorList shape mismatch: "),Jt(r)}),this.idTensor=be(0),this.maxNumElements=a,Jt(this.idTensor)}get id(){return this.idTensor.id}copy(){return new Nl([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.id))&&t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,n=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(n!==-1&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);Ea(e,this.elementShape,"TensorList shape mismatch: ");let a=$p(this.elementShape,this.tensors,e);return P(()=>{let r=this.tensors.map(s=>W(s,a));return Ft(r,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error("Trying to pop from an empty list.");let n=$p(this.elementShape,this.tensors,e),a=this.tensors.pop();return a.kept=!1,Ea(a.shape,e,"TensorList shape mismatch: "),W(a,n)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(Ea(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");Jt(e),this.tensors.push(e)}resize(e){if(e<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${e}`);if(this.maxNumElements!==-1&&e>this.maxNumElements)throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);let t=new Nl([],this.elementShape,this.elementDtype,this.maxNumElements);t.tensors.length=e;for(let n=0;n<Math.min(this.tensors.length,e);++n)t.tensors[n]=this.tensors[n];return t}getItem(e,t,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(this.tensors[e]==null)throw new Error(`element at index ${e} is null.`);Ea(this.tensors[e].shape,t,"TensorList shape mismatch: ");let a=$p(this.elementShape,this.tensors,t);return W(this.tensors[e],a)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||this.maxNumElements!==-1&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);Ea(this.elementShape,t.shape,"TensorList shape mismatch: "),Jt(t),this.tensors[e]!=null&&(this.tensors[e].kept=!1),this.tensors[e]=t}gather(e,t,n){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);Ea(this.elementShape,n,"TensorList shape mismatch: "),e=e.slice(0,this.size());let a=$p(this.elementShape,this.tensors,n);return e.length===0?In([],[0].concat(a)):P(()=>{let r=e.map(s=>W(this.tensors[s],a));return Ft(r,0)})}concat(e,t){if(!!e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);Ea(this.elementShape,t,"TensorList shape mismatch: ");let n=$p(this.elementShape,this.tensors,t);return this.size()===0?In([],[0].concat(n)):P(()=>{let a=this.tensors.map(r=>W(r,n));return Ze(a,0)})}};function r6(e,t,n){let a=e.dtype;if(e.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${e.shape}`);if(e.dtype!==n)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);let r=e.shape.slice(1);Ea(r,t,"TensorList shape mismatch: ");let s=ct(e);return new Nl(s,t,a)}function s6(e,t,n,a){return new Nl([],e,t,a)}function i6(e,t,n,a){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let r=Math.max(...t);if(a!=null&&a!==-1&&r>=a)throw new Error(`Max index must be < array size (${r}  vs. ${a})`);let s=new Nl([],n,e.dtype,a),i=ct(e,0);return t.forEach((o,l)=>{s.setItem(o,i[l])}),s}function o6(e,t,n){let a=0,r=t.map(p=>(a+=p,a));if(a!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to
+        ${n}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);let s=n===0?0:t.size/n,a=[];O(()=>{t=W(t,[1,n,s]);for(let i=0;i<e.length;++i){let u=[0,i===0?0:r[i-1],0],l=[1,e[i],s];a[i]=W(We(t,u,l),this.elementShape)}return a});let o=[];for(let i=0;i<e.length;i++)o[i]=i;this.writeMany(o,a)}},Cc=class{constructor(e,t,n,r=-1){this.tensors=e,this.elementShape=t,this.elementDtype=n,e!=null&&e.forEach(s=>{if(n!==s.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${s.dtype}`);Ar(t,s.shape,"TensorList shape mismatch: "),Jt(s)}),this.idTensor=xe(0),this.maxNumElements=r,Jt(this.idTensor)}get id(){return this.idTensor.id}copy(){return new Cc([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.id))&&t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,n=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(n!==-1&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);Ar(e,this.elementShape,"TensorList shape mismatch: ");let r=$l(this.elementShape,this.tensors,e);return O(()=>{let s=this.tensors.map(a=>W(a,r));return Ft(s,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error("Trying to pop from an empty list.");let n=$l(this.elementShape,this.tensors,e),r=this.tensors.pop();return r.kept=!1,Ar(r.shape,e,"TensorList shape mismatch: "),W(r,n)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(Ar(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");Jt(e),this.tensors.push(e)}resize(e){if(e<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${e}`);if(this.maxNumElements!==-1&&e>this.maxNumElements)throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);let t=new Cc([],this.elementShape,this.elementDtype,this.maxNumElements);t.tensors.length=e;for(let n=0;n<Math.min(this.tensors.length,e);++n)t.tensors[n]=this.tensors[n];return t}getItem(e,t,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(this.tensors[e]==null)throw new Error(`element at index ${e} is null.`);Ar(this.tensors[e].shape,t,"TensorList shape mismatch: ");let r=$l(this.elementShape,this.tensors,t);return W(this.tensors[e],r)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||this.maxNumElements!==-1&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);Ar(this.elementShape,t.shape,"TensorList shape mismatch: "),Jt(t),this.tensors[e]!=null&&(this.tensors[e].kept=!1),this.tensors[e]=t}gather(e,t,n){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);Ar(this.elementShape,n,"TensorList shape mismatch: "),e=e.slice(0,this.size());let r=$l(this.elementShape,this.tensors,n);return e.length===0?Cn([],[0].concat(r)):O(()=>{let s=e.map(a=>W(this.tensors[a],r));return Ft(s,0)})}concat(e,t){if(!!e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);Ar(this.elementShape,t,"TensorList shape mismatch: ");let n=$l(this.elementShape,this.tensors,t);return this.size()===0?Cn([],[0].concat(n)):O(()=>{let r=this.tensors.map(s=>W(s,n));return Je(r,0)})}};function i6(e,t,n){let r=e.dtype;if(e.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${e.shape}`);if(e.dtype!==n)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);let s=e.shape.slice(1);Ar(s,t,"TensorList shape mismatch: ");let a=lt(e);return new Cc(a,t,r)}function c6(e,t,n,r){return new Cc([],e,t,r)}function u6(e,t,n,r){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let s=Math.max(...t);if(r!=null&&r!==-1&&s>=r)throw new Error(`Max index must be < array size (${s}  vs. ${r})`);let a=new Cc([],n,e.dtype,r),o=lt(e,0);return t.forEach((i,c)=>{a.setItem(i,o[c])}),a}function l6(e,t,n){let r=0,s=t.map(l=>(r+=l,r));if(r!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${a}, and tensor's shape is: ${e.shape}`);let s=e.shape.slice(1),i=ux(s,n),o=a===0?0:e.size/a,l=P(()=>{let p=[];e=W(e,[1,a,o]);for(let d=0;d<t.length;++d){let c=d===0?0:r[d-1],h=[0,c,0],m=[1,t[d],o];p[d]=W(Be(e,h,m),i)}return e.dispose(),p}),u=new Nl([],n,e.dtype,t.length);for(let p=0;p<l.length;p++)u.setItem(p,l[p]);return u}var l6=async(e,t,n)=>{switch(e.op){case"If":case"StatelessIf":{let a=k("thenBranch",e,t,n),r=k("elseBranch",e,t,n),s=k("cond",e,t,n),i=k("args",e,t,n);return(await s.data())[0]?n.functionMap[a].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap):n.functionMap[r].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap)}case"While":case"StatelessWhile":{let a=k("body",e,t,n),r=k("cond",e,t,n),s=k("args",e,t,n),i=await n.functionMap[r].executeFunctionAsync(s,n.tensorArrayMap,n.tensorListMap),o=s.map(p=>p.id),l=await i[0].data();i.forEach(p=>{!p.kept&&o.indexOf(p.id)===-1&&p.dispose()});let u=s;for(;l[0];){let p=u;u=await n.functionMap[a].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);let d=u.map(h=>h.id);p.forEach(h=>{!h.kept&&o.indexOf(h.id)===-1&&d.indexOf(h.id)===-1&&h.dispose()});let c=await n.functionMap[r].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);l=await c[0].data(),c.forEach(h=>{!h.kept&&o.indexOf(h.id)===-1&&d.indexOf(h.id)===-1&&h.dispose()})}return u}case"LoopCond":{let a=k("pred",e,t,n);return[Tr(a)]}case"Switch":{let a=k("pred",e,t,n),r=k("data",e,t,n);return r.kept||(r=Tr(r)),(await a.data())[0]?[void 0,r]:[r,void 0]}case"Merge":{let a=e.inputNames.find(r=>wn(r,t,n)!==void 0);if(a){let r=wn(a,t,n);return[Tr(r)]}return}case"Enter":{let a=k("frameName",e,t,n),r=k("tensor",e,t,n);return n.enterFrame(a),[Tr(r)]}case"Exit":{let a=k("tensor",e,t,n);return n.exitFrame(),[Tr(a)]}case"NextIteration":{let a=k("tensor",e,t,n);return n.nextIteration(),[Tr(a)]}case"TensorArrayV3":{let a=k("size",e,t,n),r=k("dtype",e,t,n),s=k("elementShape",e,t,n),i=k("dynamicSize",e,t,n),o=k("clearAfterRead",e,t,n),l=k("identicalElementShapes",e,t,n),u=k("name",e,t,n),p=new a6(u,r,a,s,l,i,o);return n.addTensorArray(p),[p.idTensor,be(1)]}case"TensorArrayWriteV3":{let a=k("tensorArrayId",e,t,n),r=k("index",e,t,n),s=k("tensor",e,t,n),i=n.getTensorArray(a.id);return i.write(r,s),[i.idTensor]}case"TensorArrayReadV3":{let a=k("tensorArrayId",e,t,n),r=k("index",e,t,n);return[n.getTensorArray(a.id).read(r)]}case"TensorArrayGatherV3":{let a=k("tensorArrayId",e,t,n),r=k("indices",e,t,n),s=k("dtype",e,t,n);return[n.getTensorArray(a.id).gather(r,s)]}case"TensorArrayScatterV3":{let a=k("tensorArrayId",e,t,n),r=k("indices",e,t,n),s=k("tensor",e,t,n),i=n.getTensorArray(a.id);return i.scatter(r,s),[i.idTensor]}case"TensorArrayConcatV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id),s=k("dtype",e,t,n);return[r.concat(s)]}case"TensorArraySplitV3":{let a=k("tensorArrayId",e,t,n),r=k("tensor",e,t,n),s=k("lengths",e,t,n),i=n.getTensorArray(a.id);return i.split(s,r),[i.idTensor]}case"TensorArraySizeV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id);return[be(r.size(),"int32")]}case"TensorArrayCloseV3":{let a=k("tensorArrayId",e,t,n),r=n.getTensorArray(a.id);return r.clearAndClose(),[r.idTensor]}case"TensorListSetItem":{let a=k("tensorListId",e,t,n),r=k("index",e,t,n),s=k("tensor",e,t,n),i=n.getTensorList(a.id);return i.setItem(r,s),[i.idTensor]}case"TensorListGetItem":{let a=k("tensorListId",e,t,n),r=k("index",e,t,n),s=k("elementShape",e,t,n),i=k("elementDType",e,t,n);return[n.getTensorList(a.id).getItem(r,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{let a=k("indices",e,t,n),r=k("tensor",e,t,n),s=k("elementShape",e,t,n),i=k("numElements",e,t,n),o=i6(r,a,s,i);return n.addTensorList(o),[o.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let a=k("elementShape",e,t,n),r=k("elementDType",e,t,n),s;e.op==="TensorListReserve"?s="numElements":s="maxNumElements";let i=k(s,e,t,n),o=e.op==="TensorListReserve"?-1:i,l=s6(a,r,i,o);return n.addTensorList(l),[l.idTensor]}case"TensorListGather":{let a=k("tensorListId",e,t,n),r=k("indices",e,t,n),s=k("elementShape",e,t,n),i=k("elementDType",e,t,n);return[n.getTensorList(a.id).gather(r,i,s)]}case"TensorListStack":{let a=k("tensorListId",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n),i=k("numElements",e,t,n);return[n.getTensorList(a.id).stack(r,s,i)]}case"TensorListFromTensor":{let a=k("tensor",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n),i=r6(a,r,s);return n.addTensorList(i),[i.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let a=k("tensorListId",e,t,n),r=n.getTensorList(a.id),s=k("dtype",e,t,n),i=k("elementShape",e,t,n);return[r.concat(s,i)]}case"TensorListPushBack":{let a=k("tensorListId",e,t,n),r=k("tensor",e,t,n),s=n.getTensorList(a.id);return s.pushBack(r),[s.idTensor]}case"TensorListPopBack":{let a=k("tensorListId",e,t,n),r=k("elementShape",e,t,n),s=k("elementDType",e,t,n);return[n.getTensorList(a.id).popBack(r,s)]}case"TensorListSplit":{let a=k("tensor",e,t,n),r=k("elementShape",e,t,n),s=k("lengths",e,t,n),i=o6(a,s,r);return n.addTensorList(i),[i.idTensor]}case"TensorListLength":{let a=k("tensorListId",e,t,n),r=n.getTensorList(a.id);return[be(r.size(),"int32")]}case"TensorListResize":{let a=k("tensorListId",e,t,n),r=k("size",e,t,n),s=n.getTensorList(a.id).resize(r);return n.addTensorList(s),[s.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Qk(e,t,n){let[a,r]=k("fusedOps",e,t,n),s=a==="biasadd",i=!s,o=r==="prelu",l=a==="fusedbatchnorm",u=k("numArgs",e,t,n);if(s){if(o&&u!==2)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&s&&u!==1)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(l)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let p=k("strides",e,t,n),d=vh(e,t,n),c=k("dataFormat",e,t,n).toUpperCase(),h=k("dilations",e,t,n),[m,f]=k("args",e,t,n);i&&(f=m,m=void 0);let g=k("leakyreluAlpha",e,t,n);return{stride:p,pad:d,dataFormat:c,dilations:h,biasArg:m,preluArg:f,activationFunc:r,leakyreluAlpha:g}}var u6=(e,t,n,a=pn)=>{switch(e.op){case"Conv1D":{let r=k("stride",e,t,n),s=k("pad",e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilation",e,t,n);return[a.conv1d(k("x",e,t,n),k("filter",e,t,n),r,s,i,o)]}case"Conv2D":{let r=k("strides",e,t,n),s=vh(e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilations",e,t,n);return[a.conv2d(k("x",e,t,n),k("filter",e,t,n),[r[1],r[2]],s,i,[o[1],o[2]])]}case"_FusedConv2D":{let{stride:r,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:p,leakyreluAlpha:d}=Qk(e,t,n);return[a.fused.conv2d({x:k("x",e,t,n),filter:k("filter",e,t,n),strides:[r[1],r[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:p,preluActivationWeights:u,leakyreluAlpha:d})]}case"FusedDepthwiseConv2dNative":{let{stride:r,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:p,leakyreluAlpha:d}=Qk(e,t,n);return[a.fused.depthwiseConv2d({x:k("x",e,t,n),filter:k("filter",e,t,n),strides:[r[1],r[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:p,preluActivationWeights:u,leakyreluAlpha:d})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let r=k("outputShape",e,t,n),s=k("strides",e,t,n),i=vh(e,t,n);return[a.conv2dTranspose(k("x",e,t,n),k("filter",e,t,n),r,[s[1],s[2]],i)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let r=k("strides",e,t,n),s=vh(e,t,n),i=k("dilations",e,t,n),o=k("dataFormat",e,t,n).toUpperCase();return[a.depthwiseConv2d(k("input",e,t,n),k("filter",e,t,n),[r[1],r[2]],s,o,[i[1],i[2]])]}case"Conv3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("dataFormat",e,t,n).toUpperCase(),o=k("dilations",e,t,n);return[a.conv3d(k("x",e,t,n),k("filter",e,t,n),[r[1],r[2],r[3]],s,i,[o[1],o[2],o[3]])]}case"AvgPool":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.avgPool(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s)]}case"MaxPool":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.maxPool(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s)]}case"MaxPoolWithArgmax":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n),o=k("includeBatchInIndex",e,t,n),{result:l,indexes:u}=a.maxPoolWithArgmax(k("x",e,t,n),[i[1],i[2]],[r[1],r[2]],s,o);return[l,u]}case"AvgPool3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.avgPool3d(k("x",e,t,n),[i[1],i[2],i[3]],[r[1],r[2],r[3]],s)]}case"MaxPool3D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("kernelSize",e,t,n);return[a.maxPool3d(k("x",e,t,n),[i[1],i[2],i[3]],[r[1],r[2],r[3]],s)]}case"Dilation2D":{let r=k("strides",e,t,n),s=k("pad",e,t,n),i=k("dilations",e,t,n),o=r[1],l=r[2],u=i[1],p=i[2];return[a.dilation2d(k("x",e,t,n),k("filter",e,t,n),[o,l],s,[u,p],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},p6=(e,t,n,a=pn)=>{switch(e.op){case"Fill":{let r=k("shape",e,t,n),s=k("dtype",e,t,n),i=k("value",e,t,n);return[a.fill(r,i,s)]}case"LinSpace":{let r=k("start",e,t,n),s=k("stop",e,t,n),i=k("num",e,t,n);return[a.linspace(r,s,i)]}case"Multinomial":{let r=k("logits",e,t,n),s=k("numSamples",e,t,n),i=k("seed",e,t,n);return[a.multinomial(r,s,i)]}case"OneHot":{let r=k("indices",e,t,n),s=k("depth",e,t,n),i=k("onValue",e,t,n),o=k("offValue",e,t,n),l=k("dtype",e,t,n);return[a.oneHot(r,s,i,o,l)]}case"Ones":return[a.ones(k("shape",e,t,n),k("dtype",e,t,n))];case"OnesLike":return[a.onesLike(k("x",e,t,n))];case"RandomStandardNormal":return[a.randomStandardNormal(k("shape",e,t,n),k("dtype",e,t,n),k("seed",e,t,n))];case"RandomUniform":return[a.randomUniform(k("shape",e,t,n),k("minval",e,t,n),k("maxval",e,t,n),k("dtype",e,t,n))];case"Range":{let r=k("start",e,t,n),s=k("stop",e,t,n),i=k("step",e,t,n);return[a.range(r,s,i,k("dtype",e,t,n))]}case"TruncatedNormal":{let r=k("shape",e,t,n),s=k("mean",e,t,n),i=k("stdDev",e,t,n),o=k("seed",e,t,n);return[a.truncatedNormal(r,s,i,k("dtype",e,t,n),o)]}case"Zeros":return[a.zeros(k("shape",e,t,n),k("dtype",e,t,n))];case"ZerosLike":return[a.zerosLike(k("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function kb(e,t,n){let a=k("boxes",e,t,n),r=k("scores",e,t,n),s=k("maxOutputSize",e,t,n),i=k("iouThreshold",e,t,n),o=k("scoreThreshold",e,t,n),l=k("softNmsSigma",e,t,n);return{boxes:a,scores:r,maxOutputSize:s,iouThreshold:i,scoreThreshold:o,softNmsSigma:l}}var c6=async(e,t,n,a,r=pn)=>{switch(e.op){case"NonMaxSuppressionV5":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u,softNmsSigma:p}=kb(e,t,n),d=await r.image.nonMaxSuppressionWithScoreAsync(s,i,o,l,u,p);return[d.selectedIndices,d.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u}=kb(e,t,n),p=k("padToMaxOutputSize",e,t,n),d=await r.image.nonMaxSuppressionPaddedAsync(s,i,o,l,u,p);return[d.selectedIndices,d.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:s,scores:i,maxOutputSize:o,iouThreshold:l,scoreThreshold:u}=kb(e,t,n);return[await r.image.nonMaxSuppressionAsync(s,i,o,l,u)]}case"Where":{let s=r.cast(k("condition",e,t,n),"bool"),i=[await r.whereAsync(s)];return s.dispose(),i}case"ListDiff":return r.setdiff1dAsync(k("x",e,t,n),k("y",e,t,n));default:throw TypeError(`Node type ${e.op} is not implemented`)}},d6=(e,t,n,a=pn)=>{switch(e.op){case"LowerBound":{let r=k("sortedSequence",e,t,n),s=k("values",e,t,n);return[a.lowerBound(r,s)]}case"TopKV2":{let r=k("x",e,t,n),s=k("k",e,t,n),i=k("sorted",e,t,n),o=a.topk(r,s,i);return[o.values,o.indices]}case"UpperBound":{let r=k("sortedSequence",e,t,n),s=k("values",e,t,n);return[a.upperBound(r,s)]}case"Unique":{let r=k("x",e,t,n),s=a.unique(r);return[s.values,s.indices]}case"UniqueV2":{let r=k("x",e,t,n),s=k("axis",e,t,n),i=a.unique(r,s);return[i.values,i.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},h6=(e,t,n,a=pn)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":let r=k("default",e,t,n);return[wn(e.name,t,n)||r];case"Placeholder":return[wn(e.name,t,n)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":{let p=k("x",e,t,n);return[Tr(p)]}case"IdentityN":return k("x",e,t,n).map(p=>Tr(p));case"Snapshot":let s=k("x",e,t,n);return[Tr(s)];case"Shape":return[a.tensor1d(k("x",e,t,n).shape,"int32")];case"ShapeN":return k("x",e,t,n).map(p=>a.tensor1d(p.shape));case"Size":return[a.scalar(k("x",e,t,n).size,"int32")];case"Rank":return[a.scalar(k("x",e,t,n).rank,"int32")];case"NoOp":return[a.scalar(1)];case"Print":let i=k("x",e,t,n),o=k("data",e,t,n),l=k("message",e,t,n),u=k("summarize",e,t,n);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(l);for(let p=0;p<o.length;p++)console.log(Array.prototype.slice.call(o[p].dataSync()).slice(0,u));return[i];default:throw TypeError(`Node type ${e.op} is not implemented`)}},m6=class{constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=be(0),this.tensorMap=new Map,Jt(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return be(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return this.tensorMap.forEach(a=>a.dispose()),this.tensorMap.clear(),P(()=>{let a=ct(t),r=n.length,s=a.length;v.assert(r===s,()=>`The number of elements doesn't match, keys has ${r} elements, the values has ${s} elements.`);for(let i=0;i<r;i++){let o=n[i],l=a[i];Jt(l),this.tensorMap.set(o,l)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return P(()=>{let a=[];for(let r=0;r<n.length;r++){let s=n[r],i=this.findWithDefault(s,t);a.push(i)}return Ft(a)})}findWithDefault(e,t){let n=this.tensorMap.get(e);return n!=null?n:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}},f6=async(e,t,n,a)=>{switch(e.op){case"HashTable":case"HashTableV2":{let r=a.getHashTableHandleByName(e.name);if(r!=null)return[r];{let s=k("keyDType",e,t,n),i=k("valueDType",e,t,n),o=new m6(s,i);return a.addHashTable(e.name,o),[o.handle]}}case"LookupTableImport":case"LookupTableImportV2":{let r=k("tableHandle",e,t,n,a),s=k("keys",e,t,n),i=k("values",e,t,n);return[await a.getHashTableById(r.id).import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{let r=k("tableHandle",e,t,n,a),s=k("keys",e,t,n),i=k("defaultValue",e,t,n);return[await a.getHashTableById(r.id).find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{let r=k("tableHandle",e,t,n,a);return[a.getHashTableById(r.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},g6=(e,t,n,a=pn)=>{switch(e.op){case"ResizeBilinear":{let r=k("images",e,t,n),s=k("size",e,t,n),i=k("alignCorners",e,t,n),o=k("halfPixelCenters",e,t,n);return[a.image.resizeBilinear(r,[s[0],s[1]],i,o)]}case"ResizeNearestNeighbor":{let r=k("images",e,t,n),s=k("size",e,t,n),i=k("alignCorners",e,t,n),o=k("halfPixelCenters",e,t,n);return[a.image.resizeNearestNeighbor(r,[s[0],s[1]],i,o)]}case"CropAndResize":{let r=k("image",e,t,n),s=k("boxes",e,t,n),i=k("boxInd",e,t,n),o=k("cropSize",e,t,n),l=k("method",e,t,n),u=k("extrapolationValue",e,t,n);return[a.image.cropAndResize(r,s,i,o,l,u)]}case"ImageProjectiveTransformV3":{let r=k("images",e,t,n),s=k("transforms",e,t,n),i=k("outputShape",e,t,n),o=k("fillValue",e,t,n),l=k("interpolation",e,t,n),u=k("fillMode",e,t,n);return[a.image.transform(r,s,l.toLowerCase(),u.toLowerCase(),o,i)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},y6=(e,t,n,a=pn)=>{switch(e.op){case"Equal":return[a.equal(k("a",e,t,n),k("b",e,t,n))];case"NotEqual":return[a.notEqual(k("a",e,t,n),k("b",e,t,n))];case"Greater":return[a.greater(k("a",e,t,n),k("b",e,t,n))];case"GreaterEqual":return[a.greaterEqual(k("a",e,t,n),k("b",e,t,n))];case"Less":return[a.less(k("a",e,t,n),k("b",e,t,n))];case"LessEqual":return[a.lessEqual(k("a",e,t,n),k("b",e,t,n))];case"LogicalAnd":return[a.logicalAnd(k("a",e,t,n),k("b",e,t,n))];case"LogicalNot":return[a.logicalNot(k("a",e,t,n))];case"LogicalOr":return[a.logicalOr(k("a",e,t,n),k("b",e,t,n))];case"Select":case"SelectV2":return[a.where(k("condition",e,t,n),k("a",e,t,n),k("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},b6=(e,t,n,a=pn)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[a.matMul(k("a",e,t,n),k("b",e,t,n),k("transposeA",e,t,n),k("transposeB",e,t,n))];case"Einsum":return[a.einsum(k("equation",e,t,n),...k("tensors",e,t,n))];case"Transpose":return[a.transpose(k("x",e,t,n),k("perm",e,t,n))];case"_FusedMatMul":let[r,s]=k("fusedOps",e,t,n),i=r==="biasadd",o=s==="prelu",l=k("numArgs",e,t,n),u=k("leakyreluAlpha",e,t,n);if(i){if(o&&l!==2)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&l!==1)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[p,d]=k("args",e,t,n);return[a.fused.matMul({a:k("a",e,t,n),b:k("b",e,t,n),transposeA:k("transposeA",e,t,n),transposeB:k("transposeB",e,t,n),bias:p,activation:s,preluActivationWeights:d,leakyreluAlpha:u})];default:throw TypeError(`Node type ${e.op} is not implemented`)}},x6=(e,t,n,a=pn)=>{switch(e.op){case"EuclideanNorm":return[a.euclideanNorm(k("x",e,t,n),k("axis",e,t,n),k("keepDims",e,t,n))];case"FusedBatchNorm":case"FusedBatchNormV2":return[a.batchNorm(k("x",e,t,n),k("mean",e,t,n),k("variance",e,t,n),k("offset",e,t,n),k("scale",e,t,n),k("epsilon",e,t,n))];case"FusedBatchNormV3":return[a.batchNorm(k("x",e,t,n),k("mean",e,t,n),k("variance",e,t,n),k("offset",e,t,n),k("scale",e,t,n),k("epsilon",e,t,n))];case"LRN":return[a.localResponseNormalization(k("x",e,t,n),k("radius",e,t,n),k("bias",e,t,n),k("alpha",e,t,n),k("beta",e,t,n))];case"Softmax":return[a.softmax(k("x",e,t,n))];case"LogSoftmax":return[a.logSoftmax(k("x",e,t,n))];case"SparseToDense":return[a.sparseToDense(k("sparseIndices",e,t,n),k("outputShape",e,t,n),k("sparseValues",e,t,n),k("defaultValue",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},v6=(e,t,n,a=pn)=>{switch(e.op){case"Max":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.max(k("x",e,t,n),o,l)]}case"Mean":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.mean(k("x",e,t,n),o,l)]}case"Min":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.min(k("x",e,t,n),o,l)]}case"Sum":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.sum(k("x",e,t,n),o,l)]}case"All":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.all(k("x",e,t,n),o,l)]}case"Any":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.any(k("x",e,t,n),o,l)]}case"ArgMax":{let o=k("axis",e,t,n);return[a.argMax(k("x",e,t,n),o)]}case"ArgMin":{let o=k("axis",e,t,n);return[a.argMin(k("x",e,t,n),o)]}case"Prod":{let o=k("axis",e,t,n),l=k("keepDims",e,t,n);return[a.prod(k("x",e,t,n),o,l)]}case"Cumprod":{let o=k("axis",e,t,n),l=k("exclusive",e,t,n),u=k("reverse",e,t,n);return[a.cumprod(k("x",e,t,n),o,l,u)]}case"Cumsum":{let o=k("axis",e,t,n),l=k("exclusive",e,t,n),u=k("reverse",e,t,n);return[a.cumsum(k("x",e,t,n),o,l,u)]}case"Bincount":let r=k("x",e,t,n),s=k("weights",e,t,n),i=k("size",e,t,n);return[a.bincount(r,s,i)];case"DenseBincount":{let o=k("x",e,t,n),l=k("weights",e,t,n),u=k("size",e,t,n),p=k("binaryOutput",e,t,n);return[a.denseBincount(o,l,u,p)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},w6=(e,t,n,a=pn)=>{switch(e.op){case"ConcatV2":case"Concat":{let r=k("n",e,t,n),s=k("axis",e,t,n),i=k("tensors",e,t,n);return i=i.slice(0,r),[a.concat(i,s)]}case"Gather":{let r=k("x",e,t,n),s=k("indices",e,t,n);return[a.gather(r,a.cast(s,"int32"),0)]}case"GatherV2":{let r=k("axis",e,t,n),s=k("batchDims",e,t,n),i=k("x",e,t,n),o=k("indices",e,t,n);return[a.gather(i,a.cast(o,"int32"),r,s)]}case"Reverse":{let r=k("dims",e,t,n),s=[];for(let o=0;o<r.length;o++)r[o]&&s.push(o);let i=k("x",e,t,n);return[a.reverse(i,s)]}case"ReverseV2":{let r=k("axis",e,t,n),s=k("x",e,t,n);return[a.reverse(s,r)]}case"Slice":{let r=k("begin",e,t,n),s=k("size",e,t,n);return[a.slice(k("x",e,t,n),r,s)]}case"StridedSlice":{let r=k("begin",e,t,n),s=k("end",e,t,n),i=k("strides",e,t,n),o=k("beginMask",e,t,n),l=k("endMask",e,t,n),u=k("ellipsisMask",e,t,n),p=k("newAxisMask",e,t,n),d=k("shrinkAxisMask",e,t,n),c=k("x",e,t,n);return[a.stridedSlice(c,r,s,i,o,l,u,p,d)]}case"Pack":return P(()=>{let r=k("axis",e,t,n),s=k("tensors",e,t,n),i=s[0].shape,o=a.squeeze(s[0]).shape,l=s.map(u=>{let p=v.arraysEqual(u.shape,i);if(!p&&!v.arraysEqual(a.squeeze(u).shape,o))throw new Error("the input tensors shape does not match");return p?u:a.reshape(u,i)});return[a.stack(l,r)]});case"Unpack":{let r=k("axis",e,t,n),s=k("tensor",e,t,n);return a.unstack(s,r)}case"Tile":{let r=k("reps",e,t,n);return[a.tile(k("x",e,t,n),r)]}case"Split":case"SplitV":{let r=k("axis",e,t,n),s=k("numOrSizeSplits",e,t,n),i=k("x",e,t,n);return a.split(i,s,r)}case"ScatterNd":{let r=k("indices",e,t,n),s=k("values",e,t,n),i=k("shape",e,t,n);return[a.scatterND(r,s,i)]}case"GatherNd":{let r=k("x",e,t,n),s=k("indices",e,t,n);return[a.gatherND(r,s)]}case"SparseToDense":{let r=k("sparseIndices",e,t,n),s=k("outputShape",e,t,n),i=k("sparseValues",e,t,n),o=k("defaultValue",e,t,n);return[a.sparseToDense(r,i,s,i.dtype===o.dtype?o:a.cast(o,i.dtype))]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},k6=(e,t,n,a=pn)=>{switch(e.op){case"SparseFillEmptyRows":{let{outputIndices:r,outputValues:s,emptyRowIndicator:i,reverseIndexMap:o}=a.sparse.sparseFillEmptyRows(k("indices",e,t,n),k("values",e,t,n),k("denseShape",e,t,n),k("defaultValue",e,t,n));return[r,s,i,o]}case"SparseReshape":{let{outputIndices:r,outputShape:s}=a.sparse.sparseReshape(k("inputIndices",e,t,n),k("inputShape",e,t,n),k("newShape",e,t,n));return[r,s]}case"SparseSegmentMean":return[a.sparse.sparseSegmentMean(k("data",e,t,n),k("indices",e,t,n),k("segmentIds",e,t,n))];case"SparseSegmentSum":return[a.sparse.sparseSegmentSum(k("data",e,t,n),k("indices",e,t,n),k("segmentIds",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},I6=(e,t,n,a=pn)=>{switch(e.op){case"FFT":return[a.fft(k("x",e,t,n))];case"IFFT":return[a.ifft(k("x",e,t,n))];case"RFFT":return[a.rfft(k("x",e,t,n))];case"IRFFT":return[a.irfft(k("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},S6=(e,t,n,a=pn)=>{switch(e.op){case"StringNGrams":{let{nGrams:r,nGramsSplits:s}=a.string.stringNGrams(k("data",e,t,n),k("dataSplits",e,t,n),k("separator",e,t,n),k("nGramWidths",e,t,n),k("leftPad",e,t,n),k("rightPad",e,t,n),k("padWidth",e,t,n),k("preserveShortSequences",e,t,n));return[r,s]}case"StringSplit":{let{indices:r,values:s,shape:i}=a.string.stringSplit(k("input",e,t,n),k("delimiter",e,t,n),k("skipEmpty",e,t,n));return[r,s,i]}case"StringToHashBucketFast":return[a.string.stringToHashBucketFast(k("input",e,t,n),k("numBuckets",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},T6=(e,t,n,a=pn)=>{switch(e.op){case"Cast":return[a.cast(k("x",e,t,n),k("dtype",e,t,n))];case"ExpandDims":{let r=k("axis",e,t,n);return[a.expandDims(k("x",e,t,n),r)]}case"Squeeze":{let r=k("axis",e,t,n);return[a.squeeze(k("x",e,t,n),r)]}case"Reshape":return[a.reshape(k("x",e,t,n),k("shape",e,t,n))];case"MirrorPad":return[a.mirrorPad(k("x",e,t,n),k("padding",e,t,n),k("mode",e,t,n))];case"PadV2":case"Pad":return[a.pad(k("x",e,t,n),k("padding",e,t,n),k("constantValue",e,t,n))];case"SpaceToBatchND":{let r=k("blockShape",e,t,n),s=k("paddings",e,t,n);return[a.spaceToBatchND(k("x",e,t,n),r,s)]}case"BatchToSpaceND":{let r=k("blockShape",e,t,n),s=k("crops",e,t,n);return[a.batchToSpaceND(k("x",e,t,n),r,s)]}case"DepthToSpace":{let r=k("blockSize",e,t,n),s=k("dataFormat",e,t,n).toUpperCase();return[a.depthToSpace(k("x",e,t,n),r,s)]}case"BroadcastTo":return[a.broadcastTo(k("x",e,t,n),k("shape",e,t,n))];case"BroadcastArgs":return[a.broadcastArgs(k("s0",e,t,n),k("s1",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function eI(e,t,n,a,r=P){let s=((i,o,l)=>{switch(i.category){case"arithmetic":return r(()=>t6(i,o,l));case"basic_math":return r(()=>n6(i,o,l));case"control":return l6(i,o,l);case"convolution":return r(()=>u6(i,o,l));case"creation":return r(()=>p6(i,o,l));case"dynamic":return c6(i,o,l);case"evaluation":return r(()=>d6(i,o,l));case"image":return r(()=>g6(i,o,l));case"graph":return r(()=>h6(i,o,l));case"logical":return r(()=>y6(i,o,l));case"matrices":return r(()=>b6(i,o,l));case"normalization":return r(()=>x6(i,o,l));case"reduction":return r(()=>v6(i,o,l));case"slice_join":return r(()=>w6(i,o,l));case"sparse":return r(()=>k6(i,o,l));case"spectral":return r(()=>I6(i,o,l));case"string":return r(()=>S6(i,o,l));case"transformation":return r(()=>T6(i,o,l));case"hash_table":return f6(i,o,l,a);case"custom":let u=w2(i.op);if(u&&u.customExecutor)return u.customExecutor(new e6(i,o,l));throw TypeError(`Custom op ${i.op} is not registered.`);default:throw TypeError(`Unknown op '${i.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,n);return v.isPromise(s)?s.then(i=>[].concat(i)):[].concat(s)}var tI=class{constructor(e={},t={},n={},a={}){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=n,this.functionMap=a,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let e=[];for(let t=0;t<this.contexts.length-1;t++){let n=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(n))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(t=>t.id===0&&t.iterationId===0?"":`${t.frameName}-${t.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error("Cannot exit frame, the context is empty")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error("Cannot increase frame iteration, the context is empty")}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(let t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(let t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function nI(e,t,n,a){let r=new Set,s=[],i=null,o=null,l=new Set,u=Object.keys(e).map(c=>Yn(c)[0]),p=[];a!=null&&(p=a.map(c=>Yn(c.name)[0]));let d=[...t];for(;d.length>0;){let c=d.pop();if((U2(c)||A6(c)||$6(c))&&i==null&&(i=c,o=i.children.map(h=>h.name).filter(h=>r.has(h))),r.add(c.name),n[c.name]==null&&u.indexOf(c.name)===-1&&p.indexOf(c.name)===-1){if(c.inputs.length===0){s.push(c.name);continue}c.inputs.forEach(h=>{l.has(h.name)||(l.add(h.name),d.push(h))})}}return{inputs:e,outputs:t,usedNodes:r,missingInputs:s,dynamicNode:i,syncInputs:o}}function N6(e,t,n){let{usedNodes:a,inputs:r}=n,s=[],i=Object.keys(r).map(p=>Yn(p)[0]).map(p=>e.nodes[p]),o=e.initNodes;i.forEach(p=>{a.has(p.name)&&s.push(p)}),e.weights.forEach(p=>{a.has(p.name)&&s.push(p)}),o!=null&&o.forEach(p=>{a.has(p.name)&&s.push(p)});let l=new Set,u=[];for(;s.length>0;){let p=s.pop();l.add(p.name),t[p.name]||u.push(p),p.children.forEach(d=>{!l.has(d.name)&&a.has(d.name)&&d.inputs.every(c=>l.has(c.name))&&s.push(d)})}return u}var C6=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],_6=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],E6=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function U2(e){return C6.indexOf(e.op)>=0}function A6(e){return _6.indexOf(e.op)>=0}function $6(e){return E6.indexOf(e.op)>=0}var px=class{constructor(e,t){this.graph=e,this.parent=t,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=e.outputs,this._inputs=e.inputs,this._initNodes=e.initNodes,this._signature=e.signature,this._functions=e.functions,e.functions!=null&&Object.keys(e.functions).forEach(n=>{this._functionExecutorMap[n]=new px(e.functions[n],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){let t=Object.keys(e).map(n=>e[n].map(a=>a.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{let t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}getCompilationKey(e,t){let n=e.map(r=>r.name).sort(),a=t.map(r=>r.name).sort();return n.join(this.SEPERATOR)+"--"+a.join(this.SEPERATOR)}compile(e,t){let n=nI(e,t,this.weightMap,this._initNodes),{missingInputs:a,dynamicNode:r,syncInputs:s}=n;if(r!=null)throw new Error(`This execution contains the node '${r.name}', which has the dynamic op '${r.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);if(a.length>0){let i=t.map(l=>l.name),o=Object.keys(e);throw new Error(`Cannot compute the outputs [${i}] from the provided inputs [${o}]. Missing the following inputs: [${a}]`)}return N6(this.graph,this.weightMap,n)}execute(e,t){e=this.mapInputs(e);let n=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);let a=n.map(p=>this.graph.nodes[Yn(p)[0]]),r=t.map(p=>Yn(p)[0]),s=r.map(p=>this.graph.nodes[p]);this.resetIntermediateTensors(),s.length===0&&(s=this._outputs);let i=this.getCompilationKey(a,s),o=this.compiledMap.get(i);o==null&&(o=this.compile(e,s),this.compiledMap.set(i,o));let l={},u={};return P(()=>{let p=new tI(this.weightMap,l,u,this.functionExecutorMap),d=Object.assign({},this.weightMap);Object.keys(e).forEach(m=>{let[f,g]=Yn(m),y=[];y[g]=e[m],d[f]=y});let c=this.getFrozenTensorIds(d),h={};for(let m=0;m<o.length;m++){let f=o[m];if(!d[f.name]){let g=eI(f,d,p,this._resourceManager);if(v.isPromise(g))throw new Error(`The execution of the op '${f.op}' returned a promise. Please use model.executeAsync() instead.`);d[f.name]=g,this.checkTensorForDisposal(f.name,f,d,p,c,r,h)}}return this.parent==null&&p.dispose(c),t.map(m=>wn(m,d,p))})}getFrozenTensorIds(e){let t=[].concat.apply([],Object.keys(e).map(n=>e[n]).map(n=>n.map(a=>a.id)));return new Set(t)}checkTensorForDisposal(e,t,n,a,r,s,i){t.category==="control"||s.indexOf(e)!==-1||(n[e].forEach(o=>{o!=null&&(i[o.id]=(i[o.id]||0)+t.children.length)}),t.inputs.forEach(o=>{if(o.category!=="control"){let l=DH(o.name,n,a);l!=null&&l.forEach(u=>{if(u&&!u.kept&&!r.has(u.id)){let p=i[u.id];if(p===1){if(!this.keepTensorForDebug)u.dispose();else{let[d,c]=sr(t.name,a);this.intermediateTensors[d]?this.intermediateTensors[d][c]=u:(this.intermediateTensors[d]=[],this.intermediateTensors[d][c]=u)}delete i[u.id]}else p!=null&&i[u.id]--}})}}))}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(e=>this.intermediateTensors[e].forEach(t=>t.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(e=>{this.tensorsMap[e].forEach(t=>{t&&!t.kept&&!t.isDisposed&&!this.keepIds.has(t.id)&&t.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let e in this.intermediateTensors)this.intermediateTensors[e].forEach(t=>t.dispose()),delete this.intermediateTensors[e]}async _executeAsync(e,t,n=!1,a={},r={}){n||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepTensorForDebug=H().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(u){console.warn(u.message)}this.resetIntermediateTensors();let s=new tI(this.weightMap,a,r,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(e,s,t,n);let i=t.map(u=>wn(u,this.tensorsMap,s)),o=i.map(u=>u.id),l=Object.keys(e).map(u=>e[u].id);return this.keepIds=new Set([...o,...l,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&s.dispose(this.keepIds),i}async executeFunctionAsync(e,t,n){let a=e.reduce((r,s,i)=>(r[this.inputs[i].name]=s,r),{});return this._executeAsync(a,this.outputNodes,!0,t,n)}async executeWithControlFlow(e,t,n,a){let r=Object.keys(e),s=r.map(b=>this.graph.nodes[Yn(b)[0]]),i=n.map(b=>Yn(b)[0]),o=i.map(b=>this.graph.nodes[b]);o.length===0&&(o=this._outputs);let{usedNodes:l,missingInputs:u,dynamicNode:p,syncInputs:d}=nI(e,o,this.weightMap,this._initNodes),c=[...s,...this.graph.weights,...this._initNodes||[]].map(b=>({node:b,contexts:t.currentContext})),h=Object.assign({},this.weightMap);Object.keys(e).forEach(b=>{let[x,w]=Yn(b),I=[];I[w]=e[b],h[x]=I});let m={},f=this.getFrozenTensorIds(h),g={};for(;c.length>0;){let b=this.processStack(s,c,t,h,g,f,i,m,l);await Promise.all(b)}p==null&&!a&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let y=o.filter(b=>!U2(b)&&!wn(b.name,h,t)).map(b=>b.name);if(y.length>0){let b="";throw p!=null&&(b=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${d}]`),new Error(`Cannot compute the outputs [${y}] from the provided inputs [${r}]. Consider providing the following inputs: [${u}]. ${b}`)}return h}processStack(e,t,n,a,r,s,i,o,l){let u=[];for(;t.length>0;){let p=t.pop();n.currentContext=p.contexts;let d="";if(p.node.op==="Enter"&&k("isConstant",p.node,a,n)&&([d]=sr(p.node.name,n)),a[p.node.name]==null){let c=eI(p.node,a,n,this._resourceManager);d||([d]=sr(p.node.name,n));let h=n.currentContext;v.isPromise(c)?u.push(c.then(m=>(a[d]=m,n.currentContext=h,this.checkTensorForDisposal(d,p.node,a,n,s,i,o),this.processChildNodes(p.node,t,n,a,r,l),m))):(a[d]=c,this.checkTensorForDisposal(d,p.node,a,n,s,i,o),this.processChildNodes(p.node,t,n,a,r,l))}else this.processChildNodes(p.node,t,n,a,r,l)}return u}processChildNodes(e,t,n,a,r,s){e.children.forEach(i=>{let[o]=sr(i.name,n);r[o]||!s.has(i.name)||(i.op==="Merge"?i.inputNames.some(l=>!!wn(l,a,n))&&(r[o]=!0,t.push({contexts:n.currentContext,node:i})):i.inputNames.every(l=>!!wn(l,a,n))&&(r[o]=!0,t.push({contexts:n.currentContext,node:i})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(t=>t.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{let n=e[t],[a]=Yn(t),r=this.graph.nodes[a];if(r.attrParams.shape&&r.attrParams.shape.value){let s=r.attrParams.shape.value,i=s.length===n.shape.length&&n.shape.every((o,l)=>s[l]===-1||s[l]===o);v.assert(i,()=>`The shape of dict['${r.name}'] provided in model.execute(dict) must be [${s}], but was [${n.shape}]`)}r.attrParams.dtype&&r.attrParams.dtype.value&&v.assert(n.dtype===r.attrParams.dtype.value,()=>`The dtype of dict['${r.name}'] provided in model.execute(dict) must be ${r.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(e){let t={};for(let n in e)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[n]!=null){let a=this._signature.inputs[n];t[a.name]=e[n]}else t[n]=e[n];return t}checkInputs(e){let t=Object.keys(e).filter(n=>{let[a]=Yn(n);return this.graph.nodes[a]==null});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(t=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[t]!=null?this._signature.outputs[t].name:t,{})}checkOutputs(e){e.forEach(t=>{let[n]=Yn(t);if(!this.graph.nodes[n])throw new Error(`The output '${t}' is not found in the graph`)})}},F6=class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(let e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(let e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}},D6="?tfjs-format=file",R6="model.json",A0=class{constructor(e,t={},n=Ut){this.modelUrl=e,this.loadOptions=t,this.version="n/a",this.io=n,t==null&&(this.loadOptions={}),this.resourceManager=new F6}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}findIOHandler(){let e=this.modelUrl;if(e.load!=null)this.handler=e;else if(this.loadOptions.requestInit!=null)this.handler=this.io.browserHTTPRequest(e,this.loadOptions);else{let t=this.io.getLoadHandlers(e,this.loadOptions);if(t.length===0)t.push(this.io.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}load(){if(this.findIOHandler(),this.handler.load==null)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let e=this.handler.load();return v.isPromise(e)?e.then(t=>this.loadSync(t)):this.loadSync(e)}loadSync(e){this.artifacts=e;let t=this.artifacts.modelTopology,n=this.artifacts.signature;if(this.artifacts.userDefinedMetadata!=null){let r=this.artifacts.userDefinedMetadata;r.signature!=null&&(n=r.signature),r.structuredOutputKeys!=null&&(this.structuredOutputKeys=r.structuredOutputKeys)}this.signature=n,this.version=`${t.versions.producer}.${t.versions.minConsumer}`;let a=this.io.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new px(Yk.Instance.transformGraph(t,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(a),this.executor.resourceManager=this.resourceManager,e.modelInitializer!=null&&e.modelInitializer.node!=null){let r=Yk.Instance.transformGraph(e.modelInitializer);this.initializer=new px(r),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=e.initializerSignature}return!0}async save(e,t){if(typeof e=="string"){let n=this.io.getSaveHandlers(e);if(n.length===0)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(n.length>1)throw new Error(`Found more than one (${n.length}) save handlers for URL '${e}'`);e=n[0]}if(e.save==null)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}predict(e,t){let n=this.execute(e,this.outputNodes);if(this.structuredOutputKeys){let a=n instanceof Te?[n]:n,r={};return a.forEach((s,i)=>r[this.structuredOutputKeys[i]]=s),r}return n}normalizeInputs(e){if(!(e instanceof Te)&&!Array.isArray(e)){if(this.signature!=null&&this.signature.inputs!=null)for(let a in this.signature.inputs){let r=this.signature.inputs[a];r.resourceId!=null&&(e[a]=this.resourceIdToCapturedInput[r.resourceId])}return e}e=Array.isArray(e)?e:[e];let t=Object.keys(this.resourceIdToCapturedInput).length;if(e.length+t!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-t} non-resource placeholders, while there are ${e.length} input tensors provided.`);let n=0;return this.inputNodes.reduce((a,r)=>{let s=this.signature?this.signature.inputs[r]:null;return s!=null&&s.resourceId!=null?a[r]=this.resourceIdToCapturedInput[s.resourceId]:a[r]=e[n++],a},{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}executeInitializerGraph(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(e){if(this.resourceIdToCapturedInput={},this.initializerSignature){let t=Object.keys(this.initializerSignature.outputs);for(let n=0;n<t.length;n++){let a=t[n],r=this.initializerSignature.outputs[a];this.resourceIdToCapturedInput[r.resourceId]=e[n]}}}execute(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=this.executor.execute(e,t);return n.length>1?n:n[0]}async executeAsync(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=await this.executor.executeAsync(e,t);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,n)=>(t[n]=[e[n]],t),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&_e(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function M6(e,t={},n=Ut){if(e==null)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");t==null&&(t={}),t.fromTFHub&&typeof e=="string"&&(e=O6(e));let a=new A0(e,t,n);return await a.load(),a}function P6(e){if(e==null)throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");let t;if(e instanceof Array){let[a,r]=e;if(!a)throw new Error("modelJSON must be the first element of the array");if(!r||!(r instanceof ArrayBuffer))throw new Error("An ArrayBuffer of weights must be the second element of the array");if(!("modelTopology"in a))throw new Error("Model JSON is missing 'modelTopology'");if(!("weightsManifest"in a))throw new Error("Model JSON is missing 'weightsManifest'");let s=Ut.getWeightSpecs(a.weightsManifest),i=Ut.getModelArtifactsForJSONSync(a,s,r);t=Ut.fromMemorySync(i)}else if("load"in e)t=e;else if("modelTopology"in e&&"weightSpecs"in e&&"weightData"in e)t=Ut.fromMemorySync(e);else throw new Error("Unknown model format");let n=new A0(t);return n.load(),n}function O6(e){return e.endsWith("/")||(e=e+"/"),`${e}${R6}${D6}`}var L6="4.0.0",G2={};Ae(G2,{CSVDataset:()=>Q2,Dataset:()=>qu,FileDataSource:()=>iC,TextLineDataset:()=>J2,URLDataSource:()=>oC,array:()=>oj,csv:()=>bj,func:()=>xj,generator:()=>vj,microphone:()=>kj,version_data:()=>Ij,webcam:()=>wj,zip:()=>lj});var z6=ms(im()),W6=ms(im());function B6(e,t){return Yh(e,t)}function Yh(e,t,n=new Map,a=new Set){if(e==null)return null;if(typeof Blob=="function"&&e instanceof Blob)return e.slice();if(a.has(e))throw new Error("Circular references are not supported.");if(n.has(e))return n.get(e);let r=t(e);if(r.recurse&&r.value!==null)throw new Error("A deep map function may not return both a value and recurse=true.");if(r.recurse)if(Cl(e)){let s=Array.isArray(e)?[]:{};a.add(e);for(let i in e){let o=e[i],l=Yh(o,t,n,a);s[i]=l}return a.delete(e),e.__proto__&&(s.__proto__=e.__proto__),s}else throw new Error(`Can't recurse into non-iterable type: ${e}`);else return n.set(e,r.value),r.value}function V6(e,t=j2){return H2(e,t)}function H2(e,t,n=new Set){let a=e[0];if(n.has(a))throw new Error("Circular references are not supported.");let r=t(e);if(r.recurse&&r.value!==null)throw new Error("A deep zip function may not return both a value and recurse=true.");if(r.recurse)if(Cl(a)){let s=Array.isArray(a)?[]:{};n.add(a);for(let i in a){let o=e.map(u=>u[i]),l=H2(o,t,n);s[i]=l}return n.delete(a),s}else throw new Error(`Can't recurse into non-iterable type: ${a}`);else return r.value}function j2(e){return e===null?null:Cl(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function q2(e,t){let n=new Map;Yh(e,t,n);for(let a of Array.from(n.keys())){let r=n.get(a);if(v.isPromise(r)){let s=await r;n.set(a,s)}}return Yh(e,t,n)}function Cl(e){let t=!1;if(H().get("IS_BROWSER"))t=e instanceof TextDecoder;else{let{StringDecoder:n}=VI();t=e instanceof n}return e!=null&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||typeof e=="object"&&!(e instanceof Te)&&!(e instanceof Promise)&&!t)}function U6(e){return e==null||G6(e)||Array.isArray(e)||typeof e=="object"&&e instanceof Te||v.isTypedArray(e)}function G6(e){return e===null||typeof e!="object"&&typeof e!="function"}function H6(e){return B6(e,j6)}function j6(e){return e instanceof Te?{value:e.clone(),recurse:!1}:Cl(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var K2=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,e==null)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(let t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.wrap(this.begin+e),n=this.get(t);return this.set(t,this.pop()),n}},$0=class extends K2{constructor(){super($0.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){let e=this.capacity*2,t=new Array(e),n=this.length();for(let a=0;a<n;a++)t[a]=this.get(this.wrap(this.begin+a));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};$0.INITIAL_CAPACITY=32;function X2(e){return new X6(e)}function F0(e){return new Y6(e)}function q6(e,t){return new Y2(e,t)}function K6(e,t=ts.FAIL){return new sj(e,t)}var tn=class{async toArray(){let e=[],t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){let e=this.prefetch(100),t=[],n=await e.next();for(;!n.done;)t.push(n.value),n=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),n=e(t.value);for(;!t.done&&n;)t=await this.next(),n=e(t.value)}handleErrors(e){return new aj(this,e)}filter(e){return new tj(this,e)}map(e){return new nj(this,e)}mapAsync(e){return new aI(this,e)}serialMapAsync(e){return new aI(this,e).serial()}flatmap(e){return new rj(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(t=>t===!0)}rowMajorBatch(e,t=!0){return new ej(this,e,t)}columnMajorBatch(e,t=!0,n=j2){return this.rowMajorBatch(e,t).map(a=>V6(a,n))}concatenate(e,t){return new Y2(X2([this,e]),t)}take(e){return e<0||e==null?this:new Q6(this,e)}skip(e){return e<0||e==null?this:new J6(this,e)}prefetch(e){return new Z2(this,e)}shuffle(e,t){return new ij(this,e,t)}serial(){return new Z6(this)}},X6=class extends tn{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let e=this.items[this.trav];return this.trav++,{value:H6(e),done:!1}}},Y6=class extends tn{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message=`Error thrown while iterating through a dataset: ${e.message}`,e}}},Z6=class extends tn{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},J6=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let e=await this.upstream.next();if(e.done)return e;_e(e.value)}return this.upstream.next()}},Q6=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},ej=class extends tn{constructor(e,t,n=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let e=[];for(;e.length<this.batchSize;){let t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},tj=class extends tn{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;_e(e.value)}}},nj=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Map`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Va.getTensorsInContainer(e.value),n=this.transform(e.value),a=Va.getTensorsInContainer(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return{value:n,done:!1}}},aj=class extends tn{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},aI=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Va.getTensorsInContainer(e.value),n=await this.transform(e.value),a=Va.getTensorsInContainer(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return{value:n,done:!1}}},D0=class extends tn{constructor(){super(),this.outputQueue=new $0,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},rj=class extends D0{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let e=await this.upstream.next();if(e.done)return!1;let t=Va.getTensorsInContainer(e.value),n=this.transform(e.value),a=Va.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let r of t)Va.isTensorInList(r,a)||r.dispose();return!0}},Y2=class extends tn{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,this.iterator==null){let n=await this.moreIterators.next();if(n.done)return{value:null,done:!0};this.iterator=n.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}},ts;(function(e){e[e.FAIL=0]="FAIL",e[e.SHORTEST=1]="SHORTEST",e[e.LONGEST=2]="LONGEST"})(ts||(ts={}));var sj=class extends tn{constructor(e,t=ts.FAIL){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,n=0;function a(s){return s instanceof tn?{value:s.next().then(i=>(t++,i.done&&n++,i.value)),recurse:!1}:{value:null,recurse:!0}}let r=await q2(this.iterators,a);if(t===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case ts.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case ts.SHORTEST:return{value:null,done:!0};case ts.LONGEST:default:}return this.count++,{value:r,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},Z2=class extends tn{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new K2(t)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},ij=class extends Z2{constructor(e,t,n){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=W6.alea(n||v.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(t.done)this.upstreamExhausted=!0;else return this.refill(),t}return{value:null,done:!0}}},qu=class{constructor(){this.size=null}batch(e,t=!0){let n=this;v.assert(e>0,()=>`batchSize needs to be positive, but it is
-      ${e}`);let a;return this.size===1/0||this.size==null?a=this.size:t?a=Math.ceil(this.size/e):a=Math.floor(this.size/e),Xn(async()=>(await n.iterator()).columnMajorBatch(e,t,uj),a)}concatenate(e){let t=this,n;return this.size===1/0||e.size===1/0?n=1/0:this.size!=null&&e.size!=null?n=this.size+e.size:n=null,Xn(async()=>(await t.iterator()).concatenate(await e.iterator()),n)}filter(e){let t=this,n;return this.size===1/0?n=1/0:n=null,Xn(async()=>(await t.iterator()).filter(a=>P(()=>e(a))),n)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){let t=this;return Xn(async()=>(await t.iterator()).map(n=>P(()=>e(n))),this.size)}mapAsync(e){let t=this;return Xn(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(e==null)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let t=this;return Xn(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){let t=this,n;return this.size!=null&&e>0?n=this.size*e:e===0?n=0:this.size!=null&&(e===void 0||e<0)?n=1/0:n=null,Xn(async()=>{let a=F0(async()=>({value:await t.iterator(),done:!1}));return q6(a.take(e))},n)}skip(e){let t=this,n;return this.size!=null&&e>=0&&this.size>=e?n=this.size-e:this.size!=null&&(this.size<e||e===void 0||e<0)?n=0:n=null,Xn(async()=>(await t.iterator()).skip(e),n)}shuffle(e,t,n=!0){if(e==null||e<0)throw this.size==null?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let a=this,r=z6.alea(t||v.now().toString());return Xn(async()=>{let s=r.int32();return n&&(s+=r.int32()),(await a.iterator()).shuffle(e,s.toString())},this.size)}take(e){let t=this,n;return this.size!=null&&this.size>e?n=e:this.size!=null&&this.size<=e?n=this.size:n=null,Xn(async()=>(await t.iterator()).take(e),n)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};qu.MAX_BUFFER_SIZE=1e4;function Xn(e,t=null){return new class extends qu{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function oj(e){return Xn(async()=>X2(e),e.length)}function lj(e){if(!Cl(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let n=0;n<e.length;n++)t=t==null?e[n].size:Math.min(t,e[n].size);else if(e instanceof Object)for(let n in e)t=t==null?e[n].size:Math.min(t,e[n].size);return Xn(async()=>{let n=await q2(e,a=>{if(a instanceof qu)return{value:a.iterator(),recurse:!1};if(Cl(a))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")});return K6(n,ts.SHORTEST)},t)}function uj(e){if(e===null)return null;let t=e[0];return U6(t)?{value:pj(e),recurse:!1}:{value:null,recurse:!0}}function pj(e){if(e.length===0)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof Te?Ft(e):In(e)}var J2=class extends qu{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`
-`).map(e=>(e.endsWith("\r")&&(e=e.slice(0,-1)),e))}},dh='"',Fp=Symbol("out"),rI=Symbol("field"),hh=Symbol("quote"),Ib=Symbol("quoteafterquote"),sI=Symbol("quoteinquote"),Q2=class extends qu{constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new J2(e),t||(t={}),this.hasHeader=t.hasHeader!==!1,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(v.assert(t.delimiter==null,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&v.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);let t=this.fullColumnNames.reduce((a,r)=>(a[r]=a[r]+1||1,a),{}),n=Object.keys(t).filter(a=>t[a]>1);if(v.assert(n.length===0,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs){for(let a of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(a)===-1)throw new Error('The key "'+a+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error("No data was found for CSV parsing.");let t=e.value;return this.parseRow(t,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(t=>this.makeDataElement(t))}makeDataElement(e){let t=this.parseRow(e),n={},a={};for(let r=0;r<this.fullColumnNames.length;r++){let s=this.fullColumnNames[r],i=this.columnConfigs?this.columnConfigs[s]:null;if(!(this.configuredColumnsOnly&&!i)){let o=t[r],l=null;if(o==="")if(i&&i.default!==void 0)l=i.default;else{if(i&&(i.required||i.isLabel))throw new Error(`Required column ${s} is empty in this line: ${e}`);l=void 0}else{let u=Number(o);if(isNaN(u))i&&i.dtype==="bool"?l=this.getBoolean(o):l=o;else if(!i||!i.dtype)l=u;else switch(i.dtype){case"float32":l=u;break;case"int32":l=Math.floor(u);break;case"bool":l=this.getBoolean(o);break;default:l=u}}i&&i.isLabel?a[s]=l:n[s]=l}}return Object.keys(a).length===0?n:{xs:n,ys:a}}getBoolean(e){return e==="1"||e.toLowerCase()==="true"?1:0}parseRow(e,t=!0){let n=[],a=0,r=e.length,s=Fp;for(let i=0;i<r;i++)switch(s){case Fp:switch(e.charAt(i)){case dh:a=i+1,s=hh;break;case this.delimiter:if(a=i+1,this.delimiter===" "&&this.delimWhitespace)break;n.push(""),s=Fp;break;default:s=rI,a=i;break}break;case rI:switch(e.charAt(i)){case this.delimiter:n.push(e.substring(a,i)),s=Fp,a=i+1;break;default:}break;case hh:switch(e.charAt(i)){case dh:s=Ib;break;default:}break;case Ib:switch(e.charAt(i)){case this.delimiter:n.push(e.substring(a,i-1)),s=Fp,a=i+1;break;case dh:s=hh;break;default:s=sI;break}break;case sI:switch(e.charAt(i)){case dh:s=hh;break;default:}break;default:}if(s===Ib?n.push(e.substring(a,r-1)):n.push(e.substring(a)),t&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}},eC=class extends tn{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;let t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=e.includeSpectrogram!==!1,this.includeWaveform=e.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(e={}){if(!H().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let t=new eC(e);return await t.start(),t}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(n){throw new Error(`Error thrown while initializing video stream: ${n.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t,n=await this.getAudioData();if(this.includeSpectrogram){let a=this.flattenQueue(n.freqDataQueue);e=this.getTensorFromAudioDataArray(a,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let a=this.flattenQueue(n.timeDataQueue);t=this.getTensorFromAudioDataArray(a,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let e=[],t=[],n=0;return new Promise(a=>{let r=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&a({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(r),a({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){let t=e[0].length,n=new Float32Array(e.length*t);return e.forEach((a,r)=>n.set(a,r*t)),n}getTensorFromAudioDataArray(e,t){let n=new Float32Array(v.sizeFromShape(t));return n.set(e,n.length-e.length),In(n,t)}},tC=class extends tn{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Ke([0],"int32"),this.webcamConfig.centerCrop){let n=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,a=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,r=(1-n)/2,s=(1-a)/2,i=r+n,o=a+s;this.cropBox=Aa([s,r,o,i],[1,4])}else this.cropBox=Aa([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(e,t={}){if(!H().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!e){if(e=document.createElement("video"),!t.resizeWidth||!t.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");e.width=t.resizeWidth,e.height=t.resizeHeight}let n=new tC(e,t);return await n.start(),n}async start(){this.webcamConfig.facingMode&&v.assert(this.webcamConfig.facingMode==="user"||this.webcamConfig.facingMode==="environment",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message=`Error thrown while initializing video stream: ${e.message}`,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=ho.fromPixels(this.webcamVideoElement)}catch(t){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(t)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(t){throw new Error(`Error thrown cropping the video: ${t.message}`)}finally{e.dispose()}else return{value:e,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(e){return P(()=>{let t=Zt(oe(e,"float32"),0),n;n=za.cropAndResize(t,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let a=n.shape;return W(n,a.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},nC=class{},aC=class extends tn{split(e){return new cj(this,e)}},cj=class extends aC{constructor(e,t){super(),this.upstream=e,this.impl=new dj(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},dj=class extends D0{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let e=await this.upstream.next();if(e.done)return this.carryover===""?!1:(this.outputQueue.push(this.carryover),this.carryover="",!0);let t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(let n of t.slice(0,-1))this.outputQueue.push(n);return this.carryover=t[t.length-1],!0}},hj=class extends tn{decodeUTF8(){return new mj(this)}},mj=class extends aC{constructor(e){super(),this.upstream=e,this.impl=new fj(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},fj=class extends D0{constructor(e){if(super(),this.upstream=e,H().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:t}=VI();this.decoder=new t("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let e=await this.upstream.next(),t;if(e.done)return!1;t=e.value;let n;return H().get("IS_BROWSER")?n=this.decoder.decode(t,{stream:!0}):n=this.decoder.write(Buffer.from(t.buffer)),this.outputQueue.push(n),!0}},rC=class extends hj{constructor(e,t={}){super(),this.file=e,this.options=t,v.assert(e instanceof Uint8Array||(H().get("IS_BROWSER")?e instanceof File||e instanceof Blob:!1),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,t)=>{let n=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,n)));else{let a=new FileReader;a.onload=s=>{let i=a.result;if(i instanceof ArrayBuffer&&(i=new Uint8Array(i)),!(i instanceof Uint8Array))return t(new TypeError("FileReader returned unknown type."));e(i)},a.onabort=s=>t(new Error("Aborted")),a.onerror=s=>t(new Error(s.type));let r=this.file.slice(this.offset,n);a.readAsArrayBuffer(r)}this.offset=n}),done:!1}}};async function gj(e,t={},n){let a,r;typeof e=="string"?a=e:(a=e.url,r=yj(e));let s=await(n||v.fetch)(a,r);if(s.ok){let i=new Uint8Array(await s.arrayBuffer());return new rC(i,t)}else throw new Error(s.statusText)}var yj=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function sC(e){return typeof e=="string"&&e.slice(0,7)==="file://"}var iC=class extends nC{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(sC(this.input)&&H().get("IS_NODE")){let e=Cx();this.input=e.readFileSync(this.input.slice(7))}return new rC(this.input,this.options)}},oC=class extends nC{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return sC(this.url)?new iC(this.url,this.fileOptions).iterator():gj(this.url,this.fileOptions)}};function bj(e,t={}){return new Q2(new oC(e),t)}function xj(e){let t=F0(e);return Xn(async()=>t)}function vj(e){return Xn(async()=>{let t=await e();return F0(()=>t.next())})}async function wj(e,t){return tC.create(e,t)}async function kj(e){return eC.create(e)}var Ij="4.0.0";function ge(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&v.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var Sj=hr.whereImpl,R0=class extends pc{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new om(this,_a())}nextDataId(){return R0.nextDataId++}write(e,t,n){this.firstUse&&(this.firstUse=!1,H().get("IS_NODE")&&N.warn(`
+        ${r}, and tensor's shape is: ${e.shape}`);let a=e.shape.slice(1),o=uv(a,n),i=r===0?0:e.size/r,c=O(()=>{let l=[];e=W(e,[1,r,i]);for(let p=0;p<t.length;++p){let h=[0,p===0?0:s[p-1],0],f=[1,t[p],i];l[p]=W(We(e,h,f),o)}return e.dispose(),l}),u=new Cc([],n,e.dtype,t.length);for(let l=0;l<c.length;l++)u.setItem(l,c[l]);return u}var d6=async(e,t,n)=>{switch(e.op){case"If":case"StatelessIf":{let r=I("thenBranch",e,t,n),s=I("elseBranch",e,t,n),a=I("cond",e,t,n),o=I("args",e,t,n);return(await a.data())[0]?n.functionMap[r].executeFunctionAsync(o,n.tensorArrayMap,n.tensorListMap):n.functionMap[s].executeFunctionAsync(o,n.tensorArrayMap,n.tensorListMap)}case"While":case"StatelessWhile":{let r=I("body",e,t,n),s=I("cond",e,t,n),a=I("args",e,t,n),o=await n.functionMap[s].executeFunctionAsync(a,n.tensorArrayMap,n.tensorListMap),i=a.map(l=>l.id),c=await o[0].data();o.forEach(l=>{!l.kept&&i.indexOf(l.id)===-1&&l.dispose()});let u=a;for(;c[0];){let l=u;u=await n.functionMap[r].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);let p=u.map(h=>h.id);l.forEach(h=>{!h.kept&&i.indexOf(h.id)===-1&&p.indexOf(h.id)===-1&&h.dispose()});let d=await n.functionMap[s].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);c=await d[0].data(),d.forEach(h=>{!h.kept&&i.indexOf(h.id)===-1&&p.indexOf(h.id)===-1&&h.dispose()})}return u}case"LoopCond":{let r=I("pred",e,t,n);return[Ts(r)]}case"Switch":{let r=I("pred",e,t,n),s=I("data",e,t,n);return s.kept||(s=Ts(s)),(await r.data())[0]?[void 0,s]:[s,void 0]}case"Merge":{let r=e.inputNames.find(s=>Sn(s,t,n)!==void 0);if(r){let s=Sn(r,t,n);return[Ts(s)]}return}case"Enter":{let r=I("frameName",e,t,n),s=I("tensor",e,t,n);return n.enterFrame(r),[Ts(s)]}case"Exit":{let r=I("tensor",e,t,n);return n.exitFrame(),[Ts(r)]}case"NextIteration":{let r=I("tensor",e,t,n);return n.nextIteration(),[Ts(r)]}case"TensorArrayV3":{let r=I("size",e,t,n),s=I("dtype",e,t,n),a=I("elementShape",e,t,n),o=I("dynamicSize",e,t,n),i=I("clearAfterRead",e,t,n),c=I("identicalElementShapes",e,t,n),u=I("name",e,t,n),l=new o6(u,s,r,a,c,o,i);return n.addTensorArray(l),[l.idTensor,xe(1)]}case"TensorArrayWriteV3":{let r=I("tensorArrayId",e,t,n),s=I("index",e,t,n),a=I("tensor",e,t,n),o=n.getTensorArray(r.id);return o.write(s,a),[o.idTensor]}case"TensorArrayReadV3":{let r=I("tensorArrayId",e,t,n),s=I("index",e,t,n);return[n.getTensorArray(r.id).read(s)]}case"TensorArrayGatherV3":{let r=I("tensorArrayId",e,t,n),s=I("indices",e,t,n),a=I("dtype",e,t,n);return[n.getTensorArray(r.id).gather(s,a)]}case"TensorArrayScatterV3":{let r=I("tensorArrayId",e,t,n),s=I("indices",e,t,n),a=I("tensor",e,t,n),o=n.getTensorArray(r.id);return o.scatter(s,a),[o.idTensor]}case"TensorArrayConcatV3":{let r=I("tensorArrayId",e,t,n),s=n.getTensorArray(r.id),a=I("dtype",e,t,n);return[s.concat(a)]}case"TensorArraySplitV3":{let r=I("tensorArrayId",e,t,n),s=I("tensor",e,t,n),a=I("lengths",e,t,n),o=n.getTensorArray(r.id);return o.split(a,s),[o.idTensor]}case"TensorArraySizeV3":{let r=I("tensorArrayId",e,t,n),s=n.getTensorArray(r.id);return[xe(s.size(),"int32")]}case"TensorArrayCloseV3":{let r=I("tensorArrayId",e,t,n),s=n.getTensorArray(r.id);return s.clearAndClose(),[s.idTensor]}case"TensorListSetItem":{let r=I("tensorListId",e,t,n),s=I("index",e,t,n),a=I("tensor",e,t,n),o=n.getTensorList(r.id);return o.setItem(s,a),[o.idTensor]}case"TensorListGetItem":{let r=I("tensorListId",e,t,n),s=I("index",e,t,n),a=I("elementShape",e,t,n),o=I("elementDType",e,t,n);return[n.getTensorList(r.id).getItem(s,a,o)]}case"TensorListScatterV2":case"TensorListScatter":{let r=I("indices",e,t,n),s=I("tensor",e,t,n),a=I("elementShape",e,t,n),o=I("numElements",e,t,n),i=u6(s,r,a,o);return n.addTensorList(i),[i.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let r=I("elementShape",e,t,n),s=I("elementDType",e,t,n),a;e.op==="TensorListReserve"?a="numElements":a="maxNumElements";let o=I(a,e,t,n),i=e.op==="TensorListReserve"?-1:o,c=c6(r,s,o,i);return n.addTensorList(c),[c.idTensor]}case"TensorListGather":{let r=I("tensorListId",e,t,n),s=I("indices",e,t,n),a=I("elementShape",e,t,n),o=I("elementDType",e,t,n);return[n.getTensorList(r.id).gather(s,o,a)]}case"TensorListStack":{let r=I("tensorListId",e,t,n),s=I("elementShape",e,t,n),a=I("elementDType",e,t,n),o=I("numElements",e,t,n);return[n.getTensorList(r.id).stack(s,a,o)]}case"TensorListFromTensor":{let r=I("tensor",e,t,n),s=I("elementShape",e,t,n),a=I("elementDType",e,t,n),o=i6(r,s,a);return n.addTensorList(o),[o.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let r=I("tensorListId",e,t,n),s=n.getTensorList(r.id),a=I("dtype",e,t,n),o=I("elementShape",e,t,n);return[s.concat(a,o)]}case"TensorListPushBack":{let r=I("tensorListId",e,t,n),s=I("tensor",e,t,n),a=n.getTensorList(r.id);return a.pushBack(s),[a.idTensor]}case"TensorListPopBack":{let r=I("tensorListId",e,t,n),s=I("elementShape",e,t,n),a=I("elementDType",e,t,n);return[n.getTensorList(r.id).popBack(s,a)]}case"TensorListSplit":{let r=I("tensor",e,t,n),s=I("elementShape",e,t,n),a=I("lengths",e,t,n),o=l6(r,a,s);return n.addTensorList(o),[o.idTensor]}case"TensorListLength":{let r=I("tensorListId",e,t,n),s=n.getTensorList(r.id);return[xe(s.size(),"int32")]}case"TensorListResize":{let r=I("tensorListId",e,t,n),s=I("size",e,t,n),o=n.getTensorList(r.id).resize(s);return n.addTensorList(o),[o.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function e1(e,t,n){let[r,s]=I("fusedOps",e,t,n),a=r==="biasadd",o=!a,i=s==="prelu",c=r==="fusedbatchnorm",u=I("numArgs",e,t,n);if(a){if(i&&u!==2)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!i&&a&&u!==1)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(c)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let l=I("strides",e,t,n),p=wh(e,t,n),d=I("dataFormat",e,t,n).toUpperCase(),h=I("dilations",e,t,n),[f,m]=I("args",e,t,n);o&&(m=f,f=void 0);let g=I("leakyreluAlpha",e,t,n);return{stride:l,pad:p,dataFormat:d,dilations:h,biasArg:f,preluArg:m,activationFunc:s,leakyreluAlpha:g}}var p6=(e,t,n,r=ln)=>{switch(e.op){case"Conv1D":{let s=I("stride",e,t,n),a=I("pad",e,t,n),o=I("dataFormat",e,t,n).toUpperCase(),i=I("dilation",e,t,n);return[r.conv1d(I("x",e,t,n),I("filter",e,t,n),s,a,o,i)]}case"Conv2D":{let s=I("strides",e,t,n),a=wh(e,t,n),o=I("dataFormat",e,t,n).toUpperCase(),i=I("dilations",e,t,n);return[r.conv2d(I("x",e,t,n),I("filter",e,t,n),[s[1],s[2]],a,o,[i[1],i[2]])]}case"_FusedConv2D":{let{stride:s,pad:a,dataFormat:o,dilations:i,biasArg:c,preluArg:u,activationFunc:l,leakyreluAlpha:p}=e1(e,t,n);return[r.fused.conv2d({x:I("x",e,t,n),filter:I("filter",e,t,n),strides:[s[1],s[2]],pad:a,dataFormat:o,dilations:[i[1],i[2]],bias:c,activation:l,preluActivationWeights:u,leakyreluAlpha:p})]}case"FusedDepthwiseConv2dNative":{let{stride:s,pad:a,dataFormat:o,dilations:i,biasArg:c,preluArg:u,activationFunc:l,leakyreluAlpha:p}=e1(e,t,n);return[r.fused.depthwiseConv2d({x:I("x",e,t,n),filter:I("filter",e,t,n),strides:[s[1],s[2]],pad:a,dataFormat:o,dilations:[i[1],i[2]],bias:c,activation:l,preluActivationWeights:u,leakyreluAlpha:p})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let s=I("outputShape",e,t,n),a=I("strides",e,t,n),o=wh(e,t,n);return[r.conv2dTranspose(I("x",e,t,n),I("filter",e,t,n),s,[a[1],a[2]],o)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let s=I("strides",e,t,n),a=wh(e,t,n),o=I("dilations",e,t,n),i=I("dataFormat",e,t,n).toUpperCase();return[r.depthwiseConv2d(I("input",e,t,n),I("filter",e,t,n),[s[1],s[2]],a,i,[o[1],o[2]])]}case"Conv3D":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("dataFormat",e,t,n).toUpperCase(),i=I("dilations",e,t,n);return[r.conv3d(I("x",e,t,n),I("filter",e,t,n),[s[1],s[2],s[3]],a,o,[i[1],i[2],i[3]])]}case"AvgPool":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("kernelSize",e,t,n);return[r.avgPool(I("x",e,t,n),[o[1],o[2]],[s[1],s[2]],a)]}case"MaxPool":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("kernelSize",e,t,n);return[r.maxPool(I("x",e,t,n),[o[1],o[2]],[s[1],s[2]],a)]}case"MaxPoolWithArgmax":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("kernelSize",e,t,n),i=I("includeBatchInIndex",e,t,n),{result:c,indexes:u}=r.maxPoolWithArgmax(I("x",e,t,n),[o[1],o[2]],[s[1],s[2]],a,i);return[c,u]}case"AvgPool3D":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("kernelSize",e,t,n);return[r.avgPool3d(I("x",e,t,n),[o[1],o[2],o[3]],[s[1],s[2],s[3]],a)]}case"MaxPool3D":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("kernelSize",e,t,n);return[r.maxPool3d(I("x",e,t,n),[o[1],o[2],o[3]],[s[1],s[2],s[3]],a)]}case"Dilation2D":{let s=I("strides",e,t,n),a=I("pad",e,t,n),o=I("dilations",e,t,n),i=s[1],c=s[2],u=o[1],l=o[2];return[r.dilation2d(I("x",e,t,n),I("filter",e,t,n),[i,c],a,[u,l],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},h6=(e,t,n,r=ln)=>{switch(e.op){case"Fill":{let s=I("shape",e,t,n),a=I("dtype",e,t,n),o=I("value",e,t,n);return[r.fill(s,o,a)]}case"LinSpace":{let s=I("start",e,t,n),a=I("stop",e,t,n),o=I("num",e,t,n);return[r.linspace(s,a,o)]}case"Multinomial":{let s=I("logits",e,t,n),a=I("numSamples",e,t,n),o=I("seed",e,t,n);return[r.multinomial(s,a,o)]}case"OneHot":{let s=I("indices",e,t,n),a=I("depth",e,t,n),o=I("onValue",e,t,n),i=I("offValue",e,t,n),c=I("dtype",e,t,n);return[r.oneHot(s,a,o,i,c)]}case"Ones":return[r.ones(I("shape",e,t,n),I("dtype",e,t,n))];case"OnesLike":return[r.onesLike(I("x",e,t,n))];case"RandomStandardNormal":return[r.randomStandardNormal(I("shape",e,t,n),I("dtype",e,t,n),I("seed",e,t,n))];case"RandomUniform":return[r.randomUniform(I("shape",e,t,n),I("minval",e,t,n),I("maxval",e,t,n),I("dtype",e,t,n))];case"Range":{let s=I("start",e,t,n),a=I("stop",e,t,n),o=I("step",e,t,n);return[r.range(s,a,o,I("dtype",e,t,n))]}case"TruncatedNormal":{let s=I("shape",e,t,n),a=I("mean",e,t,n),o=I("stdDev",e,t,n),i=I("seed",e,t,n);return[r.truncatedNormal(s,a,o,I("dtype",e,t,n),i)]}case"Zeros":return[r.zeros(I("shape",e,t,n),I("dtype",e,t,n))];case"ZerosLike":return[r.zerosLike(I("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function ky(e,t,n){let r=I("boxes",e,t,n),s=I("scores",e,t,n),a=I("maxOutputSize",e,t,n),o=I("iouThreshold",e,t,n),i=I("scoreThreshold",e,t,n),c=I("softNmsSigma",e,t,n);return{boxes:r,scores:s,maxOutputSize:a,iouThreshold:o,scoreThreshold:i,softNmsSigma:c}}var f6=async(e,t,n,r,s=ln)=>{switch(e.op){case"NonMaxSuppressionV5":{let{boxes:a,scores:o,maxOutputSize:i,iouThreshold:c,scoreThreshold:u,softNmsSigma:l}=ky(e,t,n),p=await s.image.nonMaxSuppressionWithScoreAsync(a,o,i,c,u,l);return[p.selectedIndices,p.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:a,scores:o,maxOutputSize:i,iouThreshold:c,scoreThreshold:u}=ky(e,t,n),l=I("padToMaxOutputSize",e,t,n),p=await s.image.nonMaxSuppressionPaddedAsync(a,o,i,c,u,l);return[p.selectedIndices,p.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:a,scores:o,maxOutputSize:i,iouThreshold:c,scoreThreshold:u}=ky(e,t,n);return[await s.image.nonMaxSuppressionAsync(a,o,i,c,u)]}case"Where":{let a=s.cast(I("condition",e,t,n),"bool"),o=[await s.whereAsync(a)];return a.dispose(),o}case"ListDiff":return s.setdiff1dAsync(I("x",e,t,n),I("y",e,t,n));default:throw TypeError(`Node type ${e.op} is not implemented`)}},m6=(e,t,n,r=ln)=>{switch(e.op){case"LowerBound":{let s=I("sortedSequence",e,t,n),a=I("values",e,t,n);return[r.lowerBound(s,a)]}case"TopKV2":{let s=I("x",e,t,n),a=I("k",e,t,n),o=I("sorted",e,t,n),i=r.topk(s,a,o);return[i.values,i.indices]}case"UpperBound":{let s=I("sortedSequence",e,t,n),a=I("values",e,t,n);return[r.upperBound(s,a)]}case"Unique":{let s=I("x",e,t,n),a=r.unique(s);return[a.values,a.indices]}case"UniqueV2":{let s=I("x",e,t,n),a=I("axis",e,t,n),o=r.unique(s,a);return[o.values,o.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},g6=(e,t,n,r=ln)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":let s=I("default",e,t,n);return[Sn(e.name,t,n)||s];case"Placeholder":return[Sn(e.name,t,n)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":{let l=I("x",e,t,n);return[Ts(l)]}case"IdentityN":return I("x",e,t,n).map(l=>Ts(l));case"Snapshot":let a=I("x",e,t,n);return[Ts(a)];case"Shape":return[r.tensor1d(I("x",e,t,n).shape,"int32")];case"ShapeN":return I("x",e,t,n).map(l=>r.tensor1d(l.shape));case"Size":return[r.scalar(I("x",e,t,n).size,"int32")];case"Rank":return[r.scalar(I("x",e,t,n).rank,"int32")];case"NoOp":return[r.scalar(1)];case"Print":let o=I("x",e,t,n),i=I("data",e,t,n),c=I("message",e,t,n),u=I("summarize",e,t,n);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(c);for(let l=0;l<i.length;l++)console.log(Array.prototype.slice.call(i[l].dataSync()).slice(0,u));return[o];default:throw TypeError(`Node type ${e.op} is not implemented`)}},b6=class{constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=xe(0),this.tensorMap=new Map,Jt(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return xe(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return this.tensorMap.forEach(r=>r.dispose()),this.tensorMap.clear(),O(()=>{let r=lt(t),s=n.length,a=r.length;w.assert(s===a,()=>`The number of elements doesn't match, keys has ${s} elements, the values has ${a} elements.`);for(let o=0;o<s;o++){let i=n[o],c=r[o];Jt(c),this.tensorMap.set(i,c)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return O(()=>{let r=[];for(let s=0;s<n.length;s++){let a=n[s],o=this.findWithDefault(a,t);r.push(o)}return Ft(r)})}findWithDefault(e,t){let n=this.tensorMap.get(e);return n!=null?n:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}},y6=async(e,t,n,r)=>{switch(e.op){case"HashTable":case"HashTableV2":{let s=r.getHashTableHandleByName(e.name);if(s!=null)return[s];{let a=I("keyDType",e,t,n),o=I("valueDType",e,t,n),i=new b6(a,o);return r.addHashTable(e.name,i),[i.handle]}}case"LookupTableImport":case"LookupTableImportV2":{let s=I("tableHandle",e,t,n,r),a=I("keys",e,t,n),o=I("values",e,t,n);return[await r.getHashTableById(s.id).import(a,o)]}case"LookupTableFind":case"LookupTableFindV2":{let s=I("tableHandle",e,t,n,r),a=I("keys",e,t,n),o=I("defaultValue",e,t,n);return[await r.getHashTableById(s.id).find(a,o)]}case"LookupTableSize":case"LookupTableSizeV2":{let s=I("tableHandle",e,t,n,r);return[r.getHashTableById(s.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},v6=(e,t,n,r=ln)=>{switch(e.op){case"ResizeBilinear":{let s=I("images",e,t,n),a=I("size",e,t,n),o=I("alignCorners",e,t,n),i=I("halfPixelCenters",e,t,n);return[r.image.resizeBilinear(s,[a[0],a[1]],o,i)]}case"ResizeNearestNeighbor":{let s=I("images",e,t,n),a=I("size",e,t,n),o=I("alignCorners",e,t,n),i=I("halfPixelCenters",e,t,n);return[r.image.resizeNearestNeighbor(s,[a[0],a[1]],o,i)]}case"CropAndResize":{let s=I("image",e,t,n),a=I("boxes",e,t,n),o=I("boxInd",e,t,n),i=I("cropSize",e,t,n),c=I("method",e,t,n),u=I("extrapolationValue",e,t,n);return[r.image.cropAndResize(s,a,o,i,c,u)]}case"ImageProjectiveTransformV3":{let s=I("images",e,t,n),a=I("transforms",e,t,n),o=I("outputShape",e,t,n),i=I("fillValue",e,t,n),c=I("interpolation",e,t,n),u=I("fillMode",e,t,n);return[r.image.transform(s,a,c.toLowerCase(),u.toLowerCase(),i,o)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},x6=(e,t,n,r=ln)=>{switch(e.op){case"Equal":return[r.equal(I("a",e,t,n),I("b",e,t,n))];case"NotEqual":return[r.notEqual(I("a",e,t,n),I("b",e,t,n))];case"Greater":return[r.greater(I("a",e,t,n),I("b",e,t,n))];case"GreaterEqual":return[r.greaterEqual(I("a",e,t,n),I("b",e,t,n))];case"Less":return[r.less(I("a",e,t,n),I("b",e,t,n))];case"LessEqual":return[r.lessEqual(I("a",e,t,n),I("b",e,t,n))];case"LogicalAnd":return[r.logicalAnd(I("a",e,t,n),I("b",e,t,n))];case"LogicalNot":return[r.logicalNot(I("a",e,t,n))];case"LogicalOr":return[r.logicalOr(I("a",e,t,n),I("b",e,t,n))];case"Select":case"SelectV2":return[r.where(I("condition",e,t,n),I("a",e,t,n),I("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},w6=(e,t,n,r=ln)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[r.matMul(I("a",e,t,n),I("b",e,t,n),I("transposeA",e,t,n),I("transposeB",e,t,n))];case"Einsum":return[r.einsum(I("equation",e,t,n),...I("tensors",e,t,n))];case"Transpose":return[r.transpose(I("x",e,t,n),I("perm",e,t,n))];case"_FusedMatMul":let[s,a]=I("fusedOps",e,t,n),o=s==="biasadd",i=a==="prelu",c=I("numArgs",e,t,n),u=I("leakyreluAlpha",e,t,n);if(o){if(i&&c!==2)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!i&&c!==1)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[l,p]=I("args",e,t,n);return[r.fused.matMul({a:I("a",e,t,n),b:I("b",e,t,n),transposeA:I("transposeA",e,t,n),transposeB:I("transposeB",e,t,n),bias:l,activation:a,preluActivationWeights:p,leakyreluAlpha:u})];default:throw TypeError(`Node type ${e.op} is not implemented`)}},I6=(e,t,n,r=ln)=>{switch(e.op){case"EuclideanNorm":return[r.euclideanNorm(I("x",e,t,n),I("axis",e,t,n),I("keepDims",e,t,n))];case"FusedBatchNorm":case"FusedBatchNormV2":return[r.batchNorm(I("x",e,t,n),I("mean",e,t,n),I("variance",e,t,n),I("offset",e,t,n),I("scale",e,t,n),I("epsilon",e,t,n))];case"FusedBatchNormV3":return[r.batchNorm(I("x",e,t,n),I("mean",e,t,n),I("variance",e,t,n),I("offset",e,t,n),I("scale",e,t,n),I("epsilon",e,t,n))];case"LRN":return[r.localResponseNormalization(I("x",e,t,n),I("radius",e,t,n),I("bias",e,t,n),I("alpha",e,t,n),I("beta",e,t,n))];case"Softmax":return[r.softmax(I("x",e,t,n))];case"LogSoftmax":return[r.logSoftmax(I("x",e,t,n))];case"SparseToDense":return[r.sparseToDense(I("sparseIndices",e,t,n),I("outputShape",e,t,n),I("sparseValues",e,t,n),I("defaultValue",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},k6=(e,t,n,r=ln)=>{switch(e.op){case"Max":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.max(I("x",e,t,n),i,c)]}case"Mean":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.mean(I("x",e,t,n),i,c)]}case"Min":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.min(I("x",e,t,n),i,c)]}case"Sum":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.sum(I("x",e,t,n),i,c)]}case"All":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.all(I("x",e,t,n),i,c)]}case"Any":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.any(I("x",e,t,n),i,c)]}case"ArgMax":{let i=I("axis",e,t,n);return[r.argMax(I("x",e,t,n),i)]}case"ArgMin":{let i=I("axis",e,t,n);return[r.argMin(I("x",e,t,n),i)]}case"Prod":{let i=I("axis",e,t,n),c=I("keepDims",e,t,n);return[r.prod(I("x",e,t,n),i,c)]}case"Cumprod":{let i=I("axis",e,t,n),c=I("exclusive",e,t,n),u=I("reverse",e,t,n);return[r.cumprod(I("x",e,t,n),i,c,u)]}case"Cumsum":{let i=I("axis",e,t,n),c=I("exclusive",e,t,n),u=I("reverse",e,t,n);return[r.cumsum(I("x",e,t,n),i,c,u)]}case"Bincount":let s=I("x",e,t,n),a=I("weights",e,t,n),o=I("size",e,t,n);return[r.bincount(s,a,o)];case"DenseBincount":{let i=I("x",e,t,n),c=I("weights",e,t,n),u=I("size",e,t,n),l=I("binaryOutput",e,t,n);return[r.denseBincount(i,c,u,l)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},S6=(e,t,n,r=ln)=>{switch(e.op){case"ConcatV2":case"Concat":{let s=I("n",e,t,n),a=I("axis",e,t,n),o=I("tensors",e,t,n);return o=o.slice(0,s),[r.concat(o,a)]}case"Gather":{let s=I("x",e,t,n),a=I("indices",e,t,n);return[r.gather(s,r.cast(a,"int32"),0)]}case"GatherV2":{let s=I("axis",e,t,n),a=I("batchDims",e,t,n),o=I("x",e,t,n),i=I("indices",e,t,n);return[r.gather(o,r.cast(i,"int32"),s,a)]}case"Reverse":{let s=I("dims",e,t,n),a=[];for(let i=0;i<s.length;i++)s[i]&&a.push(i);let o=I("x",e,t,n);return[r.reverse(o,a)]}case"ReverseV2":{let s=I("axis",e,t,n),a=I("x",e,t,n);return[r.reverse(a,s)]}case"Slice":{let s=I("begin",e,t,n),a=I("size",e,t,n);return[r.slice(I("x",e,t,n),s,a)]}case"StridedSlice":{let s=I("begin",e,t,n),a=I("end",e,t,n),o=I("strides",e,t,n),i=I("beginMask",e,t,n),c=I("endMask",e,t,n),u=I("ellipsisMask",e,t,n),l=I("newAxisMask",e,t,n),p=I("shrinkAxisMask",e,t,n),d=I("x",e,t,n);return[r.stridedSlice(d,s,a,o,i,c,u,l,p)]}case"Pack":return O(()=>{let s=I("axis",e,t,n),a=I("tensors",e,t,n),o=a[0].shape,i=r.squeeze(a[0]).shape,c=a.map(u=>{let l=w.arraysEqual(u.shape,o);if(!l&&!w.arraysEqual(r.squeeze(u).shape,i))throw new Error("the input tensors shape does not match");return l?u:r.reshape(u,o)});return[r.stack(c,s)]});case"Unpack":{let s=I("axis",e,t,n),a=I("tensor",e,t,n);return r.unstack(a,s)}case"Tile":{let s=I("reps",e,t,n);return[r.tile(I("x",e,t,n),s)]}case"Split":case"SplitV":{let s=I("axis",e,t,n),a=I("numOrSizeSplits",e,t,n),o=I("x",e,t,n);return r.split(o,a,s)}case"ScatterNd":{let s=I("indices",e,t,n),a=I("values",e,t,n),o=I("shape",e,t,n);return[r.scatterND(s,a,o)]}case"GatherNd":{let s=I("x",e,t,n),a=I("indices",e,t,n);return[r.gatherND(s,a)]}case"SparseToDense":{let s=I("sparseIndices",e,t,n),a=I("outputShape",e,t,n),o=I("sparseValues",e,t,n),i=I("defaultValue",e,t,n);return[r.sparseToDense(s,o,a,o.dtype===i.dtype?i:r.cast(i,o.dtype))]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},T6=(e,t,n,r=ln)=>{switch(e.op){case"SparseFillEmptyRows":{let{outputIndices:s,outputValues:a,emptyRowIndicator:o,reverseIndexMap:i}=r.sparse.sparseFillEmptyRows(I("indices",e,t,n),I("values",e,t,n),I("denseShape",e,t,n),I("defaultValue",e,t,n));return[s,a,o,i]}case"SparseReshape":{let{outputIndices:s,outputShape:a}=r.sparse.sparseReshape(I("inputIndices",e,t,n),I("inputShape",e,t,n),I("newShape",e,t,n));return[s,a]}case"SparseSegmentMean":return[r.sparse.sparseSegmentMean(I("data",e,t,n),I("indices",e,t,n),I("segmentIds",e,t,n))];case"SparseSegmentSum":return[r.sparse.sparseSegmentSum(I("data",e,t,n),I("indices",e,t,n),I("segmentIds",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},C6=(e,t,n,r=ln)=>{switch(e.op){case"FFT":return[r.fft(I("x",e,t,n))];case"IFFT":return[r.ifft(I("x",e,t,n))];case"RFFT":return[r.rfft(I("x",e,t,n))];case"IRFFT":return[r.irfft(I("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},N6=(e,t,n,r=ln)=>{switch(e.op){case"StringNGrams":{let{nGrams:s,nGramsSplits:a}=r.string.stringNGrams(I("data",e,t,n),I("dataSplits",e,t,n),I("separator",e,t,n),I("nGramWidths",e,t,n),I("leftPad",e,t,n),I("rightPad",e,t,n),I("padWidth",e,t,n),I("preserveShortSequences",e,t,n));return[s,a]}case"StringSplit":{let{indices:s,values:a,shape:o}=r.string.stringSplit(I("input",e,t,n),I("delimiter",e,t,n),I("skipEmpty",e,t,n));return[s,a,o]}case"StringToHashBucketFast":return[r.string.stringToHashBucketFast(I("input",e,t,n),I("numBuckets",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},_6=(e,t,n,r=ln)=>{switch(e.op){case"Cast":return[r.cast(I("x",e,t,n),I("dtype",e,t,n))];case"ExpandDims":{let s=I("axis",e,t,n);return[r.expandDims(I("x",e,t,n),s)]}case"Squeeze":{let s=I("axis",e,t,n);return[r.squeeze(I("x",e,t,n),s)]}case"Reshape":return[r.reshape(I("x",e,t,n),I("shape",e,t,n))];case"MirrorPad":return[r.mirrorPad(I("x",e,t,n),I("padding",e,t,n),I("mode",e,t,n))];case"PadV2":case"Pad":return[r.pad(I("x",e,t,n),I("padding",e,t,n),I("constantValue",e,t,n))];case"SpaceToBatchND":{let s=I("blockShape",e,t,n),a=I("paddings",e,t,n);return[r.spaceToBatchND(I("x",e,t,n),s,a)]}case"BatchToSpaceND":{let s=I("blockShape",e,t,n),a=I("crops",e,t,n);return[r.batchToSpaceND(I("x",e,t,n),s,a)]}case"DepthToSpace":{let s=I("blockSize",e,t,n),a=I("dataFormat",e,t,n).toUpperCase();return[r.depthToSpace(I("x",e,t,n),s,a)]}case"BroadcastTo":return[r.broadcastTo(I("x",e,t,n),I("shape",e,t,n))];case"BroadcastArgs":return[r.broadcastArgs(I("s0",e,t,n),I("s1",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function t1(e,t,n,r,s=O){let a=((o,i,c)=>{switch(o.category){case"arithmetic":return s(()=>s6(o,i,c));case"basic_math":return s(()=>a6(o,i,c));case"control":return d6(o,i,c);case"convolution":return s(()=>p6(o,i,c));case"creation":return s(()=>h6(o,i,c));case"dynamic":return f6(o,i,c);case"evaluation":return s(()=>m6(o,i,c));case"image":return s(()=>v6(o,i,c));case"graph":return s(()=>g6(o,i,c));case"logical":return s(()=>x6(o,i,c));case"matrices":return s(()=>w6(o,i,c));case"normalization":return s(()=>I6(o,i,c));case"reduction":return s(()=>k6(o,i,c));case"slice_join":return s(()=>S6(o,i,c));case"sparse":return s(()=>T6(o,i,c));case"spectral":return s(()=>C6(o,i,c));case"string":return s(()=>N6(o,i,c));case"transformation":return s(()=>_6(o,i,c));case"hash_table":return y6(o,i,c,r);case"custom":let u=kN(o.op);if(u&&u.customExecutor)return u.customExecutor(new r6(o,i,c));throw TypeError(`Custom op ${o.op} is not registered.`);default:throw TypeError(`Unknown op '${o.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,n);return w.isPromise(a)?a.then(o=>[].concat(o)):[].concat(a)}var n1=class{constructor(e={},t={},n={},r={}){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=n,this.functionMap=r,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let e=[];for(let t=0;t<this.contexts.length-1;t++){let n=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(n))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(t=>t.id===0&&t.iterationId===0?"":`${t.frameName}-${t.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error("Cannot exit frame, the context is empty")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error("Cannot increase frame iteration, the context is empty")}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(let t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(let t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function r1(e,t,n,r){let s=new Set,a=[],o=null,i=null,c=new Set,u=Object.keys(e).map(d=>Zn(d)[0]),l=[];r!=null&&(l=r.map(d=>Zn(d.name)[0]));let p=[...t];for(;p.length>0;){let d=p.pop();if((HN(d)||F6(d)||R6(d))&&o==null&&(o=d,i=o.children.map(h=>h.name).filter(h=>s.has(h))),s.add(d.name),n[d.name]==null&&u.indexOf(d.name)===-1&&l.indexOf(d.name)===-1){if(d.inputs.length===0){a.push(d.name);continue}d.inputs.forEach(h=>{c.has(h.name)||(c.add(h.name),p.push(h))})}}return{inputs:e,outputs:t,usedNodes:s,missingInputs:a,dynamicNode:o,syncInputs:i}}function E6(e,t,n){let{usedNodes:r,inputs:s}=n,a=[],o=Object.keys(s).map(l=>Zn(l)[0]).map(l=>e.nodes[l]),i=e.initNodes;o.forEach(l=>{r.has(l.name)&&a.push(l)}),e.weights.forEach(l=>{r.has(l.name)&&a.push(l)}),i!=null&&i.forEach(l=>{r.has(l.name)&&a.push(l)});let c=new Set,u=[];for(;a.length>0;){let l=a.pop();c.add(l.name),t[l.name]||u.push(l),l.children.forEach(p=>{!c.has(p.name)&&r.has(p.name)&&p.inputs.every(d=>c.has(d.name))&&a.push(p)})}return u}var A6=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],$6=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],D6=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function HN(e){return A6.indexOf(e.op)>=0}function F6(e){return $6.indexOf(e.op)>=0}function R6(e){return D6.indexOf(e.op)>=0}var lv=class{constructor(e,t){this.graph=e,this.parent=t,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=e.outputs,this._inputs=e.inputs,this._initNodes=e.initNodes,this._signature=e.signature,this._functions=e.functions,e.functions!=null&&Object.keys(e.functions).forEach(n=>{this._functionExecutorMap[n]=new lv(e.functions[n],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){let t=Object.keys(e).map(n=>e[n].map(r=>r.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{let t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}getCompilationKey(e,t){let n=e.map(s=>s.name).sort(),r=t.map(s=>s.name).sort();return n.join(this.SEPERATOR)+"--"+r.join(this.SEPERATOR)}compile(e,t){let n=r1(e,t,this.weightMap,this._initNodes),{missingInputs:r,dynamicNode:s,syncInputs:a}=n;if(s!=null)throw new Error(`This execution contains the node '${s.name}', which has the dynamic op '${s.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${a}]`);if(r.length>0){let o=t.map(c=>c.name),i=Object.keys(e);throw new Error(`Cannot compute the outputs [${o}] from the provided inputs [${i}]. Missing the following inputs: [${r}]`)}return E6(this.graph,this.weightMap,n)}execute(e,t){e=this.mapInputs(e);let n=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);let r=n.map(l=>this.graph.nodes[Zn(l)[0]]),s=t.map(l=>Zn(l)[0]),a=s.map(l=>this.graph.nodes[l]);this.resetIntermediateTensors(),a.length===0&&(a=this._outputs);let o=this.getCompilationKey(r,a),i=this.compiledMap.get(o);i==null&&(i=this.compile(e,a),this.compiledMap.set(o,i));let c={},u={};return O(()=>{let l=new n1(this.weightMap,c,u,this.functionExecutorMap),p=Object.assign({},this.weightMap);Object.keys(e).forEach(f=>{let[m,g]=Zn(f),b=[];b[g]=e[f],p[m]=b});let d=this.getFrozenTensorIds(p),h={};for(let f=0;f<i.length;f++){let m=i[f];if(!p[m.name]){let g=t1(m,p,l,this._resourceManager);if(w.isPromise(g))throw new Error(`The execution of the op '${m.op}' returned a promise. Please use model.executeAsync() instead.`);p[m.name]=g,this.checkTensorForDisposal(m.name,m,p,l,d,s,h)}}return this.parent==null&&l.dispose(d),t.map(f=>Sn(f,p,l))})}getFrozenTensorIds(e){let t=[].concat.apply([],Object.keys(e).map(n=>e[n]).map(n=>n.map(r=>r.id)));return new Set(t)}checkTensorForDisposal(e,t,n,r,s,a,o){t.category==="control"||a.indexOf(e)!==-1||(n[e].forEach(i=>{i!=null&&(o[i.id]=(o[i.id]||0)+t.children.length)}),t.inputs.forEach(i=>{if(i.category!=="control"){let c=OH(i.name,n,r);c!=null&&c.forEach(u=>{if(u&&!u.kept&&!s.has(u.id)){let l=o[u.id];if(l===1){if(!this.keepTensorForDebug)u.dispose();else{let[p,d]=os(t.name,r);this.intermediateTensors[p]?this.intermediateTensors[p][d]=u:(this.intermediateTensors[p]=[],this.intermediateTensors[p][d]=u)}delete o[u.id]}else l!=null&&o[u.id]--}})}}))}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(e=>this.intermediateTensors[e].forEach(t=>t.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(e=>{this.tensorsMap[e].forEach(n=>{n&&!n.kept&&!n.isDisposed&&!this.keepIds.has(n.id)&&n.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let e in this.intermediateTensors)this.intermediateTensors[e].forEach(t=>t.dispose()),delete this.intermediateTensors[e]}async _executeAsync(e,t,n=!1,r={},s={}){n||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepTensorForDebug=q().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(u){console.warn(u.message)}this.resetIntermediateTensors();let a=new n1(this.weightMap,r,s,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(e,a,t,n);let o=t.map(u=>Sn(u,this.tensorsMap,a)),i=o.map(u=>u.id),c=Object.keys(e).map(u=>e[u].id);return this.keepIds=new Set([...i,...c,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&a.dispose(this.keepIds),o}async executeFunctionAsync(e,t,n){let r=e.reduce((s,a,o)=>(s[this.inputs[o].name]=a,s),{});return this._executeAsync(r,this.outputNodes,!0,t,n)}async executeWithControlFlow(e,t,n,r){let s=Object.keys(e),a=s.map(y=>this.graph.nodes[Zn(y)[0]]),o=n.map(y=>Zn(y)[0]),i=o.map(y=>this.graph.nodes[y]);i.length===0&&(i=this._outputs);let{usedNodes:c,missingInputs:u,dynamicNode:l,syncInputs:p}=r1(e,i,this.weightMap,this._initNodes),d=[...a,...this.graph.weights,...this._initNodes||[]].map(y=>({node:y,contexts:t.currentContext})),h=Object.assign({},this.weightMap);Object.keys(e).forEach(y=>{let[v,x]=Zn(y),k=[];k[x]=e[y],h[v]=k});let f={},m=this.getFrozenTensorIds(h),g={};for(;d.length>0;){let y=this.processStack(a,d,t,h,g,m,o,f,c);await Promise.all(y)}l==null&&!r&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let b=i.filter(y=>!HN(y)&&!Sn(y.name,h,t)).map(y=>y.name);if(b.length>0){let y="";throw l!=null&&(y=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${p}]`),new Error(`Cannot compute the outputs [${b}] from the provided inputs [${s}]. Consider providing the following inputs: [${u}]. ${y}`)}return h}processStack(e,t,n,r,s,a,o,i,c){let u=[];for(;t.length>0;){let l=t.pop();n.currentContext=l.contexts;let p="";if(l.node.op==="Enter"&&I("isConstant",l.node,r,n)&&([p]=os(l.node.name,n)),r[l.node.name]==null){let d=t1(l.node,r,n,this._resourceManager);p||([p]=os(l.node.name,n));let h=n.currentContext;w.isPromise(d)?u.push(d.then(f=>(r[p]=f,n.currentContext=h,this.checkTensorForDisposal(p,l.node,r,n,a,o,i),this.processChildNodes(l.node,t,n,r,s,c),f))):(r[p]=d,this.checkTensorForDisposal(p,l.node,r,n,a,o,i),this.processChildNodes(l.node,t,n,r,s,c))}else this.processChildNodes(l.node,t,n,r,s,c)}return u}processChildNodes(e,t,n,r,s,a){e.children.forEach(o=>{let[i]=os(o.name,n);s[i]||!a.has(o.name)||(o.op==="Merge"?o.inputNames.some(c=>!!Sn(c,r,n))&&(s[i]=!0,t.push({contexts:n.currentContext,node:o})):o.inputNames.every(c=>!!Sn(c,r,n))&&(s[i]=!0,t.push({contexts:n.currentContext,node:o})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(t=>t.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{let n=e[t],[r]=Zn(t),s=this.graph.nodes[r];if(s.attrParams.shape&&s.attrParams.shape.value){let a=s.attrParams.shape.value,o=a.length===n.shape.length&&n.shape.every((i,c)=>a[c]===-1||a[c]===i);w.assert(o,()=>`The shape of dict['${s.name}'] provided in model.execute(dict) must be [${a}], but was [${n.shape}]`)}s.attrParams.dtype&&s.attrParams.dtype.value&&w.assert(n.dtype===s.attrParams.dtype.value,()=>`The dtype of dict['${s.name}'] provided in model.execute(dict) must be ${s.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(e){let t={};for(let n in e)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[n]!=null){let r=this._signature.inputs[n];t[r.name]=e[n]}else t[n]=e[n];return t}checkInputs(e){let t=Object.keys(e).filter(n=>{let[r]=Zn(n);return this.graph.nodes[r]==null});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(t=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[t]!=null?this._signature.outputs[t].name:t,{})}checkOutputs(e){e.forEach(t=>{let[n]=Zn(t);if(!this.graph.nodes[n])throw new Error(`The output '${t}' is not found in the graph`)})}},P6=class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(let e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(let e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}},O6="?tfjs-format=file",M6="model.json",A0=class{constructor(e,t={},n=Ut){this.modelUrl=e,this.loadOptions=t,this.version="n/a",this.io=n,t==null&&(this.loadOptions={}),this.resourceManager=new P6}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}findIOHandler(){let e=this.modelUrl;if(e.load!=null)this.handler=e;else if(this.loadOptions.requestInit!=null)this.handler=this.io.browserHTTPRequest(e,this.loadOptions);else{let t=this.io.getLoadHandlers(e,this.loadOptions);if(t.length===0)t.push(this.io.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}load(){if(this.findIOHandler(),this.handler.load==null)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let e=this.handler.load();return w.isPromise(e)?e.then(t=>this.loadSync(t)):this.loadSync(e)}loadSync(e){this.artifacts=e;let t=this.artifacts.modelTopology,n=this.artifacts.signature;if(this.artifacts.userDefinedMetadata!=null){let s=this.artifacts.userDefinedMetadata;s.signature!=null&&(n=s.signature),s.structuredOutputKeys!=null&&(this.structuredOutputKeys=s.structuredOutputKeys)}this.signature=n,this.version=`${t.versions.producer}.${t.versions.minConsumer}`;let r=this.io.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new lv(Zk.Instance.transformGraph(t,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(r),this.executor.resourceManager=this.resourceManager,e.modelInitializer!=null&&e.modelInitializer.node!=null){let s=Zk.Instance.transformGraph(e.modelInitializer);this.initializer=new lv(s),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=e.initializerSignature}return!0}async save(e,t){if(typeof e=="string"){let n=this.io.getSaveHandlers(e);if(n.length===0)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(n.length>1)throw new Error(`Found more than one (${n.length}) save handlers for URL '${e}'`);e=n[0]}if(e.save==null)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}predict(e,t){let n=this.execute(e,this.outputNodes);if(this.structuredOutputKeys){let r=n instanceof Te?[n]:n,s={};return r.forEach((a,o)=>s[this.structuredOutputKeys[o]]=a),s}return n}normalizeInputs(e){if(!(e instanceof Te)&&!Array.isArray(e)){if(this.signature!=null&&this.signature.inputs!=null)for(let r in this.signature.inputs){let s=this.signature.inputs[r];s.resourceId!=null&&(e[r]=this.resourceIdToCapturedInput[s.resourceId])}return e}e=Array.isArray(e)?e:[e];let t=Object.keys(this.resourceIdToCapturedInput).length;if(e.length+t!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-t} non-resource placeholders, while there are ${e.length} input tensors provided.`);let n=0;return this.inputNodes.reduce((r,s)=>{let a=this.signature?this.signature.inputs[s]:null;return a!=null&&a.resourceId!=null?r[s]=this.resourceIdToCapturedInput[a.resourceId]:r[s]=e[n++],r},{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}executeInitializerGraph(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(e){if(this.resourceIdToCapturedInput={},this.initializerSignature){let t=Object.keys(this.initializerSignature.outputs);for(let n=0;n<t.length;n++){let r=t[n],s=this.initializerSignature.outputs[r];this.resourceIdToCapturedInput[s.resourceId]=e[n]}}}execute(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=this.executor.execute(e,t);return n.length>1?n:n[0]}async executeAsync(e,t){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=await this.executor.executeAsync(e,t);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,n)=>(t[n]=[e[n]],t),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&_e(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function L6(e,t={},n=Ut){if(e==null)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");t==null&&(t={}),t.fromTFHub&&typeof e=="string"&&(e=B6(e));let r=new A0(e,t,n);return await r.load(),r}function z6(e){if(e==null)throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");let t;if(e instanceof Array){let[r,s]=e;if(!r)throw new Error("modelJSON must be the first element of the array");if(!s||!(s instanceof ArrayBuffer))throw new Error("An ArrayBuffer of weights must be the second element of the array");if(!("modelTopology"in r))throw new Error("Model JSON is missing 'modelTopology'");if(!("weightsManifest"in r))throw new Error("Model JSON is missing 'weightsManifest'");let a=Ut.getWeightSpecs(r.weightsManifest),o=Ut.getModelArtifactsForJSONSync(r,a,s);t=Ut.fromMemorySync(o)}else if("load"in e)t=e;else if("modelTopology"in e&&"weightSpecs"in e&&"weightData"in e)t=Ut.fromMemorySync(e);else throw new Error("Unknown model format");let n=new A0(t);return n.load(),n}function B6(e){return e.endsWith("/")||(e=e+"/"),`${e}${M6}${O6}`}var W6="4.0.0",qN={};Ae(qN,{CSVDataset:()=>t_,Dataset:()=>ju,FileDataSource:()=>c_,TextLineDataset:()=>e_,URLDataSource:()=>u_,array:()=>lq,csv:()=>wq,func:()=>Iq,generator:()=>kq,microphone:()=>Tq,version_data:()=>Cq,webcam:()=>Sq,zip:()=>dq});var V6=ma(cf()),U6=ma(cf());function G6(e,t){return Zh(e,t)}function Zh(e,t,n=new Map,r=new Set){if(e==null)return null;if(typeof Blob=="function"&&e instanceof Blob)return e.slice();if(r.has(e))throw new Error("Circular references are not supported.");if(n.has(e))return n.get(e);let s=t(e);if(s.recurse&&s.value!==null)throw new Error("A deep map function may not return both a value and recurse=true.");if(s.recurse)if(Nc(e)){let a=Array.isArray(e)?[]:{};r.add(e);for(let o in e){let i=e[o],c=Zh(i,t,n,r);a[o]=c}return r.delete(e),e.__proto__&&(a.__proto__=e.__proto__),a}else throw new Error(`Can't recurse into non-iterable type: ${e}`);else return n.set(e,s.value),s.value}function H6(e,t=KN){return jN(e,t)}function jN(e,t,n=new Set){let r=e[0];if(n.has(r))throw new Error("Circular references are not supported.");let s=t(e);if(s.recurse&&s.value!==null)throw new Error("A deep zip function may not return both a value and recurse=true.");if(s.recurse)if(Nc(r)){let a=Array.isArray(r)?[]:{};n.add(r);for(let o in r){let i=e.map(u=>u[o]),c=jN(i,t,n);a[o]=c}return n.delete(r),a}else throw new Error(`Can't recurse into non-iterable type: ${r}`);else return s.value}function KN(e){return e===null?null:Nc(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function XN(e,t){let n=new Map;Zh(e,t,n);for(let s of Array.from(n.keys())){let a=n.get(s);if(w.isPromise(a)){let o=await a;n.set(s,o)}}return Zh(e,t,n)}function Nc(e){let t=!1;if(q().get("IS_BROWSER"))t=e instanceof TextDecoder;else{let{StringDecoder:n}=G1();t=e instanceof n}return e!=null&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||typeof e=="object"&&!(e instanceof Te)&&!(e instanceof Promise)&&!t)}function q6(e){return e==null||j6(e)||Array.isArray(e)||typeof e=="object"&&e instanceof Te||w.isTypedArray(e)}function j6(e){return e===null||typeof e!="object"&&typeof e!="function"}function K6(e){return G6(e,X6)}function X6(e){return e instanceof Te?{value:e.clone(),recurse:!1}:Nc(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var YN=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,e==null)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(let t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.wrap(this.begin+e),n=this.get(t);return this.set(t,this.pop()),n}},$0=class extends YN{constructor(){super($0.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){let e=this.capacity*2,t=new Array(e),n=this.length();for(let r=0;r<n;r++)t[r]=this.get(this.wrap(this.begin+r));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};$0.INITIAL_CAPACITY=32;function ZN(e){return new J6(e)}function D0(e){return new Q6(e)}function Y6(e,t){return new JN(e,t)}function Z6(e,t=na.FAIL){return new cq(e,t)}var tn=class{async toArray(){let e=[],t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){let e=this.prefetch(100),t=[],n=await e.next();for(;!n.done;)t.push(n.value),n=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),n=e(t.value);for(;!t.done&&n;)t=await this.next(),n=e(t.value)}handleErrors(e){return new oq(this,e)}filter(e){return new sq(this,e)}map(e){return new aq(this,e)}mapAsync(e){return new s1(this,e)}serialMapAsync(e){return new s1(this,e).serial()}flatmap(e){return new iq(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(t=>t===!0)}rowMajorBatch(e,t=!0){return new rq(this,e,t)}columnMajorBatch(e,t=!0,n=KN){return this.rowMajorBatch(e,t).map(s=>H6(s,n))}concatenate(e,t){return new JN(ZN([this,e]),t)}take(e){return e<0||e==null?this:new nq(this,e)}skip(e){return e<0||e==null?this:new tq(this,e)}prefetch(e){return new QN(this,e)}shuffle(e,t){return new uq(this,e,t)}serial(){return new eq(this)}},J6=class extends tn{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let e=this.items[this.trav];return this.trav++,{value:K6(e),done:!1}}},Q6=class extends tn{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message=`Error thrown while iterating through a dataset: ${e.message}`,e}}},eq=class extends tn{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},tq=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let e=await this.upstream.next();if(e.done)return e;_e(e.value)}return this.upstream.next()}},nq=class extends tn{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},rq=class extends tn{constructor(e,t,n=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let e=[];for(;e.length<this.batchSize;){let t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},sq=class extends tn{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;_e(e.value)}}},aq=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Map`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Ur.getTensorsInContainer(e.value),n=this.transform(e.value),r=Ur.getTensorsInContainer(n);for(let s of t)Ur.isTensorInList(s,r)||s.dispose();return{value:n,done:!1}}},oq=class extends tn{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},s1=class extends tn{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=Ur.getTensorsInContainer(e.value),n=await this.transform(e.value),r=Ur.getTensorsInContainer(n);for(let s of t)Ur.isTensorInList(s,r)||s.dispose();return{value:n,done:!1}}},F0=class extends tn{constructor(){super(),this.outputQueue=new $0,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},iq=class extends F0{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let e=await this.upstream.next();if(e.done)return!1;let t=Ur.getTensorsInContainer(e.value),n=this.transform(e.value),r=Ur.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let s of t)Ur.isTensorInList(s,r)||s.dispose();return!0}},JN=class extends tn{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,this.iterator==null){let n=await this.moreIterators.next();if(n.done)return{value:null,done:!0};this.iterator=n.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}},na;(function(e){e[e.FAIL=0]="FAIL",e[e.SHORTEST=1]="SHORTEST",e[e.LONGEST=2]="LONGEST"})(na||(na={}));var cq=class extends tn{constructor(e,t=na.FAIL){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,n=0;function r(a){return a instanceof tn?{value:a.next().then(i=>(t++,i.done&&n++,i.value)),recurse:!1}:{value:null,recurse:!0}}let s=await XN(this.iterators,r);if(t===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case na.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case na.SHORTEST:return{value:null,done:!0};case na.LONGEST:default:}return this.count++,{value:s,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},QN=class extends tn{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new YN(t)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},uq=class extends QN{constructor(e,t,n){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=U6.alea(n||w.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(t.done)this.upstreamExhausted=!0;else return this.refill(),t}return{value:null,done:!0}}},ju=class{constructor(){this.size=null}batch(e,t=!0){let n=this;w.assert(e>0,()=>`batchSize needs to be positive, but it is
+      ${e}`);let r;return this.size===1/0||this.size==null?r=this.size:t?r=Math.ceil(this.size/e):r=Math.floor(this.size/e),Yn(async()=>(await n.iterator()).columnMajorBatch(e,t,pq),r)}concatenate(e){let t=this,n;return this.size===1/0||e.size===1/0?n=1/0:this.size!=null&&e.size!=null?n=this.size+e.size:n=null,Yn(async()=>(await t.iterator()).concatenate(await e.iterator()),n)}filter(e){let t=this,n;return this.size===1/0?n=1/0:n=null,Yn(async()=>(await t.iterator()).filter(r=>O(()=>e(r))),n)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){let t=this;return Yn(async()=>(await t.iterator()).map(n=>O(()=>e(n))),this.size)}mapAsync(e){let t=this;return Yn(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(e==null)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let t=this;return Yn(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){let t=this,n;return this.size!=null&&e>0?n=this.size*e:e===0?n=0:this.size!=null&&(e===void 0||e<0)?n=1/0:n=null,Yn(async()=>{let r=D0(async()=>({value:await t.iterator(),done:!1}));return Y6(r.take(e))},n)}skip(e){let t=this,n;return this.size!=null&&e>=0&&this.size>=e?n=this.size-e:this.size!=null&&(this.size<e||e===void 0||e<0)?n=0:n=null,Yn(async()=>(await t.iterator()).skip(e),n)}shuffle(e,t,n=!0){if(e==null||e<0)throw this.size==null?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let r=this,s=V6.alea(t||w.now().toString());return Yn(async()=>{let a=s.int32();return n&&(a+=s.int32()),(await r.iterator()).shuffle(e,a.toString())},this.size)}take(e){let t=this,n;return this.size!=null&&this.size>e?n=e:this.size!=null&&this.size<=e?n=this.size:n=null,Yn(async()=>(await t.iterator()).take(e),n)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};ju.MAX_BUFFER_SIZE=1e4;function Yn(e,t=null){return new class extends ju{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function lq(e){return Yn(async()=>ZN(e),e.length)}function dq(e){if(!Nc(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let n=0;n<e.length;n++)t=t==null?e[n].size:Math.min(t,e[n].size);else if(e instanceof Object)for(let n in e)t=t==null?e[n].size:Math.min(t,e[n].size);return Yn(async()=>{let n=await XN(e,r=>{if(r instanceof ju)return{value:r.iterator(),recurse:!1};if(Nc(r))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")});return Z6(n,na.SHORTEST)},t)}function pq(e){if(e===null)return null;let t=e[0];return q6(t)?{value:hq(e),recurse:!1}:{value:null,recurse:!0}}function hq(e){if(e.length===0)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof Te?Ft(e):Cn(e)}var e_=class extends ju{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`
+`).map(r=>(r.endsWith("\r")&&(r=r.slice(0,-1)),r))}},ph='"',Dl=Symbol("out"),a1=Symbol("field"),hh=Symbol("quote"),Sy=Symbol("quoteafterquote"),o1=Symbol("quoteinquote"),t_=class extends ju{constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new e_(e),t||(t={}),this.hasHeader=t.hasHeader!==!1,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(w.assert(t.delimiter==null,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&w.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);let t=this.fullColumnNames.reduce((r,s)=>(r[s]=r[s]+1||1,r),{}),n=Object.keys(t).filter(r=>t[r]>1);if(w.assert(n.length===0,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs){for(let r of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(r)===-1)throw new Error('The key "'+r+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let t=await(await this.base.iterator()).next();if(t.done)throw new Error("No data was found for CSV parsing.");let n=t.value;return this.parseRow(n,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(t=>this.makeDataElement(t))}makeDataElement(e){let t=this.parseRow(e),n={},r={};for(let s=0;s<this.fullColumnNames.length;s++){let a=this.fullColumnNames[s],o=this.columnConfigs?this.columnConfigs[a]:null;if(!(this.configuredColumnsOnly&&!o)){let i=t[s],c=null;if(i==="")if(o&&o.default!==void 0)c=o.default;else{if(o&&(o.required||o.isLabel))throw new Error(`Required column ${a} is empty in this line: ${e}`);c=void 0}else{let u=Number(i);if(isNaN(u))o&&o.dtype==="bool"?c=this.getBoolean(i):c=i;else if(!o||!o.dtype)c=u;else switch(o.dtype){case"float32":c=u;break;case"int32":c=Math.floor(u);break;case"bool":c=this.getBoolean(i);break;default:c=u}}o&&o.isLabel?r[a]=c:n[a]=c}}return Object.keys(r).length===0?n:{xs:n,ys:r}}getBoolean(e){return e==="1"||e.toLowerCase()==="true"?1:0}parseRow(e,t=!0){let n=[],r=0,s=e.length,a=Dl;for(let o=0;o<s;o++)switch(a){case Dl:switch(e.charAt(o)){case ph:r=o+1,a=hh;break;case this.delimiter:if(r=o+1,this.delimiter===" "&&this.delimWhitespace)break;n.push(""),a=Dl;break;default:a=a1,r=o;break}break;case a1:switch(e.charAt(o)){case this.delimiter:n.push(e.substring(r,o)),a=Dl,r=o+1;break;default:}break;case hh:switch(e.charAt(o)){case ph:a=Sy;break;default:}break;case Sy:switch(e.charAt(o)){case this.delimiter:n.push(e.substring(r,o-1)),a=Dl,r=o+1;break;case ph:a=hh;break;default:a=o1;break}break;case o1:switch(e.charAt(o)){case ph:a=hh;break;default:}break;default:}if(a===Sy?n.push(e.substring(r,s-1)):n.push(e.substring(r)),t&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}},n_=class extends tn{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;let t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=e.includeSpectrogram!==!1,this.includeWaveform=e.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(e={}){if(!q().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let t=new n_(e);return await t.start(),t}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(n){throw new Error(`Error thrown while initializing video stream: ${n.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t,n=await this.getAudioData();if(this.includeSpectrogram){let r=this.flattenQueue(n.freqDataQueue);e=this.getTensorFromAudioDataArray(r,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let r=this.flattenQueue(n.timeDataQueue);t=this.getTensorFromAudioDataArray(r,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let e=[],t=[],n=0;return new Promise(r=>{let s=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&r({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(s),r({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){let t=e[0].length,n=new Float32Array(e.length*t);return e.forEach((r,s)=>n.set(r,s*t)),n}getTensorFromAudioDataArray(e,t){let n=new Float32Array(w.sizeFromShape(t));return n.set(e,n.length-e.length),Cn(n,t)}},r_=class extends tn{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Ke([0],"int32"),this.webcamConfig.centerCrop){let n=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,r=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,s=(1-n)/2,a=(1-r)/2,o=s+n,i=r+a;this.cropBox=$r([a,s,i,o],[1,4])}else this.cropBox=$r([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(e,t={}){if(!q().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!e){if(e=document.createElement("video"),!t.resizeWidth||!t.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");e.width=t.resizeWidth,e.height=t.resizeHeight}let n=new r_(e,t);return await n.start(),n}async start(){this.webcamConfig.facingMode&&w.assert(this.webcamConfig.facingMode==="user"||this.webcamConfig.facingMode==="environment",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message=`Error thrown while initializing video stream: ${e.message}`,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=fi.fromPixels(this.webcamVideoElement)}catch(t){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(t)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(t){throw new Error(`Error thrown cropping the video: ${t.message}`)}finally{e.dispose()}else return{value:e,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(e){return O(()=>{let t=Zt(ce(e,"float32"),0),n;n=Br.cropAndResize(t,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let r=n.shape;return W(n,r.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(t=>t.stop());try{this.webcamVideoElement.srcObject=null}catch(t){console.log(t),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},s_=class{},a_=class extends tn{split(e){return new fq(this,e)}},fq=class extends a_{constructor(e,t){super(),this.upstream=e,this.impl=new mq(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},mq=class extends F0{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let e=await this.upstream.next();if(e.done)return this.carryover===""?!1:(this.outputQueue.push(this.carryover),this.carryover="",!0);let t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(let n of t.slice(0,-1))this.outputQueue.push(n);return this.carryover=t[t.length-1],!0}},gq=class extends tn{decodeUTF8(){return new bq(this)}},bq=class extends a_{constructor(e){super(),this.upstream=e,this.impl=new yq(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},yq=class extends F0{constructor(e){if(super(),this.upstream=e,q().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:t}=G1();this.decoder=new t("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let e=await this.upstream.next(),t;if(e.done)return!1;t=e.value;let n;return q().get("IS_BROWSER")?n=this.decoder.decode(t,{stream:!0}):n=this.decoder.write(Buffer.from(t.buffer)),this.outputQueue.push(n),!0}},o_=class extends gq{constructor(e,t={}){super(),this.file=e,this.options=t,w.assert(e instanceof Uint8Array||(q().get("IS_BROWSER")?e instanceof File||e instanceof Blob:!1),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((t,n)=>{let r=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)t(new Uint8Array(this.file.slice(this.offset,r)));else{let s=new FileReader;s.onload=o=>{let i=s.result;if(i instanceof ArrayBuffer&&(i=new Uint8Array(i)),!(i instanceof Uint8Array))return n(new TypeError("FileReader returned unknown type."));t(i)},s.onabort=o=>n(new Error("Aborted")),s.onerror=o=>n(new Error(o.type));let a=this.file.slice(this.offset,r);s.readAsArrayBuffer(a)}this.offset=r}),done:!1}}};async function vq(e,t={},n){let r,s;typeof e=="string"?r=e:(r=e.url,s=xq(e));let a=await(n||w.fetch)(r,s);if(a.ok){let o=new Uint8Array(await a.arrayBuffer());return new o_(o,t)}else throw new Error(a.statusText)}var xq=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function i_(e){return typeof e=="string"&&e.slice(0,7)==="file://"}var c_=class extends s_{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(i_(this.input)&&q().get("IS_NODE")){let e=Nv();this.input=e.readFileSync(this.input.slice(7))}return new o_(this.input,this.options)}},u_=class extends s_{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return i_(this.url)?new c_(this.url,this.fileOptions).iterator():vq(this.url,this.fileOptions)}};function wq(e,t={}){return new t_(new u_(e),t)}function Iq(e){let t=D0(e);return Yn(async()=>t)}function kq(e){return Yn(async()=>{let t=await e();return D0(()=>t.next())})}async function Sq(e,t){return r_.create(e,t)}async function Tq(e){return n_.create(e)}var Cq="4.0.0";function be(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&w.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var Nq=fs.whereImpl,R0=class extends ld{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new uf(this,Er())}nextDataId(){return R0.nextDataId++}write(e,t,n){this.firstUse&&(this.firstUse=!1,q().get("IS_NODE")&&N.warn(`
 ============================
 Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. 
-============================`));let a={id:this.nextDataId()};return this.data.set(a,{values:e,dtype:n,refCount:1}),a}makeTensorInfo(e,t,n){let a;if(t==="string"&&n!=null&&n.length>0&&v.isString(n[0])){let r=n.map(s=>v.encodeString(s));a=this.write(r,e,t)}else a=this.write(n,e,t);return{dataId:a,shape:e,dtype:t}}refCount(e){return this.data.has(e)?this.data.get(e).refCount:0}incRef(e){let t=this.data.get(e);t.refCount++}decRef(e){if(this.data.has(e)){let t=this.data.get(e);t.refCount--}}move(e,t,n,a,r){this.data.set(e,{values:t,dtype:a,refCount:r})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){let{dtype:t,complexTensorInfos:n}=this.data.get(e);if(t==="complex64"){let a=this.readSync(n.real.dataId),r=this.readSync(n.imag.dataId);return N.mergeRealAndImagArrays(a,r)}return this.data.get(e).values}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(a=>v.decodeString(a));return Oe(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Oe(e.shape,e.dtype,t)}makeOutput(e,t,n){return _a().makeTensorFromTensorInfo(this.makeTensorInfo(t,n,e),this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(e);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){ge([e],"where");let t=this.readSync(e.dataId);return Sj(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};R0.nextDataId=0;var lC={};Ae(lC,{addImpl:()=>cC,bincountImpl:()=>P0,bincountReduceImpl:()=>dC,castImpl:()=>pC,ceilImpl:()=>hC,concatImpl:()=>O0,equalImpl:()=>mC,expImpl:()=>gC,expm1Impl:()=>bC,floorImpl:()=>xC,gatherNdImpl:()=>vC,gatherV2Impl:()=>wC,greaterEqualImpl:()=>IC,greaterImpl:()=>kC,lessEqualImpl:()=>TC,lessImpl:()=>SC,linSpaceImpl:()=>NC,logImpl:()=>CC,maxImpl:()=>_C,maximumImpl:()=>EC,minimumImpl:()=>AC,multiplyImpl:()=>L0,negImpl:()=>$C,notEqualImpl:()=>FC,prodImpl:()=>DC,raggedGatherImpl:()=>RC,raggedRangeImpl:()=>MC,raggedTensorToTensorImpl:()=>PC,rangeImpl:()=>W0,rsqrtImpl:()=>OC,scatterImpl:()=>ll,sigmoidImpl:()=>bq,simpleAbsImpl:()=>uC,sliceImpl:()=>Jh,sparseFillEmptyRowsImpl:()=>zC,sparseReshapeImpl:()=>WC,sparseSegmentReductionImpl:()=>B0,sqrtImpl:()=>wq,squaredDifferenceImpl:()=>BC,stridedSliceImpl:()=>VC,stringNGramsImpl:()=>V0,stringSplitImpl:()=>U0,stringToHashBucketFastImpl:()=>G0,subImpl:()=>UC,tileImpl:()=>GC,topKImpl:()=>jC,transposeImpl:()=>z0,uniqueImpl:()=>qC});function uC(e){let t=new Float32Array(e.length);for(let n=0;n<e.length;++n)t[n]=Math.abs(e[n]);return t}var Tj=e=>{let{x:t}=e.inputs,n=e.backend;ge(t,"abs");let a=new Float32Array(v.sizeFromShape(t.shape)),r=n.data.get(t.dataId).values;return a=uC(r),n.makeOutput(a,t.shape,t.dtype)},Nj={kernelName:Dl,backendName:"cpu",kernelFunc:Tj};function Vt(e){return(t,n,a,r,s)=>{let i=N.assertAndGetBroadcastShape(t,n),o=i.length,l=v.computeStrides(i),u=v.sizeFromShape(i),p=v.getTypedArrayFromDType(s,u),d=t.length,c=n.length,h=v.computeStrides(t),m=v.computeStrides(n),f=N.getBroadcastDims(t,i),g=N.getBroadcastDims(n,i);if(f.length+g.length===0)for(let y=0;y<p.length;++y)p[y]=e(a[y%a.length],r[y%r.length]);else for(let y=0;y<p.length;++y){let b=v.indexToLoc(y,o,l),x=b.slice(-d);f.forEach(C=>x[C]=0);let w=v.locToIndex(x,d,h),I=b.slice(-c);g.forEach(C=>I[C]=0);let T=v.locToIndex(I,c,m);p[y]=e(a[w],r[T])}return[p,i]}}function Zn(e){let{inputs:t,backend:n}=e,{real:a,imag:r}=t,s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,o=n.makeTensorInfo(a.shape,"complex64"),l=n.data.get(o.dataId);return l.complexTensorInfos={real:n.makeTensorInfo(a.shape,"float32",s),imag:n.makeTensorInfo(r.shape,"float32",i)},o}var Cj={kernelName:mm,backendName:"cpu",kernelFunc:Zn};function Zh(e,t,n="float32"){if(n==="complex64"){let r=Zh(e,t,"float32"),s=Zh(e,t,"float32");return Zn({inputs:{real:r,imag:s},backend:e})}let a=v.makeZerosTypedArray(v.sizeFromShape(t),n);return e.makeTensorInfo(t,n,a)}function pr(e){let{inputs:t,backend:n}=e,{x:a}=t;return n.incRef(a.dataId),{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}var _j={kernelName:Ri,backendName:"cpu",kernelFunc:pr};function ui(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.data.get(a.dataId).complexTensorInfos.real,s=n.data.get(r.dataId).values;return n.makeTensorInfo(r.shape,r.dtype,s)}var Ej={kernelName:Pm,backendName:"cpu",kernelFunc:ui};function pC(e,t,n,a){if(a==="int32"){let r=Int32Array.from(e);return[t,"int32",r]}if(a==="bool"){let r=v.toTypedArray([0],n),[s,i]=Vt((o,l)=>o!==l?1:0)(t,[],e,r,"bool");return[i,"bool",s]}throw new Error(`Error in Cast: failed to cast ${n} to ${a}`)}function ds(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dtype:s}=a;if(s==="complex64"){if(r.dtype==="complex64")return pr({inputs:{x:r},backend:n});let p=Zh(n,r.shape,r.dtype),d=ds({inputs:{x:r},backend:n,attrs:{dtype:"float32"}}),c=Zn({inputs:{real:d,imag:p},backend:n});return n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),c}if(r.dtype==="complex64"){let p=ui({inputs:{input:r},backend:n}),d=ds({inputs:{x:p},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(p),d}if(!v.hasEncodingLoss(r.dtype,s)){let p=pr({inputs:{x:r},backend:n});return{dataId:p.dataId,shape:p.shape,dtype:s}}let i=n.data.get(r.dataId).values,[o,l,u]=pC(i,r.shape,r.dtype,s);return n.makeTensorInfo(o,l,u)}var Aj={kernelName:xi,backendName:"cpu",kernelFunc:ds};function nn(e,t,n,a){return n==null?({inputs:r,backend:s})=>{let{a:i,b:o}=r,l=s;ge([i,o],e);let u=l.data.get(i.dataId).values,p=l.data.get(o.dataId).values,d=i.dtype==="string"?N.fromUint8ToStringArray(u):u,c=i.dtype==="string"?N.fromUint8ToStringArray(p):p,h=a||i.dtype,[m,f]=t(i.shape,o.shape,d,c,h);return l.makeTensorInfo(f,h,m)}:({inputs:r,backend:s})=>{let{a:i,b:o}=r,l=s;if(i.dtype==="complex64"||o.dtype==="complex64"){let u=ds({inputs:{x:i},backend:l,attrs:{dtype:"complex64"}}),p=l.data.get(u.dataId),d=p.complexTensorInfos.real,c=p.complexTensorInfos.imag,h=l.data.get(d.dataId).values,m=l.data.get(c.dataId).values,f=ds({inputs:{x:o},backend:l,attrs:{dtype:"complex64"}}),g=l.data.get(f.dataId),y=g.complexTensorInfos.real,b=g.complexTensorInfos.imag,x=l.data.get(y.dataId).values,w=l.data.get(b.dataId).values,[I,T,C]=n(i.shape,o.shape,h,m,x,w),E=l.makeTensorInfo(C,"float32",I),A=l.makeTensorInfo(C,"float32",T),R=Zn({inputs:{real:E,imag:A},backend:l});return l.disposeIntermediateTensorInfo(u),l.disposeIntermediateTensorInfo(f),l.disposeIntermediateTensorInfo(E),l.disposeIntermediateTensorInfo(A),R}else{let u=l.data.get(i.dataId).values,p=l.data.get(o.dataId).values,d=a||i.dtype,[c,h]=t(i.shape,o.shape,u,p,d);return l.makeTensorInfo(h,d,c)}}}function M0(e){return(t,n,a,r,s,i)=>{let o=N.assertAndGetBroadcastShape(t,n),l=v.sizeFromShape(o),u=o.length,p=v.computeStrides(o),d=v.getTypedArrayFromDType("float32",l),c=v.getTypedArrayFromDType("float32",l),h=N.getBroadcastDims(t,o),m=N.getBroadcastDims(n,o),f=N.mergeRealAndImagArrays(a,r),g=N.mergeRealAndImagArrays(s,i),y=t.length,b=v.computeStrides(t),x=n.length,w=v.computeStrides(n);if(h.length+m.length===0)for(let I=0;I<d.length;I++){let T=I%f.length,C=I%g.length,E=e(f[T*2],f[T*2+1],g[C*2],g[C*2+1]);d[I]=E.real,c[I]=E.imag}else for(let I=0;I<d.length;I++){let T=v.indexToLoc(I,u,p),C=T.slice(-y);h.forEach(S=>C[S]=0);let E=v.locToIndex(C,y,b),A=T.slice(-x);m.forEach(S=>A[S]=0);let R=v.locToIndex(A,x,w),F=e(f[E*2],f[E*2+1],g[R*2],g[R*2+1]);d[I]=F.real,c[I]=F.imag}return[d,c,o]}}var cC=Vt((e,t)=>e+t),$j=M0((e,t,n,a)=>({real:e+n,imag:t+a})),_l=nn(gs,cC,$j),Fj={kernelName:gs,backendName:"cpu",kernelFunc:_l};function P0(e,t,n,a,r){let s=v.sizeFromShape(a),i=v.makeZerosTypedArray(r,n);for(let o=0;o<e.length;o++){let l=e[o];if(l<0)throw new Error("Input x must be non-negative!");l>=r||(s>0?i[l]+=t[o]:i[l]+=1)}return i}function dC(e,t,n,a=!1){let r=e.shape[0],s=e.shape[1],i=Oe([r,n],t.dtype);for(let o=0;o<r;o++)for(let l=0;l<s;l++){let u=e.get(o,l);if(u<0)throw new Error("Input x must be non-negative!");u>=n||(a?i.set(1,o,u):t.size>0?i.set(i.get(o,u)+t.get(o,l),o,u):i.set(i.get(o,u)+1,o,u))}return i}function Cs(e){return(t,n,a)=>{let r=v.getTypedArrayFromDType(n,t.length);for(let s=0;s<t.length;++s)r[s]=e(t[s],a);return r}}function rt(e,t,n){return({inputs:a,attrs:r,backend:s})=>{let{x:i}=a;if(ge(i,e),i.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=v.sizeFromShape(i.shape),p=n||i.dtype,d=v.getArrayFromDType(p,u);for(let c=0;c<u;++c)d[c]=t(l[c],r);return o.makeTensorInfo(i.shape,p,d)}}function Ku(e,t,n){return({inputs:a,attrs:r,backend:s})=>{let{x:i}=a;if(ge(i,e),i.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=n||i.dtype,p=t(l,u,r);return o.makeTensorInfo(i.shape,u,p)}}var hC=Cs(e=>Math.ceil(e)),Dj=Ku(vi,hC),Rj={kernelName:vi,backendName:"cpu",kernelFunc:Dj};function O0(e,t,n,a){let r=v.getArrayFromDType(n,v.sizeFromShape(t));if(a&&n!=="string"){let s=0;e.forEach(i=>{let o=v.sizeFromShape(i.shape);r.set(i.vals,s),s+=o})}else{let s=0;e.forEach(i=>{let o=n==="string"?N.fromUint8ToStringArray(i.vals):i.vals,l=0;for(let u=0;u<i.shape[0];++u){let p=u*t[1]+s;for(let d=0;d<i.shape[1];++d)r[p+d]=o[l++]}s+=i.shape[1]})}return r}var mC=Vt((e,t)=>e===t?1:0),fC=nn(Xl,mC,null,"bool"),Mj={kernelName:Xl,backendName:"cpu",kernelFunc:fC},gC=Cs(e=>Math.exp(e)),yC=Ku(Ei,gC,"float32"),Pj={kernelName:Ei,backendName:"cpu",kernelFunc:yC},bC=Cs(e=>Math.expm1(e)),Oj=Ku(Zl,bC),Lj={kernelName:Zl,backendName:"cpu",kernelFunc:Oj},xC=Cs(e=>Math.floor(e)),zj=Ku(Ai,xC),Wj={kernelName:Ai,backendName:"cpu",kernelFunc:zj};function vC(e,t,n,a,r,s,i,o,l){let u=Oe([a,s],n);for(let p=0;p<a;p++){let d=[],c=0;for(let h=0;h<r;h++){let m=e[p*r+h];c+=m*i[h],d.push(m)}if(c<0||c>=l/s)throw new Error(`Invalid indices: ${d} does not index into ${o}`);for(let h=0;h<s;h++)u.values[p*s+h]=t.get(...t.indexToLoc(c*s+h))}return u}function wC(e,t,n){let a=Oe(n,e.dtype);for(let r=0;r<a.size;++r){let s=a.indexToLoc(r).slice(),i=s[0],o=s[2],l=t.locToIndex([i,o]);s[2]=t.values[l];let u=e.locToIndex(s);0<=u&&u<e.values.length&&(a.values[r]=e.values[u])}return a}var kC=Vt((e,t)=>e>t?1:0),Bj=nn(tu,kC,null,"bool"),Vj={kernelName:tu,backendName:"cpu",kernelFunc:Bj},IC=Vt((e,t)=>e>=t?1:0),Uj=nn(Di,IC,null,"bool"),Gj={kernelName:Di,backendName:"cpu",kernelFunc:Uj},SC=Vt((e,t)=>e<t?1:0),Hj=nn(su,SC,null,"bool"),jj={kernelName:su,backendName:"cpu",kernelFunc:Hj},TC=Vt((e,t)=>e<=t?1:0),qj=nn(iu,TC,null,"bool"),Kj={kernelName:iu,backendName:"cpu",kernelFunc:qj};function NC(e,t,n){let a=(t-e)/(n-1),r=v.makeZerosTypedArray(n,"float32");r[0]=e;for(let s=1;s<r.length;s++)r[s]=r[s-1]+a;return r}var CC=Cs(e=>Math.log(e)),Xj=Ku(Pi,CC),Yj={kernelName:Pi,backendName:"cpu",kernelFunc:Xj};function _C(e,t,n,a){let r=v.getTypedArrayFromDType(a,v.sizeFromShape(n));for(let s=0;s<r.length;++s){let i=s*t,o=e[i];for(let l=0;l<t;++l){let u=e[i+l];(Number.isNaN(u)||u>o)&&(o=u)}r[s]=o}return r}var EC=Vt((e,t)=>Math.max(e,t)),Zj=nn(Li,EC),Jj={kernelName:Li,backendName:"cpu",kernelFunc:Zj},AC=Vt((e,t)=>Math.min(e,t)),Qj=nn(Vi,AC),eq={kernelName:Vi,backendName:"cpu",kernelFunc:Qj},L0=Vt((e,t)=>e*t),tq=M0((e,t,n,a)=>({real:e*n-t*a,imag:e*a+t*n})),qf=nn(Gi,L0,tq),nq={kernelName:Gi,backendName:"cpu",kernelFunc:qf};function $C(e,t,n){let a=v.createScalarValue(-1,n);return L0([],t,a,e,n)}function aq(e){let{inputs:t,backend:n}=e,{x:a}=t;ge(a,"neg");let r=n.data.get(a.dataId).values,[s,i]=$C(r,a.shape,a.dtype);return n.makeTensorInfo(i,a.dtype,s)}var rq={kernelName:du,backendName:"cpu",kernelFunc:aq},FC=Vt((e,t)=>e!==t?1:0),sq=nn(hu,FC,null,"bool"),iq={kernelName:hu,backendName:"cpu",kernelFunc:sq};function z0(e,t,n,a,r){let s=t.length,i=v.sizeFromShape(t),o=v.computeStrides(t),l=v.computeStrides(r),u=v.getTypedArrayFromDType(n,v.sizeFromShape(r));for(let p=0;p<i;++p){let d=v.indexToLoc(p,s,o),c=new Array(d.length);for(let m=0;m<c.length;m++)c[m]=d[a[m]];let h=v.locToIndex(c,s,l);u[h]=e[p]}return u}function Vn(e){let{inputs:t,attrs:n,backend:a}=e,{x:r}=t,{perm:s}=n;ge(r,"transpose");let i=r.shape.length,o=new Array(i);for(let p=0;p<o.length;p++)o[p]=r.shape[s[p]];let l=a.data.get(r.dataId).values,u=z0(l,r.shape,r.dtype,s,o);return{dataId:a.write(u,o,r.dtype),shape:o,dtype:r.dtype}}var oq={kernelName:_r,backendName:"cpu",kernelFunc:Vn};function DC(e,t,n,a){let[r,s]=N.computeOutAndReduceShapes(e,a),i=fa(t,"int32"),o=v.makeZerosTypedArray(v.sizeFromShape(r),i),l=v.sizeFromShape(s);for(let u=0;u<o.length;++u){let p=u*l,d=1;for(let c=0;c<l;++c)d*=n[p+c];o[u]=d}return{outVals:o,outShape:r,outDtype:i}}function lq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"prod");let o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=N.getAxesPermutation(l,o),p=l,d=r,c=[];u!=null&&(d=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),c.push(d),p=N.getInnerMostAxes(p.length,o));let h=n.data.get(d.dataId).values,{outVals:m,outShape:f,outDtype:g}=DC(d.shape,d.dtype,h,p),y=f;return i&&(y=N.expandShapeToKeepDim(f,l)),c.forEach(b=>n.disposeIntermediateTensorInfo(b)),n.makeTensorInfo(y,g,m)}var uq={kernelName:Xi,backendName:"cpu",kernelFunc:lq};function pq(e,t,n){e.forEach((a,r)=>{if(a<0||a>=n){let s=v.indexToLoc(r,t.length,v.computeStrides(t)).join(",");throw new Error(`indices[${s}] = ${a} is not in [0, ${n})`)}})}function cq(e,t){for(let n=0;n<e.length;++n){let a=e[n],r=n===e.length-1?t:e[n+1].length;if(a.length===0)throw new Error("Ragged splits may not be empty");if(a[0]<0)throw new Error("Ragged splits must be non-negative");if(a[a.length-1]>r)throw new Error("Ragged splits must not point past values");for(let s=1;s<a.length;++s)if(a[s-1]>a[s])throw new Error("Ragged splits must be sorted in ascending order")}}function dq(e,t,n,a){let r=[],s=0,i=t.length-1+n.length,o=new Array(i).fill(null).map(()=>[0]);cq(n,a);let l=1;for(let u=0;u<t.length-1;++u){l*=t[u];let p=t[u+1];for(let d=1;d<l+1;++d)o[u].push(d*p)}for(let u=0;u<e.length;++u){let p=e[u],d=e[u]+1;for(let c=0;c<n.length;++c){let h=n[c],m=c+t.length-1;if(m>=0){let f=o[m],g=f[f.length-1]-h[p];for(let y=p;y<d;++y)o[m].push(h[y+1]+g)}p=h[p],d=h[d]}d!==p&&(r.push([p,d]),s+=d-p)}return{outSplits:o,valueSlices:r,numValues:s}}function hq(e){let t=[];for(let n=0;n<e.length;++n){let a=e[n].length,r=v.getArrayFromDType("int32",a);t.push(r),e[n].forEach((s,i)=>r[i]=s)}return t}function iI(e,t){let n=e.slice(0,t);for(;n.length<t;)n.push(1);for(let a=t;a<e.length;a++)n[t-1]*=e[a];return n}function mq(e,t,n,a,r,s){let i=iI(t,2)[1],o=iI(s,2)[1],l=0;for(let u of n)for(let p=u[0];p<u[1];++p){for(let d=0;d<a;++d)r[l*o+d]=e[p*i+d];++l}}function fq(e,t,n,a,r){let s=t.slice();s[0]=r;let i=v.getArrayFromDType(n,v.sizeFromShape(s)),o=e.length,l=o===0?0:o/t[0];return mq(e,t,a,l,i,s),[i,s]}function RC(e,t,n,a,r,s,i,o){if(e.length===0)throw new Error("paramsNestedSplits must be non empty");if(t[0].length===0)throw new Error("Split tensors must not be scalars");let l=t[0][0]-1;if(pq(s,i,l),a.length===0)throw new Error("params.rank must be nonzero");let u=a[0],{outSplits:p,valueSlices:d,numValues:c}=dq(s,i,e,u),h=hq(p),m=fq(n,a,r,d,c);return[h,m[0],m[1]]}var oI=2147483647;function MC(e,t,n,a,r,s,i){if(t.length>1)throw new Error("starts must be a scalar or vector");if(r.length>1)throw new Error("limits must be a scalar or vector");if(i.length>1)throw new Error("deltas must be a scalar or vector");let o=t.length===0,l=r.length===0,u=i.length===0,p=[];o||p.push(t[0]),l||p.push(r[0]),u||p.push(i[0]);for(let g=1;g<p.length;++g)if(p[g]!==p[g-1])throw new Error("starts, limits, and deltas must have the same shape");let d=p.length===0?1:p[0],c=v.getArrayFromDType("int32",d+1);c[0]=0;for(let g=0;g<d;++g){let y=o?e[0]:e[g],b=l?a[0]:a[g],x=u?s[0]:s[g];if(x===0)throw new Error("Requires delta != 0");let w;if(x>0&&b<y||x<0&&b>y)w=0;else if(w=Math.ceil(Math.abs((b-y)/x)),w>oI)throw new Error(`Requires ((limit - start) / delta) <= ${oI}`);c[g+1]=c[g]+w}let h=c[d],m=v.getArrayFromDType(n,h),f=0;for(let g=0;g<d;++g){let y=c[g+1]-c[g],b=o?e[0]:e[g],x=u?s[0]:s[g];for(let w=0;w<y;++w)m[f++]=b,b+=x}return[c,m]}var Ta=N.RowPartitionType,cx=class{constructor(e,t,n,a,r,s,i,o,l,u){this.shape=e,this.shapeShape=t,this.values=n,this.valuesShape=a,this.valuesDType=r,this.defaultValue=s,this.defaultValueShape=i,this.rowPartitionValues=o,this.rowPartitionValuesShapes=l,this.rowPartitionTypes=N.getRowPartitionTypesHelper(u),this.raggedRank=N.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(e){return this.rowPartitionTypes[0]===Ta.FIRST_DIM_SIZE?this.rowPartitionTypes[e+1]:this.rowPartitionTypes[e]}getRowPartitionTensor(e){return this.rowPartitionTypes[0]===Ta.FIRST_DIM_SIZE?this.rowPartitionValues[e+1]:this.rowPartitionValues[e]}getMaxWidth(e){let t=this.getRowPartitionTensor(e-1);switch(this.getRowPartitionTypeByDimension(e-1)){case Ta.VALUE_ROWIDS:return cx.getMaxWidthValueRowID(t);case Ta.ROW_SPLITS:return cx.getMaxWidthRowSplit(t);default:throw new Error(`Cannot handle partition type ${Ta[this.getRowPartitionTypeByDimension(e-1)]}`)}}static getMaxWidthRowSplit(e){let t=e.length;if(t===0||t===1)return 0;let n=0;for(let a=0;a<t-1;++a){let r=e[a+1]-e[a];r>n&&(n=r)}return n}static getMaxWidthValueRowID(e){let t=e.length;if(t===0)return 0;let n=0,a=e[0],r=0;for(let s=1;s<t;++s){let i=e[s];i!==a&&(a=i,r=Math.max(s-n,r),n=s)}return Math.max(t-n,r)}tensorShapeFromTensor(e,t,n=!0){if(t.length===0){if(e[0]===-1)return[];throw new Error("The only valid scalar shape tensor is the fully unknown shape specified as -1.")}return uI(e,n)}calculateOutputSize(e){let t=this.valuesShape,n=this.defaultValueShape;N.validateDefaultValueShape(n,t);let a=this.tensorShapeFromTensor(this.shape,this.shapeShape),r=N.combineRaggedTensorToTensorShapes(this.raggedRank,a,t);r[0]<0&&(r[0]=e);for(let s=1;s<=this.raggedRank;++s)r[s]<0&&(r[s]=this.getMaxWidth(s));return r}calculateFirstParentOutputIndex(e,t,n){let a=Math.min(e,n),r=[],s=0;for(let i=0;i<a;++i,s+=t)r.push(s);for(let i=a;i<e;++i)r.push(-1);return v.assert(r.length===e,()=>"Final length of result must be equal to firstDimension."),r}calculateOutputIndexRowSplit(e,t,n,a){let r=e.length,s=[];for(let i=0;i<r-1;++i){let o=e[i+1]-e[i],l=Math.min(a,o),u=t[i];u===-1&&(l=0);for(let p=0;p<l;++p)s.push(u),u+=n;for(let p=0;p<o-l;++p)s.push(-1)}if(r>0&&s.length!==e[r-1])throw new Error("Invalid row split size.");return s}calculateOutputIndexValueRowID(e,t,n,a){let r=e.length,s=[];if(r===0)return[];let i=0,o=e[0];if(o>=t.length)throw new Error(`Got currentValueRowId=${o}, which is not less than ${t.length}`);let l=t[o];s.push(l);for(let u=1;u<r;++u){let p=e[u];if(p===o)l>=0&&(++i,i<a?l+=n:l=-1);else{if(i=0,o=p,p>=t.length)throw new Error(`Got nextValueRowId=${p} which is not less than ${t.length}`);l=t[p]}s.push(l)}if(s.length!==e.length)throw new Error("Invalid row ids.");return s}calculateOutputIndex(e,t,n,a){let r=this.getRowPartitionTensor(e),s=this.getRowPartitionTypeByDimension(e);switch(s){case Ta.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(r,t,n,a);case Ta.ROW_SPLITS:if(r.length-1>t.length)throw new Error(`Row partition size is greater than output size: ${r.length-1} > ${t.length}`);return this.calculateOutputIndexRowSplit(r,t,n,a);default:throw new Error(`Unsupported partition type: ${Ta[s]}`)}}getFirstDimensionSize(){let e=this.rowPartitionValues[0];if(this.rowPartitionTypes.length===0)throw new Error("No row_partition_types given.");let t=this.rowPartitionTypes[0];switch(t){case Ta.FIRST_DIM_SIZE:return e[0];case Ta.VALUE_ROWIDS:throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");case Ta.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${Ta[t]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error("Invalid first partition input. Tensor requires at least one element.");let e=this.getFirstDimensionSize(),t=this.calculateOutputSize(e),n=new Array(this.raggedRank+1);n[n.length-1]=1;for(let s=n.length-2;s>=0;--s)n[s]=n[s+1]*t[s+1];let a=uI(t,!1),r=v.getArrayFromDType(this.valuesDType,v.sizeFromShape(a));if(n[0]*t[0]>0){let s=this.calculateFirstParentOutputIndex(e,n[0],t[0]);for(let i=1;i<=this.raggedRank;++i)s=this.calculateOutputIndex(i-1,s,n[i],t[i]);this.setOutput(this.raggedRank,s,r,a)}return[a,r]}setOutput(e,t,n,a){if(n.length===0)return;let r=this.values,s=n,i=a.slice();i=i.slice(e+1);let o=v.sizeFromShape(i),l=t.length,u=this.defaultValue;if(u.length!==o&&u.length!==1){let h=this.defaultValueShape;P(()=>{let m=W(u,h);u=Xs(m,i).dataSync()})}let p=0,d=0,c=0;for(let h=0;h<=l;++h){let m=h<l?t[h]:-1;if(m===c){++c;continue}if(d<c){let f=r.subarray(p*o),g=s.subarray(d*o),y=(c-d)*o;lI(g,f,y)}if(h>=l){let f=n.length;m=Math.floor(f/o)}if(m>c)if(this.defaultValue.length===1)s.subarray(c*o,m*o).fill(this.defaultValue[0]),c=m;else for(;m>c;){let f=s.slice(c*o);lI(f,u,o),++c}m<0?(p=h+1,d=c):(p=h,d=c,c=d+1)}}};function lI(e,t,n){for(let a=0;a<n;a++)e[a]=t[a]}function uI(e,t){let n=[];for(let a of e){if(a<0){if(!t)throw new Error(`Dimension ${a} must be >= 0`);if(a<-1)throw new Error(`Dimension ${a} must be >= -1`);a=-1}n.push(a)}return n}function PC(e,t,n,a,r,s,i,o,l,u){return new cx(e,t,n,a,r,s,i,o,l,u).compute()}function W0(e,t,n,a){let r=e===t,s=e<t&&n<0,i=t<e&&n>1;if(r||s||i)return v.makeZerosTypedArray(0,a);let o=Math.abs(Math.ceil((t-e)/n)),l=v.makeZerosTypedArray(o,a);t<e&&n===1&&(n=-1),l[0]=e;for(let u=1;u<l.length;u++)l[u]=l[u-1]+n;return l}var OC=Cs(e=>1/Math.sqrt(e)),gq=Ku(no,OC),yq={kernelName:no,backendName:"cpu",kernelFunc:gq};function ll(e,t,n,a,r,s,i,o,l,u){let p=[a/r,r],d=e.values,c=t.values;if(a===0)return Oe(n,t.dtype);let h=Oe(p,t.dtype);typeof l=="string"||typeof l=="number"?h.values.fill(l):typeof l=="boolean"&&h.values.fill(+l);for(let m=0;m<s;m++){let f=[],g=0;for(let y=0;y<i;y++){let b=d[m*i+y];f.push(b),g+=b*o[y]}if(g<0||g>=a/r)throw new Error(`Invalid indices: ${f} does not index into ${n}`);for(let y=0;y<r;y++)u?h.values[g*r+y]+=c[m*r+y]:h.values[g*r+y]=t.rank===0?c[0]:c[m*r+y]}return h}var bq=Cs(e=>1/(1+Math.exp(-e))),LC=rt(ro,e=>1/(1+Math.exp(-e))),xq={kernelName:ro,backendName:"cpu",kernelFunc:LC};function Jh(e,t,n,a,r){let s=jt.isSliceContinous(a,t,n),i=v.sizeFromShape(n),o=v.computeStrides(a);if(s){let d=jt.computeFlatOffset(t,o);return r==="string"?e.slice(d,d+i):e.subarray(d,d+i)}let l=r==="string"?N.fromUint8ToStringArray(e):e,u=Oe(a,r,l),p=Oe(n,r);for(let d=0;d<p.size;++d){let c=p.indexToLoc(d),h=c.map((m,f)=>m+t[f]);p.set(u.get(...h),...c)}return r==="string"?N.fromStringArrayToUint8(p.values):p.values}function pi(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,size:i}=a;ge(r,"slice");let[o,l]=jt.parseSliceParams(r,s,i);jt.assertParamsValid(r,o,l);let u=n.data.get(r.dataId).values,p=Jh(u,o,l,r.shape,r.dtype);return n.makeTensorInfo(l,r.dtype,p)}var vq={kernelName:Su,backendName:"cpu",kernelFunc:pi};function zC(e,t,n,a,r,s,i){let o=t[0],l=s[0],u=new Array(l),p=new Array(o),d=t[1];if(l===0){if(o!==0)throw new Error(N.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));let g=v.getArrayFromDType(n,0),y=v.getArrayFromDType(r,0);return[g,[0,d],y,u,p]}let c=!0,h=0,m=new Array(l).fill(0);for(let g=0;g<o;++g){let y=e[g*d];if(y<0)throw new Error(N.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,y));if(y>=l)throw new Error(N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,y,l));++m[y],c=c&&y>=h,h=y}let f=!0;for(let g=0;g<l;++g){let y=m[g]===0;u[g]=y,f=f&&!y,m[g]=Math.max(m[g],1),g>0&&(m[g]+=m[g-1])}if(f&&c){let g=e,y=a;for(let b=0;b<o;++b)p[b]=b;return[g,[o,d],y,u,p]}else{let g=m[l-1],y=v.getArrayFromDType(n,g*d),b=v.getArrayFromDType(r,g),x=new Array(l).fill(0);for(let w=0;w<o;++w){let I=e[w*d],T=x[I],C=(I===0?0:m[I-1])+T;x[I]++;for(let E=0;E<d;++E)y[C*d+E]=e[w*d+E];b[C]=a[w],p[w]=C}for(let w=0;w<l;++w)if(x[w]===0){let I=w===0?0:m[w-1];y[I*d+0]=w;for(let T=1;T<d;++T)y[I*d+T]=0;b[I]=i}return[y,[g,d],b,u,p]}}function WC(e,t,n,a,r){let s=v.sizeFromShape(a),i=t[0],o=r.length,l=[],u=1,p=-1;for(let f=0;f<o;++f){let g=r[f];if(g===-1){if(p!==-1)throw new Error(N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(p,f));p=f,l.push(1)}else{if(g<0)throw new Error(N.getSparseReshapeNegativeOutputDimErrorMessage(f,g));u*=g,l.push(g)}}if(p!==-1){if(u<=0)throw new Error(N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let f=Math.trunc(s/u);if(u*f!==s)throw new Error(N.getSparseReshapeInputOutputMultipleErrorMessage(a,l));l[p]=f}if(v.sizeFromShape(l)!==s)throw new Error(N.getSparseReshapeInputOutputMismatchErrorMessage(a,l));let d=a.length,c=[];if(d>0){c[d-1]=1;for(let f=d-2;f>=0;--f)c[f]=c[f+1]*a[f+1]}let h=[];if(o>0){h[o-1]=1;for(let f=o-2;f>=0;--f)h[f]=h[f+1]*l[f+1]}let m=v.getArrayFromDType(n,i*o);for(let f=0;f<i;++f){let g=0;for(let y=0;y<d;++y)g+=e[f*d+y]*c[y];for(let y=0;y<o;++y)m[f*o+y]=Math.trunc(g/h[y]),g%=h[y]}return[m,[i,o],l]}function B0(e,t,n,a,r,s=!1,i=0){let o=a.length,l=[t[0],e.length/t[0]],u=l[1],p=o>0?r[o-1]+1:0;if(p<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let d=t.slice();d[0]=p;let c=d.reduce((b,x)=>b*x,1),h=v.getArrayFromDType(n,c);if(o===0)return p>0&&h.fill(i),[h,d];if(p<=0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let m=0,f=1,g=0,y=r[m];for(;;){let b=0;if(f<o){if(b=r[f],y===b){++f;continue}if(y>=b)throw new Error(N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=p)throw new Error(N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,p));y>g&&h.fill(i,g*u,y*u);for(let x=m;x<f;++x){let w=a[x];if(w<0||w>=l[0])throw new Error(N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(x,a[x],l[0]));for(let I=0;I<u;I++)h[y*u+I]+=e[w*u+I]}if(s)for(let x=0;x<u;x++)h[y*u+x]/=f-m;if(m=f,++f,g=y+1,y=b,f>o)break}return g<p&&h.fill(i,g*u,p*u),[h,d]}var wq=Cs(e=>Math.sqrt(e)),kq=rt(so,e=>Math.sqrt(e)),Iq={kernelName:so,backendName:"cpu",kernelFunc:kq},BC=Vt((e,t)=>{let n=e-t;return n*n}),Sq=nn(lo,BC),Tq={kernelName:lo,backendName:"cpu",kernelFunc:Sq};function VC(e,t,n,a){let r=Oe(e,t.dtype);for(let s=0;s<r.size;s++){let i=r.indexToLoc(s),o=new Array(i.length);for(let l=0;l<o.length;l++)o[l]=i[l]*n[l]+a[l];r.set(t.get(...o),...i)}return r}var Nq=class{constructor(e,t,n,a,r,s){this.separator=v.encodeString(e),this.nGramWidths=t,this.leftPad=v.encodeString(n),this.rightPad=v.encodeString(a),this.padWidth=r,this.preserveShort=s}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){let n=this.getPadWidth(t);return Math.max(0,e+2*n-t+1)}createNGrams(e,t,n,a,r,s){for(let i=0;i<r;++i){let o=this.getPadWidth(s),l=Math.max(0,o-i),u=Math.max(0,o-(r-(i+1))),p=s-(l+u),d=t+(l>0?0:i-o),c=0;c+=l*this.leftPad.length;for(let g=0;g<p;++g)c+=e[d+g].length;c+=u*this.rightPad.length,c+=(l+u+p-1)*this.separator.length,n[a+i]=new Uint8Array(c);let h=n[a+i],m=0,f=g=>g.forEach(y=>h[m++]=y);for(let g=0;g<l;++g)f(this.leftPad),f(this.separator);for(let g=0;g<p-1;++g)f(e[d+g]),f(this.separator);if(p>0){f(e[d+p-1]);for(let g=0;g<u;++g)f(this.separator),f(this.rightPad)}else{for(let g=0;g<u-1;++g)f(this.rightPad),f(this.separator);f(this.rightPad)}}}compute(e,t){let n=e.length,a=t.length;if(a>0){let o=t[0];if(o!==0)throw new Error(`First split value must be 0, got ${o}`);for(let l=1;l<a;++l){let u=t[l]>=o;if(u=u&&t[l]<=n,!u)throw new Error(`Invalid split value ${t[l]}, must be in [${o}, ${n}]`);o=t[l]}if(o!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${o}`)}let r=a-1,s=v.getArrayFromDType("int32",a);if(n===0||a===0){let o=new Array(n);for(let l=0;l<=r;++l)s[l]=0;return[o,s]}s[0]=0;for(let o=1;o<=r;++o){let l=t[o]-t[o-1],u=0;this.nGramWidths.forEach(p=>{u+=this.getNumNGrams(l,p)}),this.preserveShort&&l>0&&u===0&&(u=1),s[o]=s[o-1]+u}let i=new Array(s[r]);for(let o=0;o<r;++o){let l=t[o],u=s[o];if(this.nGramWidths.forEach(p=>{let d=t[o+1]-t[o],c=this.getNumNGrams(d,p);this.createNGrams(e,l,i,u,c,p),u+=c}),this.preserveShort&&u===s[o]){let p=t[o+1]-t[o];if(p===0)continue;let d=p+2*this.padWidth,c=1;this.createNGrams(e,l,i,u,c,d)}}return[i,s]}};function V0(e,t,n,a,r,s,i,o){return new Nq(n,a,r,s,i,o).compute(e,t)}function Cq(e,t,n,a){if(!e.length)return;if(t.length===0){for(let s=0;s<e.length;++s)a.push(e.subarray(s,s+1));return}if(t.length===1){let s=t[0],i=e.indexOf(s);for(;i!==-1;){let o=e.subarray(0,i);(!n||o.length!==0)&&a.push(o),e=e.subarray(i+1),i=e.indexOf(s)}(!n||e.length!==0)&&a.push(e);return}let r=0;for(let s=0;s<e.length+1;s++)if(s===e.length||t.indexOf(e[s])!==-1){let i=e.subarray(r,s);(!n||i.length!==0)&&a.push(i),r=s+1}}function U0(e,t,n){let a=e.length,r=[],s=0,i=0,o=new Array(a);for(let c=0;c<a;++c){let h=r.length;Cq(e[c],t,n,r);let m=r.length-h;o[c]=m,s+=m,i=Math.max(i,m)}let l=v.getArrayFromDType("int32",s*2),u=new Array(s),p=[a,i],d=0;for(let c=0;c<a;++c)for(let h=0;h<o[c];++h)l[d*2]=c,l[d*2+1]=h,u[d]=r[d],++d;return[l,u,p]}function G0(e,t){let n=v.getArrayFromDType("int32",e.length);for(let a=0;a<e.length;++a)n[a]=v.fingerPrint64(e[a]).modulo(t).getLowBitsUnsigned();return n}var UC=Vt((e,t)=>e-t),_q=M0((e,t,n,a)=>({real:e-n,imag:t-a})),H0=nn(uo,UC,_q),Eq={kernelName:uo,backendName:"cpu",kernelFunc:H0};function GC(e,t){let n=new Array(e.rank);for(let r=0;r<n.length;r++)n[r]=e.shape[r]*t[r];let a=Oe(n,e.dtype);for(let r=0;r<a.values.length;++r){let s=a.indexToLoc(r),i=new Array(e.rank);for(let l=0;l<i.length;l++)i[l]=s[l]%e.shape[l];let o=e.locToIndex(i);a.values[r]=e.values[o]}return a}var Op=(e,t)=>{let n=t.value-e.value;return n===0?e.index-t.index:n};function HC(e,t,n=0,a=e.length-1){for(;a>n;){if(a-n>600){let o=a-n+1,l=t-n+1,u=Math.log(o),p=.5*Math.exp(2*u/3),d=.5*Math.sqrt(u*p*(o-p)/o)*Math.sign(l-o/2),c=Math.max(n,Math.floor(t-l*p/o+d)),h=Math.min(a,Math.floor(t+(o-l)*p/o+d));HC(e,t,c,h)}let r=e[t],s=n,i=a;for(v.swap(e,n,t),Op(e[a],r)>0&&v.swap(e,n,a);s<i;){for(v.swap(e,s,i),s++,i--;Op(e[s],r)<0;)s=s+1;for(;Op(e[i],r)>0;)i=i-1}Op(e[n],r)===0?v.swap(e,n,i):(i=i+1,v.swap(e,i,a)),i<=t&&(n=i+1),t<=i&&(a=i-1)}}function jC(e,t,n,a,r){let s=t[t.length-1],[i,o]=[e.length/s,s],l=v.getTypedArrayFromDType(n,i*a),u=v.getTypedArrayFromDType("int32",i*a);for(let d=0;d<i;d++){let c=d*o,h=e.subarray(c,c+o),m=new Array(h.length);h.forEach((b,x)=>m[x]={value:b,index:x}),a<m.length&&(HC(m,a),m=m.slice(0,a)),r&&m.sort(Op);let f=d*a,g=l.subarray(f,f+a),y=u.subarray(f,f+a);for(let b=0;b<a;b++)g[b]=m[b].value,y[b]=m[b].index}let p=t.slice();return p[p.length-1]=a,[Oe(p,n,l),Oe(p,"int32",u)]}function qC(e,t,n,a){let r=v.parseAxisParam(t,n)[0],s=[1,n[0],1];for(let m=0;m<r;m++)s[0]*=n[m];s[1]=n[r];for(let m=r+1;m<n.length;m++)s[2]*=n[m];let i={},o=new Int32Array(n[r]),l=new Ht(s,a,e),u=[],p=s[0]===1&&s[2]===1;for(let m=0;m<n[r];m++){let f;if(p)f=e[m].toString();else{let g=[];for(let y=0;y<s[0];y++)for(let b=0;b<s[2];b++)g.push(l.get(y,m,b));f=g.join(",")}if(i[f]!==void 0)o[m]=i[f];else{let g=Object.keys(i).length;i[f]=g,o[m]=g,u.push(m)}}let d=s.slice();d[1]=Object.keys(i).length;let c=new Ht(d,a);u.forEach((m,f)=>{for(let g=0;g<s[0];g++)for(let y=0;y<s[2];y++)c.set(l.get(g,m,y),g,f,y)});let h=n.slice();return h[r]=d[1],{outputValues:c.values,outputShape:h,indices:o}}Gm("cpu",()=>new R0,1);var KC=rt(_i,e=>e>=0?e:Math.exp(e)-1),Aq={kernelName:_i,backendName:"cpu",kernelFunc:KC};function XC(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{alpha:s}=a;ge([r],"leakyRelu");let i=v.sizeFromShape(r.shape),o=n.data.get(r.dataId).values,l=v.getTypedArrayFromDType("float32",i);for(let u=0;u<o.length;u++)l[u]=o[u]<0?s*o[u]:o[u];return n.makeTensorInfo(r.shape,"float32",l)}var $q={kernelName:Mi,backendName:"cpu",kernelFunc:XC},Fq=Vt((e,t)=>e<0?t*e:e);function YC(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t;ge([a,r],"prelu");let s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,[o,l]=Fq(a.shape,r.shape,s,i,"float32");return n.makeTensorInfo(l,"float32",o)}var Dq={kernelName:Ki,backendName:"cpu",kernelFunc:YC},ZC=rt(Yi,e=>Math.max(0,e)),Rq={kernelName:Yi,backendName:"cpu",kernelFunc:ZC},JC=rt(Qi,e=>Math.min(Math.max(0,e),6)),Mq={kernelName:Qi,backendName:"cpu",kernelFunc:JC};function Qh(e,t,n,a,r){if(n==="linear")return pr({inputs:{x:t},backend:e});if(n==="relu")return ZC({inputs:{x:t},backend:e});if(n==="elu")return KC({inputs:{x:t},backend:e});if(n==="relu6")return JC({inputs:{x:t},backend:e});if(n==="prelu")return YC({inputs:{x:t,alpha:a},backend:e});if(n==="leakyrelu")return XC({inputs:{x:t},backend:e,attrs:{alpha:r}});if(n==="sigmoid")return LC({inputs:{x:t},backend:e});throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)}function ft(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{shape:s}=a,i=v.sizeFromShape(r.shape),o=v.inferFromImplicitShape(s,i),l=v.sizeFromShape(o);v.assert(i===l,()=>`The new shape (${o}) has ${l} elements and the old shape (${r.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),n.incRef(r.dataId);let u=n.data.get(r.dataId);if(u.complexTensorInfos!=null){let p=u.complexTensorInfos.real,d=u.complexTensorInfos.imag;p.shape=o,d.shape=o}return{dataId:r.dataId,shape:o,dtype:r.dtype}}var Pq={kernelName:vu,backendName:"cpu",kernelFunc:ft};function QC(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;ge([r,s],"matMul");let l=r.shape.length,u=s.shape.length,p=i?r.shape[l-2]:r.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],c=i?r.shape[l-1]:r.shape[l-2],h=o?s.shape[u-2]:s.shape[u-1],m=r.shape.slice(0,-2),f=s.shape.slice(0,-2),g=v.sizeFromShape(m),y=v.sizeFromShape(f),b=Ou.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)).concat([c,h]);v.assert(p===d,()=>`Error in matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${r.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,p,c]:[g,c,p],w=o?[y,h,d]:[y,d,h],I=ft({inputs:{x:r},backend:n,attrs:{shape:x}}),T=ft({inputs:{x:s},backend:n,attrs:{shape:w}}),C=i?I.shape[1]:I.shape[2],E=i?I.shape[2]:I.shape[1],A=o?T.shape[1]:T.shape[2],R=Math.max(g,y),F=n.data.get(I.dataId).values,S=n.data.get(T.dataId).values,M=v.computeStrides(I.shape),B=v.computeStrides(T.shape),[U,G,q]=i?[M[0],1,M[1]]:[M[0],M[1],1],[K,Z,Q]=o?[1,B[1],B[0]]:[B[1],1,B[0]],ee=E*A,ae=Oe([R,E,A],I.dtype),te=ae.values,le=n.blockSize;for(let ie=0;ie<R;ie++)for(let ye=0;ye<E;ye+=le)for(let ue=0;ue<A;ue+=le)for(let xe=0;xe<C;xe+=le){let Ie=Math.min(ye+le,E),Se=Math.min(ue+le,A),Le=Math.min(xe+le,C);for(let Ve=ye;Ve<Ie;Ve++)for(let nt=ue;nt<Se;nt++){let it=0;for(let et=xe;et<Le;et++){let at=Math.min(ie,g-1)*U,ze=Math.min(ie,y-1)*Q,dt=F[at+Ve*G+et*q],jn=S[et*K+nt*Z+ze];it+=dt*jn}te[ie*ee+(Ve*A+nt)]+=it}}return n.disposeIntermediateTensorInfo(I),n.disposeIntermediateTensorInfo(T),n.makeTensorInfo(b,ae.dtype,ae.values)}var Oq={kernelName:bi,backendName:"cpu",kernelFunc:QC};function Lq(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a,c,h,m,f=[];c=QC({inputs:{a:r,b:s},attrs:{transposeA:l,transposeB:u},backend:n}),i&&(h=_l({inputs:{a:c,b:i},backend:n}),f.push(c),c=h),p&&(m=Qh(n,c,p,o,d),f.push(c),c=m);for(let g of f)n.disposeIntermediateTensorInfo(g);return c}var zq={kernelName:Qs,backendName:"cpu",kernelFunc:Lq},Wq=rt(Rl,e=>Math.acos(e)),Bq={kernelName:Rl,backendName:"cpu",kernelFunc:Wq},Vq=rt(Ml,e=>Math.acosh(e)),Uq={kernelName:Ml,backendName:"cpu",kernelFunc:Vq};function Gq(e){let{inputs:t,backend:n}=e,a=t;ge(t,"addN");let r=a.map(o=>n.data.get(o.dataId).values),s=Oe(a[0].shape,a[0].dtype),i=s.values;for(let o=0;o<a.length;o++){let l=r[o];for(let u=0;u<i.length;u++)i[u]+=l[u]}return n.makeTensorInfo(s.shape,s.dtype,s.values)}var Hq={kernelName:fi,backendName:"cpu",kernelFunc:Gq};function jq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"all");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("all",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];x=x&&I}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var qq={kernelName:Pl,backendName:"cpu",kernelFunc:jq};function Kq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"any");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("any",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];x=x||I}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var Xq={kernelName:Ol,backendName:"cpu",kernelFunc:Kq};function Yq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a;ge(r,"argMax");let i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Vn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],N.assertAxesAreInnerMostDims("argMax",i,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,i),c=v.sizeFromShape(p),h=v.makeZerosTypedArray(c,"int32"),m=v.sizeFromShape(d),f=n.data.get(l.dataId).values;for(let g=0;g<h.length;++g){let y=g*m,b=f[y],x=0;for(let w=0;w<m;++w){let I=f[y+w];I>b&&(b=I,x=w)}h[g]=x}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(p,"int32",h)}var Zq={kernelName:gi,backendName:"cpu",kernelFunc:Yq};function Jq(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a;ge(r,"argMin");let i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Vn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],N.assertAxesAreInnerMostDims("argMin",i,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,i),c=v.sizeFromShape(p),h=v.makeZerosTypedArray(c,"int32"),m=v.sizeFromShape(d),f=n.data.get(l.dataId).values;for(let g=0;g<h.length;++g){let y=g*m,b=f[y],x=0;for(let w=0;w<m;++w){let I=f[y+w];I<b&&(b=I,x=w)}h[g]=x}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(p,"int32",h)}var Qq={kernelName:cc,backendName:"cpu",kernelFunc:Jq},e5=rt(Ll,e=>Math.asin(e)),t5={kernelName:Ll,backendName:"cpu",kernelFunc:e5},n5=rt(zl,e=>Math.asinh(e)),a5={kernelName:zl,backendName:"cpu",kernelFunc:n5},r5=rt(Wl,e=>Math.atan(e)),s5={kernelName:Wl,backendName:"cpu",kernelFunc:r5},i5=Vt((e,t)=>Math.atan2(e,t)),o5=nn(Vl,i5),l5={kernelName:Vl,backendName:"cpu",kernelFunc:o5},u5=rt(Bl,e=>Math.atanh(e)),p5={kernelName:Bl,backendName:"cpu",kernelFunc:u5};function j0(e,t,n,a,r,s){let i=r.strideHeight,o=r.strideWidth,l=r.dilationHeight,u=r.dilationWidth,p=r.effectiveFilterHeight,d=r.effectiveFilterWidth,c=r.padInfo.top,h=r.padInfo.left,m=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,f=Oe(r.outShape,n),g=f.values,y=r.outShape[1]*r.outShape[2]*r.outShape[3],b=r.outShape[2]*r.outShape[3],x=r.outShape[3];for(let w=0;w<r.batchSize;++w){let I=w*y,T=w*a[0];for(let C=0;C<r.inChannels;++C)for(let E=0;E<r.outHeight;++E){let A=E*i-c,R=Math.max(0,A),F=Math.min(r.inHeight,p+A),S=I+E*b;for(let M=0;M<r.outWidth;++M){let B=M*o-h,U=Math.max(0,B),G=Math.min(r.inWidth,d+B),q=m,K=0,Z=0;for(let ee=R;ee<F;ee+=l){let ae=T+ee*a[1];for(let te=U;te<G;te+=u){let le=ae+te*a[2],ie=e[le+C];s==="max"&&ie>q?q=ie:s==="avg"&&(K+=ie,Z++)}if(isNaN(q))break}let Q=S+M*x+C;g[Q]=s==="avg"?K/Z:q}}}return f}function e_(e,t,n,a,r=!1,s=!1){let i=Oe(a.outShape,"int32"),o=a.strideHeight,l=a.strideWidth,u=a.dilationHeight,p=a.dilationWidth,d=a.effectiveFilterHeight,c=a.effectiveFilterWidth,h=a.padInfo.top,m=a.padInfo.left,f=Oe(t,n,e);for(let g=0;g<a.batchSize;++g)for(let y=0;y<a.inChannels;++y)for(let b=0;b<a.outHeight;++b){let x=b*o-h,w=x;for(;w<0;)w+=u;let I=Math.min(a.inHeight,d+x);for(let T=0;T<a.outWidth;++T){let C=T*l-m,E=C;for(;E<0;)E+=p;let A=Math.min(a.inWidth,c+C),R=Number.NEGATIVE_INFINITY,F=-1;for(let S=w;S<I;S+=u){let M=S-x;for(let B=E;B<A;B+=p){let U=B-C,G=f.get(g,S,B,y);G>R&&(R=G,r?F=s?((g*a.inHeight+S)*a.inWidth+B)*a.inChannels+y:(S*a.inWidth+B)*a.inChannels+y:F=M*c+U)}}i.set(F,g,b,T,y)}}return i}function t_(e,t,n,a,r,s){let i=r.strideDepth,o=r.strideHeight,l=r.strideWidth,u=r.dilationDepth,p=r.dilationHeight,d=r.dilationWidth,c=r.effectiveFilterDepth,h=r.effectiveFilterHeight,m=r.effectiveFilterWidth,f=r.padInfo.front,g=r.padInfo.top,y=r.padInfo.left,b=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,x=Oe(r.outShape,n),w=x.values,I=r.outShape[1]*r.outShape[2]*r.outShape[3]*r.outShape[4],T=r.outShape[2]*r.outShape[3]*r.outShape[4],C=r.outShape[3]*r.outShape[4],E=r.outShape[4];for(let A=0;A<r.batchSize;++A){let R=A*I,F=A*a[0];for(let S=0;S<r.inChannels;++S)for(let M=0;M<r.outDepth;++M){let B=M*i-f,U=B;for(;U<0;)U+=u;let G=Math.min(r.inDepth,c+B),q=R+M*T;for(let K=0;K<r.outHeight;++K){let Z=K*o-g,Q=Z;for(;Q<0;)Q+=p;let ee=Math.min(r.inHeight,h+Z),ae=q+K*C;for(let te=0;te<r.outWidth;++te){let le=te*l-y,ie=le;for(;ie<0;)ie+=d;let ye=Math.min(r.inWidth,m+le),ue=ae+te*E,xe=b,Ie=0,Se=0;for(let Ve=U;Ve<G;Ve+=u){let nt=F+Ve*a[1];for(let it=Q;it<ee;it+=p){let et=nt+it*a[2];for(let at=ie;at<ye;at+=d){let ze=et+at*a[3],dt=e[ze+S];if(s==="max"&&dt>xe?xe=dt:s==="avg"&&(Ie+=dt,Se++),isNaN(xe))break}if(isNaN(xe))break}if(isNaN(xe))break}let Le=ue+S;w[Le]=s==="avg"?Ie/Se:xe}}}}return x}function c5(e,t){let n=Oe(t.outShape,"int32"),a=t.strideDepth,r=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,o=t.dilationHeight,l=t.dilationWidth,u=t.effectiveFilterDepth,p=t.effectiveFilterHeight,d=t.effectiveFilterWidth,c=t.padInfo.front,h=t.padInfo.top,m=t.padInfo.left;for(let f=0;f<t.batchSize;++f)for(let g=0;g<t.inChannels;++g)for(let y=0;y<t.outDepth;++y){let b=y*a-c,x=b;for(;x<0;)x+=i;let w=Math.min(t.inDepth,u+b);for(let I=0;I<t.outHeight;++I){let T=I*r-h,C=T;for(;C<0;)C+=o;let E=Math.min(t.inHeight,p+T);for(let A=0;A<t.outWidth;++A){let R=A*s-m,F=R;for(;F<0;)F+=l;let S=Math.min(t.inWidth,d+R),M=Number.NEGATIVE_INFINITY,B=-1;for(let U=x;U<w;U+=i){let G=U-b;for(let q=C;q<E;q+=o){let K=q-T;for(let Z=F;Z<S;Z+=l){let Q=Z-R,ee=e.get(f,U,q,Z,g);ee>=M&&(M=ee,B=G*p*d+K*p+Q)}}}n.set(B,f,y,I,A,g)}}}return n}function d5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;ge(r,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l),d;if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))d=pr({inputs:{x:r},backend:n});else{let c=n.data.get(r.dataId).values,h=v.computeStrides(r.shape),m=j0(c,r.shape,r.dtype,h,p,"avg");d=n.makeTensorInfo(p.outShape,r.dtype,m.values)}return d}var h5={kernelName:yi,backendName:"cpu",kernelFunc:d5};function m5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a;ge(r,"avgPool3d");let p=N.computePool3DInfo(r.shape,s,i,1,o,l,u),d=n.data.get(r.dataId).values,c=t_(d,r.shape,r.dtype,v.computeStrides(r.shape),p,"avg");return n.makeTensorInfo(c.shape,"float32",c.values)}var f5={kernelName:dc,backendName:"cpu",kernelFunc:m5};function g5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=a;ge([r,s],"avgPool3DGrad");let p=N.computePool3DInfo(s.shape,i,o,1,l,u),d=p.strideDepth,c=p.strideHeight,h=p.strideWidth,m=p.filterDepth,f=p.filterHeight,g=p.filterWidth,y=p.dilationDepth,b=p.dilationHeight,x=p.dilationWidth,w=p.effectiveFilterDepth,I=p.effectiveFilterHeight,T=p.effectiveFilterWidth,C=w-1-p.padInfo.front,E=T-1-p.padInfo.left,A=I-1-p.padInfo.top,R=Oe(s.shape,"float32"),F=1/(m*f*g),S=n.bufferSync(r);for(let M=0;M<p.batchSize;++M)for(let B=0;B<p.inChannels;++B)for(let U=0;U<p.inDepth;++U)for(let G=0;G<p.inHeight;++G)for(let q=0;q<p.inWidth;++q){let K=U-C,Z=G-A,Q=q-E,ee=0;for(let ae=0;ae<w;ae+=y){let te=(K+ae)/d;if(!(te<0||te>=p.outDepth||Math.floor(te)!==te))for(let le=0;le<I;le+=b){let ie=(Z+le)/c;if(!(ie<0||ie>=p.outHeight||Math.floor(ie)!==ie))for(let ye=0;ye<T;ye+=x){let ue=(Q+ye)/h;ue<0||ue>=p.outWidth||Math.floor(ue)!==ue||(ee+=S.get(M,te,ie,ue,B))}}}R.set(ee*F,M,U,G,q,B)}return n.makeTensorInfo(R.shape,R.dtype,R.values)}var y5={kernelName:cm,backendName:"cpu",kernelFunc:g5};function b5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s;ge([r,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=a,p=N.computePool2DInfo(i.shape,o,l,1,u),d=p.strideHeight,c=p.strideWidth,h=p.filterHeight,m=p.filterWidth,f=p.dilationHeight,g=p.dilationWidth,y=p.effectiveFilterHeight,b=p.effectiveFilterWidth,x=b-1-p.padInfo.left,w=y-1-p.padInfo.top,I=Oe(i.shape,"float32"),T=1/(h*m),C=n.data.get(r.dataId).values,E=Oe(r.shape,"float32",C);for(let A=0;A<p.batchSize;++A)for(let R=0;R<p.inChannels;++R)for(let F=0;F<p.inHeight;++F)for(let S=0;S<p.inWidth;++S){let M=F-w,B=S-x,U=0;for(let G=0;G<y;G+=f){let q=(M+G)/d;if(!(q<0||q>=p.outHeight||Math.floor(q)!==q))for(let K=0;K<b;K+=g){let Z=(B+K)/c;Z<0||Z>=p.outWidth||Math.floor(Z)!==Z||(U+=E.get(A,q,Z,R))}}I.set(U*T,A,F,S,R)}return n.makeTensorInfo(I.shape,I.dtype,I.values)}var x5={kernelName:pm,backendName:"cpu",kernelFunc:b5};function v5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,scale:s,offset:i,mean:o,variance:l}=t;v.assert(o.shape.length===l.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(s==null||o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),ge([r,o,l,s,i],"batchNorm");let{varianceEpsilon:u}=a;u==null&&(u=.001);let p=n.data.get(r.dataId).values,d=n.data.get(o.dataId).values,c=n.data.get(l.dataId).values,h=s?n.data.get(s.dataId).values:new Float32Array([1]),m=i?n.data.get(i.dataId).values:new Float32Array([0]),f=new Float32Array(p.length),g=m.length,y=h.length,b=c.length,x=d.length,w=0,I=0,T=0,C=0;for(let E=0;E<p.length;++E)f[E]=m[w++]+(p[E]-d[I++])*h[T++]/Math.sqrt(c[C++]+u),w>=g&&(w=0),I>=x&&(I=0),T>=y&&(T=0),C>=b&&(C=0);return n.makeTensorInfo(r.shape,r.dtype,f)}var w5={kernelName:Fi,backendName:"cpu",kernelFunc:v5};function k5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a;ge([r],"batchToSpaceND");let o=s.reduce((y,b)=>y*b),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=ft({inputs:{x:r},backend:n,attrs:{shape:l}}),m=Vn({inputs:{x:h},backend:n,attrs:{perm:u}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:p}}),g=pi({inputs:{x:f},backend:n,attrs:{begin:d,size:c}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),g}var I5={kernelName:Ul,backendName:"cpu",kernelFunc:k5};function S5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i}=a,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,u=P0(o,l,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,u)}var T5={kernelName:dm,backendName:"cpu",kernelFunc:S5};function N5(e){let{inputs:t,backend:n}=e,{s0:a,s1:r}=t,s=n.data.get(a.dataId).values,i=n.data.get(r.dataId).values,o=N.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var C5={kernelName:hm,backendName:"cpu",kernelFunc:N5},_5=rt(ys,(e,t)=>{let n=t;return e>n.clipValueMax?n.clipValueMax:e<n.clipValueMin?n.clipValueMin:e}),E5={kernelName:ys,backendName:"cpu",kernelFunc:_5},A5=e=>{let{x:t}=e.inputs,n=e.backend,a=new Float32Array(v.sizeFromShape(t.shape)),r=n.data.get(t.dataId),s=r.complexTensorInfos.real,i=r.complexTensorInfos.imag,o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values;for(let u=0;u<o.length;u++){let p=o[u],d=l[u];a[u]=Math.hypot(p,d)}return n.makeOutput(a,t.shape,"float32")},$5={kernelName:hc,backendName:"cpu",kernelFunc:A5};function El(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.data.get(a.dataId).complexTensorInfos.imag,s=n.data.get(r.dataId).values;return n.makeTensorInfo(r.shape,r.dtype,s)}var F5={kernelName:Nm,backendName:"cpu",kernelFunc:El};function Al(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a,s=v.parseAxisParam(r,t[0].shape)[0],i=t.map(f=>f.shape);N.assertParamsConsistent(i,s);let o=N.computeOutShape(t.map(f=>f.shape),s);if(v.sizeFromShape(o)===0)return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(f=>v.sizeFromShape(f.shape)>0);if(l.length===1)return pr({inputs:{x:l[0]},backend:n});if(l[0].dtype==="complex64"){let f=l.map(w=>ui({inputs:{input:w},backend:n})),g=l.map(w=>El({inputs:{input:w},backend:n})),y=Al({inputs:f,backend:n,attrs:{axis:s}}),b=Al({inputs:g,backend:n,attrs:{axis:s}}),x=Zn({inputs:{real:y,imag:b},backend:n});return f.forEach(w=>n.disposeIntermediateTensorInfo(w)),g.forEach(w=>n.disposeIntermediateTensorInfo(w)),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(b),x}let u=l.map(f=>{let g=v.sizeFromShape(f.shape.slice(s));return ft({inputs:{x:f},backend:n,attrs:{shape:[-1,g]}})}),p=u.map(f=>({vals:n.data.get(f.dataId).values,shape:f.shape}));o=N.computeOutShape(u.map(f=>f.shape),1);let d=u[0].shape[0]===1,c=O0(p,o,t[0].dtype,d),h=N.computeOutShape(l.map(f=>f.shape),s),m=n.makeTensorInfo(h,t[0].dtype,c);return u.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var D5={kernelName:Gl,backendName:"cpu",kernelFunc:Al};function n_(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:p}=a;ge([r,s],"conv2d");let d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,s.shape,i,u,o,p,!1,d),h=c.filterHeight,m=c.filterWidth,f=c.dilationHeight,g=c.dilationWidth,y=c.padInfo.left,b=c.padInfo.top,x=c.dataFormat==="channelsLast",w=new Ht(c.outShape,r.dtype),I=v.computeStrides(r.shape),T=v.computeStrides(s.shape),C=I[0],E=x?I[1]:I[2],A=x?I[2]:1,R=x?1:I[1],F=w.strides[0],S=x?w.strides[1]:w.strides[2],M=x?w.strides[2]:1,B=x?1:w.strides[1],U=n.data.get(r.dataId).values,G=n.data.get(s.dataId).values,q=w.values;for(let K=0;K<c.batchSize;++K){let Z=K*C,Q=K*F;for(let ee=0;ee<c.outHeight;++ee){let ae=Q+ee*S,te=ee*c.strideHeight-b;for(let le=0;le<h;++le){let ie=te+le*f;if(ie<0||ie>=c.inHeight)continue;let ye=le*T[0],ue=Z+ie*E;for(let xe=0;xe<c.outWidth;++xe){let Ie=ae+xe*M,Se=xe*c.strideWidth-y;for(let Le=0;Le<m;++Le){let Ve=Se+Le*g;if(Ve<0||Ve>=c.inWidth)continue;let nt=ye+Le*T[1],it=ue+Ve*A,et=nt;for(let at=0;at<c.inChannels;++at){let ze=U[it+at*R];for(let dt=0;dt<c.outChannels;++dt)q[Ie+dt*B]+=ze*G[et+dt];et+=c.outChannels}}}}}}return n.makeTensorInfo(w.shape,w.dtype,q)}var R5={kernelName:wi,backendName:"cpu",kernelFunc:n_};function M5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:p}=a;ge([r,s],"conv2dBackpropFilter");let d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,p,i,1,o,u,!1,d),{strideHeight:h,strideWidth:m,filterHeight:f,filterWidth:g}=c,y=c.dataFormat==="channelsLast",b=new Ht(c.filterShape,"float32"),x=c.padInfo.left,w=c.padInfo.top,I=n.data.get(r.dataId).values,T=n.data.get(s.dataId).values,C=new Ht(r.shape,r.dtype,I),E=new Ht(s.shape,s.dtype,T);for(let A=0;A<f;++A){let R=Math.max(0,Math.ceil((w-A)/h)),F=Math.min(c.outHeight,(c.inHeight+w-A)/h);for(let S=0;S<g;++S){let M=Math.max(0,Math.ceil((x-S)/m)),B=Math.min(c.outWidth,(c.inWidth+x-S)/m);for(let U=0;U<c.inChannels;++U)for(let G=0;G<c.outChannels;++G){let q=0;for(let K=0;K<c.batchSize;++K)for(let Z=R;Z<F;++Z){let Q=A+Z*h-w;for(let ee=M;ee<B;++ee){let ae=S+ee*m-x;y?q+=C.get(K,Q,ae,U)*E.get(K,Z,ee,G):q+=C.get(K,U,Q,ae)*E.get(K,G,Z,ee)}}b.set(q,A,S,U,G)}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}var P5={kernelName:fm,backendName:"cpu",kernelFunc:M5};function O5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:p}=a;ge([r,s],"conv2dBackpropInput");let d=v.computeStrides(s.shape),c=v.computeStrides(r.shape),h=N.convertConv2DDataFormat(u),m=N.computeConv2DInfo(i,s.shape,o,1,l,p,!1,h),f=new Ht(m.inShape,"float32"),g=f.values,y=n.data.get(r.dataId).values,b=n.data.get(s.dataId).values,[x,w,I]=d,{batchSize:T,filterHeight:C,filterWidth:E,inChannels:A,inHeight:R,inWidth:F,outChannels:S,outHeight:M,outWidth:B,strideHeight:U,strideWidth:G}=m;h=m.dataFormat;let q=C-1-m.padInfo.top,K=E-1-m.padInfo.left,Z=h==="channelsLast",Q=f.strides[0],ee=Z?f.strides[1]:f.strides[2],ae=Z?f.strides[2]:1,te=Z?1:f.strides[1],le=c[0],ie=Z?c[1]:c[2],ye=Z?c[2]:1,ue=Z?1:c[1];for(let xe=0;xe<T;++xe)for(let Ie=0;Ie<A;++Ie)for(let Se=0;Se<R;++Se){let Le=Se-q,Ve=Math.max(0,Math.ceil(Le/U)),nt=Math.min(M,(C+Le)/U);for(let it=0;it<F;++it){let et=it-K,at=Math.max(0,Math.ceil(et/G)),ze=Math.min(B,(E+et)/G),dt=0;for(let Mt=Ve;Mt<nt;++Mt){let ia=Mt*U-Le;for(let on=at;on<ze;++on){let Fn=on*G-et,oa=le*xe+ie*Mt+ye*on,Dn=x*(C-1-ia)+w*(E-1-Fn)+I*Ie;for(let ot=0;ot<S;++ot){let Rn=y[oa+ue*ot],qn=b[Dn+ot];dt+=Rn*qn}}}let jn=Q*xe+ee*Se+ae*it+te*Ie;g[jn]=dt}}return n.makeTensorInfo(f.shape,f.dtype,f.values)}var L5={kernelName:ki,backendName:"cpu",kernelFunc:O5};function z5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a;ge([r,s],"conv3d");let u=N.computeConv3DInfo(r.shape,s.shape,i,l,o),{filterDepth:p,filterHeight:d,filterWidth:c,dilationDepth:h,dilationHeight:m,dilationWidth:f,padInfo:g}=u,y=g.front,b=g.left,x=g.top,w=new Ht(u.outShape,r.dtype),I=n.data.get(r.dataId).values,T=n.data.get(s.dataId).values,C=w.values,E=v.computeStrides(r.shape),A=v.computeStrides(s.shape);for(let R=0;R<u.batchSize;++R){let F=R*E[0],S=R*w.strides[0];for(let M=0;M<u.outDepth;++M){let B=S+M*w.strides[1],U=M*u.strideDepth-y;for(let G=0;G<p;++G){let q=U+G*h;if(q<0||q>=u.inDepth)continue;let K=G*A[0],Z=F+q*E[1];for(let Q=0;Q<u.outHeight;++Q){let ee=B+Q*w.strides[2],ae=Q*u.strideHeight-x;for(let te=0;te<d;++te){let le=ae+te*m;if(le<0||le>=u.inHeight)continue;let ie=K+te*A[1],ye=Z+le*E[2];for(let ue=0;ue<u.outWidth;++ue){let xe=ee+ue*u.outChannels,Ie=ue*u.strideWidth-b;for(let Se=0;Se<c;++Se){let Le=Ie+Se*f;if(Le<0||Le>=u.inWidth)continue;let Ve=ie+Se*A[2],nt=ye+Le*u.inChannels,it=Ve;for(let et=0;et<u.inChannels;++et){let at=I[nt+et];for(let ze=0;ze<u.outChannels;++ze)C[xe+ze]+=at*T[it+ze];it+=u.outChannels}}}}}}}}return n.makeTensorInfo(w.shape,w.dtype,w.values)}var W5={kernelName:mc,backendName:"cpu",kernelFunc:z5};function B5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,filterShape:l}=a;ge([r,s],"conv3dBackpropFilterV2");let u=v.computeStrides(r.shape),p=v.computeStrides(s.shape),d=N.computeConv3DInfo(r.shape,l,i,1,o),c=d.strideDepth,h=d.strideHeight,m=d.strideWidth,f=d.filterDepth,g=d.filterHeight,y=d.filterWidth,b=new Ht(d.filterShape,"float32"),x=b.values,[w,I,T,C]=b.strides,E=n.data.get(s.dataId).values,[A,R,F,S]=p,M=n.data.get(r.dataId).values,[B,U,G,q]=u,K=d.padInfo.front,Z=d.padInfo.left,Q=d.padInfo.top;for(let ee=0;ee<f;++ee){let ae=Math.max(0,Math.ceil((K-ee)/c)),te=Math.min(d.outDepth,(d.inDepth+K-ee)/c),le=ee*w;for(let ie=0;ie<g;++ie){let ye=Math.max(0,Math.ceil((Q-ie)/h)),ue=Math.min(d.outHeight,(d.inHeight+Q-ie)/h),xe=ie*I+le;for(let Ie=0;Ie<y;++Ie){let Se=Math.max(0,Math.ceil((Z-Ie)/m)),Le=Math.min(d.outWidth,(d.inWidth+Z-Ie)/m),Ve=Ie*T+xe;for(let nt=0;nt<d.inChannels;++nt){let it=nt*C+Ve;for(let et=0;et<d.outChannels;++et){let at=0;for(let ze=0;ze<d.batchSize;++ze){let dt=ze*B,jn=ze*A;for(let Mt=ae;Mt<te;++Mt){let ia=(ee+Mt*c-K)*U+dt,on=Mt*R+jn;for(let Fn=ye;Fn<ue;++Fn){let oa=(ie+Fn*h-Q)*G+ia,Dn=Fn*F+on;for(let ot=Se;ot<Le;++ot){let Rn=(Ie+ot*m-Z)*q+oa,qn=ot*S+Dn;at+=M[Rn+nt]*E[qn+et]}}}}x[it+et]=at}}}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}var V5={kernelName:gm,backendName:"cpu",kernelFunc:B5};function U5(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{pad:i,strides:o,inputShape:l}=a;ge([r],"conv3dBackpropInputV2");let u=v.computeStrides(r.shape),p=v.computeStrides(s.shape),d=N.computeConv3DInfo(l,s.shape,o,1,i),c=new Ht(d.inShape,"float32"),h=c.values,[m,f,g,y]=c.strides,b=n.data.get(r.dataId).values,[x,w,I,T]=u,C=n.data.get(s.dataId).values,[E,A,R,F]=p,{batchSize:S,filterDepth:M,filterHeight:B,filterWidth:U,inChannels:G,inDepth:q,inHeight:K,inWidth:Z,outChannels:Q,outDepth:ee,outHeight:ae,outWidth:te,strideDepth:le,strideHeight:ie,strideWidth:ye}=d,ue=M-1-d.padInfo.front,xe=B-1-d.padInfo.top,Ie=U-1-d.padInfo.left;for(let Se=0;Se<S;++Se)for(let Le=0;Le<G;++Le)for(let Ve=0;Ve<q;++Ve){let nt=Ve-ue,it=Math.max(0,Math.ceil(nt/le)),et=Math.min(ee,(M+nt)/le);for(let at=0;at<K;++at){let ze=at-xe,dt=Math.max(0,Math.ceil(ze/ie)),jn=Math.min(ae,(B+ze)/ie);for(let Mt=0;Mt<Z;++Mt){let ia=Mt-Ie,on=Math.max(0,Math.ceil(ia/ye)),Fn=Math.min(te,(U+ia)/ye),oa=0;for(let Dn=it;Dn<et;++Dn){let ot=Dn*le-nt;for(let Rn=dt;Rn<jn;++Rn){let qn=Rn*ie-ze;for(let xr=on;xr<Fn;++xr){let qo=xr*ye-ia,er=x*Se+w*Dn+I*Rn+T*xr,vp=E*(M-1-ot)+A*(B-1-qn)+R*(U-1-qo)+F*Le;for(let Ia=0;Ia<Q;++Ia){let Ko=b[er+Ia],Xt=C[vp+Ia];oa+=Ko*Xt}}}}h[m*Se+f*Ve+g*at+y*Mt+Le]=oa}}}return n.makeTensorInfo(c.shape,c.dtype,c.values)}var G5={kernelName:ym,backendName:"cpu",kernelFunc:U5},H5=rt(Ii,e=>Math.cos(e)),j5={kernelName:Ii,backendName:"cpu",kernelFunc:H5},q5=rt(Si,e=>Math.cosh(e)),K5={kernelName:Si,backendName:"cpu",kernelFunc:q5};function X5(e){let{inputs:t,backend:n,attrs:a}=e,{image:r,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=a,[p,d,c,h]=r.shape,m=s.shape[0],[f,g]=o,y=Oe([m,f,g,h],"float32"),b=n.data.get(s.dataId).values,x=n.data.get(i.dataId).values,w=n.data.get(r.dataId).values,I=v.computeStrides(r.shape),T=v.computeStrides(y.shape);for(let C=0;C<m;C++){let E=C*4,A=b[E],R=b[E+1],F=b[E+2],S=b[E+3],M=x[C];if(M>=p)continue;let B=f>1?(F-A)*(d-1)/(f-1):0,U=g>1?(S-R)*(c-1)/(g-1):0;for(let G=0;G<f;G++){let q=f>1?A*(d-1)+G*B:.5*(A+F)*(d-1);if(q<0||q>d-1){for(let K=0;K<g;K++)for(let Z=0;Z<h;Z++){let Q=Z+K*T[2]+G*T[1]+C*T[0];y.values[Q]=u}continue}if(l==="bilinear"){let K=Math.floor(q),Z=Math.ceil(q),Q=q-K;for(let ee=0;ee<g;ee++){let ae=g>1?R*(c-1)+ee*U:.5*(R+S)*(c-1);if(ae<0||ae>c-1){for(let ye=0;ye<h;ye++){let ue=ye+ee*T[2]+G*T[1]+C*T[0];y.values[ue]=u}continue}let te=Math.floor(ae),le=Math.ceil(ae),ie=ae-te;for(let ye=0;ye<h;ye++){let ue=ye+te*I[2]+K*I[1]+M*I[0],xe=w[ue];ue=ye+le*I[2]+K*I[1]+M*I[0];let Ie=w[ue];ue=ye+te*I[2]+Z*I[1]+M*I[0];let Se=w[ue];ue=ye+le*I[2]+Z*I[1]+M*I[0];let Le=w[ue],Ve=xe+(Ie-xe)*ie,nt=Se+(Le-Se)*ie;ue=ye+ee*T[2]+G*T[1]+C*T[0],y.values[ue]=Ve+(nt-Ve)*Q}}}else for(let K=0;K<g;++K){let Z=g>1?R*(c-1)+K*U:.5*(R+S)*(c-1);if(Z<0||Z>c-1){for(let ae=0;ae<h;ae++){let te=ae+K*T[2]+G*T[1]+C*T[0];y.values[te]=u}continue}let Q=Math.round(Z),ee=Math.round(q);for(let ae=0;ae<h;ae++){let te=ae+Q*I[2]+ee*I[1]+M*I[0],le=ae+K*T[2]+G*T[1]+C*T[0];y.values[le]=w[te]}}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}var Y5={kernelName:jl,backendName:"cpu",kernelFunc:X5};function Z5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;ge(r,"cumprod");let l=N.getAxesPermutation([s],r.shape.length),u=r;l!=null&&(u=Vn({inputs:{x:r},backend:n,attrs:{perm:l}}));let p=N.getInnerMostAxes(1,r.shape.length)[0];if(p!==u.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${p}`);let d=fa(u.dtype,"int32"),c=v.makeOnesTypedArray(v.sizeFromShape(u.shape),d),h=n.data.get(u.dataId).values,m=u.shape[u.shape.length-1],f=o?(y,b)=>y+m-b-1:(y,b)=>y+b;for(let y=0;y<h.length;y+=m)for(let b=0;b<m;b++){let x=f(y,b);if(b===0)c[x]=i?1:h[x];else{let w=f(y,b-1);c[x]=i?h[w]*c[w]:h[x]*c[w]}}let g=n.makeTensorInfo(u.shape,d,c);if(l!=null){let y=N.getUndoAxesPermutation(l),b=Vn({inputs:{x:g},backend:n,attrs:{perm:y}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),b}return g}var J5={kernelName:Hl,backendName:"cpu",kernelFunc:Z5};function Q5(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;ge(r,"cumsum");let l=N.getAxesPermutation([s],r.shape.length),u=r;l!=null&&(u=Vn({inputs:{x:r},backend:n,attrs:{perm:l}}));let p=N.getInnerMostAxes(1,r.shape.length)[0];if(p!==u.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${p}`);let d=fa(u.dtype,"int32"),c=v.makeZerosTypedArray(v.sizeFromShape(u.shape),d),h=n.data.get(u.dataId).values,m=u.shape[u.shape.length-1],f=o?(y,b)=>y+m-b-1:(y,b)=>y+b;for(let y=0;y<h.length;y+=m)for(let b=0;b<m;b++){let x=f(y,b);if(b===0)c[x]=i?0:h[x];else{let w=f(y,b-1);c[x]=i?h[w]+c[w]:h[x]+c[w]}}let g=n.makeTensorInfo(u.shape,d,c);if(l!=null){let y=N.getUndoAxesPermutation(l),b=Vn({inputs:{x:g},backend:n,attrs:{perm:y}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),b}return g}var e8={kernelName:Ti,backendName:"cpu",kernelFunc:Q5};function t8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i,binaryOutput:o}=a;if(r.shape.length===1){let l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=P0(l,u,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,p)}else if(r.shape.length===2){let l=n.bufferSync(r),u=n.bufferSync(s),p=dC(l,u,i,o);return n.makeTensorInfo(p.shape,s.dtype,p.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${r.shape.length}.`)}var n8={kernelName:bm,backendName:"cpu",kernelFunc:t8};function a8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockSize:s,dataFormat:i}=a;v.assert(i==="NHWC",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let o=r.shape[0],l=r.shape[1],u=r.shape[2],p=r.shape[3],d=l*s,c=u*s,h=p/(s*s),m=n.data.get(r.dataId).values,f=new Float32Array(o*d*c*h),g=0;for(let y=0;y<o;++y)for(let b=0;b<d;++b){let x=Math.floor(b/s),w=b%s;for(let I=0;I<c;++I){let T=Math.floor(I/s),C=I%s,E=(w*s+C)*h;for(let A=0;A<h;++A){let R=A+E+p*(T+u*(x+l*y));f[g++]=m[R]}}}return n.makeTensorInfo([o,d,c,h],r.dtype,f)}var r8={kernelName:ql,backendName:"cpu",kernelFunc:a8};function a_(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=a;ge([r,s],"depthwiseConv2DNative");let p=v.computeStrides(r.shape),d=v.computeStrides(s.shape),c=l;c==null&&(c=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(i,c),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);let h=N.computeConv2DInfo(r.shape,s.shape,i,c,o,u,!0),{filterHeight:m,filterWidth:f,dilationHeight:g,dilationWidth:y,padInfo:b}=h,x=b.left,w=b.top,I=h.outChannels/h.inChannels,T=new Ht(h.outShape,r.dtype),C=n.data.get(r.dataId).values,E=n.data.get(s.dataId).values,A=T.values;for(let R=0;R<h.batchSize;++R){let F=R*p[0],S=R*T.strides[0];for(let M=0;M<h.outHeight;++M){let B=S+M*T.strides[1],U=M*h.strideHeight-w;for(let G=0;G<m;++G){let q=U+G*g;if(q<0||q>=h.inHeight)continue;let K=G*d[0],Z=F+q*p[1];for(let Q=0;Q<h.outWidth;++Q){let ee=B+Q*T.strides[2],ae=Q*h.strideWidth-x;for(let te=0;te<f;++te){let le=ae+te*y;if(le<0||le>=h.inWidth)continue;let ie=K+te*d[1],ye=Z+le*h.inChannels,ue=ee,xe=ie;for(let Ie=0;Ie<h.inChannels;++Ie){let Se=C[ye+Ie];for(let Le=0;Le<I;++Le)A[ue+Le]+=Se*E[xe+Le];ue+=I,xe+=I}}}}}}return n.makeTensorInfo(T.shape,T.dtype,T.values)}var s8={kernelName:Ni,backendName:"cpu",kernelFunc:a_};function i8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:p}=a;ge([r,s],"depthwiseConv2dNativeBackpropFilter");let d=N.computeConv2DInfo(r.shape,p,i,o,l,u,!0),{strideHeight:c,strideWidth:h,filterHeight:m,filterWidth:f}=d,g=new Ht(d.filterShape,"float32"),y=d.padInfo.left,b=d.padInfo.top,x=d.outChannels/d.inChannels,w=n.data.get(r.dataId).values,I=new Ht(r.shape,r.dtype,w),T=n.data.get(s.dataId).values,C=new Ht(s.shape,s.dtype,T);for(let E=0;E<m;++E){let A=Math.max(0,Math.ceil((b-E)/c)),R=Math.min(d.outHeight,(d.inHeight+b-E)/c);for(let F=0;F<f;++F){let S=Math.max(0,Math.ceil((y-F)/h)),M=Math.min(d.outWidth,(d.inWidth+y-F)/h);for(let B=0;B<d.outChannels;++B){let U=Math.trunc(B/x),G=B%x,q=0;for(let K=0;K<d.batchSize;++K)for(let Z=A;Z<R;++Z){let Q=E+Z*c-b;for(let ee=S;ee<M;++ee){let ae=F+ee*h-y;q+=I.get(K,Q,ae,U)*C.get(K,Z,ee,B)}}g.set(q,E,F,U,G)}}}return n.makeTensorInfo(g.shape,g.dtype,g.values)}var o8={kernelName:xm,backendName:"cpu",kernelFunc:i8};function l8(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:p}=a;ge([r,s],"depthwiseConv2DNativeBackpropInput");let d=v.computeStrides(r.shape),c=v.computeStrides(s.shape),h=N.computeConv2DInfo(p,s.shape,i,o,l,u,!0),m=new Ht(h.inShape,"float32"),f=m.values,[g,y,b]=m.strides,x=n.data.get(r.dataId).values,[w,I,T]=d,C=n.data.get(s.dataId).values,[E,A,R]=c,{batchSize:F,filterHeight:S,filterWidth:M,inChannels:B,inHeight:U,inWidth:G,outChannels:q,outHeight:K,outWidth:Z,strideHeight:Q,strideWidth:ee}=h,ae=S-1-h.padInfo.top,te=M-1-h.padInfo.left,le=q/B;for(let ie=0;ie<F;++ie)for(let ye=0;ye<B;++ye)for(let ue=0;ue<U;++ue){let xe=ue-ae,Ie=Math.max(0,Math.ceil(xe/Q)),Se=Math.min(K,(S+xe)/Q);for(let Le=0;Le<G;++Le){let Ve=Le-te,nt=Math.max(0,Math.ceil(Ve/ee)),it=Math.min(Z,(M+Ve)/ee),et=0;for(let at=Ie;at<Se;++at){let ze=at*Q-xe;for(let dt=nt;dt<it;++dt){let jn=dt*ee-Ve,Mt=w*ie+I*at+T*dt,ia=E*(S-1-ze)+A*(M-1-jn)+R*ye;for(let on=0;on<le;++on){let Fn=ye*le+on,oa=x[Mt+Fn],Dn=C[ia+on];et+=oa*Dn}}}f[g*ie+y*ue+b*Le+ye]=et}}return n.makeTensorInfo(m.shape,m.dtype,m.values)}var u8={kernelName:vm,backendName:"cpu",kernelFunc:l8};function p8(e){let{inputs:t,backend:n}=e,{x:a}=t,r=v.sizeFromShape(a.shape),s=n.data.get(a.dataId).values,i=Oe([r,r],a.dtype),o=i.values;for(let u=0;u<s.length;u++)o[u*r+u]=s[u];let l=[...a.shape,...a.shape];return n.makeTensorInfo(l,i.dtype,i.values)}var c8={kernelName:wm,backendName:"cpu",kernelFunc:p8},d8={kernelName:fc,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r}=e,{strides:s,pad:i,dilations:o}=n,l=t,u=l.data.get(a.dataId).values,p=a.shape.length,d=l.data.get(r.dataId).values,c=r.shape.length,{batchSize:h,inHeight:m,inWidth:f,inChannels:g,outHeight:y,outWidth:b,padInfo:x,strideHeight:w,strideWidth:I,filterHeight:T,filterWidth:C,dilationHeight:E,dilationWidth:A,outShape:R}=N.computeDilation2DInfo(a.shape,r.shape,s,i,"NHWC",o),F=v.sizeFromShape(R),S=R.length,M=v.getArrayFromDType(a.dtype,F);for(let B=0;B<h;++B)for(let U=0;U<y;++U){let G=U*w-x.top;for(let q=0;q<b;++q){let K=q*I-x.left;for(let Z=0;Z<g;++Z){let Q=Number.MIN_SAFE_INTEGER;for(let ae=0;ae<T;++ae){let te=G+ae*E;if(te>=0&&te<m)for(let le=0;le<C;++le){let ie=K+le*A;if(ie>=0&&ie<f){let ye=v.locToIndex([B,te,ie,Z],p,v.computeStrides(a.shape)),ue=v.locToIndex([ae,le,Z],c,v.computeStrides(r.shape)),xe=u[ye]+d[ue];xe>Q&&(Q=xe)}}}let ee=v.locToIndex([B,U,q,Z],S,v.computeStrides(R));M[ee]=Q}}}return{dataId:l.write(v.toTypedArray(M,a.dtype),R,a.dtype),shape:R,dtype:a.dtype}}},h8={kernelName:$h,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,p=v.toNestedArray(a.shape,u.data.get(a.dataId).values),d=v.toNestedArray(r.shape,u.data.get(r.dataId).values),{batchSize:c,inHeight:h,inWidth:m,inChannels:f,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:w,filterHeight:I,filterWidth:T,dilationHeight:C,dilationWidth:E,outShape:A}=N.computeDilation2DInfo(a.shape,r.shape,i,o,"NHWC",l);v.assert(s.rank===A.length,()=>`Error in ${$h}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);let R=v.toNestedArray(A,u.data.get(s.dataId).values),F=v.makeZerosNestedTypedArray(r.shape,r.dtype);for(let S=0;S<c;++S)for(let M=0;M<g;++M){let B=M*x-b.top;for(let U=0;U<y;++U){let G=U*w-b.left;for(let q=0;q<f;++q){let K=Number.MIN_SAFE_INTEGER,Z=0,Q=0;for(let ee=0;ee<I;++ee){let ae=B+ee*C;if(ae>=0&&ae<h)for(let te=0;te<T;++te){let le=G+te*E;if(le>=0&&le<m){let ie=p[S][ae][le][q]+d[ee][te][q];ie>K&&(K=ie,Z=ee,Q=te)}}}F[Z][Q][q]+=R[S][M][U][q]}}}return{dataId:u.write(v.toTypedArray(F,a.dtype),r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},m8={kernelName:Ah,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:a,filter:r,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,p=v.toNestedArray(a.shape,u.data.get(a.dataId).values),d=v.toNestedArray(r.shape,u.data.get(r.dataId).values),{batchSize:c,inHeight:h,inWidth:m,inChannels:f,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:w,filterHeight:I,filterWidth:T,dilationHeight:C,dilationWidth:E,outShape:A}=N.computeDilation2DInfo(a.shape,r.shape,i,o,"NHWC",l);v.assert(s.rank===A.length,()=>`Error in ${Ah}, dy must have the same rank as output ${A.length}, but got ${s.rank}`);let R=v.toNestedArray(A,u.data.get(s.dataId).values),F=v.makeZerosNestedTypedArray(a.shape,a.dtype);for(let S=0;S<c;++S)for(let M=0;M<g;++M){let B=M*x-b.top;for(let U=0;U<y;++U){let G=U*w-b.left;for(let q=0;q<f;++q){let K=Number.MIN_SAFE_INTEGER,Z=B<0?0:B,Q=G<0?0:G;for(let ee=0;ee<I;++ee){let ae=B+ee*C;if(ae>=0&&ae<h)for(let te=0;te<T;++te){let le=G+te*E;if(le>=0&&le<m){let ie=p[S][ae][le][q]+d[ee][te][q];ie>K&&(K=ie,Z=ae,Q=le)}}}F[S][Z][Q][q]+=R[S][M][U][q]}}}return{dataId:u.write(v.toTypedArray(F,a.dtype),a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}};function sd(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"sum");let o;r.dtype==="bool"?o=ds({inputs:{x:r},backend:n,attrs:{dtype:"int32"}}):o=pr({inputs:{x:r},backend:n});let l=o.shape.length,u=v.parseAxisParam(s,o.shape),p=N.getAxesPermutation(u,l),d=u,c=o;p!=null&&(c=Vn({inputs:{x:o},backend:n,attrs:{perm:p}}),d=N.getInnerMostAxes(d.length,l)),N.assertAxesAreInnerMostDims("sum",d,c.shape.length);let[h,m]=N.computeOutAndReduceShapes(c.shape,d),f=N.upcastType(c.dtype,"int32"),g=Zh(n,h,f),y=v.sizeFromShape(m),b=n.data.get(g.dataId).values,x=n.data.get(c.dataId).values;for(let w=0;w<b.length;++w){let I=w*y,T=0;for(let C=0;C<y;++C)T+=x[I+C];b[w]=T}if(i){let w=N.expandShapeToKeepDim(g.shape,u),I=g;g=ft({inputs:{x:g},backend:n,attrs:{shape:w}}),n.disposeIntermediateTensorInfo(I)}return n.disposeIntermediateTensorInfo(o),p!=null&&n.disposeIntermediateTensorInfo(c),g}var f8={kernelName:io,backendName:"cpu",kernelFunc:sd};function g8(e){let{inputs:t,backend:n,attrs:a}=e,{equation:r}=a,s=t,{allDims:i,summedDims:o,idDims:l}=N.decodeEinsumEquation(r,s.length);N.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:p}=N.getEinsumComputePath(o,l),d=p.length,c=null,h=i.length,m=[];for(let f=0;f<d;++f){for(let g of p[f]){let{permutationIndices:y,expandDims:b}=N.getEinsumPermutation(h,l[g]),x;N.isIdentityPermutation(y)?x=s[g]:(x=Vn({inputs:{x:s[g]},backend:n,attrs:{perm:y}}),m.push(x));let w=x.shape.slice();for(let I=0;I<b.length;++I)w.splice(b[I],0,1);v.arraysEqual(x.shape,w)||(x=ft({inputs:{x},backend:n,attrs:{shape:w}}),m.push(x)),c===null?c=x:(c=qf({inputs:{a:x,b:c},backend:n}),m.push(c))}f<d-1&&(u[f]>=0&&(c=sd({inputs:{x:c},backend:n,attrs:{axis:u[f]-(i.length-h),keepDims:!1}}),m.push(c)),h--)}for(let f of m)f!==c&&n.disposeIntermediateTensorInfo(f);return c}var y8={kernelName:km,backendName:"cpu",kernelFunc:g8};function b8(e){let{inputs:t,backend:n}=e,{dy:a,y:r}=t;ge([a,r],"eluGrad");let s=new Float32Array(v.sizeFromShape(r.shape)),i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values;for(let l=0;l<i.length;++l){let u=i[l];u>=1?s[l]=o[l]:s[l]=o[l]*(u+1)}return n.makeTensorInfo(r.shape,"float32",s)}var x8={kernelName:Im,backendName:"cpu",kernelFunc:b8},v8=N.ERF_P,w8=N.ERF_A1,k8=N.ERF_A2,I8=N.ERF_A3,S8=N.ERF_A4,T8=N.ERF_A5,N8=rt(Kl,e=>{let t=Math.sign(e),n=Math.abs(e),a=1/(1+v8*n);return t*(1-((((T8*a+S8)*a+I8)*a+k8)*a+w8)*a*Math.exp(-n*n))}),C8={kernelName:Kl,backendName:"cpu",kernelFunc:N8};function em(e){let{inputs:t,backend:n,attrs:a}=e,{input:r}=t,{dim:s}=a,i=r.shape.length,o=r.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),ft({inputs:{x:r},backend:n,attrs:{shape:o}})}var _8={kernelName:Yl,backendName:"cpu",kernelFunc:em},E8=Vt((e,t)=>e/t),q0=nn(Ci,E8),dx={kernelName:Ci,backendName:"cpu",kernelFunc:q0};function r_(e,t,n){let a=e.shape,r=a[0],s=a[1],i=n.data.get(e.dataId),o=i.complexTensorInfos.real,l=i.complexTensorInfos.imag,u=[r,s],p=v.sizeFromShape(u),d=v.getTypedArrayFromDType("float32",p),c=v.getTypedArrayFromDType("float32",p);for(let g=0;g<r;g++){let y=pi({inputs:{x:o},backend:n,attrs:{begin:[g,0],size:[1,s]}}),b=pi({inputs:{x:l},backend:n,attrs:{begin:[g,0],size:[1,s]}}),x=Zn({inputs:{real:y,imag:b},backend:n}),{real:w,imag:I}=A8(x,t,n),T=N.mergeRealAndImagArrays(w,I);for(let C=0;C<s;C++){let E=N.getComplexWithIndex(T,C);d[g*s+C]=E.real,c[g*s+C]=E.imag}n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(b),n.disposeIntermediateTensorInfo(x)}let h=n.makeTensorInfo(u,"float32",d),m=n.makeTensorInfo(u,"float32",c),f=Zn({inputs:{real:h,imag:m},backend:n});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}function A8(e,t,n){let a=v.sizeFromShape(e.shape),r=n.data.get(e.dataId),s=n.data.get(r.complexTensorInfos.real.dataId).values,i=n.data.get(r.complexTensorInfos.imag.dataId).values;if($8(a)){let o=hx(s,i,a,t,n),l=[e.shape[0],e.shape[1]];if(t){let u=n.makeTensorInfo(l,"float32",o.real),p=n.makeTensorInfo(l,"float32",o.imag),d=n.makeTensorInfo([],"float32",v.createScalarValue(a,"float32")),c=pr({inputs:{x:d},backend:n}),h=dx.kernelFunc({inputs:{a:u,b:d},backend:n}),m=dx.kernelFunc({inputs:{a:p,b:c},backend:n}),f=n.data.get(h.dataId).values,g=n.data.get(m.dataId).values;return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),{real:f,imag:g}}return o}else{let o=N.mergeRealAndImagArrays(s,i),l=F8(o,a,t);return N.splitRealAndImagArrays(l)}}function $8(e){return(e&e-1)===0}function hx(e,t,n,a,r){if(n===1)return{real:e,imag:t};let s=N.mergeRealAndImagArrays(e,t),i=n/2,o=N.complexWithEvenIndex(s),l=o.real,u=o.imag,p=[l.length],d=r.makeTensorInfo(p,"float32",l),c=r.makeTensorInfo(p,"float32",u),h=Zn({inputs:{real:d,imag:c},backend:r}),m=N.complexWithOddIndex(s),f=m.real,g=m.imag,y=[f.length],b=r.makeTensorInfo(y,"float32",f),x=r.makeTensorInfo(y,"float32",g),w=Zn({inputs:{real:b,imag:x},backend:r}),I=hx(l,u,i,a,r),T=I.real,C=I.imag,E=[T.length],A=r.makeTensorInfo(E,"float32",T),R=r.makeTensorInfo(E,"float32",C),F=Zn({inputs:{real:A,imag:R},backend:r}),S=hx(f,g,i,a,r),M=S.real,B=S.imag,U=[M.length],G=r.makeTensorInfo(U,"float32",M),q=r.makeTensorInfo(U,"float32",B),K=Zn({inputs:{real:G,imag:q},backend:r}),Z=N.exponents(n,a),Q=[Z.real.length],ee=r.makeTensorInfo(Q,"float32",Z.real),ae=r.makeTensorInfo(Q,"float32",Z.imag),te=Zn({inputs:{real:ee,imag:ae},backend:r}),le=qf({inputs:{a:te,b:K},backend:r}),ie=_l({inputs:{a:F,b:le},backend:r}),ye=H0({inputs:{a:F,b:le},backend:r}),ue=ui({inputs:{input:ie},backend:r}),xe=ui({inputs:{input:ye},backend:r}),Ie=El({inputs:{input:ie},backend:r}),Se=El({inputs:{input:ye},backend:r}),Le=Al({inputs:[ue,xe],backend:r,attrs:{axis:0}}),Ve=Al({inputs:[Ie,Se],backend:r,attrs:{axis:0}}),nt=r.data.get(Le.dataId).values,it=r.data.get(Ve.dataId).values;return r.disposeIntermediateTensorInfo(d),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(b),r.disposeIntermediateTensorInfo(x),r.disposeIntermediateTensorInfo(w),r.disposeIntermediateTensorInfo(A),r.disposeIntermediateTensorInfo(R),r.disposeIntermediateTensorInfo(F),r.disposeIntermediateTensorInfo(G),r.disposeIntermediateTensorInfo(q),r.disposeIntermediateTensorInfo(K),r.disposeIntermediateTensorInfo(ee),r.disposeIntermediateTensorInfo(ae),r.disposeIntermediateTensorInfo(te),r.disposeIntermediateTensorInfo(le),r.disposeIntermediateTensorInfo(ie),r.disposeIntermediateTensorInfo(ye),r.disposeIntermediateTensorInfo(ue),r.disposeIntermediateTensorInfo(Ie),r.disposeIntermediateTensorInfo(xe),r.disposeIntermediateTensorInfo(Se),r.disposeIntermediateTensorInfo(Le),r.disposeIntermediateTensorInfo(Ve),{real:nt,imag:it}}function F8(e,t,n){let a=new Float32Array(t*2);for(let r=0;r<t;r++){let s=0,i=0;for(let o=0;o<t;o++){let l=N.exponent(r*o,t,n),u=N.getComplexWithIndex(e,o);s+=u.real*l.real-u.imag*l.imag,i+=u.real*l.imag+u.imag*l.real}n&&(s/=t,i/=t),N.assignToTypedArray(a,s,i,r)}return a}function D8(e){let{inputs:t,backend:n}=e,{input:a}=t,r=v.sizeFromShape(a.shape),s=a.shape[a.shape.length-1],i=r/s,o=ft({inputs:{x:a},backend:n,attrs:{shape:[i,s]}}),l=r_(o,!1,n),u=ft({inputs:{x:l},backend:n,attrs:{shape:a.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(l),u}var R8={kernelName:Sm,backendName:"cpu",kernelFunc:D8};function K0(e){let{backend:t,attrs:n}=e,{shape:a,value:r,dtype:s}=n,i=s||v.inferDtype(r),o=v.getArrayFromDType(i,v.sizeFromShape(a));return P8(o,r,i),t.makeTensorInfo(a,i,o)}var M8={kernelName:gc,backendName:"cpu",kernelFunc:K0};function P8(e,t,n){e.fill(t)}var O8={kernelName:Jl,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,r=n,s=v.getTypedArrayFromDType(a.dtype,v.sizeFromShape(a.shape)),[i,o,l,u]=a.shape,p=r.data.get(a.dataId).values;for(let d=0;d<i;d++){let c=d*l*o*u;for(let h=0;h<o;h++){let m=h*(l*u);for(let f=0;f<l;f++){let g=f*u;for(let y=0;y<u;y++){let b=Math.round(l-f-1),x=c+m+g+y,w=p[x];if(b>=0&&b<l){let I=b*u,T=c+m+I+y;w=p[T]}s[x]=w}}}}return{dataId:r.write(s,a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},L8=Vt((e,t)=>Math.floor(e/t)),z8=nn($i,L8,null,"int32"),W8={kernelName:$i,backendName:"cpu",kernelFunc:z8};function B8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=n_({inputs:{x:r,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c}});if(i){let g=f;if(p==="NCHW"&&i.shape.length===1&&i.shape[0]!==1){let y=ft({inputs:{x:i},backend:n,attrs:{shape:[i.shape[0],1,1]}});f=_l({inputs:{a:f,b:y},backend:n}),n.disposeIntermediateTensorInfo(y)}else f=_l({inputs:{a:f,b:i},backend:n});n.disposeIntermediateTensorInfo(g)}if(h){let g=f;if(p==="NCHW"&&h==="prelu"&&o.shape.length===1&&o.shape[0]!==1){let y=ft({inputs:{x:o},backend:n,attrs:{shape:[o.shape[0],1,1]}});f=Qh(n,f,h,y,m),n.disposeIntermediateTensorInfo(y)}else f=Qh(n,f,h,o,m);n.disposeIntermediateTensorInfo(g)}return f}var V8={kernelName:ei,backendName:"cpu",kernelFunc:B8};function U8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=a_({inputs:{x:r,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c}});if(i){let g=f;f=_l({inputs:{a:f,b:i},backend:n}),n.disposeIntermediateTensorInfo(g)}if(h){let g=f;f=Qh(n,f,h,o,m),n.disposeIntermediateTensorInfo(g)}return f}var G8={kernelName:ti,backendName:"cpu",kernelFunc:U8};function H8(e){let{inputs:t,backend:n}=e,{params:a,indices:r}=t,s=v.sizeFromShape(a.shape),i=r.shape,o=i[i.length-1],[l,u,p,d]=N.prepareAndValidate(a,r);if(u===0)return n.makeTensorInfo(l,a.dtype,[]);let c=n.data.get(r.dataId).values,h=n.bufferSync(a),m=vC(c,h,a.dtype,u,o,p,d,a.shape,s);return n.makeTensorInfo(l,a.dtype,m.values)}var j8={kernelName:eu,backendName:"cpu",kernelFunc:H8};function q8(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,indices:s}=t,{axis:i,batchDims:o}=a;ge([r,s],"gatherV2");let l=v.parseAxisParam(i,r.shape)[0],u=n.data.get(s.dataId).values,p=r.shape[l];for(let w=0;w<u.length;++w){let I=u[w];v.assert(I<=p-1&&I>=0,()=>`GatherV2: the index value ${I} is not in [0, ${p-1}]`)}let d=o;o==null&&(d=0);let c=v.sizeFromShape(s.shape),h=N.segment_util.collectGatherOpShapeInfo(r,s,l,d),m=ft({inputs:{x:r},backend:n,attrs:{shape:[h.batchSize,h.outerSize,h.dimSize,h.sliceSize]}}),f=ft({inputs:{x:s},backend:n,attrs:{shape:[h.batchSize,c/h.batchSize]}}),g=[h.batchSize,h.outerSize,c/h.batchSize,h.sliceSize],y=n.bufferSync(f),b=n.bufferSync(m),x=wC(b,y,g);return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),n.makeTensorInfo(h.outputShape,x.dtype,x.values)}var K8={kernelName:Ql,backendName:"cpu",kernelFunc:q8};function X8(e){let{inputs:t,backend:n}=e,{input:a}=t,r=v.sizeFromShape(a.shape),s=a.shape[a.shape.length-1],i=r/s,o=ft({inputs:{x:a},backend:n,attrs:{shape:[i,s]}}),l=r_(o,!0,n),u=ft({inputs:{x:l},backend:n,attrs:{shape:a.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(l),u}var Y8={kernelName:Tm,backendName:"cpu",kernelFunc:X8},Z8=rt(nu,e=>Number.isFinite(e)?1:0,"bool"),J8={kernelName:nu,backendName:"cpu",kernelFunc:Z8},Q8=rt(au,e=>Math.abs(e)===1/0?1:0,"bool"),eK={kernelName:au,backendName:"cpu",kernelFunc:Q8},tK=rt(ru,e=>Number.isNaN(e)?1:0,"bool"),nK={kernelName:ru,backendName:"cpu",kernelFunc:tK};function aK(e){let{backend:t,attrs:n}=e,{start:a,stop:r,num:s}=n,i=NC(a,r,s);return t.makeTensorInfo([i.length],"float32",i)}var rK={kernelName:Cm,backendName:"cpu",kernelFunc:aK},sK=rt(ou,e=>Math.log1p(e)),iK={kernelName:ou,backendName:"cpu",kernelFunc:sK},oK=Vt((e,t)=>e&&t),lK=nn(lu,oK,null,"bool"),uK={kernelName:lu,backendName:"cpu",kernelFunc:lK},pK=rt(uu,e=>e?0:1,"bool"),cK={kernelName:uu,backendName:"cpu",kernelFunc:pK},dK=Vt((e,t)=>e||t),hK=nn(pu,dK,null,"bool"),mK={kernelName:pu,backendName:"cpu",kernelFunc:hK};function fK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=a;ge(r,"LRN");let u=r.shape[3],p=u-1,d=n.data.get(r.dataId).values,c=v.sizeFromShape(r.shape),h=new Float32Array(c);function m(f){let g=f%u,y=f-g+Math.max(0,g-s),b=f-g+Math.min(g+s,p),x=0;for(;y<=b;y++){let w=d[y];x+=w*w}return x}for(let f=0;f<c;f++){let g=m(f),y=d[f]*Math.pow(i+o*g,-l);h[f]=y}return n.makeTensorInfo(r.shape,r.dtype,h)}var gK={kernelName:yc,backendName:"cpu",kernelFunc:fK};function yK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:p}=a;ge(i,"LRNGrad");let d=v.sizeFromShape(i.shape),c=i.shape[3],h=n.data.get(i.dataId).values,m=n.data.get(r.dataId).values,f=n.data.get(s.dataId).values,g=new Float32Array(d),y=d;for(let b=0;b<y;b++){let x=b%c,w=b-x+Math.max(0,x-o),I=b-x+Math.min(c,x+o+1),T=0;for(let C=w;C<I;C++)T+=Math.pow(m[C],2);T=u*T+l;for(let C=w;C<I;C++){let E=-2*u*p*m[C]*f[b]/T;b===C&&(E+=Math.pow(T,-p)),E*=h[b],g[C]+=E}}return n.makeTensorInfo(i.shape,r.dtype,g)}var bK={kernelName:_m,backendName:"cpu",kernelFunc:yK};function s_(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reductionIndices:s,keepDims:i}=a,o=n,l=r.shape,u=l.length,p=v.parseAxisParam(s,l),d=p,c=N.getAxesPermutation(d,u),h=o.data.get(r.dataId).values;if(c!=null){let w=new Array(u);for(let I=0;I<w.length;I++)w[I]=l[c[I]];h=z0(h,l,r.dtype,c,w),d=N.getInnerMostAxes(d.length,u),l=w}ge(r,"max"),N.assertAxesAreInnerMostDims("max",d,u);let[m,f]=N.computeOutAndReduceShapes(l,d),g=v.sizeFromShape(f),y=_C(h,g,m,r.dtype),b=o.write(y,m,r.dtype),x=m;return i&&(x=N.expandShapeToKeepDim(m,p)),{dataId:b,shape:x,dtype:r.dtype}}var xK={kernelName:Oi,backendName:"cpu",kernelFunc:s_};function vK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;ge(r,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l),d;if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))d=pr({inputs:{x:r},backend:n});else{let c=n.data.get(r.dataId).values,h=v.computeStrides(r.shape),m=j0(c,r.shape,r.dtype,h,p,"max");d=n.makeTensorInfo(p.outShape,r.dtype,m.values)}return d}var wK={kernelName:zi,backendName:"cpu",kernelFunc:vK};function kK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a;ge(r,"maxPool3d");let p=N.computePool3DInfo(r.shape,s,i,1,o,l,u),d=n.data.get(r.dataId).values,c=t_(d,r.shape,r.dtype,v.computeStrides(r.shape),p,"max");return n.makeTensorInfo(c.shape,"float32",c.values)}var IK={kernelName:bc,backendName:"cpu",kernelFunc:kK};function SK(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=a;ge([r,s],"maxPool3DGrad");let p=N.computePool3DInfo(s.shape,i,o,1,l,u),d=n.bufferSync(s),c=c5(d,p),h=p.strideDepth,m=p.strideHeight,f=p.strideWidth,g=p.dilationDepth,y=p.dilationHeight,b=p.dilationWidth,x=p.effectiveFilterDepth,w=p.effectiveFilterHeight,I=p.effectiveFilterWidth,T=x-1-p.padInfo.front,C=I-1-p.padInfo.left,E=w-1-p.padInfo.top,A=Oe(s.shape,"float32"),R=n.bufferSync(r);for(let F=0;F<p.batchSize;++F)for(let S=0;S<p.inChannels;++S)for(let M=0;M<p.inDepth;++M)for(let B=0;B<p.inHeight;++B)for(let U=0;U<p.inWidth;++U){let G=M-T,q=B-E,K=U-C,Z=0;for(let Q=0;Q<x;Q+=g){let ee=(G+Q)/h;if(!(ee<0||ee>=p.outDepth||Math.floor(ee)!==ee))for(let ae=0;ae<w;ae+=y){let te=(q+ae)/m;if(!(te<0||te>=p.outHeight||Math.floor(te)!==te))for(let le=0;le<I;le+=b){let ie=(K+le)/f;if(ie<0||ie>=p.outWidth||Math.floor(ie)!==ie)continue;let ye=x*w*I-1-c.get(F,ee,te,ie,S),ue=Q*w*I+ae*I+le,xe=ye===ue?1:0;xe!==0&&(Z+=R.get(F,ee,te,ie,S)*xe)}}}A.set(Z,F,M,B,U,S)}return n.makeTensorInfo(A.shape,A.dtype,A.values)}var TK={kernelName:Am,backendName:"cpu",kernelFunc:SK};function NK(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s,output:i}=t,o=s;ge([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:p,dimRoundingMode:d}=a,c=N.computePool2DInfo(o.shape,l,u,1,p,d),h=n.data.get(o.dataId).values,m=Oe(c.outShape,o.dtype,e_(h,o.shape,o.dtype,c).values),f=c.strideHeight,g=c.strideWidth,y=c.dilationHeight,b=c.dilationWidth,x=c.effectiveFilterHeight,w=c.effectiveFilterWidth,I=w-1-c.padInfo.left,T=x-1-c.padInfo.top,C=Oe(o.shape,"float32"),E=n.data.get(r.dataId).values,A=Oe(r.shape,"float32",E);for(let R=0;R<c.batchSize;++R)for(let F=0;F<c.inChannels;++F)for(let S=0;S<c.inHeight;++S)for(let M=0;M<c.inWidth;++M){let B=S-T,U=M-I,G=0;for(let q=0;q<x;q+=y){let K=(B+q)/f;if(!(K<0||K>=c.outHeight||Math.floor(K)!==K))for(let Z=0;Z<w;Z+=b){let Q=(U+Z)/g;if(Q<0||Q>=c.outWidth||Math.floor(Q)!==Q)continue;let ee=x*w-1-m.get(R,K,Q,F),ae=q*w+Z,te=ee===ae?1:0;te!==0&&(G+=A.get(R,K,Q,F)*te)}}C.set(G,R,S,M,F)}return n.makeTensorInfo(C.shape,C.dtype,C.values)}var CK={kernelName:Em,backendName:"cpu",kernelFunc:NK};function _K(e,t,n,a,r){let s=v.computeStrides(t),i=j0(e,t,n,s,r,"max"),o=e_(e,t,n,r,!0,a);return[i.values,o.values]}var EK={kernelName:$m,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{filterSize:r,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;ge(a,"MaxPoolWithArgmax");let u=l.data.get(a.dataId).values,p=N.computePool2DInfo(a.shape,r,s,[1,1],i),[d,c]=_K(u,a.shape,a.dtype,o,p),h=l.write(d,p.outShape,a.dtype),m=l.write(c,p.outShape,a.dtype);return[{dataId:h,shape:p.outShape,dtype:a.dtype},{dataId:m,shape:p.outShape,dtype:"int32"}]}};function AK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=v.parseAxisParam(s,r.shape),l=N.computeOutAndReduceShapes(r.shape,o)[1],u=v.sizeFromShape(l),p=[],d=n.makeTensorInfo([],"float32",new Float32Array([u]));p.push(d);let c=ds({inputs:{x:r},backend:n,attrs:{dtype:"float32"}});p.push(c);let h=q0({inputs:{a:c,b:d},backend:n});p.push(h);let m=sd({inputs:{x:h},backend:n,attrs:{axis:s,keepDims:i}});return p.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var $K={kernelName:Wi,backendName:"cpu",kernelFunc:AK};function FK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;ge(r,"min");let o=v.parseAxisParam(s,r.shape),l=o,u=N.getAxesPermutation(l,r.shape.length),p=r;u!=null&&(p=Vn({inputs:{x:r},backend:n,attrs:{perm:u}}),l=N.getInnerMostAxes(l.length,r.shape.length)),N.assertAxesAreInnerMostDims("min",l,p.shape.length);let[d,c]=N.computeOutAndReduceShapes(p.shape,l),h=v.sizeFromShape(c),m=v.makeZerosTypedArray(v.sizeFromShape(d),p.dtype),f=n.data.get(p.dataId).values;for(let y=0;y<m.length;++y){let b=y*h,x=f[b];for(let w=0;w<h;++w){let I=f[b+w];(Number.isNaN(I)||I<x)&&(x=I)}m[y]=x}u!=null&&n.disposeIntermediateTensorInfo(p);let g=n.makeTensorInfo(d,p.dtype,m);if(i){let y=N.expandShapeToKeepDim(d,o),b=ft({inputs:{x:g},backend:n,attrs:{shape:y}});return n.disposeIntermediateTensorInfo(g),b}return g}var DK={kernelName:Bi,backendName:"cpu",kernelFunc:FK};function RK(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,mode:i}=a;ge(r,"mirrorPad");let o=s.map((b,x)=>b[0]+r.shape[x]+b[1]),l=s.map(b=>b[0]),u=s.map((b,x)=>b[0]+r.shape[x]),p=i==="reflect"?0:1,d=n.data.get(r.dataId).values,c=r.shape.length,h=v.computeStrides(r.shape),m=v.sizeFromShape(o),f=o.length,g=v.computeStrides(o),y=v.getTypedArrayFromDType(r.dtype,m);for(let b=0;b<m;b++){let x=v.indexToLoc(b,f,g);for(let I=0;I<f;I++)x[I]<l[I]?x[I]=l[I]*2-x[I]-p:x[I]>=u[I]&&(x[I]=(u[I]-1)*2-x[I]+p);x=x.map((I,T)=>I-l[T]);let w=v.locToIndex(x,c,h);y[b]=d[w]}return{dataId:n.write(y,o,r.dtype),shape:o,dtype:r.dtype}}var MK={kernelName:Ui,backendName:"cpu",kernelFunc:RK},PK=Vt((e,t)=>{let n=e%t;return e<0&&t<0||e>=0&&t>=0?n:(n+t)%t}),OK=nn(cu,PK),LK={kernelName:cu,backendName:"cpu",kernelFunc:OK},zK=ms(im());function i_(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{dim:s}=a,i=r.shape.length,o=s;if(o===-1&&(o=i-1),o!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);let l=v.parseAxisParam([o],r.shape),u=s_({inputs:{x:r},backend:n,attrs:{reductionIndices:l,keepDims:!1}}),p=N.expandShapeToKeepDim(u.shape,l),d=ft({inputs:{x:u},backend:n,attrs:{shape:p}}),c=H0({inputs:{a:r,b:d},backend:n}),h=yC({inputs:{x:c},backend:n}),m=sd({inputs:{x:h},backend:n,attrs:{axis:l,keepDims:!1}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:p}}),g=q0({inputs:{a:h,b:f},backend:n});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(f),g}var WK={kernelName:oo,backendName:"cpu",kernelFunc:i_};function BK(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{numSamples:s,seed:i,normalized:o}=a;ge(r,"multinomial");let l=o?r:i_({inputs:{logits:r},backend:n,attrs:{dim:-1}}),u=l.shape[0],p=l.shape[1],d=n.data.get(l.dataId).values,c=[u,s],h=v.makeZerosTypedArray(v.sizeFromShape(c),"int32");for(let m=0;m<u;++m){let f=m*p,g=new Float32Array(p-1);g[0]=d[f];for(let x=1;x<g.length;++x)g[x]=g[x-1]+d[f+x];let y=zK.alea(i.toString()),b=m*s;for(let x=0;x<s;++x){let w=y();h[b+x]=g.length;for(let I=0;I<g.length;I++)if(w<g[I]){h[b+x]=I;break}}}return o||n.disposeIntermediateTensorInfo(l),n.makeTensorInfo(c,"int32",h)}var VK={kernelName:Fm,backendName:"cpu",kernelFunc:BK},UK=hr.nonMaxSuppressionV3Impl;function GK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=a;ge(r,"NonMaxSuppression");let u=n.data.get(r.dataId).values,p=n.data.get(s.dataId).values,{selectedIndices:d}=UK(u,p,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}var HK={kernelName:mu,backendName:"cpu",kernelFunc:GK},jK=hr.nonMaxSuppressionV4Impl;function qK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=a;ge(r,"NonMaxSuppressionPadded");let p=n.data.get(r.dataId).values,d=n.data.get(s.dataId).values,{selectedIndices:c,validOutputs:h}=jK(p,d,i,o,l,u);return[n.makeTensorInfo([c.length],"int32",new Int32Array(c)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var KK={kernelName:fu,backendName:"cpu",kernelFunc:qK},XK=hr.nonMaxSuppressionV5Impl;function YK(e){let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=a;ge(r,"NonMaxSuppressionWithScore");let p=n.data.get(r.dataId).values,d=n.data.get(s.dataId).values,c=i,h=o,m=l,f=u,{selectedIndices:g,selectedScores:y}=XK(p,d,c,h,m,f);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var ZK={kernelName:gu,backendName:"cpu",kernelFunc:YK};function JK(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a;ge(r,"oneHot");let u=v.sizeFromShape(r.shape),p=new Float32Array(u*i);p.fill(l);let d=n.data.get(r.dataId).values;for(let c=0;c<u;++c)d[c]>=0&&d[c]<i&&(p[c*i+d[c]]=o);return n.makeTensorInfo([...r.shape,i],s,p)}var QK={kernelName:Hi,backendName:"cpu",kernelFunc:JK};function tm(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("zerosLike is not supported for string tensors");if(a.dtype==="complex64"){let r=ui({inputs:{input:a},backend:n}),s=tm({inputs:{x:r},backend:n}),i=El({inputs:{input:a},backend:n}),o=tm({inputs:{x:i},backend:n}),l=Zn({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return K0({backend:n,attrs:{shape:a.shape,value:0,dtype:a.dtype}})}var eX={kernelName:Mu,backendName:"cpu",kernelFunc:tm};function o_(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("onesLike is not supported for string tensors");if(a.dtype==="complex64"){let r=ui({inputs:{input:a},backend:n}),s=o_({inputs:{x:r},backend:n}),i=El({inputs:{input:a},backend:n}),o=tm({inputs:{x:i},backend:n}),l=Zn({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return K0({backend:n,attrs:{shape:a.shape,value:1,dtype:a.dtype}})}var tX={kernelName:yu,backendName:"cpu",kernelFunc:o_};function l_(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return em({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=em({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=Al({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeIntermediateTensorInfo(p)),u}var nX={kernelName:bu,backendName:"cpu",kernelFunc:l_};function aX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,constantValue:i}=a;ge(r,"pad");let o=s.map((y,b)=>y[0]+r.shape[b]+y[1]),l=s.map(y=>y[0]),u=n.data.get(r.dataId).values,p=v.sizeFromShape(r.shape),d=r.shape.length,c=v.computeStrides(r.shape),h=v.sizeFromShape(o),m=o.length,f=v.computeStrides(o),g=v.getTypedArrayFromDType(r.dtype,h);i!==0&&g.fill(i);for(let y=0;y<p;y++){let b=v.indexToLoc(y,d,c).map((w,I)=>w+l[I]),x=v.locToIndex(b,m,f);g[x]=u[y]}return{dataId:n.write(g,o,r.dtype),shape:o,dtype:r.dtype}}var u_={kernelName:ji,backendName:"cpu",kernelFunc:aX},rX=Vt((e,t)=>Math.pow(e,t)),sX=nn(qi,rX),iX={kernelName:qi,backendName:"cpu",kernelFunc:sX};function oX(e){let{inputs:t,backend:n,attrs:a}=e,{paramsNestedSplits:r,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=a,l=r.map(y=>n.data.get(y.dataId).values),u=r.map(y=>y.shape),p=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,[c,h,m]=RC(l,u,p,s.shape,s.dtype,d,i.shape,o),f=c.map(y=>n.makeTensorInfo([y.length],"int32",y)),g=n.makeTensorInfo(m,s.dtype,h);return f.concat([g])}var lX={kernelName:Dm,backendName:"cpu",kernelFunc:oX};function uX(e){let{inputs:t,backend:n}=e,{starts:a,limits:r,deltas:s}=t,i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=MC(i,a.shape,a.dtype,o,r.shape,l,s.shape),d=n.makeTensorInfo([u.length],"int32",u),c=n.makeTensorInfo([p.length],a.dtype,p);return[d,c]}var pX={kernelName:Rm,backendName:"cpu",kernelFunc:uX};function cX(e){let{inputs:t,backend:n,attrs:a}=e,{shape:r,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=a,u=n.data.get(r.dataId).values,p=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,c=o.map(g=>n.data.get(g.dataId).values),h=o.map(g=>g.shape),[m,f]=PC(u,r.shape,p,s.shape,s.dtype,d,i.shape,c,h,l);return n.makeTensorInfo(m,s.dtype,f)}var dX={kernelName:Mm,backendName:"cpu",kernelFunc:cX};function hX(e){let{backend:t,attrs:n}=e,{start:a,stop:r,dtype:s,step:i}=n,o=W0(a,r,i,s);return t.makeTensorInfo([o.length],s,o)}var mX={kernelName:xc,backendName:"cpu",kernelFunc:hX},fX=rt(xu,e=>1/e),gX={kernelName:xu,backendName:"cpu",kernelFunc:fX};function yX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a;ge(r,"resizeBilinear");let l=v.computeStrides(r.shape),[u,p]=o,[d,c,h,m]=r.shape,f=n.data.get(r.dataId).values,g=new Float32Array(v.sizeFromShape([d,u,p,m])),y=[s&&u>1?c-1:c,s&&p>1?h-1:h],b=[s&&u>1?u-1:u,s&&p>1?p-1:p],x=0,w=y[0]/b[0],I=y[1]/b[1];for(let T=0;T<d;T++)for(let C=0;C<u;C++){let E;i?E=w*(C+.5)-.5:E=w*C;let A=Math.max(0,Math.floor(E)),R=E-A,F=Math.min(c-1,Math.ceil(E)),S=T*l[0]+A*l[1],M=T*l[0]+F*l[1];for(let B=0;B<p;B++){let U;i?U=I*(B+.5)-.5:U=I*B;let G=Math.max(0,Math.floor(U)),q=U-G,K=Math.min(h-1,Math.ceil(U)),Z=S+G*l[2],Q=M+G*l[2],ee=S+K*l[2],ae=M+K*l[2];for(let te=0;te<m;te++){let le=f[Z+te],ie=f[Q+te],ye=f[ee+te],ue=f[ae+te],xe=le+(ye-le)*q,Ie=ie+(ue-ie)*q,Se=xe+(Ie-xe)*R;g[x++]=Se}}}return n.makeTensorInfo([d,u,p,m],"float32",g)}var bX={kernelName:Ji,backendName:"cpu",kernelFunc:yX};function xX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a;ge([s,r],"resizeBilinearGrad");let o=v.computeStrides(r.shape),[l,u,p,d]=r.shape,[,c,h]=s.shape,m=new Float32Array(l*u*p*d),f=[i&&c>1?u-1:u,i&&h>1?p-1:p],g=[i&&c>1?c-1:c,i&&h>1?h-1:h],y=f[0]/g[0],b=f[1]/g[1],x=n.data.get(s.dataId).values,w=0;for(let I=0;I<l;I++){let T=I*o[0];for(let C=0;C<c;C++){let E=C*y,A=Math.floor(E),R=Math.min(Math.ceil(E),u-1),F=T+A*o[1],S=T+R*o[1],M=E-A,B=1-M;for(let U=0;U<h;U++){let G=U*b,q=Math.floor(G),K=Math.min(Math.ceil(G),p-1),Z=G-q,Q=1-Z,ee=F+q*o[2],ae=F+K*o[2],te=S+q*o[2],le=S+K*o[2],ie=B*Q,ye=B*Z,ue=M*Q,xe=M*Z;for(let Ie=0;Ie<d;Ie++){let Se=x[w++];m[ee+Ie]+=Se*ie,m[ae+Ie]+=Se*ye,m[te+Ie]+=Se*ue,m[le+Ie]+=Se*xe}}}}return n.makeTensorInfo([l,p,u,d],"float32",m)}var vX={kernelName:Lm,backendName:"cpu",kernelFunc:xX};function wX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a;ge(r,"resizeNearestNeighbor");let l=v.computeStrides(r.shape),[u,p]=o,[d,c,h,m]=r.shape,f=n.data.get(r.dataId).values,g=new Float32Array(d*u*p*m),y=[s&&u>1?c-1:c,s&&p>1?h-1:h],b=[s&&u>1?u-1:u,s&&p>1?p-1:p],x=y[0]/b[0],w=y[1]/b[1],I=0;for(let T=0;T<d;T++){let C=T*l[0];for(let E=0;E<u;E++){let A=i?x*(E+.5):x*E,R=Math.min(c-1,s?Math.round(A):Math.floor(A));i&&(R=Math.max(0,R));let F=C+R*l[1];for(let S=0;S<p;S++){let M=i?w*(S+.5):w*S,B=Math.min(h-1,s?Math.round(M):Math.floor(M));i&&(B=Math.max(0,B));let U=F+B*l[2];for(let G=0;G<m;G++){let q=f[U+G];g[I++]=q}}}}return n.makeTensorInfo([d,u,p,m],r.dtype,g)}var kX={kernelName:Zi,backendName:"cpu",kernelFunc:wX};function IX(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a;ge([s,r],"resizeNearestNeighborGrad");let o=v.computeStrides(r.shape),l=v.computeStrides(s.shape),[u,p,d,c]=r.shape,[,h,m]=s.shape,f=new Float32Array(u*p*d*c),g=n.data.get(s.dataId).values,y=[i&&h>1?p-1:p,i&&m>1?d-1:d],b=[i&&h>1?h-1:h,i&&m>1?m-1:m],x=y[0]/b[0],w=y[1]/b[1],I=1/x,T=1/w,C=Math.ceil(I)*2+2,E=Math.ceil(T)*2+2;for(let A=0;A<u;A++){let R=A*o[0];for(let F=0;F<p;F++){let S=R+F*o[1],M=Math.floor(F*I),B=Math.floor(M-C/2);for(let U=0;U<d;U++){let G=S+U*o[2],q=Math.floor(U*T),K=Math.floor(q-E/2);for(let Z=0;Z<c;Z++){let Q=0;for(let ee=0;ee<C;ee++){let ae=ee+B;if(ae<0||ae>=h)continue;let te=R+ae*l[1],le=ae*x,ie=Math.min(p-1,i?Math.round(le):Math.floor(le));if(F===ie)for(let ye=0;ye<E;ye++){let ue=ye+K;if(ue<0||ue>=m)continue;let xe=te+ue*l[2],Ie=ue*w,Se=Math.min(d-1,i?Math.round(Ie):Math.floor(Ie));U===Se&&(Q+=g[xe+Z])}}f[G+Z]=Q}}}}return n.makeTensorInfo(r.shape,r.dtype,f)}var SX={kernelName:Om,backendName:"cpu",kernelFunc:IX};function TX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a;ge(r,"reverse");let i=r.shape.length,o=v.parseAxisParam(s,r.shape);if(i===0)return pr({inputs:{x:r},backend:n});let l=new Ht(r.shape,r.dtype),u=n.bufferSync(r);for(let p=0;p<l.size;p++){let d=l.indexToLoc(p),c=d.slice();o.forEach(h=>c[h]=r.shape[h]-1-c[h]),l.set(u.get(...c),...d)}return n.makeTensorInfo(l.shape,l.dtype,l.values)}var NX={kernelName:eo,backendName:"cpu",kernelFunc:TX},CX={kernelName:Pu,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,{radians:r,fillValue:s,center:i}=t,o=n,l=v.getTypedArrayFromDType(a.dtype,v.sizeFromShape(a.shape)),[u,p,d,c]=a.shape,[h,m]=N.getImageCenter(i,p,d),f=255,g=Math.sin(r),y=Math.cos(r),b=o.data.get(a.dataId).values;for(let x=0;x<u;x++){let w=x*d*p*c;for(let I=0;I<p;I++){let T=I*(d*c);for(let C=0;C<d;C++){let E=C*c;for(let A=0;A<c;A++){let R=[u,I,C,A],F=R[2],S=R[1],M=(F-h)*y-(S-m)*g,B=(F-h)*g+(S-m)*y;M=Math.round(M+h),B=Math.round(B+m);let U=s;if(typeof s!="number"&&(A===3?U=f:U=s[A]),M>=0&&M<d&&B>=0&&B<p){let q=B*(d*c),K=M*c,Z=w+q+K+A;U=b[Z]}let G=w+T+E+A;l[G]=U}}}}return{dataId:o.write(l,a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},_X=rt(to,e=>{let t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2===0?t:t+1}),EX={kernelName:to,backendName:"cpu",kernelFunc:_X};function AX(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r,updates:s}=t,{shape:i}=a,{sliceRank:o,numUpdates:l,sliceSize:u,strides:p,outputSize:d}=N.calculateShapes(s,r,i),c=!0,h=n.bufferSync(r),m=n.bufferSync(s),f=ll(h,m,i,d,u,l,o,p,0,c);return n.makeTensorInfo(i,f.dtype,f.values)}var $X={kernelName:wu,backendName:"cpu",kernelFunc:AX};function FX(e,t){let n=0,a=e.length,r=0;for(;n<a;)r=Math.floor((n+a)/2),e[r]<t?n=r+1:a=r;return a}function DX(e,t){let n=0,a=e.length,r=0;for(;n<a;)r=Math.floor((n+a)/2),e[r]<=t?n=r+1:a=r;return a}function RX(e,t,n,a,r,s){let i=v.getArrayFromDType("int32",n*r);for(let o=0;o<n;++o){let l=e.slice(o*a,(o+1)*a),u=o*r;for(let p=0;p<r;++p)i[u+p]=s==="left"?FX(l,t[p+u]):DX(l,t[p+u])}return i}function MX(e){let{inputs:t,backend:n,attrs:a}=e,{sortedSequence:r,values:s}=t,{side:i}=a,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,u=RX(o,l,r.shape[0],r.shape[1],s.shape[1],i);return n.makeTensorInfo(s.shape,"int32",u)}var PX={kernelName:zm,backendName:"cpu",kernelFunc:MX};function OX(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t;ge([a,r,s],"select");let i=a.shape.length,o=n.data.get(a.dataId).values,l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=fa(r.dtype,s.dtype),d=v.makeZerosTypedArray(v.sizeFromShape(r.shape),p),c=0,h=i===0||i>1||r.shape.length===1?1:v.sizeFromShape(r.shape.slice(1));for(let m=0;m<o.length;m++)for(let f=0;f<h;f++)o[m]===1?d[c++]=l[m]:d[c++]=u[m];return n.makeTensorInfo(r.shape,p,d)}var LX={kernelName:ku,backendName:"cpu",kernelFunc:OX},zX=N.SELU_SCALEALPHA,WX=N.SELU_SCALE,BX=rt(Iu,e=>e>=0?WX*e:zX*(Math.exp(e)-1)),VX={kernelName:Iu,backendName:"cpu",kernelFunc:BX},UX=rt(Nu,e=>e<0?-1:e>0?1:0),GX={kernelName:Nu,backendName:"cpu",kernelFunc:UX},HX=rt(ao,e=>Math.sin(e)),jX={kernelName:ao,backendName:"cpu",kernelFunc:HX},qX=rt(Tu,e=>Math.sinh(e)),KX={kernelName:Tu,backendName:"cpu",kernelFunc:qX},XX=11920928955078125e-23,pI=Math.log(XX)+2,YX=rt(Cu,e=>{let t=e>-pI,n=e<pI,a=Math.exp(e),r;return n?r=a:t?r=e:r=Math.log(1+a),r}),ZX={kernelName:Cu,backendName:"cpu",kernelFunc:YX};function JX(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a;ge([r],"spaceToBatchND");let o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<r.shape.length;++g)l.push([0,0]);let u=u_.kernelFunc({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),p=N.getReshaped(u.shape,s,o,!1),d=N.getPermuted(p.length,s.length,!1),c=N.getReshapedPermuted(u.shape,s,o,!1),h=ft({inputs:{x:u},backend:n,attrs:{shape:p}}),m=Vn({inputs:{x:h},backend:n,attrs:{perm:d}}),f=ft({inputs:{x:m},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}var QX={kernelName:_u,backendName:"cpu",kernelFunc:JX};function eY(e){let{inputs:t,backend:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-        ${s.shape}`);if(a.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Values must be a vector, saw:
-        ${r.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let o=n.data.get(a.dataId).values,l=n.data.get(r.dataId).values,u=n.data.get(s.dataId).values,p=n.data.get(i.dataId).values[0],[d,c,h,m,f]=zC(o,a.shape,a.dtype,l,r.dtype,u,p);return[n.makeTensorInfo(c,a.dtype,d),n.makeTensorInfo([c[0]],r.dtype,h),n.makeTensorInfo([m.length],"bool",new Uint8Array(m.map(g=>Number(g)))),n.makeTensorInfo([f.length],a.dtype,new Int32Array(f))]}var tY={kernelName:vc,backendName:"cpu",kernelFunc:eY};function nY(e){let{inputs:t,backend:n}=e,{inputIndices:a,inputShape:r,newShape:s}=t;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.data.get(r.dataId).values),o=n.data.get(a.dataId).values,l=Array.from(n.data.get(s.dataId).values),[u,p,d]=WC(o,a.shape,a.dtype,i,l);return[n.makeTensorInfo(p,a.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}var aY={kernelName:Au,backendName:"cpu",kernelFunc:nY};function rY(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-          ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-          ${s.shape}`);if(r.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=B0(i,a.shape,a.dtype,o,l,!0);return n.makeTensorInfo(p,a.dtype,u)}var sY={kernelName:wc,backendName:"cpu",kernelFunc:rY};function iY(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-         ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-         ${s.shape}`);if(r.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values,l=n.data.get(s.dataId).values,[u,p]=B0(i,a.shape,a.dtype,o,l);return n.makeTensorInfo(p,a.dtype,u)}var oY={kernelName:kc,backendName:"cpu",kernelFunc:iY};function lY(e){let{inputs:t,backend:n,attrs:a}=e,{sparseIndices:r,sparseValues:s,defaultValue:i}=t,{outputShape:o}=a,{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=N.calculateShapes(s,r,o),h=!1,m=n.bufferSync(r),f;switch(s.dtype){case"bool":{let g=n.bufferSync(s),y=Boolean(n.data.get(i.dataId).values[0]);f=ll(m,g,o,c,p,u,l,d,y,h);break}case"float32":{let g=n.bufferSync(s),y=n.data.get(i.dataId).values[0];f=ll(m,g,o,c,p,u,l,d,y,h);break}case"int32":{let g=n.bufferSync(s),y=n.data.get(i.dataId).values[0];f=ll(m,g,o,c,p,u,l,d,y,h);break}case"string":{let g=n.bufferSync(s),y=v.decodeString(n.data.get(i.dataId).values[0]);f=ll(m,g,o,c,p,u,l,d,y,h);break}default:throw new Error(`Unsupported type ${s.dtype}`)}return n.makeTensorInfo(o,f.dtype,f.values)}var uY={kernelName:Wm,backendName:"cpu",kernelFunc:lY};function pY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=a,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=new Array(r.shape.length).fill(0),p=r.shape.slice();return l.map(d=>{let c=[...p];c[o]=d;let h=pi({inputs:{x:r},backend:n,attrs:{begin:u,size:c}});return u[o]+=d,h})}var cY={kernelName:Eu,backendName:"cpu",kernelFunc:pY},dY={kernelName:Ic,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{let{x:n}=e,a=t;ge(n,"square");let r=a.data.get(n.dataId).values,s=new Float32Array(r.length);for(let i=0;i<r.length;++i){let o=r[i];s[i]=o*o}return{dataId:a.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},hY=rt(xs,(e,t)=>{let n=t;return isNaN(e)?NaN:e>0?1:n.alpha}),mY={kernelName:xs,backendName:"cpu",kernelFunc:hY};function fY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a;ge(r,"stridedSlice");let{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=ft({inputs:{x:r},backend:n,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let T=jt.computeOutShape(b,x,w),C=pi({inputs:{x:r},backend:n,attrs:{begin:b,size:T}});I=ft({inputs:{x:C},backend:n,attrs:{shape:m}}),n.disposeIntermediateTensorInfo(C)}else{let T=n.bufferSync(r),C=VC(h,T,w,b);I=n.makeTensorInfo(m,C.dtype,C.values)}return I}var gY={kernelName:$u,backendName:"cpu",kernelFunc:fY};function yY(e){let{inputs:t,backend:n,attrs:a}=e,{separator:r,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=a,{data:p,dataSplits:d}=t,c=n.data.get(p.dataId).values,h=n.data.get(d.dataId).values,[m,f]=V0(c,h,r,s,i,o,l,u);return[n.makeTensorInfo([m.length],"string",m),n.makeTensorInfo(d.shape,"int32",f)]}var bY={kernelName:Sc,backendName:"cpu",kernelFunc:yY};function xY(e){let{inputs:t,backend:n,attrs:a}=e,{skipEmpty:r}=a,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values[0],[u,p,d]=U0(o,l,r),c=p.length;return[n.makeTensorInfo([c,2],"int32",u),n.makeTensorInfo([c],"string",p),n.makeTensorInfo([2],"int32",new Int32Array(d))]}var vY={kernelName:Tc,backendName:"cpu",kernelFunc:xY};function wY(e){let{inputs:t,backend:n,attrs:a}=e,{numBuckets:r}=a,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(r<=0)throw new Error("Number of buckets must be at least 1");let i=n.data.get(s.dataId).values,o=G0(i,r);return n.makeTensorInfo(s.shape,"int32",o)}var kY={kernelName:Nc,backendName:"cpu",kernelFunc:wY},IY=rt(po,e=>Math.tan(e)),SY={kernelName:po,backendName:"cpu",kernelFunc:IY},TY=rt(co,e=>Math.tanh(e)),NY={kernelName:co,backendName:"cpu",kernelFunc:TY};function CY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reps:s}=a;ge(r,"tile");let i=GC(n.bufferSync(r),s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}var _Y={kernelName:bs,backendName:"cpu",kernelFunc:CY};function EY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{k:s,sorted:i}=a;ge(r,"topk");let o=n.data.get(r.dataId).values,[l,u]=jC(o,r.shape,r.dtype,s,i);return[n.makeTensorInfo(l.shape,l.dtype,l.values),n.makeTensorInfo(u.shape,u.dtype,u.values)]}var AY={kernelName:Fu,backendName:"cpu",kernelFunc:EY};function $Y(e){let{inputs:t,attrs:n,backend:a}=e,{image:r,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=v.computeStrides(r.shape),b=y[0],x=y[1],w=y[2],I=v.computeStrides(g),T=I[0],C=I[1],E=I[2],A=v.getTypedArrayFromDType(r.dtype,v.sizeFromShape(g));A.fill(l);let R=a.data.get(r.dataId).values,F=a.data.get(s.dataId).values;for(let S=0;S<p;++S){let M=s.shape[0]===1?F:F.subarray(S*8,S*8+8);for(let B=0;B<m;++B)for(let U=0;U<f;++U)for(let G=0;G<h;++G){let q,K=M[6]*U+M[7]*B+1;if(K===0)continue;let Z=(M[0]*U+M[1]*B+M[2])/K,Q=(M[3]*U+M[4]*B+M[5])/K,ee=cI(Z,c,o),ae=cI(Q,d,o);switch(i){case"nearest":q=OY(R,d,c,b,x,w,S,ae,ee,G,l);break;case"bilinear":q=LY(R,d,c,b,x,w,S,ae,ee,G,l);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}let te=S*T+B*C+U*E+G;A[te]=q}return a.makeTensorInfo(g,r.dtype,A)}return{dataId:a.write(A,g,r.dtype),shape:r.shape,dtype:r.dtype}}var FY={kernelName:Du,backendName:"cpu",kernelFunc:$Y};function cI(e,t,n){switch(n){case"reflect":return DY(e,t);case"wrap":return RY(e,t);case"nearest":return PY(e,t);case"constant":default:return MY(e,t)}}function DY(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let a=2*t;n<a&&(n=a*Math.trunc(-n/a)+n),n=n<-t?n+a:-n-1}else if(n>t-1)if(t<=1)n=0;else{let a=2*t;n-=a*Math.trunc(n/a),n>=t&&(n=a-n-1)}return v.clamp(0,n,t-1)}function RY(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let a=t-1;n+=t*(Math.trunc(-n/a)+1)}else if(n>t-1)if(t<=1)n=0;else{let a=t-1;n-=t*Math.trunc(n/a)}return v.clamp(0,n,t-1)}function MY(e,t){return e}function PY(e,t){return v.clamp(0,e,t-1)}function Lp(e,t,n,a,r,s,i,o,l,u,p){let d=i*a+o*r+l*s+u;return 0<=o&&o<t&&0<=l&&l<n?e[d]:p}function OY(e,t,n,a,r,s,i,o,l,u,p){let d=Math.round(o),c=Math.round(l);return Lp(e,t,n,a,r,s,i,d,c,u,p)}function LY(e,t,n,a,r,s,i,o,l,u,p){let d=Math.floor(o),c=Math.floor(l),h=d+1,m=c+1,f=(m-l)*Lp(e,t,n,a,r,s,i,d,c,u,p)+(l-c)*Lp(e,t,n,a,r,s,i,d,m,u,p),g=(m-l)*Lp(e,t,n,a,r,s,i,h,c,u,p)+(l-c)*Lp(e,t,n,a,r,s,i,h,m,u,p);return(h-o)*f+(o-d)*g}function zY(e){let{inputs:t,attrs:n,backend:a}=e,{axis:r}=n,{x:s}=t;ge(s,"unique");let i=a.data.get(s.dataId).values,{outputValues:o,outputShape:l,indices:u}=qC(i,r,s.shape,s.dtype);return[a.makeTensorInfo(l,s.dtype,o),a.makeTensorInfo([u.length],"int32",u)]}var WY={kernelName:Bm,backendName:"cpu",kernelFunc:zY};function BY(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r.shape.length,o=r.shape[s],l=new Array(i-1),u=0;for(let h=0;h<i;h++)h!==s&&(l[u++]=r.shape[h]);let p=new Array(i).fill(0),d=r.shape.slice();d[s]=1;let c=new Array(o);for(let h=0;h<c.length;h++){p[s]=h;let m=pi({inputs:{x:r},backend:n,attrs:{begin:p,size:d}});c[h]=ft({inputs:{x:m},backend:n,attrs:{shape:l}}),n.disposeIntermediateTensorInfo(m)}return c}var VY={kernelName:Ru,backendName:"cpu",kernelFunc:BY};function UY(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,segmentIds:s}=t,{numSegments:i}=a;ge(r,"unsortedSegmentSum");let o=r.shape.length,l=s.shape.length,u=[],p=[],d=o-l,c=s;for(let m=0;m<d;++m){let f=em({inputs:{input:c},backend:n,attrs:{dim:m+1}});c=f,p.push(f)}for(let m=0;m<i;++m){let f=v.createScalarValue(m,"int32"),g=n.makeTensorInfo([],"int32",f),y=fC({inputs:{a:g,b:c},backend:n}),b=ds({inputs:{x:y},backend:n,attrs:{dtype:"float32"}}),x=qf({inputs:{a:b,b:r},backend:n}),w=sd({inputs:{x},backend:n,attrs:{axis:0,keepDims:!1}});u.push(w),p.push(g),p.push(y),p.push(b),p.push(x),p.push(w)}let h=l_({inputs:u,backend:n,attrs:{axis:0}});return p.forEach(m=>n.disposeIntermediateTensorInfo(m)),h}var GY={kernelName:Cc,backendName:"cpu",kernelFunc:UY},HY=[zq,Nj,Bq,Uq,Fj,Hq,qq,Xq,Zq,Qq,t5,a5,s5,l5,p5,h5,f5,y5,x5,Oq,w5,I5,T5,C5,Aj,Rj,E5,Cj,$5,D5,R5,P5,L5,W5,V5,G5,j5,K5,Y5,J5,e8,n8,r8,s8,o8,u8,c8,d8,h8,m8,y8,Aq,x8,Mj,C8,Pj,_8,Lj,R8,M8,O8,Wj,W8,V8,G8,j8,K8,Vj,Gj,_j,Y8,F5,J8,eK,nK,$q,jj,Kj,rK,Yj,iK,uK,cK,mK,gK,bK,xK,Jj,wK,IK,TK,CK,EK,$K,DK,eq,MK,LK,VK,nq,rq,HK,KK,ZK,iq,QK,tX,nX,u_,iX,Dq,uq,lX,pX,dX,mX,Ej,dx,gX,Rq,Mq,Pq,bX,vX,kX,SX,NX,CX,EX,yq,$X,PX,LX,VX,xq,GX,jX,KX,vq,WK,ZX,QX,tY,aY,sY,oY,uY,cY,Iq,dY,Tq,mY,gY,bY,vY,kY,Eq,f8,SY,NY,_Y,AY,FY,oq,WY,VY,GY,eX];for(let e of HY)_c(e);var p_={};Ae(p_,{assertNotComplex:()=>Yu,bindCanvasToFramebuffer:()=>nZ,bindColorTextureToFramebuffer:()=>kh,bindTextureToProgramUniformSampler:()=>N_,bindTextureUnit:()=>I_,bindVertexBufferToProgramAttribute:()=>mx,callAndCheck:()=>me,canBeRepresented:()=>d_,createFragmentShader:()=>f_,createFramebuffer:()=>k_,createProgram:()=>g_,createStaticIndexBuffer:()=>x_,createStaticVertexBuffer:()=>b_,createTexture:()=>v_,createVertexShader:()=>m_,getBatchDim:()=>ci,getExtensionOrThrow:()=>zp,getFramebufferErrorMessage:()=>C_,getMaxTexturesInShader:()=>$_,getNumChannels:()=>eZ,getProgramUniformLocation:()=>T_,getProgramUniformLocationOrThrow:()=>S_,getRowsCols:()=>di,getShapeAs3D:()=>Bp,getTextureShapeFromLogicalShape:()=>E_,getWebGLDisjointQueryTimerVersion:()=>F_,getWebGLErrorMessage:()=>h_,getWebGLMaxTextureSize:()=>A_,hasExtension:()=>da,isCapableOfRenderingToFloatTexture:()=>D_,isDownloadFloatTextureEnabled:()=>R_,isReshapeFree:()=>sc,isWebGLFenceEnabled:()=>M_,isWebGLVersionEnabled:()=>gx,linkProgram:()=>y_,logShaderSourceAndInfoLog:()=>Y0,resetMaxTextureSize:()=>aZ,resetMaxTexturesInShader:()=>rZ,unbindColorTextureFromFramebuffer:()=>fx,unbindTextureUnit:()=>tZ,validateFramebuffer:()=>Wp,validateProgram:()=>wh,validateTextureSize:()=>w_});var Gs={},mh={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function c_(e,t){Gs[e]=t}function qa(e,t){if(!(e in Gs)||t!=null){let a=qY(e,t);if(a!==null)Gs[e]=a;else return console.log("Could not get context for WebGL version",e),null}let n=Gs[e];return n==null||n.isContextLost()?(delete Gs[e],qa(e)):(n.disable(n.DEPTH_TEST),n.disable(n.STENCIL_TEST),n.disable(n.BLEND),n.disable(n.DITHER),n.disable(n.POLYGON_OFFSET_FILL),n.disable(n.SAMPLE_COVERAGE),n.enable(n.SCISSOR_TEST),n.enable(n.CULL_FACE),n.cullFace(n.BACK),Gs[e])}function jY(e){if(typeof OffscreenCanvas!="undefined"&&e===2)return new OffscreenCanvas(300,150);if(typeof document!="undefined")return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}function qY(e,t){if(e!==1&&e!==2)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let n=t==null?jY(e):t;return n.addEventListener("webglcontextlost",a=>{a.preventDefault(),delete Gs[e]},!1),H().getBool("SOFTWARE_WEBGL_ENABLED")&&(mh.failIfMajorPerformanceCaveat=!1),e===1?n.getContext("webgl",mh)||n.getContext("experimental-webgl",mh):n.getContext("webgl2",mh)}var rc;(function(e){e[e.DENSE=0]="DENSE",e[e.SHARED_BATCH=1]="SHARED_BATCH"})(rc||(rc={}));var ca;(function(e){e[e.RENDER=0]="RENDER",e[e.UPLOAD=1]="UPLOAD",e[e.PIXELS=2]="PIXELS",e[e.DOWNLOAD=3]="DOWNLOAD"})(ca||(ca={}));var ln;(function(e){e[e.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",e[e.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",e[e.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",e[e.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",e[e.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16"})(ln||(ln={}));function id(e,t){return[t,e]}function KY(e,t){return e*t}function fh(e){let t=v.sizeFromShape(e),n=Math.ceil(t/4);return v.sizeToSquarishShape(n)}function Xu(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function XY(e,t){let[n,a]=Xu(e,t);return n*a*4}function X0(e,t){let n=e,a,r,s,i,o,l,u,p,d,c;return H().getNumber("WEBGL_VERSION")===2?(a=n.R32F,r=n.R16F,s=n.RGBA16F,i=n.RGBA32F,o=n.RED,u=4,p=1,d=n.HALF_FLOAT,c=n.FLOAT,l=n.RGBA8):(a=e.RGBA,r=e.RGBA,s=e.RGBA,i=n.RGBA,o=e.RGBA,u=4,p=4,d=t!=null?t.HALF_FLOAT_OES:null,c=e.FLOAT,l=e.RGBA),{internalFormatFloat:a,internalFormatHalfFloat:r,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:o,downloadTextureFormat:l,downloadUnpackNumChannels:u,defaultNumChannels:p,textureTypeHalfFloat:d,textureTypeFloat:c}}function me(e,t){let n=t();return H().getBool("DEBUG")&&YY(e),n}function YY(e){let t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+h_(e,t))}var ZY=596e-10,JY=65504;function d_(e){return!!(H().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||e===0||ZY<Math.abs(e)&&Math.abs(e)<JY)}function h_(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function zp(e,t){return Mr(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function m_(e,t){let n=Mr(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw console.log(e.getShaderInfoLog(n)),new Error("Failed to compile vertex shader.");return n}function f_(e,t){let n=Mr(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),H().get("ENGINE_COMPILE_ONLY"))return n;if(e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw Y0(t,e.getShaderInfoLog(n)),new Error("Failed to compile fragment shader.");return n}var QY=/ERROR: [0-9]+:([0-9]+):/g;function Y0(e,t){let n=QY.exec(t);if(n==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(e);return}let a=+n[1],r=e.split(`
-`),s=r.length.toString().length+2,i=r.map((d,c)=>v.rightPad((c+1).toString(),s)+d),o=0;for(let d=0;d<i.length;d++)o=Math.max(i[d].length,o);let l=i.slice(0,a-1),u=i.slice(a-1,a),p=i.slice(a);console.log(l.join(`
+============================`));let r={id:this.nextDataId()};return this.data.set(r,{values:e,dtype:n,refCount:1}),r}makeTensorInfo(e,t,n){let r;if(t==="string"&&n!=null&&n.length>0&&w.isString(n[0])){let s=n.map(a=>w.encodeString(a));r=this.write(s,e,t)}else r=this.write(n,e,t);return{dataId:r,shape:e,dtype:t}}refCount(e){return this.data.has(e)?this.data.get(e).refCount:0}incRef(e){let t=this.data.get(e);t.refCount++}decRef(e){if(this.data.has(e)){let t=this.data.get(e);t.refCount--}}move(e,t,n,r,s){this.data.set(e,{values:t,dtype:r,refCount:s})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){let{dtype:t,complexTensorInfos:n}=this.data.get(e);if(t==="complex64"){let r=this.readSync(n.real.dataId),s=this.readSync(n.imag.dataId);return N.mergeRealAndImagArrays(r,s)}return this.data.get(e).values}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(r=>w.decodeString(r));return Me(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Me(e.shape,e.dtype,t)}makeOutput(e,t,n){return Er().makeTensorFromTensorInfo(this.makeTensorInfo(t,n,e),this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(e);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){let t=w.now();return e(),{kernelMs:w.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){be([e],"where");let t=this.readSync(e.dataId);return Nq(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};R0.nextDataId=0;var l_={};Ae(l_,{addImpl:()=>h_,bincountImpl:()=>O0,bincountReduceImpl:()=>f_,castImpl:()=>p_,ceilImpl:()=>m_,concatImpl:()=>M0,equalImpl:()=>g_,expImpl:()=>y_,expm1Impl:()=>x_,floorImpl:()=>w_,gatherNdImpl:()=>I_,gatherV2Impl:()=>k_,greaterEqualImpl:()=>T_,greaterImpl:()=>S_,lessEqualImpl:()=>N_,lessImpl:()=>C_,linSpaceImpl:()=>__,logImpl:()=>E_,maxImpl:()=>A_,maximumImpl:()=>$_,minimumImpl:()=>D_,multiplyImpl:()=>L0,negImpl:()=>F_,notEqualImpl:()=>R_,prodImpl:()=>P_,raggedGatherImpl:()=>O_,raggedRangeImpl:()=>M_,raggedTensorToTensorImpl:()=>L_,rangeImpl:()=>B0,rsqrtImpl:()=>z_,scatterImpl:()=>cc,sigmoidImpl:()=>w5,simpleAbsImpl:()=>d_,sliceImpl:()=>Qh,sparseFillEmptyRowsImpl:()=>W_,sparseReshapeImpl:()=>V_,sparseSegmentReductionImpl:()=>W0,sqrtImpl:()=>S5,squaredDifferenceImpl:()=>U_,stridedSliceImpl:()=>G_,stringNGramsImpl:()=>V0,stringSplitImpl:()=>U0,stringToHashBucketFastImpl:()=>G0,subImpl:()=>H_,tileImpl:()=>q_,topKImpl:()=>K_,transposeImpl:()=>z0,uniqueImpl:()=>X_});function d_(e){let t=new Float32Array(e.length);for(let n=0;n<e.length;++n)t[n]=Math.abs(e[n]);return t}var _q=e=>{let{x:t}=e.inputs,n=e.backend;be(t,"abs");let r=new Float32Array(w.sizeFromShape(t.shape)),s=n.data.get(t.dataId).values;return r=d_(s),n.makeOutput(r,t.shape,t.dtype)},Eq={kernelName:Fc,backendName:"cpu",kernelFunc:_q};function Vt(e){return(t,n,r,s,a)=>{let o=N.assertAndGetBroadcastShape(t,n),i=o.length,c=w.computeStrides(o),u=w.sizeFromShape(o),l=w.getTypedArrayFromDType(a,u),p=t.length,d=n.length,h=w.computeStrides(t),f=w.computeStrides(n),m=N.getBroadcastDims(t,o),g=N.getBroadcastDims(n,o);if(m.length+g.length===0)for(let b=0;b<l.length;++b)l[b]=e(r[b%r.length],s[b%s.length]);else for(let b=0;b<l.length;++b){let y=w.indexToLoc(b,i,c),v=y.slice(-p);m.forEach(C=>v[C]=0);let x=w.locToIndex(v,p,h),k=y.slice(-d);g.forEach(C=>k[C]=0);let S=w.locToIndex(k,d,f);l[b]=e(r[x],s[S])}return[l,o]}}function Jn(e){let{inputs:t,backend:n}=e,{real:r,imag:s}=t,a=n.data.get(r.dataId).values,o=n.data.get(s.dataId).values,i=n.makeTensorInfo(r.shape,"complex64"),c=n.data.get(i.dataId);return c.complexTensorInfos={real:n.makeTensorInfo(r.shape,"float32",a),imag:n.makeTensorInfo(s.shape,"float32",o)},i}var Aq={kernelName:gf,backendName:"cpu",kernelFunc:Jn};function Jh(e,t,n="float32"){if(n==="complex64"){let s=Jh(e,t,"float32"),a=Jh(e,t,"float32");return Jn({inputs:{real:s,imag:a},backend:e})}let r=w.makeZerosTypedArray(w.sizeFromShape(t),n);return e.makeTensorInfo(t,n,r)}function ds(e){let{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var $q={kernelName:Oo,backendName:"cpu",kernelFunc:ds};function lo(e){let{inputs:t,backend:n}=e,{input:r}=t,s=n.data.get(r.dataId).complexTensorInfos.real,a=n.data.get(s.dataId).values;return n.makeTensorInfo(s.shape,s.dtype,a)}var Dq={kernelName:Lf,backendName:"cpu",kernelFunc:lo};function p_(e,t,n,r){if(r==="int32"){let s=Int32Array.from(e);return[t,"int32",s]}if(r==="bool"){let s=w.toTypedArray([0],n),[a,o]=Vt((i,c)=>i!==c?1:0)(t,[],e,s,"bool");return[o,"bool",a]}throw new Error(`Error in Cast: failed to cast ${n} to ${r}`)}function ha(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{dtype:a}=r;if(a==="complex64"){if(s.dtype==="complex64")return ds({inputs:{x:s},backend:n});let l=Jh(n,s.shape,s.dtype),p=ha({inputs:{x:s},backend:n,attrs:{dtype:"float32"}}),d=Jn({inputs:{real:p,imag:l},backend:n});return n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(p),d}if(s.dtype==="complex64"){let l=lo({inputs:{input:s},backend:n}),p=ha({inputs:{x:l},backend:n,attrs:{dtype:a}});return n.disposeIntermediateTensorInfo(l),p}if(!w.hasEncodingLoss(s.dtype,a)){let l=ds({inputs:{x:s},backend:n});return{dataId:l.dataId,shape:l.shape,dtype:a}}let o=n.data.get(s.dataId).values,[i,c,u]=p_(o,s.shape,s.dtype,a);return n.makeTensorInfo(i,c,u)}var Fq={kernelName:wo,backendName:"cpu",kernelFunc:ha};function nn(e,t,n,r){return n==null?({inputs:s,backend:a})=>{let{a:o,b:i}=s,c=a;be([o,i],e);let u=c.data.get(o.dataId).values,l=c.data.get(i.dataId).values,p=o.dtype==="string"?N.fromUint8ToStringArray(u):u,d=o.dtype==="string"?N.fromUint8ToStringArray(l):l,h=r||o.dtype,[f,m]=t(o.shape,i.shape,p,d,h);return c.makeTensorInfo(m,h,f)}:({inputs:s,backend:a})=>{let{a:o,b:i}=s,c=a;if(o.dtype==="complex64"||i.dtype==="complex64"){let u=ha({inputs:{x:o},backend:c,attrs:{dtype:"complex64"}}),l=c.data.get(u.dataId),p=l.complexTensorInfos.real,d=l.complexTensorInfos.imag,h=c.data.get(p.dataId).values,f=c.data.get(d.dataId).values,m=ha({inputs:{x:i},backend:c,attrs:{dtype:"complex64"}}),g=c.data.get(m.dataId),b=g.complexTensorInfos.real,y=g.complexTensorInfos.imag,v=c.data.get(b.dataId).values,x=c.data.get(y.dataId).values,[k,S,C]=n(o.shape,i.shape,h,f,v,x),E=c.makeTensorInfo(C,"float32",k),$=c.makeTensorInfo(C,"float32",S),F=Jn({inputs:{real:E,imag:$},backend:c});return c.disposeIntermediateTensorInfo(u),c.disposeIntermediateTensorInfo(m),c.disposeIntermediateTensorInfo(E),c.disposeIntermediateTensorInfo($),F}else{let u=c.data.get(o.dataId).values,l=c.data.get(i.dataId).values,p=r||o.dtype,[d,h]=t(o.shape,i.shape,u,l,p);return c.makeTensorInfo(h,p,d)}}}function P0(e){return(t,n,r,s,a,o)=>{let i=N.assertAndGetBroadcastShape(t,n),c=w.sizeFromShape(i),u=i.length,l=w.computeStrides(i),p=w.getTypedArrayFromDType("float32",c),d=w.getTypedArrayFromDType("float32",c),h=N.getBroadcastDims(t,i),f=N.getBroadcastDims(n,i),m=N.mergeRealAndImagArrays(r,s),g=N.mergeRealAndImagArrays(a,o),b=t.length,y=w.computeStrides(t),v=n.length,x=w.computeStrides(n);if(h.length+f.length===0)for(let k=0;k<p.length;k++){let S=k%m.length,C=k%g.length,E=e(m[S*2],m[S*2+1],g[C*2],g[C*2+1]);p[k]=E.real,d[k]=E.imag}else for(let k=0;k<p.length;k++){let S=w.indexToLoc(k,u,l),C=S.slice(-b);h.forEach(R=>C[R]=0);let E=w.locToIndex(C,b,y),$=S.slice(-v);f.forEach(R=>$[R]=0);let F=w.locToIndex($,v,x),A=e(m[E*2],m[E*2+1],g[F*2],g[F*2+1]);p[k]=A.real,d[k]=A.imag}return[p,d,i]}}var h_=Vt((e,t)=>e+t),Rq=P0((e,t,n,r)=>({real:e+n,imag:t+r})),_c=nn(ba,h_,Rq),Pq={kernelName:ba,backendName:"cpu",kernelFunc:_c};function O0(e,t,n,r,s){let a=w.sizeFromShape(r),o=w.makeZerosTypedArray(s,n);for(let i=0;i<e.length;i++){let c=e[i];if(c<0)throw new Error("Input x must be non-negative!");c>=s||(a>0?o[c]+=t[i]:o[c]+=1)}return o}function f_(e,t,n,r=!1){let s=e.shape[0],a=e.shape[1],o=Me([s,n],t.dtype);for(let i=0;i<s;i++)for(let c=0;c<a;c++){let u=e.get(i,c);if(u<0)throw new Error("Input x must be non-negative!");u>=n||(r?o.set(1,i,u):t.size>0?o.set(o.get(i,u)+t.get(i,c),i,u):o.set(o.get(i,u)+1,i,u))}return o}function _a(e){return(t,n,r)=>{let s=w.getTypedArrayFromDType(n,t.length);for(let a=0;a<t.length;++a)s[a]=e(t[a],r);return s}}function at(e,t,n){return({inputs:r,attrs:s,backend:a})=>{let{x:o}=r;if(be(o,e),o.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let i=a,c=i.data.get(o.dataId).values,u=w.sizeFromShape(o.shape),l=n||o.dtype,p=w.getArrayFromDType(l,u);for(let d=0;d<u;++d)p[d]=t(c[d],s);return i.makeTensorInfo(o.shape,l,p)}}function Ku(e,t,n){return({inputs:r,attrs:s,backend:a})=>{let{x:o}=r;if(be(o,e),o.dtype==="string"||n==="string")throw new Error("unaryKernelFunc does not support string input/output");let i=a,c=i.data.get(o.dataId).values,u=n||o.dtype,l=t(c,u,s);return i.makeTensorInfo(o.shape,u,l)}}var m_=_a(e=>Math.ceil(e)),Oq=Ku(Io,m_),Mq={kernelName:Io,backendName:"cpu",kernelFunc:Oq};function M0(e,t,n,r){let s=w.getArrayFromDType(n,w.sizeFromShape(t));if(r&&n!=="string"){let a=0;e.forEach(o=>{let i=w.sizeFromShape(o.shape);s.set(o.vals,a),a+=i})}else{let a=0;e.forEach(o=>{let i=n==="string"?N.fromUint8ToStringArray(o.vals):o.vals,c=0;for(let u=0;u<o.shape[0];++u){let l=u*t[1]+a;for(let p=0;p<o.shape[1];++p)s[l+p]=i[c++]}a+=o.shape[1]})}return s}var g_=Vt((e,t)=>e===t?1:0),b_=nn(Xc,g_,null,"bool"),Lq={kernelName:Xc,backendName:"cpu",kernelFunc:b_},y_=_a(e=>Math.exp(e)),v_=Ku($o,y_,"float32"),zq={kernelName:$o,backendName:"cpu",kernelFunc:v_},x_=_a(e=>Math.expm1(e)),Bq=Ku(Zc,x_),Wq={kernelName:Zc,backendName:"cpu",kernelFunc:Bq},w_=_a(e=>Math.floor(e)),Vq=Ku(Do,w_),Uq={kernelName:Do,backendName:"cpu",kernelFunc:Vq};function I_(e,t,n,r,s,a,o,i,c){let u=Me([r,a],n);for(let l=0;l<r;l++){let p=[],d=0;for(let h=0;h<s;h++){let f=e[l*s+h];d+=f*o[h],p.push(f)}if(d<0||d>=c/a)throw new Error(`Invalid indices: ${p} does not index into ${i}`);for(let h=0;h<a;h++)u.values[l*a+h]=t.get(...t.indexToLoc(d*a+h))}return u}function k_(e,t,n){let r=Me(n,e.dtype);for(let s=0;s<r.size;++s){let o=r.indexToLoc(s).slice(),i=o[0],c=o[2],u=t.locToIndex([i,c]);o[2]=t.values[u];let l=e.locToIndex(o);0<=l&&l<e.values.length&&(r.values[s]=e.values[l])}return r}var S_=Vt((e,t)=>e>t?1:0),Gq=nn(tu,S_,null,"bool"),Hq={kernelName:tu,backendName:"cpu",kernelFunc:Gq},T_=Vt((e,t)=>e>=t?1:0),qq=nn(Po,T_,null,"bool"),jq={kernelName:Po,backendName:"cpu",kernelFunc:qq},C_=Vt((e,t)=>e<t?1:0),Kq=nn(au,C_,null,"bool"),Xq={kernelName:au,backendName:"cpu",kernelFunc:Kq},N_=Vt((e,t)=>e<=t?1:0),Yq=nn(ou,N_,null,"bool"),Zq={kernelName:ou,backendName:"cpu",kernelFunc:Yq};function __(e,t,n){let r=(t-e)/(n-1),s=w.makeZerosTypedArray(n,"float32");s[0]=e;for(let a=1;a<s.length;a++)s[a]=s[a-1]+r;return s}var E_=_a(e=>Math.log(e)),Jq=Ku(Lo,E_),Qq={kernelName:Lo,backendName:"cpu",kernelFunc:Jq};function A_(e,t,n,r){let s=w.getTypedArrayFromDType(r,w.sizeFromShape(n));for(let a=0;a<s.length;++a){let o=a*t,i=e[o];for(let c=0;c<t;++c){let u=e[o+c];(Number.isNaN(u)||u>i)&&(i=u)}s[a]=i}return s}var $_=Vt((e,t)=>Math.max(e,t)),e5=nn(Bo,$_),t5={kernelName:Bo,backendName:"cpu",kernelFunc:e5},D_=Vt((e,t)=>Math.min(e,t)),n5=nn(Go,D_),r5={kernelName:Go,backendName:"cpu",kernelFunc:n5},L0=Vt((e,t)=>e*t),s5=P0((e,t,n,r)=>({real:e*n-t*r,imag:e*r+t*n})),Km=nn(qo,L0,s5),a5={kernelName:qo,backendName:"cpu",kernelFunc:Km};function F_(e,t,n){let r=w.createScalarValue(-1,n);return L0([],t,r,e,n)}function o5(e){let{inputs:t,backend:n}=e,{x:r}=t;be(r,"neg");let s=n.data.get(r.dataId).values,[a,o]=F_(s,r.shape,r.dtype);return n.makeTensorInfo(o,r.dtype,a)}var i5={kernelName:pu,backendName:"cpu",kernelFunc:o5},R_=Vt((e,t)=>e!==t?1:0),c5=nn(hu,R_,null,"bool"),u5={kernelName:hu,backendName:"cpu",kernelFunc:c5};function z0(e,t,n,r,s){let a=t.length,o=w.sizeFromShape(t),i=w.computeStrides(t),c=w.computeStrides(s),u=w.getTypedArrayFromDType(n,w.sizeFromShape(s));for(let l=0;l<o;++l){let p=w.indexToLoc(l,a,i),d=new Array(p.length);for(let f=0;f<d.length;f++)d[f]=p[r[f]];let h=w.locToIndex(d,a,c);u[h]=e[l]}return u}function Vn(e){let{inputs:t,attrs:n,backend:r}=e,{x:s}=t,{perm:a}=n;be(s,"transpose");let o=s.shape.length,i=new Array(o);for(let p=0;p<i.length;p++)i[p]=s.shape[a[p]];let c=r.data.get(s.dataId).values,u=z0(c,s.shape,s.dtype,a,i);return{dataId:r.write(u,i,s.dtype),shape:i,dtype:s.dtype}}var l5={kernelName:_s,backendName:"cpu",kernelFunc:Vn};function P_(e,t,n,r){let[s,a]=N.computeOutAndReduceShapes(e,r),o=hr(t,"int32"),i=w.makeZerosTypedArray(w.sizeFromShape(s),o),c=w.sizeFromShape(a);for(let u=0;u<i.length;++u){let l=u*c,p=1;for(let d=0;d<c;++d)p*=n[l+d];i[u]=p}return{outVals:i,outShape:s,outDtype:o}}function d5(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;be(s,"prod");let i=s.shape.length,c=w.parseAxisParam(a,s.shape),u=N.getAxesPermutation(c,i),l=c,p=s,d=[];u!=null&&(p=Vn({inputs:{x:s},backend:n,attrs:{perm:u}}),d.push(p),l=N.getInnerMostAxes(l.length,i));let h=n.data.get(p.dataId).values,{outVals:f,outShape:m,outDtype:g}=P_(p.shape,p.dtype,h,l),b=m;return o&&(b=N.expandShapeToKeepDim(m,c)),d.forEach(y=>n.disposeIntermediateTensorInfo(y)),n.makeTensorInfo(b,g,f)}var p5={kernelName:Zo,backendName:"cpu",kernelFunc:d5};function h5(e,t,n){e.forEach((r,s)=>{if(r<0||r>=n){let a=w.indexToLoc(s,t.length,w.computeStrides(t)).join(",");throw new Error(`indices[${a}] = ${r} is not in [0, ${n})`)}})}function f5(e,t){for(let n=0;n<e.length;++n){let r=e[n],s=n===e.length-1?t:e[n+1].length;if(r.length===0)throw new Error("Ragged splits may not be empty");if(r[0]<0)throw new Error("Ragged splits must be non-negative");if(r[r.length-1]>s)throw new Error("Ragged splits must not point past values");for(let a=1;a<r.length;++a)if(r[a-1]>r[a])throw new Error("Ragged splits must be sorted in ascending order")}}function m5(e,t,n,r){let s=[],a=0,o=t.length-1+n.length,i=new Array(o).fill(null).map(()=>[0]);f5(n,r);let c=1;for(let u=0;u<t.length-1;++u){c*=t[u];let l=t[u+1];for(let p=1;p<c+1;++p)i[u].push(p*l)}for(let u=0;u<e.length;++u){let l=e[u],p=e[u]+1;for(let d=0;d<n.length;++d){let h=n[d],f=d+t.length-1;if(f>=0){let m=i[f],g=m[m.length-1]-h[l];for(let b=l;b<p;++b)i[f].push(h[b+1]+g)}l=h[l],p=h[p]}p!==l&&(s.push([l,p]),a+=p-l)}return{outSplits:i,valueSlices:s,numValues:a}}function g5(e){let t=[];for(let n=0;n<e.length;++n){let r=e[n].length,s=w.getArrayFromDType("int32",r);t.push(s),e[n].forEach((a,o)=>s[o]=a)}return t}function i1(e,t){let n=e.slice(0,t);for(;n.length<t;)n.push(1);for(let r=t;r<e.length;r++)n[t-1]*=e[r];return n}function b5(e,t,n,r,s,a){let o=i1(t,2)[1],i=i1(a,2)[1],c=0;for(let u of n)for(let l=u[0];l<u[1];++l){for(let p=0;p<r;++p)s[c*i+p]=e[l*o+p];++c}}function y5(e,t,n,r,s){let a=t.slice();a[0]=s;let o=w.getArrayFromDType(n,w.sizeFromShape(a)),i=e.length,c=i===0?0:i/t[0];return b5(e,t,r,c,o,a),[o,a]}function O_(e,t,n,r,s,a,o,i){if(e.length===0)throw new Error("paramsNestedSplits must be non empty");if(t[0].length===0)throw new Error("Split tensors must not be scalars");let c=t[0][0]-1;if(h5(a,o,c),r.length===0)throw new Error("params.rank must be nonzero");let u=r[0],{outSplits:l,valueSlices:p,numValues:d}=m5(a,o,e,u),h=g5(l),f=y5(n,r,s,p,d);return[h,f[0],f[1]]}var c1=2147483647;function M_(e,t,n,r,s,a,o){if(t.length>1)throw new Error("starts must be a scalar or vector");if(s.length>1)throw new Error("limits must be a scalar or vector");if(o.length>1)throw new Error("deltas must be a scalar or vector");let i=t.length===0,c=s.length===0,u=o.length===0,l=[];i||l.push(t[0]),c||l.push(s[0]),u||l.push(o[0]);for(let g=1;g<l.length;++g)if(l[g]!==l[g-1])throw new Error("starts, limits, and deltas must have the same shape");let p=l.length===0?1:l[0],d=w.getArrayFromDType("int32",p+1);d[0]=0;for(let g=0;g<p;++g){let b=i?e[0]:e[g],y=c?r[0]:r[g],v=u?a[0]:a[g];if(v===0)throw new Error("Requires delta != 0");let x;if(v>0&&y<b||v<0&&y>b)x=0;else if(x=Math.ceil(Math.abs((y-b)/v)),x>c1)throw new Error(`Requires ((limit - start) / delta) <= ${c1}`);d[g+1]=d[g]+x}let h=d[p],f=w.getArrayFromDType(n,h),m=0;for(let g=0;g<p;++g){let b=d[g+1]-d[g],y=i?e[0]:e[g],v=u?a[0]:a[g];for(let x=0;x<b;++x)f[m++]=y,y+=v}return[d,f]}var Cr=N.RowPartitionType,dv=class{constructor(e,t,n,r,s,a,o,i,c,u){this.shape=e,this.shapeShape=t,this.values=n,this.valuesShape=r,this.valuesDType=s,this.defaultValue=a,this.defaultValueShape=o,this.rowPartitionValues=i,this.rowPartitionValuesShapes=c,this.rowPartitionTypes=N.getRowPartitionTypesHelper(u),this.raggedRank=N.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(e){return this.rowPartitionTypes[0]===Cr.FIRST_DIM_SIZE?this.rowPartitionTypes[e+1]:this.rowPartitionTypes[e]}getRowPartitionTensor(e){return this.rowPartitionTypes[0]===Cr.FIRST_DIM_SIZE?this.rowPartitionValues[e+1]:this.rowPartitionValues[e]}getMaxWidth(e){let t=this.getRowPartitionTensor(e-1);switch(this.getRowPartitionTypeByDimension(e-1)){case Cr.VALUE_ROWIDS:return dv.getMaxWidthValueRowID(t);case Cr.ROW_SPLITS:return dv.getMaxWidthRowSplit(t);default:throw new Error(`Cannot handle partition type ${Cr[this.getRowPartitionTypeByDimension(e-1)]}`)}}static getMaxWidthRowSplit(e){let t=e.length;if(t===0||t===1)return 0;let n=0;for(let r=0;r<t-1;++r){let s=e[r+1]-e[r];s>n&&(n=s)}return n}static getMaxWidthValueRowID(e){let t=e.length;if(t===0)return 0;let n=0,r=e[0],s=0;for(let a=1;a<t;++a){let o=e[a];o!==r&&(r=o,s=Math.max(a-n,s),n=a)}return Math.max(t-n,s)}tensorShapeFromTensor(e,t,n=!0){if(t.length===0){if(e[0]===-1)return[];throw new Error("The only valid scalar shape tensor is the fully unknown shape specified as -1.")}return l1(e,n)}calculateOutputSize(e){let t=this.valuesShape,n=this.defaultValueShape;N.validateDefaultValueShape(n,t);let r=this.tensorShapeFromTensor(this.shape,this.shapeShape),a=N.combineRaggedTensorToTensorShapes(this.raggedRank,r,t);a[0]<0&&(a[0]=e);for(let o=1;o<=this.raggedRank;++o)a[o]<0&&(a[o]=this.getMaxWidth(o));return a}calculateFirstParentOutputIndex(e,t,n){let r=Math.min(e,n),s=[],a=0;for(let o=0;o<r;++o,a+=t)s.push(a);for(let o=r;o<e;++o)s.push(-1);return w.assert(s.length===e,()=>"Final length of result must be equal to firstDimension."),s}calculateOutputIndexRowSplit(e,t,n,r){let s=e.length,a=[];for(let o=0;o<s-1;++o){let i=e[o+1]-e[o],c=Math.min(r,i),u=t[o];u===-1&&(c=0);for(let l=0;l<c;++l)a.push(u),u+=n;for(let l=0;l<i-c;++l)a.push(-1)}if(s>0&&a.length!==e[s-1])throw new Error("Invalid row split size.");return a}calculateOutputIndexValueRowID(e,t,n,r){let s=e.length,a=[];if(s===0)return[];let o=0,i=e[0];if(i>=t.length)throw new Error(`Got currentValueRowId=${i}, which is not less than ${t.length}`);let c=t[i];a.push(c);for(let u=1;u<s;++u){let l=e[u];if(l===i)c>=0&&(++o,o<r?c+=n:c=-1);else{if(o=0,i=l,l>=t.length)throw new Error(`Got nextValueRowId=${l} which is not less than ${t.length}`);c=t[l]}a.push(c)}if(a.length!==e.length)throw new Error("Invalid row ids.");return a}calculateOutputIndex(e,t,n,r){let s=this.getRowPartitionTensor(e),a=this.getRowPartitionTypeByDimension(e);switch(a){case Cr.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(s,t,n,r);case Cr.ROW_SPLITS:if(s.length-1>t.length)throw new Error(`Row partition size is greater than output size: ${s.length-1} > ${t.length}`);return this.calculateOutputIndexRowSplit(s,t,n,r);default:throw new Error(`Unsupported partition type: ${Cr[a]}`)}}getFirstDimensionSize(){let e=this.rowPartitionValues[0];if(this.rowPartitionTypes.length===0)throw new Error("No row_partition_types given.");let t=this.rowPartitionTypes[0];switch(t){case Cr.FIRST_DIM_SIZE:return e[0];case Cr.VALUE_ROWIDS:throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");case Cr.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${Cr[t]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error("Invalid first partition input. Tensor requires at least one element.");let t=this.getFirstDimensionSize(),n=this.calculateOutputSize(t),r=new Array(this.raggedRank+1);r[r.length-1]=1;for(let i=r.length-2;i>=0;--i)r[i]=r[i+1]*n[i+1];let s=l1(n,!1),a=w.getArrayFromDType(this.valuesDType,w.sizeFromShape(s));if(r[0]*n[0]>0){let i=this.calculateFirstParentOutputIndex(t,r[0],n[0]);for(let c=1;c<=this.raggedRank;++c)i=this.calculateOutputIndex(c-1,i,r[c],n[c]);this.setOutput(this.raggedRank,i,a,s)}return[s,a]}setOutput(e,t,n,r){if(n.length===0)return;let s=this.values,a=n,o=r.slice();o=o.slice(e+1);let i=w.sizeFromShape(o),c=t.length,u=this.defaultValue;if(u.length!==i&&u.length!==1){let h=this.defaultValueShape;O(()=>{let f=W(u,h);u=Ya(f,o).dataSync()})}let l=0,p=0,d=0;for(let h=0;h<=c;++h){let f=h<c?t[h]:-1;if(f===d){++d;continue}if(p<d){let m=s.subarray(l*i),g=a.subarray(p*i),b=(d-p)*i;u1(g,m,b)}if(h>=c){let m=n.length;f=Math.floor(m/i)}if(f>d)if(this.defaultValue.length===1)a.subarray(d*i,f*i).fill(this.defaultValue[0]),d=f;else for(;f>d;){let m=a.slice(d*i);u1(m,u,i),++d}f<0?(l=h+1,p=d):(l=h,p=d,d=p+1)}}};function u1(e,t,n){for(let r=0;r<n;r++)e[r]=t[r]}function l1(e,t){let n=[];for(let r of e){if(r<0){if(!t)throw new Error(`Dimension ${r} must be >= 0`);if(r<-1)throw new Error(`Dimension ${r} must be >= -1`);r=-1}n.push(r)}return n}function L_(e,t,n,r,s,a,o,i,c,u){return new dv(e,t,n,r,s,a,o,i,c,u).compute()}function B0(e,t,n,r){let s=e===t,a=e<t&&n<0,o=t<e&&n>1;if(s||a||o)return w.makeZerosTypedArray(0,r);let i=Math.abs(Math.ceil((t-e)/n)),c=w.makeZerosTypedArray(i,r);t<e&&n===1&&(n=-1),c[0]=e;for(let u=1;u<c.length;u++)c[u]=c[u-1]+n;return c}var z_=_a(e=>1/Math.sqrt(e)),v5=Ku(si,z_),x5={kernelName:si,backendName:"cpu",kernelFunc:v5};function cc(e,t,n,r,s,a,o,i,c,u){let l=[r/s,s],p=e.values,d=t.values;if(r===0)return Me(n,t.dtype);let h=Me(l,t.dtype);typeof c=="string"||typeof c=="number"?h.values.fill(c):typeof c=="boolean"&&h.values.fill(+c);for(let f=0;f<a;f++){let m=[],g=0;for(let b=0;b<o;b++){let y=p[f*o+b];m.push(y),g+=y*i[b]}if(g<0||g>=r/s)throw new Error(`Invalid indices: ${m} does not index into ${n}`);for(let b=0;b<s;b++)u?h.values[g*s+b]+=d[f*s+b]:h.values[g*s+b]=t.rank===0?d[0]:d[f*s+b]}return h}var w5=_a(e=>1/(1+Math.exp(-e))),B_=at(oi,e=>1/(1+Math.exp(-e))),I5={kernelName:oi,backendName:"cpu",kernelFunc:B_};function Qh(e,t,n,r,s){let a=qt.isSliceContinous(r,t,n),o=w.sizeFromShape(n),i=w.computeStrides(r);if(a){let p=qt.computeFlatOffset(t,i);return s==="string"?e.slice(p,p+o):e.subarray(p,p+o)}let c=s==="string"?N.fromUint8ToStringArray(e):e,u=Me(r,s,c),l=Me(n,s);for(let p=0;p<l.size;++p){let d=l.indexToLoc(p),h=d.map((f,m)=>f+t[m]);l.set(u.get(...h),...d)}return s==="string"?N.fromStringArrayToUint8(l.values):l.values}function po(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{begin:a,size:o}=r;be(s,"slice");let[i,c]=qt.parseSliceParams(s,a,o);qt.assertParamsValid(s,i,c);let u=n.data.get(s.dataId).values,l=Qh(u,i,c,s.shape,s.dtype);return n.makeTensorInfo(c,s.dtype,l)}var k5={kernelName:Su,backendName:"cpu",kernelFunc:po};function W_(e,t,n,r,s,a,o){let i=t[0],c=a[0],u=new Array(c),l=new Array(i),p=t[1];if(c===0){if(i!==0)throw new Error(N.getSparseFillEmptyRowsIndicesDenseShapeMismatch(i));let g=w.getArrayFromDType(n,0),b=w.getArrayFromDType(s,0);return[g,[0,p],b,u,l]}let d=!0,h=0,f=new Array(c).fill(0);for(let g=0;g<i;++g){let b=e[g*p];if(b<0)throw new Error(N.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,b));if(b>=c)throw new Error(N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,b,c));++f[b],d=d&&b>=h,h=b}let m=!0;for(let g=0;g<c;++g){let b=f[g]===0;u[g]=b,m=m&&!b,f[g]=Math.max(f[g],1),g>0&&(f[g]+=f[g-1])}if(m&&d){let g=e,b=r;for(let y=0;y<i;++y)l[y]=y;return[g,[i,p],b,u,l]}else{let g=f[c-1],b=w.getArrayFromDType(n,g*p),y=w.getArrayFromDType(s,g),v=new Array(c).fill(0);for(let x=0;x<i;++x){let k=e[x*p],S=v[k],C=(k===0?0:f[k-1])+S;v[k]++;for(let E=0;E<p;++E)b[C*p+E]=e[x*p+E];y[C]=r[x],l[x]=C}for(let x=0;x<c;++x)if(v[x]===0){let S=x===0?0:f[x-1];b[S*p+0]=x;for(let C=1;C<p;++C)b[S*p+C]=0;y[S]=o}return[b,[g,p],y,u,l]}}function V_(e,t,n,r,s){let a=w.sizeFromShape(r),o=t[0],i=s.length,c=[],u=1,l=-1;for(let g=0;g<i;++g){let b=s[g];if(b===-1){if(l!==-1)throw new Error(N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(l,g));l=g,c.push(1)}else{if(b<0)throw new Error(N.getSparseReshapeNegativeOutputDimErrorMessage(g,b));u*=b,c.push(b)}}if(l!==-1){if(u<=0)throw new Error(N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let g=Math.trunc(a/u);if(u*g!==a)throw new Error(N.getSparseReshapeInputOutputMultipleErrorMessage(r,c));c[l]=g}if(w.sizeFromShape(c)!==a)throw new Error(N.getSparseReshapeInputOutputMismatchErrorMessage(r,c));let d=r.length,h=[];if(d>0){h[d-1]=1;for(let g=d-2;g>=0;--g)h[g]=h[g+1]*r[g+1]}let f=[];if(i>0){f[i-1]=1;for(let g=i-2;g>=0;--g)f[g]=f[g+1]*c[g+1]}let m=w.getArrayFromDType(n,o*i);for(let g=0;g<o;++g){let b=0;for(let y=0;y<d;++y)b+=e[g*d+y]*h[y];for(let y=0;y<i;++y)m[g*i+y]=Math.trunc(b/f[y]),b%=f[y]}return[m,[o,i],c]}function W0(e,t,n,r,s,a=!1,o=0){let i=r.length,c=[t[0],e.length/t[0]],u=c[1],p=i>0?s[i-1]+1:0;if(p<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let d=t.slice();d[0]=p;let h=d.reduce((v,x)=>v*x,1),f=w.getArrayFromDType(n,h);if(i===0)return p>0&&f.fill(o),[f,d];if(p<=0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let m=0,g=1,b=0,y=s[m];for(;;){let v=0;if(g<i){if(v=s[g],y===v){++g;continue}if(y>=v)throw new Error(N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=p)throw new Error(N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,p));y>b&&f.fill(o,b*u,y*u);for(let x=m;x<g;++x){let k=r[x];if(k<0||k>=c[0])throw new Error(N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(x,r[x],c[0]));for(let S=0;S<u;S++)f[y*u+S]+=e[k*u+S]}if(a)for(let x=0;x<u;x++)f[y*u+x]/=g-m;if(m=g,++g,b=y+1,y=v,g>i)break}return b<p&&f.fill(o,b*u,p*u),[f,d]}var S5=_a(e=>Math.sqrt(e)),T5=at(ii,e=>Math.sqrt(e)),C5={kernelName:ii,backendName:"cpu",kernelFunc:T5},U_=Vt((e,t)=>{let n=e-t;return n*n}),N5=nn(li,U_),_5={kernelName:li,backendName:"cpu",kernelFunc:N5};function G_(e,t,n,r){let s=Me(e,t.dtype);for(let a=0;a<s.size;a++){let o=s.indexToLoc(a),i=new Array(o.length);for(let c=0;c<i.length;c++)i[c]=o[c]*n[c]+r[c];s.set(t.get(...i),...o)}return s}var E5=class{constructor(e,t,n,r,s,a){this.separator=w.encodeString(e),this.nGramWidths=t,this.leftPad=w.encodeString(n),this.rightPad=w.encodeString(r),this.padWidth=s,this.preserveShort=a}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){let n=this.getPadWidth(t);return Math.max(0,e+2*n-t+1)}createNGrams(e,t,n,r,s,a){for(let o=0;o<s;++o){let i=this.getPadWidth(a),c=Math.max(0,i-o),u=Math.max(0,i-(s-(o+1))),l=a-(c+u),p=t+(c>0?0:o-i),d=0;d+=c*this.leftPad.length;for(let b=0;b<l;++b)d+=e[p+b].length;d+=u*this.rightPad.length;let h=c+u+l-1;d+=h*this.separator.length,n[r+o]=new Uint8Array(d);let f=n[r+o],m=0,g=b=>b.forEach(y=>f[m++]=y);for(let b=0;b<c;++b)g(this.leftPad),g(this.separator);for(let b=0;b<l-1;++b)g(e[p+b]),g(this.separator);if(l>0){g(e[p+l-1]);for(let b=0;b<u;++b)g(this.separator),g(this.rightPad)}else{for(let b=0;b<u-1;++b)g(this.rightPad),g(this.separator);g(this.rightPad)}}}compute(e,t){let n=e.length,r=t.length;if(r>0){let i=t[0];if(i!==0)throw new Error(`First split value must be 0, got ${i}`);for(let c=1;c<r;++c){let u=t[c]>=i;if(u=u&&t[c]<=n,!u)throw new Error(`Invalid split value ${t[c]}, must be in [${i}, ${n}]`);i=t[c]}if(i!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${i}`)}let s=r-1,a=w.getArrayFromDType("int32",r);if(n===0||r===0){let i=new Array(n);for(let c=0;c<=s;++c)a[c]=0;return[i,a]}a[0]=0;for(let i=1;i<=s;++i){let c=t[i]-t[i-1],u=0;this.nGramWidths.forEach(l=>{u+=this.getNumNGrams(c,l)}),this.preserveShort&&c>0&&u===0&&(u=1),a[i]=a[i-1]+u}let o=new Array(a[s]);for(let i=0;i<s;++i){let c=t[i],u=a[i];if(this.nGramWidths.forEach(l=>{let p=t[i+1]-t[i],d=this.getNumNGrams(p,l);this.createNGrams(e,c,o,u,d,l),u+=d}),this.preserveShort&&u===a[i]){let l=t[i+1]-t[i];if(l===0)continue;let p=l+2*this.padWidth,d=1;this.createNGrams(e,c,o,u,d,p)}}return[o,a]}};function V0(e,t,n,r,s,a,o,i){return new E5(n,r,s,a,o,i).compute(e,t)}function A5(e,t,n,r){if(!e.length)return;if(t.length===0){for(let a=0;a<e.length;++a)r.push(e.subarray(a,a+1));return}if(t.length===1){let a=t[0],o=e.indexOf(a);for(;o!==-1;){let i=e.subarray(0,o);(!n||i.length!==0)&&r.push(i),e=e.subarray(o+1),o=e.indexOf(a)}(!n||e.length!==0)&&r.push(e);return}let s=0;for(let a=0;a<e.length+1;a++)if(a===e.length||t.indexOf(e[a])!==-1){let o=e.subarray(s,a);(!n||o.length!==0)&&r.push(o),s=a+1}}function U0(e,t,n){let r=e.length,s=[],a=0,o=0,i=new Array(r);for(let d=0;d<r;++d){let h=s.length;A5(e[d],t,n,s);let f=s.length-h;i[d]=f,a+=f,o=Math.max(o,f)}let c=w.getArrayFromDType("int32",a*2),u=new Array(a),l=[r,o],p=0;for(let d=0;d<r;++d)for(let h=0;h<i[d];++h)c[p*2]=d,c[p*2+1]=h,u[p]=s[p],++p;return[c,u,l]}function G0(e,t){let n=w.getArrayFromDType("int32",e.length);for(let r=0;r<e.length;++r)n[r]=w.fingerPrint64(e[r]).modulo(t).getLowBitsUnsigned();return n}var H_=Vt((e,t)=>e-t),$5=P0((e,t,n,r)=>({real:e-n,imag:t-r})),H0=nn(di,H_,$5),D5={kernelName:di,backendName:"cpu",kernelFunc:H0};function q_(e,t){let n=new Array(e.rank);for(let s=0;s<n.length;s++)n[s]=e.shape[s]*t[s];let r=Me(n,e.dtype);for(let s=0;s<r.values.length;++s){let a=r.indexToLoc(s),o=new Array(e.rank);for(let c=0;c<o.length;c++)o[c]=a[c]%e.shape[c];let i=e.locToIndex(o);r.values[s]=e.values[i]}return r}var Ml=(e,t)=>{let n=t.value-e.value;return n===0?e.index-t.index:n};function j_(e,t,n=0,r=e.length-1){for(;r>n;){if(r-n>600){let i=r-n+1,c=t-n+1,u=Math.log(i),l=.5*Math.exp(2*u/3),p=.5*Math.sqrt(u*l*(i-l)/i)*Math.sign(c-i/2),d=Math.max(n,Math.floor(t-c*l/i+p)),h=Math.min(r,Math.floor(t+(i-c)*l/i+p));j_(e,t,d,h)}let s=e[t],a=n,o=r;for(w.swap(e,n,t),Ml(e[r],s)>0&&w.swap(e,n,r);a<o;){for(w.swap(e,a,o),a++,o--;Ml(e[a],s)<0;)a=a+1;for(;Ml(e[o],s)>0;)o=o-1}Ml(e[n],s)===0?w.swap(e,n,o):(o=o+1,w.swap(e,o,r)),o<=t&&(n=o+1),t<=o&&(r=o-1)}}function K_(e,t,n,r,s){let a=t[t.length-1],[o,i]=[e.length/a,a],c=w.getTypedArrayFromDType(n,o*r),u=w.getTypedArrayFromDType("int32",o*r);for(let p=0;p<o;p++){let d=p*i,h=e.subarray(d,d+i),f=new Array(h.length);h.forEach((y,v)=>f[v]={value:y,index:v}),r<f.length&&(j_(f,r),f=f.slice(0,r)),s&&f.sort(Ml);let m=p*r,g=c.subarray(m,m+r),b=u.subarray(m,m+r);for(let y=0;y<r;y++)g[y]=f[y].value,b[y]=f[y].index}let l=t.slice();return l[l.length-1]=r,[Me(l,n,c),Me(l,"int32",u)]}function X_(e,t,n,r){let s=w.parseAxisParam(t,n)[0],a=[1,n[0],1];for(let f=0;f<s;f++)a[0]*=n[f];a[1]=n[s];for(let f=s+1;f<n.length;f++)a[2]*=n[f];let o={},i=new Int32Array(n[s]),c=new Ht(a,r,e),u=[],l=a[0]===1&&a[2]===1;for(let f=0;f<n[s];f++){let m;if(l)m=e[f].toString();else{let g=[];for(let b=0;b<a[0];b++)for(let y=0;y<a[2];y++)g.push(c.get(b,f,y));m=g.join(",")}if(o[m]!==void 0)i[f]=o[m];else{let g=Object.keys(o).length;o[m]=g,i[f]=g,u.push(f)}}let p=a.slice();p[1]=Object.keys(o).length;let d=new Ht(p,r);u.forEach((f,m)=>{for(let g=0;g<a[0];g++)for(let b=0;b<a[2];b++)d.set(c.get(g,f,b),g,m,b)});let h=n.slice();return h[s]=p[1],{outputValues:d.values,outputShape:h,indices:i}}qf("cpu",()=>new R0,1);var Y_=at(Ao,e=>e>=0?e:Math.exp(e)-1),F5={kernelName:Ao,backendName:"cpu",kernelFunc:Y_};function Z_(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{alpha:a}=r;be([s],"leakyRelu");let o=w.sizeFromShape(s.shape),i=n.data.get(s.dataId).values,c=w.getTypedArrayFromDType("float32",o);for(let u=0;u<i.length;u++)c[u]=i[u]<0?a*i[u]:i[u];return n.makeTensorInfo(s.shape,"float32",c)}var R5={kernelName:Mo,backendName:"cpu",kernelFunc:Z_},P5=Vt((e,t)=>e<0?t*e:e);function J_(e){let{inputs:t,backend:n}=e,{x:r,alpha:s}=t;be([r,s],"prelu");let a=n.data.get(r.dataId).values,o=n.data.get(s.dataId).values,[i,c]=P5(r.shape,s.shape,a,o,"float32");return n.makeTensorInfo(c,"float32",i)}var O5={kernelName:Yo,backendName:"cpu",kernelFunc:J_},Q_=at(Jo,e=>Math.max(0,e)),M5={kernelName:Jo,backendName:"cpu",kernelFunc:Q_},e2=at(ti,e=>Math.min(Math.max(0,e),6)),L5={kernelName:ti,backendName:"cpu",kernelFunc:e2};function ef(e,t,n,r,s){if(n==="linear")return ds({inputs:{x:t},backend:e});if(n==="relu")return Q_({inputs:{x:t},backend:e});if(n==="elu")return Y_({inputs:{x:t},backend:e});if(n==="relu6")return e2({inputs:{x:t},backend:e});if(n==="prelu")return J_({inputs:{x:t,alpha:r},backend:e});if(n==="leakyrelu")return Z_({inputs:{x:t},backend:e,attrs:{alpha:s}});if(n==="sigmoid")return B_({inputs:{x:t},backend:e});throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)}function ft(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{shape:a}=r,o=w.sizeFromShape(s.shape),i=w.inferFromImplicitShape(a,o),c=w.sizeFromShape(i);w.assert(o===c,()=>`The new shape (${i}) has ${c} elements and the old shape (${s.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`),n.incRef(s.dataId);let u=n.data.get(s.dataId);if(u.complexTensorInfos!=null){let l=u.complexTensorInfos.real,p=u.complexTensorInfos.imag;l.shape=i,p.shape=i}return{dataId:s.dataId,shape:i,dtype:s.dtype}}var z5={kernelName:xu,backendName:"cpu",kernelFunc:ft};function t2(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a}=t,{transposeA:o,transposeB:i}=r;be([s,a],"matMul");let c=s.shape.length,u=a.shape.length,l=o?s.shape[c-2]:s.shape[c-1],p=i?a.shape[u-1]:a.shape[u-2],d=o?s.shape[c-1]:s.shape[c-2],h=i?a.shape[u-2]:a.shape[u-1],f=s.shape.slice(0,-2),m=a.shape.slice(0,-2),g=w.sizeFromShape(f),b=w.sizeFromShape(m),v=Mu.assertAndGetBroadcastShape(s.shape.slice(0,-2),a.shape.slice(0,-2)).concat([d,h]);w.assert(l===p,()=>`Error in matMul: inner shapes (${l}) and (${p}) of Tensors with shapes ${s.shape} and ${a.shape} and transposeA=${o} and transposeB=${i} must match.`);let x=o?[g,l,d]:[g,d,l],k=i?[b,h,p]:[b,p,h],S=ft({inputs:{x:s},backend:n,attrs:{shape:x}}),C=ft({inputs:{x:a},backend:n,attrs:{shape:k}}),E=o?S.shape[1]:S.shape[2],$=o?S.shape[2]:S.shape[1],F=i?C.shape[1]:C.shape[2],A=Math.max(g,b),R=n.data.get(S.dataId).values,T=n.data.get(C.dataId).values,L=w.computeStrides(S.shape),V=w.computeStrides(C.shape),[G,j,H]=o?[L[0],1,L[1]]:[L[0],L[1],1],[Z,J,ee]=i?[1,V[1],V[0]]:[V[1],1,V[0]],re=$*F,te=Me([A,$,F],S.dtype),ie=te.values,ne=n.blockSize;for(let le=0;le<A;le++)for(let ue=0;ue<$;ue+=ne)for(let ye=0;ye<F;ye+=ne)for(let ve=0;ve<E;ve+=ne){let Se=Math.min(ue+ne,$),Le=Math.min(ye+ne,F),je=Math.min(ve+ne,E);for(let nt=ue;nt<Se;nt++)for(let rt=ye;rt<Le;rt++){let st=0;for(let Ze=ve;Ze<je;Ze++){let ze=Math.min(le,g-1)*G,dt=Math.min(le,b-1)*ee,qn=R[ze+nt*j+Ze*H],At=T[Ze*Z+rt*J+dt];st+=qn*At}ie[le*re+(nt*F+rt)]+=st}}return n.disposeIntermediateTensorInfo(S),n.disposeIntermediateTensorInfo(C),n.makeTensorInfo(v,te.dtype,te.values)}var B5={kernelName:xo,backendName:"cpu",kernelFunc:t2};function W5(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a,bias:o,preluActivationWeights:i}=t,{transposeA:c,transposeB:u,activation:l,leakyreluAlpha:p}=r,d,h,f,m=[];d=t2({inputs:{a:s,b:a},attrs:{transposeA:c,transposeB:u},backend:n}),o&&(h=_c({inputs:{a:d,b:o},backend:n}),m.push(d),d=h),l&&(f=ef(n,d,l,i,p),m.push(d),d=f);for(let b of m)n.disposeIntermediateTensorInfo(b);return d}var V5={kernelName:eo,backendName:"cpu",kernelFunc:W5},U5=at(Rc,e=>Math.acos(e)),G5={kernelName:Rc,backendName:"cpu",kernelFunc:U5},H5=at(Pc,e=>Math.acosh(e)),q5={kernelName:Pc,backendName:"cpu",kernelFunc:H5};function j5(e){let{inputs:t,backend:n}=e,r=t;be(t,"addN");let s=r.map(i=>n.data.get(i.dataId).values),a=Me(r[0].shape,r[0].dtype),o=a.values;for(let i=0;i<r.length;i++){let c=s[i];for(let u=0;u<o.length;u++)o[u]+=c[u]}return n.makeTensorInfo(a.shape,a.dtype,a.values)}var K5={kernelName:bo,backendName:"cpu",kernelFunc:j5};function X5(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;be(s,"all");let i=w.parseAxisParam(a,s.shape),c=i,u=N.getAxesPermutation(c,s.shape.length),l=s;u!=null&&(l=Vn({inputs:{x:s},backend:n,attrs:{perm:u}}),c=N.getInnerMostAxes(c.length,s.shape.length)),N.assertAxesAreInnerMostDims("all",c,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,c),h=w.sizeFromShape(d),f=w.makeZerosTypedArray(w.sizeFromShape(p),l.dtype),m=n.data.get(l.dataId).values;for(let b=0;b<f.length;++b){let y=b*h,v=m[y];for(let x=0;x<h;++x){let k=m[y+x];v=v&&k}f[b]=v}u!=null&&n.disposeIntermediateTensorInfo(l);let g=n.makeTensorInfo(p,l.dtype,f);if(o){let b=N.expandShapeToKeepDim(p,i),y=ft({inputs:{x:g},backend:n,attrs:{shape:b}});return n.disposeIntermediateTensorInfo(g),y}return g}var Y5={kernelName:Oc,backendName:"cpu",kernelFunc:X5};function Z5(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;be(s,"any");let i=w.parseAxisParam(a,s.shape),c=i,u=N.getAxesPermutation(c,s.shape.length),l=s;u!=null&&(l=Vn({inputs:{x:s},backend:n,attrs:{perm:u}}),c=N.getInnerMostAxes(c.length,s.shape.length)),N.assertAxesAreInnerMostDims("any",c,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,c),h=w.sizeFromShape(d),f=w.makeZerosTypedArray(w.sizeFromShape(p),l.dtype),m=n.data.get(l.dataId).values;for(let b=0;b<f.length;++b){let y=b*h,v=m[y];for(let x=0;x<h;++x){let k=m[y+x];v=v||k}f[b]=v}u!=null&&n.disposeIntermediateTensorInfo(l);let g=n.makeTensorInfo(p,l.dtype,f);if(o){let b=N.expandShapeToKeepDim(p,i),y=ft({inputs:{x:g},backend:n,attrs:{shape:b}});return n.disposeIntermediateTensorInfo(g),y}return g}var J5={kernelName:Mc,backendName:"cpu",kernelFunc:Z5};function Q5(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a}=r;be(s,"argMax");let o=w.parseAxisParam(a,s.shape),i=N.getAxesPermutation(o,s.shape.length),c=s,u=[];i!=null&&(c=Vn({inputs:{x:s},backend:n,attrs:{perm:i}}),u.push(c),o=N.getInnerMostAxes(o.length,c.shape.length)),o=[o[0]],N.assertAxesAreInnerMostDims("argMax",o,c.shape.length);let[l,p]=N.computeOutAndReduceShapes(c.shape,o),d=w.sizeFromShape(l),h=w.makeZerosTypedArray(d,"int32"),f=w.sizeFromShape(p),m=n.data.get(c.dataId).values;for(let g=0;g<h.length;++g){let b=g*f,y=m[b],v=0;for(let x=0;x<f;++x){let k=m[b+x];k>y&&(y=k,v=x)}h[g]=v}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(l,"int32",h)}var ej={kernelName:yo,backendName:"cpu",kernelFunc:Q5};function tj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a}=r;be(s,"argMin");let o=w.parseAxisParam(a,s.shape),i=N.getAxesPermutation(o,s.shape.length),c=s,u=[];i!=null&&(c=Vn({inputs:{x:s},backend:n,attrs:{perm:i}}),u.push(c),o=N.getInnerMostAxes(o.length,c.shape.length)),o=[o[0]],N.assertAxesAreInnerMostDims("argMin",o,c.shape.length);let[l,p]=N.computeOutAndReduceShapes(c.shape,o),d=w.sizeFromShape(l),h=w.makeZerosTypedArray(d,"int32"),f=w.sizeFromShape(p),m=n.data.get(c.dataId).values;for(let g=0;g<h.length;++g){let b=g*f,y=m[b],v=0;for(let x=0;x<f;++x){let k=m[b+x];k<y&&(y=k,v=x)}h[g]=v}return u.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.makeTensorInfo(l,"int32",h)}var nj={kernelName:dd,backendName:"cpu",kernelFunc:tj},rj=at(Lc,e=>Math.asin(e)),sj={kernelName:Lc,backendName:"cpu",kernelFunc:rj},aj=at(zc,e=>Math.asinh(e)),oj={kernelName:zc,backendName:"cpu",kernelFunc:aj},ij=at(Bc,e=>Math.atan(e)),cj={kernelName:Bc,backendName:"cpu",kernelFunc:ij},uj=Vt((e,t)=>Math.atan2(e,t)),lj=nn(Vc,uj),dj={kernelName:Vc,backendName:"cpu",kernelFunc:lj},pj=at(Wc,e=>Math.atanh(e)),hj={kernelName:Wc,backendName:"cpu",kernelFunc:pj};function q0(e,t,n,r,s,a){let o=s.strideHeight,i=s.strideWidth,c=s.dilationHeight,u=s.dilationWidth,l=s.effectiveFilterHeight,p=s.effectiveFilterWidth,d=s.padInfo.top,h=s.padInfo.left,f=a==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,m=Me(s.outShape,n),g=m.values,b=s.outShape[1]*s.outShape[2]*s.outShape[3],y=s.outShape[2]*s.outShape[3],v=s.outShape[3];for(let x=0;x<s.batchSize;++x){let k=x*b,S=x*r[0];for(let C=0;C<s.inChannels;++C)for(let E=0;E<s.outHeight;++E){let $=E*o-d,F=Math.max(0,$),A=Math.min(s.inHeight,l+$),R=k+E*y;for(let T=0;T<s.outWidth;++T){let L=T*i-h,V=Math.max(0,L),G=Math.min(s.inWidth,p+L),j=f,H=0,Z=0;for(let ee=F;ee<A;ee+=c){let re=S+ee*r[1];for(let te=V;te<G;te+=u){let ie=re+te*r[2],ne=e[ie+C];a==="max"&&ne>j?j=ne:a==="avg"&&(H+=ne,Z++)}if(isNaN(j))break}let J=R+T*v+C;g[J]=a==="avg"?H/Z:j}}}return m}function n2(e,t,n,r,s=!1,a=!1){let o=Me(r.outShape,"int32"),i=r.strideHeight,c=r.strideWidth,u=r.dilationHeight,l=r.dilationWidth,p=r.effectiveFilterHeight,d=r.effectiveFilterWidth,h=r.padInfo.top,f=r.padInfo.left,m=Me(t,n,e);for(let g=0;g<r.batchSize;++g)for(let b=0;b<r.inChannels;++b)for(let y=0;y<r.outHeight;++y){let v=y*i-h,x=v;for(;x<0;)x+=u;let k=Math.min(r.inHeight,p+v);for(let S=0;S<r.outWidth;++S){let C=S*c-f,E=C;for(;E<0;)E+=l;let $=Math.min(r.inWidth,d+C),F=Number.NEGATIVE_INFINITY,A=-1;for(let R=x;R<k;R+=u){let T=R-v;for(let L=E;L<$;L+=l){let V=L-C,G=m.get(g,R,L,b);G>F&&(F=G,s?A=a?((g*r.inHeight+R)*r.inWidth+L)*r.inChannels+b:(R*r.inWidth+L)*r.inChannels+b:A=T*d+V)}}o.set(A,g,y,S,b)}}return o}function r2(e,t,n,r,s,a){let o=s.strideDepth,i=s.strideHeight,c=s.strideWidth,u=s.dilationDepth,l=s.dilationHeight,p=s.dilationWidth,d=s.effectiveFilterDepth,h=s.effectiveFilterHeight,f=s.effectiveFilterWidth,m=s.padInfo.front,g=s.padInfo.top,b=s.padInfo.left,y=a==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,v=Me(s.outShape,n),x=v.values,k=s.outShape[1]*s.outShape[2]*s.outShape[3]*s.outShape[4],S=s.outShape[2]*s.outShape[3]*s.outShape[4],C=s.outShape[3]*s.outShape[4],E=s.outShape[4];for(let $=0;$<s.batchSize;++$){let F=$*k,A=$*r[0];for(let R=0;R<s.inChannels;++R)for(let T=0;T<s.outDepth;++T){let L=T*o-m,V=L;for(;V<0;)V+=u;let G=Math.min(s.inDepth,d+L),j=F+T*S;for(let H=0;H<s.outHeight;++H){let Z=H*i-g,J=Z;for(;J<0;)J+=l;let ee=Math.min(s.inHeight,h+Z),re=j+H*C;for(let te=0;te<s.outWidth;++te){let ie=te*c-b,ne=ie;for(;ne<0;)ne+=p;let le=Math.min(s.inWidth,f+ie),ue=re+te*E,ye=y,ve=0,Se=0;for(let je=V;je<G;je+=u){let nt=A+je*r[1];for(let rt=J;rt<ee;rt+=l){let st=nt+rt*r[2];for(let Ze=ne;Ze<le;Ze+=p){let ze=st+Ze*r[3],dt=e[ze+R];if(a==="max"&&dt>ye?ye=dt:a==="avg"&&(ve+=dt,Se++),isNaN(ye))break}if(isNaN(ye))break}if(isNaN(ye))break}let Le=ue+R;x[Le]=a==="avg"?ve/Se:ye}}}}return v}function fj(e,t){let n=Me(t.outShape,"int32"),r=t.strideDepth,s=t.strideHeight,a=t.strideWidth,o=t.dilationDepth,i=t.dilationHeight,c=t.dilationWidth,u=t.effectiveFilterDepth,l=t.effectiveFilterHeight,p=t.effectiveFilterWidth,d=t.padInfo.front,h=t.padInfo.top,f=t.padInfo.left;for(let m=0;m<t.batchSize;++m)for(let g=0;g<t.inChannels;++g)for(let b=0;b<t.outDepth;++b){let y=b*r-d,v=y;for(;v<0;)v+=o;let x=Math.min(t.inDepth,u+y);for(let k=0;k<t.outHeight;++k){let S=k*s-h,C=S;for(;C<0;)C+=i;let E=Math.min(t.inHeight,l+S);for(let $=0;$<t.outWidth;++$){let F=$*a-f,A=F;for(;A<0;)A+=c;let R=Math.min(t.inWidth,p+F),T=Number.NEGATIVE_INFINITY,L=-1;for(let V=v;V<x;V+=o){let G=V-y;for(let j=C;j<E;j+=i){let H=j-S;for(let Z=A;Z<R;Z+=c){let J=Z-F,ee=e.get(m,V,j,Z,g);ee>=T&&(T=ee,L=G*l*p+H*l+J)}}}n.set(L,m,b,k,$,g)}}}return n}function mj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t;be(s,"avgPool");let{filterSize:a,strides:o,pad:i,dimRoundingMode:c}=r,u=1;w.assert(N.eitherStridesOrDilationsAreOne(o,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${o} and dilations '${u}'`);let l=N.computePool2DInfo(s.shape,a,o,u,i,c),p;if(l.filterWidth===1&&l.filterHeight===1&&w.arraysEqual(l.inShape,l.outShape))p=ds({inputs:{x:s},backend:n});else{let d=n.data.get(s.dataId).values,h=w.computeStrides(s.shape),f=q0(d,s.shape,s.dtype,h,l,"avg");p=n.makeTensorInfo(l.outShape,s.dtype,f.values)}return p}var gj={kernelName:vo,backendName:"cpu",kernelFunc:mj};function bj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{filterSize:a,strides:o,pad:i,dimRoundingMode:c,dataFormat:u}=r;be(s,"avgPool3d");let l=N.computePool3DInfo(s.shape,a,o,1,i,c,u),p=n.data.get(s.dataId).values,d=r2(p,s.shape,s.dtype,w.computeStrides(s.shape),l,"avg");return n.makeTensorInfo(d.shape,"float32",d.values)}var yj={kernelName:pd,backendName:"cpu",kernelFunc:bj};function vj(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,{filterSize:o,strides:i,pad:c,dimRoundingMode:u}=r;be([s,a],"avgPool3DGrad");let l=N.computePool3DInfo(a.shape,o,i,1,c,u),p=l.strideDepth,d=l.strideHeight,h=l.strideWidth,f=l.filterDepth,m=l.filterHeight,g=l.filterWidth,b=l.dilationDepth,y=l.dilationHeight,v=l.dilationWidth,x=l.effectiveFilterDepth,k=l.effectiveFilterHeight,S=l.effectiveFilterWidth,C=x-1-l.padInfo.front,E=S-1-l.padInfo.left,$=k-1-l.padInfo.top,F=Me(a.shape,"float32"),A=1/(f*m*g),R=n.bufferSync(s);for(let T=0;T<l.batchSize;++T)for(let L=0;L<l.inChannels;++L)for(let V=0;V<l.inDepth;++V)for(let G=0;G<l.inHeight;++G)for(let j=0;j<l.inWidth;++j){let H=V-C,Z=G-$,J=j-E,ee=0;for(let re=0;re<x;re+=b){let te=(H+re)/p;if(!(te<0||te>=l.outDepth||Math.floor(te)!==te))for(let ie=0;ie<k;ie+=y){let ne=(Z+ie)/d;if(!(ne<0||ne>=l.outHeight||Math.floor(ne)!==ne))for(let le=0;le<S;le+=v){let ue=(J+le)/h;if(ue<0||ue>=l.outWidth||Math.floor(ue)!==ue)continue;let ye=R.get(T,te,ne,ue,L);ee+=ye}}}F.set(ee*A,T,V,G,j,L)}return n.makeTensorInfo(F.shape,F.dtype,F.values)}var xj={kernelName:hf,backendName:"cpu",kernelFunc:vj};function wj(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,o=a;be([s,a],"avgPoolGrad");let{filterSize:i,strides:c,pad:u}=r,l=N.computePool2DInfo(o.shape,i,c,1,u),p=l.strideHeight,d=l.strideWidth,h=l.filterHeight,f=l.filterWidth,m=l.dilationHeight,g=l.dilationWidth,b=l.effectiveFilterHeight,y=l.effectiveFilterWidth,v=y-1-l.padInfo.left,x=b-1-l.padInfo.top,k=Me(o.shape,"float32"),S=1/(h*f),C=n.data.get(s.dataId).values,E=Me(s.shape,"float32",C);for(let $=0;$<l.batchSize;++$)for(let F=0;F<l.inChannels;++F)for(let A=0;A<l.inHeight;++A)for(let R=0;R<l.inWidth;++R){let T=A-x,L=R-v,V=0;for(let G=0;G<b;G+=m){let j=(T+G)/p;if(!(j<0||j>=l.outHeight||Math.floor(j)!==j))for(let H=0;H<y;H+=g){let Z=(L+H)/d;if(Z<0||Z>=l.outWidth||Math.floor(Z)!==Z)continue;let J=E.get($,j,Z,F);V+=J}}k.set(V*S,$,A,R,F)}return n.makeTensorInfo(k.shape,k.dtype,k.values)}var Ij={kernelName:pf,backendName:"cpu",kernelFunc:wj};function kj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,scale:a,offset:o,mean:i,variance:c}=t;w.assert(i.shape.length===c.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),w.assert(o==null||i.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),w.assert(a==null||i.shape.length===a.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),be([s,i,c,a,o],"batchNorm");let{varianceEpsilon:u}=r;u==null&&(u=.001);let l=n.data.get(s.dataId).values,p=n.data.get(i.dataId).values,d=n.data.get(c.dataId).values,h=a?n.data.get(a.dataId).values:new Float32Array([1]),f=o?n.data.get(o.dataId).values:new Float32Array([0]),m=new Float32Array(l.length),g=f.length,b=h.length,y=d.length,v=p.length,x=0,k=0,S=0,C=0;for(let E=0;E<l.length;++E)m[E]=f[x++]+(l[E]-p[k++])*h[S++]/Math.sqrt(d[C++]+u),x>=g&&(x=0),k>=v&&(k=0),S>=b&&(S=0),C>=y&&(C=0);return n.makeTensorInfo(s.shape,s.dtype,m)}var Sj={kernelName:Ro,backendName:"cpu",kernelFunc:kj};function Tj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,crops:o}=r;be([s],"batchToSpaceND");let i=a.reduce((b,y)=>b*y),c=N.getReshaped(s.shape,a,i),u=N.getPermuted(c.length,a.length),l=N.getReshapedPermuted(s.shape,a,i),p=N.getSliceBeginCoords(o,a.length),d=N.getSliceSize(l,o,a.length),h=ft({inputs:{x:s},backend:n,attrs:{shape:c}}),f=Vn({inputs:{x:h},backend:n,attrs:{perm:u}}),m=ft({inputs:{x:f},backend:n,attrs:{shape:l}}),g=po({inputs:{x:m},backend:n,attrs:{begin:p,size:d}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}var Cj={kernelName:Uc,backendName:"cpu",kernelFunc:Tj};function Nj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,weights:a}=t,{size:o}=r,i=n.data.get(s.dataId).values,c=n.data.get(a.dataId).values,u=O0(i,c,a.dtype,a.shape,o);return n.makeTensorInfo([o],a.dtype,u)}var _j={kernelName:ff,backendName:"cpu",kernelFunc:Nj};function Ej(e){let{inputs:t,backend:n}=e,{s0:r,s1:s}=t,a=n.data.get(r.dataId).values,o=n.data.get(s.dataId).values,i=N.assertAndGetBroadcastShape(Array.from(a),Array.from(o));return n.makeTensorInfo([i.length],"int32",Int32Array.from(i))}var Aj={kernelName:mf,backendName:"cpu",kernelFunc:Ej},$j=at(ya,(e,t)=>{let n=t;return e>n.clipValueMax?n.clipValueMax:e<n.clipValueMin?n.clipValueMin:e}),Dj={kernelName:ya,backendName:"cpu",kernelFunc:$j},Fj=e=>{let{x:t}=e.inputs,n=e.backend,r=new Float32Array(w.sizeFromShape(t.shape)),s=n.data.get(t.dataId),a=s.complexTensorInfos.real,o=s.complexTensorInfos.imag,i=n.data.get(a.dataId).values,c=n.data.get(o.dataId).values;for(let u=0;u<i.length;u++){let l=i[u],p=c[u];r[u]=Math.hypot(l,p)}return n.makeOutput(r,t.shape,"float32")},Rj={kernelName:hd,backendName:"cpu",kernelFunc:Fj};function Ec(e){let{inputs:t,backend:n}=e,{input:r}=t,s=n.data.get(r.dataId).complexTensorInfos.imag,a=n.data.get(s.dataId).values;return n.makeTensorInfo(s.shape,s.dtype,a)}var Pj={kernelName:_f,backendName:"cpu",kernelFunc:Ec};function Ac(e){let{inputs:t,backend:n,attrs:r}=e,{axis:s}=r,a=w.parseAxisParam(s,t[0].shape)[0],o=t.map(m=>m.shape);N.assertParamsConsistent(o,a);let i=N.computeOutShape(t.map(m=>m.shape),a);if(w.sizeFromShape(i)===0)return n.makeTensorInfo(i,t[0].dtype,[]);let c=t.filter(m=>w.sizeFromShape(m.shape)>0);if(c.length===1)return ds({inputs:{x:c[0]},backend:n});if(c[0].dtype==="complex64"){let m=c.map(x=>lo({inputs:{input:x},backend:n})),g=c.map(x=>Ec({inputs:{input:x},backend:n})),b=Ac({inputs:m,backend:n,attrs:{axis:a}}),y=Ac({inputs:g,backend:n,attrs:{axis:a}}),v=Jn({inputs:{real:b,imag:y},backend:n});return m.forEach(x=>n.disposeIntermediateTensorInfo(x)),g.forEach(x=>n.disposeIntermediateTensorInfo(x)),n.disposeIntermediateTensorInfo(b),n.disposeIntermediateTensorInfo(y),v}let u=c.map(m=>{let b=[-1,w.sizeFromShape(m.shape.slice(a))];return ft({inputs:{x:m},backend:n,attrs:{shape:b}})}),l=u.map(m=>({vals:n.data.get(m.dataId).values,shape:m.shape}));i=N.computeOutShape(u.map(m=>m.shape),1);let p=u[0].shape[0]===1,d=M0(l,i,t[0].dtype,p),h=N.computeOutShape(c.map(m=>m.shape),a),f=n.makeTensorInfo(h,t[0].dtype,d);return u.forEach(m=>n.disposeIntermediateTensorInfo(m)),f}var Oj={kernelName:Gc,backendName:"cpu",kernelFunc:Ac};function s2(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dataFormat:c,dilations:u,dimRoundingMode:l}=r;be([s,a],"conv2d");let p=N.convertConv2DDataFormat(c),d=N.computeConv2DInfo(s.shape,a.shape,o,u,i,l,!1,p),h=d.filterHeight,f=d.filterWidth,m=d.dilationHeight,g=d.dilationWidth,b=d.padInfo.left,y=d.padInfo.top,v=d.dataFormat==="channelsLast",x=new Ht(d.outShape,s.dtype),k=w.computeStrides(s.shape),S=w.computeStrides(a.shape),C=k[0],E=v?k[1]:k[2],$=v?k[2]:1,F=v?1:k[1],A=x.strides[0],R=v?x.strides[1]:x.strides[2],T=v?x.strides[2]:1,L=v?1:x.strides[1],V=n.data.get(s.dataId).values,G=n.data.get(a.dataId).values,j=x.values;for(let H=0;H<d.batchSize;++H){let Z=H*C,J=H*A;for(let ee=0;ee<d.outHeight;++ee){let re=J+ee*R,te=ee*d.strideHeight-y;for(let ie=0;ie<h;++ie){let ne=te+ie*m;if(ne<0||ne>=d.inHeight)continue;let le=ie*S[0],ue=Z+ne*E;for(let ye=0;ye<d.outWidth;++ye){let ve=re+ye*T,Se=ye*d.strideWidth-b;for(let Le=0;Le<f;++Le){let je=Se+Le*g;if(je<0||je>=d.inWidth)continue;let nt=le+Le*S[1],rt=ue+je*$,st=nt;for(let Ze=0;Ze<d.inChannels;++Ze){let ze=V[rt+Ze*F];for(let dt=0;dt<d.outChannels;++dt)j[ve+dt*L]+=ze*G[st+dt];st+=d.outChannels}}}}}}return n.makeTensorInfo(x.shape,x.dtype,j)}var Mj={kernelName:ko,backendName:"cpu",kernelFunc:s2};function Lj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,pad:i,dataFormat:c,dimRoundingMode:u,filterShape:l}=r;be([s,a],"conv2dBackpropFilter");let p=N.convertConv2DDataFormat(c),d=N.computeConv2DInfo(s.shape,l,o,1,i,u,!1,p),{strideHeight:h,strideWidth:f,filterHeight:m,filterWidth:g}=d,b=d.dataFormat==="channelsLast",y=new Ht(d.filterShape,"float32"),v=d.padInfo.left,x=d.padInfo.top,k=n.data.get(s.dataId).values,S=n.data.get(a.dataId).values,C=new Ht(s.shape,s.dtype,k),E=new Ht(a.shape,a.dtype,S);for(let $=0;$<m;++$){let F=Math.max(0,Math.ceil((x-$)/h)),A=Math.min(d.outHeight,(d.inHeight+x-$)/h);for(let R=0;R<g;++R){let T=Math.max(0,Math.ceil((v-R)/f)),L=Math.min(d.outWidth,(d.inWidth+v-R)/f);for(let V=0;V<d.inChannels;++V)for(let G=0;G<d.outChannels;++G){let j=0;for(let H=0;H<d.batchSize;++H)for(let Z=F;Z<A;++Z){let J=$+Z*h-x;for(let ee=T;ee<L;++ee){let re=R+ee*f-v;b?j+=C.get(H,J,re,V)*E.get(H,Z,ee,G):j+=C.get(H,V,J,re)*E.get(H,G,Z,ee)}}y.set(j,$,R,V,G)}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}var zj={kernelName:bf,backendName:"cpu",kernelFunc:Lj};function Bj(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{inputShape:o,strides:i,pad:c,dataFormat:u,dimRoundingMode:l}=r;be([s,a],"conv2dBackpropInput");let p=w.computeStrides(a.shape),d=w.computeStrides(s.shape),h=N.convertConv2DDataFormat(u),f=N.computeConv2DInfo(o,a.shape,i,1,c,l,!1,h),m=new Ht(f.inShape,"float32"),g=m.values,b=n.data.get(s.dataId).values,y=n.data.get(a.dataId).values,[v,x,k]=p,{batchSize:S,filterHeight:C,filterWidth:E,inChannels:$,inHeight:F,inWidth:A,outChannels:R,outHeight:T,outWidth:L,strideHeight:V,strideWidth:G}=f;h=f.dataFormat;let j=C-1-f.padInfo.top,H=E-1-f.padInfo.left,Z=h==="channelsLast",J=m.strides[0],ee=Z?m.strides[1]:m.strides[2],re=Z?m.strides[2]:1,te=Z?1:m.strides[1],ie=d[0],ne=Z?d[1]:d[2],le=Z?d[2]:1,ue=Z?1:d[1];for(let ye=0;ye<S;++ye)for(let ve=0;ve<$;++ve)for(let Se=0;Se<F;++Se){let Le=Se-j,je=Math.max(0,Math.ceil(Le/V)),nt=Math.min(T,(C+Le)/V);for(let rt=0;rt<A;++rt){let st=rt-H,Ze=Math.max(0,Math.ceil(st/G)),ze=Math.min(L,(E+st)/G),dt=0;for(let At=je;At<nt;++At){let wr=At*V-Le;for(let on=Ze;on<ze;++on){let Ir=on*G-st,vn=ie*ye+ne*At+le*on,jn=v*(C-1-wr)+x*(E-1-Ir)+k*ve;for(let yt=0;yt<R;++yt){let Pn=b[vn+ue*yt],hn=y[jn+yt];dt+=Pn*hn}}}let qn=J*ye+ee*Se+re*rt+te*ve;g[qn]=dt}}return n.makeTensorInfo(m.shape,m.dtype,m.values)}var Wj={kernelName:So,backendName:"cpu",kernelFunc:Bj};function Vj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dilations:c}=r;be([s,a],"conv3d");let u=N.computeConv3DInfo(s.shape,a.shape,o,c,i),{filterDepth:l,filterHeight:p,filterWidth:d,dilationDepth:h,dilationHeight:f,dilationWidth:m,padInfo:g}=u,b=g.front,y=g.left,v=g.top,x=new Ht(u.outShape,s.dtype),k=n.data.get(s.dataId).values,S=n.data.get(a.dataId).values,C=x.values,E=w.computeStrides(s.shape),$=w.computeStrides(a.shape);for(let F=0;F<u.batchSize;++F){let A=F*E[0],R=F*x.strides[0];for(let T=0;T<u.outDepth;++T){let L=R+T*x.strides[1],V=T*u.strideDepth-b;for(let G=0;G<l;++G){let j=V+G*h;if(j<0||j>=u.inDepth)continue;let H=G*$[0],Z=A+j*E[1];for(let J=0;J<u.outHeight;++J){let ee=L+J*x.strides[2],re=J*u.strideHeight-v;for(let te=0;te<p;++te){let ie=re+te*f;if(ie<0||ie>=u.inHeight)continue;let ne=H+te*$[1],le=Z+ie*E[2];for(let ue=0;ue<u.outWidth;++ue){let ye=ee+ue*u.outChannels,ve=ue*u.strideWidth-y;for(let Se=0;Se<d;++Se){let Le=ve+Se*m;if(Le<0||Le>=u.inWidth)continue;let je=ne+Se*$[2],nt=le+Le*u.inChannels,rt=je;for(let st=0;st<u.inChannels;++st){let Ze=k[nt+st];for(let ze=0;ze<u.outChannels;++ze)C[ye+ze]+=Ze*S[rt+ze];rt+=u.outChannels}}}}}}}}return n.makeTensorInfo(x.shape,x.dtype,x.values)}var Uj={kernelName:fd,backendName:"cpu",kernelFunc:Vj};function Gj(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,pad:i,filterShape:c}=r;be([s,a],"conv3dBackpropFilterV2");let u=w.computeStrides(s.shape),l=w.computeStrides(a.shape),p=N.computeConv3DInfo(s.shape,c,o,1,i),d=p.strideDepth,h=p.strideHeight,f=p.strideWidth,m=p.filterDepth,g=p.filterHeight,b=p.filterWidth,y=new Ht(p.filterShape,"float32"),v=y.values,[x,k,S,C]=y.strides,E=n.data.get(a.dataId).values,[$,F,A,R]=l,T=n.data.get(s.dataId).values,[L,V,G,j]=u,H=p.padInfo.front,Z=p.padInfo.left,J=p.padInfo.top;for(let ee=0;ee<m;++ee){let re=Math.max(0,Math.ceil((H-ee)/d)),te=Math.min(p.outDepth,(p.inDepth+H-ee)/d),ie=ee*x;for(let ne=0;ne<g;++ne){let le=Math.max(0,Math.ceil((J-ne)/h)),ue=Math.min(p.outHeight,(p.inHeight+J-ne)/h),ye=ne*k+ie;for(let ve=0;ve<b;++ve){let Se=Math.max(0,Math.ceil((Z-ve)/f)),Le=Math.min(p.outWidth,(p.inWidth+Z-ve)/f),je=ve*S+ye;for(let nt=0;nt<p.inChannels;++nt){let rt=nt*C+je;for(let st=0;st<p.outChannels;++st){let Ze=0;for(let ze=0;ze<p.batchSize;++ze){let dt=ze*L,qn=ze*$;for(let At=re;At<te;++At){let on=(ee+At*d-H)*V+dt,Ir=At*F+qn;for(let vn=le;vn<ue;++vn){let yt=(ne+vn*h-J)*G+on,Pn=vn*A+Ir;for(let hn=Se;hn<Le;++hn){let qs=(ve+hn*f-Z)*j+yt,kr=hn*R+Pn;Ze+=T[qs+nt]*E[kr+st]}}}}v[rt+st]=Ze}}}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}var Hj={kernelName:yf,backendName:"cpu",kernelFunc:Gj};function qj(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{pad:o,strides:i,inputShape:c}=r;be([s],"conv3dBackpropInputV2");let u=w.computeStrides(s.shape),l=w.computeStrides(a.shape),p=N.computeConv3DInfo(c,a.shape,i,1,o),d=new Ht(p.inShape,"float32"),h=d.values,[f,m,g,b]=d.strides,y=n.data.get(s.dataId).values,[v,x,k,S]=u,C=n.data.get(a.dataId).values,[E,$,F,A]=l,{batchSize:R,filterDepth:T,filterHeight:L,filterWidth:V,inChannels:G,inDepth:j,inHeight:H,inWidth:Z,outChannels:J,outDepth:ee,outHeight:re,outWidth:te,strideDepth:ie,strideHeight:ne,strideWidth:le}=p,ue=T-1-p.padInfo.front,ye=L-1-p.padInfo.top,ve=V-1-p.padInfo.left;for(let Se=0;Se<R;++Se)for(let Le=0;Le<G;++Le)for(let je=0;je<j;++je){let nt=je-ue,rt=Math.max(0,Math.ceil(nt/ie)),st=Math.min(ee,(T+nt)/ie);for(let Ze=0;Ze<H;++Ze){let ze=Ze-ye,dt=Math.max(0,Math.ceil(ze/ne)),qn=Math.min(re,(L+ze)/ne);for(let At=0;At<Z;++At){let wr=At-ve,on=Math.max(0,Math.ceil(wr/le)),Ir=Math.min(te,(V+wr)/le),vn=0;for(let jn=rt;jn<st;++jn){let yt=jn*ie-nt;for(let Pn=dt;Pn<qn;++Pn){let hn=Pn*ne-ze;for(let ts=on;ts<Ir;++ts){let qs=ts*le-wr,kr=v*Se+x*jn+k*Pn+S*ts,xl=E*(T-1-yt)+$*(L-1-hn)+F*(V-1-qs)+A*Le;for(let Sr=0;Sr<J;++Sr){let Ki=y[kr+Sr],Xt=C[xl+Sr];vn+=Ki*Xt}}}}h[f*Se+m*je+g*Ze+b*At+Le]=vn}}}return n.makeTensorInfo(d.shape,d.dtype,d.values)}var jj={kernelName:vf,backendName:"cpu",kernelFunc:qj},Kj=at(To,e=>Math.cos(e)),Xj={kernelName:To,backendName:"cpu",kernelFunc:Kj},Yj=at(Co,e=>Math.cosh(e)),Zj={kernelName:Co,backendName:"cpu",kernelFunc:Yj};function Jj(e){let{inputs:t,backend:n,attrs:r}=e,{image:s,boxes:a,boxInd:o}=t,{cropSize:i,method:c,extrapolationValue:u}=r,[l,p,d,h]=s.shape,f=a.shape[0],[m,g]=i,b=Me([f,m,g,h],"float32"),y=n.data.get(a.dataId).values,v=n.data.get(o.dataId).values,x=n.data.get(s.dataId).values,k=w.computeStrides(s.shape),S=w.computeStrides(b.shape);for(let C=0;C<f;C++){let E=C*4,$=y[E],F=y[E+1],A=y[E+2],R=y[E+3],T=v[C];if(T>=l)continue;let L=m>1?(A-$)*(p-1)/(m-1):0,V=g>1?(R-F)*(d-1)/(g-1):0;for(let G=0;G<m;G++){let j=m>1?$*(p-1)+G*L:.5*($+A)*(p-1);if(j<0||j>p-1){for(let H=0;H<g;H++)for(let Z=0;Z<h;Z++){let J=Z+H*S[2]+G*S[1]+C*S[0];b.values[J]=u}continue}if(c==="bilinear"){let H=Math.floor(j),Z=Math.ceil(j),J=j-H;for(let ee=0;ee<g;ee++){let re=g>1?F*(d-1)+ee*V:.5*(F+R)*(d-1);if(re<0||re>d-1){for(let le=0;le<h;le++){let ue=le+ee*S[2]+G*S[1]+C*S[0];b.values[ue]=u}continue}let te=Math.floor(re),ie=Math.ceil(re),ne=re-te;for(let le=0;le<h;le++){let ue=le+te*k[2]+H*k[1]+T*k[0],ye=x[ue];ue=le+ie*k[2]+H*k[1]+T*k[0];let ve=x[ue];ue=le+te*k[2]+Z*k[1]+T*k[0];let Se=x[ue];ue=le+ie*k[2]+Z*k[1]+T*k[0];let Le=x[ue],je=ye+(ve-ye)*ne,nt=Se+(Le-Se)*ne;ue=le+ee*S[2]+G*S[1]+C*S[0],b.values[ue]=je+(nt-je)*J}}}else for(let H=0;H<g;++H){let Z=g>1?F*(d-1)+H*V:.5*(F+R)*(d-1);if(Z<0||Z>d-1){for(let re=0;re<h;re++){let te=re+H*S[2]+G*S[1]+C*S[0];b.values[te]=u}continue}let J=Math.round(Z),ee=Math.round(j);for(let re=0;re<h;re++){let te=re+J*k[2]+ee*k[1]+T*k[0],ie=re+H*S[2]+G*S[1]+C*S[0];b.values[ie]=x[te]}}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}var Qj={kernelName:qc,backendName:"cpu",kernelFunc:Jj};function e8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r;be(s,"cumprod");let c=N.getAxesPermutation([a],s.shape.length),u=s;c!=null&&(u=Vn({inputs:{x:s},backend:n,attrs:{perm:c}}));let l=N.getInnerMostAxes(1,s.shape.length)[0];if(l!==u.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${l}`);let p=hr(u.dtype,"int32"),d=w.makeOnesTypedArray(w.sizeFromShape(u.shape),p),h=n.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=i?(b,y)=>b+f-y-1:(b,y)=>b+y;for(let b=0;b<h.length;b+=f)for(let y=0;y<f;y++){let v=m(b,y);if(y===0)d[v]=o?1:h[v];else{let x=m(b,y-1);d[v]=o?h[x]*d[x]:h[v]*d[x]}}let g=n.makeTensorInfo(u.shape,p,d);if(c!=null){let b=N.getUndoAxesPermutation(c),y=Vn({inputs:{x:g},backend:n,attrs:{perm:b}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),y}return g}var t8={kernelName:Hc,backendName:"cpu",kernelFunc:e8};function n8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r;be(s,"cumsum");let c=N.getAxesPermutation([a],s.shape.length),u=s;c!=null&&(u=Vn({inputs:{x:s},backend:n,attrs:{perm:c}}));let l=N.getInnerMostAxes(1,s.shape.length)[0];if(l!==u.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${l}`);let p=hr(u.dtype,"int32"),d=w.makeZerosTypedArray(w.sizeFromShape(u.shape),p),h=n.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=i?(b,y)=>b+f-y-1:(b,y)=>b+y;for(let b=0;b<h.length;b+=f)for(let y=0;y<f;y++){let v=m(b,y);if(y===0)d[v]=o?0:h[v];else{let x=m(b,y-1);d[v]=o?h[x]+d[x]:h[v]+d[x]}}let g=n.makeTensorInfo(u.shape,p,d);if(c!=null){let b=N.getUndoAxesPermutation(c),y=Vn({inputs:{x:g},backend:n,attrs:{perm:b}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),y}return g}var r8={kernelName:No,backendName:"cpu",kernelFunc:n8};function s8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,weights:a}=t,{size:o,binaryOutput:i}=r;if(s.shape.length===1){let c=n.data.get(s.dataId).values,u=n.data.get(a.dataId).values,l=O0(c,u,a.dtype,a.shape,o);return n.makeTensorInfo([o],a.dtype,l)}else if(s.shape.length===2){let c=n.bufferSync(s),u=n.bufferSync(a),l=f_(c,u,o,i);return n.makeTensorInfo(l.shape,a.dtype,l.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${s.shape.length}.`)}var a8={kernelName:xf,backendName:"cpu",kernelFunc:s8};function o8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockSize:a,dataFormat:o}=r;w.assert(o==="NHWC",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${o}`);let i=s.shape[0],c=s.shape[1],u=s.shape[2],l=s.shape[3],p=c*a,d=u*a,h=l/(a*a),f=n.data.get(s.dataId).values,m=new Float32Array(i*p*d*h),g=0;for(let b=0;b<i;++b)for(let y=0;y<p;++y){let v=Math.floor(y/a),x=y%a;for(let k=0;k<d;++k){let S=Math.floor(k/a),C=k%a,E=(x*a+C)*h;for(let $=0;$<h;++$){let A=$+E+l*(S+u*(v+c*b));m[g++]=f[A]}}}return n.makeTensorInfo([i,p,d,h],s.dtype,m)}var i8={kernelName:jc,backendName:"cpu",kernelFunc:o8};function a2(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dilations:c,dimRoundingMode:u}=r;be([s,a],"depthwiseConv2DNative");let l=w.computeStrides(s.shape),p=w.computeStrides(a.shape),d=c;d==null&&(d=[1,1]),w.assert(N.eitherStridesOrDilationsAreOne(o,d),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${o} and dilations '${d}'`);let h=N.computeConv2DInfo(s.shape,a.shape,o,d,i,u,!0),{filterHeight:f,filterWidth:m,dilationHeight:g,dilationWidth:b,padInfo:y}=h,v=y.left,x=y.top,k=h.outChannels/h.inChannels,S=new Ht(h.outShape,s.dtype),C=n.data.get(s.dataId).values,E=n.data.get(a.dataId).values,$=S.values;for(let F=0;F<h.batchSize;++F){let A=F*l[0],R=F*S.strides[0];for(let T=0;T<h.outHeight;++T){let L=R+T*S.strides[1],V=T*h.strideHeight-x;for(let G=0;G<f;++G){let j=V+G*g;if(j<0||j>=h.inHeight)continue;let H=G*p[0],Z=A+j*l[1];for(let J=0;J<h.outWidth;++J){let ee=L+J*S.strides[2],re=J*h.strideWidth-v;for(let te=0;te<m;++te){let ie=re+te*b;if(ie<0||ie>=h.inWidth)continue;let ne=H+te*p[1],le=Z+ie*h.inChannels,ue=ee,ye=ne;for(let ve=0;ve<h.inChannels;++ve){let Se=C[le+ve];for(let Le=0;Le<k;++Le)$[ue+Le]+=Se*E[ye+Le];ue+=k,ye+=k}}}}}}return n.makeTensorInfo(S.shape,S.dtype,S.values)}var c8={kernelName:_o,backendName:"cpu",kernelFunc:a2};function u8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,dilations:i,pad:c,dimRoundingMode:u,filterShape:l}=r;be([s,a],"depthwiseConv2dNativeBackpropFilter");let p=N.computeConv2DInfo(s.shape,l,o,i,c,u,!0),{strideHeight:d,strideWidth:h,filterHeight:f,filterWidth:m}=p,g=new Ht(p.filterShape,"float32"),b=p.padInfo.left,y=p.padInfo.top,v=p.outChannels/p.inChannels,x=n.data.get(s.dataId).values,k=new Ht(s.shape,s.dtype,x),S=n.data.get(a.dataId).values,C=new Ht(a.shape,a.dtype,S);for(let E=0;E<f;++E){let $=Math.max(0,Math.ceil((y-E)/d)),F=Math.min(p.outHeight,(p.inHeight+y-E)/d);for(let A=0;A<m;++A){let R=Math.max(0,Math.ceil((b-A)/h)),T=Math.min(p.outWidth,(p.inWidth+b-A)/h);for(let L=0;L<p.outChannels;++L){let V=Math.trunc(L/v),G=L%v,j=0;for(let H=0;H<p.batchSize;++H)for(let Z=$;Z<F;++Z){let J=E+Z*d-y;for(let ee=R;ee<T;++ee){let re=A+ee*h-b;j+=k.get(H,J,re,V)*C.get(H,Z,ee,L)}}g.set(j,E,A,V,G)}}}return n.makeTensorInfo(g.shape,g.dtype,g.values)}var l8={kernelName:wf,backendName:"cpu",kernelFunc:u8};function d8(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{strides:o,dilations:i,pad:c,dimRoundingMode:u,inputShape:l}=r;be([s,a],"depthwiseConv2DNativeBackpropInput");let p=w.computeStrides(s.shape),d=w.computeStrides(a.shape),h=N.computeConv2DInfo(l,a.shape,o,i,c,u,!0),f=new Ht(h.inShape,"float32"),m=f.values,[g,b,y]=f.strides,v=n.data.get(s.dataId).values,[x,k,S]=p,C=n.data.get(a.dataId).values,[E,$,F]=d,{batchSize:A,filterHeight:R,filterWidth:T,inChannels:L,inHeight:V,inWidth:G,outChannels:j,outHeight:H,outWidth:Z,strideHeight:J,strideWidth:ee}=h,re=R-1-h.padInfo.top,te=T-1-h.padInfo.left,ie=j/L;for(let ne=0;ne<A;++ne)for(let le=0;le<L;++le)for(let ue=0;ue<V;++ue){let ye=ue-re,ve=Math.max(0,Math.ceil(ye/J)),Se=Math.min(H,(R+ye)/J);for(let Le=0;Le<G;++Le){let je=Le-te,nt=Math.max(0,Math.ceil(je/ee)),rt=Math.min(Z,(T+je)/ee),st=0;for(let Ze=ve;Ze<Se;++Ze){let ze=Ze*J-ye;for(let dt=nt;dt<rt;++dt){let qn=dt*ee-je,At=x*ne+k*Ze+S*dt,wr=E*(R-1-ze)+$*(T-1-qn)+F*le;for(let on=0;on<ie;++on){let Ir=le*ie+on,vn=v[At+Ir],jn=C[wr+on];st+=vn*jn}}}m[g*ne+b*ue+y*Le+le]=st}}return n.makeTensorInfo(f.shape,f.dtype,f.values)}var p8={kernelName:If,backendName:"cpu",kernelFunc:d8};function h8(e){let{inputs:t,backend:n}=e,{x:r}=t,s=w.sizeFromShape(r.shape),a=n.data.get(r.dataId).values,o=Me([s,s],r.dtype),i=o.values;for(let u=0;u<a.length;u++)i[u*s+u]=a[u];let c=[...r.shape,...r.shape];return n.makeTensorInfo(c,o.dtype,o.values)}var f8={kernelName:kf,backendName:"cpu",kernelFunc:h8},m8={kernelName:md,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:s}=e,{strides:a,pad:o,dilations:i}=n,c=t,u=c.data.get(r.dataId).values,l=r.shape.length,p=c.data.get(s.dataId).values,d=s.shape.length,{batchSize:h,inHeight:f,inWidth:m,inChannels:g,outHeight:b,outWidth:y,padInfo:v,strideHeight:x,strideWidth:k,filterHeight:S,filterWidth:C,dilationHeight:E,dilationWidth:$,outShape:F}=N.computeDilation2DInfo(r.shape,s.shape,a,o,"NHWC",i),A=w.sizeFromShape(F),R=F.length,T=w.getArrayFromDType(r.dtype,A);for(let V=0;V<h;++V)for(let G=0;G<b;++G){let j=G*x-v.top;for(let H=0;H<y;++H){let Z=H*k-v.left;for(let J=0;J<g;++J){let ee=Number.MIN_SAFE_INTEGER;for(let te=0;te<S;++te){let ie=j+te*E;if(ie>=0&&ie<f)for(let ne=0;ne<C;++ne){let le=Z+ne*$;if(le>=0&&le<m){let ue=w.locToIndex([V,ie,le,J],l,w.computeStrides(r.shape)),ye=w.locToIndex([te,ne,J],d,w.computeStrides(s.shape)),ve=u[ue]+p[ye];ve>ee&&(ee=ve)}}}let re=w.locToIndex([V,G,H,J],R,w.computeStrides(F));T[re]=ee}}}return{dataId:c.write(w.toTypedArray(T,r.dtype),F,r.dtype),shape:F,dtype:r.dtype}}},g8={kernelName:Dh,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:s,dy:a}=e,{strides:o,pad:i,dilations:c}=n,u=t,l=w.toNestedArray(r.shape,u.data.get(r.dataId).values),p=w.toNestedArray(s.shape,u.data.get(s.dataId).values),{batchSize:d,inHeight:h,inWidth:f,inChannels:m,outHeight:g,outWidth:b,padInfo:y,strideHeight:v,strideWidth:x,filterHeight:k,filterWidth:S,dilationHeight:C,dilationWidth:E,outShape:$}=N.computeDilation2DInfo(r.shape,s.shape,o,i,"NHWC",c);w.assert(a.rank===$.length,()=>`Error in ${Dh}, dy must have the same rank as output ${$.length}, but got ${a.rank}`);let F=w.toNestedArray($,u.data.get(a.dataId).values),A=w.makeZerosNestedTypedArray(s.shape,s.dtype);for(let T=0;T<d;++T)for(let L=0;L<g;++L){let V=L*v-y.top;for(let G=0;G<b;++G){let j=G*x-y.left;for(let H=0;H<m;++H){let Z=Number.MIN_SAFE_INTEGER,J=0,ee=0;for(let re=0;re<k;++re){let te=V+re*C;if(te>=0&&te<h)for(let ie=0;ie<S;++ie){let ne=j+ie*E;if(ne>=0&&ne<f){let le=l[T][te][ne][H]+p[re][ie][H];le>Z&&(Z=le,J=re,ee=ie)}}}A[J][ee][H]+=F[T][L][G][H]}}}return{dataId:u.write(w.toTypedArray(A,r.dtype),s.shape,s.dtype),shape:s.shape,dtype:s.dtype}}},b8={kernelName:$h,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:s,dy:a}=e,{strides:o,pad:i,dilations:c}=n,u=t,l=w.toNestedArray(r.shape,u.data.get(r.dataId).values),p=w.toNestedArray(s.shape,u.data.get(s.dataId).values),{batchSize:d,inHeight:h,inWidth:f,inChannels:m,outHeight:g,outWidth:b,padInfo:y,strideHeight:v,strideWidth:x,filterHeight:k,filterWidth:S,dilationHeight:C,dilationWidth:E,outShape:$}=N.computeDilation2DInfo(r.shape,s.shape,o,i,"NHWC",c);w.assert(a.rank===$.length,()=>`Error in ${$h}, dy must have the same rank as output ${$.length}, but got ${a.rank}`);let F=w.toNestedArray($,u.data.get(a.dataId).values),A=w.makeZerosNestedTypedArray(r.shape,r.dtype);for(let T=0;T<d;++T)for(let L=0;L<g;++L){let V=L*v-y.top;for(let G=0;G<b;++G){let j=G*x-y.left;for(let H=0;H<m;++H){let Z=Number.MIN_SAFE_INTEGER,J=V<0?0:V,ee=j<0?0:j;for(let re=0;re<k;++re){let te=V+re*C;if(te>=0&&te<h)for(let ie=0;ie<S;++ie){let ne=j+ie*E;if(ne>=0&&ne<f){let le=l[T][te][ne][H]+p[re][ie][H];le>Z&&(Z=le,J=te,ee=ne)}}}A[T][J][ee][H]+=F[T][L][G][H]}}}return{dataId:u.write(w.toTypedArray(A,r.dtype),r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}};function ap(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;be(s,"sum");let i;s.dtype==="bool"?i=ha({inputs:{x:s},backend:n,attrs:{dtype:"int32"}}):i=ds({inputs:{x:s},backend:n});let c=i.shape.length,u=w.parseAxisParam(a,i.shape),l=N.getAxesPermutation(u,c),p=u,d=i;l!=null&&(d=Vn({inputs:{x:i},backend:n,attrs:{perm:l}}),p=N.getInnerMostAxes(p.length,c)),N.assertAxesAreInnerMostDims("sum",p,d.shape.length);let[h,f]=N.computeOutAndReduceShapes(d.shape,p),m=N.upcastType(d.dtype,"int32"),g=Jh(n,h,m),b=w.sizeFromShape(f),y=n.data.get(g.dataId).values,v=n.data.get(d.dataId).values;for(let x=0;x<y.length;++x){let k=x*b,S=0;for(let C=0;C<b;++C)S+=v[k+C];y[x]=S}if(o){let x=N.expandShapeToKeepDim(g.shape,u),k=g;g=ft({inputs:{x:g},backend:n,attrs:{shape:x}}),n.disposeIntermediateTensorInfo(k)}return n.disposeIntermediateTensorInfo(i),l!=null&&n.disposeIntermediateTensorInfo(d),g}var y8={kernelName:ci,backendName:"cpu",kernelFunc:ap};function v8(e){let{inputs:t,backend:n,attrs:r}=e,{equation:s}=r,a=t,{allDims:o,summedDims:i,idDims:c}=N.decodeEinsumEquation(s,a.length);N.checkEinsumDimSizes(o.length,c,a);let{path:u,steps:l}=N.getEinsumComputePath(i,c),p=l.length,d=null,h=o.length,f=[];for(let m=0;m<p;++m){for(let g of l[m]){let{permutationIndices:b,expandDims:y}=N.getEinsumPermutation(h,c[g]),v;N.isIdentityPermutation(b)?v=a[g]:(v=Vn({inputs:{x:a[g]},backend:n,attrs:{perm:b}}),f.push(v));let x=v.shape.slice();for(let k=0;k<y.length;++k)x.splice(y[k],0,1);w.arraysEqual(v.shape,x)||(v=ft({inputs:{x:v},backend:n,attrs:{shape:x}}),f.push(v)),d===null?d=v:(d=Km({inputs:{a:v,b:d},backend:n}),f.push(d))}m<p-1&&(u[m]>=0&&(d=ap({inputs:{x:d},backend:n,attrs:{axis:u[m]-(o.length-h),keepDims:!1}}),f.push(d)),h--)}for(let m of f)m!==d&&n.disposeIntermediateTensorInfo(m);return d}var x8={kernelName:Sf,backendName:"cpu",kernelFunc:v8};function w8(e){let{inputs:t,backend:n}=e,{dy:r,y:s}=t;be([r,s],"eluGrad");let a=new Float32Array(w.sizeFromShape(s.shape)),o=n.data.get(s.dataId).values,i=n.data.get(r.dataId).values;for(let c=0;c<o.length;++c){let u=o[c];u>=1?a[c]=i[c]:a[c]=i[c]*(u+1)}return n.makeTensorInfo(s.shape,"float32",a)}var I8={kernelName:Tf,backendName:"cpu",kernelFunc:w8},k8=N.ERF_P,S8=N.ERF_A1,T8=N.ERF_A2,C8=N.ERF_A3,N8=N.ERF_A4,_8=N.ERF_A5,E8=at(Kc,e=>{let t=Math.sign(e),n=Math.abs(e),r=1/(1+k8*n);return t*(1-((((_8*r+N8)*r+C8)*r+T8)*r+S8)*r*Math.exp(-n*n))}),A8={kernelName:Kc,backendName:"cpu",kernelFunc:E8};function tf(e){let{inputs:t,backend:n,attrs:r}=e,{input:s}=t,{dim:a}=r,o=s.shape.length,i=s.shape.slice(),c=a;return a<0&&(w.assert(-(o+1)<=a,()=>`Axis must be in the interval [${-(o+1)}, ${o}]`),c=o+a+1),i.splice(c,0,1),ft({inputs:{x:s},backend:n,attrs:{shape:i}})}var $8={kernelName:Yc,backendName:"cpu",kernelFunc:tf},D8=Vt((e,t)=>e/t),j0=nn(Eo,D8),pv={kernelName:Eo,backendName:"cpu",kernelFunc:j0};function o2(e,t,n){let r=e.shape,s=r[0],a=r[1],o=n.data.get(e.dataId),i=o.complexTensorInfos.real,c=o.complexTensorInfos.imag,u=[s,a],l=w.sizeFromShape(u),p=w.getTypedArrayFromDType("float32",l),d=w.getTypedArrayFromDType("float32",l);for(let g=0;g<s;g++){let b=po({inputs:{x:i},backend:n,attrs:{begin:[g,0],size:[1,a]}}),y=po({inputs:{x:c},backend:n,attrs:{begin:[g,0],size:[1,a]}}),v=Jn({inputs:{real:b,imag:y},backend:n}),{real:x,imag:k}=F8(v,t,n),S=N.mergeRealAndImagArrays(x,k);for(let C=0;C<a;C++){let E=N.getComplexWithIndex(S,C);p[g*a+C]=E.real,d[g*a+C]=E.imag}n.disposeIntermediateTensorInfo(b),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(v)}let h=n.makeTensorInfo(u,"float32",p),f=n.makeTensorInfo(u,"float32",d),m=Jn({inputs:{real:h,imag:f},backend:n});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),m}function F8(e,t,n){let r=w.sizeFromShape(e.shape),s=n.data.get(e.dataId),a=n.data.get(s.complexTensorInfos.real.dataId).values,o=n.data.get(s.complexTensorInfos.imag.dataId).values;if(R8(r)){let i=hv(a,o,r,t,n),c=[e.shape[0],e.shape[1]];if(t){let u=n.makeTensorInfo(c,"float32",i.real),l=n.makeTensorInfo(c,"float32",i.imag),p=n.makeTensorInfo([],"float32",w.createScalarValue(r,"float32")),d=ds({inputs:{x:p},backend:n}),h=pv.kernelFunc({inputs:{a:u,b:p},backend:n}),f=pv.kernelFunc({inputs:{a:l,b:d},backend:n}),m=n.data.get(h.dataId).values,g=n.data.get(f.dataId).values;return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),{real:m,imag:g}}return i}else{let i=N.mergeRealAndImagArrays(a,o),c=P8(i,r,t);return N.splitRealAndImagArrays(c)}}function R8(e){return(e&e-1)===0}function hv(e,t,n,r,s){if(n===1)return{real:e,imag:t};let a=N.mergeRealAndImagArrays(e,t),o=n/2,i=N.complexWithEvenIndex(a),c=i.real,u=i.imag,l=[c.length],p=s.makeTensorInfo(l,"float32",c),d=s.makeTensorInfo(l,"float32",u),h=Jn({inputs:{real:p,imag:d},backend:s}),f=N.complexWithOddIndex(a),m=f.real,g=f.imag,b=[m.length],y=s.makeTensorInfo(b,"float32",m),v=s.makeTensorInfo(b,"float32",g),x=Jn({inputs:{real:y,imag:v},backend:s}),k=hv(c,u,o,r,s),S=k.real,C=k.imag,E=[S.length],$=s.makeTensorInfo(E,"float32",S),F=s.makeTensorInfo(E,"float32",C),A=Jn({inputs:{real:$,imag:F},backend:s}),R=hv(m,g,o,r,s),T=R.real,L=R.imag,V=[T.length],G=s.makeTensorInfo(V,"float32",T),j=s.makeTensorInfo(V,"float32",L),H=Jn({inputs:{real:G,imag:j},backend:s}),Z=N.exponents(n,r),J=[Z.real.length],ee=s.makeTensorInfo(J,"float32",Z.real),re=s.makeTensorInfo(J,"float32",Z.imag),te=Jn({inputs:{real:ee,imag:re},backend:s}),ie=Km({inputs:{a:te,b:H},backend:s}),ne=_c({inputs:{a:A,b:ie},backend:s}),le=H0({inputs:{a:A,b:ie},backend:s}),ue=lo({inputs:{input:ne},backend:s}),ye=lo({inputs:{input:le},backend:s}),ve=Ec({inputs:{input:ne},backend:s}),Se=Ec({inputs:{input:le},backend:s}),Le=Ac({inputs:[ue,ye],backend:s,attrs:{axis:0}}),je=Ac({inputs:[ve,Se],backend:s,attrs:{axis:0}}),nt=s.data.get(Le.dataId).values,rt=s.data.get(je.dataId).values;return s.disposeIntermediateTensorInfo(p),s.disposeIntermediateTensorInfo(d),s.disposeIntermediateTensorInfo(h),s.disposeIntermediateTensorInfo(y),s.disposeIntermediateTensorInfo(v),s.disposeIntermediateTensorInfo(x),s.disposeIntermediateTensorInfo($),s.disposeIntermediateTensorInfo(F),s.disposeIntermediateTensorInfo(A),s.disposeIntermediateTensorInfo(G),s.disposeIntermediateTensorInfo(j),s.disposeIntermediateTensorInfo(H),s.disposeIntermediateTensorInfo(ee),s.disposeIntermediateTensorInfo(re),s.disposeIntermediateTensorInfo(te),s.disposeIntermediateTensorInfo(ie),s.disposeIntermediateTensorInfo(ne),s.disposeIntermediateTensorInfo(le),s.disposeIntermediateTensorInfo(ue),s.disposeIntermediateTensorInfo(ve),s.disposeIntermediateTensorInfo(ye),s.disposeIntermediateTensorInfo(Se),s.disposeIntermediateTensorInfo(Le),s.disposeIntermediateTensorInfo(je),{real:nt,imag:rt}}function P8(e,t,n){let r=new Float32Array(t*2);for(let s=0;s<t;s++){let a=0,o=0;for(let i=0;i<t;i++){let c=N.exponent(s*i,t,n),u=N.getComplexWithIndex(e,i);a+=u.real*c.real-u.imag*c.imag,o+=u.real*c.imag+u.imag*c.real}n&&(a/=t,o/=t),N.assignToTypedArray(r,a,o,s)}return r}function O8(e){let{inputs:t,backend:n}=e,{input:r}=t,s=w.sizeFromShape(r.shape),a=r.shape[r.shape.length-1],o=s/a,i=ft({inputs:{x:r},backend:n,attrs:{shape:[o,a]}}),c=o2(i,!1,n),u=ft({inputs:{x:c},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(c),u}var M8={kernelName:Cf,backendName:"cpu",kernelFunc:O8};function K0(e){let{backend:t,attrs:n}=e,{shape:r,value:s,dtype:a}=n,o=a||w.inferDtype(s),i=w.getArrayFromDType(o,w.sizeFromShape(r));return z8(i,s,o),t.makeTensorInfo(r,o,i)}var L8={kernelName:gd,backendName:"cpu",kernelFunc:K0};function z8(e,t,n){e.fill(t)}var B8={kernelName:Jc,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,s=n,a=w.getTypedArrayFromDType(r.dtype,w.sizeFromShape(r.shape)),[o,i,c,u]=r.shape,l=s.data.get(r.dataId).values;for(let d=0;d<o;d++){let h=d*c*i*u;for(let f=0;f<i;f++){let m=f*(c*u);for(let g=0;g<c;g++){let b=g*u;for(let y=0;y<u;y++){let v=Math.round(c-g-1),x=h+m+b+y,k=l[x];if(v>=0&&v<c){let S=v*u,C=h+m+S+y;k=l[C]}a[x]=k}}}}return{dataId:s.write(a,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},W8=Vt((e,t)=>Math.floor(e/t)),V8=nn(Fo,W8,null,"int32"),U8={kernelName:Fo,backendName:"cpu",kernelFunc:V8};function G8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dataFormat:l,dilations:p,dimRoundingMode:d,activation:h,leakyreluAlpha:f}=r,m=s2({inputs:{x:s,filter:a},backend:n,attrs:{strides:c,pad:u,dataFormat:l,dilations:p,dimRoundingMode:d}});if(o){let g=m;if(l==="NCHW"&&o.shape.length===1&&o.shape[0]!==1){let b=ft({inputs:{x:o},backend:n,attrs:{shape:[o.shape[0],1,1]}});m=_c({inputs:{a:m,b},backend:n}),n.disposeIntermediateTensorInfo(b)}else m=_c({inputs:{a:m,b:o},backend:n});n.disposeIntermediateTensorInfo(g)}if(h){let g=m;if(l==="NCHW"&&h==="prelu"&&i.shape.length===1&&i.shape[0]!==1){let b=ft({inputs:{x:i},backend:n,attrs:{shape:[i.shape[0],1,1]}});m=ef(n,m,h,b,f),n.disposeIntermediateTensorInfo(b)}else m=ef(n,m,h,i,f);n.disposeIntermediateTensorInfo(g)}return m}var H8={kernelName:to,backendName:"cpu",kernelFunc:G8};function q8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dataFormat:l,dilations:p,dimRoundingMode:d,activation:h,leakyreluAlpha:f}=r,m=a2({inputs:{x:s,filter:a},backend:n,attrs:{strides:c,pad:u,dataFormat:l,dilations:p,dimRoundingMode:d}});if(o){let g=m;m=_c({inputs:{a:m,b:o},backend:n}),n.disposeIntermediateTensorInfo(g)}if(h){let g=m;m=ef(n,m,h,i,f),n.disposeIntermediateTensorInfo(g)}return m}var j8={kernelName:no,backendName:"cpu",kernelFunc:q8};function K8(e){let{inputs:t,backend:n}=e,{params:r,indices:s}=t,a=w.sizeFromShape(r.shape),o=s.shape,i=o[o.length-1],[c,u,l,p]=N.prepareAndValidate(r,s);if(u===0)return n.makeTensorInfo(c,r.dtype,[]);let d=n.data.get(s.dataId).values,h=n.bufferSync(r),f=I_(d,h,r.dtype,u,i,l,p,r.shape,a);return n.makeTensorInfo(c,r.dtype,f.values)}var X8={kernelName:eu,backendName:"cpu",kernelFunc:K8};function Y8(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,indices:a}=t,{axis:o,batchDims:i}=r;be([s,a],"gatherV2");let c=w.parseAxisParam(o,s.shape)[0],u=n.data.get(a.dataId).values,l=s.shape[c];for(let x=0;x<u.length;++x){let k=u[x];w.assert(k<=l-1&&k>=0,()=>`GatherV2: the index value ${k} is not in [0, ${l-1}]`)}let p=i;i==null&&(p=0);let d=w.sizeFromShape(a.shape),h=N.segment_util.collectGatherOpShapeInfo(s,a,c,p),f=ft({inputs:{x:s},backend:n,attrs:{shape:[h.batchSize,h.outerSize,h.dimSize,h.sliceSize]}}),m=ft({inputs:{x:a},backend:n,attrs:{shape:[h.batchSize,d/h.batchSize]}}),g=[h.batchSize,h.outerSize,d/h.batchSize,h.sliceSize],b=n.bufferSync(m),y=n.bufferSync(f),v=k_(y,b,g);return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),n.makeTensorInfo(h.outputShape,v.dtype,v.values)}var Z8={kernelName:Qc,backendName:"cpu",kernelFunc:Y8};function J8(e){let{inputs:t,backend:n}=e,{input:r}=t,s=w.sizeFromShape(r.shape),a=r.shape[r.shape.length-1],o=s/a,i=ft({inputs:{x:r},backend:n,attrs:{shape:[o,a]}}),c=o2(i,!0,n),u=ft({inputs:{x:c},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(c),u}var Q8={kernelName:Nf,backendName:"cpu",kernelFunc:J8},eK=at(nu,e=>Number.isFinite(e)?1:0,"bool"),tK={kernelName:nu,backendName:"cpu",kernelFunc:eK},nK=at(ru,e=>Math.abs(e)===1/0?1:0,"bool"),rK={kernelName:ru,backendName:"cpu",kernelFunc:nK},sK=at(su,e=>Number.isNaN(e)?1:0,"bool"),aK={kernelName:su,backendName:"cpu",kernelFunc:sK};function oK(e){let{backend:t,attrs:n}=e,{start:r,stop:s,num:a}=n,o=__(r,s,a);return t.makeTensorInfo([o.length],"float32",o)}var iK={kernelName:Ef,backendName:"cpu",kernelFunc:oK},cK=at(iu,e=>Math.log1p(e)),uK={kernelName:iu,backendName:"cpu",kernelFunc:cK},lK=Vt((e,t)=>e&&t),dK=nn(cu,lK,null,"bool"),pK={kernelName:cu,backendName:"cpu",kernelFunc:dK},hK=at(uu,e=>e?0:1,"bool"),fK={kernelName:uu,backendName:"cpu",kernelFunc:hK},mK=Vt((e,t)=>e||t),gK=nn(lu,mK,null,"bool"),bK={kernelName:lu,backendName:"cpu",kernelFunc:gK};function yK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{depthRadius:a,bias:o,alpha:i,beta:c}=r;be(s,"LRN");let u=s.shape[3],l=u-1,p=n.data.get(s.dataId).values,d=w.sizeFromShape(s.shape),h=new Float32Array(d);function f(m){let g=m%u,b=m-g+Math.max(0,g-a),y=m-g+Math.min(g+a,l),v=0;for(;b<=y;b++){let x=p[b];v+=x*x}return v}for(let m=0;m<d;m++){let g=f(m),b=p[m]*Math.pow(o+i*g,-c);h[m]=b}return n.makeTensorInfo(s.shape,s.dtype,h)}var vK={kernelName:bd,backendName:"cpu",kernelFunc:yK};function xK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,y:a,dy:o}=t,{depthRadius:i,bias:c,alpha:u,beta:l}=r;be(o,"LRNGrad");let p=w.sizeFromShape(o.shape),d=o.shape[3],h=n.data.get(o.dataId).values,f=n.data.get(s.dataId).values,m=n.data.get(a.dataId).values,g=new Float32Array(p),b=p;for(let y=0;y<b;y++){let v=y%d,x=y-v+Math.max(0,v-i),k=y-v+Math.min(d,v+i+1),S=0;for(let C=x;C<k;C++)S+=Math.pow(f[C],2);S=u*S+c;for(let C=x;C<k;C++){let E=-2*u*l*f[C]*m[y]/S;y===C&&(E+=Math.pow(S,-l)),E*=h[y],g[C]+=E}}return n.makeTensorInfo(o.shape,s.dtype,g)}var wK={kernelName:Af,backendName:"cpu",kernelFunc:xK};function i2(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{reductionIndices:a,keepDims:o}=r,i=n,c=s.shape,u=c.length,l=w.parseAxisParam(a,c),p=l,d=N.getAxesPermutation(p,u),h=i.data.get(s.dataId).values;if(d!=null){let x=new Array(u);for(let k=0;k<x.length;k++)x[k]=c[d[k]];h=z0(h,c,s.dtype,d,x),p=N.getInnerMostAxes(p.length,u),c=x}be(s,"max"),N.assertAxesAreInnerMostDims("max",p,u);let[f,m]=N.computeOutAndReduceShapes(c,p),g=w.sizeFromShape(m),b=A_(h,g,f,s.dtype),y=i.write(b,f,s.dtype),v=f;return o&&(v=N.expandShapeToKeepDim(f,l)),{dataId:y,shape:v,dtype:s.dtype}}var IK={kernelName:zo,backendName:"cpu",kernelFunc:i2};function kK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t;be(s,"maxPool");let{filterSize:a,strides:o,pad:i,dimRoundingMode:c}=r,u=1;w.assert(N.eitherStridesOrDilationsAreOne(o,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${o} and dilations '${u}'`);let l=N.computePool2DInfo(s.shape,a,o,u,i,c),p;if(l.filterWidth===1&&l.filterHeight===1&&w.arraysEqual(l.inShape,l.outShape))p=ds({inputs:{x:s},backend:n});else{let d=n.data.get(s.dataId).values,h=w.computeStrides(s.shape),f=q0(d,s.shape,s.dtype,h,l,"max");p=n.makeTensorInfo(l.outShape,s.dtype,f.values)}return p}var SK={kernelName:Wo,backendName:"cpu",kernelFunc:kK};function TK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{filterSize:a,strides:o,pad:i,dimRoundingMode:c,dataFormat:u}=r;be(s,"maxPool3d");let l=N.computePool3DInfo(s.shape,a,o,1,i,c,u),p=n.data.get(s.dataId).values,d=r2(p,s.shape,s.dtype,w.computeStrides(s.shape),l,"max");return n.makeTensorInfo(d.shape,"float32",d.values)}var CK={kernelName:yd,backendName:"cpu",kernelFunc:TK};function NK(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,{filterSize:o,strides:i,pad:c,dimRoundingMode:u}=r;be([s,a],"maxPool3DGrad");let l=N.computePool3DInfo(a.shape,o,i,1,c,u),p=n.bufferSync(a),d=fj(p,l),h=l.strideDepth,f=l.strideHeight,m=l.strideWidth,g=l.dilationDepth,b=l.dilationHeight,y=l.dilationWidth,v=l.effectiveFilterDepth,x=l.effectiveFilterHeight,k=l.effectiveFilterWidth,S=v-1-l.padInfo.front,C=k-1-l.padInfo.left,E=x-1-l.padInfo.top,$=Me(a.shape,"float32"),F=n.bufferSync(s);for(let A=0;A<l.batchSize;++A)for(let R=0;R<l.inChannels;++R)for(let T=0;T<l.inDepth;++T)for(let L=0;L<l.inHeight;++L)for(let V=0;V<l.inWidth;++V){let G=T-S,j=L-E,H=V-C,Z=0;for(let J=0;J<v;J+=g){let ee=(G+J)/h;if(!(ee<0||ee>=l.outDepth||Math.floor(ee)!==ee))for(let re=0;re<x;re+=b){let te=(j+re)/f;if(!(te<0||te>=l.outHeight||Math.floor(te)!==te))for(let ie=0;ie<k;ie+=y){let ne=(H+ie)/m;if(ne<0||ne>=l.outWidth||Math.floor(ne)!==ne)continue;let le=v*x*k-1-d.get(A,ee,te,ne,R),ue=J*x*k+re*k+ie,ye=le===ue?1:0;if(ye===0)continue;let ve=F.get(A,ee,te,ne,R);Z+=ve*ye}}}$.set(Z,A,T,L,V,R)}return n.makeTensorInfo($.shape,$.dtype,$.values)}var _K={kernelName:Df,backendName:"cpu",kernelFunc:NK};function EK(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a,output:o}=t,i=a;be([a,o],"maxPoolGrad");let{filterSize:c,strides:u,pad:l,dimRoundingMode:p}=r,d=N.computePool2DInfo(i.shape,c,u,1,l,p),h=n.data.get(i.dataId).values,f=Me(d.outShape,i.dtype,n2(h,i.shape,i.dtype,d).values),m=d.strideHeight,g=d.strideWidth,b=d.dilationHeight,y=d.dilationWidth,v=d.effectiveFilterHeight,x=d.effectiveFilterWidth,k=x-1-d.padInfo.left,S=v-1-d.padInfo.top,C=Me(i.shape,"float32"),E=n.data.get(s.dataId).values,$=Me(s.shape,"float32",E);for(let F=0;F<d.batchSize;++F)for(let A=0;A<d.inChannels;++A)for(let R=0;R<d.inHeight;++R)for(let T=0;T<d.inWidth;++T){let L=R-S,V=T-k,G=0;for(let j=0;j<v;j+=b){let H=(L+j)/m;if(!(H<0||H>=d.outHeight||Math.floor(H)!==H))for(let Z=0;Z<x;Z+=y){let J=(V+Z)/g;if(J<0||J>=d.outWidth||Math.floor(J)!==J)continue;let ee=v*x-1-f.get(F,H,J,A),re=j*x+Z,te=ee===re?1:0;if(te===0)continue;let ie=$.get(F,H,J,A);G+=ie*te}}C.set(G,F,R,T,A)}return n.makeTensorInfo(C.shape,C.dtype,C.values)}var AK={kernelName:$f,backendName:"cpu",kernelFunc:EK};function $K(e,t,n,r,s){let a=w.computeStrides(t),o=q0(e,t,n,a,s,"max"),i=n2(e,t,n,s,!0,r);return[o.values,i.values]}var DK={kernelName:Ff,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{filterSize:s,strides:a,pad:o,includeBatchInIndex:i}=t,c=n;be(r,"MaxPoolWithArgmax");let u=c.data.get(r.dataId).values,l=N.computePool2DInfo(r.shape,s,a,[1,1],o),[p,d]=$K(u,r.shape,r.dtype,i,l),h=c.write(p,l.outShape,r.dtype),f=c.write(d,l.outShape,r.dtype);return[{dataId:h,shape:l.outShape,dtype:r.dtype},{dataId:f,shape:l.outShape,dtype:"int32"}]}};function FK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r,i=w.parseAxisParam(a,s.shape),u=N.computeOutAndReduceShapes(s.shape,i)[1],l=w.sizeFromShape(u),p=[],d=n.makeTensorInfo([],"float32",new Float32Array([l]));p.push(d);let h=ha({inputs:{x:s},backend:n,attrs:{dtype:"float32"}});p.push(h);let f=j0({inputs:{a:h,b:d},backend:n});p.push(f);let m=ap({inputs:{x:f},backend:n,attrs:{axis:a,keepDims:o}});return p.forEach(g=>n.disposeIntermediateTensorInfo(g)),m}var RK={kernelName:Vo,backendName:"cpu",kernelFunc:FK};function PK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;be(s,"min");let i=w.parseAxisParam(a,s.shape),c=i,u=N.getAxesPermutation(c,s.shape.length),l=s;u!=null&&(l=Vn({inputs:{x:s},backend:n,attrs:{perm:u}}),c=N.getInnerMostAxes(c.length,s.shape.length)),N.assertAxesAreInnerMostDims("min",c,l.shape.length);let[p,d]=N.computeOutAndReduceShapes(l.shape,c),h=w.sizeFromShape(d),f=w.makeZerosTypedArray(w.sizeFromShape(p),l.dtype),m=n.data.get(l.dataId).values;for(let b=0;b<f.length;++b){let y=b*h,v=m[y];for(let x=0;x<h;++x){let k=m[y+x];(Number.isNaN(k)||k<v)&&(v=k)}f[b]=v}u!=null&&n.disposeIntermediateTensorInfo(l);let g=n.makeTensorInfo(p,l.dtype,f);if(o){let b=N.expandShapeToKeepDim(p,i),y=ft({inputs:{x:g},backend:n,attrs:{shape:b}});return n.disposeIntermediateTensorInfo(g),y}return g}var OK={kernelName:Uo,backendName:"cpu",kernelFunc:PK};function MK(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{paddings:a,mode:o}=r;be(s,"mirrorPad");let i=a.map((v,x)=>v[0]+s.shape[x]+v[1]),c=a.map(v=>v[0]),u=a.map((v,x)=>v[0]+s.shape[x]),l=o==="reflect"?0:1,p=n.data.get(s.dataId).values,d=s.shape.length,h=w.computeStrides(s.shape),f=w.sizeFromShape(i),m=i.length,g=w.computeStrides(i),b=w.getTypedArrayFromDType(s.dtype,f);for(let v=0;v<f;v++){let x=w.indexToLoc(v,m,g);for(let S=0;S<m;S++)x[S]<c[S]?x[S]=c[S]*2-x[S]-l:x[S]>=u[S]&&(x[S]=(u[S]-1)*2-x[S]+l);x=x.map((S,C)=>S-c[C]);let k=w.locToIndex(x,d,h);b[v]=p[k]}return{dataId:n.write(b,i,s.dtype),shape:i,dtype:s.dtype}}var LK={kernelName:Ho,backendName:"cpu",kernelFunc:MK},zK=Vt((e,t)=>{let n=e%t;return e<0&&t<0||e>=0&&t>=0?n:(n+t)%t}),BK=nn(du,zK),WK={kernelName:du,backendName:"cpu",kernelFunc:BK},VK=ma(cf());function c2(e){let{inputs:t,backend:n,attrs:r}=e,{logits:s}=t,{dim:a}=r,o=s.shape.length,i=a;if(i===-1&&(i=o-1),i!==o-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${o} and dim was ${i}`);let c=w.parseAxisParam([i],s.shape),u=i2({inputs:{x:s},backend:n,attrs:{reductionIndices:c,keepDims:!1}}),l=N.expandShapeToKeepDim(u.shape,c),p=ft({inputs:{x:u},backend:n,attrs:{shape:l}}),d=H0({inputs:{a:s,b:p},backend:n}),h=v_({inputs:{x:d},backend:n}),f=ap({inputs:{x:h},backend:n,attrs:{axis:c,keepDims:!1}}),m=ft({inputs:{x:f},backend:n,attrs:{shape:l}}),g=j0({inputs:{a:h,b:m},backend:n});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}var UK={kernelName:ui,backendName:"cpu",kernelFunc:c2};function GK(e){let{inputs:t,backend:n,attrs:r}=e,{logits:s}=t,{numSamples:a,seed:o,normalized:i}=r;be(s,"multinomial");let c=i?s:c2({inputs:{logits:s},backend:n,attrs:{dim:-1}}),u=c.shape[0],l=c.shape[1],p=n.data.get(c.dataId).values,d=[u,a],h=w.makeZerosTypedArray(w.sizeFromShape(d),"int32");for(let f=0;f<u;++f){let m=f*l,g=new Float32Array(l-1);g[0]=p[m];for(let v=1;v<g.length;++v)g[v]=g[v-1]+p[m+v];let b=VK.alea(o.toString()),y=f*a;for(let v=0;v<a;++v){let x=b();h[y+v]=g.length;for(let k=0;k<g.length;k++)if(x<g[k]){h[y+v]=k;break}}}return i||n.disposeIntermediateTensorInfo(c),n.makeTensorInfo(d,"int32",h)}var HK={kernelName:Rf,backendName:"cpu",kernelFunc:GK},qK=fs.nonMaxSuppressionV3Impl;function jK(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c}=r;be(s,"NonMaxSuppression");let u=n.data.get(s.dataId).values,l=n.data.get(a.dataId).values,{selectedIndices:p}=qK(u,l,o,i,c);return n.makeTensorInfo([p.length],"int32",new Int32Array(p))}var KK={kernelName:fu,backendName:"cpu",kernelFunc:jK},XK=fs.nonMaxSuppressionV4Impl;function YK(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c,padToMaxOutputSize:u}=r;be(s,"NonMaxSuppressionPadded");let l=n.data.get(s.dataId).values,p=n.data.get(a.dataId).values,{selectedIndices:d,validOutputs:h}=XK(l,p,o,i,c,u);return[n.makeTensorInfo([d.length],"int32",new Int32Array(d)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var ZK={kernelName:mu,backendName:"cpu",kernelFunc:YK},JK=fs.nonMaxSuppressionV5Impl;function QK(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c,softNmsSigma:u}=r;be(s,"NonMaxSuppressionWithScore");let l=n.data.get(s.dataId).values,p=n.data.get(a.dataId).values,d=o,h=i,f=c,m=u,{selectedIndices:g,selectedScores:b}=JK(l,p,d,h,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([b.length],"float32",new Float32Array(b))]}var eX={kernelName:gu,backendName:"cpu",kernelFunc:QK};function tX(e){let{inputs:t,backend:n,attrs:r}=e,{indices:s}=t,{dtype:a,depth:o,onValue:i,offValue:c}=r;be(s,"oneHot");let u=w.sizeFromShape(s.shape),l=new Float32Array(u*o);l.fill(c);let p=n.data.get(s.dataId).values;for(let d=0;d<u;++d)p[d]>=0&&p[d]<o&&(l[d*o+p[d]]=i);return n.makeTensorInfo([...s.shape,o],a,l)}var nX={kernelName:jo,backendName:"cpu",kernelFunc:tX};function nf(e){let{inputs:t,backend:n}=e,{x:r}=t;if(r.dtype==="string")throw new Error("zerosLike is not supported for string tensors");if(r.dtype==="complex64"){let s=lo({inputs:{input:r},backend:n}),a=nf({inputs:{x:s},backend:n}),o=Ec({inputs:{input:r},backend:n}),i=nf({inputs:{x:o},backend:n}),c=Jn({inputs:{real:a,imag:i},backend:n});return n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(i),c}else return K0({backend:n,attrs:{shape:r.shape,value:0,dtype:r.dtype}})}var rX={kernelName:Pu,backendName:"cpu",kernelFunc:nf};function u2(e){let{inputs:t,backend:n}=e,{x:r}=t;if(r.dtype==="string")throw new Error("onesLike is not supported for string tensors");if(r.dtype==="complex64"){let s=lo({inputs:{input:r},backend:n}),a=u2({inputs:{x:s},backend:n}),o=Ec({inputs:{input:r},backend:n}),i=nf({inputs:{x:o},backend:n}),c=Jn({inputs:{real:a,imag:i},backend:n});return n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(i),c}else return K0({backend:n,attrs:{shape:r.shape,value:1,dtype:r.dtype}})}var sX={kernelName:bu,backendName:"cpu",kernelFunc:u2};function l2(e){let{inputs:t,backend:n,attrs:r}=e,{axis:s}=r;if(t.length===1)return tf({inputs:{input:t[0]},backend:n,attrs:{dim:s}});let a=t[0].shape,o=t[0].dtype;t.forEach(l=>{w.assertShapesMatch(a,l.shape,"All tensors passed to stack must have matching shapes"),w.assert(o===l.dtype,()=>"All tensors passed to stack must have matching dtypes")});let i=[],c=t.map(l=>{let p=tf({inputs:{input:l},backend:n,attrs:{dim:s}});return i.push(p),p}),u=Ac({inputs:c,backend:n,attrs:{axis:s}});return i.forEach(l=>n.disposeIntermediateTensorInfo(l)),u}var aX={kernelName:yu,backendName:"cpu",kernelFunc:l2};function oX(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{paddings:a,constantValue:o}=r;be(s,"pad");let i=a.map((y,v)=>y[0]+s.shape[v]+y[1]),c=a.map(y=>y[0]),u=n.data.get(s.dataId).values,l=w.sizeFromShape(s.shape),p=s.shape.length,d=w.computeStrides(s.shape),h=w.sizeFromShape(i),f=i.length,m=w.computeStrides(i),g=w.getTypedArrayFromDType(s.dtype,h);o!==0&&g.fill(o);for(let y=0;y<l;y++){let x=w.indexToLoc(y,p,d).map((S,C)=>S+c[C]),k=w.locToIndex(x,f,m);g[k]=u[y]}return{dataId:n.write(g,i,s.dtype),shape:i,dtype:s.dtype}}var d2={kernelName:Ko,backendName:"cpu",kernelFunc:oX},iX=Vt((e,t)=>Math.pow(e,t)),cX=nn(Xo,iX),uX={kernelName:Xo,backendName:"cpu",kernelFunc:cX};function lX(e){let{inputs:t,backend:n,attrs:r}=e,{paramsNestedSplits:s,paramsDenseValues:a,indices:o}=t,{outputRaggedRank:i}=r,c=s.map(b=>n.data.get(b.dataId).values),u=s.map(b=>b.shape),l=n.data.get(a.dataId).values,p=n.data.get(o.dataId).values,[d,h,f]=O_(c,u,l,a.shape,a.dtype,p,o.shape,i),m=d.map(b=>n.makeTensorInfo([b.length],"int32",b)),g=n.makeTensorInfo(f,a.dtype,h);return m.concat([g])}var dX={kernelName:Pf,backendName:"cpu",kernelFunc:lX};function pX(e){let{inputs:t,backend:n}=e,{starts:r,limits:s,deltas:a}=t,o=n.data.get(r.dataId).values,i=n.data.get(s.dataId).values,c=n.data.get(a.dataId).values,[u,l]=M_(o,r.shape,r.dtype,i,s.shape,c,a.shape),p=n.makeTensorInfo([u.length],"int32",u),d=n.makeTensorInfo([l.length],r.dtype,l);return[p,d]}var hX={kernelName:Of,backendName:"cpu",kernelFunc:pX};function fX(e){let{inputs:t,backend:n,attrs:r}=e,{shape:s,values:a,defaultValue:o,rowPartitionTensors:i}=t,{rowPartitionTypes:c}=r,u=n.data.get(s.dataId).values,l=n.data.get(a.dataId).values,p=n.data.get(o.dataId).values,d=i.map(g=>n.data.get(g.dataId).values),h=i.map(g=>g.shape),[f,m]=L_(u,s.shape,l,a.shape,a.dtype,p,o.shape,d,h,c);return n.makeTensorInfo(f,a.dtype,m)}var mX={kernelName:Mf,backendName:"cpu",kernelFunc:fX};function gX(e){let{backend:t,attrs:n}=e,{start:r,stop:s,dtype:a,step:o}=n,i=B0(r,s,o,a);return t.makeTensorInfo([i.length],a,i)}var bX={kernelName:vd,backendName:"cpu",kernelFunc:gX},yX=at(vu,e=>1/e),vX={kernelName:vu,backendName:"cpu",kernelFunc:yX};function xX(e){let{inputs:t,backend:n,attrs:r}=e,{images:s}=t,{alignCorners:a,halfPixelCenters:o,size:i}=r;be(s,"resizeBilinear");let c=w.computeStrides(s.shape),[u,l]=i,[p,d,h,f]=s.shape,m=n.data.get(s.dataId).values,g=new Float32Array(w.sizeFromShape([p,u,l,f])),b=[a&&u>1?d-1:d,a&&l>1?h-1:h],y=[a&&u>1?u-1:u,a&&l>1?l-1:l],v=0,x=b[0]/y[0],k=b[1]/y[1];for(let S=0;S<p;S++)for(let C=0;C<u;C++){let E;o?E=x*(C+.5)-.5:E=x*C;let $=Math.max(0,Math.floor(E)),F=E-$,A=Math.min(d-1,Math.ceil(E)),R=S*c[0]+$*c[1],T=S*c[0]+A*c[1];for(let L=0;L<l;L++){let V;o?V=k*(L+.5)-.5:V=k*L;let G=Math.max(0,Math.floor(V)),j=V-G,H=Math.min(h-1,Math.ceil(V)),Z=R+G*c[2],J=T+G*c[2],ee=R+H*c[2],re=T+H*c[2];for(let te=0;te<f;te++){let ie=m[Z+te],ne=m[J+te],le=m[ee+te],ue=m[re+te],ye=ie+(le-ie)*j,ve=ne+(ue-ne)*j,Se=ye+(ve-ye)*F;g[v++]=Se}}}return n.makeTensorInfo([p,u,l,f],"float32",g)}var wX={kernelName:ei,backendName:"cpu",kernelFunc:xX};function IX(e){let{inputs:t,backend:n,attrs:r}=e,{images:s,dy:a}=t,{alignCorners:o}=r;be([a,s],"resizeBilinearGrad");let i=w.computeStrides(s.shape),[c,u,l,p]=s.shape,[,d,h]=a.shape,f=new Float32Array(c*u*l*p),m=[o&&d>1?u-1:u,o&&h>1?l-1:l],g=[o&&d>1?d-1:d,o&&h>1?h-1:h],b=m[0]/g[0],y=m[1]/g[1],v=n.data.get(a.dataId).values,x=0;for(let k=0;k<c;k++){let S=k*i[0];for(let C=0;C<d;C++){let E=C*b,$=Math.floor(E),F=Math.min(Math.ceil(E),u-1),A=S+$*i[1],R=S+F*i[1],T=E-$,L=1-T;for(let V=0;V<h;V++){let G=V*y,j=Math.floor(G),H=Math.min(Math.ceil(G),l-1),Z=G-j,J=1-Z,ee=A+j*i[2],re=A+H*i[2],te=R+j*i[2],ie=R+H*i[2],ne=L*J,le=L*Z,ue=T*J,ye=T*Z;for(let ve=0;ve<p;ve++){let Se=v[x++];f[ee+ve]+=Se*ne,f[re+ve]+=Se*le,f[te+ve]+=Se*ue,f[ie+ve]+=Se*ye}}}}return n.makeTensorInfo([c,l,u,p],"float32",f)}var kX={kernelName:Bf,backendName:"cpu",kernelFunc:IX};function SX(e){let{inputs:t,backend:n,attrs:r}=e,{images:s}=t,{alignCorners:a,halfPixelCenters:o,size:i}=r;be(s,"resizeNearestNeighbor");let c=w.computeStrides(s.shape),[u,l]=i,[p,d,h,f]=s.shape,m=n.data.get(s.dataId).values,g=new Float32Array(p*u*l*f),b=[a&&u>1?d-1:d,a&&l>1?h-1:h],y=[a&&u>1?u-1:u,a&&l>1?l-1:l],v=b[0]/y[0],x=b[1]/y[1],k=0;for(let S=0;S<p;S++){let C=S*c[0];for(let E=0;E<u;E++){let $=o?v*(E+.5):v*E,F=Math.min(d-1,a?Math.round($):Math.floor($));o&&(F=Math.max(0,F));let A=C+F*c[1];for(let R=0;R<l;R++){let T=o?x*(R+.5):x*R,L=Math.min(h-1,a?Math.round(T):Math.floor(T));o&&(L=Math.max(0,L));let V=A+L*c[2];for(let G=0;G<f;G++){let j=m[V+G];g[k++]=j}}}}return n.makeTensorInfo([p,u,l,f],s.dtype,g)}var TX={kernelName:Qo,backendName:"cpu",kernelFunc:SX};function CX(e){let{inputs:t,backend:n,attrs:r}=e,{images:s,dy:a}=t,{alignCorners:o}=r;be([a,s],"resizeNearestNeighborGrad");let i=w.computeStrides(s.shape),c=w.computeStrides(a.shape),[u,l,p,d]=s.shape,[,h,f]=a.shape,m=new Float32Array(u*l*p*d),g=n.data.get(a.dataId).values,b=[o&&h>1?l-1:l,o&&f>1?p-1:p],y=[o&&h>1?h-1:h,o&&f>1?f-1:f],v=b[0]/y[0],x=b[1]/y[1],k=1/v,S=1/x,C=Math.ceil(k)*2+2,E=Math.ceil(S)*2+2;for(let $=0;$<u;$++){let F=$*i[0];for(let A=0;A<l;A++){let R=F+A*i[1],T=Math.floor(A*k),L=Math.floor(T-C/2);for(let V=0;V<p;V++){let G=R+V*i[2],j=Math.floor(V*S),H=Math.floor(j-E/2);for(let Z=0;Z<d;Z++){let J=0;for(let ee=0;ee<C;ee++){let re=ee+L;if(re<0||re>=h)continue;let te=F+re*c[1],ie=re*v,ne=Math.min(l-1,o?Math.round(ie):Math.floor(ie));if(A===ne)for(let le=0;le<E;le++){let ue=le+H;if(ue<0||ue>=f)continue;let ye=te+ue*c[2],ve=ue*x,Se=Math.min(p-1,o?Math.round(ve):Math.floor(ve));V===Se&&(J+=g[ye+Z])}}m[G+Z]=J}}}}return n.makeTensorInfo(s.shape,s.dtype,m)}var NX={kernelName:zf,backendName:"cpu",kernelFunc:CX};function _X(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{dims:a}=r;be(s,"reverse");let o=s.shape.length,i=w.parseAxisParam(a,s.shape);if(o===0)return ds({inputs:{x:s},backend:n});let c=new Ht(s.shape,s.dtype),u=n.bufferSync(s);for(let l=0;l<c.size;l++){let p=c.indexToLoc(l),d=p.slice();i.forEach(h=>d[h]=s.shape[h]-1-d[h]),c.set(u.get(...d),...p)}return n.makeTensorInfo(c.shape,c.dtype,c.values)}var EX={kernelName:ni,backendName:"cpu",kernelFunc:_X},AX={kernelName:Ou,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,{radians:s,fillValue:a,center:o}=t,i=n,c=w.getTypedArrayFromDType(r.dtype,w.sizeFromShape(r.shape)),[u,l,p,d]=r.shape,[h,f]=N.getImageCenter(o,l,p),m=255,g=Math.sin(s),b=Math.cos(s),y=i.data.get(r.dataId).values;for(let x=0;x<u;x++){let k=x*p*l*d;for(let S=0;S<l;S++){let C=S*(p*d);for(let E=0;E<p;E++){let $=E*d;for(let F=0;F<d;F++){let A=[u,S,E,F],R=A[2],T=A[1],L=(R-h)*b-(T-f)*g,V=(R-h)*g+(T-f)*b;L=Math.round(L+h),V=Math.round(V+f);let G=a;if(typeof a!="number"&&(F===3?G=m:G=a[F]),L>=0&&L<p&&V>=0&&V<l){let H=V*(p*d),Z=L*d,J=k+H+Z+F;G=y[J]}let j=k+C+$+F;c[j]=G}}}}return{dataId:i.write(c,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},$X=at(ri,e=>{let t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2===0?t:t+1}),DX={kernelName:ri,backendName:"cpu",kernelFunc:$X};function FX(e){let{inputs:t,backend:n,attrs:r}=e,{indices:s,updates:a}=t,{shape:o}=r,{sliceRank:i,numUpdates:c,sliceSize:u,strides:l,outputSize:p}=N.calculateShapes(a,s,o),d=!0,h=n.bufferSync(s),f=n.bufferSync(a),m=cc(h,f,o,p,u,c,i,l,0,d);return n.makeTensorInfo(o,m.dtype,m.values)}var RX={kernelName:wu,backendName:"cpu",kernelFunc:FX};function PX(e,t){let n=0,r=e.length,s=0;for(;n<r;)s=Math.floor((n+r)/2),e[s]<t?n=s+1:r=s;return r}function OX(e,t){let n=0,r=e.length,s=0;for(;n<r;)s=Math.floor((n+r)/2),e[s]<=t?n=s+1:r=s;return r}function MX(e,t,n,r,s,a){let o=w.getArrayFromDType("int32",n*s);for(let i=0;i<n;++i){let c=e.slice(i*r,(i+1)*r),u=i*s;for(let l=0;l<s;++l)o[u+l]=a==="left"?PX(c,t[l+u]):OX(c,t[l+u])}return o}function LX(e){let{inputs:t,backend:n,attrs:r}=e,{sortedSequence:s,values:a}=t,{side:o}=r,i=n.data.get(s.dataId).values,c=n.data.get(a.dataId).values,u=MX(i,c,s.shape[0],s.shape[1],a.shape[1],o);return n.makeTensorInfo(a.shape,"int32",u)}var zX={kernelName:Wf,backendName:"cpu",kernelFunc:LX};function BX(e){let{inputs:t,backend:n}=e,{condition:r,t:s,e:a}=t;be([r,s,a],"select");let o=r.shape.length,i=n.data.get(r.dataId).values,c=n.data.get(s.dataId).values,u=n.data.get(a.dataId).values,l=hr(s.dtype,a.dtype),p=w.makeZerosTypedArray(w.sizeFromShape(s.shape),l),d=0,h=o===0||o>1||s.shape.length===1?1:w.sizeFromShape(s.shape.slice(1));for(let f=0;f<i.length;f++)for(let m=0;m<h;m++)i[f]===1?p[d++]=c[f]:p[d++]=u[f];return n.makeTensorInfo(s.shape,l,p)}var WX={kernelName:Iu,backendName:"cpu",kernelFunc:BX},VX=N.SELU_SCALEALPHA,UX=N.SELU_SCALE,GX=at(ku,e=>e>=0?UX*e:VX*(Math.exp(e)-1)),HX={kernelName:ku,backendName:"cpu",kernelFunc:GX},qX=at(Cu,e=>e<0?-1:e>0?1:0),jX={kernelName:Cu,backendName:"cpu",kernelFunc:qX},KX=at(ai,e=>Math.sin(e)),XX={kernelName:ai,backendName:"cpu",kernelFunc:KX},YX=at(Tu,e=>Math.sinh(e)),ZX={kernelName:Tu,backendName:"cpu",kernelFunc:YX},JX=11920928955078125e-23,d1=Math.log(JX)+2,QX=at(Nu,e=>{let t=e>-d1,n=e<d1,r=Math.exp(e),s;return n?s=r:t?s=e:s=Math.log(1+r),s}),e7={kernelName:Nu,backendName:"cpu",kernelFunc:QX};function t7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,paddings:o}=r;be([s],"spaceToBatchND");let i=w.sizeFromShape(a),c=[[0,0]];c.push(...o);for(let S=1+a.length;S<s.shape.length;++S)c.push([0,0]);let u=d2.kernelFunc({inputs:{x:s},backend:n,attrs:{paddings:c,constantValue:0}}),l=N.getReshaped(u.shape,a,i,!1),p=N.getPermuted(l.length,a.length,!1),d=N.getReshapedPermuted(u.shape,a,i,!1),m=ft({inputs:{x:u},backend:n,attrs:{shape:l}}),y=Vn({inputs:{x:m},backend:n,attrs:{perm:p}}),k=ft({inputs:{x:y},backend:n,attrs:{shape:d}});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(y),k}var n7={kernelName:_u,backendName:"cpu",kernelFunc:t7};function r7(e){let{inputs:t,backend:n}=e,{indices:r,values:s,denseShape:a,defaultValue:o}=t;if(a.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
+        ${a.shape}`);if(r.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
+        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Values must be a vector, saw:
+        ${s.shape}`);if(o.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
+        ${o.shape}`);let i=n.data.get(r.dataId).values,c=n.data.get(s.dataId).values,u=n.data.get(a.dataId).values,l=n.data.get(o.dataId).values[0],[p,d,h,f,m]=W_(i,r.shape,r.dtype,c,s.dtype,u,l);return[n.makeTensorInfo(d,r.dtype,p),n.makeTensorInfo([d[0]],s.dtype,h),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(g=>Number(g)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}var s7={kernelName:xd,backendName:"cpu",kernelFunc:r7};function a7(e){let{inputs:t,backend:n}=e,{inputIndices:r,inputShape:s,newShape:a}=t;if(r.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
+        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
+        ${s.shape}`);if(a.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${a.shape}`);let o=Array.from(n.data.get(s.dataId).values),i=n.data.get(r.dataId).values,c=Array.from(n.data.get(a.dataId).values),[u,l,p]=V_(i,r.shape,r.dtype,o,c);return[n.makeTensorInfo(l,r.dtype,u),n.makeTensorInfo([p.length],a.dtype,new Int32Array(p))]}var o7={kernelName:Au,backendName:"cpu",kernelFunc:a7};function i7(e){let{inputs:t,backend:n}=e,{data:r,indices:s,segmentIds:a}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.shape.length!==1)throw new Error(`Indices should be a vector but received shape
+          ${s.shape}`);if(a.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
+          ${a.shape}`);if(s.shape[0]!==a.shape[0])throw new Error("segmentIds and indices should have same size.");let o=n.data.get(r.dataId).values,i=n.data.get(s.dataId).values,c=n.data.get(a.dataId).values,[u,l]=W0(o,r.shape,r.dtype,i,c,!0);return n.makeTensorInfo(l,r.dtype,u)}var c7={kernelName:wd,backendName:"cpu",kernelFunc:i7};function u7(e){let{inputs:t,backend:n}=e,{data:r,indices:s,segmentIds:a}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.shape.length!==1)throw new Error(`Indices should be a vector but received shape
+         ${s.shape}`);if(a.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
+         ${a.shape}`);if(s.shape[0]!==a.shape[0])throw new Error("segmentIds and indices should have same size.");let o=n.data.get(r.dataId).values,i=n.data.get(s.dataId).values,c=n.data.get(a.dataId).values,[u,l]=W0(o,r.shape,r.dtype,i,c);return n.makeTensorInfo(l,r.dtype,u)}var l7={kernelName:Id,backendName:"cpu",kernelFunc:u7};function d7(e){let{inputs:t,backend:n,attrs:r}=e,{sparseIndices:s,sparseValues:a,defaultValue:o}=t,{outputShape:i}=r,{sliceRank:c,numUpdates:u,sliceSize:l,strides:p,outputSize:d}=N.calculateShapes(a,s,i),h=!1,f=n.bufferSync(s),m;switch(a.dtype){case"bool":{let g=n.bufferSync(a),b=Boolean(n.data.get(o.dataId).values[0]);m=cc(f,g,i,d,l,u,c,p,b,h);break}case"float32":{let g=n.bufferSync(a),b=n.data.get(o.dataId).values[0];m=cc(f,g,i,d,l,u,c,p,b,h);break}case"int32":{let g=n.bufferSync(a),b=n.data.get(o.dataId).values[0];m=cc(f,g,i,d,l,u,c,p,b,h);break}case"string":{let g=n.bufferSync(a),b=w.decodeString(n.data.get(o.dataId).values[0]);m=cc(f,g,i,d,l,u,c,p,b,h);break}default:throw new Error(`Unsupported type ${a.dtype}`)}return n.makeTensorInfo(i,m.dtype,m.values)}var p7={kernelName:Vf,backendName:"cpu",kernelFunc:d7};function h7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{numOrSizeSplits:a,axis:o}=r,i=w.parseAxisParam(o,s.shape)[0],c=N.prepareSplitSize(s,a,i),u=new Array(s.shape.length).fill(0),l=s.shape.slice();return c.map(p=>{let d=[...l];d[i]=p;let h=po({inputs:{x:s},backend:n,attrs:{begin:u,size:d}});return u[i]+=p,h})}var f7={kernelName:Eu,backendName:"cpu",kernelFunc:h7},m7={kernelName:kd,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{let{x:n}=e,r=t;be(n,"square");let s=r.data.get(n.dataId).values,a=new Float32Array(s.length);for(let i=0;i<s.length;++i){let c=s[i];a[i]=c*c}return{dataId:r.write(a,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},g7=at(xa,(e,t)=>{let n=t;return isNaN(e)?NaN:e>0?1:n.alpha}),b7={kernelName:xa,backendName:"cpu",kernelFunc:g7};function y7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{begin:a,end:o,strides:i,beginMask:c,endMask:u,ellipsisMask:l,newAxisMask:p,shrinkAxisMask:d}=r;be(s,"stridedSlice");let{finalShapeSparse:h,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:b,begin:y,end:v,strides:x}=qt.sliceInfo(s.shape,a,o,i,c,u,l,p,d),k;if(m)k=ft({inputs:{x:s},backend:n,attrs:{shape:f}});else if(g||b){w.assert(s.shape.length>=1,()=>`Input must have rank at least 1, got: ${s.shape.length}`);let S=qt.computeOutShape(y,v,x),C=po({inputs:{x:s},backend:n,attrs:{begin:y,size:S}});k=ft({inputs:{x:C},backend:n,attrs:{shape:f}}),n.disposeIntermediateTensorInfo(C)}else{let S=n.bufferSync(s),C=G_(h,S,x,y);k=n.makeTensorInfo(f,C.dtype,C.values)}return k}var v7={kernelName:$u,backendName:"cpu",kernelFunc:y7};function x7(e){let{inputs:t,backend:n,attrs:r}=e,{separator:s,nGramWidths:a,leftPad:o,rightPad:i,padWidth:c,preserveShortSequences:u}=r,{data:l,dataSplits:p}=t,d=n.data.get(l.dataId).values,h=n.data.get(p.dataId).values,[f,m]=V0(d,h,s,a,o,i,c,u);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(p.shape,"int32",m)]}var w7={kernelName:Sd,backendName:"cpu",kernelFunc:x7};function I7(e){let{inputs:t,backend:n,attrs:r}=e,{skipEmpty:s}=r,{input:a,delimiter:o}=t;if(a.dtype!=="string")throw new Error("Input must be of datatype string");if(a.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${a.shape}`);if(o.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${o.shape}`);let i=n.data.get(a.dataId).values,c=n.data.get(o.dataId).values[0],[u,l,p]=U0(i,c,s),d=l.length;return[n.makeTensorInfo([d,2],"int32",u),n.makeTensorInfo([d],"string",l),n.makeTensorInfo([2],"int32",new Int32Array(p))]}var k7={kernelName:Td,backendName:"cpu",kernelFunc:I7};function S7(e){let{inputs:t,backend:n,attrs:r}=e,{numBuckets:s}=r,{input:a}=t;if(a.dtype!=="string")throw new Error("Input must be of datatype string");if(s<=0)throw new Error("Number of buckets must be at least 1");let o=n.data.get(a.dataId).values,i=G0(o,s);return n.makeTensorInfo(a.shape,"int32",i)}var T7={kernelName:Cd,backendName:"cpu",kernelFunc:S7},C7=at(pi,e=>Math.tan(e)),N7={kernelName:pi,backendName:"cpu",kernelFunc:C7},_7=at(hi,e=>Math.tanh(e)),E7={kernelName:hi,backendName:"cpu",kernelFunc:_7};function A7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{reps:a}=r;be(s,"tile");let o=q_(n.bufferSync(s),a);return n.makeTensorInfo(o.shape,o.dtype,o.values)}var $7={kernelName:va,backendName:"cpu",kernelFunc:A7};function D7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{k:a,sorted:o}=r;be(s,"topk");let i=n.data.get(s.dataId).values,[c,u]=K_(i,s.shape,s.dtype,a,o);return[n.makeTensorInfo(c.shape,c.dtype,c.values),n.makeTensorInfo(u.shape,u.dtype,u.values)]}var F7={kernelName:Du,backendName:"cpu",kernelFunc:D7};function R7(e){let{inputs:t,attrs:n,backend:r}=e,{image:s,transforms:a}=t,{interpolation:o,fillMode:i,fillValue:c,outputShape:u}=n,[l,p,d,h]=s.shape,[f,m]=u!=null?u:[p,d],g=[l,f,m,h],b=w.computeStrides(s.shape),y=b[0],v=b[1],x=b[2],k=w.computeStrides(g),S=k[0],C=k[1],E=k[2],$=w.getTypedArrayFromDType(s.dtype,w.sizeFromShape(g));$.fill(c);let F=r.data.get(s.dataId).values,A=r.data.get(a.dataId).values;for(let T=0;T<l;++T){let L=a.shape[0]===1?A:A.subarray(T*8,T*8+8);for(let V=0;V<f;++V)for(let G=0;G<m;++G)for(let j=0;j<h;++j){let H,Z=L[6]*G+L[7]*V+1;if(Z===0)continue;let J=(L[0]*G+L[1]*V+L[2])/Z,ee=(L[3]*G+L[4]*V+L[5])/Z,re=p1(J,d,i),te=p1(ee,p,i);switch(o){case"nearest":H=B7(F,p,d,y,v,x,T,te,re,j,c);break;case"bilinear":H=W7(F,p,d,y,v,x,T,te,re,j,c);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${o}`)}let ie=T*S+V*C+G*E+j;$[ie]=H}return r.makeTensorInfo(g,s.dtype,$)}return{dataId:r.write($,g,s.dtype),shape:s.shape,dtype:s.dtype}}var P7={kernelName:Fu,backendName:"cpu",kernelFunc:R7};function p1(e,t,n){switch(n){case"reflect":return O7(e,t);case"wrap":return M7(e,t);case"nearest":return z7(e,t);case"constant":default:return L7(e,t)}}function O7(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let r=2*t;n<r&&(n=r*Math.trunc(-n/r)+n),n=n<-t?n+r:-n-1}else if(n>t-1)if(t<=1)n=0;else{let r=2*t;n-=r*Math.trunc(n/r),n>=t&&(n=r-n-1)}return w.clamp(0,n,t-1)}function M7(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let r=t-1;n+=t*(Math.trunc(-n/r)+1)}else if(n>t-1)if(t<=1)n=0;else{let r=t-1;n-=t*Math.trunc(n/r)}return w.clamp(0,n,t-1)}function L7(e,t){return e}function z7(e,t){return w.clamp(0,e,t-1)}function Ll(e,t,n,r,s,a,o,i,c,u,l){let p=o*r+i*s+c*a+u;return 0<=i&&i<t&&0<=c&&c<n?e[p]:l}function B7(e,t,n,r,s,a,o,i,c,u,l){let p=Math.round(i),d=Math.round(c);return Ll(e,t,n,r,s,a,o,p,d,u,l)}function W7(e,t,n,r,s,a,o,i,c,u,l){let p=Math.floor(i),d=Math.floor(c),h=p+1,f=d+1,m=(f-c)*Ll(e,t,n,r,s,a,o,p,d,u,l)+(c-d)*Ll(e,t,n,r,s,a,o,p,f,u,l),g=(f-c)*Ll(e,t,n,r,s,a,o,h,d,u,l)+(c-d)*Ll(e,t,n,r,s,a,o,h,f,u,l);return(h-i)*m+(i-p)*g}function V7(e){let{inputs:t,attrs:n,backend:r}=e,{axis:s}=n,{x:a}=t;be(a,"unique");let o=r.data.get(a.dataId).values,{outputValues:i,outputShape:c,indices:u}=X_(o,s,a.shape,a.dtype);return[r.makeTensorInfo(c,a.dtype,i),r.makeTensorInfo([u.length],"int32",u)]}var U7={kernelName:Uf,backendName:"cpu",kernelFunc:V7};function G7(e){let{inputs:t,backend:n,attrs:r}=e,{value:s}=t,{axis:a}=r;a<0&&(a+=s.shape.length);let o=s.shape.length,i=s.shape[a],c=new Array(o-1),u=0;for(let h=0;h<o;h++)h!==a&&(c[u++]=s.shape[h]);let l=new Array(o).fill(0),p=s.shape.slice();p[a]=1;let d=new Array(i);for(let h=0;h<d.length;h++){l[a]=h;let f=po({inputs:{x:s},backend:n,attrs:{begin:l,size:p}});d[h]=ft({inputs:{x:f},backend:n,attrs:{shape:c}}),n.disposeIntermediateTensorInfo(f)}return d}var H7={kernelName:Ru,backendName:"cpu",kernelFunc:G7};function q7(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,segmentIds:a}=t,{numSegments:o}=r;be(s,"unsortedSegmentSum");let i=s.shape.length,c=a.shape.length,u=[],l=[],p=i-c,d=a;for(let f=0;f<p;++f){let m=tf({inputs:{input:d},backend:n,attrs:{dim:f+1}});d=m,l.push(m)}for(let f=0;f<o;++f){let m=w.createScalarValue(f,"int32"),g=n.makeTensorInfo([],"int32",m),b=b_({inputs:{a:g,b:d},backend:n}),y=ha({inputs:{x:b},backend:n,attrs:{dtype:"float32"}}),v=Km({inputs:{a:y,b:s},backend:n}),x=ap({inputs:{x:v},backend:n,attrs:{axis:0,keepDims:!1}});u.push(x),l.push(g),l.push(b),l.push(y),l.push(v),l.push(x)}let h=l2({inputs:u,backend:n,attrs:{axis:0}});return l.forEach(f=>n.disposeIntermediateTensorInfo(f)),h}var j7={kernelName:Nd,backendName:"cpu",kernelFunc:q7},K7=[V5,Eq,G5,q5,Pq,K5,Y5,J5,ej,nj,sj,oj,cj,dj,hj,gj,yj,xj,Ij,B5,Sj,Cj,_j,Aj,Fq,Mq,Dj,Aq,Rj,Oj,Mj,zj,Wj,Uj,Hj,jj,Xj,Zj,Qj,t8,r8,a8,i8,c8,l8,p8,f8,m8,g8,b8,x8,F5,I8,Lq,A8,zq,$8,Wq,M8,L8,B8,Uq,U8,H8,j8,X8,Z8,Hq,jq,$q,Q8,Pj,tK,rK,aK,R5,Xq,Zq,iK,Qq,uK,pK,fK,bK,vK,wK,IK,t5,SK,CK,_K,AK,DK,RK,OK,r5,LK,WK,HK,a5,i5,KK,ZK,eX,u5,nX,sX,aX,d2,uX,O5,p5,dX,hX,mX,bX,Dq,pv,vX,M5,L5,z5,wX,kX,TX,NX,EX,AX,DX,x5,RX,zX,WX,HX,I5,jX,XX,ZX,k5,UK,e7,n7,s7,o7,c7,l7,p7,f7,C5,m7,_5,b7,v7,w7,k7,T7,D5,y8,N7,E7,$7,F7,P7,l5,U7,H7,j7,rX];for(let e of K7)_d(e);var p2={};Ae(p2,{assertNotComplex:()=>Yu,bindCanvasToFramebuffer:()=>aY,bindColorTextureToFramebuffer:()=>kh,bindTextureToProgramUniformSampler:()=>_2,bindTextureUnit:()=>T2,bindVertexBufferToProgramAttribute:()=>fv,callAndCheck:()=>me,canBeRepresented:()=>f2,createFragmentShader:()=>b2,createFramebuffer:()=>S2,createProgram:()=>y2,createStaticIndexBuffer:()=>w2,createStaticVertexBuffer:()=>x2,createTexture:()=>I2,createVertexShader:()=>g2,getBatchDim:()=>ho,getExtensionOrThrow:()=>zl,getFramebufferErrorMessage:()=>E2,getMaxTexturesInShader:()=>F2,getNumChannels:()=>rY,getProgramUniformLocation:()=>N2,getProgramUniformLocationOrThrow:()=>C2,getRowsCols:()=>fo,getShapeAs3D:()=>Wl,getTextureShapeFromLogicalShape:()=>$2,getWebGLDisjointQueryTimerVersion:()=>R2,getWebGLErrorMessage:()=>m2,getWebGLMaxTextureSize:()=>D2,hasExtension:()=>lr,isCapableOfRenderingToFloatTexture:()=>P2,isDownloadFloatTextureEnabled:()=>O2,isReshapeFree:()=>ad,isWebGLFenceEnabled:()=>M2,isWebGLVersionEnabled:()=>gv,linkProgram:()=>v2,logShaderSourceAndInfoLog:()=>Y0,resetMaxTextureSize:()=>oY,resetMaxTexturesInShader:()=>iY,unbindColorTextureFromFramebuffer:()=>mv,unbindTextureUnit:()=>sY,validateFramebuffer:()=>Bl,validateProgram:()=>Ih,validateTextureSize:()=>k2});var Ha={},fh={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function h2(e,t){Ha[e]=t}function Kr(e,t){if(!(e in Ha)||t!=null){let r=Y7(e,t);if(r!==null)Ha[e]=r;else return console.log("Could not get context for WebGL version",e),null}let n=Ha[e];return n==null||n.isContextLost()?(delete Ha[e],Kr(e)):(n.disable(n.DEPTH_TEST),n.disable(n.STENCIL_TEST),n.disable(n.BLEND),n.disable(n.DITHER),n.disable(n.POLYGON_OFFSET_FILL),n.disable(n.SAMPLE_COVERAGE),n.enable(n.SCISSOR_TEST),n.enable(n.CULL_FACE),n.cullFace(n.BACK),Ha[e])}function X7(e){if(typeof OffscreenCanvas!="undefined"&&e===2)return new OffscreenCanvas(300,150);if(typeof document!="undefined")return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}function Y7(e,t){if(e!==1&&e!==2)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let n=t==null?X7(e):t;return n.addEventListener("webglcontextlost",r=>{r.preventDefault(),delete Ha[e]},!1),q().getBool("SOFTWARE_WEBGL_ENABLED")&&(fh.failIfMajorPerformanceCaveat=!1),e===1?n.getContext("webgl",fh)||n.getContext("experimental-webgl",fh):n.getContext("webgl2",fh)}var sd;(function(e){e[e.DENSE=0]="DENSE",e[e.SHARED_BATCH=1]="SHARED_BATCH"})(sd||(sd={}));var ur;(function(e){e[e.RENDER=0]="RENDER",e[e.UPLOAD=1]="UPLOAD",e[e.PIXELS=2]="PIXELS",e[e.DOWNLOAD=3]="DOWNLOAD"})(ur||(ur={}));var cn;(function(e){e[e.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",e[e.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",e[e.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",e[e.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",e[e.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16"})(cn||(cn={}));function op(e,t){return[t,e]}function Z7(e,t){return e*t}function mh(e){let t=w.sizeFromShape(e),n=Math.ceil(t/4);return w.sizeToSquarishShape(n)}function Xu(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function J7(e,t){let[n,r]=Xu(e,t);return n*r*4}function X0(e,t){let n=e,r,s,a,o,i,c,u,l,p,d;return q().getNumber("WEBGL_VERSION")===2?(r=n.R32F,s=n.R16F,a=n.RGBA16F,o=n.RGBA32F,i=n.RED,u=4,l=1,p=n.HALF_FLOAT,d=n.FLOAT,c=n.RGBA8):(r=e.RGBA,s=e.RGBA,a=e.RGBA,o=n.RGBA,i=e.RGBA,u=4,l=4,p=t!=null?t.HALF_FLOAT_OES:null,d=e.FLOAT,c=e.RGBA),{internalFormatFloat:r,internalFormatHalfFloat:s,internalFormatPackedHalfFloat:a,internalFormatPackedFloat:o,textureFormatFloat:i,downloadTextureFormat:c,downloadUnpackNumChannels:u,defaultNumChannels:l,textureTypeHalfFloat:p,textureTypeFloat:d}}function me(e,t){let n=t();return q().getBool("DEBUG")&&Q7(e),n}function Q7(e){let t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+m2(e,t))}var eY=596e-10,tY=65504;function f2(e){return!!(q().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||e===0||eY<Math.abs(e)&&Math.abs(e)<tY)}function m2(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function zl(e,t){return Ps(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function g2(e,t){let n=Ps(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw console.log(e.getShaderInfoLog(n)),new Error("Failed to compile vertex shader.");return n}function b2(e,t){let n=Ps(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(me(e,()=>e.shaderSource(n,t)),me(e,()=>e.compileShader(n)),q().get("ENGINE_COMPILE_ONLY"))return n;if(e.getShaderParameter(n,e.COMPILE_STATUS)===!1)throw Y0(t,e.getShaderInfoLog(n)),new Error("Failed to compile fragment shader.");return n}var nY=/ERROR: [0-9]+:([0-9]+):/g;function Y0(e,t){let n=nY.exec(t);if(n==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(e);return}let r=+n[1],s=e.split(`
+`),a=s.length.toString().length+2,o=s.map((p,d)=>w.rightPad((d+1).toString(),a)+p),i=0;for(let p=0;p<o.length;p++)i=Math.max(o[p].length,i);let c=o.slice(0,r-1),u=o.slice(r-1,r),l=o.slice(r);console.log(c.join(`
 `)),console.log(t.split(`
-`)[0]),console.log(`%c ${v.rightPad(u[0],o)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(p.join(`
-`))}function g_(e){return Mr(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function y_(e,t){if(me(e,()=>e.linkProgram(t)),!H().get("ENGINE_COMPILE_ONLY")&&e.getProgramParameter(t,e.LINK_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function wh(e,t){if(me(e,()=>e.validateProgram(t)),e.getProgramParameter(t,e.VALIDATE_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function b_(e,t){let n=Mr(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function x_(e,t){let n=Mr(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function eZ(){return H().getNumber("WEBGL_VERSION")===2?1:4}function v_(e){return Mr(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function w_(e,t){let n=H().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){let a=`[${e}x${t}]`;throw new Error("Requested texture size "+a+" is invalid.")}if(e>n||t>n){let a=`[${e}x${t}]`,r=`[${n}x${n}]`;throw new Error("Requested texture size "+a+" greater than WebGL maximum on this browser / GPU "+r+".")}}function k_(e){return Mr(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function mx(e,t,n,a,r,s,i){let o=e.getAttribLocation(t,n);return o===-1?!1:(me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,a)),me(e,()=>e.vertexAttribPointer(o,r,e.FLOAT,!1,s,i)),me(e,()=>e.enableVertexAttribArray(o)),!0)}function I_(e,t,n){__(e,n),me(e,()=>e.activeTexture(e.TEXTURE0+n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function tZ(e,t){__(e,t),me(e,()=>e.activeTexture(e.TEXTURE0+t)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function S_(e,t,n){return Mr(e,()=>e.getUniformLocation(t,n),'uniform "'+n+'" not present in program.')}function T_(e,t,n){return e.getUniformLocation(t,n)}function N_(e,t,n,a){me(e,()=>I_(e,t,a)),me(e,()=>e.uniform1i(n,a))}function nZ(e){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),me(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function kh(e,t,n){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,n)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function fx(e,t){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function Wp(e){let t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+C_(e,t))}function C_(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function Mr(e,t,n){let a=me(e,()=>t());if(a==null)throw new Error(n);return a}function __(e,t){let n=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,a=t+e.TEXTURE0;if(a<e.TEXTURE0||a>n){let r=`[gl.TEXTURE0, gl.TEXTURE${n}]`;throw new Error(`textureUnit must be in ${r}.`)}}function ci(e,t=2){return v.sizeFromShape(e.slice(0,e.length-t))}function di(e){if(e.length===0)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function Bp(e){let t=[1,1,1];return e.length===0||e.length===1&&e[0]===1||(t=[ci(e),...di(e)]),t}function E_(e,t=!1){let n=H().getNumber("WEBGL_MAX_TEXTURE_SIZE"),a=H().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");a===1/0&&H().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")&&(a=n/2),t&&(n=n*2,a=a*2,e=e.map((o,l)=>l>=e.length-2?v.nearestLargerEven(e[l]):e[l]),e.length===1&&(e=[2,e[0]])),e.length!==2&&(e=v.squeezeShape(e).newShape);let r=v.sizeFromShape(e),s=null;e.length<=1&&r<=n?s=[1,r]:e.length===2&&e[0]<=n&&e[1]<=n?s=e:e.length===3&&e[0]*e[1]<=n&&e[2]<=n?s=[e[0]*e[1],e[2]]:e.length===3&&e[0]<=n&&e[1]*e[2]<=n?s=[e[0],e[1]*e[2]]:e.length===4&&e[0]*e[1]*e[2]<=n&&e[3]<=n?s=[e[0]*e[1]*e[2],e[3]]:e.length===4&&e[0]<=n&&e[1]*e[2]*e[3]<=n&&(s=[e[0],e[1]*e[2]*e[3]]);let i=s!=null&&Math.max(...s)>a&&Math.min(...s)<=(t?2:1)&&Math.min(...s)>0;if(s==null||i)if(t){let o=ci(e),l=2,u=2;e.length&&([l,u]=di(e)),r=o*(l/2)*(u/2),s=v.sizeToSquarishShape(r).map(p=>p*2)}else s=v.sizeToSquarishShape(r);return s}function gh(e){return e%2===0}function sc(e,t){if(e=e.slice(-2),t=t.slice(-2),v.arraysEqual(e,t)||!e.length||!t.length||e[0]===0||e[1]===0||t[0]===0||t[1]===0)return!0;if(e.length!==t.length){let n=e.slice(-1)[0],a=t.slice(-1)[0];if(n===a||gh(n)&&gh(a)&&(e[0]===1||t[0]===1))return!0}return e[1]===t[1]&&gh(e[0])&&gh(t[0])}var Ih,Sh;function A_(e){if(Ih==null){let t=qa(e);Ih=t.getParameter(t.MAX_TEXTURE_SIZE)}return Ih}function aZ(){Ih=null}function rZ(){Sh=null}function $_(e){if(Sh==null){let t=qa(e);Sh=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,Sh)}function F_(e){if(e===0)return 0;let t,n=qa(e);return da(n,"EXT_disjoint_timer_query_webgl2")&&e===2?t=2:da(n,"EXT_disjoint_timer_query")?t=1:t=0,t}function da(e,t){return e.getExtension(t)!=null}function gx(e){try{if(qa(e)!=null)return!0}catch(t){return console.log("Error when getting WebGL context: ",t),!1}return!1}function D_(e){if(e===0)return!1;let t=qa(e);if(e===1){if(!da(t,"OES_texture_float"))return!1}else if(!da(t,"EXT_color_buffer_float"))return!1;return yx(t)}function R_(e){if(e===0)return!1;let t=qa(e);if(e===1){if(!da(t,"OES_texture_float")||!da(t,"WEBGL_color_buffer_float"))return!1}else{if(da(t,"EXT_color_buffer_float"))return yx(t);let n="EXT_color_buffer_half_float";if(da(t,n)){let a=t.getExtension(n);return sZ(t,a)}return!1}return yx(t)}function yx(e){let t=X0(e),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n);let a=1,r=1;e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,a,r,0,t.textureFormatFloat,t.textureTypeFloat,null);let s=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,s),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);let i=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(s),i}function sZ(e,t){let n=X0(e,t),a=e.createTexture();e.bindTexture(e.TEXTURE_2D,a);let r=1,s=1;e.texImage2D(e.TEXTURE_2D,0,n.internalFormatHalfFloat,r,s,0,n.textureFormatFloat,n.textureTypeHalfFloat,null);let i=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,i),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,a,0);let o=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(a),e.deleteFramebuffer(i),o}function M_(e){return e!==2?!1:qa(e).fenceSync!=null}function Yu(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&v.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var ve=H();ve.registerFlag("HAS_WEBGL",()=>ve.getNumber("WEBGL_VERSION")>0);ve.registerFlag("WEBGL_VERSION",()=>gx(2)?2:gx(1)?1:0);ve.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1);ve.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>ve.get("WEBGL_VERSION")===2);ve.registerFlag("WEBGL_CPU_FORWARD",()=>!0);ve.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1);ve.registerFlag("WEBGL_PACK",()=>ve.getBool("HAS_WEBGL"));ve.registerFlag("WEBGL_PACK_NORMALIZATION",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_CLIP",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_PACK_REDUCE",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_LAZILY_UNPACK",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_CONV_IM2COL",()=>ve.getBool("WEBGL_PACK"));ve.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>A_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>$_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let e=ve.getNumber("WEBGL_VERSION");return e===0?0:F_(e)});ve.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>ve.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!$c.isMobile());ve.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>D_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>ve.getBool("WEBGL_FORCE_F16_TEXTURES")?!1:ve.getBool("WEBGL_RENDER_FLOAT32_CAPABLE"));ve.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>R_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_FENCE_API_ENABLED",()=>M_(ve.getNumber("WEBGL_VERSION")));ve.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>ve.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0);ve.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)});ve.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>$c.isMobile()?1:-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)});ve.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128);ve.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1);ve.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5);ve.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);ve.registerFlag("WEBGL_EXP_CONV",()=>!1);ve.registerFlag("SOFTWARE_WEBGL_ENABLED",()=>ve.getBool("IS_TEST"));ve.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE",()=>1/0);ve.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE",()=>!1);ve.registerFlag("WEBGL2_ISNAN_CUSTOM",()=>!1);function _n(){let e,t,n,a,r,s,i,o,l,u;return H().getNumber("WEBGL_VERSION")===2?(e="#version 300 es",t="in",n="out",a="in",r="texture",s="outputColor",i="out vec4 outputColor;",o=H().getBool("WEBGL2_ISNAN_CUSTOM")?`
+`)[0]),console.log(`%c ${w.rightPad(u[0],i)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(l.join(`
+`))}function y2(e){return Ps(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function v2(e,t){if(me(e,()=>e.linkProgram(t)),!q().get("ENGINE_COMPILE_ONLY")&&e.getProgramParameter(t,e.LINK_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function Ih(e,t){if(me(e,()=>e.validateProgram(t)),e.getProgramParameter(t,e.VALIDATE_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function x2(e,t){let n=Ps(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function w2(e,t){let n=Ps(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,n)),me(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function rY(){return q().getNumber("WEBGL_VERSION")===2?1:4}function I2(e){return Ps(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function k2(e,t){let n=q().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){let r=`[${e}x${t}]`;throw new Error("Requested texture size "+r+" is invalid.")}if(e>n||t>n){let r=`[${e}x${t}]`,s=`[${n}x${n}]`;throw new Error("Requested texture size "+r+" greater than WebGL maximum on this browser / GPU "+s+".")}}function S2(e){return Ps(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function fv(e,t,n,r,s,a,o){let i=e.getAttribLocation(t,n);return i===-1?!1:(me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,r)),me(e,()=>e.vertexAttribPointer(i,s,e.FLOAT,!1,a,o)),me(e,()=>e.enableVertexAttribArray(i)),!0)}function T2(e,t,n){A2(e,n),me(e,()=>e.activeTexture(e.TEXTURE0+n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function sY(e,t){A2(e,t),me(e,()=>e.activeTexture(e.TEXTURE0+t)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function C2(e,t,n){return Ps(e,()=>e.getUniformLocation(t,n),'uniform "'+n+'" not present in program.')}function N2(e,t,n){return e.getUniformLocation(t,n)}function _2(e,t,n,r){me(e,()=>T2(e,t,r)),me(e,()=>e.uniform1i(n,r))}function aY(e){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),me(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function kh(e,t,n){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,n)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function mv(e,t){me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),me(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function Bl(e){let t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+E2(e,t))}function E2(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function Ps(e,t,n){let r=me(e,()=>t());if(r==null)throw new Error(n);return r}function A2(e,t){let n=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,r=t+e.TEXTURE0;if(r<e.TEXTURE0||r>n){let s=`[gl.TEXTURE0, gl.TEXTURE${n}]`;throw new Error(`textureUnit must be in ${s}.`)}}function ho(e,t=2){return w.sizeFromShape(e.slice(0,e.length-t))}function fo(e){if(e.length===0)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function Wl(e){let t=[1,1,1];return e.length===0||e.length===1&&e[0]===1||(t=[ho(e),...fo(e)]),t}function $2(e,t=!1){let n=q().getNumber("WEBGL_MAX_TEXTURE_SIZE"),r=q().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");r===1/0&&q().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")&&(r=n/2),t&&(n=n*2,r=r*2,e=e.map((i,c)=>c>=e.length-2?w.nearestLargerEven(e[c]):e[c]),e.length===1&&(e=[2,e[0]])),e.length!==2&&(e=w.squeezeShape(e).newShape);let s=w.sizeFromShape(e),a=null;e.length<=1&&s<=n?a=[1,s]:e.length===2&&e[0]<=n&&e[1]<=n?a=e:e.length===3&&e[0]*e[1]<=n&&e[2]<=n?a=[e[0]*e[1],e[2]]:e.length===3&&e[0]<=n&&e[1]*e[2]<=n?a=[e[0],e[1]*e[2]]:e.length===4&&e[0]*e[1]*e[2]<=n&&e[3]<=n?a=[e[0]*e[1]*e[2],e[3]]:e.length===4&&e[0]<=n&&e[1]*e[2]*e[3]<=n&&(a=[e[0],e[1]*e[2]*e[3]]);let o=a!=null&&Math.max(...a)>r&&Math.min(...a)<=(t?2:1)&&Math.min(...a)>0;if(a==null||o)if(t){let i=ho(e),c=2,u=2;e.length&&([c,u]=fo(e)),s=i*(c/2)*(u/2),a=w.sizeToSquarishShape(s).map(l=>l*2)}else a=w.sizeToSquarishShape(s);return a}function gh(e){return e%2===0}function ad(e,t){if(e=e.slice(-2),t=t.slice(-2),w.arraysEqual(e,t)||!e.length||!t.length||e[0]===0||e[1]===0||t[0]===0||t[1]===0)return!0;if(e.length!==t.length){let n=e.slice(-1)[0],r=t.slice(-1)[0];if(n===r||gh(n)&&gh(r)&&(e[0]===1||t[0]===1))return!0}return e[1]===t[1]&&gh(e[0])&&gh(t[0])}var Sh,Th;function D2(e){if(Sh==null){let t=Kr(e);Sh=t.getParameter(t.MAX_TEXTURE_SIZE)}return Sh}function oY(){Sh=null}function iY(){Th=null}function F2(e){if(Th==null){let t=Kr(e);Th=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,Th)}function R2(e){if(e===0)return 0;let t,n=Kr(e);return lr(n,"EXT_disjoint_timer_query_webgl2")&&e===2?t=2:lr(n,"EXT_disjoint_timer_query")?t=1:t=0,t}function lr(e,t){return e.getExtension(t)!=null}function gv(e){try{if(Kr(e)!=null)return!0}catch(t){return console.log("Error when getting WebGL context: ",t),!1}return!1}function P2(e){if(e===0)return!1;let t=Kr(e);if(e===1){if(!lr(t,"OES_texture_float"))return!1}else if(!lr(t,"EXT_color_buffer_float"))return!1;return bv(t)}function O2(e){if(e===0)return!1;let t=Kr(e);if(e===1){if(!lr(t,"OES_texture_float")||!lr(t,"WEBGL_color_buffer_float"))return!1}else{if(lr(t,"EXT_color_buffer_float"))return bv(t);let r="EXT_color_buffer_half_float";if(lr(t,r)){let s=t.getExtension(r);return cY(t,s)}return!1}return bv(t)}function bv(e){let t=X0(e),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n);let r=1,s=1;e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,r,s,0,t.textureFormatFloat,t.textureTypeFloat,null);let a=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,a),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);let o=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(a),o}function cY(e,t){let n=X0(e,t),r=e.createTexture();e.bindTexture(e.TEXTURE_2D,r);let s=1,a=1;e.texImage2D(e.TEXTURE_2D,0,n.internalFormatHalfFloat,s,a,0,n.textureFormatFloat,n.textureTypeHalfFloat,null);let o=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,o),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,r,0);let i=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(r),e.deleteFramebuffer(o),i}function M2(e){return e!==2?!1:Kr(e).fenceSync!=null}function Yu(e,t){Array.isArray(e)||(e=[e]),e.forEach(n=>{n!=null&&w.assert(n.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var we=q();we.registerFlag("HAS_WEBGL",()=>we.getNumber("WEBGL_VERSION")>0);we.registerFlag("WEBGL_VERSION",()=>gv(2)?2:gv(1)?1:0);we.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1);we.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>we.get("WEBGL_VERSION")===2);we.registerFlag("WEBGL_CPU_FORWARD",()=>!0);we.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1);we.registerFlag("WEBGL_PACK",()=>we.getBool("HAS_WEBGL"));we.registerFlag("WEBGL_PACK_NORMALIZATION",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_CLIP",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_PACK_REDUCE",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_LAZILY_UNPACK",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_CONV_IM2COL",()=>we.getBool("WEBGL_PACK"));we.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>D2(we.getNumber("WEBGL_VERSION")));we.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>F2(we.getNumber("WEBGL_VERSION")));we.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let e=we.getNumber("WEBGL_VERSION");return e===0?0:R2(e)});we.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>we.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!$d.isMobile());we.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>P2(we.getNumber("WEBGL_VERSION")));we.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>we.getBool("WEBGL_FORCE_F16_TEXTURES")?!1:we.getBool("WEBGL_RENDER_FLOAT32_CAPABLE"));we.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>O2(we.getNumber("WEBGL_VERSION")));we.registerFlag("WEBGL_FENCE_API_ENABLED",()=>M2(we.getNumber("WEBGL_VERSION")));we.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>we.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0);we.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)});we.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>$d.isMobile()?1:-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)});we.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128);we.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1);we.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5);we.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);we.registerFlag("WEBGL_EXP_CONV",()=>!1);we.registerFlag("SOFTWARE_WEBGL_ENABLED",()=>we.getBool("IS_TEST"));we.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE",()=>1/0);we.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE",()=>!1);we.registerFlag("WEBGL2_ISNAN_CUSTOM",()=>!1);function $n(){let e,t,n,r,s,a,o,i,c,u;return q().getNumber("WEBGL_VERSION")===2?(e="#version 300 es",t="in",n="out",r="in",s="texture",a="outputColor",o="out vec4 outputColor;",i=q().getBool("WEBGL2_ISNAN_CUSTOM")?`
       bool isnan_custom(float val) {
         uint floatToUint = floatBitsToUint(val);
         return (floatToUint & 0x7fffffffu) > 0x7f800000u;
@@ -74,7 +74,7 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       }
 
       #define isnan(value) isnan_custom(value)
-    `:"",l="",u=`
+    `:"",c="",u=`
       #define round(value) newRound(value)
       int newRound(float value) {
         return int(floor(value + 0.5));
@@ -83,7 +83,7 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 newRound(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `):(e="",t="attribute",n="varying",a="varying",r="texture2D",s="gl_FragColor",i="",o=`
+    `):(e="",t="attribute",n="varying",r="varying",s="texture2D",a="gl_FragColor",o="",i=`
       #define isnan(value) isnan_custom(value)
       bool isnan_custom(float val) {
         return (val > 0. || val < 1. || val == 0.) ? false : true;
@@ -91,7 +91,7 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       bvec4 isnan_custom(vec4 val) {
         return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));
       }
-    `,l=`
+    `,c=`
       uniform float INFINITY;
 
       bool isinf(float val) {
@@ -108,7 +108,7 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 round(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `),{version:e,attribute:t,varyingVs:n,varyingFs:a,texture2D:r,output:s,defineOutput:i,defineSpecialNaN:o,defineSpecialInf:l,defineRound:u}}function wo(e,t,n="index"){let a=v.computeStrides(t);return a.map((r,s)=>{let i=`int ${e[s]} = ${n} / ${r}`,o=s===a.length-1?`int ${e[s+1]} = ${n} - ${e[s]} * ${r}`:`index -= ${e[s]} * ${r}`;return`${i}; ${o};`}).join("")}function Kf(e,t,n="index"){let a=v.computeStrides(t);return a.map((r,s)=>{let i=`int ${e[s]} = ${n} / outShapeStrides[${s}]`,o=s===a.length-1?`int ${e[s+1]} = ${n} - ${e[s]} * outShapeStrides[${s}]`:`index -= ${e[s]} * outShapeStrides[${s}]`;return`${i}; ${o};`}).join("")}function iZ(e,t){let n=e.length,a=e.map(s=>`${t}[${s}]`),r=new Array(n-1);r[n-2]=a[n-1];for(let s=n-3;s>=0;--s)r[s]=`(${r[s+1]} * ${a[s+1]})`;return r}function oZ(e,t,n="index"){let a=e.map((s,i)=>i),r=iZ(a,t);return r.map((s,i)=>{let o=`int ${e[i]} = ${n} / ${r[i]}`,l=i===r.length-1?`int ${e[i+1]} = ${n} - ${e[i]} * ${r[i]}`:`index -= ${e[i]} * ${r[i]}`;return`${o}; ${l};`}).join("")}function Z0(e){let t=v.computeStrides(e).map(n=>n.toString());return`
+    `),{version:e,attribute:t,varyingVs:n,varyingFs:r,texture2D:s,output:a,defineOutput:o,defineSpecialNaN:i,defineSpecialInf:c,defineRound:u}}function Ii(e,t,n="index"){let r=w.computeStrides(t);return r.map((s,a)=>{let o=`int ${e[a]} = ${n} / ${s}`,i=a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${s}`:`index -= ${e[a]} * ${s}`;return`${o}; ${i};`}).join("")}function Xm(e,t,n="index"){let r=w.computeStrides(t);return r.map((s,a)=>{let o=`int ${e[a]} = ${n} / outShapeStrides[${a}]`,i=a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * outShapeStrides[${a}]`:`index -= ${e[a]} * outShapeStrides[${a}]`;return`${o}; ${i};`}).join("")}function uY(e,t){let n=e.length,r=e.map(a=>`${t}[${a}]`),s=new Array(n-1);s[n-2]=r[n-1];for(let a=n-3;a>=0;--a)s[a]=`(${s[a+1]} * ${r[a+1]})`;return s}function lY(e,t,n="index"){let r=e.map((a,o)=>o),s=uY(r,t);return s.map((a,o)=>{let i=`int ${e[o]} = ${n} / ${s[o]}`,c=o===s.length-1?`int ${e[o+1]} = ${n} - ${e[o]} * ${s[o]}`:`index -= ${e[o]} * ${s[o]}`;return`${i}; ${c};`}).join("")}function Z0(e){let t=w.computeStrides(e).map(n=>n.toString());return`
   int getFlatIndex(ivec3 coords) {
     return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;
   }
@@ -116,7 +116,7 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
   int getFlatIndex(ivec3 coords) {
     return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;
   }
-`}var P_=`
+`}var L2=`
   const float FLOAT_MAX = 1.70141184e38;
   const float FLOAT_MIN = 1.17549435e-38;
 
@@ -155,22 +155,22 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
 
     return c / 255.0;
   }
-`,{getBroadcastDims:O_}=N;function lZ(e,t,n){let a=[];if(e.forEach(c=>{let h=v.sizeFromShape(c.shapeInfo.logicalShape);if(c.shapeInfo.isUniform?a.push(`uniform float ${c.name}${h>1?`[${h}]`:""};`):(a.push(`uniform sampler2D ${c.name};`),a.push(`uniform int offset${c.name};`)),n.enableShapeUniforms){let{uniformShape:m}=Q0(n.packedInputs,c.shapeInfo.logicalShape,c.shapeInfo.texShape);switch(m.length){case 1:a.push(`uniform int ${c.name}Shape;`);break;case 2:a.push(`uniform ivec2 ${c.name}Shape;`);break;case 3:a.push(`uniform ivec3 ${c.name}Shape;`);break;case 4:a.push(`uniform ivec4 ${c.name}Shape;`);break;default:break}a.push(`uniform ivec2 ${c.name}TexShape;`)}}),n.enableShapeUniforms){switch(t.logicalShape.length){case 1:a.push("uniform int outShape;");break;case 2:a.push("uniform ivec2 outShape;"),a.push("uniform int outShapeStrides;");break;case 3:a.push("uniform ivec3 outShape;"),a.push("uniform ivec2 outShapeStrides;");break;case 4:a.push("uniform ivec4 outShape;"),a.push("uniform ivec3 outShapeStrides;");break;default:break}a.push("uniform ivec2 outTexShape;")}n.customUniforms&&n.customUniforms.forEach(c=>{a.push(`uniform ${c.type} ${c.name}${c.arrayIndex?`[${c.arrayIndex}]`:""};`)});let r=a.join(`
-`),s=e.map(c=>uZ(c,t,n.packedInputs,n.enableShapeUniforms)).join(`
-`),i=t.texShape,o=_n(),l=dZ(o),u,p,d=fZ(o);return t.isPacked?(u=pZ(t.logicalShape,i,n.enableShapeUniforms),p=mZ(o)):(u=cZ(t.logicalShape,i,n.enableShapeUniforms),p=hZ(o)),n.packedInputs&&(d+=xZ),[d,l,p,r,u,s,n.userCode].join(`
-`)}function Zu(e,t=!1){let n=e.shapeInfo.logicalShape;switch(n.length){case 0:return $Z(e,t);case 1:return DZ(e,t);case 2:return MZ(e,t);case 3:return OZ(e,t);case 4:return zZ(e,t);case 5:return WZ(e);case 6:return BZ(e);default:throw new Error(`${n.length}-D input sampling is not yet supported`)}}function L_(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return AZ(e);case 1:return FZ(e,t);case 2:return RZ(e,t);case 3:return PZ(e,t);default:return LZ(e,t)}}function uZ(e,t,n=!1,a){let r="";n?r+=L_(e,a):r+=Zu(e,a);let s=e.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(n?r+=VZ(e,t):r+=UZ(e,t)),r}function pZ(e,t,n){switch(e.length){case 0:return z_();case 1:return vZ(e,t,n);case 2:return _Z(e,t,n);case 3:return kZ(e,t,n);default:return SZ(e,t,n)}}function cZ(e,t,n){switch(e.length){case 0:return z_();case 1:return wZ(e,t,n);case 2:return EZ(e,t,n);case 3:return IZ(e,t,n);case 4:return TZ(e,t,n);case 5:return NZ(e,t);case 6:return CZ(e,t);default:throw new Error(`${e.length}-D output sampling is not yet supported`)}}function dZ(e){return`
+`,{getBroadcastDims:z2}=N;function dY(e,t,n){let r=[];if(e.forEach(h=>{let f=w.sizeFromShape(h.shapeInfo.logicalShape);if(h.shapeInfo.isUniform?r.push(`uniform float ${h.name}${f>1?`[${f}]`:""};`):(r.push(`uniform sampler2D ${h.name};`),r.push(`uniform int offset${h.name};`)),n.enableShapeUniforms){let{uniformShape:m}=Q0(n.packedInputs,h.shapeInfo.logicalShape,h.shapeInfo.texShape);switch(m.length){case 1:r.push(`uniform int ${h.name}Shape;`);break;case 2:r.push(`uniform ivec2 ${h.name}Shape;`);break;case 3:r.push(`uniform ivec3 ${h.name}Shape;`);break;case 4:r.push(`uniform ivec4 ${h.name}Shape;`);break;default:break}r.push(`uniform ivec2 ${h.name}TexShape;`)}}),n.enableShapeUniforms){switch(t.logicalShape.length){case 1:r.push("uniform int outShape;");break;case 2:r.push("uniform ivec2 outShape;"),r.push("uniform int outShapeStrides;");break;case 3:r.push("uniform ivec3 outShape;"),r.push("uniform ivec2 outShapeStrides;");break;case 4:r.push("uniform ivec4 outShape;"),r.push("uniform ivec3 outShapeStrides;");break;default:break}r.push("uniform ivec2 outTexShape;")}n.customUniforms&&n.customUniforms.forEach(h=>{r.push(`uniform ${h.type} ${h.name}${h.arrayIndex?`[${h.arrayIndex}]`:""};`)});let s=r.join(`
+`),a=e.map(h=>pY(h,t,n.packedInputs,n.enableShapeUniforms)).join(`
+`),o=t.texShape,i=$n(),c=mY(i),u,l,p=yY(i);return t.isPacked?(u=hY(t.logicalShape,o,n.enableShapeUniforms),l=bY(i)):(u=fY(t.logicalShape,o,n.enableShapeUniforms),l=gY(i)),n.packedInputs&&(p+=IY),[p,c,l,s,u,a,n.userCode].join(`
+`)}function Zu(e,t=!1){let n=e.shapeInfo.logicalShape;switch(n.length){case 0:return RY(e,t);case 1:return OY(e,t);case 2:return LY(e,t);case 3:return BY(e,t);case 4:return VY(e,t);case 5:return UY(e);case 6:return GY(e);default:throw new Error(`${n.length}-D input sampling is not yet supported`)}}function B2(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return FY(e);case 1:return PY(e,t);case 2:return MY(e,t);case 3:return zY(e,t);default:return WY(e,t)}}function pY(e,t,n=!1,r){let s="";n?s+=B2(e,r):s+=Zu(e,r);let a=e.shapeInfo.logicalShape,o=t.logicalShape;return a.length<=o.length&&(n?s+=HY(e,t):s+=qY(e,t)),s}function hY(e,t,n){switch(e.length){case 0:return W2();case 1:return kY(e,t,n);case 2:return $Y(e,t,n);case 3:return TY(e,t,n);default:return NY(e,t,n)}}function fY(e,t,n){switch(e.length){case 0:return W2();case 1:return SY(e,t,n);case 2:return DY(e,t,n);case 3:return CY(e,t,n);case 4:return _Y(e,t,n);case 5:return EY(e,t);case 6:return AY(e,t);default:throw new Error(`${e.length}-D output sampling is not yet supported`)}}function mY(e){return`
     float sampleTexture(sampler2D textureSampler, vec2 uv) {
       return ${e.texture2D}(textureSampler, uv).r;
     }
-  `}function hZ(e){return`
+  `}function gY(e){return`
     void setOutput(float val) {
       ${e.output} = vec4(val, 0, 0, 0);
     }
-  `}function mZ(e){return`
+  `}function bY(e){return`
     void setOutput(vec4 val) {
       ${e.output} = val;
     }
-  `}function fZ(e){return`${e.version}
+  `}function yY(e){return`${e.version}
     precision highp float;
     precision highp int;
     precision highp sampler2D;
@@ -225,10 +225,10 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       return fract((p3.x + p3.y) * p3.z);
     }
 
-    ${gZ}
-    ${yZ}
-    ${bZ}
-  `}var gZ=`
+    ${vY}
+    ${xY}
+    ${wY}
+  `}var vY=`
 vec2 uvFromFlat(int texNumR, int texNumC, int index) {
   int texR = index / texNumC;
   int texC = index - texR * texNumC;
@@ -240,7 +240,7 @@ vec2 packedUVfrom1D(int texNumR, int texNumC, int index) {
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,yZ=`
+`,xY=`
 vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texNumC, int row, int col) {
   int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);
@@ -248,7 +248,7 @@ vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,bZ=`
+`,wY=`
 vec2 packedUVfrom3D(int texNumR, int texNumC,
     int texelsInBatch, int texelsInLogicalRow, int b,
     int row, int col) {
@@ -257,7 +257,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   int texC = index - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,xZ=`
+`,IY=`
   float getChannel(vec4 frag, vec2 innerDims) {
     vec2 modCoord = mod(innerDims, 2.);
     return modCoord.x == 0. ?
@@ -268,25 +268,25 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
     float modCoord = mod(float(dim), 2.);
     return modCoord == 0. ? frag.r : frag.g;
   }
-`;function z_(){return`
+`;function W2(){return`
     int getOutputCoords() {
       return 0;
     }
-  `}function vZ(e,t,n){let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return a[0]===1?n?`
+  `}function kY(e,t,n){let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return r[0]===1?n?`
       int getOutputCoords() {
         return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));
       }
     `:`
       int getOutputCoords() {
-        return 2 * int(resultUV.x * ${a[1]}.0);
+        return 2 * int(resultUV.x * ${r[1]}.0);
       }
-    `:a[1]===1?n?`
+    `:r[1]===1?n?`
       int getOutputCoords() {
         return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));
       }
     `:`
       int getOutputCoords() {
-        return 2 * int(resultUV.y * ${a[0]}.0);
+        return 2 * int(resultUV.y * ${r[0]}.0);
       }
     `:n?`
     int getOutputCoords() {
@@ -298,10 +298,10 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   `:`
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      return 2 * (resTexRC.x * ${a[1]} + resTexRC.y);
+                             vec2(${r[0]}, ${r[1]}));
+      return 2 * (resTexRC.x * ${r[1]} + resTexRC.y);
     }
-  `}function wZ(e,t,n){return t[0]===1?n?`
+  `}function SY(e,t,n){return t[0]===1?n?`
       int getOutputCoords() {
         return int(resultUV.x * float(outTexShape[1]));
       }
@@ -329,7 +329,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
                              vec2(${t[0]}, ${t[1]}));
       return resTexRC.x * ${t[1]} + resTexRC.y;
     }
-  `}function kZ(e,t,n){if(n)return`
+  `}function TY(e,t,n){if(n)return`
     ivec3 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));
@@ -346,37 +346,37 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec3(b, r, c);
     }
-  `;let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],r=Math.ceil(e[2]/2),s=r*Math.ceil(e[1]/2);return`
+  `;let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],s=Math.ceil(e[2]/2),a=s*Math.ceil(e[1]/2);return`
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
+                             vec2(${r[0]}, ${r[1]}));
+      int index = resTexRC.x * ${r[1]} + resTexRC.y;
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${a};
+      index -= b * ${a};
 
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int r = 2 * (index / ${s});
+      int c = imod(index, ${s}) * 2;
 
       return ivec3(b, r, c);
     }
-  `}function IZ(e,t,n){if(n)return`
+  `}function CY(e,t,n){if(n)return`
   ivec3 getOutputCoords() {
     ivec2 resTexRC = ivec2(resultUV.yx *
                            vec2(outTexShape[0], outTexShape[1]));
     int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-    ${Kf(["r","c","d"],e)}
+    ${Xm(["r","c","d"],e)}
     return ivec3(r, c, d);
   }
-`;let a=wo(["r","c","d"],e);return`
+`;let r=Ii(["r","c","d"],e);return`
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(${t[0]}, ${t[1]}));
       int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${a}
+      ${r}
       return ivec3(r, c, d);
     }
-  `}function SZ(e,t,n){if(n)return`
+  `}function NY(e,t,n){if(n)return`
     ivec4 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       ivec2 resTexRC = ivec2(resultUV.yx *
@@ -398,42 +398,42 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec4(b2, b, r, c);
     }
-  `;let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],r=Math.ceil(e[e.length-1]/2),s=r*Math.ceil(e[e.length-2]/2),i=s,o="",l="b, r, c";for(let u=2;u<e.length-1;u++)i*=e[e.length-u-1],o=`
-      int b${u} = index / ${i};
-      index -= b${u} * ${i};
-    `+o,l=`b${u}, `+l;return`
+  `;let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],s=Math.ceil(e[e.length-1]/2),a=s*Math.ceil(e[e.length-2]/2),o=a,i="",c="b, r, c";for(let u=2;u<e.length-1;u++)o*=e[e.length-u-1],i=`
+      int b${u} = index / ${o};
+      index -= b${u} * ${o};
+    `+i,c=`b${u}, `+c;return`
     ivec${e.length} getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
+                             vec2(${r[0]}, ${r[1]}));
+      int index = resTexRC.x * ${r[1]} + resTexRC.y;
 
-      ${o}
+      ${i}
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${a};
+      index -= b * ${a};
 
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int r = 2 * (index / ${s});
+      int c = imod(index, ${s}) * 2;
 
-      return ivec${e.length}(${l});
+      return ivec${e.length}(${c});
     }
-  `}function TZ(e,t,n){if(n)return`
+  `}function _Y(e,t,n){if(n)return`
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
         vec2(outTexShape[0], outTexShape[1]));
       int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-      ${Kf(["r","c","d","d2"],e)}
+      ${Xm(["r","c","d","d2"],e)}
       return ivec4(r, c, d, d2);
     }
-  `;let a=wo(["r","c","d","d2"],e);return`
+  `;let r=Ii(["r","c","d","d2"],e);return`
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
         vec2(${t[0]}, ${t[1]}));
       int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${a}
+      ${r}
       return ivec4(r, c, d, d2);
     }
-  `}function NZ(e,t){let n=wo(["r","c","d","d2","d3"],e);return`
+  `}function EY(e,t){let n=Ii(["r","c","d","d2","d3"],e);return`
     ivec5 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},
                              ${t[1]}));
@@ -445,7 +445,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       ivec5 outShape = ivec5(r, c, d, d2, d3);
       return outShape;
     }
-  `}function CZ(e,t){let n=wo(["r","c","d","d2","d3","d4"],e);return`
+  `}function AY(e,t){let n=Ii(["r","c","d","d2","d3","d4"],e);return`
     ivec6 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
         vec2(${t[0]}, ${t[1]}));
@@ -456,16 +456,16 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       ivec6 result = ivec6(r, c, d, d2, d3, d4);
       return result;
     }
-  `}function _Z(e,t,n){let a=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(v.arraysEqual(e,t))return n?`
+  `}function $Y(e,t,n){let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(w.arraysEqual(e,t))return n?`
       ivec2 getOutputCoords() {
         ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
         return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));
       }
     `:`
       ivec2 getOutputCoords() {
-        return 2 * ivec2(resultUV.yx * vec2(${a[0]}, ${a[1]}));
+        return 2 * ivec2(resultUV.yx * vec2(${r[0]}, ${r[1]}));
       }
-    `;let r=Math.ceil(e[1]/2);return n?`
+    `;let s=Math.ceil(e[1]/2);return n?`
     ivec2 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));
@@ -481,15 +481,15 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   `:`
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${a[0]}, ${a[1]}));
+                             vec2(${r[0]}, ${r[1]}));
 
-      int index = resTexRC.x * ${a[1]} + resTexRC.y;
-      int r = 2 * (index / ${r});
-      int c = imod(index, ${r}) * 2;
+      int index = resTexRC.x * ${r[1]} + resTexRC.y;
+      int r = 2 * (index / ${s});
+      int c = imod(index, ${s}) * 2;
 
       return ivec2(r, c);
     }
-  `}function EZ(e,t,n){return v.arraysEqual(e,t)?n?`
+  `}function DY(e,t,n){return w.arraysEqual(e,t)?n?`
       ivec2 getOutputCoords() {
         return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));
       }
@@ -543,237 +543,237 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       int c = index - r * ${e[1]};
       return ivec2(r, c);
     }
-  `}function ko(e){return`offset${e}`}function AZ(e){let t=e.name,n="get"+t.charAt(0).toUpperCase()+t.slice(1),a=_n();return`
+  `}function ki(e){return`offset${e}`}function FY(e){let t=e.name,n="get"+t.charAt(0).toUpperCase()+t.slice(1),r=$n();return`
     vec4 ${n}() {
-      return ${a.texture2D}(${t}, halfCR);
+      return ${r.texture2D}(${t}, halfCR);
     }
-  `}function $Z(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`float ${a}() {return ${n};}`;let[r,s]=e.shapeInfo.texShape;if(r===1&&s===1)return`
-      float ${a}() {
+  `}function RY(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`float ${r}() {return ${n};}`;let[s,a]=e.shapeInfo.texShape;if(s===1&&a===1)return`
+      float ${r}() {
         return sampleTexture(${n}, halfCR);
       }
-    `;let i=ko(n);if(t)return`
-    float ${a}() {
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${i});
+    `;let o=ki(n);if(t)return`
+    float ${r}() {
+      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${o});
       return sampleTexture(${n}, uv);
     }
-  `;let[o,l]=e.shapeInfo.texShape;return`
-    float ${a}() {
-      vec2 uv = uvFromFlat(${o}, ${l}, ${i});
+  `;let[i,c]=e.shapeInfo.texShape;return`
+    float ${r}() {
+      vec2 uv = uvFromFlat(${i}, ${c}, ${o});
       return sampleTexture(${n}, uv);
     }
-  `}function FZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=e.shapeInfo.texShape,s=_n();if(t)return`
-    vec4 ${a}(int index) {
+  `}function PY(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e.shapeInfo.texShape,a=$n();if(t)return`
+    vec4 ${r}(int index) {
       ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
       vec2 uv = packedUVfrom1D(
         packedTexShape[0], packedTexShape[1], index);
-      return ${s.texture2D}(${n}, uv);
+      return ${a.texture2D}(${n}, uv);
     }
-  `;let i=[Math.ceil(r[0]/2),Math.ceil(r[1]/2)];return`
-    vec4 ${a}(int index) {
+  `;let o=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];return`
+    vec4 ${r}(int index) {
       vec2 uv = packedUVfrom1D(
-        ${i[0]}, ${i[1]}, index);
-      return ${s.texture2D}(${n}, uv);
+        ${o[0]}, ${o[1]}, index);
+      return ${a.texture2D}(${n}, uv);
     }
-  `}function DZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`
-      float ${a}(int index) {
+  `}function OY(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`
+      float ${r}(int index) {
         ${Ju(e)}
       }
-    `;let r=e.shapeInfo.texShape,s=r[0],i=r[1];if(i===1&&s===1)return`
-      float ${a}(int index) {
+    `;let s=e.shapeInfo.texShape,a=s[0],o=s[1];if(o===1&&a===1)return`
+      float ${r}(int index) {
         return sampleTexture(${n}, halfCR);
       }
-    `;let o=ko(n);return i===1?t?`
-      float ${a}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${n}TexShape[0]));
+    `;let i=ki(n);return o===1?t?`
+      float ${r}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${i}) + 0.5) / float(${n}TexShape[0]));
         return sampleTexture(${n}, uv);
       }
     `:`
-      float ${a}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);
+      float ${r}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${i}) + 0.5) / ${a}.0);
         return sampleTexture(${n}, uv);
       }
-    `:s===1?t?`
-      float ${a}(int index) {
-        vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${n}TexShape[1]), 0.5);
+    `:a===1?t?`
+      float ${r}(int index) {
+        vec2 uv = vec2((float(index + ${i}) + 0.5) / float(${n}TexShape[1]), 0.5);
         return sampleTexture(${n}, uv);
       }
     `:`
-      float ${a}(int index) {
-        vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);
+      float ${r}(int index) {
+        vec2 uv = vec2((float(index + ${i}) + 0.5) / ${o}.0, 0.5);
         return sampleTexture(${n}, uv);
       }
     `:t?`
-    float ${a}(int index) {
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${o});
+    float ${r}(int index) {
+      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${i});
       return sampleTexture(${n}, uv);
     }
   `:`
-    float ${a}(int index) {
-      vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});
+    float ${r}(int index) {
+      vec2 uv = uvFromFlat(${a}, ${o}, index + ${i});
       return sampleTexture(${n}, uv);
     }
-  `}function RZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape,i=s[0],o=s[1],l=_n();if(s!=null&&v.arraysEqual(n,s))return t?`
-      vec4 ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
+  `}function MY(e,t){let n=e.shapeInfo.logicalShape,r=e.name,s="get"+r.charAt(0).toUpperCase()+r.slice(1),a=e.shapeInfo.texShape,o=a[0],i=a[1],c=$n();if(a!=null&&w.arraysEqual(n,a))return t?`
+      vec4 ${s}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);
 
-        return ${l.texture2D}(${a}, uv);
+        return ${c.texture2D}(${r}, uv);
       }
     `:`
-      vec4 ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);
+      vec4 ${s}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${i}.0, ${o}.0);
 
-        return ${l.texture2D}(${a}, uv);
+        return ${c.texture2D}(${r}, uv);
       }
     `;if(t)return`
-    vec4 ${r}(int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${a}TexShape[0]) / 2.0), ceil(float(${a}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${a}Shape[1]) / 2.0));
+    vec4 ${s}(int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${r}Shape[1]) / 2.0));
       vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
-      return ${l.texture2D}(${a}, uv);
+      return ${c.texture2D}(${r}, uv);
     }
-  `;let u=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],p=Math.ceil(n[1]/2);return`
-    vec4 ${r}(int row, int col) {
-      vec2 uv = packedUVfrom2D(${p}, ${u[0]}, ${u[1]}, row, col);
-      return ${l.texture2D}(${a}, uv);
+  `;let u=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)],l=Math.ceil(n[1]/2);return`
+    vec4 ${s}(int row, int col) {
+      vec2 uv = packedUVfrom2D(${l}, ${u[0]}, ${u[1]}, row, col);
+      return ${c.texture2D}(${r}, uv);
     }
-  `}function MZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape;if(s!=null&&v.arraysEqual(n,s)){if(t)return`
-      float ${r}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+  `}function LY(e,t){let n=e.shapeInfo.logicalShape,r=e.name,s="get"+r.charAt(0).toUpperCase()+r.slice(1),a=e.shapeInfo.texShape;if(a!=null&&w.arraysEqual(n,a)){if(t)return`
+      float ${s}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);
+        return sampleTexture(${r}, uv);
       }
-    `;let c=s[0],h=s[1];return`
-    float ${r}(int row, int col) {
-      vec2 uv = (vec2(col, row) + halfCR) / vec2(${h}.0, ${c}.0);
-      return sampleTexture(${a}, uv);
+    `;let d=a[0],h=a[1];return`
+    float ${s}(int row, int col) {
+      vec2 uv = (vec2(col, row) + halfCR) / vec2(${h}.0, ${d}.0);
+      return sampleTexture(${r}, uv);
     }
-  `}let{newShape:i,keptDims:o}=v.squeezeShape(n),l=i;if(l.length<n.length){let c=Qu(e,l),h=["row","col"];return`
-      ${Zu(c,t)}
-      float ${r}(int row, int col) {
-        return ${r}(${ep(h,o)});
+  `}let{newShape:o,keptDims:i}=w.squeezeShape(n),c=o;if(c.length<n.length){let d=Qu(e,c),h=["row","col"];return`
+      ${Zu(d,t)}
+      float ${s}(int row, int col) {
+        return ${s}(${el(h,i)});
       }
     `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col) {
+      float ${s}(int row, int col) {
         int index = round(dot(vec2(row, col), vec2(${n[1]}, 1)));
         ${Ju(e)}
       }
-    `;let u=s[0],p=s[1],d=ko(a);return p===1?t?`
-      float ${r}(int row, int col) {
-        float index = dot(vec3(row, col, ${d}), vec3(${a}Shape[1], 1, 1));
-        vec2 uv = vec2(0.5, (index + 0.5) / float(${a}TexShape[0]));
-        return sampleTexture(${a}, uv);
+    `;let u=a[0],l=a[1],p=ki(r);return l===1?t?`
+      float ${s}(int row, int col) {
+        float index = dot(vec3(row, col, ${p}), vec3(${r}Shape[1], 1, 1));
+        vec2 uv = vec2(0.5, (index + 0.5) / float(${r}TexShape[0]));
+        return sampleTexture(${r}, uv);
       }
     `:`
-    float ${r}(int row, int col) {
-      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));
+    float ${s}(int row, int col) {
+      float index = dot(vec3(row, col, ${p}), vec3(${n[1]}, 1, 1));
       vec2 uv = vec2(0.5, (index + 0.5) / ${u}.0);
-      return sampleTexture(${a}, uv);
+      return sampleTexture(${r}, uv);
     }
   `:u===1?t?`
-      float ${r}(int row, int col) {
-        float index = dot(vec3(row, col, ${d}), vec3(${a}Shape[1], 1, 1));
-        vec2 uv = vec2((index + 0.5) / float(${a}TexShape[1]), 0.5);
-        return sampleTexture(${a}, uv);
+      float ${s}(int row, int col) {
+        float index = dot(vec3(row, col, ${p}), vec3(${r}Shape[1], 1, 1));
+        vec2 uv = vec2((index + 0.5) / float(${r}TexShape[1]), 0.5);
+        return sampleTexture(${r}, uv);
       }
     `:`
-    float ${r}(int row, int col) {
-      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));
-      vec2 uv = vec2((index + 0.5) / ${p}.0, 0.5);
-      return sampleTexture(${a}, uv);
+    float ${s}(int row, int col) {
+      float index = dot(vec3(row, col, ${p}), vec3(${n[1]}, 1, 1));
+      vec2 uv = vec2((index + 0.5) / ${l}.0, 0.5);
+      return sampleTexture(${r}, uv);
     }
   `:t?`
-      float ${r}(int row, int col) {
+      float ${s}(int row, int col) {
         // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${a}Shape[1] + col + ${d};
-        vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index);
-        return sampleTexture(${a}, uv);
+        int index = row * ${r}Shape[1] + col + ${p};
+        vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);
+        return sampleTexture(${r}, uv);
       }
     `:`
-  float ${r}(int row, int col) {
+  float ${s}(int row, int col) {
     // Explicitly use integer operations as dot() only works on floats.
-    int index = row * ${n[1]} + col + ${d};
-    vec2 uv = uvFromFlat(${u}, ${p}, index);
-    return sampleTexture(${a}, uv);
+    int index = row * ${n[1]} + col + ${p};
+    vec2 uv = uvFromFlat(${u}, ${l}, index);
+    return sampleTexture(${r}, uv);
   }
-`}function PZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=e.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(n[0]===1){let c=n.slice(1),h=[1,2],m=Qu(e,c),f=["b","row","col"];return`
-        ${L_(m,t)}
-        vec4 ${r}(int b, int row, int col) {
-          return ${r}(${ep(f,h)});
+`}function zY(e,t){let n=e.shapeInfo.logicalShape,r=e.name,s="get"+r.charAt(0).toUpperCase()+r.slice(1),a=e.shapeInfo.texShape,o=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)];if(n[0]===1){let d=n.slice(1),h=[1,2],f=Qu(e,d),m=["b","row","col"];return`
+        ${B2(f,t)}
+        vec4 ${s}(int b, int row, int col) {
+          return ${s}(${el(m,h)});
         }
-      `}let o=_n();if(t)return`
-    vec4 ${r}(int b, int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${a}TexShape[0]) / 2.0), ceil(float(${a}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${a}Shape[2]) / 2.0));
-      int texelsInBatch = valuesPerRow * int(ceil(float(${a}Shape[1]) / 2.0));
+      `}let i=$n();if(t)return`
+    vec4 ${s}(int b, int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${r}Shape[2]) / 2.0));
+      int texelsInBatch = valuesPerRow * int(ceil(float(${r}Shape[1]) / 2.0));
       vec2 uv = packedUVfrom3D(
         packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);
-      return ${o.texture2D}(${a}, uv);
+      return ${i.texture2D}(${r}, uv);
     }
-  `;let l=i[0],u=i[1],p=Math.ceil(n[2]/2),d=p*Math.ceil(n[1]/2);return`
-    vec4 ${r}(int b, int row, int col) {
+  `;let c=o[0],u=o[1],l=Math.ceil(n[2]/2),p=l*Math.ceil(n[1]/2);return`
+    vec4 ${s}(int b, int row, int col) {
       vec2 uv = packedUVfrom3D(
-        ${l}, ${u}, ${d}, ${p}, b, row, col);
-      return ${o.texture2D}(${a}, uv);
+        ${c}, ${u}, ${p}, ${l}, b, row, col);
+      return ${i.texture2D}(${r}, uv);
     }
-  `}function OZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=n[1]*n[2],i=n[2],{newShape:o,keptDims:l}=v.squeezeShape(n),u=o;if(u.length<n.length){let f=Qu(e,u),g=["row","col","depth"];return`
-        ${Zu(f,t)}
-        float ${r}(int row, int col, int depth) {
-          return ${r}(${ep(g,l)});
+  `}function BY(e,t){let n=e.shapeInfo.logicalShape,r=e.name,s="get"+r.charAt(0).toUpperCase()+r.slice(1),a=n[1]*n[2],o=n[2],{newShape:i,keptDims:c}=w.squeezeShape(n),u=i;if(u.length<n.length){let m=Qu(e,u),g=["row","col","depth"];return`
+        ${Zu(m,t)}
+        float ${s}(int row, int col, int depth) {
+          return ${s}(${el(g,c)});
         }
       `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col, int depth) {
+      float ${s}(int row, int col, int depth) {
         int index = round(dot(vec3(row, col, depth),
-                          vec3(${s}, ${i}, 1)));
+                          vec3(${a}, ${o}, 1)));
         ${Ju(e)}
       }
-    `;let p=e.shapeInfo.texShape,d=p[0],c=p[1],h=e.shapeInfo.flatOffset;if(c===s&&h==null)return t?`
-      float ${r}(int row, int col, int depth) {
-        int stride1 = ${a}Shape[2];
+    `;let l=e.shapeInfo.texShape,p=l[0],d=l[1],h=e.shapeInfo.flatOffset;if(d===a&&h==null)return t?`
+      float ${s}(int row, int col, int depth) {
+        int stride1 = ${r}Shape[2];
         float texR = float(row);
         float texC = dot(vec2(col, depth), vec2(stride1, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                   vec2(${r}TexShape[1], ${r}TexShape[0]);
+        return sampleTexture(${r}, uv);
       }
     `:`
-        float ${r}(int row, int col, int depth) {
+        float ${s}(int row, int col, int depth) {
           float texR = float(row);
-          float texC = dot(vec2(col, depth), vec2(${i}, 1));
+          float texC = dot(vec2(col, depth), vec2(${o}, 1));
           vec2 uv = (vec2(texC, texR) + halfCR) /
-                     vec2(${c}.0, ${d}.0);
-          return sampleTexture(${a}, uv);
+                     vec2(${d}.0, ${p}.0);
+          return sampleTexture(${r}, uv);
         }
-      `;if(c===i&&h==null)return t?`
-      float ${r}(int row, int col, int depth) {
-        float texR = dot(vec2(row, col), vec2(${a}Shape[1], 1));
+      `;if(d===o&&h==null)return t?`
+      float ${s}(int row, int col, int depth) {
+        float texR = dot(vec2(row, col), vec2(${r}Shape[1], 1));
         float texC = float(depth);
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);
+        return sampleTexture(${r}, uv);
       }
     `:`
-    float ${r}(int row, int col, int depth) {
+    float ${s}(int row, int col, int depth) {
       float texR = dot(vec2(row, col), vec2(${n[1]}, 1));
       float texC = float(depth);
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}.0, ${d}.0);
-      return sampleTexture(${a}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${d}.0, ${p}.0);
+      return sampleTexture(${r}, uv);
     }
-  `;let m=ko(a);return t?`
-    float ${r}(int row, int col, int depth) {
+  `;let f=ki(r);return t?`
+    float ${s}(int row, int col, int depth) {
       // Explicitly use integer operations as dot() only works on floats.
-      int stride0 = ${a}Shape[1] * ${a}Shape[2];
-      int stride1 = ${a}Shape[2];
-      int index = row * stride0 + col * stride1 + depth + ${m};
-      vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index);
-      return sampleTexture(${a}, uv);
+      int stride0 = ${r}Shape[1] * ${r}Shape[2];
+      int stride1 = ${r}Shape[2];
+      int index = row * stride0 + col * stride1 + depth + ${f};
+      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);
+      return sampleTexture(${r}, uv);
     }
     `:`
-      float ${r}(int row, int col, int depth) {
+      float ${s}(int row, int col, int depth) {
         // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${s} + col * ${i} + depth + ${m};
-        vec2 uv = uvFromFlat(${d}, ${c}, index);
-        return sampleTexture(${a}, uv);
+        int index = row * ${a} + col * ${o} + depth + ${f};
+        vec2 uv = uvFromFlat(${p}, ${d}, index);
+        return sampleTexture(${r}, uv);
       }
-  `}function LZ(e,t){let n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=_n();if(t)return`
-    vec4 ${a}(int b2, int b, int row, int col) {
+  `}function WY(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),s=$n();if(t)return`
+    vec4 ${r}(int b2, int b, int row, int col) {
       int valuesPerRow = int(ceil(float(${n}Shape[3]) / 2.0));
       int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[2]) / 2.0));
       int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);
@@ -782,157 +782,157 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
       int texR = index / packedTexShape[1];
       int texC = index - texR * packedTexShape[1];
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${r.texture2D}(${n}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${s.texture2D}(${n}, uv);
     }
-  `;let s=e.shapeInfo.logicalShape,i=s.length,o=e.shapeInfo.texShape,l=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)],u=l[0],p=l[1],d=Math.ceil(s[i-1]/2),c=d*Math.ceil(s[i-2]/2),h="int b, int row, int col",m=`b * ${c} + (row / 2) * ${d} + (col / 2)`;for(let f=2;f<i-1;f++)h=`int b${f}, `+h,c*=s[i-f-1],m=`b${f} * ${c} + `+m;return`
-    vec4 ${a}(${h}) {
-      int index = ${m};
-      int texR = index / ${p};
-      int texC = index - texR * ${p};
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}, ${u});
-      return ${r.texture2D}(${n}, uv);
+  `;let a=e.shapeInfo.logicalShape,o=a.length,i=e.shapeInfo.texShape,c=[Math.ceil(i[0]/2),Math.ceil(i[1]/2)],u=c[0],l=c[1],p=Math.ceil(a[o-1]/2),d=p*Math.ceil(a[o-2]/2),h="int b, int row, int col",f=`b * ${d} + (row / 2) * ${p} + (col / 2)`;for(let m=2;m<o-1;m++)h=`int b${m}, `+h,d*=a[o-m-1],f=`b${m} * ${d} + `+f;return`
+    vec4 ${r}(${h}) {
+      int index = ${f};
+      int texR = index / ${l};
+      int texC = index - texR * ${l};
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${l}, ${u});
+      return ${s.texture2D}(${n}, uv);
     }
-  `}function zZ(e,t){let n=e.shapeInfo.logicalShape,a=e.name,r="get"+a.charAt(0).toUpperCase()+a.slice(1),s=n[3],i=n[2]*s,o=n[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(n);if(l.length<n.length){let b=Qu(e,l),x=["row","col","depth","depth2"];return`
-      ${Zu(b,t)}
-      float ${r}(int row, int col, int depth, int depth2) {
-        return ${r}(${ep(x,u)});
+  `}function VY(e,t){let n=e.shapeInfo.logicalShape,r=e.name,s="get"+r.charAt(0).toUpperCase()+r.slice(1),a=n[3],o=n[2]*a,i=n[1]*o,{newShape:c,keptDims:u}=w.squeezeShape(n);if(c.length<n.length){let y=Qu(e,c),v=["row","col","depth","depth2"];return`
+      ${Zu(y,t)}
+      float ${s}(int row, int col, int depth, int depth2) {
+        return ${s}(${el(v,u)});
       }
     `}if(e.shapeInfo.isUniform)return`
-      float ${r}(int row, int col, int depth, int depth2) {
+      float ${s}(int row, int col, int depth, int depth2) {
         int index = round(dot(vec4(row, col, depth, depth2),
-                          vec4(${o}, ${i}, ${s}, 1)));
+                          vec4(${i}, ${o}, ${a}, 1)));
         ${Ju(e)}
       }
-    `;let p=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,c=d[0],h=d[1],m=`int stride2 = ${a}Shape[3];`,f=`int stride1 = ${a}Shape[2] * stride2;`,g=`int stride0 = ${a}Shape[1] * stride1;`;if(h===o&&p==null)return t?`
-      float ${r}(int row, int col, int depth, int depth2) {
-        ${m}
+    `;let l=e.shapeInfo.flatOffset,p=e.shapeInfo.texShape,d=p[0],h=p[1],f=`int stride2 = ${r}Shape[3];`,m=`int stride1 = ${r}Shape[2] * stride2;`,g=`int stride0 = ${r}Shape[1] * stride1;`;if(h===i&&l==null)return t?`
+      float ${s}(int row, int col, int depth, int depth2) {
         ${f}
+        ${m}
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
                 vec3(stride1, stride2, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                   vec2(${r}TexShape[1], ${r}TexShape[0]);
+        return sampleTexture(${r}, uv);
       }
     `:`
-      float ${r}(int row, int col, int depth, int depth2) {
+      float ${s}(int row, int col, int depth, int depth2) {
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
-                vec3(${i}, ${s}, 1));
+                vec3(${o}, ${a}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${h}.0, ${c}.0);
-        return sampleTexture(${a}, uv);
+                   vec2(${h}.0, ${d}.0);
+        return sampleTexture(${r}, uv);
       }
-    `;if(h===s&&p==null)return t?`
-      float ${r}(int row, int col, int depth, int depth2) {
+    `;if(h===a&&l==null)return t?`
+      float ${s}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
-                         vec3(${a}Shape[1] * ${a}Shape[2], ${a}Shape[2], 1));
+                         vec3(${r}Shape[1] * ${r}Shape[2], ${r}Shape[2], 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${a}TexShape[1], ${a}TexShape[0]);
-        return sampleTexture(${a}, uv);
+                  vec2(${r}TexShape[1], ${r}TexShape[0]);
+        return sampleTexture(${r}, uv);
       }
     `:`
-      float ${r}(int row, int col, int depth, int depth2) {
+      float ${s}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
                          vec3(${n[1]*n[2]}, ${n[2]}, 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${h}.0, ${c}.0);
-        return sampleTexture(${a}, uv);
+                  vec2(${h}.0, ${d}.0);
+        return sampleTexture(${r}, uv);
       }
-    `;let y=ko(a);return t?`
-    float ${r}(int row, int col, int depth, int depth2) {
+    `;let b=ki(r);return t?`
+    float ${s}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      ${m}
       ${f}
+      ${m}
       ${g}
       int index = row * stride0 + col * stride1 +
           depth * stride2 + depth2;
-      vec2 uv = uvFromFlat(${a}TexShape[0], ${a}TexShape[1], index + ${y});
-      return sampleTexture(${a}, uv);
+      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index + ${b});
+      return sampleTexture(${r}, uv);
     }
   `:`
-    float ${r}(int row, int col, int depth, int depth2) {
+    float ${s}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${o} + col * ${i} +
-          depth * ${s} + depth2;
-      vec2 uv = uvFromFlat(${c}, ${h}, index + ${y});
-      return sampleTexture(${a}, uv);
+      int index = row * ${i} + col * ${o} +
+          depth * ${a} + depth2;
+      vec2 uv = uvFromFlat(${d}, ${h}, index + ${b});
+      return sampleTexture(${r}, uv);
     }
-  `}function WZ(e){let t=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),r=t[4],s=t[3]*r,i=t[2]*s,o=t[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(t);if(l.length<t.length){let f=Qu(e,l),g=["row","col","depth","depth2","depth3"];return`
-      ${Zu(f)}
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
-        return ${a}(${ep(g,u)});
+  `}function UY(e){let t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),s=t[4],a=t[3]*s,o=t[2]*a,i=t[1]*o,{newShape:c,keptDims:u}=w.squeezeShape(t);if(c.length<t.length){let m=Qu(e,c),g=["row","col","depth","depth2","depth3"];return`
+      ${Zu(m)}
+      float ${r}(int row, int col, int depth, int depth2, int depth3) {
+        return ${r}(${el(g,u)});
       }
     `}if(e.shapeInfo.isUniform)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+      float ${r}(int row, int col, int depth, int depth2, int depth3) {
         float index = dot(
           vec4(row, col, depth, depth2),
-          vec4(${o}, ${i}, ${s}, ${r})) +
+          vec4(${i}, ${o}, ${a}, ${s})) +
           depth3;
         ${Ju(e)}
       }
-    `;let p=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,c=d[0],h=d[1];if(h===o&&p==null)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;let l=e.shapeInfo.flatOffset,p=e.shapeInfo.texShape,d=p[0],h=p[1];if(h===i&&l==null)return`
+      float ${r}(int row, int col, int depth, int depth2, int depth3) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-                         vec4(${i}, ${s}, ${r}, 1));
+                         vec4(${o}, ${a}, ${s}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${h}.0, ${c}.0);
+                   vec2(${h}.0, ${d}.0);
         return sampleTexture(${n}, uv);
       }
-    `;if(h===r&&p==null)return`
-      float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;if(h===s&&l==null)return`
+      float ${r}(int row, int col, int depth, int depth2, int depth3) {
         float texR = dot(
           vec4(row, col, depth, depth2),
           vec4(${t[1]*t[2]*t[3]},
                ${t[2]*t[3]}, ${t[3]}, 1));
         int texC = depth3;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${h}.0, ${c}.0);
+                  vec2(${h}.0, ${d}.0);
         return sampleTexture(${n}, uv);
       }
-    `;let m=ko(n);return`
-    float ${a}(int row, int col, int depth, int depth2, int depth3) {
+    `;let f=ki(n);return`
+    float ${r}(int row, int col, int depth, int depth2, int depth3) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${o} + col * ${i} + depth * ${s} +
-          depth2 * ${r} + depth3 + ${m};
-      vec2 uv = uvFromFlat(${c}, ${h}, index);
+      int index = row * ${i} + col * ${o} + depth * ${a} +
+          depth2 * ${s} + depth3 + ${f};
+      vec2 uv = uvFromFlat(${d}, ${h}, index);
       return sampleTexture(${n}, uv);
     }
-  `}function BZ(e){let t=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),{newShape:r,keptDims:s}=v.squeezeShape(t);if(r.length<t.length){let g=Qu(e,r),y=["row","col","depth","depth2","depth3","depth4"];return`
+  `}function GY(e){let t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),{newShape:s,keptDims:a}=w.squeezeShape(t);if(s.length<t.length){let g=Qu(e,s),b=["row","col","depth","depth2","depth3","depth4"];return`
       ${Zu(g)}
-      float ${a}(int row, int col, int depth,
+      float ${r}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
-        return ${a}(${ep(y,s)});
+        return ${r}(${el(b,a)});
       }
-    `}let i=t[5],o=t[4]*i,l=t[3]*o,u=t[2]*l,p=t[1]*u;if(e.shapeInfo.isUniform)return`
-      float ${a}(int row, int col, int depth,
+    `}let o=t[5],i=t[4]*o,c=t[3]*i,u=t[2]*c,l=t[1]*u;if(e.shapeInfo.isUniform)return`
+      float ${r}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
         int index = round(dot(
           vec4(row, col, depth, depth2),
-          vec4(${p}, ${u}, ${l}, ${o})) +
+          vec4(${l}, ${u}, ${c}, ${i})) +
           dot(
             vec2(depth3, depth4),
-            vec2(${i}, 1)));
+            vec2(${o}, 1)));
         ${Ju(e)}
       }
-    `;let d=e.shapeInfo.flatOffset,c=e.shapeInfo.texShape,h=c[0],m=c[1];if(m===p&&d==null)return`
-      float ${a}(int row, int col, int depth,
+    `;let p=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,h=d[0],f=d[1];if(f===l&&p==null)return`
+      float ${r}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-          vec4(${u}, ${l}, ${o}, ${i})) +
+          vec4(${u}, ${c}, ${i}, ${o})) +
                float(depth4);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${m}.0, ${h}.0);
+                   vec2(${f}.0, ${h}.0);
         return sampleTexture(${n}, uv);
       }
-    `;if(m===i&&d==null)return`
-      float ${a}(int row, int col, int depth,
+    `;if(f===o&&p==null)return`
+      float ${r}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         float texR = dot(vec4(row, col, depth, depth2),
           vec4(${t[1]*t[2]*t[3]*t[4]},
@@ -941,52 +941,52 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
                ${t[4]})) + float(depth3);
         int texC = depth4;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${m}.0, ${h}.0);
+                  vec2(${f}.0, ${h}.0);
         return sampleTexture(${n}, uv);
       }
-    `;let f=ko(n);return`
-    float ${a}(int row, int col, int depth,
+    `;let m=ki(n);return`
+    float ${r}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${p} + col * ${u} + depth * ${l} +
-          depth2 * ${o} + depth3 * ${i} + depth4 + ${f};
-      vec2 uv = uvFromFlat(${h}, ${m}, index);
+      int index = row * ${l} + col * ${u} + depth * ${c} +
+          depth2 * ${i} + depth3 * ${o} + depth4 + ${m};
+      vec2 uv = uvFromFlat(${h}, ${f}, index);
       return sampleTexture(${n}, uv);
     }
-  `}function Ju(e){let t=e.name,n=v.sizeFromShape(e.shapeInfo.logicalShape);return n<2?`return ${t};`:`
+  `}function Ju(e){let t=e.name,n=w.sizeFromShape(e.shapeInfo.logicalShape);return n<2?`return ${t};`:`
     for (int i = 0; i < ${n}; i++) {
       if (i == index) {
         return ${t}[i];
       }
     }
-  `}function VZ(e,t){let n=e.name,a=n.charAt(0).toUpperCase()+n.slice(1),r="get"+a+"AtOutCoords",s=e.shapeInfo.logicalShape.length,i=t.logicalShape.length,o=O_(e.shapeInfo.logicalShape,t.logicalShape),l=gt(i),u=i-s,p,d=["x","y","z","w","u","v"];s===0?p="":i<2&&o.length>=1?p="coords = 0;":p=o.map(g=>`coords.${d[g+u]} = 0;`).join(`
-`);let c="";i<2&&s>0?c="coords":c=e.shapeInfo.logicalShape.map((g,y)=>`coords.${d[y+u]}`).join(", ");let h="return outputValue;",m=v.sizeFromShape(e.shapeInfo.logicalShape)===1,f=v.sizeFromShape(t.logicalShape)===1;if(s===1&&!m&&!f)h=`
+  `}function HY(e,t){let n=e.name,r=n.charAt(0).toUpperCase()+n.slice(1),s="get"+r+"AtOutCoords",a=e.shapeInfo.logicalShape.length,o=t.logicalShape.length,i=z2(e.shapeInfo.logicalShape,t.logicalShape),c=mt(o),u=o-a,l,p=["x","y","z","w","u","v"];a===0?l="":o<2&&i.length>=1?l="coords = 0;":l=i.map(y=>`coords.${p[y+u]} = 0;`).join(`
+`);let d="";o<2&&a>0?d="coords":d=e.shapeInfo.logicalShape.map((y,v)=>`coords.${p[v+u]}`).join(", ");let h="return outputValue;",m=w.sizeFromShape(e.shapeInfo.logicalShape)===1,b=w.sizeFromShape(t.logicalShape)===1;if(a===1&&!m&&!b)h=`
       return vec4(outputValue.xy, outputValue.xy);
-    `;else if(m&&!f)i===1?h=`
+    `;else if(m&&!b)o===1?h=`
         return vec4(outputValue.x, outputValue.x, 0., 0.);
       `:h=`
         return vec4(outputValue.x);
-      `;else if(o.length){let g=s-2,y=s-1;o.indexOf(g)>-1&&o.indexOf(y)>-1?h="return vec4(outputValue.x);":o.indexOf(g)>-1?h="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":o.indexOf(y)>-1&&(h="return vec4(outputValue.xx, outputValue.zz);")}return`
-    vec4 ${r}() {
-      ${l} coords = getOutputCoords();
-      ${p}
-      vec4 outputValue = get${a}(${c});
+      `;else if(i.length){let y=a-2,v=a-1;i.indexOf(y)>-1&&i.indexOf(v)>-1?h="return vec4(outputValue.x);":i.indexOf(y)>-1?h="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":i.indexOf(v)>-1&&(h="return vec4(outputValue.xx, outputValue.zz);")}return`
+    vec4 ${s}() {
+      ${c} coords = getOutputCoords();
+      ${l}
+      vec4 outputValue = get${r}(${d});
       ${h}
     }
-  `}function UZ(e,t){let n=e.name,a=n.charAt(0).toUpperCase()+n.slice(1),r="get"+a+"AtOutCoords",s=t.texShape,i=e.shapeInfo.texShape,o=e.shapeInfo.logicalShape.length,l=t.logicalShape.length;if(!e.shapeInfo.isUniform&&o===l&&e.shapeInfo.flatOffset==null&&v.arraysEqual(i,s))return`
-      float ${r}() {
+  `}function qY(e,t){let n=e.name,r=n.charAt(0).toUpperCase()+n.slice(1),s="get"+r+"AtOutCoords",a=t.texShape,o=e.shapeInfo.texShape,i=e.shapeInfo.logicalShape.length,c=t.logicalShape.length;if(!e.shapeInfo.isUniform&&i===c&&e.shapeInfo.flatOffset==null&&w.arraysEqual(o,a))return`
+      float ${s}() {
         return sampleTexture(${n}, resultUV);
       }
-    `;let u=gt(l),p=O_(e.shapeInfo.logicalShape,t.logicalShape),d=l-o,c,h=["x","y","z","w","u","v"];o===0?c="":l<2&&p.length>=1?c="coords = 0;":c=p.map(f=>`coords.${h[f+d]} = 0;`).join(`
-`);let m="";return l<2&&o>0?m="coords":m=e.shapeInfo.logicalShape.map((f,g)=>`coords.${h[g+d]}`).join(", "),`
-    float ${r}() {
+    `;let u=mt(c),l=z2(e.shapeInfo.logicalShape,t.logicalShape),p=c-i,d,h=["x","y","z","w","u","v"];i===0?d="":c<2&&l.length>=1?d="coords = 0;":d=l.map(m=>`coords.${h[m+p]} = 0;`).join(`
+`);let f="";return c<2&&i>0?f="coords":f=e.shapeInfo.logicalShape.map((m,g)=>`coords.${h[g+p]}`).join(", "),`
+    float ${s}() {
       ${u} coords = getOutputCoords();
-      ${c}
-      return get${a}(${m});
+      ${d}
+      return get${r}(${f});
     }
-  `}function gt(e){if(e<=1)return"int";if(e===2)return"ivec2";if(e===3)return"ivec3";if(e===4)return"ivec4";if(e===5)return"ivec5";if(e===6)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function Q0(e,t,n){let{newShape:a,keptDims:r}=v.squeezeShape(t),s=t.length,i=e&&s===3&&t[0]===1,o=i?t.slice(1):a,l=!e&&s>1&&!v.arraysEqual(t,n)&&a.length<s||i;return{useSqueezeShape:l,uniformShape:l?o:t,keptDims:r}}function Qu(e,t){let n=JSON.parse(JSON.stringify(e));return n.shapeInfo.logicalShape=t,n}function ep(e,t){return t.map(n=>e[n]).join(", ")}function GZ(e,t,n,a){let r=n.map((p,d)=>{let c={logicalShape:p.shape,texShape:p.isUniform?null:p.texData.texShape,isUniform:p.isUniform,isPacked:p.isUniform?!1:p.texData.isPacked,flatOffset:null};return p.texData!=null&&p.texData.slice!=null&&p.texData.slice.flatOffset>0&&(c.flatOffset=p.texData.slice.flatOffset),{name:t.variableNames[d],shapeInfo:c}}),s=r.map(p=>p.shapeInfo),i={logicalShape:a.shape,texShape:a.texData.texShape,isUniform:!1,isPacked:a.texData.isPacked,flatOffset:null},o=lZ(r,i,t),l=f_(e.gl,o),u=e.createProgram(l);return H().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:Object.assign({program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i},W_(e,t,u))}function W_(e,t,n){let a={},r={},s={},i=[],o,l,u,p=null,d=null;d=e.getUniformLocation(n,"NAN",!1),H().getNumber("WEBGL_VERSION")===1&&(p=e.getUniformLocation(n,"INFINITY",!1));let c=!1;for(let h=0;h<t.variableNames.length;h++){let m=t.variableNames[h];a[m]=e.getUniformLocation(n,m,c),a[`offset${m}`]=e.getUniformLocation(n,`offset${m}`,c),t.enableShapeUniforms&&(r[`${m}Shape`]=e.getUniformLocation(n,`${m}Shape`,c),s[`${m}TexShape`]=e.getUniformLocation(n,`${m}TexShape`,c))}return t.enableShapeUniforms&&(o=e.getUniformLocation(n,"outShape",c),u=e.getUniformLocation(n,"outShapeStrides",c),l=e.getUniformLocation(n,"outTexShape",c)),t.customUniforms&&t.customUniforms.forEach((h,m)=>{i[m]=e.getUniformLocation(n,h.name,c)}),{uniformLocations:a,customUniformLocations:i,infLoc:p,nanLoc:d,inShapesLocations:r,inTexShapesLocations:s,outShapeLocation:o,outShapeStridesLocation:u,outTexShapeLocation:l}}function dI(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((n,a)=>{let r=n.logicalShape,s=t[a],i=s.shape;if(!v.arraysEqual(r,i))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${r} and ${i} must match`);if(n.isUniform&&s.isUniform)return;let o=n.texShape,l=s.isUniform?null:s.texData.texShape;if(!v.arraysEqual(o,l))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${o} and ${l} must match`)})}function HZ(e,t,n,a,r){t.program.enableShapeUniforms||(dI(t.inShapeInfos,n),dI([t.outShapeInfo],[a]));let s=a.texData.texture,i=a.texData.texShape;a.texData.isPacked?e.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):e.setOutputMatrixTexture(s.texture,i[0],i[1]),e.setProgram(t.webGLProgram),H().getNumber("WEBGL_VERSION")===1&&t.infLoc!==null&&e.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&e.gl.uniform1f(t.nanLoc,NaN),n.forEach((l,u)=>{let p=t.program.variableNames[u],d=t.uniformLocations[p],c=t.uniformLocations[`offset${p}`],h=t.inShapesLocations[`${p}Shape`],m=t.inTexShapesLocations[`${p}TexShape`];if(h){let{uniformShape:f}=Q0(t.program.packedInputs,l.shape,l.texData.texShape);switch(f.length){case 1:e.gl.uniform1iv(h,new Int32Array(f));break;case 2:e.gl.uniform2iv(h,new Int32Array(f));break;case 3:e.gl.uniform3iv(h,new Int32Array(f));break;case 4:e.gl.uniform4iv(h,new Int32Array(f));break;default:break}}if(m&&e.gl.uniform2i(m,l.texData.texShape[0],l.texData.texShape[1]),d!=null){if(l.isUniform){if(v.sizeFromShape(l.shape)<2)e.gl.uniform1f(d,l.uniformValues[0]);else{let f=l.uniformValues;f instanceof Float32Array||(f=new Float32Array(f)),e.gl.uniform1fv(d,f)}return}l.texData.slice!=null&&c!=null&&e.gl.uniform1i(c,l.texData.slice.flatOffset),e.setInputMatrixTexture(l.texData.texture.texture,d,u)}});let o=t.outShapeLocation;if(o)switch(a.shape.length){case 1:e.gl.uniform1iv(o,new Int32Array(a.shape));break;case 2:e.gl.uniform2iv(o,new Int32Array(a.shape));break;case 3:e.gl.uniform3iv(o,new Int32Array(a.shape));break;case 4:e.gl.uniform4iv(o,new Int32Array(a.shape));break;default:break}if(t.outShapeStridesLocation){let l=v.computeStrides(a.shape);switch(a.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(l));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(l));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(l));break;default:break}}t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,a.texData.texShape[0],a.texData.texShape[1]),t.program.customUniforms&&r&&t.program.customUniforms.forEach((l,u)=>{let p=t.customUniformLocations[u],d=r[u];if(l.type==="float")e.gl.uniform1fv(p,d);else if(l.type==="vec2")e.gl.uniform2fv(p,d);else if(l.type==="vec3")e.gl.uniform3fv(p,d);else if(l.type==="vec4")e.gl.uniform4fv(p,d);else if(l.type==="int")e.gl.uniform1iv(p,d);else if(l.type==="ivec2")e.gl.uniform2iv(p,d);else if(l.type==="ivec3")e.gl.uniform3iv(p,d);else if(l.type==="ivec4")e.gl.uniform4iv(p,d);else throw Error(`uniform type ${l.type} is not supported yet.`)}),e.executeProgram()}function jZ(e,t,n){let a="";t.concat(n).forEach(i=>{let o=i.texData!=null&&i.texData.slice!=null&&i.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!i.isUniform){let l=i.texData.texShape,{useSqueezeShape:u,uniformShape:p,keptDims:d}=Q0(e.packedInputs,i.shape,l),c="",h="",m="";if(p.length===1&&e.packedInputs){let I=[Math.ceil(l[0]/2),Math.ceil(l[1]/2)];c=`${I[0]>1}_${I[1]>1}`}else if(p.length===2&&!e.packedInputs)h=`${p[0]>1}_${p[1]>1}`;else if(p.length>2&&!e.packedInputs){let I=v.computeStrides(p);m=`${I[0]===l[1]}_${I[I.length-1]===l[1]}`}let f=i.shape.length,g=p.length===2&&v.arraysEqual(i.shape,l),y=v.sizeFromShape(i.shape)===1,b=N.getBroadcastDims(i.shape,n.shape),x=!e.packedInputs&&f===n.shape.length&&v.arraysEqual(l,n.texData.texShape),w=e.packedInputs||p.length>2?"":`${l[0]>1}_${l[1]>1}`;a+=`${f}_${x}_${u?d:""}_${p.length}_${y}_${b}_${g}_${c}_${h}_${m}_${w}_${o}`}else{let l=i.isUniform?"uniform":i.texData.texShape;a+=`${i.shape}_${l}_${o}`}});let r=e.userCode,s=e.constructor.name;return s+="_"+a+"_"+r+`${H().getNumber("WEBGL_VERSION")}`,s}function En(e){return H().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var qZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=rc.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=_n();this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length),this.userCode=`
+  `}function mt(e){if(e<=1)return"int";if(e===2)return"ivec2";if(e===3)return"ivec3";if(e===4)return"ivec4";if(e===5)return"ivec5";if(e===6)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function Q0(e,t,n){let{newShape:r,keptDims:s}=w.squeezeShape(t),a=t.length,o=e&&a===3&&t[0]===1,i=o?t.slice(1):r,c=!e&&a>1&&!w.arraysEqual(t,n)&&r.length<a||o;return{useSqueezeShape:c,uniformShape:c?i:t,keptDims:s}}function Qu(e,t){let n=JSON.parse(JSON.stringify(e));return n.shapeInfo.logicalShape=t,n}function el(e,t){return t.map(n=>e[n]).join(", ")}function jY(e,t,n,r){let s=n.map((l,p)=>{let d={logicalShape:l.shape,texShape:l.isUniform?null:l.texData.texShape,isUniform:l.isUniform,isPacked:l.isUniform?!1:l.texData.isPacked,flatOffset:null};return l.texData!=null&&l.texData.slice!=null&&l.texData.slice.flatOffset>0&&(d.flatOffset=l.texData.slice.flatOffset),{name:t.variableNames[p],shapeInfo:d}}),a=s.map(l=>l.shapeInfo),o={logicalShape:r.shape,texShape:r.texData.texShape,isUniform:!1,isPacked:r.texData.isPacked,flatOffset:null},i=dY(s,o,t),c=b2(e.gl,i),u=e.createProgram(c);return q().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:c,source:i,webGLProgram:u,inShapeInfos:a,outShapeInfo:o,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:Object.assign({program:t,fragmentShader:c,source:i,webGLProgram:u,inShapeInfos:a,outShapeInfo:o},V2(e,t,u))}function V2(e,t,n){let r={},s={},a={},o=[],i,c,u,l=null,p=null;p=e.getUniformLocation(n,"NAN",!1),q().getNumber("WEBGL_VERSION")===1&&(l=e.getUniformLocation(n,"INFINITY",!1));let d=!1;for(let h=0;h<t.variableNames.length;h++){let f=t.variableNames[h];r[f]=e.getUniformLocation(n,f,d),r[`offset${f}`]=e.getUniformLocation(n,`offset${f}`,d),t.enableShapeUniforms&&(s[`${f}Shape`]=e.getUniformLocation(n,`${f}Shape`,d),a[`${f}TexShape`]=e.getUniformLocation(n,`${f}TexShape`,d))}return t.enableShapeUniforms&&(i=e.getUniformLocation(n,"outShape",d),u=e.getUniformLocation(n,"outShapeStrides",d),c=e.getUniformLocation(n,"outTexShape",d)),t.customUniforms&&t.customUniforms.forEach((h,f)=>{o[f]=e.getUniformLocation(n,h.name,d)}),{uniformLocations:r,customUniformLocations:o,infLoc:l,nanLoc:p,inShapesLocations:s,inTexShapesLocations:a,outShapeLocation:i,outShapeStridesLocation:u,outTexShapeLocation:c}}function h1(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((n,r)=>{let s=n.logicalShape,a=t[r],o=a.shape;if(!w.arraysEqual(s,o))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${s} and ${o} must match`);if(n.isUniform&&a.isUniform)return;let i=n.texShape,c=a.isUniform?null:a.texData.texShape;if(!w.arraysEqual(i,c))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${i} and ${c} must match`)})}function KY(e,t,n,r,s){t.program.enableShapeUniforms||(h1(t.inShapeInfos,n),h1([t.outShapeInfo],[r]));let a=r.texData.texture,o=r.texData.texShape;r.texData.isPacked?e.setOutputPackedMatrixTexture(a.texture,o[0],o[1]):e.setOutputMatrixTexture(a.texture,o[0],o[1]),e.setProgram(t.webGLProgram),q().getNumber("WEBGL_VERSION")===1&&t.infLoc!==null&&e.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&e.gl.uniform1f(t.nanLoc,NaN),n.forEach((c,u)=>{let l=t.program.variableNames[u],p=t.uniformLocations[l],d=t.uniformLocations[`offset${l}`],h=t.inShapesLocations[`${l}Shape`],f=t.inTexShapesLocations[`${l}TexShape`];if(h){let{uniformShape:m}=Q0(t.program.packedInputs,c.shape,c.texData.texShape);switch(m.length){case 1:e.gl.uniform1iv(h,new Int32Array(m));break;case 2:e.gl.uniform2iv(h,new Int32Array(m));break;case 3:e.gl.uniform3iv(h,new Int32Array(m));break;case 4:e.gl.uniform4iv(h,new Int32Array(m));break;default:break}}if(f&&e.gl.uniform2i(f,c.texData.texShape[0],c.texData.texShape[1]),p!=null){if(c.isUniform){if(w.sizeFromShape(c.shape)<2)e.gl.uniform1f(p,c.uniformValues[0]);else{let m=c.uniformValues;m instanceof Float32Array||(m=new Float32Array(m)),e.gl.uniform1fv(p,m)}return}c.texData.slice!=null&&d!=null&&e.gl.uniform1i(d,c.texData.slice.flatOffset),e.setInputMatrixTexture(c.texData.texture.texture,p,u)}});let i=t.outShapeLocation;if(i)switch(r.shape.length){case 1:e.gl.uniform1iv(i,new Int32Array(r.shape));break;case 2:e.gl.uniform2iv(i,new Int32Array(r.shape));break;case 3:e.gl.uniform3iv(i,new Int32Array(r.shape));break;case 4:e.gl.uniform4iv(i,new Int32Array(r.shape));break;default:break}if(t.outShapeStridesLocation){let c=w.computeStrides(r.shape);switch(r.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(c));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(c));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(c));break;default:break}}t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,r.texData.texShape[0],r.texData.texShape[1]),t.program.customUniforms&&s&&t.program.customUniforms.forEach((c,u)=>{let l=t.customUniformLocations[u],p=s[u];if(c.type==="float")e.gl.uniform1fv(l,p);else if(c.type==="vec2")e.gl.uniform2fv(l,p);else if(c.type==="vec3")e.gl.uniform3fv(l,p);else if(c.type==="vec4")e.gl.uniform4fv(l,p);else if(c.type==="int")e.gl.uniform1iv(l,p);else if(c.type==="ivec2")e.gl.uniform2iv(l,p);else if(c.type==="ivec3")e.gl.uniform3iv(l,p);else if(c.type==="ivec4")e.gl.uniform4iv(l,p);else throw Error(`uniform type ${c.type} is not supported yet.`)}),e.executeProgram()}function XY(e,t,n){let r="";t.concat(n).forEach(o=>{let i=o.texData!=null&&o.texData.slice!=null&&o.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!o.isUniform){let c=o.texData.texShape,{useSqueezeShape:u,uniformShape:l,keptDims:p}=Q0(e.packedInputs,o.shape,c),d="",h="",f="";if(l.length===1&&e.packedInputs){let k=[Math.ceil(c[0]/2),Math.ceil(c[1]/2)];d=`${k[0]>1}_${k[1]>1}`}else if(l.length===2&&!e.packedInputs)h=`${l[0]>1}_${l[1]>1}`;else if(l.length>2&&!e.packedInputs){let k=w.computeStrides(l);f=`${k[0]===c[1]}_${k[k.length-1]===c[1]}`}let m=o.shape.length,g=l.length===2&&w.arraysEqual(o.shape,c),b=w.sizeFromShape(o.shape)===1,y=N.getBroadcastDims(o.shape,n.shape),v=!e.packedInputs&&m===n.shape.length&&w.arraysEqual(c,n.texData.texShape),x=e.packedInputs||l.length>2?"":`${c[0]>1}_${c[1]>1}`;r+=`${m}_${v}_${u?p:""}_${l.length}_${b}_${y}_${g}_${d}_${h}_${f}_${x}_${i}`}else{let c=o.isUniform?"uniform":o.texData.texShape;r+=`${o.shape}_${c}_${i}`}});let s=e.userCode,a=e.constructor.name;return a+="_"+r+"_"+s+`${q().getNumber("WEBGL_VERSION")}`,a}function Dn(e){return q().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var YY=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=sd.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=$n();this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length),this.userCode=`
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Kf(["r","c","d"],e):wo(["r","c","d"],e)}
+        ${this.enableShapeUniforms?Xm(["r","c","d"],e):Ii(["r","c","d"],e)}
         return ivec3(r, c, d);
       }
 
@@ -1004,9 +1004,9 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         ${t.output} = result;
       }
-    `}},KZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=rc.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=_n();this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length),this.userCode=`
+    `}},ZY=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=sd.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=$n();this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length),this.userCode=`
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Kf(["r","c","d"],e):wo(["r","c","d"],e)}
+        ${this.enableShapeUniforms?Xm(["r","c","d"],e):Ii(["r","c","d"],e)}
         return ivec3(r, c, d);
       }
 
@@ -1024,24 +1024,24 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         ${t.output} = result;
       }
-    `}},XZ=class{constructor(e){this.variableNames=["A"],this.outTexUsage=ca.DOWNLOAD;let t=_n();this.outputShape=e,this.userCode=`
-      ${P_}
+    `}},JY=class{constructor(e){this.variableNames=["A"],this.outTexUsage=ur.DOWNLOAD;let t=$n();this.outputShape=e,this.userCode=`
+      ${L2}
 
       void main() {
         float x = getAAtOutCoords();
         ${t.output} = encode_float(x);
       }
-    `}},YZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=ca.DOWNLOAD;let t=_n();this.outputShape=e,this.userCode=`
-      ${P_}
+    `}},QY=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=ur.DOWNLOAD;let t=$n();this.outputShape=e,this.userCode=`
+      ${L2}
 
       void main() {
         ivec3 coords = getOutputCoords();
         float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));
         ${t.output} = encode_float(x);
       }
-    `}},ZZ={R:0,G:1,B:2,A:3},hI=class{constructor(e,t=!1,n="RGBA"){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let a=_n();this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length);let r="result";t&&(r="floor(result * 255. + 0.5)");let s="";for(let i=0;i<n.length;i++){let o=n[i];s+=`
-          if(offset == ${i}) {
-            result = values[${ZZ[o]}];
+    `}},e9={R:0,G:1,B:2,A:3},f1=class{constructor(e,t=!1,n="RGBA"){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let r=$n();this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length);let s="result";t&&(s="floor(result * 255. + 0.5)");let a="";for(let o=0;o<n.length;o++){let i=n[o];a+=`
+          if(offset == ${o}) {
+            result = values[${e9[i]}];
           }`}this.userCode=`
       ${this.enableShapeUniforms?J0():Z0(e)}
 
@@ -1057,17 +1057,17 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         if (r < texShape[0]) {
           int c = imod(flatIndex, texShape[1]);
           vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
-          vec4 values = ${a.texture2D}(A, uv);
-          ${s}
+          vec4 values = ${r.texture2D}(A, uv);
+          ${a}
         }
-        ${a.output} = vec4(${r}, 0., 0., 0.);
+        ${r.output} = vec4(${s}, 0., 0., 0.);
       }
-    `}},JZ=class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let n=_n();this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length);let a="",r="result";t&&(r="floor(result * 255. + 0.5)");for(let s=0;s<=1;s++)for(let i=0;i<=1;i++){let o=s*2+i;a+=`
+    `}},t9=class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let n=$n();this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length);let r="",s="result";t&&(s="floor(result * 255. + 0.5)");for(let a=0;a<=1;a++)for(let o=0;o<=1;o++){let i=a*2+o;r+=`
           localCoords = coords;
-          if(localCoords[2] + ${i} < ${this.enableShapeUniforms?"outShape[2]":`${e[2]}`}) {
-          localCoords[2] += ${i};
-          if (localCoords[1] + ${s} < ${this.enableShapeUniforms?"outShape[1]":`${e[1]}`}) {
-            localCoords[1] += ${s};
+          if(localCoords[2] + ${o} < ${this.enableShapeUniforms?"outShape[2]":`${e[2]}`}) {
+          localCoords[2] += ${o};
+          if (localCoords[1] + ${a} < ${this.enableShapeUniforms?"outShape[1]":`${e[1]}`}) {
+            localCoords[1] += ${a};
 
             flatIndex = getFlatIndex(localCoords);
             offset = imod(flatIndex, 4);
@@ -1080,13 +1080,13 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             values = ${n.texture2D}(A, uv);
 
             if (offset == 0) {
-              result[${o}] = values[0];
+              result[${i}] = values[0];
             } else if (offset == 1) {
-              result[${o}] = values[1];
+              result[${i}] = values[1];
             } else if (offset == 2) {
-              result[${o}] = values[2];
+              result[${i}] = values[2];
             } else {
-              result[${o}] = values[3];
+              result[${i}] = values[3];
             }
           }
         }
@@ -1102,11 +1102,11 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           vec2 uv;
           vec4 values;
 
-          ${a}
+          ${r}
 
-          ${n.output} = ${r};
+          ${n.output} = ${s};
         }
-    `}},B_={};Ae(B_,{bindVertexProgramAttributeStreams:()=>Y_,createBufferFromOutputTexture:()=>Q_,createFloat16MatrixTexture:()=>j_,createFloat16PackedMatrixTexture:()=>X_,createFloat32MatrixTexture:()=>H_,createIndexBuffer:()=>G_,createPackedMatrixTexture:()=>K_,createUnsignedBytesMatrixTexture:()=>q_,createVertexBuffer:()=>U_,createVertexShader:()=>V_,downloadByteEncodedFloatMatrixFromOutputTexture:()=>tE,downloadFloat32MatrixFromBuffer:()=>eE,downloadMatrixFromPackedOutputTexture:()=>aE,downloadPackedMatrixFromBuffer:()=>nE,getInternalFormatForFloat16MatrixTexture:()=>t1,getInternalFormatForFloat16PackedMatrixTexture:()=>r1,getInternalFormatForFloat32MatrixTexture:()=>e1,getInternalFormatForPackedMatrixTexture:()=>a1,getInternalFormatForUnsignedBytesMatrixTexture:()=>n1,uploadDenseMatrixToTexture:()=>Z_,uploadPixelDataToTexture:()=>J_});function V_(e){let t=_n(),n=`${t.version}
+    `}},U2={};Ae(U2,{bindVertexProgramAttributeStreams:()=>J2,createBufferFromOutputTexture:()=>tE,createFloat16MatrixTexture:()=>K2,createFloat16PackedMatrixTexture:()=>Z2,createFloat32MatrixTexture:()=>j2,createIndexBuffer:()=>q2,createPackedMatrixTexture:()=>Y2,createUnsignedBytesMatrixTexture:()=>X2,createVertexBuffer:()=>H2,createVertexShader:()=>G2,downloadByteEncodedFloatMatrixFromOutputTexture:()=>rE,downloadFloat32MatrixFromBuffer:()=>nE,downloadMatrixFromPackedOutputTexture:()=>aE,downloadPackedMatrixFromBuffer:()=>sE,getInternalFormatForFloat16MatrixTexture:()=>tI,getInternalFormatForFloat16PackedMatrixTexture:()=>sI,getInternalFormatForFloat32MatrixTexture:()=>eI,getInternalFormatForPackedMatrixTexture:()=>rI,getInternalFormatForUnsignedBytesMatrixTexture:()=>nI,uploadDenseMatrixToTexture:()=>Q2,uploadPixelDataToTexture:()=>eE});function G2(e){let t=$n(),n=`${t.version}
     precision highp float;
     ${t.attribute} vec3 clipSpacePos;
     ${t.attribute} vec2 uv;
@@ -1115,46 +1115,46 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
     void main() {
       gl_Position = vec4(clipSpacePos, 1);
       resultUV = uv;
-    }`;return m_(e,n)}function U_(e){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return b_(e,t)}function G_(e){let t=new Uint16Array([0,1,2,2,1,3]);return x_(e,t)}function od(e,t,n,a,r,s){w_(t,n);let i=v_(e),o=e.TEXTURE_2D;return me(e,()=>e.bindTexture(o,i)),me(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(o,e.TEXTURE_MIN_FILTER,e.NEAREST)),me(e,()=>e.texParameteri(o,e.TEXTURE_MAG_FILTER,e.NEAREST)),H().getNumber("WEBGL_VERSION")===1?me(e,()=>e.texImage2D(o,0,a,t,n,0,r,s,null)):me(e,()=>e.texStorage2D(o,1,a,t,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:i,texShape:[n,t]}}function e1(e){return e.internalFormatFloat}function H_(e,t,n,a){let[r,s]=id(t,n);return od(e,r,s,e1(a),a.textureFormatFloat,e.FLOAT)}function t1(e){return e.internalFormatHalfFloat}function j_(e,t,n,a){let[r,s]=id(t,n);return od(e,r,s,t1(a),a.textureFormatFloat,a.textureTypeHalfFloat)}function n1(e){return e.downloadTextureFormat}function q_(e,t,n,a){let[r,s]=id(t,n);return od(e,r,s,n1(a),e.RGBA,e.UNSIGNED_BYTE)}function a1(e){return e.internalFormatPackedFloat}function K_(e,t,n,a){let[r,s]=Xu(t,n);return od(e,r,s,a1(a),e.RGBA,e.FLOAT)}function r1(e){return e.internalFormatPackedHalfFloat}function X_(e,t,n,a){let[r,s]=Xu(t,n);return od(e,r,s,r1(a),e.RGBA,a.textureTypeHalfFloat)}function Y_(e,t,n){return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),mx(e,t,"clipSpacePos",n,3,20,0)&&mx(e,t,"uv",n,2,20,12)}function Z_(e,t,n,a,r,s){me(e,()=>e.bindTexture(e.TEXTURE_2D,t));let i,o,l;r instanceof Uint8Array?(i=new Uint8Array(n*a*4),o=e.UNSIGNED_BYTE,l=e.RGBA):(i=new Float32Array(n*a*4),o=e.FLOAT,l=s.internalFormatPackedFloat),i.set(r),H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n,a,e.RGBA,o,i)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,l,n,a,0,e.RGBA,o,i)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function J_(e,t,n){me(e,()=>e.bindTexture(e.TEXTURE_2D,t)),n.data instanceof Uint8Array?H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n.width,n.height,e.RGBA,e.UNSIGNED_BYTE,n.data)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,n.width,n.height,0,e.RGBA,e.UNSIGNED_BYTE,n.data)):H().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,n)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function Q_(e,t,n,a){let r=e.createBuffer();me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,r));let s=4*4*t*n;return me(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,s,e.STREAM_READ)),me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,0)),me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),r}function eE(e,t,n){let a=e,r=new Float32Array(n);return a.bindBuffer(a.PIXEL_PACK_BUFFER,t),a.getBufferSubData(a.PIXEL_PACK_BUFFER,0,r),a.bindBuffer(a.PIXEL_PACK_BUFFER,null),r}function tE(e,t,n,a){let[r,s]=id(t,n),i=4,o=new Uint8Array(KY(t*n,i));return me(e,()=>e.readPixels(0,0,r,s,a.downloadTextureFormat,e.UNSIGNED_BYTE,o)),new Float32Array(o.buffer)}function nE(e,t,n,a,r,s,i,o){let l=e,u=new Float32Array(XY(s,i));return l.bindBuffer(l.PIXEL_PACK_BUFFER,t),l.getBufferSubData(l.PIXEL_PACK_BUFFER,0,u),l.bindBuffer(l.PIXEL_PACK_BUFFER,null),u}function aE(e,t,n){let a=new Float32Array(t*n*4);return me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,a)),a}var Th=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let t=H().getNumber("WEBGL_VERSION");e!=null?(this.gl=e,c_(t,e)):this.gl=qa(t);let n="WEBGL_color_buffer_float",a="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),H().getNumber("WEBGL_VERSION")===1){let r="OES_texture_float",s="OES_texture_half_float";if(this.textureFloatExtension=zp(this.gl,r),da(this.gl,s))this.textureHalfFloatExtension=zp(this.gl,s);else if(H().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),da(this.gl,a))this.colorBufferHalfFloatExtension=zp(this.gl,a);else if(H().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",da(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else if(da(this.gl,a))this.colorBufferHalfFloatExtension=this.gl.getExtension(a);else throw new Error("GL context does not support color renderable floats");this.vertexBuffer=U_(this.gl),this.indexBuffer=G_(this.gl),this.framebuffer=k_(this.gl),this.textureConfig=X0(this.gl,this.textureHalfFloatExtension)}get debug(){return H().getBool("DEBUG")}dispose(){if(this.disposed)return;this.program!=null&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),this.outputTexture!=null&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let e=this.gl;me(e,()=>e.finish()),me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.deleteFramebuffer(this.framebuffer)),me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),me(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),H_(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),j_(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),q_(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),J_(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,n,a){this.throwIfDisposed(),Z_(this.gl,e,t,n,a,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),X_(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),K_(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(fx(this.gl,this.framebuffer),this.outputTexture=null),me(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,n){return this.downloadMatrixDriver(e,()=>tE(this.gl,t,n,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,n,a,r,s){return nE(this.gl,e,t,n,a,r,s,this.textureConfig)}downloadFloat32MatrixFromBuffer(e,t){return eE(this.gl,e,t)}createBufferFromTexture(e,t,n){this.bindTextureToFrameBuffer(e);let a=Q_(this.gl,t,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),a}createAndWaitForFence(){let e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,n;if(H().getBool("WEBGL_FENCE_API_ENABLED")){let a=e,r=a.fenceSync(a.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),n=()=>{let s=a.clientWaitSync(r,0,0);return s===a.ALREADY_SIGNALED||s===a.CONDITION_SATISFIED},t=r}else H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(t,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:t,isFencePassed:n}}downloadMatrixFromPackedTexture(e,t,n){return this.downloadMatrixDriver(e,()=>aE(this.gl,t,n))}createProgram(e){this.throwIfDisposed();let t=this.gl;this.vertexShader==null&&(this.vertexShader=V_(t));let n=g_(t);return me(t,()=>t.attachShader(n,this.vertexShader)),me(t,()=>t.attachShader(n,e)),y_(t,n),this.debug&&wh(t,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=Y_(t,this.program,this.vertexBuffer)),n}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),e!=null&&me(this.gl,()=>this.gl.deleteProgram(e))}setProgram(e){this.throwIfDisposed(),this.program=e,this.program!=null&&this.debug&&wh(this.gl,this.program),me(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,n=!0){return this.throwIfDisposed(),n?S_(this.gl,e,t):T_(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),me(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,n){this.throwIfDisposed(),this.throwIfNoProgram(),N_(this.gl,e,t,n)}setOutputMatrixTexture(e,t,n){this.setOutputMatrixTextureDriver(e,n,t)}setOutputPackedMatrixTexture(e,t,n){this.throwIfDisposed();let[a,r]=Xu(t,n);this.setOutputMatrixTextureDriver(e,a,r)}setOutputMatrixWriteRegion(e,t,n,a){this.setOutputMatrixWriteRegionDriver(n,e,a,t)}setOutputPackedMatrixWriteRegion(e,t,n,a){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){this.program!=null&&wh(this.gl,this.program),Wp(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let e=this.gl;this.debug&&this.debugValidate(),me(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),me(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=zp(this.gl,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let n=this.gl,a=this.getQueryTimerExtensionWebGL2(),r=n.createQuery();return n.beginQuery(a.TIME_ELAPSED_EXT,r),r}let e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let t=this.gl,n=this.getQueryTimerExtensionWebGL2();t.endQuery(n.TIME_ELAPSED_EXT);return}let e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await v.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(t===0)return null;if(t===2){let n=this.gl;return n.getQueryParameter(e,n.QUERY_RESULT)/1e6}else{let n=this.getQueryTimerExtensionWebGL1();return n.getQueryObjectEXT(e,n.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(t===0)return!0;if(t===2){let n=this.gl,a=this.getQueryTimerExtensionWebGL2(),r=n.getQueryParameter(e,n.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(a.GPU_DISJOINT_EXT)),r&&!this.disjoint}else{let n=this.getQueryTimerExtensionWebGL1(),a=n.getQueryObjectEXT(e,n.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),a&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){let e=QZ(this.itemsToPoll.map(t=>t.isDoneFn));for(let t=0;t<=e;++t){let{resolveFn:n}=this.itemsToPoll[t];n()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){if(this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),this.itemsToPoll.length>1)return;let n;"setTimeoutCustom"in H().platform&&(n=H().platform.setTimeoutCustom.bind(H().platform)),v.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0),()=>0,null,n)}bindTextureToFrameBuffer(e){this.throwIfDisposed(),kh(this.gl,e,this.framebuffer),this.debug&&Wp(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(kh(this.gl,this.outputTexture,this.framebuffer),this.debug&&Wp(this.gl)):fx(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);let n=t();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(e,t,n){this.throwIfDisposed();let a=this.gl;kh(a,e,this.framebuffer),this.debug&&Wp(a),this.outputTexture=e,me(a,()=>a.viewport(0,0,t,n)),me(a,()=>a.scissor(0,0,t,n))}setOutputMatrixWriteRegionDriver(e,t,n,a){this.throwIfDisposed(),me(this.gl,()=>this.gl.scissor(e,t,n,a))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(this.program==null)throw new Error("No GPU program is currently set.")}};function QZ(e){let t=0;for(;t<e.length&&e[t]();++t);return t-1}var{addImpl:e7,bincountImpl:rE,bincountReduceImpl:t7,castImpl:n7,ceilImpl:a7,concatImpl:r7,equalImpl:s7,expImpl:i7,expm1Impl:o7,floorImpl:l7,gatherNdImpl:u7,gatherV2Impl:p7,greaterImpl:c7,greaterEqualImpl:d7,lessImpl:h7,lessEqualImpl:m7,linSpaceImpl:f7,logImpl:g7,maxImpl:y7,maximumImpl:b7,minimumImpl:x7,multiplyImpl:v7,negImpl:w7,notEqualImpl:k7,prodImpl:I7,raggedGatherImpl:S7,raggedRangeImpl:T7,raggedTensorToTensorImpl:N7,rangeImpl:C7,rsqrtImpl:_7,scatterImpl:E7,sigmoidImpl:A7,simpleAbsImpl:sE,sliceImpl:$7,sparseFillEmptyRowsImpl:F7,sparseReshapeImpl:D7,sparseSegmentReductionImpl:iE,sqrtImpl:R7,stridedSliceImpl:M7,stringNGramsImpl:P7,stringSplitImpl:O7,stringToHashBucketFastImpl:L7,subImpl:z7,tileImpl:W7,topKImpl:B7,transposeImpl:s1,uniqueImpl:V7}=lC;function oE(e,t){return["x","y","z","w","u","v"].slice(0,t).map(n=>`${e}.${n}`)}function kn(e,t){return t===1?[e]:oE(e,t)}function U7(e,t){if(e===1)return"rc";let n="";for(let a=0;a<e;a++)n+=t[a],a<e-1&&(n+=",");return n}var G7=class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=En(this.outputShape.length),this.rank===0)this.userCode=`
+    }`;return g2(e,n)}function H2(e){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return x2(e,t)}function q2(e){let t=new Uint16Array([0,1,2,2,1,3]);return w2(e,t)}function ip(e,t,n,r,s,a){k2(t,n);let o=I2(e),i=e.TEXTURE_2D;return me(e,()=>e.bindTexture(i,o)),me(e,()=>e.texParameteri(i,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(i,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),me(e,()=>e.texParameteri(i,e.TEXTURE_MIN_FILTER,e.NEAREST)),me(e,()=>e.texParameteri(i,e.TEXTURE_MAG_FILTER,e.NEAREST)),q().getNumber("WEBGL_VERSION")===1?me(e,()=>e.texImage2D(i,0,r,t,n,0,s,a,null)):me(e,()=>e.texStorage2D(i,1,r,t,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:o,texShape:[n,t]}}function eI(e){return e.internalFormatFloat}function j2(e,t,n,r){let[s,a]=op(t,n);return ip(e,s,a,eI(r),r.textureFormatFloat,e.FLOAT)}function tI(e){return e.internalFormatHalfFloat}function K2(e,t,n,r){let[s,a]=op(t,n);return ip(e,s,a,tI(r),r.textureFormatFloat,r.textureTypeHalfFloat)}function nI(e){return e.downloadTextureFormat}function X2(e,t,n,r){let[s,a]=op(t,n);return ip(e,s,a,nI(r),e.RGBA,e.UNSIGNED_BYTE)}function rI(e){return e.internalFormatPackedFloat}function Y2(e,t,n,r){let[s,a]=Xu(t,n);return ip(e,s,a,rI(r),e.RGBA,e.FLOAT)}function sI(e){return e.internalFormatPackedHalfFloat}function Z2(e,t,n,r){let[s,a]=Xu(t,n);return ip(e,s,a,sI(r),e.RGBA,r.textureTypeHalfFloat)}function J2(e,t,n){return me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),fv(e,t,"clipSpacePos",n,3,20,0)&&fv(e,t,"uv",n,2,20,12)}function Q2(e,t,n,r,s,a){me(e,()=>e.bindTexture(e.TEXTURE_2D,t));let o,i,c;s instanceof Uint8Array?(o=new Uint8Array(n*r*4),i=e.UNSIGNED_BYTE,c=e.RGBA):(o=new Float32Array(n*r*4),i=e.FLOAT,c=a.internalFormatPackedFloat),o.set(s),q().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n,r,e.RGBA,i,o)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,c,n,r,0,e.RGBA,i,o)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function eE(e,t,n){me(e,()=>e.bindTexture(e.TEXTURE_2D,t)),n.data instanceof Uint8Array?q().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n.width,n.height,e.RGBA,e.UNSIGNED_BYTE,n.data)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,n.width,n.height,0,e.RGBA,e.UNSIGNED_BYTE,n.data)):q().getNumber("WEBGL_VERSION")===2?me(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,n)):me(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,n)),me(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function tE(e,t,n,r){let s=e.createBuffer();me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,s));let i=4*4*t*n;return me(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,i,e.STREAM_READ)),me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,0)),me(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),s}function nE(e,t,n){let r=e,s=new Float32Array(n);return r.bindBuffer(r.PIXEL_PACK_BUFFER,t),r.getBufferSubData(r.PIXEL_PACK_BUFFER,0,s),r.bindBuffer(r.PIXEL_PACK_BUFFER,null),s}function rE(e,t,n,r){let[s,a]=op(t,n),o=4,i=new Uint8Array(Z7(t*n,o));return me(e,()=>e.readPixels(0,0,s,a,r.downloadTextureFormat,e.UNSIGNED_BYTE,i)),new Float32Array(i.buffer)}function sE(e,t,n,r,s,a,o,i){let c=e,u=new Float32Array(J7(a,o));return c.bindBuffer(c.PIXEL_PACK_BUFFER,t),c.getBufferSubData(c.PIXEL_PACK_BUFFER,0,u),c.bindBuffer(c.PIXEL_PACK_BUFFER,null),u}function aE(e,t,n){let r=new Float32Array(t*n*4);return me(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,r)),r}var Ch=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let t=q().getNumber("WEBGL_VERSION");e!=null?(this.gl=e,h2(t,e)):this.gl=Kr(t);let n="WEBGL_color_buffer_float",r="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),q().getNumber("WEBGL_VERSION")===1){let s="OES_texture_float",a="OES_texture_half_float";if(this.textureFloatExtension=zl(this.gl,s),lr(this.gl,a))this.textureHalfFloatExtension=zl(this.gl,a);else if(q().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),lr(this.gl,r))this.colorBufferHalfFloatExtension=zl(this.gl,r);else if(q().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",lr(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else if(lr(this.gl,r))this.colorBufferHalfFloatExtension=this.gl.getExtension(r);else throw new Error("GL context does not support color renderable floats");this.vertexBuffer=H2(this.gl),this.indexBuffer=q2(this.gl),this.framebuffer=S2(this.gl),this.textureConfig=X0(this.gl,this.textureHalfFloatExtension)}get debug(){return q().getBool("DEBUG")}dispose(){if(this.disposed)return;this.program!=null&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),this.outputTexture!=null&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let e=this.gl;me(e,()=>e.finish()),me(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),me(e,()=>e.deleteFramebuffer(this.framebuffer)),me(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),me(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),me(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),j2(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),K2(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),X2(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),eE(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,n,r){this.throwIfDisposed(),Q2(this.gl,e,t,n,r,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),Z2(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),Y2(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(mv(this.gl,this.framebuffer),this.outputTexture=null),me(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,n){return this.downloadMatrixDriver(e,()=>rE(this.gl,t,n,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,n,r,s,a){return sE(this.gl,e,t,n,r,s,a,this.textureConfig)}downloadFloat32MatrixFromBuffer(e,t){return nE(this.gl,e,t)}createBufferFromTexture(e,t,n){this.bindTextureToFrameBuffer(e);let r=tE(this.gl,t,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),r}createAndWaitForFence(){let e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,n;if(q().getBool("WEBGL_FENCE_API_ENABLED")){let r=e,s=r.fenceSync(r.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),n=()=>{let a=r.clientWaitSync(s,0,0);return a===r.ALREADY_SIGNALED||a===r.CONDITION_SATISFIED},t=s}else q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(t,q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:t,isFencePassed:n}}downloadMatrixFromPackedTexture(e,t,n){return this.downloadMatrixDriver(e,()=>aE(this.gl,t,n))}createProgram(e){this.throwIfDisposed();let t=this.gl;this.vertexShader==null&&(this.vertexShader=G2(t));let n=y2(t);return me(t,()=>t.attachShader(n,this.vertexShader)),me(t,()=>t.attachShader(n,e)),v2(t,n),this.debug&&Ih(t,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=J2(t,this.program,this.vertexBuffer)),n}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),e!=null&&me(this.gl,()=>this.gl.deleteProgram(e))}setProgram(e){this.throwIfDisposed(),this.program=e,this.program!=null&&this.debug&&Ih(this.gl,this.program),me(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,n=!0){return this.throwIfDisposed(),n?C2(this.gl,e,t):N2(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),me(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,n){this.throwIfDisposed(),this.throwIfNoProgram(),_2(this.gl,e,t,n)}setOutputMatrixTexture(e,t,n){this.setOutputMatrixTextureDriver(e,n,t)}setOutputPackedMatrixTexture(e,t,n){this.throwIfDisposed();let[r,s]=Xu(t,n);this.setOutputMatrixTextureDriver(e,r,s)}setOutputMatrixWriteRegion(e,t,n,r){this.setOutputMatrixWriteRegionDriver(n,e,r,t)}setOutputPackedMatrixWriteRegion(e,t,n,r){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){this.program!=null&&Ih(this.gl,this.program),Bl(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let e=this.gl;this.debug&&this.debugValidate(),me(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),me(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=zl(this.gl,q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let n=this.gl,r=this.getQueryTimerExtensionWebGL2(),s=n.createQuery();return n.beginQuery(r.TIME_ELAPSED_EXT,s),s}let e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let t=this.gl,n=this.getQueryTimerExtensionWebGL2();t.endQuery(n.TIME_ELAPSED_EXT);return}let e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await w.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(t===0)return null;if(t===2){let n=this.gl;return n.getQueryParameter(e,n.QUERY_RESULT)/1e6}else{let n=this.getQueryTimerExtensionWebGL1();return n.getQueryObjectEXT(e,n.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(t===0)return!0;if(t===2){let n=this.gl,r=this.getQueryTimerExtensionWebGL2(),s=n.getQueryParameter(e,n.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(r.GPU_DISJOINT_EXT)),s&&!this.disjoint}else{let n=this.getQueryTimerExtensionWebGL1(),r=n.getQueryObjectEXT(e,n.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),r&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){let e=n9(this.itemsToPoll.map(t=>t.isDoneFn));for(let t=0;t<=e;++t){let{resolveFn:n}=this.itemsToPoll[t];n()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){if(this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),this.itemsToPoll.length>1)return;let n;"setTimeoutCustom"in q().platform&&(n=q().platform.setTimeoutCustom.bind(q().platform)),w.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0),()=>0,null,n)}bindTextureToFrameBuffer(e){this.throwIfDisposed(),kh(this.gl,e,this.framebuffer),this.debug&&Bl(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(kh(this.gl,this.outputTexture,this.framebuffer),this.debug&&Bl(this.gl)):mv(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);let n=t();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(e,t,n){this.throwIfDisposed();let r=this.gl;kh(r,e,this.framebuffer),this.debug&&Bl(r),this.outputTexture=e,me(r,()=>r.viewport(0,0,t,n)),me(r,()=>r.scissor(0,0,t,n))}setOutputMatrixWriteRegionDriver(e,t,n,r){this.throwIfDisposed(),me(this.gl,()=>this.gl.scissor(e,t,n,r))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(this.program==null)throw new Error("No GPU program is currently set.")}};function n9(e){let t=0;for(;t<e.length&&e[t]();++t);return t-1}var{addImpl:r9,bincountImpl:oE,bincountReduceImpl:s9,castImpl:a9,ceilImpl:o9,concatImpl:i9,equalImpl:c9,expImpl:u9,expm1Impl:l9,floorImpl:d9,gatherNdImpl:p9,gatherV2Impl:h9,greaterImpl:f9,greaterEqualImpl:m9,lessImpl:g9,lessEqualImpl:b9,linSpaceImpl:y9,logImpl:v9,maxImpl:x9,maximumImpl:w9,minimumImpl:I9,multiplyImpl:k9,negImpl:S9,notEqualImpl:T9,prodImpl:C9,raggedGatherImpl:N9,raggedRangeImpl:_9,raggedTensorToTensorImpl:E9,rangeImpl:A9,rsqrtImpl:$9,scatterImpl:D9,sigmoidImpl:F9,simpleAbsImpl:iE,sliceImpl:R9,sparseFillEmptyRowsImpl:P9,sparseReshapeImpl:O9,sparseSegmentReductionImpl:cE,sqrtImpl:M9,stridedSliceImpl:L9,stringNGramsImpl:z9,stringSplitImpl:B9,stringToHashBucketFastImpl:W9,subImpl:V9,tileImpl:U9,topKImpl:G9,transposeImpl:aI,uniqueImpl:H9}=l_;function uE(e,t){return["x","y","z","w","u","v"].slice(0,t).map(n=>`${e}.${n}`)}function Tn(e,t){return t===1?[e]:uE(e,t)}function q9(e,t){if(e===1)return"rc";let n="";for(let r=0;r<e;r++)n+=t[r],r<e-1&&(n+=",");return n}var j9=class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=Dn(this.outputShape.length),this.rank===0)this.userCode=`
         void main() {
           setOutput(vec4(getA(), 0., 0., 0.));
         }
-      `;else{let t=kn("rc",this.rank),n=gt(this.rank),a=this.getOutOfBoundsCondition(t),r=this.getSetup(t),s=this.getOutput(t);this.userCode=`
+      `;else{let t=Tn("rc",this.rank),n=mt(this.rank),r=this.getOutOfBoundsCondition(t),s=this.getSetup(t),a=this.getOutput(t);this.userCode=`
         void main() {
           ${n} rc = getOutputCoords();
 
-          if(${a}) {
+          if(${r}) {
             setOutput(vec4(0));
           } else {
-            ${r}
+            ${s}
 
-            setOutput(vec4(${s}));
+            setOutput(vec4(${a}));
           }
         }
-      `}}getSourceCoordsArr(e){let t=[];for(let n=0;n<=1;n++)for(let a=0;a<=1;a++){let r=`${n===0?"r":"rp1"}, ${a===0?"c":"cp1"}`;for(let s=2;s<this.rank;s++)r=`${e[e.length-1-s]},`+r;t.push(r)}return t}getOutOfBoundsCondition(e){if(this.rank===1)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let t="";for(let n=this.rank-2;n<this.rank;n++)t+=`${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(t+="||");return t}getSetup(e){if(this.rank===1)return"";let t=e.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],a=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`
+      `}}getSourceCoordsArr(e){let t=[];for(let n=0;n<=1;n++)for(let r=0;r<=1;r++){let s=`${n===0?"r":"rp1"}, ${r===0?"c":"cp1"}`;for(let a=2;a<this.rank;a++)s=`${e[e.length-1-a]},`+s;t.push(s)}return t}getOutOfBoundsCondition(e){if(this.rank===1)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let t="";for(let n=this.rank-2;n<this.rank;n++)t+=`${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(t+="||");return t}getSetup(e){if(this.rank===1)return"";let t=e.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],r=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`
       int r = ${t[0]};
       int c = ${t[1]};
       int rp1 = r + 1;
       int cp1 = c + 1;
 
       bool cEdge = cp1 >= ${n};
-      bool rEdge = rp1 >= ${a};
+      bool rEdge = rp1 >= ${r};
     `}getOutput(e){let t=this.getSourceCoordsArr(e);return this.rank===1?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${t[0]}),
             cEdge ? 0. : getA(${t[1]}),
             rEdge ? 0. : getA(${t[2]}),
-            rEdge || cEdge ? 0. : getA(${t[3]})`}},lE=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length);let n="";for(let a=0;a<4;a++){let r="thisRC = rc;";a%2===1&&(r+="thisRC.z += 1;"),a>1&&(r+="thisRC.y += 1;"),n+=`
-        ${r}
-        ${a>0?"if(thisRC.y < rows && thisRC.z < cols){":""}
+            rEdge || cEdge ? 0. : getA(${t[3]})`}},lE=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length);let n="";for(let r=0;r<4;r++){let s="thisRC = rc;";r%2===1&&(s+="thisRC.z += 1;"),r>1&&(s+="thisRC.y += 1;"),n+=`
+        ${s}
+        ${r>0?"if(thisRC.y < rows && thisRC.z < cols){":""}
           int flatIndex = getFlatIndex(thisRC);
 
           ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);
           vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));
 
-          result[${a}] =
+          result[${r}] =
             getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);
-        ${a>0?"}":""}
+        ${r>0?"}":""}
       `}this.userCode=`
-      ${H7(t,this.enableShapeUniforms)}
+      ${K9(t,this.enableShapeUniforms)}
       ${this.enableShapeUniforms?J0():Z0(e)}
 
       void main() {
@@ -1170,12 +1170,12 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(result);
       }
-    `}};function H7(e,t){return`
+    `}};function K9(e,t){return`
     ivec3 inputCoordsFromReshapedOutCoords(int index) {
-      ${t?oZ(["r","c","d"],"inputShape"):wo(["r","c","d"],e)}
+      ${t?lY(["r","c","d"],"inputShape"):Ii(["r","c","d"],e)}
       return ivec3(r, c, d);
     }
-  `}var j7=class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(e,t,n){let a=fI(t,n),r=gI(e,a,n);r in this.freeTextures||(this.freeTextures[r]=[]),r in this.usedTextures||(this.usedTextures[r]=[]);let s=mI(e,a,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[r].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=s,this.log();let o=this.freeTextures[r].shift();return this.usedTextures[r].push(o),o}let i;return a===ln.PACKED_2X2_FLOAT32?i=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):a===ln.PACKED_2X2_FLOAT16?i=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):a===ln.UNPACKED_FLOAT32?i=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):a===ln.UNPACKED_FLOAT16?i=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):a===ln.PACKED_4X1_UNSIGNED_BYTE&&(i=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[r].push(i),this.numUsedTextures++,this._numBytesAllocated+=s,this.log(),i}releaseTexture(e,t,n,a){if(this.freeTextures==null)return;let r=fI(n,a),s=gI(t,r,a);s in this.freeTextures||(this.freeTextures[s]=[]);let i=mI(t,r,this.gpgpu.gl,this.gpgpu.textureConfig,a),o=H().get("WEBGL_DELETE_TEXTURE_THRESHOLD");o!==-1&&this._numBytesAllocated>o?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=i):(this.freeTextures[s].push(e),this.numFreeTextures++,this._numBytesFree+=i),this.numUsedTextures--;let l=this.usedTextures[s],u=l.indexOf(e);if(u<0)throw new Error("Cannot release a texture that was never provided by this texture manager");l.splice(u,1),this.log()}log(){if(!this.logEnabled)return;let e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);let t=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let e in this.freeTextures)this.freeTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});for(let e in this.usedTextures)this.usedTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function q7(e,t){let n=e;if(t===n.R32F)return 4;if(t===n.R16F)return 2;if(t===n.RGBA32F||t===e.RGBA)return 16;if(t===n.RGBA16F)return 8;if(t===n.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function mI(e,t,n,a,r){let s=K7(t,a),i;if(r){let[l,u]=Xu(e[0],e[1]);i=l*u}else{let[l,u]=id(e[0],e[1]);i=l*u}let o=q7(n,s);return i*o}function K7(e,t){switch(e){case ln.PACKED_2X2_FLOAT32:return a1(t);case ln.PACKED_2X2_FLOAT16:return r1(t);case ln.UNPACKED_FLOAT32:return e1(t);case ln.UNPACKED_FLOAT16:return t1(t);case ln.PACKED_4X1_UNSIGNED_BYTE:return n1(t);default:throw new Error(`Unknown physical texture type ${e}`)}}function X7(e){return H().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?ln.PACKED_2X2_FLOAT32:ln.UNPACKED_FLOAT32:e?ln.PACKED_2X2_FLOAT16:ln.UNPACKED_FLOAT16}function fI(e,t){if(e===ca.UPLOAD)return ln.PACKED_2X2_FLOAT32;if(e===ca.RENDER||e==null)return X7(t);if(e===ca.DOWNLOAD||e===ca.PIXELS)return ln.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${e}`)}function gI(e,t,n){return`${e[0]}_${e[1]}_${t}_${n}`}var Cr=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length),this.userCode=`
+  `}var X9=class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(e,t,n){let r=g1(t,n),s=b1(e,r,n);s in this.freeTextures||(this.freeTextures[s]=[]),s in this.usedTextures||(this.usedTextures[s]=[]);let a=m1(e,r,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[s].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=a,this.log();let i=this.freeTextures[s].shift();return this.usedTextures[s].push(i),i}let o;return r===cn.PACKED_2X2_FLOAT32?o=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):r===cn.PACKED_2X2_FLOAT16?o=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):r===cn.UNPACKED_FLOAT32?o=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):r===cn.UNPACKED_FLOAT16?o=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):r===cn.PACKED_4X1_UNSIGNED_BYTE&&(o=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[s].push(o),this.numUsedTextures++,this._numBytesAllocated+=a,this.log(),o}releaseTexture(e,t,n,r){if(this.freeTextures==null)return;let s=g1(n,r),a=b1(t,s,r);a in this.freeTextures||(this.freeTextures[a]=[]);let o=m1(t,s,this.gpgpu.gl,this.gpgpu.textureConfig,r),i=q().get("WEBGL_DELETE_TEXTURE_THRESHOLD");i!==-1&&this._numBytesAllocated>i?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=o):(this.freeTextures[a].push(e),this.numFreeTextures++,this._numBytesFree+=o),this.numUsedTextures--;let c=this.usedTextures[a],u=c.indexOf(e);if(u<0)throw new Error("Cannot release a texture that was never provided by this texture manager");c.splice(u,1),this.log()}log(){if(!this.logEnabled)return;let e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);let t=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let e in this.freeTextures)this.freeTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});for(let e in this.usedTextures)this.usedTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function Y9(e,t){let n=e;if(t===n.R32F)return 4;if(t===n.R16F)return 2;if(t===n.RGBA32F)return 16;if(t===e.RGBA)return 16;if(t===n.RGBA16F)return 8;if(t===n.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function m1(e,t,n,r,s){let a=Z9(t,r),o;if(s){let[c,u]=Xu(e[0],e[1]);o=c*u}else{let[c,u]=op(e[0],e[1]);o=c*u}let i=Y9(n,a);return o*i}function Z9(e,t){switch(e){case cn.PACKED_2X2_FLOAT32:return rI(t);case cn.PACKED_2X2_FLOAT16:return sI(t);case cn.UNPACKED_FLOAT32:return eI(t);case cn.UNPACKED_FLOAT16:return tI(t);case cn.PACKED_4X1_UNSIGNED_BYTE:return nI(t);default:throw new Error(`Unknown physical texture type ${e}`)}}function J9(e){return q().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?cn.PACKED_2X2_FLOAT32:cn.UNPACKED_FLOAT32:e?cn.PACKED_2X2_FLOAT16:cn.UNPACKED_FLOAT16}function g1(e,t){if(e===ur.UPLOAD)return cn.PACKED_2X2_FLOAT32;if(e===ur.RENDER||e==null)return J9(t);if(e===ur.DOWNLOAD||e===ur.PIXELS)return cn.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${e}`)}function b1(e,t,n){return`${e[0]}_${e[1]}_${t}_${n}`}var Ns=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length),this.userCode=`
       float unaryOperation(float x) {
         ${t}
       }
@@ -1186,11 +1186,11 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}},Ma="if (isnan(x)) return x;",Y7="return x;",yI="return abs(x);",Z7="return (x >= 0.0) ? x : (exp(x) - 1.0);",J7=Ma+`
+    `}},Or="if (isnan(x)) return x;",Q9="return x;",y1="return abs(x);",eZ="return (x >= 0.0) ? x : (exp(x) - 1.0);",tZ=Or+`
   return (x < 0.0) ? 0.0 : x;
-`,Q7=Ma+`
+`,nZ=Or+`
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,rl="return x;",eJ="return 1.0 / (1.0 + exp(-1.0 * x));",tJ="return x;",nJ=`
+`,sc="return x;",rZ="return 1.0 / (1.0 + exp(-1.0 * x));",sZ="return x;",aZ=`
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -1199,7 +1199,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,aJ=`
+`,oZ=`
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1209,7 +1209,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,rJ=`
+`,iZ=`
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1219,7 +1219,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,sJ="return 1.0 / (1.0 + exp(-1.0 * x));",qs=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length),this.userCode=`
+`,cZ="return 1.0 / (1.0 + exp(-1.0 * x));",Ka=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length),this.userCode=`
       vec4 unaryOperation(vec4 x) {
         ${t}
       }
@@ -1230,17 +1230,17 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}},iJ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length);let t=e.length,n=kn("rc",t),a=gt(t),r=U7(t,n),s=n.slice(-2),i=t<=1?"rc":`vec2(${s.join(",")})`;this.userCode=`
+    `}},uZ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length);let t=e.length,n=Tn("rc",t),r=mt(t),s=q9(t,n),a=n.slice(-2),o=t<=1?"rc":`vec2(${a.join(",")})`;this.userCode=`
       void main() {
-        ${a} rc = getOutputCoords();
-        vec4 packedInput = getA(${r});
+        ${r} rc = getOutputCoords();
+        vec4 packedInput = getA(${s});
 
-        setOutput(getChannel(packedInput, ${i}));
+        setOutput(getChannel(packedInput, ${o}));
       }
-    `}},oJ=hr.whereImpl,lJ=1e-7,uJ=1e-4,Sb={};function pJ(e){return e in Sb||(Sb[e]={}),Sb[e]}var cJ=H().getNumber("CPU_HANDOFF_SIZE_THRESHOLD"),dJ=600;function hJ(){return H().global.screen==null?1024:H().global.screen.height*H().global.screen.width*window.devicePixelRatio*dJ/1024/1024}var Xf=class extends pc{constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!H().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(e!=null){if(e instanceof Th)t=e;else{let n=qa(H().getNumber("WEBGL_VERSION"),e);t=new Th(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let n=qa(H().getNumber("WEBGL_VERSION"));t=new Th(n),this.binaryCache=pJ(H().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new j7(this.gpgpu),this.numMBBeforeWarning=hJ(),this.texData=new om(this,_a())}nextDataId(){return Xf.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(e,t,n,a,r,s){let i=this.makeTensorInfo(t,n),o=this.texData.get(i.dataId);o.isPacked=!1,o.texture={texture:e,texShape:[a,r]},o.texShape=[a,r];let l=Bp(t),u=new hI(l,!1,s),p=this.runWebGLProgram(u,[i],n,[[a,r]]);return p.shape=t,o.texture=null,this.disposeIntermediateTensorInfo(i),p.dataId}write(e,t,n){if((H().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||H().getBool("DEBUG"))&&this.checkNumericalProblems(e),n==="complex64"&&e!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let a={id:this.nextDataId()};return this.texData.set(a,{shape:t,dtype:n,values:e,usage:ca.UPLOAD,refCount:1}),a}refCount(e){return this.texData.has(e)?this.texData.get(e).refCount:0}incRef(e){let t=this.texData.get(e);t.refCount++}decRef(e){if(this.texData.has(e)){let t=this.texData.get(e);t.refCount--}}move(e,t,n,a,r){if(H().getBool("DEBUG")&&this.checkNumericalProblems(t),a==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:n,dtype:a,values:t,usage:ca.UPLOAD,refCount:r})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){let t=this.texData.get(e),{values:n,dtype:a,complexTensorInfos:r,slice:s,shape:i,isPacked:o}=t;if(s!=null){let d;o?d=new qs(i,rl):d=new Cr(i,rl);let c=this.runWebGLProgram(d,[{dataId:e,shape:i,dtype:a}],a),h=this.readSync(c.dataId);return this.disposeIntermediateTensorInfo(c),h}if(n!=null)return this.convertAndCacheOnCPU(e);if(a==="string")return n;let l=this.activeTimers!=null,u;l&&(u=v.now());let p;if(a==="complex64"){let d=this.readSync(r.real.dataId),c=this.readSync(r.imag.dataId);p=N.mergeRealAndImagArrays(d,c)}else p=this.getValuesFromTexture(e);return l&&(this.downloadWaitMs+=v.now()-u),this.convertAndCacheOnCPU(e,p)}async read(e){if(this.pendingRead.has(e)){let h=this.pendingRead.get(e);return new Promise(m=>h.push(m))}let t=this.texData.get(e),{values:n,shape:a,slice:r,dtype:s,complexTensorInfos:i,isPacked:o}=t;if(r!=null){let h;o?h=new qs(a,rl):h=new Cr(a,rl);let m=this.runWebGLProgram(h,[{dataId:e,shape:a,dtype:s}],s),f=this.read(m.dataId);return this.disposeIntermediateTensorInfo(m),f}if(n!=null)return this.convertAndCacheOnCPU(e);if(H().getBool("DEBUG")&&!H().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&H().getNumber("WEBGL_VERSION")===2)throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let l=null,u;if(s!=="complex64"&&H().get("WEBGL_BUFFER_SUPPORTED")){u=this.decode(e);let h=this.texData.get(u.dataId);l=this.gpgpu.createBufferFromTexture(h.texture.texture,...fh(a))}this.pendingRead.set(e,[]),s!=="complex64"&&await this.gpgpu.createAndWaitForFence();let p;if(s==="complex64"){let h=await Promise.all([this.read(i.real.dataId),this.read(i.imag.dataId)]),m=h[0],f=h[1];p=N.mergeRealAndImagArrays(m,f)}else if(l==null)p=this.getValuesFromTexture(e);else{let h=v.sizeFromShape(a);p=this.gpgpu.downloadFloat32MatrixFromBuffer(l,h)}if(u!=null&&this.disposeIntermediateTensorInfo(u),l!=null){let h=this.gpgpu.gl;me(h,()=>h.deleteBuffer(l))}let d=this.convertAndCacheOnCPU(e,p),c=this.pendingRead.get(e);return this.pendingRead.delete(e),c.forEach(h=>h(d)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&_a().removeDataId(e,this),this.pendingDeletes--),d}readToGPU(e,t={}){let n=this.texData.get(e),{values:a,shape:r,slice:s,dtype:i,isPacked:o,texture:l}=n;if(i==="complex64")throw new Error("Does not support reading texture for complex64 dtype.");if(s!=null){let c;o?c=new qs(r,rl):c=new Cr(r,rl);let h=this.runWebGLProgram(c,[{dataId:e,shape:r,dtype:i}],i),m=this.readToGPU(h,t);return this.disposeIntermediateTensorInfo(h),m}if(l==null)throw a!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let u=this.decode(e,t.customTexShape),p=_a().makeTensorFromTensorInfo(u),d=this.texData.get(u.dataId);return Object.assign({tensorRef:p},d.texture)}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(a=>v.decodeString(a));return Oe(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Oe(e.shape,e.dtype,t)}checkNumericalProblems(e){if(e!=null)for(let t=0;t<e.length;t++){let n=e[t];if(!d_(n))throw H().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(e){let{shape:t,dtype:n,isPacked:a}=this.texData.get(e),r=v.sizeFromShape(t);if(H().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let d=this.decode(e),c=this.texData.get(d.dataId),h=this.gpgpu.downloadMatrixFromPackedTexture(c.texture.texture,...fh(t)).subarray(0,r);return this.disposeIntermediateTensorInfo(d),h}let s=H().getBool("WEBGL_PACK")&&a===!0,i=s?Bp(t):t,o=s?new YZ(i):new XZ(i),l=this.runWebGLProgram(o,[{shape:i,dtype:n,dataId:e}],"float32"),u=this.texData.get(l.dataId),p=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(u.texture.texture,u.texShape[0],u.texShape[1]).subarray(0,r);return this.disposeIntermediateTensorInfo(l),p}timerAvailable(){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){let t=this.activeTimers,n=[],a=!1;this.programTimersStack==null?(this.programTimersStack=n,a=!0):this.activeTimers.push(n),this.activeTimers=n,e();let r=v.flatten(this.activeTimers.map(o=>o.query)).filter(o=>o!=null),s=v.flatten(this.activeTimers.map(o=>o.name)).filter(o=>o!=null);this.activeTimers=t,a&&(this.programTimersStack=null);let i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let o=await Promise.all(r);i.kernelMs=v.sum(o),i.getExtraProfileInfo=()=>o.map((l,u)=>({name:s[u],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(", ")}else i.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,i})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:v.now(),endMs:null}}endTimer(e){return H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=v.now(),e)}async getQueryTime(e){if(H().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);let t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);let{complexTensorInfos:n}=this.texData.get(e);return n!=null&&(this.disposeData(n.real.dataId,t),this.disposeData(n.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){let{texture:t,dtype:n,texShape:a,usage:r,isPacked:s,slice:i}=this.texData.get(e),o=i&&i.origDataId||e,l=this.dataRefCount.get(o);l>1?this.dataRefCount.set(o,l-1):(this.dataRefCount.delete(o),t!=null&&(this.numBytesInGPU-=this.computeBytes(a,n),this.textureManager.releaseTexture(t,a,r,s)));let u=this.texData.get(e);u.texture=null,u.texShape=null,u.isPacked=!1,u.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=cJ){return H().getBool("WEBGL_CPU_FORWARD")&&e.every(n=>this.texData.get(n.dataId).texture==null&&v.sizeFromShape(n.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){N.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let t=e.dataSync();return oJ(e.shape,t)}packedUnaryOp(e,t,n){let a=new qs(e.shape,t),r=this.compileAndRun(a,[e],n);return _a().makeTensorFromTensorInfo(r)}abs(e){if(this.shouldExecuteOnCPU([e])&&e.dtype!=="complex64"){let a=sE(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,a)}if(H().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,yI,e.dtype);let t=new Cr(e.shape,yI),n=this.compileAndRun(t,[e]);return _a().makeTensorFromTensorInfo(n)}makeTensorInfo(e,t,n){let a;if(t==="string"&&n!=null&&n.length>0&&v.isString(n[0])){let r=n.map(s=>v.encodeString(s));a=this.write(r,e,t)}else a=this.write(n,e,t);return this.texData.get(a).usage=null,{dataId:a,shape:e,dtype:t}}makeOutput(e,t,n){return _a().makeTensorFromTensorInfo(this.makeTensorInfo(e,t,n),this)}unpackTensor(e){let t=new iJ(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){let t=new G7(e.shape),n=!0;return this.runWebGLProgram(t,[e],e.dtype,null,n)}packedReshape(e,t){let n=[ci(e.shape),...di(e.shape)],a={dtype:e.dtype,shape:n,dataId:e.dataId},r=[ci(t),...di(t)],s=new lE(r,n),i=!0,o=[n],l=this.runWebGLProgram(s,[a],e.dtype,o,i);return{dataId:l.dataId,shape:t,dtype:l.dtype}}decode(e,t){let n=this.texData.get(e),{isPacked:a,shape:r,dtype:s}=n;if(t!=null){let d=v.sizeFromShape(r),c=t[0]*t[1]*4;v.assert(d<=c,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let i=Bp(r),o;a?o=new KZ(i):o=new qZ(i);let l=!0,u=[t!=null?t:fh(i)],p=this.runWebGLProgram(o,[{shape:i,dtype:s,dataId:e}],s,u,l,t);return{dtype:s,shape:r,dataId:p.dataId}}runWebGLProgram(e,t,n,a,r=!1,s){let i=this.makeTensorInfo(e.outputShape,n),o=this.texData.get(i.dataId);if(e.packedOutput&&(o.isPacked=!0),e.outPackingScheme===rc.DENSE){let g=s!=null?s:fh(e.outputShape);o.texShape=g.map(y=>y*2)}if(e.outTexUsage!=null&&(o.usage=e.outTexUsage),v.sizeFromShape(i.shape)===0)return o.values=v.getTypedArrayFromDType(i.dtype,0),i;let l=[],u=t.map(g=>{if(g.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let y=this.texData.get(g.dataId);if(y.texture==null){if(!e.packedInputs&&v.sizeFromShape(g.shape)<=H().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:g.shape,texData:null,isUniform:!0,uniformValues:y.values};e.packedInputs&&(y.isPacked=!0,y.shape=g.shape)}if(this.uploadToGPU(g.dataId),!!y.isPacked!=!!e.packedInputs)g=y.isPacked?this.unpackTensor(g):this.packTensor(g),l.push(g),y=this.texData.get(g.dataId);else if(y.isPacked&&!sc(y.shape,g.shape)){let b=g,x=g.shape;g.shape=y.shape,g=this.packedReshape(g,x),l.push(g),y=this.texData.get(g.dataId),b.shape=x}return{shape:g.shape,texData:y,isUniform:!1}});this.uploadToGPU(i.dataId);let p={shape:i.shape,texData:o,isUniform:!1},d=jZ(e,u,p),c=this.getAndSaveBinary(d,()=>GZ(this.gpgpu,e,u,p)),h=this.activeTimers!=null,m;h&&(m=this.startTimer()),H().get("ENGINE_COMPILE_ONLY")||HZ(this.gpgpu,c,u,p,a),l.forEach(g=>this.disposeIntermediateTensorInfo(g)),h&&(m=this.endTimer(m),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(m)}));let f=H().get("WEBGL_FLUSH_THRESHOLD");if(f>0){let g=v.now();g-this.lastGlFlushTime>f&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=g)}if(!H().getBool("WEBGL_LAZILY_UNPACK")&&o.isPacked&&r===!1){let g=this.unpackTensor(i);return this.disposeIntermediateTensorInfo(i),g}return i}compileAndRun(e,t,n,a,r=!1){return n=n||t[0].dtype,this.runWebGLProgram(e,t,n,a,r)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(H().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!="undefined"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=P(()=>{if(!H().get("WEBGL_RENDER_FLOAT32_ENABLED")){let e=H().getBool("DEBUG");H().set("DEBUG",!1);let t=this.abs(be(1e-8)).dataSync()[0];if(H().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?lJ:uJ}uploadToGPU(e){let t=this.texData.get(e),{shape:n,dtype:a,values:r,texture:s,usage:i,isPacked:o}=t;if(s!=null)return;let l=this.activeTimers!=null,u;l&&(u=v.now());let p=t.texShape;if(p==null&&(p=E_(n,o),t.texShape=p),r!=null){let d=Bp(n),c,h=p[1],m=p[0],f=r instanceof Uint8Array||r instanceof Uint8ClampedArray;(o||!f)&&([h,m]=Xu(p[0],p[1])),o?c=new JZ(d,f):c=new hI(d,f);let g=f?[m,h]:p,y=this.makeTensorInfo(g,a),b=this.texData.get(y.dataId);f?b.usage=ca.PIXELS:b.usage=ca.UPLOAD,b.texShape=g,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(y.dataId),h,m,r);let x=[[m,h]],w=!0,I=this.runWebGLProgram(c,[y],a,x,w),T=this.texData.get(I.dataId);t.texShape=T.texShape,t.isPacked=T.isPacked,t.usage=T.usage,H().get("ENGINE_COMPILE_ONLY")?this.disposeData(I.dataId):(t.texture=T.texture,t.values=null,this.texData.delete(I.dataId)),this.disposeIntermediateTensorInfo(y),l&&(this.uploadWaitMs+=v.now()-u)}else{let d=this.acquireTexture(p,i,a,o);t.texture=d}}convertAndCacheOnCPU(e,t){let n=this.texData.get(e),{dtype:a}=n;return this.releaseGPUData(e),t!=null&&(n.values=mJ(t,a)),n.values}acquireTexture(e,t,n,a){if(this.numBytesInGPU+=this.computeBytes(e,n),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let r=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${r} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,a)}computeBytes(e,t){return e[0]*e[1]*v.bytesPerElement(t)}checkCompileCompletion(){for(let[,e]of Object.entries(this.binaryCache))this.checkCompletion_(e)}async checkCompileCompletionAsync(){let e=[];if(this.gpgpu.parallelCompilationExtension){for(let[,t]of Object.entries(this.binaryCache))e.push(this.checkCompletionAsync_(t));return Promise.all(e)}else{for(let[,t]of Object.entries(this.binaryCache)){let n=new Promise(a=>{try{this.checkCompletion_(t),a(!0)}catch(r){throw r}});e.push(n)}return Promise.all(e)}}async checkCompletionAsync_(e){return this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(e):(await Jv(),this.checkCompletionAsync_(e))}checkCompletion_(e){if(this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(e.webGLProgram)),this.gpgpu.gl.getShaderParameter(e.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(Y0(e.source,this.gpgpu.gl.getShaderInfoLog(e.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let[,e]of Object.entries(this.binaryCache)){let{uniformLocations:t,customUniformLocations:n,infLoc:a,nanLoc:r,inShapesLocations:s,inTexShapesLocations:i,outShapeLocation:o,outShapeStridesLocation:l,outTexShapeLocation:u}=W_(this.gpgpu,e.program,e.webGLProgram);e.uniformLocations=t,e.customUniformLocations=n,e.infLoc=a,e.nanLoc=r,e.inShapesLocations=s,e.inTexShapesLocations=i,e.outShapeLocation=o,e.outShapeStridesLocation=l,e.outTexShapeLocation=u}}createTensorFromTexture(e,t,n){let{texture:a,height:r,width:s,channels:i}=e,o=_a().backend;if(!o.gpgpu.gl.isTexture(a))throw new Error("The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.");let l=o.writeTexture(a,t,n,r,s,i);return _a().makeTensorFromDataId(l,t,n,o)}};Xf.nextDataId=0;function mJ(e,t){if(t==="float32"||t==="complex64")return e;if(t==="int32"||t==="bool"){let n=t==="int32"?new Int32Array(e.length):new Uint8Array(e.length);for(let a=0;a<n.length;++a)n[a]=Math.round(e[a]);return n}else throw new Error(`Unknown dtype ${t}`)}var fJ="4.0.0";function uE(){H().set("WEBGL_FORCE_F16_TEXTURES",!0)}$c.isBrowser()&&Gm("webgl",()=>new Xf,2);var gJ={forceHalfFloat:uE},i1=`
+    `}},lZ=fs.whereImpl,dZ=1e-7,pZ=1e-4,bh={};function hZ(e){return e in bh||(bh[e]={}),bh[e]}var fZ=q().getNumber("CPU_HANDOFF_SIZE_THRESHOLD"),mZ=600;function gZ(){return q().global.screen==null?1024:q().global.screen.height*q().global.screen.width*window.devicePixelRatio*mZ/1024/1024}var Ym=class extends ld{constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!q().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(e!=null){if(e instanceof Ch)t=e;else{let n=Kr(q().getNumber("WEBGL_VERSION"),e);t=new Ch(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let n=Kr(q().getNumber("WEBGL_VERSION"));t=new Ch(n),this.binaryCache=hZ(q().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new X9(this.gpgpu),this.numMBBeforeWarning=gZ(),this.texData=new uf(this,Er())}nextDataId(){return Ym.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(e,t,n,r,s,a){let o=this.makeTensorInfo(t,n),i=this.texData.get(o.dataId);i.isPacked=!1,i.texture={texture:e,texShape:[r,s]},i.texShape=[r,s];let c=Wl(t),u=new f1(c,!1,a),l=this.runWebGLProgram(u,[o],n,[[r,s]]);return l.shape=t,i.texture=null,this.disposeIntermediateTensorInfo(o),l.dataId}write(e,t,n){if((q().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||q().getBool("DEBUG"))&&this.checkNumericalProblems(e),n==="complex64"&&e!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let r={id:this.nextDataId()};return this.texData.set(r,{shape:t,dtype:n,values:e,usage:ur.UPLOAD,refCount:1}),r}refCount(e){return this.texData.has(e)?this.texData.get(e).refCount:0}incRef(e){let t=this.texData.get(e);t.refCount++}decRef(e){if(this.texData.has(e)){let t=this.texData.get(e);t.refCount--}}move(e,t,n,r,s){if(q().getBool("DEBUG")&&this.checkNumericalProblems(t),r==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:n,dtype:r,values:t,usage:ur.UPLOAD,refCount:s})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){let t=this.texData.get(e),{values:n,dtype:r,complexTensorInfos:s,slice:a,shape:o,isPacked:i}=t;if(a!=null){let p;i?p=new Ka(o,sc):p=new Ns(o,sc);let d=this.runWebGLProgram(p,[{dataId:e,shape:o,dtype:r}],r),h=this.readSync(d.dataId);return this.disposeIntermediateTensorInfo(d),h}if(n!=null)return this.convertAndCacheOnCPU(e);if(r==="string")return n;let c=this.activeTimers!=null,u;c&&(u=w.now());let l;if(r==="complex64"){let p=this.readSync(s.real.dataId),d=this.readSync(s.imag.dataId);l=N.mergeRealAndImagArrays(p,d)}else l=this.getValuesFromTexture(e);return c&&(this.downloadWaitMs+=w.now()-u),this.convertAndCacheOnCPU(e,l)}async read(e){if(this.pendingRead.has(e)){let h=this.pendingRead.get(e);return new Promise(f=>h.push(f))}let t=this.texData.get(e),{values:n,shape:r,slice:s,dtype:a,complexTensorInfos:o,isPacked:i}=t;if(s!=null){let h;i?h=new Ka(r,sc):h=new Ns(r,sc);let f=this.runWebGLProgram(h,[{dataId:e,shape:r,dtype:a}],a),m=this.read(f.dataId);return this.disposeIntermediateTensorInfo(f),m}if(n!=null)return this.convertAndCacheOnCPU(e);if(q().getBool("DEBUG")&&!q().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&q().getNumber("WEBGL_VERSION")===2)throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let c=null,u;if(a!=="complex64"&&q().get("WEBGL_BUFFER_SUPPORTED")){u=this.decode(e);let h=this.texData.get(u.dataId);c=this.gpgpu.createBufferFromTexture(h.texture.texture,...mh(r))}this.pendingRead.set(e,[]),a!=="complex64"&&await this.gpgpu.createAndWaitForFence();let l;if(a==="complex64"){let h=await Promise.all([this.read(o.real.dataId),this.read(o.imag.dataId)]),f=h[0],m=h[1];l=N.mergeRealAndImagArrays(f,m)}else if(c==null)l=this.getValuesFromTexture(e);else{let h=w.sizeFromShape(r);l=this.gpgpu.downloadFloat32MatrixFromBuffer(c,h)}if(u!=null&&this.disposeIntermediateTensorInfo(u),c!=null){let h=this.gpgpu.gl;me(h,()=>h.deleteBuffer(c))}let p=this.convertAndCacheOnCPU(e,l),d=this.pendingRead.get(e);return this.pendingRead.delete(e),d.forEach(h=>h(p)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&Er().removeDataId(e,this),this.pendingDeletes--),p}readToGPU(e,t={}){let n=this.texData.get(e),{values:r,shape:s,slice:a,dtype:o,isPacked:i,texture:c}=n;if(o==="complex64")throw new Error("Does not support reading texture for complex64 dtype.");if(a!=null){let d;i?d=new Ka(s,sc):d=new Ns(s,sc);let h=this.runWebGLProgram(d,[{dataId:e,shape:s,dtype:o}],o),f=this.readToGPU(h,t);return this.disposeIntermediateTensorInfo(h),f}if(c==null)throw r!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let u=this.decode(e,t.customTexShape),l=Er().makeTensorFromTensorInfo(u),p=this.texData.get(u.dataId);return Object.assign({tensorRef:l},p.texture)}bufferSync(e){let t=this.readSync(e.dataId);if(e.dtype==="string")try{let n=t.map(r=>w.decodeString(r));return Me(e.shape,e.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return Me(e.shape,e.dtype,t)}checkNumericalProblems(e){if(e!=null)for(let t=0;t<e.length;t++){let n=e[t];if(!f2(n))throw q().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(e){let{shape:t,dtype:n,isPacked:r}=this.texData.get(e),s=w.sizeFromShape(t);if(q().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let p=this.decode(e),d=this.texData.get(p.dataId),h=this.gpgpu.downloadMatrixFromPackedTexture(d.texture.texture,...mh(t)).subarray(0,s);return this.disposeIntermediateTensorInfo(p),h}let a=q().getBool("WEBGL_PACK")&&r===!0,o=a?Wl(t):t,i=a?new QY(o):new JY(o),c=this.runWebGLProgram(i,[{shape:o,dtype:n,dataId:e}],"float32"),u=this.texData.get(c.dataId),l=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(u.texture.texture,u.texShape[0],u.texShape[1]).subarray(0,s);return this.disposeIntermediateTensorInfo(c),l}timerAvailable(){return q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){let t=this.activeTimers,n=[],r=!1;this.programTimersStack==null?(this.programTimersStack=n,r=!0):this.activeTimers.push(n),this.activeTimers=n,e();let s=w.flatten(this.activeTimers.map(i=>i.query)).filter(i=>i!=null),a=w.flatten(this.activeTimers.map(i=>i.name)).filter(i=>i!=null);this.activeTimers=t,r&&(this.programTimersStack=null);let o={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let i=await Promise.all(s);o.kernelMs=w.sum(i),o.getExtraProfileInfo=()=>i.map((c,u)=>({name:a[u],ms:c})).map(c=>`${c.name}: ${c.ms}`).join(", ")}else o.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,o})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:w.now(),endMs:null}}endTimer(e){return q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=w.now(),e)}async getQueryTime(e){if(q().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);let t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);let{complexTensorInfos:n}=this.texData.get(e);return n!=null&&(this.disposeData(n.real.dataId,t),this.disposeData(n.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){let{texture:t,dtype:n,texShape:r,usage:s,isPacked:a,slice:o}=this.texData.get(e),i=o&&o.origDataId||e,c=this.dataRefCount.get(i);c>1?this.dataRefCount.set(i,c-1):(this.dataRefCount.delete(i),t!=null&&(this.numBytesInGPU-=this.computeBytes(r,n),this.textureManager.releaseTexture(t,r,s,a)));let u=this.texData.get(e);u.texture=null,u.texShape=null,u.isPacked=!1,u.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=fZ){return q().getBool("WEBGL_CPU_FORWARD")&&e.every(n=>this.texData.get(n.dataId).texture==null&&w.sizeFromShape(n.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){N.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let t=e.dataSync();return lZ(e.shape,t)}packedUnaryOp(e,t,n){let r=new Ka(e.shape,t),s=this.compileAndRun(r,[e],n);return Er().makeTensorFromTensorInfo(s)}abs(e){if(this.shouldExecuteOnCPU([e])&&e.dtype!=="complex64"){let r=iE(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,r)}if(q().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,y1,e.dtype);let t=new Ns(e.shape,y1),n=this.compileAndRun(t,[e]);return Er().makeTensorFromTensorInfo(n)}makeTensorInfo(e,t,n){let r;if(t==="string"&&n!=null&&n.length>0&&w.isString(n[0])){let s=n.map(a=>w.encodeString(a));r=this.write(s,e,t)}else r=this.write(n,e,t);return this.texData.get(r).usage=null,{dataId:r,shape:e,dtype:t}}makeOutput(e,t,n){return Er().makeTensorFromTensorInfo(this.makeTensorInfo(e,t,n),this)}unpackTensor(e){let t=new uZ(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){let t=new j9(e.shape),n=!0;return this.runWebGLProgram(t,[e],e.dtype,null,n)}packedReshape(e,t){let n=[ho(e.shape),...fo(e.shape)],r={dtype:e.dtype,shape:n,dataId:e.dataId},s=[ho(t),...fo(t)],a=new lE(s,n),o=!0,i=[n],c=this.runWebGLProgram(a,[r],e.dtype,i,o);return{dataId:c.dataId,shape:t,dtype:c.dtype}}decode(e,t){let n=this.texData.get(e),{isPacked:r,shape:s,dtype:a}=n;if(t!=null){let p=w.sizeFromShape(s),d=t[0]*t[1]*4;w.assert(p<=d,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let o=Wl(s),i;r?i=new ZY(o):i=new YY(o);let c=!0,u=[t!=null?t:mh(o)],l=this.runWebGLProgram(i,[{shape:o,dtype:a,dataId:e}],a,u,c,t);return{dtype:a,shape:s,dataId:l.dataId}}runWebGLProgram(e,t,n,r,s=!1,a){let o=this.makeTensorInfo(e.outputShape,n),i=this.texData.get(o.dataId);if(e.packedOutput&&(i.isPacked=!0),e.outPackingScheme===sd.DENSE){let g=a!=null?a:mh(e.outputShape);i.texShape=g.map(b=>b*2)}if(e.outTexUsage!=null&&(i.usage=e.outTexUsage),w.sizeFromShape(o.shape)===0)return i.values=w.getTypedArrayFromDType(o.dtype,0),o;let c=[],u=t.map(g=>{if(g.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let b=this.texData.get(g.dataId);if(b.texture==null){if(!e.packedInputs&&w.sizeFromShape(g.shape)<=q().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:g.shape,texData:null,isUniform:!0,uniformValues:b.values};e.packedInputs&&(b.isPacked=!0,b.shape=g.shape)}if(this.uploadToGPU(g.dataId),!!b.isPacked!=!!e.packedInputs)g=b.isPacked?this.unpackTensor(g):this.packTensor(g),c.push(g),b=this.texData.get(g.dataId);else if(b.isPacked&&!ad(b.shape,g.shape)){let y=g,v=g.shape;g.shape=b.shape,g=this.packedReshape(g,v),c.push(g),b=this.texData.get(g.dataId),y.shape=v}return{shape:g.shape,texData:b,isUniform:!1}});this.uploadToGPU(o.dataId);let l={shape:o.shape,texData:i,isUniform:!1},p=XY(e,u,l),d=this.getAndSaveBinary(p,()=>jY(this.gpgpu,e,u,l)),h=this.activeTimers!=null,f;h&&(f=this.startTimer()),q().get("ENGINE_COMPILE_ONLY")||KY(this.gpgpu,d,u,l,r),c.forEach(g=>this.disposeIntermediateTensorInfo(g)),h&&(f=this.endTimer(f),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(f)}));let m=q().get("WEBGL_FLUSH_THRESHOLD");if(m>0){let g=w.now();g-this.lastGlFlushTime>m&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=g)}if(!q().getBool("WEBGL_LAZILY_UNPACK")&&i.isPacked&&s===!1){let g=this.unpackTensor(o);return this.disposeIntermediateTensorInfo(o),g}return o}compileAndRun(e,t,n,r,s=!1){return n=n||t[0].dtype,this.runWebGLProgram(e,t,n,r,s)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(q().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(t=>{this.gpgpu.deleteProgram(this.binaryCache[t].webGLProgram),delete this.binaryCache[t]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!="undefined"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=O(()=>{if(!q().get("WEBGL_RENDER_FLOAT32_ENABLED")){let e=q().getBool("DEBUG");q().set("DEBUG",!1);let t=this.abs(xe(1e-8)).dataSync()[0];if(q().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?dZ:pZ}uploadToGPU(e){let t=this.texData.get(e),{shape:n,dtype:r,values:s,texture:a,usage:o,isPacked:i}=t;if(a!=null)return;let c=this.activeTimers!=null,u;c&&(u=w.now());let l=t.texShape;if(l==null&&(l=$2(n,i),t.texShape=l),s!=null){let p=Wl(n),d,h=l[1],f=l[0],m=s instanceof Uint8Array||s instanceof Uint8ClampedArray;(i||!m)&&([h,f]=Xu(l[0],l[1])),i?d=new t9(p,m):d=new f1(p,m);let g=m?[f,h]:l,b=this.makeTensorInfo(g,r),y=this.texData.get(b.dataId);m?y.usage=ur.PIXELS:y.usage=ur.UPLOAD,y.texShape=g,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(b.dataId),h,f,s);let v=[[f,h]],x=!0,k=this.runWebGLProgram(d,[b],r,v,x),S=this.texData.get(k.dataId);t.texShape=S.texShape,t.isPacked=S.isPacked,t.usage=S.usage,q().get("ENGINE_COMPILE_ONLY")?this.disposeData(k.dataId):(t.texture=S.texture,t.values=null,this.texData.delete(k.dataId)),this.disposeIntermediateTensorInfo(b),c&&(this.uploadWaitMs+=w.now()-u)}else{let p=this.acquireTexture(l,o,r,i);t.texture=p}}convertAndCacheOnCPU(e,t){let n=this.texData.get(e),{dtype:r}=n;return this.releaseGPUData(e),t!=null&&(n.values=bZ(t,r)),n.values}acquireTexture(e,t,n,r){if(this.numBytesInGPU+=this.computeBytes(e,n),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let s=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${s} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,r)}computeBytes(e,t){return e[0]*e[1]*w.bytesPerElement(t)}checkCompileCompletion(){for(let[,e]of Object.entries(this.binaryCache))this.checkCompletion_(e)}async checkCompileCompletionAsync(){let e=[];if(this.gpgpu.parallelCompilationExtension){for(let[,t]of Object.entries(this.binaryCache))e.push(this.checkCompletionAsync_(t));return Promise.all(e)}else{for(let[,t]of Object.entries(this.binaryCache)){let n=new Promise(r=>{try{this.checkCompletion_(t),r(!0)}catch(s){throw s}});e.push(n)}return Promise.all(e)}}async checkCompletionAsync_(e){return this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(e):(await Jx(),this.checkCompletionAsync_(e))}checkCompletion_(e){if(this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(e.webGLProgram)),this.gpgpu.gl.getShaderParameter(e.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(Y0(e.source,this.gpgpu.gl.getShaderInfoLog(e.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let[,e]of Object.entries(this.binaryCache)){let{uniformLocations:t,customUniformLocations:n,infLoc:r,nanLoc:s,inShapesLocations:a,inTexShapesLocations:o,outShapeLocation:i,outShapeStridesLocation:c,outTexShapeLocation:u}=V2(this.gpgpu,e.program,e.webGLProgram);e.uniformLocations=t,e.customUniformLocations=n,e.infLoc=r,e.nanLoc=s,e.inShapesLocations=a,e.inTexShapesLocations=o,e.outShapeLocation=i,e.outShapeStridesLocation=c,e.outTexShapeLocation=u}}createTensorFromTexture(e,t,n){let{texture:r,height:s,width:a,channels:o}=e,i=Er().backend;if(!i.gpgpu.gl.isTexture(r))throw new Error("The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.");let c=i.writeTexture(r,t,n,s,a,o);return Er().makeTensorFromDataId(c,t,n,i)}};Ym.nextDataId=0;function bZ(e,t){if(t==="float32"||t==="complex64")return e;if(t==="int32"||t==="bool"){let n=t==="int32"?new Int32Array(e.length):new Uint8Array(e.length);for(let r=0;r<n.length;++r)n[r]=Math.round(e[r]);return n}else throw new Error(`Unknown dtype ${t}`)}var yZ="4.0.0";function dE(){q().set("WEBGL_FORCE_F16_TEXTURES",!0)}$d.isBrowser()&&qf("webgl",()=>new Ym,2);var vZ={forceHalfFloat:dE},oI=`
   if (isnan(a)) return a;
   if (isnan(b)) return b;
-`,$l=class{constructor(e,t,n){this.variableNames=["A","B"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.enableShapeUniforms=En(this.outputShape.length),this.userCode=`
+`,$c=class{constructor(e,t,n){this.variableNames=["A","B"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.enableShapeUniforms=Dn(this.outputShape.length),this.userCode=`
       float binaryOperation(float a, float b) {
         ${e}
       }
@@ -1250,38 +1250,38 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float b = getBAtOutCoords();
         setOutput(binaryOperation(a, b));
       }
-    `}},ld=`
+    `}},cp=`
   result.r = isNaN.r ? NAN : result.r;
   result.g = isNaN.g ? NAN : result.g;
   result.b = isNaN.b ? NAN : result.b;
   result.a = isNaN.a ? NAN : result.a;
-`,ud=class{constructor(e,t,n,a=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=N.assertAndGetBroadcastShape(t,n);let r=this.outputShape.length;this.enableShapeUniforms=En(r);let s="";if(a)if(r===0||v.sizeFromShape(this.outputShape)===1)s=`
+`,up=class{constructor(e,t,n,r=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=N.assertAndGetBroadcastShape(t,n);let s=this.outputShape.length;this.enableShapeUniforms=Dn(s);let a="";if(r)if(s===0||w.sizeFromShape(this.outputShape)===1)a=`
           result.y = 0.;
           result.z = 0.;
           result.w = 0.;
-        `;else if(s=`
-          ${gt(r)} coords = getOutputCoords();
-        `,r===1)this.enableShapeUniforms?s+=`
+        `;else if(a=`
+          ${mt(s)} coords = getOutputCoords();
+        `,s===1)this.enableShapeUniforms?a+=`
             result.y = (coords + 1) >= outShape ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `:s+=`
+          `:a+=`
             result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `;else{let i=kn("coords",r);this.enableShapeUniforms?s+=`
+          `;else{let i=Tn("coords",s);this.enableShapeUniforms?a+=`
             bool nextRowOutOfBounds =
-              (${i[r-2]} + 1) >= outShape[${r} - 2];
+              (${i[s-2]} + 1) >= outShape[${s} - 2];
             bool nextColOutOfBounds =
-              (${i[r-1]} + 1) >= outShape[${r} - 1];
+              (${i[s-1]} + 1) >= outShape[${s} - 1];
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
-          `:s+=`
+          `:a+=`
             bool nextRowOutOfBounds =
-              (${i[r-2]} + 1) >= ${this.outputShape[r-2]};
+              (${i[s-2]} + 1) >= ${this.outputShape[s-2]};
             bool nextColOutOfBounds =
-              (${i[r-1]} + 1) >= ${this.outputShape[r-1]};
+              (${i[s-1]} + 1) >= ${this.outputShape[s-1]};
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
@@ -1295,41 +1295,41 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         vec4 b = getBAtOutCoords();
 
         vec4 result = binaryOperation(a, b);
-        ${s}
+        ${a}
 
         setOutput(result);
       }
-    `}};function aa(e){let{inputs:t,backend:n}=e,{x:a}=t;return n.incRef(a.dataId),{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}var yJ={kernelName:Ri,backendName:"webgl",kernelFunc:aa};function _s(e){let{inputs:t,backend:n}=e,{real:a,imag:r}=t,s=n.makeTensorInfo(a.shape,"complex64"),i=n.texData.get(s.dataId),o=aa({inputs:{x:a},backend:n}),l=aa({inputs:{x:r},backend:n});return i.complexTensorInfos={real:o,imag:l},s}var bJ={kernelName:mm,backendName:"webgl",kernelFunc:_s},pE="return (a < 0.) ? b * a : a;",cE=`
+    `}};function sr(e){let{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var xZ={kernelName:Oo,backendName:"webgl",kernelFunc:sr};function Ea(e){let{inputs:t,backend:n}=e,{real:r,imag:s}=t,a=n.makeTensorInfo(r.shape,"complex64"),o=n.texData.get(a.dataId),i=sr({inputs:{x:r},backend:n}),c=sr({inputs:{x:s},backend:n});return o.complexTensorInfos={real:i,imag:c},a}var wZ={kernelName:gf,backendName:"webgl",kernelFunc:Ea},pE="return (a < 0.) ? b * a : a;",hE=`
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function xJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{alpha:s}=a,i=n.makeTensorInfo([],"float32",v.createScalarValue(s,"float32")),o=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ud(cE,r.shape,i.shape):new $l(pE,r.shape,i.shape),l=n.runWebGLProgram(o,[r,i],"float32");return n.disposeIntermediateTensorInfo(i),l}var vJ={kernelName:Mi,backendName:"webgl",kernelFunc:xJ},dE="return (a < 0.) ? b * a : a;",hE=`
+`;function IZ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{alpha:a}=r,o=n.makeTensorInfo([],"float32",w.createScalarValue(a,"float32")),i=q().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new up(hE,s.shape,o.shape):new $c(pE,s.shape,o.shape),c=n.runWebGLProgram(i,[s,o],"float32");return n.disposeIntermediateTensorInfo(o),c}var kZ={kernelName:Mo,backendName:"webgl",kernelFunc:IZ},fE="return (a < 0.) ? b * a : a;",mE=`
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function wJ(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t,s=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ud(hE,a.shape,r.shape):new $l(dE,a.shape,r.shape);return n.runWebGLProgram(s,[a,r],"float32")}var kJ={kernelName:Ki,backendName:"webgl",kernelFunc:wJ},tp="if (isnan(x)) return x;";function Ye({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:n,dtype:a}){return({inputs:r,backend:s})=>{let{x:i}=r,o=s,l=a||i.dtype;if(o.shouldExecuteOnCPU([i])&&n!=null){let d=o.texData.get(i.dataId),c=n(d.values,l);return o.makeTensorInfo(i.shape,l,c)}let u=H().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&t!=null,p;return u?p=new qs(i.shape,t):p=new Cr(i.shape,e),o.runWebGLProgram(p,[i],l)}}function cn({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:n=!1,supportsComplex:a=!1,cpuKernelImpl:r,dtype:s}){return({inputs:i,backend:o})=>{let{a:l,b:u}=i,p=o;if(a&&l.dtype==="complex64"){let m=p.texData.get(l.dataId),f=p.texData.get(u.dataId),[g,y]=[[m.complexTensorInfos.real,f.complexTensorInfos.real],[m.complexTensorInfos.imag,f.complexTensorInfos.imag]].map(x=>{let[w,I]=x,T={dataId:w.dataId,dtype:w.dtype,shape:l.shape},C={dataId:I.dataId,dtype:I.dtype,shape:u.shape},E=new $l(e,l.shape,u.shape);return p.runWebGLProgram(E,[T,C],fa(w.dtype,I.dtype))}),b=_s({inputs:{real:g,imag:y},backend:p});return p.disposeIntermediateTensorInfo(g),p.disposeIntermediateTensorInfo(y),b}let d=s||fa(l.dtype,u.dtype);if((l.dtype==="string"||u.dtype==="string"||p.shouldExecuteOnCPU([l,u]))&&r!=null){let m=p.texData.get(l.dataId).values,f=p.texData.get(u.dataId).values,g=l.dtype==="string"?N.fromUint8ToStringArray(m):m,y=l.dtype==="string"?N.fromUint8ToStringArray(f):f,[b,x]=r(l.shape,u.shape,g,y,d),w=p.makeTensorInfo(x,d),I=p.texData.get(w.dataId);return I.values=b,w}let c=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&t!=null,h;return c?h=new ud(t,l.shape,u.shape,n):h=new $l(e,l.shape,u.shape),p.runWebGLProgram(h,[l,u],d)}}function ic(e,t=!1){if(e==="linear")return t?tJ:Y7;if(e==="relu")return t?aJ:J7;if(e==="elu")return t?nJ:Z7;if(e==="relu6")return t?rJ:Q7;if(e==="prelu")return t?hE:dE;if(e==="leakyrelu")return t?cE:pE;if(e==="sigmoid")return t?sJ:eJ;throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var mE=class{constructor(e,t,n,a=!1,r=!1,s=!1,i=null,o=!1,l=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=En(this.outputShape.length);let u=a?e[1]:e[2],p=Math.ceil(u/2),d=a?"i * 2, rc.y":"rc.y, i * 2",c=r?"rc.z, i * 2":"i * 2, rc.z",h=a?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],m=r?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],f="",g="";i&&(o?f=`vec4 activation(vec4 a) {
+`;function SZ(e){let{inputs:t,backend:n}=e,{x:r,alpha:s}=t,a=q().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new up(mE,r.shape,s.shape):new $c(fE,r.shape,s.shape);return n.runWebGLProgram(a,[r,s],"float32")}var TZ={kernelName:Yo,backendName:"webgl",kernelFunc:SZ},tl="if (isnan(x)) return x;";function Ye({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:n,dtype:r}){return({inputs:s,backend:a})=>{let{x:o}=s,i=a,c=r||o.dtype;if(i.shouldExecuteOnCPU([o])&&n!=null){let p=i.texData.get(o.dataId),d=n(p.values,c);return i.makeTensorInfo(o.shape,c,d)}let u=q().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&t!=null,l;return u?l=new Ka(o.shape,t):l=new Ns(o.shape,e),i.runWebGLProgram(l,[o],c)}}function dn({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:n=!1,supportsComplex:r=!1,cpuKernelImpl:s,dtype:a}){return({inputs:o,backend:i})=>{let{a:c,b:u}=o,l=i;if(r&&c.dtype==="complex64"){let f=l.texData.get(c.dataId),m=l.texData.get(u.dataId),[g,b]=[[f.complexTensorInfos.real,m.complexTensorInfos.real],[f.complexTensorInfos.imag,m.complexTensorInfos.imag]].map(v=>{let[x,k]=v,S={dataId:x.dataId,dtype:x.dtype,shape:c.shape},C={dataId:k.dataId,dtype:k.dtype,shape:u.shape},E=new $c(e,c.shape,u.shape);return l.runWebGLProgram(E,[S,C],hr(x.dtype,k.dtype))}),y=Ea({inputs:{real:g,imag:b},backend:l});return l.disposeIntermediateTensorInfo(g),l.disposeIntermediateTensorInfo(b),y}let p=a||hr(c.dtype,u.dtype);if((c.dtype==="string"||u.dtype==="string"||l.shouldExecuteOnCPU([c,u]))&&s!=null){let f=l.texData.get(c.dataId).values,m=l.texData.get(u.dataId).values,g=c.dtype==="string"?N.fromUint8ToStringArray(f):f,b=c.dtype==="string"?N.fromUint8ToStringArray(m):m,[y,v]=s(c.shape,u.shape,g,b,p),x=l.makeTensorInfo(v,p),k=l.texData.get(x.dataId);return k.values=y,x}let d=q().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&t!=null,h;return d?h=new up(t,c.shape,u.shape,n):h=new $c(e,c.shape,u.shape),l.runWebGLProgram(h,[c,u],p)}}function od(e,t=!1){if(e==="linear")return t?sZ:Q9;if(e==="relu")return t?oZ:tZ;if(e==="elu")return t?aZ:eZ;if(e==="relu6")return t?iZ:nZ;if(e==="prelu")return t?mE:fE;if(e==="leakyrelu")return t?hE:pE;if(e==="sigmoid")return t?cZ:rZ;throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var gE=class{constructor(e,t,n,r=!1,s=!1,a=!1,o=null,i=!1,c=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=Dn(this.outputShape.length);let u=r?e[1]:e[2],l=Math.ceil(u/2),p=r?"i * 2, rc.y":"rc.y, i * 2",d=s?"rc.z, i * 2":"i * 2, rc.z",h=r?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],f=s?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],m="",g="";o&&(i?m=`vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
-          ${i}
-        }`:l?f=`vec4 activation(vec4 a) {
+          ${o}
+        }`:c?m=`vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
-          ${i}
-        }`:f=`vec4 activation(vec4 x) {
-          ${i}
-        }`,g="result = activation(result);");let y=s?"result += getBiasAtOutCoords();":"";s&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),l&&this.variableNames.push("leakyreluAlpha");let b="rc.x",x="rc.x";e[0]<t[0]?b=`int(min(float(rc.x), ${e[0]-1}.))`:t[0]<e[0]&&(x=`int(min(float(rc.x), ${t[0]-1}.))`),this.userCode=`
-      ${f}
+          ${o}
+        }`:m=`vec4 activation(vec4 x) {
+          ${o}
+        }`,g="result = activation(result);");let b=a?"result += getBiasAtOutCoords();":"";a&&this.variableNames.push("bias"),i&&this.variableNames.push("preluActivationWeights"),c&&this.variableNames.push("leakyreluAlpha");let y="rc.x",v="rc.x";e[0]<t[0]?y=`int(min(float(rc.x), ${e[0]-1}.))`:t[0]<e[0]&&(v=`int(min(float(rc.x), ${t[0]-1}.))`),this.userCode=`
+      ${m}
       // Don't use uniform for sharedDimensionPacked for performance.
-      const float sharedDimension = ${p}.0;
+      const float sharedDimension = ${l}.0;
 
       vec4 dot2x2ARowBCol(ivec3 rc) {
         vec4 result = vec4(0);
-        for (int i = 0; i < ${p}; i++) {
-          int batchA = ${b};
-          int batchB = ${x};
-          vec4 a = getMatrixA(batchA, ${d});
-          vec4 b = getMatrixB(batchB, ${c});
+        for (int i = 0; i < ${l}; i++) {
+          int batchA = ${y};
+          int batchB = ${v};
+          vec4 a = getMatrixA(batchA, ${p});
+          vec4 b = getMatrixB(batchB, ${d});
 
           // These swizzled products need to be separately added.
           // See: https://github.com/tensorflow/tfjs/issues/1735
-          result += (${h[0]} * ${m[0]});
-          result += (${h[1]} * ${m[1]});
+          result += (${h[0]} * ${f[0]});
+          result += (${h[1]} * ${f[1]});
         }
         return result;
       }
@@ -1338,13 +1338,13 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec3 rc = getOutputCoords();
         vec4 result = dot2x2ARowBCol(rc);
 
-        ${y}
+        ${b}
 
         ${g}
 
         setOutput(result);
       }
-    `}},bI={REAL:"return areal * breal - aimag * bimag;",IMAG:"return areal * bimag + aimag * breal;"},xI=class{constructor(e,t,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.userCode=`
+    `}},v1={REAL:"return areal * breal - aimag * bimag;",IMAG:"return areal * bimag + aimag * breal;"},x1=class{constructor(e,t,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=N.assertAndGetBroadcastShape(t,n),this.userCode=`
       float binaryOpComplex(
           float areal, float aimag, float breal, float bimag) {
         ${e}
@@ -1357,8 +1357,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float bimag = getBImagAtOutCoords();
         setOutput(binaryOpComplex(areal, aimag, breal, bimag));
       }
-    `}},vI="return a * b;";function o1(e){let{inputs:t,backend:n}=e,{a,b:r}=t,s=N.upcastType(a.dtype,r.dtype);if(a.dtype==="complex64"){let o=n.texData.get(a.dataId),l=n.texData.get(r.dataId),u=new xI(bI.REAL,a.shape,r.shape),p=new xI(bI.IMAG,a.shape,r.shape),d=[{dataId:o.complexTensorInfos.real.dataId,dtype:o.complexTensorInfos.real.dtype,shape:a.shape},{dataId:o.complexTensorInfos.imag.dataId,dtype:o.complexTensorInfos.imag.dtype,shape:a.shape},{dataId:l.complexTensorInfos.real.dataId,dtype:l.complexTensorInfos.real.dtype,shape:r.shape},{dataId:l.complexTensorInfos.imag.dataId,dtype:l.complexTensorInfos.imag.dtype,shape:r.shape}],c=n.runWebGLProgram(u,d,"float32"),h=n.runWebGLProgram(p,d,"float32"),m=_s({inputs:{real:c,imag:h},backend:n});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),m}if(n.shouldExecuteOnCPU([a,r])){let o=n.texData.get(a.dataId),l=n.texData.get(r.dataId),[u,p]=v7(a.shape,r.shape,o.values,l.values,s),d=n.makeTensorInfo(p,s),c=n.texData.get(d.dataId);return c.values=u,d}let i;return H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?i=new ud(vI,a.shape,r.shape):i=new $l(vI,a.shape,r.shape),n.runWebGLProgram(i,[a,r],s)}var IJ={kernelName:Gi,backendName:"webgl",kernelFunc:o1};function SJ(e,t,n){let a=[ci(e.shape),...di(e.shape)],r={dtype:e.dtype,shape:a,dataId:e.dataId},s=[ci(t),...di(t)],i=new lE(s,a),o=!0,l=[a],u=n.runWebGLProgram(i,[r],e.dtype,l,o);return{dataId:u.dataId,shape:t,dtype:u.dtype}}function de(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{shape:s}=a,i=n,o=v.sizeFromShape(r.shape),l=v.inferFromImplicitShape(s,o),u=v.sizeFromShape(l);v.assert(o===u,()=>`The new shape (${l}) has ${u} elements and the old shape (${r.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);let p=i.texData.get(r.dataId);return p.isPacked&&!sc(r.shape,l)&&!(p.texture!==null&&sc(p.shape,l))?SJ(r,l,i):(i.incRef(r.dataId),{dataId:r.dataId,shape:l,dtype:r.dtype})}var TJ={kernelName:vu,backendName:"webgl",kernelFunc:de},wI=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:a,inSize:r,outSize:s}=e;this.outputShape=[a,s];let i=Math.floor(n/4)*4,o=n%4,l="sumValue += dot(values, ones);";if(t!=null){let p=1/t;l=`sumValue += dot(values * ${v.isInt(p)?p.toPrecision(2):p}, ones);`}let u="";r%n>0&&(u=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+    `}},w1="return a * b;";function iI(e){let{inputs:t,backend:n}=e,{a:r,b:s}=t,a=N.upcastType(r.dtype,s.dtype);if(r.dtype==="complex64"){let i=n.texData.get(r.dataId),c=n.texData.get(s.dataId),u=new x1(v1.REAL,r.shape,s.shape),l=new x1(v1.IMAG,r.shape,s.shape),p=[{dataId:i.complexTensorInfos.real.dataId,dtype:i.complexTensorInfos.real.dtype,shape:r.shape},{dataId:i.complexTensorInfos.imag.dataId,dtype:i.complexTensorInfos.imag.dtype,shape:r.shape},{dataId:c.complexTensorInfos.real.dataId,dtype:c.complexTensorInfos.real.dtype,shape:s.shape},{dataId:c.complexTensorInfos.imag.dataId,dtype:c.complexTensorInfos.imag.dtype,shape:s.shape}],d=n.runWebGLProgram(u,p,"float32"),h=n.runWebGLProgram(l,p,"float32"),f=Ea({inputs:{real:d,imag:h},backend:n});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h),f}if(n.shouldExecuteOnCPU([r,s])){let i=n.texData.get(r.dataId),c=n.texData.get(s.dataId),[u,l]=k9(r.shape,s.shape,i.values,c.values,a),p=n.makeTensorInfo(l,a),d=n.texData.get(p.dataId);return d.values=u,p}let o;return q().getBool("WEBGL_PACK_BINARY_OPERATIONS")?o=new up(w1,r.shape,s.shape):o=new $c(w1,r.shape,s.shape),n.runWebGLProgram(o,[r,s],a)}var CZ={kernelName:qo,backendName:"webgl",kernelFunc:iI};function NZ(e,t,n){let r=[ho(e.shape),...fo(e.shape)],s={dtype:e.dtype,shape:r,dataId:e.dataId},a=[ho(t),...fo(t)],o=new lE(a,r),i=!0,c=[r],u=n.runWebGLProgram(o,[s],e.dtype,c,i);return{dataId:u.dataId,shape:t,dtype:u.dtype}}function he(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{shape:a}=r,o=n,i=w.sizeFromShape(s.shape),c=w.inferFromImplicitShape(a,i),u=w.sizeFromShape(c);w.assert(i===u,()=>`The new shape (${c}) has ${u} elements and the old shape (${s.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`);let l=o.texData.get(s.dataId);return l.isPacked&&!ad(s.shape,c)&&!(l.texture!==null&&ad(l.shape,c))?NZ(s,c,o):(o.incRef(s.dataId),{dataId:s.dataId,shape:c,dtype:s.dtype})}var _Z={kernelName:xu,backendName:"webgl",kernelFunc:he},I1=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:r,inSize:s,outSize:a}=e;this.outputShape=[r,a];let o=Math.floor(n/4)*4,i=n%4,c="sumValue += dot(values, ones);";if(t!=null){let l=1/t;c=`sumValue += dot(values * ${w.isInt(l)?l.toPrecision(2):l}, ones);`}let u="";s%n>0&&(u=`
+        if (inIdx < 0 || inIdx >= ${s}) {
           return 0.0;
         }
       `),this.userCode=`
@@ -1377,7 +1377,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         float sumValue = 0.0;
 
-        for (int i = 0; i < ${i}; i += 4) {
+        for (int i = 0; i < ${o}; i += 4) {
           int inIdx = inOffset + i;
           vec4 values = vec4(
             getValue(batch, inIdx),
@@ -1386,60 +1386,60 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             getValue(batch, inIdx + 3)
           );
 
-          ${l}
+          ${c}
         }
 
-        int inIdx = inOffset + ${i};
-        if (${o===1}) {
+        int inIdx = inOffset + ${o};
+        if (${i===1}) {
           vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);
 
-          ${l}
-        } else if (${o===2}) {
+          ${c}
+        } else if (${i===2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1), 0.0, 0.0);
 
-          ${l}
-        } else if (${o===3}) {
+          ${c}
+        } else if (${i===3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2), 0.0);
 
-          ${l}
+          ${c}
         }
         setOutput(sumValue);
       }
-    `}},NJ=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:a,inSize:r,outSize:s}=e;this.outputShape=[a,s];let i="0.0",o="";t==="prod"?i="1.0":t==="min"?(i="1.0 / 1e-20",o="min"):t==="max"&&(i="-1.0 / 1e-20",o="max");let l=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="sum"?l="sumValue":t==="prod"?l="prodValue":t==="all"?l="allValue":t==="any"&&(l="anyValue");let u=Math.floor(n/4)*4,p=n%4,d=`
+    `}},EZ=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:r,inSize:s,outSize:a}=e;this.outputShape=[r,a];let o="0.0",i="";t==="prod"?o="1.0":t==="min"?(o="1.0 / 1e-20",i="min"):t==="max"&&(o="-1.0 / 1e-20",i="max");let c=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="sum"?c="sumValue":t==="prod"?c="prodValue":t==="all"?c="allValue":t==="any"&&(c="anyValue");let u=Math.floor(n/4)*4,l=n%4,p=`
       if (${t==="sum"}) {
         sumValue += dot(values, ones);
       } else if (${t==="prod"}) {
         vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);
         prodValue *= tmp[0] * tmp[1];
       } else {
-        minMaxValue = ${o}(values, minMaxValue);
+        minMaxValue = ${i}(values, minMaxValue);
         if (${t==="min"} || ${t==="max"}) {
-          minMaxValue = ${o}(values, minMaxValue);
+          minMaxValue = ${i}(values, minMaxValue);
           bvec4 isNaN = isnan(values);
           if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {
             minMaxValue = vec4(NAN);
           }
         }
       }
-    `,c="vec4";t==="all"?(i="1.0",d=`
+    `,d="vec4";t==="all"?(o="1.0",p=`
         bool reducedAllValue = all(values);
         float floatedReducedAllValue = float(reducedAllValue);
         allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);
-      `,c="bvec4"):t==="any"&&(i="0.0",d=`
+      `,d="bvec4"):t==="any"&&(o="0.0",p=`
         bool reducedAnyValue = any(values);
         float floatedReducedAnyValue = float(reducedAnyValue);
         anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);
-      `,c="bvec4");let h="";r%n>0&&(h=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+      `,d="bvec4");let h="";s%n>0&&(h=`
+        if (inIdx < 0 || inIdx >= ${s}) {
           return initializationValue;
         }
       `),this.userCode=`
-      const float initializationValue = ${i};
+      const float initializationValue = ${o};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float getValue(int batch, int inIdx) {
@@ -1453,7 +1453,7 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         int outIdx = coords[1];
         int inOffset = outIdx * ${n};
 
-        vec4 minMaxValue = vec4(${i});
+        vec4 minMaxValue = vec4(${o});
         float prodValue = 1.0;
         float sumValue = 0.0;
         float allValue = 1.0;
@@ -1461,164 +1461,164 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         for (int i = 0; i < ${u}; i += 4) {
           int inIdx = inOffset + i;
-          ${c} values = ${c}(
+          ${d} values = ${d}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             getValue(batch, inIdx + 3)
           );
 
-          ${d}
+          ${p}
         }
 
         int inIdx = inOffset + ${u};
-        if (${p===1}) {
-          ${c} values = ${c}(
+        if (${l===1}) {
+          ${d} values = ${d}(
             getValue(batch, inIdx),
             initializationValue,
             initializationValue,
             initializationValue
           );
 
-          ${d}
-        } else if (${p===2}) {
-          ${c} values = ${c}(
+          ${p}
+        } else if (${l===2}) {
+          ${d} values = ${d}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             initializationValue,
             initializationValue
           );
 
-          ${d}
-        } else if (${p===3}) {
-          ${c} values = ${c}(
+          ${p}
+        } else if (${l===3}) {
+          ${d} values = ${d}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             initializationValue
           );
 
-          ${d}
+          ${p}
         }
-        setOutput(${l});
+        setOutput(${c});
       }
-    `}};function CJ(e){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let n=t.length?t[t.length-1].outSize:e[1],a=N.computeOptimalWindowSize(n);t.push({inSize:n,windowSize:a,outSize:Math.ceil(n/a)})}return t}function Io(e,t,n,a){let r=CJ(e.shape),s=e;for(let i=0;i<r.length;i++){let{inSize:o,windowSize:l,outSize:u}=r[i],p,d;n==="mean"?p=i===0?new wI({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},o):new wI({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u}):p=new NJ({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},n),d=s,s=a.runWebGLProgram(p,[s],t),d.dataId!==e.dataId&&a.disposeIntermediateTensorInfo(d)}return s}var _J=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let s=0;s<n.length;s++)n[s]=e[t[s]];this.outputShape=n,this.rank=n.length;let a=gt(this.rank),r=EJ(t);this.userCode=`
+    `}};function AZ(e){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let n=t.length?t[t.length-1].outSize:e[1],r=N.computeOptimalWindowSize(n);t.push({inSize:n,windowSize:r,outSize:Math.ceil(n/r)})}return t}function Si(e,t,n,r){let s=AZ(e.shape),a=e;for(let o=0;o<s.length;o++){let{inSize:i,windowSize:c,outSize:u}=s[o],l,p;n==="mean"?l=o===0?new I1({windowSize:c,inSize:i,batchSize:e.shape[0],outSize:u},i):new I1({windowSize:c,inSize:i,batchSize:e.shape[0],outSize:u}):l=new EZ({windowSize:c,inSize:i,batchSize:e.shape[0],outSize:u},n),p=a,a=r.runWebGLProgram(l,[a],t),p.dataId!==e.dataId&&r.disposeIntermediateTensorInfo(p)}return a}var $Z=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let a=0;a<n.length;a++)n[a]=e[t[a]];this.outputShape=n,this.rank=n.length;let r=mt(this.rank),s=DZ(t);this.userCode=`
     void main() {
-      ${a} resRC = getOutputCoords();
-      setOutput(getA(${r}));
+      ${r} resRC = getOutputCoords();
+      setOutput(getA(${s}));
     }
-    `}};function EJ(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],a=new Array(t);for(let r=0;r<e.length;r++)a[e[r]]=n[r];return a.join()}var AJ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(e.length);for(let u=0;u<n.length;u++)n[u]=e[t[u]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let a=gt(this.rank),r=oE("rc",this.rank),s=new Array(this.rank);for(let u=0;u<t.length;u++)s[t[u]]=r[u];let i=`vec2(${s.slice(-2).join()})`,o=`++${r[this.rank-1]} < ${n[this.rank-1]}`,l=`getChannel(getA(${s.join()}), ${i})`;this.userCode=`
+    `}};function DZ(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],r=new Array(t);for(let s=0;s<e.length;s++)r[e[s]]=n[s];return r.join()}var FZ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(e.length);for(let u=0;u<n.length;u++)n[u]=e[t[u]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let r=mt(this.rank),s=uE("rc",this.rank),a=new Array(this.rank);for(let u=0;u<t.length;u++)a[t[u]]=s[u];let o=`vec2(${a.slice(-2).join()})`,i=`++${s[this.rank-1]} < ${n[this.rank-1]}`,c=`getChannel(getA(${a.join()}), ${o})`;this.userCode=`
     void main() {
-      ${a} rc = getOutputCoords();
+      ${r} rc = getOutputCoords();
       vec4 result = vec4(0.);
-      result[0] = ${l};
-      if(${o}) {
-        result[1] = ${l};
+      result[0] = ${c};
+      if(${i}) {
+        result[1] = ${c};
       }
-      --${r[this.rank-1]};
-      if(++${r[this.rank-2]} < ${n[this.rank-2]}) {
-        result[2] = ${l};
-        if(${o}) {
-          result[3] = ${l};
+      --${s[this.rank-1]};
+      if(++${s[this.rank-2]} < ${n[this.rank-2]}) {
+        result[2] = ${c};
+        if(${i}) {
+          result[3] = ${c};
         }
       }
       setOutput(result);
     }
-    `}};function Yf(e,t,n){let a=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new AJ(e.shape,t):new _J(e.shape,t);return n.runWebGLProgram(a,[e],e.dtype)}function $J(e,t,n,a){let r=t,s=e.shape.length,i=v.parseAxisParam(r,e.shape),o=i,l=N.getAxesPermutation(o,s),u=l!=null,p=e;u&&(p=Yf(e,l,a),o=N.getInnerMostAxes(o.length,s)),N.assertAxesAreInnerMostDims("sum",o,s);let[d,c]=N.computeOutAndReduceShapes(p.shape,o),h=d;n&&(h=N.expandShapeToKeepDim(d,i));let m=v.sizeFromShape(c),f=v.sizeFromShape(e.shape)/m,g=de({inputs:{x:p},attrs:{shape:[f,m]},backend:a}),y=Um(e.dtype),b=Io(g,y,"sum",a),x=de({inputs:{x:b},attrs:{shape:h},backend:a});return a.disposeIntermediateTensorInfo(g),a.disposeIntermediateTensorInfo(b),u&&a.disposeIntermediateTensorInfo(p),x}function Zf(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a;return $J(r,s,i,n)}var FJ={kernelName:io,backendName:"webgl",kernelFunc:Zf};function Sn(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{perm:s}=a,i=n,o=r.shape.length,l=new Array(o);for(let p=0;p<l.length;p++)l[p]=r.shape[s[p]];let u;if(i.shouldExecuteOnCPU([r])){let p=i.texData.get(r.dataId).values,d=s1(p,r.shape,r.dtype,s,l);u=i.makeTensorInfo(l,r.dtype);let c=i.texData.get(u.dataId);c.values=d}else u=Yf(r,s,i);return u}var DJ={kernelName:_r,backendName:"webgl",kernelFunc:Sn},fE=1e3;function nm({a:e,b:t,transposeA:n,transposeB:a,backend:r,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:l=null}){let u=e.shape.length,p=t.shape.length,d=n?e.shape[u-2]:e.shape[u-1],c=a?t.shape[p-1]:t.shape[p-2],h=n?e.shape[u-1]:e.shape[u-2],m=a?t.shape[p-2]:t.shape[p-1],f=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=v.sizeFromShape(f),b=v.sizeFromShape(g),x=Ou.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([h,m]);v.assert(d===c,()=>`Error in matMul: inner shapes (${d}) and (${c}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${a} must match.`);let w=n?[y,d,h]:[y,h,d],I=a?[b,m,c]:[b,c,m],T=de({inputs:{x:e},backend:r,attrs:{shape:w}}),C=de({inputs:{x:t},backend:r,attrs:{shape:I}}),E=[T,C],A=Math.max(y,b),R=n?T.shape[1]:T.shape[2],F=s!=null,S=i!=null,M=l==="leakyrelu",B=l!=null?ic(l,!0):null,U=F||S||M||B!=null,G;if((h===1||m===1)&&R>fE&&U===!1){let K=T,Z=C;n&&(K=Sn({inputs:{x:T},backend:r,attrs:{perm:[0,2,1]}}),E.push(K)),a&&(Z=Sn({inputs:{x:C},backend:r,attrs:{perm:[0,2,1]}}),E.push(Z));let Q=m!==1,ee=m===1,ae=K;Q&&(ae=de({inputs:{x:K},backend:r,attrs:{shape:[A,R,1]}}),E.push(ae));let te=m===1?2:1,le=Z;ee&&(le=de({inputs:{x:Z},backend:r,attrs:{shape:[A,1,R]}}),E.push(le));let ie=o1({inputs:{a:ae,b:le},backend:r});G=Zf({inputs:{x:ie},backend:r,attrs:{axis:te,keepDims:!0}}),E.push(ie)}else{let K=fa(e.dtype,t.dtype),Z=new mE(w,I,[A,h,m],n,a,F,B,S,M),Q=[T,C];if(s!=null&&Q.push(s),S&&Q.push(i),M){let ee=r.makeTensorInfo([],"float32",v.createScalarValue(o,"float32"));Q.push(ee),E.push(ee)}G=r.runWebGLProgram(Z,Q,K)}let q=de({inputs:{x:G},backend:r,attrs:{shape:x}});E.push(G);for(let K of E)r.disposeIntermediateTensorInfo(K);return q}function RJ(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a;return nm({a:r,b:s,transposeA:l,transposeB:u,backend:n,bias:i,preluActivationWeights:o,leakyreluAlpha:d,activation:p})}var MJ={kernelName:Qs,backendName:"webgl",kernelFunc:RJ},kI="return abs(x);";function PJ(e){let{inputs:t,backend:n}=e,{x:a}=t;if(n.shouldExecuteOnCPU([a])&&a.dtype!=="complex64"){let s=n.texData.get(a.dataId),i=sE(s.values);return n.makeTensorInfo(a.shape,a.dtype,i)}let r;return H().getBool("WEBGL_PACK_UNARY_OPERATIONS")?r=new qs(a.shape,kI):r=new Cr(a.shape,kI),n.runWebGLProgram(r,[a],a.dtype)}var OJ={kernelName:Dl,backendName:"webgl",kernelFunc:PJ},LJ=Ma+`
+    `}};function Zm(e,t,n){let r=q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new FZ(e.shape,t):new $Z(e.shape,t);return n.runWebGLProgram(r,[e],e.dtype)}function RZ(e,t,n,r){let s=t,a=e.shape.length,o=w.parseAxisParam(s,e.shape),i=o,c=N.getAxesPermutation(i,a),u=c!=null,l=e;u&&(l=Zm(e,c,r),i=N.getInnerMostAxes(i.length,a)),N.assertAxesAreInnerMostDims("sum",i,a);let[p,d]=N.computeOutAndReduceShapes(l.shape,i),h=p;n&&(h=N.expandShapeToKeepDim(p,o));let f=w.sizeFromShape(d),g=w.sizeFromShape(e.shape)/f,b=he({inputs:{x:l},attrs:{shape:[g,f]},backend:r}),y=Hf(e.dtype),v=Si(b,y,"sum",r),x=he({inputs:{x:v},attrs:{shape:h},backend:r});return r.disposeIntermediateTensorInfo(b),r.disposeIntermediateTensorInfo(v),u&&r.disposeIntermediateTensorInfo(l),x}function Jm(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r;return RZ(s,a,o,n)}var PZ={kernelName:ci,backendName:"webgl",kernelFunc:Jm};function Nn(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{perm:a}=r,o=n,i=s.shape.length,c=new Array(i);for(let l=0;l<c.length;l++)c[l]=s.shape[a[l]];let u;if(o.shouldExecuteOnCPU([s])){let p=o.texData.get(s.dataId).values,d=aI(p,s.shape,s.dtype,a,c);u=o.makeTensorInfo(c,s.dtype);let h=o.texData.get(u.dataId);h.values=d}else u=Zm(s,a,o);return u}var OZ={kernelName:_s,backendName:"webgl",kernelFunc:Nn},bE=1e3;function rf({a:e,b:t,transposeA:n,transposeB:r,backend:s,bias:a=null,preluActivationWeights:o=null,leakyreluAlpha:i=0,activation:c=null}){let u=e.shape.length,l=t.shape.length,p=n?e.shape[u-2]:e.shape[u-1],d=r?t.shape[l-1]:t.shape[l-2],h=n?e.shape[u-1]:e.shape[u-2],f=r?t.shape[l-2]:t.shape[l-1],m=e.shape.slice(0,-2),g=t.shape.slice(0,-2),b=w.sizeFromShape(m),y=w.sizeFromShape(g),x=Mu.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([h,f]);w.assert(p===d,()=>`Error in matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${r} must match.`);let k=n?[b,p,h]:[b,h,p],S=r?[y,f,d]:[y,d,f],C=he({inputs:{x:e},backend:s,attrs:{shape:k}}),E=he({inputs:{x:t},backend:s,attrs:{shape:S}}),$=[C,E],F=Math.max(b,y),A=n?C.shape[1]:C.shape[2],R=a!=null,T=o!=null,L=c==="leakyrelu",V=c!=null?od(c,!0):null,G=R||T||L||V!=null,j;if((h===1||f===1)&&A>bE&&G===!1){let Z=C,J=E;n&&(Z=Nn({inputs:{x:C},backend:s,attrs:{perm:[0,2,1]}}),$.push(Z)),r&&(J=Nn({inputs:{x:E},backend:s,attrs:{perm:[0,2,1]}}),$.push(J));let ee=f!==1,re=f===1,te=Z;ee&&(te=he({inputs:{x:Z},backend:s,attrs:{shape:[F,A,1]}}),$.push(te));let ie=f===1?2:1,ne=J;re&&(ne=he({inputs:{x:J},backend:s,attrs:{shape:[F,1,A]}}),$.push(ne));let le=iI({inputs:{a:te,b:ne},backend:s});j=Jm({inputs:{x:le},backend:s,attrs:{axis:ie,keepDims:!0}}),$.push(le)}else{let Z=hr(e.dtype,t.dtype),J=new gE(k,S,[F,h,f],n,r,R,V,T,L),ee=[C,E];if(a!=null&&ee.push(a),T&&ee.push(o),L){let re=s.makeTensorInfo([],"float32",w.createScalarValue(i,"float32"));ee.push(re),$.push(re)}j=s.runWebGLProgram(J,ee,Z)}let H=he({inputs:{x:j},backend:s,attrs:{shape:x}});$.push(j);for(let Z of $)s.disposeIntermediateTensorInfo(Z);return H}function MZ(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a,bias:o,preluActivationWeights:i}=t,{transposeA:c,transposeB:u,activation:l,leakyreluAlpha:p}=r;return rf({a:s,b:a,transposeA:c,transposeB:u,backend:n,bias:o,preluActivationWeights:i,leakyreluAlpha:p,activation:l})}var LZ={kernelName:eo,backendName:"webgl",kernelFunc:MZ},k1="return abs(x);";function zZ(e){let{inputs:t,backend:n}=e,{x:r}=t;if(n.shouldExecuteOnCPU([r])&&r.dtype!=="complex64"){let a=n.texData.get(r.dataId),o=iE(a.values);return n.makeTensorInfo(r.shape,r.dtype,o)}let s;return q().getBool("WEBGL_PACK_UNARY_OPERATIONS")?s=new Ka(r.shape,k1):s=new Ns(r.shape,k1),n.runWebGLProgram(s,[r],r.dtype)}var BZ={kernelName:Fc,backendName:"webgl",kernelFunc:zZ},WZ=Or+`
   if (abs(x) > 1.) {
     return NAN;
   }
   return acos(x);
-`,zJ=Ye({opSnippet:LJ}),WJ={kernelName:Rl,backendName:"webgl",kernelFunc:zJ},BJ=Ma+`
+`,VZ=Ye({opSnippet:WZ}),UZ={kernelName:Rc,backendName:"webgl",kernelFunc:VZ},GZ=Or+`
   if (x < 1.0) return NAN;
-return log(x + sqrt(x * x - 1.0));`,VJ=Ye({opSnippet:BJ}),UJ={kernelName:Ml,backendName:"webgl",kernelFunc:VJ},II="return a + b;",GJ=cn({opSnippet:II,packedOpSnippet:II,supportsComplex:!0,cpuKernelImpl:e7}),HJ={kernelName:gs,backendName:"webgl",kernelFunc:GJ},jJ=class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((r,s)=>`T${s}`);let n=[];this.variableNames.forEach(r=>{n.push(`float v${r} = get${r}AtOutCoords();`)});let a=this.variableNames.map(r=>`v${r}`).join(" + ");this.userCode=`
+return log(x + sqrt(x * x - 1.0));`,HZ=Ye({opSnippet:GZ}),qZ={kernelName:Pc,backendName:"webgl",kernelFunc:HZ},S1="return a + b;",jZ=dn({opSnippet:S1,packedOpSnippet:S1,supportsComplex:!0,cpuKernelImpl:r9}),KZ={kernelName:ba,backendName:"webgl",kernelFunc:jZ},XZ=class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((s,a)=>`T${a}`);let n=[];this.variableNames.forEach(s=>{n.push(`float v${s} = get${s}AtOutCoords();`)});let r=this.variableNames.map(s=>`v${s}`).join(" + ");this.userCode=`
       void main() {
         ${n.join(`
         `)}
 
-        float result = ${a};
+        float result = ${r};
         setOutput(result);
       }
-    `}},qJ=class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((r,s)=>`T${s}`);let n=[];this.variableNames.forEach(r=>{n.push(`vec4 v${r} = get${r}AtOutCoords();`)});let a=this.variableNames.map(r=>`v${r}`).join(" + ");this.userCode=`
+    `}},YZ=class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((s,a)=>`T${a}`);let n=[];this.variableNames.forEach(s=>{n.push(`vec4 v${s} = get${s}AtOutCoords();`)});let r=this.variableNames.map(s=>`v${s}`).join(" + ");this.userCode=`
       void main() {
         ${n.join(`
         `)}
 
-        vec4 result = ${a};
+        vec4 result = ${r};
         setOutput(result);
       }
-    `}};function Nh(e){let{inputs:t,backend:n}=e,a=t;if(a.length===1)return aa({inputs:{x:a[0]},backend:n});if(a.length>H().get("WEBGL_MAX_TEXTURES_IN_SHADER")){let o=Math.floor(a.length/2),l=Nh({inputs:a.slice(0,o),backend:n}),u=Nh({inputs:a.slice(o),backend:n});return Nh({inputs:[l,u],backend:n})}let r=a.map(o=>o.dtype).reduce((o,l)=>fa(o,l)),s=a.map(o=>o.shape),i=H().getBool("WEBGL_PACK")?new qJ(a[0].shape,s):new jJ(a[0].shape,s);return n.runWebGLProgram(i,a,r)}var KJ={kernelName:fi,backendName:"webgl",kernelFunc:Nh};function XJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=Sn({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,o)),N.assertAxesAreInnerMostDims("all",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=Io(f,f.dtype,"all",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var YJ={kernelName:Pl,backendName:"webgl",kernelFunc:XJ};function ZJ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=Sn({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,o)),N.assertAxesAreInnerMostDims("any",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=Io(f,f.dtype,"any",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var JJ={kernelName:Ol,backendName:"webgl",kernelFunc:ZJ},QJ=class{constructor(e,t,n){this.variableNames=["A"];let{windowSize:a,batchSize:r,outSize:s}=e;n||this.variableNames.push("bestIndicesA"),this.outputShape=[r,s];let i=t==="max"?">":"<",o=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`
+    `}};function Nh(e){let{inputs:t,backend:n}=e,r=t;if(r.length===1)return sr({inputs:{x:r[0]},backend:n});if(r.length>q().get("WEBGL_MAX_TEXTURES_IN_SHADER")){let c=Math.floor(r.length/2),u=Nh({inputs:r.slice(0,c),backend:n}),l=Nh({inputs:r.slice(c),backend:n});return Nh({inputs:[u,l],backend:n})}let s=r.map(c=>c.dtype).reduce((c,u)=>hr(c,u)),a=r.map(c=>c.shape),i=q().getBool("WEBGL_PACK")?new YZ(r[0].shape,a):new XZ(r[0].shape,a);return n.runWebGLProgram(i,r,s)}var ZZ={kernelName:bo,backendName:"webgl",kernelFunc:Nh};function JZ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r,i=s.shape.length,c=w.parseAxisParam(a,s.shape),u=c,l=N.getAxesPermutation(u,i),p=s;l!=null&&(p=Nn({inputs:{x:s},backend:n,attrs:{perm:l}}),u=N.getInnerMostAxes(u.length,i)),N.assertAxesAreInnerMostDims("all",u,i);let[d,h]=N.computeOutAndReduceShapes(p.shape,u),f=w.sizeFromShape(h),m=he({inputs:{x:p},backend:n,attrs:{shape:[-1,f]}}),g=Si(m,m.dtype,"all",n),b;if(o){let y=N.expandShapeToKeepDim(d,c);b=he({inputs:{x:g},backend:n,attrs:{shape:y}})}else b=he({inputs:{x:g},backend:n,attrs:{shape:d}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),l!=null&&n.disposeIntermediateTensorInfo(p),b}var QZ={kernelName:Oc,backendName:"webgl",kernelFunc:JZ};function eJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r,i=s.shape.length,c=w.parseAxisParam(a,s.shape),u=c,l=N.getAxesPermutation(u,i),p=s;l!=null&&(p=Nn({inputs:{x:s},backend:n,attrs:{perm:l}}),u=N.getInnerMostAxes(u.length,i)),N.assertAxesAreInnerMostDims("any",u,i);let[d,h]=N.computeOutAndReduceShapes(p.shape,u),f=w.sizeFromShape(h),m=he({inputs:{x:p},backend:n,attrs:{shape:[-1,f]}}),g=Si(m,m.dtype,"any",n),b;if(o){let y=N.expandShapeToKeepDim(d,c);b=he({inputs:{x:g},backend:n,attrs:{shape:y}})}else b=he({inputs:{x:g},backend:n,attrs:{shape:d}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),l!=null&&n.disposeIntermediateTensorInfo(p),b}var tJ={kernelName:Mc,backendName:"webgl",kernelFunc:eJ},nJ=class{constructor(e,t,n){this.variableNames=["A"];let{windowSize:r,batchSize:s,outSize:a}=e;n||this.variableNames.push("bestIndicesA"),this.outputShape=[s,a];let o=t==="max"?">":"<",i=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${a};
+        int inOffset = outIdx * ${r};
 
         int bestIndex = inOffset;
         float bestValue = getA(batch, bestIndex);
 
-        for (int i = 0; i < ${a}; i++) {
-          int inIdx = ${o};
+        for (int i = 0; i < ${r}; i++) {
+          int inIdx = ${i};
           float candidate = getA(batch, inIdx);
-          if (candidate ${i} bestValue) {
+          if (candidate ${o} bestValue) {
             bestValue = candidate;
             bestIndex = inIdx;
           }
         }
         setOutput(float(bestIndex));
       }
-    `}},e9=class{constructor(e,t,n,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,v.assert(e.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let r=e[e.length-1],s=Math.ceil(r/t);this.outputShape=e.slice(0,-1),s>1&&this.outputShape.push(s),a||this.variableNames.push("bestIndicesA");let i=this.outputShape,o=i.length,l=gt(o),u=kn("coords",o),p,d;if(s===1){d=o+1;let C=gt(d);p=`
+    `}},rJ=class{constructor(e,t,n,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,w.assert(e.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let s=e[e.length-1],a=Math.ceil(s/t);this.outputShape=e.slice(0,-1),a>1&&this.outputShape.push(a),r||this.variableNames.push("bestIndicesA");let o=this.outputShape,i=o.length,c=mt(i),u=Tn("coords",i),l,p;if(a===1){p=i+1;let C=mt(p);l=`
         ${C} sourceLocR = ${C}(${u.join()}, 0);
-        ++${u[o-1]};
+        ++${u[i-1]};
         ${C} sourceLocG = ${C}(${u.join()}, 0);
-        ++${u[o-2]};
+        ++${u[i-2]};
         ${C} sourceLocA = ${C}(${u.join()}, 0);
-        --${u[o-1]};
+        --${u[i-1]};
         ${C} sourceLocB = ${C}(${u.join()}, 0);
-        --${u[o-2]};`}else d=o,p=`
-        ${l} sourceLocR = coords;
-        ++${u[o-1]};
-        ${l} sourceLocG = coords;
-        ++${u[o-2]};
-        ${l} sourceLocA = coords;
-        --${u[o-1]};
-        ${l} sourceLocB = coords;
-        --${u[o-2]};`;let c=["x","y","z","w","u","v"].slice(0,d),h="."+c[d-1],m=c.map(C=>"int "+C),f=kn("sourceLocR",d-1).concat("inIdx.r"),g=kn("sourceLocG",d-1).concat("inIdx.g"),y=kn("sourceLocB",d-1).concat("inIdx.b"),b=kn("sourceLocA",d-1).concat("inIdx.a"),x=n==="max"?"greaterThan":"lessThan",w=a?"":`
-          inIdx = round(vec4(getBestIndicesAChannel(${f.join()}),
+        --${u[i-2]};`}else p=i,l=`
+        ${c} sourceLocR = coords;
+        ++${u[i-1]};
+        ${c} sourceLocG = coords;
+        ++${u[i-2]};
+        ${c} sourceLocA = coords;
+        --${u[i-1]};
+        ${c} sourceLocB = coords;
+        --${u[i-2]};`;let d=["x","y","z","w","u","v"].slice(0,p),h="."+d[p-1],f=d.map(C=>"int "+C),m=Tn("sourceLocR",p-1).concat("inIdx.r"),g=Tn("sourceLocG",p-1).concat("inIdx.g"),b=Tn("sourceLocB",p-1).concat("inIdx.b"),y=Tn("sourceLocA",p-1).concat("inIdx.a"),v=n==="max"?"greaterThan":"lessThan",x=r?"":`
+          inIdx = round(vec4(getBestIndicesAChannel(${m.join()}),
                              getBestIndicesAChannel(${g.join()}),
-                             getBestIndicesAChannel(${y.join()}),
-                             getBestIndicesAChannel(${b.join()})));`,I=`vec4(
-            getAChannel(${f.join()}),
+                             getBestIndicesAChannel(${b.join()}),
+                             getBestIndicesAChannel(${y.join()})));`,k=`vec4(
+            getAChannel(${m.join()}),
             hasNextCol ? getAChannel(${g.join()}) : 0.,
-            hasNextRow ? getAChannel(${y.join()}) : 0.,
-            hasNextRow && hasNextCol ? getAChannel(${b.join()}) : 0.)`,T=a?"":`
-      float getBestIndicesAChannel(${m.join()}) {
-        return getChannel(getBestIndicesA(${c.join()}),
-                                          vec2(${c.slice(-2).join()}));
+            hasNextRow ? getAChannel(${b.join()}) : 0.,
+            hasNextRow && hasNextCol ? getAChannel(${y.join()}) : 0.)`,S=r?"":`
+      float getBestIndicesAChannel(${f.join()}) {
+        return getChannel(getBestIndicesA(${d.join()}),
+                                          vec2(${d.slice(-2).join()}));
       }`;this.userCode=`
-      float getAChannel(${m.join()}) {
-        return getChannel(getA(${c.join()}),
-                               vec2(${c.slice(-2).join()}));
+      float getAChannel(${f.join()}) {
+        return getChannel(getA(${d.join()}),
+                               vec2(${d.slice(-2).join()}));
       }
-      ${T}
+      ${S}
       void main() {
-        ${l} coords = getOutputCoords();
-        bool hasNextCol = ${u[o-1]} < ${i[o-1]-1};
-        bool hasNextRow = ${u[o-2]} < ${i[o-2]-1};
-        ${p}
+        ${c} coords = getOutputCoords();
+        bool hasNextCol = ${u[i-1]} < ${o[i-1]-1};
+        bool hasNextRow = ${u[i-2]} < ${o[i-2]-1};
+        ${l}
         ivec4 srcIdx = ivec4(sourceLocR${h}, sourceLocG${h},
           sourceLocB${h}, sourceLocA${h}) * ${t};
         ivec4 inIdx = srcIdx;
         vec4 bestIndex = vec4(inIdx);
-        vec4 bestValue = ${I};
+        vec4 bestValue = ${k};
 
         for (int i = 0; i < ${t}; i++) {
           inIdx = srcIdx;
-          ${w}
-          vec4 candidate = ${I};
+          ${x}
+          vec4 candidate = ${k};
           bvec4 nan = isnan(candidate);
           bvec4 replace = bvec4(
-            vec4(${x}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
+            vec4(${v}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
 
           bestValue = vec4(replace.x  ? candidate.x : bestValue.x,
                            replace.y  ? candidate.y : bestValue.y,
@@ -1629,27 +1629,27 @@ return log(x + sqrt(x * x - 1.0));`,VJ=Ye({opSnippet:BJ}),UJ={kernelName:Ml,back
         }
         setOutput(bestIndex);
       }
-    `}};function gE(e,t,n,a=null){let r=t.shape[0],s=t.shape[1];a!=null&&(r=a.shape[0],s=a.shape[1]);let i=N.computeOptimalWindowSize(s),o={windowSize:i,inSize:s,batchSize:r,outSize:Math.ceil(s/i)},l=new QJ(o,n,a==null),u=[t];a!=null&&u.push(a);let p=e.runWebGLProgram(l,u,"int32");if(p.shape[1]===1)return p;let d=gE(e,t,n,p);return e.disposeIntermediateTensorInfo(p),d}function yE(e,t,n,a=null){let r=a!=null?a.shape:t.shape,s=r[r.length-1],i=N.computeOptimalWindowSize(s),o=new e9(r,i,n,a==null),l=a==null?[t]:[t,a],u=e.runWebGLProgram(o,l,"int32");if(u.shape.length===t.shape.length){let p=yE(e,t,n,u);return e.disposeIntermediateTensorInfo(u),p}return u}function bE(e,t,n,a){let r=[n];if(N.assertAxesAreInnerMostDims("arg"+a.charAt(0).toUpperCase()+a.slice(1),r,t.shape.length),!H().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let s=[],i=e.texData.get(t.dataId),o=i!==null&&i.isPacked,l=t;o&&(l=e.unpackTensor(t),s.push(l));let[u,p]=N.computeOutAndReduceShapes(l.shape,r),d=v.sizeFromShape(p),c=de({inputs:{x:l},backend:e,attrs:{shape:[-1,d]}});s.push(c);let h=gE(e,c,a);s.push(h);let m=de({inputs:{x:h},backend:e,attrs:{shape:u}});return s.forEach(f=>e.disposeIntermediateTensorInfo(f)),m}return yE(e,t,a)}function t9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a,i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Sn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),N.assertAxesAreInnerMostDims("argMax",[i[0]],l.shape.length);let p=bE(n,l,i[0],"max");return u.forEach(d=>n.disposeIntermediateTensorInfo(d)),p}var n9={kernelName:gi,backendName:"webgl",kernelFunc:t9};function a9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s}=a,i=v.parseAxisParam(s,r.shape),o=N.getAxesPermutation(i,r.shape.length),l=r,u=[];o!=null&&(l=Sn({inputs:{x:r},backend:n,attrs:{perm:o}}),u.push(l),i=N.getInnerMostAxes(i.length,l.shape.length)),N.assertAxesAreInnerMostDims("argMin",[i[0]],l.shape.length);let p=bE(n,l,i[0],"min");return u.forEach(d=>n.disposeIntermediateTensorInfo(d)),p}var r9={kernelName:cc,backendName:"webgl",kernelFunc:a9},s9=Ma+`
+    `}};function yE(e,t,n,r=null){let s=t.shape[0],a=t.shape[1];r!=null&&(s=r.shape[0],a=r.shape[1]);let o=N.computeOptimalWindowSize(a),i={windowSize:o,inSize:a,batchSize:s,outSize:Math.ceil(a/o)},c=new nJ(i,n,r==null),u=[t];r!=null&&u.push(r);let l=e.runWebGLProgram(c,u,"int32");if(l.shape[1]===1)return l;let p=yE(e,t,n,l);return e.disposeIntermediateTensorInfo(l),p}function vE(e,t,n,r=null){let s=r!=null?r.shape:t.shape,a=s[s.length-1],o=N.computeOptimalWindowSize(a),i=new rJ(s,o,n,r==null),c=r==null?[t]:[t,r],u=e.runWebGLProgram(i,c,"int32");if(u.shape.length===t.shape.length){let l=vE(e,t,n,u);return e.disposeIntermediateTensorInfo(u),l}return u}function xE(e,t,n,r){let s=[n];if(N.assertAxesAreInnerMostDims("arg"+r.charAt(0).toUpperCase()+r.slice(1),s,t.shape.length),!q().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let a=[],o=e.texData.get(t.dataId),i=o!==null&&o.isPacked,c=t;i&&(c=e.unpackTensor(t),a.push(c));let[u,l]=N.computeOutAndReduceShapes(c.shape,s),p=w.sizeFromShape(l),d=he({inputs:{x:c},backend:e,attrs:{shape:[-1,p]}});a.push(d);let h=yE(e,d,r);a.push(h);let f=he({inputs:{x:h},backend:e,attrs:{shape:u}});return a.forEach(m=>e.disposeIntermediateTensorInfo(m)),f}return vE(e,t,r)}function sJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a}=r,o=w.parseAxisParam(a,s.shape),i=N.getAxesPermutation(o,s.shape.length),c=s,u=[];i!=null&&(c=Nn({inputs:{x:s},backend:n,attrs:{perm:i}}),u.push(c),o=N.getInnerMostAxes(o.length,c.shape.length)),N.assertAxesAreInnerMostDims("argMax",[o[0]],c.shape.length);let l=xE(n,c,o[0],"max");return u.forEach(p=>n.disposeIntermediateTensorInfo(p)),l}var aJ={kernelName:yo,backendName:"webgl",kernelFunc:sJ};function oJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a}=r,o=w.parseAxisParam(a,s.shape),i=N.getAxesPermutation(o,s.shape.length),c=s,u=[];i!=null&&(c=Nn({inputs:{x:s},backend:n,attrs:{perm:i}}),u.push(c),o=N.getInnerMostAxes(o.length,c.shape.length)),N.assertAxesAreInnerMostDims("argMin",[o[0]],c.shape.length);let l=xE(n,c,o[0],"min");return u.forEach(p=>n.disposeIntermediateTensorInfo(p)),l}var iJ={kernelName:dd,backendName:"webgl",kernelFunc:oJ},cJ=Or+`
   if (abs(x) > 1.) {
     return NAN;
   }
   return asin(x);
-`,i9=Ye({opSnippet:s9}),o9={kernelName:Ll,backendName:"webgl",kernelFunc:i9},l9=Ma+"return log(x + sqrt(x * x + 1.0));",u9=Ye({opSnippet:l9}),p9={kernelName:zl,backendName:"webgl",kernelFunc:u9},c9=Ma+`
+`,uJ=Ye({opSnippet:cJ}),lJ={kernelName:Lc,backendName:"webgl",kernelFunc:uJ},dJ=Or+"return log(x + sqrt(x * x + 1.0));",pJ=Ye({opSnippet:dJ}),hJ={kernelName:zc,backendName:"webgl",kernelFunc:pJ},fJ=Or+`
   return atan(x);
-`,d9=Ye({opSnippet:c9}),h9={kernelName:Wl,backendName:"webgl",kernelFunc:d9},m9=i1+`
+`,mJ=Ye({opSnippet:fJ}),gJ={kernelName:Bc,backendName:"webgl",kernelFunc:mJ},bJ=oI+`
   return atan(a, b);
-`,f9=`
+`,yJ=`
   vec4 result = atan(a, b);
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+ld+`
+  `+cp+`
   return result;
-`,g9=cn({opSnippet:m9,packedOpSnippet:f9}),y9={kernelName:Vl,backendName:"webgl",kernelFunc:g9},b9=Ma+`
+`,vJ=dn({opSnippet:bJ,packedOpSnippet:yJ}),xJ={kernelName:Vc,backendName:"webgl",kernelFunc:vJ},wJ=Or+`
   if ((x < -1.0) || (x > 1.0)) return NAN;
-return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelName:Bl,backendName:"webgl",kernelFunc:x9},oc=class{constructor(e,t,n,a=!1,r=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideHeight,o=e.strideWidth,l=e.dilationHeight,u=e.dilationWidth,p=e.effectiveFilterHeight,d=e.effectiveFilterWidth,c=e.padInfo.top,h=e.padInfo.left;this.outputShape=e.outShape;let m=t==="avg",f=`((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`,y="0.0";if(m||(y="-1.0 / 1e-20"),n){let C=">=";this.userCode=`
-        const ivec2 strides = ivec2(${i}, ${o});
-        const ivec2 pads = ivec2(${c}, ${h});
+return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,IJ=Ye({opSnippet:wJ}),kJ={kernelName:Wc,backendName:"webgl",kernelFunc:IJ},id=class{constructor(e,t,n,r=!1,s=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let a=e.filterWidth,o=e.strideHeight,i=e.strideWidth,c=e.dilationHeight,u=e.dilationWidth,l=e.effectiveFilterHeight,p=e.effectiveFilterWidth,d=e.padInfo.top,h=e.padInfo.left;this.outputShape=e.outShape;let f=t==="avg",m=`((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`,b="0.0";if(f||(b="-1.0 / 1e-20"),n){let C=">=";this.userCode=`
+        const ivec2 strides = ivec2(${o}, ${i});
+        const ivec2 pads = ivec2(${d}, ${h});
 
         void main() {
           ivec4 coords = getOutputCoords();
@@ -1667,15 +1667,15 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           int minMaxPosition = 0;
           float avgValue = 0.0;
 
-          for (int wR = 0; wR < ${p};
-              wR += ${l}) {
+          for (int wR = 0; wR < ${l};
+              wR += ${c}) {
             int xR = xRCorner + wR;
 
             if (xR < 0 || xR >= ${e.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${d};
+            for (int wC = 0; wC < ${p};
                 wC += ${u}) {
               int xC = xCCorner + wC;
 
@@ -1692,22 +1692,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               if (value ${C} currMinMaxValue) {
                 minMaxValue = value;
                 minMaxValueFound = 1.0;
-                minMaxPosition = ${a?r?f:g:`wR * ${d} + wC`};
+                minMaxPosition = ${r?s?m:g:`wR * ${p} + wC`};
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let b="max",x=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(x="avgValue / count");let w=Math.floor(s/4)*4,I=s%4,T=`
-      if (${m}) {
+      `;return}let y="max",v=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(v="avgValue / count");let x=Math.floor(a/4)*4,k=a%4,S=`
+      if (${f}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${b}(values, minMaxValue);
+        minMaxValue = ${y}(values, minMaxValue);
       }
     `;this.userCode=`
-      const ivec2 strides = ivec2(${i}, ${o});
-      const ivec2 pads = ivec2(${c}, ${h});
-      const float initializationValue = ${y};
+      const ivec2 strides = ivec2(${o}, ${i});
+      const ivec2 pads = ivec2(${d}, ${h});
+      const float initializationValue = ${b};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
@@ -1731,19 +1731,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         // max/min x(?, ?, d) to get y(yR, yC, d).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${y});
+        vec4 minMaxValue = vec4(${b});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wR = 0; wR < ${p};
-            wR += ${l}) {
+        for (int wR = 0; wR < ${l};
+            wR += ${c}) {
           int xR = xRCorner + wR;
 
           if (xR < 0 || xR >= ${e.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${w}; wC += 4) {
+          for (int wC = 0; wC < ${x}; wC += 4) {
             int xC = xCCorner + wC * ${u};
 
             vec4 values = vec4(
@@ -1753,11 +1753,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               getValue(batch, xR, xC + 3 * ${u}, d)
             );
 
-            ${T}
+            ${S}
           }
 
-          int xC = xCCorner + ${w};
-          if (${I===1}) {
+          int xC = xCCorner + ${x};
+          if (${k===1}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
               initializationValue,
@@ -1765,8 +1765,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               initializationValue
             );
 
-            ${T}
-          } else if (${I===2}) {
+            ${S}
+          } else if (${k===2}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
               getValue(batch, xR, xC + ${u}, d),
@@ -1774,8 +1774,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               initializationValue
             );
 
-            ${T}
-          } else if (${I===3}) {
+            ${S}
+          } else if (${k===3}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
               getValue(batch, xR, xC + ${u}, d),
@@ -1783,15 +1783,15 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               initializationValue
             );
 
-            ${T}
+            ${S}
           }
         }
-        setOutput(${x});
+        setOutput(${v});
       }
-    `}},l1=class{constructor(e,t,n,a=!1,r=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideDepth,o=e.strideHeight,l=e.strideWidth,u=e.dilationDepth,p=e.dilationHeight,d=e.dilationWidth,c=e.effectiveFilterDepth,h=e.effectiveFilterHeight,m=e.effectiveFilterWidth,f=e.padInfo.front,g=e.padInfo.top,y=e.padInfo.left;this.outputShape=e.outShape;let b=t==="avg",x="0.0";if(b||(x="-1.0 / 1e-20"),n){let A=">=";this.userCode=`
+    `}},cI=class{constructor(e,t,n,r=!1,s=!1){if(this.variableNames=["x"],t==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let a=e.filterWidth,o=e.strideDepth,i=e.strideHeight,c=e.strideWidth,u=e.dilationDepth,l=e.dilationHeight,p=e.dilationWidth,d=e.effectiveFilterDepth,h=e.effectiveFilterHeight,f=e.effectiveFilterWidth,m=e.padInfo.front,g=e.padInfo.top,b=e.padInfo.left;this.outputShape=e.outShape;let y=t==="avg",v="0.0";if(y||(v="-1.0 / 1e-20"),n){let $=">=";this.userCode=`
         const ivec3 strides =
-            ivec3(${i}, ${o}, ${l});
-        const ivec3 pads = ivec3(${f}, ${g}, ${y});
+            ivec3(${o}, ${i}, ${c});
+        const ivec3 pads = ivec3(${m}, ${g}, ${b});
 
         void main() {
           ivec5 coords = getOutputCoords();
@@ -1809,7 +1809,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           float minMaxValueFound = 0.0;
           int minMaxPosition = 0;
 
-          for (int wD = 0; wD < ${c};
+          for (int wD = 0; wD < ${d};
               wD += ${u}) {
             int xD = xDCorner + wD;
 
@@ -1818,15 +1818,15 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             }
 
             for (int wR = 0; wR < ${h};
-                wR += ${p}) {
+                wR += ${l}) {
               int xR = xRCorner + wR;
 
               if (xR < 0 || xR >= ${e.inHeight}) {
                 continue;
               }
 
-              for (int wC = 0; wC < ${m};
-                  wC += ${d}) {
+              for (int wC = 0; wC < ${f};
+                  wC += ${p}) {
                 int xC = xCCorner + wC;
 
                 if (xC < 0 || xC >= ${e.inWidth}) {
@@ -1839,28 +1839,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                 // use the current value.
                 float currMinMaxValue = mix(
                     value, minMaxValue, minMaxValueFound);
-                if (value ${A} currMinMaxValue) {
+                if (value ${$} currMinMaxValue) {
                   minMaxValue = value;
                   minMaxValueFound = 1.0;
-                  minMaxPosition = ${a?r?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${h} * ${m} +
-                      wR * ${m} + wC`};
+                  minMaxPosition = ${r?s?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${h} * ${f} +
+                      wR * ${f} + wC`};
                 }
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let w="max",I=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(I="avgValue / count");let T=Math.floor(s/4)*4,C=s%4,E=`
-      if (${b}) {
+      `;return}let x="max",k=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(k="avgValue / count");let S=Math.floor(a/4)*4,C=a%4,E=`
+      if (${y}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${w}(values, minMaxValue);
+        minMaxValue = ${x}(values, minMaxValue);
       }
     `;this.userCode=`
       const ivec3 strides =
-        ivec3(${i}, ${o}, ${l});
-      const ivec3 pads = ivec3(${f}, ${g}, ${y});
-      const float initializationValue = ${x};
+        ivec3(${o}, ${i}, ${c});
+      const ivec3 pads = ivec3(${m}, ${g}, ${b});
+      const float initializationValue = ${v};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
@@ -1885,11 +1885,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${x});
+        vec4 minMaxValue = vec4(${v});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wD = 0; wD < ${c};
+        for (int wD = 0; wD < ${d};
             wD += ${u}) {
           int xD = xDCorner + wD;
 
@@ -1898,27 +1898,27 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           }
 
           for (int wR = 0; wR < ${h};
-            wR += ${p}) {
+            wR += ${l}) {
             int xR = xRCorner + wR;
 
             if (xR < 0 || xR >= ${e.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${T}; wC += 4) {
-              int xC = xCCorner + wC * ${d};
+            for (int wC = 0; wC < ${S}; wC += 4) {
+              int xC = xCCorner + wC * ${p};
 
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${d}, ch),
-                getValue(batch, xD, xR, xC + 3 * ${d}, ch)
+                getValue(batch, xD, xR, xC + ${p}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${p}, ch),
+                getValue(batch, xD, xR, xC + 3 * ${p}, ch)
               );
 
               ${E}
             }
 
-            int xC = xCCorner + ${T};
+            int xC = xCCorner + ${S};
             if (${C===1}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
@@ -1931,7 +1931,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             } else if (${C===2}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
+                getValue(batch, xD, xR, xC + ${p}, ch),
                 initializationValue,
                 initializationValue
               );
@@ -1940,20 +1940,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             } else if (${C===3}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${d}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${d}, ch),
+                getValue(batch, xD, xR, xC + ${p}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${p}, ch),
                 initializationValue
               );
 
               ${E}
             }
           }
-          setOutput(${I});
+          setOutput(${k});
         }
       }
-    `}};function w9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;Yu(r,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l);if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))return aa({inputs:{x:r},backend:n});let d=new oc(p,"avg",!1);return n.runWebGLProgram(d,[r],"float32")}var k9={kernelName:yi,backendName:"webgl",kernelFunc:w9};function I9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=a,p=[1,1,1],d=N.computePool3DInfo(r.shape,s,i,p,o,l,u),c=new l1(d,"avg",!1);return n.runWebGLProgram(c,[r],"float32")}var S9={kernelName:dc,backendName:"webgl",kernelFunc:I9},T9=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterHeight,l=e.effectiveFilterWidth,u=o-1-e.padInfo.top,p=l-1-e.padInfo.left,d=1/(t*n);this.userCode=`
-      const ivec2 pads = ivec2(${u}, ${p});
-      const float avgMultiplier = float(${d});
+    `}};function SJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t;Yu(s,"avgPool");let{filterSize:a,strides:o,pad:i,dimRoundingMode:c}=r,u=1;w.assert(N.eitherStridesOrDilationsAreOne(o,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${o} and dilations '${u}'`);let l=N.computePool2DInfo(s.shape,a,o,u,i,c);if(l.filterWidth===1&&l.filterHeight===1&&w.arraysEqual(l.inShape,l.outShape))return sr({inputs:{x:s},backend:n});let p=new id(l,"avg",!1);return n.runWebGLProgram(p,[s],"float32")}var TJ={kernelName:vo,backendName:"webgl",kernelFunc:SJ};function CJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{filterSize:a,strides:o,pad:i,dimRoundingMode:c,dataFormat:u}=r,l=[1,1,1],p=N.computePool3DInfo(s.shape,a,o,l,i,c,u),d=new cI(p,"avg",!1);return n.runWebGLProgram(d,[s],"float32")}var NJ={kernelName:pd,backendName:"webgl",kernelFunc:CJ},_J=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,s=e.strideWidth,a=e.dilationHeight,o=e.dilationWidth,i=e.effectiveFilterHeight,c=e.effectiveFilterWidth,u=i-1-e.padInfo.top,l=c-1-e.padInfo.left,p=1/(t*n);this.userCode=`
+      const ivec2 pads = ivec2(${u}, ${l});
+      const float avgMultiplier = float(${p});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -1967,18 +1967,18 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${o};
-            wR += ${s}) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+        for (int wR = 0; wR < ${i};
+            wR += ${a}) {
+          float dyR = float(dyRCorner + wR) / ${r}.0;
 
           if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${l};
-            wC+= ${i}) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+          for (int wC = 0; wC < ${c};
+            wC+= ${o}) {
+            float dyC = float(dyCCorner + wC) / ${s}.0;
 
             if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                 fract(dyC) > 0.0) {
@@ -1993,8 +1993,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},N9=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,a=e.filterWidth,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,p=e.effectiveFilterDepth,d=e.effectiveFilterHeight,c=e.effectiveFilterWidth,h=p-1-e.padInfo.front,m=d-1-e.padInfo.top,f=c-1-e.padInfo.left,g=1/(t*n*a);this.userCode=`
-      const ivec3 pads = ivec3(${h}, ${m}, ${f});
+    `}},EJ=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,s=e.strideDepth,a=e.strideHeight,o=e.strideWidth,i=e.dilationDepth,c=e.dilationHeight,u=e.dilationWidth,l=e.effectiveFilterDepth,p=e.effectiveFilterHeight,d=e.effectiveFilterWidth,h=l-1-e.padInfo.front,f=p-1-e.padInfo.top,m=d-1-e.padInfo.left,g=1/(t*n*r);this.userCode=`
+      const ivec3 pads = ivec3(${h}, ${f}, ${m});
       const float avgMultiplier = float(${g});
 
       void main() {
@@ -2012,18 +2012,18 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${p};
-            wD += ${o}) {
-          float dyD = float(dyDCorner + wD) / ${r}.0;
+        for (int wD = 0; wD < ${l};
+            wD += ${i}) {
+          float dyD = float(dyDCorner + wD) / ${s}.0;
 
           if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
             continue;
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${d};
-              wR += ${l}) {
-            float dyR = float(dyRCorner + wR) / ${s}.0;
+          for (int wR = 0; wR < ${p};
+              wR += ${c}) {
+            float dyR = float(dyRCorner + wR) / ${a}.0;
 
             if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
                 fract(dyR) > 0.0) {
@@ -2031,9 +2031,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             }
             int idyR = int(dyR);
 
-            for (int wC = 0; wC < ${c};
+            for (int wC = 0; wC < ${d};
                 wC += ${u}) {
-              float dyC = float(dyCCorner + wC) / ${i}.0;
+              float dyC = float(dyCCorner + wC) / ${o}.0;
 
               if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                   fract(dyC) > 0.0) {
@@ -2049,69 +2049,69 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function C9(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:p}=a,d=[1,1,1],c=N.computePool3DInfo(i.shape,o,l,d,u,p),h=new N9(c);return n.runWebGLProgram(h,[r],i.dtype)}var _9={kernelName:cm,backendName:"webgl",kernelFunc:C9};function E9(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s;Yu([r,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=a,p=N.computePool2DInfo(i.shape,o,l,1,u),d=new T9(p);return n.runWebGLProgram(d,[r],i.dtype)}var A9={kernelName:pm,backendName:"webgl",kernelFunc:E9};function $9(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;return nm({a:r,b:s,transposeA:i,transposeB:o,backend:n})}var F9={kernelName:bi,backendName:"webgl",kernelFunc:$9},D9=class{constructor(e,t,n,a,r,s){this.outputShape=[],this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let i="0.0";a!=null&&(N.assertAndGetBroadcastShape(e,a),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="1.0";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
+    `}};function AJ(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,o=a,{filterSize:i,strides:c,pad:u,dimRoundingMode:l}=r,p=[1,1,1],d=N.computePool3DInfo(o.shape,i,c,p,u,l),h=new EJ(d);return n.runWebGLProgram(h,[s],o.dtype)}var $J={kernelName:hf,backendName:"webgl",kernelFunc:AJ};function DJ(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,o=a;Yu([s,a],"avgPoolGrad");let{filterSize:i,strides:c,pad:u}=r,l=N.computePool2DInfo(o.shape,i,c,1,u),p=new _J(l);return n.runWebGLProgram(p,[s],o.dtype)}var FJ={kernelName:pf,backendName:"webgl",kernelFunc:DJ};function RJ(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a}=t,{transposeA:o,transposeB:i}=r;return rf({a:s,b:a,transposeA:o,transposeB:i,backend:n})}var PJ={kernelName:xo,backendName:"webgl",kernelFunc:RJ},OJ=class{constructor(e,t,n,r,s,a){this.outputShape=[],this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let o="0.0";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),o="getOffsetAtOutCoords()");let i="1.0";s!=null&&(N.assertAndGetBroadcastShape(e,s),this.variableNames.push("scale"),i="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
       void main() {
         float x = getXAtOutCoords();
         float mean = getMeanAtOutCoords();
         float variance = getVarianceAtOutCoords();
-        float offset = ${i};
-        float scale = ${o};
-        float inv = scale * inversesqrt(variance + float(${s}));
+        float offset = ${o};
+        float scale = ${i};
+        float inv = scale * inversesqrt(variance + float(${a}));
         setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));
       }
-    `}},R9=class{constructor(e,t,n,a,r,s){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let i="vec4(0.0)";a!=null&&(N.assertAndGetBroadcastShape(e,a),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="vec4(1.0)";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
+    `}},MJ=class{constructor(e,t,n,r,s,a){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],N.assertAndGetBroadcastShape(e,t),N.assertAndGetBroadcastShape(e,n);let o="vec4(0.0)";r!=null&&(N.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),o="getOffsetAtOutCoords()");let i="vec4(1.0)";s!=null&&(N.assertAndGetBroadcastShape(e,s),this.variableNames.push("scale"),i="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
       void main() {
-        vec4 offset = ${i};
-        vec4 scale = ${o};
+        vec4 offset = ${o};
+        vec4 scale = ${i};
 
         vec4 x = getXAtOutCoords();
         vec4 mean = getMeanAtOutCoords();
         vec4 variance = getVarianceAtOutCoords();
 
-        vec4 inv = scale * inversesqrt(variance + vec4(${s}));
+        vec4 inv = scale * inversesqrt(variance + vec4(${a}));
 
         setOutput((x - mean) * inv + offset);
       }
-    `}},M9=({inputs:e,backend:t,attrs:n})=>{let{x:a,mean:r,variance:s,offset:i,scale:o}=e;v.assert(r.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||r.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(o==null||r.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:l}=n;l==null&&(l=.001);let u=[a,r,s],p=null;i!=null&&(p=i.shape,u.push(i));let d=null;o!=null&&(d=o.shape,u.push(o));let c=H().getBool("WEBGL_PACK_NORMALIZATION")?new R9(a.shape,r.shape,s.shape,p,d,l):new D9(a.shape,r.shape,s.shape,p,d,l);return t.runWebGLProgram(c,u,u[0].dtype)},P9={kernelName:Fi,backendName:"webgl",kernelFunc:M9},O9=class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;let t=gt(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let n=L9(this.rank),a,r=e.map((s,i)=>`sourceLoc.${bx[i]} = start[${i}] + coords.${bx[i]};`);a=`
+    `}},LJ=({inputs:e,backend:t,attrs:n})=>{let{x:r,mean:s,variance:a,offset:o,scale:i}=e;w.assert(s.shape.length===a.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),w.assert(o==null||s.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),w.assert(i==null||s.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:c}=n;c==null&&(c=.001);let u=[r,s,a],l=null;o!=null&&(l=o.shape,u.push(o));let p=null;i!=null&&(p=i.shape,u.push(i));let d=q().getBool("WEBGL_PACK_NORMALIZATION")?new MJ(r.shape,s.shape,a.shape,l,p,c):new OJ(r.shape,s.shape,a.shape,l,p,c);return t.runWebGLProgram(d,u,u[0].dtype)},zJ={kernelName:Ro,backendName:"webgl",kernelFunc:LJ},BJ=class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;let t=mt(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let n=WJ(this.rank),r,s=e.map((a,o)=>`sourceLoc.${yv[o]} = start[${o}] + coords.${yv[o]};`);r=`
         ${t} sourceLoc;
         ${t} coords = getOutputCoords();
-        ${r.join(`
+        ${s.join(`
 `)}
       `,this.userCode=`
       void main() {
-        ${a}
+        ${r}
         setOutput(getSource(${n}));
       }
-    `}},bx=["x","y","z","w","u","v"];function L9(e){if(e===1)return"sourceLoc";if(e<=6)return bx.slice(0,e).map(t=>"sourceLoc."+t).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}var z9=class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let t=gt(this.rank),n=kn("coords",this.rank),a=kn("sourceLoc",this.rank),r=this.rank===1?"sourceLoc":`vec2(${a.slice(-2).join()})`,s=`getChannel(getSource(${a.join()}), ${r})`,i=`
-      result.x = ${s};
+    `}},yv=["x","y","z","w","u","v"];function WJ(e){if(e===1)return"sourceLoc";if(e<=6)return yv.slice(0,e).map(t=>"sourceLoc."+t).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}var VJ=class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let t=mt(this.rank),n=Tn("coords",this.rank),r=Tn("sourceLoc",this.rank),s=this.rank===1?"sourceLoc":`vec2(${r.slice(-2).join()})`,a=`getChannel(getSource(${r.join()}), ${s})`,o=`
+      result.x = ${a};
       if (++${n[this.rank-1]} < ${e[this.rank-1]}) {
-        ++${a[this.rank-1]};
-        result.y = ${s};
-        --${a[this.rank-1]};
+        ++${r[this.rank-1]};
+        result.y = ${a};
+        --${r[this.rank-1]};
       }
-    `,o=this.rank===1?"":`
+    `,i=this.rank===1?"":`
       --${n[this.rank-1]};
       if (++${n[this.rank-2]} < ${e[this.rank-2]}) {
-        ++${a[this.rank-2]};
-        result.z = ${s};
+        ++${r[this.rank-2]};
+        result.z = ${a};
         if (++${n[this.rank-1]} < ${e[this.rank-1]}) {
-          ++${a[this.rank-1]};
-          result.w = ${s};
+          ++${r[this.rank-1]};
+          result.w = ${a};
         }
       }
-    `,l=this.rank<=4?`sourceLoc = coords +
-            ${t}(${e.map((u,p)=>`start[${p}]`).join()});`:e.map((u,p)=>`${a[p]} = ${n[p]} + start[${p}];`).join(`
+    `,c=this.rank<=4?`sourceLoc = coords +
+            ${t}(${e.map((u,l)=>`start[${l}]`).join()});`:e.map((u,l)=>`${r[l]} = ${n[l]} + start[${l}];`).join(`
 `);this.userCode=`
       void main() {
         ${t} coords = getOutputCoords();
         ${t} sourceLoc;
-        ${l}
+        ${c}
         vec4 result = vec4(0.);
-        ${i}
         ${o}
+        ${i}
         setOutput(result);
       }
-    `}};function W9(e,t,n,a){let r=a.texData.get(e.dataId),s=a.makeTensorInfo(n,e.dtype),i=a.texData.get(s.dataId);Object.assign(i,r),i.refCount=1,i.shape=n,i.dtype=e.dtype;let o=jt.computeFlatOffset(t,v.computeStrides(e.shape));r.slice&&(o+=r.slice.flatOffset),i.slice={flatOffset:o,origDataId:r.slice&&r.slice.origDataId||e.dataId};let l=a.dataRefCount.get(i.slice.origDataId)||1;return a.dataRefCount.set(i.slice.origDataId,l+1),s}function np(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,size:i}=a,[o,l]=jt.parseSliceParams(r,s,i);if(jt.assertParamsValid(r,o,l),v.sizeFromShape(l)===0)return n.makeTensorInfo(l,r.dtype,[]);if(n.shouldExecuteOnCPU([r])||r.dtype==="string"){let d=n.texData.get(r.dataId),c=$7(d.values,o,l,r.shape,r.dtype);return n.makeTensorInfo(l,r.dtype,c)}let{isPacked:u}=n.texData.get(r.dataId),p=jt.isSliceContinous(r.shape,o,l);if(u||!p){let d=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new z9(l):new O9(l),c=[o];return n.runWebGLProgram(d,[r],r.dtype,c)}return n.uploadToGPU(r.dataId),W9(r,o,l,n)}var B9={kernelName:Su,backendName:"webgl",kernelFunc:np},V9=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a;v.assert(r.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((b,x)=>b*x),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=[],m=de({inputs:{x:r},backend:n,attrs:{shape:l}}),f=Sn({inputs:{x:m},backend:n,attrs:{perm:u}}),g=de({inputs:{x:f},backend:n,attrs:{shape:p}}),y=np({inputs:{x:g},backend:n,attrs:{begin:d,size:c}});return h.push(m),h.push(f),h.push(g),h.forEach(b=>n.disposeIntermediateTensorInfo(b)),y},U9={kernelName:Ul,backendName:"webgl",kernelFunc:V9};function G9(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i}=a,o=n.readSync(r.dataId),l=n.readSync(s.dataId),u=rE(o,l,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,u)}var H9={kernelName:dm,backendName:"webgl",kernelFunc:G9};function j9(e){let{inputs:t,backend:n}=e,{s0:a,s1:r}=t,s=n.readSync(a.dataId),i=n.readSync(r.dataId),o=N.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var q9={kernelName:hm,backendName:"webgl",kernelFunc:j9},K9="return float(a != b);",xE=cn({opSnippet:K9,cpuKernelImpl:k7,dtype:"bool"}),X9={kernelName:hu,backendName:"webgl",kernelFunc:xE};function pd(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.texData.get(a.dataId);return aa({inputs:{x:r.complexTensorInfos.real},backend:n})}var Y9={kernelName:Pm,backendName:"webgl",kernelFunc:pd},Z9="return float(int(x));";function J9(e,t){let n=new Cr(e.shape,Z9),a=t.runWebGLProgram(n,[e],"int32");return{dataId:a.dataId,shape:a.shape,dtype:a.dtype}}function xx(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dtype:s}=a;if(s==="complex64"){if(r.dtype==="complex64")return aa({inputs:{x:r},backend:n});let i=It(r.shape),o=xx({inputs:{x:r},backend:n,attrs:{dtype:"float32"}}),l=_s({inputs:{real:o,imag:i},backend:n});return i.dispose(),n.disposeIntermediateTensorInfo(o),l}if(r.dtype==="complex64"){let i=pd({inputs:{input:r},backend:n}),o=xx({inputs:{x:i},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(i),o}if(!v.hasEncodingLoss(r.dtype,s)){let i=aa({inputs:{x:r},backend:n});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(n.shouldExecuteOnCPU([r])){let i=n.texData.get(r.dataId).values,[o,l,u]=n7(i,r.shape,r.dtype,s);return n.makeTensorInfo(o,l,u)}if(s==="int32")return J9(r,n);if(s==="bool"){let i=n.makeTensorInfo([],"bool",v.getTypedArrayFromDType("bool",1)),o=xE({inputs:{a:r,b:i},backend:n});return n.disposeIntermediateTensorInfo(i),o}throw new Error(`Error in Cast: failed to cast ${r.dtype} to ${s}`)}var Q9={kernelName:xi,backendName:"webgl",kernelFunc:xx},SI="return ceil(x);",eQ=Ye({opSnippet:SI,packedOpSnippet:SI,cpuKernelImpl:a7}),tQ={kernelName:vi,backendName:"webgl",kernelFunc:eQ},nQ=class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
+    `}};function UJ(e,t,n,r){let s=r.texData.get(e.dataId),a=r.makeTensorInfo(n,e.dtype),o=r.texData.get(a.dataId);Object.assign(o,s),o.refCount=1,o.shape=n,o.dtype=e.dtype;let i=qt.computeFlatOffset(t,w.computeStrides(e.shape));s.slice&&(i+=s.slice.flatOffset),o.slice={flatOffset:i,origDataId:s.slice&&s.slice.origDataId||e.dataId};let c=r.dataRefCount.get(o.slice.origDataId)||1;return r.dataRefCount.set(o.slice.origDataId,c+1),a}function nl(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{begin:a,size:o}=r,[i,c]=qt.parseSliceParams(s,a,o);if(qt.assertParamsValid(s,i,c),w.sizeFromShape(c)===0)return n.makeTensorInfo(c,s.dtype,[]);if(n.shouldExecuteOnCPU([s])||s.dtype==="string"){let p=n.texData.get(s.dataId),d=R9(p.values,i,c,s.shape,s.dtype);return n.makeTensorInfo(c,s.dtype,d)}let{isPacked:u}=n.texData.get(s.dataId),l=qt.isSliceContinous(s.shape,i,c);if(u||!l){let p=q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new VJ(c):new BJ(c),d=[i];return n.runWebGLProgram(p,[s],s.dtype,d)}return n.uploadToGPU(s.dataId),UJ(s,i,c,n)}var GJ={kernelName:Su,backendName:"webgl",kernelFunc:nl},HJ=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,crops:o}=r;w.assert(s.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let i=a.reduce((y,v)=>y*v),c=N.getReshaped(s.shape,a,i),u=N.getPermuted(c.length,a.length),l=N.getReshapedPermuted(s.shape,a,i),p=N.getSliceBeginCoords(o,a.length),d=N.getSliceSize(l,o,a.length),h=[],f=he({inputs:{x:s},backend:n,attrs:{shape:c}}),m=Nn({inputs:{x:f},backend:n,attrs:{perm:u}}),g=he({inputs:{x:m},backend:n,attrs:{shape:l}}),b=nl({inputs:{x:g},backend:n,attrs:{begin:p,size:d}});return h.push(f),h.push(m),h.push(g),h.forEach(y=>n.disposeIntermediateTensorInfo(y)),b},qJ={kernelName:Uc,backendName:"webgl",kernelFunc:HJ};function jJ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,weights:a}=t,{size:o}=r,i=n.readSync(s.dataId),c=n.readSync(a.dataId),u=oE(i,c,a.dtype,a.shape,o);return n.makeTensorInfo([o],a.dtype,u)}var KJ={kernelName:ff,backendName:"webgl",kernelFunc:jJ};function XJ(e){let{inputs:t,backend:n}=e,{s0:r,s1:s}=t,a=n.readSync(r.dataId),o=n.readSync(s.dataId),i=N.assertAndGetBroadcastShape(Array.from(a),Array.from(o));return n.makeTensorInfo([i.length],"int32",Int32Array.from(i))}var YJ={kernelName:mf,backendName:"webgl",kernelFunc:XJ},ZJ="return float(a != b);",wE=dn({opSnippet:ZJ,cpuKernelImpl:T9,dtype:"bool"}),JJ={kernelName:hu,backendName:"webgl",kernelFunc:wE};function lp(e){let{inputs:t,backend:n}=e,{input:r}=t,s=n.texData.get(r.dataId);return sr({inputs:{x:s.complexTensorInfos.real},backend:n})}var QJ={kernelName:Lf,backendName:"webgl",kernelFunc:lp},eQ="return float(int(x));";function tQ(e,t){let n=new Ns(e.shape,eQ),r=t.runWebGLProgram(n,[e],"int32");return{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}function vv(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{dtype:a}=r;if(a==="complex64"){if(s.dtype==="complex64")return sr({inputs:{x:s},backend:n});let o=kt(s.shape),i=vv({inputs:{x:s},backend:n,attrs:{dtype:"float32"}}),c=Ea({inputs:{real:i,imag:o},backend:n});return o.dispose(),n.disposeIntermediateTensorInfo(i),c}if(s.dtype==="complex64"){let o=lp({inputs:{input:s},backend:n}),i=vv({inputs:{x:o},backend:n,attrs:{dtype:a}});return n.disposeIntermediateTensorInfo(o),i}if(!w.hasEncodingLoss(s.dtype,a)){let o=sr({inputs:{x:s},backend:n});return{dataId:o.dataId,shape:o.shape,dtype:a}}if(n.shouldExecuteOnCPU([s])){let o=n.texData.get(s.dataId).values,[i,c,u]=a9(o,s.shape,s.dtype,a);return n.makeTensorInfo(i,c,u)}if(a==="int32")return tQ(s,n);if(a==="bool"){let o=n.makeTensorInfo([],"bool",w.getTypedArrayFromDType("bool",1)),c=wE({inputs:{a:s,b:o},backend:n});return n.disposeIntermediateTensorInfo(o),c}throw new Error(`Error in Cast: failed to cast ${s.dtype} to ${a}`)}var nQ={kernelName:wo,backendName:"webgl",kernelFunc:vv},T1="return ceil(x);",rQ=Ye({opSnippet:T1,packedOpSnippet:T1,cpuKernelImpl:o9}),sQ={kernelName:Io,backendName:"webgl",kernelFunc:rQ},aQ=class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
 
       void main() {
         float value = getAAtOutCoords();
@@ -2122,7 +2122,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         setOutput(clamp(value, minVal, maxVal));
       }
-    `}},aQ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
+    `}},oQ=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
       void main() {
         vec4 value = getAAtOutCoords();
 
@@ -2133,7 +2133,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         setOutput(clamp(value, vec4(minVal), vec4(maxVal)));
       }
-    `}};function rQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{clipValueMin:s,clipValueMax:i}=a,o;H().getBool("WEBGL_PACK_CLIP")?o=new aQ(r.shape):o=new nQ(r.shape);let l=[[s],[i]];return n.runWebGLProgram(o,[r],r.dtype,l)}var sQ={kernelName:ys,backendName:"webgl",kernelFunc:rQ},iQ=class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode=`
+    `}};function iQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{clipValueMin:a,clipValueMax:o}=r,i;q().getBool("WEBGL_PACK_CLIP")?i=new oQ(s.shape):i=new aQ(s.shape);let c=[[a],[o]];return n.runWebGLProgram(i,[s],s.dtype,c)}var cQ={kernelName:ya,backendName:"webgl",kernelFunc:iQ},uQ=class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode=`
       void main() {
         float re = abs(getRealAtOutCoords());
         float im = abs(getImagAtOutCoords());
@@ -2146,7 +2146,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))
         );
       }
-    `}};function TI(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}function oQ(e){let{inputs:t,backend:n}=e,{x:a}=t,r=n.texData.get(a.dataId),s=new iQ(a.shape),i=[TI(a,r.complexTensorInfos.real),TI(a,r.complexTensorInfos.imag)];return n.runWebGLProgram(s,i,i[0].dtype)}var lQ={kernelName:hc,backendName:"webgl",kernelFunc:oQ},uQ=class{constructor(e){this.outputShape=[],this.outputShape=N.computeOutShape(e,1),this.variableNames=e.map((s,i)=>`T${i}`);let t=new Array(e.length-1);t[0]=e[0][1];for(let s=1;s<t.length;s++)t[s]=t[s-1]+e[s][1];let n=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let s=1;s<t.length;s++){let i=t[s-1];n.push(`else if (yC < ${t[s]}) setOutput(getT${s}(yR, yC-${i}));`)}let a=t.length,r=t[t.length-1];n.push(`else setOutput(getT${a}(yR, yC-${r}));`),this.userCode=`
+    `}};function C1(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}function lQ(e){let{inputs:t,backend:n}=e,{x:r}=t,s=n.texData.get(r.dataId),a=new uQ(r.shape),o=[C1(r,s.complexTensorInfos.real),C1(r,s.complexTensorInfos.imag)];return n.runWebGLProgram(a,o,o[0].dtype)}var dQ={kernelName:hd,backendName:"webgl",kernelFunc:lQ},pQ=class{constructor(e){this.outputShape=[],this.outputShape=N.computeOutShape(e,1),this.variableNames=e.map((a,o)=>`T${o}`);let t=new Array(e.length-1);t[0]=e[0][1];for(let a=1;a<t.length;a++)t[a]=t[a-1]+e[a][1];let n=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let a=1;a<t.length;a++){let o=t[a-1];n.push(`else if (yC < ${t[a]}) setOutput(getT${a}(yR, yC-${o}));`)}let r=t.length,s=t[t.length-1];n.push(`else setOutput(getT${r}(yR, yC-${s}));`),this.userCode=`
       void main() {
         ivec2 coords = getOutputCoords();
         int yR = coords.x;
@@ -2155,81 +2155,81 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         ${n.join(`
         `)}
       }
-    `}},pQ=class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=N.computeOutShape(e,t);let n=this.outputShape,a=n.length,r=gt(a),s=kn("coords",a),i=["x","y","z","w","u","v"].slice(0,a);this.variableNames=e.map((m,f)=>`T${f}`);let o=new Array(e.length-1);o[0]=e[0][t];for(let m=1;m<o.length;m++)o[m]=o[m-1]+e[m][t];let l=i[t],u=i.slice(-2),p=i.join(),d=`if (${l} < ${o[0]}) {
+    `}},hQ=class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=N.computeOutShape(e,t);let n=this.outputShape,r=n.length,s=mt(r),a=Tn("coords",r),o=["x","y","z","w","u","v"].slice(0,r);this.variableNames=e.map((f,m)=>`T${m}`);let i=new Array(e.length-1);i[0]=e[0][t];for(let f=1;f<i.length;f++)i[f]=i[f-1]+e[f][t];let c=o[t],u=o.slice(-2),l=o.join(),p=`if (${c} < ${i[0]}) {
         return getChannel(
-            getT0(${p}), vec2(${u.join()}));
-        }`;for(let m=1;m<o.length;m++){let f=o[m-1];d+=`
-        if (${l} < ${o[m]}  && ${l} >= ${o[m-1]}) {
+            getT0(${l}), vec2(${u.join()}));
+        }`;for(let f=1;f<i.length;f++){let m=i[f-1];p+=`
+        if (${c} < ${i[f]}  && ${c} >= ${i[f-1]}) {
           return getChannel(
-            getT${m}(${yh(i,l,f)}),
-            vec2(${yh(u,l,f)}));
-        }`}let c=o.length,h=o[o.length-1];d+=`
+            getT${f}(${yh(o,c,m)}),
+            vec2(${yh(u,c,m)}));
+        }`}let d=i.length,h=i[i.length-1];p+=`
         return getChannel(
-          getT${c}(${yh(i,l,h)}),
-          vec2(${yh(u,l,h)}));`,this.userCode=`
-      float getValue(${i.map(m=>"int "+m)}) {
-        ${d}
+          getT${d}(${yh(o,c,h)}),
+          vec2(${yh(u,c,h)}));`,this.userCode=`
+      float getValue(${o.map(f=>"int "+f)}) {
+        ${p}
       }
 
       void main() {
-        ${r} coords = getOutputCoords();
-        vec4 result = vec4(getValue(${s}), 0., 0., 0.);
+        ${s} coords = getOutputCoords();
+        vec4 result = vec4(getValue(${a}), 0., 0., 0.);
 
-        ${s[a-1]} = ${s[a-1]} + 1;
-        if (${s[a-1]} < ${n[a-1]}) {
-          result.g = getValue(${s});
+        ${a[r-1]} = ${a[r-1]} + 1;
+        if (${a[r-1]} < ${n[r-1]}) {
+          result.g = getValue(${a});
         }
 
-        ${s[a-2]} = ${s[a-2]} + 1;
-        if (${s[a-2]} < ${n[a-2]}) {
-          result.a = getValue(${s});
+        ${a[r-2]} = ${a[r-2]} + 1;
+        if (${a[r-2]} < ${n[r-2]}) {
+          result.a = getValue(${a});
         }
 
-        ${s[a-1]} = ${s[a-1]} - 1;
-        if (${s[a-2]} < ${n[a-2]} &&
-            ${s[a-1]} < ${n[a-1]}) {
-          result.b = getValue(${s});
+        ${a[r-1]} = ${a[r-1]} - 1;
+        if (${a[r-2]} < ${n[r-2]} &&
+            ${a[r-1]} < ${n[r-1]}) {
+          result.b = getValue(${a});
         }
         setOutput(result);
       }
-    `}};function yh(e,t,n){let a=e.indexOf(t);return e.map((r,s)=>s===a?`${r} - ${n}`:r).join()}function Jf(e){let{inputs:t,backend:n}=e,{input:a}=t,r=n.texData.get(a.dataId);return aa({inputs:{x:r.complexTensorInfos.imag},backend:n})}var cQ={kernelName:Nm,backendName:"webgl",kernelFunc:Jf};function Vp(e,t,n){let a=e[0].dtype;if(a==="complex64"){let d=e.map(g=>pd({inputs:{input:g},backend:n})),c=e.map(g=>Jf({inputs:{input:g},backend:n})),h=Vp(d,t,n),m=Vp(c,t,n),f=_s({inputs:{real:h,imag:m},backend:n});return d.forEach(g=>n.disposeIntermediateTensorInfo(g)),c.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),f}let r=n.shouldExecuteOnCPU(e);if(a==="string"&&(r=!0),r){let d=e.map(b=>{let x=v.sizeFromShape(b.shape.slice(t));return de({inputs:{x:b},backend:n,attrs:{shape:[-1,x]}})}),c=d.map(b=>({vals:n.readSync(b.dataId),shape:b.shape})),h=N.computeOutShape(d.map(b=>b.shape),1),m=d[0].shape[0]===1,f=r7(c,h,a,m),g=N.computeOutShape(e.map(b=>b.shape),t),y=n.makeTensorInfo(g,a,f);return d.forEach(b=>n.disposeIntermediateTensorInfo(b)),y}let s=H().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");if(e.length>s){let d=[];for(let h=0;h<e.length;h+=s){let m=e.slice(h,h+s);d.push(Vp(m,t,n))}let c=Vp(d,t,n);for(let h of d)n.disposeIntermediateTensorInfo(h);return c}if(H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&e[0].shape.length>1){let d=new pQ(e.map(c=>c.shape),t);return n.runWebGLProgram(d,e,a)}let{tensors2D:i,outShape:o}=dQ(e,t,n),l=new uQ(i.map(d=>d.shape)),u=n.runWebGLProgram(l,i,a);i.forEach(d=>n.disposeIntermediateTensorInfo(d));let p=de({inputs:{x:u},attrs:{shape:o},backend:n});return n.disposeIntermediateTensorInfo(u),p}function dQ(e,t,n){let a=N.computeOutShape(e.map(r=>r.shape),t);return{tensors2D:e.map(r=>de({inputs:{x:r},attrs:{shape:[-1,v.sizeFromShape(r.shape.slice(t))]},backend:n})),outShape:a}}function vE(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a,s=v.parseAxisParam(r,t[0].shape)[0],i=t.map(u=>u.shape);N.assertParamsConsistent(i,s);let o=N.computeOutShape(t.map(u=>u.shape),s);if(v.sizeFromShape(o)===0)return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(u=>v.sizeFromShape(u.shape)>0);return l.length===1?aa({inputs:{x:l[0]},backend:n}):Vp(l,s,n)}var hQ={kernelName:Gl,backendName:"webgl",kernelFunc:vE},wE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;let s=e.padInfo.top,i=e.padInfo.left,o=e.strideHeight,l=e.strideWidth,u=e.dilationHeight,p=e.dilationWidth,d=e.filterHeight,c=e.filterWidth,h=Math.floor(e.inChannels/4)*4,m=e.inChannels%4,f=e.dataFormat==="channelsLast",g=f?1:2,y=f?2:3,b=f?3:1,x="",w="";n&&(a?x=`float activation(float a) {
+    `}};function yh(e,t,n){let r=e.indexOf(t);return e.map((a,o)=>o===r?`${a} - ${n}`:a).join()}function Qm(e){let{inputs:t,backend:n}=e,{input:r}=t,s=n.texData.get(r.dataId);return sr({inputs:{x:s.complexTensorInfos.imag},backend:n})}var fQ={kernelName:_f,backendName:"webgl",kernelFunc:Qm};function Vl(e,t,n){let r=e[0].dtype;if(r==="complex64"){let p=e.map(g=>lp({inputs:{input:g},backend:n})),d=e.map(g=>Qm({inputs:{input:g},backend:n})),h=Vl(p,t,n),f=Vl(d,t,n),m=Ea({inputs:{real:h,imag:f},backend:n});return p.forEach(g=>n.disposeIntermediateTensorInfo(g)),d.forEach(g=>n.disposeIntermediateTensorInfo(g)),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),m}let s=n.shouldExecuteOnCPU(e);if(r==="string"&&(s=!0),s){let p=e.map(y=>{let x=[-1,w.sizeFromShape(y.shape.slice(t))];return he({inputs:{x:y},backend:n,attrs:{shape:x}})}),d=p.map(y=>({vals:n.readSync(y.dataId),shape:y.shape})),h=N.computeOutShape(p.map(y=>y.shape),1),f=p[0].shape[0]===1,m=i9(d,h,r,f),g=N.computeOutShape(e.map(y=>y.shape),t),b=n.makeTensorInfo(g,r,m);return p.forEach(y=>n.disposeIntermediateTensorInfo(y)),b}let a=q().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");if(e.length>a){let p=[];for(let h=0;h<e.length;h+=a){let f=e.slice(h,h+a);p.push(Vl(f,t,n))}let d=Vl(p,t,n);for(let h of p)n.disposeIntermediateTensorInfo(h);return d}if(q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&e[0].shape.length>1){let p=new hQ(e.map(d=>d.shape),t);return n.runWebGLProgram(p,e,r)}let{tensors2D:o,outShape:i}=mQ(e,t,n),c=new pQ(o.map(p=>p.shape)),u=n.runWebGLProgram(c,o,r);o.forEach(p=>n.disposeIntermediateTensorInfo(p));let l=he({inputs:{x:u},attrs:{shape:i},backend:n});return n.disposeIntermediateTensorInfo(u),l}function mQ(e,t,n){let r=N.computeOutShape(e.map(a=>a.shape),t);return{tensors2D:e.map(a=>he({inputs:{x:a},attrs:{shape:[-1,w.sizeFromShape(a.shape.slice(t))]},backend:n})),outShape:r}}function IE(e){let{inputs:t,backend:n,attrs:r}=e,{axis:s}=r,a=w.parseAxisParam(s,t[0].shape)[0],o=t.map(u=>u.shape);N.assertParamsConsistent(o,a);let i=N.computeOutShape(t.map(u=>u.shape),a);if(w.sizeFromShape(i)===0)return n.makeTensorInfo(i,t[0].dtype,[]);let c=t.filter(u=>w.sizeFromShape(u.shape)>0);return c.length===1?sr({inputs:{x:c[0]},backend:n}):Vl(c,a,n)}var gQ={kernelName:Gc,backendName:"webgl",kernelFunc:IE},kE=class{constructor(e,t=!1,n=null,r=!1,s=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;let a=e.padInfo.top,o=e.padInfo.left,i=e.strideHeight,c=e.strideWidth,u=e.dilationHeight,l=e.dilationWidth,p=e.filterHeight,d=e.filterWidth,h=Math.floor(e.inChannels/4)*4,f=e.inChannels%4,m=e.dataFormat==="channelsLast",g=m?1:2,b=m?2:3,y=m?3:1,v="",x="";n&&(r?v=`float activation(float a) {
           float b = getPreluActivationWeightsAtOutCoords();
           ${n}
-        }`:r?x=`float activation(float a) {
+        }`:s?v=`float activation(float a) {
           float b = getLeakyreluAlphaAtOutCoords();
           ${n}
-        }`:x=`
+        }`:v=`
           float activation(float x) {
             ${n}
           }
-        `,w="result = activation(result);");let I=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${x}
+        `,x="result = activation(result);");let k=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
+      ${v}
 
-      const ivec2 strides = ivec2(${o}, ${l});
-      const ivec2 pads = ivec2(${s}, ${i});
+      const ivec2 strides = ivec2(${i}, ${c});
+      const ivec2 pads = ivec2(${a}, ${o});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d2 = coords[${b}];
+        int d2 = coords[${y}];
 
         ivec2 xRCCorner =
-            ivec2(coords[${g}], coords[${y}]) * strides - pads;
+            ivec2(coords[${g}], coords[${b}]) * strides - pads;
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${d}; wR++) {
+        for (int wR = 0; wR < ${p}; wR++) {
           int xR = xRCorner + wR * ${u};
 
           if (xR < 0 || xR >= ${e.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${c}; wC++) {
-            int xC = xCCorner + wC * ${p};
+          for (int wC = 0; wC < ${d}; wC++) {
+            int xC = xCCorner + wC * ${l};
 
             if (xC < 0 || xC >= ${e.inWidth}) {
               continue;
@@ -2243,7 +2243,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                 getW(wR, wC, d1 + 3, d2)
               );
 
-              if (${f}) {
+              if (${m}) {
                 vec4 xValues = vec4(
                   getX(batch, xR, xC, d1),
                   getX(batch, xR, xC, d1 + 1),
@@ -2262,9 +2262,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               }
             }
 
-            if (${m===1}) {
+            if (${f===1}) {
 
-              if (${f}) {
+              if (${m}) {
                 dotProd +=
                     getX(batch, xR, xC, ${h}) *
                     getW(wR, wC, ${h}, d2);
@@ -2274,13 +2274,13 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                     getW(wR, wC, ${h}, d2);
               }
 
-            } else if (${m===2}) {
+            } else if (${f===2}) {
               vec2 wValues = vec2(
                 getW(wR, wC, ${h}, d2),
                 getW(wR, wC, ${h} + 1, d2)
               );
 
-              if (${f}) {
+              if (${m}) {
                 vec2 xValues = vec2(
                   getX(batch, xR, xC, ${h}),
                   getX(batch, xR, xC, ${h} + 1)
@@ -2294,14 +2294,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                 dotProd += dot(xValues, wValues);
               }
 
-            } else if (${m===3}) {
+            } else if (${f===3}) {
               vec3 wValues = vec3(
                 getW(wR, wC, ${h}, d2),
                 getW(wR, wC, ${h} + 1, d2),
                 getW(wR, wC, ${h} + 2, d2)
               );
 
-              if (${f}) {
+              if (${m}) {
                 vec3 xValues = vec3(
                   getX(batch, xR, xC, ${h}),
                   getX(batch, xR, xC, ${h} + 1),
@@ -2322,13 +2322,13 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
 
         float result = dotProd;
-        ${I}
-        ${w}
+        ${k}
+        ${x}
         setOutput(result);
       }
-    `}},mQ=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let t=e.padInfo.front,n=e.padInfo.top,a=e.padInfo.left,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,p=e.filterDepth,d=e.filterHeight,c=e.filterWidth,h=Math.floor(e.inChannels/4)*4,m=e.inChannels%4;this.userCode=`
-      const ivec3 strides = ivec3(${r}, ${s}, ${i});
-      const ivec3 pads = ivec3(${t}, ${n}, ${a});
+    `}},bQ=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let t=e.padInfo.front,n=e.padInfo.top,r=e.padInfo.left,s=e.strideDepth,a=e.strideHeight,o=e.strideWidth,i=e.dilationDepth,c=e.dilationHeight,u=e.dilationWidth,l=e.filterDepth,p=e.filterHeight,d=e.filterWidth,h=Math.floor(e.inChannels/4)*4,f=e.inChannels%4;this.userCode=`
+      const ivec3 strides = ivec3(${s}, ${a}, ${o});
+      const ivec3 pads = ivec3(${t}, ${n}, ${r});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2344,21 +2344,21 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // y(yF, yR, yC, d2). ? = to be determined. : = across all
         // values in that axis.
         float dotProd = 0.0;
-        for (int wF = 0; wF < ${p}; wF++) {
-          int xF = xFCorner + wF * ${o};
+        for (int wF = 0; wF < ${l}; wF++) {
+          int xF = xFCorner + wF * ${i};
 
           if (xF < 0 || xF >= ${e.inDepth}) {
             continue;
           }
 
-          for (int wR = 0; wR < ${d}; wR++) {
-            int xR = xRCorner + wR * ${l};
+          for (int wR = 0; wR < ${p}; wR++) {
+            int xR = xRCorner + wR * ${c};
 
             if (xR < 0 || xR >= ${e.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${c}; wC++) {
+            for (int wC = 0; wC < ${d}; wC++) {
               int xC = xCCorner + wC * ${u};
 
               if (xC < 0 || xC >= ${e.inWidth}) {
@@ -2382,11 +2382,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                 dotProd += dot(xValues, wValues);
               }
 
-              if (${m===1}) {
+              if (${f===1}) {
                 dotProd +=
                   getX(batch, xF, xR, xC, ${h}) *
                   getW(wF, wR, wC, ${h}, d2);
-              } else if (${m===2}) {
+              } else if (${f===2}) {
                 vec2 xValues = vec2(
                   getX(batch, xF, xR, xC, ${h}),
                   getX(batch, xF, xR, xC, ${h} + 1)
@@ -2396,7 +2396,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                   getW(wF, wR, wC, ${h} + 1, d2)
                 );
                 dotProd += dot(xValues, wValues);
-              } else if (${m===3}) {
+              } else if (${f===3}) {
                 vec3 xValues = vec3(
                   getX(batch, xF, xR, xC, ${h}),
                   getX(batch, xF, xR, xC, ${h} + 1),
@@ -2414,27 +2414,27 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},kE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=En(this.outputShape.length);let s=e.padInfo.left,i=e.strideWidth,o=e.dilationWidth,l=e.filterHeight,u=e.filterWidth,p=u,d=`
+    `}},SE=class{constructor(e,t=!1,n=null,r=!1,s=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=Dn(this.outputShape.length);let a=e.padInfo.left,o=e.strideWidth,i=e.dilationWidth,c=e.filterHeight,u=e.filterWidth,l=u,p=`
        int xR; int xC; int xCOffset;
-       vec4 wTexel; vec4 previous; vec4 final;`;for(let f=0;f<u;f++)d+=`
-           vec4 xTexelC${f*2};
-           int xTexelC${f*2}Ready;
-           vec4 xTexelC${f*2+1};
-           int xTexelC${f*2+1}Ready;
-           vec4 xC${f};`;d+=`
-     for (int r = 0; r < ${l}; r++) {
+       vec4 wTexel; vec4 previous; vec4 final;`;for(let m=0;m<u;m++)p+=`
+           vec4 xTexelC${m*2};
+           int xTexelC${m*2}Ready;
+           vec4 xTexelC${m*2+1};
+           int xTexelC${m*2+1}Ready;
+           vec4 xC${m};`;p+=`
+     for (int r = 0; r < ${c}; r++) {
       for (int d1 = 0; d1 < ${e.inChannels}; d1 += 2) {
-       `;for(let f=0;f<u;f++)d+=`
-           xTexelC${f*2} = vec4(0.0);
-           xTexelC${f*2}Ready = 0;
-           xTexelC${f*2+1} = vec4(0.0);
-           xTexelC${f*2+1}Ready = 0;
-           xC${f} = vec4(0.0);`;d+=`
+       `;for(let m=0;m<u;m++)p+=`
+           xTexelC${m*2} = vec4(0.0);
+           xTexelC${m*2}Ready = 0;
+           xTexelC${m*2+1} = vec4(0.0);
+           xTexelC${m*2+1}Ready = 0;
+           xC${m} = vec4(0.0);`;p+=`
          xR = xRCorner + r * dilations[0];
          if (xR >=0 && xR < inDims[0]) {
-       `;for(let f=0;f<(p+1)/2;f++){let g=f*2;if(d+=`
-           xC = xCCorner + ${g*o};
-           `,i===1){if(g<u&&(s%2===1?(d+=`
+       `;for(let m=0;m<(l+1)/2;m++){let g=m*2;if(p+=`
+           xC = xCCorner + ${g*i};
+           `,o===1){if(g<u&&(a%2===1?(p+=`
                  xCOffset = xC + 1;
                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g}Ready == 0) {
                    xTexelC${g} = getX(batch, xR, xCOffset, d1);
@@ -2446,9 +2446,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                    }
                    xTexelC${g}Ready = 1;
                  }
-               `,o===1&&g>0?d+=`
+               `,i===1&&g>0?p+=`
                  xC${g} = vec4(xTexelC${g-2}.zw, xTexelC${g}.xy);
-                 `:d+=`
+                 `:p+=`
                    xCOffset = xC + 1 - 2;
 
                    if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -2464,7 +2464,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                    } else {
                      xC${g} = vec4(0.0, 0.0, xTexelC${g}.xy);
                    }
-                   `):d+=`
+                   `):p+=`
                  if (xC >= 0 && xC < inDims[1] && xTexelC${g}Ready == 0) {
                    xTexelC${g} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
@@ -2474,8 +2474,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                  }
 
                  xC${g} = xTexelC${g};
-                 `,g+1<u)){let y=s%2===0?v.nearestLargerEven(o):o;o%2===0&&s%2===1||o%2!==0&&s%2!==1?(d+=`
-                   xCOffset = xC + imod(pads[1], 2) + ${y};
+                 `,g+1<u)){let b=a%2===0?w.nearestLargerEven(i):i;i%2===0&&a%2===1||i%2!==0&&a%2!==1?(p+=`
+                   xCOffset = xC + imod(pads[1], 2) + ${b};
 
                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g+1}Ready == 0) {
                      xTexelC${g+1} = getX(batch, xR, xCOffset, d1);
@@ -2487,7 +2487,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                      }
                      xTexelC${g+1}Ready = 1;
                    }
-                   `,o>1?d+=`
+                   `,i>1?p+=`
                      xCOffset -= 2;
                      if (xCOffset >= 0 && xCOffset < inDims[1]) {
                       previous = getX(batch, xR, xCOffset, d1);
@@ -2495,12 +2495,12 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                      } else {
                       xC${g+1} = vec4(0.0, 0.0, xTexelC${g+1}.xy);
                      }
-                     `:d+=`
+                     `:p+=`
                      xC${g+1} = vec4(xTexelC${g}.zw, xTexelC${g+1}.xy);
-                     `):y===1?d+=`
+                     `):b===1?p+=`
                      xC${g+1} = xTexelC${g};
-                     `:d+=`
-                     xCOffset = xC + ${y};
+                     `:p+=`
+                     xCOffset = xC + ${b};
 
                      if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g+1}Ready == 0) {
                        xTexelC${g+1} = getX(batch, xR, xCOffset, d1);
@@ -2511,7 +2511,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                      }
 
                      xC${g+1} = xTexelC${g+1};
-                     `}}else g<u&&(s%2===1?(d+=`
+                     `}}else g<u&&(a%2===1?(p+=`
                  xCOffset = xC + 1 - strides[1];
                  if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${g}Ready == 0) {
                    xTexelC${g} = getX(batch, xR, xCOffset, d1);
@@ -2534,14 +2534,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                  }
 
                  xC${g} = vec4(xTexelC${g}.zw, xTexelC${g+1}.zw);
-               `,g+1<u&&(d+=`
+               `,g+1<u&&(p+=`
                    final = vec4(0.0);
                    xCOffset = xC + 1 + strides[1];
                    if(xCOffset >= 0 && xCOffset < inDims[1]) {
                      final = getX(batch, xR, xCOffset, d1);
                    }
                    xC${g+1} = vec4(xTexelC${g+1}.xy, final.xy);
-                 `)):(d+=`
+                 `)):(p+=`
                  if(xC >= 0 && xC < inDims[1] && xTexelC${g}Ready == 0) {
                    xTexelC${g} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
@@ -2561,36 +2561,36 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
                  xC${g} = vec4(
                    xTexelC${g}.xy, xTexelC${g+1}.xy);
-               `,g+1<u&&(d+=`
+               `,g+1<u&&(p+=`
                    xC${g+1} = vec4(xTexelC${g}.zw, xTexelC${g+1}.zw);
-                 `)));g<u&&(d+=`
+                 `)));g<u&&(p+=`
              wTexel = getW(r, ${g}, d1, d2);
              dotProd += xC${g}.xxzz * vec4(wTexel.xy, wTexel.xy);
              if(d1 + 1 < ${e.inChannels}) {
                dotProd += xC${g}.yyww * vec4(wTexel.zw, wTexel.zw);
              }
-           `,g+1<u&&(d+=`
+           `,g+1<u&&(p+=`
                wTexel = getW(r, ${g+1}, d1, d2);
                dotProd += xC${g+1}.xxzz * vec4(wTexel.xy, wTexel.xy);
                if(d1 + 1 < ${e.inChannels}) {
                  dotProd += xC${g+1}.yyww * vec4(wTexel.zw, wTexel.zw);
                }
-             `))}d+=`
+             `))}p+=`
      }
-   `,d+=`
+   `,p+=`
      }
-   `,d+=`
+   `,p+=`
      }
-   `;let c="",h="";n&&(a?c=`vec4 activation(vec4 a) {
+   `;let d="",h="";n&&(r?d=`vec4 activation(vec4 a) {
            vec4 b = getPreluActivationWeightsAtOutCoords();
            ${n}
-         }`:r?c=`vec4 activation(vec4 a) {
+         }`:s?d=`vec4 activation(vec4 a) {
            vec4 b = getLeakyreluAlphaAtOutCoords();
            ${n}
-         }`:c=`vec4 activation(vec4 x) {
+         }`:d=`vec4 activation(vec4 x) {
            ${n}
-         }`,h="result = activation(result);");let m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-       ${c}
+         }`,h="result = activation(result);");let f=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
+       ${d}
 
        void main() {
          ivec4 coords = getOutputCoords();
@@ -2603,22 +2603,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
          //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
          vec4 dotProd = vec4(0.000000000000001);
 
-         ${d}
+         ${p}
 
          vec4 result = dotProd - vec4(0.000000000000001);
-         ${m}
+         ${f}
          ${h}
          setOutput(result);
        }
-     `}},fQ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec4"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=En(this.outputShape.length);let{dataFormat:n}=t,a=_n(),r=n==="channelsLast",s=r?1:2,i=r?2:3,o=this.enableShapeUniforms?"if(blockIndex < outShape[2] && pos < outShape[1]) {":`if(blockIndex < ${e[2]} && pos < ${e[1]}) {`,l="";for(let u=0;u<=1;u++)for(let p=0;p<=1;p++)l+=`
-          blockIndex = rc.z + ${p};
+     `}},yQ=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec4"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=Dn(this.outputShape.length);let{dataFormat:n}=t,r=$n(),s=n==="channelsLast",a=s?1:2,o=s?2:3,i=this.enableShapeUniforms?"if(blockIndex < outShape[2] && pos < outShape[1]) {":`if(blockIndex < ${e[2]} && pos < ${e[1]}) {`,c="";for(let u=0;u<=1;u++)for(let l=0;l<=1;l++)c+=`
+          blockIndex = rc.z + ${l};
           pos = rc.y + ${u};
 
-          ${o}
+          ${i}
             offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];
             d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);
 
-            if(d0 < inputShape[${s}] && d0 >= 0) {
+            if(d0 < inputShape[${a}] && d0 >= 0) {
               // Use custom imod instead mod. On Intel GPU, mod may generate
               // unexpected value.
               // https://github.com/tensorflow/tfjs/issues/5447
@@ -2626,18 +2626,18 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /
                   inChannels);
 
-              if(d1 < inputShape[${i}] && d1 >= 0) {
+              if(d1 < inputShape[${o}] && d1 >= 0) {
 
                 ch = imod(pos, inChannels);
 
-                if (${r}) {
+                if (${s}) {
                   innerDims = vec2(d1, ch);
-                  result[${u*2+p}] = getChannel(
+                  result[${u*2+l}] = getChannel(
                     getA(rc.x, d0, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 } else {
                   innerDims = vec2(d0, d1);
-                  result[${u*2+p}] = getChannel(
+                  result[${u*2+l}] = getChannel(
                     getA(rc.x, ch, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 }
@@ -2653,11 +2653,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         int blockIndex, pos, offsetY, d0, offsetX, d1, ch;
         vec2 innerDims;
 
-        ${l}
+        ${c}
 
-        ${a.output} = result;
+        ${r.output} = result;
       }
-    `}};function am(e,t){let n=e.length;return n>=3?t?[...e.slice(0,-3),e[n-3]*e[n-2],e[n-1]]:[...e.slice(0,-3),e[n-3],e[n-2]*e[n-1]]:!t&&n===1&&e[0]>1?[e[0],1]:null}function IE({x:e,filter:t,convInfo:n,backend:a,bias:r=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l=e.shape,u=a.texData.get(e.dataId),p=n.inChannels,d=l[0]*l[1]*l[2],c=n.outChannels,h=n.dataFormat==="channelsLast",m=!1,f=!1,g,y=[];if(s!=null){let b=am(s.shape,h);b!=null&&(s=de({inputs:{x:s},backend:a,attrs:{shape:b}}),y.push(s))}if(r!=null){let b=am(r.shape,h);b!=null&&(r=de({inputs:{x:r},backend:a,attrs:{shape:b}}),y.push(r))}if(!((d===1||c===1)&&p>fE)&&u.isPacked&&h&&u.texture!=null&&l[2]%2!==0&&v.arraysEqual(u.shape.slice(-3),l.slice(-3))){let b=l[0]*l[1]*(l[2]+1),x={dataId:e.dataId,shape:[1,b,n.inChannels],dtype:e.dtype},w=u.shape;u.shape=u.shape.slice(),u.shape[u.shape.length-2]++,v.assert(sc(u.shape,x.shape),()=>`packed reshape ${u.shape} to ${x.shape} isn't free`);let I=de({inputs:{x:t},backend:a,attrs:{shape:[1,n.inChannels,n.outChannels]}});y.push(I);let T=nm({a:x,b:I,backend:a,transposeA:m,transposeB:f,bias:r,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),C=a.texData.get(T.dataId);v.assert(C.isPacked,()=>"batchMatMul result is expected to be packed"),u.shape=w,C.shape=n.outShape,g=aa({inputs:{x:T},backend:a}),g.shape=n.outShape,y.push(T)}else{let b=n.outHeight*n.outWidth,x=de({inputs:{x:e},backend:a,attrs:{shape:h?[n.batchSize,b,n.inChannels]:[n.batchSize,n.inChannels,b]}}),w=de({inputs:{x:t},backend:a,attrs:{shape:[1,n.inChannels,n.outChannels]}}),I=nm({a:h?x:w,b:h?w:x,transposeA:!h,transposeB:f,backend:a,bias:r,activation:o,preluActivationWeights:s,leakyreluAlpha:i});g=de({inputs:{x:I},backend:a,attrs:{shape:n.outShape}}),y.push(x),y.push(w),y.push(I)}for(let b of y)a.disposeIntermediateTensorInfo(b);return g}function SE({x:e,filter:t,convInfo:n,backend:a,bias:r=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let{filterWidth:l,filterHeight:u,inChannels:p,outWidth:d,outHeight:c,dataFormat:h}=n,m=h==="channelsLast",f=l*u*p,g=c*d,y=[n.batchSize,f,g],b=!0,x=!1,w=[];if(s!=null){let K=am(s.shape,m);K!=null&&(s=de({inputs:{x:s},backend:a,attrs:{shape:K}}),w.push(s))}if(r!=null){let K=am(r.shape,m);K!=null&&(r=de({inputs:{x:r},backend:a,attrs:{shape:K}}),w.push(r))}let I=de({inputs:{x:t},backend:a,attrs:{shape:[1,f,v.sizeFromShape(t.shape)/f]}});w.push(I);let T=new fQ(y,n),C=[e.shape,[n.padInfo.top,n.padInfo.left],[n.strideHeight,n.strideWidth],[n.dilationHeight,n.dilationWidth],[n.inChannels],[n.filterWidth*n.inChannels],[n.outWidth]],E=a.runWebGLProgram(T,[e],"float32",C),A=de({inputs:{x:E},backend:a,attrs:{shape:y}});w.push(E),w.push(A);let R=r!=null,F=s!=null,S=o==="leakyrelu",M=o?ic(o,!0):null,B=new mE(m?A.shape:I.shape,m?I.shape:A.shape,m?[n.batchSize,g,n.outChannels]:[n.batchSize,n.outChannels,g],b,x,R,M,F,S),U=m?[A,I]:[I,A];if(r&&U.push(r),F&&U.push(s),S){let K=a.makeTensorInfo([],"float32",v.createScalarValue(i,"float32"));U.push(K),w.push(K)}let G=a.runWebGLProgram(B,U,"float32"),q=de({inputs:{x:G},backend:a,attrs:{shape:n.outShape}});w.push(G);for(let K of w)a.disposeIntermediateTensorInfo(K);return q}function gQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:p}=a,d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,s.shape,i,u,o,p,!1,d),h;if(c.filterHeight===1&&c.filterWidth===1&&c.dilationHeight===1&&c.dilationWidth===1&&c.strideHeight===1&&c.strideWidth===1&&(c.padInfo.type==="SAME"||c.padInfo.type==="VALID"))h=IE({x:r,filter:s,convInfo:c,backend:n});else if(c.strideWidth<=2&&d==="channelsLast"&&H().getBool("WEBGL_EXP_CONV")){let f=new kE(c),g=[[c.padInfo.top,c.padInfo.left],[c.strideHeight,c.strideWidth],[c.dilationHeight,c.dilationWidth],[c.inHeight,c.inWidth]];h=n.runWebGLProgram(f,[r,s],"float32",g)}else if(H().getBool("WEBGL_CONV_IM2COL"))h=SE({x:r,filter:s,convInfo:c,backend:n});else{let f=new wE(c);h=n.runWebGLProgram(f,[r,s],"float32")}let m=de({inputs:{x:h},backend:n,attrs:{shape:c.outShape}});return n.disposeIntermediateTensorInfo(h),m}var yQ={kernelName:wi,backendName:"webgl",kernelFunc:gQ},bQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,a=e.padInfo.top,r=e.padInfo.left,s=e.dataFormat==="channelsLast";this.userCode=`
+    `}};function sf(e,t){let n=e.length;return n>=3?t?[...e.slice(0,-3),e[n-3]*e[n-2],e[n-1]]:[...e.slice(0,-3),e[n-3],e[n-2]*e[n-1]]:!t&&n===1&&e[0]>1?[e[0],1]:null}function TE({x:e,filter:t,convInfo:n,backend:r,bias:s=null,preluActivationWeights:a=null,leakyreluAlpha:o=0,activation:i=null}){let c=e.shape,u=r.texData.get(e.dataId),l=n.inChannels,p=c[0]*c[1]*c[2],d=n.outChannels,h=n.dataFormat==="channelsLast",f=!1,m=!1,g,b=[];if(a!=null){let x=sf(a.shape,h);x!=null&&(a=he({inputs:{x:a},backend:r,attrs:{shape:x}}),b.push(a))}if(s!=null){let x=sf(s.shape,h);x!=null&&(s=he({inputs:{x:s},backend:r,attrs:{shape:x}}),b.push(s))}if(!((p===1||d===1)&&l>bE)&&u.isPacked&&h&&u.texture!=null&&c[2]%2!==0&&w.arraysEqual(u.shape.slice(-3),c.slice(-3))){let x=c[0]*c[1]*(c[2]+1),k={dataId:e.dataId,shape:[1,x,n.inChannels],dtype:e.dtype},S=u.shape;u.shape=u.shape.slice(),u.shape[u.shape.length-2]++,w.assert(ad(u.shape,k.shape),()=>`packed reshape ${u.shape} to ${k.shape} isn't free`);let C=he({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}});b.push(C);let E=rf({a:k,b:C,backend:r,transposeA:f,transposeB:m,bias:s,activation:i,preluActivationWeights:a,leakyreluAlpha:o}),$=r.texData.get(E.dataId);w.assert($.isPacked,()=>"batchMatMul result is expected to be packed"),u.shape=S,$.shape=n.outShape,g=sr({inputs:{x:E},backend:r}),g.shape=n.outShape,b.push(E)}else{let x=n.outHeight*n.outWidth,k=he({inputs:{x:e},backend:r,attrs:{shape:h?[n.batchSize,x,n.inChannels]:[n.batchSize,n.inChannels,x]}}),S=he({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}}),C=rf({a:h?k:S,b:h?S:k,transposeA:!h,transposeB:m,backend:r,bias:s,activation:i,preluActivationWeights:a,leakyreluAlpha:o});g=he({inputs:{x:C},backend:r,attrs:{shape:n.outShape}}),b.push(k),b.push(S),b.push(C)}for(let x of b)r.disposeIntermediateTensorInfo(x);return g}function CE({x:e,filter:t,convInfo:n,backend:r,bias:s=null,preluActivationWeights:a=null,leakyreluAlpha:o=0,activation:i=null}){let{filterWidth:c,filterHeight:u,inChannels:l,outWidth:p,outHeight:d,dataFormat:h}=n,f=h==="channelsLast",m=c*u*l,g=d*p,b=[n.batchSize,m,g],y=!0,v=!1,x=[];if(a!=null){let H=sf(a.shape,f);H!=null&&(a=he({inputs:{x:a},backend:r,attrs:{shape:H}}),x.push(a))}if(s!=null){let H=sf(s.shape,f);H!=null&&(s=he({inputs:{x:s},backend:r,attrs:{shape:H}}),x.push(s))}let k=he({inputs:{x:t},backend:r,attrs:{shape:[1,m,w.sizeFromShape(t.shape)/m]}});x.push(k);let S=new yQ(b,n),C=[e.shape,[n.padInfo.top,n.padInfo.left],[n.strideHeight,n.strideWidth],[n.dilationHeight,n.dilationWidth],[n.inChannels],[n.filterWidth*n.inChannels],[n.outWidth]],E=r.runWebGLProgram(S,[e],"float32",C),$=he({inputs:{x:E},backend:r,attrs:{shape:b}});x.push(E),x.push($);let F=s!=null,A=a!=null,R=i==="leakyrelu",T=i?od(i,!0):null,L=new gE(f?$.shape:k.shape,f?k.shape:$.shape,f?[n.batchSize,g,n.outChannels]:[n.batchSize,n.outChannels,g],y,v,F,T,A,R),V=f?[$,k]:[k,$];if(s&&V.push(s),A&&V.push(a),R){let H=r.makeTensorInfo([],"float32",w.createScalarValue(o,"float32"));V.push(H),x.push(H)}let G=r.runWebGLProgram(L,V,"float32"),j=he({inputs:{x:G},backend:r,attrs:{shape:n.outShape}});x.push(G);for(let H of x)r.disposeIntermediateTensorInfo(H);return j}function vQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dataFormat:c,dilations:u,dimRoundingMode:l}=r,p=N.convertConv2DDataFormat(c),d=N.computeConv2DInfo(s.shape,a.shape,o,u,i,l,!1,p),h;if(d.filterHeight===1&&d.filterWidth===1&&d.dilationHeight===1&&d.dilationWidth===1&&d.strideHeight===1&&d.strideWidth===1&&(d.padInfo.type==="SAME"||d.padInfo.type==="VALID"))h=TE({x:s,filter:a,convInfo:d,backend:n});else if(d.strideWidth<=2&&p==="channelsLast"&&q().getBool("WEBGL_EXP_CONV")){let m=new SE(d),g=[[d.padInfo.top,d.padInfo.left],[d.strideHeight,d.strideWidth],[d.dilationHeight,d.dilationWidth],[d.inHeight,d.inWidth]];h=n.runWebGLProgram(m,[s,a],"float32",g)}else if(q().getBool("WEBGL_CONV_IM2COL"))h=CE({x:s,filter:a,convInfo:d,backend:n});else{let m=new kE(d);h=n.runWebGLProgram(m,[s,a],"float32")}let f=he({inputs:{x:h},backend:n,attrs:{shape:d.outShape}});return n.disposeIntermediateTensorInfo(h),f}var xQ={kernelName:ko,backendName:"webgl",kernelFunc:vQ},wQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,s=e.padInfo.left,a=e.dataFormat==="channelsLast";this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
@@ -2671,20 +2671,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         for (int b = 0; b < ${e.batchSize}; b++) {
           for (int yR = 0; yR < ${e.outHeight}; yR++) {
-            int xR = wR + yR * ${t} - ${a};
+            int xR = wR + yR * ${t} - ${r};
 
             if (xR < 0 || xR >= ${e.inHeight}) {
               continue;
             }
 
             for (int yC = 0; yC < ${e.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${r};
+              int xC = wC + yC * ${n} - ${s};
 
               if (xC < 0 || xC >= ${e.inWidth}) {
                 continue;
               }
 
-              if (${s}) {
+              if (${a}) {
                 float dyValue = getDy(b, yR, yC, d2);
                 float xValue = getX(b, xR, xC, d1);
                 dotProd += (xValue * dyValue);
@@ -2699,15 +2699,15 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},xQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=e.dataFormat==="channelsLast",i=t-1-e.padInfo.top,o=n-1-e.padInfo.left,l=s?1:2,u=s?2:3,p=s?3:1;this.userCode=`
-      const ivec2 pads = ivec2(${i}, ${o});
+    `}},IQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,s=e.strideWidth,a=e.dataFormat==="channelsLast",o=t-1-e.padInfo.top,i=n-1-e.padInfo.left,c=a?1:2,u=a?2:3,l=a?3:1;this.userCode=`
+      const ivec2 pads = ivec2(${o}, ${i});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d1 = coords[${p}];
+        int d1 = coords[${l}];
 
-        ivec2 dyCorner = ivec2(coords[${l}], coords[${u}]) - pads;
+        ivec2 dyCorner = ivec2(coords[${c}], coords[${u}]) - pads;
         int dyRCorner = dyCorner.x;
         int dyCCorner = dyCorner.y;
 
@@ -2715,7 +2715,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
         for (int wR = 0; wR < ${t}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+          float dyR = float(dyRCorner + wR) / ${r}.0;
 
           if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
@@ -2725,7 +2725,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           int wRPerm = ${t} - 1 - wR;
 
           for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+            float dyC = float(dyCCorner + wC) / ${s}.0;
 
             if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                 fract(dyC) > 0.0) {
@@ -2737,7 +2737,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
             for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
 
-              if (${s}) {
+              if (${a}) {
                 float xValue = getDy(batch, idyR, idyC, d2);
                 float wValue = getW(wRPerm, wCPerm, d1, d2);
                 dotProd += xValue * wValue;
@@ -2752,7 +2752,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},vQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideDepth,n=e.strideHeight,a=e.strideWidth,r=e.padInfo.front,s=e.padInfo.top,i=e.padInfo.left;this.userCode=`
+    `}},kQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,s=e.padInfo.front,a=e.padInfo.top,o=e.padInfo.left;this.userCode=`
       void main() {
         ivec5 coords = getOutputCoords();
         int wF = coords.x;
@@ -2765,21 +2765,21 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         for (int b = 0; b < ${e.batchSize}; b++) {
           for (int yF = 0; yF < ${e.outDepth}; yF++) {
-            int xF = wF + yF * ${t} - ${r};
+            int xF = wF + yF * ${t} - ${s};
 
             if (xF < 0 || xF >= ${e.inDepth}) {
               continue;
             }
 
             for (int yR = 0; yR < ${e.outHeight}; yR++) {
-              int xR = wR + yR * ${n} - ${s};
+              int xR = wR + yR * ${n} - ${a};
 
               if (xR < 0 || xR >= ${e.inHeight}) {
                 continue;
               }
 
               for (int yC = 0; yC < ${e.outWidth}; yC++) {
-                int xC = wC + yC * ${a} - ${i};
+                int xC = wC + yC * ${r} - ${o};
 
                 if (xC < 0 || xC >= ${e.inWidth}) {
                   continue;
@@ -2794,8 +2794,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},wQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,a=e.filterWidth,r=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=t-1-e.padInfo.front,l=n-1-e.padInfo.top,u=a-1-e.padInfo.left;this.userCode=`
-      const ivec3 pads = ivec3(${o}, ${l}, ${u});
+    `}},SQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,s=e.strideDepth,a=e.strideHeight,o=e.strideWidth,i=t-1-e.padInfo.front,c=n-1-e.padInfo.top,u=r-1-e.padInfo.left;this.userCode=`
+      const ivec3 pads = ivec3(${i}, ${c}, ${u});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2810,7 +2810,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         float dotProd = 0.0;
         for (int wF = 0; wF < ${t}; wF++) {
-          float dyF = float(dyFCorner + wF) / ${r}.0;
+          float dyF = float(dyFCorner + wF) / ${s}.0;
 
           if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {
             continue;
@@ -2820,7 +2820,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           int wFPerm = ${t} - 1 - wF;
 
           for (int wR = 0; wR < ${n}; wR++) {
-            float dyR = float(dyRCorner + wR) / ${s}.0;
+            float dyR = float(dyRCorner + wR) / ${a}.0;
 
             if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
               fract(dyR) > 0.0) {
@@ -2830,8 +2830,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
             int wRPerm = ${n} - 1 - wR;
 
-            for (int wC = 0; wC < ${a}; wC++) {
-              float dyC = float(dyCCorner + wC) / ${i}.0;
+            for (int wC = 0; wC < ${r}; wC++) {
+              float dyC = float(dyCCorner + wC) / ${o}.0;
 
               if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                   fract(dyC) > 0.0) {
@@ -2839,7 +2839,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               }
               int idyC = int(dyC);
 
-              int wCPerm = ${a} - 1 - wC;
+              int wCPerm = ${r} - 1 - wC;
 
               for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
                 float xValue = getDy(batch, idyF, idyR, idyC, d2);
@@ -2851,14 +2851,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function kQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:p}=a,d=N.convertConv2DDataFormat(l),c=N.computeConv2DInfo(r.shape,p,i,1,o,u,!1,d),h=new bQ(c);return n.runWebGLProgram(h,[r,s],"float32")}var IQ={kernelName:fm,backendName:"webgl",kernelFunc:kQ};function SQ(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:p}=a,d=N.convertConv2DDataFormat(u),c=N.computeConv2DInfo(i,s.shape,o,1,l,p,!1,d),h=new xQ(c);return n.runWebGLProgram(h,[r,s],"float32")}var TQ={kernelName:ki,backendName:"webgl",kernelFunc:SQ};function NQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a,u=N.computeConv3DInfo(r.shape,s.shape,i,l,o),p=new mQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var CQ={kernelName:mc,backendName:"webgl",kernelFunc:NQ};function _Q(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,pad:o,filterShape:l}=a,u=N.computeConv3DInfo(r.shape,l,i,1,o),p=new vQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var EQ={kernelName:gm,backendName:"webgl",kernelFunc:_Q};function AQ(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{pad:i,strides:o,inputShape:l}=a,u=N.computeConv3DInfo(l,s.shape,o,1,i),p=new wQ(u);return n.runWebGLProgram(p,[r,s],"float32")}var $Q={kernelName:ym,backendName:"webgl",kernelFunc:AQ},FQ=tp+`
+    `}};function TQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,pad:i,dataFormat:c,dimRoundingMode:u,filterShape:l}=r,p=N.convertConv2DDataFormat(c),d=N.computeConv2DInfo(s.shape,l,o,1,i,u,!1,p),h=new wQ(d);return n.runWebGLProgram(h,[s,a],"float32")}var CQ={kernelName:bf,backendName:"webgl",kernelFunc:TQ};function NQ(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{inputShape:o,strides:i,pad:c,dataFormat:u,dimRoundingMode:l}=r,p=N.convertConv2DDataFormat(u),d=N.computeConv2DInfo(o,a.shape,i,1,c,l,!1,p),h=new IQ(d);return n.runWebGLProgram(h,[s,a],"float32")}var _Q={kernelName:So,backendName:"webgl",kernelFunc:NQ};function EQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dilations:c}=r,u=N.computeConv3DInfo(s.shape,a.shape,o,c,i),l=new bQ(u);return n.runWebGLProgram(l,[s,a],"float32")}var AQ={kernelName:fd,backendName:"webgl",kernelFunc:EQ};function $Q(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,pad:i,filterShape:c}=r,u=N.computeConv3DInfo(s.shape,c,o,1,i),l=new kQ(u);return n.runWebGLProgram(l,[s,a],"float32")}var DQ={kernelName:yf,backendName:"webgl",kernelFunc:$Q};function FQ(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{pad:o,strides:i,inputShape:c}=r,u=N.computeConv3DInfo(c,a.shape,i,1,o),l=new SQ(u);return n.runWebGLProgram(l,[s,a],"float32")}var RQ={kernelName:vf,backendName:"webgl",kernelFunc:FQ},PQ=tl+`
   return cos(x);
-`,DQ=Ye({opSnippet:FQ}),RQ={kernelName:Ii,backendName:"webgl",kernelFunc:DQ},MQ=`
+`,OQ=Ye({opSnippet:PQ}),MQ={kernelName:To,backendName:"webgl",kernelFunc:OQ},LQ=`
   float e2x = exp(-x);
   return (e2x + 1.0 / e2x) / 2.0;
-`,PQ=Ye({opSnippet:MQ}),OQ={kernelName:Si,backendName:"webgl",kernelFunc:PQ},LQ=class{constructor(e,t,n,a,r){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[s,i,o,l]=e,[u]=t,[p,d]=n;this.outputShape=[u,p,d,l];let c=a==="bilinear"?1:0,[h,m]=[`${i-1}.0`,`${o-1}.0`],[f,g,y]=p>1?[`${(i-1)/(p-1)}`,"(y2-y1) * height_ratio",`y1*${h} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${h}`],[b,x,w]=d>1?[`${(o-1)/(d-1)}`,"(x2-x1) * width_ratio",`x1*${m} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${m}`];this.userCode=`
-      const float height_ratio = float(${f});
-      const float width_ratio = float(${b});
+`,zQ=Ye({opSnippet:LQ}),BQ={kernelName:Co,backendName:"webgl",kernelFunc:zQ},WQ=class{constructor(e,t,n,r,s){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[a,o,i,c]=e,[u]=t,[l,p]=n;this.outputShape=[u,l,p,c];let d=r==="bilinear"?1:0,[h,f]=[`${o-1}.0`,`${i-1}.0`],[m,g,b]=l>1?[`${(o-1)/(l-1)}`,"(y2-y1) * height_ratio",`y1*${h} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${h}`],[y,v,x]=p>1?[`${(i-1)/(p-1)}`,"(x2-x1) * width_ratio",`x1*${f} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${f}`];this.userCode=`
+      const float height_ratio = float(${m});
+      const float width_ratio = float(${y});
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -2874,26 +2874,26 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         // get image in batch index
         int bInd = round(getBoxInd(b));
-        if(bInd < 0 || bInd >= ${s}) {
+        if(bInd < 0 || bInd >= ${a}) {
           return;
         }
 
         float height_scale = ${g};
-        float width_scale = ${x};
+        float width_scale = ${v};
 
-        float in_y = ${y};
+        float in_y = ${b};
         if( in_y < 0.0 || in_y > ${h} ) {
-          setOutput(float(${r}));
+          setOutput(float(${s}));
           return;
         }
-        float in_x = ${w};
-        if( in_x < 0.0 || in_x > ${m} ) {
-          setOutput(float(${r}));
+        float in_x = ${x};
+        if( in_x < 0.0 || in_x > ${f} ) {
+          setOutput(float(${s}));
           return;
         }
 
         vec2 sourceFracIndexCR = vec2(in_x,in_y);
-        if(${c} == 1) {
+        if(${d} == 1) {
           // Compute the four integer indices.
           ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);
           ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));
@@ -2917,20 +2917,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           setOutput(newValue);
         }
       }
-    `}},zQ=e=>{let{inputs:t,backend:n,attrs:a}=e,{image:r,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=a,p=new LQ(r.shape,s.shape,o,l,u);return n.runWebGLProgram(p,[r,s,i],"float32")},WQ={kernelName:jl,backendName:"webgl",kernelFunc:zQ},lc;(function(e){e.Prod="*",e.Sum="+"})(lc||(lc={}));var NI=class{constructor(e,t,n,a){this.op=e,this.outputShape=t,this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}];let r=this.outputShape.length,s=this.op===lc.Prod?"1.0":"0.0",i=n?s:`getX(${CI(r,"coords",this.op)})`,o=this.outputShape[this.outputShape.length-1],l="",u="";n?(l=a?`end != ${o-1}`:"end != 0",u=a?"end + 1":"end - 1"):(l=a?`end + pow2 < ${o}`:"end >= pow2",u=a?"end + pow2":"end - pow2"),this.userCode=`
+    `}},VQ=e=>{let{inputs:t,backend:n,attrs:r}=e,{image:s,boxes:a,boxInd:o}=t,{cropSize:i,method:c,extrapolationValue:u}=r,l=new WQ(s.shape,a.shape,i,c,u);return n.runWebGLProgram(l,[s,a,o],"float32")},UQ={kernelName:qc,backendName:"webgl",kernelFunc:VQ},cd;(function(e){e.Prod="*",e.Sum="+"})(cd||(cd={}));var N1=class{constructor(e,t,n,r){this.op=e,this.outputShape=t,this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}];let s=this.outputShape.length,a=this.op===cd.Prod?"1.0":"0.0",o=n?a:`getX(${_1(s,"coords",this.op)})`,i=this.outputShape[this.outputShape.length-1],c="",u="";n?(c=r?`end != ${i-1}`:"end != 0",u=r?"end + 1":"end - 1"):(c=r?`end + pow2 < ${i}`:"end >= pow2",u=r?"end + pow2":"end - pow2"),this.userCode=`
       void main() {
-        ${gt(r)} coords = getOutputCoords();
-        int end = ${_I(r,"coords",this.op)};
-        float val = ${i};
+        ${mt(s)} coords = getOutputCoords();
+        int end = ${E1(s,"coords",this.op)};
+        float val = ${o};
         int pow2 = int(pow(2.0, index));
-        if (${l}) {
+        if (${c}) {
           int idx = ${u};
-          ${_I(r,"coords",this.op)} = idx;
-          val ${this.op}= getX(${CI(r,"coords",this.op)});
+          ${E1(s,"coords",this.op)} = idx;
+          val ${this.op}= getX(${_1(s,"coords",this.op)});
         }
         setOutput(val);
       }
-    `}};function CI(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.x, ${t}.y`;if(e===3)return`${t}.x, ${t}.y, ${t}.z`;if(e===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function _I(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.y`;if(e===3)return`${t}.z`;if(e===4)return`${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function TE(e,t,n,a,r,s){let i=t.shape.length,o=N.getAxesPermutation([a],i),l=t;o!=null&&(l=Sn({inputs:{x:t},backend:n,attrs:{perm:o}}));let u=N.getInnerMostAxes(1,i)[0];if(u!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${a}`);let p=l.shape[u],d=aa({inputs:{x:l},backend:n});for(let c=0;c<=Math.ceil(Math.log2(p))-1;c++){let h=new NI(e,l.shape,!1,s),m=[[c]],f=d;d=n.runWebGLProgram(h,[d],d.dtype,m),n.disposeIntermediateTensorInfo(f)}if(r){let c=new NI(e,l.shape,r,s),h=d;d=n.runWebGLProgram(c,[d],d.dtype),n.disposeIntermediateTensorInfo(h)}if(o!=null){let c=N.getUndoAxesPermutation(o),h=Sn({inputs:{x:d},backend:n,attrs:{perm:c}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(l),h}return d}function BQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;return TE(lc.Prod,r,n,s,i,o)}var VQ={kernelName:Hl,backendName:"webgl",kernelFunc:BQ};function UQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a;return TE(lc.Sum,r,n,s,i,o)}var GQ={kernelName:Ti,backendName:"webgl",kernelFunc:UQ};function HQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,weights:s}=t,{size:i,binaryOutput:o}=a;if(r.shape.length===1){let l=n.readSync(r.dataId),u=n.readSync(s.dataId),p=rE(l,u,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,p)}else if(r.shape.length===2){let l=n.bufferSync(r),u=n.bufferSync(s),p=t7(l,u,i,o);return n.makeTensorInfo(p.shape,s.dtype,p.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${r.shape.length}.`)}var jQ={kernelName:bm,backendName:"webgl",kernelFunc:HQ},qQ=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=n,this.userCode=`
+    `}};function _1(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.x, ${t}.y`;if(e===3)return`${t}.x, ${t}.y, ${t}.z`;if(e===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function E1(e,t,n){if(e===1)return`${t}`;if(e===2)return`${t}.y`;if(e===3)return`${t}.z`;if(e===4)return`${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function NE(e,t,n,r,s,a){let o=t.shape.length,i=N.getAxesPermutation([r],o),c=t;i!=null&&(c=Nn({inputs:{x:t},backend:n,attrs:{perm:i}}));let u=N.getInnerMostAxes(1,o)[0];if(u!==o-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${r}`);let l=c.shape[u],p=sr({inputs:{x:c},backend:n});for(let d=0;d<=Math.ceil(Math.log2(l))-1;d++){let h=new N1(e,c.shape,!1,a),f=[[d]],m=p;p=n.runWebGLProgram(h,[p],p.dtype,f),n.disposeIntermediateTensorInfo(m)}if(s){let d=new N1(e,c.shape,s,a),h=p;p=n.runWebGLProgram(d,[p],p.dtype),n.disposeIntermediateTensorInfo(h)}if(i!=null){let d=N.getUndoAxesPermutation(i),h=Nn({inputs:{x:p},backend:n,attrs:{perm:d}});return n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(c),h}return p}function GQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r;return NE(cd.Prod,s,n,a,o,i)}var HQ={kernelName:Hc,backendName:"webgl",kernelFunc:GQ};function qQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r;return NE(cd.Sum,s,n,a,o,i)}var jQ={kernelName:No,backendName:"webgl",kernelFunc:qQ};function KQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,weights:a}=t,{size:o,binaryOutput:i}=r;if(s.shape.length===1){let c=n.readSync(s.dataId),u=n.readSync(a.dataId),l=oE(c,u,a.dtype,a.shape,o);return n.makeTensorInfo([o],a.dtype,l)}else if(s.shape.length===2){let c=n.bufferSync(s),u=n.bufferSync(a),l=s9(c,u,o,i);return n.makeTensorInfo(l.shape,a.dtype,l.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${s.shape.length}.`)}var XQ={kernelName:xf,backendName:"webgl",kernelFunc:KQ},YQ=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=n,this.userCode=`
     void main() {
       ivec4 coords = getOutputCoords();
       int b = coords[0];
@@ -2949,26 +2949,26 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
       float result = ${this.getInputSamplingString()};
       setOutput(result);
     }
-  `}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function KQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockSize:s,dataFormat:i}=a,o=r.shape[0],l=i==="NHWC"?r.shape[1]:r.shape[2],u=i==="NHWC"?r.shape[2]:r.shape[3],p=i==="NHWC"?r.shape[3]:r.shape[1],d=l*s,c=u*s,h=p/(s*s),m=i==="NHWC"?[o,d,c,h]:[o,h,d,c],f=new qQ(m,s,i);return n.runWebGLProgram(f,[r],r.dtype)}var XQ={kernelName:ql,backendName:"webgl",kernelFunc:KQ},NE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=En(this.outputShape.length);let s=e.filterHeight,i=e.filterWidth,o=e.outChannels/e.inChannels,l="",u="";n&&(a?l=`float activation(float a) {
+  `}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function ZQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockSize:a,dataFormat:o}=r,i=s.shape[0],c=o==="NHWC"?s.shape[1]:s.shape[2],u=o==="NHWC"?s.shape[2]:s.shape[3],l=o==="NHWC"?s.shape[3]:s.shape[1],p=c*a,d=u*a,h=l/(a*a),f=o==="NHWC"?[i,p,d,h]:[i,h,p,d],m=new YQ(f,a,o);return n.runWebGLProgram(m,[s],s.dtype)}var JQ={kernelName:jc,backendName:"webgl",kernelFunc:ZQ},_E=class{constructor(e,t=!1,n=null,r=!1,s=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=Dn(this.outputShape.length);let a=e.filterHeight,o=e.filterWidth,i=e.outChannels/e.inChannels,c="",u="";n&&(r?c=`float activation(float a) {
           float b = getPreluActivationWeightsAtOutCoords();
           ${n}
-        }`:r?l=`float activation(float a) {
+        }`:s?c=`float activation(float a) {
           float b = getLeakyreluAlphaAtOutCoords();
           ${n}
-        }`:l=`
+        }`:c=`
           float activation(float x) {
             ${n}
           }
-        `,u="result = activation(result);");let p=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${l}
+        `,u="result = activation(result);");let l=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
+      ${c}
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${o};
-        int q = d2 - d1 * ${o};
+        int d1 = d2 / ${i};
+        int q = d2 - d1 * ${i};
 
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
@@ -2977,14 +2977,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
         // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.
-        for (int wR = 0; wR < ${s}; wR++) {
+        for (int wR = 0; wR < ${a}; wR++) {
           int xR = xRCorner + wR * dilations[0];
 
           if (xR < 0 || xR >= inDims[0]) {
             continue;
           }
 
-          for (int wC = 0; wC < ${i}; wC++) {
+          for (int wC = 0; wC < ${o}; wC++) {
             int xC = xCCorner + wC * dilations[1];
 
             if (xC < 0 || xC >= inDims[1]) {
@@ -2998,44 +2998,44 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
 
         float result = dotProd;
-        ${p}
+        ${l}
         ${u}
         setOutput(result);
       }
-    `}},CE=class{constructor(e,t=!1,n=null,a=!1,r=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=En(this.outputShape.length);let s=e.outChannels/e.inChannels,i=e.padInfo.left,o=e.strideWidth,l=e.dilationWidth,u=e.filterHeight,p=e.filterWidth,d=p,c=`
+    `}},EE=class{constructor(e,t=!1,n=null,r=!1,s=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=Dn(this.outputShape.length);let a=e.outChannels/e.inChannels,o=e.padInfo.left,i=e.strideWidth,c=e.dilationWidth,u=e.filterHeight,l=e.filterWidth,p=l,d=`
       int xR; int xC; int xCOffset;
-      vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<p;g++)c+=`
+      vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<l;g++)d+=`
           vec4 xTexelC${g*2};
           int xTexelC${g*2}Ready;
           vec4 xTexelC${g*2+1};
           int xTexelC${g*2+1}Ready;
-          vec4 xC${g};`;c+=`
+          vec4 xC${g};`;d+=`
     for (int r = 0; r < ${u}; r++) {
-      `;for(let g=0;g<p;g++)c+=`
+      `;for(let g=0;g<l;g++)d+=`
           xTexelC${g*2} = vec4(0.0);
           xTexelC${g*2}Ready = 0;
           xTexelC${g*2+1} = vec4(0.0);
           xTexelC${g*2+1}Ready = 0;
-          xC${g} = vec4(0.0);`;c+=`
+          xC${g} = vec4(0.0);`;d+=`
         xR = xRCorner + r * dilations[0];
         if (xR >=0 && xR < inDims[0]) {
-      `;for(let g=0;g<(d+1)/2;g++){let y=g*2;if(c+=`
-          xC = xCCorner + ${y*l};
-          `,o===1){if(y<p&&(i%2===1?(c+=`
+      `;for(let g=0;g<(p+1)/2;g++){let b=g*2;if(d+=`
+          xC = xCCorner + ${b*c};
+          `,i===1){if(b<l&&(o%2===1?(d+=`
                 xCOffset = xC + 1;
-                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xCOffset, d1);
+                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {
+                  xTexelC${b} = getX(batch, xR, xCOffset, d1);
 
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${b}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${b}Ready = 1;
                 }
-              `,l===1&&y>0?c+=`
-                xC${y} = vec4(xTexelC${y-2}.zw, xTexelC${y}.xy);
-                `:c+=`
+              `,c===1&&b>0?d+=`
+                xC${b} = vec4(xTexelC${b-2}.zw, xTexelC${b}.xy);
+                `:d+=`
                   xCOffset = xC + 1 - 2;
 
                   if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -3047,128 +3047,128 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
                       previous.zw = vec2(0.0);
                     }
 
-                    xC${y} = vec4(previous.zw, xTexelC${y}.xy);
+                    xC${b} = vec4(previous.zw, xTexelC${b}.xy);
                   } else {
-                    xC${y} = vec4(0.0, 0.0, xTexelC${y}.xy);
+                    xC${b} = vec4(0.0, 0.0, xTexelC${b}.xy);
                   }
-                  `):c+=`
-                if (xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xC, d1);
+                  `):d+=`
+                if (xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {
+                  xTexelC${b} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${b}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${b}Ready = 1;
                 }
 
-                xC${y} = xTexelC${y};
-                `,y+1<p)){let b=i%2===0?v.nearestLargerEven(l):l;l%2===0&&i%2===1||l%2!==0&&i%2!==1?(c+=`
-                  xCOffset = xC + imod(pads[1], 2) + ${b};
+                xC${b} = xTexelC${b};
+                `,b+1<l)){let y=o%2===0?w.nearestLargerEven(c):c;c%2===0&&o%2===1||c%2!==0&&o%2!==1?(d+=`
+                  xCOffset = xC + imod(pads[1], 2) + ${y};
 
-                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                    xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
+                    xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
 
                     // Need to manually clear unused channels in case
                     // we're reading from recycled texture.
                     if (xCOffset + 1 >= inDims[1]) {
-                      xTexelC${y+1}.zw = vec2(0.0);
+                      xTexelC${b+1}.zw = vec2(0.0);
                     }
-                    xTexelC${y+1}Ready = 1;
+                    xTexelC${b+1}Ready = 1;
                   }
-                  `,l>1?c+=`
+                  `,c>1?d+=`
                     xCOffset -= 2;
                     if (xCOffset >= 0 && xCOffset < inDims[1]) {
                      previous = getX(batch, xR, xCOffset, d1);
-                     xC${y+1} = vec4(previous.zw, xTexelC${y+1}.xy);
+                     xC${b+1} = vec4(previous.zw, xTexelC${b+1}.xy);
                     } else {
-                     xC${y+1} = vec4(0.0, 0.0, xTexelC${y+1}.xy);
+                     xC${b+1} = vec4(0.0, 0.0, xTexelC${b+1}.xy);
                     }
-                    `:c+=`
-                    xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.xy);
-                    `):b===1?c+=`
-                    xC${y+1} = xTexelC${y};
-                    `:c+=`
-                    xCOffset = xC + ${b};
+                    `:d+=`
+                    xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.xy);
+                    `):y===1?d+=`
+                    xC${b+1} = xTexelC${b};
+                    `:d+=`
+                    xCOffset = xC + ${y};
 
-                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                      xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
+                      xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
                       if (xCOffset + 1 >= inDims[1]) {
-                        xTexelC${y+1}.zw = vec2(0.0);
+                        xTexelC${b+1}.zw = vec2(0.0);
                       }
-                      xTexelC${y+1}Ready = 1;
+                      xTexelC${b+1}Ready = 1;
                     }
 
-                    xC${y+1} = xTexelC${y+1};
-                    `}}else y<p&&(i%2===1?(c+=`
+                    xC${b+1} = xTexelC${b+1};
+                    `}}else b<l&&(o%2===1?(d+=`
                 xCOffset = xC + 1 - strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {
+                  xTexelC${b} = getX(batch, xR, xCOffset, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${b}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${b}Ready = 1;
                 }
 
-                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${y+1}Ready == 0) {
-                  xTexelC${y+1} = getX(batch, xR, xC + 1, d1);
+                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${b+1}Ready == 0) {
+                  xTexelC${b+1} = getX(batch, xR, xC + 1, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xC + 2 >= inDims[1]) {
-                    xTexelC${y+1}.zw = vec2(0.0);
+                    xTexelC${b+1}.zw = vec2(0.0);
                   }
-                  xTexelC${y+1}Ready = 1;
+                  xTexelC${b+1}Ready = 1;
                 }
 
-                xC${y} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
-              `,y+1<p&&(c+=`
+                xC${b} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);
+              `,b+1<l&&(d+=`
                   final = vec4(0.0);
                   xCOffset = xC + 1 + strides[1];
                   if(xCOffset >= 0 && xCOffset < inDims[1]) {
                     final = getX(batch, xR, xCOffset, d1);
                   }
-                  xC${y+1} = vec4(xTexelC${y+1}.xy, final.xy);
-                `)):(c+=`
-                if(xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
-                  xTexelC${y} = getX(batch, xR, xC, d1);
+                  xC${b+1} = vec4(xTexelC${b+1}.xy, final.xy);
+                `)):(d+=`
+                if(xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {
+                  xTexelC${b} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${y}.zw = vec2(0.0);
+                    xTexelC${b}.zw = vec2(0.0);
                   }
-                  xTexelC${y}Ready = 1;
+                  xTexelC${b}Ready = 1;
                 }
 
                 xCOffset = xC + strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
-                  xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
+                  xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${y+1}.zw = vec2(0.);
+                    xTexelC${b+1}.zw = vec2(0.);
                   }
-                  xTexelC${y+1}Ready = 1;
+                  xTexelC${b+1}Ready = 1;
                 }
 
-                xC${y} = vec4(
-                  xTexelC${y}.xy, xTexelC${y+1}.xy);
-              `,y+1<p&&(c+=`
-                  xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
-                `)));y<p&&(c+=`
-            wTexel = getW(r, ${y}, d1, q);
-            dotProd += xC${y} * vec4(wTexel.xz, wTexel.xz);
-          `,y+1<p&&(c+=`
-              wTexel = getW(r, ${y+1}, d1, q);
-              dotProd += xC${y+1} * vec4(wTexel.xz, wTexel.xz);
-            `))}c+=`
+                xC${b} = vec4(
+                  xTexelC${b}.xy, xTexelC${b+1}.xy);
+              `,b+1<l&&(d+=`
+                  xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);
+                `)));b<l&&(d+=`
+            wTexel = getW(r, ${b}, d1, q);
+            dotProd += xC${b} * vec4(wTexel.xz, wTexel.xz);
+          `,b+1<l&&(d+=`
+              wTexel = getW(r, ${b+1}, d1, q);
+              dotProd += xC${b+1} * vec4(wTexel.xz, wTexel.xz);
+            `))}d+=`
     }
-  `,c+=`
+  `,d+=`
       }
-    `;let h="",m="";n&&(a?h=`vec4 activation(vec4 a) {
+    `;let h="",f="";n&&(r?h=`vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
           ${n}
-        }`:r?h=`vec4 activation(vec4 a) {
+        }`:s?h=`vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
           ${n}
         }`:h=`vec4 activation(vec4 x) {
           ${n}
-        }`,m="result = activation(result);");let f=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),a&&this.variableNames.push("preluActivationWeights"),r&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
+        }`,f="result = activation(result);");let m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
       ${h}
 
       void main() {
@@ -3176,43 +3176,43 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${s};
-        int q = d2 - d1 * ${s};
+        int d1 = d2 / ${a};
+        int q = d2 - d1 * ${a};
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
         vec4 dotProd = vec4(0.000000000000001);
 
-        ${c}
+        ${d}
 
         vec4 result = dotProd - vec4(0.000000000000001);
-        ${f}
         ${m}
+        ${f}
         setOutput(result);
       }
-    `}};function YQ(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=a,p=l;p==null&&(p=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(i,p),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);let d=N.computeConv2DInfo(r.shape,s.shape,i,p,o,u,!0),c;H().getBool("WEBGL_PACK_DEPTHWISECONV")&&d.strideWidth<=2&&d.outChannels/d.inChannels===1?c=new CE(d):c=new NE(d);let h=[[d.padInfo.top,d.padInfo.left],[d.strideHeight,d.strideWidth],[d.dilationHeight,d.dilationWidth],[d.inHeight,d.inWidth]];return n.runWebGLProgram(c,[r,s],"float32",h)}var ZQ={kernelName:Ni,backendName:"webgl",kernelFunc:YQ},JQ=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,a=e.padInfo.top,r=e.padInfo.left,s=e.outChannels/e.inChannels;this.userCode=`
+    `}};function QQ(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dilations:c,dimRoundingMode:u}=r,l=c;l==null&&(l=[1,1]),w.assert(N.eitherStridesOrDilationsAreOne(o,l),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${o} and dilations '${l}'`);let p=N.computeConv2DInfo(s.shape,a.shape,o,l,i,u,!0),d;q().getBool("WEBGL_PACK_DEPTHWISECONV")&&p.strideWidth<=2&&p.outChannels/p.inChannels===1?d=new EE(p):d=new _E(p);let h=[[p.padInfo.top,p.padInfo.left],[p.strideHeight,p.strideWidth],[p.dilationHeight,p.dilationWidth],[p.inHeight,p.inWidth]];return n.runWebGLProgram(d,[s,a],"float32",h)}var eee={kernelName:_o,backendName:"webgl",kernelFunc:QQ},tee=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,s=e.padInfo.left,a=e.outChannels/e.inChannels;this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
         int wC = coords.y;
         int d1 = coords.z;
         int dm = coords.w;
-        int d2 = d1 * ${s} + dm;
+        int d2 = d1 * ${a} + dm;
 
         float dotProd = 0.0;
 
         // TO DO: Vec4 over the batch size
         for (int b = 0; b < ${e.batchSize}; b++) {
           for (int yR = 0; yR < ${e.outHeight}; yR++) {
-            int xR = wR + yR * ${t} - ${a};
+            int xR = wR + yR * ${t} - ${r};
 
             if (xR < 0 || xR >= ${e.inHeight}) {
               continue;
             }
 
             for (int yC = 0; yC < ${e.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${r};
+              int xC = wC + yC * ${n} - ${s};
 
               if (xC < 0 || xC >= ${e.inWidth}) {
                 continue;
@@ -3226,8 +3226,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},QQ=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,a=e.strideHeight,r=e.strideWidth,s=t-1-e.padInfo.top,i=n-1-e.padInfo.left,o=e.outChannels/e.inChannels;this.userCode=`
-      const ivec2 pads = ivec2(${s}, ${i});
+    `}},nee=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,s=e.strideWidth,a=t-1-e.padInfo.top,o=n-1-e.padInfo.left,i=e.outChannels/e.inChannels;this.userCode=`
+      const ivec2 pads = ivec2(${a}, ${o});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3240,7 +3240,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         float dotProd = 0.0;
 
         for (int wR = 0; wR < ${t}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${a}.0;
+          float dyR = float(dyRCorner + wR) / ${r}.0;
 
           if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
@@ -3250,7 +3250,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           int wRPerm = ${t} - 1 - wR;
 
           for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${r}.0;
+            float dyC = float(dyCCorner + wC) / ${s}.0;
 
             if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                 fract(dyC) > 0.0) {
@@ -3261,8 +3261,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             int wCPerm = ${n} - 1 - wC;
 
             // TO DO: Vec4 over the channelMul
-            for (int dm = 0; dm < ${o}; dm++) {
-              int d2 = d1 * ${o} + dm;
+            for (int dm = 0; dm < ${i}; dm++) {
+              int d2 = d1 * ${i} + dm;
               float xValue = getDy(batch, idyR, idyC, d2);
               float wValue = getW(wRPerm, wCPerm, d1, dm);
               dotProd += xValue * wValue;
@@ -3271,15 +3271,15 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function eee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:p}=a,d=N.computeConv2DInfo(r.shape,p,i,o,l,u,!0),c=new JQ(d);return n.runWebGLProgram(c,[r,s],"float32")}var tee={kernelName:xm,backendName:"webgl",kernelFunc:eee};function nee(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:p}=a,d=N.computeConv2DInfo(p,s.shape,i,o,l,u,!0),c=new QQ(d);return n.runWebGLProgram(c,[r,s],"float32")}var aee={kernelName:vm,backendName:"webgl",kernelFunc:nee},ree=class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode=`
+    `}};function ree(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,dy:a}=t,{strides:o,dilations:i,pad:c,dimRoundingMode:u,filterShape:l}=r,p=N.computeConv2DInfo(s.shape,l,o,i,c,u,!0),d=new tee(p);return n.runWebGLProgram(d,[s,a],"float32")}var see={kernelName:wf,backendName:"webgl",kernelFunc:ree};function aee(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,filter:a}=t,{strides:o,dilations:i,pad:c,dimRoundingMode:u,inputShape:l}=r,p=N.computeConv2DInfo(l,a.shape,o,i,c,u,!0),d=new nee(p);return n.runWebGLProgram(d,[s,a],"float32")}var oee={kernelName:If,backendName:"webgl",kernelFunc:aee},iee=class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode=`
       void main() {
           ivec2 coords = getOutputCoords();
           float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;
           setOutput(val);
       }
-    `}};function see(e){let{inputs:t,backend:n}=e,{x:a}=t,r=[...a.shape,...a.shape],s=v.sizeFromShape(a.shape),i=de({inputs:{x:a},backend:n,attrs:{shape:[s]}}),o=new ree(s),l=n.runWebGLProgram(o,[i],i.dtype),u=de({inputs:{x:l},backend:n,attrs:{shape:r}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(l),u}var iee={kernelName:wm,backendName:"webgl",kernelFunc:see},oee=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let{inHeight:t,inWidth:n,padInfo:a,strideHeight:r,strideWidth:s,filterHeight:i,filterWidth:o,dilationHeight:l,dilationWidth:u}=e,{top:p,left:d}=a;this.userCode=`
-      const ivec2 strides = ivec2(${r}, ${s});
-      const ivec2 pads = ivec2(${p}, ${d});
+    `}};function cee(e){let{inputs:t,backend:n}=e,{x:r}=t,s=[...r.shape,...r.shape],a=w.sizeFromShape(r.shape),o=he({inputs:{x:r},backend:n,attrs:{shape:[a]}}),i=new iee(a),c=n.runWebGLProgram(i,[o],o.dtype),u=he({inputs:{x:c},backend:n,attrs:{shape:s}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(c),u}var uee={kernelName:kf,backendName:"webgl",kernelFunc:cee},lee=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let{inHeight:t,inWidth:n,padInfo:r,strideHeight:s,strideWidth:a,filterHeight:o,filterWidth:i,dilationHeight:c,dilationWidth:u}=e,{top:l,left:p}=r;this.userCode=`
+      const ivec2 strides = ivec2(${s}, ${a});
+      const ivec2 pads = ivec2(${l}, ${p});
       const float neg_infinity = -3.4e38;
 
       void main() {
@@ -3292,11 +3292,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         int wBeg = outTopLeftCorner.y;
 
         float curVal = neg_infinity;
-        for (int h = 0; h < ${i}; h++) {
-          int hIn = hBeg + h * ${l};
+        for (int h = 0; h < ${o}; h++) {
+          int hIn = hBeg + h * ${c};
 
           if (hIn >= 0 && hIn < ${t}) {
-            for (int w = 0; w < ${o}; w++) {
+            for (int w = 0; w < ${i}; w++) {
               int wIn = wBeg + w * ${u};
 
               if (wIn >= 0 && wIn < ${n}) {
@@ -3315,7 +3315,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         float result = curVal;
         setOutput(result);
       }
-    `}};function lee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s}=t,{strides:i,pad:o,dilations:l}=a,u=N.computeDilation2DInfo(r.shape,s.shape,i,o,"NHWC",l),p,d=new oee(u);p=n.runWebGLProgram(d,[r,s],"float32");let c=de({inputs:{x:p},backend:n,attrs:{shape:u.outShape}});return n.disposeIntermediateTensorInfo(p),c}var uee={kernelName:fc,backendName:"webgl",kernelFunc:lee};function pee(e){let{inputs:t,backend:n,attrs:a}=e,{equation:r}=a,s=t,{allDims:i,summedDims:o,idDims:l}=N.decodeEinsumEquation(r,s.length);N.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:p}=N.getEinsumComputePath(o,l),d=p.length,c=null,h=i.length,m=[];for(let f=0;f<d;++f){for(let g of p[f]){let{permutationIndices:y,expandDims:b}=N.getEinsumPermutation(h,l[g]),x;N.isIdentityPermutation(y)?x=s[g]:(x=Sn({inputs:{x:s[g]},backend:n,attrs:{perm:y}}),m.push(x));let w=x.shape.slice();for(let I=0;I<b.length;++I)w.splice(b[I],0,1);v.arraysEqual(x.shape,w)||(x=de({inputs:{x},backend:n,attrs:{shape:w}}),m.push(x)),c===null?c=x:(c=o1({inputs:{a:x,b:c},backend:n}),m.push(c))}f<d-1&&(u[f]>=0&&(c=Zf({inputs:{x:c},backend:n,attrs:{axis:u[f]-(i.length-h),keepDims:!1}}),m.push(c)),h--)}for(let f of m)f!==c&&n.disposeIntermediateTensorInfo(f);return c}var cee={kernelName:km,backendName:"webgl",kernelFunc:pee},dee="return (x >= 0.0) ? x : (exp(x) - 1.0);",hee=`
+    `}};function dee(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a}=t,{strides:o,pad:i,dilations:c}=r,u=N.computeDilation2DInfo(s.shape,a.shape,o,i,"NHWC",c),l,p=new lee(u);l=n.runWebGLProgram(p,[s,a],"float32");let d=he({inputs:{x:l},backend:n,attrs:{shape:u.outShape}});return n.disposeIntermediateTensorInfo(l),d}var pee={kernelName:md,backendName:"webgl",kernelFunc:dee};function hee(e){let{inputs:t,backend:n,attrs:r}=e,{equation:s}=r,a=t,{allDims:o,summedDims:i,idDims:c}=N.decodeEinsumEquation(s,a.length);N.checkEinsumDimSizes(o.length,c,a);let{path:u,steps:l}=N.getEinsumComputePath(i,c),p=l.length,d=null,h=o.length,f=[];for(let m=0;m<p;++m){for(let g of l[m]){let{permutationIndices:b,expandDims:y}=N.getEinsumPermutation(h,c[g]),v;N.isIdentityPermutation(b)?v=a[g]:(v=Nn({inputs:{x:a[g]},backend:n,attrs:{perm:b}}),f.push(v));let x=v.shape.slice();for(let k=0;k<y.length;++k)x.splice(y[k],0,1);w.arraysEqual(v.shape,x)||(v=he({inputs:{x:v},backend:n,attrs:{shape:x}}),f.push(v)),d===null?d=v:(d=iI({inputs:{a:v,b:d},backend:n}),f.push(d))}m<p-1&&(u[m]>=0&&(d=Jm({inputs:{x:d},backend:n,attrs:{axis:u[m]-(o.length-h),keepDims:!1}}),f.push(d)),h--)}for(let m of f)m!==d&&n.disposeIntermediateTensorInfo(m);return d}var fee={kernelName:Sf,backendName:"webgl",kernelFunc:hee},mee="return (x >= 0.0) ? x : (exp(x) - 1.0);",gee=`
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -3324,12 +3324,12 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,mee=Ye({opSnippet:dee,packedOpSnippet:hee}),fee={kernelName:_i,backendName:"webgl",kernelFunc:mee},gee="return (b >= 1.0) ? a : a * (b + 1.0);",yee=`
+`,bee=Ye({opSnippet:mee,packedOpSnippet:gee}),yee={kernelName:Ao,backendName:"webgl",kernelFunc:bee},vee="return (b >= 1.0) ? a : a * (b + 1.0);",xee=`
   vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));
   return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));
-`,bee=e=>{let{inputs:t,backend:n}=e,{dy:a,y:r}=t,s=H().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new ud(yee,a.shape,r.shape):new $l(gee,a.shape,r.shape);return n.runWebGLProgram(s,[a,r],a.dtype)},xee={kernelName:Im,backendName:"webgl",kernelFunc:bee},vee=`
+`,wee=e=>{let{inputs:t,backend:n}=e,{dy:r,y:s}=t,a=q().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new up(xee,r.shape,s.shape):new $c(vee,r.shape,s.shape);return n.runWebGLProgram(a,[r,s],r.dtype)},Iee={kernelName:Tf,backendName:"webgl",kernelFunc:wee},kee=`
   return vec4(equal(a, b));
-`,wee="return float(a == b);",kee=cn({opSnippet:wee,packedOpSnippet:vee,dtype:"bool",cpuKernelImpl:s7}),Iee={kernelName:Xl,backendName:"webgl",kernelFunc:kee},See=`
+`,See="return float(a == b);",Tee=dn({opSnippet:See,packedOpSnippet:kee,dtype:"bool",cpuKernelImpl:c9}),Cee={kernelName:Xc,backendName:"webgl",kernelFunc:Tee},Nee=`
   // Error function is calculated approximately with elementary function.
   // See "Handbook of Mathematical Functions with Formulas,
   // Graphs, and Mathematical Tables", Abramowitz and Stegun.
@@ -3344,9 +3344,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
   x = abs(x);
   float t = 1.0 / (1.0 + p * x);
   return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));
-`,Tee=Ye({opSnippet:See}),Nee={kernelName:Kl,backendName:"webgl",kernelFunc:Tee},Cee=tp+`
+`,_ee=Ye({opSnippet:Nee}),Eee={kernelName:Kc,backendName:"webgl",kernelFunc:_ee},Aee=tl+`
   return exp(x);
-`,_ee=`
+`,$ee=`
   vec4 result = exp(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : result.r;
@@ -3355,21 +3355,21 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,_E=Ye({opSnippet:Cee,packedOpSnippet:_ee,cpuKernelImpl:i7,dtype:"float32"}),Eee={kernelName:Ei,backendName:"webgl",kernelFunc:_E};function vx(e){let{inputs:t,attrs:n,backend:a}=e,{dim:r}=n,{input:s}=t,i=s.shape.length,o=s.shape.slice(),l=r;return r<0&&(v.assert(-(i+1)<=r,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+r+1),o.splice(l,0,1),de({inputs:{x:s},backend:a,attrs:{shape:o}})}var Aee={kernelName:Yl,backendName:"webgl",kernelFunc:vx},EI="return exp(x) - 1.0;",$ee=Ye({opSnippet:EI,packedOpSnippet:EI,cpuKernelImpl:o7}),Fee={kernelName:Zl,backendName:"webgl",kernelFunc:$ee},AI=class{constructor(e,t,n){this.variableNames=["real","imag"];let a=t[1];this.outputShape=t;let r=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,s=n?`${a}.0`:"1.0",i;if(e==="real")i="return real * expR - imag * expI;";else if(e==="imag")i="return real * expI + imag * expR;";else throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);this.userCode=`
-      const float exponentMultiplier = ${r};
+`,AE=Ye({opSnippet:Aee,packedOpSnippet:$ee,cpuKernelImpl:u9,dtype:"float32"}),Dee={kernelName:$o,backendName:"webgl",kernelFunc:AE};function xv(e){let{inputs:t,attrs:n,backend:r}=e,{dim:s}=n,{input:a}=t,o=a.shape.length,i=a.shape.slice(),c=s;return s<0&&(w.assert(-(o+1)<=s,()=>`Axis must be in the interval [${-(o+1)}, ${o}]`),c=o+s+1),i.splice(c,0,1),he({inputs:{x:a},backend:r,attrs:{shape:i}})}var Fee={kernelName:Yc,backendName:"webgl",kernelFunc:xv},A1="return exp(x) - 1.0;",Ree=Ye({opSnippet:A1,packedOpSnippet:A1,cpuKernelImpl:l9}),Pee={kernelName:Zc,backendName:"webgl",kernelFunc:Ree},$1=class{constructor(e,t,n){this.variableNames=["real","imag"];let r=t[1];this.outputShape=t;let s=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,a=n?`${r}.0`:"1.0",o;if(e==="real")o="return real * expR - imag * expI;";else if(e==="imag")o="return real * expI + imag * expR;";else throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);this.userCode=`
+      const float exponentMultiplier = ${s};
 
       float unaryOpComplex(float real, float expR, float imag, float expI) {
-        ${i}
+        ${o}
       }
 
       float mulMatDFT(int batch, int index) {
-        float indexRatio = float(index) / float(${a});
+        float indexRatio = float(index) / float(${r});
         float exponentMultiplierTimesIndexRatio =
             exponentMultiplier * indexRatio;
 
         float result = 0.0;
 
-        for (int i = 0; i < ${a}; i++) {
+        for (int i = 0; i < ${r}; i++) {
           // x = (-2|2 * PI / N) * index * i;
           float x = exponentMultiplierTimesIndexRatio * float(i);
           float expR = cos(x);
@@ -3378,7 +3378,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           float imag = getImag(batch, i);
 
           result +=
-              unaryOpComplex(real, expR, imag, expI) / ${s};
+              unaryOpComplex(real, expR, imag, expI) / ${a};
         }
 
         return result;
@@ -3388,12 +3388,12 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         ivec2 coords = getOutputCoords();
         setOutput(mulMatDFT(coords[0], coords[1]));
       }
-    `}};function EE(e,t,n){let a=n.texData.get(e.dataId),r=v.sizeFromShape(e.shape),s=e.shape[e.shape.length-1],i=r/s,o=de({inputs:{x:e},backend:n,attrs:{shape:[i,s]}}),l=o.shape,u=new AI("real",l,t),p=new AI("imag",l,t),d=[{dataId:a.complexTensorInfos.real.dataId,dtype:a.complexTensorInfos.real.dtype,shape:l},{dataId:a.complexTensorInfos.imag.dataId,dtype:a.complexTensorInfos.imag.dtype,shape:l}],c=n.runWebGLProgram(u,d,"float32"),h=n.runWebGLProgram(p,d,"float32"),m=_s({inputs:{real:c,imag:h},backend:n});n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h);let f=de({inputs:{x:m},backend:n,attrs:{shape:e.shape}});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(m),f}function Dee(e){let{inputs:t,backend:n}=e,{input:a}=t;return EE(a,!1,n)}var Ree={kernelName:Sm,backendName:"webgl",kernelFunc:Dee},Mee=class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode=`
+    `}};function $E(e,t,n){let r=n.texData.get(e.dataId),s=w.sizeFromShape(e.shape),a=e.shape[e.shape.length-1],o=s/a,i=he({inputs:{x:e},backend:n,attrs:{shape:[o,a]}}),c=i.shape,u=new $1("real",c,t),l=new $1("imag",c,t),p=[{dataId:r.complexTensorInfos.real.dataId,dtype:r.complexTensorInfos.real.dtype,shape:c},{dataId:r.complexTensorInfos.imag.dataId,dtype:r.complexTensorInfos.imag.dtype,shape:c}],d=n.runWebGLProgram(u,p,"float32"),h=n.runWebGLProgram(l,p,"float32"),f=Ea({inputs:{real:d,imag:h},backend:n});n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h);let m=he({inputs:{x:f},backend:n,attrs:{shape:e.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(f),m}function Oee(e){let{inputs:t,backend:n}=e,{input:r}=t;return $E(r,!1,n)}var Mee={kernelName:Cf,backendName:"webgl",kernelFunc:Oee},Lee=class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode=`
       void main() {
         // Input can be obtained from uniform value.
         setOutput(value);
       }
-    `}};function cd(e){let{backend:t,attrs:n}=e,{shape:a,value:r}=n,{dtype:s}=n;if(s=s||v.inferDtype(r),s==="string"){let i=v.getArrayFromDType(s,v.sizeFromShape(a));return i.fill(r),t.makeTensorInfo(a,s,i)}else{let i=new Mee(a,r),o=[[r]];return t.runWebGLProgram(i,[],s,o)}}var Pee={kernelName:gc,backendName:"webgl",kernelFunc:cd},Oee=class{constructor(e){this.variableNames=["Image"],this.outputShape=[];let t=e[2];this.outputShape=e,this.userCode=`
+    `}};function dp(e){let{backend:t,attrs:n}=e,{shape:r,value:s}=n,{dtype:a}=n;if(a=a||w.inferDtype(s),a==="string"){let o=w.getArrayFromDType(a,w.sizeFromShape(r));return o.fill(s),t.makeTensorInfo(r,a,o)}else{let o=new Lee(r,s),i=[[s]];return t.runWebGLProgram(o,[],a,i)}}var zee={kernelName:gd,backendName:"webgl",kernelFunc:dp},Bee=class{constructor(e){this.variableNames=["Image"],this.outputShape=[];let t=e[2];this.outputShape=e,this.userCode=`
         void main() {
           ivec4 coords = getOutputCoords();
           int x = coords[2];
@@ -3407,7 +3407,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           }
           setOutput(outputValue);
         }
-    `}},Lee={kernelName:Jl,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{let{image:n}=e,a=t,r=new Oee(n.shape);return a.runWebGLProgram(r,[n],n.dtype)}},$I="return floor(x);",zee=Ye({opSnippet:$I,packedOpSnippet:$I,cpuKernelImpl:l7}),Wee={kernelName:Ai,backendName:"webgl",kernelFunc:zee},Bee=`
+    `}},Wee={kernelName:Jc,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{let{image:n}=e,r=t,s=new Bee(n.shape);return r.runWebGLProgram(s,[n],n.dtype)}},D1="return floor(x);",Vee=Ye({opSnippet:D1,packedOpSnippet:D1,cpuKernelImpl:d9}),Uee={kernelName:Do,backendName:"webgl",kernelFunc:Vee},Gee=`
   float s = sign(a) * sign(b);
   int ia = round(a);
   int ib = round(b);
@@ -3417,7 +3417,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
   } else {
     return NAN;
   }
-`,Vee=`
+`,Hee=`
   ivec4 ia = round(a);
   ivec4 ib = round(b);
   bvec4 cond = notEqual(ib, ivec4(0));
@@ -3438,13 +3438,13 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
     result[3] = idiv(ia[3], ib[3], s[3]);
   }
   return vec4(result);
-`,Uee=cn({opSnippet:Bee,packedOpSnippet:Vee,dtype:"int32"}),Gee={kernelName:$i,backendName:"webgl",kernelFunc:Uee},Hee=class{constructor(e){this.variableNames=["A"];let t=_n(),[n,a]=e;this.outputShape=e,this.userCode=`
+`,qee=dn({opSnippet:Gee,packedOpSnippet:Hee,dtype:"int32"}),jee={kernelName:Fo,backendName:"webgl",kernelFunc:qee},Kee=class{constructor(e){this.variableNames=["A"];let t=$n(),[n,r]=e;this.outputShape=e,this.userCode=`
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
         int texC = coords[1];
         int depth = coords[2];
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${a}.0, ${n}.0);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}.0, ${n}.0);
 
         vec4 values = ${t.texture2D}(A, uv);
         float value;
@@ -3460,7 +3460,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         setOutput(floor(value * 255.0 + 0.5));
       }
-    `}},jee=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let t=_n(),[n,a]=e;this.outputShape=e,this.userCode=`
+    `}},Xee=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let t=$n(),[n,r]=e;this.outputShape=e,this.userCode=`
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
@@ -3475,7 +3475,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             depth = coords[2] + col;
 
             vec2 uv = (vec2(texC, texR) + halfCR) /
-                       vec2(${a}.0, ${n}.0);
+                       vec2(${r}.0, ${n}.0);
             vec4 values = ${t.texture2D}(A, uv);
             float value;
             if (depth == 0) {
@@ -3494,39 +3494,39 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
 
         ${t.output} = result;
       }
-    `}},qee={kernelName:Fh,backendName:"webgl",kernelFunc:Kee},sl,Tb=H().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");function Kee(e){let{inputs:t,backend:n,attrs:a}=e,{pixels:r}=t,{numChannels:s}=a,i=typeof HTMLVideoElement!="undefined"&&r instanceof HTMLVideoElement,o=typeof HTMLImageElement!="undefined"&&r instanceof HTMLImageElement,[l,u]=i?[r.videoWidth,r.videoHeight]:[r.width,r.height],p=[u,l],d=[u,l,s];if(o||i){let f=H().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");(sl==null||f!==Tb)&&(Tb=f,sl=document.createElement("canvas").getContext("2d",{willReadFrequently:Tb})),sl.canvas.width=l,sl.canvas.height=u,sl.drawImage(r,0,0,l,u),r=sl.canvas}let c=n.makeTensorInfo(p,"int32");n.texData.get(c.dataId).usage=ca.PIXELS,n.gpgpu.uploadPixelDataToTexture(n.getTexture(c.dataId),r);let h=H().getBool("WEBGL_PACK")?new jee(d):new Hee(d),m=n.runWebGLProgram(h,[c],"int32");return n.disposeData(c.dataId),m}function Xee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:p,dilations:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=a,f=N.convertConv2DDataFormat(p),g=N.computeConv2DInfo(r.shape,s.shape,l,d,u,c,!1,f),y,b=[],x=i!=null,w=o!=null,I=h==="leakyrelu",T=()=>{let E=[r,s],A=(R,F)=>{if(F==="NCHW"&&R.shape.length===1&&R.shape[0]!==1){let S=de({inputs:{x:R},backend:n,attrs:{shape:[R.shape[0],1,1]}});return b.push(S),S}return R};if(x&&E.push(A(i,p)),w&&E.push(A(o,p)),I){let R=n.makeTensorInfo([],"float32",v.createScalarValue(m,"float32"));E.push(R),b.push(R)}return E};if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type==="SAME"||g.padInfo.type==="VALID"))y=IE({x:r,filter:s,convInfo:g,backend:n,bias:i,activation:h,preluActivationWeights:o,leakyreluAlpha:m});else if(g.strideWidth<=2&&f==="channelsLast"&&H().getBool("WEBGL_EXP_CONV")){let E=h?ic(h,!0):null,A=new kE(g,x,E,w,I),R=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],F=T();y=n.runWebGLProgram(A,F,"float32",R)}else if(H().getBool("WEBGL_CONV_IM2COL"))y=SE({x:r,filter:s,convInfo:g,backend:n,bias:i,activation:h,preluActivationWeights:o,leakyreluAlpha:m});else{let E=h?ic(h,!1):null,A=new wE(g,x,E,w,I),R=T();y=n.runWebGLProgram(A,R,"float32")}let C=de({inputs:{x:y},backend:n,attrs:{shape:g.outShape}});return b.push(y),b.forEach(E=>n.disposeIntermediateTensorInfo(E)),C}var Yee={kernelName:ei,backendName:"webgl",kernelFunc:Xee};function Zee(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dimRoundingMode:d,activation:c,leakyreluAlpha:h}=a,m=[],f=p;f==null&&(f=[1,1]),v.assert(N.eitherStridesOrDilationsAreOne(l,f),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${l} and dilations '${f}'`);let g=N.computeConv2DInfo(r.shape,s.shape,l,f,u,d,!0),y=H().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,b=c?ic(c,y):null,x=[r,s],w=i!=null,I=o!=null,T=c==="leakyrelu";if(w&&x.push(i),I&&x.push(o),T){let R=n.makeTensorInfo([],"float32",v.createScalarValue(h,"float32"));x.push(R),m.push(R)}let C;y?C=new CE(g,w,b,I,T):C=new NE(g,w,b,I,T);let E=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],A=n.runWebGLProgram(C,x,"float32",E);return m.forEach(R=>n.disposeIntermediateTensorInfo(R)),A}var Jee={kernelName:ti,backendName:"webgl",kernelFunc:Zee},Qee=class{constructor(e,t,n,a){this.sliceDim=e,this.strides=t,this.paramsShape=a,this.variableNames=["x","indices"],this.outputShape=n;let r=gt(n.length),s=`
-    int index;`;for(let i=0;i<this.sliceDim;i++)s+=`
-          index = round(getIndices(coords[0], ${i}));
+    `}},Yee={kernelName:Fh,backendName:"webgl",kernelFunc:Zee},ac,Ty=q().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");function Zee(e){let{inputs:t,backend:n,attrs:r}=e,{pixels:s}=t,{numChannels:a}=r,o=typeof HTMLVideoElement!="undefined"&&s instanceof HTMLVideoElement,i=typeof HTMLImageElement!="undefined"&&s instanceof HTMLImageElement,[c,u]=o?[s.videoWidth,s.videoHeight]:[s.width,s.height],l=[u,c],p=[u,c,a];if(i||o){let m=q().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");(ac==null||m!==Ty)&&(Ty=m,ac=document.createElement("canvas").getContext("2d",{willReadFrequently:Ty})),ac.canvas.width=c,ac.canvas.height=u,ac.drawImage(s,0,0,c,u),s=ac.canvas}let d=n.makeTensorInfo(l,"int32");n.texData.get(d.dataId).usage=ur.PIXELS,n.gpgpu.uploadPixelDataToTexture(n.getTexture(d.dataId),s);let h=q().getBool("WEBGL_PACK")?new Xee(p):new Kee(p),f=n.runWebGLProgram(h,[d],"int32");return n.disposeData(d.dataId),f}function Jee(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dataFormat:l,dilations:p,dimRoundingMode:d,activation:h,leakyreluAlpha:f}=r,m=N.convertConv2DDataFormat(l),g=N.computeConv2DInfo(s.shape,a.shape,c,p,u,d,!1,m),b,y=[],v=o!=null,x=i!=null,k=h==="leakyrelu",S=()=>{let E=[s,a],$=(F,A)=>{if(A==="NCHW"&&F.shape.length===1&&F.shape[0]!==1){let R=he({inputs:{x:F},backend:n,attrs:{shape:[F.shape[0],1,1]}});return y.push(R),R}return F};if(v&&E.push($(o,l)),x&&E.push($(i,l)),k){let F=n.makeTensorInfo([],"float32",w.createScalarValue(f,"float32"));E.push(F),y.push(F)}return E};if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type==="SAME"||g.padInfo.type==="VALID"))b=TE({x:s,filter:a,convInfo:g,backend:n,bias:o,activation:h,preluActivationWeights:i,leakyreluAlpha:f});else if(g.strideWidth<=2&&m==="channelsLast"&&q().getBool("WEBGL_EXP_CONV")){let E=h?od(h,!0):null,$=new SE(g,v,E,x,k),F=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],A=S();b=n.runWebGLProgram($,A,"float32",F)}else if(q().getBool("WEBGL_CONV_IM2COL"))b=CE({x:s,filter:a,convInfo:g,backend:n,bias:o,activation:h,preluActivationWeights:i,leakyreluAlpha:f});else{let E=h?od(h,!1):null,$=new kE(g,v,E,x,k),F=S();b=n.runWebGLProgram($,F,"float32")}let C=he({inputs:{x:b},backend:n,attrs:{shape:g.outShape}});return y.push(b),y.forEach(E=>n.disposeIntermediateTensorInfo(E)),C}var Qee={kernelName:to,backendName:"webgl",kernelFunc:Jee};function ete(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dilations:l,dimRoundingMode:p,activation:d,leakyreluAlpha:h}=r,f=[],m=l;m==null&&(m=[1,1]),w.assert(N.eitherStridesOrDilationsAreOne(c,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${c} and dilations '${m}'`);let g=N.computeConv2DInfo(s.shape,a.shape,c,m,u,p,!0),b=q().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,y=d?od(d,b):null,v=[s,a],x=o!=null,k=i!=null,S=d==="leakyrelu";if(x&&v.push(o),k&&v.push(i),S){let F=n.makeTensorInfo([],"float32",w.createScalarValue(h,"float32"));v.push(F),f.push(F)}let C;b?C=new EE(g,x,y,k,S):C=new _E(g,x,y,k,S);let E=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],$=n.runWebGLProgram(C,v,"float32",E);return f.forEach(F=>n.disposeIntermediateTensorInfo(F)),$}var tte={kernelName:no,backendName:"webgl",kernelFunc:ete},nte=class{constructor(e,t,n,r){this.sliceDim=e,this.strides=t,this.paramsShape=r,this.variableNames=["x","indices"],this.outputShape=n;let s=mt(n.length),a=`
+    int index;`;for(let o=0;o<this.sliceDim;o++)a+=`
+          index = round(getIndices(coords[0], ${o}));
           out_of_bounds = out_of_bounds || index < 0;
-          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[i]};
-          flattenIndex += index * ${this.strides[i]};`;this.userCode=`
+          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[o]};
+          flattenIndex += index * ${this.strides[o]};`;this.userCode=`
          void main() {
-          ${r} coords = getOutputCoords();
+          ${s} coords = getOutputCoords();
           int flattenIndex = 0;
           bool out_of_bounds = false;
 
-          ${s}
+          ${a}
 
           setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));
         }
-      `}};function ete(e){let{inputs:t,backend:n}=e,{params:a,indices:r}=t,s=r.shape,i=s[s.length-1],o=v.sizeFromShape(a.shape),[l,u,p,d]=N.prepareAndValidate(a,r),c=de({inputs:{x:r},backend:n,attrs:{shape:[u,i]}}),h=de({inputs:{x:a},backend:n,attrs:{shape:[v.sizeFromShape(a.shape)/p,p]}});if(n.shouldExecuteOnCPU([a,r])||a.dtype==="string"){let y=n.readSync(r.dataId),b=n.bufferSync(a),x=u7(y,b,a.dtype,u,i,p,d,a.shape,o);return n.makeTensorInfo(l,a.dtype,x.values)}let m=new Qee(i,d,[u,p],a.shape),f=n.runWebGLProgram(m,[h,c],h.dtype),g=de({inputs:{x:f},backend:n,attrs:{shape:l}});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),g}var tte={kernelName:eu,backendName:"webgl",kernelFunc:ete},nte=class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;let n=gt(this.rank),a=ate(e,2);this.userCode=`
+      `}};function rte(e){let{inputs:t,backend:n}=e,{params:r,indices:s}=t,a=s.shape,o=a[a.length-1],i=w.sizeFromShape(r.shape),[c,u,l,p]=N.prepareAndValidate(r,s),d=he({inputs:{x:s},backend:n,attrs:{shape:[u,o]}}),h=he({inputs:{x:r},backend:n,attrs:{shape:[w.sizeFromShape(r.shape)/l,l]}});if(n.shouldExecuteOnCPU([r,s])||r.dtype==="string"){let b=n.readSync(s.dataId),y=n.bufferSync(r),v=p9(b,y,r.dtype,u,o,l,p,r.shape,i);return n.makeTensorInfo(c,r.dtype,v.values)}let f=new nte(o,p,[u,l],r.shape),m=n.runWebGLProgram(f,[h,d],h.dtype),g=he({inputs:{x:m},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),g}var ste={kernelName:eu,backendName:"webgl",kernelFunc:rte},ate=class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;let n=mt(this.rank),r=ote(e,2);this.userCode=`
       void main() {
         ${n} resRC = getOutputCoords();
         int index = int(getIndices(resRC.x, resRC.z));
         float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;
-        setOutput(inBounds * getA(${a}));
+        setOutput(inBounds * getA(${r}));
       }
-    `}};function ate(e,t){let n=["resRC.x","resRC.y","resRC.z","resRC.w"],a=[];for(let r=0;r<e.length;r++)r===2?a.push("index"):a.push(`${n[r]}`);return a.join()}function AE(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,indices:s}=t,{axis:i,batchDims:o}=a,l=v.parseAxisParam(i,r.shape)[0];if(H().get("DEBUG")){let b=n.readSync(s.dataId),x=r.shape[l];for(let w=0;w<b.length;++w){let I=b[w];v.assert(I<=x-1&&I>=0,()=>`GatherV2: the index value ${I} is not in [0, ${x-1}]`)}}let u=N.segment_util.collectGatherOpShapeInfo(r,s,l,o),p=v.sizeFromShape(s.shape),d=[],c=de({inputs:{x:r},backend:n,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),h=de({inputs:{x:s},backend:n,attrs:{shape:[u.batchSize,p/u.batchSize]}});d.push(c),d.push(h);let m=[u.batchSize,u.outerSize,p/u.batchSize,u.sliceSize];if(n.shouldExecuteOnCPU([r,s])||r.dtype==="string"){let b=n.bufferSync(h),x=n.bufferSync(c),w=p7(x,b,m);return d.forEach(I=>n.disposeIntermediateTensorInfo(I)),n.makeTensorInfo(u.outputShape,w.dtype,w.values)}let f=new nte(c.shape,m),g=n.runWebGLProgram(f,[c,h],c.dtype);d.push(g);let y=de({inputs:{x:g},backend:n,attrs:{shape:u.outputShape}});return d.forEach(b=>n.disposeIntermediateTensorInfo(b)),y}var rte={kernelName:Ql,backendName:"webgl",kernelFunc:AE},ste="return float(a > b);",ite=`
+    `}};function ote(e,t){let n=["resRC.x","resRC.y","resRC.z","resRC.w"],r=[];for(let s=0;s<e.length;s++)s===2?r.push("index"):r.push(`${n[s]}`);return r.join()}function DE(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,indices:a}=t,{axis:o,batchDims:i}=r,c=w.parseAxisParam(o,s.shape)[0];if(q().get("DEBUG")){let y=n.readSync(a.dataId),v=s.shape[c];for(let x=0;x<y.length;++x){let k=y[x];w.assert(k<=v-1&&k>=0,()=>`GatherV2: the index value ${k} is not in [0, ${v-1}]`)}}let u=N.segment_util.collectGatherOpShapeInfo(s,a,c,i),l=w.sizeFromShape(a.shape),p=[],d=he({inputs:{x:s},backend:n,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),h=he({inputs:{x:a},backend:n,attrs:{shape:[u.batchSize,l/u.batchSize]}});p.push(d),p.push(h);let f=[u.batchSize,u.outerSize,l/u.batchSize,u.sliceSize];if(n.shouldExecuteOnCPU([s,a])||s.dtype==="string"){let y=n.bufferSync(h),v=n.bufferSync(d),x=h9(v,y,f);return p.forEach(k=>n.disposeIntermediateTensorInfo(k)),n.makeTensorInfo(u.outputShape,x.dtype,x.values)}let m=new ate(d.shape,f),g=n.runWebGLProgram(m,[d,h],d.dtype);p.push(g);let b=he({inputs:{x:g},backend:n,attrs:{shape:u.outputShape}});return p.forEach(y=>n.disposeIntermediateTensorInfo(y)),b}var ite={kernelName:Qc,backendName:"webgl",kernelFunc:DE},cte="return float(a > b);",ute=`
   return vec4(greaterThan(a, b));
-`,ote=cn({opSnippet:ste,packedOpSnippet:ite,cpuKernelImpl:c7,dtype:"bool"}),lte={kernelName:tu,backendName:"webgl",kernelFunc:ote},ute="return float(a >= b);",pte=`
+`,lte=dn({opSnippet:cte,packedOpSnippet:ute,cpuKernelImpl:f9,dtype:"bool"}),dte={kernelName:tu,backendName:"webgl",kernelFunc:lte},pte="return float(a >= b);",hte=`
   return vec4(greaterThanEqual(a, b));
-`,cte=cn({opSnippet:ute,packedOpSnippet:pte,dtype:"bool",cpuKernelImpl:d7}),dte={kernelName:Di,backendName:"webgl",kernelFunc:cte};function hte(e){let{inputs:t,backend:n}=e,{input:a}=t;return EE(a,!0,n)}var mte={kernelName:Tm,backendName:"webgl",kernelFunc:hte},fte="return float(!isnan(x) && !isinf(x));",gte=Ye({opSnippet:fte,dtype:"bool"}),yte={kernelName:nu,backendName:"webgl",kernelFunc:gte},bte="return float(isinf(x));",xte=Ye({opSnippet:bte,dtype:"bool"}),vte={kernelName:au,backendName:"webgl",kernelFunc:xte},wte="return float(isnan(x));",kte=Ye({opSnippet:wte,dtype:"bool"}),Ite={kernelName:ru,backendName:"webgl",kernelFunc:kte},Ste="return float(a < b);",Tte=`
+`,fte=dn({opSnippet:pte,packedOpSnippet:hte,dtype:"bool",cpuKernelImpl:m9}),mte={kernelName:Po,backendName:"webgl",kernelFunc:fte};function gte(e){let{inputs:t,backend:n}=e,{input:r}=t;return $E(r,!0,n)}var bte={kernelName:Nf,backendName:"webgl",kernelFunc:gte},yte="return float(!isnan(x) && !isinf(x));",vte=Ye({opSnippet:yte,dtype:"bool"}),xte={kernelName:nu,backendName:"webgl",kernelFunc:vte},wte="return float(isinf(x));",Ite=Ye({opSnippet:wte,dtype:"bool"}),kte={kernelName:ru,backendName:"webgl",kernelFunc:Ite},Ste="return float(isnan(x));",Tte=Ye({opSnippet:Ste,dtype:"bool"}),Cte={kernelName:su,backendName:"webgl",kernelFunc:Tte},Nte="return float(a < b);",_te=`
   return vec4(lessThan(a, b));
-`,Nte=cn({opSnippet:Ste,packedOpSnippet:Tte,cpuKernelImpl:h7,dtype:"bool"}),Cte={kernelName:su,backendName:"webgl",kernelFunc:Nte},_te="return float(a <= b);",Ete=`
+`,Ete=dn({opSnippet:Nte,packedOpSnippet:_te,cpuKernelImpl:g9,dtype:"bool"}),Ate={kernelName:au,backendName:"webgl",kernelFunc:Ete},$te="return float(a <= b);",Dte=`
   return vec4(lessThanEqual(a, b));
-`,Ate=cn({opSnippet:_te,packedOpSnippet:Ete,cpuKernelImpl:m7,dtype:"bool"}),$te={kernelName:iu,backendName:"webgl",kernelFunc:Ate};function Fte(e){let{backend:t,attrs:n}=e,{start:a,stop:r,num:s}=n,i=f7(a,r,s);return t.makeTensorInfo([i.length],"float32",i)}var Dte={kernelName:Cm,backendName:"webgl",kernelFunc:Fte},Rte=tp+`
+`,Fte=dn({opSnippet:$te,packedOpSnippet:Dte,cpuKernelImpl:b9,dtype:"bool"}),Rte={kernelName:ou,backendName:"webgl",kernelFunc:Fte};function Pte(e){let{backend:t,attrs:n}=e,{start:r,stop:s,num:a}=n,o=y9(r,s,a);return t.makeTensorInfo([o.length],"float32",o)}var Ote={kernelName:Ef,backendName:"webgl",kernelFunc:Pte},Mte=tl+`
   return x < 0.0 ? 0./0. : log(x);
-`,Mte=`
+`,Lte=`
   vec4 result = log(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);
@@ -3534,18 +3534,18 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
   result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);
   result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);
   return result;
-`,Pte=Ye({opSnippet:Rte,packedOpSnippet:Mte,cpuKernelImpl:g7}),Ote={kernelName:Pi,backendName:"webgl",kernelFunc:Pte},Lte=tp+`
+`,zte=Ye({opSnippet:Mte,packedOpSnippet:Lte,cpuKernelImpl:v9}),Bte={kernelName:Lo,backendName:"webgl",kernelFunc:zte},Wte=tl+`
   return log(1.0 + x);
-`,zte=Ye({opSnippet:Lte}),Wte={kernelName:ou,backendName:"webgl",kernelFunc:zte},Bte="return float(a >= 1.0 && b >= 1.0);",Vte=`
+`,Vte=Ye({opSnippet:Wte}),Ute={kernelName:iu,backendName:"webgl",kernelFunc:Vte},Gte="return float(a >= 1.0 && b >= 1.0);",Hte=`
   return vec4(
     vec4(greaterThanEqual(a, vec4(1.0))) *
     vec4(greaterThanEqual(b, vec4(1.0))));
-`,Ute=cn({opSnippet:Bte,packedOpSnippet:Vte,dtype:"bool"}),Gte={kernelName:lu,backendName:"webgl",kernelFunc:Ute},Hte="return float(!(x >= 1.0));",jte=Ye({opSnippet:Hte}),qte={kernelName:uu,backendName:"webgl",kernelFunc:jte},Kte="return float(a >= 1.0 || b >= 1.0);",Xte=`
+`,qte=dn({opSnippet:Gte,packedOpSnippet:Hte,dtype:"bool"}),jte={kernelName:cu,backendName:"webgl",kernelFunc:qte},Kte="return float(!(x >= 1.0));",Xte=Ye({opSnippet:Kte}),Yte={kernelName:uu,backendName:"webgl",kernelFunc:Xte},Zte="return float(a >= 1.0 || b >= 1.0);",Jte=`
   return min(
     vec4(greaterThanEqual(a, vec4(1.0))) +
     vec4(greaterThanEqual(b, vec4(1.0))),
     vec4(1.0));
-`,Yte=cn({opSnippet:Kte,packedOpSnippet:Xte,dtype:"bool"}),Zte={kernelName:pu,backendName:"webgl",kernelFunc:Yte},Jte=class{constructor(e,t,n,a,r){this.variableNames=["x"],this.outputShape=[];let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${a}) * sum`;r===.5?o=`inversesqrt(${l})`:r===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${r}));`,this.userCode=`
+`,Qte=dn({opSnippet:Zte,packedOpSnippet:Jte,dtype:"bool"}),ene={kernelName:lu,backendName:"webgl",kernelFunc:Qte},tne=class{constructor(e,t,n,r,s){this.variableNames=["x"],this.outputShape=[];let a=t,o=e[3]-1;this.outputShape=e;let i,c=`float(${n}) + float(${r}) * sum`;s===.5?i=`inversesqrt(${c})`:s===1?i=`1.0/(${c})`:i=`exp(log(${c}) * float(-${s}));`,this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3554,17 +3554,17 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         int d = coords[3];
         float x = getX(b, r, c, d);
         float sum = 0.0;
-        for (int j = -${s}; j <= ${s}; j++) {
+        for (int j = -${a}; j <= ${a}; j++) {
           int idx = d + j;
-          if (idx >= 0 && idx <=  ${i}) {
+          if (idx >= 0 && idx <=  ${o}) {
             float z = getX(b, r, c, idx);
             sum += z * z;
           }
         }
-        float val = x * ${o};
+        float val = x * ${i};
         setOutput(val);
       }
-    `}},Qte=class{constructor(e,t,n,a,r){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${a}) * sum`;r===.5?o=`inversesqrt(${l})`:r===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${r}));`,this.userCode=`
+    `}},nne=class{constructor(e,t,n,r,s){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let a=t,o=e[3]-1;this.outputShape=e;let i,c=`float(${n}) + float(${r}) * sum`;s===.5?i=`inversesqrt(${c})`:s===1?i=`1.0/(${c})`:i=`exp(log(${c}) * float(-${s}));`,this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords.x;
@@ -3588,7 +3588,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0
         );
 
-        int firstChannel = d - ${s};
+        int firstChannel = d - ${a};
         vec2 cache = vec2(0.);
         if(firstChannel >= 0){
           vec4 firstChannelFrag = getX(b, r, c, firstChannel);
@@ -3599,10 +3599,10 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
 
         ivec2 depth = ivec2(d, d + 1);
-        for (int j = - ${s}; j <= ${s}; j++) {
+        for (int j = - ${a}; j <= ${a}; j++) {
           ivec2 idx = depth + j;
           bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));
-          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));
+          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${o}));
 
           bool depthInRange = aboveLowerBound.x && belowUpperBound.x;
           bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;
@@ -3623,10 +3623,10 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             sum += z * z;
           }
         }
-        vec4 result = xAtOutputCoords * ${o};
+        vec4 result = xAtOutputCoords * ${i};
         setOutput(result);
       }
-    `}},ene=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=a,u=H().getBool("WEBGL_PACK_NORMALIZATION")?new Qte(r.shape,s,i,o,l):new Jte(r.shape,s,i,o,l);return n.runWebGLProgram(u,[r],r.dtype)},tne={kernelName:yc,backendName:"webgl",kernelFunc:ene},nne=class{constructor(e,t,n,a,r){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=n,this.alpha=a,this.beta=r,this.userCode=`
+    `}},rne=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{depthRadius:a,bias:o,alpha:i,beta:c}=r,u=q().getBool("WEBGL_PACK_NORMALIZATION")?new nne(s.shape,a,o,i,c):new tne(s.shape,a,o,i,c);return n.runWebGLProgram(u,[s],s.dtype)},sne={kernelName:bd,backendName:"webgl",kernelFunc:rne},ane=class{constructor(e,t,n,r,s){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=n,this.alpha=r,this.beta=s,this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3655,19 +3655,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             }
           }
 
-          norm = float(${a}) * norm + float(${n});
+          norm = float(${r}) * norm + float(${n});
 
           for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){
             if (k < depthBegin){
               continue;
             }
             else if (k >= depthBegin && k < depthEnd){
-              float dyi = -2.0 * float(${a})
-                * float(${r})
+              float dyi = -2.0 * float(${r})
+                * float(${s})
                 * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)
                 / norm;
               if (k == d) {
-                dyi += pow(norm, -1.0 * ${r});
+                dyi += pow(norm, -1.0 * ${s});
               }
               if (k == coords[3]) {
                 dyi *= getDy(b, r, c, d);
@@ -3681,17 +3681,17 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
       }
       setOutput(result);
       }
-    `}},ane=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:p}=a,d=new nne(r.shape,o,l,u,p);return n.runWebGLProgram(d,[r,s,i],r.dtype)},rne={kernelName:_m,backendName:"webgl",kernelFunc:ane};function sne(e,t,n,a){let r=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/r,i=de({inputs:{x:e},attrs:{shape:[s,r]},backend:a}),o=Io(i,e.dtype,"max",a),l=de({inputs:{x:o},attrs:{shape:n},backend:a});return a.disposeIntermediateTensorInfo(i),a.disposeIntermediateTensorInfo(o),l}function $E(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reductionIndices:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=p!=null,c=n.shouldExecuteOnCPU([r]),h=r;if(d){if(c){let b=n.texData.get(h.dataId).values,x=new Array(o);for(let T=0;T<x.length;T++)x[T]=r.shape[p[T]];let w=s1(b,r.shape,r.dtype,p,x);h=n.makeTensorInfo(x,r.dtype);let I=n.texData.get(h.dataId);I.values=w}else h=Yf(r,p,n);u=N.getInnerMostAxes(u.length,o)}N.assertAxesAreInnerMostDims("max",u,o);let[m,f]=N.computeOutAndReduceShapes(h.shape,u),g=m;i&&(g=N.expandShapeToKeepDim(m,l));let y;if(c){let b=n.texData.get(h.dataId).values,x=y7(b,v.sizeFromShape(f),g,r.dtype);y=n.makeTensorInfo(g,r.dtype);let w=n.texData.get(y.dataId);w.values=x}else y=sne(h,f,g,n);return d&&n.disposeIntermediateTensorInfo(h),y}var ine={kernelName:Oi,backendName:"webgl",kernelFunc:$E},one=i1+`
+    `}},one=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:s,y:a,dy:o}=t,{depthRadius:i,bias:c,alpha:u,beta:l}=r,p=new ane(s.shape,i,c,u,l);return n.runWebGLProgram(p,[s,a,o],s.dtype)},ine={kernelName:Af,backendName:"webgl",kernelFunc:one};function cne(e,t,n,r){let s=w.sizeFromShape(t),o=w.sizeFromShape(e.shape)/s,i=he({inputs:{x:e},attrs:{shape:[o,s]},backend:r}),c=Si(i,e.dtype,"max",r),u=he({inputs:{x:c},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(c),u}function FE(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{reductionIndices:a,keepDims:o}=r,i=s.shape.length,c=w.parseAxisParam(a,s.shape),u=c,l=N.getAxesPermutation(u,i),p=l!=null,d=n.shouldExecuteOnCPU([s]),h=s;if(p){if(d){let v=n.texData.get(h.dataId).values,x=new Array(i);for(let C=0;C<x.length;C++)x[C]=s.shape[l[C]];let k=aI(v,s.shape,s.dtype,l,x);h=n.makeTensorInfo(x,s.dtype);let S=n.texData.get(h.dataId);S.values=k}else h=Zm(s,l,n);u=N.getInnerMostAxes(u.length,i)}N.assertAxesAreInnerMostDims("max",u,i);let[f,m]=N.computeOutAndReduceShapes(h.shape,u),g=f;o&&(g=N.expandShapeToKeepDim(f,c));let b;if(d){let v=n.texData.get(h.dataId).values,x=x9(v,w.sizeFromShape(m),g,s.dtype);b=n.makeTensorInfo(g,s.dtype);let k=n.texData.get(b.dataId);k.values=x}else b=cne(h,m,g,n);return p&&n.disposeIntermediateTensorInfo(h),b}var une={kernelName:zo,backendName:"webgl",kernelFunc:FE},lne=oI+`
   return max(a, b);
-`,lne=`
+`,dne=`
   vec4 result = vec4(max(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+ld+`
+  `+cp+`
   return result;
-`,une=cn({opSnippet:one,packedOpSnippet:lne,cpuKernelImpl:b7}),pne={kernelName:Li,backendName:"webgl",kernelFunc:une};function cne(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t;Yu(r,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=a,u=1;v.assert(N.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let p=N.computePool2DInfo(r.shape,s,i,u,o,l);if(p.filterWidth===1&&p.filterHeight===1&&v.arraysEqual(p.inShape,p.outShape))return aa({inputs:{x:r},backend:n});let d=new oc(p,"max",!1);return n.runWebGLProgram(d,[r],r.dtype)}var dne={kernelName:zi,backendName:"webgl",kernelFunc:cne};function hne(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{filterSize:s,strides:i,pad:o,dataFormat:l,dimRoundingMode:u}=a,p=[1,1,1],d=N.computePool3DInfo(r.shape,s,i,p,o,u,l),c=new l1(d,"max",!1);return n.runWebGLProgram(c,[r],r.dtype)}var mne={kernelName:bc,backendName:"webgl",kernelFunc:hne},fne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideHeight,n=e.strideWidth,a=e.dilationHeight,r=e.effectiveFilterHeight,s=e.effectiveFilterWidth,i=r-1-e.padInfo.top,o=s-1-e.padInfo.left,l=r*s-1;this.userCode=`
-      const ivec2 pads = ivec2(${i}, ${o});
+`,pne=dn({opSnippet:lne,packedOpSnippet:dne,cpuKernelImpl:w9}),hne={kernelName:Bo,backendName:"webgl",kernelFunc:pne};function fne(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t;Yu(s,"maxPool");let{filterSize:a,strides:o,pad:i,dimRoundingMode:c}=r,u=1;w.assert(N.eitherStridesOrDilationsAreOne(o,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${o} and dilations '${u}'`);let l=N.computePool2DInfo(s.shape,a,o,u,i,c);if(l.filterWidth===1&&l.filterHeight===1&&w.arraysEqual(l.inShape,l.outShape))return sr({inputs:{x:s},backend:n});let p=new id(l,"max",!1);return n.runWebGLProgram(p,[s],s.dtype)}var mne={kernelName:Wo,backendName:"webgl",kernelFunc:fne};function gne(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{filterSize:a,strides:o,pad:i,dataFormat:c,dimRoundingMode:u}=r,l=[1,1,1],p=N.computePool3DInfo(s.shape,a,o,l,i,u,c),d=new cI(p,"max",!1);return n.runWebGLProgram(d,[s],s.dtype)}var bne={kernelName:yd,backendName:"webgl",kernelFunc:gne},yne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideHeight,n=e.strideWidth,r=e.dilationHeight,s=e.effectiveFilterHeight,a=e.effectiveFilterWidth,o=s-1-e.padInfo.top,i=a-1-e.padInfo.left,c=s*a-1;this.userCode=`
+      const ivec2 pads = ivec2(${o}, ${i});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3705,8 +3705,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${r};
-          wR += ${a}) {
+        for (int wR = 0; wR < ${s};
+          wR += ${r}) {
           float dyR = float(dyRCorner + wR) / ${t}.0;
 
           if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
@@ -3714,7 +3714,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${s}; wC++) {
+          for (int wC = 0; wC < ${a}; wC++) {
             float dyC = float(dyCCorner + wC) / ${n}.0;
 
             if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
@@ -3724,11 +3724,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             int idyC = int(dyC);
 
             float dyValue = getDy(b, idyR, idyC, d);
-            int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));
+            int maxPosValue = ${c} - int(getMaxPos(b, idyR, idyC, d));
 
             // Get the current value, check it against the value from the
             // position matrix.
-            int curPosValue = wR * ${s} + wC;
+            int curPosValue = wR * ${a} + wC;
             float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
             dotProd += dyValue * mask;
@@ -3736,8 +3736,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}},gne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideDepth,n=e.strideHeight,a=e.strideWidth,r=e.dilationDepth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterDepth,l=e.effectiveFilterHeight,u=e.effectiveFilterWidth,p=o-1-e.padInfo.front,d=l-1-e.padInfo.top,c=u-1-e.padInfo.left,h=o*l*u-1;this.userCode=`
-      const ivec3 pads = ivec3(${p}, ${d}, ${c});
+    `}},vne=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,s=e.dilationDepth,a=e.dilationHeight,o=e.dilationWidth,i=e.effectiveFilterDepth,c=e.effectiveFilterHeight,u=e.effectiveFilterWidth,l=i-1-e.padInfo.front,p=c-1-e.padInfo.top,d=u-1-e.padInfo.left,h=i*c*u-1;this.userCode=`
+      const ivec3 pads = ivec3(${l}, ${p}, ${d});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -3754,8 +3754,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${o};
-           wD += ${r}) {
+        for (int wD = 0; wD < ${i};
+           wD += ${s}) {
           float dyD = float(dyDCorner + wD) / ${t}.0;
 
           if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
@@ -3763,8 +3763,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${l};
-              wR += ${s}) {
+          for (int wR = 0; wR < ${c};
+              wR += ${a}) {
             float dyR = float(dyRCorner + wR) / ${n}.0;
 
             if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
@@ -3774,8 +3774,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
             int idyR = int(dyR);
 
             for (int wC = 0; wC < ${u};
-                wC += ${i}) {
-              float dyC = float(dyCCorner + wC) / ${a}.0;
+                wC += ${o}) {
+              float dyC = float(dyCCorner + wC) / ${r}.0;
 
               if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
                   fract(dyC) > 0.0) {
@@ -3790,7 +3790,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
               // Get the current value, check it against the value from the
               // position matrix.
               int curPosValue =
-                  wD * ${l} * ${u} +
+                  wD * ${c} * ${u} +
                   wR * ${u} + wC;
               float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
@@ -3800,107 +3800,107 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         }
         setOutput(dotProd);
       }
-    `}};function yne(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:p}=a,d=[1,1,1],c=N.computePool3DInfo(i.shape,o,l,d,u,p),h=new l1(c,"max",!0),m=n.runWebGLProgram(h,[i],i.dtype),f=new gne(c),g=n.runWebGLProgram(f,[r,m],i.dtype);return n.disposeIntermediateTensorInfo(m),g}var bne={kernelName:Am,backendName:"webgl",kernelFunc:yne};function xne(e){let{inputs:t,backend:n,attrs:a}=e,{dy:r,input:s,output:i}=t,o=s;Yu([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:p,dimRoundingMode:d}=a,c=N.computePool2DInfo(o.shape,l,u,1,p,d),h=!0,m=new oc(c,"max",h),f=n.runWebGLProgram(m,[o],o.dtype),g=new fne(c),y=n.runWebGLProgram(g,[r,f],o.dtype);return n.disposeIntermediateTensorInfo(f),y}var vne={kernelName:Em,backendName:"webgl",kernelFunc:xne};function wne(e,t,n,a){let r=new oc(n,"max",!1),s=a.runWebGLProgram(r,[e],"float32");r=new oc(n,"max",!0,!0,t);let i=a.runWebGLProgram(r,[e],"float32");return[s,i]}var kne={kernelName:$m,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{filterSize:r,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;v.assert(a.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${a.shape.length}.`);let u=[1,1];v.assert(N.eitherStridesOrDilationsAreOne(s,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${u}'`);let p=N.computePool2DInfo(a.shape,r,s,u,i),[d,c]=wne(a,o,p,l);return[d,c]}};function Ine(e,t,n,a){let r=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/r,i=de({inputs:{x:e},attrs:{shape:[s,r]},backend:a}),o=Io(i,"float32","mean",a),l=de({inputs:{x:o},attrs:{shape:n},backend:a});return a.disposeIntermediateTensorInfo(i),a.disposeIntermediateTensorInfo(o),l}var Sne={kernelName:Wi,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:a}=e,{keepDims:r,axis:s}=t,i=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,p=N.getAxesPermutation(u,o),d=p!=null,c=i.shouldExecuteOnCPU([a]),h=[],m=a;if(d){if(c){let x=i.texData.get(m.dataId).values,w=new Array(o);for(let C=0;C<w.length;C++)w[C]=a.shape[p[C]];let I=s1(x,a.shape,a.dtype,p,w);m=i.makeTensorInfo(w,a.dtype);let T=i.texData.get(m.dataId);T.values=I}else m=Yf(a,p,i);h.push(m),u=N.getInnerMostAxes(u.length,o)}N.assertAxesAreInnerMostDims("sum",u,o);let[f,g]=N.computeOutAndReduceShapes(m.shape,u),y=f;r&&(y=N.expandShapeToKeepDim(f,l));let b=Ine(m,g,y,i);for(let x of h)i.disposeIntermediateTensorInfo(x);return b}};function Tne(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=v.parseAxisParam(s,r.shape),u=l,p=N.getAxesPermutation(u,o),d=r;p!=null&&(d=Sn({inputs:{x:r},backend:n,attrs:{perm:p}}),u=N.getInnerMostAxes(u.length,r.shape.length)),N.assertAxesAreInnerMostDims("min",u,o);let[c,h]=N.computeOutAndReduceShapes(d.shape,u),m=v.sizeFromShape(h),f=de({inputs:{x:d},backend:n,attrs:{shape:[-1,m]}}),g=Io(f,f.dtype,"min",n),y;if(i){let b=N.expandShapeToKeepDim(c,l);y=de({inputs:{x:g},backend:n,attrs:{shape:b}})}else y=de({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(g),p!=null&&n.disposeIntermediateTensorInfo(d),y}var Nne={kernelName:Bi,backendName:"webgl",kernelFunc:Tne},Cne=i1+`
+    `}};function xne(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a}=t,o=a,{filterSize:i,strides:c,pad:u,dimRoundingMode:l}=r,p=[1,1,1],d=N.computePool3DInfo(o.shape,i,c,p,u,l),h=new cI(d,"max",!0),f=n.runWebGLProgram(h,[o],o.dtype),m=new vne(d),g=n.runWebGLProgram(m,[s,f],o.dtype);return n.disposeIntermediateTensorInfo(f),g}var wne={kernelName:Df,backendName:"webgl",kernelFunc:xne};function Ine(e){let{inputs:t,backend:n,attrs:r}=e,{dy:s,input:a,output:o}=t,i=a;Yu([a,o],"maxPoolGrad");let{filterSize:c,strides:u,pad:l,dimRoundingMode:p}=r,d=N.computePool2DInfo(i.shape,c,u,1,l,p),h=!0,f=new id(d,"max",h),m=n.runWebGLProgram(f,[i],i.dtype),g=new yne(d),b=n.runWebGLProgram(g,[s,m],i.dtype);return n.disposeIntermediateTensorInfo(m),b}var kne={kernelName:$f,backendName:"webgl",kernelFunc:Ine};function Sne(e,t,n,r){let s=new id(n,"max",!1),a=r.runWebGLProgram(s,[e],"float32");s=new id(n,"max",!0,!0,t);let o=r.runWebGLProgram(s,[e],"float32");return[a,o]}var Tne={kernelName:Ff,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{filterSize:s,strides:a,pad:o,includeBatchInIndex:i}=t,c=n;w.assert(r.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${r.shape.length}.`);let u=[1,1];w.assert(N.eitherStridesOrDilationsAreOne(a,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${a} and dilations '${u}'`);let l=N.computePool2DInfo(r.shape,s,a,u,o),[p,d]=Sne(r,i,l,c);return[p,d]}};function Cne(e,t,n,r){let s=w.sizeFromShape(t),o=w.sizeFromShape(e.shape)/s,i=he({inputs:{x:e},attrs:{shape:[o,s]},backend:r}),c=Si(i,"float32","mean",r),u=he({inputs:{x:c},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(c),u}var Nne={kernelName:Vo,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{keepDims:s,axis:a}=t,o=n,i=r.shape.length,c=w.parseAxisParam(a,r.shape),u=c,l=N.getAxesPermutation(u,i),p=l!=null,d=o.shouldExecuteOnCPU([r]),h=[],f=r;if(p){if(d){let x=o.texData.get(f.dataId).values,k=new Array(i);for(let E=0;E<k.length;E++)k[E]=r.shape[l[E]];let S=aI(x,r.shape,r.dtype,l,k);f=o.makeTensorInfo(k,r.dtype);let C=o.texData.get(f.dataId);C.values=S}else f=Zm(r,l,o);h.push(f),u=N.getInnerMostAxes(u.length,i)}N.assertAxesAreInnerMostDims("sum",u,i);let[m,g]=N.computeOutAndReduceShapes(f.shape,u),b=m;s&&(b=N.expandShapeToKeepDim(m,c));let y=Cne(f,g,b,o);for(let v of h)o.disposeIntermediateTensorInfo(v);return y}};function _ne(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r,i=s.shape.length,c=w.parseAxisParam(a,s.shape),u=c,l=N.getAxesPermutation(u,i),p=s;l!=null&&(p=Nn({inputs:{x:s},backend:n,attrs:{perm:l}}),u=N.getInnerMostAxes(u.length,s.shape.length)),N.assertAxesAreInnerMostDims("min",u,i);let[d,h]=N.computeOutAndReduceShapes(p.shape,u),f=w.sizeFromShape(h),m=he({inputs:{x:p},backend:n,attrs:{shape:[-1,f]}}),g=Si(m,m.dtype,"min",n),b;if(o){let y=N.expandShapeToKeepDim(d,c);b=he({inputs:{x:g},backend:n,attrs:{shape:y}})}else b=he({inputs:{x:g},backend:n,attrs:{shape:d}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),l!=null&&n.disposeIntermediateTensorInfo(p),b}var Ene={kernelName:Uo,backendName:"webgl",kernelFunc:_ne},Ane=oI+`
   return min(a, b);
-`,_ne=`
+`,$ne=`
   vec4 result = vec4(min(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+ld+`
+  `+cp+`
   return result;
-`,Ene=cn({opSnippet:Cne,packedOpSnippet:_ne,cpuKernelImpl:x7}),Ane={kernelName:Vi,backendName:"webgl",kernelFunc:Ene},$ne=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=t.map((u,p)=>u[0]+e[p]+u[1]);let a=e.length,r=gt(a),s=t.map(u=>u[0]).join(","),i=t.map((u,p)=>u[0]+e[p]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,a),l=n==="reflect"?0:1;if(a===1){this.userCode=`
-        int start = ${s};
-        int end = ${i};
+`,Dne=dn({opSnippet:Ane,packedOpSnippet:$ne,cpuKernelImpl:I9}),Fne={kernelName:Go,backendName:"webgl",kernelFunc:Dne},Rne=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=t.map((u,l)=>u[0]+e[l]+u[1]);let r=e.length,s=mt(r),a=t.map(u=>u[0]).join(","),o=t.map((u,l)=>u[0]+e[l]).join(","),i=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r),c=n==="reflect"?0:1;if(r===1){this.userCode=`
+        int start = ${a};
+        int end = ${o};
 
         void main() {
           int outC = getOutputCoords();
           if (outC < start) {
-            outC = start * 2 - outC - ${l};
+            outC = start * 2 - outC - ${c};
           } else if(outC >= end) {
-            outC = (end - 1) * 2 - outC + ${l};
+            outC = (end - 1) * 2 - outC + ${c};
           }
           setOutput(getX(outC - start));
         }
       `;return}this.userCode=`
-      ${r} start = ${r}(${s});
-      ${r} end = ${r}(${i});
+      ${s} start = ${s}(${a});
+      ${s} end = ${s}(${o});
 
       void main() {
-        ${r} outC = getOutputCoords();
-        for (int i = 0; i < ${a}; i++) {
+        ${s} outC = getOutputCoords();
+        for (int i = 0; i < ${r}; i++) {
           if (outC[i] < start[i]) {
-            outC[i] = start[i] * 2 - outC[i] - ${l};
+            outC[i] = start[i] * 2 - outC[i] - ${c};
           } else if(outC[i] >= end[i]) {
-            outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};
+            outC[i] = (end[i] - 1) * 2 - outC[i] + ${c};
           }
         }
-        ${r} coords = outC - start;
-        setOutput(getX(${o}));
+        ${s} coords = outC - start;
+        setOutput(getX(${i}));
       }
-    `}},Fne=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((h,m)=>h[0]+e[m]+h[1]);let a=e.length,r=gt(a),s=t.map(h=>h[0]).join(","),i=t.map((h,m)=>h[0]+e[m]).join(","),o=kn("rc",a),l=kn("source",a),u=`${o[a-1]} < ${this.outputShape[a-1]}`,p=a===1?"source":`vec2(${l.slice(-2).join()})`,d=n==="reflect"?0:1,c="";if(a===1){let h=`
-        ${r} source = rc;
+    `}},Pne=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((h,f)=>h[0]+e[f]+h[1]);let r=e.length,s=mt(r),a=t.map(h=>h[0]).join(","),o=t.map((h,f)=>h[0]+e[f]).join(","),i=Tn("rc",r),c=Tn("source",r),u=`${i[r-1]} < ${this.outputShape[r-1]}`,l=r===1?"source":`vec2(${c.slice(-2).join()})`,p=n==="reflect"?0:1,d="";if(r===1){let h=`
+        ${s} source = rc;
         if (source < start) {
-          source = start * 2 - source - ${d};
+          source = start * 2 - source - ${p};
         } else if (source >= end) {
-          source = (end - 1) * 2 - source + ${d};
+          source = (end - 1) * 2 - source + ${p};
         }
         source -= start;
-      `;c=`
-        ${r} rc = outputLoc;
+      `;d=`
+        ${s} rc = outputLoc;
         ${h}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${o[a-1]} += 1;
+        result[0] = getChannel(getX(${c.join()}), ${l});
+        ${i[r-1]} += 1;
         if(${u}) {
           ${h}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+          result[1] = getChannel(getX(${c.join()}), ${l});
         }
       `}else{let h=`
-        ${r} source = rc;
-        ${r} lt = ${r}(lessThan(source, start));
-        ${r} gte = ${r}(greaterThanEqual(source, end));
-        ${r} orig = 1 - (lt + gte);
+        ${s} source = rc;
+        ${s} lt = ${s}(lessThan(source, start));
+        ${s} gte = ${s}(greaterThanEqual(source, end));
+        ${s} orig = 1 - (lt + gte);
         source = orig * source +
-                lt * (start * 2 - source - ${d}) +
-                gte * ((end - 1) * 2 - source + ${d});
+                lt * (start * 2 - source - ${p}) +
+                gte * ((end - 1) * 2 - source + ${p});
         source -= start;
-      `;c=`
-        ${r} rc = outputLoc;
+      `;d=`
+        ${s} rc = outputLoc;
         ${h}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${o[a-1]} += 1;
+        result[0] = getChannel(getX(${c.join()}), ${l});
+        ${i[r-1]} += 1;
         if(${u}) {
           ${h}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+          result[1] = getChannel(getX(${c.join()}), ${l});
         }
         rc = outputLoc;
-        ${o[a-2]} += 1;
-        if(${o[a-2]} < ${this.outputShape[a-2]}) {
+        ${i[r-2]} += 1;
+        if(${i[r-2]} < ${this.outputShape[r-2]}) {
           ${h}
-          result[2] = getChannel(getX(${l.join()}), ${p});
-          ${o[a-1]} += 1;
+          result[2] = getChannel(getX(${c.join()}), ${l});
+          ${i[r-1]} += 1;
           if(${u}) {
             ${h}
-            result[3] = getChannel(getX(${l.join()}), ${p});
+            result[3] = getChannel(getX(${c.join()}), ${l});
           }
         }
       `}this.userCode=`
-      const ${r} start = ${r}(${s});
-      const ${r} end = ${r}(${i});
+      const ${s} start = ${s}(${a});
+      const ${s} end = ${s}(${o});
 
       void main() {
-        ${r} outputLoc = getOutputCoords();
+        ${s} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
-        ${c}
+        ${d}
         setOutput(result);
       }
-    `}},Dne=({inputs:e,backend:t,attrs:n})=>{let{x:a}=e,{paddings:r,mode:s}=n,i=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new Fne(a.shape,r,s):new $ne(a.shape,r,s);return t.runWebGLProgram(i,[a],a.dtype)},Rne={kernelName:Ui,backendName:"webgl",kernelFunc:Dne},Mne=`if (b == 0.0) return NAN;
-  return mod(a, b);`,Pne=`
+    `}},One=({inputs:e,backend:t,attrs:n})=>{let{x:r}=e,{paddings:s,mode:a}=n,o=q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new Pne(r.shape,s,a):new Rne(r.shape,s,a);return t.runWebGLProgram(o,[r],r.dtype)},Mne={kernelName:Ho,backendName:"webgl",kernelFunc:One},Lne=`if (b == 0.0) return NAN;
+  return mod(a, b);`,zne=`
   vec4 result = mod(a, b);
   bvec4 isNaN = equal(b, vec4(0.0));
-  `+ld+`
+  `+cp+`
   return result;
-`,One=cn({opSnippet:Mne,packedOpSnippet:Pne}),Lne={kernelName:cu,backendName:"webgl",kernelFunc:One},zne=class{constructor(e,t,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,n],this.userCode=`
+`,Bne=dn({opSnippet:Lne,packedOpSnippet:zne}),Wne={kernelName:du,backendName:"webgl",kernelFunc:Bne},Vne=class{constructor(e,t,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,n],this.userCode=`
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
@@ -3920,11 +3920,11 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,x9=Ye({opSnippet:b9}),v9={kernelNam
         // If no other event happened, last event happened.
         setOutput(float(${t-1}));
       }
-    `}},Wne=`
+    `}},Une=`
 if (a == b) {
   return 1.0;
 };
-return a / b;`,Bne=`
+return a / b;`,Gne=`
   // vec4 one = vec4(equal(a, b));
   // return one + (vec4(1.0) - one) * a / b;
   vec4 result = a / b;
@@ -3942,9 +3942,9 @@ return a / b;`,Bne=`
   }
 
   return result;
-`,FE=cn({opSnippet:Wne,packedOpSnippet:Bne,checkOutOfBounds:!0}),Vne={kernelName:Ci,backendName:"webgl",kernelFunc:FE},FI="return a - b;",DE=cn({opSnippet:FI,packedOpSnippet:FI,supportsComplex:!0,cpuKernelImpl:z7}),Une={kernelName:uo,backendName:"webgl",kernelFunc:DE};function RE(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{dim:s}=a,i=v.parseAxisParam([s],r.shape),o=$E({inputs:{x:r},backend:n,attrs:{reductionIndices:i,keepDims:!1}}),l=N.expandShapeToKeepDim(o.shape,i),u=de({inputs:{x:o},backend:n,attrs:{shape:l}}),p=DE({inputs:{a:r,b:u},backend:n}),d=_E({inputs:{x:p},backend:n}),c=Zf({inputs:{x:d},backend:n,attrs:{axis:i,keepDims:!1}}),h=de({inputs:{x:c},backend:n,attrs:{shape:l}}),m=FE({inputs:{a:d,b:h},backend:n});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(h),m}var Gne={kernelName:oo,backendName:"webgl",kernelFunc:RE};function Hne(e){let{inputs:t,backend:n,attrs:a}=e,{logits:r}=t,{numSamples:s,seed:i,normalized:o}=a,l=o?r:RE({inputs:{logits:r},backend:n,attrs:{dim:r.shape.length-1}}),u=l.shape[0],p=l.shape[1],d=new zne(u,p,s),c=[[i]],h=n.runWebGLProgram(d,[l],"int32",c);return o||n.disposeIntermediateTensorInfo(l),h}var jne={kernelName:Fm,backendName:"webgl",kernelFunc:Hne},qne=Ma+`
+`,RE=dn({opSnippet:Une,packedOpSnippet:Gne,checkOutOfBounds:!0}),Hne={kernelName:Eo,backendName:"webgl",kernelFunc:RE},F1="return a - b;",PE=dn({opSnippet:F1,packedOpSnippet:F1,supportsComplex:!0,cpuKernelImpl:V9}),qne={kernelName:di,backendName:"webgl",kernelFunc:PE};function OE(e){let{inputs:t,backend:n,attrs:r}=e,{logits:s}=t,{dim:a}=r,o=w.parseAxisParam([a],s.shape),i=FE({inputs:{x:s},backend:n,attrs:{reductionIndices:o,keepDims:!1}}),c=N.expandShapeToKeepDim(i.shape,o),u=he({inputs:{x:i},backend:n,attrs:{shape:c}}),l=PE({inputs:{a:s,b:u},backend:n}),p=AE({inputs:{x:l},backend:n}),d=Jm({inputs:{x:p},backend:n,attrs:{axis:o,keepDims:!1}}),h=he({inputs:{x:d},backend:n,attrs:{shape:c}}),f=RE({inputs:{a:p,b:h},backend:n});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(h),f}var jne={kernelName:ui,backendName:"webgl",kernelFunc:OE};function Kne(e){let{inputs:t,backend:n,attrs:r}=e,{logits:s}=t,{numSamples:a,seed:o,normalized:i}=r,c=i?s:OE({inputs:{logits:s},backend:n,attrs:{dim:s.shape.length-1}}),u=c.shape[0],l=c.shape[1],p=new Vne(u,l,a),d=[[o]],h=n.runWebGLProgram(p,[c],"int32",d);return i||n.disposeIntermediateTensorInfo(c),h}var Xne={kernelName:Rf,backendName:"webgl",kernelFunc:Kne},Yne=Or+`
   return -x;
-`,Kne=`
+`,Zne=`
   vec4 result = -x;
   bvec4 isNaN = isnan(x);
 
@@ -3954,16 +3954,16 @@ return a / b;`,Bne=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`;function Xne(e){let{inputs:t,backend:n}=e,{x:a}=t;if(n.shouldExecuteOnCPU([a])){let s=n.texData.get(a.dataId),[i,o]=w7(s.values,a.shape,a.dtype);return n.makeTensorInfo(o,a.dtype,i)}let r;return H().getBool("WEBGL_PACK_UNARY_OPERATIONS")?r=new qs(a.shape,Kne):r=new Cr(a.shape,qne),n.runWebGLProgram(r,[a],a.dtype)}var Yne={kernelName:du,backendName:"webgl",kernelFunc:Xne},Zne=hr.nonMaxSuppressionV3Impl;function Jne(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=a,u=n.readSync(r.dataId),p=n.readSync(s.dataId),{selectedIndices:d}=Zne(u,p,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}var Qne={kernelName:mu,backendName:"webgl",kernelFunc:Jne},eae=hr.nonMaxSuppressionV4Impl;function tae(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=a,p=n.readSync(r.dataId),d=n.readSync(s.dataId),{selectedIndices:c,validOutputs:h}=eae(p,d,i,o,l,u);return[n.makeTensorInfo([c.length],"int32",new Int32Array(c)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var nae={kernelName:fu,backendName:"webgl",kernelFunc:tae},aae=hr.nonMaxSuppressionV5Impl;function rae(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:a}=e,{boxes:r,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=a,p=n.readSync(r.dataId),d=n.readSync(s.dataId),c=i,h=o,m=l,f=u,{selectedIndices:g,selectedScores:y}=aae(p,d,c,h,m,f);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var sae={kernelName:gu,backendName:"webgl",kernelFunc:rae},iae=class{constructor(e,t,n,a){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`
+`;function Jne(e){let{inputs:t,backend:n}=e,{x:r}=t;if(n.shouldExecuteOnCPU([r])){let a=n.texData.get(r.dataId),[o,i]=S9(a.values,r.shape,r.dtype);return n.makeTensorInfo(i,r.dtype,o)}let s;return q().getBool("WEBGL_PACK_UNARY_OPERATIONS")?s=new Ka(r.shape,Zne):s=new Ns(r.shape,Yne),n.runWebGLProgram(s,[r],r.dtype)}var Qne={kernelName:pu,backendName:"webgl",kernelFunc:Jne},ere=fs.nonMaxSuppressionV3Impl;function tre(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c}=r,u=n.readSync(s.dataId),l=n.readSync(a.dataId),{selectedIndices:p}=ere(u,l,o,i,c);return n.makeTensorInfo([p.length],"int32",new Int32Array(p))}var nre={kernelName:fu,backendName:"webgl",kernelFunc:tre},rre=fs.nonMaxSuppressionV4Impl;function sre(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c,padToMaxOutputSize:u}=r,l=n.readSync(s.dataId),p=n.readSync(a.dataId),{selectedIndices:d,validOutputs:h}=rre(l,p,o,i,c,u);return[n.makeTensorInfo([d.length],"int32",new Int32Array(d)),n.makeTensorInfo([],"int32",new Int32Array([h]))]}var are={kernelName:mu,backendName:"webgl",kernelFunc:sre},ore=fs.nonMaxSuppressionV5Impl;function ire(e){N.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:s,scores:a}=t,{maxOutputSize:o,iouThreshold:i,scoreThreshold:c,softNmsSigma:u}=r,l=n.readSync(s.dataId),p=n.readSync(a.dataId),d=o,h=i,f=c,m=u,{selectedIndices:g,selectedScores:b}=ore(l,p,d,h,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([b.length],"float32",new Float32Array(b))]}var cre={kernelName:gu,backendName:"webgl",kernelFunc:ire},ure=class{constructor(e,t,n,r){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`
       void main() {
         ivec2 coords = getOutputCoords();
         int index = round(getIndices(coords.x));
-        setOutput(mix(float(${a}), float(${n}),
+        setOutput(mix(float(${r}), float(${n}),
                       float(index == coords.y)));
       }
-    `}},oae=e=>{let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a,u=v.sizeFromShape(r.shape),p=new iae(u,i,o,l),d=de({inputs:{x:r},backend:n,attrs:{shape:[u]}}),c=n.runWebGLProgram(p,[d],s);n.disposeIntermediateTensorInfo(d);let h=[...r.shape,i],m=de({inputs:{x:c},backend:n,attrs:{shape:h}});return n.disposeIntermediateTensorInfo(c),m},lae={kernelName:Hi,backendName:"webgl",kernelFunc:oae};function rm(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="complex64"){let r=pd({inputs:{input:a},backend:n}),s=rm({inputs:{x:r},backend:n}),i=Jf({inputs:{input:a},backend:n}),o=rm({inputs:{x:i},backend:n}),l=_s({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return cd({attrs:{shape:a.shape,dtype:a.dtype,value:a.dtype==="string"?"":0},backend:n})}var uae={kernelName:Mu,backendName:"webgl",kernelFunc:rm};function ME(e){let{inputs:t,backend:n}=e,{x:a}=t;if(a.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(a.dtype==="complex64"){let r=pd({inputs:{input:a},backend:n}),s=ME({inputs:{x:r},backend:n}),i=Jf({inputs:{input:a},backend:n}),o=rm({inputs:{x:i},backend:n}),l=_s({inputs:{real:s,imag:o},backend:n});return n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}else return cd({attrs:{shape:a.shape,dtype:a.dtype,value:1},backend:n})}var pae={kernelName:yu,backendName:"webgl",kernelFunc:ME};function cae(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return vx({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=vx({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=vE({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeIntermediateTensorInfo(p)),u}var dae={kernelName:bu,backendName:"webgl",kernelFunc:cae},hae=class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((l,u)=>l[0]+e[u]+l[1]);let a=e.length,r=gt(a),s=t.map(l=>l[0]).join(","),i=t.map((l,u)=>l[0]+e[u]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,a);if(a===1){this.userCode=`
-        int start = ${s};
-        int end = ${i};
+    `}},lre=e=>{let{inputs:t,backend:n,attrs:r}=e,{indices:s}=t,{dtype:a,depth:o,onValue:i,offValue:c}=r,u=w.sizeFromShape(s.shape),l=new ure(u,o,i,c),p=he({inputs:{x:s},backend:n,attrs:{shape:[u]}}),d=n.runWebGLProgram(l,[p],a);n.disposeIntermediateTensorInfo(p);let h=[...s.shape,o],f=he({inputs:{x:d},backend:n,attrs:{shape:h}});return n.disposeIntermediateTensorInfo(d),f},dre={kernelName:jo,backendName:"webgl",kernelFunc:lre};function af(e){let{inputs:t,backend:n}=e,{x:r}=t;if(r.dtype==="complex64"){let s=lp({inputs:{input:r},backend:n}),a=af({inputs:{x:s},backend:n}),o=Qm({inputs:{input:r},backend:n}),i=af({inputs:{x:o},backend:n}),c=Ea({inputs:{real:a,imag:i},backend:n});return n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(i),c}else return dp({attrs:{shape:r.shape,dtype:r.dtype,value:r.dtype==="string"?"":0},backend:n})}var pre={kernelName:Pu,backendName:"webgl",kernelFunc:af};function ME(e){let{inputs:t,backend:n}=e,{x:r}=t;if(r.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(r.dtype==="complex64"){let s=lp({inputs:{input:r},backend:n}),a=ME({inputs:{x:s},backend:n}),o=Qm({inputs:{input:r},backend:n}),i=af({inputs:{x:o},backend:n}),c=Ea({inputs:{real:a,imag:i},backend:n});return n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(i),c}else return dp({attrs:{shape:r.shape,dtype:r.dtype,value:1},backend:n})}var hre={kernelName:bu,backendName:"webgl",kernelFunc:ME};function fre(e){let{inputs:t,backend:n,attrs:r}=e,{axis:s}=r;if(t.length===1)return xv({inputs:{input:t[0]},backend:n,attrs:{dim:s}});let a=t[0].shape,o=t[0].dtype;t.forEach(l=>{w.assertShapesMatch(a,l.shape,"All tensors passed to stack must have matching shapes"),w.assert(o===l.dtype,()=>"All tensors passed to stack must have matching dtypes")});let i=[],c=t.map(l=>{let p=xv({inputs:{input:l},backend:n,attrs:{dim:s}});return i.push(p),p}),u=IE({inputs:c,backend:n,attrs:{axis:s}});return i.forEach(l=>n.disposeIntermediateTensorInfo(l)),u}var mre={kernelName:yu,backendName:"webgl",kernelFunc:fre},gre=class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((c,u)=>c[0]+e[u]+c[1]);let r=e.length,s=mt(r),a=t.map(c=>c[0]).join(","),o=t.map((c,u)=>c[0]+e[u]).join(","),i=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r);if(r===1){this.userCode=`
+        int start = ${a};
+        int end = ${o};
 
         void main() {
           int outC = getOutputCoords();
@@ -3974,43 +3974,43 @@ return a / b;`,Bne=`
           }
         }
       `;return}this.userCode=`
-      ${r} start = ${r}(${s});
-      ${r} end = ${r}(${i});
+      ${s} start = ${s}(${a});
+      ${s} end = ${s}(${o});
 
       void main() {
-        ${r} outC = getOutputCoords();
+        ${s} outC = getOutputCoords();
         if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {
           setOutput(value);
         } else {
-          ${r} coords = outC - start;
-          setOutput(getX(${o}));
+          ${s} coords = outC - start;
+          setOutput(getX(${i}));
         }
       }
-    `}},mae=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((m,f)=>m[0]+e[f]+m[1]);let a=e.length,r=gt(a),s=t.map(m=>m[0]).join(","),i=t.map((m,f)=>m[0]+e[f]).join(","),o=kn("rc",a),l=kn("source",a),u=`${o[a-1]} < ${this.outputShape[a-1]}`,p=a===1?"source":`vec2(${l.slice(-2).join()})`,d=[`${r} rc = outputLoc;`,`${o[a-1]} += 1;
+    `}},bre=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((f,m)=>f[0]+e[m]+f[1]);let r=e.length,s=mt(r),a=t.map(f=>f[0]).join(","),o=t.map((f,m)=>f[0]+e[m]).join(","),i=Tn("rc",r),c=Tn("source",r),u=`${i[r-1]} < ${this.outputShape[r-1]}`,l=r===1?"source":`vec2(${c.slice(-2).join()})`,p=[`${s} rc = outputLoc;`,`${i[r-1]} += 1;
        if(${u}) {
-      `,a===1?"":`}
+      `,r===1?"":`}
        rc = outputLoc;
-       ${o[a-2]} += 1;
-       if(${o[a-2]} < ${this.outputShape[a-2]}) {`,a===1?"":`  ${o[a-1]} += 1;
-         if(${u}) {`],c=a===1?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",h="";for(let m=0,f=a===1?2:4;m<f;m++)h+=`
-        ${d[m]}
-        if (${c}) {
-          result[${m}] = float(value);
+       ${i[r-2]} += 1;
+       if(${i[r-2]} < ${this.outputShape[r-2]}) {`,r===1?"":`  ${i[r-1]} += 1;
+         if(${u}) {`],d=r===1?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",h="";for(let f=0,m=r===1?2:4;f<m;f++)h+=`
+        ${p[f]}
+        if (${d}) {
+          result[${f}] = float(value);
         } else {
-          ${r} source = rc - start;
-          result[${m}] = getChannel(getX(${l.join()}), ${p});
+          ${s} source = rc - start;
+          result[${f}] = getChannel(getX(${c.join()}), ${l});
         }
-      `;h+=a===1?"} ":"}}",this.userCode=`
-      const ${r} start = ${r}(${s});
-      const ${r} end = ${r}(${i});
+      `;h+=r===1?"} ":"}}",this.userCode=`
+      const ${s} start = ${s}(${a});
+      const ${s} end = ${s}(${o});
 
       void main() {
-        ${r} outputLoc = getOutputCoords();
+        ${s} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
         ${h}
         setOutput(result);
       }
-    `}},PE=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{paddings:s,constantValue:i}=a;if(v.sizeFromShape(r.shape)===0){let u=s.map((p,d)=>p[0]+r.shape[d]+p[1]);return cd({backend:n,attrs:{shape:u,value:i,dtype:r.dtype}})}let o=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new mae(r.shape,s,i):new hae(r.shape,s,i),l=[[i]];return n.runWebGLProgram(o,[r],r.dtype,l)},fae={kernelName:ji,backendName:"webgl",kernelFunc:PE},gae=`
+    `}},LE=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{paddings:a,constantValue:o}=r;if(w.sizeFromShape(s.shape)===0){let u=a.map((l,p)=>l[0]+s.shape[p]+l[1]);return dp({backend:n,attrs:{shape:u,value:o,dtype:s.dtype}})}let i=q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new bre(s.shape,a,o):new gre(s.shape,a,o),c=[[o]];return n.runWebGLProgram(i,[s],s.dtype,c)},yre={kernelName:Ko,backendName:"webgl",kernelFunc:LE},vre=`
   if(a < 0.0 && floor(b) < b){
     return NAN;
   }
@@ -4019,7 +4019,7 @@ return a / b;`,Bne=`
   }
   return (round(mod(b, 2.0)) != 1) ?
       pow(abs(a), b) : sign(a) * pow(abs(a), b);
-`,yae=`
+`,xre=`
   // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.
   vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));
   vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);
@@ -4035,11 +4035,11 @@ return a / b;`,Bne=`
   bvec4 isNaN1 = lessThan(a, vec4(0.0));
   bvec4 isNaN2 = lessThan(floor(b), b);
   bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);
-  `+ld+`
+  `+cp+`
   return result;
-`,bae=cn({opSnippet:gae,packedOpSnippet:yae}),xae={kernelName:qi,backendName:"webgl",kernelFunc:bae};function vae(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,keepDims:i}=a,o=r.shape.length,l=[],u=v.parseAxisParam(s,r.shape),p=u,d=N.getAxesPermutation(p,o),c=r;d!=null&&(c=Sn({inputs:{x:r},backend:n,attrs:{perm:d}}),p=N.getInnerMostAxes(p.length,o),l.push(c)),N.assertAxesAreInnerMostDims("prod",p,o);let h;if(n.shouldExecuteOnCPU([c])){let m=n.texData.get(c.dataId).values,{outVals:f,outShape:g,outDtype:y}=I7(c.shape,c.dtype,m,p);h=n.makeTensorInfo(g,y,f)}else{let[m,f]=N.computeOutAndReduceShapes(c.shape,p),g=v.sizeFromShape(f),y=de({inputs:{x:c},backend:n,attrs:{shape:[-1,g]}}),b=Um(r.dtype),x=Io(y,b,"prod",n);h=de({inputs:{x},backend:n,attrs:{shape:m}}),l.push(y),l.push(x)}if(i){l.push(h);let m=N.expandShapeToKeepDim(h.shape,u);h=de({inputs:{x:h},backend:n,attrs:{shape:m}})}return l.forEach(m=>n.disposeIntermediateTensorInfo(m)),h}var wae={kernelName:Xi,backendName:"webgl",kernelFunc:vae};function kae(e){let{inputs:t,backend:n,attrs:a}=e,{paramsNestedSplits:r,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=a,l=r.map(y=>n.readSync(y.dataId)),u=r.map(y=>y.shape),p=n.readSync(s.dataId),d=n.readSync(i.dataId),[c,h,m]=S7(l,u,p,s.shape,s.dtype,d,i.shape,o),f=c.map(y=>n.makeTensorInfo([y.length],"int32",y)),g=n.makeTensorInfo(m,s.dtype,h);return f.concat([g])}var Iae={kernelName:Dm,backendName:"webgl",kernelFunc:kae};function Sae(e){let{inputs:t,backend:n}=e,{starts:a,limits:r,deltas:s}=t,i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=T7(i,a.shape,a.dtype,o,r.shape,l,s.shape),d=n.makeTensorInfo([u.length],"int32",u),c=n.makeTensorInfo([p.length],a.dtype,p);return[d,c]}var Tae={kernelName:Rm,backendName:"webgl",kernelFunc:Sae};function Nae(e){let{inputs:t,backend:n,attrs:a}=e,{shape:r,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=a,u=n.readSync(r.dataId),p=n.readSync(s.dataId),d=n.readSync(i.dataId),c=o.map(g=>n.readSync(g.dataId)),h=o.map(g=>g.shape),[m,f]=N7(u,r.shape,p,s.shape,s.dtype,d,i.shape,c,h,l);return n.makeTensorInfo(m,s.dtype,f)}var Cae={kernelName:Mm,backendName:"webgl",kernelFunc:Nae},OE=e=>{let{backend:t,attrs:n}=e,{start:a,stop:r,step:s,dtype:i}=n,o=C7(a,r,s,i);return t.makeTensorInfo([o.length],i,o)},_ae={kernelName:xc,backendName:"webgl",kernelFunc:OE},Eae="return 1.0 / x;",Aae=Ye({opSnippet:Eae}),$ae={kernelName:xu,backendName:"webgl",kernelFunc:Aae},Fae=Ma+`
+`,wre=dn({opSnippet:vre,packedOpSnippet:xre}),Ire={kernelName:Xo,backendName:"webgl",kernelFunc:wre};function kre(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,keepDims:o}=r,i=s.shape.length,c=[],u=w.parseAxisParam(a,s.shape),l=u,p=N.getAxesPermutation(l,i),d=s;p!=null&&(d=Nn({inputs:{x:s},backend:n,attrs:{perm:p}}),l=N.getInnerMostAxes(l.length,i),c.push(d)),N.assertAxesAreInnerMostDims("prod",l,i);let h;if(n.shouldExecuteOnCPU([d])){let f=n.texData.get(d.dataId).values,{outVals:m,outShape:g,outDtype:b}=C9(d.shape,d.dtype,f,l);h=n.makeTensorInfo(g,b,m)}else{let[f,m]=N.computeOutAndReduceShapes(d.shape,l),g=w.sizeFromShape(m),b=he({inputs:{x:d},backend:n,attrs:{shape:[-1,g]}}),y=Hf(s.dtype),v=Si(b,y,"prod",n);h=he({inputs:{x:v},backend:n,attrs:{shape:f}}),c.push(b),c.push(v)}if(o){c.push(h);let f=N.expandShapeToKeepDim(h.shape,u);h=he({inputs:{x:h},backend:n,attrs:{shape:f}})}return c.forEach(f=>n.disposeIntermediateTensorInfo(f)),h}var Sre={kernelName:Zo,backendName:"webgl",kernelFunc:kre};function Tre(e){let{inputs:t,backend:n,attrs:r}=e,{paramsNestedSplits:s,paramsDenseValues:a,indices:o}=t,{outputRaggedRank:i}=r,c=s.map(b=>n.readSync(b.dataId)),u=s.map(b=>b.shape),l=n.readSync(a.dataId),p=n.readSync(o.dataId),[d,h,f]=N9(c,u,l,a.shape,a.dtype,p,o.shape,i),m=d.map(b=>n.makeTensorInfo([b.length],"int32",b)),g=n.makeTensorInfo(f,a.dtype,h);return m.concat([g])}var Cre={kernelName:Pf,backendName:"webgl",kernelFunc:Tre};function Nre(e){let{inputs:t,backend:n}=e,{starts:r,limits:s,deltas:a}=t,o=n.readSync(r.dataId),i=n.readSync(s.dataId),c=n.readSync(a.dataId),[u,l]=_9(o,r.shape,r.dtype,i,s.shape,c,a.shape),p=n.makeTensorInfo([u.length],"int32",u),d=n.makeTensorInfo([l.length],r.dtype,l);return[p,d]}var _re={kernelName:Of,backendName:"webgl",kernelFunc:Nre};function Ere(e){let{inputs:t,backend:n,attrs:r}=e,{shape:s,values:a,defaultValue:o,rowPartitionTensors:i}=t,{rowPartitionTypes:c}=r,u=n.readSync(s.dataId),l=n.readSync(a.dataId),p=n.readSync(o.dataId),d=i.map(g=>n.readSync(g.dataId)),h=i.map(g=>g.shape),[f,m]=E9(u,s.shape,l,a.shape,a.dtype,p,o.shape,d,h,c);return n.makeTensorInfo(f,a.dtype,m)}var Are={kernelName:Mf,backendName:"webgl",kernelFunc:Ere},zE=e=>{let{backend:t,attrs:n}=e,{start:r,stop:s,step:a,dtype:o}=n,i=A9(r,s,a,o);return t.makeTensorInfo([i.length],o,i)},$re={kernelName:vd,backendName:"webgl",kernelFunc:zE},Dre="return 1.0 / x;",Fre=Ye({opSnippet:Dre}),Rre={kernelName:vu,backendName:"webgl",kernelFunc:Fre},Pre=Or+`
   return (x < 0.0) ? 0.0 : x;
-`,Dae=`
+`,Ore=`
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4049,9 +4049,9 @@ return a / b;`,Bne=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Rae=Ye({opSnippet:Fae,packedOpSnippet:Dae}),Mae={kernelName:Yi,backendName:"webgl",kernelFunc:Rae},Pae=Ma+`
+`,Mre=Ye({opSnippet:Pre,packedOpSnippet:Ore}),Lre={kernelName:Jo,backendName:"webgl",kernelFunc:Mre},zre=Or+`
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,Oae=`
+`,Bre=`
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4061,11 +4061,11 @@ return a / b;`,Bne=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Lae=Ye({opSnippet:Pae,packedOpSnippet:Oae}),zae={kernelName:Qi,backendName:"webgl",kernelFunc:Lae},Wae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d;r?d="(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":d="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+`,Wre=Ye({opSnippet:zre,packedOpSnippet:Bre}),Vre={kernelName:ti,backendName:"webgl",kernelFunc:Wre},Ure=class{constructor(e,t,n,r,s){this.variableNames=["A"],this.outputShape=[];let[a,o,i,c]=e;this.outputShape=[a,t,n,c];let u=[r&&t>1?o-1:o,r&&n>1?i-1:i],l=[r&&t>1?t-1:t,r&&n>1?n-1:n],p;s?p="(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":p="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
+          ${u[0]/l[0]},
+          ${u[1]/l[1]});
+      const vec2 inputShapeRC = vec2(${o}.0, ${i}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4074,7 +4074,7 @@ return a / b;`,Bne=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${d};
+        vec2 sourceFracIndexRC = ${p};
 
         // Compute the four integer indices.
         ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));
@@ -4094,13 +4094,13 @@ return a / b;`,Bne=`
 
         setOutput(newValue);
       }
-    `}},Bae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d;r?d="(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":d="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `}},Gre=class{constructor(e,t,n,r,s){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[a,o,i,c]=e;this.outputShape=[a,t,n,c];let u=[r&&t>1?o-1:o,r&&n>1?i-1:i],l=[r&&t>1?t-1:t,r&&n>1?n-1:n],p;s?p="(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":p="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]},
-          ${u[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
-                                     ${o}.0);
+          ${u[0]/l[0]},
+          ${u[1]/l[1]},
+          ${u[1]/l[1]});
+      const vec3 inputShapeRC = vec3(${o}.0, ${i}.0,
+                                     ${i}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4114,7 +4114,7 @@ return a / b;`,Bne=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${d};
+        vec3 sourceFracIndexRC = ${p};
 
         // Compute the four integer indices.
         ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));
@@ -4122,7 +4122,7 @@ return a / b;`,Bne=`
           min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
+        bool hasNextCol = d < ${c-1};
         bool hasNextRow = coords.z < ${n-1};
 
         // In parallel, construct four corners for all four components in
@@ -4171,7 +4171,7 @@ return a / b;`,Bne=`
 
         setOutput(newValue);
       }
-    `}};function Vae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,p=H().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Bae(r.shape,l,u,s,i):new Wae(r.shape,l,u,s,i);return n.runWebGLProgram(p,[r],"float32")}var Uae={kernelName:Ji,backendName:"webgl",kernelFunc:Vae},Gae=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,a,r]=t,[,s,i]=e,o=[n&&s>1?a-1:a,n&&i>1?r-1:r],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],p=o[1]/l[1],d=1/u,c=1/p,h=Math.ceil(d)*2+2,m=Math.ceil(c)*2+2;this.userCode=`
+    `}};function Hre(e){let{inputs:t,backend:n,attrs:r}=e,{images:s}=t,{alignCorners:a,halfPixelCenters:o,size:i}=r,[c,u]=i,l=q().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Gre(s.shape,c,u,a,o):new Ure(s.shape,c,u,a,o);return n.runWebGLProgram(l,[s],"float32")}var qre={kernelName:ei,backendName:"webgl",kernelFunc:Hre},jre=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,r,s]=t,[,a,o]=e,i=[n&&a>1?r-1:r,n&&o>1?s-1:s],c=[n&&a>1?a-1:a,n&&o>1?o-1:o],u=i[0]/c[0],l=i[1]/c[1],p=1/u,d=1/l,h=Math.ceil(p)*2+2,f=Math.ceil(d)*2+2;this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4182,13 +4182,13 @@ return a / b;`,Bne=`
         float accumulator = 0.0;
 
         const float heightScale = float(${u});
-        const float widthScale = float(${p});
+        const float widthScale = float(${l});
 
-        const float invHeightScale = float(${d});
-        const float invWidthScale = float(${c});
+        const float invHeightScale = float(${p});
+        const float invWidthScale = float(${d});
 
         const int winHeight = int(${h});
-        const int winWidth = int(${m});
+        const int winWidth = int(${f});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4202,7 +4202,7 @@ return a / b;`,Bne=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${s}) {
+          if (dyR < 0 || dyR >= ${a}) {
             continue;
           }
 
@@ -4210,19 +4210,19 @@ return a / b;`,Bne=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${i}) {
+            if (dyC < 0 || dyC >= ${o}) {
               continue;
             }
 
             float dxR = float(dyR) * heightScale;
             int topDxRIndex = int(floor(dxR));
-            int bottomDxRIndex = int(min(ceil(dxR), ${a-1}.0));
+            int bottomDxRIndex = int(min(ceil(dxR), ${r-1}.0));
             float dxRLerp = dxR - float(topDxRIndex);
             float inverseDxRLerp = 1.0 - dxRLerp;
 
             float dxC = float(dyC) * widthScale;
             int leftDxCIndex = int(floor(dxC));
-            int rightDxCIndex = int(min(ceil(dxC), ${r-1}.0));
+            int rightDxCIndex = int(min(ceil(dxC), ${s-1}.0));
             float dxCLerp = dxC - float(leftDxCIndex);
             float inverseDxCLerp = 1.0 - dxCLerp;
 
@@ -4252,11 +4252,11 @@ return a / b;`,Bne=`
 
         setOutput(accumulator);
       }
-    `}};function Hae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a,o=new Gae(s.shape,r.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}var jae={kernelName:Lm,backendName:"webgl",kernelFunc:Hae},qae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d=a?"0.5":"0.0",c;r?c="max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":c="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `}};function Kre(e){let{inputs:t,backend:n,attrs:r}=e,{images:s,dy:a}=t,{alignCorners:o}=r,i=new jre(a.shape,s.shape,o);return n.runWebGLProgram(i,[a],a.dtype)}var Xre={kernelName:Bf,backendName:"webgl",kernelFunc:Kre},Yre=class{constructor(e,t,n,r,s){this.variableNames=["A"],this.outputShape=[];let[a,o,i,c]=e;this.outputShape=[a,t,n,c];let u=[r&&t>1?o-1:o,r&&n>1?i-1:i],l=[r&&t>1?t-1:t,r&&n>1?n-1:n],p=r?"0.5":"0.0",d;s?d="max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":d="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
+          ${u[0]/l[0]},
+          ${u[1]/l[1]});
+      const vec2 inputShapeRC = vec2(${o}.0, ${i}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4265,22 +4265,22 @@ return a / b;`,Bne=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${c};
+        vec2 sourceFracIndexRC = ${d};
 
         // Compute the coordinators of nearest neighbor point.
         ivec2 sourceNearestRC = ivec2(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${d})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${p})));
         float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);
 
         setOutput(newValue);
       }
-    `}},Kae=class{constructor(e,t,n,a,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u=[a&&t>1?i-1:i,a&&n>1?o-1:o],p=[a&&t>1?t-1:t,a&&n>1?n-1:n],d=a?"0.5":"0.0",c;r?c="max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":c="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `}},Zre=class{constructor(e,t,n,r,s){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[a,o,i,c]=e;this.outputShape=[a,t,n,c];let u=[r&&t>1?o-1:o,r&&n>1?i-1:i],l=[r&&t>1?t-1:t,r&&n>1?n-1:n],p=r?"0.5":"0.0",d;s?d="max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":d="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${u[0]/p[0]},
-          ${u[1]/p[1]},
-          ${u[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
-                                     ${o}.0);
+          ${u[0]/l[0]},
+          ${u[1]/l[1]},
+          ${u[1]/l[1]});
+      const vec3 inputShapeRC = vec3(${o}.0, ${i}.0,
+                                     ${i}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4294,14 +4294,14 @@ return a / b;`,Bne=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${c};
+        vec3 sourceFracIndexRC = ${d};
 
         // Compute the coordinators of nearest neighbor point.
         ivec3 sourceNearestRC = ivec3(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${d})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${p})));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
+        bool hasNextCol = d < ${c-1};
         bool hasNextRow = coords.z < ${n-1};
 
         vec4 newValue = vec4(
@@ -4315,7 +4315,7 @@ return a / b;`,Bne=`
 
         setOutput(newValue);
       }
-    `}};function Xae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r}=t,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,p=H().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Kae(r.shape,l,u,s,i):new qae(r.shape,l,u,s,i);return n.runWebGLProgram(p,[r],r.dtype)}var Yae={kernelName:Zi,backendName:"webgl",kernelFunc:Xae},Zae=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,a,r]=t,[,s,i]=e,o=[n&&s>1?a-1:a,n&&i>1?r-1:r],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],p=o[1]/l[1],d=1/u,c=1/p,h=Math.ceil(d)*2+2,m=Math.ceil(c)*2+2;this.userCode=`
+    `}};function Jre(e){let{inputs:t,backend:n,attrs:r}=e,{images:s}=t,{alignCorners:a,halfPixelCenters:o,size:i}=r,[c,u]=i,l=q().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new Zre(s.shape,c,u,a,o):new Yre(s.shape,c,u,a,o);return n.runWebGLProgram(l,[s],s.dtype)}var Qre={kernelName:Qo,backendName:"webgl",kernelFunc:Jre},ese=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,r,s]=t,[,a,o]=e,i=[n&&a>1?r-1:r,n&&o>1?s-1:s],c=[n&&a>1?a-1:a,n&&o>1?o-1:o],u=i[0]/c[0],l=i[1]/c[1],p=1/u,d=1/l,h=Math.ceil(p)*2+2,f=Math.ceil(d)*2+2;this.userCode=`
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4326,13 +4326,13 @@ return a / b;`,Bne=`
         float accumulator = 0.0;
 
         const float heightScale = float(${u});
-        const float widthScale = float(${p});
+        const float widthScale = float(${l});
 
-        const float invHeightScale = float(${d});
-        const float invWidthScale = float(${c});
+        const float invHeightScale = float(${p});
+        const float invWidthScale = float(${d});
 
         const int winHeight = int(${h});
-        const int winWidth = int(${m});
+        const int winWidth = int(${f});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4346,7 +4346,7 @@ return a / b;`,Bne=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${s}) {
+          if (dyR < 0 || dyR >= ${a}) {
             continue;
           }
 
@@ -4354,25 +4354,25 @@ return a / b;`,Bne=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${i}) {
+            if (dyC < 0 || dyC >= ${o}) {
               continue;
             }
 
             float sourceFracRow =
-              float(${o[0]}) *
-                (float(dyR) / float(${l[0]}));
+              float(${i[0]}) *
+                (float(dyR) / float(${c[0]}));
 
             float sourceFracCol =
-                float(${o[1]}) *
-                  (float(dyC) / float(${l[1]}));
+                float(${i[1]}) *
+                  (float(dyC) / float(${c[1]}));
 
             int sourceNearestRow = int(min(
-                float(int(${a}) - 1),
+                float(int(${r}) - 1),
                 ${n} ? float(round(sourceFracRow)) :
                                   float(floor(sourceFracRow))));
 
             int sourceNearestCol = int(min(
-                float(int(${r}) - 1),
+                float(int(${s}) - 1),
                 ${n} ? float(round(sourceFracCol)) :
                                   float(floor(sourceFracCol))));
 
@@ -4385,23 +4385,23 @@ return a / b;`,Bne=`
 
         setOutput(accumulator);
       }
-    `}};function Jae(e){let{inputs:t,backend:n,attrs:a}=e,{images:r,dy:s}=t,{alignCorners:i}=a,o=new Zae(s.shape,r.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}var Qae={kernelName:Om,backendName:"webgl",kernelFunc:Jae},ere=class{constructor(e,t){this.variableNames=["x"];let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=e,n===1){this.userCode=`
+    `}};function tse(e){let{inputs:t,backend:n,attrs:r}=e,{images:s,dy:a}=t,{alignCorners:o}=r,i=new ese(a.shape,s.shape,o);return n.runWebGLProgram(i,[a],a.dtype)}var nse={kernelName:zf,backendName:"webgl",kernelFunc:tse},rse=class{constructor(e,t){this.variableNames=["x"];let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=e,n===1){this.userCode=`
         void main() {
           int coord = getOutputCoords();
           setOutput(getX(${e[0]} - coord - 1));
         }
-      `;return}let a=i=>t.indexOf(i)!==-1&&e[i]!==1?`${e[i]} - coords[${i}] - 1`:`coords[${i}]`,r=e.map((i,o)=>a(o)).join(","),s=gt(n);this.userCode=`
+      `;return}let r=o=>t.indexOf(o)!==-1&&e[o]!==1?`${e[o]} - coords[${o}] - 1`:`coords[${o}]`,s=e.map((o,i)=>r(i)).join(","),a=mt(n);this.userCode=`
       void main() {
-        ${s} coords = getOutputCoords();
-        setOutput(getX(${r}));
+        ${a} coords = getOutputCoords();
+        setOutput(getX(${s}));
       }
-    `}},tre=class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=e;let a=kn("rc",n),r=`${a[n-1]} + 1 < ${this.outputShape[n-1]}`,s=`${a[n-2]} + 1 < ${this.outputShape[n-2]}`,i=gt(n);n===1?this.userCode=`
+    `}},sse=class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=e;let r=Tn("rc",n),s=`${r[n-1]} + 1 < ${this.outputShape[n-1]}`,a=`${r[n-2]} + 1 < ${this.outputShape[n-2]}`,o=mt(n);n===1?this.userCode=`
         void main(){
           int rc = getOutputCoords();
           vec4 result = vec4(0.);
           result.r = getChannel(getX(${e[0]} - rc - 1),
             ${e[0]} - rc - 1);
-          if(${r}){
+          if(${s}){
               result.g = getChannel(getX(${e[0]} - (rc  + 1) - 1),
                 ${e[0]} - (rc  + 1) - 1);
           }
@@ -4409,21 +4409,21 @@ return a / b;`,Bne=`
         }
       `:this.userCode=`
         void main() {
-          ${i} rc = getOutputCoords();
+          ${o} rc = getOutputCoords();
           vec4 result = vec4(0.);
-          result.r = ${o(a.slice())};
-          if(${r}){
-            result.g = ${l(a.slice())};
+          result.r = ${i(r.slice())};
+          if(${s}){
+            result.g = ${c(r.slice())};
           }
-          if(${s}) {
-            result.b = ${u(a.slice())};
-            if(${r}) {
-              result.a = ${p(a.slice())};
+          if(${a}) {
+            result.b = ${u(r.slice())};
+            if(${s}) {
+              result.a = ${l(r.slice())};
             }
           }
           setOutput(result);
         }
-    `;function o(h){return d(h)}function l(h){return h[n-1]="("+h[n-1]+" + 1)",d(h)}function u(h){return h[n-2]="("+h[n-2]+" + 1)",d(h)}function p(h){return h[n-1]="("+h[n-1]+" + 1)",h[n-2]="("+h[n-2]+" + 1)",d(h)}function d(h){let m=e.map((y,b)=>c(b,h)),f=m.join(","),g=m.slice(-2).join(",");return`getChannel(getX(${f}), vec2(${g}))`}function c(h,m){return t.indexOf(h)!==-1&&e[h]!==1?`${e[h]} - ${m[h]} - 1`:`${m[h]}`}}};function nre(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a,i=r.shape.length,o=v.parseAxisParam(s,r.shape);if(i===0)return aa({inputs:{x:r},backend:n});let l=H().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new tre(r.shape,o):new ere(r.shape,o);return n.runWebGLProgram(l,[r],r.dtype)}var are={kernelName:eo,backendName:"webgl",kernelFunc:nre},rre=class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let n=e[1],a=e[2];this.outputShape=e;let r="";typeof t=="number"?r=`float outputValue = ${t.toFixed(2)};`:r=`
+    `;function i(h){return p(h)}function c(h){return h[n-1]="("+h[n-1]+" + 1)",p(h)}function u(h){return h[n-2]="("+h[n-2]+" + 1)",p(h)}function l(h){return h[n-1]="("+h[n-1]+" + 1)",h[n-2]="("+h[n-2]+" + 1)",p(h)}function p(h){let f=e.map((b,y)=>d(y,h)),m=f.join(","),g=f.slice(-2).join(",");return`getChannel(getX(${m}), vec2(${g}))`}function d(h,f){return t.indexOf(h)!==-1&&e[h]!==1?`${e[h]} - ${f[h]} - 1`:`${f[h]}`}}};function ase(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{dims:a}=r,o=s.shape.length,i=w.parseAxisParam(a,s.shape);if(o===0)return sr({inputs:{x:s},backend:n});let c=q().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new sse(s.shape,i):new rse(s.shape,i);return n.runWebGLProgram(c,[s],s.dtype)}var ose={kernelName:ni,backendName:"webgl",kernelFunc:ase},ise=class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let n=e[1],r=e[2];this.outputShape=e;let s="";typeof t=="number"?s=`float outputValue = ${t.toFixed(2)};`:s=`
         vec3 fill = vec3(${t.join(",")});
         float outputValue = fill[coords[3]];`,this.userCode=`
         void main() {
@@ -4436,13 +4436,13 @@ return a / b;`,Bne=`
             (float(y) - params[1]) * params[3];
           int coordX = int(round(coordXFloat + params[0]));
           int coordY = int(round(coordYFloat + params[1]));
-          ${r}
-          if(coordX >= 0 && coordX < ${a} && coordY >= 0 && coordY < ${n}) {
+          ${s}
+          if(coordX >= 0 && coordX < ${r} && coordY >= 0 && coordY < ${n}) {
             outputValue = getImage(coords[0], coordY, coordX, coords[3]);
           }
           setOutput(outputValue);
         }
-    `}},sre={kernelName:Pu,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:a}=e,{radians:r,fillValue:s,center:i}=t,o=n,l=new rre(a.shape,s),[u,p]=N.getImageCenter(i,a.shape[1],a.shape[2]),d=[[u,p,Math.sin(r),Math.cos(r)]];return o.runWebGLProgram(l,[a],a.dtype,d)}},ire=`
+    `}},cse={kernelName:Ou,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,{radians:s,fillValue:a,center:o}=t,i=n,c=new ise(r.shape,a),[u,l]=N.getImageCenter(o,r.shape[1],r.shape[2]),p=[[u,l,Math.sin(s),Math.cos(s)]];return i.runWebGLProgram(c,[r],r.dtype,p)}},use=`
   // OpenGL ES does not support round function.
   // The algorithm is based on banker's rounding.
   float base = floor(x);
@@ -4457,34 +4457,34 @@ return a / b;`,Bne=`
       return base + 1.0;
     }
   }
-`,ore=Ye({opSnippet:ire}),lre={kernelName:to,backendName:"webgl",kernelFunc:ore},ure="return inversesqrt(x);",pre=Ye({opSnippet:ure,cpuKernelImpl:_7}),cre={kernelName:no,backendName:"webgl",kernelFunc:pre},LE=class{constructor(e,t,n,a,r,s,i=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=s;let o=gt(r.length),l=gt(s.length),u="";n===1?u="i":n===2&&(u="i, j");let p=`getIndices(${u})`,d="";a===1?d="i":a===2&&(d="i, coords[1]");let c=`getUpdates(${d})`,h=t>1?"strides[j]":"strides";this.userCode=`
-        ${o} strides = ${o}(${r});
+`,lse=Ye({opSnippet:use}),dse={kernelName:ri,backendName:"webgl",kernelFunc:lse},pse="return inversesqrt(x);",hse=Ye({opSnippet:pse,cpuKernelImpl:$9}),fse={kernelName:si,backendName:"webgl",kernelFunc:hse},BE=class{constructor(e,t,n,r,s,a,o=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=a;let i=mt(s.length),c=mt(a.length),u="";n===1?u="i":n===2&&(u="i, j");let l=`getIndices(${u})`,p="";r===1?p="i":r===2&&(p="i, coords[1]");let d=`getUpdates(${p})`,h=t>1?"strides[j]":"strides";this.userCode=`
+        ${i} strides = ${i}(${s});
 
         void main() {
-          ${l} coords = getOutputCoords();
+          ${c} coords = getOutputCoords();
           float sum = 0.0;
           bool found = false;
           for (int i = 0; i < ${e}; i++) {
             int flattenedIndex = 0;
             for (int j = 0; j < ${t}; j++) {
-              int index = round(${p});
+              int index = round(${l});
               flattenedIndex += index * ${h};
             }
             if (flattenedIndex == coords[0]) {
-              sum += ${c};
+              sum += ${d};
               found = true;
             }
           }
           setOutput(mix(getDefaultValue(), sum, float(found)));
         }
-      `}};function dre(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r,updates:s}=t,{shape:i}=a,{sliceRank:o,numUpdates:l,sliceSize:u,strides:p,outputSize:d}=N.calculateShapes(s,r,i),c=[d/u,u];if(d===0)return n.makeTensorInfo(i,r.dtype);let h=de({inputs:{x:r},backend:n,attrs:{shape:[l,o]}}),m=de({inputs:{x:s},backend:n,attrs:{shape:[l,u]}}),f=n.makeTensorInfo([],"float32",new Float32Array([0])),g=new LE(l,o,h.shape.length,m.shape.length,p,c),y=n.runWebGLProgram(g,[m,h,f],m.dtype),b=de({inputs:{x:y},backend:n,attrs:{shape:i}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(f),b}var hre={kernelName:wu,backendName:"webgl",kernelFunc:dre},mre=class{constructor(e,t,n,a){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[e,n];let r="while (left < right) {",s=`for (int i = 0; i < ${Math.ceil(Math.log2(t+1))}; ++i) { if (left >= right) break;`,i=H().getNumber("WEBGL_VERSION")===2?r:s,o=a==="left"?"<":"<=";this.userCode=`
+      `}};function mse(e){let{inputs:t,backend:n,attrs:r}=e,{indices:s,updates:a}=t,{shape:o}=r,{sliceRank:i,numUpdates:c,sliceSize:u,strides:l,outputSize:p}=N.calculateShapes(a,s,o),d=[p/u,u];if(p===0)return n.makeTensorInfo(o,s.dtype);let h=he({inputs:{x:s},backend:n,attrs:{shape:[c,i]}}),f=he({inputs:{x:a},backend:n,attrs:{shape:[c,u]}}),m=n.makeTensorInfo([],"float32",new Float32Array([0])),g=new BE(c,i,h.shape.length,f.shape.length,l,d),b=n.runWebGLProgram(g,[f,h,m],f.dtype),y=he({inputs:{x:b},backend:n,attrs:{shape:o}});return n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(b),n.disposeIntermediateTensorInfo(m),y}var gse={kernelName:wu,backendName:"webgl",kernelFunc:mse},bse=class{constructor(e,t,n,r){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[e,n];let s="while (left < right) {",a=`for (int i = 0; i < ${Math.ceil(Math.log2(t+1))}; ++i) { if (left >= right) break;`,o=q().getNumber("WEBGL_VERSION")===2?s:a,i=r==="left"?"<":"<=";this.userCode=`
        int findBound(int batch, float value) {
          int left = 0;
          int right = numInputs;
          int mid;
-         ${i}
+         ${o}
            mid = (left + right) / 2;
-           if (getSortedSequence(batch, mid) ${o} value) {
+           if (getSortedSequence(batch, mid) ${i} value) {
              left = mid + 1;
            } else {
              right = mid;
@@ -4502,25 +4502,25 @@ return a / b;`,Bne=`
 
          setOutput(float(findBound(batch, value)));
        }
-     `}};function fre(e){let{inputs:t,backend:n,attrs:a}=e,{sortedSequence:r,values:s}=t,{side:i}=a,o=new mre(r.shape[0],r.shape[1],s.shape[1],i),l=[[r.shape[1]]];return n.runWebGLProgram(o,[r,s],"int32",l)}var gre={kernelName:zm,backendName:"webgl",kernelFunc:fre},yre=class{constructor(e,t,n){this.variableNames=["c","a","b"],this.outputShape=t;let a,r;if(n>4)throw Error(`Where for rank ${n} is not yet supported`);if(n===1)r="resRC",a="resRC";else{let i=["resRC.x","resRC.y","resRC.z","resRC.w"],o=[],l=[];for(let u=0;u<t.length;u++)l.push(`${i[u]}`),u<e&&o.push(`${i[u]}`);a=o.join(),r=l.join()}let s=gt(n);this.userCode=`
+     `}};function yse(e){let{inputs:t,backend:n,attrs:r}=e,{sortedSequence:s,values:a}=t,{side:o}=r,i=new bse(s.shape[0],s.shape[1],a.shape[1],o),c=[[s.shape[1]]];return n.runWebGLProgram(i,[s,a],"int32",c)}var vse={kernelName:Wf,backendName:"webgl",kernelFunc:yse},xse=class{constructor(e,t,n){this.variableNames=["c","a","b"],this.outputShape=t;let r,s;if(n>4)throw Error(`Where for rank ${n} is not yet supported`);if(n===1)s="resRC",r="resRC";else{let o=["resRC.x","resRC.y","resRC.z","resRC.w"],i=[],c=[];for(let u=0;u<t.length;u++)c.push(`${o[u]}`),u<e&&i.push(`${o[u]}`);r=i.join(),s=c.join()}let a=mt(n);this.userCode=`
       void main() {
-        ${s} resRC = getOutputCoords();
-        float cVal = getC(${a});
+        ${a} resRC = getOutputCoords();
+        float cVal = getC(${r});
         if (cVal >= 1.0) {
-          setOutput(getA(${r}));
+          setOutput(getA(${s}));
         } else {
-          setOutput(getB(${r}));
+          setOutput(getB(${s}));
         }
       }
-    `}};function bre(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t,i=new yre(a.shape.length,r.shape,r.shape.length);return n.runWebGLProgram(i,[a,r,s],fa(r.dtype,s.dtype))}var xre={kernelName:ku,backendName:"webgl",kernelFunc:bre},vre=`
+    `}};function wse(e){let{inputs:t,backend:n}=e,{condition:r,t:s,e:a}=t,o=new xse(r.shape.length,s.shape,s.shape.length);return n.runWebGLProgram(o,[r,s,a],hr(s.dtype,a.dtype))}var Ise={kernelName:Iu,backendName:"webgl",kernelFunc:wse},kse=`
   // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.
   // see: https://arxiv.org/abs/1706.02515
   float scaleAlpha = ${N.SELU_SCALEALPHA};
   float scale = ${N.SELU_SCALE};
   return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);
-`,wre=Ye({opSnippet:vre}),kre={kernelName:Iu,backendName:"webgl",kernelFunc:wre},Ire=tp+`
+`,Sse=Ye({opSnippet:kse}),Tse={kernelName:ku,backendName:"webgl",kernelFunc:Sse},Cse=tl+`
   return 1.0 / (1.0 + exp(-1.0 * x));
-`,Sre=`
+`,Nse=`
   vec4 result = 1.0 / (1.0 + exp(-1.0 * x));
   bvec4 isNaN = isnan(x);
 
@@ -4530,15 +4530,15 @@ return a / b;`,Bne=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Tre=Ye({opSnippet:Ire,packedOpSnippet:Sre,cpuKernelImpl:A7}),Nre={kernelName:ro,backendName:"webgl",kernelFunc:Tre},Cre=`
+`,_se=Ye({opSnippet:Cse,packedOpSnippet:Nse,cpuKernelImpl:F9}),Ese={kernelName:oi,backendName:"webgl",kernelFunc:_se},Ase=`
   if (isnan(x)) { return 0.0; }
   return sign(x);
-`,_re=Ye({opSnippet:Cre}),Ere={kernelName:Nu,backendName:"webgl",kernelFunc:_re},Are=tp+`
+`,$se=Ye({opSnippet:Ase}),Dse={kernelName:Cu,backendName:"webgl",kernelFunc:$se},Fse=tl+`
   return sin(x);
-`,$re=Ye({opSnippet:Are}),Fre={kernelName:ao,backendName:"webgl",kernelFunc:$re},Dre=`
+`,Rse=Ye({opSnippet:Fse}),Pse={kernelName:ai,backendName:"webgl",kernelFunc:Rse},Ose=`
   float e2x = exp(x);
   return (e2x - 1.0 / e2x) / 2.0;
-`,Rre=Ye({opSnippet:Dre}),Mre={kernelName:Tu,backendName:"webgl",kernelFunc:Rre},Pre=`
+`,Mse=Ye({opSnippet:Ose}),Lse={kernelName:Tu,backendName:"webgl",kernelFunc:Mse},zse=`
   float epsilon = 1.1920928955078125e-7;
   float threshold = log(epsilon) + 2.0;
 
@@ -4558,33 +4558,33 @@ return a / b;`,Bne=`
     result = log(exp_x + 1.0);
   }
   return result;
-`,Ore=Ye({opSnippet:Pre}),Lre={kernelName:Cu,backendName:"webgl",kernelFunc:Ore},zre=e=>{let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a;v.assert(r.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((y,b)=>y*b),l=[[0,0]];l.push(...i);for(let y=1+s.length;y<r.shape.length;++y)l.push([0,0]);let u=[],p=PE({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),d=N.getReshaped(p.shape,s,o,!1),c=N.getPermuted(d.length,s.length,!1),h=N.getReshapedPermuted(p.shape,s,o,!1),m=de({inputs:{x:p},backend:n,attrs:{shape:d}}),f=Sn({inputs:{x:m},backend:n,attrs:{perm:c}}),g=de({inputs:{x:f},backend:n,attrs:{shape:h}});return u.push(p),u.push(m),u.push(f),u.forEach(y=>n.disposeIntermediateTensorInfo(y)),g},Wre={kernelName:_u,backendName:"webgl",kernelFunc:zre};function Bre(e){let{inputs:t,backend:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-         ${s.shape}`);if(a.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-         ${a.shape}`);if(r.shape.length!==1)throw new Error(`Values must be a vector, saw:
-         ${r.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let o=n.readSync(a.dataId),l=n.readSync(r.dataId),u=n.readSync(s.dataId),p=n.readSync(i.dataId)[0],[d,c,h,m,f]=F7(o,a.shape,a.dtype,l,r.dtype,u,p);return[n.makeTensorInfo(c,a.dtype,d),n.makeTensorInfo([c[0]],r.dtype,h),n.makeTensorInfo([m.length],"bool",new Uint8Array(m.map(g=>Number(g)))),n.makeTensorInfo([f.length],a.dtype,new Int32Array(f))]}var Vre={kernelName:vc,backendName:"webgl",kernelFunc:Bre};function Ure(e){let{inputs:t,backend:n}=e,{inputIndices:a,inputShape:r,newShape:s}=t;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.readSync(r.dataId)),o=n.readSync(a.dataId),l=Array.from(n.readSync(s.dataId)),[u,p,d]=D7(o,a.shape,a.dtype,i,l);return[n.makeTensorInfo(p,a.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}var Gre={kernelName:Au,backendName:"webgl",kernelFunc:Ure};function Hre(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-              ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-              ${s.shape}`);let i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=iE(i,a.shape,a.dtype,o,l,!0);return n.makeTensorInfo(p,a.dtype,u)}var jre={kernelName:wc,backendName:"webgl",kernelFunc:Hre};function qre(e){let{inputs:t,backend:n}=e,{data:a,indices:r,segmentIds:s}=t;if(a.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(r.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-             ${r.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-             ${s.shape}`);let i=n.readSync(a.dataId),o=n.readSync(r.dataId),l=n.readSync(s.dataId),[u,p]=iE(i,a.shape,a.dtype,o,l);return n.makeTensorInfo(p,a.dtype,u)}var Kre={kernelName:kc,backendName:"webgl",kernelFunc:qre};function Xre(e){let{inputs:t,backend:n,attrs:a}=e,{sparseIndices:r,sparseValues:s,defaultValue:i}=t,{outputShape:o}=a,{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=N.calculateShapes(s,r,o),h=!1;if(s.dtype==="string"){let y=n.bufferSync(r),b=n.bufferSync(s),x=v.decodeString(n.readSync(i.dataId)[0]),w=E7(y,b,o,c,p,u,l,d,x,h);return n.makeTensorInfo(o,w.dtype,w.values)}let m=new LE(u,l,r.shape.length,s.shape.length,d,[c,1],h),f=n.runWebGLProgram(m,[s,r,i],s.dtype),g=de({inputs:{x:f},backend:n,attrs:{shape:o}});return n.disposeIntermediateTensorInfo(f),g}var Yre={kernelName:Wm,backendName:"webgl",kernelFunc:Xre};function Zre(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=a,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=r.shape.length,p=new Array(u).fill(0),d=r.shape.slice();return l.map(c=>{let h=[...d];h[o]=c;let m=np({inputs:{x:r},backend:n,attrs:{begin:p,size:h}});return p[o]+=c,m})}var Jre={kernelName:Eu,backendName:"webgl",kernelFunc:Zre},DI="return sqrt(x);",Qre=Ye({opSnippet:DI,packedOpSnippet:DI,cpuKernelImpl:R7}),ese={kernelName:so,backendName:"webgl",kernelFunc:Qre},tse="return x * x;",nse=Ye({opSnippet:tse}),ase={kernelName:Ic,backendName:"webgl",kernelFunc:nse},RI="return (a - b) * (a - b);",rse=cn({opSnippet:RI,packedOpSnippet:RI}),sse={kernelName:lo,backendName:"webgl",kernelFunc:rse};function ise({inputs:e,attrs:t,backend:n}){let{x:a}=e,r=Ma+`
+`,Bse=Ye({opSnippet:zse}),Wse={kernelName:Nu,backendName:"webgl",kernelFunc:Bse},Vse=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,paddings:o}=r;w.assert(s.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let i=a.reduce((b,y)=>b*y),c=[[0,0]];c.push(...o);for(let b=1+a.length;b<s.shape.length;++b)c.push([0,0]);let u=[],l=LE({inputs:{x:s},backend:n,attrs:{paddings:c,constantValue:0}}),p=N.getReshaped(l.shape,a,i,!1),d=N.getPermuted(p.length,a.length,!1),h=N.getReshapedPermuted(l.shape,a,i,!1),f=he({inputs:{x:l},backend:n,attrs:{shape:p}}),m=Nn({inputs:{x:f},backend:n,attrs:{perm:d}}),g=he({inputs:{x:m},backend:n,attrs:{shape:h}});return u.push(l),u.push(f),u.push(m),u.forEach(b=>n.disposeIntermediateTensorInfo(b)),g},Use={kernelName:_u,backendName:"webgl",kernelFunc:Vse};function Gse(e){let{inputs:t,backend:n}=e,{indices:r,values:s,denseShape:a,defaultValue:o}=t;if(a.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
+         ${a.shape}`);if(r.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
+         ${r.shape}`);if(s.shape.length!==1)throw new Error(`Values must be a vector, saw:
+         ${s.shape}`);if(o.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
+        ${o.shape}`);let i=n.readSync(r.dataId),c=n.readSync(s.dataId),u=n.readSync(a.dataId),l=n.readSync(o.dataId)[0],[p,d,h,f,m]=P9(i,r.shape,r.dtype,c,s.dtype,u,l);return[n.makeTensorInfo(d,r.dtype,p),n.makeTensorInfo([d[0]],s.dtype,h),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(g=>Number(g)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}var Hse={kernelName:xd,backendName:"webgl",kernelFunc:Gse};function qse(e){let{inputs:t,backend:n}=e,{inputIndices:r,inputShape:s,newShape:a}=t;if(r.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${r.shape}`);if(s.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${s.shape}`);if(a.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${a.shape}`);let o=Array.from(n.readSync(s.dataId)),i=n.readSync(r.dataId),c=Array.from(n.readSync(a.dataId)),[u,l,p]=O9(i,r.shape,r.dtype,o,c);return[n.makeTensorInfo(l,r.dtype,u),n.makeTensorInfo([p.length],a.dtype,new Int32Array(p))]}var jse={kernelName:Au,backendName:"webgl",kernelFunc:qse};function Kse(e){let{inputs:t,backend:n}=e,{data:r,indices:s,segmentIds:a}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.shape.length!==1)throw new Error(`Indices should be a vector but received shape
+              ${s.shape}`);if(a.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
+              ${a.shape}`);let o=n.readSync(r.dataId),i=n.readSync(s.dataId),c=n.readSync(a.dataId),[u,l]=cE(o,r.shape,r.dtype,i,c,!0);return n.makeTensorInfo(l,r.dtype,u)}var Xse={kernelName:wd,backendName:"webgl",kernelFunc:Kse};function Yse(e){let{inputs:t,backend:n}=e,{data:r,indices:s,segmentIds:a}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(s.shape.length!==1)throw new Error(`Indices should be a vector but received shape
+             ${s.shape}`);if(a.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
+             ${a.shape}`);let o=n.readSync(r.dataId),i=n.readSync(s.dataId),c=n.readSync(a.dataId),[u,l]=cE(o,r.shape,r.dtype,i,c);return n.makeTensorInfo(l,r.dtype,u)}var Zse={kernelName:Id,backendName:"webgl",kernelFunc:Yse};function Jse(e){let{inputs:t,backend:n,attrs:r}=e,{sparseIndices:s,sparseValues:a,defaultValue:o}=t,{outputShape:i}=r,{sliceRank:c,numUpdates:u,sliceSize:l,strides:p,outputSize:d}=N.calculateShapes(a,s,i),h=!1;if(a.dtype==="string"){let b=n.bufferSync(s),y=n.bufferSync(a),v=w.decodeString(n.readSync(o.dataId)[0]),x=D9(b,y,i,d,l,u,c,p,v,h);return n.makeTensorInfo(i,x.dtype,x.values)}let f=new BE(u,c,s.shape.length,a.shape.length,p,[d,1],h),m=n.runWebGLProgram(f,[a,s,o],a.dtype),g=he({inputs:{x:m},backend:n,attrs:{shape:i}});return n.disposeIntermediateTensorInfo(m),g}var Qse={kernelName:Vf,backendName:"webgl",kernelFunc:Jse};function eae(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{numOrSizeSplits:a,axis:o}=r,i=w.parseAxisParam(o,s.shape)[0],c=N.prepareSplitSize(s,a,i),u=s.shape.length,l=new Array(u).fill(0),p=s.shape.slice();return c.map(d=>{let h=[...p];h[i]=d;let f=nl({inputs:{x:s},backend:n,attrs:{begin:l,size:h}});return l[i]+=d,f})}var tae={kernelName:Eu,backendName:"webgl",kernelFunc:eae},R1="return sqrt(x);",nae=Ye({opSnippet:R1,packedOpSnippet:R1,cpuKernelImpl:M9}),rae={kernelName:ii,backendName:"webgl",kernelFunc:nae},sae="return x * x;",aae=Ye({opSnippet:sae}),oae={kernelName:kd,backendName:"webgl",kernelFunc:aae},P1="return (a - b) * (a - b);",iae=dn({opSnippet:P1,packedOpSnippet:P1}),cae={kernelName:li,backendName:"webgl",kernelFunc:iae};function uae({inputs:e,attrs:t,backend:n}){let{x:r}=e,s=Or+`
     return x > 0.0 ? 1.0 : float(${t.alpha});
-  `,s=new Cr(a.shape,r);return n.runWebGLProgram(s,[a],a.dtype)}var ose={kernelName:xs,backendName:"webgl",kernelFunc:ise},lse=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=n;let a=n.length,r=gt(n.length),s=gt(n.length),i="";if(a===1)i="coords * strides + begin";else{let o=0;i=n.map((l,u)=>(o++,n.length===1?`coords * strides[${u}] + begin[${u}]`:`coords[${o-1}] * strides[${u}] + begin[${u}]`)).join(",")}this.userCode=`
-      ${r} begin = ${r}(${e});
-      ${r} strides = ${r}(${t});
+  `,a=new Ns(r.shape,s);return n.runWebGLProgram(a,[r],r.dtype)}var lae={kernelName:xa,backendName:"webgl",kernelFunc:uae},dae=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=n;let r=n.length,s=mt(n.length),a=mt(n.length),o="";if(r===1)o="coords * strides + begin";else{let i=0;o=n.map((c,u)=>(i++,n.length===1?`coords * strides[${u}] + begin[${u}]`:`coords[${i-1}] * strides[${u}] + begin[${u}]`)).join(",")}this.userCode=`
+      ${s} begin = ${s}(${e});
+      ${s} strides = ${s}(${t});
 
       void main() {
-        ${s} coords = getOutputCoords();
-        setOutput(getX(${i}));
+        ${a} coords = getOutputCoords();
+        setOutput(getX(${o}));
       }
-    `}};function use(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a,{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=de({inputs:{x:r},backend:n,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let C=jt.computeOutShape(b,x,w),E=np({inputs:{x:r},backend:n,attrs:{begin:b,size:C}});I=de({inputs:{x:E},backend:n,attrs:{shape:m}}),n.disposeIntermediateTensorInfo(E)}else if(n.shouldExecuteOnCPU([r])){let C=n.readSync(r.dataId),E=Oe(r.shape,r.dtype,C),A=M7(h,E,w,b);I=n.makeTensorInfo(m,r.dtype,A.values)}else{let C=new lse(b,w,h);I=n.runWebGLProgram(C,[r],r.dtype)}let T=de({inputs:{x:I},backend:n,attrs:{shape:m}});return n.disposeIntermediateTensorInfo(I),T}var pse={kernelName:$u,backendName:"webgl",kernelFunc:use};function cse(e){let{inputs:t,backend:n,attrs:a}=e,{separator:r,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=a,{data:p,dataSplits:d}=t,c=n.readSync(p.dataId),h=n.readSync(d.dataId),[m,f]=P7(c,h,r,s,i,o,l,u);return[n.makeTensorInfo([m.length],"string",m),n.makeTensorInfo(d.shape,"int32",f)]}var dse={kernelName:Sc,backendName:"webgl",kernelFunc:cse};function hse(e){let{inputs:t,backend:n,attrs:a}=e,{skipEmpty:r}=a,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.readSync(s.dataId),l=n.readSync(i.dataId)[0],[u,p,d]=O7(o,l,r),c=p.length;return[n.makeTensorInfo([c,2],"int32",u),n.makeTensorInfo([c],"string",p),n.makeTensorInfo([2],"int32",new Int32Array(d))]}var mse={kernelName:Tc,backendName:"webgl",kernelFunc:hse};function fse(e){let{inputs:t,backend:n,attrs:a}=e,{numBuckets:r}=a,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(r<=0)throw new Error("Number of buckets must be at least 1");let i=n.readSync(s.dataId),o=L7(i,r);return n.makeTensorInfo(s.shape,"int32",o)}var gse={kernelName:Nc,backendName:"webgl",kernelFunc:fse},yse="return tan(x);",bse=Ye({opSnippet:yse}),xse={kernelName:po,backendName:"webgl",kernelFunc:bse},vse=`
+    `}};function pae(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{begin:a,end:o,strides:i,beginMask:c,endMask:u,ellipsisMask:l,newAxisMask:p,shrinkAxisMask:d}=r,{finalShapeSparse:h,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:b,begin:y,end:v,strides:x}=qt.sliceInfo(s.shape,a,o,i,c,u,l,p,d),k;if(m)k=he({inputs:{x:s},backend:n,attrs:{shape:f}});else if(g||b){w.assert(s.shape.length>=1,()=>`Input must have rank at least 1, got: ${s.shape.length}`);let C=qt.computeOutShape(y,v,x),E=nl({inputs:{x:s},backend:n,attrs:{begin:y,size:C}});k=he({inputs:{x:E},backend:n,attrs:{shape:f}}),n.disposeIntermediateTensorInfo(E)}else if(n.shouldExecuteOnCPU([s])){let E=n.readSync(s.dataId),$=Me(s.shape,s.dtype,E),F=L9(h,$,x,y);k=n.makeTensorInfo(f,s.dtype,F.values)}else{let E=new dae(y,x,h);k=n.runWebGLProgram(E,[s],s.dtype)}let S=he({inputs:{x:k},backend:n,attrs:{shape:f}});return n.disposeIntermediateTensorInfo(k),S}var hae={kernelName:$u,backendName:"webgl",kernelFunc:pae};function fae(e){let{inputs:t,backend:n,attrs:r}=e,{separator:s,nGramWidths:a,leftPad:o,rightPad:i,padWidth:c,preserveShortSequences:u}=r,{data:l,dataSplits:p}=t,d=n.readSync(l.dataId),h=n.readSync(p.dataId),[f,m]=z9(d,h,s,a,o,i,c,u);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(p.shape,"int32",m)]}var mae={kernelName:Sd,backendName:"webgl",kernelFunc:fae};function gae(e){let{inputs:t,backend:n,attrs:r}=e,{skipEmpty:s}=r,{input:a,delimiter:o}=t;if(a.dtype!=="string")throw new Error("Input must be of datatype string");if(a.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${a.shape}`);if(o.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${o.shape}`);let i=n.readSync(a.dataId),c=n.readSync(o.dataId)[0],[u,l,p]=B9(i,c,s),d=l.length;return[n.makeTensorInfo([d,2],"int32",u),n.makeTensorInfo([d],"string",l),n.makeTensorInfo([2],"int32",new Int32Array(p))]}var bae={kernelName:Td,backendName:"webgl",kernelFunc:gae};function yae(e){let{inputs:t,backend:n,attrs:r}=e,{numBuckets:s}=r,{input:a}=t;if(a.dtype!=="string")throw new Error("Input must be of datatype string");if(s<=0)throw new Error("Number of buckets must be at least 1");let o=n.readSync(a.dataId),i=W9(o,s);return n.makeTensorInfo(a.shape,"int32",i)}var vae={kernelName:Cd,backendName:"webgl",kernelFunc:yae},xae="return tan(x);",wae=Ye({opSnippet:xae}),Iae={kernelName:pi,backendName:"webgl",kernelFunc:wae},kae=`
   float e2x = exp(-2.0 * abs(x));
   return sign(x) * (1.0 - e2x) / (1.0 + e2x);
-`,wse=Ye({opSnippet:vse}),kse={kernelName:co,backendName:"webgl",kernelFunc:wse},Ise=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let s=0;s<n.length;s++)n[s]=e[s]*t[s];this.outputShape=n,this.rank=n.length;let a=gt(this.rank),r=Sse(e);this.userCode=`
+`,Sae=Ye({opSnippet:kae}),Tae={kernelName:hi,backendName:"webgl",kernelFunc:Sae},Cae=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let a=0;a<n.length;a++)n[a]=e[a]*t[a];this.outputShape=n,this.rank=n.length;let r=mt(this.rank),s=Nae(e);this.userCode=`
       void main() {
-        ${a} resRC = getOutputCoords();
-        setOutput(getA(${r}));
+        ${r} resRC = getOutputCoords();
+        setOutput(getA(${s}));
       }
-    `}};function Sse(e){let t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${e[0]})`;let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],a=[];for(let r=0;r<e.length;r++)a.push(`imod(${n[r]}, ${e[r]})`);return a.join()}function zE(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{reps:s}=a;if(r.dtype==="string"||r.shape.length>5){let o=n.readSync(r.dataId),l=r.dtype==="string"?o.map(d=>v.decodeString(d)):o,u=Oe(r.shape,r.dtype,l),p=W7(u,s);return n.makeTensorInfo(p.shape,p.dtype,p.values)}let i=new Ise(r.shape,s);return n.runWebGLProgram(i,[r],r.dtype)}var Tse={kernelName:bs,backendName:"webgl",kernelFunc:zE},Nse=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode=`
+    `}};function Nae(e){let t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${e[0]})`;let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],r=[];for(let s=0;s<e.length;s++)r.push(`imod(${n[s]}, ${e[s]})`);return r.join()}function WE(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{reps:a}=r;if(s.dtype==="string"||s.shape.length>5){let c=n.readSync(s.dataId),u=s.dtype==="string"?c.map(d=>w.decodeString(d)):c,l=Me(s.shape,s.dtype,u),p=U9(l,a);return n.makeTensorInfo(p.shape,p.dtype,p.values)}let o=new Cae(s.shape,a);return n.runWebGLProgram(o,[s],s.dtype)}var _ae={kernelName:va,backendName:"webgl",kernelFunc:WE},Eae=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode=`
        void main() {
          ivec2 coords = getOutputCoords();
          int batch = coords[0];
@@ -4624,7 +4624,7 @@ return a / b;`,Bne=`
             setOutput(float(i1));
          }
        }
-     `}},Cse=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode=`
+     `}},Aae=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode=`
     void main() {
          // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...
          ivec2 coords = getOutputCoords();
@@ -4658,10 +4658,10 @@ return a / b;`,Bne=`
 
          setOutput(x0 >= x1 ? float(i0) : float(i1));
        }
-     `}};function Ls(e,t){t!==null&&e.disposeIntermediateTensorInfo(t)}function MI(e){let t=1;for(;t<e;)t*=2;return t}function _se(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{k:s,sorted:i}=a,o=H().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),l=H().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),u=r.shape,p=u[u.length-1];if(n.shouldExecuteOnCPU([r])||p<o||s>l){let A=n.readSync(r.dataId),[R,F]=B7(A,u,r.dtype,s,i);return[n.makeTensorInfo(R.shape,R.dtype,R.values),n.makeTensorInfo(F.shape,F.dtype,F.values)]}if(s===0)return u[u.length-1]=0,[n.makeTensorInfo(u,r.dtype,[]),n.makeTensorInfo(u,"int32",[])];if(p===1)return[r,cd({attrs:{shape:u,dtype:"int32",value:0},backend:n})];let d=n.texData.get(r.dataId),c=d!==null&&d.isPacked,h=c?n.unpackTensor(r):r,m=v.sizeFromShape(u)/p,f=de({inputs:{x:h},attrs:{shape:[m,p]},backend:n});c&&Ls(n,h);let g=MI(s),y=MI(p),b=null,x=()=>b===null?[f,f]:[f,b],w=(A,R,F)=>{let S=x(),M=new Nse(F),B=[[p],[b===null?1:0],[Number.NEGATIVE_INFINITY],[A],[R]],U=b;b=n.runWebGLProgram(M,S,"int32",B),Ls(n,U)};for(let A=1;A<g;A*=2){let R=A*2;for(let F=A;F>=1;F/=2)w(R,F,[m,y])}for(let A=y;A>g;A/=2){let R=x(),F=new Cse([m,A/2]),S=[[p],[b===null?1:0],[g]],M=b;b=n.runWebGLProgram(F,R,"int32",S),Ls(n,M);let B=g/2,U=B*2;for(let G=B;G>=1;G/=2)w(U,G,b.shape)}let I=b;b=np({inputs:{x:b},backend:n,attrs:{begin:0,size:[m,s]}}),Ls(n,I);let T=AE({inputs:{x:f,indices:b},backend:n,attrs:{axis:1,batchDims:1}});Ls(n,f);let C=u.slice(0,-1);C.push(s),I=b,b=de({inputs:{x:b},attrs:{shape:C},backend:n}),Ls(n,I);let E=T;return T=de({inputs:{x:T},attrs:{shape:C},backend:n}),Ls(n,E),[T,b]}var Ese={kernelName:Fu,backendName:"webgl",kernelFunc:_se},Ase=class{constructor(e,t,n,a,r,s){this.variableNames=["Image","Transforms"],this.outputShape=s;let i=n==="nearest"?1:2,o;switch(a){case"constant":o=1;break;case"reflect":o=2;break;case"wrap":o=3;break;case"nearest":o=4;break;default:o=1;break}this.userCode=`
+     `}};function za(e,t){t!==null&&e.disposeIntermediateTensorInfo(t)}function O1(e){let t=1;for(;t<e;)t*=2;return t}function $ae(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{k:a,sorted:o}=r,i=q().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),c=q().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),u=s.shape,l=u[u.length-1];if(n.shouldExecuteOnCPU([s])||l<i||a>c){let F=n.readSync(s.dataId),[A,R]=G9(F,u,s.dtype,a,o);return[n.makeTensorInfo(A.shape,A.dtype,A.values),n.makeTensorInfo(R.shape,R.dtype,R.values)]}if(a===0)return u[u.length-1]=0,[n.makeTensorInfo(u,s.dtype,[]),n.makeTensorInfo(u,"int32",[])];if(l===1)return[s,dp({attrs:{shape:u,dtype:"int32",value:0},backend:n})];let p=n.texData.get(s.dataId),d=p!==null&&p.isPacked,h=d?n.unpackTensor(s):s,m=w.sizeFromShape(u)/l,g=he({inputs:{x:h},attrs:{shape:[m,l]},backend:n});d&&za(n,h);let b=O1(a),y=O1(l),v=null,x=()=>v===null?[g,g]:[g,v],k=(F,A,R)=>{let T=x(),L=new Eae(R),G=[[l],[v===null?1:0],[Number.NEGATIVE_INFINITY],[F],[A]],j=v;v=n.runWebGLProgram(L,T,"int32",G),za(n,j)};for(let F=1;F<b;F*=2){let A=F*2;for(let R=F;R>=1;R/=2)k(A,R,[m,y])}for(let F=y;F>b;F/=2){let A=x(),R=new Aae([m,F/2]),L=[[l],[v===null?1:0],[b]],V=v;v=n.runWebGLProgram(R,A,"int32",L),za(n,V);let G=b/2,j=G*2;for(let H=G;H>=1;H/=2)k(j,H,v.shape)}let S=v;v=nl({inputs:{x:v},backend:n,attrs:{begin:0,size:[m,a]}}),za(n,S);let C=DE({inputs:{x:g,indices:v},backend:n,attrs:{axis:1,batchDims:1}});za(n,g);let E=u.slice(0,-1);E.push(a),S=v,v=he({inputs:{x:v},attrs:{shape:E},backend:n}),za(n,S);let $=C;return C=he({inputs:{x:C},attrs:{shape:E},backend:n}),za(n,$),[C,v]}var Dae={kernelName:Du,backendName:"webgl",kernelFunc:$ae},Fae=class{constructor(e,t,n,r,s,a){this.variableNames=["Image","Transforms"],this.outputShape=a;let o=n==="nearest"?1:2,i;switch(r){case"constant":i=1;break;case"reflect":i=2;break;case"wrap":i=3;break;case"nearest":i=4;break;default:i=1;break}this.userCode=`
             float mapCoord(float outCoord, float len) {
               float inCoord = outCoord;
-              if(${o} == 2) {
+              if(${i} == 2) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4685,7 +4685,7 @@ return a / b;`,Bne=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${o} == 3) {
+              } else if (${i} == 3) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4702,7 +4702,7 @@ return a / b;`,Bne=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${o} == 4) {
+              } else if (${i} == 4) {
                 return clamp(outCoord, 0.0, len - 1.0);
               } else {
                 return outCoord;
@@ -4715,7 +4715,7 @@ return a / b;`,Bne=`
               if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {
                   outputValue = getImage(batch, coordY, coordX, channel);
               } else {
-                outputValue = float(${r});
+                outputValue = float(${s});
               }
               return outputValue;
             }
@@ -4739,14 +4739,14 @@ return a / b;`,Bne=`
               float c2 = getTransforms(batch, 7);
               float projection = c1 * xf + c2 * yf + 1.0;
               if (projection == 0.0) {
-                outputValue = float(${r});
+                outputValue = float(${s});
               } else {
                 float inX = (a1 * xf + a2 * yf + a3) / projection;
                 float inY = (b1 * xf + b2 * yf + b3) / projection;
                 float mapX = mapCoord(inX, float(${t}));
                 float mapY = mapCoord(inY, float(${e}));
 
-                if (${i} == 1) {
+                if (${o} == 1) {
                   int coordY = int(round(mapY));
                   int coordX = int(round(mapX));
                   outputValue = readWithFillValue(batch, coordY, coordX,
@@ -4770,21 +4770,21 @@ return a / b;`,Bne=`
               }
               setOutput(outputValue);
             }
-        `}};function $se(e){let{inputs:t,backend:n,attrs:a}=e,{image:r,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=a,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=new Ase(d,c,i,o,l,g);return n.runWebGLProgram(y,[r,s],"float32")}var Fse={kernelName:Du,backendName:"webgl",kernelFunc:$se};function Dse(e){let{inputs:t,attrs:n,backend:a}=e,{axis:r}=n,{x:s}=t;Yu(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let i=a.readSync(s.dataId),{outputValues:o,outputShape:l,indices:u}=V7(i,r,s.shape,s.dtype);return[a.makeTensorInfo(l,s.dtype,o),a.makeTensorInfo([u.length],"int32",u)]}var Rse={kernelName:Bm,backendName:"webgl",kernelFunc:Dse};function Mse(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r,o=i.shape.length,l=r.shape[s],u=new Array(o-1),p=0;for(let f=0;f<o;f++)f!==s&&(u[p++]=i.shape[f]);let d=[],c=new Array(o).fill(0),h=i.shape.slice();h[s]=1;let m=new Array(l);for(let f=0;f<m.length;f++){c[s]=f;let g=np({inputs:{x:i},backend:n,attrs:{begin:c,size:h}}),y=de({inputs:{x:g},backend:n,attrs:{shape:u}});m[f]=y,d.push(g)}return d.forEach(f=>n.disposeIntermediateTensorInfo(f)),m}var Pse={kernelName:Ru,backendName:"webgl",kernelFunc:Mse},Ose=class{constructor(e,t){this.variableNames=["x","segmentIds"];let n=e.windowSize,a=e.batchSize,r=e.inSize,s=e.numSegments,i=s*Math.ceil(r/n);this.outputShape=[a,i];let o="0.0",l="sumValue",u=Math.floor(n/4)*4,p=n%4,d=`
+        `}};function Rae(e){let{inputs:t,backend:n,attrs:r}=e,{image:s,transforms:a}=t,{interpolation:o,fillMode:i,fillValue:c,outputShape:u}=r,[l,p,d,h]=s.shape,[f,m]=u!=null?u:[p,d],g=[l,f,m,h],b=new Fae(p,d,o,i,c,g);return n.runWebGLProgram(b,[s,a],"float32")}var Pae={kernelName:Fu,backendName:"webgl",kernelFunc:Rae};function Oae(e){let{inputs:t,attrs:n,backend:r}=e,{axis:s}=n,{x:a}=t;Yu(a,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let o=r.readSync(a.dataId),{outputValues:i,outputShape:c,indices:u}=H9(o,s,a.shape,a.dtype);return[r.makeTensorInfo(c,a.dtype,i),r.makeTensorInfo([u.length],"int32",u)]}var Mae={kernelName:Uf,backendName:"webgl",kernelFunc:Oae};function Lae(e){let{inputs:t,backend:n,attrs:r}=e,{value:s}=t,{axis:a}=r;a<0&&(a+=s.shape.length);let o=s,i=o.shape.length,c=s.shape[a],u=new Array(i-1),l=0;for(let m=0;m<i;m++)m!==a&&(u[l++]=o.shape[m]);let p=[],d=new Array(i).fill(0),h=o.shape.slice();h[a]=1;let f=new Array(c);for(let m=0;m<f.length;m++){d[a]=m;let g=nl({inputs:{x:o},backend:n,attrs:{begin:d,size:h}}),b=he({inputs:{x:g},backend:n,attrs:{shape:u}});f[m]=b,p.push(g)}return p.forEach(m=>n.disposeIntermediateTensorInfo(m)),f}var zae={kernelName:Ru,backendName:"webgl",kernelFunc:Lae},Bae=class{constructor(e,t){this.variableNames=["x","segmentIds"];let n=e.windowSize,r=e.batchSize,s=e.inSize,a=e.numSegments,o=a*Math.ceil(s/n);this.outputShape=[r,o];let i="0.0",c="sumValue",u=Math.floor(n/4)*4,l=n%4,p=`
         sumValue += dot(values, segFilter);
-    `,c="";r%n>0&&(c=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+    `,d="";s%n>0&&(d=`
+        if (inIdx < 0 || inIdx >= ${s}) {
           return initializationValue;
         }
-      `);let h="";r%n>0&&(h=`
-        if (inIdx < 0 || inIdx >= ${r}) {
+      `);let h="";s%n>0&&(h=`
+        if (inIdx < 0 || inIdx >= ${s}) {
           return -1.0;
         }
       `),this.userCode=`
-      const float initializationValue = ${o};
+      const float initializationValue = ${i};
 
       float getValue(int batch, int inIdx) {
-        ${c}
+        ${d}
         return getX(batch, inIdx);
       }
 
@@ -4798,8 +4798,8 @@ return a / b;`,Bne=`
         int batch = coords[0];
         int outIdx = coords[1];
         int inOffset = int(floor(float(outIdx) / float(
-          ${s})) * float(${n}));
-        int currentSeg = int(mod(float(outIdx), float(${s})));
+          ${a})) * float(${n}));
+        int currentSeg = int(mod(float(outIdx), float(${a})));
 
         float sumValue = 0.0;
 
@@ -4819,11 +4819,11 @@ return a / b;`,Bne=`
             int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0
           );
 
-          ${d}
+          ${p}
         }
 
         int inIdx = inOffset + ${u};
-        if (${p===1}) {
+        if (${l===1}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             initializationValue,
@@ -4840,8 +4840,8 @@ return a / b;`,Bne=`
             0
           );
 
-          ${d}
-        } else if (${p===2}) {
+          ${p}
+        } else if (${l===2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4856,8 +4856,8 @@ return a / b;`,Bne=`
               0
           );
 
-          ${d}
-        } else if (${p===3}) {
+          ${p}
+        } else if (${l===3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4872,10 +4872,10 @@ return a / b;`,Bne=`
             0
           );
 
-          ${d}
+          ${p}
         }
-        setOutput(${l});
+        setOutput(${c});
       }
-    `}};function Lse(e){let{inputs:t,backend:n,attrs:a}=e,{x:r,segmentIds:s}=t,{numSegments:i}=a,o=r.shape.length,l=[],u=0,p=N.getAxesPermutation([u],o),d=r;p!=null&&(d=Sn({inputs:{x:r},backend:n,attrs:{perm:p}}),l.push(d),u=N.getInnerMostAxes(1,o)[0]);let c=N.segment_util.computeOutShape(d.shape,u,i),h=v.sizeFromShape([d.shape[u]]),m=de({inputs:{x:d},backend:n,attrs:{shape:[-1,h]}});l.push(m);let f=Um(r.dtype),g=(w,I,T,C,E)=>{let A=w.shape[0],R=w.shape[1],F=N.segment_util.segOpComputeOptimalWindowSize(R,E),S={windowSize:F,inSize:R,batchSize:A,numSegments:E},M=new Ose(S,I),B=n.compileAndRun(M,[w,T],C);if(l.push(B),B.shape[1]===E)return B;let U=OE({backend:n,attrs:{start:0,stop:E,step:1,dtype:"float32"}}),G=zE({inputs:{x:U},backend:n,attrs:{reps:[R/F]}});return l.push(U),l.push(G),g(B,I,G,C,E)},y=g(m,"unsortedSegmentSum",s,f,i),b=de({inputs:{x:y},backend:n,attrs:{shape:c}}),x=b;if(p!=null){l.push(b);let w=N.getUndoAxesPermutation(p);x=Sn({inputs:{x},backend:n,attrs:{perm:w}})}return l.forEach(w=>n.disposeIntermediateTensorInfo(w)),x}var zse={kernelName:Cc,backendName:"webgl",kernelFunc:Lse},Wse=[MJ,OJ,WJ,UJ,HJ,KJ,YJ,JJ,n9,r9,o9,p9,h9,y9,v9,k9,S9,_9,A9,F9,P9,U9,H9,q9,Q9,tQ,sQ,bJ,lQ,hQ,yQ,IQ,TQ,CQ,EQ,$Q,RQ,OQ,WQ,VQ,GQ,jQ,XQ,ZQ,tee,aee,iee,uee,cee,fee,xee,Iee,Nee,Eee,Aee,Fee,Ree,Pee,Lee,Wee,Gee,qee,Yee,Jee,tte,rte,lte,dte,yJ,mte,cQ,yte,vte,Ite,vJ,Cte,$te,Dte,Ote,Wte,Gte,qte,Zte,tne,rne,ine,pne,dne,mne,bne,vne,kne,Sne,Nne,Ane,Rne,Lne,jne,IJ,Yne,Qne,nae,sae,X9,lae,pae,dae,fae,xae,kJ,wae,Iae,Tae,Cae,_ae,Y9,Vne,$ae,Mae,zae,TJ,Uae,jae,Yae,Qae,are,sre,lre,cre,hre,gre,xre,kre,Nre,Ere,Fre,Mre,B9,Gne,Lre,Wre,Vre,Gre,jre,Kre,Yre,Jre,ese,ase,sse,ose,pse,dse,mse,gse,Une,FJ,xse,kse,Tse,Ese,Fse,DJ,Rse,Pse,zse,uae];for(let e of Wse)_c(e);var Et;(function(e){e[e.float32=0]="float32",e[e.int32=1]="int32",e[e.bool=2]="bool",e[e.string=3]="string",e[e.complex64=4]="complex64"})(Et||(Et={}));var uc;(function(e){e[e.linear=0]="linear",e[e.relu=1]="relu",e[e.relu6=2]="relu6",e[e.prelu=3]="prelu",e[e.leakyrelu=4]="leakyrelu",e[e.sigmoid=5]="sigmoid",e[e.elu=6]="elu"})(uc||(uc={}));var WE;function Bse(e){WE=e.wasm.cwrap(Qs,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])}function Vse(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s,bias:i,preluActivationWeights:o}=t;if(r.dtype!=="float32"||s.dtype!=="float32")throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:l,transposeB:u,activation:p,leakyreluAlpha:d}=a,c=n.dataIdMap.get(r.dataId).id,h=n.dataIdMap.get(s.dataId).id,m=0;if(i!=null){let E=n.dataIdMap.get(i.dataId);if(E.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${E.shape.length}.`);m=E.id}let f=o==null?0:n.dataIdMap.get(o.dataId).id,g=uc[p];if(g==null)throw new Error(`${p} activation not yet supported for FusedConv2D in the wasm backend.`);let y=l?r.shape[2]:r.shape[1],b=u?s.shape[1]:s.shape[2],x=Ou.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)),w=n.makeOutput([...x,y,b],r.dtype),I=n.dataIdMap.get(w.dataId).id,T=new Uint8Array(new Int32Array(r.shape).buffer),C=new Uint8Array(new Int32Array(s.shape).buffer);return WE(c,T,r.shape.length,h,C,s.shape.length,l,u,g,m,f,d||0,I),w}var Use={kernelName:Qs,backendName:"wasm",setupFunc:Bse,kernelFunc:Vse};function an(e,t){let n;function a(s){n=s.wasm.cwrap(e,null,["number","number","number"])}function r(s){let{backend:i,inputs:{x:o}}=s,l=i.dataIdMap.get(o.dataId).id,u=i.makeOutput(o.shape,t||o.dtype),p=i.dataIdMap.get(u.dataId).id;return v.sizeFromShape(u.shape)===0||n(l,Et[o.dtype],p),u}return{kernelName:e,backendName:"wasm",setupFunc:a,kernelFunc:r}}var Gse=an(Dl);function dn(e,t,n){let a;function r(i){a=i.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])}function s(i){let{backend:o,inputs:l}=i,{a:u,b:p}=l,d=o.dataIdMap.get(u.dataId).id,c=o.dataIdMap.get(p.dataId).id,h=n!=null?n:u.dtype,m=N.assertAndGetBroadcastShape(u.shape,p.shape),f=o.makeOutput(m,h);if(v.sizeFromShape(m)===0)return f;let g=new Uint8Array(new Int32Array(u.shape).buffer),y=new Uint8Array(new Int32Array(p.shape).buffer),b=o.dataIdMap.get(f.dataId).id;return a(d,g,u.shape.length,c,y,p.shape.length,Et[u.dtype],b),f}return{kernelName:e,backendName:"wasm",setupFunc:r,kernelFunc:s}}var Hse=!0,jse=dn(gs,Hse),BE;function qse(e){BE=e.wasm.cwrap(fi,null,["array","number","number","number"])}function Kse(e){let{inputs:t,backend:n}=e,a=n.makeOutput(t[0].shape,t[0].dtype);if(v.sizeFromShape(a.shape)===0)return a;let r=t.map(o=>n.dataIdMap.get(o.dataId).id),s=new Uint8Array(new Int32Array(r).buffer),i=n.dataIdMap.get(a.dataId).id;return BE(s,r.length,Et[a.dtype],i),a}var Xse={kernelName:fi,backendName:"wasm",setupFunc:qse,kernelFunc:Kse};function Qf(e){let{inputs:{x:t},backend:n}=e;if(t.dtype==="string")return In(n.readSync(t.dataId),t.shape,t.dtype);let a=n.makeOutput(t.shape,t.dtype),r=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(a).set(r),a}var Yse={kernelName:Ri,backendName:"wasm",kernelFunc:Qf},VE;function Zse(e){VE=e.wasm.cwrap(_r,null,["number","array","number","number","number","array","number"])}function hs(e){let{inputs:t,backend:n,attrs:a}=e,[r,s]=Qse(t.x.shape,a.perm),i=!0;for(let m=0;m<s.length;m++)s[m]!==m&&(i=!1);let o=Jse(t.x.shape,a.perm),l={dataId:t.x.dataId,shape:r,dtype:t.x.dtype};if(i){let m=Qf({inputs:t,backend:n});return m.shape=o,m}let u=n.makeOutput(o,l.dtype),p=n.dataIdMap.get(l.dataId).id,d=n.dataIdMap.get(u.dataId).id,c=new Uint8Array(new Int32Array(s).buffer),h=new Uint8Array(new Int32Array(l.shape).buffer);return VE(p,h,l.shape.length,Et[l.dtype],d,c,s.length),u}function Jse(e,t){let n=new Array(e.length);for(let a=0;a<n.length;a++)n[a]=e[t[a]];return n}function Qse(e,t){let n=[],a=[];for(let r=0;r<e.length;++r)e[r]!==1&&n.push(e[r]),e[t[r]]!==1&&a.push(t[r]);for(let r=0;r<a.length;++r){let s=-1;for(let i=0;i<a.length;++i)a[i]>=r&&(s===-1||a[s]>a[i])&&(s=i);a[s]=r}return[n,a]}var eie={kernelName:_r,backendName:"wasm",kernelFunc:hs,setupFunc:Zse};function Es(e,t,n){let a=e.shape,r=e.shape.length,s=v.parseAxisParam(t,a),i=s,o=N.getAxesPermutation(i,r),l=null,u=!1;if(o!=null){let p=new Array(r);for(let c=0;c<p.length;c++)p[c]=a[o[c]];i=N.getInnerMostAxes(i.length,r),l=hs({inputs:{x:e},attrs:{perm:o},backend:n});let d=n.dataIdMap.get(e.dataId).id;n.dataIdMap.get(l.dataId).id!==d&&(u=!0)}return{transposed:l,originalAxes:s,axes:i,inputWasTransposed:u}}var UE;function tie(e){UE=e.wasm.cwrap(Pl,null,["number, number, number"])}function nie(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Es(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("all",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;UE(o,g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var aie={kernelName:Pl,backendName:"wasm",setupFunc:tie,kernelFunc:nie},GE;function rie(e){GE=e.wasm.cwrap(Ol,null,["number, number, number"])}function sie(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Es(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("any",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;GE(o,g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var iie={kernelName:Ol,backendName:"wasm",setupFunc:rie,kernelFunc:sie},HE;function oie(e){HE=e.wasm.cwrap(gi,null,["number","number","number","number","number"])}function lie(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r}=a,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=i,l=s,{transposed:u,axes:p,inputWasTransposed:d}=Es(s,r,t);if(d){let y=t.dataIdMap.get(u.dataId).id;y!==i&&(l=u,o=y)}let c=l.shape.slice(0,-1),h=t.makeOutput(c,"int32"),m=t.dataIdMap.get(h.dataId).id,f=v.sizeFromShape(h.shape),g=l.shape[p[0]];return HE(o,Et[l.dtype],f,g,m),d&&t.disposeData(u.dataId),h}var uie={kernelName:gi,backendName:"wasm",kernelFunc:lie,setupFunc:oie},jE;function pie(e){jE=e.wasm.cwrap(yi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function cie(e){let{inputs:t,attrs:n,backend:a}=e,r=t.x,s=a.dataIdMap.get(r.dataId).id,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,p=N.computePool2DInfo(r.shape,i,o,1,l,u),d=p.filterHeight,c=p.filterWidth,h=p.padInfo.top,m=p.padInfo.right,f=p.padInfo.bottom,g=p.padInfo.left,y=p.strideHeight,b=p.strideWidth,x=p.inChannels;if(p.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${p.dataFormat}'. Please use 'channelsLast'.`);if(p.dilationWidth!==1||p.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${p.dilationHeight}, ${p.dilationWidth}].`);let w=a.makeOutput(p.outShape,"float32"),I=a.dataIdMap.get(w.dataId).id;return jE(s,r.shape[0],r.shape[1],r.shape[2],d,c,h,m,f,g,y,b,x,I),w}var die={kernelName:yi,backendName:"wasm",setupFunc:pie,kernelFunc:cie};function Wn(e){let{inputs:t,attrs:n}=e,{x:a}=t,{shape:r}=n,s=v.sizeFromShape(a.shape),i=v.inferFromImplicitShape(r,s);return v.assert(s===v.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${a.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(a.dataId),{dataId:a.dataId,shape:i,dtype:a.dtype}}var hie={kernelName:vu,backendName:"wasm",kernelFunc:Wn},qE;function mie(e){qE=e.wasm.cwrap(bi,null,["number","array","number","number","array","number","number","number","number"])}function fie(e){let{inputs:t,backend:n,attrs:a}=e,{a:r,b:s}=t,{transposeA:i,transposeB:o}=a;if(r.dtype!=="float32"||s.dtype!=="float32")throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let l=r.shape.length,u=s.shape.length,p=i?r.shape[l-2]:r.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],c=i?r.shape[l-1]:r.shape[l-2],h=o?s.shape[u-2]:s.shape[u-1],m=r.shape.slice(0,-2),f=s.shape.slice(0,-2),g=v.sizeFromShape(m),y=v.sizeFromShape(f),b=Ou.assertAndGetBroadcastShape(r.shape.slice(0,-2),s.shape.slice(0,-2)).concat([c,h]);v.assert(p===d,()=>`Error in matMul: inner shapes (${p}) and (${d}) of Tensors with shapes ${r.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,p,c]:[g,c,p],w=o?[y,h,d]:[y,d,h],I=Wn({inputs:{x:r},backend:n,attrs:{shape:x}}),T=Wn({inputs:{x:s},backend:n,attrs:{shape:w}}),C=n.dataIdMap.get(I.dataId).id,E=n.dataIdMap.get(T.dataId).id,A=i?I.shape[2]:I.shape[1],R=o?T.shape[1]:T.shape[2],F=Math.max(g,y),S=n.makeOutput([F,A,R],I.dtype),M=n.dataIdMap.get(S.dataId).id,B=new Uint8Array(new Int32Array(I.shape).buffer),U=new Uint8Array(new Int32Array(T.shape).buffer);return qE(C,B,I.shape.length,E,U,T.shape.length,i,o,M),n.disposeData(I.dataId),n.disposeData(T.dataId),S.shape=b,S}var gie={kernelName:bi,backendName:"wasm",setupFunc:mie,kernelFunc:fie};function hi(e){let{inputs:{x:t},attrs:{begin:n,size:a},backend:r}=e,[s,i]=jt.parseSliceParams(t,n,a),o=jt.isSliceContinous(t.shape,s,i),l=r.readSync(t.dataId),u=r.makeOutput(i,t.dtype),p=v.computeStrides(t.shape),d=r.dataIdMap.get(u.dataId);if(o){let m=jt.computeFlatOffset(s,p);return t.dtype==="string"?d.stringBytes=l.slice(m,m+v.sizeFromShape(i)):r.typedArrayFromHeap(u).set(l.subarray(m,m+v.sizeFromShape(i))),u}if(t.dtype==="string"){let m=Jh(l,s,i,t.shape,t.dtype);return d.stringBytes=m,u}let c=r.typedArrayFromHeap(u),h=t.shape.length;if(h===2)yie(l,p[0],c,s,i);else if(h===3)bie(l,p[0],p[1],c,s,i);else if(h===4)xie(l,p[0],p[1],p[2],c,s,i);else{let m=Jh(l,s,i,t.shape,t.dtype);c.set(m)}return u}function yie(e,t,n,a,r){let s=0,i=a[0],o=a[1],l=i+r[0];for(let u=i;u<l;u++){let p=u*t+o;n.set(e.subarray(p,p+r[1]),s),s+=r[1]}}function bie(e,t,n,a,r,s){let i=0,o=r[0],l=r[1],u=r[2],p=o+s[0],d=l+s[1];for(let c=o;c<p;c++)for(let h=l;h<d;h++){let m=c*t+h*n+u;a.set(e.subarray(m,m+s[2]),i),i+=s[2]}}function xie(e,t,n,a,r,s,i){let o=0,l=s[0],u=s[1],p=s[2],d=l+i[0],c=u+i[1],h=p+i[2],m=s[3];for(let f=l;f<d;f++)for(let g=u;g<c;g++)for(let y=p;y<h;y++){let b=f*t+g*n+y*a+m;r.set(e.subarray(b,b+i[3]),o),o+=i[3]}}var vie={kernelName:Su,backendName:"wasm",kernelFunc:hi};function wie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,crops:i}=a,o=s.reduce((y,b)=>y*b),l=N.getReshaped(r.shape,s,o),u=N.getPermuted(l.length,s.length),p=N.getReshapedPermuted(r.shape,s,o),d=N.getSliceBeginCoords(i,s.length),c=N.getSliceSize(p,i,s.length),h=Wn({inputs:{x:r},backend:n,attrs:{shape:l}}),m=hs({inputs:{x:h},backend:n,attrs:{perm:u}}),f=Wn({inputs:{x:m},backend:n,attrs:{shape:p}}),g=hi({inputs:{x:f},backend:n,attrs:{begin:d,size:c}});return n.disposeData(h.dataId),n.disposeData(m.dataId),n.disposeData(h.dataId),g}var kie={kernelName:Ul,backendName:"wasm",kernelFunc:wie};function ap(e){let{inputs:{x:t},attrs:{dtype:n},backend:a}=e,r=a.makeOutput(t.shape,n),s=a.typedArrayFromHeap(t);return a.typedArrayFromHeap(r).set(s),r}var Iie={kernelName:xi,backendName:"wasm",kernelFunc:ap},Sie=an(vi),KE;function Tie(e){KE=e.wasm.cwrap(ys,null,["number","number","number","number"])}function Nie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{clipValueMin:s,clipValueMax:i}=a,o=n.dataIdMap.get(r.dataId).id,l=n.makeOutput(r.shape,r.dtype),u=n.dataIdMap.get(l.dataId).id;return KE(o,s,i,u),l}var Cie={kernelName:ys,backendName:"wasm",setupFunc:Tie,kernelFunc:Nie};function XE(e){let{inputs:t,backend:n}=e,a=v.parseAxisParam(e.attrs.axis,t[0].shape)[0],r=t.map(h=>h.shape);N.assertParamsConsistent(r,a);let s=N.computeOutShape(t.map(h=>h.shape),a),i=t.filter(h=>v.sizeFromShape(h.shape)>0);if(i.length===1)return Qf({inputs:{x:i[0]},backend:n});let o=n.makeOutput(s,t[0].dtype);if(v.sizeFromShape(s)===0)return o;if(i[0].dtype==="string"){let h=i.map(x=>{let w=v.sizeFromShape(x.shape.slice(a));return Wn({inputs:{x},backend:n,attrs:{shape:[-1,w]}})}),m=h.map(x=>({vals:n.readSync(x.dataId),shape:x.shape}));s=N.computeOutShape(h.map(x=>x.shape),1);let f=h[0].shape[0]===1,g=O0(m,s,t[0].dtype,f),y=N.computeOutShape(i.map(x=>x.shape),a);o.shape=y;let b=n.dataIdMap.get(o.dataId);return b.stringBytes=N.fromStringArrayToUint8(g),h.forEach(x=>n.disposeData(x.dataId)),o}let l=v.sizeFromShape(i[0].shape.slice(0,a)),u=0,p=i.map(h=>{let m=v.sizeFromShape(h.shape.slice(a));return u+=m,m}),d=i.map(h=>n.typedArrayFromHeap(h)),c=n.typedArrayFromHeap(o);for(let h=0;h<l;h++){let m=h*u;for(let f=0;f<d.length;f++){let g=p[f],y=h*g,b=d[f].subarray(y,y+g);c.set(b,m),m+=g}}return o}var _ie={kernelName:Gl,backendName:"wasm",kernelFunc:XE},YE;function Eie(e){YE=e.wasm.cwrap(wi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Aie(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s}=t,i=a.dataIdMap.get(r.dataId).id,o=a.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:p,dimRoundingMode:d,dataFormat:c}=n,h=N.convertConv2DDataFormat(c),m=N.computeConv2DInfo(r.shape,s.shape,l,u,p,d,!1,h),f=m.filterHeight,g=m.filterWidth,y=m.padInfo.top,b=m.padInfo.right,x=m.padInfo.bottom,w=m.padInfo.left,I=m.dilationHeight,T=m.dilationWidth,C=m.strideHeight,E=m.strideWidth,A=m.inChannels,R=m.outChannels,F=m.padInfo.type==="SAME"?1:0;if(m.dataFormat!=="channelsLast")throw new Error(`wasm backend Conv2D does not support dataFormat:'${m.dataFormat}'. Please use 'channelsLast'.`);let S=a.makeOutput(m.outShape,"float32"),M=a.dataIdMap.get(S.dataId).id;return YE(i,r.shape[0],r.shape[1],r.shape[2],o,f,g,y,b,x,w,F,I,T,C,E,A,R,M),S}var $ie={kernelName:wi,backendName:"wasm",setupFunc:Eie,kernelFunc:Aie},ZE;function Fie(e){ZE=e.wasm.cwrap(ki,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Die(e){let{backend:t,inputs:n,attrs:a}=e,{dy:r,filter:s}=n,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,inputShape:p}=a,d=1,c=N.convertConv2DDataFormat(l),h=N.computeConv2DInfo(p,s.shape,i,d,o,u,!1,c),{batchSize:m,filterHeight:f,filterWidth:g,inChannels:y,inHeight:b,inWidth:x,outChannels:w,outHeight:I,outWidth:T,strideHeight:C,strideWidth:E}=h,A=f-1-h.padInfo.top,R=g-1-h.padInfo.left,F=h.dataFormat==="channelsLast",S=v.computeStrides(h.inShape),M=v.computeStrides(r.shape),[B,U,G]=v.computeStrides(s.shape),q=S[0],K=F?S[1]:S[2],Z=F?S[2]:1,Q=F?1:S[1],ee=M[0],ae=F?M[1]:M[2],te=F?M[2]:1,le=F?1:M[1],ie=t.makeOutput(h.inShape,"float32"),ye=t.dataIdMap.get(ie.dataId).id,ue=t.dataIdMap.get(r.dataId).id,xe=t.dataIdMap.get(s.dataId).id;return ZE(ue,xe,m,f,g,b,x,y,I,T,w,C,E,A,R,B,U,G,q,K,Z,Q,ee,ae,te,le,ye),ie}var Rie={kernelName:ki,backendName:"wasm",setupFunc:Fie,kernelFunc:Die},Mie=an(Ii),Pie=an(Si),wx;(function(e){e[e.bilinear=0]="bilinear",e[e.nearest=1]="nearest"})(wx||(wx={}));var JE;function Oie(e){JE=e.wasm.cwrap(jl,null,["number","number","number","number","array","number","number","number","number","number"])}function Lie(e){let{backend:t,inputs:n,attrs:a}=e,{method:r,extrapolationValue:s,cropSize:i}=a,{image:o,boxes:l,boxInd:u}=n,p=l.shape[0],[d,c]=i,h=[p,d,c,o.shape[3]],m=t.dataIdMap.get(o.dataId),f;o.dtype!=="float32"&&(f=ap({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),m=t.dataIdMap.get(f.dataId));let g=m.id,y=t.dataIdMap.get(l.dataId).id,b=t.dataIdMap.get(u.dataId).id,x=t.makeOutput(h,"float32"),w=t.dataIdMap.get(x.dataId).id,I=new Uint8Array(new Int32Array(o.shape).buffer);return JE(g,y,b,p,I,d,c,wx[r],s,w),f!=null&&t.disposeData(f.dataId),x}var zie={kernelName:jl,backendName:"wasm",setupFunc:Oie,kernelFunc:Lie},QE;function Wie(e){QE=e.wasm.cwrap(Hl,null,["number","number","number","number","number","number"])}function Bie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a,l=r.shape.length;v.assert(r.dtype==="float32"||r.dtype==="int32",()=>`cumprod does not support ${r.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([s],l),p=r;u!==null&&(p=hs({inputs:{x:r},attrs:{perm:u},backend:n}));let d=N.getInnerMostAxes(1,l)[0];N.assertAxesAreInnerMostDims("cumprod",[d],l);let c=n.makeOutput(p.shape,p.dtype),h=p.shape[d],m=n.dataIdMap.get(p.dataId).id,f=n.dataIdMap.get(c.dataId).id;QE(m,i?1:0,o?1:0,h,f,Et[r.dtype]);let g=c;if(u!==null){let y=N.getUndoAxesPermutation(u);g=hs({inputs:{x:c},attrs:{perm:y},backend:n}),n.disposeData(p.dataId),n.disposeData(c.dataId)}return g}var Vie={kernelName:Hl,backendName:"wasm",setupFunc:Wie,kernelFunc:Bie},eA;function Uie(e){eA=e.wasm.cwrap(Ti,null,["number","number","number","number","number","number"])}function Gie(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{axis:s,exclusive:i,reverse:o}=a,l=r.shape.length;v.assert(r.dtype==="float32"||r.dtype==="int32",()=>`cumsum does not support ${r.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([s],l),p=r;u!==null&&(p=hs({inputs:{x:r},attrs:{perm:u},backend:n}));let d=N.getInnerMostAxes(1,l)[0];N.assertAxesAreInnerMostDims("cumsum",[d],l);let c=n.makeOutput(p.shape,p.dtype),h=p.shape[d],m=n.dataIdMap.get(p.dataId).id,f=n.dataIdMap.get(c.dataId).id;eA(m,i?1:0,o?1:0,h,f,Et[r.dtype]);let g=c;if(u!==null){let y=N.getUndoAxesPermutation(u);g=hs({inputs:{x:c},attrs:{perm:y},backend:n}),n.disposeData(p.dataId),n.disposeData(c.dataId)}return g}var Hie={kernelName:Ti,backendName:"wasm",setupFunc:Uie,kernelFunc:Gie},tA;function jie(e){tA=e.wasm.cwrap(ql,null,["number","number","number","array","number","array","array","number","number"])}function qie(e){let{backend:t,inputs:n,attrs:a}=e,{x:r}=n,{blockSize:s,dataFormat:i}=a,o=r.shape[0],l=i==="NHWC"?r.shape[1]:r.shape[2],u=i==="NHWC"?r.shape[2]:r.shape[3],p=i==="NHWC"?r.shape[3]:r.shape[1],d=l*s,c=u*s,h=p/(s*s),m=i==="NHWC"?[o,d,c,h]:[o,h,d,c],f=t.makeOutput(m,"float32"),g=t.dataIdMap.get(r.dataId).id,y=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),b=new Uint8Array(new Int32Array(m).buffer),x=new Uint8Array(new Int32Array(v.computeStrides(m)).buffer),w=t.dataIdMap.get(f.dataId).id;return tA(g,s,i==="NHWC"?1:0,y,r.shape.length-1,b,x,m.length,w),f}var Kie={kernelName:ql,backendName:"wasm",setupFunc:jie,kernelFunc:qie},nA;function Xie(e){nA=e.wasm.cwrap(Ni,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Yie(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s}=t,i=a.dataIdMap.get(r.dataId).id,o=a.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:p,dimRoundingMode:d}=n,c=u==null?[1,1]:u,h=N.computeConv2DInfo(r.shape,s.shape,l,c,p,d,!0),m=h.filterHeight,f=h.filterWidth,g=h.padInfo.top,y=h.padInfo.right,b=h.padInfo.bottom,x=h.padInfo.left,w=h.dilationHeight,I=h.dilationWidth,T=h.strideHeight,C=h.strideWidth,E=h.inChannels,A=h.outChannels,R=h.padInfo.type==="SAME"?1:0;if(h.dataFormat!=="channelsLast")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${h.dataFormat}'. Please use 'channelsLast'.`);let F=a.makeOutput(h.outShape,"float32"),S=a.dataIdMap.get(F.dataId).id;return nA(i,r.shape[0],r.shape[1],r.shape[2],o,m,f,g,y,b,x,R,w,I,T,C,E,A,S),F}var Zie={kernelName:Ni,backendName:"wasm",setupFunc:Xie,kernelFunc:Yie},Jie=an(_i),Qie=!1,eoe=dn(Xl,Qie,"bool"),toe=an(Ei,"float32");function kx(e){let{inputs:t,attrs:n,backend:a}=e,{input:r}=t,{dim:s}=n,i=r.shape.length,o=r.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),Wn({inputs:{x:r},backend:a,attrs:{shape:o}})}var noe={kernelName:Yl,backendName:"wasm",kernelFunc:kx};function aA(e){let{attrs:{shape:t,value:n,dtype:a},backend:r}=e,s=r.makeOutput(t,a);return r.typedArrayFromHeap(s).fill(n),s}var aoe={kernelName:gc,backendName:"wasm",kernelFunc:aA},rA;function roe(e){rA=e.wasm.cwrap(Jl,null,["number","number","number","number","number","number"])}function soe(e){let{inputs:t,backend:n}=e,{image:a}=t,r=n.makeOutput(a.shape,a.dtype),s=n.dataIdMap.get(a.dataId).id,i=n.dataIdMap.get(r.dataId).id,[o,l,u,p]=a.shape;return rA(s,o,l,u,p,i),r}var ioe={kernelName:Jl,backendName:"wasm",kernelFunc:soe,setupFunc:roe},ooe=an(Ai),loe=!1,uoe=dn($i,loe),sA;function poe(e){sA=e.wasm.cwrap(Fi,null,["number","number","number","number","number","number","number"])}function coe(e){let{backend:t,inputs:n,attrs:a}=e,{varianceEpsilon:r}=a,{x:s,mean:i,variance:o,offset:l,scale:u}=n,p=t.dataIdMap.get(s.dataId).id,d=t.dataIdMap.get(i.dataId).id,c=t.dataIdMap.get(o.dataId).id,h=l!=null?t.dataIdMap.get(l.dataId).id:0,m=u!=null?t.dataIdMap.get(u.dataId).id:0,f=t.makeOutput(s.shape,s.dtype);if(v.sizeFromShape(s.shape)===0)return f;let g=t.dataIdMap.get(f.dataId).id;return sA(p,d,c,h,m,r,g),f}var doe={kernelName:Fi,backendName:"wasm",setupFunc:poe,kernelFunc:coe},iA;function hoe(e){iA=e.wasm.cwrap(ei,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function moe(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dataFormat:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=n,f=N.computeConv2DInfo(r.shape,s.shape,l,p,u,c),g=uc[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedConv2D in the wasm backend.`);let y=a.dataIdMap.get(r.dataId).id,b=a.dataIdMap.get(s.dataId).id,x=f.outChannels,w=0;if(i!=null){let te=a.dataIdMap.get(i.dataId);if(te.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==x)throw new Error(`FusedConv2D bias shape (${te.shape}) does not match the number of output channels (${x})`);w=te.id}let I=f.filterHeight,T=f.filterWidth,C=f.padInfo.top,E=f.padInfo.right,A=f.padInfo.bottom,R=f.padInfo.left,F=f.dilationHeight,S=f.dilationWidth,M=f.strideHeight,B=f.strideWidth,U=f.inChannels,G=f.padInfo.type==="SAME"?1:0,q=f.batchSize,K=f.inHeight,Z=f.inWidth;if(d!=="NHWC")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let Q=a.makeOutput(f.outShape,"float32"),ee=a.dataIdMap.get(Q.dataId).id,ae=o==null?0:a.dataIdMap.get(o.dataId).id;return iA(y,q,K,Z,b,I,T,w,C,E,A,R,G,F,S,M,B,U,x,g,ae,m||0,ee),Q}var foe={kernelName:ei,backendName:"wasm",setupFunc:hoe,kernelFunc:moe},oA;function goe(e){oA=e.wasm.cwrap(ti,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function yoe(e){let{inputs:t,attrs:n,backend:a}=e,{x:r,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:p,dataFormat:d,dimRoundingMode:c,activation:h,leakyreluAlpha:m}=n,f=N.computeConv2DInfo(r.shape,s.shape,l,p,u,c,!0),g=uc[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let y=a.dataIdMap.get(r.dataId).id,b=a.dataIdMap.get(s.dataId).id,x=f.outChannels,w=0;if(i!=null){let te=a.dataIdMap.get(i.dataId);if(te.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==x)throw new Error(`FusedDepthwiseConv2D bias shape (${te.shape}) does not match the number of output channels (${x})`);w=te.id}let I=f.filterHeight,T=f.filterWidth,C=f.padInfo.top,E=f.padInfo.right,A=f.padInfo.bottom,R=f.padInfo.left,F=f.dilationHeight,S=f.dilationWidth,M=f.strideHeight,B=f.strideWidth,U=f.inChannels,G=f.padInfo.type==="SAME"?1:0,q=f.batchSize,K=f.inHeight,Z=f.inWidth;if(d!=="NHWC")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let Q=a.makeOutput(f.outShape,"float32"),ee=a.dataIdMap.get(Q.dataId).id,ae=o==null?0:a.dataIdMap.get(o.dataId).id;return oA(y,q,K,Z,b,I,T,w,C,E,A,R,G,F,S,M,B,U,x,g,ae,m||0,ee),Q}var boe={kernelName:ti,backendName:"wasm",setupFunc:goe,kernelFunc:yoe},lA;function xoe(e){lA=e.wasm.cwrap(eu,null,["number","number","number","number","number","number","array","number"])}function voe(e){let{backend:t,inputs:n}=e,{params:a,indices:r}=n,[s,i,o,l]=Ux.prepareAndValidate(a,r),u=t.makeOutput(s,a.dtype);if(i===0)return u;let p=r.shape,d=p[p.length-1],c=t.dataIdMap.get(a.dataId).id,h=t.dataIdMap.get(r.dataId).id,m=new Uint8Array(new Int32Array(l).buffer),f=t.dataIdMap.get(u.dataId).id;return lA(c,Et[a.dtype],h,i,d,o,m,f),u}var woe={kernelName:eu,backendName:"wasm",setupFunc:xoe,kernelFunc:voe},uA;function koe(e){uA=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])}function Ioe(e){let{backend:t,inputs:n,attrs:a}=e,{x:r,indices:s}=n,{axis:i,batchDims:o}=a,l=v.parseAxisParam(i,r.shape)[0],u=t.readSync(s.dataId),p=r.shape[l];for(let C=0;C<u.length;++C){let E=u[C];v.assert(E<=p-1&&E>=0,()=>`GatherV2: the index value ${E} is not in [0, ${p-1}]`)}let d=N.segment_util.collectGatherOpShapeInfo(r,s,l,o),c=Wn({inputs:{x:r},attrs:{shape:[d.batchSize,d.outerSize,d.dimSize,d.sliceSize]},backend:t}),h=v.sizeFromShape(s.shape),m=Wn({inputs:{x:s},attrs:{shape:[d.batchSize,h/d.batchSize]},backend:t}),f=[d.batchSize,d.outerSize,h/d.batchSize,d.sliceSize],g=t.makeOutput(f,r.dtype);if(v.sizeFromShape(r.shape)===0)return g;let y=c.shape.length-1,b=t.dataIdMap.get(c.dataId).id,x=t.dataIdMap.get(m.dataId).id,w=t.dataIdMap.get(g.dataId).id,I=new Uint8Array(new Int32Array(v.computeStrides(c.shape)).buffer),T=new Uint8Array(new Int32Array(v.computeStrides(f)).buffer);return uA(b,Et[r.dtype],I,y,x,d.batchSize,T,w),t.disposeData(c.dataId),t.disposeData(m.dataId),g.shape=d.outputShape,g}var Soe={kernelName:Ql,backendName:"wasm",setupFunc:koe,kernelFunc:Ioe},Toe=!1,Noe=dn(tu,Toe,"bool"),Coe=!1,_oe=dn(Di,Coe,"bool"),pA;function Eoe(e){pA=e.wasm.cwrap(Mi,null,["number","number","number","number"])}function Aoe(e){let{inputs:{x:t},attrs:{alpha:n},backend:a}=e,r=a.dataIdMap.get(t.dataId).id,s=a.makeOutput(t.shape,"float32");if(v.sizeFromShape(t.shape)!==0){let i=a.dataIdMap.get(s.dataId).id;pA(r,Et[t.dtype],n,i)}return s}var $oe={kernelName:Mi,backendName:"wasm",setupFunc:Eoe,kernelFunc:Aoe},Foe=!1,Doe=dn(su,Foe,"bool"),Roe=!1,Moe=dn(iu,Roe,"bool"),Poe=an(Pi),Ooe=!1,Loe=dn(lu,Ooe,"bool"),zoe=an(uu),Woe=!1,Boe=dn(pu,Woe,"bool"),Voe=!1,Uoe=dn(nS,Voe,"bool"),cA;function Goe(e){cA=e.wasm.cwrap(Oi,null,["number","number","number","number"])}function Hoe(e){let{backend:t,inputs:n,attrs:a}=e,{reductionIndices:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:p,originalAxes:d,inputWasTransposed:c}=Es(i,r,t);if(c){let b=t.dataIdMap.get(u.dataId).id;l=u,o=b}let h=l.shape.length;N.assertAxesAreInnerMostDims("max",p,h);let[m,f]=N.computeOutAndReduceShapes(l.shape,p),g=v.sizeFromShape(f),y=t.makeOutput(m,i.dtype);if(v.sizeFromShape(l.shape)!==0){let b=t.dataIdMap.get(y.dataId).id;cA(o,Et[i.dtype],g,b)}if(c&&t.disposeData(u.dataId),s){let b=N.expandShapeToKeepDim(y.shape,d);y.shape=b}return y}var joe={kernelName:Oi,backendName:"wasm",setupFunc:Goe,kernelFunc:Hoe},qoe=!1,Koe=dn(Li,qoe),dA;function Xoe(e){dA=e.wasm.cwrap(zi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Yoe(e){let{inputs:t,attrs:n,backend:a}=e,r=t.x,s=a.dataIdMap.get(r.dataId).id;v.assert(r.dtype==="float32",()=>`Error in MaxPool: only float32 input is supported. Got ${r.dtype}.`);let{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,p=N.computePool2DInfo(r.shape,i,o,1,l,u),d=p.filterHeight,c=p.filterWidth,h=p.padInfo.top,m=p.padInfo.right,f=p.padInfo.bottom,g=p.padInfo.left,y=p.dilationHeight,b=p.dilationWidth,x=p.strideHeight,w=p.strideWidth,I=p.inChannels,T=p.outChannels;if(p.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${p.dataFormat}'. Please use 'channelsLast'.`);let C=a.makeOutput(p.outShape,"float32"),E=a.dataIdMap.get(C.dataId).id;return dA(s,r.shape[0],r.shape[1],r.shape[2],d,c,h,m,f,g,y,b,x,w,I,T,E),C}var Zoe={kernelName:zi,backendName:"wasm",setupFunc:Xoe,kernelFunc:Yoe},hA;function Joe(e){hA=e.wasm.cwrap(Wi,null,["number, number, number"])}function Qoe(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Es(i,r,t),m=d;if(h){let w=t.dataIdMap.get(p.dataId).id;w!==o&&(u=p,l=w,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("mean",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=u;u.dtype!=="float32"&&(b=ap({backend:t,inputs:{x:u},attrs:{dtype:"float32"}}),l=t.dataIdMap.get(b.dataId).id);let x=t.makeOutput(f,"float32");if(v.sizeFromShape(u.shape)!==0){let w=t.dataIdMap.get(x.dataId).id;hA(l,y,w)}if(h&&t.disposeData(p.dataId),s){let w=N.expandShapeToKeepDim(x.shape,c);x.shape=w}return u.dtype!=="float32"&&t.disposeData(b.dataId),x}var ele={kernelName:Wi,backendName:"wasm",setupFunc:Joe,kernelFunc:Qoe},mA;function tle(e){mA=e.wasm.cwrap(Bi,null,["number","number","number","number"])}function nle(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Es(i,r,t);if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x)}let m=u.shape.length;N.assertAxesAreInnerMostDims("min",d,m);let[f,g]=N.computeOutAndReduceShapes(u.shape,d),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;mA(l,Et[i.dtype],y,x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var ale={kernelName:Bi,backendName:"wasm",setupFunc:tle,kernelFunc:nle},rle=!1,sle=dn(Vi,rle),Ix;(function(e){e[e.reflect=0]="reflect",e[e.symmetric=1]="symmetric"})(Ix||(Ix={}));var fA;function ile(e){fA=e.wasm.cwrap(Ui,null,["number","array","number","number","array","array","number","number"])}function ole(e){let{inputs:{x:t},backend:n,attrs:{paddings:a,mode:r}}=e,s=a.map((m,f)=>m[0]+t.shape[f]+m[1]),i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),p=a.map(m=>m[0]),d=a.map(m=>m[1]),c=new Uint8Array(new Int32Array(p).buffer),h=new Uint8Array(new Int32Array(d).buffer);return fA(i,u,t.shape.length,Et[t.dtype],c,h,Ix[r],l),o}var lle={kernelName:Ui,backendName:"wasm",kernelFunc:ole,setupFunc:ile},ule=!0,ple=dn(Gi,ule),cle=an(du);function u1(e,t){let n=new Int32Array(e.wasm.HEAPU8.buffer,t,4),a=n[0],r=n[1],s=n[2],i=n[3];return e.wasm._free(t),{pSelectedIndices:a,selectedSize:r,pSelectedScores:s,pValidOutputs:i}}var gA;function dle(e){gA=e.wasm.cwrap(mu,"number",["number","number","number","number","number"])}function hle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i}=a,{boxes:o,scores:l}=n,u=t.dataIdMap.get(o.dataId).id,p=t.dataIdMap.get(l.dataId).id,d=gA(u,p,s,r,i),{pSelectedIndices:c,selectedSize:h,pSelectedScores:m,pValidOutputs:f}=u1(t,d);return t.wasm._free(m),t.wasm._free(f),t.makeOutput([h],"int32",c)}var mle={kernelName:mu,backendName:"wasm",setupFunc:dle,kernelFunc:hle},yA;function fle(e){yA=e.wasm.cwrap(fu,"number",["number","number","number","number","number","bool"])}function gle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:o}=a,{boxes:l,scores:u}=n,p=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,c=yA(p,d,s,r,i,o),{pSelectedIndices:h,selectedSize:m,pSelectedScores:f,pValidOutputs:g}=u1(t,c);t.wasm._free(f);let y=t.makeOutput([m],"int32",h),b=t.makeOutput([],"int32",g);return[y,b]}var yle={kernelName:fu,backendName:"wasm",setupFunc:fle,kernelFunc:gle},bA;function ble(e){bA=e.wasm.cwrap(gu,"number",["number","number","number","number","number","number"])}function xle(e){let{backend:t,inputs:n,attrs:a}=e,{iouThreshold:r,maxOutputSize:s,scoreThreshold:i,softNmsSigma:o}=a,{boxes:l,scores:u}=n,p=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,c=bA(p,d,s,r,i,o),{pSelectedIndices:h,selectedSize:m,pSelectedScores:f,pValidOutputs:g}=u1(t,c);t.wasm._free(g);let y=t.makeOutput([m],"int32",h),b=t.makeOutput([m],"float32",f);return[y,b]}var vle={kernelName:gu,backendName:"wasm",setupFunc:ble,kernelFunc:xle},wle=!1,kle=dn(hu,wle,"bool"),xA;function Ile(e){xA=e.wasm.cwrap(Hi,null,["number","number","number","number","number"])}function Sle(e){let{inputs:t,backend:n,attrs:a}=e,{indices:r}=t,{dtype:s,depth:i,onValue:o,offValue:l}=a,u=n.makeOutput([...r.shape,i],s),p=n.dataIdMap.get(u.dataId).id,d=n.dataIdMap.get(r.dataId).id;return xA(d,i,o,l,p),u}var Tle={kernelName:Hi,backendName:"wasm",setupFunc:Ile,kernelFunc:Sle};function Nle(e){let{inputs:{x:t},backend:n}=e,a=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(a).fill(1),a}var Cle={kernelName:yu,backendName:"wasm",kernelFunc:Nle};function _le(e){let{inputs:t,backend:n,attrs:a}=e,{axis:r}=a;if(t.length===1)return kx({inputs:{input:t[0]},backend:n,attrs:{dim:r}});let s=t[0].shape,i=t[0].dtype;t.forEach(p=>{v.assertShapesMatch(s,p.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===p.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(p=>{let d=kx({inputs:{input:p},backend:n,attrs:{dim:r}});return o.push(d),d}),u=XE({inputs:l,backend:n,attrs:{axis:r}});return o.forEach(p=>n.disposeData(p.dataId)),u}var Ele={kernelName:bu,backendName:"wasm",kernelFunc:_le},vA;function Ale(e){vA=e.wasm.cwrap(ji,null,["number","array","number","number","array","array","number","number"])}function $le(e){let{inputs:{x:t},backend:n,attrs:{paddings:a,constantValue:r}}=e,s=a.map((m,f)=>m[0]+t.shape[f]+m[1]);if(v.sizeFromShape(t.shape)===0)return aA({backend:n,attrs:{shape:s,value:r,dtype:t.dtype}});let i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),p=a.map(m=>m[0]),d=a.map(m=>m[1]),c=new Uint8Array(new Int32Array(p).buffer),h=new Uint8Array(new Int32Array(d).buffer);return vA(i,u,t.shape.length,Et[t.dtype],c,h,r,l),o}var wA={kernelName:ji,backendName:"wasm",kernelFunc:$le,setupFunc:Ale},Fle=!1,Dle=dn(qi,Fle),kA;function Rle(e){kA=e.wasm.cwrap(Ki,null,["number","number","number"])}function Mle(e){let{inputs:t,backend:n}=e,{x:a,alpha:r}=t,s=n.dataIdMap.get(a.dataId).id,i=n.dataIdMap.get(r.dataId).id,o=s,l=a,u=l;l.dtype!=="float32"&&(u=ap({backend:n,inputs:{x:a},attrs:{dtype:"float32"}}),o=n.dataIdMap.get(u.dataId).id);let p=n.makeOutput(a.shape,"float32"),d=n.dataIdMap.get(p.dataId).id;return kA(o,i,d),l.dtype!=="float32"&&n.disposeData(u.dataId),p}var Ple={kernelName:Ki,backendName:"wasm",setupFunc:Rle,kernelFunc:Mle},IA;function Ole(e){IA=e.wasm.cwrap(Xi,null,["number","number","number","number"])}function Lle(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Es(i,r,t),m=d;if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("prod",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;IA(l,y,Et[b.dtype],x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var zle={kernelName:Xi,backendName:"wasm",setupFunc:Ole,kernelFunc:Lle},Wle=e=>{let{backend:t,attrs:n}=e,{start:a,stop:r,step:s,dtype:i}=n,o=W0(a,r,s,i),l=t.makeOutput([o.length],i);return t.typedArrayFromHeap(l).set(o),l},Ble={kernelName:xc,backendName:"wasm",kernelFunc:Wle},Vle=!0,Ule=dn(Ci,Vle),Gle=an(Yi),Hle=an(Qi),SA;function jle(e){SA=e.wasm.cwrap(Ji,null,["number","number","number","number","number","number","number","number","number","number"])}function qle(e){let{backend:t,inputs:n,attrs:a}=e,{images:r}=n,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,[p,d,c,h]=r.shape,m=[p,l,u,h],f=t.dataIdMap.get(r.dataId),g;f.dtype!=="float32"&&(g=ap({backend:t,inputs:{x:r},attrs:{dtype:"float32"}}),f=t.dataIdMap.get(g.dataId));let y=f.id,b=t.makeOutput(m,"float32");if(v.sizeFromShape(r.shape)===0)return b;let x=t.dataIdMap.get(b.dataId).id;return SA(y,p,d,c,h,l,u,s?1:0,i?1:0,x),g!=null&&t.disposeData(g.dataId),b}var Kle={kernelName:Ji,backendName:"wasm",setupFunc:jle,kernelFunc:qle},TA;function Xle(e){TA=e.wasm.cwrap(Zi,null,["number","number","number","number","number","number","number","number","number","number"])}function Yle(e){let{backend:t,inputs:n,attrs:a}=e,{images:r}=n,{alignCorners:s,halfPixelCenters:i,size:o}=a,[l,u]=o,[p,d,c,h]=r.shape,m=[p,l,u,h],f=t.makeOutput(m,"float32");if(v.sizeFromShape(r.shape)===0)return f;let g=t.dataIdMap.get(r.dataId),y;g.dtype!=="float32"&&(y=ap({backend:t,inputs:{x:r},attrs:{dtype:"float32"}}),g=t.dataIdMap.get(y.dataId));let b=g.id,x=t.dataIdMap.get(f.dataId).id;return TA(b,p,d,c,h,l,u,s?1:0,i?1:0,x),y!=null&&t.disposeData(y.dataId),f}var Zle={kernelName:Zi,backendName:"wasm",setupFunc:Xle,kernelFunc:Yle},NA;function Jle(e){NA=e.wasm.cwrap(eo,null,["number","array","number","array","number","number"])}function Qle(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{dims:s}=a,i=v.parseAxisParam(s,r.shape);if(r.shape.length===0)return Qf({inputs:{x:r},backend:n});let o=n.makeOutput(r.shape,r.dtype),l=n.dataIdMap.get(r.dataId).id,u=n.dataIdMap.get(o.dataId).id,p=new Uint8Array(new Int32Array(i).buffer),d=new Uint8Array(new Int32Array(r.shape).buffer);NA(l,p,i.length,d,r.shape.length,u);let c=Wn({inputs:{x:o},attrs:{shape:r.shape},backend:n});return n.disposeData(o.dataId),c}var eue={kernelName:eo,backendName:"wasm",kernelFunc:Qle,setupFunc:Jle},CA;function tue(e){CA=e.wasm.cwrap(Pu,null,["number","number","number","number","number","number","number","number","array","number","number"])}function nue(e){let{inputs:t,backend:n,attrs:a}=e,{image:r}=t,{radians:s,fillValue:i,center:o}=a,l=n.makeOutput(r.shape,r.dtype),u=n.dataIdMap.get(r.dataId).id,p=n.dataIdMap.get(l.dataId).id,[d,c,h,m]=r.shape,[f,g]=N.getImageCenter(o,c,h),y=i===0,b=255,x=typeof i=="number"?[i,i,i,y?0:b]:[...i,b],w=new Uint8Array(new Int32Array(x).buffer);return CA(u,d,c,h,m,s,f,g,w,x.length,p),l}var aue={kernelName:Pu,backendName:"wasm",kernelFunc:nue,setupFunc:tue},rue=an(to),sue=an(no),_A;function iue(e){_A=e.wasm.cwrap(wu,null,["number","number","number","number","number","number","array","number","number"])}function oue(e){let{backend:t,inputs:n,attrs:a}=e,{indices:r,updates:s}=n,{shape:i}=a,o=t.makeOutput(i,s.dtype);if(v.sizeFromShape(i)===0)return o;let{sliceRank:l,numUpdates:u,sliceSize:p,strides:d,outputSize:c}=Gx.calculateShapes(s,r,i),h=t.dataIdMap.get(r.dataId).id,m=t.dataIdMap.get(s.dataId).id,f=new Uint8Array(new Int32Array(d).buffer),g=t.dataIdMap.get(o.dataId).id;return _A(h,m,Et[s.dtype],l,u,p,f,c,g),o}var lue={kernelName:wu,backendName:"wasm",setupFunc:iue,kernelFunc:oue},EA;function uue(e){EA=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}function pue(e){let{inputs:t,backend:n}=e,{condition:a,t:r,e:s}=t,i=n.dataIdMap.get(a.dataId).id,o=n.dataIdMap.get(r.dataId).id,l=n.dataIdMap.get(s.dataId).id,u=n.makeOutput(r.shape,r.dtype),p=n.dataIdMap.get(u.dataId).id,d=a.shape.length,c=r.shape.length,h=d===0||d>1||c===1?1:v.sizeFromShape(r.shape.slice(1));return EA(i,o,l,h,p),u}var cue={kernelName:ku,backendName:"wasm",kernelFunc:pue,setupFunc:uue},AA;function due(e){AA=e.wasm.cwrap(ro,null,["number","number"])}function hue(e){let{backend:t,inputs:{x:n}}=e,a=t.dataIdMap.get(n.dataId).id,r=t.makeOutput(n.shape,n.dtype),s=t.dataIdMap.get(r.dataId).id;return v.sizeFromShape(r.shape)===0||AA(a,s),r}var mue={kernelName:"Sigmoid",backendName:"wasm",setupFunc:due,kernelFunc:hue},fue=an(ao),$A;function gue(e){$A=e.wasm.cwrap(oo,null,["number","number","number","number"])}function yue(e){let{backend:t,inputs:{logits:n},attrs:{dim:a}}=e,r=t.dataIdMap.get(n.dataId).id,s=t.makeOutput(n.shape,n.dtype),i=t.dataIdMap.get(s.dataId).id,o=n.shape[a],l=v.sizeFromShape(n.shape)/o;return v.sizeFromShape(s.shape)===0||$A(r,i,o,l),s}var bue={kernelName:oo,backendName:"wasm",setupFunc:gue,kernelFunc:yue};function xue(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,{blockShape:s,paddings:i}=a,o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<r.shape.length;++g)l.push([0,0]);let u=wA.kernelFunc({inputs:{x:r},backend:n,attrs:{paddings:l,constantValue:0}}),p=N.getReshaped(u.shape,s,o,!1),d=N.getPermuted(p.length,s.length,!1),c=N.getReshapedPermuted(u.shape,s,o,!1),h=Wn({inputs:{x:u},backend:n,attrs:{shape:p}}),m=hs({inputs:{x:h},backend:n,attrs:{perm:d}}),f=Wn({inputs:{x:m},backend:n,attrs:{shape:c}});return n.disposeData(u.dataId),n.disposeData(h.dataId),n.disposeData(m.dataId),f}var vue={kernelName:_u,backendName:"wasm",kernelFunc:xue},FA;function wue(e){FA=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])}function kue(e){let{backend:t,inputs:n}=e,{indices:a,values:r,denseShape:s,defaultValue:i}=n,o=a.shape[0],l=a.shape[1],u=t.readSync(s.dataId)[0],p=[o+u,l],d=t.dataIdMap.get(a.dataId).id,c=t.dataIdMap.get(r.dataId).id,h=t.dataIdMap.get(i.dataId).id,m=t.makeOutput(p,a.dtype),f=t.dataIdMap.get(m.dataId).id,g=t.makeOutput(p.slice(0,1),r.dtype),y=t.dataIdMap.get(g.dataId).id,b=t.makeOutput([u],"bool"),x=t.dataIdMap.get(b.dataId).id,w=t.makeOutput([o],a.dtype),I=t.dataIdMap.get(w.dataId).id,T=t.makeOutput([4],"int32"),C=t.dataIdMap.get(T.dataId).id,E=FA(d,c,Et[r.dtype],o,u,l,h,f,y,x,I,C),A=t.readSync(T.dataId),R;switch(A[0]){case 1:{R=N.getSparseFillEmptyRowsIndicesDenseShapeMismatch(A[1]);break}case 2:{R=N.getSparseFillEmptyRowsNegativeIndexErrorMessage(A[1],A[2]);break}case 3:R=N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(A[1],A[2],A[3]);break;default:R=""}if(t.disposeData(T.dataId),R)throw t.disposeData(m.dataId),t.disposeData(g.dataId),t.disposeData(b.dataId),t.disposeData(w.dataId),new Error(R);let F=m,S=g;return E!==p[0]&&(F=hi({inputs:{x:m},attrs:{begin:0,size:[E,l]},backend:t}),S=hi({inputs:{x:g},attrs:{begin:0,size:E},backend:t}),t.disposeData(m.dataId),t.disposeData(g.dataId)),[F,S,b,w]}var Iue={kernelName:vc,backendName:"wasm",setupFunc:wue,kernelFunc:kue},DA;function Sue(e){DA=e.wasm.cwrap(Au,null,["number","number","number","number","number","number","number"])}function Tue(e){let{backend:t,inputs:n}=e,{inputIndices:a,inputShape:r,newShape:s}=n;if(a.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${a.shape}`);if(r.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(a.dataId).id,o=t.dataIdMap.get(r.dataId).id,l=t.dataIdMap.get(s.dataId).id,u=a.shape[0],p=v.sizeFromShape(s.shape),d=t.makeOutput([u,p],a.dtype),c=t.dataIdMap.get(d.dataId).id,h=t.makeOutput([p],s.dtype),m=t.dataIdMap.get(h.dataId).id,f=t.makeOutput([3],"int32"),g=t.dataIdMap.get(f.dataId).id;DA(i,o,l,u,c,m,g);let y=t.readSync(f.dataId),b;switch(y[0]){case 0:{b=N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(y[1],y[2]);break}case 1:{b=N.getSparseReshapeNegativeOutputDimErrorMessage(y[1],y[2]);break}case 2:b=N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let x=Array.from(t.readSync(r.dataId)),w=Array.from(t.readSync(h.dataId));b=N.getSparseReshapeInputOutputMultipleErrorMessage(x,w);break}case 4:{let x=Array.from(t.readSync(r.dataId)),w=Array.from(t.readSync(h.dataId));b=N.getSparseReshapeInputOutputMismatchErrorMessage(x,w);break}default:b=""}if(t.disposeData(f.dataId),b)throw t.disposeData(d.dataId),t.disposeData(h.dataId),new Error(b);return[d,h]}var Nue={kernelName:Au,backendName:"wasm",setupFunc:Sue,kernelFunc:Tue},RA;function MA(e){RA=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function PA(e,t){let{backend:n,inputs:a}=e,{data:r,indices:s,segmentIds:i}=a,o=s.shape[0],l=n.readSync(i.dataId,o-1,o)[0],u=o>0?l+1:0;if(u<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let p=r.shape.slice();p[0]=u;let d=n.dataIdMap.get(r.dataId).id,c=n.dataIdMap.get(s.dataId).id,h=n.dataIdMap.get(i.dataId).id,m=n.makeOutput(p,r.dtype),f=n.dataIdMap.get(m.dataId).id,g=n.makeOutput([4],"int32"),y=n.dataIdMap.get(g.dataId).id;RA(d,Et[r.dtype],r.shape[0],c,h,f,y,t,0);let b=n.readSync(g.dataId),x;switch(b[0]){case 0:{x=N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{x=N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:x=N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b[1],b[2]);break;case 3:x=N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(b[1],b[2],b[3]);break;default:x=""}if(n.disposeData(g.dataId),x)throw n.disposeData(m.dataId),new Error(x);return m}function Cue(e){return PA(e,!0)}var _ue={kernelName:wc,backendName:"wasm",setupFunc:MA,kernelFunc:Cue};function Eue(e){return PA(e,!1)}var Aue={kernelName:kc,backendName:"wasm",setupFunc:MA,kernelFunc:Eue};function $ue(e){let{inputs:t,attrs:n,backend:a}=e,{x:r}=t,{numOrSizeSplits:s,axis:i}=n,o=v.parseAxisParam(i,r.shape)[0],l=N.prepareSplitSize(r,s,o),u=new Array(r.shape.length).fill(0),p=r.shape.slice();return l.map(d=>{let c=[...p];c[o]=d;let h=hi({inputs:{x:r},attrs:{begin:u,size:c},backend:a});return u[o]+=d,h})}var Fue={kernelName:Eu,backendName:"wasm",kernelFunc:$ue},Due=an(so),Rue=an(Ic),Mue=!0,Pue=dn(lo,Mue),OA;function Oue(e){OA=e.wasm.cwrap(xs,null,["number","number","number","number"])}function Lue(e){let{backend:t,inputs:n,attrs:a}=e,{alpha:r}=a,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=t.makeOutput(s.shape,s.dtype),l=t.dataIdMap.get(o.dataId).id;return OA(i,r,Et[s.dtype],l),o}var zue={kernelName:xs,backendName:"wasm",setupFunc:Oue,kernelFunc:Lue},LA;function Wue(e){LA=e.wasm.cwrap($u,null,["number","array","number","array","array","array","array","array","number","number"])}function Bue(e){let{backend:t,inputs:n,attrs:a}=e,{x:r}=n,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:p,newAxisMask:d,shrinkAxisMask:c}=a,{finalShapeSparse:h,finalShape:m,isIdentity:f,sliceDim0:g,isSimpleSlice:y,begin:b,end:x,strides:w}=jt.sliceInfo(r.shape,s,i,o,l,u,p,d,c),I;if(f)I=Wn({inputs:{x:r},backend:t,attrs:{shape:m}});else if(g||y){v.assert(r.shape.length>=1,()=>`Input must have rank at least 1, got: ${r.shape.length}`);let T=jt.computeOutShape(b,x,w),C=hi({inputs:{x:r},backend:t,attrs:{begin:b,size:T}});I=Wn({inputs:{x:C},backend:t,attrs:{shape:m}}),t.disposeData(C.dataId)}else{let T=t.makeOutput(h,"float32"),C=t.dataIdMap.get(r.dataId).id,E=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),A=new Uint8Array(new Int32Array(b).buffer),R=new Uint8Array(new Int32Array(x).buffer),F=new Uint8Array(new Int32Array(w).buffer),S=new Uint8Array(new Int32Array(h).buffer),M=new Uint8Array(new Int32Array(v.computeStrides(h)).buffer),B=t.dataIdMap.get(T.dataId).id;LA(C,E,r.shape.length,A,R,F,S,M,h.length,B),I=Wn({inputs:{x:T},backend:t,attrs:{shape:m}}),t.disposeData(T.dataId)}return I}var Vue={kernelName:$u,backendName:"wasm",setupFunc:Wue,kernelFunc:Bue};function Uue(e){let{backend:t,inputs:n,attrs:a}=e,{data:r,dataSplits:s}=n,{separator:i,nGramWidths:o,leftPad:l,rightPad:u,padWidth:p,preserveShortSequences:d}=a,c=t.readSync(r.dataId),h=t.readSync(s.dataId),[m,f]=V0(c,h,i,o,l,u,p,d),g=t.makeOutput([m.length],"string"),y=t.dataIdMap.get(g.dataId);y.stringBytes=m;let b=t.makeOutput(s.shape,"int32");return t.typedArrayFromHeap(b).set(f),[g,b]}var Gue={kernelName:Sc,backendName:"wasm",kernelFunc:Uue};function Hue(e){let{backend:t,inputs:n,attrs:a}=e,{input:r,delimiter:s}=n,{skipEmpty:i}=a,o=t.readSync(r.dataId),l=t.readSync(s.dataId),[u,p,d]=U0(o,l[0],i),c=p.length,h=t.makeOutput([c,2],"int32");t.typedArrayFromHeap(h).set(u);let m=t.makeOutput([c],"string"),f=t.dataIdMap.get(m.dataId);f.stringBytes=p;let g=t.makeOutput([2],"int32");return t.typedArrayFromHeap(g).set(d),[h,m,g]}var jue={kernelName:Tc,backendName:"wasm",kernelFunc:Hue};function que(e){let{backend:t,inputs:n,attrs:a}=e,{input:r}=n,{numBuckets:s}=a,i=t.readSync(r.dataId),o=G0(i,s),l=t.makeOutput(r.shape,"int32");return t.typedArrayFromHeap(l).set(o),l}var Kue={kernelName:Nc,backendName:"wasm",kernelFunc:que},Xue=!0,Yue=dn(uo,Xue),zA;function Zue(e){zA=e.wasm.cwrap(io,null,["number","number","number","number"])}function Jue(e){let{backend:t,inputs:n,attrs:a}=e,{axis:r,keepDims:s}=a,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:p,axes:d,originalAxes:c,inputWasTransposed:h}=Es(i,r,t),m=d;if(h){let x=t.dataIdMap.get(p.dataId).id;x!==o&&(u=p,l=x,m=N.getInnerMostAxes(m.length,u.shape.length))}N.assertAxesAreInnerMostDims("sum",m,u.shape.length);let[f,g]=N.computeOutAndReduceShapes(u.shape,m),y=v.sizeFromShape(g),b=t.makeOutput(f,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(b.dataId).id;zA(l,y,Et[b.dtype],x)}if(h&&t.disposeData(p.dataId),s){let x=N.expandShapeToKeepDim(b.shape,c);b.shape=x}return b}var Que={kernelName:io,backendName:"wasm",setupFunc:Zue,kernelFunc:Jue},epe=an(po),tpe=an(co),WA;function npe(e){WA=e.wasm.cwrap(bs,null,["number","array","number","array","number","number"])}function ape(e){let{inputs:t,backend:n,attrs:a}=e,{x:r}=t,s=n.dataIdMap.get(r.dataId).id,{reps:i}=a,o=new Array(r.shape.length);for(let c=0;c<o.length;c++)o[c]=r.shape[c]*i[c];let l=new Uint8Array(new Int32Array(r.shape).buffer),u=new Uint8Array(new Int32Array(o).buffer),p=n.makeOutput(o,r.dtype),d=n.dataIdMap.get(p.dataId).id;return WA(s,l,r.shape.length,u,o.length,Et[p.dtype],d),p}var rpe={kernelName:bs,backendName:"wasm",setupFunc:npe,kernelFunc:ape},BA;function spe(e){BA=e.wasm.cwrap(Fu,null,["number","array","number","number","number","bool","number","number"])}var ipe=({inputs:e,backend:t,attrs:n})=>{let{x:a}=e,{k:r,sorted:s}=n,i=t.dataIdMap.get(a.dataId).id,o=new Uint8Array(new Int32Array(a.shape).buffer),l=a.shape.slice();l[l.length-1]=r;let u=t.makeOutput(l,a.dtype),p=t.dataIdMap.get(u.dataId).id,d=t.makeOutput(l,"int32"),c=t.dataIdMap.get(d.dataId).id;return BA(i,o,a.shape.length,Et[a.dtype],r,s,p,c),[u,d]},ope={kernelName:Fu,backendName:"wasm",setupFunc:spe,kernelFunc:ipe},VA;function lpe(e){VA=e.wasm.cwrap(Du,null,["number","number","bool","number","number","number","number","number","number","array","number","array","number","number","number","number","number"])}function upe(e){let{backend:t,inputs:n,attrs:a}=e,{image:r,transforms:s}=n,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=a,[p,d,c,h]=r.shape,[m,f]=u!=null?u:[d,c],g=[p,m,f,h],y=new Uint8Array(new Int32Array(v.computeStrides(r.shape)).buffer),b=new Uint8Array(new Int32Array(v.computeStrides(g)).buffer),x=t.makeOutput(g,r.dtype),w=t.dataIdMap.get(x.dataId).id,I=t.dataIdMap.get(r.dataId).id,T=t.dataIdMap.get(s.dataId).id,C=i==="nearest"?1:2,E;switch(o){case"constant":E=1;break;case"reflect":E=2;break;case"wrap":E=3;break;case"nearest":E=4;break;default:E=1;break}return VA(I,T,s.shape[0]>1,p,m,f,h,c,d,y,r.shape.length-1,b,g.length-1,C,E,l,w),x}var ppe={kernelName:Du,backendName:"wasm",setupFunc:lpe,kernelFunc:upe};function cpe(e){let{inputs:t,backend:n,attrs:a}=e,{value:r}=t,{axis:s}=a;s<0&&(s+=r.shape.length);let i=r.shape[s],o=r.shape.length,l=new Array(o-1),u=0;for(let h=0;h<o;h++)h!==s&&(l[u++]=r.shape[h]);let p=new Array(i),d=new Array(o).fill(0),c=r.shape.slice();c[s]=1;for(let h=0;h<p.length;h++)d[s]=h,p[h]=hi({inputs:{x:r},attrs:{begin:d,size:c},backend:n});return p.map(({dataId:h,dtype:m})=>({dataId:h,dtype:m,shape:l}))}var dpe={kernelName:Ru,backendName:"wasm",kernelFunc:cpe};function hpe(e){let{inputs:{x:t},backend:n}=e,a=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(a).fill(0),a}var mpe={kernelName:Mu,backendName:"wasm",kernelFunc:hpe},fpe=[Use,Gse,jse,Xse,aie,iie,uie,die,gie,kie,Iie,Sie,Cie,_ie,$ie,Rie,Mie,Pie,zie,Vie,Hie,Kie,Zie,Jie,eoe,toe,noe,aoe,ioe,ooe,uoe,doe,foe,boe,woe,Soe,Noe,_oe,Yse,$oe,Doe,Moe,Poe,Loe,zoe,Boe,Uoe,joe,Koe,Zoe,ele,ale,sle,lle,ple,cle,mle,yle,vle,kle,Tle,Cle,Ele,wA,Dle,Ple,zle,Ble,Ule,Gle,Hle,hie,Kle,Zle,eue,aue,rue,sue,lue,cue,mue,fue,vie,bue,vue,Iue,Nue,_ue,Aue,Fue,Due,Rue,Pue,zue,Vue,Gue,jue,Kue,Yue,Que,epe,tpe,rpe,ope,ppe,eie,dpe,mpe];for(let e of fpe)_c(e);var Sx=H();Sx.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(e){return!1}});Sx.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(Sx.get("IS_NODE"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var PI=ms(MF()),gpe=ms(PF()),OI=ms(OF()),LI=PI.default||PI,ype=OI.default||OI,UA=class extends pc{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(GA),Tx=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new om(this,_a())}write(e,t,n){let a={id:this.dataIdNextNumber++};return this.move(a,e,t,n,1),a}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}move(e,t,n,a,r){let s=this.dataIdNextNumber++;if(a==="string"){let u=t;this.dataIdMap.set(e,{id:s,stringBytes:u,shape:n,dtype:a,memoryOffset:null,refCount:r});return}let i=v.sizeFromShape(n),o=i*v.bytesPerElement(a),l=this.wasm._malloc(o);this.dataIdMap.set(e,{id:s,memoryOffset:l,shape:n,dtype:a,refCount:r}),this.wasm.tfjs.registerTensor(s,i,l),t!=null&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,o),l)}async read(e){return this.readSync(e)}readSync(e,t,n){let{memoryOffset:a,dtype:r,shape:s,stringBytes:i}=this.dataIdMap.get(e);if(r==="string")return(t==null||t===0)&&(n==null||n>=i.length)?i:i.slice(t,n);t=t||0,n=n||v.sizeFromShape(s);let o=v.bytesPerElement(r),l=this.wasm.HEAPU8.slice(a+t*o,a+n*o);return vpe(l.buffer,r)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){let n=this.dataIdMap.get(e);if(n.refCount--,!t&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(e)}return!0}refCount(e){return this.dataIdMap.has(e)?this.dataIdMap.get(e).refCount:0}incRef(e){let t=this.dataIdMap.get(e);t!=null&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,n){let a;if(n==null)a=this.write(null,e,t);else{let r=this.dataIdNextNumber++;a={id:r},this.dataIdMap.set(a,{id:r,memoryOffset:n,shape:e,dtype:t,refCount:1});let s=v.sizeFromShape(e);this.wasm.tfjs.registerTensor(r,s,n)}return{dataId:a,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:n}){let a=this.wasm.HEAPU8.buffer,{memoryOffset:r}=this.dataIdMap.get(n),s=v.sizeFromShape(e);switch(t){case"float32":return new Float32Array(a,r,s);case"int32":return new Int32Array(a,r,s);case"bool":return new Uint8Array(a,r,s);default:throw new Error(`Unknown dtype ${t}`)}}};function bpe(e){return(t,n)=>(v.fetch(e,{credentials:"same-origin"}).then(a=>{a.ok||t.env.a(`failed to load wasm binary file at '${e}'`),a.arrayBuffer().then(r=>{WebAssembly.instantiate(r,t).then(s=>{n(s.instance,s.module)})})}),{})}function zI(e,t,n){if(sm!=null)return sm;let a="tfjs-backend-wasm.wasm";return e&&t?a="tfjs-backend-wasm-threaded-simd.wasm":e&&(a="tfjs-backend-wasm-simd.wasm"),jp!=null&&jp[a]!=null?jp[a]:n+a}async function xpe(){let[e,t]=await Promise.all([H().getAsync("WASM_HAS_SIMD_SUPPORT"),H().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((n,a)=>{let r={};r.locateFile=(o,l)=>{if(o.endsWith(".worker.js")){let u=gpe.wasmWorkerContents.replace(/\n/g,"\\n"),p=new Blob([u],{type:"application/javascript"});return URL.createObjectURL(p)}return o.endsWith(".wasm")?zI(e,t,Up!=null?Up:l):l+o},p1&&(r.instantiateWasm=bpe(zI(e,t,Up!=null?Up:"")));let s=!1;r.onAbort=()=>{s||qp||(qp=!0,a({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"}))};let i;t&&e&&sm==null?(r.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+LI.toString()],{type:"text/javascript"}),i=LI(r)):i=ype(r),i.then(o=>{s=!0,qp=!1;let l=null;o.tfjs={init:o.cwrap("init",null,[]),initWithThreadsCount:o.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:o.cwrap("get_threads_count","number",[]),registerTensor:o.cwrap("register_tensor",null,["number","number","number"]),disposeData:o.cwrap("dispose_data",l,["number"]),dispose:o.cwrap("dispose",l,[])},n({wasm:o})}).catch(a)})}function vpe(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error(`Unknown dtype ${t}`)}}var wpe=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],sm=null,Up=null,jp={},qp=!1,p1=!1;function kpe(e,t=!1){if(Vx("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),qp)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");sm=e,p1=t}function Ipe(e,t=!1){if(qp)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if(typeof e=="string")Up=e;else{jp=e;let n=wpe.filter(a=>jp[a]==null);if(n.length>0)throw new Error(`There were no entries found for the following binaries: ${n.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}p1=t}var GA=-1,Tx=-1;function Spe(e){GA=e}function Tpe(){if(Tx===-1)throw new Error("WASM backend not initialized.");return Tx}var Npe="4.0.0",Cpe=2;Gm("wasm",async()=>{let{wasm:e}=await xpe();return new UA(e)},Cpe);var WI="4.0.0",_pe="4.0.0",Epe="4.0.0",Ape="4.0.0",$pe="4.0.0",Fpe={tfjs:WI,"tfjs-core":WI,"tfjs-converter":_pe,"tfjs-backend-cpu":Epe,"tfjs-backend-webgl":Ape,"tfjs-backend-wasm":$pe};var M1={};rh(M1,{AnchorPosition:()=>N1,DrawBox:()=>md,DrawBoxOptions:()=>ag,DrawFaceLandmarks:()=>mg,DrawFaceLandmarksOptions:()=>hg,DrawTextField:()=>zr,DrawTextFieldOptions:()=>op,drawContour:()=>Pr,drawDetections:()=>Bpe,drawFaceExpressions:()=>jpe,drawFaceLandmarks:()=>Kpe});function Pr(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:a,y:r},s)=>{let i=t[s];e.moveTo(i.x,i.y),e.lineTo(a,r)}),n){let a=t[t.length-1],r=t[0];if(!a||!r)return;e.moveTo(a.x,a.y),e.lineTo(r.x,r.y)}e.stroke()}var m1={};rh(m1,{computeReshapedDimensions:()=>h1,getCenterPoint:()=>No,isDimensions:()=>tg,isEven:()=>eg,isFloat:()=>d1,isTensor:()=>So,isTensor1D:()=>Dpe,isTensor2D:()=>c1,isTensor3D:()=>Or,isTensor4D:()=>va,isValidNumber:()=>Ya,isValidProbablitiy:()=>rp,range:()=>fr,round:()=>To});var yn=class{constructor(t,n){if(!Ya(t)||!Ya(n))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:n})}`);this._width=t,this._height=n}get width(){return this._width}get height(){return this._height}reverse(){return new yn(1/this.width,1/this.height)}};function So(e,t){return e instanceof Te&&e.shape.length===t}function Dpe(e){return So(e,1)}function c1(e){return So(e,2)}function Or(e){return So(e,3)}function va(e){return So(e,4)}function d1(e){return e%1!==0}function eg(e){return e%2===0}function To(e,t=2){let n=10**t;return Math.floor(e*n)/n}function tg(e){return e&&e.width&&e.height}function h1({width:e,height:t},n){let a=n/Math.max(t,e);return new yn(Math.round(e*a),Math.round(t*a))}function No(e){return e.reduce((t,n)=>t.add(n),new De(0,0)).div(new De(e.length,e.length))}function fr(e,t,n){return Array(e).fill(0).map((a,r)=>t+r*n)}function Ya(e){return!!e&&e!==1/0&&e!==-1/0&&!Number.isNaN(e)||e===0}function rp(e){return Ya(e)&&e>=0&&e<=1}var De=class{constructor(t,n){this._x=t,this._y=n}get x(){return this._x}get y(){return this._y}add(t){return new De(this.x+t.x,this.y+t.y)}sub(t){return new De(this.x-t.x,this.y-t.y)}mul(t){return new De(this.x*t.x,this.y*t.y)}div(t){return new De(this.x/t.x,this.y/t.y)}abs(){return new De(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new De(Math.floor(this.x),Math.floor(this.y))}};var st=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(Ya)}static assertIsValidBox(t,n,a=!1){if(!st.isRect(t))throw new Error(`${n} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!a&&(t.width<0||t.height<0))throw new Error(`${n} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,n=!0){let a=t||{},r=[a.left,a.top,a.right,a.bottom].every(Ya),s=[a.x,a.y,a.width,a.height].every(Ya);if(!s&&!r)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(a)}`);let[i,o,l,u]=s?[a.x,a.y,a.width,a.height]:[a.left,a.top,a.right-a.left,a.bottom-a.top];st.assertIsValidBox({x:i,y:o,width:l,height:u},"Box.constructor",n),this._x=i,this._y=o,this._width=l,this._height=u}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new De(this.left,this.top)}get topRight(){return new De(this.right,this.top)}get bottomLeft(){return new De(this.left,this.bottom)}get bottomRight(){return new De(this.right,this.bottom)}round(){let[t,n,a,r]=[this.x,this.y,this.width,this.height].map(s=>Math.round(s));return new st({x:t,y:n,width:a,height:r})}floor(){let[t,n,a,r]=[this.x,this.y,this.width,this.height].map(s=>Math.floor(s));return new st({x:t,y:n,width:a,height:r})}toSquare(){let{x:t,y:n,width:a,height:r}=this,s=Math.abs(a-r);return a<r&&(t-=s/2,a+=s),r<a&&(n-=s/2,r+=s),new st({x:t,y:n,width:a,height:r})}rescale(t){let n=tg(t)?t.width:t,a=tg(t)?t.height:t;return new st({x:this.x*n,y:this.y*a,width:this.width*n,height:this.height*a})}pad(t,n){let[a,r,s,i]=[this.x-t/2,this.y-n/2,this.width+t,this.height+n];return new st({x:a,y:r,width:s,height:i})}clipAtImageBorders(t,n){let{x:a,y:r,right:s,bottom:i}=this,o=Math.max(a,0),l=Math.max(r,0),u=s-o,p=i-l,d=Math.min(u,t-o),c=Math.min(p,n-l);return new st({x:o,y:l,width:d,height:c}).floor()}shift(t,n){let{width:a,height:r}=this,s=this.x+t,i=this.y+n;return new st({x:s,y:i,width:a,height:r})}padAtBorders(t,n){let a=this.width+1,r=this.height+1,s=1,i=1,o=a,l=r,u=this.left,p=this.top,d=this.right,c=this.bottom;return d>n&&(o=-d+n+a,d=n),c>t&&(l=-c+t+r,c=t),u<1&&(l=2-u,u=1),p<1&&(l=2-p,p=1),{dy:i,edy:l,dx:s,edx:o,y:p,ey:c,x:u,ex:d,w:a,h:r}}calibrate(t){return new st({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var Co=class extends st{constructor(t,n,a,r,s=!1){super({left:t,top:n,right:a,bottom:r},s)}};var Lr=class{constructor(t,n,a,r,s){this._imageDims=new yn(s.width,s.height),this._score=t,this._classScore=n,this._className=a,this._box=new st(r).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new st(this._box).rescale(this.imageDims.reverse())}forSize(t,n){return new Lr(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:n})}};var yt=class extends Lr{constructor(t,n,a){super(t,t,"",n,a)}forSize(t,n){let{score:a,relativeBox:r,imageDims:s}=super.forSize(t,n);return new yt(a,r,s)}};function f1(e,t,n=!0){let a=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left)),r=Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top)),s=a*r;return n?s/(e.area+t.area-s):s/Math.min(e.area,t.area)}function g1(e){let t=e.map(o=>o.x),n=e.map(o=>o.y),a=t.reduce((o,l)=>l<o?l:o,1/0),r=n.reduce((o,l)=>l<o?l:o,1/0),s=t.reduce((o,l)=>o<l?l:o,0),i=n.reduce((o,l)=>o<l?l:o,0);return new Co(a,r,s,i)}function y1(e,t,n,a=!0){let r=t.map((i,o)=>({score:i,boxIndex:o})).sort((i,o)=>i.score-o.score).map(i=>i.boxIndex),s=[];for(;r.length>0;){let i=r.pop();s.push(i);let o=r,l=[];for(let u=0;u<o.length;u++){let p=o[u],d=e[i],c=e[p];l.push(f1(d,c,a))}r=r.filter((u,p)=>l[p]<=n)}return s}function Za(e,t){return P(()=>{let[n,a,r]=t,s=gn([...e.shape.slice(0,3),1],n,"float32"),i=gn([...e.shape.slice(0,3),1],a,"float32"),o=gn([...e.shape.slice(0,3),1],r,"float32"),l=Ze([s,i,o],3);return pe(e,l)})}function b1(e,t=!1){return P(()=>{let[n,a]=e.shape.slice(1);if(n===a)return e;let r=Math.abs(n-a),s=Math.round(r*(t?.5:1)),i=n>a?2:1,o=c=>{let h=e.shape.slice();return h[i]=c,gn(h,0,"float32")},l=o(s),u=r-l.shape[i],d=[t&&u?o(u):null,e,l].filter(c=>!!c).map(c=>oe(c,"float32"));return Ze(d,i)})}function Rpe(e){let t=e.slice();for(let n=t.length-1;n>0;n--){let a=Math.floor(Math.random()*(n+1)),r=t[n];t[n]=t[a],t[a]=r}return t}function dd(e){return 1/(1+Math.exp(-e))}function Mpe(e){return Math.log(e/(1-e))}var _o=class extends st{constructor(t,n,a,r,s=!1){super({x:t,y:n,width:a,height:r},s)}};var Ppe=.5,Ope=.43,Lpe=.45,ra=class{constructor(t,n,a=new De(0,0)){let{width:r,height:s}=n;this._imgDims=new yn(r,s),this._shift=a,this._positions=t.map(i=>i.mul(new De(r,s)).add(a))}get shift(){return new De(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new De(this.imageWidth,this.imageHeight)))}forSize(t,n){return new this.constructor(this.relativePositions,{width:t,height:n})}shiftBy(t,n){return new this.constructor(this.relativePositions,this._imgDims,new De(t,n))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,n={}){if(t){let s=t instanceof yt?t.box.floor():new st(t);return this.shiftBy(s.x,s.y).align(null,n)}let{useDlibAlignment:a,minBoxPadding:r}={useDlibAlignment:!1,minBoxPadding:.2,...n};return a?this.alignDlib():this.alignMinBbox(r)}alignDlib(){let t=this.getRefPointsForAlignment(),[n,a,r]=t,s=d=>r.sub(d).magnitude(),i=(s(n)+s(a))/2,o=Math.floor(i/Lpe),l=No(t),u=Math.floor(Math.max(0,l.x-Ppe*o)),p=Math.floor(Math.max(0,l.y-Ope*o));return new _o(u,p,Math.min(o,this.imageWidth+u),Math.min(o,this.imageHeight+p))}alignMinBbox(t){let n=g1(this.positions);return n.pad(n.width*t,n.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var x1=class extends ra{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],No([t[3],t[4]])]}};var Eo=class extends ra{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(No)}};var sp=class{constructor(t,n){this._label=t,this._distance=n}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${To(this.distance)})`:""}`}};var ip=class extends st{constructor(n,a){super(n);this._label=a}static assertIsValidLabeledBox(n,a){if(st.assertIsValidBox(n,a),!Ya(n.label))throw new Error(`${a} - expected property label (${n.label}) to be a number`)}get label(){return this._label}};var gr=class{constructor(t,n){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(n)||n.some(a=>!(a instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=n}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let n=t.descriptors.map(a=>new Float32Array(a));return new gr(t.label,n)}};var v1=class extends ip{constructor(n,a,r,s){super(n,a);this._score=r,this._classScore=s}static assertIsValidPredictedBox(n,a){if(ip.assertIsValidLabeledBox(n,a),!rp(n.score)||!rp(n.classScore))throw new Error(`${a} - expected properties score (${n.score}) and (${n.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function yr(e){return e.detection instanceof yt}function Ao(e,t){return{...e,...{detection:t}}}function w1(){let e=window.fetch;if(!e)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:e,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function hd(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function ng(e){let t="";if(!e&&hd())try{e=Xr("fs")}catch(a){t=a.toString()}return{readFile:e?a=>new Promise((r,s)=>{e.readFile(a,(i,o)=>i?s(i):r(o))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function k1(){let e=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,n=global.Video||global.HTMLVideoElement,a=()=>{if(e)return new e;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},r=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},s=()=>{if(n)return new n;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},i=global.fetch,o=ng();return{Canvas:e||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:a,createImageElement:r,createVideoElement:s,fetch:i,...o}}function I1(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var rn;function zpe(){if(!rn)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return rn}function S1(e){rn=e}function T1(){return I1()?S1(w1()):hd()?S1(k1()):null}function Wpe(e){if(rn||T1(),!rn)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=rn.Canvas,Image:n=rn.Image}=e;rn.Canvas=t,rn.Image=n,rn.createCanvasElement=e.createCanvasElement||(()=>new t),rn.createImageElement=e.createImageElement||(()=>new n),rn.ImageData=e.ImageData||rn.ImageData,rn.Video=e.Video||rn.Video,rn.fetch=e.fetch||rn.fetch,rn.readFile=e.readFile||rn.readFile}var Je={getEnv:zpe,setEnv:S1,initialize:T1,createBrowserEnv:w1,createFileSystem:ng,createNodejsEnv:k1,monkeyPatch:Wpe,isBrowser:I1,isNodejs:hd};T1();function $o(e){return!Je.isNodejs()&&typeof e=="string"?document.getElementById(e):e}function Gn(e){let{Canvas:t,CanvasRenderingContext2D:n}=Je.getEnv();if(e instanceof n)return e;let a=$o(e);if(!(a instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let r=a.getContext("2d");if(!r)throw new Error("resolveContext2d - canvas 2d context is null");return r}var N1=(r=>(r.TOP_LEFT="TOP_LEFT",r.TOP_RIGHT="TOP_RIGHT",r.BOTTOM_LEFT="BOTTOM_LEFT",r.BOTTOM_RIGHT="BOTTOM_RIGHT",r))(N1||{}),op=class{constructor(t={}){let{anchorPosition:n,backgroundColor:a,fontColor:r,fontSize:s,fontStyle:i,padding:o}=t;this.anchorPosition=n||"TOP_LEFT",this.backgroundColor=a||"rgba(0, 0, 0, 0.5)",this.fontColor=r||"rgba(255, 255, 255, 1)",this.fontSize=s||14,this.fontStyle=i||"Georgia",this.padding=o||4}},zr=class{constructor(t,n,a={}){this.text=typeof t=="string"?[t]:t instanceof zr?t.text:t,this.anchor=n,this.options=new op(a)}measureWidth(t){let{padding:n}=this.options;return this.text.map(a=>t.measureText(a).width).reduce((a,r)=>a<r?r:a,0)+2*n}measureHeight(){let{fontSize:t,padding:n}=this.options;return this.text.length*t+2*n}getUpperLeft(t,n){let{anchorPosition:a}=this.options,r=a==="BOTTOM_RIGHT"||a==="TOP_RIGHT",s=a==="BOTTOM_LEFT"||a==="BOTTOM_RIGHT",i=this.measureWidth(t),o=this.measureHeight(),l=r?this.anchor.x-i:this.anchor.x,u=s?this.anchor.y-o:this.anchor.y;if(n){let{width:p,height:d}=n,c=Math.max(Math.min(l,p-i),0),h=Math.max(Math.min(u,d-o),0);return{x:c,y:h}}return{x:l,y:u}}draw(t){let n=$o(t),a=Gn(n),{backgroundColor:r,fontColor:s,fontSize:i,fontStyle:o,padding:l}=this.options;a.font=`${i}px ${o}`;let u=this.measureWidth(a),p=this.measureHeight();a.fillStyle=r;let d=this.getUpperLeft(a,n);a.fillRect(d.x,d.y,u,p),a.fillStyle=s,this.text.forEach((c,h)=>{let m=l+d.x,f=l+d.y+(h+1)*i;a.fillText(c,m,f)})}};var ag=class{constructor(t={}){let{boxColor:n,lineWidth:a,label:r,drawLabelOptions:s}=t;this.boxColor=n||"rgba(0, 0, 255, 1)",this.lineWidth=a||2,this.label=r;let i={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new op({...i,...s})}},md=class{constructor(t,n={}){this.box=new st(t),this.options=new ag(n)}draw(t){let n=Gn(t),{boxColor:a,lineWidth:r}=this.options,{x:s,y:i,width:o,height:l}=this.box;n.strokeStyle=a,n.lineWidth=r,n.strokeRect(s,i,o,l);let{label:u}=this.options;u&&new zr([u],{x:s-r/2,y:i},this.options.drawLabelOptions).draw(t)}};function Bpe(e,t){(Array.isArray(t)?t:[t]).forEach(a=>{let r=a instanceof yt?a.score:yr(a)?a.detection.score:void 0,s=a instanceof yt?a.box:yr(a)?a.detection.box:new st(a),i=r?`${To(r)}`:void 0;new md(s,{label:i}).draw(e)})}function fd(e){let{Image:t,Video:n}=Je.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}function C1(e){return new Promise((t,n)=>{(e instanceof Je.getEnv().Canvas||fd(e))&&t(null);function a(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",r),s.currentTarget.removeEventListener("error",a),n(s))}function r(s){!s.currentTarget||(s.currentTarget.removeEventListener("load",r),s.currentTarget.removeEventListener("error",a),t(s))}e.addEventListener("load",r),e.addEventListener("error",a)})}function _1(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToImage - expected buf to be of type: Blob"));let a=new FileReader;a.onload=()=>{typeof a.result!="string"&&n(new Error("bufferToImage - expected reader.result to be a string, in onload"));let r=Je.getEnv().createImageElement();r.onload=()=>t(r),r.onerror=n,r.src=a.result},a.onerror=n,a.readAsDataURL(e)})}function Fo(e){let{Image:t,Video:n}=Je.getEnv();return e instanceof t?new yn(e.naturalWidth,e.naturalHeight):e instanceof n?new yn(e.videoWidth,e.videoHeight):new yn(e.width,e.height)}function Do({width:e,height:t}){let{createCanvasElement:n}=Je.getEnv(),a=n();return a.width=e,a.height=t,a}function gd(e,t){let{ImageData:n}=Je.getEnv();if(!(e instanceof n)&&!fd(e))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:a,height:r}=t||Fo(e),s=Do({width:a,height:r});return e instanceof n?Gn(s).putImageData(e,0,0):Gn(s).drawImage(e,0,0,a,r),s}async function E1(e,t){let n=t||Je.getEnv().createCanvasElement(),[a,r,s]=e.shape.slice(va(e)?1:0),i=P(()=>e.as3D(a,r,s).toInt());return await ho.toPixels(i,n),i.dispose(),n}function rg(e){let{Image:t,Canvas:n,Video:a}=Je.getEnv();return e instanceof t||e instanceof n||e instanceof a}function A1(e,t,n=!1){let{Image:a,Canvas:r}=Je.getEnv();if(!(e instanceof a||e instanceof r))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Do({width:1,height:1});let s=Fo(e),i=t/Math.max(s.height,s.width),o=i*s.width,l=i*s.height,u=Do({width:t,height:t}),p=e instanceof r?e:gd(e),d=Math.abs(o-l)/2,c=n&&o<l?d:0,h=n&&l<o?d:0;return p.width>0&&p.height>0&&Gn(u).drawImage(p,c,h,o,l),u}var br=class{constructor(t,n=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=n,this._batchSize=t.length,t.forEach((a,r)=>{if(Or(a)){this._imageTensors[r]=a,this._inputDimensions[r]=a.shape;return}if(va(a)){let i=a.shape[0];if(i!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${i} passed, but not supported in input array`);this._imageTensors[r]=a,this._inputDimensions[r]=a.shape.slice(1);return}let s=a instanceof Je.getEnv().Canvas?a:gd(a);this._canvases[r]=s,this._inputDimensions[r]=[s.height,s.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return fr(this.batchSize,0,1).map((t,n)=>this.getReshapedInputDimensions(n))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let n=this.getInputWidth(t),a=this.getInputHeight(t);return h1({width:n,height:a},this.inputSize)}toBatchTensor(t,n=!0){return this._inputSize=t,P(()=>{let a=fr(this.batchSize,0,1).map(s=>{let i=this.getInput(s);if(i instanceof Te){let o=va(i)?i:Zt(i);return o=b1(o,n),(o.shape[1]!==t||o.shape[2]!==t)&&(o=za.resizeBilinear(o,[t,t],!1,!1)),o.as3D(t,t,3)}if(i instanceof Je.getEnv().Canvas)return ho.fromPixels(A1(i,t,n));throw new Error(`toBatchTensor - at batchIdx ${s}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${i}`)});return Ft(a.map(s=>oe(s,"float32"))).as4D(this.batchSize,t,t,3)})}};async function bt(e){if(e instanceof br)return e;let t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");let n=r=>Array.isArray(e)?` at input index ${r}:`:"",a=t.map($o);return a.forEach((r,s)=>{if(!rg(r)&&!Or(r)&&!va(r))throw typeof t[s]=="string"?new Error(`toNetInput -${n(s)} string passed, but could not resolve HTMLElement for element id ${t[s]}`):new Error(`toNetInput -${n(s)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(va(r)){let i=r.shape[0];if(i!==1)throw new Error(`toNetInput -${n(s)} tf.Tensor4D with batchSize ${i} passed, but not supported in input array`)}}),await Promise.all(a.map(r=>rg(r)&&C1(r))),new br(a,Array.isArray(e))}async function lp(e,t){let{Canvas:n}=Je.getEnv(),a=e;if(!(e instanceof n)){let i=await bt(e);if(i.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let o=i.getInput(0);a=o instanceof n?o:await E1(o)}let r=Gn(a);return t.map(i=>i instanceof yt?i.forSize(a.width,a.height).box.floor():i).map(i=>i.clipAtImageBorders(a.width,a.height)).map(({x:i,y:o,width:l,height:u})=>{let p=Do({width:l,height:u});return l>0&&u>0&&Gn(p).putImageData(r.getImageData(i,o,l,u),0,0),p})}async function up(e,t){if(!Or(e)&&!va(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(va(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return P(()=>{let[n,a,r]=e.shape.slice(va(e)?1:0);return t.map(o=>o instanceof yt?o.forSize(a,n).box:o).map(o=>o.clipAtImageBorders(a,n)).filter(o=>o.width>0&&o.height>0).map(({x:o,y:l,width:u,height:p})=>fo(e.as3D(n,a,r),[l,o,0],[p,u,r]))})}async function Wr(e,t){let{fetch:n}=Je.getEnv(),a=await n(e,t);if(!(a.status<400))throw new Error(`failed to fetch: (${a.status}) ${a.statusText}, from url: ${a.url}`);return a}async function Vpe(e){let t=await Wr(e),n=await t.blob();if(!n.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${n.type}, for url: ${t.url}`);return _1(n)}async function $1(e){return(await Wr(e)).json()}async function Upe(e){return new Float32Array(await(await Wr(e)).arrayBuffer())}function HA(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToVideo - expected buf to be of type: Blob"));let a=Je.getEnv().createVideoElement();a.oncanplay=()=>t(a),a.onerror=n,a.playsInline=!0,a.muted=!0,a.src=URL.createObjectURL(e),a.play()})}async function Gpe(e){let t=await Wr(e),n=await t.blob();if(!n.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${n.type}, for url: ${t.url}`);return HA(n)}function sg(e,t){let n=`${t}-weights_manifest.json`;if(!e)return{modelBaseUri:"",manifestUri:n};if(e==="/")return{modelBaseUri:"/",manifestUri:`/${n}`};let a=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"";e=e.replace(a,"");let r=e.split("/").filter(o=>o),s=e.endsWith(".json")?r[r.length-1]:n,i=a+(e.endsWith(".json")?r.slice(0,r.length-1):r).join("/");return i=e.startsWith("/")?`/${i}`:i,{modelBaseUri:i,manifestUri:i==="/"?`/${s}`:`${i}/${s}`}}async function F1(e,t){let{manifestUri:n,modelBaseUri:a}=sg(e,t),r=await $1(n);return Ut.loadWeights(r,a)}function Hpe(e,t,n=!1){let{width:a,height:r}=n?Fo(t):t;return e.width=a,e.height=r,{width:a,height:r}}var sn=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:n,objProp:a}=this.traversePropertyPath(t);return n[a]}reassignParamFromPath(t,n){let{obj:a,objProp:r}=this.traversePropertyPath(t);a[r].dispose(),a[r]=n}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof is)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof is))}variable(){this.getFrozenParams().forEach(({path:t,tensor:n})=>{this.reassignParamFromPath(t,n.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:n})=>{let a=In(n.dataSync());n.dispose(),this.reassignParamFromPath(t,a)})}dispose(t=!0){this.getParamList().forEach(n=>{if(t&&n.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${n.path}`);n.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,n)=>t.concat(n)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let n=await F1(t,this.getDefaultModelName());this.loadFromWeightMap(n)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:n}=Je.getEnv(),{manifestUri:a,modelBaseUri:r}=sg(t,this.getDefaultModelName()),s=u=>Promise.all(u.map(p=>n(p).then(d=>d.buffer))),i=Ut.weightsLoaderFactory(s),o=JSON.parse((await n(a)).toString()),l=await i(o,r);this.loadFromWeightMap(l)}loadFromWeightMap(t){let{paramMappings:n,params:a}=this.extractParamsFromWeightMap(t);this._paramMappings=n,this._params=a}extractWeights(t){let{paramMappings:n,params:a}=this.extractParams(t);this._paramMappings=n,this._params=a}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let n=t.split("/").reduce((s,i)=>{if(!s.nextObj.hasOwnProperty(i))throw new Error(`traversePropertyPath - object does not have property ${i}, for path ${t}`);return{obj:s.nextObj,objProp:i,nextObj:s.nextObj[i]}},{nextObj:this.params}),{obj:a,objProp:r}=n;if(!a||!r||!(a[r]instanceof Te))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:a,objProp:r}}};function Hn(e,t,n){return P(()=>{let a=Ts(e,t.depthwise_filter,t.pointwise_filter,n,"same");return a=Y(a,t.bias),a})}function ig(e,t,n=!1){return P(()=>{let a=Xe(n?Y($t(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):Hn(e,t.conv0,[2,2])),r=Hn(a,t.conv1,[1,1]),s=Xe(Y(a,r)),i=Hn(s,t.conv2,[1,1]);return Xe(Y(a,Y(r,i)))})}function yd(e,t,n=!1,a=!0){return P(()=>{let r=Xe(n?Y($t(e,t.conv0.filters,a?[2,2]:[1,1],"same"),t.conv0.bias):Hn(e,t.conv0,a?[2,2]:[1,1])),s=Hn(r,t.conv1,[1,1]),i=Xe(Y(r,s)),o=Hn(i,t.conv2,[1,1]),l=Xe(Y(r,Y(s,o))),u=Hn(l,t.conv3,[1,1]);return Xe(Y(r,Y(s,Y(o,u))))})}function Ro(e,t,n="same",a=!1){return P(()=>{let r=Y($t(e,t.filters,[1,1],n),t.bias);return a?Xe(r):r})}function An(e,t){Object.keys(e).forEach(n=>{t.some(a=>a.originalPath===n)||e[n].dispose()})}function pp(e,t){return(n,a,r,s)=>{let i=Da(e(n*a*r*r),[r,r,n,a]),o=Ke(e(a));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/bias`}),{filters:i,bias:o}}}function og(e,t){return(n,a,r)=>{let s=Aa(e(n*a),[n,a]),i=Ke(e(a));return t.push({paramPath:`${r}/weights`},{paramPath:`${r}/bias`}),{weights:s,bias:i}}}var bd=class{constructor(t,n,a){this.depthwise_filter=t;this.pointwise_filter=n;this.bias=a}};function cp(e,t){return(n,a,r)=>{let s=Da(e(9*n),[3,3,n,1]),i=Da(e(n*a),[1,1,n,a]),o=Ke(e(a));return t.push({paramPath:`${r}/depthwise_filter`},{paramPath:`${r}/pointwise_filter`},{paramPath:`${r}/bias`}),new bd(s,i,o)}}function dp(e){return t=>{let n=e(`${t}/depthwise_filter`,4),a=e(`${t}/pointwise_filter`,4),r=e(`${t}/bias`,1);return new bd(n,a,r)}}function sa(e,t){return(n,a,r)=>{let s=e[n];if(!So(s,a))throw new Error(`expected weightMap[${n}] to be a Tensor${a}D, instead have ${s}`);return t.push({originalPath:n,paramPath:r||n}),s}}function $n(e){let t=e;function n(r){let s=t.slice(0,r);return t=t.slice(r),s}function a(){return t}return{extractWeights:n,getRemainingWeights:a}}function lg(e,t){let n=pp(e,t),a=cp(e,t);function r(i,o,l,u=!1){let p=u?n(i,o,3,`${l}/conv0`):a(i,o,`${l}/conv0`),d=a(o,o,`${l}/conv1`),c=a(o,o,`${l}/conv2`);return{conv0:p,conv1:d,conv2:c}}function s(i,o,l,u=!1){let{conv0:p,conv1:d,conv2:c}=r(i,o,l,u),h=a(o,o,`${l}/conv3`);return{conv0:p,conv1:d,conv2:c,conv3:h}}return{extractDenseBlock3Params:r,extractDenseBlock4Params:s}}function jA(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractDenseBlock4Params:r}=lg(n,t),s=r(3,32,"dense0",!0),i=r(32,64,"dense1"),o=r(64,128,"dense2"),l=r(128,256,"dense3");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o,dense3:l}}}function ug(e){return t=>{let n=e(`${t}/filters`,4),a=e(`${t}/bias`,1);return{filters:n,bias:a}}}function pg(e,t){let n=sa(e,t),a=ug(n),r=dp(n);function s(o,l=!1){let u=l?a(`${o}/conv0`):r(`${o}/conv0`),p=r(`${o}/conv1`),d=r(`${o}/conv2`);return{conv0:u,conv1:p,conv2:d}}function i(o,l=!1){let u=l?a(`${o}/conv0`):r(`${o}/conv0`),p=r(`${o}/conv1`),d=r(`${o}/conv2`),c=r(`${o}/conv3`);return{conv0:u,conv1:p,conv2:d,conv3:c}}return{extractDenseBlock3Params:s,extractDenseBlock4Params:i}}function qA(e){let t=[],{extractDenseBlock4Params:n}=pg(e,t),a={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return An(e,t),{params:a,paramMappings:t}}var hp=class extends sn{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceFeatureExtractor - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(112,!0),"float32"),s=Za(a,[122.782,117.001,104.298]).div(255),i=yd(s,n.dense0,!0);return i=yd(i,n.dense1),i=yd(i,n.dense2),i=yd(i,n.dense3),i=ya(i,[7,7],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await bt(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return qA(t)}extractParams(t){return jA(t)}};function xd(e,t){return P(()=>Y(Fe(e,t.weights),t.bias))}function KA(e,t,n){let a=[],{extractWeights:r,getRemainingWeights:s}=$n(e),o=og(r,a)(t,n,"fc");if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{paramMappings:a,params:{fc:o}}}function XA(e){let t=[],n=sa(e,t);function a(s){let i=n(`${s}/weights`,2),o=n(`${s}/bias`,1);return{weights:i,bias:o}}let r={fc:a("fc")};return An(e,t),{params:r,paramMappings:t}}function cg(e){let t={},n={};return Object.keys(e).forEach(a=>{let r=a.startsWith("fc")?n:t;r[a]=e[a]}),{featureExtractorMap:t,classifierMap:n}}var mp=class extends sn{constructor(n,a){super(n);this._faceFeatureExtractor=a}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:a}=this;if(!a)throw new Error(`${this._name} - load model before inference`);return P(()=>{let r=n instanceof br?this.faceFeatureExtractor.forwardInput(n):n;return xd(r.as2D(r.shape[0],-1),a.fc)})}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:a,paramMappings:r}=this.extractClassifierParams(n);this._params=a,this._paramMappings=r}extractClassifierParams(n){return KA(n,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(n){let{featureExtractorMap:a,classifierMap:r}=cg(n);return this.faceFeatureExtractor.loadFromWeightMap(a),XA(r)}extractParams(n){let a=this.getClassifierChannelsIn(),r=this.getClassifierChannelsOut(),s=r*a+r,i=n.slice(0,n.length-s),o=n.slice(n.length-s);return this.faceFeatureExtractor.extractWeights(i),this.extractClassifierParams(o)}};var D1=["neutral","happy","sad","angry","fearful","disgusted","surprised"],Br=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);D1.forEach((n,a)=>{this[n]=t[a]})}asSortedArray(){return D1.map(t=>({expression:t,probability:this[t]})).sort((t,n)=>n.probability-t.probability)}};var vd=class extends mp{constructor(t=new hp){super("FaceExpressionNet",t)}forwardInput(t){return P(()=>Ka(this.runNet(t)))}async forward(t){return this.forwardInput(await bt(t))}async predictExpressions(t){let n=await bt(t),a=await this.forwardInput(n),r=await Promise.all(ct(a).map(async i=>{let o=i.dataSync();return i.dispose(),o}));a.dispose();let s=r.map(i=>new Br(i));return n.isBatchInput?s:s[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function R1(e){return e.expressions instanceof Br}function dg(e,t){return{...e,...{expressions:t}}}function jpe(e,t,n=.1,a){(Array.isArray(t)?t:[t]).forEach(s=>{let i=s instanceof Br?s:R1(s)?s.expressions:void 0;if(!i)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let l=i.asSortedArray().filter(d=>d.probability>n),u=yr(s)?s.detection.box.bottomLeft:a||new De(0,0);new zr(l.map(d=>`${d.expression} (${To(d.probability)})`),u).draw(e)})}function Mo(e){return yr(e)&&e.landmarks instanceof ra&&e.unshiftedLandmarks instanceof ra&&e.alignedRect instanceof yt}function qpe(e){let t=l=>l*180/Math.PI,n=(l,u)=>Math.sqrt((l._x-u._x)**2+(l._y-u._y)**2),a={roll:void 0,pitch:void 0,yaw:void 0},r=(l,u,p)=>{let d=Math.floor(l._x-u._x),c=Math.floor(u._x-p._x);return d-c},s=(l,u)=>{let p=Math.hypot(u._x-l._x,u._y-l._y),d=u._y-l._y,c=Math.asin(d/p),h=t(c),m=Math.floor(90-h),f=u._x-l._x<0?-1:1;return m*f},i=(l,u,p)=>{let d=n(l,p),c={_x:(l._x+p._x)/2,_y:(l._y+p._y)/2},h=n(u,c),m=Math.atan(h/d),f=Math.floor(t(m)),g=c._y-u._y<0?-1:1;return f*g};if(!e||!e._positions||e._positions.length!==68)return a;let o=e._positions;return a.roll=s(o[27],o[66]),a.pitch=i(o[14],o[30],o[2]),a.yaw=r(o[14],o[33],o[2]),a}function fp(e,t){let{box:n}=e.detection,a=t.shiftBy(n.x,n.y),r=a.align(),{imageDims:s}=e.detection,i=new yt(e.detection.score,r.rescale(s.reverse()),s),o=qpe(t);return{...e,...{landmarks:a,unshiftedLandmarks:t,alignedRect:i,angle:o}}}var hg=class{constructor(t={}){let{drawLines:n=!0,drawPoints:a=!0,lineWidth:r,lineColor:s,pointSize:i,pointColor:o}=t;this.drawLines=n,this.drawPoints=a,this.lineWidth=r||1,this.pointSize=i||2,this.lineColor=s||"rgba(0, 255, 255, 1)",this.pointColor=o||"rgba(255, 0, 255, 1)"}},mg=class{constructor(t,n={}){this.faceLandmarks=t,this.options=new hg(n)}draw(t){let n=Gn(t),{drawLines:a,drawPoints:r,lineWidth:s,lineColor:i,pointSize:o,pointColor:l}=this.options;if(a&&this.faceLandmarks instanceof Eo&&(n.strokeStyle=i,n.lineWidth=s,Pr(n,this.faceLandmarks.getJawOutline()),Pr(n,this.faceLandmarks.getLeftEyeBrow()),Pr(n,this.faceLandmarks.getRightEyeBrow()),Pr(n,this.faceLandmarks.getNose()),Pr(n,this.faceLandmarks.getLeftEye(),!0),Pr(n,this.faceLandmarks.getRightEye(),!0),Pr(n,this.faceLandmarks.getMouth(),!0)),r){n.strokeStyle=l,n.fillStyle=l;let u=p=>{n.beginPath(),n.arc(p.x,p.y,o,0,2*Math.PI),n.fill()};this.faceLandmarks.positions.forEach(u)}}};function Kpe(e,t){(Array.isArray(t)?t:[t]).forEach(a=>{let r=a instanceof ra?a:Mo(a)?a.landmarks:void 0;if(!r)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new mg(r).draw(e)})}var YA="1.7.6";function Zpe(e,t){let n=pp(e,t),a=cp(e,t);function r(i,o,l){let u=a(i,o,`${l}/separable_conv0`),p=a(o,o,`${l}/separable_conv1`),d=n(i,o,1,`${l}/expansion_conv`);return{separable_conv0:u,separable_conv1:p,expansion_conv:d}}function s(i,o){let l=a(i,i,`${o}/separable_conv0`),u=a(i,i,`${o}/separable_conv1`),p=a(i,i,`${o}/separable_conv2`);return{separable_conv0:l,separable_conv1:u,separable_conv2:p}}return{extractConvParams:n,extractSeparableConvParams:a,extractReductionBlockParams:r,extractMainBlockParams:s}}function ZA(e,t){let n=[],{extractWeights:a,getRemainingWeights:r}=$n(e),{extractConvParams:s,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:l}=Zpe(a,n),u=s(3,32,3,"entry_flow/conv_in"),p=o(32,64,"entry_flow/reduction_block_0"),d=o(64,128,"entry_flow/reduction_block_1"),c={conv_in:u,reduction_block_0:p,reduction_block_1:d},h={};fr(t,0,1).forEach(y=>{h[`main_block_${y}`]=l(128,`middle_flow/main_block_${y}`)});let m=o(128,256,"exit_flow/reduction_block"),f=i(256,512,"exit_flow/separable_conv"),g={reduction_block:m,separable_conv:f};if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:n,params:{entry_flow:c,middle_flow:h,exit_flow:g}}}function Jpe(e,t){let n=sa(e,t),a=ug(n),r=dp(n);function s(o){let l=r(`${o}/separable_conv0`),u=r(`${o}/separable_conv1`),p=a(`${o}/expansion_conv`);return{separable_conv0:l,separable_conv1:u,expansion_conv:p}}function i(o){let l=r(`${o}/separable_conv0`),u=r(`${o}/separable_conv1`),p=r(`${o}/separable_conv2`);return{separable_conv0:l,separable_conv1:u,separable_conv2:p}}return{extractConvParams:a,extractSeparableConvParams:r,extractReductionBlockParams:s,extractMainBlockParams:i}}function JA(e,t){let n=[],{extractConvParams:a,extractSeparableConvParams:r,extractReductionBlockParams:s,extractMainBlockParams:i}=Jpe(e,n),o=a("entry_flow/conv_in"),l=s("entry_flow/reduction_block_0"),u=s("entry_flow/reduction_block_1"),p={conv_in:o,reduction_block_0:l,reduction_block_1:u},d={};fr(t,0,1).forEach(f=>{d[`main_block_${f}`]=i(`middle_flow/main_block_${f}`)});let c=s("exit_flow/reduction_block"),h=r("exit_flow/separable_conv"),m={reduction_block:c,separable_conv:h};return An(e,n),{params:{entry_flow:p,middle_flow:d,exit_flow:m},paramMappings:n}}function QA(e,t,n){return Y($t(e,t.filters,n,"same"),t.bias)}function P1(e,t,n=!0){let a=n?Xe(e):e;return a=Hn(a,t.separable_conv0,[1,1]),a=Hn(Xe(a),t.separable_conv1,[1,1]),a=Dt(a,[3,3],[2,2],"same"),a=Y(a,QA(e,t.expansion_conv,[2,2])),a}function Qpe(e,t){let n=Hn(Xe(e),t.separable_conv0,[1,1]);return n=Hn(Xe(n),t.separable_conv1,[1,1]),n=Hn(Xe(n),t.separable_conv2,[1,1]),n=Y(n,e),n}var fg=class extends sn{constructor(n){super("TinyXception");this._numMainBlocks=n}forwardInput(n){let{params:a}=this;if(!a)throw new Error("TinyXception - load model before inference");return P(()=>{let r=oe(n.toBatchTensor(112,!0),"float32"),i=Za(r,[122.782,117.001,104.298]).div(255),o=Xe(QA(i,a.entry_flow.conv_in,[2,2]));return o=P1(o,a.entry_flow.reduction_block_0,!1),o=P1(o,a.entry_flow.reduction_block_1),fr(this._numMainBlocks,0,1).forEach(l=>{o=Qpe(o,a.middle_flow[`main_block_${l}`])}),o=P1(o,a.exit_flow.reduction_block),o=Xe(Hn(o,a.exit_flow.separable_conv,[1,1])),o})}async forward(n){return this.forwardInput(await bt(n))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(n){return JA(n,this._numMainBlocks)}extractParams(n){return ZA(n,this._numMainBlocks)}};function e$(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),r=og(n,t),s=r(512,1,"fc/age"),i=r(512,2,"fc/gender");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{fc:{age:s,gender:i}}}}function t$(e){let t=[],n=sa(e,t);function a(s){let i=n(`${s}/weights`,2),o=n(`${s}/bias`,1);return{weights:i,bias:o}}let r={fc:{age:a("fc/age"),gender:a("fc/gender")}};return An(e,t),{params:r,paramMappings:t}}var gg=(n=>(n.FEMALE="female",n.MALE="male",n))(gg||{});var wd=class extends sn{constructor(n=new fg(2)){super("AgeGenderNet");this._faceFeatureExtractor=n}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:a}=this;if(!a)throw new Error(`${this._name} - load model before inference`);return P(()=>{let r=n instanceof br?this.faceFeatureExtractor.forwardInput(n):n,s=ya(r,[7,7],[2,2],"valid").as2D(r.shape[0],-1),i=xd(s,a.fc.age).as1D(),o=xd(s,a.fc.gender);return{age:i,gender:o}})}forwardInput(n){return P(()=>{let{age:a,gender:r}=this.runNet(n);return{age:a,gender:Ka(r)}})}async forward(n){return this.forwardInput(await bt(n))}async predictAgeAndGender(n){let a=await bt(n),r=await this.forwardInput(a),s=ct(r.age),i=ct(r.gender),o=s.map((u,p)=>({ageTensor:u,genderTensor:i[p]})),l=await Promise.all(o.map(async({ageTensor:u,genderTensor:p})=>{let d=u.dataSync()[0],c=p.dataSync()[0],h=c>.5,m=h?"male":"female",f=h?c:1-c;return u.dispose(),p.dispose(),{age:d,gender:m,genderProbability:f}}));return r.age.dispose(),r.gender.dispose(),a.isBatchInput?l:l[0]}getDefaultModelName(){return"age_gender_model"}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:a,paramMappings:r}=this.extractClassifierParams(n);this._params=a,this._paramMappings=r}extractClassifierParams(n){return e$(n)}extractParamsFromWeightMap(n){let{featureExtractorMap:a,classifierMap:r}=cg(n);return this.faceFeatureExtractor.loadFromWeightMap(a),t$(r)}extractParams(n){let r=n.slice(0,n.length-1539),s=n.slice(n.length-1539);return this.faceFeatureExtractor.extractWeights(r),this.extractClassifierParams(s)}};var gp=class extends mp{postProcess(t,n,a){let r=a.map(({width:i,height:o})=>{let l=n/Math.max(o,i);return{width:i*l,height:o*l}}),s=r.length;return P(()=>{let i=(d,c)=>Ft([gn([68],d,"float32"),gn([68],c,"float32")],1).as2D(1,136).as1D(),o=(d,c)=>{let{width:h,height:m}=r[d];return c(h,m)?Math.abs(h-m)/2:0},l=d=>o(d,(c,h)=>c<h),u=d=>o(d,(c,h)=>h<c);return t.mul(gn([s,136],n,"float32")).sub(Ft(Array.from(Array(s),(d,c)=>i(l(c),u(c))))).div(Ft(Array.from(Array(s),(d,c)=>i(r[c].width,r[c].height))))})}forwardInput(t){return P(()=>{let n=this.runNet(t);return this.postProcess(n,t.inputSize,t.inputDimensions.map(([a,r])=>({height:a,width:r})))})}async forward(t){return this.forwardInput(await bt(t))}async detectLandmarks(t){let n=await bt(t),a=P(()=>ct(this.forwardInput(n))),r=await Promise.all(a.map(async(s,i)=>{let o=Array.from(s.dataSync()),l=o.filter((p,d)=>eg(d)),u=o.filter((p,d)=>!eg(d));return new Eo(Array(68).fill(0).map((p,d)=>new De(l[d],u[d])),{height:n.getInputHeight(i),width:n.getInputWidth(i)})}));return a.forEach(s=>s.dispose()),n.isBatchInput?r:r[0]}getClassifierChannelsOut(){return 136}};var Po=class extends gp{constructor(t=new hp){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};function n$(e){let t=[],{extractDenseBlock3Params:n}=pg(e,t),a={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return An(e,t),{params:a,paramMappings:t}}function a$(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractDenseBlock3Params:r}=lg(n,t),s=r(3,32,"dense0",!0),i=r(32,64,"dense1"),o=r(64,128,"dense2");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o}}}var yg=class extends sn{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("TinyFaceFeatureExtractor - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(112,!0),"float32"),s=Za(a,[122.782,117.001,104.298]).div(255),i=ig(s,n.dense0,!0);return i=ig(i,n.dense1),i=ig(i,n.dense2),i=ya(i,[14,14],[2,2],"valid"),i})}async forward(t){return this.forwardInput(await bt(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return n$(t)}extractParams(t){return a$(t)}};var kd=class extends gp{constructor(t=new yg){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var O1=class extends Po{};function r$(e,t){return Y(z(e,t.weights),t.biases)}function L1(e,t,n,a,r="same"){let{filters:s,bias:i}=t.conv,o=$t(e,s,n,r);return o=Y(o,i),o=r$(o,t.scale),a?Xe(o):o}function s$(e,t){return L1(e,t,[1,1],!0)}function z1(e,t){return L1(e,t,[1,1],!1)}function bg(e,t){return L1(e,t,[2,2],!0,"valid")}function ece(e,t){function n(o,l,u){let p=e(o),d=p.length/(l*u*u);if(d1(d))throw new Error(`depth has to be an integer: ${d}, weights.length: ${p.length}, numFilters: ${l}, filterSize: ${u}`);return P(()=>Ee(Da(p,[l,d,u,u]),[2,3,1,0]))}function a(o,l,u,p){let d=n(o,l,u),c=Ke(e(l));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/bias`}),{filters:d,bias:c}}function r(o,l){let u=Ke(e(o)),p=Ke(e(o));return t.push({paramPath:`${l}/weights`},{paramPath:`${l}/biases`}),{weights:u,biases:p}}function s(o,l,u,p){let d=a(o,l,u,`${p}/conv`),c=r(l,`${p}/scale`);return{conv:d,scale:c}}function i(o,l,u,p,d=!1){let c=s((d?.5:1)*o,l,u,`${p}/conv1`),h=s(o,l,u,`${p}/conv2`);return{conv1:c,conv2:h}}return{extractConvLayerParams:s,extractResidualLayerParams:i}}function i$(e){let{extractWeights:t,getRemainingWeights:n}=$n(e),a=[],{extractConvLayerParams:r,extractResidualLayerParams:s}=ece(t,a),i=r(4704,32,7,"conv32_down"),o=s(9216,32,3,"conv32_1"),l=s(9216,32,3,"conv32_2"),u=s(9216,32,3,"conv32_3"),p=s(36864,64,3,"conv64_down",!0),d=s(36864,64,3,"conv64_1"),c=s(36864,64,3,"conv64_2"),h=s(36864,64,3,"conv64_3"),m=s(147456,128,3,"conv128_down",!0),f=s(147456,128,3,"conv128_1"),g=s(147456,128,3,"conv128_2"),y=s(589824,256,3,"conv256_down",!0),b=s(589824,256,3,"conv256_1"),x=s(589824,256,3,"conv256_2"),w=s(589824,256,3,"conv256_down_out"),I=P(()=>Ee(Aa(t(256*128),[128,256]),[1,0]));if(a.push({paramPath:"fc"}),n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{params:{conv32_down:i,conv32_1:o,conv32_2:l,conv32_3:u,conv64_down:p,conv64_1:d,conv64_2:c,conv64_3:h,conv128_down:m,conv128_1:f,conv128_2:g,conv256_down:y,conv256_1:b,conv256_2:x,conv256_down_out:w,fc:I},paramMappings:a}}function tce(e,t){let n=sa(e,t);function a(i){let o=n(`${i}/scale/weights`,1),l=n(`${i}/scale/biases`,1);return{weights:o,biases:l}}function r(i){let o=n(`${i}/conv/filters`,4),l=n(`${i}/conv/bias`,1),u=a(i);return{conv:{filters:o,bias:l},scale:u}}function s(i){return{conv1:r(`${i}/conv1`),conv2:r(`${i}/conv2`)}}return{extractConvLayerParams:r,extractResidualLayerParams:s}}function o$(e){let t=[],{extractConvLayerParams:n,extractResidualLayerParams:a}=tce(e,t),r=n("conv32_down"),s=a("conv32_1"),i=a("conv32_2"),o=a("conv32_3"),l=a("conv64_down"),u=a("conv64_1"),p=a("conv64_2"),d=a("conv64_3"),c=a("conv128_down"),h=a("conv128_1"),m=a("conv128_2"),f=a("conv256_down"),g=a("conv256_1"),y=a("conv256_2"),b=a("conv256_down_out"),{fc:x}=e;if(t.push({originalPath:"fc",paramPath:"fc"}),!c1(x))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${x}`);let w={conv32_down:r,conv32_1:s,conv32_2:i,conv32_3:o,conv64_down:l,conv64_1:u,conv64_2:p,conv64_3:d,conv128_down:c,conv128_1:h,conv128_2:m,conv256_down:f,conv256_1:g,conv256_2:y,conv256_down_out:b,fc:x};return An(e,t),{params:w,paramMappings:t}}function Ja(e,t){let n=s$(e,t.conv1);return n=z1(n,t.conv2),n=Y(n,e),n=Xe(n),n}function Id(e,t){let n=bg(e,t.conv1);n=z1(n,t.conv2);let a=ya(e,2,2,"valid"),r=It(a.shape),s=a.shape[3]!==n.shape[3];if(a.shape[1]!==n.shape[1]||a.shape[2]!==n.shape[2]){let o=[...n.shape];o[1]=1;let l=It(o);n=Ze([n,l],1);let u=[...n.shape];u[2]=1;let p=It(u);n=Ze([n,p],2)}return a=s?Ze([a,r],3):a,n=Y(a,n),n=Xe(n),n}var Oo=class extends sn{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceRecognitionNet - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(150,!0),"float32"),s=Za(a,[122.782,117.001,104.298]).div(255),i=bg(s,n.conv32_down);i=Dt(i,3,2,"valid"),i=Ja(i,n.conv32_1),i=Ja(i,n.conv32_2),i=Ja(i,n.conv32_3),i=Id(i,n.conv64_down),i=Ja(i,n.conv64_1),i=Ja(i,n.conv64_2),i=Ja(i,n.conv64_3),i=Id(i,n.conv128_down),i=Ja(i,n.conv128_1),i=Ja(i,n.conv128_2),i=Id(i,n.conv256_down),i=Ja(i,n.conv256_1),i=Ja(i,n.conv256_2),i=Id(i,n.conv256_down_out);let o=i.mean([1,2]);return Fe(o,n.fc)})}async forward(t){return this.forwardInput(await bt(t))}async computeFaceDescriptor(t){var s;if((s=t==null?void 0:t.shape)!=null&&s.some(i=>i<=0))return new Float32Array(128);let n=await bt(t),a=P(()=>ct(this.forwardInput(n))),r=await Promise.all(a.map(i=>i.data()));return a.forEach(i=>i.dispose()),n.isBatchInput?r:r[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return o$(t)}extractParams(t){return i$(t)}};function nce(e){let t=new Oo;return t.extractWeights(e),t}function xg(e,t){return{...e,...{descriptor:t}}}function ace(e){return typeof e.age=="number"}function vg(e,t){return{...e,...{age:t}}}function rce(e){return(e.gender==="male"||e.gender==="female")&&rp(e.genderProbability)}function wg(e,t,n){return{...e,...{gender:t,genderProbability:n}}}function sce(e,t){function n(l,u){let p=Da(e(9*l),[3,3,l,1]),d=Ke(e(l)),c=Ke(e(l)),h=Ke(e(l)),m=Ke(e(l));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/batch_norm_scale`},{paramPath:`${u}/batch_norm_offset`},{paramPath:`${u}/batch_norm_mean`},{paramPath:`${u}/batch_norm_variance`}),{filters:p,batch_norm_scale:d,batch_norm_offset:c,batch_norm_mean:h,batch_norm_variance:m}}function a(l,u,p,d,c){let h=Da(e(l*u*p*p),[p,p,l,u]),m=Ke(e(u));return t.push({paramPath:`${d}/filters`},{paramPath:`${d}/${c?"batch_norm_offset":"bias"}`}),{filters:h,bias:m}}function r(l,u,p,d){let{filters:c,bias:h}=a(l,u,p,d,!0);return{filters:c,batch_norm_offset:h}}function s(l,u,p){let d=n(l,`${p}/depthwise_conv`),c=r(l,u,1,`${p}/pointwise_conv`);return{depthwise_conv:d,pointwise_conv:c}}function i(){let l=r(3,32,3,"mobilenetv1/conv_0"),u=s(32,64,"mobilenetv1/conv_1"),p=s(64,128,"mobilenetv1/conv_2"),d=s(128,128,"mobilenetv1/conv_3"),c=s(128,256,"mobilenetv1/conv_4"),h=s(256,256,"mobilenetv1/conv_5"),m=s(256,512,"mobilenetv1/conv_6"),f=s(512,512,"mobilenetv1/conv_7"),g=s(512,512,"mobilenetv1/conv_8"),y=s(512,512,"mobilenetv1/conv_9"),b=s(512,512,"mobilenetv1/conv_10"),x=s(512,512,"mobilenetv1/conv_11"),w=s(512,1024,"mobilenetv1/conv_12"),I=s(1024,1024,"mobilenetv1/conv_13");return{conv_0:l,conv_1:u,conv_2:p,conv_3:d,conv_4:c,conv_5:h,conv_6:m,conv_7:f,conv_8:g,conv_9:y,conv_10:b,conv_11:x,conv_12:w,conv_13:I}}function o(){let l=r(1024,256,1,"prediction_layer/conv_0"),u=r(256,512,3,"prediction_layer/conv_1"),p=r(512,128,1,"prediction_layer/conv_2"),d=r(128,256,3,"prediction_layer/conv_3"),c=r(256,128,1,"prediction_layer/conv_4"),h=r(128,256,3,"prediction_layer/conv_5"),m=r(256,64,1,"prediction_layer/conv_6"),f=r(64,128,3,"prediction_layer/conv_7"),g=a(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),y=a(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),b=a(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),x=a(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),w=a(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),I=a(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),T=a(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),C=a(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),E=a(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),A=a(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),R=a(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),F=a(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:l,conv_1:u,conv_2:p,conv_3:d,conv_4:c,conv_5:h,conv_6:m,conv_7:f,box_predictor_0:{box_encoding_predictor:g,class_predictor:y},box_predictor_1:{box_encoding_predictor:b,class_predictor:x},box_predictor_2:{box_encoding_predictor:w,class_predictor:I},box_predictor_3:{box_encoding_predictor:T,class_predictor:C},box_predictor_4:{box_encoding_predictor:E,class_predictor:A},box_predictor_5:{box_encoding_predictor:R,class_predictor:F}}}return{extractMobilenetV1Params:i,extractPredictionLayerParams:o}}function l$(e){let t=[],{extractWeights:n,getRemainingWeights:a}=$n(e),{extractMobilenetV1Params:r,extractPredictionLayerParams:s}=sce(n,t),i=r(),o=s(),u={extra_dim:Rc(n(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:{mobilenetv1:i,prediction_layer:o,output_layer:u},paramMappings:t}}function ice(e,t){let n=sa(e,t);function a(u,p,d){let c=n(`${u}/Conv2d_${p}_pointwise/weights`,4,`${d}/filters`),h=n(`${u}/Conv2d_${p}_pointwise/convolution_bn_offset`,1,`${d}/batch_norm_offset`);return{filters:c,batch_norm_offset:h}}function r(u){let p=`mobilenetv1/conv_${u}`,d=`MobilenetV1/Conv2d_${u}_depthwise`,c=`${p}/depthwise_conv`,h=`${p}/pointwise_conv`,m=n(`${d}/depthwise_weights`,4,`${c}/filters`),f=n(`${d}/BatchNorm/gamma`,1,`${c}/batch_norm_scale`),g=n(`${d}/BatchNorm/beta`,1,`${c}/batch_norm_offset`),y=n(`${d}/BatchNorm/moving_mean`,1,`${c}/batch_norm_mean`),b=n(`${d}/BatchNorm/moving_variance`,1,`${c}/batch_norm_variance`);return{depthwise_conv:{filters:m,batch_norm_scale:f,batch_norm_offset:g,batch_norm_mean:y,batch_norm_variance:b},pointwise_conv:a("MobilenetV1",u,h)}}function s(){return{conv_0:a("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:r(1),conv_2:r(2),conv_3:r(3),conv_4:r(4),conv_5:r(5),conv_6:r(6),conv_7:r(7),conv_8:r(8),conv_9:r(9),conv_10:r(10),conv_11:r(11),conv_12:r(12),conv_13:r(13)}}function i(u,p){let d=n(`${u}/weights`,4,`${p}/filters`),c=n(`${u}/biases`,1,`${p}/bias`);return{filters:d,bias:c}}function o(u){let p=i(`Prediction/BoxPredictor_${u}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${u}/box_encoding_predictor`),d=i(`Prediction/BoxPredictor_${u}/ClassPredictor`,`prediction_layer/box_predictor_${u}/class_predictor`);return{box_encoding_predictor:p,class_predictor:d}}function l(){return{conv_0:a("Prediction",0,"prediction_layer/conv_0"),conv_1:a("Prediction",1,"prediction_layer/conv_1"),conv_2:a("Prediction",2,"prediction_layer/conv_2"),conv_3:a("Prediction",3,"prediction_layer/conv_3"),conv_4:a("Prediction",4,"prediction_layer/conv_4"),conv_5:a("Prediction",5,"prediction_layer/conv_5"),conv_6:a("Prediction",6,"prediction_layer/conv_6"),conv_7:a("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:o(0),box_predictor_1:o(1),box_predictor_2:o(2),box_predictor_3:o(3),box_predictor_4:o(4),box_predictor_5:o(5)}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:l}}function u$(e){let t=[],{extractMobilenetV1Params:n,extractPredictionLayerParams:a}=ice(e,t),r=e["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!Or(r))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${r}`);let s={mobilenetv1:n(),prediction_layer:a(),output_layer:{extra_dim:r}};return An(e,t),{params:s,paramMappings:t}}function Pa(e,t,n){return P(()=>{let a=$t(e,t.filters,n,"same");return a=Y(a,t.batch_norm_offset),en(a,0,6)})}var oce=.0010000000474974513;function lce(e,t,n){return P(()=>{let a=Is(e,t.filters,n,"same");return a=ks(a,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,oce),en(a,0,6)})}function uce(e){return[2,4,6,12].some(t=>t===e)?[2,2]:[1,1]}function p$(e,t){return P(()=>{let n,a=Pa(e,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((s,i)=>{let o=i+1,l=uce(o);a=lce(a,s.depthwise_conv,l),a=Pa(a,s.pointwise_conv,[1,1]),o===11&&(n=a)}),n===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:a,conv11:n}})}function pce(e,t,n){let a=e.arraySync(),r=Math.min(a[t][0],a[t][2]),s=Math.min(a[t][1],a[t][3]),i=Math.max(a[t][0],a[t][2]),o=Math.max(a[t][1],a[t][3]),l=Math.min(a[n][0],a[n][2]),u=Math.min(a[n][1],a[n][3]),p=Math.max(a[n][0],a[n][2]),d=Math.max(a[n][1],a[n][3]),c=(i-r)*(o-s),h=(p-l)*(d-u);if(c<=0||h<=0)return 0;let m=Math.max(r,l),f=Math.max(s,u),g=Math.min(i,p),y=Math.min(o,d),b=Math.max(g-m,0)*Math.max(y-f,0);return b/(c+h-b)}function c$(e,t,n,a,r){let s=e.shape[0],i=Math.min(n,s),o=t.map((p,d)=>({score:p,boxIndex:d})).filter(p=>p.score>r).sort((p,d)=>d.score-p.score),l=p=>p<=a?1:0,u=[];return o.forEach(p=>{if(u.length>=i)return;let d=p.score;for(let c=u.length-1;c>=0;--c){let h=pce(e,p.boxIndex,u[c]);if(h!==0&&(p.score*=l(h),p.score<=r))break}d===p.score&&u.push(p.boxIndex)}),u}function cce(e){let t=ct(Ee(e,[1,0])),n=[pe(t[2],t[0]),pe(t[3],t[1])],a=[Y(t[0],he(n[0],2)),Y(t[1],he(n[1],2))];return{sizes:n,centers:a}}function dce(e,t){let{sizes:n,centers:a}=cce(e),r=ct(Ee(t,[1,0])),s=he(z(fn(he(r[2],5)),n[0]),2),i=Y(z(he(r[0],10),n[0]),a[0]),o=he(z(fn(he(r[3],5)),n[1]),2),l=Y(z(he(r[1],10),n[1]),a[1]);return Ee(Ft([pe(i,s),pe(l,o),Y(i,s),Y(l,o)]),[1,0])}function d$(e,t,n){return P(()=>{let a=e.shape[0],r=dce(W(Ln(n.extra_dim,[a,1,1]),[-1,4]),W(e,[-1,4]));r=W(r,[a,r.shape[0]/a,4]);let s=ha(Be(t,[0,0,1],[-1,-1,-1])),i=Be(s,[0,0,0],[-1,-1,1]);i=W(i,[a,i.shape[1]]);let o=ct(r),l=ct(i);return{boxes:o,scores:l}})}function Lo(e,t){return P(()=>{let n=e.shape[0],a=W(Ro(e,t.box_encoding_predictor),[n,-1,1,4]),r=W(Ro(e,t.class_predictor),[n,-1,3]);return{boxPredictionEncoding:a,classPrediction:r}})}function h$(e,t,n){return P(()=>{let a=Pa(e,n.conv_0,[1,1]),r=Pa(a,n.conv_1,[2,2]),s=Pa(r,n.conv_2,[1,1]),i=Pa(s,n.conv_3,[2,2]),o=Pa(i,n.conv_4,[1,1]),l=Pa(o,n.conv_5,[2,2]),u=Pa(l,n.conv_6,[1,1]),p=Pa(u,n.conv_7,[2,2]),d=Lo(t,n.box_predictor_0),c=Lo(e,n.box_predictor_1),h=Lo(r,n.box_predictor_2),m=Lo(i,n.box_predictor_3),f=Lo(l,n.box_predictor_4),g=Lo(p,n.box_predictor_5),y=Ze([d.boxPredictionEncoding,c.boxPredictionEncoding,h.boxPredictionEncoding,m.boxPredictionEncoding,f.boxPredictionEncoding,g.boxPredictionEncoding],1),b=Ze([d.classPrediction,c.classPrediction,h.classPrediction,m.classPrediction,f.classPrediction,g.classPrediction],1);return{boxPredictions:y,classPredictions:b}})}var wa=class{constructor({minConfidence:t,maxResults:n}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=n||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var As=class extends sn{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("SsdMobilenetv1 - load model before inference");return P(()=>{let a=oe(t.toBatchTensor(512,!1),"float32"),r=pe(he(a,127.5),1),s=p$(r,n.mobilenetv1),{boxPredictions:i,classPredictions:o}=h$(s.out,s.conv11,n.prediction_layer);return d$(i,o,n.output_layer)})}async forward(t){return this.forwardInput(await bt(t))}async locateFaces(t,n={}){let{maxResults:a,minConfidence:r}=new wa(n),s=await bt(t),{boxes:i,scores:o}=this.forwardInput(s),l=i[0],u=o[0];for(let x=1;x<i.length;x++)i[x].dispose(),o[x].dispose();let p=Array.from(u.dataSync()),c=c$(l,p,a,.5,r),h=s.getReshapedInputDimensions(0),m=s.inputSize,f=m/h.width,g=m/h.height,y=l.arraySync(),b=c.map(x=>{let[w,I]=[Math.max(0,y[x][0]),Math.min(1,y[x][2])].map(E=>E*g),[T,C]=[Math.max(0,y[x][1]),Math.min(1,y[x][3])].map(E=>E*f);return new yt(p[x],new _o(T,w,C-T,I-w),{height:s.getInputHeight(0),width:s.getInputWidth(0)})});return l.dispose(),u.dispose(),b}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return u$(t)}extractParams(t){return l$(t)}};function m$(e){let t=new As;return t.extractWeights(e),t}function hce(e){return m$(e)}var W1=class extends As{};var f$=.4,g$=[new De(.738768,.874946),new De(2.42204,2.65704),new De(4.30971,7.04493),new De(10.246,4.59428),new De(12.6868,11.8741)],y$=[new De(1.603231,2.094468),new De(6.041143,7.080126),new De(2.882459,3.518061),new De(4.266906,5.178857),new De(9.041765,10.66308)],b$=[117.001,114.697,97.404],x$="tiny_yolov2_model",v$="tiny_yolov2_separable_conv_model";var kg=e=>typeof e=="number";function B1(e){if(!e)throw new Error(`invalid config: ${e}`);if(typeof e.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${e.withSeparableConvs}`);if(!kg(e.iouThreshold)||e.iouThreshold<0||e.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${e.iouThreshold}`);if(!Array.isArray(e.classes)||!e.classes.length||!e.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(e.classes)}`);if(!Array.isArray(e.anchors)||!e.anchors.length||!e.anchors.map(t=>t||{}).every(t=>kg(t.x)&&kg(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(e.anchors)}`);if(e.meanRgb&&(!Array.isArray(e.meanRgb)||e.meanRgb.length!==3||!e.meanRgb.every(kg)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(e.meanRgb)}`)}function yp(e){return P(()=>{let t=z(e,be(.10000000149011612));return Y(Xe(pe(e,t)),t)})}function Vr(e,t){return P(()=>{let n=ba(e,[[0,0],[1,1],[1,1],[0,0]]);return n=$t(n,t.conv.filters,[1,1],"valid"),n=pe(n,t.bn.sub),n=z(n,t.bn.truediv),n=Y(n,t.conv.bias),yp(n)})}function Ur(e,t){return P(()=>{let n=ba(e,[[0,0],[1,1],[1,1],[0,0]]);return n=Ts(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=Y(n,t.bias),yp(n)})}function mce(e,t){let n=pp(e,t);function a(i,o){let l=Ke(e(i)),u=Ke(e(i));return t.push({paramPath:`${o}/sub`},{paramPath:`${o}/truediv`}),{sub:l,truediv:u}}function r(i,o,l){let u=n(i,o,3,`${l}/conv`),p=a(o,`${l}/bn`);return{conv:u,bn:p}}let s=cp(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:r,extractSeparableConvParams:s}}function w$(e,t,n,a){let{extractWeights:r,getRemainingWeights:s}=$n(e),i=[],{extractConvParams:o,extractConvWithBatchNormParams:l,extractSeparableConvParams:u}=mce(r,i),p;if(t.withSeparableConvs){let[d,c,h,m,f,g,y,b,x]=a,w=t.isFirstLayerConv2d?o(d,c,3,"conv0"):u(d,c,"conv0"),I=u(c,h,"conv1"),T=u(h,m,"conv2"),C=u(m,f,"conv3"),E=u(f,g,"conv4"),A=u(g,y,"conv5"),R=b?u(y,b,"conv6"):void 0,F=x?u(b,x,"conv7"):void 0,S=o(x||b||y,5*n,1,"conv8");p={conv0:w,conv1:I,conv2:T,conv3:C,conv4:E,conv5:A,conv6:R,conv7:F,conv8:S}}else{let[d,c,h,m,f,g,y,b,x]=a,w=l(d,c,"conv0"),I=l(c,h,"conv1"),T=l(h,m,"conv2"),C=l(m,f,"conv3"),E=l(f,g,"conv4"),A=l(g,y,"conv5"),R=l(y,b,"conv6"),F=l(b,x,"conv7"),S=o(x,5*n,1,"conv8");p={conv0:w,conv1:I,conv2:T,conv3:C,conv4:E,conv5:A,conv6:R,conv7:F,conv8:S}}if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{params:p,paramMappings:i}}function fce(e,t){let n=sa(e,t);function a(o){let l=n(`${o}/sub`,1),u=n(`${o}/truediv`,1);return{sub:l,truediv:u}}function r(o){let l=n(`${o}/filters`,4),u=n(`${o}/bias`,1);return{filters:l,bias:u}}function s(o){let l=r(`${o}/conv`),u=a(`${o}/bn`);return{conv:l,bn:u}}let i=dp(n);return{extractConvParams:r,extractConvWithBatchNormParams:s,extractSeparableConvParams:i}}function k$(e,t){let n=[],{extractConvParams:a,extractConvWithBatchNormParams:r,extractSeparableConvParams:s}=fce(e,n),i;if(t.withSeparableConvs){let o=t.filterSizes&&t.filterSizes.length||9;i={conv0:t.isFirstLayerConv2d?a("conv0"):s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:o>7?s("conv6"):void 0,conv7:o>8?s("conv7"):void 0,conv8:a("conv8")}}else i={conv0:r("conv0"),conv1:r("conv1"),conv2:r("conv2"),conv3:r("conv3"),conv4:r("conv4"),conv5:r("conv5"),conv6:r("conv6"),conv7:r("conv7"),conv8:a("conv8")};return An(e,n),{params:i,paramMappings:n}}var Qa=class{constructor({inputSize:t,scoreThreshold:n}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=n||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var V1=class extends sn{constructor(n){super("TinyYolov2");B1(n),this._config=n}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(n,a){let r=Vr(n,a.conv0);return r=Dt(r,[2,2],[2,2],"same"),r=Vr(r,a.conv1),r=Dt(r,[2,2],[2,2],"same"),r=Vr(r,a.conv2),r=Dt(r,[2,2],[2,2],"same"),r=Vr(r,a.conv3),r=Dt(r,[2,2],[2,2],"same"),r=Vr(r,a.conv4),r=Dt(r,[2,2],[2,2],"same"),r=Vr(r,a.conv5),r=Dt(r,[2,2],[1,1],"same"),r=Vr(r,a.conv6),r=Vr(r,a.conv7),Ro(r,a.conv8,"valid",!1)}runMobilenet(n,a){let r=this.config.isFirstLayerConv2d?yp(Ro(n,a.conv0,"valid",!1)):Ur(n,a.conv0);return r=Dt(r,[2,2],[2,2],"same"),r=Ur(r,a.conv1),r=Dt(r,[2,2],[2,2],"same"),r=Ur(r,a.conv2),r=Dt(r,[2,2],[2,2],"same"),r=Ur(r,a.conv3),r=Dt(r,[2,2],[2,2],"same"),r=Ur(r,a.conv4),r=Dt(r,[2,2],[2,2],"same"),r=Ur(r,a.conv5),r=Dt(r,[2,2],[1,1],"same"),r=a.conv6?Ur(r,a.conv6):r,r=a.conv7?Ur(r,a.conv7):r,Ro(r,a.conv8,"valid",!1)}forwardInput(n,a){let{params:r}=this;if(!r)throw new Error("TinyYolov2 - load model before inference");return P(()=>{let s=oe(n.toBatchTensor(a,!1),"float32");return s=this.config.meanRgb?Za(s,this.config.meanRgb):s,s=s.div(255),this.config.withSeparableConvs?this.runMobilenet(s,r):this.runTinyYolov2(s,r)})}async forward(n,a){return this.forwardInput(await bt(n),a)}async detect(n,a={}){let{inputSize:r,scoreThreshold:s}=new Qa(a),i=await bt(n),o=await this.forwardInput(i,r),l=P(()=>ct(o)[0].expandDims()),u={width:i.getInputWidth(0),height:i.getInputHeight(0)},p=await this.extractBoxes(l,i.getReshapedInputDimensions(0),s);o.dispose(),l.dispose();let d=p.map(y=>y.box),c=p.map(y=>y.score),h=p.map(y=>y.classScore),m=p.map(y=>this.config.classes[y.label]);return y1(d.map(y=>y.rescale(r)),c,this.config.iouThreshold,!0).map(y=>new Lr(c[y],h[y],m[y],d[y],u))}getDefaultModelName(){return""}extractParamsFromWeightMap(n){return k$(n,this.config)}extractParams(n){let a=this.config.filterSizes||V1.DEFAULT_FILTER_SIZES,r=a?a.length:void 0;if(r!==7&&r!==8&&r!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${r} filterSizes in config`);return w$(n,this.config,this.boxEncodingSize,a)}async extractBoxes(n,a,r){let{width:s,height:i}=a,o=Math.max(s,i),l=o/s,u=o/i,p=n.shape[1],d=this.config.anchors.length,[c,h,m]=P(()=>{let b=n.reshape([p,p,d,this.boxEncodingSize]),x=b.slice([0,0,0,0],[p,p,d,4]),w=b.slice([0,0,0,4],[p,p,d,1]),I=this.withClassScores?Ka(b.slice([0,0,0,5],[p,p,d,this.config.classes.length]),3):be(0);return[x,w,I]}),f=[],g=await h.array(),y=await c.array();for(let b=0;b<p;b++)for(let x=0;x<p;x++)for(let w=0;w<d;w++){let I=dd(g[b][x][w][0]);if(!r||I>r){let T=(x+dd(y[b][x][w][0]))/p*l,C=(b+dd(y[b][x][w][1]))/p*u,E=Math.exp(y[b][x][w][2])*this.config.anchors[w].x/p*l,A=Math.exp(y[b][x][w][3])*this.config.anchors[w].y/p*u,R=T-E/2,F=C-A/2,S={row:b,col:x,anchor:w},{classScore:M,label:B}=this.withClassScores?await this.extractPredictedClass(m,S):{classScore:1,label:0};f.push({box:new Co(R,F,R+E,F+A),score:I,classScore:I*M,label:B,...S})}}return c.dispose(),h.dispose(),m.dispose(),f}async extractPredictedClass(n,a){let{row:r,col:s,anchor:i}=a,o=await n.array();return Array(this.config.classes.length).fill(0).map((l,u)=>o[r][s][i][u]).map((l,u)=>({classScore:l,label:u})).reduce((l,u)=>l.classScore>u.classScore?l:u)}},zo=V1;zo.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var Wo=class extends zo{constructor(t=!0){let n={withSeparableConvs:t,iouThreshold:f$,classes:["face"],...t?{anchors:y$,meanRgb:b$}:{anchors:g$,withClassScores:!0}};super(n)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(r=>new yt(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?v$:x$}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function gce(e,t=!0){let n=new Wo(t);return n.extractWeights(e),n}var Sd=class extends Qa{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var ka=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function Bo(e,t,n,a,r=({alignedRect:s})=>s){let s=e.map(l=>Mo(l)?r(l):l.detection),i=a||(t instanceof Te?await up(t,s):await lp(t,s)),o=await n(i);return i.forEach(l=>l instanceof Te&&l.dispose()),o}async function bp(e,t,n,a,r){return Bo([e],t,async s=>n(s[0]),a,r)}var I$=.4,S$=[new De(1.603231,2.094468),new De(6.041143,7.080126),new De(2.882459,3.518061),new De(4.266906,5.178857),new De(9.041765,10.66308)],T$=[117.001,114.697,97.404];var Vo=class extends zo{constructor(){let t={withSeparableConvs:!0,iouThreshold:I$,classes:["face"],anchors:S$,meanRgb:T$,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(r=>new yt(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var Qe={ssdMobilenetv1:new As,tinyFaceDetector:new Vo,tinyYolov2:new Wo,faceLandmark68Net:new Po,faceLandmark68TinyNet:new kd,faceRecognitionNet:new Oo,faceExpressionNet:new vd,ageGenderNet:new wd},N$=(e,t)=>Qe.ssdMobilenetv1.locateFaces(e,t),yce=(e,t)=>Qe.tinyFaceDetector.locateFaces(e,t),bce=(e,t)=>Qe.tinyYolov2.locateFaces(e,t),C$=e=>Qe.faceLandmark68Net.detectLandmarks(e),xce=e=>Qe.faceLandmark68TinyNet.detectLandmarks(e),vce=e=>Qe.faceRecognitionNet.computeFaceDescriptor(e),wce=e=>Qe.faceExpressionNet.predictExpressions(e),kce=e=>Qe.ageGenderNet.predictAgeAndGender(e),_$=e=>Qe.ssdMobilenetv1.load(e),Ice=e=>Qe.tinyFaceDetector.load(e),Sce=e=>Qe.tinyYolov2.load(e),Tce=e=>Qe.faceLandmark68Net.load(e),Nce=e=>Qe.faceLandmark68TinyNet.load(e),Cce=e=>Qe.faceRecognitionNet.load(e),_ce=e=>Qe.faceExpressionNet.load(e),Ece=e=>Qe.ageGenderNet.load(e),Ace=_$,$ce=N$,Fce=C$;var Ig=class extends ka{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.extractedFaces=r}},Uo=class extends Ig{async run(){let t=await this.parentTask,n=await Bo(t,this.input,async a=>Promise.all(a.map(r=>Qe.faceExpressionNet.predictExpressions(r))),this.extractedFaces);return t.map((a,r)=>dg(a,n[r]))}withAgeAndGender(){return new Ho(this,this.input)}},Go=class extends Ig{async run(){let t=await this.parentTask;if(!t)return;let n=await bp(t,this.input,a=>Qe.faceExpressionNet.predictExpressions(a),this.extractedFaces);return dg(t,n)}withAgeAndGender(){return new jo(this,this.input)}},$s=class extends Uo{withAgeAndGender(){return new Ds(this,this.input)}withFaceDescriptors(){return new Gr(this,this.input)}},Fs=class extends Go{withAgeAndGender(){return new Rs(this,this.input)}withFaceDescriptor(){return new Hr(this,this.input)}};var Sg=class extends ka{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.extractedFaces=r}},Ho=class extends Sg{async run(){let t=await this.parentTask,n=await Bo(t,this.input,async a=>Promise.all(a.map(r=>Qe.ageGenderNet.predictAgeAndGender(r))),this.extractedFaces);return t.map((a,r)=>{let{age:s,gender:i,genderProbability:o}=n[r];return vg(wg(a,i,o),s)})}withFaceExpressions(){return new Uo(this,this.input)}},jo=class extends Sg{async run(){let t=await this.parentTask;if(!t)return;let{age:n,gender:a,genderProbability:r}=await bp(t,this.input,s=>Qe.ageGenderNet.predictAgeAndGender(s),this.extractedFaces);return vg(wg(t,a,r),n)}withFaceExpressions(){return new Go(this,this.input)}},Ds=class extends Ho{withFaceExpressions(){return new $s(this,this.input)}withFaceDescriptors(){return new Gr(this,this.input)}},Rs=class extends jo{withFaceExpressions(){return new Fs(this,this.input)}withFaceDescriptor(){return new Hr(this,this.input)}};var Td=class extends ka{constructor(n,a){super();this.parentTask=n;this.input=a}},Gr=class extends Td{async run(){let t=await this.parentTask;return(await Bo(t,this.input,a=>Promise.all(a.map(r=>Qe.faceRecognitionNet.computeFaceDescriptor(r))),null,a=>a.landmarks.align(null,{useDlibAlignment:!0}))).map((a,r)=>xg(t[r],a))}withFaceExpressions(){return new $s(this,this.input)}withAgeAndGender(){return new Ds(this,this.input)}},Hr=class extends Td{async run(){let t=await this.parentTask;if(!t)return;let n=await bp(t,this.input,a=>Qe.faceRecognitionNet.computeFaceDescriptor(a),null,a=>a.landmarks.align(null,{useDlibAlignment:!0}));return xg(t,n)}withFaceExpressions(){return new Fs(this,this.input)}withAgeAndGender(){return new Rs(this,this.input)}};var Nd=class extends ka{constructor(n,a,r){super();this.parentTask=n;this.input=a;this.useTinyLandmarkNet=r}get landmarkNet(){return this.useTinyLandmarkNet?Qe.faceLandmark68TinyNet:Qe.faceLandmark68Net}},Cd=class extends Nd{async run(){let t=await this.parentTask,n=t.map(i=>i.detection),a=this.input instanceof Te?await up(this.input,n):await lp(this.input,n),r=await Promise.all(a.map(i=>this.landmarkNet.detectLandmarks(i)));return a.forEach(i=>i instanceof Te&&i.dispose()),t.filter((i,o)=>r[o]).map((i,o)=>fp(i,r[o]))}withFaceExpressions(){return new $s(this,this.input)}withAgeAndGender(){return new Ds(this,this.input)}withFaceDescriptors(){return new Gr(this,this.input)}},_d=class extends Nd{async run(){let t=await this.parentTask;if(!t)return;let{detection:n}=t,a=this.input instanceof Te?await up(this.input,[n]):await lp(this.input,[n]),r=await this.landmarkNet.detectLandmarks(a[0]);return a.forEach(s=>s instanceof Te&&s.dispose()),fp(t,r)}withFaceExpressions(){return new Fs(this,this.input)}withAgeAndGender(){return new Rs(this,this.input)}withFaceDescriptor(){return new Hr(this,this.input)}};var Ed=class extends ka{constructor(n,a=new wa){super();this.input=n;this.options=a}},xp=class extends Ed{async run(){let{input:t,options:n}=this,a;if(n instanceof Sd)a=Qe.tinyFaceDetector.locateFaces(t,n);else if(n instanceof wa)a=Qe.ssdMobilenetv1.locateFaces(t,n);else if(n instanceof Qa)a=Qe.tinyYolov2.locateFaces(t,n);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return a}runAndExtendWithFaceDetections(){return new Promise((t,n)=>{this.run().then(a=>t(a.map(r=>Ao({},r)))).catch(a=>n(a))})}withFaceLandmarks(t=!1){return new Cd(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new Uo(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new Ho(this.runAndExtendWithFaceDetections(),this.input)}},Ad=class extends Ed{async run(){let t=await new xp(this.input,this.options),n=t[0];return t.forEach(a=>{a.score>n.score&&(n=a)}),n}runAndExtendWithFaceDetection(){return new Promise(async t=>{let n=await this.run();t(n?Ao({},n):void 0)})}withFaceLandmarks(t=!1){return new _d(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new Go(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new jo(this.runAndExtendWithFaceDetection(),this.input)}};function Dce(e,t=new wa){return new Ad(e,t)}function Tg(e,t=new wa){return new xp(e,t)}async function E$(e,t){return Tg(e,new wa(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function Rce(e,t={}){return Tg(e,new Qa(t)).withFaceLandmarks().withFaceDescriptors()}var Mce=E$;function U1(e,t){if(e.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let n=Array.from(e),a=Array.from(t);return Math.sqrt(n.map((r,s)=>r-a[s]).reduce((r,s)=>r+s*s,0))}var $d=class{constructor(t,n=.6){this._distanceThreshold=n;let a=Array.isArray(t)?t:[t];if(!a.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let r=1,s=()=>`person ${r++}`;this._labeledDescriptors=a.map(i=>{if(i instanceof gr)return i;if(i instanceof Float32Array)return new gr(s(),[i]);if(i.descriptor&&i.descriptor instanceof Float32Array)return new gr(s(),[i.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,n){return n.map(a=>U1(a,t)).reduce((a,r)=>a+r,0)/(n.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:n,label:a})=>new sp(a,this.computeMeanDistance(t,n))).reduce((n,a)=>n.distance<a.distance?n:a)}findBestMatch(t){let n=this.matchDescriptor(t);return n.distance<this._distanceThreshold?n:new sp("unknown",n.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let n=t.labeledDescriptors.map(a=>gr.fromJSON(a));return new $d(n,t.distanceThreshold)}};function Pce(e){let t=new Vo;return t.extractWeights(e),t}function A$(e,t){let{width:n,height:a}=new yn(t.width,t.height);if(n<=0||a<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:n,height:a})}`);if(Array.isArray(e))return e.map(r=>A$(r,{width:n,height:a}));if(Mo(e)){let r=e.detection.forSize(n,a),s=e.unshiftedLandmarks.forSize(r.box.width,r.box.height);return fp(Ao(e,r),s)}return yr(e)?Ao(e,e.detection.forSize(n,a)):e instanceof ra||e instanceof yt?e.forSize(n,a):e}var Oce=YA;return mF(Lce);})();
+    `}};function Wae(e){let{inputs:t,backend:n,attrs:r}=e,{x:s,segmentIds:a}=t,{numSegments:o}=r,i=s.shape.length,c=[],u=0,l=N.getAxesPermutation([u],i),p=s;l!=null&&(p=Nn({inputs:{x:s},backend:n,attrs:{perm:l}}),c.push(p),u=N.getInnerMostAxes(1,i)[0]);let d=N.segment_util.computeOutShape(p.shape,u,o),h=w.sizeFromShape([p.shape[u]]),f=he({inputs:{x:p},backend:n,attrs:{shape:[-1,h]}});c.push(f);let m=Hf(s.dtype),g=(x,k,S,C,E)=>{let $=x.shape[0],F=x.shape[1],A=N.segment_util.segOpComputeOptimalWindowSize(F,E),R={windowSize:A,inSize:F,batchSize:$,numSegments:E},T=new Bae(R,k),L=n.compileAndRun(T,[x,S],C);if(c.push(L),L.shape[1]===E)return L;let V=zE({backend:n,attrs:{start:0,stop:E,step:1,dtype:"float32"}}),G=WE({inputs:{x:V},backend:n,attrs:{reps:[F/A]}});return c.push(V),c.push(G),g(L,k,G,C,E)},b=g(f,"unsortedSegmentSum",a,m,o),y=he({inputs:{x:b},backend:n,attrs:{shape:d}}),v=y;if(l!=null){c.push(y);let x=N.getUndoAxesPermutation(l);v=Nn({inputs:{x:v},backend:n,attrs:{perm:x}})}return c.forEach(x=>n.disposeIntermediateTensorInfo(x)),v}var Vae={kernelName:Nd,backendName:"webgl",kernelFunc:Wae},Uae=[LZ,BZ,UZ,qZ,KZ,ZZ,QZ,tJ,aJ,iJ,lJ,hJ,gJ,xJ,kJ,TJ,NJ,$J,FJ,PJ,zJ,qJ,KJ,YJ,nQ,sQ,cQ,wZ,dQ,gQ,xQ,CQ,_Q,AQ,DQ,RQ,MQ,BQ,UQ,HQ,jQ,XQ,JQ,eee,see,oee,uee,pee,fee,yee,Iee,Cee,Eee,Dee,Fee,Pee,Mee,zee,Wee,Uee,jee,Yee,Qee,tte,ste,ite,dte,mte,xZ,bte,fQ,xte,kte,Cte,kZ,Ate,Rte,Ote,Bte,Ute,jte,Yte,ene,sne,ine,une,hne,mne,bne,wne,kne,Tne,Nne,Ene,Fne,Mne,Wne,Xne,CZ,Qne,nre,are,cre,JJ,dre,hre,mre,yre,Ire,TZ,Sre,Cre,_re,Are,$re,QJ,Hne,Rre,Lre,Vre,_Z,qre,Xre,Qre,nse,ose,cse,dse,fse,gse,vse,Ise,Tse,Ese,Dse,Pse,Lse,GJ,jne,Wse,Use,Hse,jse,Xse,Zse,Qse,tae,rae,oae,cae,lae,hae,mae,bae,vae,qne,PZ,Iae,Tae,_ae,Dae,Pae,OZ,Mae,zae,Vae,pre];for(let e of Uae)_d(e);var Et;(function(e){e[e.float32=0]="float32",e[e.int32=1]="int32",e[e.bool=2]="bool",e[e.string=3]="string",e[e.complex64=4]="complex64"})(Et||(Et={}));var ud;(function(e){e[e.linear=0]="linear",e[e.relu=1]="relu",e[e.relu6=2]="relu6",e[e.prelu=3]="prelu",e[e.leakyrelu=4]="leakyrelu",e[e.sigmoid=5]="sigmoid",e[e.elu=6]="elu"})(ud||(ud={}));var VE;function Gae(e){VE=e.wasm.cwrap(eo,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])}function Hae(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a,bias:o,preluActivationWeights:i}=t;if(s.dtype!=="float32"||a.dtype!=="float32")throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:c,transposeB:u,activation:l,leakyreluAlpha:p}=r,d=n.dataIdMap.get(s.dataId).id,h=n.dataIdMap.get(a.dataId).id,f=0;if(o!=null){let E=n.dataIdMap.get(o.dataId);if(E.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${E.shape.length}.`);f=E.id}let m=i==null?0:n.dataIdMap.get(i.dataId).id,g=ud[l];if(g==null)throw new Error(`${l} activation not yet supported for FusedConv2D in the wasm backend.`);let b=c?s.shape[2]:s.shape[1],y=u?a.shape[1]:a.shape[2],v=Mu.assertAndGetBroadcastShape(s.shape.slice(0,-2),a.shape.slice(0,-2)),x=n.makeOutput([...v,b,y],s.dtype),k=n.dataIdMap.get(x.dataId).id,S=new Uint8Array(new Int32Array(s.shape).buffer),C=new Uint8Array(new Int32Array(a.shape).buffer);return VE(d,S,s.shape.length,h,C,a.shape.length,c,u,g,f,m,p||0,k),x}var qae={kernelName:eo,backendName:"wasm",setupFunc:Gae,kernelFunc:Hae};function rn(e,t){let n;function r(a){n=a.wasm.cwrap(e,null,["number","number","number"])}function s(a){let{backend:o,inputs:{x:i}}=a,c=o.dataIdMap.get(i.dataId).id,u=o.makeOutput(i.shape,t||i.dtype),l=o.dataIdMap.get(u.dataId).id;return w.sizeFromShape(u.shape)===0||n(c,Et[i.dtype],l),u}return{kernelName:e,backendName:"wasm",setupFunc:r,kernelFunc:s}}var jae=rn(Fc);function pn(e,t,n){let r;function s(o){r=o.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])}function a(o){let{backend:i,inputs:c}=o,{a:u,b:l}=c,p=i.dataIdMap.get(u.dataId).id,d=i.dataIdMap.get(l.dataId).id,h=n!=null?n:u.dtype,f=N.assertAndGetBroadcastShape(u.shape,l.shape),m=i.makeOutput(f,h);if(w.sizeFromShape(f)===0)return m;let g=new Uint8Array(new Int32Array(u.shape).buffer),b=new Uint8Array(new Int32Array(l.shape).buffer),y=i.dataIdMap.get(m.dataId).id;return(()=>r(p,g,u.shape.length,d,b,l.shape.length,Et[u.dtype],y))(),m}return{kernelName:e,backendName:"wasm",setupFunc:s,kernelFunc:a}}var Kae=!0,Xae=pn(ba,Kae),UE;function Yae(e){UE=e.wasm.cwrap(bo,null,["array","number","number","number"])}function Zae(e){let{inputs:t,backend:n}=e,r=n.makeOutput(t[0].shape,t[0].dtype);if(w.sizeFromShape(r.shape)===0)return r;let s=t.map(i=>n.dataIdMap.get(i.dataId).id),a=new Uint8Array(new Int32Array(s).buffer),o=n.dataIdMap.get(r.dataId).id;return UE(a,s.length,Et[r.dtype],o),r}var Jae={kernelName:bo,backendName:"wasm",setupFunc:Yae,kernelFunc:Zae};function eg(e){let{inputs:{x:t},backend:n}=e;if(t.dtype==="string")return Cn(n.readSync(t.dataId),t.shape,t.dtype);let r=n.makeOutput(t.shape,t.dtype),s=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(r).set(s),r}var Qae={kernelName:Oo,backendName:"wasm",kernelFunc:eg},GE;function eoe(e){GE=e.wasm.cwrap(_s,null,["number","array","number","number","number","array","number"])}function fa(e){let{inputs:t,backend:n,attrs:r}=e,[s,a]=noe(t.x.shape,r.perm),o=!0;for(let f=0;f<a.length;f++)a[f]!==f&&(o=!1);let i=toe(t.x.shape,r.perm),c={dataId:t.x.dataId,shape:s,dtype:t.x.dtype};if(o){let f=eg({inputs:t,backend:n});return f.shape=i,f}let u=n.makeOutput(i,c.dtype),l=n.dataIdMap.get(c.dataId).id,p=n.dataIdMap.get(u.dataId).id,d=new Uint8Array(new Int32Array(a).buffer),h=new Uint8Array(new Int32Array(c.shape).buffer);return GE(l,h,c.shape.length,Et[c.dtype],p,d,a.length),u}function toe(e,t){let n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];return n}function noe(e,t){let n=[],r=[];for(let s=0;s<e.length;++s)e[s]!==1&&n.push(e[s]),e[t[s]]!==1&&r.push(t[s]);for(let s=0;s<r.length;++s){let a=-1;for(let o=0;o<r.length;++o)r[o]>=s&&(a===-1||r[a]>r[o])&&(a=o);r[a]=s}return[n,r]}var roe={kernelName:_s,backendName:"wasm",kernelFunc:fa,setupFunc:eoe};function Aa(e,t,n){let r=e.shape,s=e.shape.length,a=w.parseAxisParam(t,r),o=a,i=N.getAxesPermutation(o,s),c=null,u=!1;if(i!=null){let l=new Array(s);for(let h=0;h<l.length;h++)l[h]=r[i[h]];o=N.getInnerMostAxes(o.length,s),c=fa({inputs:{x:e},attrs:{perm:i},backend:n});let p=n.dataIdMap.get(e.dataId).id;n.dataIdMap.get(c.dataId).id!==p&&(u=!0)}return{transposed:c,originalAxes:a,axes:o,inputWasTransposed:u}}var HE;function soe(e){HE=e.wasm.cwrap(Oc,null,["number, number, number"])}function aoe(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,c=t.dataIdMap.get(o.dataId).id,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t);if(h){let v=t.dataIdMap.get(l.dataId).id;u=l,c=v}let f=u.shape.length;N.assertAxesAreInnerMostDims("all",p,f);let[m,g]=N.computeOutAndReduceShapes(u.shape,p),b=w.sizeFromShape(g),y=t.makeOutput(m,o.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;HE(c,b,v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var ooe={kernelName:Oc,backendName:"wasm",setupFunc:soe,kernelFunc:aoe},qE;function ioe(e){qE=e.wasm.cwrap(Mc,null,["number, number, number"])}function coe(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,c=t.dataIdMap.get(o.dataId).id,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t);if(h){let v=t.dataIdMap.get(l.dataId).id;u=l,c=v}let f=u.shape.length;N.assertAxesAreInnerMostDims("any",p,f);let[m,g]=N.computeOutAndReduceShapes(u.shape,p),b=w.sizeFromShape(g),y=t.makeOutput(m,o.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;qE(c,b,v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var uoe={kernelName:Mc,backendName:"wasm",setupFunc:ioe,kernelFunc:coe},jE;function loe(e){jE=e.wasm.cwrap(yo,null,["number","number","number","number","number"])}function doe(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s}=r,{x:a}=n,o=t.dataIdMap.get(a.dataId).id,i=o,c=a,{transposed:u,axes:l,inputWasTransposed:p}=Aa(a,s,t);if(p){let b=t.dataIdMap.get(u.dataId).id;b!==o&&(c=u,i=b)}let d=c.shape.slice(0,-1),h=t.makeOutput(d,"int32"),f=t.dataIdMap.get(h.dataId).id,m=w.sizeFromShape(h.shape),g=c.shape[l[0]];return jE(i,Et[c.dtype],m,g,f),p&&t.disposeData(u.dataId),h}var poe={kernelName:yo,backendName:"wasm",kernelFunc:doe,setupFunc:loe},KE;function hoe(e){KE=e.wasm.cwrap(vo,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function foe(e){let{inputs:t,attrs:n,backend:r}=e,s=t.x,a=r.dataIdMap.get(s.dataId).id,{filterSize:o,strides:i,pad:c,dimRoundingMode:u}=n,l=N.computePool2DInfo(s.shape,o,i,1,c,u),p=l.filterHeight,d=l.filterWidth,h=l.padInfo.top,f=l.padInfo.right,m=l.padInfo.bottom,g=l.padInfo.left,b=l.strideHeight,y=l.strideWidth,v=l.inChannels;if(l.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${l.dataFormat}'. Please use 'channelsLast'.`);if(l.dilationWidth!==1||l.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${l.dilationHeight}, ${l.dilationWidth}].`);let x=r.makeOutput(l.outShape,"float32"),k=r.dataIdMap.get(x.dataId).id;return KE(a,s.shape[0],s.shape[1],s.shape[2],p,d,h,f,m,g,b,y,v,k),x}var moe={kernelName:vo,backendName:"wasm",setupFunc:hoe,kernelFunc:foe};function Bn(e){let{inputs:t,attrs:n}=e,{x:r}=t,{shape:s}=n,a=w.sizeFromShape(r.shape),o=w.inferFromImplicitShape(s,a);return w.assert(a===w.sizeFromShape(o),()=>`new shape: ${o}, old shape: ${r.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(r.dataId),{dataId:r.dataId,shape:o,dtype:r.dtype}}var goe={kernelName:xu,backendName:"wasm",kernelFunc:Bn},XE;function boe(e){XE=e.wasm.cwrap(xo,null,["number","array","number","number","array","number","number","number","number"])}function yoe(e){let{inputs:t,backend:n,attrs:r}=e,{a:s,b:a}=t,{transposeA:o,transposeB:i}=r;if(s.dtype!=="float32"||a.dtype!=="float32")throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let c=s.shape.length,u=a.shape.length,l=o?s.shape[c-2]:s.shape[c-1],p=i?a.shape[u-1]:a.shape[u-2],d=o?s.shape[c-1]:s.shape[c-2],h=i?a.shape[u-2]:a.shape[u-1],f=s.shape.slice(0,-2),m=a.shape.slice(0,-2),g=w.sizeFromShape(f),b=w.sizeFromShape(m),v=Mu.assertAndGetBroadcastShape(s.shape.slice(0,-2),a.shape.slice(0,-2)).concat([d,h]);w.assert(l===p,()=>`Error in matMul: inner shapes (${l}) and (${p}) of Tensors with shapes ${s.shape} and ${a.shape} and transposeA=${o} and transposeB=${i} must match.`);let x=o?[g,l,d]:[g,d,l],k=i?[b,h,p]:[b,p,h],S=Bn({inputs:{x:s},backend:n,attrs:{shape:x}}),C=Bn({inputs:{x:a},backend:n,attrs:{shape:k}}),E=n.dataIdMap.get(S.dataId).id,$=n.dataIdMap.get(C.dataId).id,F=o?S.shape[2]:S.shape[1],A=i?C.shape[1]:C.shape[2],R=Math.max(g,b),T=n.makeOutput([R,F,A],S.dtype),L=n.dataIdMap.get(T.dataId).id,V=new Uint8Array(new Int32Array(S.shape).buffer),G=new Uint8Array(new Int32Array(C.shape).buffer);return XE(E,V,S.shape.length,$,G,C.shape.length,o,i,L),n.disposeData(S.dataId),n.disposeData(C.dataId),T.shape=v,T}var voe={kernelName:xo,backendName:"wasm",setupFunc:boe,kernelFunc:yoe};function mo(e){let{inputs:{x:t},attrs:{begin:n,size:r},backend:s}=e,[a,o]=qt.parseSliceParams(t,n,r),i=qt.isSliceContinous(t.shape,a,o),c=s.readSync(t.dataId),u=s.makeOutput(o,t.dtype),l=w.computeStrides(t.shape),p=s.dataIdMap.get(u.dataId);if(i){let f=qt.computeFlatOffset(a,l);return t.dtype==="string"?p.stringBytes=c.slice(f,f+w.sizeFromShape(o)):s.typedArrayFromHeap(u).set(c.subarray(f,f+w.sizeFromShape(o))),u}if(t.dtype==="string"){let f=Qh(c,a,o,t.shape,t.dtype);return p.stringBytes=f,u}let d=s.typedArrayFromHeap(u),h=t.shape.length;if(h===2)xoe(c,l[0],d,a,o);else if(h===3)woe(c,l[0],l[1],d,a,o);else if(h===4)Ioe(c,l[0],l[1],l[2],d,a,o);else{let f=Qh(c,a,o,t.shape,t.dtype);d.set(f)}return u}function xoe(e,t,n,r,s){let a=0,o=r[0],i=r[1],c=o+s[0];for(let u=o;u<c;u++){let l=u*t+i;n.set(e.subarray(l,l+s[1]),a),a+=s[1]}}function woe(e,t,n,r,s,a){let o=0,i=s[0],c=s[1],u=s[2],l=i+a[0],p=c+a[1];for(let d=i;d<l;d++)for(let h=c;h<p;h++){let f=d*t+h*n+u;r.set(e.subarray(f,f+a[2]),o),o+=a[2]}}function Ioe(e,t,n,r,s,a,o){let i=0,c=a[0],u=a[1],l=a[2],p=c+o[0],d=u+o[1],h=l+o[2],f=a[3];for(let m=c;m<p;m++)for(let g=u;g<d;g++)for(let b=l;b<h;b++){let y=m*t+g*n+b*r+f;s.set(e.subarray(y,y+o[3]),i),i+=o[3]}}var koe={kernelName:Su,backendName:"wasm",kernelFunc:mo};function Soe(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,crops:o}=r,i=a.reduce((b,y)=>b*y),c=N.getReshaped(s.shape,a,i),u=N.getPermuted(c.length,a.length),l=N.getReshapedPermuted(s.shape,a,i),p=N.getSliceBeginCoords(o,a.length),d=N.getSliceSize(l,o,a.length),h=Bn({inputs:{x:s},backend:n,attrs:{shape:c}}),f=fa({inputs:{x:h},backend:n,attrs:{perm:u}}),m=Bn({inputs:{x:f},backend:n,attrs:{shape:l}}),g=mo({inputs:{x:m},backend:n,attrs:{begin:p,size:d}});return n.disposeData(h.dataId),n.disposeData(f.dataId),n.disposeData(h.dataId),g}var Toe={kernelName:Uc,backendName:"wasm",kernelFunc:Soe};function rl(e){let{inputs:{x:t},attrs:{dtype:n},backend:r}=e,s=r.makeOutput(t.shape,n),a=r.typedArrayFromHeap(t);return r.typedArrayFromHeap(s).set(a),s}var Coe={kernelName:wo,backendName:"wasm",kernelFunc:rl},Noe=rn(Io),YE;function _oe(e){YE=e.wasm.cwrap(ya,null,["number","number","number","number"])}function Eoe(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{clipValueMin:a,clipValueMax:o}=r,i=n.dataIdMap.get(s.dataId).id,c=n.makeOutput(s.shape,s.dtype),u=n.dataIdMap.get(c.dataId).id;return YE(i,a,o,u),c}var Aoe={kernelName:ya,backendName:"wasm",setupFunc:_oe,kernelFunc:Eoe};function ZE(e){let{inputs:t,backend:n}=e,r=w.parseAxisParam(e.attrs.axis,t[0].shape)[0],s=t.map(h=>h.shape);N.assertParamsConsistent(s,r);let a=N.computeOutShape(t.map(h=>h.shape),r),o=t.filter(h=>w.sizeFromShape(h.shape)>0);if(o.length===1)return eg({inputs:{x:o[0]},backend:n});let i=n.makeOutput(a,t[0].dtype);if(w.sizeFromShape(a)===0)return i;if(o[0].dtype==="string"){let h=o.map(v=>{let k=[-1,w.sizeFromShape(v.shape.slice(r))];return Bn({inputs:{x:v},backend:n,attrs:{shape:k}})}),f=h.map(v=>({vals:n.readSync(v.dataId),shape:v.shape}));a=N.computeOutShape(h.map(v=>v.shape),1);let m=h[0].shape[0]===1,g=M0(f,a,t[0].dtype,m),b=N.computeOutShape(o.map(v=>v.shape),r);i.shape=b;let y=n.dataIdMap.get(i.dataId);return y.stringBytes=N.fromStringArrayToUint8(g),h.forEach(v=>n.disposeData(v.dataId)),i}let c=w.sizeFromShape(o[0].shape.slice(0,r)),u=0,l=o.map(h=>{let f=w.sizeFromShape(h.shape.slice(r));return u+=f,f}),p=o.map(h=>n.typedArrayFromHeap(h)),d=n.typedArrayFromHeap(i);for(let h=0;h<c;h++){let f=h*u;for(let m=0;m<p.length;m++){let g=l[m],b=h*g,y=p[m].subarray(b,b+g);d.set(y,f),f+=g}}return i}var $oe={kernelName:Gc,backendName:"wasm",kernelFunc:ZE},JE;function Doe(e){JE=e.wasm.cwrap(ko,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Foe(e){let{inputs:t,attrs:n,backend:r}=e,{x:s,filter:a}=t,o=r.dataIdMap.get(s.dataId).id,i=r.dataIdMap.get(a.dataId).id,{strides:c,dilations:u,pad:l,dimRoundingMode:p,dataFormat:d}=n,h=N.convertConv2DDataFormat(d),f=N.computeConv2DInfo(s.shape,a.shape,c,u,l,p,!1,h),m=f.filterHeight,g=f.filterWidth,b=f.padInfo.top,y=f.padInfo.right,v=f.padInfo.bottom,x=f.padInfo.left,k=f.dilationHeight,S=f.dilationWidth,C=f.strideHeight,E=f.strideWidth,$=f.inChannels,F=f.outChannels,A=f.padInfo.type==="SAME"?1:0;if(f.dataFormat!=="channelsLast")throw new Error(`wasm backend Conv2D does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);let R=r.makeOutput(f.outShape,"float32"),T=r.dataIdMap.get(R.dataId).id;return JE(o,s.shape[0],s.shape[1],s.shape[2],i,m,g,b,y,v,x,A,k,S,C,E,$,F,T),R}var Roe={kernelName:ko,backendName:"wasm",setupFunc:Doe,kernelFunc:Foe},QE;function Poe(e){QE=e.wasm.cwrap(So,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Ooe(e){let{backend:t,inputs:n,attrs:r}=e,{dy:s,filter:a}=n,{strides:o,pad:i,dataFormat:c,dimRoundingMode:u,inputShape:l}=r,p=1,d=N.convertConv2DDataFormat(c),h=N.computeConv2DInfo(l,a.shape,o,p,i,u,!1,d),{batchSize:f,filterHeight:m,filterWidth:g,inChannels:b,inHeight:y,inWidth:v,outChannels:x,outHeight:k,outWidth:S,strideHeight:C,strideWidth:E}=h,$=m-1-h.padInfo.top,F=g-1-h.padInfo.left,A=h.dataFormat==="channelsLast",R=w.computeStrides(h.inShape),T=w.computeStrides(s.shape),[L,V,G]=w.computeStrides(a.shape),j=R[0],H=A?R[1]:R[2],Z=A?R[2]:1,J=A?1:R[1],ee=T[0],re=A?T[1]:T[2],te=A?T[2]:1,ie=A?1:T[1],ne=t.makeOutput(h.inShape,"float32"),le=t.dataIdMap.get(ne.dataId).id,ue=t.dataIdMap.get(s.dataId).id,ye=t.dataIdMap.get(a.dataId).id;return QE(ue,ye,f,m,g,y,v,b,k,S,x,C,E,$,F,L,V,G,j,H,Z,J,ee,re,te,ie,le),ne}var Moe={kernelName:So,backendName:"wasm",setupFunc:Poe,kernelFunc:Ooe},Loe=rn(To),zoe=rn(Co),wv;(function(e){e[e.bilinear=0]="bilinear",e[e.nearest=1]="nearest"})(wv||(wv={}));var eA;function Boe(e){eA=e.wasm.cwrap(qc,null,["number","number","number","number","array","number","number","number","number","number"])}function Woe(e){let{backend:t,inputs:n,attrs:r}=e,{method:s,extrapolationValue:a,cropSize:o}=r,{image:i,boxes:c,boxInd:u}=n,l=c.shape[0],[p,d]=o,h=[l,p,d,i.shape[3]],f=t.dataIdMap.get(i.dataId),m;i.dtype!=="float32"&&(m=rl({backend:t,inputs:{x:i},attrs:{dtype:"float32"}}),f=t.dataIdMap.get(m.dataId));let g=f.id,b=t.dataIdMap.get(c.dataId).id,y=t.dataIdMap.get(u.dataId).id,v=t.makeOutput(h,"float32"),x=t.dataIdMap.get(v.dataId).id,k=new Uint8Array(new Int32Array(i.shape).buffer);return eA(g,b,y,l,k,p,d,wv[s],a,x),m!=null&&t.disposeData(m.dataId),v}var Voe={kernelName:qc,backendName:"wasm",setupFunc:Boe,kernelFunc:Woe},tA;function Uoe(e){tA=e.wasm.cwrap(Hc,null,["number","number","number","number","number","number"])}function Goe(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r,c=s.shape.length;w.assert(s.dtype==="float32"||s.dtype==="int32",()=>`cumprod does not support ${s.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([a],c),l=s;u!==null&&(l=fa({inputs:{x:s},attrs:{perm:u},backend:n}));let p=N.getInnerMostAxes(1,c)[0];N.assertAxesAreInnerMostDims("cumprod",[p],c);let d=n.makeOutput(l.shape,l.dtype),h=l.shape[p],f=n.dataIdMap.get(l.dataId).id,m=n.dataIdMap.get(d.dataId).id;tA(f,o?1:0,i?1:0,h,m,Et[s.dtype]);let g=d;if(u!==null){let b=N.getUndoAxesPermutation(u);g=fa({inputs:{x:d},attrs:{perm:b},backend:n}),n.disposeData(l.dataId),n.disposeData(d.dataId)}return g}var Hoe={kernelName:Hc,backendName:"wasm",setupFunc:Uoe,kernelFunc:Goe},nA;function qoe(e){nA=e.wasm.cwrap(No,null,["number","number","number","number","number","number"])}function joe(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{axis:a,exclusive:o,reverse:i}=r,c=s.shape.length;w.assert(s.dtype==="float32"||s.dtype==="int32",()=>`cumsum does not support ${s.dtype} tensors in the WASM backend`);let u=N.getAxesPermutation([a],c),l=s;u!==null&&(l=fa({inputs:{x:s},attrs:{perm:u},backend:n}));let p=N.getInnerMostAxes(1,c)[0];N.assertAxesAreInnerMostDims("cumsum",[p],c);let d=n.makeOutput(l.shape,l.dtype),h=l.shape[p],f=n.dataIdMap.get(l.dataId).id,m=n.dataIdMap.get(d.dataId).id;nA(f,o?1:0,i?1:0,h,m,Et[s.dtype]);let g=d;if(u!==null){let b=N.getUndoAxesPermutation(u);g=fa({inputs:{x:d},attrs:{perm:b},backend:n}),n.disposeData(l.dataId),n.disposeData(d.dataId)}return g}var Koe={kernelName:No,backendName:"wasm",setupFunc:qoe,kernelFunc:joe},rA;function Xoe(e){rA=e.wasm.cwrap(jc,null,["number","number","number","array","number","array","array","number","number"])}function Yoe(e){let{backend:t,inputs:n,attrs:r}=e,{x:s}=n,{blockSize:a,dataFormat:o}=r,i=s.shape[0],c=o==="NHWC"?s.shape[1]:s.shape[2],u=o==="NHWC"?s.shape[2]:s.shape[3],l=o==="NHWC"?s.shape[3]:s.shape[1],p=c*a,d=u*a,h=l/(a*a),f=o==="NHWC"?[i,p,d,h]:[i,h,p,d],m=t.makeOutput(f,"float32"),b=t.dataIdMap.get(s.dataId).id,y=new Uint8Array(new Int32Array(w.computeStrides(s.shape)).buffer),v=new Uint8Array(new Int32Array(f).buffer),x=new Uint8Array(new Int32Array(w.computeStrides(f)).buffer),k=t.dataIdMap.get(m.dataId).id;return rA(b,a,o==="NHWC"?1:0,y,s.shape.length-1,v,x,f.length,k),m}var Zoe={kernelName:jc,backendName:"wasm",setupFunc:Xoe,kernelFunc:Yoe},sA;function Joe(e){sA=e.wasm.cwrap(_o,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Qoe(e){let{inputs:t,attrs:n,backend:r}=e,{x:s,filter:a}=t,o=r.dataIdMap.get(s.dataId).id,i=r.dataIdMap.get(a.dataId).id,{strides:c,dilations:u,pad:l,dimRoundingMode:p}=n,d=u==null?[1,1]:u,h=N.computeConv2DInfo(s.shape,a.shape,c,d,l,p,!0),f=h.filterHeight,m=h.filterWidth,g=h.padInfo.top,b=h.padInfo.right,y=h.padInfo.bottom,v=h.padInfo.left,x=h.dilationHeight,k=h.dilationWidth,S=h.strideHeight,C=h.strideWidth,E=h.inChannels,$=h.outChannels,F=h.padInfo.type==="SAME"?1:0;if(h.dataFormat!=="channelsLast")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${h.dataFormat}'. Please use 'channelsLast'.`);let A=r.makeOutput(h.outShape,"float32"),R=r.dataIdMap.get(A.dataId).id;return sA(o,s.shape[0],s.shape[1],s.shape[2],i,f,m,g,b,y,v,F,x,k,S,C,E,$,R),A}var eie={kernelName:_o,backendName:"wasm",setupFunc:Joe,kernelFunc:Qoe},tie=rn(Ao),nie=!1,rie=pn(Xc,nie,"bool"),sie=rn($o,"float32");function Iv(e){let{inputs:t,attrs:n,backend:r}=e,{input:s}=t,{dim:a}=n,o=s.shape.length,i=s.shape.slice(),c=a;return a<0&&(w.assert(-(o+1)<=a,()=>`Axis must be in the interval [${-(o+1)}, ${o}]`),c=o+a+1),i.splice(c,0,1),Bn({inputs:{x:s},backend:r,attrs:{shape:i}})}var aie={kernelName:Yc,backendName:"wasm",kernelFunc:Iv};function aA(e){let{attrs:{shape:t,value:n,dtype:r},backend:s}=e,a=s.makeOutput(t,r);return s.typedArrayFromHeap(a).fill(n),a}var oie={kernelName:gd,backendName:"wasm",kernelFunc:aA},oA;function iie(e){oA=e.wasm.cwrap(Jc,null,["number","number","number","number","number","number"])}function cie(e){let{inputs:t,backend:n}=e,{image:r}=t,s=n.makeOutput(r.shape,r.dtype),a=n.dataIdMap.get(r.dataId).id,o=n.dataIdMap.get(s.dataId).id,[i,c,u,l]=r.shape;return oA(a,i,c,u,l,o),s}var uie={kernelName:Jc,backendName:"wasm",kernelFunc:cie,setupFunc:iie},lie=rn(Do),die=!1,pie=pn(Fo,die),iA;function hie(e){iA=e.wasm.cwrap(Ro,null,["number","number","number","number","number","number","number"])}function fie(e){let{backend:t,inputs:n,attrs:r}=e,{varianceEpsilon:s}=r,{x:a,mean:o,variance:i,offset:c,scale:u}=n,l=t.dataIdMap.get(a.dataId).id,p=t.dataIdMap.get(o.dataId).id,d=t.dataIdMap.get(i.dataId).id,h=c!=null?t.dataIdMap.get(c.dataId).id:0,f=u!=null?t.dataIdMap.get(u.dataId).id:0,m=t.makeOutput(a.shape,a.dtype);if(w.sizeFromShape(a.shape)===0)return m;let g=t.dataIdMap.get(m.dataId).id;return iA(l,p,d,h,f,s,g),m}var mie={kernelName:Ro,backendName:"wasm",setupFunc:hie,kernelFunc:fie},cA;function gie(e){cA=e.wasm.cwrap(to,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function bie(e){let{inputs:t,attrs:n,backend:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dilations:l,dataFormat:p,dimRoundingMode:d,activation:h,leakyreluAlpha:f}=n,m=N.computeConv2DInfo(s.shape,a.shape,c,l,u,d),g=ud[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedConv2D in the wasm backend.`);let b=r.dataIdMap.get(s.dataId).id,y=r.dataIdMap.get(a.dataId).id,v=m.outChannels,x=0;if(o!=null){let te=r.dataIdMap.get(o.dataId);if(te.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==v)throw new Error(`FusedConv2D bias shape (${te.shape}) does not match the number of output channels (${v})`);x=te.id}let k=m.filterHeight,S=m.filterWidth,C=m.padInfo.top,E=m.padInfo.right,$=m.padInfo.bottom,F=m.padInfo.left,A=m.dilationHeight,R=m.dilationWidth,T=m.strideHeight,L=m.strideWidth,V=m.inChannels,G=m.padInfo.type==="SAME"?1:0,j=m.batchSize,H=m.inHeight,Z=m.inWidth;if(p!=="NHWC")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let J=r.makeOutput(m.outShape,"float32"),ee=r.dataIdMap.get(J.dataId).id,re=i==null?0:r.dataIdMap.get(i.dataId).id;return cA(b,j,H,Z,y,k,S,x,C,E,$,F,G,A,R,T,L,V,v,g,re,f||0,ee),J}var yie={kernelName:to,backendName:"wasm",setupFunc:gie,kernelFunc:bie},uA;function vie(e){uA=e.wasm.cwrap(no,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function xie(e){let{inputs:t,attrs:n,backend:r}=e,{x:s,filter:a,bias:o,preluActivationWeights:i}=t,{strides:c,pad:u,dilations:l,dataFormat:p,dimRoundingMode:d,activation:h,leakyreluAlpha:f}=n,m=N.computeConv2DInfo(s.shape,a.shape,c,l,u,d,!0),g=ud[h];if(g==null)throw new Error(`${h} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let b=r.dataIdMap.get(s.dataId).id,y=r.dataIdMap.get(a.dataId).id,v=m.outChannels,x=0;if(o!=null){let te=r.dataIdMap.get(o.dataId);if(te.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${te.shape.length}.`);if(te.shape[0]!==v)throw new Error(`FusedDepthwiseConv2D bias shape (${te.shape}) does not match the number of output channels (${v})`);x=te.id}let k=m.filterHeight,S=m.filterWidth,C=m.padInfo.top,E=m.padInfo.right,$=m.padInfo.bottom,F=m.padInfo.left,A=m.dilationHeight,R=m.dilationWidth,T=m.strideHeight,L=m.strideWidth,V=m.inChannels,G=m.padInfo.type==="SAME"?1:0,j=m.batchSize,H=m.inHeight,Z=m.inWidth;if(p!=="NHWC")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let J=r.makeOutput(m.outShape,"float32"),ee=r.dataIdMap.get(J.dataId).id,re=i==null?0:r.dataIdMap.get(i.dataId).id;return uA(b,j,H,Z,y,k,S,x,C,E,$,F,G,A,R,T,L,V,v,g,re,f||0,ee),J}var wie={kernelName:no,backendName:"wasm",setupFunc:vie,kernelFunc:xie},lA;function Iie(e){lA=e.wasm.cwrap(eu,null,["number","number","number","number","number","number","array","number"])}function kie(e){let{backend:t,inputs:n}=e,{params:r,indices:s}=n,[a,o,i,c]=Uv.prepareAndValidate(r,s),u=t.makeOutput(a,r.dtype);if(o===0)return u;let l=s.shape,p=l[l.length-1],h=t.dataIdMap.get(r.dataId).id,m=t.dataIdMap.get(s.dataId).id,g=new Uint8Array(new Int32Array(c).buffer),b=t.dataIdMap.get(u.dataId).id;return lA(h,Et[r.dtype],m,o,p,i,g,b),u}var Sie={kernelName:eu,backendName:"wasm",setupFunc:Iie,kernelFunc:kie},dA;function Tie(e){dA=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])}function Cie(e){let{backend:t,inputs:n,attrs:r}=e,{x:s,indices:a}=n,{axis:o,batchDims:i}=r,c=w.parseAxisParam(o,s.shape)[0],u=t.readSync(a.dataId),l=s.shape[c];for(let $=0;$<u.length;++$){let F=u[$];w.assert(F<=l-1&&F>=0,()=>`GatherV2: the index value ${F} is not in [0, ${l-1}]`)}let p=N.segment_util.collectGatherOpShapeInfo(s,a,c,i),d=Bn({inputs:{x:s},attrs:{shape:[p.batchSize,p.outerSize,p.dimSize,p.sliceSize]},backend:t}),h=w.sizeFromShape(a.shape),f=Bn({inputs:{x:a},attrs:{shape:[p.batchSize,h/p.batchSize]},backend:t}),m=[p.batchSize,p.outerSize,h/p.batchSize,p.sliceSize],g=t.makeOutput(m,s.dtype);if(w.sizeFromShape(s.shape)===0)return g;let b=d.shape.length-1,v=t.dataIdMap.get(d.dataId).id,k=t.dataIdMap.get(f.dataId).id,S=t.dataIdMap.get(g.dataId).id,C=new Uint8Array(new Int32Array(w.computeStrides(d.shape)).buffer),E=new Uint8Array(new Int32Array(w.computeStrides(m)).buffer);return dA(v,Et[s.dtype],C,b,k,p.batchSize,E,S),t.disposeData(d.dataId),t.disposeData(f.dataId),g.shape=p.outputShape,g}var Nie={kernelName:Qc,backendName:"wasm",setupFunc:Tie,kernelFunc:Cie},_ie=!1,Eie=pn(tu,_ie,"bool"),Aie=!1,$ie=pn(Po,Aie,"bool"),pA;function Die(e){pA=e.wasm.cwrap(Mo,null,["number","number","number","number"])}function Fie(e){let{inputs:{x:t},attrs:{alpha:n},backend:r}=e,s=r.dataIdMap.get(t.dataId).id,a=r.makeOutput(t.shape,"float32");if(w.sizeFromShape(t.shape)!==0){let o=r.dataIdMap.get(a.dataId).id;pA(s,Et[t.dtype],n,o)}return a}var Rie={kernelName:Mo,backendName:"wasm",setupFunc:Die,kernelFunc:Fie},Pie=!1,Oie=pn(au,Pie,"bool"),Mie=!1,Lie=pn(ou,Mie,"bool"),zie=rn(Lo),Bie=!1,Wie=pn(cu,Bie,"bool"),Vie=rn(uu),Uie=!1,Gie=pn(lu,Uie,"bool"),Hie=!1,qie=pn(sS,Hie,"bool"),hA;function jie(e){hA=e.wasm.cwrap(zo,null,["number","number","number","number"])}function Kie(e){let{backend:t,inputs:n,attrs:r}=e,{reductionIndices:s,keepDims:a}=r,{x:o}=n,c=t.dataIdMap.get(o.dataId).id,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t);if(h){let v=t.dataIdMap.get(l.dataId).id;u=l,c=v}let f=u.shape.length;N.assertAxesAreInnerMostDims("max",p,f);let[m,g]=N.computeOutAndReduceShapes(u.shape,p),b=w.sizeFromShape(g),y=t.makeOutput(m,o.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;hA(c,Et[o.dtype],b,v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var Xie={kernelName:zo,backendName:"wasm",setupFunc:jie,kernelFunc:Kie},Yie=!1,Zie=pn(Bo,Yie),fA;function Jie(e){fA=e.wasm.cwrap(Wo,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Qie(e){let{inputs:t,attrs:n,backend:r}=e,s=t.x,a=r.dataIdMap.get(s.dataId).id;w.assert(s.dtype==="float32",()=>`Error in MaxPool: only float32 input is supported. Got ${s.dtype}.`);let{filterSize:o,strides:i,pad:c,dimRoundingMode:u}=n,l=N.computePool2DInfo(s.shape,o,i,1,c,u),p=l.filterHeight,d=l.filterWidth,h=l.padInfo.top,f=l.padInfo.right,m=l.padInfo.bottom,g=l.padInfo.left,b=l.dilationHeight,y=l.dilationWidth,v=l.strideHeight,x=l.strideWidth,k=l.inChannels,S=l.outChannels;if(l.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${l.dataFormat}'. Please use 'channelsLast'.`);let C=r.makeOutput(l.outShape,"float32"),E=r.dataIdMap.get(C.dataId).id;return fA(a,s.shape[0],s.shape[1],s.shape[2],p,d,h,f,m,g,b,y,v,x,k,S,E),C}var ece={kernelName:Wo,backendName:"wasm",setupFunc:Jie,kernelFunc:Qie},mA;function tce(e){mA=e.wasm.cwrap(Vo,null,["number, number, number"])}function nce(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,i=t.dataIdMap.get(o.dataId).id,c=i,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t),f=p;if(h){let x=t.dataIdMap.get(l.dataId).id;x!==i&&(u=l,c=x,f=N.getInnerMostAxes(f.length,u.shape.length))}N.assertAxesAreInnerMostDims("mean",f,u.shape.length);let[m,g]=N.computeOutAndReduceShapes(u.shape,f),b=w.sizeFromShape(g),y=u;u.dtype!=="float32"&&(y=rl({backend:t,inputs:{x:u},attrs:{dtype:"float32"}}),c=t.dataIdMap.get(y.dataId).id);let v=t.makeOutput(m,"float32");if(w.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(v.dataId).id;mA(c,b,x)}if(h&&t.disposeData(l.dataId),a){let x=N.expandShapeToKeepDim(v.shape,d);v.shape=x}return u.dtype!=="float32"&&t.disposeData(y.dataId),v}var rce={kernelName:Vo,backendName:"wasm",setupFunc:tce,kernelFunc:nce},gA;function sce(e){gA=e.wasm.cwrap(Uo,null,["number","number","number","number"])}function ace(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,i=t.dataIdMap.get(o.dataId).id,c=i,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t);if(h){let v=t.dataIdMap.get(l.dataId).id;v!==i&&(u=l,c=v)}let f=u.shape.length;N.assertAxesAreInnerMostDims("min",p,f);let[m,g]=N.computeOutAndReduceShapes(u.shape,p),b=w.sizeFromShape(g),y=t.makeOutput(m,u.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;gA(c,Et[o.dtype],b,v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var oce={kernelName:Uo,backendName:"wasm",setupFunc:sce,kernelFunc:ace},ice=!1,cce=pn(Go,ice),kv;(function(e){e[e.reflect=0]="reflect",e[e.symmetric=1]="symmetric"})(kv||(kv={}));var bA;function uce(e){bA=e.wasm.cwrap(Ho,null,["number","array","number","number","array","array","number","number"])}function lce(e){let{inputs:{x:t},backend:n,attrs:{paddings:r,mode:s}}=e,a=r.map((f,m)=>f[0]+t.shape[m]+f[1]),o=n.dataIdMap.get(t.dataId).id,i=n.makeOutput(a,t.dtype),c=n.dataIdMap.get(i.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),l=r.map(f=>f[0]),p=r.map(f=>f[1]),d=new Uint8Array(new Int32Array(l).buffer),h=new Uint8Array(new Int32Array(p).buffer);return bA(o,u,t.shape.length,Et[t.dtype],d,h,kv[s],c),i}var dce={kernelName:Ho,backendName:"wasm",kernelFunc:lce,setupFunc:uce},pce=!0,hce=pn(qo,pce),fce=rn(pu);function uI(e,t){let n=new Int32Array(e.wasm.HEAPU8.buffer,t,4),r=n[0],s=n[1],a=n[2],o=n[3];return e.wasm._free(t),{pSelectedIndices:r,selectedSize:s,pSelectedScores:a,pValidOutputs:o}}var yA;function mce(e){yA=e.wasm.cwrap(fu,"number",["number","number","number","number","number"])}function gce(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:s,maxOutputSize:a,scoreThreshold:o}=r,{boxes:i,scores:c}=n,u=t.dataIdMap.get(i.dataId).id,l=t.dataIdMap.get(c.dataId).id,p=yA(u,l,a,s,o),{pSelectedIndices:d,selectedSize:h,pSelectedScores:f,pValidOutputs:m}=uI(t,p);return t.wasm._free(f),t.wasm._free(m),t.makeOutput([h],"int32",d)}var bce={kernelName:fu,backendName:"wasm",setupFunc:mce,kernelFunc:gce},vA;function yce(e){vA=e.wasm.cwrap(mu,"number",["number","number","number","number","number","bool"])}function vce(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:s,maxOutputSize:a,scoreThreshold:o,padToMaxOutputSize:i}=r,{boxes:c,scores:u}=n,l=t.dataIdMap.get(c.dataId).id,p=t.dataIdMap.get(u.dataId).id,d=vA(l,p,a,s,o,i),{pSelectedIndices:h,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=uI(t,d);t.wasm._free(m);let b=t.makeOutput([f],"int32",h),y=t.makeOutput([],"int32",g);return[b,y]}var xce={kernelName:mu,backendName:"wasm",setupFunc:yce,kernelFunc:vce},xA;function wce(e){xA=e.wasm.cwrap(gu,"number",["number","number","number","number","number","number"])}function Ice(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:s,maxOutputSize:a,scoreThreshold:o,softNmsSigma:i}=r,{boxes:c,scores:u}=n,l=t.dataIdMap.get(c.dataId).id,p=t.dataIdMap.get(u.dataId).id,d=xA(l,p,a,s,o,i),{pSelectedIndices:h,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=uI(t,d);t.wasm._free(g);let b=t.makeOutput([f],"int32",h),y=t.makeOutput([f],"float32",m);return[b,y]}var kce={kernelName:gu,backendName:"wasm",setupFunc:wce,kernelFunc:Ice},Sce=!1,Tce=pn(hu,Sce,"bool"),wA;function Cce(e){wA=e.wasm.cwrap(jo,null,["number","number","number","number","number"])}function Nce(e){let{inputs:t,backend:n,attrs:r}=e,{indices:s}=t,{dtype:a,depth:o,onValue:i,offValue:c}=r,u=n.makeOutput([...s.shape,o],a),l=n.dataIdMap.get(u.dataId).id,d=n.dataIdMap.get(s.dataId).id;return wA(d,o,i,c,l),u}var _ce={kernelName:jo,backendName:"wasm",setupFunc:Cce,kernelFunc:Nce};function Ece(e){let{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(1),r}var Ace={kernelName:bu,backendName:"wasm",kernelFunc:Ece};function $ce(e){let{inputs:t,backend:n,attrs:r}=e,{axis:s}=r;if(t.length===1)return Iv({inputs:{input:t[0]},backend:n,attrs:{dim:s}});let a=t[0].shape,o=t[0].dtype;t.forEach(l=>{w.assertShapesMatch(a,l.shape,"All tensors passed to stack must have matching shapes"),w.assert(o===l.dtype,()=>"All tensors passed to stack must have matching dtypes")});let i=[],c=t.map(l=>{let p=Iv({inputs:{input:l},backend:n,attrs:{dim:s}});return i.push(p),p}),u=ZE({inputs:c,backend:n,attrs:{axis:s}});return i.forEach(l=>n.disposeData(l.dataId)),u}var Dce={kernelName:yu,backendName:"wasm",kernelFunc:$ce},IA;function Fce(e){IA=e.wasm.cwrap(Ko,null,["number","array","number","number","array","array","number","number"])}function Rce(e){let{inputs:{x:t},backend:n,attrs:{paddings:r,constantValue:s}}=e,a=r.map((m,g)=>m[0]+t.shape[g]+m[1]);if(w.sizeFromShape(t.shape)===0)return aA({backend:n,attrs:{shape:a,value:s,dtype:t.dtype}});let o=n.dataIdMap.get(t.dataId).id,i=n.makeOutput(a,t.dtype),u=n.dataIdMap.get(i.dataId).id,l=new Uint8Array(new Int32Array(t.shape).buffer),p=r.map(m=>m[0]),d=r.map(m=>m[1]),h=new Uint8Array(new Int32Array(p).buffer),f=new Uint8Array(new Int32Array(d).buffer);return IA(o,l,t.shape.length,Et[t.dtype],h,f,s,u),i}var kA={kernelName:Ko,backendName:"wasm",kernelFunc:Rce,setupFunc:Fce},Pce=!1,Oce=pn(Xo,Pce),SA;function Mce(e){SA=e.wasm.cwrap(Yo,null,["number","number","number"])}function Lce(e){let{inputs:t,backend:n}=e,{x:r,alpha:s}=t,a=n.dataIdMap.get(r.dataId).id,o=n.dataIdMap.get(s.dataId).id,i=a,c=r,u=c;c.dtype!=="float32"&&(u=rl({backend:n,inputs:{x:r},attrs:{dtype:"float32"}}),i=n.dataIdMap.get(u.dataId).id);let l=n.makeOutput(r.shape,"float32"),p=n.dataIdMap.get(l.dataId).id;return SA(i,o,p),c.dtype!=="float32"&&n.disposeData(u.dataId),l}var zce={kernelName:Yo,backendName:"wasm",setupFunc:Mce,kernelFunc:Lce},TA;function Bce(e){TA=e.wasm.cwrap(Zo,null,["number","number","number","number"])}function Wce(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,i=t.dataIdMap.get(o.dataId).id,c=i,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t),f=p;if(h){let v=t.dataIdMap.get(l.dataId).id;v!==i&&(u=l,c=v,f=N.getInnerMostAxes(f.length,u.shape.length))}N.assertAxesAreInnerMostDims("prod",f,u.shape.length);let[m,g]=N.computeOutAndReduceShapes(u.shape,f),b=w.sizeFromShape(g),y=t.makeOutput(m,u.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;TA(c,b,Et[y.dtype],v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var Vce={kernelName:Zo,backendName:"wasm",setupFunc:Bce,kernelFunc:Wce},Uce=e=>{let{backend:t,attrs:n}=e,{start:r,stop:s,step:a,dtype:o}=n,i=B0(r,s,a,o),c=t.makeOutput([i.length],o);return t.typedArrayFromHeap(c).set(i),c},Gce={kernelName:vd,backendName:"wasm",kernelFunc:Uce},Hce=!0,qce=pn(Eo,Hce),jce=rn(Jo),Kce=rn(ti),CA;function Xce(e){CA=e.wasm.cwrap(ei,null,["number","number","number","number","number","number","number","number","number","number"])}function Yce(e){let{backend:t,inputs:n,attrs:r}=e,{images:s}=n,{alignCorners:a,halfPixelCenters:o,size:i}=r,[c,u]=i,[l,p,d,h]=s.shape,f=[l,c,u,h],m=t.dataIdMap.get(s.dataId),g;m.dtype!=="float32"&&(g=rl({backend:t,inputs:{x:s},attrs:{dtype:"float32"}}),m=t.dataIdMap.get(g.dataId));let b=m.id,y=t.makeOutput(f,"float32");if(w.sizeFromShape(s.shape)===0)return y;let v=t.dataIdMap.get(y.dataId).id;return CA(b,l,p,d,h,c,u,a?1:0,o?1:0,v),g!=null&&t.disposeData(g.dataId),y}var Zce={kernelName:ei,backendName:"wasm",setupFunc:Xce,kernelFunc:Yce},NA;function Jce(e){NA=e.wasm.cwrap(Qo,null,["number","number","number","number","number","number","number","number","number","number"])}function Qce(e){let{backend:t,inputs:n,attrs:r}=e,{images:s}=n,{alignCorners:a,halfPixelCenters:o,size:i}=r,[c,u]=i,[l,p,d,h]=s.shape,f=[l,c,u,h],m=t.makeOutput(f,"float32");if(w.sizeFromShape(s.shape)===0)return m;let g=t.dataIdMap.get(s.dataId),b;g.dtype!=="float32"&&(b=rl({backend:t,inputs:{x:s},attrs:{dtype:"float32"}}),g=t.dataIdMap.get(b.dataId));let y=g.id,v=t.dataIdMap.get(m.dataId).id;return NA(y,l,p,d,h,c,u,a?1:0,o?1:0,v),b!=null&&t.disposeData(b.dataId),m}var eue={kernelName:Qo,backendName:"wasm",setupFunc:Jce,kernelFunc:Qce},_A;function tue(e){_A=e.wasm.cwrap(ni,null,["number","array","number","array","number","number"])}function nue(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{dims:a}=r,o=w.parseAxisParam(a,s.shape);if(s.shape.length===0)return eg({inputs:{x:s},backend:n});let i=n.makeOutput(s.shape,s.dtype),c=n.dataIdMap.get(s.dataId).id,u=n.dataIdMap.get(i.dataId).id,l=new Uint8Array(new Int32Array(o).buffer),p=new Uint8Array(new Int32Array(s.shape).buffer);_A(c,l,o.length,p,s.shape.length,u);let d=Bn({inputs:{x:i},attrs:{shape:s.shape},backend:n});return n.disposeData(i.dataId),d}var rue={kernelName:ni,backendName:"wasm",kernelFunc:nue,setupFunc:tue},EA;function sue(e){EA=e.wasm.cwrap(Ou,null,["number","number","number","number","number","number","number","number","array","number","number"])}function aue(e){let{inputs:t,backend:n,attrs:r}=e,{image:s}=t,{radians:a,fillValue:o,center:i}=r,c=n.makeOutput(s.shape,s.dtype),u=n.dataIdMap.get(s.dataId).id,l=n.dataIdMap.get(c.dataId).id,[p,d,h,f]=s.shape,[m,g]=N.getImageCenter(i,d,h),b=o===0,y=255,v=typeof o=="number"?[o,o,o,b?0:y]:[...o,y],x=new Uint8Array(new Int32Array(v).buffer);return EA(u,p,d,h,f,a,m,g,x,v.length,l),c}var oue={kernelName:Ou,backendName:"wasm",kernelFunc:aue,setupFunc:sue},iue=rn(ri),cue=rn(si),AA;function uue(e){AA=e.wasm.cwrap(wu,null,["number","number","number","number","number","number","array","number","number"])}function lue(e){let{backend:t,inputs:n,attrs:r}=e,{indices:s,updates:a}=n,{shape:o}=r,i=t.makeOutput(o,a.dtype);if(w.sizeFromShape(o)===0)return i;let{sliceRank:c,numUpdates:u,sliceSize:l,strides:p,outputSize:d}=Gv.calculateShapes(a,s,o),f=t.dataIdMap.get(s.dataId).id,g=t.dataIdMap.get(a.dataId).id,b=new Uint8Array(new Int32Array(p).buffer),y=t.dataIdMap.get(i.dataId).id;return AA(f,g,Et[a.dtype],c,u,l,b,d,y),i}var due={kernelName:wu,backendName:"wasm",setupFunc:uue,kernelFunc:lue},$A;function pue(e){$A=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}function hue(e){let{inputs:t,backend:n}=e,{condition:r,t:s,e:a}=t,o=n.dataIdMap.get(r.dataId).id,i=n.dataIdMap.get(s.dataId).id,c=n.dataIdMap.get(a.dataId).id,u=n.makeOutput(s.shape,s.dtype),l=n.dataIdMap.get(u.dataId).id,p=r.shape.length,d=s.shape.length,h=p===0||p>1||d===1?1:w.sizeFromShape(s.shape.slice(1));return $A(o,i,c,h,l),u}var fue={kernelName:Iu,backendName:"wasm",kernelFunc:hue,setupFunc:pue},DA;function mue(e){DA=e.wasm.cwrap(oi,null,["number","number"])}function gue(e){let{backend:t,inputs:{x:n}}=e,r=t.dataIdMap.get(n.dataId).id,s=t.makeOutput(n.shape,n.dtype),a=t.dataIdMap.get(s.dataId).id;return w.sizeFromShape(s.shape)===0||DA(r,a),s}var bue={kernelName:"Sigmoid",backendName:"wasm",setupFunc:mue,kernelFunc:gue},yue=rn(ai),FA;function vue(e){FA=e.wasm.cwrap(ui,null,["number","number","number","number"])}function xue(e){let{backend:t,inputs:{logits:n},attrs:{dim:r}}=e,s=t.dataIdMap.get(n.dataId).id,a=t.makeOutput(n.shape,n.dtype),o=t.dataIdMap.get(a.dataId).id,i=n.shape[r],c=w.sizeFromShape(n.shape)/i;return w.sizeFromShape(a.shape)===0||FA(s,o,i,c),a}var wue={kernelName:ui,backendName:"wasm",setupFunc:vue,kernelFunc:xue};function Iue(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,{blockShape:a,paddings:o}=r,i=w.sizeFromShape(a),c=[[0,0]];c.push(...o);for(let S=1+a.length;S<s.shape.length;++S)c.push([0,0]);let u=kA.kernelFunc({inputs:{x:s},backend:n,attrs:{paddings:c,constantValue:0}}),l=N.getReshaped(u.shape,a,i,!1),p=N.getPermuted(l.length,a.length,!1),d=N.getReshapedPermuted(u.shape,a,i,!1),m=Bn({inputs:{x:u},backend:n,attrs:{shape:l}}),y=fa({inputs:{x:m},backend:n,attrs:{perm:p}}),k=Bn({inputs:{x:y},backend:n,attrs:{shape:d}});return n.disposeData(u.dataId),n.disposeData(m.dataId),n.disposeData(y.dataId),k}var kue={kernelName:_u,backendName:"wasm",kernelFunc:Iue},RA;function Sue(e){RA=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])}function Tue(e){let{backend:t,inputs:n}=e,{indices:r,values:s,denseShape:a,defaultValue:o}=n,i=r.shape[0],c=r.shape[1],u=t.readSync(a.dataId)[0],l=[i+u,c],p=t.dataIdMap.get(r.dataId).id,d=t.dataIdMap.get(s.dataId).id,h=t.dataIdMap.get(o.dataId).id,f=t.makeOutput(l,r.dtype),m=t.dataIdMap.get(f.dataId).id,g=t.makeOutput(l.slice(0,1),s.dtype),b=t.dataIdMap.get(g.dataId).id,y=t.makeOutput([u],"bool"),v=t.dataIdMap.get(y.dataId).id,x=t.makeOutput([i],r.dtype),k=t.dataIdMap.get(x.dataId).id,S=t.makeOutput([4],"int32"),C=t.dataIdMap.get(S.dataId).id,E=RA(p,d,Et[s.dtype],i,u,c,h,m,b,v,k,C),$=t.readSync(S.dataId),F;switch($[0]){case 1:{F=N.getSparseFillEmptyRowsIndicesDenseShapeMismatch($[1]);break}case 2:{F=N.getSparseFillEmptyRowsNegativeIndexErrorMessage($[1],$[2]);break}case 3:F=N.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage($[1],$[2],$[3]);break;default:F=""}if(t.disposeData(S.dataId),F)throw t.disposeData(f.dataId),t.disposeData(g.dataId),t.disposeData(y.dataId),t.disposeData(x.dataId),new Error(F);let A=f,R=g;return E!==l[0]&&(A=mo({inputs:{x:f},attrs:{begin:0,size:[E,c]},backend:t}),R=mo({inputs:{x:g},attrs:{begin:0,size:E},backend:t}),t.disposeData(f.dataId),t.disposeData(g.dataId)),[A,R,y,x]}var Cue={kernelName:xd,backendName:"wasm",setupFunc:Sue,kernelFunc:Tue},PA;function Nue(e){PA=e.wasm.cwrap(Au,null,["number","number","number","number","number","number","number"])}function _ue(e){let{backend:t,inputs:n}=e,{inputIndices:r,inputShape:s,newShape:a}=n;if(r.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
+        ${r.shape}`);if(s.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
+        ${s.shape}`);if(a.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${a.shape}`);let o=t.dataIdMap.get(r.dataId).id,i=t.dataIdMap.get(s.dataId).id,c=t.dataIdMap.get(a.dataId).id,u=r.shape[0],l=w.sizeFromShape(a.shape),p=t.makeOutput([u,l],r.dtype),d=t.dataIdMap.get(p.dataId).id,h=t.makeOutput([l],a.dtype),f=t.dataIdMap.get(h.dataId).id,m=t.makeOutput([3],"int32"),g=t.dataIdMap.get(m.dataId).id;PA(o,i,c,u,d,f,g);let b=t.readSync(m.dataId),y;switch(b[0]){case 0:{y=N.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(b[1],b[2]);break}case 1:{y=N.getSparseReshapeNegativeOutputDimErrorMessage(b[1],b[2]);break}case 2:y=N.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let v=Array.from(t.readSync(s.dataId)),x=Array.from(t.readSync(h.dataId));y=N.getSparseReshapeInputOutputMultipleErrorMessage(v,x);break}case 4:{let v=Array.from(t.readSync(s.dataId)),x=Array.from(t.readSync(h.dataId));y=N.getSparseReshapeInputOutputMismatchErrorMessage(v,x);break}default:y=""}if(t.disposeData(m.dataId),y)throw t.disposeData(p.dataId),t.disposeData(h.dataId),new Error(y);return[p,h]}var Eue={kernelName:Au,backendName:"wasm",setupFunc:Nue,kernelFunc:_ue},OA;function MA(e){OA=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function LA(e,t){let{backend:n,inputs:r}=e,{data:s,indices:a,segmentIds:o}=r,i=a.shape[0],c=n.readSync(o.dataId,i-1,i)[0],l=i>0?c+1:0;if(l<0)throw new Error(N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let p=s.shape.slice();p[0]=l;let d=n.dataIdMap.get(s.dataId).id,h=n.dataIdMap.get(a.dataId).id,f=n.dataIdMap.get(o.dataId).id,m=n.makeOutput(p,s.dtype),g=n.dataIdMap.get(m.dataId).id,b=n.makeOutput([4],"int32"),y=n.dataIdMap.get(b.dataId).id;OA(d,Et[s.dtype],s.shape[0],h,f,g,y,t,0);let v=n.readSync(b.dataId),x;switch(v[0]){case 0:{x=N.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{x=N.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:x=N.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(v[1],v[2]);break;case 3:x=N.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(v[1],v[2],v[3]);break;default:x=""}if(n.disposeData(b.dataId),x)throw n.disposeData(m.dataId),new Error(x);return m}function Aue(e){return LA(e,!0)}var $ue={kernelName:wd,backendName:"wasm",setupFunc:MA,kernelFunc:Aue};function Due(e){return LA(e,!1)}var Fue={kernelName:Id,backendName:"wasm",setupFunc:MA,kernelFunc:Due};function Rue(e){let{inputs:t,attrs:n,backend:r}=e,{x:s}=t,{numOrSizeSplits:a,axis:o}=n,i=w.parseAxisParam(o,s.shape)[0],c=N.prepareSplitSize(s,a,i),u=new Array(s.shape.length).fill(0),l=s.shape.slice();return c.map(p=>{let d=[...l];d[i]=p;let h=mo({inputs:{x:s},attrs:{begin:u,size:d},backend:r});return u[i]+=p,h})}var Pue={kernelName:Eu,backendName:"wasm",kernelFunc:Rue},Oue=rn(ii),Mue=rn(kd),Lue=!0,zue=pn(li,Lue),zA;function Bue(e){zA=e.wasm.cwrap(xa,null,["number","number","number","number"])}function Wue(e){let{backend:t,inputs:n,attrs:r}=e,{alpha:s}=r,{x:a}=n,o=t.dataIdMap.get(a.dataId).id,i=t.makeOutput(a.shape,a.dtype),c=t.dataIdMap.get(i.dataId).id;return zA(o,s,Et[a.dtype],c),i}var Vue={kernelName:xa,backendName:"wasm",setupFunc:Bue,kernelFunc:Wue},BA;function Uue(e){BA=e.wasm.cwrap($u,null,["number","array","number","array","array","array","array","array","number","number"])}function Gue(e){let{backend:t,inputs:n,attrs:r}=e,{x:s}=n,{begin:a,end:o,strides:i,beginMask:c,endMask:u,ellipsisMask:l,newAxisMask:p,shrinkAxisMask:d}=r,{finalShapeSparse:h,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:b,begin:y,end:v,strides:x}=qt.sliceInfo(s.shape,a,o,i,c,u,l,p,d),k;if(m)k=Bn({inputs:{x:s},backend:t,attrs:{shape:f}});else if(g||b){w.assert(s.shape.length>=1,()=>`Input must have rank at least 1, got: ${s.shape.length}`);let S=qt.computeOutShape(y,v,x),C=mo({inputs:{x:s},backend:t,attrs:{begin:y,size:S}});k=Bn({inputs:{x:C},backend:t,attrs:{shape:f}}),t.disposeData(C.dataId)}else{let S=t.makeOutput(h,"float32"),C=t.dataIdMap.get(s.dataId).id,E=new Uint8Array(new Int32Array(w.computeStrides(s.shape)).buffer),$=new Uint8Array(new Int32Array(y).buffer),F=new Uint8Array(new Int32Array(v).buffer),A=new Uint8Array(new Int32Array(x).buffer),R=new Uint8Array(new Int32Array(h).buffer),T=new Uint8Array(new Int32Array(w.computeStrides(h)).buffer),L=t.dataIdMap.get(S.dataId).id;BA(C,E,s.shape.length,$,F,A,R,T,h.length,L),k=Bn({inputs:{x:S},backend:t,attrs:{shape:f}}),t.disposeData(S.dataId)}return k}var Hue={kernelName:$u,backendName:"wasm",setupFunc:Uue,kernelFunc:Gue};function que(e){let{backend:t,inputs:n,attrs:r}=e,{data:s,dataSplits:a}=n,{separator:o,nGramWidths:i,leftPad:c,rightPad:u,padWidth:l,preserveShortSequences:p}=r,d=t.readSync(s.dataId),h=t.readSync(a.dataId),[f,m]=V0(d,h,o,i,c,u,l,p),g=t.makeOutput([f.length],"string"),b=t.dataIdMap.get(g.dataId);b.stringBytes=f;let y=t.makeOutput(a.shape,"int32");return t.typedArrayFromHeap(y).set(m),[g,y]}var jue={kernelName:Sd,backendName:"wasm",kernelFunc:que};function Kue(e){let{backend:t,inputs:n,attrs:r}=e,{input:s,delimiter:a}=n,{skipEmpty:o}=r,i=t.readSync(s.dataId),c=t.readSync(a.dataId),[u,l,p]=U0(i,c[0],o),d=l.length,h=t.makeOutput([d,2],"int32");t.typedArrayFromHeap(h).set(u);let m=t.makeOutput([d],"string"),g=t.dataIdMap.get(m.dataId);g.stringBytes=l;let b=t.makeOutput([2],"int32");return t.typedArrayFromHeap(b).set(p),[h,m,b]}var Xue={kernelName:Td,backendName:"wasm",kernelFunc:Kue};function Yue(e){let{backend:t,inputs:n,attrs:r}=e,{input:s}=n,{numBuckets:a}=r,o=t.readSync(s.dataId),i=G0(o,a),c=t.makeOutput(s.shape,"int32");return t.typedArrayFromHeap(c).set(i),c}var Zue={kernelName:Cd,backendName:"wasm",kernelFunc:Yue},Jue=!0,Que=pn(di,Jue),WA;function ele(e){WA=e.wasm.cwrap(ci,null,["number","number","number","number"])}function tle(e){let{backend:t,inputs:n,attrs:r}=e,{axis:s,keepDims:a}=r,{x:o}=n,i=t.dataIdMap.get(o.dataId).id,c=i,u=o,{transposed:l,axes:p,originalAxes:d,inputWasTransposed:h}=Aa(o,s,t),f=p;if(h){let v=t.dataIdMap.get(l.dataId).id;v!==i&&(u=l,c=v,f=N.getInnerMostAxes(f.length,u.shape.length))}N.assertAxesAreInnerMostDims("sum",f,u.shape.length);let[m,g]=N.computeOutAndReduceShapes(u.shape,f),b=w.sizeFromShape(g),y=t.makeOutput(m,u.dtype);if(w.sizeFromShape(u.shape)!==0){let v=t.dataIdMap.get(y.dataId).id;WA(c,b,Et[y.dtype],v)}if(h&&t.disposeData(l.dataId),a){let v=N.expandShapeToKeepDim(y.shape,d);y.shape=v}return y}var nle={kernelName:ci,backendName:"wasm",setupFunc:ele,kernelFunc:tle},rle=rn(pi),sle=rn(hi),VA;function ale(e){VA=e.wasm.cwrap(va,null,["number","array","number","array","number","number"])}function ole(e){let{inputs:t,backend:n,attrs:r}=e,{x:s}=t,a=n.dataIdMap.get(s.dataId).id,{reps:o}=r,i=new Array(s.shape.length);for(let d=0;d<i.length;d++)i[d]=s.shape[d]*o[d];let c=new Uint8Array(new Int32Array(s.shape).buffer),u=new Uint8Array(new Int32Array(i).buffer),l=n.makeOutput(i,s.dtype),p=n.dataIdMap.get(l.dataId).id;return VA(a,c,s.shape.length,u,i.length,Et[l.dtype],p),l}var ile={kernelName:va,backendName:"wasm",setupFunc:ale,kernelFunc:ole},UA;function cle(e){UA=e.wasm.cwrap(Du,null,["number","array","number","number","number","bool","number","number"])}var ule=({inputs:e,backend:t,attrs:n})=>{let{x:r}=e,{k:s,sorted:a}=n,o=t.dataIdMap.get(r.dataId).id,i=new Uint8Array(new Int32Array(r.shape).buffer),c=r.shape.slice();c[c.length-1]=s;let u=t.makeOutput(c,r.dtype),l=t.dataIdMap.get(u.dataId).id,p=t.makeOutput(c,"int32"),d=t.dataIdMap.get(p.dataId).id;return UA(o,i,r.shape.length,Et[r.dtype],s,a,l,d),[u,p]},lle={kernelName:Du,backendName:"wasm",setupFunc:cle,kernelFunc:ule},GA;function dle(e){GA=e.wasm.cwrap(Fu,null,["number","number","bool","number","number","number","number","number","number","array","number","array","number","number","number","number","number"])}function ple(e){let{backend:t,inputs:n,attrs:r}=e,{image:s,transforms:a}=n,{interpolation:o,fillMode:i,fillValue:c,outputShape:u}=r,[l,p,d,h]=s.shape,[f,m]=u!=null?u:[p,d],g=[l,f,m,h],b=new Uint8Array(new Int32Array(w.computeStrides(s.shape)).buffer),y=new Uint8Array(new Int32Array(w.computeStrides(g)).buffer),v=t.makeOutput(g,s.dtype),x=t.dataIdMap.get(v.dataId).id,S=t.dataIdMap.get(s.dataId).id,E=t.dataIdMap.get(a.dataId).id,$=o==="nearest"?1:2,F;switch(i){case"constant":F=1;break;case"reflect":F=2;break;case"wrap":F=3;break;case"nearest":F=4;break;default:F=1;break}return GA(S,E,a.shape[0]>1,l,f,m,h,d,p,b,s.shape.length-1,y,g.length-1,$,F,c,x),v}var hle={kernelName:Fu,backendName:"wasm",setupFunc:dle,kernelFunc:ple};function fle(e){let{inputs:t,backend:n,attrs:r}=e,{value:s}=t,{axis:a}=r;a<0&&(a+=s.shape.length);let o=s.shape[a],i=s.shape.length,c=new Array(i-1),u=0;for(let h=0;h<i;h++)h!==a&&(c[u++]=s.shape[h]);let l=new Array(o),p=new Array(i).fill(0),d=s.shape.slice();d[a]=1;for(let h=0;h<l.length;h++)p[a]=h,l[h]=mo({inputs:{x:s},attrs:{begin:p,size:d},backend:n});return l.map(({dataId:h,dtype:f})=>({dataId:h,dtype:f,shape:c}))}var mle={kernelName:Ru,backendName:"wasm",kernelFunc:fle};function gle(e){let{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(0),r}var ble={kernelName:Pu,backendName:"wasm",kernelFunc:gle},yle=[qae,jae,Xae,Jae,ooe,uoe,poe,moe,voe,Toe,Coe,Noe,Aoe,$oe,Roe,Moe,Loe,zoe,Voe,Hoe,Koe,Zoe,eie,tie,rie,sie,aie,oie,uie,lie,pie,mie,yie,wie,Sie,Nie,Eie,$ie,Qae,Rie,Oie,Lie,zie,Wie,Vie,Gie,qie,Xie,Zie,ece,rce,oce,cce,dce,hce,fce,bce,xce,kce,Tce,_ce,Ace,Dce,kA,Oce,zce,Vce,Gce,qce,jce,Kce,goe,Zce,eue,rue,oue,iue,cue,due,fue,bue,yue,koe,wue,kue,Cue,Eue,$ue,Fue,Pue,Oue,Mue,zue,Vue,Hue,jue,Xue,Zue,Que,nle,rle,sle,ile,lle,hle,roe,mle,ble];for(let e of yle)_d(e);var Sv=q();Sv.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(e){return!1}});Sv.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(Sv.get("IS_NODE"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var M1=ma(OD()),vle=ma(MD()),L1=ma(LD()),z1=M1.default||M1,xle=L1.default||L1,HA=class extends ld{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(qA),Tv=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new uf(this,Er())}write(e,t,n){let r={id:this.dataIdNextNumber++};return this.move(r,e,t,n,1),r}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){let t=w.now();return e(),{kernelMs:w.now()-t}}move(e,t,n,r,s){let a=this.dataIdNextNumber++;if(r==="string"){let u=t;this.dataIdMap.set(e,{id:a,stringBytes:u,shape:n,dtype:r,memoryOffset:null,refCount:s});return}let o=w.sizeFromShape(n),i=o*w.bytesPerElement(r),c=this.wasm._malloc(i);this.dataIdMap.set(e,{id:a,memoryOffset:c,shape:n,dtype:r,refCount:s}),this.wasm.tfjs.registerTensor(a,o,c),t!=null&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,i),c)}async read(e){return this.readSync(e)}readSync(e,t,n){let{memoryOffset:r,dtype:s,shape:a,stringBytes:o}=this.dataIdMap.get(e);if(s==="string")return(t==null||t===0)&&(n==null||n>=o.length)?o:o.slice(t,n);t=t||0,n=n||w.sizeFromShape(a);let i=w.bytesPerElement(s),c=this.wasm.HEAPU8.slice(r+t*i,r+n*i);return kle(c.buffer,s)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){let n=this.dataIdMap.get(e);if(n.refCount--,!t&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(e)}return!0}refCount(e){return this.dataIdMap.has(e)?this.dataIdMap.get(e).refCount:0}incRef(e){let t=this.dataIdMap.get(e);t!=null&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,n){let r;if(n==null)r=this.write(null,e,t);else{let s=this.dataIdNextNumber++;r={id:s},this.dataIdMap.set(r,{id:s,memoryOffset:n,shape:e,dtype:t,refCount:1});let a=w.sizeFromShape(e);this.wasm.tfjs.registerTensor(s,a,n)}return{dataId:r,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:n}){let r=this.wasm.HEAPU8.buffer,{memoryOffset:s}=this.dataIdMap.get(n),a=w.sizeFromShape(e);switch(t){case"float32":return new Float32Array(r,s,a);case"int32":return new Int32Array(r,s,a);case"bool":return new Uint8Array(r,s,a);default:throw new Error(`Unknown dtype ${t}`)}}};function wle(e){return(t,n)=>(w.fetch(e,{credentials:"same-origin"}).then(r=>{r.ok||t.env.a(`failed to load wasm binary file at '${e}'`),r.arrayBuffer().then(s=>{WebAssembly.instantiate(s,t).then(a=>{n(a.instance,a.module)})})}),{})}function B1(e,t,n){if(of!=null)return of;let r="tfjs-backend-wasm.wasm";return e&&t?r="tfjs-backend-wasm-threaded-simd.wasm":e&&(r="tfjs-backend-wasm-simd.wasm"),ql!=null&&ql[r]!=null?ql[r]:n+r}async function Ile(){let[e,t]=await Promise.all([q().getAsync("WASM_HAS_SIMD_SUPPORT"),q().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((n,r)=>{let s={};s.locateFile=(i,c)=>{if(i.endsWith(".worker.js")){let u=vle.wasmWorkerContents.replace(/\n/g,"\\n"),l=new Blob([u],{type:"application/javascript"});return URL.createObjectURL(l)}return i.endsWith(".wasm")?B1(e,t,Ul!=null?Ul:c):c+i},lI&&(s.instantiateWasm=wle(B1(e,t,Ul!=null?Ul:"")));let a=!1;s.onAbort=()=>{if(a||jl)return;jl=!0,r({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"})};let o;t&&e&&of==null?(s.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+z1.toString()],{type:"text/javascript"}),o=z1(s)):o=xle(s),o.then(i=>{a=!0,jl=!1;let c=null;i.tfjs={init:i.cwrap("init",null,[]),initWithThreadsCount:i.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:i.cwrap("get_threads_count","number",[]),registerTensor:i.cwrap("register_tensor",null,["number","number","number"]),disposeData:i.cwrap("dispose_data",c,["number"]),dispose:i.cwrap("dispose",c,[])},n({wasm:i})}).catch(r)})}function kle(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error(`Unknown dtype ${t}`)}}var Sle=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],of=null,Ul=null,ql={},jl=!1,lI=!1;function Tle(e,t=!1){if(Vv("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),jl)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");of=e,lI=t}function Cle(e,t=!1){if(jl)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if(typeof e=="string")Ul=e;else{ql=e;let n=Sle.filter(r=>ql[r]==null);if(n.length>0)throw new Error(`There were no entries found for the following binaries: ${n.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}lI=t}var qA=-1,Tv=-1;function Nle(e){qA=e}function _le(){if(Tv===-1)throw new Error("WASM backend not initialized.");return Tv}var Ele="4.0.0",Ale=2;qf("wasm",async()=>{let{wasm:e}=await Ile();return new HA(e)},Ale);var W1="4.0.0",$le="4.0.0",Dle="4.0.0",Fle="4.0.0",Rle="4.0.0",Ple={tfjs:W1,"tfjs-core":W1,"tfjs-converter":$le,"tfjs-backend-cpu":Dle,"tfjs-backend-webgl":Fle,"tfjs-backend-wasm":Rle};var PI={};sh(PI,{AnchorPosition:()=>CI,DrawBox:()=>fp,DrawBoxOptions:()=>sg,DrawFaceLandmarks:()=>mg,DrawFaceLandmarksOptions:()=>fg,DrawTextField:()=>zs,DrawTextFieldOptions:()=>il,drawContour:()=>Os,drawDetections:()=>Gle,drawFaceExpressions:()=>Xle,drawFaceLandmarks:()=>Zle});function Os(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:r,y:s},a)=>{let o=t[a];e.moveTo(o.x,o.y),e.lineTo(r,s)}),n){let r=t[t.length-1],s=t[0];if(!r||!s)return;e.moveTo(r.x,r.y),e.lineTo(s.x,s.y)}e.stroke()}var fI={};sh(fI,{computeReshapedDimensions:()=>hI,getCenterPoint:()=>Ni,isDimensions:()=>ng,isEven:()=>tg,isFloat:()=>pI,isTensor:()=>Ti,isTensor1D:()=>Ole,isTensor2D:()=>dI,isTensor3D:()=>Ms,isTensor4D:()=>yr,isValidNumber:()=>Zr,isValidProbablitiy:()=>sl,range:()=>gs,round:()=>Ci});var yn=class{constructor(t,n){if(!Zr(t)||!Zr(n))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:n})}`);this._width=t,this._height=n}get width(){return this._width}get height(){return this._height}reverse(){return new yn(1/this.width,1/this.height)}};function Ti(e,t){return e instanceof Te&&e.shape.length===t}function Ole(e){return Ti(e,1)}function dI(e){return Ti(e,2)}function Ms(e){return Ti(e,3)}function yr(e){return Ti(e,4)}function pI(e){return e%1!==0}function tg(e){return e%2===0}function Ci(e,t=2){let n=10**t;return Math.floor(e*n)/n}function ng(e){return e&&e.width&&e.height}function hI({width:e,height:t},n){let r=n/Math.max(t,e);return new yn(Math.round(e*r),Math.round(t*r))}function Ni(e){return e.reduce((t,n)=>t.add(n),new Fe(0,0)).div(new Fe(e.length,e.length))}function gs(e,t,n){return Array(e).fill(0).map((r,s)=>t+s*n)}function Zr(e){return!!e&&e!==1/0&&e!==-1/0&&!Number.isNaN(e)||e===0}function sl(e){return Zr(e)&&e>=0&&e<=1}var Fe=class{constructor(t,n){this._x=t,this._y=n}get x(){return this._x}get y(){return this._y}add(t){return new Fe(this.x+t.x,this.y+t.y)}sub(t){return new Fe(this.x-t.x,this.y-t.y)}mul(t){return new Fe(this.x*t.x,this.y*t.y)}div(t){return new Fe(this.x/t.x,this.y/t.y)}abs(){return new Fe(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new Fe(Math.floor(this.x),Math.floor(this.y))}};var ot=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(Zr)}static assertIsValidBox(t,n,r=!1){if(!ot.isRect(t))throw new Error(`${n} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${n} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,n=!0){let r=t||{},s=[r.left,r.top,r.right,r.bottom].every(Zr),a=[r.x,r.y,r.width,r.height].every(Zr);if(!a&&!s)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[o,i,c,u]=a?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];ot.assertIsValidBox({x:o,y:i,width:c,height:u},"Box.constructor",n),this._x=o,this._y=i,this._width=c,this._height=u}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new Fe(this.left,this.top)}get topRight(){return new Fe(this.right,this.top)}get bottomLeft(){return new Fe(this.left,this.bottom)}get bottomRight(){return new Fe(this.right,this.bottom)}round(){let[t,n,r,s]=[this.x,this.y,this.width,this.height].map(a=>Math.round(a));return new ot({x:t,y:n,width:r,height:s})}floor(){let[t,n,r,s]=[this.x,this.y,this.width,this.height].map(a=>Math.floor(a));return new ot({x:t,y:n,width:r,height:s})}toSquare(){let{x:t,y:n,width:r,height:s}=this,a=Math.abs(r-s);return r<s&&(t-=a/2,r+=a),s<r&&(n-=a/2,s+=a),new ot({x:t,y:n,width:r,height:s})}rescale(t){let n=ng(t)?t.width:t,r=ng(t)?t.height:t;return new ot({x:this.x*n,y:this.y*r,width:this.width*n,height:this.height*r})}pad(t,n){let[r,s,a,o]=[this.x-t/2,this.y-n/2,this.width+t,this.height+n];return new ot({x:r,y:s,width:a,height:o})}clipAtImageBorders(t,n){let{x:r,y:s,right:a,bottom:o}=this,i=Math.max(r,0),c=Math.max(s,0),u=a-i,l=o-c,p=Math.min(u,t-i),d=Math.min(l,n-c);return new ot({x:i,y:c,width:p,height:d}).floor()}shift(t,n){let{width:r,height:s}=this,a=this.x+t,o=this.y+n;return new ot({x:a,y:o,width:r,height:s})}padAtBorders(t,n){let r=this.width+1,s=this.height+1,a=1,o=1,i=r,c=s,u=this.left,l=this.top,p=this.right,d=this.bottom;return p>n&&(i=-p+n+r,p=n),d>t&&(c=-d+t+s,d=t),u<1&&(c=2-u,u=1),l<1&&(c=2-l,l=1),{dy:o,edy:c,dx:a,edx:i,y:l,ey:d,x:u,ex:p,w:r,h:s}}calibrate(t){return new ot({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var _i=class extends ot{constructor(t,n,r,s,a=!1){super({left:t,top:n,right:r,bottom:s},a)}};var Ls=class{constructor(t,n,r,s,a){this._imageDims=new yn(a.width,a.height),this._score=t,this._classScore=n,this._className=r,this._box=new ot(s).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new ot(this._box).rescale(this.imageDims.reverse())}forSize(t,n){return new Ls(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:n})}};var gt=class extends Ls{constructor(t,n,r){super(t,t,"",n,r)}forSize(t,n){let{score:r,relativeBox:s,imageDims:a}=super.forSize(t,n);return new gt(r,s,a)}};function mI(e,t,n=!0){let r=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left)),s=Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top)),a=r*s;return n?a/(e.area+t.area-a):a/Math.min(e.area,t.area)}function gI(e){let t=e.map(i=>i.x),n=e.map(i=>i.y),r=t.reduce((i,c)=>c<i?c:i,1/0),s=n.reduce((i,c)=>c<i?c:i,1/0),a=t.reduce((i,c)=>i<c?c:i,0),o=n.reduce((i,c)=>i<c?c:i,0);return new _i(r,s,a,o)}function bI(e,t,n,r=!0){let s=t.map((o,i)=>({score:o,boxIndex:i})).sort((o,i)=>o.score-i.score).map(o=>o.boxIndex),a=[];for(;s.length>0;){let o=s.pop();a.push(o);let i=s,c=[];for(let u=0;u<i.length;u++){let l=i[u],p=e[o],d=e[l];c.push(mI(p,d,r))}s=s.filter((u,l)=>c[l]<=n)}return a}function Jr(e,t){return O(()=>{let[n,r,s]=t,a=bn([...e.shape.slice(0,3),1],n,"float32"),o=bn([...e.shape.slice(0,3),1],r,"float32"),i=bn([...e.shape.slice(0,3),1],s,"float32"),c=Je([a,o,i],3);return de(e,c)})}function yI(e,t=!1){return O(()=>{let[n,r]=e.shape.slice(1);if(n===r)return e;let s=Math.abs(n-r),a=Math.round(s*(t?.5:1)),o=n>r?2:1,i=d=>{let h=e.shape.slice();return h[o]=d,bn(h,0,"float32")},c=i(a),u=s-c.shape[o],p=[t&&u?i(u):null,e,c].filter(d=>!!d).map(d=>ce(d,"float32"));return Je(p,o)})}function Mle(e){let t=e.slice();for(let n=t.length-1;n>0;n--){let r=Math.floor(Math.random()*(n+1)),s=t[n];t[n]=t[r],t[r]=s}return t}function pp(e){return 1/(1+Math.exp(-e))}function Lle(e){return Math.log(e/(1-e))}var Ei=class extends ot{constructor(t,n,r,s,a=!1){super({x:t,y:n,width:r,height:s},a)}};var zle=.5,Ble=.43,Wle=.45,ar=class{constructor(t,n,r=new Fe(0,0)){let{width:s,height:a}=n;this._imgDims=new yn(s,a),this._shift=r,this._positions=t.map(o=>o.mul(new Fe(s,a)).add(r))}get shift(){return new Fe(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new Fe(this.imageWidth,this.imageHeight)))}forSize(t,n){return new this.constructor(this.relativePositions,{width:t,height:n})}shiftBy(t,n){return new this.constructor(this.relativePositions,this._imgDims,new Fe(t,n))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,n={}){if(t){let a=t instanceof gt?t.box.floor():new ot(t);return this.shiftBy(a.x,a.y).align(null,n)}let{useDlibAlignment:r,minBoxPadding:s}={useDlibAlignment:!1,minBoxPadding:.2,...n};return r?this.alignDlib():this.alignMinBbox(s)}alignDlib(){let t=this.getRefPointsForAlignment(),[n,r,s]=t,a=p=>s.sub(p).magnitude(),o=(a(n)+a(r))/2,i=Math.floor(o/Wle),c=Ni(t),u=Math.floor(Math.max(0,c.x-zle*i)),l=Math.floor(Math.max(0,c.y-Ble*i));return new Ei(u,l,Math.min(i,this.imageWidth+u),Math.min(i,this.imageHeight+l))}alignMinBbox(t){let n=gI(this.positions);return n.pad(n.width*t,n.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var vI=class extends ar{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],Ni([t[3],t[4]])]}};var Ai=class extends ar{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(Ni)}};var al=class{constructor(t,n){this._label=t,this._distance=n}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${Ci(this.distance)})`:""}`}};var ol=class extends ot{constructor(n,r){super(n);this._label=r}static assertIsValidLabeledBox(n,r){if(ot.assertIsValidBox(n,r),!Zr(n.label))throw new Error(`${r} - expected property label (${n.label}) to be a number`)}get label(){return this._label}};var bs=class{constructor(t,n){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(n)||n.some(r=>!(r instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=n}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let n=t.descriptors.map(r=>new Float32Array(r));return new bs(t.label,n)}};var xI=class extends ol{constructor(n,r,s,a){super(n,r);this._score=s,this._classScore=a}static assertIsValidPredictedBox(n,r){if(ol.assertIsValidLabeledBox(n,r),!sl(n.score)||!sl(n.classScore))throw new Error(`${r} - expected properties score (${n.score}) and (${n.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function ys(e){return e.detection instanceof gt}function $i(e,t){return{...e,...{detection:t}}}function wI(){let e=window.fetch;if(!e)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:e,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function hp(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function rg(e){let t="";if(!e&&hp())try{e=Ys("fs")}catch(r){t=r.toString()}return{readFile:e?r=>new Promise((s,a)=>{e.readFile(r,(o,i)=>o?a(o):s(i))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function II(){let e=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,n=global.Video||global.HTMLVideoElement,r=()=>{if(e)return new e;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},s=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},a=()=>{if(n)return new n;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},o=global.fetch,i=rg();return{Canvas:e||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:r,createImageElement:s,createVideoElement:a,fetch:o,...i}}function kI(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var sn;function Vle(){if(!sn)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return sn}function SI(e){sn=e}function TI(){return kI()?SI(wI()):hp()?SI(II()):null}function Ule(e){if(sn||TI(),!sn)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=sn.Canvas,Image:n=sn.Image}=e;sn.Canvas=t,sn.Image=n,sn.createCanvasElement=e.createCanvasElement||(()=>new t),sn.createImageElement=e.createImageElement||(()=>new n),sn.ImageData=e.ImageData||sn.ImageData,sn.Video=e.Video||sn.Video,sn.fetch=e.fetch||sn.fetch,sn.readFile=e.readFile||sn.readFile}var Qe={getEnv:Vle,setEnv:SI,initialize:TI,createBrowserEnv:wI,createFileSystem:rg,createNodejsEnv:II,monkeyPatch:Ule,isBrowser:kI,isNodejs:hp};TI();function Di(e){return!Qe.isNodejs()&&typeof e=="string"?document.getElementById(e):e}function Gn(e){let{Canvas:t,CanvasRenderingContext2D:n}=Qe.getEnv();if(e instanceof n)return e;let r=Di(e);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let s=r.getContext("2d");if(!s)throw new Error("resolveContext2d - canvas 2d context is null");return s}var CI=(s=>(s.TOP_LEFT="TOP_LEFT",s.TOP_RIGHT="TOP_RIGHT",s.BOTTOM_LEFT="BOTTOM_LEFT",s.BOTTOM_RIGHT="BOTTOM_RIGHT",s))(CI||{}),il=class{constructor(t={}){let{anchorPosition:n,backgroundColor:r,fontColor:s,fontSize:a,fontStyle:o,padding:i}=t;this.anchorPosition=n||"TOP_LEFT",this.backgroundColor=r||"rgba(0, 0, 0, 0.5)",this.fontColor=s||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=o||"Georgia",this.padding=i||4}},zs=class{constructor(t,n,r={}){this.text=typeof t=="string"?[t]:t instanceof zs?t.text:t,this.anchor=n,this.options=new il(r)}measureWidth(t){let{padding:n}=this.options;return this.text.map(r=>t.measureText(r).width).reduce((r,s)=>r<s?s:r,0)+2*n}measureHeight(){let{fontSize:t,padding:n}=this.options;return this.text.length*t+2*n}getUpperLeft(t,n){let{anchorPosition:r}=this.options,s=r==="BOTTOM_RIGHT"||r==="TOP_RIGHT",a=r==="BOTTOM_LEFT"||r==="BOTTOM_RIGHT",o=this.measureWidth(t),i=this.measureHeight(),c=s?this.anchor.x-o:this.anchor.x,u=a?this.anchor.y-i:this.anchor.y;if(n){let{width:l,height:p}=n,d=Math.max(Math.min(c,l-o),0),h=Math.max(Math.min(u,p-i),0);return{x:d,y:h}}return{x:c,y:u}}draw(t){let n=Di(t),r=Gn(n),{backgroundColor:s,fontColor:a,fontSize:o,fontStyle:i,padding:c}=this.options;r.font=`${o}px ${i}`;let u=this.measureWidth(r),l=this.measureHeight();r.fillStyle=s;let p=this.getUpperLeft(r,n);r.fillRect(p.x,p.y,u,l),r.fillStyle=a,this.text.forEach((d,h)=>{let f=c+p.x,m=c+p.y+(h+1)*o;r.fillText(d,f,m)})}};var sg=class{constructor(t={}){let{boxColor:n,lineWidth:r,label:s,drawLabelOptions:a}=t;this.boxColor=n||"rgba(0, 0, 255, 1)",this.lineWidth=r||2,this.label=s;let o={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new il({...o,...a})}},fp=class{constructor(t,n={}){this.box=new ot(t),this.options=new sg(n)}draw(t){let n=Gn(t),{boxColor:r,lineWidth:s}=this.options,{x:a,y:o,width:i,height:c}=this.box;n.strokeStyle=r,n.lineWidth=s,n.strokeRect(a,o,i,c);let{label:u}=this.options;u&&new zs([u],{x:a-s/2,y:o},this.options.drawLabelOptions).draw(t)}};function Gle(e,t){(Array.isArray(t)?t:[t]).forEach(r=>{let s=r instanceof gt?r.score:ys(r)?r.detection.score:void 0,a=r instanceof gt?r.box:ys(r)?r.detection.box:new ot(r),o=s?`${Ci(s)}`:void 0;new fp(a,{label:o}).draw(e)})}function mp(e){let{Image:t,Video:n}=Qe.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}function NI(e){return new Promise((t,n)=>{(e instanceof Qe.getEnv().Canvas||mp(e))&&t(null);function r(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",s),a.currentTarget.removeEventListener("error",r),n(a))}function s(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",s),a.currentTarget.removeEventListener("error",r),t(a))}e.addEventListener("load",s),e.addEventListener("error",r)})}function _I(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToImage - expected buf to be of type: Blob"));let r=new FileReader;r.onload=()=>{typeof r.result!="string"&&n(new Error("bufferToImage - expected reader.result to be a string, in onload"));let s=Qe.getEnv().createImageElement();s.onload=()=>t(s),s.onerror=n,s.src=r.result},r.onerror=n,r.readAsDataURL(e)})}function Fi(e){let{Image:t,Video:n}=Qe.getEnv();return e instanceof t?new yn(e.naturalWidth,e.naturalHeight):e instanceof n?new yn(e.videoWidth,e.videoHeight):new yn(e.width,e.height)}function Ri({width:e,height:t}){let{createCanvasElement:n}=Qe.getEnv(),r=n();return r.width=e,r.height=t,r}function gp(e,t){let{ImageData:n}=Qe.getEnv();if(!(e instanceof n)&&!mp(e))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:s}=t||Fi(e),a=Ri({width:r,height:s});return e instanceof n?Gn(a).putImageData(e,0,0):Gn(a).drawImage(e,0,0,r,s),a}async function EI(e,t){let n=t||Qe.getEnv().createCanvasElement(),[r,s,a]=e.shape.slice(yr(e)?1:0),o=O(()=>e.as3D(r,s,a).toInt());return await fi.toPixels(o,n),o.dispose(),n}function ag(e){let{Image:t,Canvas:n,Video:r}=Qe.getEnv();return e instanceof t||e instanceof n||e instanceof r}function AI(e,t,n=!1){let{Image:r,Canvas:s}=Qe.getEnv();if(!(e instanceof r||e instanceof s))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Ri({width:1,height:1});let a=Fi(e),o=t/Math.max(a.height,a.width),i=o*a.width,c=o*a.height,u=Ri({width:t,height:t}),l=e instanceof s?e:gp(e),p=Math.abs(i-c)/2,d=n&&i<c?p:0,h=n&&c<i?p:0;return l.width>0&&l.height>0&&Gn(u).drawImage(l,d,h,i,c),u}var vs=class{constructor(t,n=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=n,this._batchSize=t.length,t.forEach((r,s)=>{if(Ms(r)){this._imageTensors[s]=r,this._inputDimensions[s]=r.shape;return}if(yr(r)){let o=r.shape[0];if(o!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${o} passed, but not supported in input array`);this._imageTensors[s]=r,this._inputDimensions[s]=r.shape.slice(1);return}let a=r instanceof Qe.getEnv().Canvas?r:gp(r);this._canvases[s]=a,this._inputDimensions[s]=[a.height,a.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return gs(this.batchSize,0,1).map((t,n)=>this.getReshapedInputDimensions(n))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let n=this.getInputWidth(t),r=this.getInputHeight(t);return hI({width:n,height:r},this.inputSize)}toBatchTensor(t,n=!0){return this._inputSize=t,O(()=>{let r=gs(this.batchSize,0,1).map(a=>{let o=this.getInput(a);if(o instanceof Te){let i=yr(o)?o:Zt(o);return i=yI(i,n),(i.shape[1]!==t||i.shape[2]!==t)&&(i=Br.resizeBilinear(i,[t,t],!1,!1)),i.as3D(t,t,3)}if(o instanceof Qe.getEnv().Canvas)return fi.fromPixels(AI(o,t,n));throw new Error(`toBatchTensor - at batchIdx ${a}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${o}`)});return Ft(r.map(a=>ce(a,"float32"))).as4D(this.batchSize,t,t,3)})}};async function bt(e){if(e instanceof vs)return e;let t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");let n=s=>Array.isArray(e)?` at input index ${s}:`:"",r=t.map(Di);return r.forEach((s,a)=>{if(!ag(s)&&!Ms(s)&&!yr(s))throw typeof t[a]=="string"?new Error(`toNetInput -${n(a)} string passed, but could not resolve HTMLElement for element id ${t[a]}`):new Error(`toNetInput -${n(a)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(yr(s)){let o=s.shape[0];if(o!==1)throw new Error(`toNetInput -${n(a)} tf.Tensor4D with batchSize ${o} passed, but not supported in input array`)}}),await Promise.all(r.map(s=>ag(s)&&NI(s))),new vs(r,Array.isArray(e))}async function cl(e,t){let{Canvas:n}=Qe.getEnv(),r=e;if(!(e instanceof n)){let o=await bt(e);if(o.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let i=o.getInput(0);r=i instanceof n?i:await EI(i)}let s=Gn(r);return t.map(o=>o instanceof gt?o.forSize(r.width,r.height).box.floor():o).map(o=>o.clipAtImageBorders(r.width,r.height)).map(({x:o,y:i,width:c,height:u})=>{let l=Ri({width:c,height:u});return c>0&&u>0&&Gn(l).putImageData(s.getImageData(o,i,c,u),0,0),l})}async function ul(e,t){if(!Ms(e)&&!yr(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(yr(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return O(()=>{let[n,r,s]=e.shape.slice(yr(e)?1:0);return t.map(i=>i instanceof gt?i.forSize(r,n).box:i).map(i=>i.clipAtImageBorders(r,n)).filter(i=>i.width>0&&i.height>0).map(({x:i,y:c,width:u,height:l})=>gi(e.as3D(n,r,s),[c,i,0],[l,u,s]))})}async function Bs(e,t){let{fetch:n}=Qe.getEnv(),r=await n(e,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}async function Hle(e){let t=await Bs(e),n=await t.blob();if(!n.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${n.type}, for url: ${t.url}`);return _I(n)}async function $I(e){return(await Bs(e)).json()}async function qle(e){return new Float32Array(await(await Bs(e)).arrayBuffer())}function jA(e){return new Promise((t,n)=>{e instanceof Blob||n(new Error("bufferToVideo - expected buf to be of type: Blob"));let r=Qe.getEnv().createVideoElement();r.oncanplay=()=>t(r),r.onerror=n,r.playsInline=!0,r.muted=!0,r.src=URL.createObjectURL(e),r.play()})}async function jle(e){let t=await Bs(e),n=await t.blob();if(!n.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${n.type}, for url: ${t.url}`);return jA(n)}function og(e,t){let n=`${t}-weights_manifest.json`;if(!e)return{modelBaseUri:"",manifestUri:n};if(e==="/")return{modelBaseUri:"/",manifestUri:`/${n}`};let r=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"";e=e.replace(r,"");let s=e.split("/").filter(i=>i),a=e.endsWith(".json")?s[s.length-1]:n,o=r+(e.endsWith(".json")?s.slice(0,s.length-1):s).join("/");return o=e.startsWith("/")?`/${o}`:o,{modelBaseUri:o,manifestUri:o==="/"?`/${a}`:`${o}/${a}`}}async function DI(e,t){let{manifestUri:n,modelBaseUri:r}=og(e,t),s=await $I(n);return Ut.loadWeights(s,r)}function Kle(e,t,n=!1){let{width:r,height:s}=n?Fi(t):t;return e.width=r,e.height=s,{width:r,height:s}}var an=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:n,objProp:r}=this.traversePropertyPath(t);return n[r]}reassignParamFromPath(t,n){let{obj:r,objProp:s}=this.traversePropertyPath(t);r[s].dispose(),r[s]=n}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof ia)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof ia))}variable(){this.getFrozenParams().forEach(({path:t,tensor:n})=>{this.reassignParamFromPath(t,n.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:n})=>{let r=Cn(n.dataSync());n.dispose(),this.reassignParamFromPath(t,r)})}dispose(t=!0){this.getParamList().forEach(n=>{if(t&&n.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${n.path}`);n.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,n)=>t.concat(n)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let n=await DI(t,this.getDefaultModelName());this.loadFromWeightMap(n)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:n}=Qe.getEnv(),{manifestUri:r,modelBaseUri:s}=og(t,this.getDefaultModelName()),a=u=>Promise.all(u.map(l=>n(l).then(p=>p.buffer))),o=Ut.weightsLoaderFactory(a),i=JSON.parse((await n(r)).toString()),c=await o(i,s);this.loadFromWeightMap(c)}loadFromWeightMap(t){let{paramMappings:n,params:r}=this.extractParamsFromWeightMap(t);this._paramMappings=n,this._params=r}extractWeights(t){let{paramMappings:n,params:r}=this.extractParams(t);this._paramMappings=n,this._params=r}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let n=t.split("/").reduce((a,o)=>{if(!a.nextObj.hasOwnProperty(o))throw new Error(`traversePropertyPath - object does not have property ${o}, for path ${t}`);return{obj:a.nextObj,objProp:o,nextObj:a.nextObj[o]}},{nextObj:this.params}),{obj:r,objProp:s}=n;if(!r||!s||!(r[s]instanceof Te))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:r,objProp:s}}};function Hn(e,t,n){return O(()=>{let r=Ca(e,t.depthwise_filter,t.pointwise_filter,n,"same");return r=Y(r,t.bias),r})}function ig(e,t,n=!1){return O(()=>{let r=Xe(n?Y(Dt(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):Hn(e,t.conv0,[2,2])),s=Hn(r,t.conv1,[1,1]),a=Xe(Y(r,s)),o=Hn(a,t.conv2,[1,1]);return Xe(Y(r,Y(s,o)))})}function bp(e,t,n=!1,r=!0){return O(()=>{let s=Xe(n?Y(Dt(e,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):Hn(e,t.conv0,r?[2,2]:[1,1])),a=Hn(s,t.conv1,[1,1]),o=Xe(Y(s,a)),i=Hn(o,t.conv2,[1,1]),c=Xe(Y(s,Y(a,i))),u=Hn(c,t.conv3,[1,1]);return Xe(Y(s,Y(a,Y(i,u))))})}function Pi(e,t,n="same",r=!1){return O(()=>{let s=Y(Dt(e,t.filters,[1,1],n),t.bias);return r?Xe(s):s})}function Fn(e,t){Object.keys(e).forEach(n=>{t.some(r=>r.originalPath===n)||e[n].dispose()})}function ll(e,t){return(n,r,s,a)=>{let o=Rr(e(n*r*s*s),[s,s,n,r]),i=Ke(e(r));return t.push({paramPath:`${a}/filters`},{paramPath:`${a}/bias`}),{filters:o,bias:i}}}function cg(e,t){return(n,r,s)=>{let a=$r(e(n*r),[n,r]),o=Ke(e(r));return t.push({paramPath:`${s}/weights`},{paramPath:`${s}/bias`}),{weights:a,bias:o}}}var yp=class{constructor(t,n,r){this.depthwise_filter=t;this.pointwise_filter=n;this.bias=r}};function dl(e,t){return(n,r,s)=>{let a=Rr(e(9*n),[3,3,n,1]),o=Rr(e(n*r),[1,1,n,r]),i=Ke(e(r));return t.push({paramPath:`${s}/depthwise_filter`},{paramPath:`${s}/pointwise_filter`},{paramPath:`${s}/bias`}),new yp(a,o,i)}}function pl(e){return t=>{let n=e(`${t}/depthwise_filter`,4),r=e(`${t}/pointwise_filter`,4),s=e(`${t}/bias`,1);return new yp(n,r,s)}}function or(e,t){return(n,r,s)=>{let a=e[n];if(!Ti(a,r))throw new Error(`expected weightMap[${n}] to be a Tensor${r}D, instead have ${a}`);return t.push({originalPath:n,paramPath:s||n}),a}}function Rn(e){let t=e;function n(s){let a=t.slice(0,s);return t=t.slice(s),a}function r(){return t}return{extractWeights:n,getRemainingWeights:r}}function ug(e,t){let n=ll(e,t),r=dl(e,t);function s(o,i,c,u=!1){let l=u?n(o,i,3,`${c}/conv0`):r(o,i,`${c}/conv0`),p=r(i,i,`${c}/conv1`),d=r(i,i,`${c}/conv2`);return{conv0:l,conv1:p,conv2:d}}function a(o,i,c,u=!1){let{conv0:l,conv1:p,conv2:d}=s(o,i,c,u),h=r(i,i,`${c}/conv3`);return{conv0:l,conv1:p,conv2:d,conv3:h}}return{extractDenseBlock3Params:s,extractDenseBlock4Params:a}}function KA(e){let t=[],{extractWeights:n,getRemainingWeights:r}=Rn(e),{extractDenseBlock4Params:s}=ug(n,t),a=s(3,32,"dense0",!0),o=s(32,64,"dense1"),i=s(64,128,"dense2"),c=s(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:o,dense2:i,dense3:c}}}function lg(e){return t=>{let n=e(`${t}/filters`,4),r=e(`${t}/bias`,1);return{filters:n,bias:r}}}function dg(e,t){let n=or(e,t),r=lg(n),s=pl(n);function a(i,c=!1){let u=c?r(`${i}/conv0`):s(`${i}/conv0`),l=s(`${i}/conv1`),p=s(`${i}/conv2`);return{conv0:u,conv1:l,conv2:p}}function o(i,c=!1){let u=c?r(`${i}/conv0`):s(`${i}/conv0`),l=s(`${i}/conv1`),p=s(`${i}/conv2`),d=s(`${i}/conv3`);return{conv0:u,conv1:l,conv2:p,conv3:d}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:o}}function XA(e){let t=[],{extractDenseBlock4Params:n}=dg(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return Fn(e,t),{params:r,paramMappings:t}}var hl=class extends an{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceFeatureExtractor - load model before inference");return O(()=>{let r=ce(t.toBatchTensor(112,!0),"float32"),a=Jr(r,[122.782,117.001,104.298]).div(255),o=bp(a,n.dense0,!0);return o=bp(o,n.dense1),o=bp(o,n.dense2),o=bp(o,n.dense3),o=mr(o,[7,7],[2,2],"valid"),o})}async forward(t){return this.forwardInput(await bt(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return XA(t)}extractParams(t){return KA(t)}};function vp(e,t){return O(()=>Y(De(e,t.weights),t.bias))}function YA(e,t,n){let r=[],{extractWeights:s,getRemainingWeights:a}=Rn(e),i=cg(s,r)(t,n,"fc");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:r,params:{fc:i}}}function ZA(e){let t=[],n=or(e,t);function r(a){let o=n(`${a}/weights`,2),i=n(`${a}/bias`,1);return{weights:o,bias:i}}let s={fc:r("fc")};return Fn(e,t),{params:s,paramMappings:t}}function pg(e){let t={},n={};return Object.keys(e).forEach(r=>{let s=r.startsWith("fc")?n:t;s[r]=e[r]}),{featureExtractorMap:t,classifierMap:n}}var fl=class extends an{constructor(n,r){super(n);this._faceFeatureExtractor=r}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return O(()=>{let s=n instanceof vs?this.faceFeatureExtractor.forwardInput(n):n;return vp(s.as2D(s.shape[0],-1),r.fc)})}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:r,paramMappings:s}=this.extractClassifierParams(n);this._params=r,this._paramMappings=s}extractClassifierParams(n){return YA(n,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(n){let{featureExtractorMap:r,classifierMap:s}=pg(n);return this.faceFeatureExtractor.loadFromWeightMap(r),ZA(s)}extractParams(n){let r=this.getClassifierChannelsIn(),s=this.getClassifierChannelsOut(),a=s*r+s,o=n.slice(0,n.length-a),i=n.slice(n.length-a);return this.faceFeatureExtractor.extractWeights(o),this.extractClassifierParams(i)}};var FI=["neutral","happy","sad","angry","fearful","disgusted","surprised"],Ws=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);FI.forEach((n,r)=>{this[n]=t[r]})}asSortedArray(){return FI.map(t=>({expression:t,probability:this[t]})).sort((t,n)=>n.probability-t.probability)}};var xp=class extends fl{constructor(t=new hl){super("FaceExpressionNet",t)}forwardInput(t){return O(()=>Xr(this.runNet(t)))}async forward(t){return this.forwardInput(await bt(t))}async predictExpressions(t){let n=await bt(t),r=await this.forwardInput(n),s=await Promise.all(lt(r).map(async o=>{let i=o.dataSync();return o.dispose(),i}));r.dispose();let a=s.map(o=>new Ws(o));return n.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function RI(e){return e.expressions instanceof Ws}function hg(e,t){return{...e,...{expressions:t}}}function Xle(e,t,n=.1,r){(Array.isArray(t)?t:[t]).forEach(a=>{let o=a instanceof Ws?a:RI(a)?a.expressions:void 0;if(!o)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let c=o.asSortedArray().filter(p=>p.probability>n),u=ys(a)?a.detection.box.bottomLeft:r||new Fe(0,0);new zs(c.map(p=>`${p.expression} (${Ci(p.probability)})`),u).draw(e)})}function Oi(e){return ys(e)&&e.landmarks instanceof ar&&e.unshiftedLandmarks instanceof ar&&e.alignedRect instanceof gt}function Yle(e){let t=c=>c*180/Math.PI,n=(c,u)=>Math.sqrt((c._x-u._x)**2+(c._y-u._y)**2),r={roll:void 0,pitch:void 0,yaw:void 0},s=(c,u,l)=>{let p=Math.floor(c._x-u._x),d=Math.floor(u._x-l._x);return p-d},a=(c,u)=>{let l=Math.hypot(u._x-c._x,u._y-c._y),p=u._y-c._y,d=Math.asin(p/l),h=t(d),f=Math.floor(90-h),m=u._x-c._x<0?-1:1;return f*m},o=(c,u,l)=>{let p=n(c,l),d={_x:(c._x+l._x)/2,_y:(c._y+l._y)/2},h=n(u,d),f=Math.atan(h/p),m=Math.floor(t(f)),g=d._y-u._y<0?-1:1;return m*g};if(!e||!e._positions||e._positions.length!==68)return r;let i=e._positions;return r.roll=a(i[27],i[66]),r.pitch=o(i[14],i[30],i[2]),r.yaw=s(i[14],i[33],i[2]),r}function ml(e,t){let{box:n}=e.detection,r=t.shiftBy(n.x,n.y),s=r.align(),{imageDims:a}=e.detection,o=new gt(e.detection.score,s.rescale(a.reverse()),a),i=Yle(t);return{...e,...{landmarks:r,unshiftedLandmarks:t,alignedRect:o,angle:i}}}var fg=class{constructor(t={}){let{drawLines:n=!0,drawPoints:r=!0,lineWidth:s,lineColor:a,pointSize:o,pointColor:i}=t;this.drawLines=n,this.drawPoints=r,this.lineWidth=s||1,this.pointSize=o||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},mg=class{constructor(t,n={}){this.faceLandmarks=t,this.options=new fg(n)}draw(t){let n=Gn(t),{drawLines:r,drawPoints:s,lineWidth:a,lineColor:o,pointSize:i,pointColor:c}=this.options;if(r&&this.faceLandmarks instanceof Ai&&(n.strokeStyle=o,n.lineWidth=a,Os(n,this.faceLandmarks.getJawOutline()),Os(n,this.faceLandmarks.getLeftEyeBrow()),Os(n,this.faceLandmarks.getRightEyeBrow()),Os(n,this.faceLandmarks.getNose()),Os(n,this.faceLandmarks.getLeftEye(),!0),Os(n,this.faceLandmarks.getRightEye(),!0),Os(n,this.faceLandmarks.getMouth(),!0)),s){n.strokeStyle=c,n.fillStyle=c;let u=l=>{n.beginPath(),n.arc(l.x,l.y,i,0,2*Math.PI),n.fill()};this.faceLandmarks.positions.forEach(u)}}};function Zle(e,t){(Array.isArray(t)?t:[t]).forEach(r=>{let s=r instanceof ar?r:Oi(r)?r.landmarks:void 0;if(!s)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new mg(s).draw(e)})}var JA="1.7.6";function ede(e,t){let n=ll(e,t),r=dl(e,t);function s(o,i,c){let u=r(o,i,`${c}/separable_conv0`),l=r(i,i,`${c}/separable_conv1`),p=n(o,i,1,`${c}/expansion_conv`);return{separable_conv0:u,separable_conv1:l,expansion_conv:p}}function a(o,i){let c=r(o,o,`${i}/separable_conv0`),u=r(o,o,`${i}/separable_conv1`),l=r(o,o,`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:u,separable_conv2:l}}return{extractConvParams:n,extractSeparableConvParams:r,extractReductionBlockParams:s,extractMainBlockParams:a}}function QA(e,t){let n=[],{extractWeights:r,getRemainingWeights:s}=Rn(e),{extractConvParams:a,extractSeparableConvParams:o,extractReductionBlockParams:i,extractMainBlockParams:c}=ede(r,n),u=a(3,32,3,"entry_flow/conv_in"),l=i(32,64,"entry_flow/reduction_block_0"),p=i(64,128,"entry_flow/reduction_block_1"),d={conv_in:u,reduction_block_0:l,reduction_block_1:p},h={};gs(t,0,1).forEach(b=>{h[`main_block_${b}`]=c(128,`middle_flow/main_block_${b}`)});let f=i(128,256,"exit_flow/reduction_block"),m=o(256,512,"exit_flow/separable_conv"),g={reduction_block:f,separable_conv:m};if(s().length!==0)throw new Error(`weights remaing after extract: ${s().length}`);return{paramMappings:n,params:{entry_flow:d,middle_flow:h,exit_flow:g}}}function tde(e,t){let n=or(e,t),r=lg(n),s=pl(n);function a(i){let c=s(`${i}/separable_conv0`),u=s(`${i}/separable_conv1`),l=r(`${i}/expansion_conv`);return{separable_conv0:c,separable_conv1:u,expansion_conv:l}}function o(i){let c=s(`${i}/separable_conv0`),u=s(`${i}/separable_conv1`),l=s(`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:u,separable_conv2:l}}return{extractConvParams:r,extractSeparableConvParams:s,extractReductionBlockParams:a,extractMainBlockParams:o}}function e$(e,t){let n=[],{extractConvParams:r,extractSeparableConvParams:s,extractReductionBlockParams:a,extractMainBlockParams:o}=tde(e,n),i=r("entry_flow/conv_in"),c=a("entry_flow/reduction_block_0"),u=a("entry_flow/reduction_block_1"),l={conv_in:i,reduction_block_0:c,reduction_block_1:u},p={};gs(t,0,1).forEach(m=>{p[`main_block_${m}`]=o(`middle_flow/main_block_${m}`)});let d=a("exit_flow/reduction_block"),h=s("exit_flow/separable_conv"),f={reduction_block:d,separable_conv:h};return Fn(e,n),{params:{entry_flow:l,middle_flow:p,exit_flow:f},paramMappings:n}}function t$(e,t,n){return Y(Dt(e,t.filters,n,"same"),t.bias)}function OI(e,t,n=!0){let r=n?Xe(e):e;return r=Hn(r,t.separable_conv0,[1,1]),r=Hn(Xe(r),t.separable_conv1,[1,1]),r=Rt(r,[3,3],[2,2],"same"),r=Y(r,t$(e,t.expansion_conv,[2,2])),r}function nde(e,t){let n=Hn(Xe(e),t.separable_conv0,[1,1]);return n=Hn(Xe(n),t.separable_conv1,[1,1]),n=Hn(Xe(n),t.separable_conv2,[1,1]),n=Y(n,e),n}var gg=class extends an{constructor(n){super("TinyXception");this._numMainBlocks=n}forwardInput(n){let{params:r}=this;if(!r)throw new Error("TinyXception - load model before inference");return O(()=>{let s=ce(n.toBatchTensor(112,!0),"float32"),o=Jr(s,[122.782,117.001,104.298]).div(255),i=Xe(t$(o,r.entry_flow.conv_in,[2,2]));return i=OI(i,r.entry_flow.reduction_block_0,!1),i=OI(i,r.entry_flow.reduction_block_1),gs(this._numMainBlocks,0,1).forEach(c=>{i=nde(i,r.middle_flow[`main_block_${c}`])}),i=OI(i,r.exit_flow.reduction_block),i=Xe(Hn(i,r.exit_flow.separable_conv,[1,1])),i})}async forward(n){return this.forwardInput(await bt(n))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(n){return e$(n,this._numMainBlocks)}extractParams(n){return QA(n,this._numMainBlocks)}};function n$(e){let t=[],{extractWeights:n,getRemainingWeights:r}=Rn(e),s=cg(n,t),a=s(512,1,"fc/age"),o=s(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:a,gender:o}}}}function r$(e){let t=[],n=or(e,t);function r(a){let o=n(`${a}/weights`,2),i=n(`${a}/bias`,1);return{weights:o,bias:i}}let s={fc:{age:r("fc/age"),gender:r("fc/gender")}};return Fn(e,t),{params:s,paramMappings:t}}var bg=(n=>(n.FEMALE="female",n.MALE="male",n))(bg||{});var wp=class extends an{constructor(n=new gg(2)){super("AgeGenderNet");this._faceFeatureExtractor=n}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(n){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return O(()=>{let s=n instanceof vs?this.faceFeatureExtractor.forwardInput(n):n,a=mr(s,[7,7],[2,2],"valid").as2D(s.shape[0],-1),o=vp(a,r.fc.age).as1D(),i=vp(a,r.fc.gender);return{age:o,gender:i}})}forwardInput(n){return O(()=>{let{age:r,gender:s}=this.runNet(n);return{age:r,gender:Xr(s)}})}async forward(n){return this.forwardInput(await bt(n))}async predictAgeAndGender(n){let r=await bt(n),s=await this.forwardInput(r),a=lt(s.age),o=lt(s.gender),i=a.map((u,l)=>({ageTensor:u,genderTensor:o[l]})),c=await Promise.all(i.map(async({ageTensor:u,genderTensor:l})=>{let p=u.dataSync()[0],d=l.dataSync()[0],h=d>.5,f=h?"male":"female",m=h?d:1-d;return u.dispose(),l.dispose(),{age:p,gender:f,genderProbability:m}}));return s.age.dispose(),s.gender.dispose(),r.isBatchInput?c:c[0]}getDefaultModelName(){return"age_gender_model"}dispose(n=!0){this.faceFeatureExtractor.dispose(n),super.dispose(n)}loadClassifierParams(n){let{params:r,paramMappings:s}=this.extractClassifierParams(n);this._params=r,this._paramMappings=s}extractClassifierParams(n){return n$(n)}extractParamsFromWeightMap(n){let{featureExtractorMap:r,classifierMap:s}=pg(n);return this.faceFeatureExtractor.loadFromWeightMap(r),r$(s)}extractParams(n){let s=n.slice(0,n.length-1539),a=n.slice(n.length-1539);return this.faceFeatureExtractor.extractWeights(s),this.extractClassifierParams(a)}};var gl=class extends fl{postProcess(t,n,r){let s=r.map(({width:o,height:i})=>{let c=n/Math.max(i,o);return{width:o*c,height:i*c}}),a=s.length;return O(()=>{let o=(p,d)=>Ft([bn([68],p,"float32"),bn([68],d,"float32")],1).as2D(1,136).as1D(),i=(p,d)=>{let{width:h,height:f}=s[p];return d(h,f)?Math.abs(h-f)/2:0},c=p=>i(p,(d,h)=>d<h),u=p=>i(p,(d,h)=>h<d);return t.mul(bn([a,136],n,"float32")).sub(Ft(Array.from(Array(a),(p,d)=>o(c(d),u(d))))).div(Ft(Array.from(Array(a),(p,d)=>o(s[d].width,s[d].height))))})}forwardInput(t){return O(()=>{let n=this.runNet(t);return this.postProcess(n,t.inputSize,t.inputDimensions.map(([r,s])=>({height:r,width:s})))})}async forward(t){return this.forwardInput(await bt(t))}async detectLandmarks(t){let n=await bt(t),r=O(()=>lt(this.forwardInput(n))),s=await Promise.all(r.map(async(a,o)=>{let i=Array.from(a.dataSync()),c=i.filter((l,p)=>tg(p)),u=i.filter((l,p)=>!tg(p));return new Ai(Array(68).fill(0).map((l,p)=>new Fe(c[p],u[p])),{height:n.getInputHeight(o),width:n.getInputWidth(o)})}));return r.forEach(a=>a.dispose()),n.isBatchInput?s:s[0]}getClassifierChannelsOut(){return 136}};var Mi=class extends gl{constructor(t=new hl){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};function s$(e){let t=[],{extractDenseBlock3Params:n}=dg(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return Fn(e,t),{params:r,paramMappings:t}}function a$(e){let t=[],{extractWeights:n,getRemainingWeights:r}=Rn(e),{extractDenseBlock3Params:s}=ug(n,t),a=s(3,32,"dense0",!0),o=s(32,64,"dense1"),i=s(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:o,dense2:i}}}var yg=class extends an{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("TinyFaceFeatureExtractor - load model before inference");return O(()=>{let r=ce(t.toBatchTensor(112,!0),"float32"),a=Jr(r,[122.782,117.001,104.298]).div(255),o=ig(a,n.dense0,!0);return o=ig(o,n.dense1),o=ig(o,n.dense2),o=mr(o,[14,14],[2,2],"valid"),o})}async forward(t){return this.forwardInput(await bt(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return s$(t)}extractParams(t){return a$(t)}};var Ip=class extends gl{constructor(t=new yg){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var MI=class extends Mi{};function o$(e,t){return Y(B(e,t.weights),t.biases)}function LI(e,t,n,r,s="same"){let{filters:a,bias:o}=t.conv,i=Dt(e,a,n,s);return i=Y(i,o),i=o$(i,t.scale),r?Xe(i):i}function i$(e,t){return LI(e,t,[1,1],!0)}function zI(e,t){return LI(e,t,[1,1],!1)}function vg(e,t){return LI(e,t,[2,2],!0,"valid")}function rde(e,t){function n(i,c,u){let l=e(i),p=l.length/(c*u*u);if(pI(p))throw new Error(`depth has to be an integer: ${p}, weights.length: ${l.length}, numFilters: ${c}, filterSize: ${u}`);return O(()=>Ee(Rr(l,[c,p,u,u]),[2,3,1,0]))}function r(i,c,u,l){let p=n(i,c,u),d=Ke(e(c));return t.push({paramPath:`${l}/filters`},{paramPath:`${l}/bias`}),{filters:p,bias:d}}function s(i,c){let u=Ke(e(i)),l=Ke(e(i));return t.push({paramPath:`${c}/weights`},{paramPath:`${c}/biases`}),{weights:u,biases:l}}function a(i,c,u,l){let p=r(i,c,u,`${l}/conv`),d=s(c,`${l}/scale`);return{conv:p,scale:d}}function o(i,c,u,l,p=!1){let d=a((p?.5:1)*i,c,u,`${l}/conv1`),h=a(i,c,u,`${l}/conv2`);return{conv1:d,conv2:h}}return{extractConvLayerParams:a,extractResidualLayerParams:o}}function c$(e){let{extractWeights:t,getRemainingWeights:n}=Rn(e),r=[],{extractConvLayerParams:s,extractResidualLayerParams:a}=rde(t,r),o=s(4704,32,7,"conv32_down"),i=a(9216,32,3,"conv32_1"),c=a(9216,32,3,"conv32_2"),u=a(9216,32,3,"conv32_3"),l=a(36864,64,3,"conv64_down",!0),p=a(36864,64,3,"conv64_1"),d=a(36864,64,3,"conv64_2"),h=a(36864,64,3,"conv64_3"),f=a(147456,128,3,"conv128_down",!0),m=a(147456,128,3,"conv128_1"),g=a(147456,128,3,"conv128_2"),b=a(589824,256,3,"conv256_down",!0),y=a(589824,256,3,"conv256_1"),v=a(589824,256,3,"conv256_2"),x=a(589824,256,3,"conv256_down_out"),k=O(()=>Ee($r(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{params:{conv32_down:o,conv32_1:i,conv32_2:c,conv32_3:u,conv64_down:l,conv64_1:p,conv64_2:d,conv64_3:h,conv128_down:f,conv128_1:m,conv128_2:g,conv256_down:b,conv256_1:y,conv256_2:v,conv256_down_out:x,fc:k},paramMappings:r}}function sde(e,t){let n=or(e,t);function r(o){let i=n(`${o}/scale/weights`,1),c=n(`${o}/scale/biases`,1);return{weights:i,biases:c}}function s(o){let i=n(`${o}/conv/filters`,4),c=n(`${o}/conv/bias`,1),u=r(o);return{conv:{filters:i,bias:c},scale:u}}function a(o){return{conv1:s(`${o}/conv1`),conv2:s(`${o}/conv2`)}}return{extractConvLayerParams:s,extractResidualLayerParams:a}}function u$(e){let t=[],{extractConvLayerParams:n,extractResidualLayerParams:r}=sde(e,t),s=n("conv32_down"),a=r("conv32_1"),o=r("conv32_2"),i=r("conv32_3"),c=r("conv64_down"),u=r("conv64_1"),l=r("conv64_2"),p=r("conv64_3"),d=r("conv128_down"),h=r("conv128_1"),f=r("conv128_2"),m=r("conv256_down"),g=r("conv256_1"),b=r("conv256_2"),y=r("conv256_down_out"),{fc:v}=e;if(t.push({originalPath:"fc",paramPath:"fc"}),!dI(v))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${v}`);let x={conv32_down:s,conv32_1:a,conv32_2:o,conv32_3:i,conv64_down:c,conv64_1:u,conv64_2:l,conv64_3:p,conv128_down:d,conv128_1:h,conv128_2:f,conv256_down:m,conv256_1:g,conv256_2:b,conv256_down_out:y,fc:v};return Fn(e,t),{params:x,paramMappings:t}}function Qr(e,t){let n=i$(e,t.conv1);return n=zI(n,t.conv2),n=Y(n,e),n=Xe(n),n}function kp(e,t){let n=vg(e,t.conv1);n=zI(n,t.conv2);let r=mr(e,2,2,"valid"),s=kt(r.shape),a=r.shape[3]!==n.shape[3];if(r.shape[1]!==n.shape[1]||r.shape[2]!==n.shape[2]){let i=[...n.shape];i[1]=1;let c=kt(i);n=Je([n,c],1);let u=[...n.shape];u[2]=1;let l=kt(u);n=Je([n,l],2)}return r=a?Je([r,s],3):r,n=Y(r,n),n=Xe(n),n}var Li=class extends an{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("FaceRecognitionNet - load model before inference");return O(()=>{let r=ce(t.toBatchTensor(150,!0),"float32"),a=Jr(r,[122.782,117.001,104.298]).div(255),o=vg(a,n.conv32_down);o=Rt(o,3,2,"valid"),o=Qr(o,n.conv32_1),o=Qr(o,n.conv32_2),o=Qr(o,n.conv32_3),o=kp(o,n.conv64_down),o=Qr(o,n.conv64_1),o=Qr(o,n.conv64_2),o=Qr(o,n.conv64_3),o=kp(o,n.conv128_down),o=Qr(o,n.conv128_1),o=Qr(o,n.conv128_2),o=kp(o,n.conv256_down),o=Qr(o,n.conv256_1),o=Qr(o,n.conv256_2),o=kp(o,n.conv256_down_out);let i=o.mean([1,2]);return De(i,n.fc)})}async forward(t){return this.forwardInput(await bt(t))}async computeFaceDescriptor(t){var a;if((a=t==null?void 0:t.shape)!=null&&a.some(o=>o<=0))return new Float32Array(128);let n=await bt(t),r=O(()=>lt(this.forwardInput(n))),s=await Promise.all(r.map(o=>o.data()));return r.forEach(o=>o.dispose()),n.isBatchInput?s:s[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return u$(t)}extractParams(t){return c$(t)}};function ade(e){let t=new Li;return t.extractWeights(e),t}function xg(e,t){return{...e,...{descriptor:t}}}function ode(e){return typeof e.age=="number"}function wg(e,t){return{...e,...{age:t}}}function ide(e){return(e.gender==="male"||e.gender==="female")&&sl(e.genderProbability)}function Ig(e,t,n){return{...e,...{gender:t,genderProbability:n}}}function cde(e,t){function n(c,u){let l=Rr(e(9*c),[3,3,c,1]),p=Ke(e(c)),d=Ke(e(c)),h=Ke(e(c)),f=Ke(e(c));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/batch_norm_scale`},{paramPath:`${u}/batch_norm_offset`},{paramPath:`${u}/batch_norm_mean`},{paramPath:`${u}/batch_norm_variance`}),{filters:l,batch_norm_scale:p,batch_norm_offset:d,batch_norm_mean:h,batch_norm_variance:f}}function r(c,u,l,p,d){let h=Rr(e(c*u*l*l),[l,l,c,u]),f=Ke(e(u));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/${d?"batch_norm_offset":"bias"}`}),{filters:h,bias:f}}function s(c,u,l,p){let{filters:d,bias:h}=r(c,u,l,p,!0);return{filters:d,batch_norm_offset:h}}function a(c,u,l){let p=n(c,`${l}/depthwise_conv`),d=s(c,u,1,`${l}/pointwise_conv`);return{depthwise_conv:p,pointwise_conv:d}}function o(){let c=s(3,32,3,"mobilenetv1/conv_0"),u=a(32,64,"mobilenetv1/conv_1"),l=a(64,128,"mobilenetv1/conv_2"),p=a(128,128,"mobilenetv1/conv_3"),d=a(128,256,"mobilenetv1/conv_4"),h=a(256,256,"mobilenetv1/conv_5"),f=a(256,512,"mobilenetv1/conv_6"),m=a(512,512,"mobilenetv1/conv_7"),g=a(512,512,"mobilenetv1/conv_8"),b=a(512,512,"mobilenetv1/conv_9"),y=a(512,512,"mobilenetv1/conv_10"),v=a(512,512,"mobilenetv1/conv_11"),x=a(512,1024,"mobilenetv1/conv_12"),k=a(1024,1024,"mobilenetv1/conv_13");return{conv_0:c,conv_1:u,conv_2:l,conv_3:p,conv_4:d,conv_5:h,conv_6:f,conv_7:m,conv_8:g,conv_9:b,conv_10:y,conv_11:v,conv_12:x,conv_13:k}}function i(){let c=s(1024,256,1,"prediction_layer/conv_0"),u=s(256,512,3,"prediction_layer/conv_1"),l=s(512,128,1,"prediction_layer/conv_2"),p=s(128,256,3,"prediction_layer/conv_3"),d=s(256,128,1,"prediction_layer/conv_4"),h=s(128,256,3,"prediction_layer/conv_5"),f=s(256,64,1,"prediction_layer/conv_6"),m=s(64,128,3,"prediction_layer/conv_7"),g=r(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),b=r(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),y=r(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),v=r(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),x=r(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),k=r(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),S=r(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),C=r(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),E=r(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),$=r(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),F=r(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),A=r(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:c,conv_1:u,conv_2:l,conv_3:p,conv_4:d,conv_5:h,conv_6:f,conv_7:m,box_predictor_0:{box_encoding_predictor:g,class_predictor:b},box_predictor_1:{box_encoding_predictor:y,class_predictor:v},box_predictor_2:{box_encoding_predictor:x,class_predictor:k},box_predictor_3:{box_encoding_predictor:S,class_predictor:C},box_predictor_4:{box_encoding_predictor:E,class_predictor:$},box_predictor_5:{box_encoding_predictor:F,class_predictor:A}}}return{extractMobilenetV1Params:o,extractPredictionLayerParams:i}}function l$(e){let t=[],{extractWeights:n,getRemainingWeights:r}=Rn(e),{extractMobilenetV1Params:s,extractPredictionLayerParams:a}=cde(n,t),o=s(),i=a(),u={extra_dim:Rd(n(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:o,prediction_layer:i,output_layer:u},paramMappings:t}}function ude(e,t){let n=or(e,t);function r(u,l,p){let d=n(`${u}/Conv2d_${l}_pointwise/weights`,4,`${p}/filters`),h=n(`${u}/Conv2d_${l}_pointwise/convolution_bn_offset`,1,`${p}/batch_norm_offset`);return{filters:d,batch_norm_offset:h}}function s(u){let l=`mobilenetv1/conv_${u}`,p=`MobilenetV1/Conv2d_${u}_depthwise`,d=`${l}/depthwise_conv`,h=`${l}/pointwise_conv`,f=n(`${p}/depthwise_weights`,4,`${d}/filters`),m=n(`${p}/BatchNorm/gamma`,1,`${d}/batch_norm_scale`),g=n(`${p}/BatchNorm/beta`,1,`${d}/batch_norm_offset`),b=n(`${p}/BatchNorm/moving_mean`,1,`${d}/batch_norm_mean`),y=n(`${p}/BatchNorm/moving_variance`,1,`${d}/batch_norm_variance`);return{depthwise_conv:{filters:f,batch_norm_scale:m,batch_norm_offset:g,batch_norm_mean:b,batch_norm_variance:y},pointwise_conv:r("MobilenetV1",u,h)}}function a(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:s(1),conv_2:s(2),conv_3:s(3),conv_4:s(4),conv_5:s(5),conv_6:s(6),conv_7:s(7),conv_8:s(8),conv_9:s(9),conv_10:s(10),conv_11:s(11),conv_12:s(12),conv_13:s(13)}}function o(u,l){let p=n(`${u}/weights`,4,`${l}/filters`),d=n(`${u}/biases`,1,`${l}/bias`);return{filters:p,bias:d}}function i(u){let l=o(`Prediction/BoxPredictor_${u}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${u}/box_encoding_predictor`),p=o(`Prediction/BoxPredictor_${u}/ClassPredictor`,`prediction_layer/box_predictor_${u}/class_predictor`);return{box_encoding_predictor:l,class_predictor:p}}function c(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}return{extractMobilenetV1Params:a,extractPredictionLayerParams:c}}function d$(e){let t=[],{extractMobilenetV1Params:n,extractPredictionLayerParams:r}=ude(e,t),s=e["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!Ms(s))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${s}`);let a={mobilenetv1:n(),prediction_layer:r(),output_layer:{extra_dim:s}};return Fn(e,t),{params:a,paramMappings:t}}function Mr(e,t,n){return O(()=>{let r=Dt(e,t.filters,n,"same");return r=Y(r,t.batch_norm_offset),en(r,0,6)})}var lde=.0010000000474974513;function dde(e,t,n){return O(()=>{let r=Sa(e,t.filters,n,"same");return r=ka(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,lde),en(r,0,6)})}function pde(e){return[2,4,6,12].some(t=>t===e)?[2,2]:[1,1]}function p$(e,t){return O(()=>{let n,r=Mr(e,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((a,o)=>{let i=o+1,c=pde(i);r=dde(r,a.depthwise_conv,c),r=Mr(r,a.pointwise_conv,[1,1]),i===11&&(n=r)}),n===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:n}})}function hde(e,t,n){let r=e.arraySync(),s=Math.min(r[t][0],r[t][2]),a=Math.min(r[t][1],r[t][3]),o=Math.max(r[t][0],r[t][2]),i=Math.max(r[t][1],r[t][3]),c=Math.min(r[n][0],r[n][2]),u=Math.min(r[n][1],r[n][3]),l=Math.max(r[n][0],r[n][2]),p=Math.max(r[n][1],r[n][3]),d=(o-s)*(i-a),h=(l-c)*(p-u);if(d<=0||h<=0)return 0;let f=Math.max(s,c),m=Math.max(a,u),g=Math.min(o,l),b=Math.min(i,p),y=Math.max(g-f,0)*Math.max(b-m,0);return y/(d+h-y)}function h$(e,t,n,r,s){let a=e.shape[0],o=Math.min(n,a),i=t.map((l,p)=>({score:l,boxIndex:p})).filter(l=>l.score>s).sort((l,p)=>p.score-l.score),c=l=>l<=r?1:0,u=[];return i.forEach(l=>{if(u.length>=o)return;let p=l.score;for(let d=u.length-1;d>=0;--d){let h=hde(e,l.boxIndex,u[d]);if(h!==0&&(l.score*=c(h),l.score<=s))break}p===l.score&&u.push(l.boxIndex)}),u}function fde(e){let t=lt(Ee(e,[1,0])),n=[de(t[2],t[0]),de(t[3],t[1])],r=[Y(t[0],fe(n[0],2)),Y(t[1],fe(n[1],2))];return{sizes:n,centers:r}}function mde(e,t){let{sizes:n,centers:r}=fde(e),s=lt(Ee(t,[1,0])),a=fe(B(gn(fe(s[2],5)),n[0]),2),o=Y(B(fe(s[0],10),n[0]),r[0]),i=fe(B(gn(fe(s[3],5)),n[1]),2),c=Y(B(fe(s[1],10),n[1]),r[1]);return Ee(Ft([de(o,a),de(c,i),Y(o,a),Y(c,i)]),[1,0])}function f$(e,t,n){return O(()=>{let r=e.shape[0],s=mde(W(Ln(n.extra_dim,[r,1,1]),[-1,4]),W(e,[-1,4]));s=W(s,[r,s.shape[0]/r,4]);let a=dr(We(t,[0,0,1],[-1,-1,-1])),o=We(a,[0,0,0],[-1,-1,1]);o=W(o,[r,o.shape[1]]);let i=lt(s),c=lt(o);return{boxes:i,scores:c}})}function zi(e,t){return O(()=>{let n=e.shape[0],r=W(Pi(e,t.box_encoding_predictor),[n,-1,1,4]),s=W(Pi(e,t.class_predictor),[n,-1,3]);return{boxPredictionEncoding:r,classPrediction:s}})}function m$(e,t,n){return O(()=>{let r=Mr(e,n.conv_0,[1,1]),s=Mr(r,n.conv_1,[2,2]),a=Mr(s,n.conv_2,[1,1]),o=Mr(a,n.conv_3,[2,2]),i=Mr(o,n.conv_4,[1,1]),c=Mr(i,n.conv_5,[2,2]),u=Mr(c,n.conv_6,[1,1]),l=Mr(u,n.conv_7,[2,2]),p=zi(t,n.box_predictor_0),d=zi(e,n.box_predictor_1),h=zi(s,n.box_predictor_2),f=zi(o,n.box_predictor_3),m=zi(c,n.box_predictor_4),g=zi(l,n.box_predictor_5),b=Je([p.boxPredictionEncoding,d.boxPredictionEncoding,h.boxPredictionEncoding,f.boxPredictionEncoding,m.boxPredictionEncoding,g.boxPredictionEncoding],1),y=Je([p.classPrediction,d.classPrediction,h.classPrediction,f.classPrediction,m.classPrediction,g.classPrediction],1);return{boxPredictions:b,classPredictions:y}})}var vr=class{constructor({minConfidence:t,maxResults:n}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=n||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var $a=class extends an{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:n}=this;if(!n)throw new Error("SsdMobilenetv1 - load model before inference");return O(()=>{let r=ce(t.toBatchTensor(512,!1),"float32"),s=de(fe(r,127.5),1),a=p$(s,n.mobilenetv1),{boxPredictions:o,classPredictions:i}=m$(a.out,a.conv11,n.prediction_layer);return f$(o,i,n.output_layer)})}async forward(t){return this.forwardInput(await bt(t))}async locateFaces(t,n={}){let{maxResults:r,minConfidence:s}=new vr(n),a=await bt(t),{boxes:o,scores:i}=this.forwardInput(a),c=o[0],u=i[0];for(let v=1;v<o.length;v++)o[v].dispose(),i[v].dispose();let l=Array.from(u.dataSync()),d=h$(c,l,r,.5,s),h=a.getReshapedInputDimensions(0),f=a.inputSize,m=f/h.width,g=f/h.height,b=c.arraySync(),y=d.map(v=>{let[x,k]=[Math.max(0,b[v][0]),Math.min(1,b[v][2])].map(E=>E*g),[S,C]=[Math.max(0,b[v][1]),Math.min(1,b[v][3])].map(E=>E*m);return new gt(l[v],new Ei(S,x,C-S,k-x),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return c.dispose(),u.dispose(),y}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return d$(t)}extractParams(t){return l$(t)}};function g$(e){let t=new $a;return t.extractWeights(e),t}function gde(e){return g$(e)}var BI=class extends $a{};var b$=.4,y$=[new Fe(.738768,.874946),new Fe(2.42204,2.65704),new Fe(4.30971,7.04493),new Fe(10.246,4.59428),new Fe(12.6868,11.8741)],v$=[new Fe(1.603231,2.094468),new Fe(6.041143,7.080126),new Fe(2.882459,3.518061),new Fe(4.266906,5.178857),new Fe(9.041765,10.66308)],x$=[117.001,114.697,97.404],w$="tiny_yolov2_model",I$="tiny_yolov2_separable_conv_model";var kg=e=>typeof e=="number";function WI(e){if(!e)throw new Error(`invalid config: ${e}`);if(typeof e.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${e.withSeparableConvs}`);if(!kg(e.iouThreshold)||e.iouThreshold<0||e.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${e.iouThreshold}`);if(!Array.isArray(e.classes)||!e.classes.length||!e.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(e.classes)}`);if(!Array.isArray(e.anchors)||!e.anchors.length||!e.anchors.map(t=>t||{}).every(t=>kg(t.x)&&kg(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(e.anchors)}`);if(e.meanRgb&&(!Array.isArray(e.meanRgb)||e.meanRgb.length!==3||!e.meanRgb.every(kg)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(e.meanRgb)}`)}function bl(e){return O(()=>{let t=B(e,xe(.10000000149011612));return Y(Xe(de(e,t)),t)})}function Vs(e,t){return O(()=>{let n=gr(e,[[0,0],[1,1],[1,1],[0,0]]);return n=Dt(n,t.conv.filters,[1,1],"valid"),n=de(n,t.bn.sub),n=B(n,t.bn.truediv),n=Y(n,t.conv.bias),bl(n)})}function Us(e,t){return O(()=>{let n=gr(e,[[0,0],[1,1],[1,1],[0,0]]);return n=Ca(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=Y(n,t.bias),bl(n)})}function bde(e,t){let n=ll(e,t);function r(o,i){let c=Ke(e(o)),u=Ke(e(o));return t.push({paramPath:`${i}/sub`},{paramPath:`${i}/truediv`}),{sub:c,truediv:u}}function s(o,i,c){let u=n(o,i,3,`${c}/conv`),l=r(i,`${c}/bn`);return{conv:u,bn:l}}let a=dl(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:s,extractSeparableConvParams:a}}function k$(e,t,n,r){let{extractWeights:s,getRemainingWeights:a}=Rn(e),o=[],{extractConvParams:i,extractConvWithBatchNormParams:c,extractSeparableConvParams:u}=bde(s,o),l;if(t.withSeparableConvs){let[p,d,h,f,m,g,b,y,v]=r,x=t.isFirstLayerConv2d?i(p,d,3,"conv0"):u(p,d,"conv0"),k=u(d,h,"conv1"),S=u(h,f,"conv2"),C=u(f,m,"conv3"),E=u(m,g,"conv4"),$=u(g,b,"conv5"),F=y?u(b,y,"conv6"):void 0,A=v?u(y,v,"conv7"):void 0,R=i(v||y||b,5*n,1,"conv8");l={conv0:x,conv1:k,conv2:S,conv3:C,conv4:E,conv5:$,conv6:F,conv7:A,conv8:R}}else{let[p,d,h,f,m,g,b,y,v]=r,x=c(p,d,"conv0"),k=c(d,h,"conv1"),S=c(h,f,"conv2"),C=c(f,m,"conv3"),E=c(m,g,"conv4"),$=c(g,b,"conv5"),F=c(b,y,"conv6"),A=c(y,v,"conv7"),R=i(v,5*n,1,"conv8");l={conv0:x,conv1:k,conv2:S,conv3:C,conv4:E,conv5:$,conv6:F,conv7:A,conv8:R}}if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:l,paramMappings:o}}function yde(e,t){let n=or(e,t);function r(i){let c=n(`${i}/sub`,1),u=n(`${i}/truediv`,1);return{sub:c,truediv:u}}function s(i){let c=n(`${i}/filters`,4),u=n(`${i}/bias`,1);return{filters:c,bias:u}}function a(i){let c=s(`${i}/conv`),u=r(`${i}/bn`);return{conv:c,bn:u}}let o=pl(n);return{extractConvParams:s,extractConvWithBatchNormParams:a,extractSeparableConvParams:o}}function S$(e,t){let n=[],{extractConvParams:r,extractConvWithBatchNormParams:s,extractSeparableConvParams:a}=yde(e,n),o;if(t.withSeparableConvs){let i=t.filterSizes&&t.filterSizes.length||9;o={conv0:t.isFirstLayerConv2d?r("conv0"):a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:i>7?a("conv6"):void 0,conv7:i>8?a("conv7"):void 0,conv8:r("conv8")}}else o={conv0:s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:s("conv6"),conv7:s("conv7"),conv8:r("conv8")};return Fn(e,n),{params:o,paramMappings:n}}var es=class{constructor({inputSize:t,scoreThreshold:n}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=n||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var VI=class extends an{constructor(n){super("TinyYolov2");WI(n),this._config=n}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(n,r){let s=Vs(n,r.conv0);return s=Rt(s,[2,2],[2,2],"same"),s=Vs(s,r.conv1),s=Rt(s,[2,2],[2,2],"same"),s=Vs(s,r.conv2),s=Rt(s,[2,2],[2,2],"same"),s=Vs(s,r.conv3),s=Rt(s,[2,2],[2,2],"same"),s=Vs(s,r.conv4),s=Rt(s,[2,2],[2,2],"same"),s=Vs(s,r.conv5),s=Rt(s,[2,2],[1,1],"same"),s=Vs(s,r.conv6),s=Vs(s,r.conv7),Pi(s,r.conv8,"valid",!1)}runMobilenet(n,r){let s=this.config.isFirstLayerConv2d?bl(Pi(n,r.conv0,"valid",!1)):Us(n,r.conv0);return s=Rt(s,[2,2],[2,2],"same"),s=Us(s,r.conv1),s=Rt(s,[2,2],[2,2],"same"),s=Us(s,r.conv2),s=Rt(s,[2,2],[2,2],"same"),s=Us(s,r.conv3),s=Rt(s,[2,2],[2,2],"same"),s=Us(s,r.conv4),s=Rt(s,[2,2],[2,2],"same"),s=Us(s,r.conv5),s=Rt(s,[2,2],[1,1],"same"),s=r.conv6?Us(s,r.conv6):s,s=r.conv7?Us(s,r.conv7):s,Pi(s,r.conv8,"valid",!1)}forwardInput(n,r){let{params:s}=this;if(!s)throw new Error("TinyYolov2 - load model before inference");return O(()=>{let a=ce(n.toBatchTensor(r,!1),"float32");return a=this.config.meanRgb?Jr(a,this.config.meanRgb):a,a=a.div(255),this.config.withSeparableConvs?this.runMobilenet(a,s):this.runTinyYolov2(a,s)})}async forward(n,r){return this.forwardInput(await bt(n),r)}async detect(n,r={}){let{inputSize:s,scoreThreshold:a}=new es(r),o=await bt(n),i=await this.forwardInput(o,s),c=O(()=>lt(i)[0].expandDims()),u={width:o.getInputWidth(0),height:o.getInputHeight(0)},l=await this.extractBoxes(c,o.getReshapedInputDimensions(0),a);i.dispose(),c.dispose();let p=l.map(b=>b.box),d=l.map(b=>b.score),h=l.map(b=>b.classScore),f=l.map(b=>this.config.classes[b.label]);return bI(p.map(b=>b.rescale(s)),d,this.config.iouThreshold,!0).map(b=>new Ls(d[b],h[b],f[b],p[b],u))}getDefaultModelName(){return""}extractParamsFromWeightMap(n){return S$(n,this.config)}extractParams(n){let r=this.config.filterSizes||VI.DEFAULT_FILTER_SIZES,s=r?r.length:void 0;if(s!==7&&s!==8&&s!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${s} filterSizes in config`);return k$(n,this.config,this.boxEncodingSize,r)}async extractBoxes(n,r,s){let{width:a,height:o}=r,i=Math.max(a,o),c=i/a,u=i/o,l=n.shape[1],p=this.config.anchors.length,[d,h,f]=O(()=>{let y=n.reshape([l,l,p,this.boxEncodingSize]),v=y.slice([0,0,0,0],[l,l,p,4]),x=y.slice([0,0,0,4],[l,l,p,1]),k=this.withClassScores?Xr(y.slice([0,0,0,5],[l,l,p,this.config.classes.length]),3):xe(0);return[v,x,k]}),m=[],g=await h.array(),b=await d.array();for(let y=0;y<l;y++)for(let v=0;v<l;v++)for(let x=0;x<p;x++){let k=pp(g[y][v][x][0]);if(!s||k>s){let S=(v+pp(b[y][v][x][0]))/l*c,C=(y+pp(b[y][v][x][1]))/l*u,E=Math.exp(b[y][v][x][2])*this.config.anchors[x].x/l*c,$=Math.exp(b[y][v][x][3])*this.config.anchors[x].y/l*u,F=S-E/2,A=C-$/2,R={row:y,col:v,anchor:x},{classScore:T,label:L}=this.withClassScores?await this.extractPredictedClass(f,R):{classScore:1,label:0};m.push({box:new _i(F,A,F+E,A+$),score:k,classScore:k*T,label:L,...R})}}return d.dispose(),h.dispose(),f.dispose(),m}async extractPredictedClass(n,r){let{row:s,col:a,anchor:o}=r,i=await n.array();return Array(this.config.classes.length).fill(0).map((c,u)=>i[s][a][o][u]).map((c,u)=>({classScore:c,label:u})).reduce((c,u)=>c.classScore>u.classScore?c:u)}},Bi=VI;Bi.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var Wi=class extends Bi{constructor(t=!0){let n={withSeparableConvs:t,iouThreshold:b$,classes:["face"],...t?{anchors:v$,meanRgb:x$}:{anchors:y$,withClassScores:!0}};super(n)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(s=>new gt(s.score,s.relativeBox,{width:s.imageWidth,height:s.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?I$:w$}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function vde(e,t=!0){let n=new Wi(t);return n.extractWeights(e),n}var Sp=class extends es{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var xr=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function Vi(e,t,n,r,s=({alignedRect:a})=>a){let a=e.map(c=>Oi(c)?s(c):c.detection),o=r||(t instanceof Te?await ul(t,a):await cl(t,a)),i=await n(o);return o.forEach(c=>c instanceof Te&&c.dispose()),i}async function yl(e,t,n,r,s){return Vi([e],t,async a=>n(a[0]),r,s)}var T$=.4,C$=[new Fe(1.603231,2.094468),new Fe(6.041143,7.080126),new Fe(2.882459,3.518061),new Fe(4.266906,5.178857),new Fe(9.041765,10.66308)],N$=[117.001,114.697,97.404];var Ui=class extends Bi{constructor(){let t={withSeparableConvs:!0,iouThreshold:T$,classes:["face"],anchors:C$,meanRgb:N$,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,n){return(await this.detect(t,n)).map(s=>new gt(s.score,s.relativeBox,{width:s.imageWidth,height:s.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var et={ssdMobilenetv1:new $a,tinyFaceDetector:new Ui,tinyYolov2:new Wi,faceLandmark68Net:new Mi,faceLandmark68TinyNet:new Ip,faceRecognitionNet:new Li,faceExpressionNet:new xp,ageGenderNet:new wp},_$=(e,t)=>et.ssdMobilenetv1.locateFaces(e,t),xde=(e,t)=>et.tinyFaceDetector.locateFaces(e,t),wde=(e,t)=>et.tinyYolov2.locateFaces(e,t),E$=e=>et.faceLandmark68Net.detectLandmarks(e),Ide=e=>et.faceLandmark68TinyNet.detectLandmarks(e),kde=e=>et.faceRecognitionNet.computeFaceDescriptor(e),Sde=e=>et.faceExpressionNet.predictExpressions(e),Tde=e=>et.ageGenderNet.predictAgeAndGender(e),A$=e=>et.ssdMobilenetv1.load(e),Cde=e=>et.tinyFaceDetector.load(e),Nde=e=>et.tinyYolov2.load(e),_de=e=>et.faceLandmark68Net.load(e),Ede=e=>et.faceLandmark68TinyNet.load(e),Ade=e=>et.faceRecognitionNet.load(e),$de=e=>et.faceExpressionNet.load(e),Dde=e=>et.ageGenderNet.load(e),Fde=A$,Rde=_$,Pde=E$;var Sg=class extends xr{constructor(n,r,s){super();this.parentTask=n;this.input=r;this.extractedFaces=s}},Gi=class extends Sg{async run(){let t=await this.parentTask,n=await Vi(t,this.input,async r=>Promise.all(r.map(s=>et.faceExpressionNet.predictExpressions(s))),this.extractedFaces);return t.map((r,s)=>hg(r,n[s]))}withAgeAndGender(){return new qi(this,this.input)}},Hi=class extends Sg{async run(){let t=await this.parentTask;if(!t)return;let n=await yl(t,this.input,r=>et.faceExpressionNet.predictExpressions(r),this.extractedFaces);return hg(t,n)}withAgeAndGender(){return new ji(this,this.input)}},Da=class extends Gi{withAgeAndGender(){return new Ra(this,this.input)}withFaceDescriptors(){return new Gs(this,this.input)}},Fa=class extends Hi{withAgeAndGender(){return new Pa(this,this.input)}withFaceDescriptor(){return new Hs(this,this.input)}};var Tg=class extends xr{constructor(n,r,s){super();this.parentTask=n;this.input=r;this.extractedFaces=s}},qi=class extends Tg{async run(){let t=await this.parentTask,n=await Vi(t,this.input,async r=>Promise.all(r.map(s=>et.ageGenderNet.predictAgeAndGender(s))),this.extractedFaces);return t.map((r,s)=>{let{age:a,gender:o,genderProbability:i}=n[s];return wg(Ig(r,o,i),a)})}withFaceExpressions(){return new Gi(this,this.input)}},ji=class extends Tg{async run(){let t=await this.parentTask;if(!t)return;let{age:n,gender:r,genderProbability:s}=await yl(t,this.input,a=>et.ageGenderNet.predictAgeAndGender(a),this.extractedFaces);return wg(Ig(t,r,s),n)}withFaceExpressions(){return new Hi(this,this.input)}},Ra=class extends qi{withFaceExpressions(){return new Da(this,this.input)}withFaceDescriptors(){return new Gs(this,this.input)}},Pa=class extends ji{withFaceExpressions(){return new Fa(this,this.input)}withFaceDescriptor(){return new Hs(this,this.input)}};var Tp=class extends xr{constructor(n,r){super();this.parentTask=n;this.input=r}},Gs=class extends Tp{async run(){let t=await this.parentTask;return(await Vi(t,this.input,r=>Promise.all(r.map(s=>et.faceRecognitionNet.computeFaceDescriptor(s))),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}))).map((r,s)=>xg(t[s],r))}withFaceExpressions(){return new Da(this,this.input)}withAgeAndGender(){return new Ra(this,this.input)}},Hs=class extends Tp{async run(){let t=await this.parentTask;if(!t)return;let n=await yl(t,this.input,r=>et.faceRecognitionNet.computeFaceDescriptor(r),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}));return xg(t,n)}withFaceExpressions(){return new Fa(this,this.input)}withAgeAndGender(){return new Pa(this,this.input)}};var Cp=class extends xr{constructor(n,r,s){super();this.parentTask=n;this.input=r;this.useTinyLandmarkNet=s}get landmarkNet(){return this.useTinyLandmarkNet?et.faceLandmark68TinyNet:et.faceLandmark68Net}},Np=class extends Cp{async run(){let t=await this.parentTask,n=t.map(o=>o.detection),r=this.input instanceof Te?await ul(this.input,n):await cl(this.input,n),s=await Promise.all(r.map(o=>this.landmarkNet.detectLandmarks(o)));return r.forEach(o=>o instanceof Te&&o.dispose()),t.filter((o,i)=>s[i]).map((o,i)=>ml(o,s[i]))}withFaceExpressions(){return new Da(this,this.input)}withAgeAndGender(){return new Ra(this,this.input)}withFaceDescriptors(){return new Gs(this,this.input)}},_p=class extends Cp{async run(){let t=await this.parentTask;if(!t)return;let{detection:n}=t,r=this.input instanceof Te?await ul(this.input,[n]):await cl(this.input,[n]),s=await this.landmarkNet.detectLandmarks(r[0]);return r.forEach(a=>a instanceof Te&&a.dispose()),ml(t,s)}withFaceExpressions(){return new Fa(this,this.input)}withAgeAndGender(){return new Pa(this,this.input)}withFaceDescriptor(){return new Hs(this,this.input)}};var Ep=class extends xr{constructor(n,r=new vr){super();this.input=n;this.options=r}},vl=class extends Ep{async run(){let{input:t,options:n}=this,r;if(n instanceof Sp)r=et.tinyFaceDetector.locateFaces(t,n);else if(n instanceof vr)r=et.ssdMobilenetv1.locateFaces(t,n);else if(n instanceof es)r=et.tinyYolov2.locateFaces(t,n);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return r}runAndExtendWithFaceDetections(){return new Promise((t,n)=>{this.run().then(r=>t(r.map(s=>$i({},s)))).catch(r=>n(r))})}withFaceLandmarks(t=!1){return new Np(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new Gi(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new qi(this.runAndExtendWithFaceDetections(),this.input)}},Ap=class extends Ep{async run(){let t=await new vl(this.input,this.options),n=t[0];return t.forEach(r=>{r.score>n.score&&(n=r)}),n}runAndExtendWithFaceDetection(){return new Promise(async t=>{let n=await this.run();t(n?$i({},n):void 0)})}withFaceLandmarks(t=!1){return new _p(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new Hi(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new ji(this.runAndExtendWithFaceDetection(),this.input)}};function Ode(e,t=new vr){return new Ap(e,t)}function Cg(e,t=new vr){return new vl(e,t)}async function $$(e,t){return Cg(e,new vr(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function Mde(e,t={}){return Cg(e,new es(t)).withFaceLandmarks().withFaceDescriptors()}var Lde=$$;function UI(e,t){if(e.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let n=Array.from(e),r=Array.from(t);return Math.sqrt(n.map((s,a)=>s-r[a]).reduce((s,a)=>s+a*a,0))}var $p=class{constructor(t,n=.6){this._distanceThreshold=n;let r=Array.isArray(t)?t:[t];if(!r.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let s=1,a=()=>`person ${s++}`;this._labeledDescriptors=r.map(o=>{if(o instanceof bs)return o;if(o instanceof Float32Array)return new bs(a(),[o]);if(o.descriptor&&o.descriptor instanceof Float32Array)return new bs(a(),[o.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,n){return n.map(r=>UI(r,t)).reduce((r,s)=>r+s,0)/(n.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:n,label:r})=>new al(r,this.computeMeanDistance(t,n))).reduce((n,r)=>n.distance<r.distance?n:r)}findBestMatch(t){let n=this.matchDescriptor(t);return n.distance<this._distanceThreshold?n:new al("unknown",n.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let n=t.labeledDescriptors.map(r=>bs.fromJSON(r));return new $p(n,t.distanceThreshold)}};function zde(e){let t=new Ui;return t.extractWeights(e),t}function D$(e,t){let{width:n,height:r}=new yn(t.width,t.height);if(n<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:n,height:r})}`);if(Array.isArray(e))return e.map(s=>D$(s,{width:n,height:r}));if(Oi(e)){let s=e.detection.forSize(n,r),a=e.unshiftedLandmarks.forSize(s.box.width,s.box.height);return ml($i(e,s),a)}return ys(e)?$i(e,e.detection.forSize(n,r)):e instanceof ar||e instanceof gt?e.forSize(n,r):e}var Bde=JA;return gD(Wde);})();
diff --git a/dist/face-api.node-gpu.js b/dist/face-api.node-gpu.js
index 1430d31..7b88333 100644
--- a/dist/face-api.node-gpu.js
+++ b/dist/face-api.node-gpu.js
@@ -4,4 +4,4926 @@
   author: <https://github.com/vladmandic>'
 */
 
-"use strict";var dn=Object.create;var er=Object.defineProperty;var hn=Object.getOwnPropertyDescriptor;var bn=Object.getOwnPropertyNames;var gn=Object.getPrototypeOf,xn=Object.prototype.hasOwnProperty;var vn=(o,t)=>()=>(t||o((t={exports:{}}).exports,t),t.exports),Lr=(o,t)=>{for(var e in t)er(o,e,{get:t[e],enumerable:!0})},To=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of bn(t))!xn.call(o,n)&&n!==e&&er(o,n,{get:()=>t[n],enumerable:!(r=hn(t,n))||r.enumerable});return o};var v=(o,t,e)=>(e=o!=null?dn(gn(o)):{},To(t||!o||!o.__esModule?er(e,"default",{value:o,enumerable:!0}):e,o)),yn=o=>To(er({},"__esModule",{value:!0}),o);var x=vn((Va,Br)=>{"use strict";var Wr=Object.defineProperty,_n=Object.getOwnPropertyDescriptor,Tn=Object.getOwnPropertyNames,wn=Object.prototype.hasOwnProperty,Pn=(o,t)=>{for(var e in t)Wr(o,e,{get:t[e],enumerable:!0})},Ar=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of Tn(t))!wn.call(o,n)&&n!==e&&Wr(o,n,{get:()=>t[n],enumerable:!(r=_n(t,n))||r.enumerable});return o},Fn=(o,t,e)=>(Ar(o,t,"default"),e&&Ar(e,t,"default")),Dn=o=>Ar(Wr({},"__esModule",{value:!0}),o),kr={};Pn(kr,{version:()=>Nn});Br.exports=Dn(kr);Fn(kr,require("@tensorflow/tfjs-node-gpu"),Br.exports);var wo="4.0.0",En="4.0.0",Mn="4.0.0",Cn="4.0.0",In="4.0.0",Nn={tfjs:wo,"tfjs-core":wo,"tfjs-converter":En,"tfjs-backend-cpu":Mn,"tfjs-backend-webgl":Cn,"tfjs-backend-wasm":In}});var Ra={};Lr(Ra,{AgeGenderNet:()=>He,BoundingBox:()=>Vt,Box:()=>D,ComposableTask:()=>J,ComputeAllFaceDescriptorsTask:()=>Pt,ComputeFaceDescriptorsTaskBase:()=>Ue,ComputeSingleFaceDescriptorTask:()=>Ft,DetectAllFaceLandmarksTask:()=>qe,DetectAllFacesTask:()=>Ie,DetectFaceLandmarksTaskBase:()=>Je,DetectFacesTaskBase:()=>Ke,DetectSingleFaceLandmarksTask:()=>Ze,DetectSingleFaceTask:()=>Qe,Dimensions:()=>k,FACE_EXPRESSION_LABELS:()=>so,FaceDetection:()=>M,FaceDetectionNet:()=>lo,FaceExpressionNet:()=>Oe,FaceExpressions:()=>yt,FaceLandmark68Net:()=>Kt,FaceLandmark68TinyNet:()=>ze,FaceLandmarkNet:()=>po,FaceLandmarks:()=>z,FaceLandmarks5:()=>jr,FaceLandmarks68:()=>Gt,FaceMatch:()=>pe,FaceMatcher:()=>tr,FaceRecognitionNet:()=>Qt,Gender:()=>Tr,LabeledBox:()=>ue,LabeledFaceDescriptors:()=>mt,NetInput:()=>ut,NeuralNetwork:()=>A,ObjectDetection:()=>bt,Point:()=>g,PredictedBox:()=>Ur,Rect:()=>Yt,SsdMobilenetv1:()=>St,SsdMobilenetv1Options:()=>X,TinyFaceDetector:()=>ne,TinyFaceDetectorOptions:()=>je,TinyYolov2:()=>re,TinyYolov2Options:()=>st,allFaces:()=>Aa,allFacesSsdMobilenetv1:()=>fn,allFacesTinyYolov2:()=>La,awaitMediaLoaded:()=>to,bufferToImage:()=>eo,computeFaceDescriptor:()=>va,createCanvas:()=>Jt,createCanvasFromMedia:()=>We,createFaceDetectionNet:()=>fa,createFaceRecognitionNet:()=>ea,createSsdMobilenetv1:()=>qo,createTinyFaceDetector:()=>Wa,createTinyYolov2:()=>ha,detectAllFaces:()=>Sr,detectFaceLandmarks:()=>pn,detectFaceLandmarksTiny:()=>xa,detectLandmarks:()=>Na,detectSingleFace:()=>Sa,draw:()=>co,env:()=>w,euclideanDistance:()=>vo,extendWithAge:()=>Er,extendWithFaceDescriptor:()=>Dr,extendWithFaceDetection:()=>jt,extendWithFaceExpressions:()=>xr,extendWithFaceLandmarks:()=>Pe,extendWithGender:()=>Mr,extractFaceTensors:()=>de,extractFaces:()=>le,fetchImage:()=>Hn,fetchJson:()=>no,fetchNetWeights:()=>zn,fetchOrThrow:()=>xt,fetchVideo:()=>Vn,getContext2dOrThrow:()=>O,getMediaDimensions:()=>Xt,imageTensorToCanvas:()=>ro,imageToSquare:()=>oo,inverseSigmoid:()=>An,iou:()=>zr,isMediaElement:()=>ir,isMediaLoaded:()=>Ae,isWithAge:()=>ra,isWithFaceDetection:()=>pt,isWithFaceExpressions:()=>io,isWithFaceLandmarks:()=>Zt,isWithGender:()=>oa,loadAgeGenderModel:()=>Ma,loadFaceDetectionModel:()=>Ca,loadFaceExpressionModel:()=>Ea,loadFaceLandmarkModel:()=>Pa,loadFaceLandmarkTinyModel:()=>Fa,loadFaceRecognitionModel:()=>Da,loadSsdMobilenetv1Model:()=>un,loadTinyFaceDetectorModel:()=>Ta,loadTinyYolov2Model:()=>wa,loadWeightMap:()=>ao,locateFaces:()=>Ia,matchDimensions:()=>Yn,minBbox:()=>Vr,nets:()=>P,nonMaxSuppression:()=>Yr,normalize:()=>rt,padToSquare:()=>Gr,predictAgeAndGender:()=>_a,recognizeFaceExpressions:()=>ya,resizeResults:()=>ln,resolveInput:()=>Ut,shuffleArray:()=>Ln,sigmoid:()=>Ne,ssdMobilenetv1:()=>mn,tf:()=>ka,tinyFaceDetector:()=>ba,tinyYolov2:()=>ga,toNetInput:()=>C,utils:()=>Hr,validateConfig:()=>ho,version:()=>Ba});module.exports=yn(Ra);var ka=v(x());var co={};Lr(co,{AnchorPosition:()=>Qr,DrawBox:()=>Le,DrawBoxOptions:()=>ar,DrawFaceLandmarks:()=>yr,DrawFaceLandmarksOptions:()=>vr,DrawTextField:()=>gt,DrawTextFieldOptions:()=>fe,drawContour:()=>dt,drawDetections:()=>On,drawFaceExpressions:()=>Gn,drawFaceLandmarks:()=>Un});function dt(o,t,e=!1){if(o.beginPath(),t.slice(1).forEach(({x:r,y:n},a)=>{let s=t[a];o.moveTo(s.x,s.y),o.lineTo(r,n)}),e){let r=t[t.length-1],n=t[0];if(!r||!n)return;o.moveTo(r.x,r.y),o.lineTo(n.x,n.y)}o.stroke()}var Hr={};Lr(Hr,{computeReshapedDimensions:()=>Or,getCenterPoint:()=>zt,isDimensions:()=>or,isEven:()=>rr,isFloat:()=>$r,isTensor:()=>Ot,isTensor1D:()=>Sn,isTensor2D:()=>Rr,isTensor3D:()=>ht,isTensor4D:()=>U,isValidNumber:()=>et,isValidProbablitiy:()=>me,range:()=>it,round:()=>Ht});var Po=v(x());var k=class{constructor(t,e){if(!et(t)||!et(e))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:e})}`);this._width=t,this._height=e}get width(){return this._width}get height(){return this._height}reverse(){return new k(1/this.width,1/this.height)}};function Ot(o,t){return o instanceof Po.Tensor&&o.shape.length===t}function Sn(o){return Ot(o,1)}function Rr(o){return Ot(o,2)}function ht(o){return Ot(o,3)}function U(o){return Ot(o,4)}function $r(o){return o%1!==0}function rr(o){return o%2===0}function Ht(o,t=2){let e=10**t;return Math.floor(o*e)/e}function or(o){return o&&o.width&&o.height}function Or({width:o,height:t},e){let r=e/Math.max(t,o);return new k(Math.round(o*r),Math.round(t*r))}function zt(o){return o.reduce((t,e)=>t.add(e),new g(0,0)).div(new g(o.length,o.length))}function it(o,t,e){return Array(o).fill(0).map((r,n)=>t+n*e)}function et(o){return!!o&&o!==1/0&&o!==-1/0&&!Number.isNaN(o)||o===0}function me(o){return et(o)&&o>=0&&o<=1}var g=class{constructor(t,e){this._x=t,this._y=e}get x(){return this._x}get y(){return this._y}add(t){return new g(this.x+t.x,this.y+t.y)}sub(t){return new g(this.x-t.x,this.y-t.y)}mul(t){return new g(this.x*t.x,this.y*t.y)}div(t){return new g(this.x/t.x,this.y/t.y)}abs(){return new g(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new g(Math.floor(this.x),Math.floor(this.y))}};var D=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(et)}static assertIsValidBox(t,e,r=!1){if(!D.isRect(t))throw new Error(`${e} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${e} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,e=!0){let r=t||{},n=[r.left,r.top,r.right,r.bottom].every(et),a=[r.x,r.y,r.width,r.height].every(et);if(!a&&!n)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[s,i,c,m]=a?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];D.assertIsValidBox({x:s,y:i,width:c,height:m},"Box.constructor",e),this._x=s,this._y=i,this._width=c,this._height=m}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new g(this.left,this.top)}get topRight(){return new g(this.right,this.top)}get bottomLeft(){return new g(this.left,this.bottom)}get bottomRight(){return new g(this.right,this.bottom)}round(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.round(a));return new D({x:t,y:e,width:r,height:n})}floor(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.floor(a));return new D({x:t,y:e,width:r,height:n})}toSquare(){let{x:t,y:e,width:r,height:n}=this,a=Math.abs(r-n);return r<n&&(t-=a/2,r+=a),n<r&&(e-=a/2,n+=a),new D({x:t,y:e,width:r,height:n})}rescale(t){let e=or(t)?t.width:t,r=or(t)?t.height:t;return new D({x:this.x*e,y:this.y*r,width:this.width*e,height:this.height*r})}pad(t,e){let[r,n,a,s]=[this.x-t/2,this.y-e/2,this.width+t,this.height+e];return new D({x:r,y:n,width:a,height:s})}clipAtImageBorders(t,e){let{x:r,y:n,right:a,bottom:s}=this,i=Math.max(r,0),c=Math.max(n,0),m=a-i,p=s-c,u=Math.min(m,t-i),f=Math.min(p,e-c);return new D({x:i,y:c,width:u,height:f}).floor()}shift(t,e){let{width:r,height:n}=this,a=this.x+t,s=this.y+e;return new D({x:a,y:s,width:r,height:n})}padAtBorders(t,e){let r=this.width+1,n=this.height+1,a=1,s=1,i=r,c=n,m=this.left,p=this.top,u=this.right,f=this.bottom;return u>e&&(i=-u+e+r,u=e),f>t&&(c=-f+t+n,f=t),m<1&&(c=2-m,m=1),p<1&&(c=2-p,p=1),{dy:s,edy:c,dx:a,edx:i,y:p,ey:f,x:m,ex:u,w:r,h:n}}calibrate(t){return new D({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var Vt=class extends D{constructor(t,e,r,n,a=!1){super({left:t,top:e,right:r,bottom:n},a)}};var bt=class{constructor(t,e,r,n,a){this._imageDims=new k(a.width,a.height),this._score=t,this._classScore=e,this._className=r,this._box=new D(n).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new D(this._box).rescale(this.imageDims.reverse())}forSize(t,e){return new bt(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:e})}};var M=class extends bt{constructor(t,e,r){super(t,t,"",e,r)}forSize(t,e){let{score:r,relativeBox:n,imageDims:a}=super.forSize(t,e);return new M(r,n,a)}};function zr(o,t,e=!0){let r=Math.max(0,Math.min(o.right,t.right)-Math.max(o.left,t.left)),n=Math.max(0,Math.min(o.bottom,t.bottom)-Math.max(o.top,t.top)),a=r*n;return e?a/(o.area+t.area-a):a/Math.min(o.area,t.area)}function Vr(o){let t=o.map(i=>i.x),e=o.map(i=>i.y),r=t.reduce((i,c)=>c<i?c:i,1/0),n=e.reduce((i,c)=>c<i?c:i,1/0),a=t.reduce((i,c)=>i<c?c:i,0),s=e.reduce((i,c)=>i<c?c:i,0);return new Vt(r,n,a,s)}function Yr(o,t,e,r=!0){let n=t.map((s,i)=>({score:s,boxIndex:i})).sort((s,i)=>s.score-i.score).map(s=>s.boxIndex),a=[];for(;n.length>0;){let s=n.pop();a.push(s);let i=n,c=[];for(let m=0;m<i.length;m++){let p=i[m],u=o[s],f=o[p];c.push(zr(u,f,r))}n=n.filter((m,p)=>c[p]<=e)}return a}var ct=v(x());function rt(o,t){return ct.tidy(()=>{let[e,r,n]=t,a=ct.fill([...o.shape.slice(0,3),1],e,"float32"),s=ct.fill([...o.shape.slice(0,3),1],r,"float32"),i=ct.fill([...o.shape.slice(0,3),1],n,"float32"),c=ct.concat([a,s,i],3);return ct.sub(o,c)})}var Ct=v(x());function Gr(o,t=!1){return Ct.tidy(()=>{let[e,r]=o.shape.slice(1);if(e===r)return o;let n=Math.abs(e-r),a=Math.round(n*(t?.5:1)),s=e>r?2:1,i=f=>{let l=o.shape.slice();return l[s]=f,Ct.fill(l,0,"float32")},c=i(a),m=n-c.shape[s],u=[t&&m?i(m):null,o,c].filter(f=>!!f).map(f=>Ct.cast(f,"float32"));return Ct.concat(u,s)})}function Ln(o){let t=o.slice();for(let e=t.length-1;e>0;e--){let r=Math.floor(Math.random()*(e+1)),n=t[e];t[e]=t[r],t[r]=n}return t}function Ne(o){return 1/(1+Math.exp(-o))}function An(o){return Math.log(o/(1-o))}var Yt=class extends D{constructor(t,e,r,n,a=!1){super({x:t,y:e,width:r,height:n},a)}};var Wn=.5,kn=.43,Bn=.45,z=class{constructor(t,e,r=new g(0,0)){let{width:n,height:a}=e;this._imgDims=new k(n,a),this._shift=r,this._positions=t.map(s=>s.mul(new g(n,a)).add(r))}get shift(){return new g(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new g(this.imageWidth,this.imageHeight)))}forSize(t,e){return new this.constructor(this.relativePositions,{width:t,height:e})}shiftBy(t,e){return new this.constructor(this.relativePositions,this._imgDims,new g(t,e))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,e={}){if(t){let a=t instanceof M?t.box.floor():new D(t);return this.shiftBy(a.x,a.y).align(null,e)}let{useDlibAlignment:r,minBoxPadding:n}={useDlibAlignment:!1,minBoxPadding:.2,...e};return r?this.alignDlib():this.alignMinBbox(n)}alignDlib(){let t=this.getRefPointsForAlignment(),[e,r,n]=t,a=u=>n.sub(u).magnitude(),s=(a(e)+a(r))/2,i=Math.floor(s/Bn),c=zt(t),m=Math.floor(Math.max(0,c.x-Wn*i)),p=Math.floor(Math.max(0,c.y-kn*i));return new Yt(m,p,Math.min(i,this.imageWidth+m),Math.min(i,this.imageHeight+p))}alignMinBbox(t){let e=Vr(this.positions);return e.pad(e.width*t,e.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var jr=class extends z{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],zt([t[3],t[4]])]}};var Gt=class extends z{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(zt)}};var pe=class{constructor(t,e){this._label=t,this._distance=e}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${Ht(this.distance)})`:""}`}};var ue=class extends D{constructor(e,r){super(e);this._label=r}static assertIsValidLabeledBox(e,r){if(D.assertIsValidBox(e,r),!et(e.label))throw new Error(`${r} - expected property label (${e.label}) to be a number`)}get label(){return this._label}};var mt=class{constructor(t,e){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(e)||e.some(r=>!(r instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=e}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let e=t.descriptors.map(r=>new Float32Array(r));return new mt(t.label,e)}};var Ur=class extends ue{constructor(e,r,n,a){super(e,r);this._score=n,this._classScore=a}static assertIsValidPredictedBox(e,r){if(ue.assertIsValidLabeledBox(e,r),!me(e.score)||!me(e.classScore))throw new Error(`${r} - expected properties score (${e.score}) and (${e.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function pt(o){return o.detection instanceof M}function jt(o,t){return{...o,...{detection:t}}}function Xr(){let o=window.fetch;if(!o)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:o,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function Se(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function nr(o){let t="";if(!o&&Se())try{o=require("fs")}catch(r){t=r.toString()}return{readFile:o?r=>new Promise((n,a)=>{o.readFile(r,(s,i)=>s?a(s):n(i))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function Jr(){let o=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,e=global.Video||global.HTMLVideoElement,r=()=>{if(o)return new o;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},n=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},a=()=>{if(e)return new e;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},s=global.fetch,i=nr();return{Canvas:o||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:r,createImageElement:n,createVideoElement:a,fetch:s,...i}}function qr(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var L;function Rn(){if(!L)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return L}function Zr(o){L=o}function Kr(){return qr()?Zr(Xr()):Se()?Zr(Jr()):null}function $n(o){if(L||Kr(),!L)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=L.Canvas,Image:e=L.Image}=o;L.Canvas=t,L.Image=e,L.createCanvasElement=o.createCanvasElement||(()=>new t),L.createImageElement=o.createImageElement||(()=>new e),L.ImageData=o.ImageData||L.ImageData,L.Video=o.Video||L.Video,L.fetch=o.fetch||L.fetch,L.readFile=o.readFile||L.readFile}var w={getEnv:Rn,setEnv:Zr,initialize:Kr,createBrowserEnv:Xr,createFileSystem:nr,createNodejsEnv:Jr,monkeyPatch:$n,isBrowser:qr,isNodejs:Se};Kr();function Ut(o){return!w.isNodejs()&&typeof o=="string"?document.getElementById(o):o}function O(o){let{Canvas:t,CanvasRenderingContext2D:e}=w.getEnv();if(o instanceof e)return o;let r=Ut(o);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let n=r.getContext("2d");if(!n)throw new Error("resolveContext2d - canvas 2d context is null");return n}var Qr=(n=>(n.TOP_LEFT="TOP_LEFT",n.TOP_RIGHT="TOP_RIGHT",n.BOTTOM_LEFT="BOTTOM_LEFT",n.BOTTOM_RIGHT="BOTTOM_RIGHT",n))(Qr||{}),fe=class{constructor(t={}){let{anchorPosition:e,backgroundColor:r,fontColor:n,fontSize:a,fontStyle:s,padding:i}=t;this.anchorPosition=e||"TOP_LEFT",this.backgroundColor=r||"rgba(0, 0, 0, 0.5)",this.fontColor=n||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=s||"Georgia",this.padding=i||4}},gt=class{constructor(t,e,r={}){this.text=typeof t=="string"?[t]:t instanceof gt?t.text:t,this.anchor=e,this.options=new fe(r)}measureWidth(t){let{padding:e}=this.options;return this.text.map(r=>t.measureText(r).width).reduce((r,n)=>r<n?n:r,0)+2*e}measureHeight(){let{fontSize:t,padding:e}=this.options;return this.text.length*t+2*e}getUpperLeft(t,e){let{anchorPosition:r}=this.options,n=r==="BOTTOM_RIGHT"||r==="TOP_RIGHT",a=r==="BOTTOM_LEFT"||r==="BOTTOM_RIGHT",s=this.measureWidth(t),i=this.measureHeight(),c=n?this.anchor.x-s:this.anchor.x,m=a?this.anchor.y-i:this.anchor.y;if(e){let{width:p,height:u}=e,f=Math.max(Math.min(c,p-s),0),l=Math.max(Math.min(m,u-i),0);return{x:f,y:l}}return{x:c,y:m}}draw(t){let e=Ut(t),r=O(e),{backgroundColor:n,fontColor:a,fontSize:s,fontStyle:i,padding:c}=this.options;r.font=`${s}px ${i}`;let m=this.measureWidth(r),p=this.measureHeight();r.fillStyle=n;let u=this.getUpperLeft(r,e);r.fillRect(u.x,u.y,m,p),r.fillStyle=a,this.text.forEach((f,l)=>{let b=c+u.x,y=c+u.y+(l+1)*s;r.fillText(f,b,y)})}};var ar=class{constructor(t={}){let{boxColor:e,lineWidth:r,label:n,drawLabelOptions:a}=t;this.boxColor=e||"rgba(0, 0, 255, 1)",this.lineWidth=r||2,this.label=n;let s={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new fe({...s,...a})}},Le=class{constructor(t,e={}){this.box=new D(t),this.options=new ar(e)}draw(t){let e=O(t),{boxColor:r,lineWidth:n}=this.options,{x:a,y:s,width:i,height:c}=this.box;e.strokeStyle=r,e.lineWidth=n,e.strokeRect(a,s,i,c);let{label:m}=this.options;m&&new gt([m],{x:a-n/2,y:s},this.options.drawLabelOptions).draw(t)}};function On(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof M?r.score:pt(r)?r.detection.score:void 0,a=r instanceof M?r.box:pt(r)?r.detection.box:new D(r),s=n?`${Ht(n)}`:void 0;new Le(a,{label:s}).draw(o)})}var we=v(x());function Ae(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t&&o.complete||o instanceof e&&o.readyState>=3}function to(o){return new Promise((t,e)=>{(o instanceof w.getEnv().Canvas||Ae(o))&&t(null);function r(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),e(a))}function n(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),t(a))}o.addEventListener("load",n),o.addEventListener("error",r)})}function eo(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToImage - expected buf to be of type: Blob"));let r=new FileReader;r.onload=()=>{typeof r.result!="string"&&e(new Error("bufferToImage - expected reader.result to be a string, in onload"));let n=w.getEnv().createImageElement();n.onload=()=>t(n),n.onerror=e,n.src=r.result},r.onerror=e,r.readAsDataURL(o)})}function Xt(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t?new k(o.naturalWidth,o.naturalHeight):o instanceof e?new k(o.videoWidth,o.videoHeight):new k(o.width,o.height)}function Jt({width:o,height:t}){let{createCanvasElement:e}=w.getEnv(),r=e();return r.width=o,r.height=t,r}function We(o,t){let{ImageData:e}=w.getEnv();if(!(o instanceof e)&&!Ae(o))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:n}=t||Xt(o),a=Jt({width:r,height:n});return o instanceof e?O(a).putImageData(o,0,0):O(a).drawImage(o,0,0,r,n),a}var sr=v(x());async function ro(o,t){let e=t||w.getEnv().createCanvasElement(),[r,n,a]=o.shape.slice(U(o)?1:0),s=sr.tidy(()=>o.as3D(r,n,a).toInt());return await sr.browser.toPixels(s,e),s.dispose(),e}function ir(o){let{Image:t,Canvas:e,Video:r}=w.getEnv();return o instanceof t||o instanceof e||o instanceof r}var V=v(x());function oo(o,t,e=!1){let{Image:r,Canvas:n}=w.getEnv();if(!(o instanceof r||o instanceof n))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Jt({width:1,height:1});let a=Xt(o),s=t/Math.max(a.height,a.width),i=s*a.width,c=s*a.height,m=Jt({width:t,height:t}),p=o instanceof n?o:We(o),u=Math.abs(i-c)/2,f=e&&i<c?u:0,l=e&&c<i?u:0;return p.width>0&&p.height>0&&O(m).drawImage(p,f,l,i,c),m}var ut=class{constructor(t,e=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=e,this._batchSize=t.length,t.forEach((r,n)=>{if(ht(r)){this._imageTensors[n]=r,this._inputDimensions[n]=r.shape;return}if(U(r)){let s=r.shape[0];if(s!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${s} passed, but not supported in input array`);this._imageTensors[n]=r,this._inputDimensions[n]=r.shape.slice(1);return}let a=r instanceof w.getEnv().Canvas?r:We(r);this._canvases[n]=a,this._inputDimensions[n]=[a.height,a.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return it(this.batchSize,0,1).map((t,e)=>this.getReshapedInputDimensions(e))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let e=this.getInputWidth(t),r=this.getInputHeight(t);return Or({width:e,height:r},this.inputSize)}toBatchTensor(t,e=!0){return this._inputSize=t,V.tidy(()=>{let r=it(this.batchSize,0,1).map(a=>{let s=this.getInput(a);if(s instanceof V.Tensor){let i=U(s)?s:V.expandDims(s);return i=Gr(i,e),(i.shape[1]!==t||i.shape[2]!==t)&&(i=V.image.resizeBilinear(i,[t,t],!1,!1)),i.as3D(t,t,3)}if(s instanceof w.getEnv().Canvas)return V.browser.fromPixels(oo(s,t,e));throw new Error(`toBatchTensor - at batchIdx ${a}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${s}`)});return V.stack(r.map(a=>V.cast(a,"float32"))).as4D(this.batchSize,t,t,3)})}};async function C(o){if(o instanceof ut)return o;let t=Array.isArray(o)?o:[o];if(!t.length)throw new Error("toNetInput - empty array passed as input");let e=n=>Array.isArray(o)?` at input index ${n}:`:"",r=t.map(Ut);return r.forEach((n,a)=>{if(!ir(n)&&!ht(n)&&!U(n))throw typeof t[a]=="string"?new Error(`toNetInput -${e(a)} string passed, but could not resolve HTMLElement for element id ${t[a]}`):new Error(`toNetInput -${e(a)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(U(n)){let s=n.shape[0];if(s!==1)throw new Error(`toNetInput -${e(a)} tf.Tensor4D with batchSize ${s} passed, but not supported in input array`)}}),await Promise.all(r.map(n=>ir(n)&&to(n))),new ut(r,Array.isArray(o))}async function le(o,t){let{Canvas:e}=w.getEnv(),r=o;if(!(o instanceof e)){let s=await C(o);if(s.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let i=s.getInput(0);r=i instanceof e?i:await ro(i)}let n=O(r);return t.map(s=>s instanceof M?s.forSize(r.width,r.height).box.floor():s).map(s=>s.clipAtImageBorders(r.width,r.height)).map(({x:s,y:i,width:c,height:m})=>{let p=Jt({width:c,height:m});return c>0&&m>0&&O(p).putImageData(n.getImageData(s,i,c,m),0,0),p})}var cr=v(x());async function de(o,t){if(!ht(o)&&!U(o))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(U(o)&&o.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return cr.tidy(()=>{let[e,r,n]=o.shape.slice(U(o)?1:0);return t.map(i=>i instanceof M?i.forSize(r,e).box:i).map(i=>i.clipAtImageBorders(r,e)).filter(i=>i.width>0&&i.height>0).map(({x:i,y:c,width:m,height:p})=>cr.slice3d(o.as3D(e,r,n),[c,i,0],[p,m,n]))})}async function xt(o,t){let{fetch:e}=w.getEnv(),r=await e(o,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}async function Hn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${e.type}, for url: ${t.url}`);return eo(e)}async function no(o){return(await xt(o)).json()}async function zn(o){return new Float32Array(await(await xt(o)).arrayBuffer())}function Fo(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToVideo - expected buf to be of type: Blob"));let r=w.getEnv().createVideoElement();r.oncanplay=()=>t(r),r.onerror=e,r.playsInline=!0,r.muted=!0,r.src=URL.createObjectURL(o),r.play()})}async function Vn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${e.type}, for url: ${t.url}`);return Fo(e)}var Do=v(x());function mr(o,t){let e=`${t}-weights_manifest.json`;if(!o)return{modelBaseUri:"",manifestUri:e};if(o==="/")return{modelBaseUri:"/",manifestUri:`/${e}`};let r=o.startsWith("http://")?"http://":o.startsWith("https://")?"https://":"";o=o.replace(r,"");let n=o.split("/").filter(i=>i),a=o.endsWith(".json")?n[n.length-1]:e,s=r+(o.endsWith(".json")?n.slice(0,n.length-1):n).join("/");return s=o.startsWith("/")?`/${s}`:s,{modelBaseUri:s,manifestUri:s==="/"?`/${a}`:`${s}/${a}`}}async function ao(o,t){let{manifestUri:e,modelBaseUri:r}=mr(o,t),n=await no(e);return Do.io.loadWeights(n,r)}function Yn(o,t,e=!1){let{width:r,height:n}=e?Xt(t):t;return o.width=r,o.height=n,{width:r,height:n}}var ye=v(x());var vt=v(x());var A=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:e,objProp:r}=this.traversePropertyPath(t);return e[r]}reassignParamFromPath(t,e){let{obj:r,objProp:n}=this.traversePropertyPath(t);r[n].dispose(),r[n]=e}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof vt.Variable)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof vt.Variable))}variable(){this.getFrozenParams().forEach(({path:t,tensor:e})=>{this.reassignParamFromPath(t,e.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:e})=>{let r=vt.tensor(e.dataSync());e.dispose(),this.reassignParamFromPath(t,r)})}dispose(t=!0){this.getParamList().forEach(e=>{if(t&&e.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${e.path}`);e.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,e)=>t.concat(e)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let e=await ao(t,this.getDefaultModelName());this.loadFromWeightMap(e)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:e}=w.getEnv(),{manifestUri:r,modelBaseUri:n}=mr(t,this.getDefaultModelName()),a=m=>Promise.all(m.map(p=>e(p).then(u=>u.buffer))),s=vt.io.weightsLoaderFactory(a),i=JSON.parse((await e(r)).toString()),c=await s(i,n);this.loadFromWeightMap(c)}loadFromWeightMap(t){let{paramMappings:e,params:r}=this.extractParamsFromWeightMap(t);this._paramMappings=e,this._params=r}extractWeights(t){let{paramMappings:e,params:r}=this.extractParams(t);this._paramMappings=e,this._params=r}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let e=t.split("/").reduce((a,s)=>{if(!a.nextObj.hasOwnProperty(s))throw new Error(`traversePropertyPath - object does not have property ${s}, for path ${t}`);return{obj:a.nextObj,objProp:s,nextObj:a.nextObj[s]}},{nextObj:this.params}),{obj:r,objProp:n}=e;if(!r||!n||!(r[n]instanceof vt.Tensor))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:r,objProp:n}}};var I=v(x());var he=v(x());function H(o,t,e){return he.tidy(()=>{let r=he.separableConv2d(o,t.depthwise_filter,t.pointwise_filter,e,"same");return r=he.add(r,t.bias),r})}function pr(o,t,e=!1){return I.tidy(()=>{let r=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,[2,2],"same"),t.conv0.bias):H(o,t.conv0,[2,2])),n=H(r,t.conv1,[1,1]),a=I.relu(I.add(r,n)),s=H(a,t.conv2,[1,1]);return I.relu(I.add(r,I.add(n,s)))})}function ke(o,t,e=!1,r=!0){return I.tidy(()=>{let n=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):H(o,t.conv0,r?[2,2]:[1,1])),a=H(n,t.conv1,[1,1]),s=I.relu(I.add(n,a)),i=H(s,t.conv2,[1,1]),c=I.relu(I.add(n,I.add(a,i))),m=H(c,t.conv3,[1,1]);return I.relu(I.add(n,I.add(a,I.add(i,m))))})}var It=v(x());function qt(o,t,e="same",r=!1){return It.tidy(()=>{let n=It.add(It.conv2d(o,t.filters,[1,1],e),t.bias);return r?It.relu(n):n})}function B(o,t){Object.keys(o).forEach(e=>{t.some(r=>r.originalPath===e)||o[e].dispose()})}var ur=v(x());function be(o,t){return(e,r,n,a)=>{let s=ur.tensor4d(o(e*r*n*n),[n,n,e,r]),i=ur.tensor1d(o(r));return t.push({paramPath:`${a}/filters`},{paramPath:`${a}/bias`}),{filters:s,bias:i}}}var fr=v(x());function lr(o,t){return(e,r,n)=>{let a=fr.tensor2d(o(e*r),[e,r]),s=fr.tensor1d(o(r));return t.push({paramPath:`${n}/weights`},{paramPath:`${n}/bias`}),{weights:a,bias:s}}}var Re=v(x());var Be=class{constructor(t,e,r){this.depthwise_filter=t;this.pointwise_filter=e;this.bias=r}};function ge(o,t){return(e,r,n)=>{let a=Re.tensor4d(o(9*e),[3,3,e,1]),s=Re.tensor4d(o(e*r),[1,1,e,r]),i=Re.tensor1d(o(r));return t.push({paramPath:`${n}/depthwise_filter`},{paramPath:`${n}/pointwise_filter`},{paramPath:`${n}/bias`}),new Be(a,s,i)}}function xe(o){return t=>{let e=o(`${t}/depthwise_filter`,4),r=o(`${t}/pointwise_filter`,4),n=o(`${t}/bias`,1);return new Be(e,r,n)}}function Y(o,t){return(e,r,n)=>{let a=o[e];if(!Ot(a,r))throw new Error(`expected weightMap[${e}] to be a Tensor${r}D, instead have ${a}`);return t.push({originalPath:e,paramPath:n||e}),a}}function R(o){let t=o;function e(n){let a=t.slice(0,n);return t=t.slice(n),a}function r(){return t}return{extractWeights:e,getRemainingWeights:r}}function dr(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c,m=!1){let p=m?e(s,i,3,`${c}/conv0`):r(s,i,`${c}/conv0`),u=r(i,i,`${c}/conv1`),f=r(i,i,`${c}/conv2`);return{conv0:p,conv1:u,conv2:f}}function a(s,i,c,m=!1){let{conv0:p,conv1:u,conv2:f}=n(s,i,c,m),l=r(i,i,`${c}/conv3`);return{conv0:p,conv1:u,conv2:f,conv3:l}}return{extractDenseBlock3Params:n,extractDenseBlock4Params:a}}function Eo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock4Params:n}=dr(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2"),c=n(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i,dense3:c}}}function hr(o){return t=>{let e=o(`${t}/filters`,4),r=o(`${t}/bias`,1);return{filters:e,bias:r}}}function br(o,t){let e=Y(o,t),r=hr(e),n=xe(e);function a(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`);return{conv0:m,conv1:p,conv2:u}}function s(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`),f=n(`${i}/conv3`);return{conv0:m,conv1:p,conv2:u,conv3:f}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:s}}function Mo(o){let t=[],{extractDenseBlock4Params:e}=br(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2"),dense3:e("dense3")};return B(o,t),{params:r,paramMappings:t}}var ve=class extends A{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceFeatureExtractor - load model before inference");return ye.tidy(()=>{let r=ye.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=ke(a,e.dense0,!0);return s=ke(s,e.dense1),s=ke(s,e.dense2),s=ke(s,e.dense3),s=ye.avgPool(s,[7,7],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return Mo(t)}extractParams(t){return Eo(t)}};var No=v(x());var _e=v(x());function $e(o,t){return _e.tidy(()=>_e.add(_e.matMul(o,t.weights),t.bias))}function Co(o,t,e){let r=[],{extractWeights:n,getRemainingWeights:a}=R(o),i=lr(n,r)(t,e,"fc");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:r,params:{fc:i}}}function Io(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:r("fc")};return B(o,t),{params:n,paramMappings:t}}function gr(o){let t={},e={};return Object.keys(o).forEach(r=>{let n=r.startsWith("fc")?e:t;n[r]=o[r]}),{featureExtractorMap:t,classifierMap:e}}var Te=class extends A{constructor(e,r){super(e);this._faceFeatureExtractor=r}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return No.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e;return $e(n.as2D(n.shape[0],-1),r.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return Co(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=gr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Io(n)}extractParams(e){let r=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),a=n*r+n,s=e.slice(0,e.length-a),i=e.slice(e.length-a);return this.faceFeatureExtractor.extractWeights(s),this.extractClassifierParams(i)}};var so=["neutral","happy","sad","angry","fearful","disgusted","surprised"],yt=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);so.forEach((e,r)=>{this[e]=t[r]})}asSortedArray(){return so.map(t=>({expression:t,probability:this[t]})).sort((t,e)=>e.probability-t.probability)}};var Oe=class extends Te{constructor(t=new ve){super("FaceExpressionNet",t)}forwardInput(t){return we.tidy(()=>we.softmax(this.runNet(t)))}async forward(t){return this.forwardInput(await C(t))}async predictExpressions(t){let e=await C(t),r=await this.forwardInput(e),n=await Promise.all(we.unstack(r).map(async s=>{let i=s.dataSync();return s.dispose(),i}));r.dispose();let a=n.map(s=>new yt(s));return e.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function io(o){return o.expressions instanceof yt}function xr(o,t){return{...o,...{expressions:t}}}function Gn(o,t,e=.1,r){(Array.isArray(t)?t:[t]).forEach(a=>{let s=a instanceof yt?a:io(a)?a.expressions:void 0;if(!s)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let c=s.asSortedArray().filter(u=>u.probability>e),m=pt(a)?a.detection.box.bottomLeft:r||new g(0,0);new gt(c.map(u=>`${u.expression} (${Ht(u.probability)})`),m).draw(o)})}function Zt(o){return pt(o)&&o.landmarks instanceof z&&o.unshiftedLandmarks instanceof z&&o.alignedRect instanceof M}function jn(o){let t=c=>c*180/Math.PI,e=(c,m)=>Math.sqrt((c._x-m._x)**2+(c._y-m._y)**2),r={roll:void 0,pitch:void 0,yaw:void 0},n=(c,m,p)=>{let u=Math.floor(c._x-m._x),f=Math.floor(m._x-p._x);return u-f},a=(c,m)=>{let p=Math.hypot(m._x-c._x,m._y-c._y),u=m._y-c._y,f=Math.asin(u/p),l=t(f),b=Math.floor(90-l),y=m._x-c._x<0?-1:1;return b*y},s=(c,m,p)=>{let u=e(c,p),f={_x:(c._x+p._x)/2,_y:(c._y+p._y)/2},l=e(m,f),b=Math.atan(l/u),y=Math.floor(t(b)),F=f._y-m._y<0?-1:1;return y*F};if(!o||!o._positions||o._positions.length!==68)return r;let i=o._positions;return r.roll=a(i[27],i[66]),r.pitch=s(i[14],i[30],i[2]),r.yaw=n(i[14],i[33],i[2]),r}function Pe(o,t){let{box:e}=o.detection,r=t.shiftBy(e.x,e.y),n=r.align(),{imageDims:a}=o.detection,s=new M(o.detection.score,n.rescale(a.reverse()),a),i=jn(t);return{...o,...{landmarks:r,unshiftedLandmarks:t,alignedRect:s,angle:i}}}var vr=class{constructor(t={}){let{drawLines:e=!0,drawPoints:r=!0,lineWidth:n,lineColor:a,pointSize:s,pointColor:i}=t;this.drawLines=e,this.drawPoints=r,this.lineWidth=n||1,this.pointSize=s||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},yr=class{constructor(t,e={}){this.faceLandmarks=t,this.options=new vr(e)}draw(t){let e=O(t),{drawLines:r,drawPoints:n,lineWidth:a,lineColor:s,pointSize:i,pointColor:c}=this.options;if(r&&this.faceLandmarks instanceof Gt&&(e.strokeStyle=s,e.lineWidth=a,dt(e,this.faceLandmarks.getJawOutline()),dt(e,this.faceLandmarks.getLeftEyeBrow()),dt(e,this.faceLandmarks.getRightEyeBrow()),dt(e,this.faceLandmarks.getNose()),dt(e,this.faceLandmarks.getLeftEye(),!0),dt(e,this.faceLandmarks.getRightEye(),!0),dt(e,this.faceLandmarks.getMouth(),!0)),n){e.strokeStyle=c,e.fillStyle=c;let m=p=>{e.beginPath(),e.arc(p.x,p.y,i,0,2*Math.PI),e.fill()};this.faceLandmarks.positions.forEach(m)}}};function Un(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof z?r:Zt(r)?r.landmarks:void 0;if(!n)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new yr(n).draw(o)})}var So="1.7.6";var ft=v(x());var S=v(x());function qn(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c){let m=r(s,i,`${c}/separable_conv0`),p=r(i,i,`${c}/separable_conv1`),u=e(s,i,1,`${c}/expansion_conv`);return{separable_conv0:m,separable_conv1:p,expansion_conv:u}}function a(s,i){let c=r(s,s,`${i}/separable_conv0`),m=r(s,s,`${i}/separable_conv1`),p=r(s,s,`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:e,extractSeparableConvParams:r,extractReductionBlockParams:n,extractMainBlockParams:a}}function Lo(o,t){let e=[],{extractWeights:r,getRemainingWeights:n}=R(o),{extractConvParams:a,extractSeparableConvParams:s,extractReductionBlockParams:i,extractMainBlockParams:c}=qn(r,e),m=a(3,32,3,"entry_flow/conv_in"),p=i(32,64,"entry_flow/reduction_block_0"),u=i(64,128,"entry_flow/reduction_block_1"),f={conv_in:m,reduction_block_0:p,reduction_block_1:u},l={};it(t,0,1).forEach(h=>{l[`main_block_${h}`]=c(128,`middle_flow/main_block_${h}`)});let b=i(128,256,"exit_flow/reduction_block"),y=s(256,512,"exit_flow/separable_conv"),F={reduction_block:b,separable_conv:y};if(n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{paramMappings:e,params:{entry_flow:f,middle_flow:l,exit_flow:F}}}function Zn(o,t){let e=Y(o,t),r=hr(e),n=xe(e);function a(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=r(`${i}/expansion_conv`);return{separable_conv0:c,separable_conv1:m,expansion_conv:p}}function s(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=n(`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}}function Ao(o,t){let e=[],{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}=Zn(o,e),i=r("entry_flow/conv_in"),c=a("entry_flow/reduction_block_0"),m=a("entry_flow/reduction_block_1"),p={conv_in:i,reduction_block_0:c,reduction_block_1:m},u={};it(t,0,1).forEach(y=>{u[`main_block_${y}`]=s(`middle_flow/main_block_${y}`)});let f=a("exit_flow/reduction_block"),l=n("exit_flow/separable_conv"),b={reduction_block:f,separable_conv:l};return B(o,e),{params:{entry_flow:p,middle_flow:u,exit_flow:b},paramMappings:e}}function Wo(o,t,e){return S.add(S.conv2d(o,t.filters,e,"same"),t.bias)}function mo(o,t,e=!0){let r=e?S.relu(o):o;return r=H(r,t.separable_conv0,[1,1]),r=H(S.relu(r),t.separable_conv1,[1,1]),r=S.maxPool(r,[3,3],[2,2],"same"),r=S.add(r,Wo(o,t.expansion_conv,[2,2])),r}function Kn(o,t){let e=H(S.relu(o),t.separable_conv0,[1,1]);return e=H(S.relu(e),t.separable_conv1,[1,1]),e=H(S.relu(e),t.separable_conv2,[1,1]),e=S.add(e,o),e}var _r=class extends A{constructor(e){super("TinyXception");this._numMainBlocks=e}forwardInput(e){let{params:r}=this;if(!r)throw new Error("TinyXception - load model before inference");return S.tidy(()=>{let n=S.cast(e.toBatchTensor(112,!0),"float32"),s=rt(n,[122.782,117.001,104.298]).div(255),i=S.relu(Wo(s,r.entry_flow.conv_in,[2,2]));return i=mo(i,r.entry_flow.reduction_block_0,!1),i=mo(i,r.entry_flow.reduction_block_1),it(this._numMainBlocks,0,1).forEach(c=>{i=Kn(i,r.middle_flow[`main_block_${c}`])}),i=mo(i,r.exit_flow.reduction_block),i=S.relu(H(i,r.exit_flow.separable_conv,[1,1])),i})}async forward(e){return this.forwardInput(await C(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return Ao(e,this._numMainBlocks)}extractParams(e){return Lo(e,this._numMainBlocks)}};function ko(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),n=lr(e,t),a=n(512,1,"fc/age"),s=n(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:a,gender:s}}}}function Bo(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:{age:r("fc/age"),gender:r("fc/gender")}};return B(o,t),{params:n,paramMappings:t}}var Tr=(e=>(e.FEMALE="female",e.MALE="male",e))(Tr||{});var He=class extends A{constructor(e=new _r(2)){super("AgeGenderNet");this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return ft.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e,a=ft.avgPool(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1),s=$e(a,r.fc.age).as1D(),i=$e(a,r.fc.gender);return{age:s,gender:i}})}forwardInput(e){return ft.tidy(()=>{let{age:r,gender:n}=this.runNet(e);return{age:r,gender:ft.softmax(n)}})}async forward(e){return this.forwardInput(await C(e))}async predictAgeAndGender(e){let r=await C(e),n=await this.forwardInput(r),a=ft.unstack(n.age),s=ft.unstack(n.gender),i=a.map((m,p)=>({ageTensor:m,genderTensor:s[p]})),c=await Promise.all(i.map(async({ageTensor:m,genderTensor:p})=>{let u=m.dataSync()[0],f=p.dataSync()[0],l=f>.5,b=l?"male":"female",y=l?f:1-f;return m.dispose(),p.dispose(),{age:u,gender:b,genderProbability:y}}));return n.age.dispose(),n.gender.dispose(),r.isBatchInput?c:c[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return ko(e)}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=gr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Bo(n)}extractParams(e){let n=e.slice(0,e.length-1539),a=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(n),this.extractClassifierParams(a)}};var G=v(x());var Fe=class extends Te{postProcess(t,e,r){let n=r.map(({width:s,height:i})=>{let c=e/Math.max(i,s);return{width:s*c,height:i*c}}),a=n.length;return G.tidy(()=>{let s=(u,f)=>G.stack([G.fill([68],u,"float32"),G.fill([68],f,"float32")],1).as2D(1,136).as1D(),i=(u,f)=>{let{width:l,height:b}=n[u];return f(l,b)?Math.abs(l-b)/2:0},c=u=>i(u,(f,l)=>f<l),m=u=>i(u,(f,l)=>l<f);return t.mul(G.fill([a,136],e,"float32")).sub(G.stack(Array.from(Array(a),(u,f)=>s(c(f),m(f))))).div(G.stack(Array.from(Array(a),(u,f)=>s(n[f].width,n[f].height))))})}forwardInput(t){return G.tidy(()=>{let e=this.runNet(t);return this.postProcess(e,t.inputSize,t.inputDimensions.map(([r,n])=>({height:r,width:n})))})}async forward(t){return this.forwardInput(await C(t))}async detectLandmarks(t){let e=await C(t),r=G.tidy(()=>G.unstack(this.forwardInput(e))),n=await Promise.all(r.map(async(a,s)=>{let i=Array.from(a.dataSync()),c=i.filter((p,u)=>rr(u)),m=i.filter((p,u)=>!rr(u));return new Gt(Array(68).fill(0).map((p,u)=>new g(c[u],m[u])),{height:e.getInputHeight(s),width:e.getInputWidth(s)})}));return r.forEach(a=>a.dispose()),e.isBatchInput?n:n[0]}getClassifierChannelsOut(){return 136}};var Kt=class extends Fe{constructor(t=new ve){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};var De=v(x());function Ro(o){let t=[],{extractDenseBlock3Params:e}=br(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2")};return B(o,t),{params:r,paramMappings:t}}function $o(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock3Params:n}=dr(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i}}}var wr=class extends A{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("TinyFaceFeatureExtractor - load model before inference");return De.tidy(()=>{let r=De.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=pr(a,e.dense0,!0);return s=pr(s,e.dense1),s=pr(s,e.dense2),s=De.avgPool(s,[14,14],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return Ro(t)}extractParams(t){return $o(t)}};var ze=class extends Fe{constructor(t=new wr){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var po=class extends Kt{};var nt=v(x());var Ee=v(x());var Pr=v(x());function Oo(o,t){return Pr.add(Pr.mul(o,t.weights),t.biases)}function uo(o,t,e,r,n="same"){let{filters:a,bias:s}=t.conv,i=Ee.conv2d(o,a,e,n);return i=Ee.add(i,s),i=Oo(i,t.scale),r?Ee.relu(i):i}function Ho(o,t){return uo(o,t,[1,1],!0)}function fo(o,t){return uo(o,t,[1,1],!1)}function Fr(o,t){return uo(o,t,[2,2],!0,"valid")}var j=v(x());function Qn(o,t){function e(i,c,m){let p=o(i),u=p.length/(c*m*m);if($r(u))throw new Error(`depth has to be an integer: ${u}, weights.length: ${p.length}, numFilters: ${c}, filterSize: ${m}`);return j.tidy(()=>j.transpose(j.tensor4d(p,[c,u,m,m]),[2,3,1,0]))}function r(i,c,m,p){let u=e(i,c,m),f=j.tensor1d(o(c));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/bias`}),{filters:u,bias:f}}function n(i,c){let m=j.tensor1d(o(i)),p=j.tensor1d(o(i));return t.push({paramPath:`${c}/weights`},{paramPath:`${c}/biases`}),{weights:m,biases:p}}function a(i,c,m,p){let u=r(i,c,m,`${p}/conv`),f=n(c,`${p}/scale`);return{conv:u,scale:f}}function s(i,c,m,p,u=!1){let f=a((u?.5:1)*i,c,m,`${p}/conv1`),l=a(i,c,m,`${p}/conv2`);return{conv1:f,conv2:l}}return{extractConvLayerParams:a,extractResidualLayerParams:s}}function zo(o){let{extractWeights:t,getRemainingWeights:e}=R(o),r=[],{extractConvLayerParams:n,extractResidualLayerParams:a}=Qn(t,r),s=n(4704,32,7,"conv32_down"),i=a(9216,32,3,"conv32_1"),c=a(9216,32,3,"conv32_2"),m=a(9216,32,3,"conv32_3"),p=a(36864,64,3,"conv64_down",!0),u=a(36864,64,3,"conv64_1"),f=a(36864,64,3,"conv64_2"),l=a(36864,64,3,"conv64_3"),b=a(147456,128,3,"conv128_down",!0),y=a(147456,128,3,"conv128_1"),F=a(147456,128,3,"conv128_2"),h=a(589824,256,3,"conv256_down",!0),T=a(589824,256,3,"conv256_1"),_=a(589824,256,3,"conv256_2"),E=a(589824,256,3,"conv256_down_out"),W=j.tidy(()=>j.transpose(j.tensor2d(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),e().length!==0)throw new Error(`weights remaing after extract: ${e().length}`);return{params:{conv32_down:s,conv32_1:i,conv32_2:c,conv32_3:m,conv64_down:p,conv64_1:u,conv64_2:f,conv64_3:l,conv128_down:b,conv128_1:y,conv128_2:F,conv256_down:h,conv256_1:T,conv256_2:_,conv256_down_out:E,fc:W},paramMappings:r}}function ta(o,t){let e=Y(o,t);function r(s){let i=e(`${s}/scale/weights`,1),c=e(`${s}/scale/biases`,1);return{weights:i,biases:c}}function n(s){let i=e(`${s}/conv/filters`,4),c=e(`${s}/conv/bias`,1),m=r(s);return{conv:{filters:i,bias:c},scale:m}}function a(s){return{conv1:n(`${s}/conv1`),conv2:n(`${s}/conv2`)}}return{extractConvLayerParams:n,extractResidualLayerParams:a}}function Vo(o){let t=[],{extractConvLayerParams:e,extractResidualLayerParams:r}=ta(o,t),n=e("conv32_down"),a=r("conv32_1"),s=r("conv32_2"),i=r("conv32_3"),c=r("conv64_down"),m=r("conv64_1"),p=r("conv64_2"),u=r("conv64_3"),f=r("conv128_down"),l=r("conv128_1"),b=r("conv128_2"),y=r("conv256_down"),F=r("conv256_1"),h=r("conv256_2"),T=r("conv256_down_out"),{fc:_}=o;if(t.push({originalPath:"fc",paramPath:"fc"}),!Rr(_))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${_}`);let E={conv32_down:n,conv32_1:a,conv32_2:s,conv32_3:i,conv64_down:c,conv64_1:m,conv64_2:p,conv64_3:u,conv128_down:f,conv128_1:l,conv128_2:b,conv256_down:y,conv256_1:F,conv256_2:h,conv256_down_out:T,fc:_};return B(o,t),{params:E,paramMappings:t}}var $=v(x());function ot(o,t){let e=Ho(o,t.conv1);return e=fo(e,t.conv2),e=$.add(e,o),e=$.relu(e),e}function Ve(o,t){let e=Fr(o,t.conv1);e=fo(e,t.conv2);let r=$.avgPool(o,2,2,"valid"),n=$.zeros(r.shape),a=r.shape[3]!==e.shape[3];if(r.shape[1]!==e.shape[1]||r.shape[2]!==e.shape[2]){let i=[...e.shape];i[1]=1;let c=$.zeros(i);e=$.concat([e,c],1);let m=[...e.shape];m[2]=1;let p=$.zeros(m);e=$.concat([e,p],2)}return r=a?$.concat([r,n],3):r,e=$.add(r,e),e=$.relu(e),e}var Qt=class extends A{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceRecognitionNet - load model before inference");return nt.tidy(()=>{let r=nt.cast(t.toBatchTensor(150,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=Fr(a,e.conv32_down);s=nt.maxPool(s,3,2,"valid"),s=ot(s,e.conv32_1),s=ot(s,e.conv32_2),s=ot(s,e.conv32_3),s=Ve(s,e.conv64_down),s=ot(s,e.conv64_1),s=ot(s,e.conv64_2),s=ot(s,e.conv64_3),s=Ve(s,e.conv128_down),s=ot(s,e.conv128_1),s=ot(s,e.conv128_2),s=Ve(s,e.conv256_down),s=ot(s,e.conv256_1),s=ot(s,e.conv256_2),s=Ve(s,e.conv256_down_out);let i=s.mean([1,2]);return nt.matMul(i,e.fc)})}async forward(t){return this.forwardInput(await C(t))}async computeFaceDescriptor(t){var a;if((a=t==null?void 0:t.shape)!=null&&a.some(s=>s<=0))return new Float32Array(128);let e=await C(t),r=nt.tidy(()=>nt.unstack(this.forwardInput(e))),n=await Promise.all(r.map(s=>s.data()));return r.forEach(s=>s.dispose()),e.isBatchInput?n:n[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return Vo(t)}extractParams(t){return zo(t)}};function ea(o){let t=new Qt;return t.extractWeights(o),t}function Dr(o,t){return{...o,...{descriptor:t}}}function ra(o){return typeof o.age=="number"}function Er(o,t){return{...o,...{age:t}}}function oa(o){return(o.gender==="male"||o.gender==="female")&&me(o.genderProbability)}function Mr(o,t,e){return{...o,...{gender:t,genderProbability:e}}}var Lt=v(x());var at=v(x());function na(o,t){function e(c,m){let p=at.tensor4d(o(9*c),[3,3,c,1]),u=at.tensor1d(o(c)),f=at.tensor1d(o(c)),l=at.tensor1d(o(c)),b=at.tensor1d(o(c));return t.push({paramPath:`${m}/filters`},{paramPath:`${m}/batch_norm_scale`},{paramPath:`${m}/batch_norm_offset`},{paramPath:`${m}/batch_norm_mean`},{paramPath:`${m}/batch_norm_variance`}),{filters:p,batch_norm_scale:u,batch_norm_offset:f,batch_norm_mean:l,batch_norm_variance:b}}function r(c,m,p,u,f){let l=at.tensor4d(o(c*m*p*p),[p,p,c,m]),b=at.tensor1d(o(m));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/${f?"batch_norm_offset":"bias"}`}),{filters:l,bias:b}}function n(c,m,p,u){let{filters:f,bias:l}=r(c,m,p,u,!0);return{filters:f,batch_norm_offset:l}}function a(c,m,p){let u=e(c,`${p}/depthwise_conv`),f=n(c,m,1,`${p}/pointwise_conv`);return{depthwise_conv:u,pointwise_conv:f}}function s(){let c=n(3,32,3,"mobilenetv1/conv_0"),m=a(32,64,"mobilenetv1/conv_1"),p=a(64,128,"mobilenetv1/conv_2"),u=a(128,128,"mobilenetv1/conv_3"),f=a(128,256,"mobilenetv1/conv_4"),l=a(256,256,"mobilenetv1/conv_5"),b=a(256,512,"mobilenetv1/conv_6"),y=a(512,512,"mobilenetv1/conv_7"),F=a(512,512,"mobilenetv1/conv_8"),h=a(512,512,"mobilenetv1/conv_9"),T=a(512,512,"mobilenetv1/conv_10"),_=a(512,512,"mobilenetv1/conv_11"),E=a(512,1024,"mobilenetv1/conv_12"),W=a(1024,1024,"mobilenetv1/conv_13");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,conv_8:F,conv_9:h,conv_10:T,conv_11:_,conv_12:E,conv_13:W}}function i(){let c=n(1024,256,1,"prediction_layer/conv_0"),m=n(256,512,3,"prediction_layer/conv_1"),p=n(512,128,1,"prediction_layer/conv_2"),u=n(128,256,3,"prediction_layer/conv_3"),f=n(256,128,1,"prediction_layer/conv_4"),l=n(128,256,3,"prediction_layer/conv_5"),b=n(256,64,1,"prediction_layer/conv_6"),y=n(64,128,3,"prediction_layer/conv_7"),F=r(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),h=r(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),T=r(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),_=r(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),E=r(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),W=r(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),tt=r(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),lt=r(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),q=r(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),Dt=r(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),Et=r(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),Mt=r(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,box_predictor_0:{box_encoding_predictor:F,class_predictor:h},box_predictor_1:{box_encoding_predictor:T,class_predictor:_},box_predictor_2:{box_encoding_predictor:E,class_predictor:W},box_predictor_3:{box_encoding_predictor:tt,class_predictor:lt},box_predictor_4:{box_encoding_predictor:q,class_predictor:Dt},box_predictor_5:{box_encoding_predictor:Et,class_predictor:Mt}}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:i}}function Yo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractMobilenetV1Params:n,extractPredictionLayerParams:a}=na(e,t),s=n(),i=a(),m={extra_dim:at.tensor3d(e(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:s,prediction_layer:i,output_layer:m},paramMappings:t}}function aa(o,t){let e=Y(o,t);function r(m,p,u){let f=e(`${m}/Conv2d_${p}_pointwise/weights`,4,`${u}/filters`),l=e(`${m}/Conv2d_${p}_pointwise/convolution_bn_offset`,1,`${u}/batch_norm_offset`);return{filters:f,batch_norm_offset:l}}function n(m){let p=`mobilenetv1/conv_${m}`,u=`MobilenetV1/Conv2d_${m}_depthwise`,f=`${p}/depthwise_conv`,l=`${p}/pointwise_conv`,b=e(`${u}/depthwise_weights`,4,`${f}/filters`),y=e(`${u}/BatchNorm/gamma`,1,`${f}/batch_norm_scale`),F=e(`${u}/BatchNorm/beta`,1,`${f}/batch_norm_offset`),h=e(`${u}/BatchNorm/moving_mean`,1,`${f}/batch_norm_mean`),T=e(`${u}/BatchNorm/moving_variance`,1,`${f}/batch_norm_variance`);return{depthwise_conv:{filters:b,batch_norm_scale:y,batch_norm_offset:F,batch_norm_mean:h,batch_norm_variance:T},pointwise_conv:r("MobilenetV1",m,l)}}function a(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:n(1),conv_2:n(2),conv_3:n(3),conv_4:n(4),conv_5:n(5),conv_6:n(6),conv_7:n(7),conv_8:n(8),conv_9:n(9),conv_10:n(10),conv_11:n(11),conv_12:n(12),conv_13:n(13)}}function s(m,p){let u=e(`${m}/weights`,4,`${p}/filters`),f=e(`${m}/biases`,1,`${p}/bias`);return{filters:u,bias:f}}function i(m){let p=s(`Prediction/BoxPredictor_${m}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${m}/box_encoding_predictor`),u=s(`Prediction/BoxPredictor_${m}/ClassPredictor`,`prediction_layer/box_predictor_${m}/class_predictor`);return{box_encoding_predictor:p,class_predictor:u}}function c(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}return{extractMobilenetV1Params:a,extractPredictionLayerParams:c}}function Go(o){let t=[],{extractMobilenetV1Params:e,extractPredictionLayerParams:r}=aa(o,t),n=o["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!ht(n))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${n}`);let a={mobilenetv1:e(),prediction_layer:r(),output_layer:{extra_dim:n}};return B(o,t),{params:a,paramMappings:t}}var _t=v(x());var Nt=v(x());function Z(o,t,e){return Nt.tidy(()=>{let r=Nt.conv2d(o,t.filters,e,"same");return r=Nt.add(r,t.batch_norm_offset),Nt.clipByValue(r,0,6)})}var sa=.0010000000474974513;function ia(o,t,e){return _t.tidy(()=>{let r=_t.depthwiseConv2d(o,t.filters,e,"same");return r=_t.batchNorm(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,sa),_t.clipByValue(r,0,6)})}function ca(o){return[2,4,6,12].some(t=>t===o)?[2,2]:[1,1]}function jo(o,t){return _t.tidy(()=>{let e,r=Z(o,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((a,s)=>{let i=s+1,c=ca(i);r=ia(r,a.depthwise_conv,c),r=Z(r,a.pointwise_conv,[1,1]),i===11&&(e=r)}),e===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:e}})}function ma(o,t,e){let r=o.arraySync(),n=Math.min(r[t][0],r[t][2]),a=Math.min(r[t][1],r[t][3]),s=Math.max(r[t][0],r[t][2]),i=Math.max(r[t][1],r[t][3]),c=Math.min(r[e][0],r[e][2]),m=Math.min(r[e][1],r[e][3]),p=Math.max(r[e][0],r[e][2]),u=Math.max(r[e][1],r[e][3]),f=(s-n)*(i-a),l=(p-c)*(u-m);if(f<=0||l<=0)return 0;let b=Math.max(n,c),y=Math.max(a,m),F=Math.min(s,p),h=Math.min(i,u),T=Math.max(F-b,0)*Math.max(h-y,0);return T/(f+l-T)}function Uo(o,t,e,r,n){let a=o.shape[0],s=Math.min(e,a),i=t.map((p,u)=>({score:p,boxIndex:u})).filter(p=>p.score>n).sort((p,u)=>u.score-p.score),c=p=>p<=r?1:0,m=[];return i.forEach(p=>{if(m.length>=s)return;let u=p.score;for(let f=m.length-1;f>=0;--f){let l=ma(o,p.boxIndex,m[f]);if(l!==0&&(p.score*=c(l),p.score<=n))break}u===p.score&&m.push(p.boxIndex)}),m}var d=v(x());function pa(o){let t=d.unstack(d.transpose(o,[1,0])),e=[d.sub(t[2],t[0]),d.sub(t[3],t[1])],r=[d.add(t[0],d.div(e[0],2)),d.add(t[1],d.div(e[1],2))];return{sizes:e,centers:r}}function ua(o,t){let{sizes:e,centers:r}=pa(o),n=d.unstack(d.transpose(t,[1,0])),a=d.div(d.mul(d.exp(d.div(n[2],5)),e[0]),2),s=d.add(d.mul(d.div(n[0],10),e[0]),r[0]),i=d.div(d.mul(d.exp(d.div(n[3],5)),e[1]),2),c=d.add(d.mul(d.div(n[1],10),e[1]),r[1]);return d.transpose(d.stack([d.sub(s,a),d.sub(c,i),d.add(s,a),d.add(c,i)]),[1,0])}function Xo(o,t,e){return d.tidy(()=>{let r=o.shape[0],n=ua(d.reshape(d.tile(e.extra_dim,[r,1,1]),[-1,4]),d.reshape(o,[-1,4]));n=d.reshape(n,[r,n.shape[0]/r,4]);let a=d.sigmoid(d.slice(t,[0,0,1],[-1,-1,-1])),s=d.slice(a,[0,0,0],[-1,-1,1]);s=d.reshape(s,[r,s.shape[1]]);let i=d.unstack(n),c=d.unstack(s);return{boxes:i,scores:c}})}var Ge=v(x());var Ye=v(x());function te(o,t){return Ye.tidy(()=>{let e=o.shape[0],r=Ye.reshape(qt(o,t.box_encoding_predictor),[e,-1,1,4]),n=Ye.reshape(qt(o,t.class_predictor),[e,-1,3]);return{boxPredictionEncoding:r,classPrediction:n}})}function Jo(o,t,e){return Ge.tidy(()=>{let r=Z(o,e.conv_0,[1,1]),n=Z(r,e.conv_1,[2,2]),a=Z(n,e.conv_2,[1,1]),s=Z(a,e.conv_3,[2,2]),i=Z(s,e.conv_4,[1,1]),c=Z(i,e.conv_5,[2,2]),m=Z(c,e.conv_6,[1,1]),p=Z(m,e.conv_7,[2,2]),u=te(t,e.box_predictor_0),f=te(o,e.box_predictor_1),l=te(n,e.box_predictor_2),b=te(s,e.box_predictor_3),y=te(c,e.box_predictor_4),F=te(p,e.box_predictor_5),h=Ge.concat([u.boxPredictionEncoding,f.boxPredictionEncoding,l.boxPredictionEncoding,b.boxPredictionEncoding,y.boxPredictionEncoding,F.boxPredictionEncoding],1),T=Ge.concat([u.classPrediction,f.classPrediction,l.classPrediction,b.classPrediction,y.classPrediction,F.classPrediction],1);return{boxPredictions:h,classPredictions:T}})}var X=class{constructor({minConfidence:t,maxResults:e}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=e||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var St=class extends A{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("SsdMobilenetv1 - load model before inference");return Lt.tidy(()=>{let r=Lt.cast(t.toBatchTensor(512,!1),"float32"),n=Lt.sub(Lt.div(r,127.5),1),a=jo(n,e.mobilenetv1),{boxPredictions:s,classPredictions:i}=Jo(a.out,a.conv11,e.prediction_layer);return Xo(s,i,e.output_layer)})}async forward(t){return this.forwardInput(await C(t))}async locateFaces(t,e={}){let{maxResults:r,minConfidence:n}=new X(e),a=await C(t),{boxes:s,scores:i}=this.forwardInput(a),c=s[0],m=i[0];for(let _=1;_<s.length;_++)s[_].dispose(),i[_].dispose();let p=Array.from(m.dataSync()),f=Uo(c,p,r,.5,n),l=a.getReshapedInputDimensions(0),b=a.inputSize,y=b/l.width,F=b/l.height,h=c.arraySync(),T=f.map(_=>{let[E,W]=[Math.max(0,h[_][0]),Math.min(1,h[_][2])].map(q=>q*F),[tt,lt]=[Math.max(0,h[_][1]),Math.min(1,h[_][3])].map(q=>q*y);return new M(p[_],new Yt(tt,E,lt-tt,W-E),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return c.dispose(),m.dispose(),T}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return Go(t)}extractParams(t){return Yo(t)}};function qo(o){let t=new St;return t.extractWeights(o),t}function fa(o){return qo(o)}var lo=class extends St{};var Zo=.4,Ko=[new g(.738768,.874946),new g(2.42204,2.65704),new g(4.30971,7.04493),new g(10.246,4.59428),new g(12.6868,11.8741)],Qo=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],tn=[117.001,114.697,97.404],en="tiny_yolov2_model",rn="tiny_yolov2_separable_conv_model";var N=v(x());var Cr=o=>typeof o=="number";function ho(o){if(!o)throw new Error(`invalid config: ${o}`);if(typeof o.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${o.withSeparableConvs}`);if(!Cr(o.iouThreshold)||o.iouThreshold<0||o.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${o.iouThreshold}`);if(!Array.isArray(o.classes)||!o.classes.length||!o.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(o.classes)}`);if(!Array.isArray(o.anchors)||!o.anchors.length||!o.anchors.map(t=>t||{}).every(t=>Cr(t.x)&&Cr(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(o.anchors)}`);if(o.meanRgb&&(!Array.isArray(o.meanRgb)||o.meanRgb.length!==3||!o.meanRgb.every(Cr)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(o.meanRgb)}`)}var Q=v(x());var K=v(x());function Me(o){return K.tidy(()=>{let t=K.mul(o,K.scalar(.10000000149011612));return K.add(K.relu(K.sub(o,t)),t)})}function Tt(o,t){return Q.tidy(()=>{let e=Q.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=Q.conv2d(e,t.conv.filters,[1,1],"valid"),e=Q.sub(e,t.bn.sub),e=Q.mul(e,t.bn.truediv),e=Q.add(e,t.conv.bias),Me(e)})}var At=v(x());function wt(o,t){return At.tidy(()=>{let e=At.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=At.separableConv2d(e,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),e=At.add(e,t.bias),Me(e)})}var bo=v(x());function la(o,t){let e=be(o,t);function r(s,i){let c=bo.tensor1d(o(s)),m=bo.tensor1d(o(s));return t.push({paramPath:`${i}/sub`},{paramPath:`${i}/truediv`}),{sub:c,truediv:m}}function n(s,i,c){let m=e(s,i,3,`${c}/conv`),p=r(i,`${c}/bn`);return{conv:m,bn:p}}let a=ge(o,t);return{extractConvParams:e,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}}function on(o,t,e,r){let{extractWeights:n,getRemainingWeights:a}=R(o),s=[],{extractConvParams:i,extractConvWithBatchNormParams:c,extractSeparableConvParams:m}=la(n,s),p;if(t.withSeparableConvs){let[u,f,l,b,y,F,h,T,_]=r,E=t.isFirstLayerConv2d?i(u,f,3,"conv0"):m(u,f,"conv0"),W=m(f,l,"conv1"),tt=m(l,b,"conv2"),lt=m(b,y,"conv3"),q=m(y,F,"conv4"),Dt=m(F,h,"conv5"),Et=T?m(h,T,"conv6"):void 0,Mt=_?m(T,_,"conv7"):void 0,$t=i(_||T||h,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}else{let[u,f,l,b,y,F,h,T,_]=r,E=c(u,f,"conv0"),W=c(f,l,"conv1"),tt=c(l,b,"conv2"),lt=c(b,y,"conv3"),q=c(y,F,"conv4"),Dt=c(F,h,"conv5"),Et=c(h,T,"conv6"),Mt=c(T,_,"conv7"),$t=i(_,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:p,paramMappings:s}}function da(o,t){let e=Y(o,t);function r(i){let c=e(`${i}/sub`,1),m=e(`${i}/truediv`,1);return{sub:c,truediv:m}}function n(i){let c=e(`${i}/filters`,4),m=e(`${i}/bias`,1);return{filters:c,bias:m}}function a(i){let c=n(`${i}/conv`),m=r(`${i}/bn`);return{conv:c,bn:m}}let s=xe(e);return{extractConvParams:n,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}}function nn(o,t){let e=[],{extractConvParams:r,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}=da(o,e),s;if(t.withSeparableConvs){let i=t.filterSizes&&t.filterSizes.length||9;s={conv0:t.isFirstLayerConv2d?r("conv0"):a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:i>7?a("conv6"):void 0,conv7:i>8?a("conv7"):void 0,conv8:r("conv8")}}else s={conv0:n("conv0"),conv1:n("conv1"),conv2:n("conv2"),conv3:n("conv3"),conv4:n("conv4"),conv5:n("conv5"),conv6:n("conv6"),conv7:n("conv7"),conv8:r("conv8")};return B(o,e),{params:s,paramMappings:e}}var st=class{constructor({inputSize:t,scoreThreshold:e}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=e||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var go=class extends A{constructor(e){super("TinyYolov2");ho(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,r){let n=Tt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=Tt(n,r.conv6),n=Tt(n,r.conv7),qt(n,r.conv8,"valid",!1)}runMobilenet(e,r){let n=this.config.isFirstLayerConv2d?Me(qt(e,r.conv0,"valid",!1)):wt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=r.conv6?wt(n,r.conv6):n,n=r.conv7?wt(n,r.conv7):n,qt(n,r.conv8,"valid",!1)}forwardInput(e,r){let{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return N.tidy(()=>{let a=N.cast(e.toBatchTensor(r,!1),"float32");return a=this.config.meanRgb?rt(a,this.config.meanRgb):a,a=a.div(255),this.config.withSeparableConvs?this.runMobilenet(a,n):this.runTinyYolov2(a,n)})}async forward(e,r){return this.forwardInput(await C(e),r)}async detect(e,r={}){let{inputSize:n,scoreThreshold:a}=new st(r),s=await C(e),i=await this.forwardInput(s,n),c=N.tidy(()=>N.unstack(i)[0].expandDims()),m={width:s.getInputWidth(0),height:s.getInputHeight(0)},p=await this.extractBoxes(c,s.getReshapedInputDimensions(0),a);i.dispose(),c.dispose();let u=p.map(h=>h.box),f=p.map(h=>h.score),l=p.map(h=>h.classScore),b=p.map(h=>this.config.classes[h.label]);return Yr(u.map(h=>h.rescale(n)),f,this.config.iouThreshold,!0).map(h=>new bt(f[h],l[h],b[h],u[h],m))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return nn(e,this.config)}extractParams(e){let r=this.config.filterSizes||go.DEFAULT_FILTER_SIZES,n=r?r.length:void 0;if(n!==7&&n!==8&&n!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);return on(e,this.config,this.boxEncodingSize,r)}async extractBoxes(e,r,n){let{width:a,height:s}=r,i=Math.max(a,s),c=i/a,m=i/s,p=e.shape[1],u=this.config.anchors.length,[f,l,b]=N.tidy(()=>{let T=e.reshape([p,p,u,this.boxEncodingSize]),_=T.slice([0,0,0,0],[p,p,u,4]),E=T.slice([0,0,0,4],[p,p,u,1]),W=this.withClassScores?N.softmax(T.slice([0,0,0,5],[p,p,u,this.config.classes.length]),3):N.scalar(0);return[_,E,W]}),y=[],F=await l.array(),h=await f.array();for(let T=0;T<p;T++)for(let _=0;_<p;_++)for(let E=0;E<u;E++){let W=Ne(F[T][_][E][0]);if(!n||W>n){let tt=(_+Ne(h[T][_][E][0]))/p*c,lt=(T+Ne(h[T][_][E][1]))/p*m,q=Math.exp(h[T][_][E][2])*this.config.anchors[E].x/p*c,Dt=Math.exp(h[T][_][E][3])*this.config.anchors[E].y/p*m,Et=tt-q/2,Mt=lt-Dt/2,$t={row:T,col:_,anchor:E},{classScore:yo,label:_o}=this.withClassScores?await this.extractPredictedClass(b,$t):{classScore:1,label:0};y.push({box:new Vt(Et,Mt,Et+q,Mt+Dt),score:W,classScore:W*yo,label:_o,...$t})}}return f.dispose(),l.dispose(),b.dispose(),y}async extractPredictedClass(e,r){let{row:n,col:a,anchor:s}=r,i=await e.array();return Array(this.config.classes.length).fill(0).map((c,m)=>i[n][a][s][m]).map((c,m)=>({classScore:c,label:m})).reduce((c,m)=>c.classScore>m.classScore?c:m)}},ee=go;ee.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var re=class extends ee{constructor(t=!0){let e={withSeparableConvs:t,iouThreshold:Zo,classes:["face"],...t?{anchors:Qo,meanRgb:tn}:{anchors:Ko,withClassScores:!0}};super(e)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?rn:en}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function ha(o,t=!0){let e=new re(t);return e.extractWeights(o),e}var je=class extends st{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var J=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};var Xe=v(x());var xo=v(x());async function oe(o,t,e,r,n=({alignedRect:a})=>a){let a=o.map(c=>Zt(c)?n(c):c.detection),s=r||(t instanceof xo.Tensor?await de(t,a):await le(t,a)),i=await e(s);return s.forEach(c=>c instanceof xo.Tensor&&c.dispose()),i}async function Ce(o,t,e,r,n){return oe([o],t,async a=>e(a[0]),r,n)}var an=.4,sn=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],cn=[117.001,114.697,97.404];var ne=class extends ee{constructor(){let t={withSeparableConvs:!0,iouThreshold:an,classes:["face"],anchors:sn,meanRgb:cn,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var P={ssdMobilenetv1:new St,tinyFaceDetector:new ne,tinyYolov2:new re,faceLandmark68Net:new Kt,faceLandmark68TinyNet:new ze,faceRecognitionNet:new Qt,faceExpressionNet:new Oe,ageGenderNet:new He},mn=(o,t)=>P.ssdMobilenetv1.locateFaces(o,t),ba=(o,t)=>P.tinyFaceDetector.locateFaces(o,t),ga=(o,t)=>P.tinyYolov2.locateFaces(o,t),pn=o=>P.faceLandmark68Net.detectLandmarks(o),xa=o=>P.faceLandmark68TinyNet.detectLandmarks(o),va=o=>P.faceRecognitionNet.computeFaceDescriptor(o),ya=o=>P.faceExpressionNet.predictExpressions(o),_a=o=>P.ageGenderNet.predictAgeAndGender(o),un=o=>P.ssdMobilenetv1.load(o),Ta=o=>P.tinyFaceDetector.load(o),wa=o=>P.tinyYolov2.load(o),Pa=o=>P.faceLandmark68Net.load(o),Fa=o=>P.faceLandmark68TinyNet.load(o),Da=o=>P.faceRecognitionNet.load(o),Ea=o=>P.faceExpressionNet.load(o),Ma=o=>P.ageGenderNet.load(o),Ca=un,Ia=mn,Na=pn;var Ir=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ae=class extends Ir{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.faceExpressionNet.predictExpressions(n))),this.extractedFaces);return t.map((r,n)=>xr(r,e[n]))}withAgeAndGender(){return new ie(this,this.input)}},se=class extends Ir{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceExpressionNet.predictExpressions(r),this.extractedFaces);return xr(t,e)}withAgeAndGender(){return new ce(this,this.input)}},Wt=class extends ae{withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},kt=class extends se{withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Nr=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ie=class extends Nr{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.ageGenderNet.predictAgeAndGender(n))),this.extractedFaces);return t.map((r,n)=>{let{age:a,gender:s,genderProbability:i}=e[n];return Er(Mr(r,s,i),a)})}withFaceExpressions(){return new ae(this,this.input)}},ce=class extends Nr{async run(){let t=await this.parentTask;if(!t)return;let{age:e,gender:r,genderProbability:n}=await Ce(t,this.input,a=>P.ageGenderNet.predictAgeAndGender(a),this.extractedFaces);return Er(Mr(t,r,n),e)}withFaceExpressions(){return new se(this,this.input)}},Bt=class extends ie{withFaceExpressions(){return new Wt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Rt=class extends ce{withFaceExpressions(){return new kt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ue=class extends J{constructor(e,r){super();this.parentTask=e;this.input=r}},Pt=class extends Ue{async run(){let t=await this.parentTask;return(await oe(t,this.input,r=>Promise.all(r.map(n=>P.faceRecognitionNet.computeFaceDescriptor(n))),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}))).map((r,n)=>Dr(t[n],r))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}},Ft=class extends Ue{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceRecognitionNet.computeFaceDescriptor(r),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}));return Dr(t,e)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}};var Je=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?P.faceLandmark68TinyNet:P.faceLandmark68Net}},qe=class extends Je{async run(){let t=await this.parentTask,e=t.map(s=>s.detection),r=this.input instanceof Xe.Tensor?await de(this.input,e):await le(this.input,e),n=await Promise.all(r.map(s=>this.landmarkNet.detectLandmarks(s)));return r.forEach(s=>s instanceof Xe.Tensor&&s.dispose()),t.filter((s,i)=>n[i]).map((s,i)=>Pe(s,n[i]))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Ze=class extends Je{async run(){let t=await this.parentTask;if(!t)return;let{detection:e}=t,r=this.input instanceof Xe.Tensor?await de(this.input,[e]):await le(this.input,[e]),n=await this.landmarkNet.detectLandmarks(r[0]);return r.forEach(a=>a instanceof Xe.Tensor&&a.dispose()),Pe(t,n)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ke=class extends J{constructor(e,r=new X){super();this.input=e;this.options=r}},Ie=class extends Ke{async run(){let{input:t,options:e}=this,r;if(e instanceof je)r=P.tinyFaceDetector.locateFaces(t,e);else if(e instanceof X)r=P.ssdMobilenetv1.locateFaces(t,e);else if(e instanceof st)r=P.tinyYolov2.locateFaces(t,e);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return r}runAndExtendWithFaceDetections(){return new Promise((t,e)=>{this.run().then(r=>t(r.map(n=>jt({},n)))).catch(r=>e(r))})}withFaceLandmarks(t=!1){return new qe(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new ae(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new ie(this.runAndExtendWithFaceDetections(),this.input)}},Qe=class extends Ke{async run(){let t=await new Ie(this.input,this.options),e=t[0];return t.forEach(r=>{r.score>e.score&&(e=r)}),e}runAndExtendWithFaceDetection(){return new Promise(async t=>{let e=await this.run();t(e?jt({},e):void 0)})}withFaceLandmarks(t=!1){return new Ze(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new se(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new ce(this.runAndExtendWithFaceDetection(),this.input)}};function Sa(o,t=new X){return new Qe(o,t)}function Sr(o,t=new X){return new Ie(o,t)}async function fn(o,t){return Sr(o,new X(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function La(o,t={}){return Sr(o,new st(t)).withFaceLandmarks().withFaceDescriptors()}var Aa=fn;function vo(o,t){if(o.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let e=Array.from(o),r=Array.from(t);return Math.sqrt(e.map((n,a)=>n-r[a]).reduce((n,a)=>n+a*a,0))}var tr=class{constructor(t,e=.6){this._distanceThreshold=e;let r=Array.isArray(t)?t:[t];if(!r.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let n=1,a=()=>`person ${n++}`;this._labeledDescriptors=r.map(s=>{if(s instanceof mt)return s;if(s instanceof Float32Array)return new mt(a(),[s]);if(s.descriptor&&s.descriptor instanceof Float32Array)return new mt(a(),[s.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,e){return e.map(r=>vo(r,t)).reduce((r,n)=>r+n,0)/(e.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:e,label:r})=>new pe(r,this.computeMeanDistance(t,e))).reduce((e,r)=>e.distance<r.distance?e:r)}findBestMatch(t){let e=this.matchDescriptor(t);return e.distance<this._distanceThreshold?e:new pe("unknown",e.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let e=t.labeledDescriptors.map(r=>mt.fromJSON(r));return new tr(e,t.distanceThreshold)}};function Wa(o){let t=new ne;return t.extractWeights(o),t}function ln(o,t){let{width:e,height:r}=new k(t.width,t.height);if(e<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:e,height:r})}`);if(Array.isArray(o))return o.map(n=>ln(n,{width:e,height:r}));if(Zt(o)){let n=o.detection.forSize(e,r),a=o.unshiftedLandmarks.forSize(n.box.width,n.box.height);return Pe(jt(o,n),a)}return pt(o)?jt(o,o.detection.forSize(e,r)):o instanceof z||o instanceof M?o.forSize(e,r):o}var Ba=So;0&&(module.exports={AgeGenderNet,BoundingBox,Box,ComposableTask,ComputeAllFaceDescriptorsTask,ComputeFaceDescriptorsTaskBase,ComputeSingleFaceDescriptorTask,DetectAllFaceLandmarksTask,DetectAllFacesTask,DetectFaceLandmarksTaskBase,DetectFacesTaskBase,DetectSingleFaceLandmarksTask,DetectSingleFaceTask,Dimensions,FACE_EXPRESSION_LABELS,FaceDetection,FaceDetectionNet,FaceExpressionNet,FaceExpressions,FaceLandmark68Net,FaceLandmark68TinyNet,FaceLandmarkNet,FaceLandmarks,FaceLandmarks5,FaceLandmarks68,FaceMatch,FaceMatcher,FaceRecognitionNet,Gender,LabeledBox,LabeledFaceDescriptors,NetInput,NeuralNetwork,ObjectDetection,Point,PredictedBox,Rect,SsdMobilenetv1,SsdMobilenetv1Options,TinyFaceDetector,TinyFaceDetectorOptions,TinyYolov2,TinyYolov2Options,allFaces,allFacesSsdMobilenetv1,allFacesTinyYolov2,awaitMediaLoaded,bufferToImage,computeFaceDescriptor,createCanvas,createCanvasFromMedia,createFaceDetectionNet,createFaceRecognitionNet,createSsdMobilenetv1,createTinyFaceDetector,createTinyYolov2,detectAllFaces,detectFaceLandmarks,detectFaceLandmarksTiny,detectLandmarks,detectSingleFace,draw,env,euclideanDistance,extendWithAge,extendWithFaceDescriptor,extendWithFaceDetection,extendWithFaceExpressions,extendWithFaceLandmarks,extendWithGender,extractFaceTensors,extractFaces,fetchImage,fetchJson,fetchNetWeights,fetchOrThrow,fetchVideo,getContext2dOrThrow,getMediaDimensions,imageTensorToCanvas,imageToSquare,inverseSigmoid,iou,isMediaElement,isMediaLoaded,isWithAge,isWithFaceDetection,isWithFaceExpressions,isWithFaceLandmarks,isWithGender,loadAgeGenderModel,loadFaceDetectionModel,loadFaceExpressionModel,loadFaceLandmarkModel,loadFaceLandmarkTinyModel,loadFaceRecognitionModel,loadSsdMobilenetv1Model,loadTinyFaceDetectorModel,loadTinyYolov2Model,loadWeightMap,locateFaces,matchDimensions,minBbox,nets,nonMaxSuppression,normalize,padToSquare,predictAgeAndGender,recognizeFaceExpressions,resizeResults,resolveInput,shuffleArray,sigmoid,ssdMobilenetv1,tf,tinyFaceDetector,tinyYolov2,toNetInput,utils,validateConfig,version});
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __commonJS = (cb, mod) => function __require() {
+  return mod || (0, cb[__getOwnPropNames(cb)[0]])((mod = { exports: {} }).exports, mod), mod.exports;
+};
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// dist/tfjs.esm.js
+var require_tfjs_esm = __commonJS({
+  "dist/tfjs.esm.js"(exports, module2) {
+    "use strict";
+    var __defProp2 = Object.defineProperty;
+    var __getOwnPropDesc2 = Object.getOwnPropertyDescriptor;
+    var __getOwnPropNames2 = Object.getOwnPropertyNames;
+    var __hasOwnProp2 = Object.prototype.hasOwnProperty;
+    var __export2 = (target, all) => {
+      for (var name in all)
+        __defProp2(target, name, { get: all[name], enumerable: true });
+    };
+    var __copyProps2 = (to, from, except, desc) => {
+      if (from && typeof from === "object" || typeof from === "function") {
+        for (let key of __getOwnPropNames2(from))
+          if (!__hasOwnProp2.call(to, key) && key !== except)
+            __defProp2(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc2(from, key)) || desc.enumerable });
+      }
+      return to;
+    };
+    var __reExport = (target, mod, secondTarget) => (__copyProps2(target, mod, "default"), secondTarget && __copyProps2(secondTarget, mod, "default"));
+    var __toCommonJS2 = (mod) => __copyProps2(__defProp2({}, "__esModule", { value: true }), mod);
+    var tf_node_gpu_exports = {};
+    __export2(tf_node_gpu_exports, {
+      version: () => version6
+    });
+    module2.exports = __toCommonJS2(tf_node_gpu_exports);
+    __reExport(tf_node_gpu_exports, require("@tensorflow/tfjs-node-gpu"), module2.exports);
+    var version3 = "4.0.0";
+    var version22 = "4.0.0";
+    var version32 = "4.0.0";
+    var version4 = "4.0.0";
+    var version5 = "4.0.0";
+    var version6 = {
+      tfjs: version3,
+      "tfjs-core": version3,
+      "tfjs-converter": version22,
+      "tfjs-backend-cpu": version32,
+      "tfjs-backend-webgl": version4,
+      "tfjs-backend-wasm": version5
+    };
+  }
+});
+
+// src/index.ts
+var src_exports = {};
+__export(src_exports, {
+  AgeGenderNet: () => AgeGenderNet,
+  BoundingBox: () => BoundingBox,
+  Box: () => Box,
+  ComposableTask: () => ComposableTask,
+  ComputeAllFaceDescriptorsTask: () => ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase: () => ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask: () => ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask: () => DetectAllFaceLandmarksTask,
+  DetectAllFacesTask: () => DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase: () => DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase: () => DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask: () => DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask: () => DetectSingleFaceTask,
+  Dimensions: () => Dimensions,
+  FACE_EXPRESSION_LABELS: () => FACE_EXPRESSION_LABELS,
+  FaceDetection: () => FaceDetection,
+  FaceDetectionNet: () => FaceDetectionNet,
+  FaceExpressionNet: () => FaceExpressionNet,
+  FaceExpressions: () => FaceExpressions,
+  FaceLandmark68Net: () => FaceLandmark68Net,
+  FaceLandmark68TinyNet: () => FaceLandmark68TinyNet,
+  FaceLandmarkNet: () => FaceLandmarkNet,
+  FaceLandmarks: () => FaceLandmarks,
+  FaceLandmarks5: () => FaceLandmarks5,
+  FaceLandmarks68: () => FaceLandmarks68,
+  FaceMatch: () => FaceMatch,
+  FaceMatcher: () => FaceMatcher,
+  FaceRecognitionNet: () => FaceRecognitionNet,
+  Gender: () => Gender,
+  LabeledBox: () => LabeledBox,
+  LabeledFaceDescriptors: () => LabeledFaceDescriptors,
+  NetInput: () => NetInput,
+  NeuralNetwork: () => NeuralNetwork,
+  ObjectDetection: () => ObjectDetection,
+  Point: () => Point,
+  PredictedBox: () => PredictedBox,
+  Rect: () => Rect,
+  SsdMobilenetv1: () => SsdMobilenetv1,
+  SsdMobilenetv1Options: () => SsdMobilenetv1Options,
+  TinyFaceDetector: () => TinyFaceDetector,
+  TinyFaceDetectorOptions: () => TinyFaceDetectorOptions,
+  TinyYolov2: () => TinyYolov2,
+  TinyYolov2Options: () => TinyYolov2Options,
+  allFaces: () => allFaces,
+  allFacesSsdMobilenetv1: () => allFacesSsdMobilenetv1,
+  allFacesTinyYolov2: () => allFacesTinyYolov2,
+  awaitMediaLoaded: () => awaitMediaLoaded,
+  bufferToImage: () => bufferToImage,
+  computeFaceDescriptor: () => computeFaceDescriptor,
+  createCanvas: () => createCanvas,
+  createCanvasFromMedia: () => createCanvasFromMedia,
+  createFaceDetectionNet: () => createFaceDetectionNet,
+  createFaceRecognitionNet: () => createFaceRecognitionNet,
+  createSsdMobilenetv1: () => createSsdMobilenetv1,
+  createTinyFaceDetector: () => createTinyFaceDetector,
+  createTinyYolov2: () => createTinyYolov2,
+  detectAllFaces: () => detectAllFaces,
+  detectFaceLandmarks: () => detectFaceLandmarks,
+  detectFaceLandmarksTiny: () => detectFaceLandmarksTiny,
+  detectLandmarks: () => detectLandmarks,
+  detectSingleFace: () => detectSingleFace,
+  draw: () => draw_exports,
+  env: () => env,
+  euclideanDistance: () => euclideanDistance,
+  extendWithAge: () => extendWithAge,
+  extendWithFaceDescriptor: () => extendWithFaceDescriptor,
+  extendWithFaceDetection: () => extendWithFaceDetection,
+  extendWithFaceExpressions: () => extendWithFaceExpressions,
+  extendWithFaceLandmarks: () => extendWithFaceLandmarks,
+  extendWithGender: () => extendWithGender,
+  extractFaceTensors: () => extractFaceTensors,
+  extractFaces: () => extractFaces,
+  fetchImage: () => fetchImage,
+  fetchJson: () => fetchJson,
+  fetchNetWeights: () => fetchNetWeights,
+  fetchOrThrow: () => fetchOrThrow,
+  fetchVideo: () => fetchVideo,
+  getContext2dOrThrow: () => getContext2dOrThrow,
+  getMediaDimensions: () => getMediaDimensions,
+  imageTensorToCanvas: () => imageTensorToCanvas,
+  imageToSquare: () => imageToSquare,
+  inverseSigmoid: () => inverseSigmoid,
+  iou: () => iou,
+  isMediaElement: () => isMediaElement,
+  isMediaLoaded: () => isMediaLoaded,
+  isWithAge: () => isWithAge,
+  isWithFaceDetection: () => isWithFaceDetection,
+  isWithFaceExpressions: () => isWithFaceExpressions,
+  isWithFaceLandmarks: () => isWithFaceLandmarks,
+  isWithGender: () => isWithGender,
+  loadAgeGenderModel: () => loadAgeGenderModel,
+  loadFaceDetectionModel: () => loadFaceDetectionModel,
+  loadFaceExpressionModel: () => loadFaceExpressionModel,
+  loadFaceLandmarkModel: () => loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel: () => loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel: () => loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model: () => loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel: () => loadTinyFaceDetectorModel,
+  loadTinyYolov2Model: () => loadTinyYolov2Model,
+  loadWeightMap: () => loadWeightMap,
+  locateFaces: () => locateFaces,
+  matchDimensions: () => matchDimensions,
+  minBbox: () => minBbox,
+  nets: () => nets,
+  nonMaxSuppression: () => nonMaxSuppression,
+  normalize: () => normalize,
+  padToSquare: () => padToSquare,
+  predictAgeAndGender: () => predictAgeAndGender,
+  recognizeFaceExpressions: () => recognizeFaceExpressions,
+  resizeResults: () => resizeResults,
+  resolveInput: () => resolveInput,
+  shuffleArray: () => shuffleArray,
+  sigmoid: () => sigmoid,
+  ssdMobilenetv1: () => ssdMobilenetv1,
+  tf: () => tf42,
+  tinyFaceDetector: () => tinyFaceDetector,
+  tinyYolov2: () => tinyYolov2,
+  toNetInput: () => toNetInput,
+  utils: () => utils_exports,
+  validateConfig: () => validateConfig,
+  version: () => version2
+});
+module.exports = __toCommonJS(src_exports);
+var tf42 = __toESM(require_tfjs_esm());
+
+// src/draw/index.ts
+var draw_exports = {};
+__export(draw_exports, {
+  AnchorPosition: () => AnchorPosition,
+  DrawBox: () => DrawBox,
+  DrawBoxOptions: () => DrawBoxOptions,
+  DrawFaceLandmarks: () => DrawFaceLandmarks,
+  DrawFaceLandmarksOptions: () => DrawFaceLandmarksOptions,
+  DrawTextField: () => DrawTextField,
+  DrawTextFieldOptions: () => DrawTextFieldOptions,
+  drawContour: () => drawContour,
+  drawDetections: () => drawDetections,
+  drawFaceExpressions: () => drawFaceExpressions,
+  drawFaceLandmarks: () => drawFaceLandmarks
+});
+
+// src/draw/drawContour.ts
+function drawContour(ctx, points, isClosed = false) {
+  ctx.beginPath();
+  points.slice(1).forEach(({ x, y }, prevIdx) => {
+    const from = points[prevIdx];
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(x, y);
+  });
+  if (isClosed) {
+    const from = points[points.length - 1];
+    const to = points[0];
+    if (!from || !to) {
+      return;
+    }
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(to.x, to.y);
+  }
+  ctx.stroke();
+}
+
+// src/utils/index.ts
+var utils_exports = {};
+__export(utils_exports, {
+  computeReshapedDimensions: () => computeReshapedDimensions,
+  getCenterPoint: () => getCenterPoint,
+  isDimensions: () => isDimensions,
+  isEven: () => isEven,
+  isFloat: () => isFloat,
+  isTensor: () => isTensor,
+  isTensor1D: () => isTensor1D,
+  isTensor2D: () => isTensor2D,
+  isTensor3D: () => isTensor3D,
+  isTensor4D: () => isTensor4D,
+  isValidNumber: () => isValidNumber,
+  isValidProbablitiy: () => isValidProbablitiy,
+  range: () => range,
+  round: () => round
+});
+var tf = __toESM(require_tfjs_esm());
+
+// src/classes/Dimensions.ts
+var Dimensions = class {
+  constructor(width, height) {
+    if (!isValidNumber(width) || !isValidNumber(height)) {
+      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);
+    }
+    this._width = width;
+    this._height = height;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  reverse() {
+    return new Dimensions(1 / this.width, 1 / this.height);
+  }
+};
+
+// src/utils/index.ts
+function isTensor(tensor2, dim) {
+  return tensor2 instanceof tf.Tensor && tensor2.shape.length === dim;
+}
+function isTensor1D(tensor2) {
+  return isTensor(tensor2, 1);
+}
+function isTensor2D(tensor2) {
+  return isTensor(tensor2, 2);
+}
+function isTensor3D(tensor2) {
+  return isTensor(tensor2, 3);
+}
+function isTensor4D(tensor2) {
+  return isTensor(tensor2, 4);
+}
+function isFloat(num) {
+  return num % 1 !== 0;
+}
+function isEven(num) {
+  return num % 2 === 0;
+}
+function round(num, prec = 2) {
+  const f = 10 ** prec;
+  return Math.floor(num * f) / f;
+}
+function isDimensions(obj) {
+  return obj && obj.width && obj.height;
+}
+function computeReshapedDimensions({ width, height }, inputSize) {
+  const scale2 = inputSize / Math.max(height, width);
+  return new Dimensions(Math.round(width * scale2), Math.round(height * scale2));
+}
+function getCenterPoint(pts) {
+  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0)).div(new Point(pts.length, pts.length));
+}
+function range(num, start, step) {
+  return Array(num).fill(0).map((_, i) => start + i * step);
+}
+function isValidNumber(num) {
+  return !!num && num !== Infinity && num !== -Infinity && !Number.isNaN(num) || num === 0;
+}
+function isValidProbablitiy(num) {
+  return isValidNumber(num) && num >= 0 && num <= 1;
+}
+
+// src/classes/Point.ts
+var Point = class {
+  constructor(x, y) {
+    this._x = x;
+    this._y = y;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  add(pt) {
+    return new Point(this.x + pt.x, this.y + pt.y);
+  }
+  sub(pt) {
+    return new Point(this.x - pt.x, this.y - pt.y);
+  }
+  mul(pt) {
+    return new Point(this.x * pt.x, this.y * pt.y);
+  }
+  div(pt) {
+    return new Point(this.x / pt.x, this.y / pt.y);
+  }
+  abs() {
+    return new Point(Math.abs(this.x), Math.abs(this.y));
+  }
+  magnitude() {
+    return Math.sqrt(this.x ** 2 + this.y ** 2);
+  }
+  floor() {
+    return new Point(Math.floor(this.x), Math.floor(this.y));
+  }
+};
+
+// src/classes/Box.ts
+var Box = class {
+  static isRect(rect) {
+    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);
+  }
+  static assertIsValidBox(box, callee, allowNegativeDimensions = false) {
+    if (!Box.isRect(box)) {
+      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);
+    }
+    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {
+      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);
+    }
+  }
+  constructor(_box, allowNegativeDimensions = true) {
+    const box = _box || {};
+    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);
+    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);
+    if (!isRect && !isBbox) {
+      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);
+    }
+    const [x, y, width, height] = isRect ? [box.x, box.y, box.width, box.height] : [box.left, box.top, box.right - box.left, box.bottom - box.top];
+    Box.assertIsValidBox({
+      x,
+      y,
+      width,
+      height
+    }, "Box.constructor", allowNegativeDimensions);
+    this._x = x;
+    this._y = y;
+    this._width = width;
+    this._height = height;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  get left() {
+    return this.x;
+  }
+  get top() {
+    return this.y;
+  }
+  get right() {
+    return this.x + this.width;
+  }
+  get bottom() {
+    return this.y + this.height;
+  }
+  get area() {
+    return this.width * this.height;
+  }
+  get topLeft() {
+    return new Point(this.left, this.top);
+  }
+  get topRight() {
+    return new Point(this.right, this.top);
+  }
+  get bottomLeft() {
+    return new Point(this.left, this.bottom);
+  }
+  get bottomRight() {
+    return new Point(this.right, this.bottom);
+  }
+  round() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.round(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  floor() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.floor(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  toSquare() {
+    let {
+      x,
+      y,
+      width,
+      height
+    } = this;
+    const diff = Math.abs(width - height);
+    if (width < height) {
+      x -= diff / 2;
+      width += diff;
+    }
+    if (height < width) {
+      y -= diff / 2;
+      height += diff;
+    }
+    return new Box({ x, y, width, height });
+  }
+  rescale(s) {
+    const scaleX = isDimensions(s) ? s.width : s;
+    const scaleY = isDimensions(s) ? s.height : s;
+    return new Box({
+      x: this.x * scaleX,
+      y: this.y * scaleY,
+      width: this.width * scaleX,
+      height: this.height * scaleY
+    });
+  }
+  pad(padX, padY) {
+    const [x, y, width, height] = [
+      this.x - padX / 2,
+      this.y - padY / 2,
+      this.width + padX,
+      this.height + padY
+    ];
+    return new Box({ x, y, width, height });
+  }
+  clipAtImageBorders(imgWidth, imgHeight) {
+    const { x, y, right, bottom } = this;
+    const clippedX = Math.max(x, 0);
+    const clippedY = Math.max(y, 0);
+    const newWidth = right - clippedX;
+    const newHeight = bottom - clippedY;
+    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);
+    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);
+    return new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight }).floor();
+  }
+  shift(sx, sy) {
+    const { width, height } = this;
+    const x = this.x + sx;
+    const y = this.y + sy;
+    return new Box({ x, y, width, height });
+  }
+  padAtBorders(imageHeight, imageWidth) {
+    const w = this.width + 1;
+    const h = this.height + 1;
+    const dx = 1;
+    const dy = 1;
+    let edx = w;
+    let edy = h;
+    let x = this.left;
+    let y = this.top;
+    let ex = this.right;
+    let ey = this.bottom;
+    if (ex > imageWidth) {
+      edx = -ex + imageWidth + w;
+      ex = imageWidth;
+    }
+    if (ey > imageHeight) {
+      edy = -ey + imageHeight + h;
+      ey = imageHeight;
+    }
+    if (x < 1) {
+      edy = 2 - x;
+      x = 1;
+    }
+    if (y < 1) {
+      edy = 2 - y;
+      y = 1;
+    }
+    return { dy, edy, dx, edx, y, ey, x, ex, w, h };
+  }
+  calibrate(region) {
+    return new Box({
+      left: this.left + region.left * this.width,
+      top: this.top + region.top * this.height,
+      right: this.right + region.right * this.width,
+      bottom: this.bottom + region.bottom * this.height
+    }).toSquare().round();
+  }
+};
+
+// src/classes/BoundingBox.ts
+var BoundingBox = class extends Box {
+  constructor(left, top, right, bottom, allowNegativeDimensions = false) {
+    super({ left, top, right, bottom }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/ObjectDetection.ts
+var ObjectDetection = class {
+  constructor(score, classScore, className, relativeBox, imageDims) {
+    this._imageDims = new Dimensions(imageDims.width, imageDims.height);
+    this._score = score;
+    this._classScore = classScore;
+    this._className = className;
+    this._box = new Box(relativeBox).rescale(this._imageDims);
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+  get className() {
+    return this._className;
+  }
+  get box() {
+    return this._box;
+  }
+  get imageDims() {
+    return this._imageDims;
+  }
+  get imageWidth() {
+    return this.imageDims.width;
+  }
+  get imageHeight() {
+    return this.imageDims.height;
+  }
+  get relativeBox() {
+    return new Box(this._box).rescale(this.imageDims.reverse());
+  }
+  forSize(width, height) {
+    return new ObjectDetection(
+      this.score,
+      this.classScore,
+      this.className,
+      this.relativeBox,
+      { width, height }
+    );
+  }
+};
+
+// src/classes/FaceDetection.ts
+var FaceDetection = class extends ObjectDetection {
+  constructor(score, relativeBox, imageDims) {
+    super(score, score, "", relativeBox, imageDims);
+  }
+  forSize(width, height) {
+    const { score, relativeBox, imageDims } = super.forSize(width, height);
+    return new FaceDetection(score, relativeBox, imageDims);
+  }
+};
+
+// src/ops/iou.ts
+function iou(box1, box2, isIOU = true) {
+  const width = Math.max(0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));
+  const height = Math.max(0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));
+  const interSection = width * height;
+  return isIOU ? interSection / (box1.area + box2.area - interSection) : interSection / Math.min(box1.area, box2.area);
+}
+
+// src/ops/minBbox.ts
+function minBbox(pts) {
+  const xs = pts.map((pt) => pt.x);
+  const ys = pts.map((pt) => pt.y);
+  const minX = xs.reduce((min, x) => x < min ? x : min, Infinity);
+  const minY = ys.reduce((min, y) => y < min ? y : min, Infinity);
+  const maxX = xs.reduce((max, x) => max < x ? x : max, 0);
+  const maxY = ys.reduce((max, y) => max < y ? y : max, 0);
+  return new BoundingBox(minX, minY, maxX, maxY);
+}
+
+// src/ops/nonMaxSuppression.ts
+function nonMaxSuppression(boxes, scores, iouThreshold, isIOU = true) {
+  let indicesSortedByScore = scores.map((score, boxIndex) => ({ score, boxIndex })).sort((c1, c2) => c1.score - c2.score).map((c) => c.boxIndex);
+  const pick = [];
+  while (indicesSortedByScore.length > 0) {
+    const curr = indicesSortedByScore.pop();
+    pick.push(curr);
+    const indices = indicesSortedByScore;
+    const outputs = [];
+    for (let i = 0; i < indices.length; i++) {
+      const idx = indices[i];
+      const currBox = boxes[curr];
+      const idxBox = boxes[idx];
+      outputs.push(iou(currBox, idxBox, isIOU));
+    }
+    indicesSortedByScore = indicesSortedByScore.filter(
+      (_, j) => outputs[j] <= iouThreshold
+    );
+  }
+  return pick;
+}
+
+// src/ops/normalize.ts
+var tf2 = __toESM(require_tfjs_esm());
+function normalize(x, meanRgb) {
+  return tf2.tidy(() => {
+    const [r, g, b] = meanRgb;
+    const avg_r = tf2.fill([...x.shape.slice(0, 3), 1], r, "float32");
+    const avg_g = tf2.fill([...x.shape.slice(0, 3), 1], g, "float32");
+    const avg_b = tf2.fill([...x.shape.slice(0, 3), 1], b, "float32");
+    const avg_rgb = tf2.concat([avg_r, avg_g, avg_b], 3);
+    return tf2.sub(x, avg_rgb);
+  });
+}
+
+// src/ops/padToSquare.ts
+var tf3 = __toESM(require_tfjs_esm());
+function padToSquare(imgTensor, isCenterImage = false) {
+  return tf3.tidy(() => {
+    const [height, width] = imgTensor.shape.slice(1);
+    if (height === width)
+      return imgTensor;
+    const dimDiff = Math.abs(height - width);
+    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));
+    const paddingAxis = height > width ? 2 : 1;
+    const createPaddingTensor = (paddingAmountLocal) => {
+      const paddingTensorShape = imgTensor.shape.slice();
+      paddingTensorShape[paddingAxis] = paddingAmountLocal;
+      return tf3.fill(paddingTensorShape, 0, "float32");
+    };
+    const paddingTensorAppend = createPaddingTensor(paddingAmount);
+    const remainingPaddingAmount = dimDiff - paddingTensorAppend.shape[paddingAxis];
+    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;
+    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend].filter((t) => !!t).map((t) => tf3.cast(t, "float32"));
+    return tf3.concat(tensorsToStack, paddingAxis);
+  });
+}
+
+// src/ops/shuffleArray.ts
+function shuffleArray(inputArray) {
+  const array = inputArray.slice();
+  for (let i = array.length - 1; i > 0; i--) {
+    const j = Math.floor(Math.random() * (i + 1));
+    const x = array[i];
+    array[i] = array[j];
+    array[j] = x;
+  }
+  return array;
+}
+
+// src/ops/index.ts
+function sigmoid(x) {
+  return 1 / (1 + Math.exp(-x));
+}
+function inverseSigmoid(x) {
+  return Math.log(x / (1 - x));
+}
+
+// src/classes/Rect.ts
+var Rect = class extends Box {
+  constructor(x, y, width, height, allowNegativeDimensions = false) {
+    super({ x, y, width, height }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/FaceLandmarks.ts
+var relX = 0.5;
+var relY = 0.43;
+var relScale = 0.45;
+var FaceLandmarks = class {
+  constructor(relativeFaceLandmarkPositions, imgDims, shift = new Point(0, 0)) {
+    const { width, height } = imgDims;
+    this._imgDims = new Dimensions(width, height);
+    this._shift = shift;
+    this._positions = relativeFaceLandmarkPositions.map(
+      (pt) => pt.mul(new Point(width, height)).add(shift)
+    );
+  }
+  get shift() {
+    return new Point(this._shift.x, this._shift.y);
+  }
+  get imageWidth() {
+    return this._imgDims.width;
+  }
+  get imageHeight() {
+    return this._imgDims.height;
+  }
+  get positions() {
+    return this._positions;
+  }
+  get relativePositions() {
+    return this._positions.map(
+      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight))
+    );
+  }
+  forSize(width, height) {
+    return new this.constructor(
+      this.relativePositions,
+      { width, height }
+    );
+  }
+  shiftBy(x, y) {
+    return new this.constructor(
+      this.relativePositions,
+      this._imgDims,
+      new Point(x, y)
+    );
+  }
+  shiftByPoint(pt) {
+    return this.shiftBy(pt.x, pt.y);
+  }
+  align(detection, options = {}) {
+    if (detection) {
+      const box = detection instanceof FaceDetection ? detection.box.floor() : new Box(detection);
+      return this.shiftBy(box.x, box.y).align(null, options);
+    }
+    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };
+    if (useDlibAlignment) {
+      return this.alignDlib();
+    }
+    return this.alignMinBbox(minBoxPadding);
+  }
+  alignDlib() {
+    const centers = this.getRefPointsForAlignment();
+    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;
+    const distToMouth = (pt) => mouthCenter.sub(pt).magnitude();
+    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;
+    const size = Math.floor(eyeToMouthDist / relScale);
+    const refPoint = getCenterPoint(centers);
+    const x = Math.floor(Math.max(0, refPoint.x - relX * size));
+    const y = Math.floor(Math.max(0, refPoint.y - relY * size));
+    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));
+  }
+  alignMinBbox(padding) {
+    const box = minBbox(this.positions);
+    return box.pad(box.width * padding, box.height * padding);
+  }
+  getRefPointsForAlignment() {
+    throw new Error("getRefPointsForAlignment not implemented by base class");
+  }
+};
+
+// src/classes/FaceLandmarks5.ts
+var FaceLandmarks5 = class extends FaceLandmarks {
+  getRefPointsForAlignment() {
+    const pts = this.positions;
+    return [
+      pts[0],
+      pts[1],
+      getCenterPoint([pts[3], pts[4]])
+    ];
+  }
+};
+
+// src/classes/FaceLandmarks68.ts
+var FaceLandmarks68 = class extends FaceLandmarks {
+  getJawOutline() {
+    return this.positions.slice(0, 17);
+  }
+  getLeftEyeBrow() {
+    return this.positions.slice(17, 22);
+  }
+  getRightEyeBrow() {
+    return this.positions.slice(22, 27);
+  }
+  getNose() {
+    return this.positions.slice(27, 36);
+  }
+  getLeftEye() {
+    return this.positions.slice(36, 42);
+  }
+  getRightEye() {
+    return this.positions.slice(42, 48);
+  }
+  getMouth() {
+    return this.positions.slice(48, 68);
+  }
+  getRefPointsForAlignment() {
+    return [
+      this.getLeftEye(),
+      this.getRightEye(),
+      this.getMouth()
+    ].map(getCenterPoint);
+  }
+};
+
+// src/classes/FaceMatch.ts
+var FaceMatch = class {
+  constructor(label, distance) {
+    this._label = label;
+    this._distance = distance;
+  }
+  get label() {
+    return this._label;
+  }
+  get distance() {
+    return this._distance;
+  }
+  toString(withDistance = true) {
+    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ""}`;
+  }
+};
+
+// src/classes/LabeledBox.ts
+var LabeledBox = class extends Box {
+  constructor(box, label) {
+    super(box);
+    this._label = label;
+  }
+  static assertIsValidLabeledBox(box, callee) {
+    Box.assertIsValidBox(box, callee);
+    if (!isValidNumber(box.label)) {
+      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);
+    }
+  }
+  get label() {
+    return this._label;
+  }
+};
+
+// src/classes/LabeledFaceDescriptors.ts
+var LabeledFaceDescriptors = class {
+  constructor(label, descriptors) {
+    if (!(typeof label === "string")) {
+      throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");
+    }
+    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {
+      throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");
+    }
+    this._label = label;
+    this._descriptors = descriptors;
+  }
+  get label() {
+    return this._label;
+  }
+  get descriptors() {
+    return this._descriptors;
+  }
+  toJSON() {
+    return {
+      label: this.label,
+      descriptors: this.descriptors.map((d) => Array.from(d))
+    };
+  }
+  static fromJSON(json) {
+    const descriptors = json.descriptors.map((d) => new Float32Array(d));
+    return new LabeledFaceDescriptors(json.label, descriptors);
+  }
+};
+
+// src/classes/PredictedBox.ts
+var PredictedBox = class extends LabeledBox {
+  constructor(box, label, score, classScore) {
+    super(box, label);
+    this._score = score;
+    this._classScore = classScore;
+  }
+  static assertIsValidPredictedBox(box, callee) {
+    LabeledBox.assertIsValidLabeledBox(box, callee);
+    if (!isValidProbablitiy(box.score) || !isValidProbablitiy(box.classScore)) {
+      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);
+    }
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+};
+
+// src/factories/WithFaceDetection.ts
+function isWithFaceDetection(obj) {
+  return obj.detection instanceof FaceDetection;
+}
+function extendWithFaceDetection(sourceObj, detection) {
+  const extension = { detection };
+  return { ...sourceObj, ...extension };
+}
+
+// src/env/createBrowserEnv.ts
+function createBrowserEnv() {
+  const fetch = window.fetch;
+  if (!fetch)
+    throw new Error("fetch - missing fetch implementation for browser environment");
+  const readFile = () => {
+    throw new Error("readFile - filesystem not available for browser environment");
+  };
+  return {
+    Canvas: HTMLCanvasElement,
+    CanvasRenderingContext2D,
+    Image: HTMLImageElement,
+    ImageData,
+    Video: HTMLVideoElement,
+    createCanvasElement: () => document.createElement("canvas"),
+    createImageElement: () => document.createElement("img"),
+    createVideoElement: () => document.createElement("video"),
+    fetch,
+    readFile
+  };
+}
+
+// src/env/isNodejs.ts
+function isNodejs() {
+  return typeof global === "object" && typeof process !== "undefined" && process.versions != null && process.versions.node != null;
+}
+
+// src/env/createFileSystem.ts
+function createFileSystem(fs) {
+  let requireFsError = "";
+  if (!fs && isNodejs()) {
+    try {
+      fs = require("fs");
+    } catch (err) {
+      requireFsError = err.toString();
+    }
+  }
+  const readFile = fs ? (filePath) => new Promise((resolve, reject) => {
+    fs.readFile(filePath, (err, buffer) => err ? reject(err) : resolve(buffer));
+  }) : () => {
+    throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`);
+  };
+  return { readFile };
+}
+
+// src/env/createNodejsEnv.ts
+function createNodejsEnv() {
+  const Canvas = global["Canvas"] || global.HTMLCanvasElement;
+  const Image = global.Image || global.HTMLImageElement;
+  const Video = global["Video"] || global.HTMLVideoElement;
+  const createCanvasElement = () => {
+    if (Canvas)
+      return new Canvas();
+    throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment");
+  };
+  const createImageElement = () => {
+    if (Image)
+      return new Image();
+    throw new Error("createImageElement - missing Image implementation for nodejs environment");
+  };
+  const createVideoElement = () => {
+    if (Video)
+      return new Video();
+    throw new Error("createVideoElement - missing Video implementation for nodejs environment");
+  };
+  const fetch = global.fetch;
+  const fileSystem = createFileSystem();
+  return {
+    Canvas: Canvas || class {
+    },
+    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {
+    },
+    Image: Image || class {
+    },
+    ImageData: global.ImageData || class {
+    },
+    Video: global.HTMLVideoElement || class {
+    },
+    createCanvasElement,
+    createImageElement,
+    createVideoElement,
+    fetch,
+    ...fileSystem
+  };
+}
+
+// src/env/isBrowser.ts
+function isBrowser() {
+  return typeof window === "object" && typeof document !== "undefined" && typeof HTMLImageElement !== "undefined" && typeof HTMLCanvasElement !== "undefined" && typeof HTMLVideoElement !== "undefined" && typeof ImageData !== "undefined" && typeof CanvasRenderingContext2D !== "undefined";
+}
+
+// src/env/index.ts
+var environment;
+function getEnv() {
+  if (!environment) {
+    throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  return environment;
+}
+function setEnv(env2) {
+  environment = env2;
+}
+function initialize() {
+  if (isBrowser())
+    return setEnv(createBrowserEnv());
+  if (isNodejs())
+    return setEnv(createNodejsEnv());
+  return null;
+}
+function monkeyPatch(env2) {
+  if (!environment) {
+    initialize();
+  }
+  if (!environment) {
+    throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  const { Canvas = environment.Canvas, Image = environment.Image } = env2;
+  environment.Canvas = Canvas;
+  environment.Image = Image;
+  environment.createCanvasElement = env2.createCanvasElement || (() => new Canvas());
+  environment.createImageElement = env2.createImageElement || (() => new Image());
+  environment.ImageData = env2.ImageData || environment.ImageData;
+  environment.Video = env2.Video || environment.Video;
+  environment.fetch = env2.fetch || environment.fetch;
+  environment.readFile = env2.readFile || environment.readFile;
+}
+var env = {
+  getEnv,
+  setEnv,
+  initialize,
+  createBrowserEnv,
+  createFileSystem,
+  createNodejsEnv,
+  monkeyPatch,
+  isBrowser,
+  isNodejs
+};
+initialize();
+
+// src/dom/resolveInput.ts
+function resolveInput(arg) {
+  if (!env.isNodejs() && typeof arg === "string") {
+    return document.getElementById(arg);
+  }
+  return arg;
+}
+
+// src/dom/getContext2dOrThrow.ts
+function getContext2dOrThrow(canvasArg) {
+  const { Canvas, CanvasRenderingContext2D: CanvasRenderingContext2D2 } = env.getEnv();
+  if (canvasArg instanceof CanvasRenderingContext2D2) {
+    return canvasArg;
+  }
+  const canvas = resolveInput(canvasArg);
+  if (!(canvas instanceof Canvas)) {
+    throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+  }
+  const ctx = canvas.getContext("2d");
+  if (!ctx) {
+    throw new Error("resolveContext2d - canvas 2d context is null");
+  }
+  return ctx;
+}
+
+// src/draw/DrawTextField.ts
+var AnchorPosition = /* @__PURE__ */ ((AnchorPosition2) => {
+  AnchorPosition2["TOP_LEFT"] = "TOP_LEFT";
+  AnchorPosition2["TOP_RIGHT"] = "TOP_RIGHT";
+  AnchorPosition2["BOTTOM_LEFT"] = "BOTTOM_LEFT";
+  AnchorPosition2["BOTTOM_RIGHT"] = "BOTTOM_RIGHT";
+  return AnchorPosition2;
+})(AnchorPosition || {});
+var DrawTextFieldOptions = class {
+  constructor(options = {}) {
+    const {
+      anchorPosition,
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = options;
+    this.anchorPosition = anchorPosition || "TOP_LEFT" /* TOP_LEFT */;
+    this.backgroundColor = backgroundColor || "rgba(0, 0, 0, 0.5)";
+    this.fontColor = fontColor || "rgba(255, 255, 255, 1)";
+    this.fontSize = fontSize || 14;
+    this.fontStyle = fontStyle || "Georgia";
+    this.padding = padding || 4;
+  }
+};
+var DrawTextField = class {
+  constructor(text, anchor, options = {}) {
+    this.text = typeof text === "string" ? [text] : text instanceof DrawTextField ? text.text : text;
+    this.anchor = anchor;
+    this.options = new DrawTextFieldOptions(options);
+  }
+  measureWidth(ctx) {
+    const { padding } = this.options;
+    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => w0 < w1 ? w1 : w0, 0) + 2 * padding;
+  }
+  measureHeight() {
+    const { fontSize, padding } = this.options;
+    return this.text.length * fontSize + 2 * padding;
+  }
+  getUpperLeft(ctx, canvasDims) {
+    const { anchorPosition } = this.options;
+    const isShiftLeft = anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */ || anchorPosition === "TOP_RIGHT" /* TOP_RIGHT */;
+    const isShiftTop = anchorPosition === "BOTTOM_LEFT" /* BOTTOM_LEFT */ || anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */;
+    const textFieldWidth = this.measureWidth(ctx);
+    const textFieldHeight = this.measureHeight();
+    const x = isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x;
+    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;
+    if (canvasDims) {
+      const { width, height } = canvasDims;
+      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);
+      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);
+      return { x: newX, y: newY };
+    }
+    return { x, y };
+  }
+  draw(canvasArg) {
+    const canvas = resolveInput(canvasArg);
+    const ctx = getContext2dOrThrow(canvas);
+    const {
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = this.options;
+    ctx.font = `${fontSize}px ${fontStyle}`;
+    const maxTextWidth = this.measureWidth(ctx);
+    const textHeight = this.measureHeight();
+    ctx.fillStyle = backgroundColor;
+    const upperLeft = this.getUpperLeft(ctx, canvas);
+    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);
+    ctx.fillStyle = fontColor;
+    this.text.forEach((textLine, i) => {
+      const x = padding + upperLeft.x;
+      const y = padding + upperLeft.y + (i + 1) * fontSize;
+      ctx.fillText(textLine, x, y);
+    });
+  }
+};
+
+// src/draw/DrawBox.ts
+var DrawBoxOptions = class {
+  constructor(options = {}) {
+    const {
+      boxColor,
+      lineWidth,
+      label,
+      drawLabelOptions
+    } = options;
+    this.boxColor = boxColor || "rgba(0, 0, 255, 1)";
+    this.lineWidth = lineWidth || 2;
+    this.label = label;
+    const defaultDrawLabelOptions = {
+      anchorPosition: "BOTTOM_LEFT" /* BOTTOM_LEFT */,
+      backgroundColor: this.boxColor
+    };
+    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });
+  }
+};
+var DrawBox = class {
+  constructor(box, options = {}) {
+    this.box = new Box(box);
+    this.options = new DrawBoxOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const { boxColor, lineWidth } = this.options;
+    const {
+      x,
+      y,
+      width,
+      height
+    } = this.box;
+    ctx.strokeStyle = boxColor;
+    ctx.lineWidth = lineWidth;
+    ctx.strokeRect(x, y, width, height);
+    const { label } = this.options;
+    if (label) {
+      new DrawTextField([label], { x: x - lineWidth / 2, y }, this.options.drawLabelOptions).draw(canvasArg);
+    }
+  }
+};
+
+// src/draw/drawDetections.ts
+function drawDetections(canvasArg, detections) {
+  const detectionsArray = Array.isArray(detections) ? detections : [detections];
+  detectionsArray.forEach((det) => {
+    const score = det instanceof FaceDetection ? det.score : isWithFaceDetection(det) ? det.detection.score : void 0;
+    const box = det instanceof FaceDetection ? det.box : isWithFaceDetection(det) ? det.detection.box : new Box(det);
+    const label = score ? `${round(score)}` : void 0;
+    new DrawBox(box, { label }).draw(canvasArg);
+  });
+}
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var tf18 = __toESM(require_tfjs_esm());
+
+// src/dom/isMediaLoaded.ts
+function isMediaLoaded(media) {
+  const { Image, Video } = env.getEnv();
+  return media instanceof Image && media.complete || media instanceof Video && media.readyState >= 3;
+}
+
+// src/dom/awaitMediaLoaded.ts
+function awaitMediaLoaded(media) {
+  return new Promise((resolve, reject) => {
+    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media))
+      resolve(null);
+    function onError(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      reject(e);
+    }
+    function onLoad(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      resolve(e);
+    }
+    media.addEventListener("load", onLoad);
+    media.addEventListener("error", onError);
+  });
+}
+
+// src/dom/bufferToImage.ts
+function bufferToImage(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToImage - expected buf to be of type: Blob"));
+    const reader = new FileReader();
+    reader.onload = () => {
+      if (typeof reader.result !== "string")
+        reject(new Error("bufferToImage - expected reader.result to be a string, in onload"));
+      const img = env.getEnv().createImageElement();
+      img.onload = () => resolve(img);
+      img.onerror = reject;
+      img.src = reader.result;
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(buf);
+  });
+}
+
+// src/dom/getMediaDimensions.ts
+function getMediaDimensions(input) {
+  const { Image, Video } = env.getEnv();
+  if (input instanceof Image) {
+    return new Dimensions(input.naturalWidth, input.naturalHeight);
+  }
+  if (input instanceof Video) {
+    return new Dimensions(input.videoWidth, input.videoHeight);
+  }
+  return new Dimensions(input.width, input.height);
+}
+
+// src/dom/createCanvas.ts
+function createCanvas({ width, height }) {
+  const { createCanvasElement } = env.getEnv();
+  const canvas = createCanvasElement();
+  canvas.width = width;
+  canvas.height = height;
+  return canvas;
+}
+function createCanvasFromMedia(media, dims) {
+  const { ImageData: ImageData2 } = env.getEnv();
+  if (!(media instanceof ImageData2) && !isMediaLoaded(media)) {
+    throw new Error("createCanvasFromMedia - media has not finished loading yet");
+  }
+  const { width, height } = dims || getMediaDimensions(media);
+  const canvas = createCanvas({ width, height });
+  if (media instanceof ImageData2) {
+    getContext2dOrThrow(canvas).putImageData(media, 0, 0);
+  } else {
+    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);
+  }
+  return canvas;
+}
+
+// src/dom/imageTensorToCanvas.ts
+var tf4 = __toESM(require_tfjs_esm());
+async function imageTensorToCanvas(imgTensor, canvas) {
+  const targetCanvas = canvas || env.getEnv().createCanvasElement();
+  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);
+  const imgTensor3D = tf4.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());
+  await tf4["browser"].toPixels(imgTensor3D, targetCanvas);
+  imgTensor3D.dispose();
+  return targetCanvas;
+}
+
+// src/dom/isMediaElement.ts
+function isMediaElement(input) {
+  const { Image, Canvas, Video } = env.getEnv();
+  return input instanceof Image || input instanceof Canvas || input instanceof Video;
+}
+
+// src/dom/NetInput.ts
+var tf5 = __toESM(require_tfjs_esm());
+
+// src/dom/imageToSquare.ts
+function imageToSquare(input, inputSize, centerImage = false) {
+  const { Image, Canvas } = env.getEnv();
+  if (!(input instanceof Image || input instanceof Canvas)) {
+    throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+  }
+  if (inputSize <= 0)
+    return createCanvas({ width: 1, height: 1 });
+  const dims = getMediaDimensions(input);
+  const scale2 = inputSize / Math.max(dims.height, dims.width);
+  const width = scale2 * dims.width;
+  const height = scale2 * dims.height;
+  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });
+  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);
+  const offset = Math.abs(width - height) / 2;
+  const dx = centerImage && width < height ? offset : 0;
+  const dy = centerImage && height < width ? offset : 0;
+  if (inputCanvas.width > 0 && inputCanvas.height > 0)
+    getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);
+  return targetCanvas;
+}
+
+// src/dom/NetInput.ts
+var NetInput = class {
+  constructor(inputs, treatAsBatchInput = false) {
+    this._imageTensors = [];
+    this._canvases = [];
+    this._treatAsBatchInput = false;
+    this._inputDimensions = [];
+    this._inputSize = 0;
+    if (!Array.isArray(inputs)) {
+      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);
+    }
+    this._treatAsBatchInput = treatAsBatchInput;
+    this._batchSize = inputs.length;
+    inputs.forEach((input, idx) => {
+      if (isTensor3D(input)) {
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape;
+        return;
+      }
+      if (isTensor4D(input)) {
+        const batchSize = input.shape[0];
+        if (batchSize !== 1) {
+          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+        }
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape.slice(1);
+        return;
+      }
+      const canvas = input instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);
+      this._canvases[idx] = canvas;
+      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];
+    });
+  }
+  get imageTensors() {
+    return this._imageTensors;
+  }
+  get canvases() {
+    return this._canvases;
+  }
+  get isBatchInput() {
+    return this.batchSize > 1 || this._treatAsBatchInput;
+  }
+  get batchSize() {
+    return this._batchSize;
+  }
+  get inputDimensions() {
+    return this._inputDimensions;
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get reshapedInputDimensions() {
+    return range(this.batchSize, 0, 1).map(
+      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx)
+    );
+  }
+  getInput(batchIdx) {
+    return this.canvases[batchIdx] || this.imageTensors[batchIdx];
+  }
+  getInputDimensions(batchIdx) {
+    return this._inputDimensions[batchIdx];
+  }
+  getInputHeight(batchIdx) {
+    return this._inputDimensions[batchIdx][0];
+  }
+  getInputWidth(batchIdx) {
+    return this._inputDimensions[batchIdx][1];
+  }
+  getReshapedInputDimensions(batchIdx) {
+    if (typeof this.inputSize !== "number") {
+      throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+    }
+    const width = this.getInputWidth(batchIdx);
+    const height = this.getInputHeight(batchIdx);
+    return computeReshapedDimensions({ width, height }, this.inputSize);
+  }
+  toBatchTensor(inputSize, isCenterInputs = true) {
+    this._inputSize = inputSize;
+    return tf5.tidy(() => {
+      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {
+        const input = this.getInput(batchIdx);
+        if (input instanceof tf5.Tensor) {
+          let imgTensor = isTensor4D(input) ? input : tf5.expandDims(input);
+          imgTensor = padToSquare(imgTensor, isCenterInputs);
+          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {
+            imgTensor = tf5["image"].resizeBilinear(imgTensor, [inputSize, inputSize], false, false);
+          }
+          return imgTensor.as3D(inputSize, inputSize, 3);
+        }
+        if (input instanceof env.getEnv().Canvas) {
+          return tf5["browser"].fromPixels(imageToSquare(input, inputSize, isCenterInputs));
+        }
+        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);
+      });
+      const batchTensor = tf5.stack(inputTensors.map((t) => tf5.cast(t, "float32"))).as4D(this.batchSize, inputSize, inputSize, 3);
+      return batchTensor;
+    });
+  }
+};
+
+// src/dom/toNetInput.ts
+async function toNetInput(inputs) {
+  if (inputs instanceof NetInput)
+    return inputs;
+  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];
+  if (!inputArgArray.length)
+    throw new Error("toNetInput - empty array passed as input");
+  const getIdxHint = (idx) => Array.isArray(inputs) ? ` at input index ${idx}:` : "";
+  const inputArray = inputArgArray.map(resolveInput);
+  inputArray.forEach((input, i) => {
+    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {
+      if (typeof inputArgArray[i] === "string")
+        throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);
+      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);
+    }
+    if (isTensor4D(input)) {
+      const batchSize = input.shape[0];
+      if (batchSize !== 1)
+        throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+    }
+  });
+  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));
+  return new NetInput(inputArray, Array.isArray(inputs));
+}
+
+// src/dom/extractFaces.ts
+async function extractFaces(input, detections) {
+  const { Canvas } = env.getEnv();
+  let canvas = input;
+  if (!(input instanceof Canvas)) {
+    const netInput = await toNetInput(input);
+    if (netInput.batchSize > 1)
+      throw new Error("extractFaces - batchSize > 1 not supported");
+    const tensorOrCanvas = netInput.getInput(0);
+    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);
+  }
+  const ctx = getContext2dOrThrow(canvas);
+  const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det).map((box) => box.clipAtImageBorders(canvas.width, canvas.height));
+  return boxes.map(({ x, y, width, height }) => {
+    const faceImg = createCanvas({ width, height });
+    if (width > 0 && height > 0)
+      getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);
+    return faceImg;
+  });
+}
+
+// src/dom/extractFaceTensors.ts
+var tf6 = __toESM(require_tfjs_esm());
+async function extractFaceTensors(imageTensor, detections) {
+  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {
+    throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+  }
+  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {
+    throw new Error("extractFaceTensors - batchSize > 1 not supported");
+  }
+  return tf6.tidy(() => {
+    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);
+    const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det).map((box) => box.clipAtImageBorders(imgWidth, imgHeight));
+    const faceTensors = boxes.filter((box) => box.width > 0 && box.height > 0).map(({ x, y, width, height }) => tf6.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));
+    return faceTensors;
+  });
+}
+
+// src/dom/fetchOrThrow.ts
+async function fetchOrThrow(url, init) {
+  const { fetch } = env.getEnv();
+  const res = await fetch(url, init);
+  if (!(res.status < 400)) {
+    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);
+  }
+  return res;
+}
+
+// src/dom/fetchImage.ts
+async function fetchImage(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("image/")) {
+    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToImage(blob);
+}
+
+// src/dom/fetchJson.ts
+async function fetchJson(uri) {
+  return (await fetchOrThrow(uri)).json();
+}
+
+// src/dom/fetchNetWeights.ts
+async function fetchNetWeights(uri) {
+  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());
+}
+
+// src/dom/bufferToVideo.ts
+function bufferToVideo(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToVideo - expected buf to be of type: Blob"));
+    const video = env.getEnv().createVideoElement();
+    video.oncanplay = () => resolve(video);
+    video.onerror = reject;
+    video.playsInline = true;
+    video.muted = true;
+    video.src = URL.createObjectURL(buf);
+    video.play();
+  });
+}
+
+// src/dom/fetchVideo.ts
+async function fetchVideo(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("video/")) {
+    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToVideo(blob);
+}
+
+// src/dom/loadWeightMap.ts
+var tf7 = __toESM(require_tfjs_esm());
+
+// src/common/getModelUris.ts
+function getModelUris(uri, defaultModelName) {
+  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;
+  if (!uri) {
+    return {
+      modelBaseUri: "",
+      manifestUri: defaultManifestFilename
+    };
+  }
+  if (uri === "/") {
+    return {
+      modelBaseUri: "/",
+      manifestUri: `/${defaultManifestFilename}`
+    };
+  }
+  const protocol = uri.startsWith("http://") ? "http://" : uri.startsWith("https://") ? "https://" : "";
+  uri = uri.replace(protocol, "");
+  const parts = uri.split("/").filter((s) => s);
+  const manifestFile = uri.endsWith(".json") ? parts[parts.length - 1] : defaultManifestFilename;
+  let modelBaseUri = protocol + (uri.endsWith(".json") ? parts.slice(0, parts.length - 1) : parts).join("/");
+  modelBaseUri = uri.startsWith("/") ? `/${modelBaseUri}` : modelBaseUri;
+  return {
+    modelBaseUri,
+    manifestUri: modelBaseUri === "/" ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`
+  };
+}
+
+// src/dom/loadWeightMap.ts
+async function loadWeightMap(uri, defaultModelName) {
+  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);
+  const manifest = await fetchJson(manifestUri);
+  return tf7["io"].loadWeights(manifest, modelBaseUri);
+}
+
+// src/dom/matchDimensions.ts
+function matchDimensions(input, reference, useMediaDimensions = false) {
+  const { width, height } = useMediaDimensions ? getMediaDimensions(reference) : reference;
+  input.width = width;
+  input.height = height;
+  return { width, height };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var tf15 = __toESM(require_tfjs_esm());
+
+// src/NeuralNetwork.ts
+var tf8 = __toESM(require_tfjs_esm());
+var NeuralNetwork = class {
+  constructor(name) {
+    this._params = void 0;
+    this._paramMappings = [];
+    this._name = name;
+  }
+  get params() {
+    return this._params;
+  }
+  get paramMappings() {
+    return this._paramMappings;
+  }
+  get isLoaded() {
+    return !!this.params;
+  }
+  getParamFromPath(paramPath) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    return obj[objProp];
+  }
+  reassignParamFromPath(paramPath, tensor2) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    obj[objProp].dispose();
+    obj[objProp] = tensor2;
+  }
+  getParamList() {
+    return this._paramMappings.map(({ paramPath }) => ({
+      path: paramPath,
+      tensor: this.getParamFromPath(paramPath)
+    }));
+  }
+  getTrainableParams() {
+    return this.getParamList().filter((param) => param.tensor instanceof tf8.Variable);
+  }
+  getFrozenParams() {
+    return this.getParamList().filter((param) => !(param.tensor instanceof tf8.Variable));
+  }
+  variable() {
+    this.getFrozenParams().forEach(({ path, tensor: tensor2 }) => {
+      this.reassignParamFromPath(path, tensor2.variable());
+    });
+  }
+  freeze() {
+    this.getTrainableParams().forEach(({ path, tensor: variable }) => {
+      const tensor2 = tf8.tensor(variable.dataSync());
+      variable.dispose();
+      this.reassignParamFromPath(path, tensor2);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.getParamList().forEach((param) => {
+      if (throwOnRedispose && param.tensor.isDisposed) {
+        throw new Error(`param tensor has already been disposed for path ${param.path}`);
+      }
+      param.tensor.dispose();
+    });
+    this._params = void 0;
+  }
+  serializeParams() {
+    return new Float32Array(
+      this.getParamList().map(({ tensor: tensor2 }) => Array.from(tensor2.dataSync())).reduce((flat, arr) => flat.concat(arr))
+    );
+  }
+  async load(weightsOrUrl) {
+    if (weightsOrUrl instanceof Float32Array) {
+      this.extractWeights(weightsOrUrl);
+      return;
+    }
+    await this.loadFromUri(weightsOrUrl);
+  }
+  async loadFromUri(uri) {
+    if (uri && typeof uri !== "string") {
+      throw new Error(`${this._name}.loadFromUri - expected model uri`);
+    }
+    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());
+    this.loadFromWeightMap(weightMap);
+  }
+  async loadFromDisk(filePath) {
+    if (filePath && typeof filePath !== "string") {
+      throw new Error(`${this._name}.loadFromDisk - expected model file path`);
+    }
+    const { readFile } = env.getEnv();
+    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());
+    const fetchWeightsFromDisk = (filePaths) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));
+    const loadWeights = tf8["io"].weightsLoaderFactory(fetchWeightsFromDisk);
+    const manifest = JSON.parse((await readFile(manifestUri)).toString());
+    const weightMap = await loadWeights(manifest, modelBaseUri);
+    this.loadFromWeightMap(weightMap);
+  }
+  loadFromWeightMap(weightMap) {
+    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  extractWeights(weights) {
+    const { paramMappings, params } = this.extractParams(weights);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  traversePropertyPath(paramPath) {
+    if (!this.params) {
+      throw new Error("traversePropertyPath - model has no loaded params");
+    }
+    const result = paramPath.split("/").reduce((res, objProp2) => {
+      if (!res.nextObj.hasOwnProperty(objProp2)) {
+        throw new Error(`traversePropertyPath - object does not have property ${objProp2}, for path ${paramPath}`);
+      }
+      return { obj: res.nextObj, objProp: objProp2, nextObj: res.nextObj[objProp2] };
+    }, { nextObj: this.params });
+    const { obj, objProp } = result;
+    if (!obj || !objProp || !(obj[objProp] instanceof tf8.Tensor)) {
+      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);
+    }
+    return { obj, objProp };
+  }
+};
+
+// src/faceFeatureExtractor/denseBlock.ts
+var tf10 = __toESM(require_tfjs_esm());
+
+// src/common/depthwiseSeparableConv.ts
+var tf9 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv(x, params, stride) {
+  return tf9.tidy(() => {
+    let out = tf9.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, "same");
+    out = tf9.add(out, params.bias);
+    return out;
+  });
+}
+
+// src/faceFeatureExtractor/denseBlock.ts
+function denseBlock3(x, denseBlockParams, isFirstLayer = false) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, [2, 2], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, [2, 2])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+  });
+}
+function denseBlock4(x, denseBlockParams, isFirstLayer = false, isScaleDown = true) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, isScaleDown ? [2, 2] : [1, 1], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, isScaleDown ? [2, 2] : [1, 1])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    const in4 = tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, tf10.add(out3, out4))));
+  });
+}
+
+// src/common/convLayer.ts
+var tf11 = __toESM(require_tfjs_esm());
+function convLayer(x, params, padding = "same", withRelu = false) {
+  return tf11.tidy(() => {
+    const out = tf11.add(
+      tf11.conv2d(x, params.filters, [1, 1], padding),
+      params.bias
+    );
+    return withRelu ? tf11.relu(out) : out;
+  });
+}
+
+// src/common/disposeUnusedWeightTensors.ts
+function disposeUnusedWeightTensors(weightMap, paramMappings) {
+  Object.keys(weightMap).forEach((path) => {
+    if (!paramMappings.some((pm) => pm.originalPath === path)) {
+      weightMap[path].dispose();
+    }
+  });
+}
+
+// src/common/extractConvParamsFactory.ts
+var tf12 = __toESM(require_tfjs_esm());
+function extractConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, filterSize, mappedPrefix) => {
+    const filters = tf12.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf12.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  };
+}
+
+// src/common/extractFCParamsFactory.ts
+var tf13 = __toESM(require_tfjs_esm());
+function extractFCParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const fc_weights = tf13.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);
+    const fc_bias = tf13.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return {
+      weights: fc_weights,
+      bias: fc_bias
+    };
+  };
+}
+
+// src/common/extractSeparableConvParamsFactory.ts
+var tf14 = __toESM(require_tfjs_esm());
+
+// src/common/types.ts
+var SeparableConvParams = class {
+  constructor(depthwise_filter, pointwise_filter, bias) {
+    this.depthwise_filter = depthwise_filter;
+    this.pointwise_filter = pointwise_filter;
+    this.bias = bias;
+  }
+};
+
+// src/common/extractSeparableConvParamsFactory.ts
+function extractSeparableConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const depthwise_filter = tf14.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);
+    const pointwise_filter = tf14.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);
+    const bias = tf14.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/depthwise_filter` },
+      { paramPath: `${mappedPrefix}/pointwise_filter` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+function loadSeparableConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const depthwise_filter = extractWeightEntry(`${prefix}/depthwise_filter`, 4);
+    const pointwise_filter = extractWeightEntry(`${prefix}/pointwise_filter`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+
+// src/common/extractWeightEntryFactory.ts
+function extractWeightEntryFactory(weightMap, paramMappings) {
+  return (originalPath, paramRank, mappedPath) => {
+    const tensor2 = weightMap[originalPath];
+    if (!isTensor(tensor2, paramRank)) {
+      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor2}`);
+    }
+    paramMappings.push(
+      { originalPath, paramPath: mappedPath || originalPath }
+    );
+    return tensor2;
+  };
+}
+
+// src/common/extractWeightsFactory.ts
+function extractWeightsFactory(weights) {
+  let remainingWeights = weights;
+  function extractWeights(numWeights) {
+    const ret = remainingWeights.slice(0, numWeights);
+    remainingWeights = remainingWeights.slice(numWeights);
+    return ret;
+  }
+  function getRemainingWeights() {
+    return remainingWeights;
+  }
+  return {
+    extractWeights,
+    getRemainingWeights
+  };
+}
+
+// src/faceFeatureExtractor/extractorsFactory.ts
+function extractorsFactory(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`) : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);
+    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);
+    const conv22 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const { conv0, conv1, conv2: conv22 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);
+    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParams.ts
+function extractParams(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock4Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock4Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock4Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock4Params(64, 128, "dense2");
+  const dense3 = extractDenseBlock4Params(128, 256, "dense3");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: {
+      dense0,
+      dense1,
+      dense2,
+      dense3
+    }
+  };
+}
+
+// src/common/loadConvParamsFactory.ts
+function loadConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  };
+}
+
+// src/faceFeatureExtractor/loadParamsFactory.ts
+function loadParamsFactory(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractDenseBlock3Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock4Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock4Params("dense0", true),
+    dense1: extractDenseBlock4Params("dense1"),
+    dense2: extractDenseBlock4Params("dense2"),
+    dense3: extractDenseBlock4Params("dense3")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var FaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("FaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceFeatureExtractor - load model before inference");
+    }
+    return tf15.tidy(() => {
+      const batchTensor = tf15.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock4(normalized, params.dense0, true);
+      out = denseBlock4(out, params.dense1);
+      out = denseBlock4(out, params.dense2);
+      out = denseBlock4(out, params.dense3);
+      out = tf15.avgPool(out, [7, 7], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams(weights);
+  }
+};
+
+// src/faceProcessor/FaceProcessor.ts
+var tf17 = __toESM(require_tfjs_esm());
+
+// src/common/fullyConnectedLayer.ts
+var tf16 = __toESM(require_tfjs_esm());
+function fullyConnectedLayer(x, params) {
+  return tf16.tidy(() => tf16.add(
+    tf16.matMul(x, params.weights),
+    params.bias
+  ));
+}
+
+// src/faceProcessor/extractParams.ts
+function extractParams2(weights, channelsIn, channelsOut) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const fc = extractFCParams(channelsIn, channelsOut, "fc");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc }
+  };
+}
+
+// src/faceProcessor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap2(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: extractFcParams("fc")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceProcessor/util.ts
+function seperateWeightMaps(weightMap) {
+  const featureExtractorMap = {};
+  const classifierMap = {};
+  Object.keys(weightMap).forEach((key) => {
+    const map = key.startsWith("fc") ? classifierMap : featureExtractorMap;
+    map[key] = weightMap[key];
+  });
+  return { featureExtractorMap, classifierMap };
+}
+
+// src/faceProcessor/FaceProcessor.ts
+var FaceProcessor = class extends NeuralNetwork {
+  constructor(_name, faceFeatureExtractor) {
+    super(_name);
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf17.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams2(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap2(classifierMap);
+  }
+  extractParams(weights) {
+    const cIn = this.getClassifierChannelsIn();
+    const cOut = this.getClassifierChannelsOut();
+    const classifierWeightSize = cOut * cIn + cOut;
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressions.ts
+var FACE_EXPRESSION_LABELS = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"];
+var FaceExpressions = class {
+  constructor(probabilities) {
+    this.neutral = 0;
+    this.happy = 0;
+    this.sad = 0;
+    this.angry = 0;
+    this.fearful = 0;
+    this.disgusted = 0;
+    this.surprised = 0;
+    if (probabilities.length !== 7) {
+      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);
+    }
+    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {
+      this[expression] = probabilities[idx];
+    });
+  }
+  asSortedArray() {
+    return FACE_EXPRESSION_LABELS.map((expression) => ({ expression, probability: this[expression] })).sort((e0, e1) => e1.probability - e0.probability);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var FaceExpressionNet = class extends FaceProcessor {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceExpressionNet", faceFeatureExtractor);
+  }
+  forwardInput(input) {
+    return tf18.tidy(() => tf18.softmax(this.runNet(input)));
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictExpressions(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const probabilitesByBatch = await Promise.all(tf18.unstack(out).map(async (t) => {
+      const data = t.dataSync();
+      t.dispose();
+      return data;
+    }));
+    out.dispose();
+    const predictionsByBatch = probabilitesByBatch.map((probabilites) => new FaceExpressions(probabilites));
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_expression_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+  getClassifierChannelsOut() {
+    return 7;
+  }
+};
+
+// src/factories/WithFaceExpressions.ts
+function isWithFaceExpressions(obj) {
+  return obj.expressions instanceof FaceExpressions;
+}
+function extendWithFaceExpressions(sourceObj, expressions) {
+  const extension = { expressions };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/drawFaceExpressions.ts
+function drawFaceExpressions(canvasArg, faceExpressions, minConfidence = 0.1, textFieldAnchor) {
+  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];
+  faceExpressionsArray.forEach((e) => {
+    const expr = e instanceof FaceExpressions ? e : isWithFaceExpressions(e) ? e.expressions : void 0;
+    if (!expr) {
+      throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+    }
+    const sorted = expr.asSortedArray();
+    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);
+    const anchor = isWithFaceDetection(e) ? e.detection.box.bottomLeft : textFieldAnchor || new Point(0, 0);
+    const drawTextField = new DrawTextField(
+      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),
+      anchor
+    );
+    drawTextField.draw(canvasArg);
+  });
+}
+
+// src/factories/WithFaceLandmarks.ts
+function isWithFaceLandmarks(obj) {
+  return isWithFaceDetection(obj) && obj["landmarks"] instanceof FaceLandmarks && obj["unshiftedLandmarks"] instanceof FaceLandmarks && obj["alignedRect"] instanceof FaceDetection;
+}
+function calculateFaceAngle(mesh) {
+  const degrees = (radians) => radians * 180 / Math.PI;
+  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);
+  const angle = {
+    roll: void 0,
+    pitch: void 0,
+    yaw: void 0
+  };
+  const calcYaw = (leftPoint, midPoint, rightPoint) => {
+    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);
+    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);
+    return leftToMidpoint - rightToMidpoint;
+  };
+  const calcRoll = (lever, pivot) => {
+    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);
+    const opposite = pivot._y - lever._y;
+    const angleInRadians = Math.asin(opposite / hypotenuse);
+    const angleInDegrees = degrees(angleInRadians);
+    const normalizeAngle = Math.floor(90 - angleInDegrees);
+    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;
+    const result = normalizeAngle * tiltDirection;
+    return result;
+  };
+  const calcPitch = (leftPoint, midPoint, rightPoint) => {
+    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);
+    const baseCoords = {
+      _x: (leftPoint._x + rightPoint._x) / 2,
+      _y: (leftPoint._y + rightPoint._y) / 2
+    };
+    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);
+    const angleInRadians = Math.atan(midToBaseLength / base);
+    const angleInDegrees = Math.floor(degrees(angleInRadians));
+    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;
+    const result = angleInDegrees * direction;
+    return result;
+  };
+  if (!mesh || !mesh._positions || mesh._positions.length !== 68)
+    return angle;
+  const pt = mesh._positions;
+  angle.roll = calcRoll(pt[27], pt[66]);
+  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);
+  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);
+  return angle;
+}
+function extendWithFaceLandmarks(sourceObj, unshiftedLandmarks) {
+  const { box: shift } = sourceObj.detection;
+  const landmarks = unshiftedLandmarks.shiftBy(shift.x, shift.y);
+  const rect = landmarks.align();
+  const { imageDims } = sourceObj.detection;
+  const alignedRect = new FaceDetection(
+    sourceObj.detection.score,
+    rect.rescale(imageDims.reverse()),
+    imageDims
+  );
+  const angle = calculateFaceAngle(unshiftedLandmarks);
+  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/DrawFaceLandmarks.ts
+var DrawFaceLandmarksOptions = class {
+  constructor(options = {}) {
+    const {
+      drawLines = true,
+      drawPoints = true,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = options;
+    this.drawLines = drawLines;
+    this.drawPoints = drawPoints;
+    this.lineWidth = lineWidth || 1;
+    this.pointSize = pointSize || 2;
+    this.lineColor = lineColor || "rgba(0, 255, 255, 1)";
+    this.pointColor = pointColor || "rgba(255, 0, 255, 1)";
+  }
+};
+var DrawFaceLandmarks = class {
+  constructor(faceLandmarks, options = {}) {
+    this.faceLandmarks = faceLandmarks;
+    this.options = new DrawFaceLandmarksOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const {
+      drawLines,
+      drawPoints,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = this.options;
+    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {
+      ctx.strokeStyle = lineColor;
+      ctx.lineWidth = lineWidth;
+      drawContour(ctx, this.faceLandmarks.getJawOutline());
+      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getNose());
+      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);
+      drawContour(ctx, this.faceLandmarks.getRightEye(), true);
+      drawContour(ctx, this.faceLandmarks.getMouth(), true);
+    }
+    if (drawPoints) {
+      ctx.strokeStyle = pointColor;
+      ctx.fillStyle = pointColor;
+      const drawPoint = (pt) => {
+        ctx.beginPath();
+        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);
+        ctx.fill();
+      };
+      this.faceLandmarks.positions.forEach(drawPoint);
+    }
+  }
+};
+function drawFaceLandmarks(canvasArg, faceLandmarks) {
+  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];
+  faceLandmarksArray.forEach((f) => {
+    const landmarks = f instanceof FaceLandmarks ? f : isWithFaceLandmarks(f) ? f.landmarks : void 0;
+    if (!landmarks) {
+      throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+    }
+    new DrawFaceLandmarks(landmarks).draw(canvasArg);
+  });
+}
+
+// package.json
+var version = "1.7.6";
+
+// src/ageGenderNet/AgeGenderNet.ts
+var tf20 = __toESM(require_tfjs_esm());
+
+// src/xception/TinyXception.ts
+var tf19 = __toESM(require_tfjs_esm());
+
+// src/xception/extractParams.ts
+function extractorsFactory2(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractReductionBlockParams(channelsIn, channelsOut, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(channels, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParams3(weights, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = extractorsFactory2(extractWeights, paramMappings);
+  const entry_flow_conv_in = extractConvParams(3, 32, 3, "entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, "entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, "entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, "exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, "exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { entry_flow, middle_flow, exit_flow }
+  };
+}
+
+// src/xception/extractParamsFromWeightMap.ts
+function loadParamsFactory2(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractReductionBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParamsFromWeightMap3(weightMap, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = loadParamsFactory2(weightMap, paramMappings);
+  const entry_flow_conv_in = extractConvParams("entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams("entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams("entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams("exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams("exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };
+}
+
+// src/xception/TinyXception.ts
+function conv(x, params, stride) {
+  return tf19.add(tf19.conv2d(x, params.filters, stride, "same"), params.bias);
+}
+function reductionBlock(x, params, isActivateInput = true) {
+  let out = isActivateInput ? tf19.relu(x) : x;
+  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = tf19.maxPool(out, [3, 3], [2, 2], "same");
+  out = tf19.add(out, conv(x, params.expansion_conv, [2, 2]));
+  return out;
+}
+function mainBlock(x, params) {
+  let out = depthwiseSeparableConv(tf19.relu(x), params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv2, [1, 1]);
+  out = tf19.add(out, x);
+  return out;
+}
+var TinyXception = class extends NeuralNetwork {
+  constructor(numMainBlocks) {
+    super("TinyXception");
+    this._numMainBlocks = numMainBlocks;
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyXception - load model before inference");
+    }
+    return tf19.tidy(() => {
+      const batchTensor = tf19.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = tf19.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));
+      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);
+      out = reductionBlock(out, params.entry_flow.reduction_block_1);
+      range(this._numMainBlocks, 0, 1).forEach((idx) => {
+        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);
+      });
+      out = reductionBlock(out, params.exit_flow.reduction_block);
+      out = tf19.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "tiny_xception_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap3(weightMap, this._numMainBlocks);
+  }
+  extractParams(weights) {
+    return extractParams3(weights, this._numMainBlocks);
+  }
+};
+
+// src/ageGenderNet/extractParams.ts
+function extractParams4(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const age = extractFCParams(512, 1, "fc/age");
+  const gender = extractFCParams(512, 2, "fc/gender");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc: { age, gender } }
+  };
+}
+
+// src/ageGenderNet/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap4(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: {
+      age: extractFcParams("fc/age"),
+      gender: extractFcParams("fc/gender")
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ageGenderNet/types.ts
+var Gender = /* @__PURE__ */ ((Gender2) => {
+  Gender2["FEMALE"] = "female";
+  Gender2["MALE"] = "male";
+  return Gender2;
+})(Gender || {});
+
+// src/ageGenderNet/AgeGenderNet.ts
+var AgeGenderNet = class extends NeuralNetwork {
+  constructor(faceFeatureExtractor = new TinyXception(2)) {
+    super("AgeGenderNet");
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf20.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      const pooled = tf20.avgPool(bottleneckFeatures, [7, 7], [2, 2], "valid").as2D(bottleneckFeatures.shape[0], -1);
+      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();
+      const gender = fullyConnectedLayer(pooled, params.fc.gender);
+      return { age, gender };
+    });
+  }
+  forwardInput(input) {
+    return tf20.tidy(() => {
+      const { age, gender } = this.runNet(input);
+      return { age, gender: tf20.softmax(gender) };
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictAgeAndGender(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const ages = tf20.unstack(out.age);
+    const genders = tf20.unstack(out.gender);
+    const ageAndGenderTensors = ages.map((ageTensor, i) => ({
+      ageTensor,
+      genderTensor: genders[i]
+    }));
+    const predictionsByBatch = await Promise.all(
+      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {
+        const age = ageTensor.dataSync()[0];
+        const probMale = genderTensor.dataSync()[0];
+        const isMale = probMale > 0.5;
+        const gender = isMale ? "male" /* MALE */ : "female" /* FEMALE */;
+        const genderProbability = isMale ? probMale : 1 - probMale;
+        ageTensor.dispose();
+        genderTensor.dispose();
+        return { age, gender, genderProbability };
+      })
+    );
+    out.age.dispose();
+    out.gender.dispose();
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "age_gender_model";
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams4(weights);
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap4(classifierMap);
+  }
+  extractParams(weights) {
+    const classifierWeightSize = 512 * 1 + 1 + (512 * 2 + 2);
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68NetBase.ts
+var tf21 = __toESM(require_tfjs_esm());
+var FaceLandmark68NetBase = class extends FaceProcessor {
+  postProcess(output, inputSize, originalDimensions) {
+    const inputDimensions = originalDimensions.map(({ width, height }) => {
+      const scale2 = inputSize / Math.max(height, width);
+      return {
+        width: width * scale2,
+        height: height * scale2
+      };
+    });
+    const batchSize = inputDimensions.length;
+    return tf21.tidy(() => {
+      const createInterleavedTensor = (fillX, fillY) => tf21.stack([tf21.fill([68], fillX, "float32"), tf21.fill([68], fillY, "float32")], 1).as2D(1, 136).as1D();
+      const getPadding = (batchIdx, cond) => {
+        const { width, height } = inputDimensions[batchIdx];
+        return cond(width, height) ? Math.abs(width - height) / 2 : 0;
+      };
+      const getPaddingX = (batchIdx) => getPadding(batchIdx, (w, h) => w < h);
+      const getPaddingY = (batchIdx) => getPadding(batchIdx, (w, h) => h < w);
+      const landmarkTensors = output.mul(tf21.fill([batchSize, 136], inputSize, "float32")).sub(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        getPaddingX(batchIdx),
+        getPaddingY(batchIdx)
+      )))).div(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        inputDimensions[batchIdx].width,
+        inputDimensions[batchIdx].height
+      ))));
+      return landmarkTensors;
+    });
+  }
+  forwardInput(input) {
+    return tf21.tidy(() => {
+      const out = this.runNet(input);
+      return this.postProcess(
+        out,
+        input.inputSize,
+        input.inputDimensions.map(([height, width]) => ({ height, width }))
+      );
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async detectLandmarks(input) {
+    const netInput = await toNetInput(input);
+    const landmarkTensors = tf21.tidy(
+      () => tf21.unstack(this.forwardInput(netInput))
+    );
+    const landmarksForBatch = await Promise.all(landmarkTensors.map(
+      async (landmarkTensor, batchIdx) => {
+        const landmarksArray = Array.from(landmarkTensor.dataSync());
+        const xCoords = landmarksArray.filter((_, i) => isEven(i));
+        const yCoords = landmarksArray.filter((_, i) => !isEven(i));
+        return new FaceLandmarks68(
+          Array(68).fill(0).map((_, i) => new Point(xCoords[i], yCoords[i])),
+          {
+            height: netInput.getInputHeight(batchIdx),
+            width: netInput.getInputWidth(batchIdx)
+          }
+        );
+      }
+    ));
+    landmarkTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? landmarksForBatch : landmarksForBatch[0];
+  }
+  getClassifierChannelsOut() {
+    return 136;
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68Net.ts
+var FaceLandmark68Net = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceLandmark68Net", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+};
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var tf22 = __toESM(require_tfjs_esm());
+
+// src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts
+function extractParamsFromWeightMapTiny(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock3Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock3Params("dense0", true),
+    dense1: extractDenseBlock3Params("dense1"),
+    dense2: extractDenseBlock3Params("dense2")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/extractParamsTiny.ts
+function extractParamsTiny(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock3Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock3Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock3Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock3Params(64, 128, "dense2");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { dense0, dense1, dense2 }
+  };
+}
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var TinyFaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("TinyFaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyFaceFeatureExtractor - load model before inference");
+    }
+    return tf22.tidy(() => {
+      const batchTensor = tf22.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock3(normalized, params.dense0, true);
+      out = denseBlock3(out, params.dense1);
+      out = denseBlock3(out, params.dense2);
+      out = tf22.avgPool(out, [14, 14], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_tiny_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMapTiny(weightMap);
+  }
+  extractParams(weights) {
+    return extractParamsTiny(weights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68TinyNet.ts
+var FaceLandmark68TinyNet = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new TinyFaceFeatureExtractor()) {
+    super("FaceLandmark68TinyNet", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_tiny_model";
+  }
+  getClassifierChannelsIn() {
+    return 128;
+  }
+};
+
+// src/faceLandmarkNet/index.ts
+var FaceLandmarkNet = class extends FaceLandmark68Net {
+};
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var tf27 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/convLayer.ts
+var tf24 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/scaleLayer.ts
+var tf23 = __toESM(require_tfjs_esm());
+function scale(x, params) {
+  return tf23.add(tf23.mul(x, params.weights), params.biases);
+}
+
+// src/faceRecognitionNet/convLayer.ts
+function convLayer2(x, params, strides, withRelu, padding = "same") {
+  const { filters, bias } = params.conv;
+  let out = tf24.conv2d(x, filters, strides, padding);
+  out = tf24.add(out, bias);
+  out = scale(out, params.scale);
+  return withRelu ? tf24.relu(out) : out;
+}
+function conv2(x, params) {
+  return convLayer2(x, params, [1, 1], true);
+}
+function convNoRelu(x, params) {
+  return convLayer2(x, params, [1, 1], false);
+}
+function convDown(x, params) {
+  return convLayer2(x, params, [2, 2], true, "valid");
+}
+
+// src/faceRecognitionNet/extractParams.ts
+var tf25 = __toESM(require_tfjs_esm());
+function extractorsFactory3(extractWeights, paramMappings) {
+  function extractFilterValues(numFilterValues, numFilters, filterSize) {
+    const weights = extractWeights(numFilterValues);
+    const depth = weights.length / (numFilters * filterSize * filterSize);
+    if (isFloat(depth)) {
+      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);
+    }
+    return tf25.tidy(
+      () => tf25.transpose(
+        tf25.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),
+        [2, 3, 1, 0]
+      )
+    );
+  }
+  function extractConvParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);
+    const bias = tf25.tensor1d(extractWeights(numFilters));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  }
+  function extractScaleLayerParams(numWeights, mappedPrefix) {
+    const weights = tf25.tensor1d(extractWeights(numWeights));
+    const biases = tf25.tensor1d(extractWeights(numWeights));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/biases` }
+    );
+    return {
+      weights,
+      biases
+    };
+  }
+  function extractConvLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const conv3 = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);
+    const scale2 = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);
+    return { conv: conv3, scale: scale2 };
+  }
+  function extractResidualLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix, isDown = false) {
+    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);
+    const conv22 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);
+    return { conv1, conv2: conv22 };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParams5(weights) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory3(extractWeights, paramMappings);
+  const conv32_down = extractConvLayerParams(4704, 32, 7, "conv32_down");
+  const conv32_1 = extractResidualLayerParams(9216, 32, 3, "conv32_1");
+  const conv32_2 = extractResidualLayerParams(9216, 32, 3, "conv32_2");
+  const conv32_3 = extractResidualLayerParams(9216, 32, 3, "conv32_3");
+  const conv64_down = extractResidualLayerParams(36864, 64, 3, "conv64_down", true);
+  const conv64_1 = extractResidualLayerParams(36864, 64, 3, "conv64_1");
+  const conv64_2 = extractResidualLayerParams(36864, 64, 3, "conv64_2");
+  const conv64_3 = extractResidualLayerParams(36864, 64, 3, "conv64_3");
+  const conv128_down = extractResidualLayerParams(147456, 128, 3, "conv128_down", true);
+  const conv128_1 = extractResidualLayerParams(147456, 128, 3, "conv128_1");
+  const conv128_2 = extractResidualLayerParams(147456, 128, 3, "conv128_2");
+  const conv256_down = extractResidualLayerParams(589824, 256, 3, "conv256_down", true);
+  const conv256_1 = extractResidualLayerParams(589824, 256, 3, "conv256_1");
+  const conv256_2 = extractResidualLayerParams(589824, 256, 3, "conv256_2");
+  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, "conv256_down_out");
+  const fc = tf25.tidy(
+    () => tf25.transpose(tf25.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0])
+  );
+  paramMappings.push({ paramPath: "fc" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/extractParamsFromWeightMap.ts
+function extractorsFactory4(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractScaleLayerParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);
+    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);
+    return { weights, biases };
+  }
+  function extractConvLayerParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);
+    const scale2 = extractScaleLayerParams(prefix);
+    return { conv: { filters, bias }, scale: scale2 };
+  }
+  function extractResidualLayerParams(prefix) {
+    return {
+      conv1: extractConvLayerParams(`${prefix}/conv1`),
+      conv2: extractConvLayerParams(`${prefix}/conv2`)
+    };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParamsFromWeightMap5(weightMap) {
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory4(weightMap, paramMappings);
+  const conv32_down = extractConvLayerParams("conv32_down");
+  const conv32_1 = extractResidualLayerParams("conv32_1");
+  const conv32_2 = extractResidualLayerParams("conv32_2");
+  const conv32_3 = extractResidualLayerParams("conv32_3");
+  const conv64_down = extractResidualLayerParams("conv64_down");
+  const conv64_1 = extractResidualLayerParams("conv64_1");
+  const conv64_2 = extractResidualLayerParams("conv64_2");
+  const conv64_3 = extractResidualLayerParams("conv64_3");
+  const conv128_down = extractResidualLayerParams("conv128_down");
+  const conv128_1 = extractResidualLayerParams("conv128_1");
+  const conv128_2 = extractResidualLayerParams("conv128_2");
+  const conv256_down = extractResidualLayerParams("conv256_down");
+  const conv256_1 = extractResidualLayerParams("conv256_1");
+  const conv256_2 = extractResidualLayerParams("conv256_2");
+  const conv256_down_out = extractResidualLayerParams("conv256_down_out");
+  const { fc } = weightMap;
+  paramMappings.push({ originalPath: "fc", paramPath: "fc" });
+  if (!isTensor2D(fc)) {
+    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/residualLayer.ts
+var tf26 = __toESM(require_tfjs_esm());
+function residual(x, params) {
+  let out = conv2(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  out = tf26.add(out, x);
+  out = tf26.relu(out);
+  return out;
+}
+function residualDown(x, params) {
+  let out = convDown(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  let pooled = tf26.avgPool(x, 2, 2, "valid");
+  const zeros2 = tf26.zeros(pooled.shape);
+  const isPad = pooled.shape[3] !== out.shape[3];
+  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];
+  if (isAdjustShape) {
+    const padShapeX = [...out.shape];
+    padShapeX[1] = 1;
+    const zerosW = tf26.zeros(padShapeX);
+    out = tf26.concat([out, zerosW], 1);
+    const padShapeY = [...out.shape];
+    padShapeY[2] = 1;
+    const zerosH = tf26.zeros(padShapeY);
+    out = tf26.concat([out, zerosH], 2);
+  }
+  pooled = isPad ? tf26.concat([pooled, zeros2], 3) : pooled;
+  out = tf26.add(pooled, out);
+  out = tf26.relu(out);
+  return out;
+}
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var FaceRecognitionNet = class extends NeuralNetwork {
+  constructor() {
+    super("FaceRecognitionNet");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceRecognitionNet - load model before inference");
+    }
+    return tf27.tidy(() => {
+      const batchTensor = tf27.cast(input.toBatchTensor(150, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = convDown(normalized, params.conv32_down);
+      out = tf27.maxPool(out, 3, 2, "valid");
+      out = residual(out, params.conv32_1);
+      out = residual(out, params.conv32_2);
+      out = residual(out, params.conv32_3);
+      out = residualDown(out, params.conv64_down);
+      out = residual(out, params.conv64_1);
+      out = residual(out, params.conv64_2);
+      out = residual(out, params.conv64_3);
+      out = residualDown(out, params.conv128_down);
+      out = residual(out, params.conv128_1);
+      out = residual(out, params.conv128_2);
+      out = residualDown(out, params.conv256_down);
+      out = residual(out, params.conv256_1);
+      out = residual(out, params.conv256_2);
+      out = residualDown(out, params.conv256_down_out);
+      const globalAvg = out.mean([1, 2]);
+      const fullyConnected = tf27.matMul(globalAvg, params.fc);
+      return fullyConnected;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async computeFaceDescriptor(input) {
+    var _a;
+    if ((_a = input == null ? void 0 : input.shape) == null ? void 0 : _a.some((dim) => dim <= 0))
+      return new Float32Array(128);
+    const netInput = await toNetInput(input);
+    const faceDescriptorTensors = tf27.tidy(() => tf27.unstack(this.forwardInput(netInput)));
+    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data()));
+    faceDescriptorTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_recognition_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap5(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams5(weights);
+  }
+};
+
+// src/faceRecognitionNet/index.ts
+function createFaceRecognitionNet(weights) {
+  const net = new FaceRecognitionNet();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/factories/WithFaceDescriptor.ts
+function extendWithFaceDescriptor(sourceObj, descriptor) {
+  const extension = { descriptor };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithAge.ts
+function isWithAge(obj) {
+  return typeof obj.age === "number";
+}
+function extendWithAge(sourceObj, age) {
+  const extension = { age };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithGender.ts
+function isWithGender(obj) {
+  return (obj.gender === "male" /* MALE */ || obj.gender === "female" /* FEMALE */) && isValidProbablitiy(obj.genderProbability);
+}
+function extendWithGender(sourceObj, gender, genderProbability) {
+  const extension = { gender, genderProbability };
+  return { ...sourceObj, ...extension };
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var tf34 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/extractParams.ts
+var tf28 = __toESM(require_tfjs_esm());
+function extractorsFactory5(extractWeights, paramMappings) {
+  function extractDepthwiseConvParams(numChannels, mappedPrefix) {
+    const filters = tf28.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);
+    const batch_norm_scale = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_offset = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_mean = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_variance = tf28.tensor1d(extractWeights(numChannels));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/batch_norm_scale` },
+      { paramPath: `${mappedPrefix}/batch_norm_offset` },
+      { paramPath: `${mappedPrefix}/batch_norm_mean` },
+      { paramPath: `${mappedPrefix}/batch_norm_variance` }
+    );
+    return {
+      filters,
+      batch_norm_scale,
+      batch_norm_offset,
+      batch_norm_mean,
+      batch_norm_variance
+    };
+  }
+  function extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, isPointwiseConv) {
+    const filters = tf28.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf28.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/${isPointwiseConv ? "batch_norm_offset" : "bias"}` }
+    );
+    return { filters, bias };
+  }
+  function extractPointwiseConvParams(channelsIn, channelsOut, filterSize, mappedPrefix) {
+    const {
+      filters,
+      bias
+    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);
+    return {
+      filters,
+      batch_norm_offset: bias
+    };
+  }
+  function extractConvPairParams(channelsIn, channelsOut, mappedPrefix) {
+    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);
+    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);
+    return { depthwise_conv, pointwise_conv };
+  }
+  function extractMobilenetV1Params() {
+    const conv_0 = extractPointwiseConvParams(3, 32, 3, "mobilenetv1/conv_0");
+    const conv_1 = extractConvPairParams(32, 64, "mobilenetv1/conv_1");
+    const conv_2 = extractConvPairParams(64, 128, "mobilenetv1/conv_2");
+    const conv_3 = extractConvPairParams(128, 128, "mobilenetv1/conv_3");
+    const conv_4 = extractConvPairParams(128, 256, "mobilenetv1/conv_4");
+    const conv_5 = extractConvPairParams(256, 256, "mobilenetv1/conv_5");
+    const conv_6 = extractConvPairParams(256, 512, "mobilenetv1/conv_6");
+    const conv_7 = extractConvPairParams(512, 512, "mobilenetv1/conv_7");
+    const conv_8 = extractConvPairParams(512, 512, "mobilenetv1/conv_8");
+    const conv_9 = extractConvPairParams(512, 512, "mobilenetv1/conv_9");
+    const conv_10 = extractConvPairParams(512, 512, "mobilenetv1/conv_10");
+    const conv_11 = extractConvPairParams(512, 512, "mobilenetv1/conv_11");
+    const conv_12 = extractConvPairParams(512, 1024, "mobilenetv1/conv_12");
+    const conv_13 = extractConvPairParams(1024, 1024, "mobilenetv1/conv_13");
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      conv_8,
+      conv_9,
+      conv_10,
+      conv_11,
+      conv_12,
+      conv_13
+    };
+  }
+  function extractPredictionLayerParams() {
+    const conv_0 = extractPointwiseConvParams(1024, 256, 1, "prediction_layer/conv_0");
+    const conv_1 = extractPointwiseConvParams(256, 512, 3, "prediction_layer/conv_1");
+    const conv_2 = extractPointwiseConvParams(512, 128, 1, "prediction_layer/conv_2");
+    const conv_3 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_3");
+    const conv_4 = extractPointwiseConvParams(256, 128, 1, "prediction_layer/conv_4");
+    const conv_5 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_5");
+    const conv_6 = extractPointwiseConvParams(256, 64, 1, "prediction_layer/conv_6");
+    const conv_7 = extractPointwiseConvParams(64, 128, 3, "prediction_layer/conv_7");
+    const box_encoding_0_predictor = extractConvParams(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor");
+    const class_predictor_0 = extractConvParams(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor");
+    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor");
+    const class_predictor_1 = extractConvParams(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor");
+    const box_encoding_2_predictor = extractConvParams(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor");
+    const class_predictor_2 = extractConvParams(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor");
+    const box_encoding_3_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor");
+    const class_predictor_3 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor");
+    const box_encoding_4_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor");
+    const class_predictor_4 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor");
+    const box_encoding_5_predictor = extractConvParams(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor");
+    const class_predictor_5 = extractConvParams(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor");
+    const box_predictor_0 = {
+      box_encoding_predictor: box_encoding_0_predictor,
+      class_predictor: class_predictor_0
+    };
+    const box_predictor_1 = {
+      box_encoding_predictor: box_encoding_1_predictor,
+      class_predictor: class_predictor_1
+    };
+    const box_predictor_2 = {
+      box_encoding_predictor: box_encoding_2_predictor,
+      class_predictor: class_predictor_2
+    };
+    const box_predictor_3 = {
+      box_encoding_predictor: box_encoding_3_predictor,
+      class_predictor: class_predictor_3
+    };
+    const box_predictor_4 = {
+      box_encoding_predictor: box_encoding_4_predictor,
+      class_predictor: class_predictor_4
+    };
+    const box_predictor_5 = {
+      box_encoding_predictor: box_encoding_5_predictor,
+      class_predictor: class_predictor_5
+    };
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      box_predictor_0,
+      box_predictor_1,
+      box_predictor_2,
+      box_predictor_3,
+      box_predictor_4,
+      box_predictor_5
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParams6(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory5(extractWeights, paramMappings);
+  const mobilenetv1 = extractMobilenetV1Params();
+  const prediction_layer = extractPredictionLayerParams();
+  const extra_dim = tf28.tensor3d(
+    extractWeights(5118 * 4),
+    [1, 5118, 4]
+  );
+  const output_layer = {
+    extra_dim
+  };
+  paramMappings.push({ paramPath: "output_layer/extra_dim" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    params: {
+      mobilenetv1,
+      prediction_layer,
+      output_layer
+    },
+    paramMappings
+  };
+}
+
+// src/ssdMobilenetv1/extractParamsFromWeightMap.ts
+function extractorsFactory6(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractPointwiseConvParams(prefix, idx, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);
+    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);
+    return { filters, batch_norm_offset };
+  }
+  function extractConvPairParams(idx) {
+    const mappedPrefix = `mobilenetv1/conv_${idx}`;
+    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;
+    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;
+    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;
+    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);
+    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);
+    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);
+    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);
+    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);
+    return {
+      depthwise_conv: {
+        filters,
+        batch_norm_scale,
+        batch_norm_offset,
+        batch_norm_mean,
+        batch_norm_variance
+      },
+      pointwise_conv: extractPointwiseConvParams("MobilenetV1", idx, mappedPrefixPointwiseConv)
+    };
+  }
+  function extractMobilenetV1Params() {
+    return {
+      conv_0: extractPointwiseConvParams("MobilenetV1", 0, "mobilenetv1/conv_0"),
+      conv_1: extractConvPairParams(1),
+      conv_2: extractConvPairParams(2),
+      conv_3: extractConvPairParams(3),
+      conv_4: extractConvPairParams(4),
+      conv_5: extractConvPairParams(5),
+      conv_6: extractConvPairParams(6),
+      conv_7: extractConvPairParams(7),
+      conv_8: extractConvPairParams(8),
+      conv_9: extractConvPairParams(9),
+      conv_10: extractConvPairParams(10),
+      conv_11: extractConvPairParams(11),
+      conv_12: extractConvPairParams(12),
+      conv_13: extractConvPairParams(13)
+    };
+  }
+  function extractConvParams(prefix, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);
+    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);
+    return { filters, bias };
+  }
+  function extractBoxPredictorParams(idx) {
+    const box_encoding_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,
+      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`
+    );
+    const class_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/ClassPredictor`,
+      `prediction_layer/box_predictor_${idx}/class_predictor`
+    );
+    return { box_encoding_predictor, class_predictor };
+  }
+  function extractPredictionLayerParams() {
+    return {
+      conv_0: extractPointwiseConvParams("Prediction", 0, "prediction_layer/conv_0"),
+      conv_1: extractPointwiseConvParams("Prediction", 1, "prediction_layer/conv_1"),
+      conv_2: extractPointwiseConvParams("Prediction", 2, "prediction_layer/conv_2"),
+      conv_3: extractPointwiseConvParams("Prediction", 3, "prediction_layer/conv_3"),
+      conv_4: extractPointwiseConvParams("Prediction", 4, "prediction_layer/conv_4"),
+      conv_5: extractPointwiseConvParams("Prediction", 5, "prediction_layer/conv_5"),
+      conv_6: extractPointwiseConvParams("Prediction", 6, "prediction_layer/conv_6"),
+      conv_7: extractPointwiseConvParams("Prediction", 7, "prediction_layer/conv_7"),
+      box_predictor_0: extractBoxPredictorParams(0),
+      box_predictor_1: extractBoxPredictorParams(1),
+      box_predictor_2: extractBoxPredictorParams(2),
+      box_predictor_3: extractBoxPredictorParams(3),
+      box_predictor_4: extractBoxPredictorParams(4),
+      box_predictor_5: extractBoxPredictorParams(5)
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParamsFromWeightMap6(weightMap) {
+  const paramMappings = [];
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory6(weightMap, paramMappings);
+  const extra_dim = weightMap["Output/extra_dim"];
+  paramMappings.push({ originalPath: "Output/extra_dim", paramPath: "output_layer/extra_dim" });
+  if (!isTensor3D(extra_dim)) {
+    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);
+  }
+  const params = {
+    mobilenetv1: extractMobilenetV1Params(),
+    prediction_layer: extractPredictionLayerParams(),
+    output_layer: {
+      extra_dim
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var tf30 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/pointwiseConvLayer.ts
+var tf29 = __toESM(require_tfjs_esm());
+function pointwiseConvLayer(x, params, strides) {
+  return tf29.tidy(() => {
+    let out = tf29.conv2d(x, params.filters, strides, "same");
+    out = tf29.add(out, params.batch_norm_offset);
+    return tf29.clipByValue(out, 0, 6);
+  });
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var epsilon = 0.0010000000474974513;
+function depthwiseConvLayer(x, params, strides) {
+  return tf30.tidy(() => {
+    let out = tf30.depthwiseConv2d(x, params.filters, strides, "same");
+    out = tf30.batchNorm(
+      out,
+      params.batch_norm_mean,
+      params.batch_norm_variance,
+      params.batch_norm_offset,
+      params.batch_norm_scale,
+      epsilon
+    );
+    return tf30.clipByValue(out, 0, 6);
+  });
+}
+function getStridesForLayerIdx(layerIdx) {
+  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];
+}
+function mobileNetV1(x, params) {
+  return tf30.tidy(() => {
+    let conv11;
+    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
+    const convPairParams = [
+      params.conv_1,
+      params.conv_2,
+      params.conv_3,
+      params.conv_4,
+      params.conv_5,
+      params.conv_6,
+      params.conv_7,
+      params.conv_8,
+      params.conv_9,
+      params.conv_10,
+      params.conv_11,
+      params.conv_12,
+      params.conv_13
+    ];
+    convPairParams.forEach((param, i) => {
+      const layerIdx = i + 1;
+      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
+      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
+      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
+      if (layerIdx === 11)
+        conv11 = out;
+    });
+    if (conv11 === null) {
+      throw new Error("mobileNetV1 - output of conv layer 11 is null");
+    }
+    return {
+      out,
+      conv11
+    };
+  });
+}
+
+// src/ssdMobilenetv1/nonMaxSuppression.ts
+function IOU(boxes, i, j) {
+  const boxesData = boxes.arraySync();
+  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);
+  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);
+  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);
+  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);
+  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);
+  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);
+  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);
+  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0)
+    return 0;
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function nonMaxSuppression2(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  const numBoxes = boxes.shape[0];
+  const outputSize = Math.min(maxOutputSize, numBoxes);
+  const candidates = scores.map((score, boxIndex) => ({ score, boxIndex })).filter((c) => c.score > scoreThreshold).sort((c1, c2) => c2.score - c1.score);
+  const suppressFunc = (x) => x <= iouThreshold ? 1 : 0;
+  const selected = [];
+  candidates.forEach((c) => {
+    if (selected.length >= outputSize)
+      return;
+    const originalScore = c.score;
+    for (let j = selected.length - 1; j >= 0; --j) {
+      const iou2 = IOU(boxes, c.boxIndex, selected[j]);
+      if (iou2 === 0)
+        continue;
+      c.score *= suppressFunc(iou2);
+      if (c.score <= scoreThreshold)
+        break;
+    }
+    if (originalScore === c.score) {
+      selected.push(c.boxIndex);
+    }
+  });
+  return selected;
+}
+
+// src/ssdMobilenetv1/outputLayer.ts
+var tf31 = __toESM(require_tfjs_esm());
+function getCenterCoordinatesAndSizesLayer(x) {
+  const vec = tf31.unstack(tf31.transpose(x, [1, 0]));
+  const sizes = [
+    tf31.sub(vec[2], vec[0]),
+    tf31.sub(vec[3], vec[1])
+  ];
+  const centers = [
+    tf31.add(vec[0], tf31.div(sizes[0], 2)),
+    tf31.add(vec[1], tf31.div(sizes[1], 2))
+  ];
+  return { sizes, centers };
+}
+function decodeBoxesLayer(x0, x1) {
+  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);
+  const vec = tf31.unstack(tf31.transpose(x1, [1, 0]));
+  const div0_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[2], 5)), sizes[0]), 2);
+  const add0_out = tf31.add(tf31.mul(tf31.div(vec[0], 10), sizes[0]), centers[0]);
+  const div1_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[3], 5)), sizes[1]), 2);
+  const add1_out = tf31.add(tf31.mul(tf31.div(vec[1], 10), sizes[1]), centers[1]);
+  return tf31.transpose(
+    tf31.stack([
+      tf31.sub(add0_out, div0_out),
+      tf31.sub(add1_out, div1_out),
+      tf31.add(add0_out, div0_out),
+      tf31.add(add1_out, div1_out)
+    ]),
+    [1, 0]
+  );
+}
+function outputLayer(boxPredictions, classPredictions, params) {
+  return tf31.tidy(() => {
+    const batchSize = boxPredictions.shape[0];
+    let boxes = decodeBoxesLayer(
+      tf31.reshape(tf31.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]),
+      tf31.reshape(boxPredictions, [-1, 4])
+    );
+    boxes = tf31.reshape(boxes, [batchSize, boxes.shape[0] / batchSize, 4]);
+    const scoresAndClasses = tf31.sigmoid(tf31.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));
+    let scores = tf31.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]);
+    scores = tf31.reshape(scores, [batchSize, scores.shape[1]]);
+    const boxesByBatch = tf31.unstack(boxes);
+    const scoresByBatch = tf31.unstack(scores);
+    return { boxes: boxesByBatch, scores: scoresByBatch };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+var tf33 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/boxPredictionLayer.ts
+var tf32 = __toESM(require_tfjs_esm());
+function boxPredictionLayer(x, params) {
+  return tf32.tidy(() => {
+    const batchSize = x.shape[0];
+    const boxPredictionEncoding = tf32.reshape(
+      convLayer(x, params.box_encoding_predictor),
+      [batchSize, -1, 1, 4]
+    );
+    const classPrediction = tf32.reshape(
+      convLayer(x, params.class_predictor),
+      [batchSize, -1, 3]
+    );
+    return { boxPredictionEncoding, classPrediction };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+function predictionLayer(x, conv11, params) {
+  return tf33.tidy(() => {
+    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);
+    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);
+    const conv22 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);
+    const conv3 = pointwiseConvLayer(conv22, params.conv_3, [2, 2]);
+    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);
+    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);
+    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);
+    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);
+    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);
+    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);
+    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);
+    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);
+    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);
+    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);
+    const boxPredictions = tf33.concat([
+      boxPrediction0.boxPredictionEncoding,
+      boxPrediction1.boxPredictionEncoding,
+      boxPrediction2.boxPredictionEncoding,
+      boxPrediction3.boxPredictionEncoding,
+      boxPrediction4.boxPredictionEncoding,
+      boxPrediction5.boxPredictionEncoding
+    ], 1);
+    const classPredictions = tf33.concat([
+      boxPrediction0.classPrediction,
+      boxPrediction1.classPrediction,
+      boxPrediction2.classPrediction,
+      boxPrediction3.classPrediction,
+      boxPrediction4.classPrediction,
+      boxPrediction5.classPrediction
+    ], 1);
+    return {
+      boxPredictions,
+      classPredictions
+    };
+  });
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1Options.ts
+var SsdMobilenetv1Options = class {
+  constructor({ minConfidence, maxResults } = {}) {
+    this._name = "SsdMobilenetv1Options";
+    this._minConfidence = minConfidence || 0.5;
+    this._maxResults = maxResults || 100;
+    if (typeof this._minConfidence !== "number" || this._minConfidence <= 0 || this._minConfidence >= 1) {
+      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);
+    }
+    if (typeof this._maxResults !== "number") {
+      throw new Error(`${this._name} - expected maxResults to be a number`);
+    }
+  }
+  get minConfidence() {
+    return this._minConfidence;
+  }
+  get maxResults() {
+    return this._maxResults;
+  }
+};
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var SsdMobilenetv1 = class extends NeuralNetwork {
+  constructor() {
+    super("SsdMobilenetv1");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params)
+      throw new Error("SsdMobilenetv1 - load model before inference");
+    return tf34.tidy(() => {
+      const batchTensor = tf34.cast(input.toBatchTensor(512, false), "float32");
+      const x = tf34.sub(tf34.div(batchTensor, 127.5), 1);
+      const features = mobileNetV1(x, params.mobilenetv1);
+      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);
+      return outputLayer(boxPredictions, classPredictions, params.output_layer);
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async locateFaces(input, options = {}) {
+    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);
+    const netInput = await toNetInput(input);
+    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);
+    const boxes = _boxes[0];
+    const scores = _scores[0];
+    for (let i = 1; i < _boxes.length; i++) {
+      _boxes[i].dispose();
+      _scores[i].dispose();
+    }
+    const scoresData = Array.from(scores.dataSync());
+    const iouThreshold = 0.5;
+    const indices = nonMaxSuppression2(boxes, scoresData, maxResults, iouThreshold, minConfidence);
+    const reshapedDims = netInput.getReshapedInputDimensions(0);
+    const inputSize = netInput.inputSize;
+    const padX = inputSize / reshapedDims.width;
+    const padY = inputSize / reshapedDims.height;
+    const boxesData = boxes.arraySync();
+    const results = indices.map((idx) => {
+      const [top, bottom] = [
+        Math.max(0, boxesData[idx][0]),
+        Math.min(1, boxesData[idx][2])
+      ].map((val) => val * padY);
+      const [left, right] = [
+        Math.max(0, boxesData[idx][1]),
+        Math.min(1, boxesData[idx][3])
+      ].map((val) => val * padX);
+      return new FaceDetection(
+        scoresData[idx],
+        new Rect(left, top, right - left, bottom - top),
+        { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) }
+      );
+    });
+    boxes.dispose();
+    scores.dispose();
+    return results;
+  }
+  getDefaultModelName() {
+    return "ssd_mobilenetv1_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap6(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams6(weights);
+  }
+};
+
+// src/ssdMobilenetv1/index.ts
+function createSsdMobilenetv1(weights) {
+  const net = new SsdMobilenetv1();
+  net.extractWeights(weights);
+  return net;
+}
+function createFaceDetectionNet(weights) {
+  return createSsdMobilenetv1(weights);
+}
+var FaceDetectionNet = class extends SsdMobilenetv1 {
+};
+
+// src/tinyYolov2/const.ts
+var IOU_THRESHOLD = 0.4;
+var BOX_ANCHORS = [
+  new Point(0.738768, 0.874946),
+  new Point(2.42204, 2.65704),
+  new Point(4.30971, 7.04493),
+  new Point(10.246, 4.59428),
+  new Point(12.6868, 11.8741)
+];
+var BOX_ANCHORS_SEPARABLE = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB_SEPARABLE = [117.001, 114.697, 97.404];
+var DEFAULT_MODEL_NAME = "tiny_yolov2_model";
+var DEFAULT_MODEL_NAME_SEPARABLE_CONV = "tiny_yolov2_separable_conv_model";
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var tf39 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/config.ts
+var isNumber = (arg) => typeof arg === "number";
+function validateConfig(config) {
+  if (!config) {
+    throw new Error(`invalid config: ${config}`);
+  }
+  if (typeof config.withSeparableConvs !== "boolean") {
+    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);
+  }
+  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1) {
+    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);
+  }
+  if (!Array.isArray(config.classes) || !config.classes.length || !config.classes.every((c) => typeof c === "string")) {
+    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);
+  }
+  if (!Array.isArray(config.anchors) || !config.anchors.length || !config.anchors.map((a) => a || {}).every((a) => isNumber(a.x) && isNumber(a.y))) {
+    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);
+  }
+  if (config.meanRgb && (!Array.isArray(config.meanRgb) || config.meanRgb.length !== 3 || !config.meanRgb.every(isNumber))) {
+    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);
+  }
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+var tf36 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/leaky.ts
+var tf35 = __toESM(require_tfjs_esm());
+function leaky(x) {
+  return tf35.tidy(() => {
+    const min = tf35.mul(x, tf35.scalar(0.10000000149011612));
+    return tf35.add(tf35.relu(tf35.sub(x, min)), min);
+  });
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+function convWithBatchNorm(x, params) {
+  return tf36.tidy(() => {
+    let out = tf36.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf36.conv2d(out, params.conv.filters, [1, 1], "valid");
+    out = tf36.sub(out, params.bn.sub);
+    out = tf36.mul(out, params.bn.truediv);
+    out = tf36.add(out, params.conv.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/depthwiseSeparableConv.ts
+var tf37 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv2(x, params) {
+  return tf37.tidy(() => {
+    let out = tf37.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf37.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], "valid");
+    out = tf37.add(out, params.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/extractParams.ts
+var tf38 = __toESM(require_tfjs_esm());
+function extractorsFactory7(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  function extractBatchNormParams(size, mappedPrefix) {
+    const sub6 = tf38.tensor1d(extractWeights(size));
+    const truediv = tf38.tensor1d(extractWeights(size));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/sub` },
+      { paramPath: `${mappedPrefix}/truediv` }
+    );
+    return { sub: sub6, truediv };
+  }
+  function extractConvWithBatchNormParams(channelsIn, channelsOut, mappedPrefix) {
+    const conv3 = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);
+    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParams7(weights, config, boxEncodingSize, filterSizes) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory7(extractWeights, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = config.isFirstLayerConv2d ? extractConvParams(s0, s1, 3, "conv0") : extractSeparableConvParams(s0, s1, "conv0");
+    const conv1 = extractSeparableConvParams(s1, s2, "conv1");
+    const conv22 = extractSeparableConvParams(s2, s3, "conv2");
+    const conv3 = extractSeparableConvParams(s3, s4, "conv3");
+    const conv4 = extractSeparableConvParams(s4, s5, "conv4");
+    const conv5 = extractSeparableConvParams(s5, s6, "conv5");
+    const conv6 = s7 ? extractSeparableConvParams(s6, s7, "conv6") : void 0;
+    const conv7 = s8 ? extractSeparableConvParams(s7, s8, "conv7") : void 0;
+    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  } else {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = extractConvWithBatchNormParams(s0, s1, "conv0");
+    const conv1 = extractConvWithBatchNormParams(s1, s2, "conv1");
+    const conv22 = extractConvWithBatchNormParams(s2, s3, "conv2");
+    const conv3 = extractConvWithBatchNormParams(s3, s4, "conv3");
+    const conv4 = extractConvWithBatchNormParams(s4, s5, "conv4");
+    const conv5 = extractConvWithBatchNormParams(s5, s6, "conv5");
+    const conv6 = extractConvWithBatchNormParams(s6, s7, "conv6");
+    const conv7 = extractConvWithBatchNormParams(s7, s8, "conv7");
+    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  }
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/extractParamsFromWeightMap.ts
+function extractorsFactory8(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractBatchNormParams(prefix) {
+    const sub6 = extractWeightEntry(`${prefix}/sub`, 1);
+    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);
+    return { sub: sub6, truediv };
+  }
+  function extractConvParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  }
+  function extractConvWithBatchNormParams(prefix) {
+    const conv3 = extractConvParams(`${prefix}/conv`);
+    const bn = extractBatchNormParams(`${prefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParamsFromWeightMap7(weightMap, config) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory8(weightMap, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const numFilters = config.filterSizes && config.filterSizes.length || 9;
+    params = {
+      conv0: config.isFirstLayerConv2d ? extractConvParams("conv0") : extractSeparableConvParams("conv0"),
+      conv1: extractSeparableConvParams("conv1"),
+      conv2: extractSeparableConvParams("conv2"),
+      conv3: extractSeparableConvParams("conv3"),
+      conv4: extractSeparableConvParams("conv4"),
+      conv5: extractSeparableConvParams("conv5"),
+      conv6: numFilters > 7 ? extractSeparableConvParams("conv6") : void 0,
+      conv7: numFilters > 8 ? extractSeparableConvParams("conv7") : void 0,
+      conv8: extractConvParams("conv8")
+    };
+  } else {
+    params = {
+      conv0: extractConvWithBatchNormParams("conv0"),
+      conv1: extractConvWithBatchNormParams("conv1"),
+      conv2: extractConvWithBatchNormParams("conv2"),
+      conv3: extractConvWithBatchNormParams("conv3"),
+      conv4: extractConvWithBatchNormParams("conv4"),
+      conv5: extractConvWithBatchNormParams("conv5"),
+      conv6: extractConvWithBatchNormParams("conv6"),
+      conv7: extractConvWithBatchNormParams("conv7"),
+      conv8: extractConvParams("conv8")
+    };
+  }
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/TinyYolov2Options.ts
+var TinyYolov2Options = class {
+  constructor({ inputSize, scoreThreshold } = {}) {
+    this._name = "TinyYolov2Options";
+    this._inputSize = inputSize || 416;
+    this._scoreThreshold = scoreThreshold || 0.5;
+    if (typeof this._inputSize !== "number" || this._inputSize % 32 !== 0) {
+      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);
+    }
+    if (typeof this._scoreThreshold !== "number" || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {
+      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);
+    }
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get scoreThreshold() {
+    return this._scoreThreshold;
+  }
+};
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var _TinyYolov2Base = class extends NeuralNetwork {
+  constructor(config) {
+    super("TinyYolov2");
+    validateConfig(config);
+    this._config = config;
+  }
+  get config() {
+    return this._config;
+  }
+  get withClassScores() {
+    return this.config.withClassScores || this.config.classes.length > 1;
+  }
+  get boxEncodingSize() {
+    return 5 + (this.withClassScores ? this.config.classes.length : 0);
+  }
+  runTinyYolov2(x, params) {
+    let out = convWithBatchNorm(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = convWithBatchNorm(out, params.conv6);
+    out = convWithBatchNorm(out, params.conv7);
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  runMobilenet(x, params) {
+    let out = this.config.isFirstLayerConv2d ? leaky(convLayer(x, params.conv0, "valid", false)) : depthwiseSeparableConv2(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = params.conv6 ? depthwiseSeparableConv2(out, params.conv6) : out;
+    out = params.conv7 ? depthwiseSeparableConv2(out, params.conv7) : out;
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  forwardInput(input, inputSize) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyYolov2 - load model before inference");
+    }
+    return tf39.tidy(() => {
+      let batchTensor = tf39.cast(input.toBatchTensor(inputSize, false), "float32");
+      batchTensor = this.config.meanRgb ? normalize(batchTensor, this.config.meanRgb) : batchTensor;
+      batchTensor = batchTensor.div(255);
+      return this.config.withSeparableConvs ? this.runMobilenet(batchTensor, params) : this.runTinyYolov2(batchTensor, params);
+    });
+  }
+  async forward(input, inputSize) {
+    return this.forwardInput(await toNetInput(input), inputSize);
+  }
+  async detect(input, forwardParams = {}) {
+    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput, inputSize);
+    const out0 = tf39.tidy(() => tf39.unstack(out)[0].expandDims());
+    const inputDimensions = {
+      width: netInput.getInputWidth(0),
+      height: netInput.getInputHeight(0)
+    };
+    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);
+    out.dispose();
+    out0.dispose();
+    const boxes = results.map((res) => res.box);
+    const scores = results.map((res) => res.score);
+    const classScores = results.map((res) => res.classScore);
+    const classNames = results.map((res) => this.config.classes[res.label]);
+    const indices = nonMaxSuppression(
+      boxes.map((box) => box.rescale(inputSize)),
+      scores,
+      this.config.iouThreshold,
+      true
+    );
+    const detections = indices.map((idx) => new ObjectDetection(
+      scores[idx],
+      classScores[idx],
+      classNames[idx],
+      boxes[idx],
+      inputDimensions
+    ));
+    return detections;
+  }
+  getDefaultModelName() {
+    return "";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap7(weightMap, this.config);
+  }
+  extractParams(weights) {
+    const filterSizes = this.config.filterSizes || _TinyYolov2Base.DEFAULT_FILTER_SIZES;
+    const numFilters = filterSizes ? filterSizes.length : void 0;
+    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {
+      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);
+    }
+    return extractParams7(weights, this.config, this.boxEncodingSize, filterSizes);
+  }
+  async extractBoxes(outputTensor, inputBlobDimensions, scoreThreshold) {
+    const { width, height } = inputBlobDimensions;
+    const inputSize = Math.max(width, height);
+    const correctionFactorX = inputSize / width;
+    const correctionFactorY = inputSize / height;
+    const numCells = outputTensor.shape[1];
+    const numBoxes = this.config.anchors.length;
+    const [boxesTensor, scoresTensor, classScoresTensor] = tf39.tidy(() => {
+      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);
+      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);
+      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);
+      const classScores = this.withClassScores ? tf39.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3) : tf39.scalar(0);
+      return [boxes, scores, classScores];
+    });
+    const results = [];
+    const scoresData = await scoresTensor.array();
+    const boxesData = await boxesTensor.array();
+    for (let row = 0; row < numCells; row++) {
+      for (let col = 0; col < numCells; col++) {
+        for (let anchor = 0; anchor < numBoxes; anchor++) {
+          const score = sigmoid(scoresData[row][col][anchor][0]);
+          if (!scoreThreshold || score > scoreThreshold) {
+            const ctX = (col + sigmoid(boxesData[row][col][anchor][0])) / numCells * correctionFactorX;
+            const ctY = (row + sigmoid(boxesData[row][col][anchor][1])) / numCells * correctionFactorY;
+            const widthLocal = Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x / numCells * correctionFactorX;
+            const heightLocal = Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y / numCells * correctionFactorY;
+            const x = ctX - widthLocal / 2;
+            const y = ctY - heightLocal / 2;
+            const pos = { row, col, anchor };
+            const { classScore, label } = this.withClassScores ? await this.extractPredictedClass(classScoresTensor, pos) : { classScore: 1, label: 0 };
+            results.push({
+              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),
+              score,
+              classScore: score * classScore,
+              label,
+              ...pos
+            });
+          }
+        }
+      }
+    }
+    boxesTensor.dispose();
+    scoresTensor.dispose();
+    classScoresTensor.dispose();
+    return results;
+  }
+  async extractPredictedClass(classesTensor, pos) {
+    const { row, col, anchor } = pos;
+    const classesData = await classesTensor.array();
+    return Array(this.config.classes.length).fill(0).map((_, i) => classesData[row][col][anchor][i]).map((classScore, label) => ({
+      classScore,
+      label
+    })).reduce((max, curr) => max.classScore > curr.classScore ? max : curr);
+  }
+};
+var TinyYolov2Base = _TinyYolov2Base;
+TinyYolov2Base.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+
+// src/tinyYolov2/TinyYolov2.ts
+var TinyYolov2 = class extends TinyYolov2Base {
+  constructor(withSeparableConvs = true) {
+    const config = {
+      withSeparableConvs,
+      iouThreshold: IOU_THRESHOLD,
+      classes: ["face"],
+      ...withSeparableConvs ? {
+        anchors: BOX_ANCHORS_SEPARABLE,
+        meanRgb: MEAN_RGB_SEPARABLE
+      } : {
+        anchors: BOX_ANCHORS,
+        withClassScores: true
+      }
+    };
+    super(config);
+  }
+  get withSeparableConvs() {
+    return this.config.withSeparableConvs;
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/tinyYolov2/index.ts
+function createTinyYolov2(weights, withSeparableConvs = true) {
+  const net = new TinyYolov2(withSeparableConvs);
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/tinyFaceDetector/TinyFaceDetectorOptions.ts
+var TinyFaceDetectorOptions = class extends TinyYolov2Options {
+  constructor() {
+    super(...arguments);
+    this._name = "TinyFaceDetectorOptions";
+  }
+};
+
+// src/globalApi/ComposableTask.ts
+var ComposableTask = class {
+  async then(onfulfilled) {
+    return onfulfilled(await this.run());
+  }
+  async run() {
+    throw new Error("ComposableTask - run is not implemented");
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var tf41 = __toESM(require_tfjs_esm());
+
+// src/globalApi/extractFacesAndComputeResults.ts
+var tf40 = __toESM(require_tfjs_esm());
+async function extractAllFacesAndComputeResults(parentResults, input, computeResults, extractedFaces, getRectForAlignment = ({ alignedRect }) => alignedRect) {
+  const faceBoxes = parentResults.map((parentResult) => isWithFaceLandmarks(parentResult) ? getRectForAlignment(parentResult) : parentResult.detection);
+  const faces = extractedFaces || (input instanceof tf40.Tensor ? await extractFaceTensors(input, faceBoxes) : await extractFaces(input, faceBoxes));
+  const results = await computeResults(faces);
+  faces.forEach((f) => f instanceof tf40.Tensor && f.dispose());
+  return results;
+}
+async function extractSingleFaceAndComputeResult(parentResult, input, computeResult, extractedFaces, getRectForAlignment) {
+  return extractAllFacesAndComputeResults(
+    [parentResult],
+    input,
+    async (faces) => computeResult(faces[0]),
+    extractedFaces,
+    getRectForAlignment
+  );
+}
+
+// src/tinyFaceDetector/const.ts
+var IOU_THRESHOLD2 = 0.4;
+var BOX_ANCHORS2 = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB = [117.001, 114.697, 97.404];
+
+// src/tinyFaceDetector/TinyFaceDetector.ts
+var TinyFaceDetector = class extends TinyYolov2Base {
+  constructor() {
+    const config = {
+      withSeparableConvs: true,
+      iouThreshold: IOU_THRESHOLD2,
+      classes: ["face"],
+      anchors: BOX_ANCHORS2,
+      meanRgb: MEAN_RGB,
+      isFirstLayerConv2d: true,
+      filterSizes: [3, 16, 32, 64, 128, 256, 512]
+    };
+    super(config);
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return "tiny_face_detector_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/globalApi/nets.ts
+var nets = {
+  ssdMobilenetv1: new SsdMobilenetv1(),
+  tinyFaceDetector: new TinyFaceDetector(),
+  tinyYolov2: new TinyYolov2(),
+  faceLandmark68Net: new FaceLandmark68Net(),
+  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),
+  faceRecognitionNet: new FaceRecognitionNet(),
+  faceExpressionNet: new FaceExpressionNet(),
+  ageGenderNet: new AgeGenderNet()
+};
+var ssdMobilenetv1 = (input, options) => nets.ssdMobilenetv1.locateFaces(input, options);
+var tinyFaceDetector = (input, options) => nets.tinyFaceDetector.locateFaces(input, options);
+var tinyYolov2 = (input, options) => nets.tinyYolov2.locateFaces(input, options);
+var detectFaceLandmarks = (input) => nets.faceLandmark68Net.detectLandmarks(input);
+var detectFaceLandmarksTiny = (input) => nets.faceLandmark68TinyNet.detectLandmarks(input);
+var computeFaceDescriptor = (input) => nets.faceRecognitionNet.computeFaceDescriptor(input);
+var recognizeFaceExpressions = (input) => nets.faceExpressionNet.predictExpressions(input);
+var predictAgeAndGender = (input) => nets.ageGenderNet.predictAgeAndGender(input);
+var loadSsdMobilenetv1Model = (url) => nets.ssdMobilenetv1.load(url);
+var loadTinyFaceDetectorModel = (url) => nets.tinyFaceDetector.load(url);
+var loadTinyYolov2Model = (url) => nets.tinyYolov2.load(url);
+var loadFaceLandmarkModel = (url) => nets.faceLandmark68Net.load(url);
+var loadFaceLandmarkTinyModel = (url) => nets.faceLandmark68TinyNet.load(url);
+var loadFaceRecognitionModel = (url) => nets.faceRecognitionNet.load(url);
+var loadFaceExpressionModel = (url) => nets.faceExpressionNet.load(url);
+var loadAgeGenderModel = (url) => nets.ageGenderNet.load(url);
+var loadFaceDetectionModel = loadSsdMobilenetv1Model;
+var locateFaces = ssdMobilenetv1;
+var detectLandmarks = detectFaceLandmarks;
+
+// src/globalApi/PredictFaceExpressionsTask.ts
+var PredictFaceExpressionsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const faceExpressionsByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(
+        faces.map((face) => nets.faceExpressionNet.predictExpressions(face))
+      ),
+      this.extractedFaces
+    );
+    return parentResults.map(
+      (parentResult, i) => extendWithFaceExpressions(parentResult, faceExpressionsByFace[i])
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const faceExpressions = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceExpressionNet.predictExpressions(face),
+      this.extractedFaces
+    );
+    return extendWithFaceExpressions(parentResult, faceExpressions);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictAllFaceExpressionsWithFaceAlignmentTask = class extends PredictAllFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsWithFaceAlignmentTask = class extends PredictSingleFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/PredictAgeAndGenderTask.ts
+var PredictAgeAndGenderTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const ageAndGenderByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face))),
+      this.extractedFaces
+    );
+    return parentResults.map((parentResult, i) => {
+      const { age, gender, genderProbability } = ageAndGenderByFace[i];
+      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+    });
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.ageGenderNet.predictAgeAndGender(face),
+      this.extractedFaces
+    );
+    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictAllAgeAndGenderWithFaceAlignmentTask = class extends PredictAllAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderWithFaceAlignmentTask = class extends PredictSingleAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/ComputeFaceDescriptorsTasks.ts
+var ComputeFaceDescriptorsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+  }
+};
+var ComputeAllFaceDescriptorsTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const descriptors = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face))),
+      null,
+      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return descriptors.map((descriptor, i) => extendWithFaceDescriptor(parentResults[i], descriptor));
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+var ComputeSingleFaceDescriptorTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const descriptor = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face),
+      null,
+      (parentResult2) => parentResult2.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return extendWithFaceDescriptor(parentResult, descriptor);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var DetectFaceLandmarksTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, useTinyLandmarkNet) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.useTinyLandmarkNet = useTinyLandmarkNet;
+  }
+  get landmarkNet() {
+    return this.useTinyLandmarkNet ? nets.faceLandmark68TinyNet : nets.faceLandmark68Net;
+  }
+};
+var DetectAllFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const detections = parentResults.map((res) => res.detection);
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, detections) : await extractFaces(this.input, detections);
+    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face)));
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    const result = parentResults.filter((_parentResult, i) => faceLandmarksByFace[i]).map((parentResult, i) => extendWithFaceLandmarks(parentResult, faceLandmarksByFace[i]));
+    return result;
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var DetectSingleFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const { detection } = parentResult;
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, [detection]) : await extractFaces(this.input, [detection]);
+    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]);
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    return extendWithFaceLandmarks(parentResult, landmarks);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFacesTasks.ts
+var DetectFacesTaskBase = class extends ComposableTask {
+  constructor(input, options = new SsdMobilenetv1Options()) {
+    super();
+    this.input = input;
+    this.options = options;
+  }
+};
+var DetectAllFacesTask = class extends DetectFacesTaskBase {
+  async run() {
+    const { input, options } = this;
+    let result;
+    if (options instanceof TinyFaceDetectorOptions)
+      result = nets.tinyFaceDetector.locateFaces(input, options);
+    else if (options instanceof SsdMobilenetv1Options)
+      result = nets.ssdMobilenetv1.locateFaces(input, options);
+    else if (options instanceof TinyYolov2Options)
+      result = nets.tinyYolov2.locateFaces(input, options);
+    else
+      throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+    return result;
+  }
+  runAndExtendWithFaceDetections() {
+    return new Promise((resolve, reject) => {
+      this.run().then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection)))).catch((err) => reject(err));
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectAllFaceLandmarksTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+};
+var DetectSingleFaceTask = class extends DetectFacesTaskBase {
+  async run() {
+    const faceDetections = await new DetectAllFacesTask(this.input, this.options);
+    let faceDetectionWithHighestScore = faceDetections[0];
+    faceDetections.forEach((faceDetection) => {
+      if (faceDetection.score > faceDetectionWithHighestScore.score)
+        faceDetectionWithHighestScore = faceDetection;
+    });
+    return faceDetectionWithHighestScore;
+  }
+  runAndExtendWithFaceDetection() {
+    return new Promise(async (resolve) => {
+      const detection = await this.run();
+      resolve(detection ? extendWithFaceDetection({}, detection) : void 0);
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectSingleFaceLandmarksTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+};
+
+// src/globalApi/detectFaces.ts
+function detectSingleFace(input, options = new SsdMobilenetv1Options()) {
+  return new DetectSingleFaceTask(input, options);
+}
+function detectAllFaces(input, options = new SsdMobilenetv1Options()) {
+  return new DetectAllFacesTask(input, options);
+}
+
+// src/globalApi/allFaces.ts
+async function allFacesSsdMobilenetv1(input, minConfidence) {
+  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {})).withFaceLandmarks().withFaceDescriptors();
+}
+async function allFacesTinyYolov2(input, forwardParams = {}) {
+  return detectAllFaces(input, new TinyYolov2Options(forwardParams)).withFaceLandmarks().withFaceDescriptors();
+}
+var allFaces = allFacesSsdMobilenetv1;
+
+// src/euclideanDistance.ts
+function euclideanDistance(arr1, arr2) {
+  if (arr1.length !== arr2.length)
+    throw new Error("euclideanDistance: arr1.length !== arr2.length");
+  const desc1 = Array.from(arr1);
+  const desc2 = Array.from(arr2);
+  return Math.sqrt(
+    desc1.map((val, i) => val - desc2[i]).reduce((res, diff) => res + diff * diff, 0)
+  );
+}
+
+// src/globalApi/FaceMatcher.ts
+var FaceMatcher = class {
+  constructor(inputs, distanceThreshold = 0.6) {
+    this._distanceThreshold = distanceThreshold;
+    const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+    if (!inputArray.length)
+      throw new Error("FaceRecognizer.constructor - expected atleast one input");
+    let count = 1;
+    const createUniqueLabel = () => `person ${count++}`;
+    this._labeledDescriptors = inputArray.map((desc) => {
+      if (desc instanceof LabeledFaceDescriptors)
+        return desc;
+      if (desc instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);
+      if (desc.descriptor && desc.descriptor instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);
+      throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>");
+    });
+  }
+  get labeledDescriptors() {
+    return this._labeledDescriptors;
+  }
+  get distanceThreshold() {
+    return this._distanceThreshold;
+  }
+  computeMeanDistance(queryDescriptor, descriptors) {
+    return descriptors.map((d) => euclideanDistance(d, queryDescriptor)).reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);
+  }
+  matchDescriptor(queryDescriptor) {
+    return this.labeledDescriptors.map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors))).reduce((best, curr) => best.distance < curr.distance ? best : curr);
+  }
+  findBestMatch(queryDescriptor) {
+    const bestMatch = this.matchDescriptor(queryDescriptor);
+    return bestMatch.distance < this._distanceThreshold ? bestMatch : new FaceMatch("unknown", bestMatch.distance);
+  }
+  toJSON() {
+    return {
+      distanceThreshold: this._distanceThreshold,
+      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON())
+    };
+  }
+  static fromJSON(json) {
+    const labeledDescriptors = json.labeledDescriptors.map((ld) => LabeledFaceDescriptors.fromJSON(ld));
+    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);
+  }
+};
+
+// src/tinyFaceDetector/index.ts
+function createTinyFaceDetector(weights) {
+  const net = new TinyFaceDetector();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/resizeResults.ts
+function resizeResults(results, dimensions) {
+  const { width, height } = new Dimensions(dimensions.width, dimensions.height);
+  if (width <= 0 || height <= 0) {
+    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);
+  }
+  if (Array.isArray(results)) {
+    return results.map((obj) => resizeResults(obj, { width, height }));
+  }
+  if (isWithFaceLandmarks(results)) {
+    const resizedDetection = results.detection.forSize(width, height);
+    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);
+    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);
+  }
+  if (isWithFaceDetection(results)) {
+    return extendWithFaceDetection(results, results.detection.forSize(width, height));
+  }
+  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {
+    return results.forSize(width, height);
+  }
+  return results;
+}
+
+// src/index.ts
+var version2 = version;
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  AgeGenderNet,
+  BoundingBox,
+  Box,
+  ComposableTask,
+  ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask,
+  DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask,
+  Dimensions,
+  FACE_EXPRESSION_LABELS,
+  FaceDetection,
+  FaceDetectionNet,
+  FaceExpressionNet,
+  FaceExpressions,
+  FaceLandmark68Net,
+  FaceLandmark68TinyNet,
+  FaceLandmarkNet,
+  FaceLandmarks,
+  FaceLandmarks5,
+  FaceLandmarks68,
+  FaceMatch,
+  FaceMatcher,
+  FaceRecognitionNet,
+  Gender,
+  LabeledBox,
+  LabeledFaceDescriptors,
+  NetInput,
+  NeuralNetwork,
+  ObjectDetection,
+  Point,
+  PredictedBox,
+  Rect,
+  SsdMobilenetv1,
+  SsdMobilenetv1Options,
+  TinyFaceDetector,
+  TinyFaceDetectorOptions,
+  TinyYolov2,
+  TinyYolov2Options,
+  allFaces,
+  allFacesSsdMobilenetv1,
+  allFacesTinyYolov2,
+  awaitMediaLoaded,
+  bufferToImage,
+  computeFaceDescriptor,
+  createCanvas,
+  createCanvasFromMedia,
+  createFaceDetectionNet,
+  createFaceRecognitionNet,
+  createSsdMobilenetv1,
+  createTinyFaceDetector,
+  createTinyYolov2,
+  detectAllFaces,
+  detectFaceLandmarks,
+  detectFaceLandmarksTiny,
+  detectLandmarks,
+  detectSingleFace,
+  draw,
+  env,
+  euclideanDistance,
+  extendWithAge,
+  extendWithFaceDescriptor,
+  extendWithFaceDetection,
+  extendWithFaceExpressions,
+  extendWithFaceLandmarks,
+  extendWithGender,
+  extractFaceTensors,
+  extractFaces,
+  fetchImage,
+  fetchJson,
+  fetchNetWeights,
+  fetchOrThrow,
+  fetchVideo,
+  getContext2dOrThrow,
+  getMediaDimensions,
+  imageTensorToCanvas,
+  imageToSquare,
+  inverseSigmoid,
+  iou,
+  isMediaElement,
+  isMediaLoaded,
+  isWithAge,
+  isWithFaceDetection,
+  isWithFaceExpressions,
+  isWithFaceLandmarks,
+  isWithGender,
+  loadAgeGenderModel,
+  loadFaceDetectionModel,
+  loadFaceExpressionModel,
+  loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel,
+  loadTinyYolov2Model,
+  loadWeightMap,
+  locateFaces,
+  matchDimensions,
+  minBbox,
+  nets,
+  nonMaxSuppression,
+  normalize,
+  padToSquare,
+  predictAgeAndGender,
+  recognizeFaceExpressions,
+  resizeResults,
+  resolveInput,
+  shuffleArray,
+  sigmoid,
+  ssdMobilenetv1,
+  tf,
+  tinyFaceDetector,
+  tinyYolov2,
+  toNetInput,
+  utils,
+  validateConfig,
+  version
+});
diff --git a/dist/face-api.node-wasm.js b/dist/face-api.node-wasm.js
index e32a71b..fbc2b75 100644
--- a/dist/face-api.node-wasm.js
+++ b/dist/face-api.node-wasm.js
@@ -4,4 +4,4927 @@
   author: <https://github.com/vladmandic>'
 */
 
-"use strict";var hn=Object.create;var er=Object.defineProperty;var bn=Object.getOwnPropertyDescriptor;var gn=Object.getOwnPropertyNames;var xn=Object.getPrototypeOf,vn=Object.prototype.hasOwnProperty;var yn=(o,t)=>()=>(t||o((t={exports:{}}).exports,t),t.exports),Wr=(o,t)=>{for(var e in t)er(o,e,{get:t[e],enumerable:!0})},To=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of gn(t))!vn.call(o,n)&&n!==e&&er(o,n,{get:()=>t[n],enumerable:!(r=bn(t,n))||r.enumerable});return o};var v=(o,t,e)=>(e=o!=null?hn(xn(o)):{},To(t||!o||!o.__esModule?er(e,"default",{value:o,enumerable:!0}):e,o)),_n=o=>To(er({},"__esModule",{value:!0}),o);var x=yn((Va,or)=>{"use strict";var Br=Object.defineProperty,Tn=Object.getOwnPropertyDescriptor,wn=Object.getOwnPropertyNames,Pn=Object.prototype.hasOwnProperty,Fn=(o,t)=>{for(var e in t)Br(o,e,{get:t[e],enumerable:!0})},kr=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of wn(t))!Pn.call(o,n)&&n!==e&&Br(o,n,{get:()=>t[n],enumerable:!(r=Tn(t,n))||r.enumerable});return o},Po=(o,t,e)=>(kr(o,t,"default"),e&&kr(e,t,"default")),Dn=o=>kr(Br({},"__esModule",{value:!0}),o),rr={};Fn(rr,{version:()=>Nn});or.exports=Dn(rr);Po(rr,require("@tensorflow/tfjs"),or.exports);Po(rr,require("@tensorflow/tfjs-backend-wasm"),or.exports);var wo="4.0.0",En="4.0.0",Mn="4.0.0",Cn="4.0.0",In="4.0.0",Nn={tfjs:wo,"tfjs-core":wo,"tfjs-converter":En,"tfjs-backend-cpu":Mn,"tfjs-backend-webgl":Cn,"tfjs-backend-wasm":In}});var Ra={};Wr(Ra,{AgeGenderNet:()=>He,BoundingBox:()=>Vt,Box:()=>D,ComposableTask:()=>J,ComputeAllFaceDescriptorsTask:()=>Pt,ComputeFaceDescriptorsTaskBase:()=>Ue,ComputeSingleFaceDescriptorTask:()=>Ft,DetectAllFaceLandmarksTask:()=>qe,DetectAllFacesTask:()=>Ie,DetectFaceLandmarksTaskBase:()=>Je,DetectFacesTaskBase:()=>Ke,DetectSingleFaceLandmarksTask:()=>Ze,DetectSingleFaceTask:()=>Qe,Dimensions:()=>k,FACE_EXPRESSION_LABELS:()=>so,FaceDetection:()=>M,FaceDetectionNet:()=>lo,FaceExpressionNet:()=>Oe,FaceExpressions:()=>yt,FaceLandmark68Net:()=>Kt,FaceLandmark68TinyNet:()=>ze,FaceLandmarkNet:()=>po,FaceLandmarks:()=>z,FaceLandmarks5:()=>jr,FaceLandmarks68:()=>Gt,FaceMatch:()=>pe,FaceMatcher:()=>tr,FaceRecognitionNet:()=>Qt,Gender:()=>Pr,LabeledBox:()=>ue,LabeledFaceDescriptors:()=>mt,NetInput:()=>ut,NeuralNetwork:()=>A,ObjectDetection:()=>bt,Point:()=>g,PredictedBox:()=>Ur,Rect:()=>Yt,SsdMobilenetv1:()=>St,SsdMobilenetv1Options:()=>X,TinyFaceDetector:()=>ne,TinyFaceDetectorOptions:()=>je,TinyYolov2:()=>re,TinyYolov2Options:()=>st,allFaces:()=>Aa,allFacesSsdMobilenetv1:()=>ln,allFacesTinyYolov2:()=>La,awaitMediaLoaded:()=>to,bufferToImage:()=>eo,computeFaceDescriptor:()=>va,createCanvas:()=>Jt,createCanvasFromMedia:()=>We,createFaceDetectionNet:()=>fa,createFaceRecognitionNet:()=>ea,createSsdMobilenetv1:()=>Zo,createTinyFaceDetector:()=>Wa,createTinyYolov2:()=>ha,detectAllFaces:()=>Ar,detectFaceLandmarks:()=>un,detectFaceLandmarksTiny:()=>xa,detectLandmarks:()=>Na,detectSingleFace:()=>Sa,draw:()=>co,env:()=>w,euclideanDistance:()=>vo,extendWithAge:()=>Cr,extendWithFaceDescriptor:()=>Mr,extendWithFaceDetection:()=>jt,extendWithFaceExpressions:()=>yr,extendWithFaceLandmarks:()=>Pe,extendWithGender:()=>Ir,extractFaceTensors:()=>de,extractFaces:()=>le,fetchImage:()=>Hn,fetchJson:()=>no,fetchNetWeights:()=>zn,fetchOrThrow:()=>xt,fetchVideo:()=>Vn,getContext2dOrThrow:()=>O,getMediaDimensions:()=>Xt,imageTensorToCanvas:()=>ro,imageToSquare:()=>oo,inverseSigmoid:()=>An,iou:()=>zr,isMediaElement:()=>mr,isMediaLoaded:()=>Ae,isWithAge:()=>ra,isWithFaceDetection:()=>pt,isWithFaceExpressions:()=>io,isWithFaceLandmarks:()=>Zt,isWithGender:()=>oa,loadAgeGenderModel:()=>Ma,loadFaceDetectionModel:()=>Ca,loadFaceExpressionModel:()=>Ea,loadFaceLandmarkModel:()=>Pa,loadFaceLandmarkTinyModel:()=>Fa,loadFaceRecognitionModel:()=>Da,loadSsdMobilenetv1Model:()=>fn,loadTinyFaceDetectorModel:()=>Ta,loadTinyYolov2Model:()=>wa,loadWeightMap:()=>ao,locateFaces:()=>Ia,matchDimensions:()=>Yn,minBbox:()=>Vr,nets:()=>P,nonMaxSuppression:()=>Yr,normalize:()=>rt,padToSquare:()=>Gr,predictAgeAndGender:()=>_a,recognizeFaceExpressions:()=>ya,resizeResults:()=>dn,resolveInput:()=>Ut,shuffleArray:()=>Ln,sigmoid:()=>Ne,ssdMobilenetv1:()=>pn,tf:()=>ka,tinyFaceDetector:()=>ba,tinyYolov2:()=>ga,toNetInput:()=>C,utils:()=>Hr,validateConfig:()=>ho,version:()=>Ba});module.exports=_n(Ra);var ka=v(x());var co={};Wr(co,{AnchorPosition:()=>Qr,DrawBox:()=>Le,DrawBoxOptions:()=>ir,DrawFaceLandmarks:()=>Tr,DrawFaceLandmarksOptions:()=>_r,DrawTextField:()=>gt,DrawTextFieldOptions:()=>fe,drawContour:()=>dt,drawDetections:()=>On,drawFaceExpressions:()=>Gn,drawFaceLandmarks:()=>Un});function dt(o,t,e=!1){if(o.beginPath(),t.slice(1).forEach(({x:r,y:n},a)=>{let s=t[a];o.moveTo(s.x,s.y),o.lineTo(r,n)}),e){let r=t[t.length-1],n=t[0];if(!r||!n)return;o.moveTo(r.x,r.y),o.lineTo(n.x,n.y)}o.stroke()}var Hr={};Wr(Hr,{computeReshapedDimensions:()=>Or,getCenterPoint:()=>zt,isDimensions:()=>ar,isEven:()=>nr,isFloat:()=>$r,isTensor:()=>Ot,isTensor1D:()=>Sn,isTensor2D:()=>Rr,isTensor3D:()=>ht,isTensor4D:()=>U,isValidNumber:()=>et,isValidProbablitiy:()=>me,range:()=>it,round:()=>Ht});var Fo=v(x());var k=class{constructor(t,e){if(!et(t)||!et(e))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:e})}`);this._width=t,this._height=e}get width(){return this._width}get height(){return this._height}reverse(){return new k(1/this.width,1/this.height)}};function Ot(o,t){return o instanceof Fo.Tensor&&o.shape.length===t}function Sn(o){return Ot(o,1)}function Rr(o){return Ot(o,2)}function ht(o){return Ot(o,3)}function U(o){return Ot(o,4)}function $r(o){return o%1!==0}function nr(o){return o%2===0}function Ht(o,t=2){let e=10**t;return Math.floor(o*e)/e}function ar(o){return o&&o.width&&o.height}function Or({width:o,height:t},e){let r=e/Math.max(t,o);return new k(Math.round(o*r),Math.round(t*r))}function zt(o){return o.reduce((t,e)=>t.add(e),new g(0,0)).div(new g(o.length,o.length))}function it(o,t,e){return Array(o).fill(0).map((r,n)=>t+n*e)}function et(o){return!!o&&o!==1/0&&o!==-1/0&&!Number.isNaN(o)||o===0}function me(o){return et(o)&&o>=0&&o<=1}var g=class{constructor(t,e){this._x=t,this._y=e}get x(){return this._x}get y(){return this._y}add(t){return new g(this.x+t.x,this.y+t.y)}sub(t){return new g(this.x-t.x,this.y-t.y)}mul(t){return new g(this.x*t.x,this.y*t.y)}div(t){return new g(this.x/t.x,this.y/t.y)}abs(){return new g(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new g(Math.floor(this.x),Math.floor(this.y))}};var D=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(et)}static assertIsValidBox(t,e,r=!1){if(!D.isRect(t))throw new Error(`${e} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${e} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,e=!0){let r=t||{},n=[r.left,r.top,r.right,r.bottom].every(et),a=[r.x,r.y,r.width,r.height].every(et);if(!a&&!n)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[s,i,c,m]=a?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];D.assertIsValidBox({x:s,y:i,width:c,height:m},"Box.constructor",e),this._x=s,this._y=i,this._width=c,this._height=m}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new g(this.left,this.top)}get topRight(){return new g(this.right,this.top)}get bottomLeft(){return new g(this.left,this.bottom)}get bottomRight(){return new g(this.right,this.bottom)}round(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.round(a));return new D({x:t,y:e,width:r,height:n})}floor(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.floor(a));return new D({x:t,y:e,width:r,height:n})}toSquare(){let{x:t,y:e,width:r,height:n}=this,a=Math.abs(r-n);return r<n&&(t-=a/2,r+=a),n<r&&(e-=a/2,n+=a),new D({x:t,y:e,width:r,height:n})}rescale(t){let e=ar(t)?t.width:t,r=ar(t)?t.height:t;return new D({x:this.x*e,y:this.y*r,width:this.width*e,height:this.height*r})}pad(t,e){let[r,n,a,s]=[this.x-t/2,this.y-e/2,this.width+t,this.height+e];return new D({x:r,y:n,width:a,height:s})}clipAtImageBorders(t,e){let{x:r,y:n,right:a,bottom:s}=this,i=Math.max(r,0),c=Math.max(n,0),m=a-i,p=s-c,u=Math.min(m,t-i),f=Math.min(p,e-c);return new D({x:i,y:c,width:u,height:f}).floor()}shift(t,e){let{width:r,height:n}=this,a=this.x+t,s=this.y+e;return new D({x:a,y:s,width:r,height:n})}padAtBorders(t,e){let r=this.width+1,n=this.height+1,a=1,s=1,i=r,c=n,m=this.left,p=this.top,u=this.right,f=this.bottom;return u>e&&(i=-u+e+r,u=e),f>t&&(c=-f+t+n,f=t),m<1&&(c=2-m,m=1),p<1&&(c=2-p,p=1),{dy:s,edy:c,dx:a,edx:i,y:p,ey:f,x:m,ex:u,w:r,h:n}}calibrate(t){return new D({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var Vt=class extends D{constructor(t,e,r,n,a=!1){super({left:t,top:e,right:r,bottom:n},a)}};var bt=class{constructor(t,e,r,n,a){this._imageDims=new k(a.width,a.height),this._score=t,this._classScore=e,this._className=r,this._box=new D(n).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new D(this._box).rescale(this.imageDims.reverse())}forSize(t,e){return new bt(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:e})}};var M=class extends bt{constructor(t,e,r){super(t,t,"",e,r)}forSize(t,e){let{score:r,relativeBox:n,imageDims:a}=super.forSize(t,e);return new M(r,n,a)}};function zr(o,t,e=!0){let r=Math.max(0,Math.min(o.right,t.right)-Math.max(o.left,t.left)),n=Math.max(0,Math.min(o.bottom,t.bottom)-Math.max(o.top,t.top)),a=r*n;return e?a/(o.area+t.area-a):a/Math.min(o.area,t.area)}function Vr(o){let t=o.map(i=>i.x),e=o.map(i=>i.y),r=t.reduce((i,c)=>c<i?c:i,1/0),n=e.reduce((i,c)=>c<i?c:i,1/0),a=t.reduce((i,c)=>i<c?c:i,0),s=e.reduce((i,c)=>i<c?c:i,0);return new Vt(r,n,a,s)}function Yr(o,t,e,r=!0){let n=t.map((s,i)=>({score:s,boxIndex:i})).sort((s,i)=>s.score-i.score).map(s=>s.boxIndex),a=[];for(;n.length>0;){let s=n.pop();a.push(s);let i=n,c=[];for(let m=0;m<i.length;m++){let p=i[m],u=o[s],f=o[p];c.push(zr(u,f,r))}n=n.filter((m,p)=>c[p]<=e)}return a}var ct=v(x());function rt(o,t){return ct.tidy(()=>{let[e,r,n]=t,a=ct.fill([...o.shape.slice(0,3),1],e,"float32"),s=ct.fill([...o.shape.slice(0,3),1],r,"float32"),i=ct.fill([...o.shape.slice(0,3),1],n,"float32"),c=ct.concat([a,s,i],3);return ct.sub(o,c)})}var Ct=v(x());function Gr(o,t=!1){return Ct.tidy(()=>{let[e,r]=o.shape.slice(1);if(e===r)return o;let n=Math.abs(e-r),a=Math.round(n*(t?.5:1)),s=e>r?2:1,i=f=>{let l=o.shape.slice();return l[s]=f,Ct.fill(l,0,"float32")},c=i(a),m=n-c.shape[s],u=[t&&m?i(m):null,o,c].filter(f=>!!f).map(f=>Ct.cast(f,"float32"));return Ct.concat(u,s)})}function Ln(o){let t=o.slice();for(let e=t.length-1;e>0;e--){let r=Math.floor(Math.random()*(e+1)),n=t[e];t[e]=t[r],t[r]=n}return t}function Ne(o){return 1/(1+Math.exp(-o))}function An(o){return Math.log(o/(1-o))}var Yt=class extends D{constructor(t,e,r,n,a=!1){super({x:t,y:e,width:r,height:n},a)}};var Wn=.5,kn=.43,Bn=.45,z=class{constructor(t,e,r=new g(0,0)){let{width:n,height:a}=e;this._imgDims=new k(n,a),this._shift=r,this._positions=t.map(s=>s.mul(new g(n,a)).add(r))}get shift(){return new g(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new g(this.imageWidth,this.imageHeight)))}forSize(t,e){return new this.constructor(this.relativePositions,{width:t,height:e})}shiftBy(t,e){return new this.constructor(this.relativePositions,this._imgDims,new g(t,e))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,e={}){if(t){let a=t instanceof M?t.box.floor():new D(t);return this.shiftBy(a.x,a.y).align(null,e)}let{useDlibAlignment:r,minBoxPadding:n}={useDlibAlignment:!1,minBoxPadding:.2,...e};return r?this.alignDlib():this.alignMinBbox(n)}alignDlib(){let t=this.getRefPointsForAlignment(),[e,r,n]=t,a=u=>n.sub(u).magnitude(),s=(a(e)+a(r))/2,i=Math.floor(s/Bn),c=zt(t),m=Math.floor(Math.max(0,c.x-Wn*i)),p=Math.floor(Math.max(0,c.y-kn*i));return new Yt(m,p,Math.min(i,this.imageWidth+m),Math.min(i,this.imageHeight+p))}alignMinBbox(t){let e=Vr(this.positions);return e.pad(e.width*t,e.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var jr=class extends z{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],zt([t[3],t[4]])]}};var Gt=class extends z{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(zt)}};var pe=class{constructor(t,e){this._label=t,this._distance=e}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${Ht(this.distance)})`:""}`}};var ue=class extends D{constructor(e,r){super(e);this._label=r}static assertIsValidLabeledBox(e,r){if(D.assertIsValidBox(e,r),!et(e.label))throw new Error(`${r} - expected property label (${e.label}) to be a number`)}get label(){return this._label}};var mt=class{constructor(t,e){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(e)||e.some(r=>!(r instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=e}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let e=t.descriptors.map(r=>new Float32Array(r));return new mt(t.label,e)}};var Ur=class extends ue{constructor(e,r,n,a){super(e,r);this._score=n,this._classScore=a}static assertIsValidPredictedBox(e,r){if(ue.assertIsValidLabeledBox(e,r),!me(e.score)||!me(e.classScore))throw new Error(`${r} - expected properties score (${e.score}) and (${e.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function pt(o){return o.detection instanceof M}function jt(o,t){return{...o,...{detection:t}}}function Xr(){let o=window.fetch;if(!o)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:o,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function Se(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function sr(o){let t="";if(!o&&Se())try{o=require("fs")}catch(r){t=r.toString()}return{readFile:o?r=>new Promise((n,a)=>{o.readFile(r,(s,i)=>s?a(s):n(i))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function Jr(){let o=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,e=global.Video||global.HTMLVideoElement,r=()=>{if(o)return new o;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},n=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},a=()=>{if(e)return new e;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},s=global.fetch,i=sr();return{Canvas:o||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:r,createImageElement:n,createVideoElement:a,fetch:s,...i}}function qr(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var L;function Rn(){if(!L)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return L}function Zr(o){L=o}function Kr(){return qr()?Zr(Xr()):Se()?Zr(Jr()):null}function $n(o){if(L||Kr(),!L)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=L.Canvas,Image:e=L.Image}=o;L.Canvas=t,L.Image=e,L.createCanvasElement=o.createCanvasElement||(()=>new t),L.createImageElement=o.createImageElement||(()=>new e),L.ImageData=o.ImageData||L.ImageData,L.Video=o.Video||L.Video,L.fetch=o.fetch||L.fetch,L.readFile=o.readFile||L.readFile}var w={getEnv:Rn,setEnv:Zr,initialize:Kr,createBrowserEnv:Xr,createFileSystem:sr,createNodejsEnv:Jr,monkeyPatch:$n,isBrowser:qr,isNodejs:Se};Kr();function Ut(o){return!w.isNodejs()&&typeof o=="string"?document.getElementById(o):o}function O(o){let{Canvas:t,CanvasRenderingContext2D:e}=w.getEnv();if(o instanceof e)return o;let r=Ut(o);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let n=r.getContext("2d");if(!n)throw new Error("resolveContext2d - canvas 2d context is null");return n}var Qr=(n=>(n.TOP_LEFT="TOP_LEFT",n.TOP_RIGHT="TOP_RIGHT",n.BOTTOM_LEFT="BOTTOM_LEFT",n.BOTTOM_RIGHT="BOTTOM_RIGHT",n))(Qr||{}),fe=class{constructor(t={}){let{anchorPosition:e,backgroundColor:r,fontColor:n,fontSize:a,fontStyle:s,padding:i}=t;this.anchorPosition=e||"TOP_LEFT",this.backgroundColor=r||"rgba(0, 0, 0, 0.5)",this.fontColor=n||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=s||"Georgia",this.padding=i||4}},gt=class{constructor(t,e,r={}){this.text=typeof t=="string"?[t]:t instanceof gt?t.text:t,this.anchor=e,this.options=new fe(r)}measureWidth(t){let{padding:e}=this.options;return this.text.map(r=>t.measureText(r).width).reduce((r,n)=>r<n?n:r,0)+2*e}measureHeight(){let{fontSize:t,padding:e}=this.options;return this.text.length*t+2*e}getUpperLeft(t,e){let{anchorPosition:r}=this.options,n=r==="BOTTOM_RIGHT"||r==="TOP_RIGHT",a=r==="BOTTOM_LEFT"||r==="BOTTOM_RIGHT",s=this.measureWidth(t),i=this.measureHeight(),c=n?this.anchor.x-s:this.anchor.x,m=a?this.anchor.y-i:this.anchor.y;if(e){let{width:p,height:u}=e,f=Math.max(Math.min(c,p-s),0),l=Math.max(Math.min(m,u-i),0);return{x:f,y:l}}return{x:c,y:m}}draw(t){let e=Ut(t),r=O(e),{backgroundColor:n,fontColor:a,fontSize:s,fontStyle:i,padding:c}=this.options;r.font=`${s}px ${i}`;let m=this.measureWidth(r),p=this.measureHeight();r.fillStyle=n;let u=this.getUpperLeft(r,e);r.fillRect(u.x,u.y,m,p),r.fillStyle=a,this.text.forEach((f,l)=>{let b=c+u.x,y=c+u.y+(l+1)*s;r.fillText(f,b,y)})}};var ir=class{constructor(t={}){let{boxColor:e,lineWidth:r,label:n,drawLabelOptions:a}=t;this.boxColor=e||"rgba(0, 0, 255, 1)",this.lineWidth=r||2,this.label=n;let s={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new fe({...s,...a})}},Le=class{constructor(t,e={}){this.box=new D(t),this.options=new ir(e)}draw(t){let e=O(t),{boxColor:r,lineWidth:n}=this.options,{x:a,y:s,width:i,height:c}=this.box;e.strokeStyle=r,e.lineWidth=n,e.strokeRect(a,s,i,c);let{label:m}=this.options;m&&new gt([m],{x:a-n/2,y:s},this.options.drawLabelOptions).draw(t)}};function On(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof M?r.score:pt(r)?r.detection.score:void 0,a=r instanceof M?r.box:pt(r)?r.detection.box:new D(r),s=n?`${Ht(n)}`:void 0;new Le(a,{label:s}).draw(o)})}var we=v(x());function Ae(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t&&o.complete||o instanceof e&&o.readyState>=3}function to(o){return new Promise((t,e)=>{(o instanceof w.getEnv().Canvas||Ae(o))&&t(null);function r(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),e(a))}function n(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),t(a))}o.addEventListener("load",n),o.addEventListener("error",r)})}function eo(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToImage - expected buf to be of type: Blob"));let r=new FileReader;r.onload=()=>{typeof r.result!="string"&&e(new Error("bufferToImage - expected reader.result to be a string, in onload"));let n=w.getEnv().createImageElement();n.onload=()=>t(n),n.onerror=e,n.src=r.result},r.onerror=e,r.readAsDataURL(o)})}function Xt(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t?new k(o.naturalWidth,o.naturalHeight):o instanceof e?new k(o.videoWidth,o.videoHeight):new k(o.width,o.height)}function Jt({width:o,height:t}){let{createCanvasElement:e}=w.getEnv(),r=e();return r.width=o,r.height=t,r}function We(o,t){let{ImageData:e}=w.getEnv();if(!(o instanceof e)&&!Ae(o))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:n}=t||Xt(o),a=Jt({width:r,height:n});return o instanceof e?O(a).putImageData(o,0,0):O(a).drawImage(o,0,0,r,n),a}var cr=v(x());async function ro(o,t){let e=t||w.getEnv().createCanvasElement(),[r,n,a]=o.shape.slice(U(o)?1:0),s=cr.tidy(()=>o.as3D(r,n,a).toInt());return await cr.browser.toPixels(s,e),s.dispose(),e}function mr(o){let{Image:t,Canvas:e,Video:r}=w.getEnv();return o instanceof t||o instanceof e||o instanceof r}var V=v(x());function oo(o,t,e=!1){let{Image:r,Canvas:n}=w.getEnv();if(!(o instanceof r||o instanceof n))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Jt({width:1,height:1});let a=Xt(o),s=t/Math.max(a.height,a.width),i=s*a.width,c=s*a.height,m=Jt({width:t,height:t}),p=o instanceof n?o:We(o),u=Math.abs(i-c)/2,f=e&&i<c?u:0,l=e&&c<i?u:0;return p.width>0&&p.height>0&&O(m).drawImage(p,f,l,i,c),m}var ut=class{constructor(t,e=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=e,this._batchSize=t.length,t.forEach((r,n)=>{if(ht(r)){this._imageTensors[n]=r,this._inputDimensions[n]=r.shape;return}if(U(r)){let s=r.shape[0];if(s!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${s} passed, but not supported in input array`);this._imageTensors[n]=r,this._inputDimensions[n]=r.shape.slice(1);return}let a=r instanceof w.getEnv().Canvas?r:We(r);this._canvases[n]=a,this._inputDimensions[n]=[a.height,a.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return it(this.batchSize,0,1).map((t,e)=>this.getReshapedInputDimensions(e))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let e=this.getInputWidth(t),r=this.getInputHeight(t);return Or({width:e,height:r},this.inputSize)}toBatchTensor(t,e=!0){return this._inputSize=t,V.tidy(()=>{let r=it(this.batchSize,0,1).map(a=>{let s=this.getInput(a);if(s instanceof V.Tensor){let i=U(s)?s:V.expandDims(s);return i=Gr(i,e),(i.shape[1]!==t||i.shape[2]!==t)&&(i=V.image.resizeBilinear(i,[t,t],!1,!1)),i.as3D(t,t,3)}if(s instanceof w.getEnv().Canvas)return V.browser.fromPixels(oo(s,t,e));throw new Error(`toBatchTensor - at batchIdx ${a}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${s}`)});return V.stack(r.map(a=>V.cast(a,"float32"))).as4D(this.batchSize,t,t,3)})}};async function C(o){if(o instanceof ut)return o;let t=Array.isArray(o)?o:[o];if(!t.length)throw new Error("toNetInput - empty array passed as input");let e=n=>Array.isArray(o)?` at input index ${n}:`:"",r=t.map(Ut);return r.forEach((n,a)=>{if(!mr(n)&&!ht(n)&&!U(n))throw typeof t[a]=="string"?new Error(`toNetInput -${e(a)} string passed, but could not resolve HTMLElement for element id ${t[a]}`):new Error(`toNetInput -${e(a)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(U(n)){let s=n.shape[0];if(s!==1)throw new Error(`toNetInput -${e(a)} tf.Tensor4D with batchSize ${s} passed, but not supported in input array`)}}),await Promise.all(r.map(n=>mr(n)&&to(n))),new ut(r,Array.isArray(o))}async function le(o,t){let{Canvas:e}=w.getEnv(),r=o;if(!(o instanceof e)){let s=await C(o);if(s.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let i=s.getInput(0);r=i instanceof e?i:await ro(i)}let n=O(r);return t.map(s=>s instanceof M?s.forSize(r.width,r.height).box.floor():s).map(s=>s.clipAtImageBorders(r.width,r.height)).map(({x:s,y:i,width:c,height:m})=>{let p=Jt({width:c,height:m});return c>0&&m>0&&O(p).putImageData(n.getImageData(s,i,c,m),0,0),p})}var pr=v(x());async function de(o,t){if(!ht(o)&&!U(o))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(U(o)&&o.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return pr.tidy(()=>{let[e,r,n]=o.shape.slice(U(o)?1:0);return t.map(i=>i instanceof M?i.forSize(r,e).box:i).map(i=>i.clipAtImageBorders(r,e)).filter(i=>i.width>0&&i.height>0).map(({x:i,y:c,width:m,height:p})=>pr.slice3d(o.as3D(e,r,n),[c,i,0],[p,m,n]))})}async function xt(o,t){let{fetch:e}=w.getEnv(),r=await e(o,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}async function Hn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${e.type}, for url: ${t.url}`);return eo(e)}async function no(o){return(await xt(o)).json()}async function zn(o){return new Float32Array(await(await xt(o)).arrayBuffer())}function Do(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToVideo - expected buf to be of type: Blob"));let r=w.getEnv().createVideoElement();r.oncanplay=()=>t(r),r.onerror=e,r.playsInline=!0,r.muted=!0,r.src=URL.createObjectURL(o),r.play()})}async function Vn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${e.type}, for url: ${t.url}`);return Do(e)}var Eo=v(x());function ur(o,t){let e=`${t}-weights_manifest.json`;if(!o)return{modelBaseUri:"",manifestUri:e};if(o==="/")return{modelBaseUri:"/",manifestUri:`/${e}`};let r=o.startsWith("http://")?"http://":o.startsWith("https://")?"https://":"";o=o.replace(r,"");let n=o.split("/").filter(i=>i),a=o.endsWith(".json")?n[n.length-1]:e,s=r+(o.endsWith(".json")?n.slice(0,n.length-1):n).join("/");return s=o.startsWith("/")?`/${s}`:s,{modelBaseUri:s,manifestUri:s==="/"?`/${a}`:`${s}/${a}`}}async function ao(o,t){let{manifestUri:e,modelBaseUri:r}=ur(o,t),n=await no(e);return Eo.io.loadWeights(n,r)}function Yn(o,t,e=!1){let{width:r,height:n}=e?Xt(t):t;return o.width=r,o.height=n,{width:r,height:n}}var ye=v(x());var vt=v(x());var A=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:e,objProp:r}=this.traversePropertyPath(t);return e[r]}reassignParamFromPath(t,e){let{obj:r,objProp:n}=this.traversePropertyPath(t);r[n].dispose(),r[n]=e}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof vt.Variable)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof vt.Variable))}variable(){this.getFrozenParams().forEach(({path:t,tensor:e})=>{this.reassignParamFromPath(t,e.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:e})=>{let r=vt.tensor(e.dataSync());e.dispose(),this.reassignParamFromPath(t,r)})}dispose(t=!0){this.getParamList().forEach(e=>{if(t&&e.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${e.path}`);e.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,e)=>t.concat(e)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let e=await ao(t,this.getDefaultModelName());this.loadFromWeightMap(e)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:e}=w.getEnv(),{manifestUri:r,modelBaseUri:n}=ur(t,this.getDefaultModelName()),a=m=>Promise.all(m.map(p=>e(p).then(u=>u.buffer))),s=vt.io.weightsLoaderFactory(a),i=JSON.parse((await e(r)).toString()),c=await s(i,n);this.loadFromWeightMap(c)}loadFromWeightMap(t){let{paramMappings:e,params:r}=this.extractParamsFromWeightMap(t);this._paramMappings=e,this._params=r}extractWeights(t){let{paramMappings:e,params:r}=this.extractParams(t);this._paramMappings=e,this._params=r}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let e=t.split("/").reduce((a,s)=>{if(!a.nextObj.hasOwnProperty(s))throw new Error(`traversePropertyPath - object does not have property ${s}, for path ${t}`);return{obj:a.nextObj,objProp:s,nextObj:a.nextObj[s]}},{nextObj:this.params}),{obj:r,objProp:n}=e;if(!r||!n||!(r[n]instanceof vt.Tensor))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:r,objProp:n}}};var I=v(x());var he=v(x());function H(o,t,e){return he.tidy(()=>{let r=he.separableConv2d(o,t.depthwise_filter,t.pointwise_filter,e,"same");return r=he.add(r,t.bias),r})}function fr(o,t,e=!1){return I.tidy(()=>{let r=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,[2,2],"same"),t.conv0.bias):H(o,t.conv0,[2,2])),n=H(r,t.conv1,[1,1]),a=I.relu(I.add(r,n)),s=H(a,t.conv2,[1,1]);return I.relu(I.add(r,I.add(n,s)))})}function ke(o,t,e=!1,r=!0){return I.tidy(()=>{let n=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):H(o,t.conv0,r?[2,2]:[1,1])),a=H(n,t.conv1,[1,1]),s=I.relu(I.add(n,a)),i=H(s,t.conv2,[1,1]),c=I.relu(I.add(n,I.add(a,i))),m=H(c,t.conv3,[1,1]);return I.relu(I.add(n,I.add(a,I.add(i,m))))})}var It=v(x());function qt(o,t,e="same",r=!1){return It.tidy(()=>{let n=It.add(It.conv2d(o,t.filters,[1,1],e),t.bias);return r?It.relu(n):n})}function B(o,t){Object.keys(o).forEach(e=>{t.some(r=>r.originalPath===e)||o[e].dispose()})}var lr=v(x());function be(o,t){return(e,r,n,a)=>{let s=lr.tensor4d(o(e*r*n*n),[n,n,e,r]),i=lr.tensor1d(o(r));return t.push({paramPath:`${a}/filters`},{paramPath:`${a}/bias`}),{filters:s,bias:i}}}var dr=v(x());function hr(o,t){return(e,r,n)=>{let a=dr.tensor2d(o(e*r),[e,r]),s=dr.tensor1d(o(r));return t.push({paramPath:`${n}/weights`},{paramPath:`${n}/bias`}),{weights:a,bias:s}}}var Re=v(x());var Be=class{constructor(t,e,r){this.depthwise_filter=t;this.pointwise_filter=e;this.bias=r}};function ge(o,t){return(e,r,n)=>{let a=Re.tensor4d(o(9*e),[3,3,e,1]),s=Re.tensor4d(o(e*r),[1,1,e,r]),i=Re.tensor1d(o(r));return t.push({paramPath:`${n}/depthwise_filter`},{paramPath:`${n}/pointwise_filter`},{paramPath:`${n}/bias`}),new Be(a,s,i)}}function xe(o){return t=>{let e=o(`${t}/depthwise_filter`,4),r=o(`${t}/pointwise_filter`,4),n=o(`${t}/bias`,1);return new Be(e,r,n)}}function Y(o,t){return(e,r,n)=>{let a=o[e];if(!Ot(a,r))throw new Error(`expected weightMap[${e}] to be a Tensor${r}D, instead have ${a}`);return t.push({originalPath:e,paramPath:n||e}),a}}function R(o){let t=o;function e(n){let a=t.slice(0,n);return t=t.slice(n),a}function r(){return t}return{extractWeights:e,getRemainingWeights:r}}function br(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c,m=!1){let p=m?e(s,i,3,`${c}/conv0`):r(s,i,`${c}/conv0`),u=r(i,i,`${c}/conv1`),f=r(i,i,`${c}/conv2`);return{conv0:p,conv1:u,conv2:f}}function a(s,i,c,m=!1){let{conv0:p,conv1:u,conv2:f}=n(s,i,c,m),l=r(i,i,`${c}/conv3`);return{conv0:p,conv1:u,conv2:f,conv3:l}}return{extractDenseBlock3Params:n,extractDenseBlock4Params:a}}function Mo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock4Params:n}=br(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2"),c=n(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i,dense3:c}}}function gr(o){return t=>{let e=o(`${t}/filters`,4),r=o(`${t}/bias`,1);return{filters:e,bias:r}}}function xr(o,t){let e=Y(o,t),r=gr(e),n=xe(e);function a(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`);return{conv0:m,conv1:p,conv2:u}}function s(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`),f=n(`${i}/conv3`);return{conv0:m,conv1:p,conv2:u,conv3:f}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:s}}function Co(o){let t=[],{extractDenseBlock4Params:e}=xr(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2"),dense3:e("dense3")};return B(o,t),{params:r,paramMappings:t}}var ve=class extends A{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceFeatureExtractor - load model before inference");return ye.tidy(()=>{let r=ye.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=ke(a,e.dense0,!0);return s=ke(s,e.dense1),s=ke(s,e.dense2),s=ke(s,e.dense3),s=ye.avgPool(s,[7,7],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return Co(t)}extractParams(t){return Mo(t)}};var So=v(x());var _e=v(x());function $e(o,t){return _e.tidy(()=>_e.add(_e.matMul(o,t.weights),t.bias))}function Io(o,t,e){let r=[],{extractWeights:n,getRemainingWeights:a}=R(o),i=hr(n,r)(t,e,"fc");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:r,params:{fc:i}}}function No(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:r("fc")};return B(o,t),{params:n,paramMappings:t}}function vr(o){let t={},e={};return Object.keys(o).forEach(r=>{let n=r.startsWith("fc")?e:t;n[r]=o[r]}),{featureExtractorMap:t,classifierMap:e}}var Te=class extends A{constructor(e,r){super(e);this._faceFeatureExtractor=r}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return So.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e;return $e(n.as2D(n.shape[0],-1),r.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return Io(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=vr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),No(n)}extractParams(e){let r=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),a=n*r+n,s=e.slice(0,e.length-a),i=e.slice(e.length-a);return this.faceFeatureExtractor.extractWeights(s),this.extractClassifierParams(i)}};var so=["neutral","happy","sad","angry","fearful","disgusted","surprised"],yt=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);so.forEach((e,r)=>{this[e]=t[r]})}asSortedArray(){return so.map(t=>({expression:t,probability:this[t]})).sort((t,e)=>e.probability-t.probability)}};var Oe=class extends Te{constructor(t=new ve){super("FaceExpressionNet",t)}forwardInput(t){return we.tidy(()=>we.softmax(this.runNet(t)))}async forward(t){return this.forwardInput(await C(t))}async predictExpressions(t){let e=await C(t),r=await this.forwardInput(e),n=await Promise.all(we.unstack(r).map(async s=>{let i=s.dataSync();return s.dispose(),i}));r.dispose();let a=n.map(s=>new yt(s));return e.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function io(o){return o.expressions instanceof yt}function yr(o,t){return{...o,...{expressions:t}}}function Gn(o,t,e=.1,r){(Array.isArray(t)?t:[t]).forEach(a=>{let s=a instanceof yt?a:io(a)?a.expressions:void 0;if(!s)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let c=s.asSortedArray().filter(u=>u.probability>e),m=pt(a)?a.detection.box.bottomLeft:r||new g(0,0);new gt(c.map(u=>`${u.expression} (${Ht(u.probability)})`),m).draw(o)})}function Zt(o){return pt(o)&&o.landmarks instanceof z&&o.unshiftedLandmarks instanceof z&&o.alignedRect instanceof M}function jn(o){let t=c=>c*180/Math.PI,e=(c,m)=>Math.sqrt((c._x-m._x)**2+(c._y-m._y)**2),r={roll:void 0,pitch:void 0,yaw:void 0},n=(c,m,p)=>{let u=Math.floor(c._x-m._x),f=Math.floor(m._x-p._x);return u-f},a=(c,m)=>{let p=Math.hypot(m._x-c._x,m._y-c._y),u=m._y-c._y,f=Math.asin(u/p),l=t(f),b=Math.floor(90-l),y=m._x-c._x<0?-1:1;return b*y},s=(c,m,p)=>{let u=e(c,p),f={_x:(c._x+p._x)/2,_y:(c._y+p._y)/2},l=e(m,f),b=Math.atan(l/u),y=Math.floor(t(b)),F=f._y-m._y<0?-1:1;return y*F};if(!o||!o._positions||o._positions.length!==68)return r;let i=o._positions;return r.roll=a(i[27],i[66]),r.pitch=s(i[14],i[30],i[2]),r.yaw=n(i[14],i[33],i[2]),r}function Pe(o,t){let{box:e}=o.detection,r=t.shiftBy(e.x,e.y),n=r.align(),{imageDims:a}=o.detection,s=new M(o.detection.score,n.rescale(a.reverse()),a),i=jn(t);return{...o,...{landmarks:r,unshiftedLandmarks:t,alignedRect:s,angle:i}}}var _r=class{constructor(t={}){let{drawLines:e=!0,drawPoints:r=!0,lineWidth:n,lineColor:a,pointSize:s,pointColor:i}=t;this.drawLines=e,this.drawPoints=r,this.lineWidth=n||1,this.pointSize=s||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},Tr=class{constructor(t,e={}){this.faceLandmarks=t,this.options=new _r(e)}draw(t){let e=O(t),{drawLines:r,drawPoints:n,lineWidth:a,lineColor:s,pointSize:i,pointColor:c}=this.options;if(r&&this.faceLandmarks instanceof Gt&&(e.strokeStyle=s,e.lineWidth=a,dt(e,this.faceLandmarks.getJawOutline()),dt(e,this.faceLandmarks.getLeftEyeBrow()),dt(e,this.faceLandmarks.getRightEyeBrow()),dt(e,this.faceLandmarks.getNose()),dt(e,this.faceLandmarks.getLeftEye(),!0),dt(e,this.faceLandmarks.getRightEye(),!0),dt(e,this.faceLandmarks.getMouth(),!0)),n){e.strokeStyle=c,e.fillStyle=c;let m=p=>{e.beginPath(),e.arc(p.x,p.y,i,0,2*Math.PI),e.fill()};this.faceLandmarks.positions.forEach(m)}}};function Un(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof z?r:Zt(r)?r.landmarks:void 0;if(!n)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new Tr(n).draw(o)})}var Lo="1.7.6";var ft=v(x());var S=v(x());function qn(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c){let m=r(s,i,`${c}/separable_conv0`),p=r(i,i,`${c}/separable_conv1`),u=e(s,i,1,`${c}/expansion_conv`);return{separable_conv0:m,separable_conv1:p,expansion_conv:u}}function a(s,i){let c=r(s,s,`${i}/separable_conv0`),m=r(s,s,`${i}/separable_conv1`),p=r(s,s,`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:e,extractSeparableConvParams:r,extractReductionBlockParams:n,extractMainBlockParams:a}}function Ao(o,t){let e=[],{extractWeights:r,getRemainingWeights:n}=R(o),{extractConvParams:a,extractSeparableConvParams:s,extractReductionBlockParams:i,extractMainBlockParams:c}=qn(r,e),m=a(3,32,3,"entry_flow/conv_in"),p=i(32,64,"entry_flow/reduction_block_0"),u=i(64,128,"entry_flow/reduction_block_1"),f={conv_in:m,reduction_block_0:p,reduction_block_1:u},l={};it(t,0,1).forEach(h=>{l[`main_block_${h}`]=c(128,`middle_flow/main_block_${h}`)});let b=i(128,256,"exit_flow/reduction_block"),y=s(256,512,"exit_flow/separable_conv"),F={reduction_block:b,separable_conv:y};if(n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{paramMappings:e,params:{entry_flow:f,middle_flow:l,exit_flow:F}}}function Zn(o,t){let e=Y(o,t),r=gr(e),n=xe(e);function a(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=r(`${i}/expansion_conv`);return{separable_conv0:c,separable_conv1:m,expansion_conv:p}}function s(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=n(`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}}function Wo(o,t){let e=[],{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}=Zn(o,e),i=r("entry_flow/conv_in"),c=a("entry_flow/reduction_block_0"),m=a("entry_flow/reduction_block_1"),p={conv_in:i,reduction_block_0:c,reduction_block_1:m},u={};it(t,0,1).forEach(y=>{u[`main_block_${y}`]=s(`middle_flow/main_block_${y}`)});let f=a("exit_flow/reduction_block"),l=n("exit_flow/separable_conv"),b={reduction_block:f,separable_conv:l};return B(o,e),{params:{entry_flow:p,middle_flow:u,exit_flow:b},paramMappings:e}}function ko(o,t,e){return S.add(S.conv2d(o,t.filters,e,"same"),t.bias)}function mo(o,t,e=!0){let r=e?S.relu(o):o;return r=H(r,t.separable_conv0,[1,1]),r=H(S.relu(r),t.separable_conv1,[1,1]),r=S.maxPool(r,[3,3],[2,2],"same"),r=S.add(r,ko(o,t.expansion_conv,[2,2])),r}function Kn(o,t){let e=H(S.relu(o),t.separable_conv0,[1,1]);return e=H(S.relu(e),t.separable_conv1,[1,1]),e=H(S.relu(e),t.separable_conv2,[1,1]),e=S.add(e,o),e}var wr=class extends A{constructor(e){super("TinyXception");this._numMainBlocks=e}forwardInput(e){let{params:r}=this;if(!r)throw new Error("TinyXception - load model before inference");return S.tidy(()=>{let n=S.cast(e.toBatchTensor(112,!0),"float32"),s=rt(n,[122.782,117.001,104.298]).div(255),i=S.relu(ko(s,r.entry_flow.conv_in,[2,2]));return i=mo(i,r.entry_flow.reduction_block_0,!1),i=mo(i,r.entry_flow.reduction_block_1),it(this._numMainBlocks,0,1).forEach(c=>{i=Kn(i,r.middle_flow[`main_block_${c}`])}),i=mo(i,r.exit_flow.reduction_block),i=S.relu(H(i,r.exit_flow.separable_conv,[1,1])),i})}async forward(e){return this.forwardInput(await C(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return Wo(e,this._numMainBlocks)}extractParams(e){return Ao(e,this._numMainBlocks)}};function Bo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),n=hr(e,t),a=n(512,1,"fc/age"),s=n(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:a,gender:s}}}}function Ro(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:{age:r("fc/age"),gender:r("fc/gender")}};return B(o,t),{params:n,paramMappings:t}}var Pr=(e=>(e.FEMALE="female",e.MALE="male",e))(Pr||{});var He=class extends A{constructor(e=new wr(2)){super("AgeGenderNet");this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return ft.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e,a=ft.avgPool(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1),s=$e(a,r.fc.age).as1D(),i=$e(a,r.fc.gender);return{age:s,gender:i}})}forwardInput(e){return ft.tidy(()=>{let{age:r,gender:n}=this.runNet(e);return{age:r,gender:ft.softmax(n)}})}async forward(e){return this.forwardInput(await C(e))}async predictAgeAndGender(e){let r=await C(e),n=await this.forwardInput(r),a=ft.unstack(n.age),s=ft.unstack(n.gender),i=a.map((m,p)=>({ageTensor:m,genderTensor:s[p]})),c=await Promise.all(i.map(async({ageTensor:m,genderTensor:p})=>{let u=m.dataSync()[0],f=p.dataSync()[0],l=f>.5,b=l?"male":"female",y=l?f:1-f;return m.dispose(),p.dispose(),{age:u,gender:b,genderProbability:y}}));return n.age.dispose(),n.gender.dispose(),r.isBatchInput?c:c[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return Bo(e)}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=vr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Ro(n)}extractParams(e){let n=e.slice(0,e.length-1539),a=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(n),this.extractClassifierParams(a)}};var G=v(x());var Fe=class extends Te{postProcess(t,e,r){let n=r.map(({width:s,height:i})=>{let c=e/Math.max(i,s);return{width:s*c,height:i*c}}),a=n.length;return G.tidy(()=>{let s=(u,f)=>G.stack([G.fill([68],u,"float32"),G.fill([68],f,"float32")],1).as2D(1,136).as1D(),i=(u,f)=>{let{width:l,height:b}=n[u];return f(l,b)?Math.abs(l-b)/2:0},c=u=>i(u,(f,l)=>f<l),m=u=>i(u,(f,l)=>l<f);return t.mul(G.fill([a,136],e,"float32")).sub(G.stack(Array.from(Array(a),(u,f)=>s(c(f),m(f))))).div(G.stack(Array.from(Array(a),(u,f)=>s(n[f].width,n[f].height))))})}forwardInput(t){return G.tidy(()=>{let e=this.runNet(t);return this.postProcess(e,t.inputSize,t.inputDimensions.map(([r,n])=>({height:r,width:n})))})}async forward(t){return this.forwardInput(await C(t))}async detectLandmarks(t){let e=await C(t),r=G.tidy(()=>G.unstack(this.forwardInput(e))),n=await Promise.all(r.map(async(a,s)=>{let i=Array.from(a.dataSync()),c=i.filter((p,u)=>nr(u)),m=i.filter((p,u)=>!nr(u));return new Gt(Array(68).fill(0).map((p,u)=>new g(c[u],m[u])),{height:e.getInputHeight(s),width:e.getInputWidth(s)})}));return r.forEach(a=>a.dispose()),e.isBatchInput?n:n[0]}getClassifierChannelsOut(){return 136}};var Kt=class extends Fe{constructor(t=new ve){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};var De=v(x());function $o(o){let t=[],{extractDenseBlock3Params:e}=xr(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2")};return B(o,t),{params:r,paramMappings:t}}function Oo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock3Params:n}=br(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i}}}var Fr=class extends A{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("TinyFaceFeatureExtractor - load model before inference");return De.tidy(()=>{let r=De.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=fr(a,e.dense0,!0);return s=fr(s,e.dense1),s=fr(s,e.dense2),s=De.avgPool(s,[14,14],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return $o(t)}extractParams(t){return Oo(t)}};var ze=class extends Fe{constructor(t=new Fr){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var po=class extends Kt{};var nt=v(x());var Ee=v(x());var Dr=v(x());function Ho(o,t){return Dr.add(Dr.mul(o,t.weights),t.biases)}function uo(o,t,e,r,n="same"){let{filters:a,bias:s}=t.conv,i=Ee.conv2d(o,a,e,n);return i=Ee.add(i,s),i=Ho(i,t.scale),r?Ee.relu(i):i}function zo(o,t){return uo(o,t,[1,1],!0)}function fo(o,t){return uo(o,t,[1,1],!1)}function Er(o,t){return uo(o,t,[2,2],!0,"valid")}var j=v(x());function Qn(o,t){function e(i,c,m){let p=o(i),u=p.length/(c*m*m);if($r(u))throw new Error(`depth has to be an integer: ${u}, weights.length: ${p.length}, numFilters: ${c}, filterSize: ${m}`);return j.tidy(()=>j.transpose(j.tensor4d(p,[c,u,m,m]),[2,3,1,0]))}function r(i,c,m,p){let u=e(i,c,m),f=j.tensor1d(o(c));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/bias`}),{filters:u,bias:f}}function n(i,c){let m=j.tensor1d(o(i)),p=j.tensor1d(o(i));return t.push({paramPath:`${c}/weights`},{paramPath:`${c}/biases`}),{weights:m,biases:p}}function a(i,c,m,p){let u=r(i,c,m,`${p}/conv`),f=n(c,`${p}/scale`);return{conv:u,scale:f}}function s(i,c,m,p,u=!1){let f=a((u?.5:1)*i,c,m,`${p}/conv1`),l=a(i,c,m,`${p}/conv2`);return{conv1:f,conv2:l}}return{extractConvLayerParams:a,extractResidualLayerParams:s}}function Vo(o){let{extractWeights:t,getRemainingWeights:e}=R(o),r=[],{extractConvLayerParams:n,extractResidualLayerParams:a}=Qn(t,r),s=n(4704,32,7,"conv32_down"),i=a(9216,32,3,"conv32_1"),c=a(9216,32,3,"conv32_2"),m=a(9216,32,3,"conv32_3"),p=a(36864,64,3,"conv64_down",!0),u=a(36864,64,3,"conv64_1"),f=a(36864,64,3,"conv64_2"),l=a(36864,64,3,"conv64_3"),b=a(147456,128,3,"conv128_down",!0),y=a(147456,128,3,"conv128_1"),F=a(147456,128,3,"conv128_2"),h=a(589824,256,3,"conv256_down",!0),T=a(589824,256,3,"conv256_1"),_=a(589824,256,3,"conv256_2"),E=a(589824,256,3,"conv256_down_out"),W=j.tidy(()=>j.transpose(j.tensor2d(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),e().length!==0)throw new Error(`weights remaing after extract: ${e().length}`);return{params:{conv32_down:s,conv32_1:i,conv32_2:c,conv32_3:m,conv64_down:p,conv64_1:u,conv64_2:f,conv64_3:l,conv128_down:b,conv128_1:y,conv128_2:F,conv256_down:h,conv256_1:T,conv256_2:_,conv256_down_out:E,fc:W},paramMappings:r}}function ta(o,t){let e=Y(o,t);function r(s){let i=e(`${s}/scale/weights`,1),c=e(`${s}/scale/biases`,1);return{weights:i,biases:c}}function n(s){let i=e(`${s}/conv/filters`,4),c=e(`${s}/conv/bias`,1),m=r(s);return{conv:{filters:i,bias:c},scale:m}}function a(s){return{conv1:n(`${s}/conv1`),conv2:n(`${s}/conv2`)}}return{extractConvLayerParams:n,extractResidualLayerParams:a}}function Yo(o){let t=[],{extractConvLayerParams:e,extractResidualLayerParams:r}=ta(o,t),n=e("conv32_down"),a=r("conv32_1"),s=r("conv32_2"),i=r("conv32_3"),c=r("conv64_down"),m=r("conv64_1"),p=r("conv64_2"),u=r("conv64_3"),f=r("conv128_down"),l=r("conv128_1"),b=r("conv128_2"),y=r("conv256_down"),F=r("conv256_1"),h=r("conv256_2"),T=r("conv256_down_out"),{fc:_}=o;if(t.push({originalPath:"fc",paramPath:"fc"}),!Rr(_))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${_}`);let E={conv32_down:n,conv32_1:a,conv32_2:s,conv32_3:i,conv64_down:c,conv64_1:m,conv64_2:p,conv64_3:u,conv128_down:f,conv128_1:l,conv128_2:b,conv256_down:y,conv256_1:F,conv256_2:h,conv256_down_out:T,fc:_};return B(o,t),{params:E,paramMappings:t}}var $=v(x());function ot(o,t){let e=zo(o,t.conv1);return e=fo(e,t.conv2),e=$.add(e,o),e=$.relu(e),e}function Ve(o,t){let e=Er(o,t.conv1);e=fo(e,t.conv2);let r=$.avgPool(o,2,2,"valid"),n=$.zeros(r.shape),a=r.shape[3]!==e.shape[3];if(r.shape[1]!==e.shape[1]||r.shape[2]!==e.shape[2]){let i=[...e.shape];i[1]=1;let c=$.zeros(i);e=$.concat([e,c],1);let m=[...e.shape];m[2]=1;let p=$.zeros(m);e=$.concat([e,p],2)}return r=a?$.concat([r,n],3):r,e=$.add(r,e),e=$.relu(e),e}var Qt=class extends A{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceRecognitionNet - load model before inference");return nt.tidy(()=>{let r=nt.cast(t.toBatchTensor(150,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=Er(a,e.conv32_down);s=nt.maxPool(s,3,2,"valid"),s=ot(s,e.conv32_1),s=ot(s,e.conv32_2),s=ot(s,e.conv32_3),s=Ve(s,e.conv64_down),s=ot(s,e.conv64_1),s=ot(s,e.conv64_2),s=ot(s,e.conv64_3),s=Ve(s,e.conv128_down),s=ot(s,e.conv128_1),s=ot(s,e.conv128_2),s=Ve(s,e.conv256_down),s=ot(s,e.conv256_1),s=ot(s,e.conv256_2),s=Ve(s,e.conv256_down_out);let i=s.mean([1,2]);return nt.matMul(i,e.fc)})}async forward(t){return this.forwardInput(await C(t))}async computeFaceDescriptor(t){var a;if((a=t==null?void 0:t.shape)!=null&&a.some(s=>s<=0))return new Float32Array(128);let e=await C(t),r=nt.tidy(()=>nt.unstack(this.forwardInput(e))),n=await Promise.all(r.map(s=>s.data()));return r.forEach(s=>s.dispose()),e.isBatchInput?n:n[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return Yo(t)}extractParams(t){return Vo(t)}};function ea(o){let t=new Qt;return t.extractWeights(o),t}function Mr(o,t){return{...o,...{descriptor:t}}}function ra(o){return typeof o.age=="number"}function Cr(o,t){return{...o,...{age:t}}}function oa(o){return(o.gender==="male"||o.gender==="female")&&me(o.genderProbability)}function Ir(o,t,e){return{...o,...{gender:t,genderProbability:e}}}var Lt=v(x());var at=v(x());function na(o,t){function e(c,m){let p=at.tensor4d(o(9*c),[3,3,c,1]),u=at.tensor1d(o(c)),f=at.tensor1d(o(c)),l=at.tensor1d(o(c)),b=at.tensor1d(o(c));return t.push({paramPath:`${m}/filters`},{paramPath:`${m}/batch_norm_scale`},{paramPath:`${m}/batch_norm_offset`},{paramPath:`${m}/batch_norm_mean`},{paramPath:`${m}/batch_norm_variance`}),{filters:p,batch_norm_scale:u,batch_norm_offset:f,batch_norm_mean:l,batch_norm_variance:b}}function r(c,m,p,u,f){let l=at.tensor4d(o(c*m*p*p),[p,p,c,m]),b=at.tensor1d(o(m));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/${f?"batch_norm_offset":"bias"}`}),{filters:l,bias:b}}function n(c,m,p,u){let{filters:f,bias:l}=r(c,m,p,u,!0);return{filters:f,batch_norm_offset:l}}function a(c,m,p){let u=e(c,`${p}/depthwise_conv`),f=n(c,m,1,`${p}/pointwise_conv`);return{depthwise_conv:u,pointwise_conv:f}}function s(){let c=n(3,32,3,"mobilenetv1/conv_0"),m=a(32,64,"mobilenetv1/conv_1"),p=a(64,128,"mobilenetv1/conv_2"),u=a(128,128,"mobilenetv1/conv_3"),f=a(128,256,"mobilenetv1/conv_4"),l=a(256,256,"mobilenetv1/conv_5"),b=a(256,512,"mobilenetv1/conv_6"),y=a(512,512,"mobilenetv1/conv_7"),F=a(512,512,"mobilenetv1/conv_8"),h=a(512,512,"mobilenetv1/conv_9"),T=a(512,512,"mobilenetv1/conv_10"),_=a(512,512,"mobilenetv1/conv_11"),E=a(512,1024,"mobilenetv1/conv_12"),W=a(1024,1024,"mobilenetv1/conv_13");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,conv_8:F,conv_9:h,conv_10:T,conv_11:_,conv_12:E,conv_13:W}}function i(){let c=n(1024,256,1,"prediction_layer/conv_0"),m=n(256,512,3,"prediction_layer/conv_1"),p=n(512,128,1,"prediction_layer/conv_2"),u=n(128,256,3,"prediction_layer/conv_3"),f=n(256,128,1,"prediction_layer/conv_4"),l=n(128,256,3,"prediction_layer/conv_5"),b=n(256,64,1,"prediction_layer/conv_6"),y=n(64,128,3,"prediction_layer/conv_7"),F=r(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),h=r(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),T=r(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),_=r(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),E=r(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),W=r(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),tt=r(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),lt=r(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),q=r(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),Dt=r(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),Et=r(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),Mt=r(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,box_predictor_0:{box_encoding_predictor:F,class_predictor:h},box_predictor_1:{box_encoding_predictor:T,class_predictor:_},box_predictor_2:{box_encoding_predictor:E,class_predictor:W},box_predictor_3:{box_encoding_predictor:tt,class_predictor:lt},box_predictor_4:{box_encoding_predictor:q,class_predictor:Dt},box_predictor_5:{box_encoding_predictor:Et,class_predictor:Mt}}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:i}}function Go(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractMobilenetV1Params:n,extractPredictionLayerParams:a}=na(e,t),s=n(),i=a(),m={extra_dim:at.tensor3d(e(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:s,prediction_layer:i,output_layer:m},paramMappings:t}}function aa(o,t){let e=Y(o,t);function r(m,p,u){let f=e(`${m}/Conv2d_${p}_pointwise/weights`,4,`${u}/filters`),l=e(`${m}/Conv2d_${p}_pointwise/convolution_bn_offset`,1,`${u}/batch_norm_offset`);return{filters:f,batch_norm_offset:l}}function n(m){let p=`mobilenetv1/conv_${m}`,u=`MobilenetV1/Conv2d_${m}_depthwise`,f=`${p}/depthwise_conv`,l=`${p}/pointwise_conv`,b=e(`${u}/depthwise_weights`,4,`${f}/filters`),y=e(`${u}/BatchNorm/gamma`,1,`${f}/batch_norm_scale`),F=e(`${u}/BatchNorm/beta`,1,`${f}/batch_norm_offset`),h=e(`${u}/BatchNorm/moving_mean`,1,`${f}/batch_norm_mean`),T=e(`${u}/BatchNorm/moving_variance`,1,`${f}/batch_norm_variance`);return{depthwise_conv:{filters:b,batch_norm_scale:y,batch_norm_offset:F,batch_norm_mean:h,batch_norm_variance:T},pointwise_conv:r("MobilenetV1",m,l)}}function a(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:n(1),conv_2:n(2),conv_3:n(3),conv_4:n(4),conv_5:n(5),conv_6:n(6),conv_7:n(7),conv_8:n(8),conv_9:n(9),conv_10:n(10),conv_11:n(11),conv_12:n(12),conv_13:n(13)}}function s(m,p){let u=e(`${m}/weights`,4,`${p}/filters`),f=e(`${m}/biases`,1,`${p}/bias`);return{filters:u,bias:f}}function i(m){let p=s(`Prediction/BoxPredictor_${m}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${m}/box_encoding_predictor`),u=s(`Prediction/BoxPredictor_${m}/ClassPredictor`,`prediction_layer/box_predictor_${m}/class_predictor`);return{box_encoding_predictor:p,class_predictor:u}}function c(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}return{extractMobilenetV1Params:a,extractPredictionLayerParams:c}}function jo(o){let t=[],{extractMobilenetV1Params:e,extractPredictionLayerParams:r}=aa(o,t),n=o["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!ht(n))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${n}`);let a={mobilenetv1:e(),prediction_layer:r(),output_layer:{extra_dim:n}};return B(o,t),{params:a,paramMappings:t}}var _t=v(x());var Nt=v(x());function Z(o,t,e){return Nt.tidy(()=>{let r=Nt.conv2d(o,t.filters,e,"same");return r=Nt.add(r,t.batch_norm_offset),Nt.clipByValue(r,0,6)})}var sa=.0010000000474974513;function ia(o,t,e){return _t.tidy(()=>{let r=_t.depthwiseConv2d(o,t.filters,e,"same");return r=_t.batchNorm(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,sa),_t.clipByValue(r,0,6)})}function ca(o){return[2,4,6,12].some(t=>t===o)?[2,2]:[1,1]}function Uo(o,t){return _t.tidy(()=>{let e,r=Z(o,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((a,s)=>{let i=s+1,c=ca(i);r=ia(r,a.depthwise_conv,c),r=Z(r,a.pointwise_conv,[1,1]),i===11&&(e=r)}),e===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:e}})}function ma(o,t,e){let r=o.arraySync(),n=Math.min(r[t][0],r[t][2]),a=Math.min(r[t][1],r[t][3]),s=Math.max(r[t][0],r[t][2]),i=Math.max(r[t][1],r[t][3]),c=Math.min(r[e][0],r[e][2]),m=Math.min(r[e][1],r[e][3]),p=Math.max(r[e][0],r[e][2]),u=Math.max(r[e][1],r[e][3]),f=(s-n)*(i-a),l=(p-c)*(u-m);if(f<=0||l<=0)return 0;let b=Math.max(n,c),y=Math.max(a,m),F=Math.min(s,p),h=Math.min(i,u),T=Math.max(F-b,0)*Math.max(h-y,0);return T/(f+l-T)}function Xo(o,t,e,r,n){let a=o.shape[0],s=Math.min(e,a),i=t.map((p,u)=>({score:p,boxIndex:u})).filter(p=>p.score>n).sort((p,u)=>u.score-p.score),c=p=>p<=r?1:0,m=[];return i.forEach(p=>{if(m.length>=s)return;let u=p.score;for(let f=m.length-1;f>=0;--f){let l=ma(o,p.boxIndex,m[f]);if(l!==0&&(p.score*=c(l),p.score<=n))break}u===p.score&&m.push(p.boxIndex)}),m}var d=v(x());function pa(o){let t=d.unstack(d.transpose(o,[1,0])),e=[d.sub(t[2],t[0]),d.sub(t[3],t[1])],r=[d.add(t[0],d.div(e[0],2)),d.add(t[1],d.div(e[1],2))];return{sizes:e,centers:r}}function ua(o,t){let{sizes:e,centers:r}=pa(o),n=d.unstack(d.transpose(t,[1,0])),a=d.div(d.mul(d.exp(d.div(n[2],5)),e[0]),2),s=d.add(d.mul(d.div(n[0],10),e[0]),r[0]),i=d.div(d.mul(d.exp(d.div(n[3],5)),e[1]),2),c=d.add(d.mul(d.div(n[1],10),e[1]),r[1]);return d.transpose(d.stack([d.sub(s,a),d.sub(c,i),d.add(s,a),d.add(c,i)]),[1,0])}function Jo(o,t,e){return d.tidy(()=>{let r=o.shape[0],n=ua(d.reshape(d.tile(e.extra_dim,[r,1,1]),[-1,4]),d.reshape(o,[-1,4]));n=d.reshape(n,[r,n.shape[0]/r,4]);let a=d.sigmoid(d.slice(t,[0,0,1],[-1,-1,-1])),s=d.slice(a,[0,0,0],[-1,-1,1]);s=d.reshape(s,[r,s.shape[1]]);let i=d.unstack(n),c=d.unstack(s);return{boxes:i,scores:c}})}var Ge=v(x());var Ye=v(x());function te(o,t){return Ye.tidy(()=>{let e=o.shape[0],r=Ye.reshape(qt(o,t.box_encoding_predictor),[e,-1,1,4]),n=Ye.reshape(qt(o,t.class_predictor),[e,-1,3]);return{boxPredictionEncoding:r,classPrediction:n}})}function qo(o,t,e){return Ge.tidy(()=>{let r=Z(o,e.conv_0,[1,1]),n=Z(r,e.conv_1,[2,2]),a=Z(n,e.conv_2,[1,1]),s=Z(a,e.conv_3,[2,2]),i=Z(s,e.conv_4,[1,1]),c=Z(i,e.conv_5,[2,2]),m=Z(c,e.conv_6,[1,1]),p=Z(m,e.conv_7,[2,2]),u=te(t,e.box_predictor_0),f=te(o,e.box_predictor_1),l=te(n,e.box_predictor_2),b=te(s,e.box_predictor_3),y=te(c,e.box_predictor_4),F=te(p,e.box_predictor_5),h=Ge.concat([u.boxPredictionEncoding,f.boxPredictionEncoding,l.boxPredictionEncoding,b.boxPredictionEncoding,y.boxPredictionEncoding,F.boxPredictionEncoding],1),T=Ge.concat([u.classPrediction,f.classPrediction,l.classPrediction,b.classPrediction,y.classPrediction,F.classPrediction],1);return{boxPredictions:h,classPredictions:T}})}var X=class{constructor({minConfidence:t,maxResults:e}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=e||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var St=class extends A{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("SsdMobilenetv1 - load model before inference");return Lt.tidy(()=>{let r=Lt.cast(t.toBatchTensor(512,!1),"float32"),n=Lt.sub(Lt.div(r,127.5),1),a=Uo(n,e.mobilenetv1),{boxPredictions:s,classPredictions:i}=qo(a.out,a.conv11,e.prediction_layer);return Jo(s,i,e.output_layer)})}async forward(t){return this.forwardInput(await C(t))}async locateFaces(t,e={}){let{maxResults:r,minConfidence:n}=new X(e),a=await C(t),{boxes:s,scores:i}=this.forwardInput(a),c=s[0],m=i[0];for(let _=1;_<s.length;_++)s[_].dispose(),i[_].dispose();let p=Array.from(m.dataSync()),f=Xo(c,p,r,.5,n),l=a.getReshapedInputDimensions(0),b=a.inputSize,y=b/l.width,F=b/l.height,h=c.arraySync(),T=f.map(_=>{let[E,W]=[Math.max(0,h[_][0]),Math.min(1,h[_][2])].map(q=>q*F),[tt,lt]=[Math.max(0,h[_][1]),Math.min(1,h[_][3])].map(q=>q*y);return new M(p[_],new Yt(tt,E,lt-tt,W-E),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return c.dispose(),m.dispose(),T}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return jo(t)}extractParams(t){return Go(t)}};function Zo(o){let t=new St;return t.extractWeights(o),t}function fa(o){return Zo(o)}var lo=class extends St{};var Ko=.4,Qo=[new g(.738768,.874946),new g(2.42204,2.65704),new g(4.30971,7.04493),new g(10.246,4.59428),new g(12.6868,11.8741)],tn=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],en=[117.001,114.697,97.404],rn="tiny_yolov2_model",on="tiny_yolov2_separable_conv_model";var N=v(x());var Nr=o=>typeof o=="number";function ho(o){if(!o)throw new Error(`invalid config: ${o}`);if(typeof o.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${o.withSeparableConvs}`);if(!Nr(o.iouThreshold)||o.iouThreshold<0||o.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${o.iouThreshold}`);if(!Array.isArray(o.classes)||!o.classes.length||!o.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(o.classes)}`);if(!Array.isArray(o.anchors)||!o.anchors.length||!o.anchors.map(t=>t||{}).every(t=>Nr(t.x)&&Nr(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(o.anchors)}`);if(o.meanRgb&&(!Array.isArray(o.meanRgb)||o.meanRgb.length!==3||!o.meanRgb.every(Nr)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(o.meanRgb)}`)}var Q=v(x());var K=v(x());function Me(o){return K.tidy(()=>{let t=K.mul(o,K.scalar(.10000000149011612));return K.add(K.relu(K.sub(o,t)),t)})}function Tt(o,t){return Q.tidy(()=>{let e=Q.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=Q.conv2d(e,t.conv.filters,[1,1],"valid"),e=Q.sub(e,t.bn.sub),e=Q.mul(e,t.bn.truediv),e=Q.add(e,t.conv.bias),Me(e)})}var At=v(x());function wt(o,t){return At.tidy(()=>{let e=At.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=At.separableConv2d(e,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),e=At.add(e,t.bias),Me(e)})}var bo=v(x());function la(o,t){let e=be(o,t);function r(s,i){let c=bo.tensor1d(o(s)),m=bo.tensor1d(o(s));return t.push({paramPath:`${i}/sub`},{paramPath:`${i}/truediv`}),{sub:c,truediv:m}}function n(s,i,c){let m=e(s,i,3,`${c}/conv`),p=r(i,`${c}/bn`);return{conv:m,bn:p}}let a=ge(o,t);return{extractConvParams:e,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}}function nn(o,t,e,r){let{extractWeights:n,getRemainingWeights:a}=R(o),s=[],{extractConvParams:i,extractConvWithBatchNormParams:c,extractSeparableConvParams:m}=la(n,s),p;if(t.withSeparableConvs){let[u,f,l,b,y,F,h,T,_]=r,E=t.isFirstLayerConv2d?i(u,f,3,"conv0"):m(u,f,"conv0"),W=m(f,l,"conv1"),tt=m(l,b,"conv2"),lt=m(b,y,"conv3"),q=m(y,F,"conv4"),Dt=m(F,h,"conv5"),Et=T?m(h,T,"conv6"):void 0,Mt=_?m(T,_,"conv7"):void 0,$t=i(_||T||h,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}else{let[u,f,l,b,y,F,h,T,_]=r,E=c(u,f,"conv0"),W=c(f,l,"conv1"),tt=c(l,b,"conv2"),lt=c(b,y,"conv3"),q=c(y,F,"conv4"),Dt=c(F,h,"conv5"),Et=c(h,T,"conv6"),Mt=c(T,_,"conv7"),$t=i(_,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:p,paramMappings:s}}function da(o,t){let e=Y(o,t);function r(i){let c=e(`${i}/sub`,1),m=e(`${i}/truediv`,1);return{sub:c,truediv:m}}function n(i){let c=e(`${i}/filters`,4),m=e(`${i}/bias`,1);return{filters:c,bias:m}}function a(i){let c=n(`${i}/conv`),m=r(`${i}/bn`);return{conv:c,bn:m}}let s=xe(e);return{extractConvParams:n,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}}function an(o,t){let e=[],{extractConvParams:r,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}=da(o,e),s;if(t.withSeparableConvs){let i=t.filterSizes&&t.filterSizes.length||9;s={conv0:t.isFirstLayerConv2d?r("conv0"):a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:i>7?a("conv6"):void 0,conv7:i>8?a("conv7"):void 0,conv8:r("conv8")}}else s={conv0:n("conv0"),conv1:n("conv1"),conv2:n("conv2"),conv3:n("conv3"),conv4:n("conv4"),conv5:n("conv5"),conv6:n("conv6"),conv7:n("conv7"),conv8:r("conv8")};return B(o,e),{params:s,paramMappings:e}}var st=class{constructor({inputSize:t,scoreThreshold:e}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=e||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var go=class extends A{constructor(e){super("TinyYolov2");ho(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,r){let n=Tt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=Tt(n,r.conv6),n=Tt(n,r.conv7),qt(n,r.conv8,"valid",!1)}runMobilenet(e,r){let n=this.config.isFirstLayerConv2d?Me(qt(e,r.conv0,"valid",!1)):wt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=r.conv6?wt(n,r.conv6):n,n=r.conv7?wt(n,r.conv7):n,qt(n,r.conv8,"valid",!1)}forwardInput(e,r){let{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return N.tidy(()=>{let a=N.cast(e.toBatchTensor(r,!1),"float32");return a=this.config.meanRgb?rt(a,this.config.meanRgb):a,a=a.div(255),this.config.withSeparableConvs?this.runMobilenet(a,n):this.runTinyYolov2(a,n)})}async forward(e,r){return this.forwardInput(await C(e),r)}async detect(e,r={}){let{inputSize:n,scoreThreshold:a}=new st(r),s=await C(e),i=await this.forwardInput(s,n),c=N.tidy(()=>N.unstack(i)[0].expandDims()),m={width:s.getInputWidth(0),height:s.getInputHeight(0)},p=await this.extractBoxes(c,s.getReshapedInputDimensions(0),a);i.dispose(),c.dispose();let u=p.map(h=>h.box),f=p.map(h=>h.score),l=p.map(h=>h.classScore),b=p.map(h=>this.config.classes[h.label]);return Yr(u.map(h=>h.rescale(n)),f,this.config.iouThreshold,!0).map(h=>new bt(f[h],l[h],b[h],u[h],m))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return an(e,this.config)}extractParams(e){let r=this.config.filterSizes||go.DEFAULT_FILTER_SIZES,n=r?r.length:void 0;if(n!==7&&n!==8&&n!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);return nn(e,this.config,this.boxEncodingSize,r)}async extractBoxes(e,r,n){let{width:a,height:s}=r,i=Math.max(a,s),c=i/a,m=i/s,p=e.shape[1],u=this.config.anchors.length,[f,l,b]=N.tidy(()=>{let T=e.reshape([p,p,u,this.boxEncodingSize]),_=T.slice([0,0,0,0],[p,p,u,4]),E=T.slice([0,0,0,4],[p,p,u,1]),W=this.withClassScores?N.softmax(T.slice([0,0,0,5],[p,p,u,this.config.classes.length]),3):N.scalar(0);return[_,E,W]}),y=[],F=await l.array(),h=await f.array();for(let T=0;T<p;T++)for(let _=0;_<p;_++)for(let E=0;E<u;E++){let W=Ne(F[T][_][E][0]);if(!n||W>n){let tt=(_+Ne(h[T][_][E][0]))/p*c,lt=(T+Ne(h[T][_][E][1]))/p*m,q=Math.exp(h[T][_][E][2])*this.config.anchors[E].x/p*c,Dt=Math.exp(h[T][_][E][3])*this.config.anchors[E].y/p*m,Et=tt-q/2,Mt=lt-Dt/2,$t={row:T,col:_,anchor:E},{classScore:yo,label:_o}=this.withClassScores?await this.extractPredictedClass(b,$t):{classScore:1,label:0};y.push({box:new Vt(Et,Mt,Et+q,Mt+Dt),score:W,classScore:W*yo,label:_o,...$t})}}return f.dispose(),l.dispose(),b.dispose(),y}async extractPredictedClass(e,r){let{row:n,col:a,anchor:s}=r,i=await e.array();return Array(this.config.classes.length).fill(0).map((c,m)=>i[n][a][s][m]).map((c,m)=>({classScore:c,label:m})).reduce((c,m)=>c.classScore>m.classScore?c:m)}},ee=go;ee.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var re=class extends ee{constructor(t=!0){let e={withSeparableConvs:t,iouThreshold:Ko,classes:["face"],...t?{anchors:tn,meanRgb:en}:{anchors:Qo,withClassScores:!0}};super(e)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?on:rn}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function ha(o,t=!0){let e=new re(t);return e.extractWeights(o),e}var je=class extends st{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var J=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};var Xe=v(x());var xo=v(x());async function oe(o,t,e,r,n=({alignedRect:a})=>a){let a=o.map(c=>Zt(c)?n(c):c.detection),s=r||(t instanceof xo.Tensor?await de(t,a):await le(t,a)),i=await e(s);return s.forEach(c=>c instanceof xo.Tensor&&c.dispose()),i}async function Ce(o,t,e,r,n){return oe([o],t,async a=>e(a[0]),r,n)}var sn=.4,cn=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],mn=[117.001,114.697,97.404];var ne=class extends ee{constructor(){let t={withSeparableConvs:!0,iouThreshold:sn,classes:["face"],anchors:cn,meanRgb:mn,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var P={ssdMobilenetv1:new St,tinyFaceDetector:new ne,tinyYolov2:new re,faceLandmark68Net:new Kt,faceLandmark68TinyNet:new ze,faceRecognitionNet:new Qt,faceExpressionNet:new Oe,ageGenderNet:new He},pn=(o,t)=>P.ssdMobilenetv1.locateFaces(o,t),ba=(o,t)=>P.tinyFaceDetector.locateFaces(o,t),ga=(o,t)=>P.tinyYolov2.locateFaces(o,t),un=o=>P.faceLandmark68Net.detectLandmarks(o),xa=o=>P.faceLandmark68TinyNet.detectLandmarks(o),va=o=>P.faceRecognitionNet.computeFaceDescriptor(o),ya=o=>P.faceExpressionNet.predictExpressions(o),_a=o=>P.ageGenderNet.predictAgeAndGender(o),fn=o=>P.ssdMobilenetv1.load(o),Ta=o=>P.tinyFaceDetector.load(o),wa=o=>P.tinyYolov2.load(o),Pa=o=>P.faceLandmark68Net.load(o),Fa=o=>P.faceLandmark68TinyNet.load(o),Da=o=>P.faceRecognitionNet.load(o),Ea=o=>P.faceExpressionNet.load(o),Ma=o=>P.ageGenderNet.load(o),Ca=fn,Ia=pn,Na=un;var Sr=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ae=class extends Sr{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.faceExpressionNet.predictExpressions(n))),this.extractedFaces);return t.map((r,n)=>yr(r,e[n]))}withAgeAndGender(){return new ie(this,this.input)}},se=class extends Sr{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceExpressionNet.predictExpressions(r),this.extractedFaces);return yr(t,e)}withAgeAndGender(){return new ce(this,this.input)}},Wt=class extends ae{withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},kt=class extends se{withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Lr=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ie=class extends Lr{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.ageGenderNet.predictAgeAndGender(n))),this.extractedFaces);return t.map((r,n)=>{let{age:a,gender:s,genderProbability:i}=e[n];return Cr(Ir(r,s,i),a)})}withFaceExpressions(){return new ae(this,this.input)}},ce=class extends Lr{async run(){let t=await this.parentTask;if(!t)return;let{age:e,gender:r,genderProbability:n}=await Ce(t,this.input,a=>P.ageGenderNet.predictAgeAndGender(a),this.extractedFaces);return Cr(Ir(t,r,n),e)}withFaceExpressions(){return new se(this,this.input)}},Bt=class extends ie{withFaceExpressions(){return new Wt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Rt=class extends ce{withFaceExpressions(){return new kt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ue=class extends J{constructor(e,r){super();this.parentTask=e;this.input=r}},Pt=class extends Ue{async run(){let t=await this.parentTask;return(await oe(t,this.input,r=>Promise.all(r.map(n=>P.faceRecognitionNet.computeFaceDescriptor(n))),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}))).map((r,n)=>Mr(t[n],r))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}},Ft=class extends Ue{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceRecognitionNet.computeFaceDescriptor(r),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}));return Mr(t,e)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}};var Je=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?P.faceLandmark68TinyNet:P.faceLandmark68Net}},qe=class extends Je{async run(){let t=await this.parentTask,e=t.map(s=>s.detection),r=this.input instanceof Xe.Tensor?await de(this.input,e):await le(this.input,e),n=await Promise.all(r.map(s=>this.landmarkNet.detectLandmarks(s)));return r.forEach(s=>s instanceof Xe.Tensor&&s.dispose()),t.filter((s,i)=>n[i]).map((s,i)=>Pe(s,n[i]))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Ze=class extends Je{async run(){let t=await this.parentTask;if(!t)return;let{detection:e}=t,r=this.input instanceof Xe.Tensor?await de(this.input,[e]):await le(this.input,[e]),n=await this.landmarkNet.detectLandmarks(r[0]);return r.forEach(a=>a instanceof Xe.Tensor&&a.dispose()),Pe(t,n)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ke=class extends J{constructor(e,r=new X){super();this.input=e;this.options=r}},Ie=class extends Ke{async run(){let{input:t,options:e}=this,r;if(e instanceof je)r=P.tinyFaceDetector.locateFaces(t,e);else if(e instanceof X)r=P.ssdMobilenetv1.locateFaces(t,e);else if(e instanceof st)r=P.tinyYolov2.locateFaces(t,e);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return r}runAndExtendWithFaceDetections(){return new Promise((t,e)=>{this.run().then(r=>t(r.map(n=>jt({},n)))).catch(r=>e(r))})}withFaceLandmarks(t=!1){return new qe(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new ae(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new ie(this.runAndExtendWithFaceDetections(),this.input)}},Qe=class extends Ke{async run(){let t=await new Ie(this.input,this.options),e=t[0];return t.forEach(r=>{r.score>e.score&&(e=r)}),e}runAndExtendWithFaceDetection(){return new Promise(async t=>{let e=await this.run();t(e?jt({},e):void 0)})}withFaceLandmarks(t=!1){return new Ze(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new se(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new ce(this.runAndExtendWithFaceDetection(),this.input)}};function Sa(o,t=new X){return new Qe(o,t)}function Ar(o,t=new X){return new Ie(o,t)}async function ln(o,t){return Ar(o,new X(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function La(o,t={}){return Ar(o,new st(t)).withFaceLandmarks().withFaceDescriptors()}var Aa=ln;function vo(o,t){if(o.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let e=Array.from(o),r=Array.from(t);return Math.sqrt(e.map((n,a)=>n-r[a]).reduce((n,a)=>n+a*a,0))}var tr=class{constructor(t,e=.6){this._distanceThreshold=e;let r=Array.isArray(t)?t:[t];if(!r.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let n=1,a=()=>`person ${n++}`;this._labeledDescriptors=r.map(s=>{if(s instanceof mt)return s;if(s instanceof Float32Array)return new mt(a(),[s]);if(s.descriptor&&s.descriptor instanceof Float32Array)return new mt(a(),[s.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,e){return e.map(r=>vo(r,t)).reduce((r,n)=>r+n,0)/(e.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:e,label:r})=>new pe(r,this.computeMeanDistance(t,e))).reduce((e,r)=>e.distance<r.distance?e:r)}findBestMatch(t){let e=this.matchDescriptor(t);return e.distance<this._distanceThreshold?e:new pe("unknown",e.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let e=t.labeledDescriptors.map(r=>mt.fromJSON(r));return new tr(e,t.distanceThreshold)}};function Wa(o){let t=new ne;return t.extractWeights(o),t}function dn(o,t){let{width:e,height:r}=new k(t.width,t.height);if(e<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:e,height:r})}`);if(Array.isArray(o))return o.map(n=>dn(n,{width:e,height:r}));if(Zt(o)){let n=o.detection.forSize(e,r),a=o.unshiftedLandmarks.forSize(n.box.width,n.box.height);return Pe(jt(o,n),a)}return pt(o)?jt(o,o.detection.forSize(e,r)):o instanceof z||o instanceof M?o.forSize(e,r):o}var Ba=Lo;0&&(module.exports={AgeGenderNet,BoundingBox,Box,ComposableTask,ComputeAllFaceDescriptorsTask,ComputeFaceDescriptorsTaskBase,ComputeSingleFaceDescriptorTask,DetectAllFaceLandmarksTask,DetectAllFacesTask,DetectFaceLandmarksTaskBase,DetectFacesTaskBase,DetectSingleFaceLandmarksTask,DetectSingleFaceTask,Dimensions,FACE_EXPRESSION_LABELS,FaceDetection,FaceDetectionNet,FaceExpressionNet,FaceExpressions,FaceLandmark68Net,FaceLandmark68TinyNet,FaceLandmarkNet,FaceLandmarks,FaceLandmarks5,FaceLandmarks68,FaceMatch,FaceMatcher,FaceRecognitionNet,Gender,LabeledBox,LabeledFaceDescriptors,NetInput,NeuralNetwork,ObjectDetection,Point,PredictedBox,Rect,SsdMobilenetv1,SsdMobilenetv1Options,TinyFaceDetector,TinyFaceDetectorOptions,TinyYolov2,TinyYolov2Options,allFaces,allFacesSsdMobilenetv1,allFacesTinyYolov2,awaitMediaLoaded,bufferToImage,computeFaceDescriptor,createCanvas,createCanvasFromMedia,createFaceDetectionNet,createFaceRecognitionNet,createSsdMobilenetv1,createTinyFaceDetector,createTinyYolov2,detectAllFaces,detectFaceLandmarks,detectFaceLandmarksTiny,detectLandmarks,detectSingleFace,draw,env,euclideanDistance,extendWithAge,extendWithFaceDescriptor,extendWithFaceDetection,extendWithFaceExpressions,extendWithFaceLandmarks,extendWithGender,extractFaceTensors,extractFaces,fetchImage,fetchJson,fetchNetWeights,fetchOrThrow,fetchVideo,getContext2dOrThrow,getMediaDimensions,imageTensorToCanvas,imageToSquare,inverseSigmoid,iou,isMediaElement,isMediaLoaded,isWithAge,isWithFaceDetection,isWithFaceExpressions,isWithFaceLandmarks,isWithGender,loadAgeGenderModel,loadFaceDetectionModel,loadFaceExpressionModel,loadFaceLandmarkModel,loadFaceLandmarkTinyModel,loadFaceRecognitionModel,loadSsdMobilenetv1Model,loadTinyFaceDetectorModel,loadTinyYolov2Model,loadWeightMap,locateFaces,matchDimensions,minBbox,nets,nonMaxSuppression,normalize,padToSquare,predictAgeAndGender,recognizeFaceExpressions,resizeResults,resolveInput,shuffleArray,sigmoid,ssdMobilenetv1,tf,tinyFaceDetector,tinyYolov2,toNetInput,utils,validateConfig,version});
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __commonJS = (cb, mod) => function __require() {
+  return mod || (0, cb[__getOwnPropNames(cb)[0]])((mod = { exports: {} }).exports, mod), mod.exports;
+};
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// dist/tfjs.esm.js
+var require_tfjs_esm = __commonJS({
+  "dist/tfjs.esm.js"(exports, module2) {
+    "use strict";
+    var __defProp2 = Object.defineProperty;
+    var __getOwnPropDesc2 = Object.getOwnPropertyDescriptor;
+    var __getOwnPropNames2 = Object.getOwnPropertyNames;
+    var __hasOwnProp2 = Object.prototype.hasOwnProperty;
+    var __export2 = (target, all) => {
+      for (var name in all)
+        __defProp2(target, name, { get: all[name], enumerable: true });
+    };
+    var __copyProps2 = (to, from, except, desc) => {
+      if (from && typeof from === "object" || typeof from === "function") {
+        for (let key of __getOwnPropNames2(from))
+          if (!__hasOwnProp2.call(to, key) && key !== except)
+            __defProp2(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc2(from, key)) || desc.enumerable });
+      }
+      return to;
+    };
+    var __reExport = (target, mod, secondTarget) => (__copyProps2(target, mod, "default"), secondTarget && __copyProps2(secondTarget, mod, "default"));
+    var __toCommonJS2 = (mod) => __copyProps2(__defProp2({}, "__esModule", { value: true }), mod);
+    var tf_node_wasm_exports = {};
+    __export2(tf_node_wasm_exports, {
+      version: () => version6
+    });
+    module2.exports = __toCommonJS2(tf_node_wasm_exports);
+    __reExport(tf_node_wasm_exports, require("@tensorflow/tfjs"), module2.exports);
+    __reExport(tf_node_wasm_exports, require("@tensorflow/tfjs-backend-wasm"), module2.exports);
+    var version3 = "4.0.0";
+    var version22 = "4.0.0";
+    var version32 = "4.0.0";
+    var version4 = "4.0.0";
+    var version5 = "4.0.0";
+    var version6 = {
+      tfjs: version3,
+      "tfjs-core": version3,
+      "tfjs-converter": version22,
+      "tfjs-backend-cpu": version32,
+      "tfjs-backend-webgl": version4,
+      "tfjs-backend-wasm": version5
+    };
+  }
+});
+
+// src/index.ts
+var src_exports = {};
+__export(src_exports, {
+  AgeGenderNet: () => AgeGenderNet,
+  BoundingBox: () => BoundingBox,
+  Box: () => Box,
+  ComposableTask: () => ComposableTask,
+  ComputeAllFaceDescriptorsTask: () => ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase: () => ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask: () => ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask: () => DetectAllFaceLandmarksTask,
+  DetectAllFacesTask: () => DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase: () => DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase: () => DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask: () => DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask: () => DetectSingleFaceTask,
+  Dimensions: () => Dimensions,
+  FACE_EXPRESSION_LABELS: () => FACE_EXPRESSION_LABELS,
+  FaceDetection: () => FaceDetection,
+  FaceDetectionNet: () => FaceDetectionNet,
+  FaceExpressionNet: () => FaceExpressionNet,
+  FaceExpressions: () => FaceExpressions,
+  FaceLandmark68Net: () => FaceLandmark68Net,
+  FaceLandmark68TinyNet: () => FaceLandmark68TinyNet,
+  FaceLandmarkNet: () => FaceLandmarkNet,
+  FaceLandmarks: () => FaceLandmarks,
+  FaceLandmarks5: () => FaceLandmarks5,
+  FaceLandmarks68: () => FaceLandmarks68,
+  FaceMatch: () => FaceMatch,
+  FaceMatcher: () => FaceMatcher,
+  FaceRecognitionNet: () => FaceRecognitionNet,
+  Gender: () => Gender,
+  LabeledBox: () => LabeledBox,
+  LabeledFaceDescriptors: () => LabeledFaceDescriptors,
+  NetInput: () => NetInput,
+  NeuralNetwork: () => NeuralNetwork,
+  ObjectDetection: () => ObjectDetection,
+  Point: () => Point,
+  PredictedBox: () => PredictedBox,
+  Rect: () => Rect,
+  SsdMobilenetv1: () => SsdMobilenetv1,
+  SsdMobilenetv1Options: () => SsdMobilenetv1Options,
+  TinyFaceDetector: () => TinyFaceDetector,
+  TinyFaceDetectorOptions: () => TinyFaceDetectorOptions,
+  TinyYolov2: () => TinyYolov2,
+  TinyYolov2Options: () => TinyYolov2Options,
+  allFaces: () => allFaces,
+  allFacesSsdMobilenetv1: () => allFacesSsdMobilenetv1,
+  allFacesTinyYolov2: () => allFacesTinyYolov2,
+  awaitMediaLoaded: () => awaitMediaLoaded,
+  bufferToImage: () => bufferToImage,
+  computeFaceDescriptor: () => computeFaceDescriptor,
+  createCanvas: () => createCanvas,
+  createCanvasFromMedia: () => createCanvasFromMedia,
+  createFaceDetectionNet: () => createFaceDetectionNet,
+  createFaceRecognitionNet: () => createFaceRecognitionNet,
+  createSsdMobilenetv1: () => createSsdMobilenetv1,
+  createTinyFaceDetector: () => createTinyFaceDetector,
+  createTinyYolov2: () => createTinyYolov2,
+  detectAllFaces: () => detectAllFaces,
+  detectFaceLandmarks: () => detectFaceLandmarks,
+  detectFaceLandmarksTiny: () => detectFaceLandmarksTiny,
+  detectLandmarks: () => detectLandmarks,
+  detectSingleFace: () => detectSingleFace,
+  draw: () => draw_exports,
+  env: () => env,
+  euclideanDistance: () => euclideanDistance,
+  extendWithAge: () => extendWithAge,
+  extendWithFaceDescriptor: () => extendWithFaceDescriptor,
+  extendWithFaceDetection: () => extendWithFaceDetection,
+  extendWithFaceExpressions: () => extendWithFaceExpressions,
+  extendWithFaceLandmarks: () => extendWithFaceLandmarks,
+  extendWithGender: () => extendWithGender,
+  extractFaceTensors: () => extractFaceTensors,
+  extractFaces: () => extractFaces,
+  fetchImage: () => fetchImage,
+  fetchJson: () => fetchJson,
+  fetchNetWeights: () => fetchNetWeights,
+  fetchOrThrow: () => fetchOrThrow,
+  fetchVideo: () => fetchVideo,
+  getContext2dOrThrow: () => getContext2dOrThrow,
+  getMediaDimensions: () => getMediaDimensions,
+  imageTensorToCanvas: () => imageTensorToCanvas,
+  imageToSquare: () => imageToSquare,
+  inverseSigmoid: () => inverseSigmoid,
+  iou: () => iou,
+  isMediaElement: () => isMediaElement,
+  isMediaLoaded: () => isMediaLoaded,
+  isWithAge: () => isWithAge,
+  isWithFaceDetection: () => isWithFaceDetection,
+  isWithFaceExpressions: () => isWithFaceExpressions,
+  isWithFaceLandmarks: () => isWithFaceLandmarks,
+  isWithGender: () => isWithGender,
+  loadAgeGenderModel: () => loadAgeGenderModel,
+  loadFaceDetectionModel: () => loadFaceDetectionModel,
+  loadFaceExpressionModel: () => loadFaceExpressionModel,
+  loadFaceLandmarkModel: () => loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel: () => loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel: () => loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model: () => loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel: () => loadTinyFaceDetectorModel,
+  loadTinyYolov2Model: () => loadTinyYolov2Model,
+  loadWeightMap: () => loadWeightMap,
+  locateFaces: () => locateFaces,
+  matchDimensions: () => matchDimensions,
+  minBbox: () => minBbox,
+  nets: () => nets,
+  nonMaxSuppression: () => nonMaxSuppression,
+  normalize: () => normalize,
+  padToSquare: () => padToSquare,
+  predictAgeAndGender: () => predictAgeAndGender,
+  recognizeFaceExpressions: () => recognizeFaceExpressions,
+  resizeResults: () => resizeResults,
+  resolveInput: () => resolveInput,
+  shuffleArray: () => shuffleArray,
+  sigmoid: () => sigmoid,
+  ssdMobilenetv1: () => ssdMobilenetv1,
+  tf: () => tf42,
+  tinyFaceDetector: () => tinyFaceDetector,
+  tinyYolov2: () => tinyYolov2,
+  toNetInput: () => toNetInput,
+  utils: () => utils_exports,
+  validateConfig: () => validateConfig,
+  version: () => version2
+});
+module.exports = __toCommonJS(src_exports);
+var tf42 = __toESM(require_tfjs_esm());
+
+// src/draw/index.ts
+var draw_exports = {};
+__export(draw_exports, {
+  AnchorPosition: () => AnchorPosition,
+  DrawBox: () => DrawBox,
+  DrawBoxOptions: () => DrawBoxOptions,
+  DrawFaceLandmarks: () => DrawFaceLandmarks,
+  DrawFaceLandmarksOptions: () => DrawFaceLandmarksOptions,
+  DrawTextField: () => DrawTextField,
+  DrawTextFieldOptions: () => DrawTextFieldOptions,
+  drawContour: () => drawContour,
+  drawDetections: () => drawDetections,
+  drawFaceExpressions: () => drawFaceExpressions,
+  drawFaceLandmarks: () => drawFaceLandmarks
+});
+
+// src/draw/drawContour.ts
+function drawContour(ctx, points, isClosed = false) {
+  ctx.beginPath();
+  points.slice(1).forEach(({ x, y }, prevIdx) => {
+    const from = points[prevIdx];
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(x, y);
+  });
+  if (isClosed) {
+    const from = points[points.length - 1];
+    const to = points[0];
+    if (!from || !to) {
+      return;
+    }
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(to.x, to.y);
+  }
+  ctx.stroke();
+}
+
+// src/utils/index.ts
+var utils_exports = {};
+__export(utils_exports, {
+  computeReshapedDimensions: () => computeReshapedDimensions,
+  getCenterPoint: () => getCenterPoint,
+  isDimensions: () => isDimensions,
+  isEven: () => isEven,
+  isFloat: () => isFloat,
+  isTensor: () => isTensor,
+  isTensor1D: () => isTensor1D,
+  isTensor2D: () => isTensor2D,
+  isTensor3D: () => isTensor3D,
+  isTensor4D: () => isTensor4D,
+  isValidNumber: () => isValidNumber,
+  isValidProbablitiy: () => isValidProbablitiy,
+  range: () => range,
+  round: () => round
+});
+var tf = __toESM(require_tfjs_esm());
+
+// src/classes/Dimensions.ts
+var Dimensions = class {
+  constructor(width, height) {
+    if (!isValidNumber(width) || !isValidNumber(height)) {
+      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);
+    }
+    this._width = width;
+    this._height = height;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  reverse() {
+    return new Dimensions(1 / this.width, 1 / this.height);
+  }
+};
+
+// src/utils/index.ts
+function isTensor(tensor2, dim) {
+  return tensor2 instanceof tf.Tensor && tensor2.shape.length === dim;
+}
+function isTensor1D(tensor2) {
+  return isTensor(tensor2, 1);
+}
+function isTensor2D(tensor2) {
+  return isTensor(tensor2, 2);
+}
+function isTensor3D(tensor2) {
+  return isTensor(tensor2, 3);
+}
+function isTensor4D(tensor2) {
+  return isTensor(tensor2, 4);
+}
+function isFloat(num) {
+  return num % 1 !== 0;
+}
+function isEven(num) {
+  return num % 2 === 0;
+}
+function round(num, prec = 2) {
+  const f = 10 ** prec;
+  return Math.floor(num * f) / f;
+}
+function isDimensions(obj) {
+  return obj && obj.width && obj.height;
+}
+function computeReshapedDimensions({ width, height }, inputSize) {
+  const scale2 = inputSize / Math.max(height, width);
+  return new Dimensions(Math.round(width * scale2), Math.round(height * scale2));
+}
+function getCenterPoint(pts) {
+  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0)).div(new Point(pts.length, pts.length));
+}
+function range(num, start, step) {
+  return Array(num).fill(0).map((_, i) => start + i * step);
+}
+function isValidNumber(num) {
+  return !!num && num !== Infinity && num !== -Infinity && !Number.isNaN(num) || num === 0;
+}
+function isValidProbablitiy(num) {
+  return isValidNumber(num) && num >= 0 && num <= 1;
+}
+
+// src/classes/Point.ts
+var Point = class {
+  constructor(x, y) {
+    this._x = x;
+    this._y = y;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  add(pt) {
+    return new Point(this.x + pt.x, this.y + pt.y);
+  }
+  sub(pt) {
+    return new Point(this.x - pt.x, this.y - pt.y);
+  }
+  mul(pt) {
+    return new Point(this.x * pt.x, this.y * pt.y);
+  }
+  div(pt) {
+    return new Point(this.x / pt.x, this.y / pt.y);
+  }
+  abs() {
+    return new Point(Math.abs(this.x), Math.abs(this.y));
+  }
+  magnitude() {
+    return Math.sqrt(this.x ** 2 + this.y ** 2);
+  }
+  floor() {
+    return new Point(Math.floor(this.x), Math.floor(this.y));
+  }
+};
+
+// src/classes/Box.ts
+var Box = class {
+  static isRect(rect) {
+    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);
+  }
+  static assertIsValidBox(box, callee, allowNegativeDimensions = false) {
+    if (!Box.isRect(box)) {
+      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);
+    }
+    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {
+      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);
+    }
+  }
+  constructor(_box, allowNegativeDimensions = true) {
+    const box = _box || {};
+    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);
+    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);
+    if (!isRect && !isBbox) {
+      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);
+    }
+    const [x, y, width, height] = isRect ? [box.x, box.y, box.width, box.height] : [box.left, box.top, box.right - box.left, box.bottom - box.top];
+    Box.assertIsValidBox({
+      x,
+      y,
+      width,
+      height
+    }, "Box.constructor", allowNegativeDimensions);
+    this._x = x;
+    this._y = y;
+    this._width = width;
+    this._height = height;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  get left() {
+    return this.x;
+  }
+  get top() {
+    return this.y;
+  }
+  get right() {
+    return this.x + this.width;
+  }
+  get bottom() {
+    return this.y + this.height;
+  }
+  get area() {
+    return this.width * this.height;
+  }
+  get topLeft() {
+    return new Point(this.left, this.top);
+  }
+  get topRight() {
+    return new Point(this.right, this.top);
+  }
+  get bottomLeft() {
+    return new Point(this.left, this.bottom);
+  }
+  get bottomRight() {
+    return new Point(this.right, this.bottom);
+  }
+  round() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.round(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  floor() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.floor(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  toSquare() {
+    let {
+      x,
+      y,
+      width,
+      height
+    } = this;
+    const diff = Math.abs(width - height);
+    if (width < height) {
+      x -= diff / 2;
+      width += diff;
+    }
+    if (height < width) {
+      y -= diff / 2;
+      height += diff;
+    }
+    return new Box({ x, y, width, height });
+  }
+  rescale(s) {
+    const scaleX = isDimensions(s) ? s.width : s;
+    const scaleY = isDimensions(s) ? s.height : s;
+    return new Box({
+      x: this.x * scaleX,
+      y: this.y * scaleY,
+      width: this.width * scaleX,
+      height: this.height * scaleY
+    });
+  }
+  pad(padX, padY) {
+    const [x, y, width, height] = [
+      this.x - padX / 2,
+      this.y - padY / 2,
+      this.width + padX,
+      this.height + padY
+    ];
+    return new Box({ x, y, width, height });
+  }
+  clipAtImageBorders(imgWidth, imgHeight) {
+    const { x, y, right, bottom } = this;
+    const clippedX = Math.max(x, 0);
+    const clippedY = Math.max(y, 0);
+    const newWidth = right - clippedX;
+    const newHeight = bottom - clippedY;
+    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);
+    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);
+    return new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight }).floor();
+  }
+  shift(sx, sy) {
+    const { width, height } = this;
+    const x = this.x + sx;
+    const y = this.y + sy;
+    return new Box({ x, y, width, height });
+  }
+  padAtBorders(imageHeight, imageWidth) {
+    const w = this.width + 1;
+    const h = this.height + 1;
+    const dx = 1;
+    const dy = 1;
+    let edx = w;
+    let edy = h;
+    let x = this.left;
+    let y = this.top;
+    let ex = this.right;
+    let ey = this.bottom;
+    if (ex > imageWidth) {
+      edx = -ex + imageWidth + w;
+      ex = imageWidth;
+    }
+    if (ey > imageHeight) {
+      edy = -ey + imageHeight + h;
+      ey = imageHeight;
+    }
+    if (x < 1) {
+      edy = 2 - x;
+      x = 1;
+    }
+    if (y < 1) {
+      edy = 2 - y;
+      y = 1;
+    }
+    return { dy, edy, dx, edx, y, ey, x, ex, w, h };
+  }
+  calibrate(region) {
+    return new Box({
+      left: this.left + region.left * this.width,
+      top: this.top + region.top * this.height,
+      right: this.right + region.right * this.width,
+      bottom: this.bottom + region.bottom * this.height
+    }).toSquare().round();
+  }
+};
+
+// src/classes/BoundingBox.ts
+var BoundingBox = class extends Box {
+  constructor(left, top, right, bottom, allowNegativeDimensions = false) {
+    super({ left, top, right, bottom }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/ObjectDetection.ts
+var ObjectDetection = class {
+  constructor(score, classScore, className, relativeBox, imageDims) {
+    this._imageDims = new Dimensions(imageDims.width, imageDims.height);
+    this._score = score;
+    this._classScore = classScore;
+    this._className = className;
+    this._box = new Box(relativeBox).rescale(this._imageDims);
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+  get className() {
+    return this._className;
+  }
+  get box() {
+    return this._box;
+  }
+  get imageDims() {
+    return this._imageDims;
+  }
+  get imageWidth() {
+    return this.imageDims.width;
+  }
+  get imageHeight() {
+    return this.imageDims.height;
+  }
+  get relativeBox() {
+    return new Box(this._box).rescale(this.imageDims.reverse());
+  }
+  forSize(width, height) {
+    return new ObjectDetection(
+      this.score,
+      this.classScore,
+      this.className,
+      this.relativeBox,
+      { width, height }
+    );
+  }
+};
+
+// src/classes/FaceDetection.ts
+var FaceDetection = class extends ObjectDetection {
+  constructor(score, relativeBox, imageDims) {
+    super(score, score, "", relativeBox, imageDims);
+  }
+  forSize(width, height) {
+    const { score, relativeBox, imageDims } = super.forSize(width, height);
+    return new FaceDetection(score, relativeBox, imageDims);
+  }
+};
+
+// src/ops/iou.ts
+function iou(box1, box2, isIOU = true) {
+  const width = Math.max(0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));
+  const height = Math.max(0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));
+  const interSection = width * height;
+  return isIOU ? interSection / (box1.area + box2.area - interSection) : interSection / Math.min(box1.area, box2.area);
+}
+
+// src/ops/minBbox.ts
+function minBbox(pts) {
+  const xs = pts.map((pt) => pt.x);
+  const ys = pts.map((pt) => pt.y);
+  const minX = xs.reduce((min, x) => x < min ? x : min, Infinity);
+  const minY = ys.reduce((min, y) => y < min ? y : min, Infinity);
+  const maxX = xs.reduce((max, x) => max < x ? x : max, 0);
+  const maxY = ys.reduce((max, y) => max < y ? y : max, 0);
+  return new BoundingBox(minX, minY, maxX, maxY);
+}
+
+// src/ops/nonMaxSuppression.ts
+function nonMaxSuppression(boxes, scores, iouThreshold, isIOU = true) {
+  let indicesSortedByScore = scores.map((score, boxIndex) => ({ score, boxIndex })).sort((c1, c2) => c1.score - c2.score).map((c) => c.boxIndex);
+  const pick = [];
+  while (indicesSortedByScore.length > 0) {
+    const curr = indicesSortedByScore.pop();
+    pick.push(curr);
+    const indices = indicesSortedByScore;
+    const outputs = [];
+    for (let i = 0; i < indices.length; i++) {
+      const idx = indices[i];
+      const currBox = boxes[curr];
+      const idxBox = boxes[idx];
+      outputs.push(iou(currBox, idxBox, isIOU));
+    }
+    indicesSortedByScore = indicesSortedByScore.filter(
+      (_, j) => outputs[j] <= iouThreshold
+    );
+  }
+  return pick;
+}
+
+// src/ops/normalize.ts
+var tf2 = __toESM(require_tfjs_esm());
+function normalize(x, meanRgb) {
+  return tf2.tidy(() => {
+    const [r, g, b] = meanRgb;
+    const avg_r = tf2.fill([...x.shape.slice(0, 3), 1], r, "float32");
+    const avg_g = tf2.fill([...x.shape.slice(0, 3), 1], g, "float32");
+    const avg_b = tf2.fill([...x.shape.slice(0, 3), 1], b, "float32");
+    const avg_rgb = tf2.concat([avg_r, avg_g, avg_b], 3);
+    return tf2.sub(x, avg_rgb);
+  });
+}
+
+// src/ops/padToSquare.ts
+var tf3 = __toESM(require_tfjs_esm());
+function padToSquare(imgTensor, isCenterImage = false) {
+  return tf3.tidy(() => {
+    const [height, width] = imgTensor.shape.slice(1);
+    if (height === width)
+      return imgTensor;
+    const dimDiff = Math.abs(height - width);
+    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));
+    const paddingAxis = height > width ? 2 : 1;
+    const createPaddingTensor = (paddingAmountLocal) => {
+      const paddingTensorShape = imgTensor.shape.slice();
+      paddingTensorShape[paddingAxis] = paddingAmountLocal;
+      return tf3.fill(paddingTensorShape, 0, "float32");
+    };
+    const paddingTensorAppend = createPaddingTensor(paddingAmount);
+    const remainingPaddingAmount = dimDiff - paddingTensorAppend.shape[paddingAxis];
+    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;
+    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend].filter((t) => !!t).map((t) => tf3.cast(t, "float32"));
+    return tf3.concat(tensorsToStack, paddingAxis);
+  });
+}
+
+// src/ops/shuffleArray.ts
+function shuffleArray(inputArray) {
+  const array = inputArray.slice();
+  for (let i = array.length - 1; i > 0; i--) {
+    const j = Math.floor(Math.random() * (i + 1));
+    const x = array[i];
+    array[i] = array[j];
+    array[j] = x;
+  }
+  return array;
+}
+
+// src/ops/index.ts
+function sigmoid(x) {
+  return 1 / (1 + Math.exp(-x));
+}
+function inverseSigmoid(x) {
+  return Math.log(x / (1 - x));
+}
+
+// src/classes/Rect.ts
+var Rect = class extends Box {
+  constructor(x, y, width, height, allowNegativeDimensions = false) {
+    super({ x, y, width, height }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/FaceLandmarks.ts
+var relX = 0.5;
+var relY = 0.43;
+var relScale = 0.45;
+var FaceLandmarks = class {
+  constructor(relativeFaceLandmarkPositions, imgDims, shift = new Point(0, 0)) {
+    const { width, height } = imgDims;
+    this._imgDims = new Dimensions(width, height);
+    this._shift = shift;
+    this._positions = relativeFaceLandmarkPositions.map(
+      (pt) => pt.mul(new Point(width, height)).add(shift)
+    );
+  }
+  get shift() {
+    return new Point(this._shift.x, this._shift.y);
+  }
+  get imageWidth() {
+    return this._imgDims.width;
+  }
+  get imageHeight() {
+    return this._imgDims.height;
+  }
+  get positions() {
+    return this._positions;
+  }
+  get relativePositions() {
+    return this._positions.map(
+      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight))
+    );
+  }
+  forSize(width, height) {
+    return new this.constructor(
+      this.relativePositions,
+      { width, height }
+    );
+  }
+  shiftBy(x, y) {
+    return new this.constructor(
+      this.relativePositions,
+      this._imgDims,
+      new Point(x, y)
+    );
+  }
+  shiftByPoint(pt) {
+    return this.shiftBy(pt.x, pt.y);
+  }
+  align(detection, options = {}) {
+    if (detection) {
+      const box = detection instanceof FaceDetection ? detection.box.floor() : new Box(detection);
+      return this.shiftBy(box.x, box.y).align(null, options);
+    }
+    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };
+    if (useDlibAlignment) {
+      return this.alignDlib();
+    }
+    return this.alignMinBbox(minBoxPadding);
+  }
+  alignDlib() {
+    const centers = this.getRefPointsForAlignment();
+    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;
+    const distToMouth = (pt) => mouthCenter.sub(pt).magnitude();
+    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;
+    const size = Math.floor(eyeToMouthDist / relScale);
+    const refPoint = getCenterPoint(centers);
+    const x = Math.floor(Math.max(0, refPoint.x - relX * size));
+    const y = Math.floor(Math.max(0, refPoint.y - relY * size));
+    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));
+  }
+  alignMinBbox(padding) {
+    const box = minBbox(this.positions);
+    return box.pad(box.width * padding, box.height * padding);
+  }
+  getRefPointsForAlignment() {
+    throw new Error("getRefPointsForAlignment not implemented by base class");
+  }
+};
+
+// src/classes/FaceLandmarks5.ts
+var FaceLandmarks5 = class extends FaceLandmarks {
+  getRefPointsForAlignment() {
+    const pts = this.positions;
+    return [
+      pts[0],
+      pts[1],
+      getCenterPoint([pts[3], pts[4]])
+    ];
+  }
+};
+
+// src/classes/FaceLandmarks68.ts
+var FaceLandmarks68 = class extends FaceLandmarks {
+  getJawOutline() {
+    return this.positions.slice(0, 17);
+  }
+  getLeftEyeBrow() {
+    return this.positions.slice(17, 22);
+  }
+  getRightEyeBrow() {
+    return this.positions.slice(22, 27);
+  }
+  getNose() {
+    return this.positions.slice(27, 36);
+  }
+  getLeftEye() {
+    return this.positions.slice(36, 42);
+  }
+  getRightEye() {
+    return this.positions.slice(42, 48);
+  }
+  getMouth() {
+    return this.positions.slice(48, 68);
+  }
+  getRefPointsForAlignment() {
+    return [
+      this.getLeftEye(),
+      this.getRightEye(),
+      this.getMouth()
+    ].map(getCenterPoint);
+  }
+};
+
+// src/classes/FaceMatch.ts
+var FaceMatch = class {
+  constructor(label, distance) {
+    this._label = label;
+    this._distance = distance;
+  }
+  get label() {
+    return this._label;
+  }
+  get distance() {
+    return this._distance;
+  }
+  toString(withDistance = true) {
+    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ""}`;
+  }
+};
+
+// src/classes/LabeledBox.ts
+var LabeledBox = class extends Box {
+  constructor(box, label) {
+    super(box);
+    this._label = label;
+  }
+  static assertIsValidLabeledBox(box, callee) {
+    Box.assertIsValidBox(box, callee);
+    if (!isValidNumber(box.label)) {
+      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);
+    }
+  }
+  get label() {
+    return this._label;
+  }
+};
+
+// src/classes/LabeledFaceDescriptors.ts
+var LabeledFaceDescriptors = class {
+  constructor(label, descriptors) {
+    if (!(typeof label === "string")) {
+      throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");
+    }
+    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {
+      throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");
+    }
+    this._label = label;
+    this._descriptors = descriptors;
+  }
+  get label() {
+    return this._label;
+  }
+  get descriptors() {
+    return this._descriptors;
+  }
+  toJSON() {
+    return {
+      label: this.label,
+      descriptors: this.descriptors.map((d) => Array.from(d))
+    };
+  }
+  static fromJSON(json) {
+    const descriptors = json.descriptors.map((d) => new Float32Array(d));
+    return new LabeledFaceDescriptors(json.label, descriptors);
+  }
+};
+
+// src/classes/PredictedBox.ts
+var PredictedBox = class extends LabeledBox {
+  constructor(box, label, score, classScore) {
+    super(box, label);
+    this._score = score;
+    this._classScore = classScore;
+  }
+  static assertIsValidPredictedBox(box, callee) {
+    LabeledBox.assertIsValidLabeledBox(box, callee);
+    if (!isValidProbablitiy(box.score) || !isValidProbablitiy(box.classScore)) {
+      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);
+    }
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+};
+
+// src/factories/WithFaceDetection.ts
+function isWithFaceDetection(obj) {
+  return obj.detection instanceof FaceDetection;
+}
+function extendWithFaceDetection(sourceObj, detection) {
+  const extension = { detection };
+  return { ...sourceObj, ...extension };
+}
+
+// src/env/createBrowserEnv.ts
+function createBrowserEnv() {
+  const fetch = window.fetch;
+  if (!fetch)
+    throw new Error("fetch - missing fetch implementation for browser environment");
+  const readFile = () => {
+    throw new Error("readFile - filesystem not available for browser environment");
+  };
+  return {
+    Canvas: HTMLCanvasElement,
+    CanvasRenderingContext2D,
+    Image: HTMLImageElement,
+    ImageData,
+    Video: HTMLVideoElement,
+    createCanvasElement: () => document.createElement("canvas"),
+    createImageElement: () => document.createElement("img"),
+    createVideoElement: () => document.createElement("video"),
+    fetch,
+    readFile
+  };
+}
+
+// src/env/isNodejs.ts
+function isNodejs() {
+  return typeof global === "object" && typeof process !== "undefined" && process.versions != null && process.versions.node != null;
+}
+
+// src/env/createFileSystem.ts
+function createFileSystem(fs) {
+  let requireFsError = "";
+  if (!fs && isNodejs()) {
+    try {
+      fs = require("fs");
+    } catch (err) {
+      requireFsError = err.toString();
+    }
+  }
+  const readFile = fs ? (filePath) => new Promise((resolve, reject) => {
+    fs.readFile(filePath, (err, buffer) => err ? reject(err) : resolve(buffer));
+  }) : () => {
+    throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`);
+  };
+  return { readFile };
+}
+
+// src/env/createNodejsEnv.ts
+function createNodejsEnv() {
+  const Canvas = global["Canvas"] || global.HTMLCanvasElement;
+  const Image = global.Image || global.HTMLImageElement;
+  const Video = global["Video"] || global.HTMLVideoElement;
+  const createCanvasElement = () => {
+    if (Canvas)
+      return new Canvas();
+    throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment");
+  };
+  const createImageElement = () => {
+    if (Image)
+      return new Image();
+    throw new Error("createImageElement - missing Image implementation for nodejs environment");
+  };
+  const createVideoElement = () => {
+    if (Video)
+      return new Video();
+    throw new Error("createVideoElement - missing Video implementation for nodejs environment");
+  };
+  const fetch = global.fetch;
+  const fileSystem = createFileSystem();
+  return {
+    Canvas: Canvas || class {
+    },
+    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {
+    },
+    Image: Image || class {
+    },
+    ImageData: global.ImageData || class {
+    },
+    Video: global.HTMLVideoElement || class {
+    },
+    createCanvasElement,
+    createImageElement,
+    createVideoElement,
+    fetch,
+    ...fileSystem
+  };
+}
+
+// src/env/isBrowser.ts
+function isBrowser() {
+  return typeof window === "object" && typeof document !== "undefined" && typeof HTMLImageElement !== "undefined" && typeof HTMLCanvasElement !== "undefined" && typeof HTMLVideoElement !== "undefined" && typeof ImageData !== "undefined" && typeof CanvasRenderingContext2D !== "undefined";
+}
+
+// src/env/index.ts
+var environment;
+function getEnv() {
+  if (!environment) {
+    throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  return environment;
+}
+function setEnv(env2) {
+  environment = env2;
+}
+function initialize() {
+  if (isBrowser())
+    return setEnv(createBrowserEnv());
+  if (isNodejs())
+    return setEnv(createNodejsEnv());
+  return null;
+}
+function monkeyPatch(env2) {
+  if (!environment) {
+    initialize();
+  }
+  if (!environment) {
+    throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  const { Canvas = environment.Canvas, Image = environment.Image } = env2;
+  environment.Canvas = Canvas;
+  environment.Image = Image;
+  environment.createCanvasElement = env2.createCanvasElement || (() => new Canvas());
+  environment.createImageElement = env2.createImageElement || (() => new Image());
+  environment.ImageData = env2.ImageData || environment.ImageData;
+  environment.Video = env2.Video || environment.Video;
+  environment.fetch = env2.fetch || environment.fetch;
+  environment.readFile = env2.readFile || environment.readFile;
+}
+var env = {
+  getEnv,
+  setEnv,
+  initialize,
+  createBrowserEnv,
+  createFileSystem,
+  createNodejsEnv,
+  monkeyPatch,
+  isBrowser,
+  isNodejs
+};
+initialize();
+
+// src/dom/resolveInput.ts
+function resolveInput(arg) {
+  if (!env.isNodejs() && typeof arg === "string") {
+    return document.getElementById(arg);
+  }
+  return arg;
+}
+
+// src/dom/getContext2dOrThrow.ts
+function getContext2dOrThrow(canvasArg) {
+  const { Canvas, CanvasRenderingContext2D: CanvasRenderingContext2D2 } = env.getEnv();
+  if (canvasArg instanceof CanvasRenderingContext2D2) {
+    return canvasArg;
+  }
+  const canvas = resolveInput(canvasArg);
+  if (!(canvas instanceof Canvas)) {
+    throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+  }
+  const ctx = canvas.getContext("2d");
+  if (!ctx) {
+    throw new Error("resolveContext2d - canvas 2d context is null");
+  }
+  return ctx;
+}
+
+// src/draw/DrawTextField.ts
+var AnchorPosition = /* @__PURE__ */ ((AnchorPosition2) => {
+  AnchorPosition2["TOP_LEFT"] = "TOP_LEFT";
+  AnchorPosition2["TOP_RIGHT"] = "TOP_RIGHT";
+  AnchorPosition2["BOTTOM_LEFT"] = "BOTTOM_LEFT";
+  AnchorPosition2["BOTTOM_RIGHT"] = "BOTTOM_RIGHT";
+  return AnchorPosition2;
+})(AnchorPosition || {});
+var DrawTextFieldOptions = class {
+  constructor(options = {}) {
+    const {
+      anchorPosition,
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = options;
+    this.anchorPosition = anchorPosition || "TOP_LEFT" /* TOP_LEFT */;
+    this.backgroundColor = backgroundColor || "rgba(0, 0, 0, 0.5)";
+    this.fontColor = fontColor || "rgba(255, 255, 255, 1)";
+    this.fontSize = fontSize || 14;
+    this.fontStyle = fontStyle || "Georgia";
+    this.padding = padding || 4;
+  }
+};
+var DrawTextField = class {
+  constructor(text, anchor, options = {}) {
+    this.text = typeof text === "string" ? [text] : text instanceof DrawTextField ? text.text : text;
+    this.anchor = anchor;
+    this.options = new DrawTextFieldOptions(options);
+  }
+  measureWidth(ctx) {
+    const { padding } = this.options;
+    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => w0 < w1 ? w1 : w0, 0) + 2 * padding;
+  }
+  measureHeight() {
+    const { fontSize, padding } = this.options;
+    return this.text.length * fontSize + 2 * padding;
+  }
+  getUpperLeft(ctx, canvasDims) {
+    const { anchorPosition } = this.options;
+    const isShiftLeft = anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */ || anchorPosition === "TOP_RIGHT" /* TOP_RIGHT */;
+    const isShiftTop = anchorPosition === "BOTTOM_LEFT" /* BOTTOM_LEFT */ || anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */;
+    const textFieldWidth = this.measureWidth(ctx);
+    const textFieldHeight = this.measureHeight();
+    const x = isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x;
+    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;
+    if (canvasDims) {
+      const { width, height } = canvasDims;
+      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);
+      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);
+      return { x: newX, y: newY };
+    }
+    return { x, y };
+  }
+  draw(canvasArg) {
+    const canvas = resolveInput(canvasArg);
+    const ctx = getContext2dOrThrow(canvas);
+    const {
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = this.options;
+    ctx.font = `${fontSize}px ${fontStyle}`;
+    const maxTextWidth = this.measureWidth(ctx);
+    const textHeight = this.measureHeight();
+    ctx.fillStyle = backgroundColor;
+    const upperLeft = this.getUpperLeft(ctx, canvas);
+    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);
+    ctx.fillStyle = fontColor;
+    this.text.forEach((textLine, i) => {
+      const x = padding + upperLeft.x;
+      const y = padding + upperLeft.y + (i + 1) * fontSize;
+      ctx.fillText(textLine, x, y);
+    });
+  }
+};
+
+// src/draw/DrawBox.ts
+var DrawBoxOptions = class {
+  constructor(options = {}) {
+    const {
+      boxColor,
+      lineWidth,
+      label,
+      drawLabelOptions
+    } = options;
+    this.boxColor = boxColor || "rgba(0, 0, 255, 1)";
+    this.lineWidth = lineWidth || 2;
+    this.label = label;
+    const defaultDrawLabelOptions = {
+      anchorPosition: "BOTTOM_LEFT" /* BOTTOM_LEFT */,
+      backgroundColor: this.boxColor
+    };
+    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });
+  }
+};
+var DrawBox = class {
+  constructor(box, options = {}) {
+    this.box = new Box(box);
+    this.options = new DrawBoxOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const { boxColor, lineWidth } = this.options;
+    const {
+      x,
+      y,
+      width,
+      height
+    } = this.box;
+    ctx.strokeStyle = boxColor;
+    ctx.lineWidth = lineWidth;
+    ctx.strokeRect(x, y, width, height);
+    const { label } = this.options;
+    if (label) {
+      new DrawTextField([label], { x: x - lineWidth / 2, y }, this.options.drawLabelOptions).draw(canvasArg);
+    }
+  }
+};
+
+// src/draw/drawDetections.ts
+function drawDetections(canvasArg, detections) {
+  const detectionsArray = Array.isArray(detections) ? detections : [detections];
+  detectionsArray.forEach((det) => {
+    const score = det instanceof FaceDetection ? det.score : isWithFaceDetection(det) ? det.detection.score : void 0;
+    const box = det instanceof FaceDetection ? det.box : isWithFaceDetection(det) ? det.detection.box : new Box(det);
+    const label = score ? `${round(score)}` : void 0;
+    new DrawBox(box, { label }).draw(canvasArg);
+  });
+}
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var tf18 = __toESM(require_tfjs_esm());
+
+// src/dom/isMediaLoaded.ts
+function isMediaLoaded(media) {
+  const { Image, Video } = env.getEnv();
+  return media instanceof Image && media.complete || media instanceof Video && media.readyState >= 3;
+}
+
+// src/dom/awaitMediaLoaded.ts
+function awaitMediaLoaded(media) {
+  return new Promise((resolve, reject) => {
+    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media))
+      resolve(null);
+    function onError(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      reject(e);
+    }
+    function onLoad(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      resolve(e);
+    }
+    media.addEventListener("load", onLoad);
+    media.addEventListener("error", onError);
+  });
+}
+
+// src/dom/bufferToImage.ts
+function bufferToImage(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToImage - expected buf to be of type: Blob"));
+    const reader = new FileReader();
+    reader.onload = () => {
+      if (typeof reader.result !== "string")
+        reject(new Error("bufferToImage - expected reader.result to be a string, in onload"));
+      const img = env.getEnv().createImageElement();
+      img.onload = () => resolve(img);
+      img.onerror = reject;
+      img.src = reader.result;
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(buf);
+  });
+}
+
+// src/dom/getMediaDimensions.ts
+function getMediaDimensions(input) {
+  const { Image, Video } = env.getEnv();
+  if (input instanceof Image) {
+    return new Dimensions(input.naturalWidth, input.naturalHeight);
+  }
+  if (input instanceof Video) {
+    return new Dimensions(input.videoWidth, input.videoHeight);
+  }
+  return new Dimensions(input.width, input.height);
+}
+
+// src/dom/createCanvas.ts
+function createCanvas({ width, height }) {
+  const { createCanvasElement } = env.getEnv();
+  const canvas = createCanvasElement();
+  canvas.width = width;
+  canvas.height = height;
+  return canvas;
+}
+function createCanvasFromMedia(media, dims) {
+  const { ImageData: ImageData2 } = env.getEnv();
+  if (!(media instanceof ImageData2) && !isMediaLoaded(media)) {
+    throw new Error("createCanvasFromMedia - media has not finished loading yet");
+  }
+  const { width, height } = dims || getMediaDimensions(media);
+  const canvas = createCanvas({ width, height });
+  if (media instanceof ImageData2) {
+    getContext2dOrThrow(canvas).putImageData(media, 0, 0);
+  } else {
+    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);
+  }
+  return canvas;
+}
+
+// src/dom/imageTensorToCanvas.ts
+var tf4 = __toESM(require_tfjs_esm());
+async function imageTensorToCanvas(imgTensor, canvas) {
+  const targetCanvas = canvas || env.getEnv().createCanvasElement();
+  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);
+  const imgTensor3D = tf4.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());
+  await tf4["browser"].toPixels(imgTensor3D, targetCanvas);
+  imgTensor3D.dispose();
+  return targetCanvas;
+}
+
+// src/dom/isMediaElement.ts
+function isMediaElement(input) {
+  const { Image, Canvas, Video } = env.getEnv();
+  return input instanceof Image || input instanceof Canvas || input instanceof Video;
+}
+
+// src/dom/NetInput.ts
+var tf5 = __toESM(require_tfjs_esm());
+
+// src/dom/imageToSquare.ts
+function imageToSquare(input, inputSize, centerImage = false) {
+  const { Image, Canvas } = env.getEnv();
+  if (!(input instanceof Image || input instanceof Canvas)) {
+    throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+  }
+  if (inputSize <= 0)
+    return createCanvas({ width: 1, height: 1 });
+  const dims = getMediaDimensions(input);
+  const scale2 = inputSize / Math.max(dims.height, dims.width);
+  const width = scale2 * dims.width;
+  const height = scale2 * dims.height;
+  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });
+  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);
+  const offset = Math.abs(width - height) / 2;
+  const dx = centerImage && width < height ? offset : 0;
+  const dy = centerImage && height < width ? offset : 0;
+  if (inputCanvas.width > 0 && inputCanvas.height > 0)
+    getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);
+  return targetCanvas;
+}
+
+// src/dom/NetInput.ts
+var NetInput = class {
+  constructor(inputs, treatAsBatchInput = false) {
+    this._imageTensors = [];
+    this._canvases = [];
+    this._treatAsBatchInput = false;
+    this._inputDimensions = [];
+    this._inputSize = 0;
+    if (!Array.isArray(inputs)) {
+      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);
+    }
+    this._treatAsBatchInput = treatAsBatchInput;
+    this._batchSize = inputs.length;
+    inputs.forEach((input, idx) => {
+      if (isTensor3D(input)) {
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape;
+        return;
+      }
+      if (isTensor4D(input)) {
+        const batchSize = input.shape[0];
+        if (batchSize !== 1) {
+          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+        }
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape.slice(1);
+        return;
+      }
+      const canvas = input instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);
+      this._canvases[idx] = canvas;
+      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];
+    });
+  }
+  get imageTensors() {
+    return this._imageTensors;
+  }
+  get canvases() {
+    return this._canvases;
+  }
+  get isBatchInput() {
+    return this.batchSize > 1 || this._treatAsBatchInput;
+  }
+  get batchSize() {
+    return this._batchSize;
+  }
+  get inputDimensions() {
+    return this._inputDimensions;
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get reshapedInputDimensions() {
+    return range(this.batchSize, 0, 1).map(
+      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx)
+    );
+  }
+  getInput(batchIdx) {
+    return this.canvases[batchIdx] || this.imageTensors[batchIdx];
+  }
+  getInputDimensions(batchIdx) {
+    return this._inputDimensions[batchIdx];
+  }
+  getInputHeight(batchIdx) {
+    return this._inputDimensions[batchIdx][0];
+  }
+  getInputWidth(batchIdx) {
+    return this._inputDimensions[batchIdx][1];
+  }
+  getReshapedInputDimensions(batchIdx) {
+    if (typeof this.inputSize !== "number") {
+      throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+    }
+    const width = this.getInputWidth(batchIdx);
+    const height = this.getInputHeight(batchIdx);
+    return computeReshapedDimensions({ width, height }, this.inputSize);
+  }
+  toBatchTensor(inputSize, isCenterInputs = true) {
+    this._inputSize = inputSize;
+    return tf5.tidy(() => {
+      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {
+        const input = this.getInput(batchIdx);
+        if (input instanceof tf5.Tensor) {
+          let imgTensor = isTensor4D(input) ? input : tf5.expandDims(input);
+          imgTensor = padToSquare(imgTensor, isCenterInputs);
+          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {
+            imgTensor = tf5["image"].resizeBilinear(imgTensor, [inputSize, inputSize], false, false);
+          }
+          return imgTensor.as3D(inputSize, inputSize, 3);
+        }
+        if (input instanceof env.getEnv().Canvas) {
+          return tf5["browser"].fromPixels(imageToSquare(input, inputSize, isCenterInputs));
+        }
+        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);
+      });
+      const batchTensor = tf5.stack(inputTensors.map((t) => tf5.cast(t, "float32"))).as4D(this.batchSize, inputSize, inputSize, 3);
+      return batchTensor;
+    });
+  }
+};
+
+// src/dom/toNetInput.ts
+async function toNetInput(inputs) {
+  if (inputs instanceof NetInput)
+    return inputs;
+  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];
+  if (!inputArgArray.length)
+    throw new Error("toNetInput - empty array passed as input");
+  const getIdxHint = (idx) => Array.isArray(inputs) ? ` at input index ${idx}:` : "";
+  const inputArray = inputArgArray.map(resolveInput);
+  inputArray.forEach((input, i) => {
+    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {
+      if (typeof inputArgArray[i] === "string")
+        throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);
+      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);
+    }
+    if (isTensor4D(input)) {
+      const batchSize = input.shape[0];
+      if (batchSize !== 1)
+        throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+    }
+  });
+  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));
+  return new NetInput(inputArray, Array.isArray(inputs));
+}
+
+// src/dom/extractFaces.ts
+async function extractFaces(input, detections) {
+  const { Canvas } = env.getEnv();
+  let canvas = input;
+  if (!(input instanceof Canvas)) {
+    const netInput = await toNetInput(input);
+    if (netInput.batchSize > 1)
+      throw new Error("extractFaces - batchSize > 1 not supported");
+    const tensorOrCanvas = netInput.getInput(0);
+    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);
+  }
+  const ctx = getContext2dOrThrow(canvas);
+  const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det).map((box) => box.clipAtImageBorders(canvas.width, canvas.height));
+  return boxes.map(({ x, y, width, height }) => {
+    const faceImg = createCanvas({ width, height });
+    if (width > 0 && height > 0)
+      getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);
+    return faceImg;
+  });
+}
+
+// src/dom/extractFaceTensors.ts
+var tf6 = __toESM(require_tfjs_esm());
+async function extractFaceTensors(imageTensor, detections) {
+  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {
+    throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+  }
+  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {
+    throw new Error("extractFaceTensors - batchSize > 1 not supported");
+  }
+  return tf6.tidy(() => {
+    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);
+    const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det).map((box) => box.clipAtImageBorders(imgWidth, imgHeight));
+    const faceTensors = boxes.filter((box) => box.width > 0 && box.height > 0).map(({ x, y, width, height }) => tf6.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));
+    return faceTensors;
+  });
+}
+
+// src/dom/fetchOrThrow.ts
+async function fetchOrThrow(url, init) {
+  const { fetch } = env.getEnv();
+  const res = await fetch(url, init);
+  if (!(res.status < 400)) {
+    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);
+  }
+  return res;
+}
+
+// src/dom/fetchImage.ts
+async function fetchImage(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("image/")) {
+    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToImage(blob);
+}
+
+// src/dom/fetchJson.ts
+async function fetchJson(uri) {
+  return (await fetchOrThrow(uri)).json();
+}
+
+// src/dom/fetchNetWeights.ts
+async function fetchNetWeights(uri) {
+  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());
+}
+
+// src/dom/bufferToVideo.ts
+function bufferToVideo(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToVideo - expected buf to be of type: Blob"));
+    const video = env.getEnv().createVideoElement();
+    video.oncanplay = () => resolve(video);
+    video.onerror = reject;
+    video.playsInline = true;
+    video.muted = true;
+    video.src = URL.createObjectURL(buf);
+    video.play();
+  });
+}
+
+// src/dom/fetchVideo.ts
+async function fetchVideo(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("video/")) {
+    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToVideo(blob);
+}
+
+// src/dom/loadWeightMap.ts
+var tf7 = __toESM(require_tfjs_esm());
+
+// src/common/getModelUris.ts
+function getModelUris(uri, defaultModelName) {
+  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;
+  if (!uri) {
+    return {
+      modelBaseUri: "",
+      manifestUri: defaultManifestFilename
+    };
+  }
+  if (uri === "/") {
+    return {
+      modelBaseUri: "/",
+      manifestUri: `/${defaultManifestFilename}`
+    };
+  }
+  const protocol = uri.startsWith("http://") ? "http://" : uri.startsWith("https://") ? "https://" : "";
+  uri = uri.replace(protocol, "");
+  const parts = uri.split("/").filter((s) => s);
+  const manifestFile = uri.endsWith(".json") ? parts[parts.length - 1] : defaultManifestFilename;
+  let modelBaseUri = protocol + (uri.endsWith(".json") ? parts.slice(0, parts.length - 1) : parts).join("/");
+  modelBaseUri = uri.startsWith("/") ? `/${modelBaseUri}` : modelBaseUri;
+  return {
+    modelBaseUri,
+    manifestUri: modelBaseUri === "/" ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`
+  };
+}
+
+// src/dom/loadWeightMap.ts
+async function loadWeightMap(uri, defaultModelName) {
+  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);
+  const manifest = await fetchJson(manifestUri);
+  return tf7["io"].loadWeights(manifest, modelBaseUri);
+}
+
+// src/dom/matchDimensions.ts
+function matchDimensions(input, reference, useMediaDimensions = false) {
+  const { width, height } = useMediaDimensions ? getMediaDimensions(reference) : reference;
+  input.width = width;
+  input.height = height;
+  return { width, height };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var tf15 = __toESM(require_tfjs_esm());
+
+// src/NeuralNetwork.ts
+var tf8 = __toESM(require_tfjs_esm());
+var NeuralNetwork = class {
+  constructor(name) {
+    this._params = void 0;
+    this._paramMappings = [];
+    this._name = name;
+  }
+  get params() {
+    return this._params;
+  }
+  get paramMappings() {
+    return this._paramMappings;
+  }
+  get isLoaded() {
+    return !!this.params;
+  }
+  getParamFromPath(paramPath) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    return obj[objProp];
+  }
+  reassignParamFromPath(paramPath, tensor2) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    obj[objProp].dispose();
+    obj[objProp] = tensor2;
+  }
+  getParamList() {
+    return this._paramMappings.map(({ paramPath }) => ({
+      path: paramPath,
+      tensor: this.getParamFromPath(paramPath)
+    }));
+  }
+  getTrainableParams() {
+    return this.getParamList().filter((param) => param.tensor instanceof tf8.Variable);
+  }
+  getFrozenParams() {
+    return this.getParamList().filter((param) => !(param.tensor instanceof tf8.Variable));
+  }
+  variable() {
+    this.getFrozenParams().forEach(({ path, tensor: tensor2 }) => {
+      this.reassignParamFromPath(path, tensor2.variable());
+    });
+  }
+  freeze() {
+    this.getTrainableParams().forEach(({ path, tensor: variable }) => {
+      const tensor2 = tf8.tensor(variable.dataSync());
+      variable.dispose();
+      this.reassignParamFromPath(path, tensor2);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.getParamList().forEach((param) => {
+      if (throwOnRedispose && param.tensor.isDisposed) {
+        throw new Error(`param tensor has already been disposed for path ${param.path}`);
+      }
+      param.tensor.dispose();
+    });
+    this._params = void 0;
+  }
+  serializeParams() {
+    return new Float32Array(
+      this.getParamList().map(({ tensor: tensor2 }) => Array.from(tensor2.dataSync())).reduce((flat, arr) => flat.concat(arr))
+    );
+  }
+  async load(weightsOrUrl) {
+    if (weightsOrUrl instanceof Float32Array) {
+      this.extractWeights(weightsOrUrl);
+      return;
+    }
+    await this.loadFromUri(weightsOrUrl);
+  }
+  async loadFromUri(uri) {
+    if (uri && typeof uri !== "string") {
+      throw new Error(`${this._name}.loadFromUri - expected model uri`);
+    }
+    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());
+    this.loadFromWeightMap(weightMap);
+  }
+  async loadFromDisk(filePath) {
+    if (filePath && typeof filePath !== "string") {
+      throw new Error(`${this._name}.loadFromDisk - expected model file path`);
+    }
+    const { readFile } = env.getEnv();
+    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());
+    const fetchWeightsFromDisk = (filePaths) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));
+    const loadWeights = tf8["io"].weightsLoaderFactory(fetchWeightsFromDisk);
+    const manifest = JSON.parse((await readFile(manifestUri)).toString());
+    const weightMap = await loadWeights(manifest, modelBaseUri);
+    this.loadFromWeightMap(weightMap);
+  }
+  loadFromWeightMap(weightMap) {
+    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  extractWeights(weights) {
+    const { paramMappings, params } = this.extractParams(weights);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  traversePropertyPath(paramPath) {
+    if (!this.params) {
+      throw new Error("traversePropertyPath - model has no loaded params");
+    }
+    const result = paramPath.split("/").reduce((res, objProp2) => {
+      if (!res.nextObj.hasOwnProperty(objProp2)) {
+        throw new Error(`traversePropertyPath - object does not have property ${objProp2}, for path ${paramPath}`);
+      }
+      return { obj: res.nextObj, objProp: objProp2, nextObj: res.nextObj[objProp2] };
+    }, { nextObj: this.params });
+    const { obj, objProp } = result;
+    if (!obj || !objProp || !(obj[objProp] instanceof tf8.Tensor)) {
+      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);
+    }
+    return { obj, objProp };
+  }
+};
+
+// src/faceFeatureExtractor/denseBlock.ts
+var tf10 = __toESM(require_tfjs_esm());
+
+// src/common/depthwiseSeparableConv.ts
+var tf9 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv(x, params, stride) {
+  return tf9.tidy(() => {
+    let out = tf9.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, "same");
+    out = tf9.add(out, params.bias);
+    return out;
+  });
+}
+
+// src/faceFeatureExtractor/denseBlock.ts
+function denseBlock3(x, denseBlockParams, isFirstLayer = false) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, [2, 2], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, [2, 2])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+  });
+}
+function denseBlock4(x, denseBlockParams, isFirstLayer = false, isScaleDown = true) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, isScaleDown ? [2, 2] : [1, 1], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, isScaleDown ? [2, 2] : [1, 1])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    const in4 = tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, tf10.add(out3, out4))));
+  });
+}
+
+// src/common/convLayer.ts
+var tf11 = __toESM(require_tfjs_esm());
+function convLayer(x, params, padding = "same", withRelu = false) {
+  return tf11.tidy(() => {
+    const out = tf11.add(
+      tf11.conv2d(x, params.filters, [1, 1], padding),
+      params.bias
+    );
+    return withRelu ? tf11.relu(out) : out;
+  });
+}
+
+// src/common/disposeUnusedWeightTensors.ts
+function disposeUnusedWeightTensors(weightMap, paramMappings) {
+  Object.keys(weightMap).forEach((path) => {
+    if (!paramMappings.some((pm) => pm.originalPath === path)) {
+      weightMap[path].dispose();
+    }
+  });
+}
+
+// src/common/extractConvParamsFactory.ts
+var tf12 = __toESM(require_tfjs_esm());
+function extractConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, filterSize, mappedPrefix) => {
+    const filters = tf12.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf12.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  };
+}
+
+// src/common/extractFCParamsFactory.ts
+var tf13 = __toESM(require_tfjs_esm());
+function extractFCParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const fc_weights = tf13.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);
+    const fc_bias = tf13.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return {
+      weights: fc_weights,
+      bias: fc_bias
+    };
+  };
+}
+
+// src/common/extractSeparableConvParamsFactory.ts
+var tf14 = __toESM(require_tfjs_esm());
+
+// src/common/types.ts
+var SeparableConvParams = class {
+  constructor(depthwise_filter, pointwise_filter, bias) {
+    this.depthwise_filter = depthwise_filter;
+    this.pointwise_filter = pointwise_filter;
+    this.bias = bias;
+  }
+};
+
+// src/common/extractSeparableConvParamsFactory.ts
+function extractSeparableConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const depthwise_filter = tf14.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);
+    const pointwise_filter = tf14.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);
+    const bias = tf14.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/depthwise_filter` },
+      { paramPath: `${mappedPrefix}/pointwise_filter` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+function loadSeparableConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const depthwise_filter = extractWeightEntry(`${prefix}/depthwise_filter`, 4);
+    const pointwise_filter = extractWeightEntry(`${prefix}/pointwise_filter`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+
+// src/common/extractWeightEntryFactory.ts
+function extractWeightEntryFactory(weightMap, paramMappings) {
+  return (originalPath, paramRank, mappedPath) => {
+    const tensor2 = weightMap[originalPath];
+    if (!isTensor(tensor2, paramRank)) {
+      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor2}`);
+    }
+    paramMappings.push(
+      { originalPath, paramPath: mappedPath || originalPath }
+    );
+    return tensor2;
+  };
+}
+
+// src/common/extractWeightsFactory.ts
+function extractWeightsFactory(weights) {
+  let remainingWeights = weights;
+  function extractWeights(numWeights) {
+    const ret = remainingWeights.slice(0, numWeights);
+    remainingWeights = remainingWeights.slice(numWeights);
+    return ret;
+  }
+  function getRemainingWeights() {
+    return remainingWeights;
+  }
+  return {
+    extractWeights,
+    getRemainingWeights
+  };
+}
+
+// src/faceFeatureExtractor/extractorsFactory.ts
+function extractorsFactory(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`) : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);
+    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);
+    const conv22 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const { conv0, conv1, conv2: conv22 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);
+    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParams.ts
+function extractParams(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock4Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock4Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock4Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock4Params(64, 128, "dense2");
+  const dense3 = extractDenseBlock4Params(128, 256, "dense3");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: {
+      dense0,
+      dense1,
+      dense2,
+      dense3
+    }
+  };
+}
+
+// src/common/loadConvParamsFactory.ts
+function loadConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  };
+}
+
+// src/faceFeatureExtractor/loadParamsFactory.ts
+function loadParamsFactory(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractDenseBlock3Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock4Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock4Params("dense0", true),
+    dense1: extractDenseBlock4Params("dense1"),
+    dense2: extractDenseBlock4Params("dense2"),
+    dense3: extractDenseBlock4Params("dense3")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var FaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("FaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceFeatureExtractor - load model before inference");
+    }
+    return tf15.tidy(() => {
+      const batchTensor = tf15.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock4(normalized, params.dense0, true);
+      out = denseBlock4(out, params.dense1);
+      out = denseBlock4(out, params.dense2);
+      out = denseBlock4(out, params.dense3);
+      out = tf15.avgPool(out, [7, 7], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams(weights);
+  }
+};
+
+// src/faceProcessor/FaceProcessor.ts
+var tf17 = __toESM(require_tfjs_esm());
+
+// src/common/fullyConnectedLayer.ts
+var tf16 = __toESM(require_tfjs_esm());
+function fullyConnectedLayer(x, params) {
+  return tf16.tidy(() => tf16.add(
+    tf16.matMul(x, params.weights),
+    params.bias
+  ));
+}
+
+// src/faceProcessor/extractParams.ts
+function extractParams2(weights, channelsIn, channelsOut) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const fc = extractFCParams(channelsIn, channelsOut, "fc");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc }
+  };
+}
+
+// src/faceProcessor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap2(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: extractFcParams("fc")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceProcessor/util.ts
+function seperateWeightMaps(weightMap) {
+  const featureExtractorMap = {};
+  const classifierMap = {};
+  Object.keys(weightMap).forEach((key) => {
+    const map = key.startsWith("fc") ? classifierMap : featureExtractorMap;
+    map[key] = weightMap[key];
+  });
+  return { featureExtractorMap, classifierMap };
+}
+
+// src/faceProcessor/FaceProcessor.ts
+var FaceProcessor = class extends NeuralNetwork {
+  constructor(_name, faceFeatureExtractor) {
+    super(_name);
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf17.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams2(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap2(classifierMap);
+  }
+  extractParams(weights) {
+    const cIn = this.getClassifierChannelsIn();
+    const cOut = this.getClassifierChannelsOut();
+    const classifierWeightSize = cOut * cIn + cOut;
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressions.ts
+var FACE_EXPRESSION_LABELS = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"];
+var FaceExpressions = class {
+  constructor(probabilities) {
+    this.neutral = 0;
+    this.happy = 0;
+    this.sad = 0;
+    this.angry = 0;
+    this.fearful = 0;
+    this.disgusted = 0;
+    this.surprised = 0;
+    if (probabilities.length !== 7) {
+      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);
+    }
+    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {
+      this[expression] = probabilities[idx];
+    });
+  }
+  asSortedArray() {
+    return FACE_EXPRESSION_LABELS.map((expression) => ({ expression, probability: this[expression] })).sort((e0, e1) => e1.probability - e0.probability);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var FaceExpressionNet = class extends FaceProcessor {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceExpressionNet", faceFeatureExtractor);
+  }
+  forwardInput(input) {
+    return tf18.tidy(() => tf18.softmax(this.runNet(input)));
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictExpressions(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const probabilitesByBatch = await Promise.all(tf18.unstack(out).map(async (t) => {
+      const data = t.dataSync();
+      t.dispose();
+      return data;
+    }));
+    out.dispose();
+    const predictionsByBatch = probabilitesByBatch.map((probabilites) => new FaceExpressions(probabilites));
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_expression_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+  getClassifierChannelsOut() {
+    return 7;
+  }
+};
+
+// src/factories/WithFaceExpressions.ts
+function isWithFaceExpressions(obj) {
+  return obj.expressions instanceof FaceExpressions;
+}
+function extendWithFaceExpressions(sourceObj, expressions) {
+  const extension = { expressions };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/drawFaceExpressions.ts
+function drawFaceExpressions(canvasArg, faceExpressions, minConfidence = 0.1, textFieldAnchor) {
+  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];
+  faceExpressionsArray.forEach((e) => {
+    const expr = e instanceof FaceExpressions ? e : isWithFaceExpressions(e) ? e.expressions : void 0;
+    if (!expr) {
+      throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+    }
+    const sorted = expr.asSortedArray();
+    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);
+    const anchor = isWithFaceDetection(e) ? e.detection.box.bottomLeft : textFieldAnchor || new Point(0, 0);
+    const drawTextField = new DrawTextField(
+      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),
+      anchor
+    );
+    drawTextField.draw(canvasArg);
+  });
+}
+
+// src/factories/WithFaceLandmarks.ts
+function isWithFaceLandmarks(obj) {
+  return isWithFaceDetection(obj) && obj["landmarks"] instanceof FaceLandmarks && obj["unshiftedLandmarks"] instanceof FaceLandmarks && obj["alignedRect"] instanceof FaceDetection;
+}
+function calculateFaceAngle(mesh) {
+  const degrees = (radians) => radians * 180 / Math.PI;
+  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);
+  const angle = {
+    roll: void 0,
+    pitch: void 0,
+    yaw: void 0
+  };
+  const calcYaw = (leftPoint, midPoint, rightPoint) => {
+    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);
+    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);
+    return leftToMidpoint - rightToMidpoint;
+  };
+  const calcRoll = (lever, pivot) => {
+    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);
+    const opposite = pivot._y - lever._y;
+    const angleInRadians = Math.asin(opposite / hypotenuse);
+    const angleInDegrees = degrees(angleInRadians);
+    const normalizeAngle = Math.floor(90 - angleInDegrees);
+    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;
+    const result = normalizeAngle * tiltDirection;
+    return result;
+  };
+  const calcPitch = (leftPoint, midPoint, rightPoint) => {
+    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);
+    const baseCoords = {
+      _x: (leftPoint._x + rightPoint._x) / 2,
+      _y: (leftPoint._y + rightPoint._y) / 2
+    };
+    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);
+    const angleInRadians = Math.atan(midToBaseLength / base);
+    const angleInDegrees = Math.floor(degrees(angleInRadians));
+    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;
+    const result = angleInDegrees * direction;
+    return result;
+  };
+  if (!mesh || !mesh._positions || mesh._positions.length !== 68)
+    return angle;
+  const pt = mesh._positions;
+  angle.roll = calcRoll(pt[27], pt[66]);
+  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);
+  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);
+  return angle;
+}
+function extendWithFaceLandmarks(sourceObj, unshiftedLandmarks) {
+  const { box: shift } = sourceObj.detection;
+  const landmarks = unshiftedLandmarks.shiftBy(shift.x, shift.y);
+  const rect = landmarks.align();
+  const { imageDims } = sourceObj.detection;
+  const alignedRect = new FaceDetection(
+    sourceObj.detection.score,
+    rect.rescale(imageDims.reverse()),
+    imageDims
+  );
+  const angle = calculateFaceAngle(unshiftedLandmarks);
+  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/DrawFaceLandmarks.ts
+var DrawFaceLandmarksOptions = class {
+  constructor(options = {}) {
+    const {
+      drawLines = true,
+      drawPoints = true,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = options;
+    this.drawLines = drawLines;
+    this.drawPoints = drawPoints;
+    this.lineWidth = lineWidth || 1;
+    this.pointSize = pointSize || 2;
+    this.lineColor = lineColor || "rgba(0, 255, 255, 1)";
+    this.pointColor = pointColor || "rgba(255, 0, 255, 1)";
+  }
+};
+var DrawFaceLandmarks = class {
+  constructor(faceLandmarks, options = {}) {
+    this.faceLandmarks = faceLandmarks;
+    this.options = new DrawFaceLandmarksOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const {
+      drawLines,
+      drawPoints,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = this.options;
+    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {
+      ctx.strokeStyle = lineColor;
+      ctx.lineWidth = lineWidth;
+      drawContour(ctx, this.faceLandmarks.getJawOutline());
+      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getNose());
+      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);
+      drawContour(ctx, this.faceLandmarks.getRightEye(), true);
+      drawContour(ctx, this.faceLandmarks.getMouth(), true);
+    }
+    if (drawPoints) {
+      ctx.strokeStyle = pointColor;
+      ctx.fillStyle = pointColor;
+      const drawPoint = (pt) => {
+        ctx.beginPath();
+        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);
+        ctx.fill();
+      };
+      this.faceLandmarks.positions.forEach(drawPoint);
+    }
+  }
+};
+function drawFaceLandmarks(canvasArg, faceLandmarks) {
+  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];
+  faceLandmarksArray.forEach((f) => {
+    const landmarks = f instanceof FaceLandmarks ? f : isWithFaceLandmarks(f) ? f.landmarks : void 0;
+    if (!landmarks) {
+      throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+    }
+    new DrawFaceLandmarks(landmarks).draw(canvasArg);
+  });
+}
+
+// package.json
+var version = "1.7.6";
+
+// src/ageGenderNet/AgeGenderNet.ts
+var tf20 = __toESM(require_tfjs_esm());
+
+// src/xception/TinyXception.ts
+var tf19 = __toESM(require_tfjs_esm());
+
+// src/xception/extractParams.ts
+function extractorsFactory2(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractReductionBlockParams(channelsIn, channelsOut, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(channels, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParams3(weights, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = extractorsFactory2(extractWeights, paramMappings);
+  const entry_flow_conv_in = extractConvParams(3, 32, 3, "entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, "entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, "entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, "exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, "exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { entry_flow, middle_flow, exit_flow }
+  };
+}
+
+// src/xception/extractParamsFromWeightMap.ts
+function loadParamsFactory2(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractReductionBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParamsFromWeightMap3(weightMap, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = loadParamsFactory2(weightMap, paramMappings);
+  const entry_flow_conv_in = extractConvParams("entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams("entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams("entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams("exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams("exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };
+}
+
+// src/xception/TinyXception.ts
+function conv(x, params, stride) {
+  return tf19.add(tf19.conv2d(x, params.filters, stride, "same"), params.bias);
+}
+function reductionBlock(x, params, isActivateInput = true) {
+  let out = isActivateInput ? tf19.relu(x) : x;
+  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = tf19.maxPool(out, [3, 3], [2, 2], "same");
+  out = tf19.add(out, conv(x, params.expansion_conv, [2, 2]));
+  return out;
+}
+function mainBlock(x, params) {
+  let out = depthwiseSeparableConv(tf19.relu(x), params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv2, [1, 1]);
+  out = tf19.add(out, x);
+  return out;
+}
+var TinyXception = class extends NeuralNetwork {
+  constructor(numMainBlocks) {
+    super("TinyXception");
+    this._numMainBlocks = numMainBlocks;
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyXception - load model before inference");
+    }
+    return tf19.tidy(() => {
+      const batchTensor = tf19.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = tf19.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));
+      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);
+      out = reductionBlock(out, params.entry_flow.reduction_block_1);
+      range(this._numMainBlocks, 0, 1).forEach((idx) => {
+        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);
+      });
+      out = reductionBlock(out, params.exit_flow.reduction_block);
+      out = tf19.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "tiny_xception_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap3(weightMap, this._numMainBlocks);
+  }
+  extractParams(weights) {
+    return extractParams3(weights, this._numMainBlocks);
+  }
+};
+
+// src/ageGenderNet/extractParams.ts
+function extractParams4(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const age = extractFCParams(512, 1, "fc/age");
+  const gender = extractFCParams(512, 2, "fc/gender");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc: { age, gender } }
+  };
+}
+
+// src/ageGenderNet/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap4(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: {
+      age: extractFcParams("fc/age"),
+      gender: extractFcParams("fc/gender")
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ageGenderNet/types.ts
+var Gender = /* @__PURE__ */ ((Gender2) => {
+  Gender2["FEMALE"] = "female";
+  Gender2["MALE"] = "male";
+  return Gender2;
+})(Gender || {});
+
+// src/ageGenderNet/AgeGenderNet.ts
+var AgeGenderNet = class extends NeuralNetwork {
+  constructor(faceFeatureExtractor = new TinyXception(2)) {
+    super("AgeGenderNet");
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf20.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      const pooled = tf20.avgPool(bottleneckFeatures, [7, 7], [2, 2], "valid").as2D(bottleneckFeatures.shape[0], -1);
+      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();
+      const gender = fullyConnectedLayer(pooled, params.fc.gender);
+      return { age, gender };
+    });
+  }
+  forwardInput(input) {
+    return tf20.tidy(() => {
+      const { age, gender } = this.runNet(input);
+      return { age, gender: tf20.softmax(gender) };
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictAgeAndGender(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const ages = tf20.unstack(out.age);
+    const genders = tf20.unstack(out.gender);
+    const ageAndGenderTensors = ages.map((ageTensor, i) => ({
+      ageTensor,
+      genderTensor: genders[i]
+    }));
+    const predictionsByBatch = await Promise.all(
+      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {
+        const age = ageTensor.dataSync()[0];
+        const probMale = genderTensor.dataSync()[0];
+        const isMale = probMale > 0.5;
+        const gender = isMale ? "male" /* MALE */ : "female" /* FEMALE */;
+        const genderProbability = isMale ? probMale : 1 - probMale;
+        ageTensor.dispose();
+        genderTensor.dispose();
+        return { age, gender, genderProbability };
+      })
+    );
+    out.age.dispose();
+    out.gender.dispose();
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "age_gender_model";
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams4(weights);
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap4(classifierMap);
+  }
+  extractParams(weights) {
+    const classifierWeightSize = 512 * 1 + 1 + (512 * 2 + 2);
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68NetBase.ts
+var tf21 = __toESM(require_tfjs_esm());
+var FaceLandmark68NetBase = class extends FaceProcessor {
+  postProcess(output, inputSize, originalDimensions) {
+    const inputDimensions = originalDimensions.map(({ width, height }) => {
+      const scale2 = inputSize / Math.max(height, width);
+      return {
+        width: width * scale2,
+        height: height * scale2
+      };
+    });
+    const batchSize = inputDimensions.length;
+    return tf21.tidy(() => {
+      const createInterleavedTensor = (fillX, fillY) => tf21.stack([tf21.fill([68], fillX, "float32"), tf21.fill([68], fillY, "float32")], 1).as2D(1, 136).as1D();
+      const getPadding = (batchIdx, cond) => {
+        const { width, height } = inputDimensions[batchIdx];
+        return cond(width, height) ? Math.abs(width - height) / 2 : 0;
+      };
+      const getPaddingX = (batchIdx) => getPadding(batchIdx, (w, h) => w < h);
+      const getPaddingY = (batchIdx) => getPadding(batchIdx, (w, h) => h < w);
+      const landmarkTensors = output.mul(tf21.fill([batchSize, 136], inputSize, "float32")).sub(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        getPaddingX(batchIdx),
+        getPaddingY(batchIdx)
+      )))).div(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        inputDimensions[batchIdx].width,
+        inputDimensions[batchIdx].height
+      ))));
+      return landmarkTensors;
+    });
+  }
+  forwardInput(input) {
+    return tf21.tidy(() => {
+      const out = this.runNet(input);
+      return this.postProcess(
+        out,
+        input.inputSize,
+        input.inputDimensions.map(([height, width]) => ({ height, width }))
+      );
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async detectLandmarks(input) {
+    const netInput = await toNetInput(input);
+    const landmarkTensors = tf21.tidy(
+      () => tf21.unstack(this.forwardInput(netInput))
+    );
+    const landmarksForBatch = await Promise.all(landmarkTensors.map(
+      async (landmarkTensor, batchIdx) => {
+        const landmarksArray = Array.from(landmarkTensor.dataSync());
+        const xCoords = landmarksArray.filter((_, i) => isEven(i));
+        const yCoords = landmarksArray.filter((_, i) => !isEven(i));
+        return new FaceLandmarks68(
+          Array(68).fill(0).map((_, i) => new Point(xCoords[i], yCoords[i])),
+          {
+            height: netInput.getInputHeight(batchIdx),
+            width: netInput.getInputWidth(batchIdx)
+          }
+        );
+      }
+    ));
+    landmarkTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? landmarksForBatch : landmarksForBatch[0];
+  }
+  getClassifierChannelsOut() {
+    return 136;
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68Net.ts
+var FaceLandmark68Net = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceLandmark68Net", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+};
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var tf22 = __toESM(require_tfjs_esm());
+
+// src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts
+function extractParamsFromWeightMapTiny(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock3Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock3Params("dense0", true),
+    dense1: extractDenseBlock3Params("dense1"),
+    dense2: extractDenseBlock3Params("dense2")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/extractParamsTiny.ts
+function extractParamsTiny(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock3Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock3Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock3Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock3Params(64, 128, "dense2");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { dense0, dense1, dense2 }
+  };
+}
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var TinyFaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("TinyFaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyFaceFeatureExtractor - load model before inference");
+    }
+    return tf22.tidy(() => {
+      const batchTensor = tf22.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock3(normalized, params.dense0, true);
+      out = denseBlock3(out, params.dense1);
+      out = denseBlock3(out, params.dense2);
+      out = tf22.avgPool(out, [14, 14], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_tiny_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMapTiny(weightMap);
+  }
+  extractParams(weights) {
+    return extractParamsTiny(weights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68TinyNet.ts
+var FaceLandmark68TinyNet = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new TinyFaceFeatureExtractor()) {
+    super("FaceLandmark68TinyNet", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_tiny_model";
+  }
+  getClassifierChannelsIn() {
+    return 128;
+  }
+};
+
+// src/faceLandmarkNet/index.ts
+var FaceLandmarkNet = class extends FaceLandmark68Net {
+};
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var tf27 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/convLayer.ts
+var tf24 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/scaleLayer.ts
+var tf23 = __toESM(require_tfjs_esm());
+function scale(x, params) {
+  return tf23.add(tf23.mul(x, params.weights), params.biases);
+}
+
+// src/faceRecognitionNet/convLayer.ts
+function convLayer2(x, params, strides, withRelu, padding = "same") {
+  const { filters, bias } = params.conv;
+  let out = tf24.conv2d(x, filters, strides, padding);
+  out = tf24.add(out, bias);
+  out = scale(out, params.scale);
+  return withRelu ? tf24.relu(out) : out;
+}
+function conv2(x, params) {
+  return convLayer2(x, params, [1, 1], true);
+}
+function convNoRelu(x, params) {
+  return convLayer2(x, params, [1, 1], false);
+}
+function convDown(x, params) {
+  return convLayer2(x, params, [2, 2], true, "valid");
+}
+
+// src/faceRecognitionNet/extractParams.ts
+var tf25 = __toESM(require_tfjs_esm());
+function extractorsFactory3(extractWeights, paramMappings) {
+  function extractFilterValues(numFilterValues, numFilters, filterSize) {
+    const weights = extractWeights(numFilterValues);
+    const depth = weights.length / (numFilters * filterSize * filterSize);
+    if (isFloat(depth)) {
+      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);
+    }
+    return tf25.tidy(
+      () => tf25.transpose(
+        tf25.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),
+        [2, 3, 1, 0]
+      )
+    );
+  }
+  function extractConvParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);
+    const bias = tf25.tensor1d(extractWeights(numFilters));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  }
+  function extractScaleLayerParams(numWeights, mappedPrefix) {
+    const weights = tf25.tensor1d(extractWeights(numWeights));
+    const biases = tf25.tensor1d(extractWeights(numWeights));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/biases` }
+    );
+    return {
+      weights,
+      biases
+    };
+  }
+  function extractConvLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const conv3 = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);
+    const scale2 = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);
+    return { conv: conv3, scale: scale2 };
+  }
+  function extractResidualLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix, isDown = false) {
+    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);
+    const conv22 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);
+    return { conv1, conv2: conv22 };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParams5(weights) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory3(extractWeights, paramMappings);
+  const conv32_down = extractConvLayerParams(4704, 32, 7, "conv32_down");
+  const conv32_1 = extractResidualLayerParams(9216, 32, 3, "conv32_1");
+  const conv32_2 = extractResidualLayerParams(9216, 32, 3, "conv32_2");
+  const conv32_3 = extractResidualLayerParams(9216, 32, 3, "conv32_3");
+  const conv64_down = extractResidualLayerParams(36864, 64, 3, "conv64_down", true);
+  const conv64_1 = extractResidualLayerParams(36864, 64, 3, "conv64_1");
+  const conv64_2 = extractResidualLayerParams(36864, 64, 3, "conv64_2");
+  const conv64_3 = extractResidualLayerParams(36864, 64, 3, "conv64_3");
+  const conv128_down = extractResidualLayerParams(147456, 128, 3, "conv128_down", true);
+  const conv128_1 = extractResidualLayerParams(147456, 128, 3, "conv128_1");
+  const conv128_2 = extractResidualLayerParams(147456, 128, 3, "conv128_2");
+  const conv256_down = extractResidualLayerParams(589824, 256, 3, "conv256_down", true);
+  const conv256_1 = extractResidualLayerParams(589824, 256, 3, "conv256_1");
+  const conv256_2 = extractResidualLayerParams(589824, 256, 3, "conv256_2");
+  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, "conv256_down_out");
+  const fc = tf25.tidy(
+    () => tf25.transpose(tf25.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0])
+  );
+  paramMappings.push({ paramPath: "fc" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/extractParamsFromWeightMap.ts
+function extractorsFactory4(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractScaleLayerParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);
+    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);
+    return { weights, biases };
+  }
+  function extractConvLayerParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);
+    const scale2 = extractScaleLayerParams(prefix);
+    return { conv: { filters, bias }, scale: scale2 };
+  }
+  function extractResidualLayerParams(prefix) {
+    return {
+      conv1: extractConvLayerParams(`${prefix}/conv1`),
+      conv2: extractConvLayerParams(`${prefix}/conv2`)
+    };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParamsFromWeightMap5(weightMap) {
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory4(weightMap, paramMappings);
+  const conv32_down = extractConvLayerParams("conv32_down");
+  const conv32_1 = extractResidualLayerParams("conv32_1");
+  const conv32_2 = extractResidualLayerParams("conv32_2");
+  const conv32_3 = extractResidualLayerParams("conv32_3");
+  const conv64_down = extractResidualLayerParams("conv64_down");
+  const conv64_1 = extractResidualLayerParams("conv64_1");
+  const conv64_2 = extractResidualLayerParams("conv64_2");
+  const conv64_3 = extractResidualLayerParams("conv64_3");
+  const conv128_down = extractResidualLayerParams("conv128_down");
+  const conv128_1 = extractResidualLayerParams("conv128_1");
+  const conv128_2 = extractResidualLayerParams("conv128_2");
+  const conv256_down = extractResidualLayerParams("conv256_down");
+  const conv256_1 = extractResidualLayerParams("conv256_1");
+  const conv256_2 = extractResidualLayerParams("conv256_2");
+  const conv256_down_out = extractResidualLayerParams("conv256_down_out");
+  const { fc } = weightMap;
+  paramMappings.push({ originalPath: "fc", paramPath: "fc" });
+  if (!isTensor2D(fc)) {
+    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/residualLayer.ts
+var tf26 = __toESM(require_tfjs_esm());
+function residual(x, params) {
+  let out = conv2(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  out = tf26.add(out, x);
+  out = tf26.relu(out);
+  return out;
+}
+function residualDown(x, params) {
+  let out = convDown(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  let pooled = tf26.avgPool(x, 2, 2, "valid");
+  const zeros2 = tf26.zeros(pooled.shape);
+  const isPad = pooled.shape[3] !== out.shape[3];
+  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];
+  if (isAdjustShape) {
+    const padShapeX = [...out.shape];
+    padShapeX[1] = 1;
+    const zerosW = tf26.zeros(padShapeX);
+    out = tf26.concat([out, zerosW], 1);
+    const padShapeY = [...out.shape];
+    padShapeY[2] = 1;
+    const zerosH = tf26.zeros(padShapeY);
+    out = tf26.concat([out, zerosH], 2);
+  }
+  pooled = isPad ? tf26.concat([pooled, zeros2], 3) : pooled;
+  out = tf26.add(pooled, out);
+  out = tf26.relu(out);
+  return out;
+}
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var FaceRecognitionNet = class extends NeuralNetwork {
+  constructor() {
+    super("FaceRecognitionNet");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceRecognitionNet - load model before inference");
+    }
+    return tf27.tidy(() => {
+      const batchTensor = tf27.cast(input.toBatchTensor(150, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = convDown(normalized, params.conv32_down);
+      out = tf27.maxPool(out, 3, 2, "valid");
+      out = residual(out, params.conv32_1);
+      out = residual(out, params.conv32_2);
+      out = residual(out, params.conv32_3);
+      out = residualDown(out, params.conv64_down);
+      out = residual(out, params.conv64_1);
+      out = residual(out, params.conv64_2);
+      out = residual(out, params.conv64_3);
+      out = residualDown(out, params.conv128_down);
+      out = residual(out, params.conv128_1);
+      out = residual(out, params.conv128_2);
+      out = residualDown(out, params.conv256_down);
+      out = residual(out, params.conv256_1);
+      out = residual(out, params.conv256_2);
+      out = residualDown(out, params.conv256_down_out);
+      const globalAvg = out.mean([1, 2]);
+      const fullyConnected = tf27.matMul(globalAvg, params.fc);
+      return fullyConnected;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async computeFaceDescriptor(input) {
+    var _a;
+    if ((_a = input == null ? void 0 : input.shape) == null ? void 0 : _a.some((dim) => dim <= 0))
+      return new Float32Array(128);
+    const netInput = await toNetInput(input);
+    const faceDescriptorTensors = tf27.tidy(() => tf27.unstack(this.forwardInput(netInput)));
+    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data()));
+    faceDescriptorTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_recognition_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap5(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams5(weights);
+  }
+};
+
+// src/faceRecognitionNet/index.ts
+function createFaceRecognitionNet(weights) {
+  const net = new FaceRecognitionNet();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/factories/WithFaceDescriptor.ts
+function extendWithFaceDescriptor(sourceObj, descriptor) {
+  const extension = { descriptor };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithAge.ts
+function isWithAge(obj) {
+  return typeof obj.age === "number";
+}
+function extendWithAge(sourceObj, age) {
+  const extension = { age };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithGender.ts
+function isWithGender(obj) {
+  return (obj.gender === "male" /* MALE */ || obj.gender === "female" /* FEMALE */) && isValidProbablitiy(obj.genderProbability);
+}
+function extendWithGender(sourceObj, gender, genderProbability) {
+  const extension = { gender, genderProbability };
+  return { ...sourceObj, ...extension };
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var tf34 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/extractParams.ts
+var tf28 = __toESM(require_tfjs_esm());
+function extractorsFactory5(extractWeights, paramMappings) {
+  function extractDepthwiseConvParams(numChannels, mappedPrefix) {
+    const filters = tf28.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);
+    const batch_norm_scale = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_offset = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_mean = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_variance = tf28.tensor1d(extractWeights(numChannels));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/batch_norm_scale` },
+      { paramPath: `${mappedPrefix}/batch_norm_offset` },
+      { paramPath: `${mappedPrefix}/batch_norm_mean` },
+      { paramPath: `${mappedPrefix}/batch_norm_variance` }
+    );
+    return {
+      filters,
+      batch_norm_scale,
+      batch_norm_offset,
+      batch_norm_mean,
+      batch_norm_variance
+    };
+  }
+  function extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, isPointwiseConv) {
+    const filters = tf28.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf28.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/${isPointwiseConv ? "batch_norm_offset" : "bias"}` }
+    );
+    return { filters, bias };
+  }
+  function extractPointwiseConvParams(channelsIn, channelsOut, filterSize, mappedPrefix) {
+    const {
+      filters,
+      bias
+    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);
+    return {
+      filters,
+      batch_norm_offset: bias
+    };
+  }
+  function extractConvPairParams(channelsIn, channelsOut, mappedPrefix) {
+    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);
+    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);
+    return { depthwise_conv, pointwise_conv };
+  }
+  function extractMobilenetV1Params() {
+    const conv_0 = extractPointwiseConvParams(3, 32, 3, "mobilenetv1/conv_0");
+    const conv_1 = extractConvPairParams(32, 64, "mobilenetv1/conv_1");
+    const conv_2 = extractConvPairParams(64, 128, "mobilenetv1/conv_2");
+    const conv_3 = extractConvPairParams(128, 128, "mobilenetv1/conv_3");
+    const conv_4 = extractConvPairParams(128, 256, "mobilenetv1/conv_4");
+    const conv_5 = extractConvPairParams(256, 256, "mobilenetv1/conv_5");
+    const conv_6 = extractConvPairParams(256, 512, "mobilenetv1/conv_6");
+    const conv_7 = extractConvPairParams(512, 512, "mobilenetv1/conv_7");
+    const conv_8 = extractConvPairParams(512, 512, "mobilenetv1/conv_8");
+    const conv_9 = extractConvPairParams(512, 512, "mobilenetv1/conv_9");
+    const conv_10 = extractConvPairParams(512, 512, "mobilenetv1/conv_10");
+    const conv_11 = extractConvPairParams(512, 512, "mobilenetv1/conv_11");
+    const conv_12 = extractConvPairParams(512, 1024, "mobilenetv1/conv_12");
+    const conv_13 = extractConvPairParams(1024, 1024, "mobilenetv1/conv_13");
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      conv_8,
+      conv_9,
+      conv_10,
+      conv_11,
+      conv_12,
+      conv_13
+    };
+  }
+  function extractPredictionLayerParams() {
+    const conv_0 = extractPointwiseConvParams(1024, 256, 1, "prediction_layer/conv_0");
+    const conv_1 = extractPointwiseConvParams(256, 512, 3, "prediction_layer/conv_1");
+    const conv_2 = extractPointwiseConvParams(512, 128, 1, "prediction_layer/conv_2");
+    const conv_3 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_3");
+    const conv_4 = extractPointwiseConvParams(256, 128, 1, "prediction_layer/conv_4");
+    const conv_5 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_5");
+    const conv_6 = extractPointwiseConvParams(256, 64, 1, "prediction_layer/conv_6");
+    const conv_7 = extractPointwiseConvParams(64, 128, 3, "prediction_layer/conv_7");
+    const box_encoding_0_predictor = extractConvParams(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor");
+    const class_predictor_0 = extractConvParams(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor");
+    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor");
+    const class_predictor_1 = extractConvParams(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor");
+    const box_encoding_2_predictor = extractConvParams(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor");
+    const class_predictor_2 = extractConvParams(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor");
+    const box_encoding_3_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor");
+    const class_predictor_3 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor");
+    const box_encoding_4_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor");
+    const class_predictor_4 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor");
+    const box_encoding_5_predictor = extractConvParams(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor");
+    const class_predictor_5 = extractConvParams(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor");
+    const box_predictor_0 = {
+      box_encoding_predictor: box_encoding_0_predictor,
+      class_predictor: class_predictor_0
+    };
+    const box_predictor_1 = {
+      box_encoding_predictor: box_encoding_1_predictor,
+      class_predictor: class_predictor_1
+    };
+    const box_predictor_2 = {
+      box_encoding_predictor: box_encoding_2_predictor,
+      class_predictor: class_predictor_2
+    };
+    const box_predictor_3 = {
+      box_encoding_predictor: box_encoding_3_predictor,
+      class_predictor: class_predictor_3
+    };
+    const box_predictor_4 = {
+      box_encoding_predictor: box_encoding_4_predictor,
+      class_predictor: class_predictor_4
+    };
+    const box_predictor_5 = {
+      box_encoding_predictor: box_encoding_5_predictor,
+      class_predictor: class_predictor_5
+    };
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      box_predictor_0,
+      box_predictor_1,
+      box_predictor_2,
+      box_predictor_3,
+      box_predictor_4,
+      box_predictor_5
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParams6(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory5(extractWeights, paramMappings);
+  const mobilenetv1 = extractMobilenetV1Params();
+  const prediction_layer = extractPredictionLayerParams();
+  const extra_dim = tf28.tensor3d(
+    extractWeights(5118 * 4),
+    [1, 5118, 4]
+  );
+  const output_layer = {
+    extra_dim
+  };
+  paramMappings.push({ paramPath: "output_layer/extra_dim" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    params: {
+      mobilenetv1,
+      prediction_layer,
+      output_layer
+    },
+    paramMappings
+  };
+}
+
+// src/ssdMobilenetv1/extractParamsFromWeightMap.ts
+function extractorsFactory6(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractPointwiseConvParams(prefix, idx, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);
+    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);
+    return { filters, batch_norm_offset };
+  }
+  function extractConvPairParams(idx) {
+    const mappedPrefix = `mobilenetv1/conv_${idx}`;
+    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;
+    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;
+    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;
+    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);
+    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);
+    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);
+    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);
+    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);
+    return {
+      depthwise_conv: {
+        filters,
+        batch_norm_scale,
+        batch_norm_offset,
+        batch_norm_mean,
+        batch_norm_variance
+      },
+      pointwise_conv: extractPointwiseConvParams("MobilenetV1", idx, mappedPrefixPointwiseConv)
+    };
+  }
+  function extractMobilenetV1Params() {
+    return {
+      conv_0: extractPointwiseConvParams("MobilenetV1", 0, "mobilenetv1/conv_0"),
+      conv_1: extractConvPairParams(1),
+      conv_2: extractConvPairParams(2),
+      conv_3: extractConvPairParams(3),
+      conv_4: extractConvPairParams(4),
+      conv_5: extractConvPairParams(5),
+      conv_6: extractConvPairParams(6),
+      conv_7: extractConvPairParams(7),
+      conv_8: extractConvPairParams(8),
+      conv_9: extractConvPairParams(9),
+      conv_10: extractConvPairParams(10),
+      conv_11: extractConvPairParams(11),
+      conv_12: extractConvPairParams(12),
+      conv_13: extractConvPairParams(13)
+    };
+  }
+  function extractConvParams(prefix, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);
+    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);
+    return { filters, bias };
+  }
+  function extractBoxPredictorParams(idx) {
+    const box_encoding_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,
+      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`
+    );
+    const class_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/ClassPredictor`,
+      `prediction_layer/box_predictor_${idx}/class_predictor`
+    );
+    return { box_encoding_predictor, class_predictor };
+  }
+  function extractPredictionLayerParams() {
+    return {
+      conv_0: extractPointwiseConvParams("Prediction", 0, "prediction_layer/conv_0"),
+      conv_1: extractPointwiseConvParams("Prediction", 1, "prediction_layer/conv_1"),
+      conv_2: extractPointwiseConvParams("Prediction", 2, "prediction_layer/conv_2"),
+      conv_3: extractPointwiseConvParams("Prediction", 3, "prediction_layer/conv_3"),
+      conv_4: extractPointwiseConvParams("Prediction", 4, "prediction_layer/conv_4"),
+      conv_5: extractPointwiseConvParams("Prediction", 5, "prediction_layer/conv_5"),
+      conv_6: extractPointwiseConvParams("Prediction", 6, "prediction_layer/conv_6"),
+      conv_7: extractPointwiseConvParams("Prediction", 7, "prediction_layer/conv_7"),
+      box_predictor_0: extractBoxPredictorParams(0),
+      box_predictor_1: extractBoxPredictorParams(1),
+      box_predictor_2: extractBoxPredictorParams(2),
+      box_predictor_3: extractBoxPredictorParams(3),
+      box_predictor_4: extractBoxPredictorParams(4),
+      box_predictor_5: extractBoxPredictorParams(5)
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParamsFromWeightMap6(weightMap) {
+  const paramMappings = [];
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory6(weightMap, paramMappings);
+  const extra_dim = weightMap["Output/extra_dim"];
+  paramMappings.push({ originalPath: "Output/extra_dim", paramPath: "output_layer/extra_dim" });
+  if (!isTensor3D(extra_dim)) {
+    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);
+  }
+  const params = {
+    mobilenetv1: extractMobilenetV1Params(),
+    prediction_layer: extractPredictionLayerParams(),
+    output_layer: {
+      extra_dim
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var tf30 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/pointwiseConvLayer.ts
+var tf29 = __toESM(require_tfjs_esm());
+function pointwiseConvLayer(x, params, strides) {
+  return tf29.tidy(() => {
+    let out = tf29.conv2d(x, params.filters, strides, "same");
+    out = tf29.add(out, params.batch_norm_offset);
+    return tf29.clipByValue(out, 0, 6);
+  });
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var epsilon = 0.0010000000474974513;
+function depthwiseConvLayer(x, params, strides) {
+  return tf30.tidy(() => {
+    let out = tf30.depthwiseConv2d(x, params.filters, strides, "same");
+    out = tf30.batchNorm(
+      out,
+      params.batch_norm_mean,
+      params.batch_norm_variance,
+      params.batch_norm_offset,
+      params.batch_norm_scale,
+      epsilon
+    );
+    return tf30.clipByValue(out, 0, 6);
+  });
+}
+function getStridesForLayerIdx(layerIdx) {
+  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];
+}
+function mobileNetV1(x, params) {
+  return tf30.tidy(() => {
+    let conv11;
+    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
+    const convPairParams = [
+      params.conv_1,
+      params.conv_2,
+      params.conv_3,
+      params.conv_4,
+      params.conv_5,
+      params.conv_6,
+      params.conv_7,
+      params.conv_8,
+      params.conv_9,
+      params.conv_10,
+      params.conv_11,
+      params.conv_12,
+      params.conv_13
+    ];
+    convPairParams.forEach((param, i) => {
+      const layerIdx = i + 1;
+      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
+      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
+      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
+      if (layerIdx === 11)
+        conv11 = out;
+    });
+    if (conv11 === null) {
+      throw new Error("mobileNetV1 - output of conv layer 11 is null");
+    }
+    return {
+      out,
+      conv11
+    };
+  });
+}
+
+// src/ssdMobilenetv1/nonMaxSuppression.ts
+function IOU(boxes, i, j) {
+  const boxesData = boxes.arraySync();
+  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);
+  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);
+  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);
+  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);
+  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);
+  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);
+  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);
+  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0)
+    return 0;
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function nonMaxSuppression2(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  const numBoxes = boxes.shape[0];
+  const outputSize = Math.min(maxOutputSize, numBoxes);
+  const candidates = scores.map((score, boxIndex) => ({ score, boxIndex })).filter((c) => c.score > scoreThreshold).sort((c1, c2) => c2.score - c1.score);
+  const suppressFunc = (x) => x <= iouThreshold ? 1 : 0;
+  const selected = [];
+  candidates.forEach((c) => {
+    if (selected.length >= outputSize)
+      return;
+    const originalScore = c.score;
+    for (let j = selected.length - 1; j >= 0; --j) {
+      const iou2 = IOU(boxes, c.boxIndex, selected[j]);
+      if (iou2 === 0)
+        continue;
+      c.score *= suppressFunc(iou2);
+      if (c.score <= scoreThreshold)
+        break;
+    }
+    if (originalScore === c.score) {
+      selected.push(c.boxIndex);
+    }
+  });
+  return selected;
+}
+
+// src/ssdMobilenetv1/outputLayer.ts
+var tf31 = __toESM(require_tfjs_esm());
+function getCenterCoordinatesAndSizesLayer(x) {
+  const vec = tf31.unstack(tf31.transpose(x, [1, 0]));
+  const sizes = [
+    tf31.sub(vec[2], vec[0]),
+    tf31.sub(vec[3], vec[1])
+  ];
+  const centers = [
+    tf31.add(vec[0], tf31.div(sizes[0], 2)),
+    tf31.add(vec[1], tf31.div(sizes[1], 2))
+  ];
+  return { sizes, centers };
+}
+function decodeBoxesLayer(x0, x1) {
+  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);
+  const vec = tf31.unstack(tf31.transpose(x1, [1, 0]));
+  const div0_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[2], 5)), sizes[0]), 2);
+  const add0_out = tf31.add(tf31.mul(tf31.div(vec[0], 10), sizes[0]), centers[0]);
+  const div1_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[3], 5)), sizes[1]), 2);
+  const add1_out = tf31.add(tf31.mul(tf31.div(vec[1], 10), sizes[1]), centers[1]);
+  return tf31.transpose(
+    tf31.stack([
+      tf31.sub(add0_out, div0_out),
+      tf31.sub(add1_out, div1_out),
+      tf31.add(add0_out, div0_out),
+      tf31.add(add1_out, div1_out)
+    ]),
+    [1, 0]
+  );
+}
+function outputLayer(boxPredictions, classPredictions, params) {
+  return tf31.tidy(() => {
+    const batchSize = boxPredictions.shape[0];
+    let boxes = decodeBoxesLayer(
+      tf31.reshape(tf31.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]),
+      tf31.reshape(boxPredictions, [-1, 4])
+    );
+    boxes = tf31.reshape(boxes, [batchSize, boxes.shape[0] / batchSize, 4]);
+    const scoresAndClasses = tf31.sigmoid(tf31.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));
+    let scores = tf31.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]);
+    scores = tf31.reshape(scores, [batchSize, scores.shape[1]]);
+    const boxesByBatch = tf31.unstack(boxes);
+    const scoresByBatch = tf31.unstack(scores);
+    return { boxes: boxesByBatch, scores: scoresByBatch };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+var tf33 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/boxPredictionLayer.ts
+var tf32 = __toESM(require_tfjs_esm());
+function boxPredictionLayer(x, params) {
+  return tf32.tidy(() => {
+    const batchSize = x.shape[0];
+    const boxPredictionEncoding = tf32.reshape(
+      convLayer(x, params.box_encoding_predictor),
+      [batchSize, -1, 1, 4]
+    );
+    const classPrediction = tf32.reshape(
+      convLayer(x, params.class_predictor),
+      [batchSize, -1, 3]
+    );
+    return { boxPredictionEncoding, classPrediction };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+function predictionLayer(x, conv11, params) {
+  return tf33.tidy(() => {
+    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);
+    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);
+    const conv22 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);
+    const conv3 = pointwiseConvLayer(conv22, params.conv_3, [2, 2]);
+    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);
+    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);
+    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);
+    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);
+    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);
+    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);
+    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);
+    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);
+    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);
+    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);
+    const boxPredictions = tf33.concat([
+      boxPrediction0.boxPredictionEncoding,
+      boxPrediction1.boxPredictionEncoding,
+      boxPrediction2.boxPredictionEncoding,
+      boxPrediction3.boxPredictionEncoding,
+      boxPrediction4.boxPredictionEncoding,
+      boxPrediction5.boxPredictionEncoding
+    ], 1);
+    const classPredictions = tf33.concat([
+      boxPrediction0.classPrediction,
+      boxPrediction1.classPrediction,
+      boxPrediction2.classPrediction,
+      boxPrediction3.classPrediction,
+      boxPrediction4.classPrediction,
+      boxPrediction5.classPrediction
+    ], 1);
+    return {
+      boxPredictions,
+      classPredictions
+    };
+  });
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1Options.ts
+var SsdMobilenetv1Options = class {
+  constructor({ minConfidence, maxResults } = {}) {
+    this._name = "SsdMobilenetv1Options";
+    this._minConfidence = minConfidence || 0.5;
+    this._maxResults = maxResults || 100;
+    if (typeof this._minConfidence !== "number" || this._minConfidence <= 0 || this._minConfidence >= 1) {
+      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);
+    }
+    if (typeof this._maxResults !== "number") {
+      throw new Error(`${this._name} - expected maxResults to be a number`);
+    }
+  }
+  get minConfidence() {
+    return this._minConfidence;
+  }
+  get maxResults() {
+    return this._maxResults;
+  }
+};
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var SsdMobilenetv1 = class extends NeuralNetwork {
+  constructor() {
+    super("SsdMobilenetv1");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params)
+      throw new Error("SsdMobilenetv1 - load model before inference");
+    return tf34.tidy(() => {
+      const batchTensor = tf34.cast(input.toBatchTensor(512, false), "float32");
+      const x = tf34.sub(tf34.div(batchTensor, 127.5), 1);
+      const features = mobileNetV1(x, params.mobilenetv1);
+      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);
+      return outputLayer(boxPredictions, classPredictions, params.output_layer);
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async locateFaces(input, options = {}) {
+    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);
+    const netInput = await toNetInput(input);
+    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);
+    const boxes = _boxes[0];
+    const scores = _scores[0];
+    for (let i = 1; i < _boxes.length; i++) {
+      _boxes[i].dispose();
+      _scores[i].dispose();
+    }
+    const scoresData = Array.from(scores.dataSync());
+    const iouThreshold = 0.5;
+    const indices = nonMaxSuppression2(boxes, scoresData, maxResults, iouThreshold, minConfidence);
+    const reshapedDims = netInput.getReshapedInputDimensions(0);
+    const inputSize = netInput.inputSize;
+    const padX = inputSize / reshapedDims.width;
+    const padY = inputSize / reshapedDims.height;
+    const boxesData = boxes.arraySync();
+    const results = indices.map((idx) => {
+      const [top, bottom] = [
+        Math.max(0, boxesData[idx][0]),
+        Math.min(1, boxesData[idx][2])
+      ].map((val) => val * padY);
+      const [left, right] = [
+        Math.max(0, boxesData[idx][1]),
+        Math.min(1, boxesData[idx][3])
+      ].map((val) => val * padX);
+      return new FaceDetection(
+        scoresData[idx],
+        new Rect(left, top, right - left, bottom - top),
+        { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) }
+      );
+    });
+    boxes.dispose();
+    scores.dispose();
+    return results;
+  }
+  getDefaultModelName() {
+    return "ssd_mobilenetv1_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap6(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams6(weights);
+  }
+};
+
+// src/ssdMobilenetv1/index.ts
+function createSsdMobilenetv1(weights) {
+  const net = new SsdMobilenetv1();
+  net.extractWeights(weights);
+  return net;
+}
+function createFaceDetectionNet(weights) {
+  return createSsdMobilenetv1(weights);
+}
+var FaceDetectionNet = class extends SsdMobilenetv1 {
+};
+
+// src/tinyYolov2/const.ts
+var IOU_THRESHOLD = 0.4;
+var BOX_ANCHORS = [
+  new Point(0.738768, 0.874946),
+  new Point(2.42204, 2.65704),
+  new Point(4.30971, 7.04493),
+  new Point(10.246, 4.59428),
+  new Point(12.6868, 11.8741)
+];
+var BOX_ANCHORS_SEPARABLE = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB_SEPARABLE = [117.001, 114.697, 97.404];
+var DEFAULT_MODEL_NAME = "tiny_yolov2_model";
+var DEFAULT_MODEL_NAME_SEPARABLE_CONV = "tiny_yolov2_separable_conv_model";
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var tf39 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/config.ts
+var isNumber = (arg) => typeof arg === "number";
+function validateConfig(config) {
+  if (!config) {
+    throw new Error(`invalid config: ${config}`);
+  }
+  if (typeof config.withSeparableConvs !== "boolean") {
+    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);
+  }
+  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1) {
+    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);
+  }
+  if (!Array.isArray(config.classes) || !config.classes.length || !config.classes.every((c) => typeof c === "string")) {
+    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);
+  }
+  if (!Array.isArray(config.anchors) || !config.anchors.length || !config.anchors.map((a) => a || {}).every((a) => isNumber(a.x) && isNumber(a.y))) {
+    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);
+  }
+  if (config.meanRgb && (!Array.isArray(config.meanRgb) || config.meanRgb.length !== 3 || !config.meanRgb.every(isNumber))) {
+    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);
+  }
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+var tf36 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/leaky.ts
+var tf35 = __toESM(require_tfjs_esm());
+function leaky(x) {
+  return tf35.tidy(() => {
+    const min = tf35.mul(x, tf35.scalar(0.10000000149011612));
+    return tf35.add(tf35.relu(tf35.sub(x, min)), min);
+  });
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+function convWithBatchNorm(x, params) {
+  return tf36.tidy(() => {
+    let out = tf36.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf36.conv2d(out, params.conv.filters, [1, 1], "valid");
+    out = tf36.sub(out, params.bn.sub);
+    out = tf36.mul(out, params.bn.truediv);
+    out = tf36.add(out, params.conv.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/depthwiseSeparableConv.ts
+var tf37 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv2(x, params) {
+  return tf37.tidy(() => {
+    let out = tf37.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf37.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], "valid");
+    out = tf37.add(out, params.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/extractParams.ts
+var tf38 = __toESM(require_tfjs_esm());
+function extractorsFactory7(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  function extractBatchNormParams(size, mappedPrefix) {
+    const sub6 = tf38.tensor1d(extractWeights(size));
+    const truediv = tf38.tensor1d(extractWeights(size));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/sub` },
+      { paramPath: `${mappedPrefix}/truediv` }
+    );
+    return { sub: sub6, truediv };
+  }
+  function extractConvWithBatchNormParams(channelsIn, channelsOut, mappedPrefix) {
+    const conv3 = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);
+    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParams7(weights, config, boxEncodingSize, filterSizes) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory7(extractWeights, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = config.isFirstLayerConv2d ? extractConvParams(s0, s1, 3, "conv0") : extractSeparableConvParams(s0, s1, "conv0");
+    const conv1 = extractSeparableConvParams(s1, s2, "conv1");
+    const conv22 = extractSeparableConvParams(s2, s3, "conv2");
+    const conv3 = extractSeparableConvParams(s3, s4, "conv3");
+    const conv4 = extractSeparableConvParams(s4, s5, "conv4");
+    const conv5 = extractSeparableConvParams(s5, s6, "conv5");
+    const conv6 = s7 ? extractSeparableConvParams(s6, s7, "conv6") : void 0;
+    const conv7 = s8 ? extractSeparableConvParams(s7, s8, "conv7") : void 0;
+    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  } else {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = extractConvWithBatchNormParams(s0, s1, "conv0");
+    const conv1 = extractConvWithBatchNormParams(s1, s2, "conv1");
+    const conv22 = extractConvWithBatchNormParams(s2, s3, "conv2");
+    const conv3 = extractConvWithBatchNormParams(s3, s4, "conv3");
+    const conv4 = extractConvWithBatchNormParams(s4, s5, "conv4");
+    const conv5 = extractConvWithBatchNormParams(s5, s6, "conv5");
+    const conv6 = extractConvWithBatchNormParams(s6, s7, "conv6");
+    const conv7 = extractConvWithBatchNormParams(s7, s8, "conv7");
+    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  }
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/extractParamsFromWeightMap.ts
+function extractorsFactory8(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractBatchNormParams(prefix) {
+    const sub6 = extractWeightEntry(`${prefix}/sub`, 1);
+    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);
+    return { sub: sub6, truediv };
+  }
+  function extractConvParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  }
+  function extractConvWithBatchNormParams(prefix) {
+    const conv3 = extractConvParams(`${prefix}/conv`);
+    const bn = extractBatchNormParams(`${prefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParamsFromWeightMap7(weightMap, config) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory8(weightMap, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const numFilters = config.filterSizes && config.filterSizes.length || 9;
+    params = {
+      conv0: config.isFirstLayerConv2d ? extractConvParams("conv0") : extractSeparableConvParams("conv0"),
+      conv1: extractSeparableConvParams("conv1"),
+      conv2: extractSeparableConvParams("conv2"),
+      conv3: extractSeparableConvParams("conv3"),
+      conv4: extractSeparableConvParams("conv4"),
+      conv5: extractSeparableConvParams("conv5"),
+      conv6: numFilters > 7 ? extractSeparableConvParams("conv6") : void 0,
+      conv7: numFilters > 8 ? extractSeparableConvParams("conv7") : void 0,
+      conv8: extractConvParams("conv8")
+    };
+  } else {
+    params = {
+      conv0: extractConvWithBatchNormParams("conv0"),
+      conv1: extractConvWithBatchNormParams("conv1"),
+      conv2: extractConvWithBatchNormParams("conv2"),
+      conv3: extractConvWithBatchNormParams("conv3"),
+      conv4: extractConvWithBatchNormParams("conv4"),
+      conv5: extractConvWithBatchNormParams("conv5"),
+      conv6: extractConvWithBatchNormParams("conv6"),
+      conv7: extractConvWithBatchNormParams("conv7"),
+      conv8: extractConvParams("conv8")
+    };
+  }
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/TinyYolov2Options.ts
+var TinyYolov2Options = class {
+  constructor({ inputSize, scoreThreshold } = {}) {
+    this._name = "TinyYolov2Options";
+    this._inputSize = inputSize || 416;
+    this._scoreThreshold = scoreThreshold || 0.5;
+    if (typeof this._inputSize !== "number" || this._inputSize % 32 !== 0) {
+      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);
+    }
+    if (typeof this._scoreThreshold !== "number" || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {
+      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);
+    }
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get scoreThreshold() {
+    return this._scoreThreshold;
+  }
+};
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var _TinyYolov2Base = class extends NeuralNetwork {
+  constructor(config) {
+    super("TinyYolov2");
+    validateConfig(config);
+    this._config = config;
+  }
+  get config() {
+    return this._config;
+  }
+  get withClassScores() {
+    return this.config.withClassScores || this.config.classes.length > 1;
+  }
+  get boxEncodingSize() {
+    return 5 + (this.withClassScores ? this.config.classes.length : 0);
+  }
+  runTinyYolov2(x, params) {
+    let out = convWithBatchNorm(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = convWithBatchNorm(out, params.conv6);
+    out = convWithBatchNorm(out, params.conv7);
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  runMobilenet(x, params) {
+    let out = this.config.isFirstLayerConv2d ? leaky(convLayer(x, params.conv0, "valid", false)) : depthwiseSeparableConv2(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = params.conv6 ? depthwiseSeparableConv2(out, params.conv6) : out;
+    out = params.conv7 ? depthwiseSeparableConv2(out, params.conv7) : out;
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  forwardInput(input, inputSize) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyYolov2 - load model before inference");
+    }
+    return tf39.tidy(() => {
+      let batchTensor = tf39.cast(input.toBatchTensor(inputSize, false), "float32");
+      batchTensor = this.config.meanRgb ? normalize(batchTensor, this.config.meanRgb) : batchTensor;
+      batchTensor = batchTensor.div(255);
+      return this.config.withSeparableConvs ? this.runMobilenet(batchTensor, params) : this.runTinyYolov2(batchTensor, params);
+    });
+  }
+  async forward(input, inputSize) {
+    return this.forwardInput(await toNetInput(input), inputSize);
+  }
+  async detect(input, forwardParams = {}) {
+    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput, inputSize);
+    const out0 = tf39.tidy(() => tf39.unstack(out)[0].expandDims());
+    const inputDimensions = {
+      width: netInput.getInputWidth(0),
+      height: netInput.getInputHeight(0)
+    };
+    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);
+    out.dispose();
+    out0.dispose();
+    const boxes = results.map((res) => res.box);
+    const scores = results.map((res) => res.score);
+    const classScores = results.map((res) => res.classScore);
+    const classNames = results.map((res) => this.config.classes[res.label]);
+    const indices = nonMaxSuppression(
+      boxes.map((box) => box.rescale(inputSize)),
+      scores,
+      this.config.iouThreshold,
+      true
+    );
+    const detections = indices.map((idx) => new ObjectDetection(
+      scores[idx],
+      classScores[idx],
+      classNames[idx],
+      boxes[idx],
+      inputDimensions
+    ));
+    return detections;
+  }
+  getDefaultModelName() {
+    return "";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap7(weightMap, this.config);
+  }
+  extractParams(weights) {
+    const filterSizes = this.config.filterSizes || _TinyYolov2Base.DEFAULT_FILTER_SIZES;
+    const numFilters = filterSizes ? filterSizes.length : void 0;
+    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {
+      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);
+    }
+    return extractParams7(weights, this.config, this.boxEncodingSize, filterSizes);
+  }
+  async extractBoxes(outputTensor, inputBlobDimensions, scoreThreshold) {
+    const { width, height } = inputBlobDimensions;
+    const inputSize = Math.max(width, height);
+    const correctionFactorX = inputSize / width;
+    const correctionFactorY = inputSize / height;
+    const numCells = outputTensor.shape[1];
+    const numBoxes = this.config.anchors.length;
+    const [boxesTensor, scoresTensor, classScoresTensor] = tf39.tidy(() => {
+      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);
+      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);
+      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);
+      const classScores = this.withClassScores ? tf39.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3) : tf39.scalar(0);
+      return [boxes, scores, classScores];
+    });
+    const results = [];
+    const scoresData = await scoresTensor.array();
+    const boxesData = await boxesTensor.array();
+    for (let row = 0; row < numCells; row++) {
+      for (let col = 0; col < numCells; col++) {
+        for (let anchor = 0; anchor < numBoxes; anchor++) {
+          const score = sigmoid(scoresData[row][col][anchor][0]);
+          if (!scoreThreshold || score > scoreThreshold) {
+            const ctX = (col + sigmoid(boxesData[row][col][anchor][0])) / numCells * correctionFactorX;
+            const ctY = (row + sigmoid(boxesData[row][col][anchor][1])) / numCells * correctionFactorY;
+            const widthLocal = Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x / numCells * correctionFactorX;
+            const heightLocal = Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y / numCells * correctionFactorY;
+            const x = ctX - widthLocal / 2;
+            const y = ctY - heightLocal / 2;
+            const pos = { row, col, anchor };
+            const { classScore, label } = this.withClassScores ? await this.extractPredictedClass(classScoresTensor, pos) : { classScore: 1, label: 0 };
+            results.push({
+              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),
+              score,
+              classScore: score * classScore,
+              label,
+              ...pos
+            });
+          }
+        }
+      }
+    }
+    boxesTensor.dispose();
+    scoresTensor.dispose();
+    classScoresTensor.dispose();
+    return results;
+  }
+  async extractPredictedClass(classesTensor, pos) {
+    const { row, col, anchor } = pos;
+    const classesData = await classesTensor.array();
+    return Array(this.config.classes.length).fill(0).map((_, i) => classesData[row][col][anchor][i]).map((classScore, label) => ({
+      classScore,
+      label
+    })).reduce((max, curr) => max.classScore > curr.classScore ? max : curr);
+  }
+};
+var TinyYolov2Base = _TinyYolov2Base;
+TinyYolov2Base.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+
+// src/tinyYolov2/TinyYolov2.ts
+var TinyYolov2 = class extends TinyYolov2Base {
+  constructor(withSeparableConvs = true) {
+    const config = {
+      withSeparableConvs,
+      iouThreshold: IOU_THRESHOLD,
+      classes: ["face"],
+      ...withSeparableConvs ? {
+        anchors: BOX_ANCHORS_SEPARABLE,
+        meanRgb: MEAN_RGB_SEPARABLE
+      } : {
+        anchors: BOX_ANCHORS,
+        withClassScores: true
+      }
+    };
+    super(config);
+  }
+  get withSeparableConvs() {
+    return this.config.withSeparableConvs;
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/tinyYolov2/index.ts
+function createTinyYolov2(weights, withSeparableConvs = true) {
+  const net = new TinyYolov2(withSeparableConvs);
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/tinyFaceDetector/TinyFaceDetectorOptions.ts
+var TinyFaceDetectorOptions = class extends TinyYolov2Options {
+  constructor() {
+    super(...arguments);
+    this._name = "TinyFaceDetectorOptions";
+  }
+};
+
+// src/globalApi/ComposableTask.ts
+var ComposableTask = class {
+  async then(onfulfilled) {
+    return onfulfilled(await this.run());
+  }
+  async run() {
+    throw new Error("ComposableTask - run is not implemented");
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var tf41 = __toESM(require_tfjs_esm());
+
+// src/globalApi/extractFacesAndComputeResults.ts
+var tf40 = __toESM(require_tfjs_esm());
+async function extractAllFacesAndComputeResults(parentResults, input, computeResults, extractedFaces, getRectForAlignment = ({ alignedRect }) => alignedRect) {
+  const faceBoxes = parentResults.map((parentResult) => isWithFaceLandmarks(parentResult) ? getRectForAlignment(parentResult) : parentResult.detection);
+  const faces = extractedFaces || (input instanceof tf40.Tensor ? await extractFaceTensors(input, faceBoxes) : await extractFaces(input, faceBoxes));
+  const results = await computeResults(faces);
+  faces.forEach((f) => f instanceof tf40.Tensor && f.dispose());
+  return results;
+}
+async function extractSingleFaceAndComputeResult(parentResult, input, computeResult, extractedFaces, getRectForAlignment) {
+  return extractAllFacesAndComputeResults(
+    [parentResult],
+    input,
+    async (faces) => computeResult(faces[0]),
+    extractedFaces,
+    getRectForAlignment
+  );
+}
+
+// src/tinyFaceDetector/const.ts
+var IOU_THRESHOLD2 = 0.4;
+var BOX_ANCHORS2 = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB = [117.001, 114.697, 97.404];
+
+// src/tinyFaceDetector/TinyFaceDetector.ts
+var TinyFaceDetector = class extends TinyYolov2Base {
+  constructor() {
+    const config = {
+      withSeparableConvs: true,
+      iouThreshold: IOU_THRESHOLD2,
+      classes: ["face"],
+      anchors: BOX_ANCHORS2,
+      meanRgb: MEAN_RGB,
+      isFirstLayerConv2d: true,
+      filterSizes: [3, 16, 32, 64, 128, 256, 512]
+    };
+    super(config);
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return "tiny_face_detector_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/globalApi/nets.ts
+var nets = {
+  ssdMobilenetv1: new SsdMobilenetv1(),
+  tinyFaceDetector: new TinyFaceDetector(),
+  tinyYolov2: new TinyYolov2(),
+  faceLandmark68Net: new FaceLandmark68Net(),
+  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),
+  faceRecognitionNet: new FaceRecognitionNet(),
+  faceExpressionNet: new FaceExpressionNet(),
+  ageGenderNet: new AgeGenderNet()
+};
+var ssdMobilenetv1 = (input, options) => nets.ssdMobilenetv1.locateFaces(input, options);
+var tinyFaceDetector = (input, options) => nets.tinyFaceDetector.locateFaces(input, options);
+var tinyYolov2 = (input, options) => nets.tinyYolov2.locateFaces(input, options);
+var detectFaceLandmarks = (input) => nets.faceLandmark68Net.detectLandmarks(input);
+var detectFaceLandmarksTiny = (input) => nets.faceLandmark68TinyNet.detectLandmarks(input);
+var computeFaceDescriptor = (input) => nets.faceRecognitionNet.computeFaceDescriptor(input);
+var recognizeFaceExpressions = (input) => nets.faceExpressionNet.predictExpressions(input);
+var predictAgeAndGender = (input) => nets.ageGenderNet.predictAgeAndGender(input);
+var loadSsdMobilenetv1Model = (url) => nets.ssdMobilenetv1.load(url);
+var loadTinyFaceDetectorModel = (url) => nets.tinyFaceDetector.load(url);
+var loadTinyYolov2Model = (url) => nets.tinyYolov2.load(url);
+var loadFaceLandmarkModel = (url) => nets.faceLandmark68Net.load(url);
+var loadFaceLandmarkTinyModel = (url) => nets.faceLandmark68TinyNet.load(url);
+var loadFaceRecognitionModel = (url) => nets.faceRecognitionNet.load(url);
+var loadFaceExpressionModel = (url) => nets.faceExpressionNet.load(url);
+var loadAgeGenderModel = (url) => nets.ageGenderNet.load(url);
+var loadFaceDetectionModel = loadSsdMobilenetv1Model;
+var locateFaces = ssdMobilenetv1;
+var detectLandmarks = detectFaceLandmarks;
+
+// src/globalApi/PredictFaceExpressionsTask.ts
+var PredictFaceExpressionsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const faceExpressionsByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(
+        faces.map((face) => nets.faceExpressionNet.predictExpressions(face))
+      ),
+      this.extractedFaces
+    );
+    return parentResults.map(
+      (parentResult, i) => extendWithFaceExpressions(parentResult, faceExpressionsByFace[i])
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const faceExpressions = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceExpressionNet.predictExpressions(face),
+      this.extractedFaces
+    );
+    return extendWithFaceExpressions(parentResult, faceExpressions);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictAllFaceExpressionsWithFaceAlignmentTask = class extends PredictAllFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsWithFaceAlignmentTask = class extends PredictSingleFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/PredictAgeAndGenderTask.ts
+var PredictAgeAndGenderTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const ageAndGenderByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face))),
+      this.extractedFaces
+    );
+    return parentResults.map((parentResult, i) => {
+      const { age, gender, genderProbability } = ageAndGenderByFace[i];
+      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+    });
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.ageGenderNet.predictAgeAndGender(face),
+      this.extractedFaces
+    );
+    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictAllAgeAndGenderWithFaceAlignmentTask = class extends PredictAllAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderWithFaceAlignmentTask = class extends PredictSingleAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/ComputeFaceDescriptorsTasks.ts
+var ComputeFaceDescriptorsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+  }
+};
+var ComputeAllFaceDescriptorsTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const descriptors = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face))),
+      null,
+      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return descriptors.map((descriptor, i) => extendWithFaceDescriptor(parentResults[i], descriptor));
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+var ComputeSingleFaceDescriptorTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const descriptor = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face),
+      null,
+      (parentResult2) => parentResult2.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return extendWithFaceDescriptor(parentResult, descriptor);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var DetectFaceLandmarksTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, useTinyLandmarkNet) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.useTinyLandmarkNet = useTinyLandmarkNet;
+  }
+  get landmarkNet() {
+    return this.useTinyLandmarkNet ? nets.faceLandmark68TinyNet : nets.faceLandmark68Net;
+  }
+};
+var DetectAllFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const detections = parentResults.map((res) => res.detection);
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, detections) : await extractFaces(this.input, detections);
+    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face)));
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    const result = parentResults.filter((_parentResult, i) => faceLandmarksByFace[i]).map((parentResult, i) => extendWithFaceLandmarks(parentResult, faceLandmarksByFace[i]));
+    return result;
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var DetectSingleFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const { detection } = parentResult;
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, [detection]) : await extractFaces(this.input, [detection]);
+    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]);
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    return extendWithFaceLandmarks(parentResult, landmarks);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFacesTasks.ts
+var DetectFacesTaskBase = class extends ComposableTask {
+  constructor(input, options = new SsdMobilenetv1Options()) {
+    super();
+    this.input = input;
+    this.options = options;
+  }
+};
+var DetectAllFacesTask = class extends DetectFacesTaskBase {
+  async run() {
+    const { input, options } = this;
+    let result;
+    if (options instanceof TinyFaceDetectorOptions)
+      result = nets.tinyFaceDetector.locateFaces(input, options);
+    else if (options instanceof SsdMobilenetv1Options)
+      result = nets.ssdMobilenetv1.locateFaces(input, options);
+    else if (options instanceof TinyYolov2Options)
+      result = nets.tinyYolov2.locateFaces(input, options);
+    else
+      throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+    return result;
+  }
+  runAndExtendWithFaceDetections() {
+    return new Promise((resolve, reject) => {
+      this.run().then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection)))).catch((err) => reject(err));
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectAllFaceLandmarksTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+};
+var DetectSingleFaceTask = class extends DetectFacesTaskBase {
+  async run() {
+    const faceDetections = await new DetectAllFacesTask(this.input, this.options);
+    let faceDetectionWithHighestScore = faceDetections[0];
+    faceDetections.forEach((faceDetection) => {
+      if (faceDetection.score > faceDetectionWithHighestScore.score)
+        faceDetectionWithHighestScore = faceDetection;
+    });
+    return faceDetectionWithHighestScore;
+  }
+  runAndExtendWithFaceDetection() {
+    return new Promise(async (resolve) => {
+      const detection = await this.run();
+      resolve(detection ? extendWithFaceDetection({}, detection) : void 0);
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectSingleFaceLandmarksTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+};
+
+// src/globalApi/detectFaces.ts
+function detectSingleFace(input, options = new SsdMobilenetv1Options()) {
+  return new DetectSingleFaceTask(input, options);
+}
+function detectAllFaces(input, options = new SsdMobilenetv1Options()) {
+  return new DetectAllFacesTask(input, options);
+}
+
+// src/globalApi/allFaces.ts
+async function allFacesSsdMobilenetv1(input, minConfidence) {
+  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {})).withFaceLandmarks().withFaceDescriptors();
+}
+async function allFacesTinyYolov2(input, forwardParams = {}) {
+  return detectAllFaces(input, new TinyYolov2Options(forwardParams)).withFaceLandmarks().withFaceDescriptors();
+}
+var allFaces = allFacesSsdMobilenetv1;
+
+// src/euclideanDistance.ts
+function euclideanDistance(arr1, arr2) {
+  if (arr1.length !== arr2.length)
+    throw new Error("euclideanDistance: arr1.length !== arr2.length");
+  const desc1 = Array.from(arr1);
+  const desc2 = Array.from(arr2);
+  return Math.sqrt(
+    desc1.map((val, i) => val - desc2[i]).reduce((res, diff) => res + diff * diff, 0)
+  );
+}
+
+// src/globalApi/FaceMatcher.ts
+var FaceMatcher = class {
+  constructor(inputs, distanceThreshold = 0.6) {
+    this._distanceThreshold = distanceThreshold;
+    const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+    if (!inputArray.length)
+      throw new Error("FaceRecognizer.constructor - expected atleast one input");
+    let count = 1;
+    const createUniqueLabel = () => `person ${count++}`;
+    this._labeledDescriptors = inputArray.map((desc) => {
+      if (desc instanceof LabeledFaceDescriptors)
+        return desc;
+      if (desc instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);
+      if (desc.descriptor && desc.descriptor instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);
+      throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>");
+    });
+  }
+  get labeledDescriptors() {
+    return this._labeledDescriptors;
+  }
+  get distanceThreshold() {
+    return this._distanceThreshold;
+  }
+  computeMeanDistance(queryDescriptor, descriptors) {
+    return descriptors.map((d) => euclideanDistance(d, queryDescriptor)).reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);
+  }
+  matchDescriptor(queryDescriptor) {
+    return this.labeledDescriptors.map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors))).reduce((best, curr) => best.distance < curr.distance ? best : curr);
+  }
+  findBestMatch(queryDescriptor) {
+    const bestMatch = this.matchDescriptor(queryDescriptor);
+    return bestMatch.distance < this._distanceThreshold ? bestMatch : new FaceMatch("unknown", bestMatch.distance);
+  }
+  toJSON() {
+    return {
+      distanceThreshold: this._distanceThreshold,
+      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON())
+    };
+  }
+  static fromJSON(json) {
+    const labeledDescriptors = json.labeledDescriptors.map((ld) => LabeledFaceDescriptors.fromJSON(ld));
+    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);
+  }
+};
+
+// src/tinyFaceDetector/index.ts
+function createTinyFaceDetector(weights) {
+  const net = new TinyFaceDetector();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/resizeResults.ts
+function resizeResults(results, dimensions) {
+  const { width, height } = new Dimensions(dimensions.width, dimensions.height);
+  if (width <= 0 || height <= 0) {
+    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);
+  }
+  if (Array.isArray(results)) {
+    return results.map((obj) => resizeResults(obj, { width, height }));
+  }
+  if (isWithFaceLandmarks(results)) {
+    const resizedDetection = results.detection.forSize(width, height);
+    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);
+    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);
+  }
+  if (isWithFaceDetection(results)) {
+    return extendWithFaceDetection(results, results.detection.forSize(width, height));
+  }
+  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {
+    return results.forSize(width, height);
+  }
+  return results;
+}
+
+// src/index.ts
+var version2 = version;
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  AgeGenderNet,
+  BoundingBox,
+  Box,
+  ComposableTask,
+  ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask,
+  DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask,
+  Dimensions,
+  FACE_EXPRESSION_LABELS,
+  FaceDetection,
+  FaceDetectionNet,
+  FaceExpressionNet,
+  FaceExpressions,
+  FaceLandmark68Net,
+  FaceLandmark68TinyNet,
+  FaceLandmarkNet,
+  FaceLandmarks,
+  FaceLandmarks5,
+  FaceLandmarks68,
+  FaceMatch,
+  FaceMatcher,
+  FaceRecognitionNet,
+  Gender,
+  LabeledBox,
+  LabeledFaceDescriptors,
+  NetInput,
+  NeuralNetwork,
+  ObjectDetection,
+  Point,
+  PredictedBox,
+  Rect,
+  SsdMobilenetv1,
+  SsdMobilenetv1Options,
+  TinyFaceDetector,
+  TinyFaceDetectorOptions,
+  TinyYolov2,
+  TinyYolov2Options,
+  allFaces,
+  allFacesSsdMobilenetv1,
+  allFacesTinyYolov2,
+  awaitMediaLoaded,
+  bufferToImage,
+  computeFaceDescriptor,
+  createCanvas,
+  createCanvasFromMedia,
+  createFaceDetectionNet,
+  createFaceRecognitionNet,
+  createSsdMobilenetv1,
+  createTinyFaceDetector,
+  createTinyYolov2,
+  detectAllFaces,
+  detectFaceLandmarks,
+  detectFaceLandmarksTiny,
+  detectLandmarks,
+  detectSingleFace,
+  draw,
+  env,
+  euclideanDistance,
+  extendWithAge,
+  extendWithFaceDescriptor,
+  extendWithFaceDetection,
+  extendWithFaceExpressions,
+  extendWithFaceLandmarks,
+  extendWithGender,
+  extractFaceTensors,
+  extractFaces,
+  fetchImage,
+  fetchJson,
+  fetchNetWeights,
+  fetchOrThrow,
+  fetchVideo,
+  getContext2dOrThrow,
+  getMediaDimensions,
+  imageTensorToCanvas,
+  imageToSquare,
+  inverseSigmoid,
+  iou,
+  isMediaElement,
+  isMediaLoaded,
+  isWithAge,
+  isWithFaceDetection,
+  isWithFaceExpressions,
+  isWithFaceLandmarks,
+  isWithGender,
+  loadAgeGenderModel,
+  loadFaceDetectionModel,
+  loadFaceExpressionModel,
+  loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel,
+  loadTinyYolov2Model,
+  loadWeightMap,
+  locateFaces,
+  matchDimensions,
+  minBbox,
+  nets,
+  nonMaxSuppression,
+  normalize,
+  padToSquare,
+  predictAgeAndGender,
+  recognizeFaceExpressions,
+  resizeResults,
+  resolveInput,
+  shuffleArray,
+  sigmoid,
+  ssdMobilenetv1,
+  tf,
+  tinyFaceDetector,
+  tinyYolov2,
+  toNetInput,
+  utils,
+  validateConfig,
+  version
+});
diff --git a/dist/face-api.node.js b/dist/face-api.node.js
index d582aff..7331a0d 100644
--- a/dist/face-api.node.js
+++ b/dist/face-api.node.js
@@ -4,4 +4,4926 @@
   author: <https://github.com/vladmandic>'
 */
 
-"use strict";var dn=Object.create;var er=Object.defineProperty;var hn=Object.getOwnPropertyDescriptor;var bn=Object.getOwnPropertyNames;var gn=Object.getPrototypeOf,xn=Object.prototype.hasOwnProperty;var vn=(o,t)=>()=>(t||o((t={exports:{}}).exports,t),t.exports),Lr=(o,t)=>{for(var e in t)er(o,e,{get:t[e],enumerable:!0})},To=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of bn(t))!xn.call(o,n)&&n!==e&&er(o,n,{get:()=>t[n],enumerable:!(r=hn(t,n))||r.enumerable});return o};var v=(o,t,e)=>(e=o!=null?dn(gn(o)):{},To(t||!o||!o.__esModule?er(e,"default",{value:o,enumerable:!0}):e,o)),yn=o=>To(er({},"__esModule",{value:!0}),o);var x=vn((Va,Br)=>{"use strict";var Wr=Object.defineProperty,_n=Object.getOwnPropertyDescriptor,Tn=Object.getOwnPropertyNames,wn=Object.prototype.hasOwnProperty,Pn=(o,t)=>{for(var e in t)Wr(o,e,{get:t[e],enumerable:!0})},Ar=(o,t,e,r)=>{if(t&&typeof t=="object"||typeof t=="function")for(let n of Tn(t))!wn.call(o,n)&&n!==e&&Wr(o,n,{get:()=>t[n],enumerable:!(r=_n(t,n))||r.enumerable});return o},Fn=(o,t,e)=>(Ar(o,t,"default"),e&&Ar(e,t,"default")),Dn=o=>Ar(Wr({},"__esModule",{value:!0}),o),kr={};Pn(kr,{version:()=>Nn});Br.exports=Dn(kr);Fn(kr,require("@tensorflow/tfjs-node"),Br.exports);var wo="4.0.0",En="4.0.0",Mn="4.0.0",Cn="4.0.0",In="4.0.0",Nn={tfjs:wo,"tfjs-core":wo,"tfjs-converter":En,"tfjs-backend-cpu":Mn,"tfjs-backend-webgl":Cn,"tfjs-backend-wasm":In}});var Ra={};Lr(Ra,{AgeGenderNet:()=>He,BoundingBox:()=>Vt,Box:()=>D,ComposableTask:()=>J,ComputeAllFaceDescriptorsTask:()=>Pt,ComputeFaceDescriptorsTaskBase:()=>Ue,ComputeSingleFaceDescriptorTask:()=>Ft,DetectAllFaceLandmarksTask:()=>qe,DetectAllFacesTask:()=>Ie,DetectFaceLandmarksTaskBase:()=>Je,DetectFacesTaskBase:()=>Ke,DetectSingleFaceLandmarksTask:()=>Ze,DetectSingleFaceTask:()=>Qe,Dimensions:()=>k,FACE_EXPRESSION_LABELS:()=>so,FaceDetection:()=>M,FaceDetectionNet:()=>lo,FaceExpressionNet:()=>Oe,FaceExpressions:()=>yt,FaceLandmark68Net:()=>Kt,FaceLandmark68TinyNet:()=>ze,FaceLandmarkNet:()=>po,FaceLandmarks:()=>z,FaceLandmarks5:()=>jr,FaceLandmarks68:()=>Gt,FaceMatch:()=>pe,FaceMatcher:()=>tr,FaceRecognitionNet:()=>Qt,Gender:()=>Tr,LabeledBox:()=>ue,LabeledFaceDescriptors:()=>mt,NetInput:()=>ut,NeuralNetwork:()=>A,ObjectDetection:()=>bt,Point:()=>g,PredictedBox:()=>Ur,Rect:()=>Yt,SsdMobilenetv1:()=>St,SsdMobilenetv1Options:()=>X,TinyFaceDetector:()=>ne,TinyFaceDetectorOptions:()=>je,TinyYolov2:()=>re,TinyYolov2Options:()=>st,allFaces:()=>Aa,allFacesSsdMobilenetv1:()=>fn,allFacesTinyYolov2:()=>La,awaitMediaLoaded:()=>to,bufferToImage:()=>eo,computeFaceDescriptor:()=>va,createCanvas:()=>Jt,createCanvasFromMedia:()=>We,createFaceDetectionNet:()=>fa,createFaceRecognitionNet:()=>ea,createSsdMobilenetv1:()=>qo,createTinyFaceDetector:()=>Wa,createTinyYolov2:()=>ha,detectAllFaces:()=>Sr,detectFaceLandmarks:()=>pn,detectFaceLandmarksTiny:()=>xa,detectLandmarks:()=>Na,detectSingleFace:()=>Sa,draw:()=>co,env:()=>w,euclideanDistance:()=>vo,extendWithAge:()=>Er,extendWithFaceDescriptor:()=>Dr,extendWithFaceDetection:()=>jt,extendWithFaceExpressions:()=>xr,extendWithFaceLandmarks:()=>Pe,extendWithGender:()=>Mr,extractFaceTensors:()=>de,extractFaces:()=>le,fetchImage:()=>Hn,fetchJson:()=>no,fetchNetWeights:()=>zn,fetchOrThrow:()=>xt,fetchVideo:()=>Vn,getContext2dOrThrow:()=>O,getMediaDimensions:()=>Xt,imageTensorToCanvas:()=>ro,imageToSquare:()=>oo,inverseSigmoid:()=>An,iou:()=>zr,isMediaElement:()=>ir,isMediaLoaded:()=>Ae,isWithAge:()=>ra,isWithFaceDetection:()=>pt,isWithFaceExpressions:()=>io,isWithFaceLandmarks:()=>Zt,isWithGender:()=>oa,loadAgeGenderModel:()=>Ma,loadFaceDetectionModel:()=>Ca,loadFaceExpressionModel:()=>Ea,loadFaceLandmarkModel:()=>Pa,loadFaceLandmarkTinyModel:()=>Fa,loadFaceRecognitionModel:()=>Da,loadSsdMobilenetv1Model:()=>un,loadTinyFaceDetectorModel:()=>Ta,loadTinyYolov2Model:()=>wa,loadWeightMap:()=>ao,locateFaces:()=>Ia,matchDimensions:()=>Yn,minBbox:()=>Vr,nets:()=>P,nonMaxSuppression:()=>Yr,normalize:()=>rt,padToSquare:()=>Gr,predictAgeAndGender:()=>_a,recognizeFaceExpressions:()=>ya,resizeResults:()=>ln,resolveInput:()=>Ut,shuffleArray:()=>Ln,sigmoid:()=>Ne,ssdMobilenetv1:()=>mn,tf:()=>ka,tinyFaceDetector:()=>ba,tinyYolov2:()=>ga,toNetInput:()=>C,utils:()=>Hr,validateConfig:()=>ho,version:()=>Ba});module.exports=yn(Ra);var ka=v(x());var co={};Lr(co,{AnchorPosition:()=>Qr,DrawBox:()=>Le,DrawBoxOptions:()=>ar,DrawFaceLandmarks:()=>yr,DrawFaceLandmarksOptions:()=>vr,DrawTextField:()=>gt,DrawTextFieldOptions:()=>fe,drawContour:()=>dt,drawDetections:()=>On,drawFaceExpressions:()=>Gn,drawFaceLandmarks:()=>Un});function dt(o,t,e=!1){if(o.beginPath(),t.slice(1).forEach(({x:r,y:n},a)=>{let s=t[a];o.moveTo(s.x,s.y),o.lineTo(r,n)}),e){let r=t[t.length-1],n=t[0];if(!r||!n)return;o.moveTo(r.x,r.y),o.lineTo(n.x,n.y)}o.stroke()}var Hr={};Lr(Hr,{computeReshapedDimensions:()=>Or,getCenterPoint:()=>zt,isDimensions:()=>or,isEven:()=>rr,isFloat:()=>$r,isTensor:()=>Ot,isTensor1D:()=>Sn,isTensor2D:()=>Rr,isTensor3D:()=>ht,isTensor4D:()=>U,isValidNumber:()=>et,isValidProbablitiy:()=>me,range:()=>it,round:()=>Ht});var Po=v(x());var k=class{constructor(t,e){if(!et(t)||!et(e))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:t,height:e})}`);this._width=t,this._height=e}get width(){return this._width}get height(){return this._height}reverse(){return new k(1/this.width,1/this.height)}};function Ot(o,t){return o instanceof Po.Tensor&&o.shape.length===t}function Sn(o){return Ot(o,1)}function Rr(o){return Ot(o,2)}function ht(o){return Ot(o,3)}function U(o){return Ot(o,4)}function $r(o){return o%1!==0}function rr(o){return o%2===0}function Ht(o,t=2){let e=10**t;return Math.floor(o*e)/e}function or(o){return o&&o.width&&o.height}function Or({width:o,height:t},e){let r=e/Math.max(t,o);return new k(Math.round(o*r),Math.round(t*r))}function zt(o){return o.reduce((t,e)=>t.add(e),new g(0,0)).div(new g(o.length,o.length))}function it(o,t,e){return Array(o).fill(0).map((r,n)=>t+n*e)}function et(o){return!!o&&o!==1/0&&o!==-1/0&&!Number.isNaN(o)||o===0}function me(o){return et(o)&&o>=0&&o<=1}var g=class{constructor(t,e){this._x=t,this._y=e}get x(){return this._x}get y(){return this._y}add(t){return new g(this.x+t.x,this.y+t.y)}sub(t){return new g(this.x-t.x,this.y-t.y)}mul(t){return new g(this.x*t.x,this.y*t.y)}div(t){return new g(this.x/t.x,this.y/t.y)}abs(){return new g(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new g(Math.floor(this.x),Math.floor(this.y))}};var D=class{static isRect(t){return!!t&&[t.x,t.y,t.width,t.height].every(et)}static assertIsValidBox(t,e,r=!1){if(!D.isRect(t))throw new Error(`${e} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${e} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,e=!0){let r=t||{},n=[r.left,r.top,r.right,r.bottom].every(et),a=[r.x,r.y,r.width,r.height].every(et);if(!a&&!n)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[s,i,c,m]=a?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];D.assertIsValidBox({x:s,y:i,width:c,height:m},"Box.constructor",e),this._x=s,this._y=i,this._width=c,this._height=m}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new g(this.left,this.top)}get topRight(){return new g(this.right,this.top)}get bottomLeft(){return new g(this.left,this.bottom)}get bottomRight(){return new g(this.right,this.bottom)}round(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.round(a));return new D({x:t,y:e,width:r,height:n})}floor(){let[t,e,r,n]=[this.x,this.y,this.width,this.height].map(a=>Math.floor(a));return new D({x:t,y:e,width:r,height:n})}toSquare(){let{x:t,y:e,width:r,height:n}=this,a=Math.abs(r-n);return r<n&&(t-=a/2,r+=a),n<r&&(e-=a/2,n+=a),new D({x:t,y:e,width:r,height:n})}rescale(t){let e=or(t)?t.width:t,r=or(t)?t.height:t;return new D({x:this.x*e,y:this.y*r,width:this.width*e,height:this.height*r})}pad(t,e){let[r,n,a,s]=[this.x-t/2,this.y-e/2,this.width+t,this.height+e];return new D({x:r,y:n,width:a,height:s})}clipAtImageBorders(t,e){let{x:r,y:n,right:a,bottom:s}=this,i=Math.max(r,0),c=Math.max(n,0),m=a-i,p=s-c,u=Math.min(m,t-i),f=Math.min(p,e-c);return new D({x:i,y:c,width:u,height:f}).floor()}shift(t,e){let{width:r,height:n}=this,a=this.x+t,s=this.y+e;return new D({x:a,y:s,width:r,height:n})}padAtBorders(t,e){let r=this.width+1,n=this.height+1,a=1,s=1,i=r,c=n,m=this.left,p=this.top,u=this.right,f=this.bottom;return u>e&&(i=-u+e+r,u=e),f>t&&(c=-f+t+n,f=t),m<1&&(c=2-m,m=1),p<1&&(c=2-p,p=1),{dy:s,edy:c,dx:a,edx:i,y:p,ey:f,x:m,ex:u,w:r,h:n}}calibrate(t){return new D({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}};var Vt=class extends D{constructor(t,e,r,n,a=!1){super({left:t,top:e,right:r,bottom:n},a)}};var bt=class{constructor(t,e,r,n,a){this._imageDims=new k(a.width,a.height),this._score=t,this._classScore=e,this._className=r,this._box=new D(n).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new D(this._box).rescale(this.imageDims.reverse())}forSize(t,e){return new bt(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:e})}};var M=class extends bt{constructor(t,e,r){super(t,t,"",e,r)}forSize(t,e){let{score:r,relativeBox:n,imageDims:a}=super.forSize(t,e);return new M(r,n,a)}};function zr(o,t,e=!0){let r=Math.max(0,Math.min(o.right,t.right)-Math.max(o.left,t.left)),n=Math.max(0,Math.min(o.bottom,t.bottom)-Math.max(o.top,t.top)),a=r*n;return e?a/(o.area+t.area-a):a/Math.min(o.area,t.area)}function Vr(o){let t=o.map(i=>i.x),e=o.map(i=>i.y),r=t.reduce((i,c)=>c<i?c:i,1/0),n=e.reduce((i,c)=>c<i?c:i,1/0),a=t.reduce((i,c)=>i<c?c:i,0),s=e.reduce((i,c)=>i<c?c:i,0);return new Vt(r,n,a,s)}function Yr(o,t,e,r=!0){let n=t.map((s,i)=>({score:s,boxIndex:i})).sort((s,i)=>s.score-i.score).map(s=>s.boxIndex),a=[];for(;n.length>0;){let s=n.pop();a.push(s);let i=n,c=[];for(let m=0;m<i.length;m++){let p=i[m],u=o[s],f=o[p];c.push(zr(u,f,r))}n=n.filter((m,p)=>c[p]<=e)}return a}var ct=v(x());function rt(o,t){return ct.tidy(()=>{let[e,r,n]=t,a=ct.fill([...o.shape.slice(0,3),1],e,"float32"),s=ct.fill([...o.shape.slice(0,3),1],r,"float32"),i=ct.fill([...o.shape.slice(0,3),1],n,"float32"),c=ct.concat([a,s,i],3);return ct.sub(o,c)})}var Ct=v(x());function Gr(o,t=!1){return Ct.tidy(()=>{let[e,r]=o.shape.slice(1);if(e===r)return o;let n=Math.abs(e-r),a=Math.round(n*(t?.5:1)),s=e>r?2:1,i=f=>{let l=o.shape.slice();return l[s]=f,Ct.fill(l,0,"float32")},c=i(a),m=n-c.shape[s],u=[t&&m?i(m):null,o,c].filter(f=>!!f).map(f=>Ct.cast(f,"float32"));return Ct.concat(u,s)})}function Ln(o){let t=o.slice();for(let e=t.length-1;e>0;e--){let r=Math.floor(Math.random()*(e+1)),n=t[e];t[e]=t[r],t[r]=n}return t}function Ne(o){return 1/(1+Math.exp(-o))}function An(o){return Math.log(o/(1-o))}var Yt=class extends D{constructor(t,e,r,n,a=!1){super({x:t,y:e,width:r,height:n},a)}};var Wn=.5,kn=.43,Bn=.45,z=class{constructor(t,e,r=new g(0,0)){let{width:n,height:a}=e;this._imgDims=new k(n,a),this._shift=r,this._positions=t.map(s=>s.mul(new g(n,a)).add(r))}get shift(){return new g(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(t=>t.sub(this._shift).div(new g(this.imageWidth,this.imageHeight)))}forSize(t,e){return new this.constructor(this.relativePositions,{width:t,height:e})}shiftBy(t,e){return new this.constructor(this.relativePositions,this._imgDims,new g(t,e))}shiftByPoint(t){return this.shiftBy(t.x,t.y)}align(t,e={}){if(t){let a=t instanceof M?t.box.floor():new D(t);return this.shiftBy(a.x,a.y).align(null,e)}let{useDlibAlignment:r,minBoxPadding:n}={useDlibAlignment:!1,minBoxPadding:.2,...e};return r?this.alignDlib():this.alignMinBbox(n)}alignDlib(){let t=this.getRefPointsForAlignment(),[e,r,n]=t,a=u=>n.sub(u).magnitude(),s=(a(e)+a(r))/2,i=Math.floor(s/Bn),c=zt(t),m=Math.floor(Math.max(0,c.x-Wn*i)),p=Math.floor(Math.max(0,c.y-kn*i));return new Yt(m,p,Math.min(i,this.imageWidth+m),Math.min(i,this.imageHeight+p))}alignMinBbox(t){let e=Vr(this.positions);return e.pad(e.width*t,e.height*t)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}};var jr=class extends z{getRefPointsForAlignment(){let t=this.positions;return[t[0],t[1],zt([t[3],t[4]])]}};var Gt=class extends z{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(zt)}};var pe=class{constructor(t,e){this._label=t,this._distance=e}get label(){return this._label}get distance(){return this._distance}toString(t=!0){return`${this.label}${t?` (${Ht(this.distance)})`:""}`}};var ue=class extends D{constructor(e,r){super(e);this._label=r}static assertIsValidLabeledBox(e,r){if(D.assertIsValidBox(e,r),!et(e.label))throw new Error(`${r} - expected property label (${e.label}) to be a number`)}get label(){return this._label}};var mt=class{constructor(t,e){if(typeof t!="string")throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(e)||e.some(r=>!(r instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=t,this._descriptors=e}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(t=>Array.from(t))}}static fromJSON(t){let e=t.descriptors.map(r=>new Float32Array(r));return new mt(t.label,e)}};var Ur=class extends ue{constructor(e,r,n,a){super(e,r);this._score=n,this._classScore=a}static assertIsValidPredictedBox(e,r){if(ue.assertIsValidLabeledBox(e,r),!me(e.score)||!me(e.classScore))throw new Error(`${r} - expected properties score (${e.score}) and (${e.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}};function pt(o){return o.detection instanceof M}function jt(o,t){return{...o,...{detection:t}}}function Xr(){let o=window.fetch;if(!o)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:o,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function Se(){return typeof global=="object"&&typeof process!="undefined"&&process.versions!=null&&process.versions.node!=null}function nr(o){let t="";if(!o&&Se())try{o=require("fs")}catch(r){t=r.toString()}return{readFile:o?r=>new Promise((n,a)=>{o.readFile(r,(s,i)=>s?a(s):n(i))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function Jr(){let o=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,e=global.Video||global.HTMLVideoElement,r=()=>{if(o)return new o;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},n=()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},a=()=>{if(e)return new e;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},s=global.fetch,i=nr();return{Canvas:o||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:r,createImageElement:n,createVideoElement:a,fetch:s,...i}}function qr(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}var L;function Rn(){if(!L)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return L}function Zr(o){L=o}function Kr(){return qr()?Zr(Xr()):Se()?Zr(Jr()):null}function $n(o){if(L||Kr(),!L)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=L.Canvas,Image:e=L.Image}=o;L.Canvas=t,L.Image=e,L.createCanvasElement=o.createCanvasElement||(()=>new t),L.createImageElement=o.createImageElement||(()=>new e),L.ImageData=o.ImageData||L.ImageData,L.Video=o.Video||L.Video,L.fetch=o.fetch||L.fetch,L.readFile=o.readFile||L.readFile}var w={getEnv:Rn,setEnv:Zr,initialize:Kr,createBrowserEnv:Xr,createFileSystem:nr,createNodejsEnv:Jr,monkeyPatch:$n,isBrowser:qr,isNodejs:Se};Kr();function Ut(o){return!w.isNodejs()&&typeof o=="string"?document.getElementById(o):o}function O(o){let{Canvas:t,CanvasRenderingContext2D:e}=w.getEnv();if(o instanceof e)return o;let r=Ut(o);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let n=r.getContext("2d");if(!n)throw new Error("resolveContext2d - canvas 2d context is null");return n}var Qr=(n=>(n.TOP_LEFT="TOP_LEFT",n.TOP_RIGHT="TOP_RIGHT",n.BOTTOM_LEFT="BOTTOM_LEFT",n.BOTTOM_RIGHT="BOTTOM_RIGHT",n))(Qr||{}),fe=class{constructor(t={}){let{anchorPosition:e,backgroundColor:r,fontColor:n,fontSize:a,fontStyle:s,padding:i}=t;this.anchorPosition=e||"TOP_LEFT",this.backgroundColor=r||"rgba(0, 0, 0, 0.5)",this.fontColor=n||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=s||"Georgia",this.padding=i||4}},gt=class{constructor(t,e,r={}){this.text=typeof t=="string"?[t]:t instanceof gt?t.text:t,this.anchor=e,this.options=new fe(r)}measureWidth(t){let{padding:e}=this.options;return this.text.map(r=>t.measureText(r).width).reduce((r,n)=>r<n?n:r,0)+2*e}measureHeight(){let{fontSize:t,padding:e}=this.options;return this.text.length*t+2*e}getUpperLeft(t,e){let{anchorPosition:r}=this.options,n=r==="BOTTOM_RIGHT"||r==="TOP_RIGHT",a=r==="BOTTOM_LEFT"||r==="BOTTOM_RIGHT",s=this.measureWidth(t),i=this.measureHeight(),c=n?this.anchor.x-s:this.anchor.x,m=a?this.anchor.y-i:this.anchor.y;if(e){let{width:p,height:u}=e,f=Math.max(Math.min(c,p-s),0),l=Math.max(Math.min(m,u-i),0);return{x:f,y:l}}return{x:c,y:m}}draw(t){let e=Ut(t),r=O(e),{backgroundColor:n,fontColor:a,fontSize:s,fontStyle:i,padding:c}=this.options;r.font=`${s}px ${i}`;let m=this.measureWidth(r),p=this.measureHeight();r.fillStyle=n;let u=this.getUpperLeft(r,e);r.fillRect(u.x,u.y,m,p),r.fillStyle=a,this.text.forEach((f,l)=>{let b=c+u.x,y=c+u.y+(l+1)*s;r.fillText(f,b,y)})}};var ar=class{constructor(t={}){let{boxColor:e,lineWidth:r,label:n,drawLabelOptions:a}=t;this.boxColor=e||"rgba(0, 0, 255, 1)",this.lineWidth=r||2,this.label=n;let s={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new fe({...s,...a})}},Le=class{constructor(t,e={}){this.box=new D(t),this.options=new ar(e)}draw(t){let e=O(t),{boxColor:r,lineWidth:n}=this.options,{x:a,y:s,width:i,height:c}=this.box;e.strokeStyle=r,e.lineWidth=n,e.strokeRect(a,s,i,c);let{label:m}=this.options;m&&new gt([m],{x:a-n/2,y:s},this.options.drawLabelOptions).draw(t)}};function On(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof M?r.score:pt(r)?r.detection.score:void 0,a=r instanceof M?r.box:pt(r)?r.detection.box:new D(r),s=n?`${Ht(n)}`:void 0;new Le(a,{label:s}).draw(o)})}var we=v(x());function Ae(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t&&o.complete||o instanceof e&&o.readyState>=3}function to(o){return new Promise((t,e)=>{(o instanceof w.getEnv().Canvas||Ae(o))&&t(null);function r(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),e(a))}function n(a){!a.currentTarget||(a.currentTarget.removeEventListener("load",n),a.currentTarget.removeEventListener("error",r),t(a))}o.addEventListener("load",n),o.addEventListener("error",r)})}function eo(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToImage - expected buf to be of type: Blob"));let r=new FileReader;r.onload=()=>{typeof r.result!="string"&&e(new Error("bufferToImage - expected reader.result to be a string, in onload"));let n=w.getEnv().createImageElement();n.onload=()=>t(n),n.onerror=e,n.src=r.result},r.onerror=e,r.readAsDataURL(o)})}function Xt(o){let{Image:t,Video:e}=w.getEnv();return o instanceof t?new k(o.naturalWidth,o.naturalHeight):o instanceof e?new k(o.videoWidth,o.videoHeight):new k(o.width,o.height)}function Jt({width:o,height:t}){let{createCanvasElement:e}=w.getEnv(),r=e();return r.width=o,r.height=t,r}function We(o,t){let{ImageData:e}=w.getEnv();if(!(o instanceof e)&&!Ae(o))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:n}=t||Xt(o),a=Jt({width:r,height:n});return o instanceof e?O(a).putImageData(o,0,0):O(a).drawImage(o,0,0,r,n),a}var sr=v(x());async function ro(o,t){let e=t||w.getEnv().createCanvasElement(),[r,n,a]=o.shape.slice(U(o)?1:0),s=sr.tidy(()=>o.as3D(r,n,a).toInt());return await sr.browser.toPixels(s,e),s.dispose(),e}function ir(o){let{Image:t,Canvas:e,Video:r}=w.getEnv();return o instanceof t||o instanceof e||o instanceof r}var V=v(x());function oo(o,t,e=!1){let{Image:r,Canvas:n}=w.getEnv();if(!(o instanceof r||o instanceof n))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return Jt({width:1,height:1});let a=Xt(o),s=t/Math.max(a.height,a.width),i=s*a.width,c=s*a.height,m=Jt({width:t,height:t}),p=o instanceof n?o:We(o),u=Math.abs(i-c)/2,f=e&&i<c?u:0,l=e&&c<i?u:0;return p.width>0&&p.height>0&&O(m).drawImage(p,f,l,i,c),m}var ut=class{constructor(t,e=!1){this._imageTensors=[];this._canvases=[];this._treatAsBatchInput=!1;this._inputDimensions=[];this._inputSize=0;if(!Array.isArray(t))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${t}`);this._treatAsBatchInput=e,this._batchSize=t.length,t.forEach((r,n)=>{if(ht(r)){this._imageTensors[n]=r,this._inputDimensions[n]=r.shape;return}if(U(r)){let s=r.shape[0];if(s!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${s} passed, but not supported in input array`);this._imageTensors[n]=r,this._inputDimensions[n]=r.shape.slice(1);return}let a=r instanceof w.getEnv().Canvas?r:We(r);this._canvases[n]=a,this._inputDimensions[n]=[a.height,a.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return it(this.batchSize,0,1).map((t,e)=>this.getReshapedInputDimensions(e))}getInput(t){return this.canvases[t]||this.imageTensors[t]}getInputDimensions(t){return this._inputDimensions[t]}getInputHeight(t){return this._inputDimensions[t][0]}getInputWidth(t){return this._inputDimensions[t][1]}getReshapedInputDimensions(t){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");let e=this.getInputWidth(t),r=this.getInputHeight(t);return Or({width:e,height:r},this.inputSize)}toBatchTensor(t,e=!0){return this._inputSize=t,V.tidy(()=>{let r=it(this.batchSize,0,1).map(a=>{let s=this.getInput(a);if(s instanceof V.Tensor){let i=U(s)?s:V.expandDims(s);return i=Gr(i,e),(i.shape[1]!==t||i.shape[2]!==t)&&(i=V.image.resizeBilinear(i,[t,t],!1,!1)),i.as3D(t,t,3)}if(s instanceof w.getEnv().Canvas)return V.browser.fromPixels(oo(s,t,e));throw new Error(`toBatchTensor - at batchIdx ${a}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${s}`)});return V.stack(r.map(a=>V.cast(a,"float32"))).as4D(this.batchSize,t,t,3)})}};async function C(o){if(o instanceof ut)return o;let t=Array.isArray(o)?o:[o];if(!t.length)throw new Error("toNetInput - empty array passed as input");let e=n=>Array.isArray(o)?` at input index ${n}:`:"",r=t.map(Ut);return r.forEach((n,a)=>{if(!ir(n)&&!ht(n)&&!U(n))throw typeof t[a]=="string"?new Error(`toNetInput -${e(a)} string passed, but could not resolve HTMLElement for element id ${t[a]}`):new Error(`toNetInput -${e(a)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(U(n)){let s=n.shape[0];if(s!==1)throw new Error(`toNetInput -${e(a)} tf.Tensor4D with batchSize ${s} passed, but not supported in input array`)}}),await Promise.all(r.map(n=>ir(n)&&to(n))),new ut(r,Array.isArray(o))}async function le(o,t){let{Canvas:e}=w.getEnv(),r=o;if(!(o instanceof e)){let s=await C(o);if(s.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let i=s.getInput(0);r=i instanceof e?i:await ro(i)}let n=O(r);return t.map(s=>s instanceof M?s.forSize(r.width,r.height).box.floor():s).map(s=>s.clipAtImageBorders(r.width,r.height)).map(({x:s,y:i,width:c,height:m})=>{let p=Jt({width:c,height:m});return c>0&&m>0&&O(p).putImageData(n.getImageData(s,i,c,m),0,0),p})}var cr=v(x());async function de(o,t){if(!ht(o)&&!U(o))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(U(o)&&o.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return cr.tidy(()=>{let[e,r,n]=o.shape.slice(U(o)?1:0);return t.map(i=>i instanceof M?i.forSize(r,e).box:i).map(i=>i.clipAtImageBorders(r,e)).filter(i=>i.width>0&&i.height>0).map(({x:i,y:c,width:m,height:p})=>cr.slice3d(o.as3D(e,r,n),[c,i,0],[p,m,n]))})}async function xt(o,t){let{fetch:e}=w.getEnv(),r=await e(o,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}async function Hn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${e.type}, for url: ${t.url}`);return eo(e)}async function no(o){return(await xt(o)).json()}async function zn(o){return new Float32Array(await(await xt(o)).arrayBuffer())}function Fo(o){return new Promise((t,e)=>{o instanceof Blob||e(new Error("bufferToVideo - expected buf to be of type: Blob"));let r=w.getEnv().createVideoElement();r.oncanplay=()=>t(r),r.onerror=e,r.playsInline=!0,r.muted=!0,r.src=URL.createObjectURL(o),r.play()})}async function Vn(o){let t=await xt(o),e=await t.blob();if(!e.type.startsWith("video/"))throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${e.type}, for url: ${t.url}`);return Fo(e)}var Do=v(x());function mr(o,t){let e=`${t}-weights_manifest.json`;if(!o)return{modelBaseUri:"",manifestUri:e};if(o==="/")return{modelBaseUri:"/",manifestUri:`/${e}`};let r=o.startsWith("http://")?"http://":o.startsWith("https://")?"https://":"";o=o.replace(r,"");let n=o.split("/").filter(i=>i),a=o.endsWith(".json")?n[n.length-1]:e,s=r+(o.endsWith(".json")?n.slice(0,n.length-1):n).join("/");return s=o.startsWith("/")?`/${s}`:s,{modelBaseUri:s,manifestUri:s==="/"?`/${a}`:`${s}/${a}`}}async function ao(o,t){let{manifestUri:e,modelBaseUri:r}=mr(o,t),n=await no(e);return Do.io.loadWeights(n,r)}function Yn(o,t,e=!1){let{width:r,height:n}=e?Xt(t):t;return o.width=r,o.height=n,{width:r,height:n}}var ye=v(x());var vt=v(x());var A=class{constructor(t){this._params=void 0;this._paramMappings=[];this._name=t}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(t){let{obj:e,objProp:r}=this.traversePropertyPath(t);return e[r]}reassignParamFromPath(t,e){let{obj:r,objProp:n}=this.traversePropertyPath(t);r[n].dispose(),r[n]=e}getParamList(){return this._paramMappings.map(({paramPath:t})=>({path:t,tensor:this.getParamFromPath(t)}))}getTrainableParams(){return this.getParamList().filter(t=>t.tensor instanceof vt.Variable)}getFrozenParams(){return this.getParamList().filter(t=>!(t.tensor instanceof vt.Variable))}variable(){this.getFrozenParams().forEach(({path:t,tensor:e})=>{this.reassignParamFromPath(t,e.variable())})}freeze(){this.getTrainableParams().forEach(({path:t,tensor:e})=>{let r=vt.tensor(e.dataSync());e.dispose(),this.reassignParamFromPath(t,r)})}dispose(t=!0){this.getParamList().forEach(e=>{if(t&&e.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${e.path}`);e.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:t})=>Array.from(t.dataSync())).reduce((t,e)=>t.concat(e)))}async load(t){if(t instanceof Float32Array){this.extractWeights(t);return}await this.loadFromUri(t)}async loadFromUri(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);let e=await ao(t,this.getDefaultModelName());this.loadFromWeightMap(e)}async loadFromDisk(t){if(t&&typeof t!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:e}=w.getEnv(),{manifestUri:r,modelBaseUri:n}=mr(t,this.getDefaultModelName()),a=m=>Promise.all(m.map(p=>e(p).then(u=>u.buffer))),s=vt.io.weightsLoaderFactory(a),i=JSON.parse((await e(r)).toString()),c=await s(i,n);this.loadFromWeightMap(c)}loadFromWeightMap(t){let{paramMappings:e,params:r}=this.extractParamsFromWeightMap(t);this._paramMappings=e,this._params=r}extractWeights(t){let{paramMappings:e,params:r}=this.extractParams(t);this._paramMappings=e,this._params=r}traversePropertyPath(t){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");let e=t.split("/").reduce((a,s)=>{if(!a.nextObj.hasOwnProperty(s))throw new Error(`traversePropertyPath - object does not have property ${s}, for path ${t}`);return{obj:a.nextObj,objProp:s,nextObj:a.nextObj[s]}},{nextObj:this.params}),{obj:r,objProp:n}=e;if(!r||!n||!(r[n]instanceof vt.Tensor))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${t}`);return{obj:r,objProp:n}}};var I=v(x());var he=v(x());function H(o,t,e){return he.tidy(()=>{let r=he.separableConv2d(o,t.depthwise_filter,t.pointwise_filter,e,"same");return r=he.add(r,t.bias),r})}function pr(o,t,e=!1){return I.tidy(()=>{let r=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,[2,2],"same"),t.conv0.bias):H(o,t.conv0,[2,2])),n=H(r,t.conv1,[1,1]),a=I.relu(I.add(r,n)),s=H(a,t.conv2,[1,1]);return I.relu(I.add(r,I.add(n,s)))})}function ke(o,t,e=!1,r=!0){return I.tidy(()=>{let n=I.relu(e?I.add(I.conv2d(o,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):H(o,t.conv0,r?[2,2]:[1,1])),a=H(n,t.conv1,[1,1]),s=I.relu(I.add(n,a)),i=H(s,t.conv2,[1,1]),c=I.relu(I.add(n,I.add(a,i))),m=H(c,t.conv3,[1,1]);return I.relu(I.add(n,I.add(a,I.add(i,m))))})}var It=v(x());function qt(o,t,e="same",r=!1){return It.tidy(()=>{let n=It.add(It.conv2d(o,t.filters,[1,1],e),t.bias);return r?It.relu(n):n})}function B(o,t){Object.keys(o).forEach(e=>{t.some(r=>r.originalPath===e)||o[e].dispose()})}var ur=v(x());function be(o,t){return(e,r,n,a)=>{let s=ur.tensor4d(o(e*r*n*n),[n,n,e,r]),i=ur.tensor1d(o(r));return t.push({paramPath:`${a}/filters`},{paramPath:`${a}/bias`}),{filters:s,bias:i}}}var fr=v(x());function lr(o,t){return(e,r,n)=>{let a=fr.tensor2d(o(e*r),[e,r]),s=fr.tensor1d(o(r));return t.push({paramPath:`${n}/weights`},{paramPath:`${n}/bias`}),{weights:a,bias:s}}}var Re=v(x());var Be=class{constructor(t,e,r){this.depthwise_filter=t;this.pointwise_filter=e;this.bias=r}};function ge(o,t){return(e,r,n)=>{let a=Re.tensor4d(o(9*e),[3,3,e,1]),s=Re.tensor4d(o(e*r),[1,1,e,r]),i=Re.tensor1d(o(r));return t.push({paramPath:`${n}/depthwise_filter`},{paramPath:`${n}/pointwise_filter`},{paramPath:`${n}/bias`}),new Be(a,s,i)}}function xe(o){return t=>{let e=o(`${t}/depthwise_filter`,4),r=o(`${t}/pointwise_filter`,4),n=o(`${t}/bias`,1);return new Be(e,r,n)}}function Y(o,t){return(e,r,n)=>{let a=o[e];if(!Ot(a,r))throw new Error(`expected weightMap[${e}] to be a Tensor${r}D, instead have ${a}`);return t.push({originalPath:e,paramPath:n||e}),a}}function R(o){let t=o;function e(n){let a=t.slice(0,n);return t=t.slice(n),a}function r(){return t}return{extractWeights:e,getRemainingWeights:r}}function dr(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c,m=!1){let p=m?e(s,i,3,`${c}/conv0`):r(s,i,`${c}/conv0`),u=r(i,i,`${c}/conv1`),f=r(i,i,`${c}/conv2`);return{conv0:p,conv1:u,conv2:f}}function a(s,i,c,m=!1){let{conv0:p,conv1:u,conv2:f}=n(s,i,c,m),l=r(i,i,`${c}/conv3`);return{conv0:p,conv1:u,conv2:f,conv3:l}}return{extractDenseBlock3Params:n,extractDenseBlock4Params:a}}function Eo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock4Params:n}=dr(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2"),c=n(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i,dense3:c}}}function hr(o){return t=>{let e=o(`${t}/filters`,4),r=o(`${t}/bias`,1);return{filters:e,bias:r}}}function br(o,t){let e=Y(o,t),r=hr(e),n=xe(e);function a(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`);return{conv0:m,conv1:p,conv2:u}}function s(i,c=!1){let m=c?r(`${i}/conv0`):n(`${i}/conv0`),p=n(`${i}/conv1`),u=n(`${i}/conv2`),f=n(`${i}/conv3`);return{conv0:m,conv1:p,conv2:u,conv3:f}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:s}}function Mo(o){let t=[],{extractDenseBlock4Params:e}=br(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2"),dense3:e("dense3")};return B(o,t),{params:r,paramMappings:t}}var ve=class extends A{constructor(){super("FaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceFeatureExtractor - load model before inference");return ye.tidy(()=>{let r=ye.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=ke(a,e.dense0,!0);return s=ke(s,e.dense1),s=ke(s,e.dense2),s=ke(s,e.dense3),s=ye.avgPool(s,[7,7],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(t){return Mo(t)}extractParams(t){return Eo(t)}};var No=v(x());var _e=v(x());function $e(o,t){return _e.tidy(()=>_e.add(_e.matMul(o,t.weights),t.bias))}function Co(o,t,e){let r=[],{extractWeights:n,getRemainingWeights:a}=R(o),i=lr(n,r)(t,e,"fc");if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:r,params:{fc:i}}}function Io(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:r("fc")};return B(o,t),{params:n,paramMappings:t}}function gr(o){let t={},e={};return Object.keys(o).forEach(r=>{let n=r.startsWith("fc")?e:t;n[r]=o[r]}),{featureExtractorMap:t,classifierMap:e}}var Te=class extends A{constructor(e,r){super(e);this._faceFeatureExtractor=r}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return No.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e;return $e(n.as2D(n.shape[0],-1),r.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return Co(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=gr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Io(n)}extractParams(e){let r=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),a=n*r+n,s=e.slice(0,e.length-a),i=e.slice(e.length-a);return this.faceFeatureExtractor.extractWeights(s),this.extractClassifierParams(i)}};var so=["neutral","happy","sad","angry","fearful","disgusted","surprised"],yt=class{constructor(t){this.neutral=0;this.happy=0;this.sad=0;this.angry=0;this.fearful=0;this.disgusted=0;this.surprised=0;if(t.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${t.length}`);so.forEach((e,r)=>{this[e]=t[r]})}asSortedArray(){return so.map(t=>({expression:t,probability:this[t]})).sort((t,e)=>e.probability-t.probability)}};var Oe=class extends Te{constructor(t=new ve){super("FaceExpressionNet",t)}forwardInput(t){return we.tidy(()=>we.softmax(this.runNet(t)))}async forward(t){return this.forwardInput(await C(t))}async predictExpressions(t){let e=await C(t),r=await this.forwardInput(e),n=await Promise.all(we.unstack(r).map(async s=>{let i=s.dataSync();return s.dispose(),i}));r.dispose();let a=n.map(s=>new yt(s));return e.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}};function io(o){return o.expressions instanceof yt}function xr(o,t){return{...o,...{expressions:t}}}function Gn(o,t,e=.1,r){(Array.isArray(t)?t:[t]).forEach(a=>{let s=a instanceof yt?a:io(a)?a.expressions:void 0;if(!s)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let c=s.asSortedArray().filter(u=>u.probability>e),m=pt(a)?a.detection.box.bottomLeft:r||new g(0,0);new gt(c.map(u=>`${u.expression} (${Ht(u.probability)})`),m).draw(o)})}function Zt(o){return pt(o)&&o.landmarks instanceof z&&o.unshiftedLandmarks instanceof z&&o.alignedRect instanceof M}function jn(o){let t=c=>c*180/Math.PI,e=(c,m)=>Math.sqrt((c._x-m._x)**2+(c._y-m._y)**2),r={roll:void 0,pitch:void 0,yaw:void 0},n=(c,m,p)=>{let u=Math.floor(c._x-m._x),f=Math.floor(m._x-p._x);return u-f},a=(c,m)=>{let p=Math.hypot(m._x-c._x,m._y-c._y),u=m._y-c._y,f=Math.asin(u/p),l=t(f),b=Math.floor(90-l),y=m._x-c._x<0?-1:1;return b*y},s=(c,m,p)=>{let u=e(c,p),f={_x:(c._x+p._x)/2,_y:(c._y+p._y)/2},l=e(m,f),b=Math.atan(l/u),y=Math.floor(t(b)),F=f._y-m._y<0?-1:1;return y*F};if(!o||!o._positions||o._positions.length!==68)return r;let i=o._positions;return r.roll=a(i[27],i[66]),r.pitch=s(i[14],i[30],i[2]),r.yaw=n(i[14],i[33],i[2]),r}function Pe(o,t){let{box:e}=o.detection,r=t.shiftBy(e.x,e.y),n=r.align(),{imageDims:a}=o.detection,s=new M(o.detection.score,n.rescale(a.reverse()),a),i=jn(t);return{...o,...{landmarks:r,unshiftedLandmarks:t,alignedRect:s,angle:i}}}var vr=class{constructor(t={}){let{drawLines:e=!0,drawPoints:r=!0,lineWidth:n,lineColor:a,pointSize:s,pointColor:i}=t;this.drawLines=e,this.drawPoints=r,this.lineWidth=n||1,this.pointSize=s||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},yr=class{constructor(t,e={}){this.faceLandmarks=t,this.options=new vr(e)}draw(t){let e=O(t),{drawLines:r,drawPoints:n,lineWidth:a,lineColor:s,pointSize:i,pointColor:c}=this.options;if(r&&this.faceLandmarks instanceof Gt&&(e.strokeStyle=s,e.lineWidth=a,dt(e,this.faceLandmarks.getJawOutline()),dt(e,this.faceLandmarks.getLeftEyeBrow()),dt(e,this.faceLandmarks.getRightEyeBrow()),dt(e,this.faceLandmarks.getNose()),dt(e,this.faceLandmarks.getLeftEye(),!0),dt(e,this.faceLandmarks.getRightEye(),!0),dt(e,this.faceLandmarks.getMouth(),!0)),n){e.strokeStyle=c,e.fillStyle=c;let m=p=>{e.beginPath(),e.arc(p.x,p.y,i,0,2*Math.PI),e.fill()};this.faceLandmarks.positions.forEach(m)}}};function Un(o,t){(Array.isArray(t)?t:[t]).forEach(r=>{let n=r instanceof z?r:Zt(r)?r.landmarks:void 0;if(!n)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new yr(n).draw(o)})}var So="1.7.6";var ft=v(x());var S=v(x());function qn(o,t){let e=be(o,t),r=ge(o,t);function n(s,i,c){let m=r(s,i,`${c}/separable_conv0`),p=r(i,i,`${c}/separable_conv1`),u=e(s,i,1,`${c}/expansion_conv`);return{separable_conv0:m,separable_conv1:p,expansion_conv:u}}function a(s,i){let c=r(s,s,`${i}/separable_conv0`),m=r(s,s,`${i}/separable_conv1`),p=r(s,s,`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:e,extractSeparableConvParams:r,extractReductionBlockParams:n,extractMainBlockParams:a}}function Lo(o,t){let e=[],{extractWeights:r,getRemainingWeights:n}=R(o),{extractConvParams:a,extractSeparableConvParams:s,extractReductionBlockParams:i,extractMainBlockParams:c}=qn(r,e),m=a(3,32,3,"entry_flow/conv_in"),p=i(32,64,"entry_flow/reduction_block_0"),u=i(64,128,"entry_flow/reduction_block_1"),f={conv_in:m,reduction_block_0:p,reduction_block_1:u},l={};it(t,0,1).forEach(h=>{l[`main_block_${h}`]=c(128,`middle_flow/main_block_${h}`)});let b=i(128,256,"exit_flow/reduction_block"),y=s(256,512,"exit_flow/separable_conv"),F={reduction_block:b,separable_conv:y};if(n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);return{paramMappings:e,params:{entry_flow:f,middle_flow:l,exit_flow:F}}}function Zn(o,t){let e=Y(o,t),r=hr(e),n=xe(e);function a(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=r(`${i}/expansion_conv`);return{separable_conv0:c,separable_conv1:m,expansion_conv:p}}function s(i){let c=n(`${i}/separable_conv0`),m=n(`${i}/separable_conv1`),p=n(`${i}/separable_conv2`);return{separable_conv0:c,separable_conv1:m,separable_conv2:p}}return{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}}function Ao(o,t){let e=[],{extractConvParams:r,extractSeparableConvParams:n,extractReductionBlockParams:a,extractMainBlockParams:s}=Zn(o,e),i=r("entry_flow/conv_in"),c=a("entry_flow/reduction_block_0"),m=a("entry_flow/reduction_block_1"),p={conv_in:i,reduction_block_0:c,reduction_block_1:m},u={};it(t,0,1).forEach(y=>{u[`main_block_${y}`]=s(`middle_flow/main_block_${y}`)});let f=a("exit_flow/reduction_block"),l=n("exit_flow/separable_conv"),b={reduction_block:f,separable_conv:l};return B(o,e),{params:{entry_flow:p,middle_flow:u,exit_flow:b},paramMappings:e}}function Wo(o,t,e){return S.add(S.conv2d(o,t.filters,e,"same"),t.bias)}function mo(o,t,e=!0){let r=e?S.relu(o):o;return r=H(r,t.separable_conv0,[1,1]),r=H(S.relu(r),t.separable_conv1,[1,1]),r=S.maxPool(r,[3,3],[2,2],"same"),r=S.add(r,Wo(o,t.expansion_conv,[2,2])),r}function Kn(o,t){let e=H(S.relu(o),t.separable_conv0,[1,1]);return e=H(S.relu(e),t.separable_conv1,[1,1]),e=H(S.relu(e),t.separable_conv2,[1,1]),e=S.add(e,o),e}var _r=class extends A{constructor(e){super("TinyXception");this._numMainBlocks=e}forwardInput(e){let{params:r}=this;if(!r)throw new Error("TinyXception - load model before inference");return S.tidy(()=>{let n=S.cast(e.toBatchTensor(112,!0),"float32"),s=rt(n,[122.782,117.001,104.298]).div(255),i=S.relu(Wo(s,r.entry_flow.conv_in,[2,2]));return i=mo(i,r.entry_flow.reduction_block_0,!1),i=mo(i,r.entry_flow.reduction_block_1),it(this._numMainBlocks,0,1).forEach(c=>{i=Kn(i,r.middle_flow[`main_block_${c}`])}),i=mo(i,r.exit_flow.reduction_block),i=S.relu(H(i,r.exit_flow.separable_conv,[1,1])),i})}async forward(e){return this.forwardInput(await C(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return Ao(e,this._numMainBlocks)}extractParams(e){return Lo(e,this._numMainBlocks)}};function ko(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),n=lr(e,t),a=n(512,1,"fc/age"),s=n(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:a,gender:s}}}}function Bo(o){let t=[],e=Y(o,t);function r(a){let s=e(`${a}/weights`,2),i=e(`${a}/bias`,1);return{weights:s,bias:i}}let n={fc:{age:r("fc/age"),gender:r("fc/gender")}};return B(o,t),{params:n,paramMappings:t}}var Tr=(e=>(e.FEMALE="female",e.MALE="male",e))(Tr||{});var He=class extends A{constructor(e=new _r(2)){super("AgeGenderNet");this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:r}=this;if(!r)throw new Error(`${this._name} - load model before inference`);return ft.tidy(()=>{let n=e instanceof ut?this.faceFeatureExtractor.forwardInput(e):e,a=ft.avgPool(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1),s=$e(a,r.fc.age).as1D(),i=$e(a,r.fc.gender);return{age:s,gender:i}})}forwardInput(e){return ft.tidy(()=>{let{age:r,gender:n}=this.runNet(e);return{age:r,gender:ft.softmax(n)}})}async forward(e){return this.forwardInput(await C(e))}async predictAgeAndGender(e){let r=await C(e),n=await this.forwardInput(r),a=ft.unstack(n.age),s=ft.unstack(n.gender),i=a.map((m,p)=>({ageTensor:m,genderTensor:s[p]})),c=await Promise.all(i.map(async({ageTensor:m,genderTensor:p})=>{let u=m.dataSync()[0],f=p.dataSync()[0],l=f>.5,b=l?"male":"female",y=l?f:1-f;return m.dispose(),p.dispose(),{age:u,gender:b,genderProbability:y}}));return n.age.dispose(),n.gender.dispose(),r.isBatchInput?c:c[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:r,paramMappings:n}=this.extractClassifierParams(e);this._params=r,this._paramMappings=n}extractClassifierParams(e){return ko(e)}extractParamsFromWeightMap(e){let{featureExtractorMap:r,classifierMap:n}=gr(e);return this.faceFeatureExtractor.loadFromWeightMap(r),Bo(n)}extractParams(e){let n=e.slice(0,e.length-1539),a=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(n),this.extractClassifierParams(a)}};var G=v(x());var Fe=class extends Te{postProcess(t,e,r){let n=r.map(({width:s,height:i})=>{let c=e/Math.max(i,s);return{width:s*c,height:i*c}}),a=n.length;return G.tidy(()=>{let s=(u,f)=>G.stack([G.fill([68],u,"float32"),G.fill([68],f,"float32")],1).as2D(1,136).as1D(),i=(u,f)=>{let{width:l,height:b}=n[u];return f(l,b)?Math.abs(l-b)/2:0},c=u=>i(u,(f,l)=>f<l),m=u=>i(u,(f,l)=>l<f);return t.mul(G.fill([a,136],e,"float32")).sub(G.stack(Array.from(Array(a),(u,f)=>s(c(f),m(f))))).div(G.stack(Array.from(Array(a),(u,f)=>s(n[f].width,n[f].height))))})}forwardInput(t){return G.tidy(()=>{let e=this.runNet(t);return this.postProcess(e,t.inputSize,t.inputDimensions.map(([r,n])=>({height:r,width:n})))})}async forward(t){return this.forwardInput(await C(t))}async detectLandmarks(t){let e=await C(t),r=G.tidy(()=>G.unstack(this.forwardInput(e))),n=await Promise.all(r.map(async(a,s)=>{let i=Array.from(a.dataSync()),c=i.filter((p,u)=>rr(u)),m=i.filter((p,u)=>!rr(u));return new Gt(Array(68).fill(0).map((p,u)=>new g(c[u],m[u])),{height:e.getInputHeight(s),width:e.getInputWidth(s)})}));return r.forEach(a=>a.dispose()),e.isBatchInput?n:n[0]}getClassifierChannelsOut(){return 136}};var Kt=class extends Fe{constructor(t=new ve){super("FaceLandmark68Net",t)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};var De=v(x());function Ro(o){let t=[],{extractDenseBlock3Params:e}=br(o,t),r={dense0:e("dense0",!0),dense1:e("dense1"),dense2:e("dense2")};return B(o,t),{params:r,paramMappings:t}}function $o(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractDenseBlock3Params:n}=dr(e,t),a=n(3,32,"dense0",!0),s=n(32,64,"dense1"),i=n(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:a,dense1:s,dense2:i}}}var wr=class extends A{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("TinyFaceFeatureExtractor - load model before inference");return De.tidy(()=>{let r=De.cast(t.toBatchTensor(112,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=pr(a,e.dense0,!0);return s=pr(s,e.dense1),s=pr(s,e.dense2),s=De.avgPool(s,[14,14],[2,2],"valid"),s})}async forward(t){return this.forwardInput(await C(t))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(t){return Ro(t)}extractParams(t){return $o(t)}};var ze=class extends Fe{constructor(t=new wr){super("FaceLandmark68TinyNet",t)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}};var po=class extends Kt{};var nt=v(x());var Ee=v(x());var Pr=v(x());function Oo(o,t){return Pr.add(Pr.mul(o,t.weights),t.biases)}function uo(o,t,e,r,n="same"){let{filters:a,bias:s}=t.conv,i=Ee.conv2d(o,a,e,n);return i=Ee.add(i,s),i=Oo(i,t.scale),r?Ee.relu(i):i}function Ho(o,t){return uo(o,t,[1,1],!0)}function fo(o,t){return uo(o,t,[1,1],!1)}function Fr(o,t){return uo(o,t,[2,2],!0,"valid")}var j=v(x());function Qn(o,t){function e(i,c,m){let p=o(i),u=p.length/(c*m*m);if($r(u))throw new Error(`depth has to be an integer: ${u}, weights.length: ${p.length}, numFilters: ${c}, filterSize: ${m}`);return j.tidy(()=>j.transpose(j.tensor4d(p,[c,u,m,m]),[2,3,1,0]))}function r(i,c,m,p){let u=e(i,c,m),f=j.tensor1d(o(c));return t.push({paramPath:`${p}/filters`},{paramPath:`${p}/bias`}),{filters:u,bias:f}}function n(i,c){let m=j.tensor1d(o(i)),p=j.tensor1d(o(i));return t.push({paramPath:`${c}/weights`},{paramPath:`${c}/biases`}),{weights:m,biases:p}}function a(i,c,m,p){let u=r(i,c,m,`${p}/conv`),f=n(c,`${p}/scale`);return{conv:u,scale:f}}function s(i,c,m,p,u=!1){let f=a((u?.5:1)*i,c,m,`${p}/conv1`),l=a(i,c,m,`${p}/conv2`);return{conv1:f,conv2:l}}return{extractConvLayerParams:a,extractResidualLayerParams:s}}function zo(o){let{extractWeights:t,getRemainingWeights:e}=R(o),r=[],{extractConvLayerParams:n,extractResidualLayerParams:a}=Qn(t,r),s=n(4704,32,7,"conv32_down"),i=a(9216,32,3,"conv32_1"),c=a(9216,32,3,"conv32_2"),m=a(9216,32,3,"conv32_3"),p=a(36864,64,3,"conv64_down",!0),u=a(36864,64,3,"conv64_1"),f=a(36864,64,3,"conv64_2"),l=a(36864,64,3,"conv64_3"),b=a(147456,128,3,"conv128_down",!0),y=a(147456,128,3,"conv128_1"),F=a(147456,128,3,"conv128_2"),h=a(589824,256,3,"conv256_down",!0),T=a(589824,256,3,"conv256_1"),_=a(589824,256,3,"conv256_2"),E=a(589824,256,3,"conv256_down_out"),W=j.tidy(()=>j.transpose(j.tensor2d(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),e().length!==0)throw new Error(`weights remaing after extract: ${e().length}`);return{params:{conv32_down:s,conv32_1:i,conv32_2:c,conv32_3:m,conv64_down:p,conv64_1:u,conv64_2:f,conv64_3:l,conv128_down:b,conv128_1:y,conv128_2:F,conv256_down:h,conv256_1:T,conv256_2:_,conv256_down_out:E,fc:W},paramMappings:r}}function ta(o,t){let e=Y(o,t);function r(s){let i=e(`${s}/scale/weights`,1),c=e(`${s}/scale/biases`,1);return{weights:i,biases:c}}function n(s){let i=e(`${s}/conv/filters`,4),c=e(`${s}/conv/bias`,1),m=r(s);return{conv:{filters:i,bias:c},scale:m}}function a(s){return{conv1:n(`${s}/conv1`),conv2:n(`${s}/conv2`)}}return{extractConvLayerParams:n,extractResidualLayerParams:a}}function Vo(o){let t=[],{extractConvLayerParams:e,extractResidualLayerParams:r}=ta(o,t),n=e("conv32_down"),a=r("conv32_1"),s=r("conv32_2"),i=r("conv32_3"),c=r("conv64_down"),m=r("conv64_1"),p=r("conv64_2"),u=r("conv64_3"),f=r("conv128_down"),l=r("conv128_1"),b=r("conv128_2"),y=r("conv256_down"),F=r("conv256_1"),h=r("conv256_2"),T=r("conv256_down_out"),{fc:_}=o;if(t.push({originalPath:"fc",paramPath:"fc"}),!Rr(_))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${_}`);let E={conv32_down:n,conv32_1:a,conv32_2:s,conv32_3:i,conv64_down:c,conv64_1:m,conv64_2:p,conv64_3:u,conv128_down:f,conv128_1:l,conv128_2:b,conv256_down:y,conv256_1:F,conv256_2:h,conv256_down_out:T,fc:_};return B(o,t),{params:E,paramMappings:t}}var $=v(x());function ot(o,t){let e=Ho(o,t.conv1);return e=fo(e,t.conv2),e=$.add(e,o),e=$.relu(e),e}function Ve(o,t){let e=Fr(o,t.conv1);e=fo(e,t.conv2);let r=$.avgPool(o,2,2,"valid"),n=$.zeros(r.shape),a=r.shape[3]!==e.shape[3];if(r.shape[1]!==e.shape[1]||r.shape[2]!==e.shape[2]){let i=[...e.shape];i[1]=1;let c=$.zeros(i);e=$.concat([e,c],1);let m=[...e.shape];m[2]=1;let p=$.zeros(m);e=$.concat([e,p],2)}return r=a?$.concat([r,n],3):r,e=$.add(r,e),e=$.relu(e),e}var Qt=class extends A{constructor(){super("FaceRecognitionNet")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("FaceRecognitionNet - load model before inference");return nt.tidy(()=>{let r=nt.cast(t.toBatchTensor(150,!0),"float32"),a=rt(r,[122.782,117.001,104.298]).div(255),s=Fr(a,e.conv32_down);s=nt.maxPool(s,3,2,"valid"),s=ot(s,e.conv32_1),s=ot(s,e.conv32_2),s=ot(s,e.conv32_3),s=Ve(s,e.conv64_down),s=ot(s,e.conv64_1),s=ot(s,e.conv64_2),s=ot(s,e.conv64_3),s=Ve(s,e.conv128_down),s=ot(s,e.conv128_1),s=ot(s,e.conv128_2),s=Ve(s,e.conv256_down),s=ot(s,e.conv256_1),s=ot(s,e.conv256_2),s=Ve(s,e.conv256_down_out);let i=s.mean([1,2]);return nt.matMul(i,e.fc)})}async forward(t){return this.forwardInput(await C(t))}async computeFaceDescriptor(t){var a;if((a=t==null?void 0:t.shape)!=null&&a.some(s=>s<=0))return new Float32Array(128);let e=await C(t),r=nt.tidy(()=>nt.unstack(this.forwardInput(e))),n=await Promise.all(r.map(s=>s.data()));return r.forEach(s=>s.dispose()),e.isBatchInput?n:n[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(t){return Vo(t)}extractParams(t){return zo(t)}};function ea(o){let t=new Qt;return t.extractWeights(o),t}function Dr(o,t){return{...o,...{descriptor:t}}}function ra(o){return typeof o.age=="number"}function Er(o,t){return{...o,...{age:t}}}function oa(o){return(o.gender==="male"||o.gender==="female")&&me(o.genderProbability)}function Mr(o,t,e){return{...o,...{gender:t,genderProbability:e}}}var Lt=v(x());var at=v(x());function na(o,t){function e(c,m){let p=at.tensor4d(o(9*c),[3,3,c,1]),u=at.tensor1d(o(c)),f=at.tensor1d(o(c)),l=at.tensor1d(o(c)),b=at.tensor1d(o(c));return t.push({paramPath:`${m}/filters`},{paramPath:`${m}/batch_norm_scale`},{paramPath:`${m}/batch_norm_offset`},{paramPath:`${m}/batch_norm_mean`},{paramPath:`${m}/batch_norm_variance`}),{filters:p,batch_norm_scale:u,batch_norm_offset:f,batch_norm_mean:l,batch_norm_variance:b}}function r(c,m,p,u,f){let l=at.tensor4d(o(c*m*p*p),[p,p,c,m]),b=at.tensor1d(o(m));return t.push({paramPath:`${u}/filters`},{paramPath:`${u}/${f?"batch_norm_offset":"bias"}`}),{filters:l,bias:b}}function n(c,m,p,u){let{filters:f,bias:l}=r(c,m,p,u,!0);return{filters:f,batch_norm_offset:l}}function a(c,m,p){let u=e(c,`${p}/depthwise_conv`),f=n(c,m,1,`${p}/pointwise_conv`);return{depthwise_conv:u,pointwise_conv:f}}function s(){let c=n(3,32,3,"mobilenetv1/conv_0"),m=a(32,64,"mobilenetv1/conv_1"),p=a(64,128,"mobilenetv1/conv_2"),u=a(128,128,"mobilenetv1/conv_3"),f=a(128,256,"mobilenetv1/conv_4"),l=a(256,256,"mobilenetv1/conv_5"),b=a(256,512,"mobilenetv1/conv_6"),y=a(512,512,"mobilenetv1/conv_7"),F=a(512,512,"mobilenetv1/conv_8"),h=a(512,512,"mobilenetv1/conv_9"),T=a(512,512,"mobilenetv1/conv_10"),_=a(512,512,"mobilenetv1/conv_11"),E=a(512,1024,"mobilenetv1/conv_12"),W=a(1024,1024,"mobilenetv1/conv_13");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,conv_8:F,conv_9:h,conv_10:T,conv_11:_,conv_12:E,conv_13:W}}function i(){let c=n(1024,256,1,"prediction_layer/conv_0"),m=n(256,512,3,"prediction_layer/conv_1"),p=n(512,128,1,"prediction_layer/conv_2"),u=n(128,256,3,"prediction_layer/conv_3"),f=n(256,128,1,"prediction_layer/conv_4"),l=n(128,256,3,"prediction_layer/conv_5"),b=n(256,64,1,"prediction_layer/conv_6"),y=n(64,128,3,"prediction_layer/conv_7"),F=r(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),h=r(512,9,1,"prediction_layer/box_predictor_0/class_predictor"),T=r(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),_=r(1024,18,1,"prediction_layer/box_predictor_1/class_predictor"),E=r(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),W=r(512,18,1,"prediction_layer/box_predictor_2/class_predictor"),tt=r(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),lt=r(256,18,1,"prediction_layer/box_predictor_3/class_predictor"),q=r(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),Dt=r(256,18,1,"prediction_layer/box_predictor_4/class_predictor"),Et=r(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),Mt=r(128,18,1,"prediction_layer/box_predictor_5/class_predictor");return{conv_0:c,conv_1:m,conv_2:p,conv_3:u,conv_4:f,conv_5:l,conv_6:b,conv_7:y,box_predictor_0:{box_encoding_predictor:F,class_predictor:h},box_predictor_1:{box_encoding_predictor:T,class_predictor:_},box_predictor_2:{box_encoding_predictor:E,class_predictor:W},box_predictor_3:{box_encoding_predictor:tt,class_predictor:lt},box_predictor_4:{box_encoding_predictor:q,class_predictor:Dt},box_predictor_5:{box_encoding_predictor:Et,class_predictor:Mt}}}return{extractMobilenetV1Params:s,extractPredictionLayerParams:i}}function Yo(o){let t=[],{extractWeights:e,getRemainingWeights:r}=R(o),{extractMobilenetV1Params:n,extractPredictionLayerParams:a}=na(e,t),s=n(),i=a(),m={extra_dim:at.tensor3d(e(5118*4),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:s,prediction_layer:i,output_layer:m},paramMappings:t}}function aa(o,t){let e=Y(o,t);function r(m,p,u){let f=e(`${m}/Conv2d_${p}_pointwise/weights`,4,`${u}/filters`),l=e(`${m}/Conv2d_${p}_pointwise/convolution_bn_offset`,1,`${u}/batch_norm_offset`);return{filters:f,batch_norm_offset:l}}function n(m){let p=`mobilenetv1/conv_${m}`,u=`MobilenetV1/Conv2d_${m}_depthwise`,f=`${p}/depthwise_conv`,l=`${p}/pointwise_conv`,b=e(`${u}/depthwise_weights`,4,`${f}/filters`),y=e(`${u}/BatchNorm/gamma`,1,`${f}/batch_norm_scale`),F=e(`${u}/BatchNorm/beta`,1,`${f}/batch_norm_offset`),h=e(`${u}/BatchNorm/moving_mean`,1,`${f}/batch_norm_mean`),T=e(`${u}/BatchNorm/moving_variance`,1,`${f}/batch_norm_variance`);return{depthwise_conv:{filters:b,batch_norm_scale:y,batch_norm_offset:F,batch_norm_mean:h,batch_norm_variance:T},pointwise_conv:r("MobilenetV1",m,l)}}function a(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:n(1),conv_2:n(2),conv_3:n(3),conv_4:n(4),conv_5:n(5),conv_6:n(6),conv_7:n(7),conv_8:n(8),conv_9:n(9),conv_10:n(10),conv_11:n(11),conv_12:n(12),conv_13:n(13)}}function s(m,p){let u=e(`${m}/weights`,4,`${p}/filters`),f=e(`${m}/biases`,1,`${p}/bias`);return{filters:u,bias:f}}function i(m){let p=s(`Prediction/BoxPredictor_${m}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${m}/box_encoding_predictor`),u=s(`Prediction/BoxPredictor_${m}/ClassPredictor`,`prediction_layer/box_predictor_${m}/class_predictor`);return{box_encoding_predictor:p,class_predictor:u}}function c(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}return{extractMobilenetV1Params:a,extractPredictionLayerParams:c}}function Go(o){let t=[],{extractMobilenetV1Params:e,extractPredictionLayerParams:r}=aa(o,t),n=o["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!ht(n))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${n}`);let a={mobilenetv1:e(),prediction_layer:r(),output_layer:{extra_dim:n}};return B(o,t),{params:a,paramMappings:t}}var _t=v(x());var Nt=v(x());function Z(o,t,e){return Nt.tidy(()=>{let r=Nt.conv2d(o,t.filters,e,"same");return r=Nt.add(r,t.batch_norm_offset),Nt.clipByValue(r,0,6)})}var sa=.0010000000474974513;function ia(o,t,e){return _t.tidy(()=>{let r=_t.depthwiseConv2d(o,t.filters,e,"same");return r=_t.batchNorm(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,sa),_t.clipByValue(r,0,6)})}function ca(o){return[2,4,6,12].some(t=>t===o)?[2,2]:[1,1]}function jo(o,t){return _t.tidy(()=>{let e,r=Z(o,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((a,s)=>{let i=s+1,c=ca(i);r=ia(r,a.depthwise_conv,c),r=Z(r,a.pointwise_conv,[1,1]),i===11&&(e=r)}),e===null)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:e}})}function ma(o,t,e){let r=o.arraySync(),n=Math.min(r[t][0],r[t][2]),a=Math.min(r[t][1],r[t][3]),s=Math.max(r[t][0],r[t][2]),i=Math.max(r[t][1],r[t][3]),c=Math.min(r[e][0],r[e][2]),m=Math.min(r[e][1],r[e][3]),p=Math.max(r[e][0],r[e][2]),u=Math.max(r[e][1],r[e][3]),f=(s-n)*(i-a),l=(p-c)*(u-m);if(f<=0||l<=0)return 0;let b=Math.max(n,c),y=Math.max(a,m),F=Math.min(s,p),h=Math.min(i,u),T=Math.max(F-b,0)*Math.max(h-y,0);return T/(f+l-T)}function Uo(o,t,e,r,n){let a=o.shape[0],s=Math.min(e,a),i=t.map((p,u)=>({score:p,boxIndex:u})).filter(p=>p.score>n).sort((p,u)=>u.score-p.score),c=p=>p<=r?1:0,m=[];return i.forEach(p=>{if(m.length>=s)return;let u=p.score;for(let f=m.length-1;f>=0;--f){let l=ma(o,p.boxIndex,m[f]);if(l!==0&&(p.score*=c(l),p.score<=n))break}u===p.score&&m.push(p.boxIndex)}),m}var d=v(x());function pa(o){let t=d.unstack(d.transpose(o,[1,0])),e=[d.sub(t[2],t[0]),d.sub(t[3],t[1])],r=[d.add(t[0],d.div(e[0],2)),d.add(t[1],d.div(e[1],2))];return{sizes:e,centers:r}}function ua(o,t){let{sizes:e,centers:r}=pa(o),n=d.unstack(d.transpose(t,[1,0])),a=d.div(d.mul(d.exp(d.div(n[2],5)),e[0]),2),s=d.add(d.mul(d.div(n[0],10),e[0]),r[0]),i=d.div(d.mul(d.exp(d.div(n[3],5)),e[1]),2),c=d.add(d.mul(d.div(n[1],10),e[1]),r[1]);return d.transpose(d.stack([d.sub(s,a),d.sub(c,i),d.add(s,a),d.add(c,i)]),[1,0])}function Xo(o,t,e){return d.tidy(()=>{let r=o.shape[0],n=ua(d.reshape(d.tile(e.extra_dim,[r,1,1]),[-1,4]),d.reshape(o,[-1,4]));n=d.reshape(n,[r,n.shape[0]/r,4]);let a=d.sigmoid(d.slice(t,[0,0,1],[-1,-1,-1])),s=d.slice(a,[0,0,0],[-1,-1,1]);s=d.reshape(s,[r,s.shape[1]]);let i=d.unstack(n),c=d.unstack(s);return{boxes:i,scores:c}})}var Ge=v(x());var Ye=v(x());function te(o,t){return Ye.tidy(()=>{let e=o.shape[0],r=Ye.reshape(qt(o,t.box_encoding_predictor),[e,-1,1,4]),n=Ye.reshape(qt(o,t.class_predictor),[e,-1,3]);return{boxPredictionEncoding:r,classPrediction:n}})}function Jo(o,t,e){return Ge.tidy(()=>{let r=Z(o,e.conv_0,[1,1]),n=Z(r,e.conv_1,[2,2]),a=Z(n,e.conv_2,[1,1]),s=Z(a,e.conv_3,[2,2]),i=Z(s,e.conv_4,[1,1]),c=Z(i,e.conv_5,[2,2]),m=Z(c,e.conv_6,[1,1]),p=Z(m,e.conv_7,[2,2]),u=te(t,e.box_predictor_0),f=te(o,e.box_predictor_1),l=te(n,e.box_predictor_2),b=te(s,e.box_predictor_3),y=te(c,e.box_predictor_4),F=te(p,e.box_predictor_5),h=Ge.concat([u.boxPredictionEncoding,f.boxPredictionEncoding,l.boxPredictionEncoding,b.boxPredictionEncoding,y.boxPredictionEncoding,F.boxPredictionEncoding],1),T=Ge.concat([u.classPrediction,f.classPrediction,l.classPrediction,b.classPrediction,y.classPrediction,F.classPrediction],1);return{boxPredictions:h,classPredictions:T}})}var X=class{constructor({minConfidence:t,maxResults:e}={}){this._name="SsdMobilenetv1Options";if(this._minConfidence=t||.5,this._maxResults=e||100,typeof this._minConfidence!="number"||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if(typeof this._maxResults!="number")throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}};var St=class extends A{constructor(){super("SsdMobilenetv1")}forwardInput(t){let{params:e}=this;if(!e)throw new Error("SsdMobilenetv1 - load model before inference");return Lt.tidy(()=>{let r=Lt.cast(t.toBatchTensor(512,!1),"float32"),n=Lt.sub(Lt.div(r,127.5),1),a=jo(n,e.mobilenetv1),{boxPredictions:s,classPredictions:i}=Jo(a.out,a.conv11,e.prediction_layer);return Xo(s,i,e.output_layer)})}async forward(t){return this.forwardInput(await C(t))}async locateFaces(t,e={}){let{maxResults:r,minConfidence:n}=new X(e),a=await C(t),{boxes:s,scores:i}=this.forwardInput(a),c=s[0],m=i[0];for(let _=1;_<s.length;_++)s[_].dispose(),i[_].dispose();let p=Array.from(m.dataSync()),f=Uo(c,p,r,.5,n),l=a.getReshapedInputDimensions(0),b=a.inputSize,y=b/l.width,F=b/l.height,h=c.arraySync(),T=f.map(_=>{let[E,W]=[Math.max(0,h[_][0]),Math.min(1,h[_][2])].map(q=>q*F),[tt,lt]=[Math.max(0,h[_][1]),Math.min(1,h[_][3])].map(q=>q*y);return new M(p[_],new Yt(tt,E,lt-tt,W-E),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return c.dispose(),m.dispose(),T}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(t){return Go(t)}extractParams(t){return Yo(t)}};function qo(o){let t=new St;return t.extractWeights(o),t}function fa(o){return qo(o)}var lo=class extends St{};var Zo=.4,Ko=[new g(.738768,.874946),new g(2.42204,2.65704),new g(4.30971,7.04493),new g(10.246,4.59428),new g(12.6868,11.8741)],Qo=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],tn=[117.001,114.697,97.404],en="tiny_yolov2_model",rn="tiny_yolov2_separable_conv_model";var N=v(x());var Cr=o=>typeof o=="number";function ho(o){if(!o)throw new Error(`invalid config: ${o}`);if(typeof o.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${o.withSeparableConvs}`);if(!Cr(o.iouThreshold)||o.iouThreshold<0||o.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${o.iouThreshold}`);if(!Array.isArray(o.classes)||!o.classes.length||!o.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(o.classes)}`);if(!Array.isArray(o.anchors)||!o.anchors.length||!o.anchors.map(t=>t||{}).every(t=>Cr(t.x)&&Cr(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(o.anchors)}`);if(o.meanRgb&&(!Array.isArray(o.meanRgb)||o.meanRgb.length!==3||!o.meanRgb.every(Cr)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(o.meanRgb)}`)}var Q=v(x());var K=v(x());function Me(o){return K.tidy(()=>{let t=K.mul(o,K.scalar(.10000000149011612));return K.add(K.relu(K.sub(o,t)),t)})}function Tt(o,t){return Q.tidy(()=>{let e=Q.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=Q.conv2d(e,t.conv.filters,[1,1],"valid"),e=Q.sub(e,t.bn.sub),e=Q.mul(e,t.bn.truediv),e=Q.add(e,t.conv.bias),Me(e)})}var At=v(x());function wt(o,t){return At.tidy(()=>{let e=At.pad(o,[[0,0],[1,1],[1,1],[0,0]]);return e=At.separableConv2d(e,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),e=At.add(e,t.bias),Me(e)})}var bo=v(x());function la(o,t){let e=be(o,t);function r(s,i){let c=bo.tensor1d(o(s)),m=bo.tensor1d(o(s));return t.push({paramPath:`${i}/sub`},{paramPath:`${i}/truediv`}),{sub:c,truediv:m}}function n(s,i,c){let m=e(s,i,3,`${c}/conv`),p=r(i,`${c}/bn`);return{conv:m,bn:p}}let a=ge(o,t);return{extractConvParams:e,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}}function on(o,t,e,r){let{extractWeights:n,getRemainingWeights:a}=R(o),s=[],{extractConvParams:i,extractConvWithBatchNormParams:c,extractSeparableConvParams:m}=la(n,s),p;if(t.withSeparableConvs){let[u,f,l,b,y,F,h,T,_]=r,E=t.isFirstLayerConv2d?i(u,f,3,"conv0"):m(u,f,"conv0"),W=m(f,l,"conv1"),tt=m(l,b,"conv2"),lt=m(b,y,"conv3"),q=m(y,F,"conv4"),Dt=m(F,h,"conv5"),Et=T?m(h,T,"conv6"):void 0,Mt=_?m(T,_,"conv7"):void 0,$t=i(_||T||h,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}else{let[u,f,l,b,y,F,h,T,_]=r,E=c(u,f,"conv0"),W=c(f,l,"conv1"),tt=c(l,b,"conv2"),lt=c(b,y,"conv3"),q=c(y,F,"conv4"),Dt=c(F,h,"conv5"),Et=c(h,T,"conv6"),Mt=c(T,_,"conv7"),$t=i(_,5*e,1,"conv8");p={conv0:E,conv1:W,conv2:tt,conv3:lt,conv4:q,conv5:Dt,conv6:Et,conv7:Mt,conv8:$t}}if(a().length!==0)throw new Error(`weights remaing after extract: ${a().length}`);return{params:p,paramMappings:s}}function da(o,t){let e=Y(o,t);function r(i){let c=e(`${i}/sub`,1),m=e(`${i}/truediv`,1);return{sub:c,truediv:m}}function n(i){let c=e(`${i}/filters`,4),m=e(`${i}/bias`,1);return{filters:c,bias:m}}function a(i){let c=n(`${i}/conv`),m=r(`${i}/bn`);return{conv:c,bn:m}}let s=xe(e);return{extractConvParams:n,extractConvWithBatchNormParams:a,extractSeparableConvParams:s}}function nn(o,t){let e=[],{extractConvParams:r,extractConvWithBatchNormParams:n,extractSeparableConvParams:a}=da(o,e),s;if(t.withSeparableConvs){let i=t.filterSizes&&t.filterSizes.length||9;s={conv0:t.isFirstLayerConv2d?r("conv0"):a("conv0"),conv1:a("conv1"),conv2:a("conv2"),conv3:a("conv3"),conv4:a("conv4"),conv5:a("conv5"),conv6:i>7?a("conv6"):void 0,conv7:i>8?a("conv7"):void 0,conv8:r("conv8")}}else s={conv0:n("conv0"),conv1:n("conv1"),conv2:n("conv2"),conv3:n("conv3"),conv4:n("conv4"),conv5:n("conv5"),conv6:n("conv6"),conv7:n("conv7"),conv8:r("conv8")};return B(o,e),{params:s,paramMappings:e}}var st=class{constructor({inputSize:t,scoreThreshold:e}={}){this._name="TinyYolov2Options";if(this._inputSize=t||416,this._scoreThreshold=e||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}};var go=class extends A{constructor(e){super("TinyYolov2");ho(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,r){let n=Tt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=Tt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=Tt(n,r.conv6),n=Tt(n,r.conv7),qt(n,r.conv8,"valid",!1)}runMobilenet(e,r){let n=this.config.isFirstLayerConv2d?Me(qt(e,r.conv0,"valid",!1)):wt(e,r.conv0);return n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv1),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv2),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv3),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv4),n=N.maxPool(n,[2,2],[2,2],"same"),n=wt(n,r.conv5),n=N.maxPool(n,[2,2],[1,1],"same"),n=r.conv6?wt(n,r.conv6):n,n=r.conv7?wt(n,r.conv7):n,qt(n,r.conv8,"valid",!1)}forwardInput(e,r){let{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return N.tidy(()=>{let a=N.cast(e.toBatchTensor(r,!1),"float32");return a=this.config.meanRgb?rt(a,this.config.meanRgb):a,a=a.div(255),this.config.withSeparableConvs?this.runMobilenet(a,n):this.runTinyYolov2(a,n)})}async forward(e,r){return this.forwardInput(await C(e),r)}async detect(e,r={}){let{inputSize:n,scoreThreshold:a}=new st(r),s=await C(e),i=await this.forwardInput(s,n),c=N.tidy(()=>N.unstack(i)[0].expandDims()),m={width:s.getInputWidth(0),height:s.getInputHeight(0)},p=await this.extractBoxes(c,s.getReshapedInputDimensions(0),a);i.dispose(),c.dispose();let u=p.map(h=>h.box),f=p.map(h=>h.score),l=p.map(h=>h.classScore),b=p.map(h=>this.config.classes[h.label]);return Yr(u.map(h=>h.rescale(n)),f,this.config.iouThreshold,!0).map(h=>new bt(f[h],l[h],b[h],u[h],m))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return nn(e,this.config)}extractParams(e){let r=this.config.filterSizes||go.DEFAULT_FILTER_SIZES,n=r?r.length:void 0;if(n!==7&&n!==8&&n!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);return on(e,this.config,this.boxEncodingSize,r)}async extractBoxes(e,r,n){let{width:a,height:s}=r,i=Math.max(a,s),c=i/a,m=i/s,p=e.shape[1],u=this.config.anchors.length,[f,l,b]=N.tidy(()=>{let T=e.reshape([p,p,u,this.boxEncodingSize]),_=T.slice([0,0,0,0],[p,p,u,4]),E=T.slice([0,0,0,4],[p,p,u,1]),W=this.withClassScores?N.softmax(T.slice([0,0,0,5],[p,p,u,this.config.classes.length]),3):N.scalar(0);return[_,E,W]}),y=[],F=await l.array(),h=await f.array();for(let T=0;T<p;T++)for(let _=0;_<p;_++)for(let E=0;E<u;E++){let W=Ne(F[T][_][E][0]);if(!n||W>n){let tt=(_+Ne(h[T][_][E][0]))/p*c,lt=(T+Ne(h[T][_][E][1]))/p*m,q=Math.exp(h[T][_][E][2])*this.config.anchors[E].x/p*c,Dt=Math.exp(h[T][_][E][3])*this.config.anchors[E].y/p*m,Et=tt-q/2,Mt=lt-Dt/2,$t={row:T,col:_,anchor:E},{classScore:yo,label:_o}=this.withClassScores?await this.extractPredictedClass(b,$t):{classScore:1,label:0};y.push({box:new Vt(Et,Mt,Et+q,Mt+Dt),score:W,classScore:W*yo,label:_o,...$t})}}return f.dispose(),l.dispose(),b.dispose(),y}async extractPredictedClass(e,r){let{row:n,col:a,anchor:s}=r,i=await e.array();return Array(this.config.classes.length).fill(0).map((c,m)=>i[n][a][s][m]).map((c,m)=>({classScore:c,label:m})).reduce((c,m)=>c.classScore>m.classScore?c:m)}},ee=go;ee.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var re=class extends ee{constructor(t=!0){let e={withSeparableConvs:t,iouThreshold:Zo,classes:["face"],...t?{anchors:Qo,meanRgb:tn}:{anchors:Ko,withClassScores:!0}};super(e)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?rn:en}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};function ha(o,t=!0){let e=new re(t);return e.extractWeights(o),e}var je=class extends st{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}};var J=class{async then(t){return t(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};var Xe=v(x());var xo=v(x());async function oe(o,t,e,r,n=({alignedRect:a})=>a){let a=o.map(c=>Zt(c)?n(c):c.detection),s=r||(t instanceof xo.Tensor?await de(t,a):await le(t,a)),i=await e(s);return s.forEach(c=>c instanceof xo.Tensor&&c.dispose()),i}async function Ce(o,t,e,r,n){return oe([o],t,async a=>e(a[0]),r,n)}var an=.4,sn=[new g(1.603231,2.094468),new g(6.041143,7.080126),new g(2.882459,3.518061),new g(4.266906,5.178857),new g(9.041765,10.66308)],cn=[117.001,114.697,97.404];var ne=class extends ee{constructor(){let t={withSeparableConvs:!0,iouThreshold:an,classes:["face"],anchors:sn,meanRgb:cn,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(t)}get anchors(){return this.config.anchors}async locateFaces(t,e){return(await this.detect(t,e)).map(n=>new M(n.score,n.relativeBox,{width:n.imageWidth,height:n.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(t){return super.extractParamsFromWeightMap(t)}};var P={ssdMobilenetv1:new St,tinyFaceDetector:new ne,tinyYolov2:new re,faceLandmark68Net:new Kt,faceLandmark68TinyNet:new ze,faceRecognitionNet:new Qt,faceExpressionNet:new Oe,ageGenderNet:new He},mn=(o,t)=>P.ssdMobilenetv1.locateFaces(o,t),ba=(o,t)=>P.tinyFaceDetector.locateFaces(o,t),ga=(o,t)=>P.tinyYolov2.locateFaces(o,t),pn=o=>P.faceLandmark68Net.detectLandmarks(o),xa=o=>P.faceLandmark68TinyNet.detectLandmarks(o),va=o=>P.faceRecognitionNet.computeFaceDescriptor(o),ya=o=>P.faceExpressionNet.predictExpressions(o),_a=o=>P.ageGenderNet.predictAgeAndGender(o),un=o=>P.ssdMobilenetv1.load(o),Ta=o=>P.tinyFaceDetector.load(o),wa=o=>P.tinyYolov2.load(o),Pa=o=>P.faceLandmark68Net.load(o),Fa=o=>P.faceLandmark68TinyNet.load(o),Da=o=>P.faceRecognitionNet.load(o),Ea=o=>P.faceExpressionNet.load(o),Ma=o=>P.ageGenderNet.load(o),Ca=un,Ia=mn,Na=pn;var Ir=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ae=class extends Ir{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.faceExpressionNet.predictExpressions(n))),this.extractedFaces);return t.map((r,n)=>xr(r,e[n]))}withAgeAndGender(){return new ie(this,this.input)}},se=class extends Ir{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceExpressionNet.predictExpressions(r),this.extractedFaces);return xr(t,e)}withAgeAndGender(){return new ce(this,this.input)}},Wt=class extends ae{withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},kt=class extends se{withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Nr=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.extractedFaces=n}},ie=class extends Nr{async run(){let t=await this.parentTask,e=await oe(t,this.input,async r=>Promise.all(r.map(n=>P.ageGenderNet.predictAgeAndGender(n))),this.extractedFaces);return t.map((r,n)=>{let{age:a,gender:s,genderProbability:i}=e[n];return Er(Mr(r,s,i),a)})}withFaceExpressions(){return new ae(this,this.input)}},ce=class extends Nr{async run(){let t=await this.parentTask;if(!t)return;let{age:e,gender:r,genderProbability:n}=await Ce(t,this.input,a=>P.ageGenderNet.predictAgeAndGender(a),this.extractedFaces);return Er(Mr(t,r,n),e)}withFaceExpressions(){return new se(this,this.input)}},Bt=class extends ie{withFaceExpressions(){return new Wt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Rt=class extends ce{withFaceExpressions(){return new kt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ue=class extends J{constructor(e,r){super();this.parentTask=e;this.input=r}},Pt=class extends Ue{async run(){let t=await this.parentTask;return(await oe(t,this.input,r=>Promise.all(r.map(n=>P.faceRecognitionNet.computeFaceDescriptor(n))),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}))).map((r,n)=>Dr(t[n],r))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}},Ft=class extends Ue{async run(){let t=await this.parentTask;if(!t)return;let e=await Ce(t,this.input,r=>P.faceRecognitionNet.computeFaceDescriptor(r),null,r=>r.landmarks.align(null,{useDlibAlignment:!0}));return Dr(t,e)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}};var Je=class extends J{constructor(e,r,n){super();this.parentTask=e;this.input=r;this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?P.faceLandmark68TinyNet:P.faceLandmark68Net}},qe=class extends Je{async run(){let t=await this.parentTask,e=t.map(s=>s.detection),r=this.input instanceof Xe.Tensor?await de(this.input,e):await le(this.input,e),n=await Promise.all(r.map(s=>this.landmarkNet.detectLandmarks(s)));return r.forEach(s=>s instanceof Xe.Tensor&&s.dispose()),t.filter((s,i)=>n[i]).map((s,i)=>Pe(s,n[i]))}withFaceExpressions(){return new Wt(this,this.input)}withAgeAndGender(){return new Bt(this,this.input)}withFaceDescriptors(){return new Pt(this,this.input)}},Ze=class extends Je{async run(){let t=await this.parentTask;if(!t)return;let{detection:e}=t,r=this.input instanceof Xe.Tensor?await de(this.input,[e]):await le(this.input,[e]),n=await this.landmarkNet.detectLandmarks(r[0]);return r.forEach(a=>a instanceof Xe.Tensor&&a.dispose()),Pe(t,n)}withFaceExpressions(){return new kt(this,this.input)}withAgeAndGender(){return new Rt(this,this.input)}withFaceDescriptor(){return new Ft(this,this.input)}};var Ke=class extends J{constructor(e,r=new X){super();this.input=e;this.options=r}},Ie=class extends Ke{async run(){let{input:t,options:e}=this,r;if(e instanceof je)r=P.tinyFaceDetector.locateFaces(t,e);else if(e instanceof X)r=P.ssdMobilenetv1.locateFaces(t,e);else if(e instanceof st)r=P.tinyYolov2.locateFaces(t,e);else throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");return r}runAndExtendWithFaceDetections(){return new Promise((t,e)=>{this.run().then(r=>t(r.map(n=>jt({},n)))).catch(r=>e(r))})}withFaceLandmarks(t=!1){return new qe(this.runAndExtendWithFaceDetections(),this.input,t)}withFaceExpressions(){return new ae(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new ie(this.runAndExtendWithFaceDetections(),this.input)}},Qe=class extends Ke{async run(){let t=await new Ie(this.input,this.options),e=t[0];return t.forEach(r=>{r.score>e.score&&(e=r)}),e}runAndExtendWithFaceDetection(){return new Promise(async t=>{let e=await this.run();t(e?jt({},e):void 0)})}withFaceLandmarks(t=!1){return new Ze(this.runAndExtendWithFaceDetection(),this.input,t)}withFaceExpressions(){return new se(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new ce(this.runAndExtendWithFaceDetection(),this.input)}};function Sa(o,t=new X){return new Qe(o,t)}function Sr(o,t=new X){return new Ie(o,t)}async function fn(o,t){return Sr(o,new X(t?{minConfidence:t}:{})).withFaceLandmarks().withFaceDescriptors()}async function La(o,t={}){return Sr(o,new st(t)).withFaceLandmarks().withFaceDescriptors()}var Aa=fn;function vo(o,t){if(o.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");let e=Array.from(o),r=Array.from(t);return Math.sqrt(e.map((n,a)=>n-r[a]).reduce((n,a)=>n+a*a,0))}var tr=class{constructor(t,e=.6){this._distanceThreshold=e;let r=Array.isArray(t)?t:[t];if(!r.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let n=1,a=()=>`person ${n++}`;this._labeledDescriptors=r.map(s=>{if(s instanceof mt)return s;if(s instanceof Float32Array)return new mt(a(),[s]);if(s.descriptor&&s.descriptor instanceof Float32Array)return new mt(a(),[s.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(t,e){return e.map(r=>vo(r,t)).reduce((r,n)=>r+n,0)/(e.length||1)}matchDescriptor(t){return this.labeledDescriptors.map(({descriptors:e,label:r})=>new pe(r,this.computeMeanDistance(t,e))).reduce((e,r)=>e.distance<r.distance?e:r)}findBestMatch(t){let e=this.matchDescriptor(t);return e.distance<this._distanceThreshold?e:new pe("unknown",e.distance)}toJSON(){return{distanceThreshold:this._distanceThreshold,labeledDescriptors:this._labeledDescriptors.map(t=>t.toJSON())}}static fromJSON(t){let e=t.labeledDescriptors.map(r=>mt.fromJSON(r));return new tr(e,t.distanceThreshold)}};function Wa(o){let t=new ne;return t.extractWeights(o),t}function ln(o,t){let{width:e,height:r}=new k(t.width,t.height);if(e<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:e,height:r})}`);if(Array.isArray(o))return o.map(n=>ln(n,{width:e,height:r}));if(Zt(o)){let n=o.detection.forSize(e,r),a=o.unshiftedLandmarks.forSize(n.box.width,n.box.height);return Pe(jt(o,n),a)}return pt(o)?jt(o,o.detection.forSize(e,r)):o instanceof z||o instanceof M?o.forSize(e,r):o}var Ba=So;0&&(module.exports={AgeGenderNet,BoundingBox,Box,ComposableTask,ComputeAllFaceDescriptorsTask,ComputeFaceDescriptorsTaskBase,ComputeSingleFaceDescriptorTask,DetectAllFaceLandmarksTask,DetectAllFacesTask,DetectFaceLandmarksTaskBase,DetectFacesTaskBase,DetectSingleFaceLandmarksTask,DetectSingleFaceTask,Dimensions,FACE_EXPRESSION_LABELS,FaceDetection,FaceDetectionNet,FaceExpressionNet,FaceExpressions,FaceLandmark68Net,FaceLandmark68TinyNet,FaceLandmarkNet,FaceLandmarks,FaceLandmarks5,FaceLandmarks68,FaceMatch,FaceMatcher,FaceRecognitionNet,Gender,LabeledBox,LabeledFaceDescriptors,NetInput,NeuralNetwork,ObjectDetection,Point,PredictedBox,Rect,SsdMobilenetv1,SsdMobilenetv1Options,TinyFaceDetector,TinyFaceDetectorOptions,TinyYolov2,TinyYolov2Options,allFaces,allFacesSsdMobilenetv1,allFacesTinyYolov2,awaitMediaLoaded,bufferToImage,computeFaceDescriptor,createCanvas,createCanvasFromMedia,createFaceDetectionNet,createFaceRecognitionNet,createSsdMobilenetv1,createTinyFaceDetector,createTinyYolov2,detectAllFaces,detectFaceLandmarks,detectFaceLandmarksTiny,detectLandmarks,detectSingleFace,draw,env,euclideanDistance,extendWithAge,extendWithFaceDescriptor,extendWithFaceDetection,extendWithFaceExpressions,extendWithFaceLandmarks,extendWithGender,extractFaceTensors,extractFaces,fetchImage,fetchJson,fetchNetWeights,fetchOrThrow,fetchVideo,getContext2dOrThrow,getMediaDimensions,imageTensorToCanvas,imageToSquare,inverseSigmoid,iou,isMediaElement,isMediaLoaded,isWithAge,isWithFaceDetection,isWithFaceExpressions,isWithFaceLandmarks,isWithGender,loadAgeGenderModel,loadFaceDetectionModel,loadFaceExpressionModel,loadFaceLandmarkModel,loadFaceLandmarkTinyModel,loadFaceRecognitionModel,loadSsdMobilenetv1Model,loadTinyFaceDetectorModel,loadTinyYolov2Model,loadWeightMap,locateFaces,matchDimensions,minBbox,nets,nonMaxSuppression,normalize,padToSquare,predictAgeAndGender,recognizeFaceExpressions,resizeResults,resolveInput,shuffleArray,sigmoid,ssdMobilenetv1,tf,tinyFaceDetector,tinyYolov2,toNetInput,utils,validateConfig,version});
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __commonJS = (cb, mod) => function __require() {
+  return mod || (0, cb[__getOwnPropNames(cb)[0]])((mod = { exports: {} }).exports, mod), mod.exports;
+};
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// dist/tfjs.esm.js
+var require_tfjs_esm = __commonJS({
+  "dist/tfjs.esm.js"(exports, module2) {
+    "use strict";
+    var __defProp2 = Object.defineProperty;
+    var __getOwnPropDesc2 = Object.getOwnPropertyDescriptor;
+    var __getOwnPropNames2 = Object.getOwnPropertyNames;
+    var __hasOwnProp2 = Object.prototype.hasOwnProperty;
+    var __export2 = (target, all) => {
+      for (var name in all)
+        __defProp2(target, name, { get: all[name], enumerable: true });
+    };
+    var __copyProps2 = (to, from, except, desc) => {
+      if (from && typeof from === "object" || typeof from === "function") {
+        for (let key of __getOwnPropNames2(from))
+          if (!__hasOwnProp2.call(to, key) && key !== except)
+            __defProp2(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc2(from, key)) || desc.enumerable });
+      }
+      return to;
+    };
+    var __reExport = (target, mod, secondTarget) => (__copyProps2(target, mod, "default"), secondTarget && __copyProps2(secondTarget, mod, "default"));
+    var __toCommonJS2 = (mod) => __copyProps2(__defProp2({}, "__esModule", { value: true }), mod);
+    var tf_node_exports = {};
+    __export2(tf_node_exports, {
+      version: () => version6
+    });
+    module2.exports = __toCommonJS2(tf_node_exports);
+    __reExport(tf_node_exports, require("@tensorflow/tfjs-node"), module2.exports);
+    var version3 = "4.0.0";
+    var version22 = "4.0.0";
+    var version32 = "4.0.0";
+    var version4 = "4.0.0";
+    var version5 = "4.0.0";
+    var version6 = {
+      tfjs: version3,
+      "tfjs-core": version3,
+      "tfjs-converter": version22,
+      "tfjs-backend-cpu": version32,
+      "tfjs-backend-webgl": version4,
+      "tfjs-backend-wasm": version5
+    };
+  }
+});
+
+// src/index.ts
+var src_exports = {};
+__export(src_exports, {
+  AgeGenderNet: () => AgeGenderNet,
+  BoundingBox: () => BoundingBox,
+  Box: () => Box,
+  ComposableTask: () => ComposableTask,
+  ComputeAllFaceDescriptorsTask: () => ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase: () => ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask: () => ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask: () => DetectAllFaceLandmarksTask,
+  DetectAllFacesTask: () => DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase: () => DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase: () => DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask: () => DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask: () => DetectSingleFaceTask,
+  Dimensions: () => Dimensions,
+  FACE_EXPRESSION_LABELS: () => FACE_EXPRESSION_LABELS,
+  FaceDetection: () => FaceDetection,
+  FaceDetectionNet: () => FaceDetectionNet,
+  FaceExpressionNet: () => FaceExpressionNet,
+  FaceExpressions: () => FaceExpressions,
+  FaceLandmark68Net: () => FaceLandmark68Net,
+  FaceLandmark68TinyNet: () => FaceLandmark68TinyNet,
+  FaceLandmarkNet: () => FaceLandmarkNet,
+  FaceLandmarks: () => FaceLandmarks,
+  FaceLandmarks5: () => FaceLandmarks5,
+  FaceLandmarks68: () => FaceLandmarks68,
+  FaceMatch: () => FaceMatch,
+  FaceMatcher: () => FaceMatcher,
+  FaceRecognitionNet: () => FaceRecognitionNet,
+  Gender: () => Gender,
+  LabeledBox: () => LabeledBox,
+  LabeledFaceDescriptors: () => LabeledFaceDescriptors,
+  NetInput: () => NetInput,
+  NeuralNetwork: () => NeuralNetwork,
+  ObjectDetection: () => ObjectDetection,
+  Point: () => Point,
+  PredictedBox: () => PredictedBox,
+  Rect: () => Rect,
+  SsdMobilenetv1: () => SsdMobilenetv1,
+  SsdMobilenetv1Options: () => SsdMobilenetv1Options,
+  TinyFaceDetector: () => TinyFaceDetector,
+  TinyFaceDetectorOptions: () => TinyFaceDetectorOptions,
+  TinyYolov2: () => TinyYolov2,
+  TinyYolov2Options: () => TinyYolov2Options,
+  allFaces: () => allFaces,
+  allFacesSsdMobilenetv1: () => allFacesSsdMobilenetv1,
+  allFacesTinyYolov2: () => allFacesTinyYolov2,
+  awaitMediaLoaded: () => awaitMediaLoaded,
+  bufferToImage: () => bufferToImage,
+  computeFaceDescriptor: () => computeFaceDescriptor,
+  createCanvas: () => createCanvas,
+  createCanvasFromMedia: () => createCanvasFromMedia,
+  createFaceDetectionNet: () => createFaceDetectionNet,
+  createFaceRecognitionNet: () => createFaceRecognitionNet,
+  createSsdMobilenetv1: () => createSsdMobilenetv1,
+  createTinyFaceDetector: () => createTinyFaceDetector,
+  createTinyYolov2: () => createTinyYolov2,
+  detectAllFaces: () => detectAllFaces,
+  detectFaceLandmarks: () => detectFaceLandmarks,
+  detectFaceLandmarksTiny: () => detectFaceLandmarksTiny,
+  detectLandmarks: () => detectLandmarks,
+  detectSingleFace: () => detectSingleFace,
+  draw: () => draw_exports,
+  env: () => env,
+  euclideanDistance: () => euclideanDistance,
+  extendWithAge: () => extendWithAge,
+  extendWithFaceDescriptor: () => extendWithFaceDescriptor,
+  extendWithFaceDetection: () => extendWithFaceDetection,
+  extendWithFaceExpressions: () => extendWithFaceExpressions,
+  extendWithFaceLandmarks: () => extendWithFaceLandmarks,
+  extendWithGender: () => extendWithGender,
+  extractFaceTensors: () => extractFaceTensors,
+  extractFaces: () => extractFaces,
+  fetchImage: () => fetchImage,
+  fetchJson: () => fetchJson,
+  fetchNetWeights: () => fetchNetWeights,
+  fetchOrThrow: () => fetchOrThrow,
+  fetchVideo: () => fetchVideo,
+  getContext2dOrThrow: () => getContext2dOrThrow,
+  getMediaDimensions: () => getMediaDimensions,
+  imageTensorToCanvas: () => imageTensorToCanvas,
+  imageToSquare: () => imageToSquare,
+  inverseSigmoid: () => inverseSigmoid,
+  iou: () => iou,
+  isMediaElement: () => isMediaElement,
+  isMediaLoaded: () => isMediaLoaded,
+  isWithAge: () => isWithAge,
+  isWithFaceDetection: () => isWithFaceDetection,
+  isWithFaceExpressions: () => isWithFaceExpressions,
+  isWithFaceLandmarks: () => isWithFaceLandmarks,
+  isWithGender: () => isWithGender,
+  loadAgeGenderModel: () => loadAgeGenderModel,
+  loadFaceDetectionModel: () => loadFaceDetectionModel,
+  loadFaceExpressionModel: () => loadFaceExpressionModel,
+  loadFaceLandmarkModel: () => loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel: () => loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel: () => loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model: () => loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel: () => loadTinyFaceDetectorModel,
+  loadTinyYolov2Model: () => loadTinyYolov2Model,
+  loadWeightMap: () => loadWeightMap,
+  locateFaces: () => locateFaces,
+  matchDimensions: () => matchDimensions,
+  minBbox: () => minBbox,
+  nets: () => nets,
+  nonMaxSuppression: () => nonMaxSuppression,
+  normalize: () => normalize,
+  padToSquare: () => padToSquare,
+  predictAgeAndGender: () => predictAgeAndGender,
+  recognizeFaceExpressions: () => recognizeFaceExpressions,
+  resizeResults: () => resizeResults,
+  resolveInput: () => resolveInput,
+  shuffleArray: () => shuffleArray,
+  sigmoid: () => sigmoid,
+  ssdMobilenetv1: () => ssdMobilenetv1,
+  tf: () => tf42,
+  tinyFaceDetector: () => tinyFaceDetector,
+  tinyYolov2: () => tinyYolov2,
+  toNetInput: () => toNetInput,
+  utils: () => utils_exports,
+  validateConfig: () => validateConfig,
+  version: () => version2
+});
+module.exports = __toCommonJS(src_exports);
+var tf42 = __toESM(require_tfjs_esm());
+
+// src/draw/index.ts
+var draw_exports = {};
+__export(draw_exports, {
+  AnchorPosition: () => AnchorPosition,
+  DrawBox: () => DrawBox,
+  DrawBoxOptions: () => DrawBoxOptions,
+  DrawFaceLandmarks: () => DrawFaceLandmarks,
+  DrawFaceLandmarksOptions: () => DrawFaceLandmarksOptions,
+  DrawTextField: () => DrawTextField,
+  DrawTextFieldOptions: () => DrawTextFieldOptions,
+  drawContour: () => drawContour,
+  drawDetections: () => drawDetections,
+  drawFaceExpressions: () => drawFaceExpressions,
+  drawFaceLandmarks: () => drawFaceLandmarks
+});
+
+// src/draw/drawContour.ts
+function drawContour(ctx, points, isClosed = false) {
+  ctx.beginPath();
+  points.slice(1).forEach(({ x, y }, prevIdx) => {
+    const from = points[prevIdx];
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(x, y);
+  });
+  if (isClosed) {
+    const from = points[points.length - 1];
+    const to = points[0];
+    if (!from || !to) {
+      return;
+    }
+    ctx.moveTo(from.x, from.y);
+    ctx.lineTo(to.x, to.y);
+  }
+  ctx.stroke();
+}
+
+// src/utils/index.ts
+var utils_exports = {};
+__export(utils_exports, {
+  computeReshapedDimensions: () => computeReshapedDimensions,
+  getCenterPoint: () => getCenterPoint,
+  isDimensions: () => isDimensions,
+  isEven: () => isEven,
+  isFloat: () => isFloat,
+  isTensor: () => isTensor,
+  isTensor1D: () => isTensor1D,
+  isTensor2D: () => isTensor2D,
+  isTensor3D: () => isTensor3D,
+  isTensor4D: () => isTensor4D,
+  isValidNumber: () => isValidNumber,
+  isValidProbablitiy: () => isValidProbablitiy,
+  range: () => range,
+  round: () => round
+});
+var tf = __toESM(require_tfjs_esm());
+
+// src/classes/Dimensions.ts
+var Dimensions = class {
+  constructor(width, height) {
+    if (!isValidNumber(width) || !isValidNumber(height)) {
+      throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({ width, height })}`);
+    }
+    this._width = width;
+    this._height = height;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  reverse() {
+    return new Dimensions(1 / this.width, 1 / this.height);
+  }
+};
+
+// src/utils/index.ts
+function isTensor(tensor2, dim) {
+  return tensor2 instanceof tf.Tensor && tensor2.shape.length === dim;
+}
+function isTensor1D(tensor2) {
+  return isTensor(tensor2, 1);
+}
+function isTensor2D(tensor2) {
+  return isTensor(tensor2, 2);
+}
+function isTensor3D(tensor2) {
+  return isTensor(tensor2, 3);
+}
+function isTensor4D(tensor2) {
+  return isTensor(tensor2, 4);
+}
+function isFloat(num) {
+  return num % 1 !== 0;
+}
+function isEven(num) {
+  return num % 2 === 0;
+}
+function round(num, prec = 2) {
+  const f = 10 ** prec;
+  return Math.floor(num * f) / f;
+}
+function isDimensions(obj) {
+  return obj && obj.width && obj.height;
+}
+function computeReshapedDimensions({ width, height }, inputSize) {
+  const scale2 = inputSize / Math.max(height, width);
+  return new Dimensions(Math.round(width * scale2), Math.round(height * scale2));
+}
+function getCenterPoint(pts) {
+  return pts.reduce((sum, pt) => sum.add(pt), new Point(0, 0)).div(new Point(pts.length, pts.length));
+}
+function range(num, start, step) {
+  return Array(num).fill(0).map((_, i) => start + i * step);
+}
+function isValidNumber(num) {
+  return !!num && num !== Infinity && num !== -Infinity && !Number.isNaN(num) || num === 0;
+}
+function isValidProbablitiy(num) {
+  return isValidNumber(num) && num >= 0 && num <= 1;
+}
+
+// src/classes/Point.ts
+var Point = class {
+  constructor(x, y) {
+    this._x = x;
+    this._y = y;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  add(pt) {
+    return new Point(this.x + pt.x, this.y + pt.y);
+  }
+  sub(pt) {
+    return new Point(this.x - pt.x, this.y - pt.y);
+  }
+  mul(pt) {
+    return new Point(this.x * pt.x, this.y * pt.y);
+  }
+  div(pt) {
+    return new Point(this.x / pt.x, this.y / pt.y);
+  }
+  abs() {
+    return new Point(Math.abs(this.x), Math.abs(this.y));
+  }
+  magnitude() {
+    return Math.sqrt(this.x ** 2 + this.y ** 2);
+  }
+  floor() {
+    return new Point(Math.floor(this.x), Math.floor(this.y));
+  }
+};
+
+// src/classes/Box.ts
+var Box = class {
+  static isRect(rect) {
+    return !!rect && [rect.x, rect.y, rect.width, rect.height].every(isValidNumber);
+  }
+  static assertIsValidBox(box, callee, allowNegativeDimensions = false) {
+    if (!Box.isRect(box)) {
+      throw new Error(`${callee} - invalid box: ${JSON.stringify(box)}, expected object with properties x, y, width, height`);
+    }
+    if (!allowNegativeDimensions && (box.width < 0 || box.height < 0)) {
+      throw new Error(`${callee} - width (${box.width}) and height (${box.height}) must be positive numbers`);
+    }
+  }
+  constructor(_box, allowNegativeDimensions = true) {
+    const box = _box || {};
+    const isBbox = [box.left, box.top, box.right, box.bottom].every(isValidNumber);
+    const isRect = [box.x, box.y, box.width, box.height].every(isValidNumber);
+    if (!isRect && !isBbox) {
+      throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(box)}`);
+    }
+    const [x, y, width, height] = isRect ? [box.x, box.y, box.width, box.height] : [box.left, box.top, box.right - box.left, box.bottom - box.top];
+    Box.assertIsValidBox({
+      x,
+      y,
+      width,
+      height
+    }, "Box.constructor", allowNegativeDimensions);
+    this._x = x;
+    this._y = y;
+    this._width = width;
+    this._height = height;
+  }
+  get x() {
+    return this._x;
+  }
+  get y() {
+    return this._y;
+  }
+  get width() {
+    return this._width;
+  }
+  get height() {
+    return this._height;
+  }
+  get left() {
+    return this.x;
+  }
+  get top() {
+    return this.y;
+  }
+  get right() {
+    return this.x + this.width;
+  }
+  get bottom() {
+    return this.y + this.height;
+  }
+  get area() {
+    return this.width * this.height;
+  }
+  get topLeft() {
+    return new Point(this.left, this.top);
+  }
+  get topRight() {
+    return new Point(this.right, this.top);
+  }
+  get bottomLeft() {
+    return new Point(this.left, this.bottom);
+  }
+  get bottomRight() {
+    return new Point(this.right, this.bottom);
+  }
+  round() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.round(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  floor() {
+    const [x, y, width, height] = [this.x, this.y, this.width, this.height].map((val) => Math.floor(val));
+    return new Box({
+      x,
+      y,
+      width,
+      height
+    });
+  }
+  toSquare() {
+    let {
+      x,
+      y,
+      width,
+      height
+    } = this;
+    const diff = Math.abs(width - height);
+    if (width < height) {
+      x -= diff / 2;
+      width += diff;
+    }
+    if (height < width) {
+      y -= diff / 2;
+      height += diff;
+    }
+    return new Box({ x, y, width, height });
+  }
+  rescale(s) {
+    const scaleX = isDimensions(s) ? s.width : s;
+    const scaleY = isDimensions(s) ? s.height : s;
+    return new Box({
+      x: this.x * scaleX,
+      y: this.y * scaleY,
+      width: this.width * scaleX,
+      height: this.height * scaleY
+    });
+  }
+  pad(padX, padY) {
+    const [x, y, width, height] = [
+      this.x - padX / 2,
+      this.y - padY / 2,
+      this.width + padX,
+      this.height + padY
+    ];
+    return new Box({ x, y, width, height });
+  }
+  clipAtImageBorders(imgWidth, imgHeight) {
+    const { x, y, right, bottom } = this;
+    const clippedX = Math.max(x, 0);
+    const clippedY = Math.max(y, 0);
+    const newWidth = right - clippedX;
+    const newHeight = bottom - clippedY;
+    const clippedWidth = Math.min(newWidth, imgWidth - clippedX);
+    const clippedHeight = Math.min(newHeight, imgHeight - clippedY);
+    return new Box({ x: clippedX, y: clippedY, width: clippedWidth, height: clippedHeight }).floor();
+  }
+  shift(sx, sy) {
+    const { width, height } = this;
+    const x = this.x + sx;
+    const y = this.y + sy;
+    return new Box({ x, y, width, height });
+  }
+  padAtBorders(imageHeight, imageWidth) {
+    const w = this.width + 1;
+    const h = this.height + 1;
+    const dx = 1;
+    const dy = 1;
+    let edx = w;
+    let edy = h;
+    let x = this.left;
+    let y = this.top;
+    let ex = this.right;
+    let ey = this.bottom;
+    if (ex > imageWidth) {
+      edx = -ex + imageWidth + w;
+      ex = imageWidth;
+    }
+    if (ey > imageHeight) {
+      edy = -ey + imageHeight + h;
+      ey = imageHeight;
+    }
+    if (x < 1) {
+      edy = 2 - x;
+      x = 1;
+    }
+    if (y < 1) {
+      edy = 2 - y;
+      y = 1;
+    }
+    return { dy, edy, dx, edx, y, ey, x, ex, w, h };
+  }
+  calibrate(region) {
+    return new Box({
+      left: this.left + region.left * this.width,
+      top: this.top + region.top * this.height,
+      right: this.right + region.right * this.width,
+      bottom: this.bottom + region.bottom * this.height
+    }).toSquare().round();
+  }
+};
+
+// src/classes/BoundingBox.ts
+var BoundingBox = class extends Box {
+  constructor(left, top, right, bottom, allowNegativeDimensions = false) {
+    super({ left, top, right, bottom }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/ObjectDetection.ts
+var ObjectDetection = class {
+  constructor(score, classScore, className, relativeBox, imageDims) {
+    this._imageDims = new Dimensions(imageDims.width, imageDims.height);
+    this._score = score;
+    this._classScore = classScore;
+    this._className = className;
+    this._box = new Box(relativeBox).rescale(this._imageDims);
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+  get className() {
+    return this._className;
+  }
+  get box() {
+    return this._box;
+  }
+  get imageDims() {
+    return this._imageDims;
+  }
+  get imageWidth() {
+    return this.imageDims.width;
+  }
+  get imageHeight() {
+    return this.imageDims.height;
+  }
+  get relativeBox() {
+    return new Box(this._box).rescale(this.imageDims.reverse());
+  }
+  forSize(width, height) {
+    return new ObjectDetection(
+      this.score,
+      this.classScore,
+      this.className,
+      this.relativeBox,
+      { width, height }
+    );
+  }
+};
+
+// src/classes/FaceDetection.ts
+var FaceDetection = class extends ObjectDetection {
+  constructor(score, relativeBox, imageDims) {
+    super(score, score, "", relativeBox, imageDims);
+  }
+  forSize(width, height) {
+    const { score, relativeBox, imageDims } = super.forSize(width, height);
+    return new FaceDetection(score, relativeBox, imageDims);
+  }
+};
+
+// src/ops/iou.ts
+function iou(box1, box2, isIOU = true) {
+  const width = Math.max(0, Math.min(box1.right, box2.right) - Math.max(box1.left, box2.left));
+  const height = Math.max(0, Math.min(box1.bottom, box2.bottom) - Math.max(box1.top, box2.top));
+  const interSection = width * height;
+  return isIOU ? interSection / (box1.area + box2.area - interSection) : interSection / Math.min(box1.area, box2.area);
+}
+
+// src/ops/minBbox.ts
+function minBbox(pts) {
+  const xs = pts.map((pt) => pt.x);
+  const ys = pts.map((pt) => pt.y);
+  const minX = xs.reduce((min, x) => x < min ? x : min, Infinity);
+  const minY = ys.reduce((min, y) => y < min ? y : min, Infinity);
+  const maxX = xs.reduce((max, x) => max < x ? x : max, 0);
+  const maxY = ys.reduce((max, y) => max < y ? y : max, 0);
+  return new BoundingBox(minX, minY, maxX, maxY);
+}
+
+// src/ops/nonMaxSuppression.ts
+function nonMaxSuppression(boxes, scores, iouThreshold, isIOU = true) {
+  let indicesSortedByScore = scores.map((score, boxIndex) => ({ score, boxIndex })).sort((c1, c2) => c1.score - c2.score).map((c) => c.boxIndex);
+  const pick = [];
+  while (indicesSortedByScore.length > 0) {
+    const curr = indicesSortedByScore.pop();
+    pick.push(curr);
+    const indices = indicesSortedByScore;
+    const outputs = [];
+    for (let i = 0; i < indices.length; i++) {
+      const idx = indices[i];
+      const currBox = boxes[curr];
+      const idxBox = boxes[idx];
+      outputs.push(iou(currBox, idxBox, isIOU));
+    }
+    indicesSortedByScore = indicesSortedByScore.filter(
+      (_, j) => outputs[j] <= iouThreshold
+    );
+  }
+  return pick;
+}
+
+// src/ops/normalize.ts
+var tf2 = __toESM(require_tfjs_esm());
+function normalize(x, meanRgb) {
+  return tf2.tidy(() => {
+    const [r, g, b] = meanRgb;
+    const avg_r = tf2.fill([...x.shape.slice(0, 3), 1], r, "float32");
+    const avg_g = tf2.fill([...x.shape.slice(0, 3), 1], g, "float32");
+    const avg_b = tf2.fill([...x.shape.slice(0, 3), 1], b, "float32");
+    const avg_rgb = tf2.concat([avg_r, avg_g, avg_b], 3);
+    return tf2.sub(x, avg_rgb);
+  });
+}
+
+// src/ops/padToSquare.ts
+var tf3 = __toESM(require_tfjs_esm());
+function padToSquare(imgTensor, isCenterImage = false) {
+  return tf3.tidy(() => {
+    const [height, width] = imgTensor.shape.slice(1);
+    if (height === width)
+      return imgTensor;
+    const dimDiff = Math.abs(height - width);
+    const paddingAmount = Math.round(dimDiff * (isCenterImage ? 0.5 : 1));
+    const paddingAxis = height > width ? 2 : 1;
+    const createPaddingTensor = (paddingAmountLocal) => {
+      const paddingTensorShape = imgTensor.shape.slice();
+      paddingTensorShape[paddingAxis] = paddingAmountLocal;
+      return tf3.fill(paddingTensorShape, 0, "float32");
+    };
+    const paddingTensorAppend = createPaddingTensor(paddingAmount);
+    const remainingPaddingAmount = dimDiff - paddingTensorAppend.shape[paddingAxis];
+    const paddingTensorPrepend = isCenterImage && remainingPaddingAmount ? createPaddingTensor(remainingPaddingAmount) : null;
+    const tensorsToStack = [paddingTensorPrepend, imgTensor, paddingTensorAppend].filter((t) => !!t).map((t) => tf3.cast(t, "float32"));
+    return tf3.concat(tensorsToStack, paddingAxis);
+  });
+}
+
+// src/ops/shuffleArray.ts
+function shuffleArray(inputArray) {
+  const array = inputArray.slice();
+  for (let i = array.length - 1; i > 0; i--) {
+    const j = Math.floor(Math.random() * (i + 1));
+    const x = array[i];
+    array[i] = array[j];
+    array[j] = x;
+  }
+  return array;
+}
+
+// src/ops/index.ts
+function sigmoid(x) {
+  return 1 / (1 + Math.exp(-x));
+}
+function inverseSigmoid(x) {
+  return Math.log(x / (1 - x));
+}
+
+// src/classes/Rect.ts
+var Rect = class extends Box {
+  constructor(x, y, width, height, allowNegativeDimensions = false) {
+    super({ x, y, width, height }, allowNegativeDimensions);
+  }
+};
+
+// src/classes/FaceLandmarks.ts
+var relX = 0.5;
+var relY = 0.43;
+var relScale = 0.45;
+var FaceLandmarks = class {
+  constructor(relativeFaceLandmarkPositions, imgDims, shift = new Point(0, 0)) {
+    const { width, height } = imgDims;
+    this._imgDims = new Dimensions(width, height);
+    this._shift = shift;
+    this._positions = relativeFaceLandmarkPositions.map(
+      (pt) => pt.mul(new Point(width, height)).add(shift)
+    );
+  }
+  get shift() {
+    return new Point(this._shift.x, this._shift.y);
+  }
+  get imageWidth() {
+    return this._imgDims.width;
+  }
+  get imageHeight() {
+    return this._imgDims.height;
+  }
+  get positions() {
+    return this._positions;
+  }
+  get relativePositions() {
+    return this._positions.map(
+      (pt) => pt.sub(this._shift).div(new Point(this.imageWidth, this.imageHeight))
+    );
+  }
+  forSize(width, height) {
+    return new this.constructor(
+      this.relativePositions,
+      { width, height }
+    );
+  }
+  shiftBy(x, y) {
+    return new this.constructor(
+      this.relativePositions,
+      this._imgDims,
+      new Point(x, y)
+    );
+  }
+  shiftByPoint(pt) {
+    return this.shiftBy(pt.x, pt.y);
+  }
+  align(detection, options = {}) {
+    if (detection) {
+      const box = detection instanceof FaceDetection ? detection.box.floor() : new Box(detection);
+      return this.shiftBy(box.x, box.y).align(null, options);
+    }
+    const { useDlibAlignment, minBoxPadding } = { useDlibAlignment: false, minBoxPadding: 0.2, ...options };
+    if (useDlibAlignment) {
+      return this.alignDlib();
+    }
+    return this.alignMinBbox(minBoxPadding);
+  }
+  alignDlib() {
+    const centers = this.getRefPointsForAlignment();
+    const [leftEyeCenter, rightEyeCenter, mouthCenter] = centers;
+    const distToMouth = (pt) => mouthCenter.sub(pt).magnitude();
+    const eyeToMouthDist = (distToMouth(leftEyeCenter) + distToMouth(rightEyeCenter)) / 2;
+    const size = Math.floor(eyeToMouthDist / relScale);
+    const refPoint = getCenterPoint(centers);
+    const x = Math.floor(Math.max(0, refPoint.x - relX * size));
+    const y = Math.floor(Math.max(0, refPoint.y - relY * size));
+    return new Rect(x, y, Math.min(size, this.imageWidth + x), Math.min(size, this.imageHeight + y));
+  }
+  alignMinBbox(padding) {
+    const box = minBbox(this.positions);
+    return box.pad(box.width * padding, box.height * padding);
+  }
+  getRefPointsForAlignment() {
+    throw new Error("getRefPointsForAlignment not implemented by base class");
+  }
+};
+
+// src/classes/FaceLandmarks5.ts
+var FaceLandmarks5 = class extends FaceLandmarks {
+  getRefPointsForAlignment() {
+    const pts = this.positions;
+    return [
+      pts[0],
+      pts[1],
+      getCenterPoint([pts[3], pts[4]])
+    ];
+  }
+};
+
+// src/classes/FaceLandmarks68.ts
+var FaceLandmarks68 = class extends FaceLandmarks {
+  getJawOutline() {
+    return this.positions.slice(0, 17);
+  }
+  getLeftEyeBrow() {
+    return this.positions.slice(17, 22);
+  }
+  getRightEyeBrow() {
+    return this.positions.slice(22, 27);
+  }
+  getNose() {
+    return this.positions.slice(27, 36);
+  }
+  getLeftEye() {
+    return this.positions.slice(36, 42);
+  }
+  getRightEye() {
+    return this.positions.slice(42, 48);
+  }
+  getMouth() {
+    return this.positions.slice(48, 68);
+  }
+  getRefPointsForAlignment() {
+    return [
+      this.getLeftEye(),
+      this.getRightEye(),
+      this.getMouth()
+    ].map(getCenterPoint);
+  }
+};
+
+// src/classes/FaceMatch.ts
+var FaceMatch = class {
+  constructor(label, distance) {
+    this._label = label;
+    this._distance = distance;
+  }
+  get label() {
+    return this._label;
+  }
+  get distance() {
+    return this._distance;
+  }
+  toString(withDistance = true) {
+    return `${this.label}${withDistance ? ` (${round(this.distance)})` : ""}`;
+  }
+};
+
+// src/classes/LabeledBox.ts
+var LabeledBox = class extends Box {
+  constructor(box, label) {
+    super(box);
+    this._label = label;
+  }
+  static assertIsValidLabeledBox(box, callee) {
+    Box.assertIsValidBox(box, callee);
+    if (!isValidNumber(box.label)) {
+      throw new Error(`${callee} - expected property label (${box.label}) to be a number`);
+    }
+  }
+  get label() {
+    return this._label;
+  }
+};
+
+// src/classes/LabeledFaceDescriptors.ts
+var LabeledFaceDescriptors = class {
+  constructor(label, descriptors) {
+    if (!(typeof label === "string")) {
+      throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");
+    }
+    if (!Array.isArray(descriptors) || descriptors.some((desc) => !(desc instanceof Float32Array))) {
+      throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");
+    }
+    this._label = label;
+    this._descriptors = descriptors;
+  }
+  get label() {
+    return this._label;
+  }
+  get descriptors() {
+    return this._descriptors;
+  }
+  toJSON() {
+    return {
+      label: this.label,
+      descriptors: this.descriptors.map((d) => Array.from(d))
+    };
+  }
+  static fromJSON(json) {
+    const descriptors = json.descriptors.map((d) => new Float32Array(d));
+    return new LabeledFaceDescriptors(json.label, descriptors);
+  }
+};
+
+// src/classes/PredictedBox.ts
+var PredictedBox = class extends LabeledBox {
+  constructor(box, label, score, classScore) {
+    super(box, label);
+    this._score = score;
+    this._classScore = classScore;
+  }
+  static assertIsValidPredictedBox(box, callee) {
+    LabeledBox.assertIsValidLabeledBox(box, callee);
+    if (!isValidProbablitiy(box.score) || !isValidProbablitiy(box.classScore)) {
+      throw new Error(`${callee} - expected properties score (${box.score}) and (${box.classScore}) to be a number between [0, 1]`);
+    }
+  }
+  get score() {
+    return this._score;
+  }
+  get classScore() {
+    return this._classScore;
+  }
+};
+
+// src/factories/WithFaceDetection.ts
+function isWithFaceDetection(obj) {
+  return obj.detection instanceof FaceDetection;
+}
+function extendWithFaceDetection(sourceObj, detection) {
+  const extension = { detection };
+  return { ...sourceObj, ...extension };
+}
+
+// src/env/createBrowserEnv.ts
+function createBrowserEnv() {
+  const fetch = window.fetch;
+  if (!fetch)
+    throw new Error("fetch - missing fetch implementation for browser environment");
+  const readFile = () => {
+    throw new Error("readFile - filesystem not available for browser environment");
+  };
+  return {
+    Canvas: HTMLCanvasElement,
+    CanvasRenderingContext2D,
+    Image: HTMLImageElement,
+    ImageData,
+    Video: HTMLVideoElement,
+    createCanvasElement: () => document.createElement("canvas"),
+    createImageElement: () => document.createElement("img"),
+    createVideoElement: () => document.createElement("video"),
+    fetch,
+    readFile
+  };
+}
+
+// src/env/isNodejs.ts
+function isNodejs() {
+  return typeof global === "object" && typeof process !== "undefined" && process.versions != null && process.versions.node != null;
+}
+
+// src/env/createFileSystem.ts
+function createFileSystem(fs) {
+  let requireFsError = "";
+  if (!fs && isNodejs()) {
+    try {
+      fs = require("fs");
+    } catch (err) {
+      requireFsError = err.toString();
+    }
+  }
+  const readFile = fs ? (filePath) => new Promise((resolve, reject) => {
+    fs.readFile(filePath, (err, buffer) => err ? reject(err) : resolve(buffer));
+  }) : () => {
+    throw new Error(`readFile - failed to require fs in nodejs environment with error: ${requireFsError}`);
+  };
+  return { readFile };
+}
+
+// src/env/createNodejsEnv.ts
+function createNodejsEnv() {
+  const Canvas = global["Canvas"] || global.HTMLCanvasElement;
+  const Image = global.Image || global.HTMLImageElement;
+  const Video = global["Video"] || global.HTMLVideoElement;
+  const createCanvasElement = () => {
+    if (Canvas)
+      return new Canvas();
+    throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment");
+  };
+  const createImageElement = () => {
+    if (Image)
+      return new Image();
+    throw new Error("createImageElement - missing Image implementation for nodejs environment");
+  };
+  const createVideoElement = () => {
+    if (Video)
+      return new Video();
+    throw new Error("createVideoElement - missing Video implementation for nodejs environment");
+  };
+  const fetch = global.fetch;
+  const fileSystem = createFileSystem();
+  return {
+    Canvas: Canvas || class {
+    },
+    CanvasRenderingContext2D: global.CanvasRenderingContext2D || class {
+    },
+    Image: Image || class {
+    },
+    ImageData: global.ImageData || class {
+    },
+    Video: global.HTMLVideoElement || class {
+    },
+    createCanvasElement,
+    createImageElement,
+    createVideoElement,
+    fetch,
+    ...fileSystem
+  };
+}
+
+// src/env/isBrowser.ts
+function isBrowser() {
+  return typeof window === "object" && typeof document !== "undefined" && typeof HTMLImageElement !== "undefined" && typeof HTMLCanvasElement !== "undefined" && typeof HTMLVideoElement !== "undefined" && typeof ImageData !== "undefined" && typeof CanvasRenderingContext2D !== "undefined";
+}
+
+// src/env/index.ts
+var environment;
+function getEnv() {
+  if (!environment) {
+    throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  return environment;
+}
+function setEnv(env2) {
+  environment = env2;
+}
+function initialize() {
+  if (isBrowser())
+    return setEnv(createBrowserEnv());
+  if (isNodejs())
+    return setEnv(createNodejsEnv());
+  return null;
+}
+function monkeyPatch(env2) {
+  if (!environment) {
+    initialize();
+  }
+  if (!environment) {
+    throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");
+  }
+  const { Canvas = environment.Canvas, Image = environment.Image } = env2;
+  environment.Canvas = Canvas;
+  environment.Image = Image;
+  environment.createCanvasElement = env2.createCanvasElement || (() => new Canvas());
+  environment.createImageElement = env2.createImageElement || (() => new Image());
+  environment.ImageData = env2.ImageData || environment.ImageData;
+  environment.Video = env2.Video || environment.Video;
+  environment.fetch = env2.fetch || environment.fetch;
+  environment.readFile = env2.readFile || environment.readFile;
+}
+var env = {
+  getEnv,
+  setEnv,
+  initialize,
+  createBrowserEnv,
+  createFileSystem,
+  createNodejsEnv,
+  monkeyPatch,
+  isBrowser,
+  isNodejs
+};
+initialize();
+
+// src/dom/resolveInput.ts
+function resolveInput(arg) {
+  if (!env.isNodejs() && typeof arg === "string") {
+    return document.getElementById(arg);
+  }
+  return arg;
+}
+
+// src/dom/getContext2dOrThrow.ts
+function getContext2dOrThrow(canvasArg) {
+  const { Canvas, CanvasRenderingContext2D: CanvasRenderingContext2D2 } = env.getEnv();
+  if (canvasArg instanceof CanvasRenderingContext2D2) {
+    return canvasArg;
+  }
+  const canvas = resolveInput(canvasArg);
+  if (!(canvas instanceof Canvas)) {
+    throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");
+  }
+  const ctx = canvas.getContext("2d");
+  if (!ctx) {
+    throw new Error("resolveContext2d - canvas 2d context is null");
+  }
+  return ctx;
+}
+
+// src/draw/DrawTextField.ts
+var AnchorPosition = /* @__PURE__ */ ((AnchorPosition2) => {
+  AnchorPosition2["TOP_LEFT"] = "TOP_LEFT";
+  AnchorPosition2["TOP_RIGHT"] = "TOP_RIGHT";
+  AnchorPosition2["BOTTOM_LEFT"] = "BOTTOM_LEFT";
+  AnchorPosition2["BOTTOM_RIGHT"] = "BOTTOM_RIGHT";
+  return AnchorPosition2;
+})(AnchorPosition || {});
+var DrawTextFieldOptions = class {
+  constructor(options = {}) {
+    const {
+      anchorPosition,
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = options;
+    this.anchorPosition = anchorPosition || "TOP_LEFT" /* TOP_LEFT */;
+    this.backgroundColor = backgroundColor || "rgba(0, 0, 0, 0.5)";
+    this.fontColor = fontColor || "rgba(255, 255, 255, 1)";
+    this.fontSize = fontSize || 14;
+    this.fontStyle = fontStyle || "Georgia";
+    this.padding = padding || 4;
+  }
+};
+var DrawTextField = class {
+  constructor(text, anchor, options = {}) {
+    this.text = typeof text === "string" ? [text] : text instanceof DrawTextField ? text.text : text;
+    this.anchor = anchor;
+    this.options = new DrawTextFieldOptions(options);
+  }
+  measureWidth(ctx) {
+    const { padding } = this.options;
+    return this.text.map((l) => ctx.measureText(l).width).reduce((w0, w1) => w0 < w1 ? w1 : w0, 0) + 2 * padding;
+  }
+  measureHeight() {
+    const { fontSize, padding } = this.options;
+    return this.text.length * fontSize + 2 * padding;
+  }
+  getUpperLeft(ctx, canvasDims) {
+    const { anchorPosition } = this.options;
+    const isShiftLeft = anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */ || anchorPosition === "TOP_RIGHT" /* TOP_RIGHT */;
+    const isShiftTop = anchorPosition === "BOTTOM_LEFT" /* BOTTOM_LEFT */ || anchorPosition === "BOTTOM_RIGHT" /* BOTTOM_RIGHT */;
+    const textFieldWidth = this.measureWidth(ctx);
+    const textFieldHeight = this.measureHeight();
+    const x = isShiftLeft ? this.anchor.x - textFieldWidth : this.anchor.x;
+    const y = isShiftTop ? this.anchor.y - textFieldHeight : this.anchor.y;
+    if (canvasDims) {
+      const { width, height } = canvasDims;
+      const newX = Math.max(Math.min(x, width - textFieldWidth), 0);
+      const newY = Math.max(Math.min(y, height - textFieldHeight), 0);
+      return { x: newX, y: newY };
+    }
+    return { x, y };
+  }
+  draw(canvasArg) {
+    const canvas = resolveInput(canvasArg);
+    const ctx = getContext2dOrThrow(canvas);
+    const {
+      backgroundColor,
+      fontColor,
+      fontSize,
+      fontStyle,
+      padding
+    } = this.options;
+    ctx.font = `${fontSize}px ${fontStyle}`;
+    const maxTextWidth = this.measureWidth(ctx);
+    const textHeight = this.measureHeight();
+    ctx.fillStyle = backgroundColor;
+    const upperLeft = this.getUpperLeft(ctx, canvas);
+    ctx.fillRect(upperLeft.x, upperLeft.y, maxTextWidth, textHeight);
+    ctx.fillStyle = fontColor;
+    this.text.forEach((textLine, i) => {
+      const x = padding + upperLeft.x;
+      const y = padding + upperLeft.y + (i + 1) * fontSize;
+      ctx.fillText(textLine, x, y);
+    });
+  }
+};
+
+// src/draw/DrawBox.ts
+var DrawBoxOptions = class {
+  constructor(options = {}) {
+    const {
+      boxColor,
+      lineWidth,
+      label,
+      drawLabelOptions
+    } = options;
+    this.boxColor = boxColor || "rgba(0, 0, 255, 1)";
+    this.lineWidth = lineWidth || 2;
+    this.label = label;
+    const defaultDrawLabelOptions = {
+      anchorPosition: "BOTTOM_LEFT" /* BOTTOM_LEFT */,
+      backgroundColor: this.boxColor
+    };
+    this.drawLabelOptions = new DrawTextFieldOptions({ ...defaultDrawLabelOptions, ...drawLabelOptions });
+  }
+};
+var DrawBox = class {
+  constructor(box, options = {}) {
+    this.box = new Box(box);
+    this.options = new DrawBoxOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const { boxColor, lineWidth } = this.options;
+    const {
+      x,
+      y,
+      width,
+      height
+    } = this.box;
+    ctx.strokeStyle = boxColor;
+    ctx.lineWidth = lineWidth;
+    ctx.strokeRect(x, y, width, height);
+    const { label } = this.options;
+    if (label) {
+      new DrawTextField([label], { x: x - lineWidth / 2, y }, this.options.drawLabelOptions).draw(canvasArg);
+    }
+  }
+};
+
+// src/draw/drawDetections.ts
+function drawDetections(canvasArg, detections) {
+  const detectionsArray = Array.isArray(detections) ? detections : [detections];
+  detectionsArray.forEach((det) => {
+    const score = det instanceof FaceDetection ? det.score : isWithFaceDetection(det) ? det.detection.score : void 0;
+    const box = det instanceof FaceDetection ? det.box : isWithFaceDetection(det) ? det.detection.box : new Box(det);
+    const label = score ? `${round(score)}` : void 0;
+    new DrawBox(box, { label }).draw(canvasArg);
+  });
+}
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var tf18 = __toESM(require_tfjs_esm());
+
+// src/dom/isMediaLoaded.ts
+function isMediaLoaded(media) {
+  const { Image, Video } = env.getEnv();
+  return media instanceof Image && media.complete || media instanceof Video && media.readyState >= 3;
+}
+
+// src/dom/awaitMediaLoaded.ts
+function awaitMediaLoaded(media) {
+  return new Promise((resolve, reject) => {
+    if (media instanceof env.getEnv().Canvas || isMediaLoaded(media))
+      resolve(null);
+    function onError(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      reject(e);
+    }
+    function onLoad(e) {
+      if (!e.currentTarget)
+        return;
+      e.currentTarget.removeEventListener("load", onLoad);
+      e.currentTarget.removeEventListener("error", onError);
+      resolve(e);
+    }
+    media.addEventListener("load", onLoad);
+    media.addEventListener("error", onError);
+  });
+}
+
+// src/dom/bufferToImage.ts
+function bufferToImage(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToImage - expected buf to be of type: Blob"));
+    const reader = new FileReader();
+    reader.onload = () => {
+      if (typeof reader.result !== "string")
+        reject(new Error("bufferToImage - expected reader.result to be a string, in onload"));
+      const img = env.getEnv().createImageElement();
+      img.onload = () => resolve(img);
+      img.onerror = reject;
+      img.src = reader.result;
+    };
+    reader.onerror = reject;
+    reader.readAsDataURL(buf);
+  });
+}
+
+// src/dom/getMediaDimensions.ts
+function getMediaDimensions(input) {
+  const { Image, Video } = env.getEnv();
+  if (input instanceof Image) {
+    return new Dimensions(input.naturalWidth, input.naturalHeight);
+  }
+  if (input instanceof Video) {
+    return new Dimensions(input.videoWidth, input.videoHeight);
+  }
+  return new Dimensions(input.width, input.height);
+}
+
+// src/dom/createCanvas.ts
+function createCanvas({ width, height }) {
+  const { createCanvasElement } = env.getEnv();
+  const canvas = createCanvasElement();
+  canvas.width = width;
+  canvas.height = height;
+  return canvas;
+}
+function createCanvasFromMedia(media, dims) {
+  const { ImageData: ImageData2 } = env.getEnv();
+  if (!(media instanceof ImageData2) && !isMediaLoaded(media)) {
+    throw new Error("createCanvasFromMedia - media has not finished loading yet");
+  }
+  const { width, height } = dims || getMediaDimensions(media);
+  const canvas = createCanvas({ width, height });
+  if (media instanceof ImageData2) {
+    getContext2dOrThrow(canvas).putImageData(media, 0, 0);
+  } else {
+    getContext2dOrThrow(canvas).drawImage(media, 0, 0, width, height);
+  }
+  return canvas;
+}
+
+// src/dom/imageTensorToCanvas.ts
+var tf4 = __toESM(require_tfjs_esm());
+async function imageTensorToCanvas(imgTensor, canvas) {
+  const targetCanvas = canvas || env.getEnv().createCanvasElement();
+  const [height, width, numChannels] = imgTensor.shape.slice(isTensor4D(imgTensor) ? 1 : 0);
+  const imgTensor3D = tf4.tidy(() => imgTensor.as3D(height, width, numChannels).toInt());
+  await tf4["browser"].toPixels(imgTensor3D, targetCanvas);
+  imgTensor3D.dispose();
+  return targetCanvas;
+}
+
+// src/dom/isMediaElement.ts
+function isMediaElement(input) {
+  const { Image, Canvas, Video } = env.getEnv();
+  return input instanceof Image || input instanceof Canvas || input instanceof Video;
+}
+
+// src/dom/NetInput.ts
+var tf5 = __toESM(require_tfjs_esm());
+
+// src/dom/imageToSquare.ts
+function imageToSquare(input, inputSize, centerImage = false) {
+  const { Image, Canvas } = env.getEnv();
+  if (!(input instanceof Image || input instanceof Canvas)) {
+    throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");
+  }
+  if (inputSize <= 0)
+    return createCanvas({ width: 1, height: 1 });
+  const dims = getMediaDimensions(input);
+  const scale2 = inputSize / Math.max(dims.height, dims.width);
+  const width = scale2 * dims.width;
+  const height = scale2 * dims.height;
+  const targetCanvas = createCanvas({ width: inputSize, height: inputSize });
+  const inputCanvas = input instanceof Canvas ? input : createCanvasFromMedia(input);
+  const offset = Math.abs(width - height) / 2;
+  const dx = centerImage && width < height ? offset : 0;
+  const dy = centerImage && height < width ? offset : 0;
+  if (inputCanvas.width > 0 && inputCanvas.height > 0)
+    getContext2dOrThrow(targetCanvas).drawImage(inputCanvas, dx, dy, width, height);
+  return targetCanvas;
+}
+
+// src/dom/NetInput.ts
+var NetInput = class {
+  constructor(inputs, treatAsBatchInput = false) {
+    this._imageTensors = [];
+    this._canvases = [];
+    this._treatAsBatchInput = false;
+    this._inputDimensions = [];
+    this._inputSize = 0;
+    if (!Array.isArray(inputs)) {
+      throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${inputs}`);
+    }
+    this._treatAsBatchInput = treatAsBatchInput;
+    this._batchSize = inputs.length;
+    inputs.forEach((input, idx) => {
+      if (isTensor3D(input)) {
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape;
+        return;
+      }
+      if (isTensor4D(input)) {
+        const batchSize = input.shape[0];
+        if (batchSize !== 1) {
+          throw new Error(`NetInput - tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+        }
+        this._imageTensors[idx] = input;
+        this._inputDimensions[idx] = input.shape.slice(1);
+        return;
+      }
+      const canvas = input instanceof env.getEnv().Canvas ? input : createCanvasFromMedia(input);
+      this._canvases[idx] = canvas;
+      this._inputDimensions[idx] = [canvas.height, canvas.width, 3];
+    });
+  }
+  get imageTensors() {
+    return this._imageTensors;
+  }
+  get canvases() {
+    return this._canvases;
+  }
+  get isBatchInput() {
+    return this.batchSize > 1 || this._treatAsBatchInput;
+  }
+  get batchSize() {
+    return this._batchSize;
+  }
+  get inputDimensions() {
+    return this._inputDimensions;
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get reshapedInputDimensions() {
+    return range(this.batchSize, 0, 1).map(
+      (_, batchIdx) => this.getReshapedInputDimensions(batchIdx)
+    );
+  }
+  getInput(batchIdx) {
+    return this.canvases[batchIdx] || this.imageTensors[batchIdx];
+  }
+  getInputDimensions(batchIdx) {
+    return this._inputDimensions[batchIdx];
+  }
+  getInputHeight(batchIdx) {
+    return this._inputDimensions[batchIdx][0];
+  }
+  getInputWidth(batchIdx) {
+    return this._inputDimensions[batchIdx][1];
+  }
+  getReshapedInputDimensions(batchIdx) {
+    if (typeof this.inputSize !== "number") {
+      throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");
+    }
+    const width = this.getInputWidth(batchIdx);
+    const height = this.getInputHeight(batchIdx);
+    return computeReshapedDimensions({ width, height }, this.inputSize);
+  }
+  toBatchTensor(inputSize, isCenterInputs = true) {
+    this._inputSize = inputSize;
+    return tf5.tidy(() => {
+      const inputTensors = range(this.batchSize, 0, 1).map((batchIdx) => {
+        const input = this.getInput(batchIdx);
+        if (input instanceof tf5.Tensor) {
+          let imgTensor = isTensor4D(input) ? input : tf5.expandDims(input);
+          imgTensor = padToSquare(imgTensor, isCenterInputs);
+          if (imgTensor.shape[1] !== inputSize || imgTensor.shape[2] !== inputSize) {
+            imgTensor = tf5["image"].resizeBilinear(imgTensor, [inputSize, inputSize], false, false);
+          }
+          return imgTensor.as3D(inputSize, inputSize, 3);
+        }
+        if (input instanceof env.getEnv().Canvas) {
+          return tf5["browser"].fromPixels(imageToSquare(input, inputSize, isCenterInputs));
+        }
+        throw new Error(`toBatchTensor - at batchIdx ${batchIdx}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${input}`);
+      });
+      const batchTensor = tf5.stack(inputTensors.map((t) => tf5.cast(t, "float32"))).as4D(this.batchSize, inputSize, inputSize, 3);
+      return batchTensor;
+    });
+  }
+};
+
+// src/dom/toNetInput.ts
+async function toNetInput(inputs) {
+  if (inputs instanceof NetInput)
+    return inputs;
+  const inputArgArray = Array.isArray(inputs) ? inputs : [inputs];
+  if (!inputArgArray.length)
+    throw new Error("toNetInput - empty array passed as input");
+  const getIdxHint = (idx) => Array.isArray(inputs) ? ` at input index ${idx}:` : "";
+  const inputArray = inputArgArray.map(resolveInput);
+  inputArray.forEach((input, i) => {
+    if (!isMediaElement(input) && !isTensor3D(input) && !isTensor4D(input)) {
+      if (typeof inputArgArray[i] === "string")
+        throw new Error(`toNetInput -${getIdxHint(i)} string passed, but could not resolve HTMLElement for element id ${inputArgArray[i]}`);
+      throw new Error(`toNetInput -${getIdxHint(i)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);
+    }
+    if (isTensor4D(input)) {
+      const batchSize = input.shape[0];
+      if (batchSize !== 1)
+        throw new Error(`toNetInput -${getIdxHint(i)} tf.Tensor4D with batchSize ${batchSize} passed, but not supported in input array`);
+    }
+  });
+  await Promise.all(inputArray.map((input) => isMediaElement(input) && awaitMediaLoaded(input)));
+  return new NetInput(inputArray, Array.isArray(inputs));
+}
+
+// src/dom/extractFaces.ts
+async function extractFaces(input, detections) {
+  const { Canvas } = env.getEnv();
+  let canvas = input;
+  if (!(input instanceof Canvas)) {
+    const netInput = await toNetInput(input);
+    if (netInput.batchSize > 1)
+      throw new Error("extractFaces - batchSize > 1 not supported");
+    const tensorOrCanvas = netInput.getInput(0);
+    canvas = tensorOrCanvas instanceof Canvas ? tensorOrCanvas : await imageTensorToCanvas(tensorOrCanvas);
+  }
+  const ctx = getContext2dOrThrow(canvas);
+  const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(canvas.width, canvas.height).box.floor() : det).map((box) => box.clipAtImageBorders(canvas.width, canvas.height));
+  return boxes.map(({ x, y, width, height }) => {
+    const faceImg = createCanvas({ width, height });
+    if (width > 0 && height > 0)
+      getContext2dOrThrow(faceImg).putImageData(ctx.getImageData(x, y, width, height), 0, 0);
+    return faceImg;
+  });
+}
+
+// src/dom/extractFaceTensors.ts
+var tf6 = __toESM(require_tfjs_esm());
+async function extractFaceTensors(imageTensor, detections) {
+  if (!isTensor3D(imageTensor) && !isTensor4D(imageTensor)) {
+    throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");
+  }
+  if (isTensor4D(imageTensor) && imageTensor.shape[0] > 1) {
+    throw new Error("extractFaceTensors - batchSize > 1 not supported");
+  }
+  return tf6.tidy(() => {
+    const [imgHeight, imgWidth, numChannels] = imageTensor.shape.slice(isTensor4D(imageTensor) ? 1 : 0);
+    const boxes = detections.map((det) => det instanceof FaceDetection ? det.forSize(imgWidth, imgHeight).box : det).map((box) => box.clipAtImageBorders(imgWidth, imgHeight));
+    const faceTensors = boxes.filter((box) => box.width > 0 && box.height > 0).map(({ x, y, width, height }) => tf6.slice3d(imageTensor.as3D(imgHeight, imgWidth, numChannels), [y, x, 0], [height, width, numChannels]));
+    return faceTensors;
+  });
+}
+
+// src/dom/fetchOrThrow.ts
+async function fetchOrThrow(url, init) {
+  const { fetch } = env.getEnv();
+  const res = await fetch(url, init);
+  if (!(res.status < 400)) {
+    throw new Error(`failed to fetch: (${res.status}) ${res.statusText}, from url: ${res.url}`);
+  }
+  return res;
+}
+
+// src/dom/fetchImage.ts
+async function fetchImage(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("image/")) {
+    throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToImage(blob);
+}
+
+// src/dom/fetchJson.ts
+async function fetchJson(uri) {
+  return (await fetchOrThrow(uri)).json();
+}
+
+// src/dom/fetchNetWeights.ts
+async function fetchNetWeights(uri) {
+  return new Float32Array(await (await fetchOrThrow(uri)).arrayBuffer());
+}
+
+// src/dom/bufferToVideo.ts
+function bufferToVideo(buf) {
+  return new Promise((resolve, reject) => {
+    if (!(buf instanceof Blob))
+      reject(new Error("bufferToVideo - expected buf to be of type: Blob"));
+    const video = env.getEnv().createVideoElement();
+    video.oncanplay = () => resolve(video);
+    video.onerror = reject;
+    video.playsInline = true;
+    video.muted = true;
+    video.src = URL.createObjectURL(buf);
+    video.play();
+  });
+}
+
+// src/dom/fetchVideo.ts
+async function fetchVideo(uri) {
+  const res = await fetchOrThrow(uri);
+  const blob = await res.blob();
+  if (!blob.type.startsWith("video/")) {
+    throw new Error(`fetchVideo - expected blob type to be of type video/*, instead have: ${blob.type}, for url: ${res.url}`);
+  }
+  return bufferToVideo(blob);
+}
+
+// src/dom/loadWeightMap.ts
+var tf7 = __toESM(require_tfjs_esm());
+
+// src/common/getModelUris.ts
+function getModelUris(uri, defaultModelName) {
+  const defaultManifestFilename = `${defaultModelName}-weights_manifest.json`;
+  if (!uri) {
+    return {
+      modelBaseUri: "",
+      manifestUri: defaultManifestFilename
+    };
+  }
+  if (uri === "/") {
+    return {
+      modelBaseUri: "/",
+      manifestUri: `/${defaultManifestFilename}`
+    };
+  }
+  const protocol = uri.startsWith("http://") ? "http://" : uri.startsWith("https://") ? "https://" : "";
+  uri = uri.replace(protocol, "");
+  const parts = uri.split("/").filter((s) => s);
+  const manifestFile = uri.endsWith(".json") ? parts[parts.length - 1] : defaultManifestFilename;
+  let modelBaseUri = protocol + (uri.endsWith(".json") ? parts.slice(0, parts.length - 1) : parts).join("/");
+  modelBaseUri = uri.startsWith("/") ? `/${modelBaseUri}` : modelBaseUri;
+  return {
+    modelBaseUri,
+    manifestUri: modelBaseUri === "/" ? `/${manifestFile}` : `${modelBaseUri}/${manifestFile}`
+  };
+}
+
+// src/dom/loadWeightMap.ts
+async function loadWeightMap(uri, defaultModelName) {
+  const { manifestUri, modelBaseUri } = getModelUris(uri, defaultModelName);
+  const manifest = await fetchJson(manifestUri);
+  return tf7["io"].loadWeights(manifest, modelBaseUri);
+}
+
+// src/dom/matchDimensions.ts
+function matchDimensions(input, reference, useMediaDimensions = false) {
+  const { width, height } = useMediaDimensions ? getMediaDimensions(reference) : reference;
+  input.width = width;
+  input.height = height;
+  return { width, height };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var tf15 = __toESM(require_tfjs_esm());
+
+// src/NeuralNetwork.ts
+var tf8 = __toESM(require_tfjs_esm());
+var NeuralNetwork = class {
+  constructor(name) {
+    this._params = void 0;
+    this._paramMappings = [];
+    this._name = name;
+  }
+  get params() {
+    return this._params;
+  }
+  get paramMappings() {
+    return this._paramMappings;
+  }
+  get isLoaded() {
+    return !!this.params;
+  }
+  getParamFromPath(paramPath) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    return obj[objProp];
+  }
+  reassignParamFromPath(paramPath, tensor2) {
+    const { obj, objProp } = this.traversePropertyPath(paramPath);
+    obj[objProp].dispose();
+    obj[objProp] = tensor2;
+  }
+  getParamList() {
+    return this._paramMappings.map(({ paramPath }) => ({
+      path: paramPath,
+      tensor: this.getParamFromPath(paramPath)
+    }));
+  }
+  getTrainableParams() {
+    return this.getParamList().filter((param) => param.tensor instanceof tf8.Variable);
+  }
+  getFrozenParams() {
+    return this.getParamList().filter((param) => !(param.tensor instanceof tf8.Variable));
+  }
+  variable() {
+    this.getFrozenParams().forEach(({ path, tensor: tensor2 }) => {
+      this.reassignParamFromPath(path, tensor2.variable());
+    });
+  }
+  freeze() {
+    this.getTrainableParams().forEach(({ path, tensor: variable }) => {
+      const tensor2 = tf8.tensor(variable.dataSync());
+      variable.dispose();
+      this.reassignParamFromPath(path, tensor2);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.getParamList().forEach((param) => {
+      if (throwOnRedispose && param.tensor.isDisposed) {
+        throw new Error(`param tensor has already been disposed for path ${param.path}`);
+      }
+      param.tensor.dispose();
+    });
+    this._params = void 0;
+  }
+  serializeParams() {
+    return new Float32Array(
+      this.getParamList().map(({ tensor: tensor2 }) => Array.from(tensor2.dataSync())).reduce((flat, arr) => flat.concat(arr))
+    );
+  }
+  async load(weightsOrUrl) {
+    if (weightsOrUrl instanceof Float32Array) {
+      this.extractWeights(weightsOrUrl);
+      return;
+    }
+    await this.loadFromUri(weightsOrUrl);
+  }
+  async loadFromUri(uri) {
+    if (uri && typeof uri !== "string") {
+      throw new Error(`${this._name}.loadFromUri - expected model uri`);
+    }
+    const weightMap = await loadWeightMap(uri, this.getDefaultModelName());
+    this.loadFromWeightMap(weightMap);
+  }
+  async loadFromDisk(filePath) {
+    if (filePath && typeof filePath !== "string") {
+      throw new Error(`${this._name}.loadFromDisk - expected model file path`);
+    }
+    const { readFile } = env.getEnv();
+    const { manifestUri, modelBaseUri } = getModelUris(filePath, this.getDefaultModelName());
+    const fetchWeightsFromDisk = (filePaths) => Promise.all(filePaths.map((fp) => readFile(fp).then((buf) => buf.buffer)));
+    const loadWeights = tf8["io"].weightsLoaderFactory(fetchWeightsFromDisk);
+    const manifest = JSON.parse((await readFile(manifestUri)).toString());
+    const weightMap = await loadWeights(manifest, modelBaseUri);
+    this.loadFromWeightMap(weightMap);
+  }
+  loadFromWeightMap(weightMap) {
+    const { paramMappings, params } = this.extractParamsFromWeightMap(weightMap);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  extractWeights(weights) {
+    const { paramMappings, params } = this.extractParams(weights);
+    this._paramMappings = paramMappings;
+    this._params = params;
+  }
+  traversePropertyPath(paramPath) {
+    if (!this.params) {
+      throw new Error("traversePropertyPath - model has no loaded params");
+    }
+    const result = paramPath.split("/").reduce((res, objProp2) => {
+      if (!res.nextObj.hasOwnProperty(objProp2)) {
+        throw new Error(`traversePropertyPath - object does not have property ${objProp2}, for path ${paramPath}`);
+      }
+      return { obj: res.nextObj, objProp: objProp2, nextObj: res.nextObj[objProp2] };
+    }, { nextObj: this.params });
+    const { obj, objProp } = result;
+    if (!obj || !objProp || !(obj[objProp] instanceof tf8.Tensor)) {
+      throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${paramPath}`);
+    }
+    return { obj, objProp };
+  }
+};
+
+// src/faceFeatureExtractor/denseBlock.ts
+var tf10 = __toESM(require_tfjs_esm());
+
+// src/common/depthwiseSeparableConv.ts
+var tf9 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv(x, params, stride) {
+  return tf9.tidy(() => {
+    let out = tf9.separableConv2d(x, params.depthwise_filter, params.pointwise_filter, stride, "same");
+    out = tf9.add(out, params.bias);
+    return out;
+  });
+}
+
+// src/faceFeatureExtractor/denseBlock.ts
+function denseBlock3(x, denseBlockParams, isFirstLayer = false) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, [2, 2], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, [2, 2])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+  });
+}
+function denseBlock4(x, denseBlockParams, isFirstLayer = false, isScaleDown = true) {
+  return tf10.tidy(() => {
+    const out1 = tf10.relu(
+      isFirstLayer ? tf10.add(
+        tf10.conv2d(x, denseBlockParams.conv0.filters, isScaleDown ? [2, 2] : [1, 1], "same"),
+        denseBlockParams.conv0.bias
+      ) : depthwiseSeparableConv(x, denseBlockParams.conv0, isScaleDown ? [2, 2] : [1, 1])
+    );
+    const out2 = depthwiseSeparableConv(out1, denseBlockParams.conv1, [1, 1]);
+    const in3 = tf10.relu(tf10.add(out1, out2));
+    const out3 = depthwiseSeparableConv(in3, denseBlockParams.conv2, [1, 1]);
+    const in4 = tf10.relu(tf10.add(out1, tf10.add(out2, out3)));
+    const out4 = depthwiseSeparableConv(in4, denseBlockParams.conv3, [1, 1]);
+    return tf10.relu(tf10.add(out1, tf10.add(out2, tf10.add(out3, out4))));
+  });
+}
+
+// src/common/convLayer.ts
+var tf11 = __toESM(require_tfjs_esm());
+function convLayer(x, params, padding = "same", withRelu = false) {
+  return tf11.tidy(() => {
+    const out = tf11.add(
+      tf11.conv2d(x, params.filters, [1, 1], padding),
+      params.bias
+    );
+    return withRelu ? tf11.relu(out) : out;
+  });
+}
+
+// src/common/disposeUnusedWeightTensors.ts
+function disposeUnusedWeightTensors(weightMap, paramMappings) {
+  Object.keys(weightMap).forEach((path) => {
+    if (!paramMappings.some((pm) => pm.originalPath === path)) {
+      weightMap[path].dispose();
+    }
+  });
+}
+
+// src/common/extractConvParamsFactory.ts
+var tf12 = __toESM(require_tfjs_esm());
+function extractConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, filterSize, mappedPrefix) => {
+    const filters = tf12.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf12.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  };
+}
+
+// src/common/extractFCParamsFactory.ts
+var tf13 = __toESM(require_tfjs_esm());
+function extractFCParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const fc_weights = tf13.tensor2d(extractWeights(channelsIn * channelsOut), [channelsIn, channelsOut]);
+    const fc_bias = tf13.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return {
+      weights: fc_weights,
+      bias: fc_bias
+    };
+  };
+}
+
+// src/common/extractSeparableConvParamsFactory.ts
+var tf14 = __toESM(require_tfjs_esm());
+
+// src/common/types.ts
+var SeparableConvParams = class {
+  constructor(depthwise_filter, pointwise_filter, bias) {
+    this.depthwise_filter = depthwise_filter;
+    this.pointwise_filter = pointwise_filter;
+    this.bias = bias;
+  }
+};
+
+// src/common/extractSeparableConvParamsFactory.ts
+function extractSeparableConvParamsFactory(extractWeights, paramMappings) {
+  return (channelsIn, channelsOut, mappedPrefix) => {
+    const depthwise_filter = tf14.tensor4d(extractWeights(3 * 3 * channelsIn), [3, 3, channelsIn, 1]);
+    const pointwise_filter = tf14.tensor4d(extractWeights(channelsIn * channelsOut), [1, 1, channelsIn, channelsOut]);
+    const bias = tf14.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/depthwise_filter` },
+      { paramPath: `${mappedPrefix}/pointwise_filter` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+function loadSeparableConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const depthwise_filter = extractWeightEntry(`${prefix}/depthwise_filter`, 4);
+    const pointwise_filter = extractWeightEntry(`${prefix}/pointwise_filter`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return new SeparableConvParams(
+      depthwise_filter,
+      pointwise_filter,
+      bias
+    );
+  };
+}
+
+// src/common/extractWeightEntryFactory.ts
+function extractWeightEntryFactory(weightMap, paramMappings) {
+  return (originalPath, paramRank, mappedPath) => {
+    const tensor2 = weightMap[originalPath];
+    if (!isTensor(tensor2, paramRank)) {
+      throw new Error(`expected weightMap[${originalPath}] to be a Tensor${paramRank}D, instead have ${tensor2}`);
+    }
+    paramMappings.push(
+      { originalPath, paramPath: mappedPath || originalPath }
+    );
+    return tensor2;
+  };
+}
+
+// src/common/extractWeightsFactory.ts
+function extractWeightsFactory(weights) {
+  let remainingWeights = weights;
+  function extractWeights(numWeights) {
+    const ret = remainingWeights.slice(0, numWeights);
+    remainingWeights = remainingWeights.slice(numWeights);
+    return ret;
+  }
+  function getRemainingWeights() {
+    return remainingWeights;
+  }
+  return {
+    extractWeights,
+    getRemainingWeights
+  };
+}
+
+// src/faceFeatureExtractor/extractorsFactory.ts
+function extractorsFactory(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv0`) : extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/conv0`);
+    const conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv1`);
+    const conv22 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer = false) {
+    const { conv0, conv1, conv2: conv22 } = extractDenseBlock3Params(channelsIn, channelsOut, mappedPrefix, isFirstLayer);
+    const conv3 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParams.ts
+function extractParams(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock4Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock4Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock4Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock4Params(64, 128, "dense2");
+  const dense3 = extractDenseBlock4Params(128, 256, "dense3");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: {
+      dense0,
+      dense1,
+      dense2,
+      dense3
+    }
+  };
+}
+
+// src/common/loadConvParamsFactory.ts
+function loadConvParamsFactory(extractWeightEntry) {
+  return (prefix) => {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  };
+}
+
+// src/faceFeatureExtractor/loadParamsFactory.ts
+function loadParamsFactory(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractDenseBlock3Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    return { conv0, conv1, conv2: conv22 };
+  }
+  function extractDenseBlock4Params(prefix, isFirstLayer = false) {
+    const conv0 = isFirstLayer ? extractConvParams(`${prefix}/conv0`) : extractSeparableConvParams(`${prefix}/conv0`);
+    const conv1 = extractSeparableConvParams(`${prefix}/conv1`);
+    const conv22 = extractSeparableConvParams(`${prefix}/conv2`);
+    const conv3 = extractSeparableConvParams(`${prefix}/conv3`);
+    return {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3
+    };
+  }
+  return {
+    extractDenseBlock3Params,
+    extractDenseBlock4Params
+  };
+}
+
+// src/faceFeatureExtractor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock4Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock4Params("dense0", true),
+    dense1: extractDenseBlock4Params("dense1"),
+    dense2: extractDenseBlock4Params("dense2"),
+    dense3: extractDenseBlock4Params("dense3")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/FaceFeatureExtractor.ts
+var FaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("FaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceFeatureExtractor - load model before inference");
+    }
+    return tf15.tidy(() => {
+      const batchTensor = tf15.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock4(normalized, params.dense0, true);
+      out = denseBlock4(out, params.dense1);
+      out = denseBlock4(out, params.dense2);
+      out = denseBlock4(out, params.dense3);
+      out = tf15.avgPool(out, [7, 7], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams(weights);
+  }
+};
+
+// src/faceProcessor/FaceProcessor.ts
+var tf17 = __toESM(require_tfjs_esm());
+
+// src/common/fullyConnectedLayer.ts
+var tf16 = __toESM(require_tfjs_esm());
+function fullyConnectedLayer(x, params) {
+  return tf16.tidy(() => tf16.add(
+    tf16.matMul(x, params.weights),
+    params.bias
+  ));
+}
+
+// src/faceProcessor/extractParams.ts
+function extractParams2(weights, channelsIn, channelsOut) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const fc = extractFCParams(channelsIn, channelsOut, "fc");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc }
+  };
+}
+
+// src/faceProcessor/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap2(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: extractFcParams("fc")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceProcessor/util.ts
+function seperateWeightMaps(weightMap) {
+  const featureExtractorMap = {};
+  const classifierMap = {};
+  Object.keys(weightMap).forEach((key) => {
+    const map = key.startsWith("fc") ? classifierMap : featureExtractorMap;
+    map[key] = weightMap[key];
+  });
+  return { featureExtractorMap, classifierMap };
+}
+
+// src/faceProcessor/FaceProcessor.ts
+var FaceProcessor = class extends NeuralNetwork {
+  constructor(_name, faceFeatureExtractor) {
+    super(_name);
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf17.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      return fullyConnectedLayer(bottleneckFeatures.as2D(bottleneckFeatures.shape[0], -1), params.fc);
+    });
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams2(weights, this.getClassifierChannelsIn(), this.getClassifierChannelsOut());
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap2(classifierMap);
+  }
+  extractParams(weights) {
+    const cIn = this.getClassifierChannelsIn();
+    const cOut = this.getClassifierChannelsOut();
+    const classifierWeightSize = cOut * cIn + cOut;
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressions.ts
+var FACE_EXPRESSION_LABELS = ["neutral", "happy", "sad", "angry", "fearful", "disgusted", "surprised"];
+var FaceExpressions = class {
+  constructor(probabilities) {
+    this.neutral = 0;
+    this.happy = 0;
+    this.sad = 0;
+    this.angry = 0;
+    this.fearful = 0;
+    this.disgusted = 0;
+    this.surprised = 0;
+    if (probabilities.length !== 7) {
+      throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${probabilities.length}`);
+    }
+    FACE_EXPRESSION_LABELS.forEach((expression, idx) => {
+      this[expression] = probabilities[idx];
+    });
+  }
+  asSortedArray() {
+    return FACE_EXPRESSION_LABELS.map((expression) => ({ expression, probability: this[expression] })).sort((e0, e1) => e1.probability - e0.probability);
+  }
+};
+
+// src/faceExpressionNet/FaceExpressionNet.ts
+var FaceExpressionNet = class extends FaceProcessor {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceExpressionNet", faceFeatureExtractor);
+  }
+  forwardInput(input) {
+    return tf18.tidy(() => tf18.softmax(this.runNet(input)));
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictExpressions(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const probabilitesByBatch = await Promise.all(tf18.unstack(out).map(async (t) => {
+      const data = t.dataSync();
+      t.dispose();
+      return data;
+    }));
+    out.dispose();
+    const predictionsByBatch = probabilitesByBatch.map((probabilites) => new FaceExpressions(probabilites));
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_expression_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+  getClassifierChannelsOut() {
+    return 7;
+  }
+};
+
+// src/factories/WithFaceExpressions.ts
+function isWithFaceExpressions(obj) {
+  return obj.expressions instanceof FaceExpressions;
+}
+function extendWithFaceExpressions(sourceObj, expressions) {
+  const extension = { expressions };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/drawFaceExpressions.ts
+function drawFaceExpressions(canvasArg, faceExpressions, minConfidence = 0.1, textFieldAnchor) {
+  const faceExpressionsArray = Array.isArray(faceExpressions) ? faceExpressions : [faceExpressions];
+  faceExpressionsArray.forEach((e) => {
+    const expr = e instanceof FaceExpressions ? e : isWithFaceExpressions(e) ? e.expressions : void 0;
+    if (!expr) {
+      throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");
+    }
+    const sorted = expr.asSortedArray();
+    const resultsToDisplay = sorted.filter((exprLocal) => exprLocal.probability > minConfidence);
+    const anchor = isWithFaceDetection(e) ? e.detection.box.bottomLeft : textFieldAnchor || new Point(0, 0);
+    const drawTextField = new DrawTextField(
+      resultsToDisplay.map((exprLocal) => `${exprLocal.expression} (${round(exprLocal.probability)})`),
+      anchor
+    );
+    drawTextField.draw(canvasArg);
+  });
+}
+
+// src/factories/WithFaceLandmarks.ts
+function isWithFaceLandmarks(obj) {
+  return isWithFaceDetection(obj) && obj["landmarks"] instanceof FaceLandmarks && obj["unshiftedLandmarks"] instanceof FaceLandmarks && obj["alignedRect"] instanceof FaceDetection;
+}
+function calculateFaceAngle(mesh) {
+  const degrees = (radians) => radians * 180 / Math.PI;
+  const calcLengthBetweenTwoPoints = (a, b) => Math.sqrt((a._x - b._x) ** 2 + (a._y - b._y) ** 2);
+  const angle = {
+    roll: void 0,
+    pitch: void 0,
+    yaw: void 0
+  };
+  const calcYaw = (leftPoint, midPoint, rightPoint) => {
+    const leftToMidpoint = Math.floor(leftPoint._x - midPoint._x);
+    const rightToMidpoint = Math.floor(midPoint._x - rightPoint._x);
+    return leftToMidpoint - rightToMidpoint;
+  };
+  const calcRoll = (lever, pivot) => {
+    const hypotenuse = Math.hypot(pivot._x - lever._x, pivot._y - lever._y);
+    const opposite = pivot._y - lever._y;
+    const angleInRadians = Math.asin(opposite / hypotenuse);
+    const angleInDegrees = degrees(angleInRadians);
+    const normalizeAngle = Math.floor(90 - angleInDegrees);
+    const tiltDirection = pivot._x - lever._x < 0 ? -1 : 1;
+    const result = normalizeAngle * tiltDirection;
+    return result;
+  };
+  const calcPitch = (leftPoint, midPoint, rightPoint) => {
+    const base = calcLengthBetweenTwoPoints(leftPoint, rightPoint);
+    const baseCoords = {
+      _x: (leftPoint._x + rightPoint._x) / 2,
+      _y: (leftPoint._y + rightPoint._y) / 2
+    };
+    const midToBaseLength = calcLengthBetweenTwoPoints(midPoint, baseCoords);
+    const angleInRadians = Math.atan(midToBaseLength / base);
+    const angleInDegrees = Math.floor(degrees(angleInRadians));
+    const direction = baseCoords._y - midPoint._y < 0 ? -1 : 1;
+    const result = angleInDegrees * direction;
+    return result;
+  };
+  if (!mesh || !mesh._positions || mesh._positions.length !== 68)
+    return angle;
+  const pt = mesh._positions;
+  angle.roll = calcRoll(pt[27], pt[66]);
+  angle.pitch = calcPitch(pt[14], pt[30], pt[2]);
+  angle.yaw = calcYaw(pt[14], pt[33], pt[2]);
+  return angle;
+}
+function extendWithFaceLandmarks(sourceObj, unshiftedLandmarks) {
+  const { box: shift } = sourceObj.detection;
+  const landmarks = unshiftedLandmarks.shiftBy(shift.x, shift.y);
+  const rect = landmarks.align();
+  const { imageDims } = sourceObj.detection;
+  const alignedRect = new FaceDetection(
+    sourceObj.detection.score,
+    rect.rescale(imageDims.reverse()),
+    imageDims
+  );
+  const angle = calculateFaceAngle(unshiftedLandmarks);
+  const extension = { landmarks, unshiftedLandmarks, alignedRect, angle };
+  return { ...sourceObj, ...extension };
+}
+
+// src/draw/DrawFaceLandmarks.ts
+var DrawFaceLandmarksOptions = class {
+  constructor(options = {}) {
+    const {
+      drawLines = true,
+      drawPoints = true,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = options;
+    this.drawLines = drawLines;
+    this.drawPoints = drawPoints;
+    this.lineWidth = lineWidth || 1;
+    this.pointSize = pointSize || 2;
+    this.lineColor = lineColor || "rgba(0, 255, 255, 1)";
+    this.pointColor = pointColor || "rgba(255, 0, 255, 1)";
+  }
+};
+var DrawFaceLandmarks = class {
+  constructor(faceLandmarks, options = {}) {
+    this.faceLandmarks = faceLandmarks;
+    this.options = new DrawFaceLandmarksOptions(options);
+  }
+  draw(canvasArg) {
+    const ctx = getContext2dOrThrow(canvasArg);
+    const {
+      drawLines,
+      drawPoints,
+      lineWidth,
+      lineColor,
+      pointSize,
+      pointColor
+    } = this.options;
+    if (drawLines && this.faceLandmarks instanceof FaceLandmarks68) {
+      ctx.strokeStyle = lineColor;
+      ctx.lineWidth = lineWidth;
+      drawContour(ctx, this.faceLandmarks.getJawOutline());
+      drawContour(ctx, this.faceLandmarks.getLeftEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getRightEyeBrow());
+      drawContour(ctx, this.faceLandmarks.getNose());
+      drawContour(ctx, this.faceLandmarks.getLeftEye(), true);
+      drawContour(ctx, this.faceLandmarks.getRightEye(), true);
+      drawContour(ctx, this.faceLandmarks.getMouth(), true);
+    }
+    if (drawPoints) {
+      ctx.strokeStyle = pointColor;
+      ctx.fillStyle = pointColor;
+      const drawPoint = (pt) => {
+        ctx.beginPath();
+        ctx.arc(pt.x, pt.y, pointSize, 0, 2 * Math.PI);
+        ctx.fill();
+      };
+      this.faceLandmarks.positions.forEach(drawPoint);
+    }
+  }
+};
+function drawFaceLandmarks(canvasArg, faceLandmarks) {
+  const faceLandmarksArray = Array.isArray(faceLandmarks) ? faceLandmarks : [faceLandmarks];
+  faceLandmarksArray.forEach((f) => {
+    const landmarks = f instanceof FaceLandmarks ? f : isWithFaceLandmarks(f) ? f.landmarks : void 0;
+    if (!landmarks) {
+      throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");
+    }
+    new DrawFaceLandmarks(landmarks).draw(canvasArg);
+  });
+}
+
+// package.json
+var version = "1.7.6";
+
+// src/ageGenderNet/AgeGenderNet.ts
+var tf20 = __toESM(require_tfjs_esm());
+
+// src/xception/TinyXception.ts
+var tf19 = __toESM(require_tfjs_esm());
+
+// src/xception/extractParams.ts
+function extractorsFactory2(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  function extractReductionBlockParams(channelsIn, channelsOut, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channelsIn, channelsOut, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channelsOut, channelsOut, `${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(channels, mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(channels, channels, `${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParams3(weights, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = extractorsFactory2(extractWeights, paramMappings);
+  const entry_flow_conv_in = extractConvParams(3, 32, 3, "entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams(32, 64, "entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams(64, 128, "entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(128, `middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams(128, 256, "exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams(256, 512, "exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { entry_flow, middle_flow, exit_flow }
+  };
+}
+
+// src/xception/extractParamsFromWeightMap.ts
+function loadParamsFactory2(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  const extractConvParams = loadConvParamsFactory(extractWeightEntry);
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  function extractReductionBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const expansion_conv = extractConvParams(`${mappedPrefix}/expansion_conv`);
+    return { separable_conv0, separable_conv1, expansion_conv };
+  }
+  function extractMainBlockParams(mappedPrefix) {
+    const separable_conv0 = extractSeparableConvParams(`${mappedPrefix}/separable_conv0`);
+    const separable_conv1 = extractSeparableConvParams(`${mappedPrefix}/separable_conv1`);
+    const separable_conv2 = extractSeparableConvParams(`${mappedPrefix}/separable_conv2`);
+    return { separable_conv0, separable_conv1, separable_conv2 };
+  }
+  return {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  };
+}
+function extractParamsFromWeightMap3(weightMap, numMainBlocks) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractSeparableConvParams,
+    extractReductionBlockParams,
+    extractMainBlockParams
+  } = loadParamsFactory2(weightMap, paramMappings);
+  const entry_flow_conv_in = extractConvParams("entry_flow/conv_in");
+  const entry_flow_reduction_block_0 = extractReductionBlockParams("entry_flow/reduction_block_0");
+  const entry_flow_reduction_block_1 = extractReductionBlockParams("entry_flow/reduction_block_1");
+  const entry_flow = {
+    conv_in: entry_flow_conv_in,
+    reduction_block_0: entry_flow_reduction_block_0,
+    reduction_block_1: entry_flow_reduction_block_1
+  };
+  const middle_flow = {};
+  range(numMainBlocks, 0, 1).forEach((idx) => {
+    middle_flow[`main_block_${idx}`] = extractMainBlockParams(`middle_flow/main_block_${idx}`);
+  });
+  const exit_flow_reduction_block = extractReductionBlockParams("exit_flow/reduction_block");
+  const exit_flow_separable_conv = extractSeparableConvParams("exit_flow/separable_conv");
+  const exit_flow = {
+    reduction_block: exit_flow_reduction_block,
+    separable_conv: exit_flow_separable_conv
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params: { entry_flow, middle_flow, exit_flow }, paramMappings };
+}
+
+// src/xception/TinyXception.ts
+function conv(x, params, stride) {
+  return tf19.add(tf19.conv2d(x, params.filters, stride, "same"), params.bias);
+}
+function reductionBlock(x, params, isActivateInput = true) {
+  let out = isActivateInput ? tf19.relu(x) : x;
+  out = depthwiseSeparableConv(out, params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = tf19.maxPool(out, [3, 3], [2, 2], "same");
+  out = tf19.add(out, conv(x, params.expansion_conv, [2, 2]));
+  return out;
+}
+function mainBlock(x, params) {
+  let out = depthwiseSeparableConv(tf19.relu(x), params.separable_conv0, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv1, [1, 1]);
+  out = depthwiseSeparableConv(tf19.relu(out), params.separable_conv2, [1, 1]);
+  out = tf19.add(out, x);
+  return out;
+}
+var TinyXception = class extends NeuralNetwork {
+  constructor(numMainBlocks) {
+    super("TinyXception");
+    this._numMainBlocks = numMainBlocks;
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyXception - load model before inference");
+    }
+    return tf19.tidy(() => {
+      const batchTensor = tf19.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = tf19.relu(conv(normalized, params.entry_flow.conv_in, [2, 2]));
+      out = reductionBlock(out, params.entry_flow.reduction_block_0, false);
+      out = reductionBlock(out, params.entry_flow.reduction_block_1);
+      range(this._numMainBlocks, 0, 1).forEach((idx) => {
+        out = mainBlock(out, params.middle_flow[`main_block_${idx}`]);
+      });
+      out = reductionBlock(out, params.exit_flow.reduction_block);
+      out = tf19.relu(depthwiseSeparableConv(out, params.exit_flow.separable_conv, [1, 1]));
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "tiny_xception_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap3(weightMap, this._numMainBlocks);
+  }
+  extractParams(weights) {
+    return extractParams3(weights, this._numMainBlocks);
+  }
+};
+
+// src/ageGenderNet/extractParams.ts
+function extractParams4(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const extractFCParams = extractFCParamsFactory(extractWeights, paramMappings);
+  const age = extractFCParams(512, 1, "fc/age");
+  const gender = extractFCParams(512, 2, "fc/gender");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { fc: { age, gender } }
+  };
+}
+
+// src/ageGenderNet/extractParamsFromWeightMap.ts
+function extractParamsFromWeightMap4(weightMap) {
+  const paramMappings = [];
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractFcParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/weights`, 2);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { weights, bias };
+  }
+  const params = {
+    fc: {
+      age: extractFcParams("fc/age"),
+      gender: extractFcParams("fc/gender")
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ageGenderNet/types.ts
+var Gender = /* @__PURE__ */ ((Gender2) => {
+  Gender2["FEMALE"] = "female";
+  Gender2["MALE"] = "male";
+  return Gender2;
+})(Gender || {});
+
+// src/ageGenderNet/AgeGenderNet.ts
+var AgeGenderNet = class extends NeuralNetwork {
+  constructor(faceFeatureExtractor = new TinyXception(2)) {
+    super("AgeGenderNet");
+    this._faceFeatureExtractor = faceFeatureExtractor;
+  }
+  get faceFeatureExtractor() {
+    return this._faceFeatureExtractor;
+  }
+  runNet(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error(`${this._name} - load model before inference`);
+    }
+    return tf20.tidy(() => {
+      const bottleneckFeatures = input instanceof NetInput ? this.faceFeatureExtractor.forwardInput(input) : input;
+      const pooled = tf20.avgPool(bottleneckFeatures, [7, 7], [2, 2], "valid").as2D(bottleneckFeatures.shape[0], -1);
+      const age = fullyConnectedLayer(pooled, params.fc.age).as1D();
+      const gender = fullyConnectedLayer(pooled, params.fc.gender);
+      return { age, gender };
+    });
+  }
+  forwardInput(input) {
+    return tf20.tidy(() => {
+      const { age, gender } = this.runNet(input);
+      return { age, gender: tf20.softmax(gender) };
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async predictAgeAndGender(input) {
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput);
+    const ages = tf20.unstack(out.age);
+    const genders = tf20.unstack(out.gender);
+    const ageAndGenderTensors = ages.map((ageTensor, i) => ({
+      ageTensor,
+      genderTensor: genders[i]
+    }));
+    const predictionsByBatch = await Promise.all(
+      ageAndGenderTensors.map(async ({ ageTensor, genderTensor }) => {
+        const age = ageTensor.dataSync()[0];
+        const probMale = genderTensor.dataSync()[0];
+        const isMale = probMale > 0.5;
+        const gender = isMale ? "male" /* MALE */ : "female" /* FEMALE */;
+        const genderProbability = isMale ? probMale : 1 - probMale;
+        ageTensor.dispose();
+        genderTensor.dispose();
+        return { age, gender, genderProbability };
+      })
+    );
+    out.age.dispose();
+    out.gender.dispose();
+    return netInput.isBatchInput ? predictionsByBatch : predictionsByBatch[0];
+  }
+  getDefaultModelName() {
+    return "age_gender_model";
+  }
+  dispose(throwOnRedispose = true) {
+    this.faceFeatureExtractor.dispose(throwOnRedispose);
+    super.dispose(throwOnRedispose);
+  }
+  loadClassifierParams(weights) {
+    const { params, paramMappings } = this.extractClassifierParams(weights);
+    this._params = params;
+    this._paramMappings = paramMappings;
+  }
+  extractClassifierParams(weights) {
+    return extractParams4(weights);
+  }
+  extractParamsFromWeightMap(weightMap) {
+    const { featureExtractorMap, classifierMap } = seperateWeightMaps(weightMap);
+    this.faceFeatureExtractor.loadFromWeightMap(featureExtractorMap);
+    return extractParamsFromWeightMap4(classifierMap);
+  }
+  extractParams(weights) {
+    const classifierWeightSize = 512 * 1 + 1 + (512 * 2 + 2);
+    const featureExtractorWeights = weights.slice(0, weights.length - classifierWeightSize);
+    const classifierWeights = weights.slice(weights.length - classifierWeightSize);
+    this.faceFeatureExtractor.extractWeights(featureExtractorWeights);
+    return this.extractClassifierParams(classifierWeights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68NetBase.ts
+var tf21 = __toESM(require_tfjs_esm());
+var FaceLandmark68NetBase = class extends FaceProcessor {
+  postProcess(output, inputSize, originalDimensions) {
+    const inputDimensions = originalDimensions.map(({ width, height }) => {
+      const scale2 = inputSize / Math.max(height, width);
+      return {
+        width: width * scale2,
+        height: height * scale2
+      };
+    });
+    const batchSize = inputDimensions.length;
+    return tf21.tidy(() => {
+      const createInterleavedTensor = (fillX, fillY) => tf21.stack([tf21.fill([68], fillX, "float32"), tf21.fill([68], fillY, "float32")], 1).as2D(1, 136).as1D();
+      const getPadding = (batchIdx, cond) => {
+        const { width, height } = inputDimensions[batchIdx];
+        return cond(width, height) ? Math.abs(width - height) / 2 : 0;
+      };
+      const getPaddingX = (batchIdx) => getPadding(batchIdx, (w, h) => w < h);
+      const getPaddingY = (batchIdx) => getPadding(batchIdx, (w, h) => h < w);
+      const landmarkTensors = output.mul(tf21.fill([batchSize, 136], inputSize, "float32")).sub(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        getPaddingX(batchIdx),
+        getPaddingY(batchIdx)
+      )))).div(tf21.stack(Array.from(Array(batchSize), (_, batchIdx) => createInterleavedTensor(
+        inputDimensions[batchIdx].width,
+        inputDimensions[batchIdx].height
+      ))));
+      return landmarkTensors;
+    });
+  }
+  forwardInput(input) {
+    return tf21.tidy(() => {
+      const out = this.runNet(input);
+      return this.postProcess(
+        out,
+        input.inputSize,
+        input.inputDimensions.map(([height, width]) => ({ height, width }))
+      );
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async detectLandmarks(input) {
+    const netInput = await toNetInput(input);
+    const landmarkTensors = tf21.tidy(
+      () => tf21.unstack(this.forwardInput(netInput))
+    );
+    const landmarksForBatch = await Promise.all(landmarkTensors.map(
+      async (landmarkTensor, batchIdx) => {
+        const landmarksArray = Array.from(landmarkTensor.dataSync());
+        const xCoords = landmarksArray.filter((_, i) => isEven(i));
+        const yCoords = landmarksArray.filter((_, i) => !isEven(i));
+        return new FaceLandmarks68(
+          Array(68).fill(0).map((_, i) => new Point(xCoords[i], yCoords[i])),
+          {
+            height: netInput.getInputHeight(batchIdx),
+            width: netInput.getInputWidth(batchIdx)
+          }
+        );
+      }
+    ));
+    landmarkTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? landmarksForBatch : landmarksForBatch[0];
+  }
+  getClassifierChannelsOut() {
+    return 136;
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68Net.ts
+var FaceLandmark68Net = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new FaceFeatureExtractor()) {
+    super("FaceLandmark68Net", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_model";
+  }
+  getClassifierChannelsIn() {
+    return 256;
+  }
+};
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var tf22 = __toESM(require_tfjs_esm());
+
+// src/faceFeatureExtractor/extractParamsFromWeightMapTiny.ts
+function extractParamsFromWeightMapTiny(weightMap) {
+  const paramMappings = [];
+  const {
+    extractDenseBlock3Params
+  } = loadParamsFactory(weightMap, paramMappings);
+  const params = {
+    dense0: extractDenseBlock3Params("dense0", true),
+    dense1: extractDenseBlock3Params("dense1"),
+    dense2: extractDenseBlock3Params("dense2")
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceFeatureExtractor/extractParamsTiny.ts
+function extractParamsTiny(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractDenseBlock3Params
+  } = extractorsFactory(extractWeights, paramMappings);
+  const dense0 = extractDenseBlock3Params(3, 32, "dense0", true);
+  const dense1 = extractDenseBlock3Params(32, 64, "dense1");
+  const dense2 = extractDenseBlock3Params(64, 128, "dense2");
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    paramMappings,
+    params: { dense0, dense1, dense2 }
+  };
+}
+
+// src/faceFeatureExtractor/TinyFaceFeatureExtractor.ts
+var TinyFaceFeatureExtractor = class extends NeuralNetwork {
+  constructor() {
+    super("TinyFaceFeatureExtractor");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyFaceFeatureExtractor - load model before inference");
+    }
+    return tf22.tidy(() => {
+      const batchTensor = tf22.cast(input.toBatchTensor(112, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = denseBlock3(normalized, params.dense0, true);
+      out = denseBlock3(out, params.dense1);
+      out = denseBlock3(out, params.dense2);
+      out = tf22.avgPool(out, [14, 14], [2, 2], "valid");
+      return out;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  getDefaultModelName() {
+    return "face_feature_extractor_tiny_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMapTiny(weightMap);
+  }
+  extractParams(weights) {
+    return extractParamsTiny(weights);
+  }
+};
+
+// src/faceLandmarkNet/FaceLandmark68TinyNet.ts
+var FaceLandmark68TinyNet = class extends FaceLandmark68NetBase {
+  constructor(faceFeatureExtractor = new TinyFaceFeatureExtractor()) {
+    super("FaceLandmark68TinyNet", faceFeatureExtractor);
+  }
+  getDefaultModelName() {
+    return "face_landmark_68_tiny_model";
+  }
+  getClassifierChannelsIn() {
+    return 128;
+  }
+};
+
+// src/faceLandmarkNet/index.ts
+var FaceLandmarkNet = class extends FaceLandmark68Net {
+};
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var tf27 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/convLayer.ts
+var tf24 = __toESM(require_tfjs_esm());
+
+// src/faceRecognitionNet/scaleLayer.ts
+var tf23 = __toESM(require_tfjs_esm());
+function scale(x, params) {
+  return tf23.add(tf23.mul(x, params.weights), params.biases);
+}
+
+// src/faceRecognitionNet/convLayer.ts
+function convLayer2(x, params, strides, withRelu, padding = "same") {
+  const { filters, bias } = params.conv;
+  let out = tf24.conv2d(x, filters, strides, padding);
+  out = tf24.add(out, bias);
+  out = scale(out, params.scale);
+  return withRelu ? tf24.relu(out) : out;
+}
+function conv2(x, params) {
+  return convLayer2(x, params, [1, 1], true);
+}
+function convNoRelu(x, params) {
+  return convLayer2(x, params, [1, 1], false);
+}
+function convDown(x, params) {
+  return convLayer2(x, params, [2, 2], true, "valid");
+}
+
+// src/faceRecognitionNet/extractParams.ts
+var tf25 = __toESM(require_tfjs_esm());
+function extractorsFactory3(extractWeights, paramMappings) {
+  function extractFilterValues(numFilterValues, numFilters, filterSize) {
+    const weights = extractWeights(numFilterValues);
+    const depth = weights.length / (numFilters * filterSize * filterSize);
+    if (isFloat(depth)) {
+      throw new Error(`depth has to be an integer: ${depth}, weights.length: ${weights.length}, numFilters: ${numFilters}, filterSize: ${filterSize}`);
+    }
+    return tf25.tidy(
+      () => tf25.transpose(
+        tf25.tensor4d(weights, [numFilters, depth, filterSize, filterSize]),
+        [2, 3, 1, 0]
+      )
+    );
+  }
+  function extractConvParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const filters = extractFilterValues(numFilterValues, numFilters, filterSize);
+    const bias = tf25.tensor1d(extractWeights(numFilters));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/bias` }
+    );
+    return { filters, bias };
+  }
+  function extractScaleLayerParams(numWeights, mappedPrefix) {
+    const weights = tf25.tensor1d(extractWeights(numWeights));
+    const biases = tf25.tensor1d(extractWeights(numWeights));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/weights` },
+      { paramPath: `${mappedPrefix}/biases` }
+    );
+    return {
+      weights,
+      biases
+    };
+  }
+  function extractConvLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix) {
+    const conv3 = extractConvParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv`);
+    const scale2 = extractScaleLayerParams(numFilters, `${mappedPrefix}/scale`);
+    return { conv: conv3, scale: scale2 };
+  }
+  function extractResidualLayerParams(numFilterValues, numFilters, filterSize, mappedPrefix, isDown = false) {
+    const conv1 = extractConvLayerParams((isDown ? 0.5 : 1) * numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv1`);
+    const conv22 = extractConvLayerParams(numFilterValues, numFilters, filterSize, `${mappedPrefix}/conv2`);
+    return { conv1, conv2: conv22 };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParams5(weights) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory3(extractWeights, paramMappings);
+  const conv32_down = extractConvLayerParams(4704, 32, 7, "conv32_down");
+  const conv32_1 = extractResidualLayerParams(9216, 32, 3, "conv32_1");
+  const conv32_2 = extractResidualLayerParams(9216, 32, 3, "conv32_2");
+  const conv32_3 = extractResidualLayerParams(9216, 32, 3, "conv32_3");
+  const conv64_down = extractResidualLayerParams(36864, 64, 3, "conv64_down", true);
+  const conv64_1 = extractResidualLayerParams(36864, 64, 3, "conv64_1");
+  const conv64_2 = extractResidualLayerParams(36864, 64, 3, "conv64_2");
+  const conv64_3 = extractResidualLayerParams(36864, 64, 3, "conv64_3");
+  const conv128_down = extractResidualLayerParams(147456, 128, 3, "conv128_down", true);
+  const conv128_1 = extractResidualLayerParams(147456, 128, 3, "conv128_1");
+  const conv128_2 = extractResidualLayerParams(147456, 128, 3, "conv128_2");
+  const conv256_down = extractResidualLayerParams(589824, 256, 3, "conv256_down", true);
+  const conv256_1 = extractResidualLayerParams(589824, 256, 3, "conv256_1");
+  const conv256_2 = extractResidualLayerParams(589824, 256, 3, "conv256_2");
+  const conv256_down_out = extractResidualLayerParams(589824, 256, 3, "conv256_down_out");
+  const fc = tf25.tidy(
+    () => tf25.transpose(tf25.tensor2d(extractWeights(256 * 128), [128, 256]), [1, 0])
+  );
+  paramMappings.push({ paramPath: "fc" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/extractParamsFromWeightMap.ts
+function extractorsFactory4(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractScaleLayerParams(prefix) {
+    const weights = extractWeightEntry(`${prefix}/scale/weights`, 1);
+    const biases = extractWeightEntry(`${prefix}/scale/biases`, 1);
+    return { weights, biases };
+  }
+  function extractConvLayerParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/conv/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/conv/bias`, 1);
+    const scale2 = extractScaleLayerParams(prefix);
+    return { conv: { filters, bias }, scale: scale2 };
+  }
+  function extractResidualLayerParams(prefix) {
+    return {
+      conv1: extractConvLayerParams(`${prefix}/conv1`),
+      conv2: extractConvLayerParams(`${prefix}/conv2`)
+    };
+  }
+  return {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  };
+}
+function extractParamsFromWeightMap5(weightMap) {
+  const paramMappings = [];
+  const {
+    extractConvLayerParams,
+    extractResidualLayerParams
+  } = extractorsFactory4(weightMap, paramMappings);
+  const conv32_down = extractConvLayerParams("conv32_down");
+  const conv32_1 = extractResidualLayerParams("conv32_1");
+  const conv32_2 = extractResidualLayerParams("conv32_2");
+  const conv32_3 = extractResidualLayerParams("conv32_3");
+  const conv64_down = extractResidualLayerParams("conv64_down");
+  const conv64_1 = extractResidualLayerParams("conv64_1");
+  const conv64_2 = extractResidualLayerParams("conv64_2");
+  const conv64_3 = extractResidualLayerParams("conv64_3");
+  const conv128_down = extractResidualLayerParams("conv128_down");
+  const conv128_1 = extractResidualLayerParams("conv128_1");
+  const conv128_2 = extractResidualLayerParams("conv128_2");
+  const conv256_down = extractResidualLayerParams("conv256_down");
+  const conv256_1 = extractResidualLayerParams("conv256_1");
+  const conv256_2 = extractResidualLayerParams("conv256_2");
+  const conv256_down_out = extractResidualLayerParams("conv256_down_out");
+  const { fc } = weightMap;
+  paramMappings.push({ originalPath: "fc", paramPath: "fc" });
+  if (!isTensor2D(fc)) {
+    throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${fc}`);
+  }
+  const params = {
+    conv32_down,
+    conv32_1,
+    conv32_2,
+    conv32_3,
+    conv64_down,
+    conv64_1,
+    conv64_2,
+    conv64_3,
+    conv128_down,
+    conv128_1,
+    conv128_2,
+    conv256_down,
+    conv256_1,
+    conv256_2,
+    conv256_down_out,
+    fc
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/faceRecognitionNet/residualLayer.ts
+var tf26 = __toESM(require_tfjs_esm());
+function residual(x, params) {
+  let out = conv2(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  out = tf26.add(out, x);
+  out = tf26.relu(out);
+  return out;
+}
+function residualDown(x, params) {
+  let out = convDown(x, params.conv1);
+  out = convNoRelu(out, params.conv2);
+  let pooled = tf26.avgPool(x, 2, 2, "valid");
+  const zeros2 = tf26.zeros(pooled.shape);
+  const isPad = pooled.shape[3] !== out.shape[3];
+  const isAdjustShape = pooled.shape[1] !== out.shape[1] || pooled.shape[2] !== out.shape[2];
+  if (isAdjustShape) {
+    const padShapeX = [...out.shape];
+    padShapeX[1] = 1;
+    const zerosW = tf26.zeros(padShapeX);
+    out = tf26.concat([out, zerosW], 1);
+    const padShapeY = [...out.shape];
+    padShapeY[2] = 1;
+    const zerosH = tf26.zeros(padShapeY);
+    out = tf26.concat([out, zerosH], 2);
+  }
+  pooled = isPad ? tf26.concat([pooled, zeros2], 3) : pooled;
+  out = tf26.add(pooled, out);
+  out = tf26.relu(out);
+  return out;
+}
+
+// src/faceRecognitionNet/FaceRecognitionNet.ts
+var FaceRecognitionNet = class extends NeuralNetwork {
+  constructor() {
+    super("FaceRecognitionNet");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("FaceRecognitionNet - load model before inference");
+    }
+    return tf27.tidy(() => {
+      const batchTensor = tf27.cast(input.toBatchTensor(150, true), "float32");
+      const meanRgb = [122.782, 117.001, 104.298];
+      const normalized = normalize(batchTensor, meanRgb).div(255);
+      let out = convDown(normalized, params.conv32_down);
+      out = tf27.maxPool(out, 3, 2, "valid");
+      out = residual(out, params.conv32_1);
+      out = residual(out, params.conv32_2);
+      out = residual(out, params.conv32_3);
+      out = residualDown(out, params.conv64_down);
+      out = residual(out, params.conv64_1);
+      out = residual(out, params.conv64_2);
+      out = residual(out, params.conv64_3);
+      out = residualDown(out, params.conv128_down);
+      out = residual(out, params.conv128_1);
+      out = residual(out, params.conv128_2);
+      out = residualDown(out, params.conv256_down);
+      out = residual(out, params.conv256_1);
+      out = residual(out, params.conv256_2);
+      out = residualDown(out, params.conv256_down_out);
+      const globalAvg = out.mean([1, 2]);
+      const fullyConnected = tf27.matMul(globalAvg, params.fc);
+      return fullyConnected;
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async computeFaceDescriptor(input) {
+    var _a;
+    if ((_a = input == null ? void 0 : input.shape) == null ? void 0 : _a.some((dim) => dim <= 0))
+      return new Float32Array(128);
+    const netInput = await toNetInput(input);
+    const faceDescriptorTensors = tf27.tidy(() => tf27.unstack(this.forwardInput(netInput)));
+    const faceDescriptorsForBatch = await Promise.all(faceDescriptorTensors.map((t) => t.data()));
+    faceDescriptorTensors.forEach((t) => t.dispose());
+    return netInput.isBatchInput ? faceDescriptorsForBatch : faceDescriptorsForBatch[0];
+  }
+  getDefaultModelName() {
+    return "face_recognition_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap5(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams5(weights);
+  }
+};
+
+// src/faceRecognitionNet/index.ts
+function createFaceRecognitionNet(weights) {
+  const net = new FaceRecognitionNet();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/factories/WithFaceDescriptor.ts
+function extendWithFaceDescriptor(sourceObj, descriptor) {
+  const extension = { descriptor };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithAge.ts
+function isWithAge(obj) {
+  return typeof obj.age === "number";
+}
+function extendWithAge(sourceObj, age) {
+  const extension = { age };
+  return { ...sourceObj, ...extension };
+}
+
+// src/factories/WithGender.ts
+function isWithGender(obj) {
+  return (obj.gender === "male" /* MALE */ || obj.gender === "female" /* FEMALE */) && isValidProbablitiy(obj.genderProbability);
+}
+function extendWithGender(sourceObj, gender, genderProbability) {
+  const extension = { gender, genderProbability };
+  return { ...sourceObj, ...extension };
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var tf34 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/extractParams.ts
+var tf28 = __toESM(require_tfjs_esm());
+function extractorsFactory5(extractWeights, paramMappings) {
+  function extractDepthwiseConvParams(numChannels, mappedPrefix) {
+    const filters = tf28.tensor4d(extractWeights(3 * 3 * numChannels), [3, 3, numChannels, 1]);
+    const batch_norm_scale = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_offset = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_mean = tf28.tensor1d(extractWeights(numChannels));
+    const batch_norm_variance = tf28.tensor1d(extractWeights(numChannels));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/batch_norm_scale` },
+      { paramPath: `${mappedPrefix}/batch_norm_offset` },
+      { paramPath: `${mappedPrefix}/batch_norm_mean` },
+      { paramPath: `${mappedPrefix}/batch_norm_variance` }
+    );
+    return {
+      filters,
+      batch_norm_scale,
+      batch_norm_offset,
+      batch_norm_mean,
+      batch_norm_variance
+    };
+  }
+  function extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, isPointwiseConv) {
+    const filters = tf28.tensor4d(
+      extractWeights(channelsIn * channelsOut * filterSize * filterSize),
+      [filterSize, filterSize, channelsIn, channelsOut]
+    );
+    const bias = tf28.tensor1d(extractWeights(channelsOut));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/filters` },
+      { paramPath: `${mappedPrefix}/${isPointwiseConv ? "batch_norm_offset" : "bias"}` }
+    );
+    return { filters, bias };
+  }
+  function extractPointwiseConvParams(channelsIn, channelsOut, filterSize, mappedPrefix) {
+    const {
+      filters,
+      bias
+    } = extractConvParams(channelsIn, channelsOut, filterSize, mappedPrefix, true);
+    return {
+      filters,
+      batch_norm_offset: bias
+    };
+  }
+  function extractConvPairParams(channelsIn, channelsOut, mappedPrefix) {
+    const depthwise_conv = extractDepthwiseConvParams(channelsIn, `${mappedPrefix}/depthwise_conv`);
+    const pointwise_conv = extractPointwiseConvParams(channelsIn, channelsOut, 1, `${mappedPrefix}/pointwise_conv`);
+    return { depthwise_conv, pointwise_conv };
+  }
+  function extractMobilenetV1Params() {
+    const conv_0 = extractPointwiseConvParams(3, 32, 3, "mobilenetv1/conv_0");
+    const conv_1 = extractConvPairParams(32, 64, "mobilenetv1/conv_1");
+    const conv_2 = extractConvPairParams(64, 128, "mobilenetv1/conv_2");
+    const conv_3 = extractConvPairParams(128, 128, "mobilenetv1/conv_3");
+    const conv_4 = extractConvPairParams(128, 256, "mobilenetv1/conv_4");
+    const conv_5 = extractConvPairParams(256, 256, "mobilenetv1/conv_5");
+    const conv_6 = extractConvPairParams(256, 512, "mobilenetv1/conv_6");
+    const conv_7 = extractConvPairParams(512, 512, "mobilenetv1/conv_7");
+    const conv_8 = extractConvPairParams(512, 512, "mobilenetv1/conv_8");
+    const conv_9 = extractConvPairParams(512, 512, "mobilenetv1/conv_9");
+    const conv_10 = extractConvPairParams(512, 512, "mobilenetv1/conv_10");
+    const conv_11 = extractConvPairParams(512, 512, "mobilenetv1/conv_11");
+    const conv_12 = extractConvPairParams(512, 1024, "mobilenetv1/conv_12");
+    const conv_13 = extractConvPairParams(1024, 1024, "mobilenetv1/conv_13");
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      conv_8,
+      conv_9,
+      conv_10,
+      conv_11,
+      conv_12,
+      conv_13
+    };
+  }
+  function extractPredictionLayerParams() {
+    const conv_0 = extractPointwiseConvParams(1024, 256, 1, "prediction_layer/conv_0");
+    const conv_1 = extractPointwiseConvParams(256, 512, 3, "prediction_layer/conv_1");
+    const conv_2 = extractPointwiseConvParams(512, 128, 1, "prediction_layer/conv_2");
+    const conv_3 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_3");
+    const conv_4 = extractPointwiseConvParams(256, 128, 1, "prediction_layer/conv_4");
+    const conv_5 = extractPointwiseConvParams(128, 256, 3, "prediction_layer/conv_5");
+    const conv_6 = extractPointwiseConvParams(256, 64, 1, "prediction_layer/conv_6");
+    const conv_7 = extractPointwiseConvParams(64, 128, 3, "prediction_layer/conv_7");
+    const box_encoding_0_predictor = extractConvParams(512, 12, 1, "prediction_layer/box_predictor_0/box_encoding_predictor");
+    const class_predictor_0 = extractConvParams(512, 9, 1, "prediction_layer/box_predictor_0/class_predictor");
+    const box_encoding_1_predictor = extractConvParams(1024, 24, 1, "prediction_layer/box_predictor_1/box_encoding_predictor");
+    const class_predictor_1 = extractConvParams(1024, 18, 1, "prediction_layer/box_predictor_1/class_predictor");
+    const box_encoding_2_predictor = extractConvParams(512, 24, 1, "prediction_layer/box_predictor_2/box_encoding_predictor");
+    const class_predictor_2 = extractConvParams(512, 18, 1, "prediction_layer/box_predictor_2/class_predictor");
+    const box_encoding_3_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_3/box_encoding_predictor");
+    const class_predictor_3 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_3/class_predictor");
+    const box_encoding_4_predictor = extractConvParams(256, 24, 1, "prediction_layer/box_predictor_4/box_encoding_predictor");
+    const class_predictor_4 = extractConvParams(256, 18, 1, "prediction_layer/box_predictor_4/class_predictor");
+    const box_encoding_5_predictor = extractConvParams(128, 24, 1, "prediction_layer/box_predictor_5/box_encoding_predictor");
+    const class_predictor_5 = extractConvParams(128, 18, 1, "prediction_layer/box_predictor_5/class_predictor");
+    const box_predictor_0 = {
+      box_encoding_predictor: box_encoding_0_predictor,
+      class_predictor: class_predictor_0
+    };
+    const box_predictor_1 = {
+      box_encoding_predictor: box_encoding_1_predictor,
+      class_predictor: class_predictor_1
+    };
+    const box_predictor_2 = {
+      box_encoding_predictor: box_encoding_2_predictor,
+      class_predictor: class_predictor_2
+    };
+    const box_predictor_3 = {
+      box_encoding_predictor: box_encoding_3_predictor,
+      class_predictor: class_predictor_3
+    };
+    const box_predictor_4 = {
+      box_encoding_predictor: box_encoding_4_predictor,
+      class_predictor: class_predictor_4
+    };
+    const box_predictor_5 = {
+      box_encoding_predictor: box_encoding_5_predictor,
+      class_predictor: class_predictor_5
+    };
+    return {
+      conv_0,
+      conv_1,
+      conv_2,
+      conv_3,
+      conv_4,
+      conv_5,
+      conv_6,
+      conv_7,
+      box_predictor_0,
+      box_predictor_1,
+      box_predictor_2,
+      box_predictor_3,
+      box_predictor_4,
+      box_predictor_5
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParams6(weights) {
+  const paramMappings = [];
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory5(extractWeights, paramMappings);
+  const mobilenetv1 = extractMobilenetV1Params();
+  const prediction_layer = extractPredictionLayerParams();
+  const extra_dim = tf28.tensor3d(
+    extractWeights(5118 * 4),
+    [1, 5118, 4]
+  );
+  const output_layer = {
+    extra_dim
+  };
+  paramMappings.push({ paramPath: "output_layer/extra_dim" });
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return {
+    params: {
+      mobilenetv1,
+      prediction_layer,
+      output_layer
+    },
+    paramMappings
+  };
+}
+
+// src/ssdMobilenetv1/extractParamsFromWeightMap.ts
+function extractorsFactory6(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractPointwiseConvParams(prefix, idx, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/weights`, 4, `${mappedPrefix}/filters`);
+    const batch_norm_offset = extractWeightEntry(`${prefix}/Conv2d_${idx}_pointwise/convolution_bn_offset`, 1, `${mappedPrefix}/batch_norm_offset`);
+    return { filters, batch_norm_offset };
+  }
+  function extractConvPairParams(idx) {
+    const mappedPrefix = `mobilenetv1/conv_${idx}`;
+    const prefixDepthwiseConv = `MobilenetV1/Conv2d_${idx}_depthwise`;
+    const mappedPrefixDepthwiseConv = `${mappedPrefix}/depthwise_conv`;
+    const mappedPrefixPointwiseConv = `${mappedPrefix}/pointwise_conv`;
+    const filters = extractWeightEntry(`${prefixDepthwiseConv}/depthwise_weights`, 4, `${mappedPrefixDepthwiseConv}/filters`);
+    const batch_norm_scale = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/gamma`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_scale`);
+    const batch_norm_offset = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/beta`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_offset`);
+    const batch_norm_mean = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_mean`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_mean`);
+    const batch_norm_variance = extractWeightEntry(`${prefixDepthwiseConv}/BatchNorm/moving_variance`, 1, `${mappedPrefixDepthwiseConv}/batch_norm_variance`);
+    return {
+      depthwise_conv: {
+        filters,
+        batch_norm_scale,
+        batch_norm_offset,
+        batch_norm_mean,
+        batch_norm_variance
+      },
+      pointwise_conv: extractPointwiseConvParams("MobilenetV1", idx, mappedPrefixPointwiseConv)
+    };
+  }
+  function extractMobilenetV1Params() {
+    return {
+      conv_0: extractPointwiseConvParams("MobilenetV1", 0, "mobilenetv1/conv_0"),
+      conv_1: extractConvPairParams(1),
+      conv_2: extractConvPairParams(2),
+      conv_3: extractConvPairParams(3),
+      conv_4: extractConvPairParams(4),
+      conv_5: extractConvPairParams(5),
+      conv_6: extractConvPairParams(6),
+      conv_7: extractConvPairParams(7),
+      conv_8: extractConvPairParams(8),
+      conv_9: extractConvPairParams(9),
+      conv_10: extractConvPairParams(10),
+      conv_11: extractConvPairParams(11),
+      conv_12: extractConvPairParams(12),
+      conv_13: extractConvPairParams(13)
+    };
+  }
+  function extractConvParams(prefix, mappedPrefix) {
+    const filters = extractWeightEntry(`${prefix}/weights`, 4, `${mappedPrefix}/filters`);
+    const bias = extractWeightEntry(`${prefix}/biases`, 1, `${mappedPrefix}/bias`);
+    return { filters, bias };
+  }
+  function extractBoxPredictorParams(idx) {
+    const box_encoding_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/BoxEncodingPredictor`,
+      `prediction_layer/box_predictor_${idx}/box_encoding_predictor`
+    );
+    const class_predictor = extractConvParams(
+      `Prediction/BoxPredictor_${idx}/ClassPredictor`,
+      `prediction_layer/box_predictor_${idx}/class_predictor`
+    );
+    return { box_encoding_predictor, class_predictor };
+  }
+  function extractPredictionLayerParams() {
+    return {
+      conv_0: extractPointwiseConvParams("Prediction", 0, "prediction_layer/conv_0"),
+      conv_1: extractPointwiseConvParams("Prediction", 1, "prediction_layer/conv_1"),
+      conv_2: extractPointwiseConvParams("Prediction", 2, "prediction_layer/conv_2"),
+      conv_3: extractPointwiseConvParams("Prediction", 3, "prediction_layer/conv_3"),
+      conv_4: extractPointwiseConvParams("Prediction", 4, "prediction_layer/conv_4"),
+      conv_5: extractPointwiseConvParams("Prediction", 5, "prediction_layer/conv_5"),
+      conv_6: extractPointwiseConvParams("Prediction", 6, "prediction_layer/conv_6"),
+      conv_7: extractPointwiseConvParams("Prediction", 7, "prediction_layer/conv_7"),
+      box_predictor_0: extractBoxPredictorParams(0),
+      box_predictor_1: extractBoxPredictorParams(1),
+      box_predictor_2: extractBoxPredictorParams(2),
+      box_predictor_3: extractBoxPredictorParams(3),
+      box_predictor_4: extractBoxPredictorParams(4),
+      box_predictor_5: extractBoxPredictorParams(5)
+    };
+  }
+  return {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  };
+}
+function extractParamsFromWeightMap6(weightMap) {
+  const paramMappings = [];
+  const {
+    extractMobilenetV1Params,
+    extractPredictionLayerParams
+  } = extractorsFactory6(weightMap, paramMappings);
+  const extra_dim = weightMap["Output/extra_dim"];
+  paramMappings.push({ originalPath: "Output/extra_dim", paramPath: "output_layer/extra_dim" });
+  if (!isTensor3D(extra_dim)) {
+    throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${extra_dim}`);
+  }
+  const params = {
+    mobilenetv1: extractMobilenetV1Params(),
+    prediction_layer: extractPredictionLayerParams(),
+    output_layer: {
+      extra_dim
+    }
+  };
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var tf30 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/pointwiseConvLayer.ts
+var tf29 = __toESM(require_tfjs_esm());
+function pointwiseConvLayer(x, params, strides) {
+  return tf29.tidy(() => {
+    let out = tf29.conv2d(x, params.filters, strides, "same");
+    out = tf29.add(out, params.batch_norm_offset);
+    return tf29.clipByValue(out, 0, 6);
+  });
+}
+
+// src/ssdMobilenetv1/mobileNetV1.ts
+var epsilon = 0.0010000000474974513;
+function depthwiseConvLayer(x, params, strides) {
+  return tf30.tidy(() => {
+    let out = tf30.depthwiseConv2d(x, params.filters, strides, "same");
+    out = tf30.batchNorm(
+      out,
+      params.batch_norm_mean,
+      params.batch_norm_variance,
+      params.batch_norm_offset,
+      params.batch_norm_scale,
+      epsilon
+    );
+    return tf30.clipByValue(out, 0, 6);
+  });
+}
+function getStridesForLayerIdx(layerIdx) {
+  return [2, 4, 6, 12].some((idx) => idx === layerIdx) ? [2, 2] : [1, 1];
+}
+function mobileNetV1(x, params) {
+  return tf30.tidy(() => {
+    let conv11;
+    let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
+    const convPairParams = [
+      params.conv_1,
+      params.conv_2,
+      params.conv_3,
+      params.conv_4,
+      params.conv_5,
+      params.conv_6,
+      params.conv_7,
+      params.conv_8,
+      params.conv_9,
+      params.conv_10,
+      params.conv_11,
+      params.conv_12,
+      params.conv_13
+    ];
+    convPairParams.forEach((param, i) => {
+      const layerIdx = i + 1;
+      const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
+      out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
+      out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
+      if (layerIdx === 11)
+        conv11 = out;
+    });
+    if (conv11 === null) {
+      throw new Error("mobileNetV1 - output of conv layer 11 is null");
+    }
+    return {
+      out,
+      conv11
+    };
+  });
+}
+
+// src/ssdMobilenetv1/nonMaxSuppression.ts
+function IOU(boxes, i, j) {
+  const boxesData = boxes.arraySync();
+  const yminI = Math.min(boxesData[i][0], boxesData[i][2]);
+  const xminI = Math.min(boxesData[i][1], boxesData[i][3]);
+  const ymaxI = Math.max(boxesData[i][0], boxesData[i][2]);
+  const xmaxI = Math.max(boxesData[i][1], boxesData[i][3]);
+  const yminJ = Math.min(boxesData[j][0], boxesData[j][2]);
+  const xminJ = Math.min(boxesData[j][1], boxesData[j][3]);
+  const ymaxJ = Math.max(boxesData[j][0], boxesData[j][2]);
+  const xmaxJ = Math.max(boxesData[j][1], boxesData[j][3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0)
+    return 0;
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function nonMaxSuppression2(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  const numBoxes = boxes.shape[0];
+  const outputSize = Math.min(maxOutputSize, numBoxes);
+  const candidates = scores.map((score, boxIndex) => ({ score, boxIndex })).filter((c) => c.score > scoreThreshold).sort((c1, c2) => c2.score - c1.score);
+  const suppressFunc = (x) => x <= iouThreshold ? 1 : 0;
+  const selected = [];
+  candidates.forEach((c) => {
+    if (selected.length >= outputSize)
+      return;
+    const originalScore = c.score;
+    for (let j = selected.length - 1; j >= 0; --j) {
+      const iou2 = IOU(boxes, c.boxIndex, selected[j]);
+      if (iou2 === 0)
+        continue;
+      c.score *= suppressFunc(iou2);
+      if (c.score <= scoreThreshold)
+        break;
+    }
+    if (originalScore === c.score) {
+      selected.push(c.boxIndex);
+    }
+  });
+  return selected;
+}
+
+// src/ssdMobilenetv1/outputLayer.ts
+var tf31 = __toESM(require_tfjs_esm());
+function getCenterCoordinatesAndSizesLayer(x) {
+  const vec = tf31.unstack(tf31.transpose(x, [1, 0]));
+  const sizes = [
+    tf31.sub(vec[2], vec[0]),
+    tf31.sub(vec[3], vec[1])
+  ];
+  const centers = [
+    tf31.add(vec[0], tf31.div(sizes[0], 2)),
+    tf31.add(vec[1], tf31.div(sizes[1], 2))
+  ];
+  return { sizes, centers };
+}
+function decodeBoxesLayer(x0, x1) {
+  const { sizes, centers } = getCenterCoordinatesAndSizesLayer(x0);
+  const vec = tf31.unstack(tf31.transpose(x1, [1, 0]));
+  const div0_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[2], 5)), sizes[0]), 2);
+  const add0_out = tf31.add(tf31.mul(tf31.div(vec[0], 10), sizes[0]), centers[0]);
+  const div1_out = tf31.div(tf31.mul(tf31.exp(tf31.div(vec[3], 5)), sizes[1]), 2);
+  const add1_out = tf31.add(tf31.mul(tf31.div(vec[1], 10), sizes[1]), centers[1]);
+  return tf31.transpose(
+    tf31.stack([
+      tf31.sub(add0_out, div0_out),
+      tf31.sub(add1_out, div1_out),
+      tf31.add(add0_out, div0_out),
+      tf31.add(add1_out, div1_out)
+    ]),
+    [1, 0]
+  );
+}
+function outputLayer(boxPredictions, classPredictions, params) {
+  return tf31.tidy(() => {
+    const batchSize = boxPredictions.shape[0];
+    let boxes = decodeBoxesLayer(
+      tf31.reshape(tf31.tile(params.extra_dim, [batchSize, 1, 1]), [-1, 4]),
+      tf31.reshape(boxPredictions, [-1, 4])
+    );
+    boxes = tf31.reshape(boxes, [batchSize, boxes.shape[0] / batchSize, 4]);
+    const scoresAndClasses = tf31.sigmoid(tf31.slice(classPredictions, [0, 0, 1], [-1, -1, -1]));
+    let scores = tf31.slice(scoresAndClasses, [0, 0, 0], [-1, -1, 1]);
+    scores = tf31.reshape(scores, [batchSize, scores.shape[1]]);
+    const boxesByBatch = tf31.unstack(boxes);
+    const scoresByBatch = tf31.unstack(scores);
+    return { boxes: boxesByBatch, scores: scoresByBatch };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+var tf33 = __toESM(require_tfjs_esm());
+
+// src/ssdMobilenetv1/boxPredictionLayer.ts
+var tf32 = __toESM(require_tfjs_esm());
+function boxPredictionLayer(x, params) {
+  return tf32.tidy(() => {
+    const batchSize = x.shape[0];
+    const boxPredictionEncoding = tf32.reshape(
+      convLayer(x, params.box_encoding_predictor),
+      [batchSize, -1, 1, 4]
+    );
+    const classPrediction = tf32.reshape(
+      convLayer(x, params.class_predictor),
+      [batchSize, -1, 3]
+    );
+    return { boxPredictionEncoding, classPrediction };
+  });
+}
+
+// src/ssdMobilenetv1/predictionLayer.ts
+function predictionLayer(x, conv11, params) {
+  return tf33.tidy(() => {
+    const conv0 = pointwiseConvLayer(x, params.conv_0, [1, 1]);
+    const conv1 = pointwiseConvLayer(conv0, params.conv_1, [2, 2]);
+    const conv22 = pointwiseConvLayer(conv1, params.conv_2, [1, 1]);
+    const conv3 = pointwiseConvLayer(conv22, params.conv_3, [2, 2]);
+    const conv4 = pointwiseConvLayer(conv3, params.conv_4, [1, 1]);
+    const conv5 = pointwiseConvLayer(conv4, params.conv_5, [2, 2]);
+    const conv6 = pointwiseConvLayer(conv5, params.conv_6, [1, 1]);
+    const conv7 = pointwiseConvLayer(conv6, params.conv_7, [2, 2]);
+    const boxPrediction0 = boxPredictionLayer(conv11, params.box_predictor_0);
+    const boxPrediction1 = boxPredictionLayer(x, params.box_predictor_1);
+    const boxPrediction2 = boxPredictionLayer(conv1, params.box_predictor_2);
+    const boxPrediction3 = boxPredictionLayer(conv3, params.box_predictor_3);
+    const boxPrediction4 = boxPredictionLayer(conv5, params.box_predictor_4);
+    const boxPrediction5 = boxPredictionLayer(conv7, params.box_predictor_5);
+    const boxPredictions = tf33.concat([
+      boxPrediction0.boxPredictionEncoding,
+      boxPrediction1.boxPredictionEncoding,
+      boxPrediction2.boxPredictionEncoding,
+      boxPrediction3.boxPredictionEncoding,
+      boxPrediction4.boxPredictionEncoding,
+      boxPrediction5.boxPredictionEncoding
+    ], 1);
+    const classPredictions = tf33.concat([
+      boxPrediction0.classPrediction,
+      boxPrediction1.classPrediction,
+      boxPrediction2.classPrediction,
+      boxPrediction3.classPrediction,
+      boxPrediction4.classPrediction,
+      boxPrediction5.classPrediction
+    ], 1);
+    return {
+      boxPredictions,
+      classPredictions
+    };
+  });
+}
+
+// src/ssdMobilenetv1/SsdMobilenetv1Options.ts
+var SsdMobilenetv1Options = class {
+  constructor({ minConfidence, maxResults } = {}) {
+    this._name = "SsdMobilenetv1Options";
+    this._minConfidence = minConfidence || 0.5;
+    this._maxResults = maxResults || 100;
+    if (typeof this._minConfidence !== "number" || this._minConfidence <= 0 || this._minConfidence >= 1) {
+      throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);
+    }
+    if (typeof this._maxResults !== "number") {
+      throw new Error(`${this._name} - expected maxResults to be a number`);
+    }
+  }
+  get minConfidence() {
+    return this._minConfidence;
+  }
+  get maxResults() {
+    return this._maxResults;
+  }
+};
+
+// src/ssdMobilenetv1/SsdMobilenetv1.ts
+var SsdMobilenetv1 = class extends NeuralNetwork {
+  constructor() {
+    super("SsdMobilenetv1");
+  }
+  forwardInput(input) {
+    const { params } = this;
+    if (!params)
+      throw new Error("SsdMobilenetv1 - load model before inference");
+    return tf34.tidy(() => {
+      const batchTensor = tf34.cast(input.toBatchTensor(512, false), "float32");
+      const x = tf34.sub(tf34.div(batchTensor, 127.5), 1);
+      const features = mobileNetV1(x, params.mobilenetv1);
+      const { boxPredictions, classPredictions } = predictionLayer(features.out, features.conv11, params.prediction_layer);
+      return outputLayer(boxPredictions, classPredictions, params.output_layer);
+    });
+  }
+  async forward(input) {
+    return this.forwardInput(await toNetInput(input));
+  }
+  async locateFaces(input, options = {}) {
+    const { maxResults, minConfidence } = new SsdMobilenetv1Options(options);
+    const netInput = await toNetInput(input);
+    const { boxes: _boxes, scores: _scores } = this.forwardInput(netInput);
+    const boxes = _boxes[0];
+    const scores = _scores[0];
+    for (let i = 1; i < _boxes.length; i++) {
+      _boxes[i].dispose();
+      _scores[i].dispose();
+    }
+    const scoresData = Array.from(scores.dataSync());
+    const iouThreshold = 0.5;
+    const indices = nonMaxSuppression2(boxes, scoresData, maxResults, iouThreshold, minConfidence);
+    const reshapedDims = netInput.getReshapedInputDimensions(0);
+    const inputSize = netInput.inputSize;
+    const padX = inputSize / reshapedDims.width;
+    const padY = inputSize / reshapedDims.height;
+    const boxesData = boxes.arraySync();
+    const results = indices.map((idx) => {
+      const [top, bottom] = [
+        Math.max(0, boxesData[idx][0]),
+        Math.min(1, boxesData[idx][2])
+      ].map((val) => val * padY);
+      const [left, right] = [
+        Math.max(0, boxesData[idx][1]),
+        Math.min(1, boxesData[idx][3])
+      ].map((val) => val * padX);
+      return new FaceDetection(
+        scoresData[idx],
+        new Rect(left, top, right - left, bottom - top),
+        { height: netInput.getInputHeight(0), width: netInput.getInputWidth(0) }
+      );
+    });
+    boxes.dispose();
+    scores.dispose();
+    return results;
+  }
+  getDefaultModelName() {
+    return "ssd_mobilenetv1_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap6(weightMap);
+  }
+  extractParams(weights) {
+    return extractParams6(weights);
+  }
+};
+
+// src/ssdMobilenetv1/index.ts
+function createSsdMobilenetv1(weights) {
+  const net = new SsdMobilenetv1();
+  net.extractWeights(weights);
+  return net;
+}
+function createFaceDetectionNet(weights) {
+  return createSsdMobilenetv1(weights);
+}
+var FaceDetectionNet = class extends SsdMobilenetv1 {
+};
+
+// src/tinyYolov2/const.ts
+var IOU_THRESHOLD = 0.4;
+var BOX_ANCHORS = [
+  new Point(0.738768, 0.874946),
+  new Point(2.42204, 2.65704),
+  new Point(4.30971, 7.04493),
+  new Point(10.246, 4.59428),
+  new Point(12.6868, 11.8741)
+];
+var BOX_ANCHORS_SEPARABLE = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB_SEPARABLE = [117.001, 114.697, 97.404];
+var DEFAULT_MODEL_NAME = "tiny_yolov2_model";
+var DEFAULT_MODEL_NAME_SEPARABLE_CONV = "tiny_yolov2_separable_conv_model";
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var tf39 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/config.ts
+var isNumber = (arg) => typeof arg === "number";
+function validateConfig(config) {
+  if (!config) {
+    throw new Error(`invalid config: ${config}`);
+  }
+  if (typeof config.withSeparableConvs !== "boolean") {
+    throw new Error(`config.withSeparableConvs has to be a boolean, have: ${config.withSeparableConvs}`);
+  }
+  if (!isNumber(config.iouThreshold) || config.iouThreshold < 0 || config.iouThreshold > 1) {
+    throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${config.iouThreshold}`);
+  }
+  if (!Array.isArray(config.classes) || !config.classes.length || !config.classes.every((c) => typeof c === "string")) {
+    throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(config.classes)}`);
+  }
+  if (!Array.isArray(config.anchors) || !config.anchors.length || !config.anchors.map((a) => a || {}).every((a) => isNumber(a.x) && isNumber(a.y))) {
+    throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(config.anchors)}`);
+  }
+  if (config.meanRgb && (!Array.isArray(config.meanRgb) || config.meanRgb.length !== 3 || !config.meanRgb.every(isNumber))) {
+    throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(config.meanRgb)}`);
+  }
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+var tf36 = __toESM(require_tfjs_esm());
+
+// src/tinyYolov2/leaky.ts
+var tf35 = __toESM(require_tfjs_esm());
+function leaky(x) {
+  return tf35.tidy(() => {
+    const min = tf35.mul(x, tf35.scalar(0.10000000149011612));
+    return tf35.add(tf35.relu(tf35.sub(x, min)), min);
+  });
+}
+
+// src/tinyYolov2/convWithBatchNorm.ts
+function convWithBatchNorm(x, params) {
+  return tf36.tidy(() => {
+    let out = tf36.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf36.conv2d(out, params.conv.filters, [1, 1], "valid");
+    out = tf36.sub(out, params.bn.sub);
+    out = tf36.mul(out, params.bn.truediv);
+    out = tf36.add(out, params.conv.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/depthwiseSeparableConv.ts
+var tf37 = __toESM(require_tfjs_esm());
+function depthwiseSeparableConv2(x, params) {
+  return tf37.tidy(() => {
+    let out = tf37.pad(x, [[0, 0], [1, 1], [1, 1], [0, 0]]);
+    out = tf37.separableConv2d(out, params.depthwise_filter, params.pointwise_filter, [1, 1], "valid");
+    out = tf37.add(out, params.bias);
+    return leaky(out);
+  });
+}
+
+// src/tinyYolov2/extractParams.ts
+var tf38 = __toESM(require_tfjs_esm());
+function extractorsFactory7(extractWeights, paramMappings) {
+  const extractConvParams = extractConvParamsFactory(extractWeights, paramMappings);
+  function extractBatchNormParams(size, mappedPrefix) {
+    const sub6 = tf38.tensor1d(extractWeights(size));
+    const truediv = tf38.tensor1d(extractWeights(size));
+    paramMappings.push(
+      { paramPath: `${mappedPrefix}/sub` },
+      { paramPath: `${mappedPrefix}/truediv` }
+    );
+    return { sub: sub6, truediv };
+  }
+  function extractConvWithBatchNormParams(channelsIn, channelsOut, mappedPrefix) {
+    const conv3 = extractConvParams(channelsIn, channelsOut, 3, `${mappedPrefix}/conv`);
+    const bn = extractBatchNormParams(channelsOut, `${mappedPrefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = extractSeparableConvParamsFactory(extractWeights, paramMappings);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParams7(weights, config, boxEncodingSize, filterSizes) {
+  const {
+    extractWeights,
+    getRemainingWeights
+  } = extractWeightsFactory(weights);
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory7(extractWeights, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = config.isFirstLayerConv2d ? extractConvParams(s0, s1, 3, "conv0") : extractSeparableConvParams(s0, s1, "conv0");
+    const conv1 = extractSeparableConvParams(s1, s2, "conv1");
+    const conv22 = extractSeparableConvParams(s2, s3, "conv2");
+    const conv3 = extractSeparableConvParams(s3, s4, "conv3");
+    const conv4 = extractSeparableConvParams(s4, s5, "conv4");
+    const conv5 = extractSeparableConvParams(s5, s6, "conv5");
+    const conv6 = s7 ? extractSeparableConvParams(s6, s7, "conv6") : void 0;
+    const conv7 = s8 ? extractSeparableConvParams(s7, s8, "conv7") : void 0;
+    const conv8 = extractConvParams(s8 || s7 || s6, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  } else {
+    const [s0, s1, s2, s3, s4, s5, s6, s7, s8] = filterSizes;
+    const conv0 = extractConvWithBatchNormParams(s0, s1, "conv0");
+    const conv1 = extractConvWithBatchNormParams(s1, s2, "conv1");
+    const conv22 = extractConvWithBatchNormParams(s2, s3, "conv2");
+    const conv3 = extractConvWithBatchNormParams(s3, s4, "conv3");
+    const conv4 = extractConvWithBatchNormParams(s4, s5, "conv4");
+    const conv5 = extractConvWithBatchNormParams(s5, s6, "conv5");
+    const conv6 = extractConvWithBatchNormParams(s6, s7, "conv6");
+    const conv7 = extractConvWithBatchNormParams(s7, s8, "conv7");
+    const conv8 = extractConvParams(s8, 5 * boxEncodingSize, 1, "conv8");
+    params = {
+      conv0,
+      conv1,
+      conv2: conv22,
+      conv3,
+      conv4,
+      conv5,
+      conv6,
+      conv7,
+      conv8
+    };
+  }
+  if (getRemainingWeights().length !== 0) {
+    throw new Error(`weights remaing after extract: ${getRemainingWeights().length}`);
+  }
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/extractParamsFromWeightMap.ts
+function extractorsFactory8(weightMap, paramMappings) {
+  const extractWeightEntry = extractWeightEntryFactory(weightMap, paramMappings);
+  function extractBatchNormParams(prefix) {
+    const sub6 = extractWeightEntry(`${prefix}/sub`, 1);
+    const truediv = extractWeightEntry(`${prefix}/truediv`, 1);
+    return { sub: sub6, truediv };
+  }
+  function extractConvParams(prefix) {
+    const filters = extractWeightEntry(`${prefix}/filters`, 4);
+    const bias = extractWeightEntry(`${prefix}/bias`, 1);
+    return { filters, bias };
+  }
+  function extractConvWithBatchNormParams(prefix) {
+    const conv3 = extractConvParams(`${prefix}/conv`);
+    const bn = extractBatchNormParams(`${prefix}/bn`);
+    return { conv: conv3, bn };
+  }
+  const extractSeparableConvParams = loadSeparableConvParamsFactory(extractWeightEntry);
+  return {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  };
+}
+function extractParamsFromWeightMap7(weightMap, config) {
+  const paramMappings = [];
+  const {
+    extractConvParams,
+    extractConvWithBatchNormParams,
+    extractSeparableConvParams
+  } = extractorsFactory8(weightMap, paramMappings);
+  let params;
+  if (config.withSeparableConvs) {
+    const numFilters = config.filterSizes && config.filterSizes.length || 9;
+    params = {
+      conv0: config.isFirstLayerConv2d ? extractConvParams("conv0") : extractSeparableConvParams("conv0"),
+      conv1: extractSeparableConvParams("conv1"),
+      conv2: extractSeparableConvParams("conv2"),
+      conv3: extractSeparableConvParams("conv3"),
+      conv4: extractSeparableConvParams("conv4"),
+      conv5: extractSeparableConvParams("conv5"),
+      conv6: numFilters > 7 ? extractSeparableConvParams("conv6") : void 0,
+      conv7: numFilters > 8 ? extractSeparableConvParams("conv7") : void 0,
+      conv8: extractConvParams("conv8")
+    };
+  } else {
+    params = {
+      conv0: extractConvWithBatchNormParams("conv0"),
+      conv1: extractConvWithBatchNormParams("conv1"),
+      conv2: extractConvWithBatchNormParams("conv2"),
+      conv3: extractConvWithBatchNormParams("conv3"),
+      conv4: extractConvWithBatchNormParams("conv4"),
+      conv5: extractConvWithBatchNormParams("conv5"),
+      conv6: extractConvWithBatchNormParams("conv6"),
+      conv7: extractConvWithBatchNormParams("conv7"),
+      conv8: extractConvParams("conv8")
+    };
+  }
+  disposeUnusedWeightTensors(weightMap, paramMappings);
+  return { params, paramMappings };
+}
+
+// src/tinyYolov2/TinyYolov2Options.ts
+var TinyYolov2Options = class {
+  constructor({ inputSize, scoreThreshold } = {}) {
+    this._name = "TinyYolov2Options";
+    this._inputSize = inputSize || 416;
+    this._scoreThreshold = scoreThreshold || 0.5;
+    if (typeof this._inputSize !== "number" || this._inputSize % 32 !== 0) {
+      throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);
+    }
+    if (typeof this._scoreThreshold !== "number" || this._scoreThreshold <= 0 || this._scoreThreshold >= 1) {
+      throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`);
+    }
+  }
+  get inputSize() {
+    return this._inputSize;
+  }
+  get scoreThreshold() {
+    return this._scoreThreshold;
+  }
+};
+
+// src/tinyYolov2/TinyYolov2Base.ts
+var _TinyYolov2Base = class extends NeuralNetwork {
+  constructor(config) {
+    super("TinyYolov2");
+    validateConfig(config);
+    this._config = config;
+  }
+  get config() {
+    return this._config;
+  }
+  get withClassScores() {
+    return this.config.withClassScores || this.config.classes.length > 1;
+  }
+  get boxEncodingSize() {
+    return 5 + (this.withClassScores ? this.config.classes.length : 0);
+  }
+  runTinyYolov2(x, params) {
+    let out = convWithBatchNorm(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = convWithBatchNorm(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = convWithBatchNorm(out, params.conv6);
+    out = convWithBatchNorm(out, params.conv7);
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  runMobilenet(x, params) {
+    let out = this.config.isFirstLayerConv2d ? leaky(convLayer(x, params.conv0, "valid", false)) : depthwiseSeparableConv2(x, params.conv0);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv1);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv2);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv3);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv4);
+    out = tf39.maxPool(out, [2, 2], [2, 2], "same");
+    out = depthwiseSeparableConv2(out, params.conv5);
+    out = tf39.maxPool(out, [2, 2], [1, 1], "same");
+    out = params.conv6 ? depthwiseSeparableConv2(out, params.conv6) : out;
+    out = params.conv7 ? depthwiseSeparableConv2(out, params.conv7) : out;
+    return convLayer(out, params.conv8, "valid", false);
+  }
+  forwardInput(input, inputSize) {
+    const { params } = this;
+    if (!params) {
+      throw new Error("TinyYolov2 - load model before inference");
+    }
+    return tf39.tidy(() => {
+      let batchTensor = tf39.cast(input.toBatchTensor(inputSize, false), "float32");
+      batchTensor = this.config.meanRgb ? normalize(batchTensor, this.config.meanRgb) : batchTensor;
+      batchTensor = batchTensor.div(255);
+      return this.config.withSeparableConvs ? this.runMobilenet(batchTensor, params) : this.runTinyYolov2(batchTensor, params);
+    });
+  }
+  async forward(input, inputSize) {
+    return this.forwardInput(await toNetInput(input), inputSize);
+  }
+  async detect(input, forwardParams = {}) {
+    const { inputSize, scoreThreshold } = new TinyYolov2Options(forwardParams);
+    const netInput = await toNetInput(input);
+    const out = await this.forwardInput(netInput, inputSize);
+    const out0 = tf39.tidy(() => tf39.unstack(out)[0].expandDims());
+    const inputDimensions = {
+      width: netInput.getInputWidth(0),
+      height: netInput.getInputHeight(0)
+    };
+    const results = await this.extractBoxes(out0, netInput.getReshapedInputDimensions(0), scoreThreshold);
+    out.dispose();
+    out0.dispose();
+    const boxes = results.map((res) => res.box);
+    const scores = results.map((res) => res.score);
+    const classScores = results.map((res) => res.classScore);
+    const classNames = results.map((res) => this.config.classes[res.label]);
+    const indices = nonMaxSuppression(
+      boxes.map((box) => box.rescale(inputSize)),
+      scores,
+      this.config.iouThreshold,
+      true
+    );
+    const detections = indices.map((idx) => new ObjectDetection(
+      scores[idx],
+      classScores[idx],
+      classNames[idx],
+      boxes[idx],
+      inputDimensions
+    ));
+    return detections;
+  }
+  getDefaultModelName() {
+    return "";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return extractParamsFromWeightMap7(weightMap, this.config);
+  }
+  extractParams(weights) {
+    const filterSizes = this.config.filterSizes || _TinyYolov2Base.DEFAULT_FILTER_SIZES;
+    const numFilters = filterSizes ? filterSizes.length : void 0;
+    if (numFilters !== 7 && numFilters !== 8 && numFilters !== 9) {
+      throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${numFilters} filterSizes in config`);
+    }
+    return extractParams7(weights, this.config, this.boxEncodingSize, filterSizes);
+  }
+  async extractBoxes(outputTensor, inputBlobDimensions, scoreThreshold) {
+    const { width, height } = inputBlobDimensions;
+    const inputSize = Math.max(width, height);
+    const correctionFactorX = inputSize / width;
+    const correctionFactorY = inputSize / height;
+    const numCells = outputTensor.shape[1];
+    const numBoxes = this.config.anchors.length;
+    const [boxesTensor, scoresTensor, classScoresTensor] = tf39.tidy(() => {
+      const reshaped = outputTensor.reshape([numCells, numCells, numBoxes, this.boxEncodingSize]);
+      const boxes = reshaped.slice([0, 0, 0, 0], [numCells, numCells, numBoxes, 4]);
+      const scores = reshaped.slice([0, 0, 0, 4], [numCells, numCells, numBoxes, 1]);
+      const classScores = this.withClassScores ? tf39.softmax(reshaped.slice([0, 0, 0, 5], [numCells, numCells, numBoxes, this.config.classes.length]), 3) : tf39.scalar(0);
+      return [boxes, scores, classScores];
+    });
+    const results = [];
+    const scoresData = await scoresTensor.array();
+    const boxesData = await boxesTensor.array();
+    for (let row = 0; row < numCells; row++) {
+      for (let col = 0; col < numCells; col++) {
+        for (let anchor = 0; anchor < numBoxes; anchor++) {
+          const score = sigmoid(scoresData[row][col][anchor][0]);
+          if (!scoreThreshold || score > scoreThreshold) {
+            const ctX = (col + sigmoid(boxesData[row][col][anchor][0])) / numCells * correctionFactorX;
+            const ctY = (row + sigmoid(boxesData[row][col][anchor][1])) / numCells * correctionFactorY;
+            const widthLocal = Math.exp(boxesData[row][col][anchor][2]) * this.config.anchors[anchor].x / numCells * correctionFactorX;
+            const heightLocal = Math.exp(boxesData[row][col][anchor][3]) * this.config.anchors[anchor].y / numCells * correctionFactorY;
+            const x = ctX - widthLocal / 2;
+            const y = ctY - heightLocal / 2;
+            const pos = { row, col, anchor };
+            const { classScore, label } = this.withClassScores ? await this.extractPredictedClass(classScoresTensor, pos) : { classScore: 1, label: 0 };
+            results.push({
+              box: new BoundingBox(x, y, x + widthLocal, y + heightLocal),
+              score,
+              classScore: score * classScore,
+              label,
+              ...pos
+            });
+          }
+        }
+      }
+    }
+    boxesTensor.dispose();
+    scoresTensor.dispose();
+    classScoresTensor.dispose();
+    return results;
+  }
+  async extractPredictedClass(classesTensor, pos) {
+    const { row, col, anchor } = pos;
+    const classesData = await classesTensor.array();
+    return Array(this.config.classes.length).fill(0).map((_, i) => classesData[row][col][anchor][i]).map((classScore, label) => ({
+      classScore,
+      label
+    })).reduce((max, curr) => max.classScore > curr.classScore ? max : curr);
+  }
+};
+var TinyYolov2Base = _TinyYolov2Base;
+TinyYolov2Base.DEFAULT_FILTER_SIZES = [3, 16, 32, 64, 128, 256, 512, 1024, 1024];
+
+// src/tinyYolov2/TinyYolov2.ts
+var TinyYolov2 = class extends TinyYolov2Base {
+  constructor(withSeparableConvs = true) {
+    const config = {
+      withSeparableConvs,
+      iouThreshold: IOU_THRESHOLD,
+      classes: ["face"],
+      ...withSeparableConvs ? {
+        anchors: BOX_ANCHORS_SEPARABLE,
+        meanRgb: MEAN_RGB_SEPARABLE
+      } : {
+        anchors: BOX_ANCHORS,
+        withClassScores: true
+      }
+    };
+    super(config);
+  }
+  get withSeparableConvs() {
+    return this.config.withSeparableConvs;
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return this.withSeparableConvs ? DEFAULT_MODEL_NAME_SEPARABLE_CONV : DEFAULT_MODEL_NAME;
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/tinyYolov2/index.ts
+function createTinyYolov2(weights, withSeparableConvs = true) {
+  const net = new TinyYolov2(withSeparableConvs);
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/tinyFaceDetector/TinyFaceDetectorOptions.ts
+var TinyFaceDetectorOptions = class extends TinyYolov2Options {
+  constructor() {
+    super(...arguments);
+    this._name = "TinyFaceDetectorOptions";
+  }
+};
+
+// src/globalApi/ComposableTask.ts
+var ComposableTask = class {
+  async then(onfulfilled) {
+    return onfulfilled(await this.run());
+  }
+  async run() {
+    throw new Error("ComposableTask - run is not implemented");
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var tf41 = __toESM(require_tfjs_esm());
+
+// src/globalApi/extractFacesAndComputeResults.ts
+var tf40 = __toESM(require_tfjs_esm());
+async function extractAllFacesAndComputeResults(parentResults, input, computeResults, extractedFaces, getRectForAlignment = ({ alignedRect }) => alignedRect) {
+  const faceBoxes = parentResults.map((parentResult) => isWithFaceLandmarks(parentResult) ? getRectForAlignment(parentResult) : parentResult.detection);
+  const faces = extractedFaces || (input instanceof tf40.Tensor ? await extractFaceTensors(input, faceBoxes) : await extractFaces(input, faceBoxes));
+  const results = await computeResults(faces);
+  faces.forEach((f) => f instanceof tf40.Tensor && f.dispose());
+  return results;
+}
+async function extractSingleFaceAndComputeResult(parentResult, input, computeResult, extractedFaces, getRectForAlignment) {
+  return extractAllFacesAndComputeResults(
+    [parentResult],
+    input,
+    async (faces) => computeResult(faces[0]),
+    extractedFaces,
+    getRectForAlignment
+  );
+}
+
+// src/tinyFaceDetector/const.ts
+var IOU_THRESHOLD2 = 0.4;
+var BOX_ANCHORS2 = [
+  new Point(1.603231, 2.094468),
+  new Point(6.041143, 7.080126),
+  new Point(2.882459, 3.518061),
+  new Point(4.266906, 5.178857),
+  new Point(9.041765, 10.66308)
+];
+var MEAN_RGB = [117.001, 114.697, 97.404];
+
+// src/tinyFaceDetector/TinyFaceDetector.ts
+var TinyFaceDetector = class extends TinyYolov2Base {
+  constructor() {
+    const config = {
+      withSeparableConvs: true,
+      iouThreshold: IOU_THRESHOLD2,
+      classes: ["face"],
+      anchors: BOX_ANCHORS2,
+      meanRgb: MEAN_RGB,
+      isFirstLayerConv2d: true,
+      filterSizes: [3, 16, 32, 64, 128, 256, 512]
+    };
+    super(config);
+  }
+  get anchors() {
+    return this.config.anchors;
+  }
+  async locateFaces(input, forwardParams) {
+    const objectDetections = await this.detect(input, forwardParams);
+    return objectDetections.map((det) => new FaceDetection(det.score, det.relativeBox, { width: det.imageWidth, height: det.imageHeight }));
+  }
+  getDefaultModelName() {
+    return "tiny_face_detector_model";
+  }
+  extractParamsFromWeightMap(weightMap) {
+    return super.extractParamsFromWeightMap(weightMap);
+  }
+};
+
+// src/globalApi/nets.ts
+var nets = {
+  ssdMobilenetv1: new SsdMobilenetv1(),
+  tinyFaceDetector: new TinyFaceDetector(),
+  tinyYolov2: new TinyYolov2(),
+  faceLandmark68Net: new FaceLandmark68Net(),
+  faceLandmark68TinyNet: new FaceLandmark68TinyNet(),
+  faceRecognitionNet: new FaceRecognitionNet(),
+  faceExpressionNet: new FaceExpressionNet(),
+  ageGenderNet: new AgeGenderNet()
+};
+var ssdMobilenetv1 = (input, options) => nets.ssdMobilenetv1.locateFaces(input, options);
+var tinyFaceDetector = (input, options) => nets.tinyFaceDetector.locateFaces(input, options);
+var tinyYolov2 = (input, options) => nets.tinyYolov2.locateFaces(input, options);
+var detectFaceLandmarks = (input) => nets.faceLandmark68Net.detectLandmarks(input);
+var detectFaceLandmarksTiny = (input) => nets.faceLandmark68TinyNet.detectLandmarks(input);
+var computeFaceDescriptor = (input) => nets.faceRecognitionNet.computeFaceDescriptor(input);
+var recognizeFaceExpressions = (input) => nets.faceExpressionNet.predictExpressions(input);
+var predictAgeAndGender = (input) => nets.ageGenderNet.predictAgeAndGender(input);
+var loadSsdMobilenetv1Model = (url) => nets.ssdMobilenetv1.load(url);
+var loadTinyFaceDetectorModel = (url) => nets.tinyFaceDetector.load(url);
+var loadTinyYolov2Model = (url) => nets.tinyYolov2.load(url);
+var loadFaceLandmarkModel = (url) => nets.faceLandmark68Net.load(url);
+var loadFaceLandmarkTinyModel = (url) => nets.faceLandmark68TinyNet.load(url);
+var loadFaceRecognitionModel = (url) => nets.faceRecognitionNet.load(url);
+var loadFaceExpressionModel = (url) => nets.faceExpressionNet.load(url);
+var loadAgeGenderModel = (url) => nets.ageGenderNet.load(url);
+var loadFaceDetectionModel = loadSsdMobilenetv1Model;
+var locateFaces = ssdMobilenetv1;
+var detectLandmarks = detectFaceLandmarks;
+
+// src/globalApi/PredictFaceExpressionsTask.ts
+var PredictFaceExpressionsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const faceExpressionsByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(
+        faces.map((face) => nets.faceExpressionNet.predictExpressions(face))
+      ),
+      this.extractedFaces
+    );
+    return parentResults.map(
+      (parentResult, i) => extendWithFaceExpressions(parentResult, faceExpressionsByFace[i])
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsTask = class extends PredictFaceExpressionsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const faceExpressions = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceExpressionNet.predictExpressions(face),
+      this.extractedFaces
+    );
+    return extendWithFaceExpressions(parentResult, faceExpressions);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(this, this.input);
+  }
+};
+var PredictAllFaceExpressionsWithFaceAlignmentTask = class extends PredictAllFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleFaceExpressionsWithFaceAlignmentTask = class extends PredictSingleFaceExpressionsTask {
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/PredictAgeAndGenderTask.ts
+var PredictAgeAndGenderTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, extractedFaces) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.extractedFaces = extractedFaces;
+  }
+};
+var PredictAllAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const ageAndGenderByFace = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      async (faces) => Promise.all(faces.map((face) => nets.ageGenderNet.predictAgeAndGender(face))),
+      this.extractedFaces
+    );
+    return parentResults.map((parentResult, i) => {
+      const { age, gender, genderProbability } = ageAndGenderByFace[i];
+      return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+    });
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderTask = class extends PredictAgeAndGenderTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const { age, gender, genderProbability } = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.ageGenderNet.predictAgeAndGender(face),
+      this.extractedFaces
+    );
+    return extendWithAge(extendWithGender(parentResult, gender, genderProbability), age);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(this, this.input);
+  }
+};
+var PredictAllAgeAndGenderWithFaceAlignmentTask = class extends PredictAllAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var PredictSingleAgeAndGenderWithFaceAlignmentTask = class extends PredictSingleAgeAndGenderTask {
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/ComputeFaceDescriptorsTasks.ts
+var ComputeFaceDescriptorsTaskBase = class extends ComposableTask {
+  constructor(parentTask, input) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+  }
+};
+var ComputeAllFaceDescriptorsTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const descriptors = await extractAllFacesAndComputeResults(
+      parentResults,
+      this.input,
+      (faces) => Promise.all(faces.map((face) => nets.faceRecognitionNet.computeFaceDescriptor(face))),
+      null,
+      (parentResult) => parentResult.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return descriptors.map((descriptor, i) => extendWithFaceDescriptor(parentResults[i], descriptor));
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+var ComputeSingleFaceDescriptorTask = class extends ComputeFaceDescriptorsTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult)
+      return void 0;
+    const descriptor = await extractSingleFaceAndComputeResult(
+      parentResult,
+      this.input,
+      (face) => nets.faceRecognitionNet.computeFaceDescriptor(face),
+      null,
+      (parentResult2) => parentResult2.landmarks.align(null, { useDlibAlignment: true })
+    );
+    return extendWithFaceDescriptor(parentResult, descriptor);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFaceLandmarksTasks.ts
+var DetectFaceLandmarksTaskBase = class extends ComposableTask {
+  constructor(parentTask, input, useTinyLandmarkNet) {
+    super();
+    this.parentTask = parentTask;
+    this.input = input;
+    this.useTinyLandmarkNet = useTinyLandmarkNet;
+  }
+  get landmarkNet() {
+    return this.useTinyLandmarkNet ? nets.faceLandmark68TinyNet : nets.faceLandmark68Net;
+  }
+};
+var DetectAllFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResults = await this.parentTask;
+    const detections = parentResults.map((res) => res.detection);
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, detections) : await extractFaces(this.input, detections);
+    const faceLandmarksByFace = await Promise.all(faces.map((face) => this.landmarkNet.detectLandmarks(face)));
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    const result = parentResults.filter((_parentResult, i) => faceLandmarksByFace[i]).map((parentResult, i) => extendWithFaceLandmarks(parentResult, faceLandmarksByFace[i]));
+    return result;
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptors() {
+    return new ComputeAllFaceDescriptorsTask(this, this.input);
+  }
+};
+var DetectSingleFaceLandmarksTask = class extends DetectFaceLandmarksTaskBase {
+  async run() {
+    const parentResult = await this.parentTask;
+    if (!parentResult) {
+      return void 0;
+    }
+    const { detection } = parentResult;
+    const faces = this.input instanceof tf41.Tensor ? await extractFaceTensors(this.input, [detection]) : await extractFaces(this.input, [detection]);
+    const landmarks = await this.landmarkNet.detectLandmarks(faces[0]);
+    faces.forEach((f) => f instanceof tf41.Tensor && f.dispose());
+    return extendWithFaceLandmarks(parentResult, landmarks);
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsWithFaceAlignmentTask(this, this.input);
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderWithFaceAlignmentTask(this, this.input);
+  }
+  withFaceDescriptor() {
+    return new ComputeSingleFaceDescriptorTask(this, this.input);
+  }
+};
+
+// src/globalApi/DetectFacesTasks.ts
+var DetectFacesTaskBase = class extends ComposableTask {
+  constructor(input, options = new SsdMobilenetv1Options()) {
+    super();
+    this.input = input;
+    this.options = options;
+  }
+};
+var DetectAllFacesTask = class extends DetectFacesTaskBase {
+  async run() {
+    const { input, options } = this;
+    let result;
+    if (options instanceof TinyFaceDetectorOptions)
+      result = nets.tinyFaceDetector.locateFaces(input, options);
+    else if (options instanceof SsdMobilenetv1Options)
+      result = nets.ssdMobilenetv1.locateFaces(input, options);
+    else if (options instanceof TinyYolov2Options)
+      result = nets.tinyYolov2.locateFaces(input, options);
+    else
+      throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");
+    return result;
+  }
+  runAndExtendWithFaceDetections() {
+    return new Promise((resolve, reject) => {
+      this.run().then((detections) => resolve(detections.map((detection) => extendWithFaceDetection({}, detection)))).catch((err) => reject(err));
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectAllFaceLandmarksTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictAllFaceExpressionsTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictAllAgeAndGenderTask(
+      this.runAndExtendWithFaceDetections(),
+      this.input
+    );
+  }
+};
+var DetectSingleFaceTask = class extends DetectFacesTaskBase {
+  async run() {
+    const faceDetections = await new DetectAllFacesTask(this.input, this.options);
+    let faceDetectionWithHighestScore = faceDetections[0];
+    faceDetections.forEach((faceDetection) => {
+      if (faceDetection.score > faceDetectionWithHighestScore.score)
+        faceDetectionWithHighestScore = faceDetection;
+    });
+    return faceDetectionWithHighestScore;
+  }
+  runAndExtendWithFaceDetection() {
+    return new Promise(async (resolve) => {
+      const detection = await this.run();
+      resolve(detection ? extendWithFaceDetection({}, detection) : void 0);
+    });
+  }
+  withFaceLandmarks(useTinyLandmarkNet = false) {
+    return new DetectSingleFaceLandmarksTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input,
+      useTinyLandmarkNet
+    );
+  }
+  withFaceExpressions() {
+    return new PredictSingleFaceExpressionsTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+  withAgeAndGender() {
+    return new PredictSingleAgeAndGenderTask(
+      this.runAndExtendWithFaceDetection(),
+      this.input
+    );
+  }
+};
+
+// src/globalApi/detectFaces.ts
+function detectSingleFace(input, options = new SsdMobilenetv1Options()) {
+  return new DetectSingleFaceTask(input, options);
+}
+function detectAllFaces(input, options = new SsdMobilenetv1Options()) {
+  return new DetectAllFacesTask(input, options);
+}
+
+// src/globalApi/allFaces.ts
+async function allFacesSsdMobilenetv1(input, minConfidence) {
+  return detectAllFaces(input, new SsdMobilenetv1Options(minConfidence ? { minConfidence } : {})).withFaceLandmarks().withFaceDescriptors();
+}
+async function allFacesTinyYolov2(input, forwardParams = {}) {
+  return detectAllFaces(input, new TinyYolov2Options(forwardParams)).withFaceLandmarks().withFaceDescriptors();
+}
+var allFaces = allFacesSsdMobilenetv1;
+
+// src/euclideanDistance.ts
+function euclideanDistance(arr1, arr2) {
+  if (arr1.length !== arr2.length)
+    throw new Error("euclideanDistance: arr1.length !== arr2.length");
+  const desc1 = Array.from(arr1);
+  const desc2 = Array.from(arr2);
+  return Math.sqrt(
+    desc1.map((val, i) => val - desc2[i]).reduce((res, diff) => res + diff * diff, 0)
+  );
+}
+
+// src/globalApi/FaceMatcher.ts
+var FaceMatcher = class {
+  constructor(inputs, distanceThreshold = 0.6) {
+    this._distanceThreshold = distanceThreshold;
+    const inputArray = Array.isArray(inputs) ? inputs : [inputs];
+    if (!inputArray.length)
+      throw new Error("FaceRecognizer.constructor - expected atleast one input");
+    let count = 1;
+    const createUniqueLabel = () => `person ${count++}`;
+    this._labeledDescriptors = inputArray.map((desc) => {
+      if (desc instanceof LabeledFaceDescriptors)
+        return desc;
+      if (desc instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc]);
+      if (desc.descriptor && desc.descriptor instanceof Float32Array)
+        return new LabeledFaceDescriptors(createUniqueLabel(), [desc.descriptor]);
+      throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>");
+    });
+  }
+  get labeledDescriptors() {
+    return this._labeledDescriptors;
+  }
+  get distanceThreshold() {
+    return this._distanceThreshold;
+  }
+  computeMeanDistance(queryDescriptor, descriptors) {
+    return descriptors.map((d) => euclideanDistance(d, queryDescriptor)).reduce((d1, d2) => d1 + d2, 0) / (descriptors.length || 1);
+  }
+  matchDescriptor(queryDescriptor) {
+    return this.labeledDescriptors.map(({ descriptors, label }) => new FaceMatch(label, this.computeMeanDistance(queryDescriptor, descriptors))).reduce((best, curr) => best.distance < curr.distance ? best : curr);
+  }
+  findBestMatch(queryDescriptor) {
+    const bestMatch = this.matchDescriptor(queryDescriptor);
+    return bestMatch.distance < this._distanceThreshold ? bestMatch : new FaceMatch("unknown", bestMatch.distance);
+  }
+  toJSON() {
+    return {
+      distanceThreshold: this._distanceThreshold,
+      labeledDescriptors: this._labeledDescriptors.map((ld) => ld.toJSON())
+    };
+  }
+  static fromJSON(json) {
+    const labeledDescriptors = json.labeledDescriptors.map((ld) => LabeledFaceDescriptors.fromJSON(ld));
+    return new FaceMatcher(labeledDescriptors, json.distanceThreshold);
+  }
+};
+
+// src/tinyFaceDetector/index.ts
+function createTinyFaceDetector(weights) {
+  const net = new TinyFaceDetector();
+  net.extractWeights(weights);
+  return net;
+}
+
+// src/resizeResults.ts
+function resizeResults(results, dimensions) {
+  const { width, height } = new Dimensions(dimensions.width, dimensions.height);
+  if (width <= 0 || height <= 0) {
+    throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({ width, height })}`);
+  }
+  if (Array.isArray(results)) {
+    return results.map((obj) => resizeResults(obj, { width, height }));
+  }
+  if (isWithFaceLandmarks(results)) {
+    const resizedDetection = results.detection.forSize(width, height);
+    const resizedLandmarks = results.unshiftedLandmarks.forSize(resizedDetection.box.width, resizedDetection.box.height);
+    return extendWithFaceLandmarks(extendWithFaceDetection(results, resizedDetection), resizedLandmarks);
+  }
+  if (isWithFaceDetection(results)) {
+    return extendWithFaceDetection(results, results.detection.forSize(width, height));
+  }
+  if (results instanceof FaceLandmarks || results instanceof FaceDetection) {
+    return results.forSize(width, height);
+  }
+  return results;
+}
+
+// src/index.ts
+var version2 = version;
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  AgeGenderNet,
+  BoundingBox,
+  Box,
+  ComposableTask,
+  ComputeAllFaceDescriptorsTask,
+  ComputeFaceDescriptorsTaskBase,
+  ComputeSingleFaceDescriptorTask,
+  DetectAllFaceLandmarksTask,
+  DetectAllFacesTask,
+  DetectFaceLandmarksTaskBase,
+  DetectFacesTaskBase,
+  DetectSingleFaceLandmarksTask,
+  DetectSingleFaceTask,
+  Dimensions,
+  FACE_EXPRESSION_LABELS,
+  FaceDetection,
+  FaceDetectionNet,
+  FaceExpressionNet,
+  FaceExpressions,
+  FaceLandmark68Net,
+  FaceLandmark68TinyNet,
+  FaceLandmarkNet,
+  FaceLandmarks,
+  FaceLandmarks5,
+  FaceLandmarks68,
+  FaceMatch,
+  FaceMatcher,
+  FaceRecognitionNet,
+  Gender,
+  LabeledBox,
+  LabeledFaceDescriptors,
+  NetInput,
+  NeuralNetwork,
+  ObjectDetection,
+  Point,
+  PredictedBox,
+  Rect,
+  SsdMobilenetv1,
+  SsdMobilenetv1Options,
+  TinyFaceDetector,
+  TinyFaceDetectorOptions,
+  TinyYolov2,
+  TinyYolov2Options,
+  allFaces,
+  allFacesSsdMobilenetv1,
+  allFacesTinyYolov2,
+  awaitMediaLoaded,
+  bufferToImage,
+  computeFaceDescriptor,
+  createCanvas,
+  createCanvasFromMedia,
+  createFaceDetectionNet,
+  createFaceRecognitionNet,
+  createSsdMobilenetv1,
+  createTinyFaceDetector,
+  createTinyYolov2,
+  detectAllFaces,
+  detectFaceLandmarks,
+  detectFaceLandmarksTiny,
+  detectLandmarks,
+  detectSingleFace,
+  draw,
+  env,
+  euclideanDistance,
+  extendWithAge,
+  extendWithFaceDescriptor,
+  extendWithFaceDetection,
+  extendWithFaceExpressions,
+  extendWithFaceLandmarks,
+  extendWithGender,
+  extractFaceTensors,
+  extractFaces,
+  fetchImage,
+  fetchJson,
+  fetchNetWeights,
+  fetchOrThrow,
+  fetchVideo,
+  getContext2dOrThrow,
+  getMediaDimensions,
+  imageTensorToCanvas,
+  imageToSquare,
+  inverseSigmoid,
+  iou,
+  isMediaElement,
+  isMediaLoaded,
+  isWithAge,
+  isWithFaceDetection,
+  isWithFaceExpressions,
+  isWithFaceLandmarks,
+  isWithGender,
+  loadAgeGenderModel,
+  loadFaceDetectionModel,
+  loadFaceExpressionModel,
+  loadFaceLandmarkModel,
+  loadFaceLandmarkTinyModel,
+  loadFaceRecognitionModel,
+  loadSsdMobilenetv1Model,
+  loadTinyFaceDetectorModel,
+  loadTinyYolov2Model,
+  loadWeightMap,
+  locateFaces,
+  matchDimensions,
+  minBbox,
+  nets,
+  nonMaxSuppression,
+  normalize,
+  padToSquare,
+  predictAgeAndGender,
+  recognizeFaceExpressions,
+  resizeResults,
+  resolveInput,
+  shuffleArray,
+  sigmoid,
+  ssdMobilenetv1,
+  tf,
+  tinyFaceDetector,
+  tinyYolov2,
+  toNetInput,
+  utils,
+  validateConfig,
+  version
+});
diff --git a/dist/tfjs.esm.js b/dist/tfjs.esm.js
index 4549cd2..c0a9ba4 100644
--- a/dist/tfjs.esm.js
+++ b/dist/tfjs.esm.js
@@ -4,65 +4,49522 @@
   author: <https://github.com/vladmandic>'
 */
 
-var DU=Object.create;var QS=Object.defineProperty;var RU=Object.getOwnPropertyDescriptor;var FU=Object.getOwnPropertyNames;var OU=Object.getPrototypeOf,PU=Object.prototype.hasOwnProperty;var Pg=(r=>typeof require!="undefined"?require:typeof Proxy!="undefined"?new Proxy(r,{get:(t,e)=>(typeof require!="undefined"?require:t)[e]}):r)(function(r){if(typeof require!="undefined")return require.apply(this,arguments);throw new Error('Dynamic require of "'+r+'" is not supported')});var gr=(r,t)=>()=>(t||r((t={exports:{}}).exports,t),t.exports),Wt=(r,t)=>{for(var e in t)QS(r,e,{get:t[e],enumerable:!0})},LU=(r,t,e,n)=>{if(t&&typeof t=="object"||typeof t=="function")for(let o of FU(t))!PU.call(r,o)&&o!==e&&QS(r,o,{get:()=>t[o],enumerable:!(n=RU(t,o))||n.enumerable});return r};var Tl=(r,t,e)=>(e=r!=null?DU(OU(r)):{},LU(t||!r||!r.__esModule?QS(e,"default",{value:r,enumerable:!0}):e,r));var T1=gr((Elt,N1)=>{N1.exports=Ue;var mo=null;try{mo=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(r){}function Ue(r,t,e){this.low=r|0,this.high=t|0,this.unsigned=!!e}Ue.prototype.__isLong__;Object.defineProperty(Ue.prototype,"__isLong__",{value:!0});function Fn(r){return(r&&r.__isLong__)===!0}Ue.isLong=Fn;var g1={},x1={};function zu(r,t){var e,n,o;return t?(r>>>=0,(o=0<=r&&r<256)&&(n=x1[r],n)?n:(e=He(r,(r|0)<0?-1:0,!0),o&&(x1[r]=e),e)):(r|=0,(o=-128<=r&&r<128)&&(n=g1[r],n)?n:(e=He(r,r<0?-1:0,!1),o&&(g1[r]=e),e))}Ue.fromInt=zu;function fo(r,t){if(isNaN(r))return t?Mu:ho;if(t){if(r<0)return Mu;if(r>=C1)return v1}else{if(r<=-b1)return Rn;if(r+1>=b1)return S1}return r<0?fo(-r,t).neg():He(r%Vp|0,r/Vp|0,t)}Ue.fromNumber=fo;function He(r,t,e){return new Ue(r,t,e)}Ue.fromBits=He;var Bg=Math.pow;function m0(r,t,e){if(r.length===0)throw Error("empty string");if(r==="NaN"||r==="Infinity"||r==="+Infinity"||r==="-Infinity")return ho;if(typeof t=="number"?(e=t,t=!1):t=!!t,e=e||10,e<2||36<e)throw RangeError("radix");var n;if((n=r.indexOf("-"))>0)throw Error("interior hyphen");if(n===0)return m0(r.substring(1),t,e).neg();for(var o=fo(Bg(e,8)),s=ho,i=0;i<r.length;i+=8){var a=Math.min(8,r.length-i),u=parseInt(r.substring(i,i+a),e);if(a<8){var l=fo(Bg(e,a));s=s.mul(l).add(fo(u))}else s=s.mul(o),s=s.add(fo(u))}return s.unsigned=t,s}Ue.fromString=m0;function Ls(r,t){return typeof r=="number"?fo(r,t):typeof r=="string"?m0(r,t):He(r.low,r.high,typeof t=="boolean"?t:r.unsigned)}Ue.fromValue=Ls;var y1=1<<16,s4=1<<24,Vp=y1*y1,C1=Vp*Vp,b1=C1/2,w1=zu(s4),ho=zu(0);Ue.ZERO=ho;var Mu=zu(0,!0);Ue.UZERO=Mu;var Bp=zu(1);Ue.ONE=Bp;var I1=zu(1,!0);Ue.UONE=I1;var p0=zu(-1);Ue.NEG_ONE=p0;var S1=He(-1,2147483647,!1);Ue.MAX_VALUE=S1;var v1=He(-1,-1,!0);Ue.MAX_UNSIGNED_VALUE=v1;var Rn=He(0,-2147483648,!1);Ue.MIN_VALUE=Rn;var gt=Ue.prototype;gt.toInt=function(){return this.unsigned?this.low>>>0:this.low};gt.toNumber=function(){return this.unsigned?(this.high>>>0)*Vp+(this.low>>>0):this.high*Vp+(this.low>>>0)};gt.toString=function(t){if(t=t||10,t<2||36<t)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative())if(this.eq(Rn)){var e=fo(t),n=this.div(e),o=n.mul(e).sub(this);return n.toString(t)+o.toInt().toString(t)}else return"-"+this.neg().toString(t);for(var s=fo(Bg(t,6),this.unsigned),i=this,a="";;){var u=i.div(s),l=i.sub(u.mul(s)).toInt()>>>0,c=l.toString(t);if(i=u,i.isZero())return c+a;for(;c.length<6;)c="0"+c;a=""+c+a}};gt.getHighBits=function(){return this.high};gt.getHighBitsUnsigned=function(){return this.high>>>0};gt.getLowBits=function(){return this.low};gt.getLowBitsUnsigned=function(){return this.low>>>0};gt.getNumBitsAbs=function(){if(this.isNegative())return this.eq(Rn)?64:this.neg().getNumBitsAbs();for(var t=this.high!=0?this.high:this.low,e=31;e>0&&(t&1<<e)==0;e--);return this.high!=0?e+33:e+1};gt.isZero=function(){return this.high===0&&this.low===0};gt.eqz=gt.isZero;gt.isNegative=function(){return!this.unsigned&&this.high<0};gt.isPositive=function(){return this.unsigned||this.high>=0};gt.isOdd=function(){return(this.low&1)===1};gt.isEven=function(){return(this.low&1)===0};gt.equals=function(t){return Fn(t)||(t=Ls(t)),this.unsigned!==t.unsigned&&this.high>>>31===1&&t.high>>>31===1?!1:this.high===t.high&&this.low===t.low};gt.eq=gt.equals;gt.notEquals=function(t){return!this.eq(t)};gt.neq=gt.notEquals;gt.ne=gt.notEquals;gt.lessThan=function(t){return this.comp(t)<0};gt.lt=gt.lessThan;gt.lessThanOrEqual=function(t){return this.comp(t)<=0};gt.lte=gt.lessThanOrEqual;gt.le=gt.lessThanOrEqual;gt.greaterThan=function(t){return this.comp(t)>0};gt.gt=gt.greaterThan;gt.greaterThanOrEqual=function(t){return this.comp(t)>=0};gt.gte=gt.greaterThanOrEqual;gt.ge=gt.greaterThanOrEqual;gt.compare=function(t){if(Fn(t)||(t=Ls(t)),this.eq(t))return 0;var e=this.isNegative(),n=t.isNegative();return e&&!n?-1:!e&&n?1:this.unsigned?t.high>>>0>this.high>>>0||t.high===this.high&&t.low>>>0>this.low>>>0?-1:1:this.sub(t).isNegative()?-1:1};gt.comp=gt.compare;gt.negate=function(){return!this.unsigned&&this.eq(Rn)?Rn:this.not().add(Bp)};gt.neg=gt.negate;gt.add=function(t){Fn(t)||(t=Ls(t));var e=this.high>>>16,n=this.high&65535,o=this.low>>>16,s=this.low&65535,i=t.high>>>16,a=t.high&65535,u=t.low>>>16,l=t.low&65535,c=0,p=0,m=0,f=0;return f+=s+l,m+=f>>>16,f&=65535,m+=o+u,p+=m>>>16,m&=65535,p+=n+a,c+=p>>>16,p&=65535,c+=e+i,c&=65535,He(m<<16|f,c<<16|p,this.unsigned)};gt.subtract=function(t){return Fn(t)||(t=Ls(t)),this.add(t.neg())};gt.sub=gt.subtract;gt.multiply=function(t){if(this.isZero())return ho;if(Fn(t)||(t=Ls(t)),mo){var e=mo.mul(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}if(t.isZero())return ho;if(this.eq(Rn))return t.isOdd()?Rn:ho;if(t.eq(Rn))return this.isOdd()?Rn:ho;if(this.isNegative())return t.isNegative()?this.neg().mul(t.neg()):this.neg().mul(t).neg();if(t.isNegative())return this.mul(t.neg()).neg();if(this.lt(w1)&&t.lt(w1))return fo(this.toNumber()*t.toNumber(),this.unsigned);var n=this.high>>>16,o=this.high&65535,s=this.low>>>16,i=this.low&65535,a=t.high>>>16,u=t.high&65535,l=t.low>>>16,c=t.low&65535,p=0,m=0,f=0,d=0;return d+=i*c,f+=d>>>16,d&=65535,f+=s*c,m+=f>>>16,f&=65535,f+=i*l,m+=f>>>16,f&=65535,m+=o*c,p+=m>>>16,m&=65535,m+=s*l,p+=m>>>16,m&=65535,m+=i*u,p+=m>>>16,m&=65535,p+=n*c+o*l+s*u+i*a,p&=65535,He(f<<16|d,p<<16|m,this.unsigned)};gt.mul=gt.multiply;gt.divide=function(t){if(Fn(t)||(t=Ls(t)),t.isZero())throw Error("division by zero");if(mo){if(!this.unsigned&&this.high===-2147483648&&t.low===-1&&t.high===-1)return this;var e=(this.unsigned?mo.div_u:mo.div_s)(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?Mu:ho;var n,o,s;if(this.unsigned){if(t.unsigned||(t=t.toUnsigned()),t.gt(this))return Mu;if(t.gt(this.shru(1)))return I1;s=Mu}else{if(this.eq(Rn)){if(t.eq(Bp)||t.eq(p0))return Rn;if(t.eq(Rn))return Bp;var i=this.shr(1);return n=i.div(t).shl(1),n.eq(ho)?t.isNegative()?Bp:p0:(o=this.sub(t.mul(n)),s=n.add(o.div(t)),s)}else if(t.eq(Rn))return this.unsigned?Mu:ho;if(this.isNegative())return t.isNegative()?this.neg().div(t.neg()):this.neg().div(t).neg();if(t.isNegative())return this.div(t.neg()).neg();s=ho}for(o=this;o.gte(t);){n=Math.max(1,Math.floor(o.toNumber()/t.toNumber()));for(var a=Math.ceil(Math.log(n)/Math.LN2),u=a<=48?1:Bg(2,a-48),l=fo(n),c=l.mul(t);c.isNegative()||c.gt(o);)n-=u,l=fo(n,this.unsigned),c=l.mul(t);l.isZero()&&(l=Bp),s=s.add(l),o=o.sub(c)}return s};gt.div=gt.divide;gt.modulo=function(t){if(Fn(t)||(t=Ls(t)),mo){var e=(this.unsigned?mo.rem_u:mo.rem_s)(this.low,this.high,t.low,t.high);return He(e,mo.get_high(),this.unsigned)}return this.sub(this.div(t).mul(t))};gt.mod=gt.modulo;gt.rem=gt.modulo;gt.not=function(){return He(~this.low,~this.high,this.unsigned)};gt.and=function(t){return Fn(t)||(t=Ls(t)),He(this.low&t.low,this.high&t.high,this.unsigned)};gt.or=function(t){return Fn(t)||(t=Ls(t)),He(this.low|t.low,this.high|t.high,this.unsigned)};gt.xor=function(t){return Fn(t)||(t=Ls(t)),He(this.low^t.low,this.high^t.high,this.unsigned)};gt.shiftLeft=function(t){return Fn(t)&&(t=t.toInt()),(t&=63)===0?this:t<32?He(this.low<<t,this.high<<t|this.low>>>32-t,this.unsigned):He(0,this.low<<t-32,this.unsigned)};gt.shl=gt.shiftLeft;gt.shiftRight=function(t){return Fn(t)&&(t=t.toInt()),(t&=63)===0?this:t<32?He(this.low>>>t|this.high<<32-t,this.high>>t,this.unsigned):He(this.high>>t-32,this.high>=0?0:-1,this.unsigned)};gt.shr=gt.shiftRight;gt.shiftRightUnsigned=function(t){if(Fn(t)&&(t=t.toInt()),t&=63,t===0)return this;var e=this.high;if(t<32){var n=this.low;return He(n>>>t|e<<32-t,e>>>t,this.unsigned)}else return t===32?He(e,0,this.unsigned):He(e>>>t-32,0,this.unsigned)};gt.shru=gt.shiftRightUnsigned;gt.shr_u=gt.shiftRightUnsigned;gt.toSigned=function(){return this.unsigned?He(this.low,this.high,!1):this};gt.toUnsigned=function(){return this.unsigned?this:He(this.low,this.high,!0)};gt.toBytes=function(t){return t?this.toBytesLE():this.toBytesBE()};gt.toBytesLE=function(){var t=this.high,e=this.low;return[e&255,e>>>8&255,e>>>16&255,e>>>24,t&255,t>>>8&255,t>>>16&255,t>>>24]};gt.toBytesBE=function(){var t=this.high,e=this.low;return[t>>>24,t>>>16&255,t>>>8&255,t&255,e>>>24,e>>>16&255,e>>>8&255,e&255]};Ue.fromBytes=function(t,e,n){return n?Ue.fromBytesLE(t,e):Ue.fromBytesBE(t,e)};Ue.fromBytesLE=function(t,e){return new Ue(t[0]|t[1]<<8|t[2]<<16|t[3]<<24,t[4]|t[5]<<8|t[6]<<16|t[7]<<24,e)};Ue.fromBytesBE=function(t,e){return new Ue(t[4]<<24|t[5]<<16|t[6]<<8|t[7],t[0]<<24|t[1]<<16|t[2]<<8|t[3],e)}});var iE=gr(()=>{});var aE=gr(()=>{});var u_=gr((l_,rv)=>{(function(r,t,e){function n(a){var u=this,l=i();u.next=function(){var c=2091639*u.s0+u.c*23283064365386963e-26;return u.s0=u.s1,u.s1=u.s2,u.s2=c-(u.c=c|0)},u.c=1,u.s0=l(" "),u.s1=l(" "),u.s2=l(" "),u.s0-=l(a),u.s0<0&&(u.s0+=1),u.s1-=l(a),u.s1<0&&(u.s1+=1),u.s2-=l(a),u.s2<0&&(u.s2+=1),l=null}function o(a,u){return u.c=a.c,u.s0=a.s0,u.s1=a.s1,u.s2=a.s2,u}function s(a,u){var l=new n(a),c=u&&u.state,p=l.next;return p.int32=function(){return l.next()*4294967296|0},p.double=function(){return p()+(p()*2097152|0)*11102230246251565e-32},p.quick=p,c&&(typeof c=="object"&&o(c,l),p.state=function(){return o(l,{})}),p}function i(){var a=4022871197,u=function(l){l=String(l);for(var c=0;c<l.length;c++){a+=l.charCodeAt(c);var p=.02519603282416938*a;a=p>>>0,p-=a,p*=a,a=p>>>0,p-=a,a+=p*4294967296}return(a>>>0)*23283064365386963e-26};return u}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.alea=s})(l_,typeof rv=="object"&&rv,typeof define=="function"&&define)});var p_=gr((c_,nv)=>{(function(r,t,e){function n(i){var a=this,u="";a.x=0,a.y=0,a.z=0,a.w=0,a.next=function(){var c=a.x^a.x<<11;return a.x=a.y,a.y=a.z,a.z=a.w,a.w^=a.w>>>19^c^c>>>8},i===(i|0)?a.x=i:u+=i;for(var l=0;l<u.length+64;l++)a.x^=u.charCodeAt(l)|0,a.next()}function o(i,a){return a.x=i.x,a.y=i.y,a.z=i.z,a.w=i.w,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l=="object"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xor128=s})(c_,typeof nv=="object"&&nv,typeof define=="function"&&define)});var f_=gr((m_,ov)=>{(function(r,t,e){function n(i){var a=this,u="";a.next=function(){var c=a.x^a.x>>>2;return a.x=a.y,a.y=a.z,a.z=a.w,a.w=a.v,(a.d=a.d+362437|0)+(a.v=a.v^a.v<<4^(c^c<<1))|0},a.x=0,a.y=0,a.z=0,a.w=0,a.v=0,i===(i|0)?a.x=i:u+=i;for(var l=0;l<u.length+64;l++)a.x^=u.charCodeAt(l)|0,l==u.length&&(a.d=a.x<<10^a.x>>>4),a.next()}function o(i,a){return a.x=i.x,a.y=i.y,a.z=i.z,a.w=i.w,a.v=i.v,a.d=i.d,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l=="object"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xorwow=s})(m_,typeof ov=="object"&&ov,typeof define=="function"&&define)});var h_=gr((d_,sv)=>{(function(r,t,e){function n(i){var a=this;a.next=function(){var l=a.x,c=a.i,p,m,f;return p=l[c],p^=p>>>7,m=p^p<<24,p=l[c+1&7],m^=p^p>>>10,p=l[c+3&7],m^=p^p>>>3,p=l[c+4&7],m^=p^p<<7,p=l[c+7&7],p=p^p<<13,m^=p^p<<9,l[c]=m,a.i=c+1&7,m};function u(l,c){var p,m,f=[];if(c===(c|0))m=f[0]=c;else for(c=""+c,p=0;p<c.length;++p)f[p&7]=f[p&7]<<15^c.charCodeAt(p)+f[p+1&7]<<13;for(;f.length<8;)f.push(0);for(p=0;p<8&&f[p]===0;++p);for(p==8?m=f[7]=-1:m=f[p],l.x=f,l.i=0,p=256;p>0;--p)l.next()}u(a,i)}function o(i,a){return a.x=i.x.slice(),a.i=i.i,a}function s(i,a){i==null&&(i=+new Date);var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(l.x&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xorshift7=s})(d_,typeof sv=="object"&&sv,typeof define=="function"&&define)});var x_=gr((g_,iv)=>{(function(r,t,e){function n(i){var a=this;a.next=function(){var l=a.w,c=a.X,p=a.i,m,f;return a.w=l=l+1640531527|0,f=c[p+34&127],m=c[p=p+1&127],f^=f<<13,m^=m<<17,f^=f>>>15,m^=m>>>12,f=c[p]=f^m,a.i=p,f+(l^l>>>16)|0};function u(l,c){var p,m,f,d,h,g=[],x=128;for(c===(c|0)?(m=c,c=null):(c=c+"\0",m=0,x=Math.max(x,c.length)),f=0,d=-32;d<x;++d)c&&(m^=c.charCodeAt((d+32)%c.length)),d===0&&(h=m),m^=m<<10,m^=m>>>15,m^=m<<4,m^=m>>>13,d>=0&&(h=h+1640531527|0,p=g[d&127]^=m+h,f=p==0?f+1:0);for(f>=128&&(g[(c&&c.length||0)&127]=-1),f=127,d=4*128;d>0;--d)m=g[f+34&127],p=g[f=f+1&127],m^=m<<13,p^=p<<17,m^=m>>>15,p^=p>>>12,g[f]=m^p;l.w=h,l.X=g,l.i=f}u(a,i)}function o(i,a){return a.i=i.i,a.w=i.w,a.X=i.X.slice(),a}function s(i,a){i==null&&(i=+new Date);var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(l.X&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.xor4096=s})(g_,typeof iv=="object"&&iv,typeof define=="function"&&define)});var b_=gr((y_,av)=>{(function(r,t,e){function n(i){var a=this,u="";a.next=function(){var c=a.b,p=a.c,m=a.d,f=a.a;return c=c<<25^c>>>7^p,p=p-m|0,m=m<<24^m>>>8^f,f=f-c|0,a.b=c=c<<20^c>>>12^p,a.c=p=p-m|0,a.d=m<<16^p>>>16^f,a.a=f-c|0},a.a=0,a.b=0,a.c=-1640531527,a.d=1367130551,i===Math.floor(i)?(a.a=i/4294967296|0,a.b=i|0):u+=i;for(var l=0;l<u.length+20;l++)a.b^=u.charCodeAt(l)|0,a.next()}function o(i,a){return a.a=i.a,a.b=i.b,a.c=i.c,a.d=i.d,a}function s(i,a){var u=new n(i),l=a&&a.state,c=function(){return(u.next()>>>0)/4294967296};return c.double=function(){do var p=u.next()>>>11,m=(u.next()>>>0)/4294967296,f=(p+m)/(1<<21);while(f===0);return f},c.int32=u.next,c.quick=c,l&&(typeof l=="object"&&o(l,u),c.state=function(){return o(u,{})}),c}t&&t.exports?t.exports=s:e&&e.amd?e(function(){return s}):this.tychei=s})(y_,typeof av=="object"&&av,typeof define=="function"&&define)});var w_=gr(()=>{});var I_=gr((C_,Yx)=>{(function(r,t,e){var n=256,o=6,s=52,i="random",a=e.pow(n,o),u=e.pow(2,s),l=u*2,c=n-1,p;function m(w,C,N){var _=[];C=C==!0?{entropy:!0}:C||{};var A=g(h(C.entropy?[w,b(t)]:w==null?x():w,3),_),$=new f(_),F=function(){for(var P=$.g(o),V=a,G=0;P<u;)P=(P+G)*n,V*=n,G=$.g(1);for(;P>=l;)P/=2,V/=2,G>>>=1;return(P+G)/V};return F.int32=function(){return $.g(4)|0},F.quick=function(){return $.g(4)/4294967296},F.double=F,g(b($.S),t),(C.pass||N||function(P,V,G,W){return W&&(W.S&&d(W,$),P.state=function(){return d($,{})}),G?(e[i]=P,V):P})(F,A,"global"in C?C.global:this==e,C.state)}function f(w){var C,N=w.length,_=this,A=0,$=_.i=_.j=0,F=_.S=[];for(N||(w=[N++]);A<n;)F[A]=A++;for(A=0;A<n;A++)F[A]=F[$=c&$+w[A%N]+(C=F[A])],F[$]=C;(_.g=function(P){for(var V,G=0,W=_.i,q=_.j,H=_.S;P--;)V=H[W=c&W+1],G=G*n+H[c&(H[W]=H[q=c&q+V])+(H[q]=V)];return _.i=W,_.j=q,G})(n)}function d(w,C){return C.i=w.i,C.j=w.j,C.S=w.S.slice(),C}function h(w,C){var N=[],_=typeof w,A;if(C&&_=="object")for(A in w)try{N.push(h(w[A],C-1))}catch($){}return N.length?N:_=="string"?w:w+"\0"}function g(w,C){for(var N=w+"",_,A=0;A<N.length;)C[c&A]=c&(_^=C[c&A]*19)+N.charCodeAt(A++);return b(C)}function x(){try{var w;return p&&(w=p.randomBytes)?w=w(n):(w=new Uint8Array(n),(r.crypto||r.msCrypto).getRandomValues(w)),b(w)}catch(_){var C=r.navigator,N=C&&C.plugins;return[+new Date,r,N,r.screen,b(t)]}}function b(w){return String.fromCharCode.apply(0,w)}if(g(e.random(),t),typeof Yx=="object"&&Yx.exports){Yx.exports=m;try{p=w_()}catch(w){}}else typeof define=="function"&&define.amd?define(function(){return m}):e["seed"+i]=m})(typeof self!="undefined"?self:C_,[],Math)});var fh=gr((fNt,S_)=>{var lj=u_(),uj=p_(),cj=f_(),pj=h_(),mj=x_(),fj=b_(),Ju=I_();Ju.alea=lj;Ju.xor128=uj;Ju.xorwow=cj;Ju.xorshift7=pj;Ju.xor4096=mj;Ju.tychei=fj;S_.exports=Ju});var gN=gr(()=>{});var Zb=gr(()=>{});var ig=gr(()=>{});var BW=gr(()=>{});var VW=gr(()=>{});var GW=gr(()=>{});var WW=gr((fI,Ok)=>{var Fk=(()=>{var r=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(r=r||__filename),function(t){t=t||{};function e(){return it.buffer!=jt&&qe(it.buffer),ke}function n(){return it.buffer!=jt&&qe(it.buffer),fe}function o(){return it.buffer!=jt&&qe(it.buffer),Ae}function s(){return it.buffer!=jt&&qe(it.buffer),_n}function i(){return it.buffer!=jt&&qe(it.buffer),or}function a(){return it.buffer!=jt&&qe(it.buffer),Hn}function u(){return it.buffer!=jt&&qe(it.buffer),Lr}var l=typeof t!="undefined"?t:{},c,p;l.ready=new Promise(function(L,U){c=L,p=U});var m;typeof process!="undefined"&&process.listeners&&(m={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var f=Object.assign({},l),d=[],h="./this.program",g=(L,U)=>{throw U},x=typeof window=="object",b=typeof importScripts=="function",w=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",C=l.ENVIRONMENT_IS_PTHREAD||!1,N="";function _(L){return l.locateFile?l.locateFile(L,N):N+L}var A,$,F,P;function V(L){if(L instanceof Ru)return;Y("exiting due to exception: "+L)}if(w){b?N=ig().dirname(N)+"/":N=__dirname+"/";var G,W;typeof Pg=="function"&&(G=Zb(),W=ig()),A=(U,ut)=>(U=W.normalize(U),G.readFileSync(U,ut?void 0:"utf8")),F=U=>{var ut=A(U,!0);return ut.buffer||(ut=new Uint8Array(ut)),ut},$=(U,ut,xt)=>{U=W.normalize(U),G.readFile(U,function(Dt,ie){Dt?xt(Dt):ut(ie.buffer)})},process.argv.length>1&&(h=process.argv[1].replace(/\\/g,"/")),d=process.argv.slice(2),process.on("uncaughtException",function(U){if(!(U instanceof Ru))throw U}),process.on("unhandledRejection",function(U){throw U}),g=(U,ut)=>{if(Mo())throw process.exitCode=U,ut;V(ut),process.exit(U)},l.inspect=function(){return"[Emscripten Module object]"};let L;try{L=BW()}catch(U){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),U}global.Worker=L.Worker}else(x||b)&&(b?N=self.location.href:typeof document!="undefined"&&document.currentScript&&(N=document.currentScript.src),typeof r!="undefined"&&r&&(N=r),N.indexOf("blob:")!==0?N=N.substr(0,N.replace(/[?#].*/,"").lastIndexOf("/")+1):N="",w||(A=L=>{var U=new XMLHttpRequest;return U.open("GET",L,!1),U.send(null),U.responseText},b&&(F=L=>{var U=new XMLHttpRequest;return U.open("GET",L,!1),U.responseType="arraybuffer",U.send(null),new Uint8Array(U.response)}),$=(L,U,ut)=>{var xt=new XMLHttpRequest;xt.open("GET",L,!0),xt.responseType="arraybuffer",xt.onload=()=>{if(xt.status==200||xt.status==0&&xt.response){U(xt.response);return}ut()},xt.onerror=ut,xt.send(null)}),P=L=>document.title=L);w&&typeof performance=="undefined"&&(global.performance=VW().performance);var q=console.log.bind(console),H=console.warn.bind(console);w&&(q=L=>G.writeSync(1,L+`
-`),H=L=>G.writeSync(2,L+`
-`));var j=l.print||q,Y=l.printErr||H;Object.assign(l,f),f=null,l.arguments&&(d=l.arguments),l.thisProgram&&(h=l.thisProgram),l.quit&&(g=l.quit);var Z=4,et=Atomics.load,rt=Atomics.store,ot=Atomics.compareExchange,at;l.wasmBinary&&(at=l.wasmBinary);var nt=l.noExitRuntime||!0;typeof WebAssembly!="object"&&Yc("no native wasm support detected");var it,dt,ht=!1,bt;function Et(L,U){L||Yc(U)}var At=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function Vt(L,U,ut){for(var xt=U+ut,Dt=U;L[Dt]&&!(Dt>=xt);)++Dt;if(Dt-U>16&&L.buffer&&At)return At.decode(L.buffer instanceof SharedArrayBuffer?L.slice(U,Dt):L.subarray(U,Dt));for(var ie="";U<Dt;){var Gt=L[U++];if(!(Gt&128)){ie+=String.fromCharCode(Gt);continue}var Xt=L[U++]&63;if((Gt&224)==192){ie+=String.fromCharCode((Gt&31)<<6|Xt);continue}var hr=L[U++]&63;if((Gt&240)==224?Gt=(Gt&15)<<12|Xt<<6|hr:Gt=(Gt&7)<<18|Xt<<12|hr<<6|L[U++]&63,Gt<65536)ie+=String.fromCharCode(Gt);else{var Xn=Gt-65536;ie+=String.fromCharCode(55296|Xn>>10,56320|Xn&1023)}}return ie}function Zt(L,U){return L?Vt(n(),L,U):""}function ce(L,U,ut,xt){if(!(xt>0))return 0;for(var Dt=ut,ie=ut+xt-1,Gt=0;Gt<L.length;++Gt){var Xt=L.charCodeAt(Gt);if(Xt>=55296&&Xt<=57343){var hr=L.charCodeAt(++Gt);Xt=65536+((Xt&1023)<<10)|hr&1023}if(Xt<=127){if(ut>=ie)break;U[ut++]=Xt}else if(Xt<=2047){if(ut+1>=ie)break;U[ut++]=192|Xt>>6,U[ut++]=128|Xt&63}else if(Xt<=65535){if(ut+2>=ie)break;U[ut++]=224|Xt>>12,U[ut++]=128|Xt>>6&63,U[ut++]=128|Xt&63}else{if(ut+3>=ie)break;U[ut++]=240|Xt>>18,U[ut++]=128|Xt>>12&63,U[ut++]=128|Xt>>6&63,U[ut++]=128|Xt&63}}return U[ut]=0,ut-Dt}function he(L,U,ut){return ce(L,n(),U,ut)}var jt,ke,fe,Ae,We,_n,or,Hn,Lr;C&&(jt=l.buffer);function qe(L){jt=L,l.HEAP8=ke=new Int8Array(L),l.HEAP16=Ae=new Int16Array(L),l.HEAP32=_n=new Int32Array(L),l.HEAPU8=fe=new Uint8Array(L),l.HEAPU16=We=new Uint16Array(L),l.HEAPU32=or=new Uint32Array(L),l.HEAPF32=Hn=new Float32Array(L),l.HEAPF64=Lr=new Float64Array(L)}var Mr=l.INITIAL_MEMORY||16777216;if(C)it=l.wasmMemory,jt=l.buffer;else if(l.wasmMemory)it=l.wasmMemory;else if(it=new WebAssembly.Memory({initial:Mr/65536,maximum:32768,shared:!0}),!(it.buffer instanceof SharedArrayBuffer))throw Y("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),w&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");it&&(jt=it.buffer),Mr=jt.byteLength,qe(jt);var zr,qn=[],Kn=[],Xr=[],Zi=!1;function Mo(){return nt}function Il(){if(l.preRun)for(typeof l.preRun=="function"&&(l.preRun=[l.preRun]);l.preRun.length;)Rd(l.preRun.shift());Md(qn)}function Tr(){Zi=!0,!C&&Md(Kn)}function Ji(){if(!C){if(l.postRun)for(typeof l.postRun=="function"&&(l.postRun=[l.postRun]);l.postRun.length;)Gk(l.postRun.shift());Md(Xr)}}function Rd(L){qn.unshift(L)}function Fd(L){Kn.unshift(L)}function Gk(L){Xr.unshift(L)}var Sl=0,Xc=null,Qi=null;function Wk(L){Sl++,l.monitorRunDependencies&&l.monitorRunDependencies(Sl)}function Uk(L){if(Sl--,l.monitorRunDependencies&&l.monitorRunDependencies(Sl),Sl==0&&(Xc!==null&&(clearInterval(Xc),Xc=null),Qi)){var U=Qi;Qi=null,U()}}function Yc(L){C?postMessage({cmd:"onAbort",arg:L}):l.onAbort&&l.onAbort(L),L="Aborted("+L+")",Y(L),ht=!0,bt=1,L+=". Build with -sASSERTIONS for more info.";var U=new WebAssembly.RuntimeError(L);throw p(U),U}var gI="data:application/octet-stream;base64,";function pg(L){return L.startsWith(gI)}function Od(L){return L.startsWith("file://")}var en;en="tfjs-backend-wasm-threaded-simd.wasm",pg(en)||(en=_(en));function mg(L){try{if(L==en&&at)return new Uint8Array(at);if(F)return F(L);throw"both async and sync fetching of the wasm failed"}catch(U){Yc(U)}}function xI(){if(!at&&(x||b)){if(typeof fetch=="function"&&!Od(en))return fetch(en,{credentials:"same-origin"}).then(function(L){if(!L.ok)throw"failed to load wasm binary file at '"+en+"'";return L.arrayBuffer()}).catch(function(){return mg(en)});if($)return new Promise(function(L,U){$(en,function(ut){L(new Uint8Array(ut))},U)})}return Promise.resolve().then(function(){return mg(en)})}function yI(){var L={env:vg,wasi_snapshot_preview1:vg};function U(Gt,Xt){var hr=Gt.exports;if(l.asm=hr,EI(l.asm._emscripten_tls_init),zr=l.asm.__indirect_function_table,Fd(l.asm.__wasm_call_ctors),dt=Xt,!C){var Xn=Kt.unusedWorkers.length;Kt.unusedWorkers.forEach(function(ea){Kt.loadWasmModuleToWorker(ea,function(){--Xn||Uk("wasm-instantiate")})})}}C||Wk("wasm-instantiate");function ut(Gt){U(Gt.instance,Gt.module)}function xt(Gt){return xI().then(function(Xt){return WebAssembly.instantiate(Xt,L)}).then(function(Xt){return Xt}).then(Gt,function(Xt){Y("failed to asynchronously prepare wasm: "+Xt),Yc(Xt)})}function Dt(){return!at&&typeof WebAssembly.instantiateStreaming=="function"&&!pg(en)&&!Od(en)&&!w&&typeof fetch=="function"?fetch(en,{credentials:"same-origin"}).then(function(Gt){var Xt=WebAssembly.instantiateStreaming(Gt,L);return Xt.then(ut,function(hr){return Y("wasm streaming compile failed: "+hr),Y("falling back to ArrayBuffer instantiation"),xt(ut)})}):xt(ut)}if(l.instantiateWasm)try{var ie=l.instantiateWasm(L,U);return ie}catch(Gt){Y("Module.instantiateWasm callback failed with error: "+Gt),p(Gt)}return Dt().catch(p),{}}var bI,Hk,wI={};function Ru(L){this.name="ExitStatus",this.message="Program terminated with exit("+L+")",this.status=L}function CI(L){var U=Kt.pthreads[L];delete Kt.pthreads[L],U.terminate(),YS(L),Kt.runningWorkers.splice(Kt.runningWorkers.indexOf(U),1),U.pthread_ptr=0}function II(L){var U=Kt.pthreads[L];U.postMessage({cmd:"cancel"})}function Pd(L){var U=Kt.pthreads[L];Et(U),Kt.returnWorkerToPool(U)}function fg(L){var U=Kt.getNewWorker();if(!U)return 6;Kt.runningWorkers.push(U),Kt.pthreads[L.pthread_ptr]=U,U.pthread_ptr=L.pthread_ptr;var ut={cmd:"run",start_routine:L.startRoutine,arg:L.arg,pthread_ptr:L.pthread_ptr};return U.runPthread=()=>{ut.time=performance.now(),U.postMessage(ut,L.transferList)},U.loaded&&(U.runPthread(),delete U.runPthread),0}var dg={varargs:void 0,get:function(){dg.varargs+=4;var L=s()[dg.varargs-4>>2];return L},getStr:function(L){var U=Zt(L);return U}};function Ld(L){if(C)return vl(1,1,L);bt=L,Mo()||(Kt.terminateAllThreads(),l.onExit&&l.onExit(L),ht=!0),g(L,new Ru(L))}function qk(L,U){if(bt=L,!U&&C)throw gg(L),"unwind";Ld(L)}var hg=qk;function SI(L){if(L instanceof Ru||L=="unwind")return bt;g(1,L)}var Kt={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){C?Kt.initWorker():Kt.initMainThread()},initMainThread:function(){for(var L=8;L--;)Kt.allocateUnusedWorker()},initWorker:function(){nt=!1},setExitStatus:function(L){bt=L},terminateAllThreads:function(){for(var L of Object.values(Kt.pthreads))Kt.returnWorkerToPool(L);for(var L of Kt.unusedWorkers)L.terminate();Kt.unusedWorkers=[]},returnWorkerToPool:function(L){var U=L.pthread_ptr;delete Kt.pthreads[U],Kt.unusedWorkers.push(L),Kt.runningWorkers.splice(Kt.runningWorkers.indexOf(L),1),L.pthread_ptr=0,YS(U)},receiveObjectTransfer:function(L){},threadInitTLS:function(){Kt.tlsInitFunctions.forEach(L=>L())},loadWasmModuleToWorker:function(L,U){L.onmessage=ut=>{var xt=ut.data,Dt=xt.cmd;if(L.pthread_ptr&&(Kt.currentProxiedOperationCallerThread=L.pthread_ptr),xt.targetThread&&xt.targetThread!=Ag()){var ie=Kt.pthreads[xt.targetThread];ie?ie.postMessage(xt,xt.transferList):Y('Internal error! Worker sent a message "'+Dt+'" to target pthread '+xt.targetThread+", but that thread no longer exists!"),Kt.currentProxiedOperationCallerThread=void 0;return}Dt==="processProxyingQueue"?zd(xt.queue):Dt==="spawnThread"?fg(xt):Dt==="cleanupThread"?Pd(xt.thread):Dt==="killThread"?CI(xt.thread):Dt==="cancelThread"?II(xt.thread):Dt==="loaded"?(L.loaded=!0,U&&U(L),L.runPthread&&(L.runPthread(),delete L.runPthread)):Dt==="print"?j("Thread "+xt.threadId+": "+xt.text):Dt==="printErr"?Y("Thread "+xt.threadId+": "+xt.text):Dt==="alert"?alert("Thread "+xt.threadId+": "+xt.text):xt.target==="setimmediate"?L.postMessage(xt):Dt==="onAbort"?l.onAbort&&l.onAbort(xt.arg):Dt&&Y("worker sent an unknown command "+Dt),Kt.currentProxiedOperationCallerThread=void 0},L.onerror=ut=>{var xt="worker sent an error!";throw Y(xt+" "+ut.filename+":"+ut.lineno+": "+ut.message),ut},w&&(L.on("message",function(ut){L.onmessage({data:ut})}),L.on("error",function(ut){L.onerror(ut)}),L.on("detachedExit",function(){})),L.postMessage({cmd:"load",urlOrBlob:l.mainScriptUrlOrBlob||r,wasmMemory:it,wasmModule:dt})},allocateUnusedWorker:function(){var L=_("tfjs-backend-wasm-threaded-simd.worker.js");Kt.unusedWorkers.push(new Worker(L))},getNewWorker:function(){return Kt.unusedWorkers.length==0&&(Kt.allocateUnusedWorker(),Kt.loadWasmModuleToWorker(Kt.unusedWorkers[0])),Kt.unusedWorkers.pop()}};l.PThread=Kt;function Md(L){for(;L.length>0;)L.shift()(l)}function vI(L){var U=ZS(),ut=L();return $g(U),ut}function Kk(L){return L}function jk(L){var U=/\b_Z[\w\d_]+/g;return L.replace(U,function(ut){var xt=ut;return ut===xt?ut:xt+" ["+ut+"]"})}function NI(){var L=Ag(),U=s()[L+44>>2],ut=s()[L+48>>2],xt=U-ut;e1(U,xt),$g(U)}l.establishStackSpace=NI;function gg(L){if(C)return vl(2,0,L);try{hg(L)}catch(U){SI(U)}}var Zc=[];function TI(L){var U=Zc[L];return U||(L>=Zc.length&&(Zc.length=L+1),Zc[L]=U=zr.get(L)),U}function kI(L,U){var ut=TI(L)(U);Mo()?Kt.setExitStatus(ut):t1(ut)}l.invokeEntryPoint=kI;function Xk(){var L=new Error;if(!L.stack){try{throw new Error}catch(U){L=U}if(!L.stack)return"(no stack trace available)"}return L.stack.toString()}function EI(L){Kt.tlsInitFunctions.push(L)}function _I(L,U){e().set(L,U)}function AI(L){Zk(L,!b,1,!x),Kt.threadInitTLS()}function $I(L){C?postMessage({cmd:"cleanupThread",thread:L}):Pd(L)}function xg(L,U,ut,xt){return C?vl(3,1,L,U,ut,xt):yg(L,U,ut,xt)}function yg(L,U,ut,xt){if(typeof SharedArrayBuffer=="undefined")return Y("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var Dt=[],ie=0;if(C&&(Dt.length===0||ie))return xg(L,U,ut,xt);if(ie)return ie;var Gt={startRoutine:ut,pthread_ptr:L,arg:xt,transferList:Dt};return C?(Gt.cmd="spawnThread",postMessage(Gt,Dt),0):fg(Gt)}function DI(){return 2097152}var RI=!0;function FI(){return RI}function zd(L){Atomics.store(s(),L>>2,1),Ag()&&Qk(L),Atomics.compareExchange(s(),L>>2,1,0)}l.executeNotifiedProxyingQueue=zd;function OI(L,U,ut,xt){if(L==U)setTimeout(()=>zd(xt));else if(C)postMessage({targetThread:L,cmd:"processProxyingQueue",queue:xt});else{var Dt=Kt.pthreads[L];if(!Dt)return;Dt.postMessage({cmd:"processProxyingQueue",queue:xt})}return 1}function PI(L,U,ut){return-1}function LI(){Yc("")}function Fu(L){Fu.shown||(Fu.shown={}),Fu.shown[L]||(Fu.shown[L]=1,w&&(L="warning: "+L),Y(L))}function MI(){w||b||Fu("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function zI(){return Date.now()}function bg(){return 2147483648}function BI(){return bg()}var Jc;w?Jc=()=>{var L=process.hrtime();return L[0]*1e3+L[1]/1e6}:C?Jc=()=>performance.now()-l.__performance_now_clock_drift:Jc=()=>performance.now();function VI(L,U,ut){n().copyWithin(L,U,U+ut)}function GI(){return w?GW().cpus().length:navigator.hardwareConcurrency}function vl(L,U){var ut=arguments.length-2,xt=arguments;return vI(()=>{for(var Dt=ut,ie=Dg(Dt*8),Gt=ie>>3,Xt=0;Xt<ut;Xt++){var hr=xt[2+Xt];u()[Gt+Xt]=hr}return Jk(L,Dt,ie,U)})}var Bd=[];function WI(L,U,ut){Bd.length=U;for(var xt=ut>>3,Dt=0;Dt<U;Dt++)Bd[Dt]=u()[xt+Dt];var ie=L<0,Gt=ie?wI[-L-1]:ZI[L];return Gt.apply(null,Bd)}function UI(L){try{return it.grow(L-jt.byteLength+65535>>>16),qe(it.buffer),1}catch(U){}}function HI(L){var U=n().length;if(L=L>>>0,L<=U)return!1;var ut=bg();if(L>ut)return!1;let xt=(hr,Xn)=>hr+(Xn-hr%Xn)%Xn;for(var Dt=1;Dt<=4;Dt*=2){var ie=U*(1+.2/Dt);ie=Math.min(ie,L+100663296);var Gt=Math.min(ut,xt(Math.max(L,ie),65536)),Xt=UI(Gt);if(Xt)return!0}return!1}function qI(){throw"unwind"}function wg(L){return C?vl(4,1,L):52}function Cg(L,U,ut,xt,Dt){return C?vl(5,1,L,U,ut,xt,Dt):70}var KI=[null,[],[]];function jI(L,U){var ut=KI[L];U===0||U===10?((L===1?j:Y)(Vt(ut,0)),ut.length=0):ut.push(U)}function Ig(L,U,ut,xt){if(C)return vl(6,1,L,U,ut,xt);for(var Dt=0,ie=0;ie<ut;ie++){var Gt=i()[U>>2],Xt=i()[U+4>>2];U+=8;for(var hr=0;hr<Xt;hr++)jI(L,n()[Gt+hr]);Dt+=Xt}return i()[xt>>2]=Dt,0}function Sg(L){var U=l["_"+L];return U}function XI(L,U,ut,xt,Dt){var ie={string:An=>{var rp=0;if(An!=null&&An!==0){var o1=(An.length<<2)+1;rp=Dg(o1),he(An,rp,o1)}return rp},array:An=>{var rp=Dg(An.length);return _I(An,rp),rp}};function Gt(An){return U==="string"?Zt(An):U==="boolean"?Boolean(An):An}var Xt=Sg(L),hr=[],Xn=0;if(xt)for(var ea=0;ea<xt.length;ea++){var n1=ie[ut[ea]];n1?(Xn===0&&(Xn=ZS()),hr[ea]=n1(xt[ea])):hr[ea]=xt[ea]}var JS=Xt.apply(null,hr);function $U(An){return Xn!==0&&$g(Xn),Gt(An)}return JS=$U(JS),JS}function YI(L,U,ut,xt){ut=ut||[];var Dt=ut.every(Gt=>Gt==="number"||Gt==="boolean"),ie=U!=="string";return ie&&Dt&&!xt?Sg(L):function(){return XI(L,U,ut,arguments,xt)}}Kt.init();var ZI=[null,Ld,gg,xg,wg,Cg,Ig],vg={__emscripten_init_main_thread_js:AI,__emscripten_thread_cleanup:$I,__pthread_create_js:yg,_emscripten_default_pthread_stack_size:DI,_emscripten_get_now_is_monotonic:FI,_emscripten_notify_task_queue:OI,_emscripten_set_offscreencanvas_size:PI,abort:LI,emscripten_check_blocking_allowed:MI,emscripten_date_now:zI,emscripten_get_heap_max:BI,emscripten_get_now:Jc,emscripten_memcpy_big:VI,emscripten_num_logical_cores:GI,emscripten_receive_on_main_thread_js:WI,emscripten_resize_heap:HI,emscripten_unwind_to_js_event_loop:qI,exit:hg,fd_close:wg,fd_seek:Cg,fd_write:Ig,memory:it||l.wasmMemory},Yk=yI(),JI=l.___wasm_call_ctors=function(){return(JI=l.___wasm_call_ctors=l.asm.__wasm_call_ctors).apply(null,arguments)},QI=l._init=function(){return(QI=l._init=l.asm.init).apply(null,arguments)},tS=l._init_with_threads_count=function(){return(tS=l._init_with_threads_count=l.asm.init_with_threads_count).apply(null,arguments)},eS=l._get_threads_count=function(){return(eS=l._get_threads_count=l.asm.get_threads_count).apply(null,arguments)},rS=l._register_tensor=function(){return(rS=l._register_tensor=l.asm.register_tensor).apply(null,arguments)},nS=l._dispose_data=function(){return(nS=l._dispose_data=l.asm.dispose_data).apply(null,arguments)},oS=l._dispose=function(){return(oS=l._dispose=l.asm.dispose).apply(null,arguments)},sS=l._Abs=function(){return(sS=l._Abs=l.asm.Abs).apply(null,arguments)},iS=l._Add=function(){return(iS=l._Add=l.asm.Add).apply(null,arguments)},aS=l._AddN=function(){return(aS=l._AddN=l.asm.AddN).apply(null,arguments)},lS=l._All=function(){return(lS=l._All=l.asm.All).apply(null,arguments)},uS=l._Any=function(){return(uS=l._Any=l.asm.Any).apply(null,arguments)},cS=l._ArgMax=function(){return(cS=l._ArgMax=l.asm.ArgMax).apply(null,arguments)},pS=l._AvgPool=function(){return(pS=l._AvgPool=l.asm.AvgPool).apply(null,arguments)},mS=l._BatchMatMul=function(){return(mS=l._BatchMatMul=l.asm.BatchMatMul).apply(null,arguments)},fS=l._Ceil=function(){return(fS=l._Ceil=l.asm.Ceil).apply(null,arguments)},dS=l._ClipByValue=function(){return(dS=l._ClipByValue=l.asm.ClipByValue).apply(null,arguments)},hS=l._Conv2D=function(){return(hS=l._Conv2D=l.asm.Conv2D).apply(null,arguments)},gS=l._Conv2DBackpropInput=function(){return(gS=l._Conv2DBackpropInput=l.asm.Conv2DBackpropInput).apply(null,arguments)},xS=l._Cos=function(){return(xS=l._Cos=l.asm.Cos).apply(null,arguments)},yS=l._Cosh=function(){return(yS=l._Cosh=l.asm.Cosh).apply(null,arguments)},bS=l._CropAndResize=function(){return(bS=l._CropAndResize=l.asm.CropAndResize).apply(null,arguments)},wS=l._Cumprod=function(){return(wS=l._Cumprod=l.asm.Cumprod).apply(null,arguments)},CS=l._Cumsum=function(){return(CS=l._Cumsum=l.asm.Cumsum).apply(null,arguments)},IS=l._DepthToSpace=function(){return(IS=l._DepthToSpace=l.asm.DepthToSpace).apply(null,arguments)},SS=l._DepthwiseConv2dNative=function(){return(SS=l._DepthwiseConv2dNative=l.asm.DepthwiseConv2dNative).apply(null,arguments)},vS=l._Elu=function(){return(vS=l._Elu=l.asm.Elu).apply(null,arguments)},NS=l._Equal=function(){return(NS=l._Equal=l.asm.Equal).apply(null,arguments)},TS=l._Exp=function(){return(TS=l._Exp=l.asm.Exp).apply(null,arguments)},kS=l._FlipLeftRight=function(){return(kS=l._FlipLeftRight=l.asm.FlipLeftRight).apply(null,arguments)},ES=l._Floor=function(){return(ES=l._Floor=l.asm.Floor).apply(null,arguments)},_S=l._FloorDiv=function(){return(_S=l._FloorDiv=l.asm.FloorDiv).apply(null,arguments)},AS=l._FusedBatchNorm=function(){return(AS=l._FusedBatchNorm=l.asm.FusedBatchNorm).apply(null,arguments)},$S=l._FusedConv2D=function(){return($S=l._FusedConv2D=l.asm.FusedConv2D).apply(null,arguments)},DS=l._FusedDepthwiseConv2D=function(){return(DS=l._FusedDepthwiseConv2D=l.asm.FusedDepthwiseConv2D).apply(null,arguments)},RS=l._Gather=function(){return(RS=l._Gather=l.asm.Gather).apply(null,arguments)},FS=l._GatherNd=function(){return(FS=l._GatherNd=l.asm.GatherNd).apply(null,arguments)},OS=l._Greater=function(){return(OS=l._Greater=l.asm.Greater).apply(null,arguments)},PS=l._GreaterEqual=function(){return(PS=l._GreaterEqual=l.asm.GreaterEqual).apply(null,arguments)},LS=l._LeakyRelu=function(){return(LS=l._LeakyRelu=l.asm.LeakyRelu).apply(null,arguments)},MS=l._Less=function(){return(MS=l._Less=l.asm.Less).apply(null,arguments)},zS=l._LessEqual=function(){return(zS=l._LessEqual=l.asm.LessEqual).apply(null,arguments)},BS=l._Log=function(){return(BS=l._Log=l.asm.Log).apply(null,arguments)},VS=l._LogicalAnd=function(){return(VS=l._LogicalAnd=l.asm.LogicalAnd).apply(null,arguments)},GS=l._LogicalNot=function(){return(GS=l._LogicalNot=l.asm.LogicalNot).apply(null,arguments)},WS=l._LogicalOr=function(){return(WS=l._LogicalOr=l.asm.LogicalOr).apply(null,arguments)},US=l._LogicalXor=function(){return(US=l._LogicalXor=l.asm.LogicalXor).apply(null,arguments)},HS=l._Max=function(){return(HS=l._Max=l.asm.Max).apply(null,arguments)},Ng=l._MaxPool=function(){return(Ng=l._MaxPool=l.asm.MaxPool).apply(null,arguments)},Tg=l._Maximum=function(){return(Tg=l._Maximum=l.asm.Maximum).apply(null,arguments)},Vd=l._Mean=function(){return(Vd=l._Mean=l.asm.Mean).apply(null,arguments)},qS=l._Min=function(){return(qS=l._Min=l.asm.Min).apply(null,arguments)},KS=l._Minimum=function(){return(KS=l._Minimum=l.asm.Minimum).apply(null,arguments)},Qc=l._MirrorPad=function(){return(Qc=l._MirrorPad=l.asm.MirrorPad).apply(null,arguments)},kg=l._Multiply=function(){return(kg=l._Multiply=l.asm.Multiply).apply(null,arguments)},tp=l._Neg=function(){return(tp=l._Neg=l.asm.Neg).apply(null,arguments)},ep=l._NonMaxSuppressionV3=function(){return(ep=l._NonMaxSuppressionV3=l.asm.NonMaxSuppressionV3).apply(null,arguments)},jS=l._NonMaxSuppressionV4=function(){return(jS=l._NonMaxSuppressionV4=l.asm.NonMaxSuppressionV4).apply(null,arguments)},K=l._NonMaxSuppressionV5=function(){return(K=l._NonMaxSuppressionV5=l.asm.NonMaxSuppressionV5).apply(null,arguments)},lt=l._NotEqual=function(){return(lt=l._NotEqual=l.asm.NotEqual).apply(null,arguments)},_t=l._OneHot=function(){return(_t=l._OneHot=l.asm.OneHot).apply(null,arguments)},re=l._PadV2=function(){return(re=l._PadV2=l.asm.PadV2).apply(null,arguments)},Ke=l._Pow=function(){return(Ke=l._Pow=l.asm.Pow).apply(null,arguments)},je=l._Prelu=function(){return(je=l._Prelu=l.asm.Prelu).apply(null,arguments)},ee=l._Prod=function(){return(ee=l._Prod=l.asm.Prod).apply(null,arguments)},te=l._RealDiv=function(){return(te=l._RealDiv=l.asm.RealDiv).apply(null,arguments)},dr=l._Relu=function(){return(dr=l._Relu=l.asm.Relu).apply(null,arguments)},jn=l._Relu6=function(){return(jn=l._Relu6=l.asm.Relu6).apply(null,arguments)},ta=l._ResizeBilinear=function(){return(ta=l._ResizeBilinear=l.asm.ResizeBilinear).apply(null,arguments)},Eg=l._ResizeNearestNeighbor=function(){return(Eg=l._ResizeNearestNeighbor=l.asm.ResizeNearestNeighbor).apply(null,arguments)},Gd=l._Reverse=function(){return(Gd=l._Reverse=l.asm.Reverse).apply(null,arguments)},XS=l._RotateWithOffset=function(){return(XS=l._RotateWithOffset=l.asm.RotateWithOffset).apply(null,arguments)},rn=l._Round=function(){return(rn=l._Round=l.asm.Round).apply(null,arguments)},Nl=l._Rsqrt=function(){return(Nl=l._Rsqrt=l.asm.Rsqrt).apply(null,arguments)},_g=l._ScatterNd=function(){return(_g=l._ScatterNd=l.asm.ScatterNd).apply(null,arguments)},QW=l._SelectV2=function(){return(QW=l._SelectV2=l.asm.SelectV2).apply(null,arguments)},tU=l._Sigmoid=function(){return(tU=l._Sigmoid=l.asm.Sigmoid).apply(null,arguments)},eU=l._Sin=function(){return(eU=l._Sin=l.asm.Sin).apply(null,arguments)},rU=l._Softmax=function(){return(rU=l._Softmax=l.asm.Softmax).apply(null,arguments)},nU=l._SparseFillEmptyRows=function(){return(nU=l._SparseFillEmptyRows=l.asm.SparseFillEmptyRows).apply(null,arguments)},oU=l._SparseReshape=function(){return(oU=l._SparseReshape=l.asm.SparseReshape).apply(null,arguments)},sU=l._SparseSegmentReduction=function(){return(sU=l._SparseSegmentReduction=l.asm.SparseSegmentReduction).apply(null,arguments)},iU=l._Sqrt=function(){return(iU=l._Sqrt=l.asm.Sqrt).apply(null,arguments)},aU=l._Square=function(){return(aU=l._Square=l.asm.Square).apply(null,arguments)},lU=l._SquaredDifference=function(){return(lU=l._SquaredDifference=l.asm.SquaredDifference).apply(null,arguments)},uU=l._Step=function(){return(uU=l._Step=l.asm.Step).apply(null,arguments)},cU=l._StridedSlice=function(){return(cU=l._StridedSlice=l.asm.StridedSlice).apply(null,arguments)},pU=l._Sub=function(){return(pU=l._Sub=l.asm.Sub).apply(null,arguments)},mU=l._Sum=function(){return(mU=l._Sum=l.asm.Sum).apply(null,arguments)},fU=l._Tan=function(){return(fU=l._Tan=l.asm.Tan).apply(null,arguments)},dU=l._Tanh=function(){return(dU=l._Tanh=l.asm.Tanh).apply(null,arguments)},hU=l._Tile=function(){return(hU=l._Tile=l.asm.Tile).apply(null,arguments)},gU=l._TopK=function(){return(gU=l._TopK=l.asm.TopK).apply(null,arguments)},xU=l._Transform=function(){return(xU=l._Transform=l.asm.Transform).apply(null,arguments)},yU=l._Transpose=function(){return(yU=l._Transpose=l.asm.Transpose).apply(null,arguments)},bU=l.__FusedMatMul=function(){return(bU=l.__FusedMatMul=l.asm._FusedMatMul).apply(null,arguments)},wU=l._malloc=function(){return(wU=l._malloc=l.asm.malloc).apply(null,arguments)},CU=l._free=function(){return(CU=l._free=l.asm.free).apply(null,arguments)},IU=l.__emscripten_tls_init=function(){return(IU=l.__emscripten_tls_init=l.asm._emscripten_tls_init).apply(null,arguments)},Ag=l._pthread_self=function(){return(Ag=l._pthread_self=l.asm.pthread_self).apply(null,arguments)},SU=l.___errno_location=function(){return(SU=l.___errno_location=l.asm.__errno_location).apply(null,arguments)},Zk=l.__emscripten_thread_init=function(){return(Zk=l.__emscripten_thread_init=l.asm._emscripten_thread_init).apply(null,arguments)},vU=l.__emscripten_thread_crashed=function(){return(vU=l.__emscripten_thread_crashed=l.asm._emscripten_thread_crashed).apply(null,arguments)},NU=l._emscripten_main_thread_process_queued_calls=function(){return(NU=l._emscripten_main_thread_process_queued_calls=l.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},TU=l._emscripten_main_browser_thread_id=function(){return(TU=l._emscripten_main_browser_thread_id=l.asm.emscripten_main_browser_thread_id).apply(null,arguments)},Jk=l._emscripten_run_in_main_runtime_thread_js=function(){return(Jk=l._emscripten_run_in_main_runtime_thread_js=l.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},kU=l._emscripten_dispatch_to_thread_=function(){return(kU=l._emscripten_dispatch_to_thread_=l.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},Qk=l.__emscripten_proxy_execute_task_queue=function(){return(Qk=l.__emscripten_proxy_execute_task_queue=l.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)},YS=l.__emscripten_thread_free_data=function(){return(YS=l.__emscripten_thread_free_data=l.asm._emscripten_thread_free_data).apply(null,arguments)},t1=l.__emscripten_thread_exit=function(){return(t1=l.__emscripten_thread_exit=l.asm._emscripten_thread_exit).apply(null,arguments)},e1=l._emscripten_stack_set_limits=function(){return(e1=l._emscripten_stack_set_limits=l.asm.emscripten_stack_set_limits).apply(null,arguments)},ZS=l.stackSave=function(){return(ZS=l.stackSave=l.asm.stackSave).apply(null,arguments)},$g=l.stackRestore=function(){return($g=l.stackRestore=l.asm.stackRestore).apply(null,arguments)},Dg=l.stackAlloc=function(){return(Dg=l.stackAlloc=l.asm.stackAlloc).apply(null,arguments)},EU=l.dynCall_iijjiiii=function(){return(EU=l.dynCall_iijjiiii=l.asm.dynCall_iijjiiii).apply(null,arguments)},_U=l.dynCall_jiji=function(){return(_U=l.dynCall_jiji=l.asm.dynCall_jiji).apply(null,arguments)};l.keepRuntimeAlive=Mo,l.wasmMemory=it,l.cwrap=YI,l.ExitStatus=Ru,l.PThread=Kt;var Rg;Qi=function L(){Rg||r1(),Rg||(Qi=L)};function r1(L){if(L=L||d,Sl>0)return;if(C){c(l),Tr(),postMessage({cmd:"loaded"});return}if(Il(),Sl>0)return;function U(){Rg||(Rg=!0,l.calledRun=!0,!ht&&(Tr(),c(l),l.onRuntimeInitialized&&l.onRuntimeInitialized(),Ji()))}l.setStatus?(l.setStatus("Running..."),setTimeout(function(){setTimeout(function(){l.setStatus("")},1),U()},1)):U()}if(l.preInit)for(typeof l.preInit=="function"&&(l.preInit=[l.preInit]);l.preInit.length>0;)l.preInit.pop()();r1();var Fg;m&&(Fg={uncaughtException:process.listeners("uncaughtException").filter(function(L){return!m.uncaughtException.indexOf(L)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(L){return!m.unhandledRejection.indexOf(L)>-1})});var Og;if(typeof WasmBackendModule!="undefined")Og=WasmBackendModule;else if(typeof t!="undefined")Og=t;else throw new Error("Could not find wasm module in post.js");if(Fg){var AU=Og._dispose;Og._dispose=function(){AU(),Fg.uncaughtException.forEach(function(L){process.removeListener("uncaughtException",L)}),Fg.unhandledRejection.forEach(function(L){process.removeListener("unhandledRejection",L)})}}return t.ready}})();typeof fI=="object"&&typeof Ok=="object"?Ok.exports=Fk:typeof define=="function"&&define.amd?define([],function(){return Fk}):typeof fI=="object"&&(fI.WasmBackendModuleThreadedSimd=Fk)});var HW=gr((G5e,UW)=>{UW.exports.wasmWorkerContents=`"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
-");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`});var qW=gr((dI,Lk)=>{var Pk=(()=>{var r=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(r=r||__filename),function(t){t=t||{};var e=typeof t!="undefined"?t:{},n,o;e.ready=new Promise(function(K,lt){n=K,o=lt});var s;typeof process!="undefined"&&process.listeners&&(s={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var i=Object.assign({},e),a=[],u="./this.program",l=(K,lt)=>{throw lt},c=typeof window=="object",p=typeof importScripts=="function",m=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",f="";function d(K){return e.locateFile?e.locateFile(K,f):f+K}var h,g,x,b;function w(K){if(K instanceof Xc)return;A("exiting due to exception: "+K)}if(m){p?f=ig().dirname(f)+"/":f=__dirname+"/";var C,N;typeof Pg=="function"&&(C=Zb(),N=ig()),h=(K,lt)=>(K=N.normalize(K),C.readFileSync(K,lt?void 0:"utf8")),x=K=>{var lt=h(K,!0);return lt.buffer||(lt=new Uint8Array(lt)),lt},g=(K,lt,_t)=>{K=N.normalize(K),C.readFile(K,function(re,Ke){re?_t(re):lt(Ke.buffer)})},process.argv.length>1&&(u=process.argv[1].replace(/\\/g,"/")),a=process.argv.slice(2),process.on("uncaughtException",function(K){if(!(K instanceof Xc))throw K}),process.on("unhandledRejection",function(K){throw K}),l=(K,lt)=>{if(fe())throw process.exitCode=K,lt;w(lt),process.exit(K)},e.inspect=function(){return"[Emscripten Module object]"}}else(c||p)&&(p?f=self.location.href:typeof document!="undefined"&&document.currentScript&&(f=document.currentScript.src),r&&(f=r),f.indexOf("blob:")!==0?f=f.substr(0,f.replace(/[?#].*/,"").lastIndexOf("/")+1):f="",h=K=>{var lt=new XMLHttpRequest;return lt.open("GET",K,!1),lt.send(null),lt.responseText},p&&(x=K=>{var lt=new XMLHttpRequest;return lt.open("GET",K,!1),lt.responseType="arraybuffer",lt.send(null),new Uint8Array(lt.response)}),g=(K,lt,_t)=>{var re=new XMLHttpRequest;re.open("GET",K,!0),re.responseType="arraybuffer",re.onload=()=>{if(re.status==200||re.status==0&&re.response){lt(re.response);return}_t()},re.onerror=_t,re.send(null)},b=K=>document.title=K);var _=e.print||console.log.bind(console),A=e.printErr||console.warn.bind(console);Object.assign(e,i),i=null,e.arguments&&(a=e.arguments),e.thisProgram&&(u=e.thisProgram),e.quit&&(l=e.quit);var $=4,F;e.wasmBinary&&(F=e.wasmBinary);var P=e.noExitRuntime||!0;typeof WebAssembly!="object"&&Xr("no native wasm support detected");var V,G=!1,W;function q(K,lt){K||Xr(lt)}var H=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function j(K,lt,_t){for(var re=lt+_t,Ke=lt;K[Ke]&&!(Ke>=re);)++Ke;if(Ke-lt>16&&K.buffer&&H)return H.decode(K.subarray(lt,Ke));for(var je="";lt<Ke;){var ee=K[lt++];if(!(ee&128)){je+=String.fromCharCode(ee);continue}var te=K[lt++]&63;if((ee&224)==192){je+=String.fromCharCode((ee&31)<<6|te);continue}var dr=K[lt++]&63;if((ee&240)==224?ee=(ee&15)<<12|te<<6|dr:ee=(ee&7)<<18|te<<12|dr<<6|K[lt++]&63,ee<65536)je+=String.fromCharCode(ee);else{var jn=ee-65536;je+=String.fromCharCode(55296|jn>>10,56320|jn&1023)}}return je}function Y(K,lt){return K?j(at,K,lt):""}function Z(K,lt,_t,re){if(!(re>0))return 0;for(var Ke=_t,je=_t+re-1,ee=0;ee<K.length;++ee){var te=K.charCodeAt(ee);if(te>=55296&&te<=57343){var dr=K.charCodeAt(++ee);te=65536+((te&1023)<<10)|dr&1023}if(te<=127){if(_t>=je)break;lt[_t++]=te}else if(te<=2047){if(_t+1>=je)break;lt[_t++]=192|te>>6,lt[_t++]=128|te&63}else if(te<=65535){if(_t+2>=je)break;lt[_t++]=224|te>>12,lt[_t++]=128|te>>6&63,lt[_t++]=128|te&63}else{if(_t+3>=je)break;lt[_t++]=240|te>>18,lt[_t++]=128|te>>12&63,lt[_t++]=128|te>>6&63,lt[_t++]=128|te&63}}return lt[_t]=0,_t-Ke}function et(K,lt,_t){return Z(K,at,lt,_t)}var rt,ot,at,nt,it,dt,ht,bt,Et;function At(K){rt=K,e.HEAP8=ot=new Int8Array(K),e.HEAP16=nt=new Int16Array(K),e.HEAP32=dt=new Int32Array(K),e.HEAPU8=at=new Uint8Array(K),e.HEAPU16=it=new Uint16Array(K),e.HEAPU32=ht=new Uint32Array(K),e.HEAPF32=bt=new Float32Array(K),e.HEAPF64=Et=new Float64Array(K)}var Vt=e.INITIAL_MEMORY||16777216,Zt,ce=[],he=[],jt=[],ke=!1;function fe(){return P}function Ae(){if(e.preRun)for(typeof e.preRun=="function"&&(e.preRun=[e.preRun]);e.preRun.length;)or(e.preRun.shift());Qi(ce)}function We(){ke=!0,Qi(he)}function _n(){if(e.postRun)for(typeof e.postRun=="function"&&(e.postRun=[e.postRun]);e.postRun.length;)Lr(e.postRun.shift());Qi(jt)}function or(K){ce.unshift(K)}function Hn(K){he.unshift(K)}function Lr(K){jt.unshift(K)}var qe=0,Mr=null,zr=null;function qn(K){qe++,e.monitorRunDependencies&&e.monitorRunDependencies(qe)}function Kn(K){if(qe--,e.monitorRunDependencies&&e.monitorRunDependencies(qe),qe==0&&(Mr!==null&&(clearInterval(Mr),Mr=null),zr)){var lt=zr;zr=null,lt()}}function Xr(K){e.onAbort&&e.onAbort(K),K="Aborted("+K+")",A(K),G=!0,W=1,K+=". Build with -sASSERTIONS for more info.";var lt=new WebAssembly.RuntimeError(K);throw o(lt),lt}var Zi="data:application/octet-stream;base64,";function Mo(K){return K.startsWith(Zi)}function Il(K){return K.startsWith("file://")}var Tr;Tr="tfjs-backend-wasm.wasm",Mo(Tr)||(Tr=d(Tr));function Ji(K){try{if(K==Tr&&F)return new Uint8Array(F);if(x)return x(K);throw"both async and sync fetching of the wasm failed"}catch(lt){Xr(lt)}}function Rd(){if(!F&&(c||p)){if(typeof fetch=="function"&&!Il(Tr))return fetch(Tr,{credentials:"same-origin"}).then(function(K){if(!K.ok)throw"failed to load wasm binary file at '"+Tr+"'";return K.arrayBuffer()}).catch(function(){return Ji(Tr)});if(g)return new Promise(function(K,lt){g(Tr,function(_t){K(new Uint8Array(_t))},lt)})}return Promise.resolve().then(function(){return Ji(Tr)})}function Fd(){var K={env:Ld,wasi_snapshot_preview1:Ld};function lt(ee,te){var dr=ee.exports;e.asm=dr,V=e.asm.memory,At(V.buffer),Zt=e.asm.__indirect_function_table,Hn(e.asm.__wasm_call_ctors),Kn("wasm-instantiate")}qn("wasm-instantiate");function _t(ee){lt(ee.instance)}function re(ee){return Rd().then(function(te){return WebAssembly.instantiate(te,K)}).then(function(te){return te}).then(ee,function(te){A("failed to asynchronously prepare wasm: "+te),Xr(te)})}function Ke(){return!F&&typeof WebAssembly.instantiateStreaming=="function"&&!Mo(Tr)&&!Il(Tr)&&!m&&typeof fetch=="function"?fetch(Tr,{credentials:"same-origin"}).then(function(ee){var te=WebAssembly.instantiateStreaming(ee,K);return te.then(_t,function(dr){return A("wasm streaming compile failed: "+dr),A("falling back to ArrayBuffer instantiation"),re(_t)})}):re(_t)}if(e.instantiateWasm)try{var je=e.instantiateWasm(K,lt);return je}catch(ee){A("Module.instantiateWasm callback failed with error: "+ee),o(ee)}return Ke().catch(o),{}}var Gk,Sl;function Xc(K){this.name="ExitStatus",this.message="Program terminated with exit("+K+")",this.status=K}function Qi(K){for(;K.length>0;)K.shift()(e)}function Wk(K){return K}function Uk(K){var lt=/\b_Z[\w\d_]+/g;return K.replace(lt,function(_t){var re=_t;return _t===re?_t:re+" ["+_t+"]"})}function Yc(){var K=new Error;if(!K.stack){try{throw new Error}catch(lt){K=lt}if(!K.stack)return"(no stack trace available)"}return K.stack.toString()}function gI(K,lt){ot.set(K,lt)}function pg(){Xr("")}function Od(){return 2147483648}function en(){return Od()}function mg(K,lt,_t){at.copyWithin(K,lt,lt+_t)}function xI(K){try{return V.grow(K-rt.byteLength+65535>>>16),At(V.buffer),1}catch(lt){}}function yI(K){var lt=at.length;K=K>>>0;var _t=Od();if(K>_t)return!1;let re=(dr,jn)=>dr+(jn-dr%jn)%jn;for(var Ke=1;Ke<=4;Ke*=2){var je=lt*(1+.2/Ke);je=Math.min(je,K+100663296);var ee=Math.min(_t,re(Math.max(K,je),65536)),te=xI(ee);if(te)return!0}return!1}var bI={varargs:void 0,get:function(){bI.varargs+=4;var K=dt[bI.varargs-4>>2];return K},getStr:function(K){var lt=Y(K);return lt}};function Hk(K){return 52}function wI(K,lt,_t,re,Ke){return 70}var Ru=[null,[],[]];function CI(K,lt){var _t=Ru[K];lt===0||lt===10?((K===1?_:A)(j(_t,0)),_t.length=0):_t.push(lt)}function II(K,lt,_t,re){for(var Ke=0,je=0;je<_t;je++){var ee=ht[lt>>2],te=ht[lt+4>>2];lt+=8;for(var dr=0;dr<te;dr++)CI(K,at[ee+dr]);Ke+=te}return ht[re>>2]=Ke,0}function Pd(K){var lt=e["_"+K];return lt}function fg(K,lt,_t,re,Ke){var je={string:rn=>{var Nl=0;if(rn!=null&&rn!==0){var _g=(rn.length<<2)+1;Nl=Vd(_g),et(rn,Nl,_g)}return Nl},array:rn=>{var Nl=Vd(rn.length);return gI(rn,Nl),Nl}};function ee(rn){return lt==="string"?Y(rn):lt==="boolean"?Boolean(rn):rn}var te=Pd(K),dr=[],jn=0;if(re)for(var ta=0;ta<re.length;ta++){var Eg=je[_t[ta]];Eg?(jn===0&&(jn=Ng()),dr[ta]=Eg(re[ta])):dr[ta]=re[ta]}var Gd=te.apply(null,dr);function XS(rn){return jn!==0&&Tg(jn),ee(rn)}return Gd=XS(Gd),Gd}function dg(K,lt,_t,re){_t=_t||[];var Ke=_t.every(ee=>ee==="number"||ee==="boolean"),je=lt!=="string";return je&&Ke&&!re?Pd(K):function(){return fg(K,lt,_t,arguments,re)}}var Ld={abort:pg,emscripten_get_heap_max:en,emscripten_memcpy_big:mg,emscripten_resize_heap:yI,fd_close:Hk,fd_seek:wI,fd_write:II},qk=Fd(),hg=e.___wasm_call_ctors=function(){return(hg=e.___wasm_call_ctors=e.asm.__wasm_call_ctors).apply(null,arguments)},SI=e._init=function(){return(SI=e._init=e.asm.init).apply(null,arguments)},Kt=e._init_with_threads_count=function(){return(Kt=e._init_with_threads_count=e.asm.init_with_threads_count).apply(null,arguments)},Md=e._get_threads_count=function(){return(Md=e._get_threads_count=e.asm.get_threads_count).apply(null,arguments)},vI=e._register_tensor=function(){return(vI=e._register_tensor=e.asm.register_tensor).apply(null,arguments)},Kk=e._dispose_data=function(){return(Kk=e._dispose_data=e.asm.dispose_data).apply(null,arguments)},jk=e._dispose=function(){return(jk=e._dispose=e.asm.dispose).apply(null,arguments)},NI=e._Abs=function(){return(NI=e._Abs=e.asm.Abs).apply(null,arguments)},gg=e._Add=function(){return(gg=e._Add=e.asm.Add).apply(null,arguments)},Zc=e._AddN=function(){return(Zc=e._AddN=e.asm.AddN).apply(null,arguments)},TI=e._All=function(){return(TI=e._All=e.asm.All).apply(null,arguments)},kI=e._Any=function(){return(kI=e._Any=e.asm.Any).apply(null,arguments)},Xk=e._ArgMax=function(){return(Xk=e._ArgMax=e.asm.ArgMax).apply(null,arguments)},EI=e._AvgPool=function(){return(EI=e._AvgPool=e.asm.AvgPool).apply(null,arguments)},_I=e._BatchMatMul=function(){return(_I=e._BatchMatMul=e.asm.BatchMatMul).apply(null,arguments)},AI=e._Ceil=function(){return(AI=e._Ceil=e.asm.Ceil).apply(null,arguments)},$I=e._ClipByValue=function(){return($I=e._ClipByValue=e.asm.ClipByValue).apply(null,arguments)},xg=e._Conv2D=function(){return(xg=e._Conv2D=e.asm.Conv2D).apply(null,arguments)},yg=e._Conv2DBackpropInput=function(){return(yg=e._Conv2DBackpropInput=e.asm.Conv2DBackpropInput).apply(null,arguments)},DI=e._Cos=function(){return(DI=e._Cos=e.asm.Cos).apply(null,arguments)},RI=e._Cosh=function(){return(RI=e._Cosh=e.asm.Cosh).apply(null,arguments)},FI=e._CropAndResize=function(){return(FI=e._CropAndResize=e.asm.CropAndResize).apply(null,arguments)},zd=e._Cumprod=function(){return(zd=e._Cumprod=e.asm.Cumprod).apply(null,arguments)},OI=e._Cumsum=function(){return(OI=e._Cumsum=e.asm.Cumsum).apply(null,arguments)},PI=e._DepthToSpace=function(){return(PI=e._DepthToSpace=e.asm.DepthToSpace).apply(null,arguments)},LI=e._DepthwiseConv2dNative=function(){return(LI=e._DepthwiseConv2dNative=e.asm.DepthwiseConv2dNative).apply(null,arguments)},Fu=e._Elu=function(){return(Fu=e._Elu=e.asm.Elu).apply(null,arguments)},MI=e._Equal=function(){return(MI=e._Equal=e.asm.Equal).apply(null,arguments)},zI=e._Exp=function(){return(zI=e._Exp=e.asm.Exp).apply(null,arguments)},bg=e._FlipLeftRight=function(){return(bg=e._FlipLeftRight=e.asm.FlipLeftRight).apply(null,arguments)},BI=e._Floor=function(){return(BI=e._Floor=e.asm.Floor).apply(null,arguments)},Jc=e._FloorDiv=function(){return(Jc=e._FloorDiv=e.asm.FloorDiv).apply(null,arguments)},VI=e._FusedBatchNorm=function(){return(VI=e._FusedBatchNorm=e.asm.FusedBatchNorm).apply(null,arguments)},GI=e._FusedConv2D=function(){return(GI=e._FusedConv2D=e.asm.FusedConv2D).apply(null,arguments)},vl=e._FusedDepthwiseConv2D=function(){return(vl=e._FusedDepthwiseConv2D=e.asm.FusedDepthwiseConv2D).apply(null,arguments)},Bd=e._Gather=function(){return(Bd=e._Gather=e.asm.Gather).apply(null,arguments)},WI=e._GatherNd=function(){return(WI=e._GatherNd=e.asm.GatherNd).apply(null,arguments)},UI=e._Greater=function(){return(UI=e._Greater=e.asm.Greater).apply(null,arguments)},HI=e._GreaterEqual=function(){return(HI=e._GreaterEqual=e.asm.GreaterEqual).apply(null,arguments)},qI=e._LeakyRelu=function(){return(qI=e._LeakyRelu=e.asm.LeakyRelu).apply(null,arguments)},wg=e._Less=function(){return(wg=e._Less=e.asm.Less).apply(null,arguments)},Cg=e._LessEqual=function(){return(Cg=e._LessEqual=e.asm.LessEqual).apply(null,arguments)},KI=e._Log=function(){return(KI=e._Log=e.asm.Log).apply(null,arguments)},jI=e._LogicalAnd=function(){return(jI=e._LogicalAnd=e.asm.LogicalAnd).apply(null,arguments)},Ig=e._LogicalNot=function(){return(Ig=e._LogicalNot=e.asm.LogicalNot).apply(null,arguments)},Sg=e._LogicalOr=function(){return(Sg=e._LogicalOr=e.asm.LogicalOr).apply(null,arguments)},XI=e._LogicalXor=function(){return(XI=e._LogicalXor=e.asm.LogicalXor).apply(null,arguments)},YI=e._Max=function(){return(YI=e._Max=e.asm.Max).apply(null,arguments)},ZI=e._MaxPool=function(){return(ZI=e._MaxPool=e.asm.MaxPool).apply(null,arguments)},vg=e._Maximum=function(){return(vg=e._Maximum=e.asm.Maximum).apply(null,arguments)},Yk=e._Mean=function(){return(Yk=e._Mean=e.asm.Mean).apply(null,arguments)},JI=e._Min=function(){return(JI=e._Min=e.asm.Min).apply(null,arguments)},QI=e._Minimum=function(){return(QI=e._Minimum=e.asm.Minimum).apply(null,arguments)},tS=e._MirrorPad=function(){return(tS=e._MirrorPad=e.asm.MirrorPad).apply(null,arguments)},eS=e._Multiply=function(){return(eS=e._Multiply=e.asm.Multiply).apply(null,arguments)},rS=e._Neg=function(){return(rS=e._Neg=e.asm.Neg).apply(null,arguments)},nS=e._NonMaxSuppressionV3=function(){return(nS=e._NonMaxSuppressionV3=e.asm.NonMaxSuppressionV3).apply(null,arguments)},oS=e._NonMaxSuppressionV4=function(){return(oS=e._NonMaxSuppressionV4=e.asm.NonMaxSuppressionV4).apply(null,arguments)},sS=e._NonMaxSuppressionV5=function(){return(sS=e._NonMaxSuppressionV5=e.asm.NonMaxSuppressionV5).apply(null,arguments)},iS=e._NotEqual=function(){return(iS=e._NotEqual=e.asm.NotEqual).apply(null,arguments)},aS=e._OneHot=function(){return(aS=e._OneHot=e.asm.OneHot).apply(null,arguments)},lS=e._PadV2=function(){return(lS=e._PadV2=e.asm.PadV2).apply(null,arguments)},uS=e._Pow=function(){return(uS=e._Pow=e.asm.Pow).apply(null,arguments)},cS=e._Prelu=function(){return(cS=e._Prelu=e.asm.Prelu).apply(null,arguments)},pS=e._Prod=function(){return(pS=e._Prod=e.asm.Prod).apply(null,arguments)},mS=e._RealDiv=function(){return(mS=e._RealDiv=e.asm.RealDiv).apply(null,arguments)},fS=e._Relu=function(){return(fS=e._Relu=e.asm.Relu).apply(null,arguments)},dS=e._Relu6=function(){return(dS=e._Relu6=e.asm.Relu6).apply(null,arguments)},hS=e._ResizeBilinear=function(){return(hS=e._ResizeBilinear=e.asm.ResizeBilinear).apply(null,arguments)},gS=e._ResizeNearestNeighbor=function(){return(gS=e._ResizeNearestNeighbor=e.asm.ResizeNearestNeighbor).apply(null,arguments)},xS=e._Reverse=function(){return(xS=e._Reverse=e.asm.Reverse).apply(null,arguments)},yS=e._RotateWithOffset=function(){return(yS=e._RotateWithOffset=e.asm.RotateWithOffset).apply(null,arguments)},bS=e._Round=function(){return(bS=e._Round=e.asm.Round).apply(null,arguments)},wS=e._Rsqrt=function(){return(wS=e._Rsqrt=e.asm.Rsqrt).apply(null,arguments)},CS=e._ScatterNd=function(){return(CS=e._ScatterNd=e.asm.ScatterNd).apply(null,arguments)},IS=e._SelectV2=function(){return(IS=e._SelectV2=e.asm.SelectV2).apply(null,arguments)},SS=e._Sigmoid=function(){return(SS=e._Sigmoid=e.asm.Sigmoid).apply(null,arguments)},vS=e._Sin=function(){return(vS=e._Sin=e.asm.Sin).apply(null,arguments)},NS=e._Softmax=function(){return(NS=e._Softmax=e.asm.Softmax).apply(null,arguments)},TS=e._SparseFillEmptyRows=function(){return(TS=e._SparseFillEmptyRows=e.asm.SparseFillEmptyRows).apply(null,arguments)},kS=e._SparseReshape=function(){return(kS=e._SparseReshape=e.asm.SparseReshape).apply(null,arguments)},ES=e._SparseSegmentReduction=function(){return(ES=e._SparseSegmentReduction=e.asm.SparseSegmentReduction).apply(null,arguments)},_S=e._Sqrt=function(){return(_S=e._Sqrt=e.asm.Sqrt).apply(null,arguments)},AS=e._Square=function(){return(AS=e._Square=e.asm.Square).apply(null,arguments)},$S=e._SquaredDifference=function(){return($S=e._SquaredDifference=e.asm.SquaredDifference).apply(null,arguments)},DS=e._Step=function(){return(DS=e._Step=e.asm.Step).apply(null,arguments)},RS=e._StridedSlice=function(){return(RS=e._StridedSlice=e.asm.StridedSlice).apply(null,arguments)},FS=e._Sub=function(){return(FS=e._Sub=e.asm.Sub).apply(null,arguments)},OS=e._Sum=function(){return(OS=e._Sum=e.asm.Sum).apply(null,arguments)},PS=e._Tan=function(){return(PS=e._Tan=e.asm.Tan).apply(null,arguments)},LS=e._Tanh=function(){return(LS=e._Tanh=e.asm.Tanh).apply(null,arguments)},MS=e._Tile=function(){return(MS=e._Tile=e.asm.Tile).apply(null,arguments)},zS=e._TopK=function(){return(zS=e._TopK=e.asm.TopK).apply(null,arguments)},BS=e._Transform=function(){return(BS=e._Transform=e.asm.Transform).apply(null,arguments)},VS=e._Transpose=function(){return(VS=e._Transpose=e.asm.Transpose).apply(null,arguments)},GS=e.__FusedMatMul=function(){return(GS=e.__FusedMatMul=e.asm._FusedMatMul).apply(null,arguments)},WS=e._malloc=function(){return(WS=e._malloc=e.asm.malloc).apply(null,arguments)},US=e._free=function(){return(US=e._free=e.asm.free).apply(null,arguments)},HS=e.___errno_location=function(){return(HS=e.___errno_location=e.asm.__errno_location).apply(null,arguments)},Ng=e.stackSave=function(){return(Ng=e.stackSave=e.asm.stackSave).apply(null,arguments)},Tg=e.stackRestore=function(){return(Tg=e.stackRestore=e.asm.stackRestore).apply(null,arguments)},Vd=e.stackAlloc=function(){return(Vd=e.stackAlloc=e.asm.stackAlloc).apply(null,arguments)},qS=e.dynCall_iijjiiii=function(){return(qS=e.dynCall_iijjiiii=e.asm.dynCall_iijjiiii).apply(null,arguments)},KS=e.dynCall_jiji=function(){return(KS=e.dynCall_jiji=e.asm.dynCall_jiji).apply(null,arguments)};e.cwrap=dg;var Qc;zr=function K(){Qc||kg(),Qc||(zr=K)};function kg(K){if(K=K||a,qe>0||(Ae(),qe>0))return;function lt(){Qc||(Qc=!0,e.calledRun=!0,!G&&(We(),n(e),e.onRuntimeInitialized&&e.onRuntimeInitialized(),_n()))}e.setStatus?(e.setStatus("Running..."),setTimeout(function(){setTimeout(function(){e.setStatus("")},1),lt()},1)):lt()}if(e.preInit)for(typeof e.preInit=="function"&&(e.preInit=[e.preInit]);e.preInit.length>0;)e.preInit.pop()();kg();var tp;s&&(tp={uncaughtException:process.listeners("uncaughtException").filter(function(K){return!s.uncaughtException.indexOf(K)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(K){return!s.unhandledRejection.indexOf(K)>-1})});var ep;if(typeof t!="undefined")ep=t;else if(typeof WasmBackendModuleThreadedSimd!="undefined")ep=WasmBackendModuleThreadedSimd;else throw new Error("Could not find wasm module in post.js");if(tp){var jS=ep._dispose;ep._dispose=function(){jS(),tp.uncaughtException.forEach(function(K){process.removeListener("uncaughtException",K)}),tp.unhandledRejection.forEach(function(K){process.removeListener("unhandledRejection",K)})}}return t.ready}})();typeof dI=="object"&&typeof Lk=="object"?Lk.exports=Pk:typeof define=="function"&&define.amd?define([],function(){return Pk}):typeof dI=="object"&&(dI.WasmBackendModule=Pk)});var ra=class{constructor(t,e){this.backend=t,this.dataMover=e,this.data=new WeakMap,this.dataIdsCount=0}get(t){return this.data.has(t)||this.dataMover.moveData(this.backend,t),this.data.get(t)}set(t,e){this.dataIdsCount++,this.data.set(t,e)}has(t){return this.data.has(t)}delete(t){return this.dataIdsCount--,this.data.delete(t)}numDataIds(){return this.dataIdsCount}},zo=class{refCount(t){return $n("refCount")}incRef(t){return $n("incRef")}timerAvailable(){return!0}time(t){return $n("time")}read(t){return $n("read")}readSync(t){return $n("readSync")}readToGPU(t,e){return $n("readToGPU")}numDataIds(){return $n("numDataIds")}disposeData(t,e){return $n("disposeData")}write(t,e,n){return $n("write")}move(t,e,n,o,s){return $n("move")}createTensorFromTexture(t,e,n){return $n("createTensorFromTexture")}memory(){return $n("memory")}floatPrecision(){return $n("floatPrecision")}epsilon(){return this.floatPrecision()===32?1e-7:1e-4}dispose(){return $n("dispose")}};function $n(r){throw new Error(`'${r}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function s1(r){let t=r.length,e=0;for(;t>0;)e=Math.random()*t|0,t--,Lg(r,t,e)}function MU(r,t){if(r.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${r.length}Second array length was ${t.length}`);let e=r.length,n=0;for(;e>0;)n=Math.random()*e|0,e--,Lg(r,e,n),Lg(t,e,n)}function np(r,t,e){return Math.max(r,Math.min(t,e))}function zU(r){return r%2===0?r:r+1}function Lg(r,t,e){let n=r[t];r[t]=r[e],r[e]=n}function BU(r){let t=0;for(let e=0;e<r.length;e++)t+=r[e];return t}function VU(r,t){let e=Math.random();return t*e+(1-e)*r}function GU(r,t){let e=0;for(let n=0;n<r.length;n++){let o=Number(r[n])-Number(t[n]);e+=o*o}return e}function E(r,t){if(!r)throw new Error(typeof t=="string"?t:t())}function $e(r,t,e=""){E(Dn(r,t),()=>e+` Shapes ${r} and ${t} must match`)}function Yn(r){E(r!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Bo(r,t=[],e=!1){if(t==null&&(t=[]),Array.isArray(r)||xr(r)&&!e)for(let n=0;n<r.length;++n)Bo(r[n],t,e);else t.push(r);return t}function Jt(r){if(r.length===0)return 1;let t=r[0];for(let e=1;e<r.length;e++)t*=r[e];return t}function WU(r){return r.length===0}function Dn(r,t){if(r===t)return!0;if(r==null||t==null||r.length!==t.length)return!1;for(let e=0;e<r.length;e++)if(r[e]!==t[e])return!1;return!0}function na(r){return r%1===0}function UU(r){if(Math.tanh!=null)return Math.tanh(r);if(r===1/0)return 1;if(r===-1/0)return-1;{let t=Math.exp(2*r);return(t-1)/(t+1)}}function HU(r){let t=Math.ceil(Math.sqrt(r));return[t,Math.ceil(r/t)]}function qU(r){let t=new Uint32Array(r);for(let e=0;e<r;++e)t[e]=e;return s1(t),t}function Pu(r,t){return t<=r.length?r:r+" ".repeat(t-r.length)}function KU(r,t=o=>0,e,n){return new Promise((o,s)=>{let i=0,a=()=>{if(r()){o();return}i++;let u=t(i);if(e!=null&&i>=e){s();return}n!=null?n(a,u):setTimeout(a,u)};a()})}function jU(r,t){let e=1,n=-1;for(let s=0;s<r.length;++s)if(r[s]>=0)e*=r[s];else if(r[s]===-1){if(n!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${n} and dim ${s}`);n=s}else if(r[s]<0)throw Error(`Shapes can not be < 0. Found ${r[s]} at dim ${s}`);if(n===-1){if(t>0&&t!==e)throw Error(`Size(${t}) must match the product of shape ${r}`);return r}if(e===0)throw Error(`Cannot infer the missing size in [${r}] when there are 0 elements`);if(t%e!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${e}`);let o=r.slice();return o[n]=t/e,o}function lr(r,t){let e=t.length;return r=r==null?t.map((n,o)=>o):[].concat(r),E(r.every(n=>n>=-e&&n<e),()=>`All values in axis param must be in range [-${e}, ${e}) but got axis ${r}`),E(r.every(n=>na(n)),()=>`All values in axis param must be integers but got axis ${r}`),r.map(n=>n<0?e+n:n)}function t0(r,t){let e=[],n=[],o=t!=null&&Array.isArray(t)&&t.length===0,s=t==null||o?null:lr(t,r).sort(),i=0;for(let a=0;a<r.length;++a){if(s!=null){if(s[i]===a&&r[a]!==1)throw new Error(`Can't squeeze axis ${a} since its dim '${r[a]}' is not 1`);(s[i]==null||s[i]>a)&&r[a]===1&&(e.push(r[a]),n.push(a)),s[i]<=a&&i++}r[a]!==1&&(e.push(r[a]),n.push(a))}return{newShape:e,keptDims:n}}function e0(r,t){let e=null;if(r==null||r==="float32")e=new Float32Array(t);else if(r==="int32")e=new Int32Array(t);else if(r==="bool")e=new Uint8Array(t);else throw new Error(`Unknown data type ${r}`);return e}function r0(r,t){let e=null;if(r==null||r==="float32")e=new Float32Array(t);else if(r==="int32")e=new Int32Array(t);else if(r==="bool")e=new Uint8Array(t);else if(r==="string")e=new Array(t);else throw new Error(`Unknown data type ${r}`);return e}function n0(r,t){for(let e=0;e<r.length;e++){let n=r[e];if(isNaN(n)||!isFinite(n))throw Error(`A tensor of type ${t} being uploaded contains ${n}.`)}}function o0(r){return r==="bool"||r==="complex64"||r==="float32"||r==="int32"||r==="string"}function XU(r,t){return!(t==="complex64"||t==="float32"&&r!=="complex64"||t==="int32"&&r!=="float32"&&r!=="complex64"||t==="bool"&&r==="bool")}function xr(r){return r instanceof Float32Array||r instanceof Int32Array||r instanceof Uint8Array||r instanceof Uint8ClampedArray}function Mg(r){if(r==="float32"||r==="int32")return 4;if(r==="complex64")return 8;if(r==="bool")return 1;throw new Error(`Unknown dtype ${r}`)}function s0(r){if(r==null)return 0;let t=0;return r.forEach(e=>t+=e.length),t}function Vo(r){return typeof r=="string"||r instanceof String}function i1(r){return typeof r=="boolean"}function a1(r){return typeof r=="number"}function op(r){return Array.isArray(r)?op(r[0]):r instanceof Float32Array?"float32":r instanceof Int32Array||r instanceof Uint8Array||r instanceof Uint8ClampedArray?"int32":a1(r)?"float32":Vo(r)?"string":i1(r)?"bool":"float32"}function oi(r){return!!(r&&r.constructor&&r.call&&r.apply)}function sp(r,t){for(let e=t;e<r;++e)if(r%e===0)return e;return r}function si(r){let t=r.length;if(t<2)return[];let e=new Array(t-1);e[t-2]=r[t-1];for(let n=t-3;n>=0;--n)e[n]=e[n+1]*r[n+1];return e}function l1(r,t,e,n=!1){let o=new Array;if(t.length===1){let s=t[0]*(n?2:1);for(let i=0;i<s;i++)o[i]=e[r+i]}else{let s=t[0],i=t.slice(1),a=i.reduce((u,l)=>u*l)*(n?2:1);for(let u=0;u<s;u++)o[u]=l1(r+u*a,i,e,n)}return o}function Ou(r,t,e=!1){if(r.length===0)return t[0];let n=r.reduce((o,s)=>o*s)*(e?2:1);if(n===0)return[];if(n!==t.length)throw new Error(`[${r}] does not match the input size ${t.length}${e?" for a complex tensor":""}.`);return l1(0,r,t,e)}function Wd(r,t){let e=ip(r,t);for(let n=0;n<e.length;n++)e[n]=1;return e}function ip(r,t){if(t==null||t==="float32"||t==="complex64")return new Float32Array(r);if(t==="int32")return new Int32Array(r);if(t==="bool")return new Uint8Array(r);throw new Error(`Unknown data type ${t}`)}function YU(r,t){let e=r.reduce((n,o)=>n*o,1);if(t==null||t==="float32")return Ou(r,new Float32Array(e));if(t==="int32")return Ou(r,new Int32Array(e));if(t==="bool")return Ou(r,new Uint8Array(e));throw new Error(`Unknown data type ${t}`)}function Ud(r){r.forEach(t=>{E(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${r}].`)})}function ZU(r,t,e){if(t===0)return 0;if(t===1)return r[0];let n=r[r.length-1];for(let o=0;o<r.length-1;++o)n+=e[o]*r[o];return n}function JU(r,t,e){if(t===0)return[];if(t===1)return[r];let n=new Array(t);for(let o=0;o<n.length-1;++o)n[o]=Math.floor(r/e[o]),r-=n[o]*e[o];return n[n.length-1]=r,n}function Hd(r){return r&&r.then&&typeof r.then=="function"}var u1="tfjsflags",qd=class{constructor(t){this.global=t,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=t4,this.populateURLFlags()}setPlatform(t,e){this.platform!=null&&(z().getBool("IS_TEST")||z().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${t}.`)),this.platformName=t,this.platform=e}registerFlag(t,e,n){if(this.flagRegistry[t]={evaluationFn:e,setHook:n},this.urlFlags[t]!=null){let o=this.urlFlags[t];z().getBool("IS_TEST")||z().getBool("PROD")||console.warn(`Setting feature override from URL ${t}: ${o}.`),this.set(t,o)}}async getAsync(t){return t in this.flags?this.flags[t]:(this.flags[t]=await this.evaluateFlag(t),this.flags[t])}get(t){if(t in this.flags)return this.flags[t];let e=this.evaluateFlag(t);if(Hd(e))throw new Error(`Flag ${t} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[t]=e,this.flags[t]}getNumber(t){return this.get(t)}getBool(t){return this.get(t)}getFlags(){return this.flags}get features(){return this.flags}set(t,e){if(this.flagRegistry[t]==null)throw new Error(`Cannot set flag ${t} as it has not been registered.`);this.flags[t]=e,this.flagRegistry[t].setHook!=null&&this.flagRegistry[t].setHook(e)}evaluateFlag(t){if(this.flagRegistry[t]==null)throw new Error(`Cannot evaluate flag '${t}': no evaluation function found.`);return this.flagRegistry[t].evaluationFn()}setFlags(t){this.flags=Object.assign({},t)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global=="undefined"||typeof this.global.location=="undefined"||typeof this.global.location.search=="undefined")return;let t=this.getQueryParams(this.global.location.search);u1 in t&&t[u1].split(",").forEach(n=>{let[o,s]=n.split(":");this.urlFlags[o]=r4(o,s)})}};function t4(r){let t={};return r.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(e,...n)=>(e4(t,n[0],n[1]),n.join("="))),t}function e4(r,t,e){r[decodeURIComponent(t)]=decodeURIComponent(e||"")}function r4(r,t){if(t=t.toLowerCase(),t==="true"||t==="false")return t==="true";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${r}.`)}function z(){return i0}var i0=null;function c1(r){i0=r}var a0;function l0(){if(a0==null){let r;if(typeof window!="undefined")r=window;else if(typeof global!="undefined")r=global;else if(typeof process!="undefined")r=process;else if(typeof self!="undefined")r=self;else throw new Error("Could not find a global object");a0=r}return a0}function n4(){let r=l0();return r._tfGlobals==null&&(r._tfGlobals=new Map),r._tfGlobals}function Kd(r,t){let e=n4();if(e.has(r))return e.get(r);{let n=t();return e.set(r,n),e.get(r)}}var ii="Abs",oa="Acos",sa="Acosh",Zn="Add",Go="AddN",ia="All",aa="Any",Wo="ArgMax",kl="ArgMin",la="Asin",ua="Asinh",ca="Atan",pa="Atanh",ma="Atan2",Uo="AvgPool",ap="AvgPoolGrad",El="AvgPool3D",lp="AvgPool3DGrad",Ho="BatchMatMul",ai="BatchToSpaceND",up="Bincount",p1="BroadcastTo",cp="BroadcastArgs",lo="Cast",qo="Ceil",uo="ClipByValue",pp="Complex",_l="ComplexAbs",li="Concat",Ko="Conv2D",mp="Conv2DBackpropFilter",jo="Conv2DBackpropInput",Al="Conv3D",fp="Conv3DBackpropFilterV2",dp="Conv3DBackpropInputV2",Xo="Cos",Yo="Cosh",fa="Cumprod",Zo="Cumsum",da="CropAndResize",hp="DenseBincount",ha="DepthToSpace",Jo="DepthwiseConv2dNative",gp="DepthwiseConv2dNativeBackpropFilter",xp="DepthwiseConv2dNativeBackpropInput",yp="Diag",$l="Dilation2D",jd="Dilation2DBackpropInput",Xd="Dilation2DBackpropFilter",Qo="RealDiv",bp="Einsum",ts="Elu",wp="EluGrad",ga="Erf",xa="Equal",es="Exp",ui="ExpandDims",ya="Expm1",Cp="FFT",Dl="Fill",ba="FlipLeftRight",rs="Floor",ns="FloorDiv",os="FusedBatchNorm",ci="GatherV2",wa="GatherNd",Ca="Greater",ss="GreaterEqual",co="Identity",Ip="IFFT",Sp="Imag",Ia="IsFinite",Sa="IsInf",va="IsNan",is="LeakyRelu",Na="Less",Ta="LessEqual",vp="LinSpace",as="Log",ka="Log1p",Ea="LogicalAnd",_a="LogicalNot",Aa="LogicalOr",m1="LogicalXor",f1="LogSoftmax",xlt="LowerBound",Rl="LRN",Np="LRNGrad",ls="Max",us="Maximum",cs="MaxPool",Tp="MaxPoolGrad",Fl="MaxPool3D",kp="MaxPool3DGrad",Ep="MaxPoolWithArgmax",ps="Mean",ms="Min",fs="Minimum",ds="MirrorPad",$a="Mod",_p="Multinomial",hs="Multiply",pi="Neg",Da="NotEqual",Ra="NonMaxSuppressionV3",Fa="NonMaxSuppressionV4",Oa="NonMaxSuppressionV5",mi="OnesLike",gs="OneHot",fi="Pack",xs="PadV2",ylt="Pool",ys="Pow",bs="Prelu",ws="Prod",Ap="RaggedGather",$p="RaggedRange",Dp="RaggedTensorToTensor",Ol="Range",Rp="Real",Pa="Reciprocal",Cs="Relu",di="Reshape",Is="ResizeNearestNeighbor",Fp="ResizeNearestNeighborGrad",Ss="ResizeBilinear",Op="ResizeBilinearGrad",vs="Relu6",Ns="Reverse",Ts="Round",ks="Rsqrt",La="ScatterNd",Pp="SearchSorted",hi="Select",Ma="Selu",gi="Slice",Es="Sin",za="Sinh",Ba="Sign",_s="Sigmoid",Va="Softplus",As="Sqrt",$s="Sum",xi="SpaceToBatchND",yi="SplitV",Ds="Softmax",Pl="SparseFillEmptyRows",Ga="SparseReshape",Ll="SparseSegmentMean",Ml="SparseSegmentSum",Lp="SparseToDense",Rs="SquaredDifference",zl="Square",Wa="StridedSlice",Bl="StringNGrams",Vl="StringSplit",Gl="StringToHashBucketFast",Fs="Sub",Os="Tan",Ps="Tanh",Jn="Tile",Ua="TopK",Ha="Transform",Qn="Transpose",Mp="Unique",bi="Unpack",Wl="UnsortedSegmentSum",blt="UpperBound",wi="ZerosLike",po="Step",Yd="FromPixels",qa="RotateWithOffset",Ci="_FusedMatMul",Ii="FusedConv2D",Si="FusedDepthwiseConv2D";function vi(...r){z().getBool("IS_TEST")||z().getBool("PROD")||console.warn(...r)}function o4(...r){z().getBool("IS_TEST")||z().getBool("PROD")||console.log(...r)}var zp=Kd("kernelRegistry",()=>new Map),Zd=Kd("gradRegistry",()=>new Map);function Jd(r,t){let e=c0(r,t);return zp.get(e)}function u0(r){return Zd.get(r)}function zg(r){let t=zp.entries(),e=[];for(;;){let{done:n,value:o}=t.next();if(n)break;let[s,i]=o,[a]=s.split("_");a===r&&e.push(i)}return e}function Lu(r){let{kernelName:t,backendName:e}=r,n=c0(t,e);zp.has(n)&&vi(`The kernel '${t}' for backend '${e}' is already registered`),zp.set(n,r)}function h1(r){let{kernelName:t}=r;Zd.has(t)&&z().getBool("DEBUG")&&vi(`Overriding the gradient for '${t}'`),Zd.set(t,r)}function vlt(r,t){let e=c0(r,t);if(!zp.has(e))throw new Error(`The kernel '${r}' for backend '${t}' is not registered`);zp.delete(e)}function Nlt(r){if(!Zd.has(r))throw new Error(`The gradient '${r}' for backend is not registered`);Zd.delete(r)}function Tlt(r,t){zg(r).forEach(n=>{let o=Object.assign({},n,{backendName:t});Lu(o)})}function c0(r,t){return`${t}_${r}`}var y={};Wt(y,{arraysEqual:()=>Dn,assert:()=>E,assertNonNegativeIntegerDimensions:()=>Ud,assertNonNull:()=>Yn,assertShapesMatch:()=>$e,bytesFromStringArray:()=>s0,bytesPerElement:()=>Mg,checkConversionForErrors:()=>n0,clamp:()=>np,computeStrides:()=>si,createScalarValue:()=>p4,createShuffledIndices:()=>qU,decodeString:()=>Wp,distSquared:()=>GU,encodeString:()=>Hl,fetch:()=>f4,fingerPrint64:()=>c4,flatten:()=>Bo,getArrayFromDType:()=>r0,getTypedArrayFromDType:()=>e0,hasEncodingLoss:()=>XU,hexToLong:()=>Qd,indexToLoc:()=>JU,inferDtype:()=>op,inferFromImplicitShape:()=>jU,isBoolean:()=>i1,isFunction:()=>oi,isInt:()=>na,isNumber:()=>a1,isPromise:()=>Hd,isScalarShape:()=>WU,isString:()=>Vo,isTypedArray:()=>xr,isValidDtype:()=>o0,locToIndex:()=>ZU,makeOnesTypedArray:()=>Wd,makeZerosNestedTypedArray:()=>YU,makeZerosTypedArray:()=>ip,nearestDivisor:()=>sp,nearestLargerEven:()=>zU,now:()=>Gu,parseAxisParam:()=>lr,randUniform:()=>VU,repeatedTry:()=>KU,rightPad:()=>Pu,shuffle:()=>s1,shuffleCombo:()=>MU,sizeFromShape:()=>Jt,sizeToSquarishShape:()=>HU,squeezeShape:()=>t0,sum:()=>BU,swap:()=>Lg,tanh:()=>UU,toNestedArray:()=>Ou,toTypedArray:()=>Gp});var d0=Tl(T1());var Vu=d0.default||d0;function Qd(r){return Vu.fromString(r,!0,16)}var E1=Qd("c3a5c85c97cb3127"),Bu=Qd("b492b66fbe98f273"),nn=Qd("9ae16a3b2f90404f");function f0(r){return r.xor(r.shru(47))}function _1(r,t,e){let n=r.slice(t,t+e);return Vu.fromBytes(Array.from(n),!0,!0)}function Pe(r,t){return _1(r,t,8)}function k1(r,t){return _1(r,t,4)}function kr(r,t){return t===0?r:r.shru(t).or(r.shl(64-t))}function Ul(r,t,e=Qd("9ddfea08eb382d69")){let n=r.xor(t).mul(e);n=n.xor(n.shru(47));let o=t.xor(n).mul(e);return o=o.xor(o.shru(47)),o=o.mul(e),o}function i4(r,t,e,n,o,s){o=o.add(r),s=kr(s.add(o).add(n),21);let i=o;return o=o.add(t),o=o.add(e),s=s.add(kr(o,44)),[o.add(n),s.add(i)]}function Vg(r,t,e,n){return i4(Pe(r,t),Pe(r,t+8),Pe(r,t+16),Pe(r,t+24),e,n)}function a4(r,t=r.length){if(t>=8){let e=nn.add(t*2),n=Pe(r,0).add(nn),o=Pe(r,t-8),s=kr(o,37).mul(e).add(n),i=kr(n,25).add(o).mul(e);return Ul(s,i,e)}if(t>=4){let e=nn.add(t*2),n=k1(r,0);return Ul(n.shl(3).add(t),k1(r,t-4),e)}if(t>0){let e=r[0],n=r[t>>1],o=r[t-1],s=e+(n<<8),i=t+(o<<2);return f0(nn.mul(s).xor(E1.mul(i))).mul(nn)}return nn}function l4(r,t=r.length){let e=nn.add(t*2),n=Pe(r,0).mul(Bu),o=Pe(r,8),s=Pe(r,t-8).mul(e),i=Pe(r,t-16).mul(nn);return Ul(kr(n.add(o),43).add(kr(s,30)).add(i),n.add(kr(o.add(nn),18)).add(s),e)}function u4(r,t=r.length){let e=nn.add(t*2),n=Pe(r,0).mul(nn),o=Pe(r,8),s=Pe(r,t-8).mul(e),i=Pe(r,t-16).mul(nn),a=kr(n.add(o),43).add(kr(s,30)).add(i),u=Ul(a,n.add(kr(o.add(nn),18)).add(s),e),l=Pe(r,16).mul(e),c=Pe(r,24),p=a.add(Pe(r,t-32)).mul(e),m=u.add(Pe(r,t-24)).mul(e);return Ul(kr(l.add(c),43).add(kr(p,30)).add(m),l.add(kr(c.add(n),18)).add(p),e)}function c4(r,t=r.length){let e=Vu.fromNumber(81,!0);if(t<=32)return t<=16?a4(r,t):l4(r,t);if(t<=64)return u4(r,t);let n=e,o=e.mul(Bu).add(113),s=f0(o.mul(nn).add(113)).mul(nn),i=[Vu.UZERO,Vu.UZERO],a=[Vu.UZERO,Vu.UZERO];n=n.mul(nn).add(Pe(r,0));let u=0,l=(t-1>>6)*64,c=l+(t-1&63)-63;do n=kr(n.add(o).add(i[0]).add(Pe(r,u+8)),37).mul(Bu),o=kr(o.add(i[1]).add(Pe(r,u+48)),42).mul(Bu),n=n.xor(a[1]),o=o.add(i[0]).add(Pe(r,u+40)),s=kr(s.add(a[0]),33).mul(Bu),i=Vg(r,u,i[1].mul(Bu),n.add(a[0])),a=Vg(r,u+32,s.add(a[1]),o.add(Pe(r,u+16))),[s,n]=[n,s],u+=64;while(u!==l);let p=Bu.add(s.and(255).shl(1));return u=c,a[0]=a[0].add(t-1&63),i[0]=i[0].add(a[0]),a[0]=a[0].add(i[0]),n=kr(n.add(o).add(i[0]).add(Pe(r,u+8)),37).mul(p),o=kr(o.add(i[1]).add(Pe(r,u+48)),42).mul(p),n=n.xor(a[1].mul(9)),o=o.add(i[0].mul(9).add(Pe(r,u+40))),s=kr(s.add(a[0]),33).mul(p),i=Vg(r,u,i[1].mul(p),n.add(a[0])),a=Vg(r,u+32,s.add(a[1]),o.add(Pe(r,u+16))),[s,n]=[n,s],Ul(Ul(i[0],a[0],p).add(f0(o).mul(E1)).add(s),Ul(i[1],a[1],p).add(n),p)}function p4(r,t){return t==="string"?Hl(r):Gp([r],t)}function m4(r,t){return r instanceof Float32Array&&t==="float32"||r instanceof Int32Array&&t==="int32"||r instanceof Uint8Array&&t==="bool"}function Gp(r,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(r)&&(r=Bo(r)),z().getBool("DEBUG")&&n0(r,t),m4(r,t))return r;if(t==null||t==="float32"||t==="complex64")return new Float32Array(r);if(t==="int32")return new Int32Array(r);if(t==="bool"){let e=new Uint8Array(r.length);for(let n=0;n<e.length;++n)Math.round(r[n])!==0&&(e[n]=1);return e}else throw new Error(`Unknown data type ${t}`)}function Gu(){return z().platform.now()}function f4(r,t){return z().platform.fetch(r,t)}function Hl(r,t="utf-8"){return t=t||"utf-8",z().platform.encode(r,t)}function Wp(r,t="utf-8"){return t=t||"utf-8",z().platform.decode(r,t)}var Gg=class{constructor(t,e){this.backendTimer=t,this.logger=e,e==null&&(this.logger=new h0)}profileKernel(t,e,n){let o,s=()=>{o=n()},i,a=Gu();if(this.backendTimer.timerAvailable())i=this.backendTimer.time(s);else{s();for(let l of o)l.dataSync();i=Promise.resolve({kernelMs:Gu()-a})}if(z().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let l=0;l<o.length;l++){let c=o[l];c.data().then(p=>{d4(p,c.dtype,t)})}return{kernelName:t,outputs:o,inputs:e,timeMs:i.then(l=>l.kernelMs),extraInfo:i.then(l=>l.getExtraProfileInfo!=null?l.getExtraProfileInfo():"")}}logKernelProfile(t){let{kernelName:e,outputs:n,timeMs:o,inputs:s,extraInfo:i}=t;n.forEach(a=>{Promise.all([a.data(),o,i]).then(u=>{this.logger.logKernelProfile(e,a,u[0],u[1],s,u[2])})})}};function d4(r,t,e){if(t!=="float32")return!1;for(let n=0;n<r.length;n++){let o=r[n];if(isNaN(o)||!isFinite(o))return console.warn(`Found ${o} in the result of '${e}'`),!0}return!1}var h0=class{logKernelProfile(t,e,n,o,s,i){let a=typeof o=="number"?Pu(`${o}ms`,9):o.error,u=Pu(t,25),l=e.rank,c=e.size,p=Pu(e.shape.toString(),14),m="";for(let f in s){let d=s[f];if(d!=null){let h=d.shape||e.shape,g=h.length;m+=`${f}: ${g}D ${g>0?h:""} `}}console.log(`%c${u}	%c${a}	%c${l}D ${p}	%c${c}	%c${m}	%c${i}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function A1(r,t,e){let n={},o={};for(let u=0;u<t.length;u++)n[t[u].id]=!0;for(let u=0;u<r.length;u++){let l=r[u],c=l.inputs;for(let p in c){let m=c[p],f=!1;for(let d=0;d<t.length;d++)if(n[m.id]){l.outputs.forEach(h=>n[h.id]=!0),f=!0,o[l.id]=!0;break}if(f)break}}let s={};s[e.id]=!0;let i={};for(let u=r.length-1;u>=0;u--){let l=r[u],c=l.inputs;for(let p=0;p<l.outputs.length;p++)if(s[l.outputs[p].id]){for(let m in c)s[c[m].id]=!0,i[l.id]=!0;break}}let a=[];for(let u=0;u<r.length;u++){let l=r[u];if(o[l.id]&&i[l.id]){let c={};for(let m in l.inputs){let f=l.inputs[m];n[f.id]&&(c[m]=f)}let p=Object.assign({},l);p.inputs=c,p.outputs=l.outputs,a.push(p)}}return a}function $1(r,t,e,n){for(let o=t.length-1;o>=0;o--){let s=t[o],i=[];if(s.outputs.forEach(u=>{let l=r[u.id];l!=null?i.push(l):i.push(null)}),s.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let a=s.gradient(i);for(let u in s.inputs){if(!(u in a))throw new Error(`Cannot backprop through input ${u}. Available gradients found: ${Object.keys(a)}.`);let l=e(()=>a[u]());if(l.dtype!=="float32")throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${u} must have 'float32' dtype, but has '${l.dtype}'`);let c=s.inputs[u];if(!Dn(l.shape,c.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${u}' has shape '${l.shape}', which does not match the shape of the input '${c.shape}'`);if(r[c.id]==null)r[c.id]=l;else{let p=r[c.id];r[c.id]=n(p,l),p.dispose()}}}}var D1=20,th=3,g0=7;function R1(r,t,e,n){let o=si(t),s=h4(r,t,e,o),i=t.length,a=Wg(r,t,e,o,s),u=["Tensor"];return n&&(u.push(`  dtype: ${e}`),u.push(`  rank: ${i}`),u.push(`  shape: [${t}]`),u.push("  values:")),u.push(a.map(l=>"    "+l).join(`
-`)),u.join(`
-`)}function h4(r,t,e,n){let o=Jt(t),s=n[n.length-1],i=new Array(s).fill(0),a=t.length,u=e==="complex64"?rh(r):r;if(a>1)for(let l=0;l<o/s;l++){let c=l*s;for(let p=0;p<s;p++)i[p]=Math.max(i[p],eh(u[c+p],0,e).length)}return i}function eh(r,t,e){let n;return Array.isArray(r)?n=`${parseFloat(r[0].toFixed(g0))} + ${parseFloat(r[1].toFixed(g0))}j`:Vo(r)?n=`'${r}'`:e==="bool"?n=F1(r):n=parseFloat(r.toFixed(g0)).toString(),Pu(n,t)}function F1(r){return r===0?"false":"true"}function Wg(r,t,e,n,o,s=!0){let i=e==="complex64"?2:1,a=t[0],u=t.length;if(u===0){if(e==="complex64"){let h=rh(r);return[eh(h[0],0,e)]}return e==="bool"?[F1(r[0])]:[r[0].toString()]}if(u===1){if(a>D1){let g=th*i,x=Array.from(r.slice(0,g)),b=Array.from(r.slice((a-th)*i,a*i));return e==="complex64"&&(x=rh(x),b=rh(b)),["["+x.map((w,C)=>eh(w,o[C],e)).join(", ")+", ..., "+b.map((w,C)=>eh(w,o[a-th+C],e)).join(", ")+"]"]}let h=e==="complex64"?rh(r):Array.from(r);return["["+h.map((g,x)=>eh(g,o[x],e)).join(", ")+"]"]}let l=t.slice(1),c=n.slice(1),p=n[0]*i,m=[];if(a>D1){for(let h=0;h<th;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,!1))}m.push("...");for(let h=a-th;h<a;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,h===a-1))}}else for(let h=0;h<a;h++){let g=h*p,x=g+p;m.push(...Wg(r.slice(g,x),l,e,c,o,h===a-1))}let f=u===2?",":"";m[0]="["+m[0]+f;for(let h=1;h<m.length-1;h++)m[h]=" "+m[h]+f;let d=`,
-`;for(let h=2;h<u;h++)d+=`
-`;return m[m.length-1]=" "+m[m.length-1]+"]"+(s?"":d),m}function rh(r){let t=[];for(let e=0;e<r.length;e+=2)t.push([r[e],r[e+1]]);return t}var pe=class{constructor(t,e,n){if(this.dtype=e,this.shape=t.slice(),this.size=Jt(t),n!=null){let o=n.length;E(o===this.size,()=>`Length of values '${o}' does not match the size inferred by the shape '${this.size}'.`)}if(e==="complex64")throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=n||r0(e,this.size),this.strides=si(t)}set(t,...e){e.length===0&&(e=[0]),E(e.length===this.rank,()=>`The number of provided coordinates (${e.length}) must match the rank (${this.rank})`);let n=this.locToIndex(e);this.values[n]=t}get(...t){t.length===0&&(t=[0]);let e=0;for(let o of t){if(o<0||o>=this.shape[e]){let s=`Requested out of range element at ${t}.   Buffer shape=${this.shape}`;throw new Error(s)}e++}let n=t[t.length-1];for(let o=0;o<t.length-1;++o)n+=this.strides[o]*t[o];return this.values[n]}locToIndex(t){if(this.rank===0)return 0;if(this.rank===1)return t[0];let e=t[t.length-1];for(let n=0;n<t.length-1;++n)e+=this.strides[n]*t[n];return e}indexToLoc(t){if(this.rank===0)return[];if(this.rank===1)return[t];let e=new Array(this.shape.length);for(let n=0;n<e.length-1;++n)e[n]=Math.floor(t/this.strides[n]),t-=e[n]*this.strides[n];return e[e.length-1]=t,e}get rank(){return this.shape.length}toTensor(){return Ms().makeTensor(this.values,this.shape,this.dtype)}},Ms=null,Up=null,g4=null;function O1(r){Ms=r}function P1(r){Up=r}function L1(r){g4=r}var Ft=class{constructor(t,e,n,o){this.kept=!1,this.isDisposedInternal=!1,this.shape=t.slice(),this.dtype=e||"float32",this.size=Jt(t),this.strides=si(t),this.dataId=n,this.id=o,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let t=await this.data();return Up.buffer(this.shape,this.dtype,t)}bufferSync(){return Up.buffer(this.shape,this.dtype,this.dataSync())}async array(){let t=await this.data();return Ou(this.shape,t,this.dtype==="complex64")}arraySync(){return Ou(this.shape,this.dataSync(),this.dtype==="complex64")}async data(){this.throwIfDisposed();let t=Ms().read(this.dataId);if(this.dtype==="string"){let e=await t;try{return e.map(n=>Wp(n))}catch(n){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return t}dataToGPU(t){return this.throwIfDisposed(),Ms().readToGPU(this.dataId,t)}dataSync(){this.throwIfDisposed();let t=Ms().readSync(this.dataId);if(this.dtype==="string")try{return t.map(e=>Wp(e))}catch(e){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return t}async bytes(){this.throwIfDisposed();let t=await Ms().read(this.dataId);return this.dtype==="string"?t:new Uint8Array(t.buffer)}dispose(){this.isDisposed||(Ms().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(t=!1){return Up.print(this,t)}clone(){return this.throwIfDisposed(),Up.clone(this)}toString(t=!1){let e=this.dataSync();return R1(e,this.shape,this.dtype,t)}cast(t){return this.throwIfDisposed(),Up.cast(this,t)}variable(t=!0,e,n){return this.throwIfDisposed(),Ms().makeVariable(this,t,e,n)}};Object.defineProperty(Ft,Symbol.hasInstance,{value:r=>!!r&&r.data!=null&&r.dataSync!=null&&r.throwIfDisposed!=null});function O(){return Kd("Tensor",()=>Ft)}O();var Ka=class extends Ft{constructor(t,e,n,o){super(t.shape,t.dtype,t.dataId,o),this.trainable=e,this.name=n}assign(t){if(t.dtype!==this.dtype)throw new Error(`dtype of the new value (${t.dtype}) and previous value (${this.dtype}) must match`);if(!Dn(t.shape,this.shape))throw new Error(`shape of the new value (${t.shape}) and previous value (${this.shape}) must match`);Ms().disposeTensor(this),this.dataId=t.dataId,Ms().incRef(this,null)}dispose(){Ms().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(Ka,Symbol.hasInstance,{value:r=>r instanceof Ft&&r.assign!=null&&r.assign instanceof Function});var go={};Wt(go,{assertTypesMatch:()=>I0,getTensorsInContainer:()=>nh,isTensorInList:()=>y4,makeTypesMatch:()=>Ut});var x0;(function(r){r.R0="R0",r.R1="R1",r.R2="R2",r.R3="R3",r.R4="R4",r.R5="R5",r.R6="R6"})(x0||(x0={}));var y0;(function(r){r.float32="float32",r.int32="int32",r.bool="int32",r.complex64="complex64"})(y0||(y0={}));var b0;(function(r){r.float32="float32",r.int32="int32",r.bool="bool",r.complex64="complex64"})(b0||(b0={}));var w0;(function(r){r.float32="float32",r.int32="float32",r.bool="float32",r.complex64="complex64"})(w0||(w0={}));var C0;(function(r){r.float32="complex64",r.int32="complex64",r.bool="complex64",r.complex64="complex64"})(C0||(C0={}));var x4={float32:w0,int32:y0,bool:b0,complex64:C0};function sr(r,t){if(r==="string"||t==="string"){if(r==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${r} with ${t}`)}return x4[r][t]}function Wu(r){return sr(r,"int32")}function Ut(r,t){if(r.dtype===t.dtype)return[r,t];let e=sr(r.dtype,t.dtype);return[r.cast(e),t.cast(e)]}function I0(r,t){E(r.dtype===t.dtype,()=>`The dtypes of the first(${r.dtype}) and second(${t.dtype}) input must match`)}function y4(r,t){return t.some(e=>e.id===r.id)}function nh(r){let t=[];return M1(r,t,new Set),t}function M1(r,t,e){if(r==null)return;if(r instanceof Ft){t.push(r);return}if(!b4(r))return;let n=r;for(let o in n){let s=n[o];e.has(s)||(e.add(s),M1(s,t,e))}}function b4(r){return Array.isArray(r)||typeof r=="object"}function S0(r){return r.kernelName!=null}var Ug=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(t=>t.name)))}}}dispose(){for(let t in this.registeredVariables)this.registeredVariables[t].dispose()}},ql=class{constructor(t){this.ENV=t,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new Ug}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let t=this.getSortedBackends();for(let e=0;e<t.length;e++){let n=t[e];if(await this.initializeBackend(n).success){await this.setBackend(n);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:t,asyncInit:e}=this.initializeBackendsAndReturnBest();if(e)throw new Error(`The highest priority backend '${t}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(t)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(t){if(!(t in this.registry))if(t in this.registryFactory){let{asyncInit:e}=this.initializeBackend(t);if(e)return null}else return null;return this.registry[t]}findBackendFactory(t){return t in this.registryFactory?this.registryFactory[t].factory:null}registerBackend(t,e,n=1){return t in this.registryFactory?(vi(`${t} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[t]={factory:e,priority:n},!0)}async setBackend(t){if(this.registryFactory[t]==null)throw new Error(`Backend name '${t}' not found in registry`);if(this.backendName=t,this.registry[t]==null){this.backendInstance=null;let{success:e,asyncInit:n}=this.initializeBackend(t);if(!(n?await e:e))return!1}return this.backendInstance=this.registry[t],this.setupRegisteredKernels(),this.profiler=new Gg(this.backendInstance),!0}setupRegisteredKernels(){zg(this.backendName).forEach(e=>{e.setupFunc!=null&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(t){zg(t).forEach(n=>{n.disposeFunc!=null&&n.disposeFunc(this.registry[t])})}initializeBackend(t){let e=this.registryFactory[t];if(e==null)throw new Error(`Cannot initialize backend ${t}, no registration found.`);try{let n=e.factory();if(n&&!(n instanceof zo)&&typeof n.then=="function"){let o=++this.pendingBackendInitId,s=n.then(i=>o<this.pendingBackendInitId?!1:(this.registry[t]=i,this.pendingBackendInit=null,!0)).catch(i=>(o<this.pendingBackendInitId||(this.pendingBackendInit=null,vi(`Initialization of backend ${t} failed`),vi(i.stack||i.message)),!1));return this.pendingBackendInit=s,{success:s,asyncInit:!0}}else return this.registry[t]=n,{success:!0,asyncInit:!1}}catch(n){return vi(`Initialization of backend ${t} failed`),vi(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(t){if(!(t in this.registryFactory))throw new Error(`${t} backend not found in registry`);this.backendName===t&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,t in this.registry&&(this.disposeRegisteredKernels(t),this.registry[t].dispose(),delete this.registry[t]),delete this.registryFactory[t],this.backendName===t&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((t,e)=>this.registryFactory[e].priority-this.registryFactory[t].priority)}initializeBackendsAndReturnBest(){let t=this.getSortedBackends();for(let e=0;e<t.length;e++){let n=t[e],{success:o,asyncInit:s}=this.initializeBackend(n);if(s||o)return{name:n,asyncInit:s}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(t,e){let n=this.state.tensorInfo.get(e),o=n.backend,s=this.readSync(e),i=o.refCount(e);o.disposeData(e,!0),n.backend=t,t.move(e,s,n.shape,n.dtype,i),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(t,e){let n=null;if(e==null){if(typeof t!="function")throw new Error("Please provide a function to tidy()");e=t}else{if(typeof t!="string"&&!(t instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if(typeof e!="function")throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");n=t}let o;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(o),()=>(o=e(),o instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),o))}scopedRun(t,e,n){t();try{let o=n();return e(),o}catch(o){throw e(),o}}nextTensorId(){return ql.nextTensorId++}nextVariableId(){return ql.nextVariableId++}clone(t){let e=k.runKernel(co,{x:t}),n={x:t},o=i=>({x:()=>{let a="float32",u={x:i},l={dtype:a};return k.runKernel(lo,u,l)}}),s=[];return this.addTapeNode(this.state.activeScope.name,n,[e],o,s,{}),e}runKernel(t,e,n){if(this.backendName==null&&this.backend,!(Jd(t,this.backendName)!=null))throw new Error(`Kernel '${t}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:t,inputs:e,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(t,e,n){let o=this.backend.numDataIds(),s=0;n.forEach(u=>{s+=u.dtype==="complex64"?3:1});let i=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],a=o-e-s-i;if(a>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${a} data ids) after running '${t}'`)}runKernelFunc(t){let e,n=[],o=this.isTapeOn(),s=this.state.numBytes,i=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let a;this.backendName==null&&this.backend;let u,l=S0(t)?t.kernelName:this.state.activeScope!=null?this.state.activeScope.name:"";if(S0(t)){let{kernelName:d,inputs:h,attrs:g}=t;this.backendName==null&&this.backend;let x=Jd(d,this.backendName);E(x!=null,()=>`Cannot find registered kernel '${d}' for backend '${this.backendName}'`),a=()=>{let b=this.backend.numDataIds();u=x.kernelFunc({inputs:h,attrs:g,backend:this.backend});let w=Array.isArray(u)?u:[u];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(d,b,w);let C=w.map(N=>N.rank!=null?N:this.makeTensorFromTensorInfo(N));if(o){let N=this.getTensorsForGradient(d,h,C);n=this.saveTensorsForBackwardMode(N)}return C}}else{let{forwardFunc:d}=t,h=g=>{!o||(n=g.map(x=>this.keep(this.clone(x))))};a=()=>{let g=this.backend.numDataIds();u=this.tidy(()=>d(this.backend,h));let x=Array.isArray(u)?u:[u];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,g,x),x}}let{inputs:c,attrs:p}=t,m=S0(t)?null:t.backwardsFunc,f;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?e=a():(f=this.profiler.profileKernel(l,c,()=>a()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(f),e=f.outputs)}),o&&this.addTapeNode(l,c,e,m,n,p),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-s,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-i,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(c).map(d=>c[d]!=null?c[d].shape:null),outputShapes:e.map(d=>d.shape),kernelTimeMs:f.timeMs,extraInfo:f.extraInfo}),Array.isArray(u)?e:e[0]}saveTensorsForBackwardMode(t){return t.map(n=>this.keep(this.clone(n)))}getTensorsForGradient(t,e,n){let o=u0(t);if(o!=null){let s=o.inputsToSave||[],i=o.outputsToSave||[],a;o.saveAllInputs?(E(Array.isArray(e),()=>"saveAllInputs is true, expected inputs to be an array."),a=Object.keys(e).map(l=>e[l])):a=s.map(l=>e[l]);let u=n.filter((l,c)=>i[c]);return a.concat(u)}return[]}makeTensor(t,e,n,o){if(t==null)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",o=o||this.backend;let s=t;n==="string"&&Vo(t[0])&&(s=t.map(u=>Hl(u)));let i=o.write(s,e,n),a=new Ft(e,n,i,this.nextTensorId());if(this.trackTensor(a,o),n==="string"){let u=this.state.tensorInfo.get(i),l=s0(s);this.state.numBytes+=l-u.bytes,u.bytes=l}return a}makeTensorFromDataId(t,e,n,o){n=n||"float32";let s={dataId:t,shape:e,dtype:n};return this.makeTensorFromTensorInfo(s,o)}makeTensorFromTensorInfo(t,e){let{dataId:n,shape:o,dtype:s}=t,i=new Ft(o,s,n,this.nextTensorId());return this.trackTensor(i,e),i}makeVariable(t,e=!0,n,o){n=n||this.nextVariableId().toString(),o!=null&&o!==t.dtype&&(t=t.cast(o));let s=new Ka(t,e,n,this.nextTensorId());if(this.state.registeredVariables[s.name]!=null)throw new Error(`Variable with name ${s.name} was already registered`);return this.state.registeredVariables[s.name]=s,this.incRef(s,this.backend),s}trackTensor(t,e){this.state.numTensors++,t.dtype==="string"&&this.state.numStringTensors++;let n=0;t.dtype!=="complex64"&&t.dtype!=="string"&&(n=t.size*Mg(t.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(t.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(t.dataId,{backend:e||this.backend,dtype:t.dtype,shape:t.shape,bytes:n})),t instanceof Ka||this.track(t)}incRef(t,e){this.trackTensor(t,e),this.backend.incRef(t.dataId)}removeDataId(t,e){this.state.tensorInfo.has(t)&&this.state.tensorInfo.get(t).backend===e&&(this.state.tensorInfo.delete(t),this.state.numDataBuffers--)}disposeTensor(t){if(!this.state.tensorInfo.has(t.dataId))return;let e=this.state.tensorInfo.get(t.dataId);if(this.state.numTensors--,t.dtype==="string"&&(this.state.numStringTensors--,this.state.numBytes-=e.bytes),t.dtype!=="complex64"&&t.dtype!=="string"){let n=t.size*Mg(t.dtype);this.state.numBytes-=n}e.backend.disposeData(t.dataId)&&this.removeDataId(t.dataId,e.backend)}disposeVariables(){for(let t in this.state.registeredVariables){let e=this.state.registeredVariables[t];this.disposeVariable(e)}}disposeVariable(t){this.disposeTensor(t),this.state.registeredVariables[t.name]!=null&&delete this.state.registeredVariables[t.name]}memory(){let t=this.backend.memory();return t.numTensors=this.state.numTensors,t.numDataBuffers=this.state.numDataBuffers,t.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(t.unreliable=!0,t.reasons==null&&(t.reasons=[]),t.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),t}async profile(t){this.state.profiling=!0;let e=this.state.numBytes,n=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await t(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(o=>o.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-e,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let o of this.state.activeProfile.kernels)o.kernelTimeMs=await o.kernelTimeMs,o.extraInfo=await o.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(t,e,n,o,s,i){let a={id:this.state.nextTapeNodeId++,kernelName:t,inputs:e,outputs:n,saved:s},u=u0(t);u!=null&&(o=u.gradFunc),o!=null&&(a.gradient=l=>(l=l.map((c,p)=>{if(c==null){let m=n[p],f=ip(m.size,m.dtype);return this.makeTensor(f,m.shape,m.dtype)}return c}),o(l.length>1?l:l[0],s,i))),this.state.activeTape.push(a)}keep(t){return t.kept=!0,t}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(t){let e={track:[],name:"unnamed scope",id:this.state.nextScopeId++};t&&(e.name=t),this.state.scopeStack.push(e),this.state.activeScope=e}endScope(t){let e=nh(t),n=new Set(e.map(s=>s.id));for(let s=0;s<this.state.activeScope.track.length;s++){let i=this.state.activeScope.track[s];!i.kept&&!n.has(i.id)&&i.dispose()}let o=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],e.forEach(s=>{!s.kept&&s.scopeId===o.id&&this.track(s)})}gradients(t,e,n,o=!1){if(E(e.length>0,()=>"gradients() received an empty list of xs."),n!=null&&n.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let s=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",t));E(s instanceof Ft,()=>"The result y returned by f() must be a tensor.");let i=A1(this.state.activeTape,e,s);if(!o&&i.length===0&&e.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let a={};a[s.id]=n==null?w4(s.shape):n,$1(a,i,l=>this.tidy(l),C4);let u=e.map(l=>a[l.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(l=>{for(let c of l.saved)c.dispose()}),this.state.activeTape=null),{value:s,grads:u}})}customGrad(t){return E(oi(t),()=>"The f passed in customGrad(f) must be a function."),(...e)=>{E(e.every(a=>a instanceof Ft),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n,o={};e.forEach((a,u)=>{o[u]=a});let s=(a,u)=>(n=t(...e,u),E(n.value instanceof Ft,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),E(oi(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),i=(a,u)=>{let l=n.gradFunc(a,u),c=Array.isArray(l)?l:[l];E(c.length===e.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),E(c.every(m=>m instanceof Ft),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let p={};return c.forEach((m,f)=>{p[f]=()=>m}),p};return this.runKernelFunc({forwardFunc:s,backwardsFunc:i,inputs:o})}}readSync(t){return this.state.tensorInfo.get(t).backend.readSync(t)}read(t){return this.state.tensorInfo.get(t).backend.read(t)}readToGPU(t,e){return this.state.tensorInfo.get(t).backend.readToGPU(t,e)}async time(t){let e=Gu(),n=await this.backend.time(t);return n.wallMs=Gu()-e,n}track(t){return this.state.activeScope!=null&&(t.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(t)),t}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new Ug;for(let t in this.registry)this.disposeRegisteredKernels(t),this.registry[t].dispose(),delete this.registry[t];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};ql.nextTensorId=0;ql.nextVariableId=0;function w4(r){let t=Wd(Jt(r),"float32");return k.makeTensor(t,r,"float32")}function v0(){let r=l0();if(r._tfengine==null){let t=new qd(r);r._tfengine=new ql(t)}return c1(r._tfengine.ENV),O1(()=>r._tfengine),r._tfengine}var k=v0();function C4(r,t){let e={a:r,b:t};return k.runKernel(Zn,e)}var Kl={};Wt(Kl,{isBrowser:()=>T0,isMobile:()=>v4,mockIsMobile:()=>S4});function I4(){return typeof navigator!="undefined"&&navigator!=null}var N0;function S4(r){N0=r}function v4(r){if(N0!==void 0)return N0;if(r||I4()){if(r||(r=navigator),r.product==="ReactNative")return!0;let t=r.userAgent||r.vendor||(typeof window!="undefined"?window.opera:"");if(!t){let e=r;return e.userAgentData&&e.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function T0(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}var On=z();On.registerFlag("DEBUG",()=>!1,r=>{r&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")});On.registerFlag("IS_BROWSER",()=>T0());On.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");On.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));On.registerFlag("PROD",()=>!1);On.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>On.getBool("DEBUG"));On.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);On.registerFlag("IS_TEST",()=>!1);On.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0);On.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);On.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);On.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU",()=>!1);On.registerFlag("USE_SETTIMEOUTCUSTOM",()=>!1);function Br(r,t){let e=r;if(xr(r))return t==="string"?[]:[r.length];if(typeof r=="object"&&"texture"in r){let o=r.channels||"RGBA";return[r.height,r.width*o.length]}if(!Array.isArray(r))return[];let n=[];for(;Array.isArray(e)||xr(e)&&t!=="string";)n.push(e.length),e=e[0];return Array.isArray(r)&&z().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&B1(r,n,[]),n}function B1(r,t,e){if(e=e||[],!Array.isArray(r)&&!xr(r)){E(t.length===0,()=>`Element arr[${e.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}E(t.length>0,()=>`Element arr[${e.join("][")}] should be a primitive, but is an array of ${r.length} elements`),E(r.length===t[0],()=>`Element arr[${e.join("][")}] should have ${t[0]} elements, but has ${r.length} elements`);let n=t.slice(1);for(let o=0;o<r.length;++o)B1(r[o],n,e.concat(o))}function z1(r,t,e,n){if(r!=="string_or_numeric"){if(r==null)throw new Error("Expected dtype cannot be null.");if(r!=="numeric"&&r!==t||r==="numeric"&&t==="string")throw new Error(`Argument '${e}' passed to '${n}' must be ${r} tensor, but got ${t} tensor`)}}function I(r,t,e,n="numeric"){if(r instanceof Ft)return z1(n,r.dtype,t,e),r;let o=op(r);if(o!=="string"&&["bool","int32","float32"].indexOf(n)>=0&&(o=n),z1(n,o,t,e),r==null||!xr(r)&&!Array.isArray(r)&&typeof r!="number"&&typeof r!="boolean"&&typeof r!="string"){let u=r==null?"null":r.constructor.name;throw new Error(`Argument '${t}' passed to '${e}' must be a Tensor or TensorLike, but got '${u}'`)}let s=Br(r,o);!xr(r)&&!Array.isArray(r)&&(r=[r]);let a=o!=="string"?Gp(r,o):Bo(r,[],!0);return k.makeTensor(a,s,o)}function ja(r,t,e,n="numeric"){if(!Array.isArray(r))throw new Error(`Argument ${t} passed to ${e} must be a \`Tensor[]\` or \`TensorLike[]\``);return r.map((s,i)=>I(s,`${t}[${i}]`,e,n))}var k0="__op";function T(r){let t=Object.keys(r);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let e=t[0],n=r[e];e.endsWith("_")&&(e=e.substring(0,e.length-1)),e=e+k0;let o=(...s)=>{k.startScope(e);try{let i=n(...s);return Hd(i)&&console.error("Cannot return a Promise inside of tidy."),k.endScope(i),i}catch(i){throw k.endScope(null),i}};return Object.defineProperty(o,"name",{value:e,configurable:!0}),o}function N4(r,t){let e=I(r,"real","complex"),n=I(t,"imag","complex");$e(e.shape,n.shape,`real and imag shapes, ${e.shape} and ${n.shape}, must match in call to tf.complex().`);let o={real:e,imag:n};return k.runKernel(pp,o)}var wn=T({complex_:N4});function on(r,t,e,n){if(n==null&&(n=op(r)),n==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(typeof r=="object"&&"texture"in r){if(n!=="float32"&&n!=="int32")throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${n}.`);return r.channels=r.channels||"RGBA",k.backend.createTensorFromTexture(r,t||e,n)}if(!xr(r)&&!Array.isArray(r)&&typeof r!="number"&&typeof r!="boolean"&&typeof r!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){Ud(t);let o=Jt(t),s=Jt(e);E(o===s,()=>`Based on the provided shape, [${t}], the tensor should have ${o} values but has ${s}`);for(let i=0;i<e.length;++i){let a=e[i],u=i===e.length-1?a!==Jt(t.slice(i)):!0;E(e[i]===t[i]||!u,()=>`Error creating a new Tensor. Inferred shape (${e}) does not match the provided shape (${t}). `)}}return!xr(r)&&!Array.isArray(r)&&(r=[r]),t=t||e,r=n!=="string"?Gp(r,n):Bo(r,[],!0),k.makeTensor(r,t,n)}function ur(r,t,e){let n=Br(r,e);return on(r,t,n,e)}var oh={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8};var Hg=4;async function G1(r,t){let e=[],n=[],o=Array.isArray(r)?r.map(i=>i.name):Object.keys(r);for(let i=0;i<o.length;++i){let a=o[i],u=Array.isArray(r)?r[i].tensor:r[a];if(u.dtype!=="float32"&&u.dtype!=="int32"&&u.dtype!=="bool"&&u.dtype!=="string"&&u.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${a}': ${u.dtype}`);let l={name:a,shape:u.shape,dtype:u.dtype};if(u.dtype==="string"){let c=new Promise(async p=>{let m=await u.bytes(),f=m.reduce((g,x)=>g+x.length,0)+Hg*m.length,d=new Uint8Array(f),h=0;for(let g=0;g<m.length;g++){let x=m[g],b=new Uint8Array(new Uint32Array([x.length]).buffer);d.set(b,h),h+=Hg,d.set(x,h),h+=x.length}p(d)});n.push(c)}else n.push(u.data());t!=null&&(l.group=t),e.push(l)}let s=await Promise.all(n);return{data:T4(s),specs:e}}function qg(r,t){let e={},n,o=0;for(let s of t){let i=s.name,a=s.dtype,u=s.shape,l=Jt(u),c;if("quantization"in s){let p=s.quantization;if(p.dtype==="uint8"||p.dtype==="uint16"){if(!("min"in p&&"scale"in p))throw new Error(`Weight ${s.name} with quantization ${p.dtype} doesn't have corresponding metadata min and scale.`)}else if(p.dtype==="float16"){if(a!=="float32")throw new Error(`Weight ${s.name} is quantized with ${p.dtype} which only supports weights of type float32 not ${a}.`)}else throw new Error(`Weight ${s.name} has unknown quantization dtype ${p.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let m=oh[p.dtype],f=r.slice(o,o+l*m),d=p.dtype==="uint8"?new Uint8Array(f):new Uint16Array(f);if(a==="float32")if(p.dtype==="uint8"||p.dtype==="uint16"){c=new Float32Array(d.length);for(let h=0;h<d.length;h++){let g=d[h];c[h]=g*p.scale+p.min}}else if(p.dtype==="float16")n===void 0&&(n=A4()),c=n(d);else throw new Error(`Unsupported quantization type ${p.dtype} for weight type float32.`);else if(a==="int32"){if(p.dtype!=="uint8"&&p.dtype!=="uint16")throw new Error(`Unsupported quantization type ${p.dtype} for weight type int32.`);c=new Int32Array(d.length);for(let h=0;h<d.length;h++){let g=d[h];c[h]=Math.round(g*p.scale+p.min)}}else throw new Error(`Unsupported dtype in weight '${i}': ${a}`);o+=l*m}else if(a==="string"){let p=Jt(s.shape);c=[];for(let m=0;m<p;m++){let f=new Uint32Array(r.slice(o,o+Hg))[0];o+=Hg;let d=new Uint8Array(r.slice(o,o+f));c.push(d),o+=f}}else{let p=oh[a],m=r.slice(o,o+l*p);if(a==="float32")c=new Float32Array(m);else if(a==="int32")c=new Int32Array(m);else if(a==="bool")c=new Uint8Array(m);else if(a==="complex64"){c=new Float32Array(m);let f=new Float32Array(c.length/2),d=new Float32Array(c.length/2);for(let x=0;x<f.length;x++)f[x]=c[x*2],d[x]=c[x*2+1];let h=ur(f,u,"float32"),g=ur(d,u,"float32");e[i]=wn(h,g),h.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${i}': ${a}`);o+=l*p}a!=="complex64"&&(e[i]=ur(c,u,a))}return e}function T4(r){if(r===null)throw new Error(`Invalid input value: ${JSON.stringify(r)}`);let t=0,e=[];r.forEach(s=>{if(t+=s.byteLength,e.push(s.byteLength===s.buffer.byteLength?s:new s.constructor(s)),!(s instanceof Float32Array||s instanceof Int32Array||s instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${s.constructor.name}`)});let n=new Uint8Array(t),o=0;return e.forEach(s=>{n.set(new Uint8Array(s.buffer),o),o+=s.byteLength}),n.buffer}var E0=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function V1(r){return E0?Buffer.byteLength(r):new Blob([r]).size}function W1(r){if(E0)return Buffer.from(r).toString("base64");let t=new Uint8Array(r),e="";for(let n=0,o=t.length;n<o;n++)e+=String.fromCharCode(t[n]);return btoa(e)}function U1(r){if(E0){let n=Buffer.from(r,"base64");return n.buffer.slice(n.byteOffset,n.byteOffset+n.byteLength)}let t=atob(r),e=new Uint8Array(t.length);for(let n=0;n<t.length;++n)e.set([t.charCodeAt(n)],n);return e.buffer}function Hp(r){if(r.length===1)return r[0];let t=0;r.forEach(o=>{t+=o.byteLength});let e=new Uint8Array(t),n=0;return r.forEach(o=>{e.set(new Uint8Array(o),n),n+=o.byteLength}),e.buffer}function _0(r){let t="/";for(r=r.trim();r.endsWith(t);)r=r.slice(0,r.length-1);let e=r.split(t);return e[e.length-1]}function Kg(r,t){let e={modelTopology:r.modelTopology,format:r.format,generatedBy:r.generatedBy,convertedBy:r.convertedBy,weightsManifest:t};return r.signature!=null&&(e.signature=r.signature),r.userDefinedMetadata!=null&&(e.userDefinedMetadata=r.userDefinedMetadata),r.modelInitializer!=null&&(e.modelInitializer=r.modelInitializer),r.initializerSignature!=null&&(e.initializerSignature=r.initializerSignature),r.trainingConfig!=null&&(e.trainingConfig=r.trainingConfig),e}function A0(r,t,e){let n={modelTopology:r.modelTopology,format:r.format,generatedBy:r.generatedBy,convertedBy:r.convertedBy};if(r.trainingConfig!=null&&(n.trainingConfig=r.trainingConfig),r.weightsManifest!=null){if(!t)throw new Error("modelJSON has weightsManifest but weightSpecs is null");if(!e)throw new Error("modelJSON has weightsManifest but weightData is null");n.weightSpecs=t,n.weightData=e}return r.signature!=null&&(n.signature=r.signature),r.userDefinedMetadata!=null&&(n.userDefinedMetadata=r.userDefinedMetadata),r.modelInitializer!=null&&(n.modelInitializer=r.modelInitializer),r.initializerSignature!=null&&(n.initializerSignature=r.initializerSignature),n}async function qp(r,t){let e,n;return r.weightsManifest!=null&&([e,n]=await t(r.weightsManifest)),A0(r,e,n)}function Ni(r){if(r.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:r.modelTopology==null?0:V1(JSON.stringify(r.modelTopology)),weightSpecsBytes:r.weightSpecs==null?0:V1(JSON.stringify(r.weightSpecs)),weightDataBytes:r.weightData==null?0:r.weightData.byteLength}}function jg(r){let t=[];for(let e of r)t.push(...e.weights);return t}function k4(){let r=e=>{let n=e<<13,o=0;for(;(n&8388608)===0;)o-=8388608,n<<=1;return n&=-8388609,o+=947912704,n|o},t=new Uint32Array(2048);t[0]=0;for(let e=1;e<1024;e++)t[e]=r(e);for(let e=1024;e<2048;e++)t[e]=939524096+(e-1024<<13);return t}function E4(){let r=new Uint32Array(64);r[0]=0,r[31]=1199570944,r[32]=2147483648,r[63]=3347054592;for(let t=1;t<31;t++)r[t]=t<<23;for(let t=33;t<63;t++)r[t]=2147483648+(t-32<<23);return r}function _4(){let r=new Uint32Array(64);for(let t=0;t<64;t++)r[t]=1024;return r[0]=r[32]=0,r}function A4(){let r=k4(),t=E4(),e=_4();return n=>{let o=new ArrayBuffer(4*n.length),s=new Uint32Array(o);for(let i=0;i<n.length;i++){let a=n[i],u=r[e[a>>10]+(a&1023)]+t[a>>10];s[i]=u}return new Float32Array(o)}}var Ce=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return Ce.instance==null&&(Ce.instance=new Ce),Ce.instance}static registerSaveRouter(t){Ce.getInstance().saveRouters.push(t)}static registerLoadRouter(t){Ce.getInstance().loadRouters.push(t)}static getSaveHandlers(t){return Ce.getHandlers(t,"save")}static getLoadHandlers(t,e){return Ce.getHandlers(t,"load",e)}static getHandlers(t,e,n){let o=[];return(e==="load"?Ce.getInstance().loadRouters:Ce.getInstance().saveRouters).forEach(i=>{let a=i(t,n);a!==null&&o.push(a)}),o}},H1=r=>Ce.registerSaveRouter(r),q1=r=>Ce.registerLoadRouter(r),K1=r=>Ce.getSaveHandlers(r),j1=(r,t)=>Ce.getLoadHandlers(r,t);var $0="tensorflowjs",D0=1,Uu="models_store",jl="model_info_store";function X1(){if(!z().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let r=typeof window=="undefined"?self:window,t=r.indexedDB||r.mozIndexedDB||r.webkitIndexedDB||r.msIndexedDB||r.shimIndexedDB;if(t==null)throw new Error("The current browser does not appear to support IndexedDB.");return t}function R0(r){let t=r.result;t.createObjectStore(Uu,{keyPath:"modelPath"}),t.createObjectStore(jl,{keyPath:"modelPath"})}var Ti=class{constructor(t){if(this.indexedDB=X1(),t==null||!t)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=t}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,t)}async load(){return this.databaseAction(this.modelPath)}databaseAction(t,e){return new Promise((n,o)=>{let s=this.indexedDB.open($0,D0);s.onupgradeneeded=()=>R0(s),s.onsuccess=()=>{let i=s.result;if(e==null){let a=i.transaction(Uu,"readonly"),l=a.objectStore(Uu).get(this.modelPath);l.onsuccess=()=>{if(l.result==null)return i.close(),o(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));n(l.result.modelArtifacts)},l.onerror=c=>(i.close(),o(l.error)),a.oncomplete=()=>i.close()}else{let a=Ni(e),u=i.transaction(jl,"readwrite"),l=u.objectStore(jl),c=l.put({modelPath:this.modelPath,modelArtifactsInfo:a}),p;c.onsuccess=()=>{p=i.transaction(Uu,"readwrite");let f=p.objectStore(Uu).put({modelPath:this.modelPath,modelArtifacts:e,modelArtifactsInfo:a});f.onsuccess=()=>n({modelArtifactsInfo:a}),f.onerror=d=>{l=u.objectStore(jl);let h=l.delete(this.modelPath);h.onsuccess=()=>(i.close(),o(f.error)),h.onerror=g=>(i.close(),o(f.error))}},c.onerror=m=>(i.close(),o(c.error)),u.oncomplete=()=>{p==null?i.close():p.oncomplete=()=>i.close()}}},s.onerror=i=>o(s.error)})}};Ti.URL_SCHEME="indexeddb://";var Y1=r=>z().getBool("IS_BROWSER")&&!Array.isArray(r)&&r.startsWith(Ti.URL_SCHEME)?$4(r.slice(Ti.URL_SCHEME.length)):null;Ce.registerSaveRouter(Y1);Ce.registerLoadRouter(Y1);function $4(r){return new Ti(r)}function D4(r){return r.startsWith(Ti.URL_SCHEME)?r.slice(Ti.URL_SCHEME.length):r}var Xg=class{constructor(){this.indexedDB=X1()}async listModels(){return new Promise((t,e)=>{let n=this.indexedDB.open($0,D0);n.onupgradeneeded=()=>R0(n),n.onsuccess=()=>{let o=n.result,s=o.transaction(jl,"readonly"),a=s.objectStore(jl).getAll();a.onsuccess=()=>{let u={};for(let l of a.result)u[l.modelPath]=l.modelArtifactsInfo;t(u)},a.onerror=u=>(o.close(),e(a.error)),s.oncomplete=()=>o.close()},n.onerror=o=>e(n.error)})}async removeModel(t){return t=D4(t),new Promise((e,n)=>{let o=this.indexedDB.open($0,D0);o.onupgradeneeded=()=>R0(o),o.onsuccess=()=>{let s=o.result,i=s.transaction(jl,"readwrite"),a=i.objectStore(jl),u=a.get(t),l;u.onsuccess=()=>{if(u.result==null)return s.close(),n(new Error(`Cannot find model with path '${t}' in IndexedDB.`));{let c=a.delete(t),p=()=>{l=s.transaction(Uu,"readwrite");let f=l.objectStore(Uu).delete(t);f.onsuccess=()=>e(u.result.modelArtifactsInfo),f.onerror=d=>n(u.error)};c.onsuccess=p,c.onerror=m=>(p(),s.close(),n(u.error))}},u.onerror=c=>(s.close(),n(u.error)),i.oncomplete=()=>{l==null?s.close():l.oncomplete=()=>s.close()}},o.onerror=s=>n(o.error)})}};var Xa="/",Kp="tensorflowjs_models",Z1="info",R4="model_topology",F4="weight_specs",O4="weight_data",P4="model_metadata";function J1(r){return{info:[Kp,r,Z1].join(Xa),topology:[Kp,r,R4].join(Xa),weightSpecs:[Kp,r,F4].join(Xa),weightData:[Kp,r,O4].join(Xa),modelMetadata:[Kp,r,P4].join(Xa)}}function Q1(r){for(let t of Object.values(r))window.localStorage.removeItem(t)}function L4(r){let t=r.split(Xa);if(t.length<3)throw new Error(`Invalid key format: ${r}`);return t.slice(1,t.length-1).join(Xa)}function M4(r){return r.startsWith(ki.URL_SCHEME)?r.slice(ki.URL_SCHEME.length):r}var ki=class{constructor(t){if(!z().getBool("IS_BROWSER")||typeof window=="undefined"||typeof window.localStorage=="undefined")throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,t==null||!t)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=t,this.keys=J1(this.modelPath)}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let e=JSON.stringify(t.modelTopology),n=JSON.stringify(t.weightSpecs),o=Ni(t);try{this.LS.setItem(this.keys.info,JSON.stringify(o)),this.LS.setItem(this.keys.topology,e),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,W1(t.weightData));let s={format:t.format,generatedBy:t.generatedBy,convertedBy:t.convertedBy,signature:t.signature!=null?t.signature:void 0,userDefinedMetadata:t.userDefinedMetadata!=null?t.userDefinedMetadata:void 0,modelInitializer:t.modelInitializer!=null?t.modelInitializer:void 0,initializerSignature:t.initializerSignature!=null?t.initializerSignature:void 0,trainingConfig:t.trainingConfig!=null?t.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(s)),{modelArtifactsInfo:o}}catch(s){throw Q1(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${o.modelTopologyBytes}, weightSpecsBytes=${o.weightSpecsBytes}, weightDataBytes=${o.weightDataBytes}.`)}}}async load(){let t=JSON.parse(this.LS.getItem(this.keys.info));if(t==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(t.modelTopologyType!=="JSON")throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let e={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(n==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);e.modelTopology=n;let o=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(o==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);e.weightSpecs=o;let s=this.LS.getItem(this.keys.modelMetadata);if(s!=null){let a=JSON.parse(s);e.format=a.format,e.generatedBy=a.generatedBy,e.convertedBy=a.convertedBy,a.signature!=null&&(e.signature=a.signature),a.userDefinedMetadata!=null&&(e.userDefinedMetadata=a.userDefinedMetadata),a.modelInitializer!=null&&(e.modelInitializer=a.modelInitializer),a.initializerSignature!=null&&(e.initializerSignature=a.initializerSignature),a.trainingConfig!=null&&(e.trainingConfig=a.trainingConfig)}let i=this.LS.getItem(this.keys.weightData);if(i==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return e.weightData=U1(i),e}};ki.URL_SCHEME="localstorage://";var tE=r=>z().getBool("IS_BROWSER")&&!Array.isArray(r)&&r.startsWith(ki.URL_SCHEME)?z4(r.slice(ki.URL_SCHEME.length)):null;Ce.registerSaveRouter(tE);Ce.registerLoadRouter(tE);function z4(r){return new ki(r)}var Yg=class{constructor(){E(z().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),E(typeof window=="undefined"||typeof window.localStorage!="undefined",()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let t={},e=Kp+Xa,n=Xa+Z1;for(let o=0;o<this.LS.length;++o){let s=this.LS.key(o);if(s.startsWith(e)&&s.endsWith(n)){let i=L4(s);t[i]=JSON.parse(this.LS.getItem(s))}}return t}async removeModel(t){t=M4(t);let e=J1(t);if(this.LS.getItem(e.info)==null)throw new Error(`Cannot find model at path '${t}'`);let n=JSON.parse(this.LS.getItem(e.info));return Q1(e),n}};var jp="://",Er=class{constructor(){this.managers={}}static getInstance(){return Er.instance==null&&(Er.instance=new Er),Er.instance}static registerManager(t,e){E(t!=null,()=>"scheme must not be undefined or null."),t.endsWith(jp)&&(t=t.slice(0,t.indexOf(jp))),E(t.length>0,()=>"scheme must not be an empty string.");let n=Er.getInstance();E(n.managers[t]==null,()=>`A model store manager is already registered for scheme '${t}'.`),n.managers[t]=e}static getManager(t){let e=Er.getInstance().managers[t];if(e==null)throw new Error(`Cannot find model manager for scheme '${t}'`);return e}static getSchemes(){return Object.keys(Er.getInstance().managers)}};function Zg(r){if(r.indexOf(jp)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Er.getSchemes().join(",")}`);return{scheme:r.split(jp)[0],path:r.split(jp)[1]}}async function eE(r,t,e=!1){E(r!==t,()=>`Old path and new path are the same: '${r}'`);let n=Ce.getLoadHandlers(r);E(n.length>0,()=>`Copying failed because no load handler is found for source URL ${r}.`),E(n.length<2,()=>`Copying failed because more than one (${n.length}) load handlers for source URL ${r}.`);let o=n[0],s=Ce.getSaveHandlers(t);E(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),E(s.length<2,()=>`Copying failed because more than one (${n.length}) save handlers for destination URL ${t}.`);let i=s[0],a=Zg(r).scheme,u=Zg(r).path,l=a===Zg(r).scheme,c=await o.load();e&&l&&await Er.getManager(a).removeModel(u);let p=await i.save(c);return e&&!l&&await Er.getManager(a).removeModel(u),p.modelArtifactsInfo}async function rE(){let r=Er.getSchemes(),t={};for(let e of r){let n=await Er.getManager(e).listModels();for(let o in n){let s=e+jp+o;t[s]=n[o]}}return t}async function nE(r){let t=Zg(r);return Er.getManager(t.scheme).removeModel(t.path)}async function oE(r,t){return eE(r,t,!1)}async function sE(r,t){return eE(r,t,!0)}var F0=class{constructor(){this.messageName="setTimeoutCustom",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(t,e){return fetch(t,e)}now(){return performance.now()}encode(t,e){if(e!=="utf-8"&&e!=="utf8")throw new Error(`Browser's encoder only supports utf-8, but got ${e}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(t)}decode(t,e){return new TextDecoder(e).decode(t)}setTimeoutCustom(t,e){if(typeof window=="undefined"||!z().getBool("USE_SETTIMEOUTCUSTOM")){setTimeout(t,e);return}this.functionRefs.push(t),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},"*")},e),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener("message",n=>{if(n.source===window&&n.data.name===this.messageName){n.stopPropagation();let o=this.functionRefs[n.data.index];o(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0)}},!0))}};if(z().get("IS_BROWSER")){z().setPlatform("browser",new F0);try{Er.registerManager(ki.URL_SCHEME,new Yg)}catch(r){}try{Er.registerManager(Ti.URL_SCHEME,new Xg)}catch(r){}}var B4={importFetch:()=>iE()},O0;var P0=class{constructor(){this.util=aE(),this.textEncoder=new this.util.TextEncoder}fetch(t,e){return z().global.fetch!=null?z().global.fetch(t,e):(O0==null&&(O0=B4.importFetch()),O0(t,e))}now(){let t=process.hrtime();return t[0]*1e3+t[1]/1e6}encode(t,e){if(e!=="utf-8"&&e!=="utf8")throw new Error(`Node built-in encoder only supports utf-8, but got ${e}`);return this.textEncoder.encode(t)}decode(t,e){return t.length===0?"":new this.util.TextDecoder(e).decode(t)}};z().get("IS_NODE")&&!z().get("IS_BROWSER")&&z().setPlatform("node",new P0);function wt(r,t="float32",e){return t=t||"float32",Ud(r),new pe(r,t,e)}function V4(r,t){let e=I(r,"x","cast");if(!o0(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&e.dtype!=="string"||t!=="string"&&e.dtype==="string")throw new Error("Only strings can be casted to strings");let n={x:e},o={dtype:t};return k.runKernel(lo,n,o)}var J=T({cast_:V4});function G4(r){let e={x:I(r,"x","clone","string_or_numeric")};return k.runKernel(co,e)}var sn=T({clone_:G4});function Jg(r,t=!1){console.log(r.toString(t))}v0();var W4={buffer:wt,cast:J,clone:sn,print:Jg};P1(W4);var _r={};Wt(_r,{browserFiles:()=>uE,browserHTTPRequest:()=>mE,concatenateArrayBuffers:()=>Hp,copyModel:()=>oE,decodeWeights:()=>qg,encodeWeights:()=>G1,fromMemory:()=>fE,fromMemorySync:()=>G0,getLoadHandlers:()=>j1,getModelArtifactsForJSON:()=>qp,getModelArtifactsForJSONSync:()=>A0,getModelArtifactsInfoForJSON:()=>Ni,getSaveHandlers:()=>K1,getWeightSpecs:()=>jg,http:()=>tx,isHTTPScheme:()=>Qg,listModels:()=>rE,loadWeights:()=>cE,moveModel:()=>sE,registerLoadRouter:()=>q1,registerSaveRouter:()=>H1,removeModel:()=>nE,weightsLoaderFactory:()=>B0,withSaveHandler:()=>dE,withSaveHandlerSync:()=>hE});var U4="model",H4=".json",q4=".weights.bin";function lE(r){return new Promise(t=>setTimeout(t)).then(r)}var Ya=class{constructor(t){if(!z().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");t.startsWith(Ya.URL_SCHEME)&&(t=t.slice(Ya.URL_SCHEME.length)),(t==null||t.length===0)&&(t=U4),this.modelJsonFileName=t+H4,this.weightDataFileName=t+q4}async save(t){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");let e=window.URL.createObjectURL(new Blob([t.weightData],{type:"application/octet-stream"}));if(t.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let n=[{paths:["./"+this.weightDataFileName],weights:t.weightSpecs}],o=Kg(t,n),s=window.URL.createObjectURL(new Blob([JSON.stringify(o)],{type:"application/json"})),i=this.modelJsonAnchor==null?document.createElement("a"):this.modelJsonAnchor;if(i.download=this.modelJsonFileName,i.href=s,await lE(()=>i.dispatchEvent(new MouseEvent("click"))),t.weightData!=null){let a=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;a.download=this.weightDataFileName,a.href=e,await lE(()=>a.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Ni(t)}}}};Ya.URL_SCHEME="downloads://";var L0=class{constructor(t){if(t==null||t.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${t}`);this.jsonFile=t[0],this.weightsFiles=t.slice(1)}async load(){return new Promise((t,e)=>{let n=new FileReader;n.onload=o=>{let s=JSON.parse(o.target.result),i=s.modelTopology;if(i==null){e(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(s.weightsManifest==null){e(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){t({modelTopology:i});return}let u=qp(s,l=>this.loadWeights(l));t(u)},n.onerror=o=>e(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(t){let e=[],n=[];for(let i of t)e.push(...i.weights),n.push(...i.paths);let o=this.checkManifestAndWeightFiles(t),s=n.map(i=>this.loadWeightsFile(i,o[i]));return Promise.all(s).then(i=>[e,Hp(i)])}loadWeightsFile(t,e){return new Promise((n,o)=>{let s=new FileReader;s.onload=i=>{let a=i.target.result;n(a)},s.onerror=i=>o(`Failed to weights data from file of path '${t}'.`),s.readAsArrayBuffer(e)})}checkManifestAndWeightFiles(t){let e=[],n=this.weightsFiles.map(s=>_0(s.name)),o={};for(let s of t)s.paths.forEach(i=>{let a=_0(i);if(e.indexOf(a)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${a}'`);if(e.push(a),n.indexOf(a)===-1)throw new Error(`Weight file with basename '${a}' is not provided.`);o[i]=this.weightsFiles[n.indexOf(a)]});if(e.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${e.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return o}},K4=r=>z().getBool("IS_BROWSER")&&!Array.isArray(r)&&r.startsWith(Ya.URL_SCHEME)?j4(r.slice(Ya.URL_SCHEME.length)):null;Ce.registerSaveRouter(K4);function j4(r="model"){return new Ya(r)}function uE(r){return new L0(r)}function M0(r,t,e,n){i(r),e=e==null?0:e,n=n==null?1:n,a(e,n);let o=0,s=u=>(u.then(l=>{let c=e+ ++o/r.length*(n-e);return t(c),l}),u);function i(u){E(u!=null&&Array.isArray(u)&&u.length>0,()=>"promises must be a none empty array")}function a(u,l){E(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${u}`),E(l>=0&&l<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${l}`),E(l>=u,()=>`startFraction must be no more than endFraction, but got startFraction ${u} and endFraction ${l}`)}return Promise.all(r.map(s))}async function z0(r,t){t==null&&(t={});let e=t.fetchFunc==null?z().platform.fetch:t.fetchFunc,n=r.map(p=>e(p,t.requestInit,{isBinary:!0})),o=0,s=.5,a=(t.onProgress==null?await Promise.all(n):await M0(n,t.onProgress,o,s)).map(p=>p.arrayBuffer()),u=.5,l=1;return t.onProgress==null?await Promise.all(a):await M0(a,t.onProgress,u,l)}async function cE(r,t="",e,n){return B0(i=>z0(i,{requestInit:n}))(r,t,e)}function B0(r){return async(t,e="",n)=>{let o=t.map(()=>!1),s={},i=n!=null?n.map(()=>!1):[],a=[];if(t.forEach((f,d)=>{let h=0;f.weights.forEach(g=>{let x="quantization"in g?g.quantization.dtype:g.dtype,b=oh[x]*Jt(g.shape),w=()=>{o[d]=!0,s[d]==null&&(s[d]=[]),s[d].push({manifestEntry:g,groupOffset:h,sizeBytes:b})};n!=null?n.forEach((C,N)=>{C===g.name&&(w(),i[N]=!0)}):w(),a.push(g.name),h+=b})}),!i.every(f=>f)){let f=n.filter((d,h)=>!i[h]);throw new Error(`Could not find weights in manifest with names: ${f.join(", ")}. 
-Manifest JSON has weights with names: ${a.join(", ")}.`)}let u=o.reduce((f,d,h)=>(d&&f.push(h),f),[]),l=[];u.forEach(f=>{t[f].paths.forEach(d=>{let h=e+(e.endsWith("/")?"":"/")+d;l.push(h)})});let c=await r(l),p={},m=0;return u.forEach(f=>{let d=t[f].paths.length,h=0;for(let C=0;C<d;C++)h+=c[m+C].byteLength;let g=new ArrayBuffer(h),x=new Uint8Array(g),b=0;for(let C=0;C<d;C++){let N=new Uint8Array(c[m+C]);x.set(N,b),b+=N.byteLength}s[f].forEach(C=>{let N=g.slice(C.groupOffset,C.groupOffset+C.sizeBytes),_=qg(N,[C.manifestEntry]);for(let A in _)p[A]=_[A]}),m+=d}),p}}var X4="application/octet-stream",Y4="application/json",sh=class{constructor(t,e){if(this.DEFAULT_METHOD="POST",e==null&&(e={}),this.weightPathPrefix=e.weightPathPrefix,this.onProgress=e.onProgress,this.weightUrlConverter=e.weightUrlConverter,e.fetchFunc!=null?(E(typeof e.fetchFunc=="function",()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=e.fetchFunc):this.fetch=z().platform.fetch,E(t!=null&&t.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(t)&&E(t.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${t.length}).`),this.path=t,e.requestInit!=null&&e.requestInit.body!=null)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=e.requestInit||{}}async save(t){if(t.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let e=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);e.body=new FormData;let n=[{paths:["./model.weights.bin"],weights:t.weightSpecs}],o=Kg(t,n);e.body.append("model.json",new Blob([JSON.stringify(o)],{type:Y4}),"model.json"),t.weightData!=null&&e.body.append("model.weights.bin",new Blob([t.weightData],{type:X4}),"model.weights.bin");let s=await this.fetch(this.path,e);if(s.ok)return{modelArtifactsInfo:Ni(t),responses:[s]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${s.status}.`)}async load(){let t=await this.fetch(this.path,this.requestInit);if(!t.ok)throw new Error(`Request to ${this.path} failed with status code ${t.status}. Please verify this URL points to the model JSON of the model to load.`);let e;try{e=await t.json()}catch(s){let i=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?i+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":i+=" Please make sure the server is serving valid JSON for this request.",new Error(i)}let n=e.modelTopology,o=e.weightsManifest;if(n==null&&o==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return qp(e,s=>this.loadWeights(s))}async loadWeights(t){let e=Array.isArray(this.path)?this.path[1]:this.path,[n,o]=Z4(e),s=this.weightPathPrefix||n,i=jg(t),a=[],u=[];for(let c of t)for(let p of c.paths)this.weightUrlConverter!=null?u.push(this.weightUrlConverter(p)):a.push(s+p+o);this.weightUrlConverter&&a.push(...await Promise.all(u));let l=await z0(a,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[i,Hp(l)]}};sh.URL_SCHEME_REGEX=/^https?:\/\//;function Z4(r){let t=r.lastIndexOf("/"),e=r.lastIndexOf("?"),n=r.substring(0,t),o=e>t?r.substring(e):"";return[n+"/",o]}function Qg(r){return r.match(sh.URL_SCHEME_REGEX)!=null}var pE=(r,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let e=!0;if(Array.isArray(r)?e=r.every(n=>Qg(n)):e=Qg(r),e)return tx(r,t)}return null};Ce.registerSaveRouter(pE);Ce.registerLoadRouter(pE);function tx(r,t){return new sh(r,t)}function mE(r,t){return tx(r,t)}var ih=class{constructor(t){this.modelArtifacts=t}load(){return this.modelArtifacts}},ex=class{constructor(t){this.saveHandler=t}save(t){return this.saveHandler(t)}},V0=class{constructor(t){t.load&&(this.load=()=>Promise.resolve(t.load())),t.save&&(this.save=e=>Promise.resolve(t.save(e)))}};function fE(r,t,e,n){let o=arguments;return new V0(G0(...o))}function G0(r,t,e,n){return arguments.length===1?r.modelTopology!=null||r.weightSpecs!=null?new ih(r):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new ih({modelTopology:r})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new ih({modelTopology:r,weightSpecs:t,weightData:e,trainingConfig:n}))}function dE(r){return new ex(r)}function hE(r){return new ex(r)}var yE={};Wt(yE,{confusionMatrix:()=>xE});function J4(r,t,e=!1,n=!1){let o=I(r,"a","matMul"),s=I(t,"b","matMul");[o,s]=Ut(o,s);let i={a:o,b:s},a={transposeA:e,transposeB:n};return k.runKernel(Ho,i,a)}var Lt=T({matMul_:J4});function Q4(r,t,e=1,n=0,o="int32"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let i={indices:I(r,"indices","oneHot","int32")},a={dtype:o,depth:t,onValue:e,offValue:n};return k.runKernel(gs,i,a)}var Ei=T({oneHot_:Q4});function dpt(){z().set("PROD",!0)}function hpt(){z().set("DEBUG",!0)}function gpt(){z().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function W0(r){z().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(r+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}L1(W0);function xpt(){k.disposeVariables()}function Pn(){return k}function ah(){return k.memory()}function ypt(r){return k.profile(r)}function B(r,t){return k.tidy(r,t)}function vt(r){nh(r).forEach(e=>e.dispose())}function De(r){return k.keep(r)}function bpt(r){return k.time(r)}function tH(r){return k.setBackend(r)}function wpt(){return k.ready()}function Cpt(){return k.backendName}function Ipt(r){k.removeBackend(r)}function Spt(r){return k.findBackend(r)}function vpt(r){return k.findBackendFactory(r)}function Xp(r,t,e=1){return k.registerBackend(r,t,e)}function gE(){return k.backend}function Npt(r,t){z().setPlatform(r,t)}function eH(r){let e={input:I(r,"input","imag")};return k.runKernel(Sp,e)}var Xl=T({imag_:eH});function rH(r){let e={x:I(r,"x","neg")};return k.runKernel(pi,e)}var Ht=T({neg_:rH});function nH(r){let e={input:I(r,"input","real")};return k.runKernel(Rp,e)}var Za=T({real_:nH});function oH(r,t,e){let n=I(r,"x","transpose");if(t==null&&(t=n.shape.map((i,a)=>a).reverse()),E(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of perm ${t}.`),t.forEach(i=>{E(i>=0&&i<n.rank,()=>`All entries in 'perm' must be between 0 and ${n.rank-1} but got ${t}`)}),n.rank<=1)return n.clone();let o={x:n},s={perm:t};return n.dtype==="complex64"?B(()=>{let i=Za(n),a=Xl(n);return i=k.runKernel(Qn,{x:i},s),a=k.runKernel(Qn,{x:a},s),e&&(a=Ht(a)),wn(i,a)}):k.runKernel(Qn,o,s)}var Ot=T({transpose_:oH});function sH(r,t,e){let n=I(r,"labels","confusionMatrix"),o=I(t,"predictions","confusionMatrix");E(e==null||e>0&&Number.isInteger(e),()=>`If provided, numClasses must be a positive integer, but got ${e}`),E(n.rank===1,()=>`Expected the rank of labels to be 1, but got ${n.rank}`),E(o.rank===1,()=>`Expected the rank of predictions to be 1, but got ${o.rank}`),E(n.shape[0]===o.shape[0],()=>`Mismatch in the number of examples: ${n.shape[0]} vs. ${o.shape[0]}. Labels and predictions should have the same number of elements.`),E(e>0&&Number.isInteger(e),()=>`numClasses is required to be a positive integer, but got ${e}`);let s=Ei(J(n,"int32"),e),i=Ei(J(o,"int32"),e),a=Ot(s),u=Lt(a,i);return J(u,"int32")}var xE=T({confusionMatrix_:sH});var Vr={};Wt(Vr,{assertAndGetBroadcastShape:()=>Pt,getBroadcastDims:()=>bE,getReductionAxes:()=>ge});function bE(r,t){let e=r.length,n=[];for(let o=0;o<e;o++){let s=e-1-o,i=r[s]||1;(t[t.length-1-o]||1)>1&&i===1&&n.unshift(s)}return n}function ge(r,t){let e=[];for(let n=0;n<t.length;n++){let o=r[r.length-n-1],s=t.length-n-1,i=t[s];(o==null||o===1&&i>1)&&e.unshift(s)}return e}function Pt(r,t){let e=[],n=Math.max(r.length,t.length);for(let o=0;o<n;o++){let s=r[r.length-o-1];s==null&&(s=1);let i=t[t.length-o-1];if(i==null&&(i=1),s===1)e.unshift(i);else if(i===1)e.unshift(s);else if(s!==i){let a=`Operands could not be broadcast together with shapes ${r} and ${t}.`;throw Error(a)}else e.unshift(s)}return e}var nx={};Wt(nx,{fromPixels:()=>mH,fromPixelsAsync:()=>cH,toPixels:()=>pH});function rx(r,t,e){if(Yn(r),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");let n=Br(r,e);if(n.length!==3&&n.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return on(r,t,n,e)}var Hu;function wE(r,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(r==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let e=!1,n=!1,o=!1,s=!1,i=!1,a=!1;if(r.data instanceof Uint8Array)e=!0;else if(typeof ImageData!="undefined"&&r instanceof ImageData)n=!0;else if(typeof HTMLVideoElement!="undefined"&&r instanceof HTMLVideoElement)o=!0;else if(typeof HTMLImageElement!="undefined"&&r instanceof HTMLImageElement)s=!0;else if(r.getContext!=null)i=!0;else if(typeof ImageBitmap!="undefined"&&r instanceof ImageBitmap)a=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${r.constructor.name}`);if(Jd(Yd,k.backendName)!=null){let d={pixels:r},h={numChannels:t};return k.runKernel(Yd,d,h)}let[l,c]=o?[r.videoWidth,r.videoHeight]:[r.width,r.height],p;if(i)p=r.getContext("2d").getImageData(0,0,l,c).data;else if(n||e)p=r.data;else if(s||o||a){if(Hu==null)if(typeof document=="undefined")if(typeof OffscreenCanvas!="undefined"&&typeof OffscreenCanvasRenderingContext2D!="undefined")Hu=new OffscreenCanvas(1,1).getContext("2d");else throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");else Hu=document.createElement("canvas").getContext("2d",{willReadFrequently:!0});Hu.canvas.width=l,Hu.canvas.height=c,Hu.drawImage(r,0,0,l,c),p=Hu.getImageData(0,0,l,c).data}let m;if(t===4)m=new Int32Array(p);else{let d=l*c;m=new Int32Array(d*t);for(let h=0;h<d;h++)for(let g=0;g<t;++g)m[h*t+g]=p[h*4+g]}return rx(m,[c,l,t],"int32")}function iH(r){return r!=null&&r.data instanceof Uint8Array}function aH(){return typeof window!="undefined"&&typeof ImageBitmap!="undefined"&&window.hasOwnProperty("createImageBitmap")}function lH(r){return r!=null&&r.width!==0&&r.height!==0}function uH(r){return aH()&&!(r instanceof ImageBitmap)&&lH(r)&&!iH(r)}async function cH(r,t=3){let e=null;if(z().getBool("WRAP_TO_IMAGEBITMAP")&&uH(r)){let n;try{n=await createImageBitmap(r,{premultiplyAlpha:"none"})}catch(o){n=null}n!=null&&n.width===r.width&&n.height===r.height?e=n:e=r}else e=r;return wE(e,t)}async function pH(r,t){let e=I(r,"img","toPixels");if(!(r instanceof Ft)){let l=e;e=J(l,"int32"),l.dispose()}if(e.rank!==2&&e.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${e.rank}.`);let[n,o]=e.shape.slice(0,2),s=e.rank===2?1:e.shape[2];if(s>4||s===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${s}`);if(e.dtype!=="float32"&&e.dtype!=="int32")throw new Error(`Unsupported type for toPixels: ${e.dtype}. Please use float32 or int32 tensors.`);let i=await e.data(),a=e.dtype==="float32"?255:1,u=new Uint8ClampedArray(o*n*4);for(let l=0;l<n*o;++l){let c=[0,0,0,255];for(let m=0;m<s;m++){let f=i[l*s+m];if(e.dtype==="float32"){if(f<0||f>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${f}.`)}else if(e.dtype==="int32"&&(f<0||f>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${f}.`);s===1?(c[0]=f*a,c[1]=f*a,c[2]=f*a):c[m]=f*a}let p=l*4;u[p+0]=Math.round(c[0]),u[p+1]=Math.round(c[1]),u[p+2]=Math.round(c[2]),u[p+3]=Math.round(c[3])}if(t!=null){t.width=o,t.height=n;let l=t.getContext("2d"),c=new ImageData(u,o,n);l.putImageData(c,0,0)}return e!==r&&e.dispose(),u}var mH=T({fromPixels_:wE});var ox={};Wt(ox,{prepareAndValidate:()=>CE});function CE(r,t){let e=r.shape.length,n=t.shape.length;if(e<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${e}.`);if(n<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${n}.`);if(t.dtype!=="int32")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[n-1]>e)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[n-1]} vs. ${e}`);if(Jt(r.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${r.shape}.`);let o=t.shape,s=o[o.length-1],i=1;for(let p=0;p<o.length-1;++p)i*=o[p];let a=r.shape,u=o.slice();u.pop();let l=1;for(let p=s;p<e;++p)l*=a[p],u.push(a[p]);let c=[...si(r.shape).map(p=>p/l),1].slice(0,s);return[u,i,l,c]}var lh={};Wt(lh,{calculateShapes:()=>IE,validateInput:()=>sx,validateUpdateShape:()=>U0});function U0(r,t,e){let n=t.rank>1?t.shape[t.rank-1]:1,o=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${e.shape}, indices.shape: ${t.shape}, shape: ${r}, sliceDim: ${n}, and batchDim: ${o}.`;if(e.rank<o)throw new Error(s+` update.rank < ${o}. `);if(r.length<n+(e.rank-o))throw new Error(s+` Output shape length < ${n+(e.rank-o)}`);if(e.rank!==o+r.length-n)throw new Error(s+` update.rank != ${o+r.length-n}`);for(let i=0;i<o;++i)if(e.shape[i]!==t.shape[i])throw new Error(s+` updates.shape[${i}] (${e.shape[i]}) != indices.shape[${i}] (${t.shape[i]}).`);for(let i=0;i<e.rank-o;++i)if(e.shape[i+o]!==r[i+n])throw new Error(s+` updates.shape[${i+o}] (${e.shape[i+o]}) != shape[${i+o}] (${r[i+o]})`)}function sx(r,t,e){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(r.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${r.rank}.`);if(t.dtype!=="int32")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(e.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${e}`);if(e.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(r.size===0)throw new Error(`Updates specified for empty output. updates shape: ${r.shape}`)}U0(e,t,r)}function IE(r,t,e){let n=t.shape.length,o=n>1?t.shape[n-1]:1,s=e.length,i=1;for(let p=o;p<s;++p)i*=e[p];let a=o<1?1:o,u=Jt(t.shape)/a,l=[...si(e.slice(0,o)),1],c=Jt(e);return{sliceRank:o,numUpdates:u,sliceSize:i,strides:l,outputSize:c}}var Le={};Wt(Le,{assertParamsValid:()=>dH,computeFlatOffset:()=>bH,computeOutShape:()=>gH,getNormalizedAxes:()=>xH,isSliceContinous:()=>yH,maskToAxes:()=>hH,parseSliceParams:()=>q0,sliceInfo:()=>wH,startForAxis:()=>AE,startIndicesWithElidedDims:()=>kE,stopForAxis:()=>$E,stopIndicesWithElidedDims:()=>EE,stridesForAxis:()=>_E,stridesWithElidedDims:()=>vE});var H0=-2,fH=-1;function dH(r,t,e){let n=r.shape.length;E(n===t.length,()=>`Error in slice${n}D: Length of begin ${t} must match the rank of the array (${n}).`),E(n===e.length,()=>`Error in slice${n}D: Length of size ${e} must match the rank of the array (${n}).`);for(let o=0;o<n;++o)E(t[o]+e[o]<=r.shape[o],()=>`Error in slice${n}D: begin[${o}] + size[${o}] (${t[o]+e[o]}) would overflow input.shape[${o}] (${r.shape[o]})`)}function hH(r){let t=[],e=0;for(;r>0;)r&1&&t.push(e),r/=2,e++;return t}function gH(r,t,e){let n=[];for(let o=0;o<r.length;o++)n[o]=Math.ceil((t[o]-r[o])/e[o]);return n}function vE(r,t,e,n){let o=[...r];for(let s=o.length;s<n.length;s++)o.push(1);for(let s=0;s<e;s++)s===0?o[t]=1:(o.splice(t,0,1),o.pop());return o}function NE(r,t,e){return e<=r?e:e-(t-1)}function TE(r,t){let e=[];for(let n=0;n<r;n++)e.push(t+n);return e}function xH(r,t,e,n,o,s,i,a,u){let l=r.length,c=new Array(l),p=new Array(l),m=new Array(l);if(t.length&&e>0){let f=t[0],d=e+1;c=kE(i,f,d,n,r),p=EE(a,f,d,o,r),m=vE(s,f,d,r)}else for(let f=0;f<l;f++)c[f]=AE(i,n,s,r,f,u),p[f]=$E(a,o,s,r,f,u),m[f]=_E(s,f,u);return{begin:c,end:p,strides:m}}function kE(r,t,e,n,o){let s=[...o],i=TE(e,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=0;else{let u=NE(t,e,a),l=n[u];r&1<<u&&(l=0),s[a]=l}return s}function EE(r,t,e,n,o){let s=[...o],i=TE(e,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=Number.MAX_SAFE_INTEGER;else{let u=NE(t,e,a),l=n[u];r&1<<u&&(l=Number.MAX_SAFE_INTEGER),s[a]=l}for(let a=0;a<s.length;a++){let u=o[a];s[a]<0&&(s[a]+=u),s[a]=np(0,s[a],o[a])}return s}function _E(r,t,e){let n=r[t];return(e&1<<t||n==null)&&(n=1),n}function AE(r,t,e,n,o,s){let i=t[o],a=e[o]||1;(r&1<<o||s&1<<o||i==null)&&(a>0?i=Number.MIN_SAFE_INTEGER:i=Number.MAX_SAFE_INTEGER);let u=n[o];return i<0&&(i+=u),i=np(0,i,u-1),i}function $E(r,t,e,n,o,s){let i=t[o],a=e[o]||1;(r&1<<o||s&1<<o||i==null)&&(a>0?i=Number.MAX_SAFE_INTEGER:i=Number.MIN_SAFE_INTEGER);let u=n[o];return i<0&&(i+=u),a>0?i=np(0,i,u):i=np(-1,i,u-1),i}function yH(r,t,e){let n=e.length;for(let o=0;o<e.length;o++)if(e[o]>1){n=o;break}for(let o=n+1;o<e.length;o++)if(t[o]>0||e[o]!==r[o])return!1;return!0}function bH(r,t){let e=r.length>0?r[r.length-1]:1;for(let n=0;n<r.length-1;n++)e+=r[n]*t[n];return e}function q0(r,t,e){let n,o=r.shape.length;typeof t=="number"?n=[t,...new Array(o-1).fill(0)]:t.length<o?n=t.concat(new Array(o-t.length).fill(0)):n=t.slice(),n.forEach(i=>{E(i!==-1,()=>"slice() does not support negative begin indexing.")});let s;return e==null?s=new Array(o).fill(-1):typeof e=="number"?s=[e,...new Array(o-1).fill(-1)]:e.length<o?s=e.concat(new Array(o-e.length).fill(-1)):s=e,s=s.map((i,a)=>i>=0?i:(E(i===-1,()=>`Negative size values should be exactly -1 but got ${i} for the slice() size at index ${a}.`),r.shape[a]-n[a])),[n,s]}function wH(r,t,e,n,o,s,i,a,u){let l;if(n==null?(l=new Array(t.length),l.fill(1)):l=n,i!=null&&(i&i-1)!==0)throw new Error("Multiple ellipses in slice is not allowed.");let c=!1,p={dims:l.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:e.slice(),strides:l.slice(),beginMask:o,endMask:s,ellipsisMask:i,newAxisMask:a,shrinkAxisMask:u};for(let w=0;w<p.dims;w++)c&&(1<<w&a)!==0&&p.numAddAxisAfterEllipsis++,1<<w&i&&(c=!0);c||(p.ellipsisMask|=1<<p.dims,p.dims++);let m={dims:r.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};CH(p,m);let f=!0,d=!0,h=!0,g=[],x=[];for(let w=0;w<r.length;++w){if(m.strides[w]===0)throw Error(`strides[${w}] must be non-zero`);let C=!!(m.shrinkAxisMask&1<<w),N=r[w];if(N===-1){g.push(C?1:-1);continue}let _=[m.beginMask&1<<w,m.endMask&1<<w],A=[m.strides[w]>0?0:-1,m.strides[w]>0?N:N-1];if(C&&m.strides[w]<=0)throw Error("only stride 1 allowed on non-range indexing.");h=h&&m.strides[w]===1;let $=!!(m.beginMask&1<<w&&m.endMask&1<<w);if(m.beginValid&&m.endValid){if(C){let G=m.begin[w]<0?N+m.begin[w]:m.begin[w];if(m.begin[w]=G,m.end[w]=m.begin[w]+1,G<0||G>=N)throw Error(`slice index ${m.begin[w]} of dimension ${w} out of bounds.`)}else m.begin[w]=SE(m.begin[w],0,m.strides[w],N,_,A),m.end[w]=SE(m.end[w],1,m.strides[w],N,_,A);let V=m.strides[w]===1&&m.begin[w]===0&&m.end[w]===N;f=f&&V,d=d&&(w===0&&m.strides[w]===1||V)}else f=f&&m.strides[w]===1&&$,d=d&&(w===0&&m.strides[w]===1||$);let F,P=!1;if(m.beginValid&&m.endValid?(F=m.end[w]-m.begin[w],P=!0):C?(F=1,P=!0):$&&N>=0&&(m.strides[w]<0?F=-N:F=N,P=!0),P){let V;F===0||F<0!=m.strides[w]<0?V=0:V=Math.trunc(F/m.strides[w])+(F%m.strides[w]!==0?1:0),g.push(V)}else g.push(-1)}for(let w=0;w<m.finalShapeGatherIndices.length;++w){let C=m.finalShapeGatherIndices[w];C>=0?x.push(g[C]):C===H0&&x.push(1)}return{finalShapeSparse:x.filter((w,C)=>m.finalShapeGatherIndices[C]!==H0),finalShape:x,isIdentity:f,sliceDim0:d,isSimpleSlice:h,begin:m.begin,end:m.end,strides:m.strides}}function CH(r,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let e=0;t.beginValid=r.begin!=null,t.endValid=r.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let n=0;n<r.dims;n++)if(1<<n&r.ellipsisMask){let o=Math.min(t.dims-(r.dims-n)+1+r.numAddAxisAfterEllipsis,t.dims);for(;e<o;e++)t.begin[e]=0,t.end[e]=0,t.strides[e]=1,t.beginMask|=1<<e,t.endMask|=1<<e,t.finalShapeGatherIndices.push(e),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[e]=n}else if(1<<n&r.newAxisMask)t.finalShapeGatherIndices.push(H0),t.finalShapeGatherIndicesSparse.push(-1);else{if(e===t.begin.length)throw Error(`Index out of range using input dim ${e}; input has only ${t.dims} dims, ${t.begin.length}.`);r.begin!=null&&(t.begin[e]=r.begin[n]),r.end!=null&&(t.end[e]=r.end[n]),t.strides[e]=r.strides[n],r.beginMask&1<<n&&(t.beginMask|=1<<e),r.endMask&1<<n&&(t.endMask|=1<<e),r.shrinkAxisMask&1<<n?(t.finalShapeGatherIndices.push(fH),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<e):(t.finalShapeGatherIndices.push(e),t.finalShapeGatherIndicesSparse.push(n)),t.inputShapeGatherIndicesSparse[e]=n,e++}}function SE(r,t,e,n,o,s){if(o[t])return e>0?s[t]:s[t+1&1];{let i=r<0?n+r:r;return i<s[0]?s[0]:i>s[1]?s[1]:i}}var Q={};Wt(Q,{Serializable:()=>uh,SerializationMap:()=>_i,registerClass:()=>Cn});var uh=class{getClassName(){return this.constructor.className}static fromConfig(t,e){return new t(e)}},_i=class{constructor(){this.classNameMap={}}static getMap(){return _i.instance==null&&(_i.instance=new _i),_i.instance}static register(t){_i.getMap().classNameMap[t.className]=[t,t.fromConfig]}};function Cn(r){E(r.className!=null,()=>"Class being registered does not have the static className property defined."),E(typeof r.className=="string",()=>"className is required to be a string, but got type "+typeof r.className),E(r.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),_i.register(r)}var OE={};Wt(OE,{TEST_EPSILON_FLOAT16:()=>DE,createVideoElement:()=>EH,encodeStrings:()=>FE,expectArrayBuffersEqual:()=>kH,expectArraysClose:()=>SH,expectArraysEqual:()=>NH,expectNumbersClose:()=>RE,expectPromiseToFail:()=>vH,expectValuesInRange:()=>TH,play:()=>_H,testEpsilon:()=>ix});var IH=.001,DE=.1;function SH(r,t,e){return e==null&&(e=ix()),K0(r,t,(n,o)=>j0(n,o,e))}function ix(){return k.backend.floatPrecision()===32?IH:DE}function K0(r,t,e){let n=!0;if((xr(r)||xr(t))&&(n=!1),xr(r)&&xr(t)&&(n=!0),n){let i=r.constructor.name,a=t.constructor.name;if(i!==a)throw new Error(`Arrays are of different type. Actual: ${i}. Expected: ${a}`)}if(Array.isArray(r)&&Array.isArray(t)){let i=Br(r),a=Br(t);if(!Dn(i,a))throw new Error(`Arrays have different shapes. Actual: [${i}]. Expected: [${a}]`)}let o=xr(r)?r:Bo(r),s=xr(t)?t:Bo(t);if(o.length!==s.length)throw new Error(`Arrays have different lengths actual: ${o.length} vs expected: ${s.length}.
-Actual:   ${o}.
-Expected: ${s}.`);for(let i=0;i<s.length;++i){let a=o[i],u=s[i];if(!e(a,u))throw new Error(`Arrays differ: actual[${i}] = ${a}, expected[${i}] = ${u}.
-Actual:   ${o}.
-Expected: ${s}.`)}typeof expect!="undefined"&&expect().nothing()}function vH(r,t){r().then(()=>t.fail(),()=>t()),typeof expect!="undefined"&&expect().nothing()}function NH(r,t){let e=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return Vo(r)||Vo(r[0])||Vo(t)||Vo(t[0])?K0(r,e,(n,o)=>n==o):K0(r,t,(n,o)=>j0(n,o,0))}function RE(r,t,e){if(e==null&&(e=ix()),!j0(r,t,e))throw new Error(`Numbers differ: actual === ${r}, expected === ${t}`);typeof expect!="undefined"&&expect().nothing()}function j0(r,t,e){return!isFinite(r)&&!isFinite(t)?!0:!(isNaN(r)||isNaN(t)||Math.abs(r-t)>e)}function TH(r,t,e){for(let n=0;n<r.length;n++)if(r[n]<t||r[n]>e)throw new Error(`Value out of range:${r[n]} low: ${t}, high: ${e}`)}function kH(r,t){let e=new Float32Array(r),n=new Float32Array(t);if(e.length!==n.length)throw new Error(`Expected ArrayBuffer to be of length ${n.length}, but it was ${e.length}`);for(let o=0;o<n.length;o++)if(e[o]!==n[o])throw new Error(`Expected ArrayBuffer value at ${o} to be ${n[o]} but got ${e[o]} instead`)}function FE(r){for(let t=0;t<r.length;t++){let e=r[t];Array.isArray(e)?FE(e):r[t]=Hl(e)}return r}function EH(r){let t=document.createElement("video");return"playsInline"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position="fixed",t.style.left="0px",t.style.top="0px",t.preload="auto",t.appendChild(r),new Promise(e=>{t.addEventListener("loadeddata",n=>e(t)),t.load()})}async function _H(r){await r.play(),"requestVideoFrameCallback"in r&&await new Promise(t=>{r.requestVideoFrameCallback(t)})}var PE="4.0.0";function AH(r,t){let e=I(r,"a","add"),n=I(t,"b","add");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(Zn,o)}var X=T({add_:AH});function $H(r,t){let e=I(r,"a","floorDiv"),n=I(t,"b","floorDiv");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ns,o)}var Yp=T({floorDiv_:$H});function DH(r,t){let e=I(r,"a","div"),n=I(t,"b","div");if([e,n]=Ut(e,n),e.dtype==="int32"&&n.dtype==="int32")return Yp(e,n);let o={a:e,b:n},s={};return k.runKernel(Qo,o,s)}var pt=T({div_:DH});function RH(r,t){let e=I(r,"a","mul"),n=I(t,"b","mul");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(hs,o)}var D=T({mul_:RH});function FH(r){let t=I(r,"x","abs");if(t.dtype==="complex64"){let e={x:t};return k.runKernel(_l,e)}else{let e={x:t};return k.runKernel(ii,e)}}var Ee=T({abs_:FH});function OH(r){let e={x:I(r,"x","acos")};return k.runKernel(oa,e)}var ax=T({acos_:OH});function PH(r){let e={x:I(r,"x","acosh")};return k.runKernel(sa,e)}var lx=T({acosh_:PH});function LH(r){E(Array.isArray(r),()=>"The argument passed to tf.addN() must be a list of tensors"),E(r.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${r.length}`);let t=r.map((o,s)=>I(o,`tensors${s}`,"addN")),e=t[0];t.forEach(o=>{if(o.dtype!==e.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(o=>{if(!Dn(o.shape,e.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let n=t;return k.runKernel(Go,n)}var LE=T({addN_:LH});function MH(r,t=null,e=!1){let o={x:I(r,"x","all","bool")},s={axis:t,keepDims:e};return k.runKernel(ia,o,s)}var Zp=T({all_:MH});function zH(r,t=null,e=!1){let o={x:I(r,"x","any","bool")},s={axis:t,keepDims:e};return k.runKernel(aa,o,s)}var qu=T({any_:zH});function BH(r,t=0){let n={x:I(r,"x","argMax")},o={axis:t};return k.runKernel(Wo,n,o)}var Ai=T({argMax_:BH});function VH(r,t=0){let n={x:I(r,"x","argMin")},o={axis:t};return k.runKernel(kl,n,o)}var ux=T({argMin_:VH});function GH(r){let e={x:I(r,"x","asin")};return k.runKernel(la,e)}var cx=T({asin_:GH});function WH(r){let e={x:I(r,"x","asinh")};return k.runKernel(ua,e)}var px=T({asinh_:WH});function UH(r){let e={x:I(r,"x","atan")};return k.runKernel(ca,e)}var mx=T({atan_:UH});function HH(r,t){let e=I(r,"a","atan2"),n=I(t,"b","atan2");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ma,o)}var fx=T({atan2_:HH});function qH(r){let e={x:I(r,"x","atanh")};return k.runKernel(pa,e)}var dx=T({atanh_:qH});function KH(r,t,e,n,o="NHWC",s){let i=r[3],a=[...t,i],u=zE(o);return ju(r,a,e,s,n,null,null,u)}function Y0(r,t,e,n,o,s,i="channelsLast"){let[a,u]=hx(t),l;if(i==="channelsLast")l=[a,u,r[3],r[3]];else if(i==="channelsFirst")l=[a,u,r[1],r[1]];else throw new Error(`Unknown dataFormat ${i}`);return ju(r,l,e,n,o,s,!1,i)}function jH(r,t,e,n,o,s,i="NDHWC"){let[a,u,l]=X0(t),c,p;if(i==="NDHWC")p="channelsLast",c=[a,u,l,r[4],r[4]];else if(i==="NCDHW")p="channelsFirst",c=[a,u,l,r[1],r[1]];else throw new Error(`Unknown dataFormat ${i}`);return ME(r,c,e,n,o,!1,p,s)}function ju(r,t,e,n,o,s,i=!1,a="channelsLast"){let[u,l,c,p]=[-1,-1,-1,-1];if(a==="channelsLast")[u,l,c,p]=r;else if(a==="channelsFirst")[u,p,l,c]=r;else throw new Error(`Unknown dataFormat ${a}`);let[m,f,,d]=t,[h,g]=hx(e),[x,b]=hx(n),w=Jp(m,x),C=Jp(f,b),{padInfo:N,outHeight:_,outWidth:A}=ZH(o,l,c,h,g,w,C,s,a),$=i?d*p:d,F;return a==="channelsFirst"?F=[u,$,_,A]:a==="channelsLast"&&(F=[u,_,A,$]),{batchSize:u,dataFormat:a,inHeight:l,inWidth:c,inChannels:p,outHeight:_,outWidth:A,outChannels:$,padInfo:N,strideHeight:h,strideWidth:g,filterHeight:m,filterWidth:f,effectiveFilterHeight:w,effectiveFilterWidth:C,dilationHeight:x,dilationWidth:b,inShape:r,outShape:F,filterShape:t}}function ME(r,t,e,n,o,s=!1,i="channelsLast",a){let[u,l,c,p,m]=[-1,-1,-1,-1,-1];if(i==="channelsLast")[u,l,c,p,m]=r;else if(i==="channelsFirst")[u,m,l,c,p]=r;else throw new Error(`Unknown dataFormat ${i}`);let[f,d,h,,g]=t,[x,b,w]=X0(e),[C,N,_]=X0(n),A=Jp(f,C),$=Jp(d,N),F=Jp(h,_),{padInfo:P,outDepth:V,outHeight:G,outWidth:W}=JH(o,l,c,p,x,b,w,A,$,F,a),q=s?g*m:g,H;return i==="channelsFirst"?H=[u,q,V,G,W]:i==="channelsLast"&&(H=[u,V,G,W,q]),{batchSize:u,dataFormat:i,inDepth:l,inHeight:c,inWidth:p,inChannels:m,outDepth:V,outHeight:G,outWidth:W,outChannels:q,padInfo:P,strideDepth:x,strideHeight:b,strideWidth:w,filterDepth:f,filterHeight:d,filterWidth:h,effectiveFilterDepth:A,effectiveFilterHeight:$,effectiveFilterWidth:F,dilationDepth:C,dilationHeight:N,dilationWidth:_,inShape:r,outShape:H,filterShape:t}}function XH(r,t,e,n,o){n==null&&(n=Z0(r,t,e));let s=r[0],i=r[1],a=Ku((s-t+2*n)/e+1,o),u=Ku((i-t+2*n)/e+1,o);return[a,u]}function YH(r,t,e,n,o,s){o==null&&(o=Z0(r,t,n));let i=r[0],a=r[1],u=r[2],l=Ku((i-t+2*o)/n+1,s),c=Ku((a-t+2*o)/n+1,s),p=Ku((u-t+2*o)/n+1,s);return[l,c,p,e]}function Z0(r,t,e,n=1){let o=Jp(t,n);return Math.floor((r[0]*(e-1)-e+o)/2)}function hx(r){return typeof r=="number"?[r,r,r]:r.length===2?[r[0],r[1],1]:r}function X0(r){return typeof r=="number"?[r,r,r]:r}function Jp(r,t){return t<=1?r:r+(r-1)*(t-1)}function ZH(r,t,e,n,o,s,i,a,u){let l,c,p;if(typeof r=="number"){l={top:r,bottom:r,left:r,right:r,type:r===0?"VALID":"NUMBER"};let f=XH([t,e],s,n,r,a);c=f[0],p=f[1]}else if(r==="same"){c=Math.ceil(t/n),p=Math.ceil(e/o);let m=Math.max(0,(c-1)*n+s-t),f=Math.max(0,(p-1)*o+i-e),d=Math.floor(m/2),h=m-d,g=Math.floor(f/2),x=f-g;l={top:d,bottom:h,left:g,right:x,type:"SAME"}}else if(r==="valid")l={top:0,bottom:0,left:0,right:0,type:"VALID"},c=Math.ceil((t-s+1)/n),p=Math.ceil((e-i+1)/o);else if(typeof r=="object"){let m=u==="channelsLast"?r[1][0]:r[2][0],f=u==="channelsLast"?r[1][1]:r[2][1],d=u==="channelsLast"?r[2][0]:r[3][0],h=u==="channelsLast"?r[2][1]:r[3][1];l={top:m,bottom:f,left:d,right:h,type:m===0&&f===0&&d===0&&h===0?"VALID":"EXPLICIT"},c=Ku((t-s+m+f)/n+1,a),p=Ku((e-i+d+h)/o+1,a)}else throw Error(`Unknown padding parameter: ${r}`);return{padInfo:l,outHeight:c,outWidth:p}}function JH(r,t,e,n,o,s,i,a,u,l,c){let p,m,f,d;if(typeof r=="number"){p={top:r,bottom:r,left:r,right:r,front:r,back:r,type:r===0?"VALID":"NUMBER"};let g=YH([t,e,n,1],a,1,o,r,c);m=g[0],f=g[1],d=g[2]}else if(r==="same"){m=Math.ceil(t/o),f=Math.ceil(e/s),d=Math.ceil(n/i);let h=(m-1)*o+a-t,g=(f-1)*s+u-e,x=(d-1)*i+l-n,b=Math.floor(h/2),w=h-b,C=Math.floor(g/2),N=g-C,_=Math.floor(x/2),A=x-_;p={top:C,bottom:N,left:_,right:A,front:b,back:w,type:"SAME"}}else if(r==="valid")p={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},m=Math.ceil((t-a+1)/o),f=Math.ceil((e-u+1)/s),d=Math.ceil((n-l+1)/i);else throw Error(`Unknown padding parameter: ${r}`);return{padInfo:p,outDepth:m,outHeight:f,outWidth:d}}function Ku(r,t){if(!t)return Math.trunc(r);switch(t){case"round":return Math.round(r);case"ceil":return Math.ceil(r);case"floor":return Math.floor(r);default:throw new Error(`Unknown roundingMode ${t}`)}}function to(r){let[t,e,n]=hx(r);return t===1&&e===1&&n===1}function Ar(r,t){return to(r)||to(t)}function zE(r){if(r==="NHWC")return"channelsLast";if(r==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${r}`)}function Ie(r,t,e){if(e!=null){if(typeof t=="string")throw Error(`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${t}.`);if(typeof t=="number")E(na(t),()=>`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${t}.`);else if(typeof t=="object")t.forEach(n=>{n.forEach(o=>{E(na(o),()=>`Error in ${r}: pad must be an integer when using dimRoundingMode ${e} but got pad ${o}.`)})});else throw Error(`Error in ${r}: Unknown padding parameter: ${t}`)}}function QH(r,t){let n={x:I(r,"x","reshape","string_or_numeric")},o={shape:t};return k.runKernel(di,n,o)}var R=T({reshape_:QH});function tq(r,t,e,n,o){let s=I(r,"x","avgPool","float32"),i=1;E(Ar(e,i),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${e} and dilations '${i}'`);let a=s,u=!1;s.rank===3&&(u=!0,a=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(a.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${a.rank}.`),Ie("avgPool",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o},p=k.runKernel(Uo,l,c);return p=J(p,s.dtype),u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var Yl=T({avgPool_:tq});function eq(r,t,e,n,o,s="NDHWC"){let i=I(r,"x","avgPool3d","float32"),a=i,u=!1;i.rank===4&&(u=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(a.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${a.rank}.`),E(s==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Ie("avgPool3d",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o,dataFormat:s},p=k.runKernel(El,l,c);return p=J(p,a.dtype),u?R(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var gx=T({avgPool3d_:eq});function rq(r,t=0){E(r.length>=1,()=>"Pass at least one tensor to concat");let e=ja(r,"tensors","concat","string_or_numeric");if(e[0].dtype==="complex64"&&e.forEach(s=>{if(s.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
-          with dtype ${s.dtype}. `)}),e.length===1)return sn(e[0]);let n=e,o={axis:t};return k.runKernel(li,n,o)}var ne=T({concat_:rq});function nq(r){let e={x:I(r,"x","sigmoid","float32")};return k.runKernel(_s,e)}var Yr=T({sigmoid_:nq});function oq(r,t,e){let n=I(r,"x","slice","string_or_numeric");if(n.rank===0)throw new Error("Slicing scalar is not possible");let o={x:n},s={begin:t,size:e};return k.runKernel(gi,o,s)}var Rt=T({slice_:oq});function sq(r){let e={x:I(r,"x","tanh","float32")};return k.runKernel(Ps,e)}var $i=T({tanh_:sq});function iq(r,t,e,n,o,s){let i=I(r,"forgetBias","basicLSTMCell"),a=I(t,"lstmKernel","basicLSTMCell"),u=I(e,"lstmBias","basicLSTMCell"),l=I(n,"data","basicLSTMCell"),c=I(o,"c","basicLSTMCell"),p=I(s,"h","basicLSTMCell"),m=ne([l,p],1),f=Lt(m,a),d=X(f,u),h=d.shape[0],g=d.shape[1]/4,x=[h,g],b=Rt(d,[0,0],x),w=Rt(d,[0,g],x),C=Rt(d,[0,g*2],x),N=Rt(d,[0,g*3],x),_=X(D(Yr(b),$i(w)),D(c,Yr(X(i,C)))),A=D($i(_),Yr(N));return[_,A]}var BE=T({basicLSTMCell_:iq});function aq(r,t,e){let n=I(r,"x","batchToSpaceND"),o=t.reduce((a,u)=>a*u);E(n.rank>=1+t.length,()=>`input rank is ${n.rank} but should be > than blockShape.length ${t.length}`),E(e.length===t.length,()=>`crops.length is ${e.length} but should be equal to blockShape.length  ${t.length}`),E(n.shape[0]%o===0,()=>`input tensor batch is ${n.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${o}`);let s={x:n},i={blockShape:t,crops:e};return k.runKernel(ai,s,i)}var Zl=T({batchToSpaceND_:aq});function VE(r){let t;return r.rank===0||r.rank===1?t=R(r,[1,1,1,r.size]):r.rank===2?t=R(r,[1,1,r.shape[0],r.shape[1]]):r.rank===3?t=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]):t=r,t}function lq(r,t,e,n,o,s){s==null&&(s=.001);let i=I(r,"x","batchNorm"),a=I(t,"mean","batchNorm"),u=I(e,"variance","batchNorm"),l;o!=null&&(l=I(o,"scale","batchNorm"));let c;n!=null&&(c=I(n,"offset","batchNorm")),E(a.rank===u.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),E(c==null||a.rank===c.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),E(l==null||a.rank===l.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let m={x:VE(i),scale:l,offset:c,mean:a,variance:u},f={varianceEpsilon:s},d=k.runKernel(os,m,f);return R(d,i.shape)}var Di=T({batchNorm_:lq});function uq(r,t,e,n,o,s){let i=I(r,"x","batchNorm"),a=I(t,"mean","batchNorm"),u=I(e,"variance","batchNorm"),l;o!=null&&(l=I(o,"scale","batchNorm"));let c;return n!=null&&(c=I(n,"offset","batchNorm")),E(i.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),E(a.rank===2||a.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${a.rank}.`),E(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===2||c.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var xx=T({batchNorm2d_:uq});function cq(r,t,e,n,o,s){let i=I(r,"x","batchNorm"),a=I(t,"mean","batchNorm"),u=I(e,"variance","batchNorm"),l;o!=null&&(l=I(o,"scale","batchNorm"));let c;return n!=null&&(c=I(n,"offset","batchNorm")),E(i.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),E(a.rank===3||a.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${a.rank}.`),E(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===3||c.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var yx=T({batchNorm3d_:cq});function pq(r,t,e,n,o,s){let i=I(r,"x","batchNorm"),a=I(t,"mean","batchNorm"),u=I(e,"variance","batchNorm"),l;o!=null&&(l=I(o,"scale","batchNorm"));let c;return n!=null&&(c=I(n,"offset","batchNorm")),E(i.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),E(a.rank===4||a.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${a.rank}.`),E(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${u.rank}.`),l!=null&&E(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${l.rank}.`),c!=null&&E(c.rank===4||c.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${c.rank}.`),Di(i,a,u,c,l,s)}var bx=T({batchNorm4d_:pq});function mq(r,t,e){let n=I(r,"x","bincount"),o=I(t,"weights","bincount");E(n.dtype==="int32",()=>`Error in bincount: input dtype must be int32, but got ${n.dtype}`),E(e>=0,()=>`size must be non-negative, but got ${e}.`),E(o.size===n.size||o.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${n.shape}, weights shape: ${o.shape}.`);let s={x:n,weights:o},i={size:e};return k.runKernel(up,s,i)}var wx=T({bincount_:mq});function fq(r,t){let e=I(r,"s0","broadcastArgs","int32"),n=I(t,"s1","broadcastArgs","int32");if(e.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${e.rank}`);if(n.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${n.rank}`);let o={s0:e,s1:n};return k.runKernel(cp,o)}var GE=T({broadcastArgs_:fq});function dq(r,t){let e=I(r,"broadcastTo","x"),n=e.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<e.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${e.rank}.`);if(t.length>e.rank){let l=e.shape.slice();for(;l.length<t.length;)l.unshift(1);e=R(e,l)}let o=e.shape,s=Array.from(t);for(let l=t.length-1;l>=0;l--)if(o[l]===t[l])s[l]=1;else if(e.shape[l]!==1)throw new Error(`broadcastTo(): [${n}] cannot be broadcast to [${t}].`);if(s.map((l,c)=>l>1?c:-1).filter(l=>l>=0).length===0)return sn(e);let a={x:e},u={reps:s};return k.runKernel(Jn,a,u)}var Ri=T({broadcastTo_:dq});function hq(r){let e={x:I(r,"x","ceil","float32")};return k.runKernel(qo,e)}var Cx=T({ceil_:hq});function xo(r,t,e){let n={shape:r,value:t,dtype:e};return k.runKernel(Dl,{},n)}function gq(r,t,e){let n=I(r,"x","clipByValue");if(E(t<=e,()=>`Error in clip: min (${t}) must be less than or equal to max (${e}).`),t===e)return xo(n.shape,t,n.dtype);let o={x:n},s={clipValueMin:t,clipValueMax:e};return k.runKernel(uo,o,s)}var Cr=T({clipByValue_:gq});function xq(r){return ne(r,0)}var Ix=T({concat1d_:xq});function yq(r,t){return ne(r,t)}var Sx=T({concat2d_:yq});function bq(r,t){return ne(r,t)}var vx=T({concat3d_:bq});function wq(r,t){return ne(r,t)}var Nx=T({concat4d_:wq});function Cq(r,t,e,n,o="NHWC",s=[1,1],i){let a=I(r,"x","conv2d","float32"),u=I(t,"filter","conv2d","float32"),l=a,c=!1;a.rank===3&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(l.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${l.rank}.`),E(u.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${u.rank}.`),Ie("conv2d",n,i);let p=o==="NHWC"?l.shape[3]:l.shape[1];E(p===u.shape[2],()=>`Error in conv2d: depth of input (${p}) must match input depth for filter ${u.shape[2]}.`),E(Ar(e,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`);let m={x:l,filter:u},f={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i},d=k.runKernel(Ko,m,f);return c?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var In=T({conv2d_:Cq});function Iq(r,t,e,n,o="NWC",s=1,i){let a=I(r,"x","conv1d"),u=I(t,"filter","conv1d"),l=a,c=!1;a.rank===2&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1]])),E(l.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${l.rank}.`),E(u.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${u.rank}.`),Ie("conv1d",n,i),E(l.shape[2]===u.shape[1],()=>`Error in conv1d: depth of input (${l.shape[2]}) must match input depth for filter ${u.shape[1]}.`),E(Ar(e,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${e} and dilation '${s}'`),E(o==="NWC",()=>`Error in conv1d: got dataFormat of ${o} but only NWC is currently supported.`);let p=R(u,[1,u.shape[0],u.shape[1],u.shape[2]]),m=R(l,[l.shape[0],1,l.shape[1],l.shape[2]]),g=In(m,p,[1,e],n,"NHWC",[1,s],i);return c?R(g,[g.shape[2],g.shape[3]]):R(g,[g.shape[0],g.shape[2],g.shape[3]])}var Qp=T({conv1d_:Iq});function Sq(r,t,e,n,o,s="NHWC",i){E(r.length===t.rank,()=>`Length of inShape (${r.length}) and rank of dy (${t.rank}) must match`);let a=r,u=t,l=!1;t.rank===3&&(l=!0,u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]),a=[1,r[0],r[1],r[2]]),E(a.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${a.length}.`),E(u.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${u.rank}`),E(e.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${e.rank}`);let c=s==="NHWC"?a[3]:a[1],p=s==="NHWC"?u.shape[3]:u.shape[1];E(c===e.shape[2],()=>`Error in conv2dDerInput: depth of input (${c}) must match input depth for filter ${e.shape[2]}.`),E(p===e.shape[3],()=>`Error in conv2dDerInput: depth of output (${p}) must match output depth for filter ${e.shape[3]}.`),Ie("conv2dDerInput",o,i);let m={dy:u,filter:e},f={strides:n,pad:o,dataFormat:s,dimRoundingMode:i,inputShape:a},d=k.runKernel(jo,m,f);return l?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var tm=T({conv2DBackpropInput_:Sq});function vq(r,t,e,n,o,s){let i=I(r,"x","conv2dTranspose"),a=I(t,"filter","conv2dTranspose");return tm(e,i,a,n,o,"NHWC",s)}var em=T({conv2dTranspose_:vq});function Nq(r,t,e,n,o="NDHWC",s=[1,1,1]){let i=I(r,"x","conv3d"),a=I(t,"filter","conv3d"),u=i,l=!1;i.rank===4&&(l=!0,u=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(u.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${u.rank}.`),E(a.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${a.rank}.`),E(u.shape[4]===a.shape[3],()=>`Error in conv3d: depth of input (${u.shape[4]}) must match input depth for filter ${a.shape[3]}.`),E(Ar(e,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`),E(o==="NDHWC",()=>`Error in conv3d: got dataFormat of ${o} but only NDHWC is currently supported.`);let c={x:u,filter:a},p={strides:e,pad:n,dataFormat:o,dilations:s},m=k.runKernel(Al,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3],m.shape[4]]):m}var Tx=T({conv3d_:Nq});function Tq(r,t,e,n,o){E(r.length===t.rank,()=>`Length of inShape (${r.length}) and rank of dy (${t.rank}) must match`);let s=r,i=t,a=!1;t.rank===4&&(a=!0,i=R(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,r[0],r[1],r[2],r[3]]);let u=s[4],l=i.shape[4];E(s.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),E(i.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),E(e.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${e.rank}`),E(u===e.shape[3],()=>`Error in conv3dDerInput: depth of input (${u}) must match input depth for filter ${e.shape[3]}.`),E(l===e.shape[4],()=>`Error in conv3dDerInput: depth of output (${l}) must match output depth for filter ${e.shape[4]}.`);let c={dy:i,filter:e},p={pad:o,strides:n,inputShape:s},m=k.runKernel(dp,c,p);return a?R(m,[m.shape[1],m.shape[2],m.shape[3],m.shape[4]]):m}var kx=T({conv3DBackpropInput_:Tq});function kq(r,t,e,n,o){let s=I(r,"x","conv3dTranspose"),i=I(t,"filter","conv3dTranspose");return kx(e,s,i,n,o)}var Ex=T({conv3dTranspose_:kq});function Eq(r){let e={x:I(r,"x","cos","float32")};return k.runKernel(Xo,e)}var Jl=T({cos_:Eq});function _q(r){let e={x:I(r,"x","cosh","float32")};return k.runKernel(Yo,e)}var rm=T({cosh_:_q});function Aq(r,t=0,e=!1,n=!1){let s={x:I(r,"x","cumprod")},i={axis:t,exclusive:e,reverse:n};return k.runKernel(fa,s,i)}var Xu=T({cumprod_:Aq});function $q(r,t=0,e=!1,n=!1){let s={x:I(r,"x","cumsum")},i={axis:t,exclusive:e,reverse:n};return k.runKernel(Zo,s,i)}var nm=T({cumsum_:$q});function Dq(r,t,e,n=!1){let o=I(r,"x","denseBincount"),s=I(t,"weights","denseBincount");E(o.dtype==="int32",()=>`Error in denseBincount: input dtype must be int32, but got ${o.dtype}`),E(o.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${o.rank}.`),E(e>=0,()=>`size must be non-negative, but got ${e}.`),E(s.size===o.size||s.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${o.shape}, weights shape: ${s.shape}.`);let i={x:o,weights:s},a={size:e,binaryOutput:n};return k.runKernel(hp,i,a)}var ch=T({denseBincount_:Dq});function Rq(r,t,e="NHWC"){let n=I(r,"x","depthToSpace","float32"),o=e==="NHWC"?n.shape[1]:n.shape[2],s=e==="NHWC"?n.shape[2]:n.shape[3],i=e==="NHWC"?n.shape[3]:n.shape[1];E(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),E(o*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${o} and ${t}  for depthToSpace with input shape
-    ${n.shape}`),E(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying
-    ${s} and ${t} for depthToSpace with input shape
-        ${n.shape}`),E(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${n.shape}`);let a={x:n},u={blockSize:t,dataFormat:e};return k.runKernel(ha,a,u)}var _x=T({depthToSpace_:Rq});function Fq(r,t,e,n,o="NHWC",s=[1,1],i){let a=I(r,"x","depthwiseConv2d","float32"),u=I(t,"filter","depthwiseConv2d","float32"),l=a,c=!1;a.rank===3&&(c=!0,l=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(l.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${l.rank}.`),E(u.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${u.rank}.`);let p=o==="NHWC"?l.shape[3]:l.shape[1];E(p===u.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${p}) must match the inChannels dimension in filter ${u.shape[2]}.`),Ie("depthwiseConv2d",n,i);let m={x:l,filter:u},f={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i},d=k.runKernel(Jo,m,f);return c?R(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var Fi=T({depthwiseConv2d_:Fq});function Oq(r){let e={x:I(r,"x","diag")};return k.runKernel(yp,e)}var WE=T({diag_:Oq});function Pq(r,t,e,n,o=[1,1],s="NHWC"){let i=I(r,"x","dilation2d"),a=I(t,"filter","dilation2d");E(i.rank===3||i.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),E(a.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${a.rank}.`),E(s==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let u=i,l=!1;i.rank===3&&(u=R(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=!0);let c={x:u,filter:a},p={strides:e,pad:n,dilations:o},m=k.runKernel($l,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var Ax=T({dilation2d_:Pq});function Lq(r,t){let e=I(r,"a","equal","string_or_numeric"),n=I(t,"b","equal","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(xa,o)}var $r=T({equal_:Lq});function Mq(r,t,e){let n=I(t,"a","where"),o=I(e,"b","where"),s=I(r,"condition","where","bool"),i=Pt(Pt(s.shape,n.shape),o.shape),a=Ri(s,i),u=Ri(n,i),l=Ri(o,i),c={condition:a,t:u,e:l};return k.runKernel(hi,c)}var _e=T({where_:Mq});function zq(r){let e={x:I(r,"x","zerosLike")};return k.runKernel(wi,e)}var It=T({zerosLike_:zq});function Bq(r,t){let e=I(r,"a","div"),n=I(t,"b","div");[e,n]=Ut(e,n);let o=pt(e,n),s=It(o),i=$r(n,s);return _e(i,s,o)}var $x=T({divNoNan_:Bq});function Vq(r,t){let e=I(r,"t1","dot"),n=I(t,"t2","dot");E((e.rank===1||e.rank===2)&&(n.rank===1||n.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${e.rank} and ${n.rank}.`);let o=e.rank===1?e.size:e.shape[1],s=n.rank===1?n.size:n.shape[0];if(E(o===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${o} and ${s}.`),e.rank===1&&n.rank===1){let i=R(e,[1,-1]),a=R(n,[-1,1]),u=Lt(i,a);return R(u,[])}else if(e.rank===1&&n.rank===2){let i=R(e,[1,-1]),a=R(n,[n.shape[0],n.shape[1]]),u=Lt(i,a);return R(u,[u.size])}else if(e.rank===2&&n.rank===1){let i=R(n,[-1,1]),a=Lt(e,i);return R(a,[a.size])}else{let i=R(n,[n.shape[0],n.shape[1]]);return Lt(e,i)}}var Dx=T({dot_:Vq});function Gq(r,...t){let e=t.map((o,s)=>I(o,`tensors${s}`,"einsum")),n={equation:r};return k.runKernel(bp,e,n)}var UE=T({einsum_:Gq});function Wq(r){let e={x:I(r,"x","elu","float32")};return k.runKernel(ts,e)}var Oi=T({elu_:Wq});function Uq(r){let t=I(r,"x","erf");E(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=J(t,"float32"));let e={x:t};return k.runKernel(ga,e)}var Rx=T({erf_:Uq});function J0(r,t){for(let e=0;e<r.length;++e)if(r[r.length-e-1]!==t-1-e)return!1;return!0}function HE(r,t,e){let n=r.length+t.length,o=[],s=0,i=0;for(let a=0;a<n;a++)e.indexOf(a)===-1?o.push(r[s++]):o.push(t[i++]);return o}function Q0(r,t){let e=[],n=r.length;for(let s=0;s<n;s++)t.indexOf(s)===-1&&e.push(r[s]);let o=t.map(s=>r[s]);return[e,o]}function yo(r,t){let e=t.map(n=>1);return HE(r,e,t)}function Hq(r,t,e){E(J0(t,e),()=>`${r} supports only inner-most axes for now. Got axes ${t} and rank-${e} input.`)}function tv(r,t){if(J0(r,t))return null;let e=[];for(let n=0;n<t;++n)r.indexOf(n)===-1&&e.push(n);return r.forEach(n=>e.push(n)),e}function ph(r){return r.map((t,e)=>[e,t]).sort((t,e)=>t[1]-e[1]).map(t=>t[0])}function qq(r,t){let e=[];for(let n=t-r;n<t;++n)e.push(n);return e}function Kq(r,t=null,e=!1){let o={x:I(r,"x","max")},s={reductionIndices:t,keepDims:e};return k.runKernel(ls,o,s)}var Ir=T({max_:Kq});function jq(r,t=null,e=!1){let o={x:I(r,"x","min")},s={axis:t,keepDims:e};return k.runKernel(ms,o,s)}var Ja=T({min_:jq});function Xq(r,t){let e=I(r,"base","pow"),n=I(t,"exp","pow");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(ys,o)}var an=T({pow_:Xq});function mt(r,t){if((xr(r)&&t!=="string"||Array.isArray(r))&&t!=="complex64")throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if(t==="string"&&xr(r)&&!(r instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return on(r,[],[],t)}function Yq(r){let e={x:I(r,"x","sqrt","float32")};return k.runKernel(As,e)}var Se=T({sqrt_:Yq});function Zq(r){let t=I(r,"x","square"),e={};return k.runKernel("Square",{x:t},e)}var Mt=T({square_:Zq});function Jq(r,t=null,e=!1){let n=I(r,"x","sum");n.dtype==="bool"&&(n=J(n,"int32"));let o={x:n},s={axis:t,keepDims:e};return k.runKernel($s,o,s)}var ft=T({sum_:Jq});function Qq(r,t="euclidean",e=null,n=!1){r=I(r,"x","norm");let o=qE(r,t,e),s=o.shape;if(n){let i=lr(e,r.shape);s=yo(o.shape,i)}return R(o,s)}function qE(r,t,e=null){if(r.rank===0)return Ee(r);if(r.rank!==1&&e===null)return qE(R(r,[-1]),t,e);if(r.rank===1||typeof e=="number"||Array.isArray(e)&&e.length===1){if(t===1)return ft(Ee(r),e);if(t===1/0)return Ir(Ee(r),e);if(t===-1/0)return Ja(Ee(r),e);if(t==="euclidean"||t===2)return Se(ft(an(Ee(r),mt(2,"int32")),e));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(e)&&e.length===2){if(t===1)return Ir(ft(Ee(r),e[0]),e[1]-1);if(t===1/0)return Ir(ft(Ee(r),e[1]),e[0]);if(t===-1/0)return Ja(ft(Ee(r),e[1]),e[0]);if(t==="fro"||t==="euclidean")return Se(ft(Mt(r),e));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${e}`)}var Qa=T({norm_:Qq});function tK(r,t=null,e=!1){return Qa(r,"euclidean",t,e)}var Fx=T({euclideanNorm_:tK});function eK(r){let e={x:I(r,"x","exp")};return k.runKernel(es,e)}var er=T({exp_:eK});function rK(r,t=0){let e=I(r,"x","expandDims","string_or_numeric");E(t<=e.rank,()=>"Axis must be <= rank of the tensor");let n={input:e},o={dim:t};return k.runKernel(ui,n,o)}var rr=T({expandDims_:rK});function nK(r){let e={x:I(r,"x","expm1")};return k.runKernel(ya,e)}var Ox=T({expm1_:nK});function oK(r,t){let e=I(r,"x","tile","string_or_numeric");E(e.rank===t.length,()=>`Error in transpose: rank of input ${e.rank} must match length of reps ${t}.`);let n={x:e},o={reps:t};return k.runKernel(Jn,n,o)}var Dr=T({tile_:oK});function sK(r,t,e,n="float32"){t==null&&(t=r);let o=wt([r,t],n),s=r<=t?r:t;for(let a=0;a<s;++a)o.set(1,a,a);let i=R(o.toTensor(),[r,t]);if(e==null)return i;if(e.length===1)return Dr(rr(i,0),[e[0],1,1]);if(e.length===2)return Dr(rr(rr(i,0),0),[e[0],e[1],1,1]);if(e.length===3)return Dr(rr(rr(rr(i,0),0),0),[e[0],e[1],e[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${e.length}D.`)}var Yu=T({eye_:sK});function iK(r){let e={x:I(r,"x","floor","float32")};return k.runKernel(rs,e)}var Pi=T({floor_:iK});function aK(r,t,e=0,n=0){let o=I(r,"x","gather"),s=I(t,"indices","gather","int32"),i={x:o,indices:s},a={axis:e,batchDims:n};return k.runKernel(ci,i,a)}var Li=T({gather_:aK});function lK(r,t){let e=I(r,"a","greater","string_or_numeric"),n=I(t,"b","greater","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ca,o)}var Re=T({greater_:lK});function uK(r,t){let e=I(r,"a","greaterEqual","string_or_numeric"),n=I(t,"b","greaterEqual","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(ss,o)}var ln=T({greaterEqual_:uK});function cK(r){let e={x:I(r,"x","isFinite")};return k.runKernel(Ia,e)}var Px=T({isFinite_:cK});function pK(r){let e={x:I(r,"x","isInf")};return k.runKernel(Sa,e)}var Lx=T({isInf_:pK});function mK(r){let e={x:I(r,"x","isNaN")};return k.runKernel(va,e)}var Mx=T({isNaN_:mK});function fK(r,t=.2){let n={x:I(r,"x","leakyRelu")},o={alpha:t};return k.runKernel(is,n,o)}var Ql=T({leakyRelu_:fK});function dK(r,t){let e=I(r,"a","less","string_or_numeric"),n=I(t,"b","less","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Na,o)}var om=T({less_:dK});function hK(r,t){let e=I(r,"a","lessEqual","string_or_numeric"),n=I(t,"b","lessEqual","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ta,o)}var Ln=T({lessEqual_:hK});function KE(r,t,e){if(e<=0)throw new Error("The number of values should be positive.");let n={start:r,stop:t,num:e};return k.runKernel(vp,{},n)}function gK(r,t=5,e=1,n=1,o=.5){let s=I(r,"x","localResponseNormalization");E(s.rank===4||s.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
-               rank ${s.rank}.`),E(na(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,a=!1;s.rank===3&&(a=!0,i=R(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let u={x:i},l={depthRadius:t,bias:e,alpha:n,beta:o},c=k.runKernel(Rl,u,l);return a?R(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var zx=T({localResponseNormalization_:gK});function xK(r){let e={x:I(r,"x","log","float32")};return k.runKernel(as,e)}var Sr=T({log_:xK});function yK(r){let e={x:I(r,"x","log1p")};return k.runKernel(ka,e)}var tu=T({log1p_:yK});function bK(r){return E(oi(r),()=>"The f passed in grad(f) must be a function"),(t,e)=>{let n=I(t,"x","tf.grad","string_or_numeric"),o=e!=null?I(e,"dy","tf.grad"):null;return k.tidy(()=>{let{value:s,grads:i}=k.gradients(()=>r(n),[n],o);return o!=null&&$e(s.shape,o.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Vx(i),i[0]})}}function wK(r){return E(oi(r),()=>"The f passed in grads(f) must be a function"),(t,e)=>{E(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let n=ja(t,"args","tf.grads","string_or_numeric"),o=e!=null?I(e,"dy","tf.grads"):null;return k.tidy(()=>{let{value:s,grads:i}=k.gradients(()=>r(...n),n,o);return o!=null&&$e(s.shape,o.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Vx(i),i})}}function CK(r){return E(oi(r),()=>"The f passed in valueAndGrad(f) must be a function"),(t,e)=>{E(t instanceof Ft,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),E(e==null||e instanceof Ft,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:n,value:o}=k.gradients(()=>r(t),[t],e);return Vx(n),{grad:n[0],value:o}}}function IK(r){return E(oi(r),()=>"The f passed in valueAndGrads(f) must be a function"),(t,e)=>{E(Array.isArray(t)&&t.every(o=>o instanceof Ft),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),E(e==null||e instanceof Ft,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let n=k.gradients(()=>r(...t),t,e);return e!=null&&$e(n.value.shape,e.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Vx(n.grads),n}}function Bx(r,t){E(oi(r),()=>"The f passed in variableGrads(f) must be a function"),E(t==null||Array.isArray(t)&&t.every(l=>l instanceof Ka),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let e=t!=null;if(!e){t=[];for(let l in k.registeredVariables)t.push(k.registeredVariables[l])}let n=e?t.filter(l=>!l.trainable):null,o=t.length;t=t.filter(l=>l.trainable),E(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${o} variables is trainable.`);let s=!0,{value:i,grads:a}=k.gradients(r,t,null,s);E(a.some(l=>l!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),E(i.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${i.rank} tensor`);let u={};return t.forEach((l,c)=>{a[c]!=null&&(u[l.name]=a[c])}),n!=null&&n.forEach(l=>u[l.name]=null),{value:i,grads:u}}function un(r){return k.customGrad(r)}function Vx(r){if(r.filter(e=>e==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
-    the f you passed encloses all operations that lead from x to y.`)}function SK(r){let e={x:I(r,"x","softplus")};return k.runKernel(Va,e)}var zs=T({softplus_:SK});function vK(r){let t=I(r,"x","logSigmoid");return un(n=>({value:Ht(zs(Ht(n))),gradFunc:i=>D(i,Yr(Ht(n)))}))(t)}var Gx=T({logSigmoid_:vK});function NK(r,t){let e=I(r,"a","sub"),n=I(t,"b","sub");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel(Fs,o)}var ct=T({sub_:NK});function TK(r,t=-1){let e=I(r,"logits","logSoftmax");if(t===-1&&(t=e.rank-1),t!==e.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${e.rank} and axis was ${t}`);return un((o,s)=>{let a=Ir(o,t,!0),u=ct(o,a),l=ct(J(u,"float32"),Sr(ft(er(u),t,!0)));return s([l]),{value:l,gradFunc:(p,m)=>{let[f]=m,d=!0,h=er(f);return ct(p,D(ft(p,t,d),h))}}})(e)}var sm=T({logSoftmax_:TK});function kK(r,t=null,e=!1){let n=I(r,"x","logSumExp"),o=lr(t,n.shape),s=Ir(n,o,!0),i=ct(n,s),a=er(i),u=ft(a,o),l=Sr(u),c=X(R(s,l.shape),l);if(e){let p=yo(c.shape,o);return R(c,p)}return c}var im=T({logSumExp_:kK});function EK(r,t){let e=I(r,"a","logicalAnd","bool"),n=I(t,"b","logicalAnd","bool");Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Ea,o)}var Rr=T({logicalAnd_:EK});function _K(r){let e={x:I(r,"x","logicalNot","bool")};return k.runKernel(_a,e)}var eu=T({logicalNot_:_K});function AK(r,t){let e=I(r,"a","logicalOr","bool"),n=I(t,"b","logicalOr","bool");Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Aa,o)}var am=T({logicalOr_:AK});function $K(r,t){let e=I(r,"a","logicalXor","bool"),n=I(t,"b","logicalXor","bool");return Pt(e.shape,n.shape),Rr(am(r,t),eu(Rr(r,t)))}var Wx=T({logicalXor_:$K});var Ux=2147483648;function DK(r,t,e="left"){let n=I(r,"sortedSequence","searchSorted"),o=I(t,"values","searchSorted"),s=n.shape[n.shape.length-1],i=o.shape[o.shape.length-1],a=R(n,[-1,s]),u=R(o,[-1,i]);if(a.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(a.shape[0]!==u.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(Jt(u.shape)>=Ux)throw new Error(`values tensor size must less than ${Ux}`);if(a.shape[1]>=Ux)throw new Error(`trailing dim_size must less than ${Ux} for int32 output type, was ${a.shape[1]}`);let l={sortedSequence:a,values:u},c={side:e};return k.runKernel(Pp,l,c)}var mh=T({searchSorted_:DK});function jE(r,t){return mh(r,t,"left")}function RK(r,t,e,n,o){let s=I(r,"x","maxPool"),i=1,a=s,u=!1;s.rank===3&&(u=!0,a=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(a.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${a.rank}.`),E(Ar(e,i),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${e} and dilations '${i}'`),Ie("maxPool",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o},p=k.runKernel(cs,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var ru=T({maxPool_:RK});function FK(r,t=[1,1,1],e,n,o,s="NDHWC"){let i=I(r,"x","maxPool3d"),a=i,u=!1;i.rank===4&&(u=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),E(a.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${a.rank}.`),E(s==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Ie("maxPool3d",n,o);let l={x:a},c={filterSize:t,strides:e,pad:n,dimRoundingMode:o,dataFormat:s},p=k.runKernel(Fl,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var Hx=T({maxPool3d_:FK});function OK(r,t,e,n,o=!1){let i={x:I(r,"x","maxPoolWithArgmax")},a={filterSize:t,strides:e,pad:n,includeBatchInIndex:o},u=k.runKernel(Ep,i,a);return{result:u[0],indexes:u[1]}}var XE=T({maxPoolWithArgmax_:OK});function PK(r,t){let e=I(r,"a","maximum"),n=I(t,"b","maximum");[e,n]=Ut(e,n),e.dtype==="bool"&&(e=J(e,"int32"),n=J(n,"int32")),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(us,o)}var Sn=T({maximum_:PK});function LK(r,t=null,e=!1){let o={x:I(r,"x","mean")},s={axis:t,keepDims:e};return k.runKernel(ps,o,s)}var ve=T({mean_:LK});function Ne(r,t="float32"){if(t==="complex64"){let n=Ne(r,"float32"),o=Ne(r,"float32");return wn(n,o)}let e=ip(Jt(r),t);return k.makeTensor(e,r,t)}function cr(r,t="float32"){if(t==="complex64"){let n=cr(r,"float32"),o=Ne(r,"float32");return wn(n,o)}let e=Wd(Jt(r),t);return k.makeTensor(e,r,t)}function YE(r,t,{indexing:e="xy"}={}){if(e!=="xy"&&e!=="ij")throw new TypeError(`${e} is not a valid third argument to meshgrid`);if(r===void 0)return[];let n=I(r,"x","meshgrid",r instanceof Ft?r.dtype:"float32");if(t===void 0)return[n];let o=I(t,"y","meshgrid",t instanceof Ft?t.dtype:"float32"),s=Jt(n.shape),i=Jt(o.shape);return e==="xy"?(n=R(n,[1,-1]),o=R(o,[-1,1]),[Lt(cr([i,1],n.dtype),n),Lt(o,cr([1,s],o.dtype))]):(n=R(n,[-1,1]),o=R(o,[1,-1]),[Lt(n,cr([1,i],n.dtype)),Lt(cr([s,1],o.dtype),o)])}function MK(r,t){let e=I(r,"a","minimum"),n=I(t,"b","minimum");[e,n]=Ut(e,n),e.dtype==="bool"&&(e=J(e,"int32"),n=J(n,"int32")),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(fs,o)}var Mi=T({minimum_:MK});function zK(r,t,e){E(e==="reflect"||e==="symmetric",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${e}.`);let n=I(r,"x","mirrorPad");if(n.rank===0)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");E(t.length===n.rank,()=>`Padding doesn't match input. Must be ${n.rank}. Got ${t.length}.`);let o=e==="reflect"?1:0;for(let a=0;a<n.rank;a++)E(t[a].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),E(t[a][0]>=0&&t[a][0]<=n.shape[a]-o&&t[a][1]>=0&&t[a][1]<=n.shape[a]-o,()=>`Padding in dimension ${a} cannot be greater than or equal to ${n.shape[a]-o} or less than 0 for input of shape ${n.shape}`);let s={paddings:t,mode:e},i={x:n};return k.runKernel(ds,i,s)}var qx=T({mirrorPad_:zK});function BK(r,t){let e=I(r,"a","mod"),n=I(t,"b","mod");[e,n]=Ut(e,n);let o={a:e,b:n};return k.runKernel($a,o)}var Kx=T({mod_:BK});function VK(r,t=null,e=!1){r=I(r,"x","moments");let n=lr(t,r.shape),o=ve(r,n,e),s=o.shape;e||(s=yo(o.shape,n));let i=Mt(ct(J(r,"float32"),R(o,s))),a=ve(i,n,e);return{mean:o,variance:a}}var Zu=T({moments_:VK});function GK(r,t,e,n){let o=I(t,"data","multiRNNCell"),s=ja(e,"c","multiRNNCell"),i=ja(n,"h","multiRNNCell"),a=o,u=[];for(let p=0;p<r.length;p++){let m=r[p](a,s[p],i[p]);u.push(m[0]),u.push(m[1]),a=m[1]}let l=[],c=[];for(let p=0;p<u.length;p+=2)l.push(u[p]),c.push(u[p+1]);return[l,c]}var ZE=T({multiRNNCell_:GK});function WK(r,t,e,n=!1){let o=I(r,"logits","multinomial"),s=o.size,i=o.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);e=e||Math.random();let u={logits:i===1?R(o,[1,-1]):o},l={numSamples:t,seed:e,normalized:n},c=k.runKernel(_p,u,l);return i===1?R(c,[c.size]):c}var JE=T({multinomial_:WK});function UK(r,t){let e=I(r,"a","notEqual","string_or_numeric"),n=I(t,"b","notEqual","string_or_numeric");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n};return k.runKernel(Da,o)}var Bs=T({notEqual_:UK});function HK(r){let e={x:I(r,"x","onesLike")};return k.runKernel(mi,e)}var yr=T({onesLike_:HK});function qK(r,t){let e=I(r,"v1","outerProduct"),n=I(t,"v2","outerProduct");E(e.rank===1&&n.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${e.rank} and ${n.rank}.`);let o=R(e,[-1,1]),s=R(n,[1,-1]);return Lt(o,s)}var QE=T({outerProduct_:qK});function KK(r,t,e=0){let n=I(r,"x","pad");if(n.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let o={paddings:t,constantValue:e},s={x:n};return k.runKernel(xs,s,o)}var cn=T({pad_:KK});function jK(r,t,e=0){return E(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),cn(r,[t],e)}var t_=T({pad1d_:jK});function XK(r,t,e=0){return E(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),cn(r,t,e)}var e_=T({pad2d_:XK});function YK(r,t,e=0){return E(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),cn(r,t,e)}var r_=T({pad3d_:YK});function ZK(r,t,e=0){return E(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),cn(r,t,e)}var n_=T({pad4d_:ZK});function JK(r,t,e){let n=I(r,"x","spaceToBatchND");E(n.rank>=1+t.length,()=>`input rank ${n.rank} should be > than [blockShape] ${t.length}`),E(e.length===t.length,()=>`paddings.shape[0] ${e.length} must be equal to [blockShape] ${t.length}`),E(n.shape.reduce((i,a,u)=>u>0&&u<=t.length?i&&(a+e[u-1][0]+e[u-1][1])%t[u-1]===0:i,!0),()=>`input spatial dimensions ${n.shape.slice(1)} with paddings ${e.toString()} must be divisible by blockShapes ${t.toString()}`);let o={x:n},s={blockShape:t,paddings:e};return k.runKernel(xi,o,s)}var nu=T({spaceToBatchND_:JK});function QK(r,t,e,n,o,s,i){o==null&&(o=[1,1]),s==null&&(s=1),n===0&&(n="valid");let a=I(r,"x","maxPool"),u=a,l=!1;a.rank===3&&(l=!0,u=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),E(Ar(s,o),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${o}'`);let c=Y0(u.shape,t,s,o,n),p=[c.dilationHeight,c.dilationWidth],m;n==="same"?m=ej([c.filterHeight,c.filterWidth],p):m=[[0,0],[0,0]];let f=p[0]===1&&p[1]===1,[d,h]=tj([c.inHeight,c.inWidth],p,m),g=f?n:"valid",x=f?u:nu(u,p,d),w=(e==="avg"?()=>Yl(x,t,s,g,i):()=>ru(x,t,s,g,i))(),C=f?w:Zl(w,p,h);return l?R(C,[C.shape[1],C.shape[2],C.shape[3]]):C}function tj(r,t,e){let n=e.map(c=>c[0]),o=e.map(c=>c[1]),s=r.concat(n,o),i=t.map((c,p)=>(c-s[p]%c)%c),a=o.map((c,p)=>c+i[p]),u=t.map((c,p)=>[n[p],a[p]]),l=t.map((c,p)=>[0,i[p]]);return[u,l]}function ej(r,t){let n=r.map((i,a)=>i+(i-1)*(t[a]-1)).map(i=>i-1),o=n.map(i=>Math.floor(i/2)),s=n.map((i,a)=>i-o[a]);return n.map((i,a)=>[o[a],s[a]])}var jx=T({pool_:QK});function rj(r,t){let e=I(r,"x","prelu"),n=I(t,"alpha","prelu"),o={x:e,alpha:n};return k.runKernel(bs,o)}var ou=T({prelu_:rj});function nj(r,t=null,e=!1){let n=I(r,"x","prod");n.dtype==="bool"&&(n=J(n,"int32"));let o={x:n},s={axis:t,keepDims:e};return k.runKernel(ws,o,s)}var Xx=T({prod_:nj});function oj(r,t,e,n){let o=r.map((c,p)=>I(c,`tensors${p}`,"raggedGather","int32")),s=I(t,"paramsDenseValues","raggedGather"),i=I(e,"indices","raggedGather","int32"),a={paramsNestedSplits:o,paramsDenseValues:s,indices:i},u={outputRaggedRank:n},l=k.runKernel(Ap,a,u);return{outputNestedSplits:l.slice(0,l.length-1),outputDenseValues:l[l.length-1]}}var o_=T({raggedGather_:oj});function sj(r,t,e){let n=I(r,"starts","raggedRange"),o=I(t,"limits","raggedRange",n.dtype),s=I(e,"deltas","raggedRange",n.dtype),i={starts:n,limits:o,deltas:s},a=k.runKernel($p,i);return{rtNestedSplits:a[0],rtDenseValues:a[1]}}var s_=T({raggedRange_:sj});function ij(r,t,e,n,o){let s=I(r,"shape","raggedTensorToTensor","int32"),i=I(t,"values","raggedTensorToTensor"),a=I(e,"defaultValue","raggedTensorToTensor",i.dtype),u=n.map((p,m)=>I(p,`tensors${m}`,"raggedTensorToTensor","int32")),l={shape:s,values:i,defaultValue:a,rowPartitionTensors:u},c={rowPartitionTypes:o};return k.runKernel(Dp,l,c)}var i_=T({raggedTensorToTensor_:ij});function aj(r,t,e){let n=Jt(r),o=null;if(e==null||e==="float32")o=new Float32Array(n);else if(e==="int32")o=new Int32Array(n);else if(e==="bool")o=new Uint8Array(n);else throw new Error(`Unknown data type ${e}`);for(let s=0;s<n;s++)o[s]=t();return k.makeTensor(o,r,e)}var a_=T({rand_:aj});var Qx=Tl(fh());var Qu=class{constructor(t,e,n,o,s){this.mean=t,this.stdDev=e,this.dtype=n,this.nextVal=NaN,this.truncated=o,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let i=s||Math.random();this.random=Qx.alea(i.toString())}nextValue(){if(!isNaN(this.nextVal)){let o=this.nextVal;return this.nextVal=NaN,o}let t,e,n=!1;for(;!n;){let o,s,i;do o=2*this.random()-1,s=2*this.random()-1,i=o*o+s*s;while(i>=1||i===0);let a=Math.sqrt(-2*Math.log(i)/i);t=this.mean+this.stdDev*o*a,e=this.mean+this.stdDev*s*a,(!this.truncated||this.isValidTruncated(t))&&(n=!0)}return(!this.truncated||this.isValidTruncated(e))&&(this.nextVal=this.convertValue(e)),this.convertValue(t)}convertValue(t){return this.dtype==null||this.dtype==="float32"?t:Math.round(t)}isValidTruncated(t){return t<=this.upper&&t>=this.lower}},Zx=class{constructor(t,e,n,o){this.alpha=t,this.beta=1/e,this.dtype=n;let s=o||Math.random();this.randu=Qx.alea(s.toString()),this.randn=new Qu(0,1,n,!1,this.randu()),t<1?this.d=t+2/3:this.d=t-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let t,e,n,o,s,i;for(;;){do o=this.randn.nextValue(),i=1+this.c*o;while(i<=0);if(i*=i*i,t=o*o,e=1-.331*t*t,n=.5*t+this.d*(1-i+Math.log(i)),s=this.randu(),s<e||Math.log(s)<n)break}return i=1/this.beta*this.d*i,this.alpha<1&&(i*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(i)}convertValue(t){return this.dtype==="float32"?t:Math.round(t)}},Jx=class{constructor(t=0,e=1,n,o){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=t,this.range=e-t,this.dtype=n,o==null&&(o=Math.random()),typeof o=="number"&&(o=o.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${t} - ${e} <= 1 and dtype is not float`);this.random=Qx.alea(o)}convertValue(t){return this.canReturnFloat()?t:Math.round(t)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function dj(r,t,e=1,n="float32",o){if(e==null&&(e=1),n==null&&(n="float32"),n!=="float32"&&n!=="int32")throw new Error(`Unsupported data type ${n}`);let s=new Zx(t,e,n,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var v_=T({randomGamma_:dj});function hj(r,t=0,e=1,n,o){if(n!=null&&n==="bool")throw new Error(`Unsupported data type ${n}`);let s=new Qu(t,e,n,!1,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var tc=T({randomNormal_:hj});function gj(r,t,e){if(t!=null&&t==="bool")throw new Error(`Unsupported data type ${t}`);return tc(r,0,1,t,e)}var N_=T({randomStandardNormal_:gj});function xj(r,t=0,e=1,n="float32",o){let s=wt(r,n),i=new Jx(t,e,null,o);for(let a=0;a<s.values.length;a++)s.values[a]=i.nextValue();return s.toTensor()}var zi=T({randomUniform_:xj});function su(r,t,e=1,n="float32"){if(e===0)throw new Error("Cannot have a step of zero");let o={start:r,stop:t,step:e,dtype:n};return k.runKernel(Ol,{},o)}function yj(r){let e={x:I(r,"x","reciprocal")};return k.runKernel(Pa,e)}var ty=T({reciprocal_:yj});function bj(r){let e={x:I(r,"x","relu")};return k.runKernel(Cs,e)}var Fr=T({relu_:bj});function wj(r){let e={x:I(r,"x","relu6")};return k.runKernel(vs,e)}var lm=T({relu6_:wj});function Cj(r,t){let n={x:I(r,"x","reverse")},o={dims:t};return k.runKernel(Ns,n,o)}var pr=T({reverse_:Cj});function Ij(r){let t=I(r,"x","reverse");return E(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),pr(t,0)}var T_=T({reverse1d_:Ij});function Sj(r,t){let e=I(r,"x","reverse");return E(e.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${e.rank}.`),pr(e,t)}var k_=T({reverse2d_:Sj});function vj(r,t){let e=I(r,"x","reverse");return E(e.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${e.rank}.`),pr(e,t)}var E_=T({reverse3d_:vj});function Nj(r,t){let e=I(r,"x","reverse");return E(e.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${e.rank}.`),pr(e,t)}var __=T({reverse4d_:Nj});function Tj(r){let e={x:I(r,"x","round")};return k.runKernel(Ts,e)}var um=T({round_:Tj});function kj(r){let e={x:I(r,"x","rsqrt","float32")};return k.runKernel(ks,e)}var cm=T({rsqrt_:kj});function Ej(r){let e={x:I(r,"x","selu")};return k.runKernel(Ma,e)}var pm=T({selu_:Ej});function _j(r,t,e,n,o,s=[1,1],i="NHWC"){let a=I(r,"x","separableConv2d"),u=I(t,"depthwiseFilter","separableConv2d"),l=I(e,"pointwiseFilter","separableConv2d"),c=a,p=!1;if(a.rank===3&&(p=!0,c=R(a,[1,a.shape[0],a.shape[1],a.shape[2]])),i==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");E(c.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${c.rank}.`),E(u.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${u.rank}.`),E(l.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${u.rank}.`),E(l.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${l.shape[0]}.`),E(l.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${l.shape[1]}.`);let m=u.shape[2],f=u.shape[3];E(l.shape[2]===m*f,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${m*f}, but got ${l.shape[2]}.`);let d=Fi(c,u,n,o,i,s),g=In(d,l,1,"valid",i);return p?R(g,[g.shape[1],g.shape[2],g.shape[3]]):g}var mm=T({separableConv2d_:_j});async function Aj(r,t){let e=I(r,"x","setdiff1d"),n=I(t,"y","setdiff1d");E(e.dtype===n.dtype,()=>`x and y should have the same dtype, but got x (${e.dtype}) and y (${n.dtype}).`),E(e.rank===1,()=>`x should be 1D tensor, but got x (${e.shape}).`),E(n.rank===1,()=>`y should be 1D tensor, but got y (${n.shape}).`);let o=await e.data(),s=await n.data(),i=new Set(s),a=0;for(let c=0;c<o.length;c++)i.has(o[c])||a++;let u=new pe([a],e.dtype),l=new pe([a],"int32");for(let c=0,p=0;c<o.length;c++)i.has(o[c])||(u.values[p]=o[c],l.values[p]=c,p++);return[u.toTensor(),l.toTensor()]}var A_=Aj;function $j(r){let e={x:I(r,"x","sign")};return k.runKernel(Ba,e)}var ey=T({sign_:$j});function Dj(r){let e={x:I(r,"x","sin","float32")};return k.runKernel(Es,e)}var fm=T({sin_:Dj});function Rj(r){let e={x:I(r,"x","sinh")};return k.runKernel(za,e)}var dm=T({sinh_:Rj});function Fj(r,t,e){let n=I(r,"x","slice1d");return E(n.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${n.rank} tensor`),Rt(n,[t],[e])}var hm=T({slice1d_:Fj});function Oj(r,t,e){let n=I(r,"x","slice2d");return E(n.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var dh=T({slice2d_:Oj});function Pj(r,t,e){let n=I(r,"x","slice3d");return E(n.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var gm=T({slice3d_:Pj});function Lj(r,t,e){let n=I(r,"x","slice4d");return E(n.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${n.rank} tensor`),Rt(n,t,e)}var ec=T({slice4d_:Lj});function Mj(r,t=-1){let e=I(r,"logits","softmax","float32");if(t===-1&&(t=e.rank-1),t!==e.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${e.rank} and dim was ${t}`);let n={logits:e},o={dim:t};return k.runKernel(Ds,n,o)}var iu=T({softmax_:Mj});function zj(r){E(r.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${r.dtype}.`);let t={input:r};return k.runKernel(Cp,t)}var au=T({fft_:zj});function Bj(r){E(r.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${r.dtype}.`);let t={input:r};return k.runKernel(Ip,t)}var tl=T({ifft_:Bj});function Vj(r){let t=r.shape[r.shape.length-1],e=r.size/t,n;if(t<=2){let o=R(r,[e,t]);n=tl(o)}else{let o=[e,2*(t-1)],s=R(Za(r),[e,t]),i=R(Xl(r),[e,t]),a=pr(Rt(s,[0,1],[e,t-2]),1),u=D(pr(Rt(i,[0,1],[e,t-2]),1),mt(-1)),l=ne([s,a],1),c=ne([i,u],1),p=R(wn(l,c),[o[0],o[1]]);n=tl(p)}if(n=Za(n),r.rank===3&&r.shape[0]!==0){let o=n,s=r.shape[0];n=R(n,[s,n.shape[0]/s,n.shape[1]]),o.dispose()}return n}var xm=T({irfft_:Vj});function Gj(r,t,e=0){let o={x:I(r,"x","split")},s={numOrSizeSplits:t,axis:e};return k.runKernel(yi,o,s)}var mr=T({split_:Gj});function Wj(r,t){E(r.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${r.dtype}`);let e=r.shape[r.shape.length-1],n=r.size/e,o;if(t!=null&&t<e){let d=r.shape.map(g=>0),h=r.shape.map(g=>g);h[r.shape.length-1]=t,o=Rt(r,d,h),e=t}else if(t!=null&&t>e){let d=r.shape.map(h=>h);d[r.shape.length-1]=t-e,o=ne([r,Ne(d)],r.shape.length-1),e=t}else o=r;let s=It(o),i=R(wn(o,s),[n,e]),a=au(i),u=Math.floor(e/2)+1,l=Za(a),c=Xl(a),p=mr(l,[u,e-u],l.shape.length-1),m=mr(c,[u,e-u],c.shape.length-1),f=o.shape.slice();return f[o.shape.length-1]=u,R(wn(p[0],m[0]),f)}var lu=T({rfft_:Wj});function Uj(r,t){let e=I(r,"a","squaredDifference"),n=I(t,"b","squaredDifference");[e,n]=Ut(e,n),Pt(e.shape,n.shape);let o={a:e,b:n},s={};return k.runKernel(Rs,o,s)}var ym=T({squaredDifference_:Uj});function Hj(r,t){let e=I(r,"x","squeeze","string_or_numeric");return R(e,t0(e.shape,t).newShape)}var Mn=T({squeeze_:Hj});function qj(r,t=0){let e=ja(r,"tensors","stack","string_or_numeric");E(e.length>=1,()=>"Pass at least one tensor to tf.stack"),e.length>0&&E(t<=e[0].rank,()=>"Axis must be <= rank of the tensor");let n=e,o={axis:t};return k.runKernel(fi,n,o)}var nr=T({stack_:qj});function Kj(r,t=0){let n={x:I(r,"x","step")},o={alpha:t};return k.runKernel(po,n,o)}var bo=T({step_:Kj});function jj(r,t,e,n,o=0,s=0,i=0,a=0,u=0){let c={x:I(r,"x","stridedSlice","string_or_numeric")},p={begin:t,end:e,strides:n,beginMask:o,endMask:s,ellipsisMask:i,newAxisMask:a,shrinkAxisMask:u};return k.runKernel(Wa,c,p)}var ry=T({stridedSlice_:jj});function Xj(r){let e={x:I(r,"x","tan","float32")};return k.runKernel(Os,e)}var ny=T({tan_:Xj});function Me(r,t){Yn(r);let e=Br(r,t);if(e.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");return on(r,null,e,t)}function Vs(r,t,e){if(Yn(r),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");let n=Br(r,e);if(n.length!==2&&n.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return on(r,t,n,e)}function $_(r,t,e){if(Yn(r),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");let n=Br(r,e);if(n.length!==4&&n.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return on(r,t,n,e)}function D_(r,t,e){if(Yn(r),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");let n=Br(r,e);if(n.length!==5&&n.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return on(r,t,n,e)}function R_(r,t,e){if(Yn(r),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");let n=Br(r,e);if(n.length!==6&&n.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||n,on(r,t,n,e)}function Yj(r,t=1,e=!0){let n=I(r,"x","topk");if(n.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");let o=n.shape[n.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>o)throw new Error(`'k' passed to topk() must be <= the last dimension (${o}) but got ${t}`);let s={x:n},i={k:t,sorted:e},[a,u]=k.runKernel(Ua,s,i);return{values:a,indices:u}}var oy=T({topk_:Yj});function Zj(r,t=0,e=1,n,o){if(n!=null&&n==="bool")throw new Error("Unsupported data type $ { dtype }");let s=new Qu(t,e,n,!0,o),i=wt(r,n);for(let a=0;a<i.values.length;a++)i.values[a]=s.nextValue();return i.toTensor()}var bm=T({truncatedNormal_:Zj});function Jj(r,t=0){let e=I(r,"x","unique","string_or_numeric");E(e.rank>0,()=>"The input tensor must be at least 1D");let n={x:e},o={axis:t},[s,i]=k.runKernel(Mp,n,o);return{values:s,indices:i}}var sy=T({unique_:Jj});function Qj(r,t,e){let n=I(r,"x","unsortedSegmentSum"),o=I(t,"segmentIds","unsortedSegmentSum","int32");E(na(e),()=>"numSegments must be of dtype int");let s={x:n,segmentIds:o},i={numSegments:e};return k.runKernel(Wl,s,i)}var wm=T({unsortedSegmentSum_:Qj});function t6(r,t=0){let e=I(r,"x","unstack","string_or_numeric");E(t>=-e.shape.length&&t<e.shape.length,()=>`Axis = ${t} is not in [-${e.shape.length}, ${e.shape.length})`);let n={value:e},o={axis:t};return k.runKernel(bi,n,o)}var vr=T({unstack_:t6});function F_(r,t){return mh(r,t,"right")}function iy(r,t=!0,e,n){return k.makeVariable(r,t,e,n)}function ay(r,t){let e=[];for(let s=0;s<t.length;s++)t[s]&&e.push(s);let n=wt(r,"int32"),o=wt([e.length,r.length],"int32");for(let s=0;s<e.length;s++){let i=n.indexToLoc(e[s]),a=s*r.length;o.values.set(i,a)}return o.toTensor()}async function e6(r){let t=I(r,"condition","whereAsync","bool"),e=await t.data(),n=ay(t.shape,e);return r!==t&&t.dispose(),n}var ly=e6;async function r6(r,t,e){let n=I(r,"tensor","boolMask"),o=I(t,"mask","boolMask","bool"),s=e==null?0:e,i=o.rank,a=n.shape;E(i>0,()=>"mask cannot be scalar"),$e(a.slice(s,s+i),o.shape,"mask's shape must match the first K dimensions of tensor's shape,");let u=1;for(let h=s;h<s+i;h++)u*=a[h];let l=a.slice(0,s).concat([u],a.slice(s+i)),c=R(n,l),p=R(o,[-1]),m=await ly(p),f=Mn(m,[1]),d=Li(c,f,s);return r!==n&&n.dispose(),t!==o&&o.dispose(),f.dispose(),c.dispose(),p.dispose(),m.dispose(),d}var n6=r6;function o6(r,t,e,n,o=!0){let s=I(r,"v","movingAverage"),i=I(t,"x","movingAverage"),a=I(e,"decay","movingAverage");I0(s,i),E(Dn(s.shape,i.shape),()=>"Shape mismatch in v and x");let u=mt(1),l=ct(u,a),c=D(ct(i,s),l);if(o){E(n!=null,()=>"When using zeroDebias: true, step is required.");let p=I(n,"step","movingAverage");c=pt(c,ct(u,an(a,p)))}return X(s,c)}var s6=T({movingAverage_:o6});function i6(r,t,e){let n=I(r,"indices","scatterND","int32"),o=I(t,"updates","scatterND");sx(o,n,e);let s={indices:n,updates:o},i={shape:e};return k.runKernel(La,s,i)}var a6=T({scatterND_:i6});function O_(r,t,e,n){if(r.dtype!=="int32")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${r.dtype}.`);if(r.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${r.shape}.`);let o=r.rank>0?r.shape[0]:1,s=r.rank>1?r.shape[1]:1;if(e.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${e.length}, should be: ${s}.`);let i=t.size;if(!(t.rank===0||t.rank===1&&i===o))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${o}]`);if(t.dtype!==n.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function u6(r,t,e,n=0){let o=I(r,"sparseIndices","sparseToDense","int32"),s=I(t,"sparseValues","sparseToDense","string_or_numeric"),i=I(n,"defaultValue","sparseToDense",s.dtype);O_(o,s,e,i);let a={sparseIndices:o,sparseValues:s,defaultValue:i},u={outputShape:e};return k.runKernel(Lp,a,u)}var c6=T({sparseToDense_:u6});function p6(r,t){let e=I(t,"indices","gatherND","int32"),o={params:I(r,"x","gatherND","string_or_numeric"),indices:e};return k.runKernel(wa,o)}var m6=T({gatherND_:p6});function P_(r,t){if(t==null)return r.shape.slice();if(Dn(r.shape,t))return t;if(r.shape.length===t.length){let e=[];for(let n=0;n<r.shape.length;n++)t[n]==null&&r.shape[n]!=null?e.push(r.shape[n]):e.push(t[n]);return e}return t}function f6(r,t,e,n){let o=I(r,"x","dropout");if(E(o.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${o.dtype} tensor instead.`),E(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return r instanceof Ft?o.clone():o;let s=P_(o,e),i=1-t,a=pt(Pi(X(zi(s,0,1,"float32",n),i)),i);return D(o,a)}var lv=T({dropout_:f6});function uv(r){return Math.floor(Math.pow(2,Math.ceil(Math.log(r)/Math.log(2))))}function hh(r,t,e){let n=1-r%2,o=new Float32Array(r);for(let s=0;s<r;++s){let i=2*Math.PI*s/(r+n-1);o[s]=t-e*Math.cos(i)}return Me(o,"float32")}async function d6(r,t,e=1){let n=I(r,"predictions","inTopK"),o=I(t,"targets","inTopK");E(n.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${n.rank}`),E(n.rank-1===o.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${n.rank} and targets rank ${o.rank}`),$e(n.shape.slice(0,n.shape.length-1),o.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let s=n.shape[n.shape.length-1];E(e>0&&e<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${e}`);let i=await n.data(),a=await o.data(),[u,l]=[i.length/s,s],c=e0("bool",u);for(let p=0;p<u;p++){let m=p*l,f=i.subarray(m,m+l),d=[];for(let h=0;h<f.length;h++)d.push({value:f[h],index:h});d.sort((h,g)=>g.value-h.value),c[p]=0;for(let h=0;h<e;h++)if(d[h].index===a[p]){c[p]=1;break}}return r!==n&&n.dispose(),t!==o&&o.dispose(),ur(c,o.shape,"bool")}var h6=d6;var uu={};Wt(uu,{conv2d:()=>L_,depthwiseConv2d:()=>M_,matMul:()=>z_});function g6(r,t,e,n,o,s="NHWC",i){let a=r;r.rank===3&&(a=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let u=t;u.rank===3&&(u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]])),E(a.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${a.shape}.`),E(u.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${u.shape}.`),E(e.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${e}.`);let l=s==="NHWC"?a.shape[3]:a.shape[1],c=s==="NHWC"?u.shape[3]:u.shape[1];E(l===e[2],()=>`Error in conv2dDerFilter: depth of input ${l}) must match input depth in filter (${e[2]}.`),E(c===e[3],()=>`Error in conv2dDerFilter: depth of dy (${c}) must match output depth for filter (${e[3]}).`),Ie("conv2dDerFilter",o,i);let p={x:a,dy:u},m={strides:n,pad:o,dataFormat:s,dimRoundingMode:i,filterShape:e};return k.runKernel(mp,p,m)}var Cm=T({conv2DBackpropFilter_:g6});function rc(r,t,e){if(e==null||e==="linear")return r;if(e==="relu")return D(r,bo(t));throw new Error(`Cannot compute gradient for fused activation ${e}.`)}function nc(r,t){let e=t,n=ge(r.shape,t.shape);return n.length>0&&(e=ft(e,n)),R(e,r.shape)}function oc(r,t,e,n){if(t==="linear")return r;if(t==="relu")return Fr(r);if(t==="elu")return Oi(r);if(t==="relu6")return lm(r);if(t==="prelu")return ou(r,e);if(t==="leakyrelu")return Ql(r,n);if(t==="sigmoid")return Yr(r);throw new Error(`Unknown fused activation ${t}.`)}var sc=(r,t)=>!(r>0)||t==="linear";function x6({x:r,filter:t,strides:e,pad:n,dataFormat:o="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:a,activation:u="linear",preluActivationWeights:l,leakyreluAlpha:c}){if(u=u||"linear",sc(k.state.gradientDepth,u)===!1){E(o==="NHWC",()=>`Error in fused conv2d: got dataFormat of ${o} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let _=In(r,t,e,n,o,s,i);return a!=null&&(_=X(_,a)),oc(_,u,l,c)}let p=I(r,"x","conv2d","float32"),m=I(t,"filter","conv2d","float32"),f=p,d=!1;p.rank===3&&(d=!0,f=R(p,[1,p.shape[0],p.shape[1],p.shape[2]])),E(f.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${f.rank}.`),E(m.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${m.rank}.`),Ie("fused conv2d",n,i);let h=o==="NHWC"?f.shape[3]:f.shape[1];E(m.shape[2]===h,()=>`Error in conv2d: depth of input (${h}) must match input depth for filter ${m.shape[2]}.`),E(Ar(e,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`);let g=ju(f.shape,m.shape,e,s,n,i),x;a!=null&&(x=I(a,"bias","fused conv2d"),[x]=Ut(x,p),o==="NHWC"?Pt(g.outShape,x.shape):(E(x.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${x.shape.length}.`),E(x.shape.length===0||x.shape[0]===g.outChannels||x.shape[0]===1,()=>`Error in fused conv2d: bias shape (${x.shape}) is not compatible with the number of output channels (${g.outChannels})`)));let b;if(l!=null){let _=l.shape;if(E(_.length<=1||_.length===3,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${_.length}.`),_.length===1)E(_[0]===1||_[0]===g.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${_}) is not compatible with the number of output channels (${g.outChannels}).`);else if(_.length===3)try{Pt(_,g.outShape)}catch(A){let $=`Error in fused conv2d: PReLU activation weights (${_}) is not compatible with the output shape of the conv2d (${g.outShape}).`;throw Error($)}b=I(l,"prelu weights","fused conv2d")}let w=(_,A)=>{E(o==="NHWC",()=>`Error in gradient of fused conv2D: got dataFormat of ${o} but only NHWC is currently supported.`);let[$,F,P,V]=A,G=rc(_,P,u);E(to(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let W=tm(F.shape,G,$,e,n),q=Cm(F,G,$.shape,e,n),H=[W,q];if(V!=null){let j=nc(V,G);H.push(j)}return H},C={x:f,filter:m,bias:x,preluActivationWeights:b},N={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};return a==null?un((A,$,F)=>{let P=k.runKernel(Ii,C,N);return F([$,A,P]),d&&(P=R(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:w}})(f,m):un((A,$,F,P)=>{let V=k.runKernel(Ii,C,N);return P([$,A,V,F]),d&&(V=R(V,[V.shape[1],V.shape[2],V.shape[3]])),{value:V,gradFunc:w}})(f,m,x)}var L_=T({fusedConv2d_:x6});function y6(r,t,e,n,o,s=[1,1],i){let a=r;r.rank===3&&(a=R(r,[1,r.shape[0],r.shape[1],r.shape[2]]));let u=t;u.rank===3&&(u=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let l={x:a,dy:u},c={strides:n,pad:o,dimRoundingMode:i,dilations:s,filterShape:e};return k.runKernel(gp,l,c)}var uy=T({depthwiseConv2dNativeBackpropFilter_:y6});function b6(r,t,e,n,o,s=[1,1],i){let a=t,u=!1;t.rank===3&&(u=!0,a=R(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let l={dy:a,filter:e},c={strides:n,pad:o,dimRoundingMode:i,dilations:s,inputShape:r},p=k.runKernel(xp,l,c);return u?R(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var cy=T({depthwiseConv2dNativeBackpropInput_:b6});function w6({x:r,filter:t,strides:e,pad:n,dataFormat:o="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:a,activation:u="linear",preluActivationWeights:l,leakyreluAlpha:c}){if(sc(k.state.gradientDepth,u)===!1){let N=Fi(r,t,e,n,o,s,i);return a!=null&&(N=X(N,a)),oc(N,u,l,c)}let p=I(r,"x","depthwiseConv2d","float32"),m=I(t,"filter","depthwiseConv2d","float32"),f=p,d=!1;p.rank===3&&(d=!0,f=R(p,[1,p.shape[0],p.shape[1],p.shape[2]])),E(f.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${f.rank}.`),E(m.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${m.rank}.`),E(f.shape[3]===m.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${f.shape[3]}) must match the inChannels dimension in filter ${m.shape[2]}.`),s==null&&(s=[1,1]),E(Ar(e,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${e} and dilations '${s}'`),Ie("fused depthwiseConv2d",n,i);let h=ju(f.shape,m.shape,e,s,n,i,!0),g;a!=null&&(g=I(a,"bias","fused conv2d"),[g]=Ut(g,p),Pt(h.outShape,g.shape));let x;l!=null&&(x=I(l,"prelu weights","fused depthwiseConv2d"));let b=(N,_)=>{E(to(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[A,$,F,P]=_,V=rc(N,F,u),G=cy($.shape,V,A,e,n,s,i),W=uy($,V,A.shape,e,n,s,i);if(P!=null){let q=nc(g,V);return[G,W,q]}return[G,W]},w={x:f,filter:m,bias:g,preluActivationWeights:x},C={strides:e,pad:n,dataFormat:o,dilations:s,dimRoundingMode:i,activation:u,leakyreluAlpha:c};return a==null?un((_,A,$)=>{let F=k.runKernel(Si,w,C);return $([A,_,F]),d&&(F=R(F,[F.shape[1],F.shape[2],F.shape[3]])),{value:F,gradFunc:b}})(f,m):un((_,A,$,F)=>{let P=k.runKernel(Si,w,C);return F([A,_,P,$]),d&&(P=R(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:b}})(f,m,g)}var M_=T({fusedDepthwiseConv2d_:w6});function C6({a:r,b:t,transposeA:e=!1,transposeB:n=!1,bias:o,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:a=.2}){if(sc(k.state.gradientDepth,s)===!1){let V=Lt(r,t,e,n);return o!=null&&(V=X(V,o)),oc(V,s,i,a)}let u=I(r,"a","fused matMul"),l=I(t,"b","fused matMul");[u,l]=Ut(u,l);let c=e?u.shape[u.rank-2]:u.shape[u.rank-1],p=n?l.shape[l.rank-1]:l.shape[l.rank-2],m=e?u.shape[u.rank-1]:u.shape[u.rank-2],f=n?l.shape[l.rank-2]:l.shape[l.rank-1],d=u.shape.slice(0,-2),h=l.shape.slice(0,-2),g=Jt(d),x=Jt(h);E(c===p,()=>`Error in fused matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${u.shape} and ${l.shape} and transposeA=${e} and transposeB=${n} must match.`);let w=Pt(u.shape.slice(0,-2),l.shape.slice(0,-2)).concat([m,f]),C=e?R(u,[g,c,m]):R(u,[g,m,c]),N=n?R(l,[x,f,p]):R(l,[x,p,f]),_;o!=null&&(_=I(o,"bias","fused matMul"),[_]=Ut(_,u),Pt(w,_.shape));let A;i!=null&&(A=I(i,"prelu weights","fused matMul"));let $=(V,G)=>{let[W,q,H,j]=G,Y=rc(R(V,H.shape),H,s),Z,et;if(!e&&!n?(Z=Lt(Y,q,!1,!0),et=Lt(W,Y,!0,!1)):!e&&n?(Z=Lt(Y,q,!1,!1),et=Lt(Y,W,!0,!1)):e&&!n?(Z=Lt(q,Y,!1,!0),et=Lt(W,Y,!1,!1)):(Z=Lt(q,Y,!0,!0),et=Lt(Y,W,!0,!0)),o!=null){let rt=nc(j,Y);return[Z,et,rt]}else return[Z,et]},F={a:C,b:N,bias:_,preluActivationWeights:A},P={transposeA:e,transposeB:n,activation:s,leakyreluAlpha:a};return o==null?un((G,W,q)=>{let H=k.runKernel(Ci,F,P);return q([G,W,H]),{value:R(H,w),gradFunc:$}})(C,N):un((G,W,q,H)=>{let j=k.runKernel(Ci,F,P);return H([G,W,j,q]),{value:R(j,w),gradFunc:$}})(C,N,_)}var z_=T({fusedMatMul_:C6});function I6(r){return hh(r,.54,.46)}var B_=T({hammingWindow_:I6});function S6(r){return hh(r,.5,.5)}var py=T({hannWindow_:S6});function v6(r,t,e,n=!1,o=0){let s=0,i=[];for(;s+t<=r.size;)i.push(Rt(r,s,t)),s+=e;if(n)for(;s<r.size;){let a=s+t-r.size,u=ne([Rt(r,s,t-a),xo([a],o)]);i.push(u),s+=e}return i.length===0?Vs([],[0,t]):R(ne(i),[i.length,t])}var my=T({frame_:v6});function N6(r,t,e,n,o=py){n==null&&(n=uv(t));let s=my(r,t,e),i=D(s,o(t));return lu(i,n)}var V_=T({stft_:N6});function T6(r,t,e,n,o="bilinear",s=0){let i=I(r,"image","cropAndResize"),a=I(t,"boxes","cropAndResize","float32"),u=I(e,"boxInd","cropAndResize","int32"),l=a.shape[0];E(i.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),E(a.rank===2&&a.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${l},4] but had shape ${a.shape}.`),E(u.rank===1&&u.shape[0]===l,()=>`Error in cropAndResize: boxInd must be have size [${l}] but had shape ${a.shape}.`),E(n.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${n.length}.`),E(n[0]>=1&&n[1]>=1,()=>`cropSize must be atleast [1,1], but was ${n}`),E(o==="bilinear"||o==="nearest",()=>`method must be bilinear or nearest, but was ${o}`);let c={image:i,boxes:a,boxInd:u},p={method:o,extrapolationValue:s,cropSize:n};return k.runKernel(da,c,p)}var G_=T({cropAndResize_:T6});function k6(r){let t=I(r,"image","flipLeftRight","float32");E(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let e={image:t};return k.runKernel(ba,e,{})}var W_=T({flipLeftRight_:k6});function E6(r){let t=I(r,"image","grayscaleToRGB"),e=t.rank-1,n=t.shape[e];E(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),E(n===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${n}.`);let o=new Array(t.rank);return o.fill(1,0,e),o[e]=3,Dr(t,o)}var U_=T({grayscaleToRGB_:E6});function _6(r,t,e=0,n=.5){let o=I(r,"image","rotateWithOffset","float32");E(o.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${o.rank}.`);let s={image:o},i={radians:t,fillValue:e,center:n};return k.runKernel(qa,s,i)}var H_=T({rotateWithOffset_:_6});function wo(r,t,e,n,o,s){n==null&&(n=.5),o==null&&(o=Number.NEGATIVE_INFINITY),s==null&&(s=0);let i=r.shape[0];return e=Math.min(e,i),E(0<=n&&n<=1,()=>`iouThreshold must be in [0, 1], but was '${n}'`),E(r.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${r.rank}'`),E(r.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${r.shape[1]}`),E(t.rank===1,()=>"scores must be a 1D tensor"),E(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),E(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:e,iouThreshold:n,scoreThreshold:o,softNmsSigma:s}}function A6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY){let s=I(r,"boxes","nonMaxSuppression","float32"),i=I(t,"scores","nonMaxSuppression","float32"),a=wo(s,i,e,n,o);e=a.maxOutputSize,n=a.iouThreshold,o=a.scoreThreshold;let u={maxOutputSize:e,iouThreshold:n,scoreThreshold:o};return k.runKernel(Ra,{boxes:s,scores:i},u)}var q_=T({nonMaxSuppression_:A6});function K_(r,t,e){let n=$6(r,t,e),o=n<0?-(n+1):n;r.splice(o,0,t)}function $6(r,t,e){return R6(r,t,e||D6)}function D6(r,t){return r>t?1:r<t?-1:0}function R6(r,t,e){let n=0,o=r.length,s=0,i=!1;for(;n<o;){s=n+(o-n>>>1);let a=e(t,r[s]);a>0?n=s+1:(o=s,i=!a)}return i?n:-n-1}function fy(r,t,e,n,o){return cv(r,t,e,n,o,0)}function dy(r,t,e,n,o,s){return cv(r,t,e,n,o,0,!1,s,!0)}function hy(r,t,e,n,o,s){return cv(r,t,e,n,o,s,!0)}function cv(r,t,e,n,o,s,i=!1,a=!1,u=!1){let l=[];for(let g=0;g<t.length;g++)t[g]>o&&l.push({score:t[g],boxIndex:g,suppressBeginIndex:0});l.sort(j_);let c=s>0?-.5/s:0,p=[],m=[];for(;p.length<e&&l.length>0;){let g=l.pop(),{score:x,boxIndex:b,suppressBeginIndex:w}=g;if(x<o)break;let C=!1;for(let N=p.length-1;N>=w;--N){let _=F6(r,b,p[N]);if(_>=n){C=!0;break}if(g.score=g.score*O6(n,c,_),g.score<=o)break}g.suppressBeginIndex=p.length,C||(g.score===x?(p.push(b),m.push(g.score)):g.score>o&&K_(l,g,j_))}let f=p.length,d=e-f;a&&d>0&&(p.push(...new Array(d).fill(0)),m.push(...new Array(d).fill(0)));let h={selectedIndices:p};return i&&(h.selectedScores=m),u&&(h.validOutputs=f),h}function F6(r,t,e){let n=r.subarray(t*4,t*4+4),o=r.subarray(e*4,e*4+4),s=Math.min(n[0],n[2]),i=Math.min(n[1],n[3]),a=Math.max(n[0],n[2]),u=Math.max(n[1],n[3]),l=Math.min(o[0],o[2]),c=Math.min(o[1],o[3]),p=Math.max(o[0],o[2]),m=Math.max(o[1],o[3]),f=(a-s)*(u-i),d=(p-l)*(m-c);if(f<=0||d<=0)return 0;let h=Math.max(s,l),g=Math.max(i,c),x=Math.min(a,p),b=Math.min(u,m),w=Math.max(x-h,0)*Math.max(b-g,0);return w/(f+d-w)}function O6(r,t,e){let n=Math.exp(t*e*e);return e<=r?n:0}function j_(r,t){return r.score-t.score||r.score===t.score&&t.boxIndex-r.boxIndex}async function P6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY){let s=I(r,"boxes","nonMaxSuppressionAsync"),i=I(t,"scores","nonMaxSuppressionAsync"),a=wo(s,i,e,n,o);e=a.maxOutputSize,n=a.iouThreshold,o=a.scoreThreshold;let u=await Promise.all([s.data(),i.data()]),l=u[0],c=u[1],{selectedIndices:p}=fy(l,c,e,n,o);return s!==r&&s.dispose(),i!==t&&i.dispose(),Me(p,"int32")}var X_=P6;function L6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=0){let i=I(r,"boxes","nonMaxSuppression"),a=I(t,"scores","nonMaxSuppression"),u=wo(i,a,e,n,o,s);e=u.maxOutputSize,n=u.iouThreshold,o=u.scoreThreshold,s=u.softNmsSigma;let l={boxes:i,scores:a},c={maxOutputSize:e,iouThreshold:n,scoreThreshold:o,softNmsSigma:s},p=k.runKernel(Oa,l,c);return{selectedIndices:p[0],selectedScores:p[1]}}var Y_=T({nonMaxSuppressionWithScore_:L6});async function M6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=0){let i=I(r,"boxes","nonMaxSuppressionAsync"),a=I(t,"scores","nonMaxSuppressionAsync"),u=wo(i,a,e,n,o,s);e=u.maxOutputSize,n=u.iouThreshold,o=u.scoreThreshold,s=u.softNmsSigma;let l=await Promise.all([i.data(),a.data()]),c=l[0],p=l[1],{selectedIndices:m,selectedScores:f}=hy(c,p,e,n,o,s);return i!==r&&i.dispose(),a!==t&&a.dispose(),{selectedIndices:Me(m,"int32"),selectedScores:Me(f)}}var Z_=M6;function z6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=!1){let i=I(r,"boxes","nonMaxSuppression"),a=I(t,"scores","nonMaxSuppression"),u=wo(i,a,e,n,o,null),l=u.maxOutputSize,c=u.iouThreshold,p=u.scoreThreshold,m={boxes:i,scores:a},f={maxOutputSize:l,iouThreshold:c,scoreThreshold:p,padToMaxOutputSize:s},d=k.runKernel(Fa,m,f);return{selectedIndices:d[0],validOutputs:d[1]}}var J_=T({nonMaxSuppressionPadded_:z6});async function B6(r,t,e,n=.5,o=Number.NEGATIVE_INFINITY,s=!1){let i=I(r,"boxes","nonMaxSuppressionAsync"),a=I(t,"scores","nonMaxSuppressionAsync"),u=wo(i,a,e,n,o,null),l=u.maxOutputSize,c=u.iouThreshold,p=u.scoreThreshold,[m,f]=await Promise.all([i.data(),a.data()]),{selectedIndices:d,validOutputs:h}=dy(m,f,l,c,p,s);return i!==r&&i.dispose(),a!==t&&a.dispose(),{selectedIndices:Me(d,"int32"),validOutputs:mt(h,"int32")}}var Q_=B6;function V6(r,t,e=!1,n=!1){let o=I(r,"images","resizeBilinear");E(o.rank===3||o.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${o.rank}.`),E(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),E(n===!1||e===!1,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=o,i=!1;o.rank===3&&(i=!0,s=R(o,[1,o.shape[0],o.shape[1],o.shape[2]]));let[]=t,a={images:s},u={alignCorners:e,halfPixelCenters:n,size:t},l=k.runKernel(Ss,a,u);return i?R(l,[l.shape[1],l.shape[2],l.shape[3]]):l}var gy=T({resizeBilinear_:V6});function G6(r,t,e=!1,n=!1){let o=I(r,"images","resizeNearestNeighbor");E(o.rank===3||o.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${o.rank}.`),E(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),E(o.dtype==="float32"||o.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype"),E(n===!1||e===!1,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=o,i=!1;o.rank===3&&(i=!0,s=R(o,[1,o.shape[0],o.shape[1],o.shape[2]]));let[]=t,a={images:s},u={alignCorners:e,halfPixelCenters:n,size:t},l=k.runKernel(Is,a,u);return i?R(l,[l.shape[1],l.shape[2],l.shape[3]]):l}var xy=T({resizeNearestNeighbor_:G6});function W6(r,t="binary",e=!1,n=.5){let o=I(r,"image","threshold"),s=.2989,i=.587,a=.114,u=o.shape[0]*o.shape[1],l=D(Me([n]),255),c,p,m,f;if(E(o.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${o.rank}.`),E(o.shape[2]===3||o.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${o.shape[2]}.`),E(o.dtype==="int32"||o.dtype==="float32",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${o.dtype}.`),E(t==="otsu"||t==="binary",()=>`Method must be binary or otsu, but was ${t}`),o.shape[2]===3){[c,p,m]=mr(o,[1,1,1],-1);let g=D(c,s),x=D(p,i),b=D(m,a);f=X(X(g,x),b)}else f=r;if(t==="otsu"){let g=wx(J(um(f),"int32"),ur([]),256);l=U6(g,u)}let d=e?Ln(f,l):Re(f,l);return J(D(d,255),"int32")}function U6(r,t){let e=Me([-1]),n=Me([0]),o=Me([0]),s,i,a,u,l,c;for(let p=0;p<r.size-1;p++){s=Rt(r,0,p+1),i=Rt(r,p+1),l=pt(ft(s),t),c=pt(ft(i),t);let m=ft(D(s,su(0,s.size)));a=pt(m,ft(s));let f=xo(i.shape,s.size),d=X(su(0,i.size),f),h=D(i,d);u=pt(ft(h),ft(i));let g=ct(a,u),x=ct(a,u),b=D(l,c);o=D(D(b,g),x);let w=Re(o,n);n=_e(w,o,n),e=_e(w,Me([p]),e)}return e}var tA=T({threshold_:W6});function H6(r,t,e="nearest",n="constant",o=0,s){let i=I(r,"image","transform","float32"),a=I(t,"transforms","transform","float32");E(i.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),E(a.rank===2&&(a.shape[0]===i.shape[0]||a.shape[0]===1)&&a.shape[1]===8,()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),E(s==null||s.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let u={image:i,transforms:a},l={interpolation:e,fillMode:n,fillValue:o,outputShape:s};return k.runKernel(Ha,u,l)}var eA=T({transform_:H6});function q6(r,t,e){E(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),E(e%1===0,()=>`bandPart(): numUpper must be an integer, got ${e}.`);let n=I(r,"a","bandPart");E(n.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${n.rank}.`);let o=n.shape,[s,i]=n.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(e<=i))throw new Error(`bandPart(): numUpper (${e}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),e<0&&(e=i);let a=R(su(0,s,1,"int32"),[-1,1]),u=su(0,i,1,"int32"),l=ct(a,u),c=Rr(Ln(l,mt(+t,"int32")),ln(l,mt(-e,"int32"))),p=Ne([s,i],n.dtype);return R(nr(vr(R(n,[-1,s,i])).map(m=>_e(c,m,p))),o)}var rA=T({bandPart_:q6});function K6(r){let t;if(Array.isArray(r)){t=!1,E(r!=null&&r.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let o=r[0].shape[0];for(let s=1;s<r.length;++s)E(r[s].shape[0]===o,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${r[s].shape[0]} vs. ${o})`)}else t=!0,r=mr(r,r.shape[0],0).map(o=>Mn(o,[0]));E(r.length<=r[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${r.length}) exceeds number of dimensions (${r[0].shape[0]}).`);let e=[],n=r;for(let o=0;o<r.length;++o)e.push(k.tidy(()=>{let s=n[o];if(o>0)for(let i=0;i<o;++i){let a=D(ft(D(e[i],s)),e[i]);s=ct(s,a)}return pt(s,Qa(s,"euclidean"))}));return t?nr(e,0):e}var nA=T({gramSchmidt_:K6});function j6(r,t=!1){if(E(r.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${r.rank}`),r.rank===2)return oA(r,t);{let e=r.shape.slice(0,r.shape.length-2).reduce((u,l)=>u*l),n=vr(R(r,[e,r.shape[r.shape.length-2],r.shape[r.shape.length-1]]),0),o=[],s=[];n.forEach(u=>{let[l,c]=oA(u,t);o.push(l),s.push(c)});let i=R(nr(o,0),r.shape),a=R(nr(s,0),r.shape);return[i,a]}}function oA(r,t=!1){return k.tidy(()=>{E(r.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${r.shape.length}D Tensor.`);let e=r.shape[0],n=r.shape[1],o=Yu(e),s=sn(r),i=Vs([[1]],[1,1]),a=sn(i),u=e>=n?n:e;for(let l=0;l<u;++l){let c=s,p=a,m=o;[a,s,o]=k.tidy(()=>{let f=Rt(s,[l,l],[e-l,1]),d=Qa(f),h=Rt(s,[l,l],[1,1]),g=_e(Re(h,0),Vs([[-1]]),Vs([[1]])),x=ct(h,D(g,d)),b=pt(f,x);b.shape[0]===1?a=sn(i):a=ne([i,Rt(b,[1,0],[b.shape[0]-1,b.shape[1]])],0);let w=Ht(pt(Lt(g,x),d)),C=Rt(s,[l,0],[e-l,n]),N=D(w,a),_=Ot(a);if(l===0)s=ct(C,Lt(N,Lt(_,C)));else{let F=ct(C,Lt(N,Lt(_,C)));s=ne([Rt(s,[0,0],[l,n]),F],0)}let A=Ot(N),$=Rt(o,[0,l],[e,o.shape[1]-l]);if(l===0)o=ct($,Lt(Lt($,a),A));else{let F=ct($,Lt(Lt($,a),A));o=ne([Rt(o,[0,0],[e,l]),F],1)}return[a,s,o]}),vt([c,p,m])}return!t&&e>n&&(o=Rt(o,[0,0],[e,n]),s=Rt(s,[0,0],[n,n])),[o,s]})}var sA=T({qr_:j6});var Xe;(function(r){r[r.NONE=0]="NONE",r[r.MEAN=1]="MEAN",r[r.SUM=2]="SUM",r[r.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS"})(Xe||(Xe={}));function X6(r,t,e=Xe.SUM_BY_NONZERO_WEIGHTS){let n=I(r,"losses","computeWeightedLoss"),o=null;t!=null&&(o=I(t,"weights","computeWeightedLoss"));let s=o==null?n:D(n,o);if(e===Xe.NONE)return s;if(e===Xe.SUM)return ft(s);if(e===Xe.MEAN){if(o==null)return ve(s);{let i=n.size/o.size,a=pt(ft(s),ft(o));return i>1?pt(a,mt(i)):a}}if(e===Xe.SUM_BY_NONZERO_WEIGHTS){if(o==null)return pt(ft(s),mt(n.size));{let i=D(o,cr(n.shape)),a=J(ft(Bs(i,mt(0))),"float32");return pt(ft(s),a)}}throw Error(`Unknown reduction: ${e}`)}var Gr=T({computeWeightedLoss_:X6});function Y6(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,"labels","absoluteDifference"),s=I(t,"predictions","absoluteDifference"),i=null;e!=null&&(i=I(e,"weights","absoluteDifference")),$e(o.shape,s.shape,"Error in absoluteDifference: ");let a=Ee(ct(o,s));return Gr(a,i,n)}var iA=T({absoluteDifference_:Y6});function Z6(r,t,e,n,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,"labels","cosineDistance"),i=I(t,"predictions","cosineDistance"),a=null;n!=null&&(a=I(n,"weights","cosineDistance")),$e(s.shape,i.shape,"Error in cosineDistance: ");let u=mt(1),l=ct(u,ft(D(s,i),e,!0));return Gr(l,a,o)}var aA=T({cosineDistance_:Z6});function J6(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,"labels","hingeLoss"),s=I(t,"predictions","hingeLoss"),i=null;e!=null&&(i=I(e,"weights","hingeLoss")),$e(o.shape,s.shape,"Error in hingeLoss: ");let a=mt(1);o=ct(D(mt(2),o),a);let u=Fr(ct(a,D(o,s)));return Gr(u,i,n)}var lA=T({hingeLoss_:J6});function Q6(r,t,e,n=1,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,"labels","huberLoss"),i=I(t,"predictions","huberLoss"),a=null;e!=null&&(a=I(e,"weights","huberLoss")),$e(s.shape,i.shape,"Error in huberLoss: ");let u=mt(n),l=Ee(ct(i,s)),c=Mi(l,u),p=ct(l,c),m=X(D(mt(.5),Mt(c)),D(u,p));return Gr(m,a,o)}var uA=T({huberLoss_:Q6});function tX(r,t,e,n=1e-7,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,"labels","logLoss"),i=I(t,"predictions","logLoss"),a=null;e!=null&&(a=I(e,"weights","logLoss")),$e(s.shape,i.shape,"Error in logLoss: ");let u=mt(1),l=mt(n),c=Ht(D(s,Sr(X(i,l)))),p=D(ct(u,s),Sr(X(ct(u,i),l))),m=ct(c,p);return Gr(m,a,o)}var cA=T({logLoss_:tX});function eX(r,t,e,n=Xe.SUM_BY_NONZERO_WEIGHTS){let o=I(r,"labels","meanSquaredError"),s=I(t,"predictions","meanSquaredError"),i=null;e!=null&&(i=I(e,"weights","meanSquaredError")),$e(o.shape,s.shape,"Error in meanSquaredError: ");let a=ym(o,s);return Gr(a,i,n)}var pA=T({meanSquaredError_:eX});function rX(r,t){let e=I(r,"labels","sigmoidCrossEntropyWithLogits"),n=I(t,"logits","sigmoidCrossEntropyWithLogits");$e(e.shape,n.shape,"Error in sigmoidCrossEntropyWithLogits: ");let o=Fr(n),s=D(n,e),i=tu(er(Ht(Ee(n))));return X(ct(o,s),i)}function nX(r,t,e,n=0,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,"multiClassLabels","sigmoidCrossEntropy"),i=I(t,"logits","sigmoidCrossEntropy"),a=null;if(e!=null&&(a=I(e,"weights","sigmoidCrossEntropy")),$e(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),n>0){let l=mt(n),c=mt(1),p=mt(.5);s=X(D(s,ct(c,l)),D(p,l))}let u=rX(s,i);return Gr(u,a,o)}var mA=T({sigmoidCrossEntropy_:nX});function oX(r,t,e=-1){if(e===-1&&(e=t.rank-1),e!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${e}`);return un((o,s,i)=>{let u=im(s,[e],!0),l=ct(J(s,"float32"),u);i([o,l]);let c=Ht(D(l,o));return{value:ft(c,[e]),gradFunc:(f,d)=>{let[h,g]=d,x=yo(f.shape,[e]);return[D(R(f,x),ct(J(h,"float32"),er(g))),D(R(f,x),ct(er(g),J(h,"float32")))]}}})(r,t)}function sX(r,t,e,n=0,o=Xe.SUM_BY_NONZERO_WEIGHTS){let s=I(r,"onehotLabels","softmaxCrossEntropy"),i=I(t,"logits","softmaxCrossEntropy"),a=null;if(e!=null&&(a=I(e,"weights","softmaxCrossEntropy")),$e(s.shape,i.shape,"Error in softmaxCrossEntropy: "),n>0){let l=mt(n),c=mt(1),p=mt(s.shape[1]);s=X(D(s,ct(c,l)),pt(l,p))}let u=oX(s,i);return Gr(u,a,o)}var fA=T({softmaxCrossEntropy_:sX});function iX(r,t,e,n){let o=I(r,"indices","sparseFillEmptyRows","int32"),s=I(t,"values","sparseFillEmptyRows"),i=I(e,"denseShape","sparseFillEmptyRows","int32"),a=I(n,"defaultValue","sparseFillEmptyRows",s.dtype);if(o.rank!==2)throw new Error(`Indices should be Tensor2D but received shape
-        ${o.shape}`);if(s.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(i.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(a.rank!==0)throw new Error(`Default value should be a scalar but received shape ${a.shape}`);let u={indices:o,values:s,denseShape:i,defaultValue:a},l=k.runKernel(Pl,u);return{outputIndices:l[0],outputValues:l[1],emptyRowIndicator:l[2],reverseIndexMap:l[3]}}var dA=T({sparseFillEmptyRows_:iX});function aX(r,t,e){let n=I(r,"inputIndices","sparseReshape","int32"),o=I(t,"inputShape","sparseReshape","int32"),s=I(e,"newShape","sparseReshape","int32");if(n.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape
-        ${n.shape}`);if(o.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${o.shape}`);if(s.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:n,inputShape:o,newShape:s},a=k.runKernel(Ga,i);return{outputIndices:a[0],outputShape:a[1]}}var hA=T({sparseReshape_:aX});function lX(r,t,e){let n=I(r,"data","sparseSegmentMean"),o=I(t,"indices","sparseSegmentMean","int32"),s=I(e,"segmentIds","sparseSegmentMean","int32");if(n.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-          ${o.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-          ${s.shape}`);let i={data:n,indices:o,segmentIds:s};return k.runKernel(Ll,i)}var gA=T({sparseSegmentMean_:lX});function uX(r,t,e){let n=I(r,"data","sparseSegmentSum"),o=I(t,"indices","sparseSegmentSum","int32"),s=I(e,"segmentIds","sparseSegmentSum","int32");if(n.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
-         ${o.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
-         ${s.shape}`);let i={data:n,indices:o,segmentIds:s};return k.runKernel(Ml,i)}var xA=T({sparseSegmentSum_:uX});function cX(r,t,e,n,o,s,i,a){let u=I(r,"data","stringNGrams","string");if(u.dtype!=="string")throw new Error("Data must be of datatype string");if(u.shape.length!==1)throw new Error(`Data must be a vector, saw: ${u.shape}`);let l=I(t,"dataSplits","stringNGrams");if(l.dtype!=="int32")throw new Error("Data splits must be of datatype int32");let c={separator:e,nGramWidths:n,leftPad:o,rightPad:s,padWidth:i,preserveShortSequences:a},p={data:u,dataSplits:l},m=k.runKernel(Bl,p,c);return{nGrams:m[0],nGramsSplits:m[1]}}var yA=T({stringNGrams_:cX});function pX(r,t,e=!0){let n=I(r,"input","stringSplit","string"),o=I(t,"delimiter","stringSplit","string");if(n.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${n.shape}`);if(o.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${o.shape}`);let s={skipEmpty:e},i={input:n,delimiter:o},a=k.runKernel(Vl,i,s);return{indices:a[0],values:a[1],shape:a[2]}}var bA=T({stringSplit_:pX});function mX(r,t){let e=I(r,"input","stringToHashBucketFast","string"),n={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let o={input:e};return k.runKernel(Gl,o,n)}var wA=T({stringToHashBucketFast_:mX});var fX={fft:au,ifft:tl,rfft:lu,irfft:xm},dX={hammingWindow:B_,hannWindow:py,frame:my,stft:V_},Gs={flipLeftRight:W_,grayscaleToRGB:U_,resizeNearestNeighbor:xy,resizeBilinear:gy,rotateWithOffset:H_,cropAndResize:G_,nonMaxSuppression:q_,nonMaxSuppressionAsync:X_,nonMaxSuppressionWithScore:Y_,nonMaxSuppressionWithScoreAsync:Z_,nonMaxSuppressionPadded:J_,nonMaxSuppressionPaddedAsync:Q_,threshold:tA,transform:eA},pv={bandPart:rA,gramSchmidt:nA,qr:sA},hX={absoluteDifference:iA,computeWeightedLoss:Gr,cosineDistance:aA,hingeLoss:lA,huberLoss:uA,logLoss:cA,meanSquaredError:pA,sigmoidCrossEntropy:mA,softmaxCrossEntropy:fA},gX={sparseFillEmptyRows:dA,sparseReshape:hA,sparseSegmentMean:gA,sparseSegmentSum:xA},xX={stringNGrams:yA,stringSplit:bA,stringToHashBucketFast:wA};var Wr=class extends uh{minimize(t,e=!1,n){let{value:o,grads:s}=this.computeGradients(t,n);if(n!=null){let i=n.map(a=>({name:a.name,tensor:s[a.name]}));this.applyGradients(i)}else this.applyGradients(s);return vt(s),e?o:(o.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(t,e){return Bx(t,e)}dispose(){this.iterations_!=null&&vt(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:mt(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(t){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(t){return this.iterations_=(await t[0].tensor.data())[0],t.slice(1)}};Object.defineProperty(Wr,Symbol.hasInstance,{value:r=>r.minimize!=null&&r.computeGradients!=null&&r.applyGradients!=null});var cu=class extends Wr{constructor(t,e,n=null){super(),this.learningRate=t,this.rho=e,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n],i=!1;this.accumulatedGrads[o]==null&&(this.accumulatedGrads[o]={originalName:`${n}/accum_grad`,variable:B(()=>It(s).variable(i))}),this.accumulatedUpdates[o]==null&&(this.accumulatedUpdates[o]={originalName:`${n}/accum_var`,variable:B(()=>It(s).variable(i))});let a=Array.isArray(t)?t[o].tensor:t[n];if(a==null)return;let u=this.accumulatedGrads[o].variable,l=this.accumulatedUpdates[o].variable;B(()=>{let c=X(D(u,this.rho),D(Mt(a),1-this.rho)),p=D(pt(Se(X(l,this.epsilon)),Se(X(u,this.epsilon))),a),m=X(D(l,this.rho),D(Mt(p),1-this.rho));u.assign(c),l.assign(m);let f=X(D(p,-this.learningRate),s);s.assign(f)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(vt(this.accumulatedGrads.map(t=>t.variable)),vt(this.accumulatedUpdates.map(t=>t.variable)))}async getWeights(){let t=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=t.length/2,n=!1;this.accumulatedGrads=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedUpdates=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(t,e){return new t(e.learningRate,e.rho,e.epsilon)}};cu.className="Adadelta";Cn(cu);var pu=class extends Wr{constructor(t,e=.1){super(),this.learningRate=t,this.initialAccumulatorValue=e,this.accumulatedGrads=[]}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n];this.accumulatedGrads[o]==null&&(this.accumulatedGrads[o]={originalName:`${n}/accumulator`,variable:B(()=>xo(s.shape,this.initialAccumulatorValue).variable(!1))});let i=Array.isArray(t)?t[o].tensor:t[n];if(i==null)return;let a=this.accumulatedGrads[o].variable;B(()=>{let u=X(a,Mt(i));a.assign(u);let l=X(D(pt(i,Se(X(u,k.backend.epsilon()))),-this.learningRate),s);s.assign(l)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&vt(this.accumulatedGrads.map(t=>t.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=!1;this.accumulatedGrads=t.map(n=>({originalName:n.name,variable:n.tensor.variable(e)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(t,e){return new t(e.learningRate,e.initialAccumulatorValue)}};pu.className="Adagrad";Cn(pu);var mu=class extends Wr{constructor(t,e,n,o=null){super(),this.learningRate=t,this.beta1=e,this.beta2=n,this.epsilon=o,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],B(()=>{this.accBeta1=mt(e).variable(),this.accBeta2=mt(n).variable()}),o==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){let e=Array.isArray(t)?t.map(n=>n.name):Object.keys(t);B(()=>{let n=ct(1,this.accBeta1),o=ct(1,this.accBeta2);e.forEach((s,i)=>{let a=k.registeredVariables[s],u=!1;this.accumulatedFirstMoment[i]==null&&(this.accumulatedFirstMoment[i]={originalName:`${s}/m`,variable:B(()=>It(a).variable(u))}),this.accumulatedSecondMoment[i]==null&&(this.accumulatedSecondMoment[i]={originalName:`${s}/v`,variable:B(()=>It(a).variable(u))});let l=Array.isArray(t)?t[i].tensor:t[s];if(l==null)return;let c=this.accumulatedFirstMoment[i].variable,p=this.accumulatedSecondMoment[i].variable,m=X(D(c,this.beta1),D(l,1-this.beta1)),f=X(D(p,this.beta2),D(Mt(l),1-this.beta2)),d=pt(m,n),h=pt(f,o);c.assign(m),p.assign(f);let g=X(D(pt(d,X(Se(h),this.epsilon)),-this.learningRate),a);a.assign(g)}),this.accBeta1.assign(D(this.accBeta1,this.beta1)),this.accBeta2.assign(D(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&vt(this.accumulatedFirstMoment.map(t=>t.variable)),this.accumulatedSecondMoment!=null&&vt(this.accumulatedSecondMoment.map(t=>t.variable))}async getWeights(){let t=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t),B(()=>{this.accBeta1.assign(an(this.beta1,this.iterations_+1)),this.accBeta2.assign(an(this.beta2,this.iterations_+1))});let e=t.length/2,n=!1;this.accumulatedFirstMoment=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedSecondMoment=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(t,e){return new t(e.learningRate,e.beta1,e.beta2,e.epsilon)}};mu.className="Adam";Cn(mu);var fu=class extends Wr{constructor(t,e,n,o=null,s=0){super(),this.learningRate=t,this.beta1=e,this.beta2=n,this.epsilon=o,this.decay=s,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],B(()=>{this.iteration=mt(0).variable(),this.accBeta1=mt(e).variable()}),o==null&&(this.epsilon=k.backend.epsilon())}applyGradients(t){let e=Array.isArray(t)?t.map(n=>n.name):Object.keys(t);B(()=>{let n=ct(1,this.accBeta1),o=pt(-this.learningRate,X(D(this.iteration,this.decay),1));e.forEach((s,i)=>{let a=k.registeredVariables[s],u=!1;this.accumulatedFirstMoment[i]==null&&(this.accumulatedFirstMoment[i]={originalName:`${s}/m`,variable:It(a).variable(u)}),this.accumulatedWeightedInfNorm[i]==null&&(this.accumulatedWeightedInfNorm[i]={originalName:`${s}/v`,variable:It(a).variable(u)});let l=Array.isArray(t)?t[i].tensor:t[s];if(l==null)return;let c=this.accumulatedFirstMoment[i].variable,p=this.accumulatedWeightedInfNorm[i].variable,m=X(D(c,this.beta1),D(l,1-this.beta1)),f=D(p,this.beta2),d=Ee(l),h=Sn(f,d);c.assign(m),p.assign(h);let g=X(D(pt(o,n),pt(m,X(h,this.epsilon))),a);a.assign(g)}),this.iteration.assign(X(this.iteration,1)),this.accBeta1.assign(D(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&vt(this.accumulatedFirstMoment.map(t=>t.variable)),this.accumulatedWeightedInfNorm!=null&&vt(this.accumulatedWeightedInfNorm.map(t=>t.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(t){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(t,e){return new t(e.learningRate,e.beta1,e.beta2,e.epsilon,e.decay)}};fu.className="Adamax";Cn(fu);var Bi=class extends Wr{constructor(t){super(),this.learningRate=t,this.setLearningRate(t)}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=Array.isArray(t)?t[o].tensor:t[n];if(s==null)return;let i=k.registeredVariables[n];B(()=>{let a=X(D(this.c,s),i);i.assign(a)})}),this.incrementIterations()}setLearningRate(t){this.learningRate=t,this.c!=null&&this.c.dispose(),this.c=De(mt(-t))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(t){if(t=await this.extractIterations(t),t.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(t,e){return new t(e.learningRate)}};Bi.className="SGD";Cn(Bi);var du=class extends Bi{constructor(t,e,n=!1){super(t),this.learningRate=t,this.momentum=e,this.useNesterov=n,this.accumulations=[],this.m=mt(this.momentum)}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n];this.accumulations[o]==null&&(this.accumulations[o]={originalName:`${n}/momentum`,variable:B(()=>It(s).variable(!1))});let i=this.accumulations[o].variable,a=Array.isArray(t)?t[o].tensor:t[n];a!=null&&B(()=>{let u,l=X(D(this.m,i),a);this.useNesterov?u=X(D(this.c,X(a,D(l,this.m))),s):u=X(D(this.c,l),s),i.assign(l),s.assign(u)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&vt(this.accumulations.map(t=>t.variable))}setMomentum(t){this.momentum=t}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=!1;this.accumulations=t.map(n=>({originalName:n.name,variable:n.tensor.variable(e)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(t,e){return new t(e.learningRate,e.momentum,e.useNesterov)}};du.className="Momentum";Cn(du);var hu=class extends Wr{constructor(t,e=.9,n=0,o=null,s=!1){if(super(),this.learningRate=t,this.decay=e,this.momentum=n,this.epsilon=o,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=s,o==null&&(this.epsilon=k.backend.epsilon()),t==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(t){(Array.isArray(t)?t.map(n=>n.name):Object.keys(t)).forEach((n,o)=>{let s=k.registeredVariables[n],i=!1;this.accumulatedMeanSquares[o]==null&&(this.accumulatedMeanSquares[o]={originalName:`${n}/rms`,variable:B(()=>It(s).variable(i))}),this.accumulatedMoments[o]==null&&(this.accumulatedMoments[o]={originalName:`${n}/momentum`,variable:B(()=>It(s).variable(i))}),this.accumulatedMeanGrads[o]==null&&this.centered&&(this.accumulatedMeanGrads[o]={originalName:`${n}/mg`,variable:B(()=>It(s).variable(i))});let a=Array.isArray(t)?t[o].tensor:t[n];if(a==null)return;let u=this.accumulatedMeanSquares[o].variable,l=this.accumulatedMoments[o].variable;B(()=>{let c=X(D(u,this.decay),D(Mt(a),1-this.decay));if(this.centered){let p=this.accumulatedMeanGrads[o].variable,m=X(D(p,this.decay),D(a,1-this.decay)),f=pt(D(a,this.learningRate),Se(ct(c,X(Mt(m),this.epsilon)))),d=X(D(l,this.momentum),f);u.assign(c),p.assign(m),l.assign(d);let h=ct(s,d);s.assign(h)}else{let p=X(D(u,this.decay),D(Mt(a),1-this.decay)),m=X(D(l,this.momentum),pt(D(a,this.learningRate),Se(X(p,this.epsilon))));u.assign(p),l.assign(m);let f=ct(s,m);s.assign(f)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&vt(this.accumulatedMeanSquares.map(t=>t.variable)),this.accumulatedMeanGrads!=null&&this.centered&&vt(this.accumulatedMeanGrads.map(t=>t.variable)),this.accumulatedMoments!=null&&vt(this.accumulatedMoments.map(t=>t.variable))}async getWeights(){let t=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&t.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(t.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(t){t=await this.extractIterations(t);let e=this.centered?t.length/3:t.length/2,n=!1;this.accumulatedMeanSquares=t.slice(0,e).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.accumulatedMoments=t.slice(e,e*2).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=t.slice(e*2,e*3).map(o=>({originalName:o.name,variable:o.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(t,e){return new t(e.learningRate,e.decay,e.momentum,e.epsilon,e.centered)}};hu.className="RMSProp";Cn(hu);var Ws=class{static sgd(t){return new Bi(t)}static momentum(t,e,n=!1){return new du(t,e,n)}static rmsprop(t,e=.9,n=0,o=null,s=!1){return new hu(t,e,n,o,s)}static adam(t=.001,e=.9,n=.999,o=null){return new mu(t,e,n,o)}static adadelta(t=.001,e=.95,n=null){return new cu(t,e,n)}static adamax(t=.002,e=.9,n=.999,o=null,s=0){return new fu(t,e,n,o,s)}static adagrad(t,e=.1){return new pu(t,e)}};var ic={sgd:Ws.sgd,momentum:Ws.momentum,adadelta:Ws.adadelta,adagrad:Ws.adagrad,rmsprop:Ws.rmsprop,adamax:Ws.adamax,adam:Ws.adam};var yX=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:r=>r())();function gh(){return new Promise(r=>yX(()=>r()))}var v={};Wt(v,{ERF_A1:()=>RX,ERF_A2:()=>FX,ERF_A3:()=>OX,ERF_A4:()=>PX,ERF_A5:()=>LX,ERF_P:()=>DX,PARALLELIZE_THRESHOLD:()=>yy,RowPartitionType:()=>Vi,SELU_SCALE:()=>fv,SELU_SCALEALPHA:()=>mv,applyActivation:()=>oc,assertAndGetBroadcastShape:()=>Pt,assertAxesAreInnerMostDims:()=>Hq,assertParamsConsistent:()=>bX,assignToTypedArray:()=>WX,axesAreInnerMostDims:()=>J0,calculateShapes:()=>IE,checkEinsumDimSizes:()=>XX,checkPadOnDimRoundingMode:()=>Ie,combineLocations:()=>HE,combineRaggedTensorToTensorShapes:()=>CX,complexWithEvenIndex:()=>BX,complexWithOddIndex:()=>VX,computeConv2DInfo:()=>ju,computeConv3DInfo:()=>ME,computeDefaultPad:()=>Z0,computeDilation2DInfo:()=>KH,computeOptimalWindowSize:()=>NX,computeOutAndReduceShapes:()=>Q0,computeOutShape:()=>wX,computePool2DInfo:()=>Y0,computePool3DInfo:()=>jH,convertConv2DDataFormat:()=>zE,decodeEinsumEquation:()=>KX,eitherStridesOrDilationsAreOne:()=>Ar,expandShapeToKeepDim:()=>yo,exponent:()=>HX,exponents:()=>UX,fromStringArrayToUint8:()=>g5,fromUint8ToStringArray:()=>h5,getAxesPermutation:()=>tv,getBroadcastDims:()=>bE,getComplexWithIndex:()=>GX,getEinsumComputePath:()=>YX,getEinsumPermutation:()=>jX,getFusedBiasGradient:()=>nc,getFusedDyActivation:()=>rc,getImageCenter:()=>TX,getInnerMostAxes:()=>qq,getPermuted:()=>EX,getRaggedRank:()=>SX,getReductionAxes:()=>ge,getReshaped:()=>kX,getReshapedPermuted:()=>_X,getRowPartitionTypesHelper:()=>IX,getSliceBeginCoords:()=>AX,getSliceSize:()=>$X,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>t5,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>e5,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>r5,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>s5,getSparseReshapeInputOutputMismatchErrorMessage:()=>a5,getSparseReshapeInputOutputMultipleErrorMessage:()=>i5,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>n5,getSparseReshapeNegativeOutputDimErrorMessage:()=>o5,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>p5,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>l5,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>u5,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>c5,getUndoAxesPermutation:()=>ph,isIdentityPermutation:()=>ZX,log:()=>o4,mergeRealAndImagArrays:()=>MX,prepareAndValidate:()=>CE,prepareSplitSize:()=>QX,segment_util:()=>hv,shouldFuse:()=>sc,slice_util:()=>Le,splitRealAndImagArrays:()=>zX,tupleValuesAreOne:()=>to,upcastType:()=>sr,validateDefaultValueShape:()=>vX,validateInput:()=>sx,validateUpdateShape:()=>U0,warn:()=>vi});function bX(r,t){let e=r[0].length;r.forEach((o,s)=>{E(o.length===e,()=>`Error in concat${e}D: rank of tensors[${s}] must be the same as the rank of the rest (${e})`)}),E(t>=0&&t<e,()=>`Error in concat${e}D: axis must be between 0 and ${e-1}.`);let n=r[0];r.forEach((o,s)=>{for(let i=0;i<e;i++)E(i===t||o[i]===n[i],()=>`Error in concat${e}D: Shape of tensors[${s}] (${o}) does not match the shape of the rest (${n}) along the non-concatenated axis ${s}.`)})}function wX(r,t){let e=r[0].slice();for(let n=1;n<r.length;n++)e[t]+=r[n][t];return e}var Vi;(function(r){r[r.FIRST_DIM_SIZE=0]="FIRST_DIM_SIZE",r[r.VALUE_ROWIDS=1]="VALUE_ROWIDS",r[r.ROW_LENGTHS=2]="ROW_LENGTHS",r[r.ROW_SPLITS=3]="ROW_SPLITS",r[r.ROW_LIMITS=4]="ROW_LIMITS",r[r.ROW_STARTS=5]="ROW_STARTS"})(Vi||(Vi={}));function CX(r,t,e){let n=new Array;if(e==null&&t==null)return n;if(t==null)for(;n.length<r+e.length;)n.push(-1);else n=t.slice();if(e==null)return n;if(r+e.length!==n.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${r+e.length}, but shape.rank = ${n.length}`);for(let o=1;o<e.length;++o){let s=e[o],i=n[n.length-e.length+o],a=n[i];if(s>=0)if(a>=0){if(a!==s)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${o+r}] = ${s} but shape[${o+r}] = ${a}`)}else n[i]=s}return n}function IX(r){let t={FIRST_DIM_SIZE:Vi.FIRST_DIM_SIZE,VALUE_ROWIDS:Vi.VALUE_ROWIDS,ROW_LENGTHS:Vi.ROW_LENGTHS,ROW_SPLITS:Vi.ROW_SPLITS,ROW_LIMITS:Vi.ROW_LIMITS,ROW_STARTS:Vi.ROW_STARTS},e=[];for(let n of r)if(n in t)e.push(t[n]);else break;return e}function SX(r){return r.length===0?0:r[0]===Vi.FIRST_DIM_SIZE?r.length-1:r.length}function vX(r,t){if(r==null||t==null)return;let e=r.length,n=t.length;if(e>=n)throw new Error(`defaultValue.shape=${r} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${e} must be less than ragged tensor input flatValues.rank = ${n})`);for(let o=0;o<Math.min(e,n-1);++o){let s=r[o],i=t[o+1];if(s>=0&&i>=0&&s!==1&&s!==i)throw new Error(`defaultValue.shape=${r}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${o-r.length}] = ${s} but ragged tensor input.flatValues.shape[${o-r.length}] = ${i}`)}}var yy=30;function NX(r){return r<=yy?r:sp(r,Math.floor(Math.sqrt(r)))}function TX(r,t,e){let n=e*(typeof r=="number"?r:r[0]),o=t*(typeof r=="number"?r:r[1]);return[n,o]}function kX(r,t,e,n=!0){let o=[];if(n)o=o.concat(t.slice(0)),o.push(r[0]/e),o=o.concat(r.slice(1));else{o=o.concat(r[0]);let s=t.length;for(let i=0;i<s;++i)o=o.concat([r[i+1]/t[i],t[i]]);o=o.concat(r.slice(s+1))}return o}function EX(r,t,e=!0){let n=[];if(e){n.push(t);for(let o=t+1;o<r;++o)o<=2*t?(n.push(o),n.push(o-(t+1))):n.push(o)}else{let o=[],s=[];for(let i=1;i<r;++i)i>=t*2+1||i%2===1?s.push(i):o.push(i);n.push(...o),n.push(0),n.push(...s)}return n}function _X(r,t,e,n=!0){let o=[];n?o.push(r[0]/e):o.push(r[0]*e);for(let s=1;s<r.length;++s)s<=t.length?n?o.push(t[s-1]*r[s]):o.push(r[s]/t[s-1]):o.push(r[s]);return o}function AX(r,t){let e=[0];for(let n=0;n<t;++n)e.push(r[n][0]);return e}function $X(r,t,e){let n=r.slice(0,1);for(let o=0;o<e;++o)n.push(r[o+1]-t[o][0]-t[o][1]);return n}var mv=1.7580993408473768,fv=1.0507009873554805;var DX=.3275911,RX=.254829592,FX=-.284496736,OX=1.421413741,PX=-1.453152027,LX=1.061405429;function MX(r,t){if(r.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${r.length}, imag: ${t.length}.`);let e=new Float32Array(r.length*2);for(let n=0;n<e.length;n+=2)e[n]=r[n/2],e[n+1]=t[n/2];return e}function zX(r){let t=new Float32Array(r.length/2),e=new Float32Array(r.length/2);for(let n=0;n<r.length;n+=2)t[n/2]=r[n],e[n/2]=r[n+1];return{real:t,imag:e}}function BX(r){let t=Math.ceil(r.length/4),e=new Float32Array(t),n=new Float32Array(t);for(let o=0;o<r.length;o+=4)e[Math.floor(o/4)]=r[o],n[Math.floor(o/4)]=r[o+1];return{real:e,imag:n}}function VX(r){let t=Math.floor(r.length/4),e=new Float32Array(t),n=new Float32Array(t);for(let o=2;o<r.length;o+=4)e[Math.floor(o/4)]=r[o],n[Math.floor(o/4)]=r[o+1];return{real:e,imag:n}}function GX(r,t){let e=r[t*2],n=r[t*2+1];return{real:e,imag:n}}function WX(r,t,e,n){r[n*2]=t,r[n*2+1]=e}function UX(r,t){let e=new Float32Array(r/2),n=new Float32Array(r/2);for(let o=0;o<Math.ceil(r/2);o++){let s=(t?2:-2)*Math.PI*(o/r);e[o]=Math.cos(s),n[o]=Math.sin(s)}return{real:e,imag:n}}function HX(r,t,e){let n=(e?2:-2)*Math.PI*(r/t),o=Math.cos(n),s=Math.sin(n);return{real:o,imag:s}}var dv="->",qX=/->/g,CA=",",IA="...";function KX(r,t){r=r.replace(/\s/g,"");let e=(r.length-r.replace(qX,"").length)/dv.length;if(e<1)throw new Error("Equations without an arrow are not supported.");if(e>1)throw new Error(`Equation must contain exactly one arrow ("${dv}").`);let[n,o]=r.split(dv);E(n.indexOf(IA)===-1,()=>`The ellipsis notation ("${IA}") is not supported yet.`);let s=n.split(CA),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let a=[];for(let m=0;m<o.length;++m){let f=o[m];if(!s.some(d=>d.indexOf(f)!==-1))throw new Error(`Output subscripts contain the label ${f} not present in the input subscripts.`);a.indexOf(f)===-1&&a.push(f)}for(let m=0;m<n.length;++m){let f=n[m];a.indexOf(f)===-1&&f!==CA&&a.push(f)}let u=new Array(s.length);for(let m=0;m<i;++m){if(new Set(s[m].split("")).size!==s[m].length)throw new Error(`Found duplicate axes in input component ${s[m]}. Support for duplicate axes in input is not implemented yet.`);u[m]=[];for(let f=0;f<s[m].length;++f)u[m].push(a.indexOf(s[m][f]))}let l=a.length,c=o.length,p=[];for(let m=c;m<l;++m)p.push(m);return{allDims:a,summedDims:p,idDims:u}}function jX(r,t){let e=new Array(r);e.fill(-1);for(let o=0;o<t.length;++o)e[t[o]]=o;let n=[];for(let o=0;o<r;++o)e[o]===-1&&n.push(o);return e=e.filter(o=>o!==-1),{permutationIndices:e,expandDims:n}}function XX(r,t,e){let n=new Array(r);for(let o=0;o<e.length;++o){let s=e[o].shape;for(let i=0;i<t[o].length;++i)n[t[o][i]]===void 0?n[t[o][i]]=s[i]:E(n[t[o][i]]===s[i],()=>`Expected dimension ${n[t[o][i]]} at axis ${i} of input shaped ${JSON.stringify(s)}, but got dimension ${s[i]}`)}}function YX(r,t){let e=r,n=[],o=0;r.length===0&&e.push(-1),o=r.length+1;for(let i=0;i<o;++i)n.push([]);let s=[];for(let i=0;i<e.length;++i){let a=e[i],u=JX(t,a);for(let l of u)s.indexOf(l)===-1&&(n[i].push(l),s.push(l))}return{path:e,steps:n}}function ZX(r){return r.every((t,e)=>t===e)}function JX(r,t){let e=[];for(let n=0;n<r.length;++n)(r[n].length===0||r[n].indexOf(t)!==-1||t===-1)&&e.push(n);return e}function QX(r,t,e=0){let n=[];if(typeof t=="number")E(r.shape[e]%t===0,()=>"Number of splits must evenly divide the axis."),n=new Array(t).fill(r.shape[e]/t);else{let o=t.reduce((i,a)=>(a===-1&&(i+=1),i),0);E(o<=1,()=>"There should be only one negative value in split array.");let s=t.indexOf(-1);if(s!==-1){let i=t.reduce((a,u)=>u>0?a+u:a);t[s]=r.shape[e]-i}E(r.shape[e]===t.reduce((i,a)=>i+a),()=>"The sum of sizes must match the size of the axis dimension."),n=t}return n}function t5(r){return`Received SparseTensor with denseShape[0] = 0 but
-  indices.shape[0] = ${r}`}function e5(r,t){return`indices(${r}, 0) is invalid: ${t} < 0`}function r5(r,t,e){return`indices(${r}, 0) is invalid: ${t} >= ${e}`}function n5(r,t){return`only one output dimension may be -1, not both ${r} and ${t}`}function o5(r,t){return`size ${r} must be non-negative, not ${t}`}function s5(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function i5(r,t){let e=Jt(r),n=Jt(t);return`Input to reshape is a SparseTensor with ${e}
-  dense values, but the requested shape requires a multiple of ${n}. inputShape=${r} outputShape= ${t}`}function a5(r,t){let e=Jt(r),n=Jt(t);return`Input to reshape is a tensor with ${e} dense values, but the requested shape has ${n}. inputShape=${r} outputShape=${t}`}function l5(){return"segment ids must be >= 0"}function u5(){return"segment ids are not increasing"}function c5(r,t){return`Segment id ${r} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function p5(r,t,e){return`Bad: indices[${r}] == ${t} out of range [0, ${e})`}var hv={};Wt(hv,{collectGatherOpShapeInfo:()=>d5,computeOutShape:()=>f5,segOpComputeOptimalWindowSize:()=>m5});function m5(r,t){let e=!1,n;for(r<=yy?(n=r,e=!0):n=sp(r,Math.floor(Math.sqrt(r)));!e;)n>t||n===r?e=!0:n=sp(r,n+1);return n}function f5(r,t,e){let n=[],o=r.length;for(let s=0;s<o;s++)s!==t?n.push(r[s]):n.push(e);return n}function d5(r,t,e,n){let o=t.shape.length,s=r.shape.length;if(n!==0&&(n<-o||n>o))throw new Error(`Expect batchDims in the range of [-${o}, ${o}], but got ${n}`);if(n<0&&(n+=o),n>s)throw new Error(`batchDims (${n}) must be less than rank(x) (
-    ${s}).`);if(e<n)throw new Error(`batchDims (${n}) must be less than or equal to axis (${e}).`);for(let p=0;p<n;++p)if(r.shape[p]!==t.shape[p])throw new Error(`x.shape[${p}]: ${r.shape[p]} should be equal to indices.shape[${p}]: ${t.shape[p]}.`);let i=r.shape[e],a=[],u=1,l=1,c=1;for(let p=0;p<n;++p)a.push(r.shape[p]),u*=r.shape[p];for(let p=n;p<e;p++)a.push(r.shape[p]),l*=r.shape[p];for(let p=n;p<o;p++)a.push(t.shape[p]);for(let p=e+1;p<s;p++)a.push(r.shape[p]),c*=r.shape[p];return{batchSize:u,sliceSize:c,outerSize:l,dimSize:i,outputShape:a}}function h5(r){try{return r.map(t=>Wp(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function g5(r){return r.map(t=>Hl(t))}var Ur={};Wt(Ur,{nonMaxSuppressionV3Impl:()=>fy,nonMaxSuppressionV4Impl:()=>dy,nonMaxSuppressionV5Impl:()=>hy,whereImpl:()=>ay});var by={kernelName:ii,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,bo(J(e,"float32"),-1))}}};var SA={kernelName:oa,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Mt(J(e,"float32")),o=Se(ct(mt(1),n));return Ht(pt(r,o))}}}};var vA={kernelName:sa,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Se(ct(Mt(J(e,"float32")),1));return pt(r,n)}}}};var NA={kernelName:Zn,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=r,u=ge(e.shape,o);return u.length>0&&(a=ft(a,u)),R(a,e.shape)},b:()=>{let a=r,u=ge(n.shape,o);return u.length>0&&(a=ft(a,u)),R(a,n.shape)}}}};var TA={kernelName:Go,saveAllInputs:!0,gradFunc:(r,t)=>{let e={};return t.forEach((n,o)=>{e[o]=()=>r.clone()}),e}};var kA={kernelName:Wo,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>It(e)}}};var EA={kernelName:kl,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>It(e)}}};var _A={kernelName:la,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Se(ct(mt(1),Mt(J(e,"float32")))))}}};var AA={kernelName:ua,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Se(X(mt(1),Mt(J(e,"float32"))));return pt(r,n)}}}};var $A={kernelName:ma,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=X(Mt(e),Mt(n)),u=D(r,pt(n,a)),l=ge(e.shape,o);return l.length>0&&(u=ft(u,l)),R(u,e.shape)},b:()=>{let a=X(Mt(e),Mt(n)),u=Ht(D(r,pt(e,a))),l=ge(n.shape,o);return l.length>0&&(u=ft(u,l)),R(u,n.shape)}}}};var DA={kernelName:ca,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,X(Mt(J(e,"float32")),1))}}};var RA={kernelName:pa,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,ct(mt(1),Mt(J(e,"float32"))))}}};function x5(r,t,e,n,o,s){let i=I(r,"dy","avgPool3dGrad"),a=I(t,"input","avgPool3dGrad"),u=i,l=a,c=!1;a.rank===4&&(c=!0,u=R(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),l=R(a,[1,a.shape[0],a.shape[1],a.shape[2],a.shape[3]])),E(u.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${u.rank}.`),E(l.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${l.rank}.`),Ie("avgPool3dGrad",o,s);let p={dy:u,input:l},m={filterSize:e,strides:n,pad:o,dimRoundingMode:s},f=k.runKernel(lp,p,m);return c?R(f,[f.shape[1],f.shape[2],f.shape[3],f.shape[4]]):f}var FA=T({avgPool3dGrad_:x5});var OA={kernelName:El,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{filterSize:o,strides:s,pad:i,dimRoundingMode:a}=e;return{x:()=>FA(r,n,o,s,i,a)}}};function y5(r,t,e,n,o){let s=I(r,"dy","avgPoolGrad"),i=I(t,"input","avgPoolGrad");E(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let a=i,u=s,l=!1;i.rank===3&&(l=!0,a=R(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=R(s,[1,s.shape[0],s.shape[1],s.shape[2]])),E(u.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${u.rank}.`),E(a.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${a.rank}.`);let c={dy:u,input:a},p={filterSize:e,strides:n,pad:o},m=k.runKernel(ap,c,p);return l?R(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var PA=T({avgPoolGrad_:y5});var LA={kernelName:Uo,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{filterSize:o,strides:s,pad:i}=e;return{x:()=>PA(r,n,o,s,i)}}};var MA={kernelName:Ho,inputsToSave:["a","b"],gradFunc:(r,t,e)=>{let[n,o]=t,{transposeA:s,transposeB:i}=e;return!s&&!i?{a:()=>Lt(r,o,!1,!0),b:()=>Lt(n,r,!0,!1)}:!s&&i?{a:()=>Lt(r,o,!1,!1),b:()=>Lt(r,n,!0,!1)}:s&&!i?{a:()=>Lt(o,r,!1,!0),b:()=>Lt(n,r,!1,!1)}:{a:()=>Lt(o,r,!0,!0),b:()=>Lt(r,n,!0,!0)}}};var zA={kernelName:ai,gradFunc:(r,t,e)=>{let{blockShape:n,crops:o}=e;return{x:()=>nu(r,n,o)}}};var BA={kernelName:p1,gradFunc:(r,t,e)=>{let n=e,o=n.inputShape,s=n.shape,i=Array.from(s);for(let u=o.length-1;u>=0;u--)if(o[u]===s[u])i[u]=1;else if(o[u]!==1)throw new Error(`broadcastTo(): [${o}] cannot be broadcast to [${s}].`);let a=[];for(let u=0;u<i.length;u++)i[u]>1&&a.push(u);return{x:()=>ft(r,a,!0)}}};var VA={kernelName:lo,gradFunc:r=>({x:()=>r.clone()})};var GA={kernelName:qo,gradFunc:r=>({x:()=>It(r)})};var WA={kernelName:uo,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{clipValueMin:o,clipValueMax:s}=e;return{x:()=>_e(Rr(ln(n,o),Ln(n,s)),r,It(r))}}};var UA={kernelName:_l,inputsToSave:["x"],gradFunc:by.gradFunc};var HA={kernelName:li,saveAllInputs:!0,gradFunc:(r,t,e)=>{let n=t.map(u=>u.shape),{axis:o}=e,s=lr(o,t[0].shape)[0],i=n.map(u=>u[s]);return mr(r,i,s).map(u=>()=>u)}};var qA={kernelName:Ko,inputsToSave:["x","filter"],gradFunc:(r,t,e)=>{let[n,o]=t,{dilations:s,strides:i,pad:a,dataFormat:u}=e;return E(to(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>tm(n.shape,r,o,i,a,u),filter:()=>Cm(n,r,o.shape,i,a,u)}}};var KA={kernelName:jo,inputsToSave:["dy","filter"],gradFunc:(r,t,e)=>{let[n,o]=t,{strides:s,pad:i,dataFormat:a,dimRoundingMode:u}=e;return{dy:()=>In(r,o,s,i,a,1,u),filter:()=>Cm(r,n,o.shape,s,i,a,u)}}};function b5(r,t,e,n,o){let s=r;r.rank===4&&(s=R(r,[1,r.shape[0],r.shape[1],r.shape[2],r.shape[3]]));let i=t;i.rank===4&&(i=R(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),E(s.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),E(i.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),E(e.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${e}.`),E(s.shape[4]===e[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${e[3]}.`),E(i.shape[4]===e[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${e[4]}).`);let a={x:s,dy:i},u={strides:n,pad:o,filterShape:e};return k.runKernel(fp,a,u)}var jA=T({conv3DBackpropFilter_:b5});var XA={kernelName:Al,inputsToSave:["x","filter"],gradFunc:(r,t,e)=>{let{dilations:n,strides:o,pad:s}=e;E(to(n),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${n}'`);let[i,a]=t;return{x:()=>kx(i.shape,r,a,o,s),filter:()=>jA(i,r,a.shape,o,s)}}};var YA={kernelName:Xo,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(Ht(fm(J(e,"float32"))),r)}}};var ZA={kernelName:Yo,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(dm(J(e,"float32")),r)}}};var JA={kernelName:Zo,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o,exclusive:s,reverse:i}=e;return{x:()=>{let a=tv([o],n.rank),u=nm(r,o,s,!i);return a!=null&&(u=Ot(u,a)),u}}}};var QA={kernelName:Jo,inputsToSave:["x","filter"],gradFunc:(r,t,e)=>{let{dilations:n,strides:o,pad:s,dimRoundingMode:i}=e,a=n==null?[1,1]:n;E(to(a),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${a}'`);let[u,l]=t;return E(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${u.rank}.`),E(l.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${l.rank}.`),E(u.shape[3]===l.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${u.shape[3]}) must match the inChannels dimension in filter ${l.shape[2]}.`),E(Ar(o,a),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${o} and dilations '${a}'.`),Ie("depthwiseConv2d",s,i),{x:()=>cy(u.shape,r,l,o,s,a,i),filter:()=>uy(u,r,l.shape,o,s,a,i)}}};var t2={kernelName:$l,inputsToSave:["x","filter"],gradFunc:(r,t,e)=>{let[n,o]=t,s={x:n,filter:o,dy:r},i={x:n,filter:o,dy:r};return{x:()=>k.runKernel(jd,s,e),filter:()=>k.runKernel(Xd,i,e)}}};var e2={kernelName:ts,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t,n={dy:r,y:e};return{x:()=>k.runKernel(wp,n)}}};var r2={kernelName:ga,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t,n=D(er(Ht(Mt(e))),2/Math.sqrt(Math.PI));return{x:()=>D(r,n)}}};var n2={kernelName:es,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,e)}}};var o2={kernelName:ui,inputsToSave:["input"],gradFunc:(r,t)=>{let[e]=t;return{input:()=>R(r,e.shape)}}};var s2={kernelName:ya,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,er(e))}}};var i2={kernelName:rs,gradFunc:r=>({x:()=>It(r)})};var a2={kernelName:ns,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=pt(r,J(n,"float32")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,"float32")),u=ge(n.shape,o);u.length>0&&(a=R(ft(a,u),n.shape));let l=Mt(n);return Ht(pt(a,J(l,"float32")))}}}};var l2={kernelName:os,inputsToSave:["x","mean","variance","scale"],gradFunc:(r,t,e)=>{let{varianceEpsilon:n}=e,[o,s,i,a]=t,u=a==null?mt(1):a,l=ge(s.shape,o.shape),c=[];if(s.rank===1){for(let C=0;C<o.shape.length-1;++C)c.push(o.shape[C]);c.push(1)}let p=ct(o,s),m=D(r,u),f=cm(X(i,mt(n))),d=D(D(D(f,f),f),mt(-.5));return{x:()=>s.rank===1?R(D(D(r,Dr(R(f,[1,1,1,s.shape[0]]),c)),u),o.shape):R(D(D(r,f),u),o.shape),mean:()=>{let C=D(D(f,mt(-1)),m);return s.rank===1&&(C=ft(C,l)),R(C,s.shape)},variance:()=>{let C=D(D(d,p),m);return s.rank===1&&(C=ft(C,l)),R(C,s.shape)},scale:()=>{let C=D(p,f),N=D(r,C);return s.rank===1&&(N=ft(N,l)),R(N,s.shape)},offset:()=>{let C=r;return s.rank===1&&(C=ft(C,l)),R(C,s.shape)}}}};var p2={kernelName:ci,inputsToSave:["x","indices"],gradFunc:(r,t,e)=>{let[n,o]=t,{axis:s}=e,i=lr(s,n.shape)[0];return{x:()=>{let u=n.shape,l=o.size,c=u.slice(0,i),p=c.length,m=u.slice(s,u.length).slice(1),f=m.length,d=u2(0,p),h=u2(p+1,p+1+f),g=c2([c,[l],m]),x=R(r,g),b=R(o,[l]),w=c2([[p],d,h]),C=Ot(x,w),N=wm(C,b,n.shape[i]),_=ph(w);return N=Ot(N,_),N},indices:()=>o}}};function u2(r,t){let e=[];for(let n=r;n<t;++n)e.push(n);return e}function c2(r){let t=[];for(let e=0;e<r.length;++e)for(let n=0;n<r[e].length;++n)t.push(r[e][n]);return t}var m2={kernelName:ss,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>It(e),b:()=>It(n)}}};var f2={kernelName:co,gradFunc:r=>({x:()=>J(r,"float32")})};var d2={kernelName:Ia,gradFunc:r=>({x:()=>It(r)})};var h2={kernelName:Sa,gradFunc:r=>({x:()=>It(r)})};var g2={kernelName:va,gradFunc:r=>({x:()=>It(r)})};var x2={kernelName:is,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{alpha:o}=e,s=Re(n,0);return{x:()=>_e(s,r,D(r,o))}}};var y2={kernelName:ka,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,X(e,1))}}};var b2={kernelName:as,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,J(e,"float32"))}}};var w2={kernelName:f1,inputsToSave:[],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e;return{logits:()=>{let i=er(n);return ct(r,D(ft(r,o,!0),i))}}}};function w5(r,t,e,n=5,o=1,s=1,i=.5){let a={x:r,y:t,dy:e},u={depthRadius:n,bias:o,alpha:s,beta:i};return k.runKernel(Np,a,u)}var C2=T({localResponseNormalizationBackprop_:w5});var I2={kernelName:Rl,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{depthRadius:s,bias:i,alpha:a,beta:u}=e;return{x:()=>C2(n,o,r,s,i,a,u)}}};function wy(r,t,e,n){return t.rank<e.rank&&(t=R(t,yo(t.shape,n))),r.rank<e.rank&&(r=R(r,yo(r.shape,n))),{x:()=>D(r,J($r(e,t),r.dtype))}}var gv={kernelName:ls,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let n=e,{reductionIndices:o}=n,s=t[0],i=t[1],a=lr(o,s.shape),u=wy(r,i,s,a);return{x:()=>u.x()}}};var S2={kernelName:us,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>D(r,J(ln(e,n),"float32")),b:()=>D(r,J(om(e,n),"float32"))}}};function C5(r,t,e,n,o,s,i){let a=I(r,"dy","maxPool3dGrad"),u=I(t,"input","maxPool3dGrad"),l=I(e,"output","maxPool3dGrad"),c=a,p=u,m=l,f=!1;u.rank===4&&(f=!0,c=R(a,[1,a.shape[0],a.shape[1],a.shape[2],a.shape[3]]),p=R(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]]),m=R(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]])),E(c.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${c.rank}.`),E(p.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${p.rank}.`),E(m.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${m.rank}.`),Ie("maxPool3dGrad",s,i);let d={dy:c,input:p,output:m},h={filterSize:n,strides:o,pad:s,dimRoundingMode:i},g=k.runKernel(kp,d,h);return f?R(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var v2=T({maxPool3dGrad_:C5});var N2={kernelName:Fl,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=e;return{x:()=>v2(r,n,o,s,i,a,u)}}};function I5(r,t,e,n,o,s,i){let a=I(r,"dy","maxPoolGrad"),u=I(t,"input","maxPoolGrad"),l=I(e,"output","maxPoolGrad");E(u.rank===a.rank,()=>`Rank of input (${u.rank}) does not match rank of dy (${a.rank})`),E(a.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${a.rank}.`),E(u.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${u.rank}.`),Ie("maxPoolGrad",s,i);let c={dy:a,input:u,output:l},p={filterSize:n,strides:o,pad:s,dimRoundingMode:i};return k.runKernel(Tp,c,p)}var T2=T({maxPoolGrad_:I5});var k2={kernelName:cs,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n,o]=t,{filterSize:s,strides:i,pad:a}=e;return{x:()=>T2(r,n,o,s,i,a)}}};var E2={kernelName:ps,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e,s=lr(o,n.shape),a=Q0(n.shape,s)[1],u=Jt(a);return{x:()=>{let c=n.shape.slice();s.forEach(f=>{c[f]=1});let p=R(r,c);return pt(D(p,cr(n.shape,"float32")),u)}}}};var _2={kernelName:ms,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(r,t,e)=>{let n=e,{axis:o}=n,[s,i]=t,a=lr(o,s.shape),u=wy(r,i,s,a);return{x:()=>u.x()}}};var A2={kernelName:fs,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t;return{a:()=>D(r,J(Ln(e,n),"float32")),b:()=>D(r,J(Re(e,n),"float32"))}}};var $2={kernelName:ds,inputsToSave:["x"],gradFunc:(r,t,e)=>{let n=t[0],{paddings:o}=e,s=o.map(i=>i[0]);return{x:()=>Rt(r,s,n.shape)}}};var D2={kernelName:$a,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=ge(e.shape,o);return a.length>0?R(ft(r,a),e.shape):r},b:()=>{let a=D(r,Ht(Pi(pt(e,n)))),u=ge(n.shape,o);return u.length>0?R(ft(a,u),n.shape):a}}}};var R2={kernelName:hs,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=D(r,J(n,"float32")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,"float32")),u=ge(n.shape,o);return u.length>0?R(ft(a,u),n.shape):a}}}};var F2={kernelName:pi,gradFunc:r=>({x:()=>Ht(r)})};var O2={kernelName:gs,inputsToSave:["indices"],gradFunc:(r,t)=>{let e=t[0];return{indices:()=>Ne(e.shape,"float32")}}};var P2={kernelName:mi,gradFunc:r=>({x:()=>It(r)})};var L2={kernelName:fi,saveAllInputs:!0,gradFunc:(r,t,e)=>{let{axis:n}=e;return vr(r,n).map(s=>()=>s)}};var xv={kernelName:xs,inputsToSave:["x"],gradFunc:(r,t,e)=>{let n=t[0],{paddings:o}=e,s=o.map(i=>i[0]);return{x:()=>Rt(r,s,n.shape)}}};var M2={kernelName:ys,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(r,t)=>{let[e,n,o]=t,s=e,i=n,a=Pt(s.shape,i.shape);return{a:()=>{let c=J(i,"float32"),p=D(r,D(c,an(s,ct(c,mt(1))))),m=ge(s.shape,a);return m.length>0&&(p=ft(p,m)),R(p,s.shape)},b:()=>{let c=Re(s,0),p=_e(c,Sr(s),It(s)),m=D(r,D(o,p)),f=ge(i.shape,a);return f.length>0&&(m=ft(m,f)),R(m,i.shape)}}}};var z2={kernelName:bs,inputsToSave:["x","alpha"],gradFunc:(r,t)=>{let[e,n]=t,o=Re(e,0);return{x:()=>_e(o,r,D(r,n)),alpha:()=>{let s=_e(o,It(r),D(r,e)),i=ge(n.shape,r.shape);return i.length>0&&(s=ft(s,i)),R(s,n.shape)}}}};function S5(r,t,e){let n=r.shape.slice();n[e]=1;let o=R(t,n),s=Xu(r,e,!0,!1),i=Xu(r,e,!0,!0),a=D(s,i);return D(o,a)}function v5(r,t,e){let n=r.shape.length,o=n-e.length,s=v.getAxesPermutation(e,n),i=r;s!=null&&(i=Ot(r,s));let a=i.shape.slice(),l=a.splice(n-e.length,e.length).reduce((m,f)=>m*f,1);a.push(l);let c=i.reshape(a),p=S5(c,t,o);if(p=p.reshape(i.shape),s!=null){let m=v.getUndoAxesPermutation(s);p=Ot(p,m)}return p}var B2={kernelName:ws,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{axis:o}=e,s=[];return o==null?s=n.shape.map((i,a)=>a):typeof o=="number"?s=[o]:s=o,{x:()=>v5(n,r,s)}}};var V2={kernelName:Qo,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=pt(r,J(n,"float32")),u=ge(e.shape,o);return u.length>0?R(ft(a,u),e.shape):a},b:()=>{let a=D(r,J(e,"float32")),u=ge(n.shape,o);u.length>0&&(a=R(ft(a,u),n.shape));let l=Mt(n);return Ht(pt(a,J(l,"float32")))}}}};var G2={kernelName:Pa,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Ht(Mt(e)))}}};var W2={kernelName:vs,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t,n=D(Ln(e,6),bo(e));return{x:()=>D(r,J(n,"float32"))}}};var U2={kernelName:Cs,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,J(bo(e),"float32"))}}};var H2={kernelName:di,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>R(r,e.shape)}}};var q2={kernelName:Ss,inputsToSave:["images"],gradFunc:(r,t,e)=>{let[n]=t,o={dy:r,images:n};return{images:()=>k.runKernel(Op,o,e)}}};var K2={kernelName:Is,inputsToSave:["images"],gradFunc:(r,t,e)=>{let[n]=t,o={dy:r,images:n};return{images:()=>k.runKernel(Fp,o,e)}}};var j2={kernelName:Ns,gradFunc:(r,t,e)=>{let{dims:n}=e,o=lr(n,r.shape);return{x:()=>pr(r,o)}}};var X2={kernelName:Ts,gradFunc:r=>({x:()=>It(r)})};var Y2={kernelName:ks,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>Ht(pt(r,D(an(e,1.5),2)))}}};var Z2={kernelName:hi,inputsToSave:["condition"],gradFunc:(r,t)=>{let[e]=t;return{condition:()=>J(It(e),"float32"),t:()=>D(r,J(e,r.dtype)),e:()=>D(r,J(eu(e),r.dtype))}}};var J2={kernelName:Ma,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>{let n=Re(e,mt(0)),o=mt(mv),s=mt(fv),i=D(r,s),a=D(D(r,o),er(J(e,"float32")));return _e(n,i,a)}}}};var Q2={kernelName:_s,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,D(e,ct(mt(1),e)))}}};var t$={kernelName:Ba,gradFunc:r=>({x:()=>It(r)})};var e$={kernelName:Es,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(Jl(J(e,"float32")),r)}}};var r$={kernelName:za,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(rm(J(e,"float32")),r)}}};var n$={kernelName:gi,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{begin:o,size:s}=e,i=n.shape,[a,u]=q0(n,o,s),l=[];for(let c=0;c<r.rank;c++)l.push([a[c],i[c]-a[c]-u[c]]);return{x:()=>cn(r,l)}}};var o$={kernelName:Ds,outputsToSave:[!0],gradFunc:(r,t,e)=>{let[n]=t,{dim:o}=e,s=!0,i=D(r,n);return{logits:()=>ct(i,D(ft(i,[o],s),n))}}};var s$={kernelName:Va,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,Yr(e))}}};var yv={kernelName:xi,gradFunc:(r,t,e)=>{let{blockShape:n,paddings:o}=e;return{x:()=>Zl(r,n,o)}}};var bv={kernelName:yi,gradFunc:(r,t,e)=>{let{axis:n}=e;return{x:()=>ne(r,n)}}};var i$={kernelName:As,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,D(Se(J(e,"float32")),2))}}};var a$={kernelName:zl,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(r,D(J(e,"float32"),2))}}};var l$={kernelName:Rs,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=mt(2);return{a:()=>D(r,D(o,ct(e,n))),b:()=>D(r,D(o,ct(n,e)))}}};var u$={kernelName:po,gradFunc:r=>({x:()=>It(r)})};var c$={kernelName:Fs,inputsToSave:["a","b"],gradFunc:(r,t)=>{let[e,n]=t,o=Pt(e.shape,n.shape);return{a:()=>{let a=r,u=ge(e.shape,o);return u.length>0&&(a=ft(a,u)),R(a,e.shape)},b:()=>{let a=r,u=ge(n.shape,o);return u.length>0&&(a=ft(a,u)),R(Ht(a),n.shape)}}}};var p$={kernelName:$s,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,o=n.shape.slice(),{axis:s}=e;lr(s,n.shape).forEach(l=>{o[l]=1});let a=R(r,o),u=D(a,cr(n.shape,"float32"));return{x:()=>u}}};var m$={kernelName:Os,inputsToSave:["x"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>pt(r,Mt(Jl(e)))}}};var f$={kernelName:Ps,outputsToSave:[!0],gradFunc:(r,t)=>{let[e]=t;return{x:()=>D(ct(mt(1),Mt(e)),r)}}};var d$={kernelName:Jn,inputsToSave:["x"],gradFunc:(r,t,e)=>{let[n]=t,{reps:o}=e;return{x:()=>{let i=It(n);if(n.rank===1)for(let a=0;a<o[0];++a)i=X(i,Rt(r,[a*n.shape[0]],[n.shape[0]]));else if(n.rank===2)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1]],[n.shape[0],n.shape[1]]));else if(n.rank===3)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)for(let l=0;l<o[2];++l)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1],l*n.shape[2]],[n.shape[0],n.shape[1],n.shape[2]]));else if(n.rank===4)for(let a=0;a<o[0];++a)for(let u=0;u<o[1];++u)for(let l=0;l<o[2];++l)for(let c=0;c<o[3];++c)i=X(i,Rt(r,[a*n.shape[0],u*n.shape[1],l*n.shape[2],c*n.shape[3]],[n.shape[0],n.shape[1],n.shape[2],n.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${n.rank} tensors yet.`);return i}}}};var h$={kernelName:Qn,gradFunc:(r,t,e)=>{let n=e,{perm:o}=n,s=ph(o);return{x:()=>Ot(r,s)}}};var g$={kernelName:bi,gradFunc:(r,t,e)=>{let n=e,{axis:o}=n;return{value:()=>nr(r,o)}}};var x$={kernelName:Wl,inputsToSave:["segmentIds"],gradFunc:(r,t)=>{let[e]=t;return{x:()=>N5(r,e)}}};function N5(r,t){let e=Sn(t,It(t)),n=Li(r,e),o=ln(t,mt(0,"int32")),s=n.rank-o.rank;for(let a=0;a<s;++a)o=rr(o,a+1);o=Rr(o,cr(n.shape,"bool"));let i=It(n);return _e(o,n,i)}var y$={kernelName:wi,gradFunc:r=>({x:()=>It(r)})};var T5=[by,SA,vA,NA,TA,kA,EA,_A,AA,$A,DA,RA,OA,LA,MA,zA,BA,VA,GA,WA,UA,HA,KA,qA,XA,YA,ZA,JA,QA,t2,V2,e2,r2,n2,o2,s2,a2,i2,l2,p2,m2,f2,d2,h2,g2,x2,y2,b2,w2,I2,gv,gv,S2,N2,k2,E2,_2,A2,$2,D2,R2,F2,O2,P2,L2,xv,xv,M2,z2,B2,G2,W2,U2,H2,q2,K2,j2,X2,Y2,Z2,J2,Q2,t$,e$,r$,n$,o$,s$,yv,yv,bv,bv,i$,l$,a$,u$,c$,p$,m$,f$,d$,h$,g$,x$,y$];for(let r of T5)h1(r);O().prototype.abs=function(){return this.throwIfDisposed(),Ee(this)};O().prototype.acos=function(){return this.throwIfDisposed(),ax(this)};O().prototype.acosh=function(){return this.throwIfDisposed(),lx(this)};O().prototype.add=function(r){return this.throwIfDisposed(),X(this,r)};O().prototype.all=function(r,t){return this.throwIfDisposed(),Zp(this,r,t)};O().prototype.any=function(r,t){return this.throwIfDisposed(),qu(this,r,t)};O().prototype.argMax=function(r){return this.throwIfDisposed(),Ai(this,r)};O().prototype.argMin=function(r){return this.throwIfDisposed(),ux(this,r)};O().prototype.asScalar=function(){return this.throwIfDisposed(),E(this.size===1,()=>"The array must have only 1 element."),R(this,[])};O().prototype.asType=function(r){return this.throwIfDisposed(),J(this,r)};O().prototype.as1D=function(){return this.throwIfDisposed(),R(this,[this.size])};O().prototype.as2D=function(r,t){return this.throwIfDisposed(),R(this,[r,t])};O().prototype.as3D=function(r,t,e){return this.throwIfDisposed(),R(this,[r,t,e])};O().prototype.as4D=function(r,t,e,n){return this.throwIfDisposed(),R(this,[r,t,e,n])};O().prototype.as5D=function(r,t,e,n,o){return this.throwIfDisposed(),R(this,[r,t,e,n,o])};O().prototype.asin=function(){return this.throwIfDisposed(),cx(this)};O().prototype.asinh=function(){return this.throwIfDisposed(),px(this)};O().prototype.atan=function(){return this.throwIfDisposed(),mx(this)};O().prototype.atan2=function(r){return this.throwIfDisposed(),fx(this,r)};O().prototype.atanh=function(){return this.throwIfDisposed(),dx(this)};O().prototype.avgPool=function(r,t,e,n){return this.throwIfDisposed(),Yl(this,r,t,e,n)};O().prototype.batchToSpaceND=function(r,t){return this.throwIfDisposed(),Zl(this,r,t)};O().prototype.batchNorm=function(r,t,e,n,o){return this.throwIfDisposed(),Di(this,r,t,e,n,o)};O().prototype.broadcastTo=function(r){return this.throwIfDisposed(),Ri(this,r)};O().prototype.cast=function(r){return this.throwIfDisposed(),J(this,r)};O().prototype.ceil=function(){return this.throwIfDisposed(),Cx(this)};O().prototype.clipByValue=function(r,t){return this.throwIfDisposed(),Cr(this,r,t)};O().prototype.concat=function(r,t){return this.throwIfDisposed(),r instanceof Ft&&(r=[r]),ne([this,...r],t)};O().prototype.conv1d=function(r,t,e,n,o,s){return this.throwIfDisposed(),Qp(this,r,t,e,n,o,s)};O().prototype.conv2dTranspose=function(r,t,e,n,o){return this.throwIfDisposed(),em(this,r,t,e,n,o)};O().prototype.conv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),In(this,r,t,e,n,o,s)};O().prototype.cos=function(){return this.throwIfDisposed(),Jl(this)};O().prototype.cosh=function(){return this.throwIfDisposed(),rm(this)};O().prototype.cumprod=function(r,t,e){return this.throwIfDisposed(),Xu(this,r,t,e)};O().prototype.cumsum=function(r,t,e){return this.throwIfDisposed(),nm(this,r,t,e)};O().prototype.depthToSpace=function(r,t){return this.throwIfDisposed(),_x(this,r,t)};O().prototype.depthwiseConv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),Fi(this,r,t,e,n,o,s)};O().prototype.dilation2d=function(r,t,e,n,o){return this.throwIfDisposed(),Ax(this,r,t,e,n,o)};O().prototype.divNoNan=function(r){return this.throwIfDisposed(),$x(this,r)};O().prototype.div=function(r){return this.throwIfDisposed(),pt(this,r)};O().prototype.dot=function(r){return this.throwIfDisposed(),Dx(this,r)};O().prototype.elu=function(){return this.throwIfDisposed(),Oi(this)};O().prototype.equal=function(r){return this.throwIfDisposed(),$r(this,r)};O().prototype.erf=function(){return this.throwIfDisposed(),Rx(this)};O().prototype.euclideanNorm=function(r,t){return this.throwIfDisposed(),Fx(this,r,t)};O().prototype.exp=function(){return this.throwIfDisposed(),er(this)};O().prototype.expandDims=function(r){return this.throwIfDisposed(),rr(this,r)};O().prototype.expm1=function(){return this.throwIfDisposed(),Ox(this)};O().prototype.fft=function(){return this.throwIfDisposed(),au(this)};O().prototype.flatten=function(){return this.throwIfDisposed(),R(this,[this.size])};O().prototype.floor=function(){return this.throwIfDisposed(),Pi(this)};O().prototype.floorDiv=function(r){return this.throwIfDisposed(),Yp(this,r)};O().prototype.gather=function(r,t){return this.throwIfDisposed(),Li(this,r,t)};O().prototype.greaterEqual=function(r){return this.throwIfDisposed(),ln(this,r)};O().prototype.greater=function(r){return this.throwIfDisposed(),Re(this,r)};O().prototype.ifft=function(){return this.throwIfDisposed(),tl(this)};O().prototype.irfft=function(){return this.throwIfDisposed(),xm(this)};O().prototype.isFinite=function(){return this.throwIfDisposed(),Px(this)};O().prototype.isInf=function(){return this.throwIfDisposed(),Lx(this)};O().prototype.isNaN=function(){return this.throwIfDisposed(),Mx(this)};O().prototype.leakyRelu=function(r){return this.throwIfDisposed(),Ql(this,r)};O().prototype.lessEqual=function(r){return this.throwIfDisposed(),Ln(this,r)};O().prototype.less=function(r){return this.throwIfDisposed(),om(this,r)};O().prototype.localResponseNormalization=function(r,t,e,n){return this.throwIfDisposed(),zx(this,r,t,e,n)};O().prototype.logSigmoid=function(){return this.throwIfDisposed(),Gx(this)};O().prototype.logSoftmax=function(r){return this.throwIfDisposed(),sm(this,r)};O().prototype.logSumExp=function(r,t){return this.throwIfDisposed(),im(this,r,t)};O().prototype.log=function(){return this.throwIfDisposed(),Sr(this)};O().prototype.log1p=function(){return this.throwIfDisposed(),tu(this)};O().prototype.logicalAnd=function(r){return this.throwIfDisposed(),Rr(this,r)};O().prototype.logicalNot=function(){return this.throwIfDisposed(),eu(this)};O().prototype.logicalOr=function(r){return this.throwIfDisposed(),am(this,r)};O().prototype.logicalXor=function(r){return this.throwIfDisposed(),Wx(this,r)};O().prototype.matMul=function(r,t,e){return this.throwIfDisposed(),Lt(this,r,t,e)};O().prototype.maxPool=function(r,t,e,n){return this.throwIfDisposed(),ru(this,r,t,e,n)};O().prototype.max=function(r,t){return this.throwIfDisposed(),Ir(this,r,t)};O().prototype.maximum=function(r){return this.throwIfDisposed(),Sn(this,r)};O().prototype.mean=function(r,t){return this.throwIfDisposed(),ve(this,r,t)};O().prototype.min=function(r,t){return this.throwIfDisposed(),Ja(this,r,t)};O().prototype.minimum=function(r){return this.throwIfDisposed(),Mi(this,r)};O().prototype.mirrorPad=function(r,t){return this.throwIfDisposed(),qx(this,r,t)};O().prototype.mod=function(r){return this.throwIfDisposed(),Kx(this,r)};O().prototype.mul=function(r){return this.throwIfDisposed(),D(this,r)};O().prototype.neg=function(){return this.throwIfDisposed(),Ht(this)};O().prototype.norm=function(r,t,e){return this.throwIfDisposed(),Qa(this,r,t,e)};O().prototype.notEqual=function(r){return this.throwIfDisposed(),Bs(this,r)};O().prototype.oneHot=function(r,t=1,e=0){return this.throwIfDisposed(),Ei(this,r,t,e)};O().prototype.onesLike=function(){return this.throwIfDisposed(),yr(this)};O().prototype.pad=function(r,t){return this.throwIfDisposed(),cn(this,r,t)};O().prototype.pool=function(r,t,e,n,o,s){return this.throwIfDisposed(),jx(this,r,t,e,n,o,s)};O().prototype.pow=function(r){return this.throwIfDisposed(),an(this,r)};O().prototype.prelu=function(r){return this.throwIfDisposed(),ou(this,r)};O().prototype.prod=function(r,t){return this.throwIfDisposed(),Xx(this,r,t)};O().prototype.reciprocal=function(){return this.throwIfDisposed(),ty(this)};O().prototype.relu=function(){return this.throwIfDisposed(),Fr(this)};O().prototype.relu6=function(){return this.throwIfDisposed(),lm(this)};O().prototype.reshapeAs=function(r){return this.throwIfDisposed(),R(this,r.shape)};O().prototype.reshape=function(r){return this.throwIfDisposed(),R(this,r)};O().prototype.resizeBilinear=function(r,t,e){return this.throwIfDisposed(),gy(this,r,t,e)};O().prototype.resizeNearestNeighbor=function(r,t,e){return this.throwIfDisposed(),xy(this,r,t,e)};O().prototype.reverse=function(r){return this.throwIfDisposed(),pr(this,r)};O().prototype.rfft=function(){return this.throwIfDisposed(),lu(this)};O().prototype.round=function(){return this.throwIfDisposed(),um(this)};O().prototype.rsqrt=function(){return this.throwIfDisposed(),cm(this)};O().prototype.selu=function(){return this.throwIfDisposed(),pm(this)};O().prototype.separableConv2d=function(r,t,e,n,o,s){return this.throwIfDisposed(),mm(this,r,t,e,n,o,s)};O().prototype.sigmoid=function(){return this.throwIfDisposed(),Yr(this)};O().prototype.sign=function(){return this.throwIfDisposed(),ey(this)};O().prototype.sin=function(){return this.throwIfDisposed(),fm(this)};O().prototype.sinh=function(){return this.throwIfDisposed(),dm(this)};O().prototype.slice=function(r,t){return this.throwIfDisposed(),Rt(this,r,t)};O().prototype.softmax=function(r){return this.throwIfDisposed(),iu(this,r)};O().prototype.softplus=function(){return this.throwIfDisposed(),zs(this)};O().prototype.spaceToBatchND=function(r,t){return this.throwIfDisposed(),nu(this,r,t)};O().prototype.split=function(r,t){return this.throwIfDisposed(),mr(this,r,t)};O().prototype.sqrt=function(){return this.throwIfDisposed(),Se(this)};O().prototype.square=function(){return this.throwIfDisposed(),Mt(this)};O().prototype.squaredDifference=function(r){return this.throwIfDisposed(),ym(this,r)};O().prototype.squeeze=function(r){return this.throwIfDisposed(),Mn(this,r)};O().prototype.stack=function(r,t){this.throwIfDisposed();let e=r instanceof Ft?[this,r]:[this,...r];return nr(e,t)};O().prototype.step=function(r){return this.throwIfDisposed(),bo(this,r)};O().prototype.stridedSlice=function(r,t,e,n,o,s,i,a){return this.throwIfDisposed(),ry(this,r,t,e,n,o,s,i,a)};O().prototype.sub=function(r){return this.throwIfDisposed(),ct(this,r)};O().prototype.sum=function(r,t){return this.throwIfDisposed(),ft(this,r,t)};O().prototype.tan=function(){return this.throwIfDisposed(),ny(this)};O().prototype.tanh=function(){return this.throwIfDisposed(),$i(this)};O().prototype.tile=function(r){return this.throwIfDisposed(),Dr(this,r)};O().prototype.toBool=function(){return this.throwIfDisposed(),J(this,"bool")};O().prototype.toFloat=function(){return this.throwIfDisposed(),J(this,"float32")};O().prototype.toInt=function(){return this.throwIfDisposed(),J(this,"int32")};O().prototype.topk=function(r,t){return this.throwIfDisposed(),oy(this,r,t)};O().prototype.transpose=function(r){return this.throwIfDisposed(),Ot(this,r)};O().prototype.unique=function(r){return this.throwIfDisposed(),sy(this,r)};O().prototype.unsortedSegmentSum=function(r,t){return this.throwIfDisposed(),wm(this,r,t)};O().prototype.unstack=function(r){return this.throwIfDisposed(),vr(this,r)};O().prototype.where=function(r,t){return this.throwIfDisposed(),_e(r,this,t)};O().prototype.zerosLike=function(){return this.throwIfDisposed(),It(this)};var vn=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,vn.prototype)}},Hr=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,Hr.prototype)}},M=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,M.prototype)}},St=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,St.prototype)}},Im=class extends Error{constructor(t){super(t),Object.setPrototypeOf(this,Im.prototype)}};var xh=class{constructor(t){this.maxEntries=t||100,this.cache=new Map}get(t){let e;return this.cache.has(t)&&(e=this.cache.get(t),this.cache.delete(t),this.cache.set(t,e)),e}put(t,e){if(this.cache.has(t))this.cache.delete(t);else if(this.cache.size>=this.maxEntries){let n=this.cache.keys().next().value;this.cache.delete(n)}this.cache.set(t,e)}getMaxEntries(){return this.maxEntries}setMaxEntries(t){if(t<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${t}.`);if(this.maxEntries>t)for(let e=0;e<this.maxEntries-t;e++){let n=this.cache.keys().next().value;this.cache.delete(n)}this.maxEntries=t}};function Io(r,t){if(Array.isArray(r)){let e=[];for(let n=0;n<t;n++)e=e.concat(r);return e}else{let e=new Array(t);return e.fill(r),e}}function ro(r,t){if(!r)throw new Im(t)}function Cv(r,t){let e=0;for(let n of r)n===t&&e++;return e}function Nr(r){return r.length===1?r[0]:r}function xe(r){return Array.isArray(r)?r:[r]}function So(r){let e=r.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return e[0]!=="_"?e:"private"+e}function el(r){return r.length<=1||r.indexOf("_")===-1?r:r.replace(/[_]+(\w|$)/g,(t,e)=>e.toUpperCase())}var Co={};function Sm(r){if(r==null)return null;let t={};return t.className=r.getClassName(),t.config=r.getConfig(),t}function wv(r){if(!(r==null||typeof r!="object"))if(Array.isArray(r))r.forEach(t=>wv(t));else{let t=Object.keys(r);for(let e of t){let n=r[e];n!=null&&typeof n=="object"&&(!Array.isArray(n)&&n.type==="ndarray"&&typeof n.value=="number"?r[e]=n.value:wv(n))}}}function Gi(r,t={},e={},n="object",o=!1){if(typeof r=="string"){let s=r,i;if(s in e)i=e[s];else if(s in Co)i=Co[s];else if(i=t[s],i==null)throw new M(`Unknown ${n}: ${r}. This may be due to one of the following reasons:
-1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return i}else{let s=r;if(s.className==null||s.config==null)throw new M(`${n}: Improper config format: ${JSON.stringify(s)}.
-'className' and 'config' must set.`);let i=s.className,a,u;if(i in e?[a,u]=e[i]:i in Co?[a,u]=Co.className:i in t&&([a,u]=t[i]),a==null)throw new M(`Unknown ${n}: ${i}. This may be due to one of the following reasons:
-1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
-2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(u!=null){let l={};for(let f of Object.keys(Co))l[f]=Co[f];for(let f of Object.keys(e))l[f]=e[f];let c=s.config;c.customObjects=l;let p=Object.assign({},Co);for(let f of Object.keys(e))Co[f]=e[f];wv(s.config);let m=u(a,s.config,e,o);return Co=Object.assign({},p),m}else{let l=Object.assign({},Co);for(let p of Object.keys(e))Co[p]=e[p];let c=new a(s.config);return Co=Object.assign({},l),c}}}function k5(r,t){return r<t?-1:r>t?1:0}function yh(r,t){return-1*k5(r,t)}function vo(r){if(r==null)return r;let t=[];for(let e of r)t.indexOf(e)===-1&&t.push(e);return t}function b$(r){if(r==null)throw new M(`Invalid value in obj: ${JSON.stringify(r)}`);for(let t in r)if(r.hasOwnProperty(t))return!1;return!0}function Wi(r,t,e){if(e!=null&&r.indexOf(e)<0)throw new M(`${e} is not a valid ${t}.  Valid values are ${r} or null/undefined.`)}function Cy(r,t,e=0,n=1/0){return ro(e>=0),ro(n>=e),Array.isArray(r)&&r.length>=e&&r.length<=n&&r.every(o=>typeof o===t)}function Ze(r,t){Array.isArray(r)?(y.assert(r.length>0,()=>`${t} is unexpectedly an empty array.`),r.forEach((e,n)=>Ze(e,`element ${n+1} of ${t}`))):y.assert(Number.isInteger(r)&&r>0,()=>`Expected ${t} to be a positive integer, but got ${w$(r)}.`)}function w$(r){return r===null?"null":Array.isArray(r)?"["+r.map(t=>w$(t)).join(",")+"]":typeof r=="string"?`"${r}"`:`${r}`}function C$(r,t,e){let n=e!=null?e():y.now(),o;return(...i)=>{let a=e!=null?e():y.now();return a-n<t||(n=a,o=r(...i)),o}}function Iy(r){return r==="relu"?"relu":r==="linear"?"linear":r==="elu"?"elu":null}var E5=0;function vy(){return E5++}var Sy={};function gu(r=""){return r in Sy||(Sy[r]=0),Sy[r]+=1,r+Sy[r].toString()}var I$=["channelsFirst","channelsLast"],S$=["nearest","bilinear"],v$=["valid","same","causal"],N$=["max","avg"],T$=["sum","mul","concat","ave"];var vm=new Map;function Fe(r){Wi(I$,"DataFormat",r)}function E$(r){Wi(S$,"InterpolationFormat",r)}function pn(r){Wi(v$,"PaddingMode",r)}function Iv(r){Wi(N$,"PoolMode",r)}var bh=[],k$="/";function Hs(r,t){bh.push(r);try{let e=t();return bh.pop(),e}catch(e){throw bh.pop(),e}}function _5(){return bh.length===0?"":bh.join(k$)+k$}function Ny(r){if(!_$(r))throw new Error("Not a valid tensor name: '"+r+"'");return _5()+r}function Ty(r){if(!_$(r))throw new Error("Not a valid tensor name: '"+r+"'");vm.has(r)||vm.set(r,0);let t=vm.get(r);if(vm.set(r,vm.get(r)+1),t>0){let e=`${r}_${t}`;return vm.set(e,1),e}else return r}var A5=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function _$(r){return!!r.match(A5)}function A$(r){return r===parseInt(r.toString(),10)}function No(r,t,e){t==null&&(t=0),e==null&&(e=r.length);let n=1;for(let o=t;o<e;++o)n*=r[o];return n}function ac(r){if(r.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let e=0;e<r.length;e++){let n=r[e];n<t&&(t=n)}return t}function qs(r){if(r.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let e=0;e<r.length;e++){let n=r[e];n>t&&(t=n)}return t}function Zr(r,t){if(t<r)throw new M(`end (${t}) < begin (${r}) is forbidden.`);let e=[];for(let n=r;n<t;++n)e.push(n);return e}var Sv;function ir(){return Sv==null&&(Sv=gE().epsilon()),Sv}function mn(){return"channelsLast"}function no(r,t){return J(r,t)}function nl(r,t=-1){let e=r.shape.slice();return t<0&&(t=e.length+t+1),e.splice(t,0,1),R(r,e)}function $$(r,t){return B(()=>{if(r.shape.length!==2)throw new M(`repeat() expects a rank-2 tensor, but received a rank-${r.shape.length} tensor.`);let e=nl(r,1);return Ey(e,[1,t,1])})}function D$(r){let t=[No(r.shape)];return R(r,t)}function R$(r){if(r.rank<=1)throw new M(`batchFlatten requires a minimum rank of 2. Got rank: ${r.rank}.`);let t=[r.shape[0],No(r.shape,1)];return R(r,t)}function rl(r,t,e){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:return dh(r,[t,0],[e,r.shape[1]]);case 3:return gm(r,[t,0,0],[e,r.shape[1],r.shape[2]]);case 4:return ec(r,[t,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3]]);case 5:return Rt(r,[t,0,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3],r.shape[4]]);case 6:return Rt(r,[t,0,0,0,0,0],[e,r.shape[1],r.shape[2],r.shape[3],r.shape[4],r.shape[5]]);default:throw new M(`sliceAlongFirstAxis() received an unsupported tensor rank: ${r.rank}`)}})}function vv(r,t,e){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:return dh(r,[0,t],[r.shape[0],e]);case 3:return gm(r,[0,0,t],[r.shape[0],r.shape[1],e]);case 4:return ec(r,[0,0,0,t],[r.shape[0],r.shape[1],r.shape[2],e]);default:throw new M(`sliceAlongLastAxis() received an unsupported tensor rank: ${r.rank}`)}})}function wh(r,t,e,n){return B(()=>{switch(r.rank){case 1:return hm(r,t,e);case 2:switch(n){case 1:return rl(r,t,e);case 2:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}case 3:switch(n){case 1:return rl(r,t,e);case 2:return gm(r,[0,t,0],[r.shape[0],e,r.shape[2]]);case 3:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}case 4:switch(n){case 1:return rl(r,t,e);case 2:return ec(r,[0,t,0,0],[r.shape[0],e,r.shape[2],r.shape[3]]);case 3:return ec(r,[0,0,t,0],[r.shape[0],r.shape[1],e,r.shape[3]]);case 4:return vv(r,t,e);default:throw new M(`The axis is not within the rank of the tensor ${n}`)}default:throw new M(`sliceAlongLastAxis() received an unsupported tensor rank: ${r.rank}`)}})}function Nm(r,t=-1){let e;return t<0&&(e=r[0].rank,e!==0?t=e:t=0),t===r[0].rank&&(t=-1),ne(r,t)}function Tv(r,t){switch(r.rank){case 1:return Ix([r,t]);case 2:return Sx([r,t],0);case 3:return vx([r,t],0);case 4:return Nx([r,t],0);default:throw new M(`concatAlongFirstAxis() received an unsupported tensor rank: ${r.rank}`)}}function Ey(r,t){if(Array.isArray(t)||(t=[t]),r.rank!==t.length)throw new M(`The length of input n (${t.length}) does not match the number of dimensions in input x (${r.rank})`);return Dr(r,t)}function Tm(r,t=0,e=1,n,o){return tc(r,t,e,n,o)}function To(r,t,e,n){if(r.rank<2||t.rank<2)throw new St(`dot requires both inputs to be rank >= 2 but got x shape = ${r.shape} and y shape = ${t.shape}`);if(t.rank>=3){let o=r.shape.slice(-1)[0],s=t.shape.slice(-2)[0];if(o!==s)throw new St(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${r.shape} and  y shape = ${t.shape}`)}if(r.rank===2&&t.rank===2)return uu.matMul({a:r,b:t,transposeA:!1,transposeB:!1,bias:n?Nv(r.rank,n,mn()):null,activation:e});{let o=r.shape.slice(),s=o.pop();r=R(r,[-1,s]);let i=t.shape.slice(),a=i.pop(),u=i.pop(),l=[...i,a],c=Array.from({length:t.rank},(d,h)=>h===0?t.rank-2:h<=t.rank-2?h-1:h);t=R(Ot(t,c),[u,-1]);let p=[...o,...l],m=!1,f=!1;return R(uu.matMul({a:r,b:t,transposeA:m,transposeB:f,bias:n?Nv(r.rank,n,mn()):null,activation:e}),p)}}function _y(r,t,e){return B(()=>(Array.isArray(t)?t=Me(t,"int32"):t=J(t,"int32"),Li(r,t,e)))}function lc(r){return D(r,r)}function Nv(r,t,e){let n=t.shape;if(t.rank!==1&&t.rank!==r)throw new M(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${r}`);if(r===5){if(e==="channelsFirst")return n.length===1?R(t,[1,n[0],1,1,1]):R(t,[1,n[3],n[0],n[1],n[2]]);if(e==="channelsLast")return n.length===1?R(t,[1,1,1,1,n[0]]):R(t,[1].concat(n))}else if(r===4){if(e==="channelsFirst")return n.length===1?R(t,[1,n[0],1,1]):R(t,[1,n[2],n[0],n[1]]);if(e==="channelsLast")return n.length===1?R(t,[1,1,1,n[0]]):R(t,[1].concat(n))}else if(r===3){if(e==="channelsFirst")return n.length===1?R(t,[1,n[0],1]):R(t,[1,n[1],n[0]]);if(e==="channelsLast")return n.length===1?R(t,[1,1,n[0]]):R(t,[1].concat(n))}else if(r<3)return t;throw new M(`Unsupported input rank by biasAdd: ${t.rank}`)}function fn(r,t,e){return B(()=>(e==null&&(e=mn()),Fe(e),X(r,Nv(r.rank,t,e))))}function F$(r,t=1){if(t!==1)throw new St(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Oi(r)}function O$(r){return B(()=>pt(r,X(Ee(r),1)))}function Ay(r,t,e,n){return B(()=>lv(r,t,e,n))}function P$(r){return B(()=>{let t=X(.5,D(.2,r));return Cr(t,0,1)})}function xu(r,t,e=!1){return e?r():t()}var L$=["fanIn","fanOut","fanAvg"],M$=["normal","uniform","truncatedNormal"];function $5(r){Wi(L$,"FanMode",r)}function D5(r){Wi(M$,"Distribution",r)}var dn=class extends Q.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},km=class extends dn{apply(t,e){return Ne(t,e)}};km.className="Zeros";Q.registerClass(km);var yu=class extends dn{apply(t,e){return cr(t,e)}};yu.className="Ones";Q.registerClass(yu);var Em=class extends dn{constructor(t){if(super(),typeof t!="object")throw new M(`Expected argument of type ConstantConfig but got ${t}`);if(t.value===void 0)throw new M(`config must have value set but got ${t}`);this.value=t.value}apply(t,e){return B(()=>D(mt(this.value),cr(t,e)))}getConfig(){return{value:this.value}}};Em.className="Constant";Q.registerClass(Em);var _m=class extends dn{constructor(t){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=t.minval||this.DEFAULT_MINVAL,this.maxval=t.maxval||this.DEFAULT_MAXVAL,this.seed=t.seed}apply(t,e){return zi(t,this.minval,this.maxval,e)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};_m.className="RandomUniform";Q.registerClass(_m);var Am=class extends dn{constructor(t){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=t.mean||this.DEFAULT_MEAN,this.stddev=t.stddev||this.DEFAULT_STDDEV,this.seed=t.seed}apply(t,e){if(e=e||"float32",e!=="float32"&&e!=="int32")throw new St(`randomNormal does not support dType ${e}.`);return Tm(t,this.mean,this.stddev,e,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};Am.className="RandomNormal";Q.registerClass(Am);var $m=class extends dn{constructor(t){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=t.mean||this.DEFAULT_MEAN,this.stddev=t.stddev||this.DEFAULT_STDDEV,this.seed=t.seed}apply(t,e){if(e=e||"float32",e!=="float32"&&e!=="int32")throw new St(`truncatedNormal does not support dType ${e}.`);return bm(t,this.mean,this.stddev,e,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};$m.className="TruncatedNormal";Q.registerClass($m);var Dm=class extends dn{constructor(t){super(),this.gain=t.gain!=null?t.gain:1}apply(t,e){return B(()=>{if(t.length!==2||t[0]!==t[1])throw new M("Identity matrix initializer can only be used for 2D square matrices.");return D(this.gain,Yu(t[0]))})}getConfig(){return{gain:this.gain}}};Dm.className="Identity";Q.registerClass(Dm);function R5(r,t="channelsLast"){let e,n;if(Fe(t),r.length===2)e=r[0],n=r[1];else if([3,4,5].indexOf(r.length)!==-1){if(t==="channelsFirst"){let o=No(r,2);e=r[1]*o,n=r[0]*o}else if(t==="channelsLast"){let o=No(r,0,r.length-2);e=r[r.length-2]*o,n=r[r.length-1]*o}}else{let o=No(r);e=Math.sqrt(o),n=Math.sqrt(o)}return[e,n]}var qr=class extends dn{constructor(t){if(super(),t.scale<0)throw new M(`scale must be a positive float. Got: ${t.scale}`);this.scale=t.scale==null?1:t.scale,this.mode=t.mode==null?"fanIn":t.mode,$5(this.mode),this.distribution=t.distribution==null?"normal":t.distribution,D5(this.distribution),this.seed=t.seed}apply(t,e){let n=R5(t),o=n[0],s=n[1],i=this.scale;if(this.mode==="fanIn"?i/=Math.max(1,o):this.mode==="fanOut"?i/=Math.max(1,s):i/=Math.max(1,(o+s)/2),this.distribution==="normal"){let a=Math.sqrt(i);if(e=e||"float32",e!=="float32"&&e!=="int32")throw new St(`${this.getClassName()} does not support dType ${e}.`);return bm(t,0,a,e,this.seed)}else{let a=Math.sqrt(3*i);return zi(t,-a,a,e)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};qr.className="VarianceScaling";Q.registerClass(qr);var uc=class extends qr{constructor(t){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:t==null?null:t.seed})}getClassName(){return qr.className}};uc.className="GlorotUniform";Q.registerClass(uc);var cc=class extends qr{constructor(t){super({scale:1,mode:"fanAvg",distribution:"normal",seed:t==null?null:t.seed})}getClassName(){return qr.className}};cc.className="GlorotNormal";Q.registerClass(cc);var pc=class extends qr{constructor(t){super({scale:2,mode:"fanIn",distribution:"normal",seed:t==null?null:t.seed})}getClassName(){return qr.className}};pc.className="HeNormal";Q.registerClass(pc);var mc=class extends qr{constructor(t){super({scale:2,mode:"fanIn",distribution:"uniform",seed:t==null?null:t.seed})}getClassName(){return qr.className}};mc.className="HeUniform";Q.registerClass(mc);var fc=class extends qr{constructor(t){super({scale:1,mode:"fanIn",distribution:"normal",seed:t==null?null:t.seed})}getClassName(){return qr.className}};fc.className="LeCunNormal";Q.registerClass(fc);var dc=class extends qr{constructor(t){super({scale:1,mode:"fanIn",distribution:"uniform",seed:t==null?null:t.seed})}getClassName(){return qr.className}};dc.className="LeCunNormal";Q.registerClass(dc);var Rm=class extends dn{constructor(t){if(super(),this.DEFAULT_GAIN=1,this.gain=t.gain==null?this.DEFAULT_GAIN:t.gain,this.seed=t.seed,this.seed!=null)throw new St("Random seed is not implemented for Orthogonal Initializer yet.")}apply(t,e){return B(()=>{if(t.length<2)throw new St("Shape must be at least 2D.");t[0]*t[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${t[0]*t[1]}) elements: Slowness may result.`);let n=t[0]>t[1]?[t[1],t[0]]:t,o=Tm(n,0,1,"float32"),s=pv.gramSchmidt(o);return t[0]>t[1]&&(s=Ot(s)),D(this.gain,s)})}getConfig(){return{gain:this.gain,seed:this.seed}}};Rm.className="Orthogonal";Q.registerClass(Rm);var z$={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function B$(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,"initializer")}function Te(r){return Sm(r)}function de(r){if(typeof r=="string"){let t=r in z$?z$[r]:r;if(t==="GlorotNormal")return new cc;if(t==="GlorotUniform")return new uc;if(t==="HeNormal")return new pc;if(t==="HeUniform")return new mc;if(t==="LeCunNormal")return new fc;if(t==="LeCunUniform")return new dc;{let e={};return e.className=t,e.config={},B$(e)}}else return r instanceof dn?r:B$(r)}function $y(r){return Array.isArray(r)&&Array.isArray(r[0])}function Fm(r){return r.length===0?[]:Array.isArray(r[0])?r:[r]}function Nt(r){let t;if(Array.isArray(r)){if(r.length!==1)throw new M(`Expected Tensor length to be 1; got ${r.length}`);t=r[0]}else t=r;return t}function Bt(r){if(Array.isArray(r)&&Array.isArray(r[0])){if(r.length===1)return r=r,r[0];throw new M(`Expected exactly 1 Shape; got ${r.length}`)}else return r}function Om(r){let t=0;for(let e of r)e.shape.length===0?t+=1:t+=e.shape.reduce((n,o)=>n*o);return t}var G$="Variable",Ch=class{constructor(t,e="float32",n=G$,o=!0,s=null){this.dtype=e==null?"float32":e,this.shape=t.shape,this.id=vy(),n=n==null?G$:n,this.originalName=Ny(n),this.name=Ty(this.originalName),this.trainable_=o,this.constraint=s,this.val=iy(t,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(t){return this.assertNotDisposed(),O5(this.val,t),this.val.id!==t.id&&(this.val.assign(t),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(t){this.trainable_=t,this.val.trainable=t}};function O5(r,t){if(r.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(r.shape)+" vs. "+JSON.stringify(t.shape))}function Ih(r){return r.map(t=>t.read())}function Pm(r){r.forEach(t=>{t[0].write(t[1])})}var ye=class{constructor(t){this.dtype=t.dtype,this.shape=t.shape,t.shape!=null?this.ndim=t.shape.length:this.ndim=t.ndim,this.maxNDim=t.maxNDim,this.minNDim=t.minNDim,this.axes=t.axes||{}}},Jr=class{constructor(t,e,n,o,s,i,a){this.dtype=t,this.shape=e,this.sourceLayer=n,this.inputs=o,this.callArgs=s,this.outputTensorIndex=a,this.id=vy(),i!=null&&(this.originalName=Ny(i),this.name=Ty(this.originalName)),this.rank=e.length}},P5=0,ol=class{constructor(t,e){this.callArgs=e,this.id=P5++,this.outboundLayer=t.outboundLayer,this.inboundLayers=t.inboundLayers,this.nodeIndices=t.nodeIndices,this.tensorIndices=t.tensorIndices,this.inputTensors=t.inputTensors,this.outputTensors=t.outputTensors,this.inputMasks=t.inputMasks,this.outputMasks=t.outputMasks,this.inputShapes=t.inputShapes,this.outputShapes=t.outputShapes;for(let n of t.inboundLayers)n!=null&&n.outboundNodes.push(this);t.outboundLayer.inboundNodes.push(this)}getConfig(){let t=[];for(let e of this.inboundLayers)e!=null?t.push(e.name):t.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:t,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},L5=0,$t=class extends Q.Serializable{constructor(t={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=L5++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let e=t.name;if(!e){let n=this.getClassName();e=So(n)+"_"+gu(n)}if(this.name=e,this.trainable_=t.trainable==null?!0:t.trainable,t.inputShape!=null||t.batchInputShape!=null){let n;if(t.batchInputShape!=null)n=t.batchInputShape;else if(t.inputShape!=null){let s=null;t.batchSize!=null&&(s=t.batchSize),n=[s].concat(t.inputShape)}this.batchInputShape=n;let o=t.dtype;o==null&&(o=t.inputDType),o==null&&(o="float32"),this.dtype=o}t.weights!=null?this.initialWeights=t.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(t,e){return t.name+"_ib-"+e.toString()}getNodeAtIndex(t,e){if(this.inboundNodes.length===0)throw new Hr(`The layer has never been called and thus has no defined ${e}.`);if(this.inboundNodes.length<=t)throw new M(`Asked to get ${e} at node ${t}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[t]}getInputAt(t){return Nr(this.getNodeAtIndex(t,"input").inputTensors)}getOutputAt(t){return Nr(this.getNodeAtIndex(t,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new vn(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(this.inboundNodes.length===0)throw new vn(`Layer ${this.name} is not connected, no input to return.`);return Nr(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new vn(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new vn(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return Nr(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(t=>t())}get updates(){return this._updates}get built(){return this._built}set built(t){this._built=t}get trainable(){return this.trainable_}set trainable(t){this._trainableWeights.forEach(e=>e.trainable=t),this.trainable_=t}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(t=>t.trainable):[]}set trainableWeights(t){this._trainableWeights=t}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(t=>!t.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(t){this._nonTrainableWeights=t}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(t){if(t=xe(t),this.inputSpec==null||this.inputSpec.length===0)return;let e=xe(this.inputSpec);if(t.length!==e.length)throw new M(`Layer ${this.name} expects ${e.length} inputs, but it received ${t.length} input tensors. Input received: ${t}`);for(let n=0;n<t.length;n++){let o=t[n],s=e[n];if(s==null)continue;let i=o.rank;if(s.ndim!=null&&i!==s.ndim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected ndim=${s.ndim}, found ndim=${i}`);if(s.maxNDim!=null&&i>s.maxNDim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected max_ndim=${s.maxNDim}, found ndim=${i}`);if(s.minNDim!=null&&i<s.minNDim)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected min_ndim=${s.minNDim}, found ndim=${i}.`);if(s.dtype!=null&&o.dtype!==s.dtype)throw new M(`Input ${n} is incompatible with layer ${this.name} : expected dtype=${s.dtype}, found dtype=${o.dtype}.`);if(s.axes){let a=o.shape;for(let u in s.axes){let l=Number(u),c=s.axes[u],p=l>=0?a[l]:a[a.length+l];if(c!=null&&[c,null].indexOf(p)===-1)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected axis ${l} of input shape to have value ${c} but got shape ${a}.`)}}if(s.shape!=null)for(let a=0;a<s.shape.length;++a){let u=s.shape[a],l=o.shape[a];if(u!=null&&l!=null&&u!==l)throw new M(`Input ${n} is incompatible with layer ${this.name}: expected shape=${s.shape}, found shape=${o.shape}.`)}}}call(t,e){return t}invokeCallHook(t,e){this._callHook!=null&&this._callHook(t,e)}setCallHook(t){this._callHook=t}clearCallHook(){this._callHook=null}apply(t,e){e=e||{},this.assertNotDisposed();let n=xe(t),o=!0;for(let i of n)if(!(i instanceof Jr)){o=!1;break}let s=!0;for(let i of n)if(i instanceof Jr){s=!1;break}if(o===s)throw new M("Arguments to apply() must be all SymbolicTensors or all Tensors");return Hs(this.name,()=>{if(!this.built){this.assertInputCompatibility(t);let i=[];for(let a of xe(t))i.push(a.shape);this.build(Nr(i)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&s&&(this._refCount=1)}if(this.assertInputCompatibility(t),s){let i=this.call(t,e),a=xe(i),u=[];for(let l of a)n.indexOf(l)!==-1&&(l=l.clone()),u.push(l);if(i=Nr(u),this.activityRegularizer!=null)throw new St("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return i}else{let i=M5(t),a=this.computeOutputShape(i),u,l=z5(t);if(this.warnOnIncompatibleInputShape(Array.isArray(t)?i[0]:i),a!=null&&a.length>0&&Array.isArray(a[0])?u=a.map((c,p)=>new Jr(l,c,this,xe(t),e,this.name,p)):u=new Jr(l,a,this,xe(t),e,this.name),this.addInboundNode(t,u,null,null,i,a,e),this._refCount++,this.activityRegularizer!=null)throw new St("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return u}})}warnOnIncompatibleInputShape(t){if(this.batchInputShape!=null)if(t.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(t)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let e=!1;this.batchInputShape.forEach((n,o)=>{n!=null&&t[o]!=null&&t[o]!==n&&(e=!0)}),e&&console.warn(`The shape of the input tensor (${JSON.stringify(t)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new vn(`The layer ${this.name} has never been called and thus has no defined output shape.`);let t=[];for(let e of this.inboundNodes){let n=JSON.stringify(e.outputShapes);t.indexOf(n)===-1&&t.push(n)}if(t.length===1){let e=this.inboundNodes[0].outputShapes;return Array.isArray(e)&&Array.isArray(e[0])&&e.length===1?e[0]:e}else throw new vn(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new Hr(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Om(this.weights)}build(t){this.built=!0}getWeights(t=!1){return Ih(t?this.trainableWeights:this.weights)}setWeights(t){B(()=>{let e=this.weights;if(e.length!==t.length)throw new M(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${t.length}, but the layer was expecting ${e.length} weights. Provided weights: ${t}...`);if(e.length===0)return;let n=[],o=Ih(e);for(let s=0;s<o.length;++s){let i=o[s],a=e[s],u=t[s];if(!y.arraysEqual(i.shape,u.shape))throw new M(`Layer weight shape ${i.shape} not compatible with provided weight shape ${u.shape}`);n.push([a,u])}Pm(n)})}addWeight(t,e,n,o,s,i,a,u){if(this._addedWeightNames.indexOf(t)!==-1)throw new M(`Duplicate weight name ${t} for layer ${this.name}`);this._addedWeightNames.push(t),n==null&&(n="float32"),this.fastWeightInitDuringBuild&&(o=u!=null?u():de("zeros"));let l=o.apply(e,n),c=new Ch(l,n,t,i,a);return l.dispose(),s!=null&&this.addLoss(()=>s.apply(c.read())),i==null&&(i=!0),i?this._trainableWeights.push(c):this._nonTrainableWeights.push(c),c}setFastWeightInitDuringBuild(t){this.fastWeightInitDuringBuild=t}addLoss(t){t==null||Array.isArray(t)&&t.length===0||(t=xe(t),this._losses!==void 0&&this._losses!==null&&this.losses.push(...t))}computeOutputShape(t){return t}computeMask(t,e){if(!this.supportsMasking){if(e!=null)if(Array.isArray(e))e.forEach(n=>{if(n!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return e}addInboundNode(t,e,n,o,s,i,a=null){let u=xe(t);e=xe(e),n=xe(n),o=xe(o),s=Fm(s),i=Fm(i);let l=[],c=[],p=[];for(let m of u)l.push(m.sourceLayer),c.push(m.nodeIndex),p.push(m.tensorIndex);new ol({outboundLayer:this,inboundLayers:l,nodeIndices:c,tensorIndices:p,inputTensors:u,outputTensors:e,inputMasks:n,outputMasks:o,inputShapes:s,outputShapes:i},a);for(let m=0;m<e.length;m++)e[m].sourceLayer=this,e[m].nodeIndex=this.inboundNodes.length-1,e[m].tensorIndex=m}getConfig(){let t={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(t.batchInputShape=this.batchInputShape),this.dtype!=null&&(t.dtype=this.dtype),t}disposeWeights(){return this.weights.forEach(t=>t.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let t=0;return--this._refCount===0&&(t=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:t}}};function M5(r){r=xe(r);let t=[];for(let e of r)t.push(e.shape);return Nr(t)}function z5(r){return"float32"}function kv(r,t,e){if((t==null||e!=null&&e>0)&&(t=r.sourceLayer,e=r.nodeIndex),t.inboundNodes.length===0)return[r];{let n=t.inboundNodes[e];if(n.inboundLayers.length===0)return n.inputTensors;{let o=[];for(let s=0;s<n.inboundLayers.length;s++){let i=n.inputTensors[s],a=n.inboundLayers[s],u=n.nodeIndices[s],l=kv(i,a,u);for(let c of l)o.indexOf(c)===-1&&o.push(c)}return o}}}var Ks=class extends $t{constructor(t){if(super({dtype:t.dtype,name:t.name!=null?t.name:gu("input").toString()}),t.batchSize==null&&(t.batchSize=null),t.sparse==null&&(t.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=t.sparse,t.inputShape!=null&&t.batchInputShape!=null)throw new M("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let e=t.batchInputShape;if(e==null){if(t.inputShape==null)throw new M("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");e=[t.batchSize].concat(t.inputShape)}else if(t.batchSize!=null)throw new M("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let n=t.dtype||"float32";this.batchInputShape=e,this.dtype=n,this.inputSpec=[{shape:e}];let o=new Jr(this.dtype,this.batchInputShape,this,[],{},this.name);o.nodeIndex=0,o.tensorIndex=0,new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[o],outputTensors:[o],inputMasks:[null],outputMasks:[null],inputShapes:[e],outputShapes:[e]})}apply(t,e){throw new M(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};Ks.className="InputLayer";Q.registerClass(Ks);function Dy(r){if(r.batchShape==null&&r.shape==null)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(r.batchShape!=null&&r.shape!=null)throw new M("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=r.batchShape;r.shape!=null&&t==null&&(t=[null].concat(r.shape));let e=r.dtype;return e==null&&(e="float32"),new Ks({batchInputShape:t,name:r.name,dtype:e,sparse:r.sparse}).inboundNodes[0].outputTensors[0]}function B5(r,t){if(r.dtype==null||r.dtype===t.dtype)return t;try{return J(t,r.dtype)}catch(e){throw new M(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${r.name}' (${r.dtype}).`)}}var ko=class{constructor(t){if(this.id2Value={},this.id2Mask={},this.name2Id={},t instanceof ko)for(let e in t.id2Value)this.id2Value[e]=t.id2Value[e],e in t.id2Mask&&(this.id2Mask[e]=t.id2Mask[e]);else{if(t==null)return;for(let e of t)this.add(e.key,e.value)}}add(t,e,n){if(this.id2Value[t.id]==null)this.id2Value[t.id]=B5(t,e),this.name2Id[t.name]=t.id,n!=null&&(this.id2Mask[t.id]=n);else throw new M(`Duplicate key: name=${t.name}, id=${t.id}`);return this}addFeed(t){this.add(t.key,t.value)}hasKey(t){return this.id2Value[t.id]!=null}names(){return Object.keys(this.name2Id)}getValue(t){if(t instanceof Jr){if(this.id2Value[t.id]==null)throw new M(`Nonexistent key: ${t.name}`);return this.id2Value[t.id]}else{let e=this.name2Id[t];if(e==null)throw new M(`Feed dict has no SymbolicTensor name: ${t}`);return this.id2Value[e]}}getMask(t){if(t instanceof Jr){if(this.id2Value[t.id]==null)throw new M(`Nonexistent key: ${t.name}`);return this.id2Mask[t.id]}else{let e=this.name2Id[t];if(e==null)throw new M(`Feed dict has no SymbolicTensor name: ${t}`);return this.id2Mask[e]}}disposeMasks(){this.id2Mask!=null&&vt(this.id2Mask)}},Ry=new xh,Fy=new xh;function U$(r){Ry!=null&&Ry.setMaxEntries(r),Fy!=null&&Fy.setMaxEntries(r)}function hc(r,t,e,n){let o=e==null?!1:e.training,s=Array.isArray(r),i=s?r:[r],a=i.map(d=>d.name),u=[],l=t.names();for(let d of a)l.indexOf(d)!==-1?u.push(t.getValue(d)):u.push(null);n!=null&&(n.maxNumTensors=-1/0,n.minNumTensors=1/0);let c=a.join(",")+"|"+t.names().sort().join(","),p=Ry.get(c),m;if(p==null){let d=V5(i,t);p=d.sorted,m=d.recipientCounts,Ry.put(c,p),Fy.put(c,m)}m={},o||Object.assign(m,Fy.get(c));let f=new ko(t);for(let d=0;d<p.length;++d){if(n!=null){let F=ah().numTensors;F>n.maxNumTensors&&(n.maxNumTensors=F),F<n.minNumTensors&&(n.minNumTensors=F)}let h=p[d],g=h.sourceLayer;if(g instanceof Ks)continue;let x=[],b=[],w=[],C=!1;for(let F of h.inputs){let P=f.getValue(F),V=f.getMask(F);x.push(P),b.push(V),V!=null&&(C=!0),o||(m[F.name]--,m[F.name]===0&&!t.hasKey(F)&&a.indexOf(F.name)===-1&&!P.isDisposed&&F.sourceLayer.stateful!==!0&&w.push(P))}C&&(e=e||{},e.mask=b[0]);let N=xe(g.apply(x,e)),_=null;g.supportsMasking&&(_=g.computeMask(x,b));let A=W5(h),$=Array.isArray(A)?A:[A];for(let F=0;F<$.length;++F){f.hasKey($[F])||f.add($[F],N[F],Array.isArray(_)?_[0]:_);let P=a.indexOf($[F].name);P!==-1&&(u[P]=N[F])}o||vt(w)}return f.disposeMasks(),s?u:u[0]}function V5(r,t){y.assert(r!=null&&r.length>0,()=>"Expected at least one fetch, got none");let e=[],n={};if(r.length===1){let o=W$(r[0],t);e=o.sorted,n=o.recipientMap}else{let o=new Set;for(let s of r){let{sorted:i,recipientMap:a}=W$(s,t);for(let u of i)o.has(u.name)||(e.push(u),o.add(u.name));for(let u in a)n[u]==null&&(n[u]=new Set),a[u].forEach(l=>n[u].add(l))}}return{sorted:e,recipientCounts:G5(n)}}function G5(r){let t={};for(let e in r)t[e]=r[e].size;return t}function W$(r,t){let e=new Set,n=[],o={};for(let a of t.names())e.add(a);let s=[],i=[];for(s.push(r);s.length>0;){let a=s[s.length-1];if(e.has(a.name)){s.pop();continue}let u=i[i.length-1]===s.length-1;if(a.inputs.length===0||u)s.pop(),n.push(a),e.add(a.name),u&&i.pop();else{i.push(s.length-1);for(let l of a.inputs)o[l.name]==null&&(o[l.name]=new Set),o[l.name].add(a.name),!e.has(l.name)&&s.push(l)}}return{sorted:n,recipientMap:o}}function W5(r){let t;if(r.sourceLayer.inboundNodes.length===1)t=r.sourceLayer.output;else{let e=null;for(let n=0;n<r.sourceLayer.inboundNodes.length;++n)for(let o of r.sourceLayer.inboundNodes[n].outputTensors)if(o.id===r.id){e=n;break}t=r.sourceLayer.getOutputAt(e)}return t}var U5=z();U5.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,U$);var K$={};Wt(K$,{maxNorm:()=>H5,minMaxNorm:()=>j5,nonNeg:()=>K5,unitNorm:()=>q5});function Ev(r,t){return B(()=>Se(ft(D(r,r),t,!0)))}var gc=class extends Q.Serializable{getConfig(){return{}}},Lm=class extends gc{constructor(t){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=t.maxValue!=null?t.maxValue:this.defaultMaxValue,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>{let e=Ev(t,this.axis),n=Cr(e,0,this.maxValue);return D(t,pt(n,X(ir(),e)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};Lm.className="MaxNorm";Q.registerClass(Lm);var Mm=class extends gc{constructor(t){super(),this.defaultAxis=0,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>pt(t,X(ir(),Ev(t,this.axis))))}getConfig(){return{axis:this.axis}}};Mm.className="UnitNorm";Q.registerClass(Mm);var zm=class extends gc{apply(t){return Fr(t)}};zm.className="NonNeg";Q.registerClass(zm);var Bm=class extends gc{constructor(t){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=t.minValue!=null?t.minValue:this.defaultMinValue,this.maxValue=t.maxValue!=null?t.maxValue:this.defaultMaxValue,this.rate=t.rate!=null?t.rate:this.defaultRate,this.axis=t.axis!=null?t.axis:this.defaultAxis}apply(t){return B(()=>{let e=Ev(t,this.axis),n=X(D(this.rate,Cr(e,this.minValue,this.maxValue)),D(1-this.rate,e));return D(t,pt(n,X(ir(),e)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};Bm.className="MinMaxNorm";Q.registerClass(Bm);var H$={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function ze(r){return Sm(r)}function q$(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,"constraint")}function Be(r){if(r==null)return null;if(typeof r=="string"){let e={className:r in H$?H$[r]:r,config:{}};return q$(e)}else return r instanceof gc?r:q$(r)}function H5(r){return new Lm(r)}function q5(r){return new Mm(r)}function K5(){return new zm}function j5(r){return new Bm(r)}var j$={};Wt(j$,{constant:()=>Z5,glorotNormal:()=>o8,glorotUniform:()=>n8,heNormal:()=>s8,heUniform:()=>i8,identity:()=>e8,leCunNormal:()=>a8,leCunUniform:()=>l8,ones:()=>Y5,orthogonal:()=>u8,randomNormal:()=>Q5,randomUniform:()=>J5,truncatedNormal:()=>t8,varianceScaling:()=>r8,zeros:()=>X5});function X5(){return new km}function Y5(){return new yu}function Z5(r){return new Em(r)}function J5(r){return new _m(r)}function Q5(r){return new Am(r)}function t8(r){return new $m(r)}function e8(r){return new Dm(r)}function r8(r){return new qr(r)}function n8(r){return new uc(r)}function o8(r){return new cc(r)}function s8(r){return new pc(r)}function i8(r){return new mc(r)}function a8(r){return new fc(r)}function l8(r){return new dc(r)}function u8(r){return new Rm(r)}var ED={};Wt(ED,{Layer:()=>$t,RNN:()=>Tn,RNNCell:()=>ll,activation:()=>NY,add:()=>FY,alphaDropout:()=>xZ,average:()=>OY,averagePooling1d:()=>Gv,averagePooling2d:()=>Wv,averagePooling3d:()=>Uv,avgPool1d:()=>UY,avgPool2d:()=>qY,avgPool3d:()=>jY,avgPooling1d:()=>HY,avgPooling2d:()=>KY,avgPooling3d:()=>XY,batchNormalization:()=>VY,bidirectional:()=>uZ,categoryEncoding:()=>CZ,concatenate:()=>PY,conv1d:()=>gY,conv2d:()=>xY,conv2dTranspose:()=>yY,conv3d:()=>bY,conv3dTranspose:()=>wY,convLstm2d:()=>sZ,convLstm2dCell:()=>iZ,cropping2D:()=>IY,dense:()=>TY,depthwiseConv2d:()=>vY,dot:()=>BY,dropout:()=>kY,elu:()=>cY,embedding:()=>RY,flatten:()=>_Y,gaussianDropout:()=>gZ,gaussianNoise:()=>hZ,globalAveragePooling1d:()=>YY,globalAveragePooling2d:()=>ZY,globalMaxPool1d:()=>pZ,globalMaxPool2d:()=>mZ,globalMaxPooling1d:()=>vD,globalMaxPooling2d:()=>ND,gru:()=>QY,gruCell:()=>tZ,input:()=>Pv,inputLayer:()=>uY,layerNormalization:()=>GY,leakyReLU:()=>mY,lstm:()=>eZ,lstmCell:()=>rZ,masking:()=>yZ,maxPool1d:()=>fZ,maxPool2d:()=>dZ,maxPooling1d:()=>TD,maxPooling2d:()=>kD,maxPooling3d:()=>JY,maximum:()=>LY,minimum:()=>MY,multiply:()=>zY,permute:()=>DY,prelu:()=>fY,reLU:()=>pY,repeatVector:()=>AY,rescaling:()=>bZ,reshape:()=>$Y,resizing:()=>wZ,rnn:()=>aZ,separableConv2d:()=>CY,simpleRNN:()=>nZ,simpleRNNCell:()=>oZ,softmax:()=>dY,spatialDropout1d:()=>EY,stackedRNNCells:()=>lZ,thresholdedReLU:()=>hY,timeDistributed:()=>cZ,upSampling2d:()=>SY,zeroPadding2d:()=>WY});async function Ui(r){if(r==null)return;let t=[],e=[],n=[];for(let o in r){let s=r[o];if(typeof s!="number"){let i=s;t.push(i.data()),e.push(o),n.push(i)}}if(t.length>0){let o=await Promise.all(t);for(let s=0;s<o.length;++s)r[e[s]]=o[s][0];vt(n)}}function Oy(r){if(r!=null)for(let t in r){let e=r[t];typeof e!="number"&&e.dispose()}}var X$;(function(r){r[r.SILENT=0]="SILENT",r[r.VERBOSE=1]="VERBOSE"})(X$||(X$={}));var c8=125,sl=class{constructor(){this.validationData=null}setParams(t){this.params=t}async onEpochBegin(t,e){}async onEpochEnd(t,e){}async onBatchBegin(t,e){}async onBatchEnd(t,e){}async onTrainBegin(t){}async onTrainEnd(t){}setModel(t){}},Py=class{constructor(t,e=10){t==null&&(t=[]),this.callbacks=t,this.queueLength=e}append(t){this.callbacks.push(t)}setParams(t){for(let e of this.callbacks)e.setParams(t)}setModel(t){for(let e of this.callbacks)e.setModel(t)}async onEpochBegin(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onEpochBegin(t,e)}async onEpochEnd(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onEpochEnd(t,e)}async onBatchBegin(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onBatchBegin(t,e)}async onBatchEnd(t,e){e==null&&(e={});for(let n of this.callbacks)await n.onBatchEnd(t,e)}async onTrainBegin(t){t==null&&(t={});for(let e of this.callbacks)await e.onTrainBegin(t)}async onTrainEnd(t){t==null&&(t={});for(let e of this.callbacks)await e.onTrainEnd(t)}},_v=class extends sl{constructor(){super()}async onEpochBegin(t){this.seen=0,this.totals={}}async onBatchEnd(t,e){e==null&&(e={});let n=e.size==null?0:e.size;this.seen+=n;for(let o in e){let s=e[o];if(typeof s=="number")this.totals.hasOwnProperty(o)||(this.totals[o]=0),this.totals[o]=this.totals[o]+s*n;else{let i;o in this.totals?i=this.totals[o]:this.totals[o]=0;let a=B(()=>X(this.totals[o],D(s,n)));this.totals[o]=a,i!=null&&i.dispose()}}}async onEpochEnd(t,e){if(e!=null)for(let n of this.params.metrics)this.totals[n]!=null&&(typeof this.totals[n]=="number"?e[n]=this.totals[n]/this.seen:B(()=>{let o=D(pt(1,this.seen),this.totals[n]);e[n]=o,this.totals[n].dispose(),De(e[n])}))}},Ly=class extends sl{async onTrainBegin(t){this.epoch=[],this.history={}}async onEpochEnd(t,e){e==null&&(e={}),this.epoch.push(t);for(let n in e)this.history[n]==null&&(this.history[n]=[]),this.history[n].push(e[n])}async syncData(){let t=[],e=[],n=[];for(let s in this.history){let i=this.history[s];for(let a=0;a<i.length;++a)if(typeof i[a]!="number"){let u=i[a];t.push(u.data()),e.push(s),n.push(a)}}let o=await Promise.all(t);for(let s=0;s<o.length;++s)this.history[e[s]][n[s]].dispose(),this.history[e[s]][n[s]]=o[s][0]}},My=class extends sl{constructor(t,e){if(super(),this.currentEpoch=0,this.nowFunc=t.nowFunc,this.nextFrameFunc=t.nextFrameFunc||gh,this.yieldEvery=e||"auto",this.yieldEvery==="auto"&&(this.yieldEvery=c8),this.yieldEvery==="never"&&t.onYield!=null)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");y.isNumber(this.yieldEvery)&&(this.maybeWait=C$(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=t.onTrainBegin,this.trainEnd=t.onTrainEnd,this.epochBegin=t.onEpochBegin,this.epochEnd=t.onEpochEnd,this.batchBegin=t.onBatchBegin,this.batchEnd=t.onBatchEnd,this.yield=t.onYield}async maybeWait(t,e,n){let o=[];this.yield!=null&&(await Ui(n),o.push(this.yield(t,e,n))),o.push(this.nextFrameFunc()),await Promise.all(o)}async onEpochBegin(t,e){this.currentEpoch=t,this.epochBegin!=null&&(await Ui(e),await this.epochBegin(t,e))}async onEpochEnd(t,e){let n=[];this.epochEnd!=null&&(await Ui(e),n.push(this.epochEnd(t,e))),this.yieldEvery==="epoch"&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(t,e){this.batchBegin!=null&&(await Ui(e),await this.batchBegin(t,e))}async onBatchEnd(t,e){let n=[];this.batchEnd!=null&&(await Ui(e),n.push(this.batchEnd(t,e))),this.yieldEvery==="batch"?n.push(this.nextFrameFunc()):y.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,t,e)),await Promise.all(n)}async onTrainBegin(t){this.trainBegin!=null&&(await Ui(t),await this.trainBegin(t))}async onTrainEnd(t){this.trainEnd!=null&&(await Ui(t),await this.trainEnd(t))}};function zy(r,t){return r==null&&(r={}),r instanceof sl?[r]:Array.isArray(r)&&r[0]instanceof sl?r:xe(r).map(n=>new My(n,t))}var hn=class{constructor(){}static registerCallbackConstructor(t,e){y.assert(t>=0&&Number.isInteger(t),()=>`Verbosity level is expected to be an integer >= 0, but got ${t}`),hn.checkForDuplicate(e),hn.constructors[t]==null&&(hn.constructors[t]=[]),hn.constructors[t].push(e)}static checkForDuplicate(t){for(let e in hn.constructors)hn.constructors[+e].forEach(o=>{if(o===t)throw new M("Duplicate callback constructor.")})}static clear(){hn.constructors={}}static createCallbacks(t){let e=[];for(let n in hn.constructors){let o=+n;t>=o&&e.push(...hn.constructors[o])}return e.map(n=>new n)}};hn.constructors={};function By(r,t,e,n,o,s,i,a,u){let l=new Ly,c=[new _v,...hn.createCallbacks(t)];r!=null&&c.push(...r),c.push(l);let p=new Py(c);return p.setParams({epochs:e,initialEpoch:n,samples:o,steps:s,batchSize:i,verbose:t,doValidation:a,metrics:u}),{callbackList:p,history:l}}function gn(r,t={},e=!1){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,"layer",e)}function Sh(r,t){return B(()=>{r.dtype!=="float32"&&(r=J(r,"float32"));let e=ft(lc(r),t,!0),n=xo(e.shape,ir()),o=Se(Sn(e,n));return pt(r,o)})}function Hi(r,t){return B(()=>ve(lc(ct(t,r)),-1))}function Vm(r,t){return B(()=>ve(Ee(ct(t,r)),-1))}function bu(r,t){return B(()=>{let e=ct(r,t),n=Cr(Ee(r),ir(),Number.MAX_VALUE),o=Ee(pt(e,n));return D(100,ve(o,-1))})}function p8(r,t){return B(()=>{let e=Cr(t,ir(),Number.MAX_VALUE),n=Sr(X(1,e)),o=Cr(r,ir(),Number.MAX_VALUE),s=Sr(X(1,o));return ve(lc(ct(n,s)),-1)})}function m8(r,t){return B(()=>{let e=Sn(0,ct(1,D(r,t)));return ve(lc(e),-1)})}function f8(r,t){return B(()=>{let e=Sn(0,ct(1,D(r,t)));return ve(e,-1)})}function d8(r,t){return B(()=>{let e=ft(D(r,t),-1),n=Ir(D(ct(1,r),t),-1);return Sn(0,X(1,ct(n,e)))})}function h8(r,t){return B(()=>{let e=Math.log(2),n=ct(t,r),o=ct(X(n,zs(D(-2,n))),e);return ve(o,-1)})}function xc(r,t,e=!1){return B(()=>{if(e)t=iu(t);else{let n=ft(t,t.shape.length-1,!0);t=pt(t,n)}return t=Cr(t,ir(),1-ir()),Ht(ft(D(J(r,"float32"),Sr(t)),t.shape.length-1))})}function Gm(r,t,e=!1){return B(()=>{let n=J(Pi(D$(r)),"int32");t=Cr(t,ir(),1-ir());let o=t.shape,s=R(Ei(n,o[o.length-1]),o);return xc(s,t,e)})}function g8(r,t){if(!y.arraysEqual(r.shape,t.shape))throw new M(`logits and labels must have the same shape, but got shapes ${JSON.stringify(r.shape)} and ${JSON.stringify(t.shape)}`);return B(()=>{let e=Fr(t),n=Ht(Ee(t));return X(ct(e,D(t,r)),tu(er(n)))})}function Wm(r,t){return B(()=>{let e;return e=Cr(t,ir(),1-ir()),e=Sr(pt(e,ct(1,e))),ve(g8(r,e),-1)})}function x8(r,t){return B(()=>{let e=Cr(r,ir(),1),n=Cr(t,ir(),1);return ft(D(r,Sr(pt(e,n))),-1)})}function y8(r,t){return B(()=>{let e=Sr(X(ir(),t));return ve(ct(t,D(r,e)),-1)})}function Nh(r,t){return B(()=>{let e=Sh(r,-1),n=Sh(t,-1),o=D(e,n);return Ht(ft(o,-1))})}var vh={meanSquaredError:Hi,meanAbsoluteError:Vm,meanAbsolutePercentageError:bu,meanSquaredLogarithmicError:p8,squaredHinge:m8,hinge:f8,categoricalHinge:d8,logcosh:h8,categoricalCrossentropy:xc,sparseCategoricalCrossentropy:Gm,binaryCrossentropy:Wm,kullbackLeiblerDivergence:x8,poisson:y8,cosineProximity:Nh};function Vy(r){if(typeof r=="string"){if(r in vh)return vh[r];let t=`Unknown loss ${r}`;throw r.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${r}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new M(t)}else return r}function Th(r,t){return B(()=>{let e=D(.5,yr(t)),n=no(Re(t,e),r.dtype);return ve($r(r,n),-1)})}function kh(r,t){return B(()=>no($r(Ai(r,-1),Ai(t,-1)),"float32"))}function Z$(r,t){return B(()=>J(ft(Rr($r(r,1),$r(t,1))),"float32"))}function b8(r,t){return B(()=>J(ft(Rr($r(r,1),$r(t,0))),"float32"))}function w8(r,t){return B(()=>J(ft(Rr($r(r,0),$r(t,1))),"float32"))}function Av(r,t){return B(()=>{let e=Z$(r,t),n=w8(r,t),o=X(e,n);return J(_e(Re(o,0),pt(e,o),0),"float32")})}function J$(r,t){return B(()=>{let e=Z$(r,t),n=b8(r,t),o=X(e,n);return J(_e(Re(o,0),pt(e,o),0),"float32")})}function Wy(r,t){return Wm(r,t)}function Uy(r,t){return r.rank===t.rank&&(r=Mn(r,[r.rank-1])),t=Ai(t,-1),t.dtype!==r.dtype&&(t=J(t,r.dtype)),J($r(r,t),"float32")}var C8=Hi,I8=Hi,S8=Vm,v8=Vm,N8=bu,T8=bu,Eh=xc,k8=Nh,$v=Gm,Gy={binaryAccuracy:Th,categoricalAccuracy:kh,precision:Av,categoricalCrossentropy:Eh,sparseCategoricalCrossentropy:$v,mse:C8,MSE:I8,mae:S8,MAE:v8,mape:N8,MAPE:T8,cosine:k8};function Q$(r){if(typeof r=="string"&&r in Gy)return Gy[r];if(typeof r!="string"&&r!=null)return r;throw new M(`Unknown metric ${r}`)}function _h(r){if(ro(r!==null,`Unknown LossOrMetricFn ${r}`),typeof r=="string")return r;{let t;for(let e of Object.keys(vh))if(vh[e]===r){t=e;break}if(t!==void 0)return t;for(let e of Object.keys(Gy))if(Gy[e]===r){t=e;break}return t!==void 0?t:r.name}}function eD(r){let t={Adagrad:()=>ic.adagrad(.01),Adadelta:()=>ic.adadelta(1,.95,ir()),Adam:()=>ic.adam(.001,.9,.999,ir()),Adamax:()=>ic.adamax(.002,.9,.999,ir(),0),RMSProp:()=>ic.rmsprop(.001,.9,0,ir()),SGD:()=>ic.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,r in t)return t[r]();throw new M(`Unknown Optimizer ${r}`)}function Rv(r,t,e=!1){if(r==null||typeof r!="object"||Object.getPrototypeOf(r)!==Object.prototype||!Dv(r))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(e){let n=JSON.stringify(r);n.length>1048576&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${1048576}.`)}}function Dv(r){if(r===null)return!0;if(typeof r=="object")if(Object.getPrototypeOf(r)===Object.prototype){let t=Object.keys(r);for(let e of t)if(typeof e!="string"||!Dv(r[e]))return!1;return!0}else if(Array.isArray(r)){for(let t of r)if(!Dv(t))return!1;return!0}else return!1;else{let t=typeof r;return t==="string"||t==="number"||t==="boolean"}}function rD(r,t,e,n=console.log){let o=A8(r),s=["Layer (type)","Input Shape","Output shape","Param #"];o?(t=t||90,e=e||[.32,.61,.89,1]):(t=t||115,e=e||[.24,.48,.7,.8,1]),e[e.length-1]<=1&&(e=e.map(c=>Math.floor(t*c)));let i;if(!o){s.push("Receives inputs"),i=[];for(let c in r.nodesByDepth)i.push(...r.nodesByDepth[c])}n("_".repeat(t)),Hy(s,e,n),n("=".repeat(t));let a=r.layers;for(let c=0;c<a.length;++c)o?$8(a[c],e,n):D8(a[c],e,i,n),n((c===a.length-1?"=":"_").repeat(t));r.checkTrainableWeightsConsistency();let u=_8(r),l=Om(r.nonTrainableWeights);n(`Total params: ${u+l}`),n(`Trainable params: ${u}`),n(`Non-trainable params: ${l}`),n("_".repeat(t))}function _8(r){let t;return r.collectedTrainableWeights!=null?t=Om(r.collectedTrainableWeights):t=Om(r.trainableWeights),t}function A8(r){let t=!0,e=[],n=[];for(let o in r.nodesByDepth)e.push(r.nodesByDepth[o]);for(let o of e){if(o.length>1||o.length===1&&o[0].inboundLayers.length>1){t=!1;break}n.push(...o)}if(t)for(let o of r.layers){let s=!1;for(let i of o.inboundNodes)if(n.indexOf(i)!==-1)if(s){t=!1;break}else s=!0;if(!t)break}return t}function Hy(r,t,e=console.log){let n="";for(let o=0;o<r.length;++o)o>0&&(n=n.slice(0,n.length-1)+" "),n+=r[o],n=n.slice(0,t[o]),n+=" ".repeat(t[o]-n.length);e(n)}function $8(r,t,e){let n,o;try{o=r.inboundNodes.map(u=>JSON.stringify(u.inputShapes)).join(",")}catch(u){o="multiple"}try{n=JSON.stringify(r.outputShape)}catch(u){n="multiple"}let s=r.name,i=r.getClassName(),a=[`${s} (${i})`,o,n,r.countParams().toString()];Hy(a,t,e)}function D8(r,t,e,n){let o,s;try{s=r.inboundNodes.map(p=>JSON.stringify(p.inputShapes)).join(",")}catch(p){s="multiple"}try{o=JSON.stringify(r.outputShape)}catch(p){o="multiple"}let i=[];for(let p of r.inboundNodes)if(!(e!=null&&e.length>0&&e.indexOf(p)===-1))for(let m=0;m<p.inboundLayers.length;++m){let f=p.inboundLayers[m].name,d=p.nodeIndices[m],h=p.tensorIndices[m];i.push(`${f}[${d}][${h}]`)}let a=r.name,u=r.getClassName(),l=i.length===0?"":i[0],c=[`${a} (${u})`,s,o,r.countParams().toString(),l];Hy(c,t,n);for(let p=1;p<i.length;++p)Hy(["","","","",i[p]],t,n)}function nD(r,t,e){return(r==="inboundNodes"||r==="outputLayers"||r==="inputLayers")&&t===0&&typeof e=="string"}function yc(r,t){if(r===null)return null;if(typeof r=="string")return el(r);if(typeof r=="number"||typeof r=="boolean")return r;if(r instanceof Array){let e=[],n=r.length;for(let o=0;o<n;++o){let s=r[o];nD(t,o,s)?e.push(s):e.push(yc(s,t))}return e}else{let e={};for(let n of Object.keys(r)){let o=r[n];if(n==="name"&&typeof o=="string")e[n]=o;else{let s=el(n);e[s]=yc(o,s)}}return e}}function qy(r,t){if(r==null)return null;if(typeof r=="string")return So(r);if(typeof r=="number"||typeof r=="boolean")return r;if(r instanceof Array){let e=[],n=r.length;for(let o=0;o<n;++o){let s=r[o];nD(t,o,s)?e.push(s):e.push(qy(s,t))}return e}else{let e={};for(let n of Object.keys(r)){let o=r[n],s=So(n);(n==="name"||n==="className")&&typeof o=="string"?e[s]=o:e[s]=qy(o,n)}return e}}var Um="4.0.0";var zn=class extends $t{constructor(t){if(super({}),this.containerNodes=new Set,this.name=t.name,this.name==null){let b=this.getClassName().toLowerCase();this.name=gu(b)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(t.inputs)?this.inputs=t.inputs.slice():this.inputs=[t.inputs],Array.isArray(t.outputs)?this.outputs=t.outputs.slice():this.outputs=[t.outputs],vo(this.inputs).length!==this.inputs.length)throw new M(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(b=>b.name)}`);vo(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(b=>b.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let b of this.outputs){let w=b.sourceLayer,C=b.nodeIndex,N=b.tensorIndex;this.outputLayers.push(w),this.outputLayersNodeIndices.push(C),this.outputLayersTensorIndices.push(N)}for(let b of this.inputs){let w=b.sourceLayer,C=b.nodeIndex,N=b.tensorIndex;ro(C===0,"input layer has >1 nodes"),ro(N===0,"input layer has >1 tensors"),this.inputLayers.push(w),this.inputLayersNodeIndices.push(C),this.inputLayersTensorIndices.push(N)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let b=0;b<this.inputLayers.length;b++){let w=this.inputLayers[b];if(!(w instanceof Ks))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${t.inputs}. Input ${b} (0-based) originates from layer type ${w.getClassName()}.`);this.inputNames.push(w.name),this.feedInputShapes.push(w.batchInputShape),this.feedInputNames.push(w.name)}for(let b of this.outputLayers)this.outputNames.push(b.name);this.internalInputShapes=this.inputs.map(b=>b.shape),this.internalOutputShapes=this.outputs.map(b=>b.shape);let e={},n={},o={},s={},i={},a=[],u=(b,w,C,N,_,A)=>{(N==null||_==null||A==null)&&(N=b.sourceLayer,_=b.nodeIndex,A=b.tensorIndex);let $=N.inboundNodes[_];if(C.indexOf($)!==-1)throw new Hr(`The tensor ${b.name} at layer "${N.name}" is part of a cycle.`);if(w.indexOf($)!==-1)return;this.containerNodes.add(zn.nodeKey(N,_)),N.id in i||(i[N.id]=Object.keys(i).length),C.indexOf($)===-1&&C.push($);let F=$.inboundLayers.length;for(let P=0;P<F;P++){let V=$.inputTensors[P],G=$.inboundLayers[P],W=$.nodeIndices[P],q=$.tensorIndices[P];u(V,w,C,G,W,q)}for(w.push($);C.indexOf($)>=0;)C.splice(C.indexOf($),1);a.push($)},l=[],c=[];for(let b of this.outputs)u(b,l,c);let p=a.slice().reverse();for(let b of p){n[b.id]=b,b.id in e||(e[b.id]=0);let w=e[b.id],C=o[b.outboundLayer.id]==null?0:o[b.outboundLayer.id];w=Math.max(w,C),o[b.outboundLayer.id]=w,s[b.outboundLayer.id]=b.outboundLayer,e[b.id]=w;for(let N=0;N<b.inboundLayers.length;N++){let _=b.inboundLayers[N],A=b.nodeIndices[N],$=_.inboundNodes[A],F=e[$.id]==null?0:e[$.id];e[$.id]=Math.max(w+1,F),n[$.id]=$}}let m={};for(let b in e){let w=e[b];w in m||(m[w]=[]),m[w].push(n[b])}let f={};for(let b in o){let w=o[b];w in f||(f[w]=[]),f[w].push(s[b])}let d=Object.keys(f).map(b=>parseInt(b,10)).sort(yh);this.layers=[];for(let b of d){let w=f[b];w.sort((C,N)=>{let _=i[C.id],A=i[N.id];return _<A?-1:_>A?1:0});for(let C of w)C instanceof zn&&this.internalContainerRefs.push(C),this.layers.push(C)}this.layersByDepth=f,d=Object.keys(m).map(b=>parseInt(b,10)).sort(yh);let h=this.inputs.slice(),g=[];for(let b of d)for(let w of m[b]){let C=w.outboundLayer;if(C!=null){for(let N of w.inputTensors)if(h.indexOf(N)===-1)throw new Hr(`Graph disconnected: cannot obtain value for tensor ${N} at layer "${C.name}". The following previous layers were accessed without issue: ${g}`);for(let N of w.outputTensors)h.push(N);g.push(C.name)}}this.nodesByDepth=m;let x=this.layers.map(b=>b.name);for(let b of x){let w=x.filter(C=>C===b).length;if(w!==1)throw new Hr(`The name "${b}" is used ${w} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(x))}this.outboundNodes=[],this.inboundNodes=[],new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(b=>null),outputMasks:this.outputs.map(b=>null),inputShapes:this.inputs.map(b=>b.shape),outputShapes:this.outputs.map(b=>b.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let t={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let e of this.layers)t.numDisposedVariables+=e.dispose().numDisposedVariables;for(let e of this.internalContainerRefs)t.numDisposedVariables+=e.dispose().numDisposedVariables}return t.refCountAfterDispose=this._refCount,t}get trainable(){return this.trainable_}set trainable(t){this.layers.forEach(e=>{e._trainableWeights.forEach(n=>n.trainable=t)}),this.trainable_=t}get trainableWeights(){if(this._trainableWeights.length>0)throw new M("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let t=[];for(let e of this.layers)t=t.concat(e.trainableWeights);return t}get nonTrainableWeights(){let t=[];for(let e of this.layers)t.push(...e.nonTrainableWeights);if(!this.trainable){let e=[];for(let n of this.layers)e.push(...n.trainableWeights);return e.concat(t)}return t}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(t,e=!0){let n={},o=0;for(let i of this.layers)for(let a of i.weights){if(n[a.originalName]!=null)throw new M(`Duplicate weight name: ${a.originalName}`);n[a.originalName]=a,o++}let s=[];for(let i in t){let a=i;if(n[i]==null){let u=i.split("/");a=u.slice(0,-2).concat([u[u.length-1]]).join("/")}if(n[a]!=null)s.push([n[a],t[i]]);else if(e)throw new M(`Provided weight data has no target variable: ${i}`);delete n[a]}if(e){let i=[];for(let a in n)i.push(a);if(i.length>0)throw new M(`${i.length} of ${o} weights are not set: ${i}`)}Pm(s)}updatedConfig(){let t=this.getConfig(),e={};return e.className=this.getClassName(),e.config=t,e.kerasVersion=`tfjs-layers ${Um}`,e.backend="TensorFlow.js",e}toJSON(t,e=!0){let n=qy(this.updatedConfig());return e?JSON.stringify(n):n}call(t,e){return B(()=>{t=xe(t);let n=new ko;for(let o=0;o<this.inputs.length;++o)n.add(this.inputs[o],t[o]);return hc(this.outputs,n,e)})}computeMask(t,e){return B(()=>{t=xe(t);let n;return e==null?n=Io(null,t.length):n=xe(e),this.runInternalGraph(t,n)[1]})}computeOutputShape(t){let e=Fm(t);if(e.length!==this.inputLayers.length)throw new M(`Invalid inputShape argument ${t}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let a=0;a<e.length;a++){let u=this.inputLayers[a],l=e[a],c=u.name+"_0_0";n[c]=l}let o=Object.keys(this.nodesByDepth).map(a=>parseInt(a,10)).sort(yh);if(o.length>1)for(let a of o){let u=this.nodesByDepth[a];for(let l of u){let c=l.outboundLayer;if(this.inputLayers.map(h=>h.id).indexOf(c.id)!==-1)continue;let p=[];for(let h=0;h<l.inboundLayers.length;h++){let g=l.inboundLayers[h],x=l.nodeIndices[h],b=l.tensorIndices[h],w=`${g.name}_${x}_${b}`,C=n[w];p.push(C)}let m=c.computeOutputShape(Nr(p)),f=Fm(m),d=c.inboundNodes.indexOf(l);for(let h=0;h<f.length;h++){let g=`${c.name}_${d}_${h}`;n[g]=f[h]}}}let s=[],i=[];for(let a=0;a<this.outputLayers.length;a++){let u=this.outputLayers[a],l=this.outputLayersNodeIndices[a],c=this.outputLayersTensorIndices[a],p=`${u.name}_${l}_${c}`;i.push(p)}for(let a=0;a<i.length;a++){let u=i[a];ro(u in n),s.push(n[u])}return Nr(s)}runInternalGraph(t,e){e==null&&(e=Io(null,t.length));let n={};for(let u=0;u<this.inputs.length;++u){let l=this.inputs[u],c=t[u],p=e[u];n[l.id]=[c,p]}let o=Object.keys(this.nodesByDepth).map(u=>parseInt(u,10)).sort(yh);for(let u of o){let l=this.nodesByDepth[u];for(let c of l){let p=c.outboundLayer,m=c.inputTensors,f=c.outputTensors,d=new Array;for(let h of m)h.id in n&&d.push(n[h.id]);if(d.length===m.length){let h={},g,x,b,w;if(c.callArgs!=null&&(h=c.callArgs),d.length===1){let[C,N]=d[0];h.mask==null&&(h.mask=N),b=xe(p.call(C,h)),w=xe(p.computeMask(C,N)),g=[C],x=[N]}else g=d.map(C=>C[0]),x=d.map(C=>C[1]),h.mask==null&&(h.mask=x),b=xe(p.call(g,h)),w=xe(p.computeMask(g,x));if(p.activityRegularizer)throw new St("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let C=0;C<f.length;++C){let N=f[C],_=b[C],A=w[C];n[N.id]=[_,A]}}}}let s=[],i=[],a=[];for(let u of this.outputs){ro(u.id in n,`Could not compute output ${u.name} : ${u.id}`);let[l,c]=n[u.id];a.push(l.shape),s.push(l),i.push(c)}return[s,i,a]}buildNodeConversionMap(t){let e={},n;for(let o of this.layers){n=o instanceof zn?1:0;for(let s=0;s<o.inboundNodes.length;s++){let i=zn.nodeKey(o,s);this.containerNodes.has(i)&&(e[i]=n,n+=1)}}return e}getLayer(t,e){if(e!=null){if(this.layers.length<=e)throw new M(`Was asked to retrieve layer at index ${e}, but model only has ${this.layers.length} layer(s).`);return this.layers[e]}else if(t==null)throw new M("Provide either a layer name or layer index");for(let n of this.layers)if(n.name===t)return n;throw new M(`No such layer: ${t}`)}calculateLosses(){return B(()=>{let t=[];for(let e of this.layers)for(let n=0;n<e.inboundNodes.length;++n){let o=zn.nodeKey(e,n);this.containerNodes.has(o)&&t.push(...e.calculateLosses())}return t})}getConfig(){let t={name:this.name},e=this.buildNodeConversionMap(this.layers),n=[];for(let i of this.layers){let a=i.getClassName(),u=i.getConfig(),l=[];for(let p=0;p<i.inboundNodes.length;p++){let m=i.inboundNodes[p],f=zn.nodeKey(i,p),d={};if(this.containerNodes.has(f)){if(m.callArgs)try{JSON.stringify(m.callArgs),d=m.callArgs}catch(h){console.warn(`Layer ${i.name} was passed non-serializable keyword arguments: ${m.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),d={}}if(m.inboundLayers.length>0){let h=[];for(let g=0;g<m.inboundLayers.length;g++){let x=m.inboundLayers[g],b=m.nodeIndices[g],w=m.tensorIndices[g],C=zn.nodeKey(x,b),N=e[C];N==null&&(N=0),h.push([x.name,N,w,d])}l.push(h)}}}let c={};c.name=i.name,c.className=a,c.config=u,c.inboundNodes=l,n.push(c)}t.layers=n;let o=[];for(let i=0;i<this.inputLayers.length;i++){let a=this.inputLayers[i],u=this.inputLayersNodeIndices[i],l=zn.nodeKey(a,u);if(!this.containerNodes.has(l))continue;let c=e[l];c==null&&(c=0);let p=this.inputLayersTensorIndices[i];o.push([a.name,c,p])}t.inputLayers=o;let s=[];for(let i=0;i<this.outputLayers.length;i++){let a=this.outputLayers[i],u=this.outputLayersNodeIndices[i],l=zn.nodeKey(a,u);if(!this.containerNodes.has(l))continue;let c=e[l];c==null&&(c=0);let p=this.outputLayersTensorIndices[i];s.push([a.name,c,p])}return t.outputLayers=s,t}static fromConfig(t,e,n={},o=!1){let s={},i={};function a(g,x){g.name in i?i[g.name].push(x):i[g.name]=[x]}function u(g,x){let b=[],w;for(let C of x){let N=C[0],_=C[1],A=C[2];if(w=C[3]==null?{}:C[3],!(N in s)){a(g,x);return}let $=s[N];if($.inboundNodes.length<=_){a(g,x);return}let F=$.inboundNodes[_];b.push(F.outputTensors[A])}b.length>0&&g.apply(Nr(b),w)}function l(g){let x=g.name,b=gn(g,e.customObjects!=null?e.customObjects:{});b.setFastWeightInitDuringBuild(o),s[x]=b,g.inboundNodes.forEach(C=>{if(!(C instanceof Array))throw new M(`Corrupted configuration, expected array for nodeData: ${C}`);a(b,C)})}let c=e.name,p=e.layers;for(let g of p)l(g);for(;!b$(i);)for(let g of p){let x=s[g.name];if(x.name in i){let b=i[x.name];delete i[x.name];for(let w of b)u(x,w)}}let m=[],f=[],d=e.inputLayers;for(let g of d){let x=g[0],b=g[1],w=g[2];ro(x in s);let N=s[x].inboundNodes[b].outputTensors;m.push(N[w])}let h=e.outputLayers;for(let g of h){let x=g[0],b=g[1],w=g[2];ro(x in s);let N=s[x].inboundNodes[b].outputTensors;f.push(N[w])}return new t({inputs:m,outputs:f,name:c})}get stateful(){if(this._stateful)throw new M("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let t of this.layers)if(t.stateful)return!0;return!1}resetStates(){B(()=>{this.layers.forEach(t=>{t.stateful&&t.resetStates()})})}};function R8(r,t,e){let n=t.length;if(r==null||Array.isArray(r)&&r.length===0)return t.map(o=>null);if(n===1)return Array.isArray(r)&&r.length===1?r:typeof r=="object"&&t[0]in r?[r[t[0]]]:[r];if(Array.isArray(r)){if(r.length!==n)throw new Error(`Provided ${e} is an array of ${r.length} element(s), but the model has ${n} outputs. Make sure a set of weights is provided for each model output.`);return r}else if(typeof r=="object"&&Object.keys(r).length>0&&typeof r[Object.keys(r)[0]]=="object"){let o=[];return t.forEach(s=>{s in r?o.push(r[s]):o.push(null)}),o}else throw new Error(`The model has multiple (${n}) outputs, so ${e} must be either an array with ${n} elements or an object with ${t} keys. Provided ${e} not understood: ${JSON.stringify(r)}`)}function Ky(r,t){return R8(r,t,"classWeight")}async function jy(r,t,e,n){if(t!=null||n!=null)throw new Error("Support sampleWeight is not implemented yet");if(e!=null){let o=B(()=>{if(r.shape.length===1)return sn(r);if(r.shape.length===2){if(r.shape[1]>1)return Ai(r,1);if(r.shape[1]===1)return R(r,[r.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${r.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${r.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),s=Array.from(await o.data());vt(o);let i=[];return s.forEach(a=>{if(e[a]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${a} exists in the data but not in classWeight`);i.push(e[a])}),Me(i,"float32")}else return null}function oD(r,t){return D(r,t)}var F8=32;function aD(r,t){let e,n,o=t;e=o.xs,n=o.ys,y.assert(e!=null&&n!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let s=sD("input",r.inputNames,e),i=sD("output",r.outputNames,n),a=s[0].shape[0];y.assert(s.length===r.inputs.length,()=>`LayersModel has ${r.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(r.inputNames)})`),y.assert(i.length===r.outputs.length,()=>`LayersModel has ${r.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(r.outputNames)})`);for(let u=0;u<s.length;u++)y.assert(s[u].shape[0]===a,()=>`Batch size mismatch: input ${r.inputNames[u]} has ${s[u].shape[0]}; expected  ${a} based on input ${r.inputNames[0]}.`);for(let u=0;u<i.length;u++)y.assert(i[u].shape[0]===a,()=>`Batch size mismatch: output ${r.outputNames[u]} has ${i[u].shape[0]}; expected  ${a} based on input ${r.inputNames[0]}.`);return{xs:s,ys:i}}function sD(r,t,e){if(e instanceof Ft)return[e];if(Array.isArray(e))return y.assert(e.length===t.length,()=>`Received an array of ${e.length} Tensors, but expected ${t.length} to match the ${r} keys ${t}.`),e;{let n=[];for(let o of t){if(e[o]==null)throw new M(`The feature data generated by the dataset lacks the required ${r} key '${o}'.`);n.push(e[o])}return n}}function O8(r){if(r.length===3)throw new St("Validation with sample weights is not implemented yet.");return{xs:r[0],ys:r[1]}}async function lD(r,t,e){let n=e.batchesPerEpoch!=null;if(y.assert(r.optimizer!=null,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),y.assert(e!=null,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),y.assert(e.epochs!=null&&e.epochs>0&&Number.isInteger(e.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${e.epochs}`),y.assert(!n||e.batchesPerEpoch>0&&Number.isInteger(e.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${e.batchesPerEpoch}`),y.assert(e.validationSplit==null,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),r.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");r.isTraining=!0;try{let o=e.validationData!=null,s,i;if(o)if(iD(e.validationData))y.assert(e.validationBatches==null||e.validationBatches>0&&Number.isInteger(e.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${e.validationBatches}`);else{let g=O8(e.validationData);s=g.xs,i=g.ys}let a=r.makeTrainFunction(),u=r.getDedupedMetricsNames(),l;o?l=u.slice().concat(u.map(g=>"val_"+g)):l=u.slice();let c=zy(e.callbacks,e.yieldEvery),p=e.verbose==null?1:e.verbose,{callbackList:m,history:f}=By(c,p,e.epochs,null,null,P8(t,e),null,o,l);m.setModel(r),r.history=f,await m.onTrainBegin(),r.stopTraining_=!1;let d=e.initialEpoch==null?0:e.initialEpoch,h=await t.iterator();for(;d<e.epochs;){let g={};await m.onEpochBegin(d);let x=0,b=0;for(n||(h=await t.iterator());!n||x<e.batchesPerEpoch;){let w=await h.next();if(n&&w.done){console.warn(`You provided \`batchesPerEpoch\` as ${e.batchesPerEpoch}, but your dataset iterator ran out of data after ${x} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${e.batchesPerEpoch*e.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(w.value!=null){let{xs:C,ys:N}=aD(r,w.value),_={};_.batch=b,_.size=C[0].shape[0],await m.onBatchBegin(b,_);let A=[];if(e.classWeight!=null){let P=Ky(e.classWeight,r.outputNames);for(let V=0;V<P.length;++V)A.push(await jy(N[V],null,P[V]))}let $=C.concat(N).concat(A),F=a($);vt($);for(let P=0;P<u.length;++P){let V=u[P],G=F[P];_[V]=G,De(G)}await m.onBatchEnd(b,_),Oy(_),b++,x++}if(n?x>=e.batchesPerEpoch:w.done){if(o){let C;iD(e.validationData)?C=xe(await r.evaluateDataset(e.validationData,{batches:e.validationBatches})):C=xe(r.evaluate(s,i,{batchSize:e.validationBatchSize==null?F8:e.validationBatchSize,verbose:0}));for(let N=0;N<r.metricsNames.length;++N)g[`val_${r.metricsNames[N]}`]=C[N]}break}if(r.stopTraining_)break}if(await m.onEpochEnd(d,g),d++,r.stopTraining_)break}return await m.onTrainEnd(),await r.history.syncData(),r.history}finally{r.isTraining=!1}}function P8(r,t){let e=null;return t.batchesPerEpoch!=null?e=t.batchesPerEpoch:Number.isFinite(r.size)&&(e=r.size),e}function iD(r){return typeof r.iterator=="function"}function L8(r){return typeof r.next=="function"}async function uD(r,t,e){e=e||{};let n=e.batches!=null,o=r.testFunction,s=[];if(e.verbose>0)throw new St("Verbose mode is not implemented yet.");y.assert(!n||e.batches>0&&Number.isInteger(e.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(e.batches)}`);let i=L8(t)?t:await t.iterator(),a=0,u=0;for(;!n||u<e.batches;){let l=await i.next();if(s=B(()=>{if(l.value){let{xs:c,ys:p}=aD(r,l.value),m=c.concat(p),f=B(()=>o(m));if(vt(m),u===0)for(let h=0;h<f.length;++h)s.push(mt(0));let d=m[0].shape[0];for(let h=0;h<f.length;++h){let g=f[h],x=s[h];s[h]=B(()=>X(s[h],D(d,g))),u>0&&vt(x)}vt(f),a+=d,++u}return s}),l.done){n&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${e.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let l=0;l<s.length;++l){let c=s[l];s[l]=pt(s[l],a),vt(c)}return Nr(s)}function Xy(r){y.assert(r>0&&Number.isInteger(r),()=>`batchSize is required to be a positive integer, but got ${r}`)}function Hm(r,t,e){return r==null?[null]:Array.isArray(r)?r.map(n=>rl(n,t,e-t)):rl(r,t,e-t)}function Yy(r,t){return B(()=>r==null?null:Array.isArray(r)?r.map(e=>Yy(e,t)):_y(r,t.dtype==="int32"?t:J(t,"int32")))}function Zy(r,t){let e=[],n=0,o=null;for(;n<r;)o=n+t,o>=r&&(o=r),e.push([n,o]),n=o;return e}async function M8(r,t,e,n,o,s,i,a,u,l,c,p,m,f,d){o==null&&(o=32),s==null&&(s=1),c==null&&(c=!0),m==null&&(m=0);let h=!1;if(u!=null&&l!=null&&(h=!0),d!=null&&(h=!0,f==null))throw new M("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let g=r.checkNumSamples(e,o,f,"steps_per_epoch"),x;g!=null&&(x=Zr(0,g)),i==null&&(i=1);let{callbackList:b,history:w}=By(a,i,s,m,g,f,o,h,p);b.setModel(r),r.history=w,await b.onTrainBegin(),r.stopTraining_=!1;for(let C=m;C<s;++C){await b.onEpochBegin(C);let N={};if(f!=null)throw new St("stepsPerEpoch mode is not implemented yet.");{if(c==="batch")throw new St("batch shuffling is not implemneted yet");c&&y.shuffle(x);let _=Me(x),A=Zy(g,o);for(let $=0;$<A.length;++$){let F={};if(await b.onBatchBegin($,F),B(()=>{let P=A[$][0],V=A[$][1],G=rl(_,P,V-P);F.batch=$,F.size=V-P;let W=Yy(e,G),q=t(W);for(let H=0;H<n.length;++H){let j=n[H],Y=q[H];F[j]=Y,De(Y)}if($===A.length-1&&h){let H=r.testLoop(u,l,o);for(let j=0;j<n.length;++j){let Y=n[j],Z=H[j];De(Z),N["val_"+Y]=Z}}}),await b.onBatchEnd($,F),Oy(F),r.stopTraining_)break}_.dispose()}if(await b.onEpochEnd(C,N),r.stopTraining_)break}return await b.onTrainEnd(),await r.history.syncData(),r.history}async function cD(r,t,e,n={}){if(r.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");r.isTraining=!0;let o,s,i,a,u,l,c,p,m;try{let f=n.batchSize==null?32:n.batchSize;Xy(f);let d=!1,h=await r.standardizeUserData(t,e,n.sampleWeight,n.classWeight,d,f);o=h[0],s=h[1],m=h[2];let g=!1,x;if(n.validationData!=null&&n.validationData.length>0){if(g=!0,n.validationData.length===2)u=n.validationData[0],l=n.validationData[1];else throw n.validationData.length===3?new St("validationData including sample weights is not supported yet."):new M(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${n.validationData} is invalid.`);let F=!0,P=await r.standardizeUserData(u,l,null,null,F,f);c=P[0],p=P[1],x=c.concat(p)}else if(n.validationSplit!=null&&n.validationSplit>0&&n.validationSplit<1){g=!0;let F=Math.floor(o[0].shape[0]*(1-n.validationSplit)),P=o[0].shape[0];c=Hm(o,F,P),i=o,o=Hm(o,0,F),p=Hm(s,F,P),a=s,s=Hm(s,0,F),x=c.concat(p)}else n.validationSteps!=null&&(g=!0);let b=o.concat(s).concat(m);r.checkTrainableWeightsConsistency();let w=r.makeTrainFunction(),C=r.getDedupedMetricsNames(),N,_;g?(r.makeTestFunction(),N=r.testFunction,_=C.slice().concat(C.map(F=>"val_"+F))):(N=null,x=[],_=C.slice());let A=zy(n.callbacks,n.yieldEvery);return await M8(r,w,b,C,f,n.epochs,n.verbose,A,N,x,n.shuffle,_,n.initialEpoch,null,null)}finally{r.isTraining=!1,Eo(o,t),Eo(s,e),Eo(i,t),Eo(a,e),Eo(c,u),Eo(p,l),m!=null&&vt(m)}}function Fv(r){let t=[];r instanceof Ft&&(r=[r]);for(let e=0;e<r.length;++e){let n=r[e];if(n.rank===1)t.push(nl(n,1));else{if(n.rank===0)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(n)}}return t}function Eo(r,t){if(r==null)return;let e=[];if(t instanceof Ft)e.push(t.id);else if(Array.isArray(t))t.forEach(o=>e.push(o.id));else if(t!=null)for(let o in t){let s=t[o];e.push(s.id)}let n=[];if(r instanceof Ft)e.indexOf(r.id)===-1&&n.push(r);else if(Array.isArray(r))r.forEach(o=>{e.indexOf(o.id)===-1&&n.push(o)});else if(r!=null)for(let o in r){let s=r[o];e.indexOf(s.id)===-1&&n.push(s)}n.forEach(o=>{o.isDisposed||o.dispose()})}function z8(r){return r instanceof Ft}function Ov(r){return Array.isArray(r)}function pD(r){return!z8(r)&&!Ov(r)}function mD(r,t,e,n=!0,o=""){if(t==null||t.length===0){if(r!=null){let i=!1;if(Ov(r)&&r.length>0)i=!0;else if(pD(r)){for(let a in r)if(r.hasOwnProperty(a)){i=!0;break}}else i=!0;if(i)throw new M(`Error when checking model ${o} expected no data, but got ${r}`)}return[]}if(r==null)return t.map(i=>null);let s;if(pD(r)){r=r,s=[];for(let i of t){if(r[i]==null)throw new M(`No data provided for "${i}". Need data for each key in: ${t}`);s.push(r[i])}}else if(Ov(r)){if(r=r,r.length!==t.length)throw new M(`Error when checking model ${o}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${r}`);s=r}else{if(r=r,t.length>1)throw new M(`The model ${o} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${r.shape}`);s=[r]}if(s=Fv(s),e!=null)for(let i=0;i<t.length;++i){if(e[i]==null)continue;let a=s[i];if(a.shape.length!==e[i].length)throw new M(`Error when checking ${o}: expected ${t[i]} to have ${e[i].length} dimension(s). but got array with shape ${a.shape}`);for(let u=0;u<e[i].length;++u){if(u===0&&!n)continue;let l=a.shape[u],c=e[i][u];if(c!=null&&c>=0&&l!==c)throw new M(`${o} expected a batch of elements where each example has shape [${e[i].slice(1,e[i].length)}] (i.e.,tensor shape [*,${e[i].slice(1,e[i].length)}]) but the ${o} received an input with ${a.shape[0]} examples, each with shape [${a.shape.slice(1,a.shape.length)}] (tensor shape [${a.shape}])`)}}return s}function B8(r,t,e){let n=vo(r.map(s=>s.shape[0]));n.sort();let o=vo(t.map(s=>s.shape[0]));if(o.sort(),n.length>1)throw new M(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(r.map(s=>s.shape))}`);if(o.length>1)throw new M(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(s=>s.shape))}`);if(n.length>0&&o.length>0&&!y.arraysEqual(n,o))throw new M(`Input Tensors should have the same number of samples as target Tensors. Found ${n[0]} input sample(s) and ${o[0]} target sample(s).`)}function V8(r,t,e){let n=[Hi,Wm,xc];for(let o=0;o<r.length;++o){let s=r[o],i=t[o],a=e[o];if(i!=null){if(i===xc&&s.shape[s.shape.length-1]===1)throw new M(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(n.indexOf(i)!==-1){let u=s.shape.slice(1),l=a.slice(1);for(let c=0;c<u.length;++c){let p=u[c],m=l[c];if(m!=null&&p!==m)throw new M(`A target Tensor with shape ${s.shape} was passed for an output of shape ${a}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function fD(r,t,e,n=!0,o=""){let s;if(Array.isArray(r)){if(r.length!==t.length)throw new M(`Error when checking model ${o}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${r.length} Tensors(s).`);s=r}else{if(t.length>1)throw new M(`The model expects ${t.length} ${o} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(r.shape)}.`);s=[r]}if(e!=null)for(let i=0;i<t.length;++i){if(e[i]==null)continue;let a=s[i];if(a.shape.length!==e[i].length)throw new M(`Error when checking ${o}: expected ${t[i]} to have ${e[i].length} dimension(s), but got array with shape ${JSON.stringify(a.shape)}`);for(let u=0;u<e[i].length;++u){if(u===0&&!n)continue;let l=a.shape[u],c=e[i][u];if(c!=null&&c!==l)throw new M(`Error when checking ${o}: expected ${t[i]} to have shape ${JSON.stringify(e[i])} but got array with shape ${JSON.stringify(a.shape)}.`)}}}function G8(r,t){if(r==null||Array.isArray(r)&&r.length===0)return t.map(n=>[]);let e;if(typeof r=="string"||typeof r=="function")e=[r];else if(Array.isArray(r)||typeof r=="object")e=r;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${r}`);if(Array.isArray(e))return t.map(n=>e);{let n=[];for(let o of t){let s=e.hasOwnProperty(o)?e[o]:[];Array.isArray(s)||(s=[s]),n.push(s)}return n}}var W8="layers-model",Bn=class extends zn{constructor(t){super(t),this.isTraining=!1}summary(t,e,n=console.log){if(!this.built)throw new M("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");rD(this,t,e,n)}compile(t){if(t.loss==null&&(t.loss=[]),this.loss=t.loss,typeof t.optimizer=="string")this.optimizer_=eD(t.optimizer),this.isOptimizerOwned=!0;else{if(!(t.optimizer instanceof Wr))throw new M("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=t.optimizer,this.isOptimizerOwned=!1}let e=[];if(!Array.isArray(t.loss)&&typeof t.loss!="string"&&typeof t.loss!="function"){t.loss=t.loss;for(let i in t.loss)if(this.outputNames.indexOf(i)===-1)throw new M(`Unknown entry in loss dictionary: "${i}". Only expected the following keys: ${this.outputNames}`);for(let i of this.outputNames)t.loss[i]==null&&console.warn(`Output "${i}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${i} during training`),e.push(Vy(t.loss[i]))}else if(Array.isArray(t.loss)){if(t.loss.length!==this.outputs.length)throw new M(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${t.loss}.`);e=t.loss.map(a=>Vy(a))}else{let i=Vy(t.loss);this.outputs.forEach(a=>{e.push(i)})}this.lossFunctions=e,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let i=0;i<this.outputs.length;++i){let a=this.internalOutputShapes[i],u=this.outputNames[i];this.feedOutputNames.push(u),this.feedOutputShapes.push(a),this.feedLossFns.push(this.lossFunctions[i])}let n=[];this.metrics=t.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Hs("loss",()=>{for(let i=0;i<this.outputs.length;++i){if(n.indexOf(i)!==-1)continue;let a=this.lossFunctions[i];this.outputs.length>1&&(this.metricsTensors.push([a,i]),this.metricsNames.push(this.outputNames[i]+"_loss"))}});let o=G8(t.metrics,this.outputNames),s=(i,a,u)=>{this.outputNames.length>1&&(a=this.outputNames[i]+"_"+a),this.metricsNames.push(a),this.metricsTensors.push([u,i])};Hs("metric",()=>{for(let i=0;i<this.outputs.length;++i){if(n.indexOf(i)!==-1)continue;let a=o[i];(l=>{let c="",p,m,f;for(let d of l){if(typeof d=="string"&&["accuracy","acc","crossentropy","ce"].indexOf(d)!==-1){let g=this.internalOutputShapes[i];g[g.length-1]===1||this.lossFunctions[i]===Wm?["accuracy","acc"].indexOf(d)!==-1?m=Th:["crossentropy","ce"].indexOf(d)!==-1&&(m=Wy):this.lossFunctions[i]===Gm?["accuracy","acc"].indexOf(d)!==-1?m=Uy:["crossentropy","ce"].indexOf(d)!==-1&&(m=$v):["accuracy","acc"].indexOf(d)!==-1?m=kh:["crossentropy","ce"].indexOf(d)!==-1&&(m=Eh);let x;["accuracy","acc"].indexOf(d)!==-1?x="acc":["crossentropy","ce"].indexOf(d)!==-1&&(x="ce"),f=m,p=c+x}else f=Q$(d),p=c+_h(d);let h;Hs(p,()=>{h=f}),s(i,p,h)}})(a)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(t,e,n={}){let o=n.batchSize==null?32:n.batchSize;Xy(o);let s=!0,i=this.standardizeUserDataXY(t,e,s,o);try{let a=i[0].concat(i[1]);this.makeTestFunction();let u=this.testFunction,l=this.testLoop(u,a,o,n.verbose,n.steps);return Nr(l)}finally{Eo(i[0],t),Eo(i[1],e)}}async evaluateDataset(t,e){return this.makeTestFunction(),uD(this,t,e)}checkNumSamples(t,e,n,o="steps"){let s;if(n!=null){if(s=null,e!=null)throw new M(`If ${o} is set, batchSize must be null or undefined.Got batchSize = ${e}`)}else if(t!=null)Array.isArray(t)?s=t[0].shape[0]:s=t.shape[0];else throw new M(`Either the input data should have a defined shape, or ${o} shoud be specified.`);return s}execute(t,e){if(Array.isArray(e)&&e.length===0)throw new M("`outputs` is an empty Array, which is not allowed.");let n=Array.isArray(e),o=n?e:[e],s=this.retrieveSymbolicTensors(o),i=new ko;if(t instanceof Ft&&(t=[t]),Array.isArray(t)){if(t.length!==this.inputs.length)throw new M(`The number of inputs provided (${t.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let u=0;u<this.inputs.length;++u)i.add(this.inputs[u],t[u])}else for(let u of this.inputs){let l=t[u.name];if(l==null)throw new M(`No value is provided for the model's input ${u.name}`);i.add(u,l)}let a=hc(s,i);return n?a:a[0]}retrieveSymbolicTensors(t){let e=Io(null,t.length),n=t.length;for(let o of this.layers){let s=Array.isArray(o.output)?o.output:[o.output],i=s.map(a=>a.name);for(let a=0;a<t.length;++a){let u=i.indexOf(t[a]);if(u!==-1&&(e[a]=s[u],n--),n===0)break}if(n===0)break}if(n>0){let o=[];throw e.forEach((s,i)=>{s==null&&o.push(t[i])}),new M(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(o)}`)}return e}predictLoop(t,e=32,n=!1){return B(()=>{let o=this.checkNumSamples(t);if(n)throw new St("Verbose predictLoop() is not implemented yet.");let s=Zy(o,e),i=this.outputs.map(a=>[]);for(let a=0;a<s.length;++a)B(()=>{let l=s[a][0],c=s[a][1],p=Hm(t,l,c),m=[];if(Array.isArray(p))for(let d=0;d<p.length;++d)m.push({key:this.inputs[d],value:p[d]});else m.push({key:this.inputs[0],value:p});let f=new ko(m);return hc(this.outputs,f)}).forEach((l,c)=>i[c].push(l));return Nr(i.map(a=>ne(a,0)))})}predict(t,e={}){let n=Fv(t);fD(n,this.inputNames,this.feedInputShapes,!1);try{let o=e.batchSize==null?32:e.batchSize;return Xy(o),this.predictLoop(n,o)}finally{Eo(n,t)}}predictOnBatch(t){fD(t,this.inputNames,this.feedInputShapes,!0);let e=(Array.isArray(t)?t[0]:t).shape[0];return this.predictLoop(t,e)}standardizeUserDataXY(t,e,n=!0,o){if(this.optimizer_==null)throw new Hr("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let s=[];for(let i=0;i<this.feedOutputShapes.length;++i){let a=this.feedOutputShapes[i];this.feedLossFns[i]===Gm?s.push(a.slice(0,a.length-1).concat([1])):s.push(a)}if(t=mD(t,this.feedInputNames,this.feedInputShapes,!1,"input"),e=mD(e,this.feedOutputNames,s,!1,"target"),B8(t,e,null),V8(e,this.feedLossFns,this.feedOutputShapes),this.stateful&&o!=null&&o>0&&t[0].shape[0]%o!==0)throw new M(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${o}. Found: ${t[0].shape[0]} sample(s).`);return[t,e]}async standardizeUserData(t,e,n,o,s=!0,i){let[a,u]=this.standardizeUserDataXY(t,e,s,i);if(n!=null)throw new Error("sample weight is not supported yet.");let l=null;if(o!=null){let c=Ky(o,this.outputNames);l=[];for(let p=0;p<c.length;++p)l.push(await jy(u[p],null,c[p]))}return[a,u,l]}testLoop(t,e,n,o=0,s){return B(()=>{let i=this.checkNumSamples(e,n,s,"steps"),a=[];if(o>0)throw new St("Verbose mode is not implemented yet.");if(s!=null)throw new St("steps mode in testLoop() is not implemented yet");{let u=Zy(i,n),l=Me(Zr(0,i));for(let c=0;c<u.length;++c){let p=u[c][0],m=u[c][1],f=rl(l,p,m-p),d=Yy(e,f),h=t(d);if(c===0)for(let g=0;g<h.length;++g)a.push(mt(0));for(let g=0;g<h.length;++g){let x=h[g];a[g]=X(a[g],D(m-p,x))}}for(let c=0;c<a.length;++c)a[c]=pt(a[c],i)}return a})}getDedupedMetricsNames(){let t=this.metricsNames,e=[];for(let n=0;n<t.length;++n){let o=t[n],s=o;Cv(t,o)>1&&(s+=`_${Cv(t.slice(0,n),o)}`),e.push(s)}return e}makeTrainFunction(){return t=>{let e=[],n=t.slice(0,this.inputs.length),o=t.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=t.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),i=[],a=()=>{let p=[];for(let h=0;h<this.inputs.length;++h)p.push({key:this.inputs[h],value:n[h]});let m=new ko(p),f=hc(this.outputs,m,{training:!0}),d;for(let h=0;h<this.lossFunctions.length;++h){let g=this.lossFunctions[h],x=g(o[h],f[h]);s[h]!=null&&(x=oD(x,s[h]));let b=ve(x);e.push(b),h===0?d=x:d=X(d,x)}for(let h=0;h<this.metricsTensors.length;++h){let g;if(this.outputs.length>1&&h<this.outputs.length)g=e[h];else{let x=this.metricsTensors[h][0],b=this.metricsTensors[h][1];g=ve(x(o[b],f[b]))}De(g),i.push(g)}return d=ve(d),this.calculateLosses().forEach(h=>{d=X(d,h)}),d},u=this.collectedTrainableWeights.map(p=>p.read()),l=!0;return[this.optimizer_.minimize(a,l,u)].concat(i)}}makeTestFunction(){this.testFunction=t=>B(()=>{let e=[],n,o=t.slice(0,this.inputs.length),s=t.slice(this.inputs.length,this.inputs.length+this.outputs.length),i=[];for(let l=0;l<this.inputs.length;++l)i.push({key:this.inputs[l],value:o[l]});let a=new ko(i),u=hc(this.outputs,a);for(let l=0;l<this.lossFunctions.length;++l){let c=this.lossFunctions[l],p=ve(c(s[l],u[l]));l===0?n=p:n=X(n,p),e.push(n)}for(let l=0;l<this.metricsTensors.length;++l){let c=this.metricsTensors[l][0],p=this.metricsTensors[l][1],m=ve(c(s[p],u[p]));e.push(m)}return e})}async fit(t,e,n={}){return cD(this,t,e,n)}async fitDataset(t,e){return lD(this,t,e)}async trainOnBatch(t,e){let n=await this.standardizeUserData(t,e),o=n[0],s=n[1],a=this.makeTrainFunction()(o.concat(s)),u=[];for(let l of a){let c=await l.data();u.push(c[0])}return vt(a),Eo(n[0],t),Eo(n[1],e),Nr(u)}getNamedWeights(t){let e=[],n=t!=null&&t.trainableOnly,o=n?this.trainableWeights:this.weights,s=this.getWeights(n);for(let i=0;i<o.length;++i)n&&!o[i].trainable||e.push({name:o[i].originalName,tensor:s[i]});return e}set stopTraining(t){this.stopTraining_=t}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(t){this.optimizer_!==t&&(this.optimizer_=t,this.isOptimizerOwned=!1)}dispose(){let t=super.dispose();if(t.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let e=ah().numTensors;this.optimizer_.dispose(),t.numDisposedVariables+=e-ah().numTensors}return t}getLossIdentifiers(){let t;if(typeof this.loss=="string")t=So(this.loss);else if(Array.isArray(this.loss)){for(let e of this.loss)if(typeof e!="string")throw new Error("Serialization of non-string loss is not supported.");t=this.loss.map(e=>So(e))}else{let e=Object.keys(this.loss);t={};let n=this.loss;for(let o of e)if(typeof n[o]=="string")t[o]=So(n[o]);else throw new Error("Serialization of non-string loss is not supported.")}return t}getMetricIdentifiers(){if(typeof this.metrics=="string"||typeof this.metrics=="function")return[So(_h(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(t=>So(_h(t)));{let t={};for(let e in this.metrics)t[e]=So(_h(this.metrics[e]));return t}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(t){if(t.weighted_metrics!=null)throw new Error("Loading weight_metrics is not supported yet.");if(t.loss_weights!=null)throw new Error("Loading loss_weights is not supported yet.");if(t.sample_weight_mode!=null)throw new Error("Loading sample_weight_mode is not supported yet.");let e=yc(t.optimizer_config),n=gn(e),o;if(typeof t.loss=="string")o=el(t.loss);else if(Array.isArray(t.loss))o=t.loss.map(i=>el(i));else if(t.loss!=null){o={};for(let i in t.loss)o[i]=el(t.loss[i])}let s;if(Array.isArray(t.metrics))s=t.metrics.map(i=>el(i));else if(t.metrics!=null){s={};for(let i in t.metrics)s[i]=el(t.metrics[i])}this.compile({loss:o,metrics:s,optimizer:n})}async save(t,e){if(typeof t=="string"){let l=_r.getSaveHandlers(t);if(l.length===0)throw new M(`Cannot find any save handlers for URL '${t}'`);if(l.length>1)throw new M(`Found more than one (${l.length}) save handlers for URL '${t}'`);t=l[0]}if(t.save==null)throw new M("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let n=await _r.encodeWeights(this.getNamedWeights(e)),o=!1,s=null,a={modelTopology:this.toJSON(s,o),format:W8,generatedBy:`TensorFlow.js tfjs-layers v${Um}`,convertedBy:null};if((e==null?!1:e.includeOptimizer)&&this.optimizer!=null){a.trainingConfig=this.getTrainingConfig();let l="optimizer",{data:c,specs:p}=await _r.encodeWeights(await this.optimizer.getWeights(),l);n.specs.push(...p),n.data=_r.concatenateArrayBuffers([n.data,c])}return this.userDefinedMetadata!=null&&(Rv(this.userDefinedMetadata,this.name,!0),a.userDefinedMetadata=this.userDefinedMetadata),a.weightData=n.data,a.weightSpecs=n.specs,t.save(a)}setUserDefinedMetadata(t){Rv(t,this.name),this.userDefinedMetadata=t}getUserDefinedMetadata(){return this.userDefinedMetadata}};Bn.className="Model";Q.registerClass(Bn);var Jy=class extends Bn{};Jy.className="Functional";Q.registerClass(Jy);async function dD(r,t){"modelTopology"in r||(r={modelTopology:r}),r=r;let e=r.modelTopology;e.model_config!=null&&(e=e.model_config);let n=yc(e),o=gn(n,t);if(r.weightsManifest!=null){let s=await _r.loadWeights(r.weightsManifest,r.pathPrefix,o.weights.map(a=>a.originalName)),i={};for(let a of o.weights)i[a.originalName]=s[a.originalName];o.loadWeights(i),vt(s)}return o}async function hD(r,t){if(t==null&&(t={}),typeof r=="string"){let e=_r.getLoadHandlers(r,t);if(e.length===0)e.push(_r.browserHTTPRequest(r,t));else if(e.length>1)throw new M(`Found more than one (${e.length}) load handlers for URL '${r}'`);r=e[0]}return U8(r,void 0,t)}async function U8(r,t,e){if(e==null&&(e={}),r.load==null)throw new M("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let n=await r.load(),o=n.modelTopology;o.model_config!=null&&(o=o.model_config);let s=e.strict==null?!0:e.strict,i=n.weightData!=null&&n.weightSpecs!=null&&s,a=gn(yc(o),t,i),u=n.trainingConfig;if(u!=null&&a.loadTrainingConfig(u),n.userDefinedMetadata!=null&&a.setUserDefinedMetadata(n.userDefinedMetadata),n.weightData!=null){if(n.weightSpecs==null)throw new M("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:l,optimizerWeights:c}=H8(n.weightData,n.weightSpecs);a.loadWeights(l,s),a.optimizer!=null&&c.length>0&&await a.optimizer.setWeights(c),vt(l),vt(c.map(p=>p.tensor))}return a}function H8(r,t){let e=_r.decodeWeights(r,t),n={},o=[];return t.forEach(s=>{s.group==="optimizer"?o.push({name:s.name,tensor:e[s.name]}):n[s.name]=e[s.name]}),{modelWeights:n,optimizerWeights:o}}var qi=class extends Bn{constructor(t){if(super({inputs:[],outputs:[]}),t=t||{},this.trainable=!0,this.built=!1,this.name=t.name!=null?t.name:gu("sequential_"),t.layers!=null)for(let e of t.layers)this.add(e)}checkShape(t){if(t.inboundNodes[0].outputTensors[0].shape.some(n=>n<0))throw new M(`Negative dimension size caused by adding layer ${t.name} with input shape [${t.inboundNodes[0].inputTensors[0].shape}]`)}add(t){let e=t instanceof qi||t instanceof Bn,n;if(e){if(n=t,n.outputs.length!==1)throw new M("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(n.inputs.length!==1)throw new M("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(this.outputs.length===0){if(t.inboundNodes.length===0){if(t.batchInputShape==null)throw new M("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let o=Dy({batchShape:t.batchInputShape,dtype:t.dtype,name:t.name+"_input"});t.apply(o)}if(e)this.outputs=n.outputs,this.inputs=n.inputs;else{if(t.inboundNodes.length!==1)throw new M(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${t.name} which has ${t.inboundNodes.length} pre-existing inbound connections.`);if(t.inboundNodes[0].outputTensors.length!==1)throw new M("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(t),this.outputs=[t.inboundNodes[0].outputTensors[0]],this.inputs=kv(this.outputs[0])}this.inboundNodes=[],new ol({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:Io(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(o=>o.shape),outputShapes:this.outputs[0].shape})}else{let o=t.apply(this.outputs[0]);if(Array.isArray(o))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(t),this.outputs=[o],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(t),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let t=this.layers.length-1;this.layers[t].outboundNodes=[],this.outputs=[this.layers[t].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(t,e){return this.model==null&&this.build(),this.model.call(t,e)}build(t){if(Bt(t),this.inputs.length===0||this.outputs.length===0)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Bn({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(t,e,n=console.log){this.built||this.build(),super.summary(t,e,n)}setWeights(t){this.model==null&&this.build(),this.model.setWeights(t)}evaluate(t,e,n={}){if(!this.built)throw new Hr("The model needs to be compiled before being used.");return this.model.evaluate(t,e,n)}async evaluateDataset(t,e){if(!this.built)throw new Hr("The model needs to be compiled before being used.");return this.model.evaluateDataset(t,e)}predict(t,e={}){return this.model==null&&this.build(),this.model.predict(t,e)}predictOnBatch(t){return this.model==null&&this.build(),this.model.predictOnBatch(t)}compile(t){this.build(),this.model.compile(t),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(t){this.model.optimizer=t}async fit(t,e,n={}){if(!this.built)throw new Hr("The model needs to be compiled before being used.");return this.model.fit(t,e,n)}async fitDataset(t,e){if(!this.built)throw new Hr("The model needs to be compiled before being used.");return this.model.fitDataset(t,e)}async trainOnBatch(t,e){return this.model.trainOnBatch(t,e)}static fromConfig(t,e,n={},o=!1){let s,i={};if(e instanceof Array){if(e[0].className==null||e[0].className==="Merge")throw new M("Legacy serialization format not supported yet.");s=e}else y.assert(e.layers!=null,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),s=e.layers,delete e.layers,i=e;let a=new t(i);if(!(a instanceof qi))throw new St(`Sequential.fromConfig called on non-Sequential input: ${a}`);for(let u of s){let c=gn(u,void 0,o);o&&c.setFastWeightInitDuringBuild(!0),a.add(c)}return a}set stopTraining(t){if(this.model==null)throw new M("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=t}get stopTraining(){if(this.model==null)throw new M("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let t=[];for(let e of this.layers){let n={};n.className=e.getClassName(),n.config=e.getConfig(),t.push(n)}return{name:this.name,layers:t}}};qi.className="Sequential";Q.registerClass(qi);function q8(r){return new Bn(r)}function K8(r){return new qi(r)}function Pv(r){return Dy(r)}function j8(r,t){hn.registerCallbackConstructor(r,t)}var Qr=class extends Q.Serializable{getConfig(){return{}}},Qy=class extends Qr{apply(t,e=1){return F$(t,e)}};Qy.className="elu";Q.registerClass(Qy);var tb=class extends Qr{apply(t){return pm(t)}};tb.className="selu";Q.registerClass(tb);var eb=class extends Qr{apply(t){return Fr(t)}};eb.className="relu";Q.registerClass(eb);var rb=class extends Qr{apply(t){return B(()=>Mi(6,Fr(t)))}};rb.className="relu6";Q.registerClass(rb);var nb=class extends Qr{apply(t){return t}};nb.className="linear";Q.registerClass(nb);var ob=class extends Qr{apply(t){return Yr(t)}};ob.className="sigmoid";Q.registerClass(ob);var sb=class extends Qr{apply(t){return P$(t)}};sb.className="hardSigmoid";Q.registerClass(sb);var ib=class extends Qr{apply(t){return zs(t)}};ib.className="softplus";Q.registerClass(ib);var ab=class extends Qr{apply(t){return O$(t)}};ab.className="softsign";Q.registerClass(ab);var lb=class extends Qr{apply(t){return $i(t)}};lb.className="tanh";Q.registerClass(lb);var qm=class extends Qr{apply(t,e=-1){return iu(t,e)}};qm.className="softmax";Q.registerClass(qm);var ub=class extends Qr{apply(t,e=-1){return sm(t,e)}};ub.className="logSoftmax";Q.registerClass(ub);var cb=class extends Qr{apply(t,e=1){return B(()=>D(Yr(D(t,e)),t))}};cb.className="swish";Q.registerClass(cb);var pb=class extends Qr{apply(t){return B(()=>D(t,$i(zs(t))))}};pb.className="mish";Q.registerClass(pb);function js(r){return r.getClassName()}function Lv(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,"activation")}function Xs(r){if(r==null){let t={};return t.className="linear",t.config={},Lv(t)}if(typeof r=="string"){let t={};return t.className=r,t.config={},Lv(t)}else return r instanceof Qr?r:Lv(r)}function Mv(r){if(r!=null&&typeof r!="object")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${r}`)}var mb=class extends Q.Serializable{},wu=class extends mb{constructor(t){super(),Mv(t),this.l1=t==null||t.l1==null?.01:t.l1,this.l2=t==null||t.l2==null?.01:t.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(t){return B(()=>{let e=Ne([1]);return this.hasL1&&(e=X(e,ft(D(this.l1,Ee(t))))),this.hasL2&&(e=X(e,ft(D(this.l2,lc(t))))),R(e,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(t,e){return new t({l1:e.l1,l2:e.l2})}};wu.className="L1L2";Q.registerClass(wu);function yD(r){return Mv(r),new wu({l1:r!=null?r.l1:null,l2:0})}function bD(r){return Mv(r),new wu({l2:r!=null?r.l2:null,l1:0})}var gD={l1l2:"L1L2"};function me(r){return Sm(r)}function xD(r,t={}){return Gi(r,Q.SerializationMap.getMap().classNameMap,t,"regularizer")}function be(r){if(r==null)return null;if(typeof r=="string"){let e={className:r in gD?gD[r]:r,config:{}};return xD(e)}else return r instanceof mb?r:xD(r)}var Km=class extends $t{constructor(t){super(t==null?{}:t),this.supportsMasking=!0,t!=null&&(this.maxValue=t.maxValue)}call(t,e){t=Nt(t);let n=Fr(t);return this.maxValue!=null&&(n=Cr(n,0,this.maxValue)),n}computeOutputShape(t){return t}getConfig(){let t={maxValue:this.maxValue},e=super.getConfig();return Object.assign(t,e),t}};Km.className="ReLU";Q.registerClass(Km);var jm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_ALPHA=.3,t==null&&(t={}),this.alpha=t.alpha==null?this.DEFAULT_ALPHA:t.alpha}call(t,e){let n=Nt(t);return Ql(n,this.alpha)}computeOutputShape(t){return t}getConfig(){let t={alpha:this.alpha},e=super.getConfig();return Object.assign(t,e),t}};jm.className="LeakyReLU";Q.registerClass(jm);var Xm=class extends $t{constructor(t){if(super(t==null?{}:t),this.DEFAULT_ALPHA_INITIALIZER="zeros",t==null&&(t={}),this.supportsMasking=!0,this.alphaInitializer=de(t.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=be(t.alphaRegularizer),this.alphaConstraint=Be(t.alphaConstraint),t.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(t.sharedAxes))this.sharedAxes=t.sharedAxes;else if(typeof t.sharedAxes=="number")this.sharedAxes=[t.sharedAxes];else throw new M(`Expected sharedAxes to be a number or an array of numbers, but got ${t.sharedAxes}`)}build(t){t=Bt(t);let e=t.slice(1);if(this.sharedAxes!=null)for(let o of this.sharedAxes)e[o-1]=1;this.alpha=this.addWeight("alpha",e,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(this.sharedAxes!=null)for(let o=1;o<t.length;++o)n[o]=t[o];this.inputSpec=[new ye({ndim:t.length,axes:n})],this.built=!0}call(t,e){return t=Nt(t),ou(t,this.alpha.read())}getConfig(){let t={alphaInitializer:Te(this.alphaInitializer),alphaRegularizer:me(this.alphaRegularizer),alphaConstraint:ze(this.alphaConstraint),sharedAxes:this.sharedAxes},e=super.getConfig();return Object.assign(t,e),t}};Xm.className="PReLU";Q.registerClass(Xm);var Ym=class extends $t{constructor(t){if(super(t==null?{}:t),this.DEFAULT_ALPHA=1,t==null&&(t={}),t.alpha!=null&&t.alpha!==this.DEFAULT_ALPHA)throw new St(`Non-default alpha value (${t.alpha}) is not supported by the ELU layer yet.`);this.alpha=t.alpha==null?this.DEFAULT_ALPHA:t.alpha}call(t,e){let n=Nt(t);return Oi(n)}computeOutputShape(t){return t}getConfig(){let t={alpha:this.alpha},e=super.getConfig();return Object.assign(t,e),t}};Ym.className="ELU";Q.registerClass(Ym);var Zm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_THETA=1,t==null&&(t={}),this.theta=t.theta==null?this.DEFAULT_THETA:t.theta}call(t,e){let n=Nt(t);return D(n,J(Re(n,this.theta),"float32"))}computeOutputShape(t){return t}getConfig(){let t={theta:this.theta},e=super.getConfig();return Object.assign(t,e),t}};Zm.className="ThresholdedReLU";Q.registerClass(Zm);var Jm=class extends $t{constructor(t){super(t==null?{}:t),this.DEFAULT_AXIS=1,t==null&&(t={}),this.softmax=new qm().apply,this.axis=t.axis==null?this.DEFAULT_AXIS:t.axis}call(t,e){let n=Nt(t);return this.softmax(n,this.axis)}computeOutputShape(t){return t}getConfig(){let t={axis:this.axis},e=super.getConfig();return Object.assign(t,e),t}};Jm.className="Softmax";Q.registerClass(Jm);function Cu(r,t,e){if(typeof r=="number")return Io(r,t);if(r.length!==t)throw new M(`The ${e} argument must be an integer or tuple of ${t} integers. Received: ${r.length} elements.`);for(let n=0;n<t;++n){let o=r[n];if(!A$(o))throw new M(`The ${e} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(r)} including a non-integer number ${o}`)}return r}function Nn(r,t,e,n,o=1){if(r==null)return r;let s=t+(t-1)*(o-1),i;return e==="same"?i=r:i=r-s+1,Math.floor((i+n-1)/n)}function Ys(r,t,e,n){if(r==null)return null;if(n==="valid")r=r*t+qs([e-t,0]);else if(n==="same")r=r*t;else throw new M(`Unsupport padding mode: ${n}.`);return r}function Ah(r,t){return B(()=>(Fe(t),t==="channelsFirst"?Ot(r,[0,2,3,1]):r))}function zv(r,t){return B(()=>(Fe(t),t==="channelsFirst"?Ot(r,[0,2,3,4,1]):r))}function Y8(r,t,e,n=1,o="valid",s,i=1){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.shape.length!==3)throw new M(`The input of a conv1dWithBias operation should be 3, but is ${r.shape.length} instead.`);if(t.shape.length!==3)throw new M(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(e!=null&&e.shape.length!==1)throw new M(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(s==="channelsFirst"&&(r=Ot(r,[0,2,1])),o==="causal")throw new St("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let a=Qp(r,t,n,o==="same"?"same":"valid","NWC",i);return e!=null&&(a=fn(a,e)),a})}function wD(r,t,e,n=[1,1],o="valid",s,i,a=null){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.rank!==3&&r.rank!==4)throw new M(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${r.rank}.`);if(t.rank!==3&&t.rank!==4)throw new M(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${r.rank}.`);let u=Ah(r,s);if(o==="causal")throw new St("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return u=uu.conv2d({x:u,filter:t,strides:n,pad:o==="same"?"same":"valid",dilations:i,dataFormat:"NHWC",bias:e,activation:a}),s==="channelsFirst"&&(u=Ot(u,[0,3,1,2])),u})}function Z8(r,t,e,n=[1,1,1],o="valid",s,i){return B(()=>{if(s==null&&(s=mn()),Fe(s),r.rank!==4&&r.rank!==5)throw new M(`conv3dWithBias expects input to be of rank 4 or 5, but received ${r.rank}.`);if(t.rank!==4&&t.rank!==5)throw new M(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${r.rank}.`);let a=zv(r,s);if(o==="causal")throw new St("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return a=Tx(a,t,n,o==="same"?"same":"valid","NDHWC",i),e!=null&&(a=fn(a,e)),s==="channelsFirst"&&(a=Ot(a,[0,4,1,2,3])),a})}var bc=class extends $t{constructor(t,e){if(super(e),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",bc.verifyArgs(e),this.rank=t,Ze(this.rank,"rank"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new St(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=Cu(e.kernelSize,t,"kernelSize"),this.strides=Cu(e.strides==null?1:e.strides,t,"strides"),this.padding=e.padding==null?"valid":e.padding,pn(this.padding),this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,Fe(this.dataFormat),this.activation=Xs(e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.biasInitializer=de(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=Be(e.biasConstraint),this.biasRegularizer=be(e.biasRegularizer),this.activityRegularizer=be(e.activityRegularizer),this.dilationRate=Cu(e.dilationRate==null?1:e.dilationRate,t,"dilationRate"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new M(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new M(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new M(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(t){if(ro("kernelSize"in t,"required key 'kernelSize' not in config"),typeof t.kernelSize!="number"&&!Cy(t.kernelSize,"number",1,3))throw new M(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(t.kernelSize)}.`)}getConfig(){let t={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:js(this.activation),useBias:this.useBias,biasInitializer:Te(this.biasInitializer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),biasConstraint:ze(this.biasConstraint)},e=super.getConfig();return Object.assign(t,e),t}},Iu=class extends bc{constructor(t,e){super(t,e),this.kernel=null,Iu.verifyArgs(e),this.filters=e.filters,Ze(this.filters,"filters"),this.kernelInitializer=de(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=Be(e.kernelConstraint),this.kernelRegularizer=be(e.kernelRegularizer)}build(t){t=Bt(t);let e=this.dataFormat==="channelsFirst"?1:t.length-1;if(t[e]==null)throw new M(`The channel dimension of the input should be defined. Found ${t[e]}`);let n=t[e],o=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",o,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[e]:n}}],this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n,o=this.bias==null?null:this.bias.read(),s=Iy(this.activation.getClassName());if(s!=null&&this.rank===2)n=wD(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate,s);else{if(this.rank===1)n=Y8(t,this.kernel.read(),o,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)n=wD(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)n=Z8(t,this.kernel.read(),o,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new St("convolutions greater than 3D are not implemented yet.");this.activation!=null&&(n=this.activation.apply(n))}return n})}computeOutputShape(t){t=Bt(t);let e=[],n=this.dataFormat==="channelsLast"?t.slice(1,t.length-1):t.slice(2);for(let s=0;s<n.length;++s){let i=Nn(n[s],this.kernelSize[s],this.padding,this.strides[s],typeof this.dilationRate=="number"?this.dilationRate:this.dilationRate[s]);e.push(i)}let o=[t[0]];return this.dataFormat==="channelsLast"?(o=o.concat(e),o.push(this.filters)):(o.push(this.filters),o=o.concat(e)),o}getConfig(){let t={filters:this.filters,kernelInitializer:Te(this.kernelInitializer),kernelRegularizer:me(this.kernelRegularizer),kernelConstraint:ze(this.kernelConstraint)},e=super.getConfig();return Object.assign(t,e),t}static verifyArgs(t){if(!("filters"in t)||typeof t.filters!="number"||t.filters<1)throw new M(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(t.filters)}`)}},il=class extends Iu{constructor(t){super(2,t),il.verifyArgs(t)}getConfig(){let t=super.getConfig();return delete t.rank,t}static verifyArgs(t){if(typeof t.kernelSize!="number"&&!Cy(t.kernelSize,"number",1,2))throw new M(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(t.kernelSize)}.`)}};il.className="Conv2D";Q.registerClass(il);var al=class extends Iu{constructor(t){super(3,t),al.verifyArgs(t)}getConfig(){let t=super.getConfig();return delete t.rank,t}static verifyArgs(t){if(typeof t.kernelSize!="number"&&!(Array.isArray(t.kernelSize)&&(t.kernelSize.length===1||t.kernelSize.length===3)))throw new M(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(t.kernelSize)}.`)}};al.className="Conv3D";Q.registerClass(al);var Qm=class extends il{constructor(t){if(super(t),this.inputSpec=[new ye({ndim:4})],this.padding!=="same"&&this.padding!=="valid")throw new M(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(t){if(t=Bt(t),t.length!==4)throw new M("Input should have rank 4; Received input shape: "+JSON.stringify(t));let e=this.dataFormat==="channelsFirst"?1:t.length-1;if(t[e]==null)throw new M("The channel dimension of the inputs should be defined. Found `None`.");let n=t[e],o=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",o,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new ye({ndim:4,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{let n=Nt(t);if(n.shape.length!==4)throw new M(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let o=n.shape,s=o[0],i,a;this.dataFormat==="channelsFirst"?(i=2,a=3):(i=1,a=2);let u=o[i],l=o[a],c=this.kernelSize[0],p=this.kernelSize[1],m=this.strides[0],f=this.strides[1],d=Ys(u,m,c,this.padding),h=Ys(l,f,p,this.padding),g=[s,d,h,this.filters];this.dataFormat!=="channelsLast"&&(n=Ot(n,[0,2,3,1]));let x=em(n,this.kernel.read(),g,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(x=Ot(x,[0,3,1,2])),this.bias!=null&&(x=fn(x,this.bias.read(),this.dataFormat)),this.activation!=null&&(x=this.activation.apply(x)),x})}computeOutputShape(t){t=Bt(t);let e=t.slice(),n,o,s;this.dataFormat==="channelsFirst"?(n=1,o=2,s=3):(n=3,o=1,s=2);let i=this.kernelSize[0],a=this.kernelSize[1],u=this.strides[0],l=this.strides[1];return e[n]=this.filters,e[o]=Ys(e[o],u,i,this.padding),e[s]=Ys(e[s],l,a,this.padding),e}getConfig(){let t=super.getConfig();return delete t.dilationRate,t}};Qm.className="Conv2DTranspose";Q.registerClass(Qm);var tf=class extends al{constructor(t){if(super(t),this.inputSpec=[new ye({ndim:5})],this.padding!=="same"&&this.padding!=="valid")throw new M(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(t){if(t=Bt(t),t.length!==5)throw new M("Input should have rank 5; Received input shape: "+JSON.stringify(t));let e=this.dataFormat==="channelsFirst"?1:t.length-1;if(t[e]==null)throw new M("The channel dimension of the inputs should be defined. Found `None`.");let n=t[e],o=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",o,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new ye({ndim:5,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{let n=Nt(t);if(n.shape.length!==5)throw new M(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${n.shape.length}`);let o=n.shape,s=o[0],i,a,u;this.dataFormat==="channelsFirst"?(u=2,i=3,a=4):(u=1,i=2,a=3);let l=o[u],c=o[i],p=o[a],m=this.kernelSize[0],f=this.kernelSize[1],d=this.kernelSize[2],h=this.strides[0],g=this.strides[1],x=this.strides[2],b=Ys(l,h,m,this.padding),w=Ys(c,g,f,this.padding),C=Ys(p,x,d,this.padding),N=[s,b,w,C,this.filters];this.dataFormat!=="channelsLast"&&(n=Ot(n,[0,2,3,4,1]));let _=Ex(n,this.kernel.read(),N,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(_=Ot(_,[0,4,1,2,3])),this.bias!==null&&(_=fn(_,this.bias.read(),this.dataFormat)),this.activation!==null&&(_=this.activation.apply(_)),_})}computeOutputShape(t){t=Bt(t);let e=t.slice(),n,o,s,i;this.dataFormat==="channelsFirst"?(n=1,o=2,s=3,i=4):(n=4,o=1,s=2,i=3);let a=this.kernelSize[0],u=this.kernelSize[1],l=this.kernelSize[2],c=this.strides[0],p=this.strides[1],m=this.strides[2];return e[n]=this.filters,e[o]=Ys(e[o],c,a,this.padding),e[s]=Ys(e[s],p,u,this.padding),e[i]=Ys(e[i],m,l,this.padding),e}getConfig(){let t=super.getConfig();return delete t.dilationRate,t}};tf.className="Conv3DTranspose";Q.registerClass(tf);var fb=class extends Iu{constructor(t,e){if(super(t,e),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,e.filters==null)throw new M("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(e.kernelInitializer!=null||e.kernelRegularizer!=null||e.kernelConstraint!=null)throw new M("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(e.padding!=null&&e.padding!=="same"&&e.padding!=="valid")throw new M(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(e.padding)}`);this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=de(e.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=be(e.depthwiseRegularizer),this.depthwiseConstraint=Be(e.depthwiseConstraint),this.pointwiseInitializer=de(e.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=be(e.pointwiseRegularizer),this.pointwiseConstraint=Be(e.pointwiseConstraint)}build(t){if(t=Bt(t),t.length<this.rank+2)throw new M(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(t)}`);let e=this.dataFormat==="channelsFirst"?1:t.length-1;if(t[e]==null||t[e]<0)throw new M(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(t[e])}`);let n=t[e],o=this.kernelSize.concat([n,this.depthMultiplier]),s=[];for(let a=0;a<this.rank;++a)s.push(1);s.push(n*this.depthMultiplier,this.filters);let i=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",o,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,i,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",s,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,i,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,i,this.biasConstraint):this.bias=null,this.inputSpec=[new ye({ndim:this.rank+2,axes:{[e]:n}})],this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n;if(this.rank===1)throw new St("1D separable convolution is not implemented yet.");return this.rank===2&&(this.dataFormat==="channelsFirst"&&(t=Ot(t,[0,2,3,1])),n=mm(t,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(n=fn(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),this.dataFormat==="channelsFirst"&&(n=Ot(n,[0,3,1,2])),n})}getConfig(){let t=super.getConfig();return delete t.rank,delete t.kernelInitializer,delete t.kernelRegularizer,delete t.kernelConstraint,t.depthwiseInitializer=Te(this.depthwiseInitializer),t.pointwiseInitializer=Te(this.pointwiseInitializer),t.depthwiseRegularizer=me(this.depthwiseRegularizer),t.pointwiseRegularizer=me(this.pointwiseRegularizer),t.depthwiseConstraint=ze(this.depthwiseConstraint),t.pointwiseConstraint=ze(this.pointwiseConstraint),t}};fb.className="SeparableConv";var ef=class extends fb{constructor(t){super(2,t)}};ef.className="SeparableConv2D";Q.registerClass(ef);var Su=class extends Iu{constructor(t){super(1,t),Su.verifyArgs(t),this.inputSpec=[{ndim:3}]}getConfig(){let t=super.getConfig();return delete t.rank,delete t.dataFormat,t}static verifyArgs(t){if(typeof t.kernelSize!="number"&&!Cy(t.kernelSize,"number",1,1))throw new M(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(t.kernelSize)}.`)}};Su.className="Conv1D";Q.registerClass(Su);var rf=class extends $t{constructor(t){super(t),typeof t.cropping=="number"?this.cropping=[[t.cropping,t.cropping],[t.cropping,t.cropping]]:typeof t.cropping[0]=="number"?this.cropping=[[t.cropping[0],t.cropping[0]],[t.cropping[1],t.cropping[1]]]:this.cropping=t.cropping,this.dataFormat=t.dataFormat===void 0?"channelsLast":t.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(t){return this.dataFormat==="channelsFirst"?[t[0],t[1],t[2]-this.cropping[0][0]-this.cropping[0][1],t[3]-this.cropping[1][0]-this.cropping[1][1]]:[t[0],t[1]-this.cropping[0][0]-this.cropping[0][1],t[2]-this.cropping[1][0]-this.cropping[1][1],t[3]]}call(t,e){return B(()=>{if(t=Nt(t),this.dataFormat==="channelsLast"){let n=wh(t,this.cropping[0][0],t.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return wh(n,this.cropping[1][0],t.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let n=wh(t,this.cropping[0][0],t.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return wh(n,this.cropping[1][0],t.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let t={cropping:this.cropping,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}};rf.className="Cropping2D";Q.registerClass(rf);var nf=class extends $t{constructor(t){super(t),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=t.size==null?this.DEFAULT_SIZE:t.size,this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Fe(this.dataFormat),this.interpolation=t.interpolation==null?"nearest":t.interpolation,E$(this.interpolation)}computeOutputShape(t){if(this.dataFormat==="channelsFirst"){let e=t[2]==null?null:this.size[0]*t[2],n=t[3]==null?null:this.size[1]*t[3];return[t[0],t[1],e,n]}else{let e=t[1]==null?null:this.size[0]*t[1],n=t[2]==null?null:this.size[1]*t[2];return[t[0],e,n,t[3]]}}call(t,e){return B(()=>{let n=Nt(t),o=n.shape;if(this.dataFormat==="channelsFirst"){n=Ot(n,[0,2,3,1]);let s=this.size[0]*o[2],i=this.size[1]*o[3],a=this.interpolation==="nearest"?Gs.resizeNearestNeighbor(n,[s,i]):Gs.resizeBilinear(n,[s,i]);return Ot(a,[0,3,1,2])}else{let s=this.size[0]*o[1],i=this.size[1]*o[2];return this.interpolation==="nearest"?Gs.resizeNearestNeighbor(n,[s,i]):Gs.resizeBilinear(n,[s,i])}})}getConfig(){let t={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},e=super.getConfig();return Object.assign(t,e),t}};nf.className="UpSampling2D";Q.registerClass(nf);function J8(r,t,e=[1,1],n="valid",o,s){return B(()=>{o==null&&(o=mn()),Fe(o);let i=Ah(r,o);if(r.rank!==4)throw new M(`Input for depthwiseConv2d is required to be 4-D, but is instead ${r.rank}-D`);if(t.rank!==4)throw new M(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=Fi(i,t,e,n==="same"?"same":"valid","NHWC",s),o==="channelsFirst"&&(i=Ot(i,[0,3,1,2])),i})}var of=class extends bc{constructor(t){super(2,t),this.depthwiseKernel=null,this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=de(t.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=Be(t.depthwiseConstraint),this.depthwiseRegularizer=be(t.depthwiseRegularizer)}build(t){if(t=Bt(t),t.length<4)throw new M(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(t)}.`);let e=this.dataFormat==="channelsFirst"?1:3;if(t[e]==null||t[e]<0)throw new M(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${t[e]}).`);let n=t[e],o=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",o,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{t=Nt(t);let n=J8(t,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(n=fn(n,this.bias.read(),this.dataFormat)),this.activation!=null&&(n=this.activation.apply(n)),n})}computeOutputShape(t){t=Bt(t);let e=this.dataFormat==="channelsFirst"?t[2]:t[1],n=this.dataFormat==="channelsFirst"?t[3]:t[2],o=this.dataFormat==="channelsFirst"?t[1]*this.depthMultiplier:t[3]*this.depthMultiplier,s=Nn(e,this.kernelSize[0],this.padding,this.strides[0]),i=Nn(n,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat==="channelsFirst"?[t[0],o,s,i]:[t[0],s,i,o]}getConfig(){let t=super.getConfig();return t.depthMultiplier=this.depthMultiplier,t.depthwiseInitializer=Te(this.depthwiseInitializer),t.depthwiseRegularizer=me(this.depthwiseRegularizer),t.depthwiseConstraint=ze(this.depthwiseRegularizer),t}};of.className="DepthwiseConv2D";Q.registerClass(of);function Bv(r,t,e,n){if(Array.isArray(r)){if(t!=null||e!=null)throw new M("When inputs is an array, neither initialState or constants should be provided");n!=null&&(e=r.slice(r.length-n,r.length),r=r.slice(0,r.length-n)),r.length>1&&(t=r.slice(1,r.length)),r=r[0]}function o(s){return s==null||Array.isArray(s)?s:[s]}return t=o(t),e=o(e),{inputs:r,initialState:t,constants:e}}function Vv(r,t,e,n=!1,o,s,i=!1,a=!1){return B(()=>{let u=t.shape.length;if(u<3)throw new M(`Input should be at least 3D, but is ${u}D.`);let l=[1,0].concat(Zr(2,u));if(t=Ot(t,l),s!=null)throw new St("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),o!=null&&(o=J(J(o,"bool"),"float32"),o.rank===u-1&&(o=rr(o,-1)),o=Ot(o,l)),n&&(t=pr(t,0),o!=null&&(o=pr(o,0)));let c=[],p,m=e,f=t.shape[0],d=vr(t),h;o!=null&&(h=vr(o));for(let x=0;x<f;++x){let b=d[x],w=B(()=>r(b,m));if(o==null)p=w[0],m=w[1];else{let C=B(()=>{let N=h[x],_=ct(yr(N),N),A=X(D(w[0],N),D(m[0],_)),$=m.map((F,P)=>X(D(w[1][P],N),D(F,_)));return{output:A,newStates:$}});p=C.output,m=C.newStates}a&&c.push(p)}let g;return a&&(g=nr(c,1)),[p,g,m]})}var Tn=class extends $t{constructor(t){super(t);let e;if(t.cell==null)throw new M("cell property is missing for the constructor of RNN.");if(Array.isArray(t.cell)?e=new Ic({cells:t.cell}):e=t.cell,e.stateSize==null)throw new M("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=e,this.returnSequences=t.returnSequences==null?!1:t.returnSequences,this.returnState=t.returnState==null?!1:t.returnState,this.goBackwards=t.goBackwards==null?!1:t.goBackwards,this._stateful=t.stateful==null?!1:t.stateful,this.unroll=t.unroll==null?!1:t.unroll,this.supportsMasking=!0,this.inputSpec=[new ye({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let t=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return Zr(0,t).map(e=>null)}else return this.states_}setStates(t){this.states_=t}computeOutputShape(t){$y(t)&&(t=t[0]),t=t;let e=this.cell.stateSize;Array.isArray(e)||(e=[e]);let n=e[0],o;if(this.returnSequences?o=[t[0],t[1],n]:o=[t[0],n],this.returnState){let s=[];for(let i of e)s.push([t[0],i]);return[o].concat(s)}else return o}computeMask(t,e){return B(()=>{Array.isArray(e)&&(e=e[0]);let n=this.returnSequences?e:null;if(this.returnState){let o=this.states.map(s=>null);return[n].concat(o)}else return n})}get states(){if(this.states_==null){let t=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,e=[];for(let n=0;n<t;++n)e.push(null);return e}else return this.states_}set states(t){this.states_=t}build(t){if(this.numConstants!=null)throw new St("Constants support is not implemented in RNN yet.");$y(t)&&(t=t[0]),t=t;let n=this.stateful?t[0]:null,o=t.slice(2);this.inputSpec[0]=new ye({shape:[n,null,...o]});let s=[t[0]].concat(t.slice(2));this.cell.build(s);let i;if(Array.isArray(this.cell.stateSize)?i=this.cell.stateSize:i=[this.cell.stateSize],this.stateSpec!=null){if(!y.arraysEqual(this.stateSpec.map(a=>a.shape[a.shape.length-1]),i))throw new M(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=i.map(a=>new ye({shape:[null,a]}));this.stateful&&this.resetStates()}resetStates(t,e=!1){B(()=>{if(!this.stateful)throw new vn("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape[0];if(n==null)throw new M("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(o=>Ne([n,o])):this.states_=[Ne([n,this.cell.stateSize])];else if(t==null)vt(this.states_),this.keptStates!=null&&(vt(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(o=>Ne([n,o])):this.states_[0]=Ne([n,this.cell.stateSize]);else{if(Array.isArray(t)||(t=[t]),t.length!==this.states_.length)throw new M(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${t.length} state value(s). Input received: ${t}`);e===!0?this.keptStates.push(this.states_.slice()):vt(this.states_);for(let o=0;o<this.states_.length;++o){let s=t[o],i=Array.isArray(this.cell.stateSize)?this.cell.stateSize[o]:this.cell.stateSize,a=[n,i];if(!y.arraysEqual(s.shape,a))throw new M(`State ${o} is incompatible with layer ${this.name}: expected shape=${a}, received shape=${s.shape}`);this.states_[o]=s}}this.states_=this.states_.map(o=>De(o.clone()))})}apply(t,e){let n=e==null?null:e.initialState,o=e==null?null:e.constants;e==null&&(e={});let s=Bv(t,n,o,this.numConstants);t=s.inputs,n=s.initialState,o=s.constants;let i=[],a=[];if(n!=null){e.initialState=n,i=i.concat(n),this.stateSpec=[];for(let l of n)this.stateSpec.push(new ye({shape:l.shape}));a=a.concat(this.stateSpec)}if(o!=null&&(e.constants=o,i=i.concat(o),this.numConstants=o.length),i[0]instanceof Jr){let l=[t].concat(i),c=this.inputSpec.concat(a),p=this.inputSpec;this.inputSpec=c;let m=super.apply(l,e);return this.inputSpec=p,m}else return super.apply(t,e)}call(t,e){return B(()=>{let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;t=Nt(t),s==null&&(this.stateful?s=this.states_:s=this.getInitialState(t));let i=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(s.length!==i)throw new M(`RNN Layer has ${i} state(s) but was passed ${s.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let a={training:o},l=Vv((d,h)=>{let g=this.cell.call([d].concat(h),a);return[g[0],g.slice(1)]},t,s,this.goBackwards,n,null,this.unroll,this.returnSequences),c=l[0],p=l[1],m=l[2];this.stateful&&this.resetStates(m,o);let f=this.returnSequences?p:c;return this.returnState?[f].concat(m):f})}getInitialState(t){return B(()=>{let e=Ne(t.shape);return e=ft(e,[1,2]),e=nl(e),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(n=>n>1?Ey(e,[1,n]):e):this.cell.stateSize>1?[Ey(e,[1,this.cell.stateSize])]:[e]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(t)}getConfig(){let t=super.getConfig(),e={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(e.numConstants=this.numConstants);let n=this.cell.getConfig();return this.getClassName()===Tn.className&&(e.cell={className:this.cell.getClassName(),config:n}),Object.assign(Object.assign(Object.assign({},n),t),e)}static fromConfig(t,e,n={}){let o=e.cell,s=gn(o,n);return new t(Object.assign(e,{cell:s}))}};Tn.className="RNN";Q.registerClass(Tn);var ll=class extends $t{},wc=class extends ll{constructor(t){super(t),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=t.units,Ze(this.units,"units"),this.activation=Xs(t.activation==null?this.DEFAULT_ACTIVATION:t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){t=Bt(t),this.kernel=this.addWeight("kernel",[t[t.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{if(t=t,t.length!==2)throw new M(`SimpleRNNCell expects 2 input Tensors, got ${t.length}.`);let n=t[1];t=t[0];let o=e.training==null?!1:e.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:o,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(n),rate:this.recurrentDropout,training:o,dropoutFunc:this.dropoutFunc}));let s,i=this.dropoutMask,a=this.recurrentDropoutMask;i!=null?s=To(D(t,i),this.kernel.read()):s=To(t,this.kernel.read()),this.bias!=null&&(s=fn(s,this.bias.read())),a!=null&&(n=D(n,a));let u=X(s,To(n,this.recurrentKernel.read()));return this.activation!=null&&(u=this.activation.apply(u)),[u,u]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},t),e)}};wc.className="SimpleRNNCell";Q.registerClass(wc);var sf=class extends Tn{constructor(t){t.cell=new wc(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return new t(e)}};sf.className="SimpleRNN";Q.registerClass(sf);var Cc=class extends ll{constructor(t){if(super(t),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",t.resetAfter)throw new M("GRUCell does not support reset_after parameter set to true.");this.units=t.units,Ze(this.units,"units"),this.activation=Xs(t.activation===void 0?this.DEFAULT_ACTIVATION:t.activation),this.recurrentActivation=Xs(t.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:t.recurrentActivation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.implementation=t.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){t=Bt(t);let e=t[t.length-1];this.kernel=this.addWeight("kernel",[e,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(t,e){return B(()=>{if(t=t,t.length!==2)throw new M(`GRUCell expects 2 input Tensors (inputs, h, c), got ${t.length}.`);let n=e.training==null?!1:e.training,o=t[1];t=t[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(o),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,i=this.recurrentDropoutMask,a,u,l;0<this.dropout&&this.dropout<1&&(t=D(t,s[0]));let c=To(t,this.kernel.read());this.useBias&&(c=fn(c,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(o=D(o,i[0]));let p=this.recurrentKernel.read(),[m,f]=mr(p,[2*this.units,this.units],p.rank-1),d=To(o,m),[h,g,x]=mr(c,3,c.rank-1),[b,w]=mr(d,2,d.rank-1);a=this.recurrentActivation.apply(X(h,b)),u=this.recurrentActivation.apply(X(g,w));let C=To(D(u,o),f);l=this.activation.apply(X(x,C));let N=X(D(a,o),D(X(1,Ht(a)),l));return[N,N]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),recurrentActivation:js(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},t),e)}};Cc.className="GRUCell";Q.registerClass(Cc);var af=class extends Tn{constructor(t){t.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),t.cell=new Cc(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return e.implmentation===0&&(e.implementation=1),new t(e)}};af.className="GRU";Q.registerClass(af);var ul=class extends ll{constructor(t){super(t),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=t.units,Ze(this.units,"units"),this.activation=Xs(t.activation===void 0?this.DEFAULT_ACTIVATION:t.activation),this.recurrentActivation=Xs(t.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:t.recurrentActivation),this.useBias=t.useBias==null?!0:t.useBias,this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=de(t.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=t.unitForgetBias,this.kernelRegularizer=be(t.kernelRegularizer),this.recurrentRegularizer=be(t.recurrentRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.kernelConstraint=Be(t.kernelConstraint),this.recurrentConstraint=Be(t.recurrentConstraint),this.biasConstraint=Be(t.biasConstraint),this.dropout=ac([1,qs([0,t.dropout==null?0:t.dropout])]),this.recurrentDropout=ac([1,qs([0,t.recurrentDropout==null?0:t.recurrentDropout])]),this.dropoutFunc=t.dropoutFunc,this.implementation=t.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(t){var e;t=Bt(t);let n=t[t.length-1];this.kernel=this.addWeight("kernel",[n,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let o;if(this.useBias){if(this.unitForgetBias){let s=this.biasInitializer,i=this.units;o=new(e=class extends dn{apply(u,l){let c=s.apply([i]),p=new yu().apply([i]),m=s.apply([i*2]);return Tv(Tv(c,p),m)}},e.className="CustomInit",e)}else o=this.biasInitializer;this.bias=this.addWeight("bias",[this.units*4],null,o,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(t,e){return B(()=>{let n=e.training==null?!1:e.training;if(t=t,t.length!==3)throw new M(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${t.length}.`);let o=t[1],s=t[2];t=t[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(t),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(o),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let i=this.dropoutMask,a=this.recurrentDropoutMask,u,l,c,p;0<this.dropout&&this.dropout<1&&(t=D(t,i[0]));let m=To(t,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(o=D(o,a[0])),m=X(m,To(o,this.recurrentKernel.read())),this.useBias&&(m=fn(m,this.bias.read()));let[f,d,h,g]=mr(m,4,m.rank-1);u=this.recurrentActivation.apply(f),l=this.recurrentActivation.apply(d),c=X(D(l,s),D(u,this.activation.apply(h))),p=this.recurrentActivation.apply(g);let x=D(p,this.activation.apply(c));return[x,x,c]})}getConfig(){let t=super.getConfig(),e={units:this.units,activation:js(this.activation),recurrentActivation:js(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),recurrentInitializer:Te(this.recurrentInitializer),biasInitializer:Te(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:me(this.kernelRegularizer),recurrentRegularizer:me(this.recurrentRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),recurrentConstraint:ze(this.recurrentConstraint),biasConstraint:ze(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},t),e)}};ul.className="LSTMCell";Q.registerClass(ul);var lf=class extends Tn{constructor(t){t.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),t.cell=new ul(t),super(t)}call(t,e){return B(()=>{this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}static fromConfig(t,e){return e.implmentation===0&&(e.implementation=1),new t(e)}};lf.className="LSTM";Q.registerClass(lf);var Ic=class extends ll{constructor(t){super(t),this.cells=t.cells}get stateSize(){let t=[];for(let e of this.cells.slice().reverse())Array.isArray(e.stateSize)?t.push(...e.stateSize):t.push(e.stateSize);return t}call(t,e){return B(()=>{t=t;let n=t.slice(1),o=[];for(let a of this.cells.slice().reverse())Array.isArray(a.stateSize)?o.push(n.splice(0,a.stateSize.length)):o.push(n.splice(0,1));o.reverse();let s=[],i;for(let a=0;a<this.cells.length;++a){let u=this.cells[a];n=o[a],a===0?i=[t[0]].concat(n):i=[i[0]].concat(n),i=u.call(i,e),s.push(i.slice(1))}n=[];for(let a of s.slice().reverse())n.push(...a);return[i[0]].concat(n)})}build(t){$y(t)&&(t=t[0]),t=t;let e;this.cells.forEach((n,o)=>{Hs(`RNNCell_${o}`,()=>{n.build(t),Array.isArray(n.stateSize)?e=n.stateSize[0]:e=n.stateSize,t=[t[0],e]})}),this.built=!0}getConfig(){let t=super.getConfig(),e=s=>({className:s.getClassName(),config:s.getConfig()}),o={cells:this.cells.map(e)};return Object.assign(Object.assign({},t),o)}static fromConfig(t,e,n={}){let o=[];for(let s of e.cells)o.push(gn(s,n));return new t({cells:o})}get trainableWeights(){if(!this.trainable)return[];let t=[];for(let e of this.cells)t.push(...e.trainableWeights);return t}get nonTrainableWeights(){let t=[];for(let e of this.cells)t.push(...e.nonTrainableWeights);if(!this.trainable){let e=[];for(let n of this.cells)e.push(...n.trainableWeights);return e.concat(t)}return t}getWeights(){let t=[];for(let e of this.cells)t.push(...e.weights);return Ih(t)}setWeights(t){let e=[];for(let n of this.cells){let o=n.weights.length,s=t.splice(o);for(let i=0;i<n.weights.length;++i)e.push([n.weights[i],s[i]])}Pm(e)}};Ic.className="StackedRNNCells";Q.registerClass(Ic);function cl(r){let{ones:t,rate:e,training:n=!1,count:o=1,dropoutFunc:s}=r,i=()=>s!=null?s(t(),e):Ay(t(),e),a=()=>xu(i,t,n);return!o||o<=1?De(a().clone()):Array(o).fill(void 0).map(a).map(l=>De(l.clone()))}var Q8=function(r,t){var e={};for(var n in r)Object.prototype.hasOwnProperty.call(r,n)&&t.indexOf(n)<0&&(e[n]=r[n]);if(r!=null&&typeof Object.getOwnPropertySymbols=="function")for(var o=0,n=Object.getOwnPropertySymbols(r);o<n.length;o++)t.indexOf(n[o])<0&&Object.prototype.propertyIsEnumerable.call(r,n[o])&&(e[n[o]]=r[n[o]]);return e};var db=class extends Tn{constructor(t){if(t.unroll)throw new St("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(t.cell))throw new St("It is not possible at the moment to stack convolutional cells.");super(t),this.inputSpec=[new ye({ndim:5})]}call(t,e){return B(()=>{if(this.cell.dropoutMask!=null&&(vt(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(vt(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),e&&e.constants)throw new M("ConvRNN2D cell does not support constants");let n=e==null?null:e.mask,o=e==null?null:e.training,s=e==null?null:e.initialState;return super.call(t,{mask:n,training:o,initialState:s})})}computeOutputShape(t){let e=this.computeSingleOutputShape(t);return this.returnSequences||(e=[e[0],...e.slice(2)]),this.returnState&&(e=[e,...Array(2).fill([t[0],...e.slice(-3)])]),e}getInitialState(t){return B(()=>{let{stateSize:e}=this.cell,n=t.shape,o=this.computeSingleOutputShape(n),s=[o[0],...o.slice(2)],i=Ne(s);return Array.isArray(e)?Array(e.length).fill(i):[i]})}resetStates(t,e=!1){B(()=>{if(!this.stateful)throw new vn("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape,o=this.computeSingleOutputShape(n),s=[o[0],...o.slice(2)];if(n[0]==null)throw new M("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Ne(s)):this.states_=[Ne(s)];else if(t==null)vt(this.states_),this.keptStates!=null&&(vt(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>Ne(s)):this.states_[0]=Ne(s);else{if(Array.isArray(t)||(t=[t]),t.length!==this.states_.length)throw new M(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${t.length} state value(s). Input received: ${t}`);e?this.keptStates.push(this.states_.slice()):vt(this.states_);for(let a=0;a<this.states_.length;++a){let u=t[a],l=s;if(!y.arraysEqual(u.shape,l))throw new M(`State ${a} is incompatible with layer ${this.name}: expected shape=${l}, received shape=${u.shape}`);this.states_[a]=u}}this.states_=this.states_.map(a=>De(a.clone()))})}computeSingleOutputShape(t){let{dataFormat:e,filters:n,kernelSize:o,padding:s,strides:i,dilationRate:a}=this.cell,u=e==="channelsFirst",l=t[u?3:2],c=t[u?4:3],p=Nn(l,o[0],s,i[0],a[0]),m=Nn(c,o[1],s,i[1],a[1]);return[...t.slice(0,2),...u?[n,p,m]:[p,m,n]]}};db.className="ConvRNN2D";var Sc=class extends ul{constructor(t){let{filters:e,kernelSize:n,strides:o,padding:s,dataFormat:i,dilationRate:a}=t;super(Object.assign(Object.assign({},t),{units:e})),this.filters=e,Ze(this.filters,"filters"),this.kernelSize=Cu(n,2,"kernelSize"),this.kernelSize.forEach(u=>Ze(u,"kernelSize")),this.strides=Cu(o||1,2,"strides"),this.strides.forEach(u=>Ze(u,"strides")),this.padding=s||"valid",pn(this.padding),this.dataFormat=i||"channelsLast",Fe(this.dataFormat),this.dilationRate=Cu(a||1,2,"dilationRate"),this.dilationRate.forEach(u=>Ze(u,"dilationRate"))}build(t){var e;t=Bt(t);let n=this.dataFormat==="channelsFirst"?1:t.length-1;if(t[n]==null)throw new M(`The channel dimension of the input should be defined. Found ${t[n]}`);let o=t[n],s=4,i=this.kernelSize.concat([o,this.filters*s]);this.kernel=this.addWeight("kernel",i,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let a=this.kernelSize.concat([this.filters,this.filters*s]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",a,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let u;if(this.unitForgetBias){let l=this.biasInitializer,c=this.filters;u=new(e=class extends dn{apply(m,f){let d=l.apply([c]),h=cr([c]),g=l.apply([c*2]);return Nm([d,h,g])}},e.className="CustomInit",e)}else u=this.biasInitializer;this.bias=this.addWeight("bias",[this.filters*s],null,u,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(t,e){return B(()=>{if(t.length!==3)throw new M(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${t.length}.`);let n=e.training||!1,o=t[0],s=t[1],i=t[2],a=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=cl({ones:()=>yr(o),rate:this.dropout,training:n,count:a,dropoutFunc:this.dropoutFunc}));let u=this.dropoutMask,l=(rt,ot,at)=>!ot||!ot[at]?rt:D(ot[at],rt),c=l(o,u,0),p=l(o,u,1),m=l(o,u,2),f=l(o,u,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=cl({ones:()=>yr(s),rate:this.recurrentDropout,training:n,count:a,dropoutFunc:this.dropoutFunc}));let d=this.recurrentDropoutMask,h=l(s,d,0),g=l(s,d,1),x=l(s,d,2),b=l(s,d,3),w=3,[C,N,_,A]=mr(this.kernel.read(),a,w),[$,F,P,V]=this.useBias?mr(this.bias.read(),a):[null,null,null,null];c=this.inputConv(c,C,$,this.padding),p=this.inputConv(p,N,F,this.padding),m=this.inputConv(m,_,P,this.padding),f=this.inputConv(f,A,V,this.padding);let[G,W,q,H]=mr(this.recurrentKernel.read(),a,w);h=this.recurrentConv(h,G),g=this.recurrentConv(g,W),x=this.recurrentConv(x,q),b=this.recurrentConv(b,H);let j=this.recurrentActivation.apply(X(c,h)),Y=this.recurrentActivation.apply(X(p,g)),Z=X(D(Y,i),D(j,this.activation.apply(X(m,x)))),et=D(this.recurrentActivation.apply(X(f,b)),this.activation.apply(Z));return[et,et,Z]})}getConfig(){let t=super.getConfig(),{units:e}=t,n=Q8(t,["units"]),o={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),o)}inputConv(t,e,n,o){let s=In(t,e,this.strides,o||"valid",this.dataFormat==="channelsFirst"?"NCHW":"NHWC",this.dilationRate);return n?fn(s,n,this.dataFormat):s}recurrentConv(t,e){return In(t,e,1,"same",this.dataFormat==="channelsFirst"?"NCHW":"NHWC")}};Sc.className="ConvLSTM2DCell";Q.registerClass(Sc);var uf=class extends db{constructor(t){let e=new Sc(t);super(Object.assign(Object.assign({},t),{cell:e}))}static fromConfig(t,e){return new t(e)}};uf.className="ConvLSTM2D";Q.registerClass(uf);var vc=class extends $t{constructor(t){super(t),this.rate=Math.max(Math.min(t.rate,1),0),this.noiseShape=t.noiseShape,this.seed=t.seed,this.supportsMasking=!0}getNoiseShape(t){if(this.noiseShape==null)return this.noiseShape;let e=t.shape,n=[];for(let o=0;o<this.noiseShape.length;++o)n.push(this.noiseShape[o]==null?e[o]:this.noiseShape[o]);return n}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);if(0<this.rate&&this.rate<1){let o=e.training==null?!1:e.training,s=this.getNoiseShape(n);return xu(()=>Ay(n,this.rate,s,this.seed),()=>n,o)}return t})}getConfig(){let t={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},e=super.getConfig();return Object.assign(t,e),t}dispose(){return super.dispose()}};vc.className="Dropout";Q.registerClass(vc);var cf=class extends vc{constructor(t){super(t),this.inputSpec=[{ndim:3}]}getNoiseShape(t){let e=t.shape;return[e[0],1,e[2]]}};cf.className="SpatialDropout1D";Q.registerClass(cf);var pf=class extends $t{constructor(t){if(super(t),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",t.batchInputShape==null&&t.inputShape==null&&t.inputDim!=null){let e=null;t.batchSize!=null&&(e=t.batchSize),this.batchInputShape=[e,t.inputDim]}this.units=t.units,Ze(this.units,"units"),this.activation=Xs(t.activation),t.useBias!=null&&(this.useBias=t.useBias),this.kernelInitializer=de(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=de(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=Be(t.kernelConstraint),this.biasConstraint=Be(t.biasConstraint),this.kernelRegularizer=be(t.kernelRegularizer),this.biasRegularizer=be(t.biasRegularizer),this.activityRegularizer=be(t.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(t){t=Bt(t);let e=t[t.length-1];this.kernel==null&&(this.kernel=this.addWeight("kernel",[e,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:e}}],this.built=!0}computeOutputShape(t){t=Bt(t);let e=t.slice();return e[e.length-1]=this.units,e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=Iy(this.activation.getClassName()),s;return o!=null?s=To(n,this.kernel.read(),o,this.bias?this.bias.read():null):(s=To(n,this.kernel.read()),this.bias!=null&&(s=fn(s,this.bias.read())),this.activation!=null&&(s=this.activation.apply(s))),s})}getConfig(){let t={units:this.units,activation:js(this.activation),useBias:this.useBias,kernelInitializer:Te(this.kernelInitializer),biasInitializer:Te(this.biasInitializer),kernelRegularizer:me(this.kernelRegularizer),biasRegularizer:me(this.biasRegularizer),activityRegularizer:me(this.activityRegularizer),kernelConstraint:ze(this.kernelConstraint),biasConstraint:ze(this.biasConstraint)},e=super.getConfig();return Object.assign(t,e),t}};pf.className="Dense";Q.registerClass(pf);var mf=class extends $t{constructor(t){t=t||{},super(t),this.inputSpec=[{minNDim:3}],this.dataFormat=t.dataFormat}computeOutputShape(t){t=Bt(t);for(let e of t.slice(1))if(e==null)throw new M(`The shape of the input to "Flatten" is not fully defined (got ${t.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[t[0],No(t,1)]}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);if(this.dataFormat==="channelsFirst"&&n.rank>1){let o=[0];for(let s=2;s<n.rank;++s)o.push(s);o.push(1),n=Ot(n,o)}return R$(n)})}getConfig(){let t={};this.dataFormat!=null&&(t.dataFormat=this.dataFormat);let e=super.getConfig();return Object.assign(t,e),t}};mf.className="Flatten";Q.registerClass(mf);var ff=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.activation=Xs(t.activation)}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return this.activation.apply(n)})}getConfig(){let t={activation:js(this.activation)},e=super.getConfig();return Object.assign(t,e),t}};ff.className="Activation";Q.registerClass(ff);var df=class extends $t{constructor(t){super(t),this.n=t.n,this.inputSpec=[{ndim:2}]}computeOutputShape(t){return[t[0],this.n,t[1]]}call(t,e){return B(()=>(t=Nt(t),$$(t,this.n)))}getConfig(){let t={n:this.n},e=super.getConfig();return Object.assign(t,e),t}};df.className="RepeatVector";Q.registerClass(df);var hf=class extends $t{constructor(t){super(t),this.targetShape=t.targetShape;for(let e=0;e<this.targetShape.length;++e)this.isUnknown(this.targetShape[e])&&(this.targetShape[e]=null)}isUnknown(t){return t<0||t==null}fixUnknownDimension(t,e){let n="Total size of new array must be unchanged.",o=e.slice(),s=1,i=null;for(let u=0;u<o.length;++u){let l=o[u];if(this.isUnknown(l))if(i===null)i=u;else throw new M("Can only specifiy one unknown dimension.");else s*=l}let a=No(t);if(i!==null){if(s===0||a%s!==0)throw new M(n);o[i]=a/s}else if(a!==s)throw new M(n);return o}computeOutputShape(t){let e=!1;for(let n=0;n<t.length;++n)if(this.isUnknown(t[n])){e=!0;break}return e?t.slice(0,1).concat(this.targetShape):t.slice(0,1).concat(this.fixUnknownDimension(t.slice(1),this.targetShape))}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=n.shape,s=o.slice(0,1).concat(this.fixUnknownDimension(o.slice(1),this.targetShape));return R(n,s)})}getConfig(){let t={targetShape:this.targetShape},e=super.getConfig();return Object.assign(t,e),t}};hf.className="Reshape";Q.registerClass(hf);var gf=class extends $t{constructor(t){if(super(t),t.dims==null)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(t.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${t.dims} instead.`);let e=Zr(1,t.dims.length+1);if(!y.arraysEqual(t.dims.slice().sort(),e))throw new Error("Invalid permutation `dims`: "+JSON.stringify(t.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=t.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new ye({ndim:this.dims.length+1})]}computeOutputShape(t){t=Bt(t);let e=t.slice();return this.dims.forEach((n,o)=>{e[o+1]=t[n]}),e}call(t,e){return Ot(Nt(t),this.dimsIncludingBatch)}getConfig(){let t={dims:this.dims},e=super.getConfig();return Object.assign(t,e),t}};gf.className="Permute";Q.registerClass(gf);var xf=class extends $t{constructor(t){super(t==null?{}:t),this.supportsMasking=!0,t!=null?this.maskValue=t.maskValue==null?0:t.maskValue:this.maskValue=0}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={maskValue:this.maskValue};return Object.assign(e,t),e}computeMask(t,e){let n=Nt(t),o=-1;return qu(Bs(n,this.maskValue),o)}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t),o=-1,s=!0,i=qu(Bs(n,this.maskValue),o,s);return D(n,J(i,n.dtype))})}};xf.className="Masking";Q.registerClass(xf);var yf=class extends $t{constructor(t){if(super(t),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",t.batchInputShape==null&&t.inputShape==null){let e=null;t.batchSize!=null&&(e=t.batchSize),t.inputLength==null?this.batchInputShape=[e,null]:this.batchInputShape=[e].concat(xe(t.inputLength))}this.inputDim=t.inputDim,Ze(this.inputDim,"inputDim"),this.outputDim=t.outputDim,Ze(this.outputDim,"outputDim"),this.embeddingsInitializer=de(t.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=be(t.embeddingsRegularizer),this.activityRegularizer=be(t.activityRegularizer),this.embeddingsConstraint=Be(t.embeddingsConstraint),this.maskZero=t.maskZero,this.supportsMasking=t.maskZero,this.inputLength=t.inputLength}build(t){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(t){}computeMask(t,e){return B(()=>this.maskZero?(t=Nt(t),Bs(t,It(t))):null)}computeOutputShape(t){if(t=Bt(t),this.inputLength==null)return[...t,this.outputDim];let e=xe(this.inputLength);if(e.length!==t.length-1)throw new M(`"inputLength" is ${this.inputLength}, but received input shape has shape ${t}`);{let n=0;for(let o=0;o<e.length;++o){let s=e[o],i=t[o+1];if(s!=null&&i!=null&&s!==i)throw new M(`"inputLength" is ${this.inputLength}, but received input shape has shape ${t}`);s==null&&(e[n]=i),n++}}return[t[0],...e,this.outputDim]}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);n.dtype!=="int32"&&(n=no(n,"int32"));let o=_y(this.embeddings.read(),R(n,[n.size]));return R(o,Bt(this.computeOutputShape(n.shape)))})}getConfig(){let t={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Te(this.embeddingsInitializer),embeddingsRegularizer:me(this.embeddingsRegularizer),activityRegularizer:me(this.activityRegularizer),embeddingsConstraint:ze(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},e=super.getConfig();return Object.assign(t,e),t}};yf.className="Embedding";Q.registerClass(yf);var pl=class extends $t{constructor(t){super(t||{}),this.supportsMasking=!0}mergeFunction(t){throw new St}computeElementwiseOpOutputShape(t,e){if(t==null||e==null)return null;if(t.length<e.length)return this.computeElementwiseOpOutputShape(e,t);if(e.length===0)return t;let n=t.slice(0,t.length-e.length);for(let o=0;o<e.length;++o){let s=t[t.length-e.length+o],i=e[o];if(s==null||i==null||s<0||i<0)n.push(null);else if(s===1)n.push(i);else if(i===1)n.push(s);else{if(s!==i)throw new M("Operands could not be broadcast together with shapes "+JSON.stringify(t)+" "+JSON.stringify(e));n.push(s)}}return n}build(t){if(Array.isArray(t)&&!Array.isArray(t[0])&&(t=[Bt(t)]),t=t,t.length<2)throw new M(`A merge layer should be called on an Array of at least 2 inputs. Got ${t.length} input(s).`);let e=[];for(let s of t)s!=null&&s[0]!==null&&e.push(s[0]);if(e=vo(e),e.length>1)throw new M(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(t)}.`);let n=t[0]==null?null:t[0].slice(1);for(let s=1;s<t.length;++s){let i=t[s]==null?null:t[s].slice(1);n=this.computeElementwiseOpOutputShape(n,i)}let o=t.map(s=>s.length);t.indexOf(null)===-1&&vo(o).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(t,e){return B(()=>{if(t=t,this.reshapeRequired){let n=[],o=t.map(s=>s.rank);if(o.indexOf(null)===-1){let s=qs(o);for(let i of t){let a=i.rank;for(let u=0;u<s-a;++u)i=nl(i,1);n.push(i)}return this.mergeFunction(n)}else{let s=!1;for(let u of t){let l=u.rank;if(l==null){let c=u.shape,p=c[0],m=c.slice(1).concat([p]),f=R(u,[p].concat(No(c.slice(1))));f=Ot(f,[1,0]),f=R(f,m),n.push(f),s=!0}else if(l>1){let c=Zr(1,l).concat([0]);n.push(Ot(u,c)),s=!0}else n.push(u)}let i=this.mergeFunction(n),a=i.rank;if(s){if(a==null){let u=i.shape,l=u.length,c=u[l-1],p=[c].concat(u.slice(0,u.length-1));i=R(Ot(R(i,[-1,c]),[1,0]),p)}else if(a>1){let u=[a-1].concat(Zr(0,a-1));i=Ot(i,u)}}return i}}else return this.mergeFunction(t)})}computeOutputShape(t){t=t;let e;t[0]==null?e=null:e=t[0].slice(1);for(let o=1;o<t.length;++o){let s=t[o]==null?null:t[o].slice(1);e=this.computeElementwiseOpOutputShape(e,s)}let n=[];for(let o of t)o!=null&&o[0]!==null&&n.push(o[0]);return n=vo(n),n.length===1?e=n.concat(e):e=[null].concat(e),e}computeMask(t,e){return B(()=>{if(e==null)return null;if(!Array.isArray(e))throw new M("`mask` should be an Array");if(!Array.isArray(t))throw new M("`inputs` should be an Array");if(e.length!==t.length)throw new M(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${t.length} vs ${e.length})`);if(e.every(o=>o==null))return null;e=e.map(o=>o==null?o:rr(o,0));let n=e[0];for(let o=1;o<e.length-1;++o)n=Rr(n,e[o]);return n})}},bf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=X(e,t[n]);return e})}};bf.className="Add";Q.registerClass(bf);var wf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=D(e,t[n]);return e})}};wf.className="Multiply";Q.registerClass(wf);var Cf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0].clone();for(let n=1;n<t.length;++n)e=X(e,t[n]);return D(1/t.length,e)})}};Cf.className="Average";Q.registerClass(Cf);var If=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0];for(let n=1;n<t.length;++n)e=Sn(e,t[n]);return e})}};If.className="Maximum";Q.registerClass(If);var Sf=class extends pl{constructor(t){super(t)}mergeFunction(t){return B(()=>{let e=t[0];for(let n=1;n<t.length;++n)e=Mi(e,t[n]);return e})}};Sf.className="Minimum";Q.registerClass(Sf);var vf=class extends pl{constructor(t){super(t),this.DEFAULT_AXIS=-1,t==null&&(t={}),this.axis=t.axis==null?this.DEFAULT_AXIS:t.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(t){if(!(Array.isArray(t)&&Array.isArray(t[0]))||t.length===1)throw new M("A `Concatenate` layer should be called on a list of at least 2 inputs");t=t;let e=!0;for(let o of t)if(o!=null){e=!1;break}if(e)return;let n=[];for(let o=0;o<t.length;++o){let s=t[o].slice();s.splice(this.axis,1);let i=!1;for(let a of n)if(y.arraysEqual(a,s)){i=!0;break}i||n.push(s)}if(n.length>1)throw new M("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(t))}mergeFunction(t){return B(()=>Nm(t,this.axis))}computeOutputShape(t){if(!(Array.isArray(t)&&Array.isArray(t[0])))throw new M("A `Concatenate` layer should be called on a list of inputs.");let e=t,n=e[0].slice(),o=this.axis<0?n.length+this.axis:this.axis;for(let s of e.slice(1)){if(n[o]==null||s[o]==null){n[o]=null;break}n[o]+=s[o]}return n}computeMask(t,e){if(e==null)return null;if(!Array.isArray(e))throw new M("`mask` should be an array for Concatenate");if(!Array.isArray(t))throw new M("`inputs` should be an array for Concatenate");if(e.length!==t.length)throw new M(`Mismatch in the length of mask (${e.length}) and the legnth of inputs (${t.length})`);return B(()=>{let n=!0;if(e.forEach(i=>{if(i!=null){n=!1;return}}),n)return null;let o=[];for(let i=0;i<t.length;++i)e[i]==null?o.push(J(yr(t[i]),"bool")):e[i].rank<t[i].rank?o.push(rr(e[i],-1)):o.push(e[i]);let s=ne(o,this.axis);return Zp(s,-1,!1)})}getConfig(){let t={axis:this.axis},e=super.getConfig();return Object.assign(t,e),t}};vf.className="Concatenate";Q.registerClass(vf);function $h(r,t){for(;r<0;)r+=t;return r}function tY(r,t,e){if(r.shape.length>3||t.shape.length>3)throw new St("batchDot is not implemented for tensors of 4D or higher rank yet");if(y.assert(r.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${r.shape.length}`),y.assert(r.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof e=="number"&&(e=[e,e]),r.dtype==="complex64"||t.dtype==="complex64")throw new St("batchDot is not implemented for complex64-type Tensors yet.");let n=r.shape.length,o=t.shape.length;e==null&&(e=[n-1,o-2]);let s=e;return B(()=>{let i;if(n>o){i=n-o;let u=[];for(let l=0;l<i;++l)u.push(1);t=R(t,t.shape.concat(u))}else if(o>n){i=o-n;let u=[];for(let l=0;l<i;++l)u.push(1);r=R(r,r.shape.concat(u))}else i=0;let a;if(r.shape.length===2&&t.shape.length===2)s[0]===s[1]?a=ft(D(r,t),s[0]):a=ft(D(Ot(r,[1,0]),t),s[1]);else{let u=s[0]!==r.shape.length-1,l=s[1]===t.shape.length-1;a=Lt(r,t,u,l)}if(i>0){let u;n>o?u=n+o-3:u=n-1;let l=[];for(let c=u;c<u+i;++c)l.push(c);a=Mn(a,l)}return a.shape.length===1&&(a=rr(a,1)),a})}var Nf=class extends pl{constructor(t){super(t),this.axes=t.axes,this.normalize=t.normalize==null?!1:t.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(t){y.assert(Array.isArray(t)&&t.length===2&&Array.isArray(t[0])&&Array.isArray(t[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let e=t[0],n=t[1];if(e.length>3||n.length>3)throw new St("Dot layer does not support tensors of 4D or higher rank yet.");let o=this.interpretAxes(e,n);if(e[o[0]]!==n[o[1]])throw new M(`Dimension incompatibility: ${e[o[0]]} !== ${n[o[1]]}`)}mergeFunction(t){if(t.length!==2)throw new M(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${t.length} input(s).`);let e=t[0],n=t[1],o;return Array.isArray(this.axes)?o=this.axes.map((s,i)=>$h(s,t[i].shape.length)):o=[$h(this.axes,e.shape.length),$h(this.axes,n.shape.length)],this.normalize&&(e=Sh(e,o[0]),n=Sh(n,o[1])),tY(e,n,o)}interpretAxes(t,e){let n;return Array.isArray(this.axes)?n=this.axes:n=[$h(this.axes,t.length),$h(this.axes,e.length)],n}computeOutputShape(t){y.assert(Array.isArray(t)&&t.length===2&&Array.isArray(t[0])&&Array.isArray(t[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let e=t[0].slice(),n=t[1].slice();if(e.length>3||n.length>3)throw new St("Dot layer does not support tensors of 4D or higher rank yet.");let o=this.interpretAxes(e,n);e.splice(o[0],1),n.splice(o[1],1),n.splice(0,1);let s=e.concat(n);return s.length===1&&s.push(1),s}computeMask(t,e){return null}getConfig(){let t={axes:this.axes,normalize:this.normalize},e=super.getConfig();return Object.assign(t,e),t}};Nf.className="Dot";Q.registerClass(Nf);var Tf=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.stddev=t.stddev}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={stddev:this.stddev};return Object.assign(e,t),e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return xu(()=>X(Tm(n.shape,0,this.stddev),n),()=>n,e.training||!1)})}};Tf.className="GaussianNoise";Q.registerClass(Tf);var kf=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.rate=t.rate}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={rate:this.rate};return Object.assign(e,t),e}call(t,e){return B(()=>{this.invokeCallHook(t,e);let n=Nt(t);return this.rate>0&&this.rate<1?xu(()=>{let s=Math.sqrt(this.rate/(1-this.rate));return D(n,Tm(n.shape,1,s))},()=>n,e.training||!1):n})}};kf.className="GaussianDropout";Q.registerClass(kf);var Ef=class extends $t{constructor(t){super(t),this.supportsMasking=!0,this.rate=t.rate,this.noiseShape=t.noiseShape}_getNoiseShape(t){return this.noiseShape||Nt(t).shape}computeOutputShape(t){return t}getConfig(){let t=super.getConfig(),e={rate:this.rate};return Object.assign(e,t),e}call(t,e){return B(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(t);return xu(()=>{let s=Nt(t),i=1.6732632423543772,a=1.0507009873554805,u=-i*a,l=ln(zi(n),this.rate);l=no(l,"float32");let c=((1-this.rate)*(1+this.rate*u**2))**-.5,p=-c*u*this.rate,m=X(D(s,l),D(X(l,-1),u));return X(D(m,c),p)},()=>Nt(t),e.training||!1)}return t})}};Ef.className="AlphaDropout";Q.registerClass(Ef);function Dh(r,t,e,n,o,s=.001){let i;if(r.rank===2)i=xx(r,t,e,n,o,s);else if(r.rank===3)i=yx(r,t,e,n,o,s);else if(r.rank===4)i=bx(r,t,e,n,o,s);else throw new St(`batchNormalization is not implemented for array of rank ${r.rank} yet`);return i}function eY(r,t,e,n,o=.001){return B(()=>{let s=Zu(r,n),i=s.mean,a=s.variance;return[Dh(r,i,a,e,t,o),i,a]})}function rY(r,t,e,n,o=.001){return B(()=>{let s=Zu(r,n),i=s.mean,a=s.variance,u=[];for(let d of Zr(0,r.rank))n.indexOf(d)!==-1?u.push(1):u.push(r.shape[d]);let l=R(i,u),c=R(a,u),p=t==null?null:R(t,u),m=e==null?null:R(e,u);return[Dh(r,l,c,m,p,o),i,a]})}function nY(r,t,e,n,o=.001){return y.arraysEqual(n.slice().sort(),Zr(0,r.rank-1))?eY(r,t,e,n,o):rY(r,t,e,n,o)}var _f=class extends $t{constructor(t){t==null&&(t={}),super(t),this.supportsMasking=!0,this.axis=t.axis==null?-1:t.axis,this.momentum=t.momentum==null?.99:t.momentum,this.epsilon=t.epsilon==null?.001:t.epsilon,this.center=t.center==null?!0:t.center,this.scale=t.scale==null?!0:t.scale,this.betaInitializer=de(t.betaInitializer||"zeros"),this.gammaInitializer=de(t.gammaInitializer||"ones"),this.movingMeanInitializer=de(t.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=de(t.movingVarianceInitializer||"ones"),this.betaConstraint=Be(t.betaConstraint),this.gammaConstraint=Be(t.gammaConstraint),this.betaRegularizer=be(t.betaRegularizer),this.gammaRegularizer=be(t.gammaRegularizer)}build(t){t=Bt(t);let e=this.axis>=0?this.axis:this.axis+t.length,n=t[e];if(n==null)throw new M(`Axis ${e} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(t)}.`);this.inputSpec=[new ye({ndim:t.length,axes:{[e]:n}})];let o=[n];this.scale&&(this.gamma=this.addWeight("gamma",o,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",o,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",o,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",o,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(t,e){return B(()=>{let n=e.training==null?!1:e.training,o=Nt(t),s=o.shape,i=s.length,a=Zr(0,i),u=this.axis>=0?this.axis:this.axis+i;a.splice(u,1);let l=Io(1,i);l[u]=s[u];let c=a.slice();c.sort();let p=!y.arraysEqual(c,Zr(0,i).slice(0,i-1)),m=()=>{if(p){let b=R(this.movingMean.read(),l),w=R(this.movingVariance.read(),l),C=this.center?R(this.beta.read(),l):null,N=this.scale?R(this.gamma.read(),l):null;return Dh(o,b,w,C,N,this.epsilon)}else return Dh(o,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!n)return m();let[f,d,h]=nY(o,this.gamma.read(),this.beta.read(),a,this.epsilon),g=(b,w,C)=>{B(()=>{let N=1-C,_=b.read(),A=D(ct(_,w),N);b.write(ct(_,A))})};return(()=>{g(this.movingMean,d,this.momentum),g(this.movingVariance,h,this.momentum)})(),f})}getConfig(){let t={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Te(this.betaInitializer),gammaInitializer:Te(this.gammaInitializer),movingMeanInitializer:Te(this.movingMeanInitializer),movingVarianceInitializer:Te(this.movingVarianceInitializer),betaRegularizer:me(this.betaRegularizer),gammaRegularizer:me(this.gammaRegularizer),betaConstraint:ze(this.betaConstraint),gammaConstraint:ze(this.gammaConstraint)},e=super.getConfig();return Object.assign(t,e),t}};_f.className="BatchNormalization";Q.registerClass(_f);var Af=class extends $t{constructor(t){if(t==null&&(t={}),super(t),this.axis=t.axis==null?-1:t.axis,typeof this.axis=="number"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let e of this.axis)if(!Number.isInteger(e))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=t.epsilon==null?.001:t.epsilon,this.center=t.center==null?!0:t.center,this.scale=t.scale==null?!0:t.scale,this.betaInitializer=de(t.betaInitializer||"zeros"),this.gammaInitializer=de(t.gammaInitializer||"ones"),this.betaRegularizer=be(t.betaRegularizer),this.gammaRegularizer=be(t.gammaRegularizer),this.supportsMasking=!0}build(t){t=Bt(t);let e=t.length;typeof this.axis=="number"&&(this.axis=[this.axis]);for(let s=0;s<this.axis.length;++s)this.axis[s]<0&&(this.axis[s]+=e);for(let s of this.axis)if(s<0||s>=e)throw new Error(`Invalid axis: ${s}`);if(this.axis.length!==vo(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(s=>t[s]),o=!0;this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,o):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,o):this.beta=null,this.built=!0}call(t,e){let n=Nt(t),o=n.shape,s=o.length;return B(()=>{let{mean:a,variance:u}=Zu(n,this.axis,!0),l=Io(1,s);for(let h of this.axis)l[h]=o[h];let c=h=>h!=null&&h.shape.length!==s?R(h,l):h,p=this.scale?c(this.gamma.read()):null,m=this.center?c(this.beta.read()):null,f=[],d=[];for(let h=0;h<s;++h)this.axis.indexOf(h)!==-1?(f.push(o[h]),d.push(1)):(f.push(1),d.push(o[h]));return a=Dr(a,f),u=Dr(u,f),p!=null&&(p=Dr(p,d)),m!=null&&(m=Dr(m,d)),Dh(n,a,u,m,p,this.epsilon)})}getConfig(){let t={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Te(this.betaInitializer),gammaInitializer:Te(this.gammaInitializer),betaRegularizer:me(this.betaRegularizer),gammaRegularizer:me(this.gammaRegularizer)},e=super.getConfig();return Object.assign(t,e),t}};Af.className="LayerNormalization";Q.registerClass(Af);function oY(r,t,e){return B(()=>{if(r.rank!==4)throw new M(`temporalPadding expects input tensor to be 4-D, but received a ${r.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new M("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(e==null&&(e=mn()),e!=="channelsLast"&&e!=="channelsFirst")throw new M(`Unknown data format: ${e}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let n;return e==="channelsFirst"?n=[[0,0],[0,0],t[0],t[1]]:n=[[0,0],t[0],t[1],[0,0]],cn(r,n)})}var $f=class extends $t{constructor(t){if(t==null&&(t={}),super(t),this.dataFormat=t.dataFormat==null?mn():t.dataFormat,t.padding==null)this.padding=[[1,1],[1,1]];else if(typeof t.padding=="number")this.padding=[[t.padding,t.padding],[t.padding,t.padding]];else{if(t.padding=t.padding,t.padding.length!==2)throw new M(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${t.padding.length} array.`);let e,n;if(typeof t.padding[0]=="number")e=[t.padding[0],t.padding[0]],n=[t.padding[1],t.padding[1]];else{if(t.padding=t.padding,t.padding[0].length!==2)throw new M(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${t.padding[0].length} array.`);if(e=t.padding[0],t.padding[1].length!==2)throw new M(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${t.padding[1].length} array.`);n=t.padding[1]}this.padding=[e,n]}this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){t=Bt(t);let e,n;return this.dataFormat==="channelsFirst"?(t[2]!=null&&t[2]>=0?e=t[2]+this.padding[0][0]+this.padding[0][1]:e=null,t[3]!=null&&t[3]>=0?n=t[3]+this.padding[1][0]+this.padding[1][1]:n=null,[t[0],t[1],e,n]):(t[1]!=null&&t[1]>=0?e=t[1]+this.padding[0][0]+this.padding[0][1]:e=null,t[2]!=null&&t[2]>=0?n=t[2]+this.padding[1][0]+this.padding[1][1]:n=null,[t[0],e,n,t[3]])}call(t,e){return B(()=>oY(Nt(t),this.padding,this.dataFormat))}getConfig(){let t={padding:this.padding,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}};$f.className="ZeroPadding2D";Q.registerClass($f);function wb(r,t,e,n,o,s){return B(()=>{Fe(o),Iv(s),pn(n),e==null&&(e=[1,1]),n==null&&(n="valid"),o==null&&(o=mn()),s==null&&(s="max"),r=Ah(r,o);let i,a=n==="same"?"same":"valid";return s==="max"?i=ru(r,t,e,a):i=Yl(r,t,e,a),o==="channelsFirst"&&(i=Ot(i,[0,3,1,2])),i})}function CD(r,t,e,n,o,s){return B(()=>{Fe(o),Iv(s),pn(n),e==null&&(e=[1,1,1]),n==null&&(n="valid"),o==null&&(o=mn()),s==null&&(s="max"),r=zv(r,o);let i,a=n==="same"?"same":"valid";return s==="max"?i=Hx(r,t,e,a):i=gx(r,t,e,a),o==="channelsFirst"&&(i=Ot(i,[0,4,1,2,3])),i})}var hb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=2),super(t),typeof t.poolSize=="number")this.poolSize=[t.poolSize];else if(Array.isArray(t.poolSize)&&t.poolSize.length===1&&typeof t.poolSize[0]=="number")this.poolSize=t.poolSize;else throw new M(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(t.poolSize)}`);if(Ze(this.poolSize,"poolSize"),t.strides==null)this.strides=this.poolSize;else if(typeof t.strides=="number")this.strides=[t.strides];else if(Array.isArray(t.strides)&&t.strides.length===1&&typeof t.strides[0]=="number")this.strides=t.strides;else throw new M(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(t.strides)}`);Ze(this.strides,"strides"),this.padding=t.padding==null?"valid":t.padding,pn(this.padding),this.inputSpec=[new ye({ndim:3})]}computeOutputShape(t){t=Bt(t);let e=Nn(t[1],this.poolSize[0],this.padding,this.strides[0]);return[t[0],e,t[2]]}call(t,e){return B(()=>{this.invokeCallHook(t,e),t=nl(Nt(t),2);let n=this.poolingFunction(Nt(t),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return Mn(n,[2])})}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides},e=super.getConfig();return Object.assign(t,e),t}},Df=class extends hb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,"max")}};Df.className="MaxPooling1D";Q.registerClass(Df);var Rf=class extends hb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,"avg")}};Rf.className="AveragePooling1D";Q.registerClass(Rf);var gb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=[2,2]),super(t),this.poolSize=Array.isArray(t.poolSize)?t.poolSize:[t.poolSize,t.poolSize],t.strides==null)this.strides=this.poolSize;else if(Array.isArray(t.strides)){if(t.strides.length!==2)throw new M(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${t.strides.length}.`);this.strides=t.strides}else this.strides=[t.strides,t.strides];Ze(this.poolSize,"poolSize"),Ze(this.strides,"strides"),this.padding=t.padding==null?"valid":t.padding,this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Fe(this.dataFormat),pn(this.padding),this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){t=Bt(t);let e=this.dataFormat==="channelsFirst"?t[2]:t[1],n=this.dataFormat==="channelsFirst"?t[3]:t[2];return e=Nn(e,this.poolSize[0],this.padding,this.strides[0]),n=Nn(n,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat==="channelsFirst"?[t[0],t[1],e,n]:[t[0],e,n,t[3]]}call(t,e){return B(()=>(this.invokeCallHook(t,e),this.poolingFunction(Nt(t),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Ff=class extends gb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,"max")}};Ff.className="MaxPooling2D";Q.registerClass(Ff);var Of=class extends gb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),wb(t,e,n,o,s,"avg")}};Of.className="AveragePooling2D";Q.registerClass(Of);var xb=class extends $t{constructor(t){if(t.poolSize==null&&(t.poolSize=[2,2,2]),super(t),this.poolSize=Array.isArray(t.poolSize)?t.poolSize:[t.poolSize,t.poolSize,t.poolSize],t.strides==null)this.strides=this.poolSize;else if(Array.isArray(t.strides)){if(t.strides.length!==3)throw new M(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${t.strides.length}.`);this.strides=t.strides}else this.strides=[t.strides,t.strides,t.strides];Ze(this.poolSize,"poolSize"),Ze(this.strides,"strides"),this.padding=t.padding==null?"valid":t.padding,this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Fe(this.dataFormat),pn(this.padding),this.inputSpec=[new ye({ndim:5})]}computeOutputShape(t){t=Bt(t);let e=this.dataFormat==="channelsFirst"?t[2]:t[1],n=this.dataFormat==="channelsFirst"?t[3]:t[2],o=this.dataFormat==="channelsFirst"?t[4]:t[3];return e=Nn(e,this.poolSize[0],this.padding,this.strides[0]),n=Nn(n,this.poolSize[1],this.padding,this.strides[1]),o=Nn(o,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat==="channelsFirst"?[t[0],t[1],e,n,o]:[t[0],e,n,o,t[4]]}call(t,e){return B(()=>(this.invokeCallHook(t,e),this.poolingFunction(Nt(t),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let t={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Pf=class extends xb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),CD(t,e,n,o,s,"max")}};Pf.className="MaxPooling3D";Q.registerClass(Pf);var Lf=class extends xb{constructor(t){super(t)}poolingFunction(t,e,n,o,s){return Fe(s),pn(o),CD(t,e,n,o,s,"avg")}};Lf.className="AveragePooling3D";Q.registerClass(Lf);var yb=class extends $t{constructor(t){super(t),this.inputSpec=[new ye({ndim:3})]}computeOutputShape(t){return[t[0],t[2]]}call(t,e){throw new St}},Mf=class extends yb{constructor(t){super(t||{})}call(t,e){return B(()=>{let n=Nt(t);return ve(n,1)})}};Mf.className="GlobalAveragePooling1D";Q.registerClass(Mf);var zf=class extends yb{constructor(t){super(t||{})}call(t,e){return B(()=>{let n=Nt(t);return Ir(n,1)})}};zf.className="GlobalMaxPooling1D";Q.registerClass(zf);var bb=class extends $t{constructor(t){super(t),this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,Fe(this.dataFormat),this.inputSpec=[new ye({ndim:4})]}computeOutputShape(t){return t=t,this.dataFormat==="channelsLast"?[t[0],t[3]]:[t[0],t[1]]}call(t,e){throw new St}getConfig(){let t={dataFormat:this.dataFormat},e=super.getConfig();return Object.assign(t,e),t}},Bf=class extends bb{call(t,e){return B(()=>{let n=Nt(t);return this.dataFormat==="channelsLast"?ve(n,[1,2]):ve(n,[2,3])})}};Bf.className="GlobalAveragePooling2D";Q.registerClass(Bf);var Vf=class extends bb{call(t,e){return B(()=>{let n=Nt(t);return this.dataFormat==="channelsLast"?Ir(n,[1,2]):Ir(n,[2,3])})}};Vf.className="GlobalMaxPooling2D";Q.registerClass(Vf);var Cb=class extends $t{constructor(t){super(t),this.layer=t.layer}build(t){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(t){this.layer!=null&&(this.layer.trainable=t)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(t){this.layer.setWeights(t)}getConfig(){let t={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},e=super.getConfig();return Object.assign(t,e),t}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(t)}static fromConfig(t,e,n={}){let o=e.layer,s=gn(o,n);delete e.layer;let i={layer:s};return Object.assign(i,e),new t(i)}},Gf=class extends Cb{constructor(t){super(t),this.supportsMasking=!0}build(t){if(t=Bt(t),t.length<3)throw new M(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(t)}`);this.inputSpec=[{shape:t}];let e=[t[0]].concat(t.slice(2));this.layer.built||(this.layer.build(e),this.layer.built=!0),super.build(t)}computeOutputShape(t){t=Bt(t);let e=[t[0]].concat(t.slice(2)),n=this.layer.computeOutputShape(e),o=t[1];return[n[0],o].concat(n.slice(1))}call(t,e){return B(()=>(t=Nt(t),Vv((i,a)=>[Nt(this.layer.call(i,e)),[]],t,[],!1,null,null,!1,!0)[1]))}};Gf.className="TimeDistributed";Q.registerClass(Gf);function sY(r){Wi(T$,"BidirectionalMergeMode",r)}var iY="concat",Wf=class extends Cb{constructor(t){super(t);let e=t.layer.getConfig(),n={};n.className=t.layer.getClassName(),n.config=e,this.forwardLayer=gn(n),e.goBackwards=e.goBackwards!==!0;let o={};if(o.className=t.layer.getClassName(),o.config=e,this.backwardLayer=gn(o),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=t.mergeMode===void 0?iY:t.mergeMode,sY(this.mergeMode),t.weights)throw new St("weights support is not implemented for Bidirectional layer yet.");this._stateful=t.layer.stateful,this.returnSequences=t.layer.returnSequences,this.returnState=t.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=t.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(t){this._trainable=t,this.forwardLayer!=null&&(this.forwardLayer.trainable=t),this.backwardLayer!=null&&(this.backwardLayer.trainable=t)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(t){let e=t.length,n=Math.floor(e/2);this.forwardLayer.setWeights(t.slice(0,n)),this.backwardLayer.setWeights(t.slice(n))}computeOutputShape(t){let e=this.forwardLayer.computeOutputShape(t);Array.isArray(e)&&Array.isArray(e[0])||(e=[e]),e=e;let n,o,s;return this.returnState&&(s=e.slice(1)),n=e[0],n=n,this.mergeMode==="concat"?(n[n.length-1]*=2,o=[n]):this.mergeMode==null?o=[n,n.slice()]:o=[n],this.returnState?this.mergeMode==null?o.concat(s).concat(s.slice()):[n].concat(s).concat(s.slice()):Nr(o)}apply(t,e){let n=e==null?null:e.initialState,o=e==null?null:e.constants;e==null&&(e={});let s=Bv(t,n,o,this.numConstants);if(t=s.inputs,n=s.initialState,o=s.constants,Array.isArray(t)&&(n=t.slice(1),t=t[0]),(n==null||n.length===0)&&o==null)return super.apply(t,e);let i=[],a=[];if(n!=null){let l=n.length;if(l%2>0)throw new M("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");e.initialState=n,i.push(...n);let c=n.map(p=>new ye({shape:p.shape}));this.forwardLayer.stateSpec=c.slice(0,l/2),this.backwardLayer.stateSpec=c.slice(l/2),a.push(...c)}if(o!=null)throw new St("Support for constants in Bidirectional layers is not implemented yet.");let u=i[0]instanceof Jr;for(let l of i)if(l instanceof Jr!==u)throw new M("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(u){let l=[t].concat(i),c=this.inputSpec.concat(a),p=this.inputSpec;this.inputSpec=c;let m=super.apply(l,e);return this.inputSpec=p,m}else return super.apply(t,e)}call(t,e){return B(()=>{let n=e.initialState,o,s;if(n==null)o=this.forwardLayer.call(t,e),s=this.backwardLayer.call(t,e);else{let u=n.slice(0,n.length/2),l=n.slice(n.length/2);o=this.forwardLayer.call(t,Object.assign(e,{initialState:u})),s=this.backwardLayer.call(t,Object.assign(e,{initialState:l}))}let i;this.returnState&&(Array.isArray(o)&&(i=o.slice(1).concat(s.slice(1))),o=o[0],s=s[0]),this.returnSequences&&(s=pr(s,1));let a;return this.mergeMode==="concat"?a=Nm([o,s]):this.mergeMode==="sum"?a=X(o,s):this.mergeMode==="ave"?a=D(.5,X(o,s)):this.mergeMode==="mul"?a=D(o,s):this.mergeMode==null&&(a=[o,s]),this.returnState?this.mergeMode==null?a.concat(i):[a].concat(i):a})}resetStates(t){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(t){Hs(this.forwardLayer.name,()=>{this.forwardLayer.build(t)}),Hs(this.backwardLayer.name,()=>{this.backwardLayer.build(t)}),this.built=!0}computeMask(t,e){Array.isArray(e)&&(e=e[0]);let n;if(this.returnSequences?this.mergeMode==null?n=[e,e]:n=e:this.mergeMode==null?n=[null,null]:n=null,this.returnState){let s=this.forwardLayer.states.map(i=>null);return Array.isArray(n)?n.concat(s).concat(s):[n].concat(s).concat(s)}else return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(t){super.setFastWeightInitDuringBuild(t),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(t),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(t)}getConfig(){let t={mergeMode:this.mergeMode},e=super.getConfig();return Object.assign(t,e),t}static fromConfig(t,e){let n=gn(e.layer);if(delete e.layer,e.numConstants!=null)throw new St("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let o=e;return o.layer=n,new t(o)}};Wf.className="Bidirectional";Q.registerClass(Wf);var Uf=class extends $t{constructor(t){super(t),this.scale=t.scale,t.offset?this.offset=t.offset:this.offset=0}getConfig(){let t={scale:this.scale,offset:this.offset},e=super.getConfig();return Object.assign(t,e),t}call(t,e){return B(()=>(t=Nt(t),t.dtype!=="float32"&&(t=no(t,"float32")),X(D(t,this.scale),this.offset)))}};Uf.className="Rescaling";Q.registerClass(Uf);var aY=["bilinear","nearest"],ID=new Set(aY),Hf=class extends $t{constructor(t){if(super(t),this.height=t.height,this.width=t.width,t.interpolation)if(ID.has(t.interpolation))this.interpolation=t.interpolation;else throw new M(`Invalid interpolation parameter: ${t.interpolation} is not implemented`);else this.interpolation="bilinear";this.cropToAspectRatio=Boolean(t.cropToAspectRatio)}computeOutputShape(t){t=Bt(t);let e=t[2];return[this.height,this.width,e]}getConfig(){let t={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},e=super.getConfig();return Object.assign(t,e),t}call(t,e){return B(()=>{let n=[this.height,this.width];if(this.interpolation==="bilinear")return Gs.resizeBilinear(t,n,!this.cropToAspectRatio);if(this.interpolation==="nearest")return Gs.resizeNearestNeighbor(t,n,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...ID]} are supported`)})}};Hf.className="Resizing";Q.registerClass(Hf);function SD(r,t,e,n){let o=Nt(r);if(o.dtype!=="int32"&&(o=no(o,"int32")),t==="int")return o;let s=o.shape;if(o.rank===0&&(o=rr(o,-1)),t==="oneHot"&&o.shape[o.shape.length-1]!==1&&(o=rr(o,-1)),o.rank>2)throw new M(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${s} which would result in output rank ${o.rank}.`);let i=["multiHot","oneHot"].includes(t),a=o,u;if(typeof n!="undefined"&&t==="count"?u=ch(a,n,e,i):u=ch(a,[],e,i),t!=="tfIdf")return u;if(n)return D(u,n);throw new M("When outputMode is 'tfIdf', weights must be provided.")}var qf=class extends $t{constructor(t){super(t),this.numTokens=t.numTokens,t.outputMode?this.outputMode=t.outputMode:this.outputMode="multiHot"}getConfig(){let t={numTokens:this.numTokens,outputMode:this.outputMode},e=super.getConfig();return Object.assign(t,e),t}computeOutputShape(t){return t=Bt(t),t==null?[this.numTokens]:this.outputMode==="oneHot"&&t[t.length-1]!==1?(t.push(this.numTokens),t):(t[t.length-1]=this.numTokens,t)}call(t,e){return B(()=>{t=Nt(t),t.dtype!=="int32"&&(t=no(t,"int32"));let n;if(typeof e.countWeights!="undefined"){if(this.outputMode!=="count")throw new M(`countWeights is not used when outputMode !== count.
-              Received countWeights=${e.countWeights}`);n=Nt(e.countWeights)}let o=Ir(t),s=Ja(t),i=Re(this.numTokens,o).bufferSync().get(0),a=ln(s,0).bufferSync().get(0);if(!(i&&a))throw new M(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return SD(t,this.outputMode,this.numTokens,n)})}};qf.className="CategoryEncoding";Q.registerClass(qf);function uY(r){return new Ks(r)}function cY(r){return new Ym(r)}function pY(r){return new Km(r)}function mY(r){return new jm(r)}function fY(r){return new Xm(r)}function dY(r){return new Jm(r)}function hY(r){return new Zm(r)}function gY(r){return new Su(r)}function xY(r){return new il(r)}function yY(r){return new Qm(r)}function bY(r){return new al(r)}function wY(r){return new tf(r)}function CY(r){return new ef(r)}function IY(r){return new rf(r)}function SY(r){return new nf(r)}function vY(r){return new of(r)}function NY(r){return new ff(r)}function TY(r){return new pf(r)}function kY(r){return new vc(r)}function EY(r){return new cf(r)}function _Y(r){return new mf(r)}function AY(r){return new df(r)}function $Y(r){return new hf(r)}function DY(r){return new gf(r)}function RY(r){return new yf(r)}function FY(r){return new bf(r)}function OY(r){return new Cf(r)}function PY(r){return new vf(r)}function LY(r){return new If(r)}function MY(r){return new Sf(r)}function zY(r){return new wf(r)}function BY(r){return new Nf(r)}function VY(r){return new _f(r)}function GY(r){return new Af(r)}function WY(r){return new $f(r)}function Gv(r){return new Rf(r)}function UY(r){return Gv(r)}function HY(r){return Gv(r)}function Wv(r){return new Of(r)}function qY(r){return Wv(r)}function KY(r){return Wv(r)}function Uv(r){return new Lf(r)}function jY(r){return Uv(r)}function XY(r){return Uv(r)}function YY(r){return new Mf(r)}function ZY(r){return new Bf(r)}function vD(r){return new zf(r)}function ND(r){return new Vf(r)}function TD(r){return new Df(r)}function kD(r){return new Ff(r)}function JY(r){return new Pf(r)}function QY(r){return new af(r)}function tZ(r){return new Cc(r)}function eZ(r){return new lf(r)}function rZ(r){return new ul(r)}function nZ(r){return new sf(r)}function oZ(r){return new wc(r)}function sZ(r){return new uf(r)}function iZ(r){return new Sc(r)}function aZ(r){return new Tn(r)}function lZ(r){return new Ic(r)}function uZ(r){return new Wf(r)}function cZ(r){return new Gf(r)}var pZ=vD,mZ=ND,fZ=TD,dZ=kD;function hZ(r){return new Tf(r)}function gZ(r){return new kf(r)}function xZ(r){return new Ef(r)}function yZ(r){return new xf(r)}function bZ(r){return new Uf(r)}function wZ(r){return new Hf(r)}function CZ(r){return new qf(r)}var _D={};Wt(_D,{MAPE:()=>DZ,MSE:()=>OZ,binaryAccuracy:()=>IZ,binaryCrossentropy:()=>SZ,categoricalAccuracy:()=>NZ,categoricalCrossentropy:()=>TZ,cosineProximity:()=>_Z,mape:()=>RZ,meanAbsoluteError:()=>AZ,meanAbsolutePercentageError:()=>$Z,meanSquaredError:()=>FZ,mse:()=>PZ,precision:()=>kZ,recall:()=>EZ,sparseCategoricalAccuracy:()=>vZ});function IZ(r,t){return Th(r,t)}function SZ(r,t){return Wy(r,t)}function vZ(r,t){return Uy(r,t)}function NZ(r,t){return kh(r,t)}function TZ(r,t){return Eh(r,t)}function kZ(r,t){return Av(r,t)}function EZ(r,t){return J$(r,t)}function _Z(r,t){return Nh(r,t)}function AZ(r,t){return Vm(r,t)}function $Z(r,t){return bu(r,t)}function DZ(r,t){return bu(r,t)}function RZ(r,t){return bu(r,t)}function FZ(r,t){return Hi(r,t)}function OZ(r,t){return Hi(r,t)}function PZ(r,t){return Hi(r,t)}var AD={};Wt(AD,{modelFromJSON:()=>dD});var $D={};Wt($D,{l1:()=>MZ,l1l2:()=>LZ,l2:()=>zZ});function LZ(r){return new wu(r)}function MZ(r){return yD(r)}function zZ(r){return bD(r)}var Sb=class extends sl{constructor(){super(...arguments),this.model=null}setModel(t){if(!(t instanceof Bn))throw new Error("model must be a LayersModel, not some other Container");this.model=t}};function Ib(r,t){return r<t}function DD(r,t){return r>t}var vb=class extends Sb{constructor(t){if(super(),t==null&&(t={}),t.restoreBestWeights)throw new St("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=t.monitor||"val_loss",this.minDelta=Math.abs(t.minDelta||0),this.patience=t.patience||0,this.verbose=t.verbose||0,this.mode=t.mode||"auto",this.baseline=t.baseline,["auto","min","max"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),this.mode==="min"?this.monitorFunc=Ib:this.mode==="max"?this.monitorFunc=DD:this.monitor.indexOf("acc")!==-1?this.monitorFunc=DD:this.monitorFunc=Ib,this.monitorFunc===Ib&&(this.minDelta*=-1)}async onTrainBegin(t){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===Ib?1/0:-1/0}async onEpochEnd(t,e){await Ui(e);let n=this.getMonitorValue(e);n!=null&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=t,this.model.stopTraining=!0)))}async onTrainEnd(t){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(t){t==null&&(t={});let e=t[this.monitor];return e==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(t)}`),e}};function BZ(r){return new vb(r)}var VZ={earlyStopping:BZ};var GZ=z();GZ.registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,r=>{r&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")});var oo;(function(r){r[r.DT_INVALID=0]="DT_INVALID",r[r.DT_FLOAT=1]="DT_FLOAT",r[r.DT_DOUBLE=2]="DT_DOUBLE",r[r.DT_INT32=3]="DT_INT32",r[r.DT_UINT8=4]="DT_UINT8",r[r.DT_INT16=5]="DT_INT16",r[r.DT_INT8=6]="DT_INT8",r[r.DT_STRING=7]="DT_STRING",r[r.DT_COMPLEX64=8]="DT_COMPLEX64",r[r.DT_INT64=9]="DT_INT64",r[r.DT_BOOL=10]="DT_BOOL",r[r.DT_QINT8=11]="DT_QINT8",r[r.DT_QUINT8=12]="DT_QUINT8",r[r.DT_QINT32=13]="DT_QINT32",r[r.DT_BFLOAT16=14]="DT_BFLOAT16",r[r.DT_QINT16=15]="DT_QINT16",r[r.DT_QUINT16=16]="DT_QUINT16",r[r.DT_UINT16=17]="DT_UINT16",r[r.DT_COMPLEX128=18]="DT_COMPLEX128",r[r.DT_HALF=19]="DT_HALF",r[r.DT_RESOURCE=20]="DT_RESOURCE",r[r.DT_VARIANT=21]="DT_VARIANT",r[r.DT_UINT32=22]="DT_UINT32",r[r.DT_UINT64=23]="DT_UINT64",r[r.DT_FLOAT_REF=101]="DT_FLOAT_REF",r[r.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",r[r.DT_INT32_REF=103]="DT_INT32_REF",r[r.DT_UINT8_REF=104]="DT_UINT8_REF",r[r.DT_INT16_REF=105]="DT_INT16_REF",r[r.DT_INT8_REF=106]="DT_INT8_REF",r[r.DT_STRING_REF=107]="DT_STRING_REF",r[r.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",r[r.DT_INT64_REF=109]="DT_INT64_REF",r[r.DT_BOOL_REF=110]="DT_BOOL_REF",r[r.DT_QINT8_REF=111]="DT_QINT8_REF",r[r.DT_QUINT8_REF=112]="DT_QUINT8_REF",r[r.DT_QINT32_REF=113]="DT_QINT32_REF",r[r.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",r[r.DT_QINT16_REF=115]="DT_QINT16_REF",r[r.DT_QUINT16_REF=116]="DT_QUINT16_REF",r[r.DT_UINT16_REF=117]="DT_UINT16_REF",r[r.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",r[r.DT_HALF_REF=119]="DT_HALF_REF",r[r.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",r[r.DT_VARIANT_REF=121]="DT_VARIANT_REF",r[r.DT_UINT32_REF=122]="DT_UINT32_REF",r[r.DT_UINT64_REF=123]="DT_UINT64_REF"})(oo||(oo={}));var RD;(function(r){let t;(function(e){e[e.LEGACY=0]="LEGACY",e[e.V1=1]="V1",e[e.V2=2]="V2"})(t=r.CheckpointFormatVersion||(r.CheckpointFormatVersion={}))})(RD||(RD={}));var Hv={};function UZ(r,t){let e={tfOpName:r,category:"custom",inputs:[],attrs:[],customExecutor:t};Hv[r]=e}function Nb(r){return Hv[r]}function HZ(r){delete Hv[r]}function S(r,t,e,n,o){let s=t.inputParams[r];if(s&&s.inputIndexStart!==void 0){let a=s.inputIndexStart,u=s.inputIndexEnd===0?void 0:s.inputIndexEnd===void 0?a+1:s.inputIndexEnd;if(s.type==="tensor")return br(t.inputNames[s.inputIndexStart],e,n,o);if(s.type==="tensors")return t.inputNames.slice(a,u).map(m=>br(m,e,n,o));let l=br(t.inputNames.slice(a)[0],e,n,o),c=l.dataSync();return s.type==="number"?c[0]:y.toNestedArray(l.shape,c)}let i=t.attrParams[r];return i&&i.value}function br(r,t,e,n){let[o,s]=xn(r);if(n!=null){let a=n.getHashTableHandleByName(o);if(a!=null)return a}let i=e.currentContextIds.find(a=>!!t[Tb(o,a)]);return i!==void 0?t[Tb(o,i)][s]:void 0}function FD(r,t,e){return t[Tb(r,e.currentContextId)]}function _o(r,t){let[e,n,o]=xn(r);return[Tb(e,t&&t.currentContextId),n,o]}function Tb(r,t){return t?`${r}-${t}`:r}function xn(r){let t=r.split(":");if(t.length===1)return[r,0,void 0];let e=t[0],n=t.length===3?t[1]:void 0,o=Number(t[t.length-1]);return[e,o,n]}function Rh(r,t,e){let n=S("pad",r,t,e);if(n==="explicit"){n=S("explicitPaddings",r,t,e);let o=[[0,0],[0,0],[0,0],[0,0]];for(let s=0;s<4;s++)o[s][0]=n[s*2],o[s][1]=n[s*2+1];return o}return n}function Zs(r){return r.kept?r:sn(r)}var qv={};Wt(qv,{json:()=>qZ});var qZ=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}];var Kv={};Wt(Kv,{json:()=>KZ});var KZ=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}];var jv={};Wt(jv,{json:()=>jZ});var jZ=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}];var Xv={};Wt(Xv,{json:()=>XZ});var XZ=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}];var Yv={};Wt(Yv,{json:()=>YZ});var YZ=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomStandardNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}];var Zv={};Wt(Zv,{json:()=>ZZ});var ZZ=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}];var Jv={};Wt(Jv,{json:()=>JZ});var JZ=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}];var Qv={};Wt(Qv,{json:()=>QZ});var QZ=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}];var tN={};Wt(tN,{json:()=>t7});var t7=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}];var eN={};Wt(eN,{json:()=>e7});var e7=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}];var rN={};Wt(rN,{json:()=>r7});var r7=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}];var nN={};Wt(nN,{json:()=>n7});var n7=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}];var oN={};Wt(oN,{json:()=>o7});var o7=[{tfOpName:"EuclideanNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",defaultValue:!1}]},{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}];var sN={};Wt(sN,{json:()=>s7});var s7=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}];var iN={};Wt(iN,{json:()=>i7});var i7=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}];var aN={};Wt(aN,{json:()=>a7});var a7=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}];var lN={};Wt(lN,{json:()=>l7});var l7=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}];var uN={};Wt(uN,{json:()=>u7});var u7=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}];var cN={};Wt(cN,{json:()=>c7});var c7=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}];var Fh=class{constructor(){let t=[qv,Kv,jv,Xv,Yv,Zv,Jv,Qv,tN,eN,rN,nN,oN,sN,iN,aN,lN,uN,cN],e=[].concat(...t.map(n=>n.json));this.opMappers=e.reduce((n,o)=>(n[o.tfOpName]=o,n),{})}static get Instance(){return this._instance||(this._instance=new this)}transformGraph(t,e={}){let n=t.node,o=[],s=[],i=[],a=n.reduce((h,g)=>(h[g.name]=this.mapNode(g),g.op.startsWith("Placeholder")?o.push(h[g.name]):g.op==="Const"?s.push(h[g.name]):(g.input==null||g.input.length===0)&&i.push(h[g.name]),h),{}),u=[],l=[],c={},p={};e!=null&&(c=this.mapSignatureEntries(e.inputs),p=this.mapSignatureEntries(e.outputs));let m=Object.keys(a);m.forEach(h=>{let g=a[h];g.inputNames.forEach((x,b)=>{let[w,,C]=_o(x),N=a[w];if(N.outputs!=null){let _=N.outputs.indexOf(C);if(_!==-1){let A=`${w}:${_}`;g.inputNames[b]=A}}g.inputs.push(N),N.children.push(g)})}),Object.keys(p).length===0?m.forEach(h=>{let g=a[h];g.children.length===0&&l.push(g)}):Object.keys(p).forEach(h=>{let[g]=_o(h),x=a[g];x!=null&&(x.signatureKey=p[h],l.push(x))}),Object.keys(c).length>0?Object.keys(c).forEach(h=>{let[g]=_o(h),x=a[g];x&&(x.signatureKey=c[h],u.push(x))}):u=o;let f={};t.library!=null&&t.library.function!=null&&(f=t.library.function.reduce((h,g)=>(h[g.signature.name]=this.mapFunction(g),h),{}));let d={nodes:a,inputs:u,outputs:l,weights:s,placeholders:o,signature:e,functions:f};return i.length>0&&(d.initNodes=i),d}mapSignatureEntries(t){return Object.keys(t||{}).reduce((e,n)=>(e[t[n].name]=n,e),{})}mapNode(t){let e=Nb(t.op)||this.opMappers[t.op]||{};t.attr==null&&(t.attr={});let n={name:t.name,op:t.op,category:e.category,inputNames:(t.input||[]).map(o=>o.startsWith("^")?o.slice(1):o),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:t.attr,outputs:e.outputs};return e.inputs!=null&&(n.inputParams=e.inputs.reduce((o,s)=>(o[s.name]={type:s.type,inputIndexStart:s.start,inputIndexEnd:s.end},o),{})),e.attrs!=null&&(n.attrParams=e.attrs.reduce((o,s)=>{let i=s.type,a;switch(s.type){case"string":a=kb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=kb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"string[]":a=Fb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Fb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"number":a=_b(t.attr,s.tfName,s.defaultValue||0),a===void 0&&!!s.tfDeprecatedName&&(a=_b(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"number[]":a=Rb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Rb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"bool":a=Eb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Eb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"bool[]":a=Pb(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Pb(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"shape":a=Db(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Db(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"shape[]":a=Ob(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Ob(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"dtype":a=Ab(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=Ab(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"dtype[]":a=$b(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=$b(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"func":a=OD(t.attr,s.tfName,s.defaultValue),a===void 0&&!!s.tfDeprecatedName&&(a=OD(t.attr,s.tfDeprecatedName,s.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${s.type} for op: ${t.op}`)}return o[s.name]={value:a,type:i},o},{})),n}mapFunction(t){let e=t.nodeDef,n=[],o=[],s={};e!=null&&(s=e.reduce((p,m)=>(p[m.name]=this.mapNode(m),m.op==="Const"&&o.push(p[m.name]),p),{}));let i=[],a=[];t.signature.inputArg.forEach(p=>{let[m]=_o(p.name),f={name:m,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:pN(p.type),type:"dtype"}},children:[]};f.signatureKey=p.name,i.push(f),s[m]=f}),Object.keys(s).forEach(p=>{let m=s[p];m.inputNames.forEach((f,d)=>{let[h,,g]=_o(f),x=s[h];if(x.outputs!=null){let b=x.outputs.indexOf(g);if(b!==-1){let w=`${h}:${b}`;m.inputNames[d]=w}}m.inputs.push(x),x.children.push(m)})});let l=t.ret;t.signature.outputArg.forEach(p=>{let[m,f]=_o(l[p.name]),d=s[m];d!=null&&(d.defaultOutput=f,a.push(d))});let c=this.mapArgsToSignature(t);return{nodes:s,inputs:i,outputs:a,weights:o,placeholders:n,signature:c}}mapArgsToSignature(t){return{methodName:t.signature.name,inputs:t.signature.inputArg.reduce((e,n)=>(e[n.name]=this.mapArgToTensorInfo(n),e),{}),outputs:t.signature.outputArg.reduce((e,n)=>(e[n.name]=this.mapArgToTensorInfo(n,t.ret),e),{})}}mapArgToTensorInfo(t,e){let n=t.name;return e!=null&&(n=e[n]),{name:n,dtype:t.type}}};function p7(r){let t=z().global;if(typeof t.atob!="undefined")return t.atob(r);if(typeof Buffer!="undefined")return new Buffer(r,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}function PD(r,t){let e=Array.isArray(r)?String.fromCharCode.apply(null,r):p7(r);return t?e:e.toLowerCase()}function kb(r,t,e,n=!1){let o=r[t];return o!=null?PD(o.s,n):e}function Eb(r,t,e){let n=r[t];return n?n.b:e}function _b(r,t,e){let n=r[t]||{},o=n.i!=null?n.i:n.f!=null?n.f:e;return typeof o=="number"?o:parseInt(o,10)}function pN(r){switch(typeof r=="string"&&(r=oo[r]),r){case oo.DT_FLOAT:case oo.DT_HALF:return"float32";case oo.DT_INT32:case oo.DT_INT64:case oo.DT_INT8:case oo.DT_UINT8:return"int32";case oo.DT_BOOL:return"bool";case oo.DT_DOUBLE:return"float32";case oo.DT_STRING:return"string";default:return null}}function OD(r,t,e){let n=r[t];return n&&n.func?n.func.name:e}function Ab(r,t,e){let n=r[t];return n&&n.type?pN(n.type):e}function $b(r,t,e){let n=r[t];return n&&n.list&&n.list.type?n.list.type.map(o=>pN(o)):e}function LD(r){if(!r.unknownRank)return r.dim!=null?r.dim.map(t=>typeof t.size=="number"?t.size:parseInt(t.size,10)):[]}function Db(r,t,e){let n=r[t];return n&&n.shape?LD(n.shape):e}function Rb(r,t,e){let n=r[t];return n?((n.list.f&&n.list.f.length?n.list.f:n.list.i)||[]).map(o=>typeof o=="number"?o:parseInt(o,10)):e}function Fb(r,t,e,n=!1){let o=r[t];return o&&o.list&&o.list.s?o.list.s.map(s=>PD(s,n)):e}function Ob(r,t,e){let n=r[t];return n&&n.list&&n.list.shape?n.list.shape.map(o=>LD(o)):e}function Pb(r,t,e){let n=r[t];return n&&n.list&&n.list.b?n.list.b:e}var Lb=class{constructor(t,e,n){this.node=t,this.tensorMap=e,this.context=n,this.inputs=[],this.attrs={},this.inputs=t.inputNames.map(o=>this.getInput(o)),t.rawAttrs!=null&&(this.attrs=Object.keys(t.rawAttrs).reduce((o,s)=>(o[s]=this.getAttr(s),o),{}))}getInput(t){return br(t,this.tensorMap,this.context)}getAttr(t,e){let n=this.node.rawAttrs[t];if(n.tensor!=null)return br(t,this.tensorMap,this.context);if(n.i!=null||n.f!=null)return _b(this.node.rawAttrs,t,e);if(n.s!=null)return kb(this.node.rawAttrs,t,e);if(n.b!=null)return Eb(this.node.rawAttrs,t,e);if(n.shape!=null)return Db(this.node.rawAttrs,t,e);if(n.type!=null)return Ab(this.node.rawAttrs,t,e);if(n.list!=null){if(n.list.i!=null||n.list.f!=null)return Rb(this.node.rawAttrs,t,e);if(n.list.s!=null)return Fb(this.node.rawAttrs,t,e);if(n.list.shape!=null)return Ob(this.node.rawAttrs,t,e);if(n.list.b!=null)return Pb(this.node.rawAttrs,t,e);if(n.list.type!=null)return $b(this.node.rawAttrs,t,e)}return e}};var ae={};Wt(ae,{OP_SCOPE_SUFFIX:()=>k0,abs:()=>Ee,acos:()=>ax,acosh:()=>lx,add:()=>X,addN:()=>LE,all:()=>Zp,any:()=>qu,argMax:()=>Ai,argMin:()=>ux,asin:()=>cx,asinh:()=>px,atan:()=>mx,atan2:()=>fx,atanh:()=>dx,avgPool:()=>Yl,avgPool3d:()=>gx,basicLSTMCell:()=>BE,batchNorm:()=>Di,batchNorm2d:()=>xx,batchNorm3d:()=>yx,batchNorm4d:()=>bx,batchToSpaceND:()=>Zl,bincount:()=>wx,booleanMaskAsync:()=>n6,broadcastArgs:()=>GE,broadcastTo:()=>Ri,buffer:()=>wt,cast:()=>J,ceil:()=>Cx,clipByValue:()=>Cr,clone:()=>sn,complex:()=>wn,concat:()=>ne,concat1d:()=>Ix,concat2d:()=>Sx,concat3d:()=>vx,concat4d:()=>Nx,conv1d:()=>Qp,conv2d:()=>In,conv2dTranspose:()=>em,conv3d:()=>Tx,conv3dTranspose:()=>Ex,cos:()=>Jl,cosh:()=>rm,cosineWindow:()=>hh,cumprod:()=>Xu,cumsum:()=>nm,denseBincount:()=>ch,depthToSpace:()=>_x,depthwiseConv2d:()=>Fi,diag:()=>WE,dilation2d:()=>Ax,div:()=>pt,divNoNan:()=>$x,dot:()=>Dx,dropout:()=>lv,einsum:()=>UE,elu:()=>Oi,enclosingPowerOfTwo:()=>uv,equal:()=>$r,erf:()=>Rx,euclideanNorm:()=>Fx,exp:()=>er,expandDims:()=>rr,expm1:()=>Ox,eye:()=>Yu,fft:()=>au,fill:()=>xo,floor:()=>Pi,floorDiv:()=>Yp,fused:()=>uu,gather:()=>Li,gatherND:()=>m6,greater:()=>Re,greaterEqual:()=>ln,ifft:()=>tl,imag:()=>Xl,image:()=>Gs,inTopKAsync:()=>h6,irfft:()=>xm,isFinite:()=>Px,isInf:()=>Lx,isNaN:()=>Mx,leakyRelu:()=>Ql,less:()=>om,lessEqual:()=>Ln,linalg:()=>pv,linspace:()=>KE,localResponseNormalization:()=>zx,log:()=>Sr,log1p:()=>tu,logSigmoid:()=>Gx,logSoftmax:()=>sm,logSumExp:()=>im,logicalAnd:()=>Rr,logicalNot:()=>eu,logicalOr:()=>am,logicalXor:()=>Wx,losses:()=>hX,lowerBound:()=>jE,matMul:()=>Lt,max:()=>Ir,maxPool:()=>ru,maxPool3d:()=>Hx,maxPoolWithArgmax:()=>XE,maximum:()=>Sn,mean:()=>ve,meshgrid:()=>YE,min:()=>Ja,minimum:()=>Mi,mirrorPad:()=>qx,mod:()=>Kx,moments:()=>Zu,movingAverage:()=>s6,mul:()=>D,multiRNNCell:()=>ZE,multinomial:()=>JE,neg:()=>Ht,norm:()=>Qa,notEqual:()=>Bs,oneHot:()=>Ei,ones:()=>cr,onesLike:()=>yr,op:()=>T,outerProduct:()=>QE,pad:()=>cn,pad1d:()=>t_,pad2d:()=>e_,pad3d:()=>r_,pad4d:()=>n_,pool:()=>jx,pow:()=>an,prelu:()=>ou,print:()=>Jg,prod:()=>Xx,raggedGather:()=>o_,raggedRange:()=>s_,raggedTensorToTensor:()=>i_,rand:()=>a_,randomGamma:()=>v_,randomNormal:()=>tc,randomStandardNormal:()=>N_,randomUniform:()=>zi,range:()=>su,real:()=>Za,reciprocal:()=>ty,relu:()=>Fr,relu6:()=>lm,reshape:()=>R,reverse:()=>pr,reverse1d:()=>T_,reverse2d:()=>k_,reverse3d:()=>E_,reverse4d:()=>__,rfft:()=>lu,round:()=>um,rsqrt:()=>cm,scalar:()=>mt,scatterND:()=>a6,searchSorted:()=>mh,selu:()=>pm,separableConv2d:()=>mm,setdiff1dAsync:()=>A_,sigmoid:()=>Yr,sign:()=>ey,signal:()=>dX,sin:()=>fm,sinh:()=>dm,slice:()=>Rt,slice1d:()=>hm,slice2d:()=>dh,slice3d:()=>gm,slice4d:()=>ec,softmax:()=>iu,softplus:()=>zs,spaceToBatchND:()=>nu,sparse:()=>gX,sparseToDense:()=>c6,spectral:()=>fX,split:()=>mr,sqrt:()=>Se,square:()=>Mt,squaredDifference:()=>ym,squeeze:()=>Mn,stack:()=>nr,step:()=>bo,stridedSlice:()=>ry,string:()=>xX,sub:()=>ct,sum:()=>ft,tan:()=>ny,tanh:()=>$i,tensor:()=>ur,tensor1d:()=>Me,tensor2d:()=>Vs,tensor3d:()=>rx,tensor4d:()=>$_,tensor5d:()=>D_,tensor6d:()=>R_,tile:()=>Dr,topk:()=>oy,transpose:()=>Ot,truncatedNormal:()=>bm,unique:()=>sy,unsortedSegmentSum:()=>wm,unstack:()=>vr,upperBound:()=>F_,variable:()=>iy,where:()=>_e,whereAsync:()=>ly,zeros:()=>Ne,zerosLike:()=>It});var MD=(r,t,e,n=ae)=>{switch(r.op){case"BiasAdd":case"AddV2":case"Add":return[n.add(S("a",r,t,e),S("b",r,t,e))];case"AddN":return[n.addN(S("tensors",r,t,e))];case"FloorMod":case"Mod":return[n.mod(S("a",r,t,e),S("b",r,t,e))];case"Mul":return[n.mul(S("a",r,t,e),S("b",r,t,e))];case"RealDiv":case"Div":return[n.div(S("a",r,t,e),S("b",r,t,e))];case"DivNoNan":return[n.divNoNan(S("a",r,t,e),S("b",r,t,e))];case"FloorDiv":return[n.floorDiv(S("a",r,t,e),S("b",r,t,e))];case"Sub":return[n.sub(S("a",r,t,e),S("b",r,t,e))];case"Minimum":return[n.minimum(S("a",r,t,e),S("b",r,t,e))];case"Maximum":return[n.maximum(S("a",r,t,e),S("b",r,t,e))];case"Pow":return[n.pow(S("a",r,t,e),S("b",r,t,e))];case"SquaredDifference":return[n.squaredDifference(S("a",r,t,e),S("b",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var zD=(r,t,e,n=ae)=>{switch(r.op){case"Abs":case"ComplexAbs":return[n.abs(S("x",r,t,e))];case"Acos":return[n.acos(S("x",r,t,e))];case"Acosh":return[n.acosh(S("x",r,t,e))];case"Asin":return[n.asin(S("x",r,t,e))];case"Asinh":return[n.asinh(S("x",r,t,e))];case"Atan":return[n.atan(S("x",r,t,e))];case"Atan2":return[n.atan2(S("x",r,t,e),S("y",r,t,e))];case"Atanh":return[n.atanh(S("x",r,t,e))];case"Ceil":return[n.ceil(S("x",r,t,e))];case"Complex":return[n.complex(S("real",r,t,e),S("imag",r,t,e))];case"Cos":return[n.cos(S("x",r,t,e))];case"Cosh":return[n.cosh(S("x",r,t,e))];case"Elu":return[n.elu(S("x",r,t,e))];case"Erf":return[n.erf(S("x",r,t,e))];case"Exp":return[n.exp(S("x",r,t,e))];case"Expm1":return[n.expm1(S("x",r,t,e))];case"Floor":return[n.floor(S("x",r,t,e))];case"Log":return[n.log(S("x",r,t,e))];case"Log1p":return[n.log1p(S("x",r,t,e))];case"Imag":return[n.imag(S("x",r,t,e))];case"Neg":return[n.neg(S("x",r,t,e))];case"Reciprocal":return[n.reciprocal(S("x",r,t,e))];case"Real":return[n.real(S("x",r,t,e))];case"Relu":return[n.relu(S("x",r,t,e))];case"Round":return[n.round(S("x",r,t,e))];case"Selu":return[n.selu(S("x",r,t,e))];case"Sigmoid":return[n.sigmoid(S("x",r,t,e))];case"Sin":return[n.sin(S("x",r,t,e))];case"Sign":return[n.sign(S("x",r,t,e))];case"Sinh":return[n.sinh(S("x",r,t,e))];case"Softplus":return[n.softplus(S("x",r,t,e))];case"Sqrt":return[n.sqrt(S("x",r,t,e))];case"Square":return[n.square(S("x",r,t,e))];case"Tanh":return[n.tanh(S("x",r,t,e))];case"Tan":return[n.tan(S("x",r,t,e))];case"ClipByValue":return[n.clipByValue(S("x",r,t,e),S("clipValueMin",r,t,e),S("clipValueMax",r,t,e))];case"Relu6":return[n.relu6(S("x",r,t,e))];case"Rsqrt":return[n.rsqrt(br(r.inputNames[0],t,e))];case"Prod":return[n.prod(S("x",r,t,e),S("axes",r,t,e))];case"LeakyRelu":return[n.leakyRelu(S("x",r,t,e),S("alpha",r,t,e))];case"Prelu":return[n.prelu(S("x",r,t,e),S("alpha",r,t,e))];case"IsNan":return[n.isNaN(br(r.inputNames[0],t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function Vn(r,t,e=""){if(!(typeof r=="number"||typeof t=="number")){y.assert(r.length===t.length,()=>e+` Shapes ${r} and ${t} must match`);for(let n=0;n<r.length;n++){let o=r[n],s=t[n];y.assert(o<0||s<0||o===s,()=>e+` Shapes ${r} and ${t} must match`)}}}function BD(r){return!(typeof r=="number"||r.some(t=>t<0))}function Kf(r,t,e){let n=Mb(r,e),o=!BD(n);if(o&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${n}`);if(o&&t.forEach(s=>{n=Mb(s.shape,n)}),!BD(n))throw new Error(`Non-fully-defined elementShape: ${n}`);return n}function Mb(r,t){if(typeof r=="number")return t;if(typeof t=="number")return r;if(r.length!==t.length)throw new Error(`Incompatible ranks during merge: ${r} vs. ${t}`);let e=[];for(let n=0;n<r.length;++n){let o=r[n],s=t[n];if(o>=0&&s>=0&&o!==s)throw new Error(`Incompatible shape during merge: ${r} vs. ${t}`);e[n]=o>=0?o:s}return e}var zb=class{constructor(t,e,n,o,s,i,a){this.name=t,this.dtype=e,this.maxSize=n,this.elementShape=o,this.identicalElementShapes=s,this.dynamicSize=i,this.clearAfterRead=a,this.tensors=[],this.closed_=!1,this.idTensor=mt(0),De(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(t){this.tensors.forEach(e=>{(t==null||!t.has(e.tensor.id))&&e.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(t<0||t>=this.size())throw new Error(`Tried to read from index ${t}, but array size is: ${this.size()}`);let e=this.tensors[t];if(e.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${t} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(e.cleared=!0),e.read=!0,e.tensor}readMany(t){return t.map(e=>this.read(e))}write(t,e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(t<0||!this.dynamicSize&&t>=this.maxSize)throw new Error(`Tried to write to index ${t}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[t]||{};if(e.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t},
-          because the value dtype is ${e.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=e.shape),Vn(this.elementShape,e.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${t}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${t}, because it has already been written.`);n.tensor=e,De(e),n.written=!0,this.tensors[t]=n}writeMany(t,e){if(t.length!==e.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${t.length} is not the same as tensors size: ${e.length}.`);t.forEach((n,o)=>this.write(n,e[o]))}gather(t,e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${e}`);if(t)t=t.slice(0,this.size());else{t=[];for(let o=0;o<this.size();o++)t.push(o)}if(t.length===0)return ur([],[0].concat(this.elementShape));let n=this.readMany(t);return Vn(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),nr(n,0)}concat(t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${t}`);if(this.size()===0)return ur([],[0].concat(this.elementShape));let e=[];for(let o=0;o<this.size();o++)e.push(o);let n=this.readMany(e);return Vn(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),ne(n,0)}scatter(t,e){if(e.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${e.dtype}`);if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let n=Math.max(...t);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(t,vr(e,0))}split(t,e){if(e.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${e.dtype}`);let n=0,o=t.map(u=>(n+=u,n));if(n!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
+}) : x)(function(x) {
+  if (typeof require !== "undefined")
+    return require.apply(this, arguments);
+  throw new Error('Dynamic require of "' + x + '" is not supported');
+});
+var __commonJS = (cb, mod4) => function __require2() {
+  return mod4 || (0, cb[__getOwnPropNames(cb)[0]])((mod4 = { exports: {} }).exports, mod4), mod4.exports;
+};
+var __export = (target, all5) => {
+  for (var name in all5)
+    __defProp(target, name, { get: all5[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod4, isNodeMode, target) => (target = mod4 != null ? __create(__getProtoOf(mod4)) : {}, __copyProps(
+  isNodeMode || !mod4 || !mod4.__esModule ? __defProp(target, "default", { value: mod4, enumerable: true }) : target,
+  mod4
+));
+
+// node_modules/long/src/long.js
+var require_long = __commonJS({
+  "node_modules/long/src/long.js"(exports, module) {
+    module.exports = Long2;
+    var wasm = null;
+    try {
+      wasm = new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([
+        0,
+        97,
+        115,
+        109,
+        1,
+        0,
+        0,
+        0,
+        1,
+        13,
+        2,
+        96,
+        0,
+        1,
+        127,
+        96,
+        4,
+        127,
+        127,
+        127,
+        127,
+        1,
+        127,
+        3,
+        7,
+        6,
+        0,
+        1,
+        1,
+        1,
+        1,
+        1,
+        6,
+        6,
+        1,
+        127,
+        1,
+        65,
+        0,
+        11,
+        7,
+        50,
+        6,
+        3,
+        109,
+        117,
+        108,
+        0,
+        1,
+        5,
+        100,
+        105,
+        118,
+        95,
+        115,
+        0,
+        2,
+        5,
+        100,
+        105,
+        118,
+        95,
+        117,
+        0,
+        3,
+        5,
+        114,
+        101,
+        109,
+        95,
+        115,
+        0,
+        4,
+        5,
+        114,
+        101,
+        109,
+        95,
+        117,
+        0,
+        5,
+        8,
+        103,
+        101,
+        116,
+        95,
+        104,
+        105,
+        103,
+        104,
+        0,
+        0,
+        10,
+        191,
+        1,
+        6,
+        4,
+        0,
+        35,
+        0,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        126,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        127,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        128,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        129,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11,
+        36,
+        1,
+        1,
+        126,
+        32,
+        0,
+        173,
+        32,
+        1,
+        173,
+        66,
+        32,
+        134,
+        132,
+        32,
+        2,
+        173,
+        32,
+        3,
+        173,
+        66,
+        32,
+        134,
+        132,
+        130,
+        34,
+        4,
+        66,
+        32,
+        135,
+        167,
+        36,
+        0,
+        32,
+        4,
+        167,
+        11
+      ])), {}).exports;
+    } catch (e) {
+    }
+    function Long2(low, high, unsigned) {
+      this.low = low | 0;
+      this.high = high | 0;
+      this.unsigned = !!unsigned;
+    }
+    Long2.prototype.__isLong__;
+    Object.defineProperty(Long2.prototype, "__isLong__", { value: true });
+    function isLong(obj) {
+      return (obj && obj["__isLong__"]) === true;
+    }
+    Long2.isLong = isLong;
+    var INT_CACHE = {};
+    var UINT_CACHE = {};
+    function fromInt(value, unsigned) {
+      var obj, cachedObj, cache;
+      if (unsigned) {
+        value >>>= 0;
+        if (cache = 0 <= value && value < 256) {
+          cachedObj = UINT_CACHE[value];
+          if (cachedObj)
+            return cachedObj;
+        }
+        obj = fromBits(value, (value | 0) < 0 ? -1 : 0, true);
+        if (cache)
+          UINT_CACHE[value] = obj;
+        return obj;
+      } else {
+        value |= 0;
+        if (cache = -128 <= value && value < 128) {
+          cachedObj = INT_CACHE[value];
+          if (cachedObj)
+            return cachedObj;
+        }
+        obj = fromBits(value, value < 0 ? -1 : 0, false);
+        if (cache)
+          INT_CACHE[value] = obj;
+        return obj;
+      }
+    }
+    Long2.fromInt = fromInt;
+    function fromNumber(value, unsigned) {
+      if (isNaN(value))
+        return unsigned ? UZERO : ZERO;
+      if (unsigned) {
+        if (value < 0)
+          return UZERO;
+        if (value >= TWO_PWR_64_DBL)
+          return MAX_UNSIGNED_VALUE;
+      } else {
+        if (value <= -TWO_PWR_63_DBL)
+          return MIN_VALUE;
+        if (value + 1 >= TWO_PWR_63_DBL)
+          return MAX_VALUE;
+      }
+      if (value < 0)
+        return fromNumber(-value, unsigned).neg();
+      return fromBits(value % TWO_PWR_32_DBL | 0, value / TWO_PWR_32_DBL | 0, unsigned);
+    }
+    Long2.fromNumber = fromNumber;
+    function fromBits(lowBits, highBits, unsigned) {
+      return new Long2(lowBits, highBits, unsigned);
+    }
+    Long2.fromBits = fromBits;
+    var pow_dbl = Math.pow;
+    function fromString(str, unsigned, radix) {
+      if (str.length === 0)
+        throw Error("empty string");
+      if (str === "NaN" || str === "Infinity" || str === "+Infinity" || str === "-Infinity")
+        return ZERO;
+      if (typeof unsigned === "number") {
+        radix = unsigned, unsigned = false;
+      } else {
+        unsigned = !!unsigned;
+      }
+      radix = radix || 10;
+      if (radix < 2 || 36 < radix)
+        throw RangeError("radix");
+      var p2;
+      if ((p2 = str.indexOf("-")) > 0)
+        throw Error("interior hyphen");
+      else if (p2 === 0) {
+        return fromString(str.substring(1), unsigned, radix).neg();
+      }
+      var radixToPower = fromNumber(pow_dbl(radix, 8));
+      var result = ZERO;
+      for (var i = 0; i < str.length; i += 8) {
+        var size = Math.min(8, str.length - i), value = parseInt(str.substring(i, i + size), radix);
+        if (size < 8) {
+          var power = fromNumber(pow_dbl(radix, size));
+          result = result.mul(power).add(fromNumber(value));
+        } else {
+          result = result.mul(radixToPower);
+          result = result.add(fromNumber(value));
+        }
+      }
+      result.unsigned = unsigned;
+      return result;
+    }
+    Long2.fromString = fromString;
+    function fromValue(val, unsigned) {
+      if (typeof val === "number")
+        return fromNumber(val, unsigned);
+      if (typeof val === "string")
+        return fromString(val, unsigned);
+      return fromBits(val.low, val.high, typeof unsigned === "boolean" ? unsigned : val.unsigned);
+    }
+    Long2.fromValue = fromValue;
+    var TWO_PWR_16_DBL = 1 << 16;
+    var TWO_PWR_24_DBL = 1 << 24;
+    var TWO_PWR_32_DBL = TWO_PWR_16_DBL * TWO_PWR_16_DBL;
+    var TWO_PWR_64_DBL = TWO_PWR_32_DBL * TWO_PWR_32_DBL;
+    var TWO_PWR_63_DBL = TWO_PWR_64_DBL / 2;
+    var TWO_PWR_24 = fromInt(TWO_PWR_24_DBL);
+    var ZERO = fromInt(0);
+    Long2.ZERO = ZERO;
+    var UZERO = fromInt(0, true);
+    Long2.UZERO = UZERO;
+    var ONE = fromInt(1);
+    Long2.ONE = ONE;
+    var UONE = fromInt(1, true);
+    Long2.UONE = UONE;
+    var NEG_ONE = fromInt(-1);
+    Long2.NEG_ONE = NEG_ONE;
+    var MAX_VALUE = fromBits(4294967295 | 0, 2147483647 | 0, false);
+    Long2.MAX_VALUE = MAX_VALUE;
+    var MAX_UNSIGNED_VALUE = fromBits(4294967295 | 0, 4294967295 | 0, true);
+    Long2.MAX_UNSIGNED_VALUE = MAX_UNSIGNED_VALUE;
+    var MIN_VALUE = fromBits(0, 2147483648 | 0, false);
+    Long2.MIN_VALUE = MIN_VALUE;
+    var LongPrototype = Long2.prototype;
+    LongPrototype.toInt = function toInt() {
+      return this.unsigned ? this.low >>> 0 : this.low;
+    };
+    LongPrototype.toNumber = function toNumber() {
+      if (this.unsigned)
+        return (this.high >>> 0) * TWO_PWR_32_DBL + (this.low >>> 0);
+      return this.high * TWO_PWR_32_DBL + (this.low >>> 0);
+    };
+    LongPrototype.toString = function toString(radix) {
+      radix = radix || 10;
+      if (radix < 2 || 36 < radix)
+        throw RangeError("radix");
+      if (this.isZero())
+        return "0";
+      if (this.isNegative()) {
+        if (this.eq(MIN_VALUE)) {
+          var radixLong = fromNumber(radix), div3 = this.div(radixLong), rem1 = div3.mul(radixLong).sub(this);
+          return div3.toString(radix) + rem1.toInt().toString(radix);
+        } else
+          return "-" + this.neg().toString(radix);
+      }
+      var radixToPower = fromNumber(pow_dbl(radix, 6), this.unsigned), rem = this;
+      var result = "";
+      while (true) {
+        var remDiv = rem.div(radixToPower), intval = rem.sub(remDiv.mul(radixToPower)).toInt() >>> 0, digits = intval.toString(radix);
+        rem = remDiv;
+        if (rem.isZero())
+          return digits + result;
+        else {
+          while (digits.length < 6)
+            digits = "0" + digits;
+          result = "" + digits + result;
+        }
+      }
+    };
+    LongPrototype.getHighBits = function getHighBits() {
+      return this.high;
+    };
+    LongPrototype.getHighBitsUnsigned = function getHighBitsUnsigned() {
+      return this.high >>> 0;
+    };
+    LongPrototype.getLowBits = function getLowBits() {
+      return this.low;
+    };
+    LongPrototype.getLowBitsUnsigned = function getLowBitsUnsigned() {
+      return this.low >>> 0;
+    };
+    LongPrototype.getNumBitsAbs = function getNumBitsAbs() {
+      if (this.isNegative())
+        return this.eq(MIN_VALUE) ? 64 : this.neg().getNumBitsAbs();
+      var val = this.high != 0 ? this.high : this.low;
+      for (var bit = 31; bit > 0; bit--)
+        if ((val & 1 << bit) != 0)
+          break;
+      return this.high != 0 ? bit + 33 : bit + 1;
+    };
+    LongPrototype.isZero = function isZero() {
+      return this.high === 0 && this.low === 0;
+    };
+    LongPrototype.eqz = LongPrototype.isZero;
+    LongPrototype.isNegative = function isNegative() {
+      return !this.unsigned && this.high < 0;
+    };
+    LongPrototype.isPositive = function isPositive() {
+      return this.unsigned || this.high >= 0;
+    };
+    LongPrototype.isOdd = function isOdd() {
+      return (this.low & 1) === 1;
+    };
+    LongPrototype.isEven = function isEven2() {
+      return (this.low & 1) === 0;
+    };
+    LongPrototype.equals = function equals(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      if (this.unsigned !== other.unsigned && this.high >>> 31 === 1 && other.high >>> 31 === 1)
+        return false;
+      return this.high === other.high && this.low === other.low;
+    };
+    LongPrototype.eq = LongPrototype.equals;
+    LongPrototype.notEquals = function notEquals(other) {
+      return !this.eq(other);
+    };
+    LongPrototype.neq = LongPrototype.notEquals;
+    LongPrototype.ne = LongPrototype.notEquals;
+    LongPrototype.lessThan = function lessThan(other) {
+      return this.comp(other) < 0;
+    };
+    LongPrototype.lt = LongPrototype.lessThan;
+    LongPrototype.lessThanOrEqual = function lessThanOrEqual(other) {
+      return this.comp(other) <= 0;
+    };
+    LongPrototype.lte = LongPrototype.lessThanOrEqual;
+    LongPrototype.le = LongPrototype.lessThanOrEqual;
+    LongPrototype.greaterThan = function greaterThan(other) {
+      return this.comp(other) > 0;
+    };
+    LongPrototype.gt = LongPrototype.greaterThan;
+    LongPrototype.greaterThanOrEqual = function greaterThanOrEqual(other) {
+      return this.comp(other) >= 0;
+    };
+    LongPrototype.gte = LongPrototype.greaterThanOrEqual;
+    LongPrototype.ge = LongPrototype.greaterThanOrEqual;
+    LongPrototype.compare = function compare(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      if (this.eq(other))
+        return 0;
+      var thisNeg = this.isNegative(), otherNeg = other.isNegative();
+      if (thisNeg && !otherNeg)
+        return -1;
+      if (!thisNeg && otherNeg)
+        return 1;
+      if (!this.unsigned)
+        return this.sub(other).isNegative() ? -1 : 1;
+      return other.high >>> 0 > this.high >>> 0 || other.high === this.high && other.low >>> 0 > this.low >>> 0 ? -1 : 1;
+    };
+    LongPrototype.comp = LongPrototype.compare;
+    LongPrototype.negate = function negate() {
+      if (!this.unsigned && this.eq(MIN_VALUE))
+        return MIN_VALUE;
+      return this.not().add(ONE);
+    };
+    LongPrototype.neg = LongPrototype.negate;
+    LongPrototype.add = function add5(addend) {
+      if (!isLong(addend))
+        addend = fromValue(addend);
+      var a48 = this.high >>> 16;
+      var a32 = this.high & 65535;
+      var a16 = this.low >>> 16;
+      var a00 = this.low & 65535;
+      var b48 = addend.high >>> 16;
+      var b32 = addend.high & 65535;
+      var b16 = addend.low >>> 16;
+      var b00 = addend.low & 65535;
+      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;
+      c00 += a00 + b00;
+      c16 += c00 >>> 16;
+      c00 &= 65535;
+      c16 += a16 + b16;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c32 += a32 + b32;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c48 += a48 + b48;
+      c48 &= 65535;
+      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);
+    };
+    LongPrototype.subtract = function subtract(subtrahend) {
+      if (!isLong(subtrahend))
+        subtrahend = fromValue(subtrahend);
+      return this.add(subtrahend.neg());
+    };
+    LongPrototype.sub = LongPrototype.subtract;
+    LongPrototype.multiply = function multiply4(multiplier) {
+      if (this.isZero())
+        return ZERO;
+      if (!isLong(multiplier))
+        multiplier = fromValue(multiplier);
+      if (wasm) {
+        var low = wasm.mul(
+          this.low,
+          this.high,
+          multiplier.low,
+          multiplier.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      if (multiplier.isZero())
+        return ZERO;
+      if (this.eq(MIN_VALUE))
+        return multiplier.isOdd() ? MIN_VALUE : ZERO;
+      if (multiplier.eq(MIN_VALUE))
+        return this.isOdd() ? MIN_VALUE : ZERO;
+      if (this.isNegative()) {
+        if (multiplier.isNegative())
+          return this.neg().mul(multiplier.neg());
+        else
+          return this.neg().mul(multiplier).neg();
+      } else if (multiplier.isNegative())
+        return this.mul(multiplier.neg()).neg();
+      if (this.lt(TWO_PWR_24) && multiplier.lt(TWO_PWR_24))
+        return fromNumber(this.toNumber() * multiplier.toNumber(), this.unsigned);
+      var a48 = this.high >>> 16;
+      var a32 = this.high & 65535;
+      var a16 = this.low >>> 16;
+      var a00 = this.low & 65535;
+      var b48 = multiplier.high >>> 16;
+      var b32 = multiplier.high & 65535;
+      var b16 = multiplier.low >>> 16;
+      var b00 = multiplier.low & 65535;
+      var c48 = 0, c32 = 0, c16 = 0, c00 = 0;
+      c00 += a00 * b00;
+      c16 += c00 >>> 16;
+      c00 &= 65535;
+      c16 += a16 * b00;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c16 += a00 * b16;
+      c32 += c16 >>> 16;
+      c16 &= 65535;
+      c32 += a32 * b00;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c32 += a16 * b16;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c32 += a00 * b32;
+      c48 += c32 >>> 16;
+      c32 &= 65535;
+      c48 += a48 * b00 + a32 * b16 + a16 * b32 + a00 * b48;
+      c48 &= 65535;
+      return fromBits(c16 << 16 | c00, c48 << 16 | c32, this.unsigned);
+    };
+    LongPrototype.mul = LongPrototype.multiply;
+    LongPrototype.divide = function divide(divisor) {
+      if (!isLong(divisor))
+        divisor = fromValue(divisor);
+      if (divisor.isZero())
+        throw Error("division by zero");
+      if (wasm) {
+        if (!this.unsigned && this.high === -2147483648 && divisor.low === -1 && divisor.high === -1) {
+          return this;
+        }
+        var low = (this.unsigned ? wasm.div_u : wasm.div_s)(
+          this.low,
+          this.high,
+          divisor.low,
+          divisor.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      if (this.isZero())
+        return this.unsigned ? UZERO : ZERO;
+      var approx, rem, res;
+      if (!this.unsigned) {
+        if (this.eq(MIN_VALUE)) {
+          if (divisor.eq(ONE) || divisor.eq(NEG_ONE))
+            return MIN_VALUE;
+          else if (divisor.eq(MIN_VALUE))
+            return ONE;
+          else {
+            var halfThis = this.shr(1);
+            approx = halfThis.div(divisor).shl(1);
+            if (approx.eq(ZERO)) {
+              return divisor.isNegative() ? ONE : NEG_ONE;
+            } else {
+              rem = this.sub(divisor.mul(approx));
+              res = approx.add(rem.div(divisor));
+              return res;
+            }
+          }
+        } else if (divisor.eq(MIN_VALUE))
+          return this.unsigned ? UZERO : ZERO;
+        if (this.isNegative()) {
+          if (divisor.isNegative())
+            return this.neg().div(divisor.neg());
+          return this.neg().div(divisor).neg();
+        } else if (divisor.isNegative())
+          return this.div(divisor.neg()).neg();
+        res = ZERO;
+      } else {
+        if (!divisor.unsigned)
+          divisor = divisor.toUnsigned();
+        if (divisor.gt(this))
+          return UZERO;
+        if (divisor.gt(this.shru(1)))
+          return UONE;
+        res = UZERO;
+      }
+      rem = this;
+      while (rem.gte(divisor)) {
+        approx = Math.max(1, Math.floor(rem.toNumber() / divisor.toNumber()));
+        var log22 = Math.ceil(Math.log(approx) / Math.LN2), delta = log22 <= 48 ? 1 : pow_dbl(2, log22 - 48), approxRes = fromNumber(approx), approxRem = approxRes.mul(divisor);
+        while (approxRem.isNegative() || approxRem.gt(rem)) {
+          approx -= delta;
+          approxRes = fromNumber(approx, this.unsigned);
+          approxRem = approxRes.mul(divisor);
+        }
+        if (approxRes.isZero())
+          approxRes = ONE;
+        res = res.add(approxRes);
+        rem = rem.sub(approxRem);
+      }
+      return res;
+    };
+    LongPrototype.div = LongPrototype.divide;
+    LongPrototype.modulo = function modulo(divisor) {
+      if (!isLong(divisor))
+        divisor = fromValue(divisor);
+      if (wasm) {
+        var low = (this.unsigned ? wasm.rem_u : wasm.rem_s)(
+          this.low,
+          this.high,
+          divisor.low,
+          divisor.high
+        );
+        return fromBits(low, wasm.get_high(), this.unsigned);
+      }
+      return this.sub(this.div(divisor).mul(divisor));
+    };
+    LongPrototype.mod = LongPrototype.modulo;
+    LongPrototype.rem = LongPrototype.modulo;
+    LongPrototype.not = function not() {
+      return fromBits(~this.low, ~this.high, this.unsigned);
+    };
+    LongPrototype.and = function and(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low & other.low, this.high & other.high, this.unsigned);
+    };
+    LongPrototype.or = function or(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low | other.low, this.high | other.high, this.unsigned);
+    };
+    LongPrototype.xor = function xor(other) {
+      if (!isLong(other))
+        other = fromValue(other);
+      return fromBits(this.low ^ other.low, this.high ^ other.high, this.unsigned);
+    };
+    LongPrototype.shiftLeft = function shiftLeft(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      if ((numBits &= 63) === 0)
+        return this;
+      else if (numBits < 32)
+        return fromBits(this.low << numBits, this.high << numBits | this.low >>> 32 - numBits, this.unsigned);
+      else
+        return fromBits(0, this.low << numBits - 32, this.unsigned);
+    };
+    LongPrototype.shl = LongPrototype.shiftLeft;
+    LongPrototype.shiftRight = function shiftRight(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      if ((numBits &= 63) === 0)
+        return this;
+      else if (numBits < 32)
+        return fromBits(this.low >>> numBits | this.high << 32 - numBits, this.high >> numBits, this.unsigned);
+      else
+        return fromBits(this.high >> numBits - 32, this.high >= 0 ? 0 : -1, this.unsigned);
+    };
+    LongPrototype.shr = LongPrototype.shiftRight;
+    LongPrototype.shiftRightUnsigned = function shiftRightUnsigned(numBits) {
+      if (isLong(numBits))
+        numBits = numBits.toInt();
+      numBits &= 63;
+      if (numBits === 0)
+        return this;
+      else {
+        var high = this.high;
+        if (numBits < 32) {
+          var low = this.low;
+          return fromBits(low >>> numBits | high << 32 - numBits, high >>> numBits, this.unsigned);
+        } else if (numBits === 32)
+          return fromBits(high, 0, this.unsigned);
+        else
+          return fromBits(high >>> numBits - 32, 0, this.unsigned);
+      }
+    };
+    LongPrototype.shru = LongPrototype.shiftRightUnsigned;
+    LongPrototype.shr_u = LongPrototype.shiftRightUnsigned;
+    LongPrototype.toSigned = function toSigned() {
+      if (!this.unsigned)
+        return this;
+      return fromBits(this.low, this.high, false);
+    };
+    LongPrototype.toUnsigned = function toUnsigned() {
+      if (this.unsigned)
+        return this;
+      return fromBits(this.low, this.high, true);
+    };
+    LongPrototype.toBytes = function toBytes(le) {
+      return le ? this.toBytesLE() : this.toBytesBE();
+    };
+    LongPrototype.toBytesLE = function toBytesLE() {
+      var hi = this.high, lo = this.low;
+      return [
+        lo & 255,
+        lo >>> 8 & 255,
+        lo >>> 16 & 255,
+        lo >>> 24,
+        hi & 255,
+        hi >>> 8 & 255,
+        hi >>> 16 & 255,
+        hi >>> 24
+      ];
+    };
+    LongPrototype.toBytesBE = function toBytesBE() {
+      var hi = this.high, lo = this.low;
+      return [
+        hi >>> 24,
+        hi >>> 16 & 255,
+        hi >>> 8 & 255,
+        hi & 255,
+        lo >>> 24,
+        lo >>> 16 & 255,
+        lo >>> 8 & 255,
+        lo & 255
+      ];
+    };
+    Long2.fromBytes = function fromBytes(bytes, unsigned, le) {
+      return le ? Long2.fromBytesLE(bytes, unsigned) : Long2.fromBytesBE(bytes, unsigned);
+    };
+    Long2.fromBytesLE = function fromBytesLE(bytes, unsigned) {
+      return new Long2(
+        bytes[0] | bytes[1] << 8 | bytes[2] << 16 | bytes[3] << 24,
+        bytes[4] | bytes[5] << 8 | bytes[6] << 16 | bytes[7] << 24,
+        unsigned
+      );
+    };
+    Long2.fromBytesBE = function fromBytesBE(bytes, unsigned) {
+      return new Long2(
+        bytes[4] << 24 | bytes[5] << 16 | bytes[6] << 8 | bytes[7],
+        bytes[0] << 24 | bytes[1] << 16 | bytes[2] << 8 | bytes[3],
+        unsigned
+      );
+    };
+  }
+});
+
+// (disabled):node_modules/node-fetch/browser.js
+var require_browser = __commonJS({
+  "(disabled):node_modules/node-fetch/browser.js"() {
+  }
+});
+
+// (disabled):util
+var require_util = __commonJS({
+  "(disabled):util"() {
+  }
+});
+
+// node_modules/seedrandom/lib/alea.js
+var require_alea = __commonJS({
+  "node_modules/seedrandom/lib/alea.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function Alea(seed) {
+        var me = this, mash = Mash();
+        me.next = function() {
+          var t = 2091639 * me.s0 + me.c * 23283064365386963e-26;
+          me.s0 = me.s1;
+          me.s1 = me.s2;
+          return me.s2 = t - (me.c = t | 0);
+        };
+        me.c = 1;
+        me.s0 = mash(" ");
+        me.s1 = mash(" ");
+        me.s2 = mash(" ");
+        me.s0 -= mash(seed);
+        if (me.s0 < 0) {
+          me.s0 += 1;
+        }
+        me.s1 -= mash(seed);
+        if (me.s1 < 0) {
+          me.s1 += 1;
+        }
+        me.s2 -= mash(seed);
+        if (me.s2 < 0) {
+          me.s2 += 1;
+        }
+        mash = null;
+      }
+      function copy(f, t) {
+        t.c = f.c;
+        t.s0 = f.s0;
+        t.s1 = f.s1;
+        t.s2 = f.s2;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new Alea(seed), state = opts && opts.state, prng = xg.next;
+        prng.int32 = function() {
+          return xg.next() * 4294967296 | 0;
+        };
+        prng.double = function() {
+          return prng() + (prng() * 2097152 | 0) * 11102230246251565e-32;
+        };
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      function Mash() {
+        var n = 4022871197;
+        var mash = function(data) {
+          data = String(data);
+          for (var i = 0; i < data.length; i++) {
+            n += data.charCodeAt(i);
+            var h = 0.02519603282416938 * n;
+            n = h >>> 0;
+            h -= n;
+            h *= n;
+            n = h >>> 0;
+            h -= n;
+            n += h * 4294967296;
+          }
+          return (n >>> 0) * 23283064365386963e-26;
+        };
+        return mash;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.alea = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// node_modules/seedrandom/lib/xor128.js
+var require_xor128 = __commonJS({
+  "node_modules/seedrandom/lib/xor128.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.x = 0;
+        me.y = 0;
+        me.z = 0;
+        me.w = 0;
+        me.next = function() {
+          var t = me.x ^ me.x << 11;
+          me.x = me.y;
+          me.y = me.z;
+          me.z = me.w;
+          return me.w ^= me.w >>> 19 ^ t ^ t >>> 8;
+        };
+        if (seed === (seed | 0)) {
+          me.x = seed;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 64; k++) {
+          me.x ^= strseed.charCodeAt(k) | 0;
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.x = f.x;
+        t.y = f.y;
+        t.z = f.z;
+        t.w = f.w;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xor128 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// node_modules/seedrandom/lib/xorwow.js
+var require_xorwow = __commonJS({
+  "node_modules/seedrandom/lib/xorwow.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.next = function() {
+          var t = me.x ^ me.x >>> 2;
+          me.x = me.y;
+          me.y = me.z;
+          me.z = me.w;
+          me.w = me.v;
+          return (me.d = me.d + 362437 | 0) + (me.v = me.v ^ me.v << 4 ^ (t ^ t << 1)) | 0;
+        };
+        me.x = 0;
+        me.y = 0;
+        me.z = 0;
+        me.w = 0;
+        me.v = 0;
+        if (seed === (seed | 0)) {
+          me.x = seed;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 64; k++) {
+          me.x ^= strseed.charCodeAt(k) | 0;
+          if (k == strseed.length) {
+            me.d = me.x << 10 ^ me.x >>> 4;
+          }
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.x = f.x;
+        t.y = f.y;
+        t.z = f.z;
+        t.w = f.w;
+        t.v = f.v;
+        t.d = f.d;
+        return t;
+      }
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xorwow = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// node_modules/seedrandom/lib/xorshift7.js
+var require_xorshift7 = __commonJS({
+  "node_modules/seedrandom/lib/xorshift7.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this;
+        me.next = function() {
+          var X = me.x, i = me.i, t, v, w;
+          t = X[i];
+          t ^= t >>> 7;
+          v = t ^ t << 24;
+          t = X[i + 1 & 7];
+          v ^= t ^ t >>> 10;
+          t = X[i + 3 & 7];
+          v ^= t ^ t >>> 3;
+          t = X[i + 4 & 7];
+          v ^= t ^ t << 7;
+          t = X[i + 7 & 7];
+          t = t ^ t << 13;
+          v ^= t ^ t << 9;
+          X[i] = v;
+          me.i = i + 1 & 7;
+          return v;
+        };
+        function init2(me2, seed2) {
+          var j, w, X = [];
+          if (seed2 === (seed2 | 0)) {
+            w = X[0] = seed2;
+          } else {
+            seed2 = "" + seed2;
+            for (j = 0; j < seed2.length; ++j) {
+              X[j & 7] = X[j & 7] << 15 ^ seed2.charCodeAt(j) + X[j + 1 & 7] << 13;
+            }
+          }
+          while (X.length < 8)
+            X.push(0);
+          for (j = 0; j < 8 && X[j] === 0; ++j)
+            ;
+          if (j == 8)
+            w = X[7] = -1;
+          else
+            w = X[j];
+          me2.x = X;
+          me2.i = 0;
+          for (j = 256; j > 0; --j) {
+            me2.next();
+          }
+        }
+        init2(me, seed);
+      }
+      function copy(f, t) {
+        t.x = f.x.slice();
+        t.i = f.i;
+        return t;
+      }
+      function impl(seed, opts) {
+        if (seed == null)
+          seed = +new Date();
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (state.x)
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xorshift7 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// node_modules/seedrandom/lib/xor4096.js
+var require_xor4096 = __commonJS({
+  "node_modules/seedrandom/lib/xor4096.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this;
+        me.next = function() {
+          var w = me.w, X = me.X, i = me.i, t, v;
+          me.w = w = w + 1640531527 | 0;
+          v = X[i + 34 & 127];
+          t = X[i = i + 1 & 127];
+          v ^= v << 13;
+          t ^= t << 17;
+          v ^= v >>> 15;
+          t ^= t >>> 12;
+          v = X[i] = v ^ t;
+          me.i = i;
+          return v + (w ^ w >>> 16) | 0;
+        };
+        function init2(me2, seed2) {
+          var t, v, i, j, w, X = [], limit = 128;
+          if (seed2 === (seed2 | 0)) {
+            v = seed2;
+            seed2 = null;
+          } else {
+            seed2 = seed2 + "\0";
+            v = 0;
+            limit = Math.max(limit, seed2.length);
+          }
+          for (i = 0, j = -32; j < limit; ++j) {
+            if (seed2)
+              v ^= seed2.charCodeAt((j + 32) % seed2.length);
+            if (j === 0)
+              w = v;
+            v ^= v << 10;
+            v ^= v >>> 15;
+            v ^= v << 4;
+            v ^= v >>> 13;
+            if (j >= 0) {
+              w = w + 1640531527 | 0;
+              t = X[j & 127] ^= v + w;
+              i = 0 == t ? i + 1 : 0;
+            }
+          }
+          if (i >= 128) {
+            X[(seed2 && seed2.length || 0) & 127] = -1;
+          }
+          i = 127;
+          for (j = 4 * 128; j > 0; --j) {
+            v = X[i + 34 & 127];
+            t = X[i = i + 1 & 127];
+            v ^= v << 13;
+            t ^= t << 17;
+            v ^= v >>> 15;
+            t ^= t >>> 12;
+            X[i] = v ^ t;
+          }
+          me2.w = w;
+          me2.X = X;
+          me2.i = i;
+        }
+        init2(me, seed);
+      }
+      function copy(f, t) {
+        t.i = f.i;
+        t.w = f.w;
+        t.X = f.X.slice();
+        return t;
+      }
+      ;
+      function impl(seed, opts) {
+        if (seed == null)
+          seed = +new Date();
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (state.X)
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.xor4096 = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// node_modules/seedrandom/lib/tychei.js
+var require_tychei = __commonJS({
+  "node_modules/seedrandom/lib/tychei.js"(exports, module) {
+    (function(global2, module2, define2) {
+      function XorGen(seed) {
+        var me = this, strseed = "";
+        me.next = function() {
+          var b = me.b, c = me.c, d = me.d, a = me.a;
+          b = b << 25 ^ b >>> 7 ^ c;
+          c = c - d | 0;
+          d = d << 24 ^ d >>> 8 ^ a;
+          a = a - b | 0;
+          me.b = b = b << 20 ^ b >>> 12 ^ c;
+          me.c = c = c - d | 0;
+          me.d = d << 16 ^ c >>> 16 ^ a;
+          return me.a = a - b | 0;
+        };
+        me.a = 0;
+        me.b = 0;
+        me.c = 2654435769 | 0;
+        me.d = 1367130551;
+        if (seed === Math.floor(seed)) {
+          me.a = seed / 4294967296 | 0;
+          me.b = seed | 0;
+        } else {
+          strseed += seed;
+        }
+        for (var k = 0; k < strseed.length + 20; k++) {
+          me.b ^= strseed.charCodeAt(k) | 0;
+          me.next();
+        }
+      }
+      function copy(f, t) {
+        t.a = f.a;
+        t.b = f.b;
+        t.c = f.c;
+        t.d = f.d;
+        return t;
+      }
+      ;
+      function impl(seed, opts) {
+        var xg = new XorGen(seed), state = opts && opts.state, prng = function() {
+          return (xg.next() >>> 0) / 4294967296;
+        };
+        prng.double = function() {
+          do {
+            var top = xg.next() >>> 11, bot = (xg.next() >>> 0) / 4294967296, result = (top + bot) / (1 << 21);
+          } while (result === 0);
+          return result;
+        };
+        prng.int32 = xg.next;
+        prng.quick = prng;
+        if (state) {
+          if (typeof state == "object")
+            copy(state, xg);
+          prng.state = function() {
+            return copy(xg, {});
+          };
+        }
+        return prng;
+      }
+      if (module2 && module2.exports) {
+        module2.exports = impl;
+      } else if (define2 && define2.amd) {
+        define2(function() {
+          return impl;
+        });
+      } else {
+        this.tychei = impl;
+      }
+    })(
+      exports,
+      typeof module == "object" && module,
+      typeof define == "function" && define
+    );
+  }
+});
+
+// (disabled):crypto
+var require_crypto = __commonJS({
+  "(disabled):crypto"() {
+  }
+});
+
+// node_modules/seedrandom/seedrandom.js
+var require_seedrandom = __commonJS({
+  "node_modules/seedrandom/seedrandom.js"(exports, module) {
+    (function(global2, pool3, math) {
+      var width = 256, chunks = 6, digits = 52, rngname = "random", startdenom = math.pow(width, chunks), significance = math.pow(2, digits), overflow = significance * 2, mask = width - 1, nodecrypto;
+      function seedrandom5(seed, options, callback) {
+        var key = [];
+        options = options == true ? { entropy: true } : options || {};
+        var shortseed = mixkey(flatten4(
+          options.entropy ? [seed, tostring(pool3)] : seed == null ? autoseed() : seed,
+          3
+        ), key);
+        var arc4 = new ARC4(key);
+        var prng = function() {
+          var n = arc4.g(chunks), d = startdenom, x = 0;
+          while (n < significance) {
+            n = (n + x) * width;
+            d *= width;
+            x = arc4.g(1);
+          }
+          while (n >= overflow) {
+            n /= 2;
+            d /= 2;
+            x >>>= 1;
+          }
+          return (n + x) / d;
+        };
+        prng.int32 = function() {
+          return arc4.g(4) | 0;
+        };
+        prng.quick = function() {
+          return arc4.g(4) / 4294967296;
+        };
+        prng.double = prng;
+        mixkey(tostring(arc4.S), pool3);
+        return (options.pass || callback || function(prng2, seed2, is_math_call, state) {
+          if (state) {
+            if (state.S) {
+              copy(state, arc4);
+            }
+            prng2.state = function() {
+              return copy(arc4, {});
+            };
+          }
+          if (is_math_call) {
+            math[rngname] = prng2;
+            return seed2;
+          } else
+            return prng2;
+        })(
+          prng,
+          shortseed,
+          "global" in options ? options.global : this == math,
+          options.state
+        );
+      }
+      function ARC4(key) {
+        var t, keylen = key.length, me = this, i = 0, j = me.i = me.j = 0, s = me.S = [];
+        if (!keylen) {
+          key = [keylen++];
+        }
+        while (i < width) {
+          s[i] = i++;
+        }
+        for (i = 0; i < width; i++) {
+          s[i] = s[j = mask & j + key[i % keylen] + (t = s[i])];
+          s[j] = t;
+        }
+        (me.g = function(count2) {
+          var t2, r = 0, i2 = me.i, j2 = me.j, s2 = me.S;
+          while (count2--) {
+            t2 = s2[i2 = mask & i2 + 1];
+            r = r * width + s2[mask & (s2[i2] = s2[j2 = mask & j2 + t2]) + (s2[j2] = t2)];
+          }
+          me.i = i2;
+          me.j = j2;
+          return r;
+        })(width);
+      }
+      function copy(f, t) {
+        t.i = f.i;
+        t.j = f.j;
+        t.S = f.S.slice();
+        return t;
+      }
+      ;
+      function flatten4(obj, depth) {
+        var result = [], typ = typeof obj, prop;
+        if (depth && typ == "object") {
+          for (prop in obj) {
+            try {
+              result.push(flatten4(obj[prop], depth - 1));
+            } catch (e) {
+            }
+          }
+        }
+        return result.length ? result : typ == "string" ? obj : obj + "\0";
+      }
+      function mixkey(seed, key) {
+        var stringseed = seed + "", smear, j = 0;
+        while (j < stringseed.length) {
+          key[mask & j] = mask & (smear ^= key[mask & j] * 19) + stringseed.charCodeAt(j++);
+        }
+        return tostring(key);
+      }
+      function autoseed() {
+        try {
+          var out;
+          if (nodecrypto && (out = nodecrypto.randomBytes)) {
+            out = out(width);
+          } else {
+            out = new Uint8Array(width);
+            (global2.crypto || global2.msCrypto).getRandomValues(out);
+          }
+          return tostring(out);
+        } catch (e) {
+          var browser = global2.navigator, plugins = browser && browser.plugins;
+          return [+new Date(), global2, plugins, global2.screen, tostring(pool3)];
+        }
+      }
+      function tostring(a) {
+        return String.fromCharCode.apply(0, a);
+      }
+      mixkey(math.random(), pool3);
+      if (typeof module == "object" && module.exports) {
+        module.exports = seedrandom5;
+        try {
+          nodecrypto = require_crypto();
+        } catch (ex) {
+        }
+      } else if (typeof define == "function" && define.amd) {
+        define(function() {
+          return seedrandom5;
+        });
+      } else {
+        math["seed" + rngname] = seedrandom5;
+      }
+    })(
+      typeof self !== "undefined" ? self : exports,
+      [],
+      Math
+    );
+  }
+});
+
+// node_modules/seedrandom/index.js
+var require_seedrandom2 = __commonJS({
+  "node_modules/seedrandom/index.js"(exports, module) {
+    var alea5 = require_alea();
+    var xor128 = require_xor128();
+    var xorwow = require_xorwow();
+    var xorshift7 = require_xorshift7();
+    var xor4096 = require_xor4096();
+    var tychei = require_tychei();
+    var sr = require_seedrandom();
+    sr.alea = alea5;
+    sr.xor128 = xor128;
+    sr.xorwow = xorwow;
+    sr.xorshift7 = xorshift7;
+    sr.xor4096 = xor4096;
+    sr.tychei = tychei;
+    module.exports = sr;
+  }
+});
+
+// (disabled):node_modules/string_decoder/lib/string_decoder.js
+var require_string_decoder = __commonJS({
+  "(disabled):node_modules/string_decoder/lib/string_decoder.js"() {
+  }
+});
+
+// (disabled):fs
+var require_fs = __commonJS({
+  "(disabled):fs"() {
+  }
+});
+
+// (disabled):path
+var require_path = __commonJS({
+  "(disabled):path"() {
+  }
+});
+
+// (disabled):worker_threads
+var require_worker_threads = __commonJS({
+  "(disabled):worker_threads"() {
+  }
+});
+
+// (disabled):perf_hooks
+var require_perf_hooks = __commonJS({
+  "(disabled):perf_hooks"() {
+  }
+});
+
+// (disabled):os
+var require_os = __commonJS({
+  "(disabled):os"() {
+  }
+});
+
+// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js
+var require_tfjs_backend_wasm_threaded_simd = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(exports, module) {
+    var WasmBackendModuleThreadedSimd2 = (() => {
+      var _scriptDir = typeof document !== "undefined" && document.currentScript ? document.currentScript.src : void 0;
+      if (typeof __filename !== "undefined")
+        _scriptDir = _scriptDir || __filename;
+      return function(WasmBackendModuleThreadedSimd3) {
+        WasmBackendModuleThreadedSimd3 = WasmBackendModuleThreadedSimd3 || {};
+        function GROWABLE_HEAP_I8() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP8;
+        }
+        function GROWABLE_HEAP_U8() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPU8;
+        }
+        function GROWABLE_HEAP_I16() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP16;
+        }
+        function GROWABLE_HEAP_I32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAP32;
+        }
+        function GROWABLE_HEAP_U32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPU32;
+        }
+        function GROWABLE_HEAP_F32() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPF32;
+        }
+        function GROWABLE_HEAP_F64() {
+          if (wasmMemory.buffer != buffer2) {
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+          }
+          return HEAPF64;
+        }
+        var Module = typeof WasmBackendModuleThreadedSimd3 != "undefined" ? WasmBackendModuleThreadedSimd3 : {};
+        var readyPromiseResolve, readyPromiseReject;
+        Module["ready"] = new Promise(function(resolve, reject) {
+          readyPromiseResolve = resolve;
+          readyPromiseReject = reject;
+        });
+        var beforeListeners;
+        if (typeof process !== "undefined" && process.listeners) {
+          beforeListeners = { uncaughtException: process.listeners("uncaughtException"), unhandledRejection: process.listeners("unhandledRejection") };
+        }
+        var moduleOverrides = Object.assign({}, Module);
+        var arguments_ = [];
+        var thisProgram = "./this.program";
+        var quit_ = (status, toThrow) => {
+          throw toThrow;
+        };
+        var ENVIRONMENT_IS_WEB = typeof window == "object";
+        var ENVIRONMENT_IS_WORKER = typeof importScripts == "function";
+        var ENVIRONMENT_IS_NODE = typeof process == "object" && typeof process.versions == "object" && typeof process.versions.node == "string";
+        var ENVIRONMENT_IS_PTHREAD = Module["ENVIRONMENT_IS_PTHREAD"] || false;
+        var scriptDirectory = "";
+        function locateFile(path) {
+          if (Module["locateFile"]) {
+            return Module["locateFile"](path, scriptDirectory);
+          }
+          return scriptDirectory + path;
+        }
+        var read_, readAsync, readBinary, setWindowTitle;
+        function logExceptionOnExit(e) {
+          if (e instanceof ExitStatus)
+            return;
+          let toLog = e;
+          err("exiting due to exception: " + toLog);
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = require_path().dirname(scriptDirectory) + "/";
+          } else {
+            scriptDirectory = __dirname + "/";
+          }
+          var fs, nodePath;
+          if (typeof __require === "function") {
+            fs = require_fs();
+            nodePath = require_path();
+          }
+          read_ = (filename, binary) => {
+            filename = nodePath["normalize"](filename);
+            return fs.readFileSync(filename, binary ? void 0 : "utf8");
+          };
+          readBinary = (filename) => {
+            var ret = read_(filename, true);
+            if (!ret.buffer) {
+              ret = new Uint8Array(ret);
+            }
+            return ret;
+          };
+          readAsync = (filename, onload, onerror) => {
+            filename = nodePath["normalize"](filename);
+            fs.readFile(filename, function(err2, data) {
+              if (err2)
+                onerror(err2);
+              else
+                onload(data.buffer);
+            });
+          };
+          if (process["argv"].length > 1) {
+            thisProgram = process["argv"][1].replace(/\\/g, "/");
+          }
+          arguments_ = process["argv"].slice(2);
+          process["on"]("uncaughtException", function(ex) {
+            if (!(ex instanceof ExitStatus)) {
+              throw ex;
+            }
+          });
+          process["on"]("unhandledRejection", function(reason) {
+            throw reason;
+          });
+          quit_ = (status, toThrow) => {
+            if (keepRuntimeAlive()) {
+              process["exitCode"] = status;
+              throw toThrow;
+            }
+            logExceptionOnExit(toThrow);
+            process["exit"](status);
+          };
+          Module["inspect"] = function() {
+            return "[Emscripten Module object]";
+          };
+          let nodeWorkerThreads;
+          try {
+            nodeWorkerThreads = require_worker_threads();
+          } catch (e) {
+            console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?');
+            throw e;
+          }
+          global.Worker = nodeWorkerThreads.Worker;
+        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = self.location.href;
+          } else if (typeof document != "undefined" && document.currentScript) {
+            scriptDirectory = document.currentScript.src;
+          }
+          if (typeof _scriptDir !== "undefined" && _scriptDir) {
+            scriptDirectory = _scriptDir;
+          }
+          if (scriptDirectory.indexOf("blob:") !== 0) {
+            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, "").lastIndexOf("/") + 1);
+          } else {
+            scriptDirectory = "";
+          }
+          if (!ENVIRONMENT_IS_NODE) {
+            read_ = (url) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, false);
+              xhr.send(null);
+              return xhr.responseText;
+            };
+            if (ENVIRONMENT_IS_WORKER) {
+              readBinary = (url) => {
+                var xhr = new XMLHttpRequest();
+                xhr.open("GET", url, false);
+                xhr.responseType = "arraybuffer";
+                xhr.send(null);
+                return new Uint8Array(xhr.response);
+              };
+            }
+            readAsync = (url, onload, onerror) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, true);
+              xhr.responseType = "arraybuffer";
+              xhr.onload = () => {
+                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {
+                  onload(xhr.response);
+                  return;
+                }
+                onerror();
+              };
+              xhr.onerror = onerror;
+              xhr.send(null);
+            };
+          }
+          setWindowTitle = (title) => document.title = title;
+        } else {
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (typeof performance == "undefined") {
+            global.performance = require_perf_hooks().performance;
+          }
+        }
+        var defaultPrint = console.log.bind(console);
+        var defaultPrintErr = console.warn.bind(console);
+        if (ENVIRONMENT_IS_NODE) {
+          defaultPrint = (str) => fs.writeSync(1, str + "\n");
+          defaultPrintErr = (str) => fs.writeSync(2, str + "\n");
+        }
+        var out = Module["print"] || defaultPrint;
+        var err = Module["printErr"] || defaultPrintErr;
+        Object.assign(Module, moduleOverrides);
+        moduleOverrides = null;
+        if (Module["arguments"])
+          arguments_ = Module["arguments"];
+        if (Module["thisProgram"])
+          thisProgram = Module["thisProgram"];
+        if (Module["quit"])
+          quit_ = Module["quit"];
+        var POINTER_SIZE = 4;
+        var Atomics_load = Atomics.load;
+        var Atomics_store = Atomics.store;
+        var Atomics_compareExchange = Atomics.compareExchange;
+        var wasmBinary;
+        if (Module["wasmBinary"])
+          wasmBinary = Module["wasmBinary"];
+        var noExitRuntime = Module["noExitRuntime"] || true;
+        if (typeof WebAssembly != "object") {
+          abort("no native wasm support detected");
+        }
+        var wasmMemory;
+        var wasmModule;
+        var ABORT = false;
+        var EXITSTATUS;
+        function assert3(condition, text) {
+          if (!condition) {
+            abort(text);
+          }
+        }
+        var UTF8Decoder = typeof TextDecoder != "undefined" ? new TextDecoder("utf8") : void 0;
+        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {
+          var endIdx = idx + maxBytesToRead;
+          var endPtr = idx;
+          while (heapOrArray[endPtr] && !(endPtr >= endIdx))
+            ++endPtr;
+          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {
+            return UTF8Decoder.decode(heapOrArray.buffer instanceof SharedArrayBuffer ? heapOrArray.slice(idx, endPtr) : heapOrArray.subarray(idx, endPtr));
+          }
+          var str = "";
+          while (idx < endPtr) {
+            var u0 = heapOrArray[idx++];
+            if (!(u0 & 128)) {
+              str += String.fromCharCode(u0);
+              continue;
+            }
+            var u1 = heapOrArray[idx++] & 63;
+            if ((u0 & 224) == 192) {
+              str += String.fromCharCode((u0 & 31) << 6 | u1);
+              continue;
+            }
+            var u2 = heapOrArray[idx++] & 63;
+            if ((u0 & 240) == 224) {
+              u0 = (u0 & 15) << 12 | u1 << 6 | u2;
+            } else {
+              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;
+            }
+            if (u0 < 65536) {
+              str += String.fromCharCode(u0);
+            } else {
+              var ch = u0 - 65536;
+              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);
+            }
+          }
+          return str;
+        }
+        function UTF8ToString(ptr, maxBytesToRead) {
+          return ptr ? UTF8ArrayToString(GROWABLE_HEAP_U8(), ptr, maxBytesToRead) : "";
+        }
+        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {
+          if (!(maxBytesToWrite > 0))
+            return 0;
+          var startIdx = outIdx;
+          var endIdx = outIdx + maxBytesToWrite - 1;
+          for (var i = 0; i < str.length; ++i) {
+            var u = str.charCodeAt(i);
+            if (u >= 55296 && u <= 57343) {
+              var u1 = str.charCodeAt(++i);
+              u = 65536 + ((u & 1023) << 10) | u1 & 1023;
+            }
+            if (u <= 127) {
+              if (outIdx >= endIdx)
+                break;
+              heap[outIdx++] = u;
+            } else if (u <= 2047) {
+              if (outIdx + 1 >= endIdx)
+                break;
+              heap[outIdx++] = 192 | u >> 6;
+              heap[outIdx++] = 128 | u & 63;
+            } else if (u <= 65535) {
+              if (outIdx + 2 >= endIdx)
+                break;
+              heap[outIdx++] = 224 | u >> 12;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            } else {
+              if (outIdx + 3 >= endIdx)
+                break;
+              heap[outIdx++] = 240 | u >> 18;
+              heap[outIdx++] = 128 | u >> 12 & 63;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            }
+          }
+          heap[outIdx] = 0;
+          return outIdx - startIdx;
+        }
+        function stringToUTF8(str, outPtr, maxBytesToWrite) {
+          return stringToUTF8Array(str, GROWABLE_HEAP_U8(), outPtr, maxBytesToWrite);
+        }
+        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;
+        if (ENVIRONMENT_IS_PTHREAD) {
+          buffer2 = Module["buffer"];
+        }
+        function updateGlobalBufferAndViews(buf) {
+          buffer2 = buf;
+          Module["HEAP8"] = HEAP8 = new Int8Array(buf);
+          Module["HEAP16"] = HEAP16 = new Int16Array(buf);
+          Module["HEAP32"] = HEAP32 = new Int32Array(buf);
+          Module["HEAPU8"] = HEAPU8 = new Uint8Array(buf);
+          Module["HEAPU16"] = HEAPU16 = new Uint16Array(buf);
+          Module["HEAPU32"] = HEAPU32 = new Uint32Array(buf);
+          Module["HEAPF32"] = HEAPF32 = new Float32Array(buf);
+          Module["HEAPF64"] = HEAPF64 = new Float64Array(buf);
+        }
+        var INITIAL_MEMORY = Module["INITIAL_MEMORY"] || 16777216;
+        if (ENVIRONMENT_IS_PTHREAD) {
+          wasmMemory = Module["wasmMemory"];
+          buffer2 = Module["buffer"];
+        } else {
+          if (Module["wasmMemory"]) {
+            wasmMemory = Module["wasmMemory"];
+          } else {
+            wasmMemory = new WebAssembly.Memory({ "initial": INITIAL_MEMORY / 65536, "maximum": 2147483648 / 65536, "shared": true });
+            if (!(wasmMemory.buffer instanceof SharedArrayBuffer)) {
+              err("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag");
+              if (ENVIRONMENT_IS_NODE) {
+                console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)");
+              }
+              throw Error("bad memory");
+            }
+          }
+        }
+        if (wasmMemory) {
+          buffer2 = wasmMemory.buffer;
+        }
+        INITIAL_MEMORY = buffer2.byteLength;
+        updateGlobalBufferAndViews(buffer2);
+        var wasmTable;
+        var __ATPRERUN__ = [];
+        var __ATINIT__ = [];
+        var __ATPOSTRUN__ = [];
+        var runtimeInitialized = false;
+        function keepRuntimeAlive() {
+          return noExitRuntime;
+        }
+        function preRun() {
+          if (Module["preRun"]) {
+            if (typeof Module["preRun"] == "function")
+              Module["preRun"] = [Module["preRun"]];
+            while (Module["preRun"].length) {
+              addOnPreRun(Module["preRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPRERUN__);
+        }
+        function initRuntime() {
+          runtimeInitialized = true;
+          if (ENVIRONMENT_IS_PTHREAD)
+            return;
+          callRuntimeCallbacks(__ATINIT__);
+        }
+        function postRun() {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return;
+          if (Module["postRun"]) {
+            if (typeof Module["postRun"] == "function")
+              Module["postRun"] = [Module["postRun"]];
+            while (Module["postRun"].length) {
+              addOnPostRun(Module["postRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPOSTRUN__);
+        }
+        function addOnPreRun(cb) {
+          __ATPRERUN__.unshift(cb);
+        }
+        function addOnInit(cb) {
+          __ATINIT__.unshift(cb);
+        }
+        function addOnPostRun(cb) {
+          __ATPOSTRUN__.unshift(cb);
+        }
+        var runDependencies = 0;
+        var runDependencyWatcher = null;
+        var dependenciesFulfilled = null;
+        function addRunDependency(id) {
+          runDependencies++;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+        }
+        function removeRunDependency(id) {
+          runDependencies--;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+          if (runDependencies == 0) {
+            if (runDependencyWatcher !== null) {
+              clearInterval(runDependencyWatcher);
+              runDependencyWatcher = null;
+            }
+            if (dependenciesFulfilled) {
+              var callback = dependenciesFulfilled;
+              dependenciesFulfilled = null;
+              callback();
+            }
+          }
+        }
+        function abort(what) {
+          if (ENVIRONMENT_IS_PTHREAD) {
+            postMessage({ "cmd": "onAbort", "arg": what });
+          } else {
+            if (Module["onAbort"]) {
+              Module["onAbort"](what);
+            }
+          }
+          what = "Aborted(" + what + ")";
+          err(what);
+          ABORT = true;
+          EXITSTATUS = 1;
+          what += ". Build with -sASSERTIONS for more info.";
+          var e = new WebAssembly.RuntimeError(what);
+          readyPromiseReject(e);
+          throw e;
+        }
+        var dataURIPrefix = "data:application/octet-stream;base64,";
+        function isDataURI(filename) {
+          return filename.startsWith(dataURIPrefix);
+        }
+        function isFileURI(filename) {
+          return filename.startsWith("file://");
+        }
+        var wasmBinaryFile;
+        wasmBinaryFile = "tfjs-backend-wasm-threaded-simd.wasm";
+        if (!isDataURI(wasmBinaryFile)) {
+          wasmBinaryFile = locateFile(wasmBinaryFile);
+        }
+        function getBinary(file) {
+          try {
+            if (file == wasmBinaryFile && wasmBinary) {
+              return new Uint8Array(wasmBinary);
+            }
+            if (readBinary) {
+              return readBinary(file);
+            }
+            throw "both async and sync fetching of the wasm failed";
+          } catch (err2) {
+            abort(err2);
+          }
+        }
+        function getBinaryPromise() {
+          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {
+            if (typeof fetch == "function" && !isFileURI(wasmBinaryFile)) {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                if (!response["ok"]) {
+                  throw "failed to load wasm binary file at '" + wasmBinaryFile + "'";
+                }
+                return response["arrayBuffer"]();
+              }).catch(function() {
+                return getBinary(wasmBinaryFile);
+              });
+            } else {
+              if (readAsync) {
+                return new Promise(function(resolve, reject) {
+                  readAsync(wasmBinaryFile, function(response) {
+                    resolve(new Uint8Array(response));
+                  }, reject);
+                });
+              }
+            }
+          }
+          return Promise.resolve().then(function() {
+            return getBinary(wasmBinaryFile);
+          });
+        }
+        function createWasm() {
+          var info = { "env": asmLibraryArg, "wasi_snapshot_preview1": asmLibraryArg };
+          function receiveInstance(instance, module2) {
+            var exports3 = instance.exports;
+            Module["asm"] = exports3;
+            registerTLSInit(Module["asm"]["_emscripten_tls_init"]);
+            wasmTable = Module["asm"]["__indirect_function_table"];
+            addOnInit(Module["asm"]["__wasm_call_ctors"]);
+            wasmModule = module2;
+            if (!ENVIRONMENT_IS_PTHREAD) {
+              var numWorkersToLoad = PThread.unusedWorkers.length;
+              PThread.unusedWorkers.forEach(function(w) {
+                PThread.loadWasmModuleToWorker(w, function() {
+                  if (!--numWorkersToLoad)
+                    removeRunDependency("wasm-instantiate");
+                });
+              });
+            }
+          }
+          if (!ENVIRONMENT_IS_PTHREAD) {
+            addRunDependency("wasm-instantiate");
+          }
+          function receiveInstantiationResult(result) {
+            receiveInstance(result["instance"], result["module"]);
+          }
+          function instantiateArrayBuffer(receiver) {
+            return getBinaryPromise().then(function(binary) {
+              return WebAssembly.instantiate(binary, info);
+            }).then(function(instance) {
+              return instance;
+            }).then(receiver, function(reason) {
+              err("failed to asynchronously prepare wasm: " + reason);
+              abort(reason);
+            });
+          }
+          function instantiateAsync() {
+            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == "function" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == "function") {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                var result = WebAssembly.instantiateStreaming(response, info);
+                return result.then(receiveInstantiationResult, function(reason) {
+                  err("wasm streaming compile failed: " + reason);
+                  err("falling back to ArrayBuffer instantiation");
+                  return instantiateArrayBuffer(receiveInstantiationResult);
+                });
+              });
+            } else {
+              return instantiateArrayBuffer(receiveInstantiationResult);
+            }
+          }
+          if (Module["instantiateWasm"]) {
+            try {
+              var exports2 = Module["instantiateWasm"](info, receiveInstance);
+              return exports2;
+            } catch (e) {
+              err("Module.instantiateWasm callback failed with error: " + e);
+              readyPromiseReject(e);
+            }
+          }
+          instantiateAsync().catch(readyPromiseReject);
+          return {};
+        }
+        var tempDouble;
+        var tempI64;
+        var ASM_CONSTS = {};
+        function ExitStatus(status) {
+          this.name = "ExitStatus";
+          this.message = "Program terminated with exit(" + status + ")";
+          this.status = status;
+        }
+        function killThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          delete PThread.pthreads[pthread_ptr];
+          worker.terminate();
+          __emscripten_thread_free_data(pthread_ptr);
+          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);
+          worker.pthread_ptr = 0;
+        }
+        function cancelThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          worker.postMessage({ "cmd": "cancel" });
+        }
+        function cleanupThread(pthread_ptr) {
+          var worker = PThread.pthreads[pthread_ptr];
+          assert3(worker);
+          PThread.returnWorkerToPool(worker);
+        }
+        function spawnThread(threadParams) {
+          var worker = PThread.getNewWorker();
+          if (!worker) {
+            return 6;
+          }
+          PThread.runningWorkers.push(worker);
+          PThread.pthreads[threadParams.pthread_ptr] = worker;
+          worker.pthread_ptr = threadParams.pthread_ptr;
+          var msg = { "cmd": "run", "start_routine": threadParams.startRoutine, "arg": threadParams.arg, "pthread_ptr": threadParams.pthread_ptr };
+          worker.runPthread = () => {
+            msg.time = performance.now();
+            worker.postMessage(msg, threadParams.transferList);
+          };
+          if (worker.loaded) {
+            worker.runPthread();
+            delete worker.runPthread;
+          }
+          return 0;
+        }
+        var SYSCALLS = { varargs: void 0, get: function() {
+          SYSCALLS.varargs += 4;
+          var ret = GROWABLE_HEAP_I32()[SYSCALLS.varargs - 4 >> 2];
+          return ret;
+        }, getStr: function(ptr) {
+          var ret = UTF8ToString(ptr);
+          return ret;
+        } };
+        function _proc_exit(code) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(1, 1, code);
+          EXITSTATUS = code;
+          if (!keepRuntimeAlive()) {
+            PThread.terminateAllThreads();
+            if (Module["onExit"])
+              Module["onExit"](code);
+            ABORT = true;
+          }
+          quit_(code, new ExitStatus(code));
+        }
+        function exitJS(status, implicit) {
+          EXITSTATUS = status;
+          if (!implicit) {
+            if (ENVIRONMENT_IS_PTHREAD) {
+              exitOnMainThread(status);
+              throw "unwind";
+            } else {
+            }
+          }
+          _proc_exit(status);
+        }
+        var _exit = exitJS;
+        function handleException(e) {
+          if (e instanceof ExitStatus || e == "unwind") {
+            return EXITSTATUS;
+          }
+          quit_(1, e);
+        }
+        var PThread = { unusedWorkers: [], runningWorkers: [], tlsInitFunctions: [], pthreads: {}, init: function() {
+          if (ENVIRONMENT_IS_PTHREAD) {
+            PThread.initWorker();
+          } else {
+            PThread.initMainThread();
+          }
+        }, initMainThread: function() {
+          var pthreadPoolSize = 8;
+          while (pthreadPoolSize--) {
+            PThread.allocateUnusedWorker();
+          }
+        }, initWorker: function() {
+          noExitRuntime = false;
+        }, setExitStatus: function(status) {
+          EXITSTATUS = status;
+        }, terminateAllThreads: function() {
+          for (var worker of Object.values(PThread.pthreads)) {
+            PThread.returnWorkerToPool(worker);
+          }
+          for (var worker of PThread.unusedWorkers) {
+            worker.terminate();
+          }
+          PThread.unusedWorkers = [];
+        }, returnWorkerToPool: function(worker) {
+          var pthread_ptr = worker.pthread_ptr;
+          delete PThread.pthreads[pthread_ptr];
+          PThread.unusedWorkers.push(worker);
+          PThread.runningWorkers.splice(PThread.runningWorkers.indexOf(worker), 1);
+          worker.pthread_ptr = 0;
+          __emscripten_thread_free_data(pthread_ptr);
+        }, receiveObjectTransfer: function(data) {
+        }, threadInitTLS: function() {
+          PThread.tlsInitFunctions.forEach((f) => f());
+        }, loadWasmModuleToWorker: function(worker, onFinishedLoading) {
+          worker.onmessage = (e) => {
+            var d = e["data"];
+            var cmd = d["cmd"];
+            if (worker.pthread_ptr)
+              PThread.currentProxiedOperationCallerThread = worker.pthread_ptr;
+            if (d["targetThread"] && d["targetThread"] != _pthread_self()) {
+              var targetWorker = PThread.pthreads[d.targetThread];
+              if (targetWorker) {
+                targetWorker.postMessage(d, d["transferList"]);
+              } else {
+                err('Internal error! Worker sent a message "' + cmd + '" to target pthread ' + d["targetThread"] + ", but that thread no longer exists!");
+              }
+              PThread.currentProxiedOperationCallerThread = void 0;
+              return;
+            }
+            if (cmd === "processProxyingQueue") {
+              executeNotifiedProxyingQueue(d["queue"]);
+            } else if (cmd === "spawnThread") {
+              spawnThread(d);
+            } else if (cmd === "cleanupThread") {
+              cleanupThread(d["thread"]);
+            } else if (cmd === "killThread") {
+              killThread(d["thread"]);
+            } else if (cmd === "cancelThread") {
+              cancelThread(d["thread"]);
+            } else if (cmd === "loaded") {
+              worker.loaded = true;
+              if (onFinishedLoading)
+                onFinishedLoading(worker);
+              if (worker.runPthread) {
+                worker.runPthread();
+                delete worker.runPthread;
+              }
+            } else if (cmd === "print") {
+              out("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (cmd === "printErr") {
+              err("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (cmd === "alert") {
+              alert("Thread " + d["threadId"] + ": " + d["text"]);
+            } else if (d.target === "setimmediate") {
+              worker.postMessage(d);
+            } else if (cmd === "onAbort") {
+              if (Module["onAbort"]) {
+                Module["onAbort"](d["arg"]);
+              }
+            } else if (cmd) {
+              err("worker sent an unknown command " + cmd);
+            }
+            PThread.currentProxiedOperationCallerThread = void 0;
+          };
+          worker.onerror = (e) => {
+            var message = "worker sent an error!";
+            err(message + " " + e.filename + ":" + e.lineno + ": " + e.message);
+            throw e;
+          };
+          if (ENVIRONMENT_IS_NODE) {
+            worker.on("message", function(data) {
+              worker.onmessage({ data });
+            });
+            worker.on("error", function(e) {
+              worker.onerror(e);
+            });
+            worker.on("detachedExit", function() {
+            });
+          }
+          worker.postMessage({ "cmd": "load", "urlOrBlob": Module["mainScriptUrlOrBlob"] || _scriptDir, "wasmMemory": wasmMemory, "wasmModule": wasmModule });
+        }, allocateUnusedWorker: function() {
+          var pthreadMainJs = locateFile("tfjs-backend-wasm-threaded-simd.worker.js");
+          PThread.unusedWorkers.push(new Worker(pthreadMainJs));
+        }, getNewWorker: function() {
+          if (PThread.unusedWorkers.length == 0) {
+            PThread.allocateUnusedWorker();
+            PThread.loadWasmModuleToWorker(PThread.unusedWorkers[0]);
+          }
+          return PThread.unusedWorkers.pop();
+        } };
+        Module["PThread"] = PThread;
+        function callRuntimeCallbacks(callbacks2) {
+          while (callbacks2.length > 0) {
+            callbacks2.shift()(Module);
+          }
+        }
+        function withStackSave(f) {
+          var stack2 = stackSave();
+          var ret = f();
+          stackRestore(stack2);
+          return ret;
+        }
+        function demangle(func2) {
+          return func2;
+        }
+        function demangleAll(text) {
+          var regex = /\b_Z[\w\d_]+/g;
+          return text.replace(regex, function(x) {
+            var y = demangle(x);
+            return x === y ? x : y + " [" + x + "]";
+          });
+        }
+        function establishStackSpace() {
+          var pthread_ptr = _pthread_self();
+          var stackTop = GROWABLE_HEAP_I32()[pthread_ptr + 44 >> 2];
+          var stackSize = GROWABLE_HEAP_I32()[pthread_ptr + 48 >> 2];
+          var stackMax = stackTop - stackSize;
+          _emscripten_stack_set_limits(stackTop, stackMax);
+          stackRestore(stackTop);
+        }
+        Module["establishStackSpace"] = establishStackSpace;
+        function exitOnMainThread(returnCode) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(2, 0, returnCode);
+          try {
+            _exit(returnCode);
+          } catch (e) {
+            handleException(e);
+          }
+        }
+        var wasmTableMirror = [];
+        function getWasmTableEntry(funcPtr) {
+          var func2 = wasmTableMirror[funcPtr];
+          if (!func2) {
+            if (funcPtr >= wasmTableMirror.length)
+              wasmTableMirror.length = funcPtr + 1;
+            wasmTableMirror[funcPtr] = func2 = wasmTable.get(funcPtr);
+          }
+          return func2;
+        }
+        function invokeEntryPoint(ptr, arg) {
+          var result = getWasmTableEntry(ptr)(arg);
+          if (keepRuntimeAlive()) {
+            PThread.setExitStatus(result);
+          } else {
+            __emscripten_thread_exit(result);
+          }
+        }
+        Module["invokeEntryPoint"] = invokeEntryPoint;
+        function jsStackTrace() {
+          var error = new Error();
+          if (!error.stack) {
+            try {
+              throw new Error();
+            } catch (e) {
+              error = e;
+            }
+            if (!error.stack) {
+              return "(no stack trace available)";
+            }
+          }
+          return error.stack.toString();
+        }
+        function registerTLSInit(tlsInitFunc) {
+          PThread.tlsInitFunctions.push(tlsInitFunc);
+        }
+        function writeArrayToMemory(array2, buffer3) {
+          GROWABLE_HEAP_I8().set(array2, buffer3);
+        }
+        function ___emscripten_init_main_thread_js(tb) {
+          __emscripten_thread_init(tb, !ENVIRONMENT_IS_WORKER, 1, !ENVIRONMENT_IS_WEB);
+          PThread.threadInitTLS();
+        }
+        function ___emscripten_thread_cleanup(thread) {
+          if (!ENVIRONMENT_IS_PTHREAD)
+            cleanupThread(thread);
+          else
+            postMessage({ "cmd": "cleanupThread", "thread": thread });
+        }
+        function pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(3, 1, pthread_ptr, attr, startRoutine, arg);
+          return ___pthread_create_js(pthread_ptr, attr, startRoutine, arg);
+        }
+        function ___pthread_create_js(pthread_ptr, attr, startRoutine, arg) {
+          if (typeof SharedArrayBuffer == "undefined") {
+            err("Current environment does not support SharedArrayBuffer, pthreads are not available!");
+            return 6;
+          }
+          var transferList = [];
+          var error = 0;
+          if (ENVIRONMENT_IS_PTHREAD && (transferList.length === 0 || error)) {
+            return pthreadCreateProxied(pthread_ptr, attr, startRoutine, arg);
+          }
+          if (error)
+            return error;
+          var threadParams = { startRoutine, pthread_ptr, arg, transferList };
+          if (ENVIRONMENT_IS_PTHREAD) {
+            threadParams.cmd = "spawnThread";
+            postMessage(threadParams, transferList);
+            return 0;
+          }
+          return spawnThread(threadParams);
+        }
+        function __emscripten_default_pthread_stack_size() {
+          return 2097152;
+        }
+        var nowIsMonotonic = true;
+        function __emscripten_get_now_is_monotonic() {
+          return nowIsMonotonic;
+        }
+        function executeNotifiedProxyingQueue(queue) {
+          Atomics.store(GROWABLE_HEAP_I32(), queue >> 2, 1);
+          if (_pthread_self()) {
+            __emscripten_proxy_execute_task_queue(queue);
+          }
+          Atomics.compareExchange(GROWABLE_HEAP_I32(), queue >> 2, 1, 0);
+        }
+        Module["executeNotifiedProxyingQueue"] = executeNotifiedProxyingQueue;
+        function __emscripten_notify_task_queue(targetThreadId, currThreadId, mainThreadId, queue) {
+          if (targetThreadId == currThreadId) {
+            setTimeout(() => executeNotifiedProxyingQueue(queue));
+          } else if (ENVIRONMENT_IS_PTHREAD) {
+            postMessage({ "targetThread": targetThreadId, "cmd": "processProxyingQueue", "queue": queue });
+          } else {
+            var worker = PThread.pthreads[targetThreadId];
+            if (!worker) {
+              return;
+            }
+            worker.postMessage({ "cmd": "processProxyingQueue", "queue": queue });
+          }
+          return 1;
+        }
+        function __emscripten_set_offscreencanvas_size(target, width, height) {
+          return -1;
+        }
+        function _abort() {
+          abort("");
+        }
+        function warnOnce(text) {
+          if (!warnOnce.shown)
+            warnOnce.shown = {};
+          if (!warnOnce.shown[text]) {
+            warnOnce.shown[text] = 1;
+            if (ENVIRONMENT_IS_NODE)
+              text = "warning: " + text;
+            err(text);
+          }
+        }
+        function _emscripten_check_blocking_allowed() {
+          if (ENVIRONMENT_IS_NODE)
+            return;
+          if (ENVIRONMENT_IS_WORKER)
+            return;
+          warnOnce("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread");
+        }
+        function _emscripten_date_now() {
+          return Date.now();
+        }
+        function getHeapMax() {
+          return 2147483648;
+        }
+        function _emscripten_get_heap_max() {
+          return getHeapMax();
+        }
+        var _emscripten_get_now;
+        if (ENVIRONMENT_IS_NODE) {
+          _emscripten_get_now = () => {
+            var t = process["hrtime"]();
+            return t[0] * 1e3 + t[1] / 1e6;
+          };
+        } else if (ENVIRONMENT_IS_PTHREAD) {
+          _emscripten_get_now = () => performance.now() - Module["__performance_now_clock_drift"];
+        } else
+          _emscripten_get_now = () => performance.now();
+        function _emscripten_memcpy_big(dest, src, num) {
+          GROWABLE_HEAP_U8().copyWithin(dest, src, src + num);
+        }
+        function _emscripten_num_logical_cores() {
+          if (ENVIRONMENT_IS_NODE)
+            return require_os().cpus().length;
+          return navigator["hardwareConcurrency"];
+        }
+        function _emscripten_proxy_to_main_thread_js(index, sync) {
+          var numCallArgs = arguments.length - 2;
+          var outerArgs = arguments;
+          return withStackSave(() => {
+            var serializedNumCallArgs = numCallArgs;
+            var args = stackAlloc(serializedNumCallArgs * 8);
+            var b = args >> 3;
+            for (var i = 0; i < numCallArgs; i++) {
+              var arg = outerArgs[2 + i];
+              GROWABLE_HEAP_F64()[b + i] = arg;
+            }
+            return _emscripten_run_in_main_runtime_thread_js(index, serializedNumCallArgs, args, sync);
+          });
+        }
+        var _emscripten_receive_on_main_thread_js_callArgs = [];
+        function _emscripten_receive_on_main_thread_js(index, numCallArgs, args) {
+          _emscripten_receive_on_main_thread_js_callArgs.length = numCallArgs;
+          var b = args >> 3;
+          for (var i = 0; i < numCallArgs; i++) {
+            _emscripten_receive_on_main_thread_js_callArgs[i] = GROWABLE_HEAP_F64()[b + i];
+          }
+          var isEmAsmConst = index < 0;
+          var func2 = !isEmAsmConst ? proxiedFunctionTable[index] : ASM_CONSTS[-index - 1];
+          return func2.apply(null, _emscripten_receive_on_main_thread_js_callArgs);
+        }
+        function emscripten_realloc_buffer(size) {
+          try {
+            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            return 1;
+          } catch (e) {
+          }
+        }
+        function _emscripten_resize_heap(requestedSize) {
+          var oldSize = GROWABLE_HEAP_U8().length;
+          requestedSize = requestedSize >>> 0;
+          if (requestedSize <= oldSize) {
+            return false;
+          }
+          var maxHeapSize = getHeapMax();
+          if (requestedSize > maxHeapSize) {
+            return false;
+          }
+          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;
+          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {
+            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);
+            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);
+            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));
+            var replacement = emscripten_realloc_buffer(newSize);
+            if (replacement) {
+              return true;
+            }
+          }
+          return false;
+        }
+        function _emscripten_unwind_to_js_event_loop() {
+          throw "unwind";
+        }
+        function _fd_close(fd) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(4, 1, fd);
+          return 52;
+        }
+        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(5, 1, fd, offset_low, offset_high, whence, newOffset);
+          return 70;
+        }
+        var printCharBuffers = [null, [], []];
+        function printChar(stream, curr) {
+          var buffer3 = printCharBuffers[stream];
+          if (curr === 0 || curr === 10) {
+            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));
+            buffer3.length = 0;
+          } else {
+            buffer3.push(curr);
+          }
+        }
+        function _fd_write(fd, iov, iovcnt, pnum) {
+          if (ENVIRONMENT_IS_PTHREAD)
+            return _emscripten_proxy_to_main_thread_js(6, 1, fd, iov, iovcnt, pnum);
+          var num = 0;
+          for (var i = 0; i < iovcnt; i++) {
+            var ptr = GROWABLE_HEAP_U32()[iov >> 2];
+            var len = GROWABLE_HEAP_U32()[iov + 4 >> 2];
+            iov += 8;
+            for (var j = 0; j < len; j++) {
+              printChar(fd, GROWABLE_HEAP_U8()[ptr + j]);
+            }
+            num += len;
+          }
+          GROWABLE_HEAP_U32()[pnum >> 2] = num;
+          return 0;
+        }
+        function getCFunc(ident) {
+          var func2 = Module["_" + ident];
+          return func2;
+        }
+        function ccall(ident, returnType, argTypes, args, opts) {
+          var toC = { "string": (str) => {
+            var ret2 = 0;
+            if (str !== null && str !== void 0 && str !== 0) {
+              var len = (str.length << 2) + 1;
+              ret2 = stackAlloc(len);
+              stringToUTF8(str, ret2, len);
+            }
+            return ret2;
+          }, "array": (arr) => {
+            var ret2 = stackAlloc(arr.length);
+            writeArrayToMemory(arr, ret2);
+            return ret2;
+          } };
+          function convertReturnValue(ret2) {
+            if (returnType === "string") {
+              return UTF8ToString(ret2);
+            }
+            if (returnType === "boolean")
+              return Boolean(ret2);
+            return ret2;
+          }
+          var func2 = getCFunc(ident);
+          var cArgs = [];
+          var stack2 = 0;
+          if (args) {
+            for (var i = 0; i < args.length; i++) {
+              var converter = toC[argTypes[i]];
+              if (converter) {
+                if (stack2 === 0)
+                  stack2 = stackSave();
+                cArgs[i] = converter(args[i]);
+              } else {
+                cArgs[i] = args[i];
+              }
+            }
+          }
+          var ret = func2.apply(null, cArgs);
+          function onDone(ret2) {
+            if (stack2 !== 0)
+              stackRestore(stack2);
+            return convertReturnValue(ret2);
+          }
+          ret = onDone(ret);
+          return ret;
+        }
+        function cwrap(ident, returnType, argTypes, opts) {
+          argTypes = argTypes || [];
+          var numericArgs = argTypes.every((type) => type === "number" || type === "boolean");
+          var numericRet = returnType !== "string";
+          if (numericRet && numericArgs && !opts) {
+            return getCFunc(ident);
+          }
+          return function() {
+            return ccall(ident, returnType, argTypes, arguments, opts);
+          };
+        }
+        PThread.init();
+        var proxiedFunctionTable = [null, _proc_exit, exitOnMainThread, pthreadCreateProxied, _fd_close, _fd_seek, _fd_write];
+        var asmLibraryArg = { "__emscripten_init_main_thread_js": ___emscripten_init_main_thread_js, "__emscripten_thread_cleanup": ___emscripten_thread_cleanup, "__pthread_create_js": ___pthread_create_js, "_emscripten_default_pthread_stack_size": __emscripten_default_pthread_stack_size, "_emscripten_get_now_is_monotonic": __emscripten_get_now_is_monotonic, "_emscripten_notify_task_queue": __emscripten_notify_task_queue, "_emscripten_set_offscreencanvas_size": __emscripten_set_offscreencanvas_size, "abort": _abort, "emscripten_check_blocking_allowed": _emscripten_check_blocking_allowed, "emscripten_date_now": _emscripten_date_now, "emscripten_get_heap_max": _emscripten_get_heap_max, "emscripten_get_now": _emscripten_get_now, "emscripten_memcpy_big": _emscripten_memcpy_big, "emscripten_num_logical_cores": _emscripten_num_logical_cores, "emscripten_receive_on_main_thread_js": _emscripten_receive_on_main_thread_js, "emscripten_resize_heap": _emscripten_resize_heap, "emscripten_unwind_to_js_event_loop": _emscripten_unwind_to_js_event_loop, "exit": _exit, "fd_close": _fd_close, "fd_seek": _fd_seek, "fd_write": _fd_write, "memory": wasmMemory || Module["wasmMemory"] };
+        var asm = createWasm();
+        var ___wasm_call_ctors = Module["___wasm_call_ctors"] = function() {
+          return (___wasm_call_ctors = Module["___wasm_call_ctors"] = Module["asm"]["__wasm_call_ctors"]).apply(null, arguments);
+        };
+        var _init = Module["_init"] = function() {
+          return (_init = Module["_init"] = Module["asm"]["init"]).apply(null, arguments);
+        };
+        var _init_with_threads_count = Module["_init_with_threads_count"] = function() {
+          return (_init_with_threads_count = Module["_init_with_threads_count"] = Module["asm"]["init_with_threads_count"]).apply(null, arguments);
+        };
+        var _get_threads_count = Module["_get_threads_count"] = function() {
+          return (_get_threads_count = Module["_get_threads_count"] = Module["asm"]["get_threads_count"]).apply(null, arguments);
+        };
+        var _register_tensor = Module["_register_tensor"] = function() {
+          return (_register_tensor = Module["_register_tensor"] = Module["asm"]["register_tensor"]).apply(null, arguments);
+        };
+        var _dispose_data = Module["_dispose_data"] = function() {
+          return (_dispose_data = Module["_dispose_data"] = Module["asm"]["dispose_data"]).apply(null, arguments);
+        };
+        var _dispose = Module["_dispose"] = function() {
+          return (_dispose = Module["_dispose"] = Module["asm"]["dispose"]).apply(null, arguments);
+        };
+        var _Abs = Module["_Abs"] = function() {
+          return (_Abs = Module["_Abs"] = Module["asm"]["Abs"]).apply(null, arguments);
+        };
+        var _Add = Module["_Add"] = function() {
+          return (_Add = Module["_Add"] = Module["asm"]["Add"]).apply(null, arguments);
+        };
+        var _AddN = Module["_AddN"] = function() {
+          return (_AddN = Module["_AddN"] = Module["asm"]["AddN"]).apply(null, arguments);
+        };
+        var _All = Module["_All"] = function() {
+          return (_All = Module["_All"] = Module["asm"]["All"]).apply(null, arguments);
+        };
+        var _Any = Module["_Any"] = function() {
+          return (_Any = Module["_Any"] = Module["asm"]["Any"]).apply(null, arguments);
+        };
+        var _ArgMax = Module["_ArgMax"] = function() {
+          return (_ArgMax = Module["_ArgMax"] = Module["asm"]["ArgMax"]).apply(null, arguments);
+        };
+        var _AvgPool = Module["_AvgPool"] = function() {
+          return (_AvgPool = Module["_AvgPool"] = Module["asm"]["AvgPool"]).apply(null, arguments);
+        };
+        var _BatchMatMul = Module["_BatchMatMul"] = function() {
+          return (_BatchMatMul = Module["_BatchMatMul"] = Module["asm"]["BatchMatMul"]).apply(null, arguments);
+        };
+        var _Ceil = Module["_Ceil"] = function() {
+          return (_Ceil = Module["_Ceil"] = Module["asm"]["Ceil"]).apply(null, arguments);
+        };
+        var _ClipByValue = Module["_ClipByValue"] = function() {
+          return (_ClipByValue = Module["_ClipByValue"] = Module["asm"]["ClipByValue"]).apply(null, arguments);
+        };
+        var _Conv2D = Module["_Conv2D"] = function() {
+          return (_Conv2D = Module["_Conv2D"] = Module["asm"]["Conv2D"]).apply(null, arguments);
+        };
+        var _Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = function() {
+          return (_Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = Module["asm"]["Conv2DBackpropInput"]).apply(null, arguments);
+        };
+        var _Cos = Module["_Cos"] = function() {
+          return (_Cos = Module["_Cos"] = Module["asm"]["Cos"]).apply(null, arguments);
+        };
+        var _Cosh = Module["_Cosh"] = function() {
+          return (_Cosh = Module["_Cosh"] = Module["asm"]["Cosh"]).apply(null, arguments);
+        };
+        var _CropAndResize = Module["_CropAndResize"] = function() {
+          return (_CropAndResize = Module["_CropAndResize"] = Module["asm"]["CropAndResize"]).apply(null, arguments);
+        };
+        var _Cumprod = Module["_Cumprod"] = function() {
+          return (_Cumprod = Module["_Cumprod"] = Module["asm"]["Cumprod"]).apply(null, arguments);
+        };
+        var _Cumsum = Module["_Cumsum"] = function() {
+          return (_Cumsum = Module["_Cumsum"] = Module["asm"]["Cumsum"]).apply(null, arguments);
+        };
+        var _DepthToSpace = Module["_DepthToSpace"] = function() {
+          return (_DepthToSpace = Module["_DepthToSpace"] = Module["asm"]["DepthToSpace"]).apply(null, arguments);
+        };
+        var _DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = function() {
+          return (_DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = Module["asm"]["DepthwiseConv2dNative"]).apply(null, arguments);
+        };
+        var _Elu = Module["_Elu"] = function() {
+          return (_Elu = Module["_Elu"] = Module["asm"]["Elu"]).apply(null, arguments);
+        };
+        var _Equal = Module["_Equal"] = function() {
+          return (_Equal = Module["_Equal"] = Module["asm"]["Equal"]).apply(null, arguments);
+        };
+        var _Exp = Module["_Exp"] = function() {
+          return (_Exp = Module["_Exp"] = Module["asm"]["Exp"]).apply(null, arguments);
+        };
+        var _FlipLeftRight = Module["_FlipLeftRight"] = function() {
+          return (_FlipLeftRight = Module["_FlipLeftRight"] = Module["asm"]["FlipLeftRight"]).apply(null, arguments);
+        };
+        var _Floor = Module["_Floor"] = function() {
+          return (_Floor = Module["_Floor"] = Module["asm"]["Floor"]).apply(null, arguments);
+        };
+        var _FloorDiv = Module["_FloorDiv"] = function() {
+          return (_FloorDiv = Module["_FloorDiv"] = Module["asm"]["FloorDiv"]).apply(null, arguments);
+        };
+        var _FusedBatchNorm = Module["_FusedBatchNorm"] = function() {
+          return (_FusedBatchNorm = Module["_FusedBatchNorm"] = Module["asm"]["FusedBatchNorm"]).apply(null, arguments);
+        };
+        var _FusedConv2D = Module["_FusedConv2D"] = function() {
+          return (_FusedConv2D = Module["_FusedConv2D"] = Module["asm"]["FusedConv2D"]).apply(null, arguments);
+        };
+        var _FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = function() {
+          return (_FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = Module["asm"]["FusedDepthwiseConv2D"]).apply(null, arguments);
+        };
+        var _Gather = Module["_Gather"] = function() {
+          return (_Gather = Module["_Gather"] = Module["asm"]["Gather"]).apply(null, arguments);
+        };
+        var _GatherNd = Module["_GatherNd"] = function() {
+          return (_GatherNd = Module["_GatherNd"] = Module["asm"]["GatherNd"]).apply(null, arguments);
+        };
+        var _Greater = Module["_Greater"] = function() {
+          return (_Greater = Module["_Greater"] = Module["asm"]["Greater"]).apply(null, arguments);
+        };
+        var _GreaterEqual = Module["_GreaterEqual"] = function() {
+          return (_GreaterEqual = Module["_GreaterEqual"] = Module["asm"]["GreaterEqual"]).apply(null, arguments);
+        };
+        var _LeakyRelu = Module["_LeakyRelu"] = function() {
+          return (_LeakyRelu = Module["_LeakyRelu"] = Module["asm"]["LeakyRelu"]).apply(null, arguments);
+        };
+        var _Less = Module["_Less"] = function() {
+          return (_Less = Module["_Less"] = Module["asm"]["Less"]).apply(null, arguments);
+        };
+        var _LessEqual = Module["_LessEqual"] = function() {
+          return (_LessEqual = Module["_LessEqual"] = Module["asm"]["LessEqual"]).apply(null, arguments);
+        };
+        var _Log = Module["_Log"] = function() {
+          return (_Log = Module["_Log"] = Module["asm"]["Log"]).apply(null, arguments);
+        };
+        var _LogicalAnd = Module["_LogicalAnd"] = function() {
+          return (_LogicalAnd = Module["_LogicalAnd"] = Module["asm"]["LogicalAnd"]).apply(null, arguments);
+        };
+        var _LogicalNot = Module["_LogicalNot"] = function() {
+          return (_LogicalNot = Module["_LogicalNot"] = Module["asm"]["LogicalNot"]).apply(null, arguments);
+        };
+        var _LogicalOr = Module["_LogicalOr"] = function() {
+          return (_LogicalOr = Module["_LogicalOr"] = Module["asm"]["LogicalOr"]).apply(null, arguments);
+        };
+        var _LogicalXor = Module["_LogicalXor"] = function() {
+          return (_LogicalXor = Module["_LogicalXor"] = Module["asm"]["LogicalXor"]).apply(null, arguments);
+        };
+        var _Max = Module["_Max"] = function() {
+          return (_Max = Module["_Max"] = Module["asm"]["Max"]).apply(null, arguments);
+        };
+        var _MaxPool = Module["_MaxPool"] = function() {
+          return (_MaxPool = Module["_MaxPool"] = Module["asm"]["MaxPool"]).apply(null, arguments);
+        };
+        var _Maximum = Module["_Maximum"] = function() {
+          return (_Maximum = Module["_Maximum"] = Module["asm"]["Maximum"]).apply(null, arguments);
+        };
+        var _Mean = Module["_Mean"] = function() {
+          return (_Mean = Module["_Mean"] = Module["asm"]["Mean"]).apply(null, arguments);
+        };
+        var _Min = Module["_Min"] = function() {
+          return (_Min = Module["_Min"] = Module["asm"]["Min"]).apply(null, arguments);
+        };
+        var _Minimum = Module["_Minimum"] = function() {
+          return (_Minimum = Module["_Minimum"] = Module["asm"]["Minimum"]).apply(null, arguments);
+        };
+        var _MirrorPad = Module["_MirrorPad"] = function() {
+          return (_MirrorPad = Module["_MirrorPad"] = Module["asm"]["MirrorPad"]).apply(null, arguments);
+        };
+        var _Multiply = Module["_Multiply"] = function() {
+          return (_Multiply = Module["_Multiply"] = Module["asm"]["Multiply"]).apply(null, arguments);
+        };
+        var _Neg = Module["_Neg"] = function() {
+          return (_Neg = Module["_Neg"] = Module["asm"]["Neg"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = function() {
+          return (_NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = Module["asm"]["NonMaxSuppressionV3"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = function() {
+          return (_NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = Module["asm"]["NonMaxSuppressionV4"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = function() {
+          return (_NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = Module["asm"]["NonMaxSuppressionV5"]).apply(null, arguments);
+        };
+        var _NotEqual = Module["_NotEqual"] = function() {
+          return (_NotEqual = Module["_NotEqual"] = Module["asm"]["NotEqual"]).apply(null, arguments);
+        };
+        var _OneHot = Module["_OneHot"] = function() {
+          return (_OneHot = Module["_OneHot"] = Module["asm"]["OneHot"]).apply(null, arguments);
+        };
+        var _PadV2 = Module["_PadV2"] = function() {
+          return (_PadV2 = Module["_PadV2"] = Module["asm"]["PadV2"]).apply(null, arguments);
+        };
+        var _Pow = Module["_Pow"] = function() {
+          return (_Pow = Module["_Pow"] = Module["asm"]["Pow"]).apply(null, arguments);
+        };
+        var _Prelu = Module["_Prelu"] = function() {
+          return (_Prelu = Module["_Prelu"] = Module["asm"]["Prelu"]).apply(null, arguments);
+        };
+        var _Prod = Module["_Prod"] = function() {
+          return (_Prod = Module["_Prod"] = Module["asm"]["Prod"]).apply(null, arguments);
+        };
+        var _RealDiv = Module["_RealDiv"] = function() {
+          return (_RealDiv = Module["_RealDiv"] = Module["asm"]["RealDiv"]).apply(null, arguments);
+        };
+        var _Relu = Module["_Relu"] = function() {
+          return (_Relu = Module["_Relu"] = Module["asm"]["Relu"]).apply(null, arguments);
+        };
+        var _Relu6 = Module["_Relu6"] = function() {
+          return (_Relu6 = Module["_Relu6"] = Module["asm"]["Relu6"]).apply(null, arguments);
+        };
+        var _ResizeBilinear = Module["_ResizeBilinear"] = function() {
+          return (_ResizeBilinear = Module["_ResizeBilinear"] = Module["asm"]["ResizeBilinear"]).apply(null, arguments);
+        };
+        var _ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = function() {
+          return (_ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = Module["asm"]["ResizeNearestNeighbor"]).apply(null, arguments);
+        };
+        var _Reverse = Module["_Reverse"] = function() {
+          return (_Reverse = Module["_Reverse"] = Module["asm"]["Reverse"]).apply(null, arguments);
+        };
+        var _RotateWithOffset = Module["_RotateWithOffset"] = function() {
+          return (_RotateWithOffset = Module["_RotateWithOffset"] = Module["asm"]["RotateWithOffset"]).apply(null, arguments);
+        };
+        var _Round = Module["_Round"] = function() {
+          return (_Round = Module["_Round"] = Module["asm"]["Round"]).apply(null, arguments);
+        };
+        var _Rsqrt = Module["_Rsqrt"] = function() {
+          return (_Rsqrt = Module["_Rsqrt"] = Module["asm"]["Rsqrt"]).apply(null, arguments);
+        };
+        var _ScatterNd = Module["_ScatterNd"] = function() {
+          return (_ScatterNd = Module["_ScatterNd"] = Module["asm"]["ScatterNd"]).apply(null, arguments);
+        };
+        var _SelectV2 = Module["_SelectV2"] = function() {
+          return (_SelectV2 = Module["_SelectV2"] = Module["asm"]["SelectV2"]).apply(null, arguments);
+        };
+        var _Sigmoid = Module["_Sigmoid"] = function() {
+          return (_Sigmoid = Module["_Sigmoid"] = Module["asm"]["Sigmoid"]).apply(null, arguments);
+        };
+        var _Sin = Module["_Sin"] = function() {
+          return (_Sin = Module["_Sin"] = Module["asm"]["Sin"]).apply(null, arguments);
+        };
+        var _Softmax = Module["_Softmax"] = function() {
+          return (_Softmax = Module["_Softmax"] = Module["asm"]["Softmax"]).apply(null, arguments);
+        };
+        var _SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = function() {
+          return (_SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = Module["asm"]["SparseFillEmptyRows"]).apply(null, arguments);
+        };
+        var _SparseReshape = Module["_SparseReshape"] = function() {
+          return (_SparseReshape = Module["_SparseReshape"] = Module["asm"]["SparseReshape"]).apply(null, arguments);
+        };
+        var _SparseSegmentReduction = Module["_SparseSegmentReduction"] = function() {
+          return (_SparseSegmentReduction = Module["_SparseSegmentReduction"] = Module["asm"]["SparseSegmentReduction"]).apply(null, arguments);
+        };
+        var _Sqrt = Module["_Sqrt"] = function() {
+          return (_Sqrt = Module["_Sqrt"] = Module["asm"]["Sqrt"]).apply(null, arguments);
+        };
+        var _Square = Module["_Square"] = function() {
+          return (_Square = Module["_Square"] = Module["asm"]["Square"]).apply(null, arguments);
+        };
+        var _SquaredDifference = Module["_SquaredDifference"] = function() {
+          return (_SquaredDifference = Module["_SquaredDifference"] = Module["asm"]["SquaredDifference"]).apply(null, arguments);
+        };
+        var _Step = Module["_Step"] = function() {
+          return (_Step = Module["_Step"] = Module["asm"]["Step"]).apply(null, arguments);
+        };
+        var _StridedSlice = Module["_StridedSlice"] = function() {
+          return (_StridedSlice = Module["_StridedSlice"] = Module["asm"]["StridedSlice"]).apply(null, arguments);
+        };
+        var _Sub = Module["_Sub"] = function() {
+          return (_Sub = Module["_Sub"] = Module["asm"]["Sub"]).apply(null, arguments);
+        };
+        var _Sum = Module["_Sum"] = function() {
+          return (_Sum = Module["_Sum"] = Module["asm"]["Sum"]).apply(null, arguments);
+        };
+        var _Tan = Module["_Tan"] = function() {
+          return (_Tan = Module["_Tan"] = Module["asm"]["Tan"]).apply(null, arguments);
+        };
+        var _Tanh = Module["_Tanh"] = function() {
+          return (_Tanh = Module["_Tanh"] = Module["asm"]["Tanh"]).apply(null, arguments);
+        };
+        var _Tile = Module["_Tile"] = function() {
+          return (_Tile = Module["_Tile"] = Module["asm"]["Tile"]).apply(null, arguments);
+        };
+        var _TopK = Module["_TopK"] = function() {
+          return (_TopK = Module["_TopK"] = Module["asm"]["TopK"]).apply(null, arguments);
+        };
+        var _Transform = Module["_Transform"] = function() {
+          return (_Transform = Module["_Transform"] = Module["asm"]["Transform"]).apply(null, arguments);
+        };
+        var _Transpose = Module["_Transpose"] = function() {
+          return (_Transpose = Module["_Transpose"] = Module["asm"]["Transpose"]).apply(null, arguments);
+        };
+        var __FusedMatMul = Module["__FusedMatMul"] = function() {
+          return (__FusedMatMul = Module["__FusedMatMul"] = Module["asm"]["_FusedMatMul"]).apply(null, arguments);
+        };
+        var _malloc = Module["_malloc"] = function() {
+          return (_malloc = Module["_malloc"] = Module["asm"]["malloc"]).apply(null, arguments);
+        };
+        var _free = Module["_free"] = function() {
+          return (_free = Module["_free"] = Module["asm"]["free"]).apply(null, arguments);
+        };
+        var __emscripten_tls_init = Module["__emscripten_tls_init"] = function() {
+          return (__emscripten_tls_init = Module["__emscripten_tls_init"] = Module["asm"]["_emscripten_tls_init"]).apply(null, arguments);
+        };
+        var _pthread_self = Module["_pthread_self"] = function() {
+          return (_pthread_self = Module["_pthread_self"] = Module["asm"]["pthread_self"]).apply(null, arguments);
+        };
+        var ___errno_location = Module["___errno_location"] = function() {
+          return (___errno_location = Module["___errno_location"] = Module["asm"]["__errno_location"]).apply(null, arguments);
+        };
+        var __emscripten_thread_init = Module["__emscripten_thread_init"] = function() {
+          return (__emscripten_thread_init = Module["__emscripten_thread_init"] = Module["asm"]["_emscripten_thread_init"]).apply(null, arguments);
+        };
+        var __emscripten_thread_crashed = Module["__emscripten_thread_crashed"] = function() {
+          return (__emscripten_thread_crashed = Module["__emscripten_thread_crashed"] = Module["asm"]["_emscripten_thread_crashed"]).apply(null, arguments);
+        };
+        var _emscripten_main_thread_process_queued_calls = Module["_emscripten_main_thread_process_queued_calls"] = function() {
+          return (_emscripten_main_thread_process_queued_calls = Module["_emscripten_main_thread_process_queued_calls"] = Module["asm"]["emscripten_main_thread_process_queued_calls"]).apply(null, arguments);
+        };
+        var _emscripten_main_browser_thread_id = Module["_emscripten_main_browser_thread_id"] = function() {
+          return (_emscripten_main_browser_thread_id = Module["_emscripten_main_browser_thread_id"] = Module["asm"]["emscripten_main_browser_thread_id"]).apply(null, arguments);
+        };
+        var _emscripten_run_in_main_runtime_thread_js = Module["_emscripten_run_in_main_runtime_thread_js"] = function() {
+          return (_emscripten_run_in_main_runtime_thread_js = Module["_emscripten_run_in_main_runtime_thread_js"] = Module["asm"]["emscripten_run_in_main_runtime_thread_js"]).apply(null, arguments);
+        };
+        var _emscripten_dispatch_to_thread_ = Module["_emscripten_dispatch_to_thread_"] = function() {
+          return (_emscripten_dispatch_to_thread_ = Module["_emscripten_dispatch_to_thread_"] = Module["asm"]["emscripten_dispatch_to_thread_"]).apply(null, arguments);
+        };
+        var __emscripten_proxy_execute_task_queue = Module["__emscripten_proxy_execute_task_queue"] = function() {
+          return (__emscripten_proxy_execute_task_queue = Module["__emscripten_proxy_execute_task_queue"] = Module["asm"]["_emscripten_proxy_execute_task_queue"]).apply(null, arguments);
+        };
+        var __emscripten_thread_free_data = Module["__emscripten_thread_free_data"] = function() {
+          return (__emscripten_thread_free_data = Module["__emscripten_thread_free_data"] = Module["asm"]["_emscripten_thread_free_data"]).apply(null, arguments);
+        };
+        var __emscripten_thread_exit = Module["__emscripten_thread_exit"] = function() {
+          return (__emscripten_thread_exit = Module["__emscripten_thread_exit"] = Module["asm"]["_emscripten_thread_exit"]).apply(null, arguments);
+        };
+        var _emscripten_stack_set_limits = Module["_emscripten_stack_set_limits"] = function() {
+          return (_emscripten_stack_set_limits = Module["_emscripten_stack_set_limits"] = Module["asm"]["emscripten_stack_set_limits"]).apply(null, arguments);
+        };
+        var stackSave = Module["stackSave"] = function() {
+          return (stackSave = Module["stackSave"] = Module["asm"]["stackSave"]).apply(null, arguments);
+        };
+        var stackRestore = Module["stackRestore"] = function() {
+          return (stackRestore = Module["stackRestore"] = Module["asm"]["stackRestore"]).apply(null, arguments);
+        };
+        var stackAlloc = Module["stackAlloc"] = function() {
+          return (stackAlloc = Module["stackAlloc"] = Module["asm"]["stackAlloc"]).apply(null, arguments);
+        };
+        var dynCall_iijjiiii = Module["dynCall_iijjiiii"] = function() {
+          return (dynCall_iijjiiii = Module["dynCall_iijjiiii"] = Module["asm"]["dynCall_iijjiiii"]).apply(null, arguments);
+        };
+        var dynCall_jiji = Module["dynCall_jiji"] = function() {
+          return (dynCall_jiji = Module["dynCall_jiji"] = Module["asm"]["dynCall_jiji"]).apply(null, arguments);
+        };
+        Module["keepRuntimeAlive"] = keepRuntimeAlive;
+        Module["wasmMemory"] = wasmMemory;
+        Module["cwrap"] = cwrap;
+        Module["ExitStatus"] = ExitStatus;
+        Module["PThread"] = PThread;
+        var calledRun;
+        dependenciesFulfilled = function runCaller() {
+          if (!calledRun)
+            run();
+          if (!calledRun)
+            dependenciesFulfilled = runCaller;
+        };
+        function run(args) {
+          args = args || arguments_;
+          if (runDependencies > 0) {
+            return;
+          }
+          if (ENVIRONMENT_IS_PTHREAD) {
+            readyPromiseResolve(Module);
+            initRuntime();
+            postMessage({ "cmd": "loaded" });
+            return;
+          }
+          preRun();
+          if (runDependencies > 0) {
+            return;
+          }
+          function doRun() {
+            if (calledRun)
+              return;
+            calledRun = true;
+            Module["calledRun"] = true;
+            if (ABORT)
+              return;
+            initRuntime();
+            readyPromiseResolve(Module);
+            if (Module["onRuntimeInitialized"])
+              Module["onRuntimeInitialized"]();
+            postRun();
+          }
+          if (Module["setStatus"]) {
+            Module["setStatus"]("Running...");
+            setTimeout(function() {
+              setTimeout(function() {
+                Module["setStatus"]("");
+              }, 1);
+              doRun();
+            }, 1);
+          } else {
+            doRun();
+          }
+        }
+        if (Module["preInit"]) {
+          if (typeof Module["preInit"] == "function")
+            Module["preInit"] = [Module["preInit"]];
+          while (Module["preInit"].length > 0) {
+            Module["preInit"].pop()();
+          }
+        }
+        run();
+        var listenersAdded;
+        if (beforeListeners) {
+          listenersAdded = { uncaughtException: process.listeners("uncaughtException").filter(function(listener) {
+            return !beforeListeners.uncaughtException.indexOf(listener) > -1;
+          }), unhandledRejection: process.listeners("unhandledRejection").filter(function(listener) {
+            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;
+          }) };
+        }
+        var actualModule;
+        if (typeof WasmBackendModule !== "undefined") {
+          actualModule = WasmBackendModule;
+        } else if (typeof WasmBackendModuleThreadedSimd3 !== "undefined") {
+          actualModule = WasmBackendModuleThreadedSimd3;
+        } else {
+          throw new Error("Could not find wasm module in post.js");
+        }
+        if (listenersAdded) {
+          var tmpDispose = actualModule["_dispose"];
+          actualModule["_dispose"] = function() {
+            tmpDispose();
+            listenersAdded.uncaughtException.forEach(function(listener) {
+              process.removeListener("uncaughtException", listener);
+            });
+            listenersAdded.unhandledRejection.forEach(function(listener) {
+              process.removeListener("unhandledRejection", listener);
+            });
+          };
+        }
+        return WasmBackendModuleThreadedSimd3.ready;
+      };
+    })();
+    if (typeof exports === "object" && typeof module === "object")
+      module.exports = WasmBackendModuleThreadedSimd2;
+    else if (typeof define === "function" && define["amd"])
+      define([], function() {
+        return WasmBackendModuleThreadedSimd2;
+      });
+    else if (typeof exports === "object")
+      exports["WasmBackendModuleThreadedSimd"] = WasmBackendModuleThreadedSimd2;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js
+var require_tfjs_backend_wasm_threaded_simd_worker = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.worker.js"(exports, module) {
+    module.exports.wasmWorkerContents = `"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
+");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};`;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js
+var require_tfjs_backend_wasm = __commonJS({
+  "node_modules/@tensorflow/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(exports, module) {
+    var WasmBackendModule2 = (() => {
+      var _scriptDir = typeof document !== "undefined" && document.currentScript ? document.currentScript.src : void 0;
+      if (typeof __filename !== "undefined")
+        _scriptDir = _scriptDir || __filename;
+      return function(WasmBackendModule3) {
+        WasmBackendModule3 = WasmBackendModule3 || {};
+        var Module = typeof WasmBackendModule3 != "undefined" ? WasmBackendModule3 : {};
+        var readyPromiseResolve, readyPromiseReject;
+        Module["ready"] = new Promise(function(resolve, reject) {
+          readyPromiseResolve = resolve;
+          readyPromiseReject = reject;
+        });
+        var beforeListeners;
+        if (typeof process !== "undefined" && process.listeners) {
+          beforeListeners = { uncaughtException: process.listeners("uncaughtException"), unhandledRejection: process.listeners("unhandledRejection") };
+        }
+        var moduleOverrides = Object.assign({}, Module);
+        var arguments_ = [];
+        var thisProgram = "./this.program";
+        var quit_ = (status, toThrow) => {
+          throw toThrow;
+        };
+        var ENVIRONMENT_IS_WEB = typeof window == "object";
+        var ENVIRONMENT_IS_WORKER = typeof importScripts == "function";
+        var ENVIRONMENT_IS_NODE = typeof process == "object" && typeof process.versions == "object" && typeof process.versions.node == "string";
+        var scriptDirectory = "";
+        function locateFile(path) {
+          if (Module["locateFile"]) {
+            return Module["locateFile"](path, scriptDirectory);
+          }
+          return scriptDirectory + path;
+        }
+        var read_, readAsync, readBinary, setWindowTitle;
+        function logExceptionOnExit(e) {
+          if (e instanceof ExitStatus)
+            return;
+          let toLog = e;
+          err("exiting due to exception: " + toLog);
+        }
+        if (ENVIRONMENT_IS_NODE) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = require_path().dirname(scriptDirectory) + "/";
+          } else {
+            scriptDirectory = __dirname + "/";
+          }
+          var fs, nodePath;
+          if (typeof __require === "function") {
+            fs = require_fs();
+            nodePath = require_path();
+          }
+          read_ = (filename, binary) => {
+            filename = nodePath["normalize"](filename);
+            return fs.readFileSync(filename, binary ? void 0 : "utf8");
+          };
+          readBinary = (filename) => {
+            var ret = read_(filename, true);
+            if (!ret.buffer) {
+              ret = new Uint8Array(ret);
+            }
+            return ret;
+          };
+          readAsync = (filename, onload, onerror) => {
+            filename = nodePath["normalize"](filename);
+            fs.readFile(filename, function(err2, data) {
+              if (err2)
+                onerror(err2);
+              else
+                onload(data.buffer);
+            });
+          };
+          if (process["argv"].length > 1) {
+            thisProgram = process["argv"][1].replace(/\\/g, "/");
+          }
+          arguments_ = process["argv"].slice(2);
+          process["on"]("uncaughtException", function(ex) {
+            if (!(ex instanceof ExitStatus)) {
+              throw ex;
+            }
+          });
+          process["on"]("unhandledRejection", function(reason) {
+            throw reason;
+          });
+          quit_ = (status, toThrow) => {
+            if (keepRuntimeAlive()) {
+              process["exitCode"] = status;
+              throw toThrow;
+            }
+            logExceptionOnExit(toThrow);
+            process["exit"](status);
+          };
+          Module["inspect"] = function() {
+            return "[Emscripten Module object]";
+          };
+        } else if (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER) {
+          if (ENVIRONMENT_IS_WORKER) {
+            scriptDirectory = self.location.href;
+          } else if (typeof document != "undefined" && document.currentScript) {
+            scriptDirectory = document.currentScript.src;
+          }
+          if (_scriptDir) {
+            scriptDirectory = _scriptDir;
+          }
+          if (scriptDirectory.indexOf("blob:") !== 0) {
+            scriptDirectory = scriptDirectory.substr(0, scriptDirectory.replace(/[?#].*/, "").lastIndexOf("/") + 1);
+          } else {
+            scriptDirectory = "";
+          }
+          {
+            read_ = (url) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, false);
+              xhr.send(null);
+              return xhr.responseText;
+            };
+            if (ENVIRONMENT_IS_WORKER) {
+              readBinary = (url) => {
+                var xhr = new XMLHttpRequest();
+                xhr.open("GET", url, false);
+                xhr.responseType = "arraybuffer";
+                xhr.send(null);
+                return new Uint8Array(xhr.response);
+              };
+            }
+            readAsync = (url, onload, onerror) => {
+              var xhr = new XMLHttpRequest();
+              xhr.open("GET", url, true);
+              xhr.responseType = "arraybuffer";
+              xhr.onload = () => {
+                if (xhr.status == 200 || xhr.status == 0 && xhr.response) {
+                  onload(xhr.response);
+                  return;
+                }
+                onerror();
+              };
+              xhr.onerror = onerror;
+              xhr.send(null);
+            };
+          }
+          setWindowTitle = (title) => document.title = title;
+        } else {
+        }
+        var out = Module["print"] || console.log.bind(console);
+        var err = Module["printErr"] || console.warn.bind(console);
+        Object.assign(Module, moduleOverrides);
+        moduleOverrides = null;
+        if (Module["arguments"])
+          arguments_ = Module["arguments"];
+        if (Module["thisProgram"])
+          thisProgram = Module["thisProgram"];
+        if (Module["quit"])
+          quit_ = Module["quit"];
+        var POINTER_SIZE = 4;
+        var wasmBinary;
+        if (Module["wasmBinary"])
+          wasmBinary = Module["wasmBinary"];
+        var noExitRuntime = Module["noExitRuntime"] || true;
+        if (typeof WebAssembly != "object") {
+          abort("no native wasm support detected");
+        }
+        var wasmMemory;
+        var ABORT = false;
+        var EXITSTATUS;
+        function assert3(condition, text) {
+          if (!condition) {
+            abort(text);
+          }
+        }
+        var UTF8Decoder = typeof TextDecoder != "undefined" ? new TextDecoder("utf8") : void 0;
+        function UTF8ArrayToString(heapOrArray, idx, maxBytesToRead) {
+          var endIdx = idx + maxBytesToRead;
+          var endPtr = idx;
+          while (heapOrArray[endPtr] && !(endPtr >= endIdx))
+            ++endPtr;
+          if (endPtr - idx > 16 && heapOrArray.buffer && UTF8Decoder) {
+            return UTF8Decoder.decode(heapOrArray.subarray(idx, endPtr));
+          }
+          var str = "";
+          while (idx < endPtr) {
+            var u0 = heapOrArray[idx++];
+            if (!(u0 & 128)) {
+              str += String.fromCharCode(u0);
+              continue;
+            }
+            var u1 = heapOrArray[idx++] & 63;
+            if ((u0 & 224) == 192) {
+              str += String.fromCharCode((u0 & 31) << 6 | u1);
+              continue;
+            }
+            var u2 = heapOrArray[idx++] & 63;
+            if ((u0 & 240) == 224) {
+              u0 = (u0 & 15) << 12 | u1 << 6 | u2;
+            } else {
+              u0 = (u0 & 7) << 18 | u1 << 12 | u2 << 6 | heapOrArray[idx++] & 63;
+            }
+            if (u0 < 65536) {
+              str += String.fromCharCode(u0);
+            } else {
+              var ch = u0 - 65536;
+              str += String.fromCharCode(55296 | ch >> 10, 56320 | ch & 1023);
+            }
+          }
+          return str;
+        }
+        function UTF8ToString(ptr, maxBytesToRead) {
+          return ptr ? UTF8ArrayToString(HEAPU8, ptr, maxBytesToRead) : "";
+        }
+        function stringToUTF8Array(str, heap, outIdx, maxBytesToWrite) {
+          if (!(maxBytesToWrite > 0))
+            return 0;
+          var startIdx = outIdx;
+          var endIdx = outIdx + maxBytesToWrite - 1;
+          for (var i = 0; i < str.length; ++i) {
+            var u = str.charCodeAt(i);
+            if (u >= 55296 && u <= 57343) {
+              var u1 = str.charCodeAt(++i);
+              u = 65536 + ((u & 1023) << 10) | u1 & 1023;
+            }
+            if (u <= 127) {
+              if (outIdx >= endIdx)
+                break;
+              heap[outIdx++] = u;
+            } else if (u <= 2047) {
+              if (outIdx + 1 >= endIdx)
+                break;
+              heap[outIdx++] = 192 | u >> 6;
+              heap[outIdx++] = 128 | u & 63;
+            } else if (u <= 65535) {
+              if (outIdx + 2 >= endIdx)
+                break;
+              heap[outIdx++] = 224 | u >> 12;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            } else {
+              if (outIdx + 3 >= endIdx)
+                break;
+              heap[outIdx++] = 240 | u >> 18;
+              heap[outIdx++] = 128 | u >> 12 & 63;
+              heap[outIdx++] = 128 | u >> 6 & 63;
+              heap[outIdx++] = 128 | u & 63;
+            }
+          }
+          heap[outIdx] = 0;
+          return outIdx - startIdx;
+        }
+        function stringToUTF8(str, outPtr, maxBytesToWrite) {
+          return stringToUTF8Array(str, HEAPU8, outPtr, maxBytesToWrite);
+        }
+        var buffer2, HEAP8, HEAPU8, HEAP16, HEAPU16, HEAP32, HEAPU32, HEAPF32, HEAPF64;
+        function updateGlobalBufferAndViews(buf) {
+          buffer2 = buf;
+          Module["HEAP8"] = HEAP8 = new Int8Array(buf);
+          Module["HEAP16"] = HEAP16 = new Int16Array(buf);
+          Module["HEAP32"] = HEAP32 = new Int32Array(buf);
+          Module["HEAPU8"] = HEAPU8 = new Uint8Array(buf);
+          Module["HEAPU16"] = HEAPU16 = new Uint16Array(buf);
+          Module["HEAPU32"] = HEAPU32 = new Uint32Array(buf);
+          Module["HEAPF32"] = HEAPF32 = new Float32Array(buf);
+          Module["HEAPF64"] = HEAPF64 = new Float64Array(buf);
+        }
+        var INITIAL_MEMORY = Module["INITIAL_MEMORY"] || 16777216;
+        var wasmTable;
+        var __ATPRERUN__ = [];
+        var __ATINIT__ = [];
+        var __ATPOSTRUN__ = [];
+        var runtimeInitialized = false;
+        function keepRuntimeAlive() {
+          return noExitRuntime;
+        }
+        function preRun() {
+          if (Module["preRun"]) {
+            if (typeof Module["preRun"] == "function")
+              Module["preRun"] = [Module["preRun"]];
+            while (Module["preRun"].length) {
+              addOnPreRun(Module["preRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPRERUN__);
+        }
+        function initRuntime() {
+          runtimeInitialized = true;
+          callRuntimeCallbacks(__ATINIT__);
+        }
+        function postRun() {
+          if (Module["postRun"]) {
+            if (typeof Module["postRun"] == "function")
+              Module["postRun"] = [Module["postRun"]];
+            while (Module["postRun"].length) {
+              addOnPostRun(Module["postRun"].shift());
+            }
+          }
+          callRuntimeCallbacks(__ATPOSTRUN__);
+        }
+        function addOnPreRun(cb) {
+          __ATPRERUN__.unshift(cb);
+        }
+        function addOnInit(cb) {
+          __ATINIT__.unshift(cb);
+        }
+        function addOnPostRun(cb) {
+          __ATPOSTRUN__.unshift(cb);
+        }
+        var runDependencies = 0;
+        var runDependencyWatcher = null;
+        var dependenciesFulfilled = null;
+        function addRunDependency(id) {
+          runDependencies++;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+        }
+        function removeRunDependency(id) {
+          runDependencies--;
+          if (Module["monitorRunDependencies"]) {
+            Module["monitorRunDependencies"](runDependencies);
+          }
+          if (runDependencies == 0) {
+            if (runDependencyWatcher !== null) {
+              clearInterval(runDependencyWatcher);
+              runDependencyWatcher = null;
+            }
+            if (dependenciesFulfilled) {
+              var callback = dependenciesFulfilled;
+              dependenciesFulfilled = null;
+              callback();
+            }
+          }
+        }
+        function abort(what) {
+          {
+            if (Module["onAbort"]) {
+              Module["onAbort"](what);
+            }
+          }
+          what = "Aborted(" + what + ")";
+          err(what);
+          ABORT = true;
+          EXITSTATUS = 1;
+          what += ". Build with -sASSERTIONS for more info.";
+          var e = new WebAssembly.RuntimeError(what);
+          readyPromiseReject(e);
+          throw e;
+        }
+        var dataURIPrefix = "data:application/octet-stream;base64,";
+        function isDataURI(filename) {
+          return filename.startsWith(dataURIPrefix);
+        }
+        function isFileURI(filename) {
+          return filename.startsWith("file://");
+        }
+        var wasmBinaryFile;
+        wasmBinaryFile = "tfjs-backend-wasm.wasm";
+        if (!isDataURI(wasmBinaryFile)) {
+          wasmBinaryFile = locateFile(wasmBinaryFile);
+        }
+        function getBinary(file) {
+          try {
+            if (file == wasmBinaryFile && wasmBinary) {
+              return new Uint8Array(wasmBinary);
+            }
+            if (readBinary) {
+              return readBinary(file);
+            }
+            throw "both async and sync fetching of the wasm failed";
+          } catch (err2) {
+            abort(err2);
+          }
+        }
+        function getBinaryPromise() {
+          if (!wasmBinary && (ENVIRONMENT_IS_WEB || ENVIRONMENT_IS_WORKER)) {
+            if (typeof fetch == "function" && !isFileURI(wasmBinaryFile)) {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                if (!response["ok"]) {
+                  throw "failed to load wasm binary file at '" + wasmBinaryFile + "'";
+                }
+                return response["arrayBuffer"]();
+              }).catch(function() {
+                return getBinary(wasmBinaryFile);
+              });
+            } else {
+              if (readAsync) {
+                return new Promise(function(resolve, reject) {
+                  readAsync(wasmBinaryFile, function(response) {
+                    resolve(new Uint8Array(response));
+                  }, reject);
+                });
+              }
+            }
+          }
+          return Promise.resolve().then(function() {
+            return getBinary(wasmBinaryFile);
+          });
+        }
+        function createWasm() {
+          var info = { "env": asmLibraryArg, "wasi_snapshot_preview1": asmLibraryArg };
+          function receiveInstance(instance, module2) {
+            var exports3 = instance.exports;
+            Module["asm"] = exports3;
+            wasmMemory = Module["asm"]["memory"];
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            wasmTable = Module["asm"]["__indirect_function_table"];
+            addOnInit(Module["asm"]["__wasm_call_ctors"]);
+            removeRunDependency("wasm-instantiate");
+          }
+          addRunDependency("wasm-instantiate");
+          function receiveInstantiationResult(result) {
+            receiveInstance(result["instance"]);
+          }
+          function instantiateArrayBuffer(receiver) {
+            return getBinaryPromise().then(function(binary) {
+              return WebAssembly.instantiate(binary, info);
+            }).then(function(instance) {
+              return instance;
+            }).then(receiver, function(reason) {
+              err("failed to asynchronously prepare wasm: " + reason);
+              abort(reason);
+            });
+          }
+          function instantiateAsync() {
+            if (!wasmBinary && typeof WebAssembly.instantiateStreaming == "function" && !isDataURI(wasmBinaryFile) && !isFileURI(wasmBinaryFile) && !ENVIRONMENT_IS_NODE && typeof fetch == "function") {
+              return fetch(wasmBinaryFile, { credentials: "same-origin" }).then(function(response) {
+                var result = WebAssembly.instantiateStreaming(response, info);
+                return result.then(receiveInstantiationResult, function(reason) {
+                  err("wasm streaming compile failed: " + reason);
+                  err("falling back to ArrayBuffer instantiation");
+                  return instantiateArrayBuffer(receiveInstantiationResult);
+                });
+              });
+            } else {
+              return instantiateArrayBuffer(receiveInstantiationResult);
+            }
+          }
+          if (Module["instantiateWasm"]) {
+            try {
+              var exports2 = Module["instantiateWasm"](info, receiveInstance);
+              return exports2;
+            } catch (e) {
+              err("Module.instantiateWasm callback failed with error: " + e);
+              readyPromiseReject(e);
+            }
+          }
+          instantiateAsync().catch(readyPromiseReject);
+          return {};
+        }
+        var tempDouble;
+        var tempI64;
+        function ExitStatus(status) {
+          this.name = "ExitStatus";
+          this.message = "Program terminated with exit(" + status + ")";
+          this.status = status;
+        }
+        function callRuntimeCallbacks(callbacks2) {
+          while (callbacks2.length > 0) {
+            callbacks2.shift()(Module);
+          }
+        }
+        function demangle(func2) {
+          return func2;
+        }
+        function demangleAll(text) {
+          var regex = /\b_Z[\w\d_]+/g;
+          return text.replace(regex, function(x) {
+            var y = demangle(x);
+            return x === y ? x : y + " [" + x + "]";
+          });
+        }
+        function jsStackTrace() {
+          var error = new Error();
+          if (!error.stack) {
+            try {
+              throw new Error();
+            } catch (e) {
+              error = e;
+            }
+            if (!error.stack) {
+              return "(no stack trace available)";
+            }
+          }
+          return error.stack.toString();
+        }
+        function writeArrayToMemory(array2, buffer3) {
+          HEAP8.set(array2, buffer3);
+        }
+        function _abort() {
+          abort("");
+        }
+        function getHeapMax() {
+          return 2147483648;
+        }
+        function _emscripten_get_heap_max() {
+          return getHeapMax();
+        }
+        function _emscripten_memcpy_big(dest, src, num) {
+          HEAPU8.copyWithin(dest, src, src + num);
+        }
+        function emscripten_realloc_buffer(size) {
+          try {
+            wasmMemory.grow(size - buffer2.byteLength + 65535 >>> 16);
+            updateGlobalBufferAndViews(wasmMemory.buffer);
+            return 1;
+          } catch (e) {
+          }
+        }
+        function _emscripten_resize_heap(requestedSize) {
+          var oldSize = HEAPU8.length;
+          requestedSize = requestedSize >>> 0;
+          var maxHeapSize = getHeapMax();
+          if (requestedSize > maxHeapSize) {
+            return false;
+          }
+          let alignUp = (x, multiple) => x + (multiple - x % multiple) % multiple;
+          for (var cutDown = 1; cutDown <= 4; cutDown *= 2) {
+            var overGrownHeapSize = oldSize * (1 + 0.2 / cutDown);
+            overGrownHeapSize = Math.min(overGrownHeapSize, requestedSize + 100663296);
+            var newSize = Math.min(maxHeapSize, alignUp(Math.max(requestedSize, overGrownHeapSize), 65536));
+            var replacement = emscripten_realloc_buffer(newSize);
+            if (replacement) {
+              return true;
+            }
+          }
+          return false;
+        }
+        var SYSCALLS = { varargs: void 0, get: function() {
+          SYSCALLS.varargs += 4;
+          var ret = HEAP32[SYSCALLS.varargs - 4 >> 2];
+          return ret;
+        }, getStr: function(ptr) {
+          var ret = UTF8ToString(ptr);
+          return ret;
+        } };
+        function _fd_close(fd) {
+          return 52;
+        }
+        function _fd_seek(fd, offset_low, offset_high, whence, newOffset) {
+          return 70;
+        }
+        var printCharBuffers = [null, [], []];
+        function printChar(stream, curr) {
+          var buffer3 = printCharBuffers[stream];
+          if (curr === 0 || curr === 10) {
+            (stream === 1 ? out : err)(UTF8ArrayToString(buffer3, 0));
+            buffer3.length = 0;
+          } else {
+            buffer3.push(curr);
+          }
+        }
+        function _fd_write(fd, iov, iovcnt, pnum) {
+          var num = 0;
+          for (var i = 0; i < iovcnt; i++) {
+            var ptr = HEAPU32[iov >> 2];
+            var len = HEAPU32[iov + 4 >> 2];
+            iov += 8;
+            for (var j = 0; j < len; j++) {
+              printChar(fd, HEAPU8[ptr + j]);
+            }
+            num += len;
+          }
+          HEAPU32[pnum >> 2] = num;
+          return 0;
+        }
+        function getCFunc(ident) {
+          var func2 = Module["_" + ident];
+          return func2;
+        }
+        function ccall(ident, returnType, argTypes, args, opts) {
+          var toC = { "string": (str) => {
+            var ret2 = 0;
+            if (str !== null && str !== void 0 && str !== 0) {
+              var len = (str.length << 2) + 1;
+              ret2 = stackAlloc(len);
+              stringToUTF8(str, ret2, len);
+            }
+            return ret2;
+          }, "array": (arr) => {
+            var ret2 = stackAlloc(arr.length);
+            writeArrayToMemory(arr, ret2);
+            return ret2;
+          } };
+          function convertReturnValue(ret2) {
+            if (returnType === "string") {
+              return UTF8ToString(ret2);
+            }
+            if (returnType === "boolean")
+              return Boolean(ret2);
+            return ret2;
+          }
+          var func2 = getCFunc(ident);
+          var cArgs = [];
+          var stack2 = 0;
+          if (args) {
+            for (var i = 0; i < args.length; i++) {
+              var converter = toC[argTypes[i]];
+              if (converter) {
+                if (stack2 === 0)
+                  stack2 = stackSave();
+                cArgs[i] = converter(args[i]);
+              } else {
+                cArgs[i] = args[i];
+              }
+            }
+          }
+          var ret = func2.apply(null, cArgs);
+          function onDone(ret2) {
+            if (stack2 !== 0)
+              stackRestore(stack2);
+            return convertReturnValue(ret2);
+          }
+          ret = onDone(ret);
+          return ret;
+        }
+        function cwrap(ident, returnType, argTypes, opts) {
+          argTypes = argTypes || [];
+          var numericArgs = argTypes.every((type) => type === "number" || type === "boolean");
+          var numericRet = returnType !== "string";
+          if (numericRet && numericArgs && !opts) {
+            return getCFunc(ident);
+          }
+          return function() {
+            return ccall(ident, returnType, argTypes, arguments, opts);
+          };
+        }
+        var asmLibraryArg = { "abort": _abort, "emscripten_get_heap_max": _emscripten_get_heap_max, "emscripten_memcpy_big": _emscripten_memcpy_big, "emscripten_resize_heap": _emscripten_resize_heap, "fd_close": _fd_close, "fd_seek": _fd_seek, "fd_write": _fd_write };
+        var asm = createWasm();
+        var ___wasm_call_ctors = Module["___wasm_call_ctors"] = function() {
+          return (___wasm_call_ctors = Module["___wasm_call_ctors"] = Module["asm"]["__wasm_call_ctors"]).apply(null, arguments);
+        };
+        var _init = Module["_init"] = function() {
+          return (_init = Module["_init"] = Module["asm"]["init"]).apply(null, arguments);
+        };
+        var _init_with_threads_count = Module["_init_with_threads_count"] = function() {
+          return (_init_with_threads_count = Module["_init_with_threads_count"] = Module["asm"]["init_with_threads_count"]).apply(null, arguments);
+        };
+        var _get_threads_count = Module["_get_threads_count"] = function() {
+          return (_get_threads_count = Module["_get_threads_count"] = Module["asm"]["get_threads_count"]).apply(null, arguments);
+        };
+        var _register_tensor = Module["_register_tensor"] = function() {
+          return (_register_tensor = Module["_register_tensor"] = Module["asm"]["register_tensor"]).apply(null, arguments);
+        };
+        var _dispose_data = Module["_dispose_data"] = function() {
+          return (_dispose_data = Module["_dispose_data"] = Module["asm"]["dispose_data"]).apply(null, arguments);
+        };
+        var _dispose = Module["_dispose"] = function() {
+          return (_dispose = Module["_dispose"] = Module["asm"]["dispose"]).apply(null, arguments);
+        };
+        var _Abs = Module["_Abs"] = function() {
+          return (_Abs = Module["_Abs"] = Module["asm"]["Abs"]).apply(null, arguments);
+        };
+        var _Add = Module["_Add"] = function() {
+          return (_Add = Module["_Add"] = Module["asm"]["Add"]).apply(null, arguments);
+        };
+        var _AddN = Module["_AddN"] = function() {
+          return (_AddN = Module["_AddN"] = Module["asm"]["AddN"]).apply(null, arguments);
+        };
+        var _All = Module["_All"] = function() {
+          return (_All = Module["_All"] = Module["asm"]["All"]).apply(null, arguments);
+        };
+        var _Any = Module["_Any"] = function() {
+          return (_Any = Module["_Any"] = Module["asm"]["Any"]).apply(null, arguments);
+        };
+        var _ArgMax = Module["_ArgMax"] = function() {
+          return (_ArgMax = Module["_ArgMax"] = Module["asm"]["ArgMax"]).apply(null, arguments);
+        };
+        var _AvgPool = Module["_AvgPool"] = function() {
+          return (_AvgPool = Module["_AvgPool"] = Module["asm"]["AvgPool"]).apply(null, arguments);
+        };
+        var _BatchMatMul = Module["_BatchMatMul"] = function() {
+          return (_BatchMatMul = Module["_BatchMatMul"] = Module["asm"]["BatchMatMul"]).apply(null, arguments);
+        };
+        var _Ceil = Module["_Ceil"] = function() {
+          return (_Ceil = Module["_Ceil"] = Module["asm"]["Ceil"]).apply(null, arguments);
+        };
+        var _ClipByValue = Module["_ClipByValue"] = function() {
+          return (_ClipByValue = Module["_ClipByValue"] = Module["asm"]["ClipByValue"]).apply(null, arguments);
+        };
+        var _Conv2D = Module["_Conv2D"] = function() {
+          return (_Conv2D = Module["_Conv2D"] = Module["asm"]["Conv2D"]).apply(null, arguments);
+        };
+        var _Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = function() {
+          return (_Conv2DBackpropInput = Module["_Conv2DBackpropInput"] = Module["asm"]["Conv2DBackpropInput"]).apply(null, arguments);
+        };
+        var _Cos = Module["_Cos"] = function() {
+          return (_Cos = Module["_Cos"] = Module["asm"]["Cos"]).apply(null, arguments);
+        };
+        var _Cosh = Module["_Cosh"] = function() {
+          return (_Cosh = Module["_Cosh"] = Module["asm"]["Cosh"]).apply(null, arguments);
+        };
+        var _CropAndResize = Module["_CropAndResize"] = function() {
+          return (_CropAndResize = Module["_CropAndResize"] = Module["asm"]["CropAndResize"]).apply(null, arguments);
+        };
+        var _Cumprod = Module["_Cumprod"] = function() {
+          return (_Cumprod = Module["_Cumprod"] = Module["asm"]["Cumprod"]).apply(null, arguments);
+        };
+        var _Cumsum = Module["_Cumsum"] = function() {
+          return (_Cumsum = Module["_Cumsum"] = Module["asm"]["Cumsum"]).apply(null, arguments);
+        };
+        var _DepthToSpace = Module["_DepthToSpace"] = function() {
+          return (_DepthToSpace = Module["_DepthToSpace"] = Module["asm"]["DepthToSpace"]).apply(null, arguments);
+        };
+        var _DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = function() {
+          return (_DepthwiseConv2dNative = Module["_DepthwiseConv2dNative"] = Module["asm"]["DepthwiseConv2dNative"]).apply(null, arguments);
+        };
+        var _Elu = Module["_Elu"] = function() {
+          return (_Elu = Module["_Elu"] = Module["asm"]["Elu"]).apply(null, arguments);
+        };
+        var _Equal = Module["_Equal"] = function() {
+          return (_Equal = Module["_Equal"] = Module["asm"]["Equal"]).apply(null, arguments);
+        };
+        var _Exp = Module["_Exp"] = function() {
+          return (_Exp = Module["_Exp"] = Module["asm"]["Exp"]).apply(null, arguments);
+        };
+        var _FlipLeftRight = Module["_FlipLeftRight"] = function() {
+          return (_FlipLeftRight = Module["_FlipLeftRight"] = Module["asm"]["FlipLeftRight"]).apply(null, arguments);
+        };
+        var _Floor = Module["_Floor"] = function() {
+          return (_Floor = Module["_Floor"] = Module["asm"]["Floor"]).apply(null, arguments);
+        };
+        var _FloorDiv = Module["_FloorDiv"] = function() {
+          return (_FloorDiv = Module["_FloorDiv"] = Module["asm"]["FloorDiv"]).apply(null, arguments);
+        };
+        var _FusedBatchNorm = Module["_FusedBatchNorm"] = function() {
+          return (_FusedBatchNorm = Module["_FusedBatchNorm"] = Module["asm"]["FusedBatchNorm"]).apply(null, arguments);
+        };
+        var _FusedConv2D = Module["_FusedConv2D"] = function() {
+          return (_FusedConv2D = Module["_FusedConv2D"] = Module["asm"]["FusedConv2D"]).apply(null, arguments);
+        };
+        var _FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = function() {
+          return (_FusedDepthwiseConv2D = Module["_FusedDepthwiseConv2D"] = Module["asm"]["FusedDepthwiseConv2D"]).apply(null, arguments);
+        };
+        var _Gather = Module["_Gather"] = function() {
+          return (_Gather = Module["_Gather"] = Module["asm"]["Gather"]).apply(null, arguments);
+        };
+        var _GatherNd = Module["_GatherNd"] = function() {
+          return (_GatherNd = Module["_GatherNd"] = Module["asm"]["GatherNd"]).apply(null, arguments);
+        };
+        var _Greater = Module["_Greater"] = function() {
+          return (_Greater = Module["_Greater"] = Module["asm"]["Greater"]).apply(null, arguments);
+        };
+        var _GreaterEqual = Module["_GreaterEqual"] = function() {
+          return (_GreaterEqual = Module["_GreaterEqual"] = Module["asm"]["GreaterEqual"]).apply(null, arguments);
+        };
+        var _LeakyRelu = Module["_LeakyRelu"] = function() {
+          return (_LeakyRelu = Module["_LeakyRelu"] = Module["asm"]["LeakyRelu"]).apply(null, arguments);
+        };
+        var _Less = Module["_Less"] = function() {
+          return (_Less = Module["_Less"] = Module["asm"]["Less"]).apply(null, arguments);
+        };
+        var _LessEqual = Module["_LessEqual"] = function() {
+          return (_LessEqual = Module["_LessEqual"] = Module["asm"]["LessEqual"]).apply(null, arguments);
+        };
+        var _Log = Module["_Log"] = function() {
+          return (_Log = Module["_Log"] = Module["asm"]["Log"]).apply(null, arguments);
+        };
+        var _LogicalAnd = Module["_LogicalAnd"] = function() {
+          return (_LogicalAnd = Module["_LogicalAnd"] = Module["asm"]["LogicalAnd"]).apply(null, arguments);
+        };
+        var _LogicalNot = Module["_LogicalNot"] = function() {
+          return (_LogicalNot = Module["_LogicalNot"] = Module["asm"]["LogicalNot"]).apply(null, arguments);
+        };
+        var _LogicalOr = Module["_LogicalOr"] = function() {
+          return (_LogicalOr = Module["_LogicalOr"] = Module["asm"]["LogicalOr"]).apply(null, arguments);
+        };
+        var _LogicalXor = Module["_LogicalXor"] = function() {
+          return (_LogicalXor = Module["_LogicalXor"] = Module["asm"]["LogicalXor"]).apply(null, arguments);
+        };
+        var _Max = Module["_Max"] = function() {
+          return (_Max = Module["_Max"] = Module["asm"]["Max"]).apply(null, arguments);
+        };
+        var _MaxPool = Module["_MaxPool"] = function() {
+          return (_MaxPool = Module["_MaxPool"] = Module["asm"]["MaxPool"]).apply(null, arguments);
+        };
+        var _Maximum = Module["_Maximum"] = function() {
+          return (_Maximum = Module["_Maximum"] = Module["asm"]["Maximum"]).apply(null, arguments);
+        };
+        var _Mean = Module["_Mean"] = function() {
+          return (_Mean = Module["_Mean"] = Module["asm"]["Mean"]).apply(null, arguments);
+        };
+        var _Min = Module["_Min"] = function() {
+          return (_Min = Module["_Min"] = Module["asm"]["Min"]).apply(null, arguments);
+        };
+        var _Minimum = Module["_Minimum"] = function() {
+          return (_Minimum = Module["_Minimum"] = Module["asm"]["Minimum"]).apply(null, arguments);
+        };
+        var _MirrorPad = Module["_MirrorPad"] = function() {
+          return (_MirrorPad = Module["_MirrorPad"] = Module["asm"]["MirrorPad"]).apply(null, arguments);
+        };
+        var _Multiply = Module["_Multiply"] = function() {
+          return (_Multiply = Module["_Multiply"] = Module["asm"]["Multiply"]).apply(null, arguments);
+        };
+        var _Neg = Module["_Neg"] = function() {
+          return (_Neg = Module["_Neg"] = Module["asm"]["Neg"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = function() {
+          return (_NonMaxSuppressionV3 = Module["_NonMaxSuppressionV3"] = Module["asm"]["NonMaxSuppressionV3"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = function() {
+          return (_NonMaxSuppressionV4 = Module["_NonMaxSuppressionV4"] = Module["asm"]["NonMaxSuppressionV4"]).apply(null, arguments);
+        };
+        var _NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = function() {
+          return (_NonMaxSuppressionV5 = Module["_NonMaxSuppressionV5"] = Module["asm"]["NonMaxSuppressionV5"]).apply(null, arguments);
+        };
+        var _NotEqual = Module["_NotEqual"] = function() {
+          return (_NotEqual = Module["_NotEqual"] = Module["asm"]["NotEqual"]).apply(null, arguments);
+        };
+        var _OneHot = Module["_OneHot"] = function() {
+          return (_OneHot = Module["_OneHot"] = Module["asm"]["OneHot"]).apply(null, arguments);
+        };
+        var _PadV2 = Module["_PadV2"] = function() {
+          return (_PadV2 = Module["_PadV2"] = Module["asm"]["PadV2"]).apply(null, arguments);
+        };
+        var _Pow = Module["_Pow"] = function() {
+          return (_Pow = Module["_Pow"] = Module["asm"]["Pow"]).apply(null, arguments);
+        };
+        var _Prelu = Module["_Prelu"] = function() {
+          return (_Prelu = Module["_Prelu"] = Module["asm"]["Prelu"]).apply(null, arguments);
+        };
+        var _Prod = Module["_Prod"] = function() {
+          return (_Prod = Module["_Prod"] = Module["asm"]["Prod"]).apply(null, arguments);
+        };
+        var _RealDiv = Module["_RealDiv"] = function() {
+          return (_RealDiv = Module["_RealDiv"] = Module["asm"]["RealDiv"]).apply(null, arguments);
+        };
+        var _Relu = Module["_Relu"] = function() {
+          return (_Relu = Module["_Relu"] = Module["asm"]["Relu"]).apply(null, arguments);
+        };
+        var _Relu6 = Module["_Relu6"] = function() {
+          return (_Relu6 = Module["_Relu6"] = Module["asm"]["Relu6"]).apply(null, arguments);
+        };
+        var _ResizeBilinear = Module["_ResizeBilinear"] = function() {
+          return (_ResizeBilinear = Module["_ResizeBilinear"] = Module["asm"]["ResizeBilinear"]).apply(null, arguments);
+        };
+        var _ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = function() {
+          return (_ResizeNearestNeighbor = Module["_ResizeNearestNeighbor"] = Module["asm"]["ResizeNearestNeighbor"]).apply(null, arguments);
+        };
+        var _Reverse = Module["_Reverse"] = function() {
+          return (_Reverse = Module["_Reverse"] = Module["asm"]["Reverse"]).apply(null, arguments);
+        };
+        var _RotateWithOffset = Module["_RotateWithOffset"] = function() {
+          return (_RotateWithOffset = Module["_RotateWithOffset"] = Module["asm"]["RotateWithOffset"]).apply(null, arguments);
+        };
+        var _Round = Module["_Round"] = function() {
+          return (_Round = Module["_Round"] = Module["asm"]["Round"]).apply(null, arguments);
+        };
+        var _Rsqrt = Module["_Rsqrt"] = function() {
+          return (_Rsqrt = Module["_Rsqrt"] = Module["asm"]["Rsqrt"]).apply(null, arguments);
+        };
+        var _ScatterNd = Module["_ScatterNd"] = function() {
+          return (_ScatterNd = Module["_ScatterNd"] = Module["asm"]["ScatterNd"]).apply(null, arguments);
+        };
+        var _SelectV2 = Module["_SelectV2"] = function() {
+          return (_SelectV2 = Module["_SelectV2"] = Module["asm"]["SelectV2"]).apply(null, arguments);
+        };
+        var _Sigmoid = Module["_Sigmoid"] = function() {
+          return (_Sigmoid = Module["_Sigmoid"] = Module["asm"]["Sigmoid"]).apply(null, arguments);
+        };
+        var _Sin = Module["_Sin"] = function() {
+          return (_Sin = Module["_Sin"] = Module["asm"]["Sin"]).apply(null, arguments);
+        };
+        var _Softmax = Module["_Softmax"] = function() {
+          return (_Softmax = Module["_Softmax"] = Module["asm"]["Softmax"]).apply(null, arguments);
+        };
+        var _SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = function() {
+          return (_SparseFillEmptyRows = Module["_SparseFillEmptyRows"] = Module["asm"]["SparseFillEmptyRows"]).apply(null, arguments);
+        };
+        var _SparseReshape = Module["_SparseReshape"] = function() {
+          return (_SparseReshape = Module["_SparseReshape"] = Module["asm"]["SparseReshape"]).apply(null, arguments);
+        };
+        var _SparseSegmentReduction = Module["_SparseSegmentReduction"] = function() {
+          return (_SparseSegmentReduction = Module["_SparseSegmentReduction"] = Module["asm"]["SparseSegmentReduction"]).apply(null, arguments);
+        };
+        var _Sqrt = Module["_Sqrt"] = function() {
+          return (_Sqrt = Module["_Sqrt"] = Module["asm"]["Sqrt"]).apply(null, arguments);
+        };
+        var _Square = Module["_Square"] = function() {
+          return (_Square = Module["_Square"] = Module["asm"]["Square"]).apply(null, arguments);
+        };
+        var _SquaredDifference = Module["_SquaredDifference"] = function() {
+          return (_SquaredDifference = Module["_SquaredDifference"] = Module["asm"]["SquaredDifference"]).apply(null, arguments);
+        };
+        var _Step = Module["_Step"] = function() {
+          return (_Step = Module["_Step"] = Module["asm"]["Step"]).apply(null, arguments);
+        };
+        var _StridedSlice = Module["_StridedSlice"] = function() {
+          return (_StridedSlice = Module["_StridedSlice"] = Module["asm"]["StridedSlice"]).apply(null, arguments);
+        };
+        var _Sub = Module["_Sub"] = function() {
+          return (_Sub = Module["_Sub"] = Module["asm"]["Sub"]).apply(null, arguments);
+        };
+        var _Sum = Module["_Sum"] = function() {
+          return (_Sum = Module["_Sum"] = Module["asm"]["Sum"]).apply(null, arguments);
+        };
+        var _Tan = Module["_Tan"] = function() {
+          return (_Tan = Module["_Tan"] = Module["asm"]["Tan"]).apply(null, arguments);
+        };
+        var _Tanh = Module["_Tanh"] = function() {
+          return (_Tanh = Module["_Tanh"] = Module["asm"]["Tanh"]).apply(null, arguments);
+        };
+        var _Tile = Module["_Tile"] = function() {
+          return (_Tile = Module["_Tile"] = Module["asm"]["Tile"]).apply(null, arguments);
+        };
+        var _TopK = Module["_TopK"] = function() {
+          return (_TopK = Module["_TopK"] = Module["asm"]["TopK"]).apply(null, arguments);
+        };
+        var _Transform = Module["_Transform"] = function() {
+          return (_Transform = Module["_Transform"] = Module["asm"]["Transform"]).apply(null, arguments);
+        };
+        var _Transpose = Module["_Transpose"] = function() {
+          return (_Transpose = Module["_Transpose"] = Module["asm"]["Transpose"]).apply(null, arguments);
+        };
+        var __FusedMatMul = Module["__FusedMatMul"] = function() {
+          return (__FusedMatMul = Module["__FusedMatMul"] = Module["asm"]["_FusedMatMul"]).apply(null, arguments);
+        };
+        var _malloc = Module["_malloc"] = function() {
+          return (_malloc = Module["_malloc"] = Module["asm"]["malloc"]).apply(null, arguments);
+        };
+        var _free = Module["_free"] = function() {
+          return (_free = Module["_free"] = Module["asm"]["free"]).apply(null, arguments);
+        };
+        var ___errno_location = Module["___errno_location"] = function() {
+          return (___errno_location = Module["___errno_location"] = Module["asm"]["__errno_location"]).apply(null, arguments);
+        };
+        var stackSave = Module["stackSave"] = function() {
+          return (stackSave = Module["stackSave"] = Module["asm"]["stackSave"]).apply(null, arguments);
+        };
+        var stackRestore = Module["stackRestore"] = function() {
+          return (stackRestore = Module["stackRestore"] = Module["asm"]["stackRestore"]).apply(null, arguments);
+        };
+        var stackAlloc = Module["stackAlloc"] = function() {
+          return (stackAlloc = Module["stackAlloc"] = Module["asm"]["stackAlloc"]).apply(null, arguments);
+        };
+        var dynCall_iijjiiii = Module["dynCall_iijjiiii"] = function() {
+          return (dynCall_iijjiiii = Module["dynCall_iijjiiii"] = Module["asm"]["dynCall_iijjiiii"]).apply(null, arguments);
+        };
+        var dynCall_jiji = Module["dynCall_jiji"] = function() {
+          return (dynCall_jiji = Module["dynCall_jiji"] = Module["asm"]["dynCall_jiji"]).apply(null, arguments);
+        };
+        Module["cwrap"] = cwrap;
+        var calledRun;
+        dependenciesFulfilled = function runCaller() {
+          if (!calledRun)
+            run();
+          if (!calledRun)
+            dependenciesFulfilled = runCaller;
+        };
+        function run(args) {
+          args = args || arguments_;
+          if (runDependencies > 0) {
+            return;
+          }
+          preRun();
+          if (runDependencies > 0) {
+            return;
+          }
+          function doRun() {
+            if (calledRun)
+              return;
+            calledRun = true;
+            Module["calledRun"] = true;
+            if (ABORT)
+              return;
+            initRuntime();
+            readyPromiseResolve(Module);
+            if (Module["onRuntimeInitialized"])
+              Module["onRuntimeInitialized"]();
+            postRun();
+          }
+          if (Module["setStatus"]) {
+            Module["setStatus"]("Running...");
+            setTimeout(function() {
+              setTimeout(function() {
+                Module["setStatus"]("");
+              }, 1);
+              doRun();
+            }, 1);
+          } else {
+            doRun();
+          }
+        }
+        if (Module["preInit"]) {
+          if (typeof Module["preInit"] == "function")
+            Module["preInit"] = [Module["preInit"]];
+          while (Module["preInit"].length > 0) {
+            Module["preInit"].pop()();
+          }
+        }
+        run();
+        var listenersAdded;
+        if (beforeListeners) {
+          listenersAdded = { uncaughtException: process.listeners("uncaughtException").filter(function(listener) {
+            return !beforeListeners.uncaughtException.indexOf(listener) > -1;
+          }), unhandledRejection: process.listeners("unhandledRejection").filter(function(listener) {
+            return !beforeListeners.unhandledRejection.indexOf(listener) > -1;
+          }) };
+        }
+        var actualModule;
+        if (typeof WasmBackendModule3 !== "undefined") {
+          actualModule = WasmBackendModule3;
+        } else if (typeof WasmBackendModuleThreadedSimd !== "undefined") {
+          actualModule = WasmBackendModuleThreadedSimd;
+        } else {
+          throw new Error("Could not find wasm module in post.js");
+        }
+        if (listenersAdded) {
+          var tmpDispose = actualModule["_dispose"];
+          actualModule["_dispose"] = function() {
+            tmpDispose();
+            listenersAdded.uncaughtException.forEach(function(listener) {
+              process.removeListener("uncaughtException", listener);
+            });
+            listenersAdded.unhandledRejection.forEach(function(listener) {
+              process.removeListener("unhandledRejection", listener);
+            });
+          };
+        }
+        return WasmBackendModule3.ready;
+      };
+    })();
+    if (typeof exports === "object" && typeof module === "object")
+      module.exports = WasmBackendModule2;
+    else if (typeof define === "function" && define["amd"])
+      define([], function() {
+        return WasmBackendModule2;
+      });
+    else if (typeof exports === "object")
+      exports["WasmBackendModule"] = WasmBackendModule2;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/backend.js
+var EPSILON_FLOAT32 = 1e-7;
+var EPSILON_FLOAT16 = 1e-4;
+var DataStorage = class {
+  constructor(backend2, dataMover) {
+    this.backend = backend2;
+    this.dataMover = dataMover;
+    this.data = /* @__PURE__ */ new WeakMap();
+    this.dataIdsCount = 0;
+  }
+  get(dataId) {
+    if (!this.data.has(dataId)) {
+      this.dataMover.moveData(this.backend, dataId);
+    }
+    return this.data.get(dataId);
+  }
+  set(dataId, value) {
+    this.dataIdsCount++;
+    this.data.set(dataId, value);
+  }
+  has(dataId) {
+    return this.data.has(dataId);
+  }
+  delete(dataId) {
+    this.dataIdsCount--;
+    return this.data.delete(dataId);
+  }
+  numDataIds() {
+    return this.dataIdsCount;
+  }
+};
+var KernelBackend = class {
+  refCount(dataId) {
+    return notYetImplemented("refCount");
+  }
+  incRef(dataId) {
+    return notYetImplemented("incRef");
+  }
+  timerAvailable() {
+    return true;
+  }
+  time(f) {
+    return notYetImplemented("time");
+  }
+  read(dataId) {
+    return notYetImplemented("read");
+  }
+  readSync(dataId) {
+    return notYetImplemented("readSync");
+  }
+  readToGPU(dataId, options) {
+    return notYetImplemented("readToGPU");
+  }
+  numDataIds() {
+    return notYetImplemented("numDataIds");
+  }
+  disposeData(dataId, force) {
+    return notYetImplemented("disposeData");
+  }
+  write(values, shape, dtype) {
+    return notYetImplemented("write");
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    return notYetImplemented("move");
+  }
+  createTensorFromTexture(values, shape, dtype) {
+    return notYetImplemented("createTensorFromTexture");
+  }
+  memory() {
+    return notYetImplemented("memory");
+  }
+  floatPrecision() {
+    return notYetImplemented("floatPrecision");
+  }
+  epsilon() {
+    return this.floatPrecision() === 32 ? EPSILON_FLOAT32 : EPSILON_FLOAT16;
+  }
+  dispose() {
+    return notYetImplemented("dispose");
+  }
+};
+function notYetImplemented(kernelName) {
+  throw new Error(`'${kernelName}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/util_base.js
+function shuffle(array2) {
+  let counter = array2.length;
+  let index = 0;
+  while (counter > 0) {
+    index = Math.random() * counter | 0;
+    counter--;
+    swap(array2, counter, index);
+  }
+}
+function shuffleCombo(array2, array22) {
+  if (array2.length !== array22.length) {
+    throw new Error(`Array sizes must match to be shuffled together First array length was ${array2.length}Second array length was ${array22.length}`);
+  }
+  let counter = array2.length;
+  let index = 0;
+  while (counter > 0) {
+    index = Math.random() * counter | 0;
+    counter--;
+    swap(array2, counter, index);
+    swap(array22, counter, index);
+  }
+}
+function clamp(min6, x, max6) {
+  return Math.max(min6, Math.min(x, max6));
+}
+function nearestLargerEven(val) {
+  return val % 2 === 0 ? val : val + 1;
+}
+function swap(object, left, right) {
+  const temp = object[left];
+  object[left] = object[right];
+  object[right] = temp;
+}
+function sum(arr) {
+  let sum6 = 0;
+  for (let i = 0; i < arr.length; i++) {
+    sum6 += arr[i];
+  }
+  return sum6;
+}
+function randUniform(a, b) {
+  const r = Math.random();
+  return b * r + (1 - r) * a;
+}
+function distSquared(a, b) {
+  let result = 0;
+  for (let i = 0; i < a.length; i++) {
+    const diff = Number(a[i]) - Number(b[i]);
+    result += diff * diff;
+  }
+  return result;
+}
+function assert(expr, msg) {
+  if (!expr) {
+    throw new Error(typeof msg === "string" ? msg : msg());
+  }
+}
+function assertShapesMatch(shapeA, shapeB, errorMessagePrefix = "") {
+  assert(arraysEqual(shapeA, shapeB), () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+}
+function assertNonNull(a) {
+  assert(a != null, () => `The input to the tensor constructor must be a non-null value.`);
+}
+function flatten(arr, result = [], skipTypedArray = false) {
+  if (result == null) {
+    result = [];
+  }
+  if (Array.isArray(arr) || isTypedArray(arr) && !skipTypedArray) {
+    for (let i = 0; i < arr.length; ++i) {
+      flatten(arr[i], result, skipTypedArray);
+    }
+  } else {
+    result.push(arr);
+  }
+  return result;
+}
+function sizeFromShape(shape) {
+  if (shape.length === 0) {
+    return 1;
+  }
+  let size = shape[0];
+  for (let i = 1; i < shape.length; i++) {
+    size *= shape[i];
+  }
+  return size;
+}
+function isScalarShape(shape) {
+  return shape.length === 0;
+}
+function arraysEqual(n1, n2) {
+  if (n1 === n2) {
+    return true;
+  }
+  if (n1 == null || n2 == null) {
+    return false;
+  }
+  if (n1.length !== n2.length) {
+    return false;
+  }
+  for (let i = 0; i < n1.length; i++) {
+    if (n1[i] !== n2[i]) {
+      return false;
+    }
+  }
+  return true;
+}
+function isInt(a) {
+  return a % 1 === 0;
+}
+function tanh(x) {
+  if (Math.tanh != null) {
+    return Math.tanh(x);
+  }
+  if (x === Infinity) {
+    return 1;
+  } else if (x === -Infinity) {
+    return -1;
+  } else {
+    const e2x = Math.exp(2 * x);
+    return (e2x - 1) / (e2x + 1);
+  }
+}
+function sizeToSquarishShape(size) {
+  const width = Math.ceil(Math.sqrt(size));
+  return [width, Math.ceil(size / width)];
+}
+function createShuffledIndices(n) {
+  const shuffledIndices = new Uint32Array(n);
+  for (let i = 0; i < n; ++i) {
+    shuffledIndices[i] = i;
+  }
+  shuffle(shuffledIndices);
+  return shuffledIndices;
+}
+function rightPad(a, size) {
+  if (size <= a.length) {
+    return a;
+  }
+  return a + " ".repeat(size - a.length);
+}
+function repeatedTry(checkFn, delayFn = (counter) => 0, maxCounter, scheduleFn) {
+  return new Promise((resolve, reject) => {
+    let tryCount = 0;
+    const tryFn = () => {
+      if (checkFn()) {
+        resolve();
+        return;
+      }
+      tryCount++;
+      const nextBackoff = delayFn(tryCount);
+      if (maxCounter != null && tryCount >= maxCounter) {
+        reject();
+        return;
+      }
+      if (scheduleFn != null) {
+        scheduleFn(tryFn, nextBackoff);
+      } else {
+        setTimeout(tryFn, nextBackoff);
+      }
+    };
+    tryFn();
+  });
+}
+function inferFromImplicitShape(shape, size) {
+  let shapeProd = 1;
+  let implicitIdx = -1;
+  for (let i = 0; i < shape.length; ++i) {
+    if (shape[i] >= 0) {
+      shapeProd *= shape[i];
+    } else if (shape[i] === -1) {
+      if (implicitIdx !== -1) {
+        throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${implicitIdx} and dim ${i}`);
+      }
+      implicitIdx = i;
+    } else if (shape[i] < 0) {
+      throw Error(`Shapes can not be < 0. Found ${shape[i]} at dim ${i}`);
+    }
+  }
+  if (implicitIdx === -1) {
+    if (size > 0 && size !== shapeProd) {
+      throw Error(`Size(${size}) must match the product of shape ${shape}`);
+    }
+    return shape;
+  }
+  if (shapeProd === 0) {
+    throw Error(`Cannot infer the missing size in [${shape}] when there are 0 elements`);
+  }
+  if (size % shapeProd !== 0) {
+    throw Error(`The implicit shape can't be a fractional number. Got ${size} / ${shapeProd}`);
+  }
+  const newShape = shape.slice();
+  newShape[implicitIdx] = size / shapeProd;
+  return newShape;
+}
+function parseAxisParam(axis, shape) {
+  const rank = shape.length;
+  axis = axis == null ? shape.map((s, i) => i) : [].concat(axis);
+  assert(axis.every((ax) => ax >= -rank && ax < rank), () => `All values in axis param must be in range [-${rank}, ${rank}) but got axis ${axis}`);
+  assert(axis.every((ax) => isInt(ax)), () => `All values in axis param must be integers but got axis ${axis}`);
+  return axis.map((a) => a < 0 ? rank + a : a);
+}
+function squeezeShape(shape, axis) {
+  const newShape = [];
+  const keptDims = [];
+  const isEmptyArray = axis != null && Array.isArray(axis) && axis.length === 0;
+  const axes = axis == null || isEmptyArray ? null : parseAxisParam(axis, shape).sort();
+  let j = 0;
+  for (let i = 0; i < shape.length; ++i) {
+    if (axes != null) {
+      if (axes[j] === i && shape[i] !== 1) {
+        throw new Error(`Can't squeeze axis ${i} since its dim '${shape[i]}' is not 1`);
+      }
+      if ((axes[j] == null || axes[j] > i) && shape[i] === 1) {
+        newShape.push(shape[i]);
+        keptDims.push(i);
+      }
+      if (axes[j] <= i) {
+        j++;
+      }
+    }
+    if (shape[i] !== 1) {
+      newShape.push(shape[i]);
+      keptDims.push(i);
+    }
+  }
+  return { newShape, keptDims };
+}
+function getTypedArrayFromDType(dtype, size) {
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  return values;
+}
+function getArrayFromDType(dtype, size) {
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else if (dtype === "string") {
+    values = new Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  return values;
+}
+function checkConversionForErrors(vals, dtype) {
+  for (let i = 0; i < vals.length; i++) {
+    const num = vals[i];
+    if (isNaN(num) || !isFinite(num)) {
+      throw Error(`A tensor of type ${dtype} being uploaded contains ${num}.`);
+    }
+  }
+}
+function isValidDtype(dtype) {
+  return dtype === "bool" || dtype === "complex64" || dtype === "float32" || dtype === "int32" || dtype === "string";
+}
+function hasEncodingLoss(oldType, newType) {
+  if (newType === "complex64") {
+    return false;
+  }
+  if (newType === "float32" && oldType !== "complex64") {
+    return false;
+  }
+  if (newType === "int32" && oldType !== "float32" && oldType !== "complex64") {
+    return false;
+  }
+  if (newType === "bool" && oldType === "bool") {
+    return false;
+  }
+  return true;
+}
+function isTypedArray(a) {
+  return a instanceof Float32Array || a instanceof Int32Array || a instanceof Uint8Array || a instanceof Uint8ClampedArray;
+}
+function bytesPerElement(dtype) {
+  if (dtype === "float32" || dtype === "int32") {
+    return 4;
+  } else if (dtype === "complex64") {
+    return 8;
+  } else if (dtype === "bool") {
+    return 1;
+  } else {
+    throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+function bytesFromStringArray(arr) {
+  if (arr == null) {
+    return 0;
+  }
+  let bytes = 0;
+  arr.forEach((x) => bytes += x.length);
+  return bytes;
+}
+function isString(value) {
+  return typeof value === "string" || value instanceof String;
+}
+function isBoolean(value) {
+  return typeof value === "boolean";
+}
+function isNumber(value) {
+  return typeof value === "number";
+}
+function inferDtype(values) {
+  if (Array.isArray(values)) {
+    return inferDtype(values[0]);
+  }
+  if (values instanceof Float32Array) {
+    return "float32";
+  } else if (values instanceof Int32Array || values instanceof Uint8Array || values instanceof Uint8ClampedArray) {
+    return "int32";
+  } else if (isNumber(values)) {
+    return "float32";
+  } else if (isString(values)) {
+    return "string";
+  } else if (isBoolean(values)) {
+    return "bool";
+  }
+  return "float32";
+}
+function isFunction(f) {
+  return !!(f && f.constructor && f.call && f.apply);
+}
+function nearestDivisor(size, start) {
+  for (let i = start; i < size; ++i) {
+    if (size % i === 0) {
+      return i;
+    }
+  }
+  return size;
+}
+function computeStrides(shape) {
+  const rank = shape.length;
+  if (rank < 2) {
+    return [];
+  }
+  const strides = new Array(rank - 1);
+  strides[rank - 2] = shape[rank - 1];
+  for (let i = rank - 3; i >= 0; --i) {
+    strides[i] = strides[i + 1] * shape[i + 1];
+  }
+  return strides;
+}
+function createNestedArray(offset, shape, a, isComplex = false) {
+  const ret = new Array();
+  if (shape.length === 1) {
+    const d = shape[0] * (isComplex ? 2 : 1);
+    for (let i = 0; i < d; i++) {
+      ret[i] = a[offset + i];
+    }
+  } else {
+    const d = shape[0];
+    const rest = shape.slice(1);
+    const len = rest.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);
+    for (let i = 0; i < d; i++) {
+      ret[i] = createNestedArray(offset + i * len, rest, a, isComplex);
+    }
+  }
+  return ret;
+}
+function toNestedArray(shape, a, isComplex = false) {
+  if (shape.length === 0) {
+    return a[0];
+  }
+  const size = shape.reduce((acc, c) => acc * c) * (isComplex ? 2 : 1);
+  if (size === 0) {
+    return [];
+  }
+  if (size !== a.length) {
+    throw new Error(`[${shape}] does not match the input size ${a.length}${isComplex ? " for a complex tensor" : ""}.`);
+  }
+  return createNestedArray(0, shape, a, isComplex);
+}
+function makeOnesTypedArray(size, dtype) {
+  const array2 = makeZerosTypedArray(size, dtype);
+  for (let i = 0; i < array2.length; i++) {
+    array2[i] = 1;
+  }
+  return array2;
+}
+function makeZerosTypedArray(size, dtype) {
+  if (dtype == null || dtype === "float32" || dtype === "complex64") {
+    return new Float32Array(size);
+  } else if (dtype === "int32") {
+    return new Int32Array(size);
+  } else if (dtype === "bool") {
+    return new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function makeZerosNestedTypedArray(shape, dtype) {
+  const size = shape.reduce((prev, curr) => prev * curr, 1);
+  if (dtype == null || dtype === "float32") {
+    return toNestedArray(shape, new Float32Array(size));
+  } else if (dtype === "int32") {
+    return toNestedArray(shape, new Int32Array(size));
+  } else if (dtype === "bool") {
+    return toNestedArray(shape, new Uint8Array(size));
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function assertNonNegativeIntegerDimensions(shape) {
+  shape.forEach((dimSize) => {
+    assert(Number.isInteger(dimSize) && dimSize >= 0, () => `Tensor must have a shape comprised of positive integers but got shape [${shape}].`);
+  });
+}
+function locToIndex(locs, rank, strides) {
+  if (rank === 0) {
+    return 0;
+  } else if (rank === 1) {
+    return locs[0];
+  }
+  let index = locs[locs.length - 1];
+  for (let i = 0; i < locs.length - 1; ++i) {
+    index += strides[i] * locs[i];
+  }
+  return index;
+}
+function indexToLoc(index, rank, strides) {
+  if (rank === 0) {
+    return [];
+  } else if (rank === 1) {
+    return [index];
+  }
+  const locs = new Array(rank);
+  for (let i = 0; i < locs.length - 1; ++i) {
+    locs[i] = Math.floor(index / strides[i]);
+    index -= locs[i] * strides[i];
+  }
+  locs[locs.length - 1] = index;
+  return locs;
+}
+function isPromise(object) {
+  return object && object.then && typeof object.then === "function";
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/environment.js
+var TENSORFLOWJS_FLAGS_PREFIX = "tfjsflags";
+var Environment = class {
+  constructor(global2) {
+    this.global = global2;
+    this.flags = {};
+    this.flagRegistry = {};
+    this.urlFlags = {};
+    this.getQueryParams = getQueryParams;
+    this.populateURLFlags();
+  }
+  setPlatform(platformName, platform) {
+    if (this.platform != null) {
+      if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+        console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${platformName}.`);
+      }
+    }
+    this.platformName = platformName;
+    this.platform = platform;
+  }
+  registerFlag(flagName, evaluationFn, setHook) {
+    this.flagRegistry[flagName] = { evaluationFn, setHook };
+    if (this.urlFlags[flagName] != null) {
+      const flagValue = this.urlFlags[flagName];
+      if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+        console.warn(`Setting feature override from URL ${flagName}: ${flagValue}.`);
+      }
+      this.set(flagName, flagValue);
+    }
+  }
+  async getAsync(flagName) {
+    if (flagName in this.flags) {
+      return this.flags[flagName];
+    }
+    this.flags[flagName] = await this.evaluateFlag(flagName);
+    return this.flags[flagName];
+  }
+  get(flagName) {
+    if (flagName in this.flags) {
+      return this.flags[flagName];
+    }
+    const flagValue = this.evaluateFlag(flagName);
+    if (isPromise(flagValue)) {
+      throw new Error(`Flag ${flagName} cannot be synchronously evaluated. Please use getAsync() instead.`);
+    }
+    this.flags[flagName] = flagValue;
+    return this.flags[flagName];
+  }
+  getNumber(flagName) {
+    return this.get(flagName);
+  }
+  getBool(flagName) {
+    return this.get(flagName);
+  }
+  getFlags() {
+    return this.flags;
+  }
+  get features() {
+    return this.flags;
+  }
+  set(flagName, value) {
+    if (this.flagRegistry[flagName] == null) {
+      throw new Error(`Cannot set flag ${flagName} as it has not been registered.`);
+    }
+    this.flags[flagName] = value;
+    if (this.flagRegistry[flagName].setHook != null) {
+      this.flagRegistry[flagName].setHook(value);
+    }
+  }
+  evaluateFlag(flagName) {
+    if (this.flagRegistry[flagName] == null) {
+      throw new Error(`Cannot evaluate flag '${flagName}': no evaluation function found.`);
+    }
+    return this.flagRegistry[flagName].evaluationFn();
+  }
+  setFlags(flags) {
+    this.flags = Object.assign({}, flags);
+  }
+  reset() {
+    this.flags = {};
+    this.urlFlags = {};
+    this.populateURLFlags();
+  }
+  populateURLFlags() {
+    if (typeof this.global === "undefined" || typeof this.global.location === "undefined" || typeof this.global.location.search === "undefined") {
+      return;
+    }
+    const urlParams = this.getQueryParams(this.global.location.search);
+    if (TENSORFLOWJS_FLAGS_PREFIX in urlParams) {
+      const keyValues = urlParams[TENSORFLOWJS_FLAGS_PREFIX].split(",");
+      keyValues.forEach((keyValue) => {
+        const [key, value] = keyValue.split(":");
+        this.urlFlags[key] = parseValue(key, value);
+      });
+    }
+  }
+};
+function getQueryParams(queryString) {
+  const params = {};
+  queryString.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g, (s, ...t) => {
+    decodeParam(params, t[0], t[1]);
+    return t.join("=");
+  });
+  return params;
+}
+function decodeParam(params, name, value) {
+  params[decodeURIComponent(name)] = decodeURIComponent(value || "");
+}
+function parseValue(flagName, value) {
+  value = value.toLowerCase();
+  if (value === "true" || value === "false") {
+    return value === "true";
+  } else if (`${+value}` === value) {
+    return +value;
+  }
+  throw new Error(`Could not parse value flag value ${value} for flag ${flagName}.`);
+}
+function env() {
+  return ENV;
+}
+var ENV = null;
+function setEnvironmentGlobal(environment) {
+  ENV = environment;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/global_util.js
+var globalNameSpace;
+function getGlobalNamespace() {
+  if (globalNameSpace == null) {
+    let ns;
+    if (typeof window !== "undefined") {
+      ns = window;
+    } else if (typeof global !== "undefined") {
+      ns = global;
+    } else if (typeof process !== "undefined") {
+      ns = process;
+    } else if (typeof self !== "undefined") {
+      ns = self;
+    } else {
+      throw new Error("Could not find a global object");
+    }
+    globalNameSpace = ns;
+  }
+  return globalNameSpace;
+}
+function getGlobalMap() {
+  const ns = getGlobalNamespace();
+  if (ns._tfGlobals == null) {
+    ns._tfGlobals = /* @__PURE__ */ new Map();
+  }
+  return ns._tfGlobals;
+}
+function getGlobal(key, init2) {
+  const globalMap = getGlobalMap();
+  if (globalMap.has(key)) {
+    return globalMap.get(key);
+  } else {
+    const singleton = init2();
+    globalMap.set(key, singleton);
+    return globalMap.get(key);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/kernel_names.js
+var Abs = "Abs";
+var Acos = "Acos";
+var Acosh = "Acosh";
+var Add = "Add";
+var AddN = "AddN";
+var All = "All";
+var Any = "Any";
+var ArgMax = "ArgMax";
+var ArgMin = "ArgMin";
+var Asin = "Asin";
+var Asinh = "Asinh";
+var Atan = "Atan";
+var Atanh = "Atanh";
+var Atan2 = "Atan2";
+var AvgPool = "AvgPool";
+var AvgPoolGrad = "AvgPoolGrad";
+var AvgPool3D = "AvgPool3D";
+var AvgPool3DGrad = "AvgPool3DGrad";
+var BatchMatMul = "BatchMatMul";
+var BatchToSpaceND = "BatchToSpaceND";
+var Bincount = "Bincount";
+var BroadcastTo = "BroadcastTo";
+var BroadcastArgs = "BroadcastArgs";
+var Cast = "Cast";
+var Ceil = "Ceil";
+var ClipByValue = "ClipByValue";
+var Complex = "Complex";
+var ComplexAbs = "ComplexAbs";
+var Concat = "Concat";
+var Conv2D = "Conv2D";
+var Conv2DBackpropFilter = "Conv2DBackpropFilter";
+var Conv2DBackpropInput = "Conv2DBackpropInput";
+var Conv3D = "Conv3D";
+var Conv3DBackpropFilterV2 = "Conv3DBackpropFilterV2";
+var Conv3DBackpropInputV2 = "Conv3DBackpropInputV2";
+var Cos = "Cos";
+var Cosh = "Cosh";
+var Cumprod = "Cumprod";
+var Cumsum = "Cumsum";
+var CropAndResize = "CropAndResize";
+var DenseBincount = "DenseBincount";
+var DepthToSpace = "DepthToSpace";
+var DepthwiseConv2dNative = "DepthwiseConv2dNative";
+var DepthwiseConv2dNativeBackpropFilter = "DepthwiseConv2dNativeBackpropFilter";
+var DepthwiseConv2dNativeBackpropInput = "DepthwiseConv2dNativeBackpropInput";
+var Diag = "Diag";
+var Dilation2D = "Dilation2D";
+var Dilation2DBackpropInput = "Dilation2DBackpropInput";
+var Dilation2DBackpropFilter = "Dilation2DBackpropFilter";
+var RealDiv = "RealDiv";
+var Einsum = "Einsum";
+var Elu = "Elu";
+var EluGrad = "EluGrad";
+var Erf = "Erf";
+var Equal = "Equal";
+var Exp = "Exp";
+var ExpandDims = "ExpandDims";
+var Expm1 = "Expm1";
+var FFT = "FFT";
+var Fill = "Fill";
+var FlipLeftRight = "FlipLeftRight";
+var Floor = "Floor";
+var FloorDiv = "FloorDiv";
+var FusedBatchNorm = "FusedBatchNorm";
+var GatherV2 = "GatherV2";
+var GatherNd = "GatherNd";
+var Greater = "Greater";
+var GreaterEqual = "GreaterEqual";
+var Identity = "Identity";
+var IFFT = "IFFT";
+var Imag = "Imag";
+var IsFinite = "IsFinite";
+var IsInf = "IsInf";
+var IsNan = "IsNan";
+var LeakyRelu = "LeakyRelu";
+var Less = "Less";
+var LessEqual = "LessEqual";
+var LinSpace = "LinSpace";
+var Log = "Log";
+var Log1p = "Log1p";
+var LogicalAnd = "LogicalAnd";
+var LogicalNot = "LogicalNot";
+var LogicalOr = "LogicalOr";
+var LogicalXor = "LogicalXor";
+var LogSoftmax = "LogSoftmax";
+var LowerBound = "LowerBound";
+var LRN = "LRN";
+var LRNGrad = "LRNGrad";
+var Max = "Max";
+var Maximum = "Maximum";
+var MaxPool = "MaxPool";
+var MaxPoolGrad = "MaxPoolGrad";
+var MaxPool3D = "MaxPool3D";
+var MaxPool3DGrad = "MaxPool3DGrad";
+var MaxPoolWithArgmax = "MaxPoolWithArgmax";
+var Mean = "Mean";
+var Min = "Min";
+var Minimum = "Minimum";
+var MirrorPad = "MirrorPad";
+var Mod = "Mod";
+var Multinomial = "Multinomial";
+var Multiply = "Multiply";
+var Neg = "Neg";
+var NotEqual = "NotEqual";
+var NonMaxSuppressionV3 = "NonMaxSuppressionV3";
+var NonMaxSuppressionV4 = "NonMaxSuppressionV4";
+var NonMaxSuppressionV5 = "NonMaxSuppressionV5";
+var OnesLike = "OnesLike";
+var OneHot = "OneHot";
+var Pack = "Pack";
+var PadV2 = "PadV2";
+var Pool = "Pool";
+var Pow = "Pow";
+var Prelu = "Prelu";
+var Prod = "Prod";
+var RaggedGather = "RaggedGather";
+var RaggedRange = "RaggedRange";
+var RaggedTensorToTensor = "RaggedTensorToTensor";
+var Range = "Range";
+var Real = "Real";
+var Reciprocal = "Reciprocal";
+var Relu = "Relu";
+var Reshape = "Reshape";
+var ResizeNearestNeighbor = "ResizeNearestNeighbor";
+var ResizeNearestNeighborGrad = "ResizeNearestNeighborGrad";
+var ResizeBilinear = "ResizeBilinear";
+var ResizeBilinearGrad = "ResizeBilinearGrad";
+var Relu6 = "Relu6";
+var Reverse = "Reverse";
+var Round = "Round";
+var Rsqrt = "Rsqrt";
+var ScatterNd = "ScatterNd";
+var SearchSorted = "SearchSorted";
+var Select = "Select";
+var Selu = "Selu";
+var Slice = "Slice";
+var Sin = "Sin";
+var Sinh = "Sinh";
+var Sign = "Sign";
+var Sigmoid = "Sigmoid";
+var Softplus = "Softplus";
+var Sqrt = "Sqrt";
+var Sum = "Sum";
+var SpaceToBatchND = "SpaceToBatchND";
+var SplitV = "SplitV";
+var Softmax = "Softmax";
+var SparseFillEmptyRows = "SparseFillEmptyRows";
+var SparseReshape = "SparseReshape";
+var SparseSegmentMean = "SparseSegmentMean";
+var SparseSegmentSum = "SparseSegmentSum";
+var SparseToDense = "SparseToDense";
+var SquaredDifference = "SquaredDifference";
+var Square = "Square";
+var StridedSlice = "StridedSlice";
+var StringNGrams = "StringNGrams";
+var StringSplit = "StringSplit";
+var StringToHashBucketFast = "StringToHashBucketFast";
+var Sub = "Sub";
+var Tan = "Tan";
+var Tanh = "Tanh";
+var Tile = "Tile";
+var TopK = "TopK";
+var Transform = "Transform";
+var Transpose = "Transpose";
+var Unique = "Unique";
+var Unpack = "Unpack";
+var UnsortedSegmentSum = "UnsortedSegmentSum";
+var UpperBound = "UpperBound";
+var ZerosLike = "ZerosLike";
+var Step = "Step";
+var FromPixels = "FromPixels";
+var RotateWithOffset = "RotateWithOffset";
+var _FusedMatMul = "_FusedMatMul";
+var FusedConv2D = "FusedConv2D";
+var FusedDepthwiseConv2D = "FusedDepthwiseConv2D";
+
+// node_modules/@tensorflow/tfjs-core/dist/log.js
+function warn(...msg) {
+  if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+    console.warn(...msg);
+  }
+}
+function log(...msg) {
+  if (!(env().getBool("IS_TEST") || env().getBool("PROD"))) {
+    console.log(...msg);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/kernel_registry.js
+var kernelRegistry = getGlobal("kernelRegistry", () => /* @__PURE__ */ new Map());
+var gradRegistry = getGlobal("gradRegistry", () => /* @__PURE__ */ new Map());
+function getKernel(kernelName, backendName) {
+  const key = makeKey(kernelName, backendName);
+  return kernelRegistry.get(key);
+}
+function getGradient(kernelName) {
+  return gradRegistry.get(kernelName);
+}
+function getKernelsForBackend(backendName) {
+  const it = kernelRegistry.entries();
+  const result = [];
+  while (true) {
+    const { done, value } = it.next();
+    if (done) {
+      break;
+    }
+    const [key, config] = value;
+    const [backend2] = key.split("_");
+    if (backend2 === backendName) {
+      result.push(config);
+    }
+  }
+  return result;
+}
+function registerKernel(config) {
+  const { kernelName, backendName } = config;
+  const key = makeKey(kernelName, backendName);
+  if (kernelRegistry.has(key)) {
+    warn(`The kernel '${kernelName}' for backend '${backendName}' is already registered`);
+  }
+  kernelRegistry.set(key, config);
+}
+function registerGradient(config) {
+  const { kernelName } = config;
+  if (gradRegistry.has(kernelName)) {
+    if (env().getBool("DEBUG")) {
+      warn(`Overriding the gradient for '${kernelName}'`);
+    }
+  }
+  gradRegistry.set(kernelName, config);
+}
+function unregisterKernel(kernelName, backendName) {
+  const key = makeKey(kernelName, backendName);
+  if (!kernelRegistry.has(key)) {
+    throw new Error(`The kernel '${kernelName}' for backend '${backendName}' is not registered`);
+  }
+  kernelRegistry.delete(key);
+}
+function unregisterGradient(kernelName) {
+  if (!gradRegistry.has(kernelName)) {
+    throw new Error(`The gradient '${kernelName}' for backend is not registered`);
+  }
+  gradRegistry.delete(kernelName);
+}
+function copyRegisteredKernels(registeredBackendName, newBackendName) {
+  const kernels = getKernelsForBackend(registeredBackendName);
+  kernels.forEach((kernelConfig) => {
+    const newKernelConfig = Object.assign({}, kernelConfig, { backendName: newBackendName });
+    registerKernel(newKernelConfig);
+  });
+}
+function makeKey(kernelName, backendName) {
+  return `${backendName}_${kernelName}`;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/util.js
+var util_exports = {};
+__export(util_exports, {
+  arraysEqual: () => arraysEqual,
+  assert: () => assert,
+  assertNonNegativeIntegerDimensions: () => assertNonNegativeIntegerDimensions,
+  assertNonNull: () => assertNonNull,
+  assertShapesMatch: () => assertShapesMatch,
+  bytesFromStringArray: () => bytesFromStringArray,
+  bytesPerElement: () => bytesPerElement,
+  checkConversionForErrors: () => checkConversionForErrors,
+  clamp: () => clamp,
+  computeStrides: () => computeStrides,
+  createScalarValue: () => createScalarValue,
+  createShuffledIndices: () => createShuffledIndices,
+  decodeString: () => decodeString,
+  distSquared: () => distSquared,
+  encodeString: () => encodeString,
+  fetch: () => fetch3,
+  fingerPrint64: () => fingerPrint64,
+  flatten: () => flatten,
+  getArrayFromDType: () => getArrayFromDType,
+  getTypedArrayFromDType: () => getTypedArrayFromDType,
+  hasEncodingLoss: () => hasEncodingLoss,
+  hexToLong: () => hexToLong,
+  indexToLoc: () => indexToLoc,
+  inferDtype: () => inferDtype,
+  inferFromImplicitShape: () => inferFromImplicitShape,
+  isBoolean: () => isBoolean,
+  isFunction: () => isFunction,
+  isInt: () => isInt,
+  isNumber: () => isNumber,
+  isPromise: () => isPromise,
+  isScalarShape: () => isScalarShape,
+  isString: () => isString,
+  isTypedArray: () => isTypedArray,
+  isValidDtype: () => isValidDtype,
+  locToIndex: () => locToIndex,
+  makeOnesTypedArray: () => makeOnesTypedArray,
+  makeZerosNestedTypedArray: () => makeZerosNestedTypedArray,
+  makeZerosTypedArray: () => makeZerosTypedArray,
+  nearestDivisor: () => nearestDivisor,
+  nearestLargerEven: () => nearestLargerEven,
+  now: () => now,
+  parseAxisParam: () => parseAxisParam,
+  randUniform: () => randUniform,
+  repeatedTry: () => repeatedTry,
+  rightPad: () => rightPad,
+  shuffle: () => shuffle,
+  shuffleCombo: () => shuffleCombo,
+  sizeFromShape: () => sizeFromShape,
+  sizeToSquarishShape: () => sizeToSquarishShape,
+  squeezeShape: () => squeezeShape,
+  sum: () => sum,
+  swap: () => swap,
+  tanh: () => tanh,
+  toNestedArray: () => toNestedArray,
+  toTypedArray: () => toTypedArray
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/hash_util.js
+var LongExports = __toESM(require_long());
+var Long = LongExports.default || LongExports;
+function hexToLong(hex) {
+  return Long.fromString(hex, true, 16);
+}
+var k0 = hexToLong("c3a5c85c97cb3127");
+var k1 = hexToLong("b492b66fbe98f273");
+var k2 = hexToLong("9ae16a3b2f90404f");
+function shiftMix(val) {
+  return val.xor(val.shru(47));
+}
+function fetch2(s, offset, numBytes) {
+  const bytes = s.slice(offset, offset + numBytes);
+  return Long.fromBytes(Array.from(bytes), true, true);
+}
+function fetch64(s, offset) {
+  return fetch2(s, offset, 8);
+}
+function fetch32(s, offset) {
+  return fetch2(s, offset, 4);
+}
+function rotate64(val, shift) {
+  return shift === 0 ? val : val.shru(shift).or(val.shl(64 - shift));
+}
+function hashLen16(u, v, mul2 = hexToLong("9ddfea08eb382d69")) {
+  let a = u.xor(v).mul(mul2);
+  a = a.xor(a.shru(47));
+  let b = v.xor(a).mul(mul2);
+  b = b.xor(b.shru(47));
+  b = b.mul(mul2);
+  return b;
+}
+function weakHashLen32WithSeeds(w, x, y, z, a, b) {
+  a = a.add(w);
+  b = rotate64(b.add(a).add(z), 21);
+  const c = a;
+  a = a.add(x);
+  a = a.add(y);
+  b = b.add(rotate64(a, 44));
+  return [a.add(z), b.add(c)];
+}
+function weakHashLen32WithSeedsStr(s, offset, a, b) {
+  return weakHashLen32WithSeeds(fetch64(s, offset), fetch64(s, offset + 8), fetch64(s, offset + 16), fetch64(s, offset + 24), a, b);
+}
+function hashLen0to16(s, len = s.length) {
+  if (len >= 8) {
+    const mul2 = k2.add(len * 2);
+    const a = fetch64(s, 0).add(k2);
+    const b = fetch64(s, len - 8);
+    const c = rotate64(b, 37).mul(mul2).add(a);
+    const d = rotate64(a, 25).add(b).mul(mul2);
+    return hashLen16(c, d, mul2);
+  }
+  if (len >= 4) {
+    const mul2 = k2.add(len * 2);
+    const a = fetch32(s, 0);
+    return hashLen16(a.shl(3).add(len), fetch32(s, len - 4), mul2);
+  }
+  if (len > 0) {
+    const a = s[0];
+    const b = s[len >> 1];
+    const c = s[len - 1];
+    const y = a + (b << 8);
+    const z = len + (c << 2);
+    return shiftMix(k2.mul(y).xor(k0.mul(z))).mul(k2);
+  }
+  return k2;
+}
+function hashLen17to32(s, len = s.length) {
+  const mul2 = k2.add(len * 2);
+  const a = fetch64(s, 0).mul(k1);
+  const b = fetch64(s, 8);
+  const c = fetch64(s, len - 8).mul(mul2);
+  const d = fetch64(s, len - 16).mul(k2);
+  return hashLen16(rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d), a.add(rotate64(b.add(k2), 18)).add(c), mul2);
+}
+function hashLen33to64(s, len = s.length) {
+  const mul2 = k2.add(len * 2);
+  const a = fetch64(s, 0).mul(k2);
+  const b = fetch64(s, 8);
+  const c = fetch64(s, len - 8).mul(mul2);
+  const d = fetch64(s, len - 16).mul(k2);
+  const y = rotate64(a.add(b), 43).add(rotate64(c, 30)).add(d);
+  const z = hashLen16(y, a.add(rotate64(b.add(k2), 18)).add(c), mul2);
+  const e = fetch64(s, 16).mul(mul2);
+  const f = fetch64(s, 24);
+  const g = y.add(fetch64(s, len - 32)).mul(mul2);
+  const h = z.add(fetch64(s, len - 24)).mul(mul2);
+  return hashLen16(rotate64(e.add(f), 43).add(rotate64(g, 30)).add(h), e.add(rotate64(f.add(a), 18)).add(g), mul2);
+}
+function fingerPrint64(s, len = s.length) {
+  const seed = Long.fromNumber(81, true);
+  if (len <= 32) {
+    if (len <= 16) {
+      return hashLen0to16(s, len);
+    } else {
+      return hashLen17to32(s, len);
+    }
+  } else if (len <= 64) {
+    return hashLen33to64(s, len);
+  }
+  let x = seed;
+  let y = seed.mul(k1).add(113);
+  let z = shiftMix(y.mul(k2).add(113)).mul(k2);
+  let v = [Long.UZERO, Long.UZERO];
+  let w = [Long.UZERO, Long.UZERO];
+  x = x.mul(k2).add(fetch64(s, 0));
+  let offset = 0;
+  const end = (len - 1 >> 6) * 64;
+  const last64 = end + (len - 1 & 63) - 63;
+  do {
+    x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(k1);
+    y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(k1);
+    x = x.xor(w[1]);
+    y = y.add(v[0]).add(fetch64(s, offset + 40));
+    z = rotate64(z.add(w[0]), 33).mul(k1);
+    v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(k1), x.add(w[0]));
+    w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));
+    [z, x] = [x, z];
+    offset += 64;
+  } while (offset !== end);
+  const mul2 = k1.add(z.and(255).shl(1));
+  offset = last64;
+  w[0] = w[0].add(len - 1 & 63);
+  v[0] = v[0].add(w[0]);
+  w[0] = w[0].add(v[0]);
+  x = rotate64(x.add(y).add(v[0]).add(fetch64(s, offset + 8)), 37).mul(mul2);
+  y = rotate64(y.add(v[1]).add(fetch64(s, offset + 48)), 42).mul(mul2);
+  x = x.xor(w[1].mul(9));
+  y = y.add(v[0].mul(9).add(fetch64(s, offset + 40)));
+  z = rotate64(z.add(w[0]), 33).mul(mul2);
+  v = weakHashLen32WithSeedsStr(s, offset, v[1].mul(mul2), x.add(w[0]));
+  w = weakHashLen32WithSeedsStr(s, offset + 32, z.add(w[1]), y.add(fetch64(s, offset + 16)));
+  [z, x] = [x, z];
+  return hashLen16(hashLen16(v[0], w[0], mul2).add(shiftMix(y).mul(k0)).add(z), hashLen16(v[1], w[1], mul2).add(x), mul2);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/util.js
+function createScalarValue(value, dtype) {
+  if (dtype === "string") {
+    return encodeString(value);
+  }
+  return toTypedArray([value], dtype);
+}
+function noConversionNeeded(a, dtype) {
+  return a instanceof Float32Array && dtype === "float32" || a instanceof Int32Array && dtype === "int32" || a instanceof Uint8Array && dtype === "bool";
+}
+function toTypedArray(a, dtype) {
+  if (dtype === "string") {
+    throw new Error("Cannot convert a string[] to a TypedArray");
+  }
+  if (Array.isArray(a)) {
+    a = flatten(a);
+  }
+  if (env().getBool("DEBUG")) {
+    checkConversionForErrors(a, dtype);
+  }
+  if (noConversionNeeded(a, dtype)) {
+    return a;
+  }
+  if (dtype == null || dtype === "float32" || dtype === "complex64") {
+    return new Float32Array(a);
+  } else if (dtype === "int32") {
+    return new Int32Array(a);
+  } else if (dtype === "bool") {
+    const bool = new Uint8Array(a.length);
+    for (let i = 0; i < bool.length; ++i) {
+      if (Math.round(a[i]) !== 0) {
+        bool[i] = 1;
+      }
+    }
+    return bool;
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+}
+function now() {
+  return env().platform.now();
+}
+function fetch3(path, requestInits) {
+  return env().platform.fetch(path, requestInits);
+}
+function encodeString(s, encoding = "utf-8") {
+  encoding = encoding || "utf-8";
+  return env().platform.encode(s, encoding);
+}
+function decodeString(bytes, encoding = "utf-8") {
+  encoding = encoding || "utf-8";
+  return env().platform.decode(bytes, encoding);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/profiler.js
+var Profiler = class {
+  constructor(backendTimer, logger) {
+    this.backendTimer = backendTimer;
+    this.logger = logger;
+    if (logger == null) {
+      this.logger = new Logger();
+    }
+  }
+  profileKernel(kernelName, inputs, f) {
+    let outputs;
+    const holdResultWrapperFn = () => {
+      outputs = f();
+    };
+    let timer;
+    const start = now();
+    if (this.backendTimer.timerAvailable()) {
+      timer = this.backendTimer.time(holdResultWrapperFn);
+    } else {
+      holdResultWrapperFn();
+      for (const output of outputs) {
+        output.dataSync();
+      }
+      timer = Promise.resolve({ kernelMs: now() - start });
+    }
+    if (env().getBool("CHECK_COMPUTATION_FOR_ERRORS")) {
+      for (let i = 0; i < outputs.length; i++) {
+        const output = outputs[i];
+        output.data().then((tensorVals) => {
+          checkComputationForErrors(tensorVals, output.dtype, kernelName);
+        });
+      }
+    }
+    const kernelProfile = {
+      kernelName,
+      outputs,
+      inputs,
+      timeMs: timer.then((timing) => timing.kernelMs),
+      extraInfo: timer.then((timing) => timing.getExtraProfileInfo != null ? timing.getExtraProfileInfo() : "")
+    };
+    return kernelProfile;
+  }
+  logKernelProfile(kernelProfile) {
+    const { kernelName, outputs, timeMs, inputs, extraInfo } = kernelProfile;
+    outputs.forEach((result) => {
+      Promise.all([result.data(), timeMs, extraInfo]).then((valueContainer) => {
+        this.logger.logKernelProfile(kernelName, result, valueContainer[0], valueContainer[1], inputs, valueContainer[2]);
+      });
+    });
+  }
+};
+function checkComputationForErrors(vals, dtype, kernelName) {
+  if (dtype !== "float32") {
+    return false;
+  }
+  for (let i = 0; i < vals.length; i++) {
+    const num = vals[i];
+    if (isNaN(num) || !isFinite(num)) {
+      console.warn(`Found ${num} in the result of '${kernelName}'`);
+      return true;
+    }
+  }
+  return false;
+}
+var Logger = class {
+  logKernelProfile(name, result, vals, timeMs, inputs, extraInfo) {
+    const time2 = typeof timeMs === "number" ? rightPad(`${timeMs}ms`, 9) : timeMs["error"];
+    const paddedName = rightPad(name, 25);
+    const rank = result.rank;
+    const size = result.size;
+    const shape = rightPad(result.shape.toString(), 14);
+    let inputShapesDescription = "";
+    for (const name2 in inputs) {
+      const input2 = inputs[name2];
+      if (input2 != null) {
+        const inputShape = input2.shape || result.shape;
+        const inputRank = inputShape.length;
+        inputShapesDescription += `${name2}: ${inputRank}D ${inputRank > 0 ? inputShape : ""} `;
+      }
+    }
+    console.log(`%c${paddedName}	%c${time2}	%c${rank}D ${shape}	%c${size}	%c${inputShapesDescription}	%c${extraInfo}`, "font-weight:bold", "color:red", "color:blue", "color: orange", "color: green", "color: steelblue");
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/tape.js
+function getFilteredNodesXToY(tape, xs, y) {
+  const tensorsFromX = {};
+  const nodesFromX = {};
+  for (let i = 0; i < xs.length; i++) {
+    tensorsFromX[xs[i].id] = true;
+  }
+  for (let i = 0; i < tape.length; i++) {
+    const node = tape[i];
+    const nodeInputs = node.inputs;
+    for (const inputName in nodeInputs) {
+      const input2 = nodeInputs[inputName];
+      let anyInputFromX = false;
+      for (let j = 0; j < xs.length; j++) {
+        if (tensorsFromX[input2.id]) {
+          node.outputs.forEach((output) => tensorsFromX[output.id] = true);
+          anyInputFromX = true;
+          nodesFromX[node.id] = true;
+          break;
+        }
+      }
+      if (anyInputFromX) {
+        break;
+      }
+    }
+  }
+  const tensorsLeadToY = {};
+  tensorsLeadToY[y.id] = true;
+  const nodesToY = {};
+  for (let i = tape.length - 1; i >= 0; i--) {
+    const node = tape[i];
+    const nodeInputs = node.inputs;
+    for (let j = 0; j < node.outputs.length; j++) {
+      if (tensorsLeadToY[node.outputs[j].id]) {
+        for (const inputName in nodeInputs) {
+          tensorsLeadToY[nodeInputs[inputName].id] = true;
+          nodesToY[node.id] = true;
+        }
+        break;
+      }
+    }
+  }
+  const filteredTape = [];
+  for (let i = 0; i < tape.length; i++) {
+    const node = tape[i];
+    if (nodesFromX[node.id] && nodesToY[node.id]) {
+      const prunedInputs = {};
+      for (const inputName in node.inputs) {
+        const nodeInput = node.inputs[inputName];
+        if (tensorsFromX[nodeInput.id]) {
+          prunedInputs[inputName] = nodeInput;
+        }
+      }
+      const prunedNode = Object.assign({}, node);
+      prunedNode.inputs = prunedInputs;
+      prunedNode.outputs = node.outputs;
+      filteredTape.push(prunedNode);
+    }
+  }
+  return filteredTape;
+}
+function backpropagateGradients(tensorAccumulatedGradientMap, filteredTape, tidy2, add5) {
+  for (let i = filteredTape.length - 1; i >= 0; i--) {
+    const node = filteredTape[i];
+    const dys = [];
+    node.outputs.forEach((o) => {
+      const gradTensor = tensorAccumulatedGradientMap[o.id];
+      if (gradTensor != null) {
+        dys.push(gradTensor);
+      } else {
+        dys.push(null);
+      }
+    });
+    if (node.gradient == null) {
+      throw new Error(`Cannot compute gradient: gradient function not found for ${node.kernelName}.`);
+    }
+    const inputGradients = node.gradient(dys);
+    for (const inputName in node.inputs) {
+      if (!(inputName in inputGradients)) {
+        throw new Error(`Cannot backprop through input ${inputName}. Available gradients found: ${Object.keys(inputGradients)}.`);
+      }
+      const dx = tidy2(() => inputGradients[inputName]());
+      if (dx.dtype !== "float32") {
+        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input ${inputName} must have 'float32' dtype, but has '${dx.dtype}'`);
+      }
+      const x = node.inputs[inputName];
+      if (!arraysEqual(dx.shape, x.shape)) {
+        throw new Error(`Error in gradient for op ${node.kernelName}. The gradient of input '${inputName}' has shape '${dx.shape}', which does not match the shape of the input '${x.shape}'`);
+      }
+      if (tensorAccumulatedGradientMap[x.id] == null) {
+        tensorAccumulatedGradientMap[x.id] = dx;
+      } else {
+        const curGradient = tensorAccumulatedGradientMap[x.id];
+        tensorAccumulatedGradientMap[x.id] = add5(curGradient, dx);
+        curGradient.dispose();
+      }
+    }
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/tensor_format.js
+var FORMAT_LIMIT_NUM_VALS = 20;
+var FORMAT_NUM_FIRST_LAST_VALS = 3;
+var FORMAT_NUM_SIG_DIGITS = 7;
+function tensorToString(vals, shape, dtype, verbose) {
+  const strides = computeStrides(shape);
+  const padPerCol = computeMaxSizePerColumn(vals, shape, dtype, strides);
+  const rank = shape.length;
+  const valsLines = subTensorToString(vals, shape, dtype, strides, padPerCol);
+  const lines = ["Tensor"];
+  if (verbose) {
+    lines.push(`  dtype: ${dtype}`);
+    lines.push(`  rank: ${rank}`);
+    lines.push(`  shape: [${shape}]`);
+    lines.push(`  values:`);
+  }
+  lines.push(valsLines.map((l) => "    " + l).join("\n"));
+  return lines.join("\n");
+}
+function computeMaxSizePerColumn(vals, shape, dtype, strides) {
+  const n = sizeFromShape(shape);
+  const numCols = strides[strides.length - 1];
+  const padPerCol = new Array(numCols).fill(0);
+  const rank = shape.length;
+  const valuesOrTuples = dtype === "complex64" ? createComplexTuples(vals) : vals;
+  if (rank > 1) {
+    for (let row = 0; row < n / numCols; row++) {
+      const offset = row * numCols;
+      for (let j = 0; j < numCols; j++) {
+        padPerCol[j] = Math.max(padPerCol[j], valToString(valuesOrTuples[offset + j], 0, dtype).length);
+      }
+    }
+  }
+  return padPerCol;
+}
+function valToString(val, pad3, dtype) {
+  let valStr;
+  if (Array.isArray(val)) {
+    valStr = `${parseFloat(val[0].toFixed(FORMAT_NUM_SIG_DIGITS))} + ${parseFloat(val[1].toFixed(FORMAT_NUM_SIG_DIGITS))}j`;
+  } else if (isString(val)) {
+    valStr = `'${val}'`;
+  } else if (dtype === "bool") {
+    valStr = boolNumToString(val);
+  } else {
+    valStr = parseFloat(val.toFixed(FORMAT_NUM_SIG_DIGITS)).toString();
+  }
+  return rightPad(valStr, pad3);
+}
+function boolNumToString(v) {
+  return v === 0 ? "false" : "true";
+}
+function subTensorToString(vals, shape, dtype, strides, padPerCol, isLast = true) {
+  const storagePerElement = dtype === "complex64" ? 2 : 1;
+  const size = shape[0];
+  const rank = shape.length;
+  if (rank === 0) {
+    if (dtype === "complex64") {
+      const complexTuple = createComplexTuples(vals);
+      return [valToString(complexTuple[0], 0, dtype)];
+    }
+    if (dtype === "bool") {
+      return [boolNumToString(vals[0])];
+    }
+    return [vals[0].toString()];
+  }
+  if (rank === 1) {
+    if (size > FORMAT_LIMIT_NUM_VALS) {
+      const firstValsSize = FORMAT_NUM_FIRST_LAST_VALS * storagePerElement;
+      let firstVals = Array.from(vals.slice(0, firstValsSize));
+      let lastVals = Array.from(vals.slice((size - FORMAT_NUM_FIRST_LAST_VALS) * storagePerElement, size * storagePerElement));
+      if (dtype === "complex64") {
+        firstVals = createComplexTuples(firstVals);
+        lastVals = createComplexTuples(lastVals);
+      }
+      return [
+        "[" + firstVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(", ") + ", ..., " + lastVals.map((x, i) => valToString(x, padPerCol[size - FORMAT_NUM_FIRST_LAST_VALS + i], dtype)).join(", ") + "]"
+      ];
+    }
+    const displayVals = dtype === "complex64" ? createComplexTuples(vals) : Array.from(vals);
+    return [
+      "[" + displayVals.map((x, i) => valToString(x, padPerCol[i], dtype)).join(", ") + "]"
+    ];
+  }
+  const subshape = shape.slice(1);
+  const substrides = strides.slice(1);
+  const stride = strides[0] * storagePerElement;
+  const lines = [];
+  if (size > FORMAT_LIMIT_NUM_VALS) {
+    for (let i = 0; i < FORMAT_NUM_FIRST_LAST_VALS; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, false));
+    }
+    lines.push("...");
+    for (let i = size - FORMAT_NUM_FIRST_LAST_VALS; i < size; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));
+    }
+  } else {
+    for (let i = 0; i < size; i++) {
+      const start = i * stride;
+      const end = start + stride;
+      lines.push(...subTensorToString(vals.slice(start, end), subshape, dtype, substrides, padPerCol, i === size - 1));
+    }
+  }
+  const sep = rank === 2 ? "," : "";
+  lines[0] = "[" + lines[0] + sep;
+  for (let i = 1; i < lines.length - 1; i++) {
+    lines[i] = " " + lines[i] + sep;
+  }
+  let newLineSep = ",\n";
+  for (let i = 2; i < rank; i++) {
+    newLineSep += "\n";
+  }
+  lines[lines.length - 1] = " " + lines[lines.length - 1] + "]" + (isLast ? "" : newLineSep);
+  return lines;
+}
+function createComplexTuples(vals) {
+  const complexTuples = [];
+  for (let i = 0; i < vals.length; i += 2) {
+    complexTuples.push([vals[i], vals[i + 1]]);
+  }
+  return complexTuples;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/tensor.js
+var TensorBuffer = class {
+  constructor(shape, dtype, values) {
+    this.dtype = dtype;
+    this.shape = shape.slice();
+    this.size = sizeFromShape(shape);
+    if (values != null) {
+      const n = values.length;
+      assert(n === this.size, () => `Length of values '${n}' does not match the size inferred by the shape '${this.size}'.`);
+    }
+    if (dtype === "complex64") {
+      throw new Error(`complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).`);
+    }
+    this.values = values || getArrayFromDType(dtype, this.size);
+    this.strides = computeStrides(shape);
+  }
+  set(value, ...locs) {
+    if (locs.length === 0) {
+      locs = [0];
+    }
+    assert(locs.length === this.rank, () => `The number of provided coordinates (${locs.length}) must match the rank (${this.rank})`);
+    const index = this.locToIndex(locs);
+    this.values[index] = value;
+  }
+  get(...locs) {
+    if (locs.length === 0) {
+      locs = [0];
+    }
+    let i = 0;
+    for (const loc of locs) {
+      if (loc < 0 || loc >= this.shape[i]) {
+        const msg = `Requested out of range element at ${locs}.   Buffer shape=${this.shape}`;
+        throw new Error(msg);
+      }
+      i++;
+    }
+    let index = locs[locs.length - 1];
+    for (let i2 = 0; i2 < locs.length - 1; ++i2) {
+      index += this.strides[i2] * locs[i2];
+    }
+    return this.values[index];
+  }
+  locToIndex(locs) {
+    if (this.rank === 0) {
+      return 0;
+    } else if (this.rank === 1) {
+      return locs[0];
+    }
+    let index = locs[locs.length - 1];
+    for (let i = 0; i < locs.length - 1; ++i) {
+      index += this.strides[i] * locs[i];
+    }
+    return index;
+  }
+  indexToLoc(index) {
+    if (this.rank === 0) {
+      return [];
+    } else if (this.rank === 1) {
+      return [index];
+    }
+    const locs = new Array(this.shape.length);
+    for (let i = 0; i < locs.length - 1; ++i) {
+      locs[i] = Math.floor(index / this.strides[i]);
+      index -= locs[i] * this.strides[i];
+    }
+    locs[locs.length - 1] = index;
+    return locs;
+  }
+  get rank() {
+    return this.shape.length;
+  }
+  toTensor() {
+    return trackerFn().makeTensor(this.values, this.shape, this.dtype);
+  }
+};
+var trackerFn = null;
+var opHandler = null;
+var deprecationWarningFn = null;
+function setTensorTracker(fn) {
+  trackerFn = fn;
+}
+function setOpHandler(handler) {
+  opHandler = handler;
+}
+function setDeprecationWarningFn(fn) {
+  deprecationWarningFn = fn;
+}
+var Tensor = class {
+  constructor(shape, dtype, dataId, id) {
+    this.kept = false;
+    this.isDisposedInternal = false;
+    this.shape = shape.slice();
+    this.dtype = dtype || "float32";
+    this.size = sizeFromShape(shape);
+    this.strides = computeStrides(shape);
+    this.dataId = dataId;
+    this.id = id;
+    this.rankType = this.rank < 5 ? this.rank.toString() : "higher";
+  }
+  get rank() {
+    return this.shape.length;
+  }
+  async buffer() {
+    const vals = await this.data();
+    return opHandler.buffer(this.shape, this.dtype, vals);
+  }
+  bufferSync() {
+    return opHandler.buffer(this.shape, this.dtype, this.dataSync());
+  }
+  async array() {
+    const vals = await this.data();
+    return toNestedArray(this.shape, vals, this.dtype === "complex64");
+  }
+  arraySync() {
+    return toNestedArray(this.shape, this.dataSync(), this.dtype === "complex64");
+  }
+  async data() {
+    this.throwIfDisposed();
+    const data = trackerFn().read(this.dataId);
+    if (this.dtype === "string") {
+      const bytes = await data;
+      try {
+        return bytes.map((b) => decodeString(b));
+      } catch (_a) {
+        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
+      }
+    }
+    return data;
+  }
+  dataToGPU(options) {
+    this.throwIfDisposed();
+    return trackerFn().readToGPU(this.dataId, options);
+  }
+  dataSync() {
+    this.throwIfDisposed();
+    const data = trackerFn().readSync(this.dataId);
+    if (this.dtype === "string") {
+      try {
+        return data.map((b) => decodeString(b));
+      } catch (_a) {
+        throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().");
+      }
+    }
+    return data;
+  }
+  async bytes() {
+    this.throwIfDisposed();
+    const data = await trackerFn().read(this.dataId);
+    if (this.dtype === "string") {
+      return data;
+    } else {
+      return new Uint8Array(data.buffer);
+    }
+  }
+  dispose() {
+    if (this.isDisposed) {
+      return;
+    }
+    trackerFn().disposeTensor(this);
+    this.isDisposedInternal = true;
+  }
+  get isDisposed() {
+    return this.isDisposedInternal;
+  }
+  throwIfDisposed() {
+    if (this.isDisposed) {
+      throw new Error(`Tensor is disposed.`);
+    }
+  }
+  print(verbose = false) {
+    return opHandler.print(this, verbose);
+  }
+  clone() {
+    this.throwIfDisposed();
+    return opHandler.clone(this);
+  }
+  toString(verbose = false) {
+    const vals = this.dataSync();
+    return tensorToString(vals, this.shape, this.dtype, verbose);
+  }
+  cast(dtype) {
+    this.throwIfDisposed();
+    return opHandler.cast(this, dtype);
+  }
+  variable(trainable = true, name, dtype) {
+    this.throwIfDisposed();
+    return trackerFn().makeVariable(this, trainable, name, dtype);
+  }
+};
+Object.defineProperty(Tensor, Symbol.hasInstance, {
+  value: (instance) => {
+    return !!instance && instance.data != null && instance.dataSync != null && instance.throwIfDisposed != null;
+  }
+});
+function getGlobalTensorClass() {
+  return getGlobal("Tensor", () => {
+    return Tensor;
+  });
+}
+getGlobalTensorClass();
+var Variable = class extends Tensor {
+  constructor(initialValue, trainable, name, tensorId) {
+    super(initialValue.shape, initialValue.dtype, initialValue.dataId, tensorId);
+    this.trainable = trainable;
+    this.name = name;
+  }
+  assign(newValue) {
+    if (newValue.dtype !== this.dtype) {
+      throw new Error(`dtype of the new value (${newValue.dtype}) and previous value (${this.dtype}) must match`);
+    }
+    if (!arraysEqual(newValue.shape, this.shape)) {
+      throw new Error(`shape of the new value (${newValue.shape}) and previous value (${this.shape}) must match`);
+    }
+    trackerFn().disposeTensor(this);
+    this.dataId = newValue.dataId;
+    trackerFn().incRef(this, null);
+  }
+  dispose() {
+    trackerFn().disposeVariable(this);
+    this.isDisposedInternal = true;
+  }
+};
+Object.defineProperty(Variable, Symbol.hasInstance, {
+  value: (instance) => {
+    return instance instanceof Tensor && instance.assign != null && instance.assign instanceof Function;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/tensor_util.js
+var tensor_util_exports = {};
+__export(tensor_util_exports, {
+  assertTypesMatch: () => assertTypesMatch,
+  getTensorsInContainer: () => getTensorsInContainer,
+  isTensorInList: () => isTensorInList,
+  makeTypesMatch: () => makeTypesMatch
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/types.js
+var Rank;
+(function(Rank2) {
+  Rank2["R0"] = "R0";
+  Rank2["R1"] = "R1";
+  Rank2["R2"] = "R2";
+  Rank2["R3"] = "R3";
+  Rank2["R4"] = "R4";
+  Rank2["R5"] = "R5";
+  Rank2["R6"] = "R6";
+})(Rank || (Rank = {}));
+var UpcastInt32AndMap;
+(function(UpcastInt32AndMap2) {
+  UpcastInt32AndMap2["float32"] = "float32";
+  UpcastInt32AndMap2["int32"] = "int32";
+  UpcastInt32AndMap2["bool"] = "int32";
+  UpcastInt32AndMap2["complex64"] = "complex64";
+})(UpcastInt32AndMap || (UpcastInt32AndMap = {}));
+var UpcastBoolAndMap;
+(function(UpcastBoolAndMap2) {
+  UpcastBoolAndMap2["float32"] = "float32";
+  UpcastBoolAndMap2["int32"] = "int32";
+  UpcastBoolAndMap2["bool"] = "bool";
+  UpcastBoolAndMap2["complex64"] = "complex64";
+})(UpcastBoolAndMap || (UpcastBoolAndMap = {}));
+var UpcastFloat32AndMap;
+(function(UpcastFloat32AndMap2) {
+  UpcastFloat32AndMap2["float32"] = "float32";
+  UpcastFloat32AndMap2["int32"] = "float32";
+  UpcastFloat32AndMap2["bool"] = "float32";
+  UpcastFloat32AndMap2["complex64"] = "complex64";
+})(UpcastFloat32AndMap || (UpcastFloat32AndMap = {}));
+var UpcastComplex64AndMap;
+(function(UpcastComplex64AndMap2) {
+  UpcastComplex64AndMap2["float32"] = "complex64";
+  UpcastComplex64AndMap2["int32"] = "complex64";
+  UpcastComplex64AndMap2["bool"] = "complex64";
+  UpcastComplex64AndMap2["complex64"] = "complex64";
+})(UpcastComplex64AndMap || (UpcastComplex64AndMap = {}));
+var upcastTypeMap = {
+  "float32": UpcastFloat32AndMap,
+  "int32": UpcastInt32AndMap,
+  "bool": UpcastBoolAndMap,
+  "complex64": UpcastComplex64AndMap
+};
+function upcastType(typeA, typeB) {
+  if (typeA === "string" || typeB === "string") {
+    if (typeA === "string" && typeB === "string") {
+      return "string";
+    }
+    throw new Error(`Can not upcast ${typeA} with ${typeB}`);
+  }
+  return upcastTypeMap[typeA][typeB];
+}
+function sumOutType(type) {
+  return upcastType(type, "int32");
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/tensor_util.js
+function makeTypesMatch(a, b) {
+  if (a.dtype === b.dtype) {
+    return [a, b];
+  }
+  const dtype = upcastType(a.dtype, b.dtype);
+  return [a.cast(dtype), b.cast(dtype)];
+}
+function assertTypesMatch(a, b) {
+  assert(a.dtype === b.dtype, () => `The dtypes of the first(${a.dtype}) and second(${b.dtype}) input must match`);
+}
+function isTensorInList(tensor2, tensorList) {
+  return tensorList.some((x) => x.id === tensor2.id);
+}
+function getTensorsInContainer(result) {
+  const list = [];
+  const seen = /* @__PURE__ */ new Set();
+  walkTensorContainer(result, list, seen);
+  return list;
+}
+function walkTensorContainer(container, list, seen) {
+  if (container == null) {
+    return;
+  }
+  if (container instanceof Tensor) {
+    list.push(container);
+    return;
+  }
+  if (!isIterable(container)) {
+    return;
+  }
+  const iterable = container;
+  for (const k in iterable) {
+    const val = iterable[k];
+    if (!seen.has(val)) {
+      seen.add(val);
+      walkTensorContainer(val, list, seen);
+    }
+  }
+}
+function isIterable(obj) {
+  return Array.isArray(obj) || typeof obj === "object";
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/engine.js
+function isRegisteredKernelInvocation(kernelInvocation) {
+  return kernelInvocation.kernelName != null;
+}
+var EngineState = class {
+  constructor() {
+    this.registeredVariables = {};
+    this.nextTapeNodeId = 0;
+    this.numBytes = 0;
+    this.numTensors = 0;
+    this.numStringTensors = 0;
+    this.numDataBuffers = 0;
+    this.gradientDepth = 0;
+    this.kernelDepth = 0;
+    this.scopeStack = [];
+    this.numDataMovesStack = [];
+    this.nextScopeId = 0;
+    this.tensorInfo = /* @__PURE__ */ new WeakMap();
+    this.profiling = false;
+    this.activeProfile = {
+      newBytes: 0,
+      newTensors: 0,
+      peakBytes: 0,
+      kernels: [],
+      result: null,
+      get kernelNames() {
+        return Array.from(new Set(this.kernels.map((k) => k.name)));
+      }
+    };
+  }
+  dispose() {
+    for (const variableName in this.registeredVariables) {
+      this.registeredVariables[variableName].dispose();
+    }
+  }
+};
+var Engine = class {
+  constructor(ENV7) {
+    this.ENV = ENV7;
+    this.registry = {};
+    this.registryFactory = {};
+    this.pendingBackendInitId = 0;
+    this.state = new EngineState();
+  }
+  async ready() {
+    if (this.pendingBackendInit != null) {
+      return this.pendingBackendInit.then(() => {
+      });
+    }
+    if (this.backendInstance != null) {
+      return;
+    }
+    const sortedBackends = this.getSortedBackends();
+    for (let i = 0; i < sortedBackends.length; i++) {
+      const backendName = sortedBackends[i];
+      const success = await this.initializeBackend(backendName).success;
+      if (success) {
+        await this.setBackend(backendName);
+        return;
+      }
+    }
+    throw new Error(`Could not initialize any backends, all backend initializations failed.`);
+  }
+  get backend() {
+    if (this.pendingBackendInit != null) {
+      throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+    }
+    if (this.backendInstance == null) {
+      const { name, asyncInit } = this.initializeBackendsAndReturnBest();
+      if (asyncInit) {
+        throw new Error(`The highest priority backend '${name}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);
+      }
+      this.setBackend(name);
+    }
+    return this.backendInstance;
+  }
+  backendNames() {
+    return Object.keys(this.registryFactory);
+  }
+  findBackend(backendName) {
+    if (!(backendName in this.registry)) {
+      if (backendName in this.registryFactory) {
+        const { asyncInit } = this.initializeBackend(backendName);
+        if (asyncInit) {
+          return null;
+        }
+      } else {
+        return null;
+      }
+    }
+    return this.registry[backendName];
+  }
+  findBackendFactory(backendName) {
+    if (!(backendName in this.registryFactory)) {
+      return null;
+    }
+    return this.registryFactory[backendName].factory;
+  }
+  registerBackend(backendName, factory, priority = 1) {
+    if (backendName in this.registryFactory) {
+      warn(`${backendName} backend was already registered. Reusing existing backend factory.`);
+      return false;
+    }
+    this.registryFactory[backendName] = { factory, priority };
+    return true;
+  }
+  async setBackend(backendName) {
+    if (this.registryFactory[backendName] == null) {
+      throw new Error(`Backend name '${backendName}' not found in registry`);
+    }
+    this.backendName = backendName;
+    if (this.registry[backendName] == null) {
+      this.backendInstance = null;
+      const { success, asyncInit } = this.initializeBackend(backendName);
+      const result = asyncInit ? await success : success;
+      if (!result) {
+        return false;
+      }
+    }
+    this.backendInstance = this.registry[backendName];
+    this.setupRegisteredKernels();
+    this.profiler = new Profiler(this.backendInstance);
+    return true;
+  }
+  setupRegisteredKernels() {
+    const kernels = getKernelsForBackend(this.backendName);
+    kernels.forEach((kernel) => {
+      if (kernel.setupFunc != null) {
+        kernel.setupFunc(this.backendInstance);
+      }
+    });
+  }
+  disposeRegisteredKernels(backendName) {
+    const kernels = getKernelsForBackend(backendName);
+    kernels.forEach((kernel) => {
+      if (kernel.disposeFunc != null) {
+        kernel.disposeFunc(this.registry[backendName]);
+      }
+    });
+  }
+  initializeBackend(backendName) {
+    const registryFactoryEntry = this.registryFactory[backendName];
+    if (registryFactoryEntry == null) {
+      throw new Error(`Cannot initialize backend ${backendName}, no registration found.`);
+    }
+    try {
+      const backend2 = registryFactoryEntry.factory();
+      if (backend2 && !(backend2 instanceof KernelBackend) && typeof backend2.then === "function") {
+        const promiseId = ++this.pendingBackendInitId;
+        const success = backend2.then((backendInstance) => {
+          if (promiseId < this.pendingBackendInitId) {
+            return false;
+          }
+          this.registry[backendName] = backendInstance;
+          this.pendingBackendInit = null;
+          return true;
+        }).catch((err) => {
+          if (promiseId < this.pendingBackendInitId) {
+            return false;
+          }
+          this.pendingBackendInit = null;
+          warn(`Initialization of backend ${backendName} failed`);
+          warn(err.stack || err.message);
+          return false;
+        });
+        this.pendingBackendInit = success;
+        return { success, asyncInit: true };
+      } else {
+        this.registry[backendName] = backend2;
+        return { success: true, asyncInit: false };
+      }
+    } catch (err) {
+      warn(`Initialization of backend ${backendName} failed`);
+      warn(err.stack || err.message);
+      return { success: false, asyncInit: false };
+    }
+  }
+  removeBackend(backendName) {
+    if (!(backendName in this.registryFactory)) {
+      throw new Error(`${backendName} backend not found in registry`);
+    }
+    if (this.backendName === backendName && this.pendingBackendInit != null) {
+      this.pendingBackendInitId++;
+    }
+    if (backendName in this.registry) {
+      this.disposeRegisteredKernels(backendName);
+      this.registry[backendName].dispose();
+      delete this.registry[backendName];
+    }
+    delete this.registryFactory[backendName];
+    if (this.backendName === backendName) {
+      this.pendingBackendInit = null;
+      this.backendName = null;
+      this.backendInstance = null;
+    }
+  }
+  getSortedBackends() {
+    if (Object.keys(this.registryFactory).length === 0) {
+      throw new Error("No backend found in registry.");
+    }
+    return Object.keys(this.registryFactory).sort((a, b) => {
+      return this.registryFactory[b].priority - this.registryFactory[a].priority;
+    });
+  }
+  initializeBackendsAndReturnBest() {
+    const sortedBackends = this.getSortedBackends();
+    for (let i = 0; i < sortedBackends.length; i++) {
+      const backendName = sortedBackends[i];
+      const { success, asyncInit } = this.initializeBackend(backendName);
+      if (asyncInit || success) {
+        return { name: backendName, asyncInit };
+      }
+    }
+    throw new Error(`Could not initialize any backends, all backend initializations failed.`);
+  }
+  moveData(backend2, dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    const srcBackend = info.backend;
+    const values = this.readSync(dataId);
+    const refCount = srcBackend.refCount(dataId);
+    srcBackend.disposeData(dataId, true);
+    info.backend = backend2;
+    backend2.move(dataId, values, info.shape, info.dtype, refCount);
+    if (this.shouldCheckForMemLeaks()) {
+      this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1]++;
+    }
+  }
+  tidy(nameOrFn, fn) {
+    let name = null;
+    if (fn == null) {
+      if (typeof nameOrFn !== "function") {
+        throw new Error("Please provide a function to tidy()");
+      }
+      fn = nameOrFn;
+    } else {
+      if (typeof nameOrFn !== "string" && !(nameOrFn instanceof String)) {
+        throw new Error("When calling with two arguments, the first argument to tidy() must be a string");
+      }
+      if (typeof fn !== "function") {
+        throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");
+      }
+      name = nameOrFn;
+    }
+    let result;
+    return this.scopedRun(() => this.startScope(name), () => this.endScope(result), () => {
+      result = fn();
+      if (result instanceof Promise) {
+        console.error("Cannot return a Promise inside of tidy.");
+      }
+      return result;
+    });
+  }
+  scopedRun(start, end, f) {
+    start();
+    try {
+      const res = f();
+      end();
+      return res;
+    } catch (ex) {
+      end();
+      throw ex;
+    }
+  }
+  nextTensorId() {
+    return Engine.nextTensorId++;
+  }
+  nextVariableId() {
+    return Engine.nextVariableId++;
+  }
+  clone(x) {
+    const y = ENGINE.runKernel(Identity, { x });
+    const inputs = { x };
+    const grad2 = (dy) => ({
+      x: () => {
+        const dtype = "float32";
+        const gradInputs = { x: dy };
+        const attrs = { dtype };
+        return ENGINE.runKernel(
+          Cast,
+          gradInputs,
+          attrs
+        );
+      }
+    });
+    const saved = [];
+    this.addTapeNode(this.state.activeScope.name, inputs, [y], grad2, saved, {});
+    return y;
+  }
+  runKernel(kernelName, inputs, attrs) {
+    if (this.backendName == null) {
+      this.backend;
+    }
+    const hasKernel = getKernel(kernelName, this.backendName) != null;
+    if (!hasKernel) {
+      throw new Error(`Kernel '${kernelName}' not registered for backend '${this.backendName}'`);
+    }
+    return this.runKernelFunc({ kernelName, inputs, attrs });
+  }
+  shouldCheckForMemLeaks() {
+    return this.ENV.getBool("IS_TEST");
+  }
+  checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos) {
+    const numDataIdsAfter = this.backend.numDataIds();
+    let numOutputDataIds = 0;
+    outInfos.forEach((info) => {
+      numOutputDataIds += info.dtype === "complex64" ? 3 : 1;
+    });
+    const numMoves = this.state.numDataMovesStack[this.state.numDataMovesStack.length - 1];
+    const dataIdsLeaked = numDataIdsAfter - numDataIdsBefore - numOutputDataIds - numMoves;
+    if (dataIdsLeaked > 0) {
+      throw new Error(`Backend '${this.backendName}' has an internal memory leak (${dataIdsLeaked} data ids) after running '${kernelName}'`);
+    }
+  }
+  runKernelFunc(kernelParams) {
+    let outputs;
+    let saved = [];
+    const isTapeOn = this.isTapeOn();
+    const startingBytecount = this.state.numBytes;
+    const startingNumTensors = this.state.numTensors;
+    if (this.shouldCheckForMemLeaks()) {
+      this.state.numDataMovesStack.push(0);
+    }
+    let kernelFunc3;
+    if (this.backendName == null) {
+      this.backend;
+    }
+    let out;
+    const kernelOrScopeName = isRegisteredKernelInvocation(kernelParams) ? kernelParams.kernelName : this.state.activeScope != null ? this.state.activeScope.name : "";
+    if (isRegisteredKernelInvocation(kernelParams)) {
+      const { kernelName, inputs: inputs2, attrs: attrs2 } = kernelParams;
+      if (this.backendName == null) {
+        this.backend;
+      }
+      const kernel = getKernel(kernelName, this.backendName);
+      assert(kernel != null, () => `Cannot find registered kernel '${kernelName}' for backend '${this.backendName}'`);
+      kernelFunc3 = () => {
+        const numDataIdsBefore = this.backend.numDataIds();
+        out = kernel.kernelFunc({ inputs: inputs2, attrs: attrs2, backend: this.backend });
+        const outInfos = Array.isArray(out) ? out : [out];
+        if (this.shouldCheckForMemLeaks()) {
+          this.checkKernelForMemLeak(kernelName, numDataIdsBefore, outInfos);
+        }
+        const outTensors = outInfos.map((outInfo) => {
+          if (outInfo.rank != null) {
+            return outInfo;
+          }
+          return this.makeTensorFromTensorInfo(outInfo);
+        });
+        if (isTapeOn) {
+          const tensorsToSave = this.getTensorsForGradient(kernelName, inputs2, outTensors);
+          saved = this.saveTensorsForBackwardMode(tensorsToSave);
+        }
+        return outTensors;
+      };
+    } else {
+      const { forwardFunc } = kernelParams;
+      const saveFunc = (tensors) => {
+        if (!isTapeOn) {
+          return;
+        }
+        saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));
+      };
+      kernelFunc3 = () => {
+        const numDataIdsBefore = this.backend.numDataIds();
+        out = this.tidy(() => forwardFunc(this.backend, saveFunc));
+        const outs = Array.isArray(out) ? out : [out];
+        if (this.shouldCheckForMemLeaks()) {
+          this.checkKernelForMemLeak(kernelOrScopeName, numDataIdsBefore, outs);
+        }
+        return outs;
+      };
+    }
+    const { inputs, attrs } = kernelParams;
+    const backwardsFunc = isRegisteredKernelInvocation(kernelParams) ? null : kernelParams.backwardsFunc;
+    let kernelProfile;
+    this.scopedRun(
+      () => this.state.kernelDepth++,
+      () => this.state.kernelDepth--,
+      () => {
+        if (!this.ENV.getBool("DEBUG") && !this.state.profiling) {
+          outputs = kernelFunc3();
+        } else {
+          kernelProfile = this.profiler.profileKernel(kernelOrScopeName, inputs, () => kernelFunc3());
+          if (this.ENV.getBool("DEBUG")) {
+            this.profiler.logKernelProfile(kernelProfile);
+          }
+          outputs = kernelProfile.outputs;
+        }
+      }
+    );
+    if (isTapeOn) {
+      this.addTapeNode(kernelOrScopeName, inputs, outputs, backwardsFunc, saved, attrs);
+    }
+    if (this.state.profiling) {
+      this.state.activeProfile.kernels.push({
+        name: kernelOrScopeName,
+        bytesAdded: this.state.numBytes - startingBytecount,
+        totalBytesSnapshot: this.state.numBytes,
+        tensorsAdded: this.state.numTensors - startingNumTensors,
+        totalTensorsSnapshot: this.state.numTensors,
+        inputShapes: Object.keys(inputs).map((key) => inputs[key] != null ? inputs[key].shape : null),
+        outputShapes: outputs.map((item) => item.shape),
+        kernelTimeMs: kernelProfile.timeMs,
+        extraInfo: kernelProfile.extraInfo
+      });
+    }
+    return Array.isArray(out) ? outputs : outputs[0];
+  }
+  saveTensorsForBackwardMode(tensors) {
+    const saved = tensors.map((tensor2) => this.keep(this.clone(tensor2)));
+    return saved;
+  }
+  getTensorsForGradient(kernelName, inputs, outputs) {
+    const gradConfig = getGradient(kernelName);
+    if (gradConfig != null) {
+      const inputsToSave = gradConfig.inputsToSave || [];
+      const outputsToSave = gradConfig.outputsToSave || [];
+      let inputTensorsToSave;
+      if (gradConfig.saveAllInputs) {
+        assert(Array.isArray(inputs), () => "saveAllInputs is true, expected inputs to be an array.");
+        inputTensorsToSave = Object.keys(inputs).map((key) => inputs[key]);
+      } else {
+        inputTensorsToSave = inputsToSave.map((inputName) => inputs[inputName]);
+      }
+      const outputTensorsToSave = outputs.filter((_, i) => outputsToSave[i]);
+      return inputTensorsToSave.concat(outputTensorsToSave);
+    }
+    return [];
+  }
+  makeTensor(values, shape, dtype, backend2) {
+    if (values == null) {
+      throw new Error("Values passed to engine.makeTensor() are null");
+    }
+    dtype = dtype || "float32";
+    backend2 = backend2 || this.backend;
+    let backendVals = values;
+    if (dtype === "string" && isString(values[0])) {
+      backendVals = values.map((d) => encodeString(d));
+    }
+    const dataId = backend2.write(backendVals, shape, dtype);
+    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());
+    this.trackTensor(t, backend2);
+    if (dtype === "string") {
+      const info = this.state.tensorInfo.get(dataId);
+      const newBytes = bytesFromStringArray(backendVals);
+      this.state.numBytes += newBytes - info.bytes;
+      info.bytes = newBytes;
+    }
+    return t;
+  }
+  makeTensorFromDataId(dataId, shape, dtype, backend2) {
+    dtype = dtype || "float32";
+    const tensorInfo = { dataId, shape, dtype };
+    return this.makeTensorFromTensorInfo(tensorInfo, backend2);
+  }
+  makeTensorFromTensorInfo(tensorInfo, backend2) {
+    const { dataId, shape, dtype } = tensorInfo;
+    const t = new Tensor(shape, dtype, dataId, this.nextTensorId());
+    this.trackTensor(t, backend2);
+    return t;
+  }
+  makeVariable(initialValue, trainable = true, name, dtype) {
+    name = name || this.nextVariableId().toString();
+    if (dtype != null && dtype !== initialValue.dtype) {
+      initialValue = initialValue.cast(dtype);
+    }
+    const v = new Variable(initialValue, trainable, name, this.nextTensorId());
+    if (this.state.registeredVariables[v.name] != null) {
+      throw new Error(`Variable with name ${v.name} was already registered`);
+    }
+    this.state.registeredVariables[v.name] = v;
+    this.incRef(v, this.backend);
+    return v;
+  }
+  trackTensor(a, backend2) {
+    this.state.numTensors++;
+    if (a.dtype === "string") {
+      this.state.numStringTensors++;
+    }
+    let bytes = 0;
+    if (a.dtype !== "complex64" && a.dtype !== "string") {
+      bytes = a.size * bytesPerElement(a.dtype);
+    }
+    this.state.numBytes += bytes;
+    if (!this.state.tensorInfo.has(a.dataId)) {
+      this.state.numDataBuffers++;
+      this.state.tensorInfo.set(a.dataId, {
+        backend: backend2 || this.backend,
+        dtype: a.dtype,
+        shape: a.shape,
+        bytes
+      });
+    }
+    if (!(a instanceof Variable)) {
+      this.track(a);
+    }
+  }
+  incRef(a, backend2) {
+    this.trackTensor(a, backend2);
+    this.backend.incRef(a.dataId);
+  }
+  removeDataId(dataId, backend2) {
+    if (this.state.tensorInfo.has(dataId) && this.state.tensorInfo.get(dataId).backend === backend2) {
+      this.state.tensorInfo.delete(dataId);
+      this.state.numDataBuffers--;
+    }
+  }
+  disposeTensor(a) {
+    if (!this.state.tensorInfo.has(a.dataId)) {
+      return;
+    }
+    const info = this.state.tensorInfo.get(a.dataId);
+    this.state.numTensors--;
+    if (a.dtype === "string") {
+      this.state.numStringTensors--;
+      this.state.numBytes -= info.bytes;
+    }
+    if (a.dtype !== "complex64" && a.dtype !== "string") {
+      const bytes = a.size * bytesPerElement(a.dtype);
+      this.state.numBytes -= bytes;
+    }
+    if (info.backend.disposeData(a.dataId)) {
+      this.removeDataId(a.dataId, info.backend);
+    }
+  }
+  disposeVariables() {
+    for (const varName in this.state.registeredVariables) {
+      const v = this.state.registeredVariables[varName];
+      this.disposeVariable(v);
+    }
+  }
+  disposeVariable(v) {
+    this.disposeTensor(v);
+    if (this.state.registeredVariables[v.name] != null) {
+      delete this.state.registeredVariables[v.name];
+    }
+  }
+  memory() {
+    const info = this.backend.memory();
+    info.numTensors = this.state.numTensors;
+    info.numDataBuffers = this.state.numDataBuffers;
+    info.numBytes = this.state.numBytes;
+    if (this.state.numStringTensors > 0) {
+      info.unreliable = true;
+      if (info.reasons == null) {
+        info.reasons = [];
+      }
+      info.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)");
+    }
+    return info;
+  }
+  async profile(query) {
+    this.state.profiling = true;
+    const startBytes = this.state.numBytes;
+    const startNumTensors = this.state.numTensors;
+    this.state.activeProfile.kernels = [];
+    this.state.activeProfile.result = await query();
+    this.state.profiling = false;
+    this.state.activeProfile.peakBytes = Math.max(...this.state.activeProfile.kernels.map((d) => d.totalBytesSnapshot));
+    this.state.activeProfile.newBytes = this.state.numBytes - startBytes;
+    this.state.activeProfile.newTensors = this.state.numTensors - startNumTensors;
+    for (const kernel of this.state.activeProfile.kernels) {
+      kernel.kernelTimeMs = await kernel.kernelTimeMs;
+      kernel.extraInfo = await kernel.extraInfo;
+    }
+    return this.state.activeProfile;
+  }
+  isTapeOn() {
+    return this.state.gradientDepth > 0 && this.state.kernelDepth === 0;
+  }
+  addTapeNode(kernelName, inputs, outputs, gradientsFunc, saved, attrs) {
+    const tapeNode = { id: this.state.nextTapeNodeId++, kernelName, inputs, outputs, saved };
+    const gradConfig = getGradient(kernelName);
+    if (gradConfig != null) {
+      gradientsFunc = gradConfig.gradFunc;
+    }
+    if (gradientsFunc != null) {
+      tapeNode.gradient = (dys) => {
+        dys = dys.map((dy, i) => {
+          if (dy == null) {
+            const output = outputs[i];
+            const vals = makeZerosTypedArray(output.size, output.dtype);
+            return this.makeTensor(vals, output.shape, output.dtype);
+          }
+          return dy;
+        });
+        return gradientsFunc(dys.length > 1 ? dys : dys[0], saved, attrs);
+      };
+    }
+    this.state.activeTape.push(tapeNode);
+  }
+  keep(result) {
+    result.kept = true;
+    return result;
+  }
+  startTape() {
+    if (this.state.gradientDepth === 0) {
+      this.state.activeTape = [];
+    }
+    this.state.gradientDepth++;
+  }
+  endTape() {
+    this.state.gradientDepth--;
+  }
+  startScope(name) {
+    const scopeInfo = {
+      track: [],
+      name: "unnamed scope",
+      id: this.state.nextScopeId++
+    };
+    if (name) {
+      scopeInfo.name = name;
+    }
+    this.state.scopeStack.push(scopeInfo);
+    this.state.activeScope = scopeInfo;
+  }
+  endScope(result) {
+    const tensorsToTrackInParent = getTensorsInContainer(result);
+    const tensorsToTrackInParentSet = new Set(tensorsToTrackInParent.map((t) => t.id));
+    for (let i = 0; i < this.state.activeScope.track.length; i++) {
+      const tensor2 = this.state.activeScope.track[i];
+      if (!tensor2.kept && !tensorsToTrackInParentSet.has(tensor2.id)) {
+        tensor2.dispose();
+      }
+    }
+    const oldScope = this.state.scopeStack.pop();
+    this.state.activeScope = this.state.scopeStack.length === 0 ? null : this.state.scopeStack[this.state.scopeStack.length - 1];
+    tensorsToTrackInParent.forEach((tensor2) => {
+      if (!tensor2.kept && tensor2.scopeId === oldScope.id) {
+        this.track(tensor2);
+      }
+    });
+  }
+  gradients(f, xs, dy, allowNoGradients = false) {
+    assert(xs.length > 0, () => "gradients() received an empty list of xs.");
+    if (dy != null && dy.dtype !== "float32") {
+      throw new Error(`dy must have 'float32' dtype, but has '${dy.dtype}'`);
+    }
+    const y = this.scopedRun(() => this.startTape(), () => this.endTape(), () => this.tidy("forward", f));
+    assert(y instanceof Tensor, () => "The result y returned by f() must be a tensor.");
+    const filteredTape = getFilteredNodesXToY(this.state.activeTape, xs, y);
+    if (!allowNoGradients && filteredTape.length === 0 && xs.length > 0) {
+      throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");
+    }
+    return this.tidy("backward", () => {
+      const accumulatedGradientMap = {};
+      accumulatedGradientMap[y.id] = dy == null ? ones(y.shape) : dy;
+      backpropagateGradients(
+        accumulatedGradientMap,
+        filteredTape,
+        (f2) => this.tidy(f2),
+        add
+      );
+      const grads2 = xs.map((x) => accumulatedGradientMap[x.id]);
+      if (this.state.gradientDepth === 0) {
+        this.state.activeTape.forEach((node) => {
+          for (const tensor2 of node.saved) {
+            tensor2.dispose();
+          }
+        });
+        this.state.activeTape = null;
+      }
+      return { value: y, grads: grads2 };
+    });
+  }
+  customGrad(f) {
+    assert(isFunction(f), () => "The f passed in customGrad(f) must be a function.");
+    return (...inputs) => {
+      assert(inputs.every((t) => t instanceof Tensor), () => "The args passed in customGrad(f)(x1, x2,...) must all be tensors");
+      let res;
+      const inputMap = {};
+      inputs.forEach((input2, i) => {
+        inputMap[i] = input2;
+      });
+      const forwardFunc = (_, save) => {
+        res = f(...[...inputs, save]);
+        assert(res.value instanceof Tensor, () => "The function f passed in customGrad(f) must return an object where `obj.value` is a tensor");
+        assert(isFunction(res.gradFunc), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function.");
+        return res.value;
+      };
+      const backwardsFunc = (dy, saved) => {
+        const gradRes = res.gradFunc(dy, saved);
+        const grads2 = Array.isArray(gradRes) ? gradRes : [gradRes];
+        assert(grads2.length === inputs.length, () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...).");
+        assert(grads2.every((t) => t instanceof Tensor), () => "The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");
+        const gradMap = {};
+        grads2.forEach((grad2, i) => {
+          gradMap[i] = () => grad2;
+        });
+        return gradMap;
+      };
+      return this.runKernelFunc({
+        forwardFunc,
+        backwardsFunc,
+        inputs: inputMap
+      });
+    };
+  }
+  readSync(dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.readSync(dataId);
+  }
+  read(dataId) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.read(dataId);
+  }
+  readToGPU(dataId, options) {
+    const info = this.state.tensorInfo.get(dataId);
+    return info.backend.readToGPU(dataId, options);
+  }
+  async time(query) {
+    const start = now();
+    const timingInfo = await this.backend.time(query);
+    timingInfo.wallMs = now() - start;
+    return timingInfo;
+  }
+  track(result) {
+    if (this.state.activeScope != null) {
+      result.scopeId = this.state.activeScope.id;
+      this.state.activeScope.track.push(result);
+    }
+    return result;
+  }
+  get registeredVariables() {
+    return this.state.registeredVariables;
+  }
+  reset() {
+    this.pendingBackendInitId++;
+    this.state.dispose();
+    this.ENV.reset();
+    this.state = new EngineState();
+    for (const backendName in this.registry) {
+      this.disposeRegisteredKernels(backendName);
+      this.registry[backendName].dispose();
+      delete this.registry[backendName];
+    }
+    this.backendName = null;
+    this.backendInstance = null;
+    this.pendingBackendInit = null;
+  }
+};
+Engine.nextTensorId = 0;
+Engine.nextVariableId = 0;
+function ones(shape) {
+  const values = makeOnesTypedArray(sizeFromShape(shape), "float32");
+  return ENGINE.makeTensor(values, shape, "float32");
+}
+function getOrMakeEngine() {
+  const ns = getGlobalNamespace();
+  if (ns._tfengine == null) {
+    const environment = new Environment(ns);
+    ns._tfengine = new Engine(environment);
+  }
+  setEnvironmentGlobal(ns._tfengine.ENV);
+  setTensorTracker(() => ns._tfengine);
+  return ns._tfengine;
+}
+var ENGINE = getOrMakeEngine();
+function add(a, b) {
+  const inputs = { a, b };
+  return ENGINE.runKernel(Add, inputs);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/device_util.js
+var device_util_exports = {};
+__export(device_util_exports, {
+  isBrowser: () => isBrowser,
+  isMobile: () => isMobile,
+  mockIsMobile: () => mockIsMobile
+});
+function _isNavigatorDefined() {
+  return typeof navigator !== "undefined" && navigator != null;
+}
+var isMobileMockValue;
+function mockIsMobile(value) {
+  isMobileMockValue = value;
+}
+function isMobile(nav) {
+  if (isMobileMockValue !== void 0) {
+    return isMobileMockValue;
+  }
+  if (nav || _isNavigatorDefined()) {
+    if (!nav) {
+      nav = navigator;
+    }
+    if (nav.product === "ReactNative") {
+      return true;
+    }
+    const a = nav.userAgent || nav.vendor || (typeof window !== "undefined" ? window.opera : "");
+    if (!a) {
+      const navAny = nav;
+      return navAny.userAgentData && navAny.userAgentData.mobile;
+    }
+    return /(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(a) || /1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(a.substr(0, 4));
+  }
+  return false;
+}
+function isBrowser() {
+  return typeof window !== "undefined" && window.document != null || typeof WorkerGlobalScope !== "undefined";
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/flags.js
+var ENV2 = env();
+ENV2.registerFlag("DEBUG", () => false, (debugValue) => {
+  if (debugValue) {
+    console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.");
+  }
+});
+ENV2.registerFlag("IS_BROWSER", () => isBrowser());
+ENV2.registerFlag("IS_NODE", () => typeof process !== "undefined" && typeof process.versions !== "undefined" && typeof process.versions.node !== "undefined");
+ENV2.registerFlag("IS_CHROME", () => typeof navigator !== "undefined" && navigator != null && navigator.userAgent != null && /Chrome/.test(navigator.userAgent) && /Google Inc/.test(navigator.vendor));
+ENV2.registerFlag("PROD", () => false);
+ENV2.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY", () => ENV2.getBool("DEBUG"));
+ENV2.registerFlag("DEPRECATION_WARNINGS_ENABLED", () => true);
+ENV2.registerFlag("IS_TEST", () => false);
+ENV2.registerFlag("CHECK_COMPUTATION_FOR_ERRORS", () => true);
+ENV2.registerFlag("WRAP_TO_IMAGEBITMAP", () => false);
+ENV2.registerFlag("ENGINE_COMPILE_ONLY", () => false);
+ENV2.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU", () => false);
+ENV2.registerFlag("USE_SETTIMEOUTCUSTOM", () => false);
+
+// node_modules/@tensorflow/tfjs-core/dist/tensor_util_env.js
+function inferShape(val, dtype) {
+  let firstElem = val;
+  if (isTypedArray(val)) {
+    return dtype === "string" ? [] : [val.length];
+  }
+  if (typeof val === "object" && "texture" in val) {
+    const usedChannels = val.channels || "RGBA";
+    return [val.height, val.width * usedChannels.length];
+  }
+  if (!Array.isArray(val)) {
+    return [];
+  }
+  const shape = [];
+  while (Array.isArray(firstElem) || isTypedArray(firstElem) && dtype !== "string") {
+    shape.push(firstElem.length);
+    firstElem = firstElem[0];
+  }
+  if (Array.isArray(val) && env().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")) {
+    deepAssertShapeConsistency(val, shape, []);
+  }
+  return shape;
+}
+function deepAssertShapeConsistency(val, shape, indices) {
+  indices = indices || [];
+  if (!Array.isArray(val) && !isTypedArray(val)) {
+    assert(shape.length === 0, () => `Element arr[${indices.join("][")}] is a primitive, but should be an array/TypedArray of ${shape[0]} elements`);
+    return;
+  }
+  assert(shape.length > 0, () => `Element arr[${indices.join("][")}] should be a primitive, but is an array of ${val.length} elements`);
+  assert(val.length === shape[0], () => `Element arr[${indices.join("][")}] should have ${shape[0]} elements, but has ${val.length} elements`);
+  const subShape = shape.slice(1);
+  for (let i = 0; i < val.length; ++i) {
+    deepAssertShapeConsistency(val[i], subShape, indices.concat(i));
+  }
+}
+function assertDtype(expectedDtype, actualDType, argName, functionName) {
+  if (expectedDtype === "string_or_numeric") {
+    return;
+  }
+  if (expectedDtype == null) {
+    throw new Error(`Expected dtype cannot be null.`);
+  }
+  if (expectedDtype !== "numeric" && expectedDtype !== actualDType || expectedDtype === "numeric" && actualDType === "string") {
+    throw new Error(`Argument '${argName}' passed to '${functionName}' must be ${expectedDtype} tensor, but got ${actualDType} tensor`);
+  }
+}
+function convertToTensor(x, argName, functionName, parseAsDtype = "numeric") {
+  if (x instanceof Tensor) {
+    assertDtype(parseAsDtype, x.dtype, argName, functionName);
+    return x;
+  }
+  let inferredDtype = inferDtype(x);
+  if (inferredDtype !== "string" && ["bool", "int32", "float32"].indexOf(parseAsDtype) >= 0) {
+    inferredDtype = parseAsDtype;
+  }
+  assertDtype(parseAsDtype, inferredDtype, argName, functionName);
+  if (x == null || !isTypedArray(x) && !Array.isArray(x) && typeof x !== "number" && typeof x !== "boolean" && typeof x !== "string") {
+    const type = x == null ? "null" : x.constructor.name;
+    throw new Error(`Argument '${argName}' passed to '${functionName}' must be a Tensor or TensorLike, but got '${type}'`);
+  }
+  const inferredShape = inferShape(x, inferredDtype);
+  if (!isTypedArray(x) && !Array.isArray(x)) {
+    x = [x];
+  }
+  const skipTypedArray = true;
+  const values = inferredDtype !== "string" ? toTypedArray(x, inferredDtype) : flatten(x, [], skipTypedArray);
+  return ENGINE.makeTensor(values, inferredShape, inferredDtype);
+}
+function convertToTensorArray(arg, argName, functionName, parseAsDtype = "numeric") {
+  if (!Array.isArray(arg)) {
+    throw new Error(`Argument ${argName} passed to ${functionName} must be a \`Tensor[]\` or \`TensorLike[]\``);
+  }
+  const tensors = arg;
+  return tensors.map((t, i) => convertToTensor(t, `${argName}[${i}]`, functionName, parseAsDtype));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/operation.js
+var OP_SCOPE_SUFFIX = "__op";
+function op(f) {
+  const keys = Object.keys(f);
+  if (keys.length !== 1) {
+    throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${keys.length} keys.`);
+  }
+  let opName = keys[0];
+  const fn = f[opName];
+  if (opName.endsWith("_")) {
+    opName = opName.substring(0, opName.length - 1);
+  }
+  opName = opName + OP_SCOPE_SUFFIX;
+  const f2 = (...args) => {
+    ENGINE.startScope(opName);
+    try {
+      const result = fn(...args);
+      if (isPromise(result)) {
+        console.error("Cannot return a Promise inside of tidy.");
+      }
+      ENGINE.endScope(result);
+      return result;
+    } catch (ex) {
+      ENGINE.endScope(null);
+      throw ex;
+    }
+  };
+  Object.defineProperty(f2, "name", { value: opName, configurable: true });
+  return f2;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/complex.js
+function complex_(real4, imag4) {
+  const $real = convertToTensor(real4, "real", "complex");
+  const $imag = convertToTensor(imag4, "imag", "complex");
+  assertShapesMatch($real.shape, $imag.shape, `real and imag shapes, ${$real.shape} and ${$imag.shape}, must match in call to tf.complex().`);
+  const inputs = { real: $real, imag: $imag };
+  return ENGINE.runKernel(Complex, inputs);
+}
+var complex = op({ complex_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor_ops_util.js
+function makeTensor(values, shape, inferredShape, dtype) {
+  if (dtype == null) {
+    dtype = inferDtype(values);
+  }
+  if (dtype === "complex64") {
+    throw new Error(`Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).`);
+  }
+  if (typeof values === "object" && "texture" in values) {
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new Error(`Creating tensor from texture only supports 'float32'|'int32' dtype, while the dtype is ${dtype}.`);
+    }
+    values.channels = values.channels || "RGBA";
+    return ENGINE.backend.createTensorFromTexture(values, shape || inferredShape, dtype);
+  }
+  if (!isTypedArray(values) && !Array.isArray(values) && typeof values !== "number" && typeof values !== "boolean" && typeof values !== "string") {
+    throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");
+  }
+  if (shape != null) {
+    assertNonNegativeIntegerDimensions(shape);
+    const providedSize = sizeFromShape(shape);
+    const inferredSize = sizeFromShape(inferredShape);
+    assert(providedSize === inferredSize, () => `Based on the provided shape, [${shape}], the tensor should have ${providedSize} values but has ${inferredSize}`);
+    for (let i = 0; i < inferredShape.length; ++i) {
+      const inferred = inferredShape[i];
+      const flatDimsDontMatch = i === inferredShape.length - 1 ? inferred !== sizeFromShape(shape.slice(i)) : true;
+      assert(inferredShape[i] === shape[i] || !flatDimsDontMatch, () => `Error creating a new Tensor. Inferred shape (${inferredShape}) does not match the provided shape (${shape}). `);
+    }
+  }
+  if (!isTypedArray(values) && !Array.isArray(values)) {
+    values = [values];
+  }
+  shape = shape || inferredShape;
+  values = dtype !== "string" ? toTypedArray(values, dtype) : flatten(values, [], true);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor.js
+function tensor(values, shape, dtype) {
+  const inferredShape = inferShape(values, dtype);
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/types.js
+var DTYPE_VALUE_SIZE_MAP = {
+  "float32": 4,
+  "float16": 2,
+  "int32": 4,
+  "uint16": 2,
+  "uint8": 1,
+  "bool": 1,
+  "complex64": 8
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/io/io_utils.js
+var NUM_BYTES_STRING_LENGTH = 4;
+async function encodeWeights(tensors, group) {
+  const specs = [];
+  const dataPromises = [];
+  const names = Array.isArray(tensors) ? tensors.map((tensor2) => tensor2.name) : Object.keys(tensors);
+  for (let i = 0; i < names.length; ++i) {
+    const name = names[i];
+    const t = Array.isArray(tensors) ? tensors[i].tensor : tensors[name];
+    if (t.dtype !== "float32" && t.dtype !== "int32" && t.dtype !== "bool" && t.dtype !== "string" && t.dtype !== "complex64") {
+      throw new Error(`Unsupported dtype in weight '${name}': ${t.dtype}`);
+    }
+    const spec = { name, shape: t.shape, dtype: t.dtype };
+    if (t.dtype === "string") {
+      const utf8bytes = new Promise(async (resolve) => {
+        const vals = await t.bytes();
+        const totalNumBytes = vals.reduce((p2, c) => p2 + c.length, 0) + NUM_BYTES_STRING_LENGTH * vals.length;
+        const bytes = new Uint8Array(totalNumBytes);
+        let offset = 0;
+        for (let i2 = 0; i2 < vals.length; i2++) {
+          const val = vals[i2];
+          const bytesOfLength = new Uint8Array(new Uint32Array([val.length]).buffer);
+          bytes.set(bytesOfLength, offset);
+          offset += NUM_BYTES_STRING_LENGTH;
+          bytes.set(val, offset);
+          offset += val.length;
+        }
+        resolve(bytes);
+      });
+      dataPromises.push(utf8bytes);
+    } else {
+      dataPromises.push(t.data());
+    }
+    if (group != null) {
+      spec.group = group;
+    }
+    specs.push(spec);
+  }
+  const tensorValues = await Promise.all(dataPromises);
+  return { data: concatenateTypedArrays(tensorValues), specs };
+}
+function decodeWeights(buffer2, specs) {
+  const out = {};
+  let float16Decode;
+  let offset = 0;
+  for (const spec of specs) {
+    const name = spec.name;
+    const dtype = spec.dtype;
+    const shape = spec.shape;
+    const size = sizeFromShape(shape);
+    let values;
+    if ("quantization" in spec) {
+      const quantization = spec.quantization;
+      if (quantization.dtype === "uint8" || quantization.dtype === "uint16") {
+        if (!("min" in quantization && "scale" in quantization)) {
+          throw new Error(`Weight ${spec.name} with quantization ${quantization.dtype} doesn't have corresponding metadata min and scale.`);
+        }
+      } else if (quantization.dtype === "float16") {
+        if (dtype !== "float32") {
+          throw new Error(`Weight ${spec.name} is quantized with ${quantization.dtype} which only supports weights of type float32 not ${dtype}.`);
+        }
+      } else {
+        throw new Error(`Weight ${spec.name} has unknown quantization dtype ${quantization.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);
+      }
+      const quantizationSizeFactor = DTYPE_VALUE_SIZE_MAP[quantization.dtype];
+      const byteBuffer = buffer2.slice(offset, offset + size * quantizationSizeFactor);
+      const quantizedArray = quantization.dtype === "uint8" ? new Uint8Array(byteBuffer) : new Uint16Array(byteBuffer);
+      if (dtype === "float32") {
+        if (quantization.dtype === "uint8" || quantization.dtype === "uint16") {
+          values = new Float32Array(quantizedArray.length);
+          for (let i = 0; i < quantizedArray.length; i++) {
+            const v = quantizedArray[i];
+            values[i] = v * quantization.scale + quantization.min;
+          }
+        } else if (quantization.dtype === "float16") {
+          if (float16Decode === void 0) {
+            float16Decode = getFloat16Decoder();
+          }
+          values = float16Decode(quantizedArray);
+        } else {
+          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type float32.`);
+        }
+      } else if (dtype === "int32") {
+        if (quantization.dtype !== "uint8" && quantization.dtype !== "uint16") {
+          throw new Error(`Unsupported quantization type ${quantization.dtype} for weight type int32.`);
+        }
+        values = new Int32Array(quantizedArray.length);
+        for (let i = 0; i < quantizedArray.length; i++) {
+          const v = quantizedArray[i];
+          values[i] = Math.round(v * quantization.scale + quantization.min);
+        }
+      } else {
+        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);
+      }
+      offset += size * quantizationSizeFactor;
+    } else if (dtype === "string") {
+      const size2 = sizeFromShape(spec.shape);
+      values = [];
+      for (let i = 0; i < size2; i++) {
+        const byteLength = new Uint32Array(buffer2.slice(offset, offset + NUM_BYTES_STRING_LENGTH))[0];
+        offset += NUM_BYTES_STRING_LENGTH;
+        const bytes = new Uint8Array(buffer2.slice(offset, offset + byteLength));
+        values.push(bytes);
+        offset += byteLength;
+      }
+    } else {
+      const dtypeFactor = DTYPE_VALUE_SIZE_MAP[dtype];
+      const byteBuffer = buffer2.slice(offset, offset + size * dtypeFactor);
+      if (dtype === "float32") {
+        values = new Float32Array(byteBuffer);
+      } else if (dtype === "int32") {
+        values = new Int32Array(byteBuffer);
+      } else if (dtype === "bool") {
+        values = new Uint8Array(byteBuffer);
+      } else if (dtype === "complex64") {
+        values = new Float32Array(byteBuffer);
+        const real4 = new Float32Array(values.length / 2);
+        const image2 = new Float32Array(values.length / 2);
+        for (let i = 0; i < real4.length; i++) {
+          real4[i] = values[i * 2];
+          image2[i] = values[i * 2 + 1];
+        }
+        const realTensor = tensor(real4, shape, "float32");
+        const imageTensor = tensor(image2, shape, "float32");
+        out[name] = complex(realTensor, imageTensor);
+        realTensor.dispose();
+        imageTensor.dispose();
+      } else {
+        throw new Error(`Unsupported dtype in weight '${name}': ${dtype}`);
+      }
+      offset += size * dtypeFactor;
+    }
+    if (dtype !== "complex64") {
+      out[name] = tensor(values, shape, dtype);
+    }
+  }
+  return out;
+}
+function concatenateTypedArrays(xs) {
+  if (xs === null) {
+    throw new Error(`Invalid input value: ${JSON.stringify(xs)}`);
+  }
+  let totalByteLength = 0;
+  const normalizedXs = [];
+  xs.forEach((x) => {
+    totalByteLength += x.byteLength;
+    normalizedXs.push(x.byteLength === x.buffer.byteLength ? x : new x.constructor(x));
+    if (!(x instanceof Float32Array || x instanceof Int32Array || x instanceof Uint8Array)) {
+      throw new Error(`Unsupported TypedArray subtype: ${x.constructor.name}`);
+    }
+  });
+  const y = new Uint8Array(totalByteLength);
+  let offset = 0;
+  normalizedXs.forEach((x) => {
+    y.set(new Uint8Array(x.buffer), offset);
+    offset += x.byteLength;
+  });
+  return y.buffer;
+}
+var useNodeBuffer = typeof Buffer !== "undefined" && (typeof Blob === "undefined" || typeof atob === "undefined" || typeof btoa === "undefined");
+function stringByteLength(str) {
+  if (useNodeBuffer) {
+    return Buffer.byteLength(str);
+  }
+  return new Blob([str]).size;
+}
+function arrayBufferToBase64String(buffer2) {
+  if (useNodeBuffer) {
+    return Buffer.from(buffer2).toString("base64");
+  }
+  const buf = new Uint8Array(buffer2);
+  let s = "";
+  for (let i = 0, l = buf.length; i < l; i++) {
+    s += String.fromCharCode(buf[i]);
+  }
+  return btoa(s);
+}
+function base64StringToArrayBuffer(str) {
+  if (useNodeBuffer) {
+    const buf = Buffer.from(str, "base64");
+    return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);
+  }
+  const s = atob(str);
+  const buffer2 = new Uint8Array(s.length);
+  for (let i = 0; i < s.length; ++i) {
+    buffer2.set([s.charCodeAt(i)], i);
+  }
+  return buffer2.buffer;
+}
+function concatenateArrayBuffers(buffers) {
+  if (buffers.length === 1) {
+    return buffers[0];
+  }
+  let totalByteLength = 0;
+  buffers.forEach((buffer2) => {
+    totalByteLength += buffer2.byteLength;
+  });
+  const temp = new Uint8Array(totalByteLength);
+  let offset = 0;
+  buffers.forEach((buffer2) => {
+    temp.set(new Uint8Array(buffer2), offset);
+    offset += buffer2.byteLength;
+  });
+  return temp.buffer;
+}
+function basename(path) {
+  const SEPARATOR = "/";
+  path = path.trim();
+  while (path.endsWith(SEPARATOR)) {
+    path = path.slice(0, path.length - 1);
+  }
+  const items = path.split(SEPARATOR);
+  return items[items.length - 1];
+}
+function getModelJSONForModelArtifacts(artifacts, manifest) {
+  const result = {
+    modelTopology: artifacts.modelTopology,
+    format: artifacts.format,
+    generatedBy: artifacts.generatedBy,
+    convertedBy: artifacts.convertedBy,
+    weightsManifest: manifest
+  };
+  if (artifacts.signature != null) {
+    result.signature = artifacts.signature;
+  }
+  if (artifacts.userDefinedMetadata != null) {
+    result.userDefinedMetadata = artifacts.userDefinedMetadata;
+  }
+  if (artifacts.modelInitializer != null) {
+    result.modelInitializer = artifacts.modelInitializer;
+  }
+  if (artifacts.initializerSignature != null) {
+    result.initializerSignature = artifacts.initializerSignature;
+  }
+  if (artifacts.trainingConfig != null) {
+    result.trainingConfig = artifacts.trainingConfig;
+  }
+  return result;
+}
+function getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData) {
+  const modelArtifacts = {
+    modelTopology: modelJSON.modelTopology,
+    format: modelJSON.format,
+    generatedBy: modelJSON.generatedBy,
+    convertedBy: modelJSON.convertedBy
+  };
+  if (modelJSON.trainingConfig != null) {
+    modelArtifacts.trainingConfig = modelJSON.trainingConfig;
+  }
+  if (modelJSON.weightsManifest != null) {
+    if (!weightSpecs) {
+      throw new Error("modelJSON has weightsManifest but weightSpecs is null");
+    }
+    if (!weightData) {
+      throw new Error("modelJSON has weightsManifest but weightData is null");
+    }
+    modelArtifacts.weightSpecs = weightSpecs;
+    modelArtifacts.weightData = weightData;
+  }
+  if (modelJSON.signature != null) {
+    modelArtifacts.signature = modelJSON.signature;
+  }
+  if (modelJSON.userDefinedMetadata != null) {
+    modelArtifacts.userDefinedMetadata = modelJSON.userDefinedMetadata;
+  }
+  if (modelJSON.modelInitializer != null) {
+    modelArtifacts.modelInitializer = modelJSON.modelInitializer;
+  }
+  if (modelJSON.initializerSignature != null) {
+    modelArtifacts.initializerSignature = modelJSON.initializerSignature;
+  }
+  return modelArtifacts;
+}
+async function getModelArtifactsForJSON(modelJSON, loadWeights2) {
+  let weightSpecs;
+  let weightData;
+  if (modelJSON.weightsManifest != null) {
+    [weightSpecs, weightData] = await loadWeights2(modelJSON.weightsManifest);
+  }
+  return getModelArtifactsForJSONSync(modelJSON, weightSpecs, weightData);
+}
+function getModelArtifactsInfoForJSON(modelArtifacts) {
+  if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+    throw new Error("Expected JSON model topology, received ArrayBuffer.");
+  }
+  return {
+    dateSaved: new Date(),
+    modelTopologyType: "JSON",
+    modelTopologyBytes: modelArtifacts.modelTopology == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.modelTopology)),
+    weightSpecsBytes: modelArtifacts.weightSpecs == null ? 0 : stringByteLength(JSON.stringify(modelArtifacts.weightSpecs)),
+    weightDataBytes: modelArtifacts.weightData == null ? 0 : modelArtifacts.weightData.byteLength
+  };
+}
+function getWeightSpecs(weightsManifest) {
+  const weightSpecs = [];
+  for (const entry of weightsManifest) {
+    weightSpecs.push(...entry.weights);
+  }
+  return weightSpecs;
+}
+function computeFloat16MantisaTable() {
+  const convertMantissa = (i) => {
+    let m = i << 13;
+    let e = 0;
+    while ((m & 8388608) === 0) {
+      e -= 8388608;
+      m <<= 1;
+    }
+    m &= ~8388608;
+    e += 947912704;
+    return m | e;
+  };
+  const mantisaTable = new Uint32Array(2048);
+  mantisaTable[0] = 0;
+  for (let i = 1; i < 1024; i++) {
+    mantisaTable[i] = convertMantissa(i);
+  }
+  for (let i = 1024; i < 2048; i++) {
+    mantisaTable[i] = 939524096 + (i - 1024 << 13);
+  }
+  return mantisaTable;
+}
+function computeFloat16ExponentTable() {
+  const exponentTable = new Uint32Array(64);
+  exponentTable[0] = 0;
+  exponentTable[31] = 1199570944;
+  exponentTable[32] = 2147483648;
+  exponentTable[63] = 3347054592;
+  for (let i = 1; i < 31; i++) {
+    exponentTable[i] = i << 23;
+  }
+  for (let i = 33; i < 63; i++) {
+    exponentTable[i] = 2147483648 + (i - 32 << 23);
+  }
+  return exponentTable;
+}
+function computeFloat16OffsetTable() {
+  const offsetTable = new Uint32Array(64);
+  for (let i = 0; i < 64; i++) {
+    offsetTable[i] = 1024;
+  }
+  offsetTable[0] = offsetTable[32] = 0;
+  return offsetTable;
+}
+function getFloat16Decoder() {
+  const mantisaTable = computeFloat16MantisaTable();
+  const exponentTable = computeFloat16ExponentTable();
+  const offsetTable = computeFloat16OffsetTable();
+  return (quantizedArray) => {
+    const buffer2 = new ArrayBuffer(4 * quantizedArray.length);
+    const bufferUint32View = new Uint32Array(buffer2);
+    for (let index = 0; index < quantizedArray.length; index++) {
+      const float16Bits = quantizedArray[index];
+      const float32Bits = mantisaTable[offsetTable[float16Bits >> 10] + (float16Bits & 1023)] + exponentTable[float16Bits >> 10];
+      bufferUint32View[index] = float32Bits;
+    }
+    return new Float32Array(buffer2);
+  };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/router_registry.js
+var IORouterRegistry = class {
+  constructor() {
+    this.saveRouters = [];
+    this.loadRouters = [];
+  }
+  static getInstance() {
+    if (IORouterRegistry.instance == null) {
+      IORouterRegistry.instance = new IORouterRegistry();
+    }
+    return IORouterRegistry.instance;
+  }
+  static registerSaveRouter(saveRouter) {
+    IORouterRegistry.getInstance().saveRouters.push(saveRouter);
+  }
+  static registerLoadRouter(loadRouter) {
+    IORouterRegistry.getInstance().loadRouters.push(loadRouter);
+  }
+  static getSaveHandlers(url) {
+    return IORouterRegistry.getHandlers(url, "save");
+  }
+  static getLoadHandlers(url, loadOptions) {
+    return IORouterRegistry.getHandlers(url, "load", loadOptions);
+  }
+  static getHandlers(url, handlerType, loadOptions) {
+    const validHandlers = [];
+    const routers = handlerType === "load" ? IORouterRegistry.getInstance().loadRouters : IORouterRegistry.getInstance().saveRouters;
+    routers.forEach((router) => {
+      const handler = router(url, loadOptions);
+      if (handler !== null) {
+        validHandlers.push(handler);
+      }
+    });
+    return validHandlers;
+  }
+};
+var registerSaveRouter = (loudRouter) => IORouterRegistry.registerSaveRouter(loudRouter);
+var registerLoadRouter = (loudRouter) => IORouterRegistry.registerLoadRouter(loudRouter);
+var getSaveHandlers = (url) => IORouterRegistry.getSaveHandlers(url);
+var getLoadHandlers = (url, loadOptions) => IORouterRegistry.getLoadHandlers(url, loadOptions);
+
+// node_modules/@tensorflow/tfjs-core/dist/io/indexed_db.js
+var DATABASE_NAME = "tensorflowjs";
+var DATABASE_VERSION = 1;
+var MODEL_STORE_NAME = "models_store";
+var INFO_STORE_NAME = "model_info_store";
+function getIndexedDBFactory() {
+  if (!env().getBool("IS_BROWSER")) {
+    throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");
+  }
+  const theWindow = typeof window === "undefined" ? self : window;
+  const factory = theWindow.indexedDB || theWindow.mozIndexedDB || theWindow.webkitIndexedDB || theWindow.msIndexedDB || theWindow.shimIndexedDB;
+  if (factory == null) {
+    throw new Error("The current browser does not appear to support IndexedDB.");
+  }
+  return factory;
+}
+function setUpDatabase(openRequest) {
+  const db = openRequest.result;
+  db.createObjectStore(MODEL_STORE_NAME, { keyPath: "modelPath" });
+  db.createObjectStore(INFO_STORE_NAME, { keyPath: "modelPath" });
+}
+var BrowserIndexedDB = class {
+  constructor(modelPath) {
+    this.indexedDB = getIndexedDBFactory();
+    if (modelPath == null || !modelPath) {
+      throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");
+    }
+    this.modelPath = modelPath;
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");
+    }
+    return this.databaseAction(this.modelPath, modelArtifacts);
+  }
+  async load() {
+    return this.databaseAction(this.modelPath);
+  }
+  databaseAction(modelPath, modelArtifacts) {
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        if (modelArtifacts == null) {
+          const modelTx = db.transaction(MODEL_STORE_NAME, "readonly");
+          const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+          const getRequest = modelStore.get(this.modelPath);
+          getRequest.onsuccess = () => {
+            if (getRequest.result == null) {
+              db.close();
+              return reject(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));
+            } else {
+              resolve(getRequest.result.modelArtifacts);
+            }
+          };
+          getRequest.onerror = (error) => {
+            db.close();
+            return reject(getRequest.error);
+          };
+          modelTx.oncomplete = () => db.close();
+        } else {
+          const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);
+          const infoTx = db.transaction(INFO_STORE_NAME, "readwrite");
+          let infoStore = infoTx.objectStore(INFO_STORE_NAME);
+          const putInfoRequest = infoStore.put({ modelPath: this.modelPath, modelArtifactsInfo });
+          let modelTx;
+          putInfoRequest.onsuccess = () => {
+            modelTx = db.transaction(MODEL_STORE_NAME, "readwrite");
+            const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+            const putModelRequest = modelStore.put({
+              modelPath: this.modelPath,
+              modelArtifacts,
+              modelArtifactsInfo
+            });
+            putModelRequest.onsuccess = () => resolve({ modelArtifactsInfo });
+            putModelRequest.onerror = (error) => {
+              infoStore = infoTx.objectStore(INFO_STORE_NAME);
+              const deleteInfoRequest = infoStore.delete(this.modelPath);
+              deleteInfoRequest.onsuccess = () => {
+                db.close();
+                return reject(putModelRequest.error);
+              };
+              deleteInfoRequest.onerror = (error2) => {
+                db.close();
+                return reject(putModelRequest.error);
+              };
+            };
+          };
+          putInfoRequest.onerror = (error) => {
+            db.close();
+            return reject(putInfoRequest.error);
+          };
+          infoTx.oncomplete = () => {
+            if (modelTx == null) {
+              db.close();
+            } else {
+              modelTx.oncomplete = () => db.close();
+            }
+          };
+        }
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+};
+BrowserIndexedDB.URL_SCHEME = "indexeddb://";
+var indexedDBRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserIndexedDB.URL_SCHEME)) {
+      return browserIndexedDB(url.slice(BrowserIndexedDB.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(indexedDBRouter);
+IORouterRegistry.registerLoadRouter(indexedDBRouter);
+function browserIndexedDB(modelPath) {
+  return new BrowserIndexedDB(modelPath);
+}
+function maybeStripScheme(key) {
+  return key.startsWith(BrowserIndexedDB.URL_SCHEME) ? key.slice(BrowserIndexedDB.URL_SCHEME.length) : key;
+}
+var BrowserIndexedDBManager = class {
+  constructor() {
+    this.indexedDB = getIndexedDBFactory();
+  }
+  async listModels() {
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        const tx = db.transaction(INFO_STORE_NAME, "readonly");
+        const store = tx.objectStore(INFO_STORE_NAME);
+        const getAllInfoRequest = store.getAll();
+        getAllInfoRequest.onsuccess = () => {
+          const out = {};
+          for (const item of getAllInfoRequest.result) {
+            out[item.modelPath] = item.modelArtifactsInfo;
+          }
+          resolve(out);
+        };
+        getAllInfoRequest.onerror = (error) => {
+          db.close();
+          return reject(getAllInfoRequest.error);
+        };
+        tx.oncomplete = () => db.close();
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+  async removeModel(path) {
+    path = maybeStripScheme(path);
+    return new Promise((resolve, reject) => {
+      const openRequest = this.indexedDB.open(DATABASE_NAME, DATABASE_VERSION);
+      openRequest.onupgradeneeded = () => setUpDatabase(openRequest);
+      openRequest.onsuccess = () => {
+        const db = openRequest.result;
+        const infoTx = db.transaction(INFO_STORE_NAME, "readwrite");
+        const infoStore = infoTx.objectStore(INFO_STORE_NAME);
+        const getInfoRequest = infoStore.get(path);
+        let modelTx;
+        getInfoRequest.onsuccess = () => {
+          if (getInfoRequest.result == null) {
+            db.close();
+            return reject(new Error(`Cannot find model with path '${path}' in IndexedDB.`));
+          } else {
+            const deleteInfoRequest = infoStore.delete(path);
+            const deleteModelData = () => {
+              modelTx = db.transaction(MODEL_STORE_NAME, "readwrite");
+              const modelStore = modelTx.objectStore(MODEL_STORE_NAME);
+              const deleteModelRequest = modelStore.delete(path);
+              deleteModelRequest.onsuccess = () => resolve(getInfoRequest.result.modelArtifactsInfo);
+              deleteModelRequest.onerror = (error) => reject(getInfoRequest.error);
+            };
+            deleteInfoRequest.onsuccess = deleteModelData;
+            deleteInfoRequest.onerror = (error) => {
+              deleteModelData();
+              db.close();
+              return reject(getInfoRequest.error);
+            };
+          }
+        };
+        getInfoRequest.onerror = (error) => {
+          db.close();
+          return reject(getInfoRequest.error);
+        };
+        infoTx.oncomplete = () => {
+          if (modelTx == null) {
+            db.close();
+          } else {
+            modelTx.oncomplete = () => db.close();
+          }
+        };
+      };
+      openRequest.onerror = (error) => reject(openRequest.error);
+    });
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/io/local_storage.js
+var PATH_SEPARATOR = "/";
+var PATH_PREFIX = "tensorflowjs_models";
+var INFO_SUFFIX = "info";
+var MODEL_TOPOLOGY_SUFFIX = "model_topology";
+var WEIGHT_SPECS_SUFFIX = "weight_specs";
+var WEIGHT_DATA_SUFFIX = "weight_data";
+var MODEL_METADATA_SUFFIX = "model_metadata";
+function getModelKeys(path) {
+  return {
+    info: [PATH_PREFIX, path, INFO_SUFFIX].join(PATH_SEPARATOR),
+    topology: [PATH_PREFIX, path, MODEL_TOPOLOGY_SUFFIX].join(PATH_SEPARATOR),
+    weightSpecs: [PATH_PREFIX, path, WEIGHT_SPECS_SUFFIX].join(PATH_SEPARATOR),
+    weightData: [PATH_PREFIX, path, WEIGHT_DATA_SUFFIX].join(PATH_SEPARATOR),
+    modelMetadata: [PATH_PREFIX, path, MODEL_METADATA_SUFFIX].join(PATH_SEPARATOR)
+  };
+}
+function removeItems(keys) {
+  for (const key of Object.values(keys)) {
+    window.localStorage.removeItem(key);
+  }
+}
+function getModelPathFromKey(key) {
+  const items = key.split(PATH_SEPARATOR);
+  if (items.length < 3) {
+    throw new Error(`Invalid key format: ${key}`);
+  }
+  return items.slice(1, items.length - 1).join(PATH_SEPARATOR);
+}
+function maybeStripScheme2(key) {
+  return key.startsWith(BrowserLocalStorage.URL_SCHEME) ? key.slice(BrowserLocalStorage.URL_SCHEME.length) : key;
+}
+var BrowserLocalStorage = class {
+  constructor(modelPath) {
+    if (!env().getBool("IS_BROWSER") || typeof window === "undefined" || typeof window.localStorage === "undefined") {
+      throw new Error("The current environment does not support local storage.");
+    }
+    this.LS = window.localStorage;
+    if (modelPath == null || !modelPath) {
+      throw new Error("For local storage, modelPath must not be null, undefined or empty.");
+    }
+    this.modelPath = modelPath;
+    this.keys = getModelKeys(this.modelPath);
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");
+    } else {
+      const topology = JSON.stringify(modelArtifacts.modelTopology);
+      const weightSpecs = JSON.stringify(modelArtifacts.weightSpecs);
+      const modelArtifactsInfo = getModelArtifactsInfoForJSON(modelArtifacts);
+      try {
+        this.LS.setItem(this.keys.info, JSON.stringify(modelArtifactsInfo));
+        this.LS.setItem(this.keys.topology, topology);
+        this.LS.setItem(this.keys.weightSpecs, weightSpecs);
+        this.LS.setItem(this.keys.weightData, arrayBufferToBase64String(modelArtifacts.weightData));
+        const metadata = {
+          format: modelArtifacts.format,
+          generatedBy: modelArtifacts.generatedBy,
+          convertedBy: modelArtifacts.convertedBy,
+          signature: modelArtifacts.signature != null ? modelArtifacts.signature : void 0,
+          userDefinedMetadata: modelArtifacts.userDefinedMetadata != null ? modelArtifacts.userDefinedMetadata : void 0,
+          modelInitializer: modelArtifacts.modelInitializer != null ? modelArtifacts.modelInitializer : void 0,
+          initializerSignature: modelArtifacts.initializerSignature != null ? modelArtifacts.initializerSignature : void 0,
+          trainingConfig: modelArtifacts.trainingConfig != null ? modelArtifacts.trainingConfig : void 0
+        };
+        this.LS.setItem(this.keys.modelMetadata, JSON.stringify(metadata));
+        return { modelArtifactsInfo };
+      } catch (err) {
+        removeItems(this.keys);
+        throw new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${modelArtifactsInfo.modelTopologyBytes}, weightSpecsBytes=${modelArtifactsInfo.weightSpecsBytes}, weightDataBytes=${modelArtifactsInfo.weightDataBytes}.`);
+      }
+    }
+  }
+  async load() {
+    const info = JSON.parse(this.LS.getItem(this.keys.info));
+    if (info == null) {
+      throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);
+    }
+    if (info.modelTopologyType !== "JSON") {
+      throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");
+    }
+    const out = {};
+    const topology = JSON.parse(this.LS.getItem(this.keys.topology));
+    if (topology == null) {
+      throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);
+    }
+    out.modelTopology = topology;
+    const weightSpecs = JSON.parse(this.LS.getItem(this.keys.weightSpecs));
+    if (weightSpecs == null) {
+      throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);
+    }
+    out.weightSpecs = weightSpecs;
+    const metadataString = this.LS.getItem(this.keys.modelMetadata);
+    if (metadataString != null) {
+      const metadata = JSON.parse(metadataString);
+      out.format = metadata.format;
+      out.generatedBy = metadata.generatedBy;
+      out.convertedBy = metadata.convertedBy;
+      if (metadata.signature != null) {
+        out.signature = metadata.signature;
+      }
+      if (metadata.userDefinedMetadata != null) {
+        out.userDefinedMetadata = metadata.userDefinedMetadata;
+      }
+      if (metadata.modelInitializer != null) {
+        out.modelInitializer = metadata.modelInitializer;
+      }
+      if (metadata.initializerSignature != null) {
+        out.initializerSignature = metadata.initializerSignature;
+      }
+      if (metadata.trainingConfig != null) {
+        out.trainingConfig = metadata.trainingConfig;
+      }
+    }
+    const weightDataBase64 = this.LS.getItem(this.keys.weightData);
+    if (weightDataBase64 == null) {
+      throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);
+    }
+    out.weightData = base64StringToArrayBuffer(weightDataBase64);
+    return out;
+  }
+};
+BrowserLocalStorage.URL_SCHEME = "localstorage://";
+var localStorageRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserLocalStorage.URL_SCHEME)) {
+      return browserLocalStorage(url.slice(BrowserLocalStorage.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(localStorageRouter);
+IORouterRegistry.registerLoadRouter(localStorageRouter);
+function browserLocalStorage(modelPath) {
+  return new BrowserLocalStorage(modelPath);
+}
+var BrowserLocalStorageManager = class {
+  constructor() {
+    assert(env().getBool("IS_BROWSER"), () => "Current environment is not a web browser");
+    assert(typeof window === "undefined" || typeof window.localStorage !== "undefined", () => "Current browser does not appear to support localStorage");
+    this.LS = window.localStorage;
+  }
+  async listModels() {
+    const out = {};
+    const prefix = PATH_PREFIX + PATH_SEPARATOR;
+    const suffix = PATH_SEPARATOR + INFO_SUFFIX;
+    for (let i = 0; i < this.LS.length; ++i) {
+      const key = this.LS.key(i);
+      if (key.startsWith(prefix) && key.endsWith(suffix)) {
+        const modelPath = getModelPathFromKey(key);
+        out[modelPath] = JSON.parse(this.LS.getItem(key));
+      }
+    }
+    return out;
+  }
+  async removeModel(path) {
+    path = maybeStripScheme2(path);
+    const keys = getModelKeys(path);
+    if (this.LS.getItem(keys.info) == null) {
+      throw new Error(`Cannot find model at path '${path}'`);
+    }
+    const info = JSON.parse(this.LS.getItem(keys.info));
+    removeItems(keys);
+    return info;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/io/model_management.js
+var URL_SCHEME_SUFFIX = "://";
+var ModelStoreManagerRegistry = class {
+  constructor() {
+    this.managers = {};
+  }
+  static getInstance() {
+    if (ModelStoreManagerRegistry.instance == null) {
+      ModelStoreManagerRegistry.instance = new ModelStoreManagerRegistry();
+    }
+    return ModelStoreManagerRegistry.instance;
+  }
+  static registerManager(scheme, manager) {
+    assert(scheme != null, () => "scheme must not be undefined or null.");
+    if (scheme.endsWith(URL_SCHEME_SUFFIX)) {
+      scheme = scheme.slice(0, scheme.indexOf(URL_SCHEME_SUFFIX));
+    }
+    assert(scheme.length > 0, () => "scheme must not be an empty string.");
+    const registry = ModelStoreManagerRegistry.getInstance();
+    assert(registry.managers[scheme] == null, () => `A model store manager is already registered for scheme '${scheme}'.`);
+    registry.managers[scheme] = manager;
+  }
+  static getManager(scheme) {
+    const manager = ModelStoreManagerRegistry.getInstance().managers[scheme];
+    if (manager == null) {
+      throw new Error(`Cannot find model manager for scheme '${scheme}'`);
+    }
+    return manager;
+  }
+  static getSchemes() {
+    return Object.keys(ModelStoreManagerRegistry.getInstance().managers);
+  }
+};
+function parseURL(url) {
+  if (url.indexOf(URL_SCHEME_SUFFIX) === -1) {
+    throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${ModelStoreManagerRegistry.getSchemes().join(",")}`);
+  }
+  return {
+    scheme: url.split(URL_SCHEME_SUFFIX)[0],
+    path: url.split(URL_SCHEME_SUFFIX)[1]
+  };
+}
+async function cloneModelInternal(sourceURL, destURL, deleteSource = false) {
+  assert(sourceURL !== destURL, () => `Old path and new path are the same: '${sourceURL}'`);
+  const loadHandlers = IORouterRegistry.getLoadHandlers(sourceURL);
+  assert(loadHandlers.length > 0, () => `Copying failed because no load handler is found for source URL ${sourceURL}.`);
+  assert(loadHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) load handlers for source URL ${sourceURL}.`);
+  const loadHandler = loadHandlers[0];
+  const saveHandlers = IORouterRegistry.getSaveHandlers(destURL);
+  assert(saveHandlers.length > 0, () => `Copying failed because no save handler is found for destination URL ${destURL}.`);
+  assert(saveHandlers.length < 2, () => `Copying failed because more than one (${loadHandlers.length}) save handlers for destination URL ${destURL}.`);
+  const saveHandler = saveHandlers[0];
+  const sourceScheme = parseURL(sourceURL).scheme;
+  const sourcePath = parseURL(sourceURL).path;
+  const sameMedium = sourceScheme === parseURL(sourceURL).scheme;
+  const modelArtifacts = await loadHandler.load();
+  if (deleteSource && sameMedium) {
+    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);
+  }
+  const saveResult = await saveHandler.save(modelArtifacts);
+  if (deleteSource && !sameMedium) {
+    await ModelStoreManagerRegistry.getManager(sourceScheme).removeModel(sourcePath);
+  }
+  return saveResult.modelArtifactsInfo;
+}
+async function listModels() {
+  const schemes = ModelStoreManagerRegistry.getSchemes();
+  const out = {};
+  for (const scheme of schemes) {
+    const schemeOut = await ModelStoreManagerRegistry.getManager(scheme).listModels();
+    for (const path in schemeOut) {
+      const url = scheme + URL_SCHEME_SUFFIX + path;
+      out[url] = schemeOut[path];
+    }
+  }
+  return out;
+}
+async function removeModel(url) {
+  const schemeAndPath = parseURL(url);
+  const manager = ModelStoreManagerRegistry.getManager(schemeAndPath.scheme);
+  return manager.removeModel(schemeAndPath.path);
+}
+async function copyModel(sourceURL, destURL) {
+  const deleteSource = false;
+  return cloneModelInternal(sourceURL, destURL, deleteSource);
+}
+async function moveModel(sourceURL, destURL) {
+  const deleteSource = true;
+  return cloneModelInternal(sourceURL, destURL, deleteSource);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/platforms/platform_browser.js
+var PlatformBrowser = class {
+  constructor() {
+    this.messageName = "setTimeoutCustom";
+    this.functionRefs = [];
+    this.handledMessageCount = 0;
+    this.hasEventListener = false;
+  }
+  fetch(path, init2) {
+    return fetch(path, init2);
+  }
+  now() {
+    return performance.now();
+  }
+  encode(text, encoding) {
+    if (encoding !== "utf-8" && encoding !== "utf8") {
+      throw new Error(`Browser's encoder only supports utf-8, but got ${encoding}`);
+    }
+    if (this.textEncoder == null) {
+      this.textEncoder = new TextEncoder();
+    }
+    return this.textEncoder.encode(text);
+  }
+  decode(bytes, encoding) {
+    return new TextDecoder(encoding).decode(bytes);
+  }
+  setTimeoutCustom(functionRef, delay) {
+    if (typeof window === "undefined" || !env().getBool("USE_SETTIMEOUTCUSTOM")) {
+      setTimeout(functionRef, delay);
+      return;
+    }
+    this.functionRefs.push(functionRef);
+    setTimeout(() => {
+      window.postMessage({ name: this.messageName, index: this.functionRefs.length - 1 }, "*");
+    }, delay);
+    if (!this.hasEventListener) {
+      this.hasEventListener = true;
+      window.addEventListener("message", (event) => {
+        if (event.source === window && event.data.name === this.messageName) {
+          event.stopPropagation();
+          const functionRef2 = this.functionRefs[event.data.index];
+          functionRef2();
+          this.handledMessageCount++;
+          if (this.handledMessageCount === this.functionRefs.length) {
+            this.functionRefs = [];
+            this.handledMessageCount = 0;
+          }
+        }
+      }, true);
+    }
+  }
+};
+if (env().get("IS_BROWSER")) {
+  env().setPlatform("browser", new PlatformBrowser());
+  try {
+    ModelStoreManagerRegistry.registerManager(BrowserLocalStorage.URL_SCHEME, new BrowserLocalStorageManager());
+  } catch (err) {
+  }
+  try {
+    ModelStoreManagerRegistry.registerManager(BrowserIndexedDB.URL_SCHEME, new BrowserIndexedDBManager());
+  } catch (err) {
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/platforms/platform_node.js
+var getNodeFetch = {
+  importFetch: () => require_browser()
+};
+var systemFetch;
+var PlatformNode = class {
+  constructor() {
+    this.util = require_util();
+    this.textEncoder = new this.util.TextEncoder();
+  }
+  fetch(path, requestInits) {
+    if (env().global.fetch != null) {
+      return env().global.fetch(path, requestInits);
+    }
+    if (systemFetch == null) {
+      systemFetch = getNodeFetch.importFetch();
+    }
+    return systemFetch(path, requestInits);
+  }
+  now() {
+    const time2 = process.hrtime();
+    return time2[0] * 1e3 + time2[1] / 1e6;
+  }
+  encode(text, encoding) {
+    if (encoding !== "utf-8" && encoding !== "utf8") {
+      throw new Error(`Node built-in encoder only supports utf-8, but got ${encoding}`);
+    }
+    return this.textEncoder.encode(text);
+  }
+  decode(bytes, encoding) {
+    if (bytes.length === 0) {
+      return "";
+    }
+    return new this.util.TextDecoder(encoding).decode(bytes);
+  }
+};
+if (env().get("IS_NODE") && !env().get("IS_BROWSER")) {
+  env().setPlatform("node", new PlatformNode());
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/buffer.js
+function buffer(shape, dtype = "float32", values) {
+  dtype = dtype || "float32";
+  assertNonNegativeIntegerDimensions(shape);
+  return new TensorBuffer(shape, dtype, values);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/cast.js
+function cast_(x, dtype) {
+  const $x = convertToTensor(x, "x", "cast");
+  if (!isValidDtype(dtype)) {
+    throw new Error(`Failed to cast to unknown dtype ${dtype}`);
+  }
+  if (dtype === "string" && $x.dtype !== "string" || dtype !== "string" && $x.dtype === "string") {
+    throw new Error("Only strings can be casted to strings");
+  }
+  const inputs = { x: $x };
+  const attrs = { dtype };
+  return ENGINE.runKernel(Cast, inputs, attrs);
+}
+var cast = op({ cast_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/clone.js
+function clone_(x) {
+  const $x = convertToTensor(x, "x", "clone", "string_or_numeric");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Identity, inputs);
+}
+var clone = op({ clone_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/print.js
+function print(x, verbose = false) {
+  console.log(x.toString(verbose));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/base_side_effects.js
+getOrMakeEngine();
+var opHandler2 = {
+  buffer,
+  cast,
+  clone,
+  print
+};
+setOpHandler(opHandler2);
+
+// node_modules/@tensorflow/tfjs-core/dist/io/io.js
+var io_exports = {};
+__export(io_exports, {
+  browserFiles: () => browserFiles,
+  browserHTTPRequest: () => browserHTTPRequest,
+  concatenateArrayBuffers: () => concatenateArrayBuffers,
+  copyModel: () => copyModel,
+  decodeWeights: () => decodeWeights,
+  encodeWeights: () => encodeWeights,
+  fromMemory: () => fromMemory,
+  fromMemorySync: () => fromMemorySync,
+  getLoadHandlers: () => getLoadHandlers,
+  getModelArtifactsForJSON: () => getModelArtifactsForJSON,
+  getModelArtifactsForJSONSync: () => getModelArtifactsForJSONSync,
+  getModelArtifactsInfoForJSON: () => getModelArtifactsInfoForJSON,
+  getSaveHandlers: () => getSaveHandlers,
+  getWeightSpecs: () => getWeightSpecs,
+  http: () => http,
+  isHTTPScheme: () => isHTTPScheme,
+  listModels: () => listModels,
+  loadWeights: () => loadWeights,
+  moveModel: () => moveModel,
+  registerLoadRouter: () => registerLoadRouter,
+  registerSaveRouter: () => registerSaveRouter,
+  removeModel: () => removeModel,
+  weightsLoaderFactory: () => weightsLoaderFactory,
+  withSaveHandler: () => withSaveHandler,
+  withSaveHandlerSync: () => withSaveHandlerSync
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/io/browser_files.js
+var DEFAULT_FILE_NAME_PREFIX = "model";
+var DEFAULT_JSON_EXTENSION_NAME = ".json";
+var DEFAULT_WEIGHT_DATA_EXTENSION_NAME = ".weights.bin";
+function defer(f) {
+  return new Promise((resolve) => setTimeout(resolve)).then(f);
+}
+var BrowserDownloads = class {
+  constructor(fileNamePrefix) {
+    if (!env().getBool("IS_BROWSER")) {
+      throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");
+    }
+    if (fileNamePrefix.startsWith(BrowserDownloads.URL_SCHEME)) {
+      fileNamePrefix = fileNamePrefix.slice(BrowserDownloads.URL_SCHEME.length);
+    }
+    if (fileNamePrefix == null || fileNamePrefix.length === 0) {
+      fileNamePrefix = DEFAULT_FILE_NAME_PREFIX;
+    }
+    this.modelJsonFileName = fileNamePrefix + DEFAULT_JSON_EXTENSION_NAME;
+    this.weightDataFileName = fileNamePrefix + DEFAULT_WEIGHT_DATA_EXTENSION_NAME;
+  }
+  async save(modelArtifacts) {
+    if (typeof document === "undefined") {
+      throw new Error("Browser downloads are not supported in this environment since `document` is not present");
+    }
+    const weightsURL = window.URL.createObjectURL(new Blob([modelArtifacts.weightData], { type: "application/octet-stream" }));
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");
+    } else {
+      const weightsManifest = [{
+        paths: ["./" + this.weightDataFileName],
+        weights: modelArtifacts.weightSpecs
+      }];
+      const modelJSON = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);
+      const modelJsonURL = window.URL.createObjectURL(new Blob([JSON.stringify(modelJSON)], { type: "application/json" }));
+      const jsonAnchor = this.modelJsonAnchor == null ? document.createElement("a") : this.modelJsonAnchor;
+      jsonAnchor.download = this.modelJsonFileName;
+      jsonAnchor.href = modelJsonURL;
+      await defer(() => jsonAnchor.dispatchEvent(new MouseEvent("click")));
+      if (modelArtifacts.weightData != null) {
+        const weightDataAnchor = this.weightDataAnchor == null ? document.createElement("a") : this.weightDataAnchor;
+        weightDataAnchor.download = this.weightDataFileName;
+        weightDataAnchor.href = weightsURL;
+        await defer(() => weightDataAnchor.dispatchEvent(new MouseEvent("click")));
+      }
+      return { modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts) };
+    }
+  }
+};
+BrowserDownloads.URL_SCHEME = "downloads://";
+var BrowserFiles = class {
+  constructor(files) {
+    if (files == null || files.length < 1) {
+      throw new Error(`When calling browserFiles, at least 1 file is required, but received ${files}`);
+    }
+    this.jsonFile = files[0];
+    this.weightsFiles = files.slice(1);
+  }
+  async load() {
+    return new Promise((resolve, reject) => {
+      const jsonReader = new FileReader();
+      jsonReader.onload = (event) => {
+        const modelJSON = JSON.parse(event.target.result);
+        const modelTopology = modelJSON.modelTopology;
+        if (modelTopology == null) {
+          reject(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));
+          return;
+        }
+        const weightsManifest = modelJSON.weightsManifest;
+        if (weightsManifest == null) {
+          reject(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));
+          return;
+        }
+        if (this.weightsFiles.length === 0) {
+          resolve({ modelTopology });
+          return;
+        }
+        const modelArtifactsPromise = getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));
+        resolve(modelArtifactsPromise);
+      };
+      jsonReader.onerror = (error) => reject(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`);
+      jsonReader.readAsText(this.jsonFile);
+    });
+  }
+  loadWeights(weightsManifest) {
+    const weightSpecs = [];
+    const paths = [];
+    for (const entry of weightsManifest) {
+      weightSpecs.push(...entry.weights);
+      paths.push(...entry.paths);
+    }
+    const pathToFile = this.checkManifestAndWeightFiles(weightsManifest);
+    const promises = paths.map((path) => this.loadWeightsFile(path, pathToFile[path]));
+    return Promise.all(promises).then((buffers) => [weightSpecs, concatenateArrayBuffers(buffers)]);
+  }
+  loadWeightsFile(path, file) {
+    return new Promise((resolve, reject) => {
+      const weightFileReader = new FileReader();
+      weightFileReader.onload = (event) => {
+        const weightData = event.target.result;
+        resolve(weightData);
+      };
+      weightFileReader.onerror = (error) => reject(`Failed to weights data from file of path '${path}'.`);
+      weightFileReader.readAsArrayBuffer(file);
+    });
+  }
+  checkManifestAndWeightFiles(manifest) {
+    const basenames = [];
+    const fileNames = this.weightsFiles.map((file) => basename(file.name));
+    const pathToFile = {};
+    for (const group of manifest) {
+      group.paths.forEach((path) => {
+        const pathBasename = basename(path);
+        if (basenames.indexOf(pathBasename) !== -1) {
+          throw new Error(`Duplicate file basename found in weights manifest: '${pathBasename}'`);
+        }
+        basenames.push(pathBasename);
+        if (fileNames.indexOf(pathBasename) === -1) {
+          throw new Error(`Weight file with basename '${pathBasename}' is not provided.`);
+        } else {
+          pathToFile[path] = this.weightsFiles[fileNames.indexOf(pathBasename)];
+        }
+      });
+    }
+    if (basenames.length !== this.weightsFiles.length) {
+      throw new Error(`Mismatch in the number of files in weights manifest (${basenames.length}) and the number of weight files provided (${this.weightsFiles.length}).`);
+    }
+    return pathToFile;
+  }
+};
+var browserDownloadsRouter = (url) => {
+  if (!env().getBool("IS_BROWSER")) {
+    return null;
+  } else {
+    if (!Array.isArray(url) && url.startsWith(BrowserDownloads.URL_SCHEME)) {
+      return browserDownloads(url.slice(BrowserDownloads.URL_SCHEME.length));
+    } else {
+      return null;
+    }
+  }
+};
+IORouterRegistry.registerSaveRouter(browserDownloadsRouter);
+function browserDownloads(fileNamePrefix = "model") {
+  return new BrowserDownloads(fileNamePrefix);
+}
+function browserFiles(files) {
+  return new BrowserFiles(files);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/progress.js
+function monitorPromisesProgress(promises, onProgress, startFraction, endFraction) {
+  checkPromises(promises);
+  startFraction = startFraction == null ? 0 : startFraction;
+  endFraction = endFraction == null ? 1 : endFraction;
+  checkFraction(startFraction, endFraction);
+  let resolvedPromise = 0;
+  const registerMonitor = (promise) => {
+    promise.then((value) => {
+      const fraction = startFraction + ++resolvedPromise / promises.length * (endFraction - startFraction);
+      onProgress(fraction);
+      return value;
+    });
+    return promise;
+  };
+  function checkPromises(promises2) {
+    assert(promises2 != null && Array.isArray(promises2) && promises2.length > 0, () => "promises must be a none empty array");
+  }
+  function checkFraction(startFraction2, endFraction2) {
+    assert(startFraction2 >= 0 && startFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got startFraction ${startFraction2}`);
+    assert(endFraction2 >= 0 && endFraction2 <= 1, () => `Progress fraction must be in range [0, 1], but got endFraction ${endFraction2}`);
+    assert(endFraction2 >= startFraction2, () => `startFraction must be no more than endFraction, but got startFraction ${startFraction2} and endFraction ${endFraction2}`);
+  }
+  return Promise.all(promises.map(registerMonitor));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/weights_loader.js
+async function loadWeightsAsArrayBuffer(fetchURLs, loadOptions) {
+  if (loadOptions == null) {
+    loadOptions = {};
+  }
+  const fetchFunc = loadOptions.fetchFunc == null ? env().platform.fetch : loadOptions.fetchFunc;
+  const requests = fetchURLs.map((fetchURL) => fetchFunc(fetchURL, loadOptions.requestInit, { isBinary: true }));
+  const fetchStartFraction = 0;
+  const fetchEndFraction = 0.5;
+  const responses = loadOptions.onProgress == null ? await Promise.all(requests) : await monitorPromisesProgress(requests, loadOptions.onProgress, fetchStartFraction, fetchEndFraction);
+  const bufferPromises = responses.map((response) => response.arrayBuffer());
+  const bufferStartFraction = 0.5;
+  const bufferEndFraction = 1;
+  const buffers = loadOptions.onProgress == null ? await Promise.all(bufferPromises) : await monitorPromisesProgress(bufferPromises, loadOptions.onProgress, bufferStartFraction, bufferEndFraction);
+  return buffers;
+}
+async function loadWeights(manifest, filePathPrefix = "", weightNames, requestInit) {
+  const fetchWeights = (fetchUrls) => loadWeightsAsArrayBuffer(fetchUrls, { requestInit });
+  const loadWeights2 = weightsLoaderFactory(fetchWeights);
+  return loadWeights2(manifest, filePathPrefix, weightNames);
+}
+function weightsLoaderFactory(fetchWeightsFunction) {
+  return async (manifest, filePathPrefix = "", weightNames) => {
+    const groupIndicesToFetchMap = manifest.map(() => false);
+    const groupWeightsToFetch = {};
+    const weightsFound = weightNames != null ? weightNames.map(() => false) : [];
+    const allManifestWeightNames = [];
+    manifest.forEach((manifestGroupConfig, groupIndex) => {
+      let groupOffset = 0;
+      manifestGroupConfig.weights.forEach((weightsEntry) => {
+        const rawDtype = "quantization" in weightsEntry ? weightsEntry.quantization.dtype : weightsEntry.dtype;
+        const weightsBytes = DTYPE_VALUE_SIZE_MAP[rawDtype] * sizeFromShape(weightsEntry.shape);
+        const enqueueWeightsForFetchingFn = () => {
+          groupIndicesToFetchMap[groupIndex] = true;
+          if (groupWeightsToFetch[groupIndex] == null) {
+            groupWeightsToFetch[groupIndex] = [];
+          }
+          groupWeightsToFetch[groupIndex].push({
+            manifestEntry: weightsEntry,
+            groupOffset,
+            sizeBytes: weightsBytes
+          });
+        };
+        if (weightNames != null) {
+          weightNames.forEach((weightName, weightIndex) => {
+            if (weightName === weightsEntry.name) {
+              enqueueWeightsForFetchingFn();
+              weightsFound[weightIndex] = true;
+            }
+          });
+        } else {
+          enqueueWeightsForFetchingFn();
+        }
+        allManifestWeightNames.push(weightsEntry.name);
+        groupOffset += weightsBytes;
+      });
+    });
+    if (!weightsFound.every((found) => found)) {
+      const weightsNotFound = weightNames.filter((_, i) => !weightsFound[i]);
+      throw new Error(`Could not find weights in manifest with names: ${weightsNotFound.join(", ")}. 
+Manifest JSON has weights with names: ${allManifestWeightNames.join(", ")}.`);
+    }
+    const groupIndicesToFetch = groupIndicesToFetchMap.reduce((accumulator, shouldFetch, i) => {
+      if (shouldFetch) {
+        accumulator.push(i);
+      }
+      return accumulator;
+    }, []);
+    const fetchUrls = [];
+    groupIndicesToFetch.forEach((i) => {
+      manifest[i].paths.forEach((filepath) => {
+        const fetchUrl = filePathPrefix + (!filePathPrefix.endsWith("/") ? "/" : "") + filepath;
+        fetchUrls.push(fetchUrl);
+      });
+    });
+    const buffers = await fetchWeightsFunction(fetchUrls);
+    const weightsTensorMap = {};
+    let bufferIndexOffset = 0;
+    groupIndicesToFetch.forEach((i) => {
+      const numBuffers = manifest[i].paths.length;
+      let groupBytes = 0;
+      for (let i2 = 0; i2 < numBuffers; i2++) {
+        groupBytes += buffers[bufferIndexOffset + i2].byteLength;
+      }
+      const groupBuffer = new ArrayBuffer(groupBytes);
+      const groupByteBuffer = new Uint8Array(groupBuffer);
+      let groupBufferOffset = 0;
+      for (let i2 = 0; i2 < numBuffers; i2++) {
+        const buffer2 = new Uint8Array(buffers[bufferIndexOffset + i2]);
+        groupByteBuffer.set(buffer2, groupBufferOffset);
+        groupBufferOffset += buffer2.byteLength;
+      }
+      const weightsEntries = groupWeightsToFetch[i];
+      weightsEntries.forEach((weightsEntry) => {
+        const byteBuffer = groupBuffer.slice(weightsEntry.groupOffset, weightsEntry.groupOffset + weightsEntry.sizeBytes);
+        const nameToTensorMap = decodeWeights(byteBuffer, [weightsEntry.manifestEntry]);
+        for (const name in nameToTensorMap) {
+          weightsTensorMap[name] = nameToTensorMap[name];
+        }
+      });
+      bufferIndexOffset += numBuffers;
+    });
+    return weightsTensorMap;
+  };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/http.js
+var OCTET_STREAM_MIME_TYPE = "application/octet-stream";
+var JSON_TYPE = "application/json";
+var HTTPRequest = class {
+  constructor(path, loadOptions) {
+    this.DEFAULT_METHOD = "POST";
+    if (loadOptions == null) {
+      loadOptions = {};
+    }
+    this.weightPathPrefix = loadOptions.weightPathPrefix;
+    this.onProgress = loadOptions.onProgress;
+    this.weightUrlConverter = loadOptions.weightUrlConverter;
+    if (loadOptions.fetchFunc != null) {
+      assert(typeof loadOptions.fetchFunc === "function", () => "Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)");
+      this.fetch = loadOptions.fetchFunc;
+    } else {
+      this.fetch = env().platform.fetch;
+    }
+    assert(path != null && path.length > 0, () => "URL path for http must not be null, undefined or empty.");
+    if (Array.isArray(path)) {
+      assert(path.length === 2, () => `URL paths for http must have a length of 2, (actual length is ${path.length}).`);
+    }
+    this.path = path;
+    if (loadOptions.requestInit != null && loadOptions.requestInit.body != null) {
+      throw new Error("requestInit is expected to have no pre-existing body, but has one.");
+    }
+    this.requestInit = loadOptions.requestInit || {};
+  }
+  async save(modelArtifacts) {
+    if (modelArtifacts.modelTopology instanceof ArrayBuffer) {
+      throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");
+    }
+    const init2 = Object.assign({ method: this.DEFAULT_METHOD }, this.requestInit);
+    init2.body = new FormData();
+    const weightsManifest = [{
+      paths: ["./model.weights.bin"],
+      weights: modelArtifacts.weightSpecs
+    }];
+    const modelTopologyAndWeightManifest = getModelJSONForModelArtifacts(modelArtifacts, weightsManifest);
+    init2.body.append("model.json", new Blob([JSON.stringify(modelTopologyAndWeightManifest)], { type: JSON_TYPE }), "model.json");
+    if (modelArtifacts.weightData != null) {
+      init2.body.append("model.weights.bin", new Blob([modelArtifacts.weightData], { type: OCTET_STREAM_MIME_TYPE }), "model.weights.bin");
+    }
+    const response = await this.fetch(this.path, init2);
+    if (response.ok) {
+      return {
+        modelArtifactsInfo: getModelArtifactsInfoForJSON(modelArtifacts),
+        responses: [response]
+      };
+    } else {
+      throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${response.status}.`);
+    }
+  }
+  async load() {
+    const modelConfigRequest = await this.fetch(this.path, this.requestInit);
+    if (!modelConfigRequest.ok) {
+      throw new Error(`Request to ${this.path} failed with status code ${modelConfigRequest.status}. Please verify this URL points to the model JSON of the model to load.`);
+    }
+    let modelJSON;
+    try {
+      modelJSON = await modelConfigRequest.json();
+    } catch (e) {
+      let message = `Failed to parse model JSON of response from ${this.path}.`;
+      if (this.path.endsWith(".pb")) {
+        message += " Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.";
+      } else {
+        message += " Please make sure the server is serving valid JSON for this request.";
+      }
+      throw new Error(message);
+    }
+    const modelTopology = modelJSON.modelTopology;
+    const weightsManifest = modelJSON.weightsManifest;
+    if (modelTopology == null && weightsManifest == null) {
+      throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);
+    }
+    return getModelArtifactsForJSON(modelJSON, (weightsManifest2) => this.loadWeights(weightsManifest2));
+  }
+  async loadWeights(weightsManifest) {
+    const weightPath = Array.isArray(this.path) ? this.path[1] : this.path;
+    const [prefix, suffix] = parseUrl(weightPath);
+    const pathPrefix = this.weightPathPrefix || prefix;
+    const weightSpecs = getWeightSpecs(weightsManifest);
+    const fetchURLs = [];
+    const urlPromises = [];
+    for (const weightsGroup of weightsManifest) {
+      for (const path of weightsGroup.paths) {
+        if (this.weightUrlConverter != null) {
+          urlPromises.push(this.weightUrlConverter(path));
+        } else {
+          fetchURLs.push(pathPrefix + path + suffix);
+        }
+      }
+    }
+    if (this.weightUrlConverter) {
+      fetchURLs.push(...await Promise.all(urlPromises));
+    }
+    const buffers = await loadWeightsAsArrayBuffer(fetchURLs, {
+      requestInit: this.requestInit,
+      fetchFunc: this.fetch,
+      onProgress: this.onProgress
+    });
+    return [weightSpecs, concatenateArrayBuffers(buffers)];
+  }
+};
+HTTPRequest.URL_SCHEME_REGEX = /^https?:\/\//;
+function parseUrl(url) {
+  const lastSlash = url.lastIndexOf("/");
+  const lastSearchParam = url.lastIndexOf("?");
+  const prefix = url.substring(0, lastSlash);
+  const suffix = lastSearchParam > lastSlash ? url.substring(lastSearchParam) : "";
+  return [prefix + "/", suffix];
+}
+function isHTTPScheme(url) {
+  return url.match(HTTPRequest.URL_SCHEME_REGEX) != null;
+}
+var httpRouter = (url, loadOptions) => {
+  if (typeof fetch === "undefined" && (loadOptions == null || loadOptions.fetchFunc == null)) {
+    return null;
+  } else {
+    let isHTTP = true;
+    if (Array.isArray(url)) {
+      isHTTP = url.every((urlItem) => isHTTPScheme(urlItem));
+    } else {
+      isHTTP = isHTTPScheme(url);
+    }
+    if (isHTTP) {
+      return http(url, loadOptions);
+    }
+  }
+  return null;
+};
+IORouterRegistry.registerSaveRouter(httpRouter);
+IORouterRegistry.registerLoadRouter(httpRouter);
+function http(path, loadOptions) {
+  return new HTTPRequest(path, loadOptions);
+}
+function browserHTTPRequest(path, loadOptions) {
+  return http(path, loadOptions);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/io/passthrough.js
+var PassthroughLoader = class {
+  constructor(modelArtifacts) {
+    this.modelArtifacts = modelArtifacts;
+  }
+  load() {
+    return this.modelArtifacts;
+  }
+};
+var PassthroughSaver = class {
+  constructor(saveHandler) {
+    this.saveHandler = saveHandler;
+  }
+  save(modelArtifacts) {
+    return this.saveHandler(modelArtifacts);
+  }
+};
+var PassthroughAsync = class {
+  constructor(handler) {
+    if (handler.load) {
+      this.load = () => Promise.resolve(handler.load());
+    }
+    if (handler.save) {
+      this.save = (modelArtifacts) => Promise.resolve(handler.save(modelArtifacts));
+    }
+  }
+};
+function fromMemory(modelArtifacts, weightSpecs, weightData, trainingConfig) {
+  const args = arguments;
+  return new PassthroughAsync(fromMemorySync(...args));
+}
+function fromMemorySync(modelArtifacts, weightSpecs, weightData, trainingConfig) {
+  if (arguments.length === 1) {
+    const isModelArtifacts = modelArtifacts.modelTopology != null || modelArtifacts.weightSpecs != null;
+    if (isModelArtifacts) {
+      return new PassthroughLoader(modelArtifacts);
+    } else {
+      console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.");
+      return new PassthroughLoader({ modelTopology: modelArtifacts });
+    }
+  } else {
+    console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release.");
+    return new PassthroughLoader({
+      modelTopology: modelArtifacts,
+      weightSpecs,
+      weightData,
+      trainingConfig
+    });
+  }
+}
+function withSaveHandler(saveHandler) {
+  return new PassthroughSaver(saveHandler);
+}
+function withSaveHandlerSync(saveHandler) {
+  return new PassthroughSaver(saveHandler);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/math.js
+var math_exports = {};
+__export(math_exports, {
+  confusionMatrix: () => confusionMatrix
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/mat_mul.js
+function matMul_(a, b, transposeA = false, transposeB = false) {
+  let $a = convertToTensor(a, "a", "matMul");
+  let $b = convertToTensor(b, "b", "matMul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  const attrs = { transposeA, transposeB };
+  return ENGINE.runKernel(BatchMatMul, inputs, attrs);
+}
+var matMul = op({ matMul_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/one_hot.js
+function oneHot_(indices, depth, onValue = 1, offValue = 0, dtype = "int32") {
+  if (depth < 2) {
+    throw new Error(`Error in oneHot: depth must be >=2, but it is ${depth}`);
+  }
+  const $indices = convertToTensor(indices, "indices", "oneHot", "int32");
+  const inputs = { indices: $indices };
+  const attrs = { dtype, depth, onValue, offValue };
+  return ENGINE.runKernel(OneHot, inputs, attrs);
+}
+var oneHot = op({ oneHot_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/globals.js
+function enableProdMode() {
+  env().set("PROD", true);
+}
+function enableDebugMode() {
+  env().set("DEBUG", true);
+}
+function disableDeprecationWarnings() {
+  env().set("DEPRECATION_WARNINGS_ENABLED", false);
+  console.warn(`TensorFlow.js deprecation warnings have been disabled.`);
+}
+function deprecationWarn(msg) {
+  if (env().getBool("DEPRECATION_WARNINGS_ENABLED")) {
+    console.warn(msg + " You can disable deprecation warnings with tf.disableDeprecationWarnings().");
+  }
+}
+setDeprecationWarningFn(deprecationWarn);
+function disposeVariables() {
+  ENGINE.disposeVariables();
+}
+function engine() {
+  return ENGINE;
+}
+function memory() {
+  return ENGINE.memory();
+}
+function profile(f) {
+  return ENGINE.profile(f);
+}
+function tidy(nameOrFn, fn) {
+  return ENGINE.tidy(nameOrFn, fn);
+}
+function dispose(container) {
+  const tensors = getTensorsInContainer(container);
+  tensors.forEach((tensor2) => tensor2.dispose());
+}
+function keep(result) {
+  return ENGINE.keep(result);
+}
+function time(f) {
+  return ENGINE.time(f);
+}
+function setBackend(backendName) {
+  return ENGINE.setBackend(backendName);
+}
+function ready() {
+  return ENGINE.ready();
+}
+function getBackend() {
+  return ENGINE.backendName;
+}
+function removeBackend(name) {
+  ENGINE.removeBackend(name);
+}
+function findBackend(name) {
+  return ENGINE.findBackend(name);
+}
+function findBackendFactory(name) {
+  return ENGINE.findBackendFactory(name);
+}
+function registerBackend(name, factory, priority = 1) {
+  return ENGINE.registerBackend(name, factory, priority);
+}
+function backend() {
+  return ENGINE.backend;
+}
+function setPlatform(platformName, platform) {
+  env().setPlatform(platformName, platform);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/imag.js
+function imag_(input2) {
+  const $input = convertToTensor(input2, "input", "imag");
+  const inputs = { input: $input };
+  return ENGINE.runKernel(Imag, inputs);
+}
+var imag = op({ imag_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/neg.js
+function neg_(x) {
+  const $x = convertToTensor(x, "x", "neg");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Neg, inputs);
+}
+var neg = op({ neg_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/real.js
+function real_(input2) {
+  const $input = convertToTensor(input2, "input", "real");
+  const inputs = { input: $input };
+  return ENGINE.runKernel(Real, inputs);
+}
+var real = op({ real_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/transpose.js
+function transpose_(x, perm, conjugate) {
+  const $x = convertToTensor(x, "x", "transpose");
+  if (perm == null) {
+    perm = $x.shape.map((s, i) => i).reverse();
+  }
+  assert($x.rank === perm.length, () => `Error in transpose: rank of input ${$x.rank} must match length of perm ${perm}.`);
+  perm.forEach((axis) => {
+    assert(axis >= 0 && axis < $x.rank, () => `All entries in 'perm' must be between 0 and ${$x.rank - 1} but got ${perm}`);
+  });
+  if ($x.rank <= 1) {
+    return $x.clone();
+  }
+  const inputs = { x: $x };
+  const attrs = { perm };
+  if ($x.dtype === "complex64") {
+    return tidy(() => {
+      let $real = real($x);
+      let $imag = imag($x);
+      $real = ENGINE.runKernel(Transpose, { x: $real }, attrs);
+      $imag = ENGINE.runKernel(Transpose, { x: $imag }, attrs);
+      if (conjugate) {
+        $imag = neg($imag);
+      }
+      return complex($real, $imag);
+    });
+  }
+  return ENGINE.runKernel(Transpose, inputs, attrs);
+}
+var transpose = op({ transpose_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/confusion_matrix.js
+function confusionMatrix_(labels, predictions, numClasses) {
+  const $labels = convertToTensor(labels, "labels", "confusionMatrix");
+  const $predictions = convertToTensor(predictions, "predictions", "confusionMatrix");
+  assert(numClasses == null || numClasses > 0 && Number.isInteger(numClasses), () => `If provided, numClasses must be a positive integer, but got ${numClasses}`);
+  assert($labels.rank === 1, () => `Expected the rank of labels to be 1, but got ${$labels.rank}`);
+  assert($predictions.rank === 1, () => `Expected the rank of predictions to be 1, but got ${$predictions.rank}`);
+  assert($labels.shape[0] === $predictions.shape[0], () => `Mismatch in the number of examples: ${$labels.shape[0]} vs. ${$predictions.shape[0]}. Labels and predictions should have the same number of elements.`);
+  assert(numClasses > 0 && Number.isInteger(numClasses), () => `numClasses is required to be a positive integer, but got ${numClasses}`);
+  const oneHotLabels = oneHot(cast($labels, "int32"), numClasses);
+  const oneHotPredictions = oneHot(cast($predictions, "int32"), numClasses);
+  const oneHotLabelsT = transpose(oneHotLabels);
+  const product = matMul(oneHotLabelsT, oneHotPredictions);
+  return cast(product, "int32");
+}
+var confusionMatrix = op({ confusionMatrix_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_util.js
+var broadcast_util_exports = {};
+__export(broadcast_util_exports, {
+  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,
+  getBroadcastDims: () => getBroadcastDims,
+  getReductionAxes: () => getReductionAxes
+});
+function getBroadcastDims(inShape, outShape) {
+  const inRank = inShape.length;
+  const dims = [];
+  for (let i = 0; i < inRank; i++) {
+    const dim = inRank - 1 - i;
+    const a = inShape[dim] || 1;
+    const b = outShape[outShape.length - 1 - i] || 1;
+    if (b > 1 && a === 1) {
+      dims.unshift(dim);
+    }
+  }
+  return dims;
+}
+function getReductionAxes(inShape, outShape) {
+  const result = [];
+  for (let i = 0; i < outShape.length; i++) {
+    const inDim = inShape[inShape.length - i - 1];
+    const outAxis = outShape.length - i - 1;
+    const outDim = outShape[outAxis];
+    if (inDim == null || inDim === 1 && outDim > 1) {
+      result.unshift(outAxis);
+    }
+  }
+  return result;
+}
+function assertAndGetBroadcastShape(shapeA, shapeB) {
+  const result = [];
+  const l = Math.max(shapeA.length, shapeB.length);
+  for (let i = 0; i < l; i++) {
+    let a = shapeA[shapeA.length - i - 1];
+    if (a == null) {
+      a = 1;
+    }
+    let b = shapeB[shapeB.length - i - 1];
+    if (b == null) {
+      b = 1;
+    }
+    if (a === 1) {
+      result.unshift(b);
+    } else if (b === 1) {
+      result.unshift(a);
+    } else if (a !== b) {
+      const errMsg = `Operands could not be broadcast together with shapes ${shapeA} and ${shapeB}.`;
+      throw Error(errMsg);
+    } else {
+      result.unshift(a);
+    }
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/browser.js
+var browser_exports = {};
+__export(browser_exports, {
+  fromPixels: () => fromPixels,
+  fromPixelsAsync: () => fromPixelsAsync,
+  toPixels: () => toPixels
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor3d.js
+function tensor3d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 3) {
+    throw new Error("tensor3d() requires shape to have three numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 3 && inferredShape.length !== 1) {
+    throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/browser.js
+var fromPixels2DContext;
+function fromPixels_(pixels, numChannels = 3) {
+  if (numChannels > 4) {
+    throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");
+  }
+  if (pixels == null) {
+    throw new Error("pixels passed to tf.browser.fromPixels() can not be null");
+  }
+  let isPixelData2 = false;
+  let isImageData = false;
+  let isVideo = false;
+  let isImage = false;
+  let isCanvasLike = false;
+  let isImageBitmap = false;
+  if (pixels.data instanceof Uint8Array) {
+    isPixelData2 = true;
+  } else if (typeof ImageData !== "undefined" && pixels instanceof ImageData) {
+    isImageData = true;
+  } else if (typeof HTMLVideoElement !== "undefined" && pixels instanceof HTMLVideoElement) {
+    isVideo = true;
+  } else if (typeof HTMLImageElement !== "undefined" && pixels instanceof HTMLImageElement) {
+    isImage = true;
+  } else if (pixels.getContext != null) {
+    isCanvasLike = true;
+  } else if (typeof ImageBitmap !== "undefined" && pixels instanceof ImageBitmap) {
+    isImageBitmap = true;
+  } else {
+    throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${pixels.constructor.name}`);
+  }
+  const kernel = getKernel(FromPixels, ENGINE.backendName);
+  if (kernel != null) {
+    const inputs = { pixels };
+    const attrs = { numChannels };
+    return ENGINE.runKernel(FromPixels, inputs, attrs);
+  }
+  const [width, height] = isVideo ? [
+    pixels.videoWidth,
+    pixels.videoHeight
+  ] : [pixels.width, pixels.height];
+  let vals;
+  if (isCanvasLike) {
+    vals = pixels.getContext("2d").getImageData(0, 0, width, height).data;
+  } else if (isImageData || isPixelData2) {
+    vals = pixels.data;
+  } else if (isImage || isVideo || isImageBitmap) {
+    if (fromPixels2DContext == null) {
+      if (typeof document === "undefined") {
+        if (typeof OffscreenCanvas !== "undefined" && typeof OffscreenCanvasRenderingContext2D !== "undefined") {
+          fromPixels2DContext = new OffscreenCanvas(1, 1).getContext("2d");
+        } else {
+          throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");
+        }
+      } else {
+        fromPixels2DContext = document.createElement("canvas").getContext("2d", { willReadFrequently: true });
+      }
+    }
+    fromPixels2DContext.canvas.width = width;
+    fromPixels2DContext.canvas.height = height;
+    fromPixels2DContext.drawImage(pixels, 0, 0, width, height);
+    vals = fromPixels2DContext.getImageData(0, 0, width, height).data;
+  }
+  let values;
+  if (numChannels === 4) {
+    values = new Int32Array(vals);
+  } else {
+    const numPixels = width * height;
+    values = new Int32Array(numPixels * numChannels);
+    for (let i = 0; i < numPixels; i++) {
+      for (let channel = 0; channel < numChannels; ++channel) {
+        values[i * numChannels + channel] = vals[i * 4 + channel];
+      }
+    }
+  }
+  const outShape = [height, width, numChannels];
+  return tensor3d(values, outShape, "int32");
+}
+function isPixelData(pixels) {
+  return pixels != null && pixels.data instanceof Uint8Array;
+}
+function isImageBitmapFullySupported() {
+  return typeof window !== "undefined" && typeof ImageBitmap !== "undefined" && window.hasOwnProperty("createImageBitmap");
+}
+function isNonEmptyPixels(pixels) {
+  return pixels != null && pixels.width !== 0 && pixels.height !== 0;
+}
+function canWrapPixelsToImageBitmap(pixels) {
+  return isImageBitmapFullySupported() && !(pixels instanceof ImageBitmap) && isNonEmptyPixels(pixels) && !isPixelData(pixels);
+}
+async function fromPixelsAsync(pixels, numChannels = 3) {
+  let inputs = null;
+  if (env().getBool("WRAP_TO_IMAGEBITMAP") && canWrapPixelsToImageBitmap(pixels)) {
+    let imageBitmap;
+    try {
+      imageBitmap = await createImageBitmap(pixels, { premultiplyAlpha: "none" });
+    } catch (e) {
+      imageBitmap = null;
+    }
+    if (imageBitmap != null && imageBitmap.width === pixels.width && imageBitmap.height === pixels.height) {
+      inputs = imageBitmap;
+    } else {
+      inputs = pixels;
+    }
+  } else {
+    inputs = pixels;
+  }
+  return fromPixels_(inputs, numChannels);
+}
+async function toPixels(img, canvas) {
+  let $img = convertToTensor(img, "img", "toPixels");
+  if (!(img instanceof Tensor)) {
+    const originalImgTensor = $img;
+    $img = cast(originalImgTensor, "int32");
+    originalImgTensor.dispose();
+  }
+  if ($img.rank !== 2 && $img.rank !== 3) {
+    throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${$img.rank}.`);
+  }
+  const [height, width] = $img.shape.slice(0, 2);
+  const depth = $img.rank === 2 ? 1 : $img.shape[2];
+  if (depth > 4 || depth === 2) {
+    throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${depth}`);
+  }
+  if ($img.dtype !== "float32" && $img.dtype !== "int32") {
+    throw new Error(`Unsupported type for toPixels: ${$img.dtype}. Please use float32 or int32 tensors.`);
+  }
+  const data = await $img.data();
+  const multiplier = $img.dtype === "float32" ? 255 : 1;
+  const bytes = new Uint8ClampedArray(width * height * 4);
+  for (let i = 0; i < height * width; ++i) {
+    const rgba = [0, 0, 0, 255];
+    for (let d = 0; d < depth; d++) {
+      const value = data[i * depth + d];
+      if ($img.dtype === "float32") {
+        if (value < 0 || value > 1) {
+          throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${value}.`);
+        }
+      } else if ($img.dtype === "int32") {
+        if (value < 0 || value > 255) {
+          throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${value}.`);
+        }
+      }
+      if (depth === 1) {
+        rgba[0] = value * multiplier;
+        rgba[1] = value * multiplier;
+        rgba[2] = value * multiplier;
+      } else {
+        rgba[d] = value * multiplier;
+      }
+    }
+    const j = i * 4;
+    bytes[j + 0] = Math.round(rgba[0]);
+    bytes[j + 1] = Math.round(rgba[1]);
+    bytes[j + 2] = Math.round(rgba[2]);
+    bytes[j + 3] = Math.round(rgba[3]);
+  }
+  if (canvas != null) {
+    canvas.width = width;
+    canvas.height = height;
+    const ctx = canvas.getContext("2d");
+    const imageData = new ImageData(bytes, width, height);
+    ctx.putImageData(imageData, 0, 0);
+  }
+  if ($img !== img) {
+    $img.dispose();
+  }
+  return bytes;
+}
+var fromPixels = op({ fromPixels_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/gather_nd_util.js
+var gather_nd_util_exports = {};
+__export(gather_nd_util_exports, {
+  prepareAndValidate: () => prepareAndValidate
+});
+function prepareAndValidate(tensor2, indices) {
+  const tensorRank = tensor2.shape.length;
+  const indicesRank = indices.shape.length;
+  if (tensorRank < 1) {
+    throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${tensorRank}.`);
+  }
+  if (indicesRank < 1) {
+    throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${indicesRank}.`);
+  }
+  if (indices.dtype !== "int32") {
+    throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${indices.dtype}.`);
+  }
+  if (indices.shape[indicesRank - 1] > tensorRank) {
+    throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${indices.shape[indicesRank - 1]} vs. ${tensorRank}`);
+  }
+  if (sizeFromShape(tensor2.shape) === 0) {
+    throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${tensor2.shape}.`);
+  }
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  let nResult = 1;
+  for (let i = 0; i < indicesShape.length - 1; ++i) {
+    nResult *= indicesShape[i];
+  }
+  const inputShape = tensor2.shape;
+  const resultShape = indicesShape.slice();
+  resultShape.pop();
+  let sliceSize = 1;
+  for (let i = sliceRank; i < tensorRank; ++i) {
+    sliceSize *= inputShape[i];
+    resultShape.push(inputShape[i]);
+  }
+  const strides = [
+    ...computeStrides(tensor2.shape).map((stride) => stride / sliceSize),
+    1
+  ].slice(0, sliceRank);
+  return [resultShape, nResult, sliceSize, strides];
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/scatter_nd_util.js
+var scatter_nd_util_exports = {};
+__export(scatter_nd_util_exports, {
+  calculateShapes: () => calculateShapes,
+  validateInput: () => validateInput,
+  validateUpdateShape: () => validateUpdateShape
+});
+function validateUpdateShape(shape, indices, updates) {
+  const sliceDim = indices.rank > 1 ? indices.shape[indices.rank - 1] : 1;
+  const batchDim = indices.rank > 1 ? indices.rank - 1 : 1;
+  const shapeError = `Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${updates.shape}, indices.shape: ${indices.shape}, shape: ${shape}, sliceDim: ${sliceDim}, and batchDim: ${batchDim}.`;
+  if (updates.rank < batchDim) {
+    throw new Error(shapeError + ` update.rank < ${batchDim}. `);
+  }
+  if (shape.length < sliceDim + (updates.rank - batchDim)) {
+    throw new Error(shapeError + ` Output shape length < ${sliceDim + (updates.rank - batchDim)}`);
+  }
+  if (updates.rank !== batchDim + shape.length - sliceDim) {
+    throw new Error(shapeError + ` update.rank != ${batchDim + shape.length - sliceDim}`);
+  }
+  for (let d = 0; d < batchDim; ++d) {
+    if (updates.shape[d] !== indices.shape[d]) {
+      throw new Error(shapeError + ` updates.shape[${d}] (${updates.shape[d]}) != indices.shape[${d}] (${indices.shape[d]}).`);
+    }
+  }
+  for (let d = 0; d < updates.rank - batchDim; ++d) {
+    if (updates.shape[d + batchDim] !== shape[d + sliceDim]) {
+      throw new Error(shapeError + ` updates.shape[${d + batchDim}] (${updates.shape[d + batchDim]}) != shape[${d + batchDim}] (${shape[d + batchDim]})`);
+    }
+  }
+}
+function validateInput(updates, indices, shape) {
+  if (indices.rank < 1) {
+    throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${indices.rank}.`);
+  }
+  if (updates.rank < 1) {
+    throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${updates.rank}.`);
+  }
+  if (indices.dtype !== "int32") {
+    throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${indices.dtype}`);
+  }
+  if (shape.length < 1) {
+    throw new Error(`Output rank must be greater or equal to 1, but got shape: ${shape}`);
+  }
+  if (shape.length === 0) {
+    if (indices.size === 0) {
+      throw new Error(`Indices specified for empty output. indices shape: ${indices.shape}`);
+    }
+    if (updates.size === 0) {
+      throw new Error(`Updates specified for empty output. updates shape: ${updates.shape}`);
+    }
+  }
+  validateUpdateShape(shape, indices, updates);
+}
+function calculateShapes(updates, indices, shape) {
+  const indicesRank = indices.shape.length;
+  const sliceRank = indicesRank > 1 ? indices.shape[indicesRank - 1] : 1;
+  const totalNd = shape.length;
+  let sliceSize = 1;
+  for (let i = sliceRank; i < totalNd; ++i) {
+    sliceSize *= shape[i];
+  }
+  const safeSliceDim = sliceRank < 1 ? 1 : sliceRank;
+  const numUpdates = sizeFromShape(indices.shape) / safeSliceDim;
+  const strides = [...computeStrides(shape.slice(0, sliceRank)), 1];
+  const outputSize = sizeFromShape(shape);
+  return { sliceRank, numUpdates, sliceSize, strides, outputSize };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice_util.js
+var slice_util_exports = {};
+__export(slice_util_exports, {
+  assertParamsValid: () => assertParamsValid,
+  computeFlatOffset: () => computeFlatOffset,
+  computeOutShape: () => computeOutShape,
+  getNormalizedAxes: () => getNormalizedAxes,
+  isSliceContinous: () => isSliceContinous,
+  maskToAxes: () => maskToAxes,
+  parseSliceParams: () => parseSliceParams,
+  sliceInfo: () => sliceInfo,
+  startForAxis: () => startForAxis,
+  startIndicesWithElidedDims: () => startIndicesWithElidedDims,
+  stopForAxis: () => stopForAxis,
+  stopIndicesWithElidedDims: () => stopIndicesWithElidedDims,
+  stridesForAxis: () => stridesForAxis,
+  stridesWithElidedDims: () => stridesWithElidedDims
+});
+var NEW_AXIS = -2;
+var SHRINK_AXIS = -1;
+function assertParamsValid(input2, begin, size) {
+  const inputRank = input2.shape.length;
+  assert(inputRank === begin.length, () => `Error in slice${inputRank}D: Length of begin ${begin} must match the rank of the array (${inputRank}).`);
+  assert(inputRank === size.length, () => `Error in slice${inputRank}D: Length of size ${size} must match the rank of the array (${inputRank}).`);
+  for (let i = 0; i < inputRank; ++i) {
+    assert(begin[i] + size[i] <= input2.shape[i], () => `Error in slice${inputRank}D: begin[${i}] + size[${i}] (${begin[i] + size[i]}) would overflow input.shape[${i}] (${input2.shape[i]})`);
+  }
+}
+function maskToAxes(mask) {
+  const axes = [];
+  let axis = 0;
+  while (mask > 0) {
+    if (mask & 1) {
+      axes.push(axis);
+    }
+    mask /= 2;
+    axis++;
+  }
+  return axes;
+}
+function computeOutShape(begin, end, strides) {
+  const size = [];
+  for (let axis = 0; axis < begin.length; axis++) {
+    size[axis] = Math.ceil((end[axis] - begin[axis]) / strides[axis]);
+  }
+  return size;
+}
+function stridesWithElidedDims(strides, ellipsisInsertionIndex, numElidedAxes, inputShape) {
+  const newStrides = [...strides];
+  for (let i = newStrides.length; i < inputShape.length; i++) {
+    newStrides.push(1);
+  }
+  for (let i = 0; i < numElidedAxes; i++) {
+    if (i === 0) {
+      newStrides[ellipsisInsertionIndex] = 1;
+    } else {
+      newStrides.splice(ellipsisInsertionIndex, 0, 1);
+      newStrides.pop();
+    }
+  }
+  return newStrides;
+}
+function unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, normalizedAxis) {
+  if (normalizedAxis <= ellipsisInsertionIndex) {
+    return normalizedAxis;
+  }
+  return normalizedAxis - (numElidedAxes - 1);
+}
+function getElidedAxes(numElidedAxes, ellipsisInsertionIndex) {
+  const elidedAxes = [];
+  for (let i = 0; i < numElidedAxes; i++) {
+    elidedAxes.push(ellipsisInsertionIndex + i);
+  }
+  return elidedAxes;
+}
+function getNormalizedAxes(inputShape, ellipsisAxes, numInterpolatedAxes, begin, end, strides, beginMask, endMask, ellipsisMask) {
+  const inputRank = inputShape.length;
+  let normalizedBegin = new Array(inputRank), normalizedEnd = new Array(inputRank), normalizedStrides = new Array(inputRank);
+  if (ellipsisAxes.length && numInterpolatedAxes > 0) {
+    const fullIndex = ellipsisAxes[0];
+    const numElidedAxes = numInterpolatedAxes + 1;
+    normalizedBegin = startIndicesWithElidedDims(beginMask, fullIndex, numElidedAxes, begin, inputShape);
+    normalizedEnd = stopIndicesWithElidedDims(endMask, fullIndex, numElidedAxes, end, inputShape);
+    normalizedStrides = stridesWithElidedDims(strides, fullIndex, numElidedAxes, inputShape);
+  } else {
+    for (let axis = 0; axis < inputRank; axis++) {
+      normalizedBegin[axis] = startForAxis(beginMask, begin, strides, inputShape, axis, ellipsisMask);
+      normalizedEnd[axis] = stopForAxis(endMask, end, strides, inputShape, axis, ellipsisMask);
+      normalizedStrides[axis] = stridesForAxis(strides, axis, ellipsisMask);
+    }
+  }
+  return {
+    begin: normalizedBegin,
+    end: normalizedEnd,
+    strides: normalizedStrides
+  };
+}
+function startIndicesWithElidedDims(beginMask, ellipsisInsertionIndex, numElidedAxes, originalBegin, inputShape) {
+  const newIndices = [...inputShape];
+  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);
+  for (let axis = 0; axis < newIndices.length; axis++) {
+    if (elidedAxes.indexOf(axis) > -1) {
+      newIndices[axis] = 0;
+    } else {
+      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);
+      let originalValue = originalBegin[originalAxis];
+      if (beginMask & 1 << originalAxis) {
+        originalValue = 0;
+      }
+      newIndices[axis] = originalValue;
+    }
+  }
+  return newIndices;
+}
+function stopIndicesWithElidedDims(endMask, ellipsisInsertionIndex, numElidedAxes, originalEnd, inputShape) {
+  const newIndices = [...inputShape];
+  const elidedAxes = getElidedAxes(numElidedAxes, ellipsisInsertionIndex);
+  for (let axis = 0; axis < newIndices.length; axis++) {
+    if (elidedAxes.indexOf(axis) > -1) {
+      newIndices[axis] = Number.MAX_SAFE_INTEGER;
+    } else {
+      const originalAxis = unnormalizeAxis(ellipsisInsertionIndex, numElidedAxes, axis);
+      let originalValue = originalEnd[originalAxis];
+      if (endMask & 1 << originalAxis) {
+        originalValue = Number.MAX_SAFE_INTEGER;
+      }
+      newIndices[axis] = originalValue;
+    }
+  }
+  for (let i = 0; i < newIndices.length; i++) {
+    const axisSize = inputShape[i];
+    if (newIndices[i] < 0) {
+      newIndices[i] += axisSize;
+    }
+    newIndices[i] = clamp(0, newIndices[i], inputShape[i]);
+  }
+  return newIndices;
+}
+function stridesForAxis(strides, axis, ellipsisMask) {
+  let stride = strides[axis];
+  if (ellipsisMask & 1 << axis || stride == null) {
+    stride = 1;
+  }
+  return stride;
+}
+function startForAxis(beginMask, startIndices, strides, inputShape, axis, ellipsisMask) {
+  let start = startIndices[axis];
+  const stride = strides[axis] || 1;
+  if (beginMask & 1 << axis || ellipsisMask & 1 << axis || start == null) {
+    if (stride > 0) {
+      start = Number.MIN_SAFE_INTEGER;
+    } else {
+      start = Number.MAX_SAFE_INTEGER;
+    }
+  }
+  const axisSize = inputShape[axis];
+  if (start < 0) {
+    start += axisSize;
+  }
+  start = clamp(0, start, axisSize - 1);
+  return start;
+}
+function stopForAxis(endMask, stopIndices, strides, inputShape, axis, ellipsisMask) {
+  let stop = stopIndices[axis];
+  const stride = strides[axis] || 1;
+  if (endMask & 1 << axis || ellipsisMask & 1 << axis || stop == null) {
+    if (stride > 0) {
+      stop = Number.MAX_SAFE_INTEGER;
+    } else {
+      stop = Number.MIN_SAFE_INTEGER;
+    }
+  }
+  const axisSize = inputShape[axis];
+  if (stop < 0) {
+    stop += axisSize;
+  }
+  if (stride > 0) {
+    stop = clamp(0, stop, axisSize);
+  } else {
+    stop = clamp(-1, stop, axisSize - 1);
+  }
+  return stop;
+}
+function isSliceContinous(shape, begin, size) {
+  let firstNonOneAxis = size.length;
+  for (let i = 0; i < size.length; i++) {
+    if (size[i] > 1) {
+      firstNonOneAxis = i;
+      break;
+    }
+  }
+  for (let i = firstNonOneAxis + 1; i < size.length; i++) {
+    if (begin[i] > 0 || size[i] !== shape[i]) {
+      return false;
+    }
+  }
+  return true;
+}
+function computeFlatOffset(begin, strides) {
+  let flatOffset = begin.length > 0 ? begin[begin.length - 1] : 1;
+  for (let i = 0; i < begin.length - 1; i++) {
+    flatOffset += begin[i] * strides[i];
+  }
+  return flatOffset;
+}
+function parseSliceParams(x, begin, size) {
+  let begin_;
+  const xRank = x.shape.length;
+  if (typeof begin === "number") {
+    begin_ = [begin, ...new Array(xRank - 1).fill(0)];
+  } else if (begin.length < xRank) {
+    begin_ = begin.concat(new Array(xRank - begin.length).fill(0));
+  } else {
+    begin_ = begin.slice();
+  }
+  begin_.forEach((d) => {
+    assert(d !== -1, () => "slice() does not support negative begin indexing.");
+  });
+  let size_;
+  if (size == null) {
+    size_ = new Array(xRank).fill(-1);
+  } else if (typeof size === "number") {
+    size_ = [size, ...new Array(xRank - 1).fill(-1)];
+  } else if (size.length < xRank) {
+    size_ = size.concat(new Array(xRank - size.length).fill(-1));
+  } else {
+    size_ = size;
+  }
+  size_ = size_.map((d, i) => {
+    if (d >= 0) {
+      return d;
+    } else {
+      assert(d === -1, () => `Negative size values should be exactly -1 but got ${d} for the slice() size at index ${i}.`);
+      return x.shape[i] - begin_[i];
+    }
+  });
+  return [begin_, size_];
+}
+function sliceInfo(xShape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {
+  let stridesNonNull;
+  if (strides == null) {
+    stridesNonNull = new Array(begin.length);
+    stridesNonNull.fill(1);
+  } else {
+    stridesNonNull = strides;
+  }
+  if (ellipsisMask != null && (ellipsisMask & ellipsisMask - 1) !== 0) {
+    throw new Error("Multiple ellipses in slice is not allowed.");
+  }
+  let ellipsisSeen = false;
+  const sparseSpec = {
+    dims: stridesNonNull.length,
+    numAddAxisAfterEllipsis: 0,
+    begin: begin.slice(),
+    end: end.slice(),
+    strides: stridesNonNull.slice(),
+    beginMask,
+    endMask,
+    ellipsisMask,
+    newAxisMask,
+    shrinkAxisMask
+  };
+  for (let i = 0; i < sparseSpec.dims; i++) {
+    if (ellipsisSeen && (1 << i & newAxisMask) !== 0) {
+      sparseSpec.numAddAxisAfterEllipsis++;
+    }
+    if (1 << i & ellipsisMask) {
+      ellipsisSeen = true;
+    }
+  }
+  if (!ellipsisSeen) {
+    sparseSpec.ellipsisMask |= 1 << sparseSpec.dims;
+    sparseSpec.dims++;
+  }
+  const denseSpec = {
+    dims: xShape.length,
+    beginMask: 0,
+    endMask: 0,
+    beginValid: false,
+    endValid: false
+  };
+  buildDenseSpec(sparseSpec, denseSpec);
+  let isIdentity = true;
+  let sliceDim0 = true;
+  let isSimpleSlice = true;
+  const processingShape = [];
+  const finalShape = [];
+  for (let i = 0; i < xShape.length; ++i) {
+    if (denseSpec.strides[i] === 0) {
+      throw Error(`strides[${i}] must be non-zero`);
+    }
+    const shrinkI = !!(denseSpec.shrinkAxisMask & 1 << i);
+    const dimI = xShape[i];
+    if (dimI === -1) {
+      processingShape.push(shrinkI ? 1 : -1);
+      continue;
+    }
+    const masks = [denseSpec.beginMask & 1 << i, denseSpec.endMask & 1 << i];
+    const validRange = [
+      denseSpec.strides[i] > 0 ? 0 : -1,
+      denseSpec.strides[i] > 0 ? dimI : dimI - 1
+    ];
+    if (shrinkI && denseSpec.strides[i] <= 0) {
+      throw Error("only stride 1 allowed on non-range indexing.");
+    }
+    isSimpleSlice = isSimpleSlice && denseSpec.strides[i] === 1;
+    const beginAndEndMasked = !!(denseSpec.beginMask & 1 << i && denseSpec.endMask & 1 << i);
+    if (denseSpec.beginValid && denseSpec.endValid) {
+      if (shrinkI) {
+        const xFwd = denseSpec.begin[i] < 0 ? dimI + denseSpec.begin[i] : denseSpec.begin[i];
+        denseSpec.begin[i] = xFwd;
+        denseSpec.end[i] = denseSpec.begin[i] + 1;
+        if (xFwd < 0 || xFwd >= dimI) {
+          throw Error(`slice index ${denseSpec.begin[i]} of dimension ${i} out of bounds.`);
+        }
+      } else {
+        denseSpec.begin[i] = canonical(denseSpec.begin[i], 0, denseSpec.strides[i], dimI, masks, validRange);
+        denseSpec.end[i] = canonical(denseSpec.end[i], 1, denseSpec.strides[i], dimI, masks, validRange);
+      }
+      const takeAllInDimension = denseSpec.strides[i] === 1 && denseSpec.begin[i] === 0 && denseSpec.end[i] === dimI;
+      isIdentity = isIdentity && takeAllInDimension;
+      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || takeAllInDimension);
+    } else {
+      isIdentity = isIdentity && (denseSpec.strides[i] === 1 && beginAndEndMasked);
+      sliceDim0 = sliceDim0 && (i === 0 && denseSpec.strides[i] === 1 || beginAndEndMasked);
+    }
+    let intervalLength;
+    let knownInterval = false;
+    if (denseSpec.beginValid && denseSpec.endValid) {
+      intervalLength = denseSpec.end[i] - denseSpec.begin[i];
+      knownInterval = true;
+    } else if (shrinkI) {
+      intervalLength = 1;
+      knownInterval = true;
+    } else if (beginAndEndMasked) {
+      if (dimI >= 0) {
+        if (denseSpec.strides[i] < 0) {
+          intervalLength = -dimI;
+        } else {
+          intervalLength = dimI;
+        }
+        knownInterval = true;
+      }
+    }
+    if (knownInterval) {
+      let sizeI;
+      if (intervalLength === 0 || intervalLength < 0 !== denseSpec.strides[i] < 0) {
+        sizeI = 0;
+      } else {
+        sizeI = Math.trunc(intervalLength / denseSpec.strides[i]) + (intervalLength % denseSpec.strides[i] !== 0 ? 1 : 0);
+      }
+      processingShape.push(sizeI);
+    } else {
+      processingShape.push(-1);
+    }
+  }
+  for (let denseDim = 0; denseDim < denseSpec.finalShapeGatherIndices.length; ++denseDim) {
+    const gatherIndex = denseSpec.finalShapeGatherIndices[denseDim];
+    if (gatherIndex >= 0) {
+      finalShape.push(processingShape[gatherIndex]);
+    } else if (gatherIndex === NEW_AXIS) {
+      finalShape.push(1);
+    }
+  }
+  const finalShapeSparse = finalShape.filter((dim, i) => denseSpec.finalShapeGatherIndices[i] !== NEW_AXIS);
+  return {
+    finalShapeSparse,
+    finalShape,
+    isIdentity,
+    sliceDim0,
+    isSimpleSlice,
+    begin: denseSpec.begin,
+    end: denseSpec.end,
+    strides: denseSpec.strides
+  };
+}
+function buildDenseSpec(sparse2, dense2) {
+  dense2.beginMask = 0;
+  dense2.endMask = 0;
+  dense2.shrinkAxisMask = 0;
+  let fullIndex = 0;
+  dense2.beginValid = sparse2.begin != null;
+  dense2.endValid = sparse2.end != null;
+  dense2.begin = new Array(dense2.dims);
+  dense2.end = new Array(dense2.dims);
+  dense2.strides = new Array(dense2.dims);
+  dense2.finalShapeGatherIndices = [];
+  dense2.finalShapeGatherIndicesSparse = [];
+  dense2.inputShapeGatherIndicesSparse = new Array(dense2.dims);
+  for (let i = 0; i < sparse2.dims; i++) {
+    if (1 << i & sparse2.ellipsisMask) {
+      const nextIndex = Math.min(dense2.dims - (sparse2.dims - i) + 1 + sparse2.numAddAxisAfterEllipsis, dense2.dims);
+      for (; fullIndex < nextIndex; fullIndex++) {
+        dense2.begin[fullIndex] = 0;
+        dense2.end[fullIndex] = 0;
+        dense2.strides[fullIndex] = 1;
+        dense2.beginMask |= 1 << fullIndex;
+        dense2.endMask |= 1 << fullIndex;
+        dense2.finalShapeGatherIndices.push(fullIndex);
+        dense2.finalShapeGatherIndicesSparse.push(-1);
+        dense2.inputShapeGatherIndicesSparse[fullIndex] = i;
+      }
+    } else if (1 << i & sparse2.newAxisMask) {
+      dense2.finalShapeGatherIndices.push(NEW_AXIS);
+      dense2.finalShapeGatherIndicesSparse.push(-1);
+    } else {
+      if (fullIndex === dense2.begin.length) {
+        throw Error(`Index out of range using input dim ${fullIndex}; input has only ${dense2.dims} dims, ${dense2.begin.length}.`);
+      }
+      if (sparse2.begin != null) {
+        dense2.begin[fullIndex] = sparse2.begin[i];
+      }
+      if (sparse2.end != null) {
+        dense2.end[fullIndex] = sparse2.end[i];
+      }
+      dense2.strides[fullIndex] = sparse2.strides[i];
+      if (sparse2.beginMask & 1 << i) {
+        dense2.beginMask |= 1 << fullIndex;
+      }
+      if (sparse2.endMask & 1 << i) {
+        dense2.endMask |= 1 << fullIndex;
+      }
+      if (sparse2.shrinkAxisMask & 1 << i) {
+        dense2.finalShapeGatherIndices.push(SHRINK_AXIS);
+        dense2.finalShapeGatherIndicesSparse.push(-1);
+        dense2.shrinkAxisMask |= 1 << fullIndex;
+      } else {
+        dense2.finalShapeGatherIndices.push(fullIndex);
+        dense2.finalShapeGatherIndicesSparse.push(i);
+      }
+      dense2.inputShapeGatherIndicesSparse[fullIndex] = i;
+      fullIndex++;
+    }
+  }
+}
+function canonical(x, c, strideI, dimI, masks, validRange) {
+  if (masks[c]) {
+    return strideI > 0 ? validRange[c] : validRange[c + 1 & 1];
+  } else {
+    const xFwd = x < 0 ? dimI + x : x;
+    return xFwd < validRange[0] ? validRange[0] : xFwd > validRange[1] ? validRange[1] : xFwd;
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/serialization.js
+var serialization_exports = {};
+__export(serialization_exports, {
+  Serializable: () => Serializable,
+  SerializationMap: () => SerializationMap,
+  registerClass: () => registerClass
+});
+var Serializable = class {
+  getClassName() {
+    return this.constructor.className;
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+var SerializationMap = class {
+  constructor() {
+    this.classNameMap = {};
+  }
+  static getMap() {
+    if (SerializationMap.instance == null) {
+      SerializationMap.instance = new SerializationMap();
+    }
+    return SerializationMap.instance;
+  }
+  static register(cls) {
+    SerializationMap.getMap().classNameMap[cls.className] = [cls, cls.fromConfig];
+  }
+};
+function registerClass(cls) {
+  assert(cls.className != null, () => `Class being registered does not have the static className property defined.`);
+  assert(typeof cls.className === "string", () => `className is required to be a string, but got type ` + typeof cls.className);
+  assert(cls.className.length > 0, () => `Class being registered has an empty-string as its className, which is disallowed.`);
+  SerializationMap.register(cls);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/test_util.js
+var test_util_exports = {};
+__export(test_util_exports, {
+  TEST_EPSILON_FLOAT16: () => TEST_EPSILON_FLOAT16,
+  createVideoElement: () => createVideoElement,
+  encodeStrings: () => encodeStrings,
+  expectArrayBuffersEqual: () => expectArrayBuffersEqual,
+  expectArraysClose: () => expectArraysClose,
+  expectArraysEqual: () => expectArraysEqual,
+  expectNumbersClose: () => expectNumbersClose,
+  expectPromiseToFail: () => expectPromiseToFail,
+  expectValuesInRange: () => expectValuesInRange,
+  play: () => play,
+  testEpsilon: () => testEpsilon
+});
+var TEST_EPSILON_FLOAT32 = 1e-3;
+var TEST_EPSILON_FLOAT16 = 0.1;
+function expectArraysClose(actual, expected, epsilon3) {
+  if (epsilon3 == null) {
+    epsilon3 = testEpsilon();
+  }
+  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, epsilon3));
+}
+function testEpsilon() {
+  return ENGINE.backend.floatPrecision() === 32 ? TEST_EPSILON_FLOAT32 : TEST_EPSILON_FLOAT16;
+}
+function expectArraysPredicate(actual, expected, predicate) {
+  let checkClassType = true;
+  if (isTypedArray(actual) || isTypedArray(expected)) {
+    checkClassType = false;
+  }
+  if (isTypedArray(actual) && isTypedArray(expected)) {
+    checkClassType = true;
+  }
+  if (checkClassType) {
+    const aType = actual.constructor.name;
+    const bType = expected.constructor.name;
+    if (aType !== bType) {
+      throw new Error(`Arrays are of different type. Actual: ${aType}. Expected: ${bType}`);
+    }
+  }
+  if (Array.isArray(actual) && Array.isArray(expected)) {
+    const actualShape = inferShape(actual);
+    const expectedShape = inferShape(expected);
+    if (!arraysEqual(actualShape, expectedShape)) {
+      throw new Error(`Arrays have different shapes. Actual: [${actualShape}]. Expected: [${expectedShape}]`);
+    }
+  }
+  const actualFlat = isTypedArray(actual) ? actual : flatten(actual);
+  const expectedFlat = isTypedArray(expected) ? expected : flatten(expected);
+  if (actualFlat.length !== expectedFlat.length) {
+    throw new Error(`Arrays have different lengths actual: ${actualFlat.length} vs expected: ${expectedFlat.length}.
+Actual:   ${actualFlat}.
+Expected: ${expectedFlat}.`);
+  }
+  for (let i = 0; i < expectedFlat.length; ++i) {
+    const a = actualFlat[i];
+    const e = expectedFlat[i];
+    if (!predicate(a, e)) {
+      throw new Error(`Arrays differ: actual[${i}] = ${a}, expected[${i}] = ${e}.
+Actual:   ${actualFlat}.
+Expected: ${expectedFlat}.`);
+    }
+  }
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function expectPromiseToFail(fn, done) {
+  fn().then(() => done.fail(), () => done());
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function expectArraysEqual(actual, expected) {
+  const exp4 = typeof expected === "string" || typeof expected === "number" || typeof expected === "boolean" ? [expected] : expected;
+  if (isString(actual) || isString(actual[0]) || isString(expected) || isString(expected[0])) {
+    return expectArraysPredicate(actual, exp4, (a, b) => a == b);
+  }
+  return expectArraysPredicate(actual, expected, (a, b) => areClose(a, b, 0));
+}
+function expectNumbersClose(a, e, epsilon3) {
+  if (epsilon3 == null) {
+    epsilon3 = testEpsilon();
+  }
+  if (!areClose(a, e, epsilon3)) {
+    throw new Error(`Numbers differ: actual === ${a}, expected === ${e}`);
+  }
+  if (typeof expect !== "undefined") {
+    expect().nothing();
+  }
+}
+function areClose(a, e, epsilon3) {
+  if (!isFinite(a) && !isFinite(e)) {
+    return true;
+  }
+  if (isNaN(a) || isNaN(e) || Math.abs(a - e) > epsilon3) {
+    return false;
+  }
+  return true;
+}
+function expectValuesInRange(actual, low, high) {
+  for (let i = 0; i < actual.length; i++) {
+    if (actual[i] < low || actual[i] > high) {
+      throw new Error(`Value out of range:${actual[i]} low: ${low}, high: ${high}`);
+    }
+  }
+}
+function expectArrayBuffersEqual(actual, expected) {
+  const actualArray = new Float32Array(actual);
+  const expectedArray = new Float32Array(expected);
+  if (actualArray.length !== expectedArray.length) {
+    throw new Error(`Expected ArrayBuffer to be of length ${expectedArray.length}, but it was ${actualArray.length}`);
+  }
+  for (let i = 0; i < expectedArray.length; i++) {
+    if (actualArray[i] !== expectedArray[i]) {
+      throw new Error(`Expected ArrayBuffer value at ${i} to be ${expectedArray[i]} but got ${actualArray[i]} instead`);
+    }
+  }
+}
+function encodeStrings(a) {
+  for (let i = 0; i < a.length; i++) {
+    const val = a[i];
+    if (Array.isArray(val)) {
+      encodeStrings(val);
+    } else {
+      a[i] = encodeString(val);
+    }
+  }
+  return a;
+}
+function createVideoElement(source) {
+  const video = document.createElement("video");
+  if ("playsInline" in video) {
+    video.playsInline = true;
+  }
+  video.muted = true;
+  video.loop = true;
+  video.style.position = "fixed";
+  video.style.left = "0px";
+  video.style.top = "0px";
+  video.preload = "auto";
+  video.appendChild(source);
+  return new Promise((resolve) => {
+    video.addEventListener("loadeddata", (_) => resolve(video));
+    video.load();
+  });
+}
+async function play(video) {
+  await video.play();
+  if ("requestVideoFrameCallback" in video) {
+    await new Promise((resolve) => {
+      video.requestVideoFrameCallback(resolve);
+    });
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/version.js
+var version = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/add.js
+function add_(a, b) {
+  let $a = convertToTensor(a, "a", "add");
+  let $b = convertToTensor(b, "b", "add");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Add, inputs);
+}
+var add2 = op({ add_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/floorDiv.js
+function floorDiv_(a, b) {
+  let $a = convertToTensor(a, "a", "floorDiv");
+  let $b = convertToTensor(b, "b", "floorDiv");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(FloorDiv, inputs);
+}
+var floorDiv = op({ floorDiv_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/div.js
+function div_(a, b) {
+  let $a = convertToTensor(a, "a", "div");
+  let $b = convertToTensor(b, "b", "div");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "int32" && $b.dtype === "int32") {
+    return floorDiv($a, $b);
+  }
+  const inputs = { a: $a, b: $b };
+  const attrs = {};
+  return ENGINE.runKernel(RealDiv, inputs, attrs);
+}
+var div = op({ div_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/mul.js
+function mul_(a, b) {
+  let $a = convertToTensor(a, "a", "mul");
+  let $b = convertToTensor(b, "b", "mul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Multiply, inputs);
+}
+var mul = op({ mul_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/abs.js
+function abs_(x) {
+  const $x = convertToTensor(x, "x", "abs");
+  if ($x.dtype === "complex64") {
+    const inputs = { x: $x };
+    return ENGINE.runKernel(ComplexAbs, inputs);
+  } else {
+    const inputs = { x: $x };
+    return ENGINE.runKernel(Abs, inputs);
+  }
+}
+var abs = op({ abs_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/acos.js
+function acos_(x) {
+  const $x = convertToTensor(x, "x", "acos");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Acos, inputs);
+}
+var acos = op({ acos_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/acosh.js
+function acosh_(x) {
+  const $x = convertToTensor(x, "x", "acosh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Acosh, inputs);
+}
+var acosh = op({ acosh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/add_n.js
+function addN_(tensors) {
+  assert(Array.isArray(tensors), () => "The argument passed to tf.addN() must be a list of tensors");
+  assert(tensors.length >= 1, () => `Must pass at least one tensor to tf.addN(), but got ${tensors.length}`);
+  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, "addN"));
+  const firstTensor = $tensors[0];
+  $tensors.forEach((t) => {
+    if (t.dtype !== firstTensor.dtype) {
+      throw new Error("All tensors passed to tf.addN() must have the same dtype");
+    }
+  });
+  $tensors.forEach((t) => {
+    if (!arraysEqual(t.shape, firstTensor.shape)) {
+      throw new Error("All tensors passed to tf.addN() must have the same shape");
+    }
+  });
+  const inputs = $tensors;
+  return ENGINE.runKernel(AddN, inputs);
+}
+var addN = op({ addN_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/all.js
+function all_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "all", "bool");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(All, inputs, attrs);
+}
+var all = op({ all_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/any.js
+function any_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "any", "bool");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Any, inputs, attrs);
+}
+var any = op({ any_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/arg_max.js
+function argMax_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "argMax");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(ArgMax, inputs, attrs);
+}
+var argMax = op({ argMax_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/arg_min.js
+function argMin_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "argMin");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(ArgMin, inputs, attrs);
+}
+var argMin = op({ argMin_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/asin.js
+function asin_(x) {
+  const $x = convertToTensor(x, "x", "asin");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Asin, inputs);
+}
+var asin = op({ asin_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/asinh.js
+function asinh_(x) {
+  const $x = convertToTensor(x, "x", "asinh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Asinh, inputs);
+}
+var asinh = op({ asinh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/atan.js
+function atan_(x) {
+  const $x = convertToTensor(x, "x", "atan");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Atan, inputs);
+}
+var atan = op({ atan_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/atan2.js
+function atan2_(a, b) {
+  let $a = convertToTensor(a, "a", "atan2");
+  let $b = convertToTensor(b, "b", "atan2");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Atan2, inputs);
+}
+var atan2 = op({ atan2_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/atanh.js
+function atanh_(x) {
+  const $x = convertToTensor(x, "x", "atanh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Atanh, inputs);
+}
+var atanh = op({ atanh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv_util.js
+function computeDilation2DInfo(inputShape, filterShape, strides, pad3, dataFormat = "NHWC", dilations) {
+  const inputChannels = inputShape[3];
+  const $filterShape = [...filterShape, inputChannels];
+  const $dataFormat = convertConv2DDataFormat(dataFormat);
+  return computeConv2DInfo(inputShape, $filterShape, strides, dilations, pad3, null, null, $dataFormat);
+}
+function computePool2DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = "channelsLast") {
+  const [filterHeight, filterWidth] = parseTupleParam(filterSize);
+  let filterShape;
+  if (dataFormat === "channelsLast") {
+    filterShape = [filterHeight, filterWidth, inShape[3], inShape[3]];
+  } else if (dataFormat === "channelsFirst") {
+    filterShape = [filterHeight, filterWidth, inShape[1], inShape[1]];
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  return computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, false, dataFormat);
+}
+function computePool3DInfo(inShape, filterSize, strides, dilations, pad3, roundingMode, dataFormat = "NDHWC") {
+  const [filterDepth, filterHeight, filterWidth] = parse3TupleParam(filterSize);
+  let filterShape;
+  let $dataFormat;
+  if (dataFormat === "NDHWC") {
+    $dataFormat = "channelsLast";
+    filterShape = [filterDepth, filterHeight, filterWidth, inShape[4], inShape[4]];
+  } else if (dataFormat === "NCDHW") {
+    $dataFormat = "channelsFirst";
+    filterShape = [filterDepth, filterHeight, filterWidth, inShape[1], inShape[1]];
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  return computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, false, $dataFormat, roundingMode);
+}
+function computeConv2DInfo(inShape, filterShape, strides, dilations, pad3, roundingMode, depthwise = false, dataFormat = "channelsLast") {
+  let [batchSize, inHeight, inWidth, inChannels] = [-1, -1, -1, -1];
+  if (dataFormat === "channelsLast") {
+    [batchSize, inHeight, inWidth, inChannels] = inShape;
+  } else if (dataFormat === "channelsFirst") {
+    [batchSize, inChannels, inHeight, inWidth] = inShape;
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  const [filterHeight, filterWidth, , filterChannels] = filterShape;
+  const [strideHeight, strideWidth] = parseTupleParam(strides);
+  const [dilationHeight, dilationWidth] = parseTupleParam(dilations);
+  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);
+  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);
+  const { padInfo, outHeight, outWidth } = getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, effectiveFilterHeight, effectiveFilterWidth, roundingMode, dataFormat);
+  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;
+  let outShape;
+  if (dataFormat === "channelsFirst") {
+    outShape = [batchSize, outChannels, outHeight, outWidth];
+  } else if (dataFormat === "channelsLast") {
+    outShape = [batchSize, outHeight, outWidth, outChannels];
+  }
+  return {
+    batchSize,
+    dataFormat,
+    inHeight,
+    inWidth,
+    inChannels,
+    outHeight,
+    outWidth,
+    outChannels,
+    padInfo,
+    strideHeight,
+    strideWidth,
+    filterHeight,
+    filterWidth,
+    effectiveFilterHeight,
+    effectiveFilterWidth,
+    dilationHeight,
+    dilationWidth,
+    inShape,
+    outShape,
+    filterShape
+  };
+}
+function computeConv3DInfo(inShape, filterShape, strides, dilations, pad3, depthwise = false, dataFormat = "channelsLast", roundingMode) {
+  let [batchSize, inDepth, inHeight, inWidth, inChannels] = [-1, -1, -1, -1, -1];
+  if (dataFormat === "channelsLast") {
+    [batchSize, inDepth, inHeight, inWidth, inChannels] = inShape;
+  } else if (dataFormat === "channelsFirst") {
+    [batchSize, inChannels, inDepth, inHeight, inWidth] = inShape;
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+  const [filterDepth, filterHeight, filterWidth, , filterChannels] = filterShape;
+  const [strideDepth, strideHeight, strideWidth] = parse3TupleParam(strides);
+  const [dilationDepth, dilationHeight, dilationWidth] = parse3TupleParam(dilations);
+  const effectiveFilterDepth = getEffectiveFilterSize(filterDepth, dilationDepth);
+  const effectiveFilterHeight = getEffectiveFilterSize(filterHeight, dilationHeight);
+  const effectiveFilterWidth = getEffectiveFilterSize(filterWidth, dilationWidth);
+  const { padInfo, outDepth, outHeight, outWidth } = get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, effectiveFilterDepth, effectiveFilterHeight, effectiveFilterWidth, roundingMode);
+  const outChannels = depthwise ? filterChannels * inChannels : filterChannels;
+  let outShape;
+  if (dataFormat === "channelsFirst") {
+    outShape = [batchSize, outChannels, outDepth, outHeight, outWidth];
+  } else if (dataFormat === "channelsLast") {
+    outShape = [batchSize, outDepth, outHeight, outWidth, outChannels];
+  }
+  return {
+    batchSize,
+    dataFormat,
+    inDepth,
+    inHeight,
+    inWidth,
+    inChannels,
+    outDepth,
+    outHeight,
+    outWidth,
+    outChannels,
+    padInfo,
+    strideDepth,
+    strideHeight,
+    strideWidth,
+    filterDepth,
+    filterHeight,
+    filterWidth,
+    effectiveFilterDepth,
+    effectiveFilterHeight,
+    effectiveFilterWidth,
+    dilationDepth,
+    dilationHeight,
+    dilationWidth,
+    inShape,
+    outShape,
+    filterShape
+  };
+}
+function computeOutputShape2D(inShape, fieldSize, stride, zeroPad, roundingMode) {
+  if (zeroPad == null) {
+    zeroPad = computeDefaultPad(inShape, fieldSize, stride);
+  }
+  const inputRows = inShape[0];
+  const inputCols = inShape[1];
+  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  return [outputRows, outputCols];
+}
+function computeOutputShape4D(inShape, fieldSize, outChannels, stride, zeroPad, roundingMode) {
+  if (zeroPad == null) {
+    zeroPad = computeDefaultPad(inShape, fieldSize, stride);
+  }
+  const inputDepth = inShape[0];
+  const inputRows = inShape[1];
+  const inputCols = inShape[2];
+  const outputDepths = round((inputDepth - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputRows = round((inputRows - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  const outputCols = round((inputCols - fieldSize + 2 * zeroPad) / stride + 1, roundingMode);
+  return [outputDepths, outputRows, outputCols, outChannels];
+}
+function computeDefaultPad(inputShape, fieldSize, stride, dilation = 1) {
+  const effectiveFieldSize = getEffectiveFilterSize(fieldSize, dilation);
+  return Math.floor((inputShape[0] * (stride - 1) - stride + effectiveFieldSize) / 2);
+}
+function parseTupleParam(param) {
+  if (typeof param === "number") {
+    return [param, param, param];
+  }
+  if (param.length === 2) {
+    return [param[0], param[1], 1];
+  }
+  return param;
+}
+function parse3TupleParam(param) {
+  return typeof param === "number" ? [param, param, param] : param;
+}
+function getEffectiveFilterSize(filterSize, dilation) {
+  if (dilation <= 1) {
+    return filterSize;
+  }
+  return filterSize + (filterSize - 1) * (dilation - 1);
+}
+function getPadAndOutInfo(pad3, inHeight, inWidth, strideHeight, strideWidth, filterHeight, filterWidth, roundingMode, dataFormat) {
+  let padInfo;
+  let outHeight;
+  let outWidth;
+  if (typeof pad3 === "number") {
+    const padType = pad3 === 0 ? "VALID" : "NUMBER";
+    padInfo = { top: pad3, bottom: pad3, left: pad3, right: pad3, type: padType };
+    const outShape = computeOutputShape2D([inHeight, inWidth], filterHeight, strideHeight, pad3, roundingMode);
+    outHeight = outShape[0];
+    outWidth = outShape[1];
+  } else if (pad3 === "same") {
+    outHeight = Math.ceil(inHeight / strideHeight);
+    outWidth = Math.ceil(inWidth / strideWidth);
+    const padAlongHeight = Math.max(0, (outHeight - 1) * strideHeight + filterHeight - inHeight);
+    const padAlongWidth = Math.max(0, (outWidth - 1) * strideWidth + filterWidth - inWidth);
+    const top = Math.floor(padAlongHeight / 2);
+    const bottom = padAlongHeight - top;
+    const left = Math.floor(padAlongWidth / 2);
+    const right = padAlongWidth - left;
+    padInfo = { top, bottom, left, right, type: "SAME" };
+  } else if (pad3 === "valid") {
+    padInfo = { top: 0, bottom: 0, left: 0, right: 0, type: "VALID" };
+    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);
+    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);
+  } else if (typeof pad3 === "object") {
+    const top = dataFormat === "channelsLast" ? pad3[1][0] : pad3[2][0];
+    const bottom = dataFormat === "channelsLast" ? pad3[1][1] : pad3[2][1];
+    const left = dataFormat === "channelsLast" ? pad3[2][0] : pad3[3][0];
+    const right = dataFormat === "channelsLast" ? pad3[2][1] : pad3[3][1];
+    const padType = top === 0 && bottom === 0 && left === 0 && right === 0 ? "VALID" : "EXPLICIT";
+    padInfo = { top, bottom, left, right, type: padType };
+    outHeight = round((inHeight - filterHeight + top + bottom) / strideHeight + 1, roundingMode);
+    outWidth = round((inWidth - filterWidth + left + right) / strideWidth + 1, roundingMode);
+  } else {
+    throw Error(`Unknown padding parameter: ${pad3}`);
+  }
+  return { padInfo, outHeight, outWidth };
+}
+function get3DPadAndOutInfo(pad3, inDepth, inHeight, inWidth, strideDepth, strideHeight, strideWidth, filterDepth, filterHeight, filterWidth, roundingMode) {
+  let padInfo;
+  let outDepth;
+  let outHeight;
+  let outWidth;
+  if (typeof pad3 === "number") {
+    const padType = pad3 === 0 ? "VALID" : "NUMBER";
+    padInfo = {
+      top: pad3,
+      bottom: pad3,
+      left: pad3,
+      right: pad3,
+      front: pad3,
+      back: pad3,
+      type: padType
+    };
+    const outShape = computeOutputShape4D([inDepth, inHeight, inWidth, 1], filterDepth, 1, strideDepth, pad3, roundingMode);
+    outDepth = outShape[0];
+    outHeight = outShape[1];
+    outWidth = outShape[2];
+  } else if (pad3 === "same") {
+    outDepth = Math.ceil(inDepth / strideDepth);
+    outHeight = Math.ceil(inHeight / strideHeight);
+    outWidth = Math.ceil(inWidth / strideWidth);
+    const padAlongDepth = (outDepth - 1) * strideDepth + filterDepth - inDepth;
+    const padAlongHeight = (outHeight - 1) * strideHeight + filterHeight - inHeight;
+    const padAlongWidth = (outWidth - 1) * strideWidth + filterWidth - inWidth;
+    const front = Math.floor(padAlongDepth / 2);
+    const back = padAlongDepth - front;
+    const top = Math.floor(padAlongHeight / 2);
+    const bottom = padAlongHeight - top;
+    const left = Math.floor(padAlongWidth / 2);
+    const right = padAlongWidth - left;
+    padInfo = { top, bottom, left, right, front, back, type: "SAME" };
+  } else if (pad3 === "valid") {
+    padInfo = {
+      top: 0,
+      bottom: 0,
+      left: 0,
+      right: 0,
+      front: 0,
+      back: 0,
+      type: "VALID"
+    };
+    outDepth = Math.ceil((inDepth - filterDepth + 1) / strideDepth);
+    outHeight = Math.ceil((inHeight - filterHeight + 1) / strideHeight);
+    outWidth = Math.ceil((inWidth - filterWidth + 1) / strideWidth);
+  } else {
+    throw Error(`Unknown padding parameter: ${pad3}`);
+  }
+  return { padInfo, outDepth, outHeight, outWidth };
+}
+function round(value, roundingMode) {
+  if (!roundingMode) {
+    return Math.trunc(value);
+  }
+  switch (roundingMode) {
+    case "round":
+      return Math.round(value);
+    case "ceil":
+      return Math.ceil(value);
+    case "floor":
+      return Math.floor(value);
+    default:
+      throw new Error(`Unknown roundingMode ${roundingMode}`);
+  }
+}
+function tupleValuesAreOne(param) {
+  const [dimA, dimB, dimC] = parseTupleParam(param);
+  return dimA === 1 && dimB === 1 && dimC === 1;
+}
+function eitherStridesOrDilationsAreOne(strides, dilations) {
+  return tupleValuesAreOne(strides) || tupleValuesAreOne(dilations);
+}
+function convertConv2DDataFormat(dataFormat) {
+  if (dataFormat === "NHWC") {
+    return "channelsLast";
+  } else if (dataFormat === "NCHW") {
+    return "channelsFirst";
+  } else {
+    throw new Error(`Unknown dataFormat ${dataFormat}`);
+  }
+}
+function checkPadOnDimRoundingMode(opDesc, pad3, dimRoundingMode) {
+  if (dimRoundingMode != null) {
+    if (typeof pad3 === "string") {
+      throw Error(`Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);
+    } else if (typeof pad3 === "number") {
+      assert(isInt(pad3), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${pad3}.`);
+    } else if (typeof pad3 === "object") {
+      pad3.forEach((p2) => {
+        p2.forEach((v) => {
+          assert(isInt(v), () => `Error in ${opDesc}: pad must be an integer when using dimRoundingMode ${dimRoundingMode} but got pad ${v}.`);
+        });
+      });
+    } else {
+      throw Error(`Error in ${opDesc}: Unknown padding parameter: ${pad3}`);
+    }
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reshape.js
+function reshape_(x, shape) {
+  const $x = convertToTensor(x, "x", "reshape", "string_or_numeric");
+  const inputs = { x: $x };
+  const attrs = { shape };
+  return ENGINE.runKernel(Reshape, inputs, attrs);
+}
+var reshape = op({ reshape_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool.js
+function avgPool_(x, filterSize, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "avgPool", "float32");
+  const dilations = 1;
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in avgPool: x must be rank 4 but got rank ${x4D.rank}.`);
+  checkPadOnDimRoundingMode("avgPool", pad3, dimRoundingMode);
+  const inputs = { x: x4D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  let res = ENGINE.runKernel(AvgPool, inputs, attrs);
+  res = cast(res, $x.dtype);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var avgPool = op({ avgPool_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_3d.js
+function avgPool3d_(x, filterSize, strides, pad3, dimRoundingMode, dataFormat = "NDHWC") {
+  const $x = convertToTensor(x, "x", "avgPool3d", "float32");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in avgPool3d: x must be rank 5 but got rank ${x5D.rank}.`);
+  assert(dataFormat === "NDHWC", () => `Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  checkPadOnDimRoundingMode("avgPool3d", pad3, dimRoundingMode);
+  const inputs = { x: x5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };
+  let res = ENGINE.runKernel(AvgPool3D, inputs, attrs);
+  res = cast(res, x5D.dtype);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var avgPool3d = op({ avgPool3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat.js
+function concat_(tensors, axis = 0) {
+  assert(tensors.length >= 1, () => "Pass at least one tensor to concat");
+  const $tensors = convertToTensorArray(tensors, "tensors", "concat", "string_or_numeric");
+  if ($tensors[0].dtype === "complex64") {
+    $tensors.forEach((tensor2) => {
+      if (tensor2.dtype !== "complex64") {
+        throw new Error(`Cannot concatenate complex64 tensors with a tensor
+          with dtype ${tensor2.dtype}. `);
+      }
+    });
+  }
+  if ($tensors.length === 1) {
+    return clone($tensors[0]);
+  }
+  const inputs = $tensors;
+  const attr = { axis };
+  return ENGINE.runKernel(Concat, inputs, attr);
+}
+var concat = op({ concat_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sigmoid.js
+function sigmoid_(x) {
+  const $x = convertToTensor(x, "x", "sigmoid", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sigmoid, inputs);
+}
+var sigmoid = op({ sigmoid_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice.js
+function slice_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice", "string_or_numeric");
+  if ($x.rank === 0) {
+    throw new Error("Slicing scalar is not possible");
+  }
+  const inputs = { x: $x };
+  const attrs = { begin, size };
+  return ENGINE.runKernel(Slice, inputs, attrs);
+}
+var slice = op({ slice_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tanh.js
+function tanh_(x) {
+  const $x = convertToTensor(x, "x", "tanh", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Tanh, inputs);
+}
+var tanh2 = op({ tanh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/basic_lstm_cell.js
+function basicLSTMCell_(forgetBias, lstmKernel, lstmBias, data, c, h) {
+  const $forgetBias = convertToTensor(forgetBias, "forgetBias", "basicLSTMCell");
+  const $lstmKernel = convertToTensor(lstmKernel, "lstmKernel", "basicLSTMCell");
+  const $lstmBias = convertToTensor(lstmBias, "lstmBias", "basicLSTMCell");
+  const $data = convertToTensor(data, "data", "basicLSTMCell");
+  const $c = convertToTensor(c, "c", "basicLSTMCell");
+  const $h = convertToTensor(h, "h", "basicLSTMCell");
+  const combined = concat([$data, $h], 1);
+  const weighted = matMul(combined, $lstmKernel);
+  const res = add2(weighted, $lstmBias);
+  const batchSize = res.shape[0];
+  const sliceCols = res.shape[1] / 4;
+  const sliceSize = [batchSize, sliceCols];
+  const i = slice(res, [0, 0], sliceSize);
+  const j = slice(res, [0, sliceCols], sliceSize);
+  const f = slice(res, [0, sliceCols * 2], sliceSize);
+  const o = slice(res, [0, sliceCols * 3], sliceSize);
+  const newC = add2(mul(sigmoid(i), tanh2(j)), mul($c, sigmoid(add2($forgetBias, f))));
+  const newH = mul(tanh2(newC), sigmoid(o));
+  return [newC, newH];
+}
+var basicLSTMCell = op({ basicLSTMCell_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batch_to_space_nd.js
+function batchToSpaceND_(x, blockShape, crops) {
+  const $x = convertToTensor(x, "x", "batchToSpaceND");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  assert($x.rank >= 1 + blockShape.length, () => `input rank is ${$x.rank} but should be > than blockShape.length ${blockShape.length}`);
+  assert(crops.length === blockShape.length, () => `crops.length is ${crops.length} but should be equal to blockShape.length  ${blockShape.length}`);
+  assert($x.shape[0] % prod5 === 0, () => `input tensor batch is ${$x.shape[0]} but is not divisible by the product of the elements of blockShape ${blockShape.join(" * ")} === ${prod5}`);
+  const inputs = { x: $x };
+  const attrs = { blockShape, crops };
+  return ENGINE.runKernel(BatchToSpaceND, inputs, attrs);
+}
+var batchToSpaceND = op({ batchToSpaceND_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm_util.js
+function xAs4D(x) {
+  let x4D;
+  if (x.rank === 0 || x.rank === 1) {
+    x4D = reshape(x, [1, 1, 1, x.size]);
+  } else if (x.rank === 2) {
+    x4D = reshape(x, [1, 1, x.shape[0], x.shape[1]]);
+  } else if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  } else {
+    x4D = x;
+  }
+  return x4D;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm.js
+function batchNorm_(x, mean4, variance, offset, scale2, varianceEpsilon) {
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale2 != null) {
+    $scale = convertToTensor(scale2, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($mean.rank === $variance.rank, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  assert($offset == null || $mean.rank === $offset.rank, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  assert($scale == null || $mean.rank === $scale.rank, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  const x4D = xAs4D($x);
+  const inputs = {
+    x: x4D,
+    scale: $scale,
+    offset: $offset,
+    mean: $mean,
+    variance: $variance
+  };
+  const attrs = { varianceEpsilon };
+  const res = ENGINE.runKernel(FusedBatchNorm, inputs, attrs);
+  return reshape(res, $x.shape);
+}
+var batchNorm = op({ batchNorm_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm2d.js
+function batchNorm2d_(x, mean4, variance, offset, scale2, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale2 != null) {
+    $scale = convertToTensor(scale2, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 2, () => `Error in batchNorm2D: x must be rank 2 but got rank ${$x.rank}.`);
+  assert($mean.rank === 2 || $mean.rank === 1, () => `Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 2 || $variance.rank === 1, () => `Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 2 || $scale.rank === 1, () => `Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 2 || $offset.rank === 1, () => `Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm2d = op({ batchNorm2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm3d.js
+function batchNorm3d_(x, mean4, variance, offset, scale2, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale2 != null) {
+    $scale = convertToTensor(scale2, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 3, () => `Error in batchNorm3D: x must be rank 3 but got rank ${$x.rank}.`);
+  assert($mean.rank === 3 || $mean.rank === 1, () => `Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 3 || $variance.rank === 1, () => `Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 3 || $scale.rank === 1, () => `Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 3 || $offset.rank === 1, () => `Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm3d = op({ batchNorm3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/batchnorm4d.js
+function batchNorm4d_(x, mean4, variance, offset, scale2, varianceEpsilon) {
+  const $x = convertToTensor(x, "x", "batchNorm");
+  const $mean = convertToTensor(mean4, "mean", "batchNorm");
+  const $variance = convertToTensor(variance, "variance", "batchNorm");
+  let $scale;
+  if (scale2 != null) {
+    $scale = convertToTensor(scale2, "scale", "batchNorm");
+  }
+  let $offset;
+  if (offset != null) {
+    $offset = convertToTensor(offset, "offset", "batchNorm");
+  }
+  assert($x.rank === 4, () => `Error in batchNorm4D: x must be rank 4 but got rank ${$x.rank}.`);
+  assert($mean.rank === 4 || $mean.rank === 1, () => `Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${$mean.rank}.`);
+  assert($variance.rank === 4 || $variance.rank === 1, () => `Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${$variance.rank}.`);
+  if ($scale != null) {
+    assert($scale.rank === 4 || $scale.rank === 1, () => `Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${$scale.rank}.`);
+  }
+  if ($offset != null) {
+    assert($offset.rank === 4 || $offset.rank === 1, () => `Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${$offset.rank}.`);
+  }
+  return batchNorm($x, $mean, $variance, $offset, $scale, varianceEpsilon);
+}
+var batchNorm4d = op({ batchNorm4d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/bincount.js
+function bincount_(x, weights, size) {
+  const $x = convertToTensor(x, "x", "bincount");
+  const $weights = convertToTensor(weights, "weights", "bincount");
+  assert($x.dtype === "int32", () => `Error in bincount: input dtype must be int32, but got ${$x.dtype}`);
+  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);
+  assert($weights.size === $x.size || $weights.size === 0, () => `Error in bincount: weights must have the same size as input or0-length, but got input shape: ${$x.shape}, weights shape: ${$weights.shape}.`);
+  const inputs = { x: $x, weights: $weights };
+  const attrs = { size };
+  return ENGINE.runKernel(Bincount, inputs, attrs);
+}
+var bincount = op({ bincount_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_args.js
+function broadcastArgs_(s0, s1) {
+  const shape1Input = convertToTensor(s0, "s0", "broadcastArgs", "int32");
+  const shape2Input = convertToTensor(s1, "s1", "broadcastArgs", "int32");
+  if (shape1Input.rank !== 1) {
+    throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${shape1Input.rank}`);
+  }
+  if (shape2Input.rank !== 1) {
+    throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${shape2Input.rank}`);
+  }
+  const inputs = { s0: shape1Input, s1: shape2Input };
+  return ENGINE.runKernel(BroadcastArgs, inputs);
+}
+var broadcastArgs = op({ broadcastArgs_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/broadcast_to.js
+function broadcastTo_(x, shape) {
+  let input2 = convertToTensor(x, "broadcastTo", "x");
+  const xShape = input2.shape;
+  if (shape.some((d) => !(d > 0) || d % 1 !== 0)) {
+    throw new Error(`broadcastTo(): Invalid broadcast shape [${shape}].`);
+  }
+  if (shape.length < input2.rank) {
+    throw new Error(`broadcastTo(): shape.length=${shape.length} < input.rank=${input2.rank}.`);
+  }
+  if (shape.length > input2.rank) {
+    const newShape = input2.shape.slice();
+    while (newShape.length < shape.length) {
+      newShape.unshift(1);
+    }
+    input2 = reshape(input2, newShape);
+  }
+  const inputShape = input2.shape;
+  const reps = Array.from(shape);
+  for (let i = shape.length - 1; i >= 0; i--) {
+    if (inputShape[i] === shape[i]) {
+      reps[i] = 1;
+    } else if (input2.shape[i] !== 1) {
+      throw new Error(`broadcastTo(): [${xShape}] cannot be broadcast to [${shape}].`);
+    }
+  }
+  const axes = reps.map((n, i) => n > 1 ? i : -1).filter((i) => i >= 0);
+  if (axes.length === 0) {
+    return clone(input2);
+  }
+  const inputs = { x: input2 };
+  const attrs = { reps };
+  return ENGINE.runKernel(Tile, inputs, attrs);
+}
+var broadcastTo = op({ broadcastTo_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ceil.js
+function ceil_(x) {
+  const $x = convertToTensor(x, "x", "ceil", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Ceil, inputs);
+}
+var ceil = op({ ceil_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fill.js
+function fill(shape, value, dtype) {
+  const attrs = { shape, value, dtype };
+  return ENGINE.runKernel(Fill, {}, attrs);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/clip_by_value.js
+function clipByValue_(x, clipValueMin, clipValueMax) {
+  const $x = convertToTensor(x, "x", "clipByValue");
+  assert(clipValueMin <= clipValueMax, () => `Error in clip: min (${clipValueMin}) must be less than or equal to max (${clipValueMax}).`);
+  if (clipValueMin === clipValueMax) {
+    return fill($x.shape, clipValueMin, $x.dtype);
+  }
+  const inputs = { x: $x };
+  const attrs = { clipValueMin, clipValueMax };
+  return ENGINE.runKernel(ClipByValue, inputs, attrs);
+}
+var clipByValue = op({ clipByValue_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat_1d.js
+function concat1d_(tensors) {
+  return concat(tensors, 0);
+}
+var concat1d = op({ concat1d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat_2d.js
+function concat2d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat2d = op({ concat2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat_3d.js
+function concat3d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat3d = op({ concat3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat_4d.js
+function concat4d_(tensors, axis) {
+  return concat(tensors, axis);
+}
+var concat4d = op({ concat4d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d.js
+function conv2d_(x, filter, strides, pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "conv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in conv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("conv2d", pad3, dimRoundingMode);
+  const inDepth = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert(inDepth === $filter.shape[2], () => `Error in conv2d: depth of input (${inDepth}) must match input depth for filter ${$filter.shape[2]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };
+  const res = ENGINE.runKernel(Conv2D, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var conv2d = op({ conv2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv1d.js
+function conv1d_(x, filter, stride, pad3, dataFormat = "NWC", dilation = 1, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv1d");
+  const $filter = convertToTensor(filter, "filter", "conv1d");
+  let x3D = $x;
+  let reshapedTo3D = false;
+  if ($x.rank === 2) {
+    reshapedTo3D = true;
+    x3D = reshape($x, [1, $x.shape[0], $x.shape[1]]);
+  }
+  assert(x3D.rank === 3, () => `Error in conv1d: input must be rank 3, but got rank ${x3D.rank}.`);
+  assert($filter.rank === 3, () => `Error in conv1d: filter must be rank 3, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("conv1d", pad3, dimRoundingMode);
+  assert(x3D.shape[2] === $filter.shape[1], () => `Error in conv1d: depth of input (${x3D.shape[2]}) must match input depth for filter ${$filter.shape[1]}.`);
+  assert(eitherStridesOrDilationsAreOne(stride, dilation), () => `Error in conv1D: Either stride or dilation must be 1. Got stride ${stride} and dilation '${dilation}'`);
+  assert(dataFormat === "NWC", () => `Error in conv1d: got dataFormat of ${dataFormat} but only NWC is currently supported.`);
+  const filter4D = reshape($filter, [1, $filter.shape[0], $filter.shape[1], $filter.shape[2]]);
+  const input4D = reshape(x3D, [x3D.shape[0], 1, x3D.shape[1], x3D.shape[2]]);
+  const strides = [1, stride];
+  const dilations = [1, dilation];
+  const conv2dDataFormat = "NHWC";
+  const res = conv2d(input4D, filter4D, strides, pad3, conv2dDataFormat, dilations, dimRoundingMode);
+  if (reshapedTo3D) {
+    return reshape(res, [res.shape[2], res.shape[3]]);
+  }
+  return reshape(res, [res.shape[0], res.shape[2], res.shape[3]]);
+}
+var conv1d = op({ conv1d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_backprop_input.js
+function conv2DBackpropInput_(xShape, dy, filter, strides, pad3, dataFormat = "NHWC", dimRoundingMode) {
+  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);
+  let xShape4D = xShape;
+  let dy4D = dy;
+  let reshapedTo4D = false;
+  if (dy.rank === 3) {
+    reshapedTo4D = true;
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+    xShape4D = [1, xShape[0], xShape[1], xShape[2]];
+  }
+  assert(xShape4D.length === 4, () => `Error in conv2dDerInput: inShape must be length 4, but got length ${xShape4D.length}.`);
+  assert(dy4D.rank === 4, () => `Error in conv2dDerInput: dy must be rank 4, but got rank ${dy4D.rank}`);
+  assert(filter.rank === 4, () => `Error in conv2dDerInput: filter must be rank 4, but got rank ${filter.rank}`);
+  const inDepth = dataFormat === "NHWC" ? xShape4D[3] : xShape4D[1];
+  const outDepth = dataFormat === "NHWC" ? dy4D.shape[3] : dy4D.shape[1];
+  assert(inDepth === filter.shape[2], () => `Error in conv2dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[2]}.`);
+  assert(outDepth === filter.shape[3], () => `Error in conv2dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[3]}.`);
+  checkPadOnDimRoundingMode("conv2dDerInput", pad3, dimRoundingMode);
+  const inputs = { dy: dy4D, filter };
+  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape: xShape4D };
+  const res = ENGINE.runKernel(Conv2DBackpropInput, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var conv2DBackpropInput = op({ conv2DBackpropInput_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_transpose.js
+function conv2dTranspose_(x, filter, outputShape, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "conv2dTranspose");
+  const $filter = convertToTensor(filter, "filter", "conv2dTranspose");
+  return conv2DBackpropInput(outputShape, $x, $filter, strides, pad3, "NHWC", dimRoundingMode);
+}
+var conv2dTranspose = op({ conv2dTranspose_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d.js
+function conv3d_(x, filter, strides, pad3, dataFormat = "NDHWC", dilations = [1, 1, 1]) {
+  const $x = convertToTensor(x, "x", "conv3d");
+  const $filter = convertToTensor(filter, "filter", "conv3d");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in conv3d: input must be rank 5, but got rank ${x5D.rank}.`);
+  assert($filter.rank === 5, () => `Error in conv3d: filter must be rank 5, but got rank ${$filter.rank}.`);
+  assert(x5D.shape[4] === $filter.shape[3], () => `Error in conv3d: depth of input (${x5D.shape[4]}) must match input depth for filter ${$filter.shape[3]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv3D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  assert(dataFormat === "NDHWC", () => `Error in conv3d: got dataFormat of ${dataFormat} but only NDHWC is currently supported.`);
+  const inputs = { x: x5D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations };
+  const res = ENGINE.runKernel(Conv3D, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var conv3d = op({ conv3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_backprop_input.js
+function conv3DBackpropInput_(xShape, dy, filter, strides, pad3) {
+  assert(xShape.length === dy.rank, () => `Length of inShape (${xShape.length}) and rank of dy (${dy.rank}) must match`);
+  let xShape5D = xShape;
+  let dy5D = dy;
+  let reshapedTo5D = false;
+  if (dy.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);
+    xShape5D = [1, xShape[0], xShape[1], xShape[2], xShape[3]];
+  }
+  const inDepth = xShape5D[4];
+  const outDepth = dy5D.shape[4];
+  assert(xShape5D.length === 5, () => `Error in conv3dDerInput: inShape must be length 5, but got length ${xShape5D.length}.`);
+  assert(dy5D.rank === 5, () => `Error in conv3dDerInput: dy must be rank 5, but got rank ${dy5D.rank}`);
+  assert(filter.rank === 5, () => `Error in conv3dDerInput: filter must be rank 5, but got rank ${filter.rank}`);
+  assert(inDepth === filter.shape[3], () => `Error in conv3dDerInput: depth of input (${inDepth}) must match input depth for filter ${filter.shape[3]}.`);
+  assert(outDepth === filter.shape[4], () => `Error in conv3dDerInput: depth of output (${outDepth}) must match output depth for filter ${filter.shape[4]}.`);
+  const inputs = { dy: dy5D, filter };
+  const attrs = { pad: pad3, strides, inputShape: xShape5D };
+  const res = ENGINE.runKernel(Conv3DBackpropInputV2, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var conv3DBackpropInput = op({ conv3DBackpropInput_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_transpose.js
+function conv3dTranspose_(x, filter, outputShape, strides, pad3) {
+  const $x = convertToTensor(x, "x", "conv3dTranspose");
+  const $filter = convertToTensor(filter, "filter", "conv3dTranspose");
+  return conv3DBackpropInput(outputShape, $x, $filter, strides, pad3);
+}
+var conv3dTranspose = op({ conv3dTranspose_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/cos.js
+function cos_(x) {
+  const $x = convertToTensor(x, "x", "cos", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Cos, inputs);
+}
+var cos = op({ cos_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/cosh.js
+function cosh_(x) {
+  const $x = convertToTensor(x, "x", "cosh", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Cosh, inputs);
+}
+var cosh = op({ cosh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/cumprod.js
+function cumprod_(x, axis = 0, exclusive = false, reverse5 = false) {
+  const $x = convertToTensor(x, "x", "cumprod");
+  const inputs = { x: $x };
+  const attrs = { axis, exclusive, reverse: reverse5 };
+  return ENGINE.runKernel(Cumprod, inputs, attrs);
+}
+var cumprod = op({ cumprod_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/cumsum.js
+function cumsum_(x, axis = 0, exclusive = false, reverse5 = false) {
+  const $x = convertToTensor(x, "x", "cumsum");
+  const inputs = { x: $x };
+  const attrs = { axis, exclusive, reverse: reverse5 };
+  return ENGINE.runKernel(Cumsum, inputs, attrs);
+}
+var cumsum = op({ cumsum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/dense_bincount.js
+function denseBincount_(x, weights, size, binaryOutput = false) {
+  const $x = convertToTensor(x, "x", "denseBincount");
+  const $weights = convertToTensor(weights, "weights", "denseBincount");
+  assert($x.dtype === "int32", () => `Error in denseBincount: input dtype must be int32, but got ${$x.dtype}`);
+  assert($x.rank <= 2, () => `Error in denseBincount: input must be at most rank 2, but got rank ${$x.rank}.`);
+  assert(size >= 0, () => `size must be non-negative, but got ${size}.`);
+  assert($weights.size === $x.size || $weights.size === 0, () => `Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${$x.shape}, weights shape: ${$weights.shape}.`);
+  const inputs = { x: $x, weights: $weights };
+  const attrs = { size, binaryOutput };
+  return ENGINE.runKernel(DenseBincount, inputs, attrs);
+}
+var denseBincount = op({ denseBincount_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/depth_to_space.js
+function depthToSpace_(x, blockSize, dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "depthToSpace", "float32");
+  const inputHeight = dataFormat === "NHWC" ? $x.shape[1] : $x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? $x.shape[2] : $x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? $x.shape[3] : $x.shape[1];
+  assert(blockSize > 1, () => `blockSize should be > 1 for depthToSpace, but was: ${blockSize}`);
+  assert(inputHeight * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying
+    ${inputHeight} and ${blockSize}  for depthToSpace with input shape
+    ${$x.shape}`);
+  assert(inputWidth * blockSize >= 0, () => `Negative dimension size caused by overflow when multiplying
+    ${inputWidth} and ${blockSize} for depthToSpace with input shape
+        ${$x.shape}`);
+  assert(inputDepth % (blockSize * blockSize) === 0, () => `Dimension size must be evenly divisible by ${blockSize * blockSize} but is ${inputDepth} for depthToSpace with input shape ${$x.shape}`);
+  const inputs = { x: $x };
+  const attrs = { blockSize, dataFormat };
+  return ENGINE.runKernel(DepthToSpace, inputs, attrs);
+}
+var depthToSpace = op({ depthToSpace_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d.js
+function depthwiseConv2d_(x, filter, strides, pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "depthwiseConv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "depthwiseConv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  const inChannels = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert(inChannels === $filter.shape[2], () => `Error in depthwiseConv2d: number of input channels (${inChannels}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);
+  checkPadOnDimRoundingMode("depthwiseConv2d", pad3, dimRoundingMode);
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dataFormat, dilations, dimRoundingMode };
+  const res = ENGINE.runKernel(DepthwiseConv2dNative, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var depthwiseConv2d = op({ depthwiseConv2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/diag.js
+function diag_(x) {
+  const $x = convertToTensor(x, "x", "diag");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Diag, inputs);
+}
+var diag = op({ diag_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/dilation2d.js
+function dilation2d_(x, filter, strides, pad3, dilations = [1, 1], dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "dilation2d");
+  const $filter = convertToTensor(filter, "filter", "dilation2d");
+  assert($x.rank === 3 || $x.rank === 4, () => `Error in dilation2d: input must be rank 3 or 4, but got rank ${$x.rank}.`);
+  assert($filter.rank === 3, () => `Error in dilation2d: filter must be rank 3, but got rank ${$filter.rank}.`);
+  assert(dataFormat === "NHWC", () => `Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+    reshapedTo4D = true;
+  }
+  const inputs = { x: x4D, filter: $filter };
+  const attrs = { strides, pad: pad3, dilations };
+  const res = ENGINE.runKernel(Dilation2D, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var dilation2d = op({ dilation2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/equal.js
+function equal_(a, b) {
+  let $a = convertToTensor(a, "a", "equal", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "equal", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Equal, inputs);
+}
+var equal = op({ equal_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/where.js
+function where_(condition, a, b) {
+  const $a = convertToTensor(a, "a", "where");
+  const $b = convertToTensor(b, "b", "where");
+  const $condition = convertToTensor(condition, "condition", "where", "bool");
+  const broadcastShape = assertAndGetBroadcastShape(assertAndGetBroadcastShape($condition.shape, $a.shape), $b.shape);
+  const $broadcastedCondition = broadcastTo($condition, broadcastShape);
+  const $broadcastedA = broadcastTo($a, broadcastShape);
+  const $broadcastedB = broadcastTo($b, broadcastShape);
+  const inputs = {
+    condition: $broadcastedCondition,
+    t: $broadcastedA,
+    e: $broadcastedB
+  };
+  return ENGINE.runKernel(Select, inputs);
+}
+var where = op({ where_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/zeros_like.js
+function zerosLike_(x) {
+  const $x = convertToTensor(x, "x", "zerosLike");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(ZerosLike, inputs);
+}
+var zerosLike = op({ zerosLike_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/div_no_nan.js
+function divNoNan_(a, b) {
+  let $a = convertToTensor(a, "a", "div");
+  let $b = convertToTensor(b, "b", "div");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const divResult = div($a, $b);
+  const zeros4 = zerosLike(divResult);
+  const bEqualsZero = equal($b, zeros4);
+  return where(bEqualsZero, zeros4, divResult);
+}
+var divNoNan = op({ divNoNan_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/dot.js
+function dot_(t1, t2) {
+  const $t1 = convertToTensor(t1, "t1", "dot");
+  const $t2 = convertToTensor(t2, "t2", "dot");
+  assert(($t1.rank === 1 || $t1.rank === 2) && ($t2.rank === 1 || $t2.rank === 2), () => `Error in dot: inputs must all be rank 1 or 2, but got ranks ${$t1.rank} and ${$t2.rank}.`);
+  const t1Inner = $t1.rank === 1 ? $t1.size : $t1.shape[1];
+  const t2Inner = $t2.rank === 1 ? $t2.size : $t2.shape[0];
+  assert(t1Inner === t2Inner, () => `Error in dot: inner dimensions of inputs must match, but got ${t1Inner} and ${t2Inner}.`);
+  if ($t1.rank === 1 && $t2.rank === 1) {
+    const t12D = reshape($t1, [1, -1]);
+    const t22D = reshape($t2, [-1, 1]);
+    const t1t2 = matMul(t12D, t22D);
+    return reshape(t1t2, []);
+  } else if ($t1.rank === 1 && $t2.rank === 2) {
+    const t12D = reshape($t1, [1, -1]);
+    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);
+    const t1t2 = matMul(t12D, t22D);
+    return reshape(t1t2, [t1t2.size]);
+  } else if ($t1.rank === 2 && $t2.rank === 1) {
+    const t22D = reshape($t2, [-1, 1]);
+    const t1t2 = matMul($t1, t22D);
+    return reshape(t1t2, [t1t2.size]);
+  } else {
+    const t22D = reshape($t2, [$t2.shape[0], $t2.shape[1]]);
+    const t1t2 = matMul($t1, t22D);
+    return t1t2;
+  }
+}
+var dot = op({ dot_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/einsum.js
+function einsum_(equation, ...tensors) {
+  const $tensors = tensors.map((t, i) => convertToTensor(t, `tensors${i}`, "einsum"));
+  const attrs = { equation };
+  return ENGINE.runKernel(Einsum, $tensors, attrs);
+}
+var einsum = op({ einsum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/elu.js
+function elu_(x) {
+  const $x = convertToTensor(x, "x", "elu", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Elu, inputs);
+}
+var elu = op({ elu_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/erf.js
+function erf_(x) {
+  let $x = convertToTensor(x, "x", "erf");
+  assert($x.dtype === "int32" || $x.dtype === "float32", () => "Input dtype must be `int32` or `float32`.");
+  if ($x.dtype === "int32") {
+    $x = cast($x, "float32");
+  }
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Erf, inputs);
+}
+var erf = op({ erf_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/axis_util.js
+function axesAreInnerMostDims(axes, rank) {
+  for (let i = 0; i < axes.length; ++i) {
+    if (axes[axes.length - i - 1] !== rank - 1 - i) {
+      return false;
+    }
+  }
+  return true;
+}
+function combineLocations(outputLoc, reduceLoc, axes) {
+  const rank = outputLoc.length + reduceLoc.length;
+  const loc = [];
+  let outIdx = 0;
+  let reduceIdx = 0;
+  for (let dim = 0; dim < rank; dim++) {
+    if (axes.indexOf(dim) === -1) {
+      loc.push(outputLoc[outIdx++]);
+    } else {
+      loc.push(reduceLoc[reduceIdx++]);
+    }
+  }
+  return loc;
+}
+function computeOutAndReduceShapes(aShape, axes) {
+  const outShape = [];
+  const rank = aShape.length;
+  for (let dim = 0; dim < rank; dim++) {
+    if (axes.indexOf(dim) === -1) {
+      outShape.push(aShape[dim]);
+    }
+  }
+  const reduceShape = axes.map((dim) => aShape[dim]);
+  return [outShape, reduceShape];
+}
+function expandShapeToKeepDim(shape, axes) {
+  const reduceSubShape = axes.map((x) => 1);
+  return combineLocations(shape, reduceSubShape, axes);
+}
+function assertAxesAreInnerMostDims(msg, axes, rank) {
+  assert(axesAreInnerMostDims(axes, rank), () => `${msg} supports only inner-most axes for now. Got axes ${axes} and rank-${rank} input.`);
+}
+function getAxesPermutation(axes, rank) {
+  if (axesAreInnerMostDims(axes, rank)) {
+    return null;
+  }
+  const result = [];
+  for (let i = 0; i < rank; ++i) {
+    if (axes.indexOf(i) === -1) {
+      result.push(i);
+    }
+  }
+  axes.forEach((axis) => result.push(axis));
+  return result;
+}
+function getUndoAxesPermutation(axes) {
+  return axes.map((axis, i) => [i, axis]).sort((a, b) => a[1] - b[1]).map((x) => x[0]);
+}
+function getInnerMostAxes(numAxes, rank) {
+  const res = [];
+  for (let i = rank - numAxes; i < rank; ++i) {
+    res.push(i);
+  }
+  return res;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max.js
+function max_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "max");
+  const inputs = { x: $x };
+  const attrs = { reductionIndices: axis, keepDims };
+  return ENGINE.runKernel(Max, inputs, attrs);
+}
+var max = op({ max_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/min.js
+function min_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "min");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Min, inputs, attrs);
+}
+var min = op({ min_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pow.js
+function pow_(base, exp4) {
+  let $base = convertToTensor(base, "base", "pow");
+  let $exp = convertToTensor(exp4, "exp", "pow");
+  [$base, $exp] = makeTypesMatch($base, $exp);
+  const inputs = { a: $base, b: $exp };
+  return ENGINE.runKernel(Pow, inputs);
+}
+var pow = op({ pow_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/scalar.js
+function scalar(value, dtype) {
+  if ((isTypedArray(value) && dtype !== "string" || Array.isArray(value)) && dtype !== "complex64") {
+    throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");
+  }
+  if (dtype === "string" && isTypedArray(value) && !(value instanceof Uint8Array)) {
+    throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");
+  }
+  const shape = [];
+  const inferredShape = [];
+  return makeTensor(value, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sqrt.js
+function sqrt_(x) {
+  const $x = convertToTensor(x, "x", "sqrt", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sqrt, inputs);
+}
+var sqrt = op({ sqrt_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/square.js
+function square_(x) {
+  const $x = convertToTensor(x, "x", "square");
+  const attrs = {};
+  return ENGINE.runKernel("Square", { x: $x }, attrs);
+}
+var square = op({ square_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sum.js
+function sum_(x, axis = null, keepDims = false) {
+  let $x = convertToTensor(x, "x", "sum");
+  if ($x.dtype === "bool") {
+    $x = cast($x, "int32");
+  }
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Sum, inputs, attrs);
+}
+var sum2 = op({ sum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/norm.js
+function norm_(x, ord = "euclidean", axis = null, keepDims = false) {
+  x = convertToTensor(x, "x", "norm");
+  const norm2 = normImpl(x, ord, axis);
+  let keepDimsShape = norm2.shape;
+  if (keepDims) {
+    const axes = parseAxisParam(axis, x.shape);
+    keepDimsShape = expandShapeToKeepDim(norm2.shape, axes);
+  }
+  return reshape(norm2, keepDimsShape);
+}
+function normImpl(x, p2, axis = null) {
+  if (x.rank === 0) {
+    return abs(x);
+  }
+  if (x.rank !== 1 && axis === null) {
+    return normImpl(reshape(x, [-1]), p2, axis);
+  }
+  if (x.rank === 1 || typeof axis === "number" || Array.isArray(axis) && axis.length === 1) {
+    if (p2 === 1) {
+      return sum2(abs(x), axis);
+    }
+    if (p2 === Infinity) {
+      return max(abs(x), axis);
+    }
+    if (p2 === -Infinity) {
+      return min(abs(x), axis);
+    }
+    if (p2 === "euclidean" || p2 === 2) {
+      return sqrt(sum2(pow(abs(x), scalar(2, "int32")), axis));
+    }
+    throw new Error(`Error in norm: invalid ord value: ${p2}`);
+  }
+  if (Array.isArray(axis) && axis.length === 2) {
+    if (p2 === 1) {
+      return max(sum2(abs(x), axis[0]), axis[1] - 1);
+    }
+    if (p2 === Infinity) {
+      return max(sum2(abs(x), axis[1]), axis[0]);
+    }
+    if (p2 === -Infinity) {
+      return min(sum2(abs(x), axis[1]), axis[0]);
+    }
+    if (p2 === "fro" || p2 === "euclidean") {
+      return sqrt(sum2(square(x), axis));
+    }
+    throw new Error(`Error in norm: invalid ord value: ${p2}`);
+  }
+  throw new Error(`Error in norm: invalid axis: ${axis}`);
+}
+var norm = op({ norm_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/euclidean_norm.js
+function euclideanNorm_(x, axis = null, keepDims = false) {
+  return norm(x, "euclidean", axis, keepDims);
+}
+var euclideanNorm = op({ euclideanNorm_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/exp.js
+function exp_(x) {
+  const $x = convertToTensor(x, "x", "exp");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Exp, inputs);
+}
+var exp = op({ exp_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/expand_dims.js
+function expandDims_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "expandDims", "string_or_numeric");
+  assert(axis <= $x.rank, () => "Axis must be <= rank of the tensor");
+  const inputs = { input: $x };
+  const attrs = { dim: axis };
+  return ENGINE.runKernel(ExpandDims, inputs, attrs);
+}
+var expandDims = op({ expandDims_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/expm1.js
+function expm1_(x) {
+  const $x = convertToTensor(x, "x", "expm1");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Expm1, inputs);
+}
+var expm1 = op({ expm1_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tile.js
+function tile_(x, reps) {
+  const $x = convertToTensor(x, "x", "tile", "string_or_numeric");
+  assert($x.rank === reps.length, () => `Error in transpose: rank of input ${$x.rank} must match length of reps ${reps}.`);
+  const inputs = { x: $x };
+  const attrs = { reps };
+  return ENGINE.runKernel(Tile, inputs, attrs);
+}
+var tile = op({ tile_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/eye.js
+function eye_(numRows, numColumns, batchShape, dtype = "float32") {
+  if (numColumns == null) {
+    numColumns = numRows;
+  }
+  const buff = buffer([numRows, numColumns], dtype);
+  const n = numRows <= numColumns ? numRows : numColumns;
+  for (let i = 0; i < n; ++i) {
+    buff.set(1, i, i);
+  }
+  const out = reshape(buff.toTensor(), [numRows, numColumns]);
+  if (batchShape == null) {
+    return out;
+  } else {
+    if (batchShape.length === 1) {
+      return tile(expandDims(out, 0), [batchShape[0], 1, 1]);
+    } else if (batchShape.length === 2) {
+      return tile(expandDims(expandDims(out, 0), 0), [batchShape[0], batchShape[1], 1, 1]);
+    } else if (batchShape.length === 3) {
+      return tile(expandDims(expandDims(expandDims(out, 0), 0), 0), [
+        batchShape[0],
+        batchShape[1],
+        batchShape[2],
+        1,
+        1
+      ]);
+    } else {
+      throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${batchShape.length}D.`);
+    }
+  }
+}
+var eye = op({ eye_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/floor.js
+function floor_(x) {
+  const $x = convertToTensor(x, "x", "floor", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Floor, inputs);
+}
+var floor = op({ floor_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/gather.js
+function gather_(x, indices, axis = 0, batchDims = 0) {
+  const $x = convertToTensor(x, "x", "gather");
+  const $indices = convertToTensor(indices, "indices", "gather", "int32");
+  const inputs = { x: $x, indices: $indices };
+  const attrs = { axis, batchDims };
+  return ENGINE.runKernel(GatherV2, inputs, attrs);
+}
+var gather = op({ gather_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/greater.js
+function greater_(a, b) {
+  let $a = convertToTensor(a, "a", "greater", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "greater", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Greater, inputs);
+}
+var greater = op({ greater_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/greater_equal.js
+function greaterEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "greaterEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "greaterEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(GreaterEqual, inputs);
+}
+var greaterEqual = op({ greaterEqual_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/is_finite.js
+function isFinite_(x) {
+  const $x = convertToTensor(x, "x", "isFinite");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsFinite, inputs);
+}
+var isFinite2 = op({ isFinite_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/is_inf.js
+function isInf_(x) {
+  const $x = convertToTensor(x, "x", "isInf");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsInf, inputs);
+}
+var isInf = op({ isInf_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/is_nan.js
+function isNaN_(x) {
+  const $x = convertToTensor(x, "x", "isNaN");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(IsNan, inputs);
+}
+var isNaN2 = op({ isNaN_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/leaky_relu.js
+function leakyRelu_(x, alpha = 0.2) {
+  const $x = convertToTensor(x, "x", "leakyRelu");
+  const inputs = { x: $x };
+  const attrs = { alpha };
+  return ENGINE.runKernel(LeakyRelu, inputs, attrs);
+}
+var leakyRelu = op({ leakyRelu_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/less.js
+function less_(a, b) {
+  let $a = convertToTensor(a, "a", "less", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "less", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Less, inputs);
+}
+var less = op({ less_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/less_equal.js
+function lessEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "lessEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "lessEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LessEqual, inputs);
+}
+var lessEqual = op({ lessEqual_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/linspace.js
+function linspace(start, stop, num) {
+  if (num <= 0) {
+    throw new Error("The number of values should be positive.");
+  }
+  const attrs = { start, stop, num };
+  return ENGINE.runKernel(LinSpace, {}, attrs);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/local_response_normalization.js
+function localResponseNormalization_(x, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {
+  const $x = convertToTensor(x, "x", "localResponseNormalization");
+  assert($x.rank === 4 || $x.rank === 3, () => `Error in localResponseNormalization: x must be rank 3 or 4 but got
+               rank ${$x.rank}.`);
+  assert(isInt(depthRadius), () => `Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${depthRadius}.`);
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  const inputs = { x: x4D };
+  const attrs = { depthRadius, bias, alpha, beta };
+  const res = ENGINE.runKernel(LRN, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  } else {
+    return res;
+  }
+}
+var localResponseNormalization = op({ localResponseNormalization_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/log.js
+function log_(x) {
+  const $x = convertToTensor(x, "x", "log", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Log, inputs);
+}
+var log2 = op({ log_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/log1p.js
+function log1p_(x) {
+  const $x = convertToTensor(x, "x", "log1p");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Log1p, inputs);
+}
+var log1p = op({ log1p_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients.js
+function grad(f) {
+  assert(isFunction(f), () => "The f passed in grad(f) must be a function");
+  return (x, dy) => {
+    const $x = convertToTensor(x, "x", "tf.grad", "string_or_numeric");
+    const $dy = dy != null ? convertToTensor(dy, "dy", "tf.grad") : null;
+    return ENGINE.tidy(() => {
+      const { value, grads: grads2 } = ENGINE.gradients(() => f($x), [$x], $dy);
+      if ($dy != null) {
+        assertShapesMatch(value.shape, $dy.shape, "The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)");
+      }
+      checkGrads(grads2);
+      return grads2[0];
+    });
+  };
+}
+function grads(f) {
+  assert(isFunction(f), () => "The f passed in grads(f) must be a function");
+  return (args, dy) => {
+    assert(Array.isArray(args), () => "The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");
+    const $args = convertToTensorArray(args, "args", "tf.grads", "string_or_numeric");
+    const $dy = dy != null ? convertToTensor(dy, "dy", "tf.grads") : null;
+    return ENGINE.tidy(() => {
+      const { value, grads: grads2 } = ENGINE.gradients(() => f(...$args), $args, $dy);
+      if ($dy != null) {
+        assertShapesMatch(value.shape, $dy.shape, "The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])");
+      }
+      checkGrads(grads2);
+      return grads2;
+    });
+  };
+}
+function valueAndGrad(f) {
+  assert(isFunction(f), () => "The f passed in valueAndGrad(f) must be a function");
+  return (x, dy) => {
+    assert(x instanceof Tensor, () => "The x passed in valueAndGrad(f)(x) must be a tensor");
+    assert(dy == null || dy instanceof Tensor, () => "The dy passed in valueAndGrad(f)(x, dy) must be a tensor");
+    const { grads: grads2, value } = ENGINE.gradients(() => f(x), [x], dy);
+    checkGrads(grads2);
+    return { grad: grads2[0], value };
+  };
+}
+function valueAndGrads(f) {
+  assert(isFunction(f), () => "The f passed in valueAndGrads(f) must be a function");
+  return (args, dy) => {
+    assert(Array.isArray(args) && args.every((arg) => arg instanceof Tensor), () => "The args passed in valueAndGrads(f)(args) must be array of tensors");
+    assert(dy == null || dy instanceof Tensor, () => "The dy passed in valueAndGrads(f)(args, dy) must be a tensor");
+    const res = ENGINE.gradients(() => f(...args), args, dy);
+    if (dy != null) {
+      assertShapesMatch(res.value.shape, dy.shape, "The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])");
+    }
+    checkGrads(res.grads);
+    return res;
+  };
+}
+function variableGrads(f, varList) {
+  assert(isFunction(f), () => "The f passed in variableGrads(f) must be a function");
+  assert(varList == null || Array.isArray(varList) && varList.every((v) => v instanceof Variable), () => "The varList passed in variableGrads(f, varList) must be an array of variables");
+  const specifiedVarList = varList != null;
+  if (!specifiedVarList) {
+    varList = [];
+    for (const varName in ENGINE.registeredVariables) {
+      varList.push(ENGINE.registeredVariables[varName]);
+    }
+  }
+  const specifiedNonTrainable = specifiedVarList ? varList.filter((variable2) => !variable2.trainable) : null;
+  const originalVarCount = varList.length;
+  varList = varList.filter((variable2) => variable2.trainable);
+  assert(varList.length > 0, () => `variableGrads() expects at least one of the input variables to be trainable, but none of the ${originalVarCount} variables is trainable.`);
+  const allowNoGradients = true;
+  const { value, grads: grads2 } = ENGINE.gradients(f, varList, null, allowNoGradients);
+  assert(grads2.some((g) => g != null), () => "Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize().");
+  assert(value.rank === 0, () => `The f passed in variableGrads(f) must return a scalar, but it returned a rank-${value.rank} tensor`);
+  const namedGrads = {};
+  varList.forEach((v, i) => {
+    if (grads2[i] != null) {
+      namedGrads[v.name] = grads2[i];
+    }
+  });
+  if (specifiedNonTrainable != null) {
+    specifiedNonTrainable.forEach((v) => namedGrads[v.name] = null);
+  }
+  return { value, grads: namedGrads };
+}
+function customGrad(f) {
+  return ENGINE.customGrad(f);
+}
+function checkGrads(grads2) {
+  const numNullGradients = grads2.filter((g) => g == null).length;
+  if (numNullGradients > 0) {
+    throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
+    the f you passed encloses all operations that lead from x to y.`);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/softplus.js
+function softplus_(x) {
+  const $x = convertToTensor(x, "x", "softplus");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Softplus, inputs);
+}
+var softplus = op({ softplus_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/log_sigmoid.js
+function logSigmoid_(x) {
+  const $x = convertToTensor(x, "x", "logSigmoid");
+  const customOp = customGrad((x2) => {
+    const value = neg(softplus(neg(x2)));
+    const gradFunc = (dy) => {
+      const derX = mul(dy, sigmoid(neg(x2)));
+      return derX;
+    };
+    return { value, gradFunc };
+  });
+  return customOp($x);
+}
+var logSigmoid = op({ logSigmoid_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sub.js
+function sub_(a, b) {
+  let $a = convertToTensor(a, "a", "sub");
+  let $b = convertToTensor(b, "b", "sub");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Sub, inputs);
+}
+var sub = op({ sub_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/log_softmax.js
+function logSoftmax_(logits, axis = -1) {
+  const $logits = convertToTensor(logits, "logits", "logSoftmax");
+  if (axis === -1) {
+    axis = $logits.rank - 1;
+  }
+  if (axis !== $logits.rank - 1) {
+    throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and axis was ${axis}`);
+  }
+  const customOp = customGrad((logits2, save) => {
+    const keepDims = true;
+    const xMax = max(logits2, axis, true);
+    const shifted = sub(logits2, xMax);
+    const value = sub(cast(shifted, "float32"), log2(sum2(exp(shifted), axis, keepDims)));
+    save([value]);
+    const gradFunc = (dy, saved) => {
+      const [value2] = saved;
+      const keepDims2 = true;
+      const softmax6 = exp(value2);
+      return sub(dy, mul(sum2(dy, axis, keepDims2), softmax6));
+    };
+    return { value, gradFunc };
+  });
+  return customOp($logits);
+}
+var logSoftmax = op({ logSoftmax_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/log_sum_exp.js
+function logSumExp_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "logSumExp");
+  const axes = parseAxisParam(axis, $x.shape);
+  const xMax = max($x, axes, true);
+  const a = sub($x, xMax);
+  const b = exp(a);
+  const c = sum2(b, axes);
+  const d = log2(c);
+  const res = add2(reshape(xMax, d.shape), d);
+  if (keepDims) {
+    const newShape = expandShapeToKeepDim(res.shape, axes);
+    return reshape(res, newShape);
+  }
+  return res;
+}
+var logSumExp = op({ logSumExp_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/logical_and.js
+function logicalAnd_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalAnd", "bool");
+  const $b = convertToTensor(b, "b", "logicalAnd", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LogicalAnd, inputs);
+}
+var logicalAnd = op({ logicalAnd_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/logical_not.js
+function logicalNot_(x) {
+  const $x = convertToTensor(x, "x", "logicalNot", "bool");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(LogicalNot, inputs);
+}
+var logicalNot = op({ logicalNot_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/logical_or.js
+function logicalOr_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalOr", "bool");
+  const $b = convertToTensor(b, "b", "logicalOr", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(LogicalOr, inputs);
+}
+var logicalOr = op({ logicalOr_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/logical_xor.js
+function logicalXor_(a, b) {
+  const $a = convertToTensor(a, "a", "logicalXor", "bool");
+  const $b = convertToTensor(b, "b", "logicalXor", "bool");
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  return logicalAnd(logicalOr(a, b), logicalNot(logicalAnd(a, b)));
+}
+var logicalXor = op({ logicalXor_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/search_sorted.js
+var INT32_MAX = 2147483648;
+function searchSorted_(sortedSequence, values, side = "left") {
+  const $sortedSequence = convertToTensor(sortedSequence, "sortedSequence", "searchSorted");
+  const $values = convertToTensor(values, "values", "searchSorted");
+  const sequenceSize = $sortedSequence.shape[$sortedSequence.shape.length - 1];
+  const valuesSize = $values.shape[$values.shape.length - 1];
+  const $sortedSequence2D = reshape($sortedSequence, [-1, sequenceSize]);
+  const $values2D = reshape($values, [-1, valuesSize]);
+  if ($sortedSequence2D.rank < 2) {
+    throw new Error(`Sorted input argument must be at least 2-dimensional`);
+  }
+  if ($sortedSequence2D.shape[0] !== $values2D.shape[0]) {
+    throw new Error(`Leading dimension of 'sortedSequence' and 'values' must match.`);
+  }
+  if (sizeFromShape($values2D.shape) >= INT32_MAX) {
+    throw new Error(`values tensor size must less than ${INT32_MAX}`);
+  }
+  if ($sortedSequence2D.shape[1] >= INT32_MAX) {
+    throw new Error(`trailing dim_size must less than ${INT32_MAX} for int32 output type, was ${$sortedSequence2D.shape[1]}`);
+  }
+  const inputs = {
+    sortedSequence: $sortedSequence2D,
+    values: $values2D
+  };
+  const attrs = { side };
+  return ENGINE.runKernel(SearchSorted, inputs, attrs);
+}
+var searchSorted = op({ searchSorted_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/lower_bound.js
+function lowerBound(sortedSequence, values) {
+  return searchSorted(sortedSequence, values, "left");
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool.js
+function maxPool_(x, filterSize, strides, pad3, dimRoundingMode) {
+  const $x = convertToTensor(x, "x", "maxPool");
+  const dilations = 1;
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x4D.rank}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  checkPadOnDimRoundingMode("maxPool", pad3, dimRoundingMode);
+  const inputs = { x: x4D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(MaxPool, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var maxPool = op({ maxPool_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_3d.js
+function maxPool3d_(x, filterSize = [1, 1, 1], strides, pad3, dimRoundingMode, dataFormat = "NDHWC") {
+  const $x = convertToTensor(x, "x", "maxPool3d");
+  let x5D = $x;
+  let reshapedTo5D = false;
+  if ($x.rank === 4) {
+    reshapedTo5D = true;
+    x5D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2], $x.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in maxPool3d: x must be rank 5 but got rank ${x5D.rank}.`);
+  assert(dataFormat === "NDHWC", () => `Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${dataFormat}`);
+  checkPadOnDimRoundingMode("maxPool3d", pad3, dimRoundingMode);
+  const inputs = { x: x5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat };
+  const res = ENGINE.runKernel(MaxPool3D, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var maxPool3d = op({ maxPool3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_with_argmax.js
+function maxPoolWithArgmax_(x, filterSize, strides, pad3, includeBatchInIndex = false) {
+  const $x = convertToTensor(x, "x", "maxPoolWithArgmax");
+  const inputs = { x: $x };
+  const attrs = { filterSize, strides, pad: pad3, includeBatchInIndex };
+  const result = ENGINE.runKernel(MaxPoolWithArgmax, inputs, attrs);
+  return { result: result[0], indexes: result[1] };
+}
+var maxPoolWithArgmax = op({ maxPoolWithArgmax_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/maximum.js
+function maximum_(a, b) {
+  let $a = convertToTensor(a, "a", "maximum");
+  let $b = convertToTensor(b, "b", "maximum");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "bool") {
+    $a = cast($a, "int32");
+    $b = cast($b, "int32");
+  }
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Maximum, inputs);
+}
+var maximum = op({ maximum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/mean.js
+function mean_(x, axis = null, keepDims = false) {
+  const $x = convertToTensor(x, "x", "mean");
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Mean, inputs, attrs);
+}
+var mean = op({ mean_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/zeros.js
+function zeros(shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = zeros(shape, "float32");
+    const imag4 = zeros(shape, "float32");
+    return complex(real4, imag4);
+  }
+  const values = makeZerosTypedArray(sizeFromShape(shape), dtype);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ones.js
+function ones2(shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = ones2(shape, "float32");
+    const imag4 = zeros(shape, "float32");
+    return complex(real4, imag4);
+  }
+  const values = makeOnesTypedArray(sizeFromShape(shape), dtype);
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/meshgrid.js
+function meshgrid(x, y, { indexing = "xy" } = {}) {
+  if (indexing !== "xy" && indexing !== "ij") {
+    throw new TypeError(`${indexing} is not a valid third argument to meshgrid`);
+  }
+  if (x === void 0) {
+    return [];
+  }
+  let $x = convertToTensor(x, "x", "meshgrid", x instanceof Tensor ? x.dtype : "float32");
+  if (y === void 0) {
+    return [$x];
+  }
+  let $y = convertToTensor(y, "y", "meshgrid", y instanceof Tensor ? y.dtype : "float32");
+  const w = sizeFromShape($x.shape);
+  const h = sizeFromShape($y.shape);
+  if (indexing === "xy") {
+    $x = reshape($x, [1, -1]);
+    $y = reshape($y, [-1, 1]);
+    return [
+      matMul(ones2([h, 1], $x.dtype), $x),
+      matMul($y, ones2([1, w], $y.dtype))
+    ];
+  }
+  $x = reshape($x, [-1, 1]);
+  $y = reshape($y, [1, -1]);
+  return [
+    matMul($x, ones2([1, h], $x.dtype)),
+    matMul(ones2([w, 1], $y.dtype), $y)
+  ];
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/minimum.js
+function minimum_(a, b) {
+  let $a = convertToTensor(a, "a", "minimum");
+  let $b = convertToTensor(b, "b", "minimum");
+  [$a, $b] = makeTypesMatch($a, $b);
+  if ($a.dtype === "bool") {
+    $a = cast($a, "int32");
+    $b = cast($b, "int32");
+  }
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Minimum, inputs);
+}
+var minimum = op({ minimum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/mirror_pad.js
+function mirrorPad_(x, paddings, mode) {
+  assert(mode === "reflect" || mode === "symmetric", () => `Invalid mode. Mode must be either reflect or symmetric. Got ${mode}.`);
+  const $x = convertToTensor(x, "x", "mirrorPad");
+  if ($x.rank === 0) {
+    throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");
+  }
+  assert(paddings.length === $x.rank, () => `Padding doesn't match input. Must be ${$x.rank}. Got ${paddings.length}.`);
+  const shapeOffset = mode === "reflect" ? 1 : 0;
+  for (let i = 0; i < $x.rank; i++) {
+    assert(paddings[i].length === 2, () => `Invalid number of paddings. Must be length of 2 each.`);
+    assert(paddings[i][0] >= 0 && paddings[i][0] <= $x.shape[i] - shapeOffset && paddings[i][1] >= 0 && paddings[i][1] <= $x.shape[i] - shapeOffset, () => `Padding in dimension ${i} cannot be greater than or equal to ${$x.shape[i] - shapeOffset} or less than 0 for input of shape ${$x.shape}`);
+  }
+  const attrs = { paddings, mode };
+  const inputs = { x: $x };
+  return ENGINE.runKernel(MirrorPad, inputs, attrs);
+}
+var mirrorPad = op({ mirrorPad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/mod.js
+function mod_(a, b) {
+  let $a = convertToTensor(a, "a", "mod");
+  let $b = convertToTensor(b, "b", "mod");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(Mod, inputs);
+}
+var mod = op({ mod_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/moments.js
+function moments_(x, axis = null, keepDims = false) {
+  x = convertToTensor(x, "x", "moments");
+  const axes = parseAxisParam(axis, x.shape);
+  const xMean = mean(x, axes, keepDims);
+  let keepDimsShape = xMean.shape;
+  if (!keepDims) {
+    keepDimsShape = expandShapeToKeepDim(xMean.shape, axes);
+  }
+  const devSquared = square(sub(cast(x, "float32"), reshape(xMean, keepDimsShape)));
+  const variance = mean(devSquared, axes, keepDims);
+  return { mean: xMean, variance };
+}
+var moments = op({ moments_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/multi_rnn_cell.js
+function multiRNNCell_(lstmCells, data, c, h) {
+  const $data = convertToTensor(data, "data", "multiRNNCell");
+  const $c = convertToTensorArray(c, "c", "multiRNNCell");
+  const $h = convertToTensorArray(h, "h", "multiRNNCell");
+  let input2 = $data;
+  const newStates = [];
+  for (let i = 0; i < lstmCells.length; i++) {
+    const output = lstmCells[i](input2, $c[i], $h[i]);
+    newStates.push(output[0]);
+    newStates.push(output[1]);
+    input2 = output[1];
+  }
+  const newC = [];
+  const newH = [];
+  for (let i = 0; i < newStates.length; i += 2) {
+    newC.push(newStates[i]);
+    newH.push(newStates[i + 1]);
+  }
+  return [newC, newH];
+}
+var multiRNNCell = op({ multiRNNCell_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/multinomial.js
+function multinomial_(logits, numSamples, seed, normalized = false) {
+  const $logits = convertToTensor(logits, "logits", "multinomial");
+  const numOutcomes = $logits.size;
+  const origRank = $logits.rank;
+  if (numOutcomes < 2) {
+    throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${numOutcomes}.`);
+  }
+  if (origRank > 2) {
+    throw new Error(`Rank of probabilities must be 1 or 2, but is ${origRank}`);
+  }
+  seed = seed || Math.random();
+  const logits2D = origRank === 1 ? reshape($logits, [1, -1]) : $logits;
+  const inputs = { logits: logits2D };
+  const attrs = { numSamples, seed, normalized };
+  const res = ENGINE.runKernel(Multinomial, inputs, attrs);
+  return origRank === 1 ? reshape(res, [res.size]) : res;
+}
+var multinomial = op({ multinomial_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/not_equal.js
+function notEqual_(a, b) {
+  let $a = convertToTensor(a, "a", "notEqual", "string_or_numeric");
+  let $b = convertToTensor(b, "b", "notEqual", "string_or_numeric");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  return ENGINE.runKernel(NotEqual, inputs);
+}
+var notEqual = op({ notEqual_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ones_like.js
+function onesLike_(x) {
+  const $x = convertToTensor(x, "x", "onesLike");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(OnesLike, inputs);
+}
+var onesLike = op({ onesLike_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/outer_product.js
+function outerProduct_(v1, v2) {
+  const $v1 = convertToTensor(v1, "v1", "outerProduct");
+  const $v2 = convertToTensor(v2, "v2", "outerProduct");
+  assert($v1.rank === 1 && $v2.rank === 1, () => `Error in outerProduct: inputs must be rank 1, but got ranks ${$v1.rank} and ${$v2.rank}.`);
+  const v12D = reshape($v1, [-1, 1]);
+  const v22D = reshape($v2, [1, -1]);
+  return matMul(v12D, v22D);
+}
+var outerProduct = op({ outerProduct_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pad.js
+function pad_(x, paddings, constantValue = 0) {
+  const $x = convertToTensor(x, "x", "pad");
+  if ($x.rank === 0) {
+    throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");
+  }
+  const attrs = { paddings, constantValue };
+  const inputs = { x: $x };
+  return ENGINE.runKernel(PadV2, inputs, attrs);
+}
+var pad = op({ pad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pad1d.js
+function pad1d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 2, () => "Invalid number of paddings. Must be length of 2.");
+  return pad(x, [paddings], constantValue);
+}
+var pad1d = op({ pad1d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pad2d.js
+function pad2d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 2 && paddings[0].length === 2 && paddings[1].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad2d = op({ pad2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pad3d.js
+function pad3d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 3 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad3d = op({ pad3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pad4d.js
+function pad4d_(x, paddings, constantValue = 0) {
+  assert(paddings.length === 4 && paddings[0].length === 2 && paddings[1].length === 2 && paddings[2].length === 2 && paddings[3].length === 2, () => "Invalid number of paddings. Must be length of 2 each.");
+  return pad(x, paddings, constantValue);
+}
+var pad4d = op({ pad4d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/space_to_batch_nd.js
+function spaceToBatchND_(x, blockShape, paddings) {
+  const $x = convertToTensor(x, "x", "spaceToBatchND");
+  assert($x.rank >= 1 + blockShape.length, () => `input rank ${$x.rank} should be > than [blockShape] ${blockShape.length}`);
+  assert(paddings.length === blockShape.length, () => `paddings.shape[0] ${paddings.length} must be equal to [blockShape] ${blockShape.length}`);
+  assert($x.shape.reduce((a, b, i) => {
+    if (i > 0 && i <= blockShape.length) {
+      return a && (b + paddings[i - 1][0] + paddings[i - 1][1]) % blockShape[i - 1] === 0;
+    }
+    return a;
+  }, true), () => `input spatial dimensions ${$x.shape.slice(1)} with paddings ${paddings.toString()} must be divisible by blockShapes ${blockShape.toString()}`);
+  const inputs = { x: $x };
+  const attrs = { blockShape, paddings };
+  return ENGINE.runKernel(SpaceToBatchND, inputs, attrs);
+}
+var spaceToBatchND = op({ spaceToBatchND_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/pool.js
+function pool_(input2, windowShape, poolingType, pad3, dilations, strides, dimRoundingMode) {
+  if (dilations == null) {
+    dilations = [1, 1];
+  }
+  if (strides == null) {
+    strides = 1;
+  }
+  if (pad3 === 0) {
+    pad3 = "valid";
+  }
+  const $x = convertToTensor(input2, "x", "maxPool");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in pool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = computePool2DInfo(x4D.shape, windowShape, strides, dilations, pad3);
+  const dilation = [convInfo.dilationHeight, convInfo.dilationWidth];
+  let basePadding;
+  if (pad3 === "same") {
+    basePadding = withSpaceToBatchBasePaddings([convInfo.filterHeight, convInfo.filterWidth], dilation);
+  } else {
+    basePadding = [[0, 0], [0, 0]];
+  }
+  const isDilationOne = dilation[0] === 1 && dilation[1] === 1;
+  const [adjustedPadding, adjustedCrops] = requiredSpaceToBatchPaddings([convInfo.inHeight, convInfo.inWidth], dilation, basePadding);
+  const convertedPad = isDilationOne ? pad3 : "valid";
+  const convertedX = isDilationOne ? x4D : spaceToBatchND(x4D, dilation, adjustedPadding);
+  const forwardOp = poolingType === "avg" ? () => avgPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode) : () => maxPool(convertedX, windowShape, strides, convertedPad, dimRoundingMode);
+  const y = forwardOp();
+  const res = isDilationOne ? y : batchToSpaceND(y, dilation, adjustedCrops);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+function requiredSpaceToBatchPaddings(inputShape, blockShape, basePadding) {
+  const padStart = basePadding.map((b) => b[0]);
+  const origPadEnd = basePadding.map((b) => b[1]);
+  const fullInputShape = inputShape.concat(padStart, origPadEnd);
+  const padEndExtra = blockShape.map((b, i) => (b - fullInputShape[i] % b) % b);
+  const padEnd = origPadEnd.map((s, i) => s + padEndExtra[i]);
+  const paddings = blockShape.map((_, i) => [padStart[i], padEnd[i]]);
+  const crops = blockShape.map((_, i) => [0, padEndExtra[i]]);
+  return [paddings, crops];
+}
+function withSpaceToBatchBasePaddings(filterShape, dilation) {
+  const dilatedFilterShape = filterShape.map((s, i) => {
+    return s + (s - 1) * (dilation[i] - 1);
+  });
+  const padExtraShape = dilatedFilterShape.map((s) => s - 1);
+  const padExtraStart = padExtraShape.map((s) => Math.floor(s / 2));
+  const padExtraEnd = padExtraShape.map((s, i) => s - padExtraStart[i]);
+  return padExtraShape.map((_, i) => {
+    return [padExtraStart[i], padExtraEnd[i]];
+  });
+}
+var pool = op({ pool_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/prelu.js
+function prelu_(x, alpha) {
+  const $x = convertToTensor(x, "x", "prelu");
+  const $alpha = convertToTensor(alpha, "alpha", "prelu");
+  const inputs = { x: $x, alpha: $alpha };
+  return ENGINE.runKernel(Prelu, inputs);
+}
+var prelu = op({ prelu_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/prod.js
+function prod_(x, axis = null, keepDims = false) {
+  let $x = convertToTensor(x, "x", "prod");
+  if ($x.dtype === "bool") {
+    $x = cast($x, "int32");
+  }
+  const inputs = { x: $x };
+  const attrs = { axis, keepDims };
+  return ENGINE.runKernel(Prod, inputs, attrs);
+}
+var prod = op({ prod_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_gather.js
+function raggedGather_(paramsNestedSplits, paramsDenseValues, indices, outputRaggedRank) {
+  const $paramsNestedSplits = paramsNestedSplits.map((t, i) => convertToTensor(t, `tensors${i}`, "raggedGather", "int32"));
+  const $paramsDenseValues = convertToTensor(paramsDenseValues, "paramsDenseValues", "raggedGather");
+  const $indices = convertToTensor(indices, "indices", "raggedGather", "int32");
+  const inputs = {
+    paramsNestedSplits: $paramsNestedSplits,
+    paramsDenseValues: $paramsDenseValues,
+    indices: $indices
+  };
+  const attrs = { outputRaggedRank };
+  const result = ENGINE.runKernel(RaggedGather, inputs, attrs);
+  return {
+    outputNestedSplits: result.slice(0, result.length - 1),
+    outputDenseValues: result[result.length - 1]
+  };
+}
+var raggedGather = op({ raggedGather_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_range.js
+function raggedRange_(starts, limits, deltas) {
+  const $starts = convertToTensor(starts, "starts", "raggedRange");
+  const $limits = convertToTensor(limits, "limits", "raggedRange", $starts.dtype);
+  const $deltas = convertToTensor(deltas, "deltas", "raggedRange", $starts.dtype);
+  const inputs = {
+    starts: $starts,
+    limits: $limits,
+    deltas: $deltas
+  };
+  const result = ENGINE.runKernel(RaggedRange, inputs);
+  return {
+    rtNestedSplits: result[0],
+    rtDenseValues: result[1]
+  };
+}
+var raggedRange = op({ raggedRange_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_tensor_to_tensor.js
+function raggedTensorToTensor_(shape, values, defaultValue, rowPartitionTensors, rowPartitionTypes) {
+  const $shape = convertToTensor(shape, "shape", "raggedTensorToTensor", "int32");
+  const $values = convertToTensor(values, "values", "raggedTensorToTensor");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "raggedTensorToTensor", $values.dtype);
+  const $rowPartitionTensors = rowPartitionTensors.map((t, i) => convertToTensor(t, `tensors${i}`, "raggedTensorToTensor", "int32"));
+  const inputs = {
+    shape: $shape,
+    values: $values,
+    defaultValue: $defaultValue,
+    rowPartitionTensors: $rowPartitionTensors
+  };
+  const attrs = { rowPartitionTypes };
+  return ENGINE.runKernel(RaggedTensorToTensor, inputs, attrs);
+}
+var raggedTensorToTensor = op({ raggedTensorToTensor_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/rand.js
+function rand_(shape, randFunction, dtype) {
+  const size = sizeFromShape(shape);
+  let values = null;
+  if (dtype == null || dtype === "float32") {
+    values = new Float32Array(size);
+  } else if (dtype === "int32") {
+    values = new Int32Array(size);
+  } else if (dtype === "bool") {
+    values = new Uint8Array(size);
+  } else {
+    throw new Error(`Unknown data type ${dtype}`);
+  }
+  for (let i = 0; i < size; i++) {
+    values[i] = randFunction();
+  }
+  return ENGINE.makeTensor(values, shape, dtype);
+}
+var rand = op({ rand_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/rand_util.js
+var seedrandom = __toESM(require_seedrandom2());
+var MPRandGauss = class {
+  constructor(mean4, stdDeviation, dtype, truncated, seed) {
+    this.mean = mean4;
+    this.stdDev = stdDeviation;
+    this.dtype = dtype;
+    this.nextVal = NaN;
+    this.truncated = truncated;
+    if (this.truncated) {
+      this.upper = this.mean + this.stdDev * 2;
+      this.lower = this.mean - this.stdDev * 2;
+    }
+    const seedValue = seed ? seed : Math.random();
+    this.random = seedrandom.alea(seedValue.toString());
+  }
+  nextValue() {
+    if (!isNaN(this.nextVal)) {
+      const value = this.nextVal;
+      this.nextVal = NaN;
+      return value;
+    }
+    let resultX, resultY;
+    let isValid = false;
+    while (!isValid) {
+      let v1, v2, s;
+      do {
+        v1 = 2 * this.random() - 1;
+        v2 = 2 * this.random() - 1;
+        s = v1 * v1 + v2 * v2;
+      } while (s >= 1 || s === 0);
+      const mul2 = Math.sqrt(-2 * Math.log(s) / s);
+      resultX = this.mean + this.stdDev * v1 * mul2;
+      resultY = this.mean + this.stdDev * v2 * mul2;
+      if (!this.truncated || this.isValidTruncated(resultX)) {
+        isValid = true;
+      }
+    }
+    if (!this.truncated || this.isValidTruncated(resultY)) {
+      this.nextVal = this.convertValue(resultY);
+    }
+    return this.convertValue(resultX);
+  }
+  convertValue(value) {
+    if (this.dtype == null || this.dtype === "float32") {
+      return value;
+    }
+    return Math.round(value);
+  }
+  isValidTruncated(value) {
+    return value <= this.upper && value >= this.lower;
+  }
+};
+var RandGamma = class {
+  constructor(alpha, beta, dtype, seed) {
+    this.alpha = alpha;
+    this.beta = 1 / beta;
+    this.dtype = dtype;
+    const seedValue = seed ? seed : Math.random();
+    this.randu = seedrandom.alea(seedValue.toString());
+    this.randn = new MPRandGauss(0, 1, dtype, false, this.randu());
+    if (alpha < 1) {
+      this.d = alpha + 2 / 3;
+    } else {
+      this.d = alpha - 1 / 3;
+    }
+    this.c = 1 / Math.sqrt(9 * this.d);
+  }
+  nextValue() {
+    let x2, v0, v1, x, u, v;
+    while (true) {
+      do {
+        x = this.randn.nextValue();
+        v = 1 + this.c * x;
+      } while (v <= 0);
+      v *= v * v;
+      x2 = x * x;
+      v0 = 1 - 0.331 * x2 * x2;
+      v1 = 0.5 * x2 + this.d * (1 - v + Math.log(v));
+      u = this.randu();
+      if (u < v0 || Math.log(u) < v1) {
+        break;
+      }
+    }
+    v = 1 / this.beta * this.d * v;
+    if (this.alpha < 1) {
+      v *= Math.pow(this.randu(), 1 / this.alpha);
+    }
+    return this.convertValue(v);
+  }
+  convertValue(value) {
+    if (this.dtype === "float32") {
+      return value;
+    }
+    return Math.round(value);
+  }
+};
+var UniformRandom = class {
+  constructor(min6 = 0, max6 = 1, dtype, seed) {
+    this.canReturnFloat = () => this.dtype == null || this.dtype === "float32";
+    this.min = min6;
+    this.range = max6 - min6;
+    this.dtype = dtype;
+    if (seed == null) {
+      seed = Math.random();
+    }
+    if (typeof seed === "number") {
+      seed = seed.toString();
+    }
+    if (!this.canReturnFloat() && this.range <= 1) {
+      throw new Error(`The difference between ${min6} - ${max6} <= 1 and dtype is not float`);
+    }
+    this.random = seedrandom.alea(seed);
+  }
+  convertValue(value) {
+    if (this.canReturnFloat()) {
+      return value;
+    }
+    return Math.round(value);
+  }
+  nextValue() {
+    return this.convertValue(this.min + this.range * this.random());
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/random_gamma.js
+function randomGamma_(shape, alpha, beta = 1, dtype = "float32", seed) {
+  if (beta == null) {
+    beta = 1;
+  }
+  if (dtype == null) {
+    dtype = "float32";
+  }
+  if (dtype !== "float32" && dtype !== "int32") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  const rgamma = new RandGamma(alpha, beta, dtype, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = rgamma.nextValue();
+  }
+  return res.toTensor();
+}
+var randomGamma = op({ randomGamma_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/random_normal.js
+function randomNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  const randGauss = new MPRandGauss(mean4, stdDev, dtype, false, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = randGauss.nextValue();
+  }
+  return res.toTensor();
+}
+var randomNormal = op({ randomNormal_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/random_standard_normal.js
+function randomStandardNormal_(shape, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type ${dtype}`);
+  }
+  return randomNormal(shape, 0, 1, dtype, seed);
+}
+var randomStandardNormal = op({ randomStandardNormal_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/random_uniform.js
+function randomUniform_(shape, minval = 0, maxval = 1, dtype = "float32", seed) {
+  const res = buffer(shape, dtype);
+  const random = new UniformRandom(minval, maxval, null, seed);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = random.nextValue();
+  }
+  return res.toTensor();
+}
+var randomUniform = op({ randomUniform_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/range.js
+function range(start, stop, step5 = 1, dtype = "float32") {
+  if (step5 === 0) {
+    throw new Error("Cannot have a step of zero");
+  }
+  const attrs = { start, stop, step: step5, dtype };
+  return ENGINE.runKernel(Range, {}, attrs);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reciprocal.js
+function reciprocal_(x) {
+  const $x = convertToTensor(x, "x", "reciprocal");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Reciprocal, inputs);
+}
+var reciprocal = op({ reciprocal_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/relu.js
+function relu_(x) {
+  const $x = convertToTensor(x, "x", "relu");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Relu, inputs);
+}
+var relu = op({ relu_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/relu6.js
+function relu6_(x) {
+  const $x = convertToTensor(x, "x", "relu6");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Relu6, inputs);
+}
+var relu6 = op({ relu6_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reverse.js
+function reverse_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  const inputs = { x: $x };
+  const attrs = { dims: axis };
+  return ENGINE.runKernel(Reverse, inputs, attrs);
+}
+var reverse = op({ reverse_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_1d.js
+function reverse1d_(x) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 1, () => `Error in reverse1D: x must be rank 1 but got rank ${$x.rank}.`);
+  return reverse($x, 0);
+}
+var reverse1d = op({ reverse1d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_2d.js
+function reverse2d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 2, () => `Error in reverse2D: x must be rank 2 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse2d = op({ reverse2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_3d.js
+function reverse3d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 3, () => `Error in reverse3D: x must be rank 3 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse3d = op({ reverse3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reverse_4d.js
+function reverse4d_(x, axis) {
+  const $x = convertToTensor(x, "x", "reverse");
+  assert($x.rank === 4, () => `Error in reverse4D: x must be rank 4 but got rank ${$x.rank}.`);
+  return reverse($x, axis);
+}
+var reverse4d = op({ reverse4d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/round.js
+function round_(x) {
+  const $x = convertToTensor(x, "x", "round");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Round, inputs);
+}
+var round2 = op({ round_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/rsqrt.js
+function rsqrt_(x) {
+  const $x = convertToTensor(x, "x", "rsqrt", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Rsqrt, inputs);
+}
+var rsqrt = op({ rsqrt_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/selu.js
+function selu_(x) {
+  const $x = convertToTensor(x, "x", "selu");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Selu, inputs);
+}
+var selu = op({ selu_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/separable_conv2d.js
+function separableConv2d_(x, depthwiseFilter, pointwiseFilter, strides, pad3, dilation = [1, 1], dataFormat = "NHWC") {
+  const $x = convertToTensor(x, "x", "separableConv2d");
+  const $depthwiseFilter = convertToTensor(depthwiseFilter, "depthwiseFilter", "separableConv2d");
+  const $pointwiseFilter = convertToTensor(pointwiseFilter, "pointwiseFilter", "separableConv2d");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  if (dataFormat === "NCHW") {
+    throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");
+  }
+  assert(x4D.rank === 4, () => `Error in separableConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($depthwiseFilter.rank === 4, () => `Error in separableConv2d: depthwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);
+  assert($pointwiseFilter.rank === 4, () => `Error in separableConv2d: pointwise filter must be rank 4, but got rank ${$depthwiseFilter.rank}.`);
+  assert($pointwiseFilter.shape[0] === 1, () => `Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${$pointwiseFilter.shape[0]}.`);
+  assert($pointwiseFilter.shape[1] === 1, () => `Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${$pointwiseFilter.shape[1]}.`);
+  const inChannels = $depthwiseFilter.shape[2];
+  const channelMultiplier = $depthwiseFilter.shape[3];
+  assert($pointwiseFilter.shape[2] === inChannels * channelMultiplier, () => `Error in separableConv2d: the third dimension of pointwise filter must be ${inChannels * channelMultiplier}, but got ${$pointwiseFilter.shape[2]}.`);
+  const depthwise = depthwiseConv2d(x4D, $depthwiseFilter, strides, pad3, dataFormat, dilation);
+  const pointwiseStride = 1;
+  const res = conv2d(depthwise, $pointwiseFilter, pointwiseStride, "valid", dataFormat);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var separableConv2d = op({ separableConv2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/setdiff1d_async.js
+async function setdiff1dAsync_(x, y) {
+  const $x = convertToTensor(x, "x", "setdiff1d");
+  const $y = convertToTensor(y, "y", "setdiff1d");
+  assert($x.dtype === $y.dtype, () => `x and y should have the same dtype, but got x (${$x.dtype}) and y (${$y.dtype}).`);
+  assert($x.rank === 1, () => `x should be 1D tensor, but got x (${$x.shape}).`);
+  assert($y.rank === 1, () => `y should be 1D tensor, but got y (${$y.shape}).`);
+  const xVals = await $x.data();
+  const yVals = await $y.data();
+  const ySet = new Set(yVals);
+  let outputSize = 0;
+  for (let i = 0; i < xVals.length; i++) {
+    if (!ySet.has(xVals[i])) {
+      outputSize++;
+    }
+  }
+  const buffer2 = new TensorBuffer([outputSize], $x.dtype);
+  const indices = new TensorBuffer([outputSize], "int32");
+  for (let i = 0, p2 = 0; i < xVals.length; i++) {
+    if (!ySet.has(xVals[i])) {
+      buffer2.values[p2] = xVals[i];
+      indices.values[p2] = i;
+      p2++;
+    }
+  }
+  return [buffer2.toTensor(), indices.toTensor()];
+}
+var setdiff1dAsync = setdiff1dAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sign.js
+function sign_(x) {
+  const $x = convertToTensor(x, "x", "sign");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sign, inputs);
+}
+var sign = op({ sign_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sin.js
+function sin_(x) {
+  const $x = convertToTensor(x, "x", "sin", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sin, inputs);
+}
+var sin = op({ sin_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sinh.js
+function sinh_(x) {
+  const $x = convertToTensor(x, "x", "sinh");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Sinh, inputs);
+}
+var sinh = op({ sinh_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice1d.js
+function slice1d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice1d");
+  assert($x.rank === 1, () => `slice1d expects a rank-1 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, [begin], [size]);
+}
+var slice1d = op({ slice1d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice2d.js
+function slice2d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice2d");
+  assert($x.rank === 2, () => `slice2d expects a rank-2 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice2d = op({ slice2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice3d.js
+function slice3d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice3d");
+  assert($x.rank === 3, () => `slice3d expects a rank-3 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice3d = op({ slice3d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/slice4d.js
+function slice4d_(x, begin, size) {
+  const $x = convertToTensor(x, "x", "slice4d");
+  assert($x.rank === 4, () => `slice4d expects a rank-4 tensor, but got a rank-${$x.rank} tensor`);
+  return slice($x, begin, size);
+}
+var slice4d = op({ slice4d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/softmax.js
+function softmax_(logits, dim = -1) {
+  const $logits = convertToTensor(logits, "logits", "softmax", "float32");
+  if (dim === -1) {
+    dim = $logits.rank - 1;
+  }
+  if (dim !== $logits.rank - 1) {
+    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${$logits.rank} and dim was ${dim}`);
+  }
+  const inputs = { logits: $logits };
+  const attrs = { dim };
+  return ENGINE.runKernel(Softmax, inputs, attrs);
+}
+var softmax = op({ softmax_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/fft.js
+function fft_(input2) {
+  assert(input2.dtype === "complex64", () => `The dtype for tf.spectral.fft() must be complex64 but got ${input2.dtype}.`);
+  const inputs = { input: input2 };
+  return ENGINE.runKernel(FFT, inputs);
+}
+var fft = op({ fft_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/ifft.js
+function ifft_(input2) {
+  assert(input2.dtype === "complex64", () => `The dtype for tf.spectral.ifft() must be complex64 but got ${input2.dtype}.`);
+  const inputs = { input: input2 };
+  return ENGINE.runKernel(IFFT, inputs);
+}
+var ifft = op({ ifft_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/irfft.js
+function irfft_(input2) {
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = input2.size / innerDimensionSize;
+  let ret;
+  if (innerDimensionSize <= 2) {
+    const complexInput = reshape(input2, [batch, innerDimensionSize]);
+    ret = ifft(complexInput);
+  } else {
+    const outputShape = [batch, 2 * (innerDimensionSize - 1)];
+    const realInput = reshape(real(input2), [batch, innerDimensionSize]);
+    const imagInput = reshape(imag(input2), [batch, innerDimensionSize]);
+    const realConjugate = reverse(slice(realInput, [0, 1], [batch, innerDimensionSize - 2]), 1);
+    const imagConjugate = mul(reverse(slice(imagInput, [0, 1], [batch, innerDimensionSize - 2]), 1), scalar(-1));
+    const r = concat([realInput, realConjugate], 1);
+    const i = concat([imagInput, imagConjugate], 1);
+    const complexInput = reshape(complex(r, i), [outputShape[0], outputShape[1]]);
+    ret = ifft(complexInput);
+  }
+  ret = real(ret);
+  if (input2.rank === 3 && input2.shape[0] !== 0) {
+    const temp = ret;
+    const batch2 = input2.shape[0];
+    ret = reshape(ret, [batch2, ret.shape[0] / batch2, ret.shape[1]]);
+    temp.dispose();
+  }
+  return ret;
+}
+var irfft = op({ irfft_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/split.js
+function split_(x, numOrSizeSplits, axis = 0) {
+  const $x = convertToTensor(x, "x", "split");
+  const inputs = { x: $x };
+  const attr = { numOrSizeSplits, axis };
+  return ENGINE.runKernel(SplitV, inputs, attr);
+}
+var split = op({ split_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/spectral/rfft.js
+function rfft_(input2, fftLength) {
+  assert(input2.dtype === "float32", () => `The dtype for rfft() must be real value but got ${input2.dtype}`);
+  let innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = input2.size / innerDimensionSize;
+  let adjustedInput;
+  if (fftLength != null && fftLength < innerDimensionSize) {
+    const begin = input2.shape.map((v) => 0);
+    const size = input2.shape.map((v) => v);
+    size[input2.shape.length - 1] = fftLength;
+    adjustedInput = slice(input2, begin, size);
+    innerDimensionSize = fftLength;
+  } else if (fftLength != null && fftLength > innerDimensionSize) {
+    const zerosShape = input2.shape.map((v) => v);
+    zerosShape[input2.shape.length - 1] = fftLength - innerDimensionSize;
+    adjustedInput = concat([input2, zeros(zerosShape)], input2.shape.length - 1);
+    innerDimensionSize = fftLength;
+  } else {
+    adjustedInput = input2;
+  }
+  const zerosInput = zerosLike(adjustedInput);
+  const complexInput = reshape(complex(adjustedInput, zerosInput), [batch, innerDimensionSize]);
+  const ret = fft(complexInput);
+  const half = Math.floor(innerDimensionSize / 2) + 1;
+  const realValues = real(ret);
+  const imagValues = imag(ret);
+  const realComplexConjugate = split(realValues, [half, innerDimensionSize - half], realValues.shape.length - 1);
+  const imagComplexConjugate = split(imagValues, [half, innerDimensionSize - half], imagValues.shape.length - 1);
+  const outputShape = adjustedInput.shape.slice();
+  outputShape[adjustedInput.shape.length - 1] = half;
+  return reshape(complex(realComplexConjugate[0], imagComplexConjugate[0]), outputShape);
+}
+var rfft = op({ rfft_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/squared_difference.js
+function squaredDifference_(a, b) {
+  let $a = convertToTensor(a, "a", "squaredDifference");
+  let $b = convertToTensor(b, "b", "squaredDifference");
+  [$a, $b] = makeTypesMatch($a, $b);
+  assertAndGetBroadcastShape($a.shape, $b.shape);
+  const inputs = { a: $a, b: $b };
+  const attrs = {};
+  return ENGINE.runKernel(SquaredDifference, inputs, attrs);
+}
+var squaredDifference = op({ squaredDifference_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/squeeze.js
+function squeeze_(x, axis) {
+  const $x = convertToTensor(x, "x", "squeeze", "string_or_numeric");
+  return reshape($x, squeezeShape($x.shape, axis).newShape);
+}
+var squeeze = op({ squeeze_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/stack.js
+function stack_(tensors, axis = 0) {
+  const $tensors = convertToTensorArray(tensors, "tensors", "stack", "string_or_numeric");
+  assert($tensors.length >= 1, () => "Pass at least one tensor to tf.stack");
+  if ($tensors.length > 0) {
+    assert(axis <= $tensors[0].rank, () => "Axis must be <= rank of the tensor");
+  }
+  const inputs = $tensors;
+  const attrs = { axis };
+  return ENGINE.runKernel(Pack, inputs, attrs);
+}
+var stack = op({ stack_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/step.js
+function step_(x, alpha = 0) {
+  const $x = convertToTensor(x, "x", "step");
+  const inputs = { x: $x };
+  const attrs = { alpha };
+  return ENGINE.runKernel(Step, inputs, attrs);
+}
+var step = op({ step_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/strided_slice.js
+function stridedSlice_(x, begin, end, strides, beginMask = 0, endMask = 0, ellipsisMask = 0, newAxisMask = 0, shrinkAxisMask = 0) {
+  const $x = convertToTensor(x, "x", "stridedSlice", "string_or_numeric");
+  const inputs = { x: $x };
+  const attrs = {
+    begin,
+    end,
+    strides,
+    beginMask,
+    endMask,
+    ellipsisMask,
+    newAxisMask,
+    shrinkAxisMask
+  };
+  return ENGINE.runKernel(StridedSlice, inputs, attrs);
+}
+var stridedSlice = op({ stridedSlice_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tan.js
+function tan_(x) {
+  const $x = convertToTensor(x, "x", "tan", "float32");
+  const inputs = { x: $x };
+  return ENGINE.runKernel(Tan, inputs);
+}
+var tan = op({ tan_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor1d.js
+function tensor1d(values, dtype) {
+  assertNonNull(values);
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 1) {
+    throw new Error("tensor1d() requires values to be a flat/TypedArray");
+  }
+  const shape = null;
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor2d.js
+function tensor2d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 2) {
+    throw new Error("tensor2d() requires shape to have two numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 2 && inferredShape.length !== 1) {
+    throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor4d.js
+function tensor4d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 4) {
+    throw new Error("tensor4d() requires shape to have four numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 4 && inferredShape.length !== 1) {
+    throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor5d.js
+function tensor5d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 5) {
+    throw new Error("tensor5d() requires shape to have five numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 5 && inferredShape.length !== 1) {
+    throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");
+  }
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/tensor6d.js
+function tensor6d(values, shape, dtype) {
+  assertNonNull(values);
+  if (shape != null && shape.length !== 6) {
+    throw new Error("tensor6d() requires shape to have six numbers");
+  }
+  const inferredShape = inferShape(values, dtype);
+  if (inferredShape.length !== 6 && inferredShape.length !== 1) {
+    throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");
+  }
+  if (inferredShape.length === 1 && shape == null) {
+    throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");
+  }
+  shape = shape || inferredShape;
+  return makeTensor(values, shape, inferredShape, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/topk.js
+function topk_(x, k = 1, sorted = true) {
+  const $x = convertToTensor(x, "x", "topk");
+  if ($x.rank === 0) {
+    throw new Error("topk() expects the input to be of rank 1 or higher");
+  }
+  const lastDim = $x.shape[$x.shape.length - 1];
+  if (k < 0) {
+    throw new Error(`'k' passed to topk() must be >= 0 but got ${k}`);
+  }
+  if (k > lastDim) {
+    throw new Error(`'k' passed to topk() must be <= the last dimension (${lastDim}) but got ${k}`);
+  }
+  const inputs = { x: $x };
+  const attrs = { k, sorted };
+  const [values, indices] = ENGINE.runKernel(TopK, inputs, attrs);
+  return { values, indices };
+}
+var topk = op({ topk_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/truncated_normal.js
+function truncatedNormal_(shape, mean4 = 0, stdDev = 1, dtype, seed) {
+  if (dtype != null && dtype === "bool") {
+    throw new Error(`Unsupported data type $ { dtype }`);
+  }
+  const randGauss = new MPRandGauss(mean4, stdDev, dtype, true, seed);
+  const res = buffer(shape, dtype);
+  for (let i = 0; i < res.values.length; i++) {
+    res.values[i] = randGauss.nextValue();
+  }
+  return res.toTensor();
+}
+var truncatedNormal = op({ truncatedNormal_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/unique.js
+function unique_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "unique", "string_or_numeric");
+  assert($x.rank > 0, () => "The input tensor must be at least 1D");
+  const inputs = { x: $x };
+  const attrs = { axis };
+  const [values, indices] = ENGINE.runKernel(Unique, inputs, attrs);
+  return { values, indices };
+}
+var unique = op({ unique_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/unsorted_segment_sum.js
+function unsortedSegmentSum_(x, segmentIds, numSegments) {
+  const $x = convertToTensor(x, "x", "unsortedSegmentSum");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "unsortedSegmentSum", "int32");
+  assert(isInt(numSegments), () => "numSegments must be of dtype int");
+  const inputs = { x: $x, segmentIds: $segmentIds };
+  const attrs = { numSegments };
+  return ENGINE.runKernel(UnsortedSegmentSum, inputs, attrs);
+}
+var unsortedSegmentSum = op({ unsortedSegmentSum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/unstack.js
+function unstack_(x, axis = 0) {
+  const $x = convertToTensor(x, "x", "unstack", "string_or_numeric");
+  assert(axis >= -$x.shape.length && axis < $x.shape.length, () => `Axis = ${axis} is not in [-${$x.shape.length}, ${$x.shape.length})`);
+  const inputs = { value: $x };
+  const attrs = { axis };
+  return ENGINE.runKernel(Unpack, inputs, attrs);
+}
+var unstack = op({ unstack_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/upper_bound.js
+function upperBound(sortedSequence, values) {
+  return searchSorted(sortedSequence, values, "right");
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/variable.js
+function variable(initialValue, trainable = true, name, dtype) {
+  return ENGINE.makeVariable(initialValue, trainable, name, dtype);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/where_impl.js
+function whereImpl(condShape, condVals) {
+  const indices = [];
+  for (let i = 0; i < condVals.length; i++) {
+    if (condVals[i]) {
+      indices.push(i);
+    }
+  }
+  const inBuffer = buffer(condShape, "int32");
+  const out = buffer([indices.length, condShape.length], "int32");
+  for (let i = 0; i < indices.length; i++) {
+    const loc = inBuffer.indexToLoc(indices[i]);
+    const offset = i * condShape.length;
+    out.values.set(loc, offset);
+  }
+  return out.toTensor();
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/where_async.js
+async function whereAsync_(condition) {
+  const $condition = convertToTensor(condition, "condition", "whereAsync", "bool");
+  const vals = await $condition.data();
+  const res = whereImpl($condition.shape, vals);
+  if (condition !== $condition) {
+    $condition.dispose();
+  }
+  return res;
+}
+var whereAsync = whereAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/boolean_mask.js
+async function booleanMaskAsync_(tensor2, mask, axis) {
+  const $tensor = convertToTensor(tensor2, "tensor", "boolMask");
+  const $mask = convertToTensor(mask, "mask", "boolMask", "bool");
+  const axisFrom = axis == null ? 0 : axis;
+  const maskDim = $mask.rank;
+  const tensorShape = $tensor.shape;
+  assert(maskDim > 0, () => "mask cannot be scalar");
+  assertShapesMatch(tensorShape.slice(axisFrom, axisFrom + maskDim), $mask.shape, `mask's shape must match the first K dimensions of tensor's shape,`);
+  let leadingSize = 1;
+  for (let i = axisFrom; i < axisFrom + maskDim; i++) {
+    leadingSize *= tensorShape[i];
+  }
+  const targetTensorShape = tensorShape.slice(0, axisFrom).concat([leadingSize], tensorShape.slice(axisFrom + maskDim));
+  const reshapedTensor = reshape($tensor, targetTensorShape);
+  const reshapedMask = reshape($mask, [-1]);
+  const positivePositions = await whereAsync(reshapedMask);
+  const indices = squeeze(positivePositions, [1]);
+  const res = gather(reshapedTensor, indices, axisFrom);
+  if (tensor2 !== $tensor) {
+    $tensor.dispose();
+  }
+  if (mask !== $mask) {
+    $mask.dispose();
+  }
+  indices.dispose();
+  reshapedTensor.dispose();
+  reshapedMask.dispose();
+  positivePositions.dispose();
+  return res;
+}
+var booleanMaskAsync = booleanMaskAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/moving_average.js
+function movingAverage_(v, x, decay, step5, zeroDebias = true) {
+  const $v = convertToTensor(v, "v", "movingAverage");
+  const $x = convertToTensor(x, "x", "movingAverage");
+  const $decay = convertToTensor(decay, "decay", "movingAverage");
+  assertTypesMatch($v, $x);
+  assert(arraysEqual($v.shape, $x.shape), () => "Shape mismatch in v and x");
+  const one = scalar(1);
+  const oneMinusDecay = sub(one, $decay);
+  let update = mul(sub($x, $v), oneMinusDecay);
+  if (zeroDebias) {
+    assert(step5 != null, () => "When using zeroDebias: true, step is required.");
+    const $step = convertToTensor(step5, "step", "movingAverage");
+    update = div(update, sub(one, pow($decay, $step)));
+  }
+  return add2($v, update);
+}
+var movingAverage = op({ movingAverage_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/scatter_nd.js
+function scatterND_(indices, updates, shape) {
+  const $indices = convertToTensor(indices, "indices", "scatterND", "int32");
+  const $updates = convertToTensor(updates, "updates", "scatterND");
+  validateInput($updates, $indices, shape);
+  const inputs = { indices: $indices, updates: $updates };
+  const attrs = { shape };
+  return ENGINE.runKernel(ScatterNd, inputs, attrs);
+}
+var scatterND = op({ scatterND_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse_to_dense_util.js
+function validateInput2(sparseIndices, sparseValues, outputShape, defaultValues) {
+  if (sparseIndices.dtype !== "int32") {
+    throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${sparseIndices.dtype}.`);
+  }
+  if (sparseIndices.rank > 2) {
+    throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${sparseIndices.shape}.`);
+  }
+  const numElems = sparseIndices.rank > 0 ? sparseIndices.shape[0] : 1;
+  const numDims = sparseIndices.rank > 1 ? sparseIndices.shape[1] : 1;
+  if (outputShape.length !== numDims) {
+    throw new Error(`outputShape has incorrect number of elements:, ${outputShape.length}, should be: ${numDims}.`);
+  }
+  const numValues = sparseValues.size;
+  if (!(sparseValues.rank === 0 || sparseValues.rank === 1 && numValues === numElems)) {
+    throw new Error(`sparseValues has incorrect shape ${sparseValues.shape}, should be [] or [${numElems}]`);
+  }
+  if (sparseValues.dtype !== defaultValues.dtype) {
+    throw new Error("sparseValues.dtype must match defaultValues.dtype");
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse_to_dense.js
+function sparseToDense_(sparseIndices, sparseValues, outputShape, defaultValue = 0) {
+  const $sparseIndices = convertToTensor(sparseIndices, "sparseIndices", "sparseToDense", "int32");
+  const $sparseValues = convertToTensor(sparseValues, "sparseValues", "sparseToDense", "string_or_numeric");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "sparseToDense", $sparseValues.dtype);
+  validateInput2($sparseIndices, $sparseValues, outputShape, $defaultValue);
+  const inputs = {
+    sparseIndices: $sparseIndices,
+    sparseValues: $sparseValues,
+    defaultValue: $defaultValue
+  };
+  const attrs = { outputShape };
+  return ENGINE.runKernel(SparseToDense, inputs, attrs);
+}
+var sparseToDense = op({ sparseToDense_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/gather_nd.js
+function gatherND_(x, indices) {
+  const $indices = convertToTensor(indices, "indices", "gatherND", "int32");
+  const $x = convertToTensor(x, "x", "gatherND", "string_or_numeric");
+  const inputs = { params: $x, indices: $indices };
+  return ENGINE.runKernel(GatherNd, inputs);
+}
+var gatherND = op({ gatherND_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/dropout_util.js
+function getNoiseShape(x, noiseShape) {
+  if (noiseShape == null) {
+    return x.shape.slice();
+  }
+  if (arraysEqual(x.shape, noiseShape)) {
+    return noiseShape;
+  }
+  if (x.shape.length === noiseShape.length) {
+    const newDimension = [];
+    for (let i = 0; i < x.shape.length; i++) {
+      if (noiseShape[i] == null && x.shape[i] != null) {
+        newDimension.push(x.shape[i]);
+      } else {
+        newDimension.push(noiseShape[i]);
+      }
+    }
+    return newDimension;
+  }
+  return noiseShape;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/dropout.js
+function dropout_(x, rate, noiseShape, seed) {
+  const $x = convertToTensor(x, "x", "dropout");
+  assert($x.dtype === "float32", () => `x has to be a floating point tensor since it's going to be scaled, but got a ${$x.dtype} tensor instead.`);
+  assert(rate >= 0 && rate < 1, () => `rate must be a float in the range [0, 1), but got ${rate}.`);
+  if (rate === 0) {
+    return x instanceof Tensor ? $x.clone() : $x;
+  }
+  const $noiseShape = getNoiseShape($x, noiseShape);
+  const keepProb = 1 - rate;
+  const multiplier = div(floor(add2(randomUniform($noiseShape, 0, 1, "float32", seed), keepProb)), keepProb);
+  return mul($x, multiplier);
+}
+var dropout = op({ dropout_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/signal_ops_util.js
+function enclosingPowerOfTwo(value) {
+  return Math.floor(Math.pow(2, Math.ceil(Math.log(value) / Math.log(2))));
+}
+function cosineWindow(windowLength, a, b) {
+  const even = 1 - windowLength % 2;
+  const newValues = new Float32Array(windowLength);
+  for (let i = 0; i < windowLength; ++i) {
+    const cosArg = 2 * Math.PI * i / (windowLength + even - 1);
+    newValues[i] = a - b * Math.cos(cosArg);
+  }
+  return tensor1d(newValues, "float32");
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/in_top_k.js
+async function inTopKAsync_(predictions, targets, k = 1) {
+  const $predictions = convertToTensor(predictions, "predictions", "inTopK");
+  const $targets = convertToTensor(targets, "targets", "inTopK");
+  assert($predictions.rank > 1, () => `inTopK() expects the predictions to be of rank 2 or higher, but got ${$predictions.rank}`);
+  assert($predictions.rank - 1 === $targets.rank, () => `predictions rank should be 1 larger than targets rank, but got predictions rank ${$predictions.rank} and targets rank ${$targets.rank}`);
+  assertShapesMatch($predictions.shape.slice(0, $predictions.shape.length - 1), $targets.shape, `predictions's shape should be align with the targets' shape, except the last dimension.`);
+  const lastDim = $predictions.shape[$predictions.shape.length - 1];
+  assert(k > 0 && k <= lastDim, () => `'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${lastDim}), but got ${k}`);
+  const predictionsVals = await $predictions.data();
+  const targetsVals = await $targets.data();
+  const [batch, size] = [predictionsVals.length / lastDim, lastDim];
+  const precision3 = getTypedArrayFromDType("bool", batch);
+  for (let b = 0; b < batch; b++) {
+    const offset = b * size;
+    const vals = predictionsVals.subarray(offset, offset + size);
+    const valAndInd = [];
+    for (let i = 0; i < vals.length; i++) {
+      valAndInd.push({ value: vals[i], index: i });
+    }
+    valAndInd.sort((a, b2) => b2.value - a.value);
+    precision3[b] = 0;
+    for (let i = 0; i < k; i++) {
+      if (valAndInd[i].index === targetsVals[b]) {
+        precision3[b] = 1;
+        break;
+      }
+    }
+  }
+  if (predictions !== $predictions) {
+    $predictions.dispose();
+  }
+  if (targets !== $targets) {
+    $targets.dispose();
+  }
+  return tensor(precision3, $targets.shape, "bool");
+}
+var inTopKAsync = inTopKAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fused_ops.js
+var fused_ops_exports = {};
+__export(fused_ops_exports, {
+  conv2d: () => conv2d2,
+  depthwiseConv2d: () => depthwiseConv2d2,
+  matMul: () => matMul2
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv2d_backprop_filter.js
+function conv2DBackpropFilter_(x, dy, filterShape, strides, pad3, dataFormat = "NHWC", dimRoundingMode) {
+  let x4D = x;
+  if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  }
+  let dy4D = dy;
+  if (dy4D.rank === 3) {
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in conv2dDerFilter: input must be rank 4, but got shape ${x4D.shape}.`);
+  assert(dy4D.rank === 4, () => `Error in conv2dDerFilter: dy must be rank 4, but got shape ${dy4D.shape}.`);
+  assert(filterShape.length === 4, () => `Error in conv2dDerFilter: filterShape must be length 4, but got ${filterShape}.`);
+  const inDepth = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  const outDepth = dataFormat === "NHWC" ? dy4D.shape[3] : dy4D.shape[1];
+  assert(inDepth === filterShape[2], () => `Error in conv2dDerFilter: depth of input ${inDepth}) must match input depth in filter (${filterShape[2]}.`);
+  assert(outDepth === filterShape[3], () => `Error in conv2dDerFilter: depth of dy (${outDepth}) must match output depth for filter (${filterShape[3]}).`);
+  checkPadOnDimRoundingMode("conv2dDerFilter", pad3, dimRoundingMode);
+  const inputs = { x: x4D, dy: dy4D };
+  const attrs = { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape };
+  return ENGINE.runKernel(Conv2DBackpropFilter, inputs, attrs);
+}
+var conv2DBackpropFilter = op({ conv2DBackpropFilter_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fused_util.js
+function getFusedDyActivation(dy, y, activation2) {
+  if (activation2 == null || activation2 === "linear") {
+    return dy;
+  }
+  if (activation2 === "relu") {
+    return mul(dy, step(y));
+  }
+  throw new Error(`Cannot compute gradient for fused activation ${activation2}.`);
+}
+function getFusedBiasGradient(bias, dyActivation) {
+  let res = dyActivation;
+  const reduceAxes = getReductionAxes(bias.shape, dyActivation.shape);
+  if (reduceAxes.length > 0) {
+    res = sum2(res, reduceAxes);
+  }
+  return reshape(res, bias.shape);
+}
+function applyActivation(x, activation2, preluActivationWeights, leakyreluAlpha) {
+  if (activation2 === "linear") {
+    return x;
+  } else if (activation2 === "relu") {
+    return relu(x);
+  } else if (activation2 === "elu") {
+    return elu(x);
+  } else if (activation2 === "relu6") {
+    return relu6(x);
+  } else if (activation2 === "prelu") {
+    return prelu(x, preluActivationWeights);
+  } else if (activation2 === "leakyrelu") {
+    return leakyRelu(x, leakyreluAlpha);
+  } else if (activation2 === "sigmoid") {
+    return sigmoid(x);
+  }
+  throw new Error(`Unknown fused activation ${activation2}.`);
+}
+var shouldFuse = (gradientDepth, activation2) => {
+  const gradientMode = gradientDepth > 0;
+  return !gradientMode || activation2 === "linear";
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fused/conv2d.js
+function fusedConv2d_({ x, filter, strides, pad: pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha }) {
+  activation2 = activation2 || "linear";
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    assert(dataFormat === "NHWC", () => `Error in fused conv2d: got dataFormat of ${dataFormat} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);
+    let result = conv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  const $x = convertToTensor(x, "x", "conv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "conv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in fused conv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in fused conv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  checkPadOnDimRoundingMode("fused conv2d", pad3, dimRoundingMode);
+  const inputChannels = dataFormat === "NHWC" ? x4D.shape[3] : x4D.shape[1];
+  assert($filter.shape[2] === inputChannels, () => `Error in conv2d: depth of input (${inputChannels}) must match input depth for filter ${$filter.shape[2]}.`);
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in conv2D: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused conv2d");
+    [$bias] = makeTypesMatch($bias, $x);
+    if (dataFormat === "NHWC") {
+      assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);
+    } else {
+      assert($bias.shape.length <= 1, () => `Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${$bias.shape.length}.`);
+      assert($bias.shape.length === 0 || $bias.shape[0] === convInfo.outChannels || $bias.shape[0] === 1, () => `Error in fused conv2d: bias shape (${$bias.shape}) is not compatible with the number of output channels (${convInfo.outChannels})`);
+    }
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    const alphaShape = preluActivationWeights.shape;
+    assert(alphaShape.length <= 1 || alphaShape.length === 3, () => `Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${alphaShape.length}.`);
+    if (alphaShape.length === 1) {
+      assert(alphaShape[0] === 1 || alphaShape[0] === convInfo.outChannels, () => `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the number of output channels (${convInfo.outChannels}).`);
+    } else if (alphaShape.length === 3) {
+      try {
+        assertAndGetBroadcastShape(alphaShape, convInfo.outShape);
+      } catch (e) {
+        const errMsg = `Error in fused conv2d: PReLU activation weights (${alphaShape}) is not compatible with the output shape of the conv2d (${convInfo.outShape}).`;
+        throw Error(errMsg);
+      }
+    }
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused conv2d");
+  }
+  const grad2 = (dy, saved) => {
+    assert(dataFormat === "NHWC", () => `Error in gradient of fused conv2D: got dataFormat of ${dataFormat} but only NHWC is currently supported.`);
+    const [$filter2, x4D2, y, $bias2] = saved;
+    const dyActivation = getFusedDyActivation(dy, y, activation2);
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    const xDer = conv2DBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3);
+    const filterDer = conv2DBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3);
+    const der = [xDer, filterDer];
+    if ($bias2 != null) {
+      const biasDer = getFusedBiasGradient($bias2, dyActivation);
+      der.push(biasDer);
+    }
+    return der;
+  };
+  const inputs = {
+    x: x4D,
+    filter: $filter,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = {
+    strides,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    dimRoundingMode,
+    activation: activation2,
+    leakyreluAlpha
+  };
+  if (bias == null) {
+    const customOp = customGrad((x4D2, filter2, save) => {
+      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);
+      save([filter2, x4D2, res]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOp(x4D, $filter);
+  } else {
+    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {
+      let res = ENGINE.runKernel(FusedConv2D, inputs, attrs);
+      save([filter2, x4D2, res, bias2]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOpWithBias(x4D, $filter, $bias);
+  }
+}
+var conv2d2 = op({ fusedConv2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d_native_backprop_filter.js
+function depthwiseConv2dNativeBackpropFilter_(x, dy, filterShape, strides, pad3, dilations = [1, 1], dimRoundingMode) {
+  let x4D = x;
+  if (x.rank === 3) {
+    x4D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2]]);
+  }
+  let dy4D = dy;
+  if (dy4D.rank === 3) {
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  const inputs = { x: x4D, dy: dy4D };
+  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, filterShape };
+  return ENGINE.runKernel(DepthwiseConv2dNativeBackpropFilter, inputs, attrs);
+}
+var depthwiseConv2dNativeBackpropFilter = op({ depthwiseConv2dNativeBackpropFilter_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/depthwise_conv2d_native_backprop_input.js
+function depthwiseConv2dNativeBackpropInput_(xShape, dy, filter, strides, pad3, dilations = [1, 1], dimRoundingMode) {
+  let dy4D = dy;
+  let reshapedTo4D = false;
+  if (dy.rank === 3) {
+    reshapedTo4D = true;
+    dy4D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2]]);
+  }
+  const inputs = { dy: dy4D, filter };
+  const attrs = { strides, pad: pad3, dimRoundingMode, dilations, inputShape: xShape };
+  const res = ENGINE.runKernel(DepthwiseConv2dNativeBackpropInput, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var depthwiseConv2dNativeBackpropInput = op({ depthwiseConv2dNativeBackpropInput_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fused/depthwise_conv2d.js
+function fusedDepthwiseConv2d_({ x, filter, strides, pad: pad3, dataFormat = "NHWC", dilations = [1, 1], dimRoundingMode, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha }) {
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    let result = depthwiseConv2d(x, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  const $x = convertToTensor(x, "x", "depthwiseConv2d", "float32");
+  const $filter = convertToTensor(filter, "filter", "depthwiseConv2d", "float32");
+  let x4D = $x;
+  let reshapedTo4D = false;
+  if ($x.rank === 3) {
+    reshapedTo4D = true;
+    x4D = reshape($x, [1, $x.shape[0], $x.shape[1], $x.shape[2]]);
+  }
+  assert(x4D.rank === 4, () => `Error in fused depthwiseConv2d: input must be rank 4, but got rank ${x4D.rank}.`);
+  assert($filter.rank === 4, () => `Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${$filter.rank}.`);
+  assert(x4D.shape[3] === $filter.shape[2], () => `Error in fused depthwiseConv2d: number of input channels (${x4D.shape[3]}) must match the inChannels dimension in filter ${$filter.shape[2]}.`);
+  if (dilations == null) {
+    dilations = [1, 1];
+  }
+  assert(eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  checkPadOnDimRoundingMode("fused depthwiseConv2d", pad3, dimRoundingMode);
+  const convInfo = computeConv2DInfo(x4D.shape, $filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused conv2d");
+    [$bias] = makeTypesMatch($bias, $x);
+    assertAndGetBroadcastShape(convInfo.outShape, $bias.shape);
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused depthwiseConv2d");
+  }
+  const grad2 = (dy, saved) => {
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${dilations}'`);
+    const [$filter2, x4D2, y, bias2] = saved;
+    const dyActivation = getFusedDyActivation(dy, y, activation2);
+    const xDer = depthwiseConv2dNativeBackpropInput(x4D2.shape, dyActivation, $filter2, strides, pad3, dilations, dimRoundingMode);
+    const filterDer = depthwiseConv2dNativeBackpropFilter(x4D2, dyActivation, $filter2.shape, strides, pad3, dilations, dimRoundingMode);
+    if (bias2 != null) {
+      const biasDer = getFusedBiasGradient($bias, dyActivation);
+      return [xDer, filterDer, biasDer];
+    }
+    return [xDer, filterDer];
+  };
+  const inputs = {
+    x: x4D,
+    filter: $filter,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = {
+    strides,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    dimRoundingMode,
+    activation: activation2,
+    leakyreluAlpha
+  };
+  if (bias == null) {
+    const customOp = customGrad((x4D2, filter2, save) => {
+      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);
+      save([filter2, x4D2, res]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOp(x4D, $filter);
+  } else {
+    const customOpWithBias = customGrad((x4D2, filter2, bias2, save) => {
+      let res = ENGINE.runKernel(FusedDepthwiseConv2D, inputs, attrs);
+      save([filter2, x4D2, res, bias2]);
+      if (reshapedTo4D) {
+        res = reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+      }
+      return { value: res, gradFunc: grad2 };
+    });
+    return customOpWithBias(x4D, $filter, $bias);
+  }
+}
+var depthwiseConv2d2 = op({ fusedDepthwiseConv2d_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/fused/mat_mul.js
+function fusedMatMul_({ a, b, transposeA = false, transposeB = false, bias, activation: activation2 = "linear", preluActivationWeights, leakyreluAlpha = 0.2 }) {
+  if (shouldFuse(ENGINE.state.gradientDepth, activation2) === false) {
+    let result = matMul(a, b, transposeA, transposeB);
+    if (bias != null) {
+      result = add2(result, bias);
+    }
+    return applyActivation(result, activation2, preluActivationWeights, leakyreluAlpha);
+  }
+  let $a = convertToTensor(a, "a", "fused matMul");
+  let $b = convertToTensor(b, "b", "fused matMul");
+  [$a, $b] = makeTypesMatch($a, $b);
+  const innerShapeA = transposeA ? $a.shape[$a.rank - 2] : $a.shape[$a.rank - 1];
+  const innerShapeB = transposeB ? $b.shape[$b.rank - 1] : $b.shape[$b.rank - 2];
+  const outerShapeA = transposeA ? $a.shape[$a.rank - 1] : $a.shape[$a.rank - 2];
+  const outerShapeB = transposeB ? $b.shape[$b.rank - 2] : $b.shape[$b.rank - 1];
+  const outerDimsA = $a.shape.slice(0, -2);
+  const outerDimsB = $b.shape.slice(0, -2);
+  const batchDimA = sizeFromShape(outerDimsA);
+  const batchDimB = sizeFromShape(outerDimsB);
+  assert(innerShapeA === innerShapeB, () => `Error in fused matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${$a.shape} and ${$b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const outShapeOuterDims = assertAndGetBroadcastShape($a.shape.slice(0, -2), $b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  const a3D = transposeA ? reshape($a, [batchDimA, innerShapeA, outerShapeA]) : reshape($a, [batchDimA, outerShapeA, innerShapeA]);
+  const b3D = transposeB ? reshape($b, [batchDimB, outerShapeB, innerShapeB]) : reshape($b, [batchDimB, innerShapeB, outerShapeB]);
+  let $bias;
+  if (bias != null) {
+    $bias = convertToTensor(bias, "bias", "fused matMul");
+    [$bias] = makeTypesMatch($bias, $a);
+    assertAndGetBroadcastShape(outShape, $bias.shape);
+  }
+  let $preluActivationWeights;
+  if (preluActivationWeights != null) {
+    $preluActivationWeights = convertToTensor(preluActivationWeights, "prelu weights", "fused matMul");
+  }
+  const grad2 = (dy, saved) => {
+    const [a3D2, b3D2, y, $bias2] = saved;
+    const dyActivation = getFusedDyActivation(reshape(dy, y.shape), y, activation2);
+    let aDer;
+    let bDer;
+    if (!transposeA && !transposeB) {
+      aDer = matMul(dyActivation, b3D2, false, true);
+      bDer = matMul(a3D2, dyActivation, true, false);
+    } else if (!transposeA && transposeB) {
+      aDer = matMul(dyActivation, b3D2, false, false);
+      bDer = matMul(dyActivation, a3D2, true, false);
+    } else if (transposeA && !transposeB) {
+      aDer = matMul(b3D2, dyActivation, false, true);
+      bDer = matMul(a3D2, dyActivation, false, false);
+    } else {
+      aDer = matMul(b3D2, dyActivation, true, true);
+      bDer = matMul(dyActivation, a3D2, true, true);
+    }
+    if (bias != null) {
+      const biasDer = getFusedBiasGradient($bias2, dyActivation);
+      return [aDer, bDer, biasDer];
+    } else {
+      return [aDer, bDer];
+    }
+  };
+  const inputs = {
+    a: a3D,
+    b: b3D,
+    bias: $bias,
+    preluActivationWeights: $preluActivationWeights
+  };
+  const attrs = { transposeA, transposeB, activation: activation2, leakyreluAlpha };
+  if (bias == null) {
+    const customOp = customGrad((a3D2, b3D2, save) => {
+      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);
+      save([a3D2, b3D2, res]);
+      return { value: reshape(res, outShape), gradFunc: grad2 };
+    });
+    return customOp(a3D, b3D);
+  } else {
+    const customOpWithBias = customGrad((a3D2, b3D2, $bias2, save) => {
+      const res = ENGINE.runKernel(_FusedMatMul, inputs, attrs);
+      save([a3D2, b3D2, res, $bias2]);
+      return { value: reshape(res, outShape), gradFunc: grad2 };
+    });
+    return customOpWithBias(a3D, b3D, $bias);
+  }
+}
+var matMul2 = op({ fusedMatMul_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/signal/hamming_window.js
+function hammingWindow_(windowLength) {
+  return cosineWindow(windowLength, 0.54, 0.46);
+}
+var hammingWindow = op({ hammingWindow_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/signal/hann_window.js
+function hannWindow_(windowLength) {
+  return cosineWindow(windowLength, 0.5, 0.5);
+}
+var hannWindow = op({ hannWindow_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/signal/frame.js
+function frame_(signal2, frameLength, frameStep, padEnd = false, padValue = 0) {
+  let start = 0;
+  const output = [];
+  while (start + frameLength <= signal2.size) {
+    output.push(slice(signal2, start, frameLength));
+    start += frameStep;
+  }
+  if (padEnd) {
+    while (start < signal2.size) {
+      const padLen = start + frameLength - signal2.size;
+      const pad3 = concat([
+        slice(signal2, start, frameLength - padLen),
+        fill([padLen], padValue)
+      ]);
+      output.push(pad3);
+      start += frameStep;
+    }
+  }
+  if (output.length === 0) {
+    return tensor2d([], [0, frameLength]);
+  }
+  return reshape(concat(output), [output.length, frameLength]);
+}
+var frame = op({ frame_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/signal/stft.js
+function stft_(signal2, frameLength, frameStep, fftLength, windowFn = hannWindow) {
+  if (fftLength == null) {
+    fftLength = enclosingPowerOfTwo(frameLength);
+  }
+  const framedSignal = frame(signal2, frameLength, frameStep);
+  const windowedSignal = mul(framedSignal, windowFn(frameLength));
+  return rfft(windowedSignal, fftLength);
+}
+var stft = op({ stft_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/crop_and_resize.js
+function cropAndResize_(image2, boxes, boxInd, cropSize, method = "bilinear", extrapolationValue = 0) {
+  const $image = convertToTensor(image2, "image", "cropAndResize");
+  const $boxes = convertToTensor(boxes, "boxes", "cropAndResize", "float32");
+  const $boxInd = convertToTensor(boxInd, "boxInd", "cropAndResize", "int32");
+  const numBoxes = $boxes.shape[0];
+  assert($image.rank === 4, () => `Error in cropAndResize: image must be rank 4,but got rank ${$image.rank}.`);
+  assert($boxes.rank === 2 && $boxes.shape[1] === 4, () => `Error in cropAndResize: boxes must be have size [${numBoxes},4] but had shape ${$boxes.shape}.`);
+  assert($boxInd.rank === 1 && $boxInd.shape[0] === numBoxes, () => `Error in cropAndResize: boxInd must be have size [${numBoxes}] but had shape ${$boxes.shape}.`);
+  assert(cropSize.length === 2, () => `Error in cropAndResize: cropSize must be of length 2, but got length ${cropSize.length}.`);
+  assert(cropSize[0] >= 1 && cropSize[1] >= 1, () => `cropSize must be atleast [1,1], but was ${cropSize}`);
+  assert(method === "bilinear" || method === "nearest", () => `method must be bilinear or nearest, but was ${method}`);
+  const inputs = { image: $image, boxes: $boxes, boxInd: $boxInd };
+  const attrs = { method, extrapolationValue, cropSize };
+  const res = ENGINE.runKernel(CropAndResize, inputs, attrs);
+  return res;
+}
+var cropAndResize = op({ cropAndResize_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/flip_left_right.js
+function flipLeftRight_(image2) {
+  const $image = convertToTensor(image2, "image", "flipLeftRight", "float32");
+  assert($image.rank === 4, () => `Error in flipLeftRight: image must be rank 4,but got rank ${$image.rank}.`);
+  const inputs = { image: $image };
+  const res = ENGINE.runKernel(FlipLeftRight, inputs, {});
+  return res;
+}
+var flipLeftRight = op({ flipLeftRight_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/grayscale_to_rgb.js
+function grayscaleToRGB_(image2) {
+  const $image = convertToTensor(image2, "image", "grayscaleToRGB");
+  const lastDimsIdx = $image.rank - 1;
+  const lastDims = $image.shape[lastDimsIdx];
+  assert($image.rank >= 2, () => `Error in grayscaleToRGB: images must be at least rank 2, but got rank ${$image.rank}.`);
+  assert(lastDims === 1, () => `Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${lastDims}.`);
+  const reps = new Array($image.rank);
+  reps.fill(1, 0, lastDimsIdx);
+  reps[lastDimsIdx] = 3;
+  return tile($image, reps);
+}
+var grayscaleToRGB = op({ grayscaleToRGB_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/rotate_with_offset.js
+function rotateWithOffset_(image2, radians, fillValue = 0, center = 0.5) {
+  const $image = convertToTensor(image2, "image", "rotateWithOffset", "float32");
+  assert($image.rank === 4, () => `Error in rotateWithOffset: image must be rank 4,but got rank ${$image.rank}.`);
+  const inputs = { image: $image };
+  const attrs = { radians, fillValue, center };
+  const res = ENGINE.runKernel(RotateWithOffset, inputs, attrs);
+  return res;
+}
+var rotateWithOffset = op({ rotateWithOffset_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/nonmax_util.js
+function nonMaxSuppSanityCheck(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {
+  if (iouThreshold == null) {
+    iouThreshold = 0.5;
+  }
+  if (scoreThreshold == null) {
+    scoreThreshold = Number.NEGATIVE_INFINITY;
+  }
+  if (softNmsSigma == null) {
+    softNmsSigma = 0;
+  }
+  const numBoxes = boxes.shape[0];
+  maxOutputSize = Math.min(maxOutputSize, numBoxes);
+  assert(0 <= iouThreshold && iouThreshold <= 1, () => `iouThreshold must be in [0, 1], but was '${iouThreshold}'`);
+  assert(boxes.rank === 2, () => `boxes must be a 2D tensor, but was of rank '${boxes.rank}'`);
+  assert(boxes.shape[1] === 4, () => `boxes must have 4 columns, but 2nd dimension was ${boxes.shape[1]}`);
+  assert(scores.rank === 1, () => "scores must be a 1D tensor");
+  assert(scores.shape[0] === numBoxes, () => `scores has incompatible shape with boxes. Expected ${numBoxes}, but was ${scores.shape[0]}`);
+  assert(0 <= softNmsSigma && softNmsSigma <= 1, () => `softNmsSigma must be in [0, 1], but was '${softNmsSigma}'`);
+  return { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression.js
+function nonMaxSuppression_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression", "float32");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression", "float32");
+  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);
+  maxOutputSize = inputs.maxOutputSize;
+  iouThreshold = inputs.iouThreshold;
+  scoreThreshold = inputs.scoreThreshold;
+  const attrs = { maxOutputSize, iouThreshold, scoreThreshold };
+  return ENGINE.runKernel(NonMaxSuppressionV3, { boxes: $boxes, scores: $scores }, attrs);
+}
+var nonMaxSuppression = op({ nonMaxSuppression_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/non_max_suppression_util.js
+function binaryInsert(arr, element, comparator) {
+  const index = binarySearch(arr, element, comparator);
+  const insertionPoint = index < 0 ? -(index + 1) : index;
+  arr.splice(insertionPoint, 0, element);
+}
+function binarySearch(arr, target, comparator) {
+  return binarySearch_(arr, target, comparator || defaultComparator);
+}
+function defaultComparator(a, b) {
+  return a > b ? 1 : a < b ? -1 : 0;
+}
+function binarySearch_(arr, target, comparator) {
+  let left = 0;
+  let right = arr.length;
+  let middle = 0;
+  let found = false;
+  while (left < right) {
+    middle = left + (right - left >>> 1);
+    const compareResult = comparator(target, arr[middle]);
+    if (compareResult > 0) {
+      left = middle + 1;
+    } else {
+      right = middle;
+      found = !compareResult;
+    }
+  }
+  return found ? left : -left - 1;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/non_max_suppression_impl.js
+function nonMaxSuppressionV3Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold) {
+  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, 0);
+}
+function nonMaxSuppressionV4Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize) {
+  return nonMaxSuppressionImpl_(
+    boxes,
+    scores,
+    maxOutputSize,
+    iouThreshold,
+    scoreThreshold,
+    0,
+    false,
+    padToMaxOutputSize,
+    true
+  );
+}
+function nonMaxSuppressionV5Impl(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma) {
+  return nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, true);
+}
+function nonMaxSuppressionImpl_(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma, returnScoresTensor = false, padToMaxOutputSize = false, returnValidOutputs = false) {
+  const candidates = [];
+  for (let i = 0; i < scores.length; i++) {
+    if (scores[i] > scoreThreshold) {
+      candidates.push({ score: scores[i], boxIndex: i, suppressBeginIndex: 0 });
+    }
+  }
+  candidates.sort(ascendingComparator);
+  const scale2 = softNmsSigma > 0 ? -0.5 / softNmsSigma : 0;
+  const selectedIndices = [];
+  const selectedScores = [];
+  while (selectedIndices.length < maxOutputSize && candidates.length > 0) {
+    const candidate = candidates.pop();
+    const { score: originalScore, boxIndex, suppressBeginIndex } = candidate;
+    if (originalScore < scoreThreshold) {
+      break;
+    }
+    let ignoreCandidate = false;
+    for (let j = selectedIndices.length - 1; j >= suppressBeginIndex; --j) {
+      const iou = intersectionOverUnion(boxes, boxIndex, selectedIndices[j]);
+      if (iou >= iouThreshold) {
+        ignoreCandidate = true;
+        break;
+      }
+      candidate.score = candidate.score * suppressWeight(iouThreshold, scale2, iou);
+      if (candidate.score <= scoreThreshold) {
+        break;
+      }
+    }
+    candidate.suppressBeginIndex = selectedIndices.length;
+    if (!ignoreCandidate) {
+      if (candidate.score === originalScore) {
+        selectedIndices.push(boxIndex);
+        selectedScores.push(candidate.score);
+      } else if (candidate.score > scoreThreshold) {
+        binaryInsert(candidates, candidate, ascendingComparator);
+      }
+    }
+  }
+  const validOutputs = selectedIndices.length;
+  const elemsToPad = maxOutputSize - validOutputs;
+  if (padToMaxOutputSize && elemsToPad > 0) {
+    selectedIndices.push(...new Array(elemsToPad).fill(0));
+    selectedScores.push(...new Array(elemsToPad).fill(0));
+  }
+  const result = { selectedIndices };
+  if (returnScoresTensor) {
+    result["selectedScores"] = selectedScores;
+  }
+  if (returnValidOutputs) {
+    result["validOutputs"] = validOutputs;
+  }
+  return result;
+}
+function intersectionOverUnion(boxes, i, j) {
+  const iCoord = boxes.subarray(i * 4, i * 4 + 4);
+  const jCoord = boxes.subarray(j * 4, j * 4 + 4);
+  const yminI = Math.min(iCoord[0], iCoord[2]);
+  const xminI = Math.min(iCoord[1], iCoord[3]);
+  const ymaxI = Math.max(iCoord[0], iCoord[2]);
+  const xmaxI = Math.max(iCoord[1], iCoord[3]);
+  const yminJ = Math.min(jCoord[0], jCoord[2]);
+  const xminJ = Math.min(jCoord[1], jCoord[3]);
+  const ymaxJ = Math.max(jCoord[0], jCoord[2]);
+  const xmaxJ = Math.max(jCoord[1], jCoord[3]);
+  const areaI = (ymaxI - yminI) * (xmaxI - xminI);
+  const areaJ = (ymaxJ - yminJ) * (xmaxJ - xminJ);
+  if (areaI <= 0 || areaJ <= 0) {
+    return 0;
+  }
+  const intersectionYmin = Math.max(yminI, yminJ);
+  const intersectionXmin = Math.max(xminI, xminJ);
+  const intersectionYmax = Math.min(ymaxI, ymaxJ);
+  const intersectionXmax = Math.min(xmaxI, xmaxJ);
+  const intersectionArea = Math.max(intersectionYmax - intersectionYmin, 0) * Math.max(intersectionXmax - intersectionXmin, 0);
+  return intersectionArea / (areaI + areaJ - intersectionArea);
+}
+function suppressWeight(iouThreshold, scale2, iou) {
+  const weight = Math.exp(scale2 * iou * iou);
+  return iou <= iouThreshold ? weight : 0;
+}
+function ascendingComparator(c1, c2) {
+  return c1.score - c2.score || c1.score === c2.score && c2.boxIndex - c1.boxIndex;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_async.js
+async function nonMaxSuppressionAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const inputs = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold);
+  maxOutputSize = inputs.maxOutputSize;
+  iouThreshold = inputs.iouThreshold;
+  scoreThreshold = inputs.scoreThreshold;
+  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);
+  const boxesVals = boxesAndScores[0];
+  const scoresVals = boxesAndScores[1];
+  const { selectedIndices } = nonMaxSuppressionV3Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return tensor1d(selectedIndices, "int32");
+}
+var nonMaxSuppressionAsync = nonMaxSuppressionAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_with_score.js
+function nonMaxSuppressionWithScore_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  maxOutputSize = params.maxOutputSize;
+  iouThreshold = params.iouThreshold;
+  scoreThreshold = params.scoreThreshold;
+  softNmsSigma = params.softNmsSigma;
+  const inputs = { boxes: $boxes, scores: $scores };
+  const attrs = { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma };
+  const result = ENGINE.runKernel(NonMaxSuppressionV5, inputs, attrs);
+  return { selectedIndices: result[0], selectedScores: result[1] };
+}
+var nonMaxSuppressionWithScore = op({ nonMaxSuppressionWithScore_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_with_score_async.js
+async function nonMaxSuppressionWithScoreAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, softNmsSigma = 0) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  maxOutputSize = params.maxOutputSize;
+  iouThreshold = params.iouThreshold;
+  scoreThreshold = params.scoreThreshold;
+  softNmsSigma = params.softNmsSigma;
+  const boxesAndScores = await Promise.all([$boxes.data(), $scores.data()]);
+  const boxesVals = boxesAndScores[0];
+  const scoresVals = boxesAndScores[1];
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return {
+    selectedIndices: tensor1d(selectedIndices, "int32"),
+    selectedScores: tensor1d(selectedScores)
+  };
+}
+var nonMaxSuppressionWithScoreAsync = nonMaxSuppressionWithScoreAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_padded.js
+function nonMaxSuppressionPadded_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppression");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppression");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);
+  const $maxOutputSize = params.maxOutputSize;
+  const $iouThreshold = params.iouThreshold;
+  const $scoreThreshold = params.scoreThreshold;
+  const inputs = { boxes: $boxes, scores: $scores };
+  const attrs = {
+    maxOutputSize: $maxOutputSize,
+    iouThreshold: $iouThreshold,
+    scoreThreshold: $scoreThreshold,
+    padToMaxOutputSize
+  };
+  const result = ENGINE.runKernel(NonMaxSuppressionV4, inputs, attrs);
+  return { selectedIndices: result[0], validOutputs: result[1] };
+}
+var nonMaxSuppressionPadded = op({ nonMaxSuppressionPadded_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/non_max_suppression_padded_async.js
+async function nonMaxSuppressionPaddedAsync_(boxes, scores, maxOutputSize, iouThreshold = 0.5, scoreThreshold = Number.NEGATIVE_INFINITY, padToMaxOutputSize = false) {
+  const $boxes = convertToTensor(boxes, "boxes", "nonMaxSuppressionAsync");
+  const $scores = convertToTensor(scores, "scores", "nonMaxSuppressionAsync");
+  const params = nonMaxSuppSanityCheck($boxes, $scores, maxOutputSize, iouThreshold, scoreThreshold, null);
+  const $maxOutputSize = params.maxOutputSize;
+  const $iouThreshold = params.iouThreshold;
+  const $scoreThreshold = params.scoreThreshold;
+  const [boxesVals, scoresVals] = await Promise.all([$boxes.data(), $scores.data()]);
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl(boxesVals, scoresVals, $maxOutputSize, $iouThreshold, $scoreThreshold, padToMaxOutputSize);
+  if ($boxes !== boxes) {
+    $boxes.dispose();
+  }
+  if ($scores !== scores) {
+    $scores.dispose();
+  }
+  return {
+    selectedIndices: tensor1d(selectedIndices, "int32"),
+    validOutputs: scalar(validOutputs, "int32")
+  };
+}
+var nonMaxSuppressionPaddedAsync = nonMaxSuppressionPaddedAsync_;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/resize_bilinear.js
+function resizeBilinear_(images, size, alignCorners = false, halfPixelCenters = false) {
+  const $images = convertToTensor(images, "images", "resizeBilinear");
+  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeBilinear: x must be rank 3 or 4, but got rank ${$images.rank}.`);
+  assert(size.length === 2, () => `Error in resizeBilinear: new shape must 2D, but got shape ${size}.`);
+  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.`);
+  let batchImages = $images;
+  let reshapedTo4D = false;
+  if ($images.rank === 3) {
+    reshapedTo4D = true;
+    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);
+  }
+  const [] = size;
+  const inputs = { images: batchImages };
+  const attrs = { alignCorners, halfPixelCenters, size };
+  const res = ENGINE.runKernel(ResizeBilinear, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var resizeBilinear = op({ resizeBilinear_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/resize_nearest_neighbor.js
+function resizeNearestNeighbor_(images, size, alignCorners = false, halfPixelCenters = false) {
+  const $images = convertToTensor(images, "images", "resizeNearestNeighbor");
+  assert($images.rank === 3 || $images.rank === 4, () => `Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${$images.rank}.`);
+  assert(size.length === 2, () => `Error in resizeNearestNeighbor: new shape must 2D, but got shape ${size}.`);
+  assert($images.dtype === "float32" || $images.dtype === "int32", () => "`images` must have `int32` or `float32` as dtype");
+  assert(halfPixelCenters === false || alignCorners === false, () => `Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.`);
+  let batchImages = $images;
+  let reshapedTo4D = false;
+  if ($images.rank === 3) {
+    reshapedTo4D = true;
+    batchImages = reshape($images, [1, $images.shape[0], $images.shape[1], $images.shape[2]]);
+  }
+  const [] = size;
+  const inputs = { images: batchImages };
+  const attrs = { alignCorners, halfPixelCenters, size };
+  const res = ENGINE.runKernel(ResizeNearestNeighbor, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var resizeNearestNeighbor = op({ resizeNearestNeighbor_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/threshold.js
+function threshold_(image2, method = "binary", inverted = false, threshValue = 0.5) {
+  const $image = convertToTensor(image2, "image", "threshold");
+  const RED_INTENCITY_COEF = 0.2989;
+  const GREEN_INTENCITY_COEF = 0.587;
+  const BLUE_INTENCITY_COEF = 0.114;
+  const totalPixelsInImage = $image.shape[0] * $image.shape[1];
+  let $threshold = mul(tensor1d([threshValue]), 255);
+  let r, g, b, grayscale;
+  assert($image.rank === 3, () => `Error in threshold: image must be rank 3,but got rank ${$image.rank}.`);
+  assert($image.shape[2] === 3 || $image.shape[2] === 1, () => `Error in threshold: image color channel must be equal to 3 or 1but got ${$image.shape[2]}.`);
+  assert($image.dtype === "int32" || $image.dtype === "float32", () => `Error in dtype: image dtype must be int32 or float32,but got dtype ${$image.dtype}.`);
+  assert(method === "otsu" || method === "binary", () => `Method must be binary or otsu, but was ${method}`);
+  if ($image.shape[2] === 3) {
+    [r, g, b] = split($image, [1, 1, 1], -1);
+    const $r = mul(r, RED_INTENCITY_COEF);
+    const $g = mul(g, GREEN_INTENCITY_COEF);
+    const $b = mul(b, BLUE_INTENCITY_COEF);
+    grayscale = add2(add2($r, $g), $b);
+  } else {
+    grayscale = image2;
+  }
+  if (method === "otsu") {
+    const $histogram = bincount(cast(round2(grayscale), "int32"), tensor([]), 256);
+    $threshold = otsu($histogram, totalPixelsInImage);
+  }
+  const invCondition = inverted ? lessEqual(grayscale, $threshold) : greater(grayscale, $threshold);
+  const result = cast(mul(invCondition, 255), "int32");
+  return result;
+}
+function otsu(histogram, total) {
+  let bestThresh = tensor1d([-1]);
+  let bestInBetVar = tensor1d([0]);
+  let cInBetVar = tensor1d([0]);
+  let classFirst, classSecond, meanFirst, meanSec, weightForeground, weightBack;
+  for (let index = 0; index < histogram.size - 1; index++) {
+    classFirst = slice(histogram, 0, index + 1);
+    classSecond = slice(histogram, index + 1);
+    weightForeground = div(sum2(classFirst), total);
+    weightBack = div(sum2(classSecond), total);
+    const meanFirstDivA = sum2(mul(classFirst, range(0, classFirst.size)));
+    meanFirst = div(meanFirstDivA, sum2(classFirst));
+    const meanSecFill = fill(classSecond.shape, classFirst.size);
+    const meanSecAdd = add2(range(0, classSecond.size), meanSecFill);
+    const meanSecMul = mul(classSecond, meanSecAdd);
+    meanSec = div(sum2(meanSecMul), sum2(classSecond));
+    const cInBetVarSubA = sub(meanFirst, meanSec);
+    const cInBetVarSubB = sub(meanFirst, meanSec);
+    const cInBetVarMul = mul(weightForeground, weightBack);
+    cInBetVar = mul(mul(cInBetVarMul, cInBetVarSubA), cInBetVarSubB);
+    const condition = greater(cInBetVar, bestInBetVar);
+    bestInBetVar = where(condition, cInBetVar, bestInBetVar);
+    bestThresh = where(condition, tensor1d([index]), bestThresh);
+  }
+  return bestThresh;
+}
+var threshold = op({ threshold_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/image/transform.js
+function transform_(image2, transforms, interpolation = "nearest", fillMode = "constant", fillValue = 0, outputShape) {
+  const $image = convertToTensor(image2, "image", "transform", "float32");
+  const $transforms = convertToTensor(transforms, "transforms", "transform", "float32");
+  assert($image.rank === 4, () => `Error in transform: image must be rank 4,but got rank ${$image.rank}.`);
+  assert($transforms.rank === 2 && ($transforms.shape[0] === $image.shape[0] || $transforms.shape[0] === 1) && $transforms.shape[1] === 8, () => `Error in transform: Input transform should be batch x 8 or 1 x 8`);
+  assert(outputShape == null || outputShape.length === 2, () => `Error in transform: outputShape must be [height, width] or null, but got ${outputShape}.`);
+  const inputs = { image: $image, transforms: $transforms };
+  const attrs = { interpolation, fillMode, fillValue, outputShape };
+  return ENGINE.runKernel(Transform, inputs, attrs);
+}
+var transform = op({ transform_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/band_part.js
+function bandPart_(a, numLower, numUpper) {
+  assert(numLower % 1 === 0, () => `bandPart(): numLower must be an integer, got ${numLower}.`);
+  assert(numUpper % 1 === 0, () => `bandPart(): numUpper must be an integer, got ${numUpper}.`);
+  const $a = convertToTensor(a, "a", "bandPart");
+  assert($a.rank >= 2, () => `bandPart(): Rank must be at least 2, got ${$a.rank}.`);
+  const shape = $a.shape;
+  const [M, N] = $a.shape.slice(-2);
+  if (!(numLower <= M)) {
+    throw new Error(`bandPart(): numLower (${numLower}) must not be greater than the number of rows (${M}).`);
+  }
+  if (!(numUpper <= N)) {
+    throw new Error(`bandPart(): numUpper (${numUpper}) must not be greater than the number of columns (${N}).`);
+  }
+  if (numLower < 0) {
+    numLower = M;
+  }
+  if (numUpper < 0) {
+    numUpper = N;
+  }
+  const i = reshape(range(0, M, 1, "int32"), [-1, 1]);
+  const j = range(0, N, 1, "int32");
+  const ij = sub(i, j);
+  const inBand = logicalAnd(lessEqual(ij, scalar(+numLower, "int32")), greaterEqual(ij, scalar(-numUpper, "int32")));
+  const zero = zeros([M, N], $a.dtype);
+  return reshape(stack(unstack(reshape($a, [-1, M, N])).map((mat) => where(inBand, mat, zero))), shape);
+}
+var bandPart = op({ bandPart_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/gram_schmidt.js
+function gramSchmidt_(xs) {
+  let inputIsTensor2D;
+  if (Array.isArray(xs)) {
+    inputIsTensor2D = false;
+    assert(xs != null && xs.length > 0, () => "Gram-Schmidt process: input must not be null, undefined, or empty");
+    const dim = xs[0].shape[0];
+    for (let i = 1; i < xs.length; ++i) {
+      assert(xs[i].shape[0] === dim, () => `Gram-Schmidt: Non-unique lengths found in the input vectors: (${xs[i].shape[0]} vs. ${dim})`);
+    }
+  } else {
+    inputIsTensor2D = true;
+    xs = split(xs, xs.shape[0], 0).map((x) => squeeze(x, [0]));
+  }
+  assert(xs.length <= xs[0].shape[0], () => `Gram-Schmidt: Number of vectors (${xs.length}) exceeds number of dimensions (${xs[0].shape[0]}).`);
+  const ys = [];
+  const xs1d = xs;
+  for (let i = 0; i < xs.length; ++i) {
+    ys.push(ENGINE.tidy(() => {
+      let x = xs1d[i];
+      if (i > 0) {
+        for (let j = 0; j < i; ++j) {
+          const proj = mul(sum2(mul(ys[j], x)), ys[j]);
+          x = sub(x, proj);
+        }
+      }
+      return div(x, norm(x, "euclidean"));
+    }));
+  }
+  if (inputIsTensor2D) {
+    return stack(ys, 0);
+  } else {
+    return ys;
+  }
+}
+var gramSchmidt = op({ gramSchmidt_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/linalg/qr.js
+function qr_(x, fullMatrices = false) {
+  assert(x.rank >= 2, () => `qr() requires input tensor to have a rank >= 2, but got rank ${x.rank}`);
+  if (x.rank === 2) {
+    return qr2d(x, fullMatrices);
+  } else {
+    const outerDimsProd = x.shape.slice(0, x.shape.length - 2).reduce((value, prev) => value * prev);
+    const x2ds = unstack(reshape(x, [
+      outerDimsProd,
+      x.shape[x.shape.length - 2],
+      x.shape[x.shape.length - 1]
+    ]), 0);
+    const q2ds = [];
+    const r2ds = [];
+    x2ds.forEach((x2d) => {
+      const [q2d, r2d] = qr2d(x2d, fullMatrices);
+      q2ds.push(q2d);
+      r2ds.push(r2d);
+    });
+    const q = reshape(stack(q2ds, 0), x.shape);
+    const r = reshape(stack(r2ds, 0), x.shape);
+    return [q, r];
+  }
+}
+function qr2d(x, fullMatrices = false) {
+  return ENGINE.tidy(() => {
+    assert(x.shape.length === 2, () => `qr2d() requires a 2D Tensor, but got a ${x.shape.length}D Tensor.`);
+    const m = x.shape[0];
+    const n = x.shape[1];
+    let q = eye(m);
+    let r = clone(x);
+    const one2D = tensor2d([[1]], [1, 1]);
+    let w = clone(one2D);
+    const iters = m >= n ? n : m;
+    for (let j = 0; j < iters; ++j) {
+      const rTemp = r;
+      const wTemp = w;
+      const qTemp = q;
+      [w, r, q] = ENGINE.tidy(() => {
+        const rjEnd1 = slice(r, [j, j], [m - j, 1]);
+        const normX = norm(rjEnd1);
+        const rjj = slice(r, [j, j], [1, 1]);
+        const s = where(greater(rjj, 0), tensor2d([[-1]]), tensor2d([[1]]));
+        const u1 = sub(rjj, mul(s, normX));
+        const wPre = div(rjEnd1, u1);
+        if (wPre.shape[0] === 1) {
+          w = clone(one2D);
+        } else {
+          w = concat([
+            one2D,
+            slice(wPre, [1, 0], [wPre.shape[0] - 1, wPre.shape[1]])
+          ], 0);
+        }
+        const tau = neg(div(matMul(s, u1), normX));
+        const rjEndAll = slice(r, [j, 0], [m - j, n]);
+        const tauTimesW = mul(tau, w);
+        const wT = transpose(w);
+        if (j === 0) {
+          r = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));
+        } else {
+          const rTimesTau = sub(rjEndAll, matMul(tauTimesW, matMul(wT, rjEndAll)));
+          r = concat([slice(r, [0, 0], [j, n]), rTimesTau], 0);
+        }
+        const tawTimesWT = transpose(tauTimesW);
+        const qAllJEnd = slice(q, [0, j], [m, q.shape[1] - j]);
+        if (j === 0) {
+          q = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));
+        } else {
+          const qTimesTau = sub(qAllJEnd, matMul(matMul(qAllJEnd, w), tawTimesWT));
+          q = concat([slice(q, [0, 0], [m, j]), qTimesTau], 1);
+        }
+        return [w, r, q];
+      });
+      dispose([rTemp, wTemp, qTemp]);
+    }
+    if (!fullMatrices && m > n) {
+      q = slice(q, [0, 0], [m, n]);
+      r = slice(r, [0, 0], [n, n]);
+    }
+    return [q, r];
+  });
+}
+var qr = op({ qr_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/loss_ops_utils.js
+var Reduction;
+(function(Reduction2) {
+  Reduction2[Reduction2["NONE"] = 0] = "NONE";
+  Reduction2[Reduction2["MEAN"] = 1] = "MEAN";
+  Reduction2[Reduction2["SUM"] = 2] = "SUM";
+  Reduction2[Reduction2["SUM_BY_NONZERO_WEIGHTS"] = 3] = "SUM_BY_NONZERO_WEIGHTS";
+})(Reduction || (Reduction = {}));
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/compute_weighted_loss.js
+function computeWeightedLoss_(losses2, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $losses = convertToTensor(losses2, "losses", "computeWeightedLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "computeWeightedLoss");
+  }
+  const weightedLoss = $weights == null ? $losses : mul($losses, $weights);
+  if (reduction === Reduction.NONE) {
+    return weightedLoss;
+  }
+  if (reduction === Reduction.SUM) {
+    return sum2(weightedLoss);
+  }
+  if (reduction === Reduction.MEAN) {
+    if ($weights == null) {
+      return mean(weightedLoss);
+    } else {
+      const broadcastFactor = $losses.size / $weights.size;
+      const result = div(sum2(weightedLoss), sum2($weights));
+      return broadcastFactor > 1 ? div(result, scalar(broadcastFactor)) : result;
+    }
+  }
+  if (reduction === Reduction.SUM_BY_NONZERO_WEIGHTS) {
+    if ($weights == null) {
+      return div(sum2(weightedLoss), scalar($losses.size));
+    } else {
+      const broadcastedWeights = mul($weights, ones2($losses.shape));
+      const numNonZeros = cast(sum2(notEqual(broadcastedWeights, scalar(0))), "float32");
+      return div(sum2(weightedLoss), numNonZeros);
+    }
+  }
+  throw Error(`Unknown reduction: ${reduction}`);
+}
+var computeWeightedLoss = op({ computeWeightedLoss_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/absolute_difference.js
+function absoluteDifference_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "absoluteDifference");
+  const $predictions = convertToTensor(predictions, "predictions", "absoluteDifference");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "absoluteDifference");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in absoluteDifference: ");
+  const losses2 = abs(sub($labels, $predictions));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var absoluteDifference = op({ absoluteDifference_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/cosine_distance.js
+function cosineDistance_(labels, predictions, axis, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "cosineDistance");
+  const $predictions = convertToTensor(predictions, "predictions", "cosineDistance");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "cosineDistance");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in cosineDistance: ");
+  const one = scalar(1);
+  const losses2 = sub(one, sum2(mul($labels, $predictions), axis, true));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var cosineDistance = op({ cosineDistance_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/hinge_loss.js
+function hingeLoss_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $labels = convertToTensor(labels, "labels", "hingeLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "hingeLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "hingeLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in hingeLoss: ");
+  const one = scalar(1);
+  $labels = sub(mul(scalar(2), $labels), one);
+  const losses2 = relu(sub(one, mul($labels, $predictions)));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var hingeLoss = op({ hingeLoss_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/huber_loss.js
+function huberLoss_(labels, predictions, weights, delta = 1, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "huberLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "huberLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "huberLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in huberLoss: ");
+  const deltaScalar = scalar(delta);
+  const error = abs(sub($predictions, $labels));
+  const quadratic = minimum(error, deltaScalar);
+  const linear = sub(error, quadratic);
+  const losses2 = add2(mul(scalar(0.5), square(quadratic)), mul(deltaScalar, linear));
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var huberLoss = op({ huberLoss_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/log_loss.js
+function logLoss_(labels, predictions, weights, epsilon3 = 1e-7, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "logLoss");
+  const $predictions = convertToTensor(predictions, "predictions", "logLoss");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "logLoss");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in logLoss: ");
+  const one = scalar(1);
+  const epsilonScalar = scalar(epsilon3);
+  const l13 = neg(mul($labels, log2(add2($predictions, epsilonScalar))));
+  const l23 = mul(sub(one, $labels), log2(add2(sub(one, $predictions), epsilonScalar)));
+  const losses2 = sub(l13, l23);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var logLoss = op({ logLoss_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/mean_squared_error.js
+function meanSquaredError_(labels, predictions, weights, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  const $labels = convertToTensor(labels, "labels", "meanSquaredError");
+  const $predictions = convertToTensor(predictions, "predictions", "meanSquaredError");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "meanSquaredError");
+  }
+  assertShapesMatch($labels.shape, $predictions.shape, "Error in meanSquaredError: ");
+  const losses2 = squaredDifference($labels, $predictions);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var meanSquaredError = op({ meanSquaredError_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/sigmoid_cross_entropy.js
+function sigmoidCrossEntropyWithLogits_(labels, logits) {
+  const $labels = convertToTensor(labels, "labels", "sigmoidCrossEntropyWithLogits");
+  const $logits = convertToTensor(logits, "logits", "sigmoidCrossEntropyWithLogits");
+  assertShapesMatch($labels.shape, $logits.shape, "Error in sigmoidCrossEntropyWithLogits: ");
+  const maxOutput = relu($logits);
+  const outputXTarget = mul($logits, $labels);
+  const sigmoidOutput = log1p(exp(neg(abs($logits))));
+  return add2(sub(maxOutput, outputXTarget), sigmoidOutput);
+}
+function sigmoidCrossEntropy_(multiClassLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $multiClassLabels = convertToTensor(multiClassLabels, "multiClassLabels", "sigmoidCrossEntropy");
+  const $logits = convertToTensor(logits, "logits", "sigmoidCrossEntropy");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "sigmoidCrossEntropy");
+  }
+  assertShapesMatch($multiClassLabels.shape, $logits.shape, "Error in sigmoidCrossEntropy: ");
+  if (labelSmoothing > 0) {
+    const labelSmoothingScalar = scalar(labelSmoothing);
+    const one = scalar(1);
+    const half = scalar(0.5);
+    $multiClassLabels = add2(mul($multiClassLabels, sub(one, labelSmoothingScalar)), mul(half, labelSmoothingScalar));
+  }
+  const losses2 = sigmoidCrossEntropyWithLogits_($multiClassLabels, $logits);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var sigmoidCrossEntropy = op({ sigmoidCrossEntropy_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/losses/softmax_cross_entropy.js
+function softmaxCrossEntropyWithLogits_(labels, logits, dim = -1) {
+  if (dim === -1) {
+    dim = logits.rank - 1;
+  }
+  if (dim !== logits.rank - 1) {
+    throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${logits.rank} and dim was ${dim}`);
+  }
+  const customOp = customGrad((labels2, logits2, save) => {
+    const keepDims = true;
+    const lse = logSumExp(logits2, [dim], keepDims);
+    const logResult = sub(cast(logits2, "float32"), lse);
+    save([labels2, logResult]);
+    const costVector = neg(mul(logResult, labels2));
+    const value = sum2(costVector, [dim]);
+    const gradFunc = (dy, saved) => {
+      const [labels3, logResult2] = saved;
+      const dyShape = expandShapeToKeepDim(dy.shape, [dim]);
+      return [
+        mul(reshape(dy, dyShape), sub(cast(labels3, "float32"), exp(logResult2))),
+        mul(reshape(dy, dyShape), sub(exp(logResult2), cast(labels3, "float32")))
+      ];
+    };
+    return { value, gradFunc };
+  });
+  return customOp(labels, logits);
+}
+function softmaxCrossEntropy_(onehotLabels, logits, weights, labelSmoothing = 0, reduction = Reduction.SUM_BY_NONZERO_WEIGHTS) {
+  let $onehotLabels = convertToTensor(onehotLabels, "onehotLabels", "softmaxCrossEntropy");
+  const $logits = convertToTensor(logits, "logits", "softmaxCrossEntropy");
+  let $weights = null;
+  if (weights != null) {
+    $weights = convertToTensor(weights, "weights", "softmaxCrossEntropy");
+  }
+  assertShapesMatch($onehotLabels.shape, $logits.shape, "Error in softmaxCrossEntropy: ");
+  if (labelSmoothing > 0) {
+    const labelSmoothingScalar = scalar(labelSmoothing);
+    const one = scalar(1);
+    const numClasses = scalar($onehotLabels.shape[1]);
+    $onehotLabels = add2(mul($onehotLabels, sub(one, labelSmoothingScalar)), div(labelSmoothingScalar, numClasses));
+  }
+  const losses2 = softmaxCrossEntropyWithLogits_($onehotLabels, $logits);
+  return computeWeightedLoss(losses2, $weights, reduction);
+}
+var softmaxCrossEntropy = op({ softmaxCrossEntropy_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_fill_empty_rows.js
+function sparseFillEmptyRows_(indices, values, denseShape, defaultValue) {
+  const $indices = convertToTensor(indices, "indices", "sparseFillEmptyRows", "int32");
+  const $values = convertToTensor(values, "values", "sparseFillEmptyRows");
+  const $denseShape = convertToTensor(denseShape, "denseShape", "sparseFillEmptyRows", "int32");
+  const $defaultValue = convertToTensor(defaultValue, "defaultValue", "sparseFillEmptyRows", $values.dtype);
+  if ($indices.rank !== 2) {
+    throw new Error(`Indices should be Tensor2D but received shape
+        ${$indices.shape}`);
+  }
+  if ($values.rank !== 1) {
+    throw new Error(`Values should be Tensor1D but received shape ${$values.shape}`);
+  }
+  if ($denseShape.rank !== 1) {
+    throw new Error(`Dense shape should be Tensor1D but received shape ${$denseShape.shape}`);
+  }
+  if ($defaultValue.rank !== 0) {
+    throw new Error(`Default value should be a scalar but received shape ${$defaultValue.shape}`);
+  }
+  const inputs = {
+    indices: $indices,
+    values: $values,
+    denseShape: $denseShape,
+    defaultValue: $defaultValue
+  };
+  const result = ENGINE.runKernel(SparseFillEmptyRows, inputs);
+  return {
+    outputIndices: result[0],
+    outputValues: result[1],
+    emptyRowIndicator: result[2],
+    reverseIndexMap: result[3]
+  };
+}
+var sparseFillEmptyRows = op({ sparseFillEmptyRows_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_reshape.js
+function sparseReshape_(inputIndices, inputShape, newShape) {
+  const $inputIndices = convertToTensor(inputIndices, "inputIndices", "sparseReshape", "int32");
+  const $inputShape = convertToTensor(inputShape, "inputShape", "sparseReshape", "int32");
+  const $newShape = convertToTensor(newShape, "newShape", "sparseReshape", "int32");
+  if ($inputIndices.rank !== 2) {
+    throw new Error(`Input indices should be Tensor2D but received shape
+        ${$inputIndices.shape}`);
+  }
+  if ($inputShape.rank !== 1) {
+    throw new Error(`Input shape should be Tensor1D but received shape ${$inputShape.shape}`);
+  }
+  if ($newShape.rank !== 1) {
+    throw new Error(`New shape should be Tensor1D but received shape ${$newShape.shape}`);
+  }
+  const inputs = {
+    inputIndices: $inputIndices,
+    inputShape: $inputShape,
+    newShape: $newShape
+  };
+  const result = ENGINE.runKernel(SparseReshape, inputs);
+  return { outputIndices: result[0], outputShape: result[1] };
+}
+var sparseReshape = op({ sparseReshape_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_mean.js
+function sparseSegmentMean_(data, indices, segmentIds) {
+  const $data = convertToTensor(data, "data", "sparseSegmentMean");
+  const $indices = convertToTensor(indices, "indices", "sparseSegmentMean", "int32");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "sparseSegmentMean", "int32");
+  if ($data.rank < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if ($indices.rank !== 1) {
+    throw new Error(`Indices should be Tensor1D but received shape
+          ${$indices.shape}`);
+  }
+  if ($segmentIds.rank !== 1) {
+    throw new Error(`Segment ids should be Tensor1D but received shape
+          ${$segmentIds.shape}`);
+  }
+  const inputs = {
+    data: $data,
+    indices: $indices,
+    segmentIds: $segmentIds
+  };
+  return ENGINE.runKernel(SparseSegmentMean, inputs);
+}
+var sparseSegmentMean = op({ sparseSegmentMean_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_sum.js
+function sparseSegmentSum_(data, indices, segmentIds) {
+  const $data = convertToTensor(data, "data", "sparseSegmentSum");
+  const $indices = convertToTensor(indices, "indices", "sparseSegmentSum", "int32");
+  const $segmentIds = convertToTensor(segmentIds, "segmentIds", "sparseSegmentSum", "int32");
+  if ($data.rank < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if ($indices.rank !== 1) {
+    throw new Error(`Indices should be Tensor1D but received shape
+         ${$indices.shape}`);
+  }
+  if ($segmentIds.rank !== 1) {
+    throw new Error(`Segment ids should be Tensor1D but received shape
+         ${$segmentIds.shape}`);
+  }
+  const inputs = {
+    data: $data,
+    indices: $indices,
+    segmentIds: $segmentIds
+  };
+  return ENGINE.runKernel(SparseSegmentSum, inputs);
+}
+var sparseSegmentSum = op({ sparseSegmentSum_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_n_grams.js
+function stringNGrams_(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+  const $data = convertToTensor(data, "data", "stringNGrams", "string");
+  if ($data.dtype !== "string") {
+    throw new Error("Data must be of datatype string");
+  }
+  if ($data.shape.length !== 1) {
+    throw new Error(`Data must be a vector, saw: ${$data.shape}`);
+  }
+  const $dataSplits = convertToTensor(dataSplits, "dataSplits", "stringNGrams");
+  if ($dataSplits.dtype !== "int32") {
+    throw new Error("Data splits must be of datatype int32");
+  }
+  const attrs = {
+    separator,
+    nGramWidths,
+    leftPad,
+    rightPad: rightPad2,
+    padWidth,
+    preserveShortSequences
+  };
+  const inputs = { data: $data, dataSplits: $dataSplits };
+  const result = ENGINE.runKernel(StringNGrams, inputs, attrs);
+  return { nGrams: result[0], nGramsSplits: result[1] };
+}
+var stringNGrams = op({ stringNGrams_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_split.js
+function stringSplit_(input2, delimiter, skipEmpty = true) {
+  const $input = convertToTensor(input2, "input", "stringSplit", "string");
+  const $delimiter = convertToTensor(delimiter, "delimiter", "stringSplit", "string");
+  if ($input.rank !== 1) {
+    throw new Error(`Input should be Tensor1D but received shape ${$input.shape}`);
+  }
+  if ($delimiter.rank !== 0) {
+    throw new Error(`Delimiter should be a scalar but received shape ${$delimiter.shape}`);
+  }
+  const attrs = { skipEmpty };
+  const inputs = { input: $input, delimiter: $delimiter };
+  const result = ENGINE.runKernel(StringSplit, inputs, attrs);
+  return { indices: result[0], values: result[1], shape: result[2] };
+}
+var stringSplit = op({ stringSplit_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/string/string_to_hash_bucket_fast.js
+function stringToHashBucketFast_(input2, numBuckets) {
+  const $input = convertToTensor(input2, "input", "stringToHashBucketFast", "string");
+  const attrs = { numBuckets };
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const inputs = { input: $input };
+  return ENGINE.runKernel(StringToHashBucketFast, inputs, attrs);
+}
+var stringToHashBucketFast = op({ stringToHashBucketFast_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ops.js
+var spectral = {
+  fft,
+  ifft,
+  rfft,
+  irfft
+};
+var signal = {
+  hammingWindow,
+  hannWindow,
+  frame,
+  stft
+};
+var image = {
+  flipLeftRight,
+  grayscaleToRGB,
+  resizeNearestNeighbor,
+  resizeBilinear,
+  rotateWithOffset,
+  cropAndResize,
+  nonMaxSuppression,
+  nonMaxSuppressionAsync,
+  nonMaxSuppressionWithScore,
+  nonMaxSuppressionWithScoreAsync,
+  nonMaxSuppressionPadded,
+  nonMaxSuppressionPaddedAsync,
+  threshold,
+  transform
+};
+var linalg = {
+  bandPart,
+  gramSchmidt,
+  qr
+};
+var losses = {
+  absoluteDifference,
+  computeWeightedLoss,
+  cosineDistance,
+  hingeLoss,
+  huberLoss,
+  logLoss,
+  meanSquaredError,
+  sigmoidCrossEntropy,
+  softmaxCrossEntropy
+};
+var sparse = {
+  sparseFillEmptyRows,
+  sparseReshape,
+  sparseSegmentMean,
+  sparseSegmentSum
+};
+var string = {
+  stringNGrams,
+  stringSplit,
+  stringToHashBucketFast
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/optimizer.js
+var Optimizer = class extends Serializable {
+  minimize(f, returnCost = false, varList) {
+    const { value, grads: grads2 } = this.computeGradients(f, varList);
+    if (varList != null) {
+      const gradArray = varList.map((v) => ({ name: v.name, tensor: grads2[v.name] }));
+      this.applyGradients(gradArray);
+    } else {
+      this.applyGradients(grads2);
+    }
+    dispose(grads2);
+    if (returnCost) {
+      return value;
+    } else {
+      value.dispose();
+      return null;
+    }
+  }
+  get iterations() {
+    if (this.iterations_ == null) {
+      this.iterations_ = 0;
+    }
+    return this.iterations_;
+  }
+  incrementIterations() {
+    this.iterations_ = this.iterations + 1;
+  }
+  computeGradients(f, varList) {
+    return variableGrads(f, varList);
+  }
+  dispose() {
+    if (this.iterations_ != null) {
+      dispose(this.iterations_);
+    }
+  }
+  async saveIterations() {
+    if (this.iterations_ == null) {
+      this.iterations_ = 0;
+    }
+    return {
+      name: "iter",
+      tensor: scalar(this.iterations_, "int32")
+    };
+  }
+  async getWeights() {
+    throw new Error("getWeights() is not implemented for this optimizer yet.");
+  }
+  async setWeights(weightValues) {
+    throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`);
+  }
+  async extractIterations(weightValues) {
+    this.iterations_ = (await weightValues[0].tensor.data())[0];
+    return weightValues.slice(1);
+  }
+};
+Object.defineProperty(Optimizer, Symbol.hasInstance, {
+  value: (instance) => {
+    return instance.minimize != null && instance.computeGradients != null && instance.applyGradients != null;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/adadelta_optimizer.js
+var AdadeltaOptimizer = class extends Optimizer {
+  constructor(learningRate, rho, epsilon3 = null) {
+    super();
+    this.learningRate = learningRate;
+    this.rho = rho;
+    this.epsilon = epsilon3;
+    this.accumulatedGrads = [];
+    this.accumulatedUpdates = [];
+    if (epsilon3 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      const trainable = false;
+      if (this.accumulatedGrads[i] == null) {
+        this.accumulatedGrads[i] = {
+          originalName: `${name}/accum_grad`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedUpdates[i] == null) {
+        this.accumulatedUpdates[i] = {
+          originalName: `${name}/accum_var`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedGrad = this.accumulatedGrads[i].variable;
+      const accumulatedUpdate = this.accumulatedUpdates[i].variable;
+      tidy(() => {
+        const newAccumulatedGrad = add2(mul(accumulatedGrad, this.rho), mul(square(gradient), 1 - this.rho));
+        const updates = mul(div(sqrt(add2(accumulatedUpdate, this.epsilon)), sqrt(add2(accumulatedGrad, this.epsilon))), gradient);
+        const newAccumulatedUpdate = add2(mul(accumulatedUpdate, this.rho), mul(square(updates), 1 - this.rho));
+        accumulatedGrad.assign(newAccumulatedGrad);
+        accumulatedUpdate.assign(newAccumulatedUpdate);
+        const newValue = add2(mul(updates, -this.learningRate), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedUpdates != null) {
+      dispose(this.accumulatedGrads.map((v) => v.variable));
+      dispose(this.accumulatedUpdates.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedGrads, ...this.accumulatedUpdates];
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const variableCount = weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedGrads = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedUpdates = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "rho": this.rho,
+      "epsilon": this.epsilon
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["rho"], config["epsilon"]);
+  }
+};
+AdadeltaOptimizer.className = "Adadelta";
+registerClass(AdadeltaOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/adagrad_optimizer.js
+var AdagradOptimizer = class extends Optimizer {
+  constructor(learningRate, initialAccumulatorValue = 0.1) {
+    super();
+    this.learningRate = learningRate;
+    this.initialAccumulatorValue = initialAccumulatorValue;
+    this.accumulatedGrads = [];
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      if (this.accumulatedGrads[i] == null) {
+        const trainable = false;
+        this.accumulatedGrads[i] = {
+          originalName: `${name}/accumulator`,
+          variable: tidy(() => fill(value.shape, this.initialAccumulatorValue).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedGrad = this.accumulatedGrads[i].variable;
+      tidy(() => {
+        const newAccumulatedGrad = add2(accumulatedGrad, square(gradient));
+        accumulatedGrad.assign(newAccumulatedGrad);
+        const newValue = add2(mul(div(gradient, sqrt(add2(newAccumulatedGrad, ENGINE.backend.epsilon()))), -this.learningRate), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedGrads != null) {
+      dispose(this.accumulatedGrads.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    return [await this.saveIterations()].concat(this.accumulatedGrads.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const trainable = false;
+    this.accumulatedGrads = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "initialAccumulatorValue": this.initialAccumulatorValue
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["initialAccumulatorValue"]);
+  }
+};
+AdagradOptimizer.className = "Adagrad";
+registerClass(AdagradOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/adam_optimizer.js
+var AdamOptimizer = class extends Optimizer {
+  constructor(learningRate, beta1, beta2, epsilon3 = null) {
+    super();
+    this.learningRate = learningRate;
+    this.beta1 = beta1;
+    this.beta2 = beta2;
+    this.epsilon = epsilon3;
+    this.accumulatedFirstMoment = [];
+    this.accumulatedSecondMoment = [];
+    tidy(() => {
+      this.accBeta1 = scalar(beta1).variable();
+      this.accBeta2 = scalar(beta2).variable();
+    });
+    if (epsilon3 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);
+    tidy(() => {
+      const oneMinusAccBeta1 = sub(1, this.accBeta1);
+      const oneMinusAccBeta2 = sub(1, this.accBeta2);
+      varNames.forEach((name, i) => {
+        const value = ENGINE.registeredVariables[name];
+        const trainable = false;
+        if (this.accumulatedFirstMoment[i] == null) {
+          this.accumulatedFirstMoment[i] = {
+            originalName: `${name}/m`,
+            variable: tidy(() => zerosLike(value).variable(trainable))
+          };
+        }
+        if (this.accumulatedSecondMoment[i] == null) {
+          this.accumulatedSecondMoment[i] = {
+            originalName: `${name}/v`,
+            variable: tidy(() => zerosLike(value).variable(trainable))
+          };
+        }
+        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+        if (gradient == null) {
+          return;
+        }
+        const firstMoment = this.accumulatedFirstMoment[i].variable;
+        const secondMoment = this.accumulatedSecondMoment[i].variable;
+        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));
+        const newSecondMoment = add2(mul(secondMoment, this.beta2), mul(square(gradient), 1 - this.beta2));
+        const biasCorrectedFirstMoment = div(newFirstMoment, oneMinusAccBeta1);
+        const biasCorrectedSecondMoment = div(newSecondMoment, oneMinusAccBeta2);
+        firstMoment.assign(newFirstMoment);
+        secondMoment.assign(newSecondMoment);
+        const newValue = add2(mul(div(biasCorrectedFirstMoment, add2(sqrt(biasCorrectedSecondMoment), this.epsilon)), -this.learningRate), value);
+        value.assign(newValue);
+      });
+      this.accBeta1.assign(mul(this.accBeta1, this.beta1));
+      this.accBeta2.assign(mul(this.accBeta2, this.beta2));
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.accBeta1.dispose();
+    this.accBeta2.dispose();
+    if (this.accumulatedFirstMoment != null) {
+      dispose(this.accumulatedFirstMoment.map((v) => v.variable));
+    }
+    if (this.accumulatedSecondMoment != null) {
+      dispose(this.accumulatedSecondMoment.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedFirstMoment, ...this.accumulatedSecondMoment];
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    tidy(() => {
+      this.accBeta1.assign(pow(this.beta1, this.iterations_ + 1));
+      this.accBeta2.assign(pow(this.beta2, this.iterations_ + 1));
+    });
+    const variableCount = weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedFirstMoment = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedSecondMoment = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "beta1": this.beta1,
+      "beta2": this.beta2,
+      "epsilon": this.epsilon
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["beta1"], config["beta2"], config["epsilon"]);
+  }
+};
+AdamOptimizer.className = "Adam";
+registerClass(AdamOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/adamax_optimizer.js
+var AdamaxOptimizer = class extends Optimizer {
+  constructor(learningRate, beta1, beta2, epsilon3 = null, decay = 0) {
+    super();
+    this.learningRate = learningRate;
+    this.beta1 = beta1;
+    this.beta2 = beta2;
+    this.epsilon = epsilon3;
+    this.decay = decay;
+    this.accumulatedFirstMoment = [];
+    this.accumulatedWeightedInfNorm = [];
+    tidy(() => {
+      this.iteration = scalar(0).variable();
+      this.accBeta1 = scalar(beta1).variable();
+    });
+    if (epsilon3 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    tidy(() => {
+      const oneMinusAccBeta1 = sub(1, this.accBeta1);
+      const lr = div(-this.learningRate, add2(mul(this.iteration, this.decay), 1));
+      variableNames.forEach((name, i) => {
+        const value = ENGINE.registeredVariables[name];
+        const trainable = false;
+        if (this.accumulatedFirstMoment[i] == null) {
+          this.accumulatedFirstMoment[i] = {
+            originalName: `${name}/m`,
+            variable: zerosLike(value).variable(trainable)
+          };
+        }
+        if (this.accumulatedWeightedInfNorm[i] == null) {
+          this.accumulatedWeightedInfNorm[i] = {
+            originalName: `${name}/v`,
+            variable: zerosLike(value).variable(trainable)
+          };
+        }
+        const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+        if (gradient == null) {
+          return;
+        }
+        const firstMoment = this.accumulatedFirstMoment[i].variable;
+        const weightedInfNorm = this.accumulatedWeightedInfNorm[i].variable;
+        const newFirstMoment = add2(mul(firstMoment, this.beta1), mul(gradient, 1 - this.beta1));
+        const ut0 = mul(weightedInfNorm, this.beta2);
+        const ut1 = abs(gradient);
+        const newWeightedInfNorm = maximum(ut0, ut1);
+        firstMoment.assign(newFirstMoment);
+        weightedInfNorm.assign(newWeightedInfNorm);
+        const newValue = add2(mul(div(lr, oneMinusAccBeta1), div(newFirstMoment, add2(newWeightedInfNorm, this.epsilon))), value);
+        value.assign(newValue);
+      });
+      this.iteration.assign(add2(this.iteration, 1));
+      this.accBeta1.assign(mul(this.accBeta1, this.beta1));
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.accBeta1.dispose();
+    this.iteration.dispose();
+    if (this.accumulatedFirstMoment != null) {
+      dispose(this.accumulatedFirstMoment.map((v) => v.variable));
+    }
+    if (this.accumulatedWeightedInfNorm != null) {
+      dispose(this.accumulatedWeightedInfNorm.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    throw new Error("getWeights() is not implemented for Adamax yet.");
+  }
+  async setWeights(weightValues) {
+    throw new Error("setWeights() is not implemented for Adamax yet.");
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "beta1": this.beta1,
+      "beta2": this.beta2,
+      "epsilon": this.epsilon,
+      "decay": this.decay
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["beta1"], config["beta2"], config["epsilon"], config["decay"]);
+  }
+};
+AdamaxOptimizer.className = "Adamax";
+registerClass(AdamaxOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/sgd_optimizer.js
+var SGDOptimizer = class extends Optimizer {
+  constructor(learningRate) {
+    super();
+    this.learningRate = learningRate;
+    this.setLearningRate(learningRate);
+  }
+  applyGradients(variableGradients) {
+    const varNames = Array.isArray(variableGradients) ? variableGradients.map((v) => v.name) : Object.keys(variableGradients);
+    varNames.forEach((name, i) => {
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const value = ENGINE.registeredVariables[name];
+      tidy(() => {
+        const newValue = add2(mul(this.c, gradient), value);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  setLearningRate(learningRate) {
+    this.learningRate = learningRate;
+    if (this.c != null) {
+      this.c.dispose();
+    }
+    this.c = keep(scalar(-learningRate));
+  }
+  dispose() {
+    this.c.dispose();
+  }
+  async getWeights() {
+    return [await this.saveIterations()];
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    if (weightValues.length !== 0) {
+      throw new Error("SGD optimizer does not have settable weights.");
+    }
+  }
+  getConfig() {
+    return { "learningRate": this.learningRate };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"]);
+  }
+};
+SGDOptimizer.className = "SGD";
+registerClass(SGDOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/momentum_optimizer.js
+var MomentumOptimizer = class extends SGDOptimizer {
+  constructor(learningRate, momentum, useNesterov = false) {
+    super(learningRate);
+    this.learningRate = learningRate;
+    this.momentum = momentum;
+    this.useNesterov = useNesterov;
+    this.accumulations = [];
+    this.m = scalar(this.momentum);
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      if (this.accumulations[i] == null) {
+        const trainable = false;
+        this.accumulations[i] = {
+          originalName: `${name}/momentum`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const accumulation = this.accumulations[i].variable;
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      tidy(() => {
+        let newValue;
+        const newAccumulation = add2(mul(this.m, accumulation), gradient);
+        if (this.useNesterov) {
+          newValue = add2(mul(this.c, add2(gradient, mul(newAccumulation, this.m))), value);
+        } else {
+          newValue = add2(mul(this.c, newAccumulation), value);
+        }
+        accumulation.assign(newAccumulation);
+        value.assign(newValue);
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    this.m.dispose();
+    if (this.accumulations != null) {
+      dispose(this.accumulations.map((v) => v.variable));
+    }
+  }
+  setMomentum(momentum) {
+    this.momentum = momentum;
+  }
+  async getWeights() {
+    return [await this.saveIterations()].concat(this.accumulations.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const trainable = false;
+    this.accumulations = weightValues.map((v) => ({ originalName: v.name, variable: v.tensor.variable(trainable) }));
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "momentum": this.momentum,
+      "useNesterov": this.useNesterov
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["momentum"], config["useNesterov"]);
+  }
+};
+MomentumOptimizer.className = "Momentum";
+registerClass(MomentumOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/rmsprop_optimizer.js
+var RMSPropOptimizer = class extends Optimizer {
+  constructor(learningRate, decay = 0.9, momentum = 0, epsilon3 = null, centered = false) {
+    super();
+    this.learningRate = learningRate;
+    this.decay = decay;
+    this.momentum = momentum;
+    this.epsilon = epsilon3;
+    this.accumulatedMeanSquares = [];
+    this.accumulatedMoments = [];
+    this.accumulatedMeanGrads = [];
+    this.centered = centered;
+    if (epsilon3 == null) {
+      this.epsilon = ENGINE.backend.epsilon();
+    }
+    if (learningRate == null) {
+      throw new Error(`learningRate for RMSPropOptimizer must be defined.`);
+    }
+  }
+  applyGradients(variableGradients) {
+    const variableNames = Array.isArray(variableGradients) ? variableGradients.map((item) => item.name) : Object.keys(variableGradients);
+    variableNames.forEach((name, i) => {
+      const value = ENGINE.registeredVariables[name];
+      const trainable = false;
+      if (this.accumulatedMeanSquares[i] == null) {
+        this.accumulatedMeanSquares[i] = {
+          originalName: `${name}/rms`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedMoments[i] == null) {
+        this.accumulatedMoments[i] = {
+          originalName: `${name}/momentum`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      if (this.accumulatedMeanGrads[i] == null && this.centered) {
+        this.accumulatedMeanGrads[i] = {
+          originalName: `${name}/mg`,
+          variable: tidy(() => zerosLike(value).variable(trainable))
+        };
+      }
+      const gradient = Array.isArray(variableGradients) ? variableGradients[i].tensor : variableGradients[name];
+      if (gradient == null) {
+        return;
+      }
+      const accumulatedMeanSquare = this.accumulatedMeanSquares[i].variable;
+      const accumulatedMoments = this.accumulatedMoments[i].variable;
+      tidy(() => {
+        const newAccumulatedMeanSquare = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));
+        if (this.centered) {
+          const accumulatedMeanGrad = this.accumulatedMeanGrads[i].variable;
+          const newAccumulatedMeanGrad = add2(mul(accumulatedMeanGrad, this.decay), mul(gradient, 1 - this.decay));
+          const gradContribution = div(mul(gradient, this.learningRate), sqrt(sub(newAccumulatedMeanSquare, add2(square(newAccumulatedMeanGrad), this.epsilon))));
+          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), gradContribution);
+          accumulatedMeanSquare.assign(newAccumulatedMeanSquare);
+          accumulatedMeanGrad.assign(newAccumulatedMeanGrad);
+          accumulatedMoments.assign(newAccumulatedMoments);
+          const newValue = sub(value, newAccumulatedMoments);
+          value.assign(newValue);
+        } else {
+          const newAccumulatedMeanSquare2 = add2(mul(accumulatedMeanSquare, this.decay), mul(square(gradient), 1 - this.decay));
+          const newAccumulatedMoments = add2(mul(accumulatedMoments, this.momentum), div(mul(gradient, this.learningRate), sqrt(add2(newAccumulatedMeanSquare2, this.epsilon))));
+          accumulatedMeanSquare.assign(newAccumulatedMeanSquare2);
+          accumulatedMoments.assign(newAccumulatedMoments);
+          const newValue = sub(value, newAccumulatedMoments);
+          value.assign(newValue);
+        }
+      });
+    });
+    this.incrementIterations();
+  }
+  dispose() {
+    if (this.accumulatedMeanSquares != null) {
+      dispose(this.accumulatedMeanSquares.map((v) => v.variable));
+    }
+    if (this.accumulatedMeanGrads != null && this.centered) {
+      dispose(this.accumulatedMeanGrads.map((v) => v.variable));
+    }
+    if (this.accumulatedMoments != null) {
+      dispose(this.accumulatedMoments.map((v) => v.variable));
+    }
+  }
+  async getWeights() {
+    const variables = [...this.accumulatedMeanSquares, ...this.accumulatedMoments];
+    if (this.centered) {
+      variables.push(...this.accumulatedMeanGrads);
+    }
+    return [await this.saveIterations()].concat(variables.map((v) => ({ name: v.originalName, tensor: v.variable })));
+  }
+  async setWeights(weightValues) {
+    weightValues = await this.extractIterations(weightValues);
+    const variableCount = this.centered ? weightValues.length / 3 : weightValues.length / 2;
+    const trainable = false;
+    this.accumulatedMeanSquares = weightValues.slice(0, variableCount).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    this.accumulatedMoments = weightValues.slice(variableCount, variableCount * 2).map((v) => ({
+      originalName: v.name,
+      variable: v.tensor.variable(trainable)
+    }));
+    if (this.centered) {
+      this.accumulatedMeanGrads = weightValues.slice(variableCount * 2, variableCount * 3).map((v) => ({
+        originalName: v.name,
+        variable: v.tensor.variable(trainable)
+      }));
+    }
+  }
+  getConfig() {
+    return {
+      "learningRate": this.learningRate,
+      "decay": this.decay,
+      "momentum": this.momentum,
+      "epsilon": this.epsilon,
+      "centered": this.centered
+    };
+  }
+  static fromConfig(cls, config) {
+    return new cls(config["learningRate"], config["decay"], config["momentum"], config["epsilon"], config["centered"]);
+  }
+};
+RMSPropOptimizer.className = "RMSProp";
+registerClass(RMSPropOptimizer);
+
+// node_modules/@tensorflow/tfjs-core/dist/optimizers/optimizer_constructors.js
+var OptimizerConstructors = class {
+  static sgd(learningRate) {
+    return new SGDOptimizer(learningRate);
+  }
+  static momentum(learningRate, momentum, useNesterov = false) {
+    return new MomentumOptimizer(learningRate, momentum, useNesterov);
+  }
+  static rmsprop(learningRate, decay = 0.9, momentum = 0, epsilon3 = null, centered = false) {
+    return new RMSPropOptimizer(learningRate, decay, momentum, epsilon3, centered);
+  }
+  static adam(learningRate = 1e-3, beta1 = 0.9, beta2 = 0.999, epsilon3 = null) {
+    return new AdamOptimizer(learningRate, beta1, beta2, epsilon3);
+  }
+  static adadelta(learningRate = 1e-3, rho = 0.95, epsilon3 = null) {
+    return new AdadeltaOptimizer(learningRate, rho, epsilon3);
+  }
+  static adamax(learningRate = 2e-3, beta1 = 0.9, beta2 = 0.999, epsilon3 = null, decay = 0) {
+    return new AdamaxOptimizer(learningRate, beta1, beta2, epsilon3, decay);
+  }
+  static adagrad(learningRate, initialAccumulatorValue = 0.1) {
+    return new AdagradOptimizer(learningRate, initialAccumulatorValue);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/train.js
+var train = {
+  sgd: OptimizerConstructors.sgd,
+  momentum: OptimizerConstructors.momentum,
+  adadelta: OptimizerConstructors.adadelta,
+  adagrad: OptimizerConstructors.adagrad,
+  rmsprop: OptimizerConstructors.rmsprop,
+  adamax: OptimizerConstructors.adamax,
+  adam: OptimizerConstructors.adam
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/browser_util.js
+var delayCallback = (() => {
+  if (typeof requestAnimationFrame !== "undefined") {
+    return requestAnimationFrame;
+  } else if (typeof setImmediate !== "undefined") {
+    return setImmediate;
+  }
+  return (f) => f();
+})();
+function nextFrame() {
+  return new Promise((resolve) => delayCallback(() => resolve()));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/backend_util.js
+var backend_util_exports = {};
+__export(backend_util_exports, {
+  ERF_A1: () => ERF_A1,
+  ERF_A2: () => ERF_A2,
+  ERF_A3: () => ERF_A3,
+  ERF_A4: () => ERF_A4,
+  ERF_A5: () => ERF_A5,
+  ERF_P: () => ERF_P,
+  PARALLELIZE_THRESHOLD: () => PARALLELIZE_THRESHOLD,
+  RowPartitionType: () => RowPartitionType,
+  SELU_SCALE: () => SELU_SCALE,
+  SELU_SCALEALPHA: () => SELU_SCALEALPHA,
+  applyActivation: () => applyActivation,
+  assertAndGetBroadcastShape: () => assertAndGetBroadcastShape,
+  assertAxesAreInnerMostDims: () => assertAxesAreInnerMostDims,
+  assertParamsConsistent: () => assertParamsConsistent,
+  assignToTypedArray: () => assignToTypedArray,
+  axesAreInnerMostDims: () => axesAreInnerMostDims,
+  calculateShapes: () => calculateShapes,
+  checkEinsumDimSizes: () => checkEinsumDimSizes,
+  checkPadOnDimRoundingMode: () => checkPadOnDimRoundingMode,
+  combineLocations: () => combineLocations,
+  combineRaggedTensorToTensorShapes: () => combineRaggedTensorToTensorShapes,
+  complexWithEvenIndex: () => complexWithEvenIndex,
+  complexWithOddIndex: () => complexWithOddIndex,
+  computeConv2DInfo: () => computeConv2DInfo,
+  computeConv3DInfo: () => computeConv3DInfo,
+  computeDefaultPad: () => computeDefaultPad,
+  computeDilation2DInfo: () => computeDilation2DInfo,
+  computeOptimalWindowSize: () => computeOptimalWindowSize,
+  computeOutAndReduceShapes: () => computeOutAndReduceShapes,
+  computeOutShape: () => computeOutShape2,
+  computePool2DInfo: () => computePool2DInfo,
+  computePool3DInfo: () => computePool3DInfo,
+  convertConv2DDataFormat: () => convertConv2DDataFormat,
+  decodeEinsumEquation: () => decodeEinsumEquation,
+  eitherStridesOrDilationsAreOne: () => eitherStridesOrDilationsAreOne,
+  expandShapeToKeepDim: () => expandShapeToKeepDim,
+  exponent: () => exponent,
+  exponents: () => exponents,
+  fromStringArrayToUint8: () => fromStringArrayToUint8,
+  fromUint8ToStringArray: () => fromUint8ToStringArray,
+  getAxesPermutation: () => getAxesPermutation,
+  getBroadcastDims: () => getBroadcastDims,
+  getComplexWithIndex: () => getComplexWithIndex,
+  getEinsumComputePath: () => getEinsumComputePath,
+  getEinsumPermutation: () => getEinsumPermutation,
+  getFusedBiasGradient: () => getFusedBiasGradient,
+  getFusedDyActivation: () => getFusedDyActivation,
+  getImageCenter: () => getImageCenter,
+  getInnerMostAxes: () => getInnerMostAxes,
+  getPermuted: () => getPermuted,
+  getRaggedRank: () => getRaggedRank,
+  getReductionAxes: () => getReductionAxes,
+  getReshaped: () => getReshaped,
+  getReshapedPermuted: () => getReshapedPermuted,
+  getRowPartitionTypesHelper: () => getRowPartitionTypesHelper,
+  getSliceBeginCoords: () => getSliceBeginCoords,
+  getSliceSize: () => getSliceSize,
+  getSparseFillEmptyRowsIndicesDenseShapeMismatch: () => getSparseFillEmptyRowsIndicesDenseShapeMismatch,
+  getSparseFillEmptyRowsNegativeIndexErrorMessage: () => getSparseFillEmptyRowsNegativeIndexErrorMessage,
+  getSparseFillEmptyRowsOutOfRangeIndexErrorMessage: () => getSparseFillEmptyRowsOutOfRangeIndexErrorMessage,
+  getSparseReshapeEmptyTensorZeroOutputDimErrorMessage: () => getSparseReshapeEmptyTensorZeroOutputDimErrorMessage,
+  getSparseReshapeInputOutputMismatchErrorMessage: () => getSparseReshapeInputOutputMismatchErrorMessage,
+  getSparseReshapeInputOutputMultipleErrorMessage: () => getSparseReshapeInputOutputMultipleErrorMessage,
+  getSparseReshapeMultipleNegativeOneOutputDimErrorMessage: () => getSparseReshapeMultipleNegativeOneOutputDimErrorMessage,
+  getSparseReshapeNegativeOutputDimErrorMessage: () => getSparseReshapeNegativeOutputDimErrorMessage,
+  getSparseSegmentReductionIndicesOutOfRangeErrorMessage: () => getSparseSegmentReductionIndicesOutOfRangeErrorMessage,
+  getSparseSegmentReductionNegativeSegmentIdsErrorMessage: () => getSparseSegmentReductionNegativeSegmentIdsErrorMessage,
+  getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage: () => getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage,
+  getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage: () => getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage,
+  getUndoAxesPermutation: () => getUndoAxesPermutation,
+  isIdentityPermutation: () => isIdentityPermutation,
+  log: () => log,
+  mergeRealAndImagArrays: () => mergeRealAndImagArrays,
+  prepareAndValidate: () => prepareAndValidate,
+  prepareSplitSize: () => prepareSplitSize,
+  segment_util: () => segment_util_exports,
+  shouldFuse: () => shouldFuse,
+  slice_util: () => slice_util_exports,
+  splitRealAndImagArrays: () => splitRealAndImagArrays,
+  tupleValuesAreOne: () => tupleValuesAreOne,
+  upcastType: () => upcastType,
+  validateDefaultValueShape: () => validateDefaultValueShape,
+  validateInput: () => validateInput,
+  validateUpdateShape: () => validateUpdateShape,
+  warn: () => warn
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/concat_util.js
+function assertParamsConsistent(shapes, axis) {
+  const rank = shapes[0].length;
+  shapes.forEach((shape, i) => {
+    assert(shape.length === rank, () => `Error in concat${rank}D: rank of tensors[${i}] must be the same as the rank of the rest (${rank})`);
+  });
+  assert(axis >= 0 && axis < rank, () => `Error in concat${rank}D: axis must be between 0 and ${rank - 1}.`);
+  const firstShape = shapes[0];
+  shapes.forEach((shape, i) => {
+    for (let r = 0; r < rank; r++) {
+      assert(r === axis || shape[r] === firstShape[r], () => `Error in concat${rank}D: Shape of tensors[${i}] (${shape}) does not match the shape of the rest (${firstShape}) along the non-concatenated axis ${i}.`);
+    }
+  });
+}
+function computeOutShape2(shapes, axis) {
+  const outputShape = shapes[0].slice();
+  for (let i = 1; i < shapes.length; i++) {
+    outputShape[axis] += shapes[i][axis];
+  }
+  return outputShape;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ragged_to_dense_util.js
+var RowPartitionType;
+(function(RowPartitionType3) {
+  RowPartitionType3[RowPartitionType3["FIRST_DIM_SIZE"] = 0] = "FIRST_DIM_SIZE";
+  RowPartitionType3[RowPartitionType3["VALUE_ROWIDS"] = 1] = "VALUE_ROWIDS";
+  RowPartitionType3[RowPartitionType3["ROW_LENGTHS"] = 2] = "ROW_LENGTHS";
+  RowPartitionType3[RowPartitionType3["ROW_SPLITS"] = 3] = "ROW_SPLITS";
+  RowPartitionType3[RowPartitionType3["ROW_LIMITS"] = 4] = "ROW_LIMITS";
+  RowPartitionType3[RowPartitionType3["ROW_STARTS"] = 5] = "ROW_STARTS";
+})(RowPartitionType || (RowPartitionType = {}));
+function combineRaggedTensorToTensorShapes(raggedRank, shape, valueShape) {
+  let outputShape = new Array();
+  if (valueShape == null && shape == null) {
+    return outputShape;
+  }
+  if (shape == null) {
+    while (outputShape.length < raggedRank + valueShape.length) {
+      outputShape.push(-1);
+    }
+  } else {
+    outputShape = shape.slice();
+  }
+  if (valueShape == null) {
+    return outputShape;
+  }
+  if (raggedRank + valueShape.length !== outputShape.length) {
+    throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.rank = ${raggedRank + valueShape.length}, but shape.rank = ${outputShape.length}`);
+  }
+  for (let i = 1; i < valueShape.length; ++i) {
+    const valueDim = valueShape[i];
+    const outputShapeDimIndex = outputShape[outputShape.length - valueShape.length + i];
+    const outputShapeDim = outputShape[outputShapeDimIndex];
+    if (valueDim >= 0) {
+      if (outputShapeDim >= 0) {
+        if (outputShapeDim !== valueDim) {
+          throw new Error(`rt input.shape and shape=${shape} are incompatible: rt input.shape[${i + raggedRank}] = ${valueDim} but shape[${i + raggedRank}] = ${outputShapeDim}`);
+        }
+      } else {
+        outputShape[outputShapeDimIndex] = valueDim;
+      }
+    }
+  }
+  return outputShape;
+}
+function getRowPartitionTypesHelper(rowPartitionTypeStrings) {
+  const stringToType = {
+    "FIRST_DIM_SIZE": RowPartitionType.FIRST_DIM_SIZE,
+    "VALUE_ROWIDS": RowPartitionType.VALUE_ROWIDS,
+    "ROW_LENGTHS": RowPartitionType.ROW_LENGTHS,
+    "ROW_SPLITS": RowPartitionType.ROW_SPLITS,
+    "ROW_LIMITS": RowPartitionType.ROW_LIMITS,
+    "ROW_STARTS": RowPartitionType.ROW_STARTS
+  };
+  const result = [];
+  for (const typeStr of rowPartitionTypeStrings) {
+    if (typeStr in stringToType) {
+      result.push(stringToType[typeStr]);
+    } else {
+      break;
+    }
+  }
+  return result;
+}
+function getRaggedRank(rowPartitionTypes) {
+  if (rowPartitionTypes.length === 0) {
+    return 0;
+  }
+  if (rowPartitionTypes[0] === RowPartitionType.FIRST_DIM_SIZE) {
+    return rowPartitionTypes.length - 1;
+  }
+  return rowPartitionTypes.length;
+}
+function validateDefaultValueShape(defaultValueShape, valueShape) {
+  if (defaultValueShape == null || valueShape == null) {
+    return;
+  }
+  const defaultNDims = defaultValueShape.length;
+  const valuesNDims = valueShape.length;
+  if (defaultNDims >= valuesNDims) {
+    throw new Error(`defaultValue.shape=${defaultValueShape} and ragged tensor flatValues.shape=${valueShape}, are incompatible: defaultValue.rank = ${defaultNDims} must be less than ragged tensor input flatValues.rank = ${valuesNDims})`);
+  }
+  for (let i = 0; i < Math.min(defaultNDims, valuesNDims - 1); ++i) {
+    const defaultDim = defaultValueShape[i];
+    const valueDim = valueShape[i + 1];
+    if (defaultDim >= 0 && valueDim >= 0 && defaultDim !== 1 && defaultDim !== valueDim) {
+      throw new Error(`defaultValue.shape=${defaultValueShape}, and ragged tensor input flatValues.shape=${valueShape} are incompatible: defaultValue.shape[${i - defaultValueShape.length}] = ${defaultDim} but ragged tensor input.flatValues.shape[${i - defaultValueShape.length}] = ${valueDim}`);
+    }
+  }
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/reduce_util.js
+var PARALLELIZE_THRESHOLD = 30;
+function computeOptimalWindowSize(inSize) {
+  if (inSize <= PARALLELIZE_THRESHOLD) {
+    return inSize;
+  }
+  return nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/rotate_util.js
+function getImageCenter(center, imageHeight, imageWidth) {
+  const centerX = imageWidth * (typeof center === "number" ? center : center[0]);
+  const centerY = imageHeight * (typeof center === "number" ? center : center[1]);
+  return [centerX, centerY];
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/array_ops_util.js
+function getReshaped(inputShape, blockShape, prod5, batchToSpace = true) {
+  let reshaped = [];
+  if (batchToSpace) {
+    reshaped = reshaped.concat(blockShape.slice(0));
+    reshaped.push(inputShape[0] / prod5);
+    reshaped = reshaped.concat(inputShape.slice(1));
+  } else {
+    reshaped = reshaped.concat(inputShape[0]);
+    const spatialLength = blockShape.length;
+    for (let i = 0; i < spatialLength; ++i) {
+      reshaped = reshaped.concat([inputShape[i + 1] / blockShape[i], blockShape[i]]);
+    }
+    reshaped = reshaped.concat(inputShape.slice(spatialLength + 1));
+  }
+  return reshaped;
+}
+function getPermuted(reshapedRank, blockShapeRank, batchToSpace = true) {
+  const permuted = [];
+  if (batchToSpace) {
+    permuted.push(blockShapeRank);
+    for (let i = blockShapeRank + 1; i < reshapedRank; ++i) {
+      if (i <= 2 * blockShapeRank) {
+        permuted.push(i);
+        permuted.push(i - (blockShapeRank + 1));
+      } else {
+        permuted.push(i);
+      }
+    }
+  } else {
+    const permutedBeforeBatch = [];
+    const permutedAfterBatch = [];
+    for (let i = 1; i < reshapedRank; ++i) {
+      if (i >= blockShapeRank * 2 + 1 || i % 2 === 1) {
+        permutedAfterBatch.push(i);
+      } else {
+        permutedBeforeBatch.push(i);
+      }
+    }
+    permuted.push(...permutedBeforeBatch);
+    permuted.push(0);
+    permuted.push(...permutedAfterBatch);
+  }
+  return permuted;
+}
+function getReshapedPermuted(inputShape, blockShape, prod5, batchToSpace = true) {
+  const reshapedPermuted = [];
+  if (batchToSpace) {
+    reshapedPermuted.push(inputShape[0] / prod5);
+  } else {
+    reshapedPermuted.push(inputShape[0] * prod5);
+  }
+  for (let i = 1; i < inputShape.length; ++i) {
+    if (i <= blockShape.length) {
+      if (batchToSpace) {
+        reshapedPermuted.push(blockShape[i - 1] * inputShape[i]);
+      } else {
+        reshapedPermuted.push(inputShape[i] / blockShape[i - 1]);
+      }
+    } else {
+      reshapedPermuted.push(inputShape[i]);
+    }
+  }
+  return reshapedPermuted;
+}
+function getSliceBeginCoords(crops, blockShape) {
+  const sliceBeginCoords = [0];
+  for (let i = 0; i < blockShape; ++i) {
+    sliceBeginCoords.push(crops[i][0]);
+  }
+  return sliceBeginCoords;
+}
+function getSliceSize(uncroppedShape, crops, blockShape) {
+  const sliceSize = uncroppedShape.slice(0, 1);
+  for (let i = 0; i < blockShape; ++i) {
+    sliceSize.push(uncroppedShape[i + 1] - crops[i][0] - crops[i][1]);
+  }
+  return sliceSize;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/selu_util.js
+var SELU_SCALEALPHA = 1.7580993408473768;
+var SELU_SCALE = 1.0507009873554805;
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/erf_util.js
+var ERF_P = 0.3275911;
+var ERF_A1 = 0.254829592;
+var ERF_A2 = -0.284496736;
+var ERF_A3 = 1.421413741;
+var ERF_A4 = -1.453152027;
+var ERF_A5 = 1.061405429;
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/complex_util.js
+function mergeRealAndImagArrays(real4, imag4) {
+  if (real4.length !== imag4.length) {
+    throw new Error(`Cannot merge real and imag arrays of different lengths. real:${real4.length}, imag: ${imag4.length}.`);
+  }
+  const result = new Float32Array(real4.length * 2);
+  for (let i = 0; i < result.length; i += 2) {
+    result[i] = real4[i / 2];
+    result[i + 1] = imag4[i / 2];
+  }
+  return result;
+}
+function splitRealAndImagArrays(complex4) {
+  const real4 = new Float32Array(complex4.length / 2);
+  const imag4 = new Float32Array(complex4.length / 2);
+  for (let i = 0; i < complex4.length; i += 2) {
+    real4[i / 2] = complex4[i];
+    imag4[i / 2] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function complexWithEvenIndex(complex4) {
+  const len = Math.ceil(complex4.length / 4);
+  const real4 = new Float32Array(len);
+  const imag4 = new Float32Array(len);
+  for (let i = 0; i < complex4.length; i += 4) {
+    real4[Math.floor(i / 4)] = complex4[i];
+    imag4[Math.floor(i / 4)] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function complexWithOddIndex(complex4) {
+  const len = Math.floor(complex4.length / 4);
+  const real4 = new Float32Array(len);
+  const imag4 = new Float32Array(len);
+  for (let i = 2; i < complex4.length; i += 4) {
+    real4[Math.floor(i / 4)] = complex4[i];
+    imag4[Math.floor(i / 4)] = complex4[i + 1];
+  }
+  return { real: real4, imag: imag4 };
+}
+function getComplexWithIndex(complex4, index) {
+  const real4 = complex4[index * 2];
+  const imag4 = complex4[index * 2 + 1];
+  return { real: real4, imag: imag4 };
+}
+function assignToTypedArray(data, real4, imag4, index) {
+  data[index * 2] = real4;
+  data[index * 2 + 1] = imag4;
+}
+function exponents(n, inverse) {
+  const real4 = new Float32Array(n / 2);
+  const imag4 = new Float32Array(n / 2);
+  for (let i = 0; i < Math.ceil(n / 2); i++) {
+    const x = (inverse ? 2 : -2) * Math.PI * (i / n);
+    real4[i] = Math.cos(x);
+    imag4[i] = Math.sin(x);
+  }
+  return { real: real4, imag: imag4 };
+}
+function exponent(k, n, inverse) {
+  const x = (inverse ? 2 : -2) * Math.PI * (k / n);
+  const real4 = Math.cos(x);
+  const imag4 = Math.sin(x);
+  return { real: real4, imag: imag4 };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/einsum_util.js
+var ARROW = "->";
+var ARROW_REGEX = /->/g;
+var COMMA = ",";
+var ELLIPSIS = "...";
+function decodeEinsumEquation(equation, numTensors) {
+  equation = equation.replace(/\s/g, "");
+  const numArrows = (equation.length - equation.replace(ARROW_REGEX, "").length) / ARROW.length;
+  if (numArrows < 1) {
+    throw new Error("Equations without an arrow are not supported.");
+  } else if (numArrows > 1) {
+    throw new Error(`Equation must contain exactly one arrow ("${ARROW}").`);
+  }
+  const [inputString, outputString] = equation.split(ARROW);
+  assert(inputString.indexOf(ELLIPSIS) === -1, () => `The ellipsis notation ("${ELLIPSIS}") is not supported yet.`);
+  const inputTerms = inputString.split(COMMA);
+  const numInputs = inputTerms.length;
+  if (numTensors !== numInputs) {
+    throw new Error(`Expected ${numInputs} input tensors, received ${numTensors}`);
+  }
+  if (numInputs > 2) {
+    throw new Error("Support for more than 2 input tensors is not implemented yet.");
+  }
+  const allDims = [];
+  for (let i = 0; i < outputString.length; ++i) {
+    const dimName = outputString[i];
+    if (!inputTerms.some((inputTerm) => inputTerm.indexOf(dimName) !== -1)) {
+      throw new Error(`Output subscripts contain the label ${dimName} not present in the input subscripts.`);
+    }
+    if (allDims.indexOf(dimName) === -1) {
+      allDims.push(dimName);
+    }
+  }
+  for (let i = 0; i < inputString.length; ++i) {
+    const dimName = inputString[i];
+    if (allDims.indexOf(dimName) === -1 && dimName !== COMMA) {
+      allDims.push(dimName);
+    }
+  }
+  const idDims = new Array(inputTerms.length);
+  for (let i = 0; i < numInputs; ++i) {
+    if (new Set(inputTerms[i].split("")).size !== inputTerms[i].length) {
+      throw new Error(`Found duplicate axes in input component ${inputTerms[i]}. Support for duplicate axes in input is not implemented yet.`);
+    }
+    idDims[i] = [];
+    for (let j = 0; j < inputTerms[i].length; ++j) {
+      idDims[i].push(allDims.indexOf(inputTerms[i][j]));
+    }
+  }
+  const numDims = allDims.length;
+  const numOutDims = outputString.length;
+  const summedDims = [];
+  for (let i = numOutDims; i < numDims; ++i) {
+    summedDims.push(i);
+  }
+  return { allDims, summedDims, idDims };
+}
+function getEinsumPermutation(nDims, idDims) {
+  let permutationIndices = new Array(nDims);
+  permutationIndices.fill(-1);
+  for (let i = 0; i < idDims.length; ++i) {
+    permutationIndices[idDims[i]] = i;
+  }
+  const expandDims6 = [];
+  for (let i = 0; i < nDims; ++i) {
+    if (permutationIndices[i] === -1) {
+      expandDims6.push(i);
+    }
+  }
+  permutationIndices = permutationIndices.filter((d) => d !== -1);
+  return { permutationIndices, expandDims: expandDims6 };
+}
+function checkEinsumDimSizes(nDims, idDims, tensors) {
+  const dimSizes = new Array(nDims);
+  for (let i = 0; i < tensors.length; ++i) {
+    const shape = tensors[i].shape;
+    for (let j = 0; j < idDims[i].length; ++j) {
+      if (dimSizes[idDims[i][j]] === void 0) {
+        dimSizes[idDims[i][j]] = shape[j];
+      } else {
+        assert(dimSizes[idDims[i][j]] === shape[j], () => `Expected dimension ${dimSizes[idDims[i][j]]} at axis ${j} of input shaped ${JSON.stringify(shape)}, but got dimension ${shape[j]}`);
+      }
+    }
+  }
+}
+function getEinsumComputePath(summedDims, idDims) {
+  const path = summedDims;
+  const steps = [];
+  let nSteps = 0;
+  if (summedDims.length === 0) {
+    path.push(-1);
+  }
+  nSteps = summedDims.length + 1;
+  for (let i = 0; i < nSteps; ++i) {
+    steps.push([]);
+  }
+  const computedTermIndices = [];
+  for (let i = 0; i < path.length; ++i) {
+    const summedDim = path[i];
+    const termIndices = findTermsWithDim(idDims, summedDim);
+    for (const termIndex of termIndices) {
+      if (computedTermIndices.indexOf(termIndex) === -1) {
+        steps[i].push(termIndex);
+        computedTermIndices.push(termIndex);
+      }
+    }
+  }
+  return { path, steps };
+}
+function isIdentityPermutation(perm) {
+  return perm.every((dim, index) => dim === index);
+}
+function findTermsWithDim(idDims, dim) {
+  const termIndices = [];
+  for (let i = 0; i < idDims.length; ++i) {
+    if (idDims[i].length === 0 || idDims[i].indexOf(dim) !== -1 || dim === -1) {
+      termIndices.push(i);
+    }
+  }
+  return termIndices;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/split_util.js
+function prepareSplitSize(x, numOrSizeSplits, axis = 0) {
+  let splitSizes = [];
+  if (typeof numOrSizeSplits === "number") {
+    assert(x.shape[axis] % numOrSizeSplits === 0, () => "Number of splits must evenly divide the axis.");
+    splitSizes = new Array(numOrSizeSplits).fill(x.shape[axis] / numOrSizeSplits);
+  } else {
+    const numOfNegs = numOrSizeSplits.reduce((count2, value) => {
+      if (value === -1) {
+        count2 += 1;
+      }
+      return count2;
+    }, 0);
+    assert(numOfNegs <= 1, () => "There should be only one negative value in split array.");
+    const negIndex = numOrSizeSplits.indexOf(-1);
+    if (negIndex !== -1) {
+      const total = numOrSizeSplits.reduce((a, b) => b > 0 ? a + b : a);
+      numOrSizeSplits[negIndex] = x.shape[axis] - total;
+    }
+    assert(x.shape[axis] === numOrSizeSplits.reduce((a, b) => a + b), () => "The sum of sizes must match the size of the axis dimension.");
+    splitSizes = numOrSizeSplits;
+  }
+  return splitSizes;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_fill_empty_rows_util.js
+function getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesLength) {
+  return `Received SparseTensor with denseShape[0] = 0 but
+  indices.shape[0] = ${indicesLength}`;
+}
+function getSparseFillEmptyRowsNegativeIndexErrorMessage(index, value) {
+  return `indices(${index}, 0) is invalid: ${value} < 0`;
+}
+function getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(index, value, limit) {
+  return `indices(${index}, 0) is invalid: ${value} >= ${limit}`;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_reshape_util.js
+function getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(dim1, dim2) {
+  return `only one output dimension may be -1, not both ${dim1} and ${dim2}`;
+}
+function getSparseReshapeNegativeOutputDimErrorMessage(dim, value) {
+  return `size ${dim} must be non-negative, not ${value}`;
+}
+function getSparseReshapeEmptyTensorZeroOutputDimErrorMessage() {
+  return "reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero";
+}
+function getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape) {
+  const inputSize = sizeFromShape(inputShape);
+  const outputSize = sizeFromShape(outputShape);
+  return `Input to reshape is a SparseTensor with ${inputSize}
+  dense values, but the requested shape requires a multiple of ${outputSize}. inputShape=${inputShape} outputShape= ${outputShape}`;
+}
+function getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape) {
+  const inputSize = sizeFromShape(inputShape);
+  const outputSize = sizeFromShape(outputShape);
+  return `Input to reshape is a tensor with ${inputSize} dense values, but the requested shape has ${outputSize}. inputShape=${inputShape} outputShape=${outputShape}`;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/sparse/sparse_segment_reduction_util.js
+function getSparseSegmentReductionNegativeSegmentIdsErrorMessage() {
+  return `segment ids must be >= 0`;
+}
+function getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage() {
+  return `segment ids are not increasing`;
+}
+function getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(segmentId, outputRows) {
+  return `Segment id ${segmentId} out of range [0, ${outputRows}), possibly because segmentIds input is not sorted.`;
+}
+function getSparseSegmentReductionIndicesOutOfRangeErrorMessage(index, indexValue, inputRows) {
+  return `Bad: indices[${index}] == ${indexValue} out of range [0, ${inputRows})`;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/segment_util.js
+var segment_util_exports = {};
+__export(segment_util_exports, {
+  collectGatherOpShapeInfo: () => collectGatherOpShapeInfo,
+  computeOutShape: () => computeOutShape3,
+  segOpComputeOptimalWindowSize: () => segOpComputeOptimalWindowSize
+});
+function segOpComputeOptimalWindowSize(inSize, numSegments) {
+  let done = false;
+  let res;
+  if (inSize <= PARALLELIZE_THRESHOLD) {
+    res = inSize;
+    done = true;
+  } else {
+    res = nearestDivisor(inSize, Math.floor(Math.sqrt(inSize)));
+  }
+  while (!done) {
+    if (res > numSegments || res === inSize) {
+      done = true;
+    } else {
+      res = nearestDivisor(inSize, res + 1);
+    }
+  }
+  return res;
+}
+function computeOutShape3(aShape, axis, numSegments) {
+  const outShape = [];
+  const rank = aShape.length;
+  for (let dim = 0; dim < rank; dim++) {
+    if (dim !== axis) {
+      outShape.push(aShape[dim]);
+    } else {
+      outShape.push(numSegments);
+    }
+  }
+  return outShape;
+}
+function collectGatherOpShapeInfo(x, indices, axis, batchDims) {
+  const indicesRank = indices.shape.length;
+  const xRank = x.shape.length;
+  if (batchDims !== 0) {
+    if (batchDims < -indicesRank || batchDims > indicesRank) {
+      throw new Error(`Expect batchDims in the range of [-${indicesRank}, ${indicesRank}], but got ${batchDims}`);
+    }
+  }
+  if (batchDims < 0) {
+    batchDims += indicesRank;
+  }
+  if (batchDims > xRank) {
+    throw new Error(`batchDims (${batchDims}) must be less than rank(x) (
+    ${xRank}).`);
+  }
+  if (axis < batchDims) {
+    throw new Error(`batchDims (${batchDims}) must be less than or equal to axis (${axis}).`);
+  }
+  for (let i = 0; i < batchDims; ++i) {
+    if (x.shape[i] !== indices.shape[i]) {
+      throw new Error(`x.shape[${i}]: ${x.shape[i]} should be equal to indices.shape[${i}]: ${indices.shape[i]}.`);
+    }
+  }
+  const dimSize = x.shape[axis];
+  const outputShape = [];
+  let batchSize = 1;
+  let outerSize = 1;
+  let sliceSize = 1;
+  for (let i = 0; i < batchDims; ++i) {
+    outputShape.push(x.shape[i]);
+    batchSize *= x.shape[i];
+  }
+  for (let i = batchDims; i < axis; i++) {
+    outputShape.push(x.shape[i]);
+    outerSize *= x.shape[i];
+  }
+  for (let i = batchDims; i < indicesRank; i++) {
+    outputShape.push(indices.shape[i]);
+  }
+  for (let i = axis + 1; i < xRank; i++) {
+    outputShape.push(x.shape[i]);
+    sliceSize *= x.shape[i];
+  }
+  return { batchSize, sliceSize, outerSize, dimSize, outputShape };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/backend_util.js
+function fromUint8ToStringArray(vals) {
+  try {
+    return vals.map((val) => decodeString(val));
+  } catch (err) {
+    throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${err}`);
+  }
+}
+function fromStringArrayToUint8(strings) {
+  return strings.map((s) => encodeString(s));
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/backends/kernel_impls.js
+var kernel_impls_exports = {};
+__export(kernel_impls_exports, {
+  nonMaxSuppressionV3Impl: () => nonMaxSuppressionV3Impl,
+  nonMaxSuppressionV4Impl: () => nonMaxSuppressionV4Impl,
+  nonMaxSuppressionV5Impl: () => nonMaxSuppressionV5Impl,
+  whereImpl: () => whereImpl
+});
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Abs_grad.js
+var absGradConfig = {
+  kernelName: Abs,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, step(cast(x, "float32"), -1)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Acos_grad.js
+var acosGradConfig = {
+  kernelName: Acos,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = square(cast(x, "float32"));
+        const b = sqrt(sub(scalar(1), a));
+        return neg(div(dy, b));
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Acosh_grad.js
+var acoshGradConfig = {
+  kernelName: Acosh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = sqrt(sub(square(cast(x, "float32")), 1));
+        return div(dy, a);
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Add_grad.js
+var addGradConfig = {
+  kernelName: Add,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/AddN_grad.js
+var addNGradConfig = {
+  kernelName: AddN,
+  saveAllInputs: true,
+  gradFunc: (dy, saved) => {
+    const ders = {};
+    saved.forEach((_, i) => {
+      ders[i] = () => dy.clone();
+    });
+    return ders;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ArgMax_grad.js
+var argMaxGradConfig = {
+  kernelName: ArgMax,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => zerosLike(x) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ArgMin_grad.js
+var argMinGradConfig = {
+  kernelName: ArgMin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => zerosLike(x) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Asin_grad.js
+var asinGradConfig = {
+  kernelName: Asin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, sqrt(sub(scalar(1), square(cast(x, "float32"))))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Asinh_grad.js
+var asinhGradConfig = {
+  kernelName: Asinh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const a = sqrt(add2(scalar(1), square(cast(x, "float32"))));
+        return div(dy, a);
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Atan2_grad.js
+var atan2GradConfig = {
+  kernelName: Atan2,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const d = add2(square(a), square(b));
+      let res = mul(dy, div(b, d));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      const d = add2(square(a), square(b));
+      let res = neg(mul(dy, div(a, d)));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Atan_grad.js
+var atanGradConfig = {
+  kernelName: Atan,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, add2(square(cast(x, "float32")), 1)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Atanh_grad.js
+var atanhGradConfig = {
+  kernelName: Atanh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, sub(scalar(1), square(cast(x, "float32")))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_3d_grad.js
+function avgPool3dGrad_(dy, input2, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "avgPool3dGrad");
+  const $input = convertToTensor(input2, "input", "avgPool3dGrad");
+  let dy5D = $dy;
+  let input5D = $input;
+  let reshapedTo5D = false;
+  if ($input.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);
+    input5D = reshape($input, [
+      1,
+      $input.shape[0],
+      $input.shape[1],
+      $input.shape[2],
+      $input.shape[3]
+    ]);
+  }
+  assert(dy5D.rank === 5, () => `Error in avgPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);
+  assert(input5D.rank === 5, () => `Error in avgPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);
+  checkPadOnDimRoundingMode("avgPool3dGrad", pad3, dimRoundingMode);
+  const inputs = { dy: dy5D, input: input5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(AvgPool3DGrad, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var avgPool3dGrad = op({ avgPool3dGrad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/AvgPool3D_grad.js
+var avgPool3DGradConfig = {
+  kernelName: AvgPool3D,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+    return {
+      x: () => avgPool3dGrad(dy, x, filterSize, strides, pad3, dimRoundingMode)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/avg_pool_grad.js
+function avgPoolGrad_(dy, input2, filterSize, strides, pad3) {
+  const $dy = convertToTensor(dy, "dy", "avgPoolGrad");
+  const $input = convertToTensor(input2, "input", "avgPoolGrad");
+  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);
+  let input4D = $input;
+  let dy4D = $dy;
+  let reshapedTo4D = false;
+  if ($input.rank === 3) {
+    reshapedTo4D = true;
+    input4D = reshape($input, [1, $input.shape[0], $input.shape[1], $input.shape[2]]);
+    dy4D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2]]);
+  }
+  assert(dy4D.rank === 4, () => `Error in avgPoolGrad: dy must be rank 4 but got rank ${dy4D.rank}.`);
+  assert(input4D.rank === 4, () => `Error in avgPoolGrad: input must be rank 4 but got rank ${input4D.rank}.`);
+  const inputs = { dy: dy4D, input: input4D };
+  const attrs = { filterSize, strides, pad: pad3 };
+  const res = ENGINE.runKernel(AvgPoolGrad, inputs, attrs);
+  if (reshapedTo4D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3]]);
+  }
+  return res;
+}
+var avgPoolGrad = op({ avgPoolGrad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/AvgPool_grad.js
+var avgPoolGradConfig = {
+  kernelName: AvgPool,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { filterSize, strides, pad: pad3 } = attrs;
+    return { x: () => avgPoolGrad(dy, x, filterSize, strides, pad3) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/BatchMatMul_grad.js
+var batchMatMulGradConfig = {
+  kernelName: BatchMatMul,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved, attrs) => {
+    const [a, b] = saved;
+    const { transposeA, transposeB } = attrs;
+    if (!transposeA && !transposeB) {
+      return {
+        a: () => matMul(dy, b, false, true),
+        b: () => matMul(a, dy, true, false)
+      };
+    } else if (!transposeA && transposeB) {
+      return {
+        a: () => matMul(dy, b, false, false),
+        b: () => matMul(dy, a, true, false)
+      };
+    } else if (transposeA && !transposeB) {
+      return {
+        a: () => matMul(b, dy, false, true),
+        b: () => matMul(a, dy, false, false)
+      };
+    } else {
+      return {
+        a: () => matMul(b, dy, true, true),
+        b: () => matMul(dy, a, true, true)
+      };
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/BatchToSpaceND_grad.js
+var batchToSpaceNDGradConfig = {
+  kernelName: BatchToSpaceND,
+  gradFunc: (dy, saved, attrs) => {
+    const { blockShape, crops } = attrs;
+    return { x: () => spaceToBatchND(dy, blockShape, crops) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/BroadcastTo_grad.js
+var broadcastToGradConfig = {
+  kernelName: BroadcastTo,
+  gradFunc: (dy, saved, attrs) => {
+    const broadCastToAttrs = attrs;
+    const inputShape = broadCastToAttrs.inputShape;
+    const outputShape = broadCastToAttrs.shape;
+    const reps = Array.from(outputShape);
+    for (let i = inputShape.length - 1; i >= 0; i--) {
+      if (inputShape[i] === outputShape[i]) {
+        reps[i] = 1;
+      } else if (inputShape[i] !== 1) {
+        throw new Error(`broadcastTo(): [${inputShape}] cannot be broadcast to [${outputShape}].`);
+      }
+    }
+    const axes = [];
+    for (let i = 0; i < reps.length; i++) {
+      if (reps[i] > 1) {
+        axes.push(i);
+      }
+    }
+    return { x: () => sum2(dy, axes, true) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Cast_grad.js
+var castGradConfig = {
+  kernelName: Cast,
+  gradFunc: (dy) => {
+    return { x: () => dy.clone() };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Ceil_grad.js
+var ceilGradConfig = {
+  kernelName: Ceil,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ClipByValue_grad.js
+var clipByValueGradConfig = {
+  kernelName: ClipByValue,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { clipValueMin, clipValueMax } = attrs;
+    return {
+      x: () => where(logicalAnd(greaterEqual(x, clipValueMin), lessEqual(x, clipValueMax)), dy, zerosLike(dy))
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ComplexAbs_grad.js
+var complexAbsGradConfig = {
+  kernelName: ComplexAbs,
+  inputsToSave: ["x"],
+  gradFunc: absGradConfig.gradFunc
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Concat_grad.js
+var concatGradConfig = {
+  kernelName: Concat,
+  saveAllInputs: true,
+  gradFunc: (dy, saved, attrs) => {
+    const shapes = saved.map((t) => t.shape);
+    const { axis } = attrs;
+    const $axis = parseAxisParam(axis, saved[0].shape)[0];
+    const sizeSplits = shapes.map((s) => s[$axis]);
+    const derTensors = split(dy, sizeSplits, $axis);
+    return derTensors.map((t) => () => t);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv2D_grad.js
+var conv2DGradConfig = {
+  kernelName: Conv2D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x4D, $filter] = saved;
+    const { dilations, strides, pad: pad3, dataFormat } = attrs;
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    return {
+      x: () => conv2DBackpropInput(x4D.shape, dy, $filter, strides, pad3, dataFormat),
+      filter: () => conv2DBackpropFilter(x4D, dy, $filter.shape, strides, pad3, dataFormat)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv2DBackpropInput_grad.js
+var conv2DBackpropInputGradConfig = {
+  kernelName: Conv2DBackpropInput,
+  inputsToSave: ["dy", "filter"],
+  gradFunc: (ddx, saved, attrs) => {
+    const [dy, filter] = saved;
+    const { strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+    return {
+      dy: () => conv2d(ddx, filter, strides, pad3, dataFormat, 1, dimRoundingMode),
+      filter: () => conv2DBackpropFilter(ddx, dy, filter.shape, strides, pad3, dataFormat, dimRoundingMode)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/conv3d_backprop_filter.js
+function conv3DBackpropFilter_(x, dy, filterShape, strides, pad3) {
+  let x5D = x;
+  if (x.rank === 4) {
+    x5D = reshape(x, [1, x.shape[0], x.shape[1], x.shape[2], x.shape[3]]);
+  }
+  let dy5D = dy;
+  if (dy5D.rank === 4) {
+    dy5D = reshape(dy, [1, dy.shape[0], dy.shape[1], dy.shape[2], dy.shape[3]]);
+  }
+  assert(x5D.rank === 5, () => `Error in conv3dDerFilter: input must be rank 5, but got shape ${x5D.shape}.`);
+  assert(dy5D.rank === 5, () => `Error in conv3dDerFilter: dy must be rank 5, but got shape ${dy5D.shape}.`);
+  assert(filterShape.length === 5, () => `Error in conv3dDerFilter: filterShape must be length 5, but got ${filterShape}.`);
+  assert(x5D.shape[4] === filterShape[3], () => `Error in conv3dDerFilter: depth of input ${x5D.shape[4]}) must match input depth in filter (${filterShape[3]}.`);
+  assert(dy5D.shape[4] === filterShape[4], () => `Error in conv3dDerFilter: depth of dy (${dy5D.shape[4]}) must match output depth for filter (${filterShape[4]}).`);
+  const inputs = { x: x5D, dy: dy5D };
+  const attrs = { strides, pad: pad3, filterShape };
+  return ENGINE.runKernel(Conv3DBackpropFilterV2, inputs, attrs);
+}
+var conv3DBackpropFilter = op({ conv3DBackpropFilter_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Conv3D_grad.js
+var conv3DGradConfig = {
+  kernelName: Conv3D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const { dilations, strides, pad: pad3 } = attrs;
+    assert(tupleValuesAreOne(dilations), () => `Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${dilations}'`);
+    const [x5D, $filter] = saved;
+    return {
+      x: () => conv3DBackpropInput(x5D.shape, dy, $filter, strides, pad3),
+      filter: () => conv3DBackpropFilter(x5D, dy, $filter.shape, strides, pad3)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Cos_grad.js
+var cosGradConfig = {
+  kernelName: Cos,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(neg(sin(cast(x, "float32"))), dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Cosh_grad.js
+var coshGradConfig = {
+  kernelName: Cosh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(sinh(cast(x, "float32")), dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Cumsum_grad.js
+var cumsumGradConfig = {
+  kernelName: Cumsum,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis, exclusive, reverse: reverse5 } = attrs;
+    return {
+      x: () => {
+        const permutation = getAxesPermutation([axis], x.rank);
+        let out = cumsum(dy, axis, exclusive, !reverse5);
+        if (permutation != null) {
+          out = transpose(out, permutation);
+        }
+        return out;
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/DepthwiseConv2dNative_grad.js
+var depthwiseConv2dNativeGradConfig = {
+  kernelName: DepthwiseConv2dNative,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const { dilations, strides, pad: pad3, dimRoundingMode } = attrs;
+    const $dilations = dilations == null ? [1, 1] : dilations;
+    assert(tupleValuesAreOne($dilations), () => `Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${$dilations}'`);
+    const [x, filter] = saved;
+    assert(x.rank === 4, () => `Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${x.rank}.`);
+    assert(filter.rank === 4, () => `Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${filter.rank}.`);
+    assert(x.shape[3] === filter.shape[2], () => `Error in gradient of depthwiseConv2d: number of input channels (${x.shape[3]}) must match the inChannels dimension in filter ${filter.shape[2]}.`);
+    assert(eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${strides} and dilations '${$dilations}'.`);
+    checkPadOnDimRoundingMode("depthwiseConv2d", pad3, dimRoundingMode);
+    return {
+      x: () => depthwiseConv2dNativeBackpropInput(x.shape, dy, filter, strides, pad3, $dilations, dimRoundingMode),
+      filter: () => depthwiseConv2dNativeBackpropFilter(x, dy, filter.shape, strides, pad3, $dilations, dimRoundingMode)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Dilation2D_grad.js
+var dilation2dGradConfig = {
+  kernelName: Dilation2D,
+  inputsToSave: ["x", "filter"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, filter] = saved;
+    const inputInputs = { x, filter, dy };
+    const filterInputs = { x, filter, dy };
+    return {
+      x: () => ENGINE.runKernel(Dilation2DBackpropInput, inputInputs, attrs),
+      filter: () => ENGINE.runKernel(Dilation2DBackpropFilter, filterInputs, attrs)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Elu_grad.js
+var eluGradConfig = {
+  kernelName: Elu,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    const inputs = { dy, y };
+    return { x: () => ENGINE.runKernel(EluGrad, inputs) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Erf_grad.js
+var erfGradConfig = {
+  kernelName: Erf,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    const a = mul(exp(neg(square(x))), 2 / Math.sqrt(Math.PI));
+    return { x: () => mul(dy, a) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Exp_grad.js
+var expGradConfig = {
+  kernelName: Exp,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(dy, y) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ExpandDims_grad.js
+var expandDimsGradConfig = {
+  kernelName: ExpandDims,
+  inputsToSave: ["input"],
+  gradFunc: (dy, saved) => {
+    const [input2] = saved;
+    return { input: () => reshape(dy, input2.shape) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Expm1_grad.js
+var expm1GradConfig = {
+  kernelName: Expm1,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, exp(x)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Floor_grad.js
+var floorGradConfig = {
+  kernelName: Floor,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/FloorDiv_grad.js
+var floorDivGradConfig = {
+  kernelName: FloorDiv,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = div(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      let res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = reshape(sum2(res, reduceAxes), b.shape);
+      }
+      const tmp = square(b);
+      return neg(div(res, cast(tmp, "float32")));
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/FusedBatchNorm_grad.js
+var fusedBatchNormGradConfig = {
+  kernelName: FusedBatchNorm,
+  inputsToSave: ["x", "mean", "variance", "scale"],
+  gradFunc: (dy, saved, attrs) => {
+    const { varianceEpsilon } = attrs;
+    const [x, mean4, variance, scale2] = saved;
+    const scaleValue = scale2 == null ? scalar(1) : scale2;
+    const reductionAxes = getReductionAxes(mean4.shape, x.shape);
+    const tileShape = [];
+    if (mean4.rank === 1) {
+      for (let i = 0; i < x.shape.length - 1; ++i) {
+        tileShape.push(x.shape[i]);
+      }
+      tileShape.push(1);
+    }
+    const xMinusMean = sub(x, mean4);
+    const dyTimesScaleValue = mul(dy, scaleValue);
+    const oneOverSqrtVariance = rsqrt(add2(variance, scalar(varianceEpsilon)));
+    const minusHalfRCube = mul(mul(mul(oneOverSqrtVariance, oneOverSqrtVariance), oneOverSqrtVariance), scalar(-0.5));
+    const derX = () => {
+      if (mean4.rank === 1) {
+        return reshape(mul(mul(dy, tile(reshape(oneOverSqrtVariance, [1, 1, 1, mean4.shape[0]]), tileShape)), scaleValue), x.shape);
+      } else {
+        return reshape(mul(mul(dy, oneOverSqrtVariance), scaleValue), x.shape);
+      }
+    };
+    const derMean = () => {
+      let meanDer = mul(mul(oneOverSqrtVariance, scalar(-1)), dyTimesScaleValue);
+      if (mean4.rank === 1) {
+        meanDer = sum2(meanDer, reductionAxes);
+      }
+      return reshape(meanDer, mean4.shape);
+    };
+    const derVariance = () => {
+      let varianceDer = mul(mul(minusHalfRCube, xMinusMean), dyTimesScaleValue);
+      if (mean4.rank === 1) {
+        varianceDer = sum2(varianceDer, reductionAxes);
+      }
+      return reshape(varianceDer, mean4.shape);
+    };
+    const derScale = () => {
+      const xMinusMean2TimesRsqrt = mul(xMinusMean, oneOverSqrtVariance);
+      let scaleDer = mul(dy, xMinusMean2TimesRsqrt);
+      if (mean4.rank === 1) {
+        scaleDer = sum2(scaleDer, reductionAxes);
+      }
+      return reshape(scaleDer, mean4.shape);
+    };
+    const derOffset = () => {
+      let offsetDer = dy;
+      if (mean4.rank === 1) {
+        offsetDer = sum2(offsetDer, reductionAxes);
+      }
+      return reshape(offsetDer, mean4.shape);
+    };
+    return {
+      x: derX,
+      mean: derMean,
+      variance: derVariance,
+      scale: derScale,
+      offset: derOffset
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/GatherV2_grad.js
+var gatherGradConfig = {
+  kernelName: GatherV2,
+  inputsToSave: ["x", "indices"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, indices] = saved;
+    const { axis } = attrs;
+    const parsedAxis = parseAxisParam(axis, x.shape)[0];
+    const derX = () => {
+      const paramsShape = x.shape;
+      const indicesSize = indices.size;
+      const outerShape = paramsShape.slice(0, parsedAxis);
+      const outerDims = outerShape.length;
+      const innerShape = paramsShape.slice(axis, paramsShape.length).slice(1);
+      const innerDims = innerShape.length;
+      const outerAxesIndices = arrayRange(0, outerDims);
+      const innerAxesIndices = arrayRange(outerDims + 1, outerDims + 1 + innerDims);
+      const valuesShape = arrayConcat([outerShape, [indicesSize], innerShape]);
+      const values = reshape(dy, valuesShape);
+      const reshapedIndices = reshape(indices, [indicesSize]);
+      const transposeDims = arrayConcat([[outerDims], outerAxesIndices, innerAxesIndices]);
+      const valuesTranspose = transpose(values, transposeDims);
+      let paramsGrad = unsortedSegmentSum(valuesTranspose, reshapedIndices, x.shape[parsedAxis]);
+      const invertTransposeDims = getUndoAxesPermutation(transposeDims);
+      paramsGrad = transpose(paramsGrad, invertTransposeDims);
+      return paramsGrad;
+    };
+    return { x: derX, indices: () => indices };
+  }
+};
+function arrayRange(start, stop) {
+  const result = [];
+  for (let i = start; i < stop; ++i) {
+    result.push(i);
+  }
+  return result;
+}
+function arrayConcat(arrays) {
+  const result = [];
+  for (let i = 0; i < arrays.length; ++i) {
+    for (let j = 0; j < arrays[i].length; ++j) {
+      result.push(arrays[i][j]);
+    }
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/GreaterEqual_grad.js
+var greaterEqualGradConfig = {
+  kernelName: GreaterEqual,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    return { a: () => zerosLike(a), b: () => zerosLike(b) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Identity_grad.js
+var identityGradConfig = {
+  kernelName: Identity,
+  gradFunc: (dy) => {
+    return { x: () => cast(dy, "float32") };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/IsFinite_grad.js
+var isFiniteGradConfig = {
+  kernelName: IsFinite,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/IsInf_grad.js
+var isInfGradConfig = {
+  kernelName: IsInf,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/IsNan_grad.js
+var isNanGradConfig = {
+  kernelName: IsNan,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/LeakyRelu_grad.js
+var leakyReluGradConfig = {
+  kernelName: LeakyRelu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { alpha } = attrs;
+    const mask = greater(x, 0);
+    return { x: () => where(mask, dy, mul(dy, alpha)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Log1p_grad.js
+var log1pGradConfig = {
+  kernelName: Log1p,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, add2(x, 1)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Log_grad.js
+var logGradConfig = {
+  kernelName: Log,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, cast(x, "float32")) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/LogSoftmax_grad.js
+var logSoftmaxGradConfig = {
+  kernelName: LogSoftmax,
+  inputsToSave: [],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [value] = saved;
+    const { axis } = attrs;
+    return {
+      logits: () => {
+        const keepDims = true;
+        const softmax6 = exp(value);
+        return sub(dy, mul(sum2(dy, axis, keepDims), softmax6));
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/local_response_normalization_backprop.js
+function localResponseNormalizationBackprop_(x, y, dy, depthRadius = 5, bias = 1, alpha = 1, beta = 0.5) {
+  const inputs = { x, y, dy };
+  const attrs = { depthRadius, bias, alpha, beta };
+  return ENGINE.runKernel(LRNGrad, inputs, attrs);
+}
+var localResponseNormalizationBackprop = op({ localResponseNormalizationBackprop_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/LRN_grad.js
+var lrnGradConfig = {
+  kernelName: LRN,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { depthRadius, bias, alpha, beta } = attrs;
+    return {
+      x: () => localResponseNormalizationBackprop(x, y, dy, depthRadius, bias, alpha, beta)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/min_max_grad_util.js
+function gradForMinAndMax(dy, y, xOrig, origAxes) {
+  if (y.rank < xOrig.rank) {
+    y = reshape(y, expandShapeToKeepDim(y.shape, origAxes));
+  }
+  if (dy.rank < xOrig.rank) {
+    dy = reshape(dy, expandShapeToKeepDim(dy.shape, origAxes));
+  }
+  return {
+    x: () => {
+      const dx = mul(dy, cast(equal(xOrig, y), dy.dtype));
+      return dx;
+    }
+  };
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Max_grad.js
+var maxGradConfig = {
+  kernelName: Max,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const maxAttrs = attrs;
+    const { reductionIndices } = maxAttrs;
+    const x = saved[0];
+    const y = saved[1];
+    const origAxes = parseAxisParam(reductionIndices, x.shape);
+    const maxGrad = gradForMinAndMax(dy, y, x, origAxes);
+    return {
+      x: () => {
+        return maxGrad["x"]();
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Maximum_grad.js
+var maximumGradConfig = {
+  kernelName: Maximum,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const derA = () => mul(dy, cast(greaterEqual(a, b), "float32"));
+    const derB = () => mul(dy, cast(less(a, b), "float32"));
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_3d_grad.js
+function maxPool3dGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "maxPool3dGrad");
+  const $input = convertToTensor(input2, "input", "maxPool3dGrad");
+  const $output = convertToTensor(output, "output", "maxPool3dGrad");
+  let dy5D = $dy;
+  let input5D = $input;
+  let output5D = $output;
+  let reshapedTo5D = false;
+  if ($input.rank === 4) {
+    reshapedTo5D = true;
+    dy5D = reshape($dy, [1, $dy.shape[0], $dy.shape[1], $dy.shape[2], $dy.shape[3]]);
+    input5D = reshape($input, [
+      1,
+      $input.shape[0],
+      $input.shape[1],
+      $input.shape[2],
+      $input.shape[3]
+    ]);
+    output5D = reshape($output, [
+      1,
+      $output.shape[0],
+      $output.shape[1],
+      $output.shape[2],
+      $output.shape[3]
+    ]);
+  }
+  assert(dy5D.rank === 5, () => `Error in maxPool3dGrad: dy must be rank 5 but got rank ${dy5D.rank}.`);
+  assert(input5D.rank === 5, () => `Error in maxPool3dGrad: input must be rank 5 but got rank ${input5D.rank}.`);
+  assert(output5D.rank === 5, () => `Error in maxPool3dGrad: output must be rank 5 but got rank ${output5D.rank}.`);
+  checkPadOnDimRoundingMode("maxPool3dGrad", pad3, dimRoundingMode);
+  const inputs = { dy: dy5D, input: input5D, output: output5D };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  const res = ENGINE.runKernel(MaxPool3DGrad, inputs, attrs);
+  if (reshapedTo5D) {
+    return reshape(res, [res.shape[1], res.shape[2], res.shape[3], res.shape[4]]);
+  }
+  return res;
+}
+var maxPool3dGrad = op({ maxPool3dGrad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/MaxPool3D_grad.js
+var maxPool3DGradConfig = {
+  kernelName: MaxPool3D,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+    return {
+      x: () => maxPool3dGrad(dy, x, y, filterSize, strides, pad3, dimRoundingMode)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/max_pool_grad.js
+function maxPoolGrad_(dy, input2, output, filterSize, strides, pad3, dimRoundingMode) {
+  const $dy = convertToTensor(dy, "dy", "maxPoolGrad");
+  const $input = convertToTensor(input2, "input", "maxPoolGrad");
+  const $output = convertToTensor(output, "output", "maxPoolGrad");
+  assert($input.rank === $dy.rank, () => `Rank of input (${$input.rank}) does not match rank of dy (${$dy.rank})`);
+  assert($dy.rank === 4, () => `Error in maxPoolGrad: dy must be rank 4 but got rank ${$dy.rank}.`);
+  assert($input.rank === 4, () => `Error in maxPoolGrad: input must be rank 4 but got rank ${$input.rank}.`);
+  checkPadOnDimRoundingMode("maxPoolGrad", pad3, dimRoundingMode);
+  const inputs = { dy: $dy, input: $input, output: $output };
+  const attrs = { filterSize, strides, pad: pad3, dimRoundingMode };
+  return ENGINE.runKernel(MaxPoolGrad, inputs, attrs);
+}
+var maxPoolGrad = op({ maxPoolGrad_ });
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/MaxPool_grad.js
+var maxPoolGradConfig = {
+  kernelName: MaxPool,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [x, y] = saved;
+    const { filterSize, strides, pad: pad3 } = attrs;
+    return {
+      x: () => maxPoolGrad(dy, x, y, filterSize, strides, pad3)
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Mean_grad.js
+var meanGradConfig = {
+  kernelName: Mean,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis } = attrs;
+    const axes = parseAxisParam(axis, x.shape);
+    const shapes = computeOutAndReduceShapes(x.shape, axes);
+    const reduceShape = shapes[1];
+    const reduceSize = sizeFromShape(reduceShape);
+    const derX = () => {
+      const expandedDyShape = x.shape.slice();
+      axes.forEach((axis2) => {
+        expandedDyShape[axis2] = 1;
+      });
+      const expandedDy = reshape(dy, expandedDyShape);
+      const res = div(mul(expandedDy, ones2(x.shape, "float32")), reduceSize);
+      return res;
+    };
+    return { x: derX };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Min_grad.js
+var minGradConfig = {
+  kernelName: Min,
+  inputsToSave: ["x"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const minAttrs = attrs;
+    const { axis } = minAttrs;
+    const [x, y] = saved;
+    const origAxes = parseAxisParam(axis, x.shape);
+    const minGrad = gradForMinAndMax(dy, y, x, origAxes);
+    return {
+      x: () => {
+        return minGrad["x"]();
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Minimum_grad.js
+var minimumGradConfig = {
+  kernelName: Minimum,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const derA = () => mul(dy, cast(lessEqual(a, b), "float32"));
+    const derB = () => mul(dy, cast(greater(a, b), "float32"));
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/MirrorPad_grad.js
+var mirrorPadGradConfig = {
+  kernelName: MirrorPad,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const x = saved[0];
+    const { paddings } = attrs;
+    const begin = paddings.map((p2) => p2[0]);
+    return { x: () => slice(dy, begin, x.shape) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Mod_grad.js
+var modGradConfig = {
+  kernelName: Mod,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(dy, reduceAxes), a.shape);
+      }
+      return dy;
+    };
+    const derB = () => {
+      const res = mul(dy, neg(floor(div(a, b))));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), b.shape);
+      }
+      return res;
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Multiply_grad.js
+var multiplyGradConfig = {
+  kernelName: Multiply,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = mul(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      const res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), b.shape);
+      }
+      return res;
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Neg_grad.js
+var negGradConfig = {
+  kernelName: Neg,
+  gradFunc: (dy) => {
+    return { x: () => neg(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/OneHot_grad.js
+var oneHotGradConfig = {
+  kernelName: OneHot,
+  inputsToSave: ["indices"],
+  gradFunc: (dy, saved) => {
+    const indices = saved[0];
+    return { indices: () => zeros(indices.shape, "float32") };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/OnesLike_grad.js
+var onesLikeGradConfig = {
+  kernelName: OnesLike,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Pack_grad.js
+var packGradConfig = {
+  kernelName: Pack,
+  saveAllInputs: true,
+  gradFunc: (dy, saved, attrs) => {
+    const { axis } = attrs;
+    const derTensors = unstack(dy, axis);
+    return derTensors.map((t) => () => t);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/PadV2_grad.js
+var padV2GradConfig = {
+  kernelName: PadV2,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const x = saved[0];
+    const { paddings } = attrs;
+    const begin = paddings.map((p2) => p2[0]);
+    return { x: () => slice(dy, begin, x.shape) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Pow_grad.js
+var powGradConfig = {
+  kernelName: Pow,
+  inputsToSave: ["a", "b"],
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [a, b, y] = saved;
+    const base = a;
+    const exp4 = b;
+    const outShape = assertAndGetBroadcastShape(base.shape, exp4.shape);
+    const derBase = () => {
+      const expFloat = cast(exp4, "float32");
+      let res = mul(dy, mul(expFloat, pow(base, sub(expFloat, scalar(1)))));
+      const reduceAxes = getReductionAxes(base.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, base.shape);
+    };
+    const derExp = () => {
+      const condition = greater(base, 0);
+      const logBase = where(condition, log2(base), zerosLike(base));
+      let res = mul(dy, mul(y, logBase));
+      const reduceAxes = getReductionAxes(exp4.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, exp4.shape);
+    };
+    return { a: derBase, b: derExp };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Prelu_grad.js
+var preluGradConfig = {
+  kernelName: Prelu,
+  inputsToSave: ["x", "alpha"],
+  gradFunc: (dy, saved) => {
+    const [x, alpha] = saved;
+    const mask = greater(x, 0);
+    return {
+      x: () => where(mask, dy, mul(dy, alpha)),
+      alpha: () => {
+        let res = where(mask, zerosLike(dy), mul(dy, x));
+        const reduceAxes = getReductionAxes(alpha.shape, dy.shape);
+        if (reduceAxes.length > 0) {
+          res = sum2(res, reduceAxes);
+        }
+        return reshape(res, alpha.shape);
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Prod_grad.js
+function prodGradFn_(x, dy, axis) {
+  const expandedYShape = x.shape.slice();
+  expandedYShape[axis] = 1;
+  const expandedDy = reshape(dy, expandedYShape);
+  const xCumProd = cumprod(x, axis, true, false);
+  const xCumRevProd = cumprod(x, axis, true, true);
+  const dx = mul(xCumProd, xCumRevProd);
+  return mul(expandedDy, dx);
+}
+function prodsGradFn_(x, dy, axis) {
+  const xRank = x.shape.length;
+  const finalProdAxis = xRank - axis.length;
+  const xPermutation = backend_util_exports.getAxesPermutation(axis, xRank);
+  let permutedX = x;
+  if (xPermutation != null) {
+    permutedX = transpose(x, xPermutation);
+  }
+  const newShape = permutedX.shape.slice();
+  const removedShape = newShape.splice(xRank - axis.length, axis.length);
+  const endPartShape = removedShape.reduce((p2, c) => p2 * c, 1);
+  newShape.push(endPartShape);
+  const reshapedPermutedX = permutedX.reshape(newShape);
+  let prodGrad = prodGradFn_(reshapedPermutedX, dy, finalProdAxis);
+  prodGrad = prodGrad.reshape(permutedX.shape);
+  if (xPermutation != null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(xPermutation);
+    prodGrad = transpose(prodGrad, undoPermutation);
+  }
+  return prodGrad;
+}
+var prodGradConfig = {
+  kernelName: Prod,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { axis } = attrs;
+    let axisArr = [];
+    if (axis === void 0 || axis === null) {
+      axisArr = x.shape.map((_, i) => i);
+    } else if (typeof axis === "number") {
+      axisArr = [axis];
+    } else {
+      axisArr = axis;
+    }
+    return { x: () => prodsGradFn_(x, dy, axisArr) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/RealDiv_grad.js
+var divGradConfig = {
+  kernelName: RealDiv,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      const res = div(dy, cast(b, "float32"));
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        return reshape(sum2(res, reduceAxes), a.shape);
+      }
+      return res;
+    };
+    const derB = () => {
+      let res = mul(dy, cast(a, "float32"));
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = reshape(sum2(res, reduceAxes), b.shape);
+      }
+      const tmp = square(b);
+      return neg(div(res, cast(tmp, "float32")));
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Reciprocal_grad.js
+var reciprocalGradConfig = {
+  kernelName: Reciprocal,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, neg(square(x))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Relu6_grad.js
+var relu6GradConfig = {
+  kernelName: Relu6,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    const mask = mul(lessEqual(x, 6), step(x));
+    return { x: () => mul(dy, cast(mask, "float32")) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Relu_grad.js
+var reluGradConfig = {
+  kernelName: Relu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, cast(step(x), "float32")) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Reshape_grad.js
+var reshapeGradConfig = {
+  kernelName: Reshape,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => reshape(dy, x.shape) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ResizeBilinear_grad.js
+var resizeBilinearGradConfig = {
+  kernelName: ResizeBilinear,
+  inputsToSave: ["images"],
+  gradFunc: (dy, saved, attrs) => {
+    const [images] = saved;
+    const inputs = { dy, images };
+    const imagesDer = () => ENGINE.runKernel(ResizeBilinearGrad, inputs, attrs);
+    return { images: imagesDer };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ResizeNearestNeighbor_grad.js
+var resizeNearestNeighborGradConfig = {
+  kernelName: ResizeNearestNeighbor,
+  inputsToSave: ["images"],
+  gradFunc: (dy, saved, attrs) => {
+    const [images] = saved;
+    const inputs = { dy, images };
+    const imagesDer = () => ENGINE.runKernel(ResizeNearestNeighborGrad, inputs, attrs);
+    return { images: imagesDer };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Reverse_grad.js
+var reverseGradConfig = {
+  kernelName: Reverse,
+  gradFunc: (dy, saved, attrs) => {
+    const { dims } = attrs;
+    const axes = parseAxisParam(dims, dy.shape);
+    return { x: () => reverse(dy, axes) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Round_grad.js
+var roundGradConfig = {
+  kernelName: Round,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Rsqrt_grad.js
+var rsqrtGradConfig = {
+  kernelName: Rsqrt,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => neg(div(dy, mul(pow(x, 1.5), 2))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Select_grad.js
+var selectGradConfig = {
+  kernelName: Select,
+  inputsToSave: ["condition"],
+  gradFunc: (dy, saved) => {
+    const [condition] = saved;
+    return {
+      condition: () => cast(zerosLike(condition), "float32"),
+      t: () => mul(dy, cast(condition, dy.dtype)),
+      e: () => mul(dy, cast(logicalNot(condition), dy.dtype))
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Selu_grad.js
+var seluGradConfig = {
+  kernelName: Selu,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return {
+      x: () => {
+        const mask = greater(x, scalar(0));
+        const scaleAlpha2 = scalar(SELU_SCALEALPHA);
+        const scale2 = scalar(SELU_SCALE);
+        const greaterThanZeroDer = mul(dy, scale2);
+        const lessEqualZeroDer = mul(mul(dy, scaleAlpha2), exp(cast(x, "float32")));
+        return where(mask, greaterThanZeroDer, lessEqualZeroDer);
+      }
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sigmoid_grad.js
+var sigmoidGradConfig = {
+  kernelName: Sigmoid,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(dy, mul(y, sub(scalar(1), y))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sign_grad.js
+var signGradConfig = {
+  kernelName: Sign,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sin_grad.js
+var sinGradConfig = {
+  kernelName: Sin,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(cos(cast(x, "float32")), dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sinh_grad.js
+var sinhGradConfig = {
+  kernelName: Sinh,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(cosh(cast(x, "float32")), dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Slice_grad.js
+var sliceGradConfig = {
+  kernelName: Slice,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { begin, size } = attrs;
+    const inputShape = x.shape;
+    const [begin_, size_] = parseSliceParams(x, begin, size);
+    const paddings = [];
+    for (let i = 0; i < dy.rank; i++) {
+      paddings.push([begin_[i], inputShape[i] - begin_[i] - size_[i]]);
+    }
+    return { x: () => pad(dy, paddings) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Softmax_grad.js
+var softmaxGradConfig = {
+  kernelName: Softmax,
+  outputsToSave: [true],
+  gradFunc: (dy, saved, attrs) => {
+    const [y] = saved;
+    const { dim } = attrs;
+    const keepDims = true;
+    const dyTimesY = mul(dy, y);
+    return {
+      logits: () => sub(dyTimesY, mul(sum2(dyTimesY, [dim], keepDims), y))
+    };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Softplus_grad.js
+var softplusGradConfig = {
+  kernelName: Softplus,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, sigmoid(x)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/SpaceToBatchND_grad.js
+var spaceToBatchNDGradConfig = {
+  kernelName: SpaceToBatchND,
+  gradFunc: (dy, saved, attrs) => {
+    const { blockShape, paddings } = attrs;
+    return { x: () => batchToSpaceND(dy, blockShape, paddings) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/SplitV_grad.js
+var splitVGradConfig = {
+  kernelName: SplitV,
+  gradFunc: (dy, saved, attrs) => {
+    const { axis } = attrs;
+    return { x: () => concat(dy, axis) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sqrt_grad.js
+var sqrtGradConfig = {
+  kernelName: Sqrt,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, mul(sqrt(cast(x, "float32")), 2)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Square_grad.js
+var squareGradConfig = {
+  kernelName: Square,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => mul(dy, mul(cast(x, "float32"), 2)) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/SquaredDifference_grad.js
+var squaredDifferenceGradConfig = {
+  kernelName: SquaredDifference,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const two = scalar(2);
+    const derA = () => mul(dy, mul(two, sub(a, b)));
+    const derB = () => mul(dy, mul(two, sub(b, a)));
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Step_grad.js
+var stepGradConfig = {
+  kernelName: Step,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sub_grad.js
+var subGradConfig = {
+  kernelName: Sub,
+  inputsToSave: ["a", "b"],
+  gradFunc: (dy, saved) => {
+    const [a, b] = saved;
+    const outShape = assertAndGetBroadcastShape(a.shape, b.shape);
+    const derA = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(a.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(res, a.shape);
+    };
+    const derB = () => {
+      let res = dy;
+      const reduceAxes = getReductionAxes(b.shape, outShape);
+      if (reduceAxes.length > 0) {
+        res = sum2(res, reduceAxes);
+      }
+      return reshape(neg(res), b.shape);
+    };
+    return { a: derA, b: derB };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Sum_grad.js
+var sumGradConfig = {
+  kernelName: Sum,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const expandedDyShape = x.shape.slice();
+    const { axis } = attrs;
+    const axes = parseAxisParam(axis, x.shape);
+    axes.forEach((axis2) => {
+      expandedDyShape[axis2] = 1;
+    });
+    const expandedDy = reshape(dy, expandedDyShape);
+    const derX = mul(expandedDy, ones2(x.shape, "float32"));
+    return { x: () => derX };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Tan_grad.js
+var tanGradConfig = {
+  kernelName: Tan,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved) => {
+    const [x] = saved;
+    return { x: () => div(dy, square(cos(x))) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Tanh_grad.js
+var tanhGradConfig = {
+  kernelName: Tanh,
+  outputsToSave: [true],
+  gradFunc: (dy, saved) => {
+    const [y] = saved;
+    return { x: () => mul(sub(scalar(1), square(y)), dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Tile_grad.js
+var tileGradConfig = {
+  kernelName: Tile,
+  inputsToSave: ["x"],
+  gradFunc: (dy, saved, attrs) => {
+    const [x] = saved;
+    const { reps } = attrs;
+    const derX = () => {
+      let xGrad = zerosLike(x);
+      if (x.rank === 1) {
+        for (let i = 0; i < reps[0]; ++i) {
+          xGrad = add2(xGrad, slice(dy, [i * x.shape[0]], [x.shape[0]]));
+        }
+      } else if (x.rank === 2) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1]], [
+              x.shape[0],
+              x.shape[1]
+            ]));
+          }
+        }
+      } else if (x.rank === 3) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            for (let k = 0; k < reps[2]; ++k) {
+              xGrad = add2(xGrad, slice(dy, [i * x.shape[0], j * x.shape[1], k * x.shape[2]], [x.shape[0], x.shape[1], x.shape[2]]));
+            }
+          }
+        }
+      } else if (x.rank === 4) {
+        for (let i = 0; i < reps[0]; ++i) {
+          for (let j = 0; j < reps[1]; ++j) {
+            for (let k = 0; k < reps[2]; ++k) {
+              for (let l = 0; l < reps[3]; ++l) {
+                xGrad = add2(xGrad, slice(dy, [
+                  i * x.shape[0],
+                  j * x.shape[1],
+                  k * x.shape[2],
+                  l * x.shape[3]
+                ], [x.shape[0], x.shape[1], x.shape[2], x.shape[3]]));
+              }
+            }
+          }
+        }
+      } else {
+        throw new Error(`Gradient for tile operation is not implemented for rank-${x.rank} tensors yet.`);
+      }
+      return xGrad;
+    };
+    return { x: derX };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Transpose_grad.js
+var transposeGradConfig = {
+  kernelName: Transpose,
+  gradFunc: (dy, saved, attrs) => {
+    const transposeAttrs = attrs;
+    const { perm } = transposeAttrs;
+    const undoPerm = getUndoAxesPermutation(perm);
+    return { x: () => transpose(dy, undoPerm) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/Unpack_grad.js
+var unpackGradConfig = {
+  kernelName: Unpack,
+  gradFunc: (dy, saved, attrs) => {
+    const unpackAttrs = attrs;
+    const { axis } = unpackAttrs;
+    return { value: () => stack(dy, axis) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/UnsortedSegmentSum_grad.js
+var unsortedSegmentSumGradConfig = {
+  kernelName: UnsortedSegmentSum,
+  inputsToSave: ["segmentIds"],
+  gradFunc: (dy, saved) => {
+    const [segmentIds] = saved;
+    const derX = () => {
+      return gatherDropNegatives(dy, segmentIds);
+    };
+    return { x: derX };
+  }
+};
+function gatherDropNegatives(x, indices) {
+  const zeroClippedIndices = maximum(indices, zerosLike(indices));
+  const gathered = gather(x, zeroClippedIndices);
+  let isPositive = greaterEqual(indices, scalar(0, "int32"));
+  const numIters = gathered.rank - isPositive.rank;
+  for (let i = 0; i < numIters; ++i) {
+    isPositive = expandDims(isPositive, i + 1);
+  }
+  isPositive = logicalAnd(isPositive, ones2(gathered.shape, "bool"));
+  const zeroSlice = zerosLike(gathered);
+  return where(isPositive, gathered, zeroSlice);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/gradients/ZerosLike_grad.js
+var zerosLikeGradConfig = {
+  kernelName: ZerosLike,
+  gradFunc: (dy) => {
+    return { x: () => zerosLike(dy) };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/register_all_gradients.js
+var gradConfigs = [
+  absGradConfig,
+  acosGradConfig,
+  acoshGradConfig,
+  addGradConfig,
+  addNGradConfig,
+  argMaxGradConfig,
+  argMinGradConfig,
+  asinGradConfig,
+  asinhGradConfig,
+  atan2GradConfig,
+  atanGradConfig,
+  atanhGradConfig,
+  avgPool3DGradConfig,
+  avgPoolGradConfig,
+  batchMatMulGradConfig,
+  batchToSpaceNDGradConfig,
+  broadcastToGradConfig,
+  castGradConfig,
+  ceilGradConfig,
+  clipByValueGradConfig,
+  complexAbsGradConfig,
+  concatGradConfig,
+  conv2DBackpropInputGradConfig,
+  conv2DGradConfig,
+  conv3DGradConfig,
+  cosGradConfig,
+  coshGradConfig,
+  cumsumGradConfig,
+  depthwiseConv2dNativeGradConfig,
+  dilation2dGradConfig,
+  divGradConfig,
+  eluGradConfig,
+  erfGradConfig,
+  expGradConfig,
+  expandDimsGradConfig,
+  expm1GradConfig,
+  floorDivGradConfig,
+  floorGradConfig,
+  fusedBatchNormGradConfig,
+  gatherGradConfig,
+  greaterEqualGradConfig,
+  identityGradConfig,
+  isFiniteGradConfig,
+  isInfGradConfig,
+  isNanGradConfig,
+  leakyReluGradConfig,
+  log1pGradConfig,
+  logGradConfig,
+  logSoftmaxGradConfig,
+  lrnGradConfig,
+  maxGradConfig,
+  maxGradConfig,
+  maximumGradConfig,
+  maxPool3DGradConfig,
+  maxPoolGradConfig,
+  meanGradConfig,
+  minGradConfig,
+  minimumGradConfig,
+  mirrorPadGradConfig,
+  modGradConfig,
+  multiplyGradConfig,
+  negGradConfig,
+  oneHotGradConfig,
+  onesLikeGradConfig,
+  packGradConfig,
+  padV2GradConfig,
+  padV2GradConfig,
+  powGradConfig,
+  preluGradConfig,
+  prodGradConfig,
+  reciprocalGradConfig,
+  relu6GradConfig,
+  reluGradConfig,
+  reshapeGradConfig,
+  resizeBilinearGradConfig,
+  resizeNearestNeighborGradConfig,
+  reverseGradConfig,
+  roundGradConfig,
+  rsqrtGradConfig,
+  selectGradConfig,
+  seluGradConfig,
+  sigmoidGradConfig,
+  signGradConfig,
+  sinGradConfig,
+  sinhGradConfig,
+  sliceGradConfig,
+  softmaxGradConfig,
+  softplusGradConfig,
+  spaceToBatchNDGradConfig,
+  spaceToBatchNDGradConfig,
+  splitVGradConfig,
+  splitVGradConfig,
+  sqrtGradConfig,
+  squaredDifferenceGradConfig,
+  squareGradConfig,
+  stepGradConfig,
+  subGradConfig,
+  sumGradConfig,
+  tanGradConfig,
+  tanhGradConfig,
+  tileGradConfig,
+  transposeGradConfig,
+  unpackGradConfig,
+  unsortedSegmentSumGradConfig,
+  zerosLikeGradConfig
+];
+for (const gradientConfig of gradConfigs) {
+  registerGradient(gradientConfig);
+}
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/abs.js
+getGlobalTensorClass().prototype.abs = function() {
+  this.throwIfDisposed();
+  return abs(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/acos.js
+getGlobalTensorClass().prototype.acos = function() {
+  this.throwIfDisposed();
+  return acos(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/acosh.js
+getGlobalTensorClass().prototype.acosh = function() {
+  this.throwIfDisposed();
+  return acosh(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/add.js
+getGlobalTensorClass().prototype.add = function(b) {
+  this.throwIfDisposed();
+  return add2(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/all.js
+getGlobalTensorClass().prototype.all = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return all(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/any.js
+getGlobalTensorClass().prototype.any = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return any(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/arg_max.js
+getGlobalTensorClass().prototype.argMax = function(axis) {
+  this.throwIfDisposed();
+  return argMax(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/arg_min.js
+getGlobalTensorClass().prototype.argMin = function(axis) {
+  this.throwIfDisposed();
+  return argMin(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as_scalar.js
+getGlobalTensorClass().prototype.asScalar = function() {
+  this.throwIfDisposed();
+  assert(this.size === 1, () => "The array must have only 1 element.");
+  return reshape(this, []);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as_type.js
+getGlobalTensorClass().prototype.asType = function(dtype) {
+  this.throwIfDisposed();
+  return cast(this, dtype);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as1d.js
+getGlobalTensorClass().prototype.as1D = function() {
+  this.throwIfDisposed();
+  return reshape(this, [this.size]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as2d.js
+getGlobalTensorClass().prototype.as2D = function(rows, columns) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as3d.js
+getGlobalTensorClass().prototype.as3D = function(rows, columns, depth) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as4d.js
+getGlobalTensorClass().prototype.as4D = function(rows, columns, depth, depth2) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth, depth2]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/as5d.js
+getGlobalTensorClass().prototype.as5D = function(rows, columns, depth, depth2, depth3) {
+  this.throwIfDisposed();
+  return reshape(this, [rows, columns, depth, depth2, depth3]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/asin.js
+getGlobalTensorClass().prototype.asin = function() {
+  this.throwIfDisposed();
+  return asin(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/asinh.js
+getGlobalTensorClass().prototype.asinh = function() {
+  this.throwIfDisposed();
+  return asinh(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atan.js
+getGlobalTensorClass().prototype.atan = function() {
+  this.throwIfDisposed();
+  return atan(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atan2.js
+getGlobalTensorClass().prototype.atan2 = function(b) {
+  this.throwIfDisposed();
+  return atan2(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/atanh.js
+getGlobalTensorClass().prototype.atanh = function() {
+  this.throwIfDisposed();
+  return atanh(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/avg_pool.js
+getGlobalTensorClass().prototype.avgPool = function(filterSize, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return avgPool(this, filterSize, strides, pad3, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/batch_to_space_nd.js
+getGlobalTensorClass().prototype.batchToSpaceND = function(blockShape, crops) {
+  this.throwIfDisposed();
+  return batchToSpaceND(this, blockShape, crops);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/batchnorm.js
+getGlobalTensorClass().prototype.batchNorm = function(mean4, variance, offset, scale2, varianceEpsilon) {
+  this.throwIfDisposed();
+  return batchNorm(this, mean4, variance, offset, scale2, varianceEpsilon);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/broadcast_to.js
+getGlobalTensorClass().prototype.broadcastTo = function(shape) {
+  this.throwIfDisposed();
+  return broadcastTo(this, shape);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cast.js
+getGlobalTensorClass().prototype.cast = function(dtype) {
+  this.throwIfDisposed();
+  return cast(this, dtype);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ceil.js
+getGlobalTensorClass().prototype.ceil = function() {
+  this.throwIfDisposed();
+  return ceil(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/clip_by_value.js
+getGlobalTensorClass().prototype.clipByValue = function(min6, max6) {
+  this.throwIfDisposed();
+  return clipByValue(this, min6, max6);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/concat.js
+getGlobalTensorClass().prototype.concat = function(x, axis) {
+  this.throwIfDisposed();
+  if (x instanceof Tensor) {
+    x = [x];
+  }
+  return concat([this, ...x], axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv1d.js
+getGlobalTensorClass().prototype.conv1d = function(filter, stride, pad3, dataFormat, dilation, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv1d(this, filter, stride, pad3, dataFormat, dilation, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv2d_transpose.js
+getGlobalTensorClass().prototype.conv2dTranspose = function(filter, outputShape, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv2dTranspose(this, filter, outputShape, strides, pad3, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/conv2d.js
+getGlobalTensorClass().prototype.conv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {
+  this.throwIfDisposed();
+  return conv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cos.js
+getGlobalTensorClass().prototype.cos = function() {
+  this.throwIfDisposed();
+  return cos(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cosh.js
+getGlobalTensorClass().prototype.cosh = function() {
+  this.throwIfDisposed();
+  return cosh(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cumprod.js
+getGlobalTensorClass().prototype.cumprod = function(axis, exclusive, reverse5) {
+  this.throwIfDisposed();
+  return cumprod(this, axis, exclusive, reverse5);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/cumsum.js
+getGlobalTensorClass().prototype.cumsum = function(axis, exclusive, reverse5) {
+  this.throwIfDisposed();
+  return cumsum(this, axis, exclusive, reverse5);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/depth_to_space.js
+getGlobalTensorClass().prototype.depthToSpace = function(blockSize, dataFormat) {
+  this.throwIfDisposed();
+  return depthToSpace(this, blockSize, dataFormat);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/depthwise_conv2d.js
+getGlobalTensorClass().prototype.depthwiseConv2d = function(filter, strides, pad3, dataFormat, dilations, dimRoundingMode) {
+  this.throwIfDisposed();
+  return depthwiseConv2d(this, filter, strides, pad3, dataFormat, dilations, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/dilation2d.js
+getGlobalTensorClass().prototype.dilation2d = function(filter, strides, pad3, dilations, dataFormat) {
+  this.throwIfDisposed();
+  return dilation2d(this, filter, strides, pad3, dilations, dataFormat);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/div_no_nan.js
+getGlobalTensorClass().prototype.divNoNan = function(b) {
+  this.throwIfDisposed();
+  return divNoNan(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/div.js
+getGlobalTensorClass().prototype.div = function(b) {
+  this.throwIfDisposed();
+  return div(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/dot.js
+getGlobalTensorClass().prototype.dot = function(b) {
+  this.throwIfDisposed();
+  return dot(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/elu.js
+getGlobalTensorClass().prototype.elu = function() {
+  this.throwIfDisposed();
+  return elu(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/equal.js
+getGlobalTensorClass().prototype.equal = function(b) {
+  this.throwIfDisposed();
+  return equal(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/erf.js
+getGlobalTensorClass().prototype.erf = function() {
+  this.throwIfDisposed();
+  return erf(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/euclidean_norm.js
+getGlobalTensorClass().prototype.euclideanNorm = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return euclideanNorm(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/exp.js
+getGlobalTensorClass().prototype.exp = function() {
+  this.throwIfDisposed();
+  return exp(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/expand_dims.js
+getGlobalTensorClass().prototype.expandDims = function(axis) {
+  this.throwIfDisposed();
+  return expandDims(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/expm1.js
+getGlobalTensorClass().prototype.expm1 = function() {
+  this.throwIfDisposed();
+  return expm1(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/fft.js
+getGlobalTensorClass().prototype.fft = function() {
+  this.throwIfDisposed();
+  return fft(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/flatten.js
+getGlobalTensorClass().prototype.flatten = function() {
+  this.throwIfDisposed();
+  return reshape(this, [this.size]);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/floor.js
+getGlobalTensorClass().prototype.floor = function() {
+  this.throwIfDisposed();
+  return floor(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/floorDiv.js
+getGlobalTensorClass().prototype.floorDiv = function(b) {
+  this.throwIfDisposed();
+  return floorDiv(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/gather.js
+getGlobalTensorClass().prototype.gather = function(indices, axis) {
+  this.throwIfDisposed();
+  return gather(this, indices, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/greater_equal.js
+getGlobalTensorClass().prototype.greaterEqual = function(b) {
+  this.throwIfDisposed();
+  return greaterEqual(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/greater.js
+getGlobalTensorClass().prototype.greater = function(b) {
+  this.throwIfDisposed();
+  return greater(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ifft.js
+getGlobalTensorClass().prototype.ifft = function() {
+  this.throwIfDisposed();
+  return ifft(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/irfft.js
+getGlobalTensorClass().prototype.irfft = function() {
+  this.throwIfDisposed();
+  return irfft(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_finite.js
+getGlobalTensorClass().prototype.isFinite = function() {
+  this.throwIfDisposed();
+  return isFinite2(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_inf.js
+getGlobalTensorClass().prototype.isInf = function() {
+  this.throwIfDisposed();
+  return isInf(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/is_nan.js
+getGlobalTensorClass().prototype.isNaN = function() {
+  this.throwIfDisposed();
+  return isNaN2(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/leaky_relu.js
+getGlobalTensorClass().prototype.leakyRelu = function(alpha) {
+  this.throwIfDisposed();
+  return leakyRelu(this, alpha);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/less_equal.js
+getGlobalTensorClass().prototype.lessEqual = function(b) {
+  this.throwIfDisposed();
+  return lessEqual(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/less.js
+getGlobalTensorClass().prototype.less = function(b) {
+  this.throwIfDisposed();
+  return less(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/local_response_normalization.js
+getGlobalTensorClass().prototype.localResponseNormalization = function(depthRadius, bias, alpha, beta) {
+  this.throwIfDisposed();
+  return localResponseNormalization(this, depthRadius, bias, alpha, beta);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_sigmoid.js
+getGlobalTensorClass().prototype.logSigmoid = function() {
+  this.throwIfDisposed();
+  return logSigmoid(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_softmax.js
+getGlobalTensorClass().prototype.logSoftmax = function(axis) {
+  this.throwIfDisposed();
+  return logSoftmax(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log_sum_exp.js
+getGlobalTensorClass().prototype.logSumExp = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return logSumExp(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log.js
+getGlobalTensorClass().prototype.log = function() {
+  this.throwIfDisposed();
+  return log2(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/log1p.js
+getGlobalTensorClass().prototype.log1p = function() {
+  this.throwIfDisposed();
+  return log1p(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_and.js
+getGlobalTensorClass().prototype.logicalAnd = function(b) {
+  this.throwIfDisposed();
+  return logicalAnd(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_not.js
+getGlobalTensorClass().prototype.logicalNot = function() {
+  this.throwIfDisposed();
+  return logicalNot(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_or.js
+getGlobalTensorClass().prototype.logicalOr = function(b) {
+  this.throwIfDisposed();
+  return logicalOr(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/logical_xor.js
+getGlobalTensorClass().prototype.logicalXor = function(b) {
+  this.throwIfDisposed();
+  return logicalXor(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mat_mul.js
+getGlobalTensorClass().prototype.matMul = function(b, transposeA, transposeB) {
+  this.throwIfDisposed();
+  return matMul(this, b, transposeA, transposeB);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/max_pool.js
+getGlobalTensorClass().prototype.maxPool = function(filterSize, strides, pad3, dimRoundingMode) {
+  this.throwIfDisposed();
+  return maxPool(this, filterSize, strides, pad3, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/max.js
+getGlobalTensorClass().prototype.max = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return max(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/maximum.js
+getGlobalTensorClass().prototype.maximum = function(b) {
+  this.throwIfDisposed();
+  return maximum(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mean.js
+getGlobalTensorClass().prototype.mean = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return mean(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/min.js
+getGlobalTensorClass().prototype.min = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return min(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/minimum.js
+getGlobalTensorClass().prototype.minimum = function(b) {
+  this.throwIfDisposed();
+  return minimum(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mirror_pad.js
+getGlobalTensorClass().prototype.mirrorPad = function(paddings, mode) {
+  this.throwIfDisposed();
+  return mirrorPad(this, paddings, mode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mod.js
+getGlobalTensorClass().prototype.mod = function(b) {
+  this.throwIfDisposed();
+  return mod(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/mul.js
+getGlobalTensorClass().prototype.mul = function(b) {
+  this.throwIfDisposed();
+  return mul(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/neg.js
+getGlobalTensorClass().prototype.neg = function() {
+  this.throwIfDisposed();
+  return neg(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/norm.js
+getGlobalTensorClass().prototype.norm = function(ord, axis, keepDims) {
+  this.throwIfDisposed();
+  return norm(this, ord, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/not_equal.js
+getGlobalTensorClass().prototype.notEqual = function(b) {
+  this.throwIfDisposed();
+  return notEqual(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/one_hot.js
+getGlobalTensorClass().prototype.oneHot = function(depth, onValue = 1, offValue = 0) {
+  this.throwIfDisposed();
+  return oneHot(this, depth, onValue, offValue);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/ones_like.js
+getGlobalTensorClass().prototype.onesLike = function() {
+  this.throwIfDisposed();
+  return onesLike(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pad.js
+getGlobalTensorClass().prototype.pad = function(paddings, constantValue) {
+  this.throwIfDisposed();
+  return pad(this, paddings, constantValue);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pool.js
+getGlobalTensorClass().prototype.pool = function(windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode) {
+  this.throwIfDisposed();
+  return pool(this, windowShape, poolingType, padding, dilationRate, strides, dimRoundingMode);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/pow.js
+getGlobalTensorClass().prototype.pow = function(exp4) {
+  this.throwIfDisposed();
+  return pow(this, exp4);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/prelu.js
+getGlobalTensorClass().prototype.prelu = function(alpha) {
+  this.throwIfDisposed();
+  return prelu(this, alpha);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/prod.js
+getGlobalTensorClass().prototype.prod = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return prod(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reciprocal.js
+getGlobalTensorClass().prototype.reciprocal = function() {
+  this.throwIfDisposed();
+  return reciprocal(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/relu.js
+getGlobalTensorClass().prototype.relu = function() {
+  this.throwIfDisposed();
+  return relu(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/relu6.js
+getGlobalTensorClass().prototype.relu6 = function() {
+  this.throwIfDisposed();
+  return relu6(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reshape_as.js
+getGlobalTensorClass().prototype.reshapeAs = function(x) {
+  this.throwIfDisposed();
+  return reshape(this, x.shape);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reshape.js
+getGlobalTensorClass().prototype.reshape = function(shape) {
+  this.throwIfDisposed();
+  return reshape(this, shape);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/resize_bilinear.js
+getGlobalTensorClass().prototype.resizeBilinear = function(newShape2D, alignCorners, halfPixelCenters) {
+  this.throwIfDisposed();
+  return resizeBilinear(this, newShape2D, alignCorners, halfPixelCenters);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/resize_nearest_neighbor.js
+getGlobalTensorClass().prototype.resizeNearestNeighbor = function(newShape2D, alignCorners, halfFloatCenters) {
+  this.throwIfDisposed();
+  return resizeNearestNeighbor(this, newShape2D, alignCorners, halfFloatCenters);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/reverse.js
+getGlobalTensorClass().prototype.reverse = function(axis) {
+  this.throwIfDisposed();
+  return reverse(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/rfft.js
+getGlobalTensorClass().prototype.rfft = function() {
+  this.throwIfDisposed();
+  return rfft(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/round.js
+getGlobalTensorClass().prototype.round = function() {
+  this.throwIfDisposed();
+  return round2(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/rsqrt.js
+getGlobalTensorClass().prototype.rsqrt = function() {
+  this.throwIfDisposed();
+  return rsqrt(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/selu.js
+getGlobalTensorClass().prototype.selu = function() {
+  this.throwIfDisposed();
+  return selu(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/separable_conv2d.js
+getGlobalTensorClass().prototype.separableConv2d = function(depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat) {
+  this.throwIfDisposed();
+  return separableConv2d(this, depthwiseFilter, pointwiseFilter, strides, pad3, dilation, dataFormat);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sigmoid.js
+getGlobalTensorClass().prototype.sigmoid = function() {
+  this.throwIfDisposed();
+  return sigmoid(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sign.js
+getGlobalTensorClass().prototype.sign = function() {
+  this.throwIfDisposed();
+  return sign(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sin.js
+getGlobalTensorClass().prototype.sin = function() {
+  this.throwIfDisposed();
+  return sin(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sinh.js
+getGlobalTensorClass().prototype.sinh = function() {
+  this.throwIfDisposed();
+  return sinh(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/slice.js
+getGlobalTensorClass().prototype.slice = function(begin, size) {
+  this.throwIfDisposed();
+  return slice(this, begin, size);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/softmax.js
+getGlobalTensorClass().prototype.softmax = function(dim) {
+  this.throwIfDisposed();
+  return softmax(this, dim);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/softplus.js
+getGlobalTensorClass().prototype.softplus = function() {
+  this.throwIfDisposed();
+  return softplus(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/space_to_batch_nd.js
+getGlobalTensorClass().prototype.spaceToBatchND = function(blockShape, paddings) {
+  this.throwIfDisposed();
+  return spaceToBatchND(this, blockShape, paddings);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/split.js
+getGlobalTensorClass().prototype.split = function(numOrSizeSplits, axis) {
+  this.throwIfDisposed();
+  return split(this, numOrSizeSplits, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sqrt.js
+getGlobalTensorClass().prototype.sqrt = function() {
+  this.throwIfDisposed();
+  return sqrt(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/square.js
+getGlobalTensorClass().prototype.square = function() {
+  this.throwIfDisposed();
+  return square(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/squared_difference.js
+getGlobalTensorClass().prototype.squaredDifference = function(b) {
+  this.throwIfDisposed();
+  return squaredDifference(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/squeeze.js
+getGlobalTensorClass().prototype.squeeze = function(axis) {
+  this.throwIfDisposed();
+  return squeeze(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/stack.js
+getGlobalTensorClass().prototype.stack = function(x, axis) {
+  this.throwIfDisposed();
+  const tensorsToBeStacked = x instanceof Tensor ? [this, x] : [this, ...x];
+  return stack(tensorsToBeStacked, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/step.js
+getGlobalTensorClass().prototype.step = function(alpha) {
+  this.throwIfDisposed();
+  return step(this, alpha);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/strided_slice.js
+getGlobalTensorClass().prototype.stridedSlice = function(begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask) {
+  this.throwIfDisposed();
+  return stridedSlice(this, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sub.js
+getGlobalTensorClass().prototype.sub = function(b) {
+  this.throwIfDisposed();
+  return sub(this, b);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/sum.js
+getGlobalTensorClass().prototype.sum = function(axis, keepDims) {
+  this.throwIfDisposed();
+  return sum2(this, axis, keepDims);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tan.js
+getGlobalTensorClass().prototype.tan = function() {
+  this.throwIfDisposed();
+  return tan(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tanh.js
+getGlobalTensorClass().prototype.tanh = function() {
+  this.throwIfDisposed();
+  return tanh2(this);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/tile.js
+getGlobalTensorClass().prototype.tile = function(reps) {
+  this.throwIfDisposed();
+  return tile(this, reps);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_bool.js
+getGlobalTensorClass().prototype.toBool = function() {
+  this.throwIfDisposed();
+  return cast(this, "bool");
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_float.js
+getGlobalTensorClass().prototype.toFloat = function() {
+  this.throwIfDisposed();
+  return cast(this, "float32");
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/to_int.js
+getGlobalTensorClass().prototype.toInt = function() {
+  this.throwIfDisposed();
+  return cast(this, "int32");
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/topk.js
+getGlobalTensorClass().prototype.topk = function(k, sorted) {
+  this.throwIfDisposed();
+  return topk(this, k, sorted);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/transpose.js
+getGlobalTensorClass().prototype.transpose = function(perm) {
+  this.throwIfDisposed();
+  return transpose(this, perm);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unique.js
+getGlobalTensorClass().prototype.unique = function(axis) {
+  this.throwIfDisposed();
+  return unique(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unsorted_segment_sum.js
+getGlobalTensorClass().prototype.unsortedSegmentSum = function(segmentIds, numSegments) {
+  this.throwIfDisposed();
+  return unsortedSegmentSum(this, segmentIds, numSegments);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/unstack.js
+getGlobalTensorClass().prototype.unstack = function(axis) {
+  this.throwIfDisposed();
+  return unstack(this, axis);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/where.js
+getGlobalTensorClass().prototype.where = function(condition, x) {
+  this.throwIfDisposed();
+  return where(condition, this, x);
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/public/chained_ops/zeros_like.js
+getGlobalTensorClass().prototype.zerosLike = function() {
+  this.throwIfDisposed();
+  return zerosLike(this);
+};
+
+// node_modules/@tensorflow/tfjs-layers/dist/errors.js
+var AttributeError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, AttributeError.prototype);
+  }
+};
+var RuntimeError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, RuntimeError.prototype);
+  }
+};
+var ValueError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, ValueError.prototype);
+  }
+};
+var NotImplementedError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, NotImplementedError.prototype);
+  }
+};
+var AssertionError = class extends Error {
+  constructor(message) {
+    super(message);
+    Object.setPrototypeOf(this, AssertionError.prototype);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/executor_utils.js
+var LruCache = class {
+  constructor(maxEntries) {
+    this.maxEntries = maxEntries || 100;
+    this.cache = /* @__PURE__ */ new Map();
+  }
+  get(key) {
+    let entry;
+    if (this.cache.has(key)) {
+      entry = this.cache.get(key);
+      this.cache.delete(key);
+      this.cache.set(key, entry);
+    }
+    return entry;
+  }
+  put(key, value) {
+    if (this.cache.has(key)) {
+      this.cache.delete(key);
+    } else if (this.cache.size >= this.maxEntries) {
+      const keyToDelete = this.cache.keys().next().value;
+      this.cache.delete(keyToDelete);
+    }
+    this.cache.set(key, value);
+  }
+  getMaxEntries() {
+    return this.maxEntries;
+  }
+  setMaxEntries(maxEntries) {
+    if (maxEntries < 0) {
+      throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${maxEntries}.`);
+    }
+    if (this.maxEntries > maxEntries) {
+      for (let i = 0; i < this.maxEntries - maxEntries; i++) {
+        const keyToDelete = this.cache.keys().next().value;
+        this.cache.delete(keyToDelete);
+      }
+    }
+    this.maxEntries = maxEntries;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/generic_utils.js
+function pyListRepeat(value, numValues) {
+  if (Array.isArray(value)) {
+    let newArray = [];
+    for (let i = 0; i < numValues; i++) {
+      newArray = newArray.concat(value);
+    }
+    return newArray;
+  } else {
+    const newArray = new Array(numValues);
+    newArray.fill(value);
+    return newArray;
+  }
+}
+function assert2(val, message) {
+  if (!val) {
+    throw new AssertionError(message);
+  }
+}
+function count(array2, refernce) {
+  let counter = 0;
+  for (const item of array2) {
+    if (item === refernce) {
+      counter++;
+    }
+  }
+  return counter;
+}
+function singletonOrArray(xs) {
+  if (xs.length === 1) {
+    return xs[0];
+  }
+  return xs;
+}
+function toList(x) {
+  if (Array.isArray(x)) {
+    return x;
+  }
+  return [x];
+}
+function toSnakeCase(name) {
+  const intermediate = name.replace(/(.)([A-Z][a-z0-9]+)/g, "$1_$2");
+  const insecure = intermediate.replace(/([a-z])([A-Z])/g, "$1_$2").toLowerCase();
+  if (insecure[0] !== "_") {
+    return insecure;
+  }
+  return "private" + insecure;
+}
+function toCamelCase(identifier) {
+  if (identifier.length <= 1) {
+    return identifier;
+  }
+  if (identifier.indexOf("_") === -1) {
+    return identifier;
+  }
+  return identifier.replace(/[_]+(\w|$)/g, (m, p1) => p1.toUpperCase());
+}
+var _GLOBAL_CUSTOM_OBJECTS = {};
+function serializeKerasObject(instance) {
+  if (instance === null || instance === void 0) {
+    return null;
+  }
+  const dict = {};
+  dict["className"] = instance.getClassName();
+  dict["config"] = instance.getConfig();
+  return dict;
+}
+function convertNDArrayScalarsInConfig(config) {
+  if (config == null || typeof config !== "object") {
+    return;
+  } else if (Array.isArray(config)) {
+    config.forEach((configItem) => convertNDArrayScalarsInConfig(configItem));
+  } else {
+    const fields = Object.keys(config);
+    for (const field of fields) {
+      const value = config[field];
+      if (value != null && typeof value === "object") {
+        if (!Array.isArray(value) && value["type"] === "ndarray" && typeof value["value"] === "number") {
+          config[field] = value["value"];
+        } else {
+          convertNDArrayScalarsInConfig(value);
+        }
+      }
+    }
+  }
+}
+function deserializeKerasObject(identifier, moduleObjects = {}, customObjects = {}, printableModuleName = "object", fastWeightInit = false) {
+  if (typeof identifier === "string") {
+    const functionName = identifier;
+    let fn;
+    if (functionName in customObjects) {
+      fn = customObjects[functionName];
+    } else if (functionName in _GLOBAL_CUSTOM_OBJECTS) {
+      fn = _GLOBAL_CUSTOM_OBJECTS[functionName];
+    } else {
+      fn = moduleObjects[functionName];
+      if (fn == null) {
+        throw new ValueError(`Unknown ${printableModuleName}: ${identifier}. This may be due to one of the following reasons:
+1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+      }
+    }
+    return fn;
+  } else {
+    const config = identifier;
+    if (config["className"] == null || config["config"] == null) {
+      throw new ValueError(`${printableModuleName}: Improper config format: ${JSON.stringify(config)}.
+'className' and 'config' must set.`);
+    }
+    const className = config["className"];
+    let cls, fromConfig;
+    if (className in customObjects) {
+      [cls, fromConfig] = customObjects[className];
+    } else if (className in _GLOBAL_CUSTOM_OBJECTS) {
+      [cls, fromConfig] = _GLOBAL_CUSTOM_OBJECTS["className"];
+    } else if (className in moduleObjects) {
+      [cls, fromConfig] = moduleObjects[className];
+    }
+    if (cls == null) {
+      throw new ValueError(`Unknown ${printableModuleName}: ${className}. This may be due to one of the following reasons:
+1. The ${printableModuleName} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
+2. The custom ${printableModuleName} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);
+    }
+    if (fromConfig != null) {
+      const customObjectsCombined = {};
+      for (const key of Object.keys(_GLOBAL_CUSTOM_OBJECTS)) {
+        customObjectsCombined[key] = _GLOBAL_CUSTOM_OBJECTS[key];
+      }
+      for (const key of Object.keys(customObjects)) {
+        customObjectsCombined[key] = customObjects[key];
+      }
+      const nestedConfig = config["config"];
+      nestedConfig["customObjects"] = customObjectsCombined;
+      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);
+      for (const key of Object.keys(customObjects)) {
+        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];
+      }
+      convertNDArrayScalarsInConfig(config["config"]);
+      const returnObj = fromConfig(cls, config["config"], customObjects, fastWeightInit);
+      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);
+      return returnObj;
+    } else {
+      const backupCustomObjects = Object.assign({}, _GLOBAL_CUSTOM_OBJECTS);
+      for (const key of Object.keys(customObjects)) {
+        _GLOBAL_CUSTOM_OBJECTS[key] = customObjects[key];
+      }
+      const returnObj = new cls(config["config"]);
+      _GLOBAL_CUSTOM_OBJECTS = Object.assign({}, backupCustomObjects);
+      return returnObj;
+    }
+  }
+}
+function numberCompare(a, b) {
+  return a < b ? -1 : a > b ? 1 : 0;
+}
+function reverseNumberCompare(a, b) {
+  return -1 * numberCompare(a, b);
+}
+function unique2(xs) {
+  if (xs == null) {
+    return xs;
+  }
+  const out = [];
+  for (const x of xs) {
+    if (out.indexOf(x) === -1) {
+      out.push(x);
+    }
+  }
+  return out;
+}
+function isObjectEmpty(obj) {
+  if (obj == null) {
+    throw new ValueError(`Invalid value in obj: ${JSON.stringify(obj)}`);
+  }
+  for (const key in obj) {
+    if (obj.hasOwnProperty(key)) {
+      return false;
+    }
+  }
+  return true;
+}
+function checkStringTypeUnionValue(values, label, value) {
+  if (value == null) {
+    return;
+  }
+  if (values.indexOf(value) < 0) {
+    throw new ValueError(`${value} is not a valid ${label}.  Valid values are ${values} or null/undefined.`);
+  }
+}
+function checkArrayTypeAndLength(x, expectedType, minLength = 0, maxLength = Infinity) {
+  assert2(minLength >= 0);
+  assert2(maxLength >= minLength);
+  return Array.isArray(x) && x.length >= minLength && x.length <= maxLength && x.every((e) => typeof e === expectedType);
+}
+function assertPositiveInteger(value, name) {
+  if (Array.isArray(value)) {
+    util_exports.assert(value.length > 0, () => `${name} is unexpectedly an empty array.`);
+    value.forEach((v, i) => assertPositiveInteger(v, `element ${i + 1} of ${name}`));
+  } else {
+    util_exports.assert(Number.isInteger(value) && value > 0, () => `Expected ${name} to be a positive integer, but got ${formatAsFriendlyString(value)}.`);
+  }
+}
+function formatAsFriendlyString(value) {
+  if (value === null) {
+    return "null";
+  } else if (Array.isArray(value)) {
+    return "[" + value.map((v) => formatAsFriendlyString(v)).join(",") + "]";
+  } else if (typeof value === "string") {
+    return `"${value}"`;
+  } else {
+    return `${value}`;
+  }
+}
+function debounce(f, waitMs, nowFunc) {
+  let lastTime = nowFunc != null ? nowFunc() : util_exports.now();
+  let lastResult;
+  const f2 = (...args) => {
+    const now2 = nowFunc != null ? nowFunc() : util_exports.now();
+    if (now2 - lastTime < waitMs) {
+      return lastResult;
+    }
+    lastTime = now2;
+    lastResult = f(...args);
+    return lastResult;
+  };
+  return f2;
+}
+function mapActivationToFusedKernel(activationName) {
+  if (activationName === "relu") {
+    return "relu";
+  }
+  if (activationName === "linear") {
+    return "linear";
+  }
+  if (activationName === "elu") {
+    return "elu";
+  }
+  return null;
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/backend/state.js
+var _nextUniqueTensorId = 0;
+function getNextUniqueTensorId() {
+  return _nextUniqueTensorId++;
+}
+var _uidPrefixes = {};
+function getUid(prefix = "") {
+  if (!(prefix in _uidPrefixes)) {
+    _uidPrefixes[prefix] = 0;
+  }
+  _uidPrefixes[prefix] += 1;
+  return prefix + _uidPrefixes[prefix].toString();
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/keras_format/common.js
+var VALID_DATA_FORMAT_VALUES = ["channelsFirst", "channelsLast"];
+var VALID_INTERPOLATION_FORMAT_VALUES = ["nearest", "bilinear"];
+var VALID_PADDING_MODE_VALUES = ["valid", "same", "causal"];
+var VALID_POOL_MODE_VALUES = ["max", "avg"];
+var VALID_BIDIRECTIONAL_MERGE_MODES = ["sum", "mul", "concat", "ave"];
+
+// node_modules/@tensorflow/tfjs-layers/dist/common.js
+var nameMap = /* @__PURE__ */ new Map();
+function checkDataFormat(value) {
+  checkStringTypeUnionValue(VALID_DATA_FORMAT_VALUES, "DataFormat", value);
+}
+function checkInterpolationFormat(value) {
+  checkStringTypeUnionValue(VALID_INTERPOLATION_FORMAT_VALUES, "InterpolationFormat", value);
+}
+function checkPaddingMode(value) {
+  checkStringTypeUnionValue(VALID_PADDING_MODE_VALUES, "PaddingMode", value);
+}
+function checkPoolMode(value) {
+  checkStringTypeUnionValue(VALID_POOL_MODE_VALUES, "PoolMode", value);
+}
+var _nameScopeStack = [];
+var _nameScopeDivider = "/";
+function nameScope(name, fn) {
+  _nameScopeStack.push(name);
+  try {
+    const val = fn();
+    _nameScopeStack.pop();
+    return val;
+  } catch (e) {
+    _nameScopeStack.pop();
+    throw e;
+  }
+}
+function currentNameScopePrefix() {
+  if (_nameScopeStack.length === 0) {
+    return "";
+  } else {
+    return _nameScopeStack.join(_nameScopeDivider) + _nameScopeDivider;
+  }
+}
+function getScopedTensorName(tensorName) {
+  if (!isValidTensorName(tensorName)) {
+    throw new Error("Not a valid tensor name: '" + tensorName + "'");
+  }
+  return currentNameScopePrefix() + tensorName;
+}
+function getUniqueTensorName(scopedName) {
+  if (!isValidTensorName(scopedName)) {
+    throw new Error("Not a valid tensor name: '" + scopedName + "'");
+  }
+  if (!nameMap.has(scopedName)) {
+    nameMap.set(scopedName, 0);
+  }
+  const index = nameMap.get(scopedName);
+  nameMap.set(scopedName, nameMap.get(scopedName) + 1);
+  if (index > 0) {
+    const result = `${scopedName}_${index}`;
+    nameMap.set(result, 1);
+    return result;
+  } else {
+    return scopedName;
+  }
+}
+var tensorNameRegex = new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);
+function isValidTensorName(name) {
+  return !!name.match(tensorNameRegex);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/math_utils.js
+function isInteger(x) {
+  return x === parseInt(x.toString(), 10);
+}
+function arrayProd(array2, begin, end) {
+  if (begin == null) {
+    begin = 0;
+  }
+  if (end == null) {
+    end = array2.length;
+  }
+  let prod5 = 1;
+  for (let i = begin; i < end; ++i) {
+    prod5 *= array2[i];
+  }
+  return prod5;
+}
+function min2(array2) {
+  if (array2.length === 0) {
+    return Number.NaN;
+  }
+  let min6 = Number.POSITIVE_INFINITY;
+  for (let i = 0; i < array2.length; i++) {
+    const value = array2[i];
+    if (value < min6) {
+      min6 = value;
+    }
+  }
+  return min6;
+}
+function max2(array2) {
+  if (array2.length === 0) {
+    return Number.NaN;
+  }
+  let max6 = Number.NEGATIVE_INFINITY;
+  for (let i = 0; i < array2.length; i++) {
+    const value = array2[i];
+    if (value > max6) {
+      max6 = value;
+    }
+  }
+  return max6;
+}
+function range2(begin, end) {
+  if (end < begin) {
+    throw new ValueError(`end (${end}) < begin (${begin}) is forbidden.`);
+  }
+  const out = [];
+  for (let i = begin; i < end; ++i) {
+    out.push(i);
+  }
+  return out;
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/backend/common.js
+var _epsilon;
+function epsilon() {
+  if (_epsilon == null) {
+    _epsilon = backend().epsilon();
+  }
+  return _epsilon;
+}
+function imageDataFormat() {
+  return "channelsLast";
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/backend/tfjs_backend.js
+function cast2(x, dtype) {
+  return cast(x, dtype);
+}
+function expandDims2(x, axis = -1) {
+  const outShape = x.shape.slice();
+  if (axis < 0) {
+    axis = outShape.length + axis + 1;
+  }
+  outShape.splice(axis, 0, 1);
+  return reshape(x, outShape);
+}
+function repeat(x, n) {
+  return tidy(() => {
+    if (x.shape.length !== 2) {
+      throw new ValueError(`repeat() expects a rank-2 tensor, but received a rank-${x.shape.length} tensor.`);
+    }
+    const y = expandDims2(x, 1);
+    return tile2(y, [1, n, 1]);
+  });
+}
+function flatten2(x) {
+  const newShape = [arrayProd(x.shape)];
+  return reshape(x, newShape);
+}
+function batchFlatten(x) {
+  if (x.rank <= 1) {
+    throw new ValueError(`batchFlatten requires a minimum rank of 2. Got rank: ${x.rank}.`);
+  }
+  const newShape = [x.shape[0], arrayProd(x.shape, 1)];
+  return reshape(x, newShape);
+}
+function sliceAlongFirstAxis(array2, start, size) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        return slice2d(array2, [start, 0], [size, array2.shape[1]]);
+      case 3:
+        return slice3d(array2, [start, 0, 0], [size, array2.shape[1], array2.shape[2]]);
+      case 4:
+        return slice4d(array2, [start, 0, 0, 0], [size, array2.shape[1], array2.shape[2], array2.shape[3]]);
+      case 5:
+        return slice(array2, [start, 0, 0, 0, 0], [
+          size,
+          array2.shape[1],
+          array2.shape[2],
+          array2.shape[3],
+          array2.shape[4]
+        ]);
+      case 6:
+        return slice(array2, [start, 0, 0, 0, 0, 0], [
+          size,
+          array2.shape[1],
+          array2.shape[2],
+          array2.shape[3],
+          array2.shape[4],
+          array2.shape[5]
+        ]);
+      default:
+        throw new ValueError(`sliceAlongFirstAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function sliceAlongLastAxis(array2, start, size) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        return slice2d(array2, [0, start], [array2.shape[0], size]);
+      case 3:
+        return slice3d(array2, [0, 0, start], [array2.shape[0], array2.shape[1], size]);
+      case 4:
+        return slice4d(array2, [0, 0, 0, start], [array2.shape[0], array2.shape[1], array2.shape[2], size]);
+      default:
+        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function sliceAlongAxis(array2, start, size, axis) {
+  return tidy(() => {
+    switch (array2.rank) {
+      case 1:
+        return slice1d(array2, start, size);
+      case 2:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      case 3:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return slice3d(array2, [0, start, 0], [array2.shape[0], size, array2.shape[2]]);
+          case 3:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      case 4:
+        switch (axis) {
+          case 1:
+            return sliceAlongFirstAxis(array2, start, size);
+          case 2:
+            return slice4d(array2, [0, start, 0, 0], [array2.shape[0], size, array2.shape[2], array2.shape[3]]);
+          case 3:
+            return slice4d(array2, [0, 0, start, 0], [array2.shape[0], array2.shape[1], size, array2.shape[3]]);
+          case 4:
+            return sliceAlongLastAxis(array2, start, size);
+          default:
+            throw new ValueError(`The axis is not within the rank of the tensor ${axis}`);
+        }
+      default:
+        throw new ValueError(`sliceAlongLastAxis() received an unsupported tensor rank: ${array2.rank}`);
+    }
+  });
+}
+function concatenate(tensors, axis = -1) {
+  let rank;
+  if (axis < 0) {
+    rank = tensors[0].rank;
+    if (rank !== 0) {
+      axis = rank;
+    } else {
+      axis = 0;
+    }
+  }
+  if (axis === tensors[0].rank) {
+    axis = -1;
+  }
+  return concat(tensors, axis);
+}
+function concatAlongFirstAxis(a, b) {
+  switch (a.rank) {
+    case 1:
+      return concat1d([a, b]);
+    case 2:
+      return concat2d([a, b], 0);
+    case 3:
+      return concat3d([a, b], 0);
+    case 4:
+      return concat4d([a, b], 0);
+    default:
+      throw new ValueError(`concatAlongFirstAxis() received an unsupported tensor rank: ${a.rank}`);
+  }
+}
+function tile2(x, n) {
+  if (!Array.isArray(n)) {
+    n = [n];
+  }
+  if (x.rank !== n.length) {
+    throw new ValueError(`The length of input n (${n.length}) does not match the number of dimensions in input x (${x.rank})`);
+  }
+  return tile(x, n);
+}
+function randomNormal2(shape, mean4 = 0, stddev = 1, dtype, seed) {
+  return randomNormal(shape, mean4, stddev, dtype, seed);
+}
+function dot2(a, b, activation2, bias) {
+  if (a.rank < 2 || b.rank < 2) {
+    throw new NotImplementedError(`dot requires both inputs to be rank >= 2 but got x shape = ${a.shape} and y shape = ${b.shape}`);
+  }
+  if (b.rank >= 3) {
+    const xLastDim = a.shape.slice(-1)[0];
+    const ySecondLastDim = b.shape.slice(-2)[0];
+    if (xLastDim !== ySecondLastDim) {
+      throw new NotImplementedError(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${a.shape} and  y shape = ${b.shape}`);
+    }
+  }
+  if (a.rank === 2 && b.rank === 2) {
+    const transposeA = false;
+    const transposeB = false;
+    return fused_ops_exports.matMul({
+      a,
+      b,
+      transposeA,
+      transposeB,
+      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,
+      activation: activation2
+    });
+  } else {
+    const aFirstDims = a.shape.slice();
+    const aLastDim = aFirstDims.pop();
+    a = reshape(a, [-1, aLastDim]);
+    const bShape = b.shape.slice();
+    const bLastDim = bShape.pop();
+    const ySecondLastDim = bShape.pop();
+    const yOtherDims = [...bShape, bLastDim];
+    const perm = Array.from({ length: b.rank }, (_, i) => {
+      if (i === 0) {
+        return b.rank - 2;
+      } else if (i <= b.rank - 2) {
+        return i - 1;
+      }
+      return i;
+    });
+    b = reshape(transpose(b, perm), [ySecondLastDim, -1]);
+    const outputShape = [...aFirstDims, ...yOtherDims];
+    const transposeA = false;
+    const transposeB = false;
+    return reshape(fused_ops_exports.matMul({
+      a,
+      b,
+      transposeA,
+      transposeB,
+      bias: bias ? reshapeBias(a.rank, bias, imageDataFormat()) : null,
+      activation: activation2
+    }), outputShape);
+  }
+}
+function gather2(reference, indices, axis) {
+  return tidy(() => {
+    if (Array.isArray(indices)) {
+      indices = tensor1d(indices, "int32");
+    } else {
+      indices = cast(indices, "int32");
+    }
+    return gather(reference, indices, axis);
+  });
+}
+function square2(x) {
+  return mul(x, x);
+}
+function reshapeBias(xRank, bias, dataFormat) {
+  const biasShape = bias.shape;
+  if (bias.rank !== 1 && bias.rank !== xRank) {
+    throw new ValueError(`Unexpected bias dimensions: ${bias.rank}; expected it to be 1 or ${xRank}`);
+  }
+  if (xRank === 5) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1, 1, 1]);
+      } else {
+        return reshape(bias, [1, biasShape[3], biasShape[0], biasShape[1], biasShape[2]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, 1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank === 4) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1, 1]);
+      } else {
+        return reshape(bias, [1, biasShape[2], biasShape[0], biasShape[1]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank === 3) {
+    if (dataFormat === "channelsFirst") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, biasShape[0], 1]);
+      } else {
+        return reshape(bias, [1, biasShape[1], biasShape[0]]);
+      }
+    } else if (dataFormat === "channelsLast") {
+      if (biasShape.length === 1) {
+        return reshape(bias, [1, 1, biasShape[0]]);
+      } else {
+        return reshape(bias, [1].concat(biasShape));
+      }
+    }
+  } else if (xRank < 3) {
+    return bias;
+  }
+  throw new ValueError(`Unsupported input rank by biasAdd: ${bias.rank}`);
+}
+function biasAdd(x, bias, dataFormat) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    return add2(x, reshapeBias(x.rank, bias, dataFormat));
+  });
+}
+function elu2(x, alpha = 1) {
+  if (alpha !== 1) {
+    throw new NotImplementedError(`Support for alpha values other than 1 (${alpha}) is not implemented yet.`);
+  }
+  return elu(x);
+}
+function softsign(x) {
+  return tidy(() => div(x, add2(abs(x), 1)));
+}
+function dropout2(x, level, noiseShape, seed) {
+  return tidy(() => dropout(x, level, noiseShape, seed));
+}
+function hardSigmoid(x) {
+  return tidy(() => {
+    const y = add2(0.5, mul(0.2, x));
+    return clipByValue(y, 0, 1);
+  });
+}
+function inTrainPhase(x, alt, training = false) {
+  return training ? x() : alt();
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/keras_format/initializer_config.js
+var VALID_FAN_MODE_VALUES = ["fanIn", "fanOut", "fanAvg"];
+var VALID_DISTRIBUTION_VALUES = ["normal", "uniform", "truncatedNormal"];
+
+// node_modules/@tensorflow/tfjs-layers/dist/initializers.js
+function checkFanMode(value) {
+  checkStringTypeUnionValue(VALID_FAN_MODE_VALUES, "FanMode", value);
+}
+function checkDistribution(value) {
+  checkStringTypeUnionValue(VALID_DISTRIBUTION_VALUES, "Distribution", value);
+}
+var Initializer = class extends serialization_exports.Serializable {
+  fromConfigUsesCustomObjects() {
+    return false;
+  }
+  getConfig() {
+    return {};
+  }
+};
+var Zeros = class extends Initializer {
+  apply(shape, dtype) {
+    return zeros(shape, dtype);
+  }
+};
+Zeros.className = "Zeros";
+serialization_exports.registerClass(Zeros);
+var Ones = class extends Initializer {
+  apply(shape, dtype) {
+    return ones2(shape, dtype);
+  }
+};
+Ones.className = "Ones";
+serialization_exports.registerClass(Ones);
+var Constant = class extends Initializer {
+  constructor(args) {
+    super();
+    if (typeof args !== "object") {
+      throw new ValueError(`Expected argument of type ConstantConfig but got ${args}`);
+    }
+    if (args.value === void 0) {
+      throw new ValueError(`config must have value set but got ${args}`);
+    }
+    this.value = args.value;
+  }
+  apply(shape, dtype) {
+    return tidy(() => mul(scalar(this.value), ones2(shape, dtype)));
+  }
+  getConfig() {
+    return {
+      value: this.value
+    };
+  }
+};
+Constant.className = "Constant";
+serialization_exports.registerClass(Constant);
+var RandomUniform = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MINVAL = -0.05;
+    this.DEFAULT_MAXVAL = 0.05;
+    this.minval = args.minval || this.DEFAULT_MINVAL;
+    this.maxval = args.maxval || this.DEFAULT_MAXVAL;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    return randomUniform(shape, this.minval, this.maxval, dtype);
+  }
+  getConfig() {
+    return { minval: this.minval, maxval: this.maxval, seed: this.seed };
+  }
+};
+RandomUniform.className = "RandomUniform";
+serialization_exports.registerClass(RandomUniform);
+var RandomNormal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MEAN = 0;
+    this.DEFAULT_STDDEV = 0.05;
+    this.mean = args.mean || this.DEFAULT_MEAN;
+    this.stddev = args.stddev || this.DEFAULT_STDDEV;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    dtype = dtype || "float32";
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new NotImplementedError(`randomNormal does not support dType ${dtype}.`);
+    }
+    return randomNormal2(shape, this.mean, this.stddev, dtype, this.seed);
+  }
+  getConfig() {
+    return { mean: this.mean, stddev: this.stddev, seed: this.seed };
+  }
+};
+RandomNormal.className = "RandomNormal";
+serialization_exports.registerClass(RandomNormal);
+var TruncatedNormal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_MEAN = 0;
+    this.DEFAULT_STDDEV = 0.05;
+    this.mean = args.mean || this.DEFAULT_MEAN;
+    this.stddev = args.stddev || this.DEFAULT_STDDEV;
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    dtype = dtype || "float32";
+    if (dtype !== "float32" && dtype !== "int32") {
+      throw new NotImplementedError(`truncatedNormal does not support dType ${dtype}.`);
+    }
+    return truncatedNormal(shape, this.mean, this.stddev, dtype, this.seed);
+  }
+  getConfig() {
+    return { mean: this.mean, stddev: this.stddev, seed: this.seed };
+  }
+};
+TruncatedNormal.className = "TruncatedNormal";
+serialization_exports.registerClass(TruncatedNormal);
+var Identity2 = class extends Initializer {
+  constructor(args) {
+    super();
+    this.gain = args.gain != null ? args.gain : 1;
+  }
+  apply(shape, dtype) {
+    return tidy(() => {
+      if (shape.length !== 2 || shape[0] !== shape[1]) {
+        throw new ValueError("Identity matrix initializer can only be used for 2D square matrices.");
+      } else {
+        return mul(this.gain, eye(shape[0]));
+      }
+    });
+  }
+  getConfig() {
+    return { gain: this.gain };
+  }
+};
+Identity2.className = "Identity";
+serialization_exports.registerClass(Identity2);
+function computeFans(shape, dataFormat = "channelsLast") {
+  let fanIn;
+  let fanOut;
+  checkDataFormat(dataFormat);
+  if (shape.length === 2) {
+    fanIn = shape[0];
+    fanOut = shape[1];
+  } else if ([3, 4, 5].indexOf(shape.length) !== -1) {
+    if (dataFormat === "channelsFirst") {
+      const receptiveFieldSize = arrayProd(shape, 2);
+      fanIn = shape[1] * receptiveFieldSize;
+      fanOut = shape[0] * receptiveFieldSize;
+    } else if (dataFormat === "channelsLast") {
+      const receptiveFieldSize = arrayProd(shape, 0, shape.length - 2);
+      fanIn = shape[shape.length - 2] * receptiveFieldSize;
+      fanOut = shape[shape.length - 1] * receptiveFieldSize;
+    }
+  } else {
+    const shapeProd = arrayProd(shape);
+    fanIn = Math.sqrt(shapeProd);
+    fanOut = Math.sqrt(shapeProd);
+  }
+  return [fanIn, fanOut];
+}
+var VarianceScaling = class extends Initializer {
+  constructor(args) {
+    super();
+    if (args.scale < 0) {
+      throw new ValueError(`scale must be a positive float. Got: ${args.scale}`);
+    }
+    this.scale = args.scale == null ? 1 : args.scale;
+    this.mode = args.mode == null ? "fanIn" : args.mode;
+    checkFanMode(this.mode);
+    this.distribution = args.distribution == null ? "normal" : args.distribution;
+    checkDistribution(this.distribution);
+    this.seed = args.seed;
+  }
+  apply(shape, dtype) {
+    const fans = computeFans(shape);
+    const fanIn = fans[0];
+    const fanOut = fans[1];
+    let scale2 = this.scale;
+    if (this.mode === "fanIn") {
+      scale2 /= Math.max(1, fanIn);
+    } else if (this.mode === "fanOut") {
+      scale2 /= Math.max(1, fanOut);
+    } else {
+      scale2 /= Math.max(1, (fanIn + fanOut) / 2);
+    }
+    if (this.distribution === "normal") {
+      const stddev = Math.sqrt(scale2);
+      dtype = dtype || "float32";
+      if (dtype !== "float32" && dtype !== "int32") {
+        throw new NotImplementedError(`${this.getClassName()} does not support dType ${dtype}.`);
+      }
+      return truncatedNormal(shape, 0, stddev, dtype, this.seed);
+    } else {
+      const limit = Math.sqrt(3 * scale2);
+      return randomUniform(shape, -limit, limit, dtype);
+    }
+  }
+  getConfig() {
+    return {
+      scale: this.scale,
+      mode: this.mode,
+      distribution: this.distribution,
+      seed: this.seed
+    };
+  }
+};
+VarianceScaling.className = "VarianceScaling";
+serialization_exports.registerClass(VarianceScaling);
+var GlorotUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanAvg",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+GlorotUniform.className = "GlorotUniform";
+serialization_exports.registerClass(GlorotUniform);
+var GlorotNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanAvg",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+GlorotNormal.className = "GlorotNormal";
+serialization_exports.registerClass(GlorotNormal);
+var HeNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 2,
+      mode: "fanIn",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+HeNormal.className = "HeNormal";
+serialization_exports.registerClass(HeNormal);
+var HeUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 2,
+      mode: "fanIn",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+HeUniform.className = "HeUniform";
+serialization_exports.registerClass(HeUniform);
+var LeCunNormal = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanIn",
+      distribution: "normal",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+LeCunNormal.className = "LeCunNormal";
+serialization_exports.registerClass(LeCunNormal);
+var LeCunUniform = class extends VarianceScaling {
+  constructor(args) {
+    super({
+      scale: 1,
+      mode: "fanIn",
+      distribution: "uniform",
+      seed: args == null ? null : args.seed
+    });
+  }
+  getClassName() {
+    return VarianceScaling.className;
+  }
+};
+LeCunUniform.className = "LeCunNormal";
+serialization_exports.registerClass(LeCunUniform);
+var Orthogonal = class extends Initializer {
+  constructor(args) {
+    super();
+    this.DEFAULT_GAIN = 1;
+    this.gain = args.gain == null ? this.DEFAULT_GAIN : args.gain;
+    this.seed = args.seed;
+    if (this.seed != null) {
+      throw new NotImplementedError("Random seed is not implemented for Orthogonal Initializer yet.");
+    }
+  }
+  apply(shape, dtype) {
+    return tidy(() => {
+      if (shape.length < 2) {
+        throw new NotImplementedError("Shape must be at least 2D.");
+      }
+      if (shape[0] * shape[1] > 2e3) {
+        console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${shape[0] * shape[1]}) elements: Slowness may result.`);
+      }
+      const normalizedShape = shape[0] > shape[1] ? [shape[1], shape[0]] : shape;
+      const a = randomNormal2(normalizedShape, 0, 1, "float32");
+      let q = linalg.gramSchmidt(a);
+      if (shape[0] > shape[1]) {
+        q = transpose(q);
+      }
+      return mul(this.gain, q);
+    });
+  }
+  getConfig() {
+    return {
+      gain: this.gain,
+      seed: this.seed
+    };
+  }
+};
+Orthogonal.className = "Orthogonal";
+serialization_exports.registerClass(Orthogonal);
+var INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "constant": "Constant",
+  "glorotNormal": "GlorotNormal",
+  "glorotUniform": "GlorotUniform",
+  "heNormal": "HeNormal",
+  "heUniform": "HeUniform",
+  "identity": "Identity",
+  "leCunNormal": "LeCunNormal",
+  "leCunUniform": "LeCunUniform",
+  "ones": "Ones",
+  "orthogonal": "Orthogonal",
+  "randomNormal": "RandomNormal",
+  "randomUniform": "RandomUniform",
+  "truncatedNormal": "TruncatedNormal",
+  "varianceScaling": "VarianceScaling",
+  "zeros": "Zeros"
+};
+function deserializeInitializer(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "initializer");
+}
+function serializeInitializer(initializer) {
+  return serializeKerasObject(initializer);
+}
+function getInitializer(identifier) {
+  if (typeof identifier === "string") {
+    const className = identifier in INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? INITIALIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    if (className === "GlorotNormal") {
+      return new GlorotNormal();
+    } else if (className === "GlorotUniform") {
+      return new GlorotUniform();
+    } else if (className === "HeNormal") {
+      return new HeNormal();
+    } else if (className === "HeUniform") {
+      return new HeUniform();
+    } else if (className === "LeCunNormal") {
+      return new LeCunNormal();
+    } else if (className === "LeCunUniform") {
+      return new LeCunUniform();
+    } else {
+      const config = {};
+      config["className"] = className;
+      config["config"] = {};
+      return deserializeInitializer(config);
+    }
+  } else if (identifier instanceof Initializer) {
+    return identifier;
+  } else {
+    return deserializeInitializer(identifier);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/types_utils.js
+function isArrayOfShapes(x) {
+  return Array.isArray(x) && Array.isArray(x[0]);
+}
+function normalizeShapeList(x) {
+  if (x.length === 0) {
+    return [];
+  }
+  if (!Array.isArray(x[0])) {
+    return [x];
+  }
+  return x;
+}
+function getExactlyOneTensor(xs) {
+  let x;
+  if (Array.isArray(xs)) {
+    if (xs.length !== 1) {
+      throw new ValueError(`Expected Tensor length to be 1; got ${xs.length}`);
+    }
+    x = xs[0];
+  } else {
+    x = xs;
+  }
+  return x;
+}
+function getExactlyOneShape(shapes) {
+  if (Array.isArray(shapes) && Array.isArray(shapes[0])) {
+    if (shapes.length === 1) {
+      shapes = shapes;
+      return shapes[0];
+    } else {
+      throw new ValueError(`Expected exactly 1 Shape; got ${shapes.length}`);
+    }
+  } else {
+    return shapes;
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/variable_utils.js
+function countParamsInWeights(weights) {
+  let count2 = 0;
+  for (const weight of weights) {
+    if (weight.shape.length === 0) {
+      count2 += 1;
+    } else {
+      count2 += weight.shape.reduce((a, b) => a * b);
+    }
+  }
+  return count2;
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/variables.js
+var DEFAULT_VARIABLE_NAME_PREFIX = "Variable";
+var LayerVariable = class {
+  constructor(val, dtype = "float32", name = DEFAULT_VARIABLE_NAME_PREFIX, trainable = true, constraint = null) {
+    this.dtype = dtype == null ? "float32" : dtype;
+    this.shape = val.shape;
+    this.id = getNextUniqueTensorId();
+    name = name == null ? DEFAULT_VARIABLE_NAME_PREFIX : name;
+    this.originalName = getScopedTensorName(name);
+    this.name = getUniqueTensorName(this.originalName);
+    this.trainable_ = trainable;
+    this.constraint = constraint;
+    this.val = variable(val, this.trainable_, this.name, this.dtype);
+  }
+  read() {
+    this.assertNotDisposed();
+    return this.val;
+  }
+  write(newVal) {
+    this.assertNotDisposed();
+    checkShapesMatch(this.val, newVal);
+    if (this.val.id !== newVal.id) {
+      this.val.assign(newVal);
+      if (this.constraint != null) {
+        this.val.assign(this.constraint.apply(this.val));
+      }
+    }
+    return this;
+  }
+  dispose() {
+    this.assertNotDisposed();
+    this.val.dispose();
+  }
+  assertNotDisposed() {
+    if (this.val.isDisposed) {
+      throw new Error(`LayersVariable ${this.name} is already disposed.`);
+    }
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this.trainable_ = trainable;
+    this.val.trainable = trainable;
+  }
+};
+function checkShapesMatch(x, y) {
+  if (x.shape.toString() !== y.shape.toString()) {
+    throw new Error("Shape mismatch: " + JSON.stringify(x.shape) + " vs. " + JSON.stringify(y.shape));
+  }
+}
+function batchGetValue(xs) {
+  return xs.map((x) => x.read());
+}
+function batchSetValue(variablesAndValues) {
+  variablesAndValues.forEach((variableAndValue) => {
+    const variable2 = variableAndValue[0];
+    variable2.write(variableAndValue[1]);
+  });
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/topology.js
+var InputSpec = class {
+  constructor(args) {
+    this.dtype = args.dtype;
+    this.shape = args.shape;
+    if (args.shape != null) {
+      this.ndim = args.shape.length;
+    } else {
+      this.ndim = args.ndim;
+    }
+    this.maxNDim = args.maxNDim;
+    this.minNDim = args.minNDim;
+    this.axes = args.axes || {};
+  }
+};
+var SymbolicTensor = class {
+  constructor(dtype, shape, sourceLayer, inputs, callArgs, name, outputTensorIndex) {
+    this.dtype = dtype;
+    this.shape = shape;
+    this.sourceLayer = sourceLayer;
+    this.inputs = inputs;
+    this.callArgs = callArgs;
+    this.outputTensorIndex = outputTensorIndex;
+    this.id = getNextUniqueTensorId();
+    if (name != null) {
+      this.originalName = getScopedTensorName(name);
+      this.name = getUniqueTensorName(this.originalName);
+    }
+    this.rank = shape.length;
+  }
+};
+var _nextNodeID = 0;
+var Node = class {
+  constructor(args, callArgs) {
+    this.callArgs = callArgs;
+    this.id = _nextNodeID++;
+    this.outboundLayer = args.outboundLayer;
+    this.inboundLayers = args.inboundLayers;
+    this.nodeIndices = args.nodeIndices;
+    this.tensorIndices = args.tensorIndices;
+    this.inputTensors = args.inputTensors;
+    this.outputTensors = args.outputTensors;
+    this.inputMasks = args.inputMasks;
+    this.outputMasks = args.outputMasks;
+    this.inputShapes = args.inputShapes;
+    this.outputShapes = args.outputShapes;
+    for (const layer of args.inboundLayers) {
+      if (layer != null) {
+        layer.outboundNodes.push(this);
+      }
+    }
+    args.outboundLayer.inboundNodes.push(this);
+  }
+  getConfig() {
+    const inboundNames = [];
+    for (const layer of this.inboundLayers) {
+      if (layer != null) {
+        inboundNames.push(layer.name);
+      } else {
+        inboundNames.push(null);
+      }
+    }
+    return {
+      outboundLayer: this.outboundLayer ? this.outboundLayer.name : null,
+      inboundLayers: inboundNames,
+      nodeIndices: this.nodeIndices,
+      tensorIndices: this.tensorIndices
+    };
+  }
+};
+var _nextLayerID = 0;
+var Layer = class extends serialization_exports.Serializable {
+  constructor(args = {}) {
+    super();
+    this._callHook = null;
+    this._addedWeightNames = [];
+    this._stateful = false;
+    this.id = _nextLayerID++;
+    this.activityRegularizer = null;
+    this.inputSpec = null;
+    this.supportsMasking = false;
+    this._trainableWeights = [];
+    this._nonTrainableWeights = [];
+    this._losses = [];
+    this._updates = [];
+    this._built = false;
+    this.inboundNodes = [];
+    this.outboundNodes = [];
+    let name = args.name;
+    if (!name) {
+      const prefix = this.getClassName();
+      name = toSnakeCase(prefix) + "_" + getUid(prefix);
+    }
+    this.name = name;
+    this.trainable_ = args.trainable == null ? true : args.trainable;
+    if (args.inputShape != null || args.batchInputShape != null) {
+      let batchInputShape;
+      if (args.batchInputShape != null) {
+        batchInputShape = args.batchInputShape;
+      } else if (args.inputShape != null) {
+        let batchSize = null;
+        if (args.batchSize != null) {
+          batchSize = args.batchSize;
+        }
+        batchInputShape = [batchSize].concat(args.inputShape);
+      }
+      this.batchInputShape = batchInputShape;
+      let dtype = args.dtype;
+      if (dtype == null) {
+        dtype = args.inputDType;
+      }
+      if (dtype == null) {
+        dtype = "float32";
+      }
+      this.dtype = dtype;
+    }
+    if (args.weights != null) {
+      this.initialWeights = args.weights;
+    } else {
+      this.initialWeights = null;
+    }
+    this._refCount = null;
+    this.fastWeightInitDuringBuild = false;
+  }
+  static nodeKey(layer, nodeIndex) {
+    return layer.name + "_ib-" + nodeIndex.toString();
+  }
+  getNodeAtIndex(nodeIndex, attrName) {
+    if (this.inboundNodes.length === 0) {
+      throw new RuntimeError(`The layer has never been called and thus has no defined ${attrName}.`);
+    }
+    if (this.inboundNodes.length <= nodeIndex) {
+      throw new ValueError(`Asked to get ${attrName} at node ${nodeIndex}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);
+    }
+    return this.inboundNodes[nodeIndex];
+  }
+  getInputAt(nodeIndex) {
+    return singletonOrArray(this.getNodeAtIndex(nodeIndex, "input").inputTensors);
+  }
+  getOutputAt(nodeIndex) {
+    return singletonOrArray(this.getNodeAtIndex(nodeIndex, "output").outputTensors);
+  }
+  get input() {
+    if (this.inboundNodes.length > 1) {
+      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);
+    } else if (this.inboundNodes.length === 0) {
+      throw new AttributeError(`Layer ${this.name} is not connected, no input to return.`);
+    }
+    return singletonOrArray(this.getNodeAtIndex(0, "input").inputTensors);
+  }
+  get output() {
+    if (this.inboundNodes.length === 0) {
+      throw new AttributeError(`Layer ${this.name} has no inbound nodes.`);
+    }
+    if (this.inboundNodes.length > 1) {
+      throw new AttributeError(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);
+    }
+    return singletonOrArray(this.getNodeAtIndex(0, "output").outputTensors);
+  }
+  get losses() {
+    return this._losses;
+  }
+  calculateLosses() {
+    return this.losses.map((lossFn) => lossFn());
+  }
+  get updates() {
+    return this._updates;
+  }
+  get built() {
+    return this._built;
+  }
+  set built(built) {
+    this._built = built;
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this._trainableWeights.forEach((w) => w.trainable = trainable);
+    this.trainable_ = trainable;
+  }
+  get trainableWeights() {
+    if (this.trainable_) {
+      return this._trainableWeights.filter((w) => w.trainable);
+    } else {
+      return [];
+    }
+  }
+  set trainableWeights(weights) {
+    this._trainableWeights = weights;
+  }
+  get nonTrainableWeights() {
+    if (this.trainable) {
+      return this._trainableWeights.filter((w) => !w.trainable).concat(this._nonTrainableWeights);
+    } else {
+      return this._trainableWeights.concat(this._nonTrainableWeights);
+    }
+  }
+  set nonTrainableWeights(weights) {
+    this._nonTrainableWeights = weights;
+  }
+  get weights() {
+    return this.trainableWeights.concat(this.nonTrainableWeights);
+  }
+  get stateful() {
+    return this._stateful;
+  }
+  resetStates() {
+    if (!this.stateful) {
+      throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.");
+    }
+  }
+  assertInputCompatibility(inputs) {
+    inputs = toList(inputs);
+    if (this.inputSpec == null || this.inputSpec.length === 0) {
+      return;
+    }
+    const inputSpec = toList(this.inputSpec);
+    if (inputs.length !== inputSpec.length) {
+      throw new ValueError(`Layer ${this.name} expects ${inputSpec.length} inputs, but it received ${inputs.length} input tensors. Input received: ${inputs}`);
+    }
+    for (let inputIndex = 0; inputIndex < inputs.length; inputIndex++) {
+      const x = inputs[inputIndex];
+      const spec = inputSpec[inputIndex];
+      if (spec == null) {
+        continue;
+      }
+      const ndim = x.rank;
+      if (spec.ndim != null) {
+        if (ndim !== spec.ndim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected ndim=${spec.ndim}, found ndim=${ndim}`);
+        }
+      }
+      if (spec.maxNDim != null) {
+        if (ndim > spec.maxNDim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected max_ndim=${spec.maxNDim}, found ndim=${ndim}`);
+        }
+      }
+      if (spec.minNDim != null) {
+        if (ndim < spec.minNDim) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected min_ndim=${spec.minNDim}, found ndim=${ndim}.`);
+        }
+      }
+      if (spec.dtype != null) {
+        if (x.dtype !== spec.dtype) {
+          throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name} : expected dtype=${spec.dtype}, found dtype=${x.dtype}.`);
+        }
+      }
+      if (spec.axes) {
+        const xShape = x.shape;
+        for (const key in spec.axes) {
+          const axis = Number(key);
+          const value = spec.axes[key];
+          const xShapeAtAxis = axis >= 0 ? xShape[axis] : xShape[xShape.length + axis];
+          if (value != null && [value, null].indexOf(xShapeAtAxis) === -1) {
+            throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected axis ${axis} of input shape to have value ${value} but got shape ${xShape}.`);
+          }
+        }
+      }
+      if (spec.shape != null) {
+        for (let i = 0; i < spec.shape.length; ++i) {
+          const specDim = spec.shape[i];
+          const dim = x.shape[i];
+          if (specDim != null && dim != null) {
+            if (specDim !== dim) {
+              throw new ValueError(`Input ${inputIndex} is incompatible with layer ${this.name}: expected shape=${spec.shape}, found shape=${x.shape}.`);
+            }
+          }
+        }
+      }
+    }
+  }
+  call(inputs, kwargs) {
+    return inputs;
+  }
+  invokeCallHook(inputs, kwargs) {
+    if (this._callHook != null) {
+      this._callHook(inputs, kwargs);
+    }
+  }
+  setCallHook(callHook) {
+    this._callHook = callHook;
+  }
+  clearCallHook() {
+    this._callHook = null;
+  }
+  apply(inputs, kwargs) {
+    kwargs = kwargs || {};
+    this.assertNotDisposed();
+    const inputsList = toList(inputs);
+    let allAreSymbolic = true;
+    for (const input2 of inputsList) {
+      if (!(input2 instanceof SymbolicTensor)) {
+        allAreSymbolic = false;
+        break;
+      }
+    }
+    let noneAreSymbolic = true;
+    for (const input2 of inputsList) {
+      if (input2 instanceof SymbolicTensor) {
+        noneAreSymbolic = false;
+        break;
+      }
+    }
+    if (allAreSymbolic === noneAreSymbolic) {
+      throw new ValueError("Arguments to apply() must be all SymbolicTensors or all Tensors");
+    }
+    return nameScope(this.name, () => {
+      if (!this.built) {
+        this.assertInputCompatibility(inputs);
+        const inputShapes = [];
+        for (const xElem of toList(inputs)) {
+          inputShapes.push(xElem.shape);
+        }
+        this.build(singletonOrArray(inputShapes));
+        this.built = true;
+        if (this.initialWeights) {
+          this.setWeights(this.initialWeights);
+        }
+        if (this._refCount === null && noneAreSymbolic) {
+          this._refCount = 1;
+        }
+      }
+      this.assertInputCompatibility(inputs);
+      if (noneAreSymbolic) {
+        let output = this.call(inputs, kwargs);
+        const outputList = toList(output);
+        const outputListCopy = [];
+        for (let x of outputList) {
+          if (inputsList.indexOf(x) !== -1) {
+            x = x.clone();
+          }
+          outputListCopy.push(x);
+        }
+        output = singletonOrArray(outputListCopy);
+        if (this.activityRegularizer != null) {
+          throw new NotImplementedError("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+        }
+        return output;
+      } else {
+        const inputShape = collectInputShape(inputs);
+        const outputShape = this.computeOutputShape(inputShape);
+        let output;
+        const outputDType = guessOutputDType(inputs);
+        this.warnOnIncompatibleInputShape(Array.isArray(inputs) ? inputShape[0] : inputShape);
+        if (outputShape != null && outputShape.length > 0 && Array.isArray(outputShape[0])) {
+          output = outputShape.map((shape, index) => new SymbolicTensor(outputDType, shape, this, toList(inputs), kwargs, this.name, index));
+        } else {
+          output = new SymbolicTensor(outputDType, outputShape, this, toList(inputs), kwargs, this.name);
+        }
+        this.addInboundNode(inputs, output, null, null, inputShape, outputShape, kwargs);
+        this._refCount++;
+        if (this.activityRegularizer != null) {
+          throw new NotImplementedError("Layer invocation in the presence of activity regularizer(s) is not supported yet.");
+        }
+        return output;
+      }
+    });
+  }
+  warnOnIncompatibleInputShape(inputShape) {
+    if (this.batchInputShape == null) {
+      return;
+    } else if (inputShape.length !== this.batchInputShape.length) {
+      console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(inputShape)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);
+    } else {
+      let dimMismatch = false;
+      this.batchInputShape.forEach((dimension, i) => {
+        if (dimension != null && inputShape[i] != null && inputShape[i] !== dimension) {
+          dimMismatch = true;
+        }
+      });
+      if (dimMismatch) {
+        console.warn(`The shape of the input tensor (${JSON.stringify(inputShape)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`);
+      }
+    }
+  }
+  get outputShape() {
+    if (this.inboundNodes == null || this.inboundNodes.length === 0) {
+      throw new AttributeError(`The layer ${this.name} has never been called and thus has no defined output shape.`);
+    }
+    const allOutputShapes = [];
+    for (const node of this.inboundNodes) {
+      const shapeString = JSON.stringify(node.outputShapes);
+      if (allOutputShapes.indexOf(shapeString) === -1) {
+        allOutputShapes.push(shapeString);
+      }
+    }
+    if (allOutputShapes.length === 1) {
+      const outputShapes = this.inboundNodes[0].outputShapes;
+      if (Array.isArray(outputShapes) && Array.isArray(outputShapes[0]) && outputShapes.length === 1) {
+        return outputShapes[0];
+      } else {
+        return outputShapes;
+      }
+    } else {
+      throw new AttributeError(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`);
+    }
+  }
+  countParams() {
+    if (!this.built) {
+      throw new RuntimeError(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);
+    }
+    return countParamsInWeights(this.weights);
+  }
+  build(inputShape) {
+    this.built = true;
+  }
+  getWeights(trainableOnly = false) {
+    return batchGetValue(trainableOnly ? this.trainableWeights : this.weights);
+  }
+  setWeights(weights) {
+    tidy(() => {
+      const params = this.weights;
+      if (params.length !== weights.length) {
+        throw new ValueError(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${weights.length}, but the layer was expecting ${params.length} weights. Provided weights: ${weights}...`);
+      }
+      if (params.length === 0) {
+        return;
+      }
+      const weightValueTuples = [];
+      const paramValues = batchGetValue(params);
+      for (let i = 0; i < paramValues.length; ++i) {
+        const pv = paramValues[i];
+        const p2 = params[i];
+        const w = weights[i];
+        if (!util_exports.arraysEqual(pv.shape, w.shape)) {
+          throw new ValueError(`Layer weight shape ${pv.shape} not compatible with provided weight shape ${w.shape}`);
+        }
+        weightValueTuples.push([p2, w]);
+      }
+      batchSetValue(weightValueTuples);
+    });
+  }
+  addWeight(name, shape, dtype, initializer, regularizer, trainable, constraint, getInitializerFunc) {
+    if (this._addedWeightNames.indexOf(name) !== -1) {
+      throw new ValueError(`Duplicate weight name ${name} for layer ${this.name}`);
+    }
+    this._addedWeightNames.push(name);
+    if (dtype == null) {
+      dtype = "float32";
+    }
+    if (this.fastWeightInitDuringBuild) {
+      initializer = getInitializerFunc != null ? getInitializerFunc() : getInitializer("zeros");
+    }
+    const initValue = initializer.apply(shape, dtype);
+    const weight = new LayerVariable(initValue, dtype, name, trainable, constraint);
+    initValue.dispose();
+    if (regularizer != null) {
+      this.addLoss(() => regularizer.apply(weight.read()));
+    }
+    if (trainable == null) {
+      trainable = true;
+    }
+    if (trainable) {
+      this._trainableWeights.push(weight);
+    } else {
+      this._nonTrainableWeights.push(weight);
+    }
+    return weight;
+  }
+  setFastWeightInitDuringBuild(value) {
+    this.fastWeightInitDuringBuild = value;
+  }
+  addLoss(losses2) {
+    if (losses2 == null || Array.isArray(losses2) && losses2.length === 0) {
+      return;
+    }
+    losses2 = toList(losses2);
+    if (this._losses !== void 0 && this._losses !== null) {
+      this.losses.push(...losses2);
+    }
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  computeMask(inputs, mask) {
+    if (!this.supportsMasking) {
+      if (mask != null) {
+        if (Array.isArray(mask)) {
+          mask.forEach((maskElement) => {
+            if (maskElement != null) {
+              throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);
+            }
+          });
+        } else {
+          throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);
+        }
+      }
+      return null;
+    }
+    return mask;
+  }
+  addInboundNode(inputTensors, outputTensors, inputMasks, outputMasks, inputShapes, outputShapes, kwargs = null) {
+    const inputTensorList = toList(inputTensors);
+    outputTensors = toList(outputTensors);
+    inputMasks = toList(inputMasks);
+    outputMasks = toList(outputMasks);
+    inputShapes = normalizeShapeList(inputShapes);
+    outputShapes = normalizeShapeList(outputShapes);
+    const inboundLayers = [];
+    const nodeIndices = [];
+    const tensorIndices = [];
+    for (const x of inputTensorList) {
+      inboundLayers.push(x.sourceLayer);
+      nodeIndices.push(x.nodeIndex);
+      tensorIndices.push(x.tensorIndex);
+    }
+    new Node({
+      outboundLayer: this,
+      inboundLayers,
+      nodeIndices,
+      tensorIndices,
+      inputTensors: inputTensorList,
+      outputTensors,
+      inputMasks,
+      outputMasks,
+      inputShapes,
+      outputShapes
+    }, kwargs);
+    for (let i = 0; i < outputTensors.length; i++) {
+      outputTensors[i].sourceLayer = this;
+      outputTensors[i].nodeIndex = this.inboundNodes.length - 1;
+      outputTensors[i].tensorIndex = i;
+    }
+  }
+  getConfig() {
+    const config = { name: this.name, trainable: this.trainable };
+    if (this.batchInputShape != null) {
+      config["batchInputShape"] = this.batchInputShape;
+    }
+    if (this.dtype != null) {
+      config["dtype"] = this.dtype;
+    }
+    return config;
+  }
+  disposeWeights() {
+    this.weights.forEach((weight) => weight.dispose());
+    return this.weights.length;
+  }
+  assertNotDisposed() {
+    if (this._refCount === 0) {
+      throw new Error(`Layer '${this.name}' is already disposed.`);
+    }
+  }
+  dispose() {
+    if (!this.built) {
+      throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);
+    }
+    if (this._refCount === null) {
+      throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);
+    }
+    this.assertNotDisposed();
+    let numDisposedVariables = 0;
+    if (--this._refCount === 0) {
+      numDisposedVariables = this.disposeWeights();
+    }
+    return { refCountAfterDispose: this._refCount, numDisposedVariables };
+  }
+};
+function collectInputShape(inputTensors) {
+  inputTensors = toList(inputTensors);
+  const shapes = [];
+  for (const x of inputTensors) {
+    shapes.push(x.shape);
+  }
+  return singletonOrArray(shapes);
+}
+function guessOutputDType(inputTensors) {
+  return "float32";
+}
+function getSourceInputs(tensor2, layer, nodeIndex) {
+  if (layer == null || nodeIndex != null && nodeIndex > 0) {
+    layer = tensor2.sourceLayer;
+    nodeIndex = tensor2.nodeIndex;
+  }
+  if (layer.inboundNodes.length === 0) {
+    return [tensor2];
+  } else {
+    const node = layer.inboundNodes[nodeIndex];
+    if (node.inboundLayers.length === 0) {
+      return node.inputTensors;
+    } else {
+      const sourceTensors = [];
+      for (let i = 0; i < node.inboundLayers.length; i++) {
+        const x = node.inputTensors[i];
+        const layer2 = node.inboundLayers[i];
+        const nodeIndex2 = node.nodeIndices[i];
+        const previousSources = getSourceInputs(x, layer2, nodeIndex2);
+        for (const x2 of previousSources) {
+          if (sourceTensors.indexOf(x2) === -1) {
+            sourceTensors.push(x2);
+          }
+        }
+      }
+      return sourceTensors;
+    }
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/input_layer.js
+var InputLayer = class extends Layer {
+  constructor(args) {
+    super({
+      dtype: args.dtype,
+      name: args.name != null ? args.name : getUid("input").toString()
+    });
+    if (args.batchSize == null) {
+      args.batchSize = null;
+    }
+    if (args.sparse == null) {
+      args.sparse = false;
+    }
+    this.trainable = false;
+    this.built = true;
+    this.sparse = args.sparse;
+    if (args.inputShape != null && args.batchInputShape != null) {
+      throw new ValueError("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");
+    }
+    let batchInputShape = args.batchInputShape;
+    if (batchInputShape == null) {
+      if (args.inputShape == null) {
+        throw new ValueError("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");
+      } else {
+        batchInputShape = [args.batchSize].concat(args.inputShape);
+      }
+    } else {
+      if (args.batchSize != null) {
+        throw new ValueError("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");
+      }
+    }
+    const dtype = args.dtype || "float32";
+    this.batchInputShape = batchInputShape;
+    this.dtype = dtype;
+    this.inputSpec = [{ shape: batchInputShape }];
+    const inputTensor = new SymbolicTensor(this.dtype, this.batchInputShape, this, [], {}, this.name);
+    inputTensor.nodeIndex = 0;
+    inputTensor.tensorIndex = 0;
+    new Node({
+      outboundLayer: this,
+      inboundLayers: [],
+      nodeIndices: [],
+      tensorIndices: [],
+      inputTensors: [inputTensor],
+      outputTensors: [inputTensor],
+      inputMasks: [null],
+      outputMasks: [null],
+      inputShapes: [batchInputShape],
+      outputShapes: [batchInputShape]
+    });
+  }
+  apply(inputs, kwargs) {
+    throw new ValueError(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`);
+  }
+  dispose() {
+    return { refCountAfterDispose: this._refCount, numDisposedVariables: 0 };
+  }
+  getConfig() {
+    return {
+      batchInputShape: this.batchInputShape,
+      dtype: this.dtype,
+      sparse: this.sparse,
+      name: this.name
+    };
+  }
+};
+InputLayer.className = "InputLayer";
+serialization_exports.registerClass(InputLayer);
+function Input(config) {
+  if (config.batchShape == null && config.shape == null) {
+    throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");
+  }
+  if (config.batchShape != null && config.shape != null) {
+    throw new ValueError("Please provide either a `shape` or `batchShape` argument to Input, but not both.");
+  }
+  let batchShape = config.batchShape;
+  if (config.shape != null && batchShape == null) {
+    batchShape = [null].concat(config.shape);
+  }
+  let dtype = config.dtype;
+  if (dtype == null) {
+    dtype = "float32";
+  }
+  const inputLayer2 = new InputLayer({
+    batchInputShape: batchShape,
+    name: config.name,
+    dtype,
+    sparse: config.sparse
+  });
+  const outputs = inputLayer2.inboundNodes[0].outputTensors;
+  return outputs[0];
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/executor.js
+function assertFeedCompatibility(key, val) {
+  if (key.dtype == null || key.dtype === val.dtype) {
+    return val;
+  }
+  try {
+    return cast(val, key.dtype);
+  } catch (err) {
+    throw new ValueError(`The dtype of the feed (${val.dtype}) can not be cast to the dtype of the key '${key.name}' (${key.dtype}).`);
+  }
+}
+var FeedDict = class {
+  constructor(feeds) {
+    this.id2Value = {};
+    this.id2Mask = {};
+    this.name2Id = {};
+    if (feeds instanceof FeedDict) {
+      for (const id in feeds.id2Value) {
+        this.id2Value[id] = feeds.id2Value[id];
+        if (id in feeds.id2Mask) {
+          this.id2Mask[id] = feeds.id2Mask[id];
+        }
+      }
+    } else {
+      if (feeds == null) {
+        return;
+      }
+      for (const feed of feeds) {
+        this.add(feed.key, feed.value);
+      }
+    }
+  }
+  add(key, value, mask) {
+    if (this.id2Value[key.id] == null) {
+      this.id2Value[key.id] = assertFeedCompatibility(key, value);
+      this.name2Id[key.name] = key.id;
+      if (mask != null) {
+        this.id2Mask[key.id] = mask;
+      }
+    } else {
+      throw new ValueError(`Duplicate key: name=${key.name}, id=${key.id}`);
+    }
+    return this;
+  }
+  addFeed(feed) {
+    this.add(feed.key, feed.value);
+  }
+  hasKey(key) {
+    return this.id2Value[key.id] != null;
+  }
+  names() {
+    return Object.keys(this.name2Id);
+  }
+  getValue(key) {
+    if (key instanceof SymbolicTensor) {
+      if (this.id2Value[key.id] == null) {
+        throw new ValueError(`Nonexistent key: ${key.name}`);
+      } else {
+        return this.id2Value[key.id];
+      }
+    } else {
+      const id = this.name2Id[key];
+      if (id == null) {
+        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);
+      }
+      return this.id2Value[id];
+    }
+  }
+  getMask(key) {
+    if (key instanceof SymbolicTensor) {
+      if (this.id2Value[key.id] == null) {
+        throw new ValueError(`Nonexistent key: ${key.name}`);
+      } else {
+        return this.id2Mask[key.id];
+      }
+    } else {
+      const id = this.name2Id[key];
+      if (id == null) {
+        throw new ValueError(`Feed dict has no SymbolicTensor name: ${key}`);
+      }
+      return this.id2Mask[id];
+    }
+  }
+  disposeMasks() {
+    if (this.id2Mask != null) {
+      dispose(this.id2Mask);
+    }
+  }
+};
+var cachedSorted = new LruCache();
+var cachedRecipientCounts = new LruCache();
+function updateCacheMaxEntries(maxEntries) {
+  if (cachedSorted != null) {
+    cachedSorted.setMaxEntries(maxEntries);
+  }
+  if (cachedRecipientCounts != null) {
+    cachedRecipientCounts.setMaxEntries(maxEntries);
+  }
+}
+function execute(fetches, feedDict, kwargs, probe) {
+  const training = kwargs == null ? false : kwargs["training"];
+  const arrayFetches = Array.isArray(fetches);
+  const fetchArray = arrayFetches ? fetches : [fetches];
+  const outputNames = fetchArray.map((t) => t.name);
+  const finalOutputs = [];
+  const feedNames = feedDict.names();
+  for (const outputName of outputNames) {
+    if (feedNames.indexOf(outputName) !== -1) {
+      finalOutputs.push(feedDict.getValue(outputName));
+    } else {
+      finalOutputs.push(null);
+    }
+  }
+  if (probe != null) {
+    probe.maxNumTensors = -Infinity;
+    probe.minNumTensors = Infinity;
+  }
+  const fetchAndFeedKey = outputNames.join(",") + "|" + feedDict.names().sort().join(",");
+  let sorted = cachedSorted.get(fetchAndFeedKey);
+  let recipientCounts;
+  if (sorted == null) {
+    const out = getTopologicalSortAndRecipientCounts(fetchArray, feedDict);
+    sorted = out.sorted;
+    recipientCounts = out.recipientCounts;
+    cachedSorted.put(fetchAndFeedKey, sorted);
+    cachedRecipientCounts.put(fetchAndFeedKey, recipientCounts);
+  }
+  recipientCounts = {};
+  if (!training) {
+    Object.assign(recipientCounts, cachedRecipientCounts.get(fetchAndFeedKey));
+  }
+  const internalFeedDict = new FeedDict(feedDict);
+  for (let i = 0; i < sorted.length; ++i) {
+    if (probe != null) {
+      const numTensors = memory().numTensors;
+      if (numTensors > probe.maxNumTensors) {
+        probe.maxNumTensors = numTensors;
+      }
+      if (numTensors < probe.minNumTensors) {
+        probe.minNumTensors = numTensors;
+      }
+    }
+    const symbolic = sorted[i];
+    const srcLayer = symbolic.sourceLayer;
+    if (srcLayer instanceof InputLayer) {
+      continue;
+    }
+    const inputValues = [];
+    const inputMasks = [];
+    const tensorsToDispose = [];
+    let maskExists = false;
+    for (const input2 of symbolic.inputs) {
+      const value = internalFeedDict.getValue(input2);
+      const mask = internalFeedDict.getMask(input2);
+      inputValues.push(value);
+      inputMasks.push(mask);
+      if (mask != null) {
+        maskExists = true;
+      }
+      if (!training) {
+        recipientCounts[input2.name]--;
+        if (recipientCounts[input2.name] === 0 && !feedDict.hasKey(input2) && outputNames.indexOf(input2.name) === -1 && !value.isDisposed && input2.sourceLayer.stateful !== true) {
+          tensorsToDispose.push(value);
+        }
+      }
+    }
+    if (maskExists) {
+      kwargs = kwargs || {};
+      kwargs["mask"] = inputMasks[0];
+    }
+    const outputTensors = toList(srcLayer.apply(inputValues, kwargs));
+    let outputMask = null;
+    if (srcLayer.supportsMasking) {
+      outputMask = srcLayer.computeMask(inputValues, inputMasks);
+    }
+    const layerOutputs = getNodeOutputs(symbolic);
+    const outputSymbolicTensors = Array.isArray(layerOutputs) ? layerOutputs : [layerOutputs];
+    for (let i2 = 0; i2 < outputSymbolicTensors.length; ++i2) {
+      if (!internalFeedDict.hasKey(outputSymbolicTensors[i2])) {
+        internalFeedDict.add(outputSymbolicTensors[i2], outputTensors[i2], Array.isArray(outputMask) ? outputMask[0] : outputMask);
+      }
+      const index = outputNames.indexOf(outputSymbolicTensors[i2].name);
+      if (index !== -1) {
+        finalOutputs[index] = outputTensors[i2];
+      }
+    }
+    if (!training) {
+      dispose(tensorsToDispose);
+    }
+  }
+  internalFeedDict.disposeMasks();
+  return arrayFetches ? finalOutputs : finalOutputs[0];
+}
+function getTopologicalSortAndRecipientCounts(fetches, feedDict) {
+  util_exports.assert(fetches != null && fetches.length > 0, () => `Expected at least one fetch, got none`);
+  let finalSorted = [];
+  let finalRecipientMap = {};
+  if (fetches.length === 1) {
+    const out = getTopologicalSortAndRecipientCountsForOneFetch(fetches[0], feedDict);
+    finalSorted = out.sorted;
+    finalRecipientMap = out.recipientMap;
+  } else {
+    const visited = /* @__PURE__ */ new Set();
+    for (const fetch4 of fetches) {
+      const { sorted, recipientMap } = getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict);
+      for (const symbolicTensor of sorted) {
+        if (!visited.has(symbolicTensor.name)) {
+          finalSorted.push(symbolicTensor);
+          visited.add(symbolicTensor.name);
+        }
+      }
+      for (const name in recipientMap) {
+        if (finalRecipientMap[name] == null) {
+          finalRecipientMap[name] = /* @__PURE__ */ new Set();
+        }
+        recipientMap[name].forEach((recipient) => finalRecipientMap[name].add(recipient));
+      }
+    }
+  }
+  return {
+    sorted: finalSorted,
+    recipientCounts: recipientMap2Counts(finalRecipientMap)
+  };
+}
+function recipientMap2Counts(recipientMap) {
+  const recipientCounts = {};
+  for (const name in recipientMap) {
+    recipientCounts[name] = recipientMap[name].size;
+  }
+  return recipientCounts;
+}
+function getTopologicalSortAndRecipientCountsForOneFetch(fetch4, feedDict) {
+  const visited = /* @__PURE__ */ new Set();
+  const sorted = [];
+  const recipientMap = {};
+  for (const key of feedDict.names()) {
+    visited.add(key);
+  }
+  const stack2 = [];
+  const marks = [];
+  stack2.push(fetch4);
+  while (stack2.length > 0) {
+    const top = stack2[stack2.length - 1];
+    if (visited.has(top.name)) {
+      stack2.pop();
+      continue;
+    }
+    const topIsMarked = marks[marks.length - 1] === stack2.length - 1;
+    if (top.inputs.length === 0 || topIsMarked) {
+      stack2.pop();
+      sorted.push(top);
+      visited.add(top.name);
+      if (topIsMarked) {
+        marks.pop();
+      }
+    } else {
+      marks.push(stack2.length - 1);
+      for (const input2 of top.inputs) {
+        if (recipientMap[input2.name] == null) {
+          recipientMap[input2.name] = /* @__PURE__ */ new Set();
+        }
+        recipientMap[input2.name].add(top.name);
+        if (visited.has(input2.name)) {
+          continue;
+        }
+        stack2.push(input2);
+      }
+    }
+  }
+  return { sorted, recipientMap };
+}
+function getNodeOutputs(fetch4) {
+  let layerOutputs;
+  if (fetch4.sourceLayer.inboundNodes.length === 1) {
+    layerOutputs = fetch4.sourceLayer.output;
+  } else {
+    let nodeIndex = null;
+    for (let i = 0; i < fetch4.sourceLayer.inboundNodes.length; ++i) {
+      for (const outputTensor of fetch4.sourceLayer.inboundNodes[i].outputTensors) {
+        if (outputTensor.id === fetch4.id) {
+          nodeIndex = i;
+          break;
+        }
+      }
+    }
+    layerOutputs = fetch4.sourceLayer.getOutputAt(nodeIndex);
+  }
+  return layerOutputs;
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/flags_layers.js
+var ENV3 = env();
+ENV3.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES", () => 100, updateCacheMaxEntries);
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_constraints.js
+var exports_constraints_exports = {};
+__export(exports_constraints_exports, {
+  maxNorm: () => maxNorm,
+  minMaxNorm: () => minMaxNorm,
+  nonNeg: () => nonNeg,
+  unitNorm: () => unitNorm
+});
+
+// node_modules/@tensorflow/tfjs-layers/dist/constraints.js
+function calcL2Norms(w, axis) {
+  return tidy(() => sqrt(sum2(mul(w, w), axis, true)));
+}
+var Constraint = class extends serialization_exports.Serializable {
+  getConfig() {
+    return {};
+  }
+};
+var MaxNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultMaxValue = 2;
+    this.defaultAxis = 0;
+    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => {
+      const norms = calcL2Norms(w, this.axis);
+      const desired = clipByValue(norms, 0, this.maxValue);
+      return mul(w, div(desired, add2(epsilon(), norms)));
+    });
+  }
+  getConfig() {
+    return { maxValue: this.maxValue, axis: this.axis };
+  }
+};
+MaxNorm.className = "MaxNorm";
+serialization_exports.registerClass(MaxNorm);
+var UnitNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultAxis = 0;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => div(w, add2(epsilon(), calcL2Norms(w, this.axis))));
+  }
+  getConfig() {
+    return { axis: this.axis };
+  }
+};
+UnitNorm.className = "UnitNorm";
+serialization_exports.registerClass(UnitNorm);
+var NonNeg = class extends Constraint {
+  apply(w) {
+    return relu(w);
+  }
+};
+NonNeg.className = "NonNeg";
+serialization_exports.registerClass(NonNeg);
+var MinMaxNorm = class extends Constraint {
+  constructor(args) {
+    super();
+    this.defaultMinValue = 0;
+    this.defaultMaxValue = 1;
+    this.defaultRate = 1;
+    this.defaultAxis = 0;
+    this.minValue = args.minValue != null ? args.minValue : this.defaultMinValue;
+    this.maxValue = args.maxValue != null ? args.maxValue : this.defaultMaxValue;
+    this.rate = args.rate != null ? args.rate : this.defaultRate;
+    this.axis = args.axis != null ? args.axis : this.defaultAxis;
+  }
+  apply(w) {
+    return tidy(() => {
+      const norms = calcL2Norms(w, this.axis);
+      const desired = add2(mul(this.rate, clipByValue(norms, this.minValue, this.maxValue)), mul(1 - this.rate, norms));
+      return mul(w, div(desired, add2(epsilon(), norms)));
+    });
+  }
+  getConfig() {
+    return {
+      minValue: this.minValue,
+      maxValue: this.maxValue,
+      rate: this.rate,
+      axis: this.axis
+    };
+  }
+};
+MinMaxNorm.className = "MinMaxNorm";
+serialization_exports.registerClass(MinMaxNorm);
+var CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "maxNorm": "MaxNorm",
+  "minMaxNorm": "MinMaxNorm",
+  "nonNeg": "NonNeg",
+  "unitNorm": "UnitNorm"
+};
+function serializeConstraint(constraint) {
+  return serializeKerasObject(constraint);
+}
+function deserializeConstraint(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "constraint");
+}
+function getConstraint(identifier) {
+  if (identifier == null) {
+    return null;
+  }
+  if (typeof identifier === "string") {
+    const className = identifier in CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP ? CONSTRAINT_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    const config = { className, config: {} };
+    return deserializeConstraint(config);
+  } else if (identifier instanceof Constraint) {
+    return identifier;
+  } else {
+    return deserializeConstraint(identifier);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_constraints.js
+function maxNorm(args) {
+  return new MaxNorm(args);
+}
+function unitNorm(args) {
+  return new UnitNorm(args);
+}
+function nonNeg() {
+  return new NonNeg();
+}
+function minMaxNorm(config) {
+  return new MinMaxNorm(config);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_initializers.js
+var exports_initializers_exports = {};
+__export(exports_initializers_exports, {
+  constant: () => constant,
+  glorotNormal: () => glorotNormal,
+  glorotUniform: () => glorotUniform,
+  heNormal: () => heNormal,
+  heUniform: () => heUniform,
+  identity: () => identity,
+  leCunNormal: () => leCunNormal,
+  leCunUniform: () => leCunUniform,
+  ones: () => ones3,
+  orthogonal: () => orthogonal,
+  randomNormal: () => randomNormal3,
+  randomUniform: () => randomUniform2,
+  truncatedNormal: () => truncatedNormal2,
+  varianceScaling: () => varianceScaling,
+  zeros: () => zeros2
+});
+function zeros2() {
+  return new Zeros();
+}
+function ones3() {
+  return new Ones();
+}
+function constant(args) {
+  return new Constant(args);
+}
+function randomUniform2(args) {
+  return new RandomUniform(args);
+}
+function randomNormal3(args) {
+  return new RandomNormal(args);
+}
+function truncatedNormal2(args) {
+  return new TruncatedNormal(args);
+}
+function identity(args) {
+  return new Identity2(args);
+}
+function varianceScaling(config) {
+  return new VarianceScaling(config);
+}
+function glorotUniform(args) {
+  return new GlorotUniform(args);
+}
+function glorotNormal(args) {
+  return new GlorotNormal(args);
+}
+function heNormal(args) {
+  return new HeNormal(args);
+}
+function heUniform(args) {
+  return new HeUniform(args);
+}
+function leCunNormal(args) {
+  return new LeCunNormal(args);
+}
+function leCunUniform(args) {
+  return new LeCunUniform(args);
+}
+function orthogonal(args) {
+  return new Orthogonal(args);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_layers.js
+var exports_layers_exports = {};
+__export(exports_layers_exports, {
+  Layer: () => Layer,
+  RNN: () => RNN,
+  RNNCell: () => RNNCell,
+  activation: () => activation,
+  add: () => add3,
+  alphaDropout: () => alphaDropout,
+  average: () => average,
+  averagePooling1d: () => averagePooling1d,
+  averagePooling2d: () => averagePooling2d,
+  averagePooling3d: () => averagePooling3d,
+  avgPool1d: () => avgPool1d,
+  avgPool2d: () => avgPool2d,
+  avgPool3d: () => avgPool3d2,
+  avgPooling1d: () => avgPooling1d,
+  avgPooling2d: () => avgPooling2d,
+  avgPooling3d: () => avgPooling3d,
+  batchNormalization: () => batchNormalization2,
+  bidirectional: () => bidirectional,
+  categoryEncoding: () => categoryEncoding,
+  concatenate: () => concatenate2,
+  conv1d: () => conv1d2,
+  conv2d: () => conv2d3,
+  conv2dTranspose: () => conv2dTranspose2,
+  conv3d: () => conv3d2,
+  conv3dTranspose: () => conv3dTranspose2,
+  convLstm2d: () => convLstm2d,
+  convLstm2dCell: () => convLstm2dCell,
+  cropping2D: () => cropping2D,
+  dense: () => dense,
+  depthwiseConv2d: () => depthwiseConv2d4,
+  dot: () => dot3,
+  dropout: () => dropout3,
+  elu: () => elu3,
+  embedding: () => embedding,
+  flatten: () => flatten3,
+  gaussianDropout: () => gaussianDropout,
+  gaussianNoise: () => gaussianNoise,
+  globalAveragePooling1d: () => globalAveragePooling1d,
+  globalAveragePooling2d: () => globalAveragePooling2d,
+  globalMaxPool1d: () => globalMaxPool1d,
+  globalMaxPool2d: () => globalMaxPool2d,
+  globalMaxPooling1d: () => globalMaxPooling1d,
+  globalMaxPooling2d: () => globalMaxPooling2d,
+  gru: () => gru,
+  gruCell: () => gruCell,
+  input: () => input,
+  inputLayer: () => inputLayer,
+  layerNormalization: () => layerNormalization,
+  leakyReLU: () => leakyReLU,
+  lstm: () => lstm,
+  lstmCell: () => lstmCell,
+  masking: () => masking,
+  maxPool1d: () => maxPool1d,
+  maxPool2d: () => maxPool2d,
+  maxPooling1d: () => maxPooling1d,
+  maxPooling2d: () => maxPooling2d,
+  maxPooling3d: () => maxPooling3d,
+  maximum: () => maximum2,
+  minimum: () => minimum2,
+  multiply: () => multiply,
+  permute: () => permute,
+  prelu: () => prelu2,
+  reLU: () => reLU,
+  repeatVector: () => repeatVector,
+  rescaling: () => rescaling,
+  reshape: () => reshape2,
+  resizing: () => resizing,
+  rnn: () => rnn2,
+  separableConv2d: () => separableConv2d2,
+  simpleRNN: () => simpleRNN,
+  simpleRNNCell: () => simpleRNNCell,
+  softmax: () => softmax2,
+  spatialDropout1d: () => spatialDropout1d,
+  stackedRNNCells: () => stackedRNNCells,
+  thresholdedReLU: () => thresholdedReLU,
+  timeDistributed: () => timeDistributed,
+  upSampling2d: () => upSampling2d,
+  zeroPadding2d: () => zeroPadding2d
+});
+
+// node_modules/@tensorflow/tfjs-layers/dist/logs.js
+async function resolveScalarsInLogs(logs) {
+  if (logs == null) {
+    return;
+  }
+  const promises = [];
+  const keys = [];
+  const scalarsToDispose = [];
+  for (const key in logs) {
+    const value = logs[key];
+    if (typeof value !== "number") {
+      const valueScalar = value;
+      promises.push(valueScalar.data());
+      keys.push(key);
+      scalarsToDispose.push(valueScalar);
+    }
+  }
+  if (promises.length > 0) {
+    const values = await Promise.all(promises);
+    for (let i = 0; i < values.length; ++i) {
+      logs[keys[i]] = values[i][0];
+    }
+    dispose(scalarsToDispose);
+  }
+}
+function disposeTensorsInLogs(logs) {
+  if (logs == null) {
+    return;
+  }
+  for (const key in logs) {
+    const value = logs[key];
+    if (typeof value !== "number") {
+      value.dispose();
+    }
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/base_callbacks.js
+var ModelLoggingVerbosity;
+(function(ModelLoggingVerbosity2) {
+  ModelLoggingVerbosity2[ModelLoggingVerbosity2["SILENT"] = 0] = "SILENT";
+  ModelLoggingVerbosity2[ModelLoggingVerbosity2["VERBOSE"] = 1] = "VERBOSE";
+})(ModelLoggingVerbosity || (ModelLoggingVerbosity = {}));
+var DEFAULT_YIELD_EVERY_MS = 125;
+var BaseCallback = class {
+  constructor() {
+    this.validationData = null;
+  }
+  setParams(params) {
+    this.params = params;
+  }
+  async onEpochBegin(epoch, logs) {
+  }
+  async onEpochEnd(epoch, logs) {
+  }
+  async onBatchBegin(batch, logs) {
+  }
+  async onBatchEnd(batch, logs) {
+  }
+  async onTrainBegin(logs) {
+  }
+  async onTrainEnd(logs) {
+  }
+  setModel(model2) {
+  }
+};
+var CallbackList = class {
+  constructor(callbacks2, queueLength = 10) {
+    if (callbacks2 == null) {
+      callbacks2 = [];
+    }
+    this.callbacks = callbacks2;
+    this.queueLength = queueLength;
+  }
+  append(callback) {
+    this.callbacks.push(callback);
+  }
+  setParams(params) {
+    for (const callback of this.callbacks) {
+      callback.setParams(params);
+    }
+  }
+  setModel(model2) {
+    for (const callback of this.callbacks) {
+      callback.setModel(model2);
+    }
+  }
+  async onEpochBegin(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onEpochBegin(epoch, logs);
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onEpochEnd(epoch, logs);
+    }
+  }
+  async onBatchBegin(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onBatchBegin(batch, logs);
+    }
+  }
+  async onBatchEnd(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onBatchEnd(batch, logs);
+    }
+  }
+  async onTrainBegin(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onTrainBegin(logs);
+    }
+  }
+  async onTrainEnd(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    for (const callback of this.callbacks) {
+      await callback.onTrainEnd(logs);
+    }
+  }
+};
+var BaseLogger = class extends BaseCallback {
+  constructor() {
+    super();
+  }
+  async onEpochBegin(epoch) {
+    this.seen = 0;
+    this.totals = {};
+  }
+  async onBatchEnd(batch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    const batchSize = logs["size"] == null ? 0 : logs["size"];
+    this.seen += batchSize;
+    for (const key in logs) {
+      const value = logs[key];
+      if (typeof value === "number") {
+        if (!this.totals.hasOwnProperty(key)) {
+          this.totals[key] = 0;
+        }
+        this.totals[key] = this.totals[key] + value * batchSize;
+      } else {
+        let oldTotalsToDispose;
+        if (key in this.totals) {
+          oldTotalsToDispose = this.totals[key];
+        } else {
+          this.totals[key] = 0;
+        }
+        const total = tidy(() => add2(this.totals[key], mul(value, batchSize)));
+        this.totals[key] = total;
+        if (oldTotalsToDispose != null) {
+          oldTotalsToDispose.dispose();
+        }
+      }
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs != null) {
+      for (const key of this.params["metrics"]) {
+        if (this.totals[key] == null) {
+          continue;
+        }
+        if (typeof this.totals[key] === "number") {
+          logs[key] = this.totals[key] / this.seen;
+        } else {
+          tidy(() => {
+            const log5 = mul(div(1, this.seen), this.totals[key]);
+            logs[key] = log5;
+            this.totals[key].dispose();
+            keep(logs[key]);
+          });
+        }
+      }
+    }
+  }
+};
+var History = class extends BaseCallback {
+  async onTrainBegin(logs) {
+    this.epoch = [];
+    this.history = {};
+  }
+  async onEpochEnd(epoch, logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    this.epoch.push(epoch);
+    for (const key in logs) {
+      if (this.history[key] == null) {
+        this.history[key] = [];
+      }
+      this.history[key].push(logs[key]);
+    }
+  }
+  async syncData() {
+    const promises = [];
+    const keys = [];
+    const indices = [];
+    for (const key in this.history) {
+      const valueArray = this.history[key];
+      for (let i = 0; i < valueArray.length; ++i) {
+        if (typeof valueArray[i] !== "number") {
+          const valueScalar = valueArray[i];
+          promises.push(valueScalar.data());
+          keys.push(key);
+          indices.push(i);
+        }
+      }
+    }
+    const values = await Promise.all(promises);
+    for (let n = 0; n < values.length; ++n) {
+      const tensorToDispose = this.history[keys[n]][indices[n]];
+      tensorToDispose.dispose();
+      this.history[keys[n]][indices[n]] = values[n][0];
+    }
+  }
+};
+var CustomCallback = class extends BaseCallback {
+  constructor(args, yieldEvery) {
+    super();
+    this.currentEpoch = 0;
+    this.nowFunc = args.nowFunc;
+    this.nextFrameFunc = args.nextFrameFunc || nextFrame;
+    this.yieldEvery = yieldEvery || "auto";
+    if (this.yieldEvery === "auto") {
+      this.yieldEvery = DEFAULT_YIELD_EVERY_MS;
+    }
+    if (this.yieldEvery === "never" && args.onYield != null) {
+      throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");
+    }
+    if (util_exports.isNumber(this.yieldEvery)) {
+      this.maybeWait = debounce(this.maybeWait.bind(this), this.yieldEvery, this.nowFunc);
+    }
+    this.trainBegin = args.onTrainBegin;
+    this.trainEnd = args.onTrainEnd;
+    this.epochBegin = args.onEpochBegin;
+    this.epochEnd = args.onEpochEnd;
+    this.batchBegin = args.onBatchBegin;
+    this.batchEnd = args.onBatchEnd;
+    this.yield = args.onYield;
+  }
+  async maybeWait(epoch, batch, logs) {
+    const ps = [];
+    if (this.yield != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.yield(epoch, batch, logs));
+    }
+    ps.push(this.nextFrameFunc());
+    await Promise.all(ps);
+  }
+  async onEpochBegin(epoch, logs) {
+    this.currentEpoch = epoch;
+    if (this.epochBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.epochBegin(epoch, logs);
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    const ps = [];
+    if (this.epochEnd != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.epochEnd(epoch, logs));
+    }
+    if (this.yieldEvery === "epoch") {
+      ps.push(this.nextFrameFunc());
+    }
+    await Promise.all(ps);
+  }
+  async onBatchBegin(batch, logs) {
+    if (this.batchBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.batchBegin(batch, logs);
+    }
+  }
+  async onBatchEnd(batch, logs) {
+    const ps = [];
+    if (this.batchEnd != null) {
+      await resolveScalarsInLogs(logs);
+      ps.push(this.batchEnd(batch, logs));
+    }
+    if (this.yieldEvery === "batch") {
+      ps.push(this.nextFrameFunc());
+    } else if (util_exports.isNumber(this.yieldEvery)) {
+      ps.push(this.maybeWait(this.currentEpoch, batch, logs));
+    }
+    await Promise.all(ps);
+  }
+  async onTrainBegin(logs) {
+    if (this.trainBegin != null) {
+      await resolveScalarsInLogs(logs);
+      await this.trainBegin(logs);
+    }
+  }
+  async onTrainEnd(logs) {
+    if (this.trainEnd != null) {
+      await resolveScalarsInLogs(logs);
+      await this.trainEnd(logs);
+    }
+  }
+};
+function standardizeCallbacks(callbacks2, yieldEvery) {
+  if (callbacks2 == null) {
+    callbacks2 = {};
+  }
+  if (callbacks2 instanceof BaseCallback) {
+    return [callbacks2];
+  }
+  if (Array.isArray(callbacks2) && callbacks2[0] instanceof BaseCallback) {
+    return callbacks2;
+  }
+  const callbackConfigs = toList(callbacks2);
+  return callbackConfigs.map((callbackConfig) => new CustomCallback(callbackConfig, yieldEvery));
+}
+var CallbackConstructorRegistry = class {
+  constructor() {
+  }
+  static registerCallbackConstructor(verbosityLevel, callbackConstructor) {
+    util_exports.assert(verbosityLevel >= 0 && Number.isInteger(verbosityLevel), () => `Verbosity level is expected to be an integer >= 0, but got ${verbosityLevel}`);
+    CallbackConstructorRegistry.checkForDuplicate(callbackConstructor);
+    if (CallbackConstructorRegistry.constructors[verbosityLevel] == null) {
+      CallbackConstructorRegistry.constructors[verbosityLevel] = [];
+    }
+    CallbackConstructorRegistry.constructors[verbosityLevel].push(callbackConstructor);
+  }
+  static checkForDuplicate(callbackConstructor) {
+    for (const levelName in CallbackConstructorRegistry.constructors) {
+      const constructors = CallbackConstructorRegistry.constructors[+levelName];
+      constructors.forEach((ctor) => {
+        if (ctor === callbackConstructor) {
+          throw new ValueError("Duplicate callback constructor.");
+        }
+      });
+    }
+  }
+  static clear() {
+    CallbackConstructorRegistry.constructors = {};
+  }
+  static createCallbacks(verbosityLevel) {
+    const constructors = [];
+    for (const levelName in CallbackConstructorRegistry.constructors) {
+      const level = +levelName;
+      if (verbosityLevel >= level) {
+        constructors.push(...CallbackConstructorRegistry.constructors[level]);
+      }
+    }
+    return constructors.map((ctor) => new ctor());
+  }
+};
+CallbackConstructorRegistry.constructors = {};
+function configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics) {
+  const history = new History();
+  const actualCallbacks = [
+    new BaseLogger(),
+    ...CallbackConstructorRegistry.createCallbacks(verbose)
+  ];
+  if (callbacks2 != null) {
+    actualCallbacks.push(...callbacks2);
+  }
+  actualCallbacks.push(history);
+  const callbackList = new CallbackList(actualCallbacks);
+  callbackList.setParams({
+    epochs,
+    initialEpoch,
+    samples: numTrainSamples,
+    steps: stepsPerEpoch,
+    batchSize,
+    verbose,
+    doValidation,
+    metrics: callbackMetrics
+  });
+  return { callbackList, history };
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/serialization.js
+function deserialize(config, customObjects = {}, fastWeightInit = false) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "layer", fastWeightInit);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/losses.js
+function l2Normalize(x, axis) {
+  return tidy(() => {
+    if (x.dtype !== "float32") {
+      x = cast(x, "float32");
+    }
+    const squareSum = sum2(square2(x), axis, true);
+    const epsilonTensor = fill(squareSum.shape, epsilon());
+    const norm2 = sqrt(maximum(squareSum, epsilonTensor));
+    return div(x, norm2);
+  });
+}
+function meanSquaredError2(yTrue, yPred) {
+  return tidy(() => mean(square2(sub(yPred, yTrue)), -1));
+}
+function meanAbsoluteError(yTrue, yPred) {
+  return tidy(() => mean(abs(sub(yPred, yTrue)), -1));
+}
+function meanAbsolutePercentageError(yTrue, yPred) {
+  return tidy(() => {
+    const diff = sub(yTrue, yPred);
+    const clippedTrue = clipByValue(abs(yTrue), epsilon(), Number.MAX_VALUE);
+    const absResult = abs(div(diff, clippedTrue));
+    return mul(100, mean(absResult, -1));
+  });
+}
+function meanSquaredLogarithmicError(yTrue, yPred) {
+  return tidy(() => {
+    const clippedPred = clipByValue(yPred, epsilon(), Number.MAX_VALUE);
+    const firstLog = log2(add2(1, clippedPred));
+    const clippedTrue = clipByValue(yTrue, epsilon(), Number.MAX_VALUE);
+    const secondLog = log2(add2(1, clippedTrue));
+    return mean(square2(sub(firstLog, secondLog)), -1);
+  });
+}
+function squaredHinge(yTrue, yPred) {
+  return tidy(() => {
+    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));
+    return mean(square2(maxResult), -1);
+  });
+}
+function hinge(yTrue, yPred) {
+  return tidy(() => {
+    const maxResult = maximum(0, sub(1, mul(yTrue, yPred)));
+    return mean(maxResult, -1);
+  });
+}
+function categoricalHinge(yTrue, yPred) {
+  return tidy(() => {
+    const pos = sum2(mul(yTrue, yPred), -1);
+    const neg4 = max(mul(sub(1, yTrue), yPred), -1);
+    return maximum(0, add2(1, sub(neg4, pos)));
+  });
+}
+function logcosh(yTrue, yPred) {
+  return tidy(() => {
+    const log22 = Math.log(2);
+    const predictionDiff = sub(yPred, yTrue);
+    const logcoshResult = sub(add2(predictionDiff, softplus(mul(-2, predictionDiff))), log22);
+    return mean(logcoshResult, -1);
+  });
+}
+function categoricalCrossentropy(target, output, fromLogits = false) {
+  return tidy(() => {
+    if (fromLogits) {
+      output = softmax(output);
+    } else {
+      const outputSum = sum2(output, output.shape.length - 1, true);
+      output = div(output, outputSum);
+    }
+    output = clipByValue(output, epsilon(), 1 - epsilon());
+    return neg(sum2(mul(cast(target, "float32"), log2(output)), output.shape.length - 1));
+  });
+}
+function sparseCategoricalCrossentropy(target, output, fromLogits = false) {
+  return tidy(() => {
+    const flatTarget = cast(floor(flatten2(target)), "int32");
+    output = clipByValue(output, epsilon(), 1 - epsilon());
+    const outputShape = output.shape;
+    const oneHotTarget = reshape(oneHot(flatTarget, outputShape[outputShape.length - 1]), outputShape);
+    return categoricalCrossentropy(oneHotTarget, output, fromLogits);
+  });
+}
+function sigmoidCrossEntropyWithLogits(labels, logits) {
+  if (!util_exports.arraysEqual(labels.shape, logits.shape)) {
+    throw new ValueError(`logits and labels must have the same shape, but got shapes ${JSON.stringify(labels.shape)} and ${JSON.stringify(logits.shape)}`);
+  }
+  return tidy(() => {
+    const reluLogits = relu(logits);
+    const negAbsLogits = neg(abs(logits));
+    return add2(sub(reluLogits, mul(logits, labels)), log1p(exp(negAbsLogits)));
+  });
+}
+function binaryCrossentropy(yTrue, yPred) {
+  return tidy(() => {
+    let y;
+    y = clipByValue(yPred, epsilon(), 1 - epsilon());
+    y = log2(div(y, sub(1, y)));
+    return mean(sigmoidCrossEntropyWithLogits(yTrue, y), -1);
+  });
+}
+function kullbackLeiblerDivergence(yTrue, yPred) {
+  return tidy(() => {
+    const clippedTrue = clipByValue(yTrue, epsilon(), 1);
+    const clippedPred = clipByValue(yPred, epsilon(), 1);
+    return sum2(mul(yTrue, log2(div(clippedTrue, clippedPred))), -1);
+  });
+}
+function poisson(yTrue, yPred) {
+  return tidy(() => {
+    const logPred = log2(add2(epsilon(), yPred));
+    return mean(sub(yPred, mul(yTrue, logPred)), -1);
+  });
+}
+function cosineProximity(yTrue, yPred) {
+  return tidy(() => {
+    const trueNormalized = l2Normalize(yTrue, -1);
+    const predNormalized = l2Normalize(yPred, -1);
+    const trueXPred = mul(trueNormalized, predNormalized);
+    return neg(sum2(trueXPred, -1));
+  });
+}
+var lossesMap = {
+  meanSquaredError: meanSquaredError2,
+  meanAbsoluteError,
+  meanAbsolutePercentageError,
+  meanSquaredLogarithmicError,
+  squaredHinge,
+  hinge,
+  categoricalHinge,
+  logcosh,
+  categoricalCrossentropy,
+  sparseCategoricalCrossentropy,
+  binaryCrossentropy,
+  kullbackLeiblerDivergence,
+  poisson,
+  cosineProximity
+};
+function get(identifierOrFn) {
+  if (typeof identifierOrFn === "string") {
+    if (identifierOrFn in lossesMap) {
+      return lossesMap[identifierOrFn];
+    }
+    let errMsg = `Unknown loss ${identifierOrFn}`;
+    if (identifierOrFn.toLowerCase().includes("softmaxcrossentropy")) {
+      errMsg = `Unknown loss ${identifierOrFn}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`;
+    }
+    throw new ValueError(errMsg);
+  } else {
+    return identifierOrFn;
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/metrics.js
+function binaryAccuracy(yTrue, yPred) {
+  return tidy(() => {
+    const threshold3 = mul(0.5, onesLike(yPred));
+    const yPredThresholded = cast2(greater(yPred, threshold3), yTrue.dtype);
+    return mean(equal(yTrue, yPredThresholded), -1);
+  });
+}
+function categoricalAccuracy(yTrue, yPred) {
+  return tidy(() => cast2(equal(argMax(yTrue, -1), argMax(yPred, -1)), "float32"));
+}
+function truePositives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 1))), "float32");
+  });
+}
+function falseNegatives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 1), equal(yPred, 0))), "float32");
+  });
+}
+function falsePositives(yTrue, yPred) {
+  return tidy(() => {
+    return cast(sum2(logicalAnd(equal(yTrue, 0), equal(yPred, 1))), "float32");
+  });
+}
+function precision(yTrue, yPred) {
+  return tidy(() => {
+    const tp = truePositives(yTrue, yPred);
+    const fp = falsePositives(yTrue, yPred);
+    const denominator = add2(tp, fp);
+    return cast(where(greater(denominator, 0), div(tp, denominator), 0), "float32");
+  });
+}
+function recall(yTrue, yPred) {
+  return tidy(() => {
+    const tp = truePositives(yTrue, yPred);
+    const fn = falseNegatives(yTrue, yPred);
+    const denominator = add2(tp, fn);
+    return cast(where(greater(denominator, 0), div(tp, denominator), 0), "float32");
+  });
+}
+function binaryCrossentropy2(yTrue, yPred) {
+  return binaryCrossentropy(yTrue, yPred);
+}
+function sparseCategoricalAccuracy(yTrue, yPred) {
+  if (yTrue.rank === yPred.rank) {
+    yTrue = squeeze(yTrue, [yTrue.rank - 1]);
+  }
+  yPred = argMax(yPred, -1);
+  if (yPred.dtype !== yTrue.dtype) {
+    yPred = cast(yPred, yTrue.dtype);
+  }
+  return cast(equal(yTrue, yPred), "float32");
+}
+var mse = meanSquaredError2;
+var MSE = meanSquaredError2;
+var mae = meanAbsoluteError;
+var MAE = meanAbsoluteError;
+var mape = meanAbsolutePercentageError;
+var MAPE = meanAbsolutePercentageError;
+var categoricalCrossentropy2 = categoricalCrossentropy;
+var cosine = cosineProximity;
+var sparseCategoricalCrossentropy2 = sparseCategoricalCrossentropy;
+var metricsMap = {
+  binaryAccuracy,
+  categoricalAccuracy,
+  precision,
+  categoricalCrossentropy: categoricalCrossentropy2,
+  sparseCategoricalCrossentropy: sparseCategoricalCrossentropy2,
+  mse,
+  MSE,
+  mae,
+  MAE,
+  mape,
+  MAPE,
+  cosine
+};
+function get2(identifier) {
+  if (typeof identifier === "string" && identifier in metricsMap) {
+    return metricsMap[identifier];
+  } else if (typeof identifier !== "string" && identifier != null) {
+    return identifier;
+  } else {
+    throw new ValueError(`Unknown metric ${identifier}`);
+  }
+}
+function getLossOrMetricName(fn) {
+  assert2(fn !== null, `Unknown LossOrMetricFn ${fn}`);
+  if (typeof fn === "string") {
+    return fn;
+  } else {
+    let fnName;
+    for (const key of Object.keys(lossesMap)) {
+      if (lossesMap[key] === fn) {
+        fnName = key;
+        break;
+      }
+    }
+    if (fnName !== void 0) {
+      return fnName;
+    }
+    for (const key of Object.keys(metricsMap)) {
+      if (metricsMap[key] === fn) {
+        fnName = key;
+        break;
+      }
+    }
+    if (fnName !== void 0) {
+      return fnName;
+    }
+    return fn.name;
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/optimizers.js
+function getOptimizer(identifier) {
+  const optimizerMap = {
+    "Adagrad": () => train.adagrad(0.01),
+    "Adadelta": () => train.adadelta(1, 0.95, epsilon()),
+    "Adam": () => train.adam(1e-3, 0.9, 0.999, epsilon()),
+    "Adamax": () => train.adamax(2e-3, 0.9, 0.999, epsilon(), 0),
+    "RMSProp": () => train.rmsprop(1e-3, 0.9, 0, epsilon()),
+    "SGD": () => train.sgd(0.01)
+  };
+  optimizerMap["adagrad"] = optimizerMap["Adagrad"];
+  optimizerMap["adadelta"] = optimizerMap["Adadelta"];
+  optimizerMap["adam"] = optimizerMap["Adam"];
+  optimizerMap["adamax"] = optimizerMap["Adamax"];
+  optimizerMap["rmsprop"] = optimizerMap["RMSProp"];
+  optimizerMap["sgd"] = optimizerMap["SGD"];
+  if (identifier in optimizerMap) {
+    return optimizerMap[identifier]();
+  }
+  throw new ValueError(`Unknown Optimizer ${identifier}`);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/user_defined_metadata.js
+var MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH = 1 * 1024 * 1024;
+function checkUserDefinedMetadata(userDefinedMetadata, modelName, checkSize = false) {
+  if (userDefinedMetadata == null || typeof userDefinedMetadata !== "object" || Object.getPrototypeOf(userDefinedMetadata) !== Object.prototype || !plainObjectCheck(userDefinedMetadata)) {
+    throw new Error("User-defined metadata is expected to be a JSON object, but is not.");
+  }
+  if (checkSize) {
+    const out = JSON.stringify(userDefinedMetadata);
+    if (out.length > MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH) {
+      console.warn(`User-defined metadata of model "${modelName}" is too large in size (length=${out.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${MAX_USER_DEFINED_METADATA_SERIALIZED_LENGTH}.`);
+    }
+  }
+}
+function plainObjectCheck(x) {
+  if (x === null) {
+    return true;
+  } else if (typeof x === "object") {
+    if (Object.getPrototypeOf(x) === Object.prototype) {
+      const keys = Object.keys(x);
+      for (const key of keys) {
+        if (typeof key !== "string") {
+          return false;
+        }
+        if (!plainObjectCheck(x[key])) {
+          return false;
+        }
+      }
+      return true;
+    } else {
+      if (Array.isArray(x)) {
+        for (const item of x) {
+          if (!plainObjectCheck(item)) {
+            return false;
+          }
+        }
+        return true;
+      } else {
+        return false;
+      }
+    }
+  } else {
+    const xType = typeof x;
+    return xType === "string" || xType === "number" || xType === "boolean";
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/layer_utils.js
+function printSummary(model2, lineLength, positions, printFn = console.log) {
+  const sequentialLike = isModelSequentialLike(model2);
+  const toDisplay = ["Layer (type)", "Input Shape", "Output shape", "Param #"];
+  if (sequentialLike) {
+    lineLength = lineLength || 90;
+    positions = positions || [0.32, 0.61, 0.89, 1];
+  } else {
+    lineLength = lineLength || 115;
+    positions = positions || [0.24, 0.48, 0.7, 0.8, 1];
+  }
+  if (positions[positions.length - 1] <= 1) {
+    positions = positions.map((p2) => Math.floor(lineLength * p2));
+  }
+  let relevantNodes;
+  if (!sequentialLike) {
+    toDisplay.push("Receives inputs");
+    relevantNodes = [];
+    for (const depth in model2.nodesByDepth) {
+      relevantNodes.push(...model2.nodesByDepth[depth]);
+    }
+  }
+  printFn("_".repeat(lineLength));
+  printRow(toDisplay, positions, printFn);
+  printFn("=".repeat(lineLength));
+  const layers = model2.layers;
+  for (let i = 0; i < layers.length; ++i) {
+    if (sequentialLike) {
+      printLayerSummary(layers[i], positions, printFn);
+    } else {
+      printLayerSummaryWithConnections(layers[i], positions, relevantNodes, printFn);
+    }
+    printFn((i === layers.length - 1 ? "=" : "_").repeat(lineLength));
+  }
+  model2.checkTrainableWeightsConsistency();
+  const trainableCount = countTrainableParams(model2);
+  const nonTrainableCount = countParamsInWeights(model2.nonTrainableWeights);
+  printFn(`Total params: ${trainableCount + nonTrainableCount}`);
+  printFn(`Trainable params: ${trainableCount}`);
+  printFn(`Non-trainable params: ${nonTrainableCount}`);
+  printFn("_".repeat(lineLength));
+}
+function countTrainableParams(model2) {
+  let trainableCount;
+  if (model2.collectedTrainableWeights != null) {
+    trainableCount = countParamsInWeights(model2.collectedTrainableWeights);
+  } else {
+    trainableCount = countParamsInWeights(model2.trainableWeights);
+  }
+  return trainableCount;
+}
+function isModelSequentialLike(model2) {
+  let sequentialLike = true;
+  const nodesByDepth = [];
+  const nodes = [];
+  for (const depth in model2.nodesByDepth) {
+    nodesByDepth.push(model2.nodesByDepth[depth]);
+  }
+  for (const depthNodes of nodesByDepth) {
+    if (depthNodes.length > 1 || depthNodes.length === 1 && depthNodes[0].inboundLayers.length > 1) {
+      sequentialLike = false;
+      break;
+    }
+    nodes.push(...depthNodes);
+  }
+  if (sequentialLike) {
+    for (const layer of model2.layers) {
+      let flag = false;
+      for (const node of layer.inboundNodes) {
+        if (nodes.indexOf(node) !== -1) {
+          if (flag) {
+            sequentialLike = false;
+            break;
+          } else {
+            flag = true;
+          }
+        }
+      }
+      if (!sequentialLike) {
+        break;
+      }
+    }
+  }
+  return sequentialLike;
+}
+function printRow(fields, positions, printFn = console.log) {
+  let line = "";
+  for (let i = 0; i < fields.length; ++i) {
+    if (i > 0) {
+      line = line.slice(0, line.length - 1) + " ";
+    }
+    line += fields[i];
+    line = line.slice(0, positions[i]);
+    line += " ".repeat(positions[i] - line.length);
+  }
+  printFn(line);
+}
+function printLayerSummary(layer, positions, printFn) {
+  let outputShape;
+  let inputShape;
+  try {
+    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(",");
+  } catch (err) {
+    inputShape = "multiple";
+  }
+  try {
+    outputShape = JSON.stringify(layer.outputShape);
+  } catch (err) {
+    outputShape = "multiple";
+  }
+  const name = layer.name;
+  const className = layer.getClassName();
+  const fields = [
+    `${name} (${className})`,
+    inputShape,
+    outputShape,
+    layer.countParams().toString()
+  ];
+  printRow(fields, positions, printFn);
+}
+function printLayerSummaryWithConnections(layer, positions, relevantNodes, printFn) {
+  let outputShape;
+  let inputShape;
+  try {
+    inputShape = layer.inboundNodes.map((x) => JSON.stringify(x.inputShapes)).join(",");
+  } catch (err) {
+    inputShape = "multiple";
+  }
+  try {
+    outputShape = JSON.stringify(layer.outputShape);
+  } catch (err) {
+    outputShape = "multiple";
+  }
+  const connections = [];
+  for (const node of layer.inboundNodes) {
+    if (relevantNodes != null && relevantNodes.length > 0 && relevantNodes.indexOf(node) === -1) {
+      continue;
+    }
+    for (let i = 0; i < node.inboundLayers.length; ++i) {
+      const inboundLayer = node.inboundLayers[i].name;
+      const inboundLayerIndex = node.nodeIndices[i];
+      const inboundTensorIndex = node.tensorIndices[i];
+      connections.push(`${inboundLayer}[${inboundLayerIndex}][${inboundTensorIndex}]`);
+    }
+  }
+  const name = layer.name;
+  const className = layer.getClassName();
+  const firstConnection = connections.length === 0 ? "" : connections[0];
+  const fields = [
+    `${name} (${className})`,
+    inputShape,
+    outputShape,
+    layer.countParams().toString(),
+    firstConnection
+  ];
+  printRow(fields, positions, printFn);
+  for (let i = 1; i < connections.length; ++i) {
+    printRow(["", "", "", "", connections[i]], positions, printFn);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/serialization_utils.js
+function isArrayItemInputOrOutputName(key, index, value) {
+  return (key === "inboundNodes" || key === "outputLayers" || key === "inputLayers") && index === 0 && typeof value === "string";
+}
+function convertPythonicToTs(pythonicConfig, key) {
+  if (pythonicConfig === null) {
+    return null;
+  } else if (typeof pythonicConfig === "string") {
+    return toCamelCase(pythonicConfig);
+  } else if (typeof pythonicConfig === "number" || typeof pythonicConfig === "boolean") {
+    return pythonicConfig;
+  } else if (pythonicConfig instanceof Array) {
+    const tsArray = [];
+    const arrayLength = pythonicConfig.length;
+    for (let i = 0; i < arrayLength; ++i) {
+      const item = pythonicConfig[i];
+      if (isArrayItemInputOrOutputName(key, i, item)) {
+        tsArray.push(item);
+      } else {
+        tsArray.push(convertPythonicToTs(item, key));
+      }
+    }
+    return tsArray;
+  } else {
+    const tsDict = {};
+    for (const pythonicKey of Object.keys(pythonicConfig)) {
+      const pythonicValue = pythonicConfig[pythonicKey];
+      if (pythonicKey === "name" && typeof pythonicValue === "string") {
+        tsDict[pythonicKey] = pythonicValue;
+      } else {
+        const tsKey = toCamelCase(pythonicKey);
+        tsDict[tsKey] = convertPythonicToTs(pythonicValue, tsKey);
+      }
+    }
+    return tsDict;
+  }
+}
+function convertTsToPythonic(tsConfig, key) {
+  if (tsConfig === null || tsConfig === void 0) {
+    return null;
+  } else if (typeof tsConfig === "string") {
+    return toSnakeCase(tsConfig);
+  } else if (typeof tsConfig === "number" || typeof tsConfig === "boolean") {
+    return tsConfig;
+  } else if (tsConfig instanceof Array) {
+    const pyArray = [];
+    const arrayLength = tsConfig.length;
+    for (let i = 0; i < arrayLength; ++i) {
+      const item = tsConfig[i];
+      if (isArrayItemInputOrOutputName(key, i, item)) {
+        pyArray.push(item);
+      } else {
+        pyArray.push(convertTsToPythonic(item, key));
+      }
+    }
+    return pyArray;
+  } else {
+    const pyDict = {};
+    for (const tsKey of Object.keys(tsConfig)) {
+      const tsValue = tsConfig[tsKey];
+      const pyKey = toSnakeCase(tsKey);
+      if ((tsKey === "name" || tsKey === "className") && typeof tsValue === "string") {
+        pyDict[pyKey] = tsValue;
+      } else {
+        pyDict[pyKey] = convertTsToPythonic(tsValue, tsKey);
+      }
+    }
+    return pyDict;
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/version.js
+var version2 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/container.js
+var Container = class extends Layer {
+  constructor(args) {
+    super({});
+    this.containerNodes = /* @__PURE__ */ new Set();
+    this.name = args.name;
+    if (this.name == null) {
+      const prefix = this.getClassName().toLowerCase();
+      this.name = getUid(prefix);
+    }
+    this.supportsMasking = false;
+    this.trainable_ = true;
+    if (Array.isArray(args.inputs)) {
+      this.inputs = args.inputs.slice();
+    } else {
+      this.inputs = [args.inputs];
+    }
+    if (Array.isArray(args.outputs)) {
+      this.outputs = args.outputs.slice();
+    } else {
+      this.outputs = [args.outputs];
+    }
+    if (unique2(this.inputs).length !== this.inputs.length) {
+      throw new ValueError(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map((x) => x.name)}`);
+    }
+    if (unique2(this.outputs).length !== this.outputs.length) {
+      console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map((x) => x.name)}`);
+    }
+    this.inputLayers = [];
+    this.inputLayersNodeIndices = [];
+    this.inputLayersTensorIndices = [];
+    this.outputLayers = [];
+    this.outputLayersNodeIndices = [];
+    this.outputLayersTensorIndices = [];
+    this.layers = [];
+    this.internalContainerRefs = [];
+    for (const x of this.outputs) {
+      const layer = x.sourceLayer;
+      const nodeIndex = x.nodeIndex;
+      const tensorIndex = x.tensorIndex;
+      this.outputLayers.push(layer);
+      this.outputLayersNodeIndices.push(nodeIndex);
+      this.outputLayersTensorIndices.push(tensorIndex);
+    }
+    for (const x of this.inputs) {
+      const layer = x.sourceLayer;
+      const nodeIndex = x.nodeIndex;
+      const tensorIndex = x.tensorIndex;
+      assert2(nodeIndex === 0, "input layer has >1 nodes");
+      assert2(tensorIndex === 0, "input layer has >1 tensors");
+      this.inputLayers.push(layer);
+      this.inputLayersNodeIndices.push(nodeIndex);
+      this.inputLayersTensorIndices.push(tensorIndex);
+    }
+    this.inputNames = [];
+    this.outputNames = [];
+    this.feedInputShapes = [];
+    this.feedInputNames = [];
+    this.feedOutputNames = [];
+    for (let i = 0; i < this.inputLayers.length; i++) {
+      const layer = this.inputLayers[i];
+      if (!(layer instanceof InputLayer)) {
+        throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${args.inputs}. Input ${i} (0-based) originates from layer type ${layer.getClassName()}.`);
+      }
+      this.inputNames.push(layer.name);
+      this.feedInputShapes.push(layer.batchInputShape);
+      this.feedInputNames.push(layer.name);
+    }
+    for (const layer of this.outputLayers) {
+      this.outputNames.push(layer.name);
+    }
+    this.internalInputShapes = this.inputs.map((x) => x.shape);
+    this.internalOutputShapes = this.outputs.map((x) => x.shape);
+    const nodesDepths = {};
+    const nodeIDToNode = {};
+    const layersDepths = {};
+    const layerIDToLayer = {};
+    const layerIndices = {};
+    const nodesInDecreasingDepth = [];
+    const buildMapOfGraph = (tensor2, finishedNodes2, nodesInProgress2, layer, nodeIndex, tensorIndex) => {
+      if (layer == null || nodeIndex == null || tensorIndex == null) {
+        layer = tensor2.sourceLayer;
+        nodeIndex = tensor2.nodeIndex;
+        tensorIndex = tensor2.tensorIndex;
+      }
+      const node = layer.inboundNodes[nodeIndex];
+      if (nodesInProgress2.indexOf(node) !== -1) {
+        throw new RuntimeError(`The tensor ${tensor2.name} at layer "${layer.name}" is part of a cycle.`);
+      }
+      if (finishedNodes2.indexOf(node) !== -1) {
+        return;
+      }
+      this.containerNodes.add(Container.nodeKey(layer, nodeIndex));
+      if (!(layer.id in layerIndices)) {
+        layerIndices[layer.id] = Object.keys(layerIndices).length;
+      }
+      if (nodesInProgress2.indexOf(node) === -1) {
+        nodesInProgress2.push(node);
+      }
+      const numInboundLayers = node.inboundLayers.length;
+      for (let i = 0; i < numInboundLayers; i++) {
+        const x = node.inputTensors[i];
+        const layer2 = node.inboundLayers[i];
+        const nodeIndex2 = node.nodeIndices[i];
+        const tensorIndex2 = node.tensorIndices[i];
+        buildMapOfGraph(x, finishedNodes2, nodesInProgress2, layer2, nodeIndex2, tensorIndex2);
+      }
+      finishedNodes2.push(node);
+      while (nodesInProgress2.indexOf(node) >= 0) {
+        nodesInProgress2.splice(nodesInProgress2.indexOf(node), 1);
+      }
+      nodesInDecreasingDepth.push(node);
+    };
+    const finishedNodes = [];
+    const nodesInProgress = [];
+    for (const x of this.outputs) {
+      buildMapOfGraph(x, finishedNodes, nodesInProgress);
+    }
+    const reversedNodesInDecreasingDepth = nodesInDecreasingDepth.slice().reverse();
+    for (const node of reversedNodesInDecreasingDepth) {
+      nodeIDToNode[node.id] = node;
+      if (!(node.id in nodesDepths)) {
+        nodesDepths[node.id] = 0;
+      }
+      let depth = nodesDepths[node.id];
+      const previousDepth = layersDepths[node.outboundLayer.id] == null ? 0 : layersDepths[node.outboundLayer.id];
+      depth = Math.max(depth, previousDepth);
+      layersDepths[node.outboundLayer.id] = depth;
+      layerIDToLayer[node.outboundLayer.id] = node.outboundLayer;
+      nodesDepths[node.id] = depth;
+      for (let i = 0; i < node.inboundLayers.length; i++) {
+        const inboundLayer = node.inboundLayers[i];
+        const nodeIndex = node.nodeIndices[i];
+        const inboundNode = inboundLayer.inboundNodes[nodeIndex];
+        const previousDepth2 = nodesDepths[inboundNode.id] == null ? 0 : nodesDepths[inboundNode.id];
+        nodesDepths[inboundNode.id] = Math.max(depth + 1, previousDepth2);
+        nodeIDToNode[inboundNode.id] = inboundNode;
+      }
+    }
+    const nodesByDepth = {};
+    for (const nodeID in nodesDepths) {
+      const depth = nodesDepths[nodeID];
+      if (!(depth in nodesByDepth)) {
+        nodesByDepth[depth] = [];
+      }
+      nodesByDepth[depth].push(nodeIDToNode[nodeID]);
+    }
+    const layersByDepth = {};
+    for (const layerID in layersDepths) {
+      const depth = layersDepths[layerID];
+      if (!(depth in layersByDepth)) {
+        layersByDepth[depth] = [];
+      }
+      layersByDepth[depth].push(layerIDToLayer[layerID]);
+    }
+    let depthKeys = Object.keys(layersByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    this.layers = [];
+    for (const depth of depthKeys) {
+      const layersForDepth = layersByDepth[depth];
+      layersForDepth.sort((a, b) => {
+        const aIndex = layerIndices[a.id];
+        const bIndex = layerIndices[b.id];
+        if (aIndex < bIndex) {
+          return -1;
+        }
+        if (aIndex > bIndex) {
+          return 1;
+        }
+        return 0;
+      });
+      for (const layer of layersForDepth) {
+        if (layer instanceof Container) {
+          this.internalContainerRefs.push(layer);
+        }
+        this.layers.push(layer);
+      }
+    }
+    this.layersByDepth = layersByDepth;
+    depthKeys = Object.keys(nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    const computableTensors = this.inputs.slice();
+    const layersWithCompleteInput = [];
+    for (const depth of depthKeys) {
+      for (const node of nodesByDepth[depth]) {
+        const layer = node.outboundLayer;
+        if (layer != null) {
+          for (const x of node.inputTensors) {
+            if (computableTensors.indexOf(x) === -1) {
+              throw new RuntimeError(`Graph disconnected: cannot obtain value for tensor ${x} at layer "${layer.name}". The following previous layers were accessed without issue: ${layersWithCompleteInput}`);
+            }
+          }
+          for (const x of node.outputTensors) {
+            computableTensors.push(x);
+          }
+          layersWithCompleteInput.push(layer.name);
+        }
+      }
+    }
+    this.nodesByDepth = nodesByDepth;
+    const allNames = this.layers.map((x) => x.name);
+    for (const name of allNames) {
+      const numOccurrences = allNames.filter((x) => x === name).length;
+      if (numOccurrences !== 1) {
+        throw new RuntimeError(`The name "${name}" is used ${numOccurrences} times in the model. All layer names should be unique. Layer names: ` + JSON.stringify(allNames));
+      }
+    }
+    this.outboundNodes = [];
+    this.inboundNodes = [];
+    new Node({
+      outboundLayer: this,
+      inboundLayers: [],
+      nodeIndices: [],
+      tensorIndices: [],
+      inputTensors: this.inputs,
+      outputTensors: this.outputs,
+      inputMasks: this.inputs.map((x) => null),
+      outputMasks: this.outputs.map((x) => null),
+      inputShapes: this.inputs.map((x) => x.shape),
+      outputShapes: this.outputs.map((x) => x.shape)
+    });
+    this.built = true;
+    this._refCount = 1;
+  }
+  assertNotDisposed() {
+    if (this._refCount === 0) {
+      throw new Error(`Container '${this.name}' is already disposed.`);
+    }
+  }
+  dispose() {
+    this.assertNotDisposed();
+    const result = { refCountAfterDispose: null, numDisposedVariables: 0 };
+    if (--this._refCount === 0) {
+      for (const layer of this.layers) {
+        result.numDisposedVariables += layer.dispose().numDisposedVariables;
+      }
+      for (const container of this.internalContainerRefs) {
+        result.numDisposedVariables += container.dispose().numDisposedVariables;
+      }
+    }
+    result.refCountAfterDispose = this._refCount;
+    return result;
+  }
+  get trainable() {
+    return this.trainable_;
+  }
+  set trainable(trainable) {
+    this.layers.forEach((layer) => {
+      layer._trainableWeights.forEach((w) => w.trainable = trainable);
+    });
+    this.trainable_ = trainable;
+  }
+  get trainableWeights() {
+    if (this._trainableWeights.length > 0) {
+      throw new ValueError("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");
+    }
+    if (!this.trainable) {
+      return [];
+    }
+    let weights = [];
+    for (const layer of this.layers) {
+      weights = weights.concat(layer.trainableWeights);
+    }
+    return weights;
+  }
+  get nonTrainableWeights() {
+    const weights = [];
+    for (const layer of this.layers) {
+      weights.push(...layer.nonTrainableWeights);
+    }
+    if (!this.trainable) {
+      const trainableWeights = [];
+      for (const layer of this.layers) {
+        trainableWeights.push(...layer.trainableWeights);
+      }
+      return trainableWeights.concat(weights);
+    }
+    return weights;
+  }
+  get weights() {
+    return this.trainableWeights.concat(this.nonTrainableWeights);
+  }
+  loadWeights(weights, strict = true) {
+    const nameToWeight = {};
+    let totalWeightsCount = 0;
+    for (const layer of this.layers) {
+      for (const weight of layer.weights) {
+        if (nameToWeight[weight.originalName] != null) {
+          throw new ValueError(`Duplicate weight name: ${weight.originalName}`);
+        }
+        nameToWeight[weight.originalName] = weight;
+        totalWeightsCount++;
+      }
+    }
+    const weightValueTuples = [];
+    for (const name in weights) {
+      let validatedName = name;
+      if (nameToWeight[name] == null) {
+        const tokens = name.split("/");
+        const shortenNameArray = tokens.slice(0, -2).concat([tokens[tokens.length - 1]]);
+        validatedName = shortenNameArray.join("/");
+      }
+      if (nameToWeight[validatedName] != null) {
+        weightValueTuples.push([nameToWeight[validatedName], weights[name]]);
+      } else if (strict) {
+        throw new ValueError(`Provided weight data has no target variable: ${name}`);
+      }
+      delete nameToWeight[validatedName];
+    }
+    if (strict) {
+      const unsetNames = [];
+      for (const name in nameToWeight) {
+        unsetNames.push(name);
+      }
+      if (unsetNames.length > 0) {
+        throw new ValueError(`${unsetNames.length} of ${totalWeightsCount} weights are not set: ${unsetNames}`);
+      }
+    }
+    batchSetValue(weightValueTuples);
+  }
+  updatedConfig() {
+    const theConfig = this.getConfig();
+    const modelConfig = {};
+    modelConfig["className"] = this.getClassName();
+    modelConfig["config"] = theConfig;
+    modelConfig["kerasVersion"] = `tfjs-layers ${version2}`;
+    modelConfig["backend"] = "TensorFlow.js";
+    return modelConfig;
+  }
+  toJSON(unused, returnString = true) {
+    const modelConfig = convertTsToPythonic(this.updatedConfig());
+    return returnString ? JSON.stringify(modelConfig) : modelConfig;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = toList(inputs);
+      const feedDict = new FeedDict();
+      for (let i = 0; i < this.inputs.length; ++i) {
+        feedDict.add(this.inputs[i], inputs[i]);
+      }
+      return execute(this.outputs, feedDict, kwargs);
+    });
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      inputs = toList(inputs);
+      let masks;
+      if (mask == null) {
+        masks = pyListRepeat(null, inputs.length);
+      } else {
+        masks = toList(mask);
+      }
+      return this.runInternalGraph(inputs, masks)[1];
+    });
+  }
+  computeOutputShape(inputShape) {
+    const inputShapes = normalizeShapeList(inputShape);
+    if (inputShapes.length !== this.inputLayers.length) {
+      throw new ValueError(`Invalid inputShape argument ${inputShape}: model has ${this.inputLayers.length} tensor inputs.`);
+    }
+    const layersToOutputShapes = {};
+    for (let i = 0; i < inputShapes.length; i++) {
+      const layer = this.inputLayers[i];
+      const inputShape2 = inputShapes[i];
+      const shapeKey = layer.name + "_0_0";
+      layersToOutputShapes[shapeKey] = inputShape2;
+    }
+    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    if (depthKeys.length > 1) {
+      for (const depth of depthKeys) {
+        const nodes = this.nodesByDepth[depth];
+        for (const node of nodes) {
+          const layer = node.outboundLayer;
+          if (this.inputLayers.map((x) => x.id).indexOf(layer.id) !== -1) {
+            continue;
+          }
+          const inputShapes2 = [];
+          for (let j = 0; j < node.inboundLayers.length; j++) {
+            const inboundLayer = node.inboundLayers[j];
+            const nodeIndex2 = node.nodeIndices[j];
+            const tensorIndex = node.tensorIndices[j];
+            const shapeKey = `${inboundLayer.name}_${nodeIndex2}_${tensorIndex}`;
+            const inputShape2 = layersToOutputShapes[shapeKey];
+            inputShapes2.push(inputShape2);
+          }
+          const outputShape = layer.computeOutputShape(singletonOrArray(inputShapes2));
+          const outputShapes2 = normalizeShapeList(outputShape);
+          const nodeIndex = layer.inboundNodes.indexOf(node);
+          for (let j = 0; j < outputShapes2.length; j++) {
+            const shapeKey = `${layer.name}_${nodeIndex}_${j}`;
+            layersToOutputShapes[shapeKey] = outputShapes2[j];
+          }
+        }
+      }
+    }
+    const outputShapes = [];
+    const outputShapeKeys = [];
+    for (let i = 0; i < this.outputLayers.length; i++) {
+      const layer = this.outputLayers[i];
+      const nodeIndex = this.outputLayersNodeIndices[i];
+      const tensorIndex = this.outputLayersTensorIndices[i];
+      const shapeKey = `${layer.name}_${nodeIndex}_${tensorIndex}`;
+      outputShapeKeys.push(shapeKey);
+    }
+    for (let i = 0; i < outputShapeKeys.length; i++) {
+      const key = outputShapeKeys[i];
+      assert2(key in layersToOutputShapes);
+      outputShapes.push(layersToOutputShapes[key]);
+    }
+    return singletonOrArray(outputShapes);
+  }
+  runInternalGraph(inputs, masks) {
+    if (masks == null) {
+      masks = pyListRepeat(null, inputs.length);
+    }
+    const tensorMap = {};
+    for (let i = 0; i < this.inputs.length; ++i) {
+      const x = this.inputs[i];
+      const y = inputs[i];
+      const mask = masks[i];
+      tensorMap[x.id] = [y, mask];
+    }
+    const depthKeys = Object.keys(this.nodesByDepth).map((x) => parseInt(x, 10)).sort(reverseNumberCompare);
+    for (const depth of depthKeys) {
+      const nodes = this.nodesByDepth[depth];
+      for (const node of nodes) {
+        const layer = node.outboundLayer;
+        const referenceInputTensors = node.inputTensors;
+        const referenceOutputTensors = node.outputTensors;
+        const computedData = new Array();
+        for (const x of referenceInputTensors) {
+          if (x.id in tensorMap) {
+            computedData.push(tensorMap[x.id]);
+          }
+        }
+        if (computedData.length === referenceInputTensors.length) {
+          let kwargs = {};
+          let computedTensors;
+          let computedMasks;
+          let outputTensors2;
+          let outputMasks2;
+          if (node.callArgs != null) {
+            kwargs = node.callArgs;
+          }
+          if (computedData.length === 1) {
+            const [computedTensor, computedMask] = computedData[0];
+            if (kwargs["mask"] == null) {
+              kwargs["mask"] = computedMask;
+            }
+            outputTensors2 = toList(layer.call(computedTensor, kwargs));
+            outputMasks2 = toList(layer.computeMask(computedTensor, computedMask));
+            computedTensors = [computedTensor];
+            computedMasks = [computedMask];
+          } else {
+            computedTensors = computedData.map((x) => x[0]);
+            computedMasks = computedData.map((x) => x[1]);
+            if (kwargs["mask"] == null) {
+              kwargs["mask"] = computedMasks;
+            }
+            outputTensors2 = toList(layer.call(computedTensors, kwargs));
+            outputMasks2 = toList(layer.computeMask(computedTensors, computedMasks));
+          }
+          if (layer.activityRegularizer) {
+            throw new NotImplementedError("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");
+          }
+          for (let i = 0; i < referenceOutputTensors.length; ++i) {
+            const x = referenceOutputTensors[i];
+            const y = outputTensors2[i];
+            const mask = outputMasks2[i];
+            tensorMap[x.id] = [y, mask];
+          }
+        }
+      }
+    }
+    const outputTensors = [];
+    const outputMasks = [];
+    const outputShapes = [];
+    for (const x of this.outputs) {
+      assert2(x.id in tensorMap, `Could not compute output ${x.name} : ${x.id}`);
+      const [tensor2, mask] = tensorMap[x.id];
+      outputShapes.push(tensor2.shape);
+      outputTensors.push(tensor2);
+      outputMasks.push(mask);
+    }
+    return [outputTensors, outputMasks, outputShapes];
+  }
+  buildNodeConversionMap(layers) {
+    const nodeConversionMap = {};
+    let keptNodes;
+    for (const layer of this.layers) {
+      keptNodes = layer instanceof Container ? 1 : 0;
+      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {
+        const nodeKey = Container.nodeKey(layer, originalNodeIndex);
+        if (this.containerNodes.has(nodeKey)) {
+          nodeConversionMap[nodeKey] = keptNodes;
+          keptNodes += 1;
+        }
+      }
+    }
+    return nodeConversionMap;
+  }
+  getLayer(name, index) {
+    if (index != null) {
+      if (this.layers.length <= index) {
+        throw new ValueError(`Was asked to retrieve layer at index ${index}, but model only has ${this.layers.length} layer(s).`);
+      } else {
+        return this.layers[index];
+      }
+    } else {
+      if (name == null) {
+        throw new ValueError("Provide either a layer name or layer index");
+      }
+    }
+    for (const layer of this.layers) {
+      if (layer.name === name) {
+        return layer;
+      }
+    }
+    throw new ValueError(`No such layer: ${name}`);
+  }
+  calculateLosses() {
+    return tidy(() => {
+      const losses2 = [];
+      for (const layer of this.layers) {
+        for (let nodeIndex = 0; nodeIndex < layer.inboundNodes.length; ++nodeIndex) {
+          const nodeKey = Container.nodeKey(layer, nodeIndex);
+          if (this.containerNodes.has(nodeKey)) {
+            losses2.push(...layer.calculateLosses());
+          }
+        }
+      }
+      return losses2;
+    });
+  }
+  getConfig() {
+    const config = { name: this.name };
+    const nodeConversionMap = this.buildNodeConversionMap(this.layers);
+    const layerConfigs = [];
+    for (const layer of this.layers) {
+      const layerClassName = layer.getClassName();
+      const layerConfig = layer.getConfig();
+      const filteredInboundNodes = [];
+      for (let originalNodeIndex = 0; originalNodeIndex < layer.inboundNodes.length; originalNodeIndex++) {
+        const node = layer.inboundNodes[originalNodeIndex];
+        const nodeKey = Container.nodeKey(layer, originalNodeIndex);
+        let kwargs = {};
+        if (this.containerNodes.has(nodeKey)) {
+          if (node.callArgs) {
+            try {
+              JSON.stringify(node.callArgs);
+              kwargs = node.callArgs;
+            } catch (err) {
+              console.warn(`Layer ${layer.name} was passed non-serializable keyword arguments: ${node.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`);
+              kwargs = {};
+            }
+          }
+          if (node.inboundLayers.length > 0) {
+            const nodeData = [];
+            for (let i = 0; i < node.inboundLayers.length; i++) {
+              const inboundLayer = node.inboundLayers[i];
+              const nodeIndex = node.nodeIndices[i];
+              const tensorIndex = node.tensorIndices[i];
+              const nodeKey2 = Container.nodeKey(inboundLayer, nodeIndex);
+              let newNodeIndex = nodeConversionMap[nodeKey2];
+              if (newNodeIndex == null) {
+                newNodeIndex = 0;
+              }
+              nodeData.push([inboundLayer.name, newNodeIndex, tensorIndex, kwargs]);
+            }
+            filteredInboundNodes.push(nodeData);
+          }
+        }
+      }
+      const dict = {};
+      dict["name"] = layer.name;
+      dict["className"] = layerClassName;
+      dict["config"] = layerConfig;
+      dict["inboundNodes"] = filteredInboundNodes;
+      layerConfigs.push(dict);
+    }
+    config["layers"] = layerConfigs;
+    const modelInputs = [];
+    for (let i = 0; i < this.inputLayers.length; i++) {
+      const layer = this.inputLayers[i];
+      const nodeIndex = this.inputLayersNodeIndices[i];
+      const nodeKey = Container.nodeKey(layer, nodeIndex);
+      if (!this.containerNodes.has(nodeKey)) {
+        continue;
+      }
+      let newNodeIndex = nodeConversionMap[nodeKey];
+      if (newNodeIndex === null || newNodeIndex === void 0) {
+        newNodeIndex = 0;
+      }
+      const tensorIndex = this.inputLayersTensorIndices[i];
+      modelInputs.push([layer.name, newNodeIndex, tensorIndex]);
+    }
+    config["inputLayers"] = modelInputs;
+    const modelOutputs = [];
+    for (let i = 0; i < this.outputLayers.length; i++) {
+      const layer = this.outputLayers[i];
+      const nodeIndex = this.outputLayersNodeIndices[i];
+      const nodeKey = Container.nodeKey(layer, nodeIndex);
+      if (!this.containerNodes.has(nodeKey)) {
+        continue;
+      }
+      let newNodeIndex = nodeConversionMap[nodeKey];
+      if (newNodeIndex === null || newNodeIndex === void 0) {
+        newNodeIndex = 0;
+      }
+      const tensorIndex = this.outputLayersTensorIndices[i];
+      modelOutputs.push([layer.name, newNodeIndex, tensorIndex]);
+    }
+    config["outputLayers"] = modelOutputs;
+    return config;
+  }
+  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {
+    const createdLayers = {};
+    const unprocessedNodes = {};
+    function addUnprocessedNode(layer, nodeData) {
+      if (!(layer.name in unprocessedNodes)) {
+        unprocessedNodes[layer.name] = [nodeData];
+      } else {
+        unprocessedNodes[layer.name].push(nodeData);
+      }
+    }
+    function processNode(layer, nodeData) {
+      const inputTensors2 = [];
+      let kwargs;
+      for (const inputData of nodeData) {
+        const inboundLayerName = inputData[0];
+        const inboundNodeIndex = inputData[1];
+        const inboundTensorIndex = inputData[2];
+        kwargs = inputData[3] == null ? {} : inputData[3];
+        if (!(inboundLayerName in createdLayers)) {
+          addUnprocessedNode(layer, nodeData);
+          return;
+        }
+        const inboundLayer = createdLayers[inboundLayerName];
+        if (inboundLayer.inboundNodes.length <= inboundNodeIndex) {
+          addUnprocessedNode(layer, nodeData);
+          return;
+        }
+        const inboundNode = inboundLayer.inboundNodes[inboundNodeIndex];
+        inputTensors2.push(inboundNode.outputTensors[inboundTensorIndex]);
+      }
+      if (inputTensors2.length > 0) {
+        layer.apply(singletonOrArray(inputTensors2), kwargs);
+      }
+    }
+    function processLayer(layerData) {
+      const layerName = layerData["name"];
+      const layer = deserialize(layerData, config["customObjects"] != null ? config["customObjects"] : {});
+      layer.setFastWeightInitDuringBuild(fastWeightInit);
+      createdLayers[layerName] = layer;
+      const inboundNodesData = layerData["inboundNodes"];
+      inboundNodesData.forEach((nodeData) => {
+        if (!(nodeData instanceof Array)) {
+          throw new ValueError(`Corrupted configuration, expected array for nodeData: ${nodeData}`);
+        }
+        addUnprocessedNode(layer, nodeData);
+      });
+    }
+    const name = config["name"];
+    const layersFromConfig = config["layers"];
+    for (const layerData of layersFromConfig) {
+      processLayer(layerData);
+    }
+    while (!isObjectEmpty(unprocessedNodes)) {
+      for (const layerData of layersFromConfig) {
+        const layer = createdLayers[layerData["name"]];
+        if (layer.name in unprocessedNodes) {
+          const currentUnprocessedNodesForLayer = unprocessedNodes[layer.name];
+          delete unprocessedNodes[layer.name];
+          for (const nodeData of currentUnprocessedNodesForLayer) {
+            processNode(layer, nodeData);
+          }
+        }
+      }
+    }
+    const inputTensors = [];
+    const outputTensors = [];
+    const inputLayersFromConfig = config["inputLayers"];
+    for (const layerData of inputLayersFromConfig) {
+      const layerName = layerData[0];
+      const nodeIndex = layerData[1];
+      const tensorIndex = layerData[2];
+      assert2(layerName in createdLayers);
+      const layer = createdLayers[layerName];
+      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;
+      inputTensors.push(layerOutputTensors[tensorIndex]);
+    }
+    const outputLayersFromConfig = config["outputLayers"];
+    for (const layerData of outputLayersFromConfig) {
+      const layerName = layerData[0];
+      const nodeIndex = layerData[1];
+      const tensorIndex = layerData[2];
+      assert2(layerName in createdLayers);
+      const layer = createdLayers[layerName];
+      const layerOutputTensors = layer.inboundNodes[nodeIndex].outputTensors;
+      outputTensors.push(layerOutputTensors[tensorIndex]);
+    }
+    return new cls({ inputs: inputTensors, outputs: outputTensors, name });
+  }
+  get stateful() {
+    if (this._stateful) {
+      throw new ValueError("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");
+    }
+    for (const layer of this.layers) {
+      if (layer.stateful) {
+        return true;
+      }
+    }
+    return false;
+  }
+  resetStates() {
+    tidy(() => {
+      this.layers.forEach((layer) => {
+        if (layer.stateful) {
+          layer.resetStates();
+        }
+      });
+    });
+  }
+};
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/training_utils.js
+function standardizeSampleOrClassWeights(xWeight, outputNames, weightType) {
+  const numOutputs = outputNames.length;
+  if (xWeight == null || Array.isArray(xWeight) && xWeight.length === 0) {
+    return outputNames.map((name) => null);
+  }
+  if (numOutputs === 1) {
+    if (Array.isArray(xWeight) && xWeight.length === 1) {
+      return xWeight;
+    } else if (typeof xWeight === "object" && outputNames[0] in xWeight) {
+      return [xWeight[outputNames[0]]];
+    } else {
+      return [xWeight];
+    }
+  }
+  if (Array.isArray(xWeight)) {
+    if (xWeight.length !== numOutputs) {
+      throw new Error(`Provided ${weightType} is an array of ${xWeight.length} element(s), but the model has ${numOutputs} outputs. Make sure a set of weights is provided for each model output.`);
+    }
+    return xWeight;
+  } else if (typeof xWeight === "object" && Object.keys(xWeight).length > 0 && typeof xWeight[Object.keys(xWeight)[0]] === "object") {
+    const output = [];
+    outputNames.forEach((outputName) => {
+      if (outputName in xWeight) {
+        output.push(xWeight[outputName]);
+      } else {
+        output.push(null);
+      }
+    });
+    return output;
+  } else {
+    throw new Error(`The model has multiple (${numOutputs}) outputs, so ${weightType} must be either an array with ${numOutputs} elements or an object with ${outputNames} keys. Provided ${weightType} not understood: ${JSON.stringify(xWeight)}`);
+  }
+}
+function standardizeClassWeights(classWeight, outputNames) {
+  return standardizeSampleOrClassWeights(classWeight, outputNames, "classWeight");
+}
+async function standardizeWeights(y, sampleWeight, classWeight, sampleWeightMode) {
+  if (sampleWeight != null || sampleWeightMode != null) {
+    throw new Error("Support sampleWeight is not implemented yet");
+  }
+  if (classWeight != null) {
+    const yClasses = tidy(() => {
+      if (y.shape.length === 1) {
+        return clone(y);
+      } else if (y.shape.length === 2) {
+        if (y.shape[1] > 1) {
+          const axis = 1;
+          return argMax(y, axis);
+        } else if (y.shape[1] === 1) {
+          return reshape(y, [y.shape[0]]);
+        } else {
+          throw new Error(`Encountered unexpected last-dimension size (${y.shape[1]}) during handling of class weights. The size is expected to be >= 1.`);
+        }
+      } else {
+        throw new Error(`Unexpected rank of target (y) tensor (${y.rank}) during handling of class weights. The rank is expected to be 1 or 2.`);
+      }
+    });
+    const yClassIndices = Array.from(await yClasses.data());
+    dispose(yClasses);
+    const classSampleWeight = [];
+    yClassIndices.forEach((classIndex) => {
+      if (classWeight[classIndex] == null) {
+        throw new Error(`classWeight must contain all classes in the training data. The class ${classIndex} exists in the data but not in classWeight`);
+      } else {
+        classSampleWeight.push(classWeight[classIndex]);
+      }
+    });
+    return tensor1d(classSampleWeight, "float32");
+  } else {
+    return null;
+  }
+}
+function computeWeightedLoss2(losses2, sampleWeights) {
+  return mul(losses2, sampleWeights);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/training_dataset.js
+var DEFAULT_VALIDATION_BATCH_SIZE = 32;
+function standardizeDataIteratorOutput(model2, iteratorOut) {
+  let xs;
+  let ys;
+  const iteratorOutObj = iteratorOut;
+  xs = iteratorOutObj["xs"];
+  ys = iteratorOutObj["ys"];
+  util_exports.assert(xs != null && ys != null, () => `A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${iteratorOut}`);
+  const flattenedXs = flattenTensorOrArrayOrMap("input", model2.inputNames, xs);
+  const flattenedYs = flattenTensorOrArrayOrMap("output", model2.outputNames, ys);
+  const batchSize = flattenedXs[0].shape[0];
+  util_exports.assert(flattenedXs.length === model2.inputs.length, () => `LayersModel has ${model2.inputs.length} inputs, but the dataset provides ${flattenedXs.length} inputs.  (Expected input keys: ${JSON.stringify(model2.inputNames)})`);
+  util_exports.assert(flattenedYs.length === model2.outputs.length, () => `LayersModel has ${model2.outputs.length} outputs, but the dataset provides ${flattenedYs.length} outputs.  (Expected output keys: ${JSON.stringify(model2.outputNames)})`);
+  for (let xIndex = 0; xIndex < flattenedXs.length; xIndex++) {
+    util_exports.assert(flattenedXs[xIndex].shape[0] === batchSize, () => `Batch size mismatch: input ${model2.inputNames[xIndex]} has ${flattenedXs[xIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);
+  }
+  for (let yIndex = 0; yIndex < flattenedYs.length; yIndex++) {
+    util_exports.assert(flattenedYs[yIndex].shape[0] === batchSize, () => `Batch size mismatch: output ${model2.outputNames[yIndex]} has ${flattenedYs[yIndex].shape[0]}; expected  ${batchSize} based on input ${model2.inputNames[0]}.`);
+  }
+  return { xs: flattenedXs, ys: flattenedYs };
+}
+function flattenTensorOrArrayOrMap(inputOrOutput, names, values) {
+  if (values instanceof Tensor) {
+    return [values];
+  } else if (Array.isArray(values)) {
+    util_exports.assert(values.length === names.length, () => `Received an array of ${values.length} Tensors, but expected ${names.length} to match the ${inputOrOutput} keys ${names}.`);
+    return values;
+  } else {
+    const result = [];
+    for (const name of names) {
+      if (values[name] == null) {
+        throw new ValueError(`The feature data generated by the dataset lacks the required ${inputOrOutput} key '${name}'.`);
+      }
+      result.push(values[name]);
+    }
+    return result;
+  }
+}
+function standardizeTensorValidationData(data) {
+  if (data.length === 3) {
+    throw new NotImplementedError("Validation with sample weights is not implemented yet.");
+  }
+  return { xs: data[0], ys: data[1] };
+}
+async function fitDataset(model2, dataset, args) {
+  const hasBatchesPerEpoch = args.batchesPerEpoch != null;
+  util_exports.assert(model2.optimizer != null, () => "You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig).");
+  util_exports.assert(args != null, () => `For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call.`);
+  util_exports.assert(args.epochs != null && args.epochs > 0 && Number.isInteger(args.epochs), () => `For fitDataset(), config.epochs is expected to be a positive integer, but got ${args.epochs}`);
+  util_exports.assert(!hasBatchesPerEpoch || args.batchesPerEpoch > 0 && Number.isInteger(args.batchesPerEpoch), () => `For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${args.batchesPerEpoch}`);
+  util_exports.assert(
+    args["validationSplit"] == null,
+    () => "`validationSplit` is not supported by `fitDataset()`. Use validationData instead."
+  );
+  if (model2.isTraining) {
+    throw new Error("Cannot start training because another fit() call is ongoing.");
+  }
+  model2.isTraining = true;
+  try {
+    const doValidation = args.validationData != null;
+    let valXs;
+    let valYs;
+    if (doValidation) {
+      if (isDatasetObject(args.validationData)) {
+        util_exports.assert(args.validationBatches == null || args.validationBatches > 0 && Number.isInteger(args.validationBatches), () => `For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${args.validationBatches}`);
+      } else {
+        const validationData = standardizeTensorValidationData(args.validationData);
+        valXs = validationData.xs;
+        valYs = validationData.ys;
+      }
+    }
+    const trainFunction = model2.makeTrainFunction();
+    const outLabels = model2.getDedupedMetricsNames();
+    let callbackMetrics;
+    if (doValidation) {
+      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => "val_" + n));
+    } else {
+      callbackMetrics = outLabels.slice();
+    }
+    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);
+    const verbose = args.verbose == null ? 1 : args.verbose;
+    const { callbackList, history } = configureCallbacks(
+      callbacks2,
+      verbose,
+      args.epochs,
+      null,
+      null,
+      getStepsPerEpoch(dataset, args),
+      null,
+      doValidation,
+      callbackMetrics
+    );
+    callbackList.setModel(model2);
+    model2.history = history;
+    await callbackList.onTrainBegin();
+    model2.stopTraining_ = false;
+    let epoch = args.initialEpoch == null ? 0 : args.initialEpoch;
+    let dataIterator = await dataset.iterator();
+    while (epoch < args.epochs) {
+      const epochLogs = {};
+      await callbackList.onEpochBegin(epoch);
+      let stepsDone = 0;
+      let batchIndex = 0;
+      if (!hasBatchesPerEpoch) {
+        dataIterator = await dataset.iterator();
+      }
+      while (hasBatchesPerEpoch ? stepsDone < args.batchesPerEpoch : true) {
+        const iteratorOut = await dataIterator.next();
+        if (hasBatchesPerEpoch && iteratorOut.done) {
+          console.warn(`You provided \`batchesPerEpoch\` as ${args.batchesPerEpoch}, but your dataset iterator ran out of data after ${stepsDone} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${args.batchesPerEpoch * args.epochs} batches). You may need to use the repeat() function when building your dataset.`);
+          break;
+        }
+        if (iteratorOut.value != null) {
+          const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);
+          const batchLogs = {};
+          batchLogs["batch"] = batchIndex;
+          batchLogs["size"] = xs[0].shape[0];
+          await callbackList.onBatchBegin(batchIndex, batchLogs);
+          const sampleWeights = [];
+          if (args.classWeight != null) {
+            const standardClassWeights = standardizeClassWeights(args.classWeight, model2.outputNames);
+            for (let i = 0; i < standardClassWeights.length; ++i) {
+              sampleWeights.push(await standardizeWeights(ys[i], null, standardClassWeights[i]));
+            }
+          }
+          const ins = xs.concat(ys).concat(sampleWeights);
+          const outs = trainFunction(ins);
+          dispose(ins);
+          for (let i = 0; i < outLabels.length; ++i) {
+            const label = outLabels[i];
+            const out = outs[i];
+            batchLogs[label] = out;
+            keep(out);
+          }
+          await callbackList.onBatchEnd(batchIndex, batchLogs);
+          disposeTensorsInLogs(batchLogs);
+          batchIndex++;
+          stepsDone++;
+        }
+        if (hasBatchesPerEpoch ? stepsDone >= args.batchesPerEpoch : iteratorOut.done) {
+          if (doValidation) {
+            let valOuts;
+            if (isDatasetObject(args.validationData)) {
+              valOuts = toList(await model2.evaluateDataset(args.validationData, { batches: args.validationBatches }));
+            } else {
+              valOuts = toList(model2.evaluate(valXs, valYs, {
+                batchSize: args.validationBatchSize == null ? DEFAULT_VALIDATION_BATCH_SIZE : args.validationBatchSize,
+                verbose: 0
+              }));
+            }
+            for (let i = 0; i < model2.metricsNames.length; ++i) {
+              epochLogs[`val_${model2.metricsNames[i]}`] = valOuts[i];
+            }
+          }
+          break;
+        }
+        if (model2.stopTraining_) {
+          break;
+        }
+      }
+      await callbackList.onEpochEnd(epoch, epochLogs);
+      epoch++;
+      if (model2.stopTraining_) {
+        break;
+      }
+    }
+    await callbackList.onTrainEnd();
+    await model2.history.syncData();
+    return model2.history;
+  } finally {
+    model2.isTraining = false;
+  }
+}
+function getStepsPerEpoch(dataset, args) {
+  let stepsPerEpoch = null;
+  if (args.batchesPerEpoch != null) {
+    stepsPerEpoch = args.batchesPerEpoch;
+  } else if (Number.isFinite(dataset.size)) {
+    stepsPerEpoch = dataset.size;
+  }
+  return stepsPerEpoch;
+}
+function isDatasetObject(dataset) {
+  return typeof dataset.iterator === "function";
+}
+function isLazyIteratorObject(iterator) {
+  return typeof iterator.next === "function";
+}
+async function evaluateDataset(model2, dataset, args) {
+  args = args || {};
+  const hasBatches = args.batches != null;
+  const f = model2.testFunction;
+  let outs = [];
+  if (args.verbose > 0) {
+    throw new NotImplementedError("Verbose mode is not implemented yet.");
+  }
+  util_exports.assert(!hasBatches || args.batches > 0 && Number.isInteger(args.batches), () => `Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(args.batches)}`);
+  const dataIterator = isLazyIteratorObject(dataset) ? dataset : await dataset.iterator();
+  let numExamples = 0;
+  let batch = 0;
+  while (hasBatches ? batch < args.batches : true) {
+    const iteratorOut = await dataIterator.next();
+    outs = tidy(() => {
+      if (iteratorOut.value) {
+        const { xs, ys } = standardizeDataIteratorOutput(model2, iteratorOut.value);
+        const xsAndYs = xs.concat(ys);
+        const batchOuts = tidy(() => f(xsAndYs));
+        dispose(xsAndYs);
+        if (batch === 0) {
+          for (let i = 0; i < batchOuts.length; ++i) {
+            outs.push(scalar(0));
+          }
+        }
+        const batchSize = xsAndYs[0].shape[0];
+        for (let i = 0; i < batchOuts.length; ++i) {
+          const batchOut = batchOuts[i];
+          const oldScalar = outs[i];
+          outs[i] = tidy(() => add2(outs[i], mul(batchSize, batchOut)));
+          if (batch > 0) {
+            dispose(oldScalar);
+          }
+        }
+        dispose(batchOuts);
+        numExamples += batchSize;
+        ++batch;
+      }
+      return outs;
+    });
+    if (iteratorOut.done) {
+      if (hasBatches) {
+        console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${args.batches} batches). You may need to use the repeat() function when building your dataset.`);
+      }
+      break;
+    }
+  }
+  for (let i = 0; i < outs.length; ++i) {
+    const oldScalar = outs[i];
+    outs[i] = div(outs[i], numExamples);
+    dispose(oldScalar);
+  }
+  return singletonOrArray(outs);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/training_tensors.js
+function checkBatchSize(batchSize) {
+  util_exports.assert(batchSize > 0 && Number.isInteger(batchSize), () => `batchSize is required to be a positive integer, but got ${batchSize}`);
+}
+function sliceArrays(arrays, start, stop) {
+  if (arrays == null) {
+    return [null];
+  } else if (Array.isArray(arrays)) {
+    return arrays.map((array2) => sliceAlongFirstAxis(array2, start, stop - start));
+  } else {
+    return sliceAlongFirstAxis(arrays, start, stop - start);
+  }
+}
+function sliceArraysByIndices(arrays, indices) {
+  return tidy(() => {
+    if (arrays == null) {
+      return null;
+    } else if (Array.isArray(arrays)) {
+      return arrays.map((array2) => sliceArraysByIndices(array2, indices));
+    } else {
+      return gather2(arrays, indices.dtype === "int32" ? indices : cast(indices, "int32"));
+    }
+  });
+}
+function makeBatches(size, batchSize) {
+  const output = [];
+  let batchStart = 0;
+  let batchEnd = null;
+  while (batchStart < size) {
+    batchEnd = batchStart + batchSize;
+    if (batchEnd >= size) {
+      batchEnd = size;
+    }
+    output.push([batchStart, batchEnd]);
+    batchStart = batchEnd;
+  }
+  return output;
+}
+async function fitLoop(model2, f, ins, outLabels, batchSize, epochs, verbose, callbacks2, valF, valIns, shuffle2, callbackMetrics, initialEpoch, stepsPerEpoch, validationSteps) {
+  if (batchSize == null) {
+    batchSize = 32;
+  }
+  if (epochs == null) {
+    epochs = 1;
+  }
+  if (shuffle2 == null) {
+    shuffle2 = true;
+  }
+  if (initialEpoch == null) {
+    initialEpoch = 0;
+  }
+  let doValidation = false;
+  if (valF != null && valIns != null) {
+    doValidation = true;
+  }
+  if (validationSteps != null) {
+    doValidation = true;
+    if (stepsPerEpoch == null) {
+      throw new ValueError("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");
+    }
+  }
+  const numTrainSamples = model2.checkNumSamples(ins, batchSize, stepsPerEpoch, "steps_per_epoch");
+  let indexArray;
+  if (numTrainSamples != null) {
+    indexArray = range2(0, numTrainSamples);
+  }
+  if (verbose == null) {
+    verbose = 1;
+  }
+  const { callbackList, history } = configureCallbacks(callbacks2, verbose, epochs, initialEpoch, numTrainSamples, stepsPerEpoch, batchSize, doValidation, callbackMetrics);
+  callbackList.setModel(model2);
+  model2.history = history;
+  await callbackList.onTrainBegin();
+  model2.stopTraining_ = false;
+  for (let epoch = initialEpoch; epoch < epochs; ++epoch) {
+    await callbackList.onEpochBegin(epoch);
+    const epochLogs = {};
+    if (stepsPerEpoch != null) {
+      throw new NotImplementedError("stepsPerEpoch mode is not implemented yet.");
+    } else {
+      if (shuffle2 === "batch") {
+        throw new NotImplementedError("batch shuffling is not implemneted yet");
+      } else if (shuffle2) {
+        util_exports.shuffle(indexArray);
+      }
+      const epochIndexArray1D = tensor1d(indexArray);
+      const batches = makeBatches(numTrainSamples, batchSize);
+      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+        const batchLogs = {};
+        await callbackList.onBatchBegin(batchIndex, batchLogs);
+        tidy(() => {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const batchIds = sliceAlongFirstAxis(epochIndexArray1D, batchStart, batchEnd - batchStart);
+          batchLogs["batch"] = batchIndex;
+          batchLogs["size"] = batchEnd - batchStart;
+          const insBatch = sliceArraysByIndices(ins, batchIds);
+          const outs = f(insBatch);
+          for (let i = 0; i < outLabels.length; ++i) {
+            const label = outLabels[i];
+            const out = outs[i];
+            batchLogs[label] = out;
+            keep(out);
+          }
+          if (batchIndex === batches.length - 1) {
+            if (doValidation) {
+              const valOuts = model2.testLoop(valF, valIns, batchSize);
+              for (let i = 0; i < outLabels.length; ++i) {
+                const label = outLabels[i];
+                const out = valOuts[i];
+                keep(out);
+                epochLogs["val_" + label] = out;
+              }
+            }
+          }
+        });
+        await callbackList.onBatchEnd(batchIndex, batchLogs);
+        disposeTensorsInLogs(batchLogs);
+        if (model2.stopTraining_) {
+          break;
+        }
+      }
+      epochIndexArray1D.dispose();
+    }
+    await callbackList.onEpochEnd(epoch, epochLogs);
+    if (model2.stopTraining_) {
+      break;
+    }
+  }
+  await callbackList.onTrainEnd();
+  await model2.history.syncData();
+  return model2.history;
+}
+async function fitTensors(model2, x, y, args = {}) {
+  if (model2.isTraining) {
+    throw new Error("Cannot start training because another fit() call is ongoing.");
+  }
+  model2.isTraining = true;
+  let inputs;
+  let targets;
+  let originalInputs;
+  let originalTargets;
+  let inputValX;
+  let inputValY;
+  let valX;
+  let valY;
+  let sampleWeights;
+  try {
+    const batchSize = args.batchSize == null ? 32 : args.batchSize;
+    checkBatchSize(batchSize);
+    const checkBatchAxis = false;
+    const standardizedOuts = await model2.standardizeUserData(x, y, args.sampleWeight, args.classWeight, checkBatchAxis, batchSize);
+    inputs = standardizedOuts[0];
+    targets = standardizedOuts[1];
+    sampleWeights = standardizedOuts[2];
+    let doValidation = false;
+    let valIns;
+    if (args.validationData != null && args.validationData.length > 0) {
+      doValidation = true;
+      if (args.validationData.length === 2) {
+        inputValX = args.validationData[0];
+        inputValY = args.validationData[1];
+      } else if (args.validationData.length === 3) {
+        throw new NotImplementedError("validationData including sample weights is not supported yet.");
+      } else {
+        throw new ValueError(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${args.validationData} is invalid.`);
+      }
+      const checkBatchAxis2 = true;
+      const valStandardized = await model2.standardizeUserData(inputValX, inputValY, null, null, checkBatchAxis2, batchSize);
+      valX = valStandardized[0];
+      valY = valStandardized[1];
+      valIns = valX.concat(valY);
+    } else if (args.validationSplit != null && args.validationSplit > 0 && args.validationSplit < 1) {
+      doValidation = true;
+      const splitAt = Math.floor(inputs[0].shape[0] * (1 - args.validationSplit));
+      const originalBatchSize = inputs[0].shape[0];
+      valX = sliceArrays(inputs, splitAt, originalBatchSize);
+      originalInputs = inputs;
+      inputs = sliceArrays(inputs, 0, splitAt);
+      valY = sliceArrays(targets, splitAt, originalBatchSize);
+      originalTargets = targets;
+      targets = sliceArrays(targets, 0, splitAt);
+      valIns = valX.concat(valY);
+    } else if (args.validationSteps != null) {
+      doValidation = true;
+    }
+    const ins = inputs.concat(targets).concat(sampleWeights);
+    model2.checkTrainableWeightsConsistency();
+    const trainFunction = model2.makeTrainFunction();
+    const outLabels = model2.getDedupedMetricsNames();
+    let valFunction;
+    let callbackMetrics;
+    if (doValidation) {
+      model2.makeTestFunction();
+      valFunction = model2.testFunction;
+      callbackMetrics = outLabels.slice().concat(outLabels.map((n) => "val_" + n));
+    } else {
+      valFunction = null;
+      valIns = [];
+      callbackMetrics = outLabels.slice();
+    }
+    const callbacks2 = standardizeCallbacks(args.callbacks, args.yieldEvery);
+    const out = await fitLoop(model2, trainFunction, ins, outLabels, batchSize, args.epochs, args.verbose, callbacks2, valFunction, valIns, args.shuffle, callbackMetrics, args.initialEpoch, null, null);
+    return out;
+  } finally {
+    model2.isTraining = false;
+    disposeNewTensors(inputs, x);
+    disposeNewTensors(targets, y);
+    disposeNewTensors(originalInputs, x);
+    disposeNewTensors(originalTargets, y);
+    disposeNewTensors(valX, inputValX);
+    disposeNewTensors(valY, inputValY);
+    if (sampleWeights != null) {
+      dispose(sampleWeights);
+    }
+  }
+}
+function ensureTensorsRank2OrHigher(tensors) {
+  const outs = [];
+  if (tensors instanceof Tensor) {
+    tensors = [tensors];
+  }
+  for (let i = 0; i < tensors.length; ++i) {
+    const tensor2 = tensors[i];
+    if (tensor2.rank === 1) {
+      outs.push(expandDims2(tensor2, 1));
+    } else if (tensor2.rank === 0) {
+      throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");
+    } else {
+      outs.push(tensor2);
+    }
+  }
+  return outs;
+}
+function disposeNewTensors(tensors, refTensors) {
+  if (tensors == null) {
+    return;
+  }
+  const oldTensorIds = [];
+  if (refTensors instanceof Tensor) {
+    oldTensorIds.push(refTensors.id);
+  } else if (Array.isArray(refTensors)) {
+    refTensors.forEach((t) => oldTensorIds.push(t.id));
+  } else if (refTensors != null) {
+    for (const name in refTensors) {
+      const oldTensor = refTensors[name];
+      oldTensorIds.push(oldTensor.id);
+    }
+  }
+  const tensorsToDispose = [];
+  if (tensors instanceof Tensor) {
+    if (oldTensorIds.indexOf(tensors.id) === -1) {
+      tensorsToDispose.push(tensors);
+    }
+  } else if (Array.isArray(tensors)) {
+    tensors.forEach((t) => {
+      if (oldTensorIds.indexOf(t.id) === -1) {
+        tensorsToDispose.push(t);
+      }
+    });
+  } else if (tensors != null) {
+    for (const name in tensors) {
+      const tensor2 = tensors[name];
+      if (oldTensorIds.indexOf(tensor2.id) === -1) {
+        tensorsToDispose.push(tensor2);
+      }
+    }
+  }
+  tensorsToDispose.forEach((t) => {
+    if (!t.isDisposed) {
+      t.dispose();
+    }
+  });
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/engine/training.js
+function isDataTensor(x) {
+  return x instanceof Tensor;
+}
+function isDataArray(x) {
+  return Array.isArray(x);
+}
+function isDataDict(x) {
+  return !isDataTensor(x) && !isDataArray(x);
+}
+function standardizeInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = "") {
+  if (names == null || names.length === 0) {
+    if (data != null) {
+      let gotUnexpectedData = false;
+      if (isDataArray(data) && data.length > 0) {
+        gotUnexpectedData = true;
+      } else if (isDataDict(data)) {
+        for (const key in data) {
+          if (data.hasOwnProperty(key)) {
+            gotUnexpectedData = true;
+            break;
+          }
+        }
+      } else {
+        gotUnexpectedData = true;
+      }
+      if (gotUnexpectedData) {
+        throw new ValueError(`Error when checking model ${exceptionPrefix} expected no data, but got ${data}`);
+      }
+    }
+    return [];
+  }
+  if (data == null) {
+    return names.map((name) => null);
+  }
+  let arrays;
+  if (isDataDict(data)) {
+    data = data;
+    arrays = [];
+    for (const name of names) {
+      if (data[name] == null) {
+        throw new ValueError(`No data provided for "${name}". Need data for each key in: ${names}`);
+      }
+      arrays.push(data[name]);
+    }
+  } else if (isDataArray(data)) {
+    data = data;
+    if (data.length !== names.length) {
+      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${names.length} Tensor(s), but instead got the following list of Tensor(s): ${data}`);
+    }
+    arrays = data;
+  } else {
+    data = data;
+    if (names.length > 1) {
+      throw new ValueError(`The model ${exceptionPrefix} expects ${names.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${data.shape}`);
+    }
+    arrays = [data];
+  }
+  arrays = ensureTensorsRank2OrHigher(arrays);
+  if (shapes != null) {
+    for (let i = 0; i < names.length; ++i) {
+      if (shapes[i] == null) {
+        continue;
+      }
+      const array2 = arrays[i];
+      if (array2.shape.length !== shapes[i].length) {
+        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s). but got array with shape ${array2.shape}`);
+      }
+      for (let j = 0; j < shapes[i].length; ++j) {
+        if (j === 0 && !checkBatchAxis) {
+          continue;
+        }
+        const dim = array2.shape[j];
+        const refDim = shapes[i][j];
+        if (refDim != null && refDim >= 0 && dim !== refDim) {
+          throw new ValueError(`${exceptionPrefix} expected a batch of elements where each example has shape [${shapes[i].slice(1, shapes[i].length)}] (i.e.,tensor shape [*,${shapes[i].slice(1, shapes[i].length)}]) but the ${exceptionPrefix} received an input with ${array2.shape[0]} examples, each with shape [${array2.shape.slice(1, array2.shape.length)}] (tensor shape [${array2.shape}])`);
+        }
+      }
+    }
+  }
+  return arrays;
+}
+function checkArrayLengths(inputs, targets, weights) {
+  const setX = unique2(inputs.map((input2) => input2.shape[0]));
+  setX.sort();
+  const setY = unique2(targets.map((target) => target.shape[0]));
+  setY.sort();
+  if (setX.length > 1) {
+    throw new ValueError(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(inputs.map((input2) => input2.shape))}`);
+  }
+  if (setY.length > 1) {
+    throw new ValueError(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(targets.map((target) => target.shape))}`);
+  }
+  if (setX.length > 0 && setY.length > 0 && !util_exports.arraysEqual(setX, setY)) {
+    throw new ValueError(`Input Tensors should have the same number of samples as target Tensors. Found ${setX[0]} input sample(s) and ${setY[0]} target sample(s).`);
+  }
+}
+function checkLossAndTargetCompatibility(targets, lossFns, outputShapes) {
+  const keyLosses = [
+    meanSquaredError2,
+    binaryCrossentropy,
+    categoricalCrossentropy
+  ];
+  for (let i = 0; i < targets.length; ++i) {
+    const y = targets[i];
+    const loss = lossFns[i];
+    const shape = outputShapes[i];
+    if (loss == null) {
+      continue;
+    }
+    if (loss === categoricalCrossentropy) {
+      if (y.shape[y.shape.length - 1] === 1) {
+        throw new ValueError(`You are passing a target array of shape ${y.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);
+      }
+    }
+    if (keyLosses.indexOf(loss) !== -1) {
+      const slicedYShape = y.shape.slice(1);
+      const slicedShape = shape.slice(1);
+      for (let j = 0; j < slicedYShape.length; ++j) {
+        const targetDim = slicedYShape[j];
+        const outDim = slicedShape[j];
+        if (outDim != null && targetDim !== outDim) {
+          throw new ValueError(`A target Tensor with shape ${y.shape} was passed for an output of shape ${shape}, while using a loss function that expects targets to have the same shape as the output.`);
+        }
+      }
+    }
+  }
+}
+function checkInputData(data, names, shapes, checkBatchAxis = true, exceptionPrefix = "") {
+  let arrays;
+  if (Array.isArray(data)) {
+    if (data.length !== names.length) {
+      throw new ValueError(`Error when checking model ${exceptionPrefix}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${names.length} Tensor(s), but instead got ${data.length} Tensors(s).`);
+    }
+    arrays = data;
+  } else {
+    if (names.length > 1) {
+      throw new ValueError(`The model expects ${names.length} ${exceptionPrefix} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(data.shape)}.`);
+    }
+    arrays = [data];
+  }
+  if (shapes != null) {
+    for (let i = 0; i < names.length; ++i) {
+      if (shapes[i] == null) {
+        continue;
+      }
+      const array2 = arrays[i];
+      if (array2.shape.length !== shapes[i].length) {
+        throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have ${shapes[i].length} dimension(s), but got array with shape ${JSON.stringify(array2.shape)}`);
+      }
+      for (let j = 0; j < shapes[i].length; ++j) {
+        if (j === 0 && !checkBatchAxis) {
+          continue;
+        }
+        const dim = array2.shape[j];
+        const refDim = shapes[i][j];
+        if (refDim != null) {
+          if (refDim !== dim) {
+            throw new ValueError(`Error when checking ${exceptionPrefix}: expected ${names[i]} to have shape ${JSON.stringify(shapes[i])} but got array with shape ${JSON.stringify(array2.shape)}.`);
+          }
+        }
+      }
+    }
+  }
+}
+function collectMetrics(metrics, outputNames) {
+  if (metrics == null || Array.isArray(metrics) && metrics.length === 0) {
+    return outputNames.map((name) => []);
+  }
+  let wrappedMetrics;
+  if (typeof metrics === "string" || typeof metrics === "function") {
+    wrappedMetrics = [metrics];
+  } else if (Array.isArray(metrics) || typeof metrics === "object") {
+    wrappedMetrics = metrics;
+  } else {
+    throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${metrics}`);
+  }
+  if (Array.isArray(wrappedMetrics)) {
+    return outputNames.map((name) => wrappedMetrics);
+  } else {
+    const nestedMetrics = [];
+    for (const name of outputNames) {
+      let outputMetrics = wrappedMetrics.hasOwnProperty(name) ? wrappedMetrics[name] : [];
+      if (!Array.isArray(outputMetrics)) {
+        outputMetrics = [outputMetrics];
+      }
+      nestedMetrics.push(outputMetrics);
+    }
+    return nestedMetrics;
+  }
+}
+var LAYERS_MODEL_FORMAT_NAME = "layers-model";
+var LayersModel = class extends Container {
+  constructor(args) {
+    super(args);
+    this.isTraining = false;
+  }
+  summary(lineLength, positions, printFn = console.log) {
+    if (!this.built) {
+      throw new ValueError(`This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).`);
+    }
+    printSummary(this, lineLength, positions, printFn);
+  }
+  compile(args) {
+    if (args.loss == null) {
+      args.loss = [];
+    }
+    this.loss = args.loss;
+    if (typeof args.optimizer === "string") {
+      this.optimizer_ = getOptimizer(args.optimizer);
+      this.isOptimizerOwned = true;
+    } else {
+      if (!(args.optimizer instanceof Optimizer)) {
+        throw new ValueError(`User-defined optimizer must be an instance of tf.Optimizer.`);
+      }
+      this.optimizer_ = args.optimizer;
+      this.isOptimizerOwned = false;
+    }
+    let lossFunctions = [];
+    if (!Array.isArray(args.loss) && typeof args.loss !== "string" && typeof args.loss !== "function") {
+      args.loss = args.loss;
+      for (const name in args.loss) {
+        if (this.outputNames.indexOf(name) === -1) {
+          throw new ValueError(`Unknown entry in loss dictionary: "${name}". Only expected the following keys: ${this.outputNames}`);
+        }
+      }
+      for (const name of this.outputNames) {
+        if (args.loss[name] == null) {
+          console.warn(`Output "${name}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${name} during training`);
+        }
+        lossFunctions.push(get(args.loss[name]));
+      }
+    } else if (Array.isArray(args.loss)) {
+      if (args.loss.length !== this.outputs.length) {
+        throw new ValueError(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${args.loss}.`);
+      }
+      const theLosses = args.loss;
+      lossFunctions = theLosses.map((l) => get(l));
+    } else {
+      const lossFunction = get(args.loss);
+      this.outputs.forEach((_) => {
+        lossFunctions.push(lossFunction);
+      });
+    }
+    this.lossFunctions = lossFunctions;
+    this.feedOutputNames = [];
+    this.feedOutputShapes = [];
+    this.feedLossFns = [];
+    for (let i = 0; i < this.outputs.length; ++i) {
+      const shape = this.internalOutputShapes[i];
+      const name = this.outputNames[i];
+      this.feedOutputNames.push(name);
+      this.feedOutputShapes.push(shape);
+      this.feedLossFns.push(this.lossFunctions[i]);
+    }
+    const skipTargetIndices = [];
+    this.metrics = args.metrics;
+    this.metricsNames = ["loss"];
+    this.metricsTensors = [];
+    nameScope("loss", () => {
+      for (let i = 0; i < this.outputs.length; ++i) {
+        if (skipTargetIndices.indexOf(i) !== -1) {
+          continue;
+        }
+        const weightedLoss = this.lossFunctions[i];
+        if (this.outputs.length > 1) {
+          this.metricsTensors.push([weightedLoss, i]);
+          this.metricsNames.push(this.outputNames[i] + "_loss");
+        }
+      }
+    });
+    const nestedMetrics = collectMetrics(args.metrics, this.outputNames);
+    const appendMetric = (outputIndex, metricName, metricTensor) => {
+      if (this.outputNames.length > 1) {
+        metricName = this.outputNames[outputIndex] + "_" + metricName;
+      }
+      this.metricsNames.push(metricName);
+      this.metricsTensors.push([metricTensor, outputIndex]);
+    };
+    nameScope("metric", () => {
+      for (let i = 0; i < this.outputs.length; ++i) {
+        if (skipTargetIndices.indexOf(i) !== -1) {
+          continue;
+        }
+        const outputMetrics = nestedMetrics[i];
+        const handleMetrics = (metrics) => {
+          const metricNamePrefix = "";
+          let metricName;
+          let accFn;
+          let weightedMetricFn;
+          for (const metric of metrics) {
+            if (typeof metric === "string" && ["accuracy", "acc", "crossentropy", "ce"].indexOf(metric) !== -1) {
+              const outputShape = this.internalOutputShapes[i];
+              if (outputShape[outputShape.length - 1] === 1 || this.lossFunctions[i] === binaryCrossentropy) {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = binaryAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = binaryCrossentropy2;
+                }
+              } else if (this.lossFunctions[i] === sparseCategoricalCrossentropy) {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = sparseCategoricalAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = sparseCategoricalCrossentropy2;
+                }
+              } else {
+                if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                  accFn = categoricalAccuracy;
+                } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                  accFn = categoricalCrossentropy2;
+                }
+              }
+              let suffix;
+              if (["accuracy", "acc"].indexOf(metric) !== -1) {
+                suffix = "acc";
+              } else if (["crossentropy", "ce"].indexOf(metric) !== -1) {
+                suffix = "ce";
+              }
+              weightedMetricFn = accFn;
+              metricName = metricNamePrefix + suffix;
+            } else {
+              const metricFn = get2(metric);
+              weightedMetricFn = metricFn;
+              metricName = metricNamePrefix + getLossOrMetricName(metric);
+            }
+            let metricResult;
+            nameScope(metricName, () => {
+              metricResult = weightedMetricFn;
+            });
+            appendMetric(i, metricName, metricResult);
+          }
+        };
+        handleMetrics(outputMetrics);
+      }
+    });
+    this.collectedTrainableWeights = this.trainableWeights;
+  }
+  checkTrainableWeightsConsistency() {
+    if (this.collectedTrainableWeights == null) {
+      return;
+    }
+    if (this.trainableWeights.length !== this.collectedTrainableWeights.length) {
+      console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?");
+    }
+  }
+  evaluate(x, y, args = {}) {
+    const batchSize = args.batchSize == null ? 32 : args.batchSize;
+    checkBatchSize(batchSize);
+    const checkBatchAxis = true;
+    const standardizedOuts = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);
+    try {
+      const ins = standardizedOuts[0].concat(standardizedOuts[1]);
+      this.makeTestFunction();
+      const f = this.testFunction;
+      const testOuts = this.testLoop(f, ins, batchSize, args.verbose, args.steps);
+      return singletonOrArray(testOuts);
+    } finally {
+      disposeNewTensors(standardizedOuts[0], x);
+      disposeNewTensors(standardizedOuts[1], y);
+    }
+  }
+  async evaluateDataset(dataset, args) {
+    this.makeTestFunction();
+    return evaluateDataset(this, dataset, args);
+  }
+  checkNumSamples(ins, batchSize, steps, stepsName = "steps") {
+    let numSamples;
+    if (steps != null) {
+      numSamples = null;
+      if (batchSize != null) {
+        throw new ValueError(`If ${stepsName} is set, batchSize must be null or undefined.Got batchSize = ${batchSize}`);
+      }
+    } else if (ins != null) {
+      if (Array.isArray(ins)) {
+        numSamples = ins[0].shape[0];
+      } else {
+        numSamples = ins.shape[0];
+      }
+    } else {
+      throw new ValueError(`Either the input data should have a defined shape, or ${stepsName} shoud be specified.`);
+    }
+    return numSamples;
+  }
+  execute(inputs, outputs) {
+    if (Array.isArray(outputs) && outputs.length === 0) {
+      throw new ValueError("`outputs` is an empty Array, which is not allowed.");
+    }
+    const outputsIsArray = Array.isArray(outputs);
+    const outputNames = outputsIsArray ? outputs : [outputs];
+    const outputSymbolicTensors = this.retrieveSymbolicTensors(outputNames);
+    const feedDict = new FeedDict();
+    if (inputs instanceof Tensor) {
+      inputs = [inputs];
+    }
+    if (Array.isArray(inputs)) {
+      if (inputs.length !== this.inputs.length) {
+        throw new ValueError(`The number of inputs provided (${inputs.length}) does not match the number of inputs of this model (${this.inputs.length}).`);
+      }
+      for (let i = 0; i < this.inputs.length; ++i) {
+        feedDict.add(this.inputs[i], inputs[i]);
+      }
+    } else {
+      for (const input2 of this.inputs) {
+        const tensorValue = inputs[input2.name];
+        if (tensorValue == null) {
+          throw new ValueError(`No value is provided for the model's input ${input2.name}`);
+        }
+        feedDict.add(input2, tensorValue);
+      }
+    }
+    const executeOutputs = execute(outputSymbolicTensors, feedDict);
+    return outputsIsArray ? executeOutputs : executeOutputs[0];
+  }
+  retrieveSymbolicTensors(symbolicTensorNames) {
+    const outputSymbolicTensors = pyListRepeat(null, symbolicTensorNames.length);
+    let outputsRemaining = symbolicTensorNames.length;
+    for (const layer of this.layers) {
+      const layerOutputs = Array.isArray(layer.output) ? layer.output : [layer.output];
+      const layerOutputNames = layerOutputs.map((output) => output.name);
+      for (let i = 0; i < symbolicTensorNames.length; ++i) {
+        const index = layerOutputNames.indexOf(symbolicTensorNames[i]);
+        if (index !== -1) {
+          outputSymbolicTensors[i] = layerOutputs[index];
+          outputsRemaining--;
+        }
+        if (outputsRemaining === 0) {
+          break;
+        }
+      }
+      if (outputsRemaining === 0) {
+        break;
+      }
+    }
+    if (outputsRemaining > 0) {
+      const remainingNames = [];
+      outputSymbolicTensors.forEach((tensor2, i) => {
+        if (tensor2 == null) {
+          remainingNames.push(symbolicTensorNames[i]);
+        }
+      });
+      throw new ValueError(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(remainingNames)}`);
+    }
+    return outputSymbolicTensors;
+  }
+  predictLoop(ins, batchSize = 32, verbose = false) {
+    return tidy(() => {
+      const numSamples = this.checkNumSamples(ins);
+      if (verbose) {
+        throw new NotImplementedError("Verbose predictLoop() is not implemented yet.");
+      }
+      const batches = makeBatches(numSamples, batchSize);
+      const outsBatches = this.outputs.map((output) => []);
+      for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+        const batchOuts = tidy(() => {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const insBatch = sliceArrays(ins, batchStart, batchEnd);
+          const feeds = [];
+          if (Array.isArray(insBatch)) {
+            for (let i = 0; i < insBatch.length; ++i) {
+              feeds.push({ key: this.inputs[i], value: insBatch[i] });
+            }
+          } else {
+            feeds.push({ key: this.inputs[0], value: insBatch });
+          }
+          const feedDict = new FeedDict(feeds);
+          return execute(this.outputs, feedDict);
+        });
+        batchOuts.forEach((batchOut, i) => outsBatches[i].push(batchOut));
+      }
+      return singletonOrArray(outsBatches.map((batches2) => concat(batches2, 0)));
+    });
+  }
+  predict(x, args = {}) {
+    const xsRank2OrHigher = ensureTensorsRank2OrHigher(x);
+    checkInputData(xsRank2OrHigher, this.inputNames, this.feedInputShapes, false);
+    try {
+      const batchSize = args.batchSize == null ? 32 : args.batchSize;
+      checkBatchSize(batchSize);
+      return this.predictLoop(xsRank2OrHigher, batchSize);
+    } finally {
+      disposeNewTensors(xsRank2OrHigher, x);
+    }
+  }
+  predictOnBatch(x) {
+    checkInputData(x, this.inputNames, this.feedInputShapes, true);
+    const batchSize = (Array.isArray(x) ? x[0] : x).shape[0];
+    return this.predictLoop(x, batchSize);
+  }
+  standardizeUserDataXY(x, y, checkBatchAxis = true, batchSize) {
+    if (this.optimizer_ == null) {
+      throw new RuntimeError("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");
+    }
+    const outputShapes = [];
+    for (let i = 0; i < this.feedOutputShapes.length; ++i) {
+      const outputShape = this.feedOutputShapes[i];
+      const lossFn = this.feedLossFns[i];
+      if (lossFn === sparseCategoricalCrossentropy) {
+        outputShapes.push(outputShape.slice(0, outputShape.length - 1).concat([1]));
+      } else {
+        outputShapes.push(outputShape);
+      }
+    }
+    x = standardizeInputData(x, this.feedInputNames, this.feedInputShapes, false, "input");
+    y = standardizeInputData(y, this.feedOutputNames, outputShapes, false, "target");
+    checkArrayLengths(x, y, null);
+    checkLossAndTargetCompatibility(y, this.feedLossFns, this.feedOutputShapes);
+    if (this.stateful && batchSize != null && batchSize > 0) {
+      if (x[0].shape[0] % batchSize !== 0) {
+        throw new ValueError(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${batchSize}. Found: ${x[0].shape[0]} sample(s).`);
+      }
+    }
+    return [x, y];
+  }
+  async standardizeUserData(x, y, sampleWeight, classWeight, checkBatchAxis = true, batchSize) {
+    const [standardXs, standardYs] = this.standardizeUserDataXY(x, y, checkBatchAxis, batchSize);
+    if (sampleWeight != null) {
+      throw new Error("sample weight is not supported yet.");
+    }
+    let standardSampleWeights = null;
+    if (classWeight != null) {
+      const classWeights = standardizeClassWeights(classWeight, this.outputNames);
+      standardSampleWeights = [];
+      for (let i = 0; i < classWeights.length; ++i) {
+        standardSampleWeights.push(await standardizeWeights(standardYs[i], null, classWeights[i]));
+      }
+    }
+    return [standardXs, standardYs, standardSampleWeights];
+  }
+  testLoop(f, ins, batchSize, verbose = 0, steps) {
+    return tidy(() => {
+      const numSamples = this.checkNumSamples(ins, batchSize, steps, "steps");
+      const outs = [];
+      if (verbose > 0) {
+        throw new NotImplementedError("Verbose mode is not implemented yet.");
+      }
+      if (steps != null) {
+        throw new NotImplementedError("steps mode in testLoop() is not implemented yet");
+      } else {
+        const batches = makeBatches(numSamples, batchSize);
+        const indexArray = tensor1d(range2(0, numSamples));
+        for (let batchIndex = 0; batchIndex < batches.length; ++batchIndex) {
+          const batchStart = batches[batchIndex][0];
+          const batchEnd = batches[batchIndex][1];
+          const batchIds = sliceAlongFirstAxis(indexArray, batchStart, batchEnd - batchStart);
+          const insBatch = sliceArraysByIndices(ins, batchIds);
+          const batchOuts = f(insBatch);
+          if (batchIndex === 0) {
+            for (let i = 0; i < batchOuts.length; ++i) {
+              outs.push(scalar(0));
+            }
+          }
+          for (let i = 0; i < batchOuts.length; ++i) {
+            const batchOut = batchOuts[i];
+            outs[i] = add2(outs[i], mul(batchEnd - batchStart, batchOut));
+          }
+        }
+        for (let i = 0; i < outs.length; ++i) {
+          outs[i] = div(outs[i], numSamples);
+        }
+      }
+      return outs;
+    });
+  }
+  getDedupedMetricsNames() {
+    const outLabels = this.metricsNames;
+    const dedupedOutLabels = [];
+    for (let i = 0; i < outLabels.length; ++i) {
+      const label = outLabels[i];
+      let newLabel = label;
+      if (count(outLabels, label) > 1) {
+        const dupIndex = count(outLabels.slice(0, i), label);
+        newLabel += `_${dupIndex}`;
+      }
+      dedupedOutLabels.push(newLabel);
+    }
+    return dedupedOutLabels;
+  }
+  makeTrainFunction() {
+    return (data) => {
+      const lossValues = [];
+      const inputs = data.slice(0, this.inputs.length);
+      const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);
+      const sampleWeights = data.slice(this.inputs.length + this.outputs.length, this.inputs.length + this.outputs.length * 2);
+      const metricsValues = [];
+      const totalLossFunction = () => {
+        const feeds = [];
+        for (let i = 0; i < this.inputs.length; ++i) {
+          feeds.push({ key: this.inputs[i], value: inputs[i] });
+        }
+        const feedDict = new FeedDict(feeds);
+        const outputs = execute(this.outputs, feedDict, { "training": true });
+        let totalLoss;
+        for (let i = 0; i < this.lossFunctions.length; ++i) {
+          const lossFunction = this.lossFunctions[i];
+          let loss = lossFunction(targets[i], outputs[i]);
+          if (sampleWeights[i] != null) {
+            loss = computeWeightedLoss2(loss, sampleWeights[i]);
+          }
+          const meanLoss = mean(loss);
+          lossValues.push(meanLoss);
+          if (i === 0) {
+            totalLoss = loss;
+          } else {
+            totalLoss = add2(totalLoss, loss);
+          }
+        }
+        for (let i = 0; i < this.metricsTensors.length; ++i) {
+          let weightedMetric;
+          if (this.outputs.length > 1 && i < this.outputs.length) {
+            weightedMetric = lossValues[i];
+          } else {
+            const metric = this.metricsTensors[i][0];
+            const outputIndex = this.metricsTensors[i][1];
+            weightedMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));
+          }
+          keep(weightedMetric);
+          metricsValues.push(weightedMetric);
+        }
+        totalLoss = mean(totalLoss);
+        this.calculateLosses().forEach((regularizerLoss) => {
+          totalLoss = add2(totalLoss, regularizerLoss);
+        });
+        return totalLoss;
+      };
+      const variables = this.collectedTrainableWeights.map((param) => param.read());
+      const returnCost = true;
+      const totalLossValue = this.optimizer_.minimize(totalLossFunction, returnCost, variables);
+      return [totalLossValue].concat(metricsValues);
+    };
+  }
+  makeTestFunction() {
+    this.testFunction = (data) => {
+      return tidy(() => {
+        const valOutputs = [];
+        let totalLoss;
+        const inputs = data.slice(0, this.inputs.length);
+        const targets = data.slice(this.inputs.length, this.inputs.length + this.outputs.length);
+        const feeds = [];
+        for (let i = 0; i < this.inputs.length; ++i) {
+          feeds.push({ key: this.inputs[i], value: inputs[i] });
+        }
+        const feedDict = new FeedDict(feeds);
+        const outputs = execute(this.outputs, feedDict);
+        for (let i = 0; i < this.lossFunctions.length; ++i) {
+          const lossFunction = this.lossFunctions[i];
+          const loss = mean(lossFunction(targets[i], outputs[i]));
+          if (i === 0) {
+            totalLoss = loss;
+          } else {
+            totalLoss = add2(totalLoss, loss);
+          }
+          valOutputs.push(totalLoss);
+        }
+        for (let i = 0; i < this.metricsTensors.length; ++i) {
+          const metric = this.metricsTensors[i][0];
+          const outputIndex = this.metricsTensors[i][1];
+          const meanMetric = mean(metric(targets[outputIndex], outputs[outputIndex]));
+          valOutputs.push(meanMetric);
+        }
+        return valOutputs;
+      });
+    };
+  }
+  async fit(x, y, args = {}) {
+    return fitTensors(this, x, y, args);
+  }
+  async fitDataset(dataset, args) {
+    return fitDataset(this, dataset, args);
+  }
+  async trainOnBatch(x, y) {
+    const standardizeOut = await this.standardizeUserData(x, y);
+    const inputs = standardizeOut[0];
+    const targets = standardizeOut[1];
+    const trainFunction = this.makeTrainFunction();
+    const losses2 = trainFunction(inputs.concat(targets));
+    const lossValues = [];
+    for (const loss of losses2) {
+      const v = await loss.data();
+      lossValues.push(v[0]);
+    }
+    dispose(losses2);
+    disposeNewTensors(standardizeOut[0], x);
+    disposeNewTensors(standardizeOut[1], y);
+    return singletonOrArray(lossValues);
+  }
+  getNamedWeights(config) {
+    const namedWeights = [];
+    const trainableOnly = config != null && config.trainableOnly;
+    const weights = trainableOnly ? this.trainableWeights : this.weights;
+    const weightValues = this.getWeights(trainableOnly);
+    for (let i = 0; i < weights.length; ++i) {
+      if (trainableOnly && !weights[i].trainable) {
+        continue;
+      }
+      namedWeights.push({ name: weights[i].originalName, tensor: weightValues[i] });
+    }
+    return namedWeights;
+  }
+  set stopTraining(stop) {
+    this.stopTraining_ = stop;
+  }
+  get stopTraining() {
+    return this.stopTraining_;
+  }
+  get optimizer() {
+    return this.optimizer_;
+  }
+  set optimizer(optimizer) {
+    if (this.optimizer_ !== optimizer) {
+      this.optimizer_ = optimizer;
+      this.isOptimizerOwned = false;
+    }
+  }
+  dispose() {
+    const result = super.dispose();
+    if (result.refCountAfterDispose === 0 && this.optimizer != null && this.isOptimizerOwned) {
+      const numTensorsBeforeOptmizerDisposal = memory().numTensors;
+      this.optimizer_.dispose();
+      result.numDisposedVariables += numTensorsBeforeOptmizerDisposal - memory().numTensors;
+    }
+    return result;
+  }
+  getLossIdentifiers() {
+    let lossNames;
+    if (typeof this.loss === "string") {
+      lossNames = toSnakeCase(this.loss);
+    } else if (Array.isArray(this.loss)) {
+      for (const loss of this.loss) {
+        if (typeof loss !== "string") {
+          throw new Error("Serialization of non-string loss is not supported.");
+        }
+      }
+      lossNames = this.loss.map((name) => toSnakeCase(name));
+    } else {
+      const outputNames = Object.keys(this.loss);
+      lossNames = {};
+      const losses2 = this.loss;
+      for (const outputName of outputNames) {
+        if (typeof losses2[outputName] === "string") {
+          lossNames[outputName] = toSnakeCase(losses2[outputName]);
+        } else {
+          throw new Error("Serialization of non-string loss is not supported.");
+        }
+      }
+    }
+    return lossNames;
+  }
+  getMetricIdentifiers() {
+    if (typeof this.metrics === "string" || typeof this.metrics === "function") {
+      return [toSnakeCase(getLossOrMetricName(this.metrics))];
+    } else if (Array.isArray(this.metrics)) {
+      return this.metrics.map((metric) => toSnakeCase(getLossOrMetricName(metric)));
+    } else {
+      const metricsIdentifiers = {};
+      for (const key in this.metrics) {
+        metricsIdentifiers[key] = toSnakeCase(getLossOrMetricName(this.metrics[key]));
+      }
+      return metricsIdentifiers;
+    }
+  }
+  getTrainingConfig() {
+    return {
+      loss: this.getLossIdentifiers(),
+      metrics: this.getMetricIdentifiers(),
+      optimizer_config: {
+        class_name: this.optimizer.getClassName(),
+        config: this.optimizer.getConfig()
+      }
+    };
+  }
+  loadTrainingConfig(trainingConfig) {
+    if (trainingConfig.weighted_metrics != null) {
+      throw new Error("Loading weight_metrics is not supported yet.");
+    }
+    if (trainingConfig.loss_weights != null) {
+      throw new Error("Loading loss_weights is not supported yet.");
+    }
+    if (trainingConfig.sample_weight_mode != null) {
+      throw new Error("Loading sample_weight_mode is not supported yet.");
+    }
+    const tsConfig = convertPythonicToTs(trainingConfig.optimizer_config);
+    const optimizer = deserialize(tsConfig);
+    let loss;
+    if (typeof trainingConfig.loss === "string") {
+      loss = toCamelCase(trainingConfig.loss);
+    } else if (Array.isArray(trainingConfig.loss)) {
+      loss = trainingConfig.loss.map((lossEntry) => toCamelCase(lossEntry));
+    } else if (trainingConfig.loss != null) {
+      loss = {};
+      for (const key in trainingConfig.loss) {
+        loss[key] = toCamelCase(trainingConfig.loss[key]);
+      }
+    }
+    let metrics;
+    if (Array.isArray(trainingConfig.metrics)) {
+      metrics = trainingConfig.metrics.map((metric) => toCamelCase(metric));
+    } else if (trainingConfig.metrics != null) {
+      metrics = {};
+      for (const key in trainingConfig.metrics) {
+        metrics[key] = toCamelCase(trainingConfig.metrics[key]);
+      }
+    }
+    this.compile({ loss, metrics, optimizer });
+  }
+  async save(handlerOrURL, config) {
+    if (typeof handlerOrURL === "string") {
+      const handlers = io_exports.getSaveHandlers(handlerOrURL);
+      if (handlers.length === 0) {
+        throw new ValueError(`Cannot find any save handlers for URL '${handlerOrURL}'`);
+      } else if (handlers.length > 1) {
+        throw new ValueError(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);
+      }
+      handlerOrURL = handlers[0];
+    }
+    if (handlerOrURL.save == null) {
+      throw new ValueError("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+    }
+    const weightDataAndSpecs = await io_exports.encodeWeights(this.getNamedWeights(config));
+    const returnString = false;
+    const unusedArg = null;
+    const modelConfig = this.toJSON(unusedArg, returnString);
+    const modelArtifacts = {
+      modelTopology: modelConfig,
+      format: LAYERS_MODEL_FORMAT_NAME,
+      generatedBy: `TensorFlow.js tfjs-layers v${version2}`,
+      convertedBy: null
+    };
+    const includeOptimizer = config == null ? false : config.includeOptimizer;
+    if (includeOptimizer && this.optimizer != null) {
+      modelArtifacts.trainingConfig = this.getTrainingConfig();
+      const weightType = "optimizer";
+      const { data: optimizerWeightData, specs: optimizerWeightSpecs } = await io_exports.encodeWeights(await this.optimizer.getWeights(), weightType);
+      weightDataAndSpecs.specs.push(...optimizerWeightSpecs);
+      weightDataAndSpecs.data = io_exports.concatenateArrayBuffers([weightDataAndSpecs.data, optimizerWeightData]);
+    }
+    if (this.userDefinedMetadata != null) {
+      const checkSize = true;
+      checkUserDefinedMetadata(this.userDefinedMetadata, this.name, checkSize);
+      modelArtifacts.userDefinedMetadata = this.userDefinedMetadata;
+    }
+    modelArtifacts.weightData = weightDataAndSpecs.data;
+    modelArtifacts.weightSpecs = weightDataAndSpecs.specs;
+    return handlerOrURL.save(modelArtifacts);
+  }
+  setUserDefinedMetadata(userDefinedMetadata) {
+    checkUserDefinedMetadata(userDefinedMetadata, this.name);
+    this.userDefinedMetadata = userDefinedMetadata;
+  }
+  getUserDefinedMetadata() {
+    return this.userDefinedMetadata;
+  }
+};
+LayersModel.className = "Model";
+serialization_exports.registerClass(LayersModel);
+var Functional = class extends LayersModel {
+};
+Functional.className = "Functional";
+serialization_exports.registerClass(Functional);
+
+// node_modules/@tensorflow/tfjs-layers/dist/models.js
+async function modelFromJSON(modelAndWeightsConfig, customObjects) {
+  if (!("modelTopology" in modelAndWeightsConfig)) {
+    modelAndWeightsConfig = { modelTopology: modelAndWeightsConfig };
+  }
+  modelAndWeightsConfig = modelAndWeightsConfig;
+  let modelTopology = modelAndWeightsConfig.modelTopology;
+  if (modelTopology["model_config"] != null) {
+    modelTopology = modelTopology["model_config"];
+  }
+  const tsConfig = convertPythonicToTs(modelTopology);
+  const model2 = deserialize(tsConfig, customObjects);
+  if (modelAndWeightsConfig.weightsManifest != null) {
+    const weightValues = await io_exports.loadWeights(modelAndWeightsConfig.weightsManifest, modelAndWeightsConfig.pathPrefix, model2.weights.map((weight) => weight.originalName));
+    const uniqueWeightValues = {};
+    for (const weight of model2.weights) {
+      uniqueWeightValues[weight.originalName] = weightValues[weight.originalName];
+    }
+    model2.loadWeights(uniqueWeightValues);
+    dispose(weightValues);
+  }
+  return model2;
+}
+async function loadLayersModel(pathOrIOHandler, options) {
+  if (options == null) {
+    options = {};
+  }
+  if (typeof pathOrIOHandler === "string") {
+    const handlers = io_exports.getLoadHandlers(pathOrIOHandler, options);
+    if (handlers.length === 0) {
+      handlers.push(io_exports.browserHTTPRequest(pathOrIOHandler, options));
+    } else if (handlers.length > 1) {
+      throw new ValueError(`Found more than one (${handlers.length}) load handlers for URL '${pathOrIOHandler}'`);
+    }
+    pathOrIOHandler = handlers[0];
+  }
+  return loadLayersModelFromIOHandler(pathOrIOHandler, void 0, options);
+}
+async function loadLayersModelFromIOHandler(handler, customObjects, options) {
+  if (options == null) {
+    options = {};
+  }
+  if (handler.load == null) {
+    throw new ValueError("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+  }
+  const artifacts = await handler.load();
+  let modelTopology = artifacts.modelTopology;
+  if (modelTopology["model_config"] != null) {
+    modelTopology = modelTopology["model_config"];
+  }
+  const strict = options.strict == null ? true : options.strict;
+  const fastWeightInit = artifacts.weightData != null && artifacts.weightSpecs != null && strict;
+  const model2 = deserialize(convertPythonicToTs(modelTopology), customObjects, fastWeightInit);
+  const trainingConfig = artifacts.trainingConfig;
+  if (trainingConfig != null) {
+    model2.loadTrainingConfig(trainingConfig);
+  }
+  if (artifacts.userDefinedMetadata != null) {
+    model2.setUserDefinedMetadata(artifacts.userDefinedMetadata);
+  }
+  if (artifacts.weightData != null) {
+    if (artifacts.weightSpecs == null) {
+      throw new ValueError("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");
+    }
+    const { modelWeights, optimizerWeights } = decodeModelAndOptimizerWeights(artifacts.weightData, artifacts.weightSpecs);
+    model2.loadWeights(modelWeights, strict);
+    if (model2.optimizer != null && optimizerWeights.length > 0) {
+      await model2.optimizer.setWeights(optimizerWeights);
+    }
+    dispose(modelWeights);
+    dispose(optimizerWeights.map((w) => w.tensor));
+  }
+  return model2;
+}
+function decodeModelAndOptimizerWeights(buffer2, specs) {
+  const name2Tensor = io_exports.decodeWeights(buffer2, specs);
+  const modelWeights = {};
+  const optimizerWeights = [];
+  specs.forEach((spec) => {
+    if (spec.group === "optimizer") {
+      optimizerWeights.push({ name: spec.name, tensor: name2Tensor[spec.name] });
+    } else {
+      modelWeights[spec.name] = name2Tensor[spec.name];
+    }
+  });
+  return { modelWeights, optimizerWeights };
+}
+var Sequential = class extends LayersModel {
+  constructor(args) {
+    super({ inputs: [], outputs: [] });
+    args = args || {};
+    this.trainable = true;
+    this.built = false;
+    this.name = args.name != null ? args.name : getUid("sequential_");
+    if (args.layers != null) {
+      for (const layer of args.layers) {
+        this.add(layer);
+      }
+    }
+  }
+  checkShape(layer) {
+    const shape = layer.inboundNodes[0].outputTensors[0].shape;
+    if (shape.some((x) => x < 0)) {
+      throw new ValueError(`Negative dimension size caused by adding layer ${layer.name} with input shape [${layer.inboundNodes[0].inputTensors[0].shape}]`);
+    }
+  }
+  add(layer) {
+    const isLayerModelInstance = layer instanceof Sequential || layer instanceof LayersModel;
+    let modelLayer;
+    if (isLayerModelInstance) {
+      modelLayer = layer;
+      if (modelLayer.outputs.length !== 1) {
+        throw new ValueError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+      }
+      if (modelLayer.inputs.length !== 1) {
+        throw new ValueError("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.");
+      }
+    }
+    if (this.outputs.length === 0) {
+      if (layer.inboundNodes.length === 0) {
+        if (layer.batchInputShape == null) {
+          throw new ValueError("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");
+        }
+        const x = Input({
+          batchShape: layer.batchInputShape,
+          dtype: layer.dtype,
+          name: layer.name + "_input"
+        });
+        layer.apply(x);
+      }
+      if (isLayerModelInstance) {
+        this.outputs = modelLayer.outputs;
+        this.inputs = modelLayer.inputs;
+      } else {
+        if (layer.inboundNodes.length !== 1) {
+          throw new ValueError(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${layer.name} which has ${layer.inboundNodes.length} pre-existing inbound connections.`);
+        }
+        if (layer.inboundNodes[0].outputTensors.length !== 1) {
+          throw new ValueError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+        }
+        this.checkShape(layer);
+        this.outputs = [layer.inboundNodes[0].outputTensors[0]];
+        this.inputs = getSourceInputs(this.outputs[0]);
+      }
+      this.inboundNodes = [];
+      new Node({
+        outboundLayer: this,
+        inboundLayers: [],
+        nodeIndices: [],
+        tensorIndices: [],
+        inputTensors: this.inputs,
+        outputTensors: this.outputs,
+        inputMasks: pyListRepeat(null, this.inputs.length),
+        outputMasks: [null],
+        inputShapes: this.inputs.map((x) => x.shape),
+        outputShapes: this.outputs[0].shape
+      });
+    } else {
+      const outputTensor = layer.apply(this.outputs[0]);
+      if (Array.isArray(outputTensor)) {
+        throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");
+      }
+      this.checkShape(layer);
+      this.outputs = [outputTensor];
+      this.inboundNodes[0].outputTensors = this.outputs;
+      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];
+    }
+    this.layers.push(layer);
+    this.built = false;
+  }
+  pop() {
+    if (this.layers.length === 0) {
+      throw new TypeError("There are no layers in the model.");
+    }
+    this.layers.pop();
+    if (this.layers.length === 0) {
+      this.outputs = [];
+      this.inboundNodes = [];
+      this.outboundNodes = [];
+    } else {
+      const lastLayerIndex = this.layers.length - 1;
+      this.layers[lastLayerIndex].outboundNodes = [];
+      this.outputs = [this.layers[lastLayerIndex].output];
+      this.inboundNodes[0].outputTensors = this.outputs;
+      this.inboundNodes[0].outputShapes = [this.outputs[0].shape];
+    }
+  }
+  call(inputs, kwargs) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.call(inputs, kwargs);
+  }
+  build(inputShape) {
+    getExactlyOneShape(inputShape);
+    if (this.inputs.length === 0 || this.outputs.length === 0) {
+      throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");
+    }
+    this.model = new LayersModel({
+      inputs: this.inputs,
+      outputs: this.outputs[0],
+      name: this.name + "_model"
+    });
+    this.model.trainable = this.trainable;
+    this.supportsMasking = this.model.supportsMasking;
+    this.inputLayers = this.model.inputLayers;
+    this.inputLayersNodeIndices = this.model.inputLayersNodeIndices;
+    this.inputLayersTensorIndices = this.model.inputLayersTensorIndices;
+    this.outputLayers = this.model.outputLayers;
+    this.outputLayersNodeIndices = this.model.outputLayersNodeIndices;
+    this.outputLayersTensorIndices = this.model.outputLayersTensorIndices;
+    this.nodesByDepth = this.model.nodesByDepth;
+    this.containerNodes = this.model.containerNodes;
+    this.outputNames = this.model.outputNames;
+    this.inputNames = this.model.inputNames;
+    this.built = true;
+  }
+  countParams() {
+    if (!this.built) {
+      this.build();
+    }
+    return super.countParams();
+  }
+  summary(lineLength, positions, printFn = console.log) {
+    if (!this.built) {
+      this.build();
+    }
+    super.summary(lineLength, positions, printFn);
+  }
+  setWeights(weights) {
+    if (this.model == null) {
+      this.build();
+    }
+    this.model.setWeights(weights);
+  }
+  evaluate(x, y, args = {}) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.evaluate(x, y, args);
+  }
+  async evaluateDataset(dataset, args) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.evaluateDataset(dataset, args);
+  }
+  predict(x, args = {}) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.predict(x, args);
+  }
+  predictOnBatch(x) {
+    if (this.model == null) {
+      this.build();
+    }
+    return this.model.predictOnBatch(x);
+  }
+  compile(args) {
+    this.build();
+    this.model.compile(args);
+    this.optimizer_ = this.model.optimizer;
+    this.isOptimizerOwned = this.model.isOptimizerOwned;
+    this.loss = this.model.loss;
+    this.metrics = this.model.metrics;
+    this.metricsTensors = this.model.metricsTensors;
+    this.metricsNames = this.model.metricsNames;
+  }
+  get optimizer() {
+    return this.model == null ? void 0 : this.model.optimizer;
+  }
+  set optimizer(optimizer) {
+    this.model.optimizer = optimizer;
+  }
+  async fit(x, y, args = {}) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.fit(x, y, args);
+  }
+  async fitDataset(dataset, args) {
+    if (!this.built) {
+      throw new RuntimeError("The model needs to be compiled before being used.");
+    }
+    return this.model.fitDataset(dataset, args);
+  }
+  async trainOnBatch(x, y) {
+    return this.model.trainOnBatch(x, y);
+  }
+  static fromConfig(cls, config, customObjects = {}, fastWeightInit = false) {
+    let configArray;
+    let extraModelConfig = {};
+    if (config instanceof Array) {
+      if (!(config[0].className != null) || config[0]["className"] === "Merge") {
+        throw new ValueError("Legacy serialization format not supported yet.");
+      }
+      configArray = config;
+    } else {
+      util_exports.assert(config["layers"] != null, () => `When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field.`);
+      configArray = config["layers"];
+      delete config["layers"];
+      extraModelConfig = config;
+    }
+    const model2 = new cls(extraModelConfig);
+    if (!(model2 instanceof Sequential)) {
+      throw new NotImplementedError(`Sequential.fromConfig called on non-Sequential input: ${model2}`);
+    }
+    for (const conf of configArray) {
+      const customObjects2 = void 0;
+      const layer = deserialize(conf, customObjects2, fastWeightInit);
+      if (fastWeightInit) {
+        layer.setFastWeightInitDuringBuild(true);
+      }
+      model2.add(layer);
+    }
+    return model2;
+  }
+  set stopTraining(stop) {
+    if (this.model == null) {
+      throw new ValueError("Cannot set the stopTraining property of a sequential model before it is compiled.");
+    }
+    this.model.stopTraining = stop;
+  }
+  get stopTraining() {
+    if (this.model == null) {
+      throw new ValueError("Cannot get the stopTraining property of a sequential model before it is compiled.");
+    }
+    return this.model.stopTraining;
+  }
+  getConfig() {
+    const layers = [];
+    for (const layer of this.layers) {
+      const dict = {};
+      dict["className"] = layer.getClassName();
+      dict["config"] = layer.getConfig();
+      layers.push(dict);
+    }
+    return { name: this.name, layers };
+  }
+};
+Sequential.className = "Sequential";
+serialization_exports.registerClass(Sequential);
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports.js
+function model(args) {
+  return new LayersModel(args);
+}
+function sequential(config) {
+  return new Sequential(config);
+}
+function input(config) {
+  return Input(config);
+}
+function registerCallbackConstructor(verbosityLevel, callbackConstructor) {
+  CallbackConstructorRegistry.registerCallbackConstructor(verbosityLevel, callbackConstructor);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/activations.js
+var Activation = class extends serialization_exports.Serializable {
+  getConfig() {
+    return {};
+  }
+};
+var Elu2 = class extends Activation {
+  apply(x, alpha = 1) {
+    return elu2(x, alpha);
+  }
+};
+Elu2.className = "elu";
+serialization_exports.registerClass(Elu2);
+var Selu2 = class extends Activation {
+  apply(x) {
+    return selu(x);
+  }
+};
+Selu2.className = "selu";
+serialization_exports.registerClass(Selu2);
+var Relu2 = class extends Activation {
+  apply(x) {
+    return relu(x);
+  }
+};
+Relu2.className = "relu";
+serialization_exports.registerClass(Relu2);
+var Relu62 = class extends Activation {
+  apply(x) {
+    return tidy(() => minimum(6, relu(x)));
+  }
+};
+Relu62.className = "relu6";
+serialization_exports.registerClass(Relu62);
+var Linear = class extends Activation {
+  apply(x) {
+    return x;
+  }
+};
+Linear.className = "linear";
+serialization_exports.registerClass(Linear);
+var Sigmoid2 = class extends Activation {
+  apply(x) {
+    return sigmoid(x);
+  }
+};
+Sigmoid2.className = "sigmoid";
+serialization_exports.registerClass(Sigmoid2);
+var HardSigmoid = class extends Activation {
+  apply(x) {
+    return hardSigmoid(x);
+  }
+};
+HardSigmoid.className = "hardSigmoid";
+serialization_exports.registerClass(HardSigmoid);
+var Softplus2 = class extends Activation {
+  apply(x) {
+    return softplus(x);
+  }
+};
+Softplus2.className = "softplus";
+serialization_exports.registerClass(Softplus2);
+var Softsign = class extends Activation {
+  apply(x) {
+    return softsign(x);
+  }
+};
+Softsign.className = "softsign";
+serialization_exports.registerClass(Softsign);
+var Tanh2 = class extends Activation {
+  apply(x) {
+    return tanh2(x);
+  }
+};
+Tanh2.className = "tanh";
+serialization_exports.registerClass(Tanh2);
+var Softmax2 = class extends Activation {
+  apply(x, axis = -1) {
+    return softmax(x, axis);
+  }
+};
+Softmax2.className = "softmax";
+serialization_exports.registerClass(Softmax2);
+var LogSoftmax2 = class extends Activation {
+  apply(x, axis = -1) {
+    return logSoftmax(x, axis);
+  }
+};
+LogSoftmax2.className = "logSoftmax";
+serialization_exports.registerClass(LogSoftmax2);
+var Swish = class extends Activation {
+  apply(x, alpha = 1) {
+    return tidy(() => mul(sigmoid(mul(x, alpha)), x));
+  }
+};
+Swish.className = "swish";
+serialization_exports.registerClass(Swish);
+var Mish = class extends Activation {
+  apply(x) {
+    return tidy(() => mul(x, tanh2(softplus(x))));
+  }
+};
+Mish.className = "mish";
+serialization_exports.registerClass(Mish);
+function serializeActivation(activation2) {
+  return activation2.getClassName();
+}
+function deserializeActivation(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "activation");
+}
+function getActivation(identifier) {
+  if (identifier == null) {
+    const config = {};
+    config["className"] = "linear";
+    config["config"] = {};
+    return deserializeActivation(config);
+  }
+  if (typeof identifier === "string") {
+    const config = {};
+    config["className"] = identifier;
+    config["config"] = {};
+    return deserializeActivation(config);
+  } else if (identifier instanceof Activation) {
+    return identifier;
+  } else {
+    return deserializeActivation(identifier);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/regularizers.js
+function assertObjectArgs(args) {
+  if (args != null && typeof args !== "object") {
+    throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${args}`);
+  }
+}
+var Regularizer = class extends serialization_exports.Serializable {
+};
+var L1L2 = class extends Regularizer {
+  constructor(args) {
+    super();
+    assertObjectArgs(args);
+    this.l1 = args == null || args.l1 == null ? 0.01 : args.l1;
+    this.l2 = args == null || args.l2 == null ? 0.01 : args.l2;
+    this.hasL1 = this.l1 !== 0;
+    this.hasL2 = this.l2 !== 0;
+  }
+  apply(x) {
+    return tidy(() => {
+      let regularization = zeros([1]);
+      if (this.hasL1) {
+        regularization = add2(regularization, sum2(mul(this.l1, abs(x))));
+      }
+      if (this.hasL2) {
+        regularization = add2(regularization, sum2(mul(this.l2, square2(x))));
+      }
+      return reshape(regularization, []);
+    });
+  }
+  getConfig() {
+    return { "l1": this.l1, "l2": this.l2 };
+  }
+  static fromConfig(cls, config) {
+    return new cls({ l1: config["l1"], l2: config["l2"] });
+  }
+};
+L1L2.className = "L1L2";
+serialization_exports.registerClass(L1L2);
+function l1(args) {
+  assertObjectArgs(args);
+  return new L1L2({ l1: args != null ? args.l1 : null, l2: 0 });
+}
+function l2(args) {
+  assertObjectArgs(args);
+  return new L1L2({ l2: args != null ? args.l2 : null, l1: 0 });
+}
+var REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP = {
+  "l1l2": "L1L2"
+};
+function serializeRegularizer(constraint) {
+  return serializeKerasObject(constraint);
+}
+function deserializeRegularizer(config, customObjects = {}) {
+  return deserializeKerasObject(config, serialization_exports.SerializationMap.getMap().classNameMap, customObjects, "regularizer");
+}
+function getRegularizer(identifier) {
+  if (identifier == null) {
+    return null;
+  }
+  if (typeof identifier === "string") {
+    const className = identifier in REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP ? REGULARIZER_IDENTIFIER_REGISTRY_SYMBOL_MAP[identifier] : identifier;
+    const config = { className, config: {} };
+    return deserializeRegularizer(config);
+  } else if (identifier instanceof Regularizer) {
+    return identifier;
+  } else {
+    return deserializeRegularizer(identifier);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/advanced_activations.js
+var ReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.supportsMasking = true;
+    if (args != null) {
+      this.maxValue = args.maxValue;
+    }
+  }
+  call(inputs, kwargs) {
+    inputs = getExactlyOneTensor(inputs);
+    let output = relu(inputs);
+    if (this.maxValue != null) {
+      output = clipByValue(output, 0, this.maxValue);
+    }
+    return output;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { maxValue: this.maxValue };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ReLU.className = "ReLU";
+serialization_exports.registerClass(ReLU);
+var LeakyReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA = 0.3;
+    if (args == null) {
+      args = {};
+    }
+    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return leakyRelu(x, this.alpha);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { alpha: this.alpha };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+LeakyReLU.className = "LeakyReLU";
+serialization_exports.registerClass(LeakyReLU);
+var PReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA_INITIALIZER = "zeros";
+    if (args == null) {
+      args = {};
+    }
+    this.supportsMasking = true;
+    this.alphaInitializer = getInitializer(args.alphaInitializer || this.DEFAULT_ALPHA_INITIALIZER);
+    this.alphaRegularizer = getRegularizer(args.alphaRegularizer);
+    this.alphaConstraint = getConstraint(args.alphaConstraint);
+    if (args.sharedAxes == null) {
+      this.sharedAxes = null;
+    } else if (Array.isArray(args.sharedAxes)) {
+      this.sharedAxes = args.sharedAxes;
+    } else if (typeof args.sharedAxes === "number") {
+      this.sharedAxes = [args.sharedAxes];
+    } else {
+      throw new ValueError(`Expected sharedAxes to be a number or an array of numbers, but got ${args.sharedAxes}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const paramShape = inputShape.slice(1);
+    if (this.sharedAxes != null) {
+      for (const i of this.sharedAxes) {
+        paramShape[i - 1] = 1;
+      }
+    }
+    this.alpha = this.addWeight("alpha", paramShape, "float32", this.alphaInitializer, this.alphaRegularizer, true, this.alphaConstraint);
+    const axes = {};
+    if (this.sharedAxes != null) {
+      for (let i = 1; i < inputShape.length; ++i) {
+        axes[i] = inputShape[i];
+      }
+    }
+    this.inputSpec = [new InputSpec({
+      ndim: inputShape.length,
+      axes
+    })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    inputs = getExactlyOneTensor(inputs);
+    return prelu(inputs, this.alpha.read());
+  }
+  getConfig() {
+    const config = {
+      alphaInitializer: serializeInitializer(this.alphaInitializer),
+      alphaRegularizer: serializeRegularizer(this.alphaRegularizer),
+      alphaConstraint: serializeConstraint(this.alphaConstraint),
+      sharedAxes: this.sharedAxes
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+PReLU.className = "PReLU";
+serialization_exports.registerClass(PReLU);
+var ELU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_ALPHA = 1;
+    if (args == null) {
+      args = {};
+    }
+    if (args.alpha != null && args.alpha !== this.DEFAULT_ALPHA) {
+      throw new NotImplementedError(`Non-default alpha value (${args.alpha}) is not supported by the ELU layer yet.`);
+    }
+    this.alpha = args.alpha == null ? this.DEFAULT_ALPHA : args.alpha;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return elu(x);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { alpha: this.alpha };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ELU.className = "ELU";
+serialization_exports.registerClass(ELU);
+var ThresholdedReLU = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_THETA = 1;
+    if (args == null) {
+      args = {};
+    }
+    this.theta = args.theta == null ? this.DEFAULT_THETA : args.theta;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return mul(x, cast(greater(x, this.theta), "float32"));
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { theta: this.theta };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ThresholdedReLU.className = "ThresholdedReLU";
+serialization_exports.registerClass(ThresholdedReLU);
+var Softmax3 = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.DEFAULT_AXIS = 1;
+    if (args == null) {
+      args = {};
+    }
+    this.softmax = new Softmax2().apply;
+    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;
+  }
+  call(inputs, kwargs) {
+    const x = getExactlyOneTensor(inputs);
+    return this.softmax(x, this.axis);
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const config = { axis: this.axis };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Softmax3.className = "Softmax";
+serialization_exports.registerClass(Softmax3);
+
+// node_modules/@tensorflow/tfjs-layers/dist/utils/conv_utils.js
+function normalizeArray(value, n, name) {
+  if (typeof value === "number") {
+    return pyListRepeat(value, n);
+  } else {
+    if (value.length !== n) {
+      throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${value.length} elements.`);
+    }
+    for (let i = 0; i < n; ++i) {
+      const singleValue = value[i];
+      if (!isInteger(singleValue)) {
+        throw new ValueError(`The ${name} argument must be an integer or tuple of ${n} integers. Received: ${JSON.stringify(value)} including a non-integer number ${singleValue}`);
+      }
+    }
+    return value;
+  }
+}
+function convOutputLength(inputLength, filterSize, padding, stride, dilation = 1) {
+  if (inputLength == null) {
+    return inputLength;
+  }
+  const dilatedFilterSize = filterSize + (filterSize - 1) * (dilation - 1);
+  let outputLength;
+  if (padding === "same") {
+    outputLength = inputLength;
+  } else {
+    outputLength = inputLength - dilatedFilterSize + 1;
+  }
+  return Math.floor((outputLength + stride - 1) / stride);
+}
+function deconvLength(dimSize, strideSize, kernelSize, padding) {
+  if (dimSize == null) {
+    return null;
+  }
+  if (padding === "valid") {
+    dimSize = dimSize * strideSize + max2([kernelSize - strideSize, 0]);
+  } else if (padding === "same") {
+    dimSize = dimSize * strideSize;
+  } else {
+    throw new ValueError(`Unsupport padding mode: ${padding}.`);
+  }
+  return dimSize;
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional.js
+function preprocessConv2DInput(x, dataFormat) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    if (dataFormat === "channelsFirst") {
+      return transpose(x, [0, 2, 3, 1]);
+    } else {
+      return x;
+    }
+  });
+}
+function preprocessConv3DInput(x, dataFormat) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    if (dataFormat === "channelsFirst") {
+      return transpose(x, [0, 2, 3, 4, 1]);
+    } else {
+      return x;
+    }
+  });
+}
+function conv1dWithBias(x, kernel, bias, strides = 1, padding = "valid", dataFormat, dilationRate = 1) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.shape.length !== 3) {
+      throw new ValueError(`The input of a conv1dWithBias operation should be 3, but is ${x.shape.length} instead.`);
+    }
+    if (kernel.shape.length !== 3) {
+      throw new ValueError(`The kernel for a conv1dWithBias operation should be 3, but is ${kernel.shape.length} instead`);
+    }
+    if (bias != null && bias.shape.length !== 1) {
+      throw new ValueError(`The bias for a conv1dWithBias operation should be 1, but is ${kernel.shape.length} instead`);
+    }
+    if (dataFormat === "channelsFirst") {
+      x = transpose(x, [0, 2, 1]);
+    }
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+    }
+    let y = conv1d(x, kernel, strides, padding === "same" ? "same" : "valid", "NWC", dilationRate);
+    if (bias != null) {
+      y = biasAdd(y, bias);
+    }
+    return y;
+  });
+}
+function conv2dWithBiasActivation(x, kernel, bias, strides = [1, 1], padding = "valid", dataFormat, dilationRate, activation2 = null) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.rank !== 3 && x.rank !== 4) {
+      throw new ValueError(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${x.rank}.`);
+    }
+    if (kernel.rank !== 3 && kernel.rank !== 4) {
+      throw new ValueError(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${x.rank}.`);
+    }
+    let y = preprocessConv2DInput(x, dataFormat);
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");
+    }
+    y = fused_ops_exports.conv2d({
+      x: y,
+      filter: kernel,
+      strides,
+      pad: padding === "same" ? "same" : "valid",
+      dilations: dilationRate,
+      dataFormat: "NHWC",
+      bias,
+      activation: activation2
+    });
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+function conv3dWithBias(x, kernel, bias, strides = [1, 1, 1], padding = "valid", dataFormat, dilationRate) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    if (x.rank !== 4 && x.rank !== 5) {
+      throw new ValueError(`conv3dWithBias expects input to be of rank 4 or 5, but received ${x.rank}.`);
+    }
+    if (kernel.rank !== 4 && kernel.rank !== 5) {
+      throw new ValueError(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${x.rank}.`);
+    }
+    let y = preprocessConv3DInput(x, dataFormat);
+    if (padding === "causal") {
+      throw new NotImplementedError("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");
+    }
+    y = conv3d(y, kernel, strides, padding === "same" ? "same" : "valid", "NDHWC", dilationRate);
+    if (bias != null) {
+      y = biasAdd(y, bias);
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 4, 1, 2, 3]);
+    }
+    return y;
+  });
+}
+var BaseConv = class extends Layer {
+  constructor(rank, args) {
+    super(args);
+    this.bias = null;
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    BaseConv.verifyArgs(args);
+    this.rank = rank;
+    assertPositiveInteger(this.rank, "rank");
+    if (this.rank !== 1 && this.rank !== 2 && this.rank !== 3) {
+      throw new NotImplementedError(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);
+    }
+    this.kernelSize = normalizeArray(args.kernelSize, rank, "kernelSize");
+    this.strides = normalizeArray(args.strides == null ? 1 : args.strides, rank, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    checkPaddingMode(this.padding);
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.activation = getActivation(args.activation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.dilationRate = normalizeArray(args.dilationRate == null ? 1 : args.dilationRate, rank, "dilationRate");
+    if (this.rank === 1 && (Array.isArray(this.dilationRate) && this.dilationRate.length !== 1)) {
+      throw new ValueError(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+    } else if (this.rank === 2) {
+      if (typeof this.dilationRate === "number") {
+        this.dilationRate = [this.dilationRate, this.dilationRate];
+      } else if (this.dilationRate.length !== 2) {
+        throw new ValueError(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+      }
+    } else if (this.rank === 3) {
+      if (typeof this.dilationRate === "number") {
+        this.dilationRate = [this.dilationRate, this.dilationRate, this.dilationRate];
+      } else if (this.dilationRate.length !== 3) {
+        throw new ValueError(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`);
+      }
+    }
+  }
+  static verifyArgs(args) {
+    assert2("kernelSize" in args, `required key 'kernelSize' not in config`);
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 3)) {
+      throw new ValueError(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+  getConfig() {
+    const config = {
+      kernelSize: this.kernelSize,
+      strides: this.strides,
+      padding: this.padding,
+      dataFormat: this.dataFormat,
+      dilationRate: this.dilationRate,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      biasConstraint: serializeConstraint(this.biasConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var Conv = class extends BaseConv {
+  constructor(rank, args) {
+    super(rank, args);
+    this.kernel = null;
+    Conv.verifyArgs(args);
+    this.filters = args.filters;
+    assertPositiveInteger(this.filters, "filters");
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([inputDim, this.filters]);
+    this.kernel = this.addWeight("kernel", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [{ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } }];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let outputs;
+      const biasValue = this.bias == null ? null : this.bias.read();
+      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());
+      if (fusedActivationName != null && this.rank === 2) {
+        outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate, fusedActivationName);
+      } else {
+        if (this.rank === 1) {
+          outputs = conv1dWithBias(inputs, this.kernel.read(), biasValue, this.strides[0], this.padding, this.dataFormat, this.dilationRate[0]);
+        } else if (this.rank === 2) {
+          outputs = conv2dWithBiasActivation(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);
+        } else if (this.rank === 3) {
+          outputs = conv3dWithBias(inputs, this.kernel.read(), biasValue, this.strides, this.padding, this.dataFormat, this.dilationRate);
+        } else {
+          throw new NotImplementedError("convolutions greater than 3D are not implemented yet.");
+        }
+        if (this.activation != null) {
+          outputs = this.activation.apply(outputs);
+        }
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const newSpace = [];
+    const space = this.dataFormat === "channelsLast" ? inputShape.slice(1, inputShape.length - 1) : inputShape.slice(2);
+    for (let i = 0; i < space.length; ++i) {
+      const newDim = convOutputLength(space[i], this.kernelSize[i], this.padding, this.strides[i], typeof this.dilationRate === "number" ? this.dilationRate : this.dilationRate[i]);
+      newSpace.push(newDim);
+    }
+    let outputShape = [inputShape[0]];
+    if (this.dataFormat === "channelsLast") {
+      outputShape = outputShape.concat(newSpace);
+      outputShape.push(this.filters);
+    } else {
+      outputShape.push(this.filters);
+      outputShape = outputShape.concat(newSpace);
+    }
+    return outputShape;
+  }
+  getConfig() {
+    const config = {
+      filters: this.filters,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  static verifyArgs(args) {
+    if (!("filters" in args) || typeof args.filters !== "number" || args.filters < 1) {
+      throw new ValueError(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(args.filters)}`);
+    }
+  }
+};
+var Conv2D2 = class extends Conv {
+  constructor(args) {
+    super(2, args);
+    Conv2D2.verifyArgs(args);
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 2)) {
+      throw new ValueError(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+};
+Conv2D2.className = "Conv2D";
+serialization_exports.registerClass(Conv2D2);
+var Conv3D2 = class extends Conv {
+  constructor(args) {
+    super(3, args);
+    Conv3D2.verifyArgs(args);
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number") {
+      if (!(Array.isArray(args.kernelSize) && (args.kernelSize.length === 1 || args.kernelSize.length === 3))) {
+        throw new ValueError(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(args.kernelSize)}.`);
+      }
+    }
+  }
+};
+Conv3D2.className = "Conv3D";
+serialization_exports.registerClass(Conv3D2);
+var Conv2DTranspose = class extends Conv2D2 {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+    if (this.padding !== "same" && this.padding !== "valid") {
+      throw new ValueError(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length !== 4) {
+      throw new ValueError("Input should have rank 4; Received input shape: " + JSON.stringify(inputShape));
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError("The channel dimension of the inputs should be defined. Found `None`.");
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);
+    this.kernel = this.addWeight("kernel", kernelShape, "float32", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [new InputSpec({ ndim: 4, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.shape.length !== 4) {
+        throw new ValueError(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);
+      }
+      const inputShape = input2.shape;
+      const batchSize = inputShape[0];
+      let hAxis;
+      let wAxis;
+      if (this.dataFormat === "channelsFirst") {
+        hAxis = 2;
+        wAxis = 3;
+      } else {
+        hAxis = 1;
+        wAxis = 2;
+      }
+      const height = inputShape[hAxis];
+      const width = inputShape[wAxis];
+      const kernelH = this.kernelSize[0];
+      const kernelW = this.kernelSize[1];
+      const strideH = this.strides[0];
+      const strideW = this.strides[1];
+      const outHeight = deconvLength(height, strideH, kernelH, this.padding);
+      const outWidth = deconvLength(width, strideW, kernelW, this.padding);
+      const outputShape = [batchSize, outHeight, outWidth, this.filters];
+      if (this.dataFormat !== "channelsLast") {
+        input2 = transpose(input2, [0, 2, 3, 1]);
+      }
+      let outputs = conv2dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);
+      if (this.dataFormat !== "channelsLast") {
+        outputs = transpose(outputs, [0, 3, 1, 2]);
+      }
+      if (this.bias != null) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    let channelAxis;
+    let heightAxis;
+    let widthAxis;
+    if (this.dataFormat === "channelsFirst") {
+      channelAxis = 1;
+      heightAxis = 2;
+      widthAxis = 3;
+    } else {
+      channelAxis = 3;
+      heightAxis = 1;
+      widthAxis = 2;
+    }
+    const kernelH = this.kernelSize[0];
+    const kernelW = this.kernelSize[1];
+    const strideH = this.strides[0];
+    const strideW = this.strides[1];
+    outputShape[channelAxis] = this.filters;
+    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);
+    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);
+    return outputShape;
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["dilationRate"];
+    return config;
+  }
+};
+Conv2DTranspose.className = "Conv2DTranspose";
+serialization_exports.registerClass(Conv2DTranspose);
+var Conv3DTranspose = class extends Conv3D2 {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+    if (this.padding !== "same" && this.padding !== "valid") {
+      throw new ValueError(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`);
+    }
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length !== 5) {
+      throw new ValueError("Input should have rank 5; Received input shape: " + JSON.stringify(inputShape));
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError("The channel dimension of the inputs should be defined. Found `None`.");
+    }
+    const inputDim = inputShape[channelAxis];
+    const kernelShape = this.kernelSize.concat([this.filters, inputDim]);
+    this.kernel = this.addWeight("kernel", kernelShape, "float32", this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.inputSpec = [new InputSpec({ ndim: 5, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.shape.length !== 5) {
+        throw new ValueError(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${input2.shape.length}`);
+      }
+      const inputShape = input2.shape;
+      const batchSize = inputShape[0];
+      let hAxis;
+      let wAxis;
+      let dAxis;
+      if (this.dataFormat === "channelsFirst") {
+        dAxis = 2;
+        hAxis = 3;
+        wAxis = 4;
+      } else {
+        dAxis = 1;
+        hAxis = 2;
+        wAxis = 3;
+      }
+      const depth = inputShape[dAxis];
+      const height = inputShape[hAxis];
+      const width = inputShape[wAxis];
+      const kernelD = this.kernelSize[0];
+      const kernelH = this.kernelSize[1];
+      const kernelW = this.kernelSize[2];
+      const strideD = this.strides[0];
+      const strideH = this.strides[1];
+      const strideW = this.strides[2];
+      const outDepth = deconvLength(depth, strideD, kernelD, this.padding);
+      const outHeight = deconvLength(height, strideH, kernelH, this.padding);
+      const outWidth = deconvLength(width, strideW, kernelW, this.padding);
+      const outputShape = [batchSize, outDepth, outHeight, outWidth, this.filters];
+      if (this.dataFormat !== "channelsLast") {
+        input2 = transpose(input2, [0, 2, 3, 4, 1]);
+      }
+      let outputs = conv3dTranspose(input2, this.kernel.read(), outputShape, this.strides, this.padding);
+      if (this.dataFormat !== "channelsLast") {
+        outputs = transpose(outputs, [0, 4, 1, 2, 3]);
+      }
+      if (this.bias !== null) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation !== null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    let channelAxis;
+    let depthAxis;
+    let heightAxis;
+    let widthAxis;
+    if (this.dataFormat === "channelsFirst") {
+      channelAxis = 1;
+      depthAxis = 2;
+      heightAxis = 3;
+      widthAxis = 4;
+    } else {
+      channelAxis = 4;
+      depthAxis = 1;
+      heightAxis = 2;
+      widthAxis = 3;
+    }
+    const kernelD = this.kernelSize[0];
+    const kernelH = this.kernelSize[1];
+    const kernelW = this.kernelSize[2];
+    const strideD = this.strides[0];
+    const strideH = this.strides[1];
+    const strideW = this.strides[2];
+    outputShape[channelAxis] = this.filters;
+    outputShape[depthAxis] = deconvLength(outputShape[depthAxis], strideD, kernelD, this.padding);
+    outputShape[heightAxis] = deconvLength(outputShape[heightAxis], strideH, kernelH, this.padding);
+    outputShape[widthAxis] = deconvLength(outputShape[widthAxis], strideW, kernelW, this.padding);
+    return outputShape;
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["dilationRate"];
+    return config;
+  }
+};
+Conv3DTranspose.className = "Conv3DTranspose";
+serialization_exports.registerClass(Conv3DTranspose);
+var SeparableConv = class extends Conv {
+  constructor(rank, config) {
+    super(rank, config);
+    this.DEFAULT_DEPTHWISE_INITIALIZER = "glorotUniform";
+    this.DEFAULT_POINTWISE_INITIALIZER = "glorotUniform";
+    this.depthwiseKernel = null;
+    this.pointwiseKernel = null;
+    if (config.filters == null) {
+      throw new ValueError("The `filters` configuration field is required by SeparableConv, but is unspecified.");
+    }
+    if (config.kernelInitializer != null || config.kernelRegularizer != null || config.kernelConstraint != null) {
+      throw new ValueError("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");
+    }
+    if (config.padding != null && config.padding !== "same" && config.padding !== "valid") {
+      throw new ValueError(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(config.padding)}`);
+    }
+    this.depthMultiplier = config.depthMultiplier == null ? 1 : config.depthMultiplier;
+    this.depthwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_DEPTHWISE_INITIALIZER);
+    this.depthwiseRegularizer = getRegularizer(config.depthwiseRegularizer);
+    this.depthwiseConstraint = getConstraint(config.depthwiseConstraint);
+    this.pointwiseInitializer = getInitializer(config.depthwiseInitializer || this.DEFAULT_POINTWISE_INITIALIZER);
+    this.pointwiseRegularizer = getRegularizer(config.pointwiseRegularizer);
+    this.pointwiseConstraint = getConstraint(config.pointwiseConstraint);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < this.rank + 2) {
+      throw new ValueError(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank + 2}, but received input shape: ${JSON.stringify(inputShape)}`);
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {
+      throw new ValueError(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(inputShape[channelAxis])}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const depthwiseKernelShape = this.kernelSize.concat([inputDim, this.depthMultiplier]);
+    const pointwiseKernelShape = [];
+    for (let i = 0; i < this.rank; ++i) {
+      pointwiseKernelShape.push(1);
+    }
+    pointwiseKernelShape.push(inputDim * this.depthMultiplier, this.filters);
+    const trainable = true;
+    this.depthwiseKernel = this.addWeight("depthwise_kernel", depthwiseKernelShape, "float32", this.depthwiseInitializer, this.depthwiseRegularizer, trainable, this.depthwiseConstraint);
+    this.pointwiseKernel = this.addWeight("pointwise_kernel", pointwiseKernelShape, "float32", this.pointwiseInitializer, this.pointwiseRegularizer, trainable, this.pointwiseConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.filters], "float32", this.biasInitializer, this.biasRegularizer, trainable, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.inputSpec = [new InputSpec({ ndim: this.rank + 2, axes: { [channelAxis]: inputDim } })];
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let output;
+      if (this.rank === 1) {
+        throw new NotImplementedError("1D separable convolution is not implemented yet.");
+      } else if (this.rank === 2) {
+        if (this.dataFormat === "channelsFirst") {
+          inputs = transpose(inputs, [0, 2, 3, 1]);
+        }
+        output = separableConv2d(inputs, this.depthwiseKernel.read(), this.pointwiseKernel.read(), this.strides, this.padding, this.dilationRate, "NHWC");
+      }
+      if (this.useBias) {
+        output = biasAdd(output, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        output = this.activation.apply(output);
+      }
+      if (this.dataFormat === "channelsFirst") {
+        output = transpose(output, [0, 3, 1, 2]);
+      }
+      return output;
+    });
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    delete config["kernelInitializer"];
+    delete config["kernelRegularizer"];
+    delete config["kernelConstraint"];
+    config["depthwiseInitializer"] = serializeInitializer(this.depthwiseInitializer);
+    config["pointwiseInitializer"] = serializeInitializer(this.pointwiseInitializer);
+    config["depthwiseRegularizer"] = serializeRegularizer(this.depthwiseRegularizer);
+    config["pointwiseRegularizer"] = serializeRegularizer(this.pointwiseRegularizer);
+    config["depthwiseConstraint"] = serializeConstraint(this.depthwiseConstraint);
+    config["pointwiseConstraint"] = serializeConstraint(this.pointwiseConstraint);
+    return config;
+  }
+};
+SeparableConv.className = "SeparableConv";
+var SeparableConv2D = class extends SeparableConv {
+  constructor(args) {
+    super(2, args);
+  }
+};
+SeparableConv2D.className = "SeparableConv2D";
+serialization_exports.registerClass(SeparableConv2D);
+var Conv1D = class extends Conv {
+  constructor(args) {
+    super(1, args);
+    Conv1D.verifyArgs(args);
+    this.inputSpec = [{ ndim: 3 }];
+  }
+  getConfig() {
+    const config = super.getConfig();
+    delete config["rank"];
+    delete config["dataFormat"];
+    return config;
+  }
+  static verifyArgs(args) {
+    if (typeof args.kernelSize !== "number" && !checkArrayTypeAndLength(args.kernelSize, "number", 1, 1)) {
+      throw new ValueError(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(args.kernelSize)}.`);
+    }
+  }
+};
+Conv1D.className = "Conv1D";
+serialization_exports.registerClass(Conv1D);
+var Cropping2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    if (typeof args.cropping === "number") {
+      this.cropping = [[args.cropping, args.cropping], [args.cropping, args.cropping]];
+    } else if (typeof args.cropping[0] === "number") {
+      this.cropping = [
+        [args.cropping[0], args.cropping[0]],
+        [args.cropping[1], args.cropping[1]]
+      ];
+    } else {
+      this.cropping = args.cropping;
+    }
+    this.dataFormat = args.dataFormat === void 0 ? "channelsLast" : args.dataFormat;
+    this.inputSpec = [{ ndim: 4 }];
+  }
+  computeOutputShape(inputShape) {
+    if (this.dataFormat === "channelsFirst") {
+      return [
+        inputShape[0],
+        inputShape[1],
+        inputShape[2] - this.cropping[0][0] - this.cropping[0][1],
+        inputShape[3] - this.cropping[1][0] - this.cropping[1][1]
+      ];
+    } else {
+      return [
+        inputShape[0],
+        inputShape[1] - this.cropping[0][0] - this.cropping[0][1],
+        inputShape[2] - this.cropping[1][0] - this.cropping[1][1],
+        inputShape[3]
+      ];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[1] - this.cropping[0][0] - this.cropping[0][1], 2);
+        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[2] - this.cropping[1][1] - this.cropping[1][0], 3);
+      } else {
+        const hSliced = sliceAlongAxis(inputs, this.cropping[0][0], inputs.shape[2] - this.cropping[0][0] - this.cropping[0][1], 3);
+        return sliceAlongAxis(hSliced, this.cropping[1][0], inputs.shape[3] - this.cropping[1][1] - this.cropping[1][0], 4);
+      }
+    });
+  }
+  getConfig() {
+    const config = { cropping: this.cropping, dataFormat: this.dataFormat };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Cropping2D.className = "Cropping2D";
+serialization_exports.registerClass(Cropping2D);
+var UpSampling2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_SIZE = [2, 2];
+    this.inputSpec = [{ ndim: 4 }];
+    this.size = args.size == null ? this.DEFAULT_SIZE : args.size;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.interpolation = args.interpolation == null ? "nearest" : args.interpolation;
+    checkInterpolationFormat(this.interpolation);
+  }
+  computeOutputShape(inputShape) {
+    if (this.dataFormat === "channelsFirst") {
+      const height = inputShape[2] == null ? null : this.size[0] * inputShape[2];
+      const width = inputShape[3] == null ? null : this.size[1] * inputShape[3];
+      return [inputShape[0], inputShape[1], height, width];
+    } else {
+      const height = inputShape[1] == null ? null : this.size[0] * inputShape[1];
+      const width = inputShape[2] == null ? null : this.size[1] * inputShape[2];
+      return [inputShape[0], height, width, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      let input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      if (this.dataFormat === "channelsFirst") {
+        input2 = transpose(input2, [0, 2, 3, 1]);
+        const height = this.size[0] * inputShape[2];
+        const width = this.size[1] * inputShape[3];
+        const resized = this.interpolation === "nearest" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);
+        return transpose(resized, [0, 3, 1, 2]);
+      } else {
+        const height = this.size[0] * inputShape[1];
+        const width = this.size[1] * inputShape[2];
+        return this.interpolation === "nearest" ? image.resizeNearestNeighbor(input2, [height, width]) : image.resizeBilinear(input2, [height, width]);
+      }
+    });
+  }
+  getConfig() {
+    const config = {
+      size: this.size,
+      dataFormat: this.dataFormat,
+      interpolation: this.interpolation
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+UpSampling2D.className = "UpSampling2D";
+serialization_exports.registerClass(UpSampling2D);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional_depthwise.js
+function depthwiseConv2d3(x, depthwiseKernel, strides = [1, 1], padding = "valid", dataFormat, dilationRate) {
+  return tidy(() => {
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    checkDataFormat(dataFormat);
+    let y = preprocessConv2DInput(x, dataFormat);
+    if (x.rank !== 4) {
+      throw new ValueError(`Input for depthwiseConv2d is required to be 4-D, but is instead ${x.rank}-D`);
+    }
+    if (depthwiseKernel.rank !== 4) {
+      throw new ValueError(`depthwiseKernel is required to be 4-D, but is instead ${depthwiseKernel.rank}-D`);
+    }
+    y = depthwiseConv2d(y, depthwiseKernel, strides, padding === "same" ? "same" : "valid", "NHWC", dilationRate);
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+var DepthwiseConv2D = class extends BaseConv {
+  constructor(args) {
+    super(2, args);
+    this.depthwiseKernel = null;
+    this.depthMultiplier = args.depthMultiplier == null ? 1 : args.depthMultiplier;
+    this.depthwiseInitializer = getInitializer(args.depthwiseInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.depthwiseConstraint = getConstraint(args.depthwiseConstraint);
+    this.depthwiseRegularizer = getRegularizer(args.depthwiseRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < 4) {
+      throw new ValueError(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(inputShape)}.`);
+    }
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : 3;
+    if (inputShape[channelAxis] == null || inputShape[channelAxis] < 0) {
+      throw new ValueError(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${inputShape[channelAxis]}).`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const depthwiseKernelShape = [
+      this.kernelSize[0],
+      this.kernelSize[1],
+      inputDim,
+      this.depthMultiplier
+    ];
+    this.depthwiseKernel = this.addWeight("depthwise_kernel", depthwiseKernelShape, null, this.depthwiseInitializer, this.depthwiseRegularizer, true, this.depthwiseConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [inputDim * this.depthMultiplier], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      let outputs = depthwiseConv2d3(inputs, this.depthwiseKernel.read(), this.strides, this.padding, this.dataFormat, null);
+      if (this.useBias) {
+        outputs = biasAdd(outputs, this.bias.read(), this.dataFormat);
+      }
+      if (this.activation != null) {
+        outputs = this.activation.apply(outputs);
+      }
+      return outputs;
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const rows = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    const cols = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    const outFilters = this.dataFormat === "channelsFirst" ? inputShape[1] * this.depthMultiplier : inputShape[3] * this.depthMultiplier;
+    const outRows = convOutputLength(rows, this.kernelSize[0], this.padding, this.strides[0]);
+    const outCols = convOutputLength(cols, this.kernelSize[1], this.padding, this.strides[1]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], outFilters, outRows, outCols];
+    } else {
+      return [inputShape[0], outRows, outCols, outFilters];
+    }
+  }
+  getConfig() {
+    const config = super.getConfig();
+    config["depthMultiplier"] = this.depthMultiplier;
+    config["depthwiseInitializer"] = serializeInitializer(this.depthwiseInitializer);
+    config["depthwiseRegularizer"] = serializeRegularizer(this.depthwiseRegularizer);
+    config["depthwiseConstraint"] = serializeConstraint(this.depthwiseRegularizer);
+    return config;
+  }
+};
+DepthwiseConv2D.className = "DepthwiseConv2D";
+serialization_exports.registerClass(DepthwiseConv2D);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/recurrent.js
+function standardizeArgs(inputs, initialState, constants, numConstants) {
+  if (Array.isArray(inputs)) {
+    if (initialState != null || constants != null) {
+      throw new ValueError("When inputs is an array, neither initialState or constants should be provided");
+    }
+    if (numConstants != null) {
+      constants = inputs.slice(inputs.length - numConstants, inputs.length);
+      inputs = inputs.slice(0, inputs.length - numConstants);
+    }
+    if (inputs.length > 1) {
+      initialState = inputs.slice(1, inputs.length);
+    }
+    inputs = inputs[0];
+  }
+  function toListOrNull(x) {
+    if (x == null || Array.isArray(x)) {
+      return x;
+    } else {
+      return [x];
+    }
+  }
+  initialState = toListOrNull(initialState);
+  constants = toListOrNull(constants);
+  return { inputs, initialState, constants };
+}
+function rnn(stepFunction, inputs, initialStates, goBackwards = false, mask, constants, unroll = false, needPerStepOutputs = false) {
+  return tidy(() => {
+    const ndim = inputs.shape.length;
+    if (ndim < 3) {
+      throw new ValueError(`Input should be at least 3D, but is ${ndim}D.`);
+    }
+    const axes = [1, 0].concat(range2(2, ndim));
+    inputs = transpose(inputs, axes);
+    if (constants != null) {
+      throw new NotImplementedError("The rnn() functoin of the deeplearn.js backend does not support constants yet.");
+    }
+    if (unroll) {
+      console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend.");
+    }
+    if (mask != null) {
+      mask = cast(cast(mask, "bool"), "float32");
+      if (mask.rank === ndim - 1) {
+        mask = expandDims(mask, -1);
+      }
+      mask = transpose(mask, axes);
+    }
+    if (goBackwards) {
+      inputs = reverse(inputs, 0);
+      if (mask != null) {
+        mask = reverse(mask, 0);
+      }
+    }
+    const perStepOutputs = [];
+    let lastOutput;
+    let states = initialStates;
+    const timeSteps = inputs.shape[0];
+    const perStepInputs = unstack(inputs);
+    let perStepMasks;
+    if (mask != null) {
+      perStepMasks = unstack(mask);
+    }
+    for (let t = 0; t < timeSteps; ++t) {
+      const currentInput = perStepInputs[t];
+      const stepOutputs = tidy(() => stepFunction(currentInput, states));
+      if (mask == null) {
+        lastOutput = stepOutputs[0];
+        states = stepOutputs[1];
+      } else {
+        const maskedOutputs = tidy(() => {
+          const stepMask = perStepMasks[t];
+          const negStepMask = sub(onesLike(stepMask), stepMask);
+          const output = add2(mul(stepOutputs[0], stepMask), mul(states[0], negStepMask));
+          const newStates = states.map((state, i) => {
+            return add2(mul(stepOutputs[1][i], stepMask), mul(state, negStepMask));
+          });
+          return { output, newStates };
+        });
+        lastOutput = maskedOutputs.output;
+        states = maskedOutputs.newStates;
+      }
+      if (needPerStepOutputs) {
+        perStepOutputs.push(lastOutput);
+      }
+    }
+    let outputs;
+    if (needPerStepOutputs) {
+      const axis = 1;
+      outputs = stack(perStepOutputs, axis);
+    }
+    return [lastOutput, outputs, states];
+  });
+}
+var RNN = class extends Layer {
+  constructor(args) {
+    super(args);
+    let cell;
+    if (args.cell == null) {
+      throw new ValueError("cell property is missing for the constructor of RNN.");
+    } else if (Array.isArray(args.cell)) {
+      cell = new StackedRNNCells({ cells: args.cell });
+    } else {
+      cell = args.cell;
+    }
+    if (cell.stateSize == null) {
+      throw new ValueError("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");
+    }
+    this.cell = cell;
+    this.returnSequences = args.returnSequences == null ? false : args.returnSequences;
+    this.returnState = args.returnState == null ? false : args.returnState;
+    this.goBackwards = args.goBackwards == null ? false : args.goBackwards;
+    this._stateful = args.stateful == null ? false : args.stateful;
+    this.unroll = args.unroll == null ? false : args.unroll;
+    this.supportsMasking = true;
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+    this.stateSpec = null;
+    this.states_ = null;
+    this.numConstants = null;
+    this.keptStates = [];
+  }
+  getStates() {
+    if (this.states_ == null) {
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      return range2(0, numStates).map((x) => null);
+    } else {
+      return this.states_;
+    }
+  }
+  setStates(states) {
+    this.states_ = states;
+  }
+  computeOutputShape(inputShape) {
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    let stateSize = this.cell.stateSize;
+    if (!Array.isArray(stateSize)) {
+      stateSize = [stateSize];
+    }
+    const outputDim = stateSize[0];
+    let outputShape;
+    if (this.returnSequences) {
+      outputShape = [inputShape[0], inputShape[1], outputDim];
+    } else {
+      outputShape = [inputShape[0], outputDim];
+    }
+    if (this.returnState) {
+      const stateShape = [];
+      for (const dim of stateSize) {
+        stateShape.push([inputShape[0], dim]);
+      }
+      return [outputShape].concat(stateShape);
+    } else {
+      return outputShape;
+    }
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (Array.isArray(mask)) {
+        mask = mask[0];
+      }
+      const outputMask = this.returnSequences ? mask : null;
+      if (this.returnState) {
+        const stateMask = this.states.map((s) => null);
+        return [outputMask].concat(stateMask);
+      } else {
+        return outputMask;
+      }
+    });
+  }
+  get states() {
+    if (this.states_ == null) {
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      const output = [];
+      for (let i = 0; i < numStates; ++i) {
+        output.push(null);
+      }
+      return output;
+    } else {
+      return this.states_;
+    }
+  }
+  set states(s) {
+    this.states_ = s;
+  }
+  build(inputShape) {
+    const constantShape = null;
+    if (this.numConstants != null) {
+      throw new NotImplementedError("Constants support is not implemented in RNN yet.");
+    }
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    const batchSize = this.stateful ? inputShape[0] : null;
+    const inputDim = inputShape.slice(2);
+    this.inputSpec[0] = new InputSpec({ shape: [batchSize, null, ...inputDim] });
+    const stepInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    if (constantShape != null) {
+      throw new NotImplementedError("Constants support is not implemented in RNN yet.");
+    } else {
+      this.cell.build(stepInputShape);
+    }
+    let stateSize;
+    if (Array.isArray(this.cell.stateSize)) {
+      stateSize = this.cell.stateSize;
+    } else {
+      stateSize = [this.cell.stateSize];
+    }
+    if (this.stateSpec != null) {
+      if (!util_exports.arraysEqual(this.stateSpec.map((spec) => spec.shape[spec.shape.length - 1]), stateSize)) {
+        throw new ValueError(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`);
+      }
+    } else {
+      this.stateSpec = stateSize.map((dim) => new InputSpec({ shape: [null, dim] }));
+    }
+    if (this.stateful) {
+      this.resetStates();
+    }
+  }
+  resetStates(states, training = false) {
+    tidy(() => {
+      if (!this.stateful) {
+        throw new AttributeError("Cannot call resetStates() on an RNN Layer that is not stateful.");
+      }
+      const batchSize = this.inputSpec[0].shape[0];
+      if (batchSize == null) {
+        throw new ValueError("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+      }
+      if (this.states_ == null) {
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));
+        } else {
+          this.states_ = [zeros([batchSize, this.cell.stateSize])];
+        }
+      } else if (states == null) {
+        dispose(this.states_);
+        if (this.keptStates != null) {
+          dispose(this.keptStates);
+          this.keptStates = [];
+        }
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map((dim) => zeros([batchSize, dim]));
+        } else {
+          this.states_[0] = zeros([batchSize, this.cell.stateSize]);
+        }
+      } else {
+        if (!Array.isArray(states)) {
+          states = [states];
+        }
+        if (states.length !== this.states_.length) {
+          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);
+        }
+        if (training === true) {
+          this.keptStates.push(this.states_.slice());
+        } else {
+          dispose(this.states_);
+        }
+        for (let index = 0; index < this.states_.length; ++index) {
+          const value = states[index];
+          const dim = Array.isArray(this.cell.stateSize) ? this.cell.stateSize[index] : this.cell.stateSize;
+          const expectedShape = [batchSize, dim];
+          if (!util_exports.arraysEqual(value.shape, expectedShape)) {
+            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);
+          }
+          this.states_[index] = value;
+        }
+      }
+      this.states_ = this.states_.map((state) => keep(state.clone()));
+    });
+  }
+  apply(inputs, kwargs) {
+    let initialState = kwargs == null ? null : kwargs["initialState"];
+    let constants = kwargs == null ? null : kwargs["constants"];
+    if (kwargs == null) {
+      kwargs = {};
+    }
+    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);
+    inputs = standardized.inputs;
+    initialState = standardized.initialState;
+    constants = standardized.constants;
+    let additionalInputs = [];
+    let additionalSpecs = [];
+    if (initialState != null) {
+      kwargs["initialState"] = initialState;
+      additionalInputs = additionalInputs.concat(initialState);
+      this.stateSpec = [];
+      for (const state of initialState) {
+        this.stateSpec.push(new InputSpec({ shape: state.shape }));
+      }
+      additionalSpecs = additionalSpecs.concat(this.stateSpec);
+    }
+    if (constants != null) {
+      kwargs["constants"] = constants;
+      additionalInputs = additionalInputs.concat(constants);
+      this.numConstants = constants.length;
+    }
+    const isTensor = additionalInputs[0] instanceof SymbolicTensor;
+    if (isTensor) {
+      const fullInput = [inputs].concat(additionalInputs);
+      const fullInputSpec = this.inputSpec.concat(additionalSpecs);
+      const originalInputSpec = this.inputSpec;
+      this.inputSpec = fullInputSpec;
+      const output = super.apply(fullInput, kwargs);
+      this.inputSpec = originalInputSpec;
+      return output;
+    } else {
+      return super.apply(inputs, kwargs);
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      let initialState = kwargs == null ? null : kwargs["initialState"];
+      inputs = getExactlyOneTensor(inputs);
+      if (initialState == null) {
+        if (this.stateful) {
+          initialState = this.states_;
+        } else {
+          initialState = this.getInitialState(inputs);
+        }
+      }
+      const numStates = Array.isArray(this.cell.stateSize) ? this.cell.stateSize.length : 1;
+      if (initialState.length !== numStates) {
+        throw new ValueError(`RNN Layer has ${numStates} state(s) but was passed ${initialState.length} initial state(s).`);
+      }
+      if (this.unroll) {
+        console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");
+      }
+      const cellCallKwargs = { training };
+      const step5 = (inputs2, states2) => {
+        const outputs2 = this.cell.call([inputs2].concat(states2), cellCallKwargs);
+        return [outputs2[0], outputs2.slice(1)];
+      };
+      const rnnOutputs = rnn(step5, inputs, initialState, this.goBackwards, mask, null, this.unroll, this.returnSequences);
+      const lastOutput = rnnOutputs[0];
+      const outputs = rnnOutputs[1];
+      const states = rnnOutputs[2];
+      if (this.stateful) {
+        this.resetStates(states, training);
+      }
+      const output = this.returnSequences ? outputs : lastOutput;
+      if (this.returnState) {
+        return [output].concat(states);
+      } else {
+        return output;
+      }
+    });
+  }
+  getInitialState(inputs) {
+    return tidy(() => {
+      let initialState = zeros(inputs.shape);
+      initialState = sum2(initialState, [1, 2]);
+      initialState = expandDims2(initialState);
+      if (Array.isArray(this.cell.stateSize)) {
+        return this.cell.stateSize.map((dim) => dim > 1 ? tile2(initialState, [1, dim]) : initialState);
+      } else {
+        return this.cell.stateSize > 1 ? [tile2(initialState, [1, this.cell.stateSize])] : [initialState];
+      }
+    });
+  }
+  get trainableWeights() {
+    if (!this.trainable) {
+      return [];
+    }
+    return this.cell.trainableWeights;
+  }
+  get nonTrainableWeights() {
+    if (!this.trainable) {
+      return this.cell.weights;
+    }
+    return this.cell.nonTrainableWeights;
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.cell != null) {
+      this.cell.setFastWeightInitDuringBuild(value);
+    }
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      returnSequences: this.returnSequences,
+      returnState: this.returnState,
+      goBackwards: this.goBackwards,
+      stateful: this.stateful,
+      unroll: this.unroll
+    };
+    if (this.numConstants != null) {
+      config["numConstants"] = this.numConstants;
+    }
+    const cellConfig = this.cell.getConfig();
+    if (this.getClassName() === RNN.className) {
+      config["cell"] = {
+        "className": this.cell.getClassName(),
+        "config": cellConfig
+      };
+    }
+    return Object.assign(Object.assign(Object.assign({}, cellConfig), baseConfig), config);
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const cellConfig = config["cell"];
+    const cell = deserialize(cellConfig, customObjects);
+    return new cls(Object.assign(config, { cell }));
+  }
+};
+RNN.className = "RNN";
+serialization_exports.registerClass(RNN);
+var RNNCell = class extends Layer {
+};
+var SimpleRNNCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    this.units = args.units;
+    assertPositiveInteger(this.units, `units`);
+    this.activation = getActivation(args.activation == null ? this.DEFAULT_ACTIVATION : args.activation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.stateSize = this.units;
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    this.kernel = this.addWeight("kernel", [inputShape[inputShape.length - 1], this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (inputs.length !== 2) {
+        throw new ValueError(`SimpleRNNCell expects 2 input Tensors, got ${inputs.length}.`);
+      }
+      let prevOutput = inputs[1];
+      inputs = inputs[0];
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(prevOutput),
+          rate: this.recurrentDropout,
+          training,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      let h;
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      if (dpMask != null) {
+        h = dot2(mul(inputs, dpMask), this.kernel.read());
+      } else {
+        h = dot2(inputs, this.kernel.read());
+      }
+      if (this.bias != null) {
+        h = biasAdd(h, this.bias.read());
+      }
+      if (recDpMask != null) {
+        prevOutput = mul(prevOutput, recDpMask);
+      }
+      let output = add2(h, dot2(prevOutput, this.recurrentKernel.read()));
+      if (this.activation != null) {
+        output = this.activation.apply(output);
+      }
+      return [output, output];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+SimpleRNNCell.className = "SimpleRNNCell";
+serialization_exports.registerClass(SimpleRNNCell);
+var SimpleRNN = class extends RNN {
+  constructor(args) {
+    args.cell = new SimpleRNNCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+SimpleRNN.className = "SimpleRNN";
+serialization_exports.registerClass(SimpleRNN);
+var GRUCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    if (args.resetAfter) {
+      throw new ValueError(`GRUCell does not support reset_after parameter set to true.`);
+    }
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);
+    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.implementation = args.implementation;
+    this.stateSize = this.units;
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const inputDim = inputShape[inputShape.length - 1];
+    this.kernel = this.addWeight("kernel", [inputDim, this.units * 3], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units * 3], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      this.bias = this.addWeight("bias", [this.units * 3], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (inputs.length !== 2) {
+        throw new ValueError(`GRUCell expects 2 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      let hTMinus1 = inputs[1];
+      inputs = inputs[0];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          count: 3,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: 3,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      let z;
+      let r;
+      let hh;
+      if (0 < this.dropout && this.dropout < 1) {
+        inputs = mul(inputs, dpMask[0]);
+      }
+      let matrixX = dot2(inputs, this.kernel.read());
+      if (this.useBias) {
+        matrixX = biasAdd(matrixX, this.bias.read());
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {
+        hTMinus1 = mul(hTMinus1, recDpMask[0]);
+      }
+      const recurrentKernelValue = this.recurrentKernel.read();
+      const [rk1, rk2] = split(recurrentKernelValue, [2 * this.units, this.units], recurrentKernelValue.rank - 1);
+      const matrixInner = dot2(hTMinus1, rk1);
+      const [xZ, xR, xH] = split(matrixX, 3, matrixX.rank - 1);
+      const [recurrentZ, recurrentR] = split(matrixInner, 2, matrixInner.rank - 1);
+      z = this.recurrentActivation.apply(add2(xZ, recurrentZ));
+      r = this.recurrentActivation.apply(add2(xR, recurrentR));
+      const recurrentH = dot2(mul(r, hTMinus1), rk2);
+      hh = this.activation.apply(add2(xH, recurrentH));
+      const h = add2(mul(z, hTMinus1), mul(add2(1, neg(z)), hh));
+      return [h, h];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      recurrentActivation: serializeActivation(this.recurrentActivation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout,
+      implementation: this.implementation,
+      resetAfter: false
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+GRUCell.className = "GRUCell";
+serialization_exports.registerClass(GRUCell);
+var GRU = class extends RNN {
+  constructor(args) {
+    if (args.implementation === 0) {
+      console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.");
+    }
+    args.cell = new GRUCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    if (config["implmentation"] === 0) {
+      config["implementation"] = 1;
+    }
+    return new cls(config);
+  }
+};
+GRU.className = "GRU";
+serialization_exports.registerClass(GRU);
+var LSTMCell = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_ACTIVATION = "tanh";
+    this.DEFAULT_RECURRENT_ACTIVATION = "hardSigmoid";
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_RECURRENT_INITIALIZER = "orthogonal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation === void 0 ? this.DEFAULT_ACTIVATION : args.activation);
+    this.recurrentActivation = getActivation(args.recurrentActivation === void 0 ? this.DEFAULT_RECURRENT_ACTIVATION : args.recurrentActivation);
+    this.useBias = args.useBias == null ? true : args.useBias;
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.recurrentInitializer = getInitializer(args.recurrentInitializer || this.DEFAULT_RECURRENT_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.unitForgetBias = args.unitForgetBias;
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.recurrentRegularizer = getRegularizer(args.recurrentRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.recurrentConstraint = getConstraint(args.recurrentConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.dropout = min2([1, max2([0, args.dropout == null ? 0 : args.dropout])]);
+    this.recurrentDropout = min2([
+      1,
+      max2([0, args.recurrentDropout == null ? 0 : args.recurrentDropout])
+    ]);
+    this.dropoutFunc = args.dropoutFunc;
+    this.implementation = args.implementation;
+    this.stateSize = [this.units, this.units];
+    this.dropoutMask = null;
+    this.recurrentDropoutMask = null;
+  }
+  build(inputShape) {
+    var _a;
+    inputShape = getExactlyOneShape(inputShape);
+    const inputDim = inputShape[inputShape.length - 1];
+    this.kernel = this.addWeight("kernel", [inputDim, this.units * 4], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", [this.units, this.units * 4], null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    let biasInitializer;
+    if (this.useBias) {
+      if (this.unitForgetBias) {
+        const capturedBiasInit = this.biasInitializer;
+        const capturedUnits = this.units;
+        biasInitializer = new (_a = class CustomInit extends Initializer {
+          apply(shape, dtype) {
+            const bI = capturedBiasInit.apply([capturedUnits]);
+            const bF = new Ones().apply([capturedUnits]);
+            const bCAndH = capturedBiasInit.apply([capturedUnits * 2]);
+            return concatAlongFirstAxis(concatAlongFirstAxis(bI, bF), bCAndH);
+          }
+        }, _a.className = "CustomInit", _a)();
+      } else {
+        biasInitializer = this.biasInitializer;
+      }
+      this.bias = this.addWeight("bias", [this.units * 4], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    } else {
+      this.bias = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      inputs = inputs;
+      if (inputs.length !== 3) {
+        throw new ValueError(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      let hTMinus1 = inputs[1];
+      const cTMinus1 = inputs[2];
+      inputs = inputs[0];
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(inputs),
+          rate: this.dropout,
+          training,
+          count: 4,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: 4,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dpMask = this.dropoutMask;
+      const recDpMask = this.recurrentDropoutMask;
+      let i;
+      let f;
+      let c;
+      let o;
+      if (0 < this.dropout && this.dropout < 1) {
+        inputs = mul(inputs, dpMask[0]);
+      }
+      let z = dot2(inputs, this.kernel.read());
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1) {
+        hTMinus1 = mul(hTMinus1, recDpMask[0]);
+      }
+      z = add2(z, dot2(hTMinus1, this.recurrentKernel.read()));
+      if (this.useBias) {
+        z = biasAdd(z, this.bias.read());
+      }
+      const [z0, z1, z2, z3] = split(z, 4, z.rank - 1);
+      i = this.recurrentActivation.apply(z0);
+      f = this.recurrentActivation.apply(z1);
+      c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(z2)));
+      o = this.recurrentActivation.apply(z3);
+      const h = mul(o, this.activation.apply(c));
+      return [h, h, c];
+    });
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      recurrentActivation: serializeActivation(this.recurrentActivation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      recurrentInitializer: serializeInitializer(this.recurrentInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      unitForgetBias: this.unitForgetBias,
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      recurrentRegularizer: serializeRegularizer(this.recurrentRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      recurrentConstraint: serializeConstraint(this.recurrentConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint),
+      dropout: this.dropout,
+      recurrentDropout: this.recurrentDropout,
+      implementation: this.implementation
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+};
+LSTMCell.className = "LSTMCell";
+serialization_exports.registerClass(LSTMCell);
+var LSTM = class extends RNN {
+  constructor(args) {
+    if (args.implementation === 0) {
+      console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call.");
+    }
+    args.cell = new LSTMCell(args);
+    super(args);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  static fromConfig(cls, config) {
+    if (config["implmentation"] === 0) {
+      config["implementation"] = 1;
+    }
+    return new cls(config);
+  }
+};
+LSTM.className = "LSTM";
+serialization_exports.registerClass(LSTM);
+var StackedRNNCells = class extends RNNCell {
+  constructor(args) {
+    super(args);
+    this.cells = args.cells;
+  }
+  get stateSize() {
+    const stateSize = [];
+    for (const cell of this.cells.slice().reverse()) {
+      if (Array.isArray(cell.stateSize)) {
+        stateSize.push(...cell.stateSize);
+      } else {
+        stateSize.push(cell.stateSize);
+      }
+    }
+    return stateSize;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      let states = inputs.slice(1);
+      const nestedStates = [];
+      for (const cell of this.cells.slice().reverse()) {
+        if (Array.isArray(cell.stateSize)) {
+          nestedStates.push(states.splice(0, cell.stateSize.length));
+        } else {
+          nestedStates.push(states.splice(0, 1));
+        }
+      }
+      nestedStates.reverse();
+      const newNestedStates = [];
+      let callInputs;
+      for (let i = 0; i < this.cells.length; ++i) {
+        const cell = this.cells[i];
+        states = nestedStates[i];
+        if (i === 0) {
+          callInputs = [inputs[0]].concat(states);
+        } else {
+          callInputs = [callInputs[0]].concat(states);
+        }
+        callInputs = cell.call(callInputs, kwargs);
+        newNestedStates.push(callInputs.slice(1));
+      }
+      states = [];
+      for (const cellStates of newNestedStates.slice().reverse()) {
+        states.push(...cellStates);
+      }
+      return [callInputs[0]].concat(states);
+    });
+  }
+  build(inputShape) {
+    if (isArrayOfShapes(inputShape)) {
+      inputShape = inputShape[0];
+    }
+    inputShape = inputShape;
+    let outputDim;
+    this.cells.forEach((cell, i) => {
+      nameScope(`RNNCell_${i}`, () => {
+        cell.build(inputShape);
+        if (Array.isArray(cell.stateSize)) {
+          outputDim = cell.stateSize[0];
+        } else {
+          outputDim = cell.stateSize;
+        }
+        inputShape = [inputShape[0], outputDim];
+      });
+    });
+    this.built = true;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const getCellConfig = (cell) => {
+      return {
+        "className": cell.getClassName(),
+        "config": cell.getConfig()
+      };
+    };
+    const cellConfigs = this.cells.map(getCellConfig);
+    const config = { "cells": cellConfigs };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const cells = [];
+    for (const cellConfig of config["cells"]) {
+      cells.push(deserialize(cellConfig, customObjects));
+    }
+    return new cls({ cells });
+  }
+  get trainableWeights() {
+    if (!this.trainable) {
+      return [];
+    }
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.trainableWeights);
+    }
+    return weights;
+  }
+  get nonTrainableWeights() {
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.nonTrainableWeights);
+    }
+    if (!this.trainable) {
+      const trainableWeights = [];
+      for (const cell of this.cells) {
+        trainableWeights.push(...cell.trainableWeights);
+      }
+      return trainableWeights.concat(weights);
+    }
+    return weights;
+  }
+  getWeights() {
+    const weights = [];
+    for (const cell of this.cells) {
+      weights.push(...cell.weights);
+    }
+    return batchGetValue(weights);
+  }
+  setWeights(weights) {
+    const tuples = [];
+    for (const cell of this.cells) {
+      const numParams = cell.weights.length;
+      const inputWeights = weights.splice(numParams);
+      for (let i = 0; i < cell.weights.length; ++i) {
+        tuples.push([cell.weights[i], inputWeights[i]]);
+      }
+    }
+    batchSetValue(tuples);
+  }
+};
+StackedRNNCells.className = "StackedRNNCells";
+serialization_exports.registerClass(StackedRNNCells);
+function generateDropoutMask(args) {
+  const { ones: ones4, rate, training = false, count: count2 = 1, dropoutFunc } = args;
+  const droppedInputs = () => dropoutFunc != null ? dropoutFunc(ones4(), rate) : dropout2(ones4(), rate);
+  const createMask = () => inTrainPhase(droppedInputs, ones4, training);
+  if (!count2 || count2 <= 1) {
+    return keep(createMask().clone());
+  }
+  const masks = Array(count2).fill(void 0).map(createMask);
+  return masks.map((m) => keep(m.clone()));
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/convolutional_recurrent.js
+var __rest = function(s, e) {
+  var t = {};
+  for (var p2 in s)
+    if (Object.prototype.hasOwnProperty.call(s, p2) && e.indexOf(p2) < 0)
+      t[p2] = s[p2];
+  if (s != null && typeof Object.getOwnPropertySymbols === "function")
+    for (var i = 0, p2 = Object.getOwnPropertySymbols(s); i < p2.length; i++) {
+      if (e.indexOf(p2[i]) < 0 && Object.prototype.propertyIsEnumerable.call(s, p2[i]))
+        t[p2[i]] = s[p2[i]];
+    }
+  return t;
+};
+var ConvRNN2D = class extends RNN {
+  constructor(args) {
+    if (args.unroll) {
+      throw new NotImplementedError("Unrolling is not possible with convolutional RNNs.");
+    }
+    if (Array.isArray(args.cell)) {
+      throw new NotImplementedError("It is not possible at the moment to stack convolutional cells.");
+    }
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.cell.dropoutMask != null) {
+        dispose(this.cell.dropoutMask);
+        this.cell.dropoutMask = null;
+      }
+      if (this.cell.recurrentDropoutMask != null) {
+        dispose(this.cell.recurrentDropoutMask);
+        this.cell.recurrentDropoutMask = null;
+      }
+      if (kwargs && kwargs["constants"]) {
+        throw new ValueError("ConvRNN2D cell does not support constants");
+      }
+      const mask = kwargs == null ? null : kwargs["mask"];
+      const training = kwargs == null ? null : kwargs["training"];
+      const initialState = kwargs == null ? null : kwargs["initialState"];
+      return super.call(inputs, { mask, training, initialState });
+    });
+  }
+  computeOutputShape(inputShape) {
+    let outShape = this.computeSingleOutputShape(inputShape);
+    if (!this.returnSequences) {
+      outShape = [outShape[0], ...outShape.slice(2)];
+    }
+    if (this.returnState) {
+      outShape = [outShape, ...Array(2).fill([inputShape[0], ...outShape.slice(-3)])];
+    }
+    return outShape;
+  }
+  getInitialState(inputs) {
+    return tidy(() => {
+      const { stateSize } = this.cell;
+      const inputShape = inputs.shape;
+      const outputShape = this.computeSingleOutputShape(inputShape);
+      const stateShape = [outputShape[0], ...outputShape.slice(2)];
+      const initialState = zeros(stateShape);
+      if (Array.isArray(stateSize)) {
+        return Array(stateSize.length).fill(initialState);
+      }
+      return [initialState];
+    });
+  }
+  resetStates(states, training = false) {
+    tidy(() => {
+      if (!this.stateful) {
+        throw new AttributeError("Cannot call resetStates() on an RNN Layer that is not stateful.");
+      }
+      const inputShape = this.inputSpec[0].shape;
+      const outputShape = this.computeSingleOutputShape(inputShape);
+      const stateShape = [outputShape[0], ...outputShape.slice(2)];
+      const batchSize = inputShape[0];
+      if (batchSize == null) {
+        throw new ValueError("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");
+      }
+      if (this.getStates() == null) {
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));
+        } else {
+          this.states_ = [zeros(stateShape)];
+        }
+      } else if (states == null) {
+        dispose(this.states_);
+        if (this.keptStates != null) {
+          dispose(this.keptStates);
+          this.keptStates = [];
+        }
+        if (Array.isArray(this.cell.stateSize)) {
+          this.states_ = this.cell.stateSize.map(() => zeros(stateShape));
+        } else {
+          this.states_[0] = zeros(stateShape);
+        }
+      } else {
+        if (!Array.isArray(states)) {
+          states = [states];
+        }
+        if (states.length !== this.states_.length) {
+          throw new ValueError(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${states.length} state value(s). Input received: ${states}`);
+        }
+        if (training) {
+          this.keptStates.push(this.states_.slice());
+        } else {
+          dispose(this.states_);
+        }
+        for (let index = 0; index < this.states_.length; ++index) {
+          const value = states[index];
+          const expectedShape = stateShape;
+          if (!util_exports.arraysEqual(value.shape, expectedShape)) {
+            throw new ValueError(`State ${index} is incompatible with layer ${this.name}: expected shape=${expectedShape}, received shape=${value.shape}`);
+          }
+          this.states_[index] = value;
+        }
+      }
+      this.states_ = this.states_.map((state) => keep(state.clone()));
+    });
+  }
+  computeSingleOutputShape(inputShape) {
+    const { dataFormat, filters, kernelSize, padding, strides, dilationRate } = this.cell;
+    const isChannelsFirst = dataFormat === "channelsFirst";
+    const h = inputShape[isChannelsFirst ? 3 : 2];
+    const w = inputShape[isChannelsFirst ? 4 : 3];
+    const hOut = convOutputLength(h, kernelSize[0], padding, strides[0], dilationRate[0]);
+    const wOut = convOutputLength(w, kernelSize[1], padding, strides[1], dilationRate[1]);
+    const outShape = [
+      ...inputShape.slice(0, 2),
+      ...isChannelsFirst ? [filters, hOut, wOut] : [hOut, wOut, filters]
+    ];
+    return outShape;
+  }
+};
+ConvRNN2D.className = "ConvRNN2D";
+var ConvLSTM2DCell = class extends LSTMCell {
+  constructor(args) {
+    const { filters, kernelSize, strides, padding, dataFormat, dilationRate } = args;
+    super(Object.assign(Object.assign({}, args), { units: filters }));
+    this.filters = filters;
+    assertPositiveInteger(this.filters, "filters");
+    this.kernelSize = normalizeArray(kernelSize, 2, "kernelSize");
+    this.kernelSize.forEach((size) => assertPositiveInteger(size, "kernelSize"));
+    this.strides = normalizeArray(strides || 1, 2, "strides");
+    this.strides.forEach((stride) => assertPositiveInteger(stride, "strides"));
+    this.padding = padding || "valid";
+    checkPaddingMode(this.padding);
+    this.dataFormat = dataFormat || "channelsLast";
+    checkDataFormat(this.dataFormat);
+    this.dilationRate = normalizeArray(dilationRate || 1, 2, "dilationRate");
+    this.dilationRate.forEach((rate) => assertPositiveInteger(rate, "dilationRate"));
+  }
+  build(inputShape) {
+    var _a;
+    inputShape = getExactlyOneShape(inputShape);
+    const channelAxis = this.dataFormat === "channelsFirst" ? 1 : inputShape.length - 1;
+    if (inputShape[channelAxis] == null) {
+      throw new ValueError(`The channel dimension of the input should be defined. Found ${inputShape[channelAxis]}`);
+    }
+    const inputDim = inputShape[channelAxis];
+    const numOfKernels = 4;
+    const kernelShape = this.kernelSize.concat([inputDim, this.filters * numOfKernels]);
+    this.kernel = this.addWeight("kernel", kernelShape, null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+    const recurrentKernelShape = this.kernelSize.concat([this.filters, this.filters * numOfKernels]);
+    this.recurrentKernel = this.addWeight("recurrent_kernel", recurrentKernelShape, null, this.recurrentInitializer, this.recurrentRegularizer, true, this.recurrentConstraint);
+    if (this.useBias) {
+      let biasInitializer;
+      if (this.unitForgetBias) {
+        const init2 = this.biasInitializer;
+        const filters = this.filters;
+        biasInitializer = new (_a = class CustomInit extends Initializer {
+          apply(shape, dtype) {
+            const biasI = init2.apply([filters]);
+            const biasF = ones2([filters]);
+            const biasCAndO = init2.apply([filters * 2]);
+            return concatenate([biasI, biasF, biasCAndO]);
+          }
+        }, _a.className = "CustomInit", _a)();
+      } else {
+        biasInitializer = this.biasInitializer;
+      }
+      this.bias = this.addWeight("bias", [this.filters * numOfKernels], null, biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (inputs.length !== 3) {
+        throw new ValueError(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${inputs.length}.`);
+      }
+      const training = kwargs["training"] || false;
+      const x = inputs[0];
+      const hTMinus1 = inputs[1];
+      const cTMinus1 = inputs[2];
+      const numOfKernels = 4;
+      if (0 < this.dropout && this.dropout < 1 && this.dropoutMask == null) {
+        this.dropoutMask = generateDropoutMask({
+          ones: () => onesLike(x),
+          rate: this.dropout,
+          training,
+          count: numOfKernels,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const dropoutMask = this.dropoutMask;
+      const applyDropout = (x2, mask, index) => {
+        if (!mask || !mask[index]) {
+          return x2;
+        }
+        return mul(mask[index], x2);
+      };
+      let xI = applyDropout(x, dropoutMask, 0);
+      let xF = applyDropout(x, dropoutMask, 1);
+      let xC = applyDropout(x, dropoutMask, 2);
+      let xO = applyDropout(x, dropoutMask, 3);
+      if (0 < this.recurrentDropout && this.recurrentDropout < 1 && this.recurrentDropoutMask == null) {
+        this.recurrentDropoutMask = generateDropoutMask({
+          ones: () => onesLike(hTMinus1),
+          rate: this.recurrentDropout,
+          training,
+          count: numOfKernels,
+          dropoutFunc: this.dropoutFunc
+        });
+      }
+      const recDropoutMask = this.recurrentDropoutMask;
+      let hI = applyDropout(hTMinus1, recDropoutMask, 0);
+      let hF = applyDropout(hTMinus1, recDropoutMask, 1);
+      let hC = applyDropout(hTMinus1, recDropoutMask, 2);
+      let hO = applyDropout(hTMinus1, recDropoutMask, 3);
+      const kernelChannelAxis = 3;
+      const [kernelI, kernelF, kernelC, kernelO] = split(this.kernel.read(), numOfKernels, kernelChannelAxis);
+      const [biasI, biasF, biasC, biasO] = this.useBias ? split(this.bias.read(), numOfKernels) : [null, null, null, null];
+      xI = this.inputConv(xI, kernelI, biasI, this.padding);
+      xF = this.inputConv(xF, kernelF, biasF, this.padding);
+      xC = this.inputConv(xC, kernelC, biasC, this.padding);
+      xO = this.inputConv(xO, kernelO, biasO, this.padding);
+      const [recKernelI, recKernelF, recKernelC, recKernelO] = split(this.recurrentKernel.read(), numOfKernels, kernelChannelAxis);
+      hI = this.recurrentConv(hI, recKernelI);
+      hF = this.recurrentConv(hF, recKernelF);
+      hC = this.recurrentConv(hC, recKernelC);
+      hO = this.recurrentConv(hO, recKernelO);
+      const i = this.recurrentActivation.apply(add2(xI, hI));
+      const f = this.recurrentActivation.apply(add2(xF, hF));
+      const c = add2(mul(f, cTMinus1), mul(i, this.activation.apply(add2(xC, hC))));
+      const h = mul(this.recurrentActivation.apply(add2(xO, hO)), this.activation.apply(c));
+      return [h, h, c];
+    });
+  }
+  getConfig() {
+    const _a = super.getConfig(), { "units": _ } = _a, baseConfig = __rest(_a, ["units"]);
+    const config = {
+      filters: this.filters,
+      kernelSize: this.kernelSize,
+      padding: this.padding,
+      dataFormat: this.dataFormat,
+      dilationRate: this.dilationRate,
+      strides: this.strides
+    };
+    return Object.assign(Object.assign({}, baseConfig), config);
+  }
+  inputConv(x, w, b, padding) {
+    const out = conv2d(x, w, this.strides, padding || "valid", this.dataFormat === "channelsFirst" ? "NCHW" : "NHWC", this.dilationRate);
+    if (b) {
+      return biasAdd(out, b, this.dataFormat);
+    }
+    return out;
+  }
+  recurrentConv(x, w) {
+    const strides = 1;
+    return conv2d(x, w, strides, "same", this.dataFormat === "channelsFirst" ? "NCHW" : "NHWC");
+  }
+};
+ConvLSTM2DCell.className = "ConvLSTM2DCell";
+serialization_exports.registerClass(ConvLSTM2DCell);
+var ConvLSTM2D = class extends ConvRNN2D {
+  constructor(args) {
+    const cell = new ConvLSTM2DCell(args);
+    super(Object.assign(Object.assign({}, args), { cell }));
+  }
+  static fromConfig(cls, config) {
+    return new cls(config);
+  }
+};
+ConvLSTM2D.className = "ConvLSTM2D";
+serialization_exports.registerClass(ConvLSTM2D);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/core.js
+var Dropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.rate = Math.max(Math.min(args.rate, 1), 0);
+    this.noiseShape = args.noiseShape;
+    this.seed = args.seed;
+    this.supportsMasking = true;
+  }
+  getNoiseShape(input2) {
+    if (this.noiseShape == null) {
+      return this.noiseShape;
+    }
+    const inputShape = input2.shape;
+    const noiseShape = [];
+    for (let i = 0; i < this.noiseShape.length; ++i) {
+      noiseShape.push(this.noiseShape[i] == null ? inputShape[i] : this.noiseShape[i]);
+    }
+    return noiseShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      if (0 < this.rate && this.rate < 1) {
+        const training = kwargs["training"] == null ? false : kwargs["training"];
+        const noiseShape = this.getNoiseShape(input2);
+        const output = inTrainPhase(() => dropout2(input2, this.rate, noiseShape, this.seed), () => input2, training);
+        return output;
+      }
+      return inputs;
+    });
+  }
+  getConfig() {
+    const config = {
+      rate: this.rate,
+      noiseShape: this.noiseShape,
+      seed: this.seed
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  dispose() {
+    return super.dispose();
+  }
+};
+Dropout.className = "Dropout";
+serialization_exports.registerClass(Dropout);
+var SpatialDropout1D = class extends Dropout {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [{ ndim: 3 }];
+  }
+  getNoiseShape(input2) {
+    const inputShape = input2.shape;
+    return [inputShape[0], 1, inputShape[2]];
+  }
+};
+SpatialDropout1D.className = "SpatialDropout1D";
+serialization_exports.registerClass(SpatialDropout1D);
+var Dense = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.activation = null;
+    this.useBias = true;
+    this.kernel = null;
+    this.bias = null;
+    this.DEFAULT_KERNEL_INITIALIZER = "glorotNormal";
+    this.DEFAULT_BIAS_INITIALIZER = "zeros";
+    if (args.batchInputShape == null && args.inputShape == null && args.inputDim != null) {
+      let batchSize = null;
+      if (args.batchSize != null) {
+        batchSize = args.batchSize;
+      }
+      this.batchInputShape = [batchSize, args.inputDim];
+    }
+    this.units = args.units;
+    assertPositiveInteger(this.units, "units");
+    this.activation = getActivation(args.activation);
+    if (args.useBias != null) {
+      this.useBias = args.useBias;
+    }
+    this.kernelInitializer = getInitializer(args.kernelInitializer || this.DEFAULT_KERNEL_INITIALIZER);
+    this.biasInitializer = getInitializer(args.biasInitializer || this.DEFAULT_BIAS_INITIALIZER);
+    this.kernelConstraint = getConstraint(args.kernelConstraint);
+    this.biasConstraint = getConstraint(args.biasConstraint);
+    this.kernelRegularizer = getRegularizer(args.kernelRegularizer);
+    this.biasRegularizer = getRegularizer(args.biasRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.supportsMasking = true;
+    this.inputSpec = [{ minNDim: 2 }];
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const inputLastDim = inputShape[inputShape.length - 1];
+    if (this.kernel == null) {
+      this.kernel = this.addWeight("kernel", [inputLastDim, this.units], null, this.kernelInitializer, this.kernelRegularizer, true, this.kernelConstraint);
+      if (this.useBias) {
+        this.bias = this.addWeight("bias", [this.units], null, this.biasInitializer, this.biasRegularizer, true, this.biasConstraint);
+      }
+    }
+    this.inputSpec = [{ minNDim: 2, axes: { [-1]: inputLastDim } }];
+    this.built = true;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    outputShape[outputShape.length - 1] = this.units;
+    return outputShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const fusedActivationName = mapActivationToFusedKernel(this.activation.getClassName());
+      let output;
+      if (fusedActivationName != null) {
+        output = dot2(input2, this.kernel.read(), fusedActivationName, this.bias ? this.bias.read() : null);
+      } else {
+        output = dot2(input2, this.kernel.read());
+        if (this.bias != null) {
+          output = biasAdd(output, this.bias.read());
+        }
+        if (this.activation != null) {
+          output = this.activation.apply(output);
+        }
+      }
+      return output;
+    });
+  }
+  getConfig() {
+    const config = {
+      units: this.units,
+      activation: serializeActivation(this.activation),
+      useBias: this.useBias,
+      kernelInitializer: serializeInitializer(this.kernelInitializer),
+      biasInitializer: serializeInitializer(this.biasInitializer),
+      kernelRegularizer: serializeRegularizer(this.kernelRegularizer),
+      biasRegularizer: serializeRegularizer(this.biasRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      kernelConstraint: serializeConstraint(this.kernelConstraint),
+      biasConstraint: serializeConstraint(this.biasConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Dense.className = "Dense";
+serialization_exports.registerClass(Dense);
+var Flatten = class extends Layer {
+  constructor(args) {
+    args = args || {};
+    super(args);
+    this.inputSpec = [{ minNDim: 3 }];
+    this.dataFormat = args.dataFormat;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    for (const dim of inputShape.slice(1)) {
+      if (dim == null) {
+        throw new ValueError(`The shape of the input to "Flatten" is not fully defined (got ${inputShape.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);
+      }
+    }
+    return [inputShape[0], arrayProd(inputShape, 1)];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      let input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsFirst" && input2.rank > 1) {
+        const permutation = [0];
+        for (let i = 2; i < input2.rank; ++i) {
+          permutation.push(i);
+        }
+        permutation.push(1);
+        input2 = transpose(input2, permutation);
+      }
+      return batchFlatten(input2);
+    });
+  }
+  getConfig() {
+    const config = {};
+    if (this.dataFormat != null) {
+      config["dataFormat"] = this.dataFormat;
+    }
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Flatten.className = "Flatten";
+serialization_exports.registerClass(Flatten);
+var Activation2 = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.activation = getActivation(args.activation);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      return this.activation.apply(input2);
+    });
+  }
+  getConfig() {
+    const config = { activation: serializeActivation(this.activation) };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Activation2.className = "Activation";
+serialization_exports.registerClass(Activation2);
+var RepeatVector = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.n = args.n;
+    this.inputSpec = [{ ndim: 2 }];
+  }
+  computeOutputShape(inputShape) {
+    return [inputShape[0], this.n, inputShape[1]];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      return repeat(inputs, this.n);
+    });
+  }
+  getConfig() {
+    const config = {
+      n: this.n
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+RepeatVector.className = "RepeatVector";
+serialization_exports.registerClass(RepeatVector);
+var Reshape2 = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.targetShape = args.targetShape;
+    for (let i = 0; i < this.targetShape.length; ++i) {
+      if (this.isUnknown(this.targetShape[i])) {
+        this.targetShape[i] = null;
+      }
+    }
+  }
+  isUnknown(dim) {
+    return dim < 0 || dim == null;
+  }
+  fixUnknownDimension(inputShape, outputShape) {
+    const errorMsg = "Total size of new array must be unchanged.";
+    const finalShape = outputShape.slice();
+    let known = 1;
+    let unknown = null;
+    for (let i = 0; i < finalShape.length; ++i) {
+      const dim = finalShape[i];
+      if (this.isUnknown(dim)) {
+        if (unknown === null) {
+          unknown = i;
+        } else {
+          throw new ValueError("Can only specifiy one unknown dimension.");
+        }
+      } else {
+        known *= dim;
+      }
+    }
+    const originalSize = arrayProd(inputShape);
+    if (unknown !== null) {
+      if (known === 0 || originalSize % known !== 0) {
+        throw new ValueError(errorMsg);
+      }
+      finalShape[unknown] = originalSize / known;
+    } else if (originalSize !== known) {
+      throw new ValueError(errorMsg);
+    }
+    return finalShape;
+  }
+  computeOutputShape(inputShape) {
+    let anyUnknownDims = false;
+    for (let i = 0; i < inputShape.length; ++i) {
+      if (this.isUnknown(inputShape[i])) {
+        anyUnknownDims = true;
+        break;
+      }
+    }
+    if (anyUnknownDims) {
+      return inputShape.slice(0, 1).concat(this.targetShape);
+    } else {
+      return inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      const outputShape = inputShape.slice(0, 1).concat(this.fixUnknownDimension(inputShape.slice(1), this.targetShape));
+      return reshape(input2, outputShape);
+    });
+  }
+  getConfig() {
+    const config = {
+      targetShape: this.targetShape
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Reshape2.className = "Reshape";
+serialization_exports.registerClass(Reshape2);
+var Permute = class extends Layer {
+  constructor(args) {
+    super(args);
+    if (args.dims == null) {
+      throw new Error("Required configuration field `dims` is missing during Permute constructor call.");
+    }
+    if (!Array.isArray(args.dims)) {
+      throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${args.dims} instead.`);
+    }
+    const expectedSortedIndices = range2(1, args.dims.length + 1);
+    if (!util_exports.arraysEqual(args.dims.slice().sort(), expectedSortedIndices)) {
+      throw new Error("Invalid permutation `dims`: " + JSON.stringify(args.dims) + " `dims` must contain consecutive integers starting from 1.");
+    }
+    this.dims = args.dims;
+    this.dimsIncludingBatch = [0].concat(this.dims);
+    this.inputSpec = [new InputSpec({ ndim: this.dims.length + 1 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const outputShape = inputShape.slice();
+    this.dims.forEach((dim, i) => {
+      outputShape[i + 1] = inputShape[dim];
+    });
+    return outputShape;
+  }
+  call(inputs, kwargs) {
+    return transpose(getExactlyOneTensor(inputs), this.dimsIncludingBatch);
+  }
+  getConfig() {
+    const config = {
+      dims: this.dims
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Permute.className = "Permute";
+serialization_exports.registerClass(Permute);
+var Masking = class extends Layer {
+  constructor(args) {
+    super(args == null ? {} : args);
+    this.supportsMasking = true;
+    if (args != null) {
+      this.maskValue = args.maskValue == null ? 0 : args.maskValue;
+    } else {
+      this.maskValue = 0;
+    }
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { maskValue: this.maskValue };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  computeMask(inputs, mask) {
+    const input2 = getExactlyOneTensor(inputs);
+    const axis = -1;
+    return any(notEqual(input2, this.maskValue), axis);
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const axis = -1;
+      const keepDims = true;
+      const booleanMask = any(notEqual(input2, this.maskValue), axis, keepDims);
+      const output = mul(input2, cast(booleanMask, input2.dtype));
+      return output;
+    });
+  }
+};
+Masking.className = "Masking";
+serialization_exports.registerClass(Masking);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/embeddings.js
+var Embedding = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.embeddings = null;
+    this.DEFAULT_EMBEDDINGS_INITIALIZER = "randomUniform";
+    if (args.batchInputShape == null && args.inputShape == null) {
+      let batchSize = null;
+      if (args.batchSize != null) {
+        batchSize = args.batchSize;
+      }
+      if (args.inputLength == null) {
+        this.batchInputShape = [batchSize, null];
+      } else {
+        this.batchInputShape = [batchSize].concat(toList(args.inputLength));
+      }
+    }
+    this.inputDim = args.inputDim;
+    assertPositiveInteger(this.inputDim, "inputDim");
+    this.outputDim = args.outputDim;
+    assertPositiveInteger(this.outputDim, "outputDim");
+    this.embeddingsInitializer = getInitializer(args.embeddingsInitializer || this.DEFAULT_EMBEDDINGS_INITIALIZER);
+    this.embeddingsRegularizer = getRegularizer(args.embeddingsRegularizer);
+    this.activityRegularizer = getRegularizer(args.activityRegularizer);
+    this.embeddingsConstraint = getConstraint(args.embeddingsConstraint);
+    this.maskZero = args.maskZero;
+    this.supportsMasking = args.maskZero;
+    this.inputLength = args.inputLength;
+  }
+  build(inputShape) {
+    this.embeddings = this.addWeight("embeddings", [this.inputDim, this.outputDim], this.dtype, this.embeddingsInitializer, this.embeddingsRegularizer, true, this.embeddingsConstraint);
+    this.built = true;
+  }
+  warnOnIncompatibleInputShape(inputShape) {
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (!this.maskZero) {
+        return null;
+      } else {
+        inputs = getExactlyOneTensor(inputs);
+        return notEqual(inputs, zerosLike(inputs));
+      }
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (this.inputLength == null) {
+      return [...inputShape, this.outputDim];
+    }
+    const inLens = toList(this.inputLength);
+    if (inLens.length !== inputShape.length - 1) {
+      throw new ValueError(`"inputLength" is ${this.inputLength}, but received input shape has shape ${inputShape}`);
+    } else {
+      let i = 0;
+      for (let k = 0; k < inLens.length; ++k) {
+        const s1 = inLens[k];
+        const s2 = inputShape[k + 1];
+        if (s1 != null && s2 != null && s1 !== s2) {
+          throw new ValueError(`"inputLength" is ${this.inputLength}, but received input shape has shape ${inputShape}`);
+        } else if (s1 == null) {
+          inLens[i] = s2;
+        }
+        i++;
+      }
+    }
+    return [inputShape[0], ...inLens, this.outputDim];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      let input2 = getExactlyOneTensor(inputs);
+      if (input2.dtype !== "int32") {
+        input2 = cast2(input2, "int32");
+      }
+      const output = gather2(this.embeddings.read(), reshape(input2, [input2.size]));
+      return reshape(output, getExactlyOneShape(this.computeOutputShape(input2.shape)));
+    });
+  }
+  getConfig() {
+    const config = {
+      inputDim: this.inputDim,
+      outputDim: this.outputDim,
+      embeddingsInitializer: serializeInitializer(this.embeddingsInitializer),
+      embeddingsRegularizer: serializeRegularizer(this.embeddingsRegularizer),
+      activityRegularizer: serializeRegularizer(this.activityRegularizer),
+      embeddingsConstraint: serializeConstraint(this.embeddingsConstraint),
+      maskZero: this.maskZero,
+      inputLength: this.inputLength
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Embedding.className = "Embedding";
+serialization_exports.registerClass(Embedding);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/merge.js
+var Merge = class extends Layer {
+  constructor(args) {
+    super(args || {});
+    this.supportsMasking = true;
+  }
+  mergeFunction(inputs) {
+    throw new NotImplementedError();
+  }
+  computeElementwiseOpOutputShape(shape1, shape2) {
+    if (shape1 == null || shape2 == null) {
+      return null;
+    } else if (shape1.length < shape2.length) {
+      return this.computeElementwiseOpOutputShape(shape2, shape1);
+    } else if (shape2.length === 0) {
+      return shape1;
+    }
+    const outputShape = shape1.slice(0, shape1.length - shape2.length);
+    for (let k = 0; k < shape2.length; ++k) {
+      const i = shape1[shape1.length - shape2.length + k];
+      const j = shape2[k];
+      if (i == null || j == null || i < 0 || j < 0) {
+        outputShape.push(null);
+      } else if (i === 1) {
+        outputShape.push(j);
+      } else if (j === 1) {
+        outputShape.push(i);
+      } else {
+        if (i !== j) {
+          throw new ValueError("Operands could not be broadcast together with shapes " + JSON.stringify(shape1) + " " + JSON.stringify(shape2));
+        }
+        outputShape.push(i);
+      }
+    }
+    return outputShape;
+  }
+  build(inputShape) {
+    if (Array.isArray(inputShape) && !Array.isArray(inputShape[0])) {
+      inputShape = [getExactlyOneShape(inputShape)];
+    }
+    inputShape = inputShape;
+    if (inputShape.length < 2) {
+      throw new ValueError(`A merge layer should be called on an Array of at least 2 inputs. Got ${inputShape.length} input(s).`);
+    }
+    let batchSizes = [];
+    for (const shape of inputShape) {
+      if (shape != null && shape[0] !== null) {
+        batchSizes.push(shape[0]);
+      }
+    }
+    batchSizes = unique2(batchSizes);
+    if (batchSizes.length > 1) {
+      throw new ValueError(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(inputShape)}.`);
+    }
+    let outputShape = inputShape[0] == null ? null : inputShape[0].slice(1);
+    for (let i = 1; i < inputShape.length; ++i) {
+      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);
+      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);
+    }
+    const allRanks = inputShape.map((shape) => shape.length);
+    if (inputShape.indexOf(null) === -1 && unique2(allRanks).length === 1) {
+      this.reshapeRequired = false;
+    } else {
+      this.reshapeRequired = true;
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = inputs;
+      if (this.reshapeRequired) {
+        const reshapedInputs = [];
+        const inputDims = inputs.map((input2) => input2.rank);
+        if (inputDims.indexOf(null) === -1) {
+          const maxNDim = max2(inputDims);
+          for (let x of inputs) {
+            const xNDim = x.rank;
+            for (let k = 0; k < maxNDim - xNDim; ++k) {
+              x = expandDims2(x, 1);
+            }
+            reshapedInputs.push(x);
+          }
+          return this.mergeFunction(reshapedInputs);
+        } else {
+          let transposed = false;
+          for (const x of inputs) {
+            const xNDim = x.rank;
+            if (xNDim == null) {
+              const xShape = x.shape;
+              const batchSize = xShape[0];
+              const newShape = xShape.slice(1).concat([batchSize]);
+              let xTransposed = reshape(x, [batchSize].concat(arrayProd(xShape.slice(1))));
+              xTransposed = transpose(xTransposed, [1, 0]);
+              xTransposed = reshape(xTransposed, newShape);
+              reshapedInputs.push(xTransposed);
+              transposed = true;
+            } else if (xNDim > 1) {
+              const dims = range2(1, xNDim).concat([0]);
+              reshapedInputs.push(transpose(x, dims));
+              transposed = true;
+            } else {
+              reshapedInputs.push(x);
+            }
+          }
+          let y = this.mergeFunction(reshapedInputs);
+          const yNDim = y.rank;
+          if (transposed) {
+            if (yNDim == null) {
+              const yShape = y.shape;
+              const yNDim2 = yShape.length;
+              const batchSize = yShape[yNDim2 - 1];
+              const newShape = [batchSize].concat(yShape.slice(0, yShape.length - 1));
+              y = reshape(transpose(reshape(y, [-1, batchSize]), [1, 0]), newShape);
+            } else if (yNDim > 1) {
+              const dims = [yNDim - 1].concat(range2(0, yNDim - 1));
+              y = transpose(y, dims);
+            }
+          }
+          return y;
+        }
+      } else {
+        return this.mergeFunction(inputs);
+      }
+    });
+  }
+  computeOutputShape(inputShape) {
+    inputShape = inputShape;
+    let outputShape;
+    if (inputShape[0] == null) {
+      outputShape = null;
+    } else {
+      outputShape = inputShape[0].slice(1);
+    }
+    for (let i = 1; i < inputShape.length; ++i) {
+      const shape = inputShape[i] == null ? null : inputShape[i].slice(1);
+      outputShape = this.computeElementwiseOpOutputShape(outputShape, shape);
+    }
+    let batchSizes = [];
+    for (const shape of inputShape) {
+      if (shape != null && shape[0] !== null) {
+        batchSizes.push(shape[0]);
+      }
+    }
+    batchSizes = unique2(batchSizes);
+    if (batchSizes.length === 1) {
+      outputShape = batchSizes.concat(outputShape);
+    } else {
+      outputShape = [null].concat(outputShape);
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    return tidy(() => {
+      if (mask == null) {
+        return null;
+      }
+      if (!Array.isArray(mask)) {
+        throw new ValueError("`mask` should be an Array");
+      }
+      if (!Array.isArray(inputs)) {
+        throw new ValueError("`inputs` should be an Array");
+      }
+      if (mask.length !== inputs.length) {
+        throw new ValueError(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${inputs.length} vs ${mask.length})`);
+      }
+      if (mask.every((m) => m == null)) {
+        return null;
+      }
+      mask = mask.map((m) => m == null ? m : expandDims(m, 0));
+      let output = mask[0];
+      for (let i = 1; i < mask.length - 1; ++i) {
+        output = logicalAnd(output, mask[i]);
+      }
+      return output;
+    });
+  }
+};
+var Add2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = add2(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Add2.className = "Add";
+serialization_exports.registerClass(Add2);
+var Multiply2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = mul(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Multiply2.className = "Multiply";
+serialization_exports.registerClass(Multiply2);
+var Average = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0].clone();
+      for (let i = 1; i < inputs.length; ++i) {
+        output = add2(output, inputs[i]);
+      }
+      return mul(1 / inputs.length, output);
+    });
+  }
+};
+Average.className = "Average";
+serialization_exports.registerClass(Average);
+var Maximum2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0];
+      for (let i = 1; i < inputs.length; ++i) {
+        output = maximum(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Maximum2.className = "Maximum";
+serialization_exports.registerClass(Maximum2);
+var Minimum2 = class extends Merge {
+  constructor(args) {
+    super(args);
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      let output = inputs[0];
+      for (let i = 1; i < inputs.length; ++i) {
+        output = minimum(output, inputs[i]);
+      }
+      return output;
+    });
+  }
+};
+Minimum2.className = "Minimum";
+serialization_exports.registerClass(Minimum2);
+var Concatenate = class extends Merge {
+  constructor(args) {
+    super(args);
+    this.DEFAULT_AXIS = -1;
+    if (args == null) {
+      args = {};
+    }
+    this.axis = args.axis == null ? this.DEFAULT_AXIS : args.axis;
+    this.supportsMasking = true;
+    this.reshapeRequired = false;
+  }
+  build(inputShape) {
+    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0])) || inputShape.length === 1) {
+      throw new ValueError("A `Concatenate` layer should be called on a list of at least 2 inputs");
+    }
+    inputShape = inputShape;
+    let allNoneShape = true;
+    for (const shape of inputShape) {
+      if (shape != null) {
+        allNoneShape = false;
+        break;
+      }
+    }
+    if (allNoneShape) {
+      return;
+    }
+    const shapeSet = [];
+    for (let i = 0; i < inputShape.length; ++i) {
+      const shapeWithoutConcatAxis = inputShape[i].slice();
+      shapeWithoutConcatAxis.splice(this.axis, 1);
+      let exists = false;
+      for (const shape of shapeSet) {
+        if (util_exports.arraysEqual(shape, shapeWithoutConcatAxis)) {
+          exists = true;
+          break;
+        }
+      }
+      if (!exists) {
+        shapeSet.push(shapeWithoutConcatAxis);
+      }
+    }
+    if (shapeSet.length > 1) {
+      throw new ValueError("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: " + JSON.stringify(inputShape));
+    }
+  }
+  mergeFunction(inputs) {
+    return tidy(() => {
+      return concatenate(inputs, this.axis);
+    });
+  }
+  computeOutputShape(inputShape) {
+    if (!(Array.isArray(inputShape) && Array.isArray(inputShape[0]))) {
+      throw new ValueError("A `Concatenate` layer should be called on a list of inputs.");
+    }
+    const inputShapes = inputShape;
+    const outputShape = inputShapes[0].slice();
+    const axis = this.axis < 0 ? outputShape.length + this.axis : this.axis;
+    for (const shape of inputShapes.slice(1)) {
+      if (outputShape[axis] == null || shape[axis] == null) {
+        outputShape[axis] = null;
+        break;
+      }
+      outputShape[axis] += shape[axis];
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    if (mask == null) {
+      return null;
+    }
+    if (!Array.isArray(mask)) {
+      throw new ValueError("`mask` should be an array for Concatenate");
+    }
+    if (!Array.isArray(inputs)) {
+      throw new ValueError("`inputs` should be an array for Concatenate");
+    }
+    if (mask.length !== inputs.length) {
+      throw new ValueError(`Mismatch in the length of mask (${mask.length}) and the legnth of inputs (${inputs.length})`);
+    }
+    return tidy(() => {
+      let allNullMasks = true;
+      mask.forEach((m) => {
+        if (m != null) {
+          allNullMasks = false;
+          return;
+        }
+      });
+      if (allNullMasks) {
+        return null;
+      }
+      const outputMasks = [];
+      for (let i = 0; i < inputs.length; ++i) {
+        if (mask[i] == null) {
+          outputMasks.push(cast(onesLike(inputs[i]), "bool"));
+        } else if (mask[i].rank < inputs[i].rank) {
+          outputMasks.push(expandDims(mask[i], -1));
+        } else {
+          outputMasks.push(mask[i]);
+        }
+      }
+      const concatenatedMasks = concat(outputMasks, this.axis);
+      return all(concatenatedMasks, -1, false);
+    });
+  }
+  getConfig() {
+    const config = {
+      "axis": this.axis
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Concatenate.className = "Concatenate";
+serialization_exports.registerClass(Concatenate);
+function interpretAxis(axis, dim) {
+  while (axis < 0) {
+    axis += dim;
+  }
+  return axis;
+}
+function batchDot(x, y, axes) {
+  if (x.shape.length > 3 || y.shape.length > 3) {
+    throw new NotImplementedError("batchDot is not implemented for tensors of 4D or higher rank yet");
+  }
+  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of x to be >= 2, but got ${x.shape.length}`);
+  util_exports.assert(x.shape.length >= 2, () => `batchDot requires the rank of y to be >= 2, but got ${y.shape.length}`);
+  if (typeof axes === "number") {
+    axes = [axes, axes];
+  }
+  if (x.dtype === "complex64" || y.dtype === "complex64") {
+    throw new NotImplementedError("batchDot is not implemented for complex64-type Tensors yet.");
+  }
+  const xNDim = x.shape.length;
+  const yNDim = y.shape.length;
+  if (axes == null) {
+    axes = [xNDim - 1, yNDim - 2];
+  }
+  const axesArray = axes;
+  return tidy(() => {
+    let diff;
+    if (xNDim > yNDim) {
+      diff = xNDim - yNDim;
+      const diffShape = [];
+      for (let i = 0; i < diff; ++i) {
+        diffShape.push(1);
+      }
+      y = reshape(y, y.shape.concat(diffShape));
+    } else if (yNDim > xNDim) {
+      diff = yNDim - xNDim;
+      const diffShape = [];
+      for (let i = 0; i < diff; ++i) {
+        diffShape.push(1);
+      }
+      x = reshape(x, x.shape.concat(diffShape));
+    } else {
+      diff = 0;
+    }
+    let out;
+    if (x.shape.length === 2 && y.shape.length === 2) {
+      if (axesArray[0] === axesArray[1]) {
+        out = sum2(mul(x, y), axesArray[0]);
+      } else {
+        out = sum2(mul(transpose(x, [1, 0]), y), axesArray[1]);
+      }
+    } else {
+      const adjX = axesArray[0] !== x.shape.length - 1;
+      const adjY = axesArray[1] === y.shape.length - 1;
+      out = matMul(x, y, adjX, adjY);
+    }
+    if (diff > 0) {
+      let idx;
+      if (xNDim > yNDim) {
+        idx = xNDim + yNDim - 3;
+      } else {
+        idx = xNDim - 1;
+      }
+      const squeezeAxes = [];
+      for (let i = idx; i < idx + diff; ++i) {
+        squeezeAxes.push(i);
+      }
+      out = squeeze(out, squeezeAxes);
+    }
+    if (out.shape.length === 1) {
+      out = expandDims(out, 1);
+    }
+    return out;
+  });
+}
+var Dot = class extends Merge {
+  constructor(args) {
+    super(args);
+    this.axes = args.axes;
+    this.normalize = args.normalize == null ? false : args.normalize;
+    this.supportsMasking = true;
+    this.reshapeRequired = false;
+  }
+  build(inputShape) {
+    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+    const shape1 = inputShape[0];
+    const shape2 = inputShape[1];
+    if (shape1.length > 3 || shape2.length > 3) {
+      throw new NotImplementedError("Dot layer does not support tensors of 4D or higher rank yet.");
+    }
+    const axes = this.interpretAxes(shape1, shape2);
+    if (shape1[axes[0]] !== shape2[axes[1]]) {
+      throw new ValueError(`Dimension incompatibility: ${shape1[axes[0]]} !== ${shape2[axes[1]]}`);
+    }
+  }
+  mergeFunction(inputs) {
+    if (inputs.length !== 2) {
+      throw new ValueError(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${inputs.length} input(s).`);
+    }
+    let x1 = inputs[0];
+    let x2 = inputs[1];
+    let axes;
+    if (!Array.isArray(this.axes)) {
+      axes = [
+        interpretAxis(this.axes, x1.shape.length),
+        interpretAxis(this.axes, x2.shape.length)
+      ];
+    } else {
+      axes = this.axes.map((axis, i) => interpretAxis(axis, inputs[i].shape.length));
+    }
+    if (this.normalize) {
+      x1 = l2Normalize(x1, axes[0]);
+      x2 = l2Normalize(x2, axes[1]);
+    }
+    return batchDot(x1, x2, axes);
+  }
+  interpretAxes(shape1, shape2) {
+    let axes;
+    if (!Array.isArray(this.axes)) {
+      axes = [
+        interpretAxis(this.axes, shape1.length),
+        interpretAxis(this.axes, shape2.length)
+      ];
+    } else {
+      axes = this.axes;
+    }
+    return axes;
+  }
+  computeOutputShape(inputShape) {
+    util_exports.assert(Array.isArray(inputShape) && inputShape.length === 2 && Array.isArray(inputShape[0]) && Array.isArray(inputShape[1]), () => "A `Dot` layer should be called on a list of exactly 2 inputs.");
+    const shape1 = inputShape[0].slice();
+    const shape2 = inputShape[1].slice();
+    if (shape1.length > 3 || shape2.length > 3) {
+      throw new NotImplementedError("Dot layer does not support tensors of 4D or higher rank yet.");
+    }
+    const axes = this.interpretAxes(shape1, shape2);
+    shape1.splice(axes[0], 1);
+    shape2.splice(axes[1], 1);
+    shape2.splice(0, 1);
+    const outputShape = shape1.concat(shape2);
+    if (outputShape.length === 1) {
+      outputShape.push(1);
+    }
+    return outputShape;
+  }
+  computeMask(inputs, mask) {
+    return null;
+  }
+  getConfig() {
+    const config = {
+      "axes": this.axes,
+      "normalize": this.normalize
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+Dot.className = "Dot";
+serialization_exports.registerClass(Dot);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/noise.js
+var GaussianNoise = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.stddev = args.stddev;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { stddev: this.stddev };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      const noised = () => add2(randomNormal2(input2.shape, 0, this.stddev), input2);
+      const output = inTrainPhase(noised, () => input2, kwargs["training"] || false);
+      return output;
+    });
+  }
+};
+GaussianNoise.className = "GaussianNoise";
+serialization_exports.registerClass(GaussianNoise);
+var GaussianDropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.rate = args.rate;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { rate: this.rate };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.rate > 0 && this.rate < 1) {
+        const noised = () => {
+          const stddev = Math.sqrt(this.rate / (1 - this.rate));
+          return mul(input2, randomNormal2(input2.shape, 1, stddev));
+        };
+        return inTrainPhase(noised, () => input2, kwargs["training"] || false);
+      }
+      return input2;
+    });
+  }
+};
+GaussianDropout.className = "GaussianDropout";
+serialization_exports.registerClass(GaussianDropout);
+var AlphaDropout = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+    this.rate = args.rate;
+    this.noiseShape = args.noiseShape;
+  }
+  _getNoiseShape(inputs) {
+    return this.noiseShape || getExactlyOneTensor(inputs).shape;
+  }
+  computeOutputShape(inputShape) {
+    return inputShape;
+  }
+  getConfig() {
+    const baseConfig = super.getConfig();
+    const config = { rate: this.rate };
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      if (this.rate < 1 && this.rate > 0) {
+        const noiseShape = this._getNoiseShape(inputs);
+        const droppedInputs = () => {
+          const input2 = getExactlyOneTensor(inputs);
+          const alpha = 1.6732632423543772;
+          const scale2 = 1.0507009873554805;
+          const alphaP = -alpha * scale2;
+          let keptIdx = greaterEqual(randomUniform(noiseShape), this.rate);
+          keptIdx = cast2(keptIdx, "float32");
+          const a = ((1 - this.rate) * (1 + this.rate * alphaP ** 2)) ** -0.5;
+          const b = -a * alphaP * this.rate;
+          const x = add2(mul(input2, keptIdx), mul(add2(keptIdx, -1), alphaP));
+          return add2(mul(x, a), b);
+        };
+        return inTrainPhase(droppedInputs, () => getExactlyOneTensor(inputs), kwargs["training"] || false);
+      }
+      return inputs;
+    });
+  }
+};
+AlphaDropout.className = "AlphaDropout";
+serialization_exports.registerClass(AlphaDropout);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/normalization.js
+function batchNormalization(x, mean4, variance, beta, gamma, epsilon3 = 1e-3) {
+  let out;
+  if (x.rank === 2) {
+    out = batchNorm2d(x, mean4, variance, beta, gamma, epsilon3);
+  } else if (x.rank === 3) {
+    out = batchNorm3d(x, mean4, variance, beta, gamma, epsilon3);
+  } else if (x.rank === 4) {
+    out = batchNorm4d(x, mean4, variance, beta, gamma, epsilon3);
+  } else {
+    throw new NotImplementedError(`batchNormalization is not implemented for array of rank ${x.rank} yet`);
+  }
+  return out;
+}
+function regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {
+  return tidy(() => {
+    const meanAndVariance = moments(x, reductionAxes);
+    const mean4 = meanAndVariance.mean;
+    const variance = meanAndVariance.variance;
+    const normed = batchNormalization(x, mean4, variance, beta, gamma, epsilon3);
+    return [normed, mean4, variance];
+  });
+}
+function broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {
+  return tidy(() => {
+    const meanAndVariance = moments(x, reductionAxes);
+    const mean4 = meanAndVariance.mean;
+    const variance = meanAndVariance.variance;
+    const targetShape = [];
+    for (const axis of range2(0, x.rank)) {
+      if (reductionAxes.indexOf(axis) !== -1) {
+        targetShape.push(1);
+      } else {
+        targetShape.push(x.shape[axis]);
+      }
+    }
+    const broadcastMean = reshape(mean4, targetShape);
+    const broadcastVariance = reshape(variance, targetShape);
+    const broadcastGamma = gamma == null ? null : reshape(gamma, targetShape);
+    const broadcastBeta = beta == null ? null : reshape(beta, targetShape);
+    const normed = batchNormalization(x, broadcastMean, broadcastVariance, broadcastBeta, broadcastGamma, epsilon3);
+    return [normed, mean4, variance];
+  });
+}
+function normalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3 = 1e-3) {
+  if (util_exports.arraysEqual(reductionAxes.slice().sort(), range2(0, x.rank - 1))) {
+    return regularNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3);
+  } else {
+    return broadcastNormalizeBatchInTraining(x, gamma, beta, reductionAxes, epsilon3);
+  }
+}
+var BatchNormalization = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.supportsMasking = true;
+    this.axis = args.axis == null ? -1 : args.axis;
+    this.momentum = args.momentum == null ? 0.99 : args.momentum;
+    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;
+    this.center = args.center == null ? true : args.center;
+    this.scale = args.scale == null ? true : args.scale;
+    this.betaInitializer = getInitializer(args.betaInitializer || "zeros");
+    this.gammaInitializer = getInitializer(args.gammaInitializer || "ones");
+    this.movingMeanInitializer = getInitializer(args.movingMeanInitializer || "zeros");
+    this.movingVarianceInitializer = getInitializer(args.movingVarianceInitializer || "ones");
+    this.betaConstraint = getConstraint(args.betaConstraint);
+    this.gammaConstraint = getConstraint(args.gammaConstraint);
+    this.betaRegularizer = getRegularizer(args.betaRegularizer);
+    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const axis = this.axis >= 0 ? this.axis : this.axis + inputShape.length;
+    const dim = inputShape[axis];
+    if (dim == null) {
+      throw new ValueError(`Axis ${axis} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(inputShape)}.`);
+    }
+    this.inputSpec = [new InputSpec({ ndim: inputShape.length, axes: { [axis]: dim } })];
+    const shape = [dim];
+    if (this.scale) {
+      this.gamma = this.addWeight("gamma", shape, null, this.gammaInitializer, this.gammaRegularizer, true, this.gammaConstraint);
+    }
+    if (this.center) {
+      this.beta = this.addWeight("beta", shape, null, this.betaInitializer, this.betaRegularizer, true, this.betaConstraint);
+    }
+    this.movingMean = this.addWeight("moving_mean", shape, null, this.movingMeanInitializer, null, false);
+    this.movingVariance = this.addWeight("moving_variance", shape, null, this.movingVarianceInitializer, null, false);
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const training = kwargs["training"] == null ? false : kwargs["training"];
+      const input2 = getExactlyOneTensor(inputs);
+      const inputShape = input2.shape;
+      const ndim = inputShape.length;
+      const reductionAxes = range2(0, ndim);
+      const axis = this.axis >= 0 ? this.axis : this.axis + ndim;
+      reductionAxes.splice(axis, 1);
+      const broadcastShape = pyListRepeat(1, ndim);
+      broadcastShape[axis] = inputShape[axis];
+      const sortedReductionAxes = reductionAxes.slice();
+      sortedReductionAxes.sort();
+      const needsBroadcasting = !util_exports.arraysEqual(sortedReductionAxes, range2(0, ndim).slice(0, ndim - 1));
+      const normalizeInference = () => {
+        if (needsBroadcasting) {
+          const broadcastMovingMean = reshape(this.movingMean.read(), broadcastShape);
+          const broadcastMovingVariance = reshape(this.movingVariance.read(), broadcastShape);
+          const broadcastBeta = this.center ? reshape(this.beta.read(), broadcastShape) : null;
+          const broadcastGamma = this.scale ? reshape(this.gamma.read(), broadcastShape) : null;
+          return batchNormalization(input2, broadcastMovingMean, broadcastMovingVariance, broadcastBeta, broadcastGamma, this.epsilon);
+        } else {
+          return batchNormalization(input2, this.movingMean.read(), this.movingVariance.read(), this.beta == null ? null : this.beta.read(), this.gamma == null ? null : this.gamma.read(), this.epsilon);
+        }
+      };
+      if (!training) {
+        return normalizeInference();
+      }
+      const [normedTraining, mean4, variance] = normalizeBatchInTraining(input2, this.gamma.read(), this.beta.read(), reductionAxes, this.epsilon);
+      const doMovingAverage = (variable2, value, momentum) => {
+        tidy(() => {
+          const decay = 1 - momentum;
+          const origValue = variable2.read();
+          const updateDelta = mul(sub(origValue, value), decay);
+          variable2.write(sub(origValue, updateDelta));
+        });
+      };
+      const updateMovingMeanAndVariance = () => {
+        doMovingAverage(this.movingMean, mean4, this.momentum);
+        doMovingAverage(this.movingVariance, variance, this.momentum);
+      };
+      updateMovingMeanAndVariance();
+      return normedTraining;
+    });
+  }
+  getConfig() {
+    const config = {
+      axis: this.axis,
+      momentum: this.momentum,
+      epsilon: this.epsilon,
+      center: this.center,
+      scale: this.scale,
+      betaInitializer: serializeInitializer(this.betaInitializer),
+      gammaInitializer: serializeInitializer(this.gammaInitializer),
+      movingMeanInitializer: serializeInitializer(this.movingMeanInitializer),
+      movingVarianceInitializer: serializeInitializer(this.movingVarianceInitializer),
+      betaRegularizer: serializeRegularizer(this.betaRegularizer),
+      gammaRegularizer: serializeRegularizer(this.gammaRegularizer),
+      betaConstraint: serializeConstraint(this.betaConstraint),
+      gammaConstraint: serializeConstraint(this.gammaConstraint)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+BatchNormalization.className = "BatchNormalization";
+serialization_exports.registerClass(BatchNormalization);
+var LayerNormalization = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.axis = args.axis == null ? -1 : args.axis;
+    if (typeof this.axis === "number") {
+      if (!Number.isInteger(this.axis)) {
+        throw new Error(`Expected axis to be an integer, but received ${this.axis}`);
+      }
+    } else if (Array.isArray(this.axis)) {
+      for (const axis of this.axis) {
+        if (!Number.isInteger(axis)) {
+          throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`);
+        }
+      }
+    } else {
+      throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);
+    }
+    this.epsilon = args.epsilon == null ? 1e-3 : args.epsilon;
+    this.center = args.center == null ? true : args.center;
+    this.scale = args.scale == null ? true : args.scale;
+    this.betaInitializer = getInitializer(args.betaInitializer || "zeros");
+    this.gammaInitializer = getInitializer(args.gammaInitializer || "ones");
+    this.betaRegularizer = getRegularizer(args.betaRegularizer);
+    this.gammaRegularizer = getRegularizer(args.gammaRegularizer);
+    this.supportsMasking = true;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const nDims = inputShape.length;
+    if (typeof this.axis === "number") {
+      this.axis = [this.axis];
+    }
+    for (let i = 0; i < this.axis.length; ++i) {
+      if (this.axis[i] < 0) {
+        this.axis[i] += nDims;
+      }
+    }
+    for (const axis of this.axis) {
+      if (axis < 0 || axis >= nDims) {
+        throw new Error(`Invalid axis: ${axis}`);
+      }
+    }
+    if (this.axis.length !== unique2(this.axis).length) {
+      throw new Error(`Found duplicate axes in: ${this.axis}`);
+    }
+    const paramShape = this.axis.map((axis) => inputShape[axis]);
+    const trainable = true;
+    if (this.scale) {
+      this.gamma = this.addWeight("gamma", paramShape, "float32", this.gammaInitializer, this.gammaRegularizer, trainable);
+    } else {
+      this.gamma = null;
+    }
+    if (this.center) {
+      this.beta = this.addWeight("beta", paramShape, "float32", this.betaInitializer, this.betaRegularizer, trainable);
+    } else {
+      this.beta = null;
+    }
+    this.built = true;
+  }
+  call(inputs, kwargs) {
+    const input2 = getExactlyOneTensor(inputs);
+    const inputShape = input2.shape;
+    const nDims = inputShape.length;
+    return tidy(() => {
+      const keepDims = true;
+      let { mean: mean4, variance } = moments(input2, this.axis, keepDims);
+      const broadcastShape = pyListRepeat(1, nDims);
+      for (const dim of this.axis) {
+        broadcastShape[dim] = inputShape[dim];
+      }
+      const broadcast = (v) => {
+        if (v != null && v.shape.length !== nDims) {
+          return reshape(v, broadcastShape);
+        } else {
+          return v;
+        }
+      };
+      let scale2 = this.scale ? broadcast(this.gamma.read()) : null;
+      let offset = this.center ? broadcast(this.beta.read()) : null;
+      const momentsTiling = [];
+      const scaleOffsetTiling = [];
+      for (let i = 0; i < nDims; ++i) {
+        if (this.axis.indexOf(i) !== -1) {
+          momentsTiling.push(inputShape[i]);
+          scaleOffsetTiling.push(1);
+        } else {
+          momentsTiling.push(1);
+          scaleOffsetTiling.push(inputShape[i]);
+        }
+      }
+      mean4 = tile(mean4, momentsTiling);
+      variance = tile(variance, momentsTiling);
+      if (scale2 != null) {
+        scale2 = tile(scale2, scaleOffsetTiling);
+      }
+      if (offset != null) {
+        offset = tile(offset, scaleOffsetTiling);
+      }
+      return batchNormalization(input2, mean4, variance, offset, scale2, this.epsilon);
+    });
+  }
+  getConfig() {
+    const config = {
+      axis: this.axis,
+      epsilon: this.epsilon,
+      center: this.center,
+      scale: this.scale,
+      betaInitializer: serializeInitializer(this.betaInitializer),
+      gammaInitializer: serializeInitializer(this.gammaInitializer),
+      betaRegularizer: serializeRegularizer(this.betaRegularizer),
+      gammaRegularizer: serializeRegularizer(this.gammaRegularizer)
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+LayerNormalization.className = "LayerNormalization";
+serialization_exports.registerClass(LayerNormalization);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/padding.js
+function spatial2dPadding(x, padding, dataFormat) {
+  return tidy(() => {
+    if (x.rank !== 4) {
+      throw new ValueError(`temporalPadding expects input tensor to be 4-D, but received a ${x.rank}-D tensor.`);
+    }
+    if (padding == null) {
+      padding = [[1, 1], [1, 1]];
+    }
+    if (padding.length !== 2 || padding[0].length !== 2 || padding[1].length !== 2) {
+      throw new ValueError("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (dataFormat !== "channelsLast" && dataFormat !== "channelsFirst") {
+      throw new ValueError(`Unknown data format: ${dataFormat}. Supported data formats are 'channelsLast' and 'channelsFirst.`);
+    }
+    let pattern;
+    if (dataFormat === "channelsFirst") {
+      pattern = [[0, 0], [0, 0], padding[0], padding[1]];
+    } else {
+      pattern = [[0, 0], padding[0], padding[1], [0, 0]];
+    }
+    return pad(x, pattern);
+  });
+}
+var ZeroPadding2D = class extends Layer {
+  constructor(args) {
+    if (args == null) {
+      args = {};
+    }
+    super(args);
+    this.dataFormat = args.dataFormat == null ? imageDataFormat() : args.dataFormat;
+    if (args.padding == null) {
+      this.padding = [[1, 1], [1, 1]];
+    } else if (typeof args.padding === "number") {
+      this.padding = [[args.padding, args.padding], [args.padding, args.padding]];
+    } else {
+      args.padding = args.padding;
+      if (args.padding.length !== 2) {
+        throw new ValueError(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${args.padding.length} array.`);
+      }
+      let heightPadding;
+      let widthPadding;
+      if (typeof args.padding[0] === "number") {
+        heightPadding = [args.padding[0], args.padding[0]];
+        widthPadding = [args.padding[1], args.padding[1]];
+      } else {
+        args.padding = args.padding;
+        if (args.padding[0].length !== 2) {
+          throw new ValueError(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${args.padding[0].length} array.`);
+        }
+        heightPadding = args.padding[0];
+        if (args.padding[1].length !== 2) {
+          throw new ValueError(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${args.padding[1].length} array.`);
+        }
+        widthPadding = args.padding[1];
+      }
+      this.padding = [heightPadding, widthPadding];
+    }
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let rows;
+    let cols;
+    if (this.dataFormat === "channelsFirst") {
+      if (inputShape[2] != null && inputShape[2] >= 0) {
+        rows = inputShape[2] + this.padding[0][0] + this.padding[0][1];
+      } else {
+        rows = null;
+      }
+      if (inputShape[3] != null && inputShape[3] >= 0) {
+        cols = inputShape[3] + this.padding[1][0] + this.padding[1][1];
+      } else {
+        cols = null;
+      }
+      return [inputShape[0], inputShape[1], rows, cols];
+    } else {
+      if (inputShape[1] != null && inputShape[1] >= 0) {
+        rows = inputShape[1] + this.padding[0][0] + this.padding[0][1];
+      } else {
+        rows = null;
+      }
+      if (inputShape[2] != null && inputShape[2] >= 0) {
+        cols = inputShape[2] + this.padding[1][0] + this.padding[1][1];
+      } else {
+        cols = null;
+      }
+      return [inputShape[0], rows, cols, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => spatial2dPadding(getExactlyOneTensor(inputs), this.padding, this.dataFormat));
+  }
+  getConfig() {
+    const config = {
+      padding: this.padding,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+ZeroPadding2D.className = "ZeroPadding2D";
+serialization_exports.registerClass(ZeroPadding2D);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/pooling.js
+function pool2d(x, poolSize, strides, padding, dataFormat, poolMode) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    checkPoolMode(poolMode);
+    checkPaddingMode(padding);
+    if (strides == null) {
+      strides = [1, 1];
+    }
+    if (padding == null) {
+      padding = "valid";
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (poolMode == null) {
+      poolMode = "max";
+    }
+    x = preprocessConv2DInput(x, dataFormat);
+    let y;
+    const paddingString = padding === "same" ? "same" : "valid";
+    if (poolMode === "max") {
+      y = maxPool(x, poolSize, strides, paddingString);
+    } else {
+      y = avgPool(
+        x,
+        poolSize,
+        strides,
+        paddingString
+      );
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 3, 1, 2]);
+    }
+    return y;
+  });
+}
+function pool3d(x, poolSize, strides, padding, dataFormat, poolMode) {
+  return tidy(() => {
+    checkDataFormat(dataFormat);
+    checkPoolMode(poolMode);
+    checkPaddingMode(padding);
+    if (strides == null) {
+      strides = [1, 1, 1];
+    }
+    if (padding == null) {
+      padding = "valid";
+    }
+    if (dataFormat == null) {
+      dataFormat = imageDataFormat();
+    }
+    if (poolMode == null) {
+      poolMode = "max";
+    }
+    x = preprocessConv3DInput(x, dataFormat);
+    let y;
+    const paddingString = padding === "same" ? "same" : "valid";
+    if (poolMode === "max") {
+      y = maxPool3d(x, poolSize, strides, paddingString);
+    } else {
+      y = avgPool3d(x, poolSize, strides, paddingString);
+    }
+    if (dataFormat === "channelsFirst") {
+      y = transpose(y, [0, 4, 1, 2, 3]);
+    }
+    return y;
+  });
+}
+var Pooling1D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = 2;
+    }
+    super(args);
+    if (typeof args.poolSize === "number") {
+      this.poolSize = [args.poolSize];
+    } else if (Array.isArray(args.poolSize) && args.poolSize.length === 1 && typeof args.poolSize[0] === "number") {
+      this.poolSize = args.poolSize;
+    } else {
+      throw new ValueError(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.poolSize)}`);
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else {
+      if (typeof args.strides === "number") {
+        this.strides = [args.strides];
+      } else if (Array.isArray(args.strides) && args.strides.length === 1 && typeof args.strides[0] === "number") {
+        this.strides = args.strides;
+      } else {
+        throw new ValueError(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(args.strides)}`);
+      }
+    }
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const length = convOutputLength(inputShape[1], this.poolSize[0], this.padding, this.strides[0]);
+    return [inputShape[0], length, inputShape[2]];
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      inputs = expandDims2(getExactlyOneTensor(inputs), 2);
+      const output = this.poolingFunction(getExactlyOneTensor(inputs), [this.poolSize[0], 1], [this.strides[0], 1], this.padding, "channelsLast");
+      return squeeze(output, [2]);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling1D = class extends Pooling1D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling1D.className = "MaxPooling1D";
+serialization_exports.registerClass(MaxPooling1D);
+var AveragePooling1D = class extends Pooling1D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling1D.className = "AveragePooling1D";
+serialization_exports.registerClass(AveragePooling1D);
+var Pooling2D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = [2, 2];
+    }
+    super(args);
+    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize];
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else if (Array.isArray(args.strides)) {
+      if (args.strides.length !== 2) {
+        throw new ValueError(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${args.strides.length}.`);
+      }
+      this.strides = args.strides;
+    } else {
+      this.strides = [args.strides, args.strides];
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let rows = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    let cols = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    rows = convOutputLength(rows, this.poolSize[0], this.padding, this.strides[0]);
+    cols = convOutputLength(cols, this.poolSize[1], this.padding, this.strides[1]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], inputShape[1], rows, cols];
+    } else {
+      return [inputShape[0], rows, cols, inputShape[3]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling2D = class extends Pooling2D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling2D.className = "MaxPooling2D";
+serialization_exports.registerClass(MaxPooling2D);
+var AveragePooling2D = class extends Pooling2D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool2d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling2D.className = "AveragePooling2D";
+serialization_exports.registerClass(AveragePooling2D);
+var Pooling3D = class extends Layer {
+  constructor(args) {
+    if (args.poolSize == null) {
+      args.poolSize = [2, 2, 2];
+    }
+    super(args);
+    this.poolSize = Array.isArray(args.poolSize) ? args.poolSize : [args.poolSize, args.poolSize, args.poolSize];
+    if (args.strides == null) {
+      this.strides = this.poolSize;
+    } else if (Array.isArray(args.strides)) {
+      if (args.strides.length !== 3) {
+        throw new ValueError(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${args.strides.length}.`);
+      }
+      this.strides = args.strides;
+    } else {
+      this.strides = [args.strides, args.strides, args.strides];
+    }
+    assertPositiveInteger(this.poolSize, "poolSize");
+    assertPositiveInteger(this.strides, "strides");
+    this.padding = args.padding == null ? "valid" : args.padding;
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    checkPaddingMode(this.padding);
+    this.inputSpec = [new InputSpec({ ndim: 5 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    let depths = this.dataFormat === "channelsFirst" ? inputShape[2] : inputShape[1];
+    let rows = this.dataFormat === "channelsFirst" ? inputShape[3] : inputShape[2];
+    let cols = this.dataFormat === "channelsFirst" ? inputShape[4] : inputShape[3];
+    depths = convOutputLength(depths, this.poolSize[0], this.padding, this.strides[0]);
+    rows = convOutputLength(rows, this.poolSize[1], this.padding, this.strides[1]);
+    cols = convOutputLength(cols, this.poolSize[2], this.padding, this.strides[2]);
+    if (this.dataFormat === "channelsFirst") {
+      return [inputShape[0], inputShape[1], depths, rows, cols];
+    } else {
+      return [inputShape[0], depths, rows, cols, inputShape[4]];
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      this.invokeCallHook(inputs, kwargs);
+      return this.poolingFunction(getExactlyOneTensor(inputs), this.poolSize, this.strides, this.padding, this.dataFormat);
+    });
+  }
+  getConfig() {
+    const config = {
+      poolSize: this.poolSize,
+      padding: this.padding,
+      strides: this.strides,
+      dataFormat: this.dataFormat
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var MaxPooling3D = class extends Pooling3D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool3d(inputs, poolSize, strides, padding, dataFormat, "max");
+  }
+};
+MaxPooling3D.className = "MaxPooling3D";
+serialization_exports.registerClass(MaxPooling3D);
+var AveragePooling3D = class extends Pooling3D {
+  constructor(args) {
+    super(args);
+  }
+  poolingFunction(inputs, poolSize, strides, padding, dataFormat) {
+    checkDataFormat(dataFormat);
+    checkPaddingMode(padding);
+    return pool3d(inputs, poolSize, strides, padding, dataFormat, "avg");
+  }
+};
+AveragePooling3D.className = "AveragePooling3D";
+serialization_exports.registerClass(AveragePooling3D);
+var GlobalPooling1D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.inputSpec = [new InputSpec({ ndim: 3 })];
+  }
+  computeOutputShape(inputShape) {
+    return [inputShape[0], inputShape[2]];
+  }
+  call(inputs, kwargs) {
+    throw new NotImplementedError();
+  }
+};
+var GlobalAveragePooling1D = class extends GlobalPooling1D {
+  constructor(args) {
+    super(args || {});
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      return mean(input2, 1);
+    });
+  }
+};
+GlobalAveragePooling1D.className = "GlobalAveragePooling1D";
+serialization_exports.registerClass(GlobalAveragePooling1D);
+var GlobalMaxPooling1D = class extends GlobalPooling1D {
+  constructor(args) {
+    super(args || {});
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      return max(input2, 1);
+    });
+  }
+};
+GlobalMaxPooling1D.className = "GlobalMaxPooling1D";
+serialization_exports.registerClass(GlobalMaxPooling1D);
+var GlobalPooling2D = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.dataFormat = args.dataFormat == null ? "channelsLast" : args.dataFormat;
+    checkDataFormat(this.dataFormat);
+    this.inputSpec = [new InputSpec({ ndim: 4 })];
+  }
+  computeOutputShape(inputShape) {
+    inputShape = inputShape;
+    if (this.dataFormat === "channelsLast") {
+      return [inputShape[0], inputShape[3]];
+    } else {
+      return [inputShape[0], inputShape[1]];
+    }
+  }
+  call(inputs, kwargs) {
+    throw new NotImplementedError();
+  }
+  getConfig() {
+    const config = { dataFormat: this.dataFormat };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+};
+var GlobalAveragePooling2D = class extends GlobalPooling2D {
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        return mean(input2, [1, 2]);
+      } else {
+        return mean(input2, [2, 3]);
+      }
+    });
+  }
+};
+GlobalAveragePooling2D.className = "GlobalAveragePooling2D";
+serialization_exports.registerClass(GlobalAveragePooling2D);
+var GlobalMaxPooling2D = class extends GlobalPooling2D {
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const input2 = getExactlyOneTensor(inputs);
+      if (this.dataFormat === "channelsLast") {
+        return max(input2, [1, 2]);
+      } else {
+        return max(input2, [2, 3]);
+      }
+    });
+  }
+};
+GlobalMaxPooling2D.className = "GlobalMaxPooling2D";
+serialization_exports.registerClass(GlobalMaxPooling2D);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/wrappers.js
+var Wrapper = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.layer = args.layer;
+  }
+  build(inputShape) {
+    this.built = true;
+  }
+  get trainable() {
+    if (this.layer != null) {
+      return this.layer.trainable;
+    } else {
+      return false;
+    }
+  }
+  set trainable(value) {
+    if (this.layer != null) {
+      this.layer.trainable = value;
+    }
+  }
+  get trainableWeights() {
+    return this.layer.trainableWeights;
+  }
+  get nonTrainableWeights() {
+    return this.layer.nonTrainableWeights;
+  }
+  get updates() {
+    return this.layer._updates;
+  }
+  get losses() {
+    return this.layer.losses;
+  }
+  getWeights() {
+    return this.layer.getWeights();
+  }
+  setWeights(weights) {
+    this.layer.setWeights(weights);
+  }
+  getConfig() {
+    const config = {
+      "layer": {
+        "className": this.layer.getClassName(),
+        "config": this.layer.getConfig()
+      }
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.layer != null) {
+      this.layer.setFastWeightInitDuringBuild(value);
+    }
+  }
+  static fromConfig(cls, config, customObjects = {}) {
+    const layerConfig = config["layer"];
+    const layer = deserialize(layerConfig, customObjects);
+    delete config["layer"];
+    const newConfig = { layer };
+    Object.assign(newConfig, config);
+    return new cls(newConfig);
+  }
+};
+var TimeDistributed = class extends Wrapper {
+  constructor(args) {
+    super(args);
+    this.supportsMasking = true;
+  }
+  build(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape.length < 3) {
+      throw new ValueError(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(inputShape)}`);
+    }
+    this.inputSpec = [{ shape: inputShape }];
+    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    if (!this.layer.built) {
+      this.layer.build(childInputShape);
+      this.layer.built = true;
+    }
+    super.build(inputShape);
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const childInputShape = [inputShape[0]].concat(inputShape.slice(2));
+    const childOutputShape = this.layer.computeOutputShape(childInputShape);
+    const timesteps = inputShape[1];
+    return [childOutputShape[0], timesteps].concat(childOutputShape.slice(1));
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      const step5 = (inputs2, states) => {
+        const output = getExactlyOneTensor(this.layer.call(inputs2, kwargs));
+        return [output, []];
+      };
+      const rnnOutputs = rnn(step5, inputs, [], false, null, null, false, true);
+      const y = rnnOutputs[1];
+      return y;
+    });
+  }
+};
+TimeDistributed.className = "TimeDistributed";
+serialization_exports.registerClass(TimeDistributed);
+function checkBidirectionalMergeMode(value) {
+  checkStringTypeUnionValue(VALID_BIDIRECTIONAL_MERGE_MODES, "BidirectionalMergeMode", value);
+}
+var DEFAULT_BIDIRECTIONAL_MERGE_MODE = "concat";
+var Bidirectional = class extends Wrapper {
+  constructor(args) {
+    super(args);
+    const layerConfig = args.layer.getConfig();
+    const forwDict = {};
+    forwDict["className"] = args.layer.getClassName();
+    forwDict["config"] = layerConfig;
+    this.forwardLayer = deserialize(forwDict);
+    layerConfig["goBackwards"] = layerConfig["goBackwards"] === true ? false : true;
+    const backDict = {};
+    backDict["className"] = args.layer.getClassName();
+    backDict["config"] = layerConfig;
+    this.backwardLayer = deserialize(backDict);
+    this.forwardLayer.name = "forward_" + this.forwardLayer.name;
+    this.backwardLayer.name = "backward_" + this.backwardLayer.name;
+    this.mergeMode = args.mergeMode === void 0 ? DEFAULT_BIDIRECTIONAL_MERGE_MODE : args.mergeMode;
+    checkBidirectionalMergeMode(this.mergeMode);
+    if (args.weights) {
+      throw new NotImplementedError("weights support is not implemented for Bidirectional layer yet.");
+    }
+    this._stateful = args.layer.stateful;
+    this.returnSequences = args.layer.returnSequences;
+    this.returnState = args.layer.returnState;
+    this.supportsMasking = true;
+    this._trainable = true;
+    this.inputSpec = args.layer.inputSpec;
+    this.numConstants = null;
+  }
+  get trainable() {
+    return this._trainable;
+  }
+  set trainable(value) {
+    this._trainable = value;
+    if (this.forwardLayer != null) {
+      this.forwardLayer.trainable = value;
+    }
+    if (this.backwardLayer != null) {
+      this.backwardLayer.trainable = value;
+    }
+  }
+  getWeights() {
+    return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights());
+  }
+  setWeights(weights) {
+    const numWeights = weights.length;
+    const numeightsOver2 = Math.floor(numWeights / 2);
+    this.forwardLayer.setWeights(weights.slice(0, numeightsOver2));
+    this.backwardLayer.setWeights(weights.slice(numeightsOver2));
+  }
+  computeOutputShape(inputShape) {
+    let layerShapes = this.forwardLayer.computeOutputShape(inputShape);
+    if (!(Array.isArray(layerShapes) && Array.isArray(layerShapes[0]))) {
+      layerShapes = [layerShapes];
+    }
+    layerShapes = layerShapes;
+    let outputShape;
+    let outputShapes;
+    let stateShape;
+    if (this.returnState) {
+      stateShape = layerShapes.slice(1);
+      outputShape = layerShapes[0];
+    } else {
+      outputShape = layerShapes[0];
+    }
+    outputShape = outputShape;
+    if (this.mergeMode === "concat") {
+      outputShape[outputShape.length - 1] *= 2;
+      outputShapes = [outputShape];
+    } else if (this.mergeMode == null) {
+      outputShapes = [outputShape, outputShape.slice()];
+    } else {
+      outputShapes = [outputShape];
+    }
+    if (this.returnState) {
+      if (this.mergeMode == null) {
+        return outputShapes.concat(stateShape).concat(stateShape.slice());
+      }
+      return [outputShape].concat(stateShape).concat(stateShape.slice());
+    }
+    return singletonOrArray(outputShapes);
+  }
+  apply(inputs, kwargs) {
+    let initialState = kwargs == null ? null : kwargs["initialState"];
+    let constants = kwargs == null ? null : kwargs["constants"];
+    if (kwargs == null) {
+      kwargs = {};
+    }
+    const standardized = standardizeArgs(inputs, initialState, constants, this.numConstants);
+    inputs = standardized.inputs;
+    initialState = standardized.initialState;
+    constants = standardized.constants;
+    if (Array.isArray(inputs)) {
+      initialState = inputs.slice(1);
+      inputs = inputs[0];
+    }
+    if ((initialState == null || initialState.length === 0) && constants == null) {
+      return super.apply(inputs, kwargs);
+    }
+    const additionalInputs = [];
+    const additionalSpecs = [];
+    if (initialState != null) {
+      const numStates = initialState.length;
+      if (numStates % 2 > 0) {
+        throw new ValueError("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");
+      }
+      kwargs["initialState"] = initialState;
+      additionalInputs.push(...initialState);
+      const stateSpecs = initialState.map((state) => new InputSpec({ shape: state.shape }));
+      this.forwardLayer.stateSpec = stateSpecs.slice(0, numStates / 2);
+      this.backwardLayer.stateSpec = stateSpecs.slice(numStates / 2);
+      additionalSpecs.push(...stateSpecs);
+    }
+    if (constants != null) {
+      throw new NotImplementedError("Support for constants in Bidirectional layers is not implemented yet.");
+    }
+    const isSymbolicTensor = additionalInputs[0] instanceof SymbolicTensor;
+    for (const tensor2 of additionalInputs) {
+      if (tensor2 instanceof SymbolicTensor !== isSymbolicTensor) {
+        throw new ValueError("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");
+      }
+    }
+    if (isSymbolicTensor) {
+      const fullInput = [inputs].concat(additionalInputs);
+      const fullInputSpec = this.inputSpec.concat(additionalSpecs);
+      const originalInputSpec = this.inputSpec;
+      this.inputSpec = fullInputSpec;
+      const output = super.apply(fullInput, kwargs);
+      this.inputSpec = originalInputSpec;
+      return output;
+    } else {
+      return super.apply(inputs, kwargs);
+    }
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const initialState = kwargs["initialState"];
+      let y;
+      let yRev;
+      if (initialState == null) {
+        y = this.forwardLayer.call(inputs, kwargs);
+        yRev = this.backwardLayer.call(inputs, kwargs);
+      } else {
+        const forwardState = initialState.slice(0, initialState.length / 2);
+        const backwardState = initialState.slice(initialState.length / 2);
+        y = this.forwardLayer.call(inputs, Object.assign(kwargs, { initialState: forwardState }));
+        yRev = this.backwardLayer.call(inputs, Object.assign(kwargs, { initialState: backwardState }));
+      }
+      let states;
+      if (this.returnState) {
+        if (Array.isArray(y)) {
+          states = y.slice(1).concat(yRev.slice(1));
+        } else {
+        }
+        y = y[0];
+        yRev = yRev[0];
+      }
+      if (this.returnSequences) {
+        yRev = reverse(yRev, 1);
+      }
+      let output;
+      if (this.mergeMode === "concat") {
+        output = concatenate([y, yRev]);
+      } else if (this.mergeMode === "sum") {
+        output = add2(y, yRev);
+      } else if (this.mergeMode === "ave") {
+        output = mul(0.5, add2(y, yRev));
+      } else if (this.mergeMode === "mul") {
+        output = mul(y, yRev);
+      } else if (this.mergeMode == null) {
+        output = [y, yRev];
+      }
+      if (this.returnState) {
+        if (this.mergeMode == null) {
+          return output.concat(states);
+        }
+        return [output].concat(states);
+      }
+      return output;
+    });
+  }
+  resetStates(states) {
+    this.forwardLayer.resetStates();
+    this.backwardLayer.resetStates();
+  }
+  build(inputShape) {
+    nameScope(this.forwardLayer.name, () => {
+      this.forwardLayer.build(inputShape);
+    });
+    nameScope(this.backwardLayer.name, () => {
+      this.backwardLayer.build(inputShape);
+    });
+    this.built = true;
+  }
+  computeMask(inputs, mask) {
+    if (Array.isArray(mask)) {
+      mask = mask[0];
+    }
+    let outputMask;
+    if (this.returnSequences) {
+      if (this.mergeMode == null) {
+        outputMask = [mask, mask];
+      } else {
+        outputMask = mask;
+      }
+    } else {
+      if (this.mergeMode == null) {
+        outputMask = [null, null];
+      } else {
+        outputMask = null;
+      }
+    }
+    if (this.returnState) {
+      const states = this.forwardLayer.states;
+      const stateMask = states.map((state) => null);
+      if (Array.isArray(outputMask)) {
+        return outputMask.concat(stateMask).concat(stateMask);
+      } else {
+        return [outputMask].concat(stateMask).concat(stateMask);
+      }
+    } else {
+      return outputMask;
+    }
+  }
+  get trainableWeights() {
+    return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights);
+  }
+  get nonTrainableWeights() {
+    return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights);
+  }
+  setFastWeightInitDuringBuild(value) {
+    super.setFastWeightInitDuringBuild(value);
+    if (this.forwardLayer != null) {
+      this.forwardLayer.setFastWeightInitDuringBuild(value);
+    }
+    if (this.backwardLayer != null) {
+      this.backwardLayer.setFastWeightInitDuringBuild(value);
+    }
+  }
+  getConfig() {
+    const config = {
+      "mergeMode": this.mergeMode
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  static fromConfig(cls, config) {
+    const rnnLayer = deserialize(config["layer"]);
+    delete config["layer"];
+    if (config["numConstants"] != null) {
+      throw new NotImplementedError(`Deserialization of a Bidirectional layer with numConstants present is not supported yet.`);
+    }
+    const newConfig = config;
+    newConfig["layer"] = rnnLayer;
+    return new cls(newConfig);
+  }
+};
+Bidirectional.className = "Bidirectional";
+serialization_exports.registerClass(Bidirectional);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/image_preprocessing.js
+var Rescaling = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.scale = args.scale;
+    if (args.offset) {
+      this.offset = args.offset;
+    } else {
+      this.offset = 0;
+    }
+  }
+  getConfig() {
+    const config = {
+      "scale": this.scale,
+      "offset": this.offset
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (inputs.dtype !== "float32") {
+        inputs = cast2(inputs, "float32");
+      }
+      return add2(mul(inputs, this.scale), this.offset);
+    });
+  }
+};
+Rescaling.className = "Rescaling";
+serialization_exports.registerClass(Rescaling);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/image_resizing.js
+var INTERPOLATION_KEYS = ["bilinear", "nearest"];
+var INTERPOLATION_METHODS = new Set(INTERPOLATION_KEYS);
+var Resizing = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.height = args.height;
+    this.width = args.width;
+    if (args.interpolation) {
+      if (INTERPOLATION_METHODS.has(args.interpolation)) {
+        this.interpolation = args.interpolation;
+      } else {
+        throw new ValueError(`Invalid interpolation parameter: ${args.interpolation} is not implemented`);
+      }
+    } else {
+      this.interpolation = "bilinear";
+    }
+    this.cropToAspectRatio = Boolean(args.cropToAspectRatio);
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    const numChannels = inputShape[2];
+    return [this.height, this.width, numChannels];
+  }
+  getConfig() {
+    const config = {
+      "height": this.height,
+      "width": this.width,
+      "interpolation": this.interpolation,
+      "cropToAspectRatio": this.cropToAspectRatio
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      const size = [this.height, this.width];
+      if (this.interpolation === "bilinear") {
+        return image.resizeBilinear(inputs, size, !this.cropToAspectRatio);
+      } else if (this.interpolation === "nearest") {
+        return image.resizeNearestNeighbor(inputs, size, !this.cropToAspectRatio);
+      } else {
+        throw new Error(`Interpolation is ${this.interpolation} but only ${[...INTERPOLATION_METHODS]} are supported`);
+      }
+    });
+  }
+};
+Resizing.className = "Resizing";
+serialization_exports.registerClass(Resizing);
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/preprocessing_utils.js
+function encodeCategoricalInputs(inputs, outputMode, depth, weights) {
+  let input2 = getExactlyOneTensor(inputs);
+  if (input2.dtype !== "int32") {
+    input2 = cast2(input2, "int32");
+  }
+  if (outputMode === "int") {
+    return input2;
+  }
+  const originalShape = input2.shape;
+  if (input2.rank === 0) {
+    input2 = expandDims(input2, -1);
+  }
+  if (outputMode === "oneHot") {
+    if (input2.shape[input2.shape.length - 1] !== 1) {
+      input2 = expandDims(input2, -1);
+    }
+  }
+  if (input2.rank > 2) {
+    throw new ValueError(`When outputMode is not int, maximum output rank is 2 Received outputMode ${outputMode} and input shape ${originalShape} which would result in output rank ${input2.rank}.`);
+  }
+  const binaryOutput = ["multiHot", "oneHot"].includes(outputMode);
+  const denseBincountInput = input2;
+  let binCounts;
+  if (typeof weights !== "undefined" && outputMode === "count") {
+    binCounts = denseBincount(denseBincountInput, weights, depth, binaryOutput);
+  } else {
+    binCounts = denseBincount(denseBincountInput, [], depth, binaryOutput);
+  }
+  if (outputMode !== "tfIdf") {
+    return binCounts;
+  }
+  if (weights) {
+    return mul(binCounts, weights);
+  } else {
+    throw new ValueError(`When outputMode is 'tfIdf', weights must be provided.`);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/layers/preprocessing/category_encoding.js
+var CategoryEncoding = class extends Layer {
+  constructor(args) {
+    super(args);
+    this.numTokens = args.numTokens;
+    if (args.outputMode) {
+      this.outputMode = args.outputMode;
+    } else {
+      this.outputMode = "multiHot";
+    }
+  }
+  getConfig() {
+    const config = {
+      "numTokens": this.numTokens,
+      "outputMode": this.outputMode
+    };
+    const baseConfig = super.getConfig();
+    Object.assign(config, baseConfig);
+    return config;
+  }
+  computeOutputShape(inputShape) {
+    inputShape = getExactlyOneShape(inputShape);
+    if (inputShape == null) {
+      return [this.numTokens];
+    }
+    if (this.outputMode === "oneHot" && inputShape[inputShape.length - 1] !== 1) {
+      inputShape.push(this.numTokens);
+      return inputShape;
+    }
+    inputShape[inputShape.length - 1] = this.numTokens;
+    return inputShape;
+  }
+  call(inputs, kwargs) {
+    return tidy(() => {
+      inputs = getExactlyOneTensor(inputs);
+      if (inputs.dtype !== "int32") {
+        inputs = cast2(inputs, "int32");
+      }
+      let countWeights;
+      if (typeof kwargs["countWeights"] !== "undefined") {
+        if (this.outputMode !== "count") {
+          throw new ValueError(`countWeights is not used when outputMode !== count.
+              Received countWeights=${kwargs["countWeights"]}`);
+        }
+        countWeights = getExactlyOneTensor(kwargs["countWeights"]);
+      }
+      const maxValue = max(inputs);
+      const minValue = min(inputs);
+      const greaterEqualMax = greater(this.numTokens, maxValue).bufferSync().get(0);
+      const greaterMin = greaterEqual(minValue, 0).bufferSync().get(0);
+      if (!(greaterEqualMax && greaterMin)) {
+        throw new ValueError(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);
+      }
+      return encodeCategoricalInputs(inputs, this.outputMode, this.numTokens, countWeights);
+    });
+  }
+};
+CategoryEncoding.className = "CategoryEncoding";
+serialization_exports.registerClass(CategoryEncoding);
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_layers.js
+function inputLayer(args) {
+  return new InputLayer(args);
+}
+function elu3(args) {
+  return new ELU(args);
+}
+function reLU(args) {
+  return new ReLU(args);
+}
+function leakyReLU(args) {
+  return new LeakyReLU(args);
+}
+function prelu2(args) {
+  return new PReLU(args);
+}
+function softmax2(args) {
+  return new Softmax3(args);
+}
+function thresholdedReLU(args) {
+  return new ThresholdedReLU(args);
+}
+function conv1d2(args) {
+  return new Conv1D(args);
+}
+function conv2d3(args) {
+  return new Conv2D2(args);
+}
+function conv2dTranspose2(args) {
+  return new Conv2DTranspose(args);
+}
+function conv3d2(args) {
+  return new Conv3D2(args);
+}
+function conv3dTranspose2(args) {
+  return new Conv3DTranspose(args);
+}
+function separableConv2d2(args) {
+  return new SeparableConv2D(args);
+}
+function cropping2D(args) {
+  return new Cropping2D(args);
+}
+function upSampling2d(args) {
+  return new UpSampling2D(args);
+}
+function depthwiseConv2d4(args) {
+  return new DepthwiseConv2D(args);
+}
+function activation(args) {
+  return new Activation2(args);
+}
+function dense(args) {
+  return new Dense(args);
+}
+function dropout3(args) {
+  return new Dropout(args);
+}
+function spatialDropout1d(args) {
+  return new SpatialDropout1D(args);
+}
+function flatten3(args) {
+  return new Flatten(args);
+}
+function repeatVector(args) {
+  return new RepeatVector(args);
+}
+function reshape2(args) {
+  return new Reshape2(args);
+}
+function permute(args) {
+  return new Permute(args);
+}
+function embedding(args) {
+  return new Embedding(args);
+}
+function add3(args) {
+  return new Add2(args);
+}
+function average(args) {
+  return new Average(args);
+}
+function concatenate2(args) {
+  return new Concatenate(args);
+}
+function maximum2(args) {
+  return new Maximum2(args);
+}
+function minimum2(args) {
+  return new Minimum2(args);
+}
+function multiply(args) {
+  return new Multiply2(args);
+}
+function dot3(args) {
+  return new Dot(args);
+}
+function batchNormalization2(args) {
+  return new BatchNormalization(args);
+}
+function layerNormalization(args) {
+  return new LayerNormalization(args);
+}
+function zeroPadding2d(args) {
+  return new ZeroPadding2D(args);
+}
+function averagePooling1d(args) {
+  return new AveragePooling1D(args);
+}
+function avgPool1d(args) {
+  return averagePooling1d(args);
+}
+function avgPooling1d(args) {
+  return averagePooling1d(args);
+}
+function averagePooling2d(args) {
+  return new AveragePooling2D(args);
+}
+function avgPool2d(args) {
+  return averagePooling2d(args);
+}
+function avgPooling2d(args) {
+  return averagePooling2d(args);
+}
+function averagePooling3d(args) {
+  return new AveragePooling3D(args);
+}
+function avgPool3d2(args) {
+  return averagePooling3d(args);
+}
+function avgPooling3d(args) {
+  return averagePooling3d(args);
+}
+function globalAveragePooling1d(args) {
+  return new GlobalAveragePooling1D(args);
+}
+function globalAveragePooling2d(args) {
+  return new GlobalAveragePooling2D(args);
+}
+function globalMaxPooling1d(args) {
+  return new GlobalMaxPooling1D(args);
+}
+function globalMaxPooling2d(args) {
+  return new GlobalMaxPooling2D(args);
+}
+function maxPooling1d(args) {
+  return new MaxPooling1D(args);
+}
+function maxPooling2d(args) {
+  return new MaxPooling2D(args);
+}
+function maxPooling3d(args) {
+  return new MaxPooling3D(args);
+}
+function gru(args) {
+  return new GRU(args);
+}
+function gruCell(args) {
+  return new GRUCell(args);
+}
+function lstm(args) {
+  return new LSTM(args);
+}
+function lstmCell(args) {
+  return new LSTMCell(args);
+}
+function simpleRNN(args) {
+  return new SimpleRNN(args);
+}
+function simpleRNNCell(args) {
+  return new SimpleRNNCell(args);
+}
+function convLstm2d(args) {
+  return new ConvLSTM2D(args);
+}
+function convLstm2dCell(args) {
+  return new ConvLSTM2DCell(args);
+}
+function rnn2(args) {
+  return new RNN(args);
+}
+function stackedRNNCells(args) {
+  return new StackedRNNCells(args);
+}
+function bidirectional(args) {
+  return new Bidirectional(args);
+}
+function timeDistributed(args) {
+  return new TimeDistributed(args);
+}
+var globalMaxPool1d = globalMaxPooling1d;
+var globalMaxPool2d = globalMaxPooling2d;
+var maxPool1d = maxPooling1d;
+var maxPool2d = maxPooling2d;
+function gaussianNoise(args) {
+  return new GaussianNoise(args);
+}
+function gaussianDropout(args) {
+  return new GaussianDropout(args);
+}
+function alphaDropout(args) {
+  return new AlphaDropout(args);
+}
+function masking(args) {
+  return new Masking(args);
+}
+function rescaling(args) {
+  return new Rescaling(args);
+}
+function resizing(args) {
+  return new Resizing(args);
+}
+function categoryEncoding(args) {
+  return new CategoryEncoding(args);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_metrics.js
+var exports_metrics_exports = {};
+__export(exports_metrics_exports, {
+  MAPE: () => MAPE2,
+  MSE: () => MSE2,
+  binaryAccuracy: () => binaryAccuracy2,
+  binaryCrossentropy: () => binaryCrossentropy3,
+  categoricalAccuracy: () => categoricalAccuracy2,
+  categoricalCrossentropy: () => categoricalCrossentropy3,
+  cosineProximity: () => cosineProximity2,
+  mape: () => mape2,
+  meanAbsoluteError: () => meanAbsoluteError2,
+  meanAbsolutePercentageError: () => meanAbsolutePercentageError2,
+  meanSquaredError: () => meanSquaredError3,
+  mse: () => mse2,
+  precision: () => precision2,
+  recall: () => recall2,
+  sparseCategoricalAccuracy: () => sparseCategoricalAccuracy2
+});
+function binaryAccuracy2(yTrue, yPred) {
+  return binaryAccuracy(yTrue, yPred);
+}
+function binaryCrossentropy3(yTrue, yPred) {
+  return binaryCrossentropy2(yTrue, yPred);
+}
+function sparseCategoricalAccuracy2(yTrue, yPred) {
+  return sparseCategoricalAccuracy(yTrue, yPred);
+}
+function categoricalAccuracy2(yTrue, yPred) {
+  return categoricalAccuracy(yTrue, yPred);
+}
+function categoricalCrossentropy3(yTrue, yPred) {
+  return categoricalCrossentropy2(yTrue, yPred);
+}
+function precision2(yTrue, yPred) {
+  return precision(yTrue, yPred);
+}
+function recall2(yTrue, yPred) {
+  return recall(yTrue, yPred);
+}
+function cosineProximity2(yTrue, yPred) {
+  return cosineProximity(yTrue, yPred);
+}
+function meanAbsoluteError2(yTrue, yPred) {
+  return meanAbsoluteError(yTrue, yPred);
+}
+function meanAbsolutePercentageError2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function MAPE2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function mape2(yTrue, yPred) {
+  return meanAbsolutePercentageError(yTrue, yPred);
+}
+function meanSquaredError3(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+function MSE2(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+function mse2(yTrue, yPred) {
+  return meanSquaredError2(yTrue, yPred);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_models.js
+var exports_models_exports = {};
+__export(exports_models_exports, {
+  modelFromJSON: () => modelFromJSON
+});
+
+// node_modules/@tensorflow/tfjs-layers/dist/exports_regularizers.js
+var exports_regularizers_exports = {};
+__export(exports_regularizers_exports, {
+  l1: () => l12,
+  l1l2: () => l1l2,
+  l2: () => l22
+});
+function l1l2(config) {
+  return new L1L2(config);
+}
+function l12(config) {
+  return l1(config);
+}
+function l22(config) {
+  return l2(config);
+}
+
+// node_modules/@tensorflow/tfjs-layers/dist/callbacks.js
+var Callback = class extends BaseCallback {
+  constructor() {
+    super(...arguments);
+    this.model = null;
+  }
+  setModel(model2) {
+    if (!(model2 instanceof LayersModel)) {
+      throw new Error("model must be a LayersModel, not some other Container");
+    }
+    this.model = model2;
+  }
+};
+function less2(currVal, prevVal) {
+  return currVal < prevVal;
+}
+function greater2(currVal, prevVal) {
+  return currVal > prevVal;
+}
+var EarlyStopping = class extends Callback {
+  constructor(args) {
+    super();
+    if (args == null) {
+      args = {};
+    }
+    if (args.restoreBestWeights) {
+      throw new NotImplementedError("restoreBestWeights = True is not implemented in EarlyStopping yet.");
+    }
+    this.monitor = args.monitor || "val_loss";
+    this.minDelta = Math.abs(args.minDelta || 0);
+    this.patience = args.patience || 0;
+    this.verbose = args.verbose || 0;
+    this.mode = args.mode || "auto";
+    this.baseline = args.baseline;
+    if (["auto", "min", "max"].indexOf(this.mode) === -1) {
+      console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`);
+      this.mode = "auto";
+    }
+    if (this.mode === "min") {
+      this.monitorFunc = less2;
+    } else if (this.mode === "max") {
+      this.monitorFunc = greater2;
+    } else {
+      if (this.monitor.indexOf("acc") !== -1) {
+        this.monitorFunc = greater2;
+      } else {
+        this.monitorFunc = less2;
+      }
+    }
+    if (this.monitorFunc === less2) {
+      this.minDelta *= -1;
+    }
+  }
+  async onTrainBegin(logs) {
+    this.wait = 0;
+    this.stoppedEpoch = 0;
+    if (this.baseline != null) {
+      this.best = this.baseline;
+    } else {
+      this.best = this.monitorFunc === less2 ? Infinity : -Infinity;
+    }
+  }
+  async onEpochEnd(epoch, logs) {
+    await resolveScalarsInLogs(logs);
+    const current = this.getMonitorValue(logs);
+    if (current == null) {
+      return;
+    }
+    if (this.monitorFunc(current - this.minDelta, this.best)) {
+      this.best = current;
+      this.wait = 0;
+    } else {
+      this.wait++;
+      if (this.wait >= this.patience) {
+        this.stoppedEpoch = epoch;
+        this.model.stopTraining = true;
+      }
+    }
+  }
+  async onTrainEnd(logs) {
+    if (this.stoppedEpoch > 0 && this.verbose) {
+      console.log(`Epoch ${this.stoppedEpoch}: early stopping.`);
+    }
+  }
+  getMonitorValue(logs) {
+    if (logs == null) {
+      logs = {};
+    }
+    const monitorValue = logs[this.monitor];
+    if (monitorValue == null) {
+      console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(logs)}`);
+    }
+    return monitorValue;
+  }
+};
+function earlyStopping(args) {
+  return new EarlyStopping(args);
+}
+var callbacks = { earlyStopping };
+
+// node_modules/@tensorflow/tfjs-converter/dist/flags.js
+var ENV4 = env();
+ENV4.registerFlag("KEEP_INTERMEDIATE_TENSORS", () => false, (debugValue) => {
+  if (debugValue) {
+    console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.");
+  }
+});
+
+// node_modules/@tensorflow/tfjs-converter/dist/data/compiled_api.js
+var DataType;
+(function(DataType2) {
+  DataType2[DataType2["DT_INVALID"] = 0] = "DT_INVALID";
+  DataType2[DataType2["DT_FLOAT"] = 1] = "DT_FLOAT";
+  DataType2[DataType2["DT_DOUBLE"] = 2] = "DT_DOUBLE";
+  DataType2[DataType2["DT_INT32"] = 3] = "DT_INT32";
+  DataType2[DataType2["DT_UINT8"] = 4] = "DT_UINT8";
+  DataType2[DataType2["DT_INT16"] = 5] = "DT_INT16";
+  DataType2[DataType2["DT_INT8"] = 6] = "DT_INT8";
+  DataType2[DataType2["DT_STRING"] = 7] = "DT_STRING";
+  DataType2[DataType2["DT_COMPLEX64"] = 8] = "DT_COMPLEX64";
+  DataType2[DataType2["DT_INT64"] = 9] = "DT_INT64";
+  DataType2[DataType2["DT_BOOL"] = 10] = "DT_BOOL";
+  DataType2[DataType2["DT_QINT8"] = 11] = "DT_QINT8";
+  DataType2[DataType2["DT_QUINT8"] = 12] = "DT_QUINT8";
+  DataType2[DataType2["DT_QINT32"] = 13] = "DT_QINT32";
+  DataType2[DataType2["DT_BFLOAT16"] = 14] = "DT_BFLOAT16";
+  DataType2[DataType2["DT_QINT16"] = 15] = "DT_QINT16";
+  DataType2[DataType2["DT_QUINT16"] = 16] = "DT_QUINT16";
+  DataType2[DataType2["DT_UINT16"] = 17] = "DT_UINT16";
+  DataType2[DataType2["DT_COMPLEX128"] = 18] = "DT_COMPLEX128";
+  DataType2[DataType2["DT_HALF"] = 19] = "DT_HALF";
+  DataType2[DataType2["DT_RESOURCE"] = 20] = "DT_RESOURCE";
+  DataType2[DataType2["DT_VARIANT"] = 21] = "DT_VARIANT";
+  DataType2[DataType2["DT_UINT32"] = 22] = "DT_UINT32";
+  DataType2[DataType2["DT_UINT64"] = 23] = "DT_UINT64";
+  DataType2[DataType2["DT_FLOAT_REF"] = 101] = "DT_FLOAT_REF";
+  DataType2[DataType2["DT_DOUBLE_REF"] = 102] = "DT_DOUBLE_REF";
+  DataType2[DataType2["DT_INT32_REF"] = 103] = "DT_INT32_REF";
+  DataType2[DataType2["DT_UINT8_REF"] = 104] = "DT_UINT8_REF";
+  DataType2[DataType2["DT_INT16_REF"] = 105] = "DT_INT16_REF";
+  DataType2[DataType2["DT_INT8_REF"] = 106] = "DT_INT8_REF";
+  DataType2[DataType2["DT_STRING_REF"] = 107] = "DT_STRING_REF";
+  DataType2[DataType2["DT_COMPLEX64_REF"] = 108] = "DT_COMPLEX64_REF";
+  DataType2[DataType2["DT_INT64_REF"] = 109] = "DT_INT64_REF";
+  DataType2[DataType2["DT_BOOL_REF"] = 110] = "DT_BOOL_REF";
+  DataType2[DataType2["DT_QINT8_REF"] = 111] = "DT_QINT8_REF";
+  DataType2[DataType2["DT_QUINT8_REF"] = 112] = "DT_QUINT8_REF";
+  DataType2[DataType2["DT_QINT32_REF"] = 113] = "DT_QINT32_REF";
+  DataType2[DataType2["DT_BFLOAT16_REF"] = 114] = "DT_BFLOAT16_REF";
+  DataType2[DataType2["DT_QINT16_REF"] = 115] = "DT_QINT16_REF";
+  DataType2[DataType2["DT_QUINT16_REF"] = 116] = "DT_QUINT16_REF";
+  DataType2[DataType2["DT_UINT16_REF"] = 117] = "DT_UINT16_REF";
+  DataType2[DataType2["DT_COMPLEX128_REF"] = 118] = "DT_COMPLEX128_REF";
+  DataType2[DataType2["DT_HALF_REF"] = 119] = "DT_HALF_REF";
+  DataType2[DataType2["DT_RESOURCE_REF"] = 120] = "DT_RESOURCE_REF";
+  DataType2[DataType2["DT_VARIANT_REF"] = 121] = "DT_VARIANT_REF";
+  DataType2[DataType2["DT_UINT32_REF"] = 122] = "DT_UINT32_REF";
+  DataType2[DataType2["DT_UINT64_REF"] = 123] = "DT_UINT64_REF";
+})(DataType || (DataType = {}));
+var SaverDef;
+(function(SaverDef2) {
+  let CheckpointFormatVersion;
+  (function(CheckpointFormatVersion2) {
+    CheckpointFormatVersion2[CheckpointFormatVersion2["LEGACY"] = 0] = "LEGACY";
+    CheckpointFormatVersion2[CheckpointFormatVersion2["V1"] = 1] = "V1";
+    CheckpointFormatVersion2[CheckpointFormatVersion2["V2"] = 2] = "V2";
+  })(CheckpointFormatVersion = SaverDef2.CheckpointFormatVersion || (SaverDef2.CheckpointFormatVersion = {}));
+})(SaverDef || (SaverDef = {}));
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/custom_op/register.js
+var CUSTOM_OPS = {};
+function registerOp(name, opFunc) {
+  const opMapper = {
+    tfOpName: name,
+    category: "custom",
+    inputs: [],
+    attrs: [],
+    customExecutor: opFunc
+  };
+  CUSTOM_OPS[name] = opMapper;
+}
+function getRegisteredOp(name) {
+  return CUSTOM_OPS[name];
+}
+function deregisterOp(name) {
+  delete CUSTOM_OPS[name];
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/utils.js
+function getParamValue(paramName, node, tensorMap, context, resourceManager) {
+  const inputParam = node.inputParams[paramName];
+  if (inputParam && inputParam.inputIndexStart !== void 0) {
+    const start = inputParam.inputIndexStart;
+    const end = inputParam.inputIndexEnd === 0 ? void 0 : inputParam.inputIndexEnd === void 0 ? start + 1 : inputParam.inputIndexEnd;
+    if (inputParam.type === "tensor") {
+      return getTensor(node.inputNames[inputParam.inputIndexStart], tensorMap, context, resourceManager);
+    }
+    if (inputParam.type === "tensors") {
+      const inputs = node.inputNames.slice(start, end);
+      return inputs.map((name) => getTensor(name, tensorMap, context, resourceManager));
+    }
+    const tensor2 = getTensor(node.inputNames.slice(start)[0], tensorMap, context, resourceManager);
+    const data = tensor2.dataSync();
+    return inputParam.type === "number" ? data[0] : util_exports.toNestedArray(tensor2.shape, data);
+  }
+  const attrParam = node.attrParams[paramName];
+  return attrParam && attrParam.value;
+}
+function getTensor(name, tensorsMap, context, resourceManager) {
+  const [nodeName, index] = parseNodeName(name);
+  if (resourceManager != null) {
+    const tensor2 = resourceManager.getHashTableHandleByName(nodeName);
+    if (tensor2 != null) {
+      return tensor2;
+    }
+  }
+  const contextId = context.currentContextIds.find((contextId2) => {
+    return !!tensorsMap[getNodeNameWithContextId(nodeName, contextId2)];
+  });
+  return contextId !== void 0 ? tensorsMap[getNodeNameWithContextId(nodeName, contextId)][index] : void 0;
+}
+function getTensorsForCurrentContenxt(name, tensorsMap, context) {
+  return tensorsMap[getNodeNameWithContextId(name, context.currentContextId)];
+}
+function getNodeNameAndIndex(inputName, context) {
+  const [nodeName, index, outputName] = parseNodeName(inputName);
+  return [
+    getNodeNameWithContextId(nodeName, context && context.currentContextId),
+    index,
+    outputName
+  ];
+}
+function getNodeNameWithContextId(name, contextId) {
+  return !!contextId ? `${name}-${contextId}` : name;
+}
+function parseNodeName(name) {
+  const parts = name.split(":");
+  if (parts.length === 1) {
+    return [name, 0, void 0];
+  }
+  const nodeName = parts[0];
+  const outputName = parts.length === 3 ? parts[1] : void 0;
+  const index = Number(parts[parts.length - 1]);
+  return [nodeName, index, outputName];
+}
+function getPadding(node, tensorMap, context) {
+  let pad3 = getParamValue("pad", node, tensorMap, context);
+  if (pad3 === "explicit") {
+    pad3 = getParamValue("explicitPaddings", node, tensorMap, context);
+    const explicitPadding = [[0, 0], [0, 0], [0, 0], [0, 0]];
+    for (let i = 0; i < 4; i++) {
+      explicitPadding[i][0] = pad3[i * 2];
+      explicitPadding[i][1] = pad3[i * 2 + 1];
+    }
+    return explicitPadding;
+  }
+  return pad3;
+}
+function cloneTensor(tensor2) {
+  return tensor2.kept ? tensor2 : clone(tensor2);
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/arithmetic.js
+var arithmetic_exports = {};
+__export(arithmetic_exports, {
+  json: () => json
+});
+var json = [
+  {
+    "tfOpName": "Add",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AddV2",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AddN",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BiasAdd",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sub",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "RealDiv",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Div",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "DivNoNan",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FloorDiv",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mul",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Maximum",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Minimum",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pow",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SquaredDifference",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mod",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FloorMod",
+    "category": "arithmetic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/basic_math.js
+var basic_math_exports = {};
+__export(basic_math_exports, {
+  json: () => json2
+});
+var json2 = [
+  {
+    "tfOpName": "Abs",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Acos",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Asin",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atan2",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "y",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Ceil",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ClipByValue",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "clipValueMin",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "clipValueMax",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Complex",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "real",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "imag",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ComplexAbs",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cos",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cosh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Elu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Exp",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Floor",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Log",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Imag",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "outputType",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Neg",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Real",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "outputType",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prelu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "alpha",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Relu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Relu6",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Selu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sigmoid",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sin",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sinh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sqrt",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Rsqrt",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Square",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tanh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sign",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Round",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Expm1",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Log1p",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reciprocal",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Softplus",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Asinh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Acosh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Atanh",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Erf",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prod",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axes",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool",
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LeakyRelu",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "alpha",
+        "name": "alpha",
+        "type": "number",
+        "defaultValue": 0.2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "IsNan",
+    "category": "basic_math",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/control.js
+var control_exports = {};
+__export(control_exports, {
+  json: () => json3
+});
+var json3 = [
+  {
+    "tfOpName": "EmptyTensorList",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 1,
+        "name": "maxNumElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LoopCond",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "pred",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Switch",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "pred",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Merge",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Enter",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "frame_name",
+        "name": "frameName",
+        "type": "string"
+      },
+      {
+        "tfName": "is_constant",
+        "name": "isConstant",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Exit",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "NextIteration",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "size",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dynamic_size",
+        "name": "dynamicSize",
+        "type": "bool"
+      },
+      {
+        "tfName": "clear_after_read",
+        "name": "clearAfterRead",
+        "type": "bool"
+      },
+      {
+        "tfName": "identical_element_shapes",
+        "name": "identicalElementShapes",
+        "type": "bool"
+      },
+      {
+        "tfName": "tensor_array_name",
+        "name": "name",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayWriteV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayReadV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayGatherV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayScatterV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayConcatV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "element_shape_except0",
+        "name": "elementShapeExcept0",
+        "type": "shape",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArraySplitV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "lengths",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArraySizeV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "flowIn",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorArrayCloseV3",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorArrayId",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StatelessIf",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "cond",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "then_branch",
+        "name": "thenBranch",
+        "type": "func"
+      },
+      {
+        "tfName": "else_branch",
+        "name": "elseBranch",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "If",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "cond",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "then_branch",
+        "name": "thenBranch",
+        "type": "func"
+      },
+      {
+        "tfName": "else_branch",
+        "name": "elseBranch",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StatelessWhile",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "cond",
+        "name": "cond",
+        "type": "func"
+      },
+      {
+        "tfName": "body",
+        "name": "body",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "While",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "cond",
+        "name": "cond",
+        "type": "func"
+      },
+      {
+        "tfName": "body",
+        "name": "body",
+        "type": "func"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListScatter",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListScatterV2",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 3,
+        "name": "numElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListGather",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListGetItem",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListSetItem",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "index",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListReserve",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 1,
+        "name": "numElements",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListFromTensor",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListStack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "num_elements",
+        "name": "numElements",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListSplit",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "start": 2,
+        "name": "lengths",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListConcat",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListConcatV2",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_shape",
+        "name": "elementShape",
+        "type": "shape"
+      },
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListPopBack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "elementShape",
+        "type": "shape"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListPushBack",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "element_dtype",
+        "name": "elementDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListLength",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TensorListResize",
+    "category": "control",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensorListId",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/convolution.js
+var convolution_exports = {};
+__export(convolution_exports, {
+  json: () => json4
+});
+var json4 = [
+  {
+    "tfOpName": "AvgPool",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPool",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": [],
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPoolWithArgmax",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "include_batch_in_index",
+        "name": "includeBatchInIndex",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "AvgPool3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MaxPool3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "ksize",
+        "name": "kernelSize",
+        "type": "number[]"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv1D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "stride",
+        "name": "stride",
+        "type": "number"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NWC"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "dilation",
+        "name": "dilation",
+        "type": "number",
+        "defaultValue": 1
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "useCudnnOnGpu",
+        "name": "useCudnnOnGpu",
+        "type": "bool"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "_FusedConv2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "use_cudnn_on_gpu",
+        "name": "useCudnnOnGpu",
+        "type": "bool",
+        "defaultValue": true
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "defaultValue": [
+          1,
+          1,
+          1,
+          1
+        ]
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-4
+      },
+      {
+        "tfName": "leakyrelu_alpha",
+        "name": "leakyreluAlpha",
+        "type": "number",
+        "defaultValue": 0.2
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv2DBackpropInput",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 2,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 0,
+        "name": "outputShape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthwiseConv2d",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthwiseConv2dNative",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedDepthwiseConv2dNative",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]",
+        "defaultValue": [
+          1,
+          1,
+          1,
+          1
+        ]
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "explicit_paddings",
+        "name": "explicitPaddings",
+        "type": "number[]",
+        "defaultValue": []
+      }
+    ]
+  },
+  {
+    "tfOpName": "Conv3D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "defaultValue": "NHWC"
+      },
+      {
+        "tfName": "dilations",
+        "name": "dilations",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Dilation2D",
+    "category": "convolution",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "filter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "strides",
+        "name": "strides",
+        "type": "number[]"
+      },
+      {
+        "tfName": "rates",
+        "name": "dilations",
+        "type": "number[]"
+      },
+      {
+        "tfName": "padding",
+        "name": "pad",
+        "type": "string"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/creation.js
+var creation_exports = {};
+__export(creation_exports, {
+  json: () => json5
+});
+var json5 = [
+  {
+    "tfOpName": "Fill",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      },
+      {
+        "start": 1,
+        "name": "value",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LinSpace",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "start",
+        "type": "number"
+      },
+      {
+        "start": 1,
+        "name": "stop",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "num",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "OneHot",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "depth",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "onValue",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "start": 3,
+        "name": "offValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Ones",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "OnesLike",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "RandomStandardNormal",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "RandomUniform",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "minval",
+        "name": "minval",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "maxval",
+        "name": "maxval",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Range",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "start",
+        "type": "number"
+      },
+      {
+        "start": 1,
+        "name": "stop",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "step",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tidx",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TruncatedNormal",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "means",
+        "name": "mean",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "stddev",
+        "name": "stdDev",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number"
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "T",
+        "name": "T",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Zeros",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ZerosLike",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Multinomial",
+    "category": "creation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "logits",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "numSamples",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "seed",
+        "name": "seed",
+        "type": "number"
+      },
+      {
+        "tfName": "seed2",
+        "name": "seed2",
+        "type": "number"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      },
+      {
+        "tfName": "output_dtype",
+        "name": "output_dtype",
+        "type": "dtype"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/dynamic.js
+var dynamic_exports = {};
+__export(dynamic_exports, {
+  json: () => json6
+});
+var json6 = [
+  {
+    "tfOpName": "NonMaxSuppressionV2",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV3",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV4",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "T_threshold",
+        "name": "threshold",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "pad_to_max_output_size",
+        "name": "padToMaxOutputSize",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "NonMaxSuppressionV5",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scores",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "maxOutputSize",
+        "type": "number"
+      },
+      {
+        "start": 3,
+        "name": "iouThreshold",
+        "type": "number"
+      },
+      {
+        "start": 4,
+        "name": "scoreThreshold",
+        "type": "number"
+      },
+      {
+        "start": 5,
+        "name": "softNmsSigma",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Where",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ListDiff",
+    "category": "dynamic",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "y",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/evaluation.js
+var evaluation_exports = {};
+__export(evaluation_exports, {
+  json: () => json7
+});
+var json7 = [
+  {
+    "tfOpName": "LowerBound",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sortedSequence",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "TopKV2",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "k",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "sorted",
+        "name": "sorted",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "UpperBound",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sortedSequence",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Unique",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "UniqueV2",
+    "category": "evaluation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/graph.js
+var graph_exports = {};
+__export(graph_exports, {
+  json: () => json8
+});
+var json8 = [
+  {
+    "tfOpName": "PlaceholderWithDefault",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "default",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "shape",
+        "name": "shape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Placeholder",
+    "category": "graph",
+    "attrs": [
+      {
+        "tfName": "shape",
+        "name": "shape",
+        "type": "shape"
+      },
+      {
+        "tfName": "dtype",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Const",
+    "category": "graph"
+  },
+  {
+    "tfOpName": "Identity",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "IdentityN",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "x",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Snapshot",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Rank",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Size",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Shape",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ShapeN",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "x",
+        "type": "tensors"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Print",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "data",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "message",
+        "name": "message",
+        "type": "string"
+      },
+      {
+        "tfName": "first_n",
+        "name": "firstN",
+        "type": "number",
+        "notSupported": true
+      },
+      {
+        "tfName": "summarize",
+        "name": "summarize",
+        "type": "number",
+        "defaultValue": 3
+      }
+    ]
+  },
+  {
+    "tfOpName": "NoOp",
+    "category": "graph",
+    "inputs": []
+  },
+  {
+    "tfOpName": "StopGradient",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "FakeQuantWithMinMaxVars",
+    "category": "graph",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "min",
+        "name": "min",
+        "type": "number"
+      },
+      {
+        "tfName": "max",
+        "name": "max",
+        "type": "number"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/hash_table.js
+var hash_table_exports = {};
+__export(hash_table_exports, {
+  json: () => json9
+});
+var json9 = [
+  {
+    "tfOpName": "HashTable",
+    "category": "hash_table",
+    "inputs": [],
+    "attrs": [
+      {
+        "tfName": "shared_name",
+        "name": "sharedName",
+        "type": "string"
+      },
+      {
+        "tfName": "use_node_name_sharing",
+        "name": "useNodeNameSharing",
+        "type": "bool"
+      },
+      {
+        "tfName": "key_dtype",
+        "name": "keyDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "value_dtype",
+        "name": "valueDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "HashTableV2",
+    "category": "hash_table",
+    "inputs": [],
+    "attrs": [
+      {
+        "tfName": "shared_name",
+        "name": "sharedName",
+        "type": "string"
+      },
+      {
+        "tfName": "use_node_name_sharing",
+        "name": "useNodeNameSharing",
+        "type": "bool"
+      },
+      {
+        "tfName": "key_dtype",
+        "name": "keyDType",
+        "type": "dtype"
+      },
+      {
+        "tfName": "value_dtype",
+        "name": "valueDType",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableImport",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "values",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableImportV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "values",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableFind",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableFindV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "keys",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "Tin",
+        "name": "tIn",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "Tout",
+        "name": "tOut",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableSize",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LookupTableSizeV2",
+    "category": "hash_table",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tableHandle",
+        "type": "tensor"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/image.js
+var image_exports = {};
+__export(image_exports, {
+  json: () => json10
+});
+var json10 = [
+  {
+    "tfOpName": "ResizeBilinear",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "align_corners",
+        "name": "alignCorners",
+        "type": "bool"
+      },
+      {
+        "tfName": "half_pixel_centers",
+        "name": "halfPixelCenters",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "ResizeNearestNeighbor",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "align_corners",
+        "name": "alignCorners",
+        "type": "bool"
+      },
+      {
+        "tfName": "half_pixel_centers",
+        "name": "halfPixelCenters",
+        "type": "bool"
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "CropAndResize",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "image",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "boxes",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "boxInd",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "cropSize",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "method",
+        "name": "method",
+        "type": "string"
+      },
+      {
+        "tfName": "extrapolation_value",
+        "name": "extrapolationValue",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ImageProjectiveTransformV3",
+    "category": "image",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "images",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "transforms",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "fillValue",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "interpolation",
+        "name": "interpolation",
+        "type": "string"
+      },
+      {
+        "tfName": "fill_mode",
+        "name": "fillMode",
+        "type": "string"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/logical.js
+var logical_exports = {};
+__export(logical_exports, {
+  json: () => json11
+});
+var json11 = [
+  {
+    "tfOpName": "Equal",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "NotEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Greater",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "GreaterEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Less",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LessEqual",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalAnd",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalNot",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogicalOr",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Select",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SelectV2",
+    "category": "logical",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "condition",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/matrices.js
+var matrices_exports = {};
+__export(matrices_exports, {
+  json: () => json12
+});
+var json12 = [
+  {
+    "tfOpName": "_FusedMatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "end": 0,
+        "name": "args",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_args",
+        "name": "numArgs",
+        "type": "number"
+      },
+      {
+        "tfName": "fused_ops",
+        "name": "fusedOps",
+        "type": "string[]",
+        "defaultValue": []
+      },
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-4
+      },
+      {
+        "tfName": "transpose_a",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "transpose_b",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "leakyrelu_alpha",
+        "name": "leakyreluAlpha",
+        "type": "number",
+        "defaultValue": 0.2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "MatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "transpose_a",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "transpose_b",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchMatMul",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "adj_x",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "adj_y",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchMatMulV2",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "a",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "b",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "adj_x",
+        "name": "transposeA",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "adj_y",
+        "name": "transposeB",
+        "type": "bool",
+        "defaultValue": false
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Transpose",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "perm",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Einsum",
+    "category": "matrices",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "equation",
+        "name": "equation",
+        "type": "string"
+      },
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      },
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/normalization.js
+var normalization_exports = {};
+__export(normalization_exports, {
+  json: () => json13
+});
+var json13 = [
+  {
+    "tfOpName": "EuclideanNorm",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool",
+        "defaultValue": false
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNorm",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNormV2",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "FusedBatchNormV3",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "scale",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "offset",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "mean",
+        "type": "tensor"
+      },
+      {
+        "start": 4,
+        "name": "variance",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "epsilon",
+        "name": "epsilon",
+        "type": "number",
+        "defaultValue": 1e-3
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "LRN",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "depth_radius",
+        "name": "radius",
+        "type": "number",
+        "defaultValue": 5
+      },
+      {
+        "tfName": "bias",
+        "name": "bias",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "alpha",
+        "name": "alpha",
+        "type": "number",
+        "defaultValue": 1
+      },
+      {
+        "tfName": "beta",
+        "name": "beta",
+        "type": "number",
+        "defaultValue": 0.5
+      }
+    ]
+  },
+  {
+    "tfOpName": "Softmax",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "LogSoftmax",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseToDense",
+    "category": "normalization",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sparseIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "sparseValues",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "defaultValue": true,
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/reduction.js
+var reduction_exports = {};
+__export(reduction_exports, {
+  json: () => json14
+});
+var json14 = [
+  {
+    "tfOpName": "Bincount",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "weights",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DenseBincount",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "size",
+        "type": "number"
+      },
+      {
+        "start": 2,
+        "name": "weights",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "binary_output",
+        "name": "binaryOutput",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Max",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Mean",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Min",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Sum",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "All",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Any",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ArgMax",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ArgMin",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Prod",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "keep_dims",
+        "name": "keepDims",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cumprod",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "exclusive",
+        "name": "exclusive",
+        "type": "bool"
+      },
+      {
+        "tfName": "reverse",
+        "name": "reverse",
+        "type": "bool"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Cumsum",
+    "category": "reduction",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "exclusive",
+        "name": "exclusive",
+        "type": "bool"
+      },
+      {
+        "tfName": "reverse",
+        "name": "reverse",
+        "type": "bool"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/slice_join.js
+var slice_join_exports = {};
+__export(slice_join_exports, {
+  json: () => json15
+});
+var json15 = [
+  {
+    "tfOpName": "ConcatV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "end": -1,
+        "name": "tensors",
+        "type": "tensors"
+      },
+      {
+        "start": -1,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      }
+    ]
+  },
+  {
+    "tfOpName": "Concat",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 1,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      },
+      {
+        "start": 0,
+        "name": "axis",
+        "type": "number"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "N",
+        "name": "n",
+        "type": "number",
+        "defaultValue": 2
+      }
+    ]
+  },
+  {
+    "tfOpName": "GatherV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "batch_dims",
+        "name": "batchDims",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Gather",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reverse",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "dims",
+        "type": "bool[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ReverseV2",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Slice",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "begin",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "size",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "StridedSlice",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "begin",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "end",
+        "type": "number[]"
+      },
+      {
+        "start": 3,
+        "name": "strides",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "begin_mask",
+        "name": "beginMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "end_mask",
+        "name": "endMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "new_axis_mask",
+        "name": "newAxisMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "ellipsis_mask",
+        "name": "ellipsisMask",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "shrink_axis_mask",
+        "name": "shrinkAxisMask",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pack",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "end": 0,
+        "name": "tensors",
+        "type": "tensors"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Unpack",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "tensor",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "tfName": "num",
+        "name": "num",
+        "type": "number",
+        "defaultValue": 0,
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "Tile",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "reps",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Split",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      },
+      {
+        "start": 1,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_split",
+        "name": "numOrSizeSplits",
+        "type": "number",
+        "defaultValue": 1
+      }
+    ]
+  },
+  {
+    "tfOpName": "SplitV",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "numOrSizeSplits",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "axis",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "ScatterNd",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "GatherNd",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseToDense",
+    "category": "slice_join",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "sparseIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "outputShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "sparseValues",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "validate_indices",
+        "name": "validateIndices",
+        "type": "bool",
+        "defaultValue": false,
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/sparse.js
+var sparse_exports = {};
+__export(sparse_exports, {
+  json: () => json16
+});
+var json16 = [
+  {
+    "tfOpName": "SparseFillEmptyRows",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "values",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "denseShape",
+        "type": "tensor"
+      },
+      {
+        "start": 3,
+        "name": "defaultValue",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseReshape",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "inputIndices",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "inputShape",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "newShape",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "T",
+        "name": "dtype",
+        "type": "dtype",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseSegmentMean",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "segmentIds",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SparseSegmentSum",
+    "category": "sparse",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "indices",
+        "type": "tensor"
+      },
+      {
+        "start": 2,
+        "name": "segmentIds",
+        "type": "tensor"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/spectral.js
+var spectral_exports = {};
+__export(spectral_exports, {
+  json: () => json17
+});
+var json17 = [
+  {
+    "tfOpName": "FFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "IFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ]
+  },
+  {
+    "tfOpName": "RFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "fft_length",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  },
+  {
+    "tfOpName": "IRFFT",
+    "category": "spectral",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "fft_length",
+        "type": "number",
+        "notSupported": true
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/string.js
+var string_exports = {};
+__export(string_exports, {
+  json: () => json18
+});
+var json18 = [
+  {
+    "tfOpName": "StringNGrams",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "data",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "dataSplits",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "separator",
+        "name": "separator",
+        "type": "string"
+      },
+      {
+        "tfName": "ngram_widths",
+        "name": "nGramWidths",
+        "type": "number[]"
+      },
+      {
+        "tfName": "left_pad",
+        "name": "leftPad",
+        "type": "string"
+      },
+      {
+        "tfName": "right_pad",
+        "name": "rightPad",
+        "type": "string"
+      },
+      {
+        "tfName": "pad_width",
+        "name": "padWidth",
+        "type": "number"
+      },
+      {
+        "tfName": "preserve_short_sequences",
+        "name": "preserveShortSequences",
+        "type": "bool"
+      }
+    ],
+    "outputs": [
+      "ngrams",
+      "ngrams_splits"
+    ]
+  },
+  {
+    "tfOpName": "StringSplit",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "delimiter",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "skip_empty",
+        "name": "skipEmpty",
+        "type": "bool"
+      }
+    ],
+    "outputs": [
+      "indices",
+      "values",
+      "shape"
+    ]
+  },
+  {
+    "tfOpName": "StringToHashBucketFast",
+    "category": "string",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "input",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "num_buckets",
+        "name": "numBuckets",
+        "type": "number"
+      }
+    ]
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/op_list/transformation.js
+var transformation_exports = {};
+__export(transformation_exports, {
+  json: () => json19
+});
+var json19 = [
+  {
+    "tfOpName": "Cast",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "SrcT",
+        "name": "sdtype",
+        "type": "dtype",
+        "notSupported": true
+      },
+      {
+        "tfName": "DstT",
+        "name": "dtype",
+        "type": "dtype"
+      }
+    ]
+  },
+  {
+    "tfOpName": "ExpandDims",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "axis",
+        "type": "number"
+      }
+    ]
+  },
+  {
+    "tfOpName": "MirrorPad",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "mode",
+        "name": "mode",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Pad",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "constant_value",
+        "name": "constantValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "PadV2",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "padding",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "constantValue",
+        "type": "number",
+        "defaultValue": 0
+      }
+    ]
+  },
+  {
+    "tfOpName": "Reshape",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "Squeeze",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "axis",
+        "tfDeprecatedName": "squeeze_dims",
+        "name": "axis",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "SpaceToBatchND",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "blockShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "paddings",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BatchToSpaceND",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "blockShape",
+        "type": "number[]"
+      },
+      {
+        "start": 2,
+        "name": "crops",
+        "type": "number[]"
+      }
+    ]
+  },
+  {
+    "tfOpName": "DepthToSpace",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      }
+    ],
+    "attrs": [
+      {
+        "tfName": "block_size",
+        "name": "blockSize",
+        "type": "number"
+      },
+      {
+        "tfName": "data_format",
+        "name": "dataFormat",
+        "type": "string"
+      }
+    ]
+  },
+  {
+    "tfOpName": "BroadcastTo",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "x",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "shape",
+        "type": "number[]"
+      }
+    ],
+    "attrs": []
+  },
+  {
+    "tfOpName": "BroadcastArgs",
+    "category": "transformation",
+    "inputs": [
+      {
+        "start": 0,
+        "name": "s0",
+        "type": "tensor"
+      },
+      {
+        "start": 1,
+        "name": "s1",
+        "type": "tensor"
+      }
+    ],
+    "attrs": []
+  }
+];
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/operation_mapper.js
+var OperationMapper = class {
+  constructor() {
+    const ops = [
+      arithmetic_exports,
+      basic_math_exports,
+      control_exports,
+      convolution_exports,
+      creation_exports,
+      dynamic_exports,
+      evaluation_exports,
+      graph_exports,
+      hash_table_exports,
+      image_exports,
+      logical_exports,
+      matrices_exports,
+      normalization_exports,
+      reduction_exports,
+      slice_join_exports,
+      sparse_exports,
+      spectral_exports,
+      string_exports,
+      transformation_exports
+    ];
+    const mappersJson = [].concat(...ops.map((op2) => op2.json));
+    this.opMappers = mappersJson.reduce((map, mapper) => {
+      map[mapper.tfOpName] = mapper;
+      return map;
+    }, {});
+  }
+  static get Instance() {
+    return this._instance || (this._instance = new this());
+  }
+  transformGraph(graph, signature = {}) {
+    const tfNodes = graph.node;
+    const placeholders = [];
+    const weights = [];
+    const initNodes = [];
+    const nodes = tfNodes.reduce((map, node) => {
+      map[node.name] = this.mapNode(node);
+      if (node.op.startsWith("Placeholder")) {
+        placeholders.push(map[node.name]);
+      } else if (node.op === "Const") {
+        weights.push(map[node.name]);
+      } else if (node.input == null || node.input.length === 0) {
+        initNodes.push(map[node.name]);
+      }
+      return map;
+    }, {});
+    let inputs = [];
+    const outputs = [];
+    let inputNodeNameToKey = {};
+    let outputNodeNameToKey = {};
+    if (signature != null) {
+      inputNodeNameToKey = this.mapSignatureEntries(signature.inputs);
+      outputNodeNameToKey = this.mapSignatureEntries(signature.outputs);
+    }
+    const allNodes = Object.keys(nodes);
+    allNodes.forEach((key) => {
+      const node = nodes[key];
+      node.inputNames.forEach((name, index) => {
+        const [nodeName, , outputName] = getNodeNameAndIndex(name);
+        const inputNode = nodes[nodeName];
+        if (inputNode.outputs != null) {
+          const outputIndex = inputNode.outputs.indexOf(outputName);
+          if (outputIndex !== -1) {
+            const inputName = `${nodeName}:${outputIndex}`;
+            node.inputNames[index] = inputName;
+          }
+        }
+        node.inputs.push(inputNode);
+        inputNode.children.push(node);
+      });
+    });
+    if (Object.keys(outputNodeNameToKey).length === 0) {
+      allNodes.forEach((key) => {
+        const node = nodes[key];
+        if (node.children.length === 0) {
+          outputs.push(node);
+        }
+      });
+    } else {
+      Object.keys(outputNodeNameToKey).forEach((name) => {
+        const [nodeName] = getNodeNameAndIndex(name);
+        const node = nodes[nodeName];
+        if (node != null) {
+          node.signatureKey = outputNodeNameToKey[name];
+          outputs.push(node);
+        }
+      });
+    }
+    if (Object.keys(inputNodeNameToKey).length > 0) {
+      Object.keys(inputNodeNameToKey).forEach((name) => {
+        const [nodeName] = getNodeNameAndIndex(name);
+        const node = nodes[nodeName];
+        if (node) {
+          node.signatureKey = inputNodeNameToKey[name];
+          inputs.push(node);
+        }
+      });
+    } else {
+      inputs = placeholders;
+    }
+    let functions = {};
+    if (graph.library != null && graph.library.function != null) {
+      functions = graph.library.function.reduce((functions2, func2) => {
+        functions2[func2.signature.name] = this.mapFunction(func2);
+        return functions2;
+      }, {});
+    }
+    const result = { nodes, inputs, outputs, weights, placeholders, signature, functions };
+    if (initNodes.length > 0) {
+      result.initNodes = initNodes;
+    }
+    return result;
+  }
+  mapSignatureEntries(entries) {
+    return Object.keys(entries || {}).reduce((prev, curr) => {
+      prev[entries[curr].name] = curr;
+      return prev;
+    }, {});
+  }
+  mapNode(node) {
+    const mapper = getRegisteredOp(node.op) || this.opMappers[node.op] || {};
+    if (node.attr == null) {
+      node.attr = {};
+    }
+    const newNode = {
+      name: node.name,
+      op: node.op,
+      category: mapper.category,
+      inputNames: (node.input || []).map((input2) => input2.startsWith("^") ? input2.slice(1) : input2),
+      inputs: [],
+      children: [],
+      inputParams: {},
+      attrParams: {},
+      rawAttrs: node.attr,
+      outputs: mapper.outputs
+    };
+    if (mapper.inputs != null) {
+      newNode.inputParams = mapper.inputs.reduce((map, param) => {
+        map[param.name] = {
+          type: param.type,
+          inputIndexStart: param.start,
+          inputIndexEnd: param.end
+        };
+        return map;
+      }, {});
+    }
+    if (mapper.attrs != null) {
+      newNode.attrParams = mapper.attrs.reduce((map, param) => {
+        const type = param.type;
+        let value = void 0;
+        switch (param.type) {
+          case "string":
+            value = getStringParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getStringParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "string[]":
+            value = getStringArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getStringArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "number":
+            value = getNumberParam(node.attr, param.tfName, param.defaultValue || 0);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getNumberParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "number[]":
+            value = getNumericArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getNumericArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "bool":
+            value = getBoolParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getBoolParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "bool[]":
+            value = getBoolArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getBoolArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "shape":
+            value = getTensorShapeParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getTensorShapeParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "shape[]":
+            value = getTensorShapeArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getTensorShapeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "dtype":
+            value = getDtypeParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getDtypeParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "dtype[]":
+            value = getDtypeArrayParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getDtypeArrayParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "func":
+            value = getFuncParam(node.attr, param.tfName, param.defaultValue);
+            if (value === void 0 && !!param.tfDeprecatedName) {
+              value = getFuncParam(node.attr, param.tfDeprecatedName, param.defaultValue);
+            }
+            break;
+          case "tensor":
+          case "tensors":
+            break;
+          default:
+            throw new Error(`Unsupported param type: ${param.type} for op: ${node.op}`);
+        }
+        map[param.name] = { value, type };
+        return map;
+      }, {});
+    }
+    return newNode;
+  }
+  mapFunction(functionDef) {
+    const tfNodes = functionDef.nodeDef;
+    const placeholders = [];
+    const weights = [];
+    let nodes = {};
+    if (tfNodes != null) {
+      nodes = tfNodes.reduce((map, node) => {
+        map[node.name] = this.mapNode(node);
+        if (node.op === "Const") {
+          weights.push(map[node.name]);
+        }
+        return map;
+      }, {});
+    }
+    const inputs = [];
+    const outputs = [];
+    functionDef.signature.inputArg.forEach((arg) => {
+      const [nodeName] = getNodeNameAndIndex(arg.name);
+      const node = {
+        name: nodeName,
+        op: "Placeholder",
+        inputs: [],
+        inputNames: [],
+        category: "graph",
+        inputParams: {},
+        attrParams: { dtype: { value: parseDtypeParam(arg.type), type: "dtype" } },
+        children: []
+      };
+      node.signatureKey = arg.name;
+      inputs.push(node);
+      nodes[nodeName] = node;
+    });
+    const allNodes = Object.keys(nodes);
+    allNodes.forEach((key) => {
+      const node = nodes[key];
+      node.inputNames.forEach((name, index) => {
+        const [nodeName, , outputName] = getNodeNameAndIndex(name);
+        const inputNode = nodes[nodeName];
+        if (inputNode.outputs != null) {
+          const outputIndex = inputNode.outputs.indexOf(outputName);
+          if (outputIndex !== -1) {
+            const inputName = `${nodeName}:${outputIndex}`;
+            node.inputNames[index] = inputName;
+          }
+        }
+        node.inputs.push(inputNode);
+        inputNode.children.push(node);
+      });
+    });
+    const returnNodeMap = functionDef.ret;
+    functionDef.signature.outputArg.forEach((output) => {
+      const [nodeName, index] = getNodeNameAndIndex(returnNodeMap[output.name]);
+      const node = nodes[nodeName];
+      if (node != null) {
+        node.defaultOutput = index;
+        outputs.push(node);
+      }
+    });
+    const signature = this.mapArgsToSignature(functionDef);
+    return { nodes, inputs, outputs, weights, placeholders, signature };
+  }
+  mapArgsToSignature(functionDef) {
+    return {
+      methodName: functionDef.signature.name,
+      inputs: functionDef.signature.inputArg.reduce((map, arg) => {
+        map[arg.name] = this.mapArgToTensorInfo(arg);
+        return map;
+      }, {}),
+      outputs: functionDef.signature.outputArg.reduce((map, arg) => {
+        map[arg.name] = this.mapArgToTensorInfo(arg, functionDef.ret);
+        return map;
+      }, {})
+    };
+  }
+  mapArgToTensorInfo(arg, nameMap2) {
+    let name = arg.name;
+    if (nameMap2 != null) {
+      name = nameMap2[name];
+    }
+    return { name, dtype: arg.type };
+  }
+};
+function decodeBase64(text) {
+  const global2 = env().global;
+  if (typeof global2.atob !== "undefined") {
+    return global2.atob(text);
+  } else if (typeof Buffer !== "undefined") {
+    return new Buffer(text, "base64").toString();
+  } else {
+    throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()");
+  }
+}
+function parseStringParam(s, keepCase) {
+  const value = Array.isArray(s) ? String.fromCharCode.apply(null, s) : decodeBase64(s);
+  return keepCase ? value : value.toLowerCase();
+}
+function getStringParam(attrs, name, def, keepCase = false) {
+  const param = attrs[name];
+  if (param != null) {
+    return parseStringParam(param.s, keepCase);
+  }
+  return def;
+}
+function getBoolParam(attrs, name, def) {
+  const param = attrs[name];
+  return param ? param.b : def;
+}
+function getNumberParam(attrs, name, def) {
+  const param = attrs[name] || {};
+  const value = param["i"] != null ? param["i"] : param["f"] != null ? param["f"] : def;
+  return typeof value === "number" ? value : parseInt(value, 10);
+}
+function parseDtypeParam(value) {
+  if (typeof value === "string") {
+    value = DataType[value];
+  }
+  switch (value) {
+    case DataType.DT_FLOAT:
+    case DataType.DT_HALF:
+      return "float32";
+    case DataType.DT_INT32:
+    case DataType.DT_INT64:
+    case DataType.DT_INT8:
+    case DataType.DT_UINT8:
+      return "int32";
+    case DataType.DT_BOOL:
+      return "bool";
+    case DataType.DT_DOUBLE:
+      return "float32";
+    case DataType.DT_STRING:
+      return "string";
+    default:
+      return null;
+  }
+}
+function getFuncParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.func) {
+    return param.func.name;
+  }
+  return def;
+}
+function getDtypeParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.type) {
+    return parseDtypeParam(param.type);
+  }
+  return def;
+}
+function getDtypeArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.type) {
+    return param.list.type.map((v) => parseDtypeParam(v));
+  }
+  return def;
+}
+function parseTensorShapeParam(shape) {
+  if (shape.unknownRank) {
+    return void 0;
+  }
+  if (shape.dim != null) {
+    return shape.dim.map((dim) => typeof dim.size === "number" ? dim.size : parseInt(dim.size, 10));
+  }
+  return [];
+}
+function getTensorShapeParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.shape) {
+    return parseTensorShapeParam(param.shape);
+  }
+  return def;
+}
+function getNumericArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param) {
+    return ((param.list.f && param.list.f.length ? param.list.f : param.list.i) || []).map((v) => typeof v === "number" ? v : parseInt(v, 10));
+  }
+  return def;
+}
+function getStringArrayParam(attrs, name, def, keepCase = false) {
+  const param = attrs[name];
+  if (param && param.list && param.list.s) {
+    return param.list.s.map((v) => {
+      return parseStringParam(v, keepCase);
+    });
+  }
+  return def;
+}
+function getTensorShapeArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.shape) {
+    return param.list.shape.map((v) => {
+      return parseTensorShapeParam(v);
+    });
+  }
+  return def;
+}
+function getBoolArrayParam(attrs, name, def) {
+  const param = attrs[name];
+  if (param && param.list && param.list.b) {
+    return param.list.b;
+  }
+  return def;
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/custom_op/node_value_impl.js
+var NodeValueImpl = class {
+  constructor(node, tensorMap, context) {
+    this.node = node;
+    this.tensorMap = tensorMap;
+    this.context = context;
+    this.inputs = [];
+    this.attrs = {};
+    this.inputs = node.inputNames.map((name) => this.getInput(name));
+    if (node.rawAttrs != null) {
+      this.attrs = Object.keys(node.rawAttrs).reduce((attrs, key) => {
+        attrs[key] = this.getAttr(key);
+        return attrs;
+      }, {});
+    }
+  }
+  getInput(name) {
+    return getTensor(name, this.tensorMap, this.context);
+  }
+  getAttr(name, defaultValue) {
+    const value = this.node.rawAttrs[name];
+    if (value.tensor != null) {
+      return getTensor(name, this.tensorMap, this.context);
+    }
+    if (value.i != null || value.f != null) {
+      return getNumberParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.s != null) {
+      return getStringParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.b != null) {
+      return getBoolParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.shape != null) {
+      return getTensorShapeParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.type != null) {
+      return getDtypeParam(this.node.rawAttrs, name, defaultValue);
+    }
+    if (value.list != null) {
+      if (value.list.i != null || value.list.f != null) {
+        return getNumericArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.s != null) {
+        return getStringArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.shape != null) {
+        return getTensorShapeArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.b != null) {
+        return getBoolArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+      if (value.list.type != null) {
+        return getDtypeArrayParam(this.node.rawAttrs, name, defaultValue);
+      }
+    }
+    return defaultValue;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-core/dist/ops/ops_for_converter.js
+var ops_for_converter_exports = {};
+__export(ops_for_converter_exports, {
+  OP_SCOPE_SUFFIX: () => OP_SCOPE_SUFFIX,
+  abs: () => abs,
+  acos: () => acos,
+  acosh: () => acosh,
+  add: () => add2,
+  addN: () => addN,
+  all: () => all,
+  any: () => any,
+  argMax: () => argMax,
+  argMin: () => argMin,
+  asin: () => asin,
+  asinh: () => asinh,
+  atan: () => atan,
+  atan2: () => atan2,
+  atanh: () => atanh,
+  avgPool: () => avgPool,
+  avgPool3d: () => avgPool3d,
+  basicLSTMCell: () => basicLSTMCell,
+  batchNorm: () => batchNorm,
+  batchNorm2d: () => batchNorm2d,
+  batchNorm3d: () => batchNorm3d,
+  batchNorm4d: () => batchNorm4d,
+  batchToSpaceND: () => batchToSpaceND,
+  bincount: () => bincount,
+  booleanMaskAsync: () => booleanMaskAsync,
+  broadcastArgs: () => broadcastArgs,
+  broadcastTo: () => broadcastTo,
+  buffer: () => buffer,
+  cast: () => cast,
+  ceil: () => ceil,
+  clipByValue: () => clipByValue,
+  clone: () => clone,
+  complex: () => complex,
+  concat: () => concat,
+  concat1d: () => concat1d,
+  concat2d: () => concat2d,
+  concat3d: () => concat3d,
+  concat4d: () => concat4d,
+  conv1d: () => conv1d,
+  conv2d: () => conv2d,
+  conv2dTranspose: () => conv2dTranspose,
+  conv3d: () => conv3d,
+  conv3dTranspose: () => conv3dTranspose,
+  cos: () => cos,
+  cosh: () => cosh,
+  cosineWindow: () => cosineWindow,
+  cumprod: () => cumprod,
+  cumsum: () => cumsum,
+  denseBincount: () => denseBincount,
+  depthToSpace: () => depthToSpace,
+  depthwiseConv2d: () => depthwiseConv2d,
+  diag: () => diag,
+  dilation2d: () => dilation2d,
+  div: () => div,
+  divNoNan: () => divNoNan,
+  dot: () => dot,
+  dropout: () => dropout,
+  einsum: () => einsum,
+  elu: () => elu,
+  enclosingPowerOfTwo: () => enclosingPowerOfTwo,
+  equal: () => equal,
+  erf: () => erf,
+  euclideanNorm: () => euclideanNorm,
+  exp: () => exp,
+  expandDims: () => expandDims,
+  expm1: () => expm1,
+  eye: () => eye,
+  fft: () => fft,
+  fill: () => fill,
+  floor: () => floor,
+  floorDiv: () => floorDiv,
+  fused: () => fused_ops_exports,
+  gather: () => gather,
+  gatherND: () => gatherND,
+  greater: () => greater,
+  greaterEqual: () => greaterEqual,
+  ifft: () => ifft,
+  imag: () => imag,
+  image: () => image,
+  inTopKAsync: () => inTopKAsync,
+  irfft: () => irfft,
+  isFinite: () => isFinite2,
+  isInf: () => isInf,
+  isNaN: () => isNaN2,
+  leakyRelu: () => leakyRelu,
+  less: () => less,
+  lessEqual: () => lessEqual,
+  linalg: () => linalg,
+  linspace: () => linspace,
+  localResponseNormalization: () => localResponseNormalization,
+  log: () => log2,
+  log1p: () => log1p,
+  logSigmoid: () => logSigmoid,
+  logSoftmax: () => logSoftmax,
+  logSumExp: () => logSumExp,
+  logicalAnd: () => logicalAnd,
+  logicalNot: () => logicalNot,
+  logicalOr: () => logicalOr,
+  logicalXor: () => logicalXor,
+  losses: () => losses,
+  lowerBound: () => lowerBound,
+  matMul: () => matMul,
+  max: () => max,
+  maxPool: () => maxPool,
+  maxPool3d: () => maxPool3d,
+  maxPoolWithArgmax: () => maxPoolWithArgmax,
+  maximum: () => maximum,
+  mean: () => mean,
+  meshgrid: () => meshgrid,
+  min: () => min,
+  minimum: () => minimum,
+  mirrorPad: () => mirrorPad,
+  mod: () => mod,
+  moments: () => moments,
+  movingAverage: () => movingAverage,
+  mul: () => mul,
+  multiRNNCell: () => multiRNNCell,
+  multinomial: () => multinomial,
+  neg: () => neg,
+  norm: () => norm,
+  notEqual: () => notEqual,
+  oneHot: () => oneHot,
+  ones: () => ones2,
+  onesLike: () => onesLike,
+  op: () => op,
+  outerProduct: () => outerProduct,
+  pad: () => pad,
+  pad1d: () => pad1d,
+  pad2d: () => pad2d,
+  pad3d: () => pad3d,
+  pad4d: () => pad4d,
+  pool: () => pool,
+  pow: () => pow,
+  prelu: () => prelu,
+  print: () => print,
+  prod: () => prod,
+  raggedGather: () => raggedGather,
+  raggedRange: () => raggedRange,
+  raggedTensorToTensor: () => raggedTensorToTensor,
+  rand: () => rand,
+  randomGamma: () => randomGamma,
+  randomNormal: () => randomNormal,
+  randomStandardNormal: () => randomStandardNormal,
+  randomUniform: () => randomUniform,
+  range: () => range,
+  real: () => real,
+  reciprocal: () => reciprocal,
+  relu: () => relu,
+  relu6: () => relu6,
+  reshape: () => reshape,
+  reverse: () => reverse,
+  reverse1d: () => reverse1d,
+  reverse2d: () => reverse2d,
+  reverse3d: () => reverse3d,
+  reverse4d: () => reverse4d,
+  rfft: () => rfft,
+  round: () => round2,
+  rsqrt: () => rsqrt,
+  scalar: () => scalar,
+  scatterND: () => scatterND,
+  searchSorted: () => searchSorted,
+  selu: () => selu,
+  separableConv2d: () => separableConv2d,
+  setdiff1dAsync: () => setdiff1dAsync,
+  sigmoid: () => sigmoid,
+  sign: () => sign,
+  signal: () => signal,
+  sin: () => sin,
+  sinh: () => sinh,
+  slice: () => slice,
+  slice1d: () => slice1d,
+  slice2d: () => slice2d,
+  slice3d: () => slice3d,
+  slice4d: () => slice4d,
+  softmax: () => softmax,
+  softplus: () => softplus,
+  spaceToBatchND: () => spaceToBatchND,
+  sparse: () => sparse,
+  sparseToDense: () => sparseToDense,
+  spectral: () => spectral,
+  split: () => split,
+  sqrt: () => sqrt,
+  square: () => square,
+  squaredDifference: () => squaredDifference,
+  squeeze: () => squeeze,
+  stack: () => stack,
+  step: () => step,
+  stridedSlice: () => stridedSlice,
+  string: () => string,
+  sub: () => sub,
+  sum: () => sum2,
+  tan: () => tan,
+  tanh: () => tanh2,
+  tensor: () => tensor,
+  tensor1d: () => tensor1d,
+  tensor2d: () => tensor2d,
+  tensor3d: () => tensor3d,
+  tensor4d: () => tensor4d,
+  tensor5d: () => tensor5d,
+  tensor6d: () => tensor6d,
+  tile: () => tile,
+  topk: () => topk,
+  transpose: () => transpose,
+  truncatedNormal: () => truncatedNormal,
+  unique: () => unique,
+  unsortedSegmentSum: () => unsortedSegmentSum,
+  unstack: () => unstack,
+  upperBound: () => upperBound,
+  variable: () => variable,
+  where: () => where,
+  whereAsync: () => whereAsync,
+  zeros: () => zeros,
+  zerosLike: () => zerosLike
+});
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/arithmetic_executor.js
+var executeOp = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "BiasAdd":
+    case "AddV2":
+    case "Add": {
+      return [ops.add(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "AddN": {
+      return [ops.addN(getParamValue("tensors", node, tensorMap, context))];
+    }
+    case "FloorMod":
+    case "Mod":
+      return [ops.mod(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    case "Mul":
+      return [ops.mul(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    case "RealDiv":
+    case "Div": {
+      return [ops.div(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "DivNoNan": {
+      return [ops.divNoNan(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "FloorDiv": {
+      return [ops.floorDiv(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Sub": {
+      return [ops.sub(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Minimum": {
+      return [ops.minimum(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Maximum": {
+      return [ops.maximum(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Pow": {
+      return [ops.pow(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "SquaredDifference": {
+      return [ops.squaredDifference(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/basic_math_executor.js
+var executeOp2 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Abs":
+    case "ComplexAbs":
+      return [ops.abs(getParamValue("x", node, tensorMap, context))];
+    case "Acos":
+      return [ops.acos(getParamValue("x", node, tensorMap, context))];
+    case "Acosh":
+      return [ops.acosh(getParamValue("x", node, tensorMap, context))];
+    case "Asin":
+      return [ops.asin(getParamValue("x", node, tensorMap, context))];
+    case "Asinh":
+      return [ops.asinh(getParamValue("x", node, tensorMap, context))];
+    case "Atan":
+      return [ops.atan(getParamValue("x", node, tensorMap, context))];
+    case "Atan2":
+      return [ops.atan2(getParamValue("x", node, tensorMap, context), getParamValue("y", node, tensorMap, context))];
+    case "Atanh":
+      return [ops.atanh(getParamValue("x", node, tensorMap, context))];
+    case "Ceil":
+      return [ops.ceil(getParamValue("x", node, tensorMap, context))];
+    case "Complex":
+      return [ops.complex(getParamValue("real", node, tensorMap, context), getParamValue("imag", node, tensorMap, context))];
+    case "Cos":
+      return [ops.cos(getParamValue("x", node, tensorMap, context))];
+    case "Cosh":
+      return [ops.cosh(getParamValue("x", node, tensorMap, context))];
+    case "Elu":
+      return [ops.elu(getParamValue("x", node, tensorMap, context))];
+    case "Erf":
+      return [ops.erf(getParamValue("x", node, tensorMap, context))];
+    case "Exp":
+      return [ops.exp(getParamValue("x", node, tensorMap, context))];
+    case "Expm1": {
+      return [ops.expm1(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Floor":
+      return [ops.floor(getParamValue("x", node, tensorMap, context))];
+    case "Log":
+      return [ops.log(getParamValue("x", node, tensorMap, context))];
+    case "Log1p": {
+      return [ops.log1p(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Imag":
+      return [ops.imag(getParamValue("x", node, tensorMap, context))];
+    case "Neg":
+      return [ops.neg(getParamValue("x", node, tensorMap, context))];
+    case "Reciprocal": {
+      return [ops.reciprocal(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Real":
+      return [ops.real(getParamValue("x", node, tensorMap, context))];
+    case "Relu":
+      return [ops.relu(getParamValue("x", node, tensorMap, context))];
+    case "Round": {
+      return [ops.round(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Selu":
+      return [ops.selu(getParamValue("x", node, tensorMap, context))];
+    case "Sigmoid":
+      return [ops.sigmoid(getParamValue("x", node, tensorMap, context))];
+    case "Sin":
+      return [ops.sin(getParamValue("x", node, tensorMap, context))];
+    case "Sign": {
+      return [ops.sign(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Sinh": {
+      return [ops.sinh(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Softplus": {
+      return [ops.softplus(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Sqrt": {
+      return [ops.sqrt(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Square": {
+      return [ops.square(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Tanh": {
+      return [ops.tanh(getParamValue("x", node, tensorMap, context))];
+    }
+    case "Tan":
+      return [ops.tan(getParamValue("x", node, tensorMap, context))];
+    case "ClipByValue":
+      return [ops.clipByValue(getParamValue("x", node, tensorMap, context), getParamValue("clipValueMin", node, tensorMap, context), getParamValue("clipValueMax", node, tensorMap, context))];
+    case "Relu6":
+      return [ops.relu6(getParamValue("x", node, tensorMap, context))];
+    case "Rsqrt":
+      return [ops.rsqrt(getTensor(node.inputNames[0], tensorMap, context))];
+    case "Prod":
+      return [ops.prod(getParamValue("x", node, tensorMap, context), getParamValue("axes", node, tensorMap, context))];
+    case "LeakyRelu":
+      return [ops.leakyRelu(getParamValue("x", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context))];
+    case "Prelu":
+      return [ops.prelu(getParamValue("x", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context))];
+    case "IsNan":
+      return [ops.isNaN(getTensor(node.inputNames[0], tensorMap, context))];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_utils.js
+function assertShapesMatchAllowUndefinedSize(shapeA, shapeB, errorMessagePrefix = "") {
+  if (typeof shapeA === "number" || typeof shapeB === "number") {
+    return;
+  }
+  util_exports.assert(shapeA.length === shapeB.length, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+  for (let i = 0; i < shapeA.length; i++) {
+    const dim0 = shapeA[i];
+    const dim1 = shapeB[i];
+    util_exports.assert(dim0 < 0 || dim1 < 0 || dim0 === dim1, () => errorMessagePrefix + ` Shapes ${shapeA} and ${shapeB} must match`);
+  }
+}
+function fullDefinedShape(elementShape) {
+  if (typeof elementShape === "number" || elementShape.some((dim) => dim < 0)) {
+    return false;
+  }
+  return true;
+}
+function inferElementShape(listElementShape, tensors, elementShape) {
+  let partialShape = mergeElementShape(listElementShape, elementShape);
+  const notfullDefinedShape = !fullDefinedShape(partialShape);
+  if (notfullDefinedShape && tensors.length === 0) {
+    throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${partialShape}`);
+  }
+  if (notfullDefinedShape) {
+    tensors.forEach((tensor2) => {
+      partialShape = mergeElementShape(tensor2.shape, partialShape);
+    });
+  }
+  if (!fullDefinedShape(partialShape)) {
+    throw new Error(`Non-fully-defined elementShape: ${partialShape}`);
+  }
+  return partialShape;
+}
+function mergeElementShape(elementShapeA, elementShapeB) {
+  if (typeof elementShapeA === "number") {
+    return elementShapeB;
+  }
+  if (typeof elementShapeB === "number") {
+    return elementShapeA;
+  }
+  if (elementShapeA.length !== elementShapeB.length) {
+    throw new Error(`Incompatible ranks during merge: ${elementShapeA} vs. ${elementShapeB}`);
+  }
+  const result = [];
+  for (let i = 0; i < elementShapeA.length; ++i) {
+    const dim0 = elementShapeA[i];
+    const dim1 = elementShapeB[i];
+    if (dim0 >= 0 && dim1 >= 0 && dim0 !== dim1) {
+      throw new Error(`Incompatible shape during merge: ${elementShapeA} vs. ${elementShapeB}`);
+    }
+    result[i] = dim0 >= 0 ? dim0 : dim1;
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_array.js
+var TensorArray = class {
+  constructor(name, dtype, maxSize, elementShape, identicalElementShapes, dynamicSize, clearAfterRead) {
+    this.name = name;
+    this.dtype = dtype;
+    this.maxSize = maxSize;
+    this.elementShape = elementShape;
+    this.identicalElementShapes = identicalElementShapes;
+    this.dynamicSize = dynamicSize;
+    this.clearAfterRead = clearAfterRead;
+    this.tensors = [];
+    this.closed_ = false;
+    this.idTensor = scalar(0);
+    keep(this.idTensor);
+  }
+  get id() {
+    return this.idTensor.id;
+  }
+  get closed() {
+    return this.closed_;
+  }
+  clearAndClose(keepIds) {
+    this.tensors.forEach((tensor2) => {
+      if (keepIds == null || !keepIds.has(tensor2.tensor.id)) {
+        tensor2.tensor.dispose();
+      }
+    });
+    this.tensors = [];
+    this.closed_ = true;
+    this.idTensor.dispose();
+  }
+  size() {
+    return this.tensors.length;
+  }
+  read(index) {
+    if (this.closed_) {
+      throw new Error(`TensorArray ${this.name} has already been closed.`);
+    }
+    if (index < 0 || index >= this.size()) {
+      throw new Error(`Tried to read from index ${index}, but array size is: ${this.size()}`);
+    }
+    const tensorWithState = this.tensors[index];
+    if (tensorWithState.cleared) {
+      throw new Error(`TensorArray ${this.name}: Could not read index ${index} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);
+    }
+    if (this.clearAfterRead) {
+      tensorWithState.cleared = true;
+    }
+    tensorWithState.read = true;
+    return tensorWithState.tensor;
+  }
+  readMany(indices) {
+    return indices.map((index) => this.read(index));
+  }
+  write(index, tensor2) {
+    if (this.closed_) {
+      throw new Error(`TensorArray ${this.name} has already been closed.`);
+    }
+    if (index < 0 || !this.dynamicSize && index >= this.maxSize) {
+      throw new Error(`Tried to write to index ${index}, but array is not resizeable and size is: ${this.maxSize}`);
+    }
+    const t = this.tensors[index] || {};
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index},
+          because the value dtype is ${tensor2.dtype}, but TensorArray dtype is ${this.dtype}.`);
+    }
+    if (this.size() === 0 && (this.elementShape == null || this.elementShape.length === 0)) {
+      this.elementShape = tensor2.shape;
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, `TensorArray ${this.name}: Could not write to TensorArray index ${index}.`);
+    if (t.read) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been read.`);
+    }
+    if (t.written) {
+      throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${index}, because it has already been written.`);
+    }
+    t.tensor = tensor2;
+    keep(tensor2);
+    t.written = true;
+    this.tensors[index] = t;
+  }
+  writeMany(indices, tensors) {
+    if (indices.length !== tensors.length) {
+      throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${indices.length} is not the same as tensors size: ${tensors.length}.`);
+    }
+    indices.forEach((i, index) => this.write(i, tensors[index]));
+  }
+  gather(indices, dtype) {
+    if (!!dtype && dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${dtype}`);
+    }
+    if (!indices) {
+      indices = [];
+      for (let i = 0; i < this.size(); i++) {
+        indices.push(i);
+      }
+    } else {
+      indices = indices.slice(0, this.size());
+    }
+    if (indices.length === 0) {
+      return tensor([], [0].concat(this.elementShape));
+    }
+    const tensors = this.readMany(indices);
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, "TensorArray shape mismatch: ");
+    return stack(tensors, 0);
+  }
+  concat(dtype) {
+    if (!!dtype && dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${dtype}`);
+    }
+    if (this.size() === 0) {
+      return tensor([], [0].concat(this.elementShape));
+    }
+    const indices = [];
+    for (let i = 0; i < this.size(); i++) {
+      indices.push(i);
+    }
+    const tensors = this.readMany(indices);
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensors[0].shape, `TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${tensors[0].shape})`);
+    return concat(tensors, 0);
+  }
+  scatter(indices, tensor2) {
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);
+    }
+    if (indices.length !== tensor2.shape[0]) {
+      throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);
+    }
+    const maxIndex = Math.max(...indices);
+    if (!this.dynamicSize && maxIndex >= this.maxSize) {
+      throw new Error(`Max index must be < array size (${maxIndex}  vs. ${this.maxSize})`);
+    }
+    this.writeMany(indices, unstack(tensor2, 0));
+  }
+  split(length, tensor2) {
+    if (tensor2.dtype !== this.dtype) {
+      throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${tensor2.dtype}`);
+    }
+    let totalLength = 0;
+    const cumulativeLengths = length.map((len) => {
+      totalLength += len;
+      return totalLength;
+    });
+    if (totalLength !== tensor2.shape[0]) {
+      throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${n}, and tensor's shape is: ${e.shape}`);if(!this.dynamicSize&&t.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${t.length}), and the TensorArray is not marked as dynamically resizeable`);let s=n===0?0:e.size/n,i=[];B(()=>{e=R(e,[1,n,s]);for(let u=0;u<t.length;++u){let l=u===0?0:o[u-1],c=[0,l,0],p=[1,t[u],s];i[u]=R(Rt(e,c,p),this.elementShape)}return i});let a=[];for(let u=0;u<t.length;u++)a[u]=u;this.writeMany(a,i)}};var ml=class{constructor(t,e,n,o=-1){this.tensors=t,this.elementShape=e,this.elementDtype=n,t!=null&&t.forEach(s=>{if(n!==s.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${s.dtype}`);Vn(e,s.shape,"TensorList shape mismatch: "),De(s)}),this.idTensor=mt(0),this.maxNumElements=o,De(this.idTensor)}get id(){return this.idTensor.id}copy(){return new ml([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(t){this.tensors.forEach(e=>{(t==null||!t.has(e.id))&&e.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(t,e,n=-1){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);if(n!==-1&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);Vn(t,this.elementShape,"TensorList shape mismatch: ");let o=Kf(this.elementShape,this.tensors,t);return B(()=>{let s=this.tensors.map(i=>R(i,o));return nr(s,0)})}popBack(t,e){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error("Trying to pop from an empty list.");let n=Kf(this.elementShape,this.tensors,t),o=this.tensors.pop();return o.kept=!1,Vn(o.shape,t,"TensorList shape mismatch: "),R(o,n)}pushBack(t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(Vn(t.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");De(t),this.tensors.push(t)}resize(t){if(t<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${t}`);if(this.maxNumElements!==-1&&t>this.maxNumElements)throw new Error(`TensorListResize input size ${t} is greater maxNumElement ${this.maxNumElements}.`);let e=new ml([],this.elementShape,this.elementDtype,this.maxNumElements);e.tensors.length=t;for(let n=0;n<Math.min(this.tensors.length,t);++n)e.tensors[n]=this.tensors[n];return e}getItem(t,e,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(t<0||t>this.tensors.length)throw new Error(`Trying to access element ${t} in a list with ${this.tensors.length} elements.`);if(this.tensors[t]==null)throw new Error(`element at index ${t} is null.`);Vn(this.tensors[t].shape,e,"TensorList shape mismatch: ");let o=Kf(this.elementShape,this.tensors,e);return R(this.tensors[t],o)}setItem(t,e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(t<0||this.maxNumElements!==-1&&t>=this.maxNumElements)throw new Error(`Trying to set element ${t} in a list with max ${this.maxNumElements} elements.`);Vn(this.elementShape,e.shape,"TensorList shape mismatch: "),De(e),this.tensors[t]!=null&&(this.tensors[t].kept=!1),this.tensors[t]=e}gather(t,e,n){if(e!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e}, but list elements ${this.elementDtype}`);Vn(this.elementShape,n,"TensorList shape mismatch: "),t=t.slice(0,this.size());let o=Kf(this.elementShape,this.tensors,n);return t.length===0?ur([],[0].concat(o)):B(()=>{let s=t.map(i=>R(this.tensors[i],o));return nr(s,0)})}concat(t,e){if(!!t&&t!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${t}`);Vn(this.elementShape,e,"TensorList shape mismatch: ");let n=Kf(this.elementShape,this.tensors,e);return this.size()===0?ur([],[0].concat(n)):B(()=>{let o=this.tensors.map(s=>R(s,n));return ne(o,0)})}};function VD(r,t,e){let n=r.dtype;if(r.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${r.shape}`);if(r.dtype!==e)throw new Error(`Invalid data types; op elements ${r.dtype}, but list elements ${e}`);let o=r.shape.slice(1);Vn(o,t,"TensorList shape mismatch: ");let s=vr(r);return new ml(s,t,n)}function GD(r,t,e,n){return new ml([],r,t,n)}function WD(r,t,e,n){if(t.length!==r.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${r.shape[0]}`);let o=Math.max(...t);if(n!=null&&n!==-1&&o>=n)throw new Error(`Max index must be < array size (${o}  vs. ${n})`);let s=new ml([],e,r.dtype,n),i=vr(r,0);return t.forEach((a,u)=>{s.setItem(a,i[u])}),s}function UD(r,t,e){let n=0,o=t.map(c=>(n+=c,n));if(n!==r.shape[0])throw new Error(`Expected sum of lengths to be equal to
+        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);
+    }
+    if (!this.dynamicSize && length.length !== this.maxSize) {
+      throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${length.length}), and the TensorArray is not marked as dynamically resizeable`);
+    }
+    const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;
+    const tensors = [];
+    tidy(() => {
+      tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);
+      for (let i = 0; i < length.length; ++i) {
+        const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];
+        const indices2 = [0, previousLength, 0];
+        const sizes = [1, length[i], elementPerRow];
+        tensors[i] = reshape(slice(tensor2, indices2, sizes), this.elementShape);
+      }
+      return tensors;
+    });
+    const indices = [];
+    for (let i = 0; i < length.length; i++) {
+      indices[i] = i;
+    }
+    this.writeMany(indices, tensors);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/tensor_list.js
+var TensorList = class {
+  constructor(tensors, elementShape, elementDtype, maxNumElements = -1) {
+    this.tensors = tensors;
+    this.elementShape = elementShape;
+    this.elementDtype = elementDtype;
+    if (tensors != null) {
+      tensors.forEach((tensor2) => {
+        if (elementDtype !== tensor2.dtype) {
+          throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${tensor2.dtype}`);
+        }
+        assertShapesMatchAllowUndefinedSize(elementShape, tensor2.shape, "TensorList shape mismatch: ");
+        keep(tensor2);
+      });
+    }
+    this.idTensor = scalar(0);
+    this.maxNumElements = maxNumElements;
+    keep(this.idTensor);
+  }
+  get id() {
+    return this.idTensor.id;
+  }
+  copy() {
+    return new TensorList([...this.tensors], this.elementShape, this.elementDtype);
+  }
+  clearAndClose(keepIds) {
+    this.tensors.forEach((tensor2) => {
+      if (keepIds == null || !keepIds.has(tensor2.id)) {
+        tensor2.dispose();
+      }
+    });
+    this.tensors.length = 0;
+    this.idTensor.dispose();
+  }
+  size() {
+    return this.tensors.length;
+  }
+  stack(elementShape, elementDtype, numElements = -1) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (numElements !== -1 && this.tensors.length !== numElements) {
+      throw new Error(`Operation expected a list with ${numElements} elements but got a list with ${this.tensors.length} elements.`);
+    }
+    assertShapesMatchAllowUndefinedSize(elementShape, this.elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    return tidy(() => {
+      const reshapedTensors = this.tensors.map((tensor2) => reshape(tensor2, outputElementShape));
+      return stack(reshapedTensors, 0);
+    });
+  }
+  popBack(elementShape, elementDtype) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (this.size() === 0) {
+      throw new Error("Trying to pop from an empty list.");
+    }
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    const tensor2 = this.tensors.pop();
+    tensor2.kept = false;
+    assertShapesMatchAllowUndefinedSize(tensor2.shape, elementShape, "TensorList shape mismatch: ");
+    return reshape(tensor2, outputElementShape);
+  }
+  pushBack(tensor2) {
+    if (tensor2.dtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(tensor2.shape, this.elementShape, "TensorList shape mismatch: ");
+    if (this.maxNumElements === this.size()) {
+      throw new Error(`Trying to push element into a full list.`);
+    }
+    keep(tensor2);
+    this.tensors.push(tensor2);
+  }
+  resize(size) {
+    if (size < 0) {
+      throw new Error(`TensorListResize expects size to be non-negative. Got: ${size}`);
+    }
+    if (this.maxNumElements !== -1 && size > this.maxNumElements) {
+      throw new Error(`TensorListResize input size ${size} is greater maxNumElement ${this.maxNumElements}.`);
+    }
+    const destTensorList = new TensorList([], this.elementShape, this.elementDtype, this.maxNumElements);
+    destTensorList.tensors.length = size;
+    for (let i = 0; i < Math.min(this.tensors.length, size); ++i) {
+      destTensorList.tensors[i] = this.tensors[i];
+    }
+    return destTensorList;
+  }
+  getItem(elementIndex, elementShape, elementDtype) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    if (elementIndex < 0 || elementIndex > this.tensors.length) {
+      throw new Error(`Trying to access element ${elementIndex} in a list with ${this.tensors.length} elements.`);
+    }
+    if (this.tensors[elementIndex] == null) {
+      throw new Error(`element at index ${elementIndex} is null.`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.tensors[elementIndex].shape, elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    return reshape(this.tensors[elementIndex], outputElementShape);
+  }
+  setItem(elementIndex, tensor2) {
+    if (tensor2.dtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${this.elementDtype}`);
+    }
+    if (elementIndex < 0 || this.maxNumElements !== -1 && elementIndex >= this.maxNumElements) {
+      throw new Error(`Trying to set element ${elementIndex} in a list with max ${this.maxNumElements} elements.`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, tensor2.shape, "TensorList shape mismatch: ");
+    keep(tensor2);
+    if (this.tensors[elementIndex] != null) {
+      this.tensors[elementIndex].kept = false;
+    }
+    this.tensors[elementIndex] = tensor2;
+  }
+  gather(indices, elementDtype, elementShape) {
+    if (elementDtype !== this.elementDtype) {
+      throw new Error(`Invalid data types; op elements ${elementDtype}, but list elements ${this.elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, "TensorList shape mismatch: ");
+    indices = indices.slice(0, this.size());
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    if (indices.length === 0) {
+      return tensor([], [0].concat(outputElementShape));
+    }
+    return tidy(() => {
+      const tensors = indices.map((i) => reshape(this.tensors[i], outputElementShape));
+      return stack(tensors, 0);
+    });
+  }
+  concat(elementDtype, elementShape) {
+    if (!!elementDtype && elementDtype !== this.elementDtype) {
+      throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${elementDtype}`);
+    }
+    assertShapesMatchAllowUndefinedSize(this.elementShape, elementShape, "TensorList shape mismatch: ");
+    const outputElementShape = inferElementShape(this.elementShape, this.tensors, elementShape);
+    if (this.size() === 0) {
+      return tensor([], [0].concat(outputElementShape));
+    }
+    return tidy(() => {
+      const tensors = this.tensors.map((t) => reshape(t, outputElementShape));
+      return concat(tensors, 0);
+    });
+  }
+};
+function fromTensor(tensor2, elementShape, elementDtype) {
+  const dtype = tensor2.dtype;
+  if (tensor2.shape.length < 1) {
+    throw new Error(`Tensor must be at least a vector, but saw shape: ${tensor2.shape}`);
+  }
+  if (tensor2.dtype !== elementDtype) {
+    throw new Error(`Invalid data types; op elements ${tensor2.dtype}, but list elements ${elementDtype}`);
+  }
+  const tensorElementShape = tensor2.shape.slice(1);
+  assertShapesMatchAllowUndefinedSize(tensorElementShape, elementShape, "TensorList shape mismatch: ");
+  const tensorList = unstack(tensor2);
+  return new TensorList(tensorList, elementShape, dtype);
+}
+function reserve(elementShape, elementDtype, numElements, maxNumElements) {
+  return new TensorList([], elementShape, elementDtype, maxNumElements);
+}
+function scatter(tensor2, indices, elementShape, numElements) {
+  if (indices.length !== tensor2.shape[0]) {
+    throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${indices.length} vs. ${tensor2.shape[0]}`);
+  }
+  const maxIndex = Math.max(...indices);
+  if (numElements != null && numElements !== -1 && maxIndex >= numElements) {
+    throw new Error(`Max index must be < array size (${maxIndex}  vs. ${numElements})`);
+  }
+  const list = new TensorList([], elementShape, tensor2.dtype, numElements);
+  const tensors = unstack(tensor2, 0);
+  indices.forEach((value, index) => {
+    list.setItem(value, tensors[index]);
+  });
+  return list;
+}
+function split2(tensor2, length, elementShape) {
+  let totalLength = 0;
+  const cumulativeLengths = length.map((len) => {
+    totalLength += len;
+    return totalLength;
+  });
+  if (totalLength !== tensor2.shape[0]) {
+    throw new Error(`Expected sum of lengths to be equal to
           tensor.shape[0], but sum of lengths is
-        ${n}, and tensor's shape is: ${r.shape}`);let s=r.shape.slice(1),i=Mb(s,e),a=n===0?0:r.size/n,u=B(()=>{let c=[];r=R(r,[1,n,a]);for(let p=0;p<t.length;++p){let m=p===0?0:o[p-1],f=[0,m,0],d=[1,t[p],a];c[p]=R(Rt(r,f,d),i)}return r.dispose(),c}),l=new ml([],e,r.dtype,t.length);for(let c=0;c<u.length;c++)l.setItem(c,u[c]);return l}var HD=async(r,t,e)=>{switch(r.op){case"If":case"StatelessIf":{let n=S("thenBranch",r,t,e),o=S("elseBranch",r,t,e),s=S("cond",r,t,e),i=S("args",r,t,e);return(await s.data())[0]?e.functionMap[n].executeFunctionAsync(i,e.tensorArrayMap,e.tensorListMap):e.functionMap[o].executeFunctionAsync(i,e.tensorArrayMap,e.tensorListMap)}case"While":case"StatelessWhile":{let n=S("body",r,t,e),o=S("cond",r,t,e),s=S("args",r,t,e),i=await e.functionMap[o].executeFunctionAsync(s,e.tensorArrayMap,e.tensorListMap),a=s.map(c=>c.id),u=await i[0].data();i.forEach(c=>{!c.kept&&a.indexOf(c.id)===-1&&c.dispose()});let l=s;for(;u[0];){let c=l;l=await e.functionMap[n].executeFunctionAsync(l,e.tensorArrayMap,e.tensorListMap);let p=l.map(f=>f.id);c.forEach(f=>{!f.kept&&a.indexOf(f.id)===-1&&p.indexOf(f.id)===-1&&f.dispose()});let m=await e.functionMap[o].executeFunctionAsync(l,e.tensorArrayMap,e.tensorListMap);u=await m[0].data(),m.forEach(f=>{!f.kept&&a.indexOf(f.id)===-1&&p.indexOf(f.id)===-1&&f.dispose()})}return l}case"LoopCond":{let n=S("pred",r,t,e);return[Zs(n)]}case"Switch":{let n=S("pred",r,t,e),o=S("data",r,t,e);return o.kept||(o=Zs(o)),(await n.data())[0]?[void 0,o]:[o,void 0]}case"Merge":{let n=r.inputNames.find(o=>br(o,t,e)!==void 0);if(n){let o=br(n,t,e);return[Zs(o)]}return}case"Enter":{let n=S("frameName",r,t,e),o=S("tensor",r,t,e);return e.enterFrame(n),[Zs(o)]}case"Exit":{let n=S("tensor",r,t,e);return e.exitFrame(),[Zs(n)]}case"NextIteration":{let n=S("tensor",r,t,e);return e.nextIteration(),[Zs(n)]}case"TensorArrayV3":{let n=S("size",r,t,e),o=S("dtype",r,t,e),s=S("elementShape",r,t,e),i=S("dynamicSize",r,t,e),a=S("clearAfterRead",r,t,e),u=S("identicalElementShapes",r,t,e),l=S("name",r,t,e),c=new zb(l,o,n,s,u,i,a);return e.addTensorArray(c),[c.idTensor,mt(1)]}case"TensorArrayWriteV3":{let n=S("tensorArrayId",r,t,e),o=S("index",r,t,e),s=S("tensor",r,t,e),i=e.getTensorArray(n.id);return i.write(o,s),[i.idTensor]}case"TensorArrayReadV3":{let n=S("tensorArrayId",r,t,e),o=S("index",r,t,e);return[e.getTensorArray(n.id).read(o)]}case"TensorArrayGatherV3":{let n=S("tensorArrayId",r,t,e),o=S("indices",r,t,e),s=S("dtype",r,t,e);return[e.getTensorArray(n.id).gather(o,s)]}case"TensorArrayScatterV3":{let n=S("tensorArrayId",r,t,e),o=S("indices",r,t,e),s=S("tensor",r,t,e),i=e.getTensorArray(n.id);return i.scatter(o,s),[i.idTensor]}case"TensorArrayConcatV3":{let n=S("tensorArrayId",r,t,e),o=e.getTensorArray(n.id),s=S("dtype",r,t,e);return[o.concat(s)]}case"TensorArraySplitV3":{let n=S("tensorArrayId",r,t,e),o=S("tensor",r,t,e),s=S("lengths",r,t,e),i=e.getTensorArray(n.id);return i.split(s,o),[i.idTensor]}case"TensorArraySizeV3":{let n=S("tensorArrayId",r,t,e),o=e.getTensorArray(n.id);return[mt(o.size(),"int32")]}case"TensorArrayCloseV3":{let n=S("tensorArrayId",r,t,e),o=e.getTensorArray(n.id);return o.clearAndClose(),[o.idTensor]}case"TensorListSetItem":{let n=S("tensorListId",r,t,e),o=S("index",r,t,e),s=S("tensor",r,t,e),i=e.getTensorList(n.id);return i.setItem(o,s),[i.idTensor]}case"TensorListGetItem":{let n=S("tensorListId",r,t,e),o=S("index",r,t,e),s=S("elementShape",r,t,e),i=S("elementDType",r,t,e);return[e.getTensorList(n.id).getItem(o,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{let n=S("indices",r,t,e),o=S("tensor",r,t,e),s=S("elementShape",r,t,e),i=S("numElements",r,t,e),a=WD(o,n,s,i);return e.addTensorList(a),[a.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let n=S("elementShape",r,t,e),o=S("elementDType",r,t,e),s;r.op==="TensorListReserve"?s="numElements":s="maxNumElements";let i=S(s,r,t,e),a=r.op==="TensorListReserve"?-1:i,u=GD(n,o,i,a);return e.addTensorList(u),[u.idTensor]}case"TensorListGather":{let n=S("tensorListId",r,t,e),o=S("indices",r,t,e),s=S("elementShape",r,t,e),i=S("elementDType",r,t,e);return[e.getTensorList(n.id).gather(o,i,s)]}case"TensorListStack":{let n=S("tensorListId",r,t,e),o=S("elementShape",r,t,e),s=S("elementDType",r,t,e),i=S("numElements",r,t,e);return[e.getTensorList(n.id).stack(o,s,i)]}case"TensorListFromTensor":{let n=S("tensor",r,t,e),o=S("elementShape",r,t,e),s=S("elementDType",r,t,e),i=VD(n,o,s);return e.addTensorList(i),[i.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let n=S("tensorListId",r,t,e),o=e.getTensorList(n.id),s=S("dtype",r,t,e),i=S("elementShape",r,t,e);return[o.concat(s,i)]}case"TensorListPushBack":{let n=S("tensorListId",r,t,e),o=S("tensor",r,t,e),s=e.getTensorList(n.id);return s.pushBack(o),[s.idTensor]}case"TensorListPopBack":{let n=S("tensorListId",r,t,e),o=S("elementShape",r,t,e),s=S("elementDType",r,t,e);return[e.getTensorList(n.id).popBack(o,s)]}case"TensorListSplit":{let n=S("tensor",r,t,e),o=S("elementShape",r,t,e),s=S("lengths",r,t,e),i=UD(n,s,o);return e.addTensorList(i),[i.idTensor]}case"TensorListLength":{let n=S("tensorListId",r,t,e),o=e.getTensorList(n.id);return[mt(o.size(),"int32")]}case"TensorListResize":{let n=S("tensorListId",r,t,e),o=S("size",r,t,e),i=e.getTensorList(n.id).resize(o);return e.addTensorList(i),[i.idTensor]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};function qD(r,t,e){let[n,o]=S("fusedOps",r,t,e),s=n==="biasadd",i=!s,a=o==="prelu",u=n==="fusedbatchnorm",l=S("numArgs",r,t,e);if(s){if(a&&l!==2)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!a&&s&&l!==1)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(u)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let c=S("strides",r,t,e),p=Rh(r,t,e),m=S("dataFormat",r,t,e).toUpperCase(),f=S("dilations",r,t,e),[d,h]=S("args",r,t,e);i&&(h=d,d=void 0);let g=S("leakyreluAlpha",r,t,e);return{stride:c,pad:p,dataFormat:m,dilations:f,biasArg:d,preluArg:h,activationFunc:o,leakyreluAlpha:g}}var KD=(r,t,e,n=ae)=>{switch(r.op){case"Conv1D":{let o=S("stride",r,t,e),s=S("pad",r,t,e),i=S("dataFormat",r,t,e).toUpperCase(),a=S("dilation",r,t,e);return[n.conv1d(S("x",r,t,e),S("filter",r,t,e),o,s,i,a)]}case"Conv2D":{let o=S("strides",r,t,e),s=Rh(r,t,e),i=S("dataFormat",r,t,e).toUpperCase(),a=S("dilations",r,t,e);return[n.conv2d(S("x",r,t,e),S("filter",r,t,e),[o[1],o[2]],s,i,[a[1],a[2]])]}case"_FusedConv2D":{let{stride:o,pad:s,dataFormat:i,dilations:a,biasArg:u,preluArg:l,activationFunc:c,leakyreluAlpha:p}=qD(r,t,e);return[n.fused.conv2d({x:S("x",r,t,e),filter:S("filter",r,t,e),strides:[o[1],o[2]],pad:s,dataFormat:i,dilations:[a[1],a[2]],bias:u,activation:c,preluActivationWeights:l,leakyreluAlpha:p})]}case"FusedDepthwiseConv2dNative":{let{stride:o,pad:s,dataFormat:i,dilations:a,biasArg:u,preluArg:l,activationFunc:c,leakyreluAlpha:p}=qD(r,t,e);return[n.fused.depthwiseConv2d({x:S("x",r,t,e),filter:S("filter",r,t,e),strides:[o[1],o[2]],pad:s,dataFormat:i,dilations:[a[1],a[2]],bias:u,activation:c,preluActivationWeights:l,leakyreluAlpha:p})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let o=S("outputShape",r,t,e),s=S("strides",r,t,e),i=Rh(r,t,e);return[n.conv2dTranspose(S("x",r,t,e),S("filter",r,t,e),o,[s[1],s[2]],i)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let o=S("strides",r,t,e),s=Rh(r,t,e),i=S("dilations",r,t,e),a=S("dataFormat",r,t,e).toUpperCase();return[n.depthwiseConv2d(S("input",r,t,e),S("filter",r,t,e),[o[1],o[2]],s,a,[i[1],i[2]])]}case"Conv3D":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("dataFormat",r,t,e).toUpperCase(),a=S("dilations",r,t,e);return[n.conv3d(S("x",r,t,e),S("filter",r,t,e),[o[1],o[2],o[3]],s,i,[a[1],a[2],a[3]])]}case"AvgPool":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("kernelSize",r,t,e);return[n.avgPool(S("x",r,t,e),[i[1],i[2]],[o[1],o[2]],s)]}case"MaxPool":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("kernelSize",r,t,e);return[n.maxPool(S("x",r,t,e),[i[1],i[2]],[o[1],o[2]],s)]}case"MaxPoolWithArgmax":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("kernelSize",r,t,e),a=S("includeBatchInIndex",r,t,e),{result:u,indexes:l}=n.maxPoolWithArgmax(S("x",r,t,e),[i[1],i[2]],[o[1],o[2]],s,a);return[u,l]}case"AvgPool3D":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("kernelSize",r,t,e);return[n.avgPool3d(S("x",r,t,e),[i[1],i[2],i[3]],[o[1],o[2],o[3]],s)]}case"MaxPool3D":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("kernelSize",r,t,e);return[n.maxPool3d(S("x",r,t,e),[i[1],i[2],i[3]],[o[1],o[2],o[3]],s)]}case"Dilation2D":{let o=S("strides",r,t,e),s=S("pad",r,t,e),i=S("dilations",r,t,e),a=o[1],u=o[2],l=i[1],c=i[2];return[n.dilation2d(S("x",r,t,e),S("filter",r,t,e),[a,u],s,[l,c],"NHWC")]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var jD=(r,t,e,n=ae)=>{switch(r.op){case"Fill":{let o=S("shape",r,t,e),s=S("dtype",r,t,e),i=S("value",r,t,e);return[n.fill(o,i,s)]}case"LinSpace":{let o=S("start",r,t,e),s=S("stop",r,t,e),i=S("num",r,t,e);return[n.linspace(o,s,i)]}case"Multinomial":{let o=S("logits",r,t,e),s=S("numSamples",r,t,e),i=S("seed",r,t,e);return[n.multinomial(o,s,i)]}case"OneHot":{let o=S("indices",r,t,e),s=S("depth",r,t,e),i=S("onValue",r,t,e),a=S("offValue",r,t,e),u=S("dtype",r,t,e);return[n.oneHot(o,s,i,a,u)]}case"Ones":return[n.ones(S("shape",r,t,e),S("dtype",r,t,e))];case"OnesLike":return[n.onesLike(S("x",r,t,e))];case"RandomStandardNormal":return[n.randomStandardNormal(S("shape",r,t,e),S("dtype",r,t,e),S("seed",r,t,e))];case"RandomUniform":return[n.randomUniform(S("shape",r,t,e),S("minval",r,t,e),S("maxval",r,t,e),S("dtype",r,t,e))];case"Range":{let o=S("start",r,t,e),s=S("stop",r,t,e),i=S("step",r,t,e);return[n.range(o,s,i,S("dtype",r,t,e))]}case"TruncatedNormal":{let o=S("shape",r,t,e),s=S("mean",r,t,e),i=S("stdDev",r,t,e),a=S("seed",r,t,e);return[n.truncatedNormal(o,s,i,S("dtype",r,t,e),a)]}case"Zeros":return[n.zeros(S("shape",r,t,e),S("dtype",r,t,e))];case"ZerosLike":return[n.zerosLike(S("x",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function mN(r,t,e){let n=S("boxes",r,t,e),o=S("scores",r,t,e),s=S("maxOutputSize",r,t,e),i=S("iouThreshold",r,t,e),a=S("scoreThreshold",r,t,e),u=S("softNmsSigma",r,t,e);return{boxes:n,scores:o,maxOutputSize:s,iouThreshold:i,scoreThreshold:a,softNmsSigma:u}}var XD=async(r,t,e,n,o=ae)=>{switch(r.op){case"NonMaxSuppressionV5":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l,softNmsSigma:c}=mN(r,t,e),p=await o.image.nonMaxSuppressionWithScoreAsync(s,i,a,u,l,c);return[p.selectedIndices,p.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l}=mN(r,t,e),c=S("padToMaxOutputSize",r,t,e),p=await o.image.nonMaxSuppressionPaddedAsync(s,i,a,u,l,c);return[p.selectedIndices,p.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:s,scores:i,maxOutputSize:a,iouThreshold:u,scoreThreshold:l}=mN(r,t,e);return[await o.image.nonMaxSuppressionAsync(s,i,a,u,l)]}case"Where":{let s=o.cast(S("condition",r,t,e),"bool"),i=[await o.whereAsync(s)];return s.dispose(),i}case"ListDiff":return o.setdiff1dAsync(S("x",r,t,e),S("y",r,t,e));default:throw TypeError(`Node type ${r.op} is not implemented`)}};var YD=(r,t,e,n=ae)=>{switch(r.op){case"LowerBound":{let o=S("sortedSequence",r,t,e),s=S("values",r,t,e);return[n.lowerBound(o,s)]}case"TopKV2":{let o=S("x",r,t,e),s=S("k",r,t,e),i=S("sorted",r,t,e),a=n.topk(o,s,i);return[a.values,a.indices]}case"UpperBound":{let o=S("sortedSequence",r,t,e),s=S("values",r,t,e);return[n.upperBound(o,s)]}case"Unique":{let o=S("x",r,t,e),s=n.unique(o);return[s.values,s.indices]}case"UniqueV2":{let o=S("x",r,t,e),s=S("axis",r,t,e),i=n.unique(o,s);return[i.values,i.indices]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var ZD=(r,t,e,n=ae)=>{switch(r.op){case"Const":return t[r.name];case"PlaceholderWithDefault":let o=S("default",r,t,e);return[br(r.name,t,e)||o];case"Placeholder":return[br(r.name,t,e)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":{let c=S("x",r,t,e);return[Zs(c)]}case"IdentityN":return S("x",r,t,e).map(c=>Zs(c));case"Snapshot":let s=S("x",r,t,e);return[Zs(s)];case"Shape":return[n.tensor1d(S("x",r,t,e).shape,"int32")];case"ShapeN":return S("x",r,t,e).map(c=>n.tensor1d(c.shape));case"Size":return[n.scalar(S("x",r,t,e).size,"int32")];case"Rank":return[n.scalar(S("x",r,t,e).rank,"int32")];case"NoOp":return[n.scalar(1)];case"Print":let i=S("x",r,t,e),a=S("data",r,t,e),u=S("message",r,t,e),l=S("summarize",r,t,e);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(u);for(let c=0;c<a.length;c++)console.log(Array.prototype.slice.call(a[c].dataSync()).slice(0,l));return[i];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var Bb=class{constructor(t,e){this.keyDType=t,this.valueDType=e,this.handle=mt(0),this.tensorMap=new Map,De(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(t=>t.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return mt(this.size(),"int32")}async import(t,e){this.checkKeyAndValueTensor(t,e);let n=await t.data();return this.tensorMap.forEach(o=>o.dispose()),this.tensorMap.clear(),B(()=>{let o=vr(e),s=n.length,i=o.length;y.assert(s===i,()=>`The number of elements doesn't match, keys has ${s} elements, the values has ${i} elements.`);for(let a=0;a<s;a++){let u=n[a],l=o[a];De(l),this.tensorMap.set(u,l)}return this.handle})}async find(t,e){this.checkKeyAndValueTensor(t,e);let n=await t.data();return B(()=>{let o=[];for(let s=0;s<n.length;s++){let i=n[s],a=this.findWithDefault(i,e);o.push(a)}return nr(o)})}findWithDefault(t,e){let n=this.tensorMap.get(t);return n!=null?n:e}checkKeyAndValueTensor(t,e){if(t.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${t.dtype}`);if(e.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${e.dtype}`)}};var JD=async(r,t,e,n)=>{switch(r.op){case"HashTable":case"HashTableV2":{let o=n.getHashTableHandleByName(r.name);if(o!=null)return[o];{let s=S("keyDType",r,t,e),i=S("valueDType",r,t,e),a=new Bb(s,i);return n.addHashTable(r.name,a),[a.handle]}}case"LookupTableImport":case"LookupTableImportV2":{let o=S("tableHandle",r,t,e,n),s=S("keys",r,t,e),i=S("values",r,t,e);return[await n.getHashTableById(o.id).import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{let o=S("tableHandle",r,t,e,n),s=S("keys",r,t,e),i=S("defaultValue",r,t,e);return[await n.getHashTableById(o.id).find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{let o=S("tableHandle",r,t,e,n);return[n.getHashTableById(o.id).tensorSize()]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var QD=(r,t,e,n=ae)=>{switch(r.op){case"ResizeBilinear":{let o=S("images",r,t,e),s=S("size",r,t,e),i=S("alignCorners",r,t,e),a=S("halfPixelCenters",r,t,e);return[n.image.resizeBilinear(o,[s[0],s[1]],i,a)]}case"ResizeNearestNeighbor":{let o=S("images",r,t,e),s=S("size",r,t,e),i=S("alignCorners",r,t,e),a=S("halfPixelCenters",r,t,e);return[n.image.resizeNearestNeighbor(o,[s[0],s[1]],i,a)]}case"CropAndResize":{let o=S("image",r,t,e),s=S("boxes",r,t,e),i=S("boxInd",r,t,e),a=S("cropSize",r,t,e),u=S("method",r,t,e),l=S("extrapolationValue",r,t,e);return[n.image.cropAndResize(o,s,i,a,u,l)]}case"ImageProjectiveTransformV3":{let o=S("images",r,t,e),s=S("transforms",r,t,e),i=S("outputShape",r,t,e),a=S("fillValue",r,t,e),u=S("interpolation",r,t,e),l=S("fillMode",r,t,e);return[n.image.transform(o,s,u.toLowerCase(),l.toLowerCase(),a,i)]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var tR=(r,t,e,n=ae)=>{switch(r.op){case"Equal":return[n.equal(S("a",r,t,e),S("b",r,t,e))];case"NotEqual":return[n.notEqual(S("a",r,t,e),S("b",r,t,e))];case"Greater":return[n.greater(S("a",r,t,e),S("b",r,t,e))];case"GreaterEqual":return[n.greaterEqual(S("a",r,t,e),S("b",r,t,e))];case"Less":return[n.less(S("a",r,t,e),S("b",r,t,e))];case"LessEqual":return[n.lessEqual(S("a",r,t,e),S("b",r,t,e))];case"LogicalAnd":return[n.logicalAnd(S("a",r,t,e),S("b",r,t,e))];case"LogicalNot":return[n.logicalNot(S("a",r,t,e))];case"LogicalOr":return[n.logicalOr(S("a",r,t,e),S("b",r,t,e))];case"Select":case"SelectV2":return[n.where(S("condition",r,t,e),S("a",r,t,e),S("b",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var eR=(r,t,e,n=ae)=>{switch(r.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[n.matMul(S("a",r,t,e),S("b",r,t,e),S("transposeA",r,t,e),S("transposeB",r,t,e))];case"Einsum":return[n.einsum(S("equation",r,t,e),...S("tensors",r,t,e))];case"Transpose":return[n.transpose(S("x",r,t,e),S("perm",r,t,e))];case"_FusedMatMul":let[o,s]=S("fusedOps",r,t,e),i=o==="biasadd",a=s==="prelu",u=S("numArgs",r,t,e),l=S("leakyreluAlpha",r,t,e);if(i){if(a&&u!==2)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!a&&u!==1)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[c,p]=S("args",r,t,e);return[n.fused.matMul({a:S("a",r,t,e),b:S("b",r,t,e),transposeA:S("transposeA",r,t,e),transposeB:S("transposeB",r,t,e),bias:c,activation:s,preluActivationWeights:p,leakyreluAlpha:l})];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var rR=(r,t,e,n=ae)=>{switch(r.op){case"EuclideanNorm":return[n.euclideanNorm(S("x",r,t,e),S("axis",r,t,e),S("keepDims",r,t,e))];case"FusedBatchNorm":case"FusedBatchNormV2":return[n.batchNorm(S("x",r,t,e),S("mean",r,t,e),S("variance",r,t,e),S("offset",r,t,e),S("scale",r,t,e),S("epsilon",r,t,e))];case"FusedBatchNormV3":return[n.batchNorm(S("x",r,t,e),S("mean",r,t,e),S("variance",r,t,e),S("offset",r,t,e),S("scale",r,t,e),S("epsilon",r,t,e))];case"LRN":return[n.localResponseNormalization(S("x",r,t,e),S("radius",r,t,e),S("bias",r,t,e),S("alpha",r,t,e),S("beta",r,t,e))];case"Softmax":return[n.softmax(S("x",r,t,e))];case"LogSoftmax":return[n.logSoftmax(S("x",r,t,e))];case"SparseToDense":return[n.sparseToDense(S("sparseIndices",r,t,e),S("outputShape",r,t,e),S("sparseValues",r,t,e),S("defaultValue",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var nR=(r,t,e,n=ae)=>{switch(r.op){case"Max":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.max(S("x",r,t,e),a,u)]}case"Mean":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.mean(S("x",r,t,e),a,u)]}case"Min":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.min(S("x",r,t,e),a,u)]}case"Sum":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.sum(S("x",r,t,e),a,u)]}case"All":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.all(S("x",r,t,e),a,u)]}case"Any":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.any(S("x",r,t,e),a,u)]}case"ArgMax":{let a=S("axis",r,t,e);return[n.argMax(S("x",r,t,e),a)]}case"ArgMin":{let a=S("axis",r,t,e);return[n.argMin(S("x",r,t,e),a)]}case"Prod":{let a=S("axis",r,t,e),u=S("keepDims",r,t,e);return[n.prod(S("x",r,t,e),a,u)]}case"Cumprod":{let a=S("axis",r,t,e),u=S("exclusive",r,t,e),l=S("reverse",r,t,e);return[n.cumprod(S("x",r,t,e),a,u,l)]}case"Cumsum":{let a=S("axis",r,t,e),u=S("exclusive",r,t,e),l=S("reverse",r,t,e);return[n.cumsum(S("x",r,t,e),a,u,l)]}case"Bincount":let o=S("x",r,t,e),s=S("weights",r,t,e),i=S("size",r,t,e);return[n.bincount(o,s,i)];case"DenseBincount":{let a=S("x",r,t,e),u=S("weights",r,t,e),l=S("size",r,t,e),c=S("binaryOutput",r,t,e);return[n.denseBincount(a,u,l,c)]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var oR=(r,t,e,n=ae)=>{switch(r.op){case"ConcatV2":case"Concat":{let o=S("n",r,t,e),s=S("axis",r,t,e),i=S("tensors",r,t,e);return i=i.slice(0,o),[n.concat(i,s)]}case"Gather":{let o=S("x",r,t,e),s=S("indices",r,t,e);return[n.gather(o,n.cast(s,"int32"),0)]}case"GatherV2":{let o=S("axis",r,t,e),s=S("batchDims",r,t,e),i=S("x",r,t,e),a=S("indices",r,t,e);return[n.gather(i,n.cast(a,"int32"),o,s)]}case"Reverse":{let o=S("dims",r,t,e),s=[];for(let a=0;a<o.length;a++)o[a]&&s.push(a);let i=S("x",r,t,e);return[n.reverse(i,s)]}case"ReverseV2":{let o=S("axis",r,t,e),s=S("x",r,t,e);return[n.reverse(s,o)]}case"Slice":{let o=S("begin",r,t,e),s=S("size",r,t,e);return[n.slice(S("x",r,t,e),o,s)]}case"StridedSlice":{let o=S("begin",r,t,e),s=S("end",r,t,e),i=S("strides",r,t,e),a=S("beginMask",r,t,e),u=S("endMask",r,t,e),l=S("ellipsisMask",r,t,e),c=S("newAxisMask",r,t,e),p=S("shrinkAxisMask",r,t,e),m=S("x",r,t,e);return[n.stridedSlice(m,o,s,i,a,u,l,c,p)]}case"Pack":return B(()=>{let o=S("axis",r,t,e),s=S("tensors",r,t,e),i=s[0].shape,a=n.squeeze(s[0]).shape,u=s.map(l=>{let c=y.arraysEqual(l.shape,i);if(!c&&!y.arraysEqual(n.squeeze(l).shape,a))throw new Error("the input tensors shape does not match");return c?l:n.reshape(l,i)});return[n.stack(u,o)]});case"Unpack":{let o=S("axis",r,t,e),s=S("tensor",r,t,e);return n.unstack(s,o)}case"Tile":{let o=S("reps",r,t,e);return[n.tile(S("x",r,t,e),o)]}case"Split":case"SplitV":{let o=S("axis",r,t,e),s=S("numOrSizeSplits",r,t,e),i=S("x",r,t,e);return n.split(i,s,o)}case"ScatterNd":{let o=S("indices",r,t,e),s=S("values",r,t,e),i=S("shape",r,t,e);return[n.scatterND(o,s,i)]}case"GatherNd":{let o=S("x",r,t,e),s=S("indices",r,t,e);return[n.gatherND(o,s)]}case"SparseToDense":{let o=S("sparseIndices",r,t,e),s=S("outputShape",r,t,e),i=S("sparseValues",r,t,e),a=S("defaultValue",r,t,e);return[n.sparseToDense(o,i,s,i.dtype===a.dtype?a:n.cast(a,i.dtype))]}default:throw TypeError(`Node type ${r.op} is not implemented`)}};var sR=(r,t,e,n=ae)=>{switch(r.op){case"SparseFillEmptyRows":{let{outputIndices:o,outputValues:s,emptyRowIndicator:i,reverseIndexMap:a}=n.sparse.sparseFillEmptyRows(S("indices",r,t,e),S("values",r,t,e),S("denseShape",r,t,e),S("defaultValue",r,t,e));return[o,s,i,a]}case"SparseReshape":{let{outputIndices:o,outputShape:s}=n.sparse.sparseReshape(S("inputIndices",r,t,e),S("inputShape",r,t,e),S("newShape",r,t,e));return[o,s]}case"SparseSegmentMean":return[n.sparse.sparseSegmentMean(S("data",r,t,e),S("indices",r,t,e),S("segmentIds",r,t,e))];case"SparseSegmentSum":return[n.sparse.sparseSegmentSum(S("data",r,t,e),S("indices",r,t,e),S("segmentIds",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var iR=(r,t,e,n=ae)=>{switch(r.op){case"FFT":return[n.fft(S("x",r,t,e))];case"IFFT":return[n.ifft(S("x",r,t,e))];case"RFFT":return[n.rfft(S("x",r,t,e))];case"IRFFT":return[n.irfft(S("x",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var aR=(r,t,e,n=ae)=>{switch(r.op){case"StringNGrams":{let{nGrams:o,nGramsSplits:s}=n.string.stringNGrams(S("data",r,t,e),S("dataSplits",r,t,e),S("separator",r,t,e),S("nGramWidths",r,t,e),S("leftPad",r,t,e),S("rightPad",r,t,e),S("padWidth",r,t,e),S("preserveShortSequences",r,t,e));return[o,s]}case"StringSplit":{let{indices:o,values:s,shape:i}=n.string.stringSplit(S("input",r,t,e),S("delimiter",r,t,e),S("skipEmpty",r,t,e));return[o,s,i]}case"StringToHashBucketFast":return[n.string.stringToHashBucketFast(S("input",r,t,e),S("numBuckets",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};var lR=(r,t,e,n=ae)=>{switch(r.op){case"Cast":return[n.cast(S("x",r,t,e),S("dtype",r,t,e))];case"ExpandDims":{let o=S("axis",r,t,e);return[n.expandDims(S("x",r,t,e),o)]}case"Squeeze":{let o=S("axis",r,t,e);return[n.squeeze(S("x",r,t,e),o)]}case"Reshape":return[n.reshape(S("x",r,t,e),S("shape",r,t,e))];case"MirrorPad":return[n.mirrorPad(S("x",r,t,e),S("padding",r,t,e),S("mode",r,t,e))];case"PadV2":case"Pad":return[n.pad(S("x",r,t,e),S("padding",r,t,e),S("constantValue",r,t,e))];case"SpaceToBatchND":{let o=S("blockShape",r,t,e),s=S("paddings",r,t,e);return[n.spaceToBatchND(S("x",r,t,e),o,s)]}case"BatchToSpaceND":{let o=S("blockShape",r,t,e),s=S("crops",r,t,e);return[n.batchToSpaceND(S("x",r,t,e),o,s)]}case"DepthToSpace":{let o=S("blockSize",r,t,e),s=S("dataFormat",r,t,e).toUpperCase();return[n.depthToSpace(S("x",r,t,e),o,s)]}case"BroadcastTo":return[n.broadcastTo(S("x",r,t,e),S("shape",r,t,e))];case"BroadcastArgs":return[n.broadcastArgs(S("s0",r,t,e),S("s1",r,t,e))];default:throw TypeError(`Node type ${r.op} is not implemented`)}};function fN(r,t,e,n,o=B){let s=((i,a,u)=>{switch(i.category){case"arithmetic":return o(()=>MD(i,a,u));case"basic_math":return o(()=>zD(i,a,u));case"control":return HD(i,a,u);case"convolution":return o(()=>KD(i,a,u));case"creation":return o(()=>jD(i,a,u));case"dynamic":return XD(i,a,u);case"evaluation":return o(()=>YD(i,a,u));case"image":return o(()=>QD(i,a,u));case"graph":return o(()=>ZD(i,a,u));case"logical":return o(()=>tR(i,a,u));case"matrices":return o(()=>eR(i,a,u));case"normalization":return o(()=>rR(i,a,u));case"reduction":return o(()=>nR(i,a,u));case"slice_join":return o(()=>oR(i,a,u));case"sparse":return o(()=>sR(i,a,u));case"spectral":return o(()=>iR(i,a,u));case"string":return o(()=>aR(i,a,u));case"transformation":return o(()=>lR(i,a,u));case"hash_table":return JD(i,a,u,n);case"custom":let l=Nb(i.op);if(l&&l.customExecutor)return l.customExecutor(new Lb(i,a,u));throw TypeError(`Custom op ${i.op} is not registered.`);default:throw TypeError(`Unknown op '${i.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(r,t,e);return y.isPromise(s)?s.then(i=>[].concat(i)):[].concat(s)}var Oh=class{constructor(t={},e={},n={},o={}){this.weightMap=t,this.tensorArrayMap=e,this.tensorListMap=n,this.functionMap=o,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(t,e){return{id:t,frameName:e,iterationId:0}}set currentContext(t){this.contexts!==t&&(this.contexts=t,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let t=[];for(let e=0;e<this.contexts.length-1;e++){let n=this.contexts.slice(0,this.contexts.length-e);t.push(this.contextIdforContexts(n))}t.push(""),this._currentContextIds=t}contextIdforContexts(t){return t?t.map(e=>e.id===0&&e.iterationId===0?"":`${e.frameName}-${e.iterationId}`).join("/"):""}enterFrame(t){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,t)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error("Cannot exit frame, the context is empty")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let t=Object.assign({},this.contexts[this.contexts.length-1]);t.iterationId+=1,t.id=this.lastId,this.contexts.splice(-1,1,t),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error("Cannot increase frame iteration, the context is empty")}getWeight(t){return this.weightMap[t]}addTensorArray(t){this.tensorArrayMap[t.id]=t}getTensorArray(t){return this.tensorArrayMap[t]}addTensorList(t){this.tensorListMap[t.id]=t}getTensorList(t){return this.tensorListMap[t]}dispose(t){for(let e in this.tensorArrayMap)this.tensorArrayMap[e].clearAndClose(t);for(let e in this.tensorListMap)this.tensorListMap[e].clearAndClose(t)}};function dN(r,t,e,n){let o=new Set,s=[],i=null,a=null,u=new Set,l=Object.keys(r).map(m=>xn(m)[0]),c=[];n!=null&&(c=n.map(m=>xn(m.name)[0]));let p=[...t];for(;p.length>0;){let m=p.pop();if((hN(m)||F7(m)||O7(m))&&i==null&&(i=m,a=i.children.map(f=>f.name).filter(f=>o.has(f))),o.add(m.name),e[m.name]==null&&l.indexOf(m.name)===-1&&c.indexOf(m.name)===-1){if(m.inputs.length===0){s.push(m.name);continue}m.inputs.forEach(f=>{u.has(f.name)||(u.add(f.name),p.push(f))})}}return{inputs:r,outputs:t,usedNodes:o,missingInputs:s,dynamicNode:i,syncInputs:a}}function uR(r,t,e){let{usedNodes:n,inputs:o}=e,s=[],i=Object.keys(o).map(c=>xn(c)[0]).map(c=>r.nodes[c]),a=r.initNodes;i.forEach(c=>{n.has(c.name)&&s.push(c)}),r.weights.forEach(c=>{n.has(c.name)&&s.push(c)}),a!=null&&a.forEach(c=>{n.has(c.name)&&s.push(c)});let u=new Set,l=[];for(;s.length>0;){let c=s.pop();u.add(c.name),t[c.name]||l.push(c),c.children.forEach(p=>{!u.has(p.name)&&n.has(p.name)&&p.inputs.every(m=>u.has(m.name))&&s.push(p)})}return l}var $7=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],D7=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],R7=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function hN(r){return $7.indexOf(r.op)>=0}function F7(r){return D7.indexOf(r.op)>=0}function O7(r){return R7.indexOf(r.op)>=0}var Nc=class{constructor(t,e){this.graph=t,this.parent=e,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=t.outputs,this._inputs=t.inputs,this._initNodes=t.initNodes,this._signature=t.signature,this._functions=t.functions,t.functions!=null&&Object.keys(t.functions).forEach(n=>{this._functionExecutorMap[n]=new Nc(t.functions[n],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(t){let e=Object.keys(t).map(n=>t[n].map(o=>o.id));this._weightIds=[].concat(...e),this._weightMap=t}set resourceManager(t){this._resourceManager=t}get inputs(){return this._inputs.map(t=>({name:t.name,shape:t.attrParams.shape?t.attrParams.shape.value:void 0,dtype:t.attrParams.dtype?t.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(t=>({name:t.name,shape:t.attrParams.shape?t.attrParams.shape.value:void 0,dtype:t.attrParams.dtype?t.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(t=>t.signatureKey||t.name)}get outputNodes(){return this._outputs.map(t=>{let e=t.signatureKey||t.name;return t.defaultOutput?`${e}:${t.defaultOutput}`:e})}get functions(){return Object.keys(this._functions).reduce((t,e)=>(t[e]=this._functions[e].signature,t),{})}getCompilationKey(t,e){let n=t.map(s=>s.name).sort(),o=e.map(s=>s.name).sort();return n.join(this.SEPERATOR)+"--"+o.join(this.SEPERATOR)}compile(t,e){let n=dN(t,e,this.weightMap,this._initNodes),{missingInputs:o,dynamicNode:s,syncInputs:i}=n;if(s!=null)throw new Error(`This execution contains the node '${s.name}', which has the dynamic op '${s.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${i}]`);if(o.length>0){let a=e.map(l=>l.name),u=Object.keys(t);throw new Error(`Cannot compute the outputs [${a}] from the provided inputs [${u}]. Missing the following inputs: [${o}]`)}return uR(this.graph,this.weightMap,n)}execute(t,e){t=this.mapInputs(t);let n=Object.keys(t).sort();this.checkInputs(t),this.checkInputShapeAndType(t),e=this.mapOutputs(e),this.checkOutputs(e);let o=n.map(p=>this.graph.nodes[xn(p)[0]]),s=e.map(p=>xn(p)[0]),i=s.map(p=>this.graph.nodes[p]);this.resetIntermediateTensors(),i.length===0&&(i=this._outputs);let a=this.getCompilationKey(o,i),u=this.compiledMap.get(a);u==null&&(u=this.compile(t,i),this.compiledMap.set(a,u));let l={},c={};return B(()=>{let p=new Oh(this.weightMap,l,c,this.functionExecutorMap),m=Object.assign({},this.weightMap);Object.keys(t).forEach(h=>{let[g,x]=xn(h),b=[];b[x]=t[h],m[g]=b});let f=this.getFrozenTensorIds(m),d={};for(let h=0;h<u.length;h++){let g=u[h];if(!m[g.name]){let x=fN(g,m,p,this._resourceManager);if(y.isPromise(x))throw new Error(`The execution of the op '${g.op}' returned a promise. Please use model.executeAsync() instead.`);m[g.name]=x,this.checkTensorForDisposal(g.name,g,m,p,f,s,d)}}return this.parent==null&&p.dispose(f),e.map(h=>br(h,m,p))})}getFrozenTensorIds(t){let e=[].concat.apply([],Object.keys(t).map(n=>t[n]).map(n=>n.map(o=>o.id)));return new Set(e)}checkTensorForDisposal(t,e,n,o,s,i,a){e.category==="control"||i.indexOf(t)!==-1||(n[t].forEach(u=>{u!=null&&(a[u.id]=(a[u.id]||0)+e.children.length)}),e.inputs.forEach(u=>{if(u.category!=="control"){let l=FD(u.name,n,o);l!=null&&l.forEach(c=>{if(c&&!c.kept&&!s.has(c.id)){let p=a[c.id];if(p===1){if(!this.keepTensorForDebug)c.dispose();else{let[m,f]=_o(e.name,o);this.intermediateTensors[m]?this.intermediateTensors[m][f]=c:(this.intermediateTensors[m]=[],this.intermediateTensors[m][f]=c)}delete a[c.id]}else p!=null&&a[c.id]--}})}}))}async executeAsync(t,e){return this._executeAsync(t,e)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(t=>this.intermediateTensors[t].forEach(e=>e.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(t=>{this.tensorsMap[t].forEach(n=>{n&&!n.kept&&!n.isDisposed&&!this.keepIds.has(n.id)&&n.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let t in this.intermediateTensors)this.intermediateTensors[t].forEach(e=>e.dispose()),delete this.intermediateTensors[t]}async _executeAsync(t,e,n=!1,o={},s={}){n||(t=this.mapInputs(t),this.checkInputs(t),this.checkInputShapeAndType(t),e=this.mapOutputs(e),this.checkOutputs(e));try{this.keepTensorForDebug=z().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(c){console.warn(c.message)}this.resetIntermediateTensors();let i=new Oh(this.weightMap,o,s,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(t,i,e,n);let a=e.map(c=>br(c,this.tensorsMap,i)),u=a.map(c=>c.id),l=Object.keys(t).map(c=>t[c].id);return this.keepIds=new Set([...u,...l,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&i.dispose(this.keepIds),a}async executeFunctionAsync(t,e,n){let o=t.reduce((s,i,a)=>(s[this.inputs[a].name]=i,s),{});return this._executeAsync(o,this.outputNodes,!0,e,n)}async executeWithControlFlow(t,e,n,o){let s=Object.keys(t),i=s.map(w=>this.graph.nodes[xn(w)[0]]),a=n.map(w=>xn(w)[0]),u=a.map(w=>this.graph.nodes[w]);u.length===0&&(u=this._outputs);let{usedNodes:l,missingInputs:c,dynamicNode:p,syncInputs:m}=dN(t,u,this.weightMap,this._initNodes),f=[...i,...this.graph.weights,...this._initNodes||[]].map(w=>({node:w,contexts:e.currentContext})),d=Object.assign({},this.weightMap);Object.keys(t).forEach(w=>{let[C,N]=xn(w),_=[];_[N]=t[w],d[C]=_});let h={},g=this.getFrozenTensorIds(d),x={};for(;f.length>0;){let w=this.processStack(i,f,e,d,x,g,a,h,l);await Promise.all(w)}p==null&&!o&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let b=u.filter(w=>!hN(w)&&!br(w.name,d,e)).map(w=>w.name);if(b.length>0){let w="";throw p!=null&&(w=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${m}]`),new Error(`Cannot compute the outputs [${b}] from the provided inputs [${s}]. Consider providing the following inputs: [${c}]. ${w}`)}return d}processStack(t,e,n,o,s,i,a,u,l){let c=[];for(;e.length>0;){let p=e.pop();n.currentContext=p.contexts;let m="";if(p.node.op==="Enter"&&S("isConstant",p.node,o,n)&&([m]=_o(p.node.name,n)),o[p.node.name]==null){let f=fN(p.node,o,n,this._resourceManager);m||([m]=_o(p.node.name,n));let d=n.currentContext;y.isPromise(f)?c.push(f.then(h=>(o[m]=h,n.currentContext=d,this.checkTensorForDisposal(m,p.node,o,n,i,a,u),this.processChildNodes(p.node,e,n,o,s,l),h))):(o[m]=f,this.checkTensorForDisposal(m,p.node,o,n,i,a,u),this.processChildNodes(p.node,e,n,o,s,l))}else this.processChildNodes(p.node,e,n,o,s,l)}return c}processChildNodes(t,e,n,o,s,i){t.children.forEach(a=>{let[u]=_o(a.name,n);s[u]||!i.has(a.name)||(a.op==="Merge"?a.inputNames.some(l=>!!br(l,o,n))&&(s[u]=!0,e.push({contexts:n.currentContext,node:a})):a.inputNames.every(l=>!!br(l,o,n))&&(s[u]=!0,e.push({contexts:n.currentContext,node:a})))})}dispose(){Object.keys(this.weightMap).forEach(t=>this.weightMap[t].forEach(e=>e.dispose()))}checkInputShapeAndType(t){Object.keys(t).forEach(e=>{let n=t[e],[o]=xn(e),s=this.graph.nodes[o];if(s.attrParams.shape&&s.attrParams.shape.value){let i=s.attrParams.shape.value,a=i.length===n.shape.length&&n.shape.every((u,l)=>i[l]===-1||i[l]===u);y.assert(a,()=>`The shape of dict['${s.name}'] provided in model.execute(dict) must be [${i}], but was [${n.shape}]`)}s.attrParams.dtype&&s.attrParams.dtype.value&&y.assert(n.dtype===s.attrParams.dtype.value,()=>`The dtype of dict['${s.name}'] provided in model.execute(dict) must be ${s.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(t){let e={};for(let n in t)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[n]!=null){let o=this._signature.inputs[n];e[o.name]=t[n]}else e[n]=t[n];return e}checkInputs(t){let e=Object.keys(t).filter(n=>{let[o]=xn(n);return this.graph.nodes[o]==null});if(e.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${e}] that are not part of graph`)}mapOutputs(t){return t.map(e=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[e]!=null?this._signature.outputs[e].name:e,{})}checkOutputs(t){t.forEach(e=>{let[n]=xn(e);if(!this.graph.nodes[n])throw new Error(`The output '${e}' is not found in the graph`)})}};var Vb=class{constructor(t={},e={}){this.hashTableNameToHandle=t,this.hashTableMap=e}addHashTable(t,e){this.hashTableNameToHandle[t]=e.handle,this.hashTableMap[e.id]=e}getHashTableHandleByName(t){return this.hashTableNameToHandle[t]}getHashTableById(t){return this.hashTableMap[t]}dispose(){for(let t in this.hashTableMap)this.hashTableMap[t].clearAndClose(),delete this.hashTableMap[t];for(let t in this.hashTableNameToHandle)this.hashTableNameToHandle[t].dispose(),delete this.hashTableNameToHandle[t]}};var P7="?tfjs-format=file",L7="model.json",Ph=class{constructor(t,e={},n=_r){this.modelUrl=t,this.loadOptions=e,this.version="n/a",this.io=n,e==null&&(this.loadOptions={}),this.resourceManager=new Vb}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}findIOHandler(){let t=this.modelUrl;if(t.load!=null)this.handler=t;else if(this.loadOptions.requestInit!=null)this.handler=this.io.browserHTTPRequest(t,this.loadOptions);else{let e=this.io.getLoadHandlers(t,this.loadOptions);if(e.length===0)e.push(this.io.browserHTTPRequest(t,this.loadOptions));else if(e.length>1)throw new Error(`Found more than one (${e.length}) load handlers for URL '${[t]}'`);this.handler=e[0]}}load(){if(this.findIOHandler(),this.handler.load==null)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let t=this.handler.load();return y.isPromise(t)?t.then(e=>this.loadSync(e)):this.loadSync(t)}loadSync(t){this.artifacts=t;let e=this.artifacts.modelTopology,n=this.artifacts.signature;if(this.artifacts.userDefinedMetadata!=null){let s=this.artifacts.userDefinedMetadata;s.signature!=null&&(n=s.signature),s.structuredOutputKeys!=null&&(this.structuredOutputKeys=s.structuredOutputKeys)}this.signature=n,this.version=`${e.versions.producer}.${e.versions.minConsumer}`;let o=this.io.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new Nc(Fh.Instance.transformGraph(e,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(o),this.executor.resourceManager=this.resourceManager,t.modelInitializer!=null&&t.modelInitializer.node!=null){let s=Fh.Instance.transformGraph(t.modelInitializer);this.initializer=new Nc(s),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=t.initializerSignature}return!0}async save(t,e){if(typeof t=="string"){let n=this.io.getSaveHandlers(t);if(n.length===0)throw new Error(`Cannot find any save handlers for URL '${t}'`);if(n.length>1)throw new Error(`Found more than one (${n.length}) save handlers for URL '${t}'`);t=n[0]}if(t.save==null)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return t.save(this.artifacts)}predict(t,e){let n=this.execute(t,this.outputNodes);if(this.structuredOutputKeys){let o=n instanceof Ft?[n]:n,s={};return o.forEach((i,a)=>s[this.structuredOutputKeys[a]]=i),s}return n}normalizeInputs(t){if(!(t instanceof Ft)&&!Array.isArray(t)){if(this.signature!=null&&this.signature.inputs!=null)for(let o in this.signature.inputs){let s=this.signature.inputs[o];s.resourceId!=null&&(t[o]=this.resourceIdToCapturedInput[s.resourceId])}return t}t=Array.isArray(t)?t:[t];let e=Object.keys(this.resourceIdToCapturedInput).length;if(t.length+e!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-e} non-resource placeholders, while there are ${t.length} input tensors provided.`);let n=0;return this.inputNodes.reduce((o,s)=>{let i=this.signature?this.signature.inputs[s]:null;return i!=null&&i.resourceId!=null?o[s]=this.resourceIdToCapturedInput[i.resourceId]:o[s]=t[n++],o},{})}normalizeOutputs(t){return t=t||this.outputNodes,Array.isArray(t)?t:[t]}executeInitializerGraph(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return this.initializer==null?[]:this.initializerSignature==null?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(t){if(this.resourceIdToCapturedInput={},this.initializerSignature){let e=Object.keys(this.initializerSignature.outputs);for(let n=0;n<e.length;n++){let o=e[n],s=this.initializerSignature.outputs[o];this.resourceIdToCapturedInput[s.resourceId]=t[n]}}}execute(t,e){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),t=this.normalizeInputs(t),e=this.normalizeOutputs(e);let n=this.executor.execute(t,e);return n.length>1?n:n[0]}async executeAsync(t,e){this.resourceIdToCapturedInput==null&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),t=this.normalizeInputs(t),e=this.normalizeOutputs(e);let n=await this.executor.executeAsync(t,e);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(t){return Object.keys(t).reduce((e,n)=>(e[n]=[t[n]],e),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&vt(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function M7(r,t={},e=_r){if(r==null)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");t==null&&(t={}),t.fromTFHub&&typeof r=="string"&&(r=B7(r));let n=new Ph(r,t,e);return await n.load(),n}function z7(r){if(r==null)throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");let t;if(r instanceof Array){let[n,o]=r;if(!n)throw new Error("modelJSON must be the first element of the array");if(!o||!(o instanceof ArrayBuffer))throw new Error("An ArrayBuffer of weights must be the second element of the array");if(!("modelTopology"in n))throw new Error("Model JSON is missing 'modelTopology'");if(!("weightsManifest"in n))throw new Error("Model JSON is missing 'weightsManifest'");let s=_r.getWeightSpecs(n.weightsManifest),i=_r.getModelArtifactsForJSONSync(n,s,o);t=_r.fromMemorySync(i)}else if("load"in r)t=r;else if("modelTopology"in r&&"weightSpecs"in r&&"weightData"in r)t=_r.fromMemorySync(r);else throw new Error("Unknown model format");let e=new Ph(t);return e.load(),e}function B7(r){return r.endsWith("/")||(r=r+"/"),`${r}${L7}${P7}`}var cR="4.0.0";var AR={};Wt(AR,{CSVDataset:()=>Yf,Dataset:()=>Js,FileDataSource:()=>ed,TextLineDataset:()=>Xf,URLDataSource:()=>rd,array:()=>wR,csv:()=>NR,func:()=>TR,generator:()=>kR,microphone:()=>_R,version_data:()=>PN,webcam:()=>ER,zip:()=>CR});var bR=Tl(fh());var gR=Tl(fh());function pR(r,t){return Gb(r,t)}function Gb(r,t,e=new Map,n=new Set){if(r==null)return null;if(typeof Blob=="function"&&r instanceof Blob)return r.slice();if(n.has(r))throw new Error("Circular references are not supported.");if(e.has(r))return e.get(r);let o=t(r);if(o.recurse&&o.value!==null)throw new Error("A deep map function may not return both a value and recurse=true.");if(o.recurse)if(vu(r)){let s=Array.isArray(r)?[]:{};n.add(r);for(let i in r){let a=r[i],u=Gb(a,t,e,n);s[i]=u}return n.delete(r),r.__proto__&&(s.__proto__=r.__proto__),s}else throw new Error(`Can't recurse into non-iterable type: ${r}`);else return e.set(r,o.value),o.value}function mR(r,t=xN){return fR(r,t)}function fR(r,t,e=new Set){let n=r[0];if(e.has(n))throw new Error("Circular references are not supported.");let o=t(r);if(o.recurse&&o.value!==null)throw new Error("A deep zip function may not return both a value and recurse=true.");if(o.recurse)if(vu(n)){let s=Array.isArray(n)?[]:{};e.add(n);for(let i in n){let a=r.map(l=>l[i]),u=fR(a,t,e);s[i]=u}return e.delete(n),s}else throw new Error(`Can't recurse into non-iterable type: ${n}`);else return o.value}function xN(r){return r===null?null:vu(r[0])?{value:null,recurse:!0}:{value:r,recurse:!1}}async function Wb(r,t){let e=new Map;Gb(r,t,e);for(let o of Array.from(e.keys())){let s=e.get(o);if(y.isPromise(s)){let i=await s;e.set(o,i)}}return Gb(r,t,e)}function vu(r){let t=!1;if(z().get("IS_BROWSER"))t=r instanceof TextDecoder;else{let{StringDecoder:e}=gN();t=r instanceof e}return r!=null&&!ArrayBuffer.isView(r)&&(Array.isArray(r)||typeof r=="object"&&!(r instanceof Ft)&&!(r instanceof Promise)&&!t)}function dR(r){return r==null||V7(r)||Array.isArray(r)||typeof r=="object"&&r instanceof Ft||y.isTypedArray(r)}function V7(r){return r===null||typeof r!="object"&&typeof r!="function"}function hR(r){return pR(r,G7)}function G7(r){return r instanceof Ft?{value:r.clone(),recurse:!1}:vu(r)?{value:null,recurse:!0}:{value:r,recurse:!1}}var jf=class{constructor(t){if(this.capacity=t,this.begin=0,this.end=0,t==null)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(t<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(t),this.doubledCapacity=2*t}wrap(t){for(;t<0;)t+=this.doubledCapacity;return t%this.doubledCapacity}get(t){if(t<0)throw new RangeError("Can't get item at a negative index.");return this.data[t%this.capacity]}set(t,e){if(t<0)throw new RangeError("Can't set item at a negative index.");this.data[t%this.capacity]=e}length(){let t=this.end-this.begin;return t<0&&(t=this.doubledCapacity+t),t}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(t){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,t),this.end=this.wrap(this.end+1)}pushAll(t){for(let e of t)this.push(e)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let t=this.get(this.end);return this.set(this.end,void 0),t}unshift(t){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,t)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),t}shuffleExcise(t){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.wrap(this.begin+t),n=this.get(e);return this.set(e,this.pop()),n}};var Tc=class extends jf{constructor(){super(Tc.INITIAL_CAPACITY)}isFull(){return!1}push(t){super.isFull()&&this.expand(),super.push(t)}unshift(t){super.isFull()&&this.expand(),super.unshift(t)}expand(){let t=this.capacity*2,e=new Array(t),n=this.length();for(let o=0;o<n;o++)e[o]=this.get(this.wrap(this.begin+o));this.data=e,this.capacity=t,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};Tc.INITIAL_CAPACITY=32;function AN(r){return new yN(r)}function Lh(r){return new bN(r)}function xR(r,t){return new Hb(r,t)}function yR(r,t=fl.FAIL){return new EN(r,t)}var Je=class{async toArray(){let t=[],e=await this.next();for(;!e.done;)t.push(e.value),e=await this.next();return t}async toArrayForTest(){let t=this.prefetch(100),e=[],n=await t.next();for(;!n.done;)e.push(n.value),n=await t.next();return e}async resolveFully(){let t=await this.next();for(;!t.done;)t=await this.next()}async resolveWhile(t){let e=await this.next(),n=t(e.value);for(;!e.done&&n;)e=await this.next(),n=t(e.value)}handleErrors(t){return new TN(this,t)}filter(t){return new vN(this,t)}map(t){return new NN(this,t)}mapAsync(t){return new Ub(this,t)}serialMapAsync(t){return new Ub(this,t).serial()}flatmap(t){return new kN(this,t)}async forEachAsync(t){return this.map(t).resolveFully()}async serialForEach(t){return this.serialMapAsync(t).resolveWhile(e=>e===!0)}rowMajorBatch(t,e=!0){return new SN(this,t,e)}columnMajorBatch(t,e=!0,n=xN){return this.rowMajorBatch(t,e).map(s=>mR(s,n))}concatenate(t,e){return new Hb(AN([this,t]),e)}take(t){return t<0||t==null?this:new IN(this,t)}skip(t){return t<0||t==null?this:new CN(this,t)}prefetch(t){return new qb(this,t)}shuffle(t,e){return new _N(this,t,e)}serial(){return new wN(this)}},yN=class extends Je{constructor(t){super(),this.items=t,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let t=this.items[this.trav];return this.trav++,{value:hR(t),done:!1}}},bN=class extends Je{constructor(t){super(),this.nextFn=t}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(t){throw t.message=`Error thrown while iterating through a dataset: ${t.message}`,t}}},wN=class extends Je{constructor(t){super(),this.upstream=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},CN=class extends Je{constructor(t,e){super(),this.upstream=t,this.maxCount=e,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let t=await this.upstream.next();if(t.done)return t;vt(t.value)}return this.upstream.next()}},IN=class extends Je{constructor(t,e){super(),this.upstream=t,this.maxCount=e,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},SN=class extends Je{constructor(t,e,n=!0){super(),this.upstream=t,this.batchSize=e,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let t=[];for(;t.length<this.batchSize;){let e=await this.upstream.next();if(e.done)return this.enableSmallLastBatch&&t.length>0?{value:t,done:!1}:{value:null,done:!0};t.push(e.value)}return{value:t,done:!1}}},vN=class extends Je{constructor(t,e){super(),this.upstream=t,this.predicate=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let t=await this.upstream.next();if(t.done||this.predicate(t.value))return t;vt(t.value)}}},NN=class extends Je{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> Map`}async next(){let t=await this.upstream.next();if(t.done)return{value:null,done:!0};let e=go.getTensorsInContainer(t.value),n=this.transform(t.value),o=go.getTensorsInContainer(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return{value:n,done:!1}}},TN=class extends Je{constructor(t,e){super(),this.upstream=t,this.handler=e,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(t){if(!this.handler(t))return{value:null,done:!0}}}},Ub=class extends Je{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let t=await this.upstream.next();if(t.done)return{value:null,done:!0};let e=go.getTensorsInContainer(t.value),n=await this.transform(t.value),o=go.getTensorsInContainer(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return{value:n,done:!1}}},kc=class extends Je{constructor(){super(),this.outputQueue=new Tc,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},kN=class extends kc{constructor(t,e){super(),this.upstream=t,this.transform=e}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let t=await this.upstream.next();if(t.done)return!1;let e=go.getTensorsInContainer(t.value),n=this.transform(t.value),o=go.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let s of e)go.isTensorInList(s,o)||s.dispose();return!0}},Hb=class extends Je{constructor(t,e){super(),this.baseErrorHandler=e,this.lastRead=null,this.iterator=null,this.moreIterators=t}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(t){if(await t,this.iterator==null){let n=await this.moreIterators.next();if(n.done)return{value:null,done:!0};this.iterator=n.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let e=await this.iterator.next();return e.done?(this.iterator=null,this.readFromChain(t)):e}},fl;(function(r){r[r.FAIL=0]="FAIL",r[r.SHORTEST=1]="SHORTEST",r[r.LONGEST=2]="LONGEST"})(fl||(fl={}));var EN=class extends Je{constructor(t,e=fl.FAIL){super(),this.iterators=t,this.mismatchMode=e,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(t){await t;let e=0,n=0;function o(i){return i instanceof Je?{value:i.next().then(u=>(e++,u.done&&n++,u.value)),recurse:!1}:{value:null,recurse:!0}}let s=await Wb(this.iterators,o);if(e===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case fl.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case fl.SHORTEST:return{value:null,done:!0};case fl.LONGEST:default:}return this.count++,{value:s,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},qb=class extends Je{constructor(t,e){super(),this.upstream=t,this.bufferSize=e,this.buffer=new jf(e)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let t=this.upstream.next();this.buffer.push(t)}}next(){return this.refill(),this.buffer.shift()}},_N=class extends qb{constructor(t,e,n){super(t,e),this.upstream=t,this.windowSize=e,this.upstreamExhausted=!1,this.random=gR.alea(n||y.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(t){return Math.floor(this.random()*t)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let t=this.chooseIndex(),e=await this.buffer.shuffleExcise(t);if(e.done)this.upstreamExhausted=!0;else return this.refill(),e}return{value:null,done:!0}}};var Js=class{constructor(){this.size=null}batch(t,e=!0){let n=this;y.assert(t>0,()=>`batchSize needs to be positive, but it is
-      ${t}`);let o;return this.size===1/0||this.size==null?o=this.size:e?o=Math.ceil(this.size/t):o=Math.floor(this.size/t),kn(async()=>(await n.iterator()).columnMajorBatch(t,e,W7),o)}concatenate(t){let e=this,n;return this.size===1/0||t.size===1/0?n=1/0:this.size!=null&&t.size!=null?n=this.size+t.size:n=null,kn(async()=>(await e.iterator()).concatenate(await t.iterator()),n)}filter(t){let e=this,n;return this.size===1/0?n=1/0:n=null,kn(async()=>(await e.iterator()).filter(o=>B(()=>t(o))),n)}async forEachAsync(t){return(await this.iterator()).forEachAsync(t)}map(t){let e=this;return kn(async()=>(await e.iterator()).map(n=>B(()=>t(n))),this.size)}mapAsync(t){let e=this;return kn(async()=>(await e.iterator()).mapAsync(t),this.size)}prefetch(t){if(t==null)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let e=this;return kn(async()=>(await e.iterator()).prefetch(t),this.size)}repeat(t){let e=this,n;return this.size!=null&&t>0?n=this.size*t:t===0?n=0:this.size!=null&&(t===void 0||t<0)?n=1/0:n=null,kn(async()=>{let o=Lh(async()=>({value:await e.iterator(),done:!1}));return xR(o.take(t))},n)}skip(t){let e=this,n;return this.size!=null&&t>=0&&this.size>=t?n=this.size-t:this.size!=null&&(this.size<t||t===void 0||t<0)?n=0:n=null,kn(async()=>(await e.iterator()).skip(t),n)}shuffle(t,e,n=!0){if(t==null||t<0)throw this.size==null?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let o=this,s=bR.alea(e||y.now().toString());return kn(async()=>{let i=s.int32();return n&&(i+=s.int32()),(await o.iterator()).shuffle(t,i.toString())},this.size)}take(t){let e=this,n;return this.size!=null&&this.size>t?n=t:this.size!=null&&this.size<=t?n=this.size:n=null,kn(async()=>(await e.iterator()).take(t),n)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};Js.MAX_BUFFER_SIZE=1e4;function kn(r,t=null){return new class extends Js{constructor(){super(...arguments),this.size=t}async iterator(){return r()}}}function wR(r){return kn(async()=>AN(r),r.length)}function CR(r){if(!vu(r))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(r))for(let e=0;e<r.length;e++)t=t==null?r[e].size:Math.min(t,r[e].size);else if(r instanceof Object)for(let e in r)t=t==null?r[e].size:Math.min(t,r[e].size);return kn(async()=>{let e=await Wb(r,n=>{if(n instanceof Js)return{value:n.iterator(),recurse:!1};if(vu(n))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")});return yR(e,fl.SHORTEST)},t)}function W7(r){if(r===null)return null;let t=r[0];return dR(t)?{value:U7(r),recurse:!1}:{value:null,recurse:!0}}function U7(r){if(r.length===0)throw new Error("Can't make a batch of zero elements.");return r[0]instanceof Ft?nr(r):ur(r)}var Xf=class extends Js{constructor(t){super(),this.input=t}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`
-`).map(o=>(o.endsWith("\r")&&(o=o.slice(0,-1)),o))}};var Kb='"',Mh=Symbol("out"),IR=Symbol("field"),jb=Symbol("quote"),$N=Symbol("quoteafterquote"),SR=Symbol("quoteinquote"),Yf=class extends Js{constructor(t,e){super(),this.input=t,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new Xf(t),e||(e={}),this.hasHeader=e.hasHeader!==!1,this.fullColumnNames=e.columnNames,this.columnConfigs=e.columnConfigs,this.configuredColumnsOnly=e.configuredColumnsOnly,e.delimWhitespace?(y.assert(e.delimiter==null,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=e.delimiter?e.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let t=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!t)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&t&&y.assert(t.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+t.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=t);let e=this.fullColumnNames.reduce((o,s)=>(o[s]=o[s]+1||1,o),{}),n=Object.keys(e).filter(o=>e[o]>1);if(y.assert(n.length===0,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs){for(let o of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(o)===-1)throw new Error('The key "'+o+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error("No data was found for CSV parsing.");let n=e.value;return this.parseRow(n,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let t=await this.base.iterator();return this.hasHeader&&(t=t.skip(1)),t.map(e=>this.makeDataElement(e))}makeDataElement(t){let e=this.parseRow(t),n={},o={};for(let s=0;s<this.fullColumnNames.length;s++){let i=this.fullColumnNames[s],a=this.columnConfigs?this.columnConfigs[i]:null;if(!(this.configuredColumnsOnly&&!a)){let u=e[s],l=null;if(u==="")if(a&&a.default!==void 0)l=a.default;else{if(a&&(a.required||a.isLabel))throw new Error(`Required column ${i} is empty in this line: ${t}`);l=void 0}else{let c=Number(u);if(isNaN(c))a&&a.dtype==="bool"?l=this.getBoolean(u):l=u;else if(!a||!a.dtype)l=c;else switch(a.dtype){case"float32":l=c;break;case"int32":l=Math.floor(c);break;case"bool":l=this.getBoolean(u);break;default:l=c}}a&&a.isLabel?o[i]=l:n[i]=l}}return Object.keys(o).length===0?n:{xs:n,ys:o}}getBoolean(t){return t==="1"||t.toLowerCase()==="true"?1:0}parseRow(t,e=!0){let n=[],o=0,s=t.length,i=Mh;for(let a=0;a<s;a++)switch(i){case Mh:switch(t.charAt(a)){case Kb:o=a+1,i=jb;break;case this.delimiter:if(o=a+1,this.delimiter===" "&&this.delimWhitespace)break;n.push(""),i=Mh;break;default:i=IR,o=a;break}break;case IR:switch(t.charAt(a)){case this.delimiter:n.push(t.substring(o,a)),i=Mh,o=a+1;break;default:}break;case jb:switch(t.charAt(a)){case Kb:i=$N;break;default:}break;case $N:switch(t.charAt(a)){case this.delimiter:n.push(t.substring(o,a-1)),i=Mh,o=a+1;break;case Kb:i=jb;break;default:i=SR;break}break;case SR:switch(t.charAt(a)){case Kb:i=jb;break;default:}break;default:}if(i===$N?n.push(t.substring(o,s-1)):n.push(t.substring(o)),e&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}};var Zf=class extends Je{constructor(t){super(),this.microphoneConfig=t,this.isClosed=!1,this.fftSize=t.fftSize||1024;let e=Math.log2(this.fftSize);if(this.fftSize<0||e<4||e>14||!Number.isInteger(e))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=t.numFramesPerSpectrogram||43,this.sampleRateHz=t.sampleRateHz,this.columnTruncateLength=t.columnTruncateLength||this.fftSize,this.audioTrackConstraints=t.audioTrackConstraints,this.smoothingTimeConstant=t.smoothingTimeConstant||0,this.includeSpectrogram=t.includeSpectrogram!==!1,this.includeWaveform=t.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(t={}){if(!z().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let e=new Zf(t);return await e.start(),e}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(n){throw new Error(`Error thrown while initializing video stream: ${n.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let t=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new t,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let e=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,e.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let t,e,n=await this.getAudioData();if(this.includeSpectrogram){let o=this.flattenQueue(n.freqDataQueue);t=this.getTensorFromAudioDataArray(o,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let o=this.flattenQueue(n.timeDataQueue);e=this.getTensorFromAudioDataArray(o,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:t,waveform:e},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let t=[],e=[],n=0;return new Promise(o=>{let s=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&o({freqDataQueue:t,timeDataQueue:e}),t.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),e.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(s),o({freqDataQueue:t,timeDataQueue:e}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(t){let e=t[0].length,n=new Float32Array(t.length*e);return t.forEach((o,s)=>n.set(o,s*e)),n}getTensorFromAudioDataArray(t,e){let n=new Float32Array(y.sizeFromShape(e));return n.set(t,n.length-t.length),ur(n,e)}};var Jf=class extends Je{constructor(t,e){if(super(),this.webcamVideoElement=t,this.webcamConfig=e,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Me([0],"int32"),this.webcamConfig.centerCrop){let n=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,o=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,s=(1-n)/2,i=(1-o)/2,a=s+n,u=o+i;this.cropBox=Vs([i,s,u,a],[1,4])}else this.cropBox=Vs([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(t,e={}){if(!z().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!t){if(t=document.createElement("video"),!e.resizeWidth||!e.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");t.width=e.resizeWidth,t.height=e.resizeHeight}let n=new Jf(t,e);return await n.start(),n}async start(){this.webcamConfig.facingMode&&y.assert(this.webcamConfig.facingMode==="user"||this.webcamConfig.facingMode==="environment",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(t){throw t.message=`Error thrown while initializing video stream: ${t.message}`,t}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(t){console.log(t),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(t=>{this.webcamVideoElement.onloadedmetadata=()=>{t()}})}async next(){if(this.isClosed)return{value:null,done:!0};let t;try{t=nx.fromPixels(this.webcamVideoElement)}catch(e){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(t),done:!1}}catch(e){throw new Error(`Error thrown cropping the video: ${e.message}`)}finally{t.dispose()}else return{value:t,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(t){return B(()=>{let e=rr(J(t,"float32"),0),n;n=Gs.cropAndResize(e,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let o=n.shape;return R(n,o.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}};var Qf=class{};var zh=class extends Je{split(t){return new DN(this,t)}},DN=class extends zh{constructor(t,e){super(),this.upstream=t,this.impl=new RN(t,e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},RN=class extends kc{constructor(t,e){super(),this.upstream=t,this.separator=e,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let t=await this.upstream.next();if(t.done)return this.carryover===""?!1:(this.outputQueue.push(this.carryover),this.carryover="",!0);let e=t.value.split(this.separator);e[0]=this.carryover+e[0];for(let n of e.slice(0,-1))this.outputQueue.push(n);return this.carryover=e[e.length-1],!0}};var Xb=class extends Je{decodeUTF8(){return new FN(this)}},FN=class extends zh{constructor(t){super(),this.upstream=t,this.impl=new ON(t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},ON=class extends kc{constructor(t){if(super(),this.upstream=t,z().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:e}=gN();this.decoder=new e("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let t=await this.upstream.next(),e;if(t.done)return!1;e=t.value;let n;return z().get("IS_BROWSER")?n=this.decoder.decode(e,{stream:!0}):n=this.decoder.write(Buffer.from(e.buffer)),this.outputQueue.push(n),!0}};var td=class extends Xb{constructor(t,e={}){super(),this.file=t,this.options=e,y.assert(t instanceof Uint8Array||(z().get("IS_BROWSER")?t instanceof File||t instanceof Blob:!1),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=e.offset||0,this.chunkSize=e.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,n)=>{let o=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,o)));else{let s=new FileReader;s.onload=a=>{let u=s.result;if(u instanceof ArrayBuffer&&(u=new Uint8Array(u)),!(u instanceof Uint8Array))return n(new TypeError("FileReader returned unknown type."));e(u)},s.onabort=a=>n(new Error("Aborted")),s.onerror=a=>n(new Error(a.type));let i=this.file.slice(this.offset,o);s.readAsArrayBuffer(i)}this.offset=o}),done:!1}}};async function vR(r,t={},e){let n,o;typeof r=="string"?n=r:(n=r.url,o=H7(r));let s=await(e||y.fetch)(n,o);if(s.ok){let i=new Uint8Array(await s.arrayBuffer());return new td(i,t)}else throw new Error(s.statusText)}var H7=r=>({method:r.method,headers:r.headers,body:r.body,mode:r.mode,credentials:r.credentials,cache:r.cache,redirect:r.redirect,referrer:r.referrer,integrity:r.integrity});function Yb(r){return typeof r=="string"&&r.slice(0,7)==="file://"}var ed=class extends Qf{constructor(t,e={}){super(),this.input=t,this.options=e}async iterator(){if(Yb(this.input)&&z().get("IS_NODE")){let t=Zb();this.input=t.readFileSync(this.input.slice(7))}return new td(this.input,this.options)}};var rd=class extends Qf{constructor(t,e={}){super(),this.url=t,this.fileOptions=e}async iterator(){return Yb(this.url)?new ed(this.url,this.fileOptions).iterator():vR(this.url,this.fileOptions)}};function NR(r,t={}){return new Yf(new rd(r),t)}function TR(r){let t=Lh(r);return kn(async()=>t)}function kR(r){return kn(async()=>{let t=await r();return Lh(()=>t.next())})}async function ER(r,t){return Jf.create(r,t)}async function _R(r){return Zf.create(r)}var PN="4.0.0";function tt(r,t){Array.isArray(r)||(r=[r]),r.forEach(e=>{e!=null&&y.assert(e.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var q7=Ur.whereImpl,Nu=class extends zo{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new ra(this,Pn())}nextDataId(){return Nu.nextDataId++}write(t,e,n){this.firstUse&&(this.firstUse=!1,z().get("IS_NODE")&&v.warn(`
-============================
-Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. 
-============================`));let o={id:this.nextDataId()};return this.data.set(o,{values:t,dtype:n,refCount:1}),o}makeTensorInfo(t,e,n){let o;if(e==="string"&&n!=null&&n.length>0&&y.isString(n[0])){let s=n.map(i=>y.encodeString(i));o=this.write(s,t,e)}else o=this.write(n,t,e);return{dataId:o,shape:t,dtype:e}}refCount(t){return this.data.has(t)?this.data.get(t).refCount:0}incRef(t){let e=this.data.get(t);e.refCount++}decRef(t){if(this.data.has(t)){let e=this.data.get(t);e.refCount--}}move(t,e,n,o,s){this.data.set(t,{values:e,dtype:o,refCount:s})}numDataIds(){return this.data.numDataIds()}async read(t){return this.readSync(t)}readSync(t){let{dtype:e,complexTensorInfos:n}=this.data.get(t);if(e==="complex64"){let o=this.readSync(n.real.dataId),s=this.readSync(n.imag.dataId);return v.mergeRealAndImagArrays(o,s)}return this.data.get(t).values}bufferSync(t){let e=this.readSync(t.dataId);if(t.dtype==="string")try{let n=e.map(o=>y.decodeString(o));return wt(t.shape,t.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return wt(t.shape,t.dtype,e)}makeOutput(t,e,n){return Pn().makeTensorFromTensorInfo(this.makeTensorInfo(e,n,t),this)}disposeData(t,e=!1){if(this.data.has(t)){if(this.data.get(t).refCount--,!e&&this.data.get(t).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(t);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(t)}return!0}disposeIntermediateTensorInfo(t){this.disposeData(t.dataId)}async time(t){let e=y.now();return t(),{kernelMs:y.now()-e}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(t){tt([t],"where");let e=this.readSync(t.dataId);return q7(t.shape,e)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};Nu.nextDataId=0;var mw={};Wt(mw,{addImpl:()=>zN,bincountImpl:()=>sd,bincountReduceImpl:()=>Jb,castImpl:()=>MN,ceilImpl:()=>BN,concatImpl:()=>Ec,equalImpl:()=>VN,expImpl:()=>WN,expm1Impl:()=>HN,floorImpl:()=>qN,gatherNdImpl:()=>Qb,gatherV2Impl:()=>tw,greaterEqualImpl:()=>jN,greaterImpl:()=>KN,lessEqualImpl:()=>YN,lessImpl:()=>XN,linSpaceImpl:()=>ew,logImpl:()=>ZN,maxImpl:()=>rw,maximumImpl:()=>JN,minimumImpl:()=>QN,multiplyImpl:()=>Bh,negImpl:()=>tT,notEqualImpl:()=>eT,prodImpl:()=>rT,raggedGatherImpl:()=>nw,raggedRangeImpl:()=>ow,raggedTensorToTensorImpl:()=>sw,rangeImpl:()=>Ac,rsqrtImpl:()=>nT,scatterImpl:()=>dl,sigmoidImpl:()=>sF,simpleAbsImpl:()=>LN,sliceImpl:()=>$c,sparseFillEmptyRowsImpl:()=>iw,sparseReshapeImpl:()=>aw,sparseSegmentReductionImpl:()=>ld,sqrtImpl:()=>lF,squaredDifferenceImpl:()=>sT,stridedSliceImpl:()=>lw,stringNGramsImpl:()=>Dc,stringSplitImpl:()=>Rc,stringToHashBucketFastImpl:()=>Fc,subImpl:()=>aT,tileImpl:()=>uw,topKImpl:()=>cw,transposeImpl:()=>id,uniqueImpl:()=>pw});function LN(r){let t=new Float32Array(r.length);for(let e=0;e<r.length;++e)t[e]=Math.abs(r[e]);return t}var K7=r=>{let{x:t}=r.inputs,e=r.backend;tt(t,"abs");let n=new Float32Array(y.sizeFromShape(t.shape)),o=e.data.get(t.dataId).values;return n=LN(o),e.makeOutput(n,t.shape,t.dtype)},$R={kernelName:ii,backendName:"cpu",kernelFunc:K7};function Qt(r){return(t,e,n,o,s)=>{let i=v.assertAndGetBroadcastShape(t,e),a=i.length,u=y.computeStrides(i),l=y.sizeFromShape(i),c=y.getTypedArrayFromDType(s,l),p=t.length,m=e.length,f=y.computeStrides(t),d=y.computeStrides(e),h=v.getBroadcastDims(t,i),g=v.getBroadcastDims(e,i);if(h.length+g.length===0)for(let x=0;x<c.length;++x)c[x]=r(n[x%n.length],o[x%o.length]);else for(let x=0;x<c.length;++x){let b=y.indexToLoc(x,a,u),w=b.slice(-p);h.forEach(A=>w[A]=0);let C=y.locToIndex(w,p,f),N=b.slice(-m);g.forEach(A=>N[A]=0);let _=y.locToIndex(N,m,d);c[x]=r(n[C],o[_])}return[c,i]}}function wr(r){let{inputs:t,backend:e}=r,{real:n,imag:o}=t,s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,a=e.makeTensorInfo(n.shape,"complex64"),u=e.data.get(a.dataId);return u.complexTensorInfos={real:e.makeTensorInfo(n.shape,"float32",s),imag:e.makeTensorInfo(o.shape,"float32",i)},a}var DR={kernelName:pp,backendName:"cpu",kernelFunc:wr};function nd(r,t,e="float32"){if(e==="complex64"){let o=nd(r,t,"float32"),s=nd(r,t,"float32");return wr({inputs:{real:o,imag:s},backend:r})}let n=y.makeZerosTypedArray(y.sizeFromShape(t),e);return r.makeTensorInfo(t,e,n)}function Kr(r){let{inputs:t,backend:e}=r,{x:n}=t;return e.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var RR={kernelName:co,backendName:"cpu",kernelFunc:Kr};function Ao(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.data.get(n.dataId).complexTensorInfos.real,s=e.data.get(o.dataId).values;return e.makeTensorInfo(o.shape,o.dtype,s)}var FR={kernelName:Rp,backendName:"cpu",kernelFunc:Ao};function MN(r,t,e,n){if(n==="int32"){let o=Int32Array.from(r);return[t,"int32",o]}if(n==="bool"){let o=y.toTypedArray([0],e),[s,i]=Qt((a,u)=>a!==u?1:0)(t,[],r,o,"bool");return[i,"bool",s]}throw new Error(`Error in Cast: failed to cast ${e} to ${n}`)}function $o(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dtype:s}=n;if(s==="complex64"){if(o.dtype==="complex64")return Kr({inputs:{x:o},backend:e});let c=nd(e,o.shape,o.dtype),p=$o({inputs:{x:o},backend:e,attrs:{dtype:"float32"}}),m=wr({inputs:{real:p,imag:c},backend:e});return e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),m}if(o.dtype==="complex64"){let c=Ao({inputs:{input:o},backend:e}),p=$o({inputs:{x:c},backend:e,attrs:{dtype:s}});return e.disposeIntermediateTensorInfo(c),p}if(!y.hasEncodingLoss(o.dtype,s)){let c=Kr({inputs:{x:o},backend:e});return{dataId:c.dataId,shape:c.shape,dtype:s}}let i=e.data.get(o.dataId).values,[a,u,l]=MN(i,o.shape,o.dtype,s);return e.makeTensorInfo(a,u,l)}var OR={kernelName:lo,backendName:"cpu",kernelFunc:$o};function oe(r,t,e,n){return e==null?({inputs:o,backend:s})=>{let{a:i,b:a}=o,u=s;tt([i,a],r);let l=u.data.get(i.dataId).values,c=u.data.get(a.dataId).values,p=i.dtype==="string"?v.fromUint8ToStringArray(l):l,m=i.dtype==="string"?v.fromUint8ToStringArray(c):c,f=n||i.dtype,[d,h]=t(i.shape,a.shape,p,m,f);return u.makeTensorInfo(h,f,d)}:({inputs:o,backend:s})=>{let{a:i,b:a}=o,u=s;if(i.dtype==="complex64"||a.dtype==="complex64"){let l=$o({inputs:{x:i},backend:u,attrs:{dtype:"complex64"}}),c=u.data.get(l.dataId),p=c.complexTensorInfos.real,m=c.complexTensorInfos.imag,f=u.data.get(p.dataId).values,d=u.data.get(m.dataId).values,h=$o({inputs:{x:a},backend:u,attrs:{dtype:"complex64"}}),g=u.data.get(h.dataId),x=g.complexTensorInfos.real,b=g.complexTensorInfos.imag,w=u.data.get(x.dataId).values,C=u.data.get(b.dataId).values,[N,_,A]=e(i.shape,a.shape,f,d,w,C),$=u.makeTensorInfo(A,"float32",N),F=u.makeTensorInfo(A,"float32",_),P=wr({inputs:{real:$,imag:F},backend:u});return u.disposeIntermediateTensorInfo(l),u.disposeIntermediateTensorInfo(h),u.disposeIntermediateTensorInfo($),u.disposeIntermediateTensorInfo(F),P}else{let l=u.data.get(i.dataId).values,c=u.data.get(a.dataId).values,p=n||i.dtype,[m,f]=t(i.shape,a.shape,l,c,p);return u.makeTensorInfo(f,p,m)}}}function od(r){return(t,e,n,o,s,i)=>{let a=v.assertAndGetBroadcastShape(t,e),u=y.sizeFromShape(a),l=a.length,c=y.computeStrides(a),p=y.getTypedArrayFromDType("float32",u),m=y.getTypedArrayFromDType("float32",u),f=v.getBroadcastDims(t,a),d=v.getBroadcastDims(e,a),h=v.mergeRealAndImagArrays(n,o),g=v.mergeRealAndImagArrays(s,i),x=t.length,b=y.computeStrides(t),w=e.length,C=y.computeStrides(e);if(f.length+d.length===0)for(let N=0;N<p.length;N++){let _=N%h.length,A=N%g.length,$=r(h[_*2],h[_*2+1],g[A*2],g[A*2+1]);p[N]=$.real,m[N]=$.imag}else for(let N=0;N<p.length;N++){let _=y.indexToLoc(N,l,c),A=_.slice(-x);f.forEach(G=>A[G]=0);let $=y.locToIndex(A,x,b),F=_.slice(-w);d.forEach(G=>F[G]=0);let P=y.locToIndex(F,w,C),V=r(h[$*2],h[$*2+1],g[P*2],g[P*2+1]);p[N]=V.real,m[N]=V.imag}return[p,m,a]}}var zN=Qt((r,t)=>r+t),j7=od((r,t,e,n)=>({real:r+e,imag:t+n})),Ki=oe(Zn,zN,j7),PR={kernelName:Zn,backendName:"cpu",kernelFunc:Ki};function sd(r,t,e,n,o){let s=y.sizeFromShape(n),i=y.makeZerosTypedArray(o,e);for(let a=0;a<r.length;a++){let u=r[a];if(u<0)throw new Error("Input x must be non-negative!");u>=o||(s>0?i[u]+=t[a]:i[u]+=1)}return i}function Jb(r,t,e,n=!1){let o=r.shape[0],s=r.shape[1],i=wt([o,e],t.dtype);for(let a=0;a<o;a++)for(let u=0;u<s;u++){let l=r.get(a,u);if(l<0)throw new Error("Input x must be non-negative!");l>=e||(n?i.set(1,a,l):t.size>0?i.set(i.get(a,l)+t.get(a,u),a,l):i.set(i.get(a,l)+1,a,l))}return i}function yn(r){return(t,e,n)=>{let o=y.getTypedArrayFromDType(e,t.length);for(let s=0;s<t.length;++s)o[s]=r(t[s],n);return o}}function kt(r,t,e){return({inputs:n,attrs:o,backend:s})=>{let{x:i}=n;if(tt(i,r),i.dtype==="string"||e==="string")throw new Error("unaryKernelFunc does not support string input/output");let a=s,u=a.data.get(i.dataId).values,l=y.sizeFromShape(i.shape),c=e||i.dtype,p=y.getArrayFromDType(c,l);for(let m=0;m<l;++m)p[m]=t(u[m],o);return a.makeTensorInfo(i.shape,c,p)}}function Do(r,t,e){return({inputs:n,attrs:o,backend:s})=>{let{x:i}=n;if(tt(i,r),i.dtype==="string"||e==="string")throw new Error("unaryKernelFunc does not support string input/output");let a=s,u=a.data.get(i.dataId).values,l=e||i.dtype,c=t(u,l,o);return a.makeTensorInfo(i.shape,l,c)}}var BN=yn(r=>Math.ceil(r)),X7=Do(qo,BN),LR={kernelName:qo,backendName:"cpu",kernelFunc:X7};function Ec(r,t,e,n){let o=y.getArrayFromDType(e,y.sizeFromShape(t));if(n&&e!=="string"){let s=0;r.forEach(i=>{let a=y.sizeFromShape(i.shape);o.set(i.vals,s),s+=a})}else{let s=0;r.forEach(i=>{let a=e==="string"?v.fromUint8ToStringArray(i.vals):i.vals,u=0;for(let l=0;l<i.shape[0];++l){let c=l*t[1]+s;for(let p=0;p<i.shape[1];++p)o[c+p]=a[u++]}s+=i.shape[1]})}return o}var VN=Qt((r,t)=>r===t?1:0),GN=oe(xa,VN,null,"bool"),MR={kernelName:xa,backendName:"cpu",kernelFunc:GN};var WN=yn(r=>Math.exp(r)),UN=Do(es,WN,"float32"),zR={kernelName:es,backendName:"cpu",kernelFunc:UN};var HN=yn(r=>Math.expm1(r)),Y7=Do(ya,HN),BR={kernelName:ya,backendName:"cpu",kernelFunc:Y7};var qN=yn(r=>Math.floor(r)),Z7=Do(rs,qN),VR={kernelName:rs,backendName:"cpu",kernelFunc:Z7};function Qb(r,t,e,n,o,s,i,a,u){let l=wt([n,s],e);for(let c=0;c<n;c++){let p=[],m=0;for(let f=0;f<o;f++){let d=r[c*o+f];m+=d*i[f],p.push(d)}if(m<0||m>=u/s)throw new Error(`Invalid indices: ${p} does not index into ${a}`);for(let f=0;f<s;f++)l.values[c*s+f]=t.get(...t.indexToLoc(m*s+f))}return l}function tw(r,t,e){let n=wt(e,r.dtype);for(let o=0;o<n.size;++o){let i=n.indexToLoc(o).slice(),a=i[0],u=i[2],l=t.locToIndex([a,u]);i[2]=t.values[l];let c=r.locToIndex(i);0<=c&&c<r.values.length&&(n.values[o]=r.values[c])}return n}var KN=Qt((r,t)=>r>t?1:0),J7=oe(Ca,KN,null,"bool"),GR={kernelName:Ca,backendName:"cpu",kernelFunc:J7};var jN=Qt((r,t)=>r>=t?1:0),Q7=oe(ss,jN,null,"bool"),WR={kernelName:ss,backendName:"cpu",kernelFunc:Q7};var XN=Qt((r,t)=>r<t?1:0),tJ=oe(Na,XN,null,"bool"),UR={kernelName:Na,backendName:"cpu",kernelFunc:tJ};var YN=Qt((r,t)=>r<=t?1:0),eJ=oe(Ta,YN,null,"bool"),HR={kernelName:Ta,backendName:"cpu",kernelFunc:eJ};function ew(r,t,e){let n=(t-r)/(e-1),o=y.makeZerosTypedArray(e,"float32");o[0]=r;for(let s=1;s<o.length;s++)o[s]=o[s-1]+n;return o}var ZN=yn(r=>Math.log(r)),rJ=Do(as,ZN),qR={kernelName:as,backendName:"cpu",kernelFunc:rJ};function rw(r,t,e,n){let o=y.getTypedArrayFromDType(n,y.sizeFromShape(e));for(let s=0;s<o.length;++s){let i=s*t,a=r[i];for(let u=0;u<t;++u){let l=r[i+u];(Number.isNaN(l)||l>a)&&(a=l)}o[s]=a}return o}var JN=Qt((r,t)=>Math.max(r,t)),nJ=oe(us,JN),KR={kernelName:us,backendName:"cpu",kernelFunc:nJ};var QN=Qt((r,t)=>Math.min(r,t)),oJ=oe(fs,QN),jR={kernelName:fs,backendName:"cpu",kernelFunc:oJ};var Bh=Qt((r,t)=>r*t),sJ=od((r,t,e,n)=>({real:r*e-t*n,imag:r*n+t*e})),_c=oe(hs,Bh,sJ),XR={kernelName:hs,backendName:"cpu",kernelFunc:_c};function tT(r,t,e){let n=y.createScalarValue(-1,e);return Bh([],t,n,r,e)}function iJ(r){let{inputs:t,backend:e}=r,{x:n}=t;tt(n,"neg");let o=e.data.get(n.dataId).values,[s,i]=tT(o,n.shape,n.dtype);return e.makeTensorInfo(i,n.dtype,s)}var YR={kernelName:pi,backendName:"cpu",kernelFunc:iJ};var eT=Qt((r,t)=>r!==t?1:0),aJ=oe(Da,eT,null,"bool"),ZR={kernelName:Da,backendName:"cpu",kernelFunc:aJ};function id(r,t,e,n,o){let s=t.length,i=y.sizeFromShape(t),a=y.computeStrides(t),u=y.computeStrides(o),l=y.getTypedArrayFromDType(e,y.sizeFromShape(o));for(let c=0;c<i;++c){let p=y.indexToLoc(c,s,a),m=new Array(p.length);for(let d=0;d<m.length;d++)m[d]=p[n[d]];let f=y.locToIndex(m,s,u);l[f]=r[c]}return l}function Ve(r){let{inputs:t,attrs:e,backend:n}=r,{x:o}=t,{perm:s}=e;tt(o,"transpose");let i=o.shape.length,a=new Array(i);for(let p=0;p<a.length;p++)a[p]=o.shape[s[p]];let u=n.data.get(o.dataId).values,l=id(u,o.shape,o.dtype,s,a);return{dataId:n.write(l,a,o.dtype),shape:a,dtype:o.dtype}}var JR={kernelName:Qn,backendName:"cpu",kernelFunc:Ve};function rT(r,t,e,n){let[o,s]=v.computeOutAndReduceShapes(r,n),i=sr(t,"int32"),a=y.makeZerosTypedArray(y.sizeFromShape(o),i),u=y.sizeFromShape(s);for(let l=0;l<a.length;++l){let c=l*u,p=1;for(let m=0;m<u;++m)p*=e[c+m];a[l]=p}return{outVals:a,outShape:o,outDtype:i}}function lJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,"prod");let a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=v.getAxesPermutation(u,a),c=u,p=o,m=[];l!=null&&(p=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),m.push(p),c=v.getInnerMostAxes(c.length,a));let f=e.data.get(p.dataId).values,{outVals:d,outShape:h,outDtype:g}=rT(p.shape,p.dtype,f,c),x=h;return i&&(x=v.expandShapeToKeepDim(h,u)),m.forEach(b=>e.disposeIntermediateTensorInfo(b)),e.makeTensorInfo(x,g,d)}var QR={kernelName:ws,backendName:"cpu",kernelFunc:lJ};function uJ(r,t,e){r.forEach((n,o)=>{if(n<0||n>=e){let s=y.indexToLoc(o,t.length,y.computeStrides(t)).join(",");throw new Error(`indices[${s}] = ${n} is not in [0, ${e})`)}})}function cJ(r,t){for(let e=0;e<r.length;++e){let n=r[e],o=e===r.length-1?t:r[e+1].length;if(n.length===0)throw new Error("Ragged splits may not be empty");if(n[0]<0)throw new Error("Ragged splits must be non-negative");if(n[n.length-1]>o)throw new Error("Ragged splits must not point past values");for(let s=1;s<n.length;++s)if(n[s-1]>n[s])throw new Error("Ragged splits must be sorted in ascending order")}}function pJ(r,t,e,n){let o=[],s=0,i=t.length-1+e.length,a=new Array(i).fill(null).map(()=>[0]);cJ(e,n);let u=1;for(let l=0;l<t.length-1;++l){u*=t[l];let c=t[l+1];for(let p=1;p<u+1;++p)a[l].push(p*c)}for(let l=0;l<r.length;++l){let c=r[l],p=r[l]+1;for(let m=0;m<e.length;++m){let f=e[m],d=m+t.length-1;if(d>=0){let h=a[d],g=h[h.length-1]-f[c];for(let x=c;x<p;++x)a[d].push(f[x+1]+g)}c=f[c],p=f[p]}p!==c&&(o.push([c,p]),s+=p-c)}return{outSplits:a,valueSlices:o,numValues:s}}function mJ(r){let t=[];for(let e=0;e<r.length;++e){let n=r[e].length,o=y.getArrayFromDType("int32",n);t.push(o),r[e].forEach((s,i)=>o[i]=s)}return t}function tF(r,t){let e=r.slice(0,t);for(;e.length<t;)e.push(1);for(let n=t;n<r.length;n++)e[t-1]*=r[n];return e}function fJ(r,t,e,n,o,s){let i=tF(t,2)[1],a=tF(s,2)[1],u=0;for(let l of e)for(let c=l[0];c<l[1];++c){for(let p=0;p<n;++p)o[u*a+p]=r[c*i+p];++u}}function dJ(r,t,e,n,o){let s=t.slice();s[0]=o;let i=y.getArrayFromDType(e,y.sizeFromShape(s)),a=r.length,u=a===0?0:a/t[0];return fJ(r,t,n,u,i,s),[i,s]}function nw(r,t,e,n,o,s,i,a){if(r.length===0)throw new Error("paramsNestedSplits must be non empty");if(t[0].length===0)throw new Error("Split tensors must not be scalars");let u=t[0][0]-1;if(uJ(s,i,u),n.length===0)throw new Error("params.rank must be nonzero");let l=n[0],{outSplits:c,valueSlices:p,numValues:m}=pJ(s,i,r,l),f=mJ(c),d=dJ(e,n,o,p,m);return[f,d[0],d[1]]}var eF=2147483647;function ow(r,t,e,n,o,s,i){if(t.length>1)throw new Error("starts must be a scalar or vector");if(o.length>1)throw new Error("limits must be a scalar or vector");if(i.length>1)throw new Error("deltas must be a scalar or vector");let a=t.length===0,u=o.length===0,l=i.length===0,c=[];a||c.push(t[0]),u||c.push(o[0]),l||c.push(i[0]);for(let g=1;g<c.length;++g)if(c[g]!==c[g-1])throw new Error("starts, limits, and deltas must have the same shape");let p=c.length===0?1:c[0],m=y.getArrayFromDType("int32",p+1);m[0]=0;for(let g=0;g<p;++g){let x=a?r[0]:r[g],b=u?n[0]:n[g],w=l?s[0]:s[g];if(w===0)throw new Error("Requires delta != 0");let C;if(w>0&&b<x||w<0&&b>x)C=0;else if(C=Math.ceil(Math.abs((b-x)/w)),C>eF)throw new Error(`Requires ((limit - start) / delta) <= ${eF}`);m[g+1]=m[g]+C}let f=m[p],d=y.getArrayFromDType(e,f),h=0;for(let g=0;g<p;++g){let x=m[g+1]-m[g],b=a?r[0]:r[g],w=l?s[0]:s[g];for(let C=0;C<x;++C)d[h++]=b,b+=w}return[m,d]}var Ro=v.RowPartitionType,ad=class{constructor(t,e,n,o,s,i,a,u,l,c){this.shape=t,this.shapeShape=e,this.values=n,this.valuesShape=o,this.valuesDType=s,this.defaultValue=i,this.defaultValueShape=a,this.rowPartitionValues=u,this.rowPartitionValuesShapes=l,this.rowPartitionTypes=v.getRowPartitionTypesHelper(c),this.raggedRank=v.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(t){return this.rowPartitionTypes[0]===Ro.FIRST_DIM_SIZE?this.rowPartitionTypes[t+1]:this.rowPartitionTypes[t]}getRowPartitionTensor(t){return this.rowPartitionTypes[0]===Ro.FIRST_DIM_SIZE?this.rowPartitionValues[t+1]:this.rowPartitionValues[t]}getMaxWidth(t){let e=this.getRowPartitionTensor(t-1);switch(this.getRowPartitionTypeByDimension(t-1)){case Ro.VALUE_ROWIDS:return ad.getMaxWidthValueRowID(e);case Ro.ROW_SPLITS:return ad.getMaxWidthRowSplit(e);default:throw new Error(`Cannot handle partition type ${Ro[this.getRowPartitionTypeByDimension(t-1)]}`)}}static getMaxWidthRowSplit(t){let e=t.length;if(e===0||e===1)return 0;let n=0;for(let o=0;o<e-1;++o){let s=t[o+1]-t[o];s>n&&(n=s)}return n}static getMaxWidthValueRowID(t){let e=t.length;if(e===0)return 0;let n=0,o=t[0],s=0;for(let i=1;i<e;++i){let a=t[i];a!==o&&(o=a,s=Math.max(i-n,s),n=i)}return Math.max(e-n,s)}tensorShapeFromTensor(t,e,n=!0){if(e.length===0){if(t[0]===-1)return[];throw new Error("The only valid scalar shape tensor is the fully unknown shape specified as -1.")}return nF(t,n)}calculateOutputSize(t){let e=this.valuesShape,n=this.defaultValueShape;v.validateDefaultValueShape(n,e);let o=this.tensorShapeFromTensor(this.shape,this.shapeShape),i=v.combineRaggedTensorToTensorShapes(this.raggedRank,o,e);i[0]<0&&(i[0]=t);for(let a=1;a<=this.raggedRank;++a)i[a]<0&&(i[a]=this.getMaxWidth(a));return i}calculateFirstParentOutputIndex(t,e,n){let o=Math.min(t,n),s=[],i=0;for(let a=0;a<o;++a,i+=e)s.push(i);for(let a=o;a<t;++a)s.push(-1);return y.assert(s.length===t,()=>"Final length of result must be equal to firstDimension."),s}calculateOutputIndexRowSplit(t,e,n,o){let s=t.length,i=[];for(let a=0;a<s-1;++a){let u=t[a+1]-t[a],l=Math.min(o,u),c=e[a];c===-1&&(l=0);for(let p=0;p<l;++p)i.push(c),c+=n;for(let p=0;p<u-l;++p)i.push(-1)}if(s>0&&i.length!==t[s-1])throw new Error("Invalid row split size.");return i}calculateOutputIndexValueRowID(t,e,n,o){let s=t.length,i=[];if(s===0)return[];let a=0,u=t[0];if(u>=e.length)throw new Error(`Got currentValueRowId=${u}, which is not less than ${e.length}`);let l=e[u];i.push(l);for(let c=1;c<s;++c){let p=t[c];if(p===u)l>=0&&(++a,a<o?l+=n:l=-1);else{if(a=0,u=p,p>=e.length)throw new Error(`Got nextValueRowId=${p} which is not less than ${e.length}`);l=e[p]}i.push(l)}if(i.length!==t.length)throw new Error("Invalid row ids.");return i}calculateOutputIndex(t,e,n,o){let s=this.getRowPartitionTensor(t),i=this.getRowPartitionTypeByDimension(t);switch(i){case Ro.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(s,e,n,o);case Ro.ROW_SPLITS:if(s.length-1>e.length)throw new Error(`Row partition size is greater than output size: ${s.length-1} > ${e.length}`);return this.calculateOutputIndexRowSplit(s,e,n,o);default:throw new Error(`Unsupported partition type: ${Ro[i]}`)}}getFirstDimensionSize(){let t=this.rowPartitionValues[0];if(this.rowPartitionTypes.length===0)throw new Error("No row_partition_types given.");let e=this.rowPartitionTypes[0];switch(e){case Ro.FIRST_DIM_SIZE:return t[0];case Ro.VALUE_ROWIDS:throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");case Ro.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${Ro[e]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error("Invalid first partition input. Tensor requires at least one element.");let e=this.getFirstDimensionSize(),n=this.calculateOutputSize(e),o=new Array(this.raggedRank+1);o[o.length-1]=1;for(let u=o.length-2;u>=0;--u)o[u]=o[u+1]*n[u+1];let s=nF(n,!1),i=y.getArrayFromDType(this.valuesDType,y.sizeFromShape(s));if(o[0]*n[0]>0){let u=this.calculateFirstParentOutputIndex(e,o[0],n[0]);for(let l=1;l<=this.raggedRank;++l)u=this.calculateOutputIndex(l-1,u,o[l],n[l]);this.setOutput(this.raggedRank,u,i,s)}return[s,i]}setOutput(t,e,n,o){if(n.length===0)return;let s=this.values,i=n,a=o.slice();a=a.slice(t+1);let u=y.sizeFromShape(a),l=e.length,c=this.defaultValue;if(c.length!==u&&c.length!==1){let d=this.defaultValueShape;B(()=>{let h=R(c,d);c=Ri(h,a).dataSync()})}let p=0,m=0,f=0;for(let d=0;d<=l;++d){let h=d<l?e[d]:-1;if(h===f){++f;continue}if(m<f){let g=s.subarray(p*u),x=i.subarray(m*u),b=(f-m)*u;rF(x,g,b)}if(d>=l){let g=n.length;h=Math.floor(g/u)}if(h>f)if(this.defaultValue.length===1)i.subarray(f*u,h*u).fill(this.defaultValue[0]),f=h;else for(;h>f;){let g=i.slice(f*u);rF(g,c,u),++f}h<0?(p=d+1,m=f):(p=d,m=f,f=m+1)}}};function rF(r,t,e){for(let n=0;n<e;n++)r[n]=t[n]}function nF(r,t){let e=[];for(let n of r){if(n<0){if(!t)throw new Error(`Dimension ${n} must be >= 0`);if(n<-1)throw new Error(`Dimension ${n} must be >= -1`);n=-1}e.push(n)}return e}function sw(r,t,e,n,o,s,i,a,u,l){return new ad(r,t,e,n,o,s,i,a,u,l).compute()}function Ac(r,t,e,n){let o=r===t,s=r<t&&e<0,i=t<r&&e>1;if(o||s||i)return y.makeZerosTypedArray(0,n);let a=Math.abs(Math.ceil((t-r)/e)),u=y.makeZerosTypedArray(a,n);t<r&&e===1&&(e=-1),u[0]=r;for(let l=1;l<u.length;l++)u[l]=u[l-1]+e;return u}var nT=yn(r=>1/Math.sqrt(r)),hJ=Do(ks,nT),oF={kernelName:ks,backendName:"cpu",kernelFunc:hJ};function dl(r,t,e,n,o,s,i,a,u,l){let c=[n/o,o],p=r.values,m=t.values;if(n===0)return wt(e,t.dtype);let f=wt(c,t.dtype);typeof u=="string"||typeof u=="number"?f.values.fill(u):typeof u=="boolean"&&f.values.fill(+u);for(let d=0;d<s;d++){let h=[],g=0;for(let x=0;x<i;x++){let b=p[d*i+x];h.push(b),g+=b*a[x]}if(g<0||g>=n/o)throw new Error(`Invalid indices: ${h} does not index into ${e}`);for(let x=0;x<o;x++)l?f.values[g*o+x]+=m[d*o+x]:f.values[g*o+x]=t.rank===0?m[0]:m[d*o+x]}return f}var sF=yn(r=>1/(1+Math.exp(-r))),oT=kt(_s,r=>1/(1+Math.exp(-r))),iF={kernelName:_s,backendName:"cpu",kernelFunc:oT};function $c(r,t,e,n,o){let s=Le.isSliceContinous(n,t,e),i=y.sizeFromShape(e),a=y.computeStrides(n);if(s){let p=Le.computeFlatOffset(t,a);return o==="string"?r.slice(p,p+i):r.subarray(p,p+i)}let u=o==="string"?v.fromUint8ToStringArray(r):r,l=wt(n,o,u),c=wt(e,o);for(let p=0;p<c.size;++p){let m=c.indexToLoc(p),f=m.map((d,h)=>d+t[h]);c.set(l.get(...f),...m)}return o==="string"?v.fromStringArrayToUint8(c.values):c.values}function Fo(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,size:i}=n;tt(o,"slice");let[a,u]=Le.parseSliceParams(o,s,i);Le.assertParamsValid(o,a,u);let l=e.data.get(o.dataId).values,c=$c(l,a,u,o.shape,o.dtype);return e.makeTensorInfo(u,o.dtype,c)}var aF={kernelName:gi,backendName:"cpu",kernelFunc:Fo};function iw(r,t,e,n,o,s,i){let a=t[0],u=s[0],l=new Array(u),c=new Array(a),p=t[1];if(u===0){if(a!==0)throw new Error(v.getSparseFillEmptyRowsIndicesDenseShapeMismatch(a));let g=y.getArrayFromDType(e,0),x=y.getArrayFromDType(o,0);return[g,[0,p],x,l,c]}let m=!0,f=0,d=new Array(u).fill(0);for(let g=0;g<a;++g){let x=r[g*p];if(x<0)throw new Error(v.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,x));if(x>=u)throw new Error(v.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,x,u));++d[x],m=m&&x>=f,f=x}let h=!0;for(let g=0;g<u;++g){let x=d[g]===0;l[g]=x,h=h&&!x,d[g]=Math.max(d[g],1),g>0&&(d[g]+=d[g-1])}if(h&&m){let g=r,x=n;for(let b=0;b<a;++b)c[b]=b;return[g,[a,p],x,l,c]}else{let g=d[u-1],x=y.getArrayFromDType(e,g*p),b=y.getArrayFromDType(o,g),w=new Array(u).fill(0);for(let C=0;C<a;++C){let N=r[C*p],_=w[N],A=(N===0?0:d[N-1])+_;w[N]++;for(let $=0;$<p;++$)x[A*p+$]=r[C*p+$];b[A]=n[C],c[C]=A}for(let C=0;C<u;++C)if(w[C]===0){let _=C===0?0:d[C-1];x[_*p+0]=C;for(let A=1;A<p;++A)x[_*p+A]=0;b[_]=i}return[x,[g,p],b,l,c]}}function aw(r,t,e,n,o){let s=y.sizeFromShape(n),i=t[0],a=o.length,u=[],l=1,c=-1;for(let g=0;g<a;++g){let x=o[g];if(x===-1){if(c!==-1)throw new Error(v.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(c,g));c=g,u.push(1)}else{if(x<0)throw new Error(v.getSparseReshapeNegativeOutputDimErrorMessage(g,x));l*=x,u.push(x)}}if(c!==-1){if(l<=0)throw new Error(v.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let g=Math.trunc(s/l);if(l*g!==s)throw new Error(v.getSparseReshapeInputOutputMultipleErrorMessage(n,u));u[c]=g}if(y.sizeFromShape(u)!==s)throw new Error(v.getSparseReshapeInputOutputMismatchErrorMessage(n,u));let m=n.length,f=[];if(m>0){f[m-1]=1;for(let g=m-2;g>=0;--g)f[g]=f[g+1]*n[g+1]}let d=[];if(a>0){d[a-1]=1;for(let g=a-2;g>=0;--g)d[g]=d[g+1]*u[g+1]}let h=y.getArrayFromDType(e,i*a);for(let g=0;g<i;++g){let x=0;for(let b=0;b<m;++b)x+=r[g*m+b]*f[b];for(let b=0;b<a;++b)h[g*a+b]=Math.trunc(x/d[b]),x%=d[b]}return[h,[i,a],u]}function ld(r,t,e,n,o,s=!1,i=0){let a=n.length,u=[t[0],r.length/t[0]],l=u[1],p=a>0?o[a-1]+1:0;if(p<0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let m=t.slice();m[0]=p;let f=m.reduce((w,C)=>w*C,1),d=y.getArrayFromDType(e,f);if(a===0)return p>0&&d.fill(i),[d,m];if(p<=0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let h=0,g=1,x=0,b=o[h];for(;;){let w=0;if(g<a){if(w=o[g],b===w){++g;continue}if(b>=w)throw new Error(v.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(b<0||b>=p)throw new Error(v.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(b,p));b>x&&d.fill(i,x*l,b*l);for(let C=h;C<g;++C){let N=n[C];if(N<0||N>=u[0])throw new Error(v.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(C,n[C],u[0]));for(let _=0;_<l;_++)d[b*l+_]+=r[N*l+_]}if(s)for(let C=0;C<l;C++)d[b*l+C]/=g-h;if(h=g,++g,x=b+1,b=w,g>a)break}return x<p&&d.fill(i,x*l,p*l),[d,m]}var lF=yn(r=>Math.sqrt(r)),gJ=kt(As,r=>Math.sqrt(r)),uF={kernelName:As,backendName:"cpu",kernelFunc:gJ};var sT=Qt((r,t)=>{let e=r-t;return e*e}),xJ=oe(Rs,sT),cF={kernelName:Rs,backendName:"cpu",kernelFunc:xJ};function lw(r,t,e,n){let o=wt(r,t.dtype);for(let s=0;s<o.size;s++){let i=o.indexToLoc(s),a=new Array(i.length);for(let u=0;u<a.length;u++)a[u]=i[u]*e[u]+n[u];o.set(t.get(...a),...i)}return o}var iT=class{constructor(t,e,n,o,s,i){this.separator=y.encodeString(t),this.nGramWidths=e,this.leftPad=y.encodeString(n),this.rightPad=y.encodeString(o),this.padWidth=s,this.preserveShort=i}getPadWidth(t){return Math.min(this.padWidth<0?t-1:this.padWidth,t-1)}getNumNGrams(t,e){let n=this.getPadWidth(e);return Math.max(0,t+2*n-e+1)}createNGrams(t,e,n,o,s,i){for(let a=0;a<s;++a){let u=this.getPadWidth(i),l=Math.max(0,u-a),c=Math.max(0,u-(s-(a+1))),p=i-(l+c),m=e+(l>0?0:a-u),f=0;f+=l*this.leftPad.length;for(let b=0;b<p;++b)f+=t[m+b].length;f+=c*this.rightPad.length,f+=(l+c+p-1)*this.separator.length,n[o+a]=new Uint8Array(f);let h=n[o+a],g=0,x=b=>b.forEach(w=>h[g++]=w);for(let b=0;b<l;++b)x(this.leftPad),x(this.separator);for(let b=0;b<p-1;++b)x(t[m+b]),x(this.separator);if(p>0){x(t[m+p-1]);for(let b=0;b<c;++b)x(this.separator),x(this.rightPad)}else{for(let b=0;b<c-1;++b)x(this.rightPad),x(this.separator);x(this.rightPad)}}}compute(t,e){let n=t.length,o=e.length;if(o>0){let u=e[0];if(u!==0)throw new Error(`First split value must be 0, got ${u}`);for(let l=1;l<o;++l){let c=e[l]>=u;if(c=c&&e[l]<=n,!c)throw new Error(`Invalid split value ${e[l]}, must be in [${u}, ${n}]`);u=e[l]}if(u!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${u}`)}let s=o-1,i=y.getArrayFromDType("int32",o);if(n===0||o===0){let u=new Array(n);for(let l=0;l<=s;++l)i[l]=0;return[u,i]}i[0]=0;for(let u=1;u<=s;++u){let l=e[u]-e[u-1],c=0;this.nGramWidths.forEach(p=>{c+=this.getNumNGrams(l,p)}),this.preserveShort&&l>0&&c===0&&(c=1),i[u]=i[u-1]+c}let a=new Array(i[s]);for(let u=0;u<s;++u){let l=e[u],c=i[u];if(this.nGramWidths.forEach(p=>{let m=e[u+1]-e[u],f=this.getNumNGrams(m,p);this.createNGrams(t,l,a,c,f,p),c+=f}),this.preserveShort&&c===i[u]){let p=e[u+1]-e[u];if(p===0)continue;let m=p+2*this.padWidth,f=1;this.createNGrams(t,l,a,c,f,m)}}return[a,i]}};function Dc(r,t,e,n,o,s,i,a){return new iT(e,n,o,s,i,a).compute(r,t)}function yJ(r,t,e,n){if(!r.length)return;if(t.length===0){for(let s=0;s<r.length;++s)n.push(r.subarray(s,s+1));return}if(t.length===1){let s=t[0],i=r.indexOf(s);for(;i!==-1;){let a=r.subarray(0,i);(!e||a.length!==0)&&n.push(a),r=r.subarray(i+1),i=r.indexOf(s)}(!e||r.length!==0)&&n.push(r);return}let o=0;for(let s=0;s<r.length+1;s++)if(s===r.length||t.indexOf(r[s])!==-1){let i=r.subarray(o,s);(!e||i.length!==0)&&n.push(i),o=s+1}}function Rc(r,t,e){let n=r.length,o=[],s=0,i=0,a=new Array(n);for(let m=0;m<n;++m){let f=o.length;yJ(r[m],t,e,o);let d=o.length-f;a[m]=d,s+=d,i=Math.max(i,d)}let u=y.getArrayFromDType("int32",s*2),l=new Array(s),c=[n,i],p=0;for(let m=0;m<n;++m)for(let f=0;f<a[m];++f)u[p*2]=m,u[p*2+1]=f,l[p]=o[p],++p;return[u,l,c]}function Fc(r,t){let e=y.getArrayFromDType("int32",r.length);for(let n=0;n<r.length;++n)e[n]=y.fingerPrint64(r[n]).modulo(t).getLowBitsUnsigned();return e}var aT=Qt((r,t)=>r-t),bJ=od((r,t,e,n)=>({real:r-e,imag:t-n})),Vh=oe(Fs,aT,bJ),pF={kernelName:Fs,backendName:"cpu",kernelFunc:Vh};function uw(r,t){let e=new Array(r.rank);for(let o=0;o<e.length;o++)e[o]=r.shape[o]*t[o];let n=wt(e,r.dtype);for(let o=0;o<n.values.length;++o){let s=n.indexToLoc(o),i=new Array(r.rank);for(let u=0;u<i.length;u++)i[u]=s[u]%r.shape[u];let a=r.locToIndex(i);n.values[o]=r.values[a]}return n}var Gh=(r,t)=>{let e=t.value-r.value;return e===0?r.index-t.index:e};function mF(r,t,e=0,n=r.length-1){for(;n>e;){if(n-e>600){let a=n-e+1,u=t-e+1,l=Math.log(a),c=.5*Math.exp(2*l/3),p=.5*Math.sqrt(l*c*(a-c)/a)*Math.sign(u-a/2),m=Math.max(e,Math.floor(t-u*c/a+p)),f=Math.min(n,Math.floor(t+(a-u)*c/a+p));mF(r,t,m,f)}let o=r[t],s=e,i=n;for(y.swap(r,e,t),Gh(r[n],o)>0&&y.swap(r,e,n);s<i;){for(y.swap(r,s,i),s++,i--;Gh(r[s],o)<0;)s=s+1;for(;Gh(r[i],o)>0;)i=i-1}Gh(r[e],o)===0?y.swap(r,e,i):(i=i+1,y.swap(r,i,n)),i<=t&&(e=i+1),t<=i&&(n=i-1)}}function cw(r,t,e,n,o){let s=t[t.length-1],[i,a]=[r.length/s,s],u=y.getTypedArrayFromDType(e,i*n),l=y.getTypedArrayFromDType("int32",i*n);for(let p=0;p<i;p++){let m=p*a,f=r.subarray(m,m+a),d=new Array(f.length);f.forEach((b,w)=>d[w]={value:b,index:w}),n<d.length&&(mF(d,n),d=d.slice(0,n)),o&&d.sort(Gh);let h=p*n,g=u.subarray(h,h+n),x=l.subarray(h,h+n);for(let b=0;b<n;b++)g[b]=d[b].value,x[b]=d[b].index}let c=t.slice();return c[c.length-1]=n,[wt(c,e,u),wt(c,"int32",l)]}function pw(r,t,e,n){let o=y.parseAxisParam(t,e)[0],s=[1,e[0],1];for(let d=0;d<o;d++)s[0]*=e[d];s[1]=e[o];for(let d=o+1;d<e.length;d++)s[2]*=e[d];let i={},a=new Int32Array(e[o]),u=new pe(s,n,r),l=[],c=s[0]===1&&s[2]===1;for(let d=0;d<e[o];d++){let h;if(c)h=r[d].toString();else{let g=[];for(let x=0;x<s[0];x++)for(let b=0;b<s[2];b++)g.push(u.get(x,d,b));h=g.join(",")}if(i[h]!==void 0)a[d]=i[h];else{let g=Object.keys(i).length;i[h]=g,a[d]=g,l.push(d)}}let p=s.slice();p[1]=Object.keys(i).length;let m=new pe(p,n);l.forEach((d,h)=>{for(let g=0;g<s[0];g++)for(let x=0;x<s[2];x++)m.set(u.get(g,d,x),g,h,x)});let f=e.slice();return f[o]=p[1],{outputValues:m.values,outputShape:f,indices:a}}Xp("cpu",()=>new Nu,1);var lT=kt(ts,r=>r>=0?r:Math.exp(r)-1),fF={kernelName:ts,backendName:"cpu",kernelFunc:lT};function uT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{alpha:s}=n;tt([o],"leakyRelu");let i=y.sizeFromShape(o.shape),a=e.data.get(o.dataId).values,u=y.getTypedArrayFromDType("float32",i);for(let l=0;l<a.length;l++)u[l]=a[l]<0?s*a[l]:a[l];return e.makeTensorInfo(o.shape,"float32",u)}var dF={kernelName:is,backendName:"cpu",kernelFunc:uT};var CJ=Qt((r,t)=>r<0?t*r:r);function cT(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t;tt([n,o],"prelu");let s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,[a,u]=CJ(n.shape,o.shape,s,i,"float32");return e.makeTensorInfo(u,"float32",a)}var hF={kernelName:bs,backendName:"cpu",kernelFunc:cT};var pT=kt(Cs,r=>Math.max(0,r)),gF={kernelName:Cs,backendName:"cpu",kernelFunc:pT};var mT=kt(vs,r=>Math.min(Math.max(0,r),6)),xF={kernelName:vs,backendName:"cpu",kernelFunc:mT};function Oc(r,t,e,n,o){if(e==="linear")return Kr({inputs:{x:t},backend:r});if(e==="relu")return pT({inputs:{x:t},backend:r});if(e==="elu")return lT({inputs:{x:t},backend:r});if(e==="relu6")return mT({inputs:{x:t},backend:r});if(e==="prelu")return cT({inputs:{x:t,alpha:n},backend:r});if(e==="leakyrelu")return uT({inputs:{x:t},backend:r,attrs:{alpha:o}});if(e==="sigmoid")return oT({inputs:{x:t},backend:r});throw new Error(`Activation ${e} has not been implemented for the CPU backend.`)}function Yt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{shape:s}=n,i=y.sizeFromShape(o.shape),a=y.inferFromImplicitShape(s,i),u=y.sizeFromShape(a);y.assert(i===u,()=>`The new shape (${a}) has ${u} elements and the old shape (${o.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),e.incRef(o.dataId);let l=e.data.get(o.dataId);if(l.complexTensorInfos!=null){let c=l.complexTensorInfos.real,p=l.complexTensorInfos.imag;c.shape=a,p.shape=a}return{dataId:o.dataId,shape:a,dtype:o.dtype}}var yF={kernelName:di,backendName:"cpu",kernelFunc:Yt};function fT(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;tt([o,s],"matMul");let u=o.shape.length,l=s.shape.length,c=i?o.shape[u-2]:o.shape[u-1],p=a?s.shape[l-1]:s.shape[l-2],m=i?o.shape[u-1]:o.shape[u-2],f=a?s.shape[l-2]:s.shape[l-1],d=o.shape.slice(0,-2),h=s.shape.slice(0,-2),g=y.sizeFromShape(d),x=y.sizeFromShape(h),w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)).concat([m,f]);y.assert(c===p,()=>`Error in matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${o.shape} and ${s.shape} and transposeA=${i} and transposeB=${a} must match.`);let C=i?[g,c,m]:[g,m,c],N=a?[x,f,p]:[x,p,f],_=Yt({inputs:{x:o},backend:e,attrs:{shape:C}}),A=Yt({inputs:{x:s},backend:e,attrs:{shape:N}}),$=i?_.shape[1]:_.shape[2],F=i?_.shape[2]:_.shape[1],P=a?A.shape[1]:A.shape[2],V=Math.max(g,x),G=e.data.get(_.dataId).values,W=e.data.get(A.dataId).values,q=y.computeStrides(_.shape),H=y.computeStrides(A.shape),[j,Y,Z]=i?[q[0],1,q[1]]:[q[0],q[1],1],[et,rt,ot]=a?[1,H[1],H[0]]:[H[1],1,H[0]],at=F*P,nt=wt([V,F,P],_.dtype),it=nt.values,dt=e.blockSize;for(let ht=0;ht<V;ht++)for(let bt=0;bt<F;bt+=dt)for(let Et=0;Et<P;Et+=dt)for(let At=0;At<$;At+=dt){let Vt=Math.min(bt+dt,F),Zt=Math.min(Et+dt,P),ce=Math.min(At+dt,$);for(let he=bt;he<Vt;he++)for(let jt=Et;jt<Zt;jt++){let ke=0;for(let fe=At;fe<ce;fe++){let Ae=Math.min(ht,g-1)*j,We=Math.min(ht,x-1)*ot,_n=G[Ae+he*Y+fe*Z],or=W[fe*et+jt*rt+We];ke+=_n*or}it[ht*at+(he*P+jt)]+=ke}}return e.disposeIntermediateTensorInfo(_),e.disposeIntermediateTensorInfo(A),e.makeTensorInfo(w,nt.dtype,nt.values)}var bF={kernelName:Ho,backendName:"cpu",kernelFunc:fT};function IJ(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n,m,f,d,h=[];m=fT({inputs:{a:o,b:s},attrs:{transposeA:u,transposeB:l},backend:e}),i&&(f=Ki({inputs:{a:m,b:i},backend:e}),h.push(m),m=f),c&&(d=Oc(e,m,c,a,p),h.push(m),m=d);for(let x of h)e.disposeIntermediateTensorInfo(x);return m}var wF={kernelName:Ci,backendName:"cpu",kernelFunc:IJ};var SJ=kt(oa,r=>Math.acos(r)),CF={kernelName:oa,backendName:"cpu",kernelFunc:SJ};var vJ=kt(sa,r=>Math.acosh(r)),IF={kernelName:sa,backendName:"cpu",kernelFunc:vJ};function NJ(r){let{inputs:t,backend:e}=r,n=t;tt(t,"addN");let o=n.map(a=>e.data.get(a.dataId).values),s=wt(n[0].shape,n[0].dtype),i=s.values;for(let a=0;a<n.length;a++){let u=o[a];for(let l=0;l<i.length;l++)i[l]+=u[l]}return e.makeTensorInfo(s.shape,s.dtype,s.values)}var SF={kernelName:Go,backendName:"cpu",kernelFunc:NJ};function TJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,"all");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims("all",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];w=w&&N}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var vF={kernelName:ia,backendName:"cpu",kernelFunc:TJ};function kJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,"any");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims("any",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];w=w||N}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var NF={kernelName:aa,backendName:"cpu",kernelFunc:kJ};function EJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n;tt(o,"argMax");let i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Ve({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],v.assertAxesAreInnerMostDims("argMax",i,u.shape.length);let[c,p]=v.computeOutAndReduceShapes(u.shape,i),m=y.sizeFromShape(c),f=y.makeZerosTypedArray(m,"int32"),d=y.sizeFromShape(p),h=e.data.get(u.dataId).values;for(let g=0;g<f.length;++g){let x=g*d,b=h[x],w=0;for(let C=0;C<d;++C){let N=h[x+C];N>b&&(b=N,w=C)}f[g]=w}return l.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.makeTensorInfo(c,"int32",f)}var TF={kernelName:Wo,backendName:"cpu",kernelFunc:EJ};function _J(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n;tt(o,"argMin");let i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Ve({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),i=[i[0]],v.assertAxesAreInnerMostDims("argMin",i,u.shape.length);let[c,p]=v.computeOutAndReduceShapes(u.shape,i),m=y.sizeFromShape(c),f=y.makeZerosTypedArray(m,"int32"),d=y.sizeFromShape(p),h=e.data.get(u.dataId).values;for(let g=0;g<f.length;++g){let x=g*d,b=h[x],w=0;for(let C=0;C<d;++C){let N=h[x+C];N<b&&(b=N,w=C)}f[g]=w}return l.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.makeTensorInfo(c,"int32",f)}var kF={kernelName:kl,backendName:"cpu",kernelFunc:_J};var AJ=kt(la,r=>Math.asin(r)),EF={kernelName:la,backendName:"cpu",kernelFunc:AJ};var $J=kt(ua,r=>Math.asinh(r)),_F={kernelName:ua,backendName:"cpu",kernelFunc:$J};var DJ=kt(ca,r=>Math.atan(r)),AF={kernelName:ca,backendName:"cpu",kernelFunc:DJ};var RJ=Qt((r,t)=>Math.atan2(r,t)),FJ=oe(ma,RJ),$F={kernelName:ma,backendName:"cpu",kernelFunc:FJ};var OJ=kt(pa,r=>Math.atanh(r)),DF={kernelName:pa,backendName:"cpu",kernelFunc:OJ};function ud(r,t,e,n,o,s){let i=o.strideHeight,a=o.strideWidth,u=o.dilationHeight,l=o.dilationWidth,c=o.effectiveFilterHeight,p=o.effectiveFilterWidth,m=o.padInfo.top,f=o.padInfo.left,d=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,h=wt(o.outShape,e),g=h.values,x=o.outShape[1]*o.outShape[2]*o.outShape[3],b=o.outShape[2]*o.outShape[3],w=o.outShape[3];for(let C=0;C<o.batchSize;++C){let N=C*x,_=C*n[0];for(let A=0;A<o.inChannels;++A)for(let $=0;$<o.outHeight;++$){let F=$*i-m,P=Math.max(0,F),V=Math.min(o.inHeight,c+F),G=N+$*b;for(let W=0;W<o.outWidth;++W){let q=W*a-f,H=Math.max(0,q),j=Math.min(o.inWidth,p+q),Y=d,Z=0,et=0;for(let ot=P;ot<V;ot+=u){let at=_+ot*n[1];for(let nt=H;nt<j;nt+=l){let it=at+nt*n[2],dt=r[it+A];s==="max"&&dt>Y?Y=dt:s==="avg"&&(Z+=dt,et++)}if(isNaN(Y))break}let rt=G+W*w+A;g[rt]=s==="avg"?Z/et:Y}}}return h}function fw(r,t,e,n,o=!1,s=!1){let i=wt(n.outShape,"int32"),a=n.strideHeight,u=n.strideWidth,l=n.dilationHeight,c=n.dilationWidth,p=n.effectiveFilterHeight,m=n.effectiveFilterWidth,f=n.padInfo.top,d=n.padInfo.left,h=wt(t,e,r);for(let g=0;g<n.batchSize;++g)for(let x=0;x<n.inChannels;++x)for(let b=0;b<n.outHeight;++b){let w=b*a-f,C=w;for(;C<0;)C+=l;let N=Math.min(n.inHeight,p+w);for(let _=0;_<n.outWidth;++_){let A=_*u-d,$=A;for(;$<0;)$+=c;let F=Math.min(n.inWidth,m+A),P=Number.NEGATIVE_INFINITY,V=-1;for(let G=C;G<N;G+=l){let W=G-w;for(let q=$;q<F;q+=c){let H=q-A,j=h.get(g,G,q,x);j>P&&(P=j,o?V=s?((g*n.inHeight+G)*n.inWidth+q)*n.inChannels+x:(G*n.inWidth+q)*n.inChannels+x:V=W*m+H)}}i.set(V,g,b,_,x)}}return i}function dw(r,t,e,n,o,s){let i=o.strideDepth,a=o.strideHeight,u=o.strideWidth,l=o.dilationDepth,c=o.dilationHeight,p=o.dilationWidth,m=o.effectiveFilterDepth,f=o.effectiveFilterHeight,d=o.effectiveFilterWidth,h=o.padInfo.front,g=o.padInfo.top,x=o.padInfo.left,b=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,w=wt(o.outShape,e),C=w.values,N=o.outShape[1]*o.outShape[2]*o.outShape[3]*o.outShape[4],_=o.outShape[2]*o.outShape[3]*o.outShape[4],A=o.outShape[3]*o.outShape[4],$=o.outShape[4];for(let F=0;F<o.batchSize;++F){let P=F*N,V=F*n[0];for(let G=0;G<o.inChannels;++G)for(let W=0;W<o.outDepth;++W){let q=W*i-h,H=q;for(;H<0;)H+=l;let j=Math.min(o.inDepth,m+q),Y=P+W*_;for(let Z=0;Z<o.outHeight;++Z){let et=Z*a-g,rt=et;for(;rt<0;)rt+=c;let ot=Math.min(o.inHeight,f+et),at=Y+Z*A;for(let nt=0;nt<o.outWidth;++nt){let it=nt*u-x,dt=it;for(;dt<0;)dt+=p;let ht=Math.min(o.inWidth,d+it),bt=at+nt*$,Et=b,At=0,Vt=0;for(let ce=H;ce<j;ce+=l){let he=V+ce*n[1];for(let jt=rt;jt<ot;jt+=c){let ke=he+jt*n[2];for(let fe=dt;fe<ht;fe+=p){let Ae=ke+fe*n[3],We=r[Ae+G];if(s==="max"&&We>Et?Et=We:s==="avg"&&(At+=We,Vt++),isNaN(Et))break}if(isNaN(Et))break}if(isNaN(Et))break}let Zt=bt+G;C[Zt]=s==="avg"?At/Vt:Et}}}}return w}function RF(r,t){let e=wt(t.outShape,"int32"),n=t.strideDepth,o=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,a=t.dilationHeight,u=t.dilationWidth,l=t.effectiveFilterDepth,c=t.effectiveFilterHeight,p=t.effectiveFilterWidth,m=t.padInfo.front,f=t.padInfo.top,d=t.padInfo.left;for(let h=0;h<t.batchSize;++h)for(let g=0;g<t.inChannels;++g)for(let x=0;x<t.outDepth;++x){let b=x*n-m,w=b;for(;w<0;)w+=i;let C=Math.min(t.inDepth,l+b);for(let N=0;N<t.outHeight;++N){let _=N*o-f,A=_;for(;A<0;)A+=a;let $=Math.min(t.inHeight,c+_);for(let F=0;F<t.outWidth;++F){let P=F*s-d,V=P;for(;V<0;)V+=u;let G=Math.min(t.inWidth,p+P),W=Number.NEGATIVE_INFINITY,q=-1;for(let H=w;H<C;H+=i){let j=H-b;for(let Y=A;Y<$;Y+=a){let Z=Y-_;for(let et=V;et<G;et+=u){let rt=et-P,ot=r.get(h,H,Y,et,g);ot>=W&&(W=ot,q=j*c*p+Z*c+rt)}}}e.set(q,h,x,N,F,g)}}}return e}function PJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;tt(o,"avgPool");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u),p;if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))p=Kr({inputs:{x:o},backend:e});else{let m=e.data.get(o.dataId).values,f=y.computeStrides(o.shape),d=ud(m,o.shape,o.dtype,f,c,"avg");p=e.makeTensorInfo(c.outShape,o.dtype,d.values)}return p}var FF={kernelName:Uo,backendName:"cpu",kernelFunc:PJ};function LJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n;tt(o,"avgPool3d");let c=v.computePool3DInfo(o.shape,s,i,1,a,u,l),p=e.data.get(o.dataId).values,m=dw(p,o.shape,o.dtype,y.computeStrides(o.shape),c,"avg");return e.makeTensorInfo(m.shape,"float32",m.values)}var OF={kernelName:El,backendName:"cpu",kernelFunc:LJ};function MJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=n;tt([o,s],"avgPool3DGrad");let c=v.computePool3DInfo(s.shape,i,a,1,u,l),p=c.strideDepth,m=c.strideHeight,f=c.strideWidth,d=c.filterDepth,h=c.filterHeight,g=c.filterWidth,x=c.dilationDepth,b=c.dilationHeight,w=c.dilationWidth,C=c.effectiveFilterDepth,N=c.effectiveFilterHeight,_=c.effectiveFilterWidth,A=C-1-c.padInfo.front,$=_-1-c.padInfo.left,F=N-1-c.padInfo.top,P=wt(s.shape,"float32"),V=1/(d*h*g),G=e.bufferSync(o);for(let W=0;W<c.batchSize;++W)for(let q=0;q<c.inChannels;++q)for(let H=0;H<c.inDepth;++H)for(let j=0;j<c.inHeight;++j)for(let Y=0;Y<c.inWidth;++Y){let Z=H-A,et=j-F,rt=Y-$,ot=0;for(let at=0;at<C;at+=x){let nt=(Z+at)/p;if(!(nt<0||nt>=c.outDepth||Math.floor(nt)!==nt))for(let it=0;it<N;it+=b){let dt=(et+it)/m;if(!(dt<0||dt>=c.outHeight||Math.floor(dt)!==dt))for(let ht=0;ht<_;ht+=w){let bt=(rt+ht)/f;if(bt<0||bt>=c.outWidth||Math.floor(bt)!==bt)continue;ot+=G.get(W,nt,dt,bt,q)}}}P.set(ot*V,W,H,j,Y,q)}return e.makeTensorInfo(P.shape,P.dtype,P.values)}var PF={kernelName:lp,backendName:"cpu",kernelFunc:MJ};function zJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s;tt([o,s],"avgPoolGrad");let{filterSize:a,strides:u,pad:l}=n,c=v.computePool2DInfo(i.shape,a,u,1,l),p=c.strideHeight,m=c.strideWidth,f=c.filterHeight,d=c.filterWidth,h=c.dilationHeight,g=c.dilationWidth,x=c.effectiveFilterHeight,b=c.effectiveFilterWidth,w=b-1-c.padInfo.left,C=x-1-c.padInfo.top,N=wt(i.shape,"float32"),_=1/(f*d),A=e.data.get(o.dataId).values,$=wt(o.shape,"float32",A);for(let F=0;F<c.batchSize;++F)for(let P=0;P<c.inChannels;++P)for(let V=0;V<c.inHeight;++V)for(let G=0;G<c.inWidth;++G){let W=V-C,q=G-w,H=0;for(let j=0;j<x;j+=h){let Y=(W+j)/p;if(!(Y<0||Y>=c.outHeight||Math.floor(Y)!==Y))for(let Z=0;Z<b;Z+=g){let et=(q+Z)/m;if(et<0||et>=c.outWidth||Math.floor(et)!==et)continue;H+=$.get(F,Y,et,P)}}N.set(H*_,F,V,G,P)}return e.makeTensorInfo(N.shape,N.dtype,N.values)}var LF={kernelName:ap,backendName:"cpu",kernelFunc:zJ};function BJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,scale:s,offset:i,mean:a,variance:u}=t;y.assert(a.shape.length===u.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),y.assert(i==null||a.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),y.assert(s==null||a.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),tt([o,a,u,s,i],"batchNorm");let{varianceEpsilon:l}=n;l==null&&(l=.001);let c=e.data.get(o.dataId).values,p=e.data.get(a.dataId).values,m=e.data.get(u.dataId).values,f=s?e.data.get(s.dataId).values:new Float32Array([1]),d=i?e.data.get(i.dataId).values:new Float32Array([0]),h=new Float32Array(c.length),g=d.length,x=f.length,b=m.length,w=p.length,C=0,N=0,_=0,A=0;for(let $=0;$<c.length;++$)h[$]=d[C++]+(c[$]-p[N++])*f[_++]/Math.sqrt(m[A++]+l),C>=g&&(C=0),N>=w&&(N=0),_>=x&&(_=0),A>=b&&(A=0);return e.makeTensorInfo(o.shape,o.dtype,h)}var MF={kernelName:os,backendName:"cpu",kernelFunc:BJ};function VJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n;tt([o],"batchToSpaceND");let a=s.reduce((x,b)=>x*b),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=Yt({inputs:{x:o},backend:e,attrs:{shape:u}}),d=Ve({inputs:{x:f},backend:e,attrs:{perm:l}}),h=Yt({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Fo({inputs:{x:h},backend:e,attrs:{begin:p,size:m}});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),g}var zF={kernelName:ai,backendName:"cpu",kernelFunc:VJ};function GJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i}=n,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,l=sd(a,u,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,l)}var BF={kernelName:up,backendName:"cpu",kernelFunc:GJ};function WJ(r){let{inputs:t,backend:e}=r,{s0:n,s1:o}=t,s=e.data.get(n.dataId).values,i=e.data.get(o.dataId).values,a=v.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return e.makeTensorInfo([a.length],"int32",Int32Array.from(a))}var VF={kernelName:cp,backendName:"cpu",kernelFunc:WJ};var UJ=kt(uo,(r,t)=>{let e=t;return r>e.clipValueMax?e.clipValueMax:r<e.clipValueMin?e.clipValueMin:r}),GF={kernelName:uo,backendName:"cpu",kernelFunc:UJ};var HJ=r=>{let{x:t}=r.inputs,e=r.backend,n=new Float32Array(y.sizeFromShape(t.shape)),o=e.data.get(t.dataId),s=o.complexTensorInfos.real,i=o.complexTensorInfos.imag,a=e.data.get(s.dataId).values,u=e.data.get(i.dataId).values;for(let l=0;l<a.length;l++){let c=a[l],p=u[l];n[l]=Math.hypot(c,p)}return e.makeOutput(n,t.shape,"float32")},WF={kernelName:_l,backendName:"cpu",kernelFunc:HJ};function ji(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.data.get(n.dataId).complexTensorInfos.imag,s=e.data.get(o.dataId).values;return e.makeTensorInfo(o.shape,o.dtype,s)}var UF={kernelName:Sp,backendName:"cpu",kernelFunc:ji};function Tu(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n,s=y.parseAxisParam(o,t[0].shape)[0],i=t.map(h=>h.shape);v.assertParamsConsistent(i,s);let a=v.computeOutShape(t.map(h=>h.shape),s);if(y.sizeFromShape(a)===0)return e.makeTensorInfo(a,t[0].dtype,[]);let u=t.filter(h=>y.sizeFromShape(h.shape)>0);if(u.length===1)return Kr({inputs:{x:u[0]},backend:e});if(u[0].dtype==="complex64"){let h=u.map(C=>Ao({inputs:{input:C},backend:e})),g=u.map(C=>ji({inputs:{input:C},backend:e})),x=Tu({inputs:h,backend:e,attrs:{axis:s}}),b=Tu({inputs:g,backend:e,attrs:{axis:s}}),w=wr({inputs:{real:x,imag:b},backend:e});return h.forEach(C=>e.disposeIntermediateTensorInfo(C)),g.forEach(C=>e.disposeIntermediateTensorInfo(C)),e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(b),w}let l=u.map(h=>{let g=y.sizeFromShape(h.shape.slice(s));return Yt({inputs:{x:h},backend:e,attrs:{shape:[-1,g]}})}),c=l.map(h=>({vals:e.data.get(h.dataId).values,shape:h.shape}));a=v.computeOutShape(l.map(h=>h.shape),1);let p=l[0].shape[0]===1,m=Ec(c,a,t[0].dtype,p),f=v.computeOutShape(u.map(h=>h.shape),s),d=e.makeTensorInfo(f,t[0].dtype,m);return l.forEach(h=>e.disposeIntermediateTensorInfo(h)),d}var HF={kernelName:li,backendName:"cpu",kernelFunc:Tu};function dT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dataFormat:u,dilations:l,dimRoundingMode:c}=n;tt([o,s],"conv2d");let p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,s.shape,i,l,a,c,!1,p),f=m.filterHeight,d=m.filterWidth,h=m.dilationHeight,g=m.dilationWidth,x=m.padInfo.left,b=m.padInfo.top,w=m.dataFormat==="channelsLast",C=new pe(m.outShape,o.dtype),N=y.computeStrides(o.shape),_=y.computeStrides(s.shape),A=N[0],$=w?N[1]:N[2],F=w?N[2]:1,P=w?1:N[1],V=C.strides[0],G=w?C.strides[1]:C.strides[2],W=w?C.strides[2]:1,q=w?1:C.strides[1],H=e.data.get(o.dataId).values,j=e.data.get(s.dataId).values,Y=C.values;for(let Z=0;Z<m.batchSize;++Z){let et=Z*A,rt=Z*V;for(let ot=0;ot<m.outHeight;++ot){let at=rt+ot*G,nt=ot*m.strideHeight-b;for(let it=0;it<f;++it){let dt=nt+it*h;if(dt<0||dt>=m.inHeight)continue;let ht=it*_[0],bt=et+dt*$;for(let Et=0;Et<m.outWidth;++Et){let At=at+Et*W,Vt=Et*m.strideWidth-x;for(let Zt=0;Zt<d;++Zt){let ce=Vt+Zt*g;if(ce<0||ce>=m.inWidth)continue;let he=ht+Zt*_[1],jt=bt+ce*F,ke=he;for(let fe=0;fe<m.inChannels;++fe){let Ae=H[jt+fe*P];for(let We=0;We<m.outChannels;++We)Y[At+We*q]+=Ae*j[ke+We];ke+=m.outChannels}}}}}}return e.makeTensorInfo(C.shape,C.dtype,Y)}var qF={kernelName:Ko,backendName:"cpu",kernelFunc:dT};function qJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,filterShape:c}=n;tt([o,s],"conv2dBackpropFilter");let p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,c,i,1,a,l,!1,p),{strideHeight:f,strideWidth:d,filterHeight:h,filterWidth:g}=m,x=m.dataFormat==="channelsLast",b=new pe(m.filterShape,"float32"),w=m.padInfo.left,C=m.padInfo.top,N=e.data.get(o.dataId).values,_=e.data.get(s.dataId).values,A=new pe(o.shape,o.dtype,N),$=new pe(s.shape,s.dtype,_);for(let F=0;F<h;++F){let P=Math.max(0,Math.ceil((C-F)/f)),V=Math.min(m.outHeight,(m.inHeight+C-F)/f);for(let G=0;G<g;++G){let W=Math.max(0,Math.ceil((w-G)/d)),q=Math.min(m.outWidth,(m.inWidth+w-G)/d);for(let H=0;H<m.inChannels;++H)for(let j=0;j<m.outChannels;++j){let Y=0;for(let Z=0;Z<m.batchSize;++Z)for(let et=P;et<V;++et){let rt=F+et*f-C;for(let ot=W;ot<q;++ot){let at=G+ot*d-w;x?Y+=A.get(Z,rt,at,H)*$.get(Z,et,ot,j):Y+=A.get(Z,H,rt,at)*$.get(Z,j,et,ot)}}b.set(Y,F,G,H,j)}}}return e.makeTensorInfo(b.shape,b.dtype,b.values)}var KF={kernelName:mp,backendName:"cpu",kernelFunc:qJ};function KJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{inputShape:i,strides:a,pad:u,dataFormat:l,dimRoundingMode:c}=n;tt([o,s],"conv2dBackpropInput");let p=y.computeStrides(s.shape),m=y.computeStrides(o.shape),f=v.convertConv2DDataFormat(l),d=v.computeConv2DInfo(i,s.shape,a,1,u,c,!1,f),h=new pe(d.inShape,"float32"),g=h.values,x=e.data.get(o.dataId).values,b=e.data.get(s.dataId).values,[w,C,N]=p,{batchSize:_,filterHeight:A,filterWidth:$,inChannels:F,inHeight:P,inWidth:V,outChannels:G,outHeight:W,outWidth:q,strideHeight:H,strideWidth:j}=d;f=d.dataFormat;let Y=A-1-d.padInfo.top,Z=$-1-d.padInfo.left,et=f==="channelsLast",rt=h.strides[0],ot=et?h.strides[1]:h.strides[2],at=et?h.strides[2]:1,nt=et?1:h.strides[1],it=m[0],dt=et?m[1]:m[2],ht=et?m[2]:1,bt=et?1:m[1];for(let Et=0;Et<_;++Et)for(let At=0;At<F;++At)for(let Vt=0;Vt<P;++Vt){let Zt=Vt-Y,ce=Math.max(0,Math.ceil(Zt/H)),he=Math.min(W,(A+Zt)/H);for(let jt=0;jt<V;++jt){let ke=jt-Z,fe=Math.max(0,Math.ceil(ke/j)),Ae=Math.min(q,($+ke)/j),We=0;for(let or=ce;or<he;++or){let Hn=or*H-Zt;for(let Lr=fe;Lr<Ae;++Lr){let qe=Lr*j-ke,Mr=it*Et+dt*or+ht*Lr,zr=w*(A-1-Hn)+C*($-1-qe)+N*At;for(let qn=0;qn<G;++qn){let Kn=x[Mr+bt*qn],Xr=b[zr+qn];We+=Kn*Xr}}}let _n=rt*Et+ot*Vt+at*jt+nt*At;g[_n]=We}}return e.makeTensorInfo(h.shape,h.dtype,h.values)}var jF={kernelName:jo,backendName:"cpu",kernelFunc:KJ};function jJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n;tt([o,s],"conv3d");let l=v.computeConv3DInfo(o.shape,s.shape,i,u,a),{filterDepth:c,filterHeight:p,filterWidth:m,dilationDepth:f,dilationHeight:d,dilationWidth:h,padInfo:g}=l,x=g.front,b=g.left,w=g.top,C=new pe(l.outShape,o.dtype),N=e.data.get(o.dataId).values,_=e.data.get(s.dataId).values,A=C.values,$=y.computeStrides(o.shape),F=y.computeStrides(s.shape);for(let P=0;P<l.batchSize;++P){let V=P*$[0],G=P*C.strides[0];for(let W=0;W<l.outDepth;++W){let q=G+W*C.strides[1],H=W*l.strideDepth-x;for(let j=0;j<c;++j){let Y=H+j*f;if(Y<0||Y>=l.inDepth)continue;let Z=j*F[0],et=V+Y*$[1];for(let rt=0;rt<l.outHeight;++rt){let ot=q+rt*C.strides[2],at=rt*l.strideHeight-w;for(let nt=0;nt<p;++nt){let it=at+nt*d;if(it<0||it>=l.inHeight)continue;let dt=Z+nt*F[1],ht=et+it*$[2];for(let bt=0;bt<l.outWidth;++bt){let Et=ot+bt*l.outChannels,At=bt*l.strideWidth-b;for(let Vt=0;Vt<m;++Vt){let Zt=At+Vt*h;if(Zt<0||Zt>=l.inWidth)continue;let ce=dt+Vt*F[2],he=ht+Zt*l.inChannels,jt=ce;for(let ke=0;ke<l.inChannels;++ke){let fe=N[he+ke];for(let Ae=0;Ae<l.outChannels;++Ae)A[Et+Ae]+=fe*_[jt+Ae];jt+=l.outChannels}}}}}}}}return e.makeTensorInfo(C.shape,C.dtype,C.values)}var XF={kernelName:Al,backendName:"cpu",kernelFunc:jJ};function XJ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,filterShape:u}=n;tt([o,s],"conv3dBackpropFilterV2");let l=y.computeStrides(o.shape),c=y.computeStrides(s.shape),p=v.computeConv3DInfo(o.shape,u,i,1,a),m=p.strideDepth,f=p.strideHeight,d=p.strideWidth,h=p.filterDepth,g=p.filterHeight,x=p.filterWidth,b=new pe(p.filterShape,"float32"),w=b.values,[C,N,_,A]=b.strides,$=e.data.get(s.dataId).values,[F,P,V,G]=c,W=e.data.get(o.dataId).values,[q,H,j,Y]=l,Z=p.padInfo.front,et=p.padInfo.left,rt=p.padInfo.top;for(let ot=0;ot<h;++ot){let at=Math.max(0,Math.ceil((Z-ot)/m)),nt=Math.min(p.outDepth,(p.inDepth+Z-ot)/m),it=ot*C;for(let dt=0;dt<g;++dt){let ht=Math.max(0,Math.ceil((rt-dt)/f)),bt=Math.min(p.outHeight,(p.inHeight+rt-dt)/f),Et=dt*N+it;for(let At=0;At<x;++At){let Vt=Math.max(0,Math.ceil((et-At)/d)),Zt=Math.min(p.outWidth,(p.inWidth+et-At)/d),ce=At*_+Et;for(let he=0;he<p.inChannels;++he){let jt=he*A+ce;for(let ke=0;ke<p.outChannels;++ke){let fe=0;for(let Ae=0;Ae<p.batchSize;++Ae){let We=Ae*q,_n=Ae*F;for(let or=at;or<nt;++or){let Lr=(ot+or*m-Z)*H+We,qe=or*P+_n;for(let Mr=ht;Mr<bt;++Mr){let qn=(dt+Mr*f-rt)*j+Lr,Kn=Mr*V+qe;for(let Xr=Vt;Xr<Zt;++Xr){let Mo=(At+Xr*d-et)*Y+qn,Il=Xr*G+Kn;fe+=W[Mo+he]*$[Il+ke]}}}}w[jt+ke]=fe}}}}}return e.makeTensorInfo(b.shape,b.dtype,b.values)}var YF={kernelName:fp,backendName:"cpu",kernelFunc:XJ};function YJ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{pad:i,strides:a,inputShape:u}=n;tt([o],"conv3dBackpropInputV2");let l=y.computeStrides(o.shape),c=y.computeStrides(s.shape),p=v.computeConv3DInfo(u,s.shape,a,1,i),m=new pe(p.inShape,"float32"),f=m.values,[d,h,g,x]=m.strides,b=e.data.get(o.dataId).values,[w,C,N,_]=l,A=e.data.get(s.dataId).values,[$,F,P,V]=c,{batchSize:G,filterDepth:W,filterHeight:q,filterWidth:H,inChannels:j,inDepth:Y,inHeight:Z,inWidth:et,outChannels:rt,outDepth:ot,outHeight:at,outWidth:nt,strideDepth:it,strideHeight:dt,strideWidth:ht}=p,bt=W-1-p.padInfo.front,Et=q-1-p.padInfo.top,At=H-1-p.padInfo.left;for(let Vt=0;Vt<G;++Vt)for(let Zt=0;Zt<j;++Zt)for(let ce=0;ce<Y;++ce){let he=ce-bt,jt=Math.max(0,Math.ceil(he/it)),ke=Math.min(ot,(W+he)/it);for(let fe=0;fe<Z;++fe){let Ae=fe-Et,We=Math.max(0,Math.ceil(Ae/dt)),_n=Math.min(at,(q+Ae)/dt);for(let or=0;or<et;++or){let Hn=or-At,Lr=Math.max(0,Math.ceil(Hn/ht)),qe=Math.min(nt,(H+Hn)/ht),Mr=0;for(let zr=jt;zr<ke;++zr){let qn=zr*it-he;for(let Kn=We;Kn<_n;++Kn){let Xr=Kn*dt-Ae;for(let Zi=Lr;Zi<qe;++Zi){let Mo=Zi*ht-Hn,Il=w*Vt+C*zr+N*Kn+_*Zi,Tr=$*(W-1-qn)+F*(q-1-Xr)+P*(H-1-Mo)+V*Zt;for(let Ji=0;Ji<rt;++Ji){let Rd=b[Il+Ji],Fd=A[Tr+Ji];Mr+=Rd*Fd}}}}f[d*Vt+h*ce+g*fe+x*or+Zt]=Mr}}}return e.makeTensorInfo(m.shape,m.dtype,m.values)}var ZF={kernelName:dp,backendName:"cpu",kernelFunc:YJ};var ZJ=kt(Xo,r=>Math.cos(r)),JF={kernelName:Xo,backendName:"cpu",kernelFunc:ZJ};var JJ=kt(Yo,r=>Math.cosh(r)),QF={kernelName:Yo,backendName:"cpu",kernelFunc:JJ};function QJ(r){let{inputs:t,backend:e,attrs:n}=r,{image:o,boxes:s,boxInd:i}=t,{cropSize:a,method:u,extrapolationValue:l}=n,[c,p,m,f]=o.shape,d=s.shape[0],[h,g]=a,x=wt([d,h,g,f],"float32"),b=e.data.get(s.dataId).values,w=e.data.get(i.dataId).values,C=e.data.get(o.dataId).values,N=y.computeStrides(o.shape),_=y.computeStrides(x.shape);for(let A=0;A<d;A++){let $=A*4,F=b[$],P=b[$+1],V=b[$+2],G=b[$+3],W=w[A];if(W>=c)continue;let q=h>1?(V-F)*(p-1)/(h-1):0,H=g>1?(G-P)*(m-1)/(g-1):0;for(let j=0;j<h;j++){let Y=h>1?F*(p-1)+j*q:.5*(F+V)*(p-1);if(Y<0||Y>p-1){for(let Z=0;Z<g;Z++)for(let et=0;et<f;et++){let rt=et+Z*_[2]+j*_[1]+A*_[0];x.values[rt]=l}continue}if(u==="bilinear"){let Z=Math.floor(Y),et=Math.ceil(Y),rt=Y-Z;for(let ot=0;ot<g;ot++){let at=g>1?P*(m-1)+ot*H:.5*(P+G)*(m-1);if(at<0||at>m-1){for(let ht=0;ht<f;ht++){let bt=ht+ot*_[2]+j*_[1]+A*_[0];x.values[bt]=l}continue}let nt=Math.floor(at),it=Math.ceil(at),dt=at-nt;for(let ht=0;ht<f;ht++){let bt=ht+nt*N[2]+Z*N[1]+W*N[0],Et=C[bt];bt=ht+it*N[2]+Z*N[1]+W*N[0];let At=C[bt];bt=ht+nt*N[2]+et*N[1]+W*N[0];let Vt=C[bt];bt=ht+it*N[2]+et*N[1]+W*N[0];let Zt=C[bt],ce=Et+(At-Et)*dt,he=Vt+(Zt-Vt)*dt;bt=ht+ot*_[2]+j*_[1]+A*_[0],x.values[bt]=ce+(he-ce)*rt}}}else for(let Z=0;Z<g;++Z){let et=g>1?P*(m-1)+Z*H:.5*(P+G)*(m-1);if(et<0||et>m-1){for(let at=0;at<f;at++){let nt=at+Z*_[2]+j*_[1]+A*_[0];x.values[nt]=l}continue}let rt=Math.round(et),ot=Math.round(Y);for(let at=0;at<f;at++){let nt=at+rt*N[2]+ot*N[1]+W*N[0],it=at+Z*_[2]+j*_[1]+A*_[0];x.values[it]=C[nt]}}}}return e.makeTensorInfo(x.shape,x.dtype,x.values)}var tO={kernelName:da,backendName:"cpu",kernelFunc:QJ};function tQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;tt(o,"cumprod");let u=v.getAxesPermutation([s],o.shape.length),l=o;u!=null&&(l=Ve({inputs:{x:o},backend:e,attrs:{perm:u}}));let c=v.getInnerMostAxes(1,o.shape.length)[0];if(c!==l.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);let p=sr(l.dtype,"int32"),m=y.makeOnesTypedArray(y.sizeFromShape(l.shape),p),f=e.data.get(l.dataId).values,d=l.shape[l.shape.length-1],h=a?(x,b)=>x+d-b-1:(x,b)=>x+b;for(let x=0;x<f.length;x+=d)for(let b=0;b<d;b++){let w=h(x,b);if(b===0)m[w]=i?1:f[w];else{let C=h(x,b-1);m[w]=i?f[C]*m[C]:f[w]*m[C]}}let g=e.makeTensorInfo(l.shape,p,m);if(u!=null){let x=v.getUndoAxesPermutation(u),b=Ve({inputs:{x:g},backend:e,attrs:{perm:x}});return e.disposeIntermediateTensorInfo(g),e.disposeIntermediateTensorInfo(l),b}return g}var eO={kernelName:fa,backendName:"cpu",kernelFunc:tQ};function eQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;tt(o,"cumsum");let u=v.getAxesPermutation([s],o.shape.length),l=o;u!=null&&(l=Ve({inputs:{x:o},backend:e,attrs:{perm:u}}));let c=v.getInnerMostAxes(1,o.shape.length)[0];if(c!==l.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${l.shape.length-1} but got axis=${c}`);let p=sr(l.dtype,"int32"),m=y.makeZerosTypedArray(y.sizeFromShape(l.shape),p),f=e.data.get(l.dataId).values,d=l.shape[l.shape.length-1],h=a?(x,b)=>x+d-b-1:(x,b)=>x+b;for(let x=0;x<f.length;x+=d)for(let b=0;b<d;b++){let w=h(x,b);if(b===0)m[w]=i?0:f[w];else{let C=h(x,b-1);m[w]=i?f[C]+m[C]:f[w]+m[C]}}let g=e.makeTensorInfo(l.shape,p,m);if(u!=null){let x=v.getUndoAxesPermutation(u),b=Ve({inputs:{x:g},backend:e,attrs:{perm:x}});return e.disposeIntermediateTensorInfo(g),e.disposeIntermediateTensorInfo(l),b}return g}var rO={kernelName:Zo,backendName:"cpu",kernelFunc:eQ};function rQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i,binaryOutput:a}=n;if(o.shape.length===1){let u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=sd(u,l,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,c)}else if(o.shape.length===2){let u=e.bufferSync(o),l=e.bufferSync(s),c=Jb(u,l,i,a);return e.makeTensorInfo(c.shape,s.dtype,c.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${o.shape.length}.`)}var nO={kernelName:hp,backendName:"cpu",kernelFunc:rQ};function nQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockSize:s,dataFormat:i}=n;y.assert(i==="NHWC",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let a=o.shape[0],u=o.shape[1],l=o.shape[2],c=o.shape[3],p=u*s,m=l*s,f=c/(s*s),d=e.data.get(o.dataId).values,h=new Float32Array(a*p*m*f),g=0;for(let x=0;x<a;++x)for(let b=0;b<p;++b){let w=Math.floor(b/s),C=b%s;for(let N=0;N<m;++N){let _=Math.floor(N/s),A=N%s,$=(C*s+A)*f;for(let F=0;F<f;++F){let V=F+$+c*(_+l*(w+u*x));h[g++]=d[V]}}}return e.makeTensorInfo([a,p,m,f],o.dtype,h)}var oO={kernelName:ha,backendName:"cpu",kernelFunc:nQ};function hT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u,dimRoundingMode:l}=n;tt([o,s],"depthwiseConv2DNative");let c=y.computeStrides(o.shape),p=y.computeStrides(s.shape),m=u;m==null&&(m=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(i,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${m}'`);let f=v.computeConv2DInfo(o.shape,s.shape,i,m,a,l,!0),{filterHeight:d,filterWidth:h,dilationHeight:g,dilationWidth:x,padInfo:b}=f,w=b.left,C=b.top,N=f.outChannels/f.inChannels,_=new pe(f.outShape,o.dtype),A=e.data.get(o.dataId).values,$=e.data.get(s.dataId).values,F=_.values;for(let P=0;P<f.batchSize;++P){let V=P*c[0],G=P*_.strides[0];for(let W=0;W<f.outHeight;++W){let q=G+W*_.strides[1],H=W*f.strideHeight-C;for(let j=0;j<d;++j){let Y=H+j*g;if(Y<0||Y>=f.inHeight)continue;let Z=j*p[0],et=V+Y*c[1];for(let rt=0;rt<f.outWidth;++rt){let ot=q+rt*_.strides[2],at=rt*f.strideWidth-w;for(let nt=0;nt<h;++nt){let it=at+nt*x;if(it<0||it>=f.inWidth)continue;let dt=Z+nt*p[1],ht=et+it*f.inChannels,bt=ot,Et=dt;for(let At=0;At<f.inChannels;++At){let Vt=A[ht+At];for(let Zt=0;Zt<N;++Zt)F[bt+Zt]+=Vt*$[Et+Zt];bt+=N,Et+=N}}}}}}return e.makeTensorInfo(_.shape,_.dtype,_.values)}var sO={kernelName:Jo,backendName:"cpu",kernelFunc:hT};function oQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,filterShape:c}=n;tt([o,s],"depthwiseConv2dNativeBackpropFilter");let p=v.computeConv2DInfo(o.shape,c,i,a,u,l,!0),{strideHeight:m,strideWidth:f,filterHeight:d,filterWidth:h}=p,g=new pe(p.filterShape,"float32"),x=p.padInfo.left,b=p.padInfo.top,w=p.outChannels/p.inChannels,C=e.data.get(o.dataId).values,N=new pe(o.shape,o.dtype,C),_=e.data.get(s.dataId).values,A=new pe(s.shape,s.dtype,_);for(let $=0;$<d;++$){let F=Math.max(0,Math.ceil((b-$)/m)),P=Math.min(p.outHeight,(p.inHeight+b-$)/m);for(let V=0;V<h;++V){let G=Math.max(0,Math.ceil((x-V)/f)),W=Math.min(p.outWidth,(p.inWidth+x-V)/f);for(let q=0;q<p.outChannels;++q){let H=Math.trunc(q/w),j=q%w,Y=0;for(let Z=0;Z<p.batchSize;++Z)for(let et=F;et<P;++et){let rt=$+et*m-b;for(let ot=G;ot<W;++ot){let at=V+ot*f-x;Y+=N.get(Z,rt,at,H)*A.get(Z,et,ot,q)}}g.set(Y,$,V,H,j)}}}return e.makeTensorInfo(g.shape,g.dtype,g.values)}var iO={kernelName:gp,backendName:"cpu",kernelFunc:oQ};function sQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,inputShape:c}=n;tt([o,s],"depthwiseConv2DNativeBackpropInput");let p=y.computeStrides(o.shape),m=y.computeStrides(s.shape),f=v.computeConv2DInfo(c,s.shape,i,a,u,l,!0),d=new pe(f.inShape,"float32"),h=d.values,[g,x,b]=d.strides,w=e.data.get(o.dataId).values,[C,N,_]=p,A=e.data.get(s.dataId).values,[$,F,P]=m,{batchSize:V,filterHeight:G,filterWidth:W,inChannels:q,inHeight:H,inWidth:j,outChannels:Y,outHeight:Z,outWidth:et,strideHeight:rt,strideWidth:ot}=f,at=G-1-f.padInfo.top,nt=W-1-f.padInfo.left,it=Y/q;for(let dt=0;dt<V;++dt)for(let ht=0;ht<q;++ht)for(let bt=0;bt<H;++bt){let Et=bt-at,At=Math.max(0,Math.ceil(Et/rt)),Vt=Math.min(Z,(G+Et)/rt);for(let Zt=0;Zt<j;++Zt){let ce=Zt-nt,he=Math.max(0,Math.ceil(ce/ot)),jt=Math.min(et,(W+ce)/ot),ke=0;for(let fe=At;fe<Vt;++fe){let Ae=fe*rt-Et;for(let We=he;We<jt;++We){let _n=We*ot-ce,or=C*dt+N*fe+_*We,Hn=$*(G-1-Ae)+F*(W-1-_n)+P*ht;for(let Lr=0;Lr<it;++Lr){let qe=ht*it+Lr,Mr=w[or+qe],zr=A[Hn+Lr];ke+=Mr*zr}}}h[g*dt+x*bt+b*Zt+ht]=ke}}return e.makeTensorInfo(d.shape,d.dtype,d.values)}var aO={kernelName:xp,backendName:"cpu",kernelFunc:sQ};function iQ(r){let{inputs:t,backend:e}=r,{x:n}=t,o=y.sizeFromShape(n.shape),s=e.data.get(n.dataId).values,i=wt([o,o],n.dtype),a=i.values;for(let l=0;l<s.length;l++)a[l*o+l]=s[l];let u=[...n.shape,...n.shape];return e.makeTensorInfo(u,i.dtype,i.values)}var lO={kernelName:yp,backendName:"cpu",kernelFunc:iQ};var uO={kernelName:$l,backendName:"cpu",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o}=r,{strides:s,pad:i,dilations:a}=e,u=t,l=u.data.get(n.dataId).values,c=n.shape.length,p=u.data.get(o.dataId).values,m=o.shape.length,{batchSize:f,inHeight:d,inWidth:h,inChannels:g,outHeight:x,outWidth:b,padInfo:w,strideHeight:C,strideWidth:N,filterHeight:_,filterWidth:A,dilationHeight:$,dilationWidth:F,outShape:P}=v.computeDilation2DInfo(n.shape,o.shape,s,i,"NHWC",a),V=y.sizeFromShape(P),G=P.length,W=y.getArrayFromDType(n.dtype,V);for(let H=0;H<f;++H)for(let j=0;j<x;++j){let Y=j*C-w.top;for(let Z=0;Z<b;++Z){let et=Z*N-w.left;for(let rt=0;rt<g;++rt){let ot=Number.MIN_SAFE_INTEGER;for(let nt=0;nt<_;++nt){let it=Y+nt*$;if(it>=0&&it<d)for(let dt=0;dt<A;++dt){let ht=et+dt*F;if(ht>=0&&ht<h){let bt=y.locToIndex([H,it,ht,rt],c,y.computeStrides(n.shape)),Et=y.locToIndex([nt,dt,rt],m,y.computeStrides(o.shape)),At=l[bt]+p[Et];At>ot&&(ot=At)}}}let at=y.locToIndex([H,j,Z,rt],G,y.computeStrides(P));W[at]=ot}}}return{dataId:u.write(y.toTypedArray(W,n.dtype),P,n.dtype),shape:P,dtype:n.dtype}}};var cO={kernelName:Xd,backendName:"cpu",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o,dy:s}=r,{strides:i,pad:a,dilations:u}=e,l=t,c=y.toNestedArray(n.shape,l.data.get(n.dataId).values),p=y.toNestedArray(o.shape,l.data.get(o.dataId).values),{batchSize:m,inHeight:f,inWidth:d,inChannels:h,outHeight:g,outWidth:x,padInfo:b,strideHeight:w,strideWidth:C,filterHeight:N,filterWidth:_,dilationHeight:A,dilationWidth:$,outShape:F}=v.computeDilation2DInfo(n.shape,o.shape,i,a,"NHWC",u);y.assert(s.rank===F.length,()=>`Error in ${Xd}, dy must have the same rank as output ${F.length}, but got ${s.rank}`);let P=y.toNestedArray(F,l.data.get(s.dataId).values),V=y.makeZerosNestedTypedArray(o.shape,o.dtype);for(let W=0;W<m;++W)for(let q=0;q<g;++q){let H=q*w-b.top;for(let j=0;j<x;++j){let Y=j*C-b.left;for(let Z=0;Z<h;++Z){let et=Number.MIN_SAFE_INTEGER,rt=0,ot=0;for(let at=0;at<N;++at){let nt=H+at*A;if(nt>=0&&nt<f)for(let it=0;it<_;++it){let dt=Y+it*$;if(dt>=0&&dt<d){let ht=c[W][nt][dt][Z]+p[at][it][Z];ht>et&&(et=ht,rt=at,ot=it)}}}V[rt][ot][Z]+=P[W][q][j][Z]}}}return{dataId:l.write(y.toTypedArray(V,n.dtype),o.shape,o.dtype),shape:o.shape,dtype:o.dtype}}};var pO={kernelName:jd,backendName:"cpu",kernelFunc:({inputs:r,backend:t,attrs:e})=>{let{x:n,filter:o,dy:s}=r,{strides:i,pad:a,dilations:u}=e,l=t,c=y.toNestedArray(n.shape,l.data.get(n.dataId).values),p=y.toNestedArray(o.shape,l.data.get(o.dataId).values),{batchSize:m,inHeight:f,inWidth:d,inChannels:h,outHeight:g,outWidth:x,padInfo:b,strideHeight:w,strideWidth:C,filterHeight:N,filterWidth:_,dilationHeight:A,dilationWidth:$,outShape:F}=v.computeDilation2DInfo(n.shape,o.shape,i,a,"NHWC",u);y.assert(s.rank===F.length,()=>`Error in ${jd}, dy must have the same rank as output ${F.length}, but got ${s.rank}`);let P=y.toNestedArray(F,l.data.get(s.dataId).values),V=y.makeZerosNestedTypedArray(n.shape,n.dtype);for(let W=0;W<m;++W)for(let q=0;q<g;++q){let H=q*w-b.top;for(let j=0;j<x;++j){let Y=j*C-b.left;for(let Z=0;Z<h;++Z){let et=Number.MIN_SAFE_INTEGER,rt=H<0?0:H,ot=Y<0?0:Y;for(let at=0;at<N;++at){let nt=H+at*A;if(nt>=0&&nt<f)for(let it=0;it<_;++it){let dt=Y+it*$;if(dt>=0&&dt<d){let ht=c[W][nt][dt][Z]+p[at][it][Z];ht>et&&(et=ht,rt=nt,ot=dt)}}}V[W][rt][ot][Z]+=P[W][q][j][Z]}}}return{dataId:l.write(y.toTypedArray(V,n.dtype),n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};function hl(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,"sum");let a;o.dtype==="bool"?a=$o({inputs:{x:o},backend:e,attrs:{dtype:"int32"}}):a=Kr({inputs:{x:o},backend:e});let u=a.shape.length,l=y.parseAxisParam(s,a.shape),c=v.getAxesPermutation(l,u),p=l,m=a;c!=null&&(m=Ve({inputs:{x:a},backend:e,attrs:{perm:c}}),p=v.getInnerMostAxes(p.length,u)),v.assertAxesAreInnerMostDims("sum",p,m.shape.length);let[f,d]=v.computeOutAndReduceShapes(m.shape,p),h=v.upcastType(m.dtype,"int32"),g=nd(e,f,h),x=y.sizeFromShape(d),b=e.data.get(g.dataId).values,w=e.data.get(m.dataId).values;for(let C=0;C<b.length;++C){let N=C*x,_=0;for(let A=0;A<x;++A)_+=w[N+A];b[C]=_}if(i){let C=v.expandShapeToKeepDim(g.shape,l),N=g;g=Yt({inputs:{x:g},backend:e,attrs:{shape:C}}),e.disposeIntermediateTensorInfo(N)}return e.disposeIntermediateTensorInfo(a),c!=null&&e.disposeIntermediateTensorInfo(m),g}var mO={kernelName:$s,backendName:"cpu",kernelFunc:hl};function aQ(r){let{inputs:t,backend:e,attrs:n}=r,{equation:o}=n,s=t,{allDims:i,summedDims:a,idDims:u}=v.decodeEinsumEquation(o,s.length);v.checkEinsumDimSizes(i.length,u,s);let{path:l,steps:c}=v.getEinsumComputePath(a,u),p=c.length,m=null,f=i.length,d=[];for(let h=0;h<p;++h){for(let g of c[h]){let{permutationIndices:x,expandDims:b}=v.getEinsumPermutation(f,u[g]),w;v.isIdentityPermutation(x)?w=s[g]:(w=Ve({inputs:{x:s[g]},backend:e,attrs:{perm:x}}),d.push(w));let C=w.shape.slice();for(let N=0;N<b.length;++N)C.splice(b[N],0,1);y.arraysEqual(w.shape,C)||(w=Yt({inputs:{x:w},backend:e,attrs:{shape:C}}),d.push(w)),m===null?m=w:(m=_c({inputs:{a:w,b:m},backend:e}),d.push(m))}h<p-1&&(l[h]>=0&&(m=hl({inputs:{x:m},backend:e,attrs:{axis:l[h]-(i.length-f),keepDims:!1}}),d.push(m)),f--)}for(let h of d)h!==m&&e.disposeIntermediateTensorInfo(h);return m}var fO={kernelName:bp,backendName:"cpu",kernelFunc:aQ};function lQ(r){let{inputs:t,backend:e}=r,{dy:n,y:o}=t;tt([n,o],"eluGrad");let s=new Float32Array(y.sizeFromShape(o.shape)),i=e.data.get(o.dataId).values,a=e.data.get(n.dataId).values;for(let u=0;u<i.length;++u){let l=i[u];l>=1?s[u]=a[u]:s[u]=a[u]*(l+1)}return e.makeTensorInfo(o.shape,"float32",s)}var dO={kernelName:wp,backendName:"cpu",kernelFunc:lQ};var uQ=v.ERF_P,cQ=v.ERF_A1,pQ=v.ERF_A2,mQ=v.ERF_A3,fQ=v.ERF_A4,dQ=v.ERF_A5,hQ=kt(ga,r=>{let t=Math.sign(r),e=Math.abs(r),n=1/(1+uQ*e);return t*(1-((((dQ*n+fQ)*n+mQ)*n+pQ)*n+cQ)*n*Math.exp(-e*e))}),hO={kernelName:ga,backendName:"cpu",kernelFunc:hQ};function cd(r){let{inputs:t,backend:e,attrs:n}=r,{input:o}=t,{dim:s}=n,i=o.shape.length,a=o.shape.slice(),u=s;return s<0&&(y.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),a.splice(u,0,1),Yt({inputs:{x:o},backend:e,attrs:{shape:a}})}var gO={kernelName:ui,backendName:"cpu",kernelFunc:cd};var gQ=Qt((r,t)=>r/t),Wh=oe(Qo,gQ),Uh={kernelName:Qo,backendName:"cpu",kernelFunc:Wh};function hw(r,t,e){let n=r.shape,o=n[0],s=n[1],i=e.data.get(r.dataId),a=i.complexTensorInfos.real,u=i.complexTensorInfos.imag,l=[o,s],c=y.sizeFromShape(l),p=y.getTypedArrayFromDType("float32",c),m=y.getTypedArrayFromDType("float32",c);for(let g=0;g<o;g++){let x=Fo({inputs:{x:a},backend:e,attrs:{begin:[g,0],size:[1,s]}}),b=Fo({inputs:{x:u},backend:e,attrs:{begin:[g,0],size:[1,s]}}),w=wr({inputs:{real:x,imag:b},backend:e}),{real:C,imag:N}=xQ(w,t,e),_=v.mergeRealAndImagArrays(C,N);for(let A=0;A<s;A++){let $=v.getComplexWithIndex(_,A);p[g*s+A]=$.real,m[g*s+A]=$.imag}e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(b),e.disposeIntermediateTensorInfo(w)}let f=e.makeTensorInfo(l,"float32",p),d=e.makeTensorInfo(l,"float32",m),h=wr({inputs:{real:f,imag:d},backend:e});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),h}function xQ(r,t,e){let n=y.sizeFromShape(r.shape),o=e.data.get(r.dataId),s=e.data.get(o.complexTensorInfos.real.dataId).values,i=e.data.get(o.complexTensorInfos.imag.dataId).values;if(yQ(n)){let a=gT(s,i,n,t,e),u=[r.shape[0],r.shape[1]];if(t){let l=e.makeTensorInfo(u,"float32",a.real),c=e.makeTensorInfo(u,"float32",a.imag),p=e.makeTensorInfo([],"float32",y.createScalarValue(n,"float32")),m=Kr({inputs:{x:p},backend:e}),f=Uh.kernelFunc({inputs:{a:l,b:p},backend:e}),d=Uh.kernelFunc({inputs:{a:c,b:m},backend:e}),h=e.data.get(f.dataId).values,g=e.data.get(d.dataId).values;return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),{real:h,imag:g}}return a}else{let a=v.mergeRealAndImagArrays(s,i),u=bQ(a,n,t);return v.splitRealAndImagArrays(u)}}function yQ(r){return(r&r-1)===0}function gT(r,t,e,n,o){if(e===1)return{real:r,imag:t};let s=v.mergeRealAndImagArrays(r,t),i=e/2,a=v.complexWithEvenIndex(s),u=a.real,l=a.imag,c=[u.length],p=o.makeTensorInfo(c,"float32",u),m=o.makeTensorInfo(c,"float32",l),f=wr({inputs:{real:p,imag:m},backend:o}),d=v.complexWithOddIndex(s),h=d.real,g=d.imag,x=[h.length],b=o.makeTensorInfo(x,"float32",h),w=o.makeTensorInfo(x,"float32",g),C=wr({inputs:{real:b,imag:w},backend:o}),N=gT(u,l,i,n,o),_=N.real,A=N.imag,$=[_.length],F=o.makeTensorInfo($,"float32",_),P=o.makeTensorInfo($,"float32",A),V=wr({inputs:{real:F,imag:P},backend:o}),G=gT(h,g,i,n,o),W=G.real,q=G.imag,H=[W.length],j=o.makeTensorInfo(H,"float32",W),Y=o.makeTensorInfo(H,"float32",q),Z=wr({inputs:{real:j,imag:Y},backend:o}),et=v.exponents(e,n),rt=[et.real.length],ot=o.makeTensorInfo(rt,"float32",et.real),at=o.makeTensorInfo(rt,"float32",et.imag),nt=wr({inputs:{real:ot,imag:at},backend:o}),it=_c({inputs:{a:nt,b:Z},backend:o}),dt=Ki({inputs:{a:V,b:it},backend:o}),ht=Vh({inputs:{a:V,b:it},backend:o}),bt=Ao({inputs:{input:dt},backend:o}),Et=Ao({inputs:{input:ht},backend:o}),At=ji({inputs:{input:dt},backend:o}),Vt=ji({inputs:{input:ht},backend:o}),Zt=Tu({inputs:[bt,Et],backend:o,attrs:{axis:0}}),ce=Tu({inputs:[At,Vt],backend:o,attrs:{axis:0}}),he=o.data.get(Zt.dataId).values,jt=o.data.get(ce.dataId).values;return o.disposeIntermediateTensorInfo(p),o.disposeIntermediateTensorInfo(m),o.disposeIntermediateTensorInfo(f),o.disposeIntermediateTensorInfo(b),o.disposeIntermediateTensorInfo(w),o.disposeIntermediateTensorInfo(C),o.disposeIntermediateTensorInfo(F),o.disposeIntermediateTensorInfo(P),o.disposeIntermediateTensorInfo(V),o.disposeIntermediateTensorInfo(j),o.disposeIntermediateTensorInfo(Y),o.disposeIntermediateTensorInfo(Z),o.disposeIntermediateTensorInfo(ot),o.disposeIntermediateTensorInfo(at),o.disposeIntermediateTensorInfo(nt),o.disposeIntermediateTensorInfo(it),o.disposeIntermediateTensorInfo(dt),o.disposeIntermediateTensorInfo(ht),o.disposeIntermediateTensorInfo(bt),o.disposeIntermediateTensorInfo(At),o.disposeIntermediateTensorInfo(Et),o.disposeIntermediateTensorInfo(Vt),o.disposeIntermediateTensorInfo(Zt),o.disposeIntermediateTensorInfo(ce),{real:he,imag:jt}}function bQ(r,t,e){let n=new Float32Array(t*2);for(let o=0;o<t;o++){let s=0,i=0;for(let a=0;a<t;a++){let u=v.exponent(o*a,t,e),l=v.getComplexWithIndex(r,a);s+=l.real*u.real-l.imag*u.imag,i+=l.real*u.imag+l.imag*u.real}e&&(s/=t,i/=t),v.assignToTypedArray(n,s,i,o)}return n}function wQ(r){let{inputs:t,backend:e}=r,{input:n}=t,o=y.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=o/s,a=Yt({inputs:{x:n},backend:e,attrs:{shape:[i,s]}}),u=hw(a,!1,e),l=Yt({inputs:{x:u},backend:e,attrs:{shape:n.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(u),l}var xO={kernelName:Cp,backendName:"cpu",kernelFunc:wQ};function Hh(r){let{backend:t,attrs:e}=r,{shape:n,value:o,dtype:s}=e,i=s||y.inferDtype(o),a=y.getArrayFromDType(i,y.sizeFromShape(n));return CQ(a,o,i),t.makeTensorInfo(n,i,a)}var yO={kernelName:Dl,backendName:"cpu",kernelFunc:Hh};function CQ(r,t,e){r.fill(t)}var bO={kernelName:ba,backendName:"cpu",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,o=e,s=y.getTypedArrayFromDType(n.dtype,y.sizeFromShape(n.shape)),[i,a,u,l]=n.shape,c=o.data.get(n.dataId).values;for(let m=0;m<i;m++){let f=m*u*a*l;for(let d=0;d<a;d++){let h=d*(u*l);for(let g=0;g<u;g++){let x=g*l;for(let b=0;b<l;b++){let w=Math.round(u-g-1),C=f+h+x+b,N=c[C];if(w>=0&&w<u){let _=w*l,A=f+h+_+b;N=c[A]}s[C]=N}}}}return{dataId:o.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};var IQ=Qt((r,t)=>Math.floor(r/t)),SQ=oe(ns,IQ,null,"int32"),wO={kernelName:ns,backendName:"cpu",kernelFunc:SQ};function vQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=dT({inputs:{x:o,filter:s},backend:e,attrs:{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m}});if(i){let g=h;if(c==="NCHW"&&i.shape.length===1&&i.shape[0]!==1){let x=Yt({inputs:{x:i},backend:e,attrs:{shape:[i.shape[0],1,1]}});h=Ki({inputs:{a:h,b:x},backend:e}),e.disposeIntermediateTensorInfo(x)}else h=Ki({inputs:{a:h,b:i},backend:e});e.disposeIntermediateTensorInfo(g)}if(f){let g=h;if(c==="NCHW"&&f==="prelu"&&a.shape.length===1&&a.shape[0]!==1){let x=Yt({inputs:{x:a},backend:e,attrs:{shape:[a.shape[0],1,1]}});h=Oc(e,h,f,x,d),e.disposeIntermediateTensorInfo(x)}else h=Oc(e,h,f,a,d);e.disposeIntermediateTensorInfo(g)}return h}var CO={kernelName:Ii,backendName:"cpu",kernelFunc:vQ};function NQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=hT({inputs:{x:o,filter:s},backend:e,attrs:{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m}});if(i){let g=h;h=Ki({inputs:{a:h,b:i},backend:e}),e.disposeIntermediateTensorInfo(g)}if(f){let g=h;h=Oc(e,h,f,a,d),e.disposeIntermediateTensorInfo(g)}return h}var IO={kernelName:Si,backendName:"cpu",kernelFunc:NQ};function TQ(r){let{inputs:t,backend:e}=r,{params:n,indices:o}=t,s=y.sizeFromShape(n.shape),i=o.shape,a=i[i.length-1],[u,l,c,p]=v.prepareAndValidate(n,o);if(l===0)return e.makeTensorInfo(u,n.dtype,[]);let m=e.data.get(o.dataId).values,f=e.bufferSync(n),d=Qb(m,f,n.dtype,l,a,c,p,n.shape,s);return e.makeTensorInfo(u,n.dtype,d.values)}var SO={kernelName:wa,backendName:"cpu",kernelFunc:TQ};function kQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,indices:s}=t,{axis:i,batchDims:a}=n;tt([o,s],"gatherV2");let u=y.parseAxisParam(i,o.shape)[0],l=e.data.get(s.dataId).values,c=o.shape[u];for(let C=0;C<l.length;++C){let N=l[C];y.assert(N<=c-1&&N>=0,()=>`GatherV2: the index value ${N} is not in [0, ${c-1}]`)}let p=a;a==null&&(p=0);let m=y.sizeFromShape(s.shape),f=v.segment_util.collectGatherOpShapeInfo(o,s,u,p),d=Yt({inputs:{x:o},backend:e,attrs:{shape:[f.batchSize,f.outerSize,f.dimSize,f.sliceSize]}}),h=Yt({inputs:{x:s},backend:e,attrs:{shape:[f.batchSize,m/f.batchSize]}}),g=[f.batchSize,f.outerSize,m/f.batchSize,f.sliceSize],x=e.bufferSync(h),b=e.bufferSync(d),w=tw(b,x,g);return e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),e.makeTensorInfo(f.outputShape,w.dtype,w.values)}var vO={kernelName:ci,backendName:"cpu",kernelFunc:kQ};function EQ(r){let{inputs:t,backend:e}=r,{input:n}=t,o=y.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=o/s,a=Yt({inputs:{x:n},backend:e,attrs:{shape:[i,s]}}),u=hw(a,!0,e),l=Yt({inputs:{x:u},backend:e,attrs:{shape:n.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(u),l}var NO={kernelName:Ip,backendName:"cpu",kernelFunc:EQ};var _Q=kt(Ia,r=>Number.isFinite(r)?1:0,"bool"),TO={kernelName:Ia,backendName:"cpu",kernelFunc:_Q};var AQ=kt(Sa,r=>Math.abs(r)===1/0?1:0,"bool"),kO={kernelName:Sa,backendName:"cpu",kernelFunc:AQ};var $Q=kt(va,r=>Number.isNaN(r)?1:0,"bool"),EO={kernelName:va,backendName:"cpu",kernelFunc:$Q};function DQ(r){let{backend:t,attrs:e}=r,{start:n,stop:o,num:s}=e,i=ew(n,o,s);return t.makeTensorInfo([i.length],"float32",i)}var _O={kernelName:vp,backendName:"cpu",kernelFunc:DQ};var RQ=kt(ka,r=>Math.log1p(r)),AO={kernelName:ka,backendName:"cpu",kernelFunc:RQ};var FQ=Qt((r,t)=>r&&t),OQ=oe(Ea,FQ,null,"bool"),$O={kernelName:Ea,backendName:"cpu",kernelFunc:OQ};var PQ=kt(_a,r=>r?0:1,"bool"),DO={kernelName:_a,backendName:"cpu",kernelFunc:PQ};var LQ=Qt((r,t)=>r||t),MQ=oe(Aa,LQ,null,"bool"),RO={kernelName:Aa,backendName:"cpu",kernelFunc:MQ};function zQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{depthRadius:s,bias:i,alpha:a,beta:u}=n;tt(o,"LRN");let l=o.shape[3],c=l-1,p=e.data.get(o.dataId).values,m=y.sizeFromShape(o.shape),f=new Float32Array(m);function d(h){let g=h%l,x=h-g+Math.max(0,g-s),b=h-g+Math.min(g+s,c),w=0;for(;x<=b;x++){let C=p[x];w+=C*C}return w}for(let h=0;h<m;h++){let g=d(h),x=p[h]*Math.pow(i+a*g,-u);f[h]=x}return e.makeTensorInfo(o.shape,o.dtype,f)}var FO={kernelName:Rl,backendName:"cpu",kernelFunc:zQ};function BQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,y:s,dy:i}=t,{depthRadius:a,bias:u,alpha:l,beta:c}=n;tt(i,"LRNGrad");let p=y.sizeFromShape(i.shape),m=i.shape[3],f=e.data.get(i.dataId).values,d=e.data.get(o.dataId).values,h=e.data.get(s.dataId).values,g=new Float32Array(p),x=p;for(let b=0;b<x;b++){let w=b%m,C=b-w+Math.max(0,w-a),N=b-w+Math.min(m,w+a+1),_=0;for(let A=C;A<N;A++)_+=Math.pow(d[A],2);_=l*_+u;for(let A=C;A<N;A++){let $=-2*l*c*d[A]*h[b]/_;b===A&&($+=Math.pow(_,-c)),$*=f[b],g[A]+=$}}return e.makeTensorInfo(i.shape,o.dtype,g)}var OO={kernelName:Np,backendName:"cpu",kernelFunc:BQ};function xT(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reductionIndices:s,keepDims:i}=n,a=e,u=o.shape,l=u.length,c=y.parseAxisParam(s,u),p=c,m=v.getAxesPermutation(p,l),f=a.data.get(o.dataId).values;if(m!=null){let C=new Array(l);for(let N=0;N<C.length;N++)C[N]=u[m[N]];f=id(f,u,o.dtype,m,C),p=v.getInnerMostAxes(p.length,l),u=C}tt(o,"max"),v.assertAxesAreInnerMostDims("max",p,l);let[d,h]=v.computeOutAndReduceShapes(u,p),g=y.sizeFromShape(h),x=rw(f,g,d,o.dtype),b=a.write(x,d,o.dtype),w=d;return i&&(w=v.expandShapeToKeepDim(d,c)),{dataId:b,shape:w,dtype:o.dtype}}var PO={kernelName:ls,backendName:"cpu",kernelFunc:xT};function VQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;tt(o,"maxPool");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u),p;if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))p=Kr({inputs:{x:o},backend:e});else{let m=e.data.get(o.dataId).values,f=y.computeStrides(o.shape),d=ud(m,o.shape,o.dtype,f,c,"max");p=e.makeTensorInfo(c.outShape,o.dtype,d.values)}return p}var LO={kernelName:cs,backendName:"cpu",kernelFunc:VQ};function GQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n;tt(o,"maxPool3d");let c=v.computePool3DInfo(o.shape,s,i,1,a,u,l),p=e.data.get(o.dataId).values,m=dw(p,o.shape,o.dtype,y.computeStrides(o.shape),c,"max");return e.makeTensorInfo(m.shape,"float32",m.values)}var MO={kernelName:Fl,backendName:"cpu",kernelFunc:GQ};function WQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=n;tt([o,s],"maxPool3DGrad");let c=v.computePool3DInfo(s.shape,i,a,1,u,l),p=e.bufferSync(s),m=RF(p,c),f=c.strideDepth,d=c.strideHeight,h=c.strideWidth,g=c.dilationDepth,x=c.dilationHeight,b=c.dilationWidth,w=c.effectiveFilterDepth,C=c.effectiveFilterHeight,N=c.effectiveFilterWidth,_=w-1-c.padInfo.front,A=N-1-c.padInfo.left,$=C-1-c.padInfo.top,F=wt(s.shape,"float32"),P=e.bufferSync(o);for(let V=0;V<c.batchSize;++V)for(let G=0;G<c.inChannels;++G)for(let W=0;W<c.inDepth;++W)for(let q=0;q<c.inHeight;++q)for(let H=0;H<c.inWidth;++H){let j=W-_,Y=q-$,Z=H-A,et=0;for(let rt=0;rt<w;rt+=g){let ot=(j+rt)/f;if(!(ot<0||ot>=c.outDepth||Math.floor(ot)!==ot))for(let at=0;at<C;at+=x){let nt=(Y+at)/d;if(!(nt<0||nt>=c.outHeight||Math.floor(nt)!==nt))for(let it=0;it<N;it+=b){let dt=(Z+it)/h;if(dt<0||dt>=c.outWidth||Math.floor(dt)!==dt)continue;let ht=w*C*N-1-m.get(V,ot,nt,dt,G),bt=rt*C*N+at*N+it,Et=ht===bt?1:0;if(Et===0)continue;et+=P.get(V,ot,nt,dt,G)*Et}}}F.set(et,V,W,q,H,G)}return e.makeTensorInfo(F.shape,F.dtype,F.values)}var zO={kernelName:kp,backendName:"cpu",kernelFunc:WQ};function UQ(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s,output:i}=t,a=s;tt([s,i],"maxPoolGrad");let{filterSize:u,strides:l,pad:c,dimRoundingMode:p}=n,m=v.computePool2DInfo(a.shape,u,l,1,c,p),f=e.data.get(a.dataId).values,d=wt(m.outShape,a.dtype,fw(f,a.shape,a.dtype,m).values),h=m.strideHeight,g=m.strideWidth,x=m.dilationHeight,b=m.dilationWidth,w=m.effectiveFilterHeight,C=m.effectiveFilterWidth,N=C-1-m.padInfo.left,_=w-1-m.padInfo.top,A=wt(a.shape,"float32"),$=e.data.get(o.dataId).values,F=wt(o.shape,"float32",$);for(let P=0;P<m.batchSize;++P)for(let V=0;V<m.inChannels;++V)for(let G=0;G<m.inHeight;++G)for(let W=0;W<m.inWidth;++W){let q=G-_,H=W-N,j=0;for(let Y=0;Y<w;Y+=x){let Z=(q+Y)/h;if(!(Z<0||Z>=m.outHeight||Math.floor(Z)!==Z))for(let et=0;et<C;et+=b){let rt=(H+et)/g;if(rt<0||rt>=m.outWidth||Math.floor(rt)!==rt)continue;let ot=w*C-1-d.get(P,Z,rt,V),at=Y*C+et,nt=ot===at?1:0;if(nt===0)continue;j+=F.get(P,Z,rt,V)*nt}}A.set(j,P,G,W,V)}return e.makeTensorInfo(A.shape,A.dtype,A.values)}var BO={kernelName:Tp,backendName:"cpu",kernelFunc:UQ};function VO(r,t,e,n,o){let s=y.computeStrides(t),i=ud(r,t,e,s,o,"max"),a=fw(r,t,e,o,!0,n);return[i.values,a.values]}var GO={kernelName:Ep,backendName:"cpu",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{filterSize:o,strides:s,pad:i,includeBatchInIndex:a}=t,u=e;tt(n,"MaxPoolWithArgmax");let l=u.data.get(n.dataId).values,c=v.computePool2DInfo(n.shape,o,s,[1,1],i),[p,m]=VO(l,n.shape,n.dtype,a,c),f=u.write(p,c.outShape,n.dtype),d=u.write(m,c.outShape,n.dtype);return[{dataId:f,shape:c.outShape,dtype:n.dtype},{dataId:d,shape:c.outShape,dtype:"int32"}]}};function HQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=y.parseAxisParam(s,o.shape),l=v.computeOutAndReduceShapes(o.shape,a)[1],c=y.sizeFromShape(l),p=[],m=e.makeTensorInfo([],"float32",new Float32Array([c]));p.push(m);let f=$o({inputs:{x:o},backend:e,attrs:{dtype:"float32"}});p.push(f);let d=Wh({inputs:{a:f,b:m},backend:e});p.push(d);let h=hl({inputs:{x:d},backend:e,attrs:{axis:s,keepDims:i}});return p.forEach(g=>e.disposeIntermediateTensorInfo(g)),h}var WO={kernelName:ps,backendName:"cpu",kernelFunc:HQ};function qQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;tt(o,"min");let a=y.parseAxisParam(s,o.shape),u=a,l=v.getAxesPermutation(u,o.shape.length),c=o;l!=null&&(c=Ve({inputs:{x:o},backend:e,attrs:{perm:l}}),u=v.getInnerMostAxes(u.length,o.shape.length)),v.assertAxesAreInnerMostDims("min",u,c.shape.length);let[p,m]=v.computeOutAndReduceShapes(c.shape,u),f=y.sizeFromShape(m),d=y.makeZerosTypedArray(y.sizeFromShape(p),c.dtype),h=e.data.get(c.dataId).values;for(let x=0;x<d.length;++x){let b=x*f,w=h[b];for(let C=0;C<f;++C){let N=h[b+C];(Number.isNaN(N)||N<w)&&(w=N)}d[x]=w}l!=null&&e.disposeIntermediateTensorInfo(c);let g=e.makeTensorInfo(p,c.dtype,d);if(i){let x=v.expandShapeToKeepDim(p,a),b=Yt({inputs:{x:g},backend:e,attrs:{shape:x}});return e.disposeIntermediateTensorInfo(g),b}return g}var UO={kernelName:ms,backendName:"cpu",kernelFunc:qQ};function KQ(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,mode:i}=n;tt(o,"mirrorPad");let a=s.map((w,C)=>w[0]+o.shape[C]+w[1]),u=s.map(w=>w[0]),l=s.map((w,C)=>w[0]+o.shape[C]),c=i==="reflect"?0:1,p=e.data.get(o.dataId).values,m=o.shape.length,f=y.computeStrides(o.shape),d=y.sizeFromShape(a),h=a.length,g=y.computeStrides(a),x=y.getTypedArrayFromDType(o.dtype,d);for(let w=0;w<d;w++){let C=y.indexToLoc(w,h,g);for(let _=0;_<h;_++)C[_]<u[_]?C[_]=u[_]*2-C[_]-c:C[_]>=l[_]&&(C[_]=(l[_]-1)*2-C[_]+c);C=C.map((_,A)=>_-u[A]);let N=y.locToIndex(C,m,f);x[w]=p[N]}return{dataId:e.write(x,a,o.dtype),shape:a,dtype:o.dtype}}var HO={kernelName:ds,backendName:"cpu",kernelFunc:KQ};var jQ=Qt((r,t)=>{let e=r%t;return r<0&&t<0||r>=0&&t>=0?e:(e+t)%t}),XQ=oe($a,jQ),qO={kernelName:$a,backendName:"cpu",kernelFunc:XQ};var jO=Tl(fh());function yT(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{dim:s}=n,i=o.shape.length,a=s;if(a===-1&&(a=i-1),a!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${a}`);let u=y.parseAxisParam([a],o.shape),l=xT({inputs:{x:o},backend:e,attrs:{reductionIndices:u,keepDims:!1}}),c=v.expandShapeToKeepDim(l.shape,u),p=Yt({inputs:{x:l},backend:e,attrs:{shape:c}}),m=Vh({inputs:{a:o,b:p},backend:e}),f=UN({inputs:{x:m},backend:e}),d=hl({inputs:{x:f},backend:e,attrs:{axis:u,keepDims:!1}}),h=Yt({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Wh({inputs:{a:f,b:h},backend:e});return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(h),g}var KO={kernelName:Ds,backendName:"cpu",kernelFunc:yT};function YQ(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{numSamples:s,seed:i,normalized:a}=n;tt(o,"multinomial");let u=a?o:yT({inputs:{logits:o},backend:e,attrs:{dim:-1}}),l=u.shape[0],c=u.shape[1],p=e.data.get(u.dataId).values,m=[l,s],f=y.makeZerosTypedArray(y.sizeFromShape(m),"int32");for(let d=0;d<l;++d){let h=d*c,g=new Float32Array(c-1);g[0]=p[h];for(let w=1;w<g.length;++w)g[w]=g[w-1]+p[h+w];let x=jO.alea(i.toString()),b=d*s;for(let w=0;w<s;++w){let C=x();f[b+w]=g.length;for(let N=0;N<g.length;N++)if(C<g[N]){f[b+w]=N;break}}}return a||e.disposeIntermediateTensorInfo(u),e.makeTensorInfo(m,"int32",f)}var XO={kernelName:_p,backendName:"cpu",kernelFunc:YQ};var ZQ=Ur.nonMaxSuppressionV3Impl;function JQ(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u}=n;tt(o,"NonMaxSuppression");let l=e.data.get(o.dataId).values,c=e.data.get(s.dataId).values,{selectedIndices:p}=ZQ(l,c,i,a,u);return e.makeTensorInfo([p.length],"int32",new Int32Array(p))}var YO={kernelName:Ra,backendName:"cpu",kernelFunc:JQ};var QQ=Ur.nonMaxSuppressionV4Impl;function t9(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,padToMaxOutputSize:l}=n;tt(o,"NonMaxSuppressionPadded");let c=e.data.get(o.dataId).values,p=e.data.get(s.dataId).values,{selectedIndices:m,validOutputs:f}=QQ(c,p,i,a,u,l);return[e.makeTensorInfo([m.length],"int32",new Int32Array(m)),e.makeTensorInfo([],"int32",new Int32Array([f]))]}var ZO={kernelName:Fa,backendName:"cpu",kernelFunc:t9};var e9=Ur.nonMaxSuppressionV5Impl;function r9(r){let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,softNmsSigma:l}=n;tt(o,"NonMaxSuppressionWithScore");let c=e.data.get(o.dataId).values,p=e.data.get(s.dataId).values,m=i,f=a,d=u,h=l,{selectedIndices:g,selectedScores:x}=e9(c,p,m,f,d,h);return[e.makeTensorInfo([g.length],"int32",new Int32Array(g)),e.makeTensorInfo([x.length],"float32",new Float32Array(x))]}var JO={kernelName:Oa,backendName:"cpu",kernelFunc:r9};function n9(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n;tt(o,"oneHot");let l=y.sizeFromShape(o.shape),c=new Float32Array(l*i);c.fill(u);let p=e.data.get(o.dataId).values;for(let m=0;m<l;++m)p[m]>=0&&p[m]<i&&(c[m*i+p[m]]=a);return e.makeTensorInfo([...o.shape,i],s,c)}var QO={kernelName:gs,backendName:"cpu",kernelFunc:n9};function qh(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype==="string")throw new Error("zerosLike is not supported for string tensors");if(n.dtype==="complex64"){let o=Ao({inputs:{input:n},backend:e}),s=qh({inputs:{x:o},backend:e}),i=ji({inputs:{input:n},backend:e}),a=qh({inputs:{x:i},backend:e}),u=wr({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Hh({backend:e,attrs:{shape:n.shape,value:0,dtype:n.dtype}})}var tP={kernelName:wi,backendName:"cpu",kernelFunc:qh};function eP(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype==="string")throw new Error("onesLike is not supported for string tensors");if(n.dtype==="complex64"){let o=Ao({inputs:{input:n},backend:e}),s=eP({inputs:{x:o},backend:e}),i=ji({inputs:{input:n},backend:e}),a=qh({inputs:{x:i},backend:e}),u=wr({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Hh({backend:e,attrs:{shape:n.shape,value:1,dtype:n.dtype}})}var rP={kernelName:mi,backendName:"cpu",kernelFunc:eP};function bT(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return cd({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,"All tensors passed to stack must have matching shapes"),y.assert(i===c.dtype,()=>"All tensors passed to stack must have matching dtypes")});let a=[],u=t.map(c=>{let p=cd({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=Tu({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeIntermediateTensorInfo(c)),l}var nP={kernelName:fi,backendName:"cpu",kernelFunc:bT};function o9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,constantValue:i}=n;tt(o,"pad");let a=s.map((b,w)=>b[0]+o.shape[w]+b[1]),u=s.map(b=>b[0]),l=e.data.get(o.dataId).values,c=y.sizeFromShape(o.shape),p=o.shape.length,m=y.computeStrides(o.shape),f=y.sizeFromShape(a),d=a.length,h=y.computeStrides(a),g=y.getTypedArrayFromDType(o.dtype,f);i!==0&&g.fill(i);for(let b=0;b<c;b++){let C=y.indexToLoc(b,p,m).map((_,A)=>_+u[A]),N=y.locToIndex(C,d,h);g[N]=l[b]}return{dataId:e.write(g,a,o.dtype),shape:a,dtype:o.dtype}}var gw={kernelName:xs,backendName:"cpu",kernelFunc:o9};var s9=Qt((r,t)=>Math.pow(r,t)),i9=oe(ys,s9),oP={kernelName:ys,backendName:"cpu",kernelFunc:i9};function a9(r){let{inputs:t,backend:e,attrs:n}=r,{paramsNestedSplits:o,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:a}=n,u=o.map(x=>e.data.get(x.dataId).values),l=o.map(x=>x.shape),c=e.data.get(s.dataId).values,p=e.data.get(i.dataId).values,[m,f,d]=nw(u,l,c,s.shape,s.dtype,p,i.shape,a),h=m.map(x=>e.makeTensorInfo([x.length],"int32",x)),g=e.makeTensorInfo(d,s.dtype,f);return h.concat([g])}var sP={kernelName:Ap,backendName:"cpu",kernelFunc:a9};function l9(r){let{inputs:t,backend:e}=r,{starts:n,limits:o,deltas:s}=t,i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ow(i,n.shape,n.dtype,a,o.shape,u,s.shape),p=e.makeTensorInfo([l.length],"int32",l),m=e.makeTensorInfo([c.length],n.dtype,c);return[p,m]}var iP={kernelName:$p,backendName:"cpu",kernelFunc:l9};function u9(r){let{inputs:t,backend:e,attrs:n}=r,{shape:o,values:s,defaultValue:i,rowPartitionTensors:a}=t,{rowPartitionTypes:u}=n,l=e.data.get(o.dataId).values,c=e.data.get(s.dataId).values,p=e.data.get(i.dataId).values,m=a.map(g=>e.data.get(g.dataId).values),f=a.map(g=>g.shape),[d,h]=sw(l,o.shape,c,s.shape,s.dtype,p,i.shape,m,f,u);return e.makeTensorInfo(d,s.dtype,h)}var aP={kernelName:Dp,backendName:"cpu",kernelFunc:u9};function c9(r){let{backend:t,attrs:e}=r,{start:n,stop:o,dtype:s,step:i}=e,a=Ac(n,o,i,s);return t.makeTensorInfo([a.length],s,a)}var lP={kernelName:Ol,backendName:"cpu",kernelFunc:c9};var p9=kt(Pa,r=>1/r),uP={kernelName:Pa,backendName:"cpu",kernelFunc:p9};function m9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n;tt(o,"resizeBilinear");let u=y.computeStrides(o.shape),[l,c]=a,[p,m,f,d]=o.shape,h=e.data.get(o.dataId).values,g=new Float32Array(y.sizeFromShape([p,l,c,d])),x=[s&&l>1?m-1:m,s&&c>1?f-1:f],b=[s&&l>1?l-1:l,s&&c>1?c-1:c],w=0,C=x[0]/b[0],N=x[1]/b[1];for(let _=0;_<p;_++)for(let A=0;A<l;A++){let $;i?$=C*(A+.5)-.5:$=C*A;let F=Math.max(0,Math.floor($)),P=$-F,V=Math.min(m-1,Math.ceil($)),G=_*u[0]+F*u[1],W=_*u[0]+V*u[1];for(let q=0;q<c;q++){let H;i?H=N*(q+.5)-.5:H=N*q;let j=Math.max(0,Math.floor(H)),Y=H-j,Z=Math.min(f-1,Math.ceil(H)),et=G+j*u[2],rt=W+j*u[2],ot=G+Z*u[2],at=W+Z*u[2];for(let nt=0;nt<d;nt++){let it=h[et+nt],dt=h[rt+nt],ht=h[ot+nt],bt=h[at+nt],Et=it+(ht-it)*Y,At=dt+(bt-dt)*Y,Vt=Et+(At-Et)*P;g[w++]=Vt}}}return e.makeTensorInfo([p,l,c,d],"float32",g)}var cP={kernelName:Ss,backendName:"cpu",kernelFunc:m9};function f9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n;tt([s,o],"resizeBilinearGrad");let a=y.computeStrides(o.shape),[u,l,c,p]=o.shape,[,m,f]=s.shape,d=new Float32Array(u*l*c*p),h=[i&&m>1?l-1:l,i&&f>1?c-1:c],g=[i&&m>1?m-1:m,i&&f>1?f-1:f],x=h[0]/g[0],b=h[1]/g[1],w=e.data.get(s.dataId).values,C=0;for(let N=0;N<u;N++){let _=N*a[0];for(let A=0;A<m;A++){let $=A*x,F=Math.floor($),P=Math.min(Math.ceil($),l-1),V=_+F*a[1],G=_+P*a[1],W=$-F,q=1-W;for(let H=0;H<f;H++){let j=H*b,Y=Math.floor(j),Z=Math.min(Math.ceil(j),c-1),et=j-Y,rt=1-et,ot=V+Y*a[2],at=V+Z*a[2],nt=G+Y*a[2],it=G+Z*a[2],dt=q*rt,ht=q*et,bt=W*rt,Et=W*et;for(let At=0;At<p;At++){let Vt=w[C++];d[ot+At]+=Vt*dt,d[at+At]+=Vt*ht,d[nt+At]+=Vt*bt,d[it+At]+=Vt*Et}}}}return e.makeTensorInfo([u,c,l,p],"float32",d)}var pP={kernelName:Op,backendName:"cpu",kernelFunc:f9};function d9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n;tt(o,"resizeNearestNeighbor");let u=y.computeStrides(o.shape),[l,c]=a,[p,m,f,d]=o.shape,h=e.data.get(o.dataId).values,g=new Float32Array(p*l*c*d),x=[s&&l>1?m-1:m,s&&c>1?f-1:f],b=[s&&l>1?l-1:l,s&&c>1?c-1:c],w=x[0]/b[0],C=x[1]/b[1],N=0;for(let _=0;_<p;_++){let A=_*u[0];for(let $=0;$<l;$++){let F=i?w*($+.5):w*$,P=Math.min(m-1,s?Math.round(F):Math.floor(F));i&&(P=Math.max(0,P));let V=A+P*u[1];for(let G=0;G<c;G++){let W=i?C*(G+.5):C*G,q=Math.min(f-1,s?Math.round(W):Math.floor(W));i&&(q=Math.max(0,q));let H=V+q*u[2];for(let j=0;j<d;j++){let Y=h[H+j];g[N++]=Y}}}}return e.makeTensorInfo([p,l,c,d],o.dtype,g)}var mP={kernelName:Is,backendName:"cpu",kernelFunc:d9};function h9(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n;tt([s,o],"resizeNearestNeighborGrad");let a=y.computeStrides(o.shape),u=y.computeStrides(s.shape),[l,c,p,m]=o.shape,[,f,d]=s.shape,h=new Float32Array(l*c*p*m),g=e.data.get(s.dataId).values,x=[i&&f>1?c-1:c,i&&d>1?p-1:p],b=[i&&f>1?f-1:f,i&&d>1?d-1:d],w=x[0]/b[0],C=x[1]/b[1],N=1/w,_=1/C,A=Math.ceil(N)*2+2,$=Math.ceil(_)*2+2;for(let F=0;F<l;F++){let P=F*a[0];for(let V=0;V<c;V++){let G=P+V*a[1],W=Math.floor(V*N),q=Math.floor(W-A/2);for(let H=0;H<p;H++){let j=G+H*a[2],Y=Math.floor(H*_),Z=Math.floor(Y-$/2);for(let et=0;et<m;et++){let rt=0;for(let ot=0;ot<A;ot++){let at=ot+q;if(at<0||at>=f)continue;let nt=P+at*u[1],it=at*w,dt=Math.min(c-1,i?Math.round(it):Math.floor(it));if(V===dt)for(let ht=0;ht<$;ht++){let bt=ht+Z;if(bt<0||bt>=d)continue;let Et=nt+bt*u[2],At=bt*C,Vt=Math.min(p-1,i?Math.round(At):Math.floor(At));H===Vt&&(rt+=g[Et+et])}}h[j+et]=rt}}}}return e.makeTensorInfo(o.shape,o.dtype,h)}var fP={kernelName:Fp,backendName:"cpu",kernelFunc:h9};function g9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n;tt(o,"reverse");let i=o.shape.length,a=y.parseAxisParam(s,o.shape);if(i===0)return Kr({inputs:{x:o},backend:e});let u=new pe(o.shape,o.dtype),l=e.bufferSync(o);for(let c=0;c<u.size;c++){let p=u.indexToLoc(c),m=p.slice();a.forEach(f=>m[f]=o.shape[f]-1-m[f]),u.set(l.get(...m),...p)}return e.makeTensorInfo(u.shape,u.dtype,u.values)}var dP={kernelName:Ns,backendName:"cpu",kernelFunc:g9};var hP={kernelName:qa,backendName:"cpu",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,{radians:o,fillValue:s,center:i}=t,a=e,u=y.getTypedArrayFromDType(n.dtype,y.sizeFromShape(n.shape)),[l,c,p,m]=n.shape,[f,d]=v.getImageCenter(i,c,p),h=255,g=Math.sin(o),x=Math.cos(o),b=a.data.get(n.dataId).values;for(let C=0;C<l;C++){let N=C*p*c*m;for(let _=0;_<c;_++){let A=_*(p*m);for(let $=0;$<p;$++){let F=$*m;for(let P=0;P<m;P++){let V=[l,_,$,P],G=V[2],W=V[1],q=(G-f)*x-(W-d)*g,H=(G-f)*g+(W-d)*x;q=Math.round(q+f),H=Math.round(H+d);let j=s;if(typeof s!="number"&&(P===3?j=h:j=s[P]),q>=0&&q<p&&H>=0&&H<c){let Z=H*(p*m),et=q*m,rt=N+Z+et+P;j=b[rt]}let Y=N+A+F+P;u[Y]=j}}}}return{dataId:a.write(u,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};var x9=kt(Ts,r=>{let t=Math.floor(r);return r-t<.5?Math.floor(r):r-t>.5?Math.ceil(r):t%2===0?t:t+1}),gP={kernelName:Ts,backendName:"cpu",kernelFunc:x9};function y9(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o,updates:s}=t,{shape:i}=n,{sliceRank:a,numUpdates:u,sliceSize:l,strides:c,outputSize:p}=v.calculateShapes(s,o,i),m=!0,f=e.bufferSync(o),d=e.bufferSync(s),h=dl(f,d,i,p,l,u,a,c,0,m);return e.makeTensorInfo(i,h.dtype,h.values)}var xP={kernelName:La,backendName:"cpu",kernelFunc:y9};function b9(r,t){let e=0,n=r.length,o=0;for(;e<n;)o=Math.floor((e+n)/2),r[o]<t?e=o+1:n=o;return n}function w9(r,t){let e=0,n=r.length,o=0;for(;e<n;)o=Math.floor((e+n)/2),r[o]<=t?e=o+1:n=o;return n}function yP(r,t,e,n,o,s){let i=y.getArrayFromDType("int32",e*o);for(let a=0;a<e;++a){let u=r.slice(a*n,(a+1)*n),l=a*o;for(let c=0;c<o;++c)i[l+c]=s==="left"?b9(u,t[c+l]):w9(u,t[c+l])}return i}function C9(r){let{inputs:t,backend:e,attrs:n}=r,{sortedSequence:o,values:s}=t,{side:i}=n,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,l=yP(a,u,o.shape[0],o.shape[1],s.shape[1],i);return e.makeTensorInfo(s.shape,"int32",l)}var bP={kernelName:Pp,backendName:"cpu",kernelFunc:C9};function I9(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t;tt([n,o,s],"select");let i=n.shape.length,a=e.data.get(n.dataId).values,u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=sr(o.dtype,s.dtype),p=y.makeZerosTypedArray(y.sizeFromShape(o.shape),c),m=0,f=i===0||i>1||o.shape.length===1?1:y.sizeFromShape(o.shape.slice(1));for(let d=0;d<a.length;d++)for(let h=0;h<f;h++)a[d]===1?p[m++]=u[d]:p[m++]=l[d];return e.makeTensorInfo(o.shape,c,p)}var wP={kernelName:hi,backendName:"cpu",kernelFunc:I9};var S9=v.SELU_SCALEALPHA,v9=v.SELU_SCALE,N9=kt(Ma,r=>r>=0?v9*r:S9*(Math.exp(r)-1)),CP={kernelName:Ma,backendName:"cpu",kernelFunc:N9};var T9=kt(Ba,r=>r<0?-1:r>0?1:0),IP={kernelName:Ba,backendName:"cpu",kernelFunc:T9};var k9=kt(Es,r=>Math.sin(r)),SP={kernelName:Es,backendName:"cpu",kernelFunc:k9};var E9=kt(za,r=>Math.sinh(r)),vP={kernelName:za,backendName:"cpu",kernelFunc:E9};var _9=11920928955078125e-23,NP=Math.log(_9)+2,A9=kt(Va,r=>{let t=r>-NP,e=r<NP,n=Math.exp(r),o;return e?o=n:t?o=r:o=Math.log(1+n),o}),TP={kernelName:Va,backendName:"cpu",kernelFunc:A9};function $9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n;tt([o],"spaceToBatchND");let a=y.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let _=1+s.length;_<o.shape.length;++_)u.push([0,0]);let l=gw.kernelFunc({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),c=v.getReshaped(l.shape,s,a,!1),p=v.getPermuted(c.length,s.length,!1),m=v.getReshapedPermuted(l.shape,s,a,!1),h=Yt({inputs:{x:l},backend:e,attrs:{shape:c}}),b=Ve({inputs:{x:h},backend:e,attrs:{perm:p}}),N=Yt({inputs:{x:b},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(b),N}var kP={kernelName:xi,backendName:"cpu",kernelFunc:$9};function D9(r){let{inputs:t,backend:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-        ${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Values must be a vector, saw:
-        ${o.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let a=e.data.get(n.dataId).values,u=e.data.get(o.dataId).values,l=e.data.get(s.dataId).values,c=e.data.get(i.dataId).values[0],[p,m,f,d,h]=iw(a,n.shape,n.dtype,u,o.dtype,l,c);return[e.makeTensorInfo(m,n.dtype,p),e.makeTensorInfo([m[0]],o.dtype,f),e.makeTensorInfo([d.length],"bool",new Uint8Array(d.map(g=>Number(g)))),e.makeTensorInfo([h.length],n.dtype,new Int32Array(h))]}var EP={kernelName:Pl,backendName:"cpu",kernelFunc:D9};function R9(r){let{inputs:t,backend:e}=r,{inputIndices:n,inputShape:o,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(e.data.get(o.dataId).values),a=e.data.get(n.dataId).values,u=Array.from(e.data.get(s.dataId).values),[l,c,p]=aw(a,n.shape,n.dtype,i,u);return[e.makeTensorInfo(c,n.dtype,l),e.makeTensorInfo([p.length],s.dtype,new Int32Array(p))]}var _P={kernelName:Ga,backendName:"cpu",kernelFunc:R9};function F9(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-          ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-          ${s.shape}`);if(o.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ld(i,n.shape,n.dtype,a,u,!0);return e.makeTensorInfo(c,n.dtype,l)}var AP={kernelName:Ll,backendName:"cpu",kernelFunc:F9};function O9(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-         ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-         ${s.shape}`);if(o.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=e.data.get(n.dataId).values,a=e.data.get(o.dataId).values,u=e.data.get(s.dataId).values,[l,c]=ld(i,n.shape,n.dtype,a,u);return e.makeTensorInfo(c,n.dtype,l)}var $P={kernelName:Ml,backendName:"cpu",kernelFunc:O9};function P9(r){let{inputs:t,backend:e,attrs:n}=r,{sparseIndices:o,sparseValues:s,defaultValue:i}=t,{outputShape:a}=n,{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=v.calculateShapes(s,o,a),f=!1,d=e.bufferSync(o),h;switch(s.dtype){case"bool":{let g=e.bufferSync(s),x=Boolean(e.data.get(i.dataId).values[0]);h=dl(d,g,a,m,c,l,u,p,x,f);break}case"float32":{let g=e.bufferSync(s),x=e.data.get(i.dataId).values[0];h=dl(d,g,a,m,c,l,u,p,x,f);break}case"int32":{let g=e.bufferSync(s),x=e.data.get(i.dataId).values[0];h=dl(d,g,a,m,c,l,u,p,x,f);break}case"string":{let g=e.bufferSync(s),x=y.decodeString(e.data.get(i.dataId).values[0]);h=dl(d,g,a,m,c,l,u,p,x,f);break}default:throw new Error(`Unsupported type ${s.dtype}`)}return e.makeTensorInfo(a,h.dtype,h.values)}var DP={kernelName:Lp,backendName:"cpu",kernelFunc:P9};function L9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=n,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=new Array(o.shape.length).fill(0),c=o.shape.slice();return u.map(p=>{let m=[...c];m[a]=p;let f=Fo({inputs:{x:o},backend:e,attrs:{begin:l,size:m}});return l[a]+=p,f})}var RP={kernelName:yi,backendName:"cpu",kernelFunc:L9};var FP={kernelName:zl,backendName:"cpu",kernelFunc:({inputs:r,backend:t})=>{let{x:e}=r,n=t;tt(e,"square");let o=n.data.get(e.dataId).values,s=new Float32Array(o.length);for(let a=0;a<o.length;++a){let u=o[a];s[a]=u*u}return{dataId:n.write(s,e.shape,e.dtype),shape:e.shape,dtype:e.dtype}}};var M9=kt(po,(r,t)=>{let e=t;return isNaN(r)?NaN:r>0?1:e.alpha}),OP={kernelName:po,backendName:"cpu",kernelFunc:M9};function z9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n;tt(o,"stridedSlice");let{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=Yt({inputs:{x:o},backend:e,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let _=Le.computeOutShape(b,w,C),A=Fo({inputs:{x:o},backend:e,attrs:{begin:b,size:_}});N=Yt({inputs:{x:A},backend:e,attrs:{shape:d}}),e.disposeIntermediateTensorInfo(A)}else{let _=e.bufferSync(o),A=lw(f,_,C,b);N=e.makeTensorInfo(d,A.dtype,A.values)}return N}var PP={kernelName:Wa,backendName:"cpu",kernelFunc:z9};function B9(r){let{inputs:t,backend:e,attrs:n}=r,{separator:o,nGramWidths:s,leftPad:i,rightPad:a,padWidth:u,preserveShortSequences:l}=n,{data:c,dataSplits:p}=t,m=e.data.get(c.dataId).values,f=e.data.get(p.dataId).values,[d,h]=Dc(m,f,o,s,i,a,u,l);return[e.makeTensorInfo([d.length],"string",d),e.makeTensorInfo(p.shape,"int32",h)]}var LP={kernelName:Bl,backendName:"cpu",kernelFunc:B9};function V9(r){let{inputs:t,backend:e,attrs:n}=r,{skipEmpty:o}=n,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let a=e.data.get(s.dataId).values,u=e.data.get(i.dataId).values[0],[l,c,p]=Rc(a,u,o),m=c.length;return[e.makeTensorInfo([m,2],"int32",l),e.makeTensorInfo([m],"string",c),e.makeTensorInfo([2],"int32",new Int32Array(p))]}var MP={kernelName:Vl,backendName:"cpu",kernelFunc:V9};function G9(r){let{inputs:t,backend:e,attrs:n}=r,{numBuckets:o}=n,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(o<=0)throw new Error("Number of buckets must be at least 1");let i=e.data.get(s.dataId).values,a=Fc(i,o);return e.makeTensorInfo(s.shape,"int32",a)}var zP={kernelName:Gl,backendName:"cpu",kernelFunc:G9};var W9=kt(Os,r=>Math.tan(r)),BP={kernelName:Os,backendName:"cpu",kernelFunc:W9};var U9=kt(Ps,r=>Math.tanh(r)),VP={kernelName:Ps,backendName:"cpu",kernelFunc:U9};function H9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reps:s}=n;tt(o,"tile");let i=uw(e.bufferSync(o),s);return e.makeTensorInfo(i.shape,i.dtype,i.values)}var GP={kernelName:Jn,backendName:"cpu",kernelFunc:H9};function q9(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{k:s,sorted:i}=n;tt(o,"topk");let a=e.data.get(o.dataId).values,[u,l]=cw(a,o.shape,o.dtype,s,i);return[e.makeTensorInfo(u.shape,u.dtype,u.values),e.makeTensorInfo(l.shape,l.dtype,l.values)]}var WP={kernelName:Ua,backendName:"cpu",kernelFunc:q9};function K9(r){let{inputs:t,attrs:e,backend:n}=r,{image:o,transforms:s}=t,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=e,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=y.computeStrides(o.shape),b=x[0],w=x[1],C=x[2],N=y.computeStrides(g),_=N[0],A=N[1],$=N[2],F=y.getTypedArrayFromDType(o.dtype,y.sizeFromShape(g));F.fill(u);let P=n.data.get(o.dataId).values,V=n.data.get(s.dataId).values;for(let W=0;W<c;++W){let q=s.shape[0]===1?V:V.subarray(W*8,W*8+8);for(let H=0;H<d;++H)for(let j=0;j<h;++j)for(let Y=0;Y<f;++Y){let Z,et=q[6]*j+q[7]*H+1;if(et===0)continue;let rt=(q[0]*j+q[1]*H+q[2])/et,ot=(q[3]*j+q[4]*H+q[5])/et,at=UP(rt,m,a),nt=UP(ot,p,a);switch(i){case"nearest":Z=J9(P,p,m,b,w,C,W,nt,at,Y,u);break;case"bilinear":Z=Q9(P,p,m,b,w,C,W,nt,at,Y,u);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}let it=W*_+H*A+j*$+Y;F[it]=Z}return n.makeTensorInfo(g,o.dtype,F)}return{dataId:n.write(F,g,o.dtype),shape:o.shape,dtype:o.dtype}}var HP={kernelName:Ha,backendName:"cpu",kernelFunc:K9};function UP(r,t,e){switch(e){case"reflect":return j9(r,t);case"wrap":return X9(r,t);case"nearest":return Z9(r,t);case"constant":default:return Y9(r,t)}}function j9(r,t){let e=r;if(e<0)if(t<=1)e=0;else{let n=2*t;e<n&&(e=n*Math.trunc(-e/n)+e),e=e<-t?e+n:-e-1}else if(e>t-1)if(t<=1)e=0;else{let n=2*t;e-=n*Math.trunc(e/n),e>=t&&(e=n-e-1)}return y.clamp(0,e,t-1)}function X9(r,t){let e=r;if(e<0)if(t<=1)e=0;else{let n=t-1;e+=t*(Math.trunc(-e/n)+1)}else if(e>t-1)if(t<=1)e=0;else{let n=t-1;e-=t*Math.trunc(e/n)}return y.clamp(0,e,t-1)}function Y9(r,t){return r}function Z9(r,t){return y.clamp(0,r,t-1)}function Kh(r,t,e,n,o,s,i,a,u,l,c){let p=i*n+a*o+u*s+l;return 0<=a&&a<t&&0<=u&&u<e?r[p]:c}function J9(r,t,e,n,o,s,i,a,u,l,c){let p=Math.round(a),m=Math.round(u);return Kh(r,t,e,n,o,s,i,p,m,l,c)}function Q9(r,t,e,n,o,s,i,a,u,l,c){let p=Math.floor(a),m=Math.floor(u),f=p+1,d=m+1,h=(d-u)*Kh(r,t,e,n,o,s,i,p,m,l,c)+(u-m)*Kh(r,t,e,n,o,s,i,p,d,l,c),g=(d-u)*Kh(r,t,e,n,o,s,i,f,m,l,c)+(u-m)*Kh(r,t,e,n,o,s,i,f,d,l,c);return(f-a)*h+(a-p)*g}function ttt(r){let{inputs:t,attrs:e,backend:n}=r,{axis:o}=e,{x:s}=t;tt(s,"unique");let i=n.data.get(s.dataId).values,{outputValues:a,outputShape:u,indices:l}=pw(i,o,s.shape,s.dtype);return[n.makeTensorInfo(u,s.dtype,a),n.makeTensorInfo([l.length],"int32",l)]}var qP={kernelName:Mp,backendName:"cpu",kernelFunc:ttt};function ett(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o.shape.length,a=o.shape[s],u=new Array(i-1),l=0;for(let f=0;f<i;f++)f!==s&&(u[l++]=o.shape[f]);let c=new Array(i).fill(0),p=o.shape.slice();p[s]=1;let m=new Array(a);for(let f=0;f<m.length;f++){c[s]=f;let d=Fo({inputs:{x:o},backend:e,attrs:{begin:c,size:p}});m[f]=Yt({inputs:{x:d},backend:e,attrs:{shape:u}}),e.disposeIntermediateTensorInfo(d)}return m}var KP={kernelName:bi,backendName:"cpu",kernelFunc:ett};function rtt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,segmentIds:s}=t,{numSegments:i}=n;tt(o,"unsortedSegmentSum");let a=o.shape.length,u=s.shape.length,l=[],c=[],p=a-u,m=s;for(let d=0;d<p;++d){let h=cd({inputs:{input:m},backend:e,attrs:{dim:d+1}});m=h,c.push(h)}for(let d=0;d<i;++d){let h=y.createScalarValue(d,"int32"),g=e.makeTensorInfo([],"int32",h),x=GN({inputs:{a:g,b:m},backend:e}),b=$o({inputs:{x},backend:e,attrs:{dtype:"float32"}}),w=_c({inputs:{a:b,b:o},backend:e}),C=hl({inputs:{x:w},backend:e,attrs:{axis:0,keepDims:!1}});l.push(C),c.push(g),c.push(x),c.push(b),c.push(w),c.push(C)}let f=bT({inputs:l,backend:e,attrs:{axis:0}});return c.forEach(d=>e.disposeIntermediateTensorInfo(d)),f}var jP={kernelName:Wl,backendName:"cpu",kernelFunc:rtt};var ntt=[wF,$R,CF,IF,PR,SF,vF,NF,TF,kF,EF,_F,AF,$F,DF,FF,OF,PF,LF,bF,MF,zF,BF,VF,OR,LR,GF,DR,WF,HF,qF,KF,jF,XF,YF,ZF,JF,QF,tO,eO,rO,nO,oO,sO,iO,aO,lO,uO,cO,pO,fO,fF,dO,MR,hO,zR,gO,BR,xO,yO,bO,VR,wO,CO,IO,SO,vO,GR,WR,RR,NO,UF,TO,kO,EO,dF,UR,HR,_O,qR,AO,$O,DO,RO,FO,OO,PO,KR,LO,MO,zO,BO,GO,WO,UO,jR,HO,qO,XO,XR,YR,YO,ZO,JO,ZR,QO,rP,nP,gw,oP,hF,QR,sP,iP,aP,lP,FR,Uh,uP,gF,xF,yF,cP,pP,mP,fP,dP,hP,gP,oF,xP,bP,wP,CP,iF,IP,SP,vP,aF,KO,TP,kP,EP,_P,AP,$P,DP,RP,uF,FP,cF,OP,PP,LP,MP,zP,pF,mO,BP,VP,GP,WP,HP,JR,qP,KP,jP,tP];for(let r of ntt)Lu(r);var dd={};Wt(dd,{assertNotComplex:()=>Qs,bindCanvasToFramebuffer:()=>mtt,bindColorTextureToFramebuffer:()=>Zh,bindTextureToProgramUniformSampler:()=>OT,bindTextureUnit:()=>JP,bindVertexBufferToProgramAttribute:()=>Iw,callAndCheck:()=>yt,canBeRepresented:()=>ST,createFragmentShader:()=>NT,createFramebuffer:()=>DT,createProgram:()=>TT,createStaticIndexBuffer:()=>_T,createStaticVertexBuffer:()=>ET,createTexture:()=>AT,createVertexShader:()=>vT,getBatchDim:()=>xl,getExtensionOrThrow:()=>pd,getFramebufferErrorMessage:()=>QP,getMaxTexturesInShader:()=>MT,getNumChannels:()=>ctt,getProgramUniformLocation:()=>FT,getProgramUniformLocationOrThrow:()=>RT,getRowsCols:()=>yl,getShapeAs3D:()=>fd,getTextureShapeFromLogicalShape:()=>PT,getWebGLDisjointQueryTimerVersion:()=>zT,getWebGLErrorMessage:()=>ZP,getWebGLMaxTextureSize:()=>LT,hasExtension:()=>Wn,isCapableOfRenderingToFloatTexture:()=>BT,isDownloadFloatTextureEnabled:()=>VT,isReshapeFree:()=>Eu,isWebGLFenceEnabled:()=>GT,isWebGLVersionEnabled:()=>vw,linkProgram:()=>kT,logShaderSourceAndInfoLog:()=>Cw,resetMaxTextureSize:()=>ftt,resetMaxTexturesInShader:()=>dtt,unbindColorTextureFromFramebuffer:()=>Sw,unbindTextureUnit:()=>ptt,validateFramebuffer:()=>md,validateProgram:()=>Yh,validateTextureSize:()=>$T});var Pc={},xw={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function wT(r,t){Pc[r]=t}function Gn(r,t){if(!(r in Pc)||t!=null){let n=stt(r,t);if(n!==null)Pc[r]=n;else return console.log("Could not get context for WebGL version",r),null}let e=Pc[r];return e==null||e.isContextLost()?(delete Pc[r],Gn(r)):(e.disable(e.DEPTH_TEST),e.disable(e.STENCIL_TEST),e.disable(e.BLEND),e.disable(e.DITHER),e.disable(e.POLYGON_OFFSET_FILL),e.disable(e.SAMPLE_COVERAGE),e.enable(e.SCISSOR_TEST),e.enable(e.CULL_FACE),e.cullFace(e.BACK),Pc[r])}function ott(r){if(typeof OffscreenCanvas!="undefined"&&r===2)return new OffscreenCanvas(300,150);if(typeof document!="undefined")return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}function stt(r,t){if(r!==1&&r!==2)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let e=t==null?ott(r):t;return e.addEventListener("webglcontextlost",n=>{n.preventDefault(),delete Pc[r]},!1),z().getBool("SOFTWARE_WEBGL_ENABLED")&&(xw.failIfMajorPerformanceCaveat=!1),r===1?e.getContext("webgl",xw)||e.getContext("experimental-webgl",xw):e.getContext("webgl2",xw)}var ku;(function(r){r[r.DENSE=0]="DENSE",r[r.SHARED_BATCH=1]="SHARED_BATCH"})(ku||(ku={}));var jr;(function(r){r[r.RENDER=0]="RENDER",r[r.UPLOAD=1]="UPLOAD",r[r.PIXELS=2]="PIXELS",r[r.DOWNLOAD=3]="DOWNLOAD"})(jr||(jr={}));var Pr;(function(r){r[r.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",r[r.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",r[r.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",r[r.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",r[r.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16"})(Pr||(Pr={}));function Lc(r,t){return[t,r]}function XP(r,t){return r*t}function jh(r){let t=y.sizeFromShape(r),e=Math.ceil(t/4);return y.sizeToSquarishShape(e)}function Xi(r,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(r/2))]}function YP(r,t){let[e,n]=Xi(r,t);return e*n*4}function Xh(r,t){let e=r,n,o,s,i,a,u,l,c,p,m;return z().getNumber("WEBGL_VERSION")===2?(n=e.R32F,o=e.R16F,s=e.RGBA16F,i=e.RGBA32F,a=e.RED,l=4,c=1,p=e.HALF_FLOAT,m=e.FLOAT,u=e.RGBA8):(n=r.RGBA,o=r.RGBA,s=r.RGBA,i=e.RGBA,a=r.RGBA,l=4,c=4,p=t!=null?t.HALF_FLOAT_OES:null,m=r.FLOAT,u=r.RGBA),{internalFormatFloat:n,internalFormatHalfFloat:o,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:a,downloadTextureFormat:u,downloadUnpackNumChannels:l,defaultNumChannels:c,textureTypeHalfFloat:p,textureTypeFloat:m}}function yt(r,t){let e=t();return z().getBool("DEBUG")&&itt(r),e}function itt(r){let t=r.getError();if(t!==r.NO_ERROR)throw new Error("WebGL Error: "+ZP(r,t))}var att=596e-10,ltt=65504;function ST(r){return!!(z().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||r===0||att<Math.abs(r)&&Math.abs(r)<ltt)}function ZP(r,t){switch(t){case r.NO_ERROR:return"NO_ERROR";case r.INVALID_ENUM:return"INVALID_ENUM";case r.INVALID_VALUE:return"INVALID_VALUE";case r.INVALID_OPERATION:return"INVALID_OPERATION";case r.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case r.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case r.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function pd(r,t){return gl(r,()=>r.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function vT(r,t){let e=gl(r,()=>r.createShader(r.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(yt(r,()=>r.shaderSource(e,t)),yt(r,()=>r.compileShader(e)),r.getShaderParameter(e,r.COMPILE_STATUS)===!1)throw console.log(r.getShaderInfoLog(e)),new Error("Failed to compile vertex shader.");return e}function NT(r,t){let e=gl(r,()=>r.createShader(r.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(yt(r,()=>r.shaderSource(e,t)),yt(r,()=>r.compileShader(e)),z().get("ENGINE_COMPILE_ONLY"))return e;if(r.getShaderParameter(e,r.COMPILE_STATUS)===!1)throw Cw(t,r.getShaderInfoLog(e)),new Error("Failed to compile fragment shader.");return e}var utt=/ERROR: [0-9]+:([0-9]+):/g;function Cw(r,t){let e=utt.exec(t);if(e==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(r);return}let n=+e[1],o=r.split(`
-`),s=o.length.toString().length+2,i=o.map((p,m)=>y.rightPad((m+1).toString(),s)+p),a=0;for(let p=0;p<i.length;p++)a=Math.max(i[p].length,a);let u=i.slice(0,n-1),l=i.slice(n-1,n),c=i.slice(n);console.log(u.join(`
-`)),console.log(t.split(`
-`)[0]),console.log(`%c ${y.rightPad(l[0],a)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(c.join(`
-`))}function TT(r){return gl(r,()=>r.createProgram(),"Unable to create WebGLProgram.")}function kT(r,t){if(yt(r,()=>r.linkProgram(t)),!z().get("ENGINE_COMPILE_ONLY")&&r.getProgramParameter(t,r.LINK_STATUS)===!1)throw console.log(r.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function Yh(r,t){if(yt(r,()=>r.validateProgram(t)),r.getProgramParameter(t,r.VALIDATE_STATUS)===!1)throw console.log(r.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function ET(r,t){let e=gl(r,()=>r.createBuffer(),"Unable to create WebGLBuffer");return yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,e)),yt(r,()=>r.bufferData(r.ARRAY_BUFFER,t,r.STATIC_DRAW)),e}function _T(r,t){let e=gl(r,()=>r.createBuffer(),"Unable to create WebGLBuffer");return yt(r,()=>r.bindBuffer(r.ELEMENT_ARRAY_BUFFER,e)),yt(r,()=>r.bufferData(r.ELEMENT_ARRAY_BUFFER,t,r.STATIC_DRAW)),e}function ctt(){return z().getNumber("WEBGL_VERSION")===2?1:4}function AT(r){return gl(r,()=>r.createTexture(),"Unable to create WebGLTexture.")}function $T(r,t){let e=z().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(r<=0||t<=0){let n=`[${r}x${t}]`;throw new Error("Requested texture size "+n+" is invalid.")}if(r>e||t>e){let n=`[${r}x${t}]`,o=`[${e}x${e}]`;throw new Error("Requested texture size "+n+" greater than WebGL maximum on this browser / GPU "+o+".")}}function DT(r){return gl(r,()=>r.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function Iw(r,t,e,n,o,s,i){let a=r.getAttribLocation(t,e);return a===-1?!1:(yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,n)),yt(r,()=>r.vertexAttribPointer(a,o,r.FLOAT,!1,s,i)),yt(r,()=>r.enableVertexAttribArray(a)),!0)}function JP(r,t,e){tL(r,e),yt(r,()=>r.activeTexture(r.TEXTURE0+e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,t))}function ptt(r,t){tL(r,t),yt(r,()=>r.activeTexture(r.TEXTURE0+t)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function RT(r,t,e){return gl(r,()=>r.getUniformLocation(t,e),'uniform "'+e+'" not present in program.')}function FT(r,t,e){return r.getUniformLocation(t,e)}function OT(r,t,e,n){yt(r,()=>JP(r,t,n)),yt(r,()=>r.uniform1i(e,n))}function mtt(r){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,null)),yt(r,()=>r.viewport(0,0,r.canvas.width,r.canvas.height)),yt(r,()=>r.scissor(0,0,r.canvas.width,r.canvas.height))}function Zh(r,t,e){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,e)),yt(r,()=>r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,t,0))}function Sw(r,t){yt(r,()=>r.bindFramebuffer(r.FRAMEBUFFER,t)),yt(r,()=>r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,null,0))}function md(r){let t=r.checkFramebufferStatus(r.FRAMEBUFFER);if(t!==r.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+QP(r,t))}function QP(r,t){switch(t){case r.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case r.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case r.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case r.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function gl(r,t,e){let n=yt(r,()=>t());if(n==null)throw new Error(e);return n}function tL(r,t){let e=r.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,n=t+r.TEXTURE0;if(n<r.TEXTURE0||n>e){let o=`[gl.TEXTURE0, gl.TEXTURE${e}]`;throw new Error(`textureUnit must be in ${o}.`)}}function xl(r,t=2){return y.sizeFromShape(r.slice(0,r.length-t))}function yl(r){if(r.length===0)throw Error("Cannot get rows and columns of an empty shape array.");return[r.length>1?r[r.length-2]:1,r[r.length-1]]}function fd(r){let t=[1,1,1];return r.length===0||r.length===1&&r[0]===1||(t=[xl(r),...yl(r)]),t}function PT(r,t=!1){let e=z().getNumber("WEBGL_MAX_TEXTURE_SIZE"),n=z().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");n===1/0&&z().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")&&(n=e/2),t&&(e=e*2,n=n*2,r=r.map((a,u)=>u>=r.length-2?y.nearestLargerEven(r[u]):r[u]),r.length===1&&(r=[2,r[0]])),r.length!==2&&(r=y.squeezeShape(r).newShape);let o=y.sizeFromShape(r),s=null;r.length<=1&&o<=e?s=[1,o]:r.length===2&&r[0]<=e&&r[1]<=e?s=r:r.length===3&&r[0]*r[1]<=e&&r[2]<=e?s=[r[0]*r[1],r[2]]:r.length===3&&r[0]<=e&&r[1]*r[2]<=e?s=[r[0],r[1]*r[2]]:r.length===4&&r[0]*r[1]*r[2]<=e&&r[3]<=e?s=[r[0]*r[1]*r[2],r[3]]:r.length===4&&r[0]<=e&&r[1]*r[2]*r[3]<=e&&(s=[r[0],r[1]*r[2]*r[3]]);let i=s!=null&&Math.max(...s)>n&&Math.min(...s)<=(t?2:1)&&Math.min(...s)>0;if(s==null||i)if(t){let a=xl(r),u=2,l=2;r.length&&([u,l]=yl(r)),o=a*(u/2)*(l/2),s=y.sizeToSquarishShape(o).map(c=>c*2)}else s=y.sizeToSquarishShape(o);return s}function yw(r){return r%2===0}function Eu(r,t){if(r=r.slice(-2),t=t.slice(-2),y.arraysEqual(r,t)||!r.length||!t.length||r[0]===0||r[1]===0||t[0]===0||t[1]===0)return!0;if(r.length!==t.length){let e=r.slice(-1)[0],n=t.slice(-1)[0];if(e===n||yw(e)&&yw(n)&&(r[0]===1||t[0]===1))return!0}return r[1]===t[1]&&yw(r[0])&&yw(t[0])}var bw,ww;function LT(r){if(bw==null){let t=Gn(r);bw=t.getParameter(t.MAX_TEXTURE_SIZE)}return bw}function ftt(){bw=null}function dtt(){ww=null}function MT(r){if(ww==null){let t=Gn(r);ww=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,ww)}function zT(r){if(r===0)return 0;let t,e=Gn(r);return Wn(e,"EXT_disjoint_timer_query_webgl2")&&r===2?t=2:Wn(e,"EXT_disjoint_timer_query")?t=1:t=0,t}function Wn(r,t){return r.getExtension(t)!=null}function vw(r){try{if(Gn(r)!=null)return!0}catch(t){return console.log("Error when getting WebGL context: ",t),!1}return!1}function BT(r){if(r===0)return!1;let t=Gn(r);if(r===1){if(!Wn(t,"OES_texture_float"))return!1}else if(!Wn(t,"EXT_color_buffer_float"))return!1;return IT(t)}function VT(r){if(r===0)return!1;let t=Gn(r);if(r===1){if(!Wn(t,"OES_texture_float")||!Wn(t,"WEBGL_color_buffer_float"))return!1}else{if(Wn(t,"EXT_color_buffer_float"))return IT(t);let n="EXT_color_buffer_half_float";if(Wn(t,n)){let o=t.getExtension(n);return htt(t,o)}return!1}return IT(t)}function IT(r){let t=Xh(r),e=r.createTexture();r.bindTexture(r.TEXTURE_2D,e);let n=1,o=1;r.texImage2D(r.TEXTURE_2D,0,t.internalFormatFloat,n,o,0,t.textureFormatFloat,t.textureTypeFloat,null);let s=r.createFramebuffer();r.bindFramebuffer(r.FRAMEBUFFER,s),r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,e,0);let i=r.checkFramebufferStatus(r.FRAMEBUFFER)===r.FRAMEBUFFER_COMPLETE;return r.bindTexture(r.TEXTURE_2D,null),r.bindFramebuffer(r.FRAMEBUFFER,null),r.deleteTexture(e),r.deleteFramebuffer(s),i}function htt(r,t){let e=Xh(r,t),n=r.createTexture();r.bindTexture(r.TEXTURE_2D,n);let o=1,s=1;r.texImage2D(r.TEXTURE_2D,0,e.internalFormatHalfFloat,o,s,0,e.textureFormatFloat,e.textureTypeHalfFloat,null);let i=r.createFramebuffer();r.bindFramebuffer(r.FRAMEBUFFER,i),r.framebufferTexture2D(r.FRAMEBUFFER,r.COLOR_ATTACHMENT0,r.TEXTURE_2D,n,0);let a=r.checkFramebufferStatus(r.FRAMEBUFFER)===r.FRAMEBUFFER_COMPLETE;return r.bindTexture(r.TEXTURE_2D,null),r.bindFramebuffer(r.FRAMEBUFFER,null),r.deleteTexture(n),r.deleteFramebuffer(i),a}function GT(r){return r!==2?!1:Gn(r).fenceSync!=null}function Qs(r,t){Array.isArray(r)||(r=[r]),r.forEach(e=>{e!=null&&y.assert(e.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var Tt=z();Tt.registerFlag("HAS_WEBGL",()=>Tt.getNumber("WEBGL_VERSION")>0);Tt.registerFlag("WEBGL_VERSION",()=>vw(2)?2:vw(1)?1:0);Tt.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1);Tt.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>Tt.get("WEBGL_VERSION")===2);Tt.registerFlag("WEBGL_CPU_FORWARD",()=>!0);Tt.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1);Tt.registerFlag("WEBGL_PACK",()=>Tt.getBool("HAS_WEBGL"));Tt.registerFlag("WEBGL_PACK_NORMALIZATION",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_CLIP",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_PACK_REDUCE",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_LAZILY_UNPACK",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_CONV_IM2COL",()=>Tt.getBool("WEBGL_PACK"));Tt.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>LT(Tt.getNumber("WEBGL_VERSION")));Tt.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>MT(Tt.getNumber("WEBGL_VERSION")));Tt.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let r=Tt.getNumber("WEBGL_VERSION");return r===0?0:zT(r)});Tt.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>Tt.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!Kl.isMobile());Tt.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>BT(Tt.getNumber("WEBGL_VERSION")));Tt.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>Tt.getBool("WEBGL_FORCE_F16_TEXTURES")?!1:Tt.getBool("WEBGL_RENDER_FLOAT32_CAPABLE"));Tt.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>VT(Tt.getNumber("WEBGL_VERSION")));Tt.registerFlag("WEBGL_FENCE_API_ENABLED",()=>GT(Tt.getNumber("WEBGL_VERSION")));Tt.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>Tt.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0);Tt.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,r=>{if(r<0&&r!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${r}.`)});Tt.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>Kl.isMobile()?1:-1,r=>{if(r<0&&r!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${r}.`)});Tt.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128);Tt.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1);Tt.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5);Tt.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);Tt.registerFlag("WEBGL_EXP_CONV",()=>!1);Tt.registerFlag("SOFTWARE_WEBGL_ENABLED",()=>Tt.getBool("IS_TEST"));Tt.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE",()=>1/0);Tt.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE",()=>!1);Tt.registerFlag("WEBGL2_ISNAN_CUSTOM",()=>!1);function Ge(){let r,t,e,n,o,s,i,a,u,l;return z().getNumber("WEBGL_VERSION")===2?(r="#version 300 es",t="in",e="out",n="in",o="texture",s="outputColor",i="out vec4 outputColor;",a=z().getBool("WEBGL2_ISNAN_CUSTOM")?`
+        ${totalLength}, and tensor's shape is: ${tensor2.shape}`);
+  }
+  const shapeWithoutFirstDim = tensor2.shape.slice(1);
+  const outputElementShape = mergeElementShape(shapeWithoutFirstDim, elementShape);
+  const elementPerRow = totalLength === 0 ? 0 : tensor2.size / totalLength;
+  const tensors = tidy(() => {
+    const tensors2 = [];
+    tensor2 = reshape(tensor2, [1, totalLength, elementPerRow]);
+    for (let i = 0; i < length.length; ++i) {
+      const previousLength = i === 0 ? 0 : cumulativeLengths[i - 1];
+      const indices = [0, previousLength, 0];
+      const sizes = [1, length[i], elementPerRow];
+      tensors2[i] = reshape(slice(tensor2, indices, sizes), outputElementShape);
+    }
+    tensor2.dispose();
+    return tensors2;
+  });
+  const list = new TensorList([], elementShape, tensor2.dtype, length.length);
+  for (let i = 0; i < tensors.length; i++) {
+    list.setItem(i, tensors[i]);
+  }
+  return list;
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/control_executor.js
+var executeOp3 = async (node, tensorMap, context) => {
+  switch (node.op) {
+    case "If":
+    case "StatelessIf": {
+      const thenFunc = getParamValue("thenBranch", node, tensorMap, context);
+      const elseFunc = getParamValue("elseBranch", node, tensorMap, context);
+      const cond = getParamValue("cond", node, tensorMap, context);
+      const args = getParamValue("args", node, tensorMap, context);
+      const condValue = await cond.data();
+      if (condValue[0]) {
+        return context.functionMap[thenFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      } else {
+        return context.functionMap[elseFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      }
+    }
+    case "While":
+    case "StatelessWhile": {
+      const bodyFunc = getParamValue("body", node, tensorMap, context);
+      const condFunc = getParamValue("cond", node, tensorMap, context);
+      const args = getParamValue("args", node, tensorMap, context);
+      const condResult = await context.functionMap[condFunc].executeFunctionAsync(args, context.tensorArrayMap, context.tensorListMap);
+      const argIds = args.map((tensor2) => tensor2.id);
+      let condValue = await condResult[0].data();
+      condResult.forEach((tensor2) => {
+        if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1) {
+          tensor2.dispose();
+        }
+      });
+      let result = args;
+      while (condValue[0]) {
+        const origResult = result;
+        result = await context.functionMap[bodyFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);
+        const resultIds = result.map((tensor2) => tensor2.id);
+        origResult.forEach((tensor2) => {
+          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {
+            tensor2.dispose();
+          }
+        });
+        const condResult2 = await context.functionMap[condFunc].executeFunctionAsync(result, context.tensorArrayMap, context.tensorListMap);
+        condValue = await condResult2[0].data();
+        condResult2.forEach((tensor2) => {
+          if (!tensor2.kept && argIds.indexOf(tensor2.id) === -1 && resultIds.indexOf(tensor2.id) === -1) {
+            tensor2.dispose();
+          }
+        });
+      }
+      return result;
+    }
+    case "LoopCond": {
+      const pred = getParamValue("pred", node, tensorMap, context);
+      return [cloneTensor(pred)];
+    }
+    case "Switch": {
+      const pred = getParamValue("pred", node, tensorMap, context);
+      let data = getParamValue("data", node, tensorMap, context);
+      if (!data.kept) {
+        data = cloneTensor(data);
+      }
+      return (await pred.data())[0] ? [void 0, data] : [data, void 0];
+    }
+    case "Merge": {
+      const inputName = node.inputNames.find((name) => getTensor(name, tensorMap, context) !== void 0);
+      if (inputName) {
+        const data = getTensor(inputName, tensorMap, context);
+        return [cloneTensor(data)];
+      }
+      return void 0;
+    }
+    case "Enter": {
+      const frameId = getParamValue("frameName", node, tensorMap, context);
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.enterFrame(frameId);
+      return [cloneTensor(data)];
+    }
+    case "Exit": {
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.exitFrame();
+      return [cloneTensor(data)];
+    }
+    case "NextIteration": {
+      const data = getParamValue("tensor", node, tensorMap, context);
+      context.nextIteration();
+      return [cloneTensor(data)];
+    }
+    case "TensorArrayV3": {
+      const size = getParamValue("size", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const dynamicSize = getParamValue("dynamicSize", node, tensorMap, context);
+      const clearAfterRead = getParamValue("clearAfterRead", node, tensorMap, context);
+      const identicalElementShapes = getParamValue("identicalElementShapes", node, tensorMap, context);
+      const name = getParamValue("name", node, tensorMap, context);
+      const tensorArray = new TensorArray(name, dtype, size, elementShape, identicalElementShapes, dynamicSize, clearAfterRead);
+      context.addTensorArray(tensorArray);
+      return [tensorArray.idTensor, scalar(1)];
+    }
+    case "TensorArrayWriteV3": {
+      const id = getParamValue("tensorArrayId", node, tensorMap, context);
+      const index = getParamValue("index", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const writeTensorArray = context.getTensorArray(id.id);
+      writeTensorArray.write(index, writeTensor);
+      return [writeTensorArray.idTensor];
+    }
+    case "TensorArrayReadV3": {
+      const readId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const readIndex = getParamValue("index", node, tensorMap, context);
+      const readTensorArray = context.getTensorArray(readId.id);
+      return [readTensorArray.read(readIndex)];
+    }
+    case "TensorArrayGatherV3": {
+      const gatherId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const gatherIndices = getParamValue("indices", node, tensorMap, context);
+      const gatherDtype = getParamValue("dtype", node, tensorMap, context);
+      const gatherTensorArray = context.getTensorArray(gatherId.id);
+      return [gatherTensorArray.gather(gatherIndices, gatherDtype)];
+    }
+    case "TensorArrayScatterV3": {
+      const scatterId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const scatterIndices = getParamValue("indices", node, tensorMap, context);
+      const scatterTensor = getParamValue("tensor", node, tensorMap, context);
+      const scatterTensorArray = context.getTensorArray(scatterId.id);
+      scatterTensorArray.scatter(scatterIndices, scatterTensor);
+      return [scatterTensorArray.idTensor];
+    }
+    case "TensorArrayConcatV3": {
+      const concatId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const concatTensorArray = context.getTensorArray(concatId.id);
+      const concatDtype = getParamValue("dtype", node, tensorMap, context);
+      return [concatTensorArray.concat(concatDtype)];
+    }
+    case "TensorArraySplitV3": {
+      const splitId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const splitTensor = getParamValue("tensor", node, tensorMap, context);
+      const lengths = getParamValue("lengths", node, tensorMap, context);
+      const splitTensorArray = context.getTensorArray(splitId.id);
+      splitTensorArray.split(lengths, splitTensor);
+      return [splitTensorArray.idTensor];
+    }
+    case "TensorArraySizeV3": {
+      const sizeId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const sizeTensorArray = context.getTensorArray(sizeId.id);
+      return [scalar(sizeTensorArray.size(), "int32")];
+    }
+    case "TensorArrayCloseV3": {
+      const closeId = getParamValue("tensorArrayId", node, tensorMap, context);
+      const closeTensorArray = context.getTensorArray(closeId.id);
+      closeTensorArray.clearAndClose();
+      return [closeTensorArray.idTensor];
+    }
+    case "TensorListSetItem": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const index = getParamValue("index", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      tensorList.setItem(index, writeTensor);
+      return [tensorList.idTensor];
+    }
+    case "TensorListGetItem": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const readIndex = getParamValue("index", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDType = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.getItem(readIndex, elementShape, elementDType)];
+    }
+    case "TensorListScatterV2":
+    case "TensorListScatter": {
+      const scatterIndices = getParamValue("indices", node, tensorMap, context);
+      const scatterTensor = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const numElements = getParamValue("numElements", node, tensorMap, context);
+      const tensorList = scatter(scatterTensor, scatterIndices, elementShape, numElements);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListReserve":
+    case "EmptyTensorList": {
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      let numElementsParam;
+      if (node.op === "TensorListReserve") {
+        numElementsParam = "numElements";
+      } else {
+        numElementsParam = "maxNumElements";
+      }
+      const numElements = getParamValue(numElementsParam, node, tensorMap, context);
+      const maxNumElements = node.op === "TensorListReserve" ? -1 : numElements;
+      const tensorList = reserve(elementShape, elementDtype, numElements, maxNumElements);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListGather": {
+      const gatherId = getParamValue("tensorListId", node, tensorMap, context);
+      const gatherIndices = getParamValue("indices", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(gatherId.id);
+      return [tensorList.gather(gatherIndices, elementDtype, elementShape)];
+    }
+    case "TensorListStack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const numElements = getParamValue("numElements", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.stack(elementShape, elementDtype, numElements)];
+    }
+    case "TensorListFromTensor": {
+      const tensor2 = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDtype = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = fromTensor(tensor2, elementShape, elementDtype);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListConcat":
+    case "TensorListConcatV2": {
+      const concatId = getParamValue("tensorListId", node, tensorMap, context);
+      const tensorList = context.getTensorList(concatId.id);
+      const concatDtype = getParamValue("dtype", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      return [tensorList.concat(concatDtype, elementShape)];
+    }
+    case "TensorListPushBack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const writeTensor = getParamValue("tensor", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      tensorList.pushBack(writeTensor);
+      return [tensorList.idTensor];
+    }
+    case "TensorListPopBack": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const elementDType = getParamValue("elementDType", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [tensorList.popBack(elementShape, elementDType)];
+    }
+    case "TensorListSplit": {
+      const splitTensor = getParamValue("tensor", node, tensorMap, context);
+      const elementShape = getParamValue("elementShape", node, tensorMap, context);
+      const lengths = getParamValue("lengths", node, tensorMap, context);
+      const tensorList = split2(splitTensor, lengths, elementShape);
+      context.addTensorList(tensorList);
+      return [tensorList.idTensor];
+    }
+    case "TensorListLength": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const tensorList = context.getTensorList(idTensor.id);
+      return [scalar(tensorList.size(), "int32")];
+    }
+    case "TensorListResize": {
+      const idTensor = getParamValue("tensorListId", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const srcTensorList = context.getTensorList(idTensor.id);
+      const destTensorList = srcTensorList.resize(size);
+      context.addTensorList(destTensorList);
+      return [destTensorList.idTensor];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/convolution_executor.js
+function fusedConvAndDepthWiseParams(node, tensorMap, context) {
+  const [extraOp, activationFunc] = getParamValue("fusedOps", node, tensorMap, context);
+  const isBiasAdd = extraOp === "biasadd";
+  const noBiasAdd = !isBiasAdd;
+  const isPrelu = activationFunc === "prelu";
+  const isBatchNorm = extraOp === "fusedbatchnorm";
+  const numArgs = getParamValue("numArgs", node, tensorMap, context);
+  if (isBiasAdd) {
+    if (isPrelu && numArgs !== 2) {
+      throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+    }
+    if (!isPrelu && isBiasAdd && numArgs !== 1) {
+      throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.");
+    }
+  }
+  if (isBatchNorm) {
+    throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");
+  }
+  const stride = getParamValue("strides", node, tensorMap, context);
+  const pad3 = getPadding(node, tensorMap, context);
+  const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+  const dilations = getParamValue("dilations", node, tensorMap, context);
+  let [biasArg, preluArg] = getParamValue("args", node, tensorMap, context);
+  if (noBiasAdd) {
+    preluArg = biasArg;
+    biasArg = void 0;
+  }
+  const leakyreluAlpha = getParamValue("leakyreluAlpha", node, tensorMap, context);
+  return {
+    stride,
+    pad: pad3,
+    dataFormat,
+    dilations,
+    biasArg,
+    preluArg,
+    activationFunc,
+    leakyreluAlpha
+  };
+}
+var executeOp4 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Conv1D": {
+      const stride = getParamValue("stride", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilation = getParamValue("dilation", node, tensorMap, context);
+      return [ops.conv1d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), stride, pad3, dataFormat, dilation)];
+    }
+    case "Conv2D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      return [ops.conv2d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];
+    }
+    case "_FusedConv2D": {
+      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);
+      return [ops.fused.conv2d({
+        x: getParamValue("x", node, tensorMap, context),
+        filter: getParamValue("filter", node, tensorMap, context),
+        strides: [stride[1], stride[2]],
+        pad: pad3,
+        dataFormat,
+        dilations: [dilations[1], dilations[2]],
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    }
+    case "FusedDepthwiseConv2dNative": {
+      const { stride, pad: pad3, dataFormat, dilations, biasArg, preluArg, activationFunc, leakyreluAlpha } = fusedConvAndDepthWiseParams(node, tensorMap, context);
+      return [ops.fused.depthwiseConv2d({
+        x: getParamValue("x", node, tensorMap, context),
+        filter: getParamValue("filter", node, tensorMap, context),
+        strides: [stride[1], stride[2]],
+        pad: pad3,
+        dataFormat,
+        dilations: [dilations[1], dilations[2]],
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    }
+    case "Conv2DBackpropInput":
+    case "Conv2dTranspose": {
+      const shape = getParamValue("outputShape", node, tensorMap, context);
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      return [ops.conv2dTranspose(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), shape, [stride[1], stride[2]], pad3)];
+    }
+    case "DepthwiseConv2dNative":
+    case "DepthwiseConv2d": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getPadding(node, tensorMap, context);
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      return [ops.depthwiseConv2d(getParamValue("input", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2]], pad3, dataFormat, [dilations[1], dilations[2]])];
+    }
+    case "Conv3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      return [ops.conv3d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [stride[1], stride[2], stride[3]], pad3, dataFormat, [dilations[1], dilations[2], dilations[3]])];
+    }
+    case "AvgPool": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.avgPool(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];
+    }
+    case "MaxPool": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.maxPool(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3)];
+    }
+    case "MaxPoolWithArgmax": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      const includeBatchInIndex = getParamValue("includeBatchInIndex", node, tensorMap, context);
+      const { result, indexes } = ops.maxPoolWithArgmax(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2]], [stride[1], stride[2]], pad3, includeBatchInIndex);
+      return [result, indexes];
+    }
+    case "AvgPool3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.avgPool3d(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];
+    }
+    case "MaxPool3D": {
+      const stride = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const kernelSize = getParamValue("kernelSize", node, tensorMap, context);
+      return [ops.maxPool3d(getParamValue("x", node, tensorMap, context), [kernelSize[1], kernelSize[2], kernelSize[3]], [stride[1], stride[2], stride[3]], pad3)];
+    }
+    case "Dilation2D": {
+      const strides = getParamValue("strides", node, tensorMap, context);
+      const pad3 = getParamValue("pad", node, tensorMap, context);
+      const dilations = getParamValue("dilations", node, tensorMap, context);
+      const strideHeight = strides[1];
+      const strideWidth = strides[2];
+      const dilationHeight = dilations[1];
+      const dilationWidth = dilations[2];
+      return [ops.dilation2d(getParamValue("x", node, tensorMap, context), getParamValue("filter", node, tensorMap, context), [strideHeight, strideWidth], pad3, [dilationHeight, dilationWidth], "NHWC")];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/creation_executor.js
+var executeOp5 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Fill": {
+      const shape = getParamValue("shape", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      const value = getParamValue("value", node, tensorMap, context);
+      return [ops.fill(shape, value, dtype)];
+    }
+    case "LinSpace": {
+      const start = getParamValue("start", node, tensorMap, context);
+      const stop = getParamValue("stop", node, tensorMap, context);
+      const num = getParamValue("num", node, tensorMap, context);
+      return [ops.linspace(start, stop, num)];
+    }
+    case "Multinomial": {
+      const logits = getParamValue("logits", node, tensorMap, context);
+      const numSamples = getParamValue("numSamples", node, tensorMap, context);
+      const seed = getParamValue("seed", node, tensorMap, context);
+      return [ops.multinomial(logits, numSamples, seed)];
+    }
+    case "OneHot": {
+      const indices = getParamValue("indices", node, tensorMap, context);
+      const depth = getParamValue("depth", node, tensorMap, context);
+      const onValue = getParamValue("onValue", node, tensorMap, context);
+      const offValue = getParamValue("offValue", node, tensorMap, context);
+      const dtype = getParamValue("dtype", node, tensorMap, context);
+      return [ops.oneHot(indices, depth, onValue, offValue, dtype)];
+    }
+    case "Ones": {
+      return [ops.ones(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "OnesLike": {
+      return [ops.onesLike(getParamValue("x", node, tensorMap, context))];
+    }
+    case "RandomStandardNormal": {
+      return [ops.randomStandardNormal(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context), getParamValue("seed", node, tensorMap, context))];
+    }
+    case "RandomUniform": {
+      return [ops.randomUniform(
+        getParamValue("shape", node, tensorMap, context),
+        getParamValue("minval", node, tensorMap, context),
+        getParamValue("maxval", node, tensorMap, context),
+        getParamValue("dtype", node, tensorMap, context)
+      )];
+    }
+    case "Range": {
+      const start = getParamValue("start", node, tensorMap, context);
+      const stop = getParamValue("stop", node, tensorMap, context);
+      const step5 = getParamValue("step", node, tensorMap, context);
+      return [ops.range(start, stop, step5, getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "TruncatedNormal": {
+      const shape = getParamValue("shape", node, tensorMap, context);
+      const mean4 = getParamValue("mean", node, tensorMap, context);
+      const stdDev = getParamValue("stdDev", node, tensorMap, context);
+      const seed = getParamValue("seed", node, tensorMap, context);
+      return [ops.truncatedNormal(shape, mean4, stdDev, getParamValue("dtype", node, tensorMap, context), seed)];
+    }
+    case "Zeros": {
+      return [ops.zeros(getParamValue("shape", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "ZerosLike": {
+      return [ops.zerosLike(getParamValue("x", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/dynamic_executor.js
+function nmsParams(node, tensorMap, context) {
+  const boxes = getParamValue("boxes", node, tensorMap, context);
+  const scores = getParamValue("scores", node, tensorMap, context);
+  const maxOutputSize = getParamValue("maxOutputSize", node, tensorMap, context);
+  const iouThreshold = getParamValue("iouThreshold", node, tensorMap, context);
+  const scoreThreshold = getParamValue("scoreThreshold", node, tensorMap, context);
+  const softNmsSigma = getParamValue("softNmsSigma", node, tensorMap, context);
+  return {
+    boxes,
+    scores,
+    maxOutputSize,
+    iouThreshold,
+    scoreThreshold,
+    softNmsSigma
+  };
+}
+var executeOp6 = async (node, tensorMap, context, resourceManager, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "NonMaxSuppressionV5": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = nmsParams(node, tensorMap, context);
+      const result = await ops.image.nonMaxSuppressionWithScoreAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+      return [result.selectedIndices, result.selectedScores];
+    }
+    case "NonMaxSuppressionV4": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);
+      const padToMaxOutputSize = getParamValue("padToMaxOutputSize", node, tensorMap, context);
+      const result = await ops.image.nonMaxSuppressionPaddedAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+      return [result.selectedIndices, result.validOutputs];
+    }
+    case "NonMaxSuppressionV3":
+    case "NonMaxSuppressionV2": {
+      const { boxes, scores, maxOutputSize, iouThreshold, scoreThreshold } = nmsParams(node, tensorMap, context);
+      return [await ops.image.nonMaxSuppressionAsync(boxes, scores, maxOutputSize, iouThreshold, scoreThreshold)];
+    }
+    case "Where": {
+      const condition = ops.cast(getParamValue("condition", node, tensorMap, context), "bool");
+      const result = [await ops.whereAsync(condition)];
+      condition.dispose();
+      return result;
+    }
+    case "ListDiff": {
+      return ops.setdiff1dAsync(getParamValue("x", node, tensorMap, context), getParamValue("y", node, tensorMap, context));
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/evaluation_executor.js
+var executeOp7 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "LowerBound": {
+      const sortedSequence = getParamValue("sortedSequence", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      return [ops.lowerBound(sortedSequence, values)];
+    }
+    case "TopKV2": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const k = getParamValue("k", node, tensorMap, context);
+      const sorted = getParamValue("sorted", node, tensorMap, context);
+      const result = ops.topk(x, k, sorted);
+      return [result.values, result.indices];
+    }
+    case "UpperBound": {
+      const sortedSequence = getParamValue("sortedSequence", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      return [ops.upperBound(sortedSequence, values)];
+    }
+    case "Unique": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const result = ops.unique(x);
+      return [result.values, result.indices];
+    }
+    case "UniqueV2": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const result = ops.unique(x, axis);
+      return [result.values, result.indices];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/graph_executor.js
+var executeOp8 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Const": {
+      return tensorMap[node.name];
+    }
+    case "PlaceholderWithDefault":
+      const def = getParamValue("default", node, tensorMap, context);
+      return [getTensor(node.name, tensorMap, context) || def];
+    case "Placeholder":
+      return [getTensor(node.name, tensorMap, context)];
+    case "Identity":
+    case "StopGradient":
+    case "FakeQuantWithMinMaxVars": {
+      const data2 = getParamValue("x", node, tensorMap, context);
+      return [cloneTensor(data2)];
+    }
+    case "IdentityN":
+      return getParamValue("x", node, tensorMap, context).map((t) => cloneTensor(t));
+    case "Snapshot":
+      const snapshot = getParamValue("x", node, tensorMap, context);
+      return [cloneTensor(snapshot)];
+    case "Shape":
+      return [ops.tensor1d(getParamValue("x", node, tensorMap, context).shape, "int32")];
+    case "ShapeN":
+      return getParamValue("x", node, tensorMap, context).map((t) => ops.tensor1d(t.shape));
+    case "Size":
+      return [ops.scalar(getParamValue("x", node, tensorMap, context).size, "int32")];
+    case "Rank":
+      return [ops.scalar(getParamValue("x", node, tensorMap, context).rank, "int32")];
+    case "NoOp":
+      return [ops.scalar(1)];
+    case "Print":
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const data = getParamValue("data", node, tensorMap, context);
+      const message = getParamValue("message", node, tensorMap, context);
+      const summarize = getParamValue("summarize", node, tensorMap, context);
+      console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance.");
+      console.log(message);
+      for (let i = 0; i < data.length; i++) {
+        console.log(Array.prototype.slice.call(data[i].dataSync()).slice(0, summarize));
+      }
+      return [input2];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/hash_table.js
+var HashTable = class {
+  constructor(keyDType, valueDType) {
+    this.keyDType = keyDType;
+    this.valueDType = valueDType;
+    this.handle = scalar(0);
+    this.tensorMap = /* @__PURE__ */ new Map();
+    keep(this.handle);
+  }
+  get id() {
+    return this.handle.id;
+  }
+  clearAndClose() {
+    this.tensorMap.forEach((value) => value.dispose());
+    this.tensorMap.clear();
+    this.handle.dispose();
+  }
+  size() {
+    return this.tensorMap.size;
+  }
+  tensorSize() {
+    return scalar(this.size(), "int32");
+  }
+  async import(keys, values) {
+    this.checkKeyAndValueTensor(keys, values);
+    const $keys = await keys.data();
+    this.tensorMap.forEach((value) => value.dispose());
+    this.tensorMap.clear();
+    return tidy(() => {
+      const $values = unstack(values);
+      const keysLength = $keys.length;
+      const valuesLength = $values.length;
+      util_exports.assert(keysLength === valuesLength, () => `The number of elements doesn't match, keys has ${keysLength} elements, the values has ${valuesLength} elements.`);
+      for (let i = 0; i < keysLength; i++) {
+        const key = $keys[i];
+        const value = $values[i];
+        keep(value);
+        this.tensorMap.set(key, value);
+      }
+      return this.handle;
+    });
+  }
+  async find(keys, defaultValue) {
+    this.checkKeyAndValueTensor(keys, defaultValue);
+    const $keys = await keys.data();
+    return tidy(() => {
+      const result = [];
+      for (let i = 0; i < $keys.length; i++) {
+        const key = $keys[i];
+        const value = this.findWithDefault(key, defaultValue);
+        result.push(value);
+      }
+      return stack(result);
+    });
+  }
+  findWithDefault(key, defaultValue) {
+    const result = this.tensorMap.get(key);
+    return result != null ? result : defaultValue;
+  }
+  checkKeyAndValueTensor(key, value) {
+    if (key.dtype !== this.keyDType) {
+      throw new Error(`Expect key dtype ${this.keyDType}, but got ${key.dtype}`);
+    }
+    if (value.dtype !== this.valueDType) {
+      throw new Error(`Expect value dtype ${this.valueDType}, but got ${value.dtype}`);
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/hash_table_executor.js
+var executeOp9 = async (node, tensorMap, context, resourceManager) => {
+  switch (node.op) {
+    case "HashTable":
+    case "HashTableV2": {
+      const existingTableHandle = resourceManager.getHashTableHandleByName(node.name);
+      if (existingTableHandle != null) {
+        return [existingTableHandle];
+      } else {
+        const keyDType = getParamValue("keyDType", node, tensorMap, context);
+        const valueDType = getParamValue("valueDType", node, tensorMap, context);
+        const hashTable = new HashTable(keyDType, valueDType);
+        resourceManager.addHashTable(node.name, hashTable);
+        return [hashTable.handle];
+      }
+    }
+    case "LookupTableImport":
+    case "LookupTableImportV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const keys = getParamValue("keys", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [await hashTable.import(keys, values)];
+    }
+    case "LookupTableFind":
+    case "LookupTableFindV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const keys = getParamValue("keys", node, tensorMap, context);
+      const defaultValue = getParamValue("defaultValue", node, tensorMap, context);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [await hashTable.find(keys, defaultValue)];
+    }
+    case "LookupTableSize":
+    case "LookupTableSizeV2": {
+      const handle = getParamValue("tableHandle", node, tensorMap, context, resourceManager);
+      const hashTable = resourceManager.getHashTableById(handle.id);
+      return [hashTable.tensorSize()];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/image_executor.js
+var executeOp10 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "ResizeBilinear": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const alignCorners = getParamValue("alignCorners", node, tensorMap, context);
+      const halfPixelCenters = getParamValue("halfPixelCenters", node, tensorMap, context);
+      return [ops.image.resizeBilinear(images, [size[0], size[1]], alignCorners, halfPixelCenters)];
+    }
+    case "ResizeNearestNeighbor": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      const alignCorners = getParamValue("alignCorners", node, tensorMap, context);
+      const halfPixelCenters = getParamValue("halfPixelCenters", node, tensorMap, context);
+      return [ops.image.resizeNearestNeighbor(images, [size[0], size[1]], alignCorners, halfPixelCenters)];
+    }
+    case "CropAndResize": {
+      const image2 = getParamValue("image", node, tensorMap, context);
+      const boxes = getParamValue("boxes", node, tensorMap, context);
+      const boxInd = getParamValue("boxInd", node, tensorMap, context);
+      const cropSize = getParamValue("cropSize", node, tensorMap, context);
+      const method = getParamValue("method", node, tensorMap, context);
+      const extrapolationValue = getParamValue("extrapolationValue", node, tensorMap, context);
+      return [ops.image.cropAndResize(image2, boxes, boxInd, cropSize, method, extrapolationValue)];
+    }
+    case "ImageProjectiveTransformV3": {
+      const images = getParamValue("images", node, tensorMap, context);
+      const transforms = getParamValue("transforms", node, tensorMap, context);
+      const outputShape = getParamValue("outputShape", node, tensorMap, context);
+      const fillValue = getParamValue("fillValue", node, tensorMap, context);
+      const interpolation = getParamValue("interpolation", node, tensorMap, context);
+      const fillMode = getParamValue("fillMode", node, tensorMap, context);
+      return [ops.image.transform(images, transforms, interpolation.toLowerCase(), fillMode.toLowerCase(), fillValue, outputShape)];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/logical_executor.js
+var executeOp11 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Equal": {
+      return [ops.equal(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "NotEqual": {
+      return [ops.notEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Greater": {
+      return [ops.greater(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "GreaterEqual": {
+      return [ops.greaterEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Less": {
+      return [ops.less(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LessEqual": {
+      return [ops.lessEqual(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LogicalAnd": {
+      return [ops.logicalAnd(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "LogicalNot": {
+      return [ops.logicalNot(getParamValue("a", node, tensorMap, context))];
+    }
+    case "LogicalOr": {
+      return [ops.logicalOr(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    case "Select":
+    case "SelectV2": {
+      return [ops.where(getParamValue("condition", node, tensorMap, context), getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/matrices_executor.js
+var executeOp12 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "BatchMatMul":
+    case "BatchMatMulV2":
+    case "MatMul":
+      return [ops.matMul(getParamValue("a", node, tensorMap, context), getParamValue("b", node, tensorMap, context), getParamValue("transposeA", node, tensorMap, context), getParamValue("transposeB", node, tensorMap, context))];
+    case "Einsum":
+      return [ops.einsum(getParamValue("equation", node, tensorMap, context), ...getParamValue("tensors", node, tensorMap, context))];
+    case "Transpose":
+      return [ops.transpose(getParamValue("x", node, tensorMap, context), getParamValue("perm", node, tensorMap, context))];
+    case "_FusedMatMul":
+      const [extraOp, activationFunc] = getParamValue("fusedOps", node, tensorMap, context);
+      const isBiasAdd = extraOp === "biasadd";
+      const isPrelu = activationFunc === "prelu";
+      const numArgs = getParamValue("numArgs", node, tensorMap, context);
+      const leakyreluAlpha = getParamValue("leakyreluAlpha", node, tensorMap, context);
+      if (isBiasAdd) {
+        if (isPrelu && numArgs !== 2) {
+          throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");
+        }
+        if (!isPrelu && numArgs !== 1) {
+          throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.");
+        }
+      }
+      const [biasArg, preluArg] = getParamValue("args", node, tensorMap, context);
+      return [ops.fused.matMul({
+        a: getParamValue("a", node, tensorMap, context),
+        b: getParamValue("b", node, tensorMap, context),
+        transposeA: getParamValue("transposeA", node, tensorMap, context),
+        transposeB: getParamValue("transposeB", node, tensorMap, context),
+        bias: biasArg,
+        activation: activationFunc,
+        preluActivationWeights: preluArg,
+        leakyreluAlpha
+      })];
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/normalization_executor.js
+var executeOp13 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "EuclideanNorm":
+      return [ops.euclideanNorm(getParamValue("x", node, tensorMap, context), getParamValue("axis", node, tensorMap, context), getParamValue("keepDims", node, tensorMap, context))];
+    case "FusedBatchNorm":
+    case "FusedBatchNormV2": {
+      return [ops.batchNorm(getParamValue("x", node, tensorMap, context), getParamValue("mean", node, tensorMap, context), getParamValue("variance", node, tensorMap, context), getParamValue("offset", node, tensorMap, context), getParamValue("scale", node, tensorMap, context), getParamValue("epsilon", node, tensorMap, context))];
+    }
+    case "FusedBatchNormV3": {
+      return [ops.batchNorm(getParamValue("x", node, tensorMap, context), getParamValue("mean", node, tensorMap, context), getParamValue("variance", node, tensorMap, context), getParamValue("offset", node, tensorMap, context), getParamValue("scale", node, tensorMap, context), getParamValue("epsilon", node, tensorMap, context))];
+    }
+    case "LRN": {
+      return [ops.localResponseNormalization(getParamValue("x", node, tensorMap, context), getParamValue("radius", node, tensorMap, context), getParamValue("bias", node, tensorMap, context), getParamValue("alpha", node, tensorMap, context), getParamValue("beta", node, tensorMap, context))];
+    }
+    case "Softmax": {
+      return [ops.softmax(getParamValue("x", node, tensorMap, context))];
+    }
+    case "LogSoftmax": {
+      return [ops.logSoftmax(getParamValue("x", node, tensorMap, context))];
+    }
+    case "SparseToDense": {
+      return [ops.sparseToDense(getParamValue("sparseIndices", node, tensorMap, context), getParamValue("outputShape", node, tensorMap, context), getParamValue("sparseValues", node, tensorMap, context), getParamValue("defaultValue", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/reduction_executor.js
+var executeOp14 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Max": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.max(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Mean": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.mean(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Min": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.min(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Sum": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.sum(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "All": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.all(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Any": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.any(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "ArgMax": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.argMax(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "ArgMin": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.argMin(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Prod": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const keepDims = getParamValue("keepDims", node, tensorMap, context);
+      return [ops.prod(getParamValue("x", node, tensorMap, context), axis, keepDims)];
+    }
+    case "Cumprod": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const exclusive = getParamValue("exclusive", node, tensorMap, context);
+      const reverse5 = getParamValue("reverse", node, tensorMap, context);
+      return [ops.cumprod(getParamValue("x", node, tensorMap, context), axis, exclusive, reverse5)];
+    }
+    case "Cumsum": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const exclusive = getParamValue("exclusive", node, tensorMap, context);
+      const reverse5 = getParamValue("reverse", node, tensorMap, context);
+      return [ops.cumsum(getParamValue("x", node, tensorMap, context), axis, exclusive, reverse5)];
+    }
+    case "Bincount":
+      const x = getParamValue("x", node, tensorMap, context);
+      const weights = getParamValue("weights", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      return [ops.bincount(x, weights, size)];
+    case "DenseBincount": {
+      const x2 = getParamValue("x", node, tensorMap, context);
+      const weights2 = getParamValue("weights", node, tensorMap, context);
+      const size2 = getParamValue("size", node, tensorMap, context);
+      const binaryOutput = getParamValue("binaryOutput", node, tensorMap, context);
+      return [ops.denseBincount(x2, weights2, size2, binaryOutput)];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/slice_join_executor.js
+var executeOp15 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "ConcatV2":
+    case "Concat": {
+      const n = getParamValue("n", node, tensorMap, context);
+      const axis = getParamValue("axis", node, tensorMap, context);
+      let inputs = getParamValue("tensors", node, tensorMap, context);
+      inputs = inputs.slice(0, n);
+      return [ops.concat(inputs, axis)];
+    }
+    case "Gather": {
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gather(input2, ops.cast(indices, "int32"), 0)];
+    }
+    case "GatherV2": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const batchDims = getParamValue("batchDims", node, tensorMap, context);
+      const input2 = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gather(input2, ops.cast(indices, "int32"), axis, batchDims)];
+    }
+    case "Reverse": {
+      const dims = getParamValue("dims", node, tensorMap, context);
+      const axis = [];
+      for (let i = 0; i < dims.length; i++) {
+        if (dims[i]) {
+          axis.push(i);
+        }
+      }
+      const input2 = getParamValue("x", node, tensorMap, context);
+      return [ops.reverse(input2, axis)];
+    }
+    case "ReverseV2": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const input2 = getParamValue("x", node, tensorMap, context);
+      return [ops.reverse(input2, axis)];
+    }
+    case "Slice": {
+      const begin = getParamValue("begin", node, tensorMap, context);
+      const size = getParamValue("size", node, tensorMap, context);
+      return [ops.slice(getParamValue("x", node, tensorMap, context), begin, size)];
+    }
+    case "StridedSlice": {
+      const begin = getParamValue("begin", node, tensorMap, context);
+      const end = getParamValue("end", node, tensorMap, context);
+      const strides = getParamValue("strides", node, tensorMap, context);
+      const beginMask = getParamValue("beginMask", node, tensorMap, context);
+      const endMask = getParamValue("endMask", node, tensorMap, context);
+      const ellipsisMask = getParamValue("ellipsisMask", node, tensorMap, context);
+      const newAxisMask = getParamValue("newAxisMask", node, tensorMap, context);
+      const shrinkAxisMask = getParamValue("shrinkAxisMask", node, tensorMap, context);
+      const tensor2 = getParamValue("x", node, tensorMap, context);
+      return [ops.stridedSlice(tensor2, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask)];
+    }
+    case "Pack": {
+      return tidy(() => {
+        const axis = getParamValue("axis", node, tensorMap, context);
+        const tensors = getParamValue("tensors", node, tensorMap, context);
+        const shape = tensors[0].shape;
+        const squeezedShape = ops.squeeze(tensors[0]).shape;
+        const mapped = tensors.map((tensor2) => {
+          const sameShape = util_exports.arraysEqual(tensor2.shape, shape);
+          if (!sameShape && !util_exports.arraysEqual(ops.squeeze(tensor2).shape, squeezedShape)) {
+            throw new Error("the input tensors shape does not match");
+          }
+          return sameShape ? tensor2 : ops.reshape(tensor2, shape);
+        });
+        return [ops.stack(mapped, axis)];
+      });
+    }
+    case "Unpack": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const tensor2 = getParamValue("tensor", node, tensorMap, context);
+      return ops.unstack(tensor2, axis);
+    }
+    case "Tile": {
+      const reps = getParamValue("reps", node, tensorMap, context);
+      return [ops.tile(getParamValue("x", node, tensorMap, context), reps)];
+    }
+    case "Split":
+    case "SplitV": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      const numOrSizeSplits = getParamValue("numOrSizeSplits", node, tensorMap, context);
+      const tensor2 = getParamValue("x", node, tensorMap, context);
+      return ops.split(tensor2, numOrSizeSplits, axis);
+    }
+    case "ScatterNd": {
+      const indices = getParamValue("indices", node, tensorMap, context);
+      const values = getParamValue("values", node, tensorMap, context);
+      const shape = getParamValue("shape", node, tensorMap, context);
+      return [ops.scatterND(indices, values, shape)];
+    }
+    case "GatherNd": {
+      const x = getParamValue("x", node, tensorMap, context);
+      const indices = getParamValue("indices", node, tensorMap, context);
+      return [ops.gatherND(x, indices)];
+    }
+    case "SparseToDense": {
+      const indices = getParamValue("sparseIndices", node, tensorMap, context);
+      const shape = getParamValue("outputShape", node, tensorMap, context);
+      const sparseValues = getParamValue("sparseValues", node, tensorMap, context);
+      const defaultValue = getParamValue("defaultValue", node, tensorMap, context);
+      return [ops.sparseToDense(indices, sparseValues, shape, sparseValues.dtype === defaultValue.dtype ? defaultValue : ops.cast(defaultValue, sparseValues.dtype))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/sparse_executor.js
+var executeOp16 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "SparseFillEmptyRows": {
+      const { outputIndices, outputValues, emptyRowIndicator, reverseIndexMap } = ops.sparse.sparseFillEmptyRows(getParamValue("indices", node, tensorMap, context), getParamValue("values", node, tensorMap, context), getParamValue("denseShape", node, tensorMap, context), getParamValue("defaultValue", node, tensorMap, context));
+      return [
+        outputIndices,
+        outputValues,
+        emptyRowIndicator,
+        reverseIndexMap
+      ];
+    }
+    case "SparseReshape": {
+      const { outputIndices, outputShape } = ops.sparse.sparseReshape(getParamValue("inputIndices", node, tensorMap, context), getParamValue("inputShape", node, tensorMap, context), getParamValue("newShape", node, tensorMap, context));
+      return [outputIndices, outputShape];
+    }
+    case "SparseSegmentMean": {
+      const outputData = ops.sparse.sparseSegmentMean(getParamValue("data", node, tensorMap, context), getParamValue("indices", node, tensorMap, context), getParamValue("segmentIds", node, tensorMap, context));
+      return [outputData];
+    }
+    case "SparseSegmentSum": {
+      const outputData = ops.sparse.sparseSegmentSum(getParamValue("data", node, tensorMap, context), getParamValue("indices", node, tensorMap, context), getParamValue("segmentIds", node, tensorMap, context));
+      return [outputData];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/spectral_executor.js
+var executeOp17 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "FFT": {
+      return [ops.fft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "IFFT": {
+      return [ops.ifft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "RFFT": {
+      return [ops.rfft(getParamValue("x", node, tensorMap, context))];
+    }
+    case "IRFFT": {
+      return [ops.irfft(getParamValue("x", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/string_executor.js
+var executeOp18 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "StringNGrams": {
+      const { nGrams, nGramsSplits } = ops.string.stringNGrams(getParamValue("data", node, tensorMap, context), getParamValue("dataSplits", node, tensorMap, context), getParamValue("separator", node, tensorMap, context), getParamValue("nGramWidths", node, tensorMap, context), getParamValue("leftPad", node, tensorMap, context), getParamValue("rightPad", node, tensorMap, context), getParamValue("padWidth", node, tensorMap, context), getParamValue("preserveShortSequences", node, tensorMap, context));
+      return [nGrams, nGramsSplits];
+    }
+    case "StringSplit": {
+      const { indices, values, shape } = ops.string.stringSplit(getParamValue("input", node, tensorMap, context), getParamValue("delimiter", node, tensorMap, context), getParamValue("skipEmpty", node, tensorMap, context));
+      return [indices, values, shape];
+    }
+    case "StringToHashBucketFast": {
+      const output = ops.string.stringToHashBucketFast(getParamValue("input", node, tensorMap, context), getParamValue("numBuckets", node, tensorMap, context));
+      return [output];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/executors/transformation_executor.js
+var executeOp19 = (node, tensorMap, context, ops = ops_for_converter_exports) => {
+  switch (node.op) {
+    case "Cast": {
+      return [ops.cast(getParamValue("x", node, tensorMap, context), getParamValue("dtype", node, tensorMap, context))];
+    }
+    case "ExpandDims": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.expandDims(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Squeeze": {
+      const axis = getParamValue("axis", node, tensorMap, context);
+      return [ops.squeeze(getParamValue("x", node, tensorMap, context), axis)];
+    }
+    case "Reshape": {
+      return [ops.reshape(getParamValue("x", node, tensorMap, context), getParamValue("shape", node, tensorMap, context))];
+    }
+    case "MirrorPad": {
+      return [ops.mirrorPad(getParamValue("x", node, tensorMap, context), getParamValue("padding", node, tensorMap, context), getParamValue("mode", node, tensorMap, context))];
+    }
+    case "PadV2":
+    case "Pad": {
+      return [ops.pad(getParamValue("x", node, tensorMap, context), getParamValue("padding", node, tensorMap, context), getParamValue("constantValue", node, tensorMap, context))];
+    }
+    case "SpaceToBatchND": {
+      const blockShape = getParamValue("blockShape", node, tensorMap, context);
+      const paddings = getParamValue("paddings", node, tensorMap, context);
+      return [ops.spaceToBatchND(getParamValue("x", node, tensorMap, context), blockShape, paddings)];
+    }
+    case "BatchToSpaceND": {
+      const blockShape = getParamValue("blockShape", node, tensorMap, context);
+      const crops = getParamValue("crops", node, tensorMap, context);
+      return [ops.batchToSpaceND(getParamValue("x", node, tensorMap, context), blockShape, crops)];
+    }
+    case "DepthToSpace": {
+      const blockSize = getParamValue("blockSize", node, tensorMap, context);
+      const dataFormat = getParamValue("dataFormat", node, tensorMap, context).toUpperCase();
+      return [ops.depthToSpace(getParamValue("x", node, tensorMap, context), blockSize, dataFormat)];
+    }
+    case "BroadcastTo": {
+      return [ops.broadcastTo(getParamValue("x", node, tensorMap, context), getParamValue("shape", node, tensorMap, context))];
+    }
+    case "BroadcastArgs": {
+      return [ops.broadcastArgs(getParamValue("s0", node, tensorMap, context), getParamValue("s1", node, tensorMap, context))];
+    }
+    default:
+      throw TypeError(`Node type ${node.op} is not implemented`);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/operations/operation_executor.js
+function executeOp20(node, tensorMap, context, resourceManager, tidy2 = tidy) {
+  const value = ((node2, tensorMap2, context2) => {
+    switch (node2.category) {
+      case "arithmetic":
+        return tidy2(() => executeOp(node2, tensorMap2, context2));
+      case "basic_math":
+        return tidy2(() => executeOp2(node2, tensorMap2, context2));
+      case "control":
+        return executeOp3(node2, tensorMap2, context2);
+      case "convolution":
+        return tidy2(() => executeOp4(node2, tensorMap2, context2));
+      case "creation":
+        return tidy2(() => executeOp5(node2, tensorMap2, context2));
+      case "dynamic":
+        return executeOp6(node2, tensorMap2, context2);
+      case "evaluation":
+        return tidy2(() => executeOp7(node2, tensorMap2, context2));
+      case "image":
+        return tidy2(() => executeOp10(node2, tensorMap2, context2));
+      case "graph":
+        return tidy2(() => executeOp8(node2, tensorMap2, context2));
+      case "logical":
+        return tidy2(() => executeOp11(node2, tensorMap2, context2));
+      case "matrices":
+        return tidy2(() => executeOp12(node2, tensorMap2, context2));
+      case "normalization":
+        return tidy2(() => executeOp13(node2, tensorMap2, context2));
+      case "reduction":
+        return tidy2(() => executeOp14(node2, tensorMap2, context2));
+      case "slice_join":
+        return tidy2(() => executeOp15(node2, tensorMap2, context2));
+      case "sparse":
+        return tidy2(() => executeOp16(node2, tensorMap2, context2));
+      case "spectral":
+        return tidy2(() => executeOp17(node2, tensorMap2, context2));
+      case "string":
+        return tidy2(() => executeOp18(node2, tensorMap2, context2));
+      case "transformation":
+        return tidy2(() => executeOp19(node2, tensorMap2, context2));
+      case "hash_table":
+        return executeOp9(node2, tensorMap2, context2, resourceManager);
+      case "custom":
+        const opMapper = getRegisteredOp(node2.op);
+        if (opMapper && opMapper.customExecutor) {
+          return opMapper.customExecutor(new NodeValueImpl(node2, tensorMap2, context2));
+        } else {
+          throw TypeError(`Custom op ${node2.op} is not registered.`);
+        }
+      default:
+        throw TypeError(`Unknown op '${node2.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`);
+    }
+  })(node, tensorMap, context);
+  if (util_exports.isPromise(value)) {
+    return value.then((data) => [].concat(data));
+  }
+  return [].concat(value);
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/execution_context.js
+var ExecutionContext = class {
+  constructor(weightMap = {}, tensorArrayMap = {}, tensorListMap = {}, functionMap = {}) {
+    this.weightMap = weightMap;
+    this.tensorArrayMap = tensorArrayMap;
+    this.tensorListMap = tensorListMap;
+    this.functionMap = functionMap;
+    this.rootContext = { id: 0, frameName: "", iterationId: 0 };
+    this.contexts = [this.rootContext];
+    this.lastId = 0;
+    this.generateCurrentContextIds();
+  }
+  newFrame(id, frameName) {
+    return { id, frameName, iterationId: 0 };
+  }
+  set currentContext(contexts2) {
+    if (this.contexts !== contexts2) {
+      this.contexts = contexts2;
+      this.generateCurrentContextIds();
+    }
+  }
+  get currentContext() {
+    return this.contexts;
+  }
+  get currentContextId() {
+    return this._currentContextIds[0];
+  }
+  get currentContextIds() {
+    return this._currentContextIds;
+  }
+  generateCurrentContextIds() {
+    const names = [];
+    for (let i = 0; i < this.contexts.length - 1; i++) {
+      const contexts2 = this.contexts.slice(0, this.contexts.length - i);
+      names.push(this.contextIdforContexts(contexts2));
+    }
+    names.push("");
+    this._currentContextIds = names;
+  }
+  contextIdforContexts(contexts2) {
+    return contexts2 ? contexts2.map((context) => context.id === 0 && context.iterationId === 0 ? "" : `${context.frameName}-${context.iterationId}`).join("/") : "";
+  }
+  enterFrame(frameId) {
+    if (this.contexts) {
+      this.lastId++;
+      this.contexts = this.contexts.slice();
+      this.contexts.push(this.newFrame(this.lastId, frameId));
+      this._currentContextIds.unshift(this.contextIdforContexts(this.contexts));
+    }
+  }
+  exitFrame() {
+    if (this.contexts && this.contexts.length > 1) {
+      this.contexts = this.contexts.slice();
+      this.contexts.splice(-1);
+      this.currentContextIds.shift();
+    } else {
+      throw new Error("Cannot exit frame, the context is empty");
+    }
+  }
+  nextIteration() {
+    if (this.contexts && this.contexts.length > 0) {
+      this.contexts = this.contexts.slice();
+      this.lastId++;
+      const context = Object.assign({}, this.contexts[this.contexts.length - 1]);
+      context.iterationId += 1;
+      context.id = this.lastId;
+      this.contexts.splice(-1, 1, context);
+      this._currentContextIds.splice(0, 1, this.contextIdforContexts(this.contexts));
+    } else {
+      throw new Error("Cannot increase frame iteration, the context is empty");
+    }
+  }
+  getWeight(name) {
+    return this.weightMap[name];
+  }
+  addTensorArray(tensorArray) {
+    this.tensorArrayMap[tensorArray.id] = tensorArray;
+  }
+  getTensorArray(id) {
+    return this.tensorArrayMap[id];
+  }
+  addTensorList(tensorList) {
+    this.tensorListMap[tensorList.id] = tensorList;
+  }
+  getTensorList(id) {
+    return this.tensorListMap[id];
+  }
+  dispose(keepIds) {
+    for (const key in this.tensorArrayMap) {
+      this.tensorArrayMap[key].clearAndClose(keepIds);
+    }
+    for (const key in this.tensorListMap) {
+      this.tensorListMap[key].clearAndClose(keepIds);
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/model_analysis.js
+function getExecutionSubgraph(inputs, outputs, weightMap, initNodes) {
+  const usedNodes = /* @__PURE__ */ new Set();
+  const missingInputs = [];
+  let dynamicNode = null;
+  let syncInputs = null;
+  const seen = /* @__PURE__ */ new Set();
+  const inputNodeNames = Object.keys(inputs).map((name) => parseNodeName(name)[0]);
+  let initNodeNames = [];
+  if (initNodes != null) {
+    initNodeNames = initNodes.map((node) => parseNodeName(node.name)[0]);
+  }
+  const frontier = [...outputs];
+  while (frontier.length > 0) {
+    const node = frontier.pop();
+    if (isControlFlow(node) || isDynamicShape(node) || isHashTable(node)) {
+      if (dynamicNode == null) {
+        dynamicNode = node;
+        syncInputs = dynamicNode.children.map((child) => child.name).filter((name) => usedNodes.has(name));
+      }
+    }
+    usedNodes.add(node.name);
+    if (weightMap[node.name] != null) {
+      continue;
+    }
+    if (inputNodeNames.indexOf(node.name) !== -1) {
+      continue;
+    }
+    if (initNodeNames.indexOf(node.name) !== -1) {
+      continue;
+    }
+    if (node.inputs.length === 0) {
+      missingInputs.push(node.name);
+      continue;
+    }
+    node.inputs.forEach((input2) => {
+      if (seen.has(input2.name)) {
+        return;
+      }
+      seen.add(input2.name);
+      frontier.push(input2);
+    });
+  }
+  return { inputs, outputs, usedNodes, missingInputs, dynamicNode, syncInputs };
+}
+function getNodesInTopologicalOrder(graph, weightMap, executionInfo) {
+  const { usedNodes, inputs } = executionInfo;
+  const frontier = [];
+  const inputNodes = Object.keys(inputs).map((name) => parseNodeName(name)[0]).map((name) => graph.nodes[name]);
+  const initNodes = graph.initNodes;
+  inputNodes.forEach((input2) => {
+    if (usedNodes.has(input2.name)) {
+      frontier.push(input2);
+    }
+  });
+  graph.weights.forEach((weight) => {
+    if (usedNodes.has(weight.name)) {
+      frontier.push(weight);
+    }
+  });
+  if (initNodes != null) {
+    initNodes.forEach((node) => {
+      if (usedNodes.has(node.name)) {
+        frontier.push(node);
+      }
+    });
+  }
+  const seen = /* @__PURE__ */ new Set();
+  const orderedNodes = [];
+  while (frontier.length > 0) {
+    const node = frontier.pop();
+    seen.add(node.name);
+    if (!weightMap[node.name]) {
+      orderedNodes.push(node);
+    }
+    node.children.forEach((child) => {
+      if (!seen.has(child.name) && usedNodes.has(child.name) && child.inputs.every((input2) => seen.has(input2.name))) {
+        frontier.push(child);
+      }
+    });
+  }
+  return orderedNodes;
+}
+var CONTROL_FLOW_OPS = [
+  "Switch",
+  "Merge",
+  "Enter",
+  "Exit",
+  "NextIteration",
+  "StatelessIf",
+  "StatelessWhile",
+  "if",
+  "While"
+];
+var DYNAMIC_SHAPE_OPS = [
+  "NonMaxSuppressionV2",
+  "NonMaxSuppressionV3",
+  "NonMaxSuppressionV5",
+  "Where"
+];
+var HASH_TABLE_OPS = [
+  "HashTable",
+  "HashTableV2",
+  "LookupTableImport",
+  "LookupTableImportV2",
+  "LookupTableFind",
+  "LookupTableFindV2",
+  "LookupTableSize",
+  "LookupTableSizeV2"
+];
+function isControlFlow(node) {
+  return CONTROL_FLOW_OPS.indexOf(node.op) >= 0;
+}
+function isDynamicShape(node) {
+  return DYNAMIC_SHAPE_OPS.indexOf(node.op) >= 0;
+}
+function isHashTable(node) {
+  return HASH_TABLE_OPS.indexOf(node.op) >= 0;
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/graph_executor.js
+var GraphExecutor = class {
+  constructor(graph, parent) {
+    this.graph = graph;
+    this.parent = parent;
+    this.compiledMap = /* @__PURE__ */ new Map();
+    this._weightMap = {};
+    this.SEPERATOR = ",";
+    this._functions = {};
+    this._functionExecutorMap = {};
+    this.intermediateTensors = {};
+    this.keepTensorForDebug = false;
+    this._outputs = graph.outputs;
+    this._inputs = graph.inputs;
+    this._initNodes = graph.initNodes;
+    this._signature = graph.signature;
+    this._functions = graph.functions;
+    if (graph.functions != null) {
+      Object.keys(graph.functions).forEach((name) => {
+        this._functionExecutorMap[name] = new GraphExecutor(graph.functions[name], this);
+      });
+    }
+  }
+  get weightIds() {
+    return this.parent ? this.parent.weightIds : this._weightIds;
+  }
+  get functionExecutorMap() {
+    return this.parent ? this.parent.functionExecutorMap : this._functionExecutorMap;
+  }
+  get weightMap() {
+    return this.parent ? this.parent.weightMap : this._weightMap;
+  }
+  set weightMap(weightMap) {
+    const weightIds = Object.keys(weightMap).map((key) => weightMap[key].map((tensor2) => tensor2.id));
+    this._weightIds = [].concat(...weightIds);
+    this._weightMap = weightMap;
+  }
+  set resourceManager(resourceManager) {
+    this._resourceManager = resourceManager;
+  }
+  get inputs() {
+    return this._inputs.map((node) => {
+      return {
+        name: node.name,
+        shape: node.attrParams["shape"] ? node.attrParams["shape"].value : void 0,
+        dtype: node.attrParams["dtype"] ? node.attrParams["dtype"].value : void 0
+      };
+    });
+  }
+  get outputs() {
+    return this._outputs.map((node) => {
+      return {
+        name: node.name,
+        shape: node.attrParams["shape"] ? node.attrParams["shape"].value : void 0,
+        dtype: node.attrParams["dtype"] ? node.attrParams["dtype"].value : void 0
+      };
+    });
+  }
+  get inputNodes() {
+    return this._inputs.map((node) => node.signatureKey || node.name);
+  }
+  get outputNodes() {
+    return this._outputs.map((node) => {
+      const name = node.signatureKey || node.name;
+      return node.defaultOutput ? `${name}:${node.defaultOutput}` : name;
+    });
+  }
+  get functions() {
+    return Object.keys(this._functions).reduce((map, key) => {
+      map[key] = this._functions[key].signature;
+      return map;
+    }, {});
+  }
+  getCompilationKey(inputs, outputs) {
+    const sortedInputs = inputs.map((node) => node.name).sort();
+    const sortedOutputs = outputs.map((node) => node.name).sort();
+    return sortedInputs.join(this.SEPERATOR) + "--" + sortedOutputs.join(this.SEPERATOR);
+  }
+  compile(inputs, outputs) {
+    const executionInfo = getExecutionSubgraph(inputs, outputs, this.weightMap, this._initNodes);
+    const { missingInputs, dynamicNode, syncInputs } = executionInfo;
+    if (dynamicNode != null) {
+      throw new Error(`This execution contains the node '${dynamicNode.name}', which has the dynamic op '${dynamicNode.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${syncInputs}]`);
+    }
+    if (missingInputs.length > 0) {
+      const outNames = outputs.map((n) => n.name);
+      const inNames = Object.keys(inputs);
+      throw new Error(`Cannot compute the outputs [${outNames}] from the provided inputs [${inNames}]. Missing the following inputs: [${missingInputs}]`);
+    }
+    return getNodesInTopologicalOrder(this.graph, this.weightMap, executionInfo);
+  }
+  execute(inputs, outputs) {
+    inputs = this.mapInputs(inputs);
+    const names = Object.keys(inputs).sort();
+    this.checkInputs(inputs);
+    this.checkInputShapeAndType(inputs);
+    outputs = this.mapOutputs(outputs);
+    this.checkOutputs(outputs);
+    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);
+    const outputNodeNames = outputs.map((name) => parseNodeName(name)[0]);
+    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);
+    this.resetIntermediateTensors();
+    if (outputNodes.length === 0) {
+      outputNodes = this._outputs;
+    }
+    const compilationKey = this.getCompilationKey(inputNodes, outputNodes);
+    let orderedNodes = this.compiledMap.get(compilationKey);
+    if (orderedNodes == null) {
+      orderedNodes = this.compile(inputs, outputNodes);
+      this.compiledMap.set(compilationKey, orderedNodes);
+    }
+    const tensorArrayMap = {};
+    const tensorListMap = {};
+    return tidy(() => {
+      const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);
+      const tensorsMap = Object.assign({}, this.weightMap);
+      Object.keys(inputs).forEach((name) => {
+        const [nodeName, index] = parseNodeName(name);
+        const tensors = [];
+        tensors[index] = inputs[name];
+        tensorsMap[nodeName] = tensors;
+      });
+      const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);
+      const intermediateTensorConsumerCount = {};
+      for (let i = 0; i < orderedNodes.length; i++) {
+        const node = orderedNodes[i];
+        if (!tensorsMap[node.name]) {
+          const tensors = executeOp20(node, tensorsMap, context, this._resourceManager);
+          if (util_exports.isPromise(tensors)) {
+            throw new Error(`The execution of the op '${node.op}' returned a promise. Please use model.executeAsync() instead.`);
+          }
+          tensorsMap[node.name] = tensors;
+          this.checkTensorForDisposal(node.name, node, tensorsMap, context, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount);
+        }
+      }
+      if (this.parent == null) {
+        context.dispose(tensorsToKeep);
+      }
+      return outputs.map((name) => getTensor(name, tensorsMap, context));
+    });
+  }
+  getFrozenTensorIds(tensorMap) {
+    const ids = [].concat.apply([], Object.keys(tensorMap).map((key) => tensorMap[key]).map((tensors) => tensors.map((tensor2) => tensor2.id)));
+    return new Set(ids);
+  }
+  checkTensorForDisposal(nodeName, node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount) {
+    if (node.category === "control" || outputNames.indexOf(nodeName) !== -1) {
+      return;
+    }
+    tensorMap[nodeName].forEach((tensor2) => {
+      if (tensor2 != null) {
+        intermediateTensorConsumerCount[tensor2.id] = (intermediateTensorConsumerCount[tensor2.id] || 0) + node.children.length;
+      }
+    });
+    node.inputs.forEach((input2) => {
+      if (input2.category !== "control") {
+        const tensors = getTensorsForCurrentContenxt(input2.name, tensorMap, context);
+        if (tensors != null) {
+          tensors.forEach((tensor2) => {
+            if (tensor2 && !tensor2.kept && !tensorsToKeep.has(tensor2.id)) {
+              const count2 = intermediateTensorConsumerCount[tensor2.id];
+              if (count2 === 1) {
+                if (!this.keepTensorForDebug) {
+                  tensor2.dispose();
+                } else {
+                  const [nodeName2, index] = getNodeNameAndIndex(node.name, context);
+                  if (this.intermediateTensors[nodeName2]) {
+                    this.intermediateTensors[nodeName2][index] = tensor2;
+                  } else {
+                    this.intermediateTensors[nodeName2] = [];
+                    this.intermediateTensors[nodeName2][index] = tensor2;
+                  }
+                }
+                delete intermediateTensorConsumerCount[tensor2.id];
+              } else if (count2 != null) {
+                intermediateTensorConsumerCount[tensor2.id]--;
+              }
+            }
+          });
+        }
+      }
+    });
+  }
+  async executeAsync(inputs, outputs) {
+    return this._executeAsync(inputs, outputs);
+  }
+  disposeIntermediateTensors() {
+    if (!this.intermediateTensors) {
+      return;
+    }
+    Object.keys(this.intermediateTensors).forEach((key) => this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose()));
+    this.disposeTensorsMap();
+  }
+  disposeTensorsMap() {
+    if (!this.tensorsMap) {
+      return;
+    }
+    Object.keys(this.tensorsMap).forEach((key) => {
+      const tensorArray = this.tensorsMap[key];
+      tensorArray.forEach((tensor2) => {
+        if (tensor2 && !tensor2.kept && !tensor2.isDisposed && !this.keepIds.has(tensor2.id)) {
+          tensor2.dispose();
+        }
+      });
+    });
+  }
+  getIntermediateTensors() {
+    return this.tensorsMap;
+  }
+  resetIntermediateTensors() {
+    for (const key in this.intermediateTensors) {
+      this.intermediateTensors[key].forEach((tensor2) => tensor2.dispose());
+      delete this.intermediateTensors[key];
+    }
+  }
+  async _executeAsync(inputs, outputs, isFunctionExecution = false, tensorArrayMap = {}, tensorListMap = {}) {
+    if (!isFunctionExecution) {
+      inputs = this.mapInputs(inputs);
+      this.checkInputs(inputs);
+      this.checkInputShapeAndType(inputs);
+      outputs = this.mapOutputs(outputs);
+      this.checkOutputs(outputs);
+    }
+    try {
+      this.keepTensorForDebug = env().getBool("KEEP_INTERMEDIATE_TENSORS");
+    } catch (e) {
+      console.warn(e.message);
+    }
+    this.resetIntermediateTensors();
+    const context = new ExecutionContext(this.weightMap, tensorArrayMap, tensorListMap, this.functionExecutorMap);
+    this.tensorsMap = await this.executeWithControlFlow(inputs, context, outputs, isFunctionExecution);
+    const results = outputs.map((name) => getTensor(name, this.tensorsMap, context));
+    const outputIds = results.map((t) => t.id);
+    const inputIds = Object.keys(inputs).map((name) => inputs[name].id);
+    this.keepIds = /* @__PURE__ */ new Set([...outputIds, ...inputIds, ...this.weightIds]);
+    if (!this.keepTensorForDebug) {
+      this.disposeTensorsMap();
+    }
+    if (this.parent == null) {
+      context.dispose(this.keepIds);
+    }
+    return results;
+  }
+  async executeFunctionAsync(inputs, tensorArrayMap, tensorListMap) {
+    const mappedInputs = inputs.reduce((map, tensor2, index) => {
+      map[this.inputs[index].name] = tensor2;
+      return map;
+    }, {});
+    return this._executeAsync(mappedInputs, this.outputNodes, true, tensorArrayMap, tensorListMap);
+  }
+  async executeWithControlFlow(inputs, context, outputNames, isFunctionExecution) {
+    const names = Object.keys(inputs);
+    const inputNodes = names.map((name) => this.graph.nodes[parseNodeName(name)[0]]);
+    const outputNodeNames = outputNames.map((name) => parseNodeName(name)[0]);
+    let outputNodes = outputNodeNames.map((name) => this.graph.nodes[name]);
+    if (outputNodes.length === 0) {
+      outputNodes = this._outputs;
+    }
+    const { usedNodes, missingInputs, dynamicNode, syncInputs } = getExecutionSubgraph(inputs, outputNodes, this.weightMap, this._initNodes);
+    const stack2 = [
+      ...inputNodes,
+      ...this.graph.weights,
+      ...this._initNodes || []
+    ].map((node) => {
+      return { node, contexts: context.currentContext };
+    });
+    const tensorsMap = Object.assign({}, this.weightMap);
+    Object.keys(inputs).forEach((name) => {
+      const [nodeName, index] = parseNodeName(name);
+      const tensors = [];
+      tensors[index] = inputs[name];
+      tensorsMap[nodeName] = tensors;
+    });
+    const intermediateTensorConsumerCount = {};
+    const tensorsToKeep = this.getFrozenTensorIds(tensorsMap);
+    const added = {};
+    while (stack2.length > 0) {
+      const promises = this.processStack(inputNodes, stack2, context, tensorsMap, added, tensorsToKeep, outputNodeNames, intermediateTensorConsumerCount, usedNodes);
+      await Promise.all(promises);
+    }
+    if (dynamicNode == null && !isFunctionExecution) {
+      console.warn(`This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.`);
+    }
+    const missingOutputs = outputNodes.filter((node) => !isControlFlow(node) && !getTensor(node.name, tensorsMap, context)).map((node) => node.name);
+    if (missingOutputs.length > 0) {
+      let alternativeMsg = "";
+      if (dynamicNode != null) {
+        alternativeMsg = `Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${syncInputs}]`;
+      }
+      throw new Error(`Cannot compute the outputs [${missingOutputs}] from the provided inputs [${names}]. Consider providing the following inputs: [${missingInputs}]. ${alternativeMsg}`);
+    }
+    return tensorsMap;
+  }
+  processStack(inputNodes, stack2, context, tensorMap, added, tensorsToKeep, outputNames, intermediateTensorConsumerCount, usedNodes) {
+    const promises = [];
+    while (stack2.length > 0) {
+      const item = stack2.pop();
+      context.currentContext = item.contexts;
+      let nodeName = "";
+      if (item.node.op === "Enter" && getParamValue("isConstant", item.node, tensorMap, context)) {
+        [nodeName] = getNodeNameAndIndex(item.node.name, context);
+      }
+      if (tensorMap[item.node.name] == null) {
+        const tensors = executeOp20(item.node, tensorMap, context, this._resourceManager);
+        if (!nodeName) {
+          [nodeName] = getNodeNameAndIndex(item.node.name, context);
+        }
+        const currentContext = context.currentContext;
+        if (util_exports.isPromise(tensors)) {
+          promises.push(tensors.then((t) => {
+            tensorMap[nodeName] = t;
+            context.currentContext = currentContext;
+            this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);
+            this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+            return t;
+          }));
+        } else {
+          tensorMap[nodeName] = tensors;
+          this.checkTensorForDisposal(nodeName, item.node, tensorMap, context, tensorsToKeep, outputNames, intermediateTensorConsumerCount);
+          this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+        }
+      } else {
+        this.processChildNodes(item.node, stack2, context, tensorMap, added, usedNodes);
+      }
+    }
+    return promises;
+  }
+  processChildNodes(node, stack2, context, tensorMap, added, usedNodes) {
+    node.children.forEach((childNode) => {
+      const [nodeName] = getNodeNameAndIndex(childNode.name, context);
+      if (added[nodeName] || !usedNodes.has(childNode.name)) {
+        return;
+      }
+      if (childNode.op === "Merge") {
+        if (childNode.inputNames.some((name) => {
+          return !!getTensor(name, tensorMap, context);
+        })) {
+          added[nodeName] = true;
+          stack2.push({ contexts: context.currentContext, node: childNode });
+        }
+      } else if (childNode.inputNames.every((name) => {
+        return !!getTensor(name, tensorMap, context);
+      })) {
+        added[nodeName] = true;
+        stack2.push({ contexts: context.currentContext, node: childNode });
+      }
+    });
+  }
+  dispose() {
+    Object.keys(this.weightMap).forEach((key) => this.weightMap[key].forEach((tensor2) => tensor2.dispose()));
+  }
+  checkInputShapeAndType(inputs) {
+    Object.keys(inputs).forEach((name) => {
+      const input2 = inputs[name];
+      const [nodeName] = parseNodeName(name);
+      const node = this.graph.nodes[nodeName];
+      if (node.attrParams["shape"] && node.attrParams["shape"].value) {
+        const shape = node.attrParams["shape"].value;
+        const match = shape.length === input2.shape.length && input2.shape.every((dim, index) => shape[index] === -1 || shape[index] === dim);
+        util_exports.assert(match, () => `The shape of dict['${node.name}'] provided in model.execute(dict) must be [${shape}], but was [${input2.shape}]`);
+      }
+      if (node.attrParams["dtype"] && node.attrParams["dtype"].value) {
+        util_exports.assert(input2.dtype === node.attrParams["dtype"].value, () => `The dtype of dict['${node.name}'] provided in model.execute(dict) must be ${node.attrParams["dtype"].value}, but was ${input2.dtype}`);
+      }
+    });
+  }
+  mapInputs(inputs) {
+    const result = {};
+    for (const inputName in inputs) {
+      if (this._signature != null && this._signature.inputs != null && this._signature.inputs[inputName] != null) {
+        const tensor2 = this._signature.inputs[inputName];
+        result[tensor2.name] = inputs[inputName];
+      } else {
+        result[inputName] = inputs[inputName];
+      }
+    }
+    return result;
+  }
+  checkInputs(inputs) {
+    const notInGraph = Object.keys(inputs).filter((name) => {
+      const [nodeName] = parseNodeName(name);
+      return this.graph.nodes[nodeName] == null;
+    });
+    if (notInGraph.length > 0) {
+      throw new Error(`The dict provided in model.execute(dict) has keys: [${notInGraph}] that are not part of graph`);
+    }
+  }
+  mapOutputs(outputs) {
+    return outputs.map((name) => {
+      if (this._signature != null && this._signature.outputs != null && this._signature.outputs[name] != null) {
+        const tensor2 = this._signature.outputs[name];
+        return tensor2.name;
+      }
+      return name;
+    }, {});
+  }
+  checkOutputs(outputs) {
+    outputs.forEach((name) => {
+      const [normalizedName] = parseNodeName(name);
+      if (!this.graph.nodes[normalizedName]) {
+        throw new Error(`The output '${name}' is not found in the graph`);
+      }
+    });
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/resource_manager.js
+var ResourceManager = class {
+  constructor(hashTableNameToHandle = {}, hashTableMap = {}) {
+    this.hashTableNameToHandle = hashTableNameToHandle;
+    this.hashTableMap = hashTableMap;
+  }
+  addHashTable(name, hashTable) {
+    this.hashTableNameToHandle[name] = hashTable.handle;
+    this.hashTableMap[hashTable.id] = hashTable;
+  }
+  getHashTableHandleByName(name) {
+    return this.hashTableNameToHandle[name];
+  }
+  getHashTableById(id) {
+    return this.hashTableMap[id];
+  }
+  dispose() {
+    for (const key in this.hashTableMap) {
+      this.hashTableMap[key].clearAndClose();
+      delete this.hashTableMap[key];
+    }
+    for (const name in this.hashTableNameToHandle) {
+      this.hashTableNameToHandle[name].dispose();
+      delete this.hashTableNameToHandle[name];
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-converter/dist/executor/graph_model.js
+var TFHUB_SEARCH_PARAM = "?tfjs-format=file";
+var DEFAULT_MODEL_NAME = "model.json";
+var GraphModel = class {
+  constructor(modelUrl, loadOptions = {}, tfio = io_exports) {
+    this.modelUrl = modelUrl;
+    this.loadOptions = loadOptions;
+    this.version = "n/a";
+    this.io = tfio;
+    if (loadOptions == null) {
+      this.loadOptions = {};
+    }
+    this.resourceManager = new ResourceManager();
+  }
+  get modelVersion() {
+    return this.version;
+  }
+  get inputNodes() {
+    return this.executor.inputNodes;
+  }
+  get outputNodes() {
+    return this.executor.outputNodes;
+  }
+  get inputs() {
+    return this.executor.inputs;
+  }
+  get outputs() {
+    return this.executor.outputs;
+  }
+  get weights() {
+    return this.executor.weightMap;
+  }
+  get metadata() {
+    return this.artifacts.userDefinedMetadata;
+  }
+  get modelSignature() {
+    return this.signature;
+  }
+  get modelStructuredOutputKeys() {
+    return this.structuredOutputKeys;
+  }
+  findIOHandler() {
+    const path = this.modelUrl;
+    if (path.load != null) {
+      this.handler = path;
+    } else if (this.loadOptions.requestInit != null) {
+      this.handler = this.io.browserHTTPRequest(path, this.loadOptions);
+    } else {
+      const handlers = this.io.getLoadHandlers(path, this.loadOptions);
+      if (handlers.length === 0) {
+        handlers.push(this.io.browserHTTPRequest(path, this.loadOptions));
+      } else if (handlers.length > 1) {
+        throw new Error(`Found more than one (${handlers.length}) load handlers for URL '${[path]}'`);
+      }
+      this.handler = handlers[0];
+    }
+  }
+  load() {
+    this.findIOHandler();
+    if (this.handler.load == null) {
+      throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");
+    }
+    const loadResult = this.handler.load();
+    if (util_exports.isPromise(loadResult)) {
+      return loadResult.then((artifacts) => this.loadSync(artifacts));
+    }
+    return this.loadSync(loadResult);
+  }
+  loadSync(artifacts) {
+    this.artifacts = artifacts;
+    const graph = this.artifacts.modelTopology;
+    let signature = this.artifacts.signature;
+    if (this.artifacts.userDefinedMetadata != null) {
+      const metadata = this.artifacts.userDefinedMetadata;
+      if (metadata.signature != null) {
+        signature = metadata.signature;
+      }
+      if (metadata.structuredOutputKeys != null) {
+        this.structuredOutputKeys = metadata.structuredOutputKeys;
+      }
+    }
+    this.signature = signature;
+    this.version = `${graph.versions.producer}.${graph.versions.minConsumer}`;
+    const weightMap = this.io.decodeWeights(this.artifacts.weightData, this.artifacts.weightSpecs);
+    this.executor = new GraphExecutor(OperationMapper.Instance.transformGraph(graph, this.signature));
+    this.executor.weightMap = this.convertTensorMapToTensorsMap(weightMap);
+    this.executor.resourceManager = this.resourceManager;
+    if (artifacts.modelInitializer != null && artifacts.modelInitializer.node != null) {
+      const initializer = OperationMapper.Instance.transformGraph(artifacts.modelInitializer);
+      this.initializer = new GraphExecutor(initializer);
+      this.initializer.weightMap = this.executor.weightMap;
+      this.initializer.resourceManager = this.resourceManager;
+      this.initializerSignature = artifacts.initializerSignature;
+    }
+    return true;
+  }
+  async save(handlerOrURL, config) {
+    if (typeof handlerOrURL === "string") {
+      const handlers = this.io.getSaveHandlers(handlerOrURL);
+      if (handlers.length === 0) {
+        throw new Error(`Cannot find any save handlers for URL '${handlerOrURL}'`);
+      } else if (handlers.length > 1) {
+        throw new Error(`Found more than one (${handlers.length}) save handlers for URL '${handlerOrURL}'`);
+      }
+      handlerOrURL = handlers[0];
+    }
+    if (handlerOrURL.save == null) {
+      throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");
+    }
+    return handlerOrURL.save(this.artifacts);
+  }
+  predict(inputs, config) {
+    const outputTensors = this.execute(inputs, this.outputNodes);
+    if (this.structuredOutputKeys) {
+      const outputTensorsArray = outputTensors instanceof Tensor ? [outputTensors] : outputTensors;
+      const outputTensorMap = {};
+      outputTensorsArray.forEach((outputTensor, i) => outputTensorMap[this.structuredOutputKeys[i]] = outputTensor);
+      return outputTensorMap;
+    }
+    return outputTensors;
+  }
+  normalizeInputs(inputs) {
+    if (!(inputs instanceof Tensor) && !Array.isArray(inputs)) {
+      if (this.signature != null && this.signature.inputs != null) {
+        for (const input2 in this.signature.inputs) {
+          const tensor2 = this.signature.inputs[input2];
+          if (tensor2.resourceId != null) {
+            inputs[input2] = this.resourceIdToCapturedInput[tensor2.resourceId];
+          }
+        }
+      }
+      return inputs;
+    }
+    inputs = Array.isArray(inputs) ? inputs : [inputs];
+    const numCapturedInputs = Object.keys(this.resourceIdToCapturedInput).length;
+    if (inputs.length + numCapturedInputs !== this.inputNodes.length) {
+      throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length - numCapturedInputs} non-resource placeholders, while there are ${inputs.length} input tensors provided.`);
+    }
+    let inputIndex = 0;
+    return this.inputNodes.reduce((map, inputName) => {
+      const signature = this.signature ? this.signature.inputs[inputName] : null;
+      if (signature != null && signature.resourceId != null) {
+        map[inputName] = this.resourceIdToCapturedInput[signature.resourceId];
+      } else {
+        map[inputName] = inputs[inputIndex++];
+      }
+      return map;
+    }, {});
+  }
+  normalizeOutputs(outputs) {
+    outputs = outputs || this.outputNodes;
+    return !Array.isArray(outputs) ? [outputs] : outputs;
+  }
+  executeInitializerGraph() {
+    if (this.initializer == null) {
+      return [];
+    }
+    if (this.initializerSignature == null) {
+      return this.initializer.execute({}, []);
+    } else {
+      return this.initializer.execute({}, Object.keys(this.initializerSignature.outputs));
+    }
+  }
+  async executeInitializerGraphAsync() {
+    if (this.initializer == null) {
+      return [];
+    }
+    if (this.initializerSignature == null) {
+      return this.initializer.executeAsync({}, []);
+    } else {
+      return this.initializer.executeAsync({}, Object.keys(this.initializerSignature.outputs));
+    }
+  }
+  setResourceIdToCapturedInput(outputs) {
+    this.resourceIdToCapturedInput = {};
+    if (this.initializerSignature) {
+      const outputNames = Object.keys(this.initializerSignature.outputs);
+      for (let i = 0; i < outputNames.length; i++) {
+        const outputName = outputNames[i];
+        const tensorInfo = this.initializerSignature.outputs[outputName];
+        this.resourceIdToCapturedInput[tensorInfo.resourceId] = outputs[i];
+      }
+    }
+  }
+  execute(inputs, outputs) {
+    if (this.resourceIdToCapturedInput == null) {
+      this.setResourceIdToCapturedInput(this.executeInitializerGraph());
+    }
+    inputs = this.normalizeInputs(inputs);
+    outputs = this.normalizeOutputs(outputs);
+    const result = this.executor.execute(inputs, outputs);
+    return result.length > 1 ? result : result[0];
+  }
+  async executeAsync(inputs, outputs) {
+    if (this.resourceIdToCapturedInput == null) {
+      this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync());
+    }
+    inputs = this.normalizeInputs(inputs);
+    outputs = this.normalizeOutputs(outputs);
+    const result = await this.executor.executeAsync(inputs, outputs);
+    return result.length > 1 ? result : result[0];
+  }
+  getIntermediateTensors() {
+    return this.executor.getIntermediateTensors();
+  }
+  disposeIntermediateTensors() {
+    this.executor.disposeIntermediateTensors();
+  }
+  convertTensorMapToTensorsMap(map) {
+    return Object.keys(map).reduce((newMap, key) => {
+      newMap[key] = [map[key]];
+      return newMap;
+    }, {});
+  }
+  dispose() {
+    this.executor.dispose();
+    if (this.initializer) {
+      this.initializer.dispose();
+      if (this.resourceIdToCapturedInput) {
+        dispose(this.resourceIdToCapturedInput);
+      }
+    }
+    this.resourceManager.dispose();
+  }
+};
+async function loadGraphModel(modelUrl, options = {}, tfio = io_exports) {
+  if (modelUrl == null) {
+    throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");
+  }
+  if (options == null) {
+    options = {};
+  }
+  if (options.fromTFHub && typeof modelUrl === "string") {
+    modelUrl = getTFHubUrl(modelUrl);
+  }
+  const model2 = new GraphModel(modelUrl, options, tfio);
+  await model2.load();
+  return model2;
+}
+function loadGraphModelSync(modelSource) {
+  if (modelSource == null) {
+    throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");
+  }
+  let ioHandler;
+  if (modelSource instanceof Array) {
+    const [modelJSON, weights] = modelSource;
+    if (!modelJSON) {
+      throw new Error("modelJSON must be the first element of the array");
+    }
+    if (!weights || !(weights instanceof ArrayBuffer)) {
+      throw new Error("An ArrayBuffer of weights must be the second element of the array");
+    }
+    if (!("modelTopology" in modelJSON)) {
+      throw new Error("Model JSON is missing 'modelTopology'");
+    }
+    if (!("weightsManifest" in modelJSON)) {
+      throw new Error("Model JSON is missing 'weightsManifest'");
+    }
+    const weightSpecs = io_exports.getWeightSpecs(modelJSON.weightsManifest);
+    const modelArtifacts = io_exports.getModelArtifactsForJSONSync(modelJSON, weightSpecs, weights);
+    ioHandler = io_exports.fromMemorySync(modelArtifacts);
+  } else if ("load" in modelSource) {
+    ioHandler = modelSource;
+  } else if ("modelTopology" in modelSource && "weightSpecs" in modelSource && "weightData" in modelSource) {
+    ioHandler = io_exports.fromMemorySync(modelSource);
+  } else {
+    throw new Error("Unknown model format");
+  }
+  const model2 = new GraphModel(ioHandler);
+  model2.load();
+  return model2;
+}
+function getTFHubUrl(modelUrl) {
+  if (!modelUrl.endsWith("/")) {
+    modelUrl = modelUrl + "/";
+  }
+  return `${modelUrl}${DEFAULT_MODEL_NAME}${TFHUB_SEARCH_PARAM}`;
+}
+
+// node_modules/@tensorflow/tfjs-converter/dist/version.js
+var version3 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-data/dist/index.js
+var dist_exports2 = {};
+__export(dist_exports2, {
+  CSVDataset: () => CSVDataset,
+  Dataset: () => Dataset,
+  FileDataSource: () => FileDataSource,
+  TextLineDataset: () => TextLineDataset,
+  URLDataSource: () => URLDataSource,
+  array: () => array,
+  csv: () => csv,
+  func: () => func,
+  generator: () => generator,
+  microphone: () => microphone,
+  version_data: () => version4,
+  webcam: () => webcam,
+  zip: () => zip
+});
+
+// node_modules/@tensorflow/tfjs-data/dist/dataset.js
+var seedrandom3 = __toESM(require_seedrandom2());
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/lazy_iterator.js
+var seedrandom2 = __toESM(require_seedrandom2());
+
+// node_modules/@tensorflow/tfjs-data/dist/util/deep_map.js
+function deepMap(input2, mapFn) {
+  return deepMapInternal(input2, mapFn);
+}
+function deepMapInternal(input2, mapFn, seen = /* @__PURE__ */ new Map(), containedIn = /* @__PURE__ */ new Set()) {
+  if (input2 == null) {
+    return null;
+  }
+  if (typeof Blob === "function" && input2 instanceof Blob) {
+    return input2.slice();
+  }
+  if (containedIn.has(input2)) {
+    throw new Error("Circular references are not supported.");
+  }
+  if (seen.has(input2)) {
+    return seen.get(input2);
+  }
+  const result = mapFn(input2);
+  if (result.recurse && result.value !== null) {
+    throw new Error("A deep map function may not return both a value and recurse=true.");
+  }
+  if (!result.recurse) {
+    seen.set(input2, result.value);
+    return result.value;
+  } else if (isIterable2(input2)) {
+    const mappedIterable = Array.isArray(input2) ? [] : {};
+    containedIn.add(input2);
+    for (const k in input2) {
+      const child = input2[k];
+      const childResult = deepMapInternal(child, mapFn, seen, containedIn);
+      mappedIterable[k] = childResult;
+    }
+    containedIn.delete(input2);
+    if (input2.__proto__) {
+      mappedIterable.__proto__ = input2.__proto__;
+    }
+    return mappedIterable;
+  } else {
+    throw new Error(`Can't recurse into non-iterable type: ${input2}`);
+  }
+}
+function deepZip(inputs, zipFn = zipToList) {
+  return deepZipInternal(inputs, zipFn);
+}
+function deepZipInternal(inputs, zipFn, containedIn = /* @__PURE__ */ new Set()) {
+  const input2 = inputs[0];
+  if (containedIn.has(input2)) {
+    throw new Error("Circular references are not supported.");
+  }
+  const result = zipFn(inputs);
+  if (result.recurse && result.value !== null) {
+    throw new Error("A deep zip function may not return both a value and recurse=true.");
+  }
+  if (!result.recurse) {
+    return result.value;
+  } else if (isIterable2(input2)) {
+    const mappedIterable = Array.isArray(input2) ? [] : {};
+    containedIn.add(input2);
+    for (const k in input2) {
+      const children = inputs.map((x) => x[k]);
+      const childResult = deepZipInternal(children, zipFn, containedIn);
+      mappedIterable[k] = childResult;
+    }
+    containedIn.delete(input2);
+    return mappedIterable;
+  } else {
+    throw new Error(`Can't recurse into non-iterable type: ${input2}`);
+  }
+}
+function zipToList(x) {
+  if (x === null) {
+    return null;
+  }
+  if (isIterable2(x[0])) {
+    return { value: null, recurse: true };
+  } else {
+    return { value: x, recurse: false };
+  }
+}
+async function deepMapAndAwaitAll(input2, mapFn) {
+  const seen = /* @__PURE__ */ new Map();
+  deepMapInternal(input2, mapFn, seen);
+  for (const key of Array.from(seen.keys())) {
+    const value = seen.get(key);
+    if (util_exports.isPromise(value)) {
+      const mappedValue = await value;
+      seen.set(key, mappedValue);
+    }
+  }
+  const result = deepMapInternal(input2, mapFn, seen);
+  return result;
+}
+function isIterable2(obj) {
+  let isTextDecoder = false;
+  if (env().get("IS_BROWSER")) {
+    isTextDecoder = obj instanceof TextDecoder;
+  } else {
+    const { StringDecoder } = require_string_decoder();
+    isTextDecoder = obj instanceof StringDecoder;
+  }
+  return obj != null && !ArrayBuffer.isView(obj) && (Array.isArray(obj) || typeof obj === "object" && !(obj instanceof Tensor) && !(obj instanceof Promise) && !isTextDecoder);
+}
+function canTensorify(obj) {
+  return obj == null || isPrimitive(obj) || Array.isArray(obj) || typeof obj === "object" && obj instanceof Tensor || util_exports.isTypedArray(obj);
+}
+function isPrimitive(value) {
+  return value === null || typeof value !== "object" && typeof value !== "function";
+}
+
+// node_modules/@tensorflow/tfjs-data/dist/util/deep_clone.js
+function deepClone(container) {
+  return deepMap(container, cloneIfTensor);
+}
+function cloneIfTensor(item) {
+  if (item instanceof Tensor) {
+    return { value: item.clone(), recurse: false };
+  } else if (isIterable2(item)) {
+    return { value: null, recurse: true };
+  } else {
+    return { value: item, recurse: false };
+  }
+}
+
+// node_modules/@tensorflow/tfjs-data/dist/util/ring_buffer.js
+var RingBuffer = class {
+  constructor(capacity) {
+    this.capacity = capacity;
+    this.begin = 0;
+    this.end = 0;
+    if (capacity == null) {
+      throw new RangeError("Can't create a ring buffer of unknown capacity.");
+    }
+    if (capacity < 1) {
+      throw new RangeError("Can't create ring buffer of capacity < 1.");
+    }
+    this.data = new Array(capacity);
+    this.doubledCapacity = 2 * capacity;
+  }
+  wrap(index) {
+    while (index < 0) {
+      index += this.doubledCapacity;
+    }
+    return index % this.doubledCapacity;
+  }
+  get(index) {
+    if (index < 0) {
+      throw new RangeError("Can't get item at a negative index.");
+    }
+    return this.data[index % this.capacity];
+  }
+  set(index, value) {
+    if (index < 0) {
+      throw new RangeError("Can't set item at a negative index.");
+    }
+    this.data[index % this.capacity] = value;
+  }
+  length() {
+    let length = this.end - this.begin;
+    if (length < 0) {
+      length = this.doubledCapacity + length;
+    }
+    return length;
+  }
+  isFull() {
+    return this.length() === this.capacity;
+  }
+  isEmpty() {
+    return this.length() === 0;
+  }
+  push(value) {
+    if (this.isFull()) {
+      throw new RangeError("Ring buffer is full.");
+    }
+    this.set(this.end, value);
+    this.end = this.wrap(this.end + 1);
+  }
+  pushAll(values) {
+    for (const value of values) {
+      this.push(value);
+    }
+  }
+  pop() {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    this.end = this.wrap(this.end - 1);
+    const result = this.get(this.end);
+    this.set(this.end, void 0);
+    return result;
+  }
+  unshift(value) {
+    if (this.isFull()) {
+      throw new RangeError("Ring buffer is full.");
+    }
+    this.begin = this.wrap(this.begin - 1);
+    this.set(this.begin, value);
+  }
+  shift() {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    const result = this.get(this.begin);
+    this.set(this.begin, void 0);
+    this.begin = this.wrap(this.begin + 1);
+    return result;
+  }
+  shuffleExcise(relativeIndex) {
+    if (this.isEmpty()) {
+      throw new RangeError("Ring buffer is empty.");
+    }
+    const index = this.wrap(this.begin + relativeIndex);
+    const result = this.get(index);
+    this.set(index, this.pop());
+    return result;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/util/growing_ring_buffer.js
+var GrowingRingBuffer = class extends RingBuffer {
+  constructor() {
+    super(GrowingRingBuffer.INITIAL_CAPACITY);
+  }
+  isFull() {
+    return false;
+  }
+  push(value) {
+    if (super.isFull()) {
+      this.expand();
+    }
+    super.push(value);
+  }
+  unshift(value) {
+    if (super.isFull()) {
+      this.expand();
+    }
+    super.unshift(value);
+  }
+  expand() {
+    const newCapacity = this.capacity * 2;
+    const newData = new Array(newCapacity);
+    const len = this.length();
+    for (let i = 0; i < len; i++) {
+      newData[i] = this.get(this.wrap(this.begin + i));
+    }
+    this.data = newData;
+    this.capacity = newCapacity;
+    this.doubledCapacity = 2 * this.capacity;
+    this.begin = 0;
+    this.end = len;
+  }
+};
+GrowingRingBuffer.INITIAL_CAPACITY = 32;
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/lazy_iterator.js
+function iteratorFromItems(items) {
+  return new ArrayIterator(items);
+}
+function iteratorFromFunction(func2) {
+  return new FunctionCallIterator(func2);
+}
+function iteratorFromConcatenated(baseIterators, baseErrorHandler) {
+  return new ChainedIterator(baseIterators, baseErrorHandler);
+}
+function iteratorFromZipped(iterators, mismatchMode = ZipMismatchMode.FAIL) {
+  return new ZipIterator(iterators, mismatchMode);
+}
+var LazyIterator = class {
+  async toArray() {
+    const result = [];
+    let x = await this.next();
+    while (!x.done) {
+      result.push(x.value);
+      x = await this.next();
+    }
+    return result;
+  }
+  async toArrayForTest() {
+    const stream = this.prefetch(100);
+    const result = [];
+    let x = await stream.next();
+    while (!x.done) {
+      result.push(x.value);
+      x = await stream.next();
+    }
+    return result;
+  }
+  async resolveFully() {
+    let x = await this.next();
+    while (!x.done) {
+      x = await this.next();
+    }
+  }
+  async resolveWhile(predicate) {
+    let x = await this.next();
+    let shouldContinue = predicate(x.value);
+    while (!x.done && shouldContinue) {
+      x = await this.next();
+      shouldContinue = predicate(x.value);
+    }
+  }
+  handleErrors(handler) {
+    return new ErrorHandlingLazyIterator(this, handler);
+  }
+  filter(predicate) {
+    return new FilterIterator(this, predicate);
+  }
+  map(transform5) {
+    return new MapIterator(this, transform5);
+  }
+  mapAsync(transform5) {
+    return new AsyncMapIterator(this, transform5);
+  }
+  serialMapAsync(transform5) {
+    return new AsyncMapIterator(this, transform5).serial();
+  }
+  flatmap(transform5) {
+    return new FlatmapIterator(this, transform5);
+  }
+  async forEachAsync(f) {
+    return this.map(f).resolveFully();
+  }
+  async serialForEach(f) {
+    return this.serialMapAsync(f).resolveWhile((x) => x === true);
+  }
+  rowMajorBatch(batchSize, smallLastBatch = true) {
+    return new RowMajorBatchIterator(this, batchSize, smallLastBatch);
+  }
+  columnMajorBatch(batchSize, smallLastBatch = true, zipFn = zipToList) {
+    const rowBatches = this.rowMajorBatch(batchSize, smallLastBatch);
+    return rowBatches.map((x) => deepZip(x, zipFn));
+  }
+  concatenate(iterator, baseErrorHandler) {
+    return new ChainedIterator(iteratorFromItems([this, iterator]), baseErrorHandler);
+  }
+  take(count2) {
+    if (count2 < 0 || count2 == null) {
+      return this;
+    }
+    return new TakeIterator(this, count2);
+  }
+  skip(count2) {
+    if (count2 < 0 || count2 == null) {
+      return this;
+    }
+    return new SkipIterator(this, count2);
+  }
+  prefetch(bufferSize) {
+    return new PrefetchIterator(this, bufferSize);
+  }
+  shuffle(windowSize, seed) {
+    return new ShuffleIterator(this, windowSize, seed);
+  }
+  serial() {
+    return new SerialIterator(this);
+  }
+};
+var ArrayIterator = class extends LazyIterator {
+  constructor(items) {
+    super();
+    this.items = items;
+    this.trav = 0;
+  }
+  summary() {
+    return `Array of ${this.items.length} items`;
+  }
+  async next() {
+    if (this.trav >= this.items.length) {
+      return { value: null, done: true };
+    }
+    const item = this.items[this.trav];
+    this.trav++;
+    return { value: deepClone(item), done: false };
+  }
+};
+var FunctionCallIterator = class extends LazyIterator {
+  constructor(nextFn) {
+    super();
+    this.nextFn = nextFn;
+  }
+  summary() {
+    return `Function call`;
+  }
+  async next() {
+    try {
+      return this.nextFn();
+    } catch (e) {
+      e.message = `Error thrown while iterating through a dataset: ${e.message}`;
+      throw e;
+    }
+  }
+};
+var SerialIterator = class extends LazyIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Serial`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    return this.upstream.next();
+  }
+};
+var SkipIterator = class extends LazyIterator {
+  constructor(upstream, maxCount) {
+    super();
+    this.upstream = upstream;
+    this.maxCount = maxCount;
+    this.count = 0;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Skip`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (this.count++ < this.maxCount) {
+      const skipped = await this.upstream.next();
+      if (skipped.done) {
+        return skipped;
+      }
+      dispose(skipped.value);
+    }
+    return this.upstream.next();
+  }
+};
+var TakeIterator = class extends LazyIterator {
+  constructor(upstream, maxCount) {
+    super();
+    this.upstream = upstream;
+    this.maxCount = maxCount;
+    this.count = 0;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Take`;
+  }
+  async next() {
+    if (this.count++ >= this.maxCount) {
+      return { value: null, done: true };
+    }
+    return this.upstream.next();
+  }
+};
+var RowMajorBatchIterator = class extends LazyIterator {
+  constructor(upstream, batchSize, enableSmallLastBatch = true) {
+    super();
+    this.upstream = upstream;
+    this.batchSize = batchSize;
+    this.enableSmallLastBatch = enableSmallLastBatch;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> RowMajorBatch`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    const batch = [];
+    while (batch.length < this.batchSize) {
+      const item = await this.upstream.next();
+      if (item.done) {
+        if (this.enableSmallLastBatch && batch.length > 0) {
+          return { value: batch, done: false };
+        }
+        return { value: null, done: true };
+      }
+      batch.push(item.value);
+    }
+    return { value: batch, done: false };
+  }
+};
+var FilterIterator = class extends LazyIterator {
+  constructor(upstream, predicate) {
+    super();
+    this.upstream = upstream;
+    this.predicate = predicate;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Filter`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (true) {
+      const item = await this.upstream.next();
+      if (item.done || this.predicate(item.value)) {
+        return item;
+      }
+      dispose(item.value);
+    }
+  }
+};
+var MapIterator = class extends LazyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Map`;
+  }
+  async next() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return { value: null, done: true };
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mapped = this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return { value: mapped, done: false };
+  }
+};
+var ErrorHandlingLazyIterator = class extends LazyIterator {
+  constructor(upstream, handler) {
+    super();
+    this.upstream = upstream;
+    this.handler = handler;
+    this.count = 0;
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  summary() {
+    return `${this.upstream.summary()} -> handleErrors`;
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (true) {
+      try {
+        return await this.upstream.next();
+      } catch (e) {
+        if (!this.handler(e)) {
+          return { value: null, done: true };
+        }
+      }
+    }
+  }
+};
+var AsyncMapIterator = class extends LazyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> AsyncMap`;
+  }
+  async next() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return { value: null, done: true };
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mapped = await this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mapped);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return { value: mapped, done: false };
+  }
+};
+var OneToManyIterator = class extends LazyIterator {
+  constructor() {
+    super();
+    this.outputQueue = new GrowingRingBuffer();
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  async serialNext() {
+    while (this.outputQueue.length() === 0) {
+      if (!await this.pump()) {
+        return { value: null, done: true };
+      }
+    }
+    return { value: this.outputQueue.shift(), done: false };
+  }
+};
+var FlatmapIterator = class extends OneToManyIterator {
+  constructor(upstream, transform5) {
+    super();
+    this.upstream = upstream;
+    this.transform = transform5;
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Flatmap`;
+  }
+  async pump() {
+    const item = await this.upstream.next();
+    if (item.done) {
+      return false;
+    }
+    const inputTensors = tensor_util_exports.getTensorsInContainer(item.value);
+    const mappedArray = this.transform(item.value);
+    const outputTensors = tensor_util_exports.getTensorsInContainer(mappedArray);
+    this.outputQueue.pushAll(mappedArray);
+    for (const t of inputTensors) {
+      if (!tensor_util_exports.isTensorInList(t, outputTensors)) {
+        t.dispose();
+      }
+    }
+    return true;
+  }
+};
+var ChainedIterator = class extends LazyIterator {
+  constructor(iterators, baseErrorHandler) {
+    super();
+    this.baseErrorHandler = baseErrorHandler;
+    this.lastRead = null;
+    this.iterator = null;
+    this.moreIterators = iterators;
+  }
+  summary() {
+    const upstreamSummaries = "TODO: fill in upstream of chained summaries";
+    return `${upstreamSummaries} -> Chained`;
+  }
+  async next() {
+    this.lastRead = this.readFromChain(this.lastRead);
+    return this.lastRead;
+  }
+  async readFromChain(lastRead) {
+    await lastRead;
+    if (this.iterator == null) {
+      const iteratorResult = await this.moreIterators.next();
+      if (iteratorResult.done) {
+        return { value: null, done: true };
+      }
+      this.iterator = iteratorResult.value;
+      if (this.baseErrorHandler != null) {
+        this.iterator = this.iterator.handleErrors(this.baseErrorHandler);
+      }
+    }
+    const itemResult = await this.iterator.next();
+    if (itemResult.done) {
+      this.iterator = null;
+      return this.readFromChain(lastRead);
+    }
+    return itemResult;
+  }
+};
+var ZipMismatchMode;
+(function(ZipMismatchMode2) {
+  ZipMismatchMode2[ZipMismatchMode2["FAIL"] = 0] = "FAIL";
+  ZipMismatchMode2[ZipMismatchMode2["SHORTEST"] = 1] = "SHORTEST";
+  ZipMismatchMode2[ZipMismatchMode2["LONGEST"] = 2] = "LONGEST";
+})(ZipMismatchMode || (ZipMismatchMode = {}));
+var ZipIterator = class extends LazyIterator {
+  constructor(iterators, mismatchMode = ZipMismatchMode.FAIL) {
+    super();
+    this.iterators = iterators;
+    this.mismatchMode = mismatchMode;
+    this.count = 0;
+    this.currentPromise = null;
+  }
+  summary() {
+    const upstreamSummaries = "TODO: fill in upstream of zip summaries";
+    return `{${upstreamSummaries}} -> Zip`;
+  }
+  async nextState(afterState) {
+    await afterState;
+    let numIterators = 0;
+    let iteratorsDone = 0;
+    function getNext(container) {
+      if (container instanceof LazyIterator) {
+        const result = container.next();
+        return {
+          value: result.then((x) => {
+            numIterators++;
+            if (x.done) {
+              iteratorsDone++;
+            }
+            return x.value;
+          }),
+          recurse: false
+        };
+      } else {
+        return { value: null, recurse: true };
+      }
+    }
+    const mapped = await deepMapAndAwaitAll(this.iterators, getNext);
+    if (numIterators === iteratorsDone) {
+      return { value: null, done: true };
+    }
+    if (iteratorsDone > 0) {
+      switch (this.mismatchMode) {
+        case ZipMismatchMode.FAIL:
+          throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);
+        case ZipMismatchMode.SHORTEST:
+          return { value: null, done: true };
+        case ZipMismatchMode.LONGEST:
+        default:
+      }
+    }
+    this.count++;
+    return { value: mapped, done: false };
+  }
+  async next() {
+    this.currentPromise = this.nextState(this.currentPromise);
+    return this.currentPromise;
+  }
+};
+var PrefetchIterator = class extends LazyIterator {
+  constructor(upstream, bufferSize) {
+    super();
+    this.upstream = upstream;
+    this.bufferSize = bufferSize;
+    this.buffer = new RingBuffer(bufferSize);
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Prefetch`;
+  }
+  refill() {
+    while (!this.buffer.isFull()) {
+      const v = this.upstream.next();
+      this.buffer.push(v);
+    }
+  }
+  next() {
+    this.refill();
+    return this.buffer.shift();
+  }
+};
+var ShuffleIterator = class extends PrefetchIterator {
+  constructor(upstream, windowSize, seed) {
+    super(upstream, windowSize);
+    this.upstream = upstream;
+    this.windowSize = windowSize;
+    this.upstreamExhausted = false;
+    this.random = seedrandom2.alea(seed || util_exports.now().toString());
+    this.lastRead = Promise.resolve({ value: null, done: false });
+  }
+  async next() {
+    this.lastRead = this.lastRead.then(() => this.serialNext());
+    return this.lastRead;
+  }
+  randomInt(max6) {
+    return Math.floor(this.random() * max6);
+  }
+  chooseIndex() {
+    return this.randomInt(this.buffer.length());
+  }
+  async serialNext() {
+    if (!this.upstreamExhausted) {
+      this.refill();
+    }
+    while (!this.buffer.isEmpty()) {
+      const chosenIndex = this.chooseIndex();
+      const result = await this.buffer.shuffleExcise(chosenIndex);
+      if (result.done) {
+        this.upstreamExhausted = true;
+      } else {
+        this.refill();
+        return result;
+      }
+    }
+    return { value: null, done: true };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/dataset.js
+var Dataset = class {
+  constructor() {
+    this.size = null;
+  }
+  batch(batchSize, smallLastBatch = true) {
+    const base = this;
+    util_exports.assert(batchSize > 0, () => `batchSize needs to be positive, but it is
+      ${batchSize}`);
+    let size;
+    if (this.size === Infinity || this.size == null) {
+      size = this.size;
+    } else if (smallLastBatch) {
+      size = Math.ceil(this.size / batchSize);
+    } else {
+      size = Math.floor(this.size / batchSize);
+    }
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).columnMajorBatch(batchSize, smallLastBatch, deepBatchConcat);
+    }, size);
+  }
+  concatenate(dataset) {
+    const base = this;
+    let size;
+    if (this.size === Infinity || dataset.size === Infinity) {
+      size = Infinity;
+    } else if (this.size != null && dataset.size != null) {
+      size = this.size + dataset.size;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).concatenate(await dataset.iterator()), size);
+  }
+  filter(predicate) {
+    const base = this;
+    let size;
+    if (this.size === Infinity) {
+      size = Infinity;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).filter((x) => tidy(() => predicate(x)));
+    }, size);
+  }
+  async forEachAsync(f) {
+    return (await this.iterator()).forEachAsync(f);
+  }
+  map(transform5) {
+    const base = this;
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).map((x) => tidy(() => transform5(x)));
+    }, this.size);
+  }
+  mapAsync(transform5) {
+    const base = this;
+    return datasetFromIteratorFn(async () => {
+      return (await base.iterator()).mapAsync(transform5);
+    }, this.size);
+  }
+  prefetch(bufferSize) {
+    if (bufferSize == null) {
+      throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");
+    }
+    const base = this;
+    return datasetFromIteratorFn(async () => (await base.iterator()).prefetch(bufferSize), this.size);
+  }
+  repeat(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && count2 > 0) {
+      size = this.size * count2;
+    } else if (count2 === 0) {
+      size = 0;
+    } else if (this.size != null && (count2 === void 0 || count2 < 0)) {
+      size = Infinity;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => {
+      const iteratorIterator = iteratorFromFunction(async () => ({ value: await base.iterator(), done: false }));
+      return iteratorFromConcatenated(iteratorIterator.take(count2));
+    }, size);
+  }
+  skip(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && count2 >= 0 && this.size >= count2) {
+      size = this.size - count2;
+    } else if (this.size != null && (this.size < count2 || count2 === void 0 || count2 < 0)) {
+      size = 0;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).skip(count2), size);
+  }
+  shuffle(bufferSize, seed, reshuffleEachIteration = true) {
+    if (bufferSize == null || bufferSize < 0) {
+      if (this.size == null) {
+        throw new RangeError("`Dataset.shuffle()` requires bufferSize to be specified.");
+      } else {
+        throw new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);
+      }
+    }
+    const base = this;
+    const random = seedrandom3.alea(seed || util_exports.now().toString());
+    return datasetFromIteratorFn(async () => {
+      let seed2 = random.int32();
+      if (reshuffleEachIteration) {
+        seed2 += random.int32();
+      }
+      return (await base.iterator()).shuffle(bufferSize, seed2.toString());
+    }, this.size);
+  }
+  take(count2) {
+    const base = this;
+    let size;
+    if (this.size != null && this.size > count2) {
+      size = count2;
+    } else if (this.size != null && this.size <= count2) {
+      size = this.size;
+    } else {
+      size = null;
+    }
+    return datasetFromIteratorFn(async () => (await base.iterator()).take(count2), size);
+  }
+  async toArray() {
+    if (this.size === Infinity) {
+      throw new Error("Can not convert infinite data stream to array.");
+    }
+    return (await this.iterator()).toArray();
+  }
+  async toArrayForTest() {
+    if (this.size === Infinity) {
+      throw new Error("Can not convert infinite data stream to array.");
+    }
+    return (await this.iterator()).toArrayForTest();
+  }
+};
+Dataset.MAX_BUFFER_SIZE = 1e4;
+function datasetFromIteratorFn(iteratorFn, size = null) {
+  return new class extends Dataset {
+    constructor() {
+      super(...arguments);
+      this.size = size;
+    }
+    async iterator() {
+      return iteratorFn();
+    }
+  }();
+}
+function array(items) {
+  return datasetFromIteratorFn(async () => iteratorFromItems(items), items.length);
+}
+function zip(datasets) {
+  if (!isIterable2(datasets)) {
+    throw new Error("The argument to zip() must be an object or array.");
+  }
+  let size;
+  if (Array.isArray(datasets)) {
+    for (let i = 0; i < datasets.length; i++) {
+      size = size == null ? datasets[i].size : Math.min(size, datasets[i].size);
+    }
+  } else if (datasets instanceof Object) {
+    for (const ds in datasets) {
+      size = size == null ? datasets[ds].size : Math.min(size, datasets[ds].size);
+    }
+  }
+  return datasetFromIteratorFn(async () => {
+    const streams = await deepMapAndAwaitAll(datasets, (d) => {
+      if (d instanceof Dataset) {
+        return { value: d.iterator(), recurse: false };
+      } else if (isIterable2(d)) {
+        return { value: null, recurse: true };
+      } else {
+        throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.");
+      }
+    });
+    return iteratorFromZipped(streams, ZipMismatchMode.SHORTEST);
+  }, size);
+}
+function deepBatchConcat(rows) {
+  if (rows === null) {
+    return null;
+  }
+  const exampleRow = rows[0];
+  if (canTensorify(exampleRow)) {
+    const value = batchConcat(rows);
+    return { value, recurse: false };
+  }
+  return { value: null, recurse: true };
+}
+function batchConcat(arrays) {
+  if (arrays.length === 0) {
+    throw new Error("Can't make a batch of zero elements.");
+  }
+  if (arrays[0] instanceof Tensor) {
+    return stack(arrays);
+  } else {
+    return tensor(arrays);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-data/dist/datasets/text_line_dataset.js
+var TextLineDataset = class extends Dataset {
+  constructor(input2) {
+    super();
+    this.input = input2;
+  }
+  async iterator() {
+    const inputIterator = await this.input.iterator();
+    const utf8Iterator = inputIterator.decodeUTF8();
+    const lineIterator = utf8Iterator.split("\n").map((line) => {
+      if (line.endsWith("\r")) {
+        line = line.slice(0, -1);
+      }
+      return line;
+    });
+    return lineIterator;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/datasets/csv_dataset.js
+var CODE_QUOTE = '"';
+var STATE_OUT = Symbol("out");
+var STATE_FIELD = Symbol("field");
+var STATE_QUOTE = Symbol("quote");
+var STATE_QUOTE_AFTER_QUOTE = Symbol("quoteafterquote");
+var STATE_WITHIN_QUOTE_IN_QUOTE = Symbol("quoteinquote");
+var CSVDataset = class extends Dataset {
+  constructor(input2, csvConfig) {
+    super();
+    this.input = input2;
+    this.hasHeader = true;
+    this.fullColumnNames = null;
+    this.columnNamesValidated = false;
+    this.columnConfigs = null;
+    this.configuredColumnsOnly = false;
+    this.delimiter = ",";
+    this.delimWhitespace = false;
+    this.base = new TextLineDataset(input2);
+    if (!csvConfig) {
+      csvConfig = {};
+    }
+    this.hasHeader = csvConfig.hasHeader === false ? false : true;
+    this.fullColumnNames = csvConfig.columnNames;
+    this.columnConfigs = csvConfig.columnConfigs;
+    this.configuredColumnsOnly = csvConfig.configuredColumnsOnly;
+    if (csvConfig.delimWhitespace) {
+      util_exports.assert(csvConfig.delimiter == null, () => "Delimiter should not be provided when delimWhitespace is true.");
+      this.delimWhitespace = true;
+      this.delimiter = " ";
+    } else {
+      this.delimiter = csvConfig.delimiter ? csvConfig.delimiter : ",";
+    }
+  }
+  async columnNames() {
+    if (!this.columnNamesValidated) {
+      await this.setColumnNames();
+    }
+    return this.configuredColumnsOnly ? Object.keys(this.columnConfigs) : this.fullColumnNames;
+  }
+  async setColumnNames() {
+    const columnNamesFromFile = await this.maybeReadHeaderLine();
+    if (!this.fullColumnNames && !columnNamesFromFile) {
+      throw new Error("Column names must be provided if there is no header line.");
+    } else if (this.fullColumnNames && columnNamesFromFile) {
+      util_exports.assert(columnNamesFromFile.length === this.fullColumnNames.length, () => "The length of provided columnNames (" + this.fullColumnNames.length.toString() + ") does not match the length of the header line read from file (" + columnNamesFromFile.length.toString() + ").");
+    }
+    if (!this.fullColumnNames) {
+      this.fullColumnNames = columnNamesFromFile;
+    }
+    const counts = this.fullColumnNames.reduce((countAcc, name) => {
+      countAcc[name] = countAcc[name] + 1 || 1;
+      return countAcc;
+    }, {});
+    const duplicateNames = Object.keys(counts).filter((name) => counts[name] > 1);
+    util_exports.assert(duplicateNames.length === 0, () => "Duplicate column names found: " + duplicateNames.toString());
+    if (this.columnConfigs) {
+      for (const key of Object.keys(this.columnConfigs)) {
+        const index = this.fullColumnNames.indexOf(key);
+        if (index === -1) {
+          throw new Error('The key "' + key + '" provided in columnConfigs does not match any of the column names (' + this.fullColumnNames.toString() + ").");
+        }
+      }
+    }
+    this.columnNamesValidated = true;
+  }
+  async maybeReadHeaderLine() {
+    if (this.hasHeader) {
+      const iter = await this.base.iterator();
+      const firstElement = await iter.next();
+      if (firstElement.done) {
+        throw new Error("No data was found for CSV parsing.");
+      }
+      const firstLine = firstElement.value;
+      const headers = this.parseRow(firstLine, false);
+      return headers;
+    } else {
+      return null;
+    }
+  }
+  async iterator() {
+    if (!this.columnNamesValidated) {
+      await this.setColumnNames();
+    }
+    let lines = await this.base.iterator();
+    if (this.hasHeader) {
+      lines = lines.skip(1);
+    }
+    return lines.map((x) => this.makeDataElement(x));
+  }
+  makeDataElement(line) {
+    const values = this.parseRow(line);
+    const features = {};
+    const labels = {};
+    for (let i = 0; i < this.fullColumnNames.length; i++) {
+      const key = this.fullColumnNames[i];
+      const config = this.columnConfigs ? this.columnConfigs[key] : null;
+      if (this.configuredColumnsOnly && !config) {
+        continue;
+      } else {
+        const value = values[i];
+        let parsedValue = null;
+        if (value === "") {
+          if (config && config.default !== void 0) {
+            parsedValue = config.default;
+          } else if (config && (config.required || config.isLabel)) {
+            throw new Error(`Required column ${key} is empty in this line: ${line}`);
+          } else {
+            parsedValue = void 0;
+          }
+        } else {
+          const valueAsNum = Number(value);
+          if (isNaN(valueAsNum)) {
+            if (config && config.dtype === "bool") {
+              parsedValue = this.getBoolean(value);
+            } else {
+              parsedValue = value;
+            }
+          } else if (!config || !config.dtype) {
+            parsedValue = valueAsNum;
+          } else {
+            switch (config.dtype) {
+              case "float32":
+                parsedValue = valueAsNum;
+                break;
+              case "int32":
+                parsedValue = Math.floor(valueAsNum);
+                break;
+              case "bool":
+                parsedValue = this.getBoolean(value);
+                break;
+              default:
+                parsedValue = valueAsNum;
+            }
+          }
+        }
+        config && config.isLabel ? labels[key] = parsedValue : features[key] = parsedValue;
+      }
+    }
+    if (Object.keys(labels).length === 0) {
+      return features;
+    } else {
+      return { xs: features, ys: labels };
+    }
+  }
+  getBoolean(value) {
+    if (value === "1" || value.toLowerCase() === "true") {
+      return 1;
+    } else {
+      return 0;
+    }
+  }
+  parseRow(line, validateElementCount = true) {
+    const result = [];
+    let readOffset = 0;
+    const readLength = line.length;
+    let currentState = STATE_OUT;
+    for (let i = 0; i < readLength; i++) {
+      switch (currentState) {
+        case STATE_OUT:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              readOffset = i + 1;
+              currentState = STATE_QUOTE;
+              break;
+            case this.delimiter:
+              readOffset = i + 1;
+              if (this.delimiter === " " && this.delimWhitespace) {
+                break;
+              }
+              result.push("");
+              currentState = STATE_OUT;
+              break;
+            default:
+              currentState = STATE_FIELD;
+              readOffset = i;
+              break;
+          }
+          break;
+        case STATE_FIELD:
+          switch (line.charAt(i)) {
+            case this.delimiter:
+              result.push(line.substring(readOffset, i));
+              currentState = STATE_OUT;
+              readOffset = i + 1;
+              break;
+            default:
+          }
+          break;
+        case STATE_QUOTE:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE_AFTER_QUOTE;
+              break;
+            default:
+          }
+          break;
+        case STATE_QUOTE_AFTER_QUOTE:
+          switch (line.charAt(i)) {
+            case this.delimiter:
+              result.push(line.substring(readOffset, i - 1));
+              currentState = STATE_OUT;
+              readOffset = i + 1;
+              break;
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE;
+              break;
+            default:
+              currentState = STATE_WITHIN_QUOTE_IN_QUOTE;
+              break;
+          }
+          break;
+        case STATE_WITHIN_QUOTE_IN_QUOTE:
+          switch (line.charAt(i)) {
+            case CODE_QUOTE:
+              currentState = STATE_QUOTE;
+              break;
+            default:
+          }
+          break;
+        default:
+      }
+    }
+    if (currentState === STATE_QUOTE_AFTER_QUOTE) {
+      result.push(line.substring(readOffset, readLength - 1));
+    } else {
+      result.push(line.substring(readOffset));
+    }
+    if (validateElementCount && result.length !== this.fullColumnNames.length) {
+      throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${result}`);
+    }
+    return result;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/microphone_iterator.js
+var MicrophoneIterator = class extends LazyIterator {
+  constructor(microphoneConfig) {
+    super();
+    this.microphoneConfig = microphoneConfig;
+    this.isClosed = false;
+    this.fftSize = microphoneConfig.fftSize || 1024;
+    const fftSizeLog2 = Math.log2(this.fftSize);
+    if (this.fftSize < 0 || fftSizeLog2 < 4 || fftSizeLog2 > 14 || !Number.isInteger(fftSizeLog2)) {
+      throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);
+    }
+    this.numFrames = microphoneConfig.numFramesPerSpectrogram || 43;
+    this.sampleRateHz = microphoneConfig.sampleRateHz;
+    this.columnTruncateLength = microphoneConfig.columnTruncateLength || this.fftSize;
+    this.audioTrackConstraints = microphoneConfig.audioTrackConstraints;
+    this.smoothingTimeConstant = microphoneConfig.smoothingTimeConstant || 0;
+    this.includeSpectrogram = microphoneConfig.includeSpectrogram === false ? false : true;
+    this.includeWaveform = microphoneConfig.includeWaveform === true ? true : false;
+    if (!this.includeSpectrogram && !this.includeWaveform) {
+      throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.");
+    }
+  }
+  summary() {
+    return `microphone`;
+  }
+  static async create(microphoneConfig = {}) {
+    if (!env().get("IS_BROWSER")) {
+      throw new Error("microphone API is only supported in browser environment.");
+    }
+    const microphoneIterator = new MicrophoneIterator(microphoneConfig);
+    await microphoneIterator.start();
+    return microphoneIterator;
+  }
+  async start() {
+    try {
+      this.stream = await navigator.mediaDevices.getUserMedia({
+        audio: this.audioTrackConstraints == null ? true : this.audioTrackConstraints,
+        video: false
+      });
+    } catch (e) {
+      throw new Error(`Error thrown while initializing video stream: ${e.message}`);
+    }
+    if (!this.stream) {
+      throw new Error("Could not obtain audio from microphone.");
+    }
+    const ctxConstructor = window.AudioContext || window.webkitAudioContext;
+    this.audioContext = new ctxConstructor();
+    if (!this.sampleRateHz) {
+      this.sampleRateHz = this.audioContext.sampleRate;
+    } else if (this.audioContext.sampleRate !== this.sampleRateHz) {
+      throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);
+    }
+    const streamSource = this.audioContext.createMediaStreamSource(this.stream);
+    this.analyser = this.audioContext.createAnalyser();
+    this.analyser.fftSize = this.fftSize * 2;
+    this.analyser.smoothingTimeConstant = this.smoothingTimeConstant;
+    streamSource.connect(this.analyser);
+    this.freqData = new Float32Array(this.fftSize);
+    this.timeData = new Float32Array(this.fftSize);
+    return;
+  }
+  async next() {
+    if (this.isClosed) {
+      return { value: null, done: true };
+    }
+    let spectrogramTensor;
+    let waveformTensor;
+    const audioDataQueue = await this.getAudioData();
+    if (this.includeSpectrogram) {
+      const freqData = this.flattenQueue(audioDataQueue.freqDataQueue);
+      spectrogramTensor = this.getTensorFromAudioDataArray(freqData, [this.numFrames, this.columnTruncateLength, 1]);
+    }
+    if (this.includeWaveform) {
+      const timeData = this.flattenQueue(audioDataQueue.timeDataQueue);
+      waveformTensor = this.getTensorFromAudioDataArray(timeData, [this.numFrames * this.fftSize, 1]);
+    }
+    return {
+      value: { "spectrogram": spectrogramTensor, "waveform": waveformTensor },
+      done: false
+    };
+  }
+  async capture() {
+    return (await this.next()).value;
+  }
+  async getAudioData() {
+    const freqDataQueue = [];
+    const timeDataQueue = [];
+    let currentFrames = 0;
+    return new Promise((resolve) => {
+      const intervalID = setInterval(() => {
+        if (this.includeSpectrogram) {
+          this.analyser.getFloatFrequencyData(this.freqData);
+          if (this.freqData[0] === -Infinity) {
+            resolve({ freqDataQueue, timeDataQueue });
+          }
+          freqDataQueue.push(this.freqData.slice(0, this.columnTruncateLength));
+        }
+        if (this.includeWaveform) {
+          this.analyser.getFloatTimeDomainData(this.timeData);
+          timeDataQueue.push(this.timeData.slice());
+        }
+        if (++currentFrames === this.numFrames) {
+          clearInterval(intervalID);
+          resolve({ freqDataQueue, timeDataQueue });
+        }
+      }, this.fftSize / this.sampleRateHz * 1e3);
+    });
+  }
+  stop() {
+    if (!this.isClosed) {
+      this.isClosed = true;
+      this.analyser.disconnect();
+      this.audioContext.close();
+      if (this.stream != null && this.stream.getTracks().length > 0) {
+        this.stream.getTracks()[0].stop();
+      }
+    }
+  }
+  toArray() {
+    throw new Error("Can not convert infinite audio stream to array.");
+  }
+  getSampleRate() {
+    return this.sampleRateHz;
+  }
+  flattenQueue(queue) {
+    const frameSize = queue[0].length;
+    const freqData = new Float32Array(queue.length * frameSize);
+    queue.forEach((data, i) => freqData.set(data, i * frameSize));
+    return freqData;
+  }
+  getTensorFromAudioDataArray(freqData, shape) {
+    const vals = new Float32Array(util_exports.sizeFromShape(shape));
+    vals.set(freqData, vals.length - freqData.length);
+    return tensor(vals, shape);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/webcam_iterator.js
+var WebcamIterator = class extends LazyIterator {
+  constructor(webcamVideoElement, webcamConfig) {
+    super();
+    this.webcamVideoElement = webcamVideoElement;
+    this.webcamConfig = webcamConfig;
+    this.isClosed = true;
+    this.resize = false;
+    if (this.needToResize()) {
+      this.resize = true;
+      this.cropSize = [this.webcamConfig.resizeHeight, this.webcamConfig.resizeWidth];
+      this.cropBoxInd = tensor1d([0], "int32");
+      if (this.webcamConfig.centerCrop) {
+        const widthCroppingRatio = this.webcamConfig.resizeWidth * 1 / this.webcamVideoElement.width;
+        const heightCroppingRatio = this.webcamConfig.resizeHeight * 1 / this.webcamVideoElement.height;
+        const widthCropStart = (1 - widthCroppingRatio) / 2;
+        const heightCropStart = (1 - heightCroppingRatio) / 2;
+        const widthCropEnd = widthCropStart + widthCroppingRatio;
+        const heightCropEnd = heightCroppingRatio + heightCropStart;
+        this.cropBox = tensor2d([heightCropStart, widthCropStart, heightCropEnd, widthCropEnd], [1, 4]);
+      } else {
+        this.cropBox = tensor2d([0, 0, 1, 1], [1, 4]);
+      }
+    }
+  }
+  summary() {
+    return `webcam`;
+  }
+  static async create(webcamVideoElement, webcamConfig = {}) {
+    if (!env().get("IS_BROWSER")) {
+      throw new Error("tf.data.webcam is only supported in browser environment.");
+    }
+    if (!webcamVideoElement) {
+      webcamVideoElement = document.createElement("video");
+      if (!webcamConfig.resizeWidth || !webcamConfig.resizeHeight) {
+        throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");
+      }
+      webcamVideoElement.width = webcamConfig.resizeWidth;
+      webcamVideoElement.height = webcamConfig.resizeHeight;
+    }
+    const webcamIterator = new WebcamIterator(webcamVideoElement, webcamConfig);
+    await webcamIterator.start();
+    return webcamIterator;
+  }
+  async start() {
+    if (this.webcamConfig.facingMode) {
+      util_exports.assert(this.webcamConfig.facingMode === "user" || this.webcamConfig.facingMode === "environment", () => `Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);
+    }
+    try {
+      this.stream = await navigator.mediaDevices.getUserMedia({
+        video: {
+          deviceId: this.webcamConfig.deviceId,
+          facingMode: this.webcamConfig.facingMode ? this.webcamConfig.facingMode : "user",
+          width: this.webcamVideoElement.width,
+          height: this.webcamVideoElement.height
+        }
+      });
+    } catch (e) {
+      e.message = `Error thrown while initializing video stream: ${e.message}`;
+      throw e;
+    }
+    if (!this.stream) {
+      throw new Error("Could not obtain video from webcam.");
+    }
+    try {
+      this.webcamVideoElement.srcObject = this.stream;
+    } catch (error) {
+      console.log(error);
+      this.webcamVideoElement.src = window.URL.createObjectURL(this.stream);
+    }
+    this.webcamVideoElement.play();
+    this.isClosed = false;
+    return new Promise((resolve) => {
+      this.webcamVideoElement.onloadedmetadata = () => {
+        resolve();
+      };
+    });
+  }
+  async next() {
+    if (this.isClosed) {
+      return { value: null, done: true };
+    }
+    let img;
+    try {
+      img = browser_exports.fromPixels(this.webcamVideoElement);
+    } catch (e) {
+      throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`);
+    }
+    if (this.resize) {
+      try {
+        return { value: this.cropAndResizeFrame(img), done: false };
+      } catch (e) {
+        throw new Error(`Error thrown cropping the video: ${e.message}`);
+      } finally {
+        img.dispose();
+      }
+    } else {
+      return { value: img, done: false };
+    }
+  }
+  needToResize() {
+    if (this.webcamConfig.resizeWidth && this.webcamConfig.resizeHeight && (this.webcamVideoElement.width !== this.webcamConfig.resizeWidth || this.webcamVideoElement.height !== this.webcamConfig.resizeHeight)) {
+      return true;
+    }
+    return false;
+  }
+  cropAndResizeFrame(img) {
+    return tidy(() => {
+      const expandedImage = expandDims(cast(img, "float32"), 0);
+      let resizedImage;
+      resizedImage = image.cropAndResize(expandedImage, this.cropBox, this.cropBoxInd, this.cropSize, "bilinear");
+      const shape = resizedImage.shape;
+      return reshape(resizedImage, shape.slice(1));
+    });
+  }
+  async capture() {
+    return (await this.next()).value;
+  }
+  stop() {
+    const tracks = this.stream.getTracks();
+    tracks.forEach((track) => track.stop());
+    try {
+      this.webcamVideoElement.srcObject = null;
+    } catch (error) {
+      console.log(error);
+      this.webcamVideoElement.src = null;
+    }
+    this.isClosed = true;
+  }
+  toArray() {
+    throw new Error("Can not convert infinite video stream to array.");
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/datasource.js
+var DataSource = class {
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/string_iterator.js
+var StringIterator = class extends LazyIterator {
+  split(separator) {
+    return new SplitIterator(this, separator);
+  }
+};
+var SplitIterator = class extends StringIterator {
+  constructor(upstream, separator) {
+    super();
+    this.upstream = upstream;
+    this.impl = new SplitIteratorImpl(upstream, separator);
+  }
+  summary() {
+    return this.impl.summary();
+  }
+  async next() {
+    return this.impl.next();
+  }
+};
+var SplitIteratorImpl = class extends OneToManyIterator {
+  constructor(upstream, separator) {
+    super();
+    this.upstream = upstream;
+    this.separator = separator;
+    this.carryover = "";
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Split('${this.separator}')`;
+  }
+  async pump() {
+    const chunkResult = await this.upstream.next();
+    if (chunkResult.done) {
+      if (this.carryover === "") {
+        return false;
+      }
+      this.outputQueue.push(this.carryover);
+      this.carryover = "";
+      return true;
+    }
+    const lines = chunkResult.value.split(this.separator);
+    lines[0] = this.carryover + lines[0];
+    for (const line of lines.slice(0, -1)) {
+      this.outputQueue.push(line);
+    }
+    this.carryover = lines[lines.length - 1];
+    return true;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/byte_chunk_iterator.js
+var ByteChunkIterator = class extends LazyIterator {
+  decodeUTF8() {
+    return new Utf8Iterator(this);
+  }
+};
+var Utf8Iterator = class extends StringIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    this.impl = new Utf8IteratorImpl(upstream);
+  }
+  summary() {
+    return this.impl.summary();
+  }
+  async next() {
+    return this.impl.next();
+  }
+};
+var Utf8IteratorImpl = class extends OneToManyIterator {
+  constructor(upstream) {
+    super();
+    this.upstream = upstream;
+    if (env().get("IS_BROWSER")) {
+      this.decoder = new TextDecoder("utf-8");
+    } else {
+      const { StringDecoder } = require_string_decoder();
+      this.decoder = new StringDecoder("utf8");
+    }
+  }
+  summary() {
+    return `${this.upstream.summary()} -> Utf8`;
+  }
+  async pump() {
+    const chunkResult = await this.upstream.next();
+    let chunk;
+    if (chunkResult.done) {
+      return false;
+    } else {
+      chunk = chunkResult.value;
+    }
+    let text;
+    if (env().get("IS_BROWSER")) {
+      text = this.decoder.decode(chunk, { stream: true });
+    } else {
+      text = this.decoder.write(Buffer.from(chunk.buffer));
+    }
+    this.outputQueue.push(text);
+    return true;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/file_chunk_iterator.js
+var FileChunkIterator = class extends ByteChunkIterator {
+  constructor(file, options = {}) {
+    super();
+    this.file = file;
+    this.options = options;
+    util_exports.assert(file instanceof Uint8Array || (env().get("IS_BROWSER") ? file instanceof File || file instanceof Blob : false), () => "FileChunkIterator only supports File, Blob and Uint8Array right now.");
+    this.offset = options.offset || 0;
+    this.chunkSize = options.chunkSize || 1024 * 1024;
+  }
+  summary() {
+    return `FileChunks ${this.file}`;
+  }
+  async next() {
+    if (this.offset >= (this.file instanceof Uint8Array ? this.file.byteLength : this.file.size)) {
+      return { value: null, done: true };
+    }
+    const chunk = new Promise((resolve, reject) => {
+      const end = this.offset + this.chunkSize;
+      if (this.file instanceof Uint8Array) {
+        resolve(new Uint8Array(this.file.slice(this.offset, end)));
+      } else {
+        const fileReader = new FileReader();
+        fileReader.onload = (event) => {
+          let data = fileReader.result;
+          if (data instanceof ArrayBuffer) {
+            data = new Uint8Array(data);
+          }
+          if (!(data instanceof Uint8Array)) {
+            return reject(new TypeError("FileReader returned unknown type."));
+          }
+          resolve(data);
+        };
+        fileReader.onabort = (event) => {
+          return reject(new Error("Aborted"));
+        };
+        fileReader.onerror = (event) => {
+          return reject(new Error(event.type));
+        };
+        const slice5 = this.file.slice(this.offset, end);
+        fileReader.readAsArrayBuffer(slice5);
+      }
+      this.offset = end;
+    });
+    return { value: await chunk, done: false };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/iterators/url_chunk_iterator.js
+async function urlChunkIterator(url, options = {}, fetchFunc) {
+  let urlString;
+  let requestInit;
+  if (typeof url === "string") {
+    urlString = url;
+  } else {
+    urlString = url.url;
+    requestInit = getRequestInitFromRequest(url);
+  }
+  const response = await (fetchFunc || util_exports.fetch)(urlString, requestInit);
+  if (response.ok) {
+    const uint8Array = new Uint8Array(await response.arrayBuffer());
+    return new FileChunkIterator(uint8Array, options);
+  } else {
+    throw new Error(response.statusText);
+  }
+}
+var getRequestInitFromRequest = (request) => {
+  const init2 = {
+    method: request.method,
+    headers: request.headers,
+    body: request.body,
+    mode: request.mode,
+    credentials: request.credentials,
+    cache: request.cache,
+    redirect: request.redirect,
+    referrer: request.referrer,
+    integrity: request.integrity
+  };
+  return init2;
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/util/source_util.js
+function isLocalPath(source) {
+  return typeof source === "string" && source.slice(0, 7) === "file://";
+}
+
+// node_modules/@tensorflow/tfjs-data/dist/sources/file_data_source.js
+var FileDataSource = class extends DataSource {
+  constructor(input2, options = {}) {
+    super();
+    this.input = input2;
+    this.options = options;
+  }
+  async iterator() {
+    if (isLocalPath(this.input) && env().get("IS_NODE")) {
+      const fs = require_fs();
+      this.input = fs.readFileSync(this.input.slice(7));
+    }
+    return new FileChunkIterator(this.input, this.options);
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/sources/url_data_source.js
+var URLDataSource = class extends DataSource {
+  constructor(url, fileOptions = {}) {
+    super();
+    this.url = url;
+    this.fileOptions = fileOptions;
+  }
+  async iterator() {
+    if (isLocalPath(this.url)) {
+      return new FileDataSource(this.url, this.fileOptions).iterator();
+    } else {
+      return urlChunkIterator(this.url, this.fileOptions);
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-data/dist/readers.js
+function csv(source, csvConfig = {}) {
+  return new CSVDataset(new URLDataSource(source), csvConfig);
+}
+function func(f) {
+  const iter = iteratorFromFunction(f);
+  return datasetFromIteratorFn(async () => iter);
+}
+function generator(generator2) {
+  return datasetFromIteratorFn(async () => {
+    const gen = await generator2();
+    return iteratorFromFunction(() => gen.next());
+  });
+}
+async function webcam(webcamVideoElement, webcamConfig) {
+  return WebcamIterator.create(webcamVideoElement, webcamConfig);
+}
+async function microphone(microphoneConfig) {
+  return MicrophoneIterator.create(microphoneConfig);
+}
+
+// node_modules/@tensorflow/tfjs-data/dist/version.js
+var version4 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/cpu_util.js
+function assertNotComplex(tensor2, opName) {
+  if (!Array.isArray(tensor2)) {
+    tensor2 = [tensor2];
+  }
+  tensor2.forEach((t) => {
+    if (t != null) {
+      util_exports.assert(t.dtype !== "complex64", () => `${opName} does not support complex64 tensors in the CPU backend.`);
+    }
+  });
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/backend_cpu.js
+var whereImpl2 = kernel_impls_exports.whereImpl;
+var MathBackendCPU = class extends KernelBackend {
+  constructor() {
+    super();
+    this.blockSize = 48;
+    this.firstUse = true;
+    this.data = new DataStorage(this, engine());
+  }
+  nextDataId() {
+    return MathBackendCPU.nextDataId++;
+  }
+  write(values, shape, dtype) {
+    if (this.firstUse) {
+      this.firstUse = false;
+      if (env().get("IS_NODE")) {
+        backend_util_exports.warn("\n============================\nHi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. \n============================");
+      }
+    }
+    const dataId = { id: this.nextDataId() };
+    this.data.set(dataId, { values, dtype, refCount: 1 });
+    return dataId;
+  }
+  makeTensorInfo(shape, dtype, values) {
+    let outId;
+    if (dtype === "string" && values != null && values.length > 0 && util_exports.isString(values[0])) {
+      const encodedValues = values.map((d) => util_exports.encodeString(d));
+      outId = this.write(encodedValues, shape, dtype);
+    } else {
+      outId = this.write(values, shape, dtype);
+    }
+    return { dataId: outId, shape, dtype };
+  }
+  refCount(dataId) {
+    if (this.data.has(dataId)) {
+      const tensorData = this.data.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const tensorData = this.data.get(dataId);
+    tensorData.refCount++;
+  }
+  decRef(dataId) {
+    if (this.data.has(dataId)) {
+      const tensorData = this.data.get(dataId);
+      tensorData.refCount--;
+    }
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    this.data.set(dataId, { values, dtype, refCount });
+  }
+  numDataIds() {
+    return this.data.numDataIds();
+  }
+  async read(dataId) {
+    return this.readSync(dataId);
+  }
+  readSync(dataId) {
+    const { dtype, complexTensorInfos } = this.data.get(dataId);
+    if (dtype === "complex64") {
+      const realValues = this.readSync(complexTensorInfos.real.dataId);
+      const imagValues = this.readSync(complexTensorInfos.imag.dataId);
+      return backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    }
+    return this.data.get(dataId).values;
+  }
+  bufferSync(t) {
+    const data = this.readSync(t.dataId);
+    if (t.dtype === "string") {
+      try {
+        const strings = data.map((d) => util_exports.decodeString(d));
+        return buffer(t.shape, t.dtype, strings);
+      } catch (_a) {
+        throw new Error("Failed to decode encoded string bytes into utf-8");
+      }
+    }
+    return buffer(t.shape, t.dtype, data);
+  }
+  makeOutput(values, shape, dtype) {
+    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);
+  }
+  disposeData(dataId, force = false) {
+    if (this.data.has(dataId)) {
+      this.data.get(dataId).refCount--;
+      if (!force && this.data.get(dataId).refCount > 0) {
+        return false;
+      }
+      const { complexTensorInfos } = this.data.get(dataId);
+      if (complexTensorInfos != null) {
+        this.disposeData(complexTensorInfos.real.dataId, true);
+        this.disposeData(complexTensorInfos.imag.dataId, true);
+      }
+      this.data.delete(dataId);
+    }
+    return true;
+  }
+  disposeIntermediateTensorInfo(tensorInfo) {
+    this.disposeData(tensorInfo.dataId);
+  }
+  async time(f) {
+    const start = util_exports.now();
+    f();
+    const kernelMs = util_exports.now() - start;
+    return { kernelMs };
+  }
+  memory() {
+    return {
+      unreliable: true,
+      reasons: ["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]
+    };
+  }
+  where(condition) {
+    assertNotComplex([condition], "where");
+    const condVals = this.readSync(condition.dataId);
+    return whereImpl2(condition.shape, condVals);
+  }
+  dispose() {
+  }
+  floatPrecision() {
+    return 32;
+  }
+  epsilon() {
+    return super.epsilon();
+  }
+};
+MathBackendCPU.nextDataId = 0;
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/shared.js
+var shared_exports = {};
+__export(shared_exports, {
+  addImpl: () => addImpl,
+  bincountImpl: () => bincountImpl,
+  bincountReduceImpl: () => bincountReduceImpl,
+  castImpl: () => castImpl,
+  ceilImpl: () => ceilImpl,
+  concatImpl: () => concatImpl,
+  equalImpl: () => equalImpl,
+  expImpl: () => expImpl,
+  expm1Impl: () => expm1Impl,
+  floorImpl: () => floorImpl,
+  gatherNdImpl: () => gatherNdImpl,
+  gatherV2Impl: () => gatherV2Impl,
+  greaterEqualImpl: () => greaterEqualImpl,
+  greaterImpl: () => greaterImpl,
+  lessEqualImpl: () => lessEqualImpl,
+  lessImpl: () => lessImpl,
+  linSpaceImpl: () => linSpaceImpl,
+  logImpl: () => logImpl,
+  maxImpl: () => maxImpl,
+  maximumImpl: () => maximumImpl,
+  minimumImpl: () => minimumImpl,
+  multiplyImpl: () => multiplyImpl,
+  negImpl: () => negImpl,
+  notEqualImpl: () => notEqualImpl,
+  prodImpl: () => prodImpl,
+  raggedGatherImpl: () => raggedGatherImpl,
+  raggedRangeImpl: () => raggedRangeImpl,
+  raggedTensorToTensorImpl: () => raggedTensorToTensorImpl,
+  rangeImpl: () => rangeImpl,
+  rsqrtImpl: () => rsqrtImpl,
+  scatterImpl: () => scatterImpl,
+  sigmoidImpl: () => sigmoidImpl,
+  simpleAbsImpl: () => simpleAbsImpl,
+  sliceImpl: () => sliceImpl,
+  sparseFillEmptyRowsImpl: () => sparseFillEmptyRowsImpl,
+  sparseReshapeImpl: () => sparseReshapeImpl,
+  sparseSegmentReductionImpl: () => sparseSegmentReductionImpl,
+  sqrtImpl: () => sqrtImpl,
+  squaredDifferenceImpl: () => squaredDifferenceImpl,
+  stridedSliceImpl: () => stridedSliceImpl,
+  stringNGramsImpl: () => stringNGramsImpl,
+  stringSplitImpl: () => stringSplitImpl,
+  stringToHashBucketFastImpl: () => stringToHashBucketFastImpl,
+  subImpl: () => subImpl,
+  tileImpl: () => tileImpl,
+  topKImpl: () => topKImpl,
+  transposeImpl: () => transposeImpl,
+  uniqueImpl: () => uniqueImpl
+});
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Abs.js
+function simpleAbsImpl(vals) {
+  const resultValues = new Float32Array(vals.length);
+  for (let i = 0; i < vals.length; ++i) {
+    resultValues[i] = Math.abs(vals[i]);
+  }
+  return resultValues;
+}
+var abs2 = (args) => {
+  const { x } = args.inputs;
+  const cpuBackend = args.backend;
+  assertNotComplex(x, "abs");
+  let resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));
+  const values = cpuBackend.data.get(x.dataId).values;
+  resultValues = simpleAbsImpl(values);
+  return cpuBackend.makeOutput(resultValues, x.shape, x.dtype);
+};
+var absConfig = {
+  kernelName: Abs,
+  backendName: "cpu",
+  kernelFunc: abs2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/binary_impl.js
+function createSimpleBinaryKernelImpl(op2) {
+  return (aShape, bShape, aVals, bVals, dtype) => {
+    const newShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const resultRank = newShape.length;
+    const resultStrides = util_exports.computeStrides(newShape);
+    const resultSize = util_exports.sizeFromShape(newShape);
+    const result = util_exports.getTypedArrayFromDType(dtype, resultSize);
+    const aRank = aShape.length;
+    const bRank = bShape.length;
+    const aStrides = util_exports.computeStrides(aShape);
+    const bStrides = util_exports.computeStrides(bShape);
+    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, newShape);
+    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, newShape);
+    if (aBroadcastDims.length + bBroadcastDims.length === 0) {
+      for (let i = 0; i < result.length; ++i) {
+        result[i] = op2(aVals[i % aVals.length], bVals[i % bVals.length]);
+      }
+    } else {
+      for (let i = 0; i < result.length; ++i) {
+        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);
+        const aLoc = loc.slice(-aRank);
+        aBroadcastDims.forEach((d) => aLoc[d] = 0);
+        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);
+        const bLoc = loc.slice(-bRank);
+        bBroadcastDims.forEach((d) => bLoc[d] = 0);
+        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);
+        result[i] = op2(aVals[aIndex], bVals[bIndex]);
+      }
+    }
+    return [result, newShape];
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Complex.js
+function complex2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { real: real4, imag: imag4 } = inputs;
+  const realVals = backend2.data.get(real4.dataId).values;
+  const imagVals = backend2.data.get(imag4.dataId).values;
+  const complexInfo = backend2.makeTensorInfo(real4.shape, "complex64");
+  const complex4 = backend2.data.get(complexInfo.dataId);
+  complex4.complexTensorInfos = {
+    real: backend2.makeTensorInfo(real4.shape, "float32", realVals),
+    imag: backend2.makeTensorInfo(imag4.shape, "float32", imagVals)
+  };
+  return complexInfo;
+}
+var complexConfig = {
+  kernelName: Complex,
+  backendName: "cpu",
+  kernelFunc: complex2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/zeros_impl.js
+function zeros3(backend2, shape, dtype = "float32") {
+  if (dtype === "complex64") {
+    const real4 = zeros3(backend2, shape, "float32");
+    const imag4 = zeros3(backend2, shape, "float32");
+    return complex2({ inputs: { real: real4, imag: imag4 }, backend: backend2 });
+  }
+  const values = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(shape), dtype);
+  return backend2.makeTensorInfo(shape, dtype, values);
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Identity.js
+function identity2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  backend2.incRef(x.dataId);
+  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };
+}
+var identityConfig = {
+  kernelName: Identity,
+  backendName: "cpu",
+  kernelFunc: identity2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Real.js
+function real2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const real4 = backend2.data.get(input2.dataId).complexTensorInfos.real;
+  const realVal = backend2.data.get(real4.dataId).values;
+  return backend2.makeTensorInfo(real4.shape, real4.dtype, realVal);
+}
+var realConfig = {
+  kernelName: Real,
+  backendName: "cpu",
+  kernelFunc: real2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cast.js
+function castImpl(values, shape, inputType, dtype) {
+  if (dtype === "int32") {
+    const resultValues = Int32Array.from(values);
+    return [shape, "int32", resultValues];
+  }
+  if (dtype === "bool") {
+    const zero = util_exports.toTypedArray([0], inputType);
+    const [resultData, resultShape] = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0)(shape, [], values, zero, "bool");
+    return [resultShape, "bool", resultData];
+  }
+  throw new Error(`Error in Cast: failed to cast ${inputType} to ${dtype}`);
+}
+function cast3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dtype } = attrs;
+  if (dtype === "complex64") {
+    if (x.dtype === "complex64") {
+      return identity2({ inputs: { x }, backend: backend2 });
+    }
+    const zerosTensorInfo = zeros3(backend2, x.shape, x.dtype);
+    const floatX = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+    const result = complex2({ inputs: { real: floatX, imag: zerosTensorInfo }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);
+    backend2.disposeIntermediateTensorInfo(floatX);
+    return result;
+  }
+  if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const result = cast3({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    return result;
+  }
+  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {
+    const result = identity2({ inputs: { x }, backend: backend2 });
+    return { dataId: result.dataId, shape: result.shape, dtype };
+  }
+  const values = backend2.data.get(x.dataId).values;
+  const [resultShape, resultType, resultData] = castImpl(values, x.shape, x.dtype, dtype);
+  return backend2.makeTensorInfo(resultShape, resultType, resultData);
+}
+var castConfig = {
+  kernelName: Cast,
+  backendName: "cpu",
+  kernelFunc: cast3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/binary_utils.js
+function binaryKernelFunc(name, simpleImpl, complexImpl, dtype) {
+  if (complexImpl == null) {
+    return ({ inputs, backend: backend2 }) => {
+      const { a, b } = inputs;
+      const cpuBackend = backend2;
+      assertNotComplex([a, b], name);
+      const aVals = cpuBackend.data.get(a.dataId).values;
+      const bVals = cpuBackend.data.get(b.dataId).values;
+      const decodedAVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;
+      const decodedBVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;
+      const $dtype = dtype || a.dtype;
+      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);
+      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);
+    };
+  }
+  return ({ inputs, backend: backend2 }) => {
+    const { a, b } = inputs;
+    const cpuBackend = backend2;
+    if (a.dtype === "complex64" || b.dtype === "complex64") {
+      const $aComplex = cast3({ inputs: { x: a }, backend: cpuBackend, attrs: { dtype: "complex64" } });
+      const $aComplexVals = cpuBackend.data.get($aComplex.dataId);
+      const aReal = $aComplexVals.complexTensorInfos.real;
+      const aImag = $aComplexVals.complexTensorInfos.imag;
+      const aRealVals = cpuBackend.data.get(aReal.dataId).values;
+      const aImagVals = cpuBackend.data.get(aImag.dataId).values;
+      const $bComplex = cast3({ inputs: { x: b }, backend: cpuBackend, attrs: { dtype: "complex64" } });
+      const $bComplexVals = cpuBackend.data.get($bComplex.dataId);
+      const bReal = $bComplexVals.complexTensorInfos.real;
+      const bImag = $bComplexVals.complexTensorInfos.imag;
+      const bRealVals = cpuBackend.data.get(bReal.dataId).values;
+      const bImagVals = cpuBackend.data.get(bImag.dataId).values;
+      const [resultRealData, resultImagData, resultShape] = complexImpl(a.shape, b.shape, aRealVals, aImagVals, bRealVals, bImagVals);
+      const resultReal = cpuBackend.makeTensorInfo(resultShape, "float32", resultRealData);
+      const resultImag = cpuBackend.makeTensorInfo(resultShape, "float32", resultImagData);
+      const result = complex2({ inputs: { real: resultReal, imag: resultImag }, backend: cpuBackend });
+      cpuBackend.disposeIntermediateTensorInfo($aComplex);
+      cpuBackend.disposeIntermediateTensorInfo($bComplex);
+      cpuBackend.disposeIntermediateTensorInfo(resultReal);
+      cpuBackend.disposeIntermediateTensorInfo(resultImag);
+      return result;
+    } else {
+      const aVals = cpuBackend.data.get(a.dataId).values;
+      const bVals = cpuBackend.data.get(b.dataId).values;
+      const $dtype = dtype || a.dtype;
+      const [resultData, resultShape] = simpleImpl(a.shape, b.shape, aVals, bVals, $dtype);
+      return cpuBackend.makeTensorInfo(resultShape, $dtype, resultData);
+    }
+  };
+}
+function createComplexBinaryKernelImpl(op2) {
+  return (aShape, bShape, aRealVals, aImagVals, bRealVals, bImagVals) => {
+    const resultShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const resultSize = util_exports.sizeFromShape(resultShape);
+    const resultRank = resultShape.length;
+    const resultStrides = util_exports.computeStrides(resultShape);
+    const resultRealVals = util_exports.getTypedArrayFromDType("float32", resultSize);
+    const resultImagVals = util_exports.getTypedArrayFromDType("float32", resultSize);
+    const aBroadcastDims = backend_util_exports.getBroadcastDims(aShape, resultShape);
+    const bBroadcastDims = backend_util_exports.getBroadcastDims(bShape, resultShape);
+    const aVals = backend_util_exports.mergeRealAndImagArrays(aRealVals, aImagVals);
+    const bVals = backend_util_exports.mergeRealAndImagArrays(bRealVals, bImagVals);
+    const aRank = aShape.length;
+    const aStrides = util_exports.computeStrides(aShape);
+    const bRank = bShape.length;
+    const bStrides = util_exports.computeStrides(bShape);
+    if (aBroadcastDims.length + bBroadcastDims.length === 0) {
+      for (let i = 0; i < resultRealVals.length; i++) {
+        const aIdx = i % aVals.length;
+        const bIdx = i % bVals.length;
+        const result = op2(aVals[aIdx * 2], aVals[aIdx * 2 + 1], bVals[bIdx * 2], bVals[bIdx * 2 + 1]);
+        resultRealVals[i] = result.real;
+        resultImagVals[i] = result.imag;
+      }
+    } else {
+      for (let i = 0; i < resultRealVals.length; i++) {
+        const loc = util_exports.indexToLoc(i, resultRank, resultStrides);
+        const aLoc = loc.slice(-aRank);
+        aBroadcastDims.forEach((d) => aLoc[d] = 0);
+        const aIndex = util_exports.locToIndex(aLoc, aRank, aStrides);
+        const bLoc = loc.slice(-bRank);
+        bBroadcastDims.forEach((d) => bLoc[d] = 0);
+        const bIndex = util_exports.locToIndex(bLoc, bRank, bStrides);
+        const opResult = op2(aVals[aIndex * 2], aVals[aIndex * 2 + 1], bVals[bIndex * 2], bVals[bIndex * 2 + 1]);
+        resultRealVals[i] = opResult.real;
+        resultImagVals[i] = opResult.imag;
+      }
+    }
+    return [resultRealVals, resultImagVals, resultShape];
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Add.js
+var addImpl = createSimpleBinaryKernelImpl((a, b) => a + b);
+var addComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return { real: aReal + bReal, imag: aImag + bImag };
+});
+var add4 = binaryKernelFunc(Add, addImpl, addComplexImpl);
+var addConfig = {
+  kernelName: Add,
+  backendName: "cpu",
+  kernelFunc: add4
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Bincount_impl.js
+function bincountImpl(xVals, weightsVals, weightsDtype, weightsShape, size) {
+  const weightsSize = util_exports.sizeFromShape(weightsShape);
+  const outVals = util_exports.makeZerosTypedArray(size, weightsDtype);
+  for (let i = 0; i < xVals.length; i++) {
+    const value = xVals[i];
+    if (value < 0) {
+      throw new Error("Input x must be non-negative!");
+    }
+    if (value >= size) {
+      continue;
+    }
+    if (weightsSize > 0) {
+      outVals[value] += weightsVals[i];
+    } else {
+      outVals[value] += 1;
+    }
+  }
+  return outVals;
+}
+function bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput = false) {
+  const numRows = xBuf.shape[0];
+  const numCols = xBuf.shape[1];
+  const outBuf = buffer([numRows, size], weightsBuf.dtype);
+  for (let i = 0; i < numRows; i++) {
+    for (let j = 0; j < numCols; j++) {
+      const value = xBuf.get(i, j);
+      if (value < 0) {
+        throw new Error("Input x must be non-negative!");
+      }
+      if (value >= size) {
+        continue;
+      }
+      if (binaryOutput) {
+        outBuf.set(1, i, value);
+      } else {
+        if (weightsBuf.size > 0) {
+          outBuf.set(outBuf.get(i, value) + weightsBuf.get(i, j), i, value);
+        } else {
+          outBuf.set(outBuf.get(i, value) + 1, i, value);
+        }
+      }
+    }
+  }
+  return outBuf;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/unary_impl.js
+function createSimpleUnaryImpl(op2) {
+  return (values, dtype, attrs) => {
+    const newValues = util_exports.getTypedArrayFromDType(dtype, values.length);
+    for (let i = 0; i < values.length; ++i) {
+      newValues[i] = op2(values[i], attrs);
+    }
+    return newValues;
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/unary_utils.js
+function unaryKernelFunc(name, op2, dtype) {
+  return ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    assertNotComplex(x, name);
+    if (x.dtype === "string" || dtype === "string") {
+      throw new Error("unaryKernelFunc does not support string input/output");
+    }
+    const cpuBackend = backend2;
+    const values = cpuBackend.data.get(x.dataId).values;
+    const xSize = util_exports.sizeFromShape(x.shape);
+    const $dtype = dtype || x.dtype;
+    const newValues = util_exports.getArrayFromDType($dtype, xSize);
+    for (let i = 0; i < xSize; ++i) {
+      newValues[i] = op2(values[i], attrs);
+    }
+    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);
+  };
+}
+function unaryKernelFuncFromImpl(name, unaryImpl, dtype) {
+  return ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    assertNotComplex(x, name);
+    if (x.dtype === "string" || dtype === "string") {
+      throw new Error("unaryKernelFunc does not support string input/output");
+    }
+    const cpuBackend = backend2;
+    const values = cpuBackend.data.get(x.dataId).values;
+    const $dtype = dtype || x.dtype;
+    const newValues = unaryImpl(values, $dtype, attrs);
+    return cpuBackend.makeTensorInfo(x.shape, $dtype, newValues);
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Ceil.js
+var ceilImpl = createSimpleUnaryImpl((xi) => Math.ceil(xi));
+var ceil2 = unaryKernelFuncFromImpl(Ceil, ceilImpl);
+var ceilConfig = {
+  kernelName: Ceil,
+  backendName: "cpu",
+  kernelFunc: ceil2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Concat_impl.js
+function concatImpl(inputs, outShape, dtype, simplyConcat) {
+  const outVals = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(outShape));
+  if (simplyConcat && dtype !== "string") {
+    let offset = 0;
+    inputs.forEach((input2) => {
+      const size = util_exports.sizeFromShape(input2.shape);
+      outVals.set(input2.vals, offset);
+      offset += size;
+    });
+  } else {
+    let colOffset = 0;
+    inputs.forEach((input2) => {
+      const decodedData = dtype === "string" ? backend_util_exports.fromUint8ToStringArray(input2.vals) : input2.vals;
+      let tIdx = 0;
+      for (let row = 0; row < input2.shape[0]; ++row) {
+        const resIdx = row * outShape[1] + colOffset;
+        for (let col = 0; col < input2.shape[1]; ++col) {
+          outVals[resIdx + col] = decodedData[tIdx++];
+        }
+      }
+      colOffset += input2.shape[1];
+    });
+  }
+  return outVals;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Equal.js
+var equalImpl = createSimpleBinaryKernelImpl((a, b) => a === b ? 1 : 0);
+var equal2 = binaryKernelFunc(Equal, equalImpl, null, "bool");
+var equalConfig = {
+  kernelName: Equal,
+  backendName: "cpu",
+  kernelFunc: equal2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Exp.js
+var expImpl = createSimpleUnaryImpl((xi) => Math.exp(xi));
+var exp2 = unaryKernelFuncFromImpl(Exp, expImpl, "float32");
+var expConfig = {
+  kernelName: Exp,
+  backendName: "cpu",
+  kernelFunc: exp2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Expm1.js
+var expm1Impl = createSimpleUnaryImpl((xi) => Math.expm1(xi));
+var expm12 = unaryKernelFuncFromImpl(Expm1, expm1Impl);
+var expm1Config = {
+  kernelName: Expm1,
+  backendName: "cpu",
+  kernelFunc: expm12
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Floor.js
+var floorImpl = createSimpleUnaryImpl((xi) => Math.floor(xi));
+var floor2 = unaryKernelFuncFromImpl(Floor, floorImpl);
+var floorConfig = {
+  kernelName: Floor,
+  backendName: "cpu",
+  kernelFunc: floor2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherNd_Impl.js
+function gatherNdImpl(indicesData, paramsBuf, dtype, numSlices, sliceRank, sliceSize, strides, paramsShape, paramsSize) {
+  const outBuf = buffer([numSlices, sliceSize], dtype);
+  for (let i = 0; i < numSlices; i++) {
+    const index = [];
+    let flattenIndex = 0;
+    for (let j = 0; j < sliceRank; j++) {
+      const dim = indicesData[i * sliceRank + j];
+      flattenIndex += dim * strides[j];
+      index.push(dim);
+    }
+    if (flattenIndex < 0 || flattenIndex >= paramsSize / sliceSize) {
+      throw new Error(`Invalid indices: ${index} does not index into ${paramsShape}`);
+    }
+    for (let k = 0; k < sliceSize; k++) {
+      outBuf.values[i * sliceSize + k] = paramsBuf.get(...paramsBuf.indexToLoc(flattenIndex * sliceSize + k));
+    }
+  }
+  return outBuf;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherV2_impl.js
+function gatherV2Impl(xBuf, indicesBuf, flattenOutputShape) {
+  const outBuf = buffer(flattenOutputShape, xBuf.dtype);
+  for (let i = 0; i < outBuf.size; ++i) {
+    const newLoc = outBuf.indexToLoc(i);
+    const originalLoc = newLoc.slice();
+    const batchIdx = originalLoc[0];
+    const indicesIdx = originalLoc[2];
+    const indicesIndex = indicesBuf.locToIndex([batchIdx, indicesIdx]);
+    originalLoc[2] = indicesBuf.values[indicesIndex];
+    const originalIndex = xBuf.locToIndex(originalLoc);
+    if (0 <= originalIndex && originalIndex < xBuf.values.length) {
+      outBuf.values[i] = xBuf.values[originalIndex];
+    }
+  }
+  return outBuf;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Greater.js
+var greaterImpl = createSimpleBinaryKernelImpl((a, b) => a > b ? 1 : 0);
+var greater3 = binaryKernelFunc(Greater, greaterImpl, null, "bool");
+var greaterConfig = {
+  kernelName: Greater,
+  backendName: "cpu",
+  kernelFunc: greater3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GreaterEqual.js
+var greaterEqualImpl = createSimpleBinaryKernelImpl((a, b) => a >= b ? 1 : 0);
+var greaterEqual2 = binaryKernelFunc(GreaterEqual, greaterEqualImpl, null, "bool");
+var greaterEqualConfig = {
+  kernelName: GreaterEqual,
+  backendName: "cpu",
+  kernelFunc: greaterEqual2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Less.js
+var lessImpl = createSimpleBinaryKernelImpl((a, b) => a < b ? 1 : 0);
+var less3 = binaryKernelFunc(Less, lessImpl, null, "bool");
+var lessConfig = {
+  kernelName: Less,
+  backendName: "cpu",
+  kernelFunc: less3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LessEqual.js
+var lessEqualImpl = createSimpleBinaryKernelImpl((a, b) => a <= b ? 1 : 0);
+var lessEqual2 = binaryKernelFunc(LessEqual, lessEqualImpl, null, "bool");
+var lessEqualConfig = {
+  kernelName: LessEqual,
+  backendName: "cpu",
+  kernelFunc: lessEqual2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LinSpace_impl.js
+function linSpaceImpl(start, stop, num) {
+  const step5 = (stop - start) / (num - 1);
+  const values = util_exports.makeZerosTypedArray(num, "float32");
+  values[0] = start;
+  for (let i = 1; i < values.length; i++) {
+    values[i] = values[i - 1] + step5;
+  }
+  return values;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Log.js
+var logImpl = createSimpleUnaryImpl((xi) => Math.log(xi));
+var log3 = unaryKernelFuncFromImpl(Log, logImpl);
+var logConfig = {
+  kernelName: Log,
+  backendName: "cpu",
+  kernelFunc: log3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Max_impl.js
+function maxImpl(aVals, reduceSize, outShape, dtype) {
+  const vals = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(outShape));
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let max6 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (Number.isNaN(value) || value > max6) {
+        max6 = value;
+      }
+    }
+    vals[i] = max6;
+  }
+  return vals;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Maximum.js
+var maximumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.max(aValue, bValue));
+var maximum3 = binaryKernelFunc(Maximum, maximumImpl);
+var maximumConfig = {
+  kernelName: Maximum,
+  backendName: "cpu",
+  kernelFunc: maximum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Minimum.js
+var minimumImpl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.min(aValue, bValue));
+var minimum3 = binaryKernelFunc(Minimum, minimumImpl);
+var minimumConfig = {
+  kernelName: Minimum,
+  backendName: "cpu",
+  kernelFunc: minimum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multiply.js
+var multiplyImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue * bValue);
+var multiplyComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return {
+    real: aReal * bReal - aImag * bImag,
+    imag: aReal * bImag + aImag * bReal
+  };
+});
+var multiply2 = binaryKernelFunc(Multiply, multiplyImpl, multiplyComplexImpl);
+var multiplyConfig = {
+  kernelName: Multiply,
+  backendName: "cpu",
+  kernelFunc: multiply2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Neg.js
+function negImpl(xVals, xShape, xDtype) {
+  const minusOne = util_exports.createScalarValue(-1, xDtype);
+  return multiplyImpl([], xShape, minusOne, xVals, xDtype);
+}
+function neg2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "neg");
+  const xVals = backend2.data.get(x.dataId).values;
+  const [res, newShape] = negImpl(xVals, x.shape, x.dtype);
+  return backend2.makeTensorInfo(newShape, x.dtype, res);
+}
+var negConfig = {
+  kernelName: Neg,
+  backendName: "cpu",
+  kernelFunc: neg2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NotEqual.js
+var notEqualImpl = createSimpleBinaryKernelImpl((a, b) => a !== b ? 1 : 0);
+var notEqual2 = binaryKernelFunc(NotEqual, notEqualImpl, null, "bool");
+var notEqualConfig = {
+  kernelName: NotEqual,
+  backendName: "cpu",
+  kernelFunc: notEqual2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transpose_impl.js
+function transposeImpl(xVals, xShape, dtype, perm, newShape) {
+  const xRank = xShape.length;
+  const xSize = util_exports.sizeFromShape(xShape);
+  const xStrides = util_exports.computeStrides(xShape);
+  const newStrides = util_exports.computeStrides(newShape);
+  const result = util_exports.getTypedArrayFromDType(dtype, util_exports.sizeFromShape(newShape));
+  for (let i = 0; i < xSize; ++i) {
+    const loc = util_exports.indexToLoc(i, xRank, xStrides);
+    const newLoc = new Array(loc.length);
+    for (let i2 = 0; i2 < newLoc.length; i2++) {
+      newLoc[i2] = loc[perm[i2]];
+    }
+    const newIndex = util_exports.locToIndex(newLoc, xRank, newStrides);
+    result[newIndex] = xVals[i];
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transpose.js
+function transpose2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x } = inputs;
+  const { perm } = attrs;
+  assertNotComplex(x, "transpose");
+  const xRank = x.shape.length;
+  const newShape = new Array(xRank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[perm[i]];
+  }
+  const values = backend2.data.get(x.dataId).values;
+  const result = transposeImpl(values, x.shape, x.dtype, perm, newShape);
+  const dataId = backend2.write(result, newShape, x.dtype);
+  return { dataId, shape: newShape, dtype: x.dtype };
+}
+var transposeConfig = {
+  kernelName: Transpose,
+  backendName: "cpu",
+  kernelFunc: transpose2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Prod.js
+function prodImpl(xShape, xDtype, xVals, reductionAxes) {
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, reductionAxes);
+  const outDtype = upcastType(xDtype, "int32");
+  const outVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), outDtype);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  for (let i = 0; i < outVals.length; ++i) {
+    const offset = i * reduceSize;
+    let prod5 = 1;
+    for (let j = 0; j < reduceSize; ++j) {
+      prod5 *= xVals[offset + j];
+    }
+    outVals[i] = prod5;
+  }
+  return { outVals, outShape, outDtype };
+}
+function prod2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "prod");
+  const xRank = x.shape.length;
+  const axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);
+  let reductionAxes = axes;
+  let permutedX = x;
+  const intermediateTensorInfos = [];
+  if (permutation != null) {
+    permutedX = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+    intermediateTensorInfos.push(permutedX);
+    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);
+  }
+  const xVals = backend2.data.get(permutedX.dataId).values;
+  const { outVals, outShape, outDtype } = prodImpl(permutedX.shape, permutedX.dtype, xVals, reductionAxes);
+  let resultShape = outShape;
+  if (keepDims) {
+    resultShape = backend_util_exports.expandShapeToKeepDim(outShape, axes);
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(resultShape, outDtype, outVals);
+}
+var prodConfig = {
+  kernelName: Prod,
+  backendName: "cpu",
+  kernelFunc: prod2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedGather_impl.js
+function validateIndices(indices, indicesShape, numParams) {
+  indices.forEach((index, i) => {
+    if (index < 0 || index >= numParams) {
+      const locString = util_exports.indexToLoc(i, indicesShape.length, util_exports.computeStrides(indicesShape)).join(",");
+      throw new Error(`indices[${locString}] = ${index} is not in [0, ${numParams})`);
+    }
+  });
+}
+function validateSplits(paramsNestedSplits, numParamsDenseValues) {
+  for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {
+    const splits = paramsNestedSplits[dim];
+    const lastSplit = dim === paramsNestedSplits.length - 1 ? numParamsDenseValues : paramsNestedSplits[dim + 1].length;
+    if (splits.length === 0) {
+      throw new Error("Ragged splits may not be empty");
+    }
+    if (splits[0] < 0) {
+      throw new Error("Ragged splits must be non-negative");
+    }
+    if (splits[splits.length - 1] > lastSplit) {
+      throw new Error("Ragged splits must not point past values");
+    }
+    for (let i = 1; i < splits.length; ++i) {
+      if (splits[i - 1] > splits[i]) {
+        throw new Error("Ragged splits must be sorted in ascending order");
+      }
+    }
+  }
+}
+function makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues) {
+  const valueSlices = [];
+  let numValues = 0;
+  const numSplits = indicesShape.length - 1 + paramsNestedSplits.length;
+  const outSplits = new Array(numSplits).fill(null).map(() => [0]);
+  validateSplits(paramsNestedSplits, numParamsDenseValues);
+  let nrows = 1;
+  for (let dim = 0; dim < indicesShape.length - 1; ++dim) {
+    nrows *= indicesShape[dim];
+    const rowLength = indicesShape[dim + 1];
+    for (let i = 1; i < nrows + 1; ++i) {
+      outSplits[dim].push(i * rowLength);
+    }
+  }
+  for (let i = 0; i < indices.length; ++i) {
+    let start = indices[i];
+    let limit = indices[i] + 1;
+    for (let dim = 0; dim < paramsNestedSplits.length; ++dim) {
+      const splits = paramsNestedSplits[dim];
+      const outDim = dim + indicesShape.length - 1;
+      if (outDim >= 0) {
+        const outSplitsOutDim = outSplits[outDim];
+        const delta = outSplitsOutDim[outSplitsOutDim.length - 1] - splits[start];
+        for (let j = start; j < limit; ++j) {
+          outSplits[outDim].push(splits[j + 1] + delta);
+        }
+      }
+      start = splits[start];
+      limit = splits[limit];
+    }
+    if (limit !== start) {
+      valueSlices.push([start, limit]);
+      numValues += limit - start;
+    }
+  }
+  return { outSplits, valueSlices, numValues };
+}
+function getSplits(outSplits) {
+  const splitsOut = [];
+  for (let i = 0; i < outSplits.length; ++i) {
+    const numSplits = outSplits[i].length;
+    const splits = util_exports.getArrayFromDType("int32", numSplits);
+    splitsOut.push(splits);
+    outSplits[i].forEach((value, j) => splits[j] = value);
+  }
+  return splitsOut;
+}
+function computeFlatOuterDims(orig, numOutDims) {
+  const outDims = orig.slice(0, numOutDims);
+  while (outDims.length < numOutDims) {
+    outDims.push(1);
+  }
+  for (let inDim = numOutDims; inDim < orig.length; inDim++) {
+    outDims[numOutDims - 1] *= orig[inDim];
+  }
+  return outDims;
+}
+function writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, values, valuesShape) {
+  const denseM = computeFlatOuterDims(paramsDenseValuesShape, 2)[1];
+  const valuesM = computeFlatOuterDims(valuesShape, 2)[1];
+  let outPos = 0;
+  for (const slice5 of valueSlices) {
+    for (let i = slice5[0]; i < slice5[1]; ++i) {
+      for (let j = 0; j < valueSize; ++j) {
+        values[outPos * valuesM + j] = paramsDenseValues[i * denseM + j];
+      }
+      ++outPos;
+    }
+  }
+}
+function getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues) {
+  const valuesShape = paramsDenseValuesShape.slice();
+  valuesShape[0] = numValues;
+  const valuesOut = util_exports.getArrayFromDType(paramsDenseValuesDType, util_exports.sizeFromShape(valuesShape));
+  const numElements = paramsDenseValues.length;
+  const valueSize = numElements === 0 ? 0 : numElements / paramsDenseValuesShape[0];
+  writeValueSlices(paramsDenseValues, paramsDenseValuesShape, valueSlices, valueSize, valuesOut, valuesShape);
+  return [valuesOut, valuesShape];
+}
+function raggedGatherImpl(paramsNestedSplits, paramsNestedSplitsShapes, paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, indices, indicesShape, outputRaggedRank) {
+  if (paramsNestedSplits.length === 0) {
+    throw new Error("paramsNestedSplits must be non empty");
+  }
+  if (paramsNestedSplitsShapes[0].length === 0) {
+    throw new Error("Split tensors must not be scalars");
+  }
+  const numParams = paramsNestedSplitsShapes[0][0] - 1;
+  validateIndices(indices, indicesShape, numParams);
+  if (paramsDenseValuesShape.length === 0) {
+    throw new Error("params.rank must be nonzero");
+  }
+  const numParamsDenseValues = paramsDenseValuesShape[0];
+  const { outSplits, valueSlices, numValues } = makeSplits(indices, indicesShape, paramsNestedSplits, numParamsDenseValues);
+  const outputNestedSplits = getSplits(outSplits);
+  const outputDenseValues = getValues(paramsDenseValues, paramsDenseValuesShape, paramsDenseValuesDType, valueSlices, numValues);
+  return [outputNestedSplits, outputDenseValues[0], outputDenseValues[1]];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedRange_impl.js
+var INT32_MAX2 = 2147483647;
+function raggedRangeImpl(starts, startsShape, startsDType, limits, limitsShape, deltas, deltasShape) {
+  if (startsShape.length > 1) {
+    throw new Error("starts must be a scalar or vector");
+  }
+  if (limitsShape.length > 1) {
+    throw new Error("limits must be a scalar or vector");
+  }
+  if (deltasShape.length > 1) {
+    throw new Error("deltas must be a scalar or vector");
+  }
+  const broadcastStarts = startsShape.length === 0;
+  const broadcastLimits = limitsShape.length === 0;
+  const broadcastDeltas = deltasShape.length === 0;
+  const inSizes = [];
+  if (!broadcastStarts) {
+    inSizes.push(startsShape[0]);
+  }
+  if (!broadcastLimits) {
+    inSizes.push(limitsShape[0]);
+  }
+  if (!broadcastDeltas) {
+    inSizes.push(deltasShape[0]);
+  }
+  for (let i = 1; i < inSizes.length; ++i) {
+    if (inSizes[i] !== inSizes[i - 1]) {
+      throw new Error("starts, limits, and deltas must have the same shape");
+    }
+  }
+  const nRows = inSizes.length === 0 ? 1 : inSizes[0];
+  const rtNestedSplits = util_exports.getArrayFromDType("int32", nRows + 1);
+  rtNestedSplits[0] = 0;
+  for (let row = 0; row < nRows; ++row) {
+    const start = broadcastStarts ? starts[0] : starts[row];
+    const limit = broadcastLimits ? limits[0] : limits[row];
+    const delta = broadcastDeltas ? deltas[0] : deltas[row];
+    if (delta === 0) {
+      throw new Error("Requires delta != 0");
+    }
+    let size;
+    if (delta > 0 && limit < start || delta < 0 && limit > start) {
+      size = 0;
+    } else {
+      size = Math.ceil(Math.abs((limit - start) / delta));
+      if (size > INT32_MAX2) {
+        throw new Error(`Requires ((limit - start) / delta) <= ${INT32_MAX2}`);
+      }
+    }
+    rtNestedSplits[row + 1] = rtNestedSplits[row] + size;
+  }
+  const nVals = rtNestedSplits[nRows];
+  const rtDenseValues = util_exports.getArrayFromDType(startsDType, nVals);
+  let valueIndex = 0;
+  for (let row = 0; row < nRows; ++row) {
+    const rowSize = rtNestedSplits[row + 1] - rtNestedSplits[row];
+    let value = broadcastStarts ? starts[0] : starts[row];
+    const delta = broadcastDeltas ? deltas[0] : deltas[row];
+    for (let i = 0; i < rowSize; ++i) {
+      rtDenseValues[valueIndex++] = value;
+      value += delta;
+    }
+  }
+  return [rtNestedSplits, rtDenseValues];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedTensorToTensor_impl.js
+var RowPartitionType2 = backend_util_exports.RowPartitionType;
+var RaggedTensorToTensorOp = class {
+  constructor(shape, shapeShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypeStrings) {
+    this.shape = shape;
+    this.shapeShape = shapeShape;
+    this.values = values;
+    this.valuesShape = valuesShape;
+    this.valuesDType = valuesDType;
+    this.defaultValue = defaultValue;
+    this.defaultValueShape = defaultValueShape;
+    this.rowPartitionValues = rowPartitionValues;
+    this.rowPartitionValuesShapes = rowPartitionValuesShapes;
+    this.rowPartitionTypes = backend_util_exports.getRowPartitionTypesHelper(rowPartitionTypeStrings);
+    this.raggedRank = backend_util_exports.getRaggedRank(this.rowPartitionTypes);
+  }
+  getRowPartitionTypeByDimension(dimension) {
+    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {
+      return this.rowPartitionTypes[dimension + 1];
+    } else {
+      return this.rowPartitionTypes[dimension];
+    }
+  }
+  getRowPartitionTensor(dimension) {
+    if (this.rowPartitionTypes[0] === RowPartitionType2.FIRST_DIM_SIZE) {
+      return this.rowPartitionValues[dimension + 1];
+    } else {
+      return this.rowPartitionValues[dimension];
+    }
+  }
+  getMaxWidth(dimension) {
+    const rowPartitionTensor = this.getRowPartitionTensor(dimension - 1);
+    switch (this.getRowPartitionTypeByDimension(dimension - 1)) {
+      case RowPartitionType2.VALUE_ROWIDS:
+        return RaggedTensorToTensorOp.getMaxWidthValueRowID(rowPartitionTensor);
+      case RowPartitionType2.ROW_SPLITS:
+        return RaggedTensorToTensorOp.getMaxWidthRowSplit(rowPartitionTensor);
+      default:
+        throw new Error(`Cannot handle partition type ${RowPartitionType2[this.getRowPartitionTypeByDimension(dimension - 1)]}`);
+    }
+  }
+  static getMaxWidthRowSplit(rowSplit) {
+    const tensorLength = rowSplit.length;
+    if (tensorLength === 0 || tensorLength === 1) {
+      return 0;
+    }
+    let maxWidth = 0;
+    for (let i = 0; i < tensorLength - 1; ++i) {
+      const currentWidth = rowSplit[i + 1] - rowSplit[i];
+      if (currentWidth > maxWidth) {
+        maxWidth = currentWidth;
+      }
+    }
+    return maxWidth;
+  }
+  static getMaxWidthValueRowID(valueRowIds) {
+    const indexLength = valueRowIds.length;
+    if (indexLength === 0) {
+      return 0;
+    }
+    let firstEqualIndex = 0;
+    let firstEqualIndexValue = valueRowIds[0];
+    let maxWidth = 0;
+    for (let i = 1; i < indexLength; ++i) {
+      const value = valueRowIds[i];
+      if (value !== firstEqualIndexValue) {
+        firstEqualIndexValue = value;
+        maxWidth = Math.max(i - firstEqualIndex, maxWidth);
+        firstEqualIndex = i;
+      }
+    }
+    return Math.max(indexLength - firstEqualIndex, maxWidth);
+  }
+  tensorShapeFromTensor(t, tShape, isPartial = true) {
+    if (tShape.length === 0) {
+      if (t[0] === -1) {
+        return [];
+      }
+      throw new Error(`The only valid scalar shape tensor is the fully unknown shape specified as -1.`);
+    }
+    return makeShape(t, isPartial);
+  }
+  calculateOutputSize(firstDim) {
+    const valueShape = this.valuesShape;
+    const defaultValueShape = this.defaultValueShape;
+    backend_util_exports.validateDefaultValueShape(defaultValueShape, valueShape);
+    const shape = this.tensorShapeFromTensor(this.shape, this.shapeShape);
+    const outputShape = backend_util_exports.combineRaggedTensorToTensorShapes(this.raggedRank, shape, valueShape);
+    const result = outputShape;
+    if (result[0] < 0) {
+      result[0] = firstDim;
+    }
+    for (let i = 1; i <= this.raggedRank; ++i) {
+      if (result[i] < 0) {
+        result[i] = this.getMaxWidth(i);
+      }
+    }
+    return result;
+  }
+  calculateFirstParentOutputIndex(firstDimension, outputIndexMultiplier, firstDimensionOutput) {
+    const minDimension = Math.min(firstDimension, firstDimensionOutput);
+    const result = [];
+    let currentOutputIndex = 0;
+    for (let i = 0; i < minDimension; ++i, currentOutputIndex += outputIndexMultiplier) {
+      result.push(currentOutputIndex);
+    }
+    for (let i = minDimension; i < firstDimension; ++i) {
+      result.push(-1);
+    }
+    util_exports.assert(result.length === firstDimension, () => "Final length of result must be equal to firstDimension.");
+    return result;
+  }
+  calculateOutputIndexRowSplit(rowSplit, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const rowSplitSize = rowSplit.length;
+    const result = [];
+    for (let i = 0; i < rowSplitSize - 1; ++i) {
+      const rowLength = rowSplit[i + 1] - rowSplit[i];
+      let realLength = Math.min(outputSize, rowLength);
+      let parentOutputIndexCurrent = parentOutputIndex[i];
+      if (parentOutputIndexCurrent === -1) {
+        realLength = 0;
+      }
+      for (let j = 0; j < realLength; ++j) {
+        result.push(parentOutputIndexCurrent);
+        parentOutputIndexCurrent += outputIndexMultiplier;
+      }
+      for (let j = 0; j < rowLength - realLength; ++j) {
+        result.push(-1);
+      }
+    }
+    if (rowSplitSize > 0 && result.length !== rowSplit[rowSplitSize - 1]) {
+      throw new Error("Invalid row split size.");
+    }
+    return result;
+  }
+  calculateOutputIndexValueRowID(valueRowIds, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const indexSize = valueRowIds.length;
+    const result = [];
+    if (indexSize === 0) {
+      return [];
+    }
+    let currentOutputColumn = 0;
+    let currentValueRowId = valueRowIds[0];
+    if (currentValueRowId >= parentOutputIndex.length) {
+      throw new Error(`Got currentValueRowId=${currentValueRowId}, which is not less than ${parentOutputIndex.length}`);
+    }
+    let currentOutputIndex = parentOutputIndex[currentValueRowId];
+    result.push(currentOutputIndex);
+    for (let i = 1; i < indexSize; ++i) {
+      const nextValueRowId = valueRowIds[i];
+      if (nextValueRowId === currentValueRowId) {
+        if (currentOutputIndex >= 0) {
+          ++currentOutputColumn;
+          if (currentOutputColumn < outputSize) {
+            currentOutputIndex += outputIndexMultiplier;
+          } else {
+            currentOutputIndex = -1;
+          }
+        }
+      } else {
+        currentOutputColumn = 0;
+        currentValueRowId = nextValueRowId;
+        if (nextValueRowId >= parentOutputIndex.length) {
+          throw new Error(`Got nextValueRowId=${nextValueRowId} which is not less than ${parentOutputIndex.length}`);
+        }
+        currentOutputIndex = parentOutputIndex[nextValueRowId];
+      }
+      result.push(currentOutputIndex);
+    }
+    if (result.length !== valueRowIds.length) {
+      throw new Error("Invalid row ids.");
+    }
+    return result;
+  }
+  calculateOutputIndex(dimension, parentOutputIndex, outputIndexMultiplier, outputSize) {
+    const rowPartitionTensor = this.getRowPartitionTensor(dimension);
+    const partitionType = this.getRowPartitionTypeByDimension(dimension);
+    switch (partitionType) {
+      case RowPartitionType2.VALUE_ROWIDS:
+        return this.calculateOutputIndexValueRowID(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);
+      case RowPartitionType2.ROW_SPLITS:
+        if (rowPartitionTensor.length - 1 > parentOutputIndex.length) {
+          throw new Error(`Row partition size is greater than output size: ${rowPartitionTensor.length - 1} > ${parentOutputIndex.length}`);
+        }
+        return this.calculateOutputIndexRowSplit(rowPartitionTensor, parentOutputIndex, outputIndexMultiplier, outputSize);
+      default:
+        throw new Error(`Unsupported partition type: ${RowPartitionType2[partitionType]}`);
+    }
+  }
+  getFirstDimensionSize() {
+    const firstPartitionTensor = this.rowPartitionValues[0];
+    if (this.rowPartitionTypes.length === 0) {
+      throw new Error("No row_partition_types given.");
+    }
+    const firstPartitionType = this.rowPartitionTypes[0];
+    switch (firstPartitionType) {
+      case RowPartitionType2.FIRST_DIM_SIZE:
+        return firstPartitionTensor[0];
+      case RowPartitionType2.VALUE_ROWIDS:
+        throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");
+      case RowPartitionType2.ROW_SPLITS:
+        return this.rowPartitionValuesShapes[0][0] - 1;
+      default:
+        throw new Error(`Cannot handle type ${RowPartitionType2[firstPartitionType]}`);
+    }
+  }
+  compute() {
+    const firstPartitionTensor = this.rowPartitionValues[0];
+    if (firstPartitionTensor.length <= 0) {
+      throw new Error("Invalid first partition input. Tensor requires at least one element.");
+    }
+    const firstDimension = this.getFirstDimensionSize();
+    const outputSize = this.calculateOutputSize(firstDimension);
+    const multiplier = new Array(this.raggedRank + 1);
+    multiplier[multiplier.length - 1] = 1;
+    for (let i = multiplier.length - 2; i >= 0; --i) {
+      multiplier[i] = multiplier[i + 1] * outputSize[i + 1];
+    }
+    const outputShape = makeShape(outputSize, false);
+    const outputTensor = util_exports.getArrayFromDType(this.valuesDType, util_exports.sizeFromShape(outputShape));
+    const fullSize = multiplier[0] * outputSize[0];
+    if (fullSize > 0) {
+      let outputIndex = this.calculateFirstParentOutputIndex(firstDimension, multiplier[0], outputSize[0]);
+      for (let i = 1; i <= this.raggedRank; ++i) {
+        const newOutputIndex = this.calculateOutputIndex(i - 1, outputIndex, multiplier[i], outputSize[i]);
+        outputIndex = newOutputIndex;
+      }
+      this.setOutput(this.raggedRank, outputIndex, outputTensor, outputShape);
+    }
+    return [outputShape, outputTensor];
+  }
+  setOutput(raggedRank, outputIndex, outputTensor, outputShape) {
+    if (outputTensor.length === 0) {
+      return;
+    }
+    const valuesBase = this.values;
+    const outputBase = outputTensor;
+    let elementShape = outputShape.slice();
+    elementShape = elementShape.slice(raggedRank + 1);
+    const valueElementSize = util_exports.sizeFromShape(elementShape);
+    const outputIndexSize = outputIndex.length;
+    let defaultValue = this.defaultValue;
+    if (defaultValue.length !== valueElementSize && defaultValue.length !== 1) {
+      const srcShape = this.defaultValueShape;
+      tidy(() => {
+        const defaultValueTensor = reshape(defaultValue, srcShape);
+        const bCastDefault = broadcastTo(defaultValueTensor, elementShape);
+        defaultValue = bCastDefault.dataSync();
+      });
+    }
+    let srcStart = 0;
+    let dstStart = 0;
+    let dstEnd = 0;
+    for (let srcI = 0; srcI <= outputIndexSize; ++srcI) {
+      let dstI = srcI < outputIndexSize ? outputIndex[srcI] : -1;
+      if (dstI === dstEnd) {
+        ++dstEnd;
+        continue;
+      }
+      if (dstStart < dstEnd) {
+        const src = valuesBase.subarray(srcStart * valueElementSize);
+        const dst = outputBase.subarray(dstStart * valueElementSize);
+        const nVals = (dstEnd - dstStart) * valueElementSize;
+        copyArray(dst, src, nVals);
+      }
+      if (srcI >= outputIndexSize) {
+        const outputSize = outputTensor.length;
+        dstI = Math.floor(outputSize / valueElementSize);
+      }
+      if (dstI > dstEnd) {
+        if (this.defaultValue.length === 1) {
+          outputBase.subarray(dstEnd * valueElementSize, dstI * valueElementSize).fill(this.defaultValue[0]);
+          dstEnd = dstI;
+        } else {
+          while (dstI > dstEnd) {
+            const dst = outputBase.slice(dstEnd * valueElementSize);
+            copyArray(dst, defaultValue, valueElementSize);
+            ++dstEnd;
+          }
+        }
+      }
+      if (dstI < 0) {
+        srcStart = srcI + 1;
+        dstStart = dstEnd;
+      } else {
+        srcStart = srcI;
+        dstStart = dstEnd;
+        dstEnd = dstStart + 1;
+      }
+    }
+  }
+};
+function copyArray(dst, src, size) {
+  for (let i = 0; i < size; i++) {
+    dst[i] = src[i];
+  }
+}
+function makeShape(shape, isPartial) {
+  const out = [];
+  for (let dim of shape) {
+    if (dim < 0) {
+      if (!isPartial) {
+        throw new Error(`Dimension ${dim} must be >= 0`);
+      }
+      if (dim < -1) {
+        throw new Error(`Dimension ${dim} must be >= -1`);
+      }
+      dim = -1;
+    }
+    out.push(dim);
+  }
+  return out;
+}
+function raggedTensorToTensorImpl(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes) {
+  return new RaggedTensorToTensorOp(shape, shapesShape, values, valuesShape, valuesDType, defaultValue, defaultValueShape, rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes).compute();
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Range_impl.js
+function rangeImpl(start, stop, step5, dtype) {
+  const sameStartStop = start === stop;
+  const increasingRangeNegativeStep = start < stop && step5 < 0;
+  const decreasingRangePositiveStep = stop < start && step5 > 1;
+  if (sameStartStop || increasingRangeNegativeStep || decreasingRangePositiveStep) {
+    return util_exports.makeZerosTypedArray(0, dtype);
+  }
+  const numElements = Math.abs(Math.ceil((stop - start) / step5));
+  const values = util_exports.makeZerosTypedArray(numElements, dtype);
+  if (stop < start && step5 === 1) {
+    step5 = -1;
+  }
+  values[0] = start;
+  for (let i = 1; i < values.length; i++) {
+    values[i] = values[i - 1] + step5;
+  }
+  return values;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Rsqrt.js
+var rsqrtImpl = createSimpleUnaryImpl((xi) => 1 / Math.sqrt(xi));
+var rsqrt2 = unaryKernelFuncFromImpl(Rsqrt, rsqrtImpl);
+var rsqrtConfig = {
+  kernelName: Rsqrt,
+  backendName: "cpu",
+  kernelFunc: rsqrt2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Scatter_impl.js
+function scatterImpl(indices, updates, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, defaultValue, sumDupeIndices) {
+  const flattenShape = [outputSize / sliceSize, sliceSize];
+  const indicesData = indices.values;
+  const updatesData = updates.values;
+  if (outputSize === 0) {
+    return buffer(shape, updates.dtype);
+  }
+  const outBuf = buffer(flattenShape, updates.dtype);
+  if (typeof defaultValue === "string") {
+    outBuf.values.fill(defaultValue);
+  } else if (typeof defaultValue === "number") {
+    outBuf.values.fill(defaultValue);
+  } else if (typeof defaultValue === "boolean") {
+    outBuf.values.fill(+defaultValue);
+  }
+  for (let i = 0; i < numUpdates; i++) {
+    const index = [];
+    let flattenIndex = 0;
+    for (let j = 0; j < sliceRank; j++) {
+      const dim = indicesData[i * sliceRank + j];
+      index.push(dim);
+      flattenIndex += dim * strides[j];
+    }
+    if (flattenIndex < 0 || flattenIndex >= outputSize / sliceSize) {
+      throw new Error(`Invalid indices: ${index} does not index into ${shape}`);
+    }
+    for (let k = 0; k < sliceSize; k++) {
+      if (sumDupeIndices) {
+        outBuf.values[flattenIndex * sliceSize + k] += updatesData[i * sliceSize + k];
+      } else {
+        outBuf.values[flattenIndex * sliceSize + k] = updates.rank === 0 ? updatesData[0] : updatesData[i * sliceSize + k];
+      }
+    }
+  }
+  return outBuf;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sigmoid.js
+var sigmoidImpl = createSimpleUnaryImpl((xi) => 1 / (1 + Math.exp(-xi)));
+var sigmoid2 = unaryKernelFunc(Sigmoid, (xi) => 1 / (1 + Math.exp(-xi)));
+var sigmoidConfig = {
+  kernelName: Sigmoid,
+  backendName: "cpu",
+  kernelFunc: sigmoid2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Slice.js
+function sliceImpl(vals, begin, size, shape, dtype) {
+  const isContinous = slice_util_exports.isSliceContinous(shape, begin, size);
+  const length = util_exports.sizeFromShape(size);
+  const xStrides = util_exports.computeStrides(shape);
+  if (isContinous) {
+    const flatOffset = slice_util_exports.computeFlatOffset(begin, xStrides);
+    if (dtype === "string") {
+      return vals.slice(flatOffset, flatOffset + length);
+    }
+    return vals.subarray(flatOffset, flatOffset + length);
+  }
+  const decodedData = dtype === "string" ? backend_util_exports.fromUint8ToStringArray(vals) : vals;
+  const inBuf = buffer(shape, dtype, decodedData);
+  const outBuf = buffer(size, dtype);
+  for (let i = 0; i < outBuf.size; ++i) {
+    const outLoc = outBuf.indexToLoc(i);
+    const inLoc = outLoc.map((idx, j) => idx + begin[j]);
+    outBuf.set(inBuf.get(...inLoc), ...outLoc);
+  }
+  if (dtype === "string") {
+    return backend_util_exports.fromStringArrayToUint8(outBuf.values);
+  }
+  return outBuf.values;
+}
+function slice2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, size } = attrs;
+  assertNotComplex(x, "slice");
+  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);
+  slice_util_exports.assertParamsValid(x, $begin, $size);
+  const vals = backend2.data.get(x.dataId).values;
+  const outVals = sliceImpl(vals, $begin, $size, x.shape, x.dtype);
+  return backend2.makeTensorInfo($size, x.dtype, outVals);
+}
+var sliceConfig = {
+  kernelName: Slice,
+  backendName: "cpu",
+  kernelFunc: slice2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseFillEmptyRows_impl.js
+function sparseFillEmptyRowsImpl(indices, indicesShape, indicesDType, values, valuesDType, denseShape, defaultValue) {
+  const indicesCount = indicesShape[0];
+  const denseRows = denseShape[0];
+  const emptyRowIndicator = new Array(denseRows);
+  const reverseIndexMap = new Array(indicesCount);
+  const rank = indicesShape[1];
+  if (denseRows === 0) {
+    if (indicesCount !== 0) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(indicesCount));
+    }
+    const outputIndices = util_exports.getArrayFromDType(indicesDType, 0);
+    const outputValues = util_exports.getArrayFromDType(valuesDType, 0);
+    return [
+      outputIndices,
+      [0, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  }
+  let rowsAreOrdered = true;
+  let lastIndicesRow = 0;
+  const csrOffset = new Array(denseRows).fill(0);
+  for (let i = 0; i < indicesCount; ++i) {
+    const row = indices[i * rank];
+    if (row < 0) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(i, row));
+    }
+    if (row >= denseRows) {
+      throw new Error(backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(i, row, denseRows));
+    }
+    ++csrOffset[row];
+    rowsAreOrdered = rowsAreOrdered && row >= lastIndicesRow;
+    lastIndicesRow = row;
+  }
+  let allRowsFull = true;
+  for (let row = 0; row < denseRows; ++row) {
+    const rowEmpty = csrOffset[row] === 0;
+    emptyRowIndicator[row] = rowEmpty;
+    allRowsFull = allRowsFull && !rowEmpty;
+    csrOffset[row] = Math.max(csrOffset[row], 1);
+    if (row > 0) {
+      csrOffset[row] += csrOffset[row - 1];
+    }
+  }
+  if (allRowsFull && rowsAreOrdered) {
+    const outputIndices = indices;
+    const outputValues = values;
+    for (let i = 0; i < indicesCount; ++i) {
+      reverseIndexMap[i] = i;
+    }
+    return [
+      outputIndices,
+      [indicesCount, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  } else {
+    const fullIndicesCount = csrOffset[denseRows - 1];
+    const outputIndices = util_exports.getArrayFromDType(indicesDType, fullIndicesCount * rank);
+    const outputValues = util_exports.getArrayFromDType(valuesDType, fullIndicesCount);
+    const filledCount = new Array(denseRows).fill(0);
+    for (let i = 0; i < indicesCount; ++i) {
+      const row = indices[i * rank];
+      const offset = filledCount[row];
+      const outputI = (row === 0 ? 0 : csrOffset[row - 1]) + offset;
+      filledCount[row]++;
+      for (let j = 0; j < rank; ++j) {
+        outputIndices[outputI * rank + j] = indices[i * rank + j];
+      }
+      outputValues[outputI] = values[i];
+      reverseIndexMap[i] = outputI;
+    }
+    for (let row = 0; row < denseRows; ++row) {
+      const rowCount = filledCount[row];
+      if (rowCount === 0) {
+        const startingIndex = row === 0 ? 0 : csrOffset[row - 1];
+        outputIndices[startingIndex * rank + 0] = row;
+        for (let col = 1; col < rank; ++col) {
+          outputIndices[startingIndex * rank + col] = 0;
+        }
+        outputValues[startingIndex] = defaultValue;
+      }
+    }
+    return [
+      outputIndices,
+      [fullIndicesCount, rank],
+      outputValues,
+      emptyRowIndicator,
+      reverseIndexMap
+    ];
+  }
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseReshape_impl.js
+function sparseReshapeImpl(inputIndices, inputIndicesShape, inputDType, inputShape, targetShape) {
+  const denseSize = util_exports.sizeFromShape(inputShape);
+  const nnz = inputIndicesShape[0];
+  const outputRank = targetShape.length;
+  const outputShape = [];
+  let product = 1;
+  let unknownIndex = -1;
+  for (let d = 0; d < outputRank; ++d) {
+    const size = targetShape[d];
+    if (size === -1) {
+      if (unknownIndex !== -1) {
+        throw new Error(backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(unknownIndex, d));
+      }
+      unknownIndex = d;
+      outputShape.push(1);
+    } else {
+      if (size < 0) {
+        throw new Error(backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(d, size));
+      }
+      product *= size;
+      outputShape.push(size);
+    }
+  }
+  if (unknownIndex !== -1) {
+    if (product <= 0) {
+      throw new Error(backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());
+    }
+    const missing = Math.trunc(denseSize / product);
+    if (product * missing !== denseSize) {
+      throw new Error(backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShape, outputShape));
+    }
+    outputShape[unknownIndex] = missing;
+  }
+  const outputSize = util_exports.sizeFromShape(outputShape);
+  if (outputSize !== denseSize) {
+    throw new Error(backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShape, outputShape));
+  }
+  const inputRank = inputShape.length;
+  const inputStrides = [];
+  if (inputRank > 0) {
+    inputStrides[inputRank - 1] = 1;
+    for (let d = inputRank - 2; d >= 0; --d) {
+      inputStrides[d] = inputStrides[d + 1] * inputShape[d + 1];
+    }
+  }
+  const outputStrides = [];
+  if (outputRank > 0) {
+    outputStrides[outputRank - 1] = 1;
+    for (let d = outputRank - 2; d >= 0; --d) {
+      outputStrides[d] = outputStrides[d + 1] * outputShape[d + 1];
+    }
+  }
+  const newIndices = util_exports.getArrayFromDType(inputDType, nnz * outputRank);
+  for (let i = 0; i < nnz; ++i) {
+    let id = 0;
+    for (let j = 0; j < inputRank; ++j) {
+      id += inputIndices[i * inputRank + j] * inputStrides[j];
+    }
+    for (let j = 0; j < outputRank; ++j) {
+      newIndices[i * outputRank + j] = Math.trunc(id / outputStrides[j]);
+      id %= outputStrides[j];
+    }
+  }
+  return [newIndices, [nnz, outputRank], outputShape];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentReduction_impl.js
+function sparseSegmentReductionImpl(input2, inputShape, inputDType, indices, segmentIds, isMean = false, defaultValue = 0) {
+  const numIndices = indices.length;
+  const inputFlat = [inputShape[0], input2.length / inputShape[0]];
+  const numCol = inputFlat[1];
+  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIds[numIndices - 1] + 1 : 0;
+  const outputRows = lastSegmentIdPlusOne;
+  if (outputRows < 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  const outputShape = inputShape.slice();
+  outputShape[0] = outputRows;
+  const outputLength = outputShape.reduce((product, value) => product * value, 1);
+  const output = util_exports.getArrayFromDType(inputDType, outputLength);
+  if (numIndices === 0) {
+    if (outputRows > 0) {
+      output.fill(defaultValue);
+    }
+    return [output, outputShape];
+  }
+  if (outputRows <= 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  let start = 0, end = 1;
+  let uninitializedIndex = 0;
+  let outIndex = segmentIds[start];
+  while (true) {
+    let nextIndex = 0;
+    if (end < numIndices) {
+      nextIndex = segmentIds[end];
+      if (outIndex === nextIndex) {
+        ++end;
+        continue;
+      }
+      if (outIndex >= nextIndex) {
+        throw new Error(backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage());
+      }
+    }
+    if (outIndex < 0 || outIndex >= outputRows) {
+      throw new Error(backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(outIndex, outputRows));
+    }
+    if (outIndex > uninitializedIndex) {
+      output.fill(defaultValue, uninitializedIndex * numCol, outIndex * numCol);
+    }
+    for (let i = start; i < end; ++i) {
+      const index = indices[i];
+      if (index < 0 || index >= inputFlat[0]) {
+        throw new Error(backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(i, indices[i], inputFlat[0]));
+      }
+      for (let j = 0; j < numCol; j++) {
+        output[outIndex * numCol + j] += input2[index * numCol + j];
+      }
+    }
+    if (isMean) {
+      for (let j = 0; j < numCol; j++) {
+        output[outIndex * numCol + j] /= end - start;
+      }
+    }
+    start = end;
+    ++end;
+    uninitializedIndex = outIndex + 1;
+    outIndex = nextIndex;
+    if (end > numIndices) {
+      break;
+    }
+  }
+  if (uninitializedIndex < outputRows) {
+    output.fill(defaultValue, uninitializedIndex * numCol, outputRows * numCol);
+  }
+  return [output, outputShape];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sqrt.js
+var sqrtImpl = createSimpleUnaryImpl((xi) => Math.sqrt(xi));
+var sqrt2 = unaryKernelFunc(Sqrt, (xi) => Math.sqrt(xi));
+var sqrtConfig = {
+  kernelName: Sqrt,
+  backendName: "cpu",
+  kernelFunc: sqrt2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SquaredDifference.js
+var squaredDifferenceImpl = createSimpleBinaryKernelImpl((a, b) => {
+  const diff = a - b;
+  return diff * diff;
+});
+var squaredDifference2 = binaryKernelFunc(SquaredDifference, squaredDifferenceImpl);
+var squaredDifferenceConfig = {
+  kernelName: SquaredDifference,
+  backendName: "cpu",
+  kernelFunc: squaredDifference2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StridedSlice_impl.js
+function stridedSliceImpl(outShape, xBuf, strides, begin) {
+  const outBuf = buffer(outShape, xBuf.dtype);
+  for (let i = 0; i < outBuf.size; i++) {
+    const loc = outBuf.indexToLoc(i);
+    const newLoc = new Array(loc.length);
+    for (let j = 0; j < newLoc.length; j++) {
+      newLoc[j] = loc[j] * strides[j] + begin[j];
+    }
+    outBuf.set(xBuf.get(...newLoc), ...loc);
+  }
+  return outBuf;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringNGrams_impl.js
+var StringNGramsOp = class {
+  constructor(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+    this.separator = util_exports.encodeString(separator);
+    this.nGramWidths = nGramWidths;
+    this.leftPad = util_exports.encodeString(leftPad);
+    this.rightPad = util_exports.encodeString(rightPad2);
+    this.padWidth = padWidth;
+    this.preserveShort = preserveShortSequences;
+  }
+  getPadWidth(nGramWidth) {
+    return Math.min(this.padWidth < 0 ? nGramWidth - 1 : this.padWidth, nGramWidth - 1);
+  }
+  getNumNGrams(length, nGramWidth) {
+    const padWidth = this.getPadWidth(nGramWidth);
+    return Math.max(0, length + 2 * padWidth - nGramWidth + 1);
+  }
+  createNGrams(data, splitIndex, output, outputStartIndex, numNGrams, nGramWidth) {
+    for (let nGramIndex = 0; nGramIndex < numNGrams; ++nGramIndex) {
+      const padWidth = this.getPadWidth(nGramWidth);
+      const leftPadding = Math.max(0, padWidth - nGramIndex);
+      const rightPadding = Math.max(0, padWidth - (numNGrams - (nGramIndex + 1)));
+      const numTokens = nGramWidth - (leftPadding + rightPadding);
+      const dataStartIndex = splitIndex + (leftPadding > 0 ? 0 : nGramIndex - padWidth);
+      let nGramSize = 0;
+      nGramSize += leftPadding * this.leftPad.length;
+      for (let n = 0; n < numTokens; ++n) {
+        nGramSize += data[dataStartIndex + n].length;
+      }
+      nGramSize += rightPadding * this.rightPad.length;
+      const numSeparators = leftPadding + rightPadding + numTokens - 1;
+      nGramSize += numSeparators * this.separator.length;
+      output[outputStartIndex + nGramIndex] = new Uint8Array(nGramSize);
+      const nGram = output[outputStartIndex + nGramIndex];
+      let nextNGramIndex = 0;
+      const appendToNGram = (str) => str.forEach((value) => nGram[nextNGramIndex++] = value);
+      for (let n = 0; n < leftPadding; ++n) {
+        appendToNGram(this.leftPad);
+        appendToNGram(this.separator);
+      }
+      for (let n = 0; n < numTokens - 1; ++n) {
+        appendToNGram(data[dataStartIndex + n]);
+        appendToNGram(this.separator);
+      }
+      if (numTokens > 0) {
+        appendToNGram(data[dataStartIndex + numTokens - 1]);
+        for (let n = 0; n < rightPadding; ++n) {
+          appendToNGram(this.separator);
+          appendToNGram(this.rightPad);
+        }
+      } else {
+        for (let n = 0; n < rightPadding - 1; ++n) {
+          appendToNGram(this.rightPad);
+          appendToNGram(this.separator);
+        }
+        appendToNGram(this.rightPad);
+      }
+    }
+  }
+  compute(data, splits) {
+    const inputDataSize = data.length;
+    const splitsSize = splits.length;
+    if (splitsSize > 0) {
+      let prevSplit = splits[0];
+      if (prevSplit !== 0) {
+        throw new Error(`First split value must be 0, got ${prevSplit}`);
+      }
+      for (let i = 1; i < splitsSize; ++i) {
+        let validSplits = splits[i] >= prevSplit;
+        validSplits = validSplits && splits[i] <= inputDataSize;
+        if (!validSplits) {
+          throw new Error(`Invalid split value ${splits[i]}, must be in [${prevSplit}, ${inputDataSize}]`);
+        }
+        prevSplit = splits[i];
+      }
+      if (prevSplit !== inputDataSize) {
+        throw new Error(`Last split value must be data size. Expected ${inputDataSize}, got ${prevSplit}`);
+      }
+    }
+    const numBatchItems = splitsSize - 1;
+    const nGramsSplits = util_exports.getArrayFromDType("int32", splitsSize);
+    if (inputDataSize === 0 || splitsSize === 0) {
+      const empty = new Array(inputDataSize);
+      for (let i = 0; i <= numBatchItems; ++i) {
+        nGramsSplits[i] = 0;
+      }
+      return [empty, nGramsSplits];
+    }
+    nGramsSplits[0] = 0;
+    for (let i = 1; i <= numBatchItems; ++i) {
+      const length = splits[i] - splits[i - 1];
+      let numNGrams = 0;
+      this.nGramWidths.forEach((nGramWidth) => {
+        numNGrams += this.getNumNGrams(length, nGramWidth);
+      });
+      if (this.preserveShort && length > 0 && numNGrams === 0) {
+        numNGrams = 1;
+      }
+      nGramsSplits[i] = nGramsSplits[i - 1] + numNGrams;
+    }
+    const nGrams = new Array(nGramsSplits[numBatchItems]);
+    for (let i = 0; i < numBatchItems; ++i) {
+      const splitIndex = splits[i];
+      let outputStartIdx = nGramsSplits[i];
+      this.nGramWidths.forEach((nGramWidth) => {
+        const length = splits[i + 1] - splits[i];
+        const numNGrams = this.getNumNGrams(length, nGramWidth);
+        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);
+        outputStartIdx += numNGrams;
+      });
+      if (this.preserveShort && outputStartIdx === nGramsSplits[i]) {
+        const dataLength = splits[i + 1] - splits[i];
+        if (dataLength === 0) {
+          continue;
+        }
+        const nGramWidth = dataLength + 2 * this.padWidth;
+        const numNGrams = 1;
+        this.createNGrams(data, splitIndex, nGrams, outputStartIdx, numNGrams, nGramWidth);
+      }
+    }
+    return [nGrams, nGramsSplits];
+  }
+};
+function stringNGramsImpl(data, dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences) {
+  return new StringNGramsOp(separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences).compute(data, dataSplits);
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringSplit_impl.js
+function split3(str, delimiters, skipEmpty, result) {
+  if (!str.length) {
+    return;
+  }
+  if (delimiters.length === 0) {
+    for (let i = 0; i < str.length; ++i) {
+      result.push(str.subarray(i, i + 1));
+    }
+    return;
+  }
+  if (delimiters.length === 1) {
+    const delimiter = delimiters[0];
+    let f = str.indexOf(delimiter);
+    while (f !== -1) {
+      const token = str.subarray(0, f);
+      if (!skipEmpty || token.length !== 0) {
+        result.push(token);
+      }
+      str = str.subarray(f + 1);
+      f = str.indexOf(delimiter);
+    }
+    if (!skipEmpty || str.length !== 0) {
+      result.push(str);
+    }
+    return;
+  }
+  let tokenStart = 0;
+  for (let i = 0; i < str.length + 1; i++) {
+    if (i === str.length || delimiters.indexOf(str[i]) !== -1) {
+      const token = str.subarray(tokenStart, i);
+      if (!skipEmpty || token.length !== 0) {
+        result.push(token);
+      }
+      tokenStart = i + 1;
+    }
+  }
+}
+function stringSplitImpl(input2, delimiter, skipEmpty) {
+  const batchSize = input2.length;
+  const tokens = [];
+  let outputSize = 0;
+  let maxNumEntries = 0;
+  const numIndices = new Array(batchSize);
+  for (let i = 0; i < batchSize; ++i) {
+    const prevTokensLength = tokens.length;
+    split3(input2[i], delimiter, skipEmpty, tokens);
+    const nEntries = tokens.length - prevTokensLength;
+    numIndices[i] = nEntries;
+    outputSize += nEntries;
+    maxNumEntries = Math.max(maxNumEntries, nEntries);
+  }
+  const indices = util_exports.getArrayFromDType("int32", outputSize * 2);
+  const values = new Array(outputSize);
+  const shape = [batchSize, maxNumEntries];
+  let c = 0;
+  for (let i = 0; i < batchSize; ++i) {
+    for (let j = 0; j < numIndices[i]; ++j) {
+      indices[c * 2] = i;
+      indices[c * 2 + 1] = j;
+      values[c] = tokens[c];
+      ++c;
+    }
+  }
+  return [indices, values, shape];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringToHashBucketFast_impl.js
+function stringToHashBucketFastImpl(input2, numBuckets) {
+  const output = util_exports.getArrayFromDType("int32", input2.length);
+  for (let i = 0; i < input2.length; ++i) {
+    output[i] = util_exports.fingerPrint64(input2[i]).modulo(numBuckets).getLowBitsUnsigned();
+  }
+  return output;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sub.js
+var subImpl = createSimpleBinaryKernelImpl((aValue, bValue) => aValue - bValue);
+var subComplexImpl = createComplexBinaryKernelImpl((aReal, aImag, bReal, bImag) => {
+  return { real: aReal - bReal, imag: aImag - bImag };
+});
+var sub2 = binaryKernelFunc(Sub, subImpl, subComplexImpl);
+var subConfig = {
+  kernelName: Sub,
+  backendName: "cpu",
+  kernelFunc: sub2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tile_impl.js
+function tileImpl(xBuf, reps) {
+  const newShape = new Array(xBuf.rank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = xBuf.shape[i] * reps[i];
+  }
+  const result = buffer(newShape, xBuf.dtype);
+  for (let i = 0; i < result.values.length; ++i) {
+    const newLoc = result.indexToLoc(i);
+    const originalLoc = new Array(xBuf.rank);
+    for (let j = 0; j < originalLoc.length; j++) {
+      originalLoc[j] = newLoc[j] % xBuf.shape[j];
+    }
+    const originalIndex = xBuf.locToIndex(originalLoc);
+    result.values[i] = xBuf.values[originalIndex];
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/TopK_impl.js
+var comparePair = (a, b) => {
+  const valueDiff = b.value - a.value;
+  return valueDiff === 0 ? a.index - b.index : valueDiff;
+};
+function select(array2, k, left = 0, right = array2.length - 1) {
+  while (right > left) {
+    if (right - left > 600) {
+      const n = right - left + 1;
+      const i2 = k - left + 1;
+      const z = Math.log(n);
+      const s = 0.5 * Math.exp(2 * z / 3);
+      const sd = 0.5 * Math.sqrt(z * s * (n - s) / n) * Math.sign(i2 - n / 2);
+      const newLeft = Math.max(left, Math.floor(k - i2 * s / n + sd));
+      const newRight = Math.min(right, Math.floor(k + (n - i2) * s / n + sd));
+      select(array2, k, newLeft, newRight);
+    }
+    const t = array2[k];
+    let i = left;
+    let j = right;
+    util_exports.swap(array2, left, k);
+    if (comparePair(array2[right], t) > 0) {
+      util_exports.swap(array2, left, right);
+    }
+    while (i < j) {
+      util_exports.swap(array2, i, j);
+      i++;
+      j--;
+      while (comparePair(array2[i], t) < 0) {
+        i = i + 1;
+      }
+      while (comparePair(array2[j], t) > 0) {
+        j = j - 1;
+      }
+    }
+    if (comparePair(array2[left], t) === 0) {
+      util_exports.swap(array2, left, j);
+    } else {
+      j = j + 1;
+      util_exports.swap(array2, j, right);
+    }
+    if (j <= k) {
+      left = j + 1;
+    }
+    if (k <= j) {
+      right = j - 1;
+    }
+  }
+}
+function topKImpl(x, xShape, xDtype, k, sorted) {
+  const lastDim = xShape[xShape.length - 1];
+  const [batch, size] = [x.length / lastDim, lastDim];
+  const allTopKVals = util_exports.getTypedArrayFromDType(xDtype, batch * k);
+  const allTopKIndices = util_exports.getTypedArrayFromDType("int32", batch * k);
+  for (let b = 0; b < batch; b++) {
+    const offset = b * size;
+    const vals = x.subarray(offset, offset + size);
+    let valAndInd = new Array(vals.length);
+    vals.forEach((value, index) => valAndInd[index] = { value, index });
+    if (k < valAndInd.length) {
+      select(valAndInd, k);
+      valAndInd = valAndInd.slice(0, k);
+    }
+    if (sorted) {
+      valAndInd.sort(comparePair);
+    }
+    const outOffset = b * k;
+    const topKVals = allTopKVals.subarray(outOffset, outOffset + k);
+    const topKIndices = allTopKIndices.subarray(outOffset, outOffset + k);
+    for (let i = 0; i < k; i++) {
+      topKVals[i] = valAndInd[i].value;
+      topKIndices[i] = valAndInd[i].index;
+    }
+  }
+  const outputShape = xShape.slice();
+  outputShape[outputShape.length - 1] = k;
+  return [
+    buffer(outputShape, xDtype, allTopKVals),
+    buffer(outputShape, "int32", allTopKIndices)
+  ];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unique_impl.js
+function uniqueImpl(values, axis, shape, dtype) {
+  const $axis = util_exports.parseAxisParam(axis, shape)[0];
+  const newShape = [1, shape[0], 1];
+  for (let i = 0; i < $axis; i++) {
+    newShape[0] *= shape[i];
+  }
+  newShape[1] = shape[$axis];
+  for (let i = $axis + 1; i < shape.length; i++) {
+    newShape[2] *= shape[i];
+  }
+  const uniqueElements = {};
+  const indices = new Int32Array(shape[$axis]);
+  const inputBuffer = new TensorBuffer(newShape, dtype, values);
+  const uniqueIndices = [];
+  const is1DTensor = newShape[0] === 1 && newShape[2] === 1;
+  for (let i = 0; i < shape[$axis]; i++) {
+    let element;
+    if (is1DTensor) {
+      element = values[i].toString();
+    } else {
+      const axisValues = [];
+      for (let m = 0; m < newShape[0]; m++) {
+        for (let n = 0; n < newShape[2]; n++) {
+          axisValues.push(inputBuffer.get(m, i, n));
+        }
+      }
+      element = axisValues.join(",");
+    }
+    if (uniqueElements[element] !== void 0) {
+      indices[i] = uniqueElements[element];
+    } else {
+      const uniqueIndex = Object.keys(uniqueElements).length;
+      uniqueElements[element] = uniqueIndex;
+      indices[i] = uniqueIndex;
+      uniqueIndices.push(i);
+    }
+  }
+  const outputTmpShape = newShape.slice();
+  outputTmpShape[1] = Object.keys(uniqueElements).length;
+  const outputBuffer = new TensorBuffer(outputTmpShape, dtype);
+  uniqueIndices.forEach((uniqueElementIndex, i) => {
+    for (let m = 0; m < newShape[0]; m++) {
+      for (let n = 0; n < newShape[2]; n++) {
+        outputBuffer.set(inputBuffer.get(m, uniqueElementIndex, n), m, i, n);
+      }
+    }
+  });
+  const outputShape = shape.slice();
+  outputShape[$axis] = outputTmpShape[1];
+  return {
+    outputValues: outputBuffer.values,
+    outputShape,
+    indices
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/base.js
+registerBackend("cpu", () => new MathBackendCPU(), 1);
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Elu.js
+var elu4 = unaryKernelFunc(Elu, (xi) => xi >= 0 ? xi : Math.exp(xi) - 1);
+var eluConfig = {
+  kernelName: Elu,
+  backendName: "cpu",
+  kernelFunc: elu4
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LeakyRelu.js
+function leakyRelu2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { alpha } = attrs;
+  assertNotComplex([x], "leakyRelu");
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xVals = backend2.data.get(x.dataId).values;
+  const outVals = util_exports.getTypedArrayFromDType("float32", xSize);
+  for (let i = 0; i < xVals.length; i++) {
+    outVals[i] = xVals[i] < 0 ? alpha * xVals[i] : xVals[i];
+  }
+  return backend2.makeTensorInfo(x.shape, "float32", outVals);
+}
+var leakyReluConfig = {
+  kernelName: LeakyRelu,
+  backendName: "cpu",
+  kernelFunc: leakyRelu2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Prelu.js
+var preluImpl = createSimpleBinaryKernelImpl((xValue, aValue) => xValue < 0 ? aValue * xValue : xValue);
+function prelu3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  assertNotComplex([x, alpha], "prelu");
+  const aVals = backend2.data.get(x.dataId).values;
+  const bVals = backend2.data.get(alpha.dataId).values;
+  const [resultData, resultShape] = preluImpl(x.shape, alpha.shape, aVals, bVals, "float32");
+  return backend2.makeTensorInfo(resultShape, "float32", resultData);
+}
+var preluConfig = {
+  kernelName: Prelu,
+  backendName: "cpu",
+  kernelFunc: prelu3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Relu.js
+var relu2 = unaryKernelFunc(Relu, (xi) => Math.max(0, xi));
+var reluConfig = {
+  kernelName: Relu,
+  backendName: "cpu",
+  kernelFunc: relu2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Relu6.js
+var relu62 = unaryKernelFunc(Relu6, (xi) => Math.min(Math.max(0, xi), 6));
+var relu6Config = {
+  kernelName: Relu6,
+  backendName: "cpu",
+  kernelFunc: relu62
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/fused_utils.js
+function applyActivation2(backend2, x, activation2, preluActivationWeights, leakyreluAlpha) {
+  if (activation2 === "linear") {
+    return identity2({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "relu") {
+    return relu2({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "elu") {
+    return elu4({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "relu6") {
+    return relu62({ inputs: { x }, backend: backend2 });
+  } else if (activation2 === "prelu") {
+    return prelu3({ inputs: { x, alpha: preluActivationWeights }, backend: backend2 });
+  } else if (activation2 === "leakyrelu") {
+    return leakyRelu2({ inputs: { x }, backend: backend2, attrs: { alpha: leakyreluAlpha } });
+  } else if (activation2 === "sigmoid") {
+    return sigmoid2({ inputs: { x }, backend: backend2 });
+  }
+  throw new Error(`Activation ${activation2} has not been implemented for the CPU backend.`);
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reshape.js
+function reshape3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  const $xSize = util_exports.sizeFromShape($shape);
+  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);
+  backend2.incRef(x.dataId);
+  const xData = backend2.data.get(x.dataId);
+  if (xData.complexTensorInfos != null) {
+    const real4 = xData.complexTensorInfos.real;
+    const imag4 = xData.complexTensorInfos.imag;
+    real4.shape = $shape;
+    imag4.shape = $shape;
+  }
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig = {
+  kernelName: Reshape,
+  backendName: "cpu",
+  kernelFunc: reshape3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchMatMul.js
+function batchMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  assertNotComplex([a, b], "matMul");
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape3({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape3({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];
+  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];
+  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const a3dValues = backend2.data.get(a3d.dataId).values;
+  const b3dValues = backend2.data.get(b3d.dataId).values;
+  const a3dStrides = util_exports.computeStrides(a3d.shape);
+  const b3dStrides = util_exports.computeStrides(b3d.shape);
+  const [aBatch, aOuterStep, aInnerStep] = transposeA ? [a3dStrides[0], 1, a3dStrides[1]] : [a3dStrides[0], a3dStrides[1], 1];
+  const [bInnerStep, bOuterStep, bBatch] = transposeB ? [1, b3dStrides[1], b3dStrides[0]] : [b3dStrides[1], 1, b3dStrides[0]];
+  const size = leftDim * rightDim;
+  const result = buffer([batchDim, leftDim, rightDim], a3d.dtype);
+  const resVals = result.values;
+  const blockSize = backend2.blockSize;
+  for (let bi = 0; bi < batchDim; bi++) {
+    for (let i0 = 0; i0 < leftDim; i0 += blockSize) {
+      for (let j0 = 0; j0 < rightDim; j0 += blockSize) {
+        for (let k02 = 0; k02 < sharedDim; k02 += blockSize) {
+          const iBlock = Math.min(i0 + blockSize, leftDim);
+          const jBlock = Math.min(j0 + blockSize, rightDim);
+          const kBlock = Math.min(k02 + blockSize, sharedDim);
+          for (let i = i0; i < iBlock; i++) {
+            for (let j = j0; j < jBlock; j++) {
+              let sum6 = 0;
+              for (let k = k02; k < kBlock; k++) {
+                const batchOffsetA = Math.min(bi, batchDimA - 1) * aBatch;
+                const batchOffsetB = Math.min(bi, batchDimB - 1) * bBatch;
+                const aVal = a3dValues[batchOffsetA + i * aOuterStep + k * aInnerStep];
+                const bVal = b3dValues[k * bInnerStep + j * bOuterStep + batchOffsetB];
+                sum6 += aVal * bVal;
+              }
+              resVals[bi * size + (i * rightDim + j)] += sum6;
+            }
+          }
+        }
+      }
+    }
+  }
+  backend2.disposeIntermediateTensorInfo(a3d);
+  backend2.disposeIntermediateTensorInfo(b3d);
+  return backend2.makeTensorInfo(outShape, result.dtype, result.values);
+}
+var batchMatMulConfig = {
+  kernelName: BatchMatMul,
+  backendName: "cpu",
+  kernelFunc: batchMatMul
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/_FusedMatMul.js
+function _fusedMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  let current;
+  let addRes;
+  let activationRes;
+  const intermediates = [];
+  const matMulRes = batchMatMul({ inputs: { a, b }, attrs: { transposeA, transposeB }, backend: backend2 });
+  current = matMulRes;
+  if (bias) {
+    addRes = add4({ inputs: { a: current, b: bias }, backend: backend2 });
+    intermediates.push(current);
+    current = addRes;
+  }
+  if (activation2) {
+    activationRes = applyActivation2(backend2, current, activation2, preluActivationWeights, leakyreluAlpha);
+    intermediates.push(current);
+    current = activationRes;
+  }
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return current;
+}
+var _fusedMatMulConfig = {
+  kernelName: _FusedMatMul,
+  backendName: "cpu",
+  kernelFunc: _fusedMatMul
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Acos.js
+var acos2 = unaryKernelFunc(Acos, (xi) => Math.acos(xi));
+var acosConfig = {
+  kernelName: Acos,
+  backendName: "cpu",
+  kernelFunc: acos2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Acosh.js
+var acosh2 = unaryKernelFunc(Acosh, (xi) => Math.acosh(xi));
+var acoshConfig = {
+  kernelName: Acosh,
+  backendName: "cpu",
+  kernelFunc: acosh2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AddN.js
+function addN2(args) {
+  const { inputs, backend: backend2 } = args;
+  const tensors = inputs;
+  assertNotComplex(inputs, "addN");
+  const vals = tensors.map((t) => backend2.data.get(t.dataId).values);
+  const outBuf = buffer(tensors[0].shape, tensors[0].dtype);
+  const outVals = outBuf.values;
+  for (let i = 0; i < tensors.length; i++) {
+    const currVals = vals[i];
+    for (let j = 0; j < outVals.length; j++) {
+      outVals[j] += currVals[j];
+    }
+  }
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var addNConfig = {
+  kernelName: AddN,
+  backendName: "cpu",
+  kernelFunc: addN2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/All.js
+function all2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "all");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let all5 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      all5 = all5 && value;
+    }
+    vals[i] = all5;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var allConfig = {
+  kernelName: All,
+  backendName: "cpu",
+  kernelFunc: all2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Any.js
+function any2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "any");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let anyVal = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      anyVal = anyVal || value;
+    }
+    vals[i] = anyVal;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var anyConfig = {
+  kernelName: Any,
+  backendName: "cpu",
+  kernelFunc: any2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ArgMax.js
+function argMax2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  assertNotComplex(x, "argMax");
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  axes = [axes[0]];
+  backend_util_exports.assertAxesAreInnerMostDims("argMax", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const outSize = util_exports.sizeFromShape(outShape);
+  const vals = util_exports.makeZerosTypedArray(outSize, "int32");
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let max6 = aVals[offset];
+    let maxIndex = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (value > max6) {
+        max6 = value;
+        maxIndex = j;
+      }
+    }
+    vals[i] = maxIndex;
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(outShape, "int32", vals);
+}
+var argMaxConfig = {
+  kernelName: ArgMax,
+  backendName: "cpu",
+  kernelFunc: argMax2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ArgMin.js
+function argMin2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  assertNotComplex(x, "argMin");
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  axes = [axes[0]];
+  backend_util_exports.assertAxesAreInnerMostDims("argMin", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const outSize = util_exports.sizeFromShape(outShape);
+  const vals = util_exports.makeZerosTypedArray(outSize, "int32");
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let min6 = aVals[offset];
+    let minIndex = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (value < min6) {
+        min6 = value;
+        minIndex = j;
+      }
+    }
+    vals[i] = minIndex;
+  }
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return backend2.makeTensorInfo(outShape, "int32", vals);
+}
+var argMinConfig = {
+  kernelName: ArgMin,
+  backendName: "cpu",
+  kernelFunc: argMin2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Asin.js
+var asin2 = unaryKernelFunc(Asin, (xi) => Math.asin(xi));
+var asinConfig = {
+  kernelName: Asin,
+  backendName: "cpu",
+  kernelFunc: asin2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Asinh.js
+var asinh2 = unaryKernelFunc(Asinh, (xi) => Math.asinh(xi));
+var asinhConfig = {
+  kernelName: Asinh,
+  backendName: "cpu",
+  kernelFunc: asinh2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atan.js
+var atan3 = unaryKernelFunc(Atan, (xi) => Math.atan(xi));
+var atanConfig = {
+  kernelName: Atan,
+  backendName: "cpu",
+  kernelFunc: atan3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atan2.js
+var atan2Impl = createSimpleBinaryKernelImpl((aValue, bValue) => Math.atan2(aValue, bValue));
+var atan22 = binaryKernelFunc(Atan2, atan2Impl);
+var atan2Config = {
+  kernelName: Atan2,
+  backendName: "cpu",
+  kernelFunc: atan22
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Atanh.js
+var atanh2 = unaryKernelFunc(Atanh, (xi) => Math.atanh(xi));
+var atanhConfig = {
+  kernelName: Atanh,
+  backendName: "cpu",
+  kernelFunc: atanh2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/pool_utils.js
+function pool2(xValues, xShape, dtype, strides, convInfo, poolType) {
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const initialValue = poolType === "max" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;
+  const output = buffer(convInfo.outShape, dtype);
+  const outputVals = output.values;
+  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3];
+  const outputRowStrides = convInfo.outShape[2] * convInfo.outShape[3];
+  const outputColStrides = convInfo.outShape[3];
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const outputBatchOffset = b * outputBatchStrides;
+    const inputBatchOffset = b * strides[0];
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+        const xRCorner = yR * strideHeight - padTop;
+        const xRMin = Math.max(0, xRCorner);
+        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);
+        const outputRowOffset = outputBatchOffset + yR * outputRowStrides;
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const xCCorner = yC * strideWidth - padLeft;
+          const xCMin = Math.max(0, xCCorner);
+          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);
+          let minMaxValue = initialValue;
+          let avgValue = 0;
+          let count2 = 0;
+          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {
+            const xROffset = inputBatchOffset + xR * strides[1];
+            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {
+              const xCOffset = xROffset + xC * strides[2];
+              const pixel = xValues[xCOffset + d];
+              if (poolType === "max" && pixel > minMaxValue) {
+                minMaxValue = pixel;
+              } else if (poolType === "avg") {
+                avgValue += pixel;
+                count2++;
+              }
+            }
+            if (isNaN(minMaxValue)) {
+              break;
+            }
+          }
+          const outputOffset = outputRowOffset + yC * outputColStrides + d;
+          outputVals[outputOffset] = poolType === "avg" ? avgValue / count2 : minMaxValue;
+        }
+      }
+    }
+  }
+  return output;
+}
+function maxPoolPositions(xValues, xShape, dtype, convInfo, flattenPositions = false, includeBatchInIndex = false) {
+  const maxPositions = buffer(convInfo.outShape, "int32");
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const xBuf = buffer(xShape, dtype, xValues);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+        const xRCorner = yR * strideHeight - padTop;
+        let xRMin = xRCorner;
+        while (xRMin < 0) {
+          xRMin += dilationHeight;
+        }
+        const xRMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRCorner);
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const xCCorner = yC * strideWidth - padLeft;
+          let xCMin = xCCorner;
+          while (xCMin < 0) {
+            xCMin += dilationWidth;
+          }
+          const xCMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xCCorner);
+          let maxValue = Number.NEGATIVE_INFINITY;
+          let maxPosition = -1;
+          for (let xR = xRMin; xR < xRMax; xR += dilationHeight) {
+            const wR = xR - xRCorner;
+            for (let xC = xCMin; xC < xCMax; xC += dilationWidth) {
+              const wC = xC - xCCorner;
+              const pixel = xBuf.get(b, xR, xC, d);
+              if (pixel > maxValue) {
+                maxValue = pixel;
+                if (flattenPositions) {
+                  maxPosition = includeBatchInIndex ? ((b * convInfo.inHeight + xR) * convInfo.inWidth + xC) * convInfo.inChannels + d : (xR * convInfo.inWidth + xC) * convInfo.inChannels + d;
+                } else {
+                  maxPosition = wR * effectiveFilterWidth + wC;
+                }
+              }
+            }
+          }
+          maxPositions.set(maxPosition, b, yR, yC, d);
+        }
+      }
+    }
+  }
+  return maxPositions;
+}
+function pool3d2(xValues, xShape, dtype, strides, convInfo, poolType) {
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = convInfo.padInfo.front;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  const initialValue = poolType === "max" ? Number.NEGATIVE_INFINITY : Number.POSITIVE_INFINITY;
+  const output = buffer(convInfo.outShape, dtype);
+  const outputVals = output.values;
+  const outputBatchStrides = convInfo.outShape[1] * convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];
+  const outputDepthStrides = convInfo.outShape[2] * convInfo.outShape[3] * convInfo.outShape[4];
+  const outputRowStrides = convInfo.outShape[3] * convInfo.outShape[4];
+  const outputColStrides = convInfo.outShape[4];
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    const outputBatchOffset = batch * outputBatchStrides;
+    const inputBatchOffset = batch * strides[0];
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {
+        const xDepthCorner = yDepth * strideDepth - padFront;
+        let xDepthMin = xDepthCorner;
+        while (xDepthMin < 0) {
+          xDepthMin += dilationDepth;
+        }
+        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);
+        const outputDepthOffset = outputBatchOffset + yDepth * outputDepthStrides;
+        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {
+          const xRowCorner = yRow * strideHeight - padTop;
+          let xRowMin = xRowCorner;
+          while (xRowMin < 0) {
+            xRowMin += dilationHeight;
+          }
+          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);
+          const outputRowOffset = outputDepthOffset + yRow * outputRowStrides;
+          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {
+            const xColCorner = yCol * strideWidth - padLeft;
+            let xColMin = xColCorner;
+            while (xColMin < 0) {
+              xColMin += dilationWidth;
+            }
+            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);
+            const outputColOffset = outputRowOffset + yCol * outputColStrides;
+            let minMaxValue = initialValue;
+            let avgValue = 0;
+            let count2 = 0;
+            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {
+              const xDepthOffset = inputBatchOffset + xDepth * strides[1];
+              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {
+                const xRowOffset = xDepthOffset + xRow * strides[2];
+                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {
+                  const xColOffset = xRowOffset + xCol * strides[3];
+                  const pixel = xValues[xColOffset + channel];
+                  if (poolType === "max" && pixel > minMaxValue) {
+                    minMaxValue = pixel;
+                  } else if (poolType === "avg") {
+                    avgValue += pixel;
+                    count2++;
+                  }
+                  if (isNaN(minMaxValue)) {
+                    break;
+                  }
+                }
+                if (isNaN(minMaxValue)) {
+                  break;
+                }
+              }
+              if (isNaN(minMaxValue)) {
+                break;
+              }
+            }
+            const outputOffset = outputColOffset + channel;
+            outputVals[outputOffset] = poolType === "avg" ? avgValue / count2 : minMaxValue;
+          }
+        }
+      }
+    }
+  }
+  return output;
+}
+function maxPool3dPositions(xBuf, convInfo) {
+  const maxPositions = buffer(convInfo.outShape, "int32");
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = convInfo.padInfo.front;
+  const padTop = convInfo.padInfo.top;
+  const padLeft = convInfo.padInfo.left;
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let yDepth = 0; yDepth < convInfo.outDepth; ++yDepth) {
+        const xDepthCorner = yDepth * strideDepth - padFront;
+        let xDepthMin = xDepthCorner;
+        while (xDepthMin < 0) {
+          xDepthMin += dilationDepth;
+        }
+        const xDepthMax = Math.min(convInfo.inDepth, effectiveFilterDepth + xDepthCorner);
+        for (let yRow = 0; yRow < convInfo.outHeight; ++yRow) {
+          const xRowCorner = yRow * strideHeight - padTop;
+          let xRowMin = xRowCorner;
+          while (xRowMin < 0) {
+            xRowMin += dilationHeight;
+          }
+          const xRowMax = Math.min(convInfo.inHeight, effectiveFilterHeight + xRowCorner);
+          for (let yCol = 0; yCol < convInfo.outWidth; ++yCol) {
+            const xColCorner = yCol * strideWidth - padLeft;
+            let xColMin = xColCorner;
+            while (xColMin < 0) {
+              xColMin += dilationWidth;
+            }
+            const xColMax = Math.min(convInfo.inWidth, effectiveFilterWidth + xColCorner);
+            let maxValue = Number.NEGATIVE_INFINITY;
+            let maxPosition = -1;
+            for (let xDepth = xDepthMin; xDepth < xDepthMax; xDepth += dilationDepth) {
+              const wDepth = xDepth - xDepthCorner;
+              for (let xRow = xRowMin; xRow < xRowMax; xRow += dilationHeight) {
+                const wRow = xRow - xRowCorner;
+                for (let xCol = xColMin; xCol < xColMax; xCol += dilationWidth) {
+                  const wCol = xCol - xColCorner;
+                  const pixel = xBuf.get(batch, xDepth, xRow, xCol, channel);
+                  if (pixel >= maxValue) {
+                    maxValue = pixel;
+                    maxPosition = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterHeight + wCol;
+                  }
+                }
+              }
+            }
+            maxPositions.set(maxPosition, batch, yDepth, yRow, yCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return maxPositions;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool.js
+function avgPool2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "avgPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  let res;
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    res = identity2({ inputs: { x }, backend: backend2 });
+  } else {
+    const xValues = backend2.data.get(x.dataId).values;
+    const strides2 = util_exports.computeStrides(x.shape);
+    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, "avg");
+    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);
+  }
+  return res;
+}
+var avgPoolConfig = {
+  kernelName: AvgPool,
+  backendName: "cpu",
+  kernelFunc: avgPool2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool3D.js
+function avgPool3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  assertNotComplex(x, "avgPool3d");
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);
+  const xValues = backend2.data.get(x.dataId).values;
+  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, "avg");
+  return backend2.makeTensorInfo(outBuf.shape, "float32", outBuf.values);
+}
+var avgPool3DConfig = {
+  kernelName: AvgPool3D,
+  backendName: "cpu",
+  kernelFunc: avgPool3D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPool3DGrad.js
+function avgPool3DGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  assertNotComplex([dy, input2], "avgPool3DGrad");
+  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterDepth = convInfo.filterDepth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(input2.shape, "float32");
+  const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);
+  const dyBuf = backend2.bufferSync(dy);
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {
+        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {
+          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {
+            const dyDepthCorner = dxDepth - padFront;
+            const dyRowCorner = dxRow - padTop;
+            const dyColCorner = dxCol - padLeft;
+            let dotProd = 0;
+            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {
+              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;
+              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {
+                continue;
+              }
+              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {
+                const dyRow = (dyRowCorner + wRow) / strideHeight;
+                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {
+                  continue;
+                }
+                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {
+                  const dyCol = (dyColCorner + wCol) / strideWidth;
+                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {
+                    continue;
+                  }
+                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  dotProd += pixel;
+                }
+              }
+            }
+            dx.set(dotProd * avgMultiplier, batch, dxDepth, dxRow, dxCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var avgPool3DGradConfig2 = {
+  kernelName: AvgPool3DGrad,
+  backendName: "cpu",
+  kernelFunc: avgPool3DGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/AvgPoolGrad.js
+function avgPoolGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  assertNotComplex([dy, input2], "avgPoolGrad");
+  const { filterSize, strides, pad: pad3 } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(x.shape, "float32");
+  const avgMultiplier = 1 / (filterHeight * filterWidth);
+  const dyData = backend2.data.get(dy.dataId).values;
+  const dyBuf = buffer(dy.shape, "float32", dyData);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {
+        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {
+          const dyRCorner = dxR - padTop;
+          const dyCCorner = dxC - padLeft;
+          let dotProd = 0;
+          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {
+            const dyR = (dyRCorner + wR) / strideHeight;
+            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {
+              continue;
+            }
+            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {
+              const dyC = (dyCCorner + wC) / strideWidth;
+              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {
+                continue;
+              }
+              const pixel = dyBuf.get(b, dyR, dyC, d);
+              dotProd += pixel;
+            }
+          }
+          dx.set(dotProd * avgMultiplier, b, dxR, dxC, d);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var avgPoolGradConfig2 = {
+  kernelName: AvgPoolGrad,
+  backendName: "cpu",
+  kernelFunc: avgPoolGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchNorm.js
+function batchNorm2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, scale: scale2, offset, mean: mean4, variance } = inputs;
+  util_exports.assert(mean4.shape.length === variance.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  util_exports.assert(scale2 == null || mean4.shape.length === scale2.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  assertNotComplex([x, mean4, variance, scale2, offset], "batchNorm");
+  let { varianceEpsilon } = attrs;
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const xVals = backend2.data.get(x.dataId).values;
+  const mVals = backend2.data.get(mean4.dataId).values;
+  const varVals = backend2.data.get(variance.dataId).values;
+  const sVals = scale2 ? backend2.data.get(scale2.dataId).values : new Float32Array([1]);
+  const offVals = offset ? backend2.data.get(offset.dataId).values : new Float32Array([0]);
+  const outVals = new Float32Array(xVals.length);
+  const offValsLength = offVals.length;
+  const sValsLength = sVals.length;
+  const varValsLength = varVals.length;
+  const mValsLength = mVals.length;
+  let offi = 0;
+  let mi = 0;
+  let si = 0;
+  let vi = 0;
+  for (let i = 0; i < xVals.length; ++i) {
+    outVals[i] = offVals[offi++] + (xVals[i] - mVals[mi++]) * sVals[si++] / Math.sqrt(varVals[vi++] + varianceEpsilon);
+    if (offi >= offValsLength) {
+      offi = 0;
+    }
+    if (mi >= mValsLength) {
+      mi = 0;
+    }
+    if (si >= sValsLength) {
+      si = 0;
+    }
+    if (vi >= varValsLength) {
+      vi = 0;
+    }
+  }
+  return backend2.makeTensorInfo(x.shape, x.dtype, outVals);
+}
+var batchNormConfig = {
+  kernelName: FusedBatchNorm,
+  backendName: "cpu",
+  kernelFunc: batchNorm2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BatchToSpaceND.js
+function batchToSpaceND2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  assertNotComplex([x], "batchToSpaceND");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const xReshaped = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const xTransposed = transpose2({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });
+  const xTransposedReshaped = reshape3({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });
+  const result = slice2({
+    inputs: { x: xTransposedReshaped },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  backend2.disposeIntermediateTensorInfo(xReshaped);
+  backend2.disposeIntermediateTensorInfo(xTransposed);
+  backend2.disposeIntermediateTensorInfo(xTransposedReshaped);
+  return result;
+}
+var batchToSpaceNDConfig = {
+  kernelName: BatchToSpaceND,
+  backendName: "cpu",
+  kernelFunc: batchToSpaceND2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Bincount.js
+function bincount2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size } = attrs;
+  const xVals = backend2.data.get(x.dataId).values;
+  const weightsVals = backend2.data.get(weights.dataId).values;
+  const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);
+  return backend2.makeTensorInfo([size], weights.dtype, outVals);
+}
+var bincountConfig = {
+  kernelName: Bincount,
+  backendName: "cpu",
+  kernelFunc: bincount2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/BroadcastArgs.js
+function broadcastArgs2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { s0, s1 } = inputs;
+  const s0Vals = backend2.data.get(s0.dataId).values;
+  const s1Vals = backend2.data.get(s1.dataId).values;
+  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));
+  return backend2.makeTensorInfo([broadcastShape.length], "int32", Int32Array.from(broadcastShape));
+}
+var broadcastArgsConfig = {
+  kernelName: BroadcastArgs,
+  backendName: "cpu",
+  kernelFunc: broadcastArgs2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ClipByValue.js
+var clipByValue2 = unaryKernelFunc(ClipByValue, (xi, attrs) => {
+  const clipAttrs = attrs;
+  if (xi > clipAttrs.clipValueMax) {
+    return clipAttrs.clipValueMax;
+  }
+  return xi < clipAttrs.clipValueMin ? clipAttrs.clipValueMin : xi;
+});
+var clipByValueConfig = {
+  kernelName: ClipByValue,
+  backendName: "cpu",
+  kernelFunc: clipByValue2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ComplexAbs.js
+var complexAbs = (args) => {
+  const { x } = args.inputs;
+  const cpuBackend = args.backend;
+  const resultValues = new Float32Array(util_exports.sizeFromShape(x.shape));
+  const complexVals = cpuBackend.data.get(x.dataId);
+  const real4 = complexVals.complexTensorInfos.real;
+  const imag4 = complexVals.complexTensorInfos.imag;
+  const realVals = cpuBackend.data.get(real4.dataId).values;
+  const imagVals = cpuBackend.data.get(imag4.dataId).values;
+  for (let i = 0; i < realVals.length; i++) {
+    const real5 = realVals[i];
+    const imag5 = imagVals[i];
+    resultValues[i] = Math.hypot(real5, imag5);
+  }
+  return cpuBackend.makeOutput(resultValues, x.shape, "float32");
+};
+var complexAbsConfig = {
+  kernelName: ComplexAbs,
+  backendName: "cpu",
+  kernelFunc: complexAbs
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Imag.js
+function imag2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const imag4 = backend2.data.get(input2.dataId).complexTensorInfos.imag;
+  const imagVal = backend2.data.get(imag4.dataId).values;
+  return backend2.makeTensorInfo(imag4.shape, imag4.dtype, imagVal);
+}
+var imagConfig = {
+  kernelName: Imag,
+  backendName: "cpu",
+  kernelFunc: imag2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Concat.js
+function concat2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, $axis);
+  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);
+  }
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity2({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  if ($inputs[0].dtype === "complex64") {
+    const reals = $inputs.map((t) => real2({ inputs: { input: t }, backend: backend2 }));
+    const imags = $inputs.map((t) => imag2({ inputs: { input: t }, backend: backend2 }));
+    const realConcated = concat2({ inputs: reals, backend: backend2, attrs: { axis: $axis } });
+    const imagConcated = concat2({ inputs: imags, backend: backend2, attrs: { axis: $axis } });
+    const result = complex2({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });
+    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));
+    backend2.disposeIntermediateTensorInfo(realConcated);
+    backend2.disposeIntermediateTensorInfo(imagConcated);
+    return result;
+  }
+  const inputs2D = $inputs.map((t) => {
+    const innerSize = util_exports.sizeFromShape(t.shape.slice($axis));
+    const shape = [-1, innerSize];
+    return reshape3({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+  });
+  const inputsValShapes = inputs2D.map((t) => {
+    return { vals: backend2.data.get(t.dataId).values, shape: t.shape };
+  });
+  outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);
+  const simplyConcat = inputs2D[0].shape[0] === 1;
+  const outVals = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);
+  const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), $axis);
+  const outInfo = backend2.makeTensorInfo(finalOutShape, inputs[0].dtype, outVals);
+  inputs2D.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return outInfo;
+}
+var concatConfig = {
+  kernelName: Concat,
+  backendName: "cpu",
+  kernelFunc: concat2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2D.js
+function conv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;
+  assertNotComplex([x, filter], "conv2d");
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const padLeft = convInfo.padInfo.left;
+  const padTop = convInfo.padInfo.top;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const xBatchStride = xStrides[0];
+  const xRowStride = isChannelsLast ? xStrides[1] : xStrides[2];
+  const xColStride = isChannelsLast ? xStrides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : xStrides[1];
+  const yBatchStride = y.strides[0];
+  const yRowStride = isChannelsLast ? y.strides[1] : y.strides[2];
+  const yColStride = isChannelsLast ? y.strides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : y.strides[1];
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xBatchStride;
+    const yOffset1 = b * yBatchStride;
+    for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+      const yOffset2 = yOffset1 + yR * yRowStride;
+      const xRCorner = yR * convInfo.strideHeight - padTop;
+      for (let wR = 0; wR < filterHeight; ++wR) {
+        const xR = xRCorner + wR * dilationHeight;
+        if (xR < 0 || xR >= convInfo.inHeight) {
+          continue;
+        }
+        const wOffset1 = wR * filterStrides[0];
+        const xOffset2 = xOffset1 + xR * xRowStride;
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const yOffset3 = yOffset2 + yC * yColStride;
+          const xCCorner = yC * convInfo.strideWidth - padLeft;
+          for (let wC = 0; wC < filterWidth; ++wC) {
+            const xC = xCCorner + wC * dilationWidth;
+            if (xC < 0 || xC >= convInfo.inWidth) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wC * filterStrides[1];
+            const xOffset3 = xOffset2 + xC * xColStride;
+            let wOffset3 = wOffset2;
+            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+              const xVal = xVals[xOffset3 + d1 * xChannelStride];
+              for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+                yVals[yOffset3 + d2 * yChannelStride] += xVal * wVals[wOffset3 + d2];
+              }
+              wOffset3 += convInfo.outChannels;
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, yVals);
+}
+var conv2DConfig = {
+  kernelName: Conv2D,
+  backendName: "cpu",
+  kernelFunc: conv2D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2DBackpropFilter.js
+function conv2DBackpropFilter2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;
+  assertNotComplex([x, dy], "conv2dBackpropFilter");
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const dW = new TensorBuffer(convInfo.filterShape, "float32");
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  const xVals = backend2.data.get(x.dataId).values;
+  const dyVals = backend2.data.get(dy.dataId).values;
+  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);
+  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);
+  for (let wR = 0; wR < filterHeight; ++wR) {
+    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+    for (let wC = 0; wC < filterWidth; ++wC) {
+      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+      for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+        for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+          let dotProd = 0;
+          for (let b = 0; b < convInfo.batchSize; ++b) {
+            for (let yR = yRMin; yR < yRMax; ++yR) {
+              const xR = wR + yR * strideHeight - topPad;
+              for (let yC = yCMin; yC < yCMax; ++yC) {
+                const xC = wC + yC * strideWidth - leftPad;
+                if (isChannelsLast) {
+                  dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);
+                } else {
+                  dotProd += xBuf.get(b, d1, xR, xC) * dyBuf.get(b, d2, yR, yC);
+                }
+              }
+            }
+          }
+          dW.set(dotProd, wR, wC, d1, d2);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);
+}
+var conv2DBackpropFilterConfig = {
+  kernelName: Conv2DBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: conv2DBackpropFilter2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv2DBackpropInput.js
+function conv2DBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  assertNotComplex([dy, filter], "conv2dBackpropInput");
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  let $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2] = filterStrides;
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  $dataFormat = convInfo.dataFormat;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const isChannelsLast = $dataFormat === "channelsLast";
+  const xBatchStride = dx.strides[0];
+  const xRowStride = isChannelsLast ? dx.strides[1] : dx.strides[2];
+  const xColStride = isChannelsLast ? dx.strides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : dx.strides[1];
+  const yBatchStride = dyStrides[0];
+  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];
+  const yColStride = isChannelsLast ? dyStrides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xR = 0; xR < inHeight; ++xR) {
+        const xRCorner = xR - topPad;
+        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+        for (let xC = 0; xC < inWidth; ++xC) {
+          const xCCorner = xC - leftPad;
+          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+          let dotProd = 0;
+          for (let yR = xRMin; yR < yRMax; ++yR) {
+            const wR = yR * strideHeight - xRCorner;
+            for (let yC = xCMin; yC < yCMax; ++yC) {
+              const wC = yC * strideWidth - xCCorner;
+              const dyOffset = yBatchStride * b + yRowStride * yR + yColStride * yC;
+              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;
+              for (let d2 = 0; d2 < outChannels; ++d2) {
+                const pixel = dyValues[dyOffset + yChannelStride * d2];
+                const weight = fltValues[fltOffset + d2];
+                dotProd += pixel * weight;
+              }
+            }
+          }
+          const dxOffset = xBatchStride * b + xRowStride * xR + xColStride * xC + xChannelStride * d1;
+          dxValues[dxOffset] = dotProd;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var conv2DBackpropInputConfig = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "cpu",
+  kernelFunc: conv2DBackpropInput2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3D.js
+function conv3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  assertNotComplex([x, filter], "conv3d");
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);
+  const { filterDepth, filterHeight, filterWidth, dilationDepth, dilationHeight, dilationWidth, padInfo } = convInfo;
+  const padFront = padInfo.front;
+  const padLeft = padInfo.left;
+  const padTop = padInfo.top;
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xStrides[0];
+    const yOffset1 = b * y.strides[0];
+    for (let yF = 0; yF < convInfo.outDepth; ++yF) {
+      const yOffset2 = yOffset1 + yF * y.strides[1];
+      const xFCorner = yF * convInfo.strideDepth - padFront;
+      for (let wF = 0; wF < filterDepth; ++wF) {
+        const xF = xFCorner + wF * dilationDepth;
+        if (xF < 0 || xF >= convInfo.inDepth) {
+          continue;
+        }
+        const wOffset1 = wF * filterStrides[0];
+        const xOffset2 = xOffset1 + xF * xStrides[1];
+        for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+          const yOffset3 = yOffset2 + yR * y.strides[2];
+          const xRCorner = yR * convInfo.strideHeight - padTop;
+          for (let wR = 0; wR < filterHeight; ++wR) {
+            const xR = xRCorner + wR * dilationHeight;
+            if (xR < 0 || xR >= convInfo.inHeight) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wR * filterStrides[1];
+            const xOffset3 = xOffset2 + xR * xStrides[2];
+            for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+              const yOffset4 = yOffset3 + yC * convInfo.outChannels;
+              const xCCorner = yC * convInfo.strideWidth - padLeft;
+              for (let wC = 0; wC < filterWidth; ++wC) {
+                const xC = xCCorner + wC * dilationWidth;
+                if (xC < 0 || xC >= convInfo.inWidth) {
+                  continue;
+                }
+                const wOffset3 = wOffset2 + wC * filterStrides[2];
+                const xOffset4 = xOffset3 + xC * convInfo.inChannels;
+                let wOffset4 = wOffset3;
+                for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+                  const xVal = xVals[xOffset4 + d1];
+                  for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+                    yVals[yOffset4 + d2] += xVal * wVals[wOffset4 + d2];
+                  }
+                  wOffset4 += convInfo.outChannels;
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);
+}
+var conv3DConfig = {
+  kernelName: Conv3D,
+  backendName: "cpu",
+  kernelFunc: conv3D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3DBackpropFilterV2.js
+function conv3DBackpropFilterV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, filterShape } = attrs;
+  assertNotComplex([x, dy], "conv3dBackpropFilterV2");
+  const xStrides = util_exports.computeStrides(x.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const filterDepth = convInfo.filterDepth;
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const dw = new TensorBuffer(convInfo.filterShape, "float32");
+  const dwValues = dw.values;
+  const [dwS0, dwS1, dwS2, dwS3] = dw.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;
+  const xValues = backend2.data.get(x.dataId).values;
+  const [xS0, xS1, xS2, xS3] = xStrides;
+  const frontPad = convInfo.padInfo.front;
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  for (let wF = 0; wF < filterDepth; ++wF) {
+    const yFMin = Math.max(0, Math.ceil((frontPad - wF) / strideDepth));
+    const yFMax = Math.min(convInfo.outDepth, (convInfo.inDepth + frontPad - wF) / strideDepth);
+    const wOffset1 = wF * dwS0;
+    for (let wR = 0; wR < filterHeight; ++wR) {
+      const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+      const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+      const wOffset2 = wR * dwS1 + wOffset1;
+      for (let wC = 0; wC < filterWidth; ++wC) {
+        const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+        const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+        const wOffset3 = wC * dwS2 + wOffset2;
+        for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+          const wOffset4 = d1 * dwS3 + wOffset3;
+          for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+            let dotProd = 0;
+            for (let b = 0; b < convInfo.batchSize; ++b) {
+              const xOffset1 = b * xS0;
+              const yOffset1 = b * dyS0;
+              for (let yF = yFMin; yF < yFMax; ++yF) {
+                const xF = wF + yF * strideDepth - frontPad;
+                const xOffset2 = xF * xS1 + xOffset1;
+                const yOffset2 = yF * dyS1 + yOffset1;
+                for (let yR = yRMin; yR < yRMax; ++yR) {
+                  const xR = wR + yR * strideHeight - topPad;
+                  const xOffset3 = xR * xS2 + xOffset2;
+                  const yOffset3 = yR * dyS2 + yOffset2;
+                  for (let yC = yCMin; yC < yCMax; ++yC) {
+                    const xC = wC + yC * strideWidth - leftPad;
+                    const xOffset4 = xC * xS3 + xOffset3;
+                    const yOffset4 = yC * dyS3 + yOffset3;
+                    dotProd += xValues[xOffset4 + d1] * dyValues[yOffset4 + d2];
+                  }
+                }
+              }
+            }
+            dwValues[wOffset4 + d2] = dotProd;
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dw.shape, dw.dtype, dw.values);
+}
+var conv3DBackpropFilterV2Config = {
+  kernelName: Conv3DBackpropFilterV2,
+  backendName: "cpu",
+  kernelFunc: conv3DBackpropFilterV2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Conv3DBackpropInputV2.js
+function conv3DBackpropInputV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { pad: pad3, strides, inputShape } = attrs;
+  assertNotComplex([dy], "conv3dBackpropInputV2");
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const [dxS0, dxS1, dxS2, dxS3] = dx.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2, dyS3] = dyStrides;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2, fltS3] = filterStrides;
+  const { batchSize, filterDepth, filterHeight, filterWidth, inChannels, inDepth, inHeight, inWidth, outChannels, outDepth, outHeight, outWidth, strideDepth, strideHeight, strideWidth } = convInfo;
+  const frontPad = filterDepth - 1 - convInfo.padInfo.front;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xF = 0; xF < inDepth; ++xF) {
+        const xFCorner = xF - frontPad;
+        const xFMin = Math.max(0, Math.ceil(xFCorner / strideDepth));
+        const yFMax = Math.min(outDepth, (filterDepth + xFCorner) / strideDepth);
+        for (let xR = 0; xR < inHeight; ++xR) {
+          const xRCorner = xR - topPad;
+          const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+          const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+          for (let xC = 0; xC < inWidth; ++xC) {
+            const xCCorner = xC - leftPad;
+            const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+            const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+            let dotProd = 0;
+            for (let yF = xFMin; yF < yFMax; ++yF) {
+              const wF = yF * strideDepth - xFCorner;
+              for (let yR = xRMin; yR < yRMax; ++yR) {
+                const wR = yR * strideHeight - xRCorner;
+                for (let yC = xCMin; yC < yCMax; ++yC) {
+                  const wC = yC * strideWidth - xCCorner;
+                  const dyOffset = dyS0 * b + dyS1 * yF + dyS2 * yR + dyS3 * yC;
+                  const fltOffset = fltS0 * (filterDepth - 1 - wF) + fltS1 * (filterHeight - 1 - wR) + fltS2 * (filterWidth - 1 - wC) + fltS3 * d1;
+                  for (let d2 = 0; d2 < outChannels; ++d2) {
+                    const pixel = dyValues[dyOffset + d2];
+                    const weight = fltValues[fltOffset + d2];
+                    dotProd += pixel * weight;
+                  }
+                }
+              }
+            }
+            dxValues[dxS0 * b + dxS1 * xF + dxS2 * xR + dxS3 * xC + d1] = dotProd;
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var conv3DBackpropInputV2Config = {
+  kernelName: Conv3DBackpropInputV2,
+  backendName: "cpu",
+  kernelFunc: conv3DBackpropInputV2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cos.js
+var cos2 = unaryKernelFunc(Cos, (xi) => Math.cos(xi));
+var cosConfig = {
+  kernelName: Cos,
+  backendName: "cpu",
+  kernelFunc: cos2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cosh.js
+var cosh2 = unaryKernelFunc(Cosh, (xi) => Math.cosh(xi));
+var coshConfig = {
+  kernelName: Cosh,
+  backendName: "cpu",
+  kernelFunc: cosh2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/CropAndResize.js
+function cropAndResize2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, boxes, boxInd } = inputs;
+  const { cropSize, method, extrapolationValue } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const numBoxes = boxes.shape[0];
+  const [cropHeight, cropWidth] = cropSize;
+  const output = buffer([numBoxes, cropHeight, cropWidth, numChannels], "float32");
+  const boxVals = backend2.data.get(boxes.dataId).values;
+  const boxIndVals = backend2.data.get(boxInd.dataId).values;
+  const imageVals = backend2.data.get(image2.dataId).values;
+  const inStride = util_exports.computeStrides(image2.shape);
+  const outStride = util_exports.computeStrides(output.shape);
+  for (let b = 0; b < numBoxes; b++) {
+    const startInd = b * 4;
+    const y1 = boxVals[startInd];
+    const x1 = boxVals[startInd + 1];
+    const y2 = boxVals[startInd + 2];
+    const x2 = boxVals[startInd + 3];
+    const bInd = boxIndVals[b];
+    if (bInd >= batch) {
+      continue;
+    }
+    const heightScale = cropHeight > 1 ? (y2 - y1) * (imageHeight - 1) / (cropHeight - 1) : 0;
+    const widthScale = cropWidth > 1 ? (x2 - x1) * (imageWidth - 1) / (cropWidth - 1) : 0;
+    for (let y = 0; y < cropHeight; y++) {
+      const yInd = cropHeight > 1 ? y1 * (imageHeight - 1) + y * heightScale : 0.5 * (y1 + y2) * (imageHeight - 1);
+      if (yInd < 0 || yInd > imageHeight - 1) {
+        for (let x = 0; x < cropWidth; x++) {
+          for (let c = 0; c < numChannels; c++) {
+            const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[ind] = extrapolationValue;
+          }
+        }
+        continue;
+      }
+      if (method === "bilinear") {
+        const topInd = Math.floor(yInd);
+        const bottomInd = Math.ceil(yInd);
+        const yLerp = yInd - topInd;
+        for (let x = 0; x < cropWidth; x++) {
+          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);
+          if (xInd < 0 || xInd > imageWidth - 1) {
+            for (let c = 0; c < numChannels; c++) {
+              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+              output.values[ind] = extrapolationValue;
+            }
+            continue;
+          }
+          const leftInd = Math.floor(xInd);
+          const rightInd = Math.ceil(xInd);
+          const xLerp = xInd - leftInd;
+          for (let c = 0; c < numChannels; c++) {
+            let ind = c + leftInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];
+            const topLeft = imageVals[ind];
+            ind = c + rightInd * inStride[2] + topInd * inStride[1] + bInd * inStride[0];
+            const topRight = imageVals[ind];
+            ind = c + leftInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];
+            const bottomLeft = imageVals[ind];
+            ind = c + rightInd * inStride[2] + bottomInd * inStride[1] + bInd * inStride[0];
+            const bottomRight = imageVals[ind];
+            const top = topLeft + (topRight - topLeft) * xLerp;
+            const bottom = bottomLeft + (bottomRight - bottomLeft) * xLerp;
+            ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[ind] = top + (bottom - top) * yLerp;
+          }
+        }
+      } else {
+        for (let x = 0; x < cropWidth; ++x) {
+          const xInd = cropWidth > 1 ? x1 * (imageWidth - 1) + x * widthScale : 0.5 * (x1 + x2) * (imageWidth - 1);
+          if (xInd < 0 || xInd > imageWidth - 1) {
+            for (let c = 0; c < numChannels; c++) {
+              const ind = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+              output.values[ind] = extrapolationValue;
+            }
+            continue;
+          }
+          const closestX = Math.round(xInd);
+          const closestY = Math.round(yInd);
+          for (let c = 0; c < numChannels; c++) {
+            const inInd = c + closestX * inStride[2] + closestY * inStride[1] + bInd * inStride[0];
+            const outInd = c + x * outStride[2] + y * outStride[1] + b * outStride[0];
+            output.values[outInd] = imageVals[inInd];
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(output.shape, output.dtype, output.values);
+}
+var cropAndResizeConfig = {
+  kernelName: CropAndResize,
+  backendName: "cpu",
+  kernelFunc: cropAndResize2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cumprod.js
+function cumprod2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  assertNotComplex(x, "cumprod");
+  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);
+  let $x = x;
+  if (permutation != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];
+  if (permutedAxis !== $x.shape.length - 1) {
+    throw new Error(`backend.cumprod in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);
+  }
+  const resultDtype = upcastType($x.dtype, "int32");
+  const vals = util_exports.makeOnesTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  const finalDim = $x.shape[$x.shape.length - 1];
+  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;
+  for (let i = 0; i < aVals.length; i += finalDim) {
+    for (let j = 0; j < finalDim; j++) {
+      const idx = indexAdjuster(i, j);
+      if (j === 0) {
+        vals[idx] = exclusive ? 1 : aVals[idx];
+      } else {
+        const prevIdx = indexAdjuster(i, j - 1);
+        vals[idx] = exclusive ? aVals[prevIdx] * vals[prevIdx] : aVals[idx] * vals[prevIdx];
+      }
+    }
+  }
+  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo($x);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+var cumprodConfig = {
+  kernelName: Cumprod,
+  backendName: "cpu",
+  kernelFunc: cumprod2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Cumsum.js
+function cumsum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  assertNotComplex(x, "cumsum");
+  const permutation = backend_util_exports.getAxesPermutation([axis], x.shape.length);
+  let $x = x;
+  if (permutation != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, x.shape.length)[0];
+  if (permutedAxis !== $x.shape.length - 1) {
+    throw new Error(`backend.cumsum in CPU expects an inner-most axis=${$x.shape.length - 1} but got axis=${permutedAxis}`);
+  }
+  const resultDtype = upcastType($x.dtype, "int32");
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape($x.shape), resultDtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  const finalDim = $x.shape[$x.shape.length - 1];
+  const indexAdjuster = reverse5 ? (i, j) => i + finalDim - j - 1 : (i, j) => i + j;
+  for (let i = 0; i < aVals.length; i += finalDim) {
+    for (let j = 0; j < finalDim; j++) {
+      const idx = indexAdjuster(i, j);
+      if (j === 0) {
+        vals[idx] = exclusive ? 0 : aVals[idx];
+      } else {
+        const prevIdx = indexAdjuster(i, j - 1);
+        vals[idx] = exclusive ? aVals[prevIdx] + vals[prevIdx] : aVals[idx] + vals[prevIdx];
+      }
+    }
+  }
+  const result = backend2.makeTensorInfo($x.shape, resultDtype, vals);
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose2({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo($x);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+var cumsumConfig = {
+  kernelName: Cumsum,
+  backendName: "cpu",
+  kernelFunc: cumsum2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DenseBincount.js
+function denseBincount2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size, binaryOutput } = attrs;
+  if (x.shape.length === 1) {
+    const xVals = backend2.data.get(x.dataId).values;
+    const weightsVals = backend2.data.get(weights.dataId).values;
+    const outVals = bincountImpl(xVals, weightsVals, weights.dtype, weights.shape, size);
+    return backend2.makeTensorInfo([size], weights.dtype, outVals);
+  } else if (x.shape.length === 2) {
+    const xBuf = backend2.bufferSync(x);
+    const weightsBuf = backend2.bufferSync(weights);
+    const outBuf = bincountReduceImpl(xBuf, weightsBuf, size, binaryOutput);
+    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);
+  }
+  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);
+}
+var denseBincountConfig = {
+  kernelName: DenseBincount,
+  backendName: "cpu",
+  kernelFunc: denseBincount2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthToSpace.js
+function depthToSpace2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  util_exports.assert(dataFormat === "NHWC", () => `Only NHWC dataFormat supported on CPU for depthToSpace. Got ${dataFormat}`);
+  const batchSize = x.shape[0];
+  const inputHeight = x.shape[1];
+  const inputWidth = x.shape[2];
+  const inputDepth = x.shape[3];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const xValues = backend2.data.get(x.dataId).values;
+  const result = new Float32Array(batchSize * outputHeight * outputWidth * outputDepth);
+  let outputIdx = 0;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let h = 0; h < outputHeight; ++h) {
+      const inH = Math.floor(h / blockSize);
+      const offsetH = h % blockSize;
+      for (let w = 0; w < outputWidth; ++w) {
+        const inW = Math.floor(w / blockSize);
+        const offsetW = w % blockSize;
+        const offsetD = (offsetH * blockSize + offsetW) * outputDepth;
+        for (let d = 0; d < outputDepth; ++d) {
+          const inD = d + offsetD;
+          const inputIdx = inD + inputDepth * (inW + inputWidth * (inH + inputHeight * b));
+          result[outputIdx++] = xValues[inputIdx];
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batchSize, outputHeight, outputWidth, outputDepth], x.dtype, result);
+}
+var depthToSpaceConfig = {
+  kernelName: DepthToSpace,
+  backendName: "cpu",
+  kernelFunc: depthToSpace2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNative.js
+function depthwiseConv2dNative(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;
+  assertNotComplex([x, filter], "depthwiseConv2DNative");
+  const xStrides = util_exports.computeStrides(x.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const { filterHeight, filterWidth, dilationHeight, dilationWidth, padInfo } = convInfo;
+  const padLeft = padInfo.left;
+  const padTop = padInfo.top;
+  const chMul = convInfo.outChannels / convInfo.inChannels;
+  const y = new TensorBuffer(convInfo.outShape, x.dtype);
+  const xVals = backend2.data.get(x.dataId).values;
+  const wVals = backend2.data.get(filter.dataId).values;
+  const yVals = y.values;
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    const xOffset1 = b * xStrides[0];
+    const yOffset1 = b * y.strides[0];
+    for (let yR = 0; yR < convInfo.outHeight; ++yR) {
+      const yOffset2 = yOffset1 + yR * y.strides[1];
+      const xRCorner = yR * convInfo.strideHeight - padTop;
+      for (let wR = 0; wR < filterHeight; ++wR) {
+        const xR = xRCorner + wR * dilationHeight;
+        if (xR < 0 || xR >= convInfo.inHeight) {
+          continue;
+        }
+        const wOffset1 = wR * filterStrides[0];
+        const xOffset2 = xOffset1 + xR * xStrides[1];
+        for (let yC = 0; yC < convInfo.outWidth; ++yC) {
+          const yOffset3 = yOffset2 + yC * y.strides[2];
+          const xCCorner = yC * convInfo.strideWidth - padLeft;
+          for (let wC = 0; wC < filterWidth; ++wC) {
+            const xC = xCCorner + wC * dilationWidth;
+            if (xC < 0 || xC >= convInfo.inWidth) {
+              continue;
+            }
+            const wOffset2 = wOffset1 + wC * filterStrides[1];
+            const xOffset3 = xOffset2 + xC * convInfo.inChannels;
+            let yOffset4 = yOffset3;
+            let wOffset3 = wOffset2;
+            for (let d1 = 0; d1 < convInfo.inChannels; ++d1) {
+              const xVal = xVals[xOffset3 + d1];
+              for (let q = 0; q < chMul; ++q) {
+                yVals[yOffset4 + q] += xVal * wVals[wOffset3 + q];
+              }
+              yOffset4 += chMul;
+              wOffset3 += chMul;
+            }
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, y.dtype, y.values);
+}
+var depthwiseConv2dNativeConfig = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNative
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNativeBackpropFilter.js
+function depthwiseConv2dNativeBackpropFilter2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;
+  assertNotComplex([x, dy], "depthwiseConv2dNativeBackpropFilter");
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);
+  const { strideHeight, strideWidth, filterHeight, filterWidth } = convInfo;
+  const dW = new TensorBuffer(convInfo.filterShape, "float32");
+  const leftPad = convInfo.padInfo.left;
+  const topPad = convInfo.padInfo.top;
+  const chMul = convInfo.outChannels / convInfo.inChannels;
+  const xVals = backend2.data.get(x.dataId).values;
+  const xBuf = new TensorBuffer(x.shape, x.dtype, xVals);
+  const dyVals = backend2.data.get(dy.dataId).values;
+  const dyBuf = new TensorBuffer(dy.shape, dy.dtype, dyVals);
+  for (let wR = 0; wR < filterHeight; ++wR) {
+    const yRMin = Math.max(0, Math.ceil((topPad - wR) / strideHeight));
+    const yRMax = Math.min(convInfo.outHeight, (convInfo.inHeight + topPad - wR) / strideHeight);
+    for (let wC = 0; wC < filterWidth; ++wC) {
+      const yCMin = Math.max(0, Math.ceil((leftPad - wC) / strideWidth));
+      const yCMax = Math.min(convInfo.outWidth, (convInfo.inWidth + leftPad - wC) / strideWidth);
+      for (let d2 = 0; d2 < convInfo.outChannels; ++d2) {
+        const d1 = Math.trunc(d2 / chMul);
+        const dm = d2 % chMul;
+        let dotProd = 0;
+        for (let b = 0; b < convInfo.batchSize; ++b) {
+          for (let yR = yRMin; yR < yRMax; ++yR) {
+            const xR = wR + yR * strideHeight - topPad;
+            for (let yC = yCMin; yC < yCMax; ++yC) {
+              const xC = wC + yC * strideWidth - leftPad;
+              dotProd += xBuf.get(b, xR, xC, d1) * dyBuf.get(b, yR, yC, d2);
+            }
+          }
+        }
+        dW.set(dotProd, wR, wC, d1, dm);
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dW.shape, dW.dtype, dW.values);
+}
+var depthwiseConv2dNativeBackpropFilterConfig = {
+  kernelName: DepthwiseConv2dNativeBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNativeBackpropFilter2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/DepthwiseConv2dNativeBackpropInput.js
+function depthwiseConv2dNativeBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;
+  assertNotComplex([dy, filter], "depthwiseConv2DNativeBackpropInput");
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const filterStrides = util_exports.computeStrides(filter.shape);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const dx = new TensorBuffer(convInfo.inShape, "float32");
+  const dxValues = dx.values;
+  const [dxS0, dxS1, dxS2] = dx.strides;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const [dyS0, dyS1, dyS2] = dyStrides;
+  const fltValues = backend2.data.get(filter.dataId).values;
+  const [fltS0, fltS1, fltS2] = filterStrides;
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const chMul = outChannels / inChannels;
+  for (let b = 0; b < batchSize; ++b) {
+    for (let d1 = 0; d1 < inChannels; ++d1) {
+      for (let xR = 0; xR < inHeight; ++xR) {
+        const xRCorner = xR - topPad;
+        const xRMin = Math.max(0, Math.ceil(xRCorner / strideHeight));
+        const yRMax = Math.min(outHeight, (filterHeight + xRCorner) / strideHeight);
+        for (let xC = 0; xC < inWidth; ++xC) {
+          const xCCorner = xC - leftPad;
+          const xCMin = Math.max(0, Math.ceil(xCCorner / strideWidth));
+          const yCMax = Math.min(outWidth, (filterWidth + xCCorner) / strideWidth);
+          let dotProd = 0;
+          for (let yR = xRMin; yR < yRMax; ++yR) {
+            const wR = yR * strideHeight - xRCorner;
+            for (let yC = xCMin; yC < yCMax; ++yC) {
+              const wC = yC * strideWidth - xCCorner;
+              const dyOffset = dyS0 * b + dyS1 * yR + dyS2 * yC;
+              const fltOffset = fltS0 * (filterHeight - 1 - wR) + fltS1 * (filterWidth - 1 - wC) + fltS2 * d1;
+              for (let dm = 0; dm < chMul; ++dm) {
+                const d2 = d1 * chMul + dm;
+                const pixel = dyValues[dyOffset + d2];
+                const weight = fltValues[fltOffset + dm];
+                dotProd += pixel * weight;
+              }
+            }
+          }
+          dxValues[dxS0 * b + dxS1 * xR + dxS2 * xC + d1] = dotProd;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var depthwiseConv2dNativeBackpropInputConfig = {
+  kernelName: DepthwiseConv2dNativeBackpropInput,
+  backendName: "cpu",
+  kernelFunc: depthwiseConv2dNativeBackpropInput2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Diag.js
+function diag2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xVals = backend2.data.get(x.dataId).values;
+  const outBuf = buffer([xSize, xSize], x.dtype);
+  const vals = outBuf.values;
+  for (let i = 0; i < xVals.length; i++) {
+    vals[i * xSize + i] = xVals[i];
+  }
+  const outShape = [...x.shape, ...x.shape];
+  return backend2.makeTensorInfo(outShape, outBuf.dtype, outBuf.values);
+}
+var diagConfig = {
+  kernelName: Diag,
+  backendName: "cpu",
+  kernelFunc: diag2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2D.js
+var dilation2DConfig = {
+  kernelName: Dilation2D,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const xVals = cpuBackend.data.get(x.dataId).values;
+    const xRank = x.shape.length;
+    const filterVals = cpuBackend.data.get(filter.dataId).values;
+    const filterRank = filter.shape.length;
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    const outSize = util_exports.sizeFromShape(outShape);
+    const outRank = outShape.length;
+    const outputVals = util_exports.getArrayFromDType(x.dtype, outSize);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const xIndex = util_exports.locToIndex([b, hIn, wIn, d], xRank, util_exports.computeStrides(x.shape));
+                    const filterIndex = util_exports.locToIndex([h, w, d], filterRank, util_exports.computeStrides(filter.shape));
+                    const val = xVals[xIndex] + filterVals[filterIndex];
+                    if (val > curVal) {
+                      curVal = val;
+                    }
+                  }
+                }
+              }
+            }
+            const outputIndex = util_exports.locToIndex([b, hOut, wOut, d], outRank, util_exports.computeStrides(outShape));
+            outputVals[outputIndex] = curVal;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(outputVals, x.dtype), outShape, x.dtype);
+    return { dataId, shape: outShape, dtype: x.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2DBackpropFilter.js
+var dilation2DBackpropFilterConfig = {
+  kernelName: Dilation2DBackpropFilter,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter, dy } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);
+    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropFilter}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);
+    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);
+    const gradients = util_exports.makeZerosNestedTypedArray(filter.shape, filter.dtype);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            let hMax = 0;
+            let wMax = 0;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];
+                    if (val > curVal) {
+                      curVal = val;
+                      hMax = h;
+                      wMax = w;
+                    }
+                  }
+                }
+              }
+            }
+            gradients[hMax][wMax][d] += $dy[b][hOut][wOut][d];
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), filter.shape, filter.dtype);
+    return { dataId, shape: filter.shape, dtype: filter.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Dilation2DBackpropInput.js
+var dilation2DBackpropInputConfig = {
+  kernelName: Dilation2DBackpropInput,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2, attrs }) => {
+    const { x, filter, dy } = inputs;
+    const { strides, pad: pad3, dilations } = attrs;
+    const cpuBackend = backend2;
+    const $x = util_exports.toNestedArray(x.shape, cpuBackend.data.get(x.dataId).values);
+    const $filter = util_exports.toNestedArray(filter.shape, cpuBackend.data.get(filter.dataId).values);
+    const { batchSize, inHeight, inWidth, inChannels, outHeight, outWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth, outShape } = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+    util_exports.assert(dy.rank === outShape.length, () => `Error in ${Dilation2DBackpropInput}, dy must have the same rank as output ${outShape.length}, but got ${dy.rank}`);
+    const $dy = util_exports.toNestedArray(outShape, cpuBackend.data.get(dy.dataId).values);
+    const gradients = util_exports.makeZerosNestedTypedArray(x.shape, x.dtype);
+    for (let b = 0; b < batchSize; ++b) {
+      for (let hOut = 0; hOut < outHeight; ++hOut) {
+        const hBeg = hOut * strideHeight - padInfo.top;
+        for (let wOut = 0; wOut < outWidth; ++wOut) {
+          const wBeg = wOut * strideWidth - padInfo.left;
+          for (let d = 0; d < inChannels; ++d) {
+            let curVal = Number.MIN_SAFE_INTEGER;
+            let hInMax = hBeg < 0 ? 0 : hBeg;
+            let wInMax = wBeg < 0 ? 0 : wBeg;
+            for (let h = 0; h < filterHeight; ++h) {
+              const hIn = hBeg + h * dilationHeight;
+              if (hIn >= 0 && hIn < inHeight) {
+                for (let w = 0; w < filterWidth; ++w) {
+                  const wIn = wBeg + w * dilationWidth;
+                  if (wIn >= 0 && wIn < inWidth) {
+                    const val = $x[b][hIn][wIn][d] + $filter[h][w][d];
+                    if (val > curVal) {
+                      curVal = val;
+                      hInMax = hIn;
+                      wInMax = wIn;
+                    }
+                  }
+                }
+              }
+            }
+            gradients[b][hInMax][wInMax][d] += $dy[b][hOut][wOut][d];
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(util_exports.toTypedArray(gradients, x.dtype), x.shape, x.dtype);
+    return { dataId, shape: x.shape, dtype: x.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sum.js
+function sum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "sum");
+  let $x;
+  if (x.dtype === "bool") {
+    $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "int32" } });
+  } else {
+    $x = identity2({ inputs: { x }, backend: backend2 });
+  }
+  const xRank = $x.shape.length;
+  const axes = util_exports.parseAxisParam(axis, $x.shape);
+  const permutation = backend_util_exports.getAxesPermutation(axes, xRank);
+  let reductionAxes = axes;
+  let permutedX = $x;
+  if (permutation != null) {
+    permutedX = transpose2({ inputs: { x: $x }, backend: backend2, attrs: { perm: permutation } });
+    reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", reductionAxes, permutedX.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, reductionAxes);
+  const resultDtype = backend_util_exports.upcastType(permutedX.dtype, "int32");
+  let result = zeros3(backend2, outShape, resultDtype);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = backend2.data.get(result.dataId).values;
+  const aVals = backend2.data.get(permutedX.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let sum6 = 0;
+    for (let j = 0; j < reduceSize; ++j) {
+      sum6 += aVals[offset + j];
+    }
+    vals[i] = sum6;
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(result.shape, axes);
+    const oldResult = result;
+    result = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: newShape } });
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  backend2.disposeIntermediateTensorInfo($x);
+  if (permutation != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return result;
+}
+var sumConfig = {
+  kernelName: Sum,
+  backendName: "cpu",
+  kernelFunc: sum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Einsum.js
+function einsum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { equation } = attrs;
+  const tensors = inputs;
+  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);
+  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);
+  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);
+  const nSteps = steps.length;
+  let out = null;
+  let numDimsRemaining = allDims.length;
+  const tensorsToDispose = [];
+  for (let i = 0; i < nSteps; ++i) {
+    for (const idTerm of steps[i]) {
+      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);
+      let x;
+      if (backend_util_exports.isIdentityPermutation(perm)) {
+        x = tensors[idTerm];
+      } else {
+        x = transpose2({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });
+        tensorsToDispose.push(x);
+      }
+      const targetShape = x.shape.slice();
+      for (let k = 0; k < dimsToExpand.length; ++k) {
+        targetShape.splice(dimsToExpand[k], 0, 1);
+      }
+      if (!util_exports.arraysEqual(x.shape, targetShape)) {
+        x = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });
+        tensorsToDispose.push(x);
+      }
+      if (out === null) {
+        out = x;
+      } else {
+        out = multiply2({ inputs: { a: x, b: out }, backend: backend2 });
+        tensorsToDispose.push(out);
+      }
+    }
+    if (i < nSteps - 1) {
+      if (path[i] >= 0) {
+        out = sum3({
+          inputs: { x: out },
+          backend: backend2,
+          attrs: {
+            axis: path[i] - (allDims.length - numDimsRemaining),
+            keepDims: false
+          }
+        });
+        tensorsToDispose.push(out);
+      }
+      numDimsRemaining--;
+    }
+  }
+  for (const tensorInfo of tensorsToDispose) {
+    if (tensorInfo === out) {
+      continue;
+    }
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+  return out;
+}
+var einsumConfig = {
+  kernelName: Einsum,
+  backendName: "cpu",
+  kernelFunc: einsum2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/EluGrad.js
+function eluGrad(args) {
+  const { inputs, backend: backend2 } = args;
+  const { dy, y } = inputs;
+  assertNotComplex([dy, y], "eluGrad");
+  const resultValues = new Float32Array(util_exports.sizeFromShape(y.shape));
+  const values = backend2.data.get(y.dataId).values;
+  const dyValues = backend2.data.get(dy.dataId).values;
+  for (let i = 0; i < values.length; ++i) {
+    const v = values[i];
+    if (v >= 1) {
+      resultValues[i] = dyValues[i];
+    } else {
+      resultValues[i] = dyValues[i] * (v + 1);
+    }
+  }
+  return backend2.makeTensorInfo(y.shape, "float32", resultValues);
+}
+var eluGradConfig2 = {
+  kernelName: EluGrad,
+  backendName: "cpu",
+  kernelFunc: eluGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Erf.js
+var p = backend_util_exports.ERF_P;
+var a1 = backend_util_exports.ERF_A1;
+var a2 = backend_util_exports.ERF_A2;
+var a3 = backend_util_exports.ERF_A3;
+var a4 = backend_util_exports.ERF_A4;
+var a5 = backend_util_exports.ERF_A5;
+var erf2 = unaryKernelFunc(Erf, (xi) => {
+  const sign4 = Math.sign(xi);
+  const v = Math.abs(xi);
+  const t = 1 / (1 + p * v);
+  return sign4 * (1 - ((((a5 * t + a4) * t + a3) * t + a2) * t + a1) * t * Math.exp(-v * v));
+});
+var erfConfig = {
+  kernelName: Erf,
+  backendName: "cpu",
+  kernelFunc: erf2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ExpandDims.js
+function expandDims3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { input: input2 } = inputs;
+  const { dim } = attrs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape3({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig = {
+  kernelName: ExpandDims,
+  backendName: "cpu",
+  kernelFunc: expandDims3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RealDiv.js
+var realDivImpl = createSimpleBinaryKernelImpl((a, b) => a / b);
+var div2 = binaryKernelFunc(RealDiv, realDivImpl);
+var realDivConfig = {
+  kernelName: RealDiv,
+  backendName: "cpu",
+  kernelFunc: div2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/utils/fft_utils.js
+function fftBatch(input2, inverse, cpuBackend) {
+  const inputShape = input2.shape;
+  const batch = inputShape[0];
+  const innerDim = inputShape[1];
+  const inputVals = cpuBackend.data.get(input2.dataId);
+  const real2D = inputVals.complexTensorInfos.real;
+  const imag2D = inputVals.complexTensorInfos.imag;
+  const resultShape = [batch, innerDim];
+  const resultSize = util_exports.sizeFromShape(resultShape);
+  const resultReal = util_exports.getTypedArrayFromDType("float32", resultSize);
+  const resultImag = util_exports.getTypedArrayFromDType("float32", resultSize);
+  for (let b = 0; b < batch; b++) {
+    const r = slice2({
+      inputs: { x: real2D },
+      backend: cpuBackend,
+      attrs: { begin: [b, 0], size: [1, innerDim] }
+    });
+    const i = slice2({
+      inputs: { x: imag2D },
+      backend: cpuBackend,
+      attrs: { begin: [b, 0], size: [1, innerDim] }
+    });
+    const input3 = complex2({ inputs: { real: r, imag: i }, backend: cpuBackend });
+    const { real: real4, imag: imag4 } = fftImpl(input3, inverse, cpuBackend);
+    const res = backend_util_exports.mergeRealAndImagArrays(real4, imag4);
+    for (let d = 0; d < innerDim; d++) {
+      const c = backend_util_exports.getComplexWithIndex(res, d);
+      resultReal[b * innerDim + d] = c.real;
+      resultImag[b * innerDim + d] = c.imag;
+    }
+    cpuBackend.disposeIntermediateTensorInfo(r);
+    cpuBackend.disposeIntermediateTensorInfo(i);
+    cpuBackend.disposeIntermediateTensorInfo(input3);
+  }
+  const $realInfo = cpuBackend.makeTensorInfo(resultShape, "float32", resultReal);
+  const $imagInfo = cpuBackend.makeTensorInfo(resultShape, "float32", resultImag);
+  const result = complex2({ inputs: { real: $realInfo, imag: $imagInfo }, backend: cpuBackend });
+  cpuBackend.disposeIntermediateTensorInfo($realInfo);
+  cpuBackend.disposeIntermediateTensorInfo($imagInfo);
+  return result;
+}
+function fftImpl(input2, inverse, cpuBackend) {
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const inputVals = cpuBackend.data.get(input2.dataId);
+  const realVals = cpuBackend.data.get(inputVals.complexTensorInfos.real.dataId).values;
+  const imagVals = cpuBackend.data.get(inputVals.complexTensorInfos.imag.dataId).values;
+  if (isExponentOf2(inputSize)) {
+    const result = fftRadix2(realVals, imagVals, inputSize, inverse, cpuBackend);
+    const resultShape = [input2.shape[0], input2.shape[1]];
+    if (inverse) {
+      const realInfo = cpuBackend.makeTensorInfo(resultShape, "float32", result.real);
+      const imagInfo = cpuBackend.makeTensorInfo(resultShape, "float32", result.imag);
+      const sizeInfo = cpuBackend.makeTensorInfo([], "float32", util_exports.createScalarValue(inputSize, "float32"));
+      const sizeInfoCopy = identity2({ inputs: { x: sizeInfo }, backend: cpuBackend });
+      const divRealInfo = realDivConfig.kernelFunc({ inputs: { a: realInfo, b: sizeInfo }, backend: cpuBackend });
+      const divImagInfo = realDivConfig.kernelFunc({ inputs: { a: imagInfo, b: sizeInfoCopy }, backend: cpuBackend });
+      const divRealVals = cpuBackend.data.get(divRealInfo.dataId).values;
+      const divImagVals = cpuBackend.data.get(divImagInfo.dataId).values;
+      cpuBackend.disposeIntermediateTensorInfo(realInfo);
+      cpuBackend.disposeIntermediateTensorInfo(imagInfo);
+      cpuBackend.disposeIntermediateTensorInfo(sizeInfo);
+      cpuBackend.disposeIntermediateTensorInfo(sizeInfoCopy);
+      cpuBackend.disposeIntermediateTensorInfo(divRealInfo);
+      cpuBackend.disposeIntermediateTensorInfo(divImagInfo);
+      return { real: divRealVals, imag: divImagVals };
+    }
+    return result;
+  } else {
+    const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);
+    const rawOutput = fourierTransformByMatmul(data, inputSize, inverse);
+    return backend_util_exports.splitRealAndImagArrays(rawOutput);
+  }
+}
+function isExponentOf2(size) {
+  return (size & size - 1) === 0;
+}
+function fftRadix2(realVals, imagVals, size, inverse, cpuBackend) {
+  if (size === 1) {
+    return { real: realVals, imag: imagVals };
+  }
+  const data = backend_util_exports.mergeRealAndImagArrays(realVals, imagVals);
+  const half = size / 2;
+  const evenComplex = backend_util_exports.complexWithEvenIndex(data);
+  const evenRealVals = evenComplex.real;
+  const evenImagVals = evenComplex.imag;
+  const evenShape = [evenRealVals.length];
+  const evenRealInfo = cpuBackend.makeTensorInfo(evenShape, "float32", evenRealVals);
+  const evenImagInfo = cpuBackend.makeTensorInfo(evenShape, "float32", evenImagVals);
+  const evenTensorInfo = complex2({ inputs: { real: evenRealInfo, imag: evenImagInfo }, backend: cpuBackend });
+  const oddComplex = backend_util_exports.complexWithOddIndex(data);
+  const oddRealVals = oddComplex.real;
+  const oddImagVals = oddComplex.imag;
+  const oddShape = [oddRealVals.length];
+  const oddRealInfo = cpuBackend.makeTensorInfo(oddShape, "float32", oddRealVals);
+  const oddImagInfo = cpuBackend.makeTensorInfo(oddShape, "float32", oddImagVals);
+  const oddTensorInfo = complex2({ inputs: { real: oddRealInfo, imag: oddImagInfo }, backend: cpuBackend });
+  const $evenComplex = fftRadix2(evenRealVals, evenImagVals, half, inverse, cpuBackend);
+  const $evenRealVals = $evenComplex.real;
+  const $evenImagVals = $evenComplex.imag;
+  const $evenShape = [$evenRealVals.length];
+  const $evenRealInfo = cpuBackend.makeTensorInfo($evenShape, "float32", $evenRealVals);
+  const $evenImagInfo = cpuBackend.makeTensorInfo($evenShape, "float32", $evenImagVals);
+  const $evenTensorInfo = complex2({
+    inputs: { real: $evenRealInfo, imag: $evenImagInfo },
+    backend: cpuBackend
+  });
+  const $oddComplex = fftRadix2(oddRealVals, oddImagVals, half, inverse, cpuBackend);
+  const $oddRealVals = $oddComplex.real;
+  const $oddImagVals = $oddComplex.imag;
+  const $oddShape = [$oddRealVals.length];
+  const $oddRealInfo = cpuBackend.makeTensorInfo($oddShape, "float32", $oddRealVals);
+  const $oddImagInfo = cpuBackend.makeTensorInfo($oddShape, "float32", $oddImagVals);
+  const $oddTensorInfo = complex2({ inputs: { real: $oddRealInfo, imag: $oddImagInfo }, backend: cpuBackend });
+  const e = backend_util_exports.exponents(size, inverse);
+  const eShape = [e.real.length];
+  const eRealInfo = cpuBackend.makeTensorInfo(eShape, "float32", e.real);
+  const eImagInfo = cpuBackend.makeTensorInfo(eShape, "float32", e.imag);
+  const complexInfo = complex2({ inputs: { real: eRealInfo, imag: eImagInfo }, backend: cpuBackend });
+  const exponentInfo = multiply2({ inputs: { a: complexInfo, b: $oddTensorInfo }, backend: cpuBackend });
+  const addPart = add4({
+    inputs: { a: $evenTensorInfo, b: exponentInfo },
+    backend: cpuBackend
+  });
+  const subPart = sub2({
+    inputs: { a: $evenTensorInfo, b: exponentInfo },
+    backend: cpuBackend
+  });
+  const addPartReal = real2({ inputs: { input: addPart }, backend: cpuBackend });
+  const subPartReal = real2({ inputs: { input: subPart }, backend: cpuBackend });
+  const addPartImag = imag2({ inputs: { input: addPart }, backend: cpuBackend });
+  const subPartImag = imag2({ inputs: { input: subPart }, backend: cpuBackend });
+  const $real = concat2({
+    inputs: [addPartReal, subPartReal],
+    backend: cpuBackend,
+    attrs: { axis: 0 }
+  });
+  const $imag = concat2({
+    inputs: [addPartImag, subPartImag],
+    backend: cpuBackend,
+    attrs: { axis: 0 }
+  });
+  const $realVals = cpuBackend.data.get($real.dataId).values;
+  const $imagVals = cpuBackend.data.get($imag.dataId).values;
+  cpuBackend.disposeIntermediateTensorInfo(evenRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(evenImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(evenTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(oddTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo($evenTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo($oddTensorInfo);
+  cpuBackend.disposeIntermediateTensorInfo(eRealInfo);
+  cpuBackend.disposeIntermediateTensorInfo(eImagInfo);
+  cpuBackend.disposeIntermediateTensorInfo(complexInfo);
+  cpuBackend.disposeIntermediateTensorInfo(exponentInfo);
+  cpuBackend.disposeIntermediateTensorInfo(addPart);
+  cpuBackend.disposeIntermediateTensorInfo(subPart);
+  cpuBackend.disposeIntermediateTensorInfo(addPartReal);
+  cpuBackend.disposeIntermediateTensorInfo(addPartImag);
+  cpuBackend.disposeIntermediateTensorInfo(subPartReal);
+  cpuBackend.disposeIntermediateTensorInfo(subPartImag);
+  cpuBackend.disposeIntermediateTensorInfo($real);
+  cpuBackend.disposeIntermediateTensorInfo($imag);
+  return { real: $realVals, imag: $imagVals };
+}
+function fourierTransformByMatmul(data, size, inverse) {
+  const ret = new Float32Array(size * 2);
+  for (let r = 0; r < size; r++) {
+    let real4 = 0;
+    let imag4 = 0;
+    for (let c = 0; c < size; c++) {
+      const e = backend_util_exports.exponent(r * c, size, inverse);
+      const term = backend_util_exports.getComplexWithIndex(data, c);
+      real4 += term.real * e.real - term.imag * e.imag;
+      imag4 += term.real * e.imag + term.imag * e.real;
+    }
+    if (inverse) {
+      real4 /= size;
+      imag4 /= size;
+    }
+    backend_util_exports.assignToTypedArray(ret, real4, imag4, r);
+  }
+  return ret;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FFT.js
+function fft2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape3({
+    inputs: { x: input2 },
+    backend: backend2,
+    attrs: { shape: [batch, innerDimensionSize] }
+  });
+  const result = fftBatch(input2D, false, backend2);
+  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var fftConfig = {
+  kernelName: FFT,
+  backendName: "cpu",
+  kernelFunc: fft2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Fill.js
+function fill2(args) {
+  const { backend: backend2, attrs } = args;
+  const { shape, value, dtype } = attrs;
+  const $dtype = dtype || util_exports.inferDtype(value);
+  const values = util_exports.getArrayFromDType($dtype, util_exports.sizeFromShape(shape));
+  fillValues(values, value, $dtype);
+  return backend2.makeTensorInfo(shape, $dtype, values);
+}
+var fillConfig = {
+  kernelName: Fill,
+  backendName: "cpu",
+  kernelFunc: fill2
+};
+function fillValues(values, value, dtype) {
+  if (dtype === "string") {
+    values.fill(value);
+  } else {
+    values.fill(value);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FlipLeftRight.js
+var flipLeftRightConfig = {
+  kernelName: FlipLeftRight,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const cpuBackend = backend2;
+    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));
+    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+    const imageVals = cpuBackend.data.get(image2.dataId).values;
+    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {
+      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;
+      for (let row = 0; row < imageHeight; row++) {
+        const rowOffset = row * (imageWidth * numChannels);
+        for (let col = 0; col < imageWidth; col++) {
+          const colOffset = col * numChannels;
+          for (let channel = 0; channel < numChannels; channel++) {
+            const coordX = Math.round(imageWidth - col - 1);
+            const outIdx = batchOffset + rowOffset + colOffset + channel;
+            let outputValue = imageVals[outIdx];
+            if (coordX >= 0 && coordX < imageWidth) {
+              const rotatedColOffset = coordX * numChannels;
+              const imageIdx = batchOffset + rowOffset + rotatedColOffset + channel;
+              outputValue = imageVals[imageIdx];
+            }
+            output[outIdx] = outputValue;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);
+    return { dataId, shape: image2.shape, dtype: image2.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FloorDiv.js
+var floorDivImpl = createSimpleBinaryKernelImpl((a, b) => Math.floor(a / b));
+var floorDiv2 = binaryKernelFunc(FloorDiv, floorDivImpl, null, "int32");
+var floorDivConfig = {
+  kernelName: FloorDiv,
+  backendName: "cpu",
+  kernelFunc: floorDiv2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FusedConv2D.js
+function fusedConv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  let result = conv2D({
+    inputs: { x, filter },
+    backend: backend2,
+    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }
+  });
+  if (bias) {
+    const resultOld = result;
+    if (dataFormat === "NCHW" && bias.shape.length === 1 && bias.shape[0] !== 1) {
+      const reshapedBias = reshape3({ inputs: { x: bias }, backend: backend2, attrs: { shape: [bias.shape[0], 1, 1] } });
+      result = add4({ inputs: { a: result, b: reshapedBias }, backend: backend2 });
+      backend2.disposeIntermediateTensorInfo(reshapedBias);
+    } else {
+      result = add4({ inputs: { a: result, b: bias }, backend: backend2 });
+    }
+    backend2.disposeIntermediateTensorInfo(resultOld);
+  }
+  if (activation2) {
+    const resultOld = result;
+    if (dataFormat === "NCHW" && activation2 === "prelu" && preluActivationWeights.shape.length === 1 && preluActivationWeights.shape[0] !== 1) {
+      const reshapedAlpha = reshape3({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: [preluActivationWeights.shape[0], 1, 1] }
+      });
+      result = applyActivation2(backend2, result, activation2, reshapedAlpha, leakyreluAlpha);
+      backend2.disposeIntermediateTensorInfo(reshapedAlpha);
+    } else {
+      result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);
+    }
+    backend2.disposeIntermediateTensorInfo(resultOld);
+  }
+  return result;
+}
+var fusedConv2DConfig = {
+  kernelName: FusedConv2D,
+  backendName: "cpu",
+  kernelFunc: fusedConv2D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/FusedDepthwiseConv2D.js
+function fusedDepthwiseConv2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  let result = depthwiseConv2dNative({
+    inputs: { x, filter },
+    backend: backend2,
+    attrs: { strides, pad: pad3, dataFormat, dilations, dimRoundingMode }
+  });
+  if (bias) {
+    const oldResult = result;
+    result = add4({ inputs: { a: result, b: bias }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  if (activation2) {
+    const oldResult = result;
+    result = applyActivation2(backend2, result, activation2, preluActivationWeights, leakyreluAlpha);
+    backend2.disposeIntermediateTensorInfo(oldResult);
+  }
+  return result;
+}
+var fusedDepthwiseConv2DConfig = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "cpu",
+  kernelFunc: fusedDepthwiseConv2D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherNd.js
+function gatherNd(args) {
+  const { inputs, backend: backend2 } = args;
+  const { params, indices } = inputs;
+  const paramsSize = util_exports.sizeFromShape(params.shape);
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);
+  if (numSlices === 0) {
+    return backend2.makeTensorInfo(resultShape, params.dtype, []);
+  }
+  const indicesData = backend2.data.get(indices.dataId).values;
+  const paramsBuf = backend2.bufferSync(params);
+  const outBuf = gatherNdImpl(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);
+  return backend2.makeTensorInfo(resultShape, params.dtype, outBuf.values);
+}
+var gatherNdConfig = {
+  kernelName: GatherNd,
+  backendName: "cpu",
+  kernelFunc: gatherNd
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/GatherV2.js
+function gatherV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  assertNotComplex([x, indices], "gatherV2");
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const indicesVals = backend2.data.get(indices.dataId).values;
+  const axisDim = x.shape[parsedAxis];
+  for (let i = 0; i < indicesVals.length; ++i) {
+    const index = indicesVals[i];
+    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+  }
+  let $batchDims = batchDims;
+  if (batchDims == null) {
+    $batchDims = 0;
+  }
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, $batchDims);
+  const flattenX = reshape3({
+    inputs: { x },
+    backend: backend2,
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    }
+  });
+  const flattenIndex = reshape3({
+    inputs: { x: indices },
+    backend: backend2,
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }
+  });
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  const indicesBuf = backend2.bufferSync(flattenIndex);
+  const xBuf = backend2.bufferSync(flattenX);
+  const outBuf = gatherV2Impl(xBuf, indicesBuf, flattenOutputShape);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(flattenIndex);
+  return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);
+}
+var gatherV2Config = {
+  kernelName: GatherV2,
+  backendName: "cpu",
+  kernelFunc: gatherV2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IFFT.js
+function ifft2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputSize = util_exports.sizeFromShape(input2.shape);
+  const innerDimensionSize = input2.shape[input2.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape3({
+    inputs: { x: input2 },
+    backend: backend2,
+    attrs: { shape: [batch, innerDimensionSize] }
+  });
+  const result = fftBatch(input2D, true, backend2);
+  const resultReshaped = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: input2.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var ifftConfig = {
+  kernelName: IFFT,
+  backendName: "cpu",
+  kernelFunc: ifft2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsFinite.js
+var isFinite3 = unaryKernelFunc(IsFinite, (xi) => Number.isFinite(xi) ? 1 : 0, "bool");
+var isFiniteConfig = {
+  kernelName: IsFinite,
+  backendName: "cpu",
+  kernelFunc: isFinite3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsInf.js
+var isInf2 = unaryKernelFunc(IsInf, (xi) => Math.abs(xi) === Infinity ? 1 : 0, "bool");
+var isInfConfig = {
+  kernelName: IsInf,
+  backendName: "cpu",
+  kernelFunc: isInf2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/IsNaN.js
+var isNaN3 = unaryKernelFunc(IsNan, (xi) => Number.isNaN(xi) ? 1 : 0, "bool");
+var isNaNConfig = {
+  kernelName: IsNan,
+  backendName: "cpu",
+  kernelFunc: isNaN3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LinSpace.js
+function linSpace(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, num } = attrs;
+  const outVals = linSpaceImpl(start, stop, num);
+  return backend2.makeTensorInfo([outVals.length], "float32", outVals);
+}
+var linSpaceConfig = {
+  kernelName: LinSpace,
+  backendName: "cpu",
+  kernelFunc: linSpace
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Log1p.js
+var log1p2 = unaryKernelFunc(Log1p, (xi) => Math.log1p(xi));
+var log1pConfig = {
+  kernelName: Log1p,
+  backendName: "cpu",
+  kernelFunc: log1p2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalAnd.js
+var logicalAndImpl = createSimpleBinaryKernelImpl((a, b) => a && b);
+var logicalAnd2 = binaryKernelFunc(LogicalAnd, logicalAndImpl, null, "bool");
+var logicalAndConfig = {
+  kernelName: LogicalAnd,
+  backendName: "cpu",
+  kernelFunc: logicalAnd2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalNot.js
+var logicalNot2 = unaryKernelFunc(LogicalNot, (xi) => xi ? 0 : 1, "bool");
+var logicalNotConfig = {
+  kernelName: LogicalNot,
+  backendName: "cpu",
+  kernelFunc: logicalNot2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LogicalOr.js
+var logicalOrImpl = createSimpleBinaryKernelImpl((a, b) => a || b);
+var logicalOr2 = binaryKernelFunc(LogicalOr, logicalOrImpl, null, "bool");
+var logicalOrConfig = {
+  kernelName: LogicalOr,
+  backendName: "cpu",
+  kernelFunc: logicalOr2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LRN.js
+function lRN(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  assertNotComplex(x, "LRN");
+  const channels = x.shape[3];
+  const maxD = channels - 1;
+  const xValues = backend2.data.get(x.dataId).values;
+  const size = util_exports.sizeFromShape(x.shape);
+  const result = new Float32Array(size);
+  function sumAcrossChannels(offset) {
+    const currentChannel = offset % channels;
+    let beginSumOffset = offset - currentChannel + Math.max(0, currentChannel - depthRadius);
+    const endSumOffset = offset - currentChannel + Math.min(currentChannel + depthRadius, maxD);
+    let sum6 = 0;
+    for (; beginSumOffset <= endSumOffset; beginSumOffset++) {
+      const z = xValues[beginSumOffset];
+      sum6 += z * z;
+    }
+    return sum6;
+  }
+  for (let offset = 0; offset < size; offset++) {
+    const sum6 = sumAcrossChannels(offset);
+    const val = xValues[offset] * Math.pow(bias + alpha * sum6, -beta);
+    result[offset] = val;
+  }
+  return backend2.makeTensorInfo(x.shape, x.dtype, result);
+}
+var LRNConfig = {
+  kernelName: LRN,
+  backendName: "cpu",
+  kernelFunc: lRN
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/LRNGrad.js
+function lRNGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, y, dy } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  assertNotComplex(dy, "LRNGrad");
+  const dySize = util_exports.sizeFromShape(dy.shape);
+  const channels = dy.shape[3];
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const xValues = backend2.data.get(x.dataId).values;
+  const yValues = backend2.data.get(y.dataId).values;
+  const result = new Float32Array(dySize);
+  const size = dySize;
+  for (let offset = 0; offset < size; offset++) {
+    const currentChannel = offset % channels;
+    const depthBegin = offset - currentChannel + Math.max(0, currentChannel - depthRadius);
+    const depthEnd = offset - currentChannel + Math.min(channels, currentChannel + depthRadius + 1);
+    let norm2 = 0;
+    for (let k = depthBegin; k < depthEnd; k++) {
+      norm2 += Math.pow(xValues[k], 2);
+    }
+    norm2 = alpha * norm2 + bias;
+    for (let k = depthBegin; k < depthEnd; k++) {
+      let dyi = -2 * alpha * beta * xValues[k] * yValues[offset] / norm2;
+      if (offset === k) {
+        dyi += Math.pow(norm2, -beta);
+      }
+      dyi *= dyValues[offset];
+      result[k] += dyi;
+    }
+  }
+  return backend2.makeTensorInfo(dy.shape, x.dtype, result);
+}
+var LRNGradConfig = {
+  kernelName: LRNGrad,
+  backendName: "cpu",
+  kernelFunc: lRNGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Max.js
+function max3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reductionIndices, keepDims } = attrs;
+  const cpuBackend = backend2;
+  let xShape = x.shape;
+  const xRank = xShape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, xShape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let xVals = cpuBackend.data.get(x.dataId).values;
+  if (permutedAxes != null) {
+    const newShape = new Array(xRank);
+    for (let i = 0; i < newShape.length; i++) {
+      newShape[i] = xShape[permutedAxes[i]];
+    }
+    xVals = transposeImpl(xVals, xShape, x.dtype, permutedAxes, newShape);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    xShape = newShape;
+  }
+  assertNotComplex(x, "max");
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, xRank);
+  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xShape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const result = maxImpl(xVals, reduceSize, maxOutShape, x.dtype);
+  const dataId = cpuBackend.write(result, maxOutShape, x.dtype);
+  let outShape = maxOutShape;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);
+    outShape = newShape;
+  }
+  return { dataId, shape: outShape, dtype: x.dtype };
+}
+var maxConfig = {
+  kernelName: Max,
+  backendName: "cpu",
+  kernelFunc: max3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool.js
+function maxPool2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex(x, "maxPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  let res;
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    res = identity2({ inputs: { x }, backend: backend2 });
+  } else {
+    const xValues = backend2.data.get(x.dataId).values;
+    const strides2 = util_exports.computeStrides(x.shape);
+    const buffer2 = pool2(xValues, x.shape, x.dtype, strides2, convInfo, "max");
+    res = backend2.makeTensorInfo(convInfo.outShape, x.dtype, buffer2.values);
+  }
+  return res;
+}
+var maxPoolConfig = {
+  kernelName: MaxPool,
+  backendName: "cpu",
+  kernelFunc: maxPool2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool3D.js
+function maxPool3D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  assertNotComplex(x, "maxPool3d");
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode, dataFormat);
+  const xValues = backend2.data.get(x.dataId).values;
+  const outBuf = pool3d2(xValues, x.shape, x.dtype, util_exports.computeStrides(x.shape), convInfo, "max");
+  return backend2.makeTensorInfo(outBuf.shape, "float32", outBuf.values);
+}
+var maxPool3DConfig = {
+  kernelName: MaxPool3D,
+  backendName: "cpu",
+  kernelFunc: maxPool3D
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPool3DGrad.js
+function maxPool3DGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  assertNotComplex([dy, input2], "maxPool3DGrad");
+  const convInfo = backend_util_exports.computePool3DInfo(input2.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const inputBuf = backend2.bufferSync(input2);
+  const maxPosBuf = maxPool3dPositions(inputBuf, convInfo);
+  const strideDepth = convInfo.strideDepth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationDepth = convInfo.dilationDepth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(input2.shape, "float32");
+  const dyBuf = backend2.bufferSync(dy);
+  for (let batch = 0; batch < convInfo.batchSize; ++batch) {
+    for (let channel = 0; channel < convInfo.inChannels; ++channel) {
+      for (let dxDepth = 0; dxDepth < convInfo.inDepth; ++dxDepth) {
+        for (let dxRow = 0; dxRow < convInfo.inHeight; ++dxRow) {
+          for (let dxCol = 0; dxCol < convInfo.inWidth; ++dxCol) {
+            const dyDepthCorner = dxDepth - padFront;
+            const dyRowCorner = dxRow - padTop;
+            const dyColCorner = dxCol - padLeft;
+            let dotProd = 0;
+            for (let wDepth = 0; wDepth < effectiveFilterDepth; wDepth += dilationDepth) {
+              const dyDepth = (dyDepthCorner + wDepth) / strideDepth;
+              if (dyDepth < 0 || dyDepth >= convInfo.outDepth || Math.floor(dyDepth) !== dyDepth) {
+                continue;
+              }
+              for (let wRow = 0; wRow < effectiveFilterHeight; wRow += dilationHeight) {
+                const dyRow = (dyRowCorner + wRow) / strideHeight;
+                if (dyRow < 0 || dyRow >= convInfo.outHeight || Math.floor(dyRow) !== dyRow) {
+                  continue;
+                }
+                for (let wCol = 0; wCol < effectiveFilterWidth; wCol += dilationWidth) {
+                  const dyCol = (dyColCorner + wCol) / strideWidth;
+                  if (dyCol < 0 || dyCol >= convInfo.outWidth || Math.floor(dyCol) !== dyCol) {
+                    continue;
+                  }
+                  const maxPos = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  const curPos = wDepth * effectiveFilterHeight * effectiveFilterWidth + wRow * effectiveFilterWidth + wCol;
+                  const mask = maxPos === curPos ? 1 : 0;
+                  if (mask === 0) {
+                    continue;
+                  }
+                  const pixel = dyBuf.get(batch, dyDepth, dyRow, dyCol, channel);
+                  dotProd += pixel * mask;
+                }
+              }
+            }
+            dx.set(dotProd, batch, dxDepth, dxRow, dxCol, channel);
+          }
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var maxPool3DGradConfig2 = {
+  kernelName: MaxPool3DGrad,
+  backendName: "cpu",
+  kernelFunc: maxPool3DGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolGrad.js
+function maxPoolGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2, output } = inputs;
+  const x = input2;
+  assertNotComplex([input2, output], "maxPoolGrad");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const xValues = backend2.data.get(x.dataId).values;
+  const maxPosBuf = buffer(convInfo.outShape, x.dtype, maxPoolPositions(xValues, x.shape, x.dtype, convInfo).values);
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+  const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+  const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+  const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+  const dx = buffer(x.shape, "float32");
+  const dyData = backend2.data.get(dy.dataId).values;
+  const dyBuf = buffer(dy.shape, "float32", dyData);
+  for (let b = 0; b < convInfo.batchSize; ++b) {
+    for (let d = 0; d < convInfo.inChannels; ++d) {
+      for (let dxR = 0; dxR < convInfo.inHeight; ++dxR) {
+        for (let dxC = 0; dxC < convInfo.inWidth; ++dxC) {
+          const dyRCorner = dxR - padTop;
+          const dyCCorner = dxC - padLeft;
+          let dotProd = 0;
+          for (let wR = 0; wR < effectiveFilterHeight; wR += dilationHeight) {
+            const dyR = (dyRCorner + wR) / strideHeight;
+            if (dyR < 0 || dyR >= convInfo.outHeight || Math.floor(dyR) !== dyR) {
+              continue;
+            }
+            for (let wC = 0; wC < effectiveFilterWidth; wC += dilationWidth) {
+              const dyC = (dyCCorner + wC) / strideWidth;
+              if (dyC < 0 || dyC >= convInfo.outWidth || Math.floor(dyC) !== dyC) {
+                continue;
+              }
+              const maxPos = effectiveFilterHeight * effectiveFilterWidth - 1 - maxPosBuf.get(b, dyR, dyC, d);
+              const curPos = wR * effectiveFilterWidth + wC;
+              const mask = maxPos === curPos ? 1 : 0;
+              if (mask === 0) {
+                continue;
+              }
+              const pixel = dyBuf.get(b, dyR, dyC, d);
+              dotProd += pixel * mask;
+            }
+          }
+          dx.set(dotProd, b, dxR, dxC, d);
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(dx.shape, dx.dtype, dx.values);
+}
+var maxPoolGradConfig2 = {
+  kernelName: MaxPoolGrad,
+  backendName: "cpu",
+  kernelFunc: maxPoolGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolWithArgmax_impl.js
+function maxPoolWithArgmaxImpl(xValues, xShape, dtype, includeBatchInIndex, convInfo) {
+  const strides = util_exports.computeStrides(xShape);
+  const maxPools = pool2(xValues, xShape, dtype, strides, convInfo, "max");
+  const maxPositions = maxPoolPositions(xValues, xShape, dtype, convInfo, true, includeBatchInIndex);
+  return [maxPools.values, maxPositions.values];
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MaxPoolWithArgmax.js
+var maxPoolWithArgmaxConfig = {
+  kernelName: MaxPoolWithArgmax,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;
+    const cpuBackend = backend2;
+    assertNotComplex(x, "MaxPoolWithArgmax");
+    const values = cpuBackend.data.get(x.dataId).values;
+    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, [1, 1], pad3);
+    const [pooled, indexes] = maxPoolWithArgmaxImpl(values, x.shape, x.dtype, includeBatchInIndex, convInfo);
+    const pooledDataId = cpuBackend.write(pooled, convInfo.outShape, x.dtype);
+    const indexesDataId = cpuBackend.write(indexes, convInfo.outShape, x.dtype);
+    return [
+      { dataId: pooledDataId, shape: convInfo.outShape, dtype: x.dtype },
+      { dataId: indexesDataId, shape: convInfo.outShape, dtype: "int32" }
+    ];
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Mean.js
+function mean2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const axes = util_exports.parseAxisParam(axis, x.shape);
+  const shapes = backend_util_exports.computeOutAndReduceShapes(x.shape, axes);
+  const reduceShape = shapes[1];
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const toDispose = [];
+  const reduceSizeScalar = backend2.makeTensorInfo([], "float32", new Float32Array([reduceSize]));
+  toDispose.push(reduceSizeScalar);
+  const $x = cast3({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+  toDispose.push($x);
+  const res = div2({ inputs: { a: $x, b: reduceSizeScalar }, backend: backend2 });
+  toDispose.push(res);
+  const result = sum3({ inputs: { x: res }, backend: backend2, attrs: { axis, keepDims } });
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var meanConfig = {
+  kernelName: Mean,
+  backendName: "cpu",
+  kernelFunc: mean2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Min.js
+function min3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  assertNotComplex(x, "min");
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  if (permutedAxes != null) {
+    $x = transpose2({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, $x.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes($x.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const vals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(outShape), $x.dtype);
+  const aVals = backend2.data.get($x.dataId).values;
+  for (let i = 0; i < vals.length; ++i) {
+    const offset = i * reduceSize;
+    let min6 = aVals[offset];
+    for (let j = 0; j < reduceSize; ++j) {
+      const value = aVals[offset + j];
+      if (Number.isNaN(value) || value < min6) {
+        min6 = value;
+      }
+    }
+    vals[i] = min6;
+  }
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo($x);
+  }
+  const result = backend2.makeTensorInfo(outShape, $x.dtype, vals);
+  if (keepDims) {
+    const expandedShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    const reshapedResult = reshape3({ inputs: { x: result }, backend: backend2, attrs: { shape: expandedShape } });
+    backend2.disposeIntermediateTensorInfo(result);
+    return reshapedResult;
+  }
+  return result;
+}
+var minConfig = {
+  kernelName: Min,
+  backendName: "cpu",
+  kernelFunc: min3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/MirrorPad.js
+function mirrorPad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, mode } = attrs;
+  assertNotComplex(x, "mirrorPad");
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const start = paddings.map((p2) => p2[0]);
+  const end = paddings.map((p2, i) => p2[0] + x.shape[i]);
+  const offset = mode === "reflect" ? 0 : 1;
+  const xVals = backend2.data.get(x.dataId).values;
+  const xRank = x.shape.length;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const resultSize = util_exports.sizeFromShape(outShape);
+  const resultRank = outShape.length;
+  const resultStrides = util_exports.computeStrides(outShape);
+  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);
+  for (let i = 0; i < resultSize; i++) {
+    let coords2 = util_exports.indexToLoc(i, resultRank, resultStrides);
+    for (let i2 = 0; i2 < resultRank; i2++) {
+      if (coords2[i2] < start[i2]) {
+        coords2[i2] = start[i2] * 2 - coords2[i2] - offset;
+      } else if (coords2[i2] >= end[i2]) {
+        coords2[i2] = (end[i2] - 1) * 2 - coords2[i2] + offset;
+      }
+    }
+    coords2 = coords2.map((c, i2) => c - start[i2]);
+    const inIndex = util_exports.locToIndex(coords2, xRank, xStrides);
+    resVals[i] = xVals[inIndex];
+  }
+  const outId = backend2.write(resVals, outShape, x.dtype);
+  return { dataId: outId, shape: outShape, dtype: x.dtype };
+}
+var mirrorPadConfig = {
+  kernelName: MirrorPad,
+  backendName: "cpu",
+  kernelFunc: mirrorPad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Mod.js
+var modImpl = createSimpleBinaryKernelImpl((aValue, bValue) => {
+  const rem = aValue % bValue;
+  if (aValue < 0 && bValue < 0 || aValue >= 0 && bValue >= 0) {
+    return rem;
+  } else {
+    return (rem + bValue) % bValue;
+  }
+});
+var mod2 = binaryKernelFunc(Mod, modImpl);
+var modConfig = {
+  kernelName: Mod,
+  backendName: "cpu",
+  kernelFunc: mod2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multinomial.js
+var seedrandom4 = __toESM(require_seedrandom2());
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Softmax.js
+function softmax3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { dim } = attrs;
+  const logitsRank = logits.shape.length;
+  let $dim = dim;
+  if ($dim === -1) {
+    $dim = logitsRank - 1;
+  }
+  if ($dim !== logitsRank - 1) {
+    throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${logitsRank} and dim was ${$dim}`);
+  }
+  const axes = util_exports.parseAxisParam([$dim], logits.shape);
+  const maxLogit = max3({
+    inputs: { x: logits },
+    backend: backend2,
+    attrs: { reductionIndices: axes, keepDims: false }
+  });
+  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);
+  const maxLogitReshaped = reshape3({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });
+  const a = sub2({ inputs: { a: logits, b: maxLogitReshaped }, backend: backend2 });
+  const b = exp2({ inputs: { x: a }, backend: backend2 });
+  const sumExp = sum3({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });
+  const sumReshaped = reshape3({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });
+  const result = div2({ inputs: { a: b, b: sumReshaped }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(maxLogit);
+  backend2.disposeIntermediateTensorInfo(maxLogitReshaped);
+  backend2.disposeIntermediateTensorInfo(a);
+  backend2.disposeIntermediateTensorInfo(b);
+  backend2.disposeIntermediateTensorInfo(sumExp);
+  backend2.disposeIntermediateTensorInfo(sumReshaped);
+  return result;
+}
+var softmaxConfig = {
+  kernelName: Softmax,
+  backendName: "cpu",
+  kernelFunc: softmax3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Multinomial.js
+function multinomial2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { numSamples, seed, normalized } = attrs;
+  assertNotComplex(logits, "multinomial");
+  const probabilities = normalized ? logits : softmax3({ inputs: { logits }, backend: backend2, attrs: { dim: -1 } });
+  const batchSize = probabilities.shape[0];
+  const numEvents = probabilities.shape[1];
+  const probVals = backend2.data.get(probabilities.dataId).values;
+  const resShape = [batchSize, numSamples];
+  const resVals = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(resShape), "int32");
+  for (let b = 0; b < batchSize; ++b) {
+    const offset = b * numEvents;
+    const cdf = new Float32Array(numEvents - 1);
+    cdf[0] = probVals[offset];
+    for (let event = 1; event < cdf.length; ++event) {
+      cdf[event] = cdf[event - 1] + probVals[offset + event];
+    }
+    const random = seedrandom4.alea(seed.toString());
+    const outOffset = b * numSamples;
+    for (let sampleId = 0; sampleId < numSamples; ++sampleId) {
+      const r = random();
+      resVals[outOffset + sampleId] = cdf.length;
+      for (let event = 0; event < cdf.length; event++) {
+        if (r < cdf[event]) {
+          resVals[outOffset + sampleId] = event;
+          break;
+        }
+      }
+    }
+  }
+  if (!normalized) {
+    backend2.disposeIntermediateTensorInfo(probabilities);
+  }
+  return backend2.makeTensorInfo(resShape, "int32", resVals);
+}
+var multinomialConfig = {
+  kernelName: Multinomial,
+  backendName: "cpu",
+  kernelFunc: multinomial2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV3.js
+var nonMaxSuppressionV3Impl2 = kernel_impls_exports.nonMaxSuppressionV3Impl;
+function nonMaxSuppressionV3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppression");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const { selectedIndices } = nonMaxSuppressionV3Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  return backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices));
+}
+var nonMaxSuppressionV3Config = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV4.js
+var nonMaxSuppressionV4Impl2 = kernel_impls_exports.nonMaxSuppressionV4Impl;
+function nonMaxSuppressionV4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppressionPadded");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl2(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([], "int32", new Int32Array([validOutputs]))
+  ];
+}
+var nonMaxSuppressionV4Config = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV4
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/NonMaxSuppressionV5.js
+var nonMaxSuppressionV5Impl2 = kernel_impls_exports.nonMaxSuppressionV5Impl;
+function nonMaxSuppressionV5(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;
+  assertNotComplex(boxes, "NonMaxSuppressionWithScore");
+  const boxesVals = backend2.data.get(boxes.dataId).values;
+  const scoresVals = backend2.data.get(scores.dataId).values;
+  const maxOutputSizeVal = maxOutputSize;
+  const iouThresholdVal = iouThreshold;
+  const scoreThresholdVal = scoreThreshold;
+  const softNmsSigmaVal = softNmsSigma;
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl2(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([selectedScores.length], "float32", new Float32Array(selectedScores))
+  ];
+}
+var nonMaxSuppressionV5Config = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "cpu",
+  kernelFunc: nonMaxSuppressionV5
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/OneHot.js
+function oneHot2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  assertNotComplex(indices, "oneHot");
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const res = new Float32Array(indicesSize * depth);
+  res.fill(offValue);
+  const indicesVal = backend2.data.get(indices.dataId).values;
+  for (let event = 0; event < indicesSize; ++event) {
+    if (indicesVal[event] >= 0 && indicesVal[event] < depth) {
+      res[event * depth + indicesVal[event]] = onValue;
+    }
+  }
+  return backend2.makeTensorInfo([...indices.shape, depth], dtype, res);
+}
+var oneHotConfig = {
+  kernelName: OneHot,
+  backendName: "cpu",
+  kernelFunc: oneHot2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ZerosLike.js
+function zerosLike2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("zerosLike is not supported for string tensors");
+  } else if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const r = zerosLike2({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 0, dtype: x.dtype } });
+  }
+}
+var zerosLikeConfig = {
+  kernelName: ZerosLike,
+  backendName: "cpu",
+  kernelFunc: zerosLike2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/OnesLike.js
+function onesLike2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("onesLike is not supported for string tensors");
+  } else if (x.dtype === "complex64") {
+    const realPart = real2({ inputs: { input: x }, backend: backend2 });
+    const r = onesLike2({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag2({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike2({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex2({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill2({ backend: backend2, attrs: { shape: x.shape, value: 1, dtype: x.dtype } });
+  }
+}
+var onesLikeConfig = {
+  kernelName: OnesLike,
+  backendName: "cpu",
+  kernelFunc: onesLike2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Pack.js
+function pack(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims3({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims3({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat2({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var packConfig = {
+  kernelName: Pack,
+  backendName: "cpu",
+  kernelFunc: pack
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/PadV2.js
+function padV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, constantValue } = attrs;
+  assertNotComplex(x, "pad");
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const start = paddings.map((p2) => p2[0]);
+  const xVals = backend2.data.get(x.dataId).values;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const xRank = x.shape.length;
+  const xStrides = util_exports.computeStrides(x.shape);
+  const resultSize = util_exports.sizeFromShape(outShape);
+  const resultRank = outShape.length;
+  const resultStrides = util_exports.computeStrides(outShape);
+  const resVals = util_exports.getTypedArrayFromDType(x.dtype, resultSize);
+  if (constantValue !== 0) {
+    resVals.fill(constantValue);
+  }
+  for (let i = 0; i < xSize; i++) {
+    const coords2 = util_exports.indexToLoc(i, xRank, xStrides);
+    const outCoords = coords2.map((c, i2) => c + start[i2]);
+    const outIndex = util_exports.locToIndex(outCoords, resultRank, resultStrides);
+    resVals[outIndex] = xVals[i];
+  }
+  const outId = backend2.write(resVals, outShape, x.dtype);
+  return { dataId: outId, shape: outShape, dtype: x.dtype };
+}
+var padV2Config = {
+  kernelName: PadV2,
+  backendName: "cpu",
+  kernelFunc: padV2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Pow.js
+var powImpl = createSimpleBinaryKernelImpl((a, b) => Math.pow(a, b));
+var pow2 = binaryKernelFunc(Pow, powImpl);
+var powConfig = {
+  kernelName: Pow,
+  backendName: "cpu",
+  kernelFunc: pow2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedGather.js
+function raggedGather2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;
+  const { outputRaggedRank } = attrs;
+  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.data.get(t.dataId).values);
+  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);
+  const $paramsDenseValues = backend2.data.get(paramsDenseValues.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImpl($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);
+  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], "int32", splits));
+  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);
+  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);
+}
+var raggedGatherConfig = {
+  kernelName: RaggedGather,
+  backendName: "cpu",
+  kernelFunc: raggedGather2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedRange.js
+function raggedRange2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { starts, limits, deltas } = inputs;
+  const $starts = backend2.data.get(starts.dataId).values;
+  const $limits = backend2.data.get(limits.dataId).values;
+  const $deltas = backend2.data.get(deltas.dataId).values;
+  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImpl($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);
+  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], "int32", rtNestedSplitsData);
+  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);
+  return [rtNestedSplits, rtDenseValues];
+}
+var raggedRangeConfig = {
+  kernelName: RaggedRange,
+  backendName: "cpu",
+  kernelFunc: raggedRange2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RaggedTensorToTensor.js
+function raggedTensorToTensor2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { shape, values, defaultValue, rowPartitionTensors } = inputs;
+  const { rowPartitionTypes } = attrs;
+  const $shape = backend2.data.get(shape.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const $defaultValue = backend2.data.get(defaultValue.dataId).values;
+  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.data.get(t.dataId).values);
+  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);
+  const [outputShape, output] = raggedTensorToTensorImpl($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);
+  return backend2.makeTensorInfo(outputShape, values.dtype, output);
+}
+var raggedTensorToTensorConfig = {
+  kernelName: RaggedTensorToTensor,
+  backendName: "cpu",
+  kernelFunc: raggedTensorToTensor2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Range.js
+function range3(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, dtype, step: step5 } = attrs;
+  const values = rangeImpl(start, stop, step5, dtype);
+  return backend2.makeTensorInfo([values.length], dtype, values);
+}
+var rangeConfig = {
+  kernelName: Range,
+  backendName: "cpu",
+  kernelFunc: range3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reciprocal.js
+var reciprocal2 = unaryKernelFunc(Reciprocal, (xi) => 1 / xi);
+var reciprocalConfig = {
+  kernelName: Reciprocal,
+  backendName: "cpu",
+  kernelFunc: reciprocal2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeBilinear.js
+function resizeBilinear2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  assertNotComplex(images, "resizeBilinear");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const xValues = backend2.data.get(images.dataId).values;
+  const result = new Float32Array(util_exports.sizeFromShape([batch, newHeight, newWidth, numChannels]));
+  const effectiveInputSize = [
+    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+  ];
+  const effectiveOutputSize = [
+    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+  ];
+  let outputIdx = 0;
+  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];
+  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];
+  for (let b = 0; b < batch; b++) {
+    for (let r = 0; r < newHeight; r++) {
+      let sourceFracRow;
+      if (halfPixelCenters) {
+        sourceFracRow = effectiveRowSizeRatio * (r + 0.5) - 0.5;
+      } else {
+        sourceFracRow = effectiveRowSizeRatio * r;
+      }
+      const sourceRowFloor = Math.max(0, Math.floor(sourceFracRow));
+      const rowFrac = sourceFracRow - sourceRowFloor;
+      const sourceRowCeil = Math.min(oldHeight - 1, Math.ceil(sourceFracRow));
+      const topRowOffset = b * imagesStrides[0] + sourceRowFloor * imagesStrides[1];
+      const botRowOffset = b * imagesStrides[0] + sourceRowCeil * imagesStrides[1];
+      for (let c = 0; c < newWidth; c++) {
+        let sourceFracCol;
+        if (halfPixelCenters) {
+          sourceFracCol = effectiveColSizeRatio * (c + 0.5) - 0.5;
+        } else {
+          sourceFracCol = effectiveColSizeRatio * c;
+        }
+        const sourceColFloor = Math.max(0, Math.floor(sourceFracCol));
+        const colFrac = sourceFracCol - sourceColFloor;
+        const sourceColCeil = Math.min(oldWidth - 1, Math.ceil(sourceFracCol));
+        const topLeftOffest = topRowOffset + sourceColFloor * imagesStrides[2];
+        const botLeftOffset = botRowOffset + sourceColFloor * imagesStrides[2];
+        const topRightOffset = topRowOffset + sourceColCeil * imagesStrides[2];
+        const botRightOffest = botRowOffset + sourceColCeil * imagesStrides[2];
+        for (let d = 0; d < numChannels; d++) {
+          const topLeft = xValues[topLeftOffest + d];
+          const bottomLeft = xValues[botLeftOffset + d];
+          const topRight = xValues[topRightOffset + d];
+          const bottomRight = xValues[botRightOffest + d];
+          const top = topLeft + (topRight - topLeft) * colFrac;
+          const bottom = bottomLeft + (bottomRight - bottomLeft) * colFrac;
+          const newValue = top + (bottom - top) * rowFrac;
+          result[outputIdx++] = newValue;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], "float32", result);
+}
+var resizeBilinearConfig = {
+  kernelName: ResizeBilinear,
+  backendName: "cpu",
+  kernelFunc: resizeBilinear2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeBilinearGrad.js
+function resizeBilinearGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  assertNotComplex([dy, images], "resizeBilinearGrad");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [batch, xHeight, xWidth, depth] = images.shape;
+  const [, yHeight, yWidth] = dy.shape;
+  const output = new Float32Array(batch * xHeight * xWidth * depth);
+  const effectiveXSize = [
+    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+  ];
+  const effectiveYSize = [
+    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+  ];
+  const heightScale = effectiveXSize[0] / effectiveYSize[0];
+  const widthScale = effectiveXSize[1] / effectiveYSize[1];
+  const dyValues = backend2.data.get(dy.dataId).values;
+  let offset = 0;
+  for (let b = 0; b < batch; b++) {
+    const bOffset = b * imagesStrides[0];
+    for (let r = 0; r < yHeight; r++) {
+      const dxR = r * heightScale;
+      const topDxRIndex = Math.floor(dxR);
+      const bottomDxRIndex = Math.min(Math.ceil(dxR), xHeight - 1);
+      const topDxROffset = bOffset + topDxRIndex * imagesStrides[1];
+      const bottomDxROffset = bOffset + bottomDxRIndex * imagesStrides[1];
+      const dxRLerp = dxR - topDxRIndex;
+      const inverseDxRLerp = 1 - dxRLerp;
+      for (let c = 0; c < yWidth; c++) {
+        const dxC = c * widthScale;
+        const leftDxCIndex = Math.floor(dxC);
+        const rightDxCIndex = Math.min(Math.ceil(dxC), xWidth - 1);
+        const dxCLerp = dxC - leftDxCIndex;
+        const inverseDxCLerp = 1 - dxCLerp;
+        const topLeftRCOffset = topDxROffset + leftDxCIndex * imagesStrides[2];
+        const topRightRCOffset = topDxROffset + rightDxCIndex * imagesStrides[2];
+        const bottomLeftRCOffset = bottomDxROffset + leftDxCIndex * imagesStrides[2];
+        const bottomRightRCOffset = bottomDxROffset + rightDxCIndex * imagesStrides[2];
+        const inverseDxRLerpTimesInverseDxCLerp = inverseDxRLerp * inverseDxCLerp;
+        const inverseDxRLerpTimesDxCLerp = inverseDxRLerp * dxCLerp;
+        const dxRLerpTimesInverseDxCLerp = dxRLerp * inverseDxCLerp;
+        const dxRLerpTimesDxCLerp = dxRLerp * dxCLerp;
+        for (let d = 0; d < depth; d++) {
+          const dyVal = dyValues[offset++];
+          output[topLeftRCOffset + d] += dyVal * inverseDxRLerpTimesInverseDxCLerp;
+          output[topRightRCOffset + d] += dyVal * inverseDxRLerpTimesDxCLerp;
+          output[bottomLeftRCOffset + d] += dyVal * dxRLerpTimesInverseDxCLerp;
+          output[bottomRightRCOffset + d] += dyVal * dxRLerpTimesDxCLerp;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, xWidth, xHeight, depth], "float32", output);
+}
+var resizeBilinearGradConfig2 = {
+  kernelName: ResizeBilinearGrad,
+  backendName: "cpu",
+  kernelFunc: resizeBilinearGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeNearestNeighbor.js
+function resizeNearestNeighbor2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  assertNotComplex(images, "resizeNearestNeighbor");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const xValues = backend2.data.get(images.dataId).values;
+  const output = new Float32Array(batch * newHeight * newWidth * numChannels);
+  const effectiveInputSize = [
+    alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+    alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+  ];
+  const effectiveOutputSize = [
+    alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+    alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+  ];
+  const effectiveRowSizeRatio = effectiveInputSize[0] / effectiveOutputSize[0];
+  const effectiveColSizeRatio = effectiveInputSize[1] / effectiveOutputSize[1];
+  let outputOffset = 0;
+  for (let b = 0; b < batch; b++) {
+    const batchOffset = b * imagesStrides[0];
+    for (let r = 0; r < newHeight; r++) {
+      const sourceFracRow = halfPixelCenters ? effectiveRowSizeRatio * (r + 0.5) : effectiveRowSizeRatio * r;
+      let sourceNearestRow = Math.min(oldHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));
+      if (halfPixelCenters) {
+        sourceNearestRow = Math.max(0, sourceNearestRow);
+      }
+      const rowOffset = batchOffset + sourceNearestRow * imagesStrides[1];
+      for (let c = 0; c < newWidth; c++) {
+        const sourceFracCol = halfPixelCenters ? effectiveColSizeRatio * (c + 0.5) : effectiveColSizeRatio * c;
+        let sourceNearestCol = Math.min(oldWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));
+        if (halfPixelCenters) {
+          sourceNearestCol = Math.max(0, sourceNearestCol);
+        }
+        const colOffset = rowOffset + sourceNearestCol * imagesStrides[2];
+        for (let d = 0; d < numChannels; d++) {
+          const newVal = xValues[colOffset + d];
+          output[outputOffset++] = newVal;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo([batch, newHeight, newWidth, numChannels], images.dtype, output);
+}
+var resizeNearestNeighborConfig = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "cpu",
+  kernelFunc: resizeNearestNeighbor2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ResizeNearestNeighborGrad.js
+function resizeNearestNeighborGrad(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  assertNotComplex([dy, images], "resizeNearestNeighborGrad");
+  const imagesStrides = util_exports.computeStrides(images.shape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const [batch, xHeight, xWidth, depth] = images.shape;
+  const [, yHeight, yWidth] = dy.shape;
+  const output = new Float32Array(batch * xHeight * xWidth * depth);
+  const dyValues = backend2.data.get(dy.dataId).values;
+  const effectiveXSize = [
+    alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+    alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+  ];
+  const effectiveYSize = [
+    alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+    alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+  ];
+  const heightScale = effectiveXSize[0] / effectiveYSize[0];
+  const widthScale = effectiveXSize[1] / effectiveYSize[1];
+  const invHeightScale = 1 / heightScale;
+  const invWidthScale = 1 / widthScale;
+  const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+  const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+  for (let b = 0; b < batch; b++) {
+    const batchOffset = b * imagesStrides[0];
+    for (let r = 0; r < xHeight; r++) {
+      const rowOffset = batchOffset + r * imagesStrides[1];
+      const startRLerp = Math.floor(r * invHeightScale);
+      const startDyR = Math.floor(startRLerp - winHeight / 2);
+      for (let c = 0; c < xWidth; c++) {
+        const colOffset = rowOffset + c * imagesStrides[2];
+        const startCLerp = Math.floor(c * invWidthScale);
+        const startDyC = Math.floor(startCLerp - winWidth / 2);
+        for (let d = 0; d < depth; d++) {
+          let accum = 0;
+          for (let dyRIndex = 0; dyRIndex < winHeight; dyRIndex++) {
+            const dyR = dyRIndex + startDyR;
+            if (dyR < 0 || dyR >= yHeight) {
+              continue;
+            }
+            const dyROffset = batchOffset + dyR * dyStrides[1];
+            const sourceFracRow = dyR * heightScale;
+            const sourceNearestRow = Math.min(xHeight - 1, alignCorners ? Math.round(sourceFracRow) : Math.floor(sourceFracRow));
+            if (r !== sourceNearestRow) {
+              continue;
+            }
+            for (let dyCIndex = 0; dyCIndex < winWidth; dyCIndex++) {
+              const dyC = dyCIndex + startDyC;
+              if (dyC < 0 || dyC >= yWidth) {
+                continue;
+              }
+              const dyCOffset = dyROffset + dyC * dyStrides[2];
+              const sourceFracCol = dyC * widthScale;
+              const sourceNearestCol = Math.min(xWidth - 1, alignCorners ? Math.round(sourceFracCol) : Math.floor(sourceFracCol));
+              if (c === sourceNearestCol) {
+                accum += dyValues[dyCOffset + d];
+              }
+            }
+          }
+          output[colOffset + d] = accum;
+        }
+      }
+    }
+  }
+  return backend2.makeTensorInfo(images.shape, images.dtype, output);
+}
+var resizeNearestNeighborGradConfig2 = {
+  kernelName: ResizeNearestNeighborGrad,
+  backendName: "cpu",
+  kernelFunc: resizeNearestNeighborGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Reverse.js
+function reverse2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  assertNotComplex(x, "reverse");
+  const xRank = x.shape.length;
+  const $dims = util_exports.parseAxisParam(dims, x.shape);
+  if (xRank === 0) {
+    return identity2({ inputs: { x }, backend: backend2 });
+  }
+  const outBuf = new TensorBuffer(x.shape, x.dtype);
+  const xBuf = backend2.bufferSync(x);
+  for (let i = 0; i < outBuf.size; i++) {
+    const outLoc = outBuf.indexToLoc(i);
+    const inLoc = outLoc.slice();
+    $dims.forEach((d) => inLoc[d] = x.shape[d] - 1 - inLoc[d]);
+    outBuf.set(xBuf.get(...inLoc), ...outLoc);
+  }
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var reverseConfig = {
+  kernelName: Reverse,
+  backendName: "cpu",
+  kernelFunc: reverse2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/RotateWithOffset.js
+var rotateWithOffsetConfig = {
+  kernelName: RotateWithOffset,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const { radians, fillValue, center } = attrs;
+    const cpuBackend = backend2;
+    const output = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(image2.shape));
+    const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+    const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);
+    const fullOpacityValue = 255;
+    const sinFactor = Math.sin(radians);
+    const cosFactor = Math.cos(radians);
+    const imageVals = cpuBackend.data.get(image2.dataId).values;
+    for (let batchIdx = 0; batchIdx < batch; batchIdx++) {
+      const batchOffset = batchIdx * imageWidth * imageHeight * numChannels;
+      for (let row = 0; row < imageHeight; row++) {
+        const rowOffset = row * (imageWidth * numChannels);
+        for (let col = 0; col < imageWidth; col++) {
+          const colOffset = col * numChannels;
+          for (let channel = 0; channel < numChannels; channel++) {
+            const coords2 = [batch, row, col, channel];
+            const x = coords2[2];
+            const y = coords2[1];
+            let coordX = (x - centerX) * cosFactor - (y - centerY) * sinFactor;
+            let coordY = (x - centerX) * sinFactor + (y - centerY) * cosFactor;
+            coordX = Math.round(coordX + centerX);
+            coordY = Math.round(coordY + centerY);
+            let outputValue = fillValue;
+            if (typeof fillValue !== "number") {
+              if (channel === 3) {
+                outputValue = fullOpacityValue;
+              } else {
+                outputValue = fillValue[channel];
+              }
+            }
+            if (coordX >= 0 && coordX < imageWidth && coordY >= 0 && coordY < imageHeight) {
+              const rotatedRowOffset = coordY * (imageWidth * numChannels);
+              const rotatedColOffset = coordX * numChannels;
+              const imageIdx = batchOffset + rotatedRowOffset + rotatedColOffset + channel;
+              outputValue = imageVals[imageIdx];
+            }
+            const outIdx = batchOffset + rowOffset + colOffset + channel;
+            output[outIdx] = outputValue;
+          }
+        }
+      }
+    }
+    const dataId = cpuBackend.write(output, image2.shape, image2.dtype);
+    return { dataId, shape: image2.shape, dtype: image2.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Round.js
+var round3 = unaryKernelFunc(Round, (xi) => {
+  const base = Math.floor(xi);
+  if (xi - base < 0.5) {
+    return Math.floor(xi);
+  } else if (xi - base > 0.5) {
+    return Math.ceil(xi);
+  } else {
+    if (base % 2 === 0) {
+      return base;
+    } else {
+      return base + 1;
+    }
+  }
+});
+var roundConfig = {
+  kernelName: Round,
+  backendName: "cpu",
+  kernelFunc: round3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/ScatterNd.js
+function scatterNd(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);
+  const sumDupeIndices = true;
+  const indicesBuf = backend2.bufferSync(indices);
+  const updatesBuf = backend2.bufferSync(updates);
+  const outBuf = scatterImpl(indicesBuf, updatesBuf, shape, outputSize, sliceSize, numUpdates, sliceRank, strides, 0, sumDupeIndices);
+  return backend2.makeTensorInfo(shape, outBuf.dtype, outBuf.values);
+}
+var scatterNdConfig = {
+  kernelName: ScatterNd,
+  backendName: "cpu",
+  kernelFunc: scatterNd
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SearchSorted_impl.js
+function lowerBound2(array2, value) {
+  let left = 0;
+  let right = array2.length;
+  let mid = 0;
+  while (left < right) {
+    mid = Math.floor((left + right) / 2);
+    if (array2[mid] < value) {
+      left = mid + 1;
+    } else {
+      right = mid;
+    }
+  }
+  return right;
+}
+function upperBound2(array2, value) {
+  let left = 0;
+  let right = array2.length;
+  let mid = 0;
+  while (left < right) {
+    mid = Math.floor((left + right) / 2);
+    if (array2[mid] <= value) {
+      left = mid + 1;
+    } else {
+      right = mid;
+    }
+  }
+  return right;
+}
+function searchSortedImpl(sortedInputs, values, batchSize, numInputs, numValues, side) {
+  const output = util_exports.getArrayFromDType("int32", batchSize * numValues);
+  for (let b = 0; b < batchSize; ++b) {
+    const sortedInputsSlice = sortedInputs.slice(b * numInputs, (b + 1) * numInputs);
+    const outputOffset = b * numValues;
+    for (let i = 0; i < numValues; ++i) {
+      output[outputOffset + i] = side === "left" ? lowerBound2(sortedInputsSlice, values[i + outputOffset]) : upperBound2(sortedInputsSlice, values[i + outputOffset]);
+    }
+  }
+  return output;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SearchSorted.js
+function searchSorted2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sortedSequence, values } = inputs;
+  const { side } = attrs;
+  const $sortedSequence = backend2.data.get(sortedSequence.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const output = searchSortedImpl($sortedSequence, $values, sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);
+  return backend2.makeTensorInfo(values.shape, "int32", output);
+}
+var searchSortedConfig = {
+  kernelName: SearchSorted,
+  backendName: "cpu",
+  kernelFunc: searchSorted2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Select.js
+function select2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  assertNotComplex([condition, t, e], "select");
+  const conditionRank = condition.shape.length;
+  const values = backend2.data.get(condition.dataId).values;
+  const tValues = backend2.data.get(t.dataId).values;
+  const eValues = backend2.data.get(e.dataId).values;
+  const resultDtype = upcastType(t.dtype, e.dtype);
+  const newValues = util_exports.makeZerosTypedArray(util_exports.sizeFromShape(t.shape), resultDtype);
+  let index = 0;
+  const offset = conditionRank === 0 || conditionRank > 1 || t.shape.length === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));
+  for (let i = 0; i < values.length; i++) {
+    for (let j = 0; j < offset; j++) {
+      if (values[i] === 1) {
+        newValues[index++] = tValues[i];
+      } else {
+        newValues[index++] = eValues[i];
+      }
+    }
+  }
+  return backend2.makeTensorInfo(t.shape, resultDtype, newValues);
+}
+var selectConfig = {
+  kernelName: Select,
+  backendName: "cpu",
+  kernelFunc: select2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Selu.js
+var scaleAlpha = backend_util_exports.SELU_SCALEALPHA;
+var scale = backend_util_exports.SELU_SCALE;
+var selu2 = unaryKernelFunc(Selu, (xi) => {
+  if (xi >= 0) {
+    return scale * xi;
+  } else {
+    return scaleAlpha * (Math.exp(xi) - 1);
+  }
+});
+var seluConfig = {
+  kernelName: Selu,
+  backendName: "cpu",
+  kernelFunc: selu2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sign.js
+var sign2 = unaryKernelFunc(Sign, (xi) => {
+  if (xi < 0) {
+    return -1;
+  } else if (xi > 0) {
+    return 1;
+  } else {
+    return 0;
+  }
+});
+var signConfig = {
+  kernelName: Sign,
+  backendName: "cpu",
+  kernelFunc: sign2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sin.js
+var sin2 = unaryKernelFunc(Sin, (xi) => Math.sin(xi));
+var sinConfig = {
+  kernelName: Sin,
+  backendName: "cpu",
+  kernelFunc: sin2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Sinh.js
+var sinh2 = unaryKernelFunc(Sinh, (xi) => Math.sinh(xi));
+var sinhConfig = {
+  kernelName: Sinh,
+  backendName: "cpu",
+  kernelFunc: sinh2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Softplus.js
+var epsilon2 = 11920928955078125e-23;
+var threshold2 = Math.log(epsilon2) + 2;
+var softplus2 = unaryKernelFunc(Softplus, (xi) => {
+  const tooLarge = xi > -threshold2;
+  const tooSmall = xi < threshold2;
+  const expX = Math.exp(xi);
+  let result;
+  if (tooSmall) {
+    result = expX;
+  } else if (tooLarge) {
+    result = xi;
+  } else {
+    result = Math.log(1 + expX);
+  }
+  return result;
+});
+var softplusConfig = {
+  kernelName: Softplus,
+  backendName: "cpu",
+  kernelFunc: softplus2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SpaceToBatchND.js
+function spaceToBatchND2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  assertNotComplex([x], "spaceToBatchND");
+  const prod5 = util_exports.sizeFromShape(blockShape);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const paddedX = padV2Config.kernelFunc({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapeInputs = { x: paddedX };
+  const reshapeAttrs = { shape: reshapedPaddedShape };
+  const paddedXReshaped = reshape3({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });
+  const transposeInputs = { x: paddedXReshaped };
+  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };
+  const paddedXT = transpose2({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });
+  const resultReshapeInputs = { x: paddedXT };
+  const resultReshapeAttrs = { shape: flattenShape };
+  const result = reshape3({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });
+  backend2.disposeIntermediateTensorInfo(paddedX);
+  backend2.disposeIntermediateTensorInfo(paddedXReshaped);
+  backend2.disposeIntermediateTensorInfo(paddedXT);
+  return result;
+}
+var spaceToBatchNDConfig = {
+  kernelName: SpaceToBatchND,
+  backendName: "cpu",
+  kernelFunc: spaceToBatchND2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseFillEmptyRows.js
+function sparseFillEmptyRows2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  if (denseShape.shape.length !== 1) {
+    throw new Error(`Dense shape must be a vector, saw:
+        ${denseShape.shape}`);
+  }
+  if (indices.shape.length !== 2) {
+    throw new Error(`Indices must be a matrix, saw:
+        ${indices.shape}`);
+  }
+  if (values.shape.length !== 1) {
+    throw new Error(`Values must be a vector, saw:
+        ${values.shape}`);
+  }
+  if (defaultValue.shape.length !== 0) {
+    throw new Error(`Default value must be a scalar, saw:
+        ${defaultValue.shape}`);
+  }
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $values = backend2.data.get(values.dataId).values;
+  const $denseShape = backend2.data.get(denseShape.dataId).values;
+  const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImpl($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);
+  return [
+    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),
+    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),
+    backend2.makeTensorInfo([emptyRowIndicator.length], "bool", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),
+    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))
+  ];
+}
+var sparseFillEmptyRowsConfig = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "cpu",
+  kernelFunc: sparseFillEmptyRows2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseReshape.js
+function sparseReshape2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape
+        ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape
+        ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const $inputShape = Array.from(backend2.data.get(inputShape.dataId).values);
+  const $inputIndices = backend2.data.get(inputIndices.dataId).values;
+  const targetShape = Array.from(backend2.data.get(newShape.dataId).values);
+  const [newIndices, indicesShape, outputShape] = sparseReshapeImpl($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);
+  return [
+    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),
+    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))
+  ];
+}
+var sparseReshapeConfig = {
+  kernelName: SparseReshape,
+  backendName: "cpu",
+  kernelFunc: sparseReshape2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentMean.js
+function sparseSegmentMean2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+          ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+          ${segmentIds.shape}`);
+  }
+  if (indices.shape[0] !== segmentIds.shape[0]) {
+    throw new Error(`segmentIds and indices should have same size.`);
+  }
+  const $data = backend2.data.get(data.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $segmentIds = backend2.data.get(segmentIds.dataId).values;
+  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds, true);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentMeanConfig = {
+  kernelName: SparseSegmentMean,
+  backendName: "cpu",
+  kernelFunc: sparseSegmentMean2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseSegmentSum.js
+function sparseSegmentSum2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+         ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+         ${segmentIds.shape}`);
+  }
+  if (indices.shape[0] !== segmentIds.shape[0]) {
+    throw new Error(`segmentIds and indices should have same size.`);
+  }
+  const $data = backend2.data.get(data.dataId).values;
+  const $indices = backend2.data.get(indices.dataId).values;
+  const $segmentIds = backend2.data.get(segmentIds.dataId).values;
+  const [outputData, outputDataShape] = sparseSegmentReductionImpl($data, data.shape, data.dtype, $indices, $segmentIds);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentSumConfig = {
+  kernelName: SparseSegmentSum,
+  backendName: "cpu",
+  kernelFunc: sparseSegmentSum2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SparseToDense.js
+function sparseToDense2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sparseIndices, sparseValues, defaultValue } = inputs;
+  const { outputShape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);
+  const sumDupeIndices = false;
+  const indicesBuf = backend2.bufferSync(sparseIndices);
+  let outBuf;
+  switch (sparseValues.dtype) {
+    case "bool": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = Boolean(backend2.data.get(defaultValue.dataId).values[0]);
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "float32": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "int32": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = backend2.data.get(defaultValue.dataId).values[0];
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    case "string": {
+      const updatesBuf = backend2.bufferSync(sparseValues);
+      const $defaultValue = util_exports.decodeString(backend2.data.get(defaultValue.dataId).values[0]);
+      outBuf = scatterImpl(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+      break;
+    }
+    default:
+      throw new Error(`Unsupported type ${sparseValues.dtype}`);
+  }
+  return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);
+}
+var sparseToDenseConfig = {
+  kernelName: SparseToDense,
+  backendName: "cpu",
+  kernelFunc: sparseToDense2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/SplitV.js
+function splitV(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const begin = new Array(x.shape.length).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const sliceSize = [...size];
+    sliceSize[$axis] = s;
+    const sliceT = slice2({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });
+    begin[$axis] += s;
+    return sliceT;
+  });
+}
+var splitVConfig = {
+  kernelName: SplitV,
+  backendName: "cpu",
+  kernelFunc: splitV
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Square.js
+var squareConfig = {
+  kernelName: Square,
+  backendName: "cpu",
+  kernelFunc: ({ inputs, backend: backend2 }) => {
+    const { x } = inputs;
+    const cpuBackend = backend2;
+    assertNotComplex(x, "square");
+    const values = cpuBackend.data.get(x.dataId).values;
+    const newValues = new Float32Array(values.length);
+    for (let i = 0; i < values.length; ++i) {
+      const value = values[i];
+      newValues[i] = value * value;
+    }
+    const dataId = cpuBackend.write(newValues, x.shape, x.dtype);
+    return { dataId, shape: x.shape, dtype: x.dtype };
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Step.js
+var step2 = unaryKernelFunc(Step, (xi, attrs) => {
+  const stepAttrs = attrs;
+  if (isNaN(xi)) {
+    return NaN;
+  } else {
+    return xi > 0 ? 1 : stepAttrs.alpha;
+  }
+});
+var stepConfig = {
+  kernelName: Step,
+  backendName: "cpu",
+  kernelFunc: step2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StridedSlice.js
+function stridedSlice2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  assertNotComplex(x, "stridedSlice");
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape3({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice2({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape3({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeIntermediateTensorInfo(sliced);
+  } else {
+    const xBuf = backend2.bufferSync(x);
+    const outBuf = stridedSliceImpl(finalShapeSparse, xBuf, $strides, $begin);
+    result = backend2.makeTensorInfo(finalShape, outBuf.dtype, outBuf.values);
+  }
+  return result;
+}
+var stridedSliceConfig = {
+  kernelName: StridedSlice,
+  backendName: "cpu",
+  kernelFunc: stridedSlice2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringNGrams.js
+function stringNGrams2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const { data, dataSplits } = inputs;
+  const $data = backend2.data.get(data.dataId).values;
+  const $dataSplits = backend2.data.get(dataSplits.dataId).values;
+  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  return [
+    backend2.makeTensorInfo([nGrams.length], "string", nGrams),
+    backend2.makeTensorInfo(dataSplits.shape, "int32", nGramsSplits)
+  ];
+}
+var stringNGramsConfig = {
+  kernelName: StringNGrams,
+  backendName: "cpu",
+  kernelFunc: stringNGrams2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringSplit.js
+function stringSplit2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { skipEmpty } = attrs;
+  const { input: input2, delimiter } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (input2.shape.length !== 1) {
+    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);
+  }
+  if (delimiter.shape.length !== 0) {
+    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);
+  }
+  const $input = backend2.data.get(input2.dataId).values;
+  const $delimiter = backend2.data.get(delimiter.dataId).values[0];
+  const [indices, values, shape] = stringSplitImpl($input, $delimiter, skipEmpty);
+  const outputSize = values.length;
+  return [
+    backend2.makeTensorInfo([outputSize, 2], "int32", indices),
+    backend2.makeTensorInfo([outputSize], "string", values),
+    backend2.makeTensorInfo([2], "int32", new Int32Array(shape))
+  ];
+}
+var stringSplitConfig = {
+  kernelName: StringSplit,
+  backendName: "cpu",
+  kernelFunc: stringSplit2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/StringToHashBucketFast.js
+function stringToHashBucketFast2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { numBuckets } = attrs;
+  const { input: input2 } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const $input = backend2.data.get(input2.dataId).values;
+  const output = stringToHashBucketFastImpl($input, numBuckets);
+  return backend2.makeTensorInfo(input2.shape, "int32", output);
+}
+var stringToHashBucketFastConfig = {
+  kernelName: StringToHashBucketFast,
+  backendName: "cpu",
+  kernelFunc: stringToHashBucketFast2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tan.js
+var tan2 = unaryKernelFunc(Tan, (xi) => Math.tan(xi));
+var tanConfig = {
+  kernelName: Tan,
+  backendName: "cpu",
+  kernelFunc: tan2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tanh.js
+var tanh3 = unaryKernelFunc(Tanh, (xi) => Math.tanh(xi));
+var tanhConfig = {
+  kernelName: Tanh,
+  backendName: "cpu",
+  kernelFunc: tanh3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Tile.js
+function tile3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reps } = attrs;
+  assertNotComplex(x, "tile");
+  const outBuf = tileImpl(backend2.bufferSync(x), reps);
+  return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+}
+var tileConfig = {
+  kernelName: Tile,
+  backendName: "cpu",
+  kernelFunc: tile3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/TopK.js
+function topK(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  assertNotComplex(x, "topk");
+  const xVals = backend2.data.get(x.dataId).values;
+  const [allTopKVals, allTopKIndices] = topKImpl(xVals, x.shape, x.dtype, k, sorted);
+  return [
+    backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),
+    backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)
+  ];
+}
+var topKConfig = {
+  kernelName: TopK,
+  backendName: "cpu",
+  kernelFunc: topK
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Transform.js
+function transform2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [batch, outHeight, outWidth, numChannels];
+  const inStrides = util_exports.computeStrides(image2.shape);
+  const batchInStride = inStrides[0];
+  const rowInStride = inStrides[1];
+  const colInStride = inStrides[2];
+  const outStrides = util_exports.computeStrides(outShape);
+  const batchOutStride = outStrides[0];
+  const rowOutStride = outStrides[1];
+  const colOutStride = outStrides[2];
+  const outVals = util_exports.getTypedArrayFromDType(image2.dtype, util_exports.sizeFromShape(outShape));
+  outVals.fill(fillValue);
+  const imageVals = backend2.data.get(image2.dataId).values;
+  const transformVals = backend2.data.get(transforms.dataId).values;
+  for (let b = 0; b < batch; ++b) {
+    const transform5 = transforms.shape[0] === 1 ? transformVals : transformVals.subarray(b * 8, b * 8 + 8);
+    for (let outY = 0; outY < outHeight; ++outY) {
+      for (let outX = 0; outX < outWidth; ++outX) {
+        for (let channel = 0; channel < numChannels; ++channel) {
+          let val;
+          const projection = transform5[6] * outX + transform5[7] * outY + 1;
+          if (projection === 0) {
+            continue;
+          }
+          const inX = (transform5[0] * outX + transform5[1] * outY + transform5[2]) / projection;
+          const inY = (transform5[3] * outX + transform5[4] * outY + transform5[5]) / projection;
+          const x = mapCoord(inX, imageWidth, fillMode);
+          const y = mapCoord(inY, imageHeight, fillMode);
+          switch (interpolation) {
+            case "nearest":
+              val = nearestInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);
+              break;
+            case "bilinear":
+              val = bilinearInterpolation(imageVals, imageHeight, imageWidth, batchInStride, rowInStride, colInStride, b, y, x, channel, fillValue);
+              break;
+            default:
+              throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${interpolation}`);
+          }
+          const ind = b * batchOutStride + outY * rowOutStride + outX * colOutStride + channel;
+          outVals[ind] = val;
+        }
+      }
+    }
+    return backend2.makeTensorInfo(outShape, image2.dtype, outVals);
+  }
+  const dataId = backend2.write(outVals, outShape, image2.dtype);
+  return { dataId, shape: image2.shape, dtype: image2.dtype };
+}
+var transformConfig = {
+  kernelName: Transform,
+  backendName: "cpu",
+  kernelFunc: transform2
+};
+function mapCoord(outCoord, len, mode) {
+  switch (mode) {
+    case "reflect":
+      return mapCoordReflect(outCoord, len);
+    case "wrap":
+      return mapCoordWrap(outCoord, len);
+    case "nearest":
+      return mapCoordNearest(outCoord, len);
+    case "constant":
+    default:
+      return mapCoordConstant(outCoord, len);
+  }
+}
+function mapCoordReflect(outCoord, len) {
+  let inCoord = outCoord;
+  if (inCoord < 0) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz2 = 2 * len;
+      if (inCoord < sz2) {
+        inCoord = sz2 * Math.trunc(-inCoord / sz2) + inCoord;
+      }
+      inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1;
+    }
+  } else if (inCoord > len - 1) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz2 = 2 * len;
+      inCoord -= sz2 * Math.trunc(inCoord / sz2);
+      if (inCoord >= len) {
+        inCoord = sz2 - inCoord - 1;
+      }
+    }
+  }
+  return util_exports.clamp(0, inCoord, len - 1);
+}
+function mapCoordWrap(outCoord, len) {
+  let inCoord = outCoord;
+  if (inCoord < 0) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz = len - 1;
+      inCoord += len * (Math.trunc(-inCoord / sz) + 1);
+    }
+  } else if (inCoord > len - 1) {
+    if (len <= 1) {
+      inCoord = 0;
+    } else {
+      const sz = len - 1;
+      inCoord -= len * Math.trunc(inCoord / sz);
+    }
+  }
+  return util_exports.clamp(0, inCoord, len - 1);
+}
+function mapCoordConstant(outCoord, len) {
+  return outCoord;
+}
+function mapCoordNearest(outCoord, len) {
+  return util_exports.clamp(0, outCoord, len - 1);
+}
+function readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const ind = batch * batchStride + y * rowStride + x * colStride + channel;
+  if (0 <= y && y < imageHeight && 0 <= x && x < imageWidth) {
+    return imageVals[ind];
+  } else {
+    return fillValue;
+  }
+}
+function nearestInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const $y = Math.round(y);
+  const $x = Math.round(x);
+  return readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, $y, $x, channel, fillValue);
+}
+function bilinearInterpolation(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, y, x, channel, fillValue) {
+  const yFloor = Math.floor(y);
+  const xFloor = Math.floor(x);
+  const yCeil = yFloor + 1;
+  const xCeil = xFloor + 1;
+  const valueYFloor = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yFloor, xCeil, channel, fillValue);
+  const valueYCeil = (xCeil - x) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xFloor, channel, fillValue) + (x - xFloor) * readWithFillValue(imageVals, imageHeight, imageWidth, batchStride, rowStride, colStride, batch, yCeil, xCeil, channel, fillValue);
+  return (yCeil - y) * valueYFloor + (y - yFloor) * valueYCeil;
+}
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unique.js
+function unique3(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  assertNotComplex(x, "unique");
+  const values = backend2.data.get(x.dataId).values;
+  const { outputValues, outputShape, indices } = uniqueImpl(values, axis, x.shape, x.dtype);
+  return [
+    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),
+    backend2.makeTensorInfo([indices.length], "int32", indices)
+  ];
+}
+var uniqueConfig = {
+  kernelName: Unique,
+  backendName: "cpu",
+  kernelFunc: unique3
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/Unpack.js
+function unpack(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const valueRank = value.shape.length;
+  const num = value.shape[axis];
+  const outShape = new Array(valueRank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < valueRank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = value.shape[i];
+    }
+  }
+  const begin = new Array(valueRank).fill(0);
+  const size = value.shape.slice();
+  size[axis] = 1;
+  const res = new Array(num);
+  for (let i = 0; i < res.length; i++) {
+    begin[axis] = i;
+    const tempRes = slice2({ inputs: { x: value }, backend: backend2, attrs: { begin, size } });
+    res[i] = reshape3({ inputs: { x: tempRes }, backend: backend2, attrs: { shape: outShape } });
+    backend2.disposeIntermediateTensorInfo(tempRes);
+  }
+  return res;
+}
+var unpackConfig = {
+  kernelName: Unpack,
+  backendName: "cpu",
+  kernelFunc: unpack
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/kernels/UnsortedSegmentSum.js
+function unsortedSegmentSum2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, segmentIds } = inputs;
+  const { numSegments } = attrs;
+  assertNotComplex(x, "unsortedSegmentSum");
+  const xRank = x.shape.length;
+  const segmentIdsRank = segmentIds.shape.length;
+  const res = [];
+  const intermediates = [];
+  const numIters = xRank - segmentIdsRank;
+  let $segmentIds = segmentIds;
+  for (let i = 0; i < numIters; ++i) {
+    const expanded = expandDims3({ inputs: { input: $segmentIds }, backend: backend2, attrs: { dim: i + 1 } });
+    $segmentIds = expanded;
+    intermediates.push(expanded);
+  }
+  for (let i = 0; i < numSegments; ++i) {
+    const scalarValue = util_exports.createScalarValue(i, "int32");
+    const segmentId = backend2.makeTensorInfo([], "int32", scalarValue);
+    const mask = equal2({ inputs: { a: segmentId, b: $segmentIds }, backend: backend2 });
+    const maskCasted = cast3({ inputs: { x: mask }, backend: backend2, attrs: { dtype: "float32" } });
+    const mul2 = multiply2({ inputs: { a: maskCasted, b: x }, backend: backend2 });
+    const sumTensorInfo = sum3({ inputs: { x: mul2 }, backend: backend2, attrs: { axis: 0, keepDims: false } });
+    res.push(sumTensorInfo);
+    intermediates.push(segmentId);
+    intermediates.push(mask);
+    intermediates.push(maskCasted);
+    intermediates.push(mul2);
+    intermediates.push(sumTensorInfo);
+  }
+  const result = pack({ inputs: res, backend: backend2, attrs: { axis: 0 } });
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var unsortedSegmentSumConfig = {
+  kernelName: UnsortedSegmentSum,
+  backendName: "cpu",
+  kernelFunc: unsortedSegmentSum2
+};
+
+// node_modules/@tensorflow/tfjs-backend-cpu/dist/register_all_kernels.js
+var kernelConfigs = [
+  _fusedMatMulConfig,
+  absConfig,
+  acosConfig,
+  acoshConfig,
+  addConfig,
+  addNConfig,
+  allConfig,
+  anyConfig,
+  argMaxConfig,
+  argMinConfig,
+  asinConfig,
+  asinhConfig,
+  atanConfig,
+  atan2Config,
+  atanhConfig,
+  avgPoolConfig,
+  avgPool3DConfig,
+  avgPool3DGradConfig2,
+  avgPoolGradConfig2,
+  batchMatMulConfig,
+  batchNormConfig,
+  batchToSpaceNDConfig,
+  bincountConfig,
+  broadcastArgsConfig,
+  castConfig,
+  ceilConfig,
+  clipByValueConfig,
+  complexConfig,
+  complexAbsConfig,
+  concatConfig,
+  conv2DConfig,
+  conv2DBackpropFilterConfig,
+  conv2DBackpropInputConfig,
+  conv3DConfig,
+  conv3DBackpropFilterV2Config,
+  conv3DBackpropInputV2Config,
+  cosConfig,
+  coshConfig,
+  cropAndResizeConfig,
+  cumprodConfig,
+  cumsumConfig,
+  denseBincountConfig,
+  depthToSpaceConfig,
+  depthwiseConv2dNativeConfig,
+  depthwiseConv2dNativeBackpropFilterConfig,
+  depthwiseConv2dNativeBackpropInputConfig,
+  diagConfig,
+  dilation2DConfig,
+  dilation2DBackpropFilterConfig,
+  dilation2DBackpropInputConfig,
+  einsumConfig,
+  eluConfig,
+  eluGradConfig2,
+  equalConfig,
+  erfConfig,
+  expConfig,
+  expandDimsConfig,
+  expm1Config,
+  fftConfig,
+  fillConfig,
+  flipLeftRightConfig,
+  floorConfig,
+  floorDivConfig,
+  fusedConv2DConfig,
+  fusedDepthwiseConv2DConfig,
+  gatherNdConfig,
+  gatherV2Config,
+  greaterConfig,
+  greaterEqualConfig,
+  identityConfig,
+  ifftConfig,
+  imagConfig,
+  isFiniteConfig,
+  isInfConfig,
+  isNaNConfig,
+  leakyReluConfig,
+  lessConfig,
+  lessEqualConfig,
+  linSpaceConfig,
+  logConfig,
+  log1pConfig,
+  logicalAndConfig,
+  logicalNotConfig,
+  logicalOrConfig,
+  LRNConfig,
+  LRNGradConfig,
+  maxConfig,
+  maximumConfig,
+  maxPoolConfig,
+  maxPool3DConfig,
+  maxPool3DGradConfig2,
+  maxPoolGradConfig2,
+  maxPoolWithArgmaxConfig,
+  meanConfig,
+  minConfig,
+  minimumConfig,
+  mirrorPadConfig,
+  modConfig,
+  multinomialConfig,
+  multiplyConfig,
+  negConfig,
+  nonMaxSuppressionV3Config,
+  nonMaxSuppressionV4Config,
+  nonMaxSuppressionV5Config,
+  notEqualConfig,
+  oneHotConfig,
+  onesLikeConfig,
+  packConfig,
+  padV2Config,
+  powConfig,
+  preluConfig,
+  prodConfig,
+  raggedGatherConfig,
+  raggedRangeConfig,
+  raggedTensorToTensorConfig,
+  rangeConfig,
+  realConfig,
+  realDivConfig,
+  reciprocalConfig,
+  reluConfig,
+  relu6Config,
+  reshapeConfig,
+  resizeBilinearConfig,
+  resizeBilinearGradConfig2,
+  resizeNearestNeighborConfig,
+  resizeNearestNeighborGradConfig2,
+  reverseConfig,
+  rotateWithOffsetConfig,
+  roundConfig,
+  rsqrtConfig,
+  scatterNdConfig,
+  searchSortedConfig,
+  selectConfig,
+  seluConfig,
+  sigmoidConfig,
+  signConfig,
+  sinConfig,
+  sinhConfig,
+  sliceConfig,
+  softmaxConfig,
+  softplusConfig,
+  spaceToBatchNDConfig,
+  sparseFillEmptyRowsConfig,
+  sparseReshapeConfig,
+  sparseSegmentMeanConfig,
+  sparseSegmentSumConfig,
+  sparseToDenseConfig,
+  splitVConfig,
+  sqrtConfig,
+  squareConfig,
+  squaredDifferenceConfig,
+  stepConfig,
+  stridedSliceConfig,
+  stringNGramsConfig,
+  stringSplitConfig,
+  stringToHashBucketFastConfig,
+  subConfig,
+  sumConfig,
+  tanConfig,
+  tanhConfig,
+  tileConfig,
+  topKConfig,
+  transformConfig,
+  transposeConfig,
+  uniqueConfig,
+  unpackConfig,
+  unsortedSegmentSumConfig,
+  zerosLikeConfig
+];
+for (const kernelConfig of kernelConfigs) {
+  registerKernel(kernelConfig);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl_util.js
+var webgl_util_exports = {};
+__export(webgl_util_exports, {
+  assertNotComplex: () => assertNotComplex2,
+  bindCanvasToFramebuffer: () => bindCanvasToFramebuffer,
+  bindColorTextureToFramebuffer: () => bindColorTextureToFramebuffer,
+  bindTextureToProgramUniformSampler: () => bindTextureToProgramUniformSampler,
+  bindTextureUnit: () => bindTextureUnit,
+  bindVertexBufferToProgramAttribute: () => bindVertexBufferToProgramAttribute,
+  callAndCheck: () => callAndCheck,
+  canBeRepresented: () => canBeRepresented,
+  createFragmentShader: () => createFragmentShader,
+  createFramebuffer: () => createFramebuffer,
+  createProgram: () => createProgram,
+  createStaticIndexBuffer: () => createStaticIndexBuffer,
+  createStaticVertexBuffer: () => createStaticVertexBuffer,
+  createTexture: () => createTexture,
+  createVertexShader: () => createVertexShader,
+  getBatchDim: () => getBatchDim,
+  getExtensionOrThrow: () => getExtensionOrThrow,
+  getFramebufferErrorMessage: () => getFramebufferErrorMessage,
+  getMaxTexturesInShader: () => getMaxTexturesInShader,
+  getNumChannels: () => getNumChannels,
+  getProgramUniformLocation: () => getProgramUniformLocation,
+  getProgramUniformLocationOrThrow: () => getProgramUniformLocationOrThrow,
+  getRowsCols: () => getRowsCols,
+  getShapeAs3D: () => getShapeAs3D,
+  getTextureShapeFromLogicalShape: () => getTextureShapeFromLogicalShape,
+  getWebGLDisjointQueryTimerVersion: () => getWebGLDisjointQueryTimerVersion,
+  getWebGLErrorMessage: () => getWebGLErrorMessage,
+  getWebGLMaxTextureSize: () => getWebGLMaxTextureSize,
+  hasExtension: () => hasExtension,
+  isCapableOfRenderingToFloatTexture: () => isCapableOfRenderingToFloatTexture,
+  isDownloadFloatTextureEnabled: () => isDownloadFloatTextureEnabled,
+  isReshapeFree: () => isReshapeFree,
+  isWebGLFenceEnabled: () => isWebGLFenceEnabled,
+  isWebGLVersionEnabled: () => isWebGLVersionEnabled,
+  linkProgram: () => linkProgram,
+  logShaderSourceAndInfoLog: () => logShaderSourceAndInfoLog,
+  resetMaxTextureSize: () => resetMaxTextureSize,
+  resetMaxTexturesInShader: () => resetMaxTexturesInShader,
+  unbindColorTextureFromFramebuffer: () => unbindColorTextureFromFramebuffer,
+  unbindTextureUnit: () => unbindTextureUnit,
+  validateFramebuffer: () => validateFramebuffer,
+  validateProgram: () => validateProgram,
+  validateTextureSize: () => validateTextureSize
+});
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/canvas_util.js
+var contexts = {};
+var WEBGL_ATTRIBUTES = {
+  alpha: false,
+  antialias: false,
+  premultipliedAlpha: false,
+  preserveDrawingBuffer: false,
+  depth: false,
+  stencil: false,
+  failIfMajorPerformanceCaveat: true
+};
+function setWebGLContext(webGLVersion, gl) {
+  contexts[webGLVersion] = gl;
+}
+function getWebGLContext(webGLVersion, customCanvas) {
+  if (!(webGLVersion in contexts) || customCanvas != null) {
+    const newCtx = getWebGLRenderingContext(webGLVersion, customCanvas);
+    if (newCtx !== null) {
+      contexts[webGLVersion] = newCtx;
+    } else {
+      console.log("Could not get context for WebGL version", webGLVersion);
+      return null;
+    }
+  }
+  const gl = contexts[webGLVersion];
+  if (gl == null || gl.isContextLost()) {
+    delete contexts[webGLVersion];
+    return getWebGLContext(webGLVersion);
+  }
+  gl.disable(gl.DEPTH_TEST);
+  gl.disable(gl.STENCIL_TEST);
+  gl.disable(gl.BLEND);
+  gl.disable(gl.DITHER);
+  gl.disable(gl.POLYGON_OFFSET_FILL);
+  gl.disable(gl.SAMPLE_COVERAGE);
+  gl.enable(gl.SCISSOR_TEST);
+  gl.enable(gl.CULL_FACE);
+  gl.cullFace(gl.BACK);
+  return contexts[webGLVersion];
+}
+function createCanvas(webGLVersion) {
+  if (typeof OffscreenCanvas !== "undefined" && webGLVersion === 2) {
+    return new OffscreenCanvas(300, 150);
+  } else if (typeof document !== "undefined") {
+    return document.createElement("canvas");
+  } else {
+    throw new Error("Cannot create a canvas in this context");
+  }
+}
+function getWebGLRenderingContext(webGLVersion, customCanvas) {
+  if (webGLVersion !== 1 && webGLVersion !== 2) {
+    throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");
+  }
+  const canvas = customCanvas == null ? createCanvas(webGLVersion) : customCanvas;
+  canvas.addEventListener("webglcontextlost", (ev) => {
+    ev.preventDefault();
+    delete contexts[webGLVersion];
+  }, false);
+  if (env().getBool("SOFTWARE_WEBGL_ENABLED")) {
+    WEBGL_ATTRIBUTES.failIfMajorPerformanceCaveat = false;
+  }
+  if (webGLVersion === 1) {
+    return canvas.getContext("webgl", WEBGL_ATTRIBUTES) || canvas.getContext("experimental-webgl", WEBGL_ATTRIBUTES);
+  }
+  return canvas.getContext("webgl2", WEBGL_ATTRIBUTES);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/tex_util.js
+var PackingScheme;
+(function(PackingScheme2) {
+  PackingScheme2[PackingScheme2["DENSE"] = 0] = "DENSE";
+  PackingScheme2[PackingScheme2["SHARED_BATCH"] = 1] = "SHARED_BATCH";
+})(PackingScheme || (PackingScheme = {}));
+var TextureUsage;
+(function(TextureUsage2) {
+  TextureUsage2[TextureUsage2["RENDER"] = 0] = "RENDER";
+  TextureUsage2[TextureUsage2["UPLOAD"] = 1] = "UPLOAD";
+  TextureUsage2[TextureUsage2["PIXELS"] = 2] = "PIXELS";
+  TextureUsage2[TextureUsage2["DOWNLOAD"] = 3] = "DOWNLOAD";
+})(TextureUsage || (TextureUsage = {}));
+var PhysicalTextureType;
+(function(PhysicalTextureType2) {
+  PhysicalTextureType2[PhysicalTextureType2["UNPACKED_FLOAT16"] = 0] = "UNPACKED_FLOAT16";
+  PhysicalTextureType2[PhysicalTextureType2["UNPACKED_FLOAT32"] = 1] = "UNPACKED_FLOAT32";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_4X1_UNSIGNED_BYTE"] = 2] = "PACKED_4X1_UNSIGNED_BYTE";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_2X2_FLOAT32"] = 3] = "PACKED_2X2_FLOAT32";
+  PhysicalTextureType2[PhysicalTextureType2["PACKED_2X2_FLOAT16"] = 4] = "PACKED_2X2_FLOAT16";
+})(PhysicalTextureType || (PhysicalTextureType = {}));
+function getUnpackedMatrixTextureShapeWidthHeight(rows, columns) {
+  return [columns, rows];
+}
+function getUnpackedArraySizeFromMatrixSize(matrixSize, channelsPerTexture) {
+  return matrixSize * channelsPerTexture;
+}
+function getDenseTexShape(shape) {
+  const size = util_exports.sizeFromShape(shape);
+  const texelsNeeded = Math.ceil(size / 4);
+  return util_exports.sizeToSquarishShape(texelsNeeded);
+}
+function getPackedMatrixTextureShapeWidthHeight(rows, columns) {
+  return [
+    Math.max(1, Math.ceil(columns / 2)),
+    Math.max(1, Math.ceil(rows / 2))
+  ];
+}
+function getPackedRGBAArraySizeFromMatrixShape(rows, columns) {
+  const [w, h] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return w * h * 4;
+}
+function getTextureConfig(gl, textureHalfFloatExtension) {
+  const glany = gl;
+  let internalFormatFloat;
+  let internalFormatHalfFloat;
+  let internalFormatPackedHalfFloat;
+  let internalFormatPackedFloat;
+  let textureFormatFloat;
+  let downloadTextureFormat;
+  let downloadUnpackNumChannels;
+  let defaultNumChannels;
+  let textureTypeHalfFloat;
+  let textureTypeFloat;
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    internalFormatFloat = glany.R32F;
+    internalFormatHalfFloat = glany.R16F;
+    internalFormatPackedHalfFloat = glany.RGBA16F;
+    internalFormatPackedFloat = glany.RGBA32F;
+    textureFormatFloat = glany.RED;
+    downloadUnpackNumChannels = 4;
+    defaultNumChannels = 1;
+    textureTypeHalfFloat = glany.HALF_FLOAT;
+    textureTypeFloat = glany.FLOAT;
+    downloadTextureFormat = glany.RGBA8;
+  } else {
+    internalFormatFloat = gl.RGBA;
+    internalFormatHalfFloat = gl.RGBA;
+    internalFormatPackedHalfFloat = gl.RGBA;
+    internalFormatPackedFloat = glany.RGBA;
+    textureFormatFloat = gl.RGBA;
+    downloadUnpackNumChannels = 4;
+    defaultNumChannels = 4;
+    textureTypeHalfFloat = textureHalfFloatExtension != null ? textureHalfFloatExtension.HALF_FLOAT_OES : null;
+    textureTypeFloat = gl.FLOAT;
+    downloadTextureFormat = gl.RGBA;
+  }
+  return {
+    internalFormatFloat,
+    internalFormatHalfFloat,
+    internalFormatPackedHalfFloat,
+    internalFormatPackedFloat,
+    textureFormatFloat,
+    downloadTextureFormat,
+    downloadUnpackNumChannels,
+    defaultNumChannels,
+    textureTypeHalfFloat,
+    textureTypeFloat
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl_util.js
+function callAndCheck(gl, func2) {
+  const returnValue = func2();
+  if (env().getBool("DEBUG")) {
+    checkWebGLError(gl);
+  }
+  return returnValue;
+}
+function checkWebGLError(gl) {
+  const error = gl.getError();
+  if (error !== gl.NO_ERROR) {
+    throw new Error("WebGL Error: " + getWebGLErrorMessage(gl, error));
+  }
+}
+var MIN_FLOAT16 = 596e-10;
+var MAX_FLOAT16 = 65504;
+function canBeRepresented(num) {
+  if (env().getBool("WEBGL_RENDER_FLOAT32_ENABLED") || num === 0 || MIN_FLOAT16 < Math.abs(num) && Math.abs(num) < MAX_FLOAT16) {
+    return true;
+  }
+  return false;
+}
+function getWebGLErrorMessage(gl, status) {
+  switch (status) {
+    case gl.NO_ERROR:
+      return "NO_ERROR";
+    case gl.INVALID_ENUM:
+      return "INVALID_ENUM";
+    case gl.INVALID_VALUE:
+      return "INVALID_VALUE";
+    case gl.INVALID_OPERATION:
+      return "INVALID_OPERATION";
+    case gl.INVALID_FRAMEBUFFER_OPERATION:
+      return "INVALID_FRAMEBUFFER_OPERATION";
+    case gl.OUT_OF_MEMORY:
+      return "OUT_OF_MEMORY";
+    case gl.CONTEXT_LOST_WEBGL:
+      return "CONTEXT_LOST_WEBGL";
+    default:
+      return `Unknown error code ${status}`;
+  }
+}
+function getExtensionOrThrow(gl, extensionName) {
+  return throwIfNull(gl, () => gl.getExtension(extensionName), 'Extension "' + extensionName + '" not supported on this browser.');
+}
+function createVertexShader(gl, vertexShaderSource) {
+  const vertexShader = throwIfNull(gl, () => gl.createShader(gl.VERTEX_SHADER), "Unable to create vertex WebGLShader.");
+  callAndCheck(gl, () => gl.shaderSource(vertexShader, vertexShaderSource));
+  callAndCheck(gl, () => gl.compileShader(vertexShader));
+  if (gl.getShaderParameter(vertexShader, gl.COMPILE_STATUS) === false) {
+    console.log(gl.getShaderInfoLog(vertexShader));
+    throw new Error("Failed to compile vertex shader.");
+  }
+  return vertexShader;
+}
+function createFragmentShader(gl, fragmentShaderSource) {
+  const fragmentShader = throwIfNull(gl, () => gl.createShader(gl.FRAGMENT_SHADER), "Unable to create fragment WebGLShader.");
+  callAndCheck(gl, () => gl.shaderSource(fragmentShader, fragmentShaderSource));
+  callAndCheck(gl, () => gl.compileShader(fragmentShader));
+  if (env().get("ENGINE_COMPILE_ONLY")) {
+    return fragmentShader;
+  }
+  if (gl.getShaderParameter(fragmentShader, gl.COMPILE_STATUS) === false) {
+    logShaderSourceAndInfoLog(fragmentShaderSource, gl.getShaderInfoLog(fragmentShader));
+    throw new Error("Failed to compile fragment shader.");
+  }
+  return fragmentShader;
+}
+var lineNumberRegex = /ERROR: [0-9]+:([0-9]+):/g;
+function logShaderSourceAndInfoLog(shaderSource, shaderInfoLog) {
+  const lineNumberRegexResult = lineNumberRegex.exec(shaderInfoLog);
+  if (lineNumberRegexResult == null) {
+    console.log(`Couldn't parse line number in error: ${shaderInfoLog}`);
+    console.log(shaderSource);
+    return;
+  }
+  const lineNumber = +lineNumberRegexResult[1];
+  const shaderLines = shaderSource.split("\n");
+  const pad3 = shaderLines.length.toString().length + 2;
+  const linesWithLineNumbers = shaderLines.map((line, lineNumber2) => util_exports.rightPad((lineNumber2 + 1).toString(), pad3) + line);
+  let maxLineLength = 0;
+  for (let i = 0; i < linesWithLineNumbers.length; i++) {
+    maxLineLength = Math.max(linesWithLineNumbers[i].length, maxLineLength);
+  }
+  const beforeErrorLines = linesWithLineNumbers.slice(0, lineNumber - 1);
+  const errorLine = linesWithLineNumbers.slice(lineNumber - 1, lineNumber);
+  const afterErrorLines = linesWithLineNumbers.slice(lineNumber);
+  console.log(beforeErrorLines.join("\n"));
+  console.log(shaderInfoLog.split("\n")[0]);
+  console.log(`%c ${util_exports.rightPad(errorLine[0], maxLineLength)}`, "border:1px solid red; background-color:#e3d2d2; color:#a61717");
+  console.log(afterErrorLines.join("\n"));
+}
+function createProgram(gl) {
+  return throwIfNull(gl, () => gl.createProgram(), "Unable to create WebGLProgram.");
+}
+function linkProgram(gl, program) {
+  callAndCheck(gl, () => gl.linkProgram(program));
+  if (env().get("ENGINE_COMPILE_ONLY")) {
+    return;
+  }
+  if (gl.getProgramParameter(program, gl.LINK_STATUS) === false) {
+    console.log(gl.getProgramInfoLog(program));
+    throw new Error("Failed to link vertex and fragment shaders.");
+  }
+}
+function validateProgram(gl, program) {
+  callAndCheck(gl, () => gl.validateProgram(program));
+  if (gl.getProgramParameter(program, gl.VALIDATE_STATUS) === false) {
+    console.log(gl.getProgramInfoLog(program));
+    throw new Error("Shader program validation failed.");
+  }
+}
+function createStaticVertexBuffer(gl, data) {
+  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), "Unable to create WebGLBuffer");
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.bufferData(gl.ARRAY_BUFFER, data, gl.STATIC_DRAW));
+  return buffer2;
+}
+function createStaticIndexBuffer(gl, data) {
+  const buffer2 = throwIfNull(gl, () => gl.createBuffer(), "Unable to create WebGLBuffer");
+  callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, data, gl.STATIC_DRAW));
+  return buffer2;
+}
+function getNumChannels() {
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    return 1;
+  }
+  return 4;
+}
+function createTexture(gl) {
+  return throwIfNull(gl, () => gl.createTexture(), "Unable to create WebGLTexture.");
+}
+function validateTextureSize(width, height) {
+  const maxTextureSize = env().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+  if (width <= 0 || height <= 0) {
+    const requested = `[${width}x${height}]`;
+    throw new Error("Requested texture size " + requested + " is invalid.");
+  }
+  if (width > maxTextureSize || height > maxTextureSize) {
+    const requested = `[${width}x${height}]`;
+    const max6 = `[${maxTextureSize}x${maxTextureSize}]`;
+    throw new Error("Requested texture size " + requested + " greater than WebGL maximum on this browser / GPU " + max6 + ".");
+  }
+}
+function createFramebuffer(gl) {
+  return throwIfNull(gl, () => gl.createFramebuffer(), "Unable to create WebGLFramebuffer.");
+}
+function bindVertexBufferToProgramAttribute(gl, program, attribute, buffer2, arrayEntriesPerItem, itemStrideInBytes, itemOffsetInBytes) {
+  const loc = gl.getAttribLocation(program, attribute);
+  if (loc === -1) {
+    return false;
+  }
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, buffer2));
+  callAndCheck(gl, () => gl.vertexAttribPointer(loc, arrayEntriesPerItem, gl.FLOAT, false, itemStrideInBytes, itemOffsetInBytes));
+  callAndCheck(gl, () => gl.enableVertexAttribArray(loc));
+  return true;
+}
+function bindTextureUnit(gl, texture, textureUnit) {
+  validateTextureUnit(gl, textureUnit);
+  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+}
+function unbindTextureUnit(gl, textureUnit) {
+  validateTextureUnit(gl, textureUnit);
+  callAndCheck(gl, () => gl.activeTexture(gl.TEXTURE0 + textureUnit));
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function getProgramUniformLocationOrThrow(gl, program, uniformName) {
+  return throwIfNull(gl, () => gl.getUniformLocation(program, uniformName), 'uniform "' + uniformName + '" not present in program.');
+}
+function getProgramUniformLocation(gl, program, uniformName) {
+  return gl.getUniformLocation(program, uniformName);
+}
+function bindTextureToProgramUniformSampler(gl, texture, uniformSamplerLocation, textureUnit) {
+  callAndCheck(gl, () => bindTextureUnit(gl, texture, textureUnit));
+  callAndCheck(gl, () => gl.uniform1i(uniformSamplerLocation, textureUnit));
+}
+function bindCanvasToFramebuffer(gl) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));
+  callAndCheck(gl, () => gl.viewport(0, 0, gl.canvas.width, gl.canvas.height));
+  callAndCheck(gl, () => gl.scissor(0, 0, gl.canvas.width, gl.canvas.height));
+}
+function bindColorTextureToFramebuffer(gl, texture, framebuffer) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));
+  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0));
+}
+function unbindColorTextureFromFramebuffer(gl, framebuffer) {
+  callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, framebuffer));
+  callAndCheck(gl, () => gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, null, 0));
+}
+function validateFramebuffer(gl) {
+  const status = gl.checkFramebufferStatus(gl.FRAMEBUFFER);
+  if (status !== gl.FRAMEBUFFER_COMPLETE) {
+    throw new Error("Error binding framebuffer: " + getFramebufferErrorMessage(gl, status));
+  }
+}
+function getFramebufferErrorMessage(gl, status) {
+  switch (status) {
+    case gl.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:
+      return "FRAMEBUFFER_INCOMPLETE_ATTACHMENT";
+    case gl.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:
+      return "FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";
+    case gl.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:
+      return "FRAMEBUFFER_INCOMPLETE_DIMENSIONS";
+    case gl.FRAMEBUFFER_UNSUPPORTED:
+      return "FRAMEBUFFER_UNSUPPORTED";
+    default:
+      return `unknown error ${status}`;
+  }
+}
+function throwIfNull(gl, returnTOrNull, failureMessage) {
+  const tOrNull = callAndCheck(gl, () => returnTOrNull());
+  if (tOrNull == null) {
+    throw new Error(failureMessage);
+  }
+  return tOrNull;
+}
+function validateTextureUnit(gl, textureUnit) {
+  const maxTextureUnit = gl.MAX_COMBINED_TEXTURE_IMAGE_UNITS - 1;
+  const glTextureUnit = textureUnit + gl.TEXTURE0;
+  if (glTextureUnit < gl.TEXTURE0 || glTextureUnit > maxTextureUnit) {
+    const textureUnitRange = `[gl.TEXTURE0, gl.TEXTURE${maxTextureUnit}]`;
+    throw new Error(`textureUnit must be in ${textureUnitRange}.`);
+  }
+}
+function getBatchDim(shape, dimsToSkip = 2) {
+  return util_exports.sizeFromShape(shape.slice(0, shape.length - dimsToSkip));
+}
+function getRowsCols(shape) {
+  if (shape.length === 0) {
+    throw Error("Cannot get rows and columns of an empty shape array.");
+  }
+  return [
+    shape.length > 1 ? shape[shape.length - 2] : 1,
+    shape[shape.length - 1]
+  ];
+}
+function getShapeAs3D(shape) {
+  let shapeAs3D = [1, 1, 1];
+  const isScalar = shape.length === 0 || shape.length === 1 && shape[0] === 1;
+  if (!isScalar) {
+    shapeAs3D = [getBatchDim(shape), ...getRowsCols(shape)];
+  }
+  return shapeAs3D;
+}
+function getTextureShapeFromLogicalShape(logShape, isPacked = false) {
+  let maxTexSize = env().getNumber("WEBGL_MAX_TEXTURE_SIZE");
+  let maxSizeForNarrowTex = env().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");
+  if (maxSizeForNarrowTex === Infinity && env().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")) {
+    maxSizeForNarrowTex = maxTexSize / 2;
+  }
+  if (isPacked) {
+    maxTexSize = maxTexSize * 2;
+    maxSizeForNarrowTex = maxSizeForNarrowTex * 2;
+    logShape = logShape.map((d, i) => i >= logShape.length - 2 ? util_exports.nearestLargerEven(logShape[i]) : logShape[i]);
+    if (logShape.length === 1) {
+      logShape = [2, logShape[0]];
+    }
+  }
+  if (logShape.length !== 2) {
+    const squeezeResult = util_exports.squeezeShape(logShape);
+    logShape = squeezeResult.newShape;
+  }
+  let size = util_exports.sizeFromShape(logShape);
+  let textureShape = null;
+  if (logShape.length <= 1 && size <= maxTexSize) {
+    textureShape = [1, size];
+  } else if (logShape.length === 2 && logShape[0] <= maxTexSize && logShape[1] <= maxTexSize) {
+    textureShape = logShape;
+  } else if (logShape.length === 3 && logShape[0] * logShape[1] <= maxTexSize && logShape[2] <= maxTexSize) {
+    textureShape = [logShape[0] * logShape[1], logShape[2]];
+  } else if (logShape.length === 3 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] <= maxTexSize) {
+    textureShape = [logShape[0], logShape[1] * logShape[2]];
+  } else if (logShape.length === 4 && logShape[0] * logShape[1] * logShape[2] <= maxTexSize && logShape[3] <= maxTexSize) {
+    textureShape = [logShape[0] * logShape[1] * logShape[2], logShape[3]];
+  } else if (logShape.length === 4 && logShape[0] <= maxTexSize && logShape[1] * logShape[2] * logShape[3] <= maxTexSize) {
+    textureShape = [logShape[0], logShape[1] * logShape[2] * logShape[3]];
+  }
+  const isLongNarrowTex = textureShape != null && Math.max(...textureShape) > maxSizeForNarrowTex && Math.min(...textureShape) <= (isPacked ? 2 : 1) && Math.min(...textureShape) > 0;
+  if (textureShape == null || isLongNarrowTex) {
+    if (isPacked) {
+      const batchDim = getBatchDim(logShape);
+      let rows = 2, cols = 2;
+      if (logShape.length) {
+        [rows, cols] = getRowsCols(logShape);
+      }
+      size = batchDim * (rows / 2) * (cols / 2);
+      textureShape = util_exports.sizeToSquarishShape(size).map((d) => d * 2);
+    } else {
+      textureShape = util_exports.sizeToSquarishShape(size);
+    }
+  }
+  return textureShape;
+}
+function isEven(n) {
+  return n % 2 === 0;
+}
+function isReshapeFree(shape1, shape2) {
+  shape1 = shape1.slice(-2);
+  shape2 = shape2.slice(-2);
+  if (util_exports.arraysEqual(shape1, shape2)) {
+    return true;
+  }
+  if (!shape1.length || !shape2.length) {
+    return true;
+  }
+  if (shape1[0] === 0 || shape1[1] === 0 || shape2[0] === 0 || shape2[1] === 0) {
+    return true;
+  }
+  if (shape1.length !== shape2.length) {
+    const shape1Cols = shape1.slice(-1)[0];
+    const shape2Cols = shape2.slice(-1)[0];
+    if (shape1Cols === shape2Cols) {
+      return true;
+    }
+    if (isEven(shape1Cols) && isEven(shape2Cols) && (shape1[0] === 1 || shape2[0] === 1)) {
+      return true;
+    }
+  }
+  return shape1[1] === shape2[1] && isEven(shape1[0]) && isEven(shape2[0]);
+}
+var MAX_TEXTURE_SIZE;
+var MAX_TEXTURES_IN_SHADER;
+function getWebGLMaxTextureSize(webGLVersion) {
+  if (MAX_TEXTURE_SIZE == null) {
+    const gl = getWebGLContext(webGLVersion);
+    MAX_TEXTURE_SIZE = gl.getParameter(gl.MAX_TEXTURE_SIZE);
+  }
+  return MAX_TEXTURE_SIZE;
+}
+function resetMaxTextureSize() {
+  MAX_TEXTURE_SIZE = null;
+}
+function resetMaxTexturesInShader() {
+  MAX_TEXTURES_IN_SHADER = null;
+}
+function getMaxTexturesInShader(webGLVersion) {
+  if (MAX_TEXTURES_IN_SHADER == null) {
+    const gl = getWebGLContext(webGLVersion);
+    MAX_TEXTURES_IN_SHADER = gl.getParameter(gl.MAX_TEXTURE_IMAGE_UNITS);
+  }
+  return Math.min(16, MAX_TEXTURES_IN_SHADER);
+}
+function getWebGLDisjointQueryTimerVersion(webGLVersion) {
+  if (webGLVersion === 0) {
+    return 0;
+  }
+  let queryTimerVersion;
+  const gl = getWebGLContext(webGLVersion);
+  if (hasExtension(gl, "EXT_disjoint_timer_query_webgl2") && webGLVersion === 2) {
+    queryTimerVersion = 2;
+  } else if (hasExtension(gl, "EXT_disjoint_timer_query")) {
+    queryTimerVersion = 1;
+  } else {
+    queryTimerVersion = 0;
+  }
+  return queryTimerVersion;
+}
+function hasExtension(gl, extensionName) {
+  const ext = gl.getExtension(extensionName);
+  return ext != null;
+}
+function isWebGLVersionEnabled(webGLVersion) {
+  try {
+    const gl = getWebGLContext(webGLVersion);
+    if (gl != null) {
+      return true;
+    }
+  } catch (e) {
+    console.log("Error when getting WebGL context: ", e);
+    return false;
+  }
+  return false;
+}
+function isCapableOfRenderingToFloatTexture(webGLVersion) {
+  if (webGLVersion === 0) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  if (webGLVersion === 1) {
+    if (!hasExtension(gl, "OES_texture_float")) {
+      return false;
+    }
+  } else {
+    if (!hasExtension(gl, "EXT_color_buffer_float")) {
+      return false;
+    }
+  }
+  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);
+  return isFrameBufferComplete;
+}
+function isDownloadFloatTextureEnabled(webGLVersion) {
+  if (webGLVersion === 0) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  if (webGLVersion === 1) {
+    if (!hasExtension(gl, "OES_texture_float")) {
+      return false;
+    }
+    if (!hasExtension(gl, "WEBGL_color_buffer_float")) {
+      return false;
+    }
+  } else {
+    if (hasExtension(gl, "EXT_color_buffer_float")) {
+      return createFloatTextureAndBindToFramebuffer(gl);
+    }
+    const COLOR_BUFFER_HALF_FLOAT = "EXT_color_buffer_half_float";
+    if (hasExtension(gl, COLOR_BUFFER_HALF_FLOAT)) {
+      const textureHalfFloatExtension = gl.getExtension(COLOR_BUFFER_HALF_FLOAT);
+      return createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension);
+    }
+    return false;
+  }
+  const isFrameBufferComplete = createFloatTextureAndBindToFramebuffer(gl);
+  return isFrameBufferComplete;
+}
+function createFloatTextureAndBindToFramebuffer(gl) {
+  const texConfig = getTextureConfig(gl);
+  const texture = gl.createTexture();
+  gl.bindTexture(gl.TEXTURE_2D, texture);
+  const width = 1;
+  const height = 1;
+  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeFloat, null);
+  const frameBuffer = gl.createFramebuffer();
+  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);
+  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);
+  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;
+  gl.bindTexture(gl.TEXTURE_2D, null);
+  gl.bindFramebuffer(gl.FRAMEBUFFER, null);
+  gl.deleteTexture(texture);
+  gl.deleteFramebuffer(frameBuffer);
+  return isFrameBufferComplete;
+}
+function createHalfFloatTextureAndBindToFramebuffer(gl, textureHalfFloatExtension) {
+  const texConfig = getTextureConfig(gl, textureHalfFloatExtension);
+  const texture = gl.createTexture();
+  gl.bindTexture(gl.TEXTURE_2D, texture);
+  const width = 1;
+  const height = 1;
+  gl.texImage2D(gl.TEXTURE_2D, 0, texConfig.internalFormatHalfFloat, width, height, 0, texConfig.textureFormatFloat, texConfig.textureTypeHalfFloat, null);
+  const frameBuffer = gl.createFramebuffer();
+  gl.bindFramebuffer(gl.FRAMEBUFFER, frameBuffer);
+  gl.framebufferTexture2D(gl.FRAMEBUFFER, gl.COLOR_ATTACHMENT0, gl.TEXTURE_2D, texture, 0);
+  const isFrameBufferComplete = gl.checkFramebufferStatus(gl.FRAMEBUFFER) === gl.FRAMEBUFFER_COMPLETE;
+  gl.bindTexture(gl.TEXTURE_2D, null);
+  gl.bindFramebuffer(gl.FRAMEBUFFER, null);
+  gl.deleteTexture(texture);
+  gl.deleteFramebuffer(frameBuffer);
+  return isFrameBufferComplete;
+}
+function isWebGLFenceEnabled(webGLVersion) {
+  if (webGLVersion !== 2) {
+    return false;
+  }
+  const gl = getWebGLContext(webGLVersion);
+  const isEnabled = gl.fenceSync != null;
+  return isEnabled;
+}
+function assertNotComplex2(tensor2, opName) {
+  if (!Array.isArray(tensor2)) {
+    tensor2 = [tensor2];
+  }
+  tensor2.forEach((t) => {
+    if (t != null) {
+      util_exports.assert(t.dtype !== "complex64", () => `${opName} does not support complex64 tensors in the WebGL backend.`);
+    }
+  });
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/flags_webgl.js
+var ENV5 = env();
+ENV5.registerFlag("HAS_WEBGL", () => ENV5.getNumber("WEBGL_VERSION") > 0);
+ENV5.registerFlag("WEBGL_VERSION", () => {
+  if (isWebGLVersionEnabled(2)) {
+    return 2;
+  } else if (isWebGLVersionEnabled(1)) {
+    return 1;
+  }
+  return 0;
+});
+ENV5.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS", () => false);
+ENV5.registerFlag("WEBGL_BUFFER_SUPPORTED", () => ENV5.get("WEBGL_VERSION") === 2);
+ENV5.registerFlag("WEBGL_CPU_FORWARD", () => true);
+ENV5.registerFlag("WEBGL_FORCE_F16_TEXTURES", () => false);
+ENV5.registerFlag("WEBGL_PACK", () => ENV5.getBool("HAS_WEBGL"));
+ENV5.registerFlag("WEBGL_PACK_NORMALIZATION", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_CLIP", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_DEPTHWISECONV", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_BINARY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_UNARY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_PACK_REDUCE", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_LAZILY_UNPACK", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_CONV_IM2COL", () => ENV5.getBool("WEBGL_PACK"));
+ENV5.registerFlag("WEBGL_MAX_TEXTURE_SIZE", () => getWebGLMaxTextureSize(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER", () => getMaxTexturesInShader(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION", () => {
+  const webGLVersion = ENV5.getNumber("WEBGL_VERSION");
+  if (webGLVersion === 0) {
+    return 0;
+  }
+  return getWebGLDisjointQueryTimerVersion(webGLVersion);
+});
+ENV5.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE", () => ENV5.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0 && !device_util_exports.isMobile());
+ENV5.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE", () => isCapableOfRenderingToFloatTexture(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED", () => {
+  return ENV5.getBool("WEBGL_FORCE_F16_TEXTURES") ? false : ENV5.getBool("WEBGL_RENDER_FLOAT32_CAPABLE");
+});
+ENV5.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED", () => isDownloadFloatTextureEnabled(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_FENCE_API_ENABLED", () => isWebGLFenceEnabled(ENV5.getNumber("WEBGL_VERSION")));
+ENV5.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM", () => {
+  const useUniforms = ENV5.getBool("WEBGL_RENDER_FLOAT32_ENABLED");
+  return useUniforms ? 4 : 0;
+});
+ENV5.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD", () => {
+  return -1;
+}, (threshold3) => {
+  if (threshold3 < 0 && threshold3 !== -1) {
+    throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${threshold3}.`);
+  }
+});
+ENV5.registerFlag("WEBGL_FLUSH_THRESHOLD", () => {
+  return device_util_exports.isMobile() ? 1 : -1;
+}, (threshold3) => {
+  if (threshold3 < 0 && threshold3 !== -1) {
+    throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${threshold3}.`);
+  }
+});
+ENV5.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD", () => 128);
+ENV5.registerFlag("WEBGL_USE_SHAPES_UNIFORMS", () => false);
+ENV5.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD", () => 1e5);
+ENV5.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD", () => 128);
+ENV5.registerFlag("WEBGL_EXP_CONV", () => false);
+ENV5.registerFlag("SOFTWARE_WEBGL_ENABLED", () => ENV5.getBool("IS_TEST"));
+ENV5.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE", () => Infinity);
+ENV5.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE", () => false);
+ENV5.registerFlag("WEBGL2_ISNAN_CUSTOM", () => false);
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/glsl_version.js
+function getGlslDifferences() {
+  let version10;
+  let attribute;
+  let varyingVs;
+  let varyingFs;
+  let texture2D;
+  let output;
+  let defineOutput;
+  let defineSpecialNaN;
+  let defineSpecialInf;
+  let defineRound;
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    version10 = "#version 300 es";
+    attribute = "in";
+    varyingVs = "out";
+    varyingFs = "in";
+    texture2D = "texture";
+    output = "outputColor";
+    defineOutput = "out vec4 outputColor;";
+    defineSpecialNaN = env().getBool("WEBGL2_ISNAN_CUSTOM") ? `
       bool isnan_custom(float val) {
         uint floatToUint = floatBitsToUint(val);
         return (floatToUint & 0x7fffffffu) > 0x7f800000u;
@@ -74,7 +49531,9 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       }
 
       #define isnan(value) isnan_custom(value)
-    `:"",u="",l=`
+    ` : "";
+    defineSpecialInf = ``;
+    defineRound = `
       #define round(value) newRound(value)
       int newRound(float value) {
         return int(floor(value + 0.5));
@@ -83,7 +49542,16 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 newRound(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `):(r="",t="attribute",e="varying",n="varying",o="texture2D",s="gl_FragColor",i="",a=`
+    `;
+  } else {
+    version10 = "";
+    attribute = "attribute";
+    varyingVs = "varying";
+    varyingFs = "varying";
+    texture2D = "texture2D";
+    output = "gl_FragColor";
+    defineOutput = "";
+    defineSpecialNaN = `
       #define isnan(value) isnan_custom(value)
       bool isnan_custom(float val) {
         return (val > 0. || val < 1. || val == 0.) ? false : true;
@@ -91,7 +49559,8 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       bvec4 isnan_custom(vec4 val) {
         return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));
       }
-    `,u=`
+    `;
+    defineSpecialInf = `
       uniform float INFINITY;
 
       bool isinf(float val) {
@@ -100,7 +49569,8 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       bvec4 isinf(vec4 val) {
         return equal(abs(val), vec4(INFINITY));
       }
-    `,l=`
+    `;
+    defineRound = `
       int round(float value) {
         return int(floor(value + 0.5));
       }
@@ -108,15 +49578,74 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       ivec4 round(vec4 value) {
         return ivec4(floor(value + vec4(0.5)));
       }
-    `),{version:r,attribute:t,varyingVs:e,varyingFs:n,texture2D:o,output:s,defineOutput:i,defineSpecialNaN:a,defineSpecialInf:u,defineRound:l}}function ti(r,t,e="index"){let n=y.computeStrides(t);return n.map((o,s)=>{let i=`int ${r[s]} = ${e} / ${o}`,a=s===n.length-1?`int ${r[s+1]} = ${e} - ${r[s]} * ${o}`:`index -= ${r[s]} * ${o}`;return`${i}; ${a};`}).join("")}function Mc(r,t,e="index"){let n=y.computeStrides(t);return n.map((o,s)=>{let i=`int ${r[s]} = ${e} / outShapeStrides[${s}]`,a=s===n.length-1?`int ${r[s+1]} = ${e} - ${r[s]} * outShapeStrides[${s}]`:`index -= ${r[s]} * outShapeStrides[${s}]`;return`${i}; ${a};`}).join("")}function gtt(r,t){let e=r.length,n=r.map(s=>`${t}[${s}]`),o=new Array(e-1);o[e-2]=n[e-1];for(let s=e-3;s>=0;--s)o[s]=`(${o[s+1]} * ${n[s+1]})`;return o}function eL(r,t,e="index"){let n=r.map((s,i)=>i),o=gtt(n,t);return o.map((s,i)=>{let a=`int ${r[i]} = ${e} / ${o[i]}`,u=i===o.length-1?`int ${r[i+1]} = ${e} - ${r[i]} * ${o[i]}`:`index -= ${r[i]} * ${o[i]}`;return`${a}; ${u};`}).join("")}function hd(r){let t=y.computeStrides(r).map(e=>e.toString());return`
-  int getFlatIndex(ivec3 coords) {
-    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;
+    `;
   }
-`}function gd(){return`
+  return {
+    version: version10,
+    attribute,
+    varyingVs,
+    varyingFs,
+    texture2D,
+    output,
+    defineOutput,
+    defineSpecialNaN,
+    defineSpecialInf,
+    defineRound
+  };
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/shader_compiler_util.js
+function getLogicalCoordinatesFromFlatIndex(coords2, shape, index = "index") {
+  const strides = util_exports.computeStrides(shape);
+  return strides.map((stride, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / ${stride}`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${stride}` : `index -= ${coords2[i]} * ${stride}`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function getOutputLogicalCoordinatesFromFlatIndexByUniform(coords2, shape, index = "index") {
+  const strides = util_exports.computeStrides(shape);
+  return strides.map((_, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / outShapeStrides[${i}]`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * outShapeStrides[${i}]` : `index -= ${coords2[i]} * outShapeStrides[${i}]`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function symbolicallyComputeStrides(indicesArr, variableName) {
+  const numCoords = indicesArr.length;
+  const shape = indicesArr.map((d) => `${variableName}[${d}]`);
+  const strides = new Array(numCoords - 1);
+  strides[numCoords - 2] = shape[numCoords - 1];
+  for (let i = numCoords - 3; i >= 0; --i) {
+    strides[i] = `(${strides[i + 1]} * ${shape[i + 1]})`;
+  }
+  return strides;
+}
+function getLogicalCoordinatesFromFlatIndexByUniform(coords2, variableName, index = "index") {
+  const indicesArray = coords2.map((_, i) => i);
+  const strides = symbolicallyComputeStrides(indicesArray, variableName);
+  return strides.map((_, i) => {
+    const line1 = `int ${coords2[i]} = ${index} / ${strides[i]}`;
+    const line2 = i === strides.length - 1 ? `int ${coords2[i + 1]} = ${index} - ${coords2[i]} * ${strides[i]}` : `index -= ${coords2[i]} * ${strides[i]}`;
+    return `${line1}; ${line2};`;
+  }).join("");
+}
+function getFlatIndexFrom3D(shape) {
+  const strides = util_exports.computeStrides(shape).map((d) => d.toString());
+  return `
+  int getFlatIndex(ivec3 coords) {
+    return coords.x * ${strides[0]} + coords.y * ${strides[1]} + coords.z;
+  }
+`;
+}
+function getFlatIndexFrom3DOutput() {
+  return `
   int getFlatIndex(ivec3 coords) {
     return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;
   }
-`}var Nw=`
+`;
+}
+var ENCODE_FLOAT_SNIPPET = `
   const float FLOAT_MAX = 1.70141184e38;
   const float FLOAT_MIN = 1.17549435e-38;
 
@@ -155,27 +49684,213 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
 
     return c / 255.0;
   }
-`;var{getBroadcastDims:rL}=v;function nL(r,t,e){let n=[];if(r.forEach(f=>{let d=y.sizeFromShape(f.shapeInfo.logicalShape);if(f.shapeInfo.isUniform?n.push(`uniform float ${f.name}${d>1?`[${d}]`:""};`):(n.push(`uniform sampler2D ${f.name};`),n.push(`uniform int offset${f.name};`)),e.enableShapeUniforms){let{uniformShape:h}=Tw(e.packedInputs,f.shapeInfo.logicalShape,f.shapeInfo.texShape);switch(h.length){case 1:n.push(`uniform int ${f.name}Shape;`);break;case 2:n.push(`uniform ivec2 ${f.name}Shape;`);break;case 3:n.push(`uniform ivec3 ${f.name}Shape;`);break;case 4:n.push(`uniform ivec4 ${f.name}Shape;`);break;default:break}n.push(`uniform ivec2 ${f.name}TexShape;`)}}),e.enableShapeUniforms){switch(t.logicalShape.length){case 1:n.push("uniform int outShape;");break;case 2:n.push("uniform ivec2 outShape;"),n.push("uniform int outShapeStrides;");break;case 3:n.push("uniform ivec3 outShape;"),n.push("uniform ivec2 outShapeStrides;");break;case 4:n.push("uniform ivec4 outShape;"),n.push("uniform ivec3 outShapeStrides;");break;default:break}n.push("uniform ivec2 outTexShape;")}e.customUniforms&&e.customUniforms.forEach(f=>{n.push(`uniform ${f.type} ${f.name}${f.arrayIndex?`[${f.arrayIndex}]`:""};`)});let o=n.join(`
-`),s=r.map(f=>xtt(f,t,e.packedInputs,e.enableShapeUniforms)).join(`
-`),i=t.texShape,a=Ge(),u=wtt(a),l,c,p=Stt(a);return t.isPacked?(l=ytt(t.logicalShape,i,e.enableShapeUniforms),c=Itt(a)):(l=btt(t.logicalShape,i,e.enableShapeUniforms),c=Ctt(a)),e.packedInputs&&(p+=ktt),[p,u,c,o,l,s,e.userCode].join(`
-`)}function yd(r,t=!1){let e=r.shapeInfo.logicalShape;switch(e.length){case 0:return ztt(r,t);case 1:return Vtt(r,t);case 2:return Wtt(r,t);case 3:return Htt(r,t);case 4:return Ktt(r,t);case 5:return jtt(r);case 6:return Xtt(r);default:throw new Error(`${e.length}-D input sampling is not yet supported`)}}function oL(r,t){switch(r.shapeInfo.logicalShape.length){case 0:return Mtt(r);case 1:return Btt(r,t);case 2:return Gtt(r,t);case 3:return Utt(r,t);default:return qtt(r,t)}}function xtt(r,t,e=!1,n){let o="";e?o+=oL(r,n):o+=yd(r,n);let s=r.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(e?o+=Ytt(r,t):o+=Ztt(r,t)),o}function ytt(r,t,e){switch(r.length){case 0:return sL();case 1:return Ett(r,t,e);case 2:return Ptt(r,t,e);case 3:return Att(r,t,e);default:return Dtt(r,t,e)}}function btt(r,t,e){switch(r.length){case 0:return sL();case 1:return _tt(r,t,e);case 2:return Ltt(r,t,e);case 3:return $tt(r,t,e);case 4:return Rtt(r,t,e);case 5:return Ftt(r,t);case 6:return Ott(r,t);default:throw new Error(`${r.length}-D output sampling is not yet supported`)}}function wtt(r){return`
+`;
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/shader_compiler.js
+var { getBroadcastDims: getBroadcastDims2 } = backend_util_exports;
+function makeShader(inputsInfo, outputShape, program) {
+  const prefixSnippets = [];
+  inputsInfo.forEach((x) => {
+    const size = util_exports.sizeFromShape(x.shapeInfo.logicalShape);
+    if (x.shapeInfo.isUniform) {
+      prefixSnippets.push(`uniform float ${x.name}${size > 1 ? `[${size}]` : ""};`);
+    } else {
+      prefixSnippets.push(`uniform sampler2D ${x.name};`);
+      prefixSnippets.push(`uniform int offset${x.name};`);
+    }
+    if (program.enableShapeUniforms) {
+      const { uniformShape } = getUniformInfoFromShape(program.packedInputs, x.shapeInfo.logicalShape, x.shapeInfo.texShape);
+      switch (uniformShape.length) {
+        case 1:
+          prefixSnippets.push(`uniform int ${x.name}Shape;`);
+          break;
+        case 2:
+          prefixSnippets.push(`uniform ivec2 ${x.name}Shape;`);
+          break;
+        case 3:
+          prefixSnippets.push(`uniform ivec3 ${x.name}Shape;`);
+          break;
+        case 4:
+          prefixSnippets.push(`uniform ivec4 ${x.name}Shape;`);
+          break;
+        default:
+          break;
+      }
+      prefixSnippets.push(`uniform ivec2 ${x.name}TexShape;`);
+    }
+  });
+  if (program.enableShapeUniforms) {
+    switch (outputShape.logicalShape.length) {
+      case 1:
+        prefixSnippets.push(`uniform int outShape;`);
+        break;
+      case 2:
+        prefixSnippets.push(`uniform ivec2 outShape;`);
+        prefixSnippets.push(`uniform int outShapeStrides;`);
+        break;
+      case 3:
+        prefixSnippets.push(`uniform ivec3 outShape;`);
+        prefixSnippets.push(`uniform ivec2 outShapeStrides;`);
+        break;
+      case 4:
+        prefixSnippets.push(`uniform ivec4 outShape;`);
+        prefixSnippets.push(`uniform ivec3 outShapeStrides;`);
+        break;
+      default:
+        break;
+    }
+    prefixSnippets.push(`uniform ivec2 outTexShape;`);
+  }
+  if (program.customUniforms) {
+    program.customUniforms.forEach((d) => {
+      prefixSnippets.push(`uniform ${d.type} ${d.name}${d.arrayIndex ? `[${d.arrayIndex}]` : ""};`);
+    });
+  }
+  const inputPrefixSnippet = prefixSnippets.join("\n");
+  const inputSamplingSnippet = inputsInfo.map((x) => getInputSamplingSnippet(x, outputShape, program.packedInputs, program.enableShapeUniforms)).join("\n");
+  const outTexShape = outputShape.texShape;
+  const glsl = getGlslDifferences();
+  const floatTextureSampleSnippet = getFloatTextureSampleSnippet(glsl);
+  let outputSamplingSnippet;
+  let floatTextureSetOutputSnippet;
+  let shaderPrefix = getShaderPrefix(glsl);
+  if (outputShape.isPacked) {
+    outputSamplingSnippet = getPackedOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);
+    floatTextureSetOutputSnippet = getFloatTextureSetRGBASnippet(glsl);
+  } else {
+    outputSamplingSnippet = getOutputSamplingSnippet(outputShape.logicalShape, outTexShape, program.enableShapeUniforms);
+    floatTextureSetOutputSnippet = getFloatTextureSetRSnippet(glsl);
+  }
+  if (program.packedInputs) {
+    shaderPrefix += SHADER_PACKED_PREFIX;
+  }
+  const source = [
+    shaderPrefix,
+    floatTextureSampleSnippet,
+    floatTextureSetOutputSnippet,
+    inputPrefixSnippet,
+    outputSamplingSnippet,
+    inputSamplingSnippet,
+    program.userCode
+  ].join("\n");
+  return source;
+}
+function getSamplerFromInInfo(inInfo, enableShapeUniforms = false) {
+  const shape = inInfo.shapeInfo.logicalShape;
+  switch (shape.length) {
+    case 0:
+      return getSamplerScalar(inInfo, enableShapeUniforms);
+    case 1:
+      return getSampler1D(inInfo, enableShapeUniforms);
+    case 2:
+      return getSampler2D(inInfo, enableShapeUniforms);
+    case 3:
+      return getSampler3D(inInfo, enableShapeUniforms);
+    case 4:
+      return getSampler4D(inInfo, enableShapeUniforms);
+    case 5:
+      return getSampler5D(inInfo);
+    case 6:
+      return getSampler6D(inInfo);
+    default:
+      throw new Error(`${shape.length}-D input sampling is not yet supported`);
+  }
+}
+function getPackedSamplerFromInInfo(inInfo, enableShapeUniforms) {
+  const shape = inInfo.shapeInfo.logicalShape;
+  switch (shape.length) {
+    case 0:
+      return getPackedSamplerScalar(inInfo);
+    case 1:
+      return getPackedSampler1D(inInfo, enableShapeUniforms);
+    case 2:
+      return getPackedSampler2D(inInfo, enableShapeUniforms);
+    case 3:
+      return getPackedSampler3D(inInfo, enableShapeUniforms);
+    default:
+      return getPackedSamplerND(inInfo, enableShapeUniforms);
+  }
+}
+function getInputSamplingSnippet(inInfo, outShapeInfo, usesPackedTextures = false, enableShapeUniforms) {
+  let res = "";
+  if (usesPackedTextures) {
+    res += getPackedSamplerFromInInfo(inInfo, enableShapeUniforms);
+  } else {
+    res += getSamplerFromInInfo(inInfo, enableShapeUniforms);
+  }
+  const inShape = inInfo.shapeInfo.logicalShape;
+  const outShape = outShapeInfo.logicalShape;
+  if (inShape.length <= outShape.length) {
+    if (usesPackedTextures) {
+      res += getPackedSamplerAtOutputCoords(inInfo, outShapeInfo);
+    } else {
+      res += getSamplerAtOutputCoords(inInfo, outShapeInfo);
+    }
+  }
+  return res;
+}
+function getPackedOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {
+  switch (outShape.length) {
+    case 0:
+      return getOutputScalarCoords();
+    case 1:
+      return getOutputPacked1DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 2:
+      return getOutputPacked2DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 3:
+      return getOutputPacked3DCoords(outShape, outTexShape, enableShapeUniforms);
+    default:
+      return getOutputPackedNDCoords(outShape, outTexShape, enableShapeUniforms);
+  }
+}
+function getOutputSamplingSnippet(outShape, outTexShape, enableShapeUniforms) {
+  switch (outShape.length) {
+    case 0:
+      return getOutputScalarCoords();
+    case 1:
+      return getOutput1DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 2:
+      return getOutput2DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 3:
+      return getOutput3DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 4:
+      return getOutput4DCoords(outShape, outTexShape, enableShapeUniforms);
+    case 5:
+      return getOutput5DCoords(outShape, outTexShape);
+    case 6:
+      return getOutput6DCoords(outShape, outTexShape);
+    default:
+      throw new Error(`${outShape.length}-D output sampling is not yet supported`);
+  }
+}
+function getFloatTextureSampleSnippet(glsl) {
+  return `
     float sampleTexture(sampler2D textureSampler, vec2 uv) {
-      return ${r.texture2D}(textureSampler, uv).r;
+      return ${glsl.texture2D}(textureSampler, uv).r;
     }
-  `}function Ctt(r){return`
+  `;
+}
+function getFloatTextureSetRSnippet(glsl) {
+  return `
     void setOutput(float val) {
-      ${r.output} = vec4(val, 0, 0, 0);
+      ${glsl.output} = vec4(val, 0, 0, 0);
     }
-  `}function Itt(r){return`
+  `;
+}
+function getFloatTextureSetRGBASnippet(glsl) {
+  return `
     void setOutput(vec4 val) {
-      ${r.output} = val;
+      ${glsl.output} = val;
     }
-  `}function Stt(r){return`${r.version}
+  `;
+}
+function getShaderPrefix(glsl) {
+  const SHADER_PREFIX = `${glsl.version}
     precision highp float;
     precision highp int;
     precision highp sampler2D;
-    ${r.varyingFs} vec2 resultUV;
-    ${r.defineOutput}
+    ${glsl.varyingFs} vec2 resultUV;
+    ${glsl.defineOutput}
     const vec2 halfCR = vec2(0.5, 0.5);
 
     struct ivec5
@@ -198,9 +49913,9 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
     };
 
     uniform float NAN;
-    ${r.defineSpecialNaN}
-    ${r.defineSpecialInf}
-    ${r.defineRound}
+    ${glsl.defineSpecialNaN}
+    ${glsl.defineSpecialInf}
+    ${glsl.defineRound}
 
     int imod(int x, int y) {
       return x - y * (x / y);
@@ -225,10 +49940,13 @@ Hi, looks like you are running TensorFlow.js in Node.js. To speed things up dram
       return fract((p3.x + p3.y) * p3.z);
     }
 
-    ${vtt}
-    ${Ntt}
-    ${Ttt}
-  `}var vtt=`
+    ${SAMPLE_1D_SNIPPET}
+    ${SAMPLE_2D_SNIPPET}
+    ${SAMPLE_3D_SNIPPET}
+  `;
+  return SHADER_PREFIX;
+}
+var SAMPLE_1D_SNIPPET = `
 vec2 uvFromFlat(int texNumR, int texNumC, int index) {
   int texR = index / texNumC;
   int texC = index - texR * texNumC;
@@ -240,7 +49958,8 @@ vec2 packedUVfrom1D(int texNumR, int texNumC, int index) {
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,Ntt=`
+`;
+var SAMPLE_2D_SNIPPET = `
 vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texNumC, int row, int col) {
   int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);
@@ -248,7 +49967,8 @@ vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
   int texC = texelIndex - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,Ttt=`
+`;
+var SAMPLE_3D_SNIPPET = `
 vec2 packedUVfrom3D(int texNumR, int texNumC,
     int texelsInBatch, int texelsInLogicalRow, int b,
     int row, int col) {
@@ -257,7 +49977,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   int texC = index - texR * texNumC;
   return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
 }
-`,ktt=`
+`;
+var SHADER_PACKED_PREFIX = `
   float getChannel(vec4 frag, vec2 innerDims) {
     vec2 modCoord = mod(innerDims, 2.);
     return modCoord.x == 0. ?
@@ -268,68 +49989,111 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
     float modCoord = mod(float(dim), 2.);
     return modCoord == 0. ? frag.r : frag.g;
   }
-`;function sL(){return`
+`;
+function getOutputScalarCoords() {
+  return `
     int getOutputCoords() {
       return 0;
     }
-  `}function Ett(r,t,e){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return n[0]===1?e?`
+  `;
+}
+function getOutputPacked1DCoords(shape, texShape, enableShapeUniforms) {
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (packedTexShape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return 2 * int(resultUV.x * ${n[1]}.0);
+        return 2 * int(resultUV.x * ${packedTexShape[1]}.0);
       }
-    `:n[1]===1?e?`
+    `;
+  }
+  if (packedTexShape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return 2 * int(resultUV.y * ${n[0]}.0);
+        return 2 * int(resultUV.y * ${packedTexShape[0]}.0);
       }
-    `:e?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     int getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(packedTexShape[0], packedTexShape[1]));
       return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);
     }
-  `:`
+  `;
+  }
+  return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${n[0]}, ${n[1]}));
-      return 2 * (resTexRC.x * ${n[1]} + resTexRC.y);
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      return 2 * (resTexRC.x * ${packedTexShape[1]} + resTexRC.y);
     }
-  `}function _tt(r,t,e){return t[0]===1?e?`
+  `;
+}
+function getOutput1DCoords(shape, texShape, enableShapeUniforms) {
+  if (texShape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return int(resultUV.x * float(outTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return int(resultUV.x * ${t[1]}.0);
+        return int(resultUV.x * ${texShape[1]}.0);
       }
-    `:t[1]===1?e?`
+    `;
+  }
+  if (texShape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       int getOutputCoords() {
         return int(resultUV.y * float(outTexShape[0]));
       }
-    `:`
+    `;
+    }
+    return `
       int getOutputCoords() {
-        return int(resultUV.y * ${t[0]}.0);
+        return int(resultUV.y * ${texShape[0]}.0);
       }
-    `:e?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(outTexShape[0], outTexShape[1]));
       return resTexRC.x * outTexShape[1] + resTexRC.y;
     }
-  `:`
+  `;
+  }
+  return `
     int getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      return resTexRC.x * ${t[1]} + resTexRC.y;
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      return resTexRC.x * ${texShape[1]} + resTexRC.y;
     }
-  `}function Att(r,t,e){if(e)return`
+  `;
+}
+function getOutputPacked3DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    return `
     ivec3 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));
@@ -346,37 +50110,54 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec3(b, r, c);
     }
-  `;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],o=Math.ceil(r[2]/2),s=o*Math.ceil(r[1]/2);return`
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texelsInLogicalRow = Math.ceil(shape[2] / 2);
+  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[1] / 2);
+  return `
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${n[0]}, ${n[1]}));
-      int index = resTexRC.x * ${n[1]} + resTexRC.y;
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${texelsInBatch};
+      index -= b * ${texelsInBatch};
 
-      int r = 2 * (index / ${o});
-      int c = imod(index, ${o}) * 2;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
       return ivec3(b, r, c);
     }
-  `}function $tt(r,t,e){if(e)return`
+  `;
+}
+function getOutput3DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], shape);
+    return `
   ivec3 getOutputCoords() {
     ivec2 resTexRC = ivec2(resultUV.yx *
                            vec2(outTexShape[0], outTexShape[1]));
     int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-    ${Mc(["r","c","d"],r)}
+    ${coordsFromIndexSnippet2}
     return ivec3(r, c, d);
   }
-`;let n=ti(["r","c","d"],r);return`
+`;
+  }
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], shape);
+  return `
     ivec3 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${n}
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      ${coordsFromIndexSnippet}
       return ivec3(r, c, d);
     }
-  `}function Dtt(r,t,e){if(e)return`
+  `;
+}
+function getOutputPackedNDCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    return `
     ivec4 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       ivec2 resTexRC = ivec2(resultUV.yx *
@@ -398,74 +50179,115 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec4(b2, b, r, c);
     }
-  `;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],o=Math.ceil(r[r.length-1]/2),s=o*Math.ceil(r[r.length-2]/2),i=s,a="",u="b, r, c";for(let l=2;l<r.length-1;l++)i*=r[r.length-l-1],a=`
-      int b${l} = index / ${i};
-      index -= b${l} * ${i};
-    `+a,u=`b${l}, `+u;return`
-    ivec${r.length} getOutputCoords() {
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texelsInLogicalRow = Math.ceil(shape[shape.length - 1] / 2);
+  const texelsInBatch = texelsInLogicalRow * Math.ceil(shape[shape.length - 2] / 2);
+  let texelsInBatchN = texelsInBatch;
+  let batches = ``;
+  let coords2 = "b, r, c";
+  for (let b = 2; b < shape.length - 1; b++) {
+    texelsInBatchN *= shape[shape.length - b - 1];
+    batches = `
+      int b${b} = index / ${texelsInBatchN};
+      index -= b${b} * ${texelsInBatchN};
+    ` + batches;
+    coords2 = `b${b}, ` + coords2;
+  }
+  return `
+    ivec${shape.length} getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${n[0]}, ${n[1]}));
-      int index = resTexRC.x * ${n[1]} + resTexRC.y;
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
 
-      ${a}
+      ${batches}
 
-      int b = index / ${s};
-      index -= b * ${s};
+      int b = index / ${texelsInBatch};
+      index -= b * ${texelsInBatch};
 
-      int r = 2 * (index / ${o});
-      int c = imod(index, ${o}) * 2;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
-      return ivec${r.length}(${u});
+      return ivec${shape.length}(${coords2});
     }
-  `}function Rtt(r,t,e){if(e)return`
+  `;
+}
+function getOutput4DCoords(shape, texShape, enableShapeUniforms) {
+  if (enableShapeUniforms) {
+    const coordsFromIndexSnippet2 = getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d", "d2"], shape);
+    return `
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
         vec2(outTexShape[0], outTexShape[1]));
       int index = resTexRC.x * outTexShape[1] + resTexRC.y;
-      ${Mc(["r","c","d","d2"],r)}
+      ${coordsFromIndexSnippet2}
       return ivec4(r, c, d, d2);
     }
-  `;let n=ti(["r","c","d","d2"],r);return`
+  `;
+  }
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2"], shape);
+  return `
     ivec4 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-        vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      ${n}
+        vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      ${coordsFromIndexSnippet}
       return ivec4(r, c, d, d2);
     }
-  `}function Ftt(r,t){let e=ti(["r","c","d","d2","d3"],r);return`
+  `;
+}
+function getOutput5DCoords(shape, texShape) {
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2", "d3"], shape);
+  return `
     ivec5 getOutputCoords() {
-      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},
-                             ${t[1]}));
+      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${texShape[0]},
+                             ${texShape[1]}));
 
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
 
-      ${e}
+      ${coordsFromIndexSnippet}
 
       ivec5 outShape = ivec5(r, c, d, d2, d3);
       return outShape;
     }
-  `}function Ott(r,t){let e=ti(["r","c","d","d2","d3","d4"],r);return`
+  `;
+}
+function getOutput6DCoords(shape, texShape) {
+  const coordsFromIndexSnippet = getLogicalCoordinatesFromFlatIndex(["r", "c", "d", "d2", "d3", "d4"], shape);
+  return `
     ivec6 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-        vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
+        vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
 
-      ${e}
+      ${coordsFromIndexSnippet}
 
       ivec6 result = ivec6(r, c, d, d2, d3, d4);
       return result;
     }
-  `}function Ptt(r,t,e){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(y.arraysEqual(r,t))return e?`
+  `;
+}
+function getOutputPacked2DCoords(shape, texShape, enableShapeUniforms) {
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
         return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
-        return 2 * ivec2(resultUV.yx * vec2(${n[0]}, ${n[1]}));
+        return 2 * ivec2(resultUV.yx * vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
       }
-    `;let o=Math.ceil(r[1]/2);return e?`
+    `;
+  }
+  const texelsInLogicalRow = Math.ceil(shape[1] / 2);
+  if (enableShapeUniforms) {
+    return `
     ivec2 getOutputCoords() {
       ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
       int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));
@@ -478,54 +50300,78 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
       return ivec2(r, c);
     }
-  `:`
+  `;
+  }
+  return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${n[0]}, ${n[1]}));
+                             vec2(${packedTexShape[0]}, ${packedTexShape[1]}));
 
-      int index = resTexRC.x * ${n[1]} + resTexRC.y;
-      int r = 2 * (index / ${o});
-      int c = imod(index, ${o}) * 2;
+      int index = resTexRC.x * ${packedTexShape[1]} + resTexRC.y;
+      int r = 2 * (index / ${texelsInLogicalRow});
+      int c = imod(index, ${texelsInLogicalRow}) * 2;
 
       return ivec2(r, c);
     }
-  `}function Ltt(r,t,e){return y.arraysEqual(r,t)?e?`
+  `;
+}
+function getOutput2DCoords(shape, texShape, enableShapeUniforms) {
+  if (util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
-        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));
+        return ivec2(resultUV.yx * vec2(${texShape[0]}, ${texShape[1]}));
       }
-    `:r[1]===1?e?`
+    `;
+  }
+  if (shape[1] === 1) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
                                vec2(outTexShape[0], outTexShape[1]));
         int index = resTexRC.x * outTexShape[1] + resTexRC.y;
         return ivec2(index, 0);
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
-                               vec2(${t[0]}, ${t[1]}));
-        int index = resTexRC.x * ${t[1]} + resTexRC.y;
+                               vec2(${texShape[0]}, ${texShape[1]}));
+        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
         return ivec2(index, 0);
       }
-    `:r[0]===1?e?`
+    `;
+  }
+  if (shape[0] === 1) {
+    if (enableShapeUniforms) {
+      return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
                                vec2(outTexShape[0], outTexShape[1]));
         int index = resTexRC.x * outTexShape[1] + resTexRC.y;
         return ivec2(0, index);
       }
-    `:`
+    `;
+    }
+    return `
       ivec2 getOutputCoords() {
         ivec2 resTexRC = ivec2(resultUV.yx *
-                               vec2(${t[0]}, ${t[1]}));
-        int index = resTexRC.x * ${t[1]} + resTexRC.y;
+                               vec2(${texShape[0]}, ${texShape[1]}));
+        int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
         return ivec2(0, index);
       }
-    `:e?`
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
                              vec2(outTexShape[0], outTexShape[1]));
@@ -534,459 +50380,1208 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
       int c = index - r * outShape[1];
       return ivec2(r, c);
     }
-  `:`
+  `;
+  }
+  return `
     ivec2 getOutputCoords() {
       ivec2 resTexRC = ivec2(resultUV.yx *
-                             vec2(${t[0]}, ${t[1]}));
-      int index = resTexRC.x * ${t[1]} + resTexRC.y;
-      int r = index / ${r[1]};
-      int c = index - r * ${r[1]};
+                             vec2(${texShape[0]}, ${texShape[1]}));
+      int index = resTexRC.x * ${texShape[1]} + resTexRC.y;
+      int r = index / ${shape[1]};
+      int c = index - r * ${shape[1]};
       return ivec2(r, c);
     }
-  `}function zc(r){return`offset${r}`}function Mtt(r){let t=r.name,e="get"+t.charAt(0).toUpperCase()+t.slice(1),n=Ge();return`
-    vec4 ${e}() {
-      return ${n.texture2D}(${t}, halfCR);
+  `;
+}
+function getFlatOffsetUniformName(texName) {
+  return `offset${texName}`;
+}
+function getPackedSamplerScalar(inputInfo) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const glsl = getGlslDifferences();
+  return `
+    vec4 ${funcName}() {
+      return ${glsl.texture2D}(${texName}, halfCR);
     }
-  `}function ztt(r,t){let e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1);if(r.shapeInfo.isUniform)return`float ${n}() {return ${e};}`;let[o,s]=r.shapeInfo.texShape;if(o===1&&s===1)return`
-      float ${n}() {
-        return sampleTexture(${e}, halfCR);
+  `;
+}
+function getSamplerScalar(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  if (inputInfo.shapeInfo.isUniform) {
+    return `float ${funcName}() {return ${texName};}`;
+  }
+  const [texNumR, texNumC] = inputInfo.shapeInfo.texShape;
+  if (texNumR === 1 && texNumC === 1) {
+    return `
+      float ${funcName}() {
+        return sampleTexture(${texName}, halfCR);
       }
-    `;let i=zc(e);if(t)return`
-    float ${n}() {
-      vec2 uv = uvFromFlat(${e}TexShape[0], ${e}TexShape[1], ${i});
-      return sampleTexture(${e}, uv);
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}() {
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `;let[a,u]=r.shapeInfo.texShape;return`
-    float ${n}() {
-      vec2 uv = uvFromFlat(${a}, ${u}, ${i});
-      return sampleTexture(${e}, uv);
+  `;
+  }
+  const [tNumR, tNumC] = inputInfo.shapeInfo.texShape;
+  return `
+    float ${funcName}() {
+      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `}function Btt(r,t){let e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1),o=r.shapeInfo.texShape,s=Ge();if(t)return`
-    vec4 ${n}(int index) {
-      ivec2 packedTexShape = ivec2(ceil(float(${e}TexShape[0]) / 2.0), ceil(float(${e}TexShape[1]) / 2.0));
+  `;
+}
+function getPackedSampler1D(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int index) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
       vec2 uv = packedUVfrom1D(
         packedTexShape[0], packedTexShape[1], index);
-      return ${s.texture2D}(${e}, uv);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let i=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)];return`
-    vec4 ${n}(int index) {
-      vec2 uv = packedUVfrom1D(
-        ${i[0]}, ${i[1]}, index);
-      return ${s.texture2D}(${e}, uv);
-    }
-  `}function Vtt(r,t){let e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1);if(r.shapeInfo.isUniform)return`
-      float ${n}(int index) {
-        ${bd(r)}
-      }
-    `;let o=r.shapeInfo.texShape,s=o[0],i=o[1];if(i===1&&s===1)return`
-      float ${n}(int index) {
-        return sampleTexture(${e}, halfCR);
-      }
-    `;let a=zc(e);return i===1?t?`
-      float ${n}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${a}) + 0.5) / float(${e}TexShape[0]));
-        return sampleTexture(${e}, uv);
-      }
-    `:`
-      float ${n}(int index) {
-        vec2 uv = vec2(0.5, (float(index + ${a}) + 0.5) / ${s}.0);
-        return sampleTexture(${e}, uv);
-      }
-    `:s===1?t?`
-      float ${n}(int index) {
-        vec2 uv = vec2((float(index + ${a}) + 0.5) / float(${e}TexShape[1]), 0.5);
-        return sampleTexture(${e}, uv);
-      }
-    `:`
-      float ${n}(int index) {
-        vec2 uv = vec2((float(index + ${a}) + 0.5) / ${i}.0, 0.5);
-        return sampleTexture(${e}, uv);
-      }
-    `:t?`
-    float ${n}(int index) {
-      vec2 uv = uvFromFlat(${e}TexShape[0], ${e}TexShape[1], index + ${a});
-      return sampleTexture(${e}, uv);
-    }
-  `:`
-    float ${n}(int index) {
-      vec2 uv = uvFromFlat(${s}, ${i}, index + ${a});
-      return sampleTexture(${e}, uv);
-    }
-  `}function Gtt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o="get"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape,i=s[0],a=s[1],u=Ge();if(s!=null&&y.arraysEqual(e,s))return t?`
-      vec4 ${o}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
-
-        return ${u.texture2D}(${n}, uv);
-      }
-    `:`
-      vec4 ${o}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${a}.0, ${i}.0);
-
-        return ${u.texture2D}(${n}, uv);
-      }
-    `;if(t)return`
-    vec4 ${o}(int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${n}Shape[1]) / 2.0));
-      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
-      return ${u.texture2D}(${n}, uv);
-    }
-  `;let l=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],c=Math.ceil(e[1]/2);return`
-    vec4 ${o}(int row, int col) {
-      vec2 uv = packedUVfrom2D(${c}, ${l[0]}, ${l[1]}, row, col);
-      return ${u.texture2D}(${n}, uv);
-    }
-  `}function Wtt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o="get"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape;if(s!=null&&y.arraysEqual(e,s)){if(t)return`
-      float ${o}(int row, int col) {
-        vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
-        return sampleTexture(${n}, uv);
-      }
-    `;let m=s[0],f=s[1];return`
-    float ${o}(int row, int col) {
-      vec2 uv = (vec2(col, row) + halfCR) / vec2(${f}.0, ${m}.0);
-      return sampleTexture(${n}, uv);
-    }
-  `}let{newShape:i,keptDims:a}=y.squeezeShape(e),u=i;if(u.length<e.length){let m=wd(r,u),f=["row","col"];return`
-      ${yd(m,t)}
-      float ${o}(int row, int col) {
-        return ${o}(${Cd(f,a)});
-      }
-    `}if(r.shapeInfo.isUniform)return`
-      float ${o}(int row, int col) {
-        int index = round(dot(vec2(row, col), vec2(${e[1]}, 1)));
-        ${bd(r)}
-      }
-    `;let l=s[0],c=s[1],p=zc(n);return c===1?t?`
-      float ${o}(int row, int col) {
-        float index = dot(vec3(row, col, ${p}), vec3(${n}Shape[1], 1, 1));
-        vec2 uv = vec2(0.5, (index + 0.5) / float(${n}TexShape[0]));
-        return sampleTexture(${n}, uv);
-      }
-    `:`
-    float ${o}(int row, int col) {
-      float index = dot(vec3(row, col, ${p}), vec3(${e[1]}, 1, 1));
-      vec2 uv = vec2(0.5, (index + 0.5) / ${l}.0);
-      return sampleTexture(${n}, uv);
-    }
-  `:l===1?t?`
-      float ${o}(int row, int col) {
-        float index = dot(vec3(row, col, ${p}), vec3(${n}Shape[1], 1, 1));
-        vec2 uv = vec2((index + 0.5) / float(${n}TexShape[1]), 0.5);
-        return sampleTexture(${n}, uv);
-      }
-    `:`
-    float ${o}(int row, int col) {
-      float index = dot(vec3(row, col, ${p}), vec3(${e[1]}, 1, 1));
-      vec2 uv = vec2((index + 0.5) / ${c}.0, 0.5);
-      return sampleTexture(${n}, uv);
-    }
-  `:t?`
-      float ${o}(int row, int col) {
-        // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${n}Shape[1] + col + ${p};
-        vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);
-        return sampleTexture(${n}, uv);
-      }
-    `:`
-  float ${o}(int row, int col) {
-    // Explicitly use integer operations as dot() only works on floats.
-    int index = row * ${e[1]} + col + ${p};
-    vec2 uv = uvFromFlat(${l}, ${c}, index);
-    return sampleTexture(${n}, uv);
+  `;
   }
-`}function Utt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o="get"+n.charAt(0).toUpperCase()+n.slice(1),s=r.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(e[0]===1){let m=e.slice(1),f=[1,2],d=wd(r,m),h=["b","row","col"];return`
-        ${oL(d,t)}
-        vec4 ${o}(int b, int row, int col) {
-          return ${o}(${Cd(h,f)});
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  return `
+    vec4 ${funcName}(int index) {
+      vec2 uv = packedUVfrom1D(
+        ${packedTexShape[0]}, ${packedTexShape[1]}, index);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+}
+function getSampler1D(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int index) {
+        ${getUniformSampler(inputInfo)}
+      }
+    `;
+  }
+  const texShape = inputInfo.shapeInfo.texShape;
+  const tNumR = texShape[0];
+  const tNumC = texShape[1];
+  if (tNumC === 1 && tNumR === 1) {
+    return `
+      float ${funcName}(int index) {
+        return sampleTexture(${texName}, halfCR);
+      }
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (tNumC === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / float(${texName}TexShape[0]));
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2(0.5, (float(index + ${offset}) + 0.5) / ${tNumR}.0);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  if (tNumR === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2((float(index + ${offset}) + 0.5) / float(${texName}TexShape[1]), 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      float ${funcName}(int index) {
+        vec2 uv = vec2((float(index + ${offset}) + 0.5) / ${tNumC}.0, 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int index) {
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  return `
+    float ${funcName}(int index) {
+      vec2 uv = uvFromFlat(${tNumR}, ${tNumC}, index + ${offset});
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+}
+function getPackedSampler2D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const glsl = getGlslDifferences();
+  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
+      vec4 ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+
+        return ${glsl.texture2D}(${texName}, uv);
+      }
+    `;
+    }
+    return `
+      vec4 ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);
+
+        return ${glsl.texture2D}(${texName}, uv);
+      }
+    `;
+  }
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${texName}Shape[1]) / 2.0));
+      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+  }
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const valuesPerRow = Math.ceil(shape[1] / 2);
+  return `
+    vec4 ${funcName}(int row, int col) {
+      vec2 uv = packedUVfrom2D(${valuesPerRow}, ${packedTexShape[0]}, ${packedTexShape[1]}, row, col);
+      return ${glsl.texture2D}(${texName}, uv);
+    }
+  `;
+}
+function getSampler2D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  if (texShape != null && util_exports.arraysEqual(shape, texShape)) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        vec2 uv = (vec2(col, row) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    const texNumR2 = texShape[0];
+    const texNumC2 = texShape[1];
+    return `
+    float ${funcName}(int row, int col) {
+      vec2 uv = (vec2(col, row) + halfCR) / vec2(${texNumC2}.0, ${texNumR2}.0);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const squeezedShape = newShape;
+  if (squeezedShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["row", "col"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+      float ${funcName}(int row, int col) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
+      }
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col) {
+        int index = round(dot(vec2(row, col), vec2(${shape[1]}, 1)));
+        ${getUniformSampler(inputInfo)}
+      }
+    `;
+  }
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const offset = getFlatOffsetUniformName(texName);
+  if (texNumC === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));
+        vec2 uv = vec2(0.5, (index + 0.5) / float(${texName}TexShape[0]));
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col) {
+      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));
+      vec2 uv = vec2(0.5, (index + 0.5) / ${texNumR}.0);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  if (texNumR === 1) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col) {
+        float index = dot(vec3(row, col, ${offset}), vec3(${texName}Shape[1], 1, 1));
+        vec2 uv = vec2((index + 0.5) / float(${texName}TexShape[1]), 0.5);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col) {
+      float index = dot(vec3(row, col, ${offset}), vec3(${shape[1]}, 1, 1));
+      vec2 uv = vec2((index + 0.5) / ${texNumC}.0, 0.5);
+      return sampleTexture(${texName}, uv);
+    }
+  `;
+  }
+  if (enableShapeUniforms) {
+    return `
+      float ${funcName}(int row, int col) {
+        // Explicitly use integer operations as dot() only works on floats.
+        int index = row * ${texName}Shape[1] + col + ${offset};
+        vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);
+        return sampleTexture(${texName}, uv);
+      }
+    `;
+  }
+  return `
+  float ${funcName}(int row, int col) {
+    // Explicitly use integer operations as dot() only works on floats.
+    int index = row * ${shape[1]} + col + ${offset};
+    vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+    return sampleTexture(${texName}, uv);
+  }
+`;
+}
+function getPackedSampler3D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const texShape = inputInfo.shapeInfo.texShape;
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  if (shape[0] === 1) {
+    const squeezedShape = shape.slice(1);
+    const keptDims = [1, 2];
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["b", "row", "col"];
+    return `
+        ${getPackedSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+        vec4 ${funcName}(int b, int row, int col) {
+          return ${funcName}(${getSqueezedParams(params, keptDims)});
         }
-      `}let a=Ge();if(t)return`
-    vec4 ${o}(int b, int row, int col) {
-      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
-      int valuesPerRow = int(ceil(float(${n}Shape[2]) / 2.0));
-      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[1]) / 2.0));
+      `;
+  }
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int b, int row, int col) {
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
+      int valuesPerRow = int(ceil(float(${texName}Shape[2]) / 2.0));
+      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[1]) / 2.0));
       vec2 uv = packedUVfrom3D(
         packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);
-      return ${a.texture2D}(${n}, uv);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let u=i[0],l=i[1],c=Math.ceil(e[2]/2),p=c*Math.ceil(e[1]/2);return`
-    vec4 ${o}(int b, int row, int col) {
+  `;
+  }
+  const texNumR = packedTexShape[0];
+  const texNumC = packedTexShape[1];
+  const valuesPerRow = Math.ceil(shape[2] / 2);
+  const texelsInBatch = valuesPerRow * Math.ceil(shape[1] / 2);
+  return `
+    vec4 ${funcName}(int b, int row, int col) {
       vec2 uv = packedUVfrom3D(
-        ${u}, ${l}, ${p}, ${c}, b, row, col);
-      return ${a.texture2D}(${n}, uv);
+        ${texNumR}, ${texNumC}, ${texelsInBatch}, ${valuesPerRow}, b, row, col);
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `}function Htt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e[1]*e[2],i=e[2],{newShape:a,keptDims:u}=y.squeezeShape(e),l=a;if(l.length<e.length){let h=wd(r,l),g=["row","col","depth"];return`
-        ${yd(h,t)}
-        float ${o}(int row, int col, int depth) {
-          return ${o}(${Cd(g,u)});
+  `;
+}
+function getSampler3D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride0 = shape[1] * shape[2];
+  const stride1 = shape[2];
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const squeezedShape = newShape;
+  if (squeezedShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, squeezedShape);
+    const params = ["row", "col", "depth"];
+    return `
+        ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+        float ${funcName}(int row, int col, int depth) {
+          return ${funcName}(${getSqueezedParams(params, keptDims)});
         }
-      `}if(r.shapeInfo.isUniform)return`
-      float ${o}(int row, int col, int depth) {
+      `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth) {
         int index = round(dot(vec3(row, col, depth),
-                          vec3(${s}, ${i}, 1)));
-        ${bd(r)}
+                          vec3(${stride0}, ${stride1}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let c=r.shapeInfo.texShape,p=c[0],m=c[1],f=r.shapeInfo.flatOffset;if(m===s&&f==null)return t?`
-      float ${o}(int row, int col, int depth) {
-        int stride1 = ${n}Shape[2];
+    `;
+  }
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  if (texNumC === stride0 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth) {
+        int stride1 = ${texName}Shape[2];
         float texR = float(row);
         float texC = dot(vec2(col, depth), vec2(stride1, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${n}TexShape[1], ${n}TexShape[0]);
-        return sampleTexture(${n}, uv);
+                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-        float ${o}(int row, int col, int depth) {
+    `;
+    }
+    return `
+        float ${funcName}(int row, int col, int depth) {
           float texR = float(row);
-          float texC = dot(vec2(col, depth), vec2(${i}, 1));
+          float texC = dot(vec2(col, depth), vec2(${stride1}, 1));
           vec2 uv = (vec2(texC, texR) + halfCR) /
-                     vec2(${m}.0, ${p}.0);
-          return sampleTexture(${n}, uv);
+                     vec2(${texNumC}.0, ${texNumR}.0);
+          return sampleTexture(${texName}, uv);
         }
-      `;if(m===i&&f==null)return t?`
-      float ${o}(int row, int col, int depth) {
-        float texR = dot(vec2(row, col), vec2(${n}Shape[1], 1));
+      `;
+  }
+  if (texNumC === stride1 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth) {
+        float texR = dot(vec2(row, col), vec2(${texName}Shape[1], 1));
         float texC = float(depth);
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
-        return sampleTexture(${n}, uv);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-    float ${o}(int row, int col, int depth) {
-      float texR = dot(vec2(row, col), vec2(${e[1]}, 1));
+    `;
+    }
+    return `
+    float ${funcName}(int row, int col, int depth) {
+      float texR = dot(vec2(row, col), vec2(${shape[1]}, 1));
       float texC = float(depth);
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${m}.0, ${p}.0);
-      return sampleTexture(${n}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}.0, ${texNumR}.0);
+      return sampleTexture(${texName}, uv);
     }
-  `;let d=zc(n);return t?`
-    float ${o}(int row, int col, int depth) {
+  `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int row, int col, int depth) {
       // Explicitly use integer operations as dot() only works on floats.
-      int stride0 = ${n}Shape[1] * ${n}Shape[2];
-      int stride1 = ${n}Shape[2];
-      int index = row * stride0 + col * stride1 + depth + ${d};
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);
-      return sampleTexture(${n}, uv);
+      int stride0 = ${texName}Shape[1] * ${texName}Shape[2];
+      int stride1 = ${texName}Shape[2];
+      int index = row * stride0 + col * stride1 + depth + ${offset};
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index);
+      return sampleTexture(${texName}, uv);
     }
-    `:`
-      float ${o}(int row, int col, int depth) {
+    `;
+  }
+  return `
+      float ${funcName}(int row, int col, int depth) {
         // Explicitly use integer operations as dot() only works on floats.
-        int index = row * ${s} + col * ${i} + depth + ${d};
-        vec2 uv = uvFromFlat(${p}, ${m}, index);
-        return sampleTexture(${n}, uv);
+        int index = row * ${stride0} + col * ${stride1} + depth + ${offset};
+        vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+        return sampleTexture(${texName}, uv);
       }
-  `}function qtt(r,t){let e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1),o=Ge();if(t)return`
-    vec4 ${n}(int b2, int b, int row, int col) {
-      int valuesPerRow = int(ceil(float(${e}Shape[3]) / 2.0));
-      int texelsInBatch = valuesPerRow * int(ceil(float(${e}Shape[2]) / 2.0));
+  `;
+}
+function getPackedSamplerND(inputInfo, enableShapeUniforms) {
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const glsl = getGlslDifferences();
+  if (enableShapeUniforms) {
+    return `
+    vec4 ${funcName}(int b2, int b, int row, int col) {
+      int valuesPerRow = int(ceil(float(${texName}Shape[3]) / 2.0));
+      int texelsInBatch = valuesPerRow * int(ceil(float(${texName}Shape[2]) / 2.0));
       int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);
-      texelsInBatch *= ${e}Shape[1];
+      texelsInBatch *= ${texName}Shape[1];
       index = b2 * texelsInBatch + index;
-      ivec2 packedTexShape = ivec2(ceil(float(${e}TexShape[0]) / 2.0), ceil(float(${e}TexShape[1]) / 2.0));
+      ivec2 packedTexShape = ivec2(ceil(float(${texName}TexShape[0]) / 2.0), ceil(float(${texName}TexShape[1]) / 2.0));
       int texR = index / packedTexShape[1];
       int texC = index - texR * packedTexShape[1];
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${o.texture2D}(${e}, uv);
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${glsl.texture2D}(${texName}, uv);
     }
-  `;let s=r.shapeInfo.logicalShape,i=s.length,a=r.shapeInfo.texShape,u=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)],l=u[0],c=u[1],p=Math.ceil(s[i-1]/2),m=p*Math.ceil(s[i-2]/2),f="int b, int row, int col",d=`b * ${m} + (row / 2) * ${p} + (col / 2)`;for(let h=2;h<i-1;h++)f=`int b${h}, `+f,m*=s[i-h-1],d=`b${h} * ${m} + `+d;return`
-    vec4 ${n}(${f}) {
-      int index = ${d};
-      int texR = index / ${c};
-      int texC = index - texR * ${c};
-      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${c}, ${l});
-      return ${o.texture2D}(${e}, uv);
+  `;
+  }
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const rank = shape.length;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const packedTexShape = [Math.ceil(texShape[0] / 2), Math.ceil(texShape[1] / 2)];
+  const texNumR = packedTexShape[0];
+  const texNumC = packedTexShape[1];
+  const valuesPerRow = Math.ceil(shape[rank - 1] / 2);
+  let texelsInBatch = valuesPerRow * Math.ceil(shape[rank - 2] / 2);
+  let params = `int b, int row, int col`;
+  let index = `b * ${texelsInBatch} + (row / 2) * ${valuesPerRow} + (col / 2)`;
+  for (let b = 2; b < rank - 1; b++) {
+    params = `int b${b}, ` + params;
+    texelsInBatch *= shape[rank - b - 1];
+    index = `b${b} * ${texelsInBatch} + ` + index;
+  }
+  return `
+    vec4 ${funcName}(${params}) {
+      int index = ${index};
+      int texR = index / ${texNumC};
+      int texC = index - texR * ${texNumC};
+      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${texNumC}, ${texNumR});
+      return ${glsl.texture2D}(${texName}, uv);
     }
-  `}function Ktt(r,t){let e=r.shapeInfo.logicalShape,n=r.name,o="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e[3],i=e[2]*s,a=e[1]*i,{newShape:u,keptDims:l}=y.squeezeShape(e);if(u.length<e.length){let b=wd(r,u),w=["row","col","depth","depth2"];return`
-      ${yd(b,t)}
-      float ${o}(int row, int col, int depth, int depth2) {
-        return ${o}(${Cd(w,l)});
+  `;
+}
+function getSampler4D(inputInfo, enableShapeUniforms) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride2 = shape[3];
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo, enableShapeUniforms)}
+      float ${funcName}(int row, int col, int depth, int depth2) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}if(r.shapeInfo.isUniform)return`
-      float ${o}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         int index = round(dot(vec4(row, col, depth, depth2),
-                          vec4(${a}, ${i}, ${s}, 1)));
-        ${bd(r)}
+                          vec4(${stride0}, ${stride1}, ${stride2}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let c=r.shapeInfo.flatOffset,p=r.shapeInfo.texShape,m=p[0],f=p[1],d=`int stride2 = ${n}Shape[3];`,h=`int stride1 = ${n}Shape[2] * stride2;`,g=`int stride0 = ${n}Shape[1] * stride1;`;if(f===a&&c==null)return t?`
-      float ${o}(int row, int col, int depth, int depth2) {
-        ${d}
-        ${h}
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  const stride2Str = `int stride2 = ${texName}Shape[3];`;
+  const stride1Str = `int stride1 = ${texName}Shape[2] * stride2;`;
+  const stride0Str = `int stride0 = ${texName}Shape[1] * stride1;`;
+  if (texNumC === stride0 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
+        ${stride2Str}
+        ${stride1Str}
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
                 vec3(stride1, stride2, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${n}TexShape[1], ${n}TexShape[0]);
-        return sampleTexture(${n}, uv);
+                   vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-      float ${o}(int row, int col, int depth, int depth2) {
+    `;
+    }
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = float(row);
         float texC =
             dot(vec3(col, depth, depth2),
-                vec3(${i}, ${s}, 1));
+                vec3(${stride1}, ${stride2}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${f}.0, ${m}.0);
-        return sampleTexture(${n}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(f===s&&c==null)return t?`
-      float ${o}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  if (texNumC === stride2 && flatOffset == null) {
+    if (enableShapeUniforms) {
+      return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
-                         vec3(${n}Shape[1] * ${n}Shape[2], ${n}Shape[2], 1));
+                         vec3(${texName}Shape[1] * ${texName}Shape[2], ${texName}Shape[2], 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${n}TexShape[1], ${n}TexShape[0]);
-        return sampleTexture(${n}, uv);
+                  vec2(${texName}TexShape[1], ${texName}TexShape[0]);
+        return sampleTexture(${texName}, uv);
       }
-    `:`
-      float ${o}(int row, int col, int depth, int depth2) {
+    `;
+    }
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2) {
         float texR = dot(vec3(row, col, depth),
-                         vec3(${e[1]*e[2]}, ${e[2]}, 1));
+                         vec3(${shape[1] * shape[2]}, ${shape[2]}, 1));
         float texC = float(depth2);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${f}.0, ${m}.0);
-        return sampleTexture(${n}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let x=zc(n);return t?`
-    float ${o}(int row, int col, int depth, int depth2) {
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  if (enableShapeUniforms) {
+    return `
+    float ${funcName}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      ${d}
-      ${h}
-      ${g}
+      ${stride2Str}
+      ${stride1Str}
+      ${stride0Str}
       int index = row * stride0 + col * stride1 +
           depth * stride2 + depth2;
-      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${x});
-      return sampleTexture(${n}, uv);
+      vec2 uv = uvFromFlat(${texName}TexShape[0], ${texName}TexShape[1], index + ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `:`
-    float ${o}(int row, int col, int depth, int depth2) {
+  `;
+  }
+  return `
+    float ${funcName}(int row, int col, int depth, int depth2) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${a} + col * ${i} +
-          depth * ${s} + depth2;
-      vec2 uv = uvFromFlat(${m}, ${f}, index + ${x});
-      return sampleTexture(${n}, uv);
+      int index = row * ${stride0} + col * ${stride1} +
+          depth * ${stride2} + depth2;
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index + ${offset});
+      return sampleTexture(${texName}, uv);
     }
-  `}function jtt(r){let t=r.shapeInfo.logicalShape,e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1),o=t[4],s=t[3]*o,i=t[2]*s,a=t[1]*i,{newShape:u,keptDims:l}=y.squeezeShape(t);if(u.length<t.length){let h=wd(r,u),g=["row","col","depth","depth2","depth3"];return`
-      ${yd(h)}
-      float ${n}(int row, int col, int depth, int depth2, int depth3) {
-        return ${n}(${Cd(g,l)});
+  `;
+}
+function getSampler5D(inputInfo) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const stride3 = shape[4];
+  const stride2 = shape[3] * stride3;
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2", "depth3"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo)}
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}if(r.shapeInfo.isUniform)return`
-      float ${n}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         float index = dot(
           vec4(row, col, depth, depth2),
-          vec4(${a}, ${i}, ${s}, ${o})) +
+          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +
           depth3;
-        ${bd(r)}
+        ${getUniformSampler(inputInfo)}
       }
-    `;let c=r.shapeInfo.flatOffset,p=r.shapeInfo.texShape,m=p[0],f=p[1];if(f===a&&c==null)return`
-      float ${n}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  if (texNumC === stride0 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-                         vec4(${i}, ${s}, ${o}, 1));
+                         vec4(${stride1}, ${stride2}, ${stride3}, 1));
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${f}.0, ${m}.0);
-        return sampleTexture(${e}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(f===o&&c==null)return`
-      float ${n}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  if (texNumC === stride3 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
         float texR = dot(
           vec4(row, col, depth, depth2),
-          vec4(${t[1]*t[2]*t[3]},
-               ${t[2]*t[3]}, ${t[3]}, 1));
+          vec4(${shape[1] * shape[2] * shape[3]},
+               ${shape[2] * shape[3]}, ${shape[3]}, 1));
         int texC = depth3;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${f}.0, ${m}.0);
-        return sampleTexture(${e}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let d=zc(e);return`
-    float ${n}(int row, int col, int depth, int depth2, int depth3) {
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  return `
+    float ${funcName}(int row, int col, int depth, int depth2, int depth3) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${a} + col * ${i} + depth * ${s} +
-          depth2 * ${o} + depth3 + ${d};
-      vec2 uv = uvFromFlat(${m}, ${f}, index);
-      return sampleTexture(${e}, uv);
+      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +
+          depth2 * ${stride3} + depth3 + ${offset};
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+      return sampleTexture(${texName}, uv);
     }
-  `}function Xtt(r){let t=r.shapeInfo.logicalShape,e=r.name,n="get"+e.charAt(0).toUpperCase()+e.slice(1),{newShape:o,keptDims:s}=y.squeezeShape(t);if(o.length<t.length){let g=wd(r,o),x=["row","col","depth","depth2","depth3","depth4"];return`
-      ${yd(g)}
-      float ${n}(int row, int col, int depth,
+  `;
+}
+function getSampler6D(inputInfo) {
+  const shape = inputInfo.shapeInfo.logicalShape;
+  const texName = inputInfo.name;
+  const funcName = "get" + texName.charAt(0).toUpperCase() + texName.slice(1);
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  if (newShape.length < shape.length) {
+    const newInputInfo = squeezeInputInfo(inputInfo, newShape);
+    const params = ["row", "col", "depth", "depth2", "depth3", "depth4"];
+    return `
+      ${getSamplerFromInInfo(newInputInfo)}
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
-        return ${n}(${Cd(x,s)});
+        return ${funcName}(${getSqueezedParams(params, keptDims)});
       }
-    `}let i=t[5],a=t[4]*i,u=t[3]*a,l=t[2]*u,c=t[1]*l;if(r.shapeInfo.isUniform)return`
-      float ${n}(int row, int col, int depth,
+    `;
+  }
+  const stride4 = shape[5];
+  const stride3 = shape[4] * stride4;
+  const stride2 = shape[3] * stride3;
+  const stride1 = shape[2] * stride2;
+  const stride0 = shape[1] * stride1;
+  if (inputInfo.shapeInfo.isUniform) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
         int index = round(dot(
           vec4(row, col, depth, depth2),
-          vec4(${c}, ${l}, ${u}, ${a})) +
+          vec4(${stride0}, ${stride1}, ${stride2}, ${stride3})) +
           dot(
             vec2(depth3, depth4),
-            vec2(${i}, 1)));
-        ${bd(r)}
+            vec2(${stride4}, 1)));
+        ${getUniformSampler(inputInfo)}
       }
-    `;let p=r.shapeInfo.flatOffset,m=r.shapeInfo.texShape,f=m[0],d=m[1];if(d===c&&p==null)return`
-      float ${n}(int row, int col, int depth,
+    `;
+  }
+  const flatOffset = inputInfo.shapeInfo.flatOffset;
+  const texShape = inputInfo.shapeInfo.texShape;
+  const texNumR = texShape[0];
+  const texNumC = texShape[1];
+  if (texNumC === stride0 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         int texR = row;
         float texC = dot(vec4(col, depth, depth2, depth3),
-          vec4(${l}, ${u}, ${a}, ${i})) +
+          vec4(${stride1}, ${stride2}, ${stride3}, ${stride4})) +
                float(depth4);
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                   vec2(${d}.0, ${f}.0);
-        return sampleTexture(${e}, uv);
+                   vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;if(d===i&&p==null)return`
-      float ${n}(int row, int col, int depth,
+    `;
+  }
+  if (texNumC === stride4 && flatOffset == null) {
+    return `
+      float ${funcName}(int row, int col, int depth,
                     int depth2, int depth3, int depth4) {
         float texR = dot(vec4(row, col, depth, depth2),
-          vec4(${t[1]*t[2]*t[3]*t[4]},
-               ${t[2]*t[3]*t[4]},
-               ${t[3]*t[4]},
-               ${t[4]})) + float(depth3);
+          vec4(${shape[1] * shape[2] * shape[3] * shape[4]},
+               ${shape[2] * shape[3] * shape[4]},
+               ${shape[3] * shape[4]},
+               ${shape[4]})) + float(depth3);
         int texC = depth4;
         vec2 uv = (vec2(texC, texR) + halfCR) /
-                  vec2(${d}.0, ${f}.0);
-        return sampleTexture(${e}, uv);
+                  vec2(${texNumC}.0, ${texNumR}.0);
+        return sampleTexture(${texName}, uv);
       }
-    `;let h=zc(e);return`
-    float ${n}(int row, int col, int depth,
+    `;
+  }
+  const offset = getFlatOffsetUniformName(texName);
+  return `
+    float ${funcName}(int row, int col, int depth,
                   int depth2, int depth3, int depth4) {
       // Explicitly use integer operations as dot() only works on floats.
-      int index = row * ${c} + col * ${l} + depth * ${u} +
-          depth2 * ${a} + depth3 * ${i} + depth4 + ${h};
-      vec2 uv = uvFromFlat(${f}, ${d}, index);
-      return sampleTexture(${e}, uv);
+      int index = row * ${stride0} + col * ${stride1} + depth * ${stride2} +
+          depth2 * ${stride3} + depth3 * ${stride4} + depth4 + ${offset};
+      vec2 uv = uvFromFlat(${texNumR}, ${texNumC}, index);
+      return sampleTexture(${texName}, uv);
     }
-  `}function bd(r){let t=r.name,e=y.sizeFromShape(r.shapeInfo.logicalShape);return e<2?`return ${t};`:`
-    for (int i = 0; i < ${e}; i++) {
+  `;
+}
+function getUniformSampler(inputInfo) {
+  const texName = inputInfo.name;
+  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);
+  if (inSize < 2) {
+    return `return ${texName};`;
+  }
+  return `
+    for (int i = 0; i < ${inSize}; i++) {
       if (i == index) {
-        return ${t}[i];
+        return ${texName}[i];
       }
     }
-  `}function Ytt(r,t){let e=r.name,n=e.charAt(0).toUpperCase()+e.slice(1),o="get"+n+"AtOutCoords",s=r.shapeInfo.logicalShape.length,i=t.logicalShape.length,a=rL(r.shapeInfo.logicalShape,t.logicalShape),u=zt(i),l=i-s,c,p=["x","y","z","w","u","v"];s===0?c="":i<2&&a.length>=1?c="coords = 0;":c=a.map(b=>`coords.${p[b+l]} = 0;`).join(`
-`);let m="";i<2&&s>0?m="coords":m=r.shapeInfo.logicalShape.map((b,w)=>`coords.${p[w+l]}`).join(", ");let f="return outputValue;",h=y.sizeFromShape(r.shapeInfo.logicalShape)===1,x=y.sizeFromShape(t.logicalShape)===1;if(s===1&&!h&&!x)f=`
+  `;
+}
+function getPackedSamplerAtOutputCoords(inputInfo, outShapeInfo) {
+  const texName = inputInfo.name;
+  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);
+  const funcName = "get" + texFuncSnippet + "AtOutCoords";
+  const inRank = inputInfo.shapeInfo.logicalShape.length;
+  const outRank = outShapeInfo.logicalShape.length;
+  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);
+  const type = getCoordsDataType(outRank);
+  const rankDiff = outRank - inRank;
+  let coordsSnippet;
+  const fields = ["x", "y", "z", "w", "u", "v"];
+  if (inRank === 0) {
+    coordsSnippet = "";
+  } else if (outRank < 2 && broadcastDims.length >= 1) {
+    coordsSnippet = "coords = 0;";
+  } else {
+    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join("\n");
+  }
+  let unpackedCoordsSnippet = "";
+  if (outRank < 2 && inRank > 0) {
+    unpackedCoordsSnippet = "coords";
+  } else {
+    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(", ");
+  }
+  let output = `return outputValue;`;
+  const inSize = util_exports.sizeFromShape(inputInfo.shapeInfo.logicalShape);
+  const isInputScalar = inSize === 1;
+  const outSize = util_exports.sizeFromShape(outShapeInfo.logicalShape);
+  const isOutputScalar = outSize === 1;
+  if (inRank === 1 && !isInputScalar && !isOutputScalar) {
+    output = `
       return vec4(outputValue.xy, outputValue.xy);
-    `;else if(h&&!x)i===1?f=`
+    `;
+  } else if (isInputScalar && !isOutputScalar) {
+    if (outRank === 1) {
+      output = `
         return vec4(outputValue.x, outputValue.x, 0., 0.);
-      `:f=`
+      `;
+    } else {
+      output = `
         return vec4(outputValue.x);
-      `;else if(a.length){let b=s-2,w=s-1;a.indexOf(b)>-1&&a.indexOf(w)>-1?f="return vec4(outputValue.x);":a.indexOf(b)>-1?f="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":a.indexOf(w)>-1&&(f="return vec4(outputValue.xx, outputValue.zz);")}return`
-    vec4 ${o}() {
-      ${u} coords = getOutputCoords();
-      ${c}
-      vec4 outputValue = get${n}(${m});
-      ${f}
+      `;
     }
-  `}function Ztt(r,t){let e=r.name,n=e.charAt(0).toUpperCase()+e.slice(1),o="get"+n+"AtOutCoords",s=t.texShape,i=r.shapeInfo.texShape,a=r.shapeInfo.logicalShape.length,u=t.logicalShape.length;if(!r.shapeInfo.isUniform&&a===u&&r.shapeInfo.flatOffset==null&&y.arraysEqual(i,s))return`
-      float ${o}() {
-        return sampleTexture(${e}, resultUV);
+  } else if (broadcastDims.length) {
+    const rows = inRank - 2;
+    const cols = inRank - 1;
+    if (broadcastDims.indexOf(rows) > -1 && broadcastDims.indexOf(cols) > -1) {
+      output = `return vec4(outputValue.x);`;
+    } else if (broadcastDims.indexOf(rows) > -1) {
+      output = `return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);`;
+    } else if (broadcastDims.indexOf(cols) > -1) {
+      output = `return vec4(outputValue.xx, outputValue.zz);`;
+    }
+  }
+  return `
+    vec4 ${funcName}() {
+      ${type} coords = getOutputCoords();
+      ${coordsSnippet}
+      vec4 outputValue = get${texFuncSnippet}(${unpackedCoordsSnippet});
+      ${output}
+    }
+  `;
+}
+function getSamplerAtOutputCoords(inputInfo, outShapeInfo) {
+  const texName = inputInfo.name;
+  const texFuncSnippet = texName.charAt(0).toUpperCase() + texName.slice(1);
+  const funcName = "get" + texFuncSnippet + "AtOutCoords";
+  const outTexShape = outShapeInfo.texShape;
+  const inTexShape = inputInfo.shapeInfo.texShape;
+  const inRank = inputInfo.shapeInfo.logicalShape.length;
+  const outRank = outShapeInfo.logicalShape.length;
+  if (!inputInfo.shapeInfo.isUniform && inRank === outRank && inputInfo.shapeInfo.flatOffset == null && util_exports.arraysEqual(inTexShape, outTexShape)) {
+    return `
+      float ${funcName}() {
+        return sampleTexture(${texName}, resultUV);
       }
-    `;let l=zt(u),c=rL(r.shapeInfo.logicalShape,t.logicalShape),p=u-a,m,f=["x","y","z","w","u","v"];a===0?m="":u<2&&c.length>=1?m="coords = 0;":m=c.map(h=>`coords.${f[h+p]} = 0;`).join(`
-`);let d="";return u<2&&a>0?d="coords":d=r.shapeInfo.logicalShape.map((h,g)=>`coords.${f[g+p]}`).join(", "),`
-    float ${o}() {
-      ${l} coords = getOutputCoords();
-      ${m}
-      return get${n}(${d});
+    `;
+  }
+  const type = getCoordsDataType(outRank);
+  const broadcastDims = getBroadcastDims2(inputInfo.shapeInfo.logicalShape, outShapeInfo.logicalShape);
+  const rankDiff = outRank - inRank;
+  let coordsSnippet;
+  const fields = ["x", "y", "z", "w", "u", "v"];
+  if (inRank === 0) {
+    coordsSnippet = "";
+  } else if (outRank < 2 && broadcastDims.length >= 1) {
+    coordsSnippet = "coords = 0;";
+  } else {
+    coordsSnippet = broadcastDims.map((d) => `coords.${fields[d + rankDiff]} = 0;`).join("\n");
+  }
+  let unpackedCoordsSnippet = "";
+  if (outRank < 2 && inRank > 0) {
+    unpackedCoordsSnippet = "coords";
+  } else {
+    unpackedCoordsSnippet = inputInfo.shapeInfo.logicalShape.map((s, i) => `coords.${fields[i + rankDiff]}`).join(", ");
+  }
+  return `
+    float ${funcName}() {
+      ${type} coords = getOutputCoords();
+      ${coordsSnippet}
+      return get${texFuncSnippet}(${unpackedCoordsSnippet});
     }
-  `}function zt(r){if(r<=1)return"int";if(r===2)return"ivec2";if(r===3)return"ivec3";if(r===4)return"ivec4";if(r===5)return"ivec5";if(r===6)return"ivec6";throw Error(`GPU for rank ${r} is not yet supported`)}function Tw(r,t,e){let{newShape:n,keptDims:o}=y.squeezeShape(t),s=t.length,i=r&&s===3&&t[0]===1,a=i?t.slice(1):n,u=!r&&s>1&&!y.arraysEqual(t,e)&&n.length<s||i;return{useSqueezeShape:u,uniformShape:u?a:t,keptDims:o}}function wd(r,t){let e=JSON.parse(JSON.stringify(r));return e.shapeInfo.logicalShape=t,e}function Cd(r,t){return t.map(e=>r[e]).join(", ")}function aL(r,t,e,n){let o=e.map((c,p)=>{let m={logicalShape:c.shape,texShape:c.isUniform?null:c.texData.texShape,isUniform:c.isUniform,isPacked:c.isUniform?!1:c.texData.isPacked,flatOffset:null};return c.texData!=null&&c.texData.slice!=null&&c.texData.slice.flatOffset>0&&(m.flatOffset=c.texData.slice.flatOffset),{name:t.variableNames[p],shapeInfo:m}}),s=o.map(c=>c.shapeInfo),i={logicalShape:n.shape,texShape:n.texData.texShape,isUniform:!1,isPacked:n.texData.isPacked,flatOffset:null},a=nL(o,i,t),u=NT(r.gl,a),l=r.createProgram(u);return z().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:u,source:a,webGLProgram:l,inShapeInfos:s,outShapeInfo:i,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:Object.assign({program:t,fragmentShader:u,source:a,webGLProgram:l,inShapeInfos:s,outShapeInfo:i},WT(r,t,l))}function WT(r,t,e){let n={},o={},s={},i=[],a,u,l,c=null,p=null;p=r.getUniformLocation(e,"NAN",!1),z().getNumber("WEBGL_VERSION")===1&&(c=r.getUniformLocation(e,"INFINITY",!1));let m=!1;for(let f=0;f<t.variableNames.length;f++){let d=t.variableNames[f];n[d]=r.getUniformLocation(e,d,m),n[`offset${d}`]=r.getUniformLocation(e,`offset${d}`,m),t.enableShapeUniforms&&(o[`${d}Shape`]=r.getUniformLocation(e,`${d}Shape`,m),s[`${d}TexShape`]=r.getUniformLocation(e,`${d}TexShape`,m))}return t.enableShapeUniforms&&(a=r.getUniformLocation(e,"outShape",m),l=r.getUniformLocation(e,"outShapeStrides",m),u=r.getUniformLocation(e,"outTexShape",m)),t.customUniforms&&t.customUniforms.forEach((f,d)=>{i[d]=r.getUniformLocation(e,f.name,m)}),{uniformLocations:n,customUniformLocations:i,infLoc:c,nanLoc:p,inShapesLocations:o,inTexShapesLocations:s,outShapeLocation:a,outShapeStridesLocation:l,outTexShapeLocation:u}}function iL(r,t){if(r.length!==t.length)throw Error(`Binary was compiled with ${r.length} inputs, but was executed with ${t.length} inputs`);r.forEach((e,n)=>{let o=e.logicalShape,s=t[n],i=s.shape;if(!y.arraysEqual(o,i))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${o} and ${i} must match`);if(e.isUniform&&s.isUniform)return;let a=e.texShape,u=s.isUniform?null:s.texData.texShape;if(!y.arraysEqual(a,u))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${a} and ${u} must match`)})}function lL(r,t,e,n,o){t.program.enableShapeUniforms||(iL(t.inShapeInfos,e),iL([t.outShapeInfo],[n]));let s=n.texData.texture,i=n.texData.texShape;n.texData.isPacked?r.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):r.setOutputMatrixTexture(s.texture,i[0],i[1]),r.setProgram(t.webGLProgram),z().getNumber("WEBGL_VERSION")===1&&t.infLoc!==null&&r.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&r.gl.uniform1f(t.nanLoc,NaN),e.forEach((u,l)=>{let c=t.program.variableNames[l],p=t.uniformLocations[c],m=t.uniformLocations[`offset${c}`],f=t.inShapesLocations[`${c}Shape`],d=t.inTexShapesLocations[`${c}TexShape`];if(f){let{uniformShape:h}=Tw(t.program.packedInputs,u.shape,u.texData.texShape);switch(h.length){case 1:r.gl.uniform1iv(f,new Int32Array(h));break;case 2:r.gl.uniform2iv(f,new Int32Array(h));break;case 3:r.gl.uniform3iv(f,new Int32Array(h));break;case 4:r.gl.uniform4iv(f,new Int32Array(h));break;default:break}}if(d&&r.gl.uniform2i(d,u.texData.texShape[0],u.texData.texShape[1]),p!=null){if(u.isUniform){if(y.sizeFromShape(u.shape)<2)r.gl.uniform1f(p,u.uniformValues[0]);else{let h=u.uniformValues;h instanceof Float32Array||(h=new Float32Array(h)),r.gl.uniform1fv(p,h)}return}u.texData.slice!=null&&m!=null&&r.gl.uniform1i(m,u.texData.slice.flatOffset),r.setInputMatrixTexture(u.texData.texture.texture,p,l)}});let a=t.outShapeLocation;if(a)switch(n.shape.length){case 1:r.gl.uniform1iv(a,new Int32Array(n.shape));break;case 2:r.gl.uniform2iv(a,new Int32Array(n.shape));break;case 3:r.gl.uniform3iv(a,new Int32Array(n.shape));break;case 4:r.gl.uniform4iv(a,new Int32Array(n.shape));break;default:break}if(t.outShapeStridesLocation){let u=y.computeStrides(n.shape);switch(n.shape.length){case 2:r.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(u));break;case 3:r.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(u));break;case 4:r.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(u));break;default:break}}t.outTexShapeLocation&&r.gl.uniform2i(t.outTexShapeLocation,n.texData.texShape[0],n.texData.texShape[1]),t.program.customUniforms&&o&&t.program.customUniforms.forEach((u,l)=>{let c=t.customUniformLocations[l],p=o[l];if(u.type==="float")r.gl.uniform1fv(c,p);else if(u.type==="vec2")r.gl.uniform2fv(c,p);else if(u.type==="vec3")r.gl.uniform3fv(c,p);else if(u.type==="vec4")r.gl.uniform4fv(c,p);else if(u.type==="int")r.gl.uniform1iv(c,p);else if(u.type==="ivec2")r.gl.uniform2iv(c,p);else if(u.type==="ivec3")r.gl.uniform3iv(c,p);else if(u.type==="ivec4")r.gl.uniform4iv(c,p);else throw Error(`uniform type ${u.type} is not supported yet.`)}),r.executeProgram()}function uL(r,t,e){let n="";t.concat(e).forEach(i=>{let a=i.texData!=null&&i.texData.slice!=null&&i.texData.slice.flatOffset>0;if(r.enableShapeUniforms&&!i.isUniform){let u=i.texData.texShape,{useSqueezeShape:l,uniformShape:c,keptDims:p}=Tw(r.packedInputs,i.shape,u),m="",f="",d="";if(c.length===1&&r.packedInputs){let N=[Math.ceil(u[0]/2),Math.ceil(u[1]/2)];m=`${N[0]>1}_${N[1]>1}`}else if(c.length===2&&!r.packedInputs)f=`${c[0]>1}_${c[1]>1}`;else if(c.length>2&&!r.packedInputs){let N=y.computeStrides(c);d=`${N[0]===u[1]}_${N[N.length-1]===u[1]}`}let h=i.shape.length,g=c.length===2&&y.arraysEqual(i.shape,u),x=y.sizeFromShape(i.shape)===1,b=v.getBroadcastDims(i.shape,e.shape),w=!r.packedInputs&&h===e.shape.length&&y.arraysEqual(u,e.texData.texShape),C=r.packedInputs||c.length>2?"":`${u[0]>1}_${u[1]>1}`;n+=`${h}_${w}_${l?p:""}_${c.length}_${x}_${b}_${g}_${m}_${f}_${d}_${C}_${a}`}else{let u=i.isUniform?"uniform":i.texData.texShape;n+=`${i.shape}_${u}_${a}`}});let o=r.userCode,s=r.constructor.name;return s+="_"+n+"_"+o+`${z().getNumber("WEBGL_VERSION")}`,s}function we(r){return z().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&r<=4}var kw=class{constructor(t){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=ku.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let e=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`
+  `;
+}
+function getCoordsDataType(rank) {
+  if (rank <= 1) {
+    return "int";
+  } else if (rank === 2) {
+    return "ivec2";
+  } else if (rank === 3) {
+    return "ivec3";
+  } else if (rank === 4) {
+    return "ivec4";
+  } else if (rank === 5) {
+    return "ivec5";
+  } else if (rank === 6) {
+    return "ivec6";
+  } else {
+    throw Error(`GPU for rank ${rank} is not yet supported`);
+  }
+}
+function getUniformInfoFromShape(isPacked, shape, texShape) {
+  const { newShape, keptDims } = util_exports.squeezeShape(shape);
+  const rank = shape.length;
+  const useSqueezePackedShape = isPacked && rank === 3 && shape[0] === 1;
+  const squeezeShape2 = useSqueezePackedShape ? shape.slice(1) : newShape;
+  const useSqueezeShape = !isPacked && rank > 1 && !util_exports.arraysEqual(shape, texShape) && newShape.length < rank || useSqueezePackedShape;
+  const uniformShape = useSqueezeShape ? squeezeShape2 : shape;
+  return { useSqueezeShape, uniformShape, keptDims };
+}
+function squeezeInputInfo(inInfo, squeezedShape) {
+  const newInputInfo = JSON.parse(JSON.stringify(inInfo));
+  newInputInfo.shapeInfo.logicalShape = squeezedShape;
+  return newInputInfo;
+}
+function getSqueezedParams(params, keptDims) {
+  return keptDims.map((d) => params[d]).join(", ");
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_math.js
+function compileProgram(gpgpu, program, inputs, output) {
+  const inputInfos = inputs.map((input2, i) => {
+    const shapeInfo = {
+      logicalShape: input2.shape,
+      texShape: input2.isUniform ? null : input2.texData.texShape,
+      isUniform: input2.isUniform,
+      isPacked: input2.isUniform ? false : input2.texData.isPacked,
+      flatOffset: null
+    };
+    if (input2.texData != null && input2.texData.slice != null && input2.texData.slice.flatOffset > 0) {
+      shapeInfo.flatOffset = input2.texData.slice.flatOffset;
+    }
+    return { name: program.variableNames[i], shapeInfo };
+  });
+  const inShapeInfos = inputInfos.map((x) => x.shapeInfo);
+  const outShapeInfo = {
+    logicalShape: output.shape,
+    texShape: output.texData.texShape,
+    isUniform: false,
+    isPacked: output.texData.isPacked,
+    flatOffset: null
+  };
+  const source = makeShader(inputInfos, outShapeInfo, program);
+  const fragmentShader = createFragmentShader(gpgpu.gl, source);
+  const webGLProgram = gpgpu.createProgram(fragmentShader);
+  if (!env().get("ENGINE_COMPILE_ONLY")) {
+    return Object.assign({
+      program,
+      fragmentShader,
+      source,
+      webGLProgram,
+      inShapeInfos,
+      outShapeInfo
+    }, getUniformLocations(gpgpu, program, webGLProgram));
+  } else {
+    return {
+      program,
+      fragmentShader,
+      source,
+      webGLProgram,
+      inShapeInfos,
+      outShapeInfo,
+      uniformLocations: null,
+      customUniformLocations: null,
+      infLoc: null,
+      nanLoc: null,
+      inShapesLocations: null,
+      inTexShapesLocations: null,
+      outShapeLocation: null,
+      outShapeStridesLocation: null,
+      outTexShapeLocation: null
+    };
+  }
+}
+function getUniformLocations(gpgpu, program, webGLProgram) {
+  const uniformLocations = {};
+  const inShapesLocations = {};
+  const inTexShapesLocations = {};
+  const customUniformLocations = [];
+  let outShapeLocation;
+  let outTexShapeLocation;
+  let outShapeStridesLocation;
+  let infLoc = null;
+  let nanLoc = null;
+  nanLoc = gpgpu.getUniformLocation(webGLProgram, "NAN", false);
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    infLoc = gpgpu.getUniformLocation(webGLProgram, "INFINITY", false);
+  }
+  const shouldThrow = false;
+  for (let i = 0; i < program.variableNames.length; i++) {
+    const varName = program.variableNames[i];
+    uniformLocations[varName] = gpgpu.getUniformLocation(webGLProgram, varName, shouldThrow);
+    uniformLocations[`offset${varName}`] = gpgpu.getUniformLocation(webGLProgram, `offset${varName}`, shouldThrow);
+    if (program.enableShapeUniforms) {
+      inShapesLocations[`${varName}Shape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}Shape`, shouldThrow);
+      inTexShapesLocations[`${varName}TexShape`] = gpgpu.getUniformLocation(webGLProgram, `${varName}TexShape`, shouldThrow);
+    }
+  }
+  if (program.enableShapeUniforms) {
+    outShapeLocation = gpgpu.getUniformLocation(webGLProgram, "outShape", shouldThrow);
+    outShapeStridesLocation = gpgpu.getUniformLocation(webGLProgram, "outShapeStrides", shouldThrow);
+    outTexShapeLocation = gpgpu.getUniformLocation(webGLProgram, "outTexShape", shouldThrow);
+  }
+  if (program.customUniforms) {
+    program.customUniforms.forEach((d, i) => {
+      customUniformLocations[i] = gpgpu.getUniformLocation(webGLProgram, d.name, shouldThrow);
+    });
+  }
+  return {
+    uniformLocations,
+    customUniformLocations,
+    infLoc,
+    nanLoc,
+    inShapesLocations,
+    inTexShapesLocations,
+    outShapeLocation,
+    outShapeStridesLocation,
+    outTexShapeLocation
+  };
+}
+function validateBinaryAndProgram(shapeInfos, inputs) {
+  if (shapeInfos.length !== inputs.length) {
+    throw Error(`Binary was compiled with ${shapeInfos.length} inputs, but was executed with ${inputs.length} inputs`);
+  }
+  shapeInfos.forEach((s, i) => {
+    const shapeA = s.logicalShape;
+    const input2 = inputs[i];
+    const shapeB = input2.shape;
+    if (!util_exports.arraysEqual(shapeA, shapeB)) {
+      throw Error(`Binary was compiled with different shapes than the current args. Shapes ${shapeA} and ${shapeB} must match`);
+    }
+    if (s.isUniform && input2.isUniform) {
+      return;
+    }
+    const texShapeA = s.texShape;
+    const texShapeB = input2.isUniform ? null : input2.texData.texShape;
+    if (!util_exports.arraysEqual(texShapeA, texShapeB)) {
+      throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${texShapeA} and ${texShapeB} must match`);
+    }
+  });
+}
+function runProgram(gpgpu, binary, inputs, output, customUniformValues) {
+  if (!binary.program.enableShapeUniforms) {
+    validateBinaryAndProgram(binary.inShapeInfos, inputs);
+    validateBinaryAndProgram([binary.outShapeInfo], [output]);
+  }
+  const outTex = output.texData.texture;
+  const outTexShape = output.texData.texShape;
+  if (output.texData.isPacked) {
+    gpgpu.setOutputPackedMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);
+  } else {
+    gpgpu.setOutputMatrixTexture(outTex.texture, outTexShape[0], outTexShape[1]);
+  }
+  gpgpu.setProgram(binary.webGLProgram);
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    if (binary.infLoc !== null) {
+      gpgpu.gl.uniform1f(binary.infLoc, Infinity);
+    }
+  }
+  if (binary.nanLoc !== null) {
+    gpgpu.gl.uniform1f(binary.nanLoc, NaN);
+  }
+  inputs.forEach((input2, i) => {
+    const varName = binary.program.variableNames[i];
+    const varLoc = binary.uniformLocations[varName];
+    const varOffsetLoc = binary.uniformLocations[`offset${varName}`];
+    const varShapeLoc = binary.inShapesLocations[`${varName}Shape`];
+    const varTexShapeLoc = binary.inTexShapesLocations[`${varName}TexShape`];
+    if (varShapeLoc) {
+      const { uniformShape } = getUniformInfoFromShape(binary.program.packedInputs, input2.shape, input2.texData.texShape);
+      switch (uniformShape.length) {
+        case 1:
+          gpgpu.gl.uniform1iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 2:
+          gpgpu.gl.uniform2iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 3:
+          gpgpu.gl.uniform3iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        case 4:
+          gpgpu.gl.uniform4iv(varShapeLoc, new Int32Array(uniformShape));
+          break;
+        default:
+          break;
+      }
+    }
+    if (varTexShapeLoc) {
+      gpgpu.gl.uniform2i(varTexShapeLoc, input2.texData.texShape[0], input2.texData.texShape[1]);
+    }
+    if (varLoc == null) {
+      return;
+    }
+    if (input2.isUniform) {
+      if (util_exports.sizeFromShape(input2.shape) < 2) {
+        gpgpu.gl.uniform1f(varLoc, input2.uniformValues[0]);
+      } else {
+        let vals = input2.uniformValues;
+        if (!(vals instanceof Float32Array)) {
+          vals = new Float32Array(vals);
+        }
+        gpgpu.gl.uniform1fv(varLoc, vals);
+      }
+      return;
+    }
+    if (input2.texData.slice != null && varOffsetLoc != null) {
+      gpgpu.gl.uniform1i(varOffsetLoc, input2.texData.slice.flatOffset);
+    }
+    gpgpu.setInputMatrixTexture(input2.texData.texture.texture, varLoc, i);
+  });
+  const outShapeLoc = binary.outShapeLocation;
+  if (outShapeLoc) {
+    switch (output.shape.length) {
+      case 1:
+        gpgpu.gl.uniform1iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 2:
+        gpgpu.gl.uniform2iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 3:
+        gpgpu.gl.uniform3iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      case 4:
+        gpgpu.gl.uniform4iv(outShapeLoc, new Int32Array(output.shape));
+        break;
+      default:
+        break;
+    }
+  }
+  if (binary.outShapeStridesLocation) {
+    const strides = util_exports.computeStrides(output.shape);
+    switch (output.shape.length) {
+      case 2:
+        gpgpu.gl.uniform1iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      case 3:
+        gpgpu.gl.uniform2iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      case 4:
+        gpgpu.gl.uniform3iv(binary.outShapeStridesLocation, new Int32Array(strides));
+        break;
+      default:
+        break;
+    }
+  }
+  if (binary.outTexShapeLocation) {
+    gpgpu.gl.uniform2i(binary.outTexShapeLocation, output.texData.texShape[0], output.texData.texShape[1]);
+  }
+  if (binary.program.customUniforms && customUniformValues) {
+    binary.program.customUniforms.forEach((d, i) => {
+      const customLoc = binary.customUniformLocations[i];
+      const customValue = customUniformValues[i];
+      if (d.type === "float") {
+        gpgpu.gl.uniform1fv(customLoc, customValue);
+      } else if (d.type === "vec2") {
+        gpgpu.gl.uniform2fv(customLoc, customValue);
+      } else if (d.type === "vec3") {
+        gpgpu.gl.uniform3fv(customLoc, customValue);
+      } else if (d.type === "vec4") {
+        gpgpu.gl.uniform4fv(customLoc, customValue);
+      } else if (d.type === "int") {
+        gpgpu.gl.uniform1iv(customLoc, customValue);
+      } else if (d.type === "ivec2") {
+        gpgpu.gl.uniform2iv(customLoc, customValue);
+      } else if (d.type === "ivec3") {
+        gpgpu.gl.uniform3iv(customLoc, customValue);
+      } else if (d.type === "ivec4") {
+        gpgpu.gl.uniform4iv(customLoc, customValue);
+      } else {
+        throw Error(`uniform type ${d.type} is not supported yet.`);
+      }
+    });
+  }
+  gpgpu.executeProgram();
+}
+function makeShaderKey(program, inputs, output) {
+  let keyInputs = "";
+  inputs.concat(output).forEach((x) => {
+    const hasOffset = x.texData != null && x.texData.slice != null && x.texData.slice.flatOffset > 0;
+    if (program.enableShapeUniforms && !x.isUniform) {
+      const xTexShape = x.texData.texShape;
+      const { useSqueezeShape, uniformShape, keptDims } = getUniformInfoFromShape(program.packedInputs, x.shape, xTexShape);
+      let rank1 = "", rank2 = "", rank34 = "";
+      if (uniformShape.length === 1 && program.packedInputs) {
+        const packedTexShape = [Math.ceil(xTexShape[0] / 2), Math.ceil(xTexShape[1] / 2)];
+        rank1 = `${packedTexShape[0] > 1}_${packedTexShape[1] > 1}`;
+      } else if (uniformShape.length === 2 && !program.packedInputs) {
+        rank2 = `${uniformShape[0] > 1}_${uniformShape[1] > 1}`;
+      } else if (uniformShape.length > 2 && !program.packedInputs) {
+        const strides = util_exports.computeStrides(uniformShape);
+        rank34 = `${strides[0] === xTexShape[1]}_${strides[strides.length - 1] === xTexShape[1]}`;
+      }
+      const xRank = x.shape.length;
+      const isLogicalShapTexShapeEqual = uniformShape.length === 2 && util_exports.arraysEqual(x.shape, xTexShape);
+      const isScalar = util_exports.sizeFromShape(x.shape) === 1;
+      const broadcastDims = backend_util_exports.getBroadcastDims(x.shape, output.shape);
+      const isInOutTexShapeEqual = !program.packedInputs && xRank === output.shape.length && util_exports.arraysEqual(xTexShape, output.texData.texShape);
+      const isTexShapeGreaterThanOne = program.packedInputs || uniformShape.length > 2 ? "" : `${xTexShape[0] > 1}_${xTexShape[1] > 1}`;
+      keyInputs += `${xRank}_${isInOutTexShapeEqual}_${useSqueezeShape ? keptDims : ""}_${uniformShape.length}_${isScalar}_${broadcastDims}_${isLogicalShapTexShapeEqual}_${rank1}_${rank2}_${rank34}_${isTexShapeGreaterThanOne}_${hasOffset}`;
+    } else {
+      const texShape = x.isUniform ? "uniform" : x.texData.texShape;
+      keyInputs += `${x.shape}_${texShape}_${hasOffset}`;
+    }
+  });
+  const keyUserCode = program.userCode;
+  let key = program.constructor.name;
+  key += "_" + keyInputs + "_" + keyUserCode + `${env().getNumber("WEBGL_VERSION")}`;
+  return key;
+}
+function useShapeUniforms(rank) {
+  return env().getBool("WEBGL_USE_SHAPES_UNIFORMS") && rank <= 4;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/decode_matrix_gpu.js
+var DecodeMatrixProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.outPackingScheme = PackingScheme.DENSE;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Mc(["r","c","d"],t):ti(["r","c","d"],t)}
+        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], outputShape) : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], outputShape)}
         return ivec3(r, c, d);
       }
 
@@ -1002,11 +51597,26 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           result[i] = getA(rc.x, rc.y, rc.z);
         }
 
-        ${e.output} = result;
+        ${glsl.output} = result;
       }
-    `}};var Ew=class{constructor(t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=ku.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let e=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/decode_matrix_packed_gpu.js
+var DecodeMatrixPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outPackingScheme = PackingScheme.DENSE;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       ivec3 outCoordsFromFlatIndex(int index) {
-        ${this.enableShapeUniforms?Mc(["r","c","d"],t):ti(["r","c","d"],t)}
+        ${this.enableShapeUniforms ? getOutputLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], outputShape) : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], outputShape)}
         return ivec3(r, c, d);
       }
 
@@ -1022,52 +51632,125 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));
         }
 
-        ${e.output} = result;
+        ${glsl.output} = result;
       }
-    `}};var _w=class{constructor(t){this.variableNames=["A"],this.outTexUsage=jr.DOWNLOAD;let e=Ge();this.outputShape=t,this.userCode=`
-      ${Nw}
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_float_gpu.js
+var EncodeFloatProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.outTexUsage = TextureUsage.DOWNLOAD;
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.userCode = `
+      ${ENCODE_FLOAT_SNIPPET}
 
       void main() {
         float x = getAAtOutCoords();
-        ${e.output} = encode_float(x);
+        ${glsl.output} = encode_float(x);
       }
-    `}};var Aw=class{constructor(t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=jr.DOWNLOAD;let e=Ge();this.outputShape=t,this.userCode=`
-      ${Nw}
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_float_packed_gpu.js
+var EncodeFloatPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = false;
+    this.outTexUsage = TextureUsage.DOWNLOAD;
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.userCode = `
+      ${ENCODE_FLOAT_SNIPPET}
 
       void main() {
         ivec3 coords = getOutputCoords();
         float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));
-        ${e.output} = encode_float(x);
+        ${glsl.output} = encode_float(x);
       }
-    `}};var tet={R:0,G:1,B:2,A:3},Jh=class{constructor(t,e=!1,n="RGBA"){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let o=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let s="result";e&&(s="floor(result * 255. + 0.5)");let i="";for(let a=0;a<n.length;a++){let u=n[a];i+=`
-          if(offset == ${a}) {
-            result = values[${tet[u]}];
-          }`}this.userCode=`
-      ${this.enableShapeUniforms?gd():hd(t)}
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_matrix_gpu.js
+var CHANNEL_CHAR_TO_INDEX_MAP = {
+  "R": 0,
+  "G": 1,
+  "B": 2,
+  "A": 3
+};
+var EncodeMatrixProgram = class {
+  constructor(outputShape, inputIsUnsignedByte = false, usedChannels = "RGBA") {
+    this.variableNames = ["A"];
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let output = `result`;
+    if (inputIsUnsignedByte) {
+      output = `floor(result * 255. + 0.5)`;
+    }
+    let mainLoop = "";
+    for (let usedChannelIndex = 0; usedChannelIndex < usedChannels.length; usedChannelIndex++) {
+      const curChannel = usedChannels[usedChannelIndex];
+      mainLoop += `
+          if(offset == ${usedChannelIndex}) {
+            result = values[${CHANNEL_CHAR_TO_INDEX_MAP[curChannel]}];
+          }`;
+    }
+    this.userCode = `
+      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
       void main() {
         ivec3 coords = getOutputCoords();
         int flatIndex = getFlatIndex(coords);
         float result = 0.;
-        int offset = imod(flatIndex, ${n.length});
+        int offset = imod(flatIndex, ${usedChannels.length});
 
-        flatIndex = idiv(flatIndex, ${n.length}, 1.);
+        flatIndex = idiv(flatIndex, ${usedChannels.length}, 1.);
 
         int r = flatIndex / texShape[1];
         if (r < texShape[0]) {
           int c = imod(flatIndex, texShape[1]);
           vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
-          vec4 values = ${o.texture2D}(A, uv);
-          ${i}
+          vec4 values = ${glsl.texture2D}(A, uv);
+          ${mainLoop}
         }
-        ${o.output} = vec4(${s}, 0., 0., 0.);
+        ${glsl.output} = vec4(${output}, 0., 0., 0.);
       }
-    `}};var $w=class{constructor(t,e=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let n=Ge();this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let o="",s="result";e&&(s="floor(result * 255. + 0.5)");for(let i=0;i<=1;i++)for(let a=0;a<=1;a++){let u=i*2+a;o+=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/encode_matrix_packed_gpu.js
+var EncodeMatrixPackedProgram = class {
+  constructor(outputShape, inputIsUnsignedByte = false) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "texShape", type: "ivec2" }];
+    const glsl = getGlslDifferences();
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let mainLoop = "";
+    let output = "result";
+    if (inputIsUnsignedByte) {
+      output = "floor(result * 255. + 0.5)";
+    }
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        const channel = row * 2 + col;
+        mainLoop += `
           localCoords = coords;
-          if(localCoords[2] + ${a} < ${this.enableShapeUniforms?"outShape[2]":`${t[2]}`}) {
-          localCoords[2] += ${a};
-          if (localCoords[1] + ${i} < ${this.enableShapeUniforms?"outShape[1]":`${t[1]}`}) {
-            localCoords[1] += ${i};
+          if(localCoords[2] + ${col} < ${this.enableShapeUniforms ? "outShape[2]" : `${outputShape[2]}`}) {
+          localCoords[2] += ${col};
+          if (localCoords[1] + ${row} < ${this.enableShapeUniforms ? "outShape[1]" : `${outputShape[1]}`}) {
+            localCoords[1] += ${row};
 
             flatIndex = getFlatIndex(localCoords);
             offset = imod(flatIndex, 4);
@@ -1077,21 +51760,24 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             int r = flatIndex / texShape[1];
             int c = imod(flatIndex, texShape[1]);
             vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
-            values = ${n.texture2D}(A, uv);
+            values = ${glsl.texture2D}(A, uv);
 
             if (offset == 0) {
-              result[${u}] = values[0];
+              result[${channel}] = values[0];
             } else if (offset == 1) {
-              result[${u}] = values[1];
+              result[${channel}] = values[1];
             } else if (offset == 2) {
-              result[${u}] = values[2];
+              result[${channel}] = values[2];
             } else {
-              result[${u}] = values[3];
+              result[${channel}] = values[3];
             }
           }
         }
-        `}this.userCode=`
-        ${this.enableShapeUniforms?gd():hd(t)}
+        `;
+      }
+    }
+    this.userCode = `
+        ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
         void main() {
           ivec3 coords = getOutputCoords();
@@ -1102,60 +51788,751 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
           vec2 uv;
           vec4 values;
 
-          ${o}
+          ${mainLoop}
 
-          ${n.output} = ${s};
+          ${glsl.output} = ${output};
         }
-    `}};var ik={};Wt(ik,{bindVertexProgramAttributeStreams:()=>JT,createBufferFromOutputTexture:()=>ek,createFloat16MatrixTexture:()=>jT,createFloat16PackedMatrixTexture:()=>ZT,createFloat32MatrixTexture:()=>KT,createIndexBuffer:()=>qT,createPackedMatrixTexture:()=>YT,createUnsignedBytesMatrixTexture:()=>XT,createVertexBuffer:()=>HT,createVertexShader:()=>UT,downloadByteEncodedFloatMatrixFromOutputTexture:()=>nk,downloadFloat32MatrixFromBuffer:()=>rk,downloadMatrixFromPackedOutputTexture:()=>sk,downloadPackedMatrixFromBuffer:()=>ok,getInternalFormatForFloat16MatrixTexture:()=>Rw,getInternalFormatForFloat16PackedMatrixTexture:()=>Pw,getInternalFormatForFloat32MatrixTexture:()=>Dw,getInternalFormatForPackedMatrixTexture:()=>Ow,getInternalFormatForUnsignedBytesMatrixTexture:()=>Fw,uploadDenseMatrixToTexture:()=>QT,uploadPixelDataToTexture:()=>tk});function UT(r){let t=Ge(),e=`${t.version}
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_util.js
+var gpgpu_util_exports = {};
+__export(gpgpu_util_exports, {
+  bindVertexProgramAttributeStreams: () => bindVertexProgramAttributeStreams,
+  createBufferFromOutputTexture: () => createBufferFromOutputTexture,
+  createFloat16MatrixTexture: () => createFloat16MatrixTexture,
+  createFloat16PackedMatrixTexture: () => createFloat16PackedMatrixTexture,
+  createFloat32MatrixTexture: () => createFloat32MatrixTexture,
+  createIndexBuffer: () => createIndexBuffer,
+  createPackedMatrixTexture: () => createPackedMatrixTexture,
+  createUnsignedBytesMatrixTexture: () => createUnsignedBytesMatrixTexture,
+  createVertexBuffer: () => createVertexBuffer,
+  createVertexShader: () => createVertexShader2,
+  downloadByteEncodedFloatMatrixFromOutputTexture: () => downloadByteEncodedFloatMatrixFromOutputTexture,
+  downloadFloat32MatrixFromBuffer: () => downloadFloat32MatrixFromBuffer,
+  downloadMatrixFromPackedOutputTexture: () => downloadMatrixFromPackedOutputTexture,
+  downloadPackedMatrixFromBuffer: () => downloadPackedMatrixFromBuffer,
+  getInternalFormatForFloat16MatrixTexture: () => getInternalFormatForFloat16MatrixTexture,
+  getInternalFormatForFloat16PackedMatrixTexture: () => getInternalFormatForFloat16PackedMatrixTexture,
+  getInternalFormatForFloat32MatrixTexture: () => getInternalFormatForFloat32MatrixTexture,
+  getInternalFormatForPackedMatrixTexture: () => getInternalFormatForPackedMatrixTexture,
+  getInternalFormatForUnsignedBytesMatrixTexture: () => getInternalFormatForUnsignedBytesMatrixTexture,
+  uploadDenseMatrixToTexture: () => uploadDenseMatrixToTexture,
+  uploadPixelDataToTexture: () => uploadPixelDataToTexture
+});
+function createVertexShader2(gl) {
+  const glsl = getGlslDifferences();
+  const vertexShaderSource = `${glsl.version}
     precision highp float;
-    ${t.attribute} vec3 clipSpacePos;
-    ${t.attribute} vec2 uv;
-    ${t.varyingVs} vec2 resultUV;
+    ${glsl.attribute} vec3 clipSpacePos;
+    ${glsl.attribute} vec2 uv;
+    ${glsl.varyingVs} vec2 resultUV;
 
     void main() {
       gl_Position = vec4(clipSpacePos, 1);
       resultUV = uv;
-    }`;return vT(r,e)}function HT(r){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return ET(r,t)}function qT(r){let t=new Uint16Array([0,1,2,2,1,3]);return _T(r,t)}function Qh(r,t,e,n,o,s){$T(t,e);let i=AT(r),a=r.TEXTURE_2D;return yt(r,()=>r.bindTexture(a,i)),yt(r,()=>r.texParameteri(a,r.TEXTURE_WRAP_S,r.CLAMP_TO_EDGE)),yt(r,()=>r.texParameteri(a,r.TEXTURE_WRAP_T,r.CLAMP_TO_EDGE)),yt(r,()=>r.texParameteri(a,r.TEXTURE_MIN_FILTER,r.NEAREST)),yt(r,()=>r.texParameteri(a,r.TEXTURE_MAG_FILTER,r.NEAREST)),z().getNumber("WEBGL_VERSION")===1?yt(r,()=>r.texImage2D(a,0,n,t,e,0,o,s,null)):yt(r,()=>r.texStorage2D(a,1,n,t,e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null)),{texture:i,texShape:[e,t]}}function Dw(r){return r.internalFormatFloat}function KT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Dw(n),n.textureFormatFloat,r.FLOAT)}function Rw(r){return r.internalFormatHalfFloat}function jT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Rw(n),n.textureFormatFloat,n.textureTypeHalfFloat)}function Fw(r){return r.downloadTextureFormat}function XT(r,t,e,n){let[o,s]=Lc(t,e);return Qh(r,o,s,Fw(n),r.RGBA,r.UNSIGNED_BYTE)}function Ow(r){return r.internalFormatPackedFloat}function YT(r,t,e,n){let[o,s]=Xi(t,e);return Qh(r,o,s,Ow(n),r.RGBA,r.FLOAT)}function Pw(r){return r.internalFormatPackedHalfFloat}function ZT(r,t,e,n){let[o,s]=Xi(t,e);return Qh(r,o,s,Pw(n),r.RGBA,n.textureTypeHalfFloat)}function JT(r,t,e){return yt(r,()=>r.bindBuffer(r.ARRAY_BUFFER,e)),Iw(r,t,"clipSpacePos",e,3,20,0)&&Iw(r,t,"uv",e,2,20,12)}function QT(r,t,e,n,o,s){yt(r,()=>r.bindTexture(r.TEXTURE_2D,t));let i,a,u;o instanceof Uint8Array?(i=new Uint8Array(e*n*4),a=r.UNSIGNED_BYTE,u=r.RGBA):(i=new Float32Array(e*n*4),a=r.FLOAT,u=s.internalFormatPackedFloat),i.set(o),z().getNumber("WEBGL_VERSION")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,e,n,r.RGBA,a,i)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,u,e,n,0,r.RGBA,a,i)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function tk(r,t,e){yt(r,()=>r.bindTexture(r.TEXTURE_2D,t)),e.data instanceof Uint8Array?z().getNumber("WEBGL_VERSION")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,e.width,e.height,r.RGBA,r.UNSIGNED_BYTE,e.data)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,r.RGBA,e.width,e.height,0,r.RGBA,r.UNSIGNED_BYTE,e.data)):z().getNumber("WEBGL_VERSION")===2?yt(r,()=>r.texSubImage2D(r.TEXTURE_2D,0,0,0,r.RGBA,r.UNSIGNED_BYTE,e)):yt(r,()=>r.texImage2D(r.TEXTURE_2D,0,r.RGBA,r.RGBA,r.UNSIGNED_BYTE,e)),yt(r,()=>r.bindTexture(r.TEXTURE_2D,null))}function ek(r,t,e,n){let o=r.createBuffer();yt(r,()=>r.bindBuffer(r.PIXEL_PACK_BUFFER,o));let a=4*4*t*e;return yt(r,()=>r.bufferData(r.PIXEL_PACK_BUFFER,a,r.STREAM_READ)),yt(r,()=>r.readPixels(0,0,e,t,r.RGBA,r.FLOAT,0)),yt(r,()=>r.bindBuffer(r.PIXEL_PACK_BUFFER,null)),o}function rk(r,t,e){let n=r,o=new Float32Array(e);return n.bindBuffer(n.PIXEL_PACK_BUFFER,t),n.getBufferSubData(n.PIXEL_PACK_BUFFER,0,o),n.bindBuffer(n.PIXEL_PACK_BUFFER,null),o}function nk(r,t,e,n){let[o,s]=Lc(t,e),i=4,a=new Uint8Array(XP(t*e,i));return yt(r,()=>r.readPixels(0,0,o,s,n.downloadTextureFormat,r.UNSIGNED_BYTE,a)),new Float32Array(a.buffer)}function ok(r,t,e,n,o,s,i,a){let u=r,l=new Float32Array(YP(s,i));return u.bindBuffer(u.PIXEL_PACK_BUFFER,t),u.getBufferSubData(u.PIXEL_PACK_BUFFER,0,l),u.bindBuffer(u.PIXEL_PACK_BUFFER,null),l}function sk(r,t,e){let n=new Float32Array(t*e*4);return yt(r,()=>r.readPixels(0,0,e,t,r.RGBA,r.FLOAT,n)),n}var Bc=class{constructor(t){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let e=z().getNumber("WEBGL_VERSION");t!=null?(this.gl=t,wT(e,t)):this.gl=Gn(e);let n="WEBGL_color_buffer_float",o="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),z().getNumber("WEBGL_VERSION")===1){let s="OES_texture_float",i="OES_texture_half_float";if(this.textureFloatExtension=pd(this.gl,s),Wn(this.gl,i))this.textureHalfFloatExtension=pd(this.gl,i);else if(z().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),Wn(this.gl,o))this.colorBufferHalfFloatExtension=pd(this.gl,o);else if(z().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",Wn(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else if(Wn(this.gl,o))this.colorBufferHalfFloatExtension=this.gl.getExtension(o);else throw new Error("GL context does not support color renderable floats");this.vertexBuffer=HT(this.gl),this.indexBuffer=qT(this.gl),this.framebuffer=DT(this.gl),this.textureConfig=Xh(this.gl,this.textureHalfFloatExtension)}get debug(){return z().getBool("DEBUG")}dispose(){if(this.disposed)return;this.program!=null&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),this.outputTexture!=null&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let t=this.gl;yt(t,()=>t.finish()),yt(t,()=>t.bindFramebuffer(t.FRAMEBUFFER,null)),yt(t,()=>t.deleteFramebuffer(this.framebuffer)),yt(t,()=>t.bindBuffer(t.ARRAY_BUFFER,null)),yt(t,()=>t.bindBuffer(t.ELEMENT_ARRAY_BUFFER,null)),yt(t,()=>t.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(t,e){return this.throwIfDisposed(),KT(this.gl,t,e,this.textureConfig)}createFloat16MatrixTexture(t,e){return this.throwIfDisposed(),jT(this.gl,t,e,this.textureConfig)}createUnsignedBytesMatrixTexture(t,e){return this.throwIfDisposed(),XT(this.gl,t,e,this.textureConfig)}uploadPixelDataToTexture(t,e){this.throwIfDisposed(),tk(this.gl,t,e)}uploadDenseMatrixToTexture(t,e,n,o){this.throwIfDisposed(),QT(this.gl,t,e,n,o,this.textureConfig)}createFloat16PackedMatrixTexture(t,e){return this.throwIfDisposed(),ZT(this.gl,t,e,this.textureConfig)}createPackedMatrixTexture(t,e){return this.throwIfDisposed(),YT(this.gl,t,e,this.textureConfig)}deleteMatrixTexture(t){this.throwIfDisposed(),this.outputTexture===t&&(Sw(this.gl,this.framebuffer),this.outputTexture=null),yt(this.gl,()=>this.gl.deleteTexture(t))}downloadByteEncodedFloatMatrixFromOutputTexture(t,e,n){return this.downloadMatrixDriver(t,()=>nk(this.gl,e,n,this.textureConfig))}downloadPackedMatrixFromBuffer(t,e,n,o,s,i){return ok(this.gl,t,e,n,o,s,i,this.textureConfig)}downloadFloat32MatrixFromBuffer(t,e){return rk(this.gl,t,e)}createBufferFromTexture(t,e,n){this.bindTextureToFrameBuffer(t);let o=ek(this.gl,e,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),o}createAndWaitForFence(){let t=this.createFence(this.gl);return this.pollFence(t)}createFence(t){let e,n;if(z().getBool("WEBGL_FENCE_API_ENABLED")){let o=t,s=o.fenceSync(o.SYNC_GPU_COMMANDS_COMPLETE,0);t.flush(),n=()=>{let i=o.clientWaitSync(s,0,0);return i===o.ALREADY_SIGNALED||i===o.CONDITION_SATISFIED},e=s}else z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(e=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(e,z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:e,isFencePassed:n}}downloadMatrixFromPackedTexture(t,e,n){return this.downloadMatrixDriver(t,()=>sk(this.gl,e,n))}createProgram(t){this.throwIfDisposed();let e=this.gl;this.vertexShader==null&&(this.vertexShader=UT(e));let n=TT(e);return yt(e,()=>e.attachShader(n,this.vertexShader)),yt(e,()=>e.attachShader(n,t)),kT(e,n),this.debug&&Yh(e,n),this.vertexAttrsAreBound||(this.setProgram(n),this.vertexAttrsAreBound=JT(e,this.program,this.vertexBuffer)),n}deleteProgram(t){this.throwIfDisposed(),t===this.program&&(this.program=null),t!=null&&yt(this.gl,()=>this.gl.deleteProgram(t))}setProgram(t){this.throwIfDisposed(),this.program=t,this.program!=null&&this.debug&&Yh(this.gl,this.program),yt(this.gl,()=>this.gl.useProgram(t))}getUniformLocation(t,e,n=!0){return this.throwIfDisposed(),n?RT(this.gl,t,e):FT(this.gl,t,e)}getAttributeLocation(t,e){return this.throwIfDisposed(),yt(this.gl,()=>this.gl.getAttribLocation(t,e))}getUniformLocationNoThrow(t,e){return this.throwIfDisposed(),this.gl.getUniformLocation(t,e)}setInputMatrixTexture(t,e,n){this.throwIfDisposed(),this.throwIfNoProgram(),OT(this.gl,t,e,n)}setOutputMatrixTexture(t,e,n){this.setOutputMatrixTextureDriver(t,n,e)}setOutputPackedMatrixTexture(t,e,n){this.throwIfDisposed();let[o,s]=Xi(e,n);this.setOutputMatrixTextureDriver(t,o,s)}setOutputMatrixWriteRegion(t,e,n,o){this.setOutputMatrixWriteRegionDriver(n,t,o,e)}setOutputPackedMatrixWriteRegion(t,e,n,o){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){this.program!=null&&Yh(this.gl,this.program),md(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let t=this.gl;this.debug&&this.debugValidate(),yt(t,()=>t.drawElements(t.TRIANGLES,6,t.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),yt(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=pd(this.gl,z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let n=this.gl,o=this.getQueryTimerExtensionWebGL2(),s=n.createQuery();return n.beginQuery(o.TIME_ELAPSED_EXT,s),s}let t=this.getQueryTimerExtensionWebGL1(),e=t.createQueryEXT();return t.beginQueryEXT(t.TIME_ELAPSED_EXT,e),e}endQuery(){if(z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let e=this.gl,n=this.getQueryTimerExtensionWebGL2();e.endQuery(n.TIME_ELAPSED_EXT);return}let t=this.getQueryTimerExtensionWebGL1();t.endQueryEXT(t.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(t){return await y.repeatedTry(()=>this.disposed||this.isQueryAvailable(t,z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(t,z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(t,e){if(e===0)return null;if(e===2){let n=this.gl;return n.getQueryParameter(t,n.QUERY_RESULT)/1e6}else{let n=this.getQueryTimerExtensionWebGL1();return n.getQueryObjectEXT(t,n.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(t,e){if(e===0)return!0;if(e===2){let n=this.gl,o=this.getQueryTimerExtensionWebGL2(),s=n.getQueryParameter(t,n.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(o.GPU_DISJOINT_EXT)),s&&!this.disjoint}else{let n=this.getQueryTimerExtensionWebGL1(),o=n.getQueryObjectEXT(t,n.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),o&&!this.disjoint}}pollFence(t){return new Promise(e=>{this.addItemToPoll(()=>t.isFencePassed(),()=>e())})}pollItems(){let t=eet(this.itemsToPoll.map(e=>e.isDoneFn));for(let e=0;e<=t;++e){let{resolveFn:n}=this.itemsToPoll[e];n()}this.itemsToPoll=this.itemsToPoll.slice(t+1)}addItemToPoll(t,e){if(this.itemsToPoll.push({isDoneFn:t,resolveFn:e}),this.itemsToPoll.length>1)return;let n;"setTimeoutCustom"in z().platform&&(n=z().platform.setTimeoutCustom.bind(z().platform)),y.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0),()=>0,null,n)}bindTextureToFrameBuffer(t){this.throwIfDisposed(),Zh(this.gl,t,this.framebuffer),this.debug&&md(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(Zh(this.gl,this.outputTexture,this.framebuffer),this.debug&&md(this.gl)):Sw(this.gl,this.framebuffer)}downloadMatrixDriver(t,e){this.bindTextureToFrameBuffer(t);let n=e();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(t,e,n){this.throwIfDisposed();let o=this.gl;Zh(o,t,this.framebuffer),this.debug&&md(o),this.outputTexture=t,yt(o,()=>o.viewport(0,0,e,n)),yt(o,()=>o.scissor(0,0,e,n))}setOutputMatrixWriteRegionDriver(t,e,n,o){this.throwIfDisposed(),yt(this.gl,()=>this.gl.scissor(t,e,n,o))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(this.program==null)throw new Error("No GPU program is currently set.")}};function eet(r){let t=0;for(;t<r.length&&r[t]();++t);return t-1}var{addImpl:cL,bincountImpl:Lw,bincountReduceImpl:pL,castImpl:mL,ceilImpl:fL,concatImpl:dL,equalImpl:hL,expImpl:gL,expm1Impl:xL,floorImpl:yL,gatherNdImpl:bL,gatherV2Impl:wL,greaterImpl:CL,greaterEqualImpl:IL,lessImpl:SL,lessEqualImpl:vL,linSpaceImpl:NL,logImpl:TL,maxImpl:kL,maximumImpl:EL,minimumImpl:_L,multiplyImpl:AL,negImpl:$L,notEqualImpl:DL,prodImpl:RL,raggedGatherImpl:FL,raggedRangeImpl:OL,raggedTensorToTensorImpl:PL,rangeImpl:LL,rsqrtImpl:ML,scatterImpl:zL,sigmoidImpl:BL,simpleAbsImpl:Mw,sliceImpl:VL,sparseFillEmptyRowsImpl:GL,sparseReshapeImpl:WL,sparseSegmentReductionImpl:zw,sqrtImpl:UL,stridedSliceImpl:HL,stringNGramsImpl:qL,stringSplitImpl:KL,stringToHashBucketFastImpl:jL,subImpl:XL,tileImpl:YL,topKImpl:ZL,transposeImpl:Vc,uniqueImpl:JL}=mw;function ak(r,t){return["x","y","z","w","u","v"].slice(0,t).map(e=>`${r}.${e}`)}function Qe(r,t){return t===1?[r]:ak(r,t)}function QL(r,t){if(r===1)return"rc";let e="";for(let n=0;n<r;n++)e+=t[n],n<r-1&&(e+=",");return e}var Bw=class{constructor(t){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=t,this.rank=t.length,this.enableShapeUniforms=we(this.outputShape.length),this.rank===0)this.userCode=`
+    }`;
+  return createVertexShader(gl, vertexShaderSource);
+}
+function createVertexBuffer(gl) {
+  const vertexArray = new Float32Array([-1, 1, 0, 0, 1, -1, -1, 0, 0, 0, 1, 1, 0, 1, 1, 1, -1, 0, 1, 0]);
+  return createStaticVertexBuffer(gl, vertexArray);
+}
+function createIndexBuffer(gl) {
+  const triangleVertexIndices = new Uint16Array([0, 1, 2, 2, 1, 3]);
+  return createStaticIndexBuffer(gl, triangleVertexIndices);
+}
+function createAndConfigureTexture(gl, width, height, internalFormat, textureFormat, textureType) {
+  validateTextureSize(width, height);
+  const texture = createTexture(gl);
+  const tex2d = gl.TEXTURE_2D;
+  callAndCheck(gl, () => gl.bindTexture(tex2d, texture));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_S, gl.CLAMP_TO_EDGE));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_WRAP_T, gl.CLAMP_TO_EDGE));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MIN_FILTER, gl.NEAREST));
+  callAndCheck(gl, () => gl.texParameteri(tex2d, gl.TEXTURE_MAG_FILTER, gl.NEAREST));
+  if (env().getNumber("WEBGL_VERSION") === 1) {
+    callAndCheck(gl, () => gl.texImage2D(tex2d, 0, internalFormat, width, height, 0, textureFormat, textureType, null));
+  } else {
+    callAndCheck(gl, () => gl.texStorage2D(tex2d, 1, internalFormat, width, height));
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+  return { texture, texShape: [height, width] };
+}
+function getInternalFormatForFloat32MatrixTexture(textureConfig) {
+  return textureConfig.internalFormatFloat;
+}
+function createFloat32MatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat32MatrixTexture(textureConfig), textureConfig.textureFormatFloat, gl.FLOAT);
+}
+function getInternalFormatForFloat16MatrixTexture(textureConfig) {
+  return textureConfig.internalFormatHalfFloat;
+}
+function createFloat16MatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16MatrixTexture(textureConfig), textureConfig.textureFormatFloat, textureConfig.textureTypeHalfFloat);
+}
+function getInternalFormatForUnsignedBytesMatrixTexture(textureConfig) {
+  return textureConfig.downloadTextureFormat;
+}
+function createUnsignedBytesMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForUnsignedBytesMatrixTexture(textureConfig), gl.RGBA, gl.UNSIGNED_BYTE);
+}
+function getInternalFormatForPackedMatrixTexture(textureConfig) {
+  return textureConfig.internalFormatPackedFloat;
+}
+function createPackedMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForPackedMatrixTexture(textureConfig), gl.RGBA, gl.FLOAT);
+}
+function getInternalFormatForFloat16PackedMatrixTexture(textureConfig) {
+  return textureConfig.internalFormatPackedHalfFloat;
+}
+function createFloat16PackedMatrixTexture(gl, rows, columns, textureConfig) {
+  const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+  return createAndConfigureTexture(gl, width, height, getInternalFormatForFloat16PackedMatrixTexture(textureConfig), gl.RGBA, textureConfig.textureTypeHalfFloat);
+}
+function bindVertexProgramAttributeStreams(gl, program, vertexBuffer) {
+  const posOffset = 0;
+  const uvOffset = 3 * 4;
+  const stride = 3 * 4 + 2 * 4;
+  callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, vertexBuffer));
+  const success = bindVertexBufferToProgramAttribute(gl, program, "clipSpacePos", vertexBuffer, 3, stride, posOffset);
+  return success && bindVertexBufferToProgramAttribute(gl, program, "uv", vertexBuffer, 2, stride, uvOffset);
+}
+function uploadDenseMatrixToTexture(gl, texture, width, height, data, textureConfig) {
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+  let dataForUpload, texelDataType, internalFormat;
+  if (data instanceof Uint8Array) {
+    dataForUpload = new Uint8Array(width * height * 4);
+    texelDataType = gl.UNSIGNED_BYTE;
+    internalFormat = gl.RGBA;
+  } else {
+    dataForUpload = new Float32Array(width * height * 4);
+    texelDataType = gl.FLOAT;
+    internalFormat = textureConfig.internalFormatPackedFloat;
+  }
+  dataForUpload.set(data);
+  if (env().getNumber("WEBGL_VERSION") === 2) {
+    callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, width, height, gl.RGBA, texelDataType, dataForUpload));
+  } else {
+    callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, internalFormat, width, height, 0, gl.RGBA, texelDataType, dataForUpload));
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function uploadPixelDataToTexture(gl, texture, pixels) {
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, texture));
+  if (pixels.data instanceof Uint8Array) {
+    if (env().getNumber("WEBGL_VERSION") === 2) {
+      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, pixels.width, pixels.height, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));
+    } else {
+      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, pixels.width, pixels.height, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels.data));
+    }
+  } else {
+    if (env().getNumber("WEBGL_VERSION") === 2) {
+      callAndCheck(gl, () => gl.texSubImage2D(gl.TEXTURE_2D, 0, 0, 0, gl.RGBA, gl.UNSIGNED_BYTE, pixels));
+    } else {
+      callAndCheck(gl, () => gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gl.RGBA, gl.UNSIGNED_BYTE, pixels));
+    }
+  }
+  callAndCheck(gl, () => gl.bindTexture(gl.TEXTURE_2D, null));
+}
+function createBufferFromOutputTexture(gl2, rows, columns, textureConfig) {
+  const buffer2 = gl2.createBuffer();
+  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2));
+  const bytesPerFloat = 4;
+  const valuesPerTexel = 4;
+  const bufferSizeBytes = bytesPerFloat * valuesPerTexel * rows * columns;
+  callAndCheck(gl2, () => gl2.bufferData(gl2.PIXEL_PACK_BUFFER, bufferSizeBytes, gl2.STREAM_READ));
+  callAndCheck(gl2, () => gl2.readPixels(0, 0, columns, rows, gl2.RGBA, gl2.FLOAT, 0));
+  callAndCheck(gl2, () => gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null));
+  return buffer2;
+}
+function downloadFloat32MatrixFromBuffer(gl, buffer2, size) {
+  const gl2 = gl;
+  const downloadTarget = new Float32Array(size);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);
+  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);
+  return downloadTarget;
+}
+function downloadByteEncodedFloatMatrixFromOutputTexture(gl, rows, columns, textureConfig) {
+  const [w, h] = getUnpackedMatrixTextureShapeWidthHeight(rows, columns);
+  const numChannels = 4;
+  const downloadTarget = new Uint8Array(getUnpackedArraySizeFromMatrixSize(rows * columns, numChannels));
+  callAndCheck(gl, () => gl.readPixels(0, 0, w, h, textureConfig.downloadTextureFormat, gl.UNSIGNED_BYTE, downloadTarget));
+  return new Float32Array(downloadTarget.buffer);
+}
+function downloadPackedMatrixFromBuffer(gl, buffer2, batch, rows, cols, physicalRows, physicalCols, textureConfig) {
+  const gl2 = gl;
+  const downloadTarget = new Float32Array(getPackedRGBAArraySizeFromMatrixShape(physicalRows, physicalCols));
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, buffer2);
+  gl2.getBufferSubData(gl2.PIXEL_PACK_BUFFER, 0, downloadTarget);
+  gl2.bindBuffer(gl2.PIXEL_PACK_BUFFER, null);
+  return downloadTarget;
+}
+function downloadMatrixFromPackedOutputTexture(gl, physicalRows, physicalCols) {
+  const packedRGBA = new Float32Array(physicalRows * physicalCols * 4);
+  callAndCheck(gl, () => gl.readPixels(0, 0, physicalCols, physicalRows, gl.RGBA, gl.FLOAT, packedRGBA));
+  return packedRGBA;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/gpgpu_context.js
+var GPGPUContext = class {
+  constructor(gl) {
+    this.outputTexture = null;
+    this.program = null;
+    this.disposed = false;
+    this.vertexAttrsAreBound = false;
+    this.itemsToPoll = [];
+    const glVersion = env().getNumber("WEBGL_VERSION");
+    if (gl != null) {
+      this.gl = gl;
+      setWebGLContext(glVersion, gl);
+    } else {
+      this.gl = getWebGLContext(glVersion);
+    }
+    let COLOR_BUFFER_FLOAT = "WEBGL_color_buffer_float";
+    const COLOR_BUFFER_HALF_FLOAT = "EXT_color_buffer_half_float";
+    this.parallelCompilationExtension = this.gl.getExtension("KHR_parallel_shader_compile");
+    if (env().getNumber("WEBGL_VERSION") === 1) {
+      const TEXTURE_FLOAT = "OES_texture_float";
+      const TEXTURE_HALF_FLOAT = "OES_texture_half_float";
+      this.textureFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_FLOAT);
+      if (hasExtension(this.gl, TEXTURE_HALF_FLOAT)) {
+        this.textureHalfFloatExtension = getExtensionOrThrow(this.gl, TEXTURE_HALF_FLOAT);
+      } else if (env().get("WEBGL_FORCE_F16_TEXTURES")) {
+        throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");
+      }
+      this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);
+      if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {
+        this.colorBufferHalfFloatExtension = getExtensionOrThrow(this.gl, COLOR_BUFFER_HALF_FLOAT);
+      } else if (env().get("WEBGL_FORCE_F16_TEXTURES")) {
+        throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");
+      }
+    } else {
+      COLOR_BUFFER_FLOAT = "EXT_color_buffer_float";
+      if (hasExtension(this.gl, COLOR_BUFFER_FLOAT)) {
+        this.colorBufferFloatExtension = this.gl.getExtension(COLOR_BUFFER_FLOAT);
+      } else if (hasExtension(this.gl, COLOR_BUFFER_HALF_FLOAT)) {
+        this.colorBufferHalfFloatExtension = this.gl.getExtension(COLOR_BUFFER_HALF_FLOAT);
+      } else {
+        throw new Error("GL context does not support color renderable floats");
+      }
+    }
+    this.vertexBuffer = createVertexBuffer(this.gl);
+    this.indexBuffer = createIndexBuffer(this.gl);
+    this.framebuffer = createFramebuffer(this.gl);
+    this.textureConfig = getTextureConfig(this.gl, this.textureHalfFloatExtension);
+  }
+  get debug() {
+    return env().getBool("DEBUG");
+  }
+  dispose() {
+    if (this.disposed) {
+      return;
+    }
+    if (this.program != null) {
+      console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing.");
+    }
+    if (this.outputTexture != null) {
+      console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");
+    }
+    const gl = this.gl;
+    callAndCheck(gl, () => gl.finish());
+    callAndCheck(gl, () => gl.bindFramebuffer(gl.FRAMEBUFFER, null));
+    callAndCheck(gl, () => gl.deleteFramebuffer(this.framebuffer));
+    callAndCheck(gl, () => gl.bindBuffer(gl.ARRAY_BUFFER, null));
+    callAndCheck(gl, () => gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, null));
+    callAndCheck(gl, () => gl.deleteBuffer(this.indexBuffer));
+    this.disposed = true;
+  }
+  createFloat32MatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat32MatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createFloat16MatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat16MatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createUnsignedBytesMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createUnsignedBytesMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  uploadPixelDataToTexture(texture, pixels) {
+    this.throwIfDisposed();
+    uploadPixelDataToTexture(this.gl, texture, pixels);
+  }
+  uploadDenseMatrixToTexture(texture, width, height, data) {
+    this.throwIfDisposed();
+    uploadDenseMatrixToTexture(this.gl, texture, width, height, data, this.textureConfig);
+  }
+  createFloat16PackedMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createFloat16PackedMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  createPackedMatrixTexture(rows, columns) {
+    this.throwIfDisposed();
+    return createPackedMatrixTexture(this.gl, rows, columns, this.textureConfig);
+  }
+  deleteMatrixTexture(texture) {
+    this.throwIfDisposed();
+    if (this.outputTexture === texture) {
+      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);
+      this.outputTexture = null;
+    }
+    callAndCheck(this.gl, () => this.gl.deleteTexture(texture));
+  }
+  downloadByteEncodedFloatMatrixFromOutputTexture(texture, rows, columns) {
+    return this.downloadMatrixDriver(texture, () => downloadByteEncodedFloatMatrixFromOutputTexture(this.gl, rows, columns, this.textureConfig));
+  }
+  downloadPackedMatrixFromBuffer(buffer2, batch, rows, columns, physicalRows, physicalCols) {
+    return downloadPackedMatrixFromBuffer(this.gl, buffer2, batch, rows, columns, physicalRows, physicalCols, this.textureConfig);
+  }
+  downloadFloat32MatrixFromBuffer(buffer2, size) {
+    return downloadFloat32MatrixFromBuffer(this.gl, buffer2, size);
+  }
+  createBufferFromTexture(texture, rows, columns) {
+    this.bindTextureToFrameBuffer(texture);
+    const result = createBufferFromOutputTexture(this.gl, rows, columns, this.textureConfig);
+    this.unbindTextureToFrameBuffer();
+    return result;
+  }
+  createAndWaitForFence() {
+    const fenceContext = this.createFence(this.gl);
+    return this.pollFence(fenceContext);
+  }
+  createFence(gl) {
+    let query;
+    let isFencePassed;
+    if (env().getBool("WEBGL_FENCE_API_ENABLED")) {
+      const gl2 = gl;
+      const sync = gl2.fenceSync(gl2.SYNC_GPU_COMMANDS_COMPLETE, 0);
+      gl.flush();
+      isFencePassed = () => {
+        const status = gl2.clientWaitSync(sync, 0, 0);
+        return status === gl2.ALREADY_SIGNALED || status === gl2.CONDITION_SATISFIED;
+      };
+      query = sync;
+    } else if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") > 0) {
+      query = this.beginQuery();
+      this.endQuery();
+      isFencePassed = () => this.isQueryAvailable(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"));
+    } else {
+      isFencePassed = () => true;
+    }
+    return { query, isFencePassed };
+  }
+  downloadMatrixFromPackedTexture(texture, physicalRows, physicalCols) {
+    return this.downloadMatrixDriver(texture, () => downloadMatrixFromPackedOutputTexture(this.gl, physicalRows, physicalCols));
+  }
+  createProgram(fragmentShader) {
+    this.throwIfDisposed();
+    const gl = this.gl;
+    if (this.vertexShader == null) {
+      this.vertexShader = createVertexShader2(gl);
+    }
+    const program = createProgram(gl);
+    callAndCheck(gl, () => gl.attachShader(program, this.vertexShader));
+    callAndCheck(gl, () => gl.attachShader(program, fragmentShader));
+    linkProgram(gl, program);
+    if (this.debug) {
+      validateProgram(gl, program);
+    }
+    if (!this.vertexAttrsAreBound) {
+      this.setProgram(program);
+      this.vertexAttrsAreBound = bindVertexProgramAttributeStreams(gl, this.program, this.vertexBuffer);
+    }
+    return program;
+  }
+  deleteProgram(program) {
+    this.throwIfDisposed();
+    if (program === this.program) {
+      this.program = null;
+    }
+    if (program != null) {
+      callAndCheck(this.gl, () => this.gl.deleteProgram(program));
+    }
+  }
+  setProgram(program) {
+    this.throwIfDisposed();
+    this.program = program;
+    if (this.program != null && this.debug) {
+      validateProgram(this.gl, this.program);
+    }
+    callAndCheck(this.gl, () => this.gl.useProgram(program));
+  }
+  getUniformLocation(program, uniformName, shouldThrow = true) {
+    this.throwIfDisposed();
+    if (shouldThrow) {
+      return getProgramUniformLocationOrThrow(this.gl, program, uniformName);
+    } else {
+      return getProgramUniformLocation(this.gl, program, uniformName);
+    }
+  }
+  getAttributeLocation(program, attribute) {
+    this.throwIfDisposed();
+    return callAndCheck(this.gl, () => this.gl.getAttribLocation(program, attribute));
+  }
+  getUniformLocationNoThrow(program, uniformName) {
+    this.throwIfDisposed();
+    return this.gl.getUniformLocation(program, uniformName);
+  }
+  setInputMatrixTexture(inputMatrixTexture, uniformLocation, textureUnit) {
+    this.throwIfDisposed();
+    this.throwIfNoProgram();
+    bindTextureToProgramUniformSampler(this.gl, inputMatrixTexture, uniformLocation, textureUnit);
+  }
+  setOutputMatrixTexture(outputMatrixTexture, rows, columns) {
+    this.setOutputMatrixTextureDriver(outputMatrixTexture, columns, rows);
+  }
+  setOutputPackedMatrixTexture(outputPackedMatrixTexture, rows, columns) {
+    this.throwIfDisposed();
+    const [width, height] = getPackedMatrixTextureShapeWidthHeight(rows, columns);
+    this.setOutputMatrixTextureDriver(outputPackedMatrixTexture, width, height);
+  }
+  setOutputMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {
+    this.setOutputMatrixWriteRegionDriver(startColumn, startRow, numColumns, numRows);
+  }
+  setOutputPackedMatrixWriteRegion(startRow, numRows, startColumn, numColumns) {
+    throw new Error("setOutputPackedMatrixWriteRegion not implemented.");
+  }
+  debugValidate() {
+    if (this.program != null) {
+      validateProgram(this.gl, this.program);
+    }
+    validateFramebuffer(this.gl);
+  }
+  executeProgram() {
+    this.throwIfDisposed();
+    this.throwIfNoProgram();
+    const gl = this.gl;
+    if (this.debug) {
+      this.debugValidate();
+    }
+    callAndCheck(gl, () => gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 0));
+  }
+  blockUntilAllProgramsCompleted() {
+    this.throwIfDisposed();
+    callAndCheck(this.gl, () => this.gl.finish());
+  }
+  getQueryTimerExtension() {
+    if (this.disjointQueryTimerExtension == null) {
+      this.disjointQueryTimerExtension = getExtensionOrThrow(this.gl, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2 ? "EXT_disjoint_timer_query_webgl2" : "EXT_disjoint_timer_query");
+    }
+    return this.disjointQueryTimerExtension;
+  }
+  getQueryTimerExtensionWebGL2() {
+    return this.getQueryTimerExtension();
+  }
+  getQueryTimerExtensionWebGL1() {
+    return this.getQueryTimerExtension();
+  }
+  beginQuery() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2) {
+      const gl2 = this.gl;
+      const ext2 = this.getQueryTimerExtensionWebGL2();
+      const query2 = gl2.createQuery();
+      gl2.beginQuery(ext2.TIME_ELAPSED_EXT, query2);
+      return query2;
+    }
+    const ext = this.getQueryTimerExtensionWebGL1();
+    const query = ext.createQueryEXT();
+    ext.beginQueryEXT(ext.TIME_ELAPSED_EXT, query);
+    return query;
+  }
+  endQuery() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION") === 2) {
+      const gl2 = this.gl;
+      const ext2 = this.getQueryTimerExtensionWebGL2();
+      gl2.endQuery(ext2.TIME_ELAPSED_EXT);
+      return;
+    }
+    const ext = this.getQueryTimerExtensionWebGL1();
+    ext.endQueryEXT(ext.TIME_ELAPSED_EXT);
+  }
+  async waitForQueryAndGetTime(query) {
+    await util_exports.repeatedTry(() => this.disposed || this.isQueryAvailable(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")));
+    return this.getQueryTime(query, env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"));
+  }
+  getQueryTime(query, queryTimerVersion) {
+    if (queryTimerVersion === 0) {
+      return null;
+    }
+    if (queryTimerVersion === 2) {
+      const gl2 = this.gl;
+      const timeElapsedNanos = gl2.getQueryParameter(query, gl2.QUERY_RESULT);
+      return timeElapsedNanos / 1e6;
+    } else {
+      const ext = this.getQueryTimerExtensionWebGL1();
+      const timeElapsedNanos = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_EXT);
+      return timeElapsedNanos / 1e6;
+    }
+  }
+  isQueryAvailable(query, queryTimerVersion) {
+    if (queryTimerVersion === 0) {
+      return true;
+    }
+    if (queryTimerVersion === 2) {
+      const gl2 = this.gl;
+      const ext = this.getQueryTimerExtensionWebGL2();
+      const available = gl2.getQueryParameter(query, gl2.QUERY_RESULT_AVAILABLE);
+      if (this.disjoint == null) {
+        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);
+      }
+      return available && !this.disjoint;
+    } else {
+      const ext = this.getQueryTimerExtensionWebGL1();
+      const available = ext.getQueryObjectEXT(query, ext.QUERY_RESULT_AVAILABLE_EXT);
+      if (this.disjoint == null) {
+        this.disjoint = this.gl.getParameter(ext.GPU_DISJOINT_EXT);
+      }
+      return available && !this.disjoint;
+    }
+  }
+  pollFence(fenceContext) {
+    return new Promise((resolve) => {
+      this.addItemToPoll(() => fenceContext.isFencePassed(), () => resolve());
+    });
+  }
+  pollItems() {
+    const index = linearSearchLastTrue(this.itemsToPoll.map((x) => x.isDoneFn));
+    for (let i = 0; i <= index; ++i) {
+      const { resolveFn } = this.itemsToPoll[i];
+      resolveFn();
+    }
+    this.itemsToPoll = this.itemsToPoll.slice(index + 1);
+  }
+  addItemToPoll(isDoneFn, resolveFn) {
+    this.itemsToPoll.push({ isDoneFn, resolveFn });
+    if (this.itemsToPoll.length > 1) {
+      return;
+    }
+    let scheduleFn = void 0;
+    if ("setTimeoutCustom" in env().platform) {
+      scheduleFn = env().platform.setTimeoutCustom.bind(env().platform);
+    }
+    util_exports.repeatedTry(() => {
+      this.pollItems();
+      return this.itemsToPoll.length === 0;
+    }, () => 0, null, scheduleFn);
+  }
+  bindTextureToFrameBuffer(texture) {
+    this.throwIfDisposed();
+    bindColorTextureToFramebuffer(this.gl, texture, this.framebuffer);
+    if (this.debug) {
+      validateFramebuffer(this.gl);
+    }
+  }
+  unbindTextureToFrameBuffer() {
+    if (this.outputTexture != null) {
+      bindColorTextureToFramebuffer(this.gl, this.outputTexture, this.framebuffer);
+      if (this.debug) {
+        validateFramebuffer(this.gl);
+      }
+    } else {
+      unbindColorTextureFromFramebuffer(this.gl, this.framebuffer);
+    }
+  }
+  downloadMatrixDriver(texture, downloadAndDecode) {
+    this.bindTextureToFrameBuffer(texture);
+    const result = downloadAndDecode();
+    this.unbindTextureToFrameBuffer();
+    return result;
+  }
+  setOutputMatrixTextureDriver(outputMatrixTextureMaybePacked, width, height) {
+    this.throwIfDisposed();
+    const gl = this.gl;
+    bindColorTextureToFramebuffer(gl, outputMatrixTextureMaybePacked, this.framebuffer);
+    if (this.debug) {
+      validateFramebuffer(gl);
+    }
+    this.outputTexture = outputMatrixTextureMaybePacked;
+    callAndCheck(gl, () => gl.viewport(0, 0, width, height));
+    callAndCheck(gl, () => gl.scissor(0, 0, width, height));
+  }
+  setOutputMatrixWriteRegionDriver(x, y, width, height) {
+    this.throwIfDisposed();
+    callAndCheck(this.gl, () => this.gl.scissor(x, y, width, height));
+  }
+  throwIfDisposed() {
+    if (this.disposed) {
+      throw new Error("Attempted to use disposed GPGPUContext.");
+    }
+  }
+  throwIfNoProgram() {
+    if (this.program == null) {
+      throw new Error("No GPU program is currently set.");
+    }
+  }
+};
+function linearSearchLastTrue(arr) {
+  let i = 0;
+  for (; i < arr.length; ++i) {
+    const isDone = arr[i]();
+    if (!isDone) {
+      break;
+    }
+  }
+  return i - 1;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/shared.js
+var { addImpl: addImplCPU, bincountImpl: bincountImplCPU, bincountReduceImpl: bincountReduceImplCPU, castImpl: castImplCPU, ceilImpl: ceilImplCPU, concatImpl: concatImplCPU, equalImpl: equalImplCPU, expImpl: expImplCPU, expm1Impl: expm1ImplCPU, floorImpl: floorImplCPU, gatherNdImpl: gatherNdImplCPU, gatherV2Impl: gatherV2ImplCPU, greaterImpl: greaterImplCPU, greaterEqualImpl: greaterEqualImplCPU, lessImpl: lessImplCPU, lessEqualImpl: lessEqualImplCPU, linSpaceImpl: linSpaceImplCPU, logImpl: logImplCPU, maxImpl: maxImplCPU, maximumImpl: maximumImplCPU, minimumImpl: minimumImplCPU, multiplyImpl: multiplyImplCPU, negImpl: negImplCPU, notEqualImpl: notEqualImplCPU, prodImpl: prodImplCPU, raggedGatherImpl: raggedGatherImplCPU, raggedRangeImpl: raggedRangeImplCPU, raggedTensorToTensorImpl: raggedTensorToTensorImplCPU, rangeImpl: rangeImplCPU, rsqrtImpl: rsqrtImplCPU, scatterImpl: scatterImplCPU, sigmoidImpl: sigmoidImplCPU, simpleAbsImpl: simpleAbsImplCPU, sliceImpl: sliceImplCPU, sparseFillEmptyRowsImpl: sparseFillEmptyRowsImplCPU, sparseReshapeImpl: sparseReshapeImplCPU, sparseSegmentReductionImpl: sparseSegmentReductionImplCPU, sqrtImpl: sqrtImplCPU, stridedSliceImpl: stridedSliceImplCPU, stringNGramsImpl: stringNGramsImplCPU, stringSplitImpl: stringSplitImplCPU, stringToHashBucketFastImpl: stringToHashBucketFastImplCPU, subImpl: subImplCPU, tileImpl: tileImplCPU, topKImpl: topKImplCPU, transposeImpl: transposeImplCPU, uniqueImpl: uniqueImplCPU } = shared_exports;
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/packing_util.js
+function getVecChannels(name, rank) {
+  return ["x", "y", "z", "w", "u", "v"].slice(0, rank).map((d) => `${name}.${d}`);
+}
+function getChannels(name, rank) {
+  if (rank === 1) {
+    return [name];
+  }
+  return getVecChannels(name, rank);
+}
+function getSourceCoords(rank, dims) {
+  if (rank === 1) {
+    return "rc";
+  }
+  let coords2 = "";
+  for (let i = 0; i < rank; i++) {
+    coords2 += dims[i];
+    if (i < rank - 1) {
+      coords2 += ",";
+    }
+  }
+  return coords2;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/pack_gpu.js
+var PackProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    if (this.rank === 0) {
+      this.userCode = `
         void main() {
           setOutput(vec4(getA(), 0., 0., 0.));
         }
-      `;else{let e=Qe("rc",this.rank),n=zt(this.rank),o=this.getOutOfBoundsCondition(e),s=this.getSetup(e),i=this.getOutput(e);this.userCode=`
+      `;
+    } else {
+      const channels = getChannels("rc", this.rank);
+      const dtype = getCoordsDataType(this.rank);
+      const outOfBoundsCondition = this.getOutOfBoundsCondition(channels);
+      const setup51 = this.getSetup(channels);
+      const output = this.getOutput(channels);
+      this.userCode = `
         void main() {
-          ${n} rc = getOutputCoords();
+          ${dtype} rc = getOutputCoords();
 
-          if(${o}) {
+          if(${outOfBoundsCondition}) {
             setOutput(vec4(0));
           } else {
-            ${s}
+            ${setup51}
 
-            setOutput(vec4(${i}));
+            setOutput(vec4(${output}));
           }
         }
-      `}}getSourceCoordsArr(t){let e=[];for(let n=0;n<=1;n++)for(let o=0;o<=1;o++){let s=`${n===0?"r":"rp1"}, ${o===0?"c":"cp1"}`;for(let i=2;i<this.rank;i++)s=`${t[t.length-1-i]},`+s;e.push(s)}return e}getOutOfBoundsCondition(t){if(this.rank===1)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let e="";for(let n=this.rank-2;n<this.rank;n++)e+=`${t[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(e+="||");return e}getSetup(t){if(this.rank===1)return"";let e=t.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],o=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`
-      int r = ${e[0]};
-      int c = ${e[1]};
+      `;
+    }
+  }
+  getSourceCoordsArr(dims) {
+    const coords2 = [];
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        let coord = `${row === 0 ? "r" : "rp1"}, ${col === 0 ? "c" : "cp1"}`;
+        for (let d = 2; d < this.rank; d++) {
+          coord = `${dims[dims.length - 1 - d]},` + coord;
+        }
+        coords2.push(coord);
+      }
+    }
+    return coords2;
+  }
+  getOutOfBoundsCondition(dims) {
+    if (this.rank === 1) {
+      return `rc > ${this.enableShapeUniforms ? "outShape" : this.outputShape[0]}`;
+    }
+    let cond = "";
+    for (let i = this.rank - 2; i < this.rank; i++) {
+      cond += `${dims[i]} >= ${this.enableShapeUniforms ? `outShape[${i}]` : this.outputShape[i]}`;
+      if (i < this.rank - 1) {
+        cond += "||";
+      }
+    }
+    return cond;
+  }
+  getSetup(dims) {
+    if (this.rank === 1) {
+      return "";
+    }
+    const innerDims = dims.slice(-2);
+    const col = this.enableShapeUniforms ? `outShape[${this.rank} - 1]` : this.outputShape[this.rank - 1];
+    const row = this.enableShapeUniforms ? `outShape[${this.rank} - 2]` : this.outputShape[this.rank - 2];
+    return `
+      int r = ${innerDims[0]};
+      int c = ${innerDims[1]};
       int rp1 = r + 1;
       int cp1 = c + 1;
 
-      bool cEdge = cp1 >= ${n};
-      bool rEdge = rp1 >= ${o};
-    `}getOutput(t){let e=this.getSourceCoordsArr(t);return this.rank===1?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${e[0]}),
-            cEdge ? 0. : getA(${e[1]}),
-            rEdge ? 0. : getA(${e[2]}),
-            rEdge || cEdge ? 0. : getA(${e[3]})`}};var Id=class{constructor(t,e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let n="";for(let o=0;o<4;o++){let s="thisRC = rc;";o%2===1&&(s+="thisRC.z += 1;"),o>1&&(s+="thisRC.y += 1;"),n+=`
-        ${s}
-        ${o>0?"if(thisRC.y < rows && thisRC.z < cols){":""}
+      bool cEdge = cp1 >= ${col};
+      bool rEdge = rp1 >= ${row};
+    `;
+  }
+  getOutput(dims) {
+    const sourceCoords = this.getSourceCoordsArr(dims);
+    if (this.rank === 1) {
+      const outShape = this.enableShapeUniforms ? "outShape" : this.outputShape[0];
+      return `getA(rc), (rc + 1 >= ${outShape} ? 0. : getA(rc + 1)), 0, 0`;
+    }
+    return `getA(${sourceCoords[0]}),
+            cEdge ? 0. : getA(${sourceCoords[1]}),
+            rEdge ? 0. : getA(${sourceCoords[2]}),
+            rEdge || cEdge ? 0. : getA(${sourceCoords[3]})`;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/reshape_packed_gpu.js
+var ReshapePackedProgram = class {
+  constructor(outputShape, inputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "inputShape", type: "ivec3" }];
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    let mainLoop = ``;
+    for (let i = 0; i < 4; i++) {
+      let thisRC = `thisRC = rc;`;
+      if (i % 2 === 1) {
+        thisRC += `thisRC.z += 1;`;
+      }
+      if (i > 1) {
+        thisRC += `thisRC.y += 1;`;
+      }
+      mainLoop += `
+        ${thisRC}
+        ${i > 0 ? `if(thisRC.y < rows && thisRC.z < cols){` : ""}
           int flatIndex = getFlatIndex(thisRC);
 
           ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);
           vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));
 
-          result[${o}] =
+          result[${i}] =
             getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);
-        ${o>0?"}":""}
-      `}this.userCode=`
-      ${ret(e,this.enableShapeUniforms)}
-      ${this.enableShapeUniforms?gd():hd(t)}
+        ${i > 0 ? "}" : ""}
+      `;
+    }
+    this.userCode = `
+      ${getReshapedInputCoords(inputShape, this.enableShapeUniforms)}
+      ${this.enableShapeUniforms ? getFlatIndexFrom3DOutput() : getFlatIndexFrom3D(outputShape)}
 
       void main() {
         ivec3 rc = getOutputCoords();
@@ -1163,21 +52540,228 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         vec4 result = vec4(0.);
 
         ivec3 thisRC;
-        int rows = ${this.enableShapeUniforms?"outShape[1]":t[1]};
-        int cols = ${this.enableShapeUniforms?"outShape[2]":t[2]};
+        int rows = ${this.enableShapeUniforms ? "outShape[1]" : outputShape[1]};
+        int cols = ${this.enableShapeUniforms ? "outShape[2]" : outputShape[2]};
 
-        ${n}
+        ${mainLoop}
 
         setOutput(result);
       }
-    `}};function ret(r,t){return`
+    `;
+  }
+};
+function getReshapedInputCoords(shape, enableShapeUniforms) {
+  const coordsFromIndexSnippet = enableShapeUniforms ? getLogicalCoordinatesFromFlatIndexByUniform(["r", "c", "d"], "inputShape") : getLogicalCoordinatesFromFlatIndex(["r", "c", "d"], shape);
+  return `
     ivec3 inputCoordsFromReshapedOutCoords(int index) {
-      ${t?eL(["r","c","d"],"inputShape"):ti(["r","c","d"],r)}
+      ${coordsFromIndexSnippet}
       return ivec3(r, c, d);
     }
-  `}var Vw=class{constructor(t){this.gpgpu=t,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(t,e,n){let o=eM(e,n),s=rM(t,o,n);s in this.freeTextures||(this.freeTextures[s]=[]),s in this.usedTextures||(this.usedTextures[s]=[]);let i=tM(t,o,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[s].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=i,this.log();let u=this.freeTextures[s].shift();return this.usedTextures[s].push(u),u}let a;return o===Pr.PACKED_2X2_FLOAT32?a=this.gpgpu.createPackedMatrixTexture(t[0],t[1]):o===Pr.PACKED_2X2_FLOAT16?a=this.gpgpu.createFloat16PackedMatrixTexture(t[0],t[1]):o===Pr.UNPACKED_FLOAT32?a=this.gpgpu.createFloat32MatrixTexture(t[0],t[1]):o===Pr.UNPACKED_FLOAT16?a=this.gpgpu.createFloat16MatrixTexture(t[0],t[1]):o===Pr.PACKED_4X1_UNSIGNED_BYTE&&(a=this.gpgpu.createUnsignedBytesMatrixTexture(t[0],t[1])),this.usedTextures[s].push(a),this.numUsedTextures++,this._numBytesAllocated+=i,this.log(),a}releaseTexture(t,e,n,o){if(this.freeTextures==null)return;let s=eM(n,o),i=rM(e,s,o);i in this.freeTextures||(this.freeTextures[i]=[]);let a=tM(e,s,this.gpgpu.gl,this.gpgpu.textureConfig,o),u=z().get("WEBGL_DELETE_TEXTURE_THRESHOLD");u!==-1&&this._numBytesAllocated>u?(this.gpgpu.deleteMatrixTexture(t.texture),this._numBytesAllocated-=a):(this.freeTextures[i].push(t),this.numFreeTextures++,this._numBytesFree+=a),this.numUsedTextures--;let l=this.usedTextures[i],c=l.indexOf(t);if(c<0)throw new Error("Cannot release a texture that was never provided by this texture manager");l.splice(c,1),this.log()}log(){if(!this.logEnabled)return;let t=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${t})`);let e=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*e)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let t in this.freeTextures)this.freeTextures[t].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});for(let t in this.usedTextures)this.usedTextures[t].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function net(r,t){let e=r;if(t===e.R32F)return 4;if(t===e.R16F)return 2;if(t===e.RGBA32F)return 16;if(t===r.RGBA)return 16;if(t===e.RGBA16F)return 8;if(t===e.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function tM(r,t,e,n,o){let s=oet(t,n),i;if(o){let[u,l]=Xi(r[0],r[1]);i=u*l}else{let[u,l]=Lc(r[0],r[1]);i=u*l}let a=net(e,s);return i*a}function oet(r,t){switch(r){case Pr.PACKED_2X2_FLOAT32:return Ow(t);case Pr.PACKED_2X2_FLOAT16:return Pw(t);case Pr.UNPACKED_FLOAT32:return Dw(t);case Pr.UNPACKED_FLOAT16:return Rw(t);case Pr.PACKED_4X1_UNSIGNED_BYTE:return Fw(t);default:throw new Error(`Unknown physical texture type ${r}`)}}function set(r){return z().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?r?Pr.PACKED_2X2_FLOAT32:Pr.UNPACKED_FLOAT32:r?Pr.PACKED_2X2_FLOAT16:Pr.UNPACKED_FLOAT16}function eM(r,t){if(r===jr.UPLOAD)return Pr.PACKED_2X2_FLOAT32;if(r===jr.RENDER||r==null)return set(t);if(r===jr.DOWNLOAD||r===jr.PIXELS)return Pr.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${r}`)}function rM(r,t,e){return`${r[0]}_${r[1]}_${t}_${e}`}var tn=class{constructor(t,e){this.variableNames=["A"],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`
+  `;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/texture_manager.js
+var TextureManager = class {
+  constructor(gpgpu) {
+    this.gpgpu = gpgpu;
+    this.numUsedTextures = 0;
+    this.numFreeTextures = 0;
+    this._numBytesAllocated = 0;
+    this._numBytesFree = 0;
+    this.freeTextures = {};
+    this.logEnabled = false;
+    this.usedTextures = {};
+  }
+  acquireTexture(shapeRC, usage, isPacked) {
+    const physicalTexType = getPhysicalFromLogicalTextureType(usage, isPacked);
+    const shapeKey = getKeyFromTextureShape(shapeRC, physicalTexType, isPacked);
+    if (!(shapeKey in this.freeTextures)) {
+      this.freeTextures[shapeKey] = [];
+    }
+    if (!(shapeKey in this.usedTextures)) {
+      this.usedTextures[shapeKey] = [];
+    }
+    const texBytes = computeBytes(shapeRC, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);
+    if (this.freeTextures[shapeKey].length > 0) {
+      this.numFreeTextures--;
+      this.numUsedTextures++;
+      this._numBytesFree -= texBytes;
+      this.log();
+      const newTexture2 = this.freeTextures[shapeKey].shift();
+      this.usedTextures[shapeKey].push(newTexture2);
+      return newTexture2;
+    }
+    let newTexture;
+    if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT32) {
+      newTexture = this.gpgpu.createPackedMatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.PACKED_2X2_FLOAT16) {
+      newTexture = this.gpgpu.createFloat16PackedMatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT32) {
+      newTexture = this.gpgpu.createFloat32MatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.UNPACKED_FLOAT16) {
+      newTexture = this.gpgpu.createFloat16MatrixTexture(shapeRC[0], shapeRC[1]);
+    } else if (physicalTexType === PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE) {
+      newTexture = this.gpgpu.createUnsignedBytesMatrixTexture(shapeRC[0], shapeRC[1]);
+    }
+    this.usedTextures[shapeKey].push(newTexture);
+    this.numUsedTextures++;
+    this._numBytesAllocated += texBytes;
+    this.log();
+    return newTexture;
+  }
+  releaseTexture(texture, shape, logicalTexType, isPacked) {
+    if (this.freeTextures == null) {
+      return;
+    }
+    const physicalTexType = getPhysicalFromLogicalTextureType(logicalTexType, isPacked);
+    const shapeKey = getKeyFromTextureShape(shape, physicalTexType, isPacked);
+    if (!(shapeKey in this.freeTextures)) {
+      this.freeTextures[shapeKey] = [];
+    }
+    const texBytes = computeBytes(shape, physicalTexType, this.gpgpu.gl, this.gpgpu.textureConfig, isPacked);
+    const deleteTexThreshold = env().get("WEBGL_DELETE_TEXTURE_THRESHOLD");
+    if (deleteTexThreshold !== -1 && this._numBytesAllocated > deleteTexThreshold) {
+      this.gpgpu.deleteMatrixTexture(texture.texture);
+      this._numBytesAllocated -= texBytes;
+    } else {
+      this.freeTextures[shapeKey].push(texture);
+      this.numFreeTextures++;
+      this._numBytesFree += texBytes;
+    }
+    this.numUsedTextures--;
+    const texList = this.usedTextures[shapeKey];
+    const texIndex = texList.indexOf(texture);
+    if (texIndex < 0) {
+      throw new Error("Cannot release a texture that was never provided by this texture manager");
+    }
+    texList.splice(texIndex, 1);
+    this.log();
+  }
+  log() {
+    if (!this.logEnabled) {
+      return;
+    }
+    const total = this.numFreeTextures + this.numUsedTextures;
+    console.log("Free/Used", `${this.numFreeTextures} / ${this.numUsedTextures}`, `(${total})`);
+    const freeRatio = this._numBytesFree / this._numBytesAllocated;
+    console.log(`Bytes allocated: ${this._numBytesAllocated}`);
+    console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100 * freeRatio)}%)`);
+  }
+  get numBytesAllocated() {
+    return this._numBytesAllocated;
+  }
+  get numBytesFree() {
+    return this._numBytesFree;
+  }
+  getNumUsedTextures() {
+    return this.numUsedTextures;
+  }
+  getNumFreeTextures() {
+    return this.numFreeTextures;
+  }
+  dispose() {
+    if (this.freeTextures == null) {
+      return;
+    }
+    for (const texShape in this.freeTextures) {
+      this.freeTextures[texShape].forEach((tex) => {
+        this.gpgpu.deleteMatrixTexture(tex.texture);
+      });
+    }
+    for (const texShape in this.usedTextures) {
+      this.usedTextures[texShape].forEach((tex) => {
+        this.gpgpu.deleteMatrixTexture(tex.texture);
+      });
+    }
+    this.freeTextures = null;
+    this.usedTextures = null;
+    this.numUsedTextures = 0;
+    this.numFreeTextures = 0;
+    this._numBytesAllocated = 0;
+    this._numBytesFree = 0;
+  }
+};
+function numBytesForInternalFormat(gl, internalFormat) {
+  const glany = gl;
+  if (internalFormat === glany.R32F) {
+    return 4;
+  } else if (internalFormat === glany.R16F) {
+    return 2;
+  } else if (internalFormat === glany.RGBA32F) {
+    return 16;
+  } else if (internalFormat === gl.RGBA) {
+    return 16;
+  } else if (internalFormat === glany.RGBA16F) {
+    return 8;
+  } else if (internalFormat === glany.RGBA8) {
+    return 4;
+  }
+  throw new Error(`Unknown internal format ${internalFormat}`);
+}
+function computeBytes(shape, physicalTexType, gl, textureConfig, isPacked) {
+  const internalFormat = internalFormatForPhysicalTexType(physicalTexType, textureConfig);
+  let numElements;
+  if (isPacked) {
+    const [packedWidth, packedHeight] = getPackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);
+    numElements = packedWidth * packedHeight;
+  } else {
+    const [width, height] = getUnpackedMatrixTextureShapeWidthHeight(shape[0], shape[1]);
+    numElements = width * height;
+  }
+  const bytesPerElement2 = numBytesForInternalFormat(gl, internalFormat);
+  return numElements * bytesPerElement2;
+}
+function internalFormatForPhysicalTexType(physicalTexType, textureConfig) {
+  switch (physicalTexType) {
+    case PhysicalTextureType.PACKED_2X2_FLOAT32:
+      return getInternalFormatForPackedMatrixTexture(textureConfig);
+    case PhysicalTextureType.PACKED_2X2_FLOAT16:
+      return getInternalFormatForFloat16PackedMatrixTexture(textureConfig);
+    case PhysicalTextureType.UNPACKED_FLOAT32:
+      return getInternalFormatForFloat32MatrixTexture(textureConfig);
+    case PhysicalTextureType.UNPACKED_FLOAT16:
+      return getInternalFormatForFloat16MatrixTexture(textureConfig);
+    case PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE:
+      return getInternalFormatForUnsignedBytesMatrixTexture(textureConfig);
+    default:
+      throw new Error(`Unknown physical texture type ${physicalTexType}`);
+  }
+}
+function getPhysicalTextureForRendering(isPacked) {
+  if (env().getBool("WEBGL_RENDER_FLOAT32_ENABLED")) {
+    if (isPacked) {
+      return PhysicalTextureType.PACKED_2X2_FLOAT32;
+    }
+    return PhysicalTextureType.UNPACKED_FLOAT32;
+  }
+  if (isPacked) {
+    return PhysicalTextureType.PACKED_2X2_FLOAT16;
+  }
+  return PhysicalTextureType.UNPACKED_FLOAT16;
+}
+function getPhysicalFromLogicalTextureType(logicalTexType, isPacked) {
+  if (logicalTexType === TextureUsage.UPLOAD) {
+    return PhysicalTextureType.PACKED_2X2_FLOAT32;
+  } else if (logicalTexType === TextureUsage.RENDER || logicalTexType == null) {
+    return getPhysicalTextureForRendering(isPacked);
+  } else if (logicalTexType === TextureUsage.DOWNLOAD || logicalTexType === TextureUsage.PIXELS) {
+    return PhysicalTextureType.PACKED_4X1_UNSIGNED_BYTE;
+  }
+  throw new Error(`Unknown logical texture type ${logicalTexType}`);
+}
+function getKeyFromTextureShape(shapeRowsCol, physicalTexType, isPacked) {
+  return `${shapeRowsCol[0]}_${shapeRowsCol[1]}_${physicalTexType}_${isPacked}`;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/unaryop_gpu.js
+var UnaryOpProgram = class {
+  constructor(aShape, opSnippet) {
+    this.variableNames = ["A"];
+    this.outputShape = aShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       float unaryOperation(float x) {
-        ${e}
+        ${opSnippet}
       }
 
       void main() {
@@ -1186,11 +52770,25 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}},fr="if (isnan(x)) return x;",nM="return x;",lk="return abs(x);";var oM="return (x >= 0.0) ? x : (exp(x) - 1.0);",sM=fr+`
+    `;
+  }
+};
+var CHECK_NAN_SNIPPET = `if (isnan(x)) return x;`;
+var LINEAR = `return x;`;
+var ABS = `return abs(x);`;
+var ELU2 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;
+var RELU = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : x;
-`,iM=fr+`
+`;
+var RELU6 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,Gc="return x;",aM="return 1.0 / (1.0 + exp(-1.0 * x));";var uM="return x;",cM=`
+`;
+var CLONE = "return x;";
+var SIGMOID = `return 1.0 / (1.0 + exp(-1.0 * x));`;
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/unaryop_packed_gpu.js
+var LINEAR2 = `return x;`;
+var ELU3 = `
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -1199,7 +52797,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,pM=`
+`;
+var RELU2 = `
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1209,7 +52808,8 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,mM=`
+`;
+var RELU62 = `
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -1219,9 +52819,18 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,fM="return 1.0 / (1.0 + exp(-1.0 * x));",so=class{constructor(t,e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`
+`;
+var SIGMOID2 = `return 1.0 / (1.0 + exp(-1.0 * x));`;
+var UnaryOpPackedProgram = class {
+  constructor(aShape, opSnippet) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = aShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       vec4 unaryOperation(vec4 x) {
-        ${e}
+        ${opSnippet}
       }
 
       void main() {
@@ -1230,19 +52839,896 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
 
         setOutput(y);
       }
-    `}};var Gw=class{constructor(t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let e=t.length,n=Qe("rc",e),o=zt(e),s=QL(e,n),i=n.slice(-2),a=e<=1?"rc":`vec2(${i.join(",")})`;this.userCode=`
-      void main() {
-        ${o} rc = getOutputCoords();
-        vec4 packedInput = getA(${s});
+    `;
+  }
+};
 
-        setOutput(getChannel(packedInput, ${a}));
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/unpack_gpu.js
+var UnpackProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = false;
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const rank = outputShape.length;
+    const channels = getChannels("rc", rank);
+    const dtype = getCoordsDataType(rank);
+    const sourceCoords = getSourceCoords(rank, channels);
+    const innerDims = channels.slice(-2);
+    const coords2 = rank <= 1 ? "rc" : `vec2(${innerDims.join(",")})`;
+    this.userCode = `
+      void main() {
+        ${dtype} rc = getOutputCoords();
+        vec4 packedInput = getA(${sourceCoords});
+
+        setOutput(getChannel(packedInput, ${coords2}));
       }
-    `}};var aet=Ur.whereImpl,uet=1e-7,cet=1e-4,Ww={};function pet(r){return r in Ww||(Ww[r]={}),Ww[r]}var met=z().getNumber("CPU_HANDOFF_SIZE_THRESHOLD"),fet=600;function det(){return z().global.screen==null?1024:z().global.screen.height*z().global.screen.width*window.devicePixelRatio*fet/1024/1024}var _u=class extends zo{constructor(t){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!z().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let e;if(t!=null){if(t instanceof Bc)e=t;else{let n=Gn(z().getNumber("WEBGL_VERSION"),t);e=new Bc(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let n=Gn(z().getNumber("WEBGL_VERSION"));e=new Bc(n),this.binaryCache=pet(z().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=e,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new Vw(this.gpgpu),this.numMBBeforeWarning=det(),this.texData=new ra(this,Pn())}nextDataId(){return _u.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(t,e,n,o,s,i){let a=this.makeTensorInfo(e,n),u=this.texData.get(a.dataId);u.isPacked=!1,u.texture={texture:t,texShape:[o,s]},u.texShape=[o,s];let l=fd(e),c=new Jh(l,!1,i),p=this.runWebGLProgram(c,[a],n,[[o,s]]);return p.shape=e,u.texture=null,this.disposeIntermediateTensorInfo(a),p.dataId}write(t,e,n){if((z().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||z().getBool("DEBUG"))&&this.checkNumericalProblems(t),n==="complex64"&&t!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let o={id:this.nextDataId()};return this.texData.set(o,{shape:e,dtype:n,values:t,usage:jr.UPLOAD,refCount:1}),o}refCount(t){return this.texData.has(t)?this.texData.get(t).refCount:0}incRef(t){let e=this.texData.get(t);e.refCount++}decRef(t){if(this.texData.has(t)){let e=this.texData.get(t);e.refCount--}}move(t,e,n,o,s){if(z().getBool("DEBUG")&&this.checkNumericalProblems(e),o==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(t,{shape:n,dtype:o,values:e,usage:jr.UPLOAD,refCount:s})}disposeIntermediateTensorInfo(t){this.disposeData(t.dataId)}readSync(t){let e=this.texData.get(t),{values:n,dtype:o,complexTensorInfos:s,slice:i,shape:a,isPacked:u}=e;if(i!=null){let m;u?m=new so(a,Gc):m=new tn(a,Gc);let f=this.runWebGLProgram(m,[{dataId:t,shape:a,dtype:o}],o),d=this.readSync(f.dataId);return this.disposeIntermediateTensorInfo(f),d}if(n!=null)return this.convertAndCacheOnCPU(t);if(o==="string")return n;let l=this.activeTimers!=null,c;l&&(c=y.now());let p;if(o==="complex64"){let m=this.readSync(s.real.dataId),f=this.readSync(s.imag.dataId);p=v.mergeRealAndImagArrays(m,f)}else p=this.getValuesFromTexture(t);return l&&(this.downloadWaitMs+=y.now()-c),this.convertAndCacheOnCPU(t,p)}async read(t){if(this.pendingRead.has(t)){let d=this.pendingRead.get(t);return new Promise(h=>d.push(h))}let e=this.texData.get(t),{values:n,shape:o,slice:s,dtype:i,complexTensorInfos:a,isPacked:u}=e;if(s!=null){let d;u?d=new so(o,Gc):d=new tn(o,Gc);let h=this.runWebGLProgram(d,[{dataId:t,shape:o,dtype:i}],i),g=this.read(h.dataId);return this.disposeIntermediateTensorInfo(h),g}if(n!=null)return this.convertAndCacheOnCPU(t);if(z().getBool("DEBUG")&&!z().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&z().getNumber("WEBGL_VERSION")===2)throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let l=null,c;if(i!=="complex64"&&z().get("WEBGL_BUFFER_SUPPORTED")){c=this.decode(t);let d=this.texData.get(c.dataId);l=this.gpgpu.createBufferFromTexture(d.texture.texture,...jh(o))}this.pendingRead.set(t,[]),i!=="complex64"&&await this.gpgpu.createAndWaitForFence();let p;if(i==="complex64"){let d=await Promise.all([this.read(a.real.dataId),this.read(a.imag.dataId)]),h=d[0],g=d[1];p=v.mergeRealAndImagArrays(h,g)}else if(l==null)p=this.getValuesFromTexture(t);else{let d=y.sizeFromShape(o);p=this.gpgpu.downloadFloat32MatrixFromBuffer(l,d)}if(c!=null&&this.disposeIntermediateTensorInfo(c),l!=null){let d=this.gpgpu.gl;yt(d,()=>d.deleteBuffer(l))}let m=this.convertAndCacheOnCPU(t,p),f=this.pendingRead.get(t);return this.pendingRead.delete(t),f.forEach(d=>d(m)),this.pendingDisposal.has(t)&&(this.pendingDisposal.delete(t),this.disposeData(t)&&Pn().removeDataId(t,this),this.pendingDeletes--),m}readToGPU(t,e={}){let n=this.texData.get(t),{values:o,shape:s,slice:i,dtype:a,isPacked:u,texture:l}=n;if(a==="complex64")throw new Error("Does not support reading texture for complex64 dtype.");if(i!=null){let f;u?f=new so(s,Gc):f=new tn(s,Gc);let d=this.runWebGLProgram(f,[{dataId:t,shape:s,dtype:a}],a),h=this.readToGPU(d,e);return this.disposeIntermediateTensorInfo(d),h}if(l==null)throw o!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let c=this.decode(t,e.customTexShape),p=Pn().makeTensorFromTensorInfo(c),m=this.texData.get(c.dataId);return Object.assign({tensorRef:p},m.texture)}bufferSync(t){let e=this.readSync(t.dataId);if(t.dtype==="string")try{let n=e.map(o=>y.decodeString(o));return wt(t.shape,t.dtype,n)}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return wt(t.shape,t.dtype,e)}checkNumericalProblems(t){if(t!=null)for(let e=0;e<t.length;e++){let n=t[e];if(!ST(n))throw z().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(t){let{shape:e,dtype:n,isPacked:o}=this.texData.get(t),s=y.sizeFromShape(e);if(z().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let m=this.decode(t),f=this.texData.get(m.dataId),d=this.gpgpu.downloadMatrixFromPackedTexture(f.texture.texture,...jh(e)).subarray(0,s);return this.disposeIntermediateTensorInfo(m),d}let i=z().getBool("WEBGL_PACK")&&o===!0,a=i?fd(e):e,u=i?new Aw(a):new _w(a),l=this.runWebGLProgram(u,[{shape:a,dtype:n,dataId:t}],"float32"),c=this.texData.get(l.dataId),p=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(c.texture.texture,c.texShape[0],c.texShape[1]).subarray(0,s);return this.disposeIntermediateTensorInfo(l),p}timerAvailable(){return z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(t){let e=this.activeTimers,n=[],o=!1;this.programTimersStack==null?(this.programTimersStack=n,o=!0):this.activeTimers.push(n),this.activeTimers=n,t();let s=y.flatten(this.activeTimers.map(u=>u.query)).filter(u=>u!=null),i=y.flatten(this.activeTimers.map(u=>u.name)).filter(u=>u!=null);this.activeTimers=e,o&&(this.programTimersStack=null);let a={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let u=await Promise.all(s);a.kernelMs=y.sum(u),a.getExtraProfileInfo=()=>u.map((l,c)=>({name:i[c],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(", ")}else a.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,a})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:y.now(),endMs:null}}endTimer(t){return z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),t):(t.endMs=y.now(),t)}async getQueryTime(t){if(z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(t);let e=t;return e.endMs-e.startMs}disposeData(t,e=!1){if(this.pendingDisposal.has(t))return!1;if(!this.texData.has(t))return!0;if(e?this.texData.get(t).refCount=0:this.texData.get(t).refCount--,!e&&this.texData.get(t).refCount>0)return!1;if(this.pendingRead.has(t))return this.pendingDisposal.add(t),this.pendingDeletes++,!1;this.releaseGPUData(t);let{complexTensorInfos:n}=this.texData.get(t);return n!=null&&(this.disposeData(n.real.dataId,e),this.disposeData(n.imag.dataId,e)),this.texData.delete(t),!0}releaseGPUData(t){let{texture:e,dtype:n,texShape:o,usage:s,isPacked:i,slice:a}=this.texData.get(t),u=a&&a.origDataId||t,l=this.dataRefCount.get(u);l>1?this.dataRefCount.set(u,l-1):(this.dataRefCount.delete(u),e!=null&&(this.numBytesInGPU-=this.computeBytes(o,n),this.textureManager.releaseTexture(e,o,s,i)));let c=this.texData.get(t);c.texture=null,c.texShape=null,c.isPacked=!1,c.slice=null}getTexture(t){return this.uploadToGPU(t),this.texData.get(t).texture.texture}getDataInfo(t){return this.texData.get(t)}shouldExecuteOnCPU(t,e=met){return z().getBool("WEBGL_CPU_FORWARD")&&t.every(n=>this.texData.get(n.dataId).texture==null&&y.sizeFromShape(n.shape)<e)}getGPGPUContext(){return this.gpgpu}where(t){v.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let e=t.dataSync();return aet(t.shape,e)}packedUnaryOp(t,e,n){let o=new so(t.shape,e),s=this.compileAndRun(o,[t],n);return Pn().makeTensorFromTensorInfo(s)}abs(t){if(this.shouldExecuteOnCPU([t])&&t.dtype!=="complex64"){let o=Mw(this.texData.get(t.dataId).values);return this.makeOutput(t.shape,t.dtype,o)}if(z().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(t,lk,t.dtype);let e=new tn(t.shape,lk),n=this.compileAndRun(e,[t]);return Pn().makeTensorFromTensorInfo(n)}makeTensorInfo(t,e,n){let o;if(e==="string"&&n!=null&&n.length>0&&y.isString(n[0])){let s=n.map(i=>y.encodeString(i));o=this.write(s,t,e)}else o=this.write(n,t,e);return this.texData.get(o).usage=null,{dataId:o,shape:t,dtype:e}}makeOutput(t,e,n){return Pn().makeTensorFromTensorInfo(this.makeTensorInfo(t,e,n),this)}unpackTensor(t){let e=new Gw(t.shape);return this.runWebGLProgram(e,[t],t.dtype)}packTensor(t){let e=new Bw(t.shape),n=!0;return this.runWebGLProgram(e,[t],t.dtype,null,n)}packedReshape(t,e){let n=[xl(t.shape),...yl(t.shape)],o={dtype:t.dtype,shape:n,dataId:t.dataId},s=[xl(e),...yl(e)],i=new Id(s,n),a=!0,u=[n],l=this.runWebGLProgram(i,[o],t.dtype,u,a);return{dataId:l.dataId,shape:e,dtype:l.dtype}}decode(t,e){let n=this.texData.get(t),{isPacked:o,shape:s,dtype:i}=n;if(e!=null){let m=y.sizeFromShape(s),f=e[0]*e[1]*4;y.assert(m<=f,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let a=fd(s),u;o?u=new Ew(a):u=new kw(a);let l=!0,c=[e!=null?e:jh(a)],p=this.runWebGLProgram(u,[{shape:a,dtype:i,dataId:t}],i,c,l,e);return{dtype:i,shape:s,dataId:p.dataId}}runWebGLProgram(t,e,n,o,s=!1,i){let a=this.makeTensorInfo(t.outputShape,n),u=this.texData.get(a.dataId);if(t.packedOutput&&(u.isPacked=!0),t.outPackingScheme===ku.DENSE){let x=i!=null?i:jh(t.outputShape);u.texShape=x.map(b=>b*2)}if(t.outTexUsage!=null&&(u.usage=t.outTexUsage),y.sizeFromShape(a.shape)===0)return u.values=y.getTypedArrayFromDType(a.dtype,0),a;let l=[],c=e.map(x=>{if(x.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let b=this.texData.get(x.dataId);if(b.texture==null){if(!t.packedInputs&&y.sizeFromShape(x.shape)<=z().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:x.shape,texData:null,isUniform:!0,uniformValues:b.values};t.packedInputs&&(b.isPacked=!0,b.shape=x.shape)}if(this.uploadToGPU(x.dataId),!!b.isPacked!=!!t.packedInputs)x=b.isPacked?this.unpackTensor(x):this.packTensor(x),l.push(x),b=this.texData.get(x.dataId);else if(b.isPacked&&!Eu(b.shape,x.shape)){let w=x,C=x.shape;x.shape=b.shape,x=this.packedReshape(x,C),l.push(x),b=this.texData.get(x.dataId),w.shape=C}return{shape:x.shape,texData:b,isUniform:!1}});this.uploadToGPU(a.dataId);let p={shape:a.shape,texData:u,isUniform:!1},m=uL(t,c,p),f=this.getAndSaveBinary(m,()=>aL(this.gpgpu,t,c,p)),d=this.activeTimers!=null,h;d&&(h=this.startTimer()),z().get("ENGINE_COMPILE_ONLY")||lL(this.gpgpu,f,c,p,o),l.forEach(x=>this.disposeIntermediateTensorInfo(x)),d&&(h=this.endTimer(h),this.activeTimers.push({name:t.constructor.name,query:this.getQueryTime(h)}));let g=z().get("WEBGL_FLUSH_THRESHOLD");if(g>0){let x=y.now();x-this.lastGlFlushTime>g&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=x)}if(!z().getBool("WEBGL_LAZILY_UNPACK")&&u.isPacked&&s===!1){let x=this.unpackTensor(a);return this.disposeIntermediateTensorInfo(a),x}return a}compileAndRun(t,e,n,o,s=!1){return n=n||e[0].dtype,this.runWebGLProgram(t,e,n,o,s)}getAndSaveBinary(t,e){return t in this.binaryCache||(this.binaryCache[t]=e()),this.binaryCache[t]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(z().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!="undefined"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=B(()=>{if(!z().get("WEBGL_RENDER_FLOAT32_ENABLED")){let t=z().getBool("DEBUG");z().set("DEBUG",!1);let e=this.abs(mt(1e-8)).dataSync()[0];if(z().set("DEBUG",t),e>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?uet:cet}uploadToGPU(t){let e=this.texData.get(t),{shape:n,dtype:o,values:s,texture:i,usage:a,isPacked:u}=e;if(i!=null)return;let l=this.activeTimers!=null,c;l&&(c=y.now());let p=e.texShape;if(p==null&&(p=PT(n,u),e.texShape=p),s!=null){let m=fd(n),f,d=p[1],h=p[0],g=s instanceof Uint8Array||s instanceof Uint8ClampedArray;(u||!g)&&([d,h]=Xi(p[0],p[1])),u?f=new $w(m,g):f=new Jh(m,g);let x=g?[h,d]:p,b=this.makeTensorInfo(x,o),w=this.texData.get(b.dataId);g?w.usage=jr.PIXELS:w.usage=jr.UPLOAD,w.texShape=x,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(b.dataId),d,h,s);let C=[[h,d]],N=!0,_=this.runWebGLProgram(f,[b],o,C,N),A=this.texData.get(_.dataId);e.texShape=A.texShape,e.isPacked=A.isPacked,e.usage=A.usage,z().get("ENGINE_COMPILE_ONLY")?this.disposeData(_.dataId):(e.texture=A.texture,e.values=null,this.texData.delete(_.dataId)),this.disposeIntermediateTensorInfo(b),l&&(this.uploadWaitMs+=y.now()-c)}else{let m=this.acquireTexture(p,a,o,u);e.texture=m}}convertAndCacheOnCPU(t,e){let n=this.texData.get(t),{dtype:o}=n;return this.releaseGPUData(t),e!=null&&(n.values=het(e,o)),n.values}acquireTexture(t,e,n,o){if(this.numBytesInGPU+=this.computeBytes(t,n),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let s=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${s} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(t,e,o)}computeBytes(t,e){return t[0]*t[1]*y.bytesPerElement(e)}checkCompileCompletion(){for(let[,t]of Object.entries(this.binaryCache))this.checkCompletion_(t)}async checkCompileCompletionAsync(){let t=[];if(this.gpgpu.parallelCompilationExtension){for(let[,e]of Object.entries(this.binaryCache))t.push(this.checkCompletionAsync_(e));return Promise.all(t)}else{for(let[,e]of Object.entries(this.binaryCache)){let n=new Promise(o=>{try{this.checkCompletion_(e),o(!0)}catch(s){throw s}});t.push(n)}return Promise.all(t)}}async checkCompletionAsync_(t){return this.gpgpu.gl.getProgramParameter(t.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(t):(await gh(),this.checkCompletionAsync_(t))}checkCompletion_(t){if(this.gpgpu.gl.getProgramParameter(t.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(t.webGLProgram)),this.gpgpu.gl.getShaderParameter(t.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(Cw(t.source,this.gpgpu.gl.getShaderInfoLog(t.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let[,t]of Object.entries(this.binaryCache)){let{uniformLocations:e,customUniformLocations:n,infLoc:o,nanLoc:s,inShapesLocations:i,inTexShapesLocations:a,outShapeLocation:u,outShapeStridesLocation:l,outTexShapeLocation:c}=WT(this.gpgpu,t.program,t.webGLProgram);t.uniformLocations=e,t.customUniformLocations=n,t.infLoc=o,t.nanLoc=s,t.inShapesLocations=i,t.inTexShapesLocations=a,t.outShapeLocation=u,t.outShapeStridesLocation=l,t.outTexShapeLocation=c}}createTensorFromTexture(t,e,n){let{texture:o,height:s,width:i,channels:a}=t,u=Pn().backend;if(!u.gpgpu.gl.isTexture(o))throw new Error("The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.");let l=u.writeTexture(o,e,n,s,i,a);return Pn().makeTensorFromDataId(l,e,n,u)}};_u.nextDataId=0;function het(r,t){if(t==="float32"||t==="complex64")return r;if(t==="int32"||t==="bool"){let e=t==="int32"?new Int32Array(r.length):new Uint8Array(r.length);for(let n=0;n<e.length;++n)e[n]=Math.round(r[n]);return e}else throw new Error(`Unknown dtype ${t}`)}var dM="4.0.0";function hM(){z().set("WEBGL_FORCE_F16_TEXTURES",!0)}Kl.isBrowser()&&Xp("webgl",()=>new _u,2);var Zke={forceHalfFloat:hM};var Sd=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/backend_webgl.js
+var whereImpl3 = kernel_impls_exports.whereImpl;
+var EPSILON_FLOAT322 = 1e-7;
+var EPSILON_FLOAT162 = 1e-4;
+var binaryCaches = {};
+function getBinaryCache(webGLVersion) {
+  if (webGLVersion in binaryCaches) {
+    return binaryCaches[webGLVersion];
+  }
+  binaryCaches[webGLVersion] = {};
+  return binaryCaches[webGLVersion];
+}
+var CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber("CPU_HANDOFF_SIZE_THRESHOLD");
+var BEFORE_PAGING_CONSTANT = 600;
+function numMBBeforeWarning() {
+  if (env().global.screen == null) {
+    return 1024;
+  }
+  return env().global.screen.height * env().global.screen.width * window.devicePixelRatio * BEFORE_PAGING_CONSTANT / 1024 / 1024;
+}
+var MathBackendWebGL = class extends KernelBackend {
+  constructor(gpuResource) {
+    super();
+    this.pendingRead = /* @__PURE__ */ new WeakMap();
+    this.pendingDisposal = /* @__PURE__ */ new WeakSet();
+    this.dataRefCount = /* @__PURE__ */ new WeakMap();
+    this.numBytesInGPU = 0;
+    this.uploadWaitMs = 0;
+    this.downloadWaitMs = 0;
+    this.lastGlFlushTime = 0;
+    this.warnedAboutMemory = false;
+    this.pendingDeletes = 0;
+    this.disposed = false;
+    if (!env().getBool("HAS_WEBGL")) {
+      throw new Error("WebGL is not supported on this device");
+    }
+    let newGPGPU;
+    if (gpuResource != null) {
+      if (gpuResource instanceof GPGPUContext) {
+        newGPGPU = gpuResource;
+      } else {
+        const gl = getWebGLContext(env().getNumber("WEBGL_VERSION"), gpuResource);
+        newGPGPU = new GPGPUContext(gl);
+      }
+      this.binaryCache = {};
+      this.gpgpuCreatedLocally = false;
+    } else {
+      const gl = getWebGLContext(env().getNumber("WEBGL_VERSION"));
+      newGPGPU = new GPGPUContext(gl);
+      this.binaryCache = getBinaryCache(env().getNumber("WEBGL_VERSION"));
+      this.gpgpuCreatedLocally = true;
+    }
+    this.gpgpu = newGPGPU;
+    this.canvas = this.gpgpu.gl.canvas;
+    this.textureManager = new TextureManager(this.gpgpu);
+    this.numMBBeforeWarning = numMBBeforeWarning();
+    this.texData = new DataStorage(this, engine());
+  }
+  nextDataId() {
+    return MathBackendWebGL.nextDataId++;
+  }
+  numDataIds() {
+    return this.texData.numDataIds() - this.pendingDeletes;
+  }
+  writeTexture(texture, shape, dtype, texHeight, texWidth, channels) {
+    const input2 = this.makeTensorInfo(shape, dtype);
+    const inData = this.texData.get(input2.dataId);
+    inData.isPacked = false;
+    inData.texture = { texture, texShape: [texHeight, texWidth] };
+    inData.texShape = [texHeight, texWidth];
+    const shapeAs3D = getShapeAs3D(shape);
+    const program = new EncodeMatrixProgram(shapeAs3D, false, channels);
+    const output = this.runWebGLProgram(program, [input2], dtype, [[texHeight, texWidth]]);
+    output.shape = shape;
+    inData.texture = null;
+    this.disposeIntermediateTensorInfo(input2);
+    return output.dataId;
+  }
+  write(values, shape, dtype) {
+    if (env().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS") || env().getBool("DEBUG")) {
+      this.checkNumericalProblems(values);
+    }
+    if (dtype === "complex64" && values != null) {
+      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);
+    }
+    const dataId = { id: this.nextDataId() };
+    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount: 1 });
+    return dataId;
+  }
+  refCount(dataId) {
+    if (this.texData.has(dataId)) {
+      const tensorData = this.texData.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const texData = this.texData.get(dataId);
+    texData.refCount++;
+  }
+  decRef(dataId) {
+    if (this.texData.has(dataId)) {
+      const texData = this.texData.get(dataId);
+      texData.refCount--;
+    }
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    if (env().getBool("DEBUG")) {
+      this.checkNumericalProblems(values);
+    }
+    if (dtype === "complex64") {
+      throw new Error(`Cannot write to a complex64 dtype. Please use tf.complex(real, imag).`);
+    }
+    this.texData.set(dataId, { shape, dtype, values, usage: TextureUsage.UPLOAD, refCount });
+  }
+  disposeIntermediateTensorInfo(tensorInfo) {
+    this.disposeData(tensorInfo.dataId);
+  }
+  readSync(dataId) {
+    const texData = this.texData.get(dataId);
+    const { values, dtype, complexTensorInfos, slice: slice5, shape, isPacked } = texData;
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const data = this.readSync(res.dataId);
+      this.disposeIntermediateTensorInfo(res);
+      return data;
+    }
+    if (values != null) {
+      return this.convertAndCacheOnCPU(dataId);
+    }
+    if (dtype === "string") {
+      return values;
+    }
+    const shouldTimeProgram = this.activeTimers != null;
+    let start;
+    if (shouldTimeProgram) {
+      start = util_exports.now();
+    }
+    let result;
+    if (dtype === "complex64") {
+      const realValues = this.readSync(complexTensorInfos.real.dataId);
+      const imagValues = this.readSync(complexTensorInfos.imag.dataId);
+      result = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    } else {
+      result = this.getValuesFromTexture(dataId);
+    }
+    if (shouldTimeProgram) {
+      this.downloadWaitMs += util_exports.now() - start;
+    }
+    return this.convertAndCacheOnCPU(dataId, result);
+  }
+  async read(dataId) {
+    if (this.pendingRead.has(dataId)) {
+      const subscribers2 = this.pendingRead.get(dataId);
+      return new Promise((resolve) => subscribers2.push(resolve));
+    }
+    const texData = this.texData.get(dataId);
+    const { values, shape, slice: slice5, dtype, complexTensorInfos, isPacked } = texData;
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const data = this.read(res.dataId);
+      this.disposeIntermediateTensorInfo(res);
+      return data;
+    }
+    if (values != null) {
+      return this.convertAndCacheOnCPU(dataId);
+    }
+    if (env().getBool("DEBUG")) {
+      if (!env().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED") && env().getNumber("WEBGL_VERSION") === 2) {
+        throw new Error(`tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.`);
+      }
+    }
+    let buffer2 = null;
+    let tmpDownloadTarget;
+    if (dtype !== "complex64" && env().get("WEBGL_BUFFER_SUPPORTED")) {
+      tmpDownloadTarget = this.decode(dataId);
+      const tmpData = this.texData.get(tmpDownloadTarget.dataId);
+      buffer2 = this.gpgpu.createBufferFromTexture(tmpData.texture.texture, ...getDenseTexShape(shape));
+    }
+    this.pendingRead.set(dataId, []);
+    if (dtype !== "complex64") {
+      await this.gpgpu.createAndWaitForFence();
+    }
+    let vals;
+    if (dtype === "complex64") {
+      const ps = await Promise.all([
+        this.read(complexTensorInfos.real.dataId),
+        this.read(complexTensorInfos.imag.dataId)
+      ]);
+      const realValues = ps[0];
+      const imagValues = ps[1];
+      vals = backend_util_exports.mergeRealAndImagArrays(realValues, imagValues);
+    } else if (buffer2 == null) {
+      vals = this.getValuesFromTexture(dataId);
+    } else {
+      const size = util_exports.sizeFromShape(shape);
+      vals = this.gpgpu.downloadFloat32MatrixFromBuffer(buffer2, size);
+    }
+    if (tmpDownloadTarget != null) {
+      this.disposeIntermediateTensorInfo(tmpDownloadTarget);
+    }
+    if (buffer2 != null) {
+      const gl = this.gpgpu.gl;
+      callAndCheck(gl, () => gl.deleteBuffer(buffer2));
+    }
+    const dTypeVals = this.convertAndCacheOnCPU(dataId, vals);
+    const subscribers = this.pendingRead.get(dataId);
+    this.pendingRead.delete(dataId);
+    subscribers.forEach((resolve) => resolve(dTypeVals));
+    if (this.pendingDisposal.has(dataId)) {
+      this.pendingDisposal.delete(dataId);
+      if (this.disposeData(dataId)) {
+        engine().removeDataId(dataId, this);
+      }
+      this.pendingDeletes--;
+    }
+    return dTypeVals;
+  }
+  readToGPU(dataId, options = {}) {
+    const texData = this.texData.get(dataId);
+    const { values, shape, slice: slice5, dtype, isPacked, texture } = texData;
+    if (dtype === "complex64") {
+      throw new Error("Does not support reading texture for complex64 dtype.");
+    }
+    if (slice5 != null) {
+      let program;
+      if (isPacked) {
+        program = new UnaryOpPackedProgram(shape, CLONE);
+      } else {
+        program = new UnaryOpProgram(shape, CLONE);
+      }
+      const res = this.runWebGLProgram(program, [{ dataId, shape, dtype }], dtype);
+      const gpuResouorce = this.readToGPU(res, options);
+      this.disposeIntermediateTensorInfo(res);
+      return gpuResouorce;
+    }
+    if (texture == null) {
+      if (values != null) {
+        throw new Error("Data is not on GPU but on CPU.");
+      } else {
+        throw new Error("There is no data on GPU or CPU.");
+      }
+    }
+    const tmpTarget = this.decode(dataId, options.customTexShape);
+    const tensorRef = engine().makeTensorFromTensorInfo(tmpTarget);
+    const tmpData = this.texData.get(tmpTarget.dataId);
+    return Object.assign({ tensorRef }, tmpData.texture);
+  }
+  bufferSync(t) {
+    const data = this.readSync(t.dataId);
+    if (t.dtype === "string") {
+      try {
+        const strings = data.map((d) => util_exports.decodeString(d));
+        return buffer(t.shape, t.dtype, strings);
+      } catch (_a) {
+        throw new Error("Failed to decode encoded string bytes into utf-8");
+      }
+    }
+    return buffer(t.shape, t.dtype, data);
+  }
+  checkNumericalProblems(values) {
+    if (values == null) {
+      return;
+    }
+    for (let i = 0; i < values.length; i++) {
+      const num = values[i];
+      if (!canBeRepresented(num)) {
+        if (env().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")) {
+          throw Error(`The value ${num} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`);
+        }
+        throw Error(`The value ${num} cannot be represented on this device.`);
+      }
+    }
+  }
+  getValuesFromTexture(dataId) {
+    const { shape, dtype, isPacked } = this.texData.get(dataId);
+    const size = util_exports.sizeFromShape(shape);
+    if (env().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")) {
+      const tmpTarget = this.decode(dataId);
+      const tmpData2 = this.texData.get(tmpTarget.dataId);
+      const vals2 = this.gpgpu.downloadMatrixFromPackedTexture(tmpData2.texture.texture, ...getDenseTexShape(shape)).subarray(0, size);
+      this.disposeIntermediateTensorInfo(tmpTarget);
+      return vals2;
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK") && isPacked === true;
+    const outputShape = shouldUsePackedProgram ? getShapeAs3D(shape) : shape;
+    const program = shouldUsePackedProgram ? new EncodeFloatPackedProgram(outputShape) : new EncodeFloatProgram(outputShape);
+    const output = this.runWebGLProgram(program, [{ shape: outputShape, dtype, dataId }], "float32");
+    const tmpData = this.texData.get(output.dataId);
+    const vals = this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(tmpData.texture.texture, tmpData.texShape[0], tmpData.texShape[1]).subarray(0, size);
+    this.disposeIntermediateTensorInfo(output);
+    return vals;
+  }
+  timerAvailable() {
+    return env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0;
+  }
+  time(f) {
+    const oldActiveTimers = this.activeTimers;
+    const newActiveTimers = [];
+    let outerMostTime = false;
+    if (this.programTimersStack == null) {
+      this.programTimersStack = newActiveTimers;
+      outerMostTime = true;
+    } else {
+      this.activeTimers.push(newActiveTimers);
+    }
+    this.activeTimers = newActiveTimers;
+    f();
+    const flattenedActiveTimerQueries = util_exports.flatten(this.activeTimers.map((d) => d.query)).filter((d) => d != null);
+    const flattenedActiveTimerNames = util_exports.flatten(this.activeTimers.map((d) => d.name)).filter((d) => d != null);
+    this.activeTimers = oldActiveTimers;
+    if (outerMostTime) {
+      this.programTimersStack = null;
+    }
+    const res = {
+      uploadWaitMs: this.uploadWaitMs,
+      downloadWaitMs: this.downloadWaitMs,
+      kernelMs: null,
+      wallMs: null
+    };
+    return (async () => {
+      if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+        const kernelMs = await Promise.all(flattenedActiveTimerQueries);
+        res["kernelMs"] = util_exports.sum(kernelMs);
+        res["getExtraProfileInfo"] = () => kernelMs.map((d, i) => ({ name: flattenedActiveTimerNames[i], ms: d })).map((d) => `${d.name}: ${d.ms}`).join(", ");
+      } else {
+        res["kernelMs"] = {
+          error: "WebGL query timers are not supported in this environment."
+        };
+      }
+      this.uploadWaitMs = 0;
+      this.downloadWaitMs = 0;
+      return res;
+    })();
+  }
+  memory() {
+    return {
+      unreliable: false,
+      numBytesInGPU: this.numBytesInGPU,
+      numBytesInGPUAllocated: this.textureManager.numBytesAllocated,
+      numBytesInGPUFree: this.textureManager.numBytesFree
+    };
+  }
+  startTimer() {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      return this.gpgpu.beginQuery();
+    }
+    return { startMs: util_exports.now(), endMs: null };
+  }
+  endTimer(query) {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      this.gpgpu.endQuery();
+      return query;
+    }
+    query.endMs = util_exports.now();
+    return query;
+  }
+  async getQueryTime(query) {
+    if (env().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE") > 0) {
+      return this.gpgpu.waitForQueryAndGetTime(query);
+    }
+    const timerQuery = query;
+    return timerQuery.endMs - timerQuery.startMs;
+  }
+  disposeData(dataId, force = false) {
+    if (this.pendingDisposal.has(dataId)) {
+      return false;
+    }
+    if (!this.texData.has(dataId)) {
+      return true;
+    }
+    if (force) {
+      this.texData.get(dataId).refCount = 0;
+    } else {
+      this.texData.get(dataId).refCount--;
+    }
+    if (!force && this.texData.get(dataId).refCount > 0) {
+      return false;
+    }
+    if (this.pendingRead.has(dataId)) {
+      this.pendingDisposal.add(dataId);
+      this.pendingDeletes++;
+      return false;
+    }
+    this.releaseGPUData(dataId);
+    const { complexTensorInfos } = this.texData.get(dataId);
+    if (complexTensorInfos != null) {
+      this.disposeData(complexTensorInfos.real.dataId, force);
+      this.disposeData(complexTensorInfos.imag.dataId, force);
+    }
+    this.texData.delete(dataId);
+    return true;
+  }
+  releaseGPUData(dataId) {
+    const { texture, dtype, texShape, usage, isPacked, slice: slice5 } = this.texData.get(dataId);
+    const key = slice5 && slice5.origDataId || dataId;
+    const refCount = this.dataRefCount.get(key);
+    if (refCount > 1) {
+      this.dataRefCount.set(key, refCount - 1);
+    } else {
+      this.dataRefCount.delete(key);
+      if (texture != null) {
+        this.numBytesInGPU -= this.computeBytes(texShape, dtype);
+        this.textureManager.releaseTexture(texture, texShape, usage, isPacked);
+      }
+    }
+    const texData = this.texData.get(dataId);
+    texData.texture = null;
+    texData.texShape = null;
+    texData.isPacked = false;
+    texData.slice = null;
+  }
+  getTexture(dataId) {
+    this.uploadToGPU(dataId);
+    return this.texData.get(dataId).texture.texture;
+  }
+  getDataInfo(dataId) {
+    return this.texData.get(dataId);
+  }
+  shouldExecuteOnCPU(inputs, sizeThreshold = CPU_HANDOFF_SIZE_THRESHOLD) {
+    return env().getBool("WEBGL_CPU_FORWARD") && inputs.every((input2) => this.texData.get(input2.dataId).texture == null && util_exports.sizeFromShape(input2.shape) < sizeThreshold);
+  }
+  getGPGPUContext() {
+    return this.gpgpu;
+  }
+  where(condition) {
+    backend_util_exports.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");
+    const condVals = condition.dataSync();
+    return whereImpl3(condition.shape, condVals);
+  }
+  packedUnaryOp(x, op2, dtype) {
+    const program = new UnaryOpPackedProgram(x.shape, op2);
+    const outInfo = this.compileAndRun(program, [x], dtype);
+    return engine().makeTensorFromTensorInfo(outInfo);
+  }
+  abs(x) {
+    if (this.shouldExecuteOnCPU([x]) && x.dtype !== "complex64") {
+      const outValues = simpleAbsImplCPU(this.texData.get(x.dataId).values);
+      return this.makeOutput(x.shape, x.dtype, outValues);
+    }
+    if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+      return this.packedUnaryOp(x, ABS, x.dtype);
+    }
+    const program = new UnaryOpProgram(x.shape, ABS);
+    const outInfo = this.compileAndRun(program, [x]);
+    return engine().makeTensorFromTensorInfo(outInfo);
+  }
+  makeTensorInfo(shape, dtype, values) {
+    let dataId;
+    if (dtype === "string" && values != null && values.length > 0 && util_exports.isString(values[0])) {
+      const encodedValues = values.map((d) => util_exports.encodeString(d));
+      dataId = this.write(encodedValues, shape, dtype);
+    } else {
+      dataId = this.write(values, shape, dtype);
+    }
+    this.texData.get(dataId).usage = null;
+    return { dataId, shape, dtype };
+  }
+  makeOutput(shape, dtype, values) {
+    return engine().makeTensorFromTensorInfo(this.makeTensorInfo(shape, dtype, values), this);
+  }
+  unpackTensor(input2) {
+    const program = new UnpackProgram(input2.shape);
+    return this.runWebGLProgram(program, [input2], input2.dtype);
+  }
+  packTensor(input2) {
+    const program = new PackProgram(input2.shape);
+    const preventEagerUnpackingOutput = true;
+    return this.runWebGLProgram(program, [input2], input2.dtype, null, preventEagerUnpackingOutput);
+  }
+  packedReshape(input2, afterShape) {
+    const input3DShape = [
+      getBatchDim(input2.shape),
+      ...getRowsCols(input2.shape)
+    ];
+    const input3D = {
+      dtype: input2.dtype,
+      shape: input3DShape,
+      dataId: input2.dataId
+    };
+    const afterShapeAs3D = [
+      getBatchDim(afterShape),
+      ...getRowsCols(afterShape)
+    ];
+    const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);
+    const preventEagerUnpackingOfOutput = true;
+    const customValues = [input3DShape];
+    const output = this.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);
+    return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };
+  }
+  decode(dataId, customTexShape) {
+    const texData = this.texData.get(dataId);
+    const { isPacked, shape, dtype } = texData;
+    if (customTexShape != null) {
+      const size = util_exports.sizeFromShape(shape);
+      const texSize = customTexShape[0] * customTexShape[1] * 4;
+      util_exports.assert(size <= texSize, () => "customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.");
+    }
+    const shapeAs3D = getShapeAs3D(shape);
+    let program;
+    if (isPacked) {
+      program = new DecodeMatrixPackedProgram(shapeAs3D);
+    } else {
+      program = new DecodeMatrixProgram(shapeAs3D);
+    }
+    const preventEagerUnpackingOfOutput = true;
+    const customValues = [customTexShape != null ? customTexShape : getDenseTexShape(shapeAs3D)];
+    const out = this.runWebGLProgram(program, [{ shape: shapeAs3D, dtype, dataId }], dtype, customValues, preventEagerUnpackingOfOutput, customTexShape);
+    return { dtype, shape, dataId: out.dataId };
+  }
+  runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false, customTexShape) {
+    const output = this.makeTensorInfo(program.outputShape, outputDtype);
+    const outData = this.texData.get(output.dataId);
+    if (program.packedOutput) {
+      outData.isPacked = true;
+    }
+    if (program.outPackingScheme === PackingScheme.DENSE) {
+      const texelShape = customTexShape != null ? customTexShape : getDenseTexShape(program.outputShape);
+      outData.texShape = texelShape.map((d) => d * 2);
+    }
+    if (program.outTexUsage != null) {
+      outData.usage = program.outTexUsage;
+    }
+    if (util_exports.sizeFromShape(output.shape) === 0) {
+      outData.values = util_exports.getTypedArrayFromDType(output.dtype, 0);
+      return output;
+    }
+    const dataToDispose = [];
+    const inputsData = inputs.map((input2) => {
+      if (input2.dtype === "complex64") {
+        throw new Error(`GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.`);
+      }
+      let texData = this.texData.get(input2.dataId);
+      if (texData.texture == null) {
+        if (!program.packedInputs && util_exports.sizeFromShape(input2.shape) <= env().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM")) {
+          return {
+            shape: input2.shape,
+            texData: null,
+            isUniform: true,
+            uniformValues: texData.values
+          };
+        }
+        if (program.packedInputs) {
+          texData.isPacked = true;
+          texData.shape = input2.shape;
+        }
+      }
+      this.uploadToGPU(input2.dataId);
+      if (!!texData.isPacked !== !!program.packedInputs) {
+        input2 = texData.isPacked ? this.unpackTensor(input2) : this.packTensor(input2);
+        dataToDispose.push(input2);
+        texData = this.texData.get(input2.dataId);
+      } else if (texData.isPacked && !isReshapeFree(texData.shape, input2.shape)) {
+        const savedInput = input2;
+        const targetShape = input2.shape;
+        input2.shape = texData.shape;
+        input2 = this.packedReshape(input2, targetShape);
+        dataToDispose.push(input2);
+        texData = this.texData.get(input2.dataId);
+        savedInput.shape = targetShape;
+      }
+      return { shape: input2.shape, texData, isUniform: false };
+    });
+    this.uploadToGPU(output.dataId);
+    const outputData = { shape: output.shape, texData: outData, isUniform: false };
+    const key = makeShaderKey(program, inputsData, outputData);
+    const binary = this.getAndSaveBinary(key, () => {
+      return compileProgram(this.gpgpu, program, inputsData, outputData);
+    });
+    const shouldTimeProgram = this.activeTimers != null;
+    let query;
+    if (shouldTimeProgram) {
+      query = this.startTimer();
+    }
+    if (!env().get("ENGINE_COMPILE_ONLY")) {
+      runProgram(this.gpgpu, binary, inputsData, outputData, customUniformValues);
+    }
+    dataToDispose.forEach((info) => this.disposeIntermediateTensorInfo(info));
+    if (shouldTimeProgram) {
+      query = this.endTimer(query);
+      this.activeTimers.push({ name: program.constructor.name, query: this.getQueryTime(query) });
+    }
+    const glFlushThreshold = env().get("WEBGL_FLUSH_THRESHOLD");
+    if (glFlushThreshold > 0) {
+      const time2 = util_exports.now();
+      if (time2 - this.lastGlFlushTime > glFlushThreshold) {
+        this.gpgpu.gl.flush();
+        this.lastGlFlushTime = time2;
+      }
+    }
+    if (!env().getBool("WEBGL_LAZILY_UNPACK") && outData.isPacked && preventEagerUnpackingOfOutput === false) {
+      const unpacked = this.unpackTensor(output);
+      this.disposeIntermediateTensorInfo(output);
+      return unpacked;
+    }
+    return output;
+  }
+  compileAndRun(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput = false) {
+    outputDtype = outputDtype || inputs[0].dtype;
+    const outInfo = this.runWebGLProgram(program, inputs, outputDtype, customUniformValues, preventEagerUnpackingOfOutput);
+    return outInfo;
+  }
+  getAndSaveBinary(key, getBinary) {
+    if (!(key in this.binaryCache)) {
+      this.binaryCache[key] = getBinary();
+    }
+    return this.binaryCache[key];
+  }
+  getTextureManager() {
+    return this.textureManager;
+  }
+  dispose() {
+    if (this.disposed) {
+      return;
+    }
+    if (!env().getBool("IS_TEST")) {
+      const allKeys = Object.keys(this.binaryCache);
+      allKeys.forEach((key) => {
+        this.gpgpu.deleteProgram(this.binaryCache[key].webGLProgram);
+        delete this.binaryCache[key];
+      });
+    }
+    this.textureManager.dispose();
+    if (this.canvas != null && (typeof HTMLCanvasElement !== "undefined" && this.canvas instanceof HTMLCanvasElement)) {
+      this.canvas.remove();
+    } else {
+      this.canvas = null;
+    }
+    if (this.gpgpuCreatedLocally) {
+      this.gpgpu.program = null;
+      this.gpgpu.dispose();
+    }
+    this.disposed = true;
+  }
+  floatPrecision() {
+    if (this.floatPrecisionValue == null) {
+      this.floatPrecisionValue = tidy(() => {
+        if (!env().get("WEBGL_RENDER_FLOAT32_ENABLED")) {
+          const debugFlag = env().getBool("DEBUG");
+          env().set("DEBUG", false);
+          const underflowCheckValue = this.abs(scalar(1e-8)).dataSync()[0];
+          env().set("DEBUG", debugFlag);
+          if (underflowCheckValue > 0) {
+            return 32;
+          }
+        }
+        return 16;
+      });
+    }
+    return this.floatPrecisionValue;
+  }
+  epsilon() {
+    return this.floatPrecision() === 32 ? EPSILON_FLOAT322 : EPSILON_FLOAT162;
+  }
+  uploadToGPU(dataId) {
+    const texData = this.texData.get(dataId);
+    const { shape, dtype, values, texture, usage, isPacked } = texData;
+    if (texture != null) {
+      return;
+    }
+    const shouldTimeProgram = this.activeTimers != null;
+    let start;
+    if (shouldTimeProgram) {
+      start = util_exports.now();
+    }
+    let texShape = texData.texShape;
+    if (texShape == null) {
+      texShape = getTextureShapeFromLogicalShape(shape, isPacked);
+      texData.texShape = texShape;
+    }
+    if (values != null) {
+      const shapeAs3D = getShapeAs3D(shape);
+      let program;
+      let width = texShape[1], height = texShape[0];
+      const isByteArray = values instanceof Uint8Array || values instanceof Uint8ClampedArray;
+      if (isPacked || !isByteArray) {
+        [width, height] = getPackedMatrixTextureShapeWidthHeight(texShape[0], texShape[1]);
+      }
+      if (isPacked) {
+        program = new EncodeMatrixPackedProgram(shapeAs3D, isByteArray);
+      } else {
+        program = new EncodeMatrixProgram(shapeAs3D, isByteArray);
+      }
+      const tempDenseInputTexShape = isByteArray ? [height, width] : texShape;
+      const tempDenseInputHandle = this.makeTensorInfo(tempDenseInputTexShape, dtype);
+      const tempDenseInputTexData = this.texData.get(tempDenseInputHandle.dataId);
+      if (isByteArray) {
+        tempDenseInputTexData.usage = TextureUsage.PIXELS;
+      } else {
+        tempDenseInputTexData.usage = TextureUsage.UPLOAD;
+      }
+      tempDenseInputTexData.texShape = tempDenseInputTexShape;
+      this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(tempDenseInputHandle.dataId), width, height, values);
+      const customValues = [[height, width]];
+      const preventEagerUnpacking = true;
+      const encodedOutputTarget = this.runWebGLProgram(program, [tempDenseInputHandle], dtype, customValues, preventEagerUnpacking);
+      const outputTexData = this.texData.get(encodedOutputTarget.dataId);
+      texData.texShape = outputTexData.texShape;
+      texData.isPacked = outputTexData.isPacked;
+      texData.usage = outputTexData.usage;
+      if (!env().get("ENGINE_COMPILE_ONLY")) {
+        texData.texture = outputTexData.texture;
+        texData.values = null;
+        this.texData.delete(encodedOutputTarget.dataId);
+      } else {
+        this.disposeData(encodedOutputTarget.dataId);
+      }
+      this.disposeIntermediateTensorInfo(tempDenseInputHandle);
+      if (shouldTimeProgram) {
+        this.uploadWaitMs += util_exports.now() - start;
+      }
+    } else {
+      const newTexture = this.acquireTexture(texShape, usage, dtype, isPacked);
+      texData.texture = newTexture;
+    }
+  }
+  convertAndCacheOnCPU(dataId, float32Values) {
+    const texData = this.texData.get(dataId);
+    const { dtype } = texData;
+    this.releaseGPUData(dataId);
+    if (float32Values != null) {
+      texData.values = float32ToTypedArray(float32Values, dtype);
+    }
+    return texData.values;
+  }
+  acquireTexture(texShape, texType, dtype, isPacked) {
+    this.numBytesInGPU += this.computeBytes(texShape, dtype);
+    if (!this.warnedAboutMemory && this.numBytesInGPU > this.numMBBeforeWarning * 1024 * 1024) {
+      const mb = (this.numBytesInGPU / 1024 / 1024).toFixed(2);
+      this.warnedAboutMemory = true;
+      console.warn(`High memory usage in GPU: ${mb} MB, most likely due to a memory leak`);
+    }
+    return this.textureManager.acquireTexture(texShape, texType, isPacked);
+  }
+  computeBytes(shape, dtype) {
+    return shape[0] * shape[1] * util_exports.bytesPerElement(dtype);
+  }
+  checkCompileCompletion() {
+    for (const [, binary] of Object.entries(this.binaryCache)) {
+      this.checkCompletion_(binary);
+    }
+  }
+  async checkCompileCompletionAsync() {
+    const ps = [];
+    if (this.gpgpu.parallelCompilationExtension) {
+      for (const [, binary] of Object.entries(this.binaryCache)) {
+        ps.push(this.checkCompletionAsync_(binary));
+      }
+      return Promise.all(ps);
+    } else {
+      for (const [, binary] of Object.entries(this.binaryCache)) {
+        const p2 = new Promise((resolve) => {
+          try {
+            this.checkCompletion_(binary);
+            resolve(true);
+          } catch (error) {
+            throw error;
+          }
+        });
+        ps.push(p2);
+      }
+      return Promise.all(ps);
+    }
+  }
+  async checkCompletionAsync_(binary) {
+    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)) {
+      return this.checkCompletion_(binary);
+    } else {
+      await nextFrame();
+      return this.checkCompletionAsync_(binary);
+    }
+  }
+  checkCompletion_(binary) {
+    if (this.gpgpu.gl.getProgramParameter(binary.webGLProgram, this.gpgpu.gl.LINK_STATUS) === false) {
+      console.log(this.gpgpu.gl.getProgramInfoLog(binary.webGLProgram));
+      if (this.gpgpu.gl.getShaderParameter(binary.fragmentShader, this.gpgpu.gl.COMPILE_STATUS) === false) {
+        logShaderSourceAndInfoLog(binary.source, this.gpgpu.gl.getShaderInfoLog(binary.fragmentShader));
+        throw new Error("Failed to compile fragment shader.");
+      }
+      throw new Error("Failed to link vertex and fragment shaders.");
+    }
+    return true;
+  }
+  getUniformLocations() {
+    for (const [, binary] of Object.entries(this.binaryCache)) {
+      const { uniformLocations, customUniformLocations, infLoc, nanLoc, inShapesLocations, inTexShapesLocations, outShapeLocation, outShapeStridesLocation, outTexShapeLocation } = getUniformLocations(this.gpgpu, binary.program, binary.webGLProgram);
+      binary.uniformLocations = uniformLocations;
+      binary.customUniformLocations = customUniformLocations;
+      binary.infLoc = infLoc;
+      binary.nanLoc = nanLoc;
+      binary.inShapesLocations = inShapesLocations;
+      binary.inTexShapesLocations = inTexShapesLocations;
+      binary.outShapeLocation = outShapeLocation;
+      binary.outShapeStridesLocation = outShapeStridesLocation;
+      binary.outTexShapeLocation = outTexShapeLocation;
+    }
+  }
+  createTensorFromTexture(values, shape, dtype) {
+    const { texture, height, width, channels } = values;
+    const backend2 = engine().backend;
+    if (!backend2.gpgpu.gl.isTexture(texture)) {
+      throw new Error(`The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.`);
+    }
+    const dataId = backend2.writeTexture(texture, shape, dtype, height, width, channels);
+    return engine().makeTensorFromDataId(dataId, shape, dtype, backend2);
+  }
+};
+MathBackendWebGL.nextDataId = 0;
+function float32ToTypedArray(a, dtype) {
+  if (dtype === "float32" || dtype === "complex64") {
+    return a;
+  } else if (dtype === "int32" || dtype === "bool") {
+    const result = dtype === "int32" ? new Int32Array(a.length) : new Uint8Array(a.length);
+    for (let i = 0; i < result.length; ++i) {
+      result[i] = Math.round(a[i]);
+    }
+    return result;
+  } else {
+    throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/version.js
+var version6 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/webgl.js
+function forceHalfFloat() {
+  env().set("WEBGL_FORCE_F16_TEXTURES", true);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/base.js
+if (device_util_exports.isBrowser()) {
+  registerBackend("webgl", () => new MathBackendWebGL(), 2);
+}
+var webgl = { forceHalfFloat };
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_gpu.js
+var CHECK_NAN_SNIPPET2 = `
   if (isnan(a)) return a;
   if (isnan(b)) return b;
-`;var io=class{constructor(t,e,n){this.variableNames=["A","B"],this.outputShape=v.assertAndGetBroadcastShape(e,n),this.enableShapeUniforms=we(this.outputShape.length),this.userCode=`
+`;
+var BinaryOpProgram = class {
+  constructor(op2, aShape, bShape) {
+    this.variableNames = ["A", "B"];
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    this.userCode = `
       float binaryOperation(float a, float b) {
-        ${t}
+        ${op2}
       }
 
       void main() {
@@ -1250,44 +53736,82 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float b = getBAtOutCoords();
         setOutput(binaryOperation(a, b));
       }
-    `}};var Yi=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_packed_gpu.js
+var CHECK_NAN_SNIPPET_PACKED = `
   result.r = isNaN.r ? NAN : result.r;
   result.g = isNaN.g ? NAN : result.g;
   result.b = isNaN.b ? NAN : result.b;
   result.a = isNaN.a ? NAN : result.a;
-`;var Oo=class{constructor(t,e,n,o=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=v.assertAndGetBroadcastShape(e,n);let s=this.outputShape.length;this.enableShapeUniforms=we(s);let i="";if(o)if(s===0||y.sizeFromShape(this.outputShape)===1)i=`
+`;
+var BinaryOpPackedProgram = class {
+  constructor(op2, aShape, bShape, checkOutOfBounds = false) {
+    this.variableNames = ["A", "B"];
+    this.supportsBroadcasting = true;
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    const rank = this.outputShape.length;
+    this.enableShapeUniforms = useShapeUniforms(rank);
+    let checkOutOfBoundsString = "";
+    if (checkOutOfBounds) {
+      if (rank === 0 || util_exports.sizeFromShape(this.outputShape) === 1) {
+        checkOutOfBoundsString = `
           result.y = 0.;
           result.z = 0.;
           result.w = 0.;
-        `;else if(i=`
-          ${zt(s)} coords = getOutputCoords();
-        `,s===1)this.enableShapeUniforms?i+=`
+        `;
+      } else {
+        const dtype = getCoordsDataType(rank);
+        checkOutOfBoundsString = `
+          ${dtype} coords = getOutputCoords();
+        `;
+        if (rank === 1) {
+          if (this.enableShapeUniforms) {
+            checkOutOfBoundsString += `
             result.y = (coords + 1) >= outShape ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `:i+=`
+          `;
+          } else {
+            checkOutOfBoundsString += `
             result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;
             result.z = 0.;
             result.w = 0.;
-          `;else{let u=Qe("coords",s);this.enableShapeUniforms?i+=`
+          `;
+          }
+        } else {
+          const channels = getChannels("coords", rank);
+          if (this.enableShapeUniforms) {
+            checkOutOfBoundsString += `
             bool nextRowOutOfBounds =
-              (${u[s-2]} + 1) >= outShape[${s} - 2];
+              (${channels[rank - 2]} + 1) >= outShape[${rank} - 2];
             bool nextColOutOfBounds =
-              (${u[s-1]} + 1) >= outShape[${s} - 1];
+              (${channels[rank - 1]} + 1) >= outShape[${rank} - 1];
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
-          `:i+=`
+          `;
+          } else {
+            checkOutOfBoundsString += `
             bool nextRowOutOfBounds =
-              (${u[s-2]} + 1) >= ${this.outputShape[s-2]};
+              (${channels[rank - 2]} + 1) >= ${this.outputShape[rank - 2]};
             bool nextColOutOfBounds =
-              (${u[s-1]} + 1) >= ${this.outputShape[s-1]};
+              (${channels[rank - 1]} + 1) >= ${this.outputShape[rank - 1]};
             result.y = nextColOutOfBounds ? 0. : result.y;
             result.z = nextRowOutOfBounds ? 0. : result.z;
             result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
-          `}this.userCode=`
+          `;
+          }
+        }
+      }
+    }
+    this.userCode = `
       vec4 binaryOperation(vec4 a, vec4 b) {
-        ${t}
+        ${op2}
       }
 
       void main() {
@@ -1295,41 +53819,265 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         vec4 b = getBAtOutCoords();
 
         vec4 result = binaryOperation(a, b);
-        ${i}
+        ${checkOutOfBoundsString}
 
         setOutput(result);
       }
-    `}};function tr(r){let{inputs:t,backend:e}=r,{x:n}=t;return e.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var gM={kernelName:co,backendName:"webgl",kernelFunc:tr};function En(r){let{inputs:t,backend:e}=r,{real:n,imag:o}=t,s=e.makeTensorInfo(n.shape,"complex64"),i=e.texData.get(s.dataId),a=tr({inputs:{x:n},backend:e}),u=tr({inputs:{x:o},backend:e});return i.complexTensorInfos={real:a,imag:u},s}var xM={kernelName:pp,backendName:"webgl",kernelFunc:En};var uk="return (a < 0.) ? b * a : a;",ck=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Identity.js
+function identity3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  backend2.incRef(x.dataId);
+  return { dataId: x.dataId, shape: x.shape, dtype: x.dtype };
+}
+var identityConfig2 = {
+  kernelName: Identity,
+  backendName: "webgl",
+  kernelFunc: identity3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Complex.js
+function complex3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { real: real4, imag: imag4 } = inputs;
+  const complexInfo = backend2.makeTensorInfo(real4.shape, "complex64");
+  const complex4 = backend2.texData.get(complexInfo.dataId);
+  const realTensorInfo = identity3({ inputs: { x: real4 }, backend: backend2 });
+  const imagTensorInfo = identity3({ inputs: { x: imag4 }, backend: backend2 });
+  complex4.complexTensorInfos = { real: realTensorInfo, imag: imagTensorInfo };
+  return complexInfo;
+}
+var complexConfig2 = {
+  kernelName: Complex,
+  backendName: "webgl",
+  kernelFunc: complex3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LeakyRelu.js
+var LEAKYRELU = `return (a < 0.) ? b * a : a;`;
+var LEAKYRELU_PACKED = `
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function get(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{alpha:s}=n,i=e.makeTensorInfo([],"float32",y.createScalarValue(s,"float32")),a=z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Oo(ck,o.shape,i.shape):new io(uk,o.shape,i.shape),u=e.runWebGLProgram(a,[o,i],"float32");return e.disposeIntermediateTensorInfo(i),u}var yM={kernelName:is,backendName:"webgl",kernelFunc:get};var pk="return (a < 0.) ? b * a : a;",mk=`
+`;
+function leakyRelu3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { alpha } = attrs;
+  const $alpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(alpha, "float32"));
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(LEAKYRELU_PACKED, x.shape, $alpha.shape) : new BinaryOpProgram(LEAKYRELU, x.shape, $alpha.shape);
+  const result = backend2.runWebGLProgram(program, [x, $alpha], "float32");
+  backend2.disposeIntermediateTensorInfo($alpha);
+  return result;
+}
+var leakyReluConfig2 = {
+  kernelName: LeakyRelu,
+  backendName: "webgl",
+  kernelFunc: leakyRelu3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Prelu.js
+var PRELU = `return (a < 0.) ? b * a : a;`;
+var PRELU_PACKED = `
   vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
   return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
-`;function xet(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t,s=z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Oo(mk,n.shape,o.shape):new io(pk,n.shape,o.shape);return e.runWebGLProgram(s,[n,o],"float32")}var bM={kernelName:bs,backendName:"webgl",kernelFunc:xet};var Po="if (isnan(x)) return x;";function Ct({opSnippet:r,packedOpSnippet:t,cpuKernelImpl:e,dtype:n}){return({inputs:o,backend:s})=>{let{x:i}=o,a=s,u=n||i.dtype;if(a.shouldExecuteOnCPU([i])&&e!=null){let p=a.texData.get(i.dataId),m=e(p.values,u);return a.makeTensorInfo(i.shape,u,m)}let l=z().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&t!=null,c;return l?c=new so(i.shape,t):c=new tn(i.shape,r),a.runWebGLProgram(c,[i],u)}}function le({opSnippet:r,packedOpSnippet:t,checkOutOfBounds:e=!1,supportsComplex:n=!1,cpuKernelImpl:o,dtype:s}){return({inputs:i,backend:a})=>{let{a:u,b:l}=i,c=a;if(n&&u.dtype==="complex64"){let d=c.texData.get(u.dataId),h=c.texData.get(l.dataId),[g,x]=[[d.complexTensorInfos.real,h.complexTensorInfos.real],[d.complexTensorInfos.imag,h.complexTensorInfos.imag]].map(w=>{let[C,N]=w,_={dataId:C.dataId,dtype:C.dtype,shape:u.shape},A={dataId:N.dataId,dtype:N.dtype,shape:l.shape},$=new io(r,u.shape,l.shape);return c.runWebGLProgram($,[_,A],sr(C.dtype,N.dtype))}),b=En({inputs:{real:g,imag:x},backend:c});return c.disposeIntermediateTensorInfo(g),c.disposeIntermediateTensorInfo(x),b}let p=s||sr(u.dtype,l.dtype);if((u.dtype==="string"||l.dtype==="string"||c.shouldExecuteOnCPU([u,l]))&&o!=null){let d=c.texData.get(u.dataId).values,h=c.texData.get(l.dataId).values,g=u.dtype==="string"?v.fromUint8ToStringArray(d):d,x=u.dtype==="string"?v.fromUint8ToStringArray(h):h,[b,w]=o(u.shape,l.shape,g,x,p),C=c.makeTensorInfo(w,p),N=c.texData.get(C.dataId);return N.values=b,C}let m=z().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&t!=null,f;return m?f=new Oo(t,u.shape,l.shape,e):f=new io(r,u.shape,l.shape),c.runWebGLProgram(f,[u,l],p)}}function bl(r,t=!1){if(r==="linear")return t?uM:nM;if(r==="relu")return t?pM:sM;if(r==="elu")return t?cM:oM;if(r==="relu6")return t?mM:iM;if(r==="prelu")return t?mk:pk;if(r==="leakyrelu")return t?ck:uk;if(r==="sigmoid")return t?fM:aM;throw new Error(`Activation ${r} has not been implemented for the WebGL backend.`)}var vd=class{constructor(t,e,n,o=!1,s=!1,i=!1,a=null,u=!1,l=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=we(this.outputShape.length);let c=o?t[1]:t[2],p=Math.ceil(c/2),m=o?"i * 2, rc.y":"rc.y, i * 2",f=s?"rc.z, i * 2":"i * 2, rc.z",d=o?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],h=s?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],g="",x="";a&&(u?g=`vec4 activation(vec4 a) {
+`;
+function prelu4(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(PRELU_PACKED, x.shape, alpha.shape) : new BinaryOpProgram(PRELU, x.shape, alpha.shape);
+  return backend2.runWebGLProgram(program, [x, alpha], "float32");
+}
+var preluConfig2 = {
+  kernelName: Prelu,
+  backendName: "webgl",
+  kernelFunc: prelu4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/kernel_funcs_utils.js
+var CHECK_NAN_SNIPPET_UNARY = `if (isnan(x)) return x;`;
+function unaryKernelFunc2({ opSnippet, packedOpSnippet, cpuKernelImpl, dtype }) {
+  return ({ inputs, backend: backend2 }) => {
+    const { x } = inputs;
+    const webglBackend = backend2;
+    const $dtype = dtype || x.dtype;
+    if (webglBackend.shouldExecuteOnCPU([x]) && cpuKernelImpl != null) {
+      const xData = webglBackend.texData.get(x.dataId);
+      const outValues = cpuKernelImpl(xData.values, $dtype);
+      return webglBackend.makeTensorInfo(x.shape, $dtype, outValues);
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK_UNARY_OPERATIONS") && packedOpSnippet != null;
+    let program;
+    if (shouldUsePackedProgram) {
+      program = new UnaryOpPackedProgram(x.shape, packedOpSnippet);
+    } else {
+      program = new UnaryOpProgram(x.shape, opSnippet);
+    }
+    return webglBackend.runWebGLProgram(program, [x], $dtype);
+  };
+}
+function binaryKernelFunc2({ opSnippet, packedOpSnippet, checkOutOfBounds = false, supportsComplex = false, cpuKernelImpl, dtype }) {
+  return ({ inputs, backend: backend2 }) => {
+    const { a, b } = inputs;
+    const webglBackend = backend2;
+    if (supportsComplex && a.dtype === "complex64") {
+      const aData = webglBackend.texData.get(a.dataId);
+      const bData = webglBackend.texData.get(b.dataId);
+      const [real4, imag4] = [
+        [aData.complexTensorInfos.real, bData.complexTensorInfos.real],
+        [aData.complexTensorInfos.imag, bData.complexTensorInfos.imag]
+      ].map((complexParts) => {
+        const [aPart, bPart] = complexParts;
+        const aHandle = {
+          dataId: aPart.dataId,
+          dtype: aPart.dtype,
+          shape: a.shape
+        };
+        const bHandle = {
+          dataId: bPart.dataId,
+          dtype: bPart.dtype,
+          shape: b.shape
+        };
+        const program2 = new BinaryOpProgram(opSnippet, a.shape, b.shape);
+        return webglBackend.runWebGLProgram(program2, [aHandle, bHandle], upcastType(aPart.dtype, bPart.dtype));
+      });
+      const complexOutput = complex3({ inputs: { real: real4, imag: imag4 }, backend: webglBackend });
+      webglBackend.disposeIntermediateTensorInfo(real4);
+      webglBackend.disposeIntermediateTensorInfo(imag4);
+      return complexOutput;
+    }
+    const $dtype = dtype || upcastType(a.dtype, b.dtype);
+    if ((a.dtype === "string" || b.dtype === "string" || webglBackend.shouldExecuteOnCPU([a, b])) && cpuKernelImpl != null) {
+      const aVals = webglBackend.texData.get(a.dataId).values;
+      const bVals = webglBackend.texData.get(b.dataId).values;
+      const decodedAVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(aVals) : aVals;
+      const decodedBVals = a.dtype === "string" ? backend_util_exports.fromUint8ToStringArray(bVals) : bVals;
+      const [outValues, outShape] = cpuKernelImpl(a.shape, b.shape, decodedAVals, decodedBVals, $dtype);
+      const out = webglBackend.makeTensorInfo(outShape, $dtype);
+      const outData = webglBackend.texData.get(out.dataId);
+      outData.values = outValues;
+      return out;
+    }
+    const shouldUsePackedProgram = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") && packedOpSnippet != null;
+    let program;
+    if (shouldUsePackedProgram) {
+      program = new BinaryOpPackedProgram(packedOpSnippet, a.shape, b.shape, checkOutOfBounds);
+    } else {
+      program = new BinaryOpProgram(opSnippet, a.shape, b.shape);
+    }
+    return webglBackend.runWebGLProgram(program, [a, b], $dtype);
+  };
+}
+function mapActivationToShaderProgram(activation2, packed = false) {
+  if (activation2 === "linear") {
+    if (packed) {
+      return LINEAR2;
+    }
+    return LINEAR;
+  } else if (activation2 === "relu") {
+    if (packed) {
+      return RELU2;
+    }
+    return RELU;
+  } else if (activation2 === "elu") {
+    if (packed) {
+      return ELU3;
+    }
+    return ELU2;
+  } else if (activation2 === "relu6") {
+    if (packed) {
+      return RELU62;
+    }
+    return RELU6;
+  } else if (activation2 === "prelu") {
+    if (packed) {
+      return PRELU_PACKED;
+    }
+    return PRELU;
+  } else if (activation2 === "leakyrelu") {
+    if (packed) {
+      return LEAKYRELU_PACKED;
+    }
+    return LEAKYRELU;
+  } else if (activation2 === "sigmoid") {
+    if (packed) {
+      return SIGMOID2;
+    }
+    return SIGMOID;
+  }
+  throw new Error(`Activation ${activation2} has not been implemented for the WebGL backend.`);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/mulmat_packed_gpu.js
+var MatMulPackedProgram = class {
+  constructor(aShape, bShape, outputShape, transposeA = false, transposeB = false, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyreluActivation = false) {
+    this.variableNames = ["matrixA", "matrixB"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const sharedDim = transposeA ? aShape[1] : aShape[2];
+    const sharedDimensionPacked = Math.ceil(sharedDim / 2);
+    const aSample = transposeA ? "i * 2, rc.y" : "rc.y, i * 2";
+    const bSample = transposeB ? "rc.z, i * 2" : "i * 2, rc.z";
+    const aSwizzle = transposeA ? ["a.xxyy", "a.zzww"] : ["a.xxzz", "a.yyww"];
+    const bSwizzle = transposeB ? ["b.xzxz", "b.ywyw"] : ["b.xyxy", "b.zwzw"];
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
-          ${a}
-        }`:l?g=`vec4 activation(vec4 a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
-          ${a}
-        }`:g=`vec4 activation(vec4 x) {
-          ${a}
-        }`,x="result = activation(result);");let b=i?"result += getBiasAtOutCoords();":"";i&&this.variableNames.push("bias"),u&&this.variableNames.push("preluActivationWeights"),l&&this.variableNames.push("leakyreluAlpha");let w="rc.x",C="rc.x";t[0]<e[0]?w=`int(min(float(rc.x), ${t[0]-1}.))`:e[0]<t[0]&&(C=`int(min(float(rc.x), ${e[0]-1}.))`),this.userCode=`
-      ${g}
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+          ${activation2}
+        }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyreluActivation) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    let batchASnippet = "rc.x";
+    let batchBSnippet = "rc.x";
+    if (aShape[0] < bShape[0]) {
+      batchASnippet = `int(min(float(rc.x), ${aShape[0] - 1}.))`;
+    } else if (bShape[0] < aShape[0]) {
+      batchBSnippet = `int(min(float(rc.x), ${bShape[0] - 1}.))`;
+    }
+    this.userCode = `
+      ${activationSnippet}
       // Don't use uniform for sharedDimensionPacked for performance.
-      const float sharedDimension = ${p}.0;
+      const float sharedDimension = ${sharedDimensionPacked}.0;
 
       vec4 dot2x2ARowBCol(ivec3 rc) {
         vec4 result = vec4(0);
-        for (int i = 0; i < ${p}; i++) {
-          int batchA = ${w};
-          int batchB = ${C};
-          vec4 a = getMatrixA(batchA, ${m});
-          vec4 b = getMatrixB(batchB, ${f});
+        for (int i = 0; i < ${sharedDimensionPacked}; i++) {
+          int batchA = ${batchASnippet};
+          int batchB = ${batchBSnippet};
+          vec4 a = getMatrixA(batchA, ${aSample});
+          vec4 b = getMatrixB(batchB, ${bSample});
 
           // These swizzled products need to be separately added.
           // See: https://github.com/tensorflow/tfjs/issues/1735
-          result += (${d[0]} * ${h[0]});
-          result += (${d[1]} * ${h[1]});
+          result += (${aSwizzle[0]} * ${bSwizzle[0]});
+          result += (${aSwizzle[1]} * ${bSwizzle[1]});
         }
         return result;
       }
@@ -1338,16 +54086,29 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec3 rc = getOutputCoords();
         vec4 result = dot2x2ARowBCol(rc);
 
-        ${b}
+        ${addBiasSnippet}
 
-        ${x}
+        ${applyActivationSnippet}
 
         setOutput(result);
       }
-    `}};var fk={REAL:"return areal * breal - aimag * bimag;",IMAG:"return areal * bimag + aimag * breal;"},tg=class{constructor(t,e,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=v.assertAndGetBroadcastShape(e,n),this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/binaryop_complex_gpu.js
+var COMPLEX_MULTIPLY = {
+  REAL: "return areal * breal - aimag * bimag;",
+  IMAG: "return areal * bimag + aimag * breal;"
+};
+var BinaryOpComplexProgram = class {
+  constructor(op2, aShape, bShape) {
+    this.variableNames = ["AReal", "AImag", "BReal", "BImag"];
+    this.outputShape = backend_util_exports.assertAndGetBroadcastShape(aShape, bShape);
+    this.userCode = `
       float binaryOpComplex(
           float areal, float aimag, float breal, float bimag) {
-        ${t}
+        ${op2}
       }
 
       void main() {
@@ -1357,15 +54118,144 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         float bimag = getBImagAtOutCoords();
         setOutput(binaryOpComplex(areal, aimag, breal, bimag));
       }
-    `}};var wM="return a * b;";function eg(r){let{inputs:t,backend:e}=r,{a:n,b:o}=t,s=v.upcastType(n.dtype,o.dtype);if(n.dtype==="complex64"){let a=e.texData.get(n.dataId),u=e.texData.get(o.dataId),l=new tg(fk.REAL,n.shape,o.shape),c=new tg(fk.IMAG,n.shape,o.shape),p=[{dataId:a.complexTensorInfos.real.dataId,dtype:a.complexTensorInfos.real.dtype,shape:n.shape},{dataId:a.complexTensorInfos.imag.dataId,dtype:a.complexTensorInfos.imag.dtype,shape:n.shape},{dataId:u.complexTensorInfos.real.dataId,dtype:u.complexTensorInfos.real.dtype,shape:o.shape},{dataId:u.complexTensorInfos.imag.dataId,dtype:u.complexTensorInfos.imag.dtype,shape:o.shape}],m=e.runWebGLProgram(l,p,"float32"),f=e.runWebGLProgram(c,p,"float32"),d=En({inputs:{real:m,imag:f},backend:e});return e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),d}if(e.shouldExecuteOnCPU([n,o])){let a=e.texData.get(n.dataId),u=e.texData.get(o.dataId),[l,c]=AL(n.shape,o.shape,a.values,u.values,s),p=e.makeTensorInfo(c,s),m=e.texData.get(p.dataId);return m.values=l,p}let i;return z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?i=new Oo(wM,n.shape,o.shape):i=new io(wM,n.shape,o.shape),e.runWebGLProgram(i,[n,o],s)}var CM={kernelName:hs,backendName:"webgl",kernelFunc:eg};function IM(r,t,e){let n=[xl(r.shape),...yl(r.shape)],o={dtype:r.dtype,shape:n,dataId:r.dataId},s=[xl(t),...yl(t)],i=new Id(s,n),a=!0,u=[n],l=e.runWebGLProgram(i,[o],r.dtype,u,a);return{dataId:l.dataId,shape:t,dtype:l.dtype}}function st(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{shape:s}=n,i=e,a=y.sizeFromShape(o.shape),u=y.inferFromImplicitShape(s,a),l=y.sizeFromShape(u);y.assert(a===l,()=>`The new shape (${u}) has ${l} elements and the old shape (${o.shape}) has ${a} elements. The new shape and old shape must have the same number of elements.`);let c=i.texData.get(o.dataId);return c.isPacked&&!Eu(o.shape,u)&&!(c.texture!==null&&Eu(c.shape,u))?IM(o,u,i):(i.incRef(o.dataId),{dataId:o.dataId,shape:u,dtype:o.dtype})}var SM={kernelName:di,backendName:"webgl",kernelFunc:st};var rg=class{constructor(t,e){this.variableNames=["x"];let{windowSize:n,batchSize:o,inSize:s,outSize:i}=t;this.outputShape=[o,i];let a=Math.floor(n/4)*4,u=n%4,l="sumValue += dot(values, ones);";if(e!=null){let p=1/e;l=`sumValue += dot(values * ${y.isInt(p)?p.toPrecision(2):p}, ones);`}let c="";s%n>0&&(c=`
-        if (inIdx < 0 || inIdx >= ${s}) {
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Multiply.js
+var MUL = "return a * b;";
+function multiply3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { a, b } = inputs;
+  const dtype = backend_util_exports.upcastType(a.dtype, b.dtype);
+  if (a.dtype === "complex64") {
+    const aData = backend2.texData.get(a.dataId);
+    const bData = backend2.texData.get(b.dataId);
+    const realProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.REAL, a.shape, b.shape);
+    const imagProgram = new BinaryOpComplexProgram(COMPLEX_MULTIPLY.IMAG, a.shape, b.shape);
+    const inputs2 = [
+      {
+        dataId: aData.complexTensorInfos.real.dataId,
+        dtype: aData.complexTensorInfos.real.dtype,
+        shape: a.shape
+      },
+      {
+        dataId: aData.complexTensorInfos.imag.dataId,
+        dtype: aData.complexTensorInfos.imag.dtype,
+        shape: a.shape
+      },
+      {
+        dataId: bData.complexTensorInfos.real.dataId,
+        dtype: bData.complexTensorInfos.real.dtype,
+        shape: b.shape
+      },
+      {
+        dataId: bData.complexTensorInfos.imag.dataId,
+        dtype: bData.complexTensorInfos.imag.dtype,
+        shape: b.shape
+      }
+    ];
+    const realPart = backend2.runWebGLProgram(realProgram, inputs2, "float32");
+    const imagPart = backend2.runWebGLProgram(imagProgram, inputs2, "float32");
+    const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    return complexOutput;
+  }
+  if (backend2.shouldExecuteOnCPU([a, b])) {
+    const aData = backend2.texData.get(a.dataId);
+    const bData = backend2.texData.get(b.dataId);
+    const [outValues, outShape] = multiplyImplCPU(a.shape, b.shape, aData.values, bData.values, dtype);
+    const out = backend2.makeTensorInfo(outShape, dtype);
+    const outData = backend2.texData.get(out.dataId);
+    outData.values = outValues;
+    return out;
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_BINARY_OPERATIONS")) {
+    program = new BinaryOpPackedProgram(MUL, a.shape, b.shape);
+  } else {
+    program = new BinaryOpProgram(MUL, a.shape, b.shape);
+  }
+  return backend2.runWebGLProgram(program, [a, b], dtype);
+}
+var multiplyConfig2 = {
+  kernelName: Multiply,
+  backendName: "webgl",
+  kernelFunc: multiply3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/reshape.js
+function packedReshape(input2, afterShape, backend2) {
+  const input3DShape = [
+    getBatchDim(input2.shape),
+    ...getRowsCols(input2.shape)
+  ];
+  const input3D = {
+    dtype: input2.dtype,
+    shape: input3DShape,
+    dataId: input2.dataId
+  };
+  const afterShapeAs3D = [
+    getBatchDim(afterShape),
+    ...getRowsCols(afterShape)
+  ];
+  const program = new ReshapePackedProgram(afterShapeAs3D, input3DShape);
+  const preventEagerUnpackingOfOutput = true;
+  const customValues = [input3DShape];
+  const output = backend2.runWebGLProgram(program, [input3D], input2.dtype, customValues, preventEagerUnpackingOfOutput);
+  return { dataId: output.dataId, shape: afterShape, dtype: output.dtype };
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reshape.js
+function reshape4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const webglBackend = backend2;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  const $xSize = util_exports.sizeFromShape($shape);
+  util_exports.assert(xSize === $xSize, () => `The new shape (${$shape}) has ${$xSize} elements and the old shape (${x.shape}) has ${xSize} elements. The new shape and old shape must have the same number of elements.`);
+  const xTexData = webglBackend.texData.get(x.dataId);
+  if (xTexData.isPacked && !isReshapeFree(x.shape, $shape) && !(xTexData.texture !== null && isReshapeFree(xTexData.shape, $shape))) {
+    return packedReshape(x, $shape, webglBackend);
+  }
+  webglBackend.incRef(x.dataId);
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig2 = {
+  kernelName: Reshape,
+  backendName: "webgl",
+  kernelFunc: reshape4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/mean_gpu.js
+var MeanProgram = class {
+  constructor(reduceInfo, divisor) {
+    this.variableNames = ["x"];
+    const { windowSize, batchSize, inSize, outSize } = reduceInfo;
+    this.outputShape = [batchSize, outSize];
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    let updateSnippet = `sumValue += dot(values, ones);`;
+    if (divisor != null) {
+      const denominator = 1 / divisor;
+      updateSnippet = `sumValue += dot(values * ${util_exports.isInt(denominator) ? denominator.toPrecision(2) : denominator}, ones);`;
+    }
+    let checkOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return 0.0;
         }
-      `),this.userCode=`
+      `;
+    }
+    this.userCode = `
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float getValue(int batch, int inIdx) {
-        ${c}
+        ${checkOutOfBounds}
         return getX(batch, inIdx);
       }
 
@@ -1373,11 +54263,11 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${n};
+        int inOffset = outIdx * ${windowSize};
 
         float sumValue = 0.0;
 
-        for (int i = 0; i < ${a}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
           vec4 values = vec4(
             getValue(batch, inIdx),
@@ -1386,64 +54276,112 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
             getValue(batch, inIdx + 3)
           );
 
-          ${l}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${a};
-        if (${u===1}) {
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
           vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);
 
-          ${l}
-        } else if (${u===2}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1), 0.0, 0.0);
 
-          ${l}
-        } else if (${u===3}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2), 0.0);
 
-          ${l}
+          ${updateSnippet}
         }
         setOutput(sumValue);
       }
-    `}};var Uw=class{constructor(t,e){this.variableNames=["x"];let{windowSize:n,batchSize:o,inSize:s,outSize:i}=t;this.outputShape=[o,i];let a="0.0",u="";e==="prod"?a="1.0":e==="min"?(a="1.0 / 1e-20",u="min"):e==="max"&&(a="-1.0 / 1e-20",u="max");let l=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e==="sum"?l="sumValue":e==="prod"?l="prodValue":e==="all"?l="allValue":e==="any"&&(l="anyValue");let c=Math.floor(n/4)*4,p=n%4,m=`
-      if (${e==="sum"}) {
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/reduce_gpu.js
+var ReduceProgram = class {
+  constructor(reduceInfo, reduceType) {
+    this.variableNames = ["x"];
+    const { windowSize, batchSize, inSize, outSize } = reduceInfo;
+    this.outputShape = [batchSize, outSize];
+    let initializationValue = "0.0";
+    let compareOp = ``;
+    if (reduceType === "prod") {
+      initializationValue = "1.0";
+    } else if (reduceType === "min") {
+      initializationValue = "1.0 / 1e-20";
+      compareOp = `min`;
+    } else if (reduceType === "max") {
+      initializationValue = "-1.0 / 1e-20";
+      compareOp = `max`;
+    }
+    let returnValue = `${reduceType}(${reduceType}(${reduceType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (reduceType === "sum") {
+      returnValue = `sumValue`;
+    } else if (reduceType === "prod") {
+      returnValue = `prodValue`;
+    } else if (reduceType === "all") {
+      returnValue = `allValue`;
+    } else if (reduceType === "any") {
+      returnValue = `anyValue`;
+    }
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    let updateSnippet = `
+      if (${reduceType === "sum"}) {
         sumValue += dot(values, ones);
-      } else if (${e==="prod"}) {
+      } else if (${reduceType === "prod"}) {
         vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);
         prodValue *= tmp[0] * tmp[1];
       } else {
-        minMaxValue = ${u}(values, minMaxValue);
-        if (${e==="min"} || ${e==="max"}) {
-          minMaxValue = ${u}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
+        if (${reduceType === "min"} || ${reduceType === "max"}) {
+          minMaxValue = ${compareOp}(values, minMaxValue);
           bvec4 isNaN = isnan(values);
           if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {
             minMaxValue = vec4(NAN);
           }
         }
       }
-    `,f="vec4";e==="all"?(a="1.0",m=`
+    `;
+    let vecType = `vec4`;
+    if (reduceType === "all") {
+      initializationValue = "1.0";
+      updateSnippet = `
         bool reducedAllValue = all(values);
         float floatedReducedAllValue = float(reducedAllValue);
         allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);
-      `,f="bvec4"):e==="any"&&(a="0.0",m=`
+      `;
+      vecType = `bvec4`;
+    } else if (reduceType === "any") {
+      initializationValue = "0.0";
+      updateSnippet = `
         bool reducedAnyValue = any(values);
         float floatedReducedAnyValue = float(reducedAnyValue);
         anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);
-      `,f="bvec4");let d="";s%n>0&&(d=`
-        if (inIdx < 0 || inIdx >= ${s}) {
+      `;
+      vecType = `bvec4`;
+    }
+    let checkOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return initializationValue;
         }
-      `),this.userCode=`
-      const float initializationValue = ${a};
+      `;
+    }
+    this.userCode = `
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float getValue(int batch, int inIdx) {
-        ${d}
+        ${checkOutOfBounds}
         return getX(batch, inIdx);
       }
 
@@ -1451,174 +54389,713 @@ vec2 packedUVfrom3D(int texNumR, int texNumC,
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${n};
+        int inOffset = outIdx * ${windowSize};
 
-        vec4 minMaxValue = vec4(${a});
+        vec4 minMaxValue = vec4(${initializationValue});
         float prodValue = 1.0;
         float sumValue = 0.0;
         float allValue = 1.0;
         float anyValue = 0.0;
 
-        for (int i = 0; i < ${c}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
-          ${f} values = ${f}(
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             getValue(batch, inIdx + 3)
           );
 
-          ${m}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${c};
-        if (${p===1}) {
-          ${f} values = ${f}(
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             initializationValue,
             initializationValue,
             initializationValue
           );
 
-          ${m}
-        } else if (${p===2}) {
-          ${f} values = ${f}(
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             initializationValue,
             initializationValue
           );
 
-          ${m}
-        } else if (${p===3}) {
-          ${f} values = ${f}(
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
+          ${vecType} values = ${vecType}(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
             getValue(batch, inIdx + 2),
             initializationValue
           );
 
-          ${m}
+          ${updateSnippet}
         }
-        setOutput(${l});
+        setOutput(${returnValue});
       }
-    `}};function bet(r){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let e=t.length?t[t.length-1].outSize:r[1],n=v.computeOptimalWindowSize(e);t.push({inSize:e,windowSize:n,outSize:Math.ceil(e/n)})}return t}function Un(r,t,e,n){let o=bet(r.shape),s=r;for(let i=0;i<o.length;i++){let{inSize:a,windowSize:u,outSize:l}=o[i],c,p;e==="mean"?c=i===0?new rg({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l},a):new rg({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l}):c=new Uw({windowSize:u,inSize:a,batchSize:r.shape[0],outSize:l},e),p=s,s=n.runWebGLProgram(c,[s],t),p.dataId!==r.dataId&&n.disposeIntermediateTensorInfo(p)}return s}var Hw=class{constructor(t,e){this.variableNames=["A"];let n=new Array(t.length);for(let i=0;i<n.length;i++)n[i]=t[e[i]];this.outputShape=n,this.rank=n.length;let o=zt(this.rank),s=wet(e);this.userCode=`
-    void main() {
-      ${o} resRC = getOutputCoords();
-      setOutput(getA(${s}));
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/reduce.js
+function getReductionStages(inShape) {
+  const stages = [];
+  while (stages.length === 0 || stages[stages.length - 1].outSize !== 1) {
+    const outSize = stages.length ? stages[stages.length - 1].outSize : inShape[1];
+    const windowSize = backend_util_exports.computeOptimalWindowSize(outSize);
+    stages.push({
+      inSize: outSize,
+      windowSize,
+      outSize: Math.ceil(outSize / windowSize)
+    });
+  }
+  return stages;
+}
+function reduce(x, dtype, reductionType, backend2) {
+  const reductionStages = getReductionStages(x.shape);
+  let result = x;
+  for (let i = 0; i < reductionStages.length; i++) {
+    const { inSize, windowSize, outSize } = reductionStages[i];
+    let program;
+    let previousResult;
+    if (reductionType === "mean") {
+      program = i === 0 ? new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, inSize) : new MeanProgram({ windowSize, inSize, batchSize: x.shape[0], outSize });
+    } else {
+      program = new ReduceProgram({ windowSize, inSize, batchSize: x.shape[0], outSize }, reductionType);
     }
-    `}};function wet(r){let t=r.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let e=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],n=new Array(t);for(let o=0;o<r.length;o++)n[r[o]]=e[o];return n.join()}var qw=class{constructor(t,e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(t.length);for(let c=0;c<n.length;c++)n[c]=t[e[c]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let o=zt(this.rank),s=ak("rc",this.rank),i=new Array(this.rank);for(let c=0;c<e.length;c++)i[e[c]]=s[c];let a=`vec2(${i.slice(-2).join()})`,u=`++${s[this.rank-1]} < ${n[this.rank-1]}`,l=`getChannel(getA(${i.join()}), ${a})`;this.userCode=`
+    previousResult = result;
+    result = backend2.runWebGLProgram(program, [result], dtype);
+    if (previousResult.dataId !== x.dataId) {
+      backend2.disposeIntermediateTensorInfo(previousResult);
+    }
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/transpose_gpu.js
+var TransposeProgram = class {
+  constructor(aShape, newDim) {
+    this.variableNames = ["A"];
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[newDim[i]];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const switched = getSwitchedCoords(newDim);
+    this.userCode = `
     void main() {
-      ${o} rc = getOutputCoords();
+      ${dtype} resRC = getOutputCoords();
+      setOutput(getA(${switched}));
+    }
+    `;
+  }
+};
+function getSwitchedCoords(newDim) {
+  const rank = newDim.length;
+  if (rank > 6) {
+    throw Error(`Transpose for rank ${rank} is not yet supported`);
+  }
+  const originalOrder = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u", "resRC.v"];
+  const switchedCoords = new Array(rank);
+  for (let i = 0; i < newDim.length; i++) {
+    switchedCoords[newDim[i]] = originalOrder[i];
+  }
+  return switchedCoords.join();
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/transpose_packed_gpu.js
+var TransposePackedProgram = class {
+  constructor(aShape, newDim) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[newDim[i]];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    if (this.rank > 6) {
+      throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);
+    }
+    const dtype = getCoordsDataType(this.rank);
+    const outputOrder = getVecChannels("rc", this.rank);
+    const switchedOrder = new Array(this.rank);
+    for (let i = 0; i < newDim.length; i++) {
+      switchedOrder[newDim[i]] = outputOrder[i];
+    }
+    const innerDims = `vec2(${switchedOrder.slice(-2).join()})`;
+    const nextColumn = `++${outputOrder[this.rank - 1]} < ${outputShape[this.rank - 1]}`;
+    const getc = `getChannel(getA(${switchedOrder.join()}), ${innerDims})`;
+    this.userCode = `
+    void main() {
+      ${dtype} rc = getOutputCoords();
       vec4 result = vec4(0.);
-      result[0] = ${l};
-      if(${u}) {
-        result[1] = ${l};
+      result[0] = ${getc};
+      if(${nextColumn}) {
+        result[1] = ${getc};
       }
-      --${s[this.rank-1]};
-      if(++${s[this.rank-2]} < ${n[this.rank-2]}) {
-        result[2] = ${l};
-        if(${u}) {
-          result[3] = ${l};
+      --${outputOrder[this.rank - 1]};
+      if(++${outputOrder[this.rank - 2]} < ${outputShape[this.rank - 2]}) {
+        result[2] = ${getc};
+        if(${nextColumn}) {
+          result[3] = ${getc};
         }
       }
       setOutput(result);
     }
-    `}};function Au(r,t,e){let n=z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new qw(r.shape,t):new Hw(r.shape,t);return e.runWebGLProgram(n,[r],r.dtype)}function vM(r,t,e,n){let o=t,s=r.shape.length,i=y.parseAxisParam(o,r.shape),a=i,u=v.getAxesPermutation(a,s),l=u!=null,c=r;l&&(c=Au(r,u,n),a=v.getInnerMostAxes(a.length,s)),v.assertAxesAreInnerMostDims("sum",a,s);let[p,m]=v.computeOutAndReduceShapes(c.shape,a),f=p;e&&(f=v.expandShapeToKeepDim(p,i));let d=y.sizeFromShape(m),g=y.sizeFromShape(r.shape)/d,x=st({inputs:{x:c},attrs:{shape:[g,d]},backend:n}),b=Wu(r.dtype),w=Un(x,b,"sum",n),C=st({inputs:{x:w},attrs:{shape:f},backend:n});return n.disposeIntermediateTensorInfo(x),n.disposeIntermediateTensorInfo(w),l&&n.disposeIntermediateTensorInfo(c),C}function Wc(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n;return vM(o,s,i,e)}var NM={kernelName:$s,backendName:"webgl",kernelFunc:Wc};function Oe(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{perm:s}=n,i=e,a=o.shape.length,u=new Array(a);for(let c=0;c<u.length;c++)u[c]=o.shape[s[c]];let l;if(i.shouldExecuteOnCPU([o])){let p=i.texData.get(o.dataId).values,m=Vc(p,o.shape,o.dtype,s,u);l=i.makeTensorInfo(u,o.dtype);let f=i.texData.get(l.dataId);f.values=m}else l=Au(o,s,i);return l}var TM={kernelName:Qn,backendName:"webgl",kernelFunc:Oe};var dk=1e3;function Uc({a:r,b:t,transposeA:e,transposeB:n,backend:o,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:a=0,activation:u=null}){let l=r.shape.length,c=t.shape.length,p=e?r.shape[l-2]:r.shape[l-1],m=n?t.shape[c-1]:t.shape[c-2],f=e?r.shape[l-1]:r.shape[l-2],d=n?t.shape[c-2]:t.shape[c-1],h=r.shape.slice(0,-2),g=t.shape.slice(0,-2),x=y.sizeFromShape(h),b=y.sizeFromShape(g),C=Vr.assertAndGetBroadcastShape(r.shape.slice(0,-2),t.shape.slice(0,-2)).concat([f,d]);y.assert(p===m,()=>`Error in matMul: inner shapes (${p}) and (${m}) of Tensors with shapes ${r.shape} and ${t.shape} and transposeA=${e} and transposeB=${n} must match.`);let N=e?[x,p,f]:[x,f,p],_=n?[b,d,m]:[b,m,d],A=st({inputs:{x:r},backend:o,attrs:{shape:N}}),$=st({inputs:{x:t},backend:o,attrs:{shape:_}}),F=[A,$],P=Math.max(x,b),V=e?A.shape[1]:A.shape[2],G=s!=null,W=i!=null,q=u==="leakyrelu",H=u!=null?bl(u,!0):null,j=G||W||q||H!=null,Y;if((f===1||d===1)&&V>dk&&j===!1){let et=A,rt=$;e&&(et=Oe({inputs:{x:A},backend:o,attrs:{perm:[0,2,1]}}),F.push(et)),n&&(rt=Oe({inputs:{x:$},backend:o,attrs:{perm:[0,2,1]}}),F.push(rt));let ot=d!==1,at=d===1,nt=et;ot&&(nt=st({inputs:{x:et},backend:o,attrs:{shape:[P,V,1]}}),F.push(nt));let it=d===1?2:1,dt=rt;at&&(dt=st({inputs:{x:rt},backend:o,attrs:{shape:[P,1,V]}}),F.push(dt));let ht=eg({inputs:{a:nt,b:dt},backend:o});Y=Wc({inputs:{x:ht},backend:o,attrs:{axis:it,keepDims:!0}}),F.push(ht)}else{let et=sr(r.dtype,t.dtype),rt=new vd(N,_,[P,f,d],e,n,G,H,W,q),ot=[A,$];if(s!=null&&ot.push(s),W&&ot.push(i),q){let at=o.makeTensorInfo([],"float32",y.createScalarValue(a,"float32"));ot.push(at),F.push(at)}Y=o.runWebGLProgram(rt,ot,et)}let Z=st({inputs:{x:Y},backend:o,attrs:{shape:C}});F.push(Y);for(let et of F)o.disposeIntermediateTensorInfo(et);return Z}function Cet(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t,{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n;return Uc({a:o,b:s,transposeA:u,transposeB:l,backend:e,bias:i,preluActivationWeights:a,leakyreluAlpha:p,activation:c})}var kM={kernelName:Ci,backendName:"webgl",kernelFunc:Cet};var EM="return abs(x);";function Iet(r){let{inputs:t,backend:e}=r,{x:n}=t;if(e.shouldExecuteOnCPU([n])&&n.dtype!=="complex64"){let s=e.texData.get(n.dataId),i=Mw(s.values);return e.makeTensorInfo(n.shape,n.dtype,i)}let o;return z().getBool("WEBGL_PACK_UNARY_OPERATIONS")?o=new so(n.shape,EM):o=new tn(n.shape,EM),e.runWebGLProgram(o,[n],n.dtype)}var _M={kernelName:ii,backendName:"webgl",kernelFunc:Iet};var vet=fr+`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transpose_impl.js
+function transposeImpl2(x, perm, backend2) {
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new TransposePackedProgram(x.shape, perm) : new TransposeProgram(x.shape, perm);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sum_impl.js
+function sumImpl(x, axis, keepDims, backend2) {
+  const reductionIndices = axis;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  const sumInputIsTransposed = permutedAxes != null;
+  let sumInput = x;
+  if (sumInputIsTransposed) {
+    sumInput = transposeImpl2(x, permutedAxes, backend2);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", axes, xRank);
+  const [sumOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(sumInput.shape, axes);
+  let outShape = sumOutShape;
+  if (keepDims) {
+    outShape = backend_util_exports.expandShapeToKeepDim(sumOutShape, origAxes);
+  }
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x: sumInput }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const outType = sumOutType(x.dtype);
+  const reduced = reduce(reshapedInput, outType, "sum", backend2);
+  const out = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (sumInputIsTransposed) {
+    backend2.disposeIntermediateTensorInfo(sumInput);
+  }
+  return out;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sum.js
+function sum4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  return sumImpl(x, axis, keepDims, backend2);
+}
+var sumConfig2 = {
+  kernelName: Sum,
+  backendName: "webgl",
+  kernelFunc: sum4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transpose.js
+function transpose3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { perm } = attrs;
+  const webglBackend = backend2;
+  const xRank = x.shape.length;
+  const newShape = new Array(xRank);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[perm[i]];
+  }
+  let out;
+  if (webglBackend.shouldExecuteOnCPU([x])) {
+    const xTexData = webglBackend.texData.get(x.dataId);
+    const values = xTexData.values;
+    const outValues = transposeImplCPU(values, x.shape, x.dtype, perm, newShape);
+    out = webglBackend.makeTensorInfo(newShape, x.dtype);
+    const outData = webglBackend.texData.get(out.dataId);
+    outData.values = outValues;
+  } else {
+    out = transposeImpl2(x, perm, webglBackend);
+  }
+  return out;
+}
+var transposeConfig2 = {
+  kernelName: Transpose,
+  backendName: "webgl",
+  kernelFunc: transpose3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchMatMul_impl.js
+var MATMUL_SHARED_DIM_THRESHOLD = 1e3;
+function batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape4({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape4({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const intermediates = [a3d, b3d];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const sharedDim = transposeA ? a3d.shape[1] : a3d.shape[2];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const fusedActivation = activation2 != null ? mapActivationToShaderProgram(activation2, true) : null;
+  const containsFusedOps = hasBias || hasPreluActivationWeights || hasLeakyreluAlpha || fusedActivation != null;
+  let out;
+  if ((outerShapeA === 1 || outerShapeB === 1) && sharedDim > MATMUL_SHARED_DIM_THRESHOLD && containsFusedOps === false) {
+    let aVec = a3d;
+    let bVec = b3d;
+    if (transposeA) {
+      aVec = transpose3({ inputs: { x: a3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });
+      intermediates.push(aVec);
+    }
+    if (transposeB) {
+      bVec = transpose3({ inputs: { x: b3d }, backend: backend2, attrs: { perm: [0, 2, 1] } });
+      intermediates.push(bVec);
+    }
+    const shouldReshapeA = outerShapeB !== 1;
+    const shouldReshapeB = outerShapeB === 1;
+    let aVec3d = aVec;
+    if (shouldReshapeA) {
+      aVec3d = reshape4({
+        inputs: { x: aVec },
+        backend: backend2,
+        attrs: { shape: [batchDim, sharedDim, 1] }
+      });
+      intermediates.push(aVec3d);
+    }
+    const axis = outerShapeB === 1 ? 2 : 1;
+    let bVec3d = bVec;
+    if (shouldReshapeB) {
+      bVec3d = reshape4({
+        inputs: { x: bVec },
+        backend: backend2,
+        attrs: { shape: [batchDim, 1, sharedDim] }
+      });
+      intermediates.push(bVec3d);
+    }
+    const product = multiply3({ inputs: { a: aVec3d, b: bVec3d }, backend: backend2 });
+    out = sum4({ inputs: { x: product }, backend: backend2, attrs: { axis, keepDims: true } });
+    intermediates.push(product);
+  } else {
+    const dtype = upcastType(a.dtype, b.dtype);
+    const program = new MatMulPackedProgram(a3dShape, b3dShape, [batchDim, outerShapeA, outerShapeB], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const inputs = [a3d, b3d];
+    if (bias != null) {
+      inputs.push(bias);
+    }
+    if (hasPreluActivationWeights) {
+      inputs.push(preluActivationWeights);
+    }
+    if (hasLeakyreluAlpha) {
+      const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+      inputs.push($leakyreluAlpha);
+      intermediates.push($leakyreluAlpha);
+    }
+    out = backend2.runWebGLProgram(program, inputs, dtype);
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: outShape } });
+  intermediates.push(out);
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return outReshaped;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/_FusedMatMul.js
+function _fusedMatMul2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  return batchMatMulImpl({
+    a,
+    b,
+    transposeA,
+    transposeB,
+    backend: backend2,
+    bias,
+    preluActivationWeights,
+    leakyreluAlpha,
+    activation: activation2
+  });
+}
+var _fusedMatMulConfig2 = {
+  kernelName: _FusedMatMul,
+  backendName: "webgl",
+  kernelFunc: _fusedMatMul2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Abs.js
+var ABS2 = `return abs(x);`;
+function abs3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (backend2.shouldExecuteOnCPU([x]) && x.dtype !== "complex64") {
+    const xData = backend2.texData.get(x.dataId);
+    const outValues = simpleAbsImplCPU(xData.values);
+    return backend2.makeTensorInfo(x.shape, x.dtype, outValues);
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+    program = new UnaryOpPackedProgram(x.shape, ABS2);
+  } else {
+    program = new UnaryOpProgram(x.shape, ABS2);
+  }
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var absConfig2 = {
+  kernelName: Abs,
+  backendName: "webgl",
+  kernelFunc: abs3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Acos.js
+var ACOS = CHECK_NAN_SNIPPET + `
   if (abs(x) > 1.) {
     return NAN;
   }
   return acos(x);
-`,Net=Ct({opSnippet:vet}),AM={kernelName:oa,backendName:"webgl",kernelFunc:Net};var Tet=fr+`
+`;
+var acos3 = unaryKernelFunc2({ opSnippet: ACOS });
+var acosConfig2 = {
+  kernelName: Acos,
+  backendName: "webgl",
+  kernelFunc: acos3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Acosh.js
+var ACOSH = CHECK_NAN_SNIPPET + `
   if (x < 1.0) return NAN;
-return log(x + sqrt(x * x - 1.0));`,ket=Ct({opSnippet:Tet}),$M={kernelName:sa,backendName:"webgl",kernelFunc:ket};var DM="return a + b;",Eet=le({opSnippet:DM,packedOpSnippet:DM,supportsComplex:!0,cpuKernelImpl:cL}),RM={kernelName:Zn,backendName:"webgl",kernelFunc:Eet};var Kw=class{constructor(t,e){this.outputShape=[],this.outputShape=t,this.variableNames=e.map((s,i)=>`T${i}`);let n=[];this.variableNames.forEach(s=>{n.push(`float v${s} = get${s}AtOutCoords();`)});let o=this.variableNames.map(s=>`v${s}`).join(" + ");this.userCode=`
-      void main() {
-        ${n.join(`
-        `)}
+return log(x + sqrt(x * x - 1.0));`;
+var acosh3 = unaryKernelFunc2({ opSnippet: ACOSH });
+var acoshConfig2 = {
+  kernelName: Acosh,
+  backendName: "webgl",
+  kernelFunc: acosh3
+};
 
-        float result = ${o};
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Add.js
+var ADD = "return a + b;";
+var addKernelFunc = binaryKernelFunc2({
+  opSnippet: ADD,
+  packedOpSnippet: ADD,
+  supportsComplex: true,
+  cpuKernelImpl: addImplCPU
+});
+var addConfig2 = {
+  kernelName: Add,
+  backendName: "webgl",
+  kernelFunc: addKernelFunc
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/addn_gpu.js
+var AddNProgram = class {
+  constructor(outputShape, shapes) {
+    this.outputShape = [];
+    this.outputShape = outputShape;
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const snippets = [];
+    this.variableNames.forEach((variable2) => {
+      snippets.push(`float v${variable2} = get${variable2}AtOutCoords();`);
+    });
+    const operation = this.variableNames.map((variable2) => {
+      return `v${variable2}`;
+    }).join(" + ");
+    this.userCode = `
+      void main() {
+        ${snippets.join("\n        ")}
+
+        float result = ${operation};
         setOutput(result);
       }
-    `}};var jw=class{constructor(t,e){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.variableNames=e.map((s,i)=>`T${i}`);let n=[];this.variableNames.forEach(s=>{n.push(`vec4 v${s} = get${s}AtOutCoords();`)});let o=this.variableNames.map(s=>`v${s}`).join(" + ");this.userCode=`
-      void main() {
-        ${n.join(`
-        `)}
+    `;
+  }
+};
 
-        vec4 result = ${o};
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/addn_packed_gpu.js
+var AddNPackedProgram = class {
+  constructor(outputShape, shapes) {
+    this.outputShape = [];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = outputShape;
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const snippets = [];
+    this.variableNames.forEach((variable2) => {
+      snippets.push(`vec4 v${variable2} = get${variable2}AtOutCoords();`);
+    });
+    const operation = this.variableNames.map((variable2) => {
+      return `v${variable2}`;
+    }).join(" + ");
+    this.userCode = `
+      void main() {
+        ${snippets.join("\n        ")}
+
+        vec4 result = ${operation};
         setOutput(result);
       }
-    `}};function Xw(r){let{inputs:t,backend:e}=r,n=t;if(n.length===1)return tr({inputs:{x:n[0]},backend:e});if(n.length>z().get("WEBGL_MAX_TEXTURES_IN_SHADER")){let u=Math.floor(n.length/2),l=Xw({inputs:n.slice(0,u),backend:e}),c=Xw({inputs:n.slice(u),backend:e});return Xw({inputs:[l,c],backend:e})}let o=n.map(u=>u.dtype).reduce((u,l)=>sr(u,l)),s=n.map(u=>u.shape),a=z().getBool("WEBGL_PACK")?new jw(n[0].shape,s):new Kw(n[0].shape,s);return e.runWebGLProgram(a,n,o)}var FM={kernelName:Go,backendName:"webgl",kernelFunc:Xw};function _et(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,a)),v.assertAxesAreInnerMostDims("all",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,"all",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var OM={kernelName:ia,backendName:"webgl",kernelFunc:_et};function Aet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,a)),v.assertAxesAreInnerMostDims("any",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,"any",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var PM={kernelName:aa,backendName:"webgl",kernelFunc:Aet};var Yw=class{constructor(t,e,n){this.variableNames=["A"];let{windowSize:o,batchSize:s,outSize:i}=t;n||this.variableNames.push("bestIndicesA"),this.outputShape=[s,i];let a=e==="max"?">":"<",u=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AddN.js
+function addN3(args) {
+  const { inputs, backend: backend2 } = args;
+  const tensors = inputs;
+  if (tensors.length === 1) {
+    return identity3({ inputs: { x: tensors[0] }, backend: backend2 });
+  }
+  if (tensors.length > env().get("WEBGL_MAX_TEXTURES_IN_SHADER")) {
+    const midIndex = Math.floor(tensors.length / 2);
+    const leftSide = addN3({ inputs: tensors.slice(0, midIndex), backend: backend2 });
+    const rightSide = addN3({ inputs: tensors.slice(midIndex), backend: backend2 });
+    return addN3({ inputs: [leftSide, rightSide], backend: backend2 });
+  }
+  const dtype = tensors.map((t) => t.dtype).reduce((d1, d2) => upcastType(d1, d2));
+  const shapes = tensors.map((t) => t.shape);
+  const usePackedOp = env().getBool("WEBGL_PACK");
+  const program = usePackedOp ? new AddNPackedProgram(tensors[0].shape, shapes) : new AddNProgram(tensors[0].shape, shapes);
+  return backend2.runWebGLProgram(program, tensors, dtype);
+}
+var addNConfig2 = {
+  kernelName: AddN,
+  backendName: "webgl",
+  kernelFunc: addN3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/All.js
+function all3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "all", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var allConfig2 = {
+  kernelName: All,
+  backendName: "webgl",
+  kernelFunc: all3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Any.js
+function any3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "any", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var anyConfig2 = {
+  kernelName: Any,
+  backendName: "webgl",
+  kernelFunc: any3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/argminmax_gpu.js
+var ArgMinMaxProgram = class {
+  constructor(reduceInfo, op2, firstPass) {
+    this.variableNames = ["A"];
+    const { windowSize, batchSize, outSize } = reduceInfo;
+    if (!firstPass) {
+      this.variableNames.push("bestIndicesA");
+    }
+    this.outputShape = [batchSize, outSize];
+    const compOp = op2 === "max" ? ">" : "<";
+    const indexSnippet = firstPass ? "inOffset + i;" : "round(getBestIndicesA(batch, inOffset + i));";
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
         int outIdx = coords[1];
-        int inOffset = outIdx * ${o};
+        int inOffset = outIdx * ${windowSize};
 
         int bestIndex = inOffset;
         float bestValue = getA(batch, bestIndex);
 
-        for (int i = 0; i < ${o}; i++) {
-          int inIdx = ${u};
+        for (int i = 0; i < ${windowSize}; i++) {
+          int inIdx = ${indexSnippet};
           float candidate = getA(batch, inIdx);
-          if (candidate ${a} bestValue) {
+          if (candidate ${compOp} bestValue) {
             bestValue = candidate;
             bestIndex = inIdx;
           }
         }
         setOutput(float(bestIndex));
       }
-    `}};var Zw=class{constructor(t,e,n,o){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,y.assert(t.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let s=t[t.length-1],i=Math.ceil(s/e);this.outputShape=t.slice(0,-1),i>1&&this.outputShape.push(i),o||this.variableNames.push("bestIndicesA");let a=this.outputShape,u=a.length,l=zt(u),c=Qe("coords",u),p,m;if(i===1){m=u+1;let $=zt(m);p=`
-        ${$} sourceLocR = ${$}(${c.join()}, 0);
-        ++${c[u-1]};
-        ${$} sourceLocG = ${$}(${c.join()}, 0);
-        ++${c[u-2]};
-        ${$} sourceLocA = ${$}(${c.join()}, 0);
-        --${c[u-1]};
-        ${$} sourceLocB = ${$}(${c.join()}, 0);
-        --${c[u-2]};`}else m=u,p=`
-        ${l} sourceLocR = coords;
-        ++${c[u-1]};
-        ${l} sourceLocG = coords;
-        ++${c[u-2]};
-        ${l} sourceLocA = coords;
-        --${c[u-1]};
-        ${l} sourceLocB = coords;
-        --${c[u-2]};`;let f=["x","y","z","w","u","v"].slice(0,m),d="."+f[m-1],h=f.map($=>"int "+$),g=Qe("sourceLocR",m-1).concat("inIdx.r"),x=Qe("sourceLocG",m-1).concat("inIdx.g"),b=Qe("sourceLocB",m-1).concat("inIdx.b"),w=Qe("sourceLocA",m-1).concat("inIdx.a"),C=n==="max"?"greaterThan":"lessThan",N=o?"":`
-          inIdx = round(vec4(getBestIndicesAChannel(${g.join()}),
-                             getBestIndicesAChannel(${x.join()}),
-                             getBestIndicesAChannel(${b.join()}),
-                             getBestIndicesAChannel(${w.join()})));`,_=`vec4(
-            getAChannel(${g.join()}),
-            hasNextCol ? getAChannel(${x.join()}) : 0.,
-            hasNextRow ? getAChannel(${b.join()}) : 0.,
-            hasNextRow && hasNextCol ? getAChannel(${w.join()}) : 0.)`,A=o?"":`
-      float getBestIndicesAChannel(${h.join()}) {
-        return getChannel(getBestIndicesA(${f.join()}),
-                                          vec2(${f.slice(-2).join()}));
-      }`;this.userCode=`
-      float getAChannel(${h.join()}) {
-        return getChannel(getA(${f.join()}),
-                               vec2(${f.slice(-2).join()}));
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/argminmax_packed_gpu.js
+var ArgMinMaxPackedProgram = class {
+  constructor(shape, windowSize, op2, firstPass) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    util_exports.assert(shape.length > 2, () => `Packed arg${op2.charAt(0).toUpperCase() + op2.slice(1)} supports only inputs with rank above 2.`);
+    const inSize = shape[shape.length - 1];
+    const outSize = Math.ceil(inSize / windowSize);
+    this.outputShape = shape.slice(0, -1);
+    if (outSize > 1) {
+      this.outputShape.push(outSize);
+    }
+    if (!firstPass) {
+      this.variableNames.push("bestIndicesA");
+    }
+    const outShape = this.outputShape;
+    const rank = outShape.length;
+    const dtype = getCoordsDataType(rank);
+    const coords2 = getChannels("coords", rank);
+    let sourceLocSetup;
+    let sourceRank;
+    if (outSize === 1) {
+      sourceRank = rank + 1;
+      const sourceLocDType = getCoordsDataType(sourceRank);
+      sourceLocSetup = `
+        ${sourceLocDType} sourceLocR = ${sourceLocDType}(${coords2.join()}, 0);
+        ++${coords2[rank - 1]};
+        ${sourceLocDType} sourceLocG = ${sourceLocDType}(${coords2.join()}, 0);
+        ++${coords2[rank - 2]};
+        ${sourceLocDType} sourceLocA = ${sourceLocDType}(${coords2.join()}, 0);
+        --${coords2[rank - 1]};
+        ${sourceLocDType} sourceLocB = ${sourceLocDType}(${coords2.join()}, 0);
+        --${coords2[rank - 2]};`;
+    } else {
+      sourceRank = rank;
+      sourceLocSetup = `
+        ${dtype} sourceLocR = coords;
+        ++${coords2[rank - 1]};
+        ${dtype} sourceLocG = coords;
+        ++${coords2[rank - 2]};
+        ${dtype} sourceLocA = coords;
+        --${coords2[rank - 1]};
+        ${dtype} sourceLocB = coords;
+        --${coords2[rank - 2]};`;
+    }
+    const channels = ["x", "y", "z", "w", "u", "v"].slice(0, sourceRank);
+    const inChannel = "." + channels[sourceRank - 1];
+    const intChannels = channels.map((x) => "int " + x);
+    const srcRCoords = getChannels("sourceLocR", sourceRank - 1).concat("inIdx.r");
+    const srcGCoords = getChannels("sourceLocG", sourceRank - 1).concat("inIdx.g");
+    const srcBCoords = getChannels("sourceLocB", sourceRank - 1).concat("inIdx.b");
+    const srcACoords = getChannels("sourceLocA", sourceRank - 1).concat("inIdx.a");
+    const compOp = op2 === "max" ? "greaterThan" : "lessThan";
+    const fetchCandidateIdx = firstPass ? "" : `
+          inIdx = round(vec4(getBestIndicesAChannel(${srcRCoords.join()}),
+                             getBestIndicesAChannel(${srcGCoords.join()}),
+                             getBestIndicesAChannel(${srcBCoords.join()}),
+                             getBestIndicesAChannel(${srcACoords.join()})));`;
+    const fetchValue = `vec4(
+            getAChannel(${srcRCoords.join()}),
+            hasNextCol ? getAChannel(${srcGCoords.join()}) : 0.,
+            hasNextRow ? getAChannel(${srcBCoords.join()}) : 0.,
+            hasNextRow && hasNextCol ? getAChannel(${srcACoords.join()}) : 0.)`;
+    const getBestIndicesAChannelSnippet = firstPass ? "" : `
+      float getBestIndicesAChannel(${intChannels.join()}) {
+        return getChannel(getBestIndicesA(${channels.join()}),
+                                          vec2(${channels.slice(-2).join()}));
+      }`;
+    this.userCode = `
+      float getAChannel(${intChannels.join()}) {
+        return getChannel(getA(${channels.join()}),
+                               vec2(${channels.slice(-2).join()}));
       }
-      ${A}
+      ${getBestIndicesAChannelSnippet}
       void main() {
-        ${l} coords = getOutputCoords();
-        bool hasNextCol = ${c[u-1]} < ${a[u-1]-1};
-        bool hasNextRow = ${c[u-2]} < ${a[u-2]-1};
-        ${p}
-        ivec4 srcIdx = ivec4(sourceLocR${d}, sourceLocG${d},
-          sourceLocB${d}, sourceLocA${d}) * ${e};
+        ${dtype} coords = getOutputCoords();
+        bool hasNextCol = ${coords2[rank - 1]} < ${outShape[rank - 1] - 1};
+        bool hasNextRow = ${coords2[rank - 2]} < ${outShape[rank - 2] - 1};
+        ${sourceLocSetup}
+        ivec4 srcIdx = ivec4(sourceLocR${inChannel}, sourceLocG${inChannel},
+          sourceLocB${inChannel}, sourceLocA${inChannel}) * ${windowSize};
         ivec4 inIdx = srcIdx;
         vec4 bestIndex = vec4(inIdx);
-        vec4 bestValue = ${_};
+        vec4 bestValue = ${fetchValue};
 
-        for (int i = 0; i < ${e}; i++) {
+        for (int i = 0; i < ${windowSize}; i++) {
           inIdx = srcIdx;
-          ${N}
-          vec4 candidate = ${_};
+          ${fetchCandidateIdx}
+          vec4 candidate = ${fetchValue};
           bvec4 nan = isnan(candidate);
           bvec4 replace = bvec4(
-            vec4(${C}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
+            vec4(${compOp}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
 
           bestValue = vec4(replace.x  ? candidate.x : bestValue.x,
                            replace.y  ? candidate.y : bestValue.y,
@@ -1629,27 +55106,215 @@ return log(x + sqrt(x * x - 1.0));`,ket=Ct({opSnippet:Tet}),$M={kernelName:sa,ba
         }
         setOutput(bestIndex);
       }
-    `}};function LM(r,t,e,n=null){let o=t.shape[0],s=t.shape[1];n!=null&&(o=n.shape[0],s=n.shape[1]);let i=v.computeOptimalWindowSize(s),a={windowSize:i,inSize:s,batchSize:o,outSize:Math.ceil(s/i)},u=new Yw(a,e,n==null),l=[t];n!=null&&l.push(n);let c=r.runWebGLProgram(u,l,"int32");if(c.shape[1]===1)return c;let p=LM(r,t,e,c);return r.disposeIntermediateTensorInfo(c),p}function MM(r,t,e,n=null){let o=n!=null?n.shape:t.shape,s=o[o.length-1],i=v.computeOptimalWindowSize(s),a=new Zw(o,i,e,n==null),u=n==null?[t]:[t,n],l=r.runWebGLProgram(a,u,"int32");if(l.shape.length===t.shape.length){let c=MM(r,t,e,l);return r.disposeIntermediateTensorInfo(l),c}return l}function Jw(r,t,e,n){let o=[e];if(v.assertAxesAreInnerMostDims("arg"+n.charAt(0).toUpperCase()+n.slice(1),o,t.shape.length),!z().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let s=[],i=r.texData.get(t.dataId),a=i!==null&&i.isPacked,u=t;a&&(u=r.unpackTensor(t),s.push(u));let[l,c]=v.computeOutAndReduceShapes(u.shape,o),p=y.sizeFromShape(c),m=st({inputs:{x:u},backend:r,attrs:{shape:[-1,p]}});s.push(m);let f=LM(r,m,n);s.push(f);let d=st({inputs:{x:f},backend:r,attrs:{shape:l}});return s.forEach(h=>r.disposeIntermediateTensorInfo(h)),d}return MM(r,t,n)}function $et(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n,i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Oe({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),v.assertAxesAreInnerMostDims("argMax",[i[0]],u.shape.length);let c=Jw(e,u,i[0],"max");return l.forEach(p=>e.disposeIntermediateTensorInfo(p)),c}var zM={kernelName:Wo,backendName:"webgl",kernelFunc:$et};function Det(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s}=n,i=y.parseAxisParam(s,o.shape),a=v.getAxesPermutation(i,o.shape.length),u=o,l=[];a!=null&&(u=Oe({inputs:{x:o},backend:e,attrs:{perm:a}}),l.push(u),i=v.getInnerMostAxes(i.length,u.shape.length)),v.assertAxesAreInnerMostDims("argMin",[i[0]],u.shape.length);let c=Jw(e,u,i[0],"min");return l.forEach(p=>e.disposeIntermediateTensorInfo(p)),c}var BM={kernelName:kl,backendName:"webgl",kernelFunc:Det};var Ret=fr+`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/arg_min_max.js
+function argReduce(backend2, x, reduceType, bestIndicesA = null) {
+  let batchSize = x.shape[0];
+  let inSize = x.shape[1];
+  if (bestIndicesA != null) {
+    batchSize = bestIndicesA.shape[0];
+    inSize = bestIndicesA.shape[1];
+  }
+  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);
+  const reduceInfo = { windowSize, inSize, batchSize, outSize: Math.ceil(inSize / windowSize) };
+  const program = new ArgMinMaxProgram(reduceInfo, reduceType, bestIndicesA == null);
+  const inputs = [x];
+  if (bestIndicesA != null) {
+    inputs.push(bestIndicesA);
+  }
+  const output = backend2.runWebGLProgram(program, inputs, "int32");
+  if (output.shape[1] === 1) {
+    return output;
+  }
+  const result = argReduce(backend2, x, reduceType, output);
+  backend2.disposeIntermediateTensorInfo(output);
+  return result;
+}
+function argReducePacked(backend2, x, reduceType, bestIndicesA = null) {
+  const inShape = bestIndicesA != null ? bestIndicesA.shape : x.shape;
+  const inSize = inShape[inShape.length - 1];
+  const windowSize = backend_util_exports.computeOptimalWindowSize(inSize);
+  const program = new ArgMinMaxPackedProgram(inShape, windowSize, reduceType, bestIndicesA == null);
+  const inputs = bestIndicesA == null ? [x] : [x, bestIndicesA];
+  const output = backend2.runWebGLProgram(program, inputs, "int32");
+  if (output.shape.length === x.shape.length) {
+    const result = argReducePacked(backend2, x, reduceType, output);
+    backend2.disposeIntermediateTensorInfo(output);
+    return result;
+  }
+  return output;
+}
+function argMinMaxReduce(backend2, x, axis, reduceType) {
+  const axes = [axis];
+  backend_util_exports.assertAxesAreInnerMostDims("arg" + reduceType.charAt(0).toUpperCase() + reduceType.slice(1), axes, x.shape.length);
+  if (!env().getBool("WEBGL_PACK_REDUCE") || x.shape.length <= 2) {
+    const intermediateTensorInfos = [];
+    const xtexData = backend2.texData.get(x.dataId);
+    const xIsPacked = xtexData !== null && xtexData.isPacked;
+    let xUnPacked = x;
+    if (xIsPacked) {
+      xUnPacked = backend2.unpackTensor(x);
+      intermediateTensorInfos.push(xUnPacked);
+    }
+    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(xUnPacked.shape, axes);
+    const inSize = util_exports.sizeFromShape(reduceShape);
+    const a2D = reshape4({ inputs: { x: xUnPacked }, backend: backend2, attrs: { shape: [-1, inSize] } });
+    intermediateTensorInfos.push(a2D);
+    const reduced = argReduce(backend2, a2D, reduceType);
+    intermediateTensorInfos.push(reduced);
+    const reshaped = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+    intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return reshaped;
+  }
+  return argReducePacked(backend2, x, reduceType);
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ArgMax.js
+function argMax3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("argMax", [axes[0]], $x.shape.length);
+  const out = argMinMaxReduce(backend2, $x, axes[0], "max");
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return out;
+}
+var argMaxConfig2 = {
+  kernelName: ArgMax,
+  backendName: "webgl",
+  kernelFunc: argMax3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ArgMin.js
+function argMin3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis } = attrs;
+  let axes = util_exports.parseAxisParam(axis, x.shape);
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, x.shape.length);
+  let $x = x;
+  const intermediateTensorInfos = [];
+  if (permutedAxes != null) {
+    $x = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    intermediateTensorInfos.push($x);
+    axes = backend_util_exports.getInnerMostAxes(axes.length, $x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("argMin", [axes[0]], $x.shape.length);
+  const out = argMinMaxReduce(backend2, $x, axes[0], "min");
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return out;
+}
+var argMinConfig2 = {
+  kernelName: ArgMin,
+  backendName: "webgl",
+  kernelFunc: argMin3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Asin.js
+var ASIN = CHECK_NAN_SNIPPET + `
   if (abs(x) > 1.) {
     return NAN;
   }
   return asin(x);
-`,Fet=Ct({opSnippet:Ret}),VM={kernelName:la,backendName:"webgl",kernelFunc:Fet};var Oet=fr+"return log(x + sqrt(x * x + 1.0));",Pet=Ct({opSnippet:Oet}),GM={kernelName:ua,backendName:"webgl",kernelFunc:Pet};var Let=fr+`
+`;
+var asin3 = unaryKernelFunc2({ opSnippet: ASIN });
+var asinConfig2 = {
+  kernelName: Asin,
+  backendName: "webgl",
+  kernelFunc: asin3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Asinh.js
+var ASINH = CHECK_NAN_SNIPPET + `return log(x + sqrt(x * x + 1.0));`;
+var asinh3 = unaryKernelFunc2({ opSnippet: ASINH });
+var asinhConfig2 = {
+  kernelName: Asinh,
+  backendName: "webgl",
+  kernelFunc: asinh3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atan.js
+var ATAN = CHECK_NAN_SNIPPET + `
   return atan(x);
-`,Met=Ct({opSnippet:Let}),WM={kernelName:ca,backendName:"webgl",kernelFunc:Met};var zet=Sd+`
+`;
+var atan4 = unaryKernelFunc2({ opSnippet: ATAN });
+var atanConfig2 = {
+  kernelName: Atan,
+  backendName: "webgl",
+  kernelFunc: atan4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atan2.js
+var ATAN2 = CHECK_NAN_SNIPPET2 + `
   return atan(a, b);
-`,Bet=`
+`;
+var ATAN2_PACKED = `
   vec4 result = atan(a, b);
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+Yi+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Vet=le({opSnippet:zet,packedOpSnippet:Bet}),UM={kernelName:ma,backendName:"webgl",kernelFunc:Vet};var Get=fr+`
+`;
+var atan23 = binaryKernelFunc2({ opSnippet: ATAN2, packedOpSnippet: ATAN2_PACKED });
+var atan2Config2 = {
+  kernelName: Atan2,
+  backendName: "webgl",
+  kernelFunc: atan23
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Atanh.js
+var ATANH = CHECK_NAN_SNIPPET + `
   if ((x < -1.0) || (x > 1.0)) return NAN;
-return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelName:pa,backendName:"webgl",kernelFunc:Wet};var ei=class{constructor(t,e,n,o=!1,s=!1){if(this.variableNames=["x"],e==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let i=t.filterWidth,a=t.strideHeight,u=t.strideWidth,l=t.dilationHeight,c=t.dilationWidth,p=t.effectiveFilterHeight,m=t.effectiveFilterWidth,f=t.padInfo.top,d=t.padInfo.left;this.outputShape=t.outShape;let h=e==="avg",g=`((batch  * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + d`,x=`(xR * ${t.inWidth} + xC) * ${t.inChannels} + d`,b="0.0";if(h||(b="-1.0 / 1e-20"),n){let $=">=";this.userCode=`
-        const ivec2 strides = ivec2(${a}, ${u});
-        const ivec2 pads = ivec2(${f}, ${d});
+return (log(1.0 + x) - log(1.0 - x)) / 2.0;`;
+var atanh3 = unaryKernelFunc2({ opSnippet: ATANH });
+var atanhConfig2 = {
+  kernelName: Atanh,
+  backendName: "webgl",
+  kernelFunc: atanh3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/pool_gpu.js
+var Pool2DProgram = class {
+  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {
+    this.variableNames = ["x"];
+    if (poolType === "avg" && computePositions) {
+      throw new Error("Cannot compute positions for average pool.");
+    }
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.outputShape = convInfo.outShape;
+    const isAvgPool = poolType === "avg";
+    const batchFlattenPositionStr = `((batch  * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;
+    const flattenPositionStr = `(xR * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + d`;
+    let initializationValue = "0.0";
+    if (!isAvgPool) {
+      initializationValue = "-1.0 / 1e-20";
+    }
+    if (computePositions) {
+      const compareOp2 = ">=";
+      this.userCode = `
+        const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+        const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
         void main() {
           ivec4 coords = getOutputCoords();
@@ -1667,19 +55332,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           int minMaxPosition = 0;
           float avgValue = 0.0;
 
-          for (int wR = 0; wR < ${p};
-              wR += ${l}) {
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
             int xR = xRCorner + wR;
 
-            if (xR < 0 || xR >= ${t.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${m};
-                wC += ${c}) {
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
               int xC = xCCorner + wC;
 
-              if (xC < 0 || xC >= ${t.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
@@ -1689,31 +55354,42 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
               // use the current value.
               float currMinMaxValue = mix(
                   value, minMaxValue, minMaxValueFound);
-              if (value ${$} currMinMaxValue) {
+              if (value ${compareOp2} currMinMaxValue) {
                 minMaxValue = value;
                 minMaxValueFound = 1.0;
-                minMaxPosition = ${o?s?g:x:`wR * ${m} + wC`};
+                minMaxPosition = ${flattenPositions ? includeBatchInIndex ? batchFlattenPositionStr : flattenPositionStr : `wR * ${effectiveFilterWidth} + wC`};
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let w="max",C=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e==="avg"&&(C="avgValue / count");let N=Math.floor(i/4)*4,_=i%4,A=`
-      if (${h}) {
+      `;
+      return;
+    }
+    const compareOp = "max";
+    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (poolType === "avg") {
+      returnValue = `avgValue / count`;
+    }
+    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;
+    const filterWidthVec4Remainder = filterWidth % 4;
+    const updateSnippet = `
+      if (${isAvgPool}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${w}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
       }
-    `;this.userCode=`
-      const ivec2 strides = ivec2(${a}, ${u});
-      const ivec2 pads = ivec2(${f}, ${d});
-      const float initializationValue = ${b};
+    `;
+    this.userCode = `
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
 
       float getValue(int batch, int xR, int xC, int d) {
-        if (xC < 0 || xC >= ${t.inWidth}) {
+        if (xC < 0 || xC >= ${convInfo.inWidth}) {
           return initializationValue;
         }
         count += 1.0;
@@ -1731,33 +55407,33 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         // max/min x(?, ?, d) to get y(yR, yC, d).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${b});
+        vec4 minMaxValue = vec4(${initializationValue});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wR = 0; wR < ${p};
-            wR += ${l}) {
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
           int xR = xRCorner + wR;
 
-          if (xR < 0 || xR >= ${t.inHeight}) {
+          if (xR < 0 || xR >= ${convInfo.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${N}; wC += 4) {
-            int xC = xCCorner + wC * ${c};
+          for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {
+            int xC = xCCorner + wC * ${dilationWidth};
 
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${c}, d),
-              getValue(batch, xR, xC + 2 * ${c}, d),
-              getValue(batch, xR, xC + 3 * ${c}, d)
+              getValue(batch, xR, xC + ${dilationWidth}, d),
+              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),
+              getValue(batch, xR, xC + 3 * ${dilationWidth}, d)
             );
 
-            ${A}
+            ${updateSnippet}
           }
 
-          int xC = xCCorner + ${N};
-          if (${_===1}) {
+          int xC = xCCorner + ${filterWidthNearestVec4};
+          if (${filterWidthVec4Remainder === 1}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
               initializationValue,
@@ -1765,33 +55441,63 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
               initializationValue
             );
 
-            ${A}
-          } else if (${_===2}) {
+            ${updateSnippet}
+          } else if (${filterWidthVec4Remainder === 2}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${c}, d),
+              getValue(batch, xR, xC + ${dilationWidth}, d),
               initializationValue,
               initializationValue
             );
 
-            ${A}
-          } else if (${_===3}) {
+            ${updateSnippet}
+          } else if (${filterWidthVec4Remainder === 3}) {
             vec4 values = vec4(
               getValue(batch, xR, xC, d),
-              getValue(batch, xR, xC + ${c}, d),
-              getValue(batch, xR, xC + 2 * ${c}, d),
+              getValue(batch, xR, xC + ${dilationWidth}, d),
+              getValue(batch, xR, xC + 2 * ${dilationWidth}, d),
               initializationValue
             );
 
-            ${A}
+            ${updateSnippet}
           }
         }
-        setOutput(${C});
+        setOutput(${returnValue});
       }
-    `}},$u=class{constructor(t,e,n,o=!1,s=!1){if(this.variableNames=["x"],e==="avg"&&n)throw new Error("Cannot compute positions for average pool.");let i=t.filterWidth,a=t.strideDepth,u=t.strideHeight,l=t.strideWidth,c=t.dilationDepth,p=t.dilationHeight,m=t.dilationWidth,f=t.effectiveFilterDepth,d=t.effectiveFilterHeight,h=t.effectiveFilterWidth,g=t.padInfo.front,x=t.padInfo.top,b=t.padInfo.left;this.outputShape=t.outShape;let w=e==="avg",C="0.0";if(w||(C="-1.0 / 1e-20"),n){let P=">=";this.userCode=`
+    `;
+  }
+};
+var Pool3DProgram = class {
+  constructor(convInfo, poolType, computePositions, flattenPositions = false, includeBatchInIndex = false) {
+    this.variableNames = ["x"];
+    if (poolType === "avg" && computePositions) {
+      throw new Error("Cannot compute positions for average pool.");
+    }
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.outputShape = convInfo.outShape;
+    const isAvgPool = poolType === "avg";
+    let initializationValue = "0.0";
+    if (!isAvgPool) {
+      initializationValue = "-1.0 / 1e-20";
+    }
+    if (computePositions) {
+      const compareOp2 = ">=";
+      this.userCode = `
         const ivec3 strides =
-            ivec3(${a}, ${u}, ${l});
-        const ivec3 pads = ivec3(${g}, ${x}, ${b});
+            ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+        const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
         void main() {
           ivec5 coords = getOutputCoords();
@@ -1809,27 +55515,27 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           float minMaxValueFound = 0.0;
           int minMaxPosition = 0;
 
-          for (int wD = 0; wD < ${f};
-              wD += ${c}) {
+          for (int wD = 0; wD < ${effectiveFilterDepth};
+              wD += ${dilationDepth}) {
             int xD = xDCorner + wD;
 
-            if (xD < 0 || xD >= ${t.inDepth}) {
+            if (xD < 0 || xD >= ${convInfo.inDepth}) {
               continue;
             }
 
-            for (int wR = 0; wR < ${d};
-                wR += ${p}) {
+            for (int wR = 0; wR < ${effectiveFilterHeight};
+                wR += ${dilationHeight}) {
               int xR = xRCorner + wR;
 
-              if (xR < 0 || xR >= ${t.inHeight}) {
+              if (xR < 0 || xR >= ${convInfo.inHeight}) {
                 continue;
               }
 
-              for (int wC = 0; wC < ${h};
-                  wC += ${m}) {
+              for (int wC = 0; wC < ${effectiveFilterWidth};
+                  wC += ${dilationWidth}) {
                 int xC = xCCorner + wC;
 
-                if (xC < 0 || xC >= ${t.inWidth}) {
+                if (xC < 0 || xC >= ${convInfo.inWidth}) {
                   continue;
                 }
 
@@ -1839,34 +55545,45 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                 // use the current value.
                 float currMinMaxValue = mix(
                     value, minMaxValue, minMaxValueFound);
-                if (value ${P} currMinMaxValue) {
+                if (value ${compareOp2} currMinMaxValue) {
                   minMaxValue = value;
                   minMaxValueFound = 1.0;
-                  minMaxPosition = ${o?s?`(((batch * ${t.inDepth} + xD) * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + ch`:`((xD * ${t.inHeight} + xR) * ${t.inWidth} + xC) * ${t.inChannels} + ch`:`wD * ${d} * ${h} +
-                      wR * ${h} + wC`};
+                  minMaxPosition = ${flattenPositions ? includeBatchInIndex ? `(((batch * ${convInfo.inDepth} + xD) * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `((xD * ${convInfo.inHeight} + xR) * ${convInfo.inWidth} + xC) * ${convInfo.inChannels} + ch` : `wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +
+                      wR * ${effectiveFilterWidth} + wC`};
                 }
               }
             }
           }
           setOutput(float(minMaxPosition));
         }
-      `;return}let N="max",_=`${e}(${e}(${e}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;e==="avg"&&(_="avgValue / count");let A=Math.floor(i/4)*4,$=i%4,F=`
-      if (${w}) {
+      `;
+      return;
+    }
+    const compareOp = "max";
+    let returnValue = `${poolType}(${poolType}(${poolType}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;
+    if (poolType === "avg") {
+      returnValue = `avgValue / count`;
+    }
+    const filterWidthNearestVec4 = Math.floor(filterWidth / 4) * 4;
+    const filterWidthVec4Remainder = filterWidth % 4;
+    const updateSnippet = `
+      if (${isAvgPool}) {
         avgValue += dot(values, ones);
       } else {
-        minMaxValue = ${N}(values, minMaxValue);
+        minMaxValue = ${compareOp}(values, minMaxValue);
       }
-    `;this.userCode=`
+    `;
+    this.userCode = `
       const ivec3 strides =
-        ivec3(${a}, ${u}, ${l});
-      const ivec3 pads = ivec3(${g}, ${x}, ${b});
-      const float initializationValue = ${C};
+        ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
+      const float initializationValue = ${initializationValue};
       const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
 
       float count = 0.0;
 
       float getValue(int batch, int xD, int xR, int xC, int ch) {
-        if (xC < 0 || xC >= ${t.inWidth}) {
+        if (xC < 0 || xC >= ${convInfo.inWidth}) {
           return initializationValue;
         }
         count += 1.0;
@@ -1885,41 +55602,41 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).
         // ? = to be determined
-        vec4 minMaxValue = vec4(${C});
+        vec4 minMaxValue = vec4(${initializationValue});
         float avgValue = 0.0;
         count = 0.0;
 
-        for (int wD = 0; wD < ${f};
-            wD += ${c}) {
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+            wD += ${dilationDepth}) {
           int xD = xDCorner + wD;
 
-          if (xD < 0 || xD >= ${t.inDepth}) {
+          if (xD < 0 || xD >= ${convInfo.inDepth}) {
             continue;
           }
 
-          for (int wR = 0; wR < ${d};
-            wR += ${p}) {
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
             int xR = xRCorner + wR;
 
-            if (xR < 0 || xR >= ${t.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${A}; wC += 4) {
-              int xC = xCCorner + wC * ${m};
+            for (int wC = 0; wC < ${filterWidthNearestVec4}; wC += 4) {
+              int xC = xCCorner + wC * ${dilationWidth};
 
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${m}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${m}, ch),
-                getValue(batch, xD, xR, xC + 3 * ${m}, ch)
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 3 * ${dilationWidth}, ch)
               );
 
-              ${F}
+              ${updateSnippet}
             }
 
-            int xC = xCCorner + ${A};
-            if (${$===1}) {
+            int xC = xCCorner + ${filterWidthNearestVec4};
+            if (${filterWidthVec4Remainder === 1}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
                 initializationValue,
@@ -1927,33 +55644,90 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                 initializationValue
               );
 
-              ${F}
-            } else if (${$===2}) {
+              ${updateSnippet}
+            } else if (${filterWidthVec4Remainder === 2}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${m}, ch),
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
                 initializationValue,
                 initializationValue
               );
 
-              ${F}
-            } else if (${$===3}) {
+              ${updateSnippet}
+            } else if (${filterWidthVec4Remainder === 3}) {
               vec4 values = vec4(
                 getValue(batch, xD, xR, xC, ch),
-                getValue(batch, xD, xR, xC + ${m}, ch),
-                getValue(batch, xD, xR, xC + 2 * ${m}, ch),
+                getValue(batch, xD, xR, xC + ${dilationWidth}, ch),
+                getValue(batch, xD, xR, xC + 2 * ${dilationWidth}, ch),
                 initializationValue
               );
 
-              ${F}
+              ${updateSnippet}
             }
           }
-          setOutput(${_});
+          setOutput(${returnValue});
         }
       }
-    `}};function Uet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;Qs(o,"avgPool");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u);if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))return tr({inputs:{x:o},backend:e});let p=new ei(c,"avg",!1);return e.runWebGLProgram(p,[o],"float32")}var qM={kernelName:Uo,backendName:"webgl",kernelFunc:Uet};function Het(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dimRoundingMode:u,dataFormat:l}=n,c=[1,1,1],p=v.computePool3DInfo(o.shape,s,i,c,a,u,l),m=new $u(p,"avg",!1);return e.runWebGLProgram(m,[o],"float32")}var KM={kernelName:El,backendName:"webgl",kernelFunc:Het};var Qw=class{constructor(t){this.variableNames=["dy"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=t.dilationHeight,a=t.dilationWidth,u=t.effectiveFilterHeight,l=t.effectiveFilterWidth,c=u-1-t.padInfo.top,p=l-1-t.padInfo.left,m=1/(e*n);this.userCode=`
-      const ivec2 pads = ivec2(${c}, ${p});
-      const float avgMultiplier = float(${m});
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool.js
+function avgPool3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex2(x, "avgPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in avgPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const avgPoolProgram = new Pool2DProgram(convInfo, "avg", false);
+  return backend2.runWebGLProgram(avgPoolProgram, [x], "float32");
+}
+var avgPoolConfig2 = {
+  kernelName: AvgPool,
+  backendName: "webgl",
+  kernelFunc: avgPool3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool3D.js
+function avgPool3D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);
+  const avgPoolProgram = new Pool3DProgram(convInfo, "avg", false);
+  return backend2.runWebGLProgram(avgPoolProgram, [x], "float32");
+}
+var avgPool3DConfig2 = {
+  kernelName: AvgPool3D,
+  backendName: "webgl",
+  kernelFunc: avgPool3D2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/avg_pool_backprop_gpu.js
+var AvgPool2DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const avgMultiplier = 1 / (filterHeight * filterWidth);
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
+      const float avgMultiplier = float(${avgMultiplier});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -1967,20 +55741,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${u};
-            wR += ${i}) {
-          float dyR = float(dyRCorner + wR) / ${o}.0;
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+            wR += ${dilationHeight}) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${l};
-            wC+= ${a}) {
-            float dyC = float(dyCCorner + wC) / ${s}.0;
+          for (int wC = 0; wC < ${effectiveFilterWidth};
+            wC+= ${dilationWidth}) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
@@ -1993,9 +55767,32 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},tC=class{constructor(t){this.variableNames=["dy"],this.outputShape=t.inShape;let e=t.filterDepth,n=t.filterHeight,o=t.filterWidth,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=t.dilationDepth,l=t.dilationHeight,c=t.dilationWidth,p=t.effectiveFilterDepth,m=t.effectiveFilterHeight,f=t.effectiveFilterWidth,d=p-1-t.padInfo.front,h=m-1-t.padInfo.top,g=f-1-t.padInfo.left,x=1/(e*n*o);this.userCode=`
-      const ivec3 pads = ivec3(${d}, ${h}, ${g});
-      const float avgMultiplier = float(${x});
+    `;
+  }
+};
+var AvgPool3DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy"];
+    this.outputShape = convInfo.inShape;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const avgMultiplier = 1 / (filterDepth * filterHeight * filterWidth);
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
+      const float avgMultiplier = float(${avgMultiplier});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2012,30 +55809,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${p};
-            wD += ${u}) {
-          float dyD = float(dyDCorner + wD) / ${s}.0;
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+            wD += ${dilationDepth}) {
+          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;
 
-          if (dyD < 0.0 || dyD >= ${t.outDepth}.0 || fract(dyD) > 0.0) {
+          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {
             continue;
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${m};
-              wR += ${l}) {
-            float dyR = float(dyRCorner + wR) / ${i}.0;
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
                 fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            for (int wC = 0; wC < ${f};
-                wC += ${c}) {
-              float dyC = float(dyCCorner + wC) / ${a}.0;
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
@@ -2049,69 +55846,463 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}};function qet(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s,{filterSize:a,strides:u,pad:l,dimRoundingMode:c}=n,p=[1,1,1],m=v.computePool3DInfo(i.shape,a,u,p,l,c),f=new tC(m);return e.runWebGLProgram(f,[o],i.dtype)}var jM={kernelName:lp,backendName:"webgl",kernelFunc:qet};function Ket(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s;Qs([o,s],"avgPoolGrad");let{filterSize:a,strides:u,pad:l}=n,c=v.computePool2DInfo(i.shape,a,u,1,l),p=new Qw(c);return e.runWebGLProgram(p,[o],i.dtype)}var XM={kernelName:ap,backendName:"webgl",kernelFunc:Ket};function jet(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;return Uc({a:o,b:s,transposeA:i,transposeB:a,backend:e})}var YM={kernelName:Ho,backendName:"webgl",kernelFunc:jet};var eC=class{constructor(t,e,n,o,s,i){this.outputShape=[],this.variableNames=["x","mean","variance"],v.assertAndGetBroadcastShape(t,e),v.assertAndGetBroadcastShape(t,n);let a="0.0";o!=null&&(v.assertAndGetBroadcastShape(t,o),this.variableNames.push("offset"),a="getOffsetAtOutCoords()");let u="1.0";s!=null&&(v.assertAndGetBroadcastShape(t,s),this.variableNames.push("scale"),u="getScaleAtOutCoords()"),this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPool3DGrad.js
+function avgPool3DGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  const avgPoolBackpropProgram = new AvgPool3DBackpropProgram(convInfo);
+  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);
+}
+var avgPool3DGradConfig3 = {
+  kernelName: AvgPool3DGrad,
+  backendName: "webgl",
+  kernelFunc: avgPool3DGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/AvgPoolGrad.js
+function avgPoolGrad3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  assertNotComplex2([dy, input2], "avgPoolGrad");
+  const { filterSize, strides, pad: pad3 } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3);
+  const avgPoolBackpropProgram = new AvgPool2DBackpropProgram(convInfo);
+  return backend2.runWebGLProgram(avgPoolBackpropProgram, [dy], x.dtype);
+}
+var avgPoolGradConfig3 = {
+  kernelName: AvgPoolGrad,
+  backendName: "webgl",
+  kernelFunc: avgPoolGrad3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchMatMul.js
+function batchMatMul2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  return batchMatMulImpl({ a, b, transposeA, transposeB, backend: backend2 });
+}
+var batchMatMulConfig2 = {
+  kernelName: BatchMatMul,
+  backendName: "webgl",
+  kernelFunc: batchMatMul2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/batchnorm_gpu.js
+var BatchNormProgram = class {
+  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {
+    this.outputShape = [];
+    this.variableNames = ["x", "mean", "variance"];
+    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);
+    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);
+    let offsetSnippet = "0.0";
+    if (offsetShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);
+      this.variableNames.push("offset");
+      offsetSnippet = "getOffsetAtOutCoords()";
+    }
+    let scaleSnippet = "1.0";
+    if (scaleShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);
+      this.variableNames.push("scale");
+      scaleSnippet = "getScaleAtOutCoords()";
+    }
+    this.outputShape = xShape;
+    this.userCode = `
       void main() {
         float x = getXAtOutCoords();
         float mean = getMeanAtOutCoords();
         float variance = getVarianceAtOutCoords();
-        float offset = ${a};
-        float scale = ${u};
-        float inv = scale * inversesqrt(variance + float(${i}));
+        float offset = ${offsetSnippet};
+        float scale = ${scaleSnippet};
+        float inv = scale * inversesqrt(variance + float(${varianceEpsilon}));
         setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));
       }
-    `}};var rC=class{constructor(t,e,n,o,s,i){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],v.assertAndGetBroadcastShape(t,e),v.assertAndGetBroadcastShape(t,n);let a="vec4(0.0)";o!=null&&(v.assertAndGetBroadcastShape(t,o),this.variableNames.push("offset"),a="getOffsetAtOutCoords()");let u="vec4(1.0)";s!=null&&(v.assertAndGetBroadcastShape(t,s),this.variableNames.push("scale"),u="getScaleAtOutCoords()"),this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/batchnorm_packed_gpu.js
+var BatchNormPackedProgram = class {
+  constructor(xShape, meanShape, varianceShape, offsetShape, scaleShape, varianceEpsilon) {
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.variableNames = ["x", "mean", "variance"];
+    backend_util_exports.assertAndGetBroadcastShape(xShape, meanShape);
+    backend_util_exports.assertAndGetBroadcastShape(xShape, varianceShape);
+    let offsetSnippet = "vec4(0.0)";
+    if (offsetShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, offsetShape);
+      this.variableNames.push("offset");
+      offsetSnippet = "getOffsetAtOutCoords()";
+    }
+    let scaleSnippet = "vec4(1.0)";
+    if (scaleShape != null) {
+      backend_util_exports.assertAndGetBroadcastShape(xShape, scaleShape);
+      this.variableNames.push("scale");
+      scaleSnippet = "getScaleAtOutCoords()";
+    }
+    this.outputShape = xShape;
+    this.userCode = `
       void main() {
-        vec4 offset = ${a};
-        vec4 scale = ${u};
+        vec4 offset = ${offsetSnippet};
+        vec4 scale = ${scaleSnippet};
 
         vec4 x = getXAtOutCoords();
         vec4 mean = getMeanAtOutCoords();
         vec4 variance = getVarianceAtOutCoords();
 
-        vec4 inv = scale * inversesqrt(variance + vec4(${i}));
+        vec4 inv = scale * inversesqrt(variance + vec4(${varianceEpsilon}));
 
         setOutput((x - mean) * inv + offset);
       }
-    `}};var Xet=({inputs:r,backend:t,attrs:e})=>{let{x:n,mean:o,variance:s,offset:i,scale:a}=r;y.assert(o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),y.assert(i==null||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),y.assert(a==null||o.shape.length===a.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:u}=e;u==null&&(u=.001);let l=[n,o,s],c=null;i!=null&&(c=i.shape,l.push(i));let p=null;a!=null&&(p=a.shape,l.push(a));let m=z().getBool("WEBGL_PACK_NORMALIZATION")?new rC(n.shape,o.shape,s.shape,c,p,u):new eC(n.shape,o.shape,s.shape,c,p,u);return t.runWebGLProgram(m,l,l[0].dtype)},ZM={kernelName:os,backendName:"webgl",kernelFunc:Xet};var nC=class{constructor(t){this.variableNames=["source"],this.outputShape=t,this.rank=t.length;let e=zt(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let n=Yet(this.rank),o,s=t.map((i,a)=>`sourceLoc.${hk[a]} = start[${a}] + coords.${hk[a]};`);o=`
-        ${e} sourceLoc;
-        ${e} coords = getOutputCoords();
-        ${s.join(`
-`)}
-      `,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchNorm.js
+var batchNorm3 = ({ inputs, backend: backend2, attrs }) => {
+  const { x, mean: mean4, variance, offset, scale: scale2 } = inputs;
+  util_exports.assert(mean4.shape.length === variance.shape.length, () => "Batch normalization gradient requires mean and variance to have equal ranks.");
+  util_exports.assert(offset == null || mean4.shape.length === offset.shape.length, () => "Batch normalization gradient requires mean and offset to have equal ranks.");
+  util_exports.assert(scale2 == null || mean4.shape.length === scale2.shape.length, () => "Batch normalization gradient requires mean and scale to have equal ranks.");
+  let { varianceEpsilon } = attrs;
+  if (varianceEpsilon == null) {
+    varianceEpsilon = 1e-3;
+  }
+  const finalInputs = [x, mean4, variance];
+  let offsetShape = null;
+  if (offset != null) {
+    offsetShape = offset.shape;
+    finalInputs.push(offset);
+  }
+  let scaleShape = null;
+  if (scale2 != null) {
+    scaleShape = scale2.shape;
+    finalInputs.push(scale2);
+  }
+  const program = env().getBool("WEBGL_PACK_NORMALIZATION") ? new BatchNormPackedProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon) : new BatchNormProgram(x.shape, mean4.shape, variance.shape, offsetShape, scaleShape, varianceEpsilon);
+  const output = backend2.runWebGLProgram(program, finalInputs, finalInputs[0].dtype);
+  return output;
+};
+var batchNormConfig2 = {
+  kernelName: FusedBatchNorm,
+  backendName: "webgl",
+  kernelFunc: batchNorm3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/slice_gpu.js
+var SliceProgram = class {
+  constructor(destSize) {
+    this.variableNames = ["source"];
+    this.outputShape = destSize;
+    this.rank = destSize.length;
+    const dtype = getCoordsDataType(this.rank);
+    this.customUniforms = [{ name: "start", arrayIndex: this.rank, type: "int" }];
+    const sourceCoords = getCoords(this.rank);
+    let body;
+    const coordSum = destSize.map((_, i) => {
+      return `sourceLoc.${coords[i]} = start[${i}] + coords.${coords[i]};`;
+    });
+    body = `
+        ${dtype} sourceLoc;
+        ${dtype} coords = getOutputCoords();
+        ${coordSum.join("\n")}
+      `;
+    this.userCode = `
       void main() {
-        ${o}
-        setOutput(getSource(${n}));
+        ${body}
+        setOutput(getSource(${sourceCoords}));
       }
-    `}},hk=["x","y","z","w","u","v"];function Yet(r){if(r===1)return"sourceLoc";if(r<=6)return hk.slice(0,r).map(t=>"sourceLoc."+t).join(",");throw Error(`Slicing for rank ${r} is not yet supported`)}var oC=class{constructor(t){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t,this.rank=t.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let e=zt(this.rank),n=Qe("coords",this.rank),o=Qe("sourceLoc",this.rank),s=this.rank===1?"sourceLoc":`vec2(${o.slice(-2).join()})`,i=`getChannel(getSource(${o.join()}), ${s})`,a=`
-      result.x = ${i};
-      if (++${n[this.rank-1]} < ${t[this.rank-1]}) {
-        ++${o[this.rank-1]};
-        result.y = ${i};
-        --${o[this.rank-1]};
+    `;
+  }
+};
+var coords = ["x", "y", "z", "w", "u", "v"];
+function getCoords(rank) {
+  if (rank === 1) {
+    return "sourceLoc";
+  } else if (rank <= 6) {
+    return coords.slice(0, rank).map((x) => "sourceLoc." + x).join(",");
+  } else {
+    throw Error(`Slicing for rank ${rank} is not yet supported`);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/slice_packed_gpu.js
+var SlicePackedProgram = class {
+  constructor(destSize) {
+    this.variableNames = ["source"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = destSize;
+    this.rank = destSize.length;
+    this.customUniforms = [{ name: "start", arrayIndex: this.rank, type: "int" }];
+    const dtype = getCoordsDataType(this.rank);
+    const coords2 = getChannels("coords", this.rank);
+    const sourceLoc = getChannels("sourceLoc", this.rank);
+    const innerDims = this.rank === 1 ? "sourceLoc" : `vec2(${sourceLoc.slice(-2).join()})`;
+    const getChannel = `getChannel(getSource(${sourceLoc.join()}), ${innerDims})`;
+    const upperRow = `
+      result.x = ${getChannel};
+      if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {
+        ++${sourceLoc[this.rank - 1]};
+        result.y = ${getChannel};
+        --${sourceLoc[this.rank - 1]};
       }
-    `,u=this.rank===1?"":`
-      --${n[this.rank-1]};
-      if (++${n[this.rank-2]} < ${t[this.rank-2]}) {
-        ++${o[this.rank-2]};
-        result.z = ${i};
-        if (++${n[this.rank-1]} < ${t[this.rank-1]}) {
-          ++${o[this.rank-1]};
-          result.w = ${i};
+    `;
+    const lowerRow = this.rank === 1 ? "" : `
+      --${coords2[this.rank - 1]};
+      if (++${coords2[this.rank - 2]} < ${destSize[this.rank - 2]}) {
+        ++${sourceLoc[this.rank - 2]};
+        result.z = ${getChannel};
+        if (++${coords2[this.rank - 1]} < ${destSize[this.rank - 1]}) {
+          ++${sourceLoc[this.rank - 1]};
+          result.w = ${getChannel};
         }
       }
-    `,l=this.rank<=4?`sourceLoc = coords +
-            ${e}(${t.map((c,p)=>`start[${p}]`).join()});`:t.map((c,p)=>`${o[p]} = ${n[p]} + start[${p}];`).join(`
-`);this.userCode=`
+    `;
+    const sourceLocSetup = this.rank <= 4 ? `sourceLoc = coords +
+            ${dtype}(${destSize.map((_, i) => `start[${i}]`).join()});` : destSize.map((_, i) => `${sourceLoc[i]} = ${coords2[i]} + start[${i}];`).join("\n");
+    this.userCode = `
       void main() {
-        ${e} coords = getOutputCoords();
-        ${e} sourceLoc;
-        ${l}
+        ${dtype} coords = getOutputCoords();
+        ${dtype} sourceLoc;
+        ${sourceLocSetup}
         vec4 result = vec4(0.);
-        ${a}
-        ${u}
+        ${upperRow}
+        ${lowerRow}
         setOutput(result);
       }
-    `}};function Zet(r,t,e,n){let o=n.texData.get(r.dataId),s=n.makeTensorInfo(e,r.dtype),i=n.texData.get(s.dataId);Object.assign(i,o),i.refCount=1,i.shape=e,i.dtype=r.dtype;let a=Le.computeFlatOffset(t,y.computeStrides(r.shape));o.slice&&(a+=o.slice.flatOffset),i.slice={flatOffset:a,origDataId:o.slice&&o.slice.origDataId||r.dataId};let u=n.dataRefCount.get(i.slice.origDataId)||1;return n.dataRefCount.set(i.slice.origDataId,u+1),s}function ri(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,size:i}=n,[a,u]=Le.parseSliceParams(o,s,i);if(Le.assertParamsValid(o,a,u),y.sizeFromShape(u)===0)return e.makeTensorInfo(u,o.dtype,[]);if(e.shouldExecuteOnCPU([o])||o.dtype==="string"){let p=e.texData.get(o.dataId),m=VL(p.values,a,u,o.shape,o.dtype);return e.makeTensorInfo(u,o.dtype,m)}let{isPacked:l}=e.texData.get(o.dataId),c=Le.isSliceContinous(o.shape,a,u);if(l||!c){let p=z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new oC(u):new nC(u),m=[a];return e.runWebGLProgram(p,[o],o.dtype,m)}return e.uploadToGPU(o.dataId),Zet(o,a,u,e)}var JM={kernelName:gi,backendName:"webgl",kernelFunc:ri};var Jet=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n;y.assert(o.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let a=s.reduce((b,w)=>b*w),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=[],d=st({inputs:{x:o},backend:e,attrs:{shape:u}}),h=Oe({inputs:{x:d},backend:e,attrs:{perm:l}}),g=st({inputs:{x:h},backend:e,attrs:{shape:c}}),x=ri({inputs:{x:g},backend:e,attrs:{begin:p,size:m}});return f.push(d),f.push(h),f.push(g),f.forEach(b=>e.disposeIntermediateTensorInfo(b)),x},QM={kernelName:ai,backendName:"webgl",kernelFunc:Jet};function Qet(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i}=n,a=e.readSync(o.dataId),u=e.readSync(s.dataId),l=Lw(a,u,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,l)}var tz={kernelName:up,backendName:"webgl",kernelFunc:Qet};function trt(r){let{inputs:t,backend:e}=r,{s0:n,s1:o}=t,s=e.readSync(n.dataId),i=e.readSync(o.dataId),a=v.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return e.makeTensorInfo([a.length],"int32",Int32Array.from(a))}var ez={kernelName:cp,backendName:"webgl",kernelFunc:trt};var ert="return float(a != b);",gk=le({opSnippet:ert,cpuKernelImpl:DL,dtype:"bool"}),rz={kernelName:Da,backendName:"webgl",kernelFunc:gk};function wl(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.texData.get(n.dataId);return tr({inputs:{x:o.complexTensorInfos.real},backend:e})}var nz={kernelName:Rp,backendName:"webgl",kernelFunc:wl};var rrt="return float(int(x));";function oz(r,t){let e=new tn(r.shape,rrt),n=t.runWebGLProgram(e,[r],"int32");return{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}function xk(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dtype:s}=n;if(s==="complex64"){if(o.dtype==="complex64")return tr({inputs:{x:o},backend:e});let i=Ne(o.shape),a=xk({inputs:{x:o},backend:e,attrs:{dtype:"float32"}}),u=En({inputs:{real:a,imag:i},backend:e});return i.dispose(),e.disposeIntermediateTensorInfo(a),u}if(o.dtype==="complex64"){let i=wl({inputs:{input:o},backend:e}),a=xk({inputs:{x:i},backend:e,attrs:{dtype:s}});return e.disposeIntermediateTensorInfo(i),a}if(!y.hasEncodingLoss(o.dtype,s)){let i=tr({inputs:{x:o},backend:e});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(e.shouldExecuteOnCPU([o])){let i=e.texData.get(o.dataId).values,[a,u,l]=mL(i,o.shape,o.dtype,s);return e.makeTensorInfo(a,u,l)}if(s==="int32")return oz(o,e);if(s==="bool"){let i=e.makeTensorInfo([],"bool",y.getTypedArrayFromDType("bool",1)),u=gk({inputs:{a:o,b:i},backend:e});return e.disposeIntermediateTensorInfo(i),u}throw new Error(`Error in Cast: failed to cast ${o.dtype} to ${s}`)}var sz={kernelName:lo,backendName:"webgl",kernelFunc:xk};var iz="return ceil(x);",nrt=Ct({opSnippet:iz,packedOpSnippet:iz,cpuKernelImpl:fL}),az={kernelName:qo,backendName:"webgl",kernelFunc:nrt};var sC=class{constructor(t){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Slice.js
+function shallowSlice(x, begin, size, backend2) {
+  const xTexData = backend2.texData.get(x.dataId);
+  const t = backend2.makeTensorInfo(size, x.dtype);
+  const newTexData = backend2.texData.get(t.dataId);
+  Object.assign(newTexData, xTexData);
+  newTexData.refCount = 1;
+  newTexData.shape = size;
+  newTexData.dtype = x.dtype;
+  let flatOffset = slice_util_exports.computeFlatOffset(begin, util_exports.computeStrides(x.shape));
+  if (xTexData.slice) {
+    flatOffset += xTexData.slice.flatOffset;
+  }
+  newTexData.slice = {
+    flatOffset,
+    origDataId: xTexData.slice && xTexData.slice.origDataId || x.dataId
+  };
+  const refCount = backend2.dataRefCount.get(newTexData.slice.origDataId) || 1;
+  backend2.dataRefCount.set(newTexData.slice.origDataId, refCount + 1);
+  return t;
+}
+function slice3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, size } = attrs;
+  const [$begin, $size] = slice_util_exports.parseSliceParams(x, begin, size);
+  slice_util_exports.assertParamsValid(x, $begin, $size);
+  if (util_exports.sizeFromShape($size) === 0) {
+    return backend2.makeTensorInfo($size, x.dtype, []);
+  }
+  if (backend2.shouldExecuteOnCPU([x]) || x.dtype === "string") {
+    const xTexData = backend2.texData.get(x.dataId);
+    const outValues = sliceImplCPU(xTexData.values, $begin, $size, x.shape, x.dtype);
+    return backend2.makeTensorInfo($size, x.dtype, outValues);
+  }
+  const { isPacked } = backend2.texData.get(x.dataId);
+  const isContinous = slice_util_exports.isSliceContinous(x.shape, $begin, $size);
+  if (isPacked || !isContinous) {
+    const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new SlicePackedProgram($size) : new SliceProgram($size);
+    const customValues = [$begin];
+    return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+  }
+  backend2.uploadToGPU(x.dataId);
+  return shallowSlice(x, $begin, $size, backend2);
+}
+var sliceConfig2 = {
+  kernelName: Slice,
+  backendName: "webgl",
+  kernelFunc: slice3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BatchToSpaceND.js
+var batchToSpaceND3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  util_exports.assert(x.shape.length <= 4, () => "batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const toDispose = [];
+  const reshapedIntermediate = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const transposedIntermediate = transpose3({ inputs: { x: reshapedIntermediate }, backend: backend2, attrs: { perm: permuted } });
+  const reshapedIntermediate2 = reshape4({
+    inputs: { x: transposedIntermediate },
+    backend: backend2,
+    attrs: { shape: reshapedPermuted }
+  });
+  const sliced = slice3({
+    inputs: { x: reshapedIntermediate2 },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  toDispose.push(reshapedIntermediate);
+  toDispose.push(transposedIntermediate);
+  toDispose.push(reshapedIntermediate2);
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return sliced;
+};
+var batchToSpaceNDConfig2 = {
+  kernelName: BatchToSpaceND,
+  backendName: "webgl",
+  kernelFunc: batchToSpaceND3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Bincount.js
+function bincount3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size } = attrs;
+  const xVals = backend2.readSync(x.dataId);
+  const weightsVals = backend2.readSync(weights.dataId);
+  const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);
+  return backend2.makeTensorInfo([size], weights.dtype, outVals);
+}
+var bincountConfig2 = {
+  kernelName: Bincount,
+  backendName: "webgl",
+  kernelFunc: bincount3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/BroadcastArgs.js
+function broadcastArgs3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { s0, s1 } = inputs;
+  const s0Vals = backend2.readSync(s0.dataId);
+  const s1Vals = backend2.readSync(s1.dataId);
+  const broadcastShape = backend_util_exports.assertAndGetBroadcastShape(Array.from(s0Vals), Array.from(s1Vals));
+  return backend2.makeTensorInfo([broadcastShape.length], "int32", Int32Array.from(broadcastShape));
+}
+var broadcastArgsConfig2 = {
+  kernelName: BroadcastArgs,
+  backendName: "webgl",
+  kernelFunc: broadcastArgs3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NotEqual.js
+var NOT_EQUAL = `return float(a != b);`;
+var notEqual3 = binaryKernelFunc2({ opSnippet: NOT_EQUAL, cpuKernelImpl: notEqualImplCPU, dtype: "bool" });
+var notEqualConfig2 = {
+  kernelName: NotEqual,
+  backendName: "webgl",
+  kernelFunc: notEqual3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Real.js
+function real3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputData = backend2.texData.get(input2.dataId);
+  return identity3({ inputs: { x: inputData.complexTensorInfos.real }, backend: backend2 });
+}
+var realConfig2 = {
+  kernelName: Real,
+  backendName: "webgl",
+  kernelFunc: real3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernel_utils/int.js
+var TO_INT = `return float(int(x));`;
+function int(input2, backend2) {
+  const program = new UnaryOpProgram(input2.shape, TO_INT);
+  const output = backend2.runWebGLProgram(program, [input2], "int32");
+  return { dataId: output.dataId, shape: output.shape, dtype: output.dtype };
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cast.js
+function cast4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dtype } = attrs;
+  if (dtype === "complex64") {
+    if (x.dtype === "complex64") {
+      return identity3({ inputs: { x }, backend: backend2 });
+    }
+    const zerosTensor = zeros(x.shape);
+    const floatX = cast4({ inputs: { x }, backend: backend2, attrs: { dtype: "float32" } });
+    const result = complex3({ inputs: { real: floatX, imag: zerosTensor }, backend: backend2 });
+    zerosTensor.dispose();
+    backend2.disposeIntermediateTensorInfo(floatX);
+    return result;
+  }
+  if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const result = cast4({ inputs: { x: realPart }, backend: backend2, attrs: { dtype } });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    return result;
+  }
+  if (!util_exports.hasEncodingLoss(x.dtype, dtype)) {
+    const result = identity3({ inputs: { x }, backend: backend2 });
+    return { dataId: result.dataId, shape: result.shape, dtype };
+  }
+  if (backend2.shouldExecuteOnCPU([x])) {
+    const values = backend2.texData.get(x.dataId).values;
+    const [resultShape, resultType, resultData] = castImplCPU(values, x.shape, x.dtype, dtype);
+    return backend2.makeTensorInfo(resultShape, resultType, resultData);
+  }
+  if (dtype === "int32") {
+    return int(x, backend2);
+  }
+  if (dtype === "bool") {
+    const zerosTensorInfo = backend2.makeTensorInfo([], "bool", util_exports.getTypedArrayFromDType("bool", 1));
+    const binaryInputs = { a: x, b: zerosTensorInfo };
+    const result = notEqual3({ inputs: binaryInputs, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(zerosTensorInfo);
+    return result;
+  }
+  throw new Error(`Error in Cast: failed to cast ${x.dtype} to ${dtype}`);
+}
+var castConfig2 = {
+  kernelName: Cast,
+  backendName: "webgl",
+  kernelFunc: cast4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Ceil.js
+var CEIL = `return ceil(x);`;
+var ceil3 = unaryKernelFunc2({ opSnippet: CEIL, packedOpSnippet: CEIL, cpuKernelImpl: ceilImplCPU });
+var ceilConfig2 = {
+  kernelName: Ceil,
+  backendName: "webgl",
+  kernelFunc: ceil3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/clip_gpu.js
+var ClipProgram = class {
+  constructor(aShape) {
+    this.variableNames = ["A"];
+    this.customUniforms = [
+      { name: "minVal", type: "float" },
+      { name: "maxVal", type: "float" }
+    ];
+    this.outputShape = aShape;
+    this.userCode = `
 
       void main() {
         float value = getAAtOutCoords();
@@ -2122,7 +56313,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         setOutput(clamp(value, minVal, maxVal));
       }
-    `}};var iC=class{constructor(t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/clip_packed_gpu.js
+var ClipPackedProgram = class {
+  constructor(aShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "minVal", type: "float" },
+      { name: "maxVal", type: "float" }
+    ];
+    this.outputShape = aShape;
+    this.userCode = `
       void main() {
         vec4 value = getAAtOutCoords();
 
@@ -2133,7 +56339,36 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         setOutput(clamp(value, vec4(minVal), vec4(maxVal)));
       }
-    `}};function ort(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{clipValueMin:s,clipValueMax:i}=n,a;z().getBool("WEBGL_PACK_CLIP")?a=new iC(o.shape):a=new sC(o.shape);let u=[[s],[i]];return e.runWebGLProgram(a,[o],o.dtype,u)}var lz={kernelName:uo,backendName:"webgl",kernelFunc:ort};var aC=class{constructor(t){this.variableNames=["real","imag"],this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ClipByValue.js
+function clipByValue3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { clipValueMin, clipValueMax } = attrs;
+  let program;
+  if (env().getBool("WEBGL_PACK_CLIP")) {
+    program = new ClipPackedProgram(x.shape);
+  } else {
+    program = new ClipProgram(x.shape);
+  }
+  const customValues = [[clipValueMin], [clipValueMax]];
+  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+}
+var clipByValueConfig2 = {
+  kernelName: ClipByValue,
+  backendName: "webgl",
+  kernelFunc: clipByValue3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/complex_abs_gpu.js
+var ComplexAbsProgram = class {
+  constructor(shape) {
+    this.variableNames = ["real", "imag"];
+    this.outputShape = shape;
+    this.userCode = `
       void main() {
         float re = abs(getRealAtOutCoords());
         float im = abs(getImagAtOutCoords());
@@ -2146,96 +56381,338 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))
         );
       }
-    `}};function uz(r,t){return{dataId:t.dataId,dtype:t.dtype,shape:r.shape}}function srt(r){let{inputs:t,backend:e}=r,{x:n}=t,o=e.texData.get(n.dataId),s=new aC(n.shape),i=[uz(n,o.complexTensorInfos.real),uz(n,o.complexTensorInfos.imag)];return e.runWebGLProgram(s,i,i[0].dtype)}var cz={kernelName:_l,backendName:"webgl",kernelFunc:srt};var lC=class{constructor(t){this.outputShape=[],this.outputShape=v.computeOutShape(t,1),this.variableNames=t.map((i,a)=>`T${a}`);let e=new Array(t.length-1);e[0]=t[0][1];for(let i=1;i<e.length;i++)e[i]=e[i-1]+t[i][1];let n=[`if (yC < ${e[0]}) setOutput(getT0(yR, yC));`];for(let i=1;i<e.length;i++){let a=e[i-1];n.push(`else if (yC < ${e[i]}) setOutput(getT${i}(yR, yC-${a}));`)}let o=e.length,s=e[e.length-1];n.push(`else setOutput(getT${o}(yR, yC-${s}));`),this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ComplexAbs.js
+function makeComplexComponentTensorInfo(complexTensor, complexPart) {
+  return {
+    dataId: complexPart.dataId,
+    dtype: complexPart.dtype,
+    shape: complexTensor.shape
+  };
+}
+function complexAbs2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const xData = backend2.texData.get(x.dataId);
+  const program = new ComplexAbsProgram(x.shape);
+  const programInputs = [
+    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.real),
+    makeComplexComponentTensorInfo(x, xData.complexTensorInfos.imag)
+  ];
+  return backend2.runWebGLProgram(program, programInputs, programInputs[0].dtype);
+}
+var complexAbsConfig2 = {
+  kernelName: ComplexAbs,
+  backendName: "webgl",
+  kernelFunc: complexAbs2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/concat_gpu.js
+var ConcatProgram = class {
+  constructor(shapes) {
+    this.outputShape = [];
+    this.outputShape = backend_util_exports.computeOutShape(shapes, 1);
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const offsets = new Array(shapes.length - 1);
+    offsets[0] = shapes[0][1];
+    for (let i = 1; i < offsets.length; i++) {
+      offsets[i] = offsets[i - 1] + shapes[i][1];
+    }
+    const snippets = [`if (yC < ${offsets[0]}) setOutput(getT0(yR, yC));`];
+    for (let i = 1; i < offsets.length; i++) {
+      const shift = offsets[i - 1];
+      snippets.push(`else if (yC < ${offsets[i]}) setOutput(getT${i}(yR, yC-${shift}));`);
+    }
+    const lastIndex = offsets.length;
+    const lastShift = offsets[offsets.length - 1];
+    snippets.push(`else setOutput(getT${lastIndex}(yR, yC-${lastShift}));`);
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int yR = coords.x;
         int yC = coords.y;
 
-        ${n.join(`
-        `)}
+        ${snippets.join("\n        ")}
       }
-    `}};var cC=class{constructor(t,e){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=v.computeOutShape(t,e);let n=this.outputShape,o=n.length,s=zt(o),i=Qe("coords",o),a=["x","y","z","w","u","v"].slice(0,o);this.variableNames=t.map((h,g)=>`T${g}`);let u=new Array(t.length-1);u[0]=t[0][e];for(let h=1;h<u.length;h++)u[h]=u[h-1]+t[h][e];let l=a[e],c=a.slice(-2),p=a.join(),m=`if (${l} < ${u[0]}) {
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/concat_packed_gpu.js
+var ConcatPackedProgram = class {
+  constructor(shapes, axis) {
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    this.outputShape = backend_util_exports.computeOutShape(shapes, axis);
+    const shape = this.outputShape;
+    const rank = shape.length;
+    const dtype = getCoordsDataType(rank);
+    const coords2 = getChannels("coords", rank);
+    const channels = ["x", "y", "z", "w", "u", "v"].slice(0, rank);
+    this.variableNames = shapes.map((_, i) => `T${i}`);
+    const offsets = new Array(shapes.length - 1);
+    offsets[0] = shapes[0][axis];
+    for (let i = 1; i < offsets.length; i++) {
+      offsets[i] = offsets[i - 1] + shapes[i][axis];
+    }
+    const channel = channels[axis];
+    const lastChannels = channels.slice(-2);
+    const allChannels = channels.join();
+    let getValueSnippet = `if (${channel} < ${offsets[0]}) {
         return getChannel(
-            getT0(${p}), vec2(${c.join()}));
-        }`;for(let h=1;h<u.length;h++){let g=u[h-1];m+=`
-        if (${l} < ${u[h]}  && ${l} >= ${u[h-1]}) {
+            getT0(${allChannels}), vec2(${lastChannels.join()}));
+        }`;
+    for (let i = 1; i < offsets.length; i++) {
+      const shift2 = offsets[i - 1];
+      getValueSnippet += `
+        if (${channel} < ${offsets[i]}  && ${channel} >= ${offsets[i - 1]}) {
           return getChannel(
-            getT${h}(${uC(a,l,g)}),
-            vec2(${uC(c,l,g)}));
-        }`}let f=u.length,d=u[u.length-1];m+=`
+            getT${i}(${shiftedChannels(channels, channel, shift2)}),
+            vec2(${shiftedChannels(lastChannels, channel, shift2)}));
+        }`;
+    }
+    const lastIndex = offsets.length;
+    const shift = offsets[offsets.length - 1];
+    getValueSnippet += `
         return getChannel(
-          getT${f}(${uC(a,l,d)}),
-          vec2(${uC(c,l,d)}));`,this.userCode=`
-      float getValue(${a.map(h=>"int "+h)}) {
-        ${m}
+          getT${lastIndex}(${shiftedChannels(channels, channel, shift)}),
+          vec2(${shiftedChannels(lastChannels, channel, shift)}));`;
+    this.userCode = `
+      float getValue(${channels.map((x) => "int " + x)}) {
+        ${getValueSnippet}
       }
 
       void main() {
-        ${s} coords = getOutputCoords();
-        vec4 result = vec4(getValue(${i}), 0., 0., 0.);
+        ${dtype} coords = getOutputCoords();
+        vec4 result = vec4(getValue(${coords2}), 0., 0., 0.);
 
-        ${i[o-1]} = ${i[o-1]} + 1;
-        if (${i[o-1]} < ${n[o-1]}) {
-          result.g = getValue(${i});
+        ${coords2[rank - 1]} = ${coords2[rank - 1]} + 1;
+        if (${coords2[rank - 1]} < ${shape[rank - 1]}) {
+          result.g = getValue(${coords2});
         }
 
-        ${i[o-2]} = ${i[o-2]} + 1;
-        if (${i[o-2]} < ${n[o-2]}) {
-          result.a = getValue(${i});
+        ${coords2[rank - 2]} = ${coords2[rank - 2]} + 1;
+        if (${coords2[rank - 2]} < ${shape[rank - 2]}) {
+          result.a = getValue(${coords2});
         }
 
-        ${i[o-1]} = ${i[o-1]} - 1;
-        if (${i[o-2]} < ${n[o-2]} &&
-            ${i[o-1]} < ${n[o-1]}) {
-          result.b = getValue(${i});
+        ${coords2[rank - 1]} = ${coords2[rank - 1]} - 1;
+        if (${coords2[rank - 2]} < ${shape[rank - 2]} &&
+            ${coords2[rank - 1]} < ${shape[rank - 1]}) {
+          result.b = getValue(${coords2});
         }
         setOutput(result);
       }
-    `}};function uC(r,t,e){let n=r.indexOf(t);return r.map((s,i)=>i===n?`${s} - ${e}`:s).join()}function Hc(r){let{inputs:t,backend:e}=r,{input:n}=t,o=e.texData.get(n.dataId);return tr({inputs:{x:o.complexTensorInfos.imag},backend:e})}var pz={kernelName:Sp,backendName:"webgl",kernelFunc:Hc};function Nd(r,t,e){let n=r[0].dtype;if(n==="complex64"){let p=r.map(g=>wl({inputs:{input:g},backend:e})),m=r.map(g=>Hc({inputs:{input:g},backend:e})),f=Nd(p,t,e),d=Nd(m,t,e),h=En({inputs:{real:f,imag:d},backend:e});return p.forEach(g=>e.disposeIntermediateTensorInfo(g)),m.forEach(g=>e.disposeIntermediateTensorInfo(g)),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),h}let o=e.shouldExecuteOnCPU(r);if(n==="string"&&(o=!0),o){let p=r.map(b=>{let w=y.sizeFromShape(b.shape.slice(t));return st({inputs:{x:b},backend:e,attrs:{shape:[-1,w]}})}),m=p.map(b=>({vals:e.readSync(b.dataId),shape:b.shape})),f=v.computeOutShape(p.map(b=>b.shape),1),d=p[0].shape[0]===1,h=dL(m,f,n,d),g=v.computeOutShape(r.map(b=>b.shape),t),x=e.makeTensorInfo(g,n,h);return p.forEach(b=>e.disposeIntermediateTensorInfo(b)),x}let s=z().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");if(r.length>s){let p=[];for(let f=0;f<r.length;f+=s){let d=r.slice(f,f+s);p.push(Nd(d,t,e))}let m=Nd(p,t,e);for(let f of p)e.disposeIntermediateTensorInfo(f);return m}if(z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&r[0].shape.length>1){let p=new cC(r.map(m=>m.shape),t);return e.runWebGLProgram(p,r,n)}let{tensors2D:i,outShape:a}=irt(r,t,e),u=new lC(i.map(p=>p.shape)),l=e.runWebGLProgram(u,i,n);i.forEach(p=>e.disposeIntermediateTensorInfo(p));let c=st({inputs:{x:l},attrs:{shape:a},backend:e});return e.disposeIntermediateTensorInfo(l),c}function irt(r,t,e){let n=v.computeOutShape(r.map(s=>s.shape),t);return{tensors2D:r.map(s=>st({inputs:{x:s},attrs:{shape:[-1,y.sizeFromShape(s.shape.slice(t))]},backend:e})),outShape:n}}function yk(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n,s=y.parseAxisParam(o,t[0].shape)[0],i=t.map(l=>l.shape);v.assertParamsConsistent(i,s);let a=v.computeOutShape(t.map(l=>l.shape),s);if(y.sizeFromShape(a)===0)return e.makeTensorInfo(a,t[0].dtype,[]);let u=t.filter(l=>y.sizeFromShape(l.shape)>0);return u.length===1?tr({inputs:{x:u[0]},backend:e}):Nd(u,s,e)}var mz={kernelName:li,backendName:"webgl",kernelFunc:yk};var Td=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=["x","W"],this.outputShape=t.outShape;let i=t.padInfo.top,a=t.padInfo.left,u=t.strideHeight,l=t.strideWidth,c=t.dilationHeight,p=t.dilationWidth,m=t.filterHeight,f=t.filterWidth,d=Math.floor(t.inChannels/4)*4,h=t.inChannels%4,g=t.dataFormat==="channelsLast",x=g?1:2,b=g?2:3,w=g?3:1,C="",N="";n&&(o?C=`float activation(float a) {
-          float b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:s?C=`float activation(float a) {
-          float b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:C=`
-          float activation(float x) {
-            ${n}
-          }
-        `,N="result = activation(result);");let _=e?"result += getBiasAtOutCoords();":"";e&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${C}
+    `;
+  }
+};
+function shiftedChannels(channels, channel, shift) {
+  const channelIdx = channels.indexOf(channel);
+  const res = channels.map((c, idx) => {
+    if (idx === channelIdx) {
+      return `${c} - ${shift}`;
+    } else {
+      return c;
+    }
+  });
+  return res.join();
+}
 
-      const ivec2 strides = ivec2(${u}, ${l});
-      const ivec2 pads = ivec2(${i}, ${a});
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Imag.js
+function imag3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const inputData = backend2.texData.get(input2.dataId);
+  return identity3({ inputs: { x: inputData.complexTensorInfos.imag }, backend: backend2 });
+}
+var imagConfig2 = {
+  kernelName: Imag,
+  backendName: "webgl",
+  kernelFunc: imag3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Concat_impl.js
+function concatImpl2(inputs, axis, backend2) {
+  const dtype = inputs[0].dtype;
+  if (dtype === "complex64") {
+    const reals = inputs.map((t) => real3({ inputs: { input: t }, backend: backend2 }));
+    const imags = inputs.map((t) => imag3({ inputs: { input: t }, backend: backend2 }));
+    const realConcated = concatImpl2(reals, axis, backend2);
+    const imagConcated = concatImpl2(imags, axis, backend2);
+    const result2 = complex3({ inputs: { real: realConcated, imag: imagConcated }, backend: backend2 });
+    reals.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+    imags.forEach((i) => backend2.disposeIntermediateTensorInfo(i));
+    backend2.disposeIntermediateTensorInfo(realConcated);
+    backend2.disposeIntermediateTensorInfo(imagConcated);
+    return result2;
+  }
+  let runOnCpu = backend2.shouldExecuteOnCPU(inputs);
+  if (dtype === "string") {
+    runOnCpu = true;
+  }
+  if (runOnCpu) {
+    const tensors2D2 = inputs.map((t) => {
+      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));
+      const shape = [-1, innerSize];
+      return reshape4({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+    });
+    const inputsValShapes = tensors2D2.map((t) => {
+      return { vals: backend2.readSync(t.dataId), shape: t.shape };
+    });
+    const outShape2 = backend_util_exports.computeOutShape(tensors2D2.map((t) => t.shape), 1);
+    const simplyConcat = tensors2D2[0].shape[0] === 1;
+    const outVals = concatImplCPU(inputsValShapes, outShape2, dtype, simplyConcat);
+    const finalOutShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+    const outInfo = backend2.makeTensorInfo(finalOutShape, dtype, outVals);
+    tensors2D2.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return outInfo;
+  }
+  const maxTexturesInShader = env().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");
+  if (inputs.length > maxTexturesInShader) {
+    const reducedInputs = [];
+    for (let i = 0; i < inputs.length; i += maxTexturesInShader) {
+      const subArray = inputs.slice(i, i + maxTexturesInShader);
+      reducedInputs.push(concatImpl2(subArray, axis, backend2));
+    }
+    const result2 = concatImpl2(reducedInputs, axis, backend2);
+    for (const i of reducedInputs) {
+      backend2.disposeIntermediateTensorInfo(i);
+    }
+    return result2;
+  }
+  if (env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") && inputs[0].shape.length > 1) {
+    const program2 = new ConcatPackedProgram(inputs.map((t) => t.shape), axis);
+    return backend2.runWebGLProgram(program2, inputs, dtype);
+  }
+  const { tensors2D, outShape } = computeTensors2D(inputs, axis, backend2);
+  const program = new ConcatProgram(tensors2D.map((t) => t.shape));
+  const result = backend2.runWebGLProgram(program, tensors2D, dtype);
+  tensors2D.forEach((r) => backend2.disposeIntermediateTensorInfo(r));
+  const reshapedResult = reshape4({ inputs: { x: result }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(result);
+  return reshapedResult;
+}
+function computeTensors2D(inputs, axis, backend2) {
+  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+  const tensors2D = inputs.map((x) => reshape4({
+    inputs: { x },
+    attrs: { shape: [-1, util_exports.sizeFromShape(x.shape.slice(axis))] },
+    backend: backend2
+  }));
+  return { tensors2D, outShape };
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Concat.js
+function concat3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, $axis);
+  const outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), $axis);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return backend2.makeTensorInfo(outShape, inputs[0].dtype, []);
+  }
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity3({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  return concatImpl2($inputs, $axis, backend2);
+}
+var concatConfig2 = {
+  kernelName: Concat,
+  backendName: "webgl",
+  kernelFunc: concat3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_gpu.js
+var Conv2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivationWeights = false, hasLeakyreluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;
+    const inputDepthVec4Remainder = convInfo.inChannels % 4;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const channelDim = isChannelsLast ? 3 : 1;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivationWeights) {
+        activationSnippet = `float activation(float a) {
+          float b = getPreluActivationWeightsAtOutCoords();
+          ${activation2}
+        }`;
+      } else if (hasLeakyreluAlpha) {
+        activationSnippet = `float activation(float a) {
+          float b = getLeakyreluAlphaAtOutCoords();
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `
+          float activation(float x) {
+            ${activation2}
+          }
+        `;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivationWeights) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyreluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
+
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d2 = coords[${w}];
+        int d2 = coords[${channelDim}];
 
         ivec2 xRCCorner =
-            ivec2(coords[${x}], coords[${b}]) * strides - pads;
+            ivec2(coords[${rowDim}], coords[${colDim}]) * strides - pads;
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${m}; wR++) {
-          int xR = xRCorner + wR * ${c};
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          int xR = xRCorner + wR * ${dilationHeight};
 
-          if (xR < 0 || xR >= ${t.inHeight}) {
+          if (xR < 0 || xR >= ${convInfo.inHeight}) {
             continue;
           }
 
-          for (int wC = 0; wC < ${f}; wC++) {
-            int xC = xCCorner + wC * ${p};
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            int xC = xCCorner + wC * ${dilationWidth};
 
-            if (xC < 0 || xC >= ${t.inWidth}) {
+            if (xC < 0 || xC >= ${convInfo.inWidth}) {
               continue;
             }
 
-            for (int d1 = 0; d1 < ${d}; d1 += 4) {
+            for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {
               vec4 wValues = vec4(
                 getW(wR, wC, d1, d2),
                 getW(wR, wC, d1 + 1, d2),
@@ -2243,7 +56720,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                 getW(wR, wC, d1 + 3, d2)
               );
 
-              if (${g}) {
+              if (${isChannelsLast}) {
                 vec4 xValues = vec4(
                   getX(batch, xR, xC, d1),
                   getX(batch, xR, xC, d1 + 1),
@@ -2262,57 +56739,57 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
               }
             }
 
-            if (${h===1}) {
+            if (${inputDepthVec4Remainder === 1}) {
 
-              if (${g}) {
+              if (${isChannelsLast}) {
                 dotProd +=
-                    getX(batch, xR, xC, ${d}) *
-                    getW(wR, wC, ${d}, d2);
+                    getX(batch, xR, xC, ${inputDepthNearestVec4}) *
+                    getW(wR, wC, ${inputDepthNearestVec4}, d2);
               } else {
                 dotProd +=
-                    getX(batch, ${d}, xR, xC) *
-                    getW(wR, wC, ${d}, d2);
+                    getX(batch, ${inputDepthNearestVec4}, xR, xC) *
+                    getW(wR, wC, ${inputDepthNearestVec4}, d2);
               }
 
-            } else if (${h===2}) {
+            } else if (${inputDepthVec4Remainder === 2}) {
               vec2 wValues = vec2(
-                getW(wR, wC, ${d}, d2),
-                getW(wR, wC, ${d} + 1, d2)
+                getW(wR, wC, ${inputDepthNearestVec4}, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2)
               );
 
-              if (${g}) {
+              if (${isChannelsLast}) {
                 vec2 xValues = vec2(
-                  getX(batch, xR, xC, ${d}),
-                  getX(batch, xR, xC, ${d} + 1)
+                  getX(batch, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1)
                 );
                 dotProd += dot(xValues, wValues);
               } else {
                 vec2 xValues = vec2(
-                  getX(batch, ${d}, xR, xC),
-                  getX(batch, ${d} + 1, xR, xC)
+                  getX(batch, ${inputDepthNearestVec4}, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC)
                 );
                 dotProd += dot(xValues, wValues);
               }
 
-            } else if (${h===3}) {
+            } else if (${inputDepthVec4Remainder === 3}) {
               vec3 wValues = vec3(
-                getW(wR, wC, ${d}, d2),
-                getW(wR, wC, ${d} + 1, d2),
-                getW(wR, wC, ${d} + 2, d2)
+                getW(wR, wC, ${inputDepthNearestVec4}, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 1, d2),
+                getW(wR, wC, ${inputDepthNearestVec4} + 2, d2)
               );
 
-              if (${g}) {
+              if (${isChannelsLast}) {
                 vec3 xValues = vec3(
-                  getX(batch, xR, xC, ${d}),
-                  getX(batch, xR, xC, ${d} + 1),
-                  getX(batch, xR, xC, ${d} + 2)
+                  getX(batch, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 1),
+                  getX(batch, xR, xC, ${inputDepthNearestVec4} + 2)
                 );
                 dotProd += dot(xValues, wValues);
               } else {
                 vec3 xValues = vec3(
-                  getX(batch, ${d}, xR, xC),
-                  getX(batch, ${d} + 1, xR, xC),
-                  getX(batch, ${d} + 2, xR, xC)
+                  getX(batch, ${inputDepthNearestVec4}, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 1, xR, xC),
+                  getX(batch, ${inputDepthNearestVec4} + 2, xR, xC)
                 );
                 dotProd += dot(xValues, wValues);
               }
@@ -2322,13 +56799,34 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
 
         float result = dotProd;
-        ${_}
-        ${N}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}},pC=class{constructor(t){this.variableNames=["x","W"],this.outputShape=t.outShape;let e=t.padInfo.front,n=t.padInfo.top,o=t.padInfo.left,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=t.dilationDepth,l=t.dilationHeight,c=t.dilationWidth,p=t.filterDepth,m=t.filterHeight,f=t.filterWidth,d=Math.floor(t.inChannels/4)*4,h=t.inChannels%4;this.userCode=`
-      const ivec3 strides = ivec3(${s}, ${i}, ${a});
-      const ivec3 pads = ivec3(${e}, ${n}, ${o});
+    `;
+  }
+};
+var Conv3DProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const inputDepthNearestVec4 = Math.floor(convInfo.inChannels / 4) * 4;
+    const inputDepthVec4Remainder = convInfo.inChannels % 4;
+    this.userCode = `
+      const ivec3 strides = ivec3(${strideDepth}, ${strideHeight}, ${strideWidth});
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2344,28 +56842,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // y(yF, yR, yC, d2). ? = to be determined. : = across all
         // values in that axis.
         float dotProd = 0.0;
-        for (int wF = 0; wF < ${p}; wF++) {
-          int xF = xFCorner + wF * ${u};
+        for (int wF = 0; wF < ${filterDepth}; wF++) {
+          int xF = xFCorner + wF * ${dilationDepth};
 
-          if (xF < 0 || xF >= ${t.inDepth}) {
+          if (xF < 0 || xF >= ${convInfo.inDepth}) {
             continue;
           }
 
-          for (int wR = 0; wR < ${m}; wR++) {
-            int xR = xRCorner + wR * ${l};
+          for (int wR = 0; wR < ${filterHeight}; wR++) {
+            int xR = xRCorner + wR * ${dilationHeight};
 
-            if (xR < 0 || xR >= ${t.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int wC = 0; wC < ${f}; wC++) {
-              int xC = xCCorner + wC * ${c};
+            for (int wC = 0; wC < ${filterWidth}; wC++) {
+              int xC = xCCorner + wC * ${dilationWidth};
 
-              if (xC < 0 || xC >= ${t.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
-              for (int d1 = 0; d1 < ${d}; d1 += 4) {
+              for (int d1 = 0; d1 < ${inputDepthNearestVec4}; d1 += 4) {
                 vec4 xValues = vec4(
                   getX(batch, xF, xR, xC, d1),
                   getX(batch, xF, xR, xC, d1 + 1),
@@ -2382,30 +56880,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                 dotProd += dot(xValues, wValues);
               }
 
-              if (${h===1}) {
+              if (${inputDepthVec4Remainder === 1}) {
                 dotProd +=
-                  getX(batch, xF, xR, xC, ${d}) *
-                  getW(wF, wR, wC, ${d}, d2);
-              } else if (${h===2}) {
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}) *
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2);
+              } else if (${inputDepthVec4Remainder === 2}) {
                 vec2 xValues = vec2(
-                  getX(batch, xF, xR, xC, ${d}),
-                  getX(batch, xF, xR, xC, ${d} + 1)
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1)
                 );
                 vec2 wValues = vec2(
-                  getW(wF, wR, wC, ${d}, d2),
-                  getW(wF, wR, wC, ${d} + 1, d2)
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2)
                 );
                 dotProd += dot(xValues, wValues);
-              } else if (${h===3}) {
+              } else if (${inputDepthVec4Remainder === 3}) {
                 vec3 xValues = vec3(
-                  getX(batch, xF, xR, xC, ${d}),
-                  getX(batch, xF, xR, xC, ${d} + 1),
-                  getX(batch, xF, xR, xC, ${d} + 2)
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4}),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 1),
+                  getX(batch, xF, xR, xC, ${inputDepthNearestVec4} + 2)
                 );
                 vec3 wValues = vec3(
-                  getW(wF, wR, wC, ${d}, d2),
-                  getW(wF, wR, wC, ${d} + 1, d2),
-                  getW(wF, wR, wC, ${d} + 2, d2)
+                  getW(wF, wR, wC, ${inputDepthNearestVec4}, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 1, d2),
+                  getW(wF, wR, wC, ${inputDepthNearestVec4} + 2, d2)
                 );
                 dotProd += dot(xValues, wValues);
               }
@@ -2414,41 +56912,84 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}};var kd=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.padInfo.left,a=t.strideWidth,u=t.dilationWidth,l=t.filterHeight,c=t.filterWidth,p=c,m=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_packed_gpu.js
+var Conv2DPackedProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const padLeft = convInfo.padInfo.left;
+    const strideWidth = convInfo.strideWidth;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const texelsAcross = filterWidth;
+    let mainLoop = `
        int xR; int xC; int xCOffset;
-       vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<c;g++)m+=`
-           vec4 xTexelC${g*2};
-           int xTexelC${g*2}Ready;
-           vec4 xTexelC${g*2+1};
-           int xTexelC${g*2+1}Ready;
-           vec4 xC${g};`;m+=`
-     for (int r = 0; r < ${l}; r++) {
-      for (int d1 = 0; d1 < ${t.inChannels}; d1 += 2) {
-       `;for(let g=0;g<c;g++)m+=`
-           xTexelC${g*2} = vec4(0.0);
-           xTexelC${g*2}Ready = 0;
-           xTexelC${g*2+1} = vec4(0.0);
-           xTexelC${g*2+1}Ready = 0;
-           xC${g} = vec4(0.0);`;m+=`
+       vec4 wTexel; vec4 previous; vec4 final;`;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+           vec4 xTexelC${c * 2};
+           int xTexelC${c * 2}Ready;
+           vec4 xTexelC${c * 2 + 1};
+           int xTexelC${c * 2 + 1}Ready;
+           vec4 xC${c};`;
+    }
+    mainLoop += `
+     for (int r = 0; r < ${filterHeight}; r++) {
+      for (int d1 = 0; d1 < ${convInfo.inChannels}; d1 += 2) {
+       `;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+           xTexelC${c * 2} = vec4(0.0);
+           xTexelC${c * 2}Ready = 0;
+           xTexelC${c * 2 + 1} = vec4(0.0);
+           xTexelC${c * 2 + 1}Ready = 0;
+           xC${c} = vec4(0.0);`;
+    }
+    mainLoop += `
          xR = xRCorner + r * dilations[0];
          if (xR >=0 && xR < inDims[0]) {
-       `;for(let g=0;g<(p+1)/2;g++){let x=g*2;if(m+=`
-           xC = xCCorner + ${x*u};
-           `,a===1){if(x<c&&(i%2===1?(m+=`
+       `;
+    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {
+      const colIndex = texelC * 2;
+      mainLoop += `
+           xC = xCCorner + ${colIndex * dilationWidth};
+           `;
+      if (strideWidth === 1) {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                  xCOffset = xC + 1;
-                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x}Ready == 0) {
-                   xTexelC${x} = getX(batch, xR, xCOffset, d1);
+                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
 
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${x}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${x}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
-               `,u===1&&x>0?m+=`
-                 xC${x} = vec4(xTexelC${x-2}.zw, xTexelC${x}.xy);
-                 `:m+=`
+               `;
+            if (dilationWidth === 1 && colIndex > 0) {
+              mainLoop += `
+                 xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);
+                 `;
+            } else {
+              mainLoop += `
                    xCOffset = xC + 1 - 2;
 
                    if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -2460,137 +57001,206 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                        previous.zw = vec2(0.0);
                      }
 
-                     xC${x} = vec4(previous.zw, xTexelC${x}.xy);
+                     xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);
                    } else {
-                     xC${x} = vec4(0.0, 0.0, xTexelC${x}.xy);
+                     xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);
                    }
-                   `):m+=`
-                 if (xC >= 0 && xC < inDims[1] && xTexelC${x}Ready == 0) {
-                   xTexelC${x} = getX(batch, xR, xC, d1);
+                   `;
+            }
+          } else {
+            mainLoop += `
+                 if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
-                     xTexelC${x}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${x}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
-                 xC${x} = xTexelC${x};
-                 `,x+1<c)){let b=i%2===0?y.nearestLargerEven(u):u;u%2===0&&i%2===1||u%2!==0&&i%2!==1?(m+=`
-                   xCOffset = xC + imod(pads[1], 2) + ${b};
+                 xC${colIndex} = xTexelC${colIndex};
+                 `;
+          }
+          if (colIndex + 1 < filterWidth) {
+            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;
+            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {
+              mainLoop += `
+                   xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};
 
-                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {
-                     xTexelC${x+1} = getX(batch, xR, xCOffset, d1);
+                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                     xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
 
                      // Need to manually clear unused channels in case
                      // we're reading from recycled texture.
                      if (xCOffset + 1 >= inDims[1]) {
-                       xTexelC${x+1}.zw = vec2(0.0);
+                       xTexelC${colIndex + 1}.zw = vec2(0.0);
                      }
-                     xTexelC${x+1}Ready = 1;
+                     xTexelC${colIndex + 1}Ready = 1;
                    }
-                   `,u>1?m+=`
+                   `;
+              if (dilationWidth > 1) {
+                mainLoop += `
                      xCOffset -= 2;
                      if (xCOffset >= 0 && xCOffset < inDims[1]) {
                       previous = getX(batch, xR, xCOffset, d1);
-                      xC${x+1} = vec4(previous.zw, xTexelC${x+1}.xy);
+                      xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);
                      } else {
-                      xC${x+1} = vec4(0.0, 0.0, xTexelC${x+1}.xy);
+                      xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);
                      }
-                     `:m+=`
-                     xC${x+1} = vec4(xTexelC${x}.zw, xTexelC${x+1}.xy);
-                     `):b===1?m+=`
-                     xC${x+1} = xTexelC${x};
-                     `:m+=`
-                     xCOffset = xC + ${b};
+                     `;
+              } else {
+                mainLoop += `
+                     xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);
+                     `;
+              }
+            } else {
+              if (nextTexelOffset === 1) {
+                mainLoop += `
+                     xC${colIndex + 1} = xTexelC${colIndex};
+                     `;
+              } else {
+                mainLoop += `
+                     xCOffset = xC + ${nextTexelOffset};
 
-                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {
-                       xTexelC${x+1} = getX(batch, xR, xCOffset, d1);
+                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                       xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                        if (xCOffset + 1 >= inDims[1]) {
-                         xTexelC${x+1}.zw = vec2(0.0);
+                         xTexelC${colIndex + 1}.zw = vec2(0.0);
                        }
-                       xTexelC${x+1}Ready = 1;
+                       xTexelC${colIndex + 1}Ready = 1;
                      }
 
-                     xC${x+1} = xTexelC${x+1};
-                     `}}else x<c&&(i%2===1?(m+=`
+                     xC${colIndex + 1} = xTexelC${colIndex + 1};
+                     `;
+              }
+            }
+          }
+        }
+      } else {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                  xCOffset = xC + 1 - strides[1];
-                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x}Ready == 0) {
-                   xTexelC${x} = getX(batch, xR, xCOffset, d1);
+                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${x}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${x}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
-                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${x+1}Ready == 0) {
-                   xTexelC${x+1} = getX(batch, xR, xC + 1, d1);
+                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                   xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);
                    // Need to manually clear unused channels in case
                    // we're reading from recycled texture.
                    if (xC + 2 >= inDims[1]) {
-                     xTexelC${x+1}.zw = vec2(0.0);
+                     xTexelC${colIndex + 1}.zw = vec2(0.0);
                    }
-                   xTexelC${x+1}Ready = 1;
+                   xTexelC${colIndex + 1}Ready = 1;
                  }
 
-                 xC${x} = vec4(xTexelC${x}.zw, xTexelC${x+1}.zw);
-               `,x+1<c&&(m+=`
+                 xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+               `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
                    final = vec4(0.0);
                    xCOffset = xC + 1 + strides[1];
                    if(xCOffset >= 0 && xCOffset < inDims[1]) {
                      final = getX(batch, xR, xCOffset, d1);
                    }
-                   xC${x+1} = vec4(xTexelC${x+1}.xy, final.xy);
-                 `)):(m+=`
-                 if(xC >= 0 && xC < inDims[1] && xTexelC${x}Ready == 0) {
-                   xTexelC${x} = getX(batch, xR, xC, d1);
+                   xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);
+                 `;
+            }
+          } else {
+            mainLoop += `
+                 if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                   xTexelC${colIndex} = getX(batch, xR, xC, d1);
                    if (xC + 1 >= inDims[1]) {
-                     xTexelC${x}.zw = vec2(0.0);
+                     xTexelC${colIndex}.zw = vec2(0.0);
                    }
-                   xTexelC${x}Ready = 1;
+                   xTexelC${colIndex}Ready = 1;
                  }
 
                  xCOffset = xC + strides[1];
-                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${x+1}Ready == 0) {
-                   xTexelC${x+1} = getX(batch, xR, xCOffset, d1);
+                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                   xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                    if (xCOffset + 1 >= inDims[1]) {
-                     xTexelC${x+1}.zw = vec2(0.);
+                     xTexelC${colIndex + 1}.zw = vec2(0.);
                    }
-                   xTexelC${x+1}Ready = 1;
+                   xTexelC${colIndex + 1}Ready = 1;
                  }
 
-                 xC${x} = vec4(
-                   xTexelC${x}.xy, xTexelC${x+1}.xy);
-               `,x+1<c&&(m+=`
-                   xC${x+1} = vec4(xTexelC${x}.zw, xTexelC${x+1}.zw);
-                 `)));x<c&&(m+=`
-             wTexel = getW(r, ${x}, d1, d2);
-             dotProd += xC${x}.xxzz * vec4(wTexel.xy, wTexel.xy);
-             if(d1 + 1 < ${t.inChannels}) {
-               dotProd += xC${x}.yyww * vec4(wTexel.zw, wTexel.zw);
+                 xC${colIndex} = vec4(
+                   xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);
+               `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
+                   xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+                 `;
+            }
+          }
+        }
+      }
+      if (colIndex < filterWidth) {
+        mainLoop += `
+             wTexel = getW(r, ${colIndex}, d1, d2);
+             dotProd += xC${colIndex}.xxzz * vec4(wTexel.xy, wTexel.xy);
+             if(d1 + 1 < ${convInfo.inChannels}) {
+               dotProd += xC${colIndex}.yyww * vec4(wTexel.zw, wTexel.zw);
              }
-           `,x+1<c&&(m+=`
-               wTexel = getW(r, ${x+1}, d1, d2);
-               dotProd += xC${x+1}.xxzz * vec4(wTexel.xy, wTexel.xy);
-               if(d1 + 1 < ${t.inChannels}) {
-                 dotProd += xC${x+1}.yyww * vec4(wTexel.zw, wTexel.zw);
+           `;
+        if (colIndex + 1 < filterWidth) {
+          mainLoop += `
+               wTexel = getW(r, ${colIndex + 1}, d1, d2);
+               dotProd += xC${colIndex + 1}.xxzz * vec4(wTexel.xy, wTexel.xy);
+               if(d1 + 1 < ${convInfo.inChannels}) {
+                 dotProd += xC${colIndex + 1}.yyww * vec4(wTexel.zw, wTexel.zw);
                }
-             `))}m+=`
+             `;
+        }
+      }
+    }
+    mainLoop += `
      }
-   `,m+=`
+   `;
+    mainLoop += `
      }
-   `,m+=`
+   `;
+    mainLoop += `
      }
-   `;let f="",d="";n&&(o?f=`vec4 activation(vec4 a) {
+   `;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
            vec4 b = getPreluActivationWeightsAtOutCoords();
-           ${n}
-         }`:s?f=`vec4 activation(vec4 a) {
+           ${activation2}
+         }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `vec4 activation(vec4 a) {
            vec4 b = getLeakyreluAlphaAtOutCoords();
-           ${n}
-         }`:f=`vec4 activation(vec4 x) {
-           ${n}
-         }`,d="result = activation(result);");let h=e?"result += getBiasAtOutCoords();":"";e&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-       ${f}
+           ${activation2}
+         }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+           ${activation2}
+         }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+       ${activationSnippet}
 
        void main() {
          ivec4 coords = getOutputCoords();
@@ -2603,22 +57213,52 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
          //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
          vec4 dotProd = vec4(0.000000000000001);
 
-         ${m}
+         ${mainLoop}
 
          vec4 result = dotProd - vec4(0.000000000000001);
-         ${h}
-         ${d}
+         ${addBiasSnippet}
+         ${applyActivationSnippet}
          setOutput(result);
        }
-     `}};var mC=class{constructor(t,e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec4"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=t,this.enableShapeUniforms=we(this.outputShape.length);let{dataFormat:n}=e,o=Ge(),s=n==="channelsLast",i=s?1:2,a=s?2:3,u=this.enableShapeUniforms?"if(blockIndex < outShape[2] && pos < outShape[1]) {":`if(blockIndex < ${t[2]} && pos < ${t[1]}) {`,l="";for(let c=0;c<=1;c++)for(let p=0;p<=1;p++)l+=`
-          blockIndex = rc.z + ${p};
-          pos = rc.y + ${c};
+     `;
+  }
+};
 
-          ${u}
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/im2col_packed_gpu.js
+var Im2ColPackedProgram = class {
+  constructor(outputShape, convInfo) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "inputShape", type: "ivec4" },
+      { name: "pad", type: "ivec2" },
+      { name: "stride", type: "ivec2" },
+      { name: "dilation", type: "ivec2" },
+      { name: "inChannels", type: "int" },
+      { name: "itemsPerBlockRow", type: "int" },
+      { name: "outWidth", type: "int" }
+    ];
+    this.outputShape = outputShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const { dataFormat } = convInfo;
+    const glsl = getGlslDifferences();
+    const isChannelsLast = dataFormat === "channelsLast";
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const boundsCheckingSnippet = this.enableShapeUniforms ? "if(blockIndex < outShape[2] && pos < outShape[1]) {" : `if(blockIndex < ${outputShape[2]} && pos < ${outputShape[1]}) {`;
+    let unrolled = ``;
+    for (let row = 0; row <= 1; row++) {
+      for (let col = 0; col <= 1; col++) {
+        unrolled += `
+          blockIndex = rc.z + ${col};
+          pos = rc.y + ${row};
+
+          ${boundsCheckingSnippet}
             offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];
             d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);
 
-            if(d0 < inputShape[${i}] && d0 >= 0) {
+            if(d0 < inputShape[${rowDim}] && d0 >= 0) {
               // Use custom imod instead mod. On Intel GPU, mod may generate
               // unexpected value.
               // https://github.com/tensorflow/tfjs/issues/5447
@@ -2626,25 +57266,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
               d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /
                   inChannels);
 
-              if(d1 < inputShape[${a}] && d1 >= 0) {
+              if(d1 < inputShape[${colDim}] && d1 >= 0) {
 
                 ch = imod(pos, inChannels);
 
-                if (${s}) {
+                if (${isChannelsLast}) {
                   innerDims = vec2(d1, ch);
-                  result[${c*2+p}] = getChannel(
+                  result[${row * 2 + col}] = getChannel(
                     getA(rc.x, d0, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 } else {
                   innerDims = vec2(d0, d1);
-                  result[${c*2+p}] = getChannel(
+                  result[${row * 2 + col}] = getChannel(
                     getA(rc.x, ch, int(innerDims.x),
                     int(innerDims.y)), innerDims);
                 }
               }
             }
           }
-        `;this.userCode=`
+        `;
+      }
+    }
+    this.userCode = `
       void main() {
         ivec3 rc = getOutputCoords();
 
@@ -2653,11 +57296,253 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         int blockIndex, pos, offsetY, d0, offsetX, d1, ch;
         vec2 innerDims;
 
-        ${l}
+        ${unrolled}
 
-        ${o.output} = result;
+        ${glsl.output} = result;
       }
-    `}};function fC(r,t){let e=r.length;return e>=3?t?[...r.slice(0,-3),r[e-3]*r[e-2],r[e-1]]:[...r.slice(0,-3),r[e-3],r[e-2]*r[e-1]]:!t&&e===1&&r[0]>1?[r[0],1]:null}function dC({x:r,filter:t,convInfo:e,backend:n,bias:o=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:a=null}){let u=r.shape,l=n.texData.get(r.dataId),c=e.inChannels,p=u[0]*u[1]*u[2],m=e.outChannels,f=e.dataFormat==="channelsLast",d=!1,h=!1,g,x=[];if(s!=null){let C=fC(s.shape,f);C!=null&&(s=st({inputs:{x:s},backend:n,attrs:{shape:C}}),x.push(s))}if(o!=null){let C=fC(o.shape,f);C!=null&&(o=st({inputs:{x:o},backend:n,attrs:{shape:C}}),x.push(o))}if(!((p===1||m===1)&&c>dk)&&l.isPacked&&f&&l.texture!=null&&u[2]%2!==0&&y.arraysEqual(l.shape.slice(-3),u.slice(-3))){let C=u[0]*u[1]*(u[2]+1),N={dataId:r.dataId,shape:[1,C,e.inChannels],dtype:r.dtype},_=l.shape;l.shape=l.shape.slice(),l.shape[l.shape.length-2]++,y.assert(Eu(l.shape,N.shape),()=>`packed reshape ${l.shape} to ${N.shape} isn't free`);let A=st({inputs:{x:t},backend:n,attrs:{shape:[1,e.inChannels,e.outChannels]}});x.push(A);let $=Uc({a:N,b:A,backend:n,transposeA:d,transposeB:h,bias:o,activation:a,preluActivationWeights:s,leakyreluAlpha:i}),F=n.texData.get($.dataId);y.assert(F.isPacked,()=>"batchMatMul result is expected to be packed"),l.shape=_,F.shape=e.outShape,g=tr({inputs:{x:$},backend:n}),g.shape=e.outShape,x.push($)}else{let C=e.outHeight*e.outWidth,N=st({inputs:{x:r},backend:n,attrs:{shape:f?[e.batchSize,C,e.inChannels]:[e.batchSize,e.inChannels,C]}}),_=st({inputs:{x:t},backend:n,attrs:{shape:[1,e.inChannels,e.outChannels]}}),A=Uc({a:f?N:_,b:f?_:N,transposeA:!f,transposeB:h,backend:n,bias:o,activation:a,preluActivationWeights:s,leakyreluAlpha:i});g=st({inputs:{x:A},backend:n,attrs:{shape:e.outShape}}),x.push(N),x.push(_),x.push(A)}for(let C of x)n.disposeIntermediateTensorInfo(C);return g}function hC({x:r,filter:t,convInfo:e,backend:n,bias:o=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:a=null}){let{filterWidth:u,filterHeight:l,inChannels:c,outWidth:p,outHeight:m,dataFormat:f}=e,d=f==="channelsLast",h=u*l*c,g=m*p,x=[e.batchSize,h,g],b=!0,w=!1,C=[];if(s!=null){let Z=fC(s.shape,d);Z!=null&&(s=st({inputs:{x:s},backend:n,attrs:{shape:Z}}),C.push(s))}if(o!=null){let Z=fC(o.shape,d);Z!=null&&(o=st({inputs:{x:o},backend:n,attrs:{shape:Z}}),C.push(o))}let N=st({inputs:{x:t},backend:n,attrs:{shape:[1,h,y.sizeFromShape(t.shape)/h]}});C.push(N);let _=new mC(x,e),A=[r.shape,[e.padInfo.top,e.padInfo.left],[e.strideHeight,e.strideWidth],[e.dilationHeight,e.dilationWidth],[e.inChannels],[e.filterWidth*e.inChannels],[e.outWidth]],$=n.runWebGLProgram(_,[r],"float32",A),F=st({inputs:{x:$},backend:n,attrs:{shape:x}});C.push($),C.push(F);let P=o!=null,V=s!=null,G=a==="leakyrelu",W=a?bl(a,!0):null,q=new vd(d?F.shape:N.shape,d?N.shape:F.shape,d?[e.batchSize,g,e.outChannels]:[e.batchSize,e.outChannels,g],b,w,P,W,V,G),H=d?[F,N]:[N,F];if(o&&H.push(o),V&&H.push(s),G){let Z=n.makeTensorInfo([],"float32",y.createScalarValue(i,"float32"));H.push(Z),C.push(Z)}let j=n.runWebGLProgram(q,H,"float32"),Y=st({inputs:{x:j},backend:n,attrs:{shape:e.outShape}});C.push(j);for(let Z of C)n.disposeIntermediateTensorInfo(Z);return Y}function art(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dataFormat:u,dilations:l,dimRoundingMode:c}=n,p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,s.shape,i,l,a,c,!1,p),f;if(m.filterHeight===1&&m.filterWidth===1&&m.dilationHeight===1&&m.dilationWidth===1&&m.strideHeight===1&&m.strideWidth===1&&(m.padInfo.type==="SAME"||m.padInfo.type==="VALID"))f=dC({x:o,filter:s,convInfo:m,backend:e});else if(m.strideWidth<=2&&p==="channelsLast"&&z().getBool("WEBGL_EXP_CONV")){let h=new kd(m),g=[[m.padInfo.top,m.padInfo.left],[m.strideHeight,m.strideWidth],[m.dilationHeight,m.dilationWidth],[m.inHeight,m.inWidth]];f=e.runWebGLProgram(h,[o,s],"float32",g)}else if(z().getBool("WEBGL_CONV_IM2COL"))f=hC({x:o,filter:s,convInfo:m,backend:e});else{let h=new Td(m);f=e.runWebGLProgram(h,[o,s],"float32")}let d=st({inputs:{x:f},backend:e,attrs:{shape:m.outShape}});return e.disposeIntermediateTensorInfo(f),d}var fz={kernelName:Ko,backendName:"webgl",kernelFunc:art};var gC=class{constructor(t){this.variableNames=["x","dy"],this.outputShape=t.filterShape;let e=t.strideHeight,n=t.strideWidth,o=t.padInfo.top,s=t.padInfo.left,i=t.dataFormat==="channelsLast";this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2D_impl.js
+function getShapeForBatchMatMul(shape, isChannelsLast) {
+  const length = shape.length;
+  if (length >= 3) {
+    return isChannelsLast ? [
+      ...shape.slice(0, -3),
+      shape[length - 3] * shape[length - 2],
+      shape[length - 1]
+    ] : [
+      ...shape.slice(0, -3),
+      shape[length - 3],
+      shape[length - 2] * shape[length - 1]
+    ];
+  } else if (!isChannelsLast && length === 1 && shape[0] > 1) {
+    return [shape[0], 1];
+  } else {
+    return null;
+  }
+}
+function conv2dByMatMul({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const xShape = x.shape;
+  const xTexData = backend2.texData.get(x.dataId);
+  const sharedMatMulDim = convInfo.inChannels;
+  const outerShapeX = xShape[0] * xShape[1] * xShape[2];
+  const outerShapeFilter = convInfo.outChannels;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const transposeA = false;
+  const transposeB = false;
+  let out;
+  const intermediates = [];
+  if (preluActivationWeights != null) {
+    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);
+    if (targetShape != null) {
+      preluActivationWeights = reshape4({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: targetShape }
+      });
+      intermediates.push(preluActivationWeights);
+    }
+  }
+  if (bias != null) {
+    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);
+    if (targetShape != null) {
+      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });
+      intermediates.push(bias);
+    }
+  }
+  const batchMatMulWillBeUnpacked = (outerShapeX === 1 || outerShapeFilter === 1) && sharedMatMulDim > MATMUL_SHARED_DIM_THRESHOLD;
+  const canOptimize = !batchMatMulWillBeUnpacked && xTexData.isPacked && isChannelsLast && xTexData.texture != null && xShape[2] % 2 !== 0 && util_exports.arraysEqual(xTexData.shape.slice(-3), xShape.slice(-3));
+  if (canOptimize) {
+    const targetShape = xShape[0] * xShape[1] * (xShape[2] + 1);
+    const xReshaped = {
+      dataId: x.dataId,
+      shape: [1, targetShape, convInfo.inChannels],
+      dtype: x.dtype
+    };
+    const originalXTexDataShape = xTexData.shape;
+    xTexData.shape = xTexData.shape.slice();
+    xTexData.shape[xTexData.shape.length - 2]++;
+    util_exports.assert(isReshapeFree(xTexData.shape, xReshaped.shape), () => `packed reshape ${xTexData.shape} to ${xReshaped.shape} isn't free`);
+    const filterReshaped = reshape4({
+      inputs: { x: filter },
+      backend: backend2,
+      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }
+    });
+    intermediates.push(filterReshaped);
+    const pointwiseConv = batchMatMulImpl({
+      a: xReshaped,
+      b: filterReshaped,
+      backend: backend2,
+      transposeA,
+      transposeB,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+    const pointwiseConvTexData = backend2.texData.get(pointwiseConv.dataId);
+    util_exports.assert(pointwiseConvTexData.isPacked, () => "batchMatMul result is expected to be packed");
+    xTexData.shape = originalXTexDataShape;
+    pointwiseConvTexData.shape = convInfo.outShape;
+    out = identity3({ inputs: { x: pointwiseConv }, backend: backend2 });
+    out.shape = convInfo.outShape;
+    intermediates.push(pointwiseConv);
+  } else {
+    const numCols = convInfo.outHeight * convInfo.outWidth;
+    const xReshaped = reshape4({
+      inputs: { x },
+      backend: backend2,
+      attrs: {
+        shape: isChannelsLast ? [convInfo.batchSize, numCols, convInfo.inChannels] : [convInfo.batchSize, convInfo.inChannels, numCols]
+      }
+    });
+    const filterReshaped = reshape4({
+      inputs: { x: filter },
+      backend: backend2,
+      attrs: { shape: [1, convInfo.inChannels, convInfo.outChannels] }
+    });
+    const result = batchMatMulImpl({
+      a: isChannelsLast ? xReshaped : filterReshaped,
+      b: isChannelsLast ? filterReshaped : xReshaped,
+      transposeA: !isChannelsLast,
+      transposeB,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+    out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: convInfo.outShape } });
+    intermediates.push(xReshaped);
+    intermediates.push(filterReshaped);
+    intermediates.push(result);
+  }
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return out;
+}
+function conv2dWithIm2Row({ x, filter, convInfo, backend: backend2, bias = null, preluActivationWeights = null, leakyreluAlpha = 0, activation: activation2 = null }) {
+  const { filterWidth, filterHeight, inChannels, outWidth, outHeight, dataFormat } = convInfo;
+  const isChannelsLast = dataFormat === "channelsLast";
+  const sharedDim = filterWidth * filterHeight * inChannels;
+  const numCols = outHeight * outWidth;
+  const x2ColShape = [convInfo.batchSize, sharedDim, numCols];
+  const transposeA = true;
+  const transposeB = false;
+  const intermediates = [];
+  if (preluActivationWeights != null) {
+    const targetShape = getShapeForBatchMatMul(preluActivationWeights.shape, isChannelsLast);
+    if (targetShape != null) {
+      preluActivationWeights = reshape4({
+        inputs: { x: preluActivationWeights },
+        backend: backend2,
+        attrs: { shape: targetShape }
+      });
+      intermediates.push(preluActivationWeights);
+    }
+  }
+  if (bias != null) {
+    const targetShape = getShapeForBatchMatMul(bias.shape, isChannelsLast);
+    if (targetShape != null) {
+      bias = reshape4({ inputs: { x: bias }, backend: backend2, attrs: { shape: targetShape } });
+      intermediates.push(bias);
+    }
+  }
+  const w2Row = reshape4({
+    inputs: { x: filter },
+    backend: backend2,
+    attrs: { shape: [1, sharedDim, util_exports.sizeFromShape(filter.shape) / sharedDim] }
+  });
+  intermediates.push(w2Row);
+  const im2ColProgram = new Im2ColPackedProgram(x2ColShape, convInfo);
+  const customValues = [
+    x.shape,
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inChannels],
+    [convInfo.filterWidth * convInfo.inChannels],
+    [convInfo.outWidth]
+  ];
+  const im2Col = backend2.runWebGLProgram(im2ColProgram, [x], "float32", customValues);
+  const im2ColReshaped = reshape4({ inputs: { x: im2Col }, backend: backend2, attrs: { shape: x2ColShape } });
+  intermediates.push(im2Col);
+  intermediates.push(im2ColReshaped);
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;
+  const matmulProgram = new MatMulPackedProgram(isChannelsLast ? im2ColReshaped.shape : w2Row.shape, isChannelsLast ? w2Row.shape : im2ColReshaped.shape, isChannelsLast ? [convInfo.batchSize, numCols, convInfo.outChannels] : [convInfo.batchSize, convInfo.outChannels, numCols], transposeA, transposeB, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  const inputs = isChannelsLast ? [im2ColReshaped, w2Row] : [w2Row, im2ColReshaped];
+  if (bias) {
+    inputs.push(bias);
+  }
+  if (hasPreluActivationWeights) {
+    inputs.push(preluActivationWeights);
+  }
+  if (hasLeakyreluAlpha) {
+    const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+    inputs.push($leakyreluAlpha);
+    intermediates.push($leakyreluAlpha);
+  }
+  const product = backend2.runWebGLProgram(matmulProgram, inputs, "float32");
+  const out = reshape4({ inputs: { x: product }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  intermediates.push(product);
+  for (const i of intermediates) {
+    backend2.disposeIntermediateTensorInfo(i);
+  }
+  return out;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2D.js
+function conv2d4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  let out;
+  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === "SAME" || convInfo.padInfo.type === "VALID")) {
+    out = conv2dByMatMul({ x, filter, convInfo, backend: backend2 });
+  } else if (convInfo.strideWidth <= 2 && $dataFormat === "channelsLast" && env().getBool("WEBGL_EXP_CONV")) {
+    const program = new Conv2DPackedProgram(convInfo);
+    const customValues = [
+      [convInfo.padInfo.top, convInfo.padInfo.left],
+      [convInfo.strideHeight, convInfo.strideWidth],
+      [convInfo.dilationHeight, convInfo.dilationWidth],
+      [convInfo.inHeight, convInfo.inWidth]
+    ];
+    out = backend2.runWebGLProgram(program, [x, filter], "float32", customValues);
+  } else if (env().getBool("WEBGL_CONV_IM2COL")) {
+    out = conv2dWithIm2Row({ x, filter, convInfo, backend: backend2 });
+  } else {
+    const program = new Conv2DProgram(convInfo);
+    out = backend2.runWebGLProgram(program, [x, filter], "float32");
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  backend2.disposeIntermediateTensorInfo(out);
+  return outReshaped;
+}
+var conv2DConfig2 = {
+  kernelName: Conv2D,
+  backendName: "webgl",
+  kernelFunc: conv2d4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_backprop_gpu.js
+var Conv2DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
@@ -2669,22 +57554,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int b = 0; b < ${t.batchSize}; b++) {
-          for (int yR = 0; yR < ${t.outHeight}; yR++) {
-            int xR = wR + yR * ${e} - ${o};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+            int xR = wR + yR * ${strideHeight} - ${padTop};
 
-            if (xR < 0 || xR >= ${t.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int yC = 0; yC < ${t.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${s};
+            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+              int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-              if (xC < 0 || xC >= ${t.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
-              if (${i}) {
+              if (${isChannelsLast}) {
                 float dyValue = getDy(b, yR, yC, d2);
                 float xValue = getX(b, xR, xC, d1);
                 dotProd += (xValue * dyValue);
@@ -2699,45 +57584,62 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},xC=class{constructor(t){this.variableNames=["dy","W"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=t.dataFormat==="channelsLast",a=e-1-t.padInfo.top,u=n-1-t.padInfo.left,l=i?1:2,c=i?2:3,p=i?3:1;this.userCode=`
-      const ivec2 pads = ivec2(${a}, ${u});
+    `;
+  }
+};
+var Conv2DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const isChannelsLast = convInfo.dataFormat === "channelsLast";
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    const rowDim = isChannelsLast ? 1 : 2;
+    const colDim = isChannelsLast ? 2 : 3;
+    const channelDim = isChannelsLast ? 3 : 1;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords[0];
-        int d1 = coords[${p}];
+        int d1 = coords[${channelDim}];
 
-        ivec2 dyCorner = ivec2(coords[${l}], coords[${c}]) - pads;
+        ivec2 dyCorner = ivec2(coords[${rowDim}], coords[${colDim}]) - pads;
         int dyRCorner = dyCorner.x;
         int dyCCorner = dyCorner.y;
 
         // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${e}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${o}.0;
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          int wRPerm = ${e} - 1 - wR;
+          int wRPerm = ${filterHeight} - 1 - wR;
 
-          for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${s}.0;
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
-            int wCPerm = ${n} - 1 - wC;
+            int wCPerm = ${filterWidth} - 1 - wC;
 
-            for (int d2 = 0; d2 < ${t.outChannels}; d2++) {
+            for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {
 
-              if (${i}) {
+              if (${isChannelsLast}) {
                 float xValue = getDy(batch, idyR, idyC, d2);
                 float wValue = getW(wRPerm, wCPerm, d1, d2);
                 dotProd += xValue * wValue;
@@ -2752,7 +57654,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},yC=class{constructor(t){this.variableNames=["x","dy"],this.outputShape=t.filterShape;let e=t.strideDepth,n=t.strideHeight,o=t.strideWidth,s=t.padInfo.front,i=t.padInfo.top,a=t.padInfo.left;this.userCode=`
+    `;
+  }
+};
+var Conv3DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padFront = convInfo.padInfo.front;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    this.userCode = `
       void main() {
         ivec5 coords = getOutputCoords();
         int wF = coords.x;
@@ -2763,25 +57678,25 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         float dotProd = 0.0;
 
-        for (int b = 0; b < ${t.batchSize}; b++) {
-          for (int yF = 0; yF < ${t.outDepth}; yF++) {
-            int xF = wF + yF * ${e} - ${s};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yF = 0; yF < ${convInfo.outDepth}; yF++) {
+            int xF = wF + yF * ${strideDepth} - ${padFront};
 
-            if (xF < 0 || xF >= ${t.inDepth}) {
+            if (xF < 0 || xF >= ${convInfo.inDepth}) {
               continue;
             }
 
-            for (int yR = 0; yR < ${t.outHeight}; yR++) {
-              int xR = wR + yR * ${n} - ${i};
+            for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+              int xR = wR + yR * ${strideHeight} - ${padTop};
 
-              if (xR < 0 || xR >= ${t.inHeight}) {
+              if (xR < 0 || xR >= ${convInfo.inHeight}) {
                 continue;
               }
 
-              for (int yC = 0; yC < ${t.outWidth}; yC++) {
-                int xC = wC + yC * ${o} - ${a};
+              for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+                int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-                if (xC < 0 || xC >= ${t.inWidth}) {
+                if (xC < 0 || xC >= ${convInfo.inWidth}) {
                   continue;
                 }
 
@@ -2794,8 +57709,24 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},bC=class{constructor(t){this.variableNames=["dy","W"],this.outputShape=t.inShape;let e=t.filterDepth,n=t.filterHeight,o=t.filterWidth,s=t.strideDepth,i=t.strideHeight,a=t.strideWidth,u=e-1-t.padInfo.front,l=n-1-t.padInfo.top,c=o-1-t.padInfo.left;this.userCode=`
-      const ivec3 pads = ivec3(${u}, ${l}, ${c});
+    `;
+  }
+};
+var Conv3DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterDepth = convInfo.filterDepth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padFront = filterDepth - 1 - convInfo.padInfo.front;
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -2809,39 +57740,39 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         int dyCCorner = dyCorner.z;
 
         float dotProd = 0.0;
-        for (int wF = 0; wF < ${e}; wF++) {
-          float dyF = float(dyFCorner + wF) / ${s}.0;
+        for (int wF = 0; wF < ${filterDepth}; wF++) {
+          float dyF = float(dyFCorner + wF) / ${strideDepth}.0;
 
-          if (dyF < 0.0 || dyF >= ${t.outDepth}.0 || fract(dyF) > 0.0) {
+          if (dyF < 0.0 || dyF >= ${convInfo.outDepth}.0 || fract(dyF) > 0.0) {
             continue;
           }
           int idyF = int(dyF);
 
-          int wFPerm = ${e} - 1 - wF;
+          int wFPerm = ${filterDepth} - 1 - wF;
 
-          for (int wR = 0; wR < ${n}; wR++) {
-            float dyR = float(dyRCorner + wR) / ${i}.0;
+          for (int wR = 0; wR < ${filterHeight}; wR++) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
               fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            int wRPerm = ${n} - 1 - wR;
+            int wRPerm = ${filterHeight} - 1 - wR;
 
-            for (int wC = 0; wC < ${o}; wC++) {
-              float dyC = float(dyCCorner + wC) / ${a}.0;
+            for (int wC = 0; wC < ${filterWidth}; wC++) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
               int idyC = int(dyC);
 
-              int wCPerm = ${o} - 1 - wC;
+              int wCPerm = ${filterWidth} - 1 - wC;
 
-              for (int d2 = 0; d2 < ${t.outChannels}; d2++) {
+              for (int d2 = 0; d2 < ${convInfo.outChannels}; d2++) {
                 float xValue = getDy(batch, idyF, idyR, idyC, d2);
                 float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);
                 dotProd += xValue * wValue;
@@ -2851,14 +57782,142 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}};function lrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,filterShape:c}=n,p=v.convertConv2DDataFormat(u),m=v.computeConv2DInfo(o.shape,c,i,1,a,l,!1,p),f=new gC(m);return e.runWebGLProgram(f,[o,s],"float32")}var dz={kernelName:mp,backendName:"webgl",kernelFunc:lrt};function urt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{inputShape:i,strides:a,pad:u,dataFormat:l,dimRoundingMode:c}=n,p=v.convertConv2DDataFormat(l),m=v.computeConv2DInfo(i,s.shape,a,1,u,c,!1,p),f=new xC(m);return e.runWebGLProgram(f,[o,s],"float32")}var hz={kernelName:jo,backendName:"webgl",kernelFunc:urt};function crt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n,l=v.computeConv3DInfo(o.shape,s.shape,i,u,a),c=new pC(l);return e.runWebGLProgram(c,[o,s],"float32")}var gz={kernelName:Al,backendName:"webgl",kernelFunc:crt};function prt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,pad:a,filterShape:u}=n,l=v.computeConv3DInfo(o.shape,u,i,1,a),c=new yC(l);return e.runWebGLProgram(c,[o,s],"float32")}var xz={kernelName:fp,backendName:"webgl",kernelFunc:prt};function mrt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{pad:i,strides:a,inputShape:u}=n,l=v.computeConv3DInfo(u,s.shape,a,1,i),c=new bC(l);return e.runWebGLProgram(c,[o,s],"float32")}var yz={kernelName:dp,backendName:"webgl",kernelFunc:mrt};var frt=Po+`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2DBackpropFilter.js
+function conv2DBackpropFilter3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, filterShape } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const program = new Conv2DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var conv2DBackpropFilterConfig2 = {
+  kernelName: Conv2DBackpropFilter,
+  backendName: "webgl",
+  kernelFunc: conv2DBackpropFilter3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv2DBackpropInput.js
+function conv2DBackpropInput3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { inputShape, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, 1, pad3, dimRoundingMode, false, $dataFormat);
+  const program = new Conv2DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var conv2DBackpropInputConfig2 = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "webgl",
+  kernelFunc: conv2DBackpropInput3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3D.js
+function conv3D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filter.shape, strides, dilations, pad3);
+  const program = new Conv3DProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, filter], "float32");
+}
+var conv3DConfig2 = {
+  kernelName: Conv3D,
+  backendName: "webgl",
+  kernelFunc: conv3D2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3DBackpropFilterV2.js
+function conv3DBackpropFilterV22(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, pad: pad3, filterShape } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(x.shape, filterShape, strides, 1, pad3);
+  const program = new Conv3DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var conv3DBackpropFilterV2Config2 = {
+  kernelName: Conv3DBackpropFilterV2,
+  backendName: "webgl",
+  kernelFunc: conv3DBackpropFilterV22
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Conv3DBackpropInputV2.js
+function conv3DBackpropInput2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { pad: pad3, strides, inputShape } = attrs;
+  const convInfo = backend_util_exports.computeConv3DInfo(inputShape, filter.shape, strides, 1, pad3);
+  const program = new Conv3DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var conv3DBackpropInputConfig = {
+  kernelName: Conv3DBackpropInputV2,
+  backendName: "webgl",
+  kernelFunc: conv3DBackpropInput2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cos.js
+var COS = CHECK_NAN_SNIPPET_UNARY + `
   return cos(x);
-`,drt=Ct({opSnippet:frt}),bz={kernelName:Xo,backendName:"webgl",kernelFunc:drt};var hrt=`
+`;
+var cos3 = unaryKernelFunc2({ opSnippet: COS });
+var cosConfig2 = {
+  kernelName: Cos,
+  backendName: "webgl",
+  kernelFunc: cos3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cosh.js
+var COSH = `
   float e2x = exp(-x);
   return (e2x + 1.0 / e2x) / 2.0;
-`,grt=Ct({opSnippet:hrt}),wz={kernelName:Yo,backendName:"webgl",kernelFunc:grt};var wC=class{constructor(t,e,n,o,s){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[i,a,u,l]=t,[c]=e,[p,m]=n;this.outputShape=[c,p,m,l];let f=o==="bilinear"?1:0,[d,h]=[`${a-1}.0`,`${u-1}.0`],[g,x,b]=p>1?[`${(a-1)/(p-1)}`,"(y2-y1) * height_ratio",`y1*${d} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${d}`],[w,C,N]=m>1?[`${(u-1)/(m-1)}`,"(x2-x1) * width_ratio",`x1*${h} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${h}`];this.userCode=`
-      const float height_ratio = float(${g});
-      const float width_ratio = float(${w});
+`;
+var cosh3 = unaryKernelFunc2({ opSnippet: COSH });
+var coshConfig2 = {
+  kernelName: Cosh,
+  backendName: "webgl",
+  kernelFunc: cosh3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/crop_and_resize_gpu.js
+var CropAndResizeProgram = class {
+  constructor(imageShape, boxShape, cropSize, method, extrapolationValue) {
+    this.variableNames = ["Image", "Boxes", "BoxInd"];
+    this.outputShape = [];
+    const [batch, imageHeight, imageWidth, depth] = imageShape;
+    const [numBoxes] = boxShape;
+    const [cropHeight, cropWidth] = cropSize;
+    this.outputShape = [numBoxes, cropHeight, cropWidth, depth];
+    const methodId = method === "bilinear" ? 1 : 0;
+    const [inputHeightFloat, inputWidthFloat] = [`${imageHeight - 1}.0`, `${imageWidth - 1}.0`];
+    const [heightRatio, heightScale, inY] = cropHeight > 1 ? [
+      `${(imageHeight - 1) / (cropHeight - 1)}`,
+      "(y2-y1) * height_ratio",
+      `y1*${inputHeightFloat} + float(y)*(height_scale)`
+    ] : [
+      "0.0",
+      "0.0",
+      `0.5 * (y1+y2) * ${inputHeightFloat}`
+    ];
+    const [widthRatio, widthScale, inX] = cropWidth > 1 ? [
+      `${(imageWidth - 1) / (cropWidth - 1)}`,
+      "(x2-x1) * width_ratio",
+      `x1*${inputWidthFloat} + float(x)*(width_scale)`
+    ] : [
+      "0.0",
+      "0.0",
+      `0.5 * (x1+x2) * ${inputWidthFloat}`
+    ];
+    this.userCode = `
+      const float height_ratio = float(${heightRatio});
+      const float width_ratio = float(${widthRatio});
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -2874,26 +57933,26 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         // get image in batch index
         int bInd = round(getBoxInd(b));
-        if(bInd < 0 || bInd >= ${i}) {
+        if(bInd < 0 || bInd >= ${batch}) {
           return;
         }
 
-        float height_scale = ${x};
-        float width_scale = ${C};
+        float height_scale = ${heightScale};
+        float width_scale = ${widthScale};
 
-        float in_y = ${b};
-        if( in_y < 0.0 || in_y > ${d} ) {
-          setOutput(float(${s}));
+        float in_y = ${inY};
+        if( in_y < 0.0 || in_y > ${inputHeightFloat} ) {
+          setOutput(float(${extrapolationValue}));
           return;
         }
-        float in_x = ${N};
-        if( in_x < 0.0 || in_x > ${h} ) {
-          setOutput(float(${s}));
+        float in_x = ${inX};
+        if( in_x < 0.0 || in_x > ${inputWidthFloat} ) {
+          setOutput(float(${extrapolationValue}));
           return;
         }
 
         vec2 sourceFracIndexCR = vec2(in_x,in_y);
-        if(${f} == 1) {
+        if(${methodId} == 1) {
           // Compute the four integer indices.
           ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);
           ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));
@@ -2917,20 +57976,188 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           setOutput(newValue);
         }
       }
-    `}};var xrt=r=>{let{inputs:t,backend:e,attrs:n}=r,{image:o,boxes:s,boxInd:i}=t,{cropSize:a,method:u,extrapolationValue:l}=n,c=new wC(o.shape,s.shape,a,u,l);return e.runWebGLProgram(c,[o,s,i],"float32")},Cz={kernelName:da,backendName:"webgl",kernelFunc:xrt};var qc;(function(r){r.Prod="*",r.Sum="+"})(qc||(qc={}));var ng=class{constructor(t,e,n,o){this.op=t,this.outputShape=e,this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}];let s=this.outputShape.length,i=this.op===qc.Prod?"1.0":"0.0",a=n?i:`getX(${Iz(s,"coords",this.op)})`,u=this.outputShape[this.outputShape.length-1],l="",c="";n?(l=o?`end != ${u-1}`:"end != 0",c=o?"end + 1":"end - 1"):(l=o?`end + pow2 < ${u}`:"end >= pow2",c=o?"end + pow2":"end - pow2"),this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/CropAndResize.js
+var cropAndResize3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, boxes, boxInd } = inputs;
+  const { cropSize, method, extrapolationValue } = attrs;
+  const program = new CropAndResizeProgram(image2.shape, boxes.shape, cropSize, method, extrapolationValue);
+  return backend2.runWebGLProgram(program, [image2, boxes, boxInd], "float32");
+};
+var cropAndResizeConfig2 = {
+  kernelName: CropAndResize,
+  backendName: "webgl",
+  kernelFunc: cropAndResize3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/cum_gpu.js
+var CumOpType;
+(function(CumOpType2) {
+  CumOpType2["Prod"] = "*";
+  CumOpType2["Sum"] = "+";
+})(CumOpType || (CumOpType = {}));
+var CumProgram = class {
+  constructor(op2, outputShape, exclusive, reverse5) {
+    this.op = op2;
+    this.outputShape = outputShape;
+    this.variableNames = ["x"];
+    this.customUniforms = [{ name: "index", type: "float" }];
+    const rank = this.outputShape.length;
+    const initVal = this.op === CumOpType.Prod ? "1.0" : "0.0";
+    const val = exclusive ? initVal : `getX(${getCoords2(rank, "coords", this.op)})`;
+    const length = this.outputShape[this.outputShape.length - 1];
+    let condition = "";
+    let idxString = "";
+    if (exclusive) {
+      condition = reverse5 ? `end != ${length - 1}` : "end != 0";
+      idxString = reverse5 ? "end + 1" : "end - 1";
+    } else {
+      condition = reverse5 ? `end + pow2 < ${length}` : "end >= pow2";
+      idxString = reverse5 ? "end + pow2" : "end - pow2";
+    }
+    this.userCode = `
       void main() {
-        ${zt(s)} coords = getOutputCoords();
-        int end = ${Sz(s,"coords",this.op)};
-        float val = ${a};
+        ${getCoordsDataType(rank)} coords = getOutputCoords();
+        int end = ${getFinalCoord(rank, "coords", this.op)};
+        float val = ${val};
         int pow2 = int(pow(2.0, index));
-        if (${l}) {
-          int idx = ${c};
-          ${Sz(s,"coords",this.op)} = idx;
-          val ${this.op}= getX(${Iz(s,"coords",this.op)});
+        if (${condition}) {
+          int idx = ${idxString};
+          ${getFinalCoord(rank, "coords", this.op)} = idx;
+          val ${this.op}= getX(${getCoords2(rank, "coords", this.op)});
         }
         setOutput(val);
       }
-    `}};function Iz(r,t,e){if(r===1)return`${t}`;if(r===2)return`${t}.x, ${t}.y`;if(r===3)return`${t}.x, ${t}.y, ${t}.z`;if(r===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${e} for rank ${r} is not yet supported`)}function Sz(r,t,e){if(r===1)return`${t}`;if(r===2)return`${t}.y`;if(r===3)return`${t}.z`;if(r===4)return`${t}.w`;throw new Error(`Cumulative ${e} for rank ${r} is not yet supported`)}function CC(r,t,e,n,o,s){let i=t.shape.length,a=v.getAxesPermutation([n],i),u=t;a!=null&&(u=Oe({inputs:{x:t},backend:e,attrs:{perm:a}}));let l=v.getInnerMostAxes(1,i)[0];if(l!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${n}`);let c=u.shape[l],p=tr({inputs:{x:u},backend:e});for(let m=0;m<=Math.ceil(Math.log2(c))-1;m++){let f=new ng(r,u.shape,!1,s),d=[[m]],h=p;p=e.runWebGLProgram(f,[p],p.dtype,d),e.disposeIntermediateTensorInfo(h)}if(o){let m=new ng(r,u.shape,o,s),f=p;p=e.runWebGLProgram(m,[p],p.dtype),e.disposeIntermediateTensorInfo(f)}if(a!=null){let m=v.getUndoAxesPermutation(a),f=Oe({inputs:{x:p},backend:e,attrs:{perm:m}});return e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(u),f}return p}function yrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;return CC(qc.Prod,o,e,s,i,a)}var vz={kernelName:fa,backendName:"webgl",kernelFunc:yrt};function brt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n;return CC(qc.Sum,o,e,s,i,a)}var Nz={kernelName:Zo,backendName:"webgl",kernelFunc:brt};function wrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,weights:s}=t,{size:i,binaryOutput:a}=n;if(o.shape.length===1){let u=e.readSync(o.dataId),l=e.readSync(s.dataId),c=Lw(u,l,s.dtype,s.shape,i);return e.makeTensorInfo([i],s.dtype,c)}else if(o.shape.length===2){let u=e.bufferSync(o),l=e.bufferSync(s),c=pL(u,l,i,a);return e.makeTensorInfo(c.shape,s.dtype,c.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${o.shape.length}.`)}var Tz={kernelName:hp,backendName:"webgl",kernelFunc:wrt};var IC=class{constructor(t,e,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=t,this.blockSize=e,this.dataFormat=n,this.userCode=`
+    `;
+  }
+};
+function getCoords2(rank, name, op2) {
+  if (rank === 1) {
+    return `${name}`;
+  } else if (rank === 2) {
+    return `${name}.x, ${name}.y`;
+  } else if (rank === 3) {
+    return `${name}.x, ${name}.y, ${name}.z`;
+  } else if (rank === 4) {
+    return `${name}.x, ${name}.y, ${name}.z, ${name}.w`;
+  } else {
+    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);
+  }
+}
+function getFinalCoord(rank, name, op2) {
+  if (rank === 1) {
+    return `${name}`;
+  } else if (rank === 2) {
+    return `${name}.y`;
+  } else if (rank === 3) {
+    return `${name}.z`;
+  } else if (rank === 4) {
+    return `${name}.w`;
+  } else {
+    throw new Error(`Cumulative ${op2} for rank ${rank} is not yet supported`);
+  }
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cum_impl.js
+function cumImpl(op2, x, backend2, axis, exclusive, reverse5) {
+  const xRank = x.shape.length;
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  if (permutedAxis !== xRank - 1) {
+    throw new Error(`WebGL cumprod shader expects an inner-most axis=${x.shape.length - 1} but got axis=${axis}`);
+  }
+  const size = permutedX.shape[permutedAxis];
+  let result = identity3({ inputs: { x: permutedX }, backend: backend2 });
+  for (let i = 0; i <= Math.ceil(Math.log2(size)) - 1; i++) {
+    const program = new CumProgram(op2, permutedX.shape, false, reverse5);
+    const customValues = [[i]];
+    const prevResult = result;
+    result = backend2.runWebGLProgram(program, [result], result.dtype, customValues);
+    backend2.disposeIntermediateTensorInfo(prevResult);
+  }
+  if (exclusive) {
+    const program = new CumProgram(op2, permutedX.shape, exclusive, reverse5);
+    const prevResult = result;
+    result = backend2.runWebGLProgram(program, [result], result.dtype);
+    backend2.disposeIntermediateTensorInfo(prevResult);
+  }
+  if (permutation != null) {
+    const reversePermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    const reverseTransposedResult = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm: reversePermutation } });
+    backend2.disposeIntermediateTensorInfo(result);
+    backend2.disposeIntermediateTensorInfo(permutedX);
+    return reverseTransposedResult;
+  }
+  return result;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cumprod.js
+function cumprod3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  return cumImpl(CumOpType.Prod, x, backend2, axis, exclusive, reverse5);
+}
+var cumprodConfig2 = {
+  kernelName: Cumprod,
+  backendName: "webgl",
+  kernelFunc: cumprod3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Cumsum.js
+function cumsum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  return cumImpl(CumOpType.Sum, x, backend2, axis, exclusive, reverse5);
+}
+var cumsumConfig2 = {
+  kernelName: Cumsum,
+  backendName: "webgl",
+  kernelFunc: cumsum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DenseBincount.js
+function denseBincount3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, weights } = inputs;
+  const { size, binaryOutput } = attrs;
+  if (x.shape.length === 1) {
+    const xVals = backend2.readSync(x.dataId);
+    const weightsVals = backend2.readSync(weights.dataId);
+    const outVals = bincountImplCPU(xVals, weightsVals, weights.dtype, weights.shape, size);
+    return backend2.makeTensorInfo([size], weights.dtype, outVals);
+  } else if (x.shape.length === 2) {
+    const xBuf = backend2.bufferSync(x);
+    const weightsBuf = backend2.bufferSync(weights);
+    const outBuf = bincountReduceImplCPU(xBuf, weightsBuf, size, binaryOutput);
+    return backend2.makeTensorInfo(outBuf.shape, weights.dtype, outBuf.values);
+  }
+  throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${x.shape.length}.`);
+}
+var denseBincountConfig2 = {
+  kernelName: DenseBincount,
+  backendName: "webgl",
+  kernelFunc: denseBincount3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/depth_to_space_gpu.js
+var DepthToSpaceProgram = class {
+  constructor(outputShape, blockSize, dataFormat) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    this.outputShape = outputShape;
+    this.blockSize = blockSize;
+    this.dataFormat = dataFormat;
+    this.userCode = `
     void main() {
       ivec4 coords = getOutputCoords();
       int b = coords[0];
@@ -2938,37 +58165,134 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
       int w = ${this.getWidthCoordString()};
       int d = ${this.getDepthCoordString()};
 
-      int in_h = h / ${e};
-      int offset_h = imod(h, ${e});
-      int in_w = w / ${e};
-      int offset_w = imod(w, ${e});
-      int offset_d = (offset_h * ${e} + offset_w) *
+      int in_h = h / ${blockSize};
+      int offset_h = imod(h, ${blockSize});
+      int in_w = w / ${blockSize};
+      int offset_w = imod(w, ${blockSize});
+      int offset_d = (offset_h * ${blockSize} + offset_w) *
         ${this.getOutputDepthSize()};
       int in_d = d + offset_d;
 
       float result = ${this.getInputSamplingString()};
       setOutput(result);
     }
-  `}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function Crt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockSize:s,dataFormat:i}=n,a=o.shape[0],u=i==="NHWC"?o.shape[1]:o.shape[2],l=i==="NHWC"?o.shape[2]:o.shape[3],c=i==="NHWC"?o.shape[3]:o.shape[1],p=u*s,m=l*s,f=c/(s*s),d=i==="NHWC"?[a,p,m,f]:[a,f,p,m],h=new IC(d,s,i);return e.runWebGLProgram(h,[o],o.dtype)}var kz={kernelName:ha,backendName:"webgl",kernelFunc:Crt};var Ed=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.filterHeight,a=t.filterWidth,u=t.outChannels/t.inChannels,l="",c="";n&&(o?l=`float activation(float a) {
+  `;
+  }
+  getHeightCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[1]`;
+    } else {
+      return `coords[2]`;
+    }
+  }
+  getWidthCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[2]`;
+    } else {
+      return `coords[3]`;
+    }
+  }
+  getDepthCoordString() {
+    if (this.dataFormat === "NHWC") {
+      return `coords[3]`;
+    } else {
+      return `coords[1]`;
+    }
+  }
+  getOutputDepthSize() {
+    if (this.dataFormat === "NHWC") {
+      return this.outputShape[3];
+    } else {
+      return this.outputShape[1];
+    }
+  }
+  getInputSamplingString() {
+    if (this.dataFormat === "NHWC") {
+      return `getX(b, in_h, in_w, in_d)`;
+    } else {
+      return `getX(b, in_d, in_h, in_w)`;
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthToSpace.js
+function depthToSpace3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  const batchSize = x.shape[0];
+  const inputHeight = dataFormat === "NHWC" ? x.shape[1] : x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? x.shape[2] : x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? x.shape[3] : x.shape[1];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const outputShape = dataFormat === "NHWC" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];
+  const program = new DepthToSpaceProgram(outputShape, blockSize, dataFormat);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var depthToSpaceConfig2 = {
+  kernelName: DepthToSpace,
+  backendName: "webgl",
+  kernelFunc: depthToSpace3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_gpu_depthwise.js
+var DepthwiseConv2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `float activation(float a) {
           float b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:s?l=`float activation(float a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `float activation(float a) {
           float b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:l=`
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `
           float activation(float x) {
-            ${n}
+            ${activation2}
           }
-        `,c="result = activation(result);");let p=e?"result += getBiasAtOutCoords();":"";e&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${l}
+        `;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${u};
-        int q = d2 - d1 * ${u};
+        int d1 = d2 / ${channelMul};
+        int q = d2 - d1 * ${channelMul};
 
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
@@ -2977,14 +58301,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
         // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.
-        for (int wR = 0; wR < ${i}; wR++) {
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
           int xR = xRCorner + wR * dilations[0];
 
           if (xR < 0 || xR >= inDims[0]) {
             continue;
           }
 
-          for (int wC = 0; wC < ${a}; wC++) {
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
             int xC = xCCorner + wC * dilations[1];
 
             if (xC < 0 || xC >= inDims[1]) {
@@ -2998,44 +58322,88 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
 
         float result = dotProd;
-        ${p}
-        ${c}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}};var _d=class{constructor(t,e=!1,n=null,o=!1,s=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=t.outShape,this.enableShapeUniforms=we(this.outputShape.length);let i=t.outChannels/t.inChannels,a=t.padInfo.left,u=t.strideWidth,l=t.dilationWidth,c=t.filterHeight,p=t.filterWidth,m=p,f=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_packed_gpu_depthwise.js
+var DepthwiseConvPacked2DProgram = class {
+  constructor(convInfo, addBias = false, activation2 = null, hasPreluActivation = false, hasLeakyReluAlpha = false) {
+    this.variableNames = ["x", "W"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [
+      { name: "pads", type: "ivec2" },
+      { name: "strides", type: "ivec2" },
+      { name: "dilations", type: "ivec2" },
+      { name: "inDims", type: "ivec2" }
+    ];
+    this.outputShape = convInfo.outShape;
+    this.enableShapeUniforms = useShapeUniforms(this.outputShape.length);
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    const padLeft = convInfo.padInfo.left;
+    const strideWidth = convInfo.strideWidth;
+    const dilationWidth = convInfo.dilationWidth;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const texelsAcross = filterWidth;
+    let mainLoop = `
       int xR; int xC; int xCOffset;
-      vec4 wTexel; vec4 previous; vec4 final;`;for(let x=0;x<p;x++)f+=`
-          vec4 xTexelC${x*2};
-          int xTexelC${x*2}Ready;
-          vec4 xTexelC${x*2+1};
-          int xTexelC${x*2+1}Ready;
-          vec4 xC${x};`;f+=`
-    for (int r = 0; r < ${c}; r++) {
-      `;for(let x=0;x<p;x++)f+=`
-          xTexelC${x*2} = vec4(0.0);
-          xTexelC${x*2}Ready = 0;
-          xTexelC${x*2+1} = vec4(0.0);
-          xTexelC${x*2+1}Ready = 0;
-          xC${x} = vec4(0.0);`;f+=`
+      vec4 wTexel; vec4 previous; vec4 final;`;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+          vec4 xTexelC${c * 2};
+          int xTexelC${c * 2}Ready;
+          vec4 xTexelC${c * 2 + 1};
+          int xTexelC${c * 2 + 1}Ready;
+          vec4 xC${c};`;
+    }
+    mainLoop += `
+    for (int r = 0; r < ${filterHeight}; r++) {
+      `;
+    for (let c = 0; c < filterWidth; c++) {
+      mainLoop += `
+          xTexelC${c * 2} = vec4(0.0);
+          xTexelC${c * 2}Ready = 0;
+          xTexelC${c * 2 + 1} = vec4(0.0);
+          xTexelC${c * 2 + 1}Ready = 0;
+          xC${c} = vec4(0.0);`;
+    }
+    mainLoop += `
         xR = xRCorner + r * dilations[0];
         if (xR >=0 && xR < inDims[0]) {
-      `;for(let x=0;x<(m+1)/2;x++){let b=x*2;if(f+=`
-          xC = xCCorner + ${b*l};
-          `,u===1){if(b<p&&(a%2===1?(f+=`
+      `;
+    for (let texelC = 0; texelC < (texelsAcross + 1) / 2; texelC++) {
+      const colIndex = texelC * 2;
+      mainLoop += `
+          xC = xCCorner + ${colIndex * dilationWidth};
+          `;
+      if (strideWidth === 1) {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                 xCOffset = xC + 1;
-                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {
-                  xTexelC${b} = getX(batch, xR, xCOffset, d1);
+                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
 
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${b}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${b}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
-              `,l===1&&b>0?f+=`
-                xC${b} = vec4(xTexelC${b-2}.zw, xTexelC${b}.xy);
-                `:f+=`
+              `;
+            if (dilationWidth === 1 && colIndex > 0) {
+              mainLoop += `
+                xC${colIndex} = vec4(xTexelC${colIndex - 2}.zw, xTexelC${colIndex}.xy);
+                `;
+            } else {
+              mainLoop += `
                   xCOffset = xC + 1 - 2;
 
                   if (xCOffset >= 0 && xCOffset < inDims[1]) {
@@ -3047,174 +58415,287 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
                       previous.zw = vec2(0.0);
                     }
 
-                    xC${b} = vec4(previous.zw, xTexelC${b}.xy);
+                    xC${colIndex} = vec4(previous.zw, xTexelC${colIndex}.xy);
                   } else {
-                    xC${b} = vec4(0.0, 0.0, xTexelC${b}.xy);
+                    xC${colIndex} = vec4(0.0, 0.0, xTexelC${colIndex}.xy);
                   }
-                  `):f+=`
-                if (xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {
-                  xTexelC${b} = getX(batch, xR, xC, d1);
+                  `;
+            }
+          } else {
+            mainLoop += `
+                if (xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${b}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${b}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
-                xC${b} = xTexelC${b};
-                `,b+1<p)){let w=a%2===0?y.nearestLargerEven(l):l;l%2===0&&a%2===1||l%2!==0&&a%2!==1?(f+=`
-                  xCOffset = xC + imod(pads[1], 2) + ${w};
+                xC${colIndex} = xTexelC${colIndex};
+                `;
+          }
+          if (colIndex + 1 < filterWidth) {
+            const nextTexelOffset = padLeft % 2 === 0 ? util_exports.nearestLargerEven(dilationWidth) : dilationWidth;
+            if (dilationWidth % 2 === 0 && padLeft % 2 === 1 || dilationWidth % 2 !== 0 && padLeft % 2 !== 1) {
+              mainLoop += `
+                  xCOffset = xC + imod(pads[1], 2) + ${nextTexelOffset};
 
-                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
-                    xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
+                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                    xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
 
                     // Need to manually clear unused channels in case
                     // we're reading from recycled texture.
                     if (xCOffset + 1 >= inDims[1]) {
-                      xTexelC${b+1}.zw = vec2(0.0);
+                      xTexelC${colIndex + 1}.zw = vec2(0.0);
                     }
-                    xTexelC${b+1}Ready = 1;
+                    xTexelC${colIndex + 1}Ready = 1;
                   }
-                  `,l>1?f+=`
+                  `;
+              if (dilationWidth > 1) {
+                mainLoop += `
                     xCOffset -= 2;
                     if (xCOffset >= 0 && xCOffset < inDims[1]) {
                      previous = getX(batch, xR, xCOffset, d1);
-                     xC${b+1} = vec4(previous.zw, xTexelC${b+1}.xy);
+                     xC${colIndex + 1} = vec4(previous.zw, xTexelC${colIndex + 1}.xy);
                     } else {
-                     xC${b+1} = vec4(0.0, 0.0, xTexelC${b+1}.xy);
+                     xC${colIndex + 1} = vec4(0.0, 0.0, xTexelC${colIndex + 1}.xy);
                     }
-                    `:f+=`
-                    xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.xy);
-                    `):w===1?f+=`
-                    xC${b+1} = xTexelC${b};
-                    `:f+=`
-                    xCOffset = xC + ${w};
+                    `;
+              } else {
+                mainLoop += `
+                    xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.xy);
+                    `;
+              }
+            } else {
+              if (nextTexelOffset === 1) {
+                mainLoop += `
+                    xC${colIndex + 1} = xTexelC${colIndex};
+                    `;
+              } else {
+                mainLoop += `
+                    xCOffset = xC + ${nextTexelOffset};
 
-                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
-                      xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
+                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                      xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                       if (xCOffset + 1 >= inDims[1]) {
-                        xTexelC${b+1}.zw = vec2(0.0);
+                        xTexelC${colIndex + 1}.zw = vec2(0.0);
                       }
-                      xTexelC${b+1}Ready = 1;
+                      xTexelC${colIndex + 1}Ready = 1;
                     }
 
-                    xC${b+1} = xTexelC${b+1};
-                    `}}else b<p&&(a%2===1?(f+=`
+                    xC${colIndex + 1} = xTexelC${colIndex + 1};
+                    `;
+              }
+            }
+          }
+        }
+      } else {
+        if (colIndex < filterWidth) {
+          if (padLeft % 2 === 1) {
+            mainLoop += `
                 xCOffset = xC + 1 - strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b}Ready == 0) {
-                  xTexelC${b} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xCOffset, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${b}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${b}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
-                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${b+1}Ready == 0) {
-                  xTexelC${b+1} = getX(batch, xR, xC + 1, d1);
+                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                  xTexelC${colIndex + 1} = getX(batch, xR, xC + 1, d1);
                   // Need to manually clear unused channels in case
                   // we're reading from recycled texture.
                   if (xC + 2 >= inDims[1]) {
-                    xTexelC${b+1}.zw = vec2(0.0);
+                    xTexelC${colIndex + 1}.zw = vec2(0.0);
                   }
-                  xTexelC${b+1}Ready = 1;
+                  xTexelC${colIndex + 1}Ready = 1;
                 }
 
-                xC${b} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);
-              `,b+1<p&&(f+=`
+                xC${colIndex} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+              `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
                   final = vec4(0.0);
                   xCOffset = xC + 1 + strides[1];
                   if(xCOffset >= 0 && xCOffset < inDims[1]) {
                     final = getX(batch, xR, xCOffset, d1);
                   }
-                  xC${b+1} = vec4(xTexelC${b+1}.xy, final.xy);
-                `)):(f+=`
-                if(xC >= 0 && xC < inDims[1] && xTexelC${b}Ready == 0) {
-                  xTexelC${b} = getX(batch, xR, xC, d1);
+                  xC${colIndex + 1} = vec4(xTexelC${colIndex + 1}.xy, final.xy);
+                `;
+            }
+          } else {
+            mainLoop += `
+                if(xC >= 0 && xC < inDims[1] && xTexelC${colIndex}Ready == 0) {
+                  xTexelC${colIndex} = getX(batch, xR, xC, d1);
                   if (xC + 1 >= inDims[1]) {
-                    xTexelC${b}.zw = vec2(0.0);
+                    xTexelC${colIndex}.zw = vec2(0.0);
                   }
-                  xTexelC${b}Ready = 1;
+                  xTexelC${colIndex}Ready = 1;
                 }
 
                 xCOffset = xC + strides[1];
-                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${b+1}Ready == 0) {
-                  xTexelC${b+1} = getX(batch, xR, xCOffset, d1);
+                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${colIndex + 1}Ready == 0) {
+                  xTexelC${colIndex + 1} = getX(batch, xR, xCOffset, d1);
                   if (xCOffset + 1 >= inDims[1]) {
-                    xTexelC${b+1}.zw = vec2(0.);
+                    xTexelC${colIndex + 1}.zw = vec2(0.);
                   }
-                  xTexelC${b+1}Ready = 1;
+                  xTexelC${colIndex + 1}Ready = 1;
                 }
 
-                xC${b} = vec4(
-                  xTexelC${b}.xy, xTexelC${b+1}.xy);
-              `,b+1<p&&(f+=`
-                  xC${b+1} = vec4(xTexelC${b}.zw, xTexelC${b+1}.zw);
-                `)));b<p&&(f+=`
-            wTexel = getW(r, ${b}, d1, q);
-            dotProd += xC${b} * vec4(wTexel.xz, wTexel.xz);
-          `,b+1<p&&(f+=`
-              wTexel = getW(r, ${b+1}, d1, q);
-              dotProd += xC${b+1} * vec4(wTexel.xz, wTexel.xz);
-            `))}f+=`
-    }
-  `,f+=`
+                xC${colIndex} = vec4(
+                  xTexelC${colIndex}.xy, xTexelC${colIndex + 1}.xy);
+              `;
+            if (colIndex + 1 < filterWidth) {
+              mainLoop += `
+                  xC${colIndex + 1} = vec4(xTexelC${colIndex}.zw, xTexelC${colIndex + 1}.zw);
+                `;
+            }
+          }
+        }
       }
-    `;let d="",h="";n&&(o?d=`vec4 activation(vec4 a) {
+      if (colIndex < filterWidth) {
+        mainLoop += `
+            wTexel = getW(r, ${colIndex}, d1, q);
+            dotProd += xC${colIndex} * vec4(wTexel.xz, wTexel.xz);
+          `;
+        if (colIndex + 1 < filterWidth) {
+          mainLoop += `
+              wTexel = getW(r, ${colIndex + 1}, d1, q);
+              dotProd += xC${colIndex + 1} * vec4(wTexel.xz, wTexel.xz);
+            `;
+        }
+      }
+    }
+    mainLoop += `
+    }
+  `;
+    mainLoop += `
+      }
+    `;
+    let activationSnippet = "", applyActivationSnippet = "";
+    if (activation2) {
+      if (hasPreluActivation) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getPreluActivationWeightsAtOutCoords();
-          ${n}
-        }`:s?d=`vec4 activation(vec4 a) {
+          ${activation2}
+        }`;
+      } else if (hasLeakyReluAlpha) {
+        activationSnippet = `vec4 activation(vec4 a) {
           vec4 b = getLeakyreluAlphaAtOutCoords();
-          ${n}
-        }`:d=`vec4 activation(vec4 x) {
-          ${n}
-        }`,h="result = activation(result);");let g=e?"result += getBiasAtOutCoords();":"";e&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),s&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
-      ${d}
+          ${activation2}
+        }`;
+      } else {
+        activationSnippet = `vec4 activation(vec4 x) {
+          ${activation2}
+        }`;
+      }
+      applyActivationSnippet = `result = activation(result);`;
+    }
+    const addBiasSnippet = addBias ? "result += getBiasAtOutCoords();" : "";
+    if (addBias) {
+      this.variableNames.push("bias");
+    }
+    if (hasPreluActivation) {
+      this.variableNames.push("preluActivationWeights");
+    }
+    if (hasLeakyReluAlpha) {
+      this.variableNames.push("leakyreluAlpha");
+    }
+    this.userCode = `
+      ${activationSnippet}
 
       void main() {
         ivec4 coords = getOutputCoords();
         int batch = coords.x;
         ivec2 xRCCorner = coords.yz * strides - pads;
         int d2 = coords.w;
-        int d1 = d2 / ${i};
-        int q = d2 - d1 * ${i};
+        int d1 = d2 / ${channelMul};
+        int q = d2 - d1 * ${channelMul};
         int xRCorner = xRCCorner.x;
         int xCCorner = xRCCorner.y;
 
         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
         vec4 dotProd = vec4(0.000000000000001);
 
-        ${f}
+        ${mainLoop}
 
         vec4 result = dotProd - vec4(0.000000000000001);
-        ${g}
-        ${h}
+        ${addBiasSnippet}
+        ${applyActivationSnippet}
         setOutput(result);
       }
-    `}};function Irt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u,dimRoundingMode:l}=n,c=u;c==null&&(c=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(i,c),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${c}'`);let p=v.computeConv2DInfo(o.shape,s.shape,i,c,a,l,!0),m;z().getBool("WEBGL_PACK_DEPTHWISECONV")&&p.strideWidth<=2&&p.outChannels/p.inChannels===1?m=new _d(p):m=new Ed(p);let f=[[p.padInfo.top,p.padInfo.left],[p.strideHeight,p.strideWidth],[p.dilationHeight,p.dilationWidth],[p.inHeight,p.inWidth]];return e.runWebGLProgram(m,[o,s],"float32",f)}var Ez={kernelName:Jo,backendName:"webgl",kernelFunc:Irt};var SC=class{constructor(t){this.variableNames=["x","dy"],this.outputShape=t.filterShape;let e=t.strideHeight,n=t.strideWidth,o=t.padInfo.top,s=t.padInfo.left,i=t.outChannels/t.inChannels;this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNative.js
+function depthwiseConv2dNative2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode } = attrs;
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  let program;
+  if (env().getBool("WEBGL_PACK_DEPTHWISECONV") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1) {
+    program = new DepthwiseConvPacked2DProgram(convInfo);
+  } else {
+    program = new DepthwiseConv2DProgram(convInfo);
+  }
+  const customValues = [
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inHeight, convInfo.inWidth]
+  ];
+  return backend2.runWebGLProgram(program, [x, filter], "float32", customValues);
+}
+var depthwiseConv2dNativeConfig2 = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNative2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/conv_backprop_gpu_depthwise.js
+var DepthwiseConv2DDerFilterProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "dy"];
+    this.outputShape = convInfo.filterShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = convInfo.padInfo.top;
+    const padLeft = convInfo.padInfo.left;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int wR = coords.x;
         int wC = coords.y;
         int d1 = coords.z;
         int dm = coords.w;
-        int d2 = d1 * ${i} + dm;
+        int d2 = d1 * ${channelMul} + dm;
 
         float dotProd = 0.0;
 
         // TO DO: Vec4 over the batch size
-        for (int b = 0; b < ${t.batchSize}; b++) {
-          for (int yR = 0; yR < ${t.outHeight}; yR++) {
-            int xR = wR + yR * ${e} - ${o};
+        for (int b = 0; b < ${convInfo.batchSize}; b++) {
+          for (int yR = 0; yR < ${convInfo.outHeight}; yR++) {
+            int xR = wR + yR * ${strideHeight} - ${padTop};
 
-            if (xR < 0 || xR >= ${t.inHeight}) {
+            if (xR < 0 || xR >= ${convInfo.inHeight}) {
               continue;
             }
 
-            for (int yC = 0; yC < ${t.outWidth}; yC++) {
-              int xC = wC + yC * ${n} - ${s};
+            for (int yC = 0; yC < ${convInfo.outWidth}; yC++) {
+              int xC = wC + yC * ${strideWidth} - ${padLeft};
 
-              if (xC < 0 || xC >= ${t.inWidth}) {
+              if (xC < 0 || xC >= ${convInfo.inWidth}) {
                 continue;
               }
 
@@ -3226,8 +58707,22 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},vC=class{constructor(t){this.variableNames=["dy","W"],this.outputShape=t.inShape;let e=t.filterHeight,n=t.filterWidth,o=t.strideHeight,s=t.strideWidth,i=e-1-t.padInfo.top,a=n-1-t.padInfo.left,u=t.outChannels/t.inChannels;this.userCode=`
-      const ivec2 pads = ivec2(${i}, ${a});
+    `;
+  }
+};
+var DepthwiseConv2DDerInputProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "W"];
+    this.outputShape = convInfo.inShape;
+    const filterHeight = convInfo.filterHeight;
+    const filterWidth = convInfo.filterWidth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const padTop = filterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = filterWidth - 1 - convInfo.padInfo.left;
+    const channelMul = convInfo.outChannels / convInfo.inChannels;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3239,30 +58734,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         float dotProd = 0.0;
 
-        for (int wR = 0; wR < ${e}; wR++) {
-          float dyR = float(dyRCorner + wR) / ${o}.0;
+        for (int wR = 0; wR < ${filterHeight}; wR++) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          int wRPerm = ${e} - 1 - wR;
+          int wRPerm = ${filterHeight} - 1 - wR;
 
-          for (int wC = 0; wC < ${n}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${s}.0;
+          for (int wC = 0; wC < ${filterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
-            int wCPerm = ${n} - 1 - wC;
+            int wCPerm = ${filterWidth} - 1 - wC;
 
             // TO DO: Vec4 over the channelMul
-            for (int dm = 0; dm < ${u}; dm++) {
-              int d2 = d1 * ${u} + dm;
+            for (int dm = 0; dm < ${channelMul}; dm++) {
+              int d2 = d1 * ${channelMul} + dm;
               float xValue = getDy(batch, idyR, idyC, d2);
               float wValue = getW(wRPerm, wCPerm, d1, dm);
               dotProd += xValue * wValue;
@@ -3271,15 +58766,85 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}};function Srt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,dy:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,filterShape:c}=n,p=v.computeConv2DInfo(o.shape,c,i,a,u,l,!0),m=new SC(p);return e.runWebGLProgram(m,[o,s],"float32")}var _z={kernelName:gp,backendName:"webgl",kernelFunc:Srt};function vrt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,filter:s}=t,{strides:i,dilations:a,pad:u,dimRoundingMode:l,inputShape:c}=n,p=v.computeConv2DInfo(c,s.shape,i,a,u,l,!0),m=new vC(p);return e.runWebGLProgram(m,[o,s],"float32")}var Az={kernelName:xp,backendName:"webgl",kernelFunc:vrt};var NC=class{constructor(t){this.variableNames=["X"],this.outputShape=[t,t],this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNativeBackpropFilter.js
+function depthwiseConv2dNativeBackpropFilter3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, dy } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, filterShape } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filterShape, strides, dilations, pad3, dimRoundingMode, true);
+  const program = new DepthwiseConv2DDerFilterProgram(convInfo);
+  return backend2.runWebGLProgram(program, [x, dy], "float32");
+}
+var depthwiseConv2dNativeBackpropFilterConfig2 = {
+  kernelName: DepthwiseConv2dNativeBackpropFilter,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNativeBackpropFilter3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/DepthwiseConv2dNativeBackpropInput.js
+function depthwiseConv2dNativeBackpropInput3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, dilations, pad: pad3, dimRoundingMode, inputShape } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const program = new DepthwiseConv2DDerInputProgram(convInfo);
+  return backend2.runWebGLProgram(program, [dy, filter], "float32");
+}
+var depthwiseConv2dNativeBackpropInputConfig2 = {
+  kernelName: DepthwiseConv2dNativeBackpropInput,
+  backendName: "webgl",
+  kernelFunc: depthwiseConv2dNativeBackpropInput3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/diag_gpu.js
+var DiagProgram = class {
+  constructor(size) {
+    this.variableNames = ["X"];
+    this.outputShape = [size, size];
+    this.userCode = `
       void main() {
           ivec2 coords = getOutputCoords();
           float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;
           setOutput(val);
       }
-    `}};function Nrt(r){let{inputs:t,backend:e}=r,{x:n}=t,o=[...n.shape,...n.shape],s=y.sizeFromShape(n.shape),i=st({inputs:{x:n},backend:e,attrs:{shape:[s]}}),a=new NC(s),u=e.runWebGLProgram(a,[i],i.dtype),l=st({inputs:{x:u},backend:e,attrs:{shape:o}});return e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(u),l}var $z={kernelName:yp,backendName:"webgl",kernelFunc:Nrt};var TC=class{constructor(t){this.variableNames=["x","W"],this.outputShape=t.outShape;let{inHeight:e,inWidth:n,padInfo:o,strideHeight:s,strideWidth:i,filterHeight:a,filterWidth:u,dilationHeight:l,dilationWidth:c}=t,{top:p,left:m}=o;this.userCode=`
-      const ivec2 strides = ivec2(${s}, ${i});
-      const ivec2 pads = ivec2(${p}, ${m});
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Diag.js
+function diag3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  const outShape = [...x.shape, ...x.shape];
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const flat = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [xSize] } });
+  const program = new DiagProgram(xSize);
+  const res = backend2.runWebGLProgram(program, [flat], flat.dtype);
+  const out = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outShape } });
+  backend2.disposeIntermediateTensorInfo(flat);
+  backend2.disposeIntermediateTensorInfo(res);
+  return out;
+}
+var diagConfig2 = {
+  kernelName: Diag,
+  backendName: "webgl",
+  kernelFunc: diag3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/dilation_gpu.js
+var Dilation2DProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["x", "W"];
+    this.outputShape = convInfo.outShape;
+    const { inHeight, inWidth, padInfo, strideHeight, strideWidth, filterHeight, filterWidth, dilationHeight, dilationWidth } = convInfo;
+    const { top: padTop, left: padLeft } = padInfo;
+    this.userCode = `
+      const ivec2 strides = ivec2(${strideHeight}, ${strideWidth});
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
       const float neg_infinity = -3.4e38;
 
       void main() {
@@ -3292,14 +58857,14 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         int wBeg = outTopLeftCorner.y;
 
         float curVal = neg_infinity;
-        for (int h = 0; h < ${a}; h++) {
-          int hIn = hBeg + h * ${l};
+        for (int h = 0; h < ${filterHeight}; h++) {
+          int hIn = hBeg + h * ${dilationHeight};
 
-          if (hIn >= 0 && hIn < ${e}) {
-            for (int w = 0; w < ${u}; w++) {
-              int wIn = wBeg + w * ${c};
+          if (hIn >= 0 && hIn < ${inHeight}) {
+            for (int w = 0; w < ${filterWidth}; w++) {
+              int wIn = wBeg + w * ${dilationWidth};
 
-              if (wIn >= 0 && wIn < ${n}) {
+              if (wIn >= 0 && wIn < ${inWidth}) {
                 float xVal = getX(batch, hIn, wIn, d1);
                 float wVal = getW(h, w, d1);
 
@@ -3315,7 +58880,98 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         float result = curVal;
         setOutput(result);
       }
-    `}};function Trt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s}=t,{strides:i,pad:a,dilations:u}=n,l=v.computeDilation2DInfo(o.shape,s.shape,i,a,"NHWC",u),c,p=new TC(l);c=e.runWebGLProgram(p,[o,s],"float32");let m=st({inputs:{x:c},backend:e,attrs:{shape:l.outShape}});return e.disposeIntermediateTensorInfo(c),m}var Dz={kernelName:$l,backendName:"webgl",kernelFunc:Trt};function krt(r){let{inputs:t,backend:e,attrs:n}=r,{equation:o}=n,s=t,{allDims:i,summedDims:a,idDims:u}=v.decodeEinsumEquation(o,s.length);v.checkEinsumDimSizes(i.length,u,s);let{path:l,steps:c}=v.getEinsumComputePath(a,u),p=c.length,m=null,f=i.length,d=[];for(let h=0;h<p;++h){for(let g of c[h]){let{permutationIndices:x,expandDims:b}=v.getEinsumPermutation(f,u[g]),w;v.isIdentityPermutation(x)?w=s[g]:(w=Oe({inputs:{x:s[g]},backend:e,attrs:{perm:x}}),d.push(w));let C=w.shape.slice();for(let N=0;N<b.length;++N)C.splice(b[N],0,1);y.arraysEqual(w.shape,C)||(w=st({inputs:{x:w},backend:e,attrs:{shape:C}}),d.push(w)),m===null?m=w:(m=eg({inputs:{a:w,b:m},backend:e}),d.push(m))}h<p-1&&(l[h]>=0&&(m=Wc({inputs:{x:m},backend:e,attrs:{axis:l[h]-(i.length-f),keepDims:!1}}),d.push(m)),f--)}for(let h of d)h!==m&&e.disposeIntermediateTensorInfo(h);return m}var Rz={kernelName:bp,backendName:"webgl",kernelFunc:krt};var Ert="return (x >= 0.0) ? x : (exp(x) - 1.0);",_rt=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Dilation2D.js
+function dilation2D(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter } = inputs;
+  const { strides, pad: pad3, dilations } = attrs;
+  const convInfo = backend_util_exports.computeDilation2DInfo(x.shape, filter.shape, strides, pad3, "NHWC", dilations);
+  let out;
+  const program = new Dilation2DProgram(convInfo);
+  out = backend2.runWebGLProgram(program, [x, filter], "float32");
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  backend2.disposeIntermediateTensorInfo(out);
+  return outReshaped;
+}
+var dilation2DConfig2 = {
+  kernelName: Dilation2D,
+  backendName: "webgl",
+  kernelFunc: dilation2D
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Einsum.js
+function einsum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { equation } = attrs;
+  const tensors = inputs;
+  const { allDims, summedDims, idDims } = backend_util_exports.decodeEinsumEquation(equation, tensors.length);
+  backend_util_exports.checkEinsumDimSizes(allDims.length, idDims, tensors);
+  const { path, steps } = backend_util_exports.getEinsumComputePath(summedDims, idDims);
+  const nSteps = steps.length;
+  let out = null;
+  let numDimsRemaining = allDims.length;
+  const tensorsToDispose = [];
+  for (let i = 0; i < nSteps; ++i) {
+    for (const idTerm of steps[i]) {
+      const { permutationIndices: perm, expandDims: dimsToExpand } = backend_util_exports.getEinsumPermutation(numDimsRemaining, idDims[idTerm]);
+      let x;
+      if (backend_util_exports.isIdentityPermutation(perm)) {
+        x = tensors[idTerm];
+      } else {
+        x = transpose3({ inputs: { x: tensors[idTerm] }, backend: backend2, attrs: { perm } });
+        tensorsToDispose.push(x);
+      }
+      const targetShape = x.shape.slice();
+      for (let k = 0; k < dimsToExpand.length; ++k) {
+        targetShape.splice(dimsToExpand[k], 0, 1);
+      }
+      if (!util_exports.arraysEqual(x.shape, targetShape)) {
+        x = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: targetShape } });
+        tensorsToDispose.push(x);
+      }
+      if (out === null) {
+        out = x;
+      } else {
+        out = multiply3({ inputs: { a: x, b: out }, backend: backend2 });
+        tensorsToDispose.push(out);
+      }
+    }
+    if (i < nSteps - 1) {
+      if (path[i] >= 0) {
+        out = sum4({
+          inputs: { x: out },
+          backend: backend2,
+          attrs: {
+            axis: path[i] - (allDims.length - numDimsRemaining),
+            keepDims: false
+          }
+        });
+        tensorsToDispose.push(out);
+      }
+      numDimsRemaining--;
+    }
+  }
+  for (const tensorInfo of tensorsToDispose) {
+    if (tensorInfo === out) {
+      continue;
+    }
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+  return out;
+}
+var einsumConfig2 = {
+  kernelName: Einsum,
+  backendName: "webgl",
+  kernelFunc: einsum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Elu.js
+var ELU4 = `return (x >= 0.0) ? x : (exp(x) - 1.0);`;
+var ELU_PACKED = `
   vec4 result;
 
   result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
@@ -3324,29 +58980,78 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
   result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
 
   return result;
-`,Art=Ct({opSnippet:Ert,packedOpSnippet:_rt}),Fz={kernelName:ts,backendName:"webgl",kernelFunc:Art};var $rt="return (b >= 1.0) ? a : a * (b + 1.0);",Drt=`
+`;
+var elu5 = unaryKernelFunc2({ opSnippet: ELU4, packedOpSnippet: ELU_PACKED });
+var eluConfig2 = {
+  kernelName: Elu,
+  backendName: "webgl",
+  kernelFunc: elu5
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/EluGrad.js
+var ELU_DER = `return (b >= 1.0) ? a : a * (b + 1.0);`;
+var ELU_DER_PACKED = `
   vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));
   return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));
-`,Rrt=r=>{let{inputs:t,backend:e}=r,{dy:n,y:o}=t,s=z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Oo(Drt,n.shape,o.shape):new io($rt,n.shape,o.shape);return e.runWebGLProgram(s,[n,o],n.dtype)},Oz={kernelName:wp,backendName:"webgl",kernelFunc:Rrt};var Frt=`
+`;
+var eluGrad2 = (args) => {
+  const { inputs, backend: backend2 } = args;
+  const { dy, y } = inputs;
+  const program = env().getBool("WEBGL_PACK_BINARY_OPERATIONS") ? new BinaryOpPackedProgram(ELU_DER_PACKED, dy.shape, y.shape) : new BinaryOpProgram(ELU_DER, dy.shape, y.shape);
+  return backend2.runWebGLProgram(program, [dy, y], dy.dtype);
+};
+var eluGradConfig3 = {
+  kernelName: EluGrad,
+  backendName: "webgl",
+  kernelFunc: eluGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Equal.js
+var PACKED_EQUAL = `
   return vec4(equal(a, b));
-`,Ort="return float(a == b);",Prt=le({opSnippet:Ort,packedOpSnippet:Frt,dtype:"bool",cpuKernelImpl:hL}),Pz={kernelName:xa,backendName:"webgl",kernelFunc:Prt};var Lrt=`
+`;
+var EQUAL = `return float(a == b);`;
+var equal3 = binaryKernelFunc2({
+  opSnippet: EQUAL,
+  packedOpSnippet: PACKED_EQUAL,
+  dtype: "bool",
+  cpuKernelImpl: equalImplCPU
+});
+var equalConfig2 = {
+  kernelName: Equal,
+  backendName: "webgl",
+  kernelFunc: equal3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Erf.js
+var ERF = `
   // Error function is calculated approximately with elementary function.
   // See "Handbook of Mathematical Functions with Formulas,
   // Graphs, and Mathematical Tables", Abramowitz and Stegun.
-  float p = ${v.ERF_P};
-  float a1 = ${v.ERF_A1};
-  float a2 = ${v.ERF_A2};
-  float a3 = ${v.ERF_A3};
-  float a4 = ${v.ERF_A4};
-  float a5 = ${v.ERF_A5};
+  float p = ${backend_util_exports.ERF_P};
+  float a1 = ${backend_util_exports.ERF_A1};
+  float a2 = ${backend_util_exports.ERF_A2};
+  float a3 = ${backend_util_exports.ERF_A3};
+  float a4 = ${backend_util_exports.ERF_A4};
+  float a5 = ${backend_util_exports.ERF_A5};
 
   float sign = sign(x);
   x = abs(x);
   float t = 1.0 / (1.0 + p * x);
   return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));
-`,Mrt=Ct({opSnippet:Lrt}),Lz={kernelName:ga,backendName:"webgl",kernelFunc:Mrt};var zrt=Po+`
+`;
+var erf3 = unaryKernelFunc2({ opSnippet: ERF });
+var erfConfig2 = {
+  kernelName: Erf,
+  backendName: "webgl",
+  kernelFunc: erf3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Exp.js
+var EXP = CHECK_NAN_SNIPPET_UNARY + `
   return exp(x);
-`,Brt=`
+`;
+var EXP_PACKED = `
   vec4 result = exp(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : result.r;
@@ -3355,21 +59060,80 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,bk=Ct({opSnippet:zrt,packedOpSnippet:Brt,cpuKernelImpl:gL,dtype:"float32"}),Mz={kernelName:es,backendName:"webgl",kernelFunc:bk};function kC(r){let{inputs:t,attrs:e,backend:n}=r,{dim:o}=e,{input:s}=t,i=s.shape.length,a=s.shape.slice(),u=o;return o<0&&(y.assert(-(i+1)<=o,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+o+1),a.splice(u,0,1),st({inputs:{x:s},backend:n,attrs:{shape:a}})}var zz={kernelName:ui,backendName:"webgl",kernelFunc:kC};var Bz="return exp(x) - 1.0;",Vrt=Ct({opSnippet:Bz,packedOpSnippet:Bz,cpuKernelImpl:xL}),Vz={kernelName:ya,backendName:"webgl",kernelFunc:Vrt};var og=class{constructor(t,e,n){this.variableNames=["real","imag"];let o=e[1];this.outputShape=e;let s=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,i=n?`${o}.0`:"1.0",a;if(t==="real")a="return real * expR - imag * expI;";else if(t==="imag")a="return real * expI + imag * expR;";else throw new Error(`FFT component must be either "real" or "imag", got ${t}.`);this.userCode=`
-      const float exponentMultiplier = ${s};
+`;
+var exp3 = unaryKernelFunc2({
+  opSnippet: EXP,
+  packedOpSnippet: EXP_PACKED,
+  cpuKernelImpl: expImplCPU,
+  dtype: "float32"
+});
+var expConfig2 = {
+  kernelName: Exp,
+  backendName: "webgl",
+  kernelFunc: exp3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ExpandDims.js
+function expandDims4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { dim } = attrs;
+  const { input: input2 } = inputs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape4({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig2 = {
+  kernelName: ExpandDims,
+  backendName: "webgl",
+  kernelFunc: expandDims4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Expm1.js
+var EXPM1 = `return exp(x) - 1.0;`;
+var expm13 = unaryKernelFunc2({ opSnippet: EXPM1, packedOpSnippet: EXPM1, cpuKernelImpl: expm1ImplCPU });
+var expm1Config2 = {
+  kernelName: Expm1,
+  backendName: "webgl",
+  kernelFunc: expm13
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/fft_gpu.js
+var FFTProgram = class {
+  constructor(component, inputShape, inverse) {
+    this.variableNames = ["real", "imag"];
+    const innerDim = inputShape[1];
+    this.outputShape = inputShape;
+    const exponentMultiplierSnippet = inverse ? `2.0 * ${Math.PI}` : `-2.0 * ${Math.PI}`;
+    const resultDenominator = inverse ? `${innerDim}.0` : "1.0";
+    let opString;
+    if (component === "real") {
+      opString = "return real * expR - imag * expI;";
+    } else if (component === "imag") {
+      opString = "return real * expI + imag * expR;";
+    } else {
+      throw new Error(`FFT component must be either "real" or "imag", got ${component}.`);
+    }
+    this.userCode = `
+      const float exponentMultiplier = ${exponentMultiplierSnippet};
 
       float unaryOpComplex(float real, float expR, float imag, float expI) {
-        ${a}
+        ${opString}
       }
 
       float mulMatDFT(int batch, int index) {
-        float indexRatio = float(index) / float(${o});
+        float indexRatio = float(index) / float(${innerDim});
         float exponentMultiplierTimesIndexRatio =
             exponentMultiplier * indexRatio;
 
         float result = 0.0;
 
-        for (int i = 0; i < ${o}; i++) {
+        for (int i = 0; i < ${innerDim}; i++) {
           // x = (-2|2 * PI / N) * index * i;
           float x = exponentMultiplierTimesIndexRatio * float(i);
           float expR = cos(x);
@@ -3378,7 +59142,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           float imag = getImag(batch, i);
 
           result +=
-              unaryOpComplex(real, expR, imag, expI) / ${i};
+              unaryOpComplex(real, expR, imag, expI) / ${resultDenominator};
         }
 
         return result;
@@ -3388,26 +59152,142 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         ivec2 coords = getOutputCoords();
         setOutput(mulMatDFT(coords[0], coords[1]));
       }
-    `}};function EC(r,t,e){let n=e.texData.get(r.dataId),o=y.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=o/s,a=st({inputs:{x:r},backend:e,attrs:{shape:[i,s]}}),u=a.shape,l=new og("real",u,t),c=new og("imag",u,t),p=[{dataId:n.complexTensorInfos.real.dataId,dtype:n.complexTensorInfos.real.dtype,shape:u},{dataId:n.complexTensorInfos.imag.dataId,dtype:n.complexTensorInfos.imag.dtype,shape:u}],m=e.runWebGLProgram(l,p,"float32"),f=e.runWebGLProgram(c,p,"float32"),d=En({inputs:{real:m,imag:f},backend:e});e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f);let h=st({inputs:{x:d},backend:e,attrs:{shape:r.shape}});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(d),h}function Grt(r){let{inputs:t,backend:e}=r,{input:n}=t;return EC(n,!1,e)}var Gz={kernelName:Cp,backendName:"webgl",kernelFunc:Grt};var _C=class{constructor(t,e){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FFT_impl.js
+function fftImpl2(x, inverse, backend2) {
+  const xData = backend2.texData.get(x.dataId);
+  const inputSize = util_exports.sizeFromShape(x.shape);
+  const innerDimensionSize = x.shape[x.shape.length - 1];
+  const batch = inputSize / innerDimensionSize;
+  const input2D = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: [batch, innerDimensionSize] } });
+  const xShape = input2D.shape;
+  const realProgram = new FFTProgram("real", xShape, inverse);
+  const imagProgram = new FFTProgram("imag", xShape, inverse);
+  const inputs = [
+    {
+      dataId: xData.complexTensorInfos.real.dataId,
+      dtype: xData.complexTensorInfos.real.dtype,
+      shape: xShape
+    },
+    {
+      dataId: xData.complexTensorInfos.imag.dataId,
+      dtype: xData.complexTensorInfos.imag.dtype,
+      shape: xShape
+    }
+  ];
+  const realPart = backend2.runWebGLProgram(realProgram, inputs, "float32");
+  const imagPart = backend2.runWebGLProgram(imagProgram, inputs, "float32");
+  const complexOutput = complex3({ inputs: { real: realPart, imag: imagPart }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(realPart);
+  backend2.disposeIntermediateTensorInfo(imagPart);
+  const complexOutputReshaped = reshape4({ inputs: { x: complexOutput }, backend: backend2, attrs: { shape: x.shape } });
+  backend2.disposeIntermediateTensorInfo(input2D);
+  backend2.disposeIntermediateTensorInfo(complexOutput);
+  return complexOutputReshaped;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FFT.js
+function fft3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  return fftImpl2(input2, false, backend2);
+}
+var fftConfig2 = {
+  kernelName: FFT,
+  backendName: "webgl",
+  kernelFunc: fft3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/fill_gpu.js
+var FillProgram = class {
+  constructor(shape, value) {
+    this.outputShape = [];
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.variableNames = ["x"];
+    this.outputShape = shape;
+    this.userCode = `
       void main() {
         // Input can be obtained from uniform value.
         setOutput(value);
       }
-    `}};function Cl(r){let{backend:t,attrs:e}=r,{shape:n,value:o}=e,{dtype:s}=e;if(s=s||y.inferDtype(o),s==="string"){let i=y.getArrayFromDType(s,y.sizeFromShape(n));return i.fill(o),t.makeTensorInfo(n,s,i)}else{let i=new _C(n,o),a=[[o]];return t.runWebGLProgram(i,[],s,a)}}var Wz={kernelName:Dl,backendName:"webgl",kernelFunc:Cl};var AC=class{constructor(t){this.variableNames=["Image"],this.outputShape=[];let e=t[2];this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Fill.js
+function fill3(args) {
+  const { backend: backend2, attrs } = args;
+  const { shape, value } = attrs;
+  let { dtype } = attrs;
+  dtype = dtype || util_exports.inferDtype(value);
+  if (dtype === "string") {
+    const values = util_exports.getArrayFromDType(dtype, util_exports.sizeFromShape(shape));
+    values.fill(value);
+    return backend2.makeTensorInfo(shape, dtype, values);
+  } else {
+    const program = new FillProgram(shape, value);
+    const customValues = [[value]];
+    return backend2.runWebGLProgram(program, [], dtype, customValues);
+  }
+}
+var fillConfig2 = {
+  kernelName: Fill,
+  backendName: "webgl",
+  kernelFunc: fill3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/flip_left_right_gpu.js
+var FlipLeftRightProgram = class {
+  constructor(imageShape) {
+    this.variableNames = ["Image"];
+    this.outputShape = [];
+    const imageWidth = imageShape[2];
+    this.outputShape = imageShape;
+    this.userCode = `
         void main() {
           ivec4 coords = getOutputCoords();
           int x = coords[2];
 
-          int coordX = ${e} - x - 1;
+          int coordX = ${imageWidth} - x - 1;
           float outputValue;
-          if(coordX >= 0 && coordX < ${e}) {
+          if(coordX >= 0 && coordX < ${imageWidth}) {
             outputValue = getImage(coords[0], coords[1], coordX, coords[3]);
           } else {
             outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);
           }
           setOutput(outputValue);
         }
-    `}};var Uz={kernelName:ba,backendName:"webgl",kernelFunc:({inputs:r,backend:t})=>{let{image:e}=r,n=t,o=new AC(e.shape);return n.runWebGLProgram(o,[e],e.dtype)}};var Hz="return floor(x);",Wrt=Ct({opSnippet:Hz,packedOpSnippet:Hz,cpuKernelImpl:yL}),qz={kernelName:rs,backendName:"webgl",kernelFunc:Wrt};var Urt=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FlipLeftRight.js
+var flipLeftRightConfig2 = {
+  kernelName: FlipLeftRight,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const webglBackend = backend2;
+    const program = new FlipLeftRightProgram(image2.shape);
+    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype);
+    return output;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Floor.js
+var FLOOR = `return floor(x);`;
+var floor3 = unaryKernelFunc2({ opSnippet: FLOOR, packedOpSnippet: FLOOR, cpuKernelImpl: floorImplCPU });
+var floorConfig2 = {
+  kernelName: Floor,
+  backendName: "webgl",
+  kernelFunc: floor3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FloorDiv.js
+var INT_DIV = `
   float s = sign(a) * sign(b);
   int ia = round(a);
   int ib = round(b);
@@ -3417,7 +59297,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
   } else {
     return NAN;
   }
-`,Hrt=`
+`;
+var INT_DIV_PACKED = `
   ivec4 ia = round(a);
   ivec4 ib = round(b);
   bvec4 cond = notEqual(ib, ivec4(0));
@@ -3438,15 +59319,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
     result[3] = idiv(ia[3], ib[3], s[3]);
   }
   return vec4(result);
-`,qrt=le({opSnippet:Urt,packedOpSnippet:Hrt,dtype:"int32"}),Kz={kernelName:ns,backendName:"webgl",kernelFunc:qrt};var $C=class{constructor(t){this.variableNames=["A"];let e=Ge(),[n,o]=t;this.outputShape=t,this.userCode=`
+`;
+var floorDiv3 = binaryKernelFunc2({ opSnippet: INT_DIV, packedOpSnippet: INT_DIV_PACKED, dtype: "int32" });
+var floorDivConfig2 = {
+  kernelName: FloorDiv,
+  backendName: "webgl",
+  kernelFunc: floorDiv3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels_utils/from_pixels_gpu.js
+var FromPixelsProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    const glsl = getGlslDifferences();
+    const [height, width] = outputShape;
+    this.outputShape = outputShape;
+    this.userCode = `
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
         int texC = coords[1];
         int depth = coords[2];
-        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${o}.0, ${n}.0);
+        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${width}.0, ${height}.0);
 
-        vec4 values = ${e.texture2D}(A, uv);
+        vec4 values = ${glsl.texture2D}(A, uv);
         float value;
         if (depth == 0) {
           value = values.r;
@@ -3460,7 +59356,20 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         setOutput(floor(value * 255.0 + 0.5));
       }
-    `}};var DC=class{constructor(t){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let e=Ge(),[n,o]=t;this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels_utils/from_pixels_packed_gpu.js
+var FromPixelsPackedProgram = class {
+  constructor(outputShape) {
+    this.variableNames = ["A"];
+    this.packedInputs = false;
+    this.packedOutput = true;
+    const glsl = getGlslDifferences();
+    const [height, width] = outputShape;
+    this.outputShape = outputShape;
+    this.userCode = `
       void main() {
         ivec3 coords = getOutputCoords();
         int texR = coords[0];
@@ -3475,8 +59384,8 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
             depth = coords[2] + col;
 
             vec2 uv = (vec2(texC, texR) + halfCR) /
-                       vec2(${o}.0, ${n}.0);
-            vec4 values = ${e.texture2D}(A, uv);
+                       vec2(${width}.0, ${height}.0);
+            vec4 values = ${glsl.texture2D}(A, uv);
             float value;
             if (depth == 0) {
               value = values.r;
@@ -3492,41 +59401,476 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
           }
         }
 
-        ${e.output} = result;
+        ${glsl.output} = result;
       }
-    `}};var jz={kernelName:Yd,backendName:"webgl",kernelFunc:Krt},Ad,wk=z().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");function Krt(r){let{inputs:t,backend:e,attrs:n}=r,{pixels:o}=t,{numChannels:s}=n,i=typeof HTMLVideoElement!="undefined"&&o instanceof HTMLVideoElement,a=typeof HTMLImageElement!="undefined"&&o instanceof HTMLImageElement,[u,l]=i?[o.videoWidth,o.videoHeight]:[o.width,o.height],c=[l,u],p=[l,u,s];if(a||i){let h=z().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");(Ad==null||h!==wk)&&(wk=h,Ad=document.createElement("canvas").getContext("2d",{willReadFrequently:wk})),Ad.canvas.width=u,Ad.canvas.height=l,Ad.drawImage(o,0,0,u,l),o=Ad.canvas}let m=e.makeTensorInfo(c,"int32");e.texData.get(m.dataId).usage=jr.PIXELS,e.gpgpu.uploadPixelDataToTexture(e.getTexture(m.dataId),o);let f=z().getBool("WEBGL_PACK")?new DC(p):new $C(p),d=e.runWebGLProgram(f,[m],"int32");return e.disposeData(m.dataId),d}function jrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dataFormat:c,dilations:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=n,h=v.convertConv2DDataFormat(c),g=v.computeConv2DInfo(o.shape,s.shape,u,p,l,m,!1,h),x,b=[],w=i!=null,C=a!=null,N=f==="leakyrelu",_=()=>{let $=[o,s],F=(P,V)=>{if(V==="NCHW"&&P.shape.length===1&&P.shape[0]!==1){let G=st({inputs:{x:P},backend:e,attrs:{shape:[P.shape[0],1,1]}});return b.push(G),G}return P};if(w&&$.push(F(i,c)),C&&$.push(F(a,c)),N){let P=e.makeTensorInfo([],"float32",y.createScalarValue(d,"float32"));$.push(P),b.push(P)}return $};if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type==="SAME"||g.padInfo.type==="VALID"))x=dC({x:o,filter:s,convInfo:g,backend:e,bias:i,activation:f,preluActivationWeights:a,leakyreluAlpha:d});else if(g.strideWidth<=2&&h==="channelsLast"&&z().getBool("WEBGL_EXP_CONV")){let $=f?bl(f,!0):null,F=new kd(g,w,$,C,N),P=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],V=_();x=e.runWebGLProgram(F,V,"float32",P)}else if(z().getBool("WEBGL_CONV_IM2COL"))x=hC({x:o,filter:s,convInfo:g,backend:e,bias:i,activation:f,preluActivationWeights:a,leakyreluAlpha:d});else{let $=f?bl(f,!1):null,F=new Td(g,w,$,C,N),P=_();x=e.runWebGLProgram(F,P,"float32")}let A=st({inputs:{x},backend:e,attrs:{shape:g.outShape}});return b.push(x),b.forEach($=>e.disposeIntermediateTensorInfo($)),A}var Xz={kernelName:Ii,backendName:"webgl",kernelFunc:jrt};function Xrt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dimRoundingMode:p,activation:m,leakyreluAlpha:f}=n,d=[],h=c;h==null&&(h=[1,1]),y.assert(v.eitherStridesOrDilationsAreOne(u,h),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${u} and dilations '${h}'`);let g=v.computeConv2DInfo(o.shape,s.shape,u,h,l,p,!0),x=z().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,b=m?bl(m,x):null,w=[o,s],C=i!=null,N=a!=null,_=m==="leakyrelu";if(C&&w.push(i),N&&w.push(a),_){let P=e.makeTensorInfo([],"float32",y.createScalarValue(f,"float32"));w.push(P),d.push(P)}let A;x?A=new _d(g,C,b,N,_):A=new Ed(g,C,b,N,_);let $=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],F=e.runWebGLProgram(A,w,"float32",$);return d.forEach(P=>e.disposeIntermediateTensorInfo(P)),F}var Yz={kernelName:Si,backendName:"webgl",kernelFunc:Xrt};var RC=class{constructor(t,e,n,o){this.sliceDim=t,this.strides=e,this.paramsShape=o,this.variableNames=["x","indices"],this.outputShape=n;let s=zt(n.length),i=`
-    int index;`;for(let a=0;a<this.sliceDim;a++)i+=`
-          index = round(getIndices(coords[0], ${a}));
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FromPixels.js
+var fromPixelsConfig = {
+  kernelName: FromPixels,
+  backendName: "webgl",
+  kernelFunc: fromPixels2
+};
+var fromPixels2DContext2;
+var willReadFrequently = env().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");
+function fromPixels2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  let { pixels } = inputs;
+  const { numChannels } = attrs;
+  const isVideo = typeof HTMLVideoElement !== "undefined" && pixels instanceof HTMLVideoElement;
+  const isImage = typeof HTMLImageElement !== "undefined" && pixels instanceof HTMLImageElement;
+  const [width, height] = isVideo ? [
+    pixels.videoWidth,
+    pixels.videoHeight
+  ] : [pixels.width, pixels.height];
+  const texShape = [height, width];
+  const outShape = [height, width, numChannels];
+  if (isImage || isVideo) {
+    const newWillReadFrequently = env().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");
+    if (fromPixels2DContext2 == null || newWillReadFrequently !== willReadFrequently) {
+      willReadFrequently = newWillReadFrequently;
+      fromPixels2DContext2 = document.createElement("canvas").getContext("2d", { willReadFrequently });
+    }
+    fromPixels2DContext2.canvas.width = width;
+    fromPixels2DContext2.canvas.height = height;
+    fromPixels2DContext2.drawImage(pixels, 0, 0, width, height);
+    pixels = fromPixels2DContext2.canvas;
+  }
+  const tempPixelHandle = backend2.makeTensorInfo(texShape, "int32");
+  backend2.texData.get(tempPixelHandle.dataId).usage = TextureUsage.PIXELS;
+  backend2.gpgpu.uploadPixelDataToTexture(backend2.getTexture(tempPixelHandle.dataId), pixels);
+  const program = env().getBool("WEBGL_PACK") ? new FromPixelsPackedProgram(outShape) : new FromPixelsProgram(outShape);
+  const res = backend2.runWebGLProgram(program, [tempPixelHandle], "int32");
+  backend2.disposeData(tempPixelHandle.dataId);
+  return res;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FusedConv2D.js
+function fusedConv2d(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dataFormat, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  let out;
+  const intermediates = [];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  const prepareInputs = () => {
+    const inputs2 = [x, filter];
+    const alignInputWithDataFormat = (input2, dataFormat2) => {
+      if (dataFormat2 === "NCHW" && input2.shape.length === 1 && input2.shape[0] !== 1) {
+        const alignedInput = reshape4({
+          inputs: { x: input2 },
+          backend: backend2,
+          attrs: { shape: [input2.shape[0], 1, 1] }
+        });
+        intermediates.push(alignedInput);
+        return alignedInput;
+      }
+      return input2;
+    };
+    if (hasBias) {
+      inputs2.push(alignInputWithDataFormat(bias, dataFormat));
+    }
+    if (hasPreluActivationWeights) {
+      inputs2.push(alignInputWithDataFormat(preluActivationWeights, dataFormat));
+    }
+    if (hasLeakyreluAlpha) {
+      const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+      inputs2.push($leakyreluAlpha);
+      intermediates.push($leakyreluAlpha);
+    }
+    return inputs2;
+  };
+  if (convInfo.filterHeight === 1 && convInfo.filterWidth === 1 && convInfo.dilationHeight === 1 && convInfo.dilationWidth === 1 && convInfo.strideHeight === 1 && convInfo.strideWidth === 1 && (convInfo.padInfo.type === "SAME" || convInfo.padInfo.type === "VALID")) {
+    out = conv2dByMatMul({
+      x,
+      filter,
+      convInfo,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+  } else if (convInfo.strideWidth <= 2 && $dataFormat === "channelsLast" && env().getBool("WEBGL_EXP_CONV")) {
+    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, true) : null;
+    const program = new Conv2DPackedProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const customValues = [
+      [convInfo.padInfo.top, convInfo.padInfo.left],
+      [convInfo.strideHeight, convInfo.strideWidth],
+      [convInfo.dilationHeight, convInfo.dilationWidth],
+      [convInfo.inHeight, convInfo.inWidth]
+    ];
+    const inputs2 = prepareInputs();
+    out = backend2.runWebGLProgram(program, inputs2, "float32", customValues);
+  } else if (env().getBool("WEBGL_CONV_IM2COL")) {
+    out = conv2dWithIm2Row({
+      x,
+      filter,
+      convInfo,
+      backend: backend2,
+      bias,
+      activation: activation2,
+      preluActivationWeights,
+      leakyreluAlpha
+    });
+  } else {
+    const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, false) : null;
+    const program = new Conv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+    const inputs2 = prepareInputs();
+    out = backend2.runWebGLProgram(program, inputs2, "float32");
+  }
+  const outReshaped = reshape4({ inputs: { x: out }, backend: backend2, attrs: { shape: convInfo.outShape } });
+  intermediates.push(out);
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return outReshaped;
+}
+var fusedConv2DConfig2 = {
+  kernelName: FusedConv2D,
+  backendName: "webgl",
+  kernelFunc: fusedConv2d
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/FusedDepthwiseConv2D.js
+function fusedDepthwiseConv2D2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const intermediates = [];
+  let $dilations = dilations;
+  if ($dilations == null) {
+    $dilations = [1, 1];
+  }
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, $dilations), () => `Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${strides} and dilations '${$dilations}'`);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const shouldPackDepthwiseConv = env().getBool("WEBGL_PACK_DEPTHWISECONV") && convInfo.strideWidth <= 2 && convInfo.outChannels / convInfo.inChannels === 1;
+  const fusedActivation = activation2 ? mapActivationToShaderProgram(activation2, shouldPackDepthwiseConv) : null;
+  const programInputs = [x, filter];
+  const hasBias = bias != null;
+  const hasPreluActivationWeights = preluActivationWeights != null;
+  const hasLeakyreluAlpha = activation2 === "leakyrelu";
+  if (hasBias) {
+    programInputs.push(bias);
+  }
+  if (hasPreluActivationWeights) {
+    programInputs.push(preluActivationWeights);
+  }
+  if (hasLeakyreluAlpha) {
+    const $leakyreluAlpha = backend2.makeTensorInfo([], "float32", util_exports.createScalarValue(leakyreluAlpha, "float32"));
+    programInputs.push($leakyreluAlpha);
+    intermediates.push($leakyreluAlpha);
+  }
+  let program;
+  if (shouldPackDepthwiseConv) {
+    program = new DepthwiseConvPacked2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  } else {
+    program = new DepthwiseConv2DProgram(convInfo, hasBias, fusedActivation, hasPreluActivationWeights, hasLeakyreluAlpha);
+  }
+  const customValues = [
+    [convInfo.padInfo.top, convInfo.padInfo.left],
+    [convInfo.strideHeight, convInfo.strideWidth],
+    [convInfo.dilationHeight, convInfo.dilationWidth],
+    [convInfo.inHeight, convInfo.inWidth]
+  ];
+  const result = backend2.runWebGLProgram(program, programInputs, "float32", customValues);
+  intermediates.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var fusedDepthwiseConv2DConfig2 = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "webgl",
+  kernelFunc: fusedDepthwiseConv2D2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/gather_nd_gpu.js
+var GatherNDProgram = class {
+  constructor(sliceDim, strides, shape, paramsShape) {
+    this.sliceDim = sliceDim;
+    this.strides = strides;
+    this.paramsShape = paramsShape;
+    this.variableNames = ["x", "indices"];
+    this.outputShape = shape;
+    const dtype = getCoordsDataType(shape.length);
+    let mainLoop = `
+    int index;`;
+    for (let j = 0; j < this.sliceDim; j++) {
+      mainLoop += `
+          index = round(getIndices(coords[0], ${j}));
           out_of_bounds = out_of_bounds || index < 0;
-          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[a]};
-          flattenIndex += index * ${this.strides[a]};`;this.userCode=`
+          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[j]};
+          flattenIndex += index * ${this.strides[j]};`;
+    }
+    this.userCode = `
          void main() {
-          ${s} coords = getOutputCoords();
+          ${dtype} coords = getOutputCoords();
           int flattenIndex = 0;
           bool out_of_bounds = false;
 
-          ${i}
+          ${mainLoop}
 
           setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));
         }
-      `}};function Yrt(r){let{inputs:t,backend:e}=r,{params:n,indices:o}=t,s=o.shape,i=s[s.length-1],a=y.sizeFromShape(n.shape),[u,l,c,p]=v.prepareAndValidate(n,o),m=st({inputs:{x:o},backend:e,attrs:{shape:[l,i]}}),f=st({inputs:{x:n},backend:e,attrs:{shape:[y.sizeFromShape(n.shape)/c,c]}});if(e.shouldExecuteOnCPU([n,o])||n.dtype==="string"){let x=e.readSync(o.dataId),b=e.bufferSync(n),w=bL(x,b,n.dtype,l,i,c,p,n.shape,a);return e.makeTensorInfo(u,n.dtype,w.values)}let d=new RC(i,p,[l,c],n.shape),h=e.runWebGLProgram(d,[f,m],f.dtype),g=st({inputs:{x:h},backend:e,attrs:{shape:u}});return e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(h),g}var Zz={kernelName:wa,backendName:"webgl",kernelFunc:Yrt};var FC=class{constructor(t,e){this.variableNames=["A","indices"],this.outputShape=e,this.rank=e.length;let n=zt(this.rank),o=Zrt(t,2);this.userCode=`
+      `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GatherNd.js
+function gatherNd2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { params, indices } = inputs;
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const paramsSize = util_exports.sizeFromShape(params.shape);
+  const [resultShape, numSlices, sliceSize, strides] = backend_util_exports.prepareAndValidate(params, indices);
+  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numSlices, sliceRank] } });
+  const flattenX = reshape4({
+    inputs: { x: params },
+    backend: backend2,
+    attrs: { shape: [util_exports.sizeFromShape(params.shape) / sliceSize, sliceSize] }
+  });
+  if (backend2.shouldExecuteOnCPU([params, indices]) || params.dtype === "string") {
+    const indicesData = backend2.readSync(indices.dataId);
+    const paramsBuf = backend2.bufferSync(params);
+    const outValue = gatherNdImplCPU(indicesData, paramsBuf, params.dtype, numSlices, sliceRank, sliceSize, strides, params.shape, paramsSize);
+    return backend2.makeTensorInfo(resultShape, params.dtype, outValue.values);
+  }
+  const program = new GatherNDProgram(sliceRank, strides, [numSlices, sliceSize], params.shape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices], flattenX.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: resultShape } });
+  backend2.disposeIntermediateTensorInfo(flattenIndices);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(res);
+  return reshaped;
+}
+var gatherNdConfig2 = {
+  kernelName: GatherNd,
+  backendName: "webgl",
+  kernelFunc: gatherNd2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/gather_gpu.js
+var GatherProgram = class {
+  constructor(aShape, outputShape) {
+    this.variableNames = ["A", "indices"];
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const sourceCoords = getSourceCoords2(aShape, 2);
+    this.userCode = `
       void main() {
-        ${n} resRC = getOutputCoords();
+        ${dtype} resRC = getOutputCoords();
         int index = int(getIndices(resRC.x, resRC.z));
-        float inBounds = (index >= 0) && (index < ${t[2]}) ? 1.0 : 0.0;
-        setOutput(inBounds * getA(${o}));
+        float inBounds = (index >= 0) && (index < ${aShape[2]}) ? 1.0 : 0.0;
+        setOutput(inBounds * getA(${sourceCoords}));
       }
-    `}};function Zrt(r,t){let e=["resRC.x","resRC.y","resRC.z","resRC.w"],n=[];for(let o=0;o<r.length;o++)o===2?n.push("index"):n.push(`${e[o]}`);return n.join()}function Ck(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,indices:s}=t,{axis:i,batchDims:a}=n,u=y.parseAxisParam(i,o.shape)[0];if(z().get("DEBUG")){let b=e.readSync(s.dataId),w=o.shape[u];for(let C=0;C<b.length;++C){let N=b[C];y.assert(N<=w-1&&N>=0,()=>`GatherV2: the index value ${N} is not in [0, ${w-1}]`)}}let l=v.segment_util.collectGatherOpShapeInfo(o,s,u,a),c=y.sizeFromShape(s.shape),p=[],m=st({inputs:{x:o},backend:e,attrs:{shape:[l.batchSize,l.outerSize,l.dimSize,l.sliceSize]}}),f=st({inputs:{x:s},backend:e,attrs:{shape:[l.batchSize,c/l.batchSize]}});p.push(m),p.push(f);let d=[l.batchSize,l.outerSize,c/l.batchSize,l.sliceSize];if(e.shouldExecuteOnCPU([o,s])||o.dtype==="string"){let b=e.bufferSync(f),w=e.bufferSync(m),C=wL(w,b,d);return p.forEach(N=>e.disposeIntermediateTensorInfo(N)),e.makeTensorInfo(l.outputShape,C.dtype,C.values)}let h=new FC(m.shape,d),g=e.runWebGLProgram(h,[m,f],m.dtype);p.push(g);let x=st({inputs:{x:g},backend:e,attrs:{shape:l.outputShape}});return p.forEach(b=>e.disposeIntermediateTensorInfo(b)),x}var Jz={kernelName:ci,backendName:"webgl",kernelFunc:Ck};var Jrt="return float(a > b);",Qrt=`
+    `;
+  }
+};
+function getSourceCoords2(aShape, axis) {
+  const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"];
+  const sourceCoords = [];
+  for (let i = 0; i < aShape.length; i++) {
+    if (i === 2) {
+      sourceCoords.push("index");
+    } else {
+      sourceCoords.push(`${currentCoords[i]}`);
+    }
+  }
+  return sourceCoords.join();
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GatherV2.js
+function gatherV22(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  if (env().get("DEBUG")) {
+    const indicesVals = backend2.readSync(indices.dataId);
+    const axisDim = x.shape[parsedAxis];
+    for (let i = 0; i < indicesVals.length; ++i) {
+      const index = indicesVals[i];
+      util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+    }
+  }
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const toDispose = [];
+  const flattenX = reshape4({
+    inputs: { x },
+    backend: backend2,
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    }
+  });
+  const flattenIndex = reshape4({
+    inputs: { x: indices },
+    backend: backend2,
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] }
+  });
+  toDispose.push(flattenX);
+  toDispose.push(flattenIndex);
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  if (backend2.shouldExecuteOnCPU([x, indices]) || x.dtype === "string") {
+    const indicesBuf = backend2.bufferSync(flattenIndex);
+    const xBuf = backend2.bufferSync(flattenX);
+    const outBuf = gatherV2ImplCPU(xBuf, indicesBuf, flattenOutputShape);
+    toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+    return backend2.makeTensorInfo(shapeInfo.outputShape, outBuf.dtype, outBuf.values);
+  }
+  const program = new GatherProgram(flattenX.shape, flattenOutputShape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndex], flattenX.dtype);
+  toDispose.push(res);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: shapeInfo.outputShape } });
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return reshaped;
+}
+var gatherV2Config2 = {
+  kernelName: GatherV2,
+  backendName: "webgl",
+  kernelFunc: gatherV22
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Greater.js
+var GREATER = `return float(a > b);`;
+var GREATER_PACKED = `
   return vec4(greaterThan(a, b));
-`,tnt=le({opSnippet:Jrt,packedOpSnippet:Qrt,cpuKernelImpl:CL,dtype:"bool"}),Qz={kernelName:Ca,backendName:"webgl",kernelFunc:tnt};var ent="return float(a >= b);",rnt=`
+`;
+var greater4 = binaryKernelFunc2({
+  opSnippet: GREATER,
+  packedOpSnippet: GREATER_PACKED,
+  cpuKernelImpl: greaterImplCPU,
+  dtype: "bool"
+});
+var greaterConfig2 = {
+  kernelName: Greater,
+  backendName: "webgl",
+  kernelFunc: greater4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/GreaterEqual.js
+var GREATER_EQUAL = `return float(a >= b);`;
+var GREATER_EQUAL_PACKED = `
   return vec4(greaterThanEqual(a, b));
-`,nnt=le({opSnippet:ent,packedOpSnippet:rnt,dtype:"bool",cpuKernelImpl:IL}),t3={kernelName:ss,backendName:"webgl",kernelFunc:nnt};function ont(r){let{inputs:t,backend:e}=r,{input:n}=t;return EC(n,!0,e)}var e3={kernelName:Ip,backendName:"webgl",kernelFunc:ont};var snt="return float(!isnan(x) && !isinf(x));",int=Ct({opSnippet:snt,dtype:"bool"}),r3={kernelName:Ia,backendName:"webgl",kernelFunc:int};var ant="return float(isinf(x));",lnt=Ct({opSnippet:ant,dtype:"bool"}),n3={kernelName:Sa,backendName:"webgl",kernelFunc:lnt};var unt="return float(isnan(x));",cnt=Ct({opSnippet:unt,dtype:"bool"}),o3={kernelName:va,backendName:"webgl",kernelFunc:cnt};var pnt="return float(a < b);",mnt=`
+`;
+var greaterEqual3 = binaryKernelFunc2({
+  opSnippet: GREATER_EQUAL,
+  packedOpSnippet: GREATER_EQUAL_PACKED,
+  dtype: "bool",
+  cpuKernelImpl: greaterEqualImplCPU
+});
+var greaterEqualConfig2 = {
+  kernelName: GreaterEqual,
+  backendName: "webgl",
+  kernelFunc: greaterEqual3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IFFT.js
+function ifft3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  return fftImpl2(input2, true, backend2);
+}
+var ifftConfig2 = {
+  kernelName: IFFT,
+  backendName: "webgl",
+  kernelFunc: ifft3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsFinite.js
+var IS_FINITE = `return float(!isnan(x) && !isinf(x));`;
+var isFinite4 = unaryKernelFunc2({ opSnippet: IS_FINITE, dtype: "bool" });
+var isFiniteConfig2 = {
+  kernelName: IsFinite,
+  backendName: "webgl",
+  kernelFunc: isFinite4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsInf.js
+var IS_INF = `return float(isinf(x));`;
+var isInf3 = unaryKernelFunc2({ opSnippet: IS_INF, dtype: "bool" });
+var isInfConfig2 = {
+  kernelName: IsInf,
+  backendName: "webgl",
+  kernelFunc: isInf3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/IsNaN.js
+var IS_NAN = `return float(isnan(x));`;
+var isNaN4 = unaryKernelFunc2({ opSnippet: IS_NAN, dtype: "bool" });
+var isNaNConfig2 = {
+  kernelName: IsNan,
+  backendName: "webgl",
+  kernelFunc: isNaN4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Less.js
+var LESS = `return float(a < b);`;
+var LESS_PACKED = `
   return vec4(lessThan(a, b));
-`,fnt=le({opSnippet:pnt,packedOpSnippet:mnt,cpuKernelImpl:SL,dtype:"bool"}),s3={kernelName:Na,backendName:"webgl",kernelFunc:fnt};var dnt="return float(a <= b);",hnt=`
+`;
+var less4 = binaryKernelFunc2({
+  opSnippet: LESS,
+  packedOpSnippet: LESS_PACKED,
+  cpuKernelImpl: lessImplCPU,
+  dtype: "bool"
+});
+var lessConfig2 = {
+  kernelName: Less,
+  backendName: "webgl",
+  kernelFunc: less4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LessEqual.js
+var LESS_EQUAL = `return float(a <= b);`;
+var LESS_EQUAL_PACKED = `
   return vec4(lessThanEqual(a, b));
-`,gnt=le({opSnippet:dnt,packedOpSnippet:hnt,cpuKernelImpl:vL,dtype:"bool"}),i3={kernelName:Ta,backendName:"webgl",kernelFunc:gnt};function xnt(r){let{backend:t,attrs:e}=r,{start:n,stop:o,num:s}=e,i=NL(n,o,s);return t.makeTensorInfo([i.length],"float32",i)}var a3={kernelName:vp,backendName:"webgl",kernelFunc:xnt};var ynt=Po+`
+`;
+var lessEqual3 = binaryKernelFunc2({
+  opSnippet: LESS_EQUAL,
+  packedOpSnippet: LESS_EQUAL_PACKED,
+  cpuKernelImpl: lessEqualImplCPU,
+  dtype: "bool"
+});
+var lessEqualConfig2 = {
+  kernelName: LessEqual,
+  backendName: "webgl",
+  kernelFunc: lessEqual3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LinSpace.js
+function linSpace2(args) {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, num } = attrs;
+  const outVals = linSpaceImplCPU(start, stop, num);
+  return backend2.makeTensorInfo([outVals.length], "float32", outVals);
+}
+var linSpaceConfig2 = {
+  kernelName: LinSpace,
+  backendName: "webgl",
+  kernelFunc: linSpace2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Log.js
+var LOG = CHECK_NAN_SNIPPET_UNARY + `
   return x < 0.0 ? 0./0. : log(x);
-`,bnt=`
+`;
+var LOG_PACKED = `
   vec4 result = log(x);
   bvec4 isNaN = isnan(x);
   result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);
@@ -3534,18 +59878,85 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
   result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);
   result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);
   return result;
-`,wnt=Ct({opSnippet:ynt,packedOpSnippet:bnt,cpuKernelImpl:TL}),l3={kernelName:as,backendName:"webgl",kernelFunc:wnt};var Cnt=Po+`
+`;
+var log4 = unaryKernelFunc2({ opSnippet: LOG, packedOpSnippet: LOG_PACKED, cpuKernelImpl: logImplCPU });
+var logConfig2 = {
+  kernelName: Log,
+  backendName: "webgl",
+  kernelFunc: log4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Log1p.js
+var LOG1P = CHECK_NAN_SNIPPET_UNARY + `
   return log(1.0 + x);
-`,Int=Ct({opSnippet:Cnt}),u3={kernelName:ka,backendName:"webgl",kernelFunc:Int};var Snt="return float(a >= 1.0 && b >= 1.0);",vnt=`
+`;
+var log1p3 = unaryKernelFunc2({ opSnippet: LOG1P });
+var log1pConfig2 = {
+  kernelName: Log1p,
+  backendName: "webgl",
+  kernelFunc: log1p3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalAnd.js
+var LOGICAL_AND = `return float(a >= 1.0 && b >= 1.0);`;
+var LOGICAL_AND_PACKED = `
   return vec4(
     vec4(greaterThanEqual(a, vec4(1.0))) *
     vec4(greaterThanEqual(b, vec4(1.0))));
-`,Nnt=le({opSnippet:Snt,packedOpSnippet:vnt,dtype:"bool"}),c3={kernelName:Ea,backendName:"webgl",kernelFunc:Nnt};var Tnt="return float(!(x >= 1.0));",knt=Ct({opSnippet:Tnt}),p3={kernelName:_a,backendName:"webgl",kernelFunc:knt};var Ent="return float(a >= 1.0 || b >= 1.0);",_nt=`
+`;
+var logicalAnd3 = binaryKernelFunc2({
+  opSnippet: LOGICAL_AND,
+  packedOpSnippet: LOGICAL_AND_PACKED,
+  dtype: "bool"
+});
+var logicalAndConfig2 = {
+  kernelName: LogicalAnd,
+  backendName: "webgl",
+  kernelFunc: logicalAnd3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalNot.js
+var LOGICAL_NOT = `return float(!(x >= 1.0));`;
+var logicalNot3 = unaryKernelFunc2({ opSnippet: LOGICAL_NOT });
+var logicalNotConfig2 = {
+  kernelName: LogicalNot,
+  backendName: "webgl",
+  kernelFunc: logicalNot3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LogicalOr.js
+var LOGICAL_OR = `return float(a >= 1.0 || b >= 1.0);`;
+var LOGICAL_OR_PACKED = `
   return min(
     vec4(greaterThanEqual(a, vec4(1.0))) +
     vec4(greaterThanEqual(b, vec4(1.0))),
     vec4(1.0));
-`,Ant=le({opSnippet:Ent,packedOpSnippet:_nt,dtype:"bool"}),m3={kernelName:Aa,backendName:"webgl",kernelFunc:Ant};var OC=class{constructor(t,e,n,o,s){this.variableNames=["x"],this.outputShape=[];let i=e,a=t[3]-1;this.outputShape=t;let u,l=`float(${n}) + float(${o}) * sum`;s===.5?u=`inversesqrt(${l})`:s===1?u=`1.0/(${l})`:u=`exp(log(${l}) * float(-${s}));`,this.userCode=`
+`;
+var logicalOr3 = binaryKernelFunc2({ opSnippet: LOGICAL_OR, packedOpSnippet: LOGICAL_OR_PACKED, dtype: "bool" });
+var logicalOrConfig2 = {
+  kernelName: LogicalOr,
+  backendName: "webgl",
+  kernelFunc: logicalOr3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_gpu.js
+var LRNProgram = class {
+  constructor(xShape, radius, bias, alpha, beta) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    const rad = radius;
+    const maxD = xShape[3] - 1;
+    this.outputShape = xShape;
+    let powOperator;
+    const basis = `float(${bias}) + float(${alpha}) * sum`;
+    if (beta === 0.5) {
+      powOperator = `inversesqrt(${basis})`;
+    } else if (beta === 1) {
+      powOperator = `1.0/(${basis})`;
+    } else {
+      powOperator = `exp(log(${basis}) * float(-${beta}));`;
+    }
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3554,17 +59965,40 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         int d = coords[3];
         float x = getX(b, r, c, d);
         float sum = 0.0;
-        for (int j = -${i}; j <= ${i}; j++) {
+        for (int j = -${rad}; j <= ${rad}; j++) {
           int idx = d + j;
-          if (idx >= 0 && idx <=  ${a}) {
+          if (idx >= 0 && idx <=  ${maxD}) {
             float z = getX(b, r, c, idx);
             sum += z * z;
           }
         }
-        float val = x * ${u};
+        float val = x * ${powOperator};
         setOutput(val);
       }
-    `}};var PC=class{constructor(t,e,n,o,s){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let i=e,a=t[3]-1;this.outputShape=t;let u,l=`float(${n}) + float(${o}) * sum`;s===.5?u=`inversesqrt(${l})`:s===1?u=`1.0/(${l})`:u=`exp(log(${l}) * float(-${s}));`,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_packed_gpu.js
+var LRNPackedProgram = class {
+  constructor(xShape, radius, bias, alpha, beta) {
+    this.variableNames = ["x"];
+    this.outputShape = [];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const rad = radius;
+    const maxD = xShape[3] - 1;
+    this.outputShape = xShape;
+    let powOperator;
+    const basis = `float(${bias}) + float(${alpha}) * sum`;
+    if (beta === 0.5) {
+      powOperator = `inversesqrt(${basis})`;
+    } else if (beta === 1) {
+      powOperator = `1.0/(${basis})`;
+    } else {
+      powOperator = `exp(log(${basis}) * float(-${beta}));`;
+    }
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords.x;
@@ -3588,7 +60022,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
             getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0
         );
 
-        int firstChannel = d - ${i};
+        int firstChannel = d - ${rad};
         vec2 cache = vec2(0.);
         if(firstChannel >= 0){
           vec4 firstChannelFrag = getX(b, r, c, firstChannel);
@@ -3599,10 +60033,10 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
 
         ivec2 depth = ivec2(d, d + 1);
-        for (int j = - ${i}; j <= ${i}; j++) {
+        for (int j = - ${rad}; j <= ${rad}; j++) {
           ivec2 idx = depth + j;
           bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));
-          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${a}));
+          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${maxD}));
 
           bool depthInRange = aboveLowerBound.x && belowUpperBound.x;
           bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;
@@ -3623,10 +60057,39 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
             sum += z * z;
           }
         }
-        vec4 result = xAtOutputCoords * ${u};
+        vec4 result = xAtOutputCoords * ${powOperator};
         setOutput(result);
       }
-    `}};var $nt=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{depthRadius:s,bias:i,alpha:a,beta:u}=n,l=z().getBool("WEBGL_PACK_NORMALIZATION")?new PC(o.shape,s,i,a,u):new OC(o.shape,s,i,a,u);return e.runWebGLProgram(l,[o],o.dtype)},f3={kernelName:Rl,backendName:"webgl",kernelFunc:$nt};var LC=class{constructor(t,e,n,o,s){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=t,this.depth=t[3],this.depthRadius=e,this.bias=n,this.alpha=o,this.beta=s,this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LRN.js
+var lrn = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  const program = env().getBool("WEBGL_PACK_NORMALIZATION") ? new LRNPackedProgram(x.shape, depthRadius, bias, alpha, beta) : new LRNProgram(x.shape, depthRadius, bias, alpha, beta);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+};
+var LRNConfig2 = {
+  kernelName: LRN,
+  backendName: "webgl",
+  kernelFunc: lrn
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/lrn_grad_gpu.js
+var LRNGradProgram = class {
+  constructor(inputShape, depthRadius, bias, alpha, beta) {
+    this.variableNames = ["inputImage", "outputImage", "dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    this.depth = inputShape[3];
+    this.depthRadius = depthRadius;
+    this.bias = bias;
+    this.alpha = alpha;
+    this.beta = beta;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -3635,9 +60098,9 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
 
         float result = 0.0;
         for (int d = 0; d < ${this.depth}; ++d) {
-          int depthBegin = int(max(0.0, float(d - ${e})));
+          int depthBegin = int(max(0.0, float(d - ${depthRadius})));
           int depthEnd = int(min(float(${this.depth}),
-              float(d + ${e} + 1)));
+              float(d + ${depthRadius} + 1)));
 
           const int MIN_DEPTH_BEGIN = 0;
           const int MAX_DEPTH_END = ${this.depth};
@@ -3655,19 +60118,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
             }
           }
 
-          norm = float(${o}) * norm + float(${n});
+          norm = float(${alpha}) * norm + float(${bias});
 
           for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){
             if (k < depthBegin){
               continue;
             }
             else if (k >= depthBegin && k < depthEnd){
-              float dyi = -2.0 * float(${o})
-                * float(${s})
+              float dyi = -2.0 * float(${alpha})
+                * float(${beta})
                 * getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)
                 / norm;
               if (k == d) {
-                dyi += pow(norm, -1.0 * ${s});
+                dyi += pow(norm, -1.0 * ${beta});
               }
               if (k == coords[3]) {
                 dyi *= getDy(b, r, c, d);
@@ -3681,17 +60144,169 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
       }
       setOutput(result);
       }
-    `}};var Dnt=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o,y:s,dy:i}=t,{depthRadius:a,bias:u,alpha:l,beta:c}=n,p=new LC(o.shape,a,u,l,c);return e.runWebGLProgram(p,[o,s,i],o.dtype)},d3={kernelName:Np,backendName:"webgl",kernelFunc:Dnt};function h3(r,t,e,n){let o=y.sizeFromShape(t),i=y.sizeFromShape(r.shape)/o,a=st({inputs:{x:r},attrs:{shape:[i,o]},backend:n}),u=Un(a,r.dtype,"max",n),l=st({inputs:{x:u},attrs:{shape:e},backend:n});return n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(u),l}function Ik(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reductionIndices:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=c!=null,m=e.shouldExecuteOnCPU([o]),f=o;if(p){if(m){let w=e.texData.get(f.dataId).values,C=new Array(a);for(let A=0;A<C.length;A++)C[A]=o.shape[c[A]];let N=Vc(w,o.shape,o.dtype,c,C);f=e.makeTensorInfo(C,o.dtype);let _=e.texData.get(f.dataId);_.values=N}else f=Au(o,c,e);l=v.getInnerMostAxes(l.length,a)}v.assertAxesAreInnerMostDims("max",l,a);let[d,h]=v.computeOutAndReduceShapes(f.shape,l),g=d;i&&(g=v.expandShapeToKeepDim(d,u));let x;if(m){let w=e.texData.get(f.dataId).values,C=kL(w,y.sizeFromShape(h),g,o.dtype);x=e.makeTensorInfo(g,o.dtype);let N=e.texData.get(x.dataId);N.values=C}else x=h3(f,h,g,e);return p&&e.disposeIntermediateTensorInfo(f),x}var g3={kernelName:ls,backendName:"webgl",kernelFunc:Ik};var Rnt=Sd+`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/LRNGrad.js
+var lrnGrad = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, y, dy } = inputs;
+  const { depthRadius, bias, alpha, beta } = attrs;
+  const program = new LRNGradProgram(x.shape, depthRadius, bias, alpha, beta);
+  return backend2.runWebGLProgram(program, [x, y, dy], x.dtype);
+};
+var LRNGradConfig2 = {
+  kernelName: LRNGrad,
+  backendName: "webgl",
+  kernelFunc: lrnGrad
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Max_impl.js
+function maxImpl2(x, reduceShape, outShape, backend2) {
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const reduced = reduce(reshapedInput, x.dtype, "max", backend2);
+  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  return reshapedOutput;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Max.js
+function max4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { reductionIndices, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(reductionIndices, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  const maxInputIsTransposed = permutedAxes != null;
+  const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);
+  let maxInput = x;
+  if (maxInputIsTransposed) {
+    if (shouldExecuteOnCPU) {
+      const xTexData = backend2.texData.get(maxInput.dataId);
+      const values = xTexData.values;
+      const newShape = new Array(xRank);
+      for (let i = 0; i < newShape.length; i++) {
+        newShape[i] = x.shape[permutedAxes[i]];
+      }
+      const maxInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);
+      maxInput = backend2.makeTensorInfo(newShape, x.dtype);
+      const maxInputData = backend2.texData.get(maxInput.dataId);
+      maxInputData.values = maxInputValues;
+    } else {
+      maxInput = transposeImpl2(x, permutedAxes, backend2);
+    }
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, xRank);
+  const [maxOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(maxInput.shape, axes);
+  let outShape = maxOutShape;
+  if (keepDims) {
+    outShape = backend_util_exports.expandShapeToKeepDim(maxOutShape, origAxes);
+  }
+  let out;
+  if (shouldExecuteOnCPU) {
+    const xTexData = backend2.texData.get(maxInput.dataId);
+    const values = xTexData.values;
+    const outValues = maxImplCPU(values, util_exports.sizeFromShape(reduceShape), outShape, x.dtype);
+    out = backend2.makeTensorInfo(outShape, x.dtype);
+    const outData = backend2.texData.get(out.dataId);
+    outData.values = outValues;
+  } else {
+    out = maxImpl2(maxInput, reduceShape, outShape, backend2);
+  }
+  if (maxInputIsTransposed) {
+    backend2.disposeIntermediateTensorInfo(maxInput);
+  }
+  return out;
+}
+var maxConfig2 = {
+  kernelName: Max,
+  backendName: "webgl",
+  kernelFunc: max4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Maximum.js
+var MAXIMUM = CHECK_NAN_SNIPPET2 + `
   return max(a, b);
-`,Fnt=`
+`;
+var MAXIMUM_PACKED = `
   vec4 result = vec4(max(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+Yi+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Ont=le({opSnippet:Rnt,packedOpSnippet:Fnt,cpuKernelImpl:EL}),x3={kernelName:us,backendName:"webgl",kernelFunc:Ont};function Pnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t;Qs(o,"maxPool");let{filterSize:s,strides:i,pad:a,dimRoundingMode:u}=n,l=1;y.assert(v.eitherStridesOrDilationsAreOne(i,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${l}'`);let c=v.computePool2DInfo(o.shape,s,i,l,a,u);if(c.filterWidth===1&&c.filterHeight===1&&y.arraysEqual(c.inShape,c.outShape))return tr({inputs:{x:o},backend:e});let p=new ei(c,"max",!1);return e.runWebGLProgram(p,[o],o.dtype)}var y3={kernelName:cs,backendName:"webgl",kernelFunc:Pnt};function Lnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{filterSize:s,strides:i,pad:a,dataFormat:u,dimRoundingMode:l}=n,c=[1,1,1],p=v.computePool3DInfo(o.shape,s,i,c,a,l,u),m=new $u(p,"max",!1);return e.runWebGLProgram(m,[o],o.dtype)}var b3={kernelName:Fl,backendName:"webgl",kernelFunc:Lnt};var MC=class{constructor(t){this.variableNames=["dy","maxPos"],this.outputShape=t.inShape;let e=t.strideHeight,n=t.strideWidth,o=t.dilationHeight,s=t.effectiveFilterHeight,i=t.effectiveFilterWidth,a=s-1-t.padInfo.top,u=i-1-t.padInfo.left,l=s*i-1;this.userCode=`
-      const ivec2 pads = ivec2(${a}, ${u});
+`;
+var maximum4 = binaryKernelFunc2({
+  opSnippet: MAXIMUM,
+  packedOpSnippet: MAXIMUM_PACKED,
+  cpuKernelImpl: maximumImplCPU
+});
+var maximumConfig2 = {
+  kernelName: Maximum,
+  backendName: "webgl",
+  kernelFunc: maximum4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool.js
+function maxPool3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  assertNotComplex2(x, "maxPool");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = 1;
+  util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  if (convInfo.filterWidth === 1 && convInfo.filterHeight === 1 && util_exports.arraysEqual(convInfo.inShape, convInfo.outShape)) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const maxPoolProgram = new Pool2DProgram(convInfo, "max", false);
+  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);
+}
+var maxPoolConfig2 = {
+  kernelName: MaxPool,
+  backendName: "webgl",
+  kernelFunc: maxPool3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool3D.js
+function maxPool3d2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { filterSize, strides, pad: pad3, dataFormat, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode, dataFormat);
+  const maxPoolProgram = new Pool3DProgram(convInfo, "max", false);
+  return backend2.runWebGLProgram(maxPoolProgram, [x], x.dtype);
+}
+var maxPool3DConfig2 = {
+  kernelName: MaxPool3D,
+  backendName: "webgl",
+  kernelFunc: maxPool3d2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/max_pool_backprop_gpu.js
+var MaxPool2DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "maxPos"];
+    this.outputShape = convInfo.inShape;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationHeight = convInfo.dilationHeight;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const lastIndex = effectiveFilterHeight * effectiveFilterWidth - 1;
+    this.userCode = `
+      const ivec2 pads = ivec2(${padTop}, ${padLeft});
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -3705,30 +60320,30 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
-        for (int wR = 0; wR < ${s};
-          wR += ${o}) {
-          float dyR = float(dyRCorner + wR) / ${e}.0;
+        for (int wR = 0; wR < ${effectiveFilterHeight};
+          wR += ${dilationHeight}) {
+          float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-          if (dyR < 0.0 || dyR >= ${t.outHeight}.0 || fract(dyR) > 0.0) {
+          if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 || fract(dyR) > 0.0) {
             continue;
           }
           int idyR = int(dyR);
 
-          for (int wC = 0; wC < ${i}; wC++) {
-            float dyC = float(dyCCorner + wC) / ${n}.0;
+          for (int wC = 0; wC < ${effectiveFilterWidth}; wC++) {
+            float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-            if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+            if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                 fract(dyC) > 0.0) {
               continue;
             }
             int idyC = int(dyC);
 
             float dyValue = getDy(b, idyR, idyC, d);
-            int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));
+            int maxPosValue = ${lastIndex} - int(getMaxPos(b, idyR, idyC, d));
 
             // Get the current value, check it against the value from the
             // position matrix.
-            int curPosValue = wR * ${i} + wC;
+            int curPosValue = wR * ${effectiveFilterWidth} + wC;
             float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
             dotProd += dyValue * mask;
@@ -3736,8 +60351,28 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}},zC=class{constructor(t){this.variableNames=["dy","maxPos"],this.outputShape=t.inShape;let e=t.strideDepth,n=t.strideHeight,o=t.strideWidth,s=t.dilationDepth,i=t.dilationHeight,a=t.dilationWidth,u=t.effectiveFilterDepth,l=t.effectiveFilterHeight,c=t.effectiveFilterWidth,p=u-1-t.padInfo.front,m=l-1-t.padInfo.top,f=c-1-t.padInfo.left,d=u*l*c-1;this.userCode=`
-      const ivec3 pads = ivec3(${p}, ${m}, ${f});
+    `;
+  }
+};
+var MaxPool3DBackpropProgram = class {
+  constructor(convInfo) {
+    this.variableNames = ["dy", "maxPos"];
+    this.outputShape = convInfo.inShape;
+    const strideDepth = convInfo.strideDepth;
+    const strideHeight = convInfo.strideHeight;
+    const strideWidth = convInfo.strideWidth;
+    const dilationDepth = convInfo.dilationDepth;
+    const dilationHeight = convInfo.dilationHeight;
+    const dilationWidth = convInfo.dilationWidth;
+    const effectiveFilterDepth = convInfo.effectiveFilterDepth;
+    const effectiveFilterHeight = convInfo.effectiveFilterHeight;
+    const effectiveFilterWidth = convInfo.effectiveFilterWidth;
+    const padFront = effectiveFilterDepth - 1 - convInfo.padInfo.front;
+    const padTop = effectiveFilterHeight - 1 - convInfo.padInfo.top;
+    const padLeft = effectiveFilterWidth - 1 - convInfo.padInfo.left;
+    const lastIndex = effectiveFilterDepth * effectiveFilterHeight * effectiveFilterWidth - 1;
+    this.userCode = `
+      const ivec3 pads = ivec3(${padFront}, ${padTop}, ${padLeft});
 
       void main() {
         ivec5 coords = getOutputCoords();
@@ -3754,44 +60389,44 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         // ? = to be determined. : = across all values in that axis.
         float dotProd = 0.0;
 
-        for (int wD = 0; wD < ${u};
-           wD += ${s}) {
-          float dyD = float(dyDCorner + wD) / ${e}.0;
+        for (int wD = 0; wD < ${effectiveFilterDepth};
+           wD += ${dilationDepth}) {
+          float dyD = float(dyDCorner + wD) / ${strideDepth}.0;
 
-          if (dyD < 0.0 || dyD >= ${t.outDepth}.0 || fract(dyD) > 0.0) {
+          if (dyD < 0.0 || dyD >= ${convInfo.outDepth}.0 || fract(dyD) > 0.0) {
             continue;
           }
           int idyD = int(dyD);
 
-          for (int wR = 0; wR < ${l};
-              wR += ${i}) {
-            float dyR = float(dyRCorner + wR) / ${n}.0;
+          for (int wR = 0; wR < ${effectiveFilterHeight};
+              wR += ${dilationHeight}) {
+            float dyR = float(dyRCorner + wR) / ${strideHeight}.0;
 
-            if (dyR < 0.0 || dyR >= ${t.outHeight}.0 ||
+            if (dyR < 0.0 || dyR >= ${convInfo.outHeight}.0 ||
                 fract(dyR) > 0.0) {
               continue;
             }
             int idyR = int(dyR);
 
-            for (int wC = 0; wC < ${c};
-                wC += ${a}) {
-              float dyC = float(dyCCorner + wC) / ${o}.0;
+            for (int wC = 0; wC < ${effectiveFilterWidth};
+                wC += ${dilationWidth}) {
+              float dyC = float(dyCCorner + wC) / ${strideWidth}.0;
 
-              if (dyC < 0.0 || dyC >= ${t.outWidth}.0 ||
+              if (dyC < 0.0 || dyC >= ${convInfo.outWidth}.0 ||
                   fract(dyC) > 0.0) {
                 continue;
               }
               int idyC = int(dyC);
 
               float dyValue = getDy(batch, idyD, idyR, idyC, ch);
-              int maxPosValue = ${d} -
+              int maxPosValue = ${lastIndex} -
                   int(getMaxPos(batch, idyD, idyR, idyC, ch));
 
               // Get the current value, check it against the value from the
               // position matrix.
               int curPosValue =
-                  wD * ${l} * ${c} +
-                  wR * ${c} + wC;
+                  wD * ${effectiveFilterHeight} * ${effectiveFilterWidth} +
+                  wR * ${effectiveFilterWidth} + wC;
               float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
 
               dotProd += dyValue * mask;
@@ -3800,107 +60435,374 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
         setOutput(dotProd);
       }
-    `}};function Mnt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s}=t,i=s,{filterSize:a,strides:u,pad:l,dimRoundingMode:c}=n,p=[1,1,1],m=v.computePool3DInfo(i.shape,a,u,p,l,c),f=new $u(m,"max",!0),d=e.runWebGLProgram(f,[i],i.dtype),h=new zC(m),g=e.runWebGLProgram(h,[o,d],i.dtype);return e.disposeIntermediateTensorInfo(d),g}var w3={kernelName:kp,backendName:"webgl",kernelFunc:Mnt};function znt(r){let{inputs:t,backend:e,attrs:n}=r,{dy:o,input:s,output:i}=t,a=s;Qs([s,i],"maxPoolGrad");let{filterSize:u,strides:l,pad:c,dimRoundingMode:p}=n,m=v.computePool2DInfo(a.shape,u,l,1,c,p),f=!0,d=new ei(m,"max",f),h=e.runWebGLProgram(d,[a],a.dtype),g=new MC(m),x=e.runWebGLProgram(g,[o,h],a.dtype);return e.disposeIntermediateTensorInfo(h),x}var C3={kernelName:Tp,backendName:"webgl",kernelFunc:znt};function I3(r,t,e,n){let o=new ei(e,"max",!1),s=n.runWebGLProgram(o,[r],"float32");o=new ei(e,"max",!0,!0,t);let i=n.runWebGLProgram(o,[r],"float32");return[s,i]}var S3={kernelName:Ep,backendName:"webgl",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{filterSize:o,strides:s,pad:i,includeBatchInIndex:a}=t,u=e;y.assert(n.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${n.shape.length}.`);let l=[1,1];y.assert(v.eitherStridesOrDilationsAreOne(s,l),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${l}'`);let c=v.computePool2DInfo(n.shape,o,s,l,i),[p,m]=I3(n,a,c,u);return[p,m]}};function v3(r,t,e,n){let o=y.sizeFromShape(t),i=y.sizeFromShape(r.shape)/o,a=st({inputs:{x:r},attrs:{shape:[i,o]},backend:n}),u=Un(a,"float32","mean",n),l=st({inputs:{x:u},attrs:{shape:e},backend:n});return n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(u),l}var N3={kernelName:ps,backendName:"webgl",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{x:n}=r,{keepDims:o,axis:s}=t,i=e,a=n.shape.length,u=y.parseAxisParam(s,n.shape),l=u,c=v.getAxesPermutation(l,a),p=c!=null,m=i.shouldExecuteOnCPU([n]),f=[],d=n;if(p){if(m){let C=i.texData.get(d.dataId).values,N=new Array(a);for(let $=0;$<N.length;$++)N[$]=n.shape[c[$]];let _=Vc(C,n.shape,n.dtype,c,N);d=i.makeTensorInfo(N,n.dtype);let A=i.texData.get(d.dataId);A.values=_}else d=Au(n,c,i);f.push(d),l=v.getInnerMostAxes(l.length,a)}v.assertAxesAreInnerMostDims("sum",l,a);let[h,g]=v.computeOutAndReduceShapes(d.shape,l),x=h;o&&(x=v.expandShapeToKeepDim(h,u));let b=v3(d,g,x,i);for(let w of f)i.disposeIntermediateTensorInfo(w);return b}};function Bnt(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=y.parseAxisParam(s,o.shape),l=u,c=v.getAxesPermutation(l,a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),l=v.getInnerMostAxes(l.length,o.shape.length)),v.assertAxesAreInnerMostDims("min",l,a);let[m,f]=v.computeOutAndReduceShapes(p.shape,l),d=y.sizeFromShape(f),h=st({inputs:{x:p},backend:e,attrs:{shape:[-1,d]}}),g=Un(h,h.dtype,"min",e),x;if(i){let b=v.expandShapeToKeepDim(m,u);x=st({inputs:{x:g},backend:e,attrs:{shape:b}})}else x=st({inputs:{x:g},backend:e,attrs:{shape:m}});return e.disposeIntermediateTensorInfo(h),e.disposeIntermediateTensorInfo(g),c!=null&&e.disposeIntermediateTensorInfo(p),x}var T3={kernelName:ms,backendName:"webgl",kernelFunc:Bnt};var Vnt=Sd+`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPool3DGrad.js
+function maxPool3DGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2 } = inputs;
+  const x = input2;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const dilations = [1, 1, 1];
+  const convInfo = backend_util_exports.computePool3DInfo(x.shape, filterSize, strides, dilations, pad3, dimRoundingMode);
+  const maxPool3dPositionsProgram = new Pool3DProgram(convInfo, "max", true);
+  const maxPool3dPositions2 = backend2.runWebGLProgram(maxPool3dPositionsProgram, [x], x.dtype);
+  const maxPoolBackpropProgram = new MaxPool3DBackpropProgram(convInfo);
+  const result = backend2.runWebGLProgram(maxPoolBackpropProgram, [dy, maxPool3dPositions2], x.dtype);
+  backend2.disposeIntermediateTensorInfo(maxPool3dPositions2);
+  return result;
+}
+var maxPool3DGradConfig3 = {
+  kernelName: MaxPool3DGrad,
+  backendName: "webgl",
+  kernelFunc: maxPool3DGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolGrad.js
+function maxPoolGrad3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { dy, input: input2, output } = inputs;
+  const x = input2;
+  assertNotComplex2([input2, output], "maxPoolGrad");
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const getPositions = true;
+  const maxPoolPositionsProgram = new Pool2DProgram(convInfo, "max", getPositions);
+  const maxPoolPositions2 = backend2.runWebGLProgram(maxPoolPositionsProgram, [x], x.dtype);
+  const maxPoolBackPropProgram = new MaxPool2DBackpropProgram(convInfo);
+  const result = backend2.runWebGLProgram(maxPoolBackPropProgram, [dy, maxPoolPositions2], x.dtype);
+  backend2.disposeIntermediateTensorInfo(maxPoolPositions2);
+  return result;
+}
+var maxPoolGradConfig3 = {
+  kernelName: MaxPoolGrad,
+  backendName: "webgl",
+  kernelFunc: maxPoolGrad3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolWithArgmax_impl.js
+function maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, backend2) {
+  let program = new Pool2DProgram(convInfo, "max", false);
+  const poolOutput = backend2.runWebGLProgram(program, [x], "float32");
+  program = new Pool2DProgram(convInfo, "max", true, true, includeBatchInIndex);
+  const indexOutput = backend2.runWebGLProgram(program, [x], "float32");
+  return [poolOutput, indexOutput];
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MaxPoolWithArgmax.js
+var maxPoolWithArgmaxConfig2 = {
+  kernelName: MaxPoolWithArgmax,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { filterSize, strides, pad: pad3, includeBatchInIndex } = attrs;
+    const webglBackend = backend2;
+    util_exports.assert(x.shape.length === 4, () => `Error in maxPool: input must be rank 4 but got rank ${x.shape.length}.`);
+    const dilations = [1, 1];
+    util_exports.assert(backend_util_exports.eitherStridesOrDilationsAreOne(strides, dilations), () => `Error in maxPool: Either strides or dilations must be 1. Got strides ${strides} and dilations '${dilations}'`);
+    const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, dilations, pad3);
+    const [result, indexes] = maxPoolWithArgmaxImpl2(x, includeBatchInIndex, convInfo, webglBackend);
+    return [result, indexes];
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mean_impl.js
+function meanImpl(x, reduceShape, outShape, backend2) {
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const batchSize = xSize / inSize;
+  const reshapedInput = reshape4({ inputs: { x }, attrs: { shape: [batchSize, inSize] }, backend: backend2 });
+  const reduced = reduce(reshapedInput, "float32", "mean", backend2);
+  const reshapedOutput = reshape4({ inputs: { x: reduced }, attrs: { shape: outShape }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(reshapedInput);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  return reshapedOutput;
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mean.js
+var meanConfig2 = {
+  kernelName: Mean,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { x } = inputs;
+    const { keepDims, axis } = attrs;
+    const webglBackend = backend2;
+    const xRank = x.shape.length;
+    const origAxes = util_exports.parseAxisParam(axis, x.shape);
+    let axes = origAxes;
+    const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+    const meanInputIsTransposed = permutedAxes != null;
+    const shouldExecuteOnCPU = webglBackend.shouldExecuteOnCPU([x]);
+    const intermediates = [];
+    let meanInput = x;
+    if (meanInputIsTransposed) {
+      if (shouldExecuteOnCPU) {
+        const xTexData = webglBackend.texData.get(meanInput.dataId);
+        const values = xTexData.values;
+        const newShape = new Array(xRank);
+        for (let i = 0; i < newShape.length; i++) {
+          newShape[i] = x.shape[permutedAxes[i]];
+        }
+        const meanInputValues = transposeImplCPU(values, x.shape, x.dtype, permutedAxes, newShape);
+        meanInput = webglBackend.makeTensorInfo(newShape, x.dtype);
+        const meanInputData = webglBackend.texData.get(meanInput.dataId);
+        meanInputData.values = meanInputValues;
+      } else {
+        meanInput = transposeImpl2(x, permutedAxes, webglBackend);
+      }
+      intermediates.push(meanInput);
+      axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    }
+    backend_util_exports.assertAxesAreInnerMostDims("sum", axes, xRank);
+    const [meanOutShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(meanInput.shape, axes);
+    let outShape = meanOutShape;
+    if (keepDims) {
+      outShape = backend_util_exports.expandShapeToKeepDim(meanOutShape, origAxes);
+    }
+    const out = meanImpl(meanInput, reduceShape, outShape, webglBackend);
+    for (const i of intermediates) {
+      webglBackend.disposeIntermediateTensorInfo(i);
+    }
+    return out;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Min.js
+function min4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, x.shape.length);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, xRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+  const inSize = util_exports.sizeFromShape(reduceShape);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  const reduced = reduce(a2D, a2D.dtype, "min", backend2);
+  let res;
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(outShape, origAxes);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: newShape } });
+  } else {
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+  }
+  backend2.disposeIntermediateTensorInfo(a2D);
+  backend2.disposeIntermediateTensorInfo(reduced);
+  if (permutedAxes != null) {
+    backend2.disposeIntermediateTensorInfo(permutedX);
+  }
+  return res;
+}
+var minConfig2 = {
+  kernelName: Min,
+  backendName: "webgl",
+  kernelFunc: min4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Minimum.js
+var MINIMUM = CHECK_NAN_SNIPPET2 + `
   return min(a, b);
-`,Gnt=`
+`;
+var MINIMUM_PACKED = `
   vec4 result = vec4(min(a, b));
   bvec4 isNaNA = isnan(a);
   bvec4 isNaNB = isnan(b);
   bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);
-  `+Yi+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Wnt=le({opSnippet:Vnt,packedOpSnippet:Gnt,cpuKernelImpl:_L}),k3={kernelName:fs,backendName:"webgl",kernelFunc:Wnt};var BC=class{constructor(t,e,n){this.variableNames=["x"],this.outputShape=e.map((c,p)=>c[0]+t[p]+c[1]);let o=t.length,s=zt(o),i=e.map(c=>c[0]).join(","),a=e.map((c,p)=>c[0]+t[p]).join(","),u=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,o),l=n==="reflect"?0:1;if(o===1){this.userCode=`
-        int start = ${i};
-        int end = ${a};
+`;
+var minimum4 = binaryKernelFunc2({
+  opSnippet: MINIMUM,
+  packedOpSnippet: MINIMUM_PACKED,
+  cpuKernelImpl: minimumImplCPU
+});
+var minimumConfig2 = {
+  kernelName: Minimum,
+  backendName: "webgl",
+  kernelFunc: minimum4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/mirror_pad_gpu.js
+var MirrorPadProgram = class {
+  constructor(xShape, paddings, mode) {
+    this.variableNames = ["x"];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const unpackedCoords = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, rank);
+    const offset = mode === "reflect" ? 0 : 1;
+    if (rank === 1) {
+      this.userCode = `
+        int start = ${start};
+        int end = ${end};
 
         void main() {
           int outC = getOutputCoords();
           if (outC < start) {
-            outC = start * 2 - outC - ${l};
+            outC = start * 2 - outC - ${offset};
           } else if(outC >= end) {
-            outC = (end - 1) * 2 - outC + ${l};
+            outC = (end - 1) * 2 - outC + ${offset};
           }
           setOutput(getX(outC - start));
         }
-      `;return}this.userCode=`
-      ${s} start = ${s}(${i});
-      ${s} end = ${s}(${a});
+      `;
+      return;
+    }
+    this.userCode = `
+      ${dtype} start = ${dtype}(${start});
+      ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${s} outC = getOutputCoords();
-        for (int i = 0; i < ${o}; i++) {
+        ${dtype} outC = getOutputCoords();
+        for (int i = 0; i < ${rank}; i++) {
           if (outC[i] < start[i]) {
-            outC[i] = start[i] * 2 - outC[i] - ${l};
+            outC[i] = start[i] * 2 - outC[i] - ${offset};
           } else if(outC[i] >= end[i]) {
-            outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};
+            outC[i] = (end[i] - 1) * 2 - outC[i] + ${offset};
           }
         }
-        ${s} coords = outC - start;
-        setOutput(getX(${u}));
+        ${dtype} coords = outC - start;
+        setOutput(getX(${unpackedCoords}));
       }
-    `}};var VC=class{constructor(t,e,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e.map((d,h)=>d[0]+t[h]+d[1]);let o=t.length,s=zt(o),i=e.map(d=>d[0]).join(","),a=e.map((d,h)=>d[0]+t[h]).join(","),u=Qe("rc",o),l=Qe("source",o),c=`${u[o-1]} < ${this.outputShape[o-1]}`,p=o===1?"source":`vec2(${l.slice(-2).join()})`,m=n==="reflect"?0:1,f="";if(o===1){let d=`
-        ${s} source = rc;
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/mirror_pad_packed_gpu.js
+var MirrorPadPackedProgram = class {
+  constructor(xShape, paddings, mode) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const coords2 = getChannels("rc", rank);
+    const source = getChannels("source", rank);
+    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;
+    const innerDims = rank === 1 ? "source" : `vec2(${source.slice(-2).join()})`;
+    const offset = mode === "reflect" ? 0 : 1;
+    let mainLoop = "";
+    if (rank === 1) {
+      const padSetup = `
+        ${dtype} source = rc;
         if (source < start) {
-          source = start * 2 - source - ${m};
+          source = start * 2 - source - ${offset};
         } else if (source >= end) {
-          source = (end - 1) * 2 - source + ${m};
+          source = (end - 1) * 2 - source + ${offset};
         }
         source -= start;
-      `;f=`
-        ${s} rc = outputLoc;
-        ${d}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${u[o-1]} += 1;
-        if(${c}) {
-          ${d}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+      `;
+      mainLoop = `
+        ${dtype} rc = outputLoc;
+        ${padSetup}
+        result[0] = getChannel(getX(${source.join()}), ${innerDims});
+        ${coords2[rank - 1]} += 1;
+        if(${cLimit}) {
+          ${padSetup}
+          result[1] = getChannel(getX(${source.join()}), ${innerDims});
         }
-      `}else{let d=`
-        ${s} source = rc;
-        ${s} lt = ${s}(lessThan(source, start));
-        ${s} gte = ${s}(greaterThanEqual(source, end));
-        ${s} orig = 1 - (lt + gte);
+      `;
+    } else {
+      const padSetup = `
+        ${dtype} source = rc;
+        ${dtype} lt = ${dtype}(lessThan(source, start));
+        ${dtype} gte = ${dtype}(greaterThanEqual(source, end));
+        ${dtype} orig = 1 - (lt + gte);
         source = orig * source +
-                lt * (start * 2 - source - ${m}) +
-                gte * ((end - 1) * 2 - source + ${m});
+                lt * (start * 2 - source - ${offset}) +
+                gte * ((end - 1) * 2 - source + ${offset});
         source -= start;
-      `;f=`
-        ${s} rc = outputLoc;
-        ${d}
-        result[0] = getChannel(getX(${l.join()}), ${p});
-        ${u[o-1]} += 1;
-        if(${c}) {
-          ${d}
-          result[1] = getChannel(getX(${l.join()}), ${p});
+      `;
+      mainLoop = `
+        ${dtype} rc = outputLoc;
+        ${padSetup}
+        result[0] = getChannel(getX(${source.join()}), ${innerDims});
+        ${coords2[rank - 1]} += 1;
+        if(${cLimit}) {
+          ${padSetup}
+          result[1] = getChannel(getX(${source.join()}), ${innerDims});
         }
         rc = outputLoc;
-        ${u[o-2]} += 1;
-        if(${u[o-2]} < ${this.outputShape[o-2]}) {
-          ${d}
-          result[2] = getChannel(getX(${l.join()}), ${p});
-          ${u[o-1]} += 1;
-          if(${c}) {
-            ${d}
-            result[3] = getChannel(getX(${l.join()}), ${p});
+        ${coords2[rank - 2]} += 1;
+        if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {
+          ${padSetup}
+          result[2] = getChannel(getX(${source.join()}), ${innerDims});
+          ${coords2[rank - 1]} += 1;
+          if(${cLimit}) {
+            ${padSetup}
+            result[3] = getChannel(getX(${source.join()}), ${innerDims});
           }
         }
-      `}this.userCode=`
-      const ${s} start = ${s}(${i});
-      const ${s} end = ${s}(${a});
+      `;
+    }
+    this.userCode = `
+      const ${dtype} start = ${dtype}(${start});
+      const ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${s} outputLoc = getOutputCoords();
+        ${dtype} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
-        ${f}
+        ${mainLoop}
         setOutput(result);
       }
-    `}};var Unt=({inputs:r,backend:t,attrs:e})=>{let{x:n}=r,{paddings:o,mode:s}=e,i=z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new VC(n.shape,o,s):new BC(n.shape,o,s);return t.runWebGLProgram(i,[n],n.dtype)},E3={kernelName:ds,backendName:"webgl",kernelFunc:Unt};var Hnt=`if (b == 0.0) return NAN;
-  return mod(a, b);`,qnt=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/MirrorPad.js
+var mirrorPadKernelFunc = ({ inputs, backend: backend2, attrs }) => {
+  const { x } = inputs;
+  const { paddings, mode } = attrs;
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new MirrorPadPackedProgram(x.shape, paddings, mode) : new MirrorPadProgram(x.shape, paddings, mode);
+  const output = backend2.runWebGLProgram(program, [x], x.dtype);
+  return output;
+};
+var mirrorPadConfig2 = {
+  kernelName: MirrorPad,
+  backendName: "webgl",
+  kernelFunc: mirrorPadKernelFunc
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Mod.js
+var MOD = `if (b == 0.0) return NAN;
+  return mod(a, b);`;
+var MOD_PACKED = `
   vec4 result = mod(a, b);
   bvec4 isNaN = equal(b, vec4(0.0));
-  `+Yi+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,Knt=le({opSnippet:Hnt,packedOpSnippet:qnt}),_3={kernelName:$a,backendName:"webgl",kernelFunc:Knt};var GC=class{constructor(t,e,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[t,n],this.userCode=`
+`;
+var mod3 = binaryKernelFunc2({
+  opSnippet: MOD,
+  packedOpSnippet: MOD_PACKED
+});
+var modConfig2 = {
+  kernelName: Mod,
+  backendName: "webgl",
+  kernelFunc: mod3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/multinomial_gpu.js
+var MultinomialProgram = class {
+  constructor(batchSize, numOutcomes, numSamples) {
+    this.variableNames = ["probs"];
+    this.customUniforms = [{ name: "seed", type: "float" }];
+    this.outputShape = [batchSize, numSamples];
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int batch = coords[0];
@@ -3908,7 +60810,7 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         float r = random(seed);
         float cdf = 0.0;
 
-        for (int i = 0; i < ${e-1}; i++) {
+        for (int i = 0; i < ${numOutcomes - 1}; i++) {
           cdf += getProbs(batch, i);
 
           if (r < cdf) {
@@ -3918,13 +60820,19 @@ return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Wet=Ct({opSnippet:Get}),HM={kernelN
         }
 
         // If no other event happened, last event happened.
-        setOutput(float(${e-1}));
+        setOutput(float(${numOutcomes - 1}));
       }
-    `}};var jnt=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RealDiv.js
+var DIV = `
 if (a == b) {
   return 1.0;
 };
-return a / b;`,Xnt=`
+return a / b;`;
+var DIV_PACKED = `
   // vec4 one = vec4(equal(a, b));
   // return one + (vec4(1.0) - one) * a / b;
   vec4 result = a / b;
@@ -3942,9 +60850,87 @@ return a / b;`,Xnt=`
   }
 
   return result;
-`,Sk=le({opSnippet:jnt,packedOpSnippet:Xnt,checkOutOfBounds:!0}),A3={kernelName:Qo,backendName:"webgl",kernelFunc:Sk};var $3="return a - b;",vk=le({opSnippet:$3,packedOpSnippet:$3,supportsComplex:!0,cpuKernelImpl:XL}),D3={kernelName:Fs,backendName:"webgl",kernelFunc:vk};function Nk(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{dim:s}=n,i=y.parseAxisParam([s],o.shape),a=Ik({inputs:{x:o},backend:e,attrs:{reductionIndices:i,keepDims:!1}}),u=v.expandShapeToKeepDim(a.shape,i),l=st({inputs:{x:a},backend:e,attrs:{shape:u}}),c=vk({inputs:{a:o,b:l},backend:e}),p=bk({inputs:{x:c},backend:e}),m=Wc({inputs:{x:p},backend:e,attrs:{axis:i,keepDims:!1}}),f=st({inputs:{x:m},backend:e,attrs:{shape:u}}),d=Sk({inputs:{a:p,b:f},backend:e});return e.disposeIntermediateTensorInfo(a),e.disposeIntermediateTensorInfo(l),e.disposeIntermediateTensorInfo(c),e.disposeIntermediateTensorInfo(p),e.disposeIntermediateTensorInfo(m),e.disposeIntermediateTensorInfo(f),d}var R3={kernelName:Ds,backendName:"webgl",kernelFunc:Nk};function Ynt(r){let{inputs:t,backend:e,attrs:n}=r,{logits:o}=t,{numSamples:s,seed:i,normalized:a}=n,u=a?o:Nk({inputs:{logits:o},backend:e,attrs:{dim:o.shape.length-1}}),l=u.shape[0],c=u.shape[1],p=new GC(l,c,s),m=[[i]],f=e.runWebGLProgram(p,[u],"int32",m);return a||e.disposeIntermediateTensorInfo(u),f}var F3={kernelName:_p,backendName:"webgl",kernelFunc:Ynt};var Znt=fr+`
+`;
+var realDiv = binaryKernelFunc2({ opSnippet: DIV, packedOpSnippet: DIV_PACKED, checkOutOfBounds: true });
+var realDivConfig2 = {
+  kernelName: RealDiv,
+  backendName: "webgl",
+  kernelFunc: realDiv
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sub.js
+var SUB = "return a - b;";
+var sub3 = binaryKernelFunc2({
+  opSnippet: SUB,
+  packedOpSnippet: SUB,
+  supportsComplex: true,
+  cpuKernelImpl: subImplCPU
+});
+var subConfig2 = {
+  kernelName: Sub,
+  backendName: "webgl",
+  kernelFunc: sub3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Softmax.js
+function softmax4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { dim } = attrs;
+  const axes = util_exports.parseAxisParam([dim], logits.shape);
+  const maxLogit = max4({
+    inputs: { x: logits },
+    backend: backend2,
+    attrs: { reductionIndices: axes, keepDims: false }
+  });
+  const expandedShape = backend_util_exports.expandShapeToKeepDim(maxLogit.shape, axes);
+  const maxLogitsReshaped = reshape4({ inputs: { x: maxLogit }, backend: backend2, attrs: { shape: expandedShape } });
+  const a = sub3({ inputs: { a: logits, b: maxLogitsReshaped }, backend: backend2 });
+  const b = exp3({ inputs: { x: a }, backend: backend2 });
+  const sumExp = sum4({ inputs: { x: b }, backend: backend2, attrs: { axis: axes, keepDims: false } });
+  const sumExpReshaped = reshape4({ inputs: { x: sumExp }, backend: backend2, attrs: { shape: expandedShape } });
+  const res = realDiv({ inputs: { a: b, b: sumExpReshaped }, backend: backend2 });
+  backend2.disposeIntermediateTensorInfo(maxLogit);
+  backend2.disposeIntermediateTensorInfo(maxLogitsReshaped);
+  backend2.disposeIntermediateTensorInfo(a);
+  backend2.disposeIntermediateTensorInfo(b);
+  backend2.disposeIntermediateTensorInfo(sumExp);
+  backend2.disposeIntermediateTensorInfo(sumExpReshaped);
+  return res;
+}
+var softmaxConfig2 = {
+  kernelName: Softmax,
+  backendName: "webgl",
+  kernelFunc: softmax4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Multinomial.js
+function multinomial3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { logits } = inputs;
+  const { numSamples, seed, normalized } = attrs;
+  const probs = normalized ? logits : softmax4({ inputs: { logits }, backend: backend2, attrs: { dim: logits.shape.length - 1 } });
+  const batchSize = probs.shape[0];
+  const numOutcomes = probs.shape[1];
+  const program = new MultinomialProgram(batchSize, numOutcomes, numSamples);
+  const customValues = [[seed]];
+  const res = backend2.runWebGLProgram(program, [probs], "int32", customValues);
+  if (!normalized) {
+    backend2.disposeIntermediateTensorInfo(probs);
+  }
+  return res;
+}
+var multinomialConfig2 = {
+  kernelName: Multinomial,
+  backendName: "webgl",
+  kernelFunc: multinomial3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Neg.js
+var NEG = CHECK_NAN_SNIPPET + `
   return -x;
-`,Jnt=`
+`;
+var NEG_PACKED = `
   vec4 result = -x;
   bvec4 isNaN = isnan(x);
 
@@ -3954,16 +60940,233 @@ return a / b;`,Xnt=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`;function Qnt(r){let{inputs:t,backend:e}=r,{x:n}=t;if(e.shouldExecuteOnCPU([n])){let s=e.texData.get(n.dataId),[i,a]=$L(s.values,n.shape,n.dtype);return e.makeTensorInfo(a,n.dtype,i)}let o;return z().getBool("WEBGL_PACK_UNARY_OPERATIONS")?o=new so(n.shape,Jnt):o=new tn(n.shape,Znt),e.runWebGLProgram(o,[n],n.dtype)}var O3={kernelName:pi,backendName:"webgl",kernelFunc:Qnt};var tot=Ur.nonMaxSuppressionV3Impl;function eot(r){v.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u}=n,l=e.readSync(o.dataId),c=e.readSync(s.dataId),{selectedIndices:p}=tot(l,c,i,a,u);return e.makeTensorInfo([p.length],"int32",new Int32Array(p))}var P3={kernelName:Ra,backendName:"webgl",kernelFunc:eot};var rot=Ur.nonMaxSuppressionV4Impl;function not(r){v.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,padToMaxOutputSize:l}=n,c=e.readSync(o.dataId),p=e.readSync(s.dataId),{selectedIndices:m,validOutputs:f}=rot(c,p,i,a,u,l);return[e.makeTensorInfo([m.length],"int32",new Int32Array(m)),e.makeTensorInfo([],"int32",new Int32Array([f]))]}var L3={kernelName:Fa,backendName:"webgl",kernelFunc:not};var oot=Ur.nonMaxSuppressionV5Impl;function sot(r){v.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:e,attrs:n}=r,{boxes:o,scores:s}=t,{maxOutputSize:i,iouThreshold:a,scoreThreshold:u,softNmsSigma:l}=n,c=e.readSync(o.dataId),p=e.readSync(s.dataId),m=i,f=a,d=u,h=l,{selectedIndices:g,selectedScores:x}=oot(c,p,m,f,d,h);return[e.makeTensorInfo([g.length],"int32",new Int32Array(g)),e.makeTensorInfo([x.length],"float32",new Float32Array(x))]}var M3={kernelName:Oa,backendName:"webgl",kernelFunc:sot};var WC=class{constructor(t,e,n,o){this.variableNames=["indices"],this.outputShape=[t,e],this.userCode=`
+`;
+function neg3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (backend2.shouldExecuteOnCPU([x])) {
+    const xData = backend2.texData.get(x.dataId);
+    const [outValues, newShape] = negImplCPU(xData.values, x.shape, x.dtype);
+    return backend2.makeTensorInfo(newShape, x.dtype, outValues);
+  }
+  let program;
+  if (env().getBool("WEBGL_PACK_UNARY_OPERATIONS")) {
+    program = new UnaryOpPackedProgram(x.shape, NEG_PACKED);
+  } else {
+    program = new UnaryOpProgram(x.shape, NEG);
+  }
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var negConfig2 = {
+  kernelName: Neg,
+  backendName: "webgl",
+  kernelFunc: neg3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV3.js
+var nonMaxSuppressionV3Impl3 = kernel_impls_exports.nonMaxSuppressionV3Impl;
+function nonMaxSuppressionV32(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const { selectedIndices } = nonMaxSuppressionV3Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold);
+  return backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices));
+}
+var nonMaxSuppressionV3Config2 = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV32
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV4.js
+var nonMaxSuppressionV4Impl3 = kernel_impls_exports.nonMaxSuppressionV4Impl;
+function nonMaxSuppressionV42(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const { selectedIndices, validOutputs } = nonMaxSuppressionV4Impl3(boxesVals, scoresVals, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([], "int32", new Int32Array([validOutputs]))
+  ];
+}
+var nonMaxSuppressionV4Config2 = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV42
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/NonMaxSuppressionV5.js
+var nonMaxSuppressionV5Impl3 = kernel_impls_exports.nonMaxSuppressionV5Impl;
+function nonMaxSuppressionV52(args) {
+  backend_util_exports.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");
+  const { inputs, backend: backend2, attrs } = args;
+  const { boxes, scores } = inputs;
+  const { maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma } = attrs;
+  const boxesVals = backend2.readSync(boxes.dataId);
+  const scoresVals = backend2.readSync(scores.dataId);
+  const maxOutputSizeVal = maxOutputSize;
+  const iouThresholdVal = iouThreshold;
+  const scoreThresholdVal = scoreThreshold;
+  const softNmsSigmaVal = softNmsSigma;
+  const { selectedIndices, selectedScores } = nonMaxSuppressionV5Impl3(boxesVals, scoresVals, maxOutputSizeVal, iouThresholdVal, scoreThresholdVal, softNmsSigmaVal);
+  return [
+    backend2.makeTensorInfo([selectedIndices.length], "int32", new Int32Array(selectedIndices)),
+    backend2.makeTensorInfo([selectedScores.length], "float32", new Float32Array(selectedScores))
+  ];
+}
+var nonMaxSuppressionV5Config2 = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "webgl",
+  kernelFunc: nonMaxSuppressionV52
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/onehot_gpu.js
+var OneHotProgram = class {
+  constructor(numIndices, depth, onValue, offValue) {
+    this.variableNames = ["indices"];
+    this.outputShape = [numIndices, depth];
+    this.userCode = `
       void main() {
         ivec2 coords = getOutputCoords();
         int index = round(getIndices(coords.x));
-        setOutput(mix(float(${o}), float(${n}),
+        setOutput(mix(float(${offValue}), float(${onValue}),
                       float(index == coords.y)));
       }
-    `}};var iot=r=>{let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n,l=y.sizeFromShape(o.shape),c=new WC(l,i,a,u),p=st({inputs:{x:o},backend:e,attrs:{shape:[l]}}),m=e.runWebGLProgram(c,[p],s);e.disposeIntermediateTensorInfo(p);let f=[...o.shape,i],d=st({inputs:{x:m},backend:e,attrs:{shape:f}});return e.disposeIntermediateTensorInfo(m),d},z3={kernelName:gs,backendName:"webgl",kernelFunc:iot};function sg(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype==="complex64"){let o=wl({inputs:{input:n},backend:e}),s=sg({inputs:{x:o},backend:e}),i=Hc({inputs:{input:n},backend:e}),a=sg({inputs:{x:i},backend:e}),u=En({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Cl({attrs:{shape:n.shape,dtype:n.dtype,value:n.dtype==="string"?"":0},backend:e})}var B3={kernelName:wi,backendName:"webgl",kernelFunc:sg};function V3(r){let{inputs:t,backend:e}=r,{x:n}=t;if(n.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(n.dtype==="complex64"){let o=wl({inputs:{input:n},backend:e}),s=V3({inputs:{x:o},backend:e}),i=Hc({inputs:{input:n},backend:e}),a=sg({inputs:{x:i},backend:e}),u=En({inputs:{real:s,imag:a},backend:e});return e.disposeIntermediateTensorInfo(o),e.disposeIntermediateTensorInfo(s),e.disposeIntermediateTensorInfo(i),e.disposeIntermediateTensorInfo(a),u}else return Cl({attrs:{shape:n.shape,dtype:n.dtype,value:1},backend:e})}var G3={kernelName:mi,backendName:"webgl",kernelFunc:V3};function aot(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return kC({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,"All tensors passed to stack must have matching shapes"),y.assert(i===c.dtype,()=>"All tensors passed to stack must have matching dtypes")});let a=[],u=t.map(c=>{let p=kC({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=yk({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeIntermediateTensorInfo(c)),l}var W3={kernelName:fi,backendName:"webgl",kernelFunc:aot};var UC=class{constructor(t,e,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=e.map((l,c)=>l[0]+t[c]+l[1]);let o=t.length,s=zt(o),i=e.map(l=>l[0]).join(","),a=e.map((l,c)=>l[0]+t[c]).join(","),u=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,o);if(o===1){this.userCode=`
-        int start = ${i};
-        int end = ${a};
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/OneHot.js
+var oneHot3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const program = new OneHotProgram(indicesSize, depth, onValue, offValue);
+  const reshaped = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [indicesSize] } });
+  const result = backend2.runWebGLProgram(program, [reshaped], dtype);
+  backend2.disposeIntermediateTensorInfo(reshaped);
+  const outShape = [...indices.shape, depth];
+  const out = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: outShape } });
+  backend2.disposeIntermediateTensorInfo(result);
+  return out;
+};
+var oneHotConfig2 = {
+  kernelName: OneHot,
+  backendName: "webgl",
+  kernelFunc: oneHot3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ZerosLike.js
+function zerosLike3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const r = zerosLike3({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill3({
+      attrs: {
+        shape: x.shape,
+        dtype: x.dtype,
+        value: x.dtype === "string" ? "" : 0
+      },
+      backend: backend2
+    });
+  }
+}
+var zerosLikeConfig2 = {
+  kernelName: ZerosLike,
+  backendName: "webgl",
+  kernelFunc: zerosLike3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/OnesLike.js
+function onesLike3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x } = inputs;
+  if (x.dtype === "string") {
+    throw new Error("onesLike is not supported under string dtype");
+  } else if (x.dtype === "complex64") {
+    const realPart = real3({ inputs: { input: x }, backend: backend2 });
+    const r = onesLike3({ inputs: { x: realPart }, backend: backend2 });
+    const imagPart = imag3({ inputs: { input: x }, backend: backend2 });
+    const i = zerosLike3({ inputs: { x: imagPart }, backend: backend2 });
+    const result = complex3({ inputs: { real: r, imag: i }, backend: backend2 });
+    backend2.disposeIntermediateTensorInfo(realPart);
+    backend2.disposeIntermediateTensorInfo(r);
+    backend2.disposeIntermediateTensorInfo(imagPart);
+    backend2.disposeIntermediateTensorInfo(i);
+    return result;
+  } else {
+    return fill3({ attrs: { shape: x.shape, dtype: x.dtype, value: 1 }, backend: backend2 });
+  }
+}
+var onesLikeConfig2 = {
+  kernelName: OnesLike,
+  backendName: "webgl",
+  kernelFunc: onesLike3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Pack.js
+function pack2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims4({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims4({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat3({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var packConfig2 = {
+  kernelName: Pack,
+  backendName: "webgl",
+  kernelFunc: pack2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/pad_gpu.js
+var PadProgram = class {
+  constructor(xShape, paddings, constantValue) {
+    this.variableNames = ["x"];
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const type = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const unpackedCoords = ["coords[0]", "coords[1]", "coords[2]", "coords[3]"].slice(0, rank);
+    if (rank === 1) {
+      this.userCode = `
+        int start = ${start};
+        int end = ${end};
 
         void main() {
           int outC = getOutputCoords();
@@ -3973,44 +61176,106 @@ return a / b;`,Xnt=`
             setOutput(getX(outC - start));
           }
         }
-      `;return}this.userCode=`
-      ${s} start = ${s}(${i});
-      ${s} end = ${s}(${a});
+      `;
+      return;
+    }
+    this.userCode = `
+      ${type} start = ${type}(${start});
+      ${type} end = ${type}(${end});
 
       void main() {
-        ${s} outC = getOutputCoords();
+        ${type} outC = getOutputCoords();
         if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {
           setOutput(value);
         } else {
-          ${s} coords = outC - start;
-          setOutput(getX(${u}));
+          ${type} coords = outC - start;
+          setOutput(getX(${unpackedCoords}));
         }
       }
-    `}};var HC=class{constructor(t,e,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=e.map((h,g)=>h[0]+t[g]+h[1]);let o=t.length,s=zt(o),i=e.map(h=>h[0]).join(","),a=e.map((h,g)=>h[0]+t[g]).join(","),u=Qe("rc",o),l=Qe("source",o),c=`${u[o-1]} < ${this.outputShape[o-1]}`,p=o===1?"source":`vec2(${l.slice(-2).join()})`,m=[`${s} rc = outputLoc;`,`${u[o-1]} += 1;
-       if(${c}) {
-      `,o===1?"":`}
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/pad_packed_gpu.js
+var PadPackedProgram = class {
+  constructor(xShape, paddings, constantValue) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.customUniforms = [{ name: "value", type: "float" }];
+    this.outputShape = paddings.map((p2, i) => p2[0] + xShape[i] + p2[1]);
+    const rank = xShape.length;
+    const dtype = getCoordsDataType(rank);
+    const start = paddings.map((p2) => p2[0]).join(",");
+    const end = paddings.map((p2, i) => p2[0] + xShape[i]).join(",");
+    const coords2 = getChannels("rc", rank);
+    const source = getChannels("source", rank);
+    const cLimit = `${coords2[rank - 1]} < ${this.outputShape[rank - 1]}`;
+    const innerDims = rank === 1 ? "source" : `vec2(${source.slice(-2).join()})`;
+    const componentSetup = [
+      `${dtype} rc = outputLoc;`,
+      `${coords2[rank - 1]} += 1;
+       if(${cLimit}) {
+      `,
+      rank === 1 ? "" : `}
        rc = outputLoc;
-       ${u[o-2]} += 1;
-       if(${u[o-2]} < ${this.outputShape[o-2]}) {`,o===1?"":`  ${u[o-1]} += 1;
-         if(${c}) {`],f=o===1?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",d="";for(let h=0,g=o===1?2:4;h<g;h++)d+=`
-        ${m[h]}
-        if (${f}) {
-          result[${h}] = float(value);
+       ${coords2[rank - 2]} += 1;
+       if(${coords2[rank - 2]} < ${this.outputShape[rank - 2]}) {`,
+      rank === 1 ? "" : `  ${coords2[rank - 1]} += 1;
+         if(${cLimit}) {`
+    ];
+    const paddingArea = rank === 1 ? "rc < start || rc >= end" : "any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))";
+    let mainLoop = "";
+    for (let i = 0, j = rank === 1 ? 2 : 4; i < j; i++) {
+      mainLoop += `
+        ${componentSetup[i]}
+        if (${paddingArea}) {
+          result[${i}] = float(value);
         } else {
-          ${s} source = rc - start;
-          result[${h}] = getChannel(getX(${l.join()}), ${p});
+          ${dtype} source = rc - start;
+          result[${i}] = getChannel(getX(${source.join()}), ${innerDims});
         }
-      `;d+=o===1?"} ":"}}",this.userCode=`
-      const ${s} start = ${s}(${i});
-      const ${s} end = ${s}(${a});
+      `;
+    }
+    mainLoop += rank === 1 ? `} ` : `}}`;
+    this.userCode = `
+      const ${dtype} start = ${dtype}(${start});
+      const ${dtype} end = ${dtype}(${end});
 
       void main() {
-        ${s} outputLoc = getOutputCoords();
+        ${dtype} outputLoc = getOutputCoords();
         vec4 result = vec4(0.);
-        ${d}
+        ${mainLoop}
         setOutput(result);
       }
-    `}};var Tk=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{paddings:s,constantValue:i}=n;if(y.sizeFromShape(o.shape)===0){let l=s.map((c,p)=>c[0]+o.shape[p]+c[1]);return Cl({backend:e,attrs:{shape:l,value:i,dtype:o.dtype}})}let a=z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new HC(o.shape,s,i):new UC(o.shape,s,i),u=[[i]];return e.runWebGLProgram(a,[o],o.dtype,u)},U3={kernelName:xs,backendName:"webgl",kernelFunc:Tk};var lot=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/PadV2.js
+var padV22 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { paddings, constantValue } = attrs;
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    const outputShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+    return fill3({
+      backend: backend2,
+      attrs: { shape: outputShape, value: constantValue, dtype: x.dtype }
+    });
+  }
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new PadPackedProgram(x.shape, paddings, constantValue) : new PadProgram(x.shape, paddings, constantValue);
+  const customValues = [[constantValue]];
+  return backend2.runWebGLProgram(program, [x], x.dtype, customValues);
+};
+var padV2Config2 = {
+  kernelName: PadV2,
+  backendName: "webgl",
+  kernelFunc: padV22
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Pow.js
+var POW = `
   if(a < 0.0 && floor(b) < b){
     return NAN;
   }
@@ -4019,7 +61284,8 @@ return a / b;`,Xnt=`
   }
   return (round(mod(b, 2.0)) != 1) ?
       pow(abs(a), b) : sign(a) * pow(abs(a), b);
-`,uot=`
+`;
+var POW_PACKED = `
   // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.
   vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));
   vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);
@@ -4035,11 +61301,146 @@ return a / b;`,Xnt=`
   bvec4 isNaN1 = lessThan(a, vec4(0.0));
   bvec4 isNaN2 = lessThan(floor(b), b);
   bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);
-  `+Yi+`
+  ` + CHECK_NAN_SNIPPET_PACKED + `
   return result;
-`,cot=le({opSnippet:lot,packedOpSnippet:uot}),H3={kernelName:ys,backendName:"webgl",kernelFunc:cot};function pot(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,keepDims:i}=n,a=o.shape.length,u=[],l=y.parseAxisParam(s,o.shape),c=l,p=v.getAxesPermutation(c,a),m=o;p!=null&&(m=Oe({inputs:{x:o},backend:e,attrs:{perm:p}}),c=v.getInnerMostAxes(c.length,a),u.push(m)),v.assertAxesAreInnerMostDims("prod",c,a);let f;if(e.shouldExecuteOnCPU([m])){let d=e.texData.get(m.dataId).values,{outVals:h,outShape:g,outDtype:x}=RL(m.shape,m.dtype,d,c);f=e.makeTensorInfo(g,x,h)}else{let[d,h]=v.computeOutAndReduceShapes(m.shape,c),g=y.sizeFromShape(h),x=st({inputs:{x:m},backend:e,attrs:{shape:[-1,g]}}),b=Wu(o.dtype),w=Un(x,b,"prod",e);f=st({inputs:{x:w},backend:e,attrs:{shape:d}}),u.push(x),u.push(w)}if(i){u.push(f);let d=v.expandShapeToKeepDim(f.shape,l);f=st({inputs:{x:f},backend:e,attrs:{shape:d}})}return u.forEach(d=>e.disposeIntermediateTensorInfo(d)),f}var q3={kernelName:ws,backendName:"webgl",kernelFunc:pot};function mot(r){let{inputs:t,backend:e,attrs:n}=r,{paramsNestedSplits:o,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:a}=n,u=o.map(x=>e.readSync(x.dataId)),l=o.map(x=>x.shape),c=e.readSync(s.dataId),p=e.readSync(i.dataId),[m,f,d]=FL(u,l,c,s.shape,s.dtype,p,i.shape,a),h=m.map(x=>e.makeTensorInfo([x.length],"int32",x)),g=e.makeTensorInfo(d,s.dtype,f);return h.concat([g])}var K3={kernelName:Ap,backendName:"webgl",kernelFunc:mot};function fot(r){let{inputs:t,backend:e}=r,{starts:n,limits:o,deltas:s}=t,i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=OL(i,n.shape,n.dtype,a,o.shape,u,s.shape),p=e.makeTensorInfo([l.length],"int32",l),m=e.makeTensorInfo([c.length],n.dtype,c);return[p,m]}var j3={kernelName:$p,backendName:"webgl",kernelFunc:fot};function dot(r){let{inputs:t,backend:e,attrs:n}=r,{shape:o,values:s,defaultValue:i,rowPartitionTensors:a}=t,{rowPartitionTypes:u}=n,l=e.readSync(o.dataId),c=e.readSync(s.dataId),p=e.readSync(i.dataId),m=a.map(g=>e.readSync(g.dataId)),f=a.map(g=>g.shape),[d,h]=PL(l,o.shape,c,s.shape,s.dtype,p,i.shape,m,f,u);return e.makeTensorInfo(d,s.dtype,h)}var X3={kernelName:Dp,backendName:"webgl",kernelFunc:dot};var kk=r=>{let{backend:t,attrs:e}=r,{start:n,stop:o,step:s,dtype:i}=e,a=LL(n,o,s,i);return t.makeTensorInfo([a.length],i,a)},Y3={kernelName:Ol,backendName:"webgl",kernelFunc:kk};var hot="return 1.0 / x;",got=Ct({opSnippet:hot}),Z3={kernelName:Pa,backendName:"webgl",kernelFunc:got};var xot=fr+`
+`;
+var pow3 = binaryKernelFunc2({ opSnippet: POW, packedOpSnippet: POW_PACKED });
+var powConfig2 = {
+  kernelName: Pow,
+  backendName: "webgl",
+  kernelFunc: pow3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Prod.js
+function prod3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, keepDims } = attrs;
+  const xRank = x.shape.length;
+  const toDispose = [];
+  const origAxes = util_exports.parseAxisParam(axis, x.shape);
+  let axes = origAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let permutedX = x;
+  if (permutedAxes != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutedAxes } });
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    toDispose.push(permutedX);
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("prod", axes, xRank);
+  let res;
+  if (backend2.shouldExecuteOnCPU([permutedX])) {
+    const xVals = backend2.texData.get(permutedX.dataId).values;
+    const { outVals, outShape, outDtype } = prodImplCPU(permutedX.shape, permutedX.dtype, xVals, axes);
+    res = backend2.makeTensorInfo(outShape, outDtype, outVals);
+  } else {
+    const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(permutedX.shape, axes);
+    const inSize = util_exports.sizeFromShape(reduceShape);
+    const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+    const outputDType = sumOutType(x.dtype);
+    const reduced = reduce(a2D, outputDType, "prod", backend2);
+    res = reshape4({ inputs: { x: reduced }, backend: backend2, attrs: { shape: outShape } });
+    toDispose.push(a2D);
+    toDispose.push(reduced);
+  }
+  if (keepDims) {
+    toDispose.push(res);
+    const newShape = backend_util_exports.expandShapeToKeepDim(res.shape, origAxes);
+    res = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: newShape } });
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return res;
+}
+var prodConfig2 = {
+  kernelName: Prod,
+  backendName: "webgl",
+  kernelFunc: prod3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedGather.js
+function raggedGather3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { paramsNestedSplits, paramsDenseValues, indices } = inputs;
+  const { outputRaggedRank } = attrs;
+  const $paramsNestedSplits = paramsNestedSplits.map((t) => backend2.readSync(t.dataId));
+  const $paramsNestedSplitsShapes = paramsNestedSplits.map((t) => t.shape);
+  const $paramsDenseValues = backend2.readSync(paramsDenseValues.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const [outputNestedSplits, outputDenseValues, outputDenseValuesShape] = raggedGatherImplCPU($paramsNestedSplits, $paramsNestedSplitsShapes, $paramsDenseValues, paramsDenseValues.shape, paramsDenseValues.dtype, $indices, indices.shape, outputRaggedRank);
+  const outputNestedSplitsTensors = outputNestedSplits.map((splits) => backend2.makeTensorInfo([splits.length], "int32", splits));
+  const outputDenseValuesTensor = backend2.makeTensorInfo(outputDenseValuesShape, paramsDenseValues.dtype, outputDenseValues);
+  return outputNestedSplitsTensors.concat([outputDenseValuesTensor]);
+}
+var raggedGatherConfig2 = {
+  kernelName: RaggedGather,
+  backendName: "webgl",
+  kernelFunc: raggedGather3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedRange.js
+function raggedRange3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { starts, limits, deltas } = inputs;
+  const $starts = backend2.readSync(starts.dataId);
+  const $limits = backend2.readSync(limits.dataId);
+  const $deltas = backend2.readSync(deltas.dataId);
+  const [rtNestedSplitsData, rtDenseValuesData] = raggedRangeImplCPU($starts, starts.shape, starts.dtype, $limits, limits.shape, $deltas, deltas.shape);
+  const rtNestedSplits = backend2.makeTensorInfo([rtNestedSplitsData.length], "int32", rtNestedSplitsData);
+  const rtDenseValues = backend2.makeTensorInfo([rtDenseValuesData.length], starts.dtype, rtDenseValuesData);
+  return [rtNestedSplits, rtDenseValues];
+}
+var raggedRangeConfig2 = {
+  kernelName: RaggedRange,
+  backendName: "webgl",
+  kernelFunc: raggedRange3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RaggedTensorToTensor.js
+function raggedTensorToTensor3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { shape, values, defaultValue, rowPartitionTensors } = inputs;
+  const { rowPartitionTypes } = attrs;
+  const $shape = backend2.readSync(shape.dataId);
+  const $values = backend2.readSync(values.dataId);
+  const $defaultValue = backend2.readSync(defaultValue.dataId);
+  const $rowPartitionValues = rowPartitionTensors.map((t) => backend2.readSync(t.dataId));
+  const rowPartitionValuesShapes = rowPartitionTensors.map((t) => t.shape);
+  const [outputShape, output] = raggedTensorToTensorImplCPU($shape, shape.shape, $values, values.shape, values.dtype, $defaultValue, defaultValue.shape, $rowPartitionValues, rowPartitionValuesShapes, rowPartitionTypes);
+  return backend2.makeTensorInfo(outputShape, values.dtype, output);
+}
+var raggedTensorToTensorConfig2 = {
+  kernelName: RaggedTensorToTensor,
+  backendName: "webgl",
+  kernelFunc: raggedTensorToTensor3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Range.js
+var range4 = (args) => {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, step: step5, dtype } = attrs;
+  const values = rangeImplCPU(start, stop, step5, dtype);
+  return backend2.makeTensorInfo([values.length], dtype, values);
+};
+var rangeConfig2 = {
+  kernelName: Range,
+  backendName: "webgl",
+  kernelFunc: range4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reciprocal.js
+var RECIPROCAL = `return 1.0 / x;`;
+var reciprocal3 = unaryKernelFunc2({ opSnippet: RECIPROCAL });
+var reciprocalConfig2 = {
+  kernelName: Reciprocal,
+  backendName: "webgl",
+  kernelFunc: reciprocal3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Relu.js
+var RELU3 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : x;
-`,yot=`
+`;
+var RELU_PACKED = `
   vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4049,9 +61450,19 @@ return a / b;`,Xnt=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,bot=Ct({opSnippet:xot,packedOpSnippet:yot}),J3={kernelName:Cs,backendName:"webgl",kernelFunc:bot};var wot=fr+`
+`;
+var relu3 = unaryKernelFunc2({ opSnippet: RELU3, packedOpSnippet: RELU_PACKED });
+var reluConfig2 = {
+  kernelName: Relu,
+  backendName: "webgl",
+  kernelFunc: relu3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Relu6.js
+var RELU63 = CHECK_NAN_SNIPPET + `
   return (x < 0.0) ? 0.0 : min(6.0, x);
-`,Cot=`
+`;
+var RELU6_PACKED = `
   vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
   bvec4 isNaN = isnan(x);
 
@@ -4061,11 +61472,40 @@ return a / b;`,Xnt=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,Iot=Ct({opSnippet:wot,packedOpSnippet:Cot}),Q3={kernelName:vs,backendName:"webgl",kernelFunc:Iot};var qC=class{constructor(t,e,n,o,s){this.variableNames=["A"],this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m;s?m="(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":m="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+`;
+var relu63 = unaryKernelFunc2({ opSnippet: RELU63, packedOpSnippet: RELU6_PACKED });
+var relu6Config2 = {
+  kernelName: Relu6,
+  backendName: "webgl",
+  kernelFunc: relu63
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_gpu.js
+var ResizeBilinearProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)`;
+    } else {
+      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${c[0]/p[0]},
-          ${c[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${a}.0, ${u}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4074,7 +61514,7 @@ return a / b;`,Xnt=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${m};
+        vec2 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the four integer indices.
         ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));
@@ -4094,13 +61534,40 @@ return a / b;`,Xnt=`
 
         setOutput(newValue);
       }
-    `}};var KC=class{constructor(t,e,n,o,s){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m;s?m="(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":m="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_packed_gpu.js
+var ResizeBilinearPackedProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)`;
+    } else {
+      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${c[0]/p[0]},
-          ${c[1]/p[1]},
-          ${c[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${a}.0, ${u}.0,
-                                     ${u}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,
+                                     ${oldWidth}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4114,7 +61581,7 @@ return a / b;`,Xnt=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${m};
+        vec3 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the four integer indices.
         ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));
@@ -4122,8 +61589,8 @@ return a / b;`,Xnt=`
           min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
-        bool hasNextRow = coords.z < ${n-1};
+        bool hasNextCol = d < ${depth - 1};
+        bool hasNextRow = coords.z < ${newWidth - 1};
 
         // In parallel, construct four corners for all four components in
         // packed 2x2 cell.
@@ -4171,7 +61638,48 @@ return a / b;`,Xnt=`
 
         setOutput(newValue);
       }
-    `}};function Sot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,c=z().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new KC(o.shape,u,l,s,i):new qC(o.shape,u,l,s,i);return e.runWebGLProgram(c,[o],"float32")}var tB={kernelName:Ss,backendName:"webgl",kernelFunc:Sot};var jC=class{constructor(t,e,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=e;let[,o,s]=e,[,i,a]=t,u=[n&&i>1?o-1:o,n&&a>1?s-1:s],l=[n&&i>1?i-1:i,n&&a>1?a-1:a],c=u[0]/l[0],p=u[1]/l[1],m=1/c,f=1/p,d=Math.ceil(m)*2+2,h=Math.ceil(f)*2+2;this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeBilinear.js
+function resizeBilinear3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const program = env().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new ResizeBilinearPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeBilinearProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);
+  return backend2.runWebGLProgram(program, [images], "float32");
+}
+var resizeBilinearConfig2 = {
+  kernelName: ResizeBilinear,
+  backendName: "webgl",
+  kernelFunc: resizeBilinear3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_bilinear_backprop_gpu.js
+var ResizeBilinearBackpropProgram = class {
+  constructor(dyShape, inputShape, alignCorners) {
+    this.variableNames = ["dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    const [, xHeight, xWidth] = inputShape;
+    const [, yHeight, yWidth] = dyShape;
+    const effectiveXSize = [
+      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+    ];
+    const effectiveYSize = [
+      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+    ];
+    const heightScale = effectiveXSize[0] / effectiveYSize[0];
+    const widthScale = effectiveXSize[1] / effectiveYSize[1];
+    const invHeightScale = 1 / heightScale;
+    const invWidthScale = 1 / widthScale;
+    const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+    const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4181,14 +61689,14 @@ return a / b;`,Xnt=`
 
         float accumulator = 0.0;
 
-        const float heightScale = float(${c});
-        const float widthScale = float(${p});
+        const float heightScale = float(${heightScale});
+        const float widthScale = float(${widthScale});
 
-        const float invHeightScale = float(${m});
-        const float invWidthScale = float(${f});
+        const float invHeightScale = float(${invHeightScale});
+        const float invWidthScale = float(${invWidthScale});
 
-        const int winHeight = int(${d});
-        const int winWidth = int(${h});
+        const int winHeight = int(${winHeight});
+        const int winWidth = int(${winWidth});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4202,7 +61710,7 @@ return a / b;`,Xnt=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${i}) {
+          if (dyR < 0 || dyR >= ${yHeight}) {
             continue;
           }
 
@@ -4210,19 +61718,19 @@ return a / b;`,Xnt=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${a}) {
+            if (dyC < 0 || dyC >= ${yWidth}) {
               continue;
             }
 
             float dxR = float(dyR) * heightScale;
             int topDxRIndex = int(floor(dxR));
-            int bottomDxRIndex = int(min(ceil(dxR), ${o-1}.0));
+            int bottomDxRIndex = int(min(ceil(dxR), ${xHeight - 1}.0));
             float dxRLerp = dxR - float(topDxRIndex);
             float inverseDxRLerp = 1.0 - dxRLerp;
 
             float dxC = float(dyC) * widthScale;
             int leftDxCIndex = int(floor(dxC));
-            int rightDxCIndex = int(min(ceil(dxC), ${s-1}.0));
+            int rightDxCIndex = int(min(ceil(dxC), ${xWidth - 1}.0));
             float dxCLerp = dxC - float(leftDxCIndex);
             float inverseDxCLerp = 1.0 - dxCLerp;
 
@@ -4252,11 +61760,51 @@ return a / b;`,Xnt=`
 
         setOutput(accumulator);
       }
-    `}};function vot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n,a=new jC(s.shape,o.shape,i);return e.runWebGLProgram(a,[s],s.dtype)}var eB={kernelName:Op,backendName:"webgl",kernelFunc:vot};var XC=class{constructor(t,e,n,o,s){this.variableNames=["A"],this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m=o?"0.5":"0.0",f;s?f="max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":f="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeBilinearGrad.js
+function resizeBilinearGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  const program = new ResizeBilinearBackpropProgram(dy.shape, images.shape, alignCorners);
+  return backend2.runWebGLProgram(program, [dy], dy.dtype);
+}
+var resizeBilinearGradConfig3 = {
+  kernelName: ResizeBilinearGrad,
+  backendName: "webgl",
+  kernelFunc: resizeBilinearGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_gpu.js
+var ResizeNearestNeighborProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    const roundBase = alignCorners ? "0.5" : "0.0";
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))`;
+    } else {
+      sourceFracIndexRC = `vec2(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec2 effectiveInputOverOutputRatioRC = vec2(
-          ${c[0]/p[0]},
-          ${c[1]/p[1]});
-      const vec2 inputShapeRC = vec2(${a}.0, ${u}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec2 inputShapeRC = vec2(${oldHeight}.0, ${oldWidth}.0);
 
       void main() {
         ivec4 coords = getOutputCoords();
@@ -4265,22 +61813,50 @@ return a / b;`,Xnt=`
         ivec2 yRC = coords.yz;
 
         // Fractional source index.
-        vec2 sourceFracIndexRC = ${f};
+        vec2 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the coordinators of nearest neighbor point.
         ivec2 sourceNearestRC = ivec2(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${m})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));
         float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);
 
         setOutput(newValue);
       }
-    `}};var YC=class{constructor(t,e,n,o,s){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[i,a,u,l]=t;this.outputShape=[i,e,n,l];let c=[o&&e>1?a-1:a,o&&n>1?u-1:u],p=[o&&e>1?e-1:e,o&&n>1?n-1:n],m=o?"0.5":"0.0",f;s?f="max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":f="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_packed_gpu.js
+var ResizeNearestNeighborPackedProgram = class {
+  constructor(inputShape, newHeight, newWidth, alignCorners, halfPixelCenters) {
+    this.variableNames = ["A"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    this.outputShape = [];
+    const [batch, oldHeight, oldWidth, depth] = inputShape;
+    this.outputShape = [batch, newHeight, newWidth, depth];
+    const effectiveInSize = [
+      alignCorners && newHeight > 1 ? oldHeight - 1 : oldHeight,
+      alignCorners && newWidth > 1 ? oldWidth - 1 : oldWidth
+    ];
+    const effectiveOutSize = [
+      alignCorners && newHeight > 1 ? newHeight - 1 : newHeight,
+      alignCorners && newWidth > 1 ? newWidth - 1 : newWidth
+    ];
+    const roundBase = alignCorners ? "0.5" : "0.0";
+    let sourceFracIndexRC;
+    if (halfPixelCenters) {
+      sourceFracIndexRC = `max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))`;
+    } else {
+      sourceFracIndexRC = `vec3(yRC) * effectiveInputOverOutputRatioRC`;
+    }
+    this.userCode = `
       const vec3 effectiveInputOverOutputRatioRC = vec3(
-          ${c[0]/p[0]},
-          ${c[1]/p[1]},
-          ${c[1]/p[1]});
-      const vec3 inputShapeRC = vec3(${a}.0, ${u}.0,
-                                     ${u}.0);
+          ${effectiveInSize[0] / effectiveOutSize[0]},
+          ${effectiveInSize[1] / effectiveOutSize[1]},
+          ${effectiveInSize[1] / effectiveOutSize[1]});
+      const vec3 inputShapeRC = vec3(${oldHeight}.0, ${oldWidth}.0,
+                                     ${oldWidth}.0);
 
       float getAValue(int b, int r, int c, int d) {
         return getChannel(getA(b, r, c, d), vec2(c, d));
@@ -4294,15 +61870,15 @@ return a / b;`,Xnt=`
         ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
 
         // Fractional source index.
-        vec3 sourceFracIndexRC = ${f};
+        vec3 sourceFracIndexRC = ${sourceFracIndexRC};
 
         // Compute the coordinators of nearest neighbor point.
         ivec3 sourceNearestRC = ivec3(
-          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${m})));
+          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${roundBase})));
 
         // Should we calculate next column and row elements in 2x2 packed cell.
-        bool hasNextCol = d < ${l-1};
-        bool hasNextRow = coords.z < ${n-1};
+        bool hasNextCol = d < ${depth - 1};
+        bool hasNextRow = coords.z < ${newWidth - 1};
 
         vec4 newValue = vec4(
           getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),
@@ -4315,7 +61891,48 @@ return a / b;`,Xnt=`
 
         setOutput(newValue);
       }
-    `}};function Not(r){let{inputs:t,backend:e,attrs:n}=r,{images:o}=t,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,c=z().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new YC(o.shape,u,l,s,i):new XC(o.shape,u,l,s,i);return e.runWebGLProgram(c,[o],o.dtype)}var rB={kernelName:Is,backendName:"webgl",kernelFunc:Not};var ZC=class{constructor(t,e,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=e;let[,o,s]=e,[,i,a]=t,u=[n&&i>1?o-1:o,n&&a>1?s-1:s],l=[n&&i>1?i-1:i,n&&a>1?a-1:a],c=u[0]/l[0],p=u[1]/l[1],m=1/c,f=1/p,d=Math.ceil(m)*2+2,h=Math.ceil(f)*2+2;this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeNearestNeighbor.js
+function resizeNearestNeighbor3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const program = env().getBool("WEBGL_PACK_IMAGE_OPERATIONS") ? new ResizeNearestNeighborPackedProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters) : new ResizeNearestNeighborProgram(images.shape, newHeight, newWidth, alignCorners, halfPixelCenters);
+  return backend2.runWebGLProgram(program, [images], images.dtype);
+}
+var resizeNearestNeighborConfig2 = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "webgl",
+  kernelFunc: resizeNearestNeighbor3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/resize_nearest_neighbor_backprop_gpu.js
+var ResizeNearestNeigborBackpropProgram = class {
+  constructor(dyShape, inputShape, alignCorners) {
+    this.variableNames = ["dy"];
+    this.outputShape = [];
+    this.outputShape = inputShape;
+    const [, xHeight, xWidth] = inputShape;
+    const [, yHeight, yWidth] = dyShape;
+    const effectiveXSize = [
+      alignCorners && yHeight > 1 ? xHeight - 1 : xHeight,
+      alignCorners && yWidth > 1 ? xWidth - 1 : xWidth
+    ];
+    const effectiveYSize = [
+      alignCorners && yHeight > 1 ? yHeight - 1 : yHeight,
+      alignCorners && yWidth > 1 ? yWidth - 1 : yWidth
+    ];
+    const heightScale = effectiveXSize[0] / effectiveYSize[0];
+    const widthScale = effectiveXSize[1] / effectiveYSize[1];
+    const invHeightScale = 1 / heightScale;
+    const invWidthScale = 1 / widthScale;
+    const winHeight = Math.ceil(invHeightScale) * 2 + 2;
+    const winWidth = Math.ceil(invWidthScale) * 2 + 2;
+    this.userCode = `
       void main() {
         ivec4 coords = getOutputCoords();
         int b = coords[0];
@@ -4325,14 +61942,14 @@ return a / b;`,Xnt=`
 
         float accumulator = 0.0;
 
-        const float heightScale = float(${c});
-        const float widthScale = float(${p});
+        const float heightScale = float(${heightScale});
+        const float widthScale = float(${widthScale});
 
-        const float invHeightScale = float(${m});
-        const float invWidthScale = float(${f});
+        const float invHeightScale = float(${invHeightScale});
+        const float invWidthScale = float(${invWidthScale});
 
-        const int winHeight = int(${d});
-        const int winWidth = int(${h});
+        const int winHeight = int(${winHeight});
+        const int winWidth = int(${winWidth});
 
         // Compute bounds for where in dy we will look
         float startRLerp = floor(float(r) * invHeightScale);
@@ -4346,7 +61963,7 @@ return a / b;`,Xnt=`
           int dyR = dyROffset + startDyR;
 
           // Guard against the window exceeding the bounds of dy
-          if (dyR < 0 || dyR >= ${i}) {
+          if (dyR < 0 || dyR >= ${yHeight}) {
             continue;
           }
 
@@ -4354,26 +61971,26 @@ return a / b;`,Xnt=`
             int dyC = dyCOffset + startDyC;
 
             // Guard against the window exceeding the bounds of dy
-            if (dyC < 0 || dyC >= ${a}) {
+            if (dyC < 0 || dyC >= ${yWidth}) {
               continue;
             }
 
             float sourceFracRow =
-              float(${u[0]}) *
-                (float(dyR) / float(${l[0]}));
+              float(${effectiveXSize[0]}) *
+                (float(dyR) / float(${effectiveYSize[0]}));
 
             float sourceFracCol =
-                float(${u[1]}) *
-                  (float(dyC) / float(${l[1]}));
+                float(${effectiveXSize[1]}) *
+                  (float(dyC) / float(${effectiveYSize[1]}));
 
             int sourceNearestRow = int(min(
-                float(int(${o}) - 1),
-                ${n} ? float(round(sourceFracRow)) :
+                float(int(${xHeight}) - 1),
+                ${alignCorners} ? float(round(sourceFracRow)) :
                                   float(floor(sourceFracRow))));
 
             int sourceNearestCol = int(min(
-                float(int(${s}) - 1),
-                ${n} ? float(round(sourceFracCol)) :
+                float(int(${xWidth}) - 1),
+                ${alignCorners} ? float(round(sourceFracCol)) :
                                   float(floor(sourceFracCol))));
 
             if (r == sourceNearestRow && c == sourceNearestCol) {
@@ -4385,47 +62002,176 @@ return a / b;`,Xnt=`
 
         setOutput(accumulator);
       }
-    `}};function Tot(r){let{inputs:t,backend:e,attrs:n}=r,{images:o,dy:s}=t,{alignCorners:i}=n,a=new ZC(s.shape,o.shape,i);return e.runWebGLProgram(a,[s],s.dtype)}var nB={kernelName:Fp,backendName:"webgl",kernelFunc:Tot};var JC=class{constructor(t,e){this.variableNames=["x"];let n=t.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=t,n===1){this.userCode=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ResizeNearestNeighborGrad.js
+function resizeNearestNeighborGrad2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { images, dy } = inputs;
+  const { alignCorners } = attrs;
+  const program = new ResizeNearestNeigborBackpropProgram(dy.shape, images.shape, alignCorners);
+  return backend2.runWebGLProgram(program, [dy], dy.dtype);
+}
+var resizeNearestNeighborGradConfig3 = {
+  kernelName: ResizeNearestNeighborGrad,
+  backendName: "webgl",
+  kernelFunc: resizeNearestNeighborGrad2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/reverse_gpu.js
+var ReverseProgram = class {
+  constructor(xShape, axis) {
+    this.variableNames = ["x"];
+    const rank = xShape.length;
+    if (rank > 4) {
+      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);
+    }
+    this.outputShape = xShape;
+    if (rank === 1) {
+      this.userCode = `
         void main() {
           int coord = getOutputCoords();
-          setOutput(getX(${t[0]} - coord - 1));
+          setOutput(getX(${xShape[0]} - coord - 1));
         }
-      `;return}let o=a=>e.indexOf(a)!==-1&&t[a]!==1?`${t[a]} - coords[${a}] - 1`:`coords[${a}]`,s=t.map((a,u)=>o(u)).join(","),i=zt(n);this.userCode=`
-      void main() {
-        ${i} coords = getOutputCoords();
-        setOutput(getX(${s}));
+      `;
+      return;
+    }
+    const getInCoord = (i) => {
+      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {
+        return `${xShape[i]} - coords[${i}] - 1`;
       }
-    `}};var QC=class{constructor(t,e){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let n=t.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=t;let o=Qe("rc",n),s=`${o[n-1]} + 1 < ${this.outputShape[n-1]}`,i=`${o[n-2]} + 1 < ${this.outputShape[n-2]}`,a=zt(n);n===1?this.userCode=`
+      return `coords[${i}]`;
+    };
+    const inCoords = xShape.map((_, i) => getInCoord(i)).join(",");
+    const type = getCoordsDataType(rank);
+    this.userCode = `
+      void main() {
+        ${type} coords = getOutputCoords();
+        setOutput(getX(${inCoords}));
+      }
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/reverse_packed_gpu.js
+var ReversePackedProgram = class {
+  constructor(xShape, axis) {
+    this.variableNames = ["x"];
+    this.packedInputs = true;
+    this.packedOutput = true;
+    const rank = xShape.length;
+    if (rank > 4) {
+      throw new Error(`WebGL backend: Reverse of rank-${rank} tensor is not yet supported`);
+    }
+    this.outputShape = xShape;
+    const channels = getChannels("rc", rank);
+    const nextColumn = `${channels[rank - 1]} + 1 < ${this.outputShape[rank - 1]}`;
+    const nextRow = `${channels[rank - 2]} + 1 < ${this.outputShape[rank - 2]}`;
+    const type = getCoordsDataType(rank);
+    if (rank === 1) {
+      this.userCode = `
         void main(){
           int rc = getOutputCoords();
           vec4 result = vec4(0.);
-          result.r = getChannel(getX(${t[0]} - rc - 1),
-            ${t[0]} - rc - 1);
-          if(${s}){
-              result.g = getChannel(getX(${t[0]} - (rc  + 1) - 1),
-                ${t[0]} - (rc  + 1) - 1);
+          result.r = getChannel(getX(${xShape[0]} - rc - 1),
+            ${xShape[0]} - rc - 1);
+          if(${nextColumn}){
+              result.g = getChannel(getX(${xShape[0]} - (rc  + 1) - 1),
+                ${xShape[0]} - (rc  + 1) - 1);
           }
           setOutput(result);
         }
-      `:this.userCode=`
+      `;
+    } else {
+      this.userCode = `
         void main() {
-          ${a} rc = getOutputCoords();
+          ${type} rc = getOutputCoords();
           vec4 result = vec4(0.);
-          result.r = ${u(o.slice())};
-          if(${s}){
-            result.g = ${l(o.slice())};
+          result.r = ${getR(channels.slice())};
+          if(${nextColumn}){
+            result.g = ${getG(channels.slice())};
           }
-          if(${i}) {
-            result.b = ${c(o.slice())};
-            if(${s}) {
-              result.a = ${p(o.slice())};
+          if(${nextRow}) {
+            result.b = ${getB(channels.slice())};
+            if(${nextColumn}) {
+              result.a = ${getA(channels.slice())};
             }
           }
           setOutput(result);
         }
-    `;function u(d){return m(d)}function l(d){return d[n-1]="("+d[n-1]+" + 1)",m(d)}function c(d){return d[n-2]="("+d[n-2]+" + 1)",m(d)}function p(d){return d[n-1]="("+d[n-1]+" + 1)",d[n-2]="("+d[n-2]+" + 1)",m(d)}function m(d){let h=t.map((b,w)=>f(w,d)),g=h.join(","),x=h.slice(-2).join(",");return`getChannel(getX(${g}), vec2(${x}))`}function f(d,h){return e.indexOf(d)!==-1&&t[d]!==1?`${t[d]} - ${h[d]} - 1`:`${h[d]}`}}};function kot(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n,i=o.shape.length,a=y.parseAxisParam(s,o.shape);if(i===0)return tr({inputs:{x:o},backend:e});let u=z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new QC(o.shape,a):new JC(o.shape,a);return e.runWebGLProgram(u,[o],o.dtype)}var oB={kernelName:Ns,backendName:"webgl",kernelFunc:kot};var tI=class{constructor(t,e){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let n=t[1],o=t[2];this.outputShape=t;let s="";typeof e=="number"?s=`float outputValue = ${e.toFixed(2)};`:s=`
-        vec3 fill = vec3(${e.join(",")});
-        float outputValue = fill[coords[3]];`,this.userCode=`
+    `;
+    }
+    function getR(channels2) {
+      return getChannel(channels2);
+    }
+    function getG(channels2) {
+      channels2[rank - 1] = "(" + channels2[rank - 1] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getB(channels2) {
+      channels2[rank - 2] = "(" + channels2[rank - 2] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getA(channels2) {
+      channels2[rank - 1] = "(" + channels2[rank - 1] + ` + 1)`;
+      channels2[rank - 2] = "(" + channels2[rank - 2] + ` + 1)`;
+      return getChannel(channels2);
+    }
+    function getChannel(channels2) {
+      const inCoordsArray = xShape.map((_, i) => getInCoord(i, channels2));
+      const inCoords = inCoordsArray.join(",");
+      const innerDims = inCoordsArray.slice(-2).join(",");
+      return `getChannel(getX(${inCoords}), vec2(${innerDims}))`;
+    }
+    function getInCoord(i, channels1) {
+      if (axis.indexOf(i) !== -1 && xShape[i] !== 1) {
+        return `${xShape[i]} - ${channels1[i]} - 1`;
+      } else {
+        return `${channels1[i]}`;
+      }
+    }
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Reverse.js
+function reverse3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  const xRank = x.shape.length;
+  const $dims = util_exports.parseAxisParam(dims, x.shape);
+  if (xRank === 0) {
+    return identity3({ inputs: { x }, backend: backend2 });
+  }
+  const program = env().getBool("WEBGL_PACK_ARRAY_OPERATIONS") ? new ReversePackedProgram(x.shape, $dims) : new ReverseProgram(x.shape, $dims);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var reverseConfig2 = {
+  kernelName: Reverse,
+  backendName: "webgl",
+  kernelFunc: reverse3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/rotate_gpu.js
+var RotateProgram = class {
+  constructor(imageShape, fillValue) {
+    this.variableNames = ["Image"];
+    this.outputShape = [];
+    this.customUniforms = [{ name: "params", type: "vec4" }];
+    const imageHeight = imageShape[1];
+    const imageWidth = imageShape[2];
+    this.outputShape = imageShape;
+    let fillSnippet = "";
+    if (typeof fillValue === "number") {
+      fillSnippet = `float outputValue = ${fillValue.toFixed(2)};`;
+    } else {
+      fillSnippet = `
+        vec3 fill = vec3(${fillValue.join(",")});
+        float outputValue = fill[coords[3]];`;
+    }
+    this.userCode = `
         void main() {
           ivec4 coords = getOutputCoords();
           int x = coords[2];
@@ -4436,13 +62182,34 @@ return a / b;`,Xnt=`
             (float(y) - params[1]) * params[3];
           int coordX = int(round(coordXFloat + params[0]));
           int coordY = int(round(coordYFloat + params[1]));
-          ${s}
-          if(coordX >= 0 && coordX < ${o} && coordY >= 0 && coordY < ${n}) {
+          ${fillSnippet}
+          if(coordX >= 0 && coordX < ${imageWidth} && coordY >= 0 && coordY < ${imageHeight}) {
             outputValue = getImage(coords[0], coordY, coordX, coords[3]);
           }
           setOutput(outputValue);
         }
-    `}};var sB={kernelName:qa,backendName:"webgl",kernelFunc:({inputs:r,attrs:t,backend:e})=>{let{image:n}=r,{radians:o,fillValue:s,center:i}=t,a=e,u=new tI(n.shape,s),[l,c]=v.getImageCenter(i,n.shape[1],n.shape[2]),p=[[l,c,Math.sin(o),Math.cos(o)]];return a.runWebGLProgram(u,[n],n.dtype,p)}};var Eot=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/RotateWithOffset.js
+var rotateWithOffsetConfig2 = {
+  kernelName: RotateWithOffset,
+  backendName: "webgl",
+  kernelFunc: ({ inputs, attrs, backend: backend2 }) => {
+    const { image: image2 } = inputs;
+    const { radians, fillValue, center } = attrs;
+    const webglBackend = backend2;
+    const program = new RotateProgram(image2.shape, fillValue);
+    const [centerX, centerY] = backend_util_exports.getImageCenter(center, image2.shape[1], image2.shape[2]);
+    const customValues = [[centerX, centerY, Math.sin(radians), Math.cos(radians)]];
+    const output = webglBackend.runWebGLProgram(program, [image2], image2.dtype, customValues);
+    return output;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Round.js
+var ROUND = `
   // OpenGL ES does not support round function.
   // The algorithm is based on banker's rounding.
   float base = floor(x);
@@ -4457,34 +62224,115 @@ return a / b;`,Xnt=`
       return base + 1.0;
     }
   }
-`,_ot=Ct({opSnippet:Eot}),iB={kernelName:Ts,backendName:"webgl",kernelFunc:_ot};var Aot="return inversesqrt(x);",$ot=Ct({opSnippet:Aot,cpuKernelImpl:ML}),aB={kernelName:ks,backendName:"webgl",kernelFunc:$ot};var $d=class{constructor(t,e,n,o,s,i,a=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=i;let u=zt(s.length),l=zt(i.length),c="";n===1?c="i":n===2&&(c="i, j");let p=`getIndices(${c})`,m="";o===1?m="i":o===2&&(m="i, coords[1]");let f=`getUpdates(${m})`,d=e>1?"strides[j]":"strides";this.userCode=`
-        ${u} strides = ${u}(${s});
+`;
+var round4 = unaryKernelFunc2({ opSnippet: ROUND });
+var roundConfig2 = {
+  kernelName: Round,
+  backendName: "webgl",
+  kernelFunc: round4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Rsqrt.js
+var RSQRT = `return inversesqrt(x);`;
+var rsqrt3 = unaryKernelFunc2({ opSnippet: RSQRT, cpuKernelImpl: rsqrtImplCPU });
+var rsqrtConfig2 = {
+  kernelName: Rsqrt,
+  backendName: "webgl",
+  kernelFunc: rsqrt3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/scatter_gpu.js
+var ScatterProgram = class {
+  constructor(updateSize, sliceDim, indicesRank, updatesRank, strides, shape, summingDupeIndex = true) {
+    this.variableNames = ["updates", "indices", "defaultValue"];
+    this.outputShape = shape;
+    const stridesType = getCoordsDataType(strides.length);
+    const dtype = getCoordsDataType(shape.length);
+    let indicesString = "";
+    if (indicesRank === 1) {
+      indicesString = "i";
+    } else if (indicesRank === 2) {
+      indicesString = "i, j";
+    }
+    const indicesSnippet = `getIndices(${indicesString})`;
+    let updatesString = "";
+    if (updatesRank === 1) {
+      updatesString = "i";
+    } else if (updatesRank === 2) {
+      updatesString = "i, coords[1]";
+    }
+    const updatesSnippet = `getUpdates(${updatesString})`;
+    const strideString = sliceDim > 1 ? "strides[j]" : "strides";
+    this.userCode = `
+        ${stridesType} strides = ${stridesType}(${strides});
 
         void main() {
-          ${l} coords = getOutputCoords();
+          ${dtype} coords = getOutputCoords();
           float sum = 0.0;
           bool found = false;
-          for (int i = 0; i < ${t}; i++) {
+          for (int i = 0; i < ${updateSize}; i++) {
             int flattenedIndex = 0;
-            for (int j = 0; j < ${e}; j++) {
-              int index = round(${p});
-              flattenedIndex += index * ${d};
+            for (int j = 0; j < ${sliceDim}; j++) {
+              int index = round(${indicesSnippet});
+              flattenedIndex += index * ${strideString};
             }
             if (flattenedIndex == coords[0]) {
-              sum += ${f};
+              sum += ${updatesSnippet};
               found = true;
             }
           }
           setOutput(mix(getDefaultValue(), sum, float(found)));
         }
-      `}};function Dot(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o,updates:s}=t,{shape:i}=n,{sliceRank:a,numUpdates:u,sliceSize:l,strides:c,outputSize:p}=v.calculateShapes(s,o,i),m=[p/l,l];if(p===0)return e.makeTensorInfo(i,o.dtype);let f=st({inputs:{x:o},backend:e,attrs:{shape:[u,a]}}),d=st({inputs:{x:s},backend:e,attrs:{shape:[u,l]}}),h=e.makeTensorInfo([],"float32",new Float32Array([0])),g=new $d(u,a,f.shape.length,d.shape.length,c,m),x=e.runWebGLProgram(g,[d,f,h],d.dtype),b=st({inputs:{x},backend:e,attrs:{shape:i}});return e.disposeIntermediateTensorInfo(f),e.disposeIntermediateTensorInfo(d),e.disposeIntermediateTensorInfo(x),e.disposeIntermediateTensorInfo(h),b}var lB={kernelName:La,backendName:"webgl",kernelFunc:Dot};var eI=class{constructor(t,e,n,o){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[t,n];let s="while (left < right) {",i=`for (int i = 0; i < ${Math.ceil(Math.log2(e+1))}; ++i) { if (left >= right) break;`,a=z().getNumber("WEBGL_VERSION")===2?s:i,u=o==="left"?"<":"<=";this.userCode=`
+      `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/ScatterNd.js
+function scatterNd2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(updates, indices, shape);
+  const flattenShape = [outputSize / sliceSize, sliceSize];
+  if (outputSize === 0) {
+    return backend2.makeTensorInfo(shape, indices.dtype);
+  }
+  const flattenIndices = reshape4({ inputs: { x: indices }, backend: backend2, attrs: { shape: [numUpdates, sliceRank] } });
+  const flattenX = reshape4({ inputs: { x: updates }, backend: backend2, attrs: { shape: [numUpdates, sliceSize] } });
+  const defaultValue = backend2.makeTensorInfo([], "float32", new Float32Array([0]));
+  const program = new ScatterProgram(numUpdates, sliceRank, flattenIndices.shape.length, flattenX.shape.length, strides, flattenShape);
+  const res = backend2.runWebGLProgram(program, [flattenX, flattenIndices, defaultValue], flattenX.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape } });
+  backend2.disposeIntermediateTensorInfo(flattenIndices);
+  backend2.disposeIntermediateTensorInfo(flattenX);
+  backend2.disposeIntermediateTensorInfo(res);
+  backend2.disposeIntermediateTensorInfo(defaultValue);
+  return reshaped;
+}
+var scatterNdConfig2 = {
+  kernelName: ScatterNd,
+  backendName: "webgl",
+  kernelFunc: scatterNd2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/search_sorted_gpu.js
+var SearchSortedProgram = class {
+  constructor(batchSize, numInputs, numValues, side) {
+    this.variableNames = ["sortedSequence", "values"];
+    this.customUniforms = [{ name: "numInputs", type: "int" }];
+    this.outputShape = [batchSize, numValues];
+    const webGL2LoopHead = "while (left < right) {";
+    const webGL1LoopHead = `for (int i = 0; i < ${Math.ceil(Math.log2(numInputs + 1))}; ++i) { if (left >= right) break;`;
+    const loopHead = env().getNumber("WEBGL_VERSION") === 2 ? webGL2LoopHead : webGL1LoopHead;
+    const boundComparator = side === "left" ? "<" : "<=";
+    this.userCode = `
        int findBound(int batch, float value) {
          int left = 0;
          int right = numInputs;
          int mid;
-         ${a}
+         ${loopHead}
            mid = (left + right) / 2;
-           if (getSortedSequence(batch, mid) ${u} value) {
+           if (getSortedSequence(batch, mid) ${boundComparator} value) {
              left = mid + 1;
            } else {
              right = mid;
@@ -4502,25 +62350,99 @@ return a / b;`,Xnt=`
 
          setOutput(float(findBound(batch, value)));
        }
-     `}};function Rot(r){let{inputs:t,backend:e,attrs:n}=r,{sortedSequence:o,values:s}=t,{side:i}=n,a=new eI(o.shape[0],o.shape[1],s.shape[1],i),u=[[o.shape[1]]];return e.runWebGLProgram(a,[o,s],"int32",u)}var uB={kernelName:Pp,backendName:"webgl",kernelFunc:Rot};var rI=class{constructor(t,e,n){this.variableNames=["c","a","b"],this.outputShape=e;let o,s;if(n>4)throw Error(`Where for rank ${n} is not yet supported`);if(n===1)s="resRC",o="resRC";else{let a=["resRC.x","resRC.y","resRC.z","resRC.w"],u=[],l=[];for(let c=0;c<e.length;c++)l.push(`${a[c]}`),c<t&&u.push(`${a[c]}`);o=u.join(),s=l.join()}let i=zt(n);this.userCode=`
-      void main() {
-        ${i} resRC = getOutputCoords();
-        float cVal = getC(${o});
-        if (cVal >= 1.0) {
-          setOutput(getA(${s}));
-        } else {
-          setOutput(getB(${s}));
+     `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SearchSorted.js
+function searchSorted3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sortedSequence, values } = inputs;
+  const { side } = attrs;
+  const program = new SearchSortedProgram(sortedSequence.shape[0], sortedSequence.shape[1], values.shape[1], side);
+  const customValues = [[sortedSequence.shape[1]]];
+  return backend2.runWebGLProgram(program, [sortedSequence, values], "int32", customValues);
+}
+var searchSortedConfig2 = {
+  kernelName: SearchSorted,
+  backendName: "webgl",
+  kernelFunc: searchSorted3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/select_gpu.js
+var SelectProgram = class {
+  constructor(cRank, shape, rank) {
+    this.variableNames = ["c", "a", "b"];
+    this.outputShape = shape;
+    let cCoords;
+    let abCoords;
+    if (rank > 4) {
+      throw Error(`Where for rank ${rank} is not yet supported`);
+    }
+    if (rank === 1) {
+      abCoords = `resRC`;
+      cCoords = `resRC`;
+    } else {
+      const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w"];
+      const cCoordVars = [];
+      const abCoordVars = [];
+      for (let i = 0; i < shape.length; i++) {
+        abCoordVars.push(`${currentCoords[i]}`);
+        if (i < cRank) {
+          cCoordVars.push(`${currentCoords[i]}`);
         }
       }
-    `}};function Fot(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t,i=new rI(n.shape.length,o.shape,o.shape.length);return e.runWebGLProgram(i,[n,o,s],sr(o.dtype,s.dtype))}var cB={kernelName:hi,backendName:"webgl",kernelFunc:Fot};var Oot=`
+      cCoords = cCoordVars.join();
+      abCoords = abCoordVars.join();
+    }
+    const dtype = getCoordsDataType(rank);
+    this.userCode = `
+      void main() {
+        ${dtype} resRC = getOutputCoords();
+        float cVal = getC(${cCoords});
+        if (cVal >= 1.0) {
+          setOutput(getA(${abCoords}));
+        } else {
+          setOutput(getB(${abCoords}));
+        }
+      }
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Select.js
+function select3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  const program = new SelectProgram(condition.shape.length, t.shape, t.shape.length);
+  return backend2.runWebGLProgram(program, [condition, t, e], upcastType(t.dtype, e.dtype));
+}
+var selectConfig2 = {
+  kernelName: Select,
+  backendName: "webgl",
+  kernelFunc: select3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Selu.js
+var SELU = `
   // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.
   // see: https://arxiv.org/abs/1706.02515
-  float scaleAlpha = ${v.SELU_SCALEALPHA};
-  float scale = ${v.SELU_SCALE};
+  float scaleAlpha = ${backend_util_exports.SELU_SCALEALPHA};
+  float scale = ${backend_util_exports.SELU_SCALE};
   return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);
-`,Pot=Ct({opSnippet:Oot}),pB={kernelName:Ma,backendName:"webgl",kernelFunc:Pot};var Lot=Po+`
+`;
+var selu3 = unaryKernelFunc2({ opSnippet: SELU });
+var seluConfig2 = {
+  kernelName: Selu,
+  backendName: "webgl",
+  kernelFunc: selu3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sigmoid.js
+var SIGMOID3 = CHECK_NAN_SNIPPET_UNARY + `
   return 1.0 / (1.0 + exp(-1.0 * x));
-`,Mot=`
+`;
+var SIGMOID_PACKED = `
   vec4 result = 1.0 / (1.0 + exp(-1.0 * x));
   bvec4 isNaN = isnan(x);
 
@@ -4530,15 +62452,55 @@ return a / b;`,Xnt=`
   result.a = isNaN.a ? x.a : result.a;
 
   return result;
-`,zot=Ct({opSnippet:Lot,packedOpSnippet:Mot,cpuKernelImpl:BL}),mB={kernelName:_s,backendName:"webgl",kernelFunc:zot};var Bot=`
+`;
+var sigmoid3 = unaryKernelFunc2({
+  opSnippet: SIGMOID3,
+  packedOpSnippet: SIGMOID_PACKED,
+  cpuKernelImpl: sigmoidImplCPU
+});
+var sigmoidConfig2 = {
+  kernelName: Sigmoid,
+  backendName: "webgl",
+  kernelFunc: sigmoid3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sign.js
+var SIGN = `
   if (isnan(x)) { return 0.0; }
   return sign(x);
-`,Vot=Ct({opSnippet:Bot}),fB={kernelName:Ba,backendName:"webgl",kernelFunc:Vot};var Got=Po+`
+`;
+var sign3 = unaryKernelFunc2({ opSnippet: SIGN });
+var signConfig2 = {
+  kernelName: Sign,
+  backendName: "webgl",
+  kernelFunc: sign3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sin.js
+var SIN = CHECK_NAN_SNIPPET_UNARY + `
   return sin(x);
-`,Wot=Ct({opSnippet:Got}),dB={kernelName:Es,backendName:"webgl",kernelFunc:Wot};var Uot=`
+`;
+var sin3 = unaryKernelFunc2({ opSnippet: SIN });
+var sinConfig2 = {
+  kernelName: Sin,
+  backendName: "webgl",
+  kernelFunc: sin3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sinh.js
+var SINH = `
   float e2x = exp(x);
   return (e2x - 1.0 / e2x) / 2.0;
-`,Hot=Ct({opSnippet:Uot}),hB={kernelName:za,backendName:"webgl",kernelFunc:Hot};var qot=`
+`;
+var sinh3 = unaryKernelFunc2({ opSnippet: SINH });
+var sinhConfig2 = {
+  kernelName: Sinh,
+  backendName: "webgl",
+  kernelFunc: sinh3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Softplus.js
+var SOFTPLUS = `
   float epsilon = 1.1920928955078125e-7;
   float threshold = log(epsilon) + 2.0;
 
@@ -4558,33 +62520,495 @@ return a / b;`,Xnt=`
     result = log(exp_x + 1.0);
   }
   return result;
-`,Kot=Ct({opSnippet:qot}),gB={kernelName:Va,backendName:"webgl",kernelFunc:Kot};var jot=r=>{let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n;y.assert(o.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let a=s.reduce((x,b)=>x*b),u=[[0,0]];u.push(...i);for(let x=1+s.length;x<o.shape.length;++x)u.push([0,0]);let l=[],c=Tk({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),p=v.getReshaped(c.shape,s,a,!1),m=v.getPermuted(p.length,s.length,!1),f=v.getReshapedPermuted(c.shape,s,a,!1),d=st({inputs:{x:c},backend:e,attrs:{shape:p}}),h=Oe({inputs:{x:d},backend:e,attrs:{perm:m}}),g=st({inputs:{x:h},backend:e,attrs:{shape:f}});return l.push(c),l.push(d),l.push(h),l.forEach(x=>e.disposeIntermediateTensorInfo(x)),g},xB={kernelName:xi,backendName:"webgl",kernelFunc:jot};function Xot(r){let{inputs:t,backend:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
-         ${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
-         ${n.shape}`);if(o.shape.length!==1)throw new Error(`Values must be a vector, saw:
-         ${o.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
-        ${i.shape}`);let a=e.readSync(n.dataId),u=e.readSync(o.dataId),l=e.readSync(s.dataId),c=e.readSync(i.dataId)[0],[p,m,f,d,h]=GL(a,n.shape,n.dtype,u,o.dtype,l,c);return[e.makeTensorInfo(m,n.dtype,p),e.makeTensorInfo([m[0]],o.dtype,f),e.makeTensorInfo([d.length],"bool",new Uint8Array(d.map(g=>Number(g)))),e.makeTensorInfo([h.length],n.dtype,new Int32Array(h))]}var yB={kernelName:Pl,backendName:"webgl",kernelFunc:Xot};function Yot(r){let{inputs:t,backend:e}=r,{inputIndices:n,inputShape:o,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(e.readSync(o.dataId)),a=e.readSync(n.dataId),u=Array.from(e.readSync(s.dataId)),[l,c,p]=WL(a,n.shape,n.dtype,i,u);return[e.makeTensorInfo(c,n.dtype,l),e.makeTensorInfo([p.length],s.dtype,new Int32Array(p))]}var bB={kernelName:Ga,backendName:"webgl",kernelFunc:Yot};function Zot(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-              ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-              ${s.shape}`);let i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=zw(i,n.shape,n.dtype,a,u,!0);return e.makeTensorInfo(c,n.dtype,l)}var wB={kernelName:Ll,backendName:"webgl",kernelFunc:Zot};function Jot(r){let{inputs:t,backend:e}=r,{data:n,indices:o,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(o.shape.length!==1)throw new Error(`Indices should be a vector but received shape
-             ${o.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
-             ${s.shape}`);let i=e.readSync(n.dataId),a=e.readSync(o.dataId),u=e.readSync(s.dataId),[l,c]=zw(i,n.shape,n.dtype,a,u);return e.makeTensorInfo(c,n.dtype,l)}var CB={kernelName:Ml,backendName:"webgl",kernelFunc:Jot};function Qot(r){let{inputs:t,backend:e,attrs:n}=r,{sparseIndices:o,sparseValues:s,defaultValue:i}=t,{outputShape:a}=n,{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=v.calculateShapes(s,o,a),f=!1;if(s.dtype==="string"){let x=e.bufferSync(o),b=e.bufferSync(s),w=y.decodeString(e.readSync(i.dataId)[0]),C=zL(x,b,a,m,c,l,u,p,w,f);return e.makeTensorInfo(a,C.dtype,C.values)}let d=new $d(l,u,o.shape.length,s.shape.length,p,[m,1],f),h=e.runWebGLProgram(d,[s,o,i],s.dtype),g=st({inputs:{x:h},backend:e,attrs:{shape:a}});return e.disposeIntermediateTensorInfo(h),g}var IB={kernelName:Lp,backendName:"webgl",kernelFunc:Qot};function tst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=n,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=o.shape.length,c=new Array(l).fill(0),p=o.shape.slice();return u.map(m=>{let f=[...p];f[a]=m;let d=ri({inputs:{x:o},backend:e,attrs:{begin:c,size:f}});return c[a]+=m,d})}var SB={kernelName:yi,backendName:"webgl",kernelFunc:tst};var vB="return sqrt(x);",est=Ct({opSnippet:vB,packedOpSnippet:vB,cpuKernelImpl:UL}),NB={kernelName:As,backendName:"webgl",kernelFunc:est};var rst="return x * x;",nst=Ct({opSnippet:rst}),TB={kernelName:zl,backendName:"webgl",kernelFunc:nst};var kB="return (a - b) * (a - b);",ost=le({opSnippet:kB,packedOpSnippet:kB}),EB={kernelName:Rs,backendName:"webgl",kernelFunc:ost};function sst({inputs:r,attrs:t,backend:e}){let{x:n}=r,o=fr+`
-    return x > 0.0 ? 1.0 : float(${t.alpha});
-  `,s=new tn(n.shape,o);return e.runWebGLProgram(s,[n],n.dtype)}var _B={kernelName:po,backendName:"webgl",kernelFunc:sst};var nI=class{constructor(t,e,n){this.variableNames=["x"],this.outputShape=n;let o=n.length,s=zt(n.length),i=zt(n.length),a="";if(o===1)a="coords * strides + begin";else{let u=0;a=n.map((l,c)=>(u++,n.length===1?`coords * strides[${c}] + begin[${c}]`:`coords[${u-1}] * strides[${c}] + begin[${c}]`)).join(",")}this.userCode=`
-      ${s} begin = ${s}(${t});
-      ${s} strides = ${s}(${e});
+`;
+var softplus3 = unaryKernelFunc2({ opSnippet: SOFTPLUS });
+var softplusConfig2 = {
+  kernelName: Softplus,
+  backendName: "webgl",
+  kernelFunc: softplus3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SpaceToBatchND.js
+var spaceToBatchND3 = (args) => {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  util_exports.assert(x.shape.length <= 4, () => "spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const toDispose = [];
+  const paddedX = padV22({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapedPaddedX = reshape4({ inputs: { x: paddedX }, backend: backend2, attrs: { shape: reshapedPaddedShape } });
+  const paddedXT = transpose3({
+    inputs: { x: reshapedPaddedX },
+    backend: backend2,
+    attrs: { perm: permutedReshapedPaddedPermutation }
+  });
+  const result = reshape4({ inputs: { x: paddedXT }, backend: backend2, attrs: { shape: flattenShape } });
+  toDispose.push(paddedX);
+  toDispose.push(reshapedPaddedX);
+  toDispose.push(paddedXT);
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+};
+var spaceToBatchNDConfig2 = {
+  kernelName: SpaceToBatchND,
+  backendName: "webgl",
+  kernelFunc: spaceToBatchND3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseFillEmptyRows.js
+function sparseFillEmptyRows3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  if (denseShape.shape.length !== 1) {
+    throw new Error(`Dense shape must be a vector, saw:
+         ${denseShape.shape}`);
+  }
+  if (indices.shape.length !== 2) {
+    throw new Error(`Indices must be a matrix, saw:
+         ${indices.shape}`);
+  }
+  if (values.shape.length !== 1) {
+    throw new Error(`Values must be a vector, saw:
+         ${values.shape}`);
+  }
+  if (defaultValue.shape.length !== 0) {
+    throw new Error(`Default value must be a scalar, saw:
+        ${defaultValue.shape}`);
+  }
+  const $indices = backend2.readSync(indices.dataId);
+  const $values = backend2.readSync(values.dataId);
+  const $denseShape = backend2.readSync(denseShape.dataId);
+  const $defaultValue = backend2.readSync(defaultValue.dataId)[0];
+  const [outputIndices, outputIndicesShape, outputValues, emptyRowIndicator, reverseIndexMap] = sparseFillEmptyRowsImplCPU($indices, indices.shape, indices.dtype, $values, values.dtype, $denseShape, $defaultValue);
+  return [
+    backend2.makeTensorInfo(outputIndicesShape, indices.dtype, outputIndices),
+    backend2.makeTensorInfo([outputIndicesShape[0]], values.dtype, outputValues),
+    backend2.makeTensorInfo([emptyRowIndicator.length], "bool", new Uint8Array(emptyRowIndicator.map((value) => Number(value)))),
+    backend2.makeTensorInfo([reverseIndexMap.length], indices.dtype, new Int32Array(reverseIndexMap))
+  ];
+}
+var sparseFillEmptyRowsConfig2 = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "webgl",
+  kernelFunc: sparseFillEmptyRows3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseReshape.js
+function sparseReshape3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const $inputShape = Array.from(backend2.readSync(inputShape.dataId));
+  const $inputIndices = backend2.readSync(inputIndices.dataId);
+  const targetShape = Array.from(backend2.readSync(newShape.dataId));
+  const [newIndices, indicesShape, outputShape] = sparseReshapeImplCPU($inputIndices, inputIndices.shape, inputIndices.dtype, $inputShape, targetShape);
+  return [
+    backend2.makeTensorInfo(indicesShape, inputIndices.dtype, newIndices),
+    backend2.makeTensorInfo([outputShape.length], newShape.dtype, new Int32Array(outputShape))
+  ];
+}
+var sparseReshapeConfig2 = {
+  kernelName: SparseReshape,
+  backendName: "webgl",
+  kernelFunc: sparseReshape3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseSegmentMean.js
+function sparseSegmentMean3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+              ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+              ${segmentIds.shape}`);
+  }
+  const $data = backend2.readSync(data.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const $segmentIds = backend2.readSync(segmentIds.dataId);
+  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds, true);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentMeanConfig2 = {
+  kernelName: SparseSegmentMean,
+  backendName: "webgl",
+  kernelFunc: sparseSegmentMean3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseSegmentSum.js
+function sparseSegmentSum3(args) {
+  const { inputs, backend: backend2 } = args;
+  const { data, indices, segmentIds } = inputs;
+  if (data.shape.length < 1) {
+    throw new Error(`Data should be at least 1 dimensional but received scalar`);
+  }
+  if (indices.shape.length !== 1) {
+    throw new Error(`Indices should be a vector but received shape
+             ${indices.shape}`);
+  }
+  if (segmentIds.shape.length !== 1) {
+    throw new Error(`Segment ids should be a vector but received shape
+             ${segmentIds.shape}`);
+  }
+  const $data = backend2.readSync(data.dataId);
+  const $indices = backend2.readSync(indices.dataId);
+  const $segmentIds = backend2.readSync(segmentIds.dataId);
+  const [outputData, outputDataShape] = sparseSegmentReductionImplCPU($data, data.shape, data.dtype, $indices, $segmentIds);
+  return backend2.makeTensorInfo(outputDataShape, data.dtype, outputData);
+}
+var sparseSegmentSumConfig2 = {
+  kernelName: SparseSegmentSum,
+  backendName: "webgl",
+  kernelFunc: sparseSegmentSum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SparseToDense.js
+function sparseToDense3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { sparseIndices, sparseValues, defaultValue } = inputs;
+  const { outputShape } = attrs;
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = backend_util_exports.calculateShapes(sparseValues, sparseIndices, outputShape);
+  const sumDupeIndices = false;
+  if (sparseValues.dtype === "string") {
+    const indicesBuf = backend2.bufferSync(sparseIndices);
+    const updatesBuf = backend2.bufferSync(sparseValues);
+    const $defaultValue = util_exports.decodeString(backend2.readSync(defaultValue.dataId)[0]);
+    const outBuf = scatterImplCPU(indicesBuf, updatesBuf, outputShape, outputSize, sliceSize, numUpdates, sliceRank, strides, $defaultValue, sumDupeIndices);
+    return backend2.makeTensorInfo(outputShape, outBuf.dtype, outBuf.values);
+  }
+  const program = new ScatterProgram(numUpdates, sliceRank, sparseIndices.shape.length, sparseValues.shape.length, strides, [outputSize, 1], sumDupeIndices);
+  const res = backend2.runWebGLProgram(program, [sparseValues, sparseIndices, defaultValue], sparseValues.dtype);
+  const reshaped = reshape4({ inputs: { x: res }, backend: backend2, attrs: { shape: outputShape } });
+  backend2.disposeIntermediateTensorInfo(res);
+  return reshaped;
+}
+var sparseToDenseConfig2 = {
+  kernelName: SparseToDense,
+  backendName: "webgl",
+  kernelFunc: sparseToDense3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SplitV.js
+function splitV2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const xRank = x.shape.length;
+  const begin = new Array(xRank).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const sliceSize = [...size];
+    sliceSize[$axis] = s;
+    const sliceT = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size: sliceSize } });
+    begin[$axis] += s;
+    return sliceT;
+  });
+}
+var splitVConfig2 = {
+  kernelName: SplitV,
+  backendName: "webgl",
+  kernelFunc: splitV2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Sqrt.js
+var SQRT = `return sqrt(x);`;
+var sqrt3 = unaryKernelFunc2({ opSnippet: SQRT, packedOpSnippet: SQRT, cpuKernelImpl: sqrtImplCPU });
+var sqrtConfig2 = {
+  kernelName: Sqrt,
+  backendName: "webgl",
+  kernelFunc: sqrt3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Square.js
+var SQUARE = `return x * x;`;
+var square3 = unaryKernelFunc2({ opSnippet: SQUARE });
+var squareConfig2 = {
+  kernelName: Square,
+  backendName: "webgl",
+  kernelFunc: square3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/SquaredDifference.js
+var SQUARED_DIFFERENCE = "return (a - b) * (a - b);";
+var squaredDifference3 = binaryKernelFunc2({ opSnippet: SQUARED_DIFFERENCE, packedOpSnippet: SQUARED_DIFFERENCE });
+var squaredDifferenceConfig2 = {
+  kernelName: SquaredDifference,
+  backendName: "webgl",
+  kernelFunc: squaredDifference3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Step.js
+function step3({ inputs, attrs, backend: backend2 }) {
+  const { x } = inputs;
+  const opSnippet = CHECK_NAN_SNIPPET + `
+    return x > 0.0 ? 1.0 : float(${attrs.alpha});
+  `;
+  const program = new UnaryOpProgram(x.shape, opSnippet);
+  return backend2.runWebGLProgram(program, [x], x.dtype);
+}
+var stepConfig2 = {
+  kernelName: Step,
+  backendName: "webgl",
+  kernelFunc: step3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/strided_slice_gpu.js
+var StridedSliceProgram = class {
+  constructor(begin, strides, size) {
+    this.variableNames = ["x"];
+    this.outputShape = size;
+    const rank = size.length;
+    const inputDtype = getCoordsDataType(size.length);
+    const dtype = getCoordsDataType(size.length);
+    let newCoords = "";
+    if (rank === 1) {
+      newCoords = "coords * strides + begin";
+    } else {
+      let outputAxis = 0;
+      newCoords = size.map((_, i) => {
+        outputAxis++;
+        return size.length === 1 ? `coords * strides[${i}] + begin[${i}]` : `coords[${outputAxis - 1}] * strides[${i}] + begin[${i}]`;
+      }).join(",");
+    }
+    this.userCode = `
+      ${inputDtype} begin = ${inputDtype}(${begin});
+      ${inputDtype} strides = ${inputDtype}(${strides});
 
       void main() {
-        ${i} coords = getOutputCoords();
-        setOutput(getX(${a}));
+        ${dtype} coords = getOutputCoords();
+        setOutput(getX(${newCoords}));
       }
-    `}};function ist(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n,{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=st({inputs:{x:o},backend:e,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let A=Le.computeOutShape(b,w,C),$=ri({inputs:{x:o},backend:e,attrs:{begin:b,size:A}});N=st({inputs:{x:$},backend:e,attrs:{shape:d}}),e.disposeIntermediateTensorInfo($)}else if(e.shouldExecuteOnCPU([o])){let $=e.readSync(o.dataId),F=wt(o.shape,o.dtype,$),P=HL(f,F,C,b);N=e.makeTensorInfo(d,o.dtype,P.values)}else{let $=new nI(b,C,f);N=e.runWebGLProgram($,[o],o.dtype)}let _=st({inputs:{x:N},backend:e,attrs:{shape:d}});return e.disposeIntermediateTensorInfo(N),_}var AB={kernelName:Wa,backendName:"webgl",kernelFunc:ist};function ast(r){let{inputs:t,backend:e,attrs:n}=r,{separator:o,nGramWidths:s,leftPad:i,rightPad:a,padWidth:u,preserveShortSequences:l}=n,{data:c,dataSplits:p}=t,m=e.readSync(c.dataId),f=e.readSync(p.dataId),[d,h]=qL(m,f,o,s,i,a,u,l);return[e.makeTensorInfo([d.length],"string",d),e.makeTensorInfo(p.shape,"int32",h)]}var $B={kernelName:Bl,backendName:"webgl",kernelFunc:ast};function lst(r){let{inputs:t,backend:e,attrs:n}=r,{skipEmpty:o}=n,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let a=e.readSync(s.dataId),u=e.readSync(i.dataId)[0],[l,c,p]=KL(a,u,o),m=c.length;return[e.makeTensorInfo([m,2],"int32",l),e.makeTensorInfo([m],"string",c),e.makeTensorInfo([2],"int32",new Int32Array(p))]}var DB={kernelName:Vl,backendName:"webgl",kernelFunc:lst};function ust(r){let{inputs:t,backend:e,attrs:n}=r,{numBuckets:o}=n,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(o<=0)throw new Error("Number of buckets must be at least 1");let i=e.readSync(s.dataId),a=jL(i,o);return e.makeTensorInfo(s.shape,"int32",a)}var RB={kernelName:Gl,backendName:"webgl",kernelFunc:ust};var cst="return tan(x);",pst=Ct({opSnippet:cst}),FB={kernelName:Os,backendName:"webgl",kernelFunc:pst};var mst=`
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StridedSlice.js
+function stridedSlice3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape4({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeIntermediateTensorInfo(sliced);
+  } else {
+    const shouldExecuteOnCPU = backend2.shouldExecuteOnCPU([x]);
+    if (shouldExecuteOnCPU) {
+      const values = backend2.readSync(x.dataId);
+      const xBuf = buffer(x.shape, x.dtype, values);
+      const resultValues = stridedSliceImplCPU(finalShapeSparse, xBuf, $strides, $begin);
+      result = backend2.makeTensorInfo(finalShape, x.dtype, resultValues.values);
+    } else {
+      const program = new StridedSliceProgram($begin, $strides, finalShapeSparse);
+      result = backend2.runWebGLProgram(program, [x], x.dtype);
+    }
+  }
+  const resultReshaped = reshape4({ inputs: { x: result }, backend: backend2, attrs: { shape: finalShape } });
+  backend2.disposeIntermediateTensorInfo(result);
+  return resultReshaped;
+}
+var stridedSliceConfig2 = {
+  kernelName: StridedSlice,
+  backendName: "webgl",
+  kernelFunc: stridedSlice3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringNGrams.js
+function stringNGrams3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const { data, dataSplits } = inputs;
+  const $data = backend2.readSync(data.dataId);
+  const $dataSplits = backend2.readSync(dataSplits.dataId);
+  const [nGrams, nGramsSplits] = stringNGramsImplCPU($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  return [
+    backend2.makeTensorInfo([nGrams.length], "string", nGrams),
+    backend2.makeTensorInfo(dataSplits.shape, "int32", nGramsSplits)
+  ];
+}
+var stringNGramsConfig2 = {
+  kernelName: StringNGrams,
+  backendName: "webgl",
+  kernelFunc: stringNGrams3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringSplit.js
+function stringSplit3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { skipEmpty } = attrs;
+  const { input: input2, delimiter } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (input2.shape.length !== 1) {
+    throw new Error(`Input must be a vector, got shape: ${input2.shape}`);
+  }
+  if (delimiter.shape.length !== 0) {
+    throw new Error(`Delimiter must be a scalar, got shape: ${delimiter.shape}`);
+  }
+  const $input = backend2.readSync(input2.dataId);
+  const $delimiter = backend2.readSync(delimiter.dataId)[0];
+  const [indices, values, shape] = stringSplitImplCPU($input, $delimiter, skipEmpty);
+  const outputSize = values.length;
+  return [
+    backend2.makeTensorInfo([outputSize, 2], "int32", indices),
+    backend2.makeTensorInfo([outputSize], "string", values),
+    backend2.makeTensorInfo([2], "int32", new Int32Array(shape))
+  ];
+}
+var stringSplitConfig2 = {
+  kernelName: StringSplit,
+  backendName: "webgl",
+  kernelFunc: stringSplit3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/StringToHashBucketFast.js
+function stringToHashBucketFast3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { numBuckets } = attrs;
+  const { input: input2 } = inputs;
+  if (input2.dtype !== "string") {
+    throw new Error("Input must be of datatype string");
+  }
+  if (numBuckets <= 0) {
+    throw new Error(`Number of buckets must be at least 1`);
+  }
+  const $input = backend2.readSync(input2.dataId);
+  const output = stringToHashBucketFastImplCPU($input, numBuckets);
+  return backend2.makeTensorInfo(input2.shape, "int32", output);
+}
+var stringToHashBucketFastConfig2 = {
+  kernelName: StringToHashBucketFast,
+  backendName: "webgl",
+  kernelFunc: stringToHashBucketFast3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tan.js
+var TAN = `return tan(x);`;
+var tan3 = unaryKernelFunc2({ opSnippet: TAN });
+var tanConfig2 = {
+  kernelName: Tan,
+  backendName: "webgl",
+  kernelFunc: tan3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tanh.js
+var TANH = `
   float e2x = exp(-2.0 * abs(x));
   return sign(x) * (1.0 - e2x) / (1.0 + e2x);
-`,fst=Ct({opSnippet:mst}),OB={kernelName:Ps,backendName:"webgl",kernelFunc:fst};var oI=class{constructor(t,e){this.variableNames=["A"];let n=new Array(t.length);for(let i=0;i<n.length;i++)n[i]=t[i]*e[i];this.outputShape=n,this.rank=n.length;let o=zt(this.rank),s=dst(t);this.userCode=`
+`;
+var tanh4 = unaryKernelFunc2({ opSnippet: TANH });
+var tanhConfig2 = {
+  kernelName: Tanh,
+  backendName: "webgl",
+  kernelFunc: tanh4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/tile_gpu.js
+var TileProgram = class {
+  constructor(aShape, reps) {
+    this.variableNames = ["A"];
+    const outputShape = new Array(aShape.length);
+    for (let i = 0; i < outputShape.length; i++) {
+      outputShape[i] = aShape[i] * reps[i];
+    }
+    this.outputShape = outputShape;
+    this.rank = outputShape.length;
+    const dtype = getCoordsDataType(this.rank);
+    const sourceCoords = getSourceCoords3(aShape);
+    this.userCode = `
       void main() {
-        ${o} resRC = getOutputCoords();
-        setOutput(getA(${s}));
+        ${dtype} resRC = getOutputCoords();
+        setOutput(getA(${sourceCoords}));
       }
-    `}};function dst(r){let t=r.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${r[0]})`;let e=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],n=[];for(let o=0;o<r.length;o++)n.push(`imod(${e[o]}, ${r[o]})`);return n.join()}function Ek(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{reps:s}=n;if(o.dtype==="string"||o.shape.length>5){let u=e.readSync(o.dataId),l=o.dtype==="string"?u.map(m=>y.decodeString(m)):u,c=wt(o.shape,o.dtype,l),p=YL(c,s);return e.makeTensorInfo(p.shape,p.dtype,p.values)}let i=new oI(o.shape,s);return e.runWebGLProgram(i,[o],o.dtype)}var PB={kernelName:Jn,backendName:"webgl",kernelFunc:Ek};var sI=class{constructor(t){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=t,this.userCode=`
+    `;
+  }
+};
+function getSourceCoords3(aShape) {
+  const rank = aShape.length;
+  if (rank > 5) {
+    throw Error(`Tile for rank ${rank} is not yet supported`);
+  }
+  if (rank === 1) {
+    return `imod(resRC, ${aShape[0]})`;
+  }
+  const currentCoords = ["resRC.x", "resRC.y", "resRC.z", "resRC.w", "resRC.u"];
+  const sourceCoords = [];
+  for (let i = 0; i < aShape.length; i++) {
+    sourceCoords.push(`imod(${currentCoords[i]}, ${aShape[i]})`);
+  }
+  return sourceCoords.join();
+}
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Tile.js
+function tile4(params) {
+  const { inputs, backend: backend2, attrs } = params;
+  const { x } = inputs;
+  const { reps } = attrs;
+  if (x.dtype === "string" || x.shape.length > 5) {
+    const data = backend2.readSync(x.dataId);
+    const value = x.dtype === "string" ? data.map((d) => util_exports.decodeString(d)) : data;
+    const buf = buffer(x.shape, x.dtype, value);
+    const outBuf = tileImplCPU(buf, reps);
+    return backend2.makeTensorInfo(outBuf.shape, outBuf.dtype, outBuf.values);
+  }
+  const program = new TileProgram(x.shape, reps);
+  const output = backend2.runWebGLProgram(program, [x], x.dtype);
+  return output;
+}
+var tileConfig2 = {
+  kernelName: Tile,
+  backendName: "webgl",
+  kernelFunc: tile4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/top_k_gpu.js
+var SwapProgram = class {
+  constructor(shape) {
+    this.variableNames = ["x", "indices"];
+    this.customUniforms = [
+      { name: "n", type: "int" },
+      { name: "firstPass", type: "int" },
+      { name: "negativeInf", type: "float" },
+      { name: "dir", type: "int" },
+      { name: "inc", type: "int" }
+    ];
+    this.outputShape = shape;
+    this.userCode = `
        void main() {
          ivec2 coords = getOutputCoords();
          int batch = coords[0];
@@ -4624,7 +63048,19 @@ return a / b;`,Xnt=`
             setOutput(float(i1));
          }
        }
-     `}},iI=class{constructor(t){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=t,this.userCode=`
+     `;
+  }
+};
+var MergeProgram = class {
+  constructor(shape) {
+    this.variableNames = ["x", "indices"];
+    this.customUniforms = [
+      { name: "n", type: "int" },
+      { name: "firstPass", type: "int" },
+      { name: "k", type: "int" }
+    ];
+    this.outputShape = shape;
+    this.userCode = `
     void main() {
          // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...
          ivec2 coords = getOutputCoords();
@@ -4658,10 +63094,143 @@ return a / b;`,Xnt=`
 
          setOutput(x0 >= x1 ? float(i0) : float(i1));
        }
-     `}};function Kc(r,t){t!==null&&r.disposeIntermediateTensorInfo(t)}function LB(r){let t=1;for(;t<r;)t*=2;return t}function hst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{k:s,sorted:i}=n,a=z().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),u=z().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),l=o.shape,c=l[l.length-1];if(e.shouldExecuteOnCPU([o])||c<a||s>u){let P=e.readSync(o.dataId),[V,G]=ZL(P,l,o.dtype,s,i);return[e.makeTensorInfo(V.shape,V.dtype,V.values),e.makeTensorInfo(G.shape,G.dtype,G.values)]}if(s===0)return l[l.length-1]=0,[e.makeTensorInfo(l,o.dtype,[]),e.makeTensorInfo(l,"int32",[])];if(c===1)return[o,Cl({attrs:{shape:l,dtype:"int32",value:0},backend:e})];let p=e.texData.get(o.dataId),m=p!==null&&p.isPacked,f=m?e.unpackTensor(o):o,h=y.sizeFromShape(l)/c,g=st({inputs:{x:f},attrs:{shape:[h,c]},backend:e});m&&Kc(e,f);let x=LB(s),b=LB(c),w=null,C=()=>w===null?[g,g]:[g,w],N=(P,V,G)=>{let W=C(),q=new sI(G),j=[[c],[w===null?1:0],[Number.NEGATIVE_INFINITY],[P],[V]],Y=w;w=e.runWebGLProgram(q,W,"int32",j),Kc(e,Y)};for(let P=1;P<x;P*=2){let V=P*2;for(let G=P;G>=1;G/=2)N(V,G,[h,b])}for(let P=b;P>x;P/=2){let V=C(),G=new iI([h,P/2]),q=[[c],[w===null?1:0],[x]],H=w;w=e.runWebGLProgram(G,V,"int32",q),Kc(e,H);let j=x/2,Y=j*2;for(let Z=j;Z>=1;Z/=2)N(Y,Z,w.shape)}let _=w;w=ri({inputs:{x:w},backend:e,attrs:{begin:0,size:[h,s]}}),Kc(e,_);let A=Ck({inputs:{x:g,indices:w},backend:e,attrs:{axis:1,batchDims:1}});Kc(e,g);let $=l.slice(0,-1);$.push(s),_=w,w=st({inputs:{x:w},attrs:{shape:$},backend:e}),Kc(e,_);let F=A;return A=st({inputs:{x:A},attrs:{shape:$},backend:e}),Kc(e,F),[A,w]}var MB={kernelName:Ua,backendName:"webgl",kernelFunc:hst};var aI=class{constructor(t,e,n,o,s,i){this.variableNames=["Image","Transforms"],this.outputShape=i;let a=n==="nearest"?1:2,u;switch(o){case"constant":u=1;break;case"reflect":u=2;break;case"wrap":u=3;break;case"nearest":u=4;break;default:u=1;break}this.userCode=`
+     `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/TopK.js
+function disposeIntermediateTensorInfoOrNull(backend2, tensorInfo) {
+  if (tensorInfo !== null) {
+    backend2.disposeIntermediateTensorInfo(tensorInfo);
+  }
+}
+function roundUpToPow2(num) {
+  let pow22 = 1;
+  while (pow22 < num) {
+    pow22 *= 2;
+  }
+  return pow22;
+}
+function topK2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  const TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD = env().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD");
+  const TOPK_K_CPU_HANDOFF_THRESHOLD = env().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD");
+  const xShape = x.shape;
+  const lastDim = xShape[xShape.length - 1];
+  if (backend2.shouldExecuteOnCPU([x]) || lastDim < TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD || k > TOPK_K_CPU_HANDOFF_THRESHOLD) {
+    const xVals = backend2.readSync(x.dataId);
+    const [allTopKVals, allTopKIndices] = topKImplCPU(xVals, xShape, x.dtype, k, sorted);
+    return [
+      backend2.makeTensorInfo(allTopKVals.shape, allTopKVals.dtype, allTopKVals.values),
+      backend2.makeTensorInfo(allTopKIndices.shape, allTopKIndices.dtype, allTopKIndices.values)
+    ];
+  }
+  if (k === 0) {
+    xShape[xShape.length - 1] = 0;
+    return [
+      backend2.makeTensorInfo(xShape, x.dtype, []),
+      backend2.makeTensorInfo(xShape, "int32", [])
+    ];
+  }
+  if (lastDim === 1) {
+    return [
+      x,
+      fill3({ attrs: { shape: xShape, dtype: "int32", value: 0 }, backend: backend2 })
+    ];
+  }
+  const xtexData = backend2.texData.get(x.dataId);
+  const xIsPacked = xtexData !== null && xtexData.isPacked;
+  const xUnPacked = xIsPacked ? backend2.unpackTensor(x) : x;
+  const xSize = util_exports.sizeFromShape(xShape);
+  const batch = xSize / lastDim;
+  const x2D = reshape4({ inputs: { x: xUnPacked }, attrs: { shape: [batch, lastDim] }, backend: backend2 });
+  if (xIsPacked) {
+    disposeIntermediateTensorInfoOrNull(backend2, xUnPacked);
+  }
+  const kPow2 = roundUpToPow2(k);
+  const lastDimPow2 = roundUpToPow2(lastDim);
+  let indices = null;
+  const getInputs = () => indices === null ? [x2D, x2D] : [x2D, indices];
+  const runSwap = (dir, inc, shape) => {
+    const inputs2 = getInputs();
+    const program = new SwapProgram(shape);
+    const fistPass = indices === null ? 1 : 0;
+    const customValues = [[lastDim], [fistPass], [Number.NEGATIVE_INFINITY], [dir], [inc]];
+    const prevIndices2 = indices;
+    indices = backend2.runWebGLProgram(program, inputs2, "int32", customValues);
+    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);
+  };
+  for (let len = 1; len < kPow2; len *= 2) {
+    const dir = len * 2;
+    for (let inc = len; inc >= 1; inc /= 2) {
+      runSwap(dir, inc, [batch, lastDimPow2]);
+    }
+  }
+  for (let indicesSize = lastDimPow2; indicesSize > kPow2; indicesSize /= 2) {
+    const inputs2 = getInputs();
+    const mergeProgram = new MergeProgram([batch, indicesSize / 2]);
+    const firstPass = indices === null ? 1 : 0;
+    const customValues = [[lastDim], [firstPass], [kPow2]];
+    const prevIndices2 = indices;
+    indices = backend2.runWebGLProgram(mergeProgram, inputs2, "int32", customValues);
+    disposeIntermediateTensorInfoOrNull(backend2, prevIndices2);
+    const len = kPow2 / 2;
+    const dir = len * 2;
+    for (let inc = len; inc >= 1; inc /= 2) {
+      runSwap(dir, inc, indices.shape);
+    }
+  }
+  let prevIndices = indices;
+  indices = slice3({ inputs: { x: indices }, backend: backend2, attrs: { begin: 0, size: [batch, k] } });
+  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);
+  let values = gatherV22({ inputs: { x: x2D, indices }, backend: backend2, attrs: { axis: 1, batchDims: 1 } });
+  disposeIntermediateTensorInfoOrNull(backend2, x2D);
+  const newShape = xShape.slice(0, -1);
+  newShape.push(k);
+  prevIndices = indices;
+  indices = reshape4({ inputs: { x: indices }, attrs: { shape: newShape }, backend: backend2 });
+  disposeIntermediateTensorInfoOrNull(backend2, prevIndices);
+  const prevValues = values;
+  values = reshape4({ inputs: { x: values }, attrs: { shape: newShape }, backend: backend2 });
+  disposeIntermediateTensorInfoOrNull(backend2, prevValues);
+  return [values, indices];
+}
+var topKConfig2 = {
+  kernelName: TopK,
+  backendName: "webgl",
+  kernelFunc: topK2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/transform_gpu.js
+var TransformProgram = class {
+  constructor(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape) {
+    this.variableNames = ["Image", "Transforms"];
+    this.outputShape = outShape;
+    const interpolationModeId = interpolation === "nearest" ? 1 : 2;
+    let fillModeId;
+    switch (fillMode) {
+      case "constant":
+        fillModeId = 1;
+        break;
+      case "reflect":
+        fillModeId = 2;
+        break;
+      case "wrap":
+        fillModeId = 3;
+        break;
+      case "nearest":
+        fillModeId = 4;
+        break;
+      default:
+        fillModeId = 1;
+        break;
+    }
+    this.userCode = `
             float mapCoord(float outCoord, float len) {
               float inCoord = outCoord;
-              if(${u} == 2) {
+              if(${fillModeId} == 2) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4685,7 +63254,7 @@ return a / b;`,Xnt=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${u} == 3) {
+              } else if (${fillModeId} == 3) {
                 if (inCoord < 0.0) {
                   if (len <= 1.0) {
                     inCoord = 0.0;
@@ -4702,7 +63271,7 @@ return a / b;`,Xnt=`
                   }
                 }
                 return clamp(inCoord, 0.0, len - 1.0);
-              } else if (${u} == 4) {
+              } else if (${fillModeId} == 4) {
                 return clamp(outCoord, 0.0, len - 1.0);
               } else {
                 return outCoord;
@@ -4712,10 +63281,10 @@ return a / b;`,Xnt=`
             float readWithFillValue(int batch, int coordY, int coordX,
               int channel) {
               float outputValue;
-              if (0 <= coordY && coordY < ${t} && 0 <= coordX && coordX < ${e}) {
+              if (0 <= coordY && coordY < ${imageHeight} && 0 <= coordX && coordX < ${imageWidth}) {
                   outputValue = getImage(batch, coordY, coordX, channel);
               } else {
-                outputValue = float(${s});
+                outputValue = float(${fillValue});
               }
               return outputValue;
             }
@@ -4739,14 +63308,14 @@ return a / b;`,Xnt=`
               float c2 = getTransforms(batch, 7);
               float projection = c1 * xf + c2 * yf + 1.0;
               if (projection == 0.0) {
-                outputValue = float(${s});
+                outputValue = float(${fillValue});
               } else {
                 float inX = (a1 * xf + a2 * yf + a3) / projection;
                 float inY = (b1 * xf + b2 * yf + b3) / projection;
-                float mapX = mapCoord(inX, float(${e}));
-                float mapY = mapCoord(inY, float(${t}));
+                float mapX = mapCoord(inX, float(${imageWidth}));
+                float mapY = mapCoord(inY, float(${imageHeight}));
 
-                if (${a} == 1) {
+                if (${interpolationModeId} == 1) {
                   int coordY = int(round(mapY));
                   int coordX = int(round(mapX));
                   outputValue = readWithFillValue(batch, coordY, coordX,
@@ -4770,26 +63339,134 @@ return a / b;`,Xnt=`
               }
               setOutput(outputValue);
             }
-        `}};function gst(r){let{inputs:t,backend:e,attrs:n}=r,{image:o,transforms:s}=t,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=n,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=new aI(p,m,i,a,u,g);return e.runWebGLProgram(x,[o,s],"float32")}var zB={kernelName:Ha,backendName:"webgl",kernelFunc:gst};function xst(r){let{inputs:t,attrs:e,backend:n}=r,{axis:o}=e,{x:s}=t;Qs(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let i=n.readSync(s.dataId),{outputValues:a,outputShape:u,indices:l}=JL(i,o,s.shape,s.dtype);return[n.makeTensorInfo(u,s.dtype,a),n.makeTensorInfo([l.length],"int32",l)]}var BB={kernelName:Mp,backendName:"webgl",kernelFunc:xst};function yst(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o,a=i.shape.length,u=o.shape[s],l=new Array(a-1),c=0;for(let h=0;h<a;h++)h!==s&&(l[c++]=i.shape[h]);let p=[],m=new Array(a).fill(0),f=i.shape.slice();f[s]=1;let d=new Array(u);for(let h=0;h<d.length;h++){m[s]=h;let g=ri({inputs:{x:i},backend:e,attrs:{begin:m,size:f}}),x=st({inputs:{x:g},backend:e,attrs:{shape:l}});d[h]=x,p.push(g)}return p.forEach(h=>e.disposeIntermediateTensorInfo(h)),d}var VB={kernelName:bi,backendName:"webgl",kernelFunc:yst};var lI=class{constructor(t,e){this.variableNames=["x","segmentIds"];let n=t.windowSize,o=t.batchSize,s=t.inSize,i=t.numSegments,a=i*Math.ceil(s/n);this.outputShape=[o,a];let u="0.0",l="sumValue",c=Math.floor(n/4)*4,p=n%4,m=`
+        `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Transform.js
+function transform3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [
+    batch,
+    outHeight,
+    outWidth,
+    numChannels
+  ];
+  const program = new TransformProgram(imageHeight, imageWidth, interpolation, fillMode, fillValue, outShape);
+  return backend2.runWebGLProgram(program, [image2, transforms], "float32");
+}
+var transformConfig2 = {
+  kernelName: Transform,
+  backendName: "webgl",
+  kernelFunc: transform3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Unique.js
+function unique4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  assertNotComplex2(x, "unique");
+  console.warn("WARNING: ", "UI might be locked temporarily as data is being downloaded");
+  const values = backend2.readSync(x.dataId);
+  const { outputValues, outputShape, indices } = uniqueImplCPU(values, axis, x.shape, x.dtype);
+  return [
+    backend2.makeTensorInfo(outputShape, x.dtype, outputValues),
+    backend2.makeTensorInfo([indices.length], "int32", indices)
+  ];
+}
+var uniqueConfig2 = {
+  kernelName: Unique,
+  backendName: "webgl",
+  kernelFunc: unique4
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/Unpack.js
+function unpack2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const x = value;
+  const xRank = x.shape.length;
+  const num = value.shape[axis];
+  const outShape = new Array(xRank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < xRank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = x.shape[i];
+    }
+  }
+  const toDispose = [];
+  const begin = new Array(xRank).fill(0);
+  const size = x.shape.slice();
+  size[axis] = 1;
+  const res = new Array(num);
+  for (let i = 0; i < res.length; i++) {
+    begin[axis] = i;
+    const sliced = slice3({ inputs: { x }, backend: backend2, attrs: { begin, size } });
+    const reshaped = reshape4({ inputs: { x: sliced }, backend: backend2, attrs: { shape: outShape } });
+    res[i] = reshaped;
+    toDispose.push(sliced);
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return res;
+}
+var unpackConfig2 = {
+  kernelName: Unpack,
+  backendName: "webgl",
+  kernelFunc: unpack2
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/segment_gpu.js
+var SegmentOpProgram = class {
+  constructor(segOpInfo, segOpType) {
+    this.variableNames = ["x", "segmentIds"];
+    const windowSize = segOpInfo.windowSize;
+    const batchSize = segOpInfo.batchSize;
+    const inSize = segOpInfo.inSize;
+    const numSegments = segOpInfo.numSegments;
+    const outSize = numSegments * Math.ceil(inSize / windowSize);
+    this.outputShape = [batchSize, outSize];
+    const initializationValue = "0.0";
+    const returnValue = `sumValue`;
+    const windowSizeNearestVec4 = Math.floor(windowSize / 4) * 4;
+    const windowSizeVec4Remainder = windowSize % 4;
+    const updateSnippet = `
         sumValue += dot(values, segFilter);
-    `,f="";s%n>0&&(f=`
-        if (inIdx < 0 || inIdx >= ${s}) {
+    `;
+    let checkValueOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkValueOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return initializationValue;
         }
-      `);let d="";s%n>0&&(d=`
-        if (inIdx < 0 || inIdx >= ${s}) {
+      `;
+    }
+    let checkSegmentIdOutOfBounds = "";
+    if (inSize % windowSize > 0) {
+      checkSegmentIdOutOfBounds = `
+        if (inIdx < 0 || inIdx >= ${inSize}) {
           return -1.0;
         }
-      `),this.userCode=`
-      const float initializationValue = ${u};
+      `;
+    }
+    this.userCode = `
+      const float initializationValue = ${initializationValue};
 
       float getValue(int batch, int inIdx) {
-        ${f}
+        ${checkValueOutOfBounds}
         return getX(batch, inIdx);
       }
 
       float getSegmentIdAtIndex(int inIdx) {
-        ${d}
+        ${checkSegmentIdOutOfBounds}
         return getSegmentIds(inIdx);
       }
 
@@ -4798,12 +63475,12 @@ return a / b;`,Xnt=`
         int batch = coords[0];
         int outIdx = coords[1];
         int inOffset = int(floor(float(outIdx) / float(
-          ${i})) * float(${n}));
-        int currentSeg = int(mod(float(outIdx), float(${i})));
+          ${numSegments})) * float(${windowSize}));
+        int currentSeg = int(mod(float(outIdx), float(${numSegments})));
 
         float sumValue = 0.0;
 
-        for (int i = 0; i < ${c}; i += 4) {
+        for (int i = 0; i < ${windowSizeNearestVec4}; i += 4) {
           int inIdx = inOffset + i;
           vec4 values = vec4(
             getValue(batch, inIdx),
@@ -4819,11 +63496,11 @@ return a / b;`,Xnt=`
             int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0
           );
 
-          ${m}
+          ${updateSnippet}
         }
 
-        int inIdx = inOffset + ${c};
-        if (${p===1}) {
+        int inIdx = inOffset + ${windowSizeNearestVec4};
+        if (${windowSizeVec4Remainder === 1}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             initializationValue,
@@ -4840,8 +63517,8 @@ return a / b;`,Xnt=`
             0
           );
 
-          ${m}
-        } else if (${p===2}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 2}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4856,8 +63533,8 @@ return a / b;`,Xnt=`
               0
           );
 
-          ${m}
-        } else if (${p===3}) {
+          ${updateSnippet}
+        } else if (${windowSizeVec4Remainder === 3}) {
           vec4 values = vec4(
             getValue(batch, inIdx),
             getValue(batch, inIdx + 1),
@@ -4872,10 +63549,4536 @@ return a / b;`,Xnt=`
             0
           );
 
-          ${m}
+          ${updateSnippet}
         }
-        setOutput(${l});
+        setOutput(${returnValue});
       }
-    `}};function bst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o,segmentIds:s}=t,{numSegments:i}=n,a=o.shape.length,u=[],l=0,c=v.getAxesPermutation([l],a),p=o;c!=null&&(p=Oe({inputs:{x:o},backend:e,attrs:{perm:c}}),u.push(p),l=v.getInnerMostAxes(1,a)[0]);let m=v.segment_util.computeOutShape(p.shape,l,i),f=y.sizeFromShape([p.shape[l]]),d=st({inputs:{x:p},backend:e,attrs:{shape:[-1,f]}});u.push(d);let h=Wu(o.dtype),g=(C,N,_,A,$)=>{let F=C.shape[0],P=C.shape[1],V=v.segment_util.segOpComputeOptimalWindowSize(P,$),G={windowSize:V,inSize:P,batchSize:F,numSegments:$},W=new lI(G,N),q=e.compileAndRun(W,[C,_],A);if(u.push(q),q.shape[1]===$)return q;let H=kk({backend:e,attrs:{start:0,stop:$,step:1,dtype:"float32"}}),j=Ek({inputs:{x:H},backend:e,attrs:{reps:[P/V]}});return u.push(H),u.push(j),g(q,N,j,A,$)},x=g(d,"unsortedSegmentSum",s,h,i),b=st({inputs:{x},backend:e,attrs:{shape:m}}),w=b;if(c!=null){u.push(b);let C=v.getUndoAxesPermutation(c);w=Oe({inputs:{x:w},backend:e,attrs:{perm:C}})}return u.forEach(C=>e.disposeIntermediateTensorInfo(C)),w}var GB={kernelName:Wl,backendName:"webgl",kernelFunc:bst};var wst=[kM,_M,AM,$M,RM,FM,OM,PM,zM,BM,VM,GM,WM,UM,HM,qM,KM,jM,XM,YM,ZM,QM,tz,ez,sz,az,lz,xM,cz,mz,fz,dz,hz,gz,xz,yz,bz,wz,Cz,vz,Nz,Tz,kz,Ez,_z,Az,$z,Dz,Rz,Fz,Oz,Pz,Lz,Mz,zz,Vz,Gz,Wz,Uz,qz,Kz,jz,Xz,Yz,Zz,Jz,Qz,t3,gM,e3,pz,r3,n3,o3,yM,s3,i3,a3,l3,u3,c3,p3,m3,f3,d3,g3,x3,y3,b3,w3,C3,S3,N3,T3,k3,E3,_3,F3,CM,O3,P3,L3,M3,rz,z3,G3,W3,U3,H3,bM,q3,K3,j3,X3,Y3,nz,A3,Z3,J3,Q3,SM,tB,eB,rB,nB,oB,sB,iB,aB,lB,uB,cB,pB,mB,fB,dB,hB,JM,R3,gB,xB,yB,bB,wB,CB,IB,SB,NB,TB,EB,_B,AB,$B,DB,RB,D3,NM,FB,OB,PB,MB,zB,TM,BB,VB,GB,B3];for(let r of wst)Lu(r);var qt;(function(r){r[r.float32=0]="float32",r[r.int32=1]="int32",r[r.bool=2]="bool",r[r.string=3]="string",r[r.complex64=4]="complex64"})(qt||(qt={}));var Du;(function(r){r[r.linear=0]="linear",r[r.relu=1]="relu",r[r.relu6=2]="relu6",r[r.prelu=3]="prelu",r[r.leakyrelu=4]="leakyrelu",r[r.sigmoid=5]="sigmoid",r[r.elu=6]="elu"})(Du||(Du={}));var WB;function Cst(r){WB=r.wasm.cwrap(Ci,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])}function Ist(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s,bias:i,preluActivationWeights:a}=t;if(o.dtype!=="float32"||s.dtype!=="float32")throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:u,transposeB:l,activation:c,leakyreluAlpha:p}=n,m=e.dataIdMap.get(o.dataId).id,f=e.dataIdMap.get(s.dataId).id,d=0;if(i!=null){let $=e.dataIdMap.get(i.dataId);if($.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${$.shape.length}.`);d=$.id}let h=a==null?0:e.dataIdMap.get(a.dataId).id,g=Du[c];if(g==null)throw new Error(`${c} activation not yet supported for FusedConv2D in the wasm backend.`);let x=u?o.shape[2]:o.shape[1],b=l?s.shape[1]:s.shape[2],w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)),C=e.makeOutput([...w,x,b],o.dtype),N=e.dataIdMap.get(C.dataId).id,_=new Uint8Array(new Int32Array(o.shape).buffer),A=new Uint8Array(new Int32Array(s.shape).buffer);return WB(m,_,o.shape.length,f,A,s.shape.length,u,l,g,d,h,p||0,N),C}var UB={kernelName:Ci,backendName:"wasm",setupFunc:Cst,kernelFunc:Ist};function se(r,t){let e;function n(s){e=s.wasm.cwrap(r,null,["number","number","number"])}function o(s){let{backend:i,inputs:{x:a}}=s,u=i.dataIdMap.get(a.dataId).id,l=i.makeOutput(a.shape,t||a.dtype),c=i.dataIdMap.get(l.dataId).id;return y.sizeFromShape(l.shape)===0||e(u,qt[a.dtype],c),l}return{kernelName:r,backendName:"wasm",setupFunc:n,kernelFunc:o}}var HB=se(ii);function ue(r,t,e){let n;function o(i){n=i.wasm.cwrap(r,null,["number","array","number","number","array","number","number","number"])}function s(i){let{backend:a,inputs:u}=i,{a:l,b:c}=u,p=a.dataIdMap.get(l.dataId).id,m=a.dataIdMap.get(c.dataId).id,f=e!=null?e:l.dtype,d=v.assertAndGetBroadcastShape(l.shape,c.shape),h=a.makeOutput(d,f);if(y.sizeFromShape(d)===0)return h;let g=new Uint8Array(new Int32Array(l.shape).buffer),x=new Uint8Array(new Int32Array(c.shape).buffer),b=a.dataIdMap.get(h.dataId).id;return(()=>n(p,g,l.shape.length,m,x,c.shape.length,qt[l.dtype],b))(),h}return{kernelName:r,backendName:"wasm",setupFunc:o,kernelFunc:s}}var Sst=!0,qB=ue(Zn,Sst);var KB;function vst(r){KB=r.wasm.cwrap(Go,null,["array","number","number","number"])}function Nst(r){let{inputs:t,backend:e}=r,n=e.makeOutput(t[0].shape,t[0].dtype);if(y.sizeFromShape(n.shape)===0)return n;let o=t.map(a=>e.dataIdMap.get(a.dataId).id),s=new Uint8Array(new Int32Array(o).buffer),i=e.dataIdMap.get(n.dataId).id;return KB(s,o.length,qt[n.dtype],i),n}var jB={kernelName:Go,backendName:"wasm",setupFunc:vst,kernelFunc:Nst};function jc(r){let{inputs:{x:t},backend:e}=r;if(t.dtype==="string")return ur(e.readSync(t.dataId),t.shape,t.dtype);let n=e.makeOutput(t.shape,t.dtype),o=e.typedArrayFromHeap(t);return e.typedArrayFromHeap(n).set(o),n}var XB={kernelName:co,backendName:"wasm",kernelFunc:jc};var YB;function Tst(r){YB=r.wasm.cwrap(Qn,null,["number","array","number","number","number","array","number"])}function ao(r){let{inputs:t,backend:e,attrs:n}=r,[o,s]=Est(t.x.shape,n.perm),i=!0;for(let d=0;d<s.length;d++)s[d]!==d&&(i=!1);let a=kst(t.x.shape,n.perm),u={dataId:t.x.dataId,shape:o,dtype:t.x.dtype};if(i){let d=jc({inputs:t,backend:e});return d.shape=a,d}let l=e.makeOutput(a,u.dtype),c=e.dataIdMap.get(u.dataId).id,p=e.dataIdMap.get(l.dataId).id,m=new Uint8Array(new Int32Array(s).buffer),f=new Uint8Array(new Int32Array(u.shape).buffer);return YB(c,f,u.shape.length,qt[u.dtype],p,m,s.length),l}function kst(r,t){let e=new Array(r.length);for(let n=0;n<e.length;n++)e[n]=r[t[n]];return e}function Est(r,t){let e=[],n=[];for(let o=0;o<r.length;++o)r[o]!==1&&e.push(r[o]),r[t[o]]!==1&&n.push(t[o]);for(let o=0;o<n.length;++o){let s=-1;for(let i=0;i<n.length;++i)n[i]>=o&&(s===-1||n[s]>n[i])&&(s=i);n[s]=o}return[e,n]}var ZB={kernelName:Qn,backendName:"wasm",kernelFunc:ao,setupFunc:Tst};function bn(r,t,e){let n=r.shape,o=r.shape.length,s=y.parseAxisParam(t,n),i=s,a=v.getAxesPermutation(i,o),u=null,l=!1;if(a!=null){let c=new Array(o);for(let f=0;f<c.length;f++)c[f]=n[a[f]];i=v.getInnerMostAxes(i.length,o),u=ao({inputs:{x:r},attrs:{perm:a},backend:e});let p=e.dataIdMap.get(r.dataId).id;e.dataIdMap.get(u.dataId).id!==p&&(l=!0)}return{transposed:u,originalAxes:s,axes:i,inputWasTransposed:l}}var JB;function _st(r){JB=r.wasm.cwrap(ia,null,["number, number, number"])}function Ast(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims("all",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;JB(u,x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var QB={kernelName:ia,backendName:"wasm",setupFunc:_st,kernelFunc:Ast};var tV;function $st(r){tV=r.wasm.cwrap(aa,null,["number, number, number"])}function Dst(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims("any",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;tV(u,x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var eV={kernelName:aa,backendName:"wasm",setupFunc:$st,kernelFunc:Dst};var rV;function Rst(r){rV=r.wasm.cwrap(Wo,null,["number","number","number","number","number"])}function Fst(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o}=n,{x:s}=e,i=t.dataIdMap.get(s.dataId).id,a=i,u=s,{transposed:l,axes:c,inputWasTransposed:p}=bn(s,o,t);if(p){let x=t.dataIdMap.get(l.dataId).id;x!==i&&(u=l,a=x)}let m=u.shape.slice(0,-1),f=t.makeOutput(m,"int32"),d=t.dataIdMap.get(f.dataId).id,h=y.sizeFromShape(f.shape),g=u.shape[c[0]];return rV(a,qt[u.dtype],h,g,d),p&&t.disposeData(l.dataId),f}var nV={kernelName:Wo,backendName:"wasm",kernelFunc:Fst,setupFunc:Rst};var oV;function Ost(r){oV=r.wasm.cwrap(Uo,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Pst(r){let{inputs:t,attrs:e,backend:n}=r,o=t.x,s=n.dataIdMap.get(o.dataId).id,{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=e,c=v.computePool2DInfo(o.shape,i,a,1,u,l),p=c.filterHeight,m=c.filterWidth,f=c.padInfo.top,d=c.padInfo.right,h=c.padInfo.bottom,g=c.padInfo.left,x=c.strideHeight,b=c.strideWidth,w=c.inChannels;if(c.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);if(c.dilationWidth!==1||c.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${c.dilationHeight}, ${c.dilationWidth}].`);let C=n.makeOutput(c.outShape,"float32"),N=n.dataIdMap.get(C.dataId).id;return oV(s,o.shape[0],o.shape[1],o.shape[2],p,m,f,d,h,g,x,b,w,N),C}var sV={kernelName:Uo,backendName:"wasm",setupFunc:Ost,kernelFunc:Pst};function ar(r){let{inputs:t,attrs:e}=r,{x:n}=t,{shape:o}=e,s=y.sizeFromShape(n.shape),i=y.inferFromImplicitShape(o,s);return y.assert(s===y.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${n.shape}. New shape and old shape must have the same number of elements.`),r.backend.incRef(n.dataId),{dataId:n.dataId,shape:i,dtype:n.dtype}}var iV={kernelName:di,backendName:"wasm",kernelFunc:ar};var aV;function Lst(r){aV=r.wasm.cwrap(Ho,null,["number","array","number","number","array","number","number","number","number"])}function Mst(r){let{inputs:t,backend:e,attrs:n}=r,{a:o,b:s}=t,{transposeA:i,transposeB:a}=n;if(o.dtype!=="float32"||s.dtype!=="float32")throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let u=o.shape.length,l=s.shape.length,c=i?o.shape[u-2]:o.shape[u-1],p=a?s.shape[l-1]:s.shape[l-2],m=i?o.shape[u-1]:o.shape[u-2],f=a?s.shape[l-2]:s.shape[l-1],d=o.shape.slice(0,-2),h=s.shape.slice(0,-2),g=y.sizeFromShape(d),x=y.sizeFromShape(h),w=Vr.assertAndGetBroadcastShape(o.shape.slice(0,-2),s.shape.slice(0,-2)).concat([m,f]);y.assert(c===p,()=>`Error in matMul: inner shapes (${c}) and (${p}) of Tensors with shapes ${o.shape} and ${s.shape} and transposeA=${i} and transposeB=${a} must match.`);let C=i?[g,c,m]:[g,m,c],N=a?[x,f,p]:[x,p,f],_=ar({inputs:{x:o},backend:e,attrs:{shape:C}}),A=ar({inputs:{x:s},backend:e,attrs:{shape:N}}),$=e.dataIdMap.get(_.dataId).id,F=e.dataIdMap.get(A.dataId).id,P=i?_.shape[2]:_.shape[1],V=a?A.shape[1]:A.shape[2],G=Math.max(g,x),W=e.makeOutput([G,P,V],_.dtype),q=e.dataIdMap.get(W.dataId).id,H=new Uint8Array(new Int32Array(_.shape).buffer),j=new Uint8Array(new Int32Array(A.shape).buffer);return aV($,H,_.shape.length,F,j,A.shape.length,i,a,q),e.disposeData(_.dataId),e.disposeData(A.dataId),W.shape=w,W}var lV={kernelName:Ho,backendName:"wasm",setupFunc:Lst,kernelFunc:Mst};function Lo(r){let{inputs:{x:t},attrs:{begin:e,size:n},backend:o}=r,[s,i]=Le.parseSliceParams(t,e,n),a=Le.isSliceContinous(t.shape,s,i),u=o.readSync(t.dataId),l=o.makeOutput(i,t.dtype),c=y.computeStrides(t.shape),p=o.dataIdMap.get(l.dataId);if(a){let d=Le.computeFlatOffset(s,c);return t.dtype==="string"?p.stringBytes=u.slice(d,d+y.sizeFromShape(i)):o.typedArrayFromHeap(l).set(u.subarray(d,d+y.sizeFromShape(i))),l}if(t.dtype==="string"){let d=$c(u,s,i,t.shape,t.dtype);return p.stringBytes=d,l}let m=o.typedArrayFromHeap(l),f=t.shape.length;if(f===2)zst(u,c[0],m,s,i);else if(f===3)Bst(u,c[0],c[1],m,s,i);else if(f===4)Vst(u,c[0],c[1],c[2],m,s,i);else{let d=$c(u,s,i,t.shape,t.dtype);m.set(d)}return l}function zst(r,t,e,n,o){let s=0,i=n[0],a=n[1],u=i+o[0];for(let l=i;l<u;l++){let c=l*t+a;e.set(r.subarray(c,c+o[1]),s),s+=o[1]}}function Bst(r,t,e,n,o,s){let i=0,a=o[0],u=o[1],l=o[2],c=a+s[0],p=u+s[1];for(let m=a;m<c;m++)for(let f=u;f<p;f++){let d=m*t+f*e+l;n.set(r.subarray(d,d+s[2]),i),i+=s[2]}}function Vst(r,t,e,n,o,s,i){let a=0,u=s[0],l=s[1],c=s[2],p=u+i[0],m=l+i[1],f=c+i[2],d=s[3];for(let h=u;h<p;h++)for(let g=l;g<m;g++)for(let x=c;x<f;x++){let b=h*t+g*e+x*n+d;o.set(r.subarray(b,b+i[3]),a),a+=i[3]}}var uV={kernelName:gi,backendName:"wasm",kernelFunc:Lo};function Gst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,crops:i}=n,a=s.reduce((x,b)=>x*b),u=v.getReshaped(o.shape,s,a),l=v.getPermuted(u.length,s.length),c=v.getReshapedPermuted(o.shape,s,a),p=v.getSliceBeginCoords(i,s.length),m=v.getSliceSize(c,i,s.length),f=ar({inputs:{x:o},backend:e,attrs:{shape:u}}),d=ao({inputs:{x:f},backend:e,attrs:{perm:l}}),h=ar({inputs:{x:d},backend:e,attrs:{shape:c}}),g=Lo({inputs:{x:h},backend:e,attrs:{begin:p,size:m}});return e.disposeData(f.dataId),e.disposeData(d.dataId),e.disposeData(f.dataId),g}var cV={kernelName:ai,backendName:"wasm",kernelFunc:Gst};function ni(r){let{inputs:{x:t},attrs:{dtype:e},backend:n}=r,o=n.makeOutput(t.shape,e),s=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(o).set(s),o}var pV={kernelName:lo,backendName:"wasm",kernelFunc:ni};var mV=se(qo);var fV;function Wst(r){fV=r.wasm.cwrap(uo,null,["number","number","number","number"])}function Ust(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{clipValueMin:s,clipValueMax:i}=n,a=e.dataIdMap.get(o.dataId).id,u=e.makeOutput(o.shape,o.dtype),l=e.dataIdMap.get(u.dataId).id;return fV(a,s,i,l),u}var dV={kernelName:uo,backendName:"wasm",setupFunc:Wst,kernelFunc:Ust};function _k(r){let{inputs:t,backend:e}=r,n=y.parseAxisParam(r.attrs.axis,t[0].shape)[0],o=t.map(f=>f.shape);v.assertParamsConsistent(o,n);let s=v.computeOutShape(t.map(f=>f.shape),n),i=t.filter(f=>y.sizeFromShape(f.shape)>0);if(i.length===1)return jc({inputs:{x:i[0]},backend:e});let a=e.makeOutput(s,t[0].dtype);if(y.sizeFromShape(s)===0)return a;if(i[0].dtype==="string"){let f=i.map(w=>{let C=y.sizeFromShape(w.shape.slice(n));return ar({inputs:{x:w},backend:e,attrs:{shape:[-1,C]}})}),d=f.map(w=>({vals:e.readSync(w.dataId),shape:w.shape}));s=v.computeOutShape(f.map(w=>w.shape),1);let h=f[0].shape[0]===1,g=Ec(d,s,t[0].dtype,h),x=v.computeOutShape(i.map(w=>w.shape),n);a.shape=x;let b=e.dataIdMap.get(a.dataId);return b.stringBytes=v.fromStringArrayToUint8(g),f.forEach(w=>e.disposeData(w.dataId)),a}let u=y.sizeFromShape(i[0].shape.slice(0,n)),l=0,c=i.map(f=>{let d=y.sizeFromShape(f.shape.slice(n));return l+=d,d}),p=i.map(f=>e.typedArrayFromHeap(f)),m=e.typedArrayFromHeap(a);for(let f=0;f<u;f++){let d=f*l;for(let h=0;h<p.length;h++){let g=c[h],x=f*g,b=p[h].subarray(x,x+g);m.set(b,d),d+=g}}return a}var hV={kernelName:li,backendName:"wasm",kernelFunc:_k};var gV;function Hst(r){gV=r.wasm.cwrap(Ko,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function qst(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s}=t,i=n.dataIdMap.get(o.dataId).id,a=n.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:p,dataFormat:m}=e,f=v.convertConv2DDataFormat(m),d=v.computeConv2DInfo(o.shape,s.shape,u,l,c,p,!1,f),h=d.filterHeight,g=d.filterWidth,x=d.padInfo.top,b=d.padInfo.right,w=d.padInfo.bottom,C=d.padInfo.left,N=d.dilationHeight,_=d.dilationWidth,A=d.strideHeight,$=d.strideWidth,F=d.inChannels,P=d.outChannels,V=d.padInfo.type==="SAME"?1:0;if(d.dataFormat!=="channelsLast")throw new Error(`wasm backend Conv2D does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);let G=n.makeOutput(d.outShape,"float32"),W=n.dataIdMap.get(G.dataId).id;return gV(i,o.shape[0],o.shape[1],o.shape[2],a,h,g,x,b,w,C,V,N,_,A,$,F,P,W),G}var xV={kernelName:Ko,backendName:"wasm",setupFunc:Hst,kernelFunc:qst};var yV;function Kst(r){yV=r.wasm.cwrap(jo,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function jst(r){let{backend:t,inputs:e,attrs:n}=r,{dy:o,filter:s}=e,{strides:i,pad:a,dataFormat:u,dimRoundingMode:l,inputShape:c}=n,p=1,m=v.convertConv2DDataFormat(u),f=v.computeConv2DInfo(c,s.shape,i,p,a,l,!1,m),{batchSize:d,filterHeight:h,filterWidth:g,inChannels:x,inHeight:b,inWidth:w,outChannels:C,outHeight:N,outWidth:_,strideHeight:A,strideWidth:$}=f,F=h-1-f.padInfo.top,P=g-1-f.padInfo.left,V=f.dataFormat==="channelsLast",G=y.computeStrides(f.inShape),W=y.computeStrides(o.shape),[q,H,j]=y.computeStrides(s.shape),Y=G[0],Z=V?G[1]:G[2],et=V?G[2]:1,rt=V?1:G[1],ot=W[0],at=V?W[1]:W[2],nt=V?W[2]:1,it=V?1:W[1],dt=t.makeOutput(f.inShape,"float32"),ht=t.dataIdMap.get(dt.dataId).id,bt=t.dataIdMap.get(o.dataId).id,Et=t.dataIdMap.get(s.dataId).id;return yV(bt,Et,d,h,g,b,w,x,N,_,C,A,$,F,P,q,H,j,Y,Z,et,rt,ot,at,nt,it,ht),dt}var bV={kernelName:jo,backendName:"wasm",setupFunc:Kst,kernelFunc:jst};var wV=se(Xo);var CV=se(Yo);var Ak;(function(r){r[r.bilinear=0]="bilinear",r[r.nearest=1]="nearest"})(Ak||(Ak={}));var IV;function Xst(r){IV=r.wasm.cwrap(da,null,["number","number","number","number","array","number","number","number","number","number"])}function Yst(r){let{backend:t,inputs:e,attrs:n}=r,{method:o,extrapolationValue:s,cropSize:i}=n,{image:a,boxes:u,boxInd:l}=e,c=u.shape[0],[p,m]=i,f=[c,p,m,a.shape[3]],d=t.dataIdMap.get(a.dataId),h;a.dtype!=="float32"&&(h=ni({backend:t,inputs:{x:a},attrs:{dtype:"float32"}}),d=t.dataIdMap.get(h.dataId));let g=d.id,x=t.dataIdMap.get(u.dataId).id,b=t.dataIdMap.get(l.dataId).id,w=t.makeOutput(f,"float32"),C=t.dataIdMap.get(w.dataId).id,N=new Uint8Array(new Int32Array(a.shape).buffer);return IV(g,x,b,c,N,p,m,Ak[o],s,C),h!=null&&t.disposeData(h.dataId),w}var SV={kernelName:da,backendName:"wasm",setupFunc:Xst,kernelFunc:Yst};var vV;function Zst(r){vV=r.wasm.cwrap(fa,null,["number","number","number","number","number","number"])}function Jst(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n,u=o.shape.length;y.assert(o.dtype==="float32"||o.dtype==="int32",()=>`cumprod does not support ${o.dtype} tensors in the WASM backend`);let l=v.getAxesPermutation([s],u),c=o;l!==null&&(c=ao({inputs:{x:o},attrs:{perm:l},backend:e}));let p=v.getInnerMostAxes(1,u)[0];v.assertAxesAreInnerMostDims("cumprod",[p],u);let m=e.makeOutput(c.shape,c.dtype),f=c.shape[p],d=e.dataIdMap.get(c.dataId).id,h=e.dataIdMap.get(m.dataId).id;vV(d,i?1:0,a?1:0,f,h,qt[o.dtype]);let g=m;if(l!==null){let x=v.getUndoAxesPermutation(l);g=ao({inputs:{x:m},attrs:{perm:x},backend:e}),e.disposeData(c.dataId),e.disposeData(m.dataId)}return g}var NV={kernelName:fa,backendName:"wasm",setupFunc:Zst,kernelFunc:Jst};var TV;function Qst(r){TV=r.wasm.cwrap(Zo,null,["number","number","number","number","number","number"])}function tit(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{axis:s,exclusive:i,reverse:a}=n,u=o.shape.length;y.assert(o.dtype==="float32"||o.dtype==="int32",()=>`cumsum does not support ${o.dtype} tensors in the WASM backend`);let l=v.getAxesPermutation([s],u),c=o;l!==null&&(c=ao({inputs:{x:o},attrs:{perm:l},backend:e}));let p=v.getInnerMostAxes(1,u)[0];v.assertAxesAreInnerMostDims("cumsum",[p],u);let m=e.makeOutput(c.shape,c.dtype),f=c.shape[p],d=e.dataIdMap.get(c.dataId).id,h=e.dataIdMap.get(m.dataId).id;TV(d,i?1:0,a?1:0,f,h,qt[o.dtype]);let g=m;if(l!==null){let x=v.getUndoAxesPermutation(l);g=ao({inputs:{x:m},attrs:{perm:x},backend:e}),e.disposeData(c.dataId),e.disposeData(m.dataId)}return g}var kV={kernelName:Zo,backendName:"wasm",setupFunc:Qst,kernelFunc:tit};var EV;function eit(r){EV=r.wasm.cwrap(ha,null,["number","number","number","array","number","array","array","number","number"])}function rit(r){let{backend:t,inputs:e,attrs:n}=r,{x:o}=e,{blockSize:s,dataFormat:i}=n,a=o.shape[0],u=i==="NHWC"?o.shape[1]:o.shape[2],l=i==="NHWC"?o.shape[2]:o.shape[3],c=i==="NHWC"?o.shape[3]:o.shape[1],p=u*s,m=l*s,f=c/(s*s),d=i==="NHWC"?[a,p,m,f]:[a,f,p,m],h=t.makeOutput(d,"float32"),x=t.dataIdMap.get(o.dataId).id,b=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),w=new Uint8Array(new Int32Array(d).buffer),C=new Uint8Array(new Int32Array(y.computeStrides(d)).buffer),N=t.dataIdMap.get(h.dataId).id;return EV(x,s,i==="NHWC"?1:0,b,o.shape.length-1,w,C,d.length,N),h}var _V={kernelName:ha,backendName:"wasm",setupFunc:eit,kernelFunc:rit};var AV;function nit(r){AV=r.wasm.cwrap(Jo,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function oit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s}=t,i=n.dataIdMap.get(o.dataId).id,a=n.dataIdMap.get(s.dataId).id,{strides:u,dilations:l,pad:c,dimRoundingMode:p}=e,m=l==null?[1,1]:l,f=v.computeConv2DInfo(o.shape,s.shape,u,m,c,p,!0),d=f.filterHeight,h=f.filterWidth,g=f.padInfo.top,x=f.padInfo.right,b=f.padInfo.bottom,w=f.padInfo.left,C=f.dilationHeight,N=f.dilationWidth,_=f.strideHeight,A=f.strideWidth,$=f.inChannels,F=f.outChannels,P=f.padInfo.type==="SAME"?1:0;if(f.dataFormat!=="channelsLast")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);let V=n.makeOutput(f.outShape,"float32"),G=n.dataIdMap.get(V.dataId).id;return AV(i,o.shape[0],o.shape[1],o.shape[2],a,d,h,g,x,b,w,P,C,N,_,A,$,F,G),V}var $V={kernelName:Jo,backendName:"wasm",setupFunc:nit,kernelFunc:oit};var DV=se(ts);var sit=!1,RV=ue(xa,sit,"bool");var FV=se(es,"float32");function uI(r){let{inputs:t,attrs:e,backend:n}=r,{input:o}=t,{dim:s}=e,i=o.shape.length,a=o.shape.slice(),u=s;return s<0&&(y.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),u=i+s+1),a.splice(u,0,1),ar({inputs:{x:o},backend:n,attrs:{shape:a}})}var OV={kernelName:ui,backendName:"wasm",kernelFunc:uI};function $k(r){let{attrs:{shape:t,value:e,dtype:n},backend:o}=r,s=o.makeOutput(t,n);return o.typedArrayFromHeap(s).fill(e),s}var PV={kernelName:Dl,backendName:"wasm",kernelFunc:$k};var LV;function iit(r){LV=r.wasm.cwrap(ba,null,["number","number","number","number","number","number"])}function ait(r){let{inputs:t,backend:e}=r,{image:n}=t,o=e.makeOutput(n.shape,n.dtype),s=e.dataIdMap.get(n.dataId).id,i=e.dataIdMap.get(o.dataId).id,[a,u,l,c]=n.shape;return LV(s,a,u,l,c,i),o}var MV={kernelName:ba,backendName:"wasm",kernelFunc:ait,setupFunc:iit};var zV=se(rs);var lit=!1,BV=ue(ns,lit);var VV;function uit(r){VV=r.wasm.cwrap(os,null,["number","number","number","number","number","number","number"])}function cit(r){let{backend:t,inputs:e,attrs:n}=r,{varianceEpsilon:o}=n,{x:s,mean:i,variance:a,offset:u,scale:l}=e,c=t.dataIdMap.get(s.dataId).id,p=t.dataIdMap.get(i.dataId).id,m=t.dataIdMap.get(a.dataId).id,f=u!=null?t.dataIdMap.get(u.dataId).id:0,d=l!=null?t.dataIdMap.get(l.dataId).id:0,h=t.makeOutput(s.shape,s.dtype);if(y.sizeFromShape(s.shape)===0)return h;let g=t.dataIdMap.get(h.dataId).id;return VV(c,p,m,f,d,o,g),h}var GV={kernelName:os,backendName:"wasm",setupFunc:uit,kernelFunc:cit};var WV;function pit(r){WV=r.wasm.cwrap(Ii,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function mit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dataFormat:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=e,h=v.computeConv2DInfo(o.shape,s.shape,u,c,l,m),g=Du[f];if(g==null)throw new Error(`${f} activation not yet supported for FusedConv2D in the wasm backend.`);let x=n.dataIdMap.get(o.dataId).id,b=n.dataIdMap.get(s.dataId).id,w=h.outChannels,C=0;if(i!=null){let nt=n.dataIdMap.get(i.dataId);if(nt.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${nt.shape.length}.`);if(nt.shape[0]!==w)throw new Error(`FusedConv2D bias shape (${nt.shape}) does not match the number of output channels (${w})`);C=nt.id}let N=h.filterHeight,_=h.filterWidth,A=h.padInfo.top,$=h.padInfo.right,F=h.padInfo.bottom,P=h.padInfo.left,V=h.dilationHeight,G=h.dilationWidth,W=h.strideHeight,q=h.strideWidth,H=h.inChannels,j=h.padInfo.type==="SAME"?1:0,Y=h.batchSize,Z=h.inHeight,et=h.inWidth;if(p!=="NHWC")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let rt=n.makeOutput(h.outShape,"float32"),ot=n.dataIdMap.get(rt.dataId).id,at=a==null?0:n.dataIdMap.get(a.dataId).id;return WV(x,Y,Z,et,b,N,_,C,A,$,F,P,j,V,G,W,q,H,w,g,at,d||0,ot),rt}var UV={kernelName:Ii,backendName:"wasm",setupFunc:pit,kernelFunc:mit};var HV;function fit(r){HV=r.wasm.cwrap(Si,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function dit(r){let{inputs:t,attrs:e,backend:n}=r,{x:o,filter:s,bias:i,preluActivationWeights:a}=t,{strides:u,pad:l,dilations:c,dataFormat:p,dimRoundingMode:m,activation:f,leakyreluAlpha:d}=e,h=v.computeConv2DInfo(o.shape,s.shape,u,c,l,m,!0),g=Du[f];if(g==null)throw new Error(`${f} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let x=n.dataIdMap.get(o.dataId).id,b=n.dataIdMap.get(s.dataId).id,w=h.outChannels,C=0;if(i!=null){let nt=n.dataIdMap.get(i.dataId);if(nt.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${nt.shape.length}.`);if(nt.shape[0]!==w)throw new Error(`FusedDepthwiseConv2D bias shape (${nt.shape}) does not match the number of output channels (${w})`);C=nt.id}let N=h.filterHeight,_=h.filterWidth,A=h.padInfo.top,$=h.padInfo.right,F=h.padInfo.bottom,P=h.padInfo.left,V=h.dilationHeight,G=h.dilationWidth,W=h.strideHeight,q=h.strideWidth,H=h.inChannels,j=h.padInfo.type==="SAME"?1:0,Y=h.batchSize,Z=h.inHeight,et=h.inWidth;if(p!=="NHWC")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${p}'. Please use 'NHWC'.`);let rt=n.makeOutput(h.outShape,"float32"),ot=n.dataIdMap.get(rt.dataId).id,at=a==null?0:n.dataIdMap.get(a.dataId).id;return HV(x,Y,Z,et,b,N,_,C,A,$,F,P,j,V,G,W,q,H,w,g,at,d||0,ot),rt}var qV={kernelName:Si,backendName:"wasm",setupFunc:fit,kernelFunc:dit};var KV;function hit(r){KV=r.wasm.cwrap(wa,null,["number","number","number","number","number","number","array","number"])}function git(r){let{backend:t,inputs:e}=r,{params:n,indices:o}=e,[s,i,a,u]=ox.prepareAndValidate(n,o),l=t.makeOutput(s,n.dtype);if(i===0)return l;let c=o.shape,p=c[c.length-1],f=t.dataIdMap.get(n.dataId).id,h=t.dataIdMap.get(o.dataId).id,g=new Uint8Array(new Int32Array(u).buffer),x=t.dataIdMap.get(l.dataId).id;return KV(f,qt[n.dtype],h,i,p,a,g,x),l}var jV={kernelName:wa,backendName:"wasm",setupFunc:hit,kernelFunc:git};var XV;function xit(r){XV=r.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])}function yit(r){let{backend:t,inputs:e,attrs:n}=r,{x:o,indices:s}=e,{axis:i,batchDims:a}=n,u=y.parseAxisParam(i,o.shape)[0],l=t.readSync(s.dataId),c=o.shape[u];for(let F=0;F<l.length;++F){let P=l[F];y.assert(P<=c-1&&P>=0,()=>`GatherV2: the index value ${P} is not in [0, ${c-1}]`)}let p=v.segment_util.collectGatherOpShapeInfo(o,s,u,a),m=ar({inputs:{x:o},attrs:{shape:[p.batchSize,p.outerSize,p.dimSize,p.sliceSize]},backend:t}),f=y.sizeFromShape(s.shape),d=ar({inputs:{x:s},attrs:{shape:[p.batchSize,f/p.batchSize]},backend:t}),h=[p.batchSize,p.outerSize,f/p.batchSize,p.sliceSize],g=t.makeOutput(h,o.dtype);if(y.sizeFromShape(o.shape)===0)return g;let x=m.shape.length-1,w=t.dataIdMap.get(m.dataId).id,N=t.dataIdMap.get(d.dataId).id,_=t.dataIdMap.get(g.dataId).id,A=new Uint8Array(new Int32Array(y.computeStrides(m.shape)).buffer),$=new Uint8Array(new Int32Array(y.computeStrides(h)).buffer);return XV(w,qt[o.dtype],A,x,N,p.batchSize,$,_),t.disposeData(m.dataId),t.disposeData(d.dataId),g.shape=p.outputShape,g}var YV={kernelName:ci,backendName:"wasm",setupFunc:xit,kernelFunc:yit};var bit=!1,ZV=ue(Ca,bit,"bool");var wit=!1,JV=ue(ss,wit,"bool");var QV;function Cit(r){QV=r.wasm.cwrap(is,null,["number","number","number","number"])}function Iit(r){let{inputs:{x:t},attrs:{alpha:e},backend:n}=r,o=n.dataIdMap.get(t.dataId).id,s=n.makeOutput(t.shape,"float32");if(y.sizeFromShape(t.shape)!==0){let i=n.dataIdMap.get(s.dataId).id;QV(o,qt[t.dtype],e,i)}return s}var tG={kernelName:is,backendName:"wasm",setupFunc:Cit,kernelFunc:Iit};var Sit=!1,eG=ue(Na,Sit,"bool");var vit=!1,rG=ue(Ta,vit,"bool");var nG=se(as);var Nit=!1,oG=ue(Ea,Nit,"bool");var sG=se(_a);var Tit=!1,iG=ue(Aa,Tit,"bool");var kit=!1,aG=ue(m1,kit,"bool");var lG;function Eit(r){lG=r.wasm.cwrap(ls,null,["number","number","number","number"])}function _it(r){let{backend:t,inputs:e,attrs:n}=r,{reductionIndices:o,keepDims:s}=n,{x:i}=e,u=t.dataIdMap.get(i.dataId).id,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;l=c,u=w}let d=l.shape.length;v.assertAxesAreInnerMostDims("max",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,i.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;lG(u,qt[i.dtype],x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var uG={kernelName:ls,backendName:"wasm",setupFunc:Eit,kernelFunc:_it};var Ait=!1,cG=ue(us,Ait);var pG;function $it(r){pG=r.wasm.cwrap(cs,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function Dit(r){let{inputs:t,attrs:e,backend:n}=r,o=t.x,s=n.dataIdMap.get(o.dataId).id;y.assert(o.dtype==="float32",()=>`Error in MaxPool: only float32 input is supported. Got ${o.dtype}.`);let{filterSize:i,strides:a,pad:u,dimRoundingMode:l}=e,c=v.computePool2DInfo(o.shape,i,a,1,u,l),p=c.filterHeight,m=c.filterWidth,f=c.padInfo.top,d=c.padInfo.right,h=c.padInfo.bottom,g=c.padInfo.left,x=c.dilationHeight,b=c.dilationWidth,w=c.strideHeight,C=c.strideWidth,N=c.inChannels,_=c.outChannels;if(c.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);let A=n.makeOutput(c.outShape,"float32"),$=n.dataIdMap.get(A.dataId).id;return pG(s,o.shape[0],o.shape[1],o.shape[2],p,m,f,d,h,g,x,b,w,C,N,_,$),A}var mG={kernelName:cs,backendName:"wasm",setupFunc:$it,kernelFunc:Dit};var fG;function Rit(r){fG=r.wasm.cwrap(ps,null,["number, number, number"])}function Fit(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let C=t.dataIdMap.get(c.dataId).id;C!==a&&(l=c,u=C,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims("mean",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=l;l.dtype!=="float32"&&(b=ni({backend:t,inputs:{x:l},attrs:{dtype:"float32"}}),u=t.dataIdMap.get(b.dataId).id);let w=t.makeOutput(h,"float32");if(y.sizeFromShape(l.shape)!==0){let C=t.dataIdMap.get(w.dataId).id;fG(u,x,C)}if(f&&t.disposeData(c.dataId),s){let C=v.expandShapeToKeepDim(w.shape,m);w.shape=C}return l.dtype!=="float32"&&t.disposeData(b.dataId),w}var dG={kernelName:ps,backendName:"wasm",setupFunc:Rit,kernelFunc:Fit};var hG;function Oit(r){hG=r.wasm.cwrap(ms,null,["number","number","number","number"])}function Pit(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t);if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w)}let d=l.shape.length;v.assertAxesAreInnerMostDims("min",p,d);let[h,g]=v.computeOutAndReduceShapes(l.shape,p),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;hG(u,qt[i.dtype],x,w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var gG={kernelName:ms,backendName:"wasm",setupFunc:Oit,kernelFunc:Pit};var Lit=!1,xG=ue(fs,Lit);var Dk;(function(r){r[r.reflect=0]="reflect",r[r.symmetric=1]="symmetric"})(Dk||(Dk={}));var yG;function Mit(r){yG=r.wasm.cwrap(ds,null,["number","array","number","number","array","array","number","number"])}function zit(r){let{inputs:{x:t},backend:e,attrs:{paddings:n,mode:o}}=r,s=n.map((d,h)=>d[0]+t.shape[h]+d[1]),i=e.dataIdMap.get(t.dataId).id,a=e.makeOutput(s,t.dtype),u=e.dataIdMap.get(a.dataId).id,l=new Uint8Array(new Int32Array(t.shape).buffer),c=n.map(d=>d[0]),p=n.map(d=>d[1]),m=new Uint8Array(new Int32Array(c).buffer),f=new Uint8Array(new Int32Array(p).buffer);return yG(i,l,t.shape.length,qt[t.dtype],m,f,Dk[o],u),a}var bG={kernelName:ds,backendName:"wasm",kernelFunc:zit,setupFunc:Mit};var Bit=!0,wG=ue(hs,Bit);var CG=se(pi);function Dd(r,t){let e=new Int32Array(r.wasm.HEAPU8.buffer,t,4),n=e[0],o=e[1],s=e[2],i=e[3];return r.wasm._free(t),{pSelectedIndices:n,selectedSize:o,pSelectedScores:s,pValidOutputs:i}}var IG;function Vit(r){IG=r.wasm.cwrap(Ra,"number",["number","number","number","number","number"])}function Git(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i}=n,{boxes:a,scores:u}=e,l=t.dataIdMap.get(a.dataId).id,c=t.dataIdMap.get(u.dataId).id,p=IG(l,c,s,o,i),{pSelectedIndices:m,selectedSize:f,pSelectedScores:d,pValidOutputs:h}=Dd(t,p);return t.wasm._free(d),t.wasm._free(h),t.makeOutput([f],"int32",m)}var SG={kernelName:Ra,backendName:"wasm",setupFunc:Vit,kernelFunc:Git};var vG;function Wit(r){vG=r.wasm.cwrap(Fa,"number",["number","number","number","number","number","bool"])}function Uit(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:a}=n,{boxes:u,scores:l}=e,c=t.dataIdMap.get(u.dataId).id,p=t.dataIdMap.get(l.dataId).id,m=vG(c,p,s,o,i,a),{pSelectedIndices:f,selectedSize:d,pSelectedScores:h,pValidOutputs:g}=Dd(t,m);t.wasm._free(h);let x=t.makeOutput([d],"int32",f),b=t.makeOutput([],"int32",g);return[x,b]}var NG={kernelName:Fa,backendName:"wasm",setupFunc:Wit,kernelFunc:Uit};var TG;function Hit(r){TG=r.wasm.cwrap(Oa,"number",["number","number","number","number","number","number"])}function qit(r){let{backend:t,inputs:e,attrs:n}=r,{iouThreshold:o,maxOutputSize:s,scoreThreshold:i,softNmsSigma:a}=n,{boxes:u,scores:l}=e,c=t.dataIdMap.get(u.dataId).id,p=t.dataIdMap.get(l.dataId).id,m=TG(c,p,s,o,i,a),{pSelectedIndices:f,selectedSize:d,pSelectedScores:h,pValidOutputs:g}=Dd(t,m);t.wasm._free(g);let x=t.makeOutput([d],"int32",f),b=t.makeOutput([d],"float32",h);return[x,b]}var kG={kernelName:Oa,backendName:"wasm",setupFunc:Hit,kernelFunc:qit};var Kit=!1,EG=ue(Da,Kit,"bool");var _G;function jit(r){_G=r.wasm.cwrap(gs,null,["number","number","number","number","number"])}function Xit(r){let{inputs:t,backend:e,attrs:n}=r,{indices:o}=t,{dtype:s,depth:i,onValue:a,offValue:u}=n,l=e.makeOutput([...o.shape,i],s),c=e.dataIdMap.get(l.dataId).id,m=e.dataIdMap.get(o.dataId).id;return _G(m,i,a,u,c),l}var AG={kernelName:gs,backendName:"wasm",setupFunc:jit,kernelFunc:Xit};function Yit(r){let{inputs:{x:t},backend:e}=r,n=e.makeOutput(t.shape,t.dtype);return e.typedArrayFromHeap(n).fill(1),n}var $G={kernelName:mi,backendName:"wasm",kernelFunc:Yit};function Zit(r){let{inputs:t,backend:e,attrs:n}=r,{axis:o}=n;if(t.length===1)return uI({inputs:{input:t[0]},backend:e,attrs:{dim:o}});let s=t[0].shape,i=t[0].dtype;t.forEach(c=>{y.assertShapesMatch(s,c.shape,"All tensors passed to stack must have matching shapes"),y.assert(i===c.dtype,()=>"All tensors passed to stack must have matching dtypes")});let a=[],u=t.map(c=>{let p=uI({inputs:{input:c},backend:e,attrs:{dim:o}});return a.push(p),p}),l=_k({inputs:u,backend:e,attrs:{axis:o}});return a.forEach(c=>e.disposeData(c.dataId)),l}var DG={kernelName:fi,backendName:"wasm",kernelFunc:Zit};var RG;function Jit(r){RG=r.wasm.cwrap(xs,null,["number","array","number","number","array","array","number","number"])}function Qit(r){let{inputs:{x:t},backend:e,attrs:{paddings:n,constantValue:o}}=r,s=n.map((h,g)=>h[0]+t.shape[g]+h[1]);if(y.sizeFromShape(t.shape)===0)return $k({backend:e,attrs:{shape:s,value:o,dtype:t.dtype}});let i=e.dataIdMap.get(t.dataId).id,a=e.makeOutput(s,t.dtype),l=e.dataIdMap.get(a.dataId).id,c=new Uint8Array(new Int32Array(t.shape).buffer),p=n.map(h=>h[0]),m=n.map(h=>h[1]),f=new Uint8Array(new Int32Array(p).buffer),d=new Uint8Array(new Int32Array(m).buffer);return RG(i,c,t.shape.length,qt[t.dtype],f,d,o,l),a}var cI={kernelName:xs,backendName:"wasm",kernelFunc:Qit,setupFunc:Jit};var tat=!1,FG=ue(ys,tat);var OG;function eat(r){OG=r.wasm.cwrap(bs,null,["number","number","number"])}function rat(r){let{inputs:t,backend:e}=r,{x:n,alpha:o}=t,s=e.dataIdMap.get(n.dataId).id,i=e.dataIdMap.get(o.dataId).id,a=s,u=n,l=u;u.dtype!=="float32"&&(l=ni({backend:e,inputs:{x:n},attrs:{dtype:"float32"}}),a=e.dataIdMap.get(l.dataId).id);let c=e.makeOutput(n.shape,"float32"),p=e.dataIdMap.get(c.dataId).id;return OG(a,i,p),u.dtype!=="float32"&&e.disposeData(l.dataId),c}var PG={kernelName:bs,backendName:"wasm",setupFunc:eat,kernelFunc:rat};var LG;function nat(r){LG=r.wasm.cwrap(ws,null,["number","number","number","number"])}function oat(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims("prod",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;LG(u,x,qt[b.dtype],w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var MG={kernelName:ws,backendName:"wasm",setupFunc:nat,kernelFunc:oat};var sat=r=>{let{backend:t,attrs:e}=r,{start:n,stop:o,step:s,dtype:i}=e,a=Ac(n,o,s,i),u=t.makeOutput([a.length],i);return t.typedArrayFromHeap(u).set(a),u},zG={kernelName:Ol,backendName:"wasm",kernelFunc:sat};var iat=!0,BG=ue(Qo,iat);var VG=se(Cs);var GG=se(vs);var WG;function aat(r){WG=r.wasm.cwrap(Ss,null,["number","number","number","number","number","number","number","number","number","number"])}function lat(r){let{backend:t,inputs:e,attrs:n}=r,{images:o}=e,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,[c,p,m,f]=o.shape,d=[c,u,l,f],h=t.dataIdMap.get(o.dataId),g;h.dtype!=="float32"&&(g=ni({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),h=t.dataIdMap.get(g.dataId));let x=h.id,b=t.makeOutput(d,"float32");if(y.sizeFromShape(o.shape)===0)return b;let w=t.dataIdMap.get(b.dataId).id;return WG(x,c,p,m,f,u,l,s?1:0,i?1:0,w),g!=null&&t.disposeData(g.dataId),b}var UG={kernelName:Ss,backendName:"wasm",setupFunc:aat,kernelFunc:lat};var HG;function uat(r){HG=r.wasm.cwrap(Is,null,["number","number","number","number","number","number","number","number","number","number"])}function cat(r){let{backend:t,inputs:e,attrs:n}=r,{images:o}=e,{alignCorners:s,halfPixelCenters:i,size:a}=n,[u,l]=a,[c,p,m,f]=o.shape,d=[c,u,l,f],h=t.makeOutput(d,"float32");if(y.sizeFromShape(o.shape)===0)return h;let g=t.dataIdMap.get(o.dataId),x;g.dtype!=="float32"&&(x=ni({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),g=t.dataIdMap.get(x.dataId));let b=g.id,w=t.dataIdMap.get(h.dataId).id;return HG(b,c,p,m,f,u,l,s?1:0,i?1:0,w),x!=null&&t.disposeData(x.dataId),h}var qG={kernelName:Is,backendName:"wasm",setupFunc:uat,kernelFunc:cat};var KG;function pat(r){KG=r.wasm.cwrap(Ns,null,["number","array","number","array","number","number"])}function mat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{dims:s}=n,i=y.parseAxisParam(s,o.shape);if(o.shape.length===0)return jc({inputs:{x:o},backend:e});let a=e.makeOutput(o.shape,o.dtype),u=e.dataIdMap.get(o.dataId).id,l=e.dataIdMap.get(a.dataId).id,c=new Uint8Array(new Int32Array(i).buffer),p=new Uint8Array(new Int32Array(o.shape).buffer);KG(u,c,i.length,p,o.shape.length,l);let m=ar({inputs:{x:a},attrs:{shape:o.shape},backend:e});return e.disposeData(a.dataId),m}var jG={kernelName:Ns,backendName:"wasm",kernelFunc:mat,setupFunc:pat};var XG;function fat(r){XG=r.wasm.cwrap(qa,null,["number","number","number","number","number","number","number","number","array","number","number"])}function dat(r){let{inputs:t,backend:e,attrs:n}=r,{image:o}=t,{radians:s,fillValue:i,center:a}=n,u=e.makeOutput(o.shape,o.dtype),l=e.dataIdMap.get(o.dataId).id,c=e.dataIdMap.get(u.dataId).id,[p,m,f,d]=o.shape,[h,g]=v.getImageCenter(a,m,f),x=i===0,b=255,w=typeof i=="number"?[i,i,i,x?0:b]:[...i,b],C=new Uint8Array(new Int32Array(w).buffer);return XG(l,p,m,f,d,s,h,g,C,w.length,c),u}var YG={kernelName:qa,backendName:"wasm",kernelFunc:dat,setupFunc:fat};var ZG=se(Ts);var JG=se(ks);var QG;function hat(r){QG=r.wasm.cwrap(La,null,["number","number","number","number","number","number","array","number","number"])}function gat(r){let{backend:t,inputs:e,attrs:n}=r,{indices:o,updates:s}=e,{shape:i}=n,a=t.makeOutput(i,s.dtype);if(y.sizeFromShape(i)===0)return a;let{sliceRank:u,numUpdates:l,sliceSize:c,strides:p,outputSize:m}=lh.calculateShapes(s,o,i),d=t.dataIdMap.get(o.dataId).id,g=t.dataIdMap.get(s.dataId).id,x=new Uint8Array(new Int32Array(p).buffer),b=t.dataIdMap.get(a.dataId).id;return QG(d,g,qt[s.dtype],u,l,c,x,m,b),a}var tW={kernelName:La,backendName:"wasm",setupFunc:hat,kernelFunc:gat};var eW;function xat(r){eW=r.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}function yat(r){let{inputs:t,backend:e}=r,{condition:n,t:o,e:s}=t,i=e.dataIdMap.get(n.dataId).id,a=e.dataIdMap.get(o.dataId).id,u=e.dataIdMap.get(s.dataId).id,l=e.makeOutput(o.shape,o.dtype),c=e.dataIdMap.get(l.dataId).id,p=n.shape.length,m=o.shape.length,f=p===0||p>1||m===1?1:y.sizeFromShape(o.shape.slice(1));return eW(i,a,u,f,c),l}var rW={kernelName:hi,backendName:"wasm",kernelFunc:yat,setupFunc:xat};var nW;function bat(r){nW=r.wasm.cwrap(_s,null,["number","number"])}function wat(r){let{backend:t,inputs:{x:e}}=r,n=t.dataIdMap.get(e.dataId).id,o=t.makeOutput(e.shape,e.dtype),s=t.dataIdMap.get(o.dataId).id;return y.sizeFromShape(o.shape)===0||nW(n,s),o}var oW={kernelName:"Sigmoid",backendName:"wasm",setupFunc:bat,kernelFunc:wat};var sW=se(Es);var iW;function Cat(r){iW=r.wasm.cwrap(Ds,null,["number","number","number","number"])}function Iat(r){let{backend:t,inputs:{logits:e},attrs:{dim:n}}=r,o=t.dataIdMap.get(e.dataId).id,s=t.makeOutput(e.shape,e.dtype),i=t.dataIdMap.get(s.dataId).id,a=e.shape[n],u=y.sizeFromShape(e.shape)/a;return y.sizeFromShape(s.shape)===0||iW(o,i,a,u),s}var aW={kernelName:Ds,backendName:"wasm",setupFunc:Cat,kernelFunc:Iat};function Sat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,{blockShape:s,paddings:i}=n,a=y.sizeFromShape(s),u=[[0,0]];u.push(...i);for(let _=1+s.length;_<o.shape.length;++_)u.push([0,0]);let l=cI.kernelFunc({inputs:{x:o},backend:e,attrs:{paddings:u,constantValue:0}}),c=v.getReshaped(l.shape,s,a,!1),p=v.getPermuted(c.length,s.length,!1),m=v.getReshapedPermuted(l.shape,s,a,!1),h=ar({inputs:{x:l},backend:e,attrs:{shape:c}}),b=ao({inputs:{x:h},backend:e,attrs:{perm:p}}),N=ar({inputs:{x:b},backend:e,attrs:{shape:m}});return e.disposeData(l.dataId),e.disposeData(h.dataId),e.disposeData(b.dataId),N}var lW={kernelName:xi,backendName:"wasm",kernelFunc:Sat};var uW;function vat(r){uW=r.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])}function Nat(r){let{backend:t,inputs:e}=r,{indices:n,values:o,denseShape:s,defaultValue:i}=e,a=n.shape[0],u=n.shape[1],l=t.readSync(s.dataId)[0],c=[a+l,u],p=t.dataIdMap.get(n.dataId).id,m=t.dataIdMap.get(o.dataId).id,f=t.dataIdMap.get(i.dataId).id,d=t.makeOutput(c,n.dtype),h=t.dataIdMap.get(d.dataId).id,g=t.makeOutput(c.slice(0,1),o.dtype),x=t.dataIdMap.get(g.dataId).id,b=t.makeOutput([l],"bool"),w=t.dataIdMap.get(b.dataId).id,C=t.makeOutput([a],n.dtype),N=t.dataIdMap.get(C.dataId).id,_=t.makeOutput([4],"int32"),A=t.dataIdMap.get(_.dataId).id,$=uW(p,m,qt[o.dtype],a,l,u,f,h,x,w,N,A),F=t.readSync(_.dataId),P;switch(F[0]){case 1:{P=v.getSparseFillEmptyRowsIndicesDenseShapeMismatch(F[1]);break}case 2:{P=v.getSparseFillEmptyRowsNegativeIndexErrorMessage(F[1],F[2]);break}case 3:P=v.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(F[1],F[2],F[3]);break;default:P=""}if(t.disposeData(_.dataId),P)throw t.disposeData(d.dataId),t.disposeData(g.dataId),t.disposeData(b.dataId),t.disposeData(C.dataId),new Error(P);let V=d,G=g;return $!==c[0]&&(V=Lo({inputs:{x:d},attrs:{begin:0,size:[$,u]},backend:t}),G=Lo({inputs:{x:g},attrs:{begin:0,size:$},backend:t}),t.disposeData(d.dataId),t.disposeData(g.dataId)),[V,G,b,C]}var cW={kernelName:Pl,backendName:"wasm",setupFunc:vat,kernelFunc:Nat};var pW;function Tat(r){pW=r.wasm.cwrap(Ga,null,["number","number","number","number","number","number","number"])}function kat(r){let{backend:t,inputs:e}=r,{inputIndices:n,inputShape:o,newShape:s}=e;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
-        ${n.shape}`);if(o.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
-        ${o.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(n.dataId).id,a=t.dataIdMap.get(o.dataId).id,u=t.dataIdMap.get(s.dataId).id,l=n.shape[0],c=y.sizeFromShape(s.shape),p=t.makeOutput([l,c],n.dtype),m=t.dataIdMap.get(p.dataId).id,f=t.makeOutput([c],s.dtype),d=t.dataIdMap.get(f.dataId).id,h=t.makeOutput([3],"int32"),g=t.dataIdMap.get(h.dataId).id;pW(i,a,u,l,m,d,g);let x=t.readSync(h.dataId),b;switch(x[0]){case 0:{b=v.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(x[1],x[2]);break}case 1:{b=v.getSparseReshapeNegativeOutputDimErrorMessage(x[1],x[2]);break}case 2:b=v.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let w=Array.from(t.readSync(o.dataId)),C=Array.from(t.readSync(f.dataId));b=v.getSparseReshapeInputOutputMultipleErrorMessage(w,C);break}case 4:{let w=Array.from(t.readSync(o.dataId)),C=Array.from(t.readSync(f.dataId));b=v.getSparseReshapeInputOutputMismatchErrorMessage(w,C);break}default:b=""}if(t.disposeData(h.dataId),b)throw t.disposeData(p.dataId),t.disposeData(f.dataId),new Error(b);return[p,f]}var mW={kernelName:Ga,backendName:"wasm",setupFunc:Tat,kernelFunc:kat};var fW;function pI(r){fW=r.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function mI(r,t){let{backend:e,inputs:n}=r,{data:o,indices:s,segmentIds:i}=n,a=s.shape[0],u=e.readSync(i.dataId,a-1,a)[0],c=a>0?u+1:0;if(c<0)throw new Error(v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let p=o.shape.slice();p[0]=c;let m=e.dataIdMap.get(o.dataId).id,f=e.dataIdMap.get(s.dataId).id,d=e.dataIdMap.get(i.dataId).id,h=e.makeOutput(p,o.dtype),g=e.dataIdMap.get(h.dataId).id,x=e.makeOutput([4],"int32"),b=e.dataIdMap.get(x.dataId).id;fW(m,qt[o.dtype],o.shape[0],f,d,g,b,t,0);let w=e.readSync(x.dataId),C;switch(w[0]){case 0:{C=v.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{C=v.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:C=v.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(w[1],w[2]);break;case 3:C=v.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(w[1],w[2],w[3]);break;default:C=""}if(e.disposeData(x.dataId),C)throw e.disposeData(h.dataId),new Error(C);return h}function Eat(r){return mI(r,!0)}var dW={kernelName:Ll,backendName:"wasm",setupFunc:pI,kernelFunc:Eat};function _at(r){return mI(r,!1)}var hW={kernelName:Ml,backendName:"wasm",setupFunc:pI,kernelFunc:_at};function Aat(r){let{inputs:t,attrs:e,backend:n}=r,{x:o}=t,{numOrSizeSplits:s,axis:i}=e,a=y.parseAxisParam(i,o.shape)[0],u=v.prepareSplitSize(o,s,a),l=new Array(o.shape.length).fill(0),c=o.shape.slice();return u.map(p=>{let m=[...c];m[a]=p;let f=Lo({inputs:{x:o},attrs:{begin:l,size:m},backend:n});return l[a]+=p,f})}var gW={kernelName:yi,backendName:"wasm",kernelFunc:Aat};var xW=se(As);var yW=se(zl);var $at=!0,bW=ue(Rs,$at);var wW;function Dat(r){wW=r.wasm.cwrap(po,null,["number","number","number","number"])}function Rat(r){let{backend:t,inputs:e,attrs:n}=r,{alpha:o}=n,{x:s}=e,i=t.dataIdMap.get(s.dataId).id,a=t.makeOutput(s.shape,s.dtype),u=t.dataIdMap.get(a.dataId).id;return wW(i,o,qt[s.dtype],u),a}var CW={kernelName:po,backendName:"wasm",setupFunc:Dat,kernelFunc:Rat};var IW;function Fat(r){IW=r.wasm.cwrap(Wa,null,["number","array","number","array","array","array","array","array","number","number"])}function Oat(r){let{backend:t,inputs:e,attrs:n}=r,{x:o}=e,{begin:s,end:i,strides:a,beginMask:u,endMask:l,ellipsisMask:c,newAxisMask:p,shrinkAxisMask:m}=n,{finalShapeSparse:f,finalShape:d,isIdentity:h,sliceDim0:g,isSimpleSlice:x,begin:b,end:w,strides:C}=Le.sliceInfo(o.shape,s,i,a,u,l,c,p,m),N;if(h)N=ar({inputs:{x:o},backend:t,attrs:{shape:d}});else if(g||x){y.assert(o.shape.length>=1,()=>`Input must have rank at least 1, got: ${o.shape.length}`);let _=Le.computeOutShape(b,w,C),A=Lo({inputs:{x:o},backend:t,attrs:{begin:b,size:_}});N=ar({inputs:{x:A},backend:t,attrs:{shape:d}}),t.disposeData(A.dataId)}else{let _=t.makeOutput(f,"float32"),A=t.dataIdMap.get(o.dataId).id,$=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),F=new Uint8Array(new Int32Array(b).buffer),P=new Uint8Array(new Int32Array(w).buffer),V=new Uint8Array(new Int32Array(C).buffer),G=new Uint8Array(new Int32Array(f).buffer),W=new Uint8Array(new Int32Array(y.computeStrides(f)).buffer),q=t.dataIdMap.get(_.dataId).id;IW(A,$,o.shape.length,F,P,V,G,W,f.length,q),N=ar({inputs:{x:_},backend:t,attrs:{shape:d}}),t.disposeData(_.dataId)}return N}var SW={kernelName:Wa,backendName:"wasm",setupFunc:Fat,kernelFunc:Oat};function Pat(r){let{backend:t,inputs:e,attrs:n}=r,{data:o,dataSplits:s}=e,{separator:i,nGramWidths:a,leftPad:u,rightPad:l,padWidth:c,preserveShortSequences:p}=n,m=t.readSync(o.dataId),f=t.readSync(s.dataId),[d,h]=Dc(m,f,i,a,u,l,c,p),g=t.makeOutput([d.length],"string"),x=t.dataIdMap.get(g.dataId);x.stringBytes=d;let b=t.makeOutput(s.shape,"int32");return t.typedArrayFromHeap(b).set(h),[g,b]}var vW={kernelName:Bl,backendName:"wasm",kernelFunc:Pat};function Lat(r){let{backend:t,inputs:e,attrs:n}=r,{input:o,delimiter:s}=e,{skipEmpty:i}=n,a=t.readSync(o.dataId),u=t.readSync(s.dataId),[l,c,p]=Rc(a,u[0],i),m=c.length,f=t.makeOutput([m,2],"int32");t.typedArrayFromHeap(f).set(l);let h=t.makeOutput([m],"string"),g=t.dataIdMap.get(h.dataId);g.stringBytes=c;let x=t.makeOutput([2],"int32");return t.typedArrayFromHeap(x).set(p),[f,h,x]}var NW={kernelName:Vl,backendName:"wasm",kernelFunc:Lat};function Mat(r){let{backend:t,inputs:e,attrs:n}=r,{input:o}=e,{numBuckets:s}=n,i=t.readSync(o.dataId),a=Fc(i,s),u=t.makeOutput(o.shape,"int32");return t.typedArrayFromHeap(u).set(a),u}var TW={kernelName:Gl,backendName:"wasm",kernelFunc:Mat};var zat=!0,kW=ue(Fs,zat);var EW;function Bat(r){EW=r.wasm.cwrap($s,null,["number","number","number","number"])}function Vat(r){let{backend:t,inputs:e,attrs:n}=r,{axis:o,keepDims:s}=n,{x:i}=e,a=t.dataIdMap.get(i.dataId).id,u=a,l=i,{transposed:c,axes:p,originalAxes:m,inputWasTransposed:f}=bn(i,o,t),d=p;if(f){let w=t.dataIdMap.get(c.dataId).id;w!==a&&(l=c,u=w,d=v.getInnerMostAxes(d.length,l.shape.length))}v.assertAxesAreInnerMostDims("sum",d,l.shape.length);let[h,g]=v.computeOutAndReduceShapes(l.shape,d),x=y.sizeFromShape(g),b=t.makeOutput(h,l.dtype);if(y.sizeFromShape(l.shape)!==0){let w=t.dataIdMap.get(b.dataId).id;EW(u,x,qt[b.dtype],w)}if(f&&t.disposeData(c.dataId),s){let w=v.expandShapeToKeepDim(b.shape,m);b.shape=w}return b}var _W={kernelName:$s,backendName:"wasm",setupFunc:Bat,kernelFunc:Vat};var AW=se(Os);var $W=se(Ps);var DW;function Gat(r){DW=r.wasm.cwrap(Jn,null,["number","array","number","array","number","number"])}function Wat(r){let{inputs:t,backend:e,attrs:n}=r,{x:o}=t,s=e.dataIdMap.get(o.dataId).id,{reps:i}=n,a=new Array(o.shape.length);for(let m=0;m<a.length;m++)a[m]=o.shape[m]*i[m];let u=new Uint8Array(new Int32Array(o.shape).buffer),l=new Uint8Array(new Int32Array(a).buffer),c=e.makeOutput(a,o.dtype),p=e.dataIdMap.get(c.dataId).id;return DW(s,u,o.shape.length,l,a.length,qt[c.dtype],p),c}var RW={kernelName:Jn,backendName:"wasm",setupFunc:Gat,kernelFunc:Wat};var FW;function Uat(r){FW=r.wasm.cwrap(Ua,null,["number","array","number","number","number","bool","number","number"])}var Hat=({inputs:r,backend:t,attrs:e})=>{let{x:n}=r,{k:o,sorted:s}=e,i=t.dataIdMap.get(n.dataId).id,a=new Uint8Array(new Int32Array(n.shape).buffer),u=n.shape.slice();u[u.length-1]=o;let l=t.makeOutput(u,n.dtype),c=t.dataIdMap.get(l.dataId).id,p=t.makeOutput(u,"int32"),m=t.dataIdMap.get(p.dataId).id;return FW(i,a,n.shape.length,qt[n.dtype],o,s,c,m),[l,p]},OW={kernelName:Ua,backendName:"wasm",setupFunc:Uat,kernelFunc:Hat};var PW;function qat(r){PW=r.wasm.cwrap(Ha,null,["number","number","bool","number","number","number","number","number","number","array","number","array","number","number","number","number","number"])}function Kat(r){let{backend:t,inputs:e,attrs:n}=r,{image:o,transforms:s}=e,{interpolation:i,fillMode:a,fillValue:u,outputShape:l}=n,[c,p,m,f]=o.shape,[d,h]=l!=null?l:[p,m],g=[c,d,h,f],x=new Uint8Array(new Int32Array(y.computeStrides(o.shape)).buffer),b=new Uint8Array(new Int32Array(y.computeStrides(g)).buffer),w=t.makeOutput(g,o.dtype),C=t.dataIdMap.get(w.dataId).id,_=t.dataIdMap.get(o.dataId).id,$=t.dataIdMap.get(s.dataId).id,F=i==="nearest"?1:2,P;switch(a){case"constant":P=1;break;case"reflect":P=2;break;case"wrap":P=3;break;case"nearest":P=4;break;default:P=1;break}return PW(_,$,s.shape[0]>1,c,d,h,f,m,p,x,o.shape.length-1,b,g.length-1,F,P,u,C),w}var LW={kernelName:Ha,backendName:"wasm",setupFunc:qat,kernelFunc:Kat};function jat(r){let{inputs:t,backend:e,attrs:n}=r,{value:o}=t,{axis:s}=n;s<0&&(s+=o.shape.length);let i=o.shape[s],a=o.shape.length,u=new Array(a-1),l=0;for(let f=0;f<a;f++)f!==s&&(u[l++]=o.shape[f]);let c=new Array(i),p=new Array(a).fill(0),m=o.shape.slice();m[s]=1;for(let f=0;f<c.length;f++)p[s]=f,c[f]=Lo({inputs:{x:o},attrs:{begin:p,size:m},backend:e});return c.map(({dataId:f,dtype:d})=>({dataId:f,dtype:d,shape:u}))}var MW={kernelName:bi,backendName:"wasm",kernelFunc:jat};function Xat(r){let{inputs:{x:t},backend:e}=r,n=e.makeOutput(t.shape,t.dtype);return e.typedArrayFromHeap(n).fill(0),n}var zW={kernelName:wi,backendName:"wasm",kernelFunc:Xat};var Yat=[UB,HB,qB,jB,QB,eV,nV,sV,lV,cV,pV,mV,dV,hV,xV,bV,wV,CV,SV,NV,kV,_V,$V,DV,RV,FV,OV,PV,MV,zV,BV,GV,UV,qV,jV,YV,ZV,JV,XB,tG,eG,rG,nG,oG,sG,iG,aG,uG,cG,mG,dG,gG,xG,bG,wG,CG,SG,NG,kG,EG,AG,$G,DG,cI,FG,PG,MG,zG,BG,VG,GG,iV,UG,qG,jG,YG,ZG,JG,tW,rW,oW,sW,uV,aW,lW,cW,mW,dW,hW,gW,xW,yW,bW,CW,SW,vW,NW,TW,kW,_W,AW,$W,RW,OW,LW,ZB,MW,zW];for(let r of Yat)Lu(r);var Rk=z();Rk.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(r){return!1}});Rk.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(Rk.get("IS_NODE"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(r){return!1}});var zk=Tl(WW()),XW=Tl(HW()),Bk=Tl(qW());var KW=zk.default||zk,Zat=Bk.default||Bk,cg=class extends zo{constructor(t){super(),this.wasm=t,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(ZW),Mk=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new ra(this,Pn())}write(t,e,n){let o={id:this.dataIdNextNumber++};return this.move(o,t,e,n,1),o}numDataIds(){return this.dataIdMap.numDataIds()}async time(t){let e=y.now();return t(),{kernelMs:y.now()-e}}move(t,e,n,o,s){let i=this.dataIdNextNumber++;if(o==="string"){let c=e;this.dataIdMap.set(t,{id:i,stringBytes:c,shape:n,dtype:o,memoryOffset:null,refCount:s});return}let a=y.sizeFromShape(n),u=a*y.bytesPerElement(o),l=this.wasm._malloc(u);this.dataIdMap.set(t,{id:i,memoryOffset:l,shape:n,dtype:o,refCount:s}),this.wasm.tfjs.registerTensor(i,a,l),e!=null&&this.wasm.HEAPU8.set(new Uint8Array(e.buffer,e.byteOffset,u),l)}async read(t){return this.readSync(t)}readSync(t,e,n){let{memoryOffset:o,dtype:s,shape:i,stringBytes:a}=this.dataIdMap.get(t);if(s==="string")return(e==null||e===0)&&(n==null||n>=a.length)?a:a.slice(e,n);e=e||0,n=n||y.sizeFromShape(i);let u=y.bytesPerElement(s),l=this.wasm.HEAPU8.slice(o+e*u,o+n*u);return Qat(l.buffer,s)}disposeData(t,e=!1){if(this.dataIdMap.has(t)){let n=this.dataIdMap.get(t);if(n.refCount--,!e&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(t)}return!0}refCount(t){return this.dataIdMap.has(t)?this.dataIdMap.get(t).refCount:0}incRef(t){let e=this.dataIdMap.get(t);e!=null&&e.refCount++}floatPrecision(){return 32}getMemoryOffset(t){return this.dataIdMap.get(t).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(t,e,n){let o;if(n==null)o=this.write(null,t,e);else{let s=this.dataIdNextNumber++;o={id:s},this.dataIdMap.set(o,{id:s,memoryOffset:n,shape:t,dtype:e,refCount:1});let i=y.sizeFromShape(t);this.wasm.tfjs.registerTensor(s,i,n)}return{dataId:o,shape:t,dtype:e}}typedArrayFromHeap({shape:t,dtype:e,dataId:n}){let o=this.wasm.HEAPU8.buffer,{memoryOffset:s}=this.dataIdMap.get(n),i=y.sizeFromShape(t);switch(e){case"float32":return new Float32Array(o,s,i);case"int32":return new Int32Array(o,s,i);case"bool":return new Uint8Array(o,s,i);default:throw new Error(`Unknown dtype ${e}`)}}};function Jat(r){return(t,e)=>(y.fetch(r,{credentials:"same-origin"}).then(n=>{n.ok||t.env.a(`failed to load wasm binary file at '${r}'`),n.arrayBuffer().then(o=>{WebAssembly.instantiate(o,t).then(s=>{e(s.instance,s.module)})})}),{})}function jW(r,t,e){if(hI!=null)return hI;let n="tfjs-backend-wasm.wasm";return r&&t?n="tfjs-backend-wasm-threaded-simd.wasm":r&&(n="tfjs-backend-wasm-simd.wasm"),lg!=null&&lg[n]!=null?lg[n]:e+n}async function YW(){let[r,t]=await Promise.all([z().getAsync("WASM_HAS_SIMD_SUPPORT"),z().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((e,n)=>{let o={};o.locateFile=(a,u)=>{if(a.endsWith(".worker.js")){let l=XW.wasmWorkerContents.replace(/\n/g,"\\n"),c=new Blob([l],{type:"application/javascript"});return URL.createObjectURL(c)}return a.endsWith(".wasm")?jW(r,t,ag!=null?ag:u):u+a},Vk&&(o.instantiateWasm=Jat(jW(r,t,ag!=null?ag:"")));let s=!1;o.onAbort=()=>{if(s||ug)return;ug=!0,n({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"})};let i;t&&r&&hI==null?(o.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+KW.toString()],{type:"text/javascript"}),i=KW(o)):i=Zat(o),i.then(a=>{s=!0,ug=!1;let u=null;a.tfjs={init:a.cwrap("init",null,[]),initWithThreadsCount:a.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:a.cwrap("get_threads_count","number",[]),registerTensor:a.cwrap("register_tensor",null,["number","number","number"]),disposeData:a.cwrap("dispose_data",u,["number"]),dispose:a.cwrap("dispose",u,[])},e({wasm:a})}).catch(n)})}function Qat(r,t){switch(t){case"float32":return new Float32Array(r);case"int32":return new Int32Array(r);case"bool":return new Uint8Array(r);default:throw new Error(`Unknown dtype ${t}`)}}var tlt=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],hI=null,ag=null,lg={},ug=!1,Vk=!1;function elt(r,t=!1){if(W0("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),ug)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");hI=r,Vk=t}function rlt(r,t=!1){if(ug)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if(typeof r=="string")ag=r;else{lg=r;let e=tlt.filter(n=>lg[n]==null);if(e.length>0)throw new Error(`There were no entries found for the following binaries: ${e.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}Vk=t}var ZW=-1,Mk=-1;function nlt(r){ZW=r}function olt(){if(Mk===-1)throw new Error("WASM backend not initialized.");return Mk}var slt="4.0.0";var ilt=2;Xp("wasm",async()=>{let{wasm:r}=await YW();return new cg(r)},ilt);var JW="4.0.0",alt="4.0.0",llt="4.0.0",ult="4.0.0",clt="4.0.0",plt={tfjs:JW,"tfjs-core":JW,"tfjs-converter":alt,"tfjs-backend-cpu":llt,"tfjs-backend-webgl":ult,"tfjs-backend-wasm":clt};export{ii as Abs,oa as Acos,sa as Acosh,cu as AdadeltaOptimizer,pu as AdagradOptimizer,mu as AdamOptimizer,fu as AdamaxOptimizer,Zn as Add,Go as AddN,ia as All,aa as Any,Wo as ArgMax,kl as ArgMin,la as Asin,ua as Asinh,ca as Atan,ma as Atan2,pa as Atanh,Uo as AvgPool,El as AvgPool3D,lp as AvgPool3DGrad,ap as AvgPoolGrad,cg as BackendWasm,Ho as BatchMatMul,ai as BatchToSpaceND,up as Bincount,cp as BroadcastArgs,p1 as BroadcastTo,Sb as Callback,Py as CallbackList,lo as Cast,qo as Ceil,uo as ClipByValue,pp as Complex,_l as ComplexAbs,li as Concat,Ko as Conv2D,mp as Conv2DBackpropFilter,jo as Conv2DBackpropInput,Al as Conv3D,fp as Conv3DBackpropFilterV2,dp as Conv3DBackpropInputV2,Xo as Cos,Yo as Cosh,da as CropAndResize,fa as Cumprod,Zo as Cumsum,My as CustomCallback,ra as DataStorage,hp as DenseBincount,ha as DepthToSpace,Jo as DepthwiseConv2dNative,gp as DepthwiseConv2dNativeBackpropFilter,xp as DepthwiseConv2dNativeBackpropInput,yp as Diag,$l as Dilation2D,Xd as Dilation2DBackpropFilter,jd as Dilation2DBackpropInput,i0 as ENV,vb as EarlyStopping,bp as Einsum,ts as Elu,wp as EluGrad,qd as Environment,xa as Equal,ga as Erf,es as Exp,ui as ExpandDims,ya as Expm1,Cp as FFT,Dl as Fill,ba as FlipLeftRight,rs as Floor,ns as FloorDiv,Yd as FromPixels,os as FusedBatchNorm,Ii as FusedConv2D,Si as FusedDepthwiseConv2D,Bc as GPGPUContext,wa as GatherNd,ci as GatherV2,Ph as GraphModel,Ca as Greater,ss as GreaterEqual,Ly as History,Ip as IFFT,co as Identity,Sp as Imag,ye as InputSpec,Ia as IsFinite,Sa as IsInf,va as IsNan,zo as KernelBackend,Rl as LRN,Np as LRNGrad,Ch as LayerVariable,Bn as LayersModel,is as LeakyRelu,Na as Less,Ta as LessEqual,vp as LinSpace,as as Log,ka as Log1p,f1 as LogSoftmax,Ea as LogicalAnd,_a as LogicalNot,Aa as LogicalOr,m1 as LogicalXor,xlt as LowerBound,_u as MathBackendWebGL,ls as Max,cs as MaxPool,Fl as MaxPool3D,kp as MaxPool3DGrad,Tp as MaxPoolGrad,Ep as MaxPoolWithArgmax,us as Maximum,ps as Mean,ms as Min,fs as Minimum,ds as MirrorPad,$a as Mod,du as MomentumOptimizer,_p as Multinomial,hs as Multiply,pi as Neg,Ra as NonMaxSuppressionV3,Fa as NonMaxSuppressionV4,Oa as NonMaxSuppressionV5,Da as NotEqual,k0 as OP_SCOPE_SUFFIX,gs as OneHot,mi as OnesLike,Wr as Optimizer,Ws as OptimizerConstructors,fi as Pack,xs as PadV2,ylt as Pool,ys as Pow,bs as Prelu,ws as Prod,hu as RMSPropOptimizer,Tn as RNN,Ap as RaggedGather,$p as RaggedRange,Dp as RaggedTensorToTensor,Ol as Range,x0 as Rank,Rp as Real,Qo as RealDiv,Pa as Reciprocal,Xe as Reduction,Cs as Relu,vs as Relu6,di as Reshape,Ss as ResizeBilinear,Op as ResizeBilinearGrad,Is as ResizeNearestNeighbor,Fp as ResizeNearestNeighborGrad,Ns as Reverse,qa as RotateWithOffset,Ts as Round,ks as Rsqrt,Bi as SGDOptimizer,La as ScatterNd,Pp as SearchSorted,hi as Select,Ma as Selu,qi as Sequential,_s as Sigmoid,Ba as Sign,Es as Sin,za as Sinh,gi as Slice,Ds as Softmax,Va as Softplus,xi as SpaceToBatchND,Pl as SparseFillEmptyRows,Ga as SparseReshape,Ll as SparseSegmentMean,Ml as SparseSegmentSum,Lp as SparseToDense,yi as SplitV,As as Sqrt,zl as Square,Rs as SquaredDifference,po as Step,Wa as StridedSlice,Bl as StringNGrams,Vl as StringSplit,Gl as StringToHashBucketFast,Fs as Sub,$s as Sum,Jr as SymbolicTensor,Os as Tan,Ps as Tanh,Ft as Tensor,pe as TensorBuffer,Jn as Tile,Ua as TopK,Ha as Transform,Qn as Transpose,Mp as Unique,bi as Unpack,Wl as UnsortedSegmentSum,blt as UpperBound,Ka as Variable,wi as ZerosLike,Ci as _FusedMatMul,Ee as abs,ax as acos,lx as acosh,X as add,LE as addN,Zp as all,qu as any,Ai as argMax,ux as argMin,cx as asin,px as asinh,mx as atan,fx as atan2,dx as atanh,Yl as avgPool,gx as avgPool3d,gE as backend,v as backend_util,BE as basicLSTMCell,Di as batchNorm,xx as batchNorm2d,yx as batchNorm3d,bx as batchNorm4d,Zl as batchToSpaceND,wx as bincount,n6 as booleanMaskAsync,GE as broadcastArgs,Ri as broadcastTo,Vr as broadcast_util,nx as browser,wt as buffer,VZ as callbacks,J as cast,Cx as ceil,Cr as clipByValue,sn as clone,wn as complex,ne as concat,Ix as concat1d,Sx as concat2d,vx as concat3d,Nx as concat4d,K$ as constraints,Qp as conv1d,In as conv2d,em as conv2dTranspose,Tx as conv3d,Ex as conv3dTranspose,Tlt as copyRegisteredKernels,Jl as cos,rm as cosh,hh as cosineWindow,Xu as cumprod,nm as cumsum,un as customGrad,AR as data,ch as denseBincount,W0 as deprecationWarn,_x as depthToSpace,Fi as depthwiseConv2d,HZ as deregisterOp,Kl as device_util,WE as diag,Ax as dilation2d,gpt as disableDeprecationWarnings,vt as dispose,xpt as disposeVariables,pt as div,$x as divNoNan,Dx as dot,lv as dropout,UE as einsum,Oi as elu,hpt as enableDebugMode,dpt as enableProdMode,uv as enclosingPowerOfTwo,Pn as engine,z as env,$r as equal,Rx as erf,Fx as euclideanNorm,er as exp,rr as expandDims,Ox as expm1,Yu as eye,au as fft,xo as fill,Spt as findBackend,vpt as findBackendFactory,Pi as floor,Yp as floorDiv,hM as forceHalfFloat,uu as fused,Li as gather,m6 as gatherND,ox as gather_util,Cpt as getBackend,u0 as getGradient,Jd as getKernel,zg as getKernelsForBackend,olt as getThreadsCount,ik as gpgpu_util,bK as grad,wK as grads,Re as greater,ln as greaterEqual,tl as ifft,Xl as imag,Gs as image,h6 as inTopKAsync,j$ as initializers,Pv as input,_r as io,xm as irfft,Px as isFinite,Lx as isInf,Mx as isNaN,De as keep,Ur as kernel_impls,ED as layers,Ql as leakyRelu,om as less,Ln as lessEqual,pv as linalg,KE as linspace,M7 as loadGraphModel,z7 as loadGraphModelSync,hD as loadLayersModel,zx as localResponseNormalization,Sr as log,tu as log1p,Gx as logSigmoid,sm as logSoftmax,im as logSumExp,Rr as logicalAnd,eu as logicalNot,am as logicalOr,Wx as logicalXor,hX as losses,jE as lowerBound,Lt as matMul,yE as math,Ir as max,ru as maxPool,Hx as maxPool3d,XE as maxPoolWithArgmax,Sn as maximum,ve as mean,ah as memory,YE as meshgrid,_D as metrics,Ja as min,Mi as minimum,qx as mirrorPad,Kx as mod,q8 as model,AD as models,Zu as moments,s6 as movingAverage,D as mul,ZE as multiRNNCell,JE as multinomial,Ht as neg,gh as nextFrame,Qa as norm,Bs as notEqual,Ei as oneHot,cr as ones,yr as onesLike,T as op,QE as outerProduct,cn as pad,t_ as pad1d,e_ as pad2d,r_ as pad3d,n_ as pad4d,jx as pool,an as pow,ou as prelu,Jg as print,Xx as prod,ypt as profile,o_ as raggedGather,s_ as raggedRange,i_ as raggedTensorToTensor,a_ as rand,v_ as randomGamma,tc as randomNormal,N_ as randomStandardNormal,zi as randomUniform,su as range,wpt as ready,Za as real,ty as reciprocal,Xp as registerBackend,j8 as registerCallbackConstructor,h1 as registerGradient,Lu as registerKernel,UZ as registerOp,$D as regularizers,Fr as relu,lm as relu6,Ipt as removeBackend,R as reshape,pr as reverse,T_ as reverse1d,k_ as reverse2d,E_ as reverse3d,__ as reverse4d,lu as rfft,um as round,cm as rsqrt,mt as scalar,a6 as scatterND,lh as scatter_util,mh as searchSorted,pm as selu,mm as separableConv2d,K8 as sequential,Q as serialization,tH as setBackend,Npt as setPlatform,nlt as setThreadsCount,elt as setWasmPath,rlt as setWasmPaths,wT as setWebGLContext,A_ as setdiff1dAsync,Yr as sigmoid,ey as sign,dX as signal,fm as sin,dm as sinh,Rt as slice,hm as slice1d,dh as slice2d,gm as slice3d,ec as slice4d,Le as slice_util,iu as softmax,zs as softplus,nu as spaceToBatchND,gX as sparse,c6 as sparseToDense,fX as spectral,mr as split,Se as sqrt,Mt as square,ym as squaredDifference,Mn as squeeze,nr as stack,bo as step,ry as stridedSlice,xX as string,ct as sub,ft as sum,Wu as sumOutType,ny as tan,$i as tanh,ur as tensor,Me as tensor1d,Vs as tensor2d,rx as tensor3d,$_ as tensor4d,D_ as tensor5d,R_ as tensor6d,go as tensor_util,OE as test_util,B as tidy,Dr as tile,bpt as time,oy as topk,ic as train,Ot as transpose,bm as truncatedNormal,sy as unique,Nlt as unregisterGradient,vlt as unregisterKernel,wm as unsortedSegmentSum,vr as unstack,sr as upcastType,F_ as upperBound,y as util,CK as valueAndGrad,IK as valueAndGrads,iy as variable,Bx as variableGrads,plt as version,cR as version_converter,PE as version_core,Um as version_layers,slt as version_wasm,dM as version_webgl,Zke as webgl,dd as webgl_util,_e as where,ly as whereAsync,Ne as zeros,It as zerosLike};
+    `;
+  }
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/kernels/UnsortedSegmentSum.js
+function unsortedSegmentSum3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x, segmentIds } = inputs;
+  const { numSegments } = attrs;
+  const xRank = x.shape.length;
+  const toDispose = [];
+  let axis = 0;
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation != null) {
+    permutedX = transpose3({ inputs: { x }, backend: backend2, attrs: { perm: permutation } });
+    toDispose.push(permutedX);
+    axis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  }
+  const outShape = backend_util_exports.segment_util.computeOutShape(permutedX.shape, axis, numSegments);
+  const inSize = util_exports.sizeFromShape([permutedX.shape[axis]]);
+  const a2D = reshape4({ inputs: { x: permutedX }, backend: backend2, attrs: { shape: [-1, inSize] } });
+  toDispose.push(a2D);
+  const outputDType = sumOutType(x.dtype);
+  const segOpCompute = (x2, segOpType, segmentIds2, dtype, numSegments2) => {
+    const batchSize = x2.shape[0];
+    const inSize2 = x2.shape[1];
+    const windowSize = backend_util_exports.segment_util.segOpComputeOptimalWindowSize(inSize2, numSegments2);
+    const segOpInfo = { windowSize, inSize: inSize2, batchSize, numSegments: numSegments2 };
+    const program = new SegmentOpProgram(segOpInfo, segOpType);
+    const output = backend2.compileAndRun(program, [x2, segmentIds2], dtype);
+    toDispose.push(output);
+    if (output.shape[1] === numSegments2) {
+      return output;
+    }
+    const rangeInfo = range4({
+      backend: backend2,
+      attrs: { start: 0, stop: numSegments2, step: 1, dtype: "float32" }
+    });
+    const tileInfo = tile4({
+      inputs: { x: rangeInfo },
+      backend: backend2,
+      attrs: { reps: [inSize2 / windowSize] }
+    });
+    toDispose.push(rangeInfo);
+    toDispose.push(tileInfo);
+    const result2 = segOpCompute(output, segOpType, tileInfo, dtype, numSegments2);
+    return result2;
+  };
+  const segOpResult = segOpCompute(a2D, "unsortedSegmentSum", segmentIds, outputDType, numSegments);
+  const reshaped = reshape4({ inputs: { x: segOpResult }, backend: backend2, attrs: { shape: outShape } });
+  let result = reshaped;
+  if (permutation != null) {
+    toDispose.push(reshaped);
+    const perm = backend_util_exports.getUndoAxesPermutation(permutation);
+    result = transpose3({ inputs: { x: result }, backend: backend2, attrs: { perm } });
+  }
+  toDispose.forEach((t) => backend2.disposeIntermediateTensorInfo(t));
+  return result;
+}
+var unsortedSegmentSumConfig2 = {
+  kernelName: UnsortedSegmentSum,
+  backendName: "webgl",
+  kernelFunc: unsortedSegmentSum3
+};
+
+// node_modules/@tensorflow/tfjs-backend-webgl/dist/register_all_kernels.js
+var kernelConfigs2 = [
+  _fusedMatMulConfig2,
+  absConfig2,
+  acosConfig2,
+  acoshConfig2,
+  addConfig2,
+  addNConfig2,
+  allConfig2,
+  anyConfig2,
+  argMaxConfig2,
+  argMinConfig2,
+  asinConfig2,
+  asinhConfig2,
+  atanConfig2,
+  atan2Config2,
+  atanhConfig2,
+  avgPoolConfig2,
+  avgPool3DConfig2,
+  avgPool3DGradConfig3,
+  avgPoolGradConfig3,
+  batchMatMulConfig2,
+  batchNormConfig2,
+  batchToSpaceNDConfig2,
+  bincountConfig2,
+  broadcastArgsConfig2,
+  castConfig2,
+  ceilConfig2,
+  clipByValueConfig2,
+  complexConfig2,
+  complexAbsConfig2,
+  concatConfig2,
+  conv2DConfig2,
+  conv2DBackpropFilterConfig2,
+  conv2DBackpropInputConfig2,
+  conv3DConfig2,
+  conv3DBackpropFilterV2Config2,
+  conv3DBackpropInputConfig,
+  cosConfig2,
+  coshConfig2,
+  cropAndResizeConfig2,
+  cumprodConfig2,
+  cumsumConfig2,
+  denseBincountConfig2,
+  depthToSpaceConfig2,
+  depthwiseConv2dNativeConfig2,
+  depthwiseConv2dNativeBackpropFilterConfig2,
+  depthwiseConv2dNativeBackpropInputConfig2,
+  diagConfig2,
+  dilation2DConfig2,
+  einsumConfig2,
+  eluConfig2,
+  eluGradConfig3,
+  equalConfig2,
+  erfConfig2,
+  expConfig2,
+  expandDimsConfig2,
+  expm1Config2,
+  fftConfig2,
+  fillConfig2,
+  flipLeftRightConfig2,
+  floorConfig2,
+  floorDivConfig2,
+  fromPixelsConfig,
+  fusedConv2DConfig2,
+  fusedDepthwiseConv2DConfig2,
+  gatherNdConfig2,
+  gatherV2Config2,
+  greaterConfig2,
+  greaterEqualConfig2,
+  identityConfig2,
+  ifftConfig2,
+  imagConfig2,
+  isFiniteConfig2,
+  isInfConfig2,
+  isNaNConfig2,
+  leakyReluConfig2,
+  lessConfig2,
+  lessEqualConfig2,
+  linSpaceConfig2,
+  logConfig2,
+  log1pConfig2,
+  logicalAndConfig2,
+  logicalNotConfig2,
+  logicalOrConfig2,
+  LRNConfig2,
+  LRNGradConfig2,
+  maxConfig2,
+  maximumConfig2,
+  maxPoolConfig2,
+  maxPool3DConfig2,
+  maxPool3DGradConfig3,
+  maxPoolGradConfig3,
+  maxPoolWithArgmaxConfig2,
+  meanConfig2,
+  minConfig2,
+  minimumConfig2,
+  mirrorPadConfig2,
+  modConfig2,
+  multinomialConfig2,
+  multiplyConfig2,
+  negConfig2,
+  nonMaxSuppressionV3Config2,
+  nonMaxSuppressionV4Config2,
+  nonMaxSuppressionV5Config2,
+  notEqualConfig2,
+  oneHotConfig2,
+  onesLikeConfig2,
+  packConfig2,
+  padV2Config2,
+  powConfig2,
+  preluConfig2,
+  prodConfig2,
+  raggedGatherConfig2,
+  raggedRangeConfig2,
+  raggedTensorToTensorConfig2,
+  rangeConfig2,
+  realConfig2,
+  realDivConfig2,
+  reciprocalConfig2,
+  reluConfig2,
+  relu6Config2,
+  reshapeConfig2,
+  resizeBilinearConfig2,
+  resizeBilinearGradConfig3,
+  resizeNearestNeighborConfig2,
+  resizeNearestNeighborGradConfig3,
+  reverseConfig2,
+  rotateWithOffsetConfig2,
+  roundConfig2,
+  rsqrtConfig2,
+  scatterNdConfig2,
+  searchSortedConfig2,
+  selectConfig2,
+  seluConfig2,
+  sigmoidConfig2,
+  signConfig2,
+  sinConfig2,
+  sinhConfig2,
+  sliceConfig2,
+  softmaxConfig2,
+  softplusConfig2,
+  spaceToBatchNDConfig2,
+  sparseFillEmptyRowsConfig2,
+  sparseReshapeConfig2,
+  sparseSegmentMeanConfig2,
+  sparseSegmentSumConfig2,
+  sparseToDenseConfig2,
+  splitVConfig2,
+  sqrtConfig2,
+  squareConfig2,
+  squaredDifferenceConfig2,
+  stepConfig2,
+  stridedSliceConfig2,
+  stringNGramsConfig2,
+  stringSplitConfig2,
+  stringToHashBucketFastConfig2,
+  subConfig2,
+  sumConfig2,
+  tanConfig2,
+  tanhConfig2,
+  tileConfig2,
+  topKConfig2,
+  transformConfig2,
+  transposeConfig2,
+  uniqueConfig2,
+  unpackConfig2,
+  unsortedSegmentSumConfig2,
+  zerosLikeConfig2
+];
+for (const kernelConfig of kernelConfigs2) {
+  registerKernel(kernelConfig);
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/types.js
+var CppDType;
+(function(CppDType2) {
+  CppDType2[CppDType2["float32"] = 0] = "float32";
+  CppDType2[CppDType2["int32"] = 1] = "int32";
+  CppDType2[CppDType2["bool"] = 2] = "bool";
+  CppDType2[CppDType2["string"] = 3] = "string";
+  CppDType2[CppDType2["complex64"] = 4] = "complex64";
+})(CppDType || (CppDType = {}));
+var FusableActivation;
+(function(FusableActivation2) {
+  FusableActivation2[FusableActivation2["linear"] = 0] = "linear";
+  FusableActivation2[FusableActivation2["relu"] = 1] = "relu";
+  FusableActivation2[FusableActivation2["relu6"] = 2] = "relu6";
+  FusableActivation2[FusableActivation2["prelu"] = 3] = "prelu";
+  FusableActivation2[FusableActivation2["leakyrelu"] = 4] = "leakyrelu";
+  FusableActivation2[FusableActivation2["sigmoid"] = 5] = "sigmoid";
+  FusableActivation2[FusableActivation2["elu"] = 6] = "elu";
+})(FusableActivation || (FusableActivation = {}));
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/_FusedMatMul.js
+var wasmFusedMatMul;
+function setup(backend2) {
+  wasmFusedMatMul = backend2.wasm.cwrap(_FusedMatMul, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedBatchMatMul(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b, bias, preluActivationWeights } = inputs;
+  if (a.dtype !== "float32" || b.dtype !== "float32") {
+    throw new Error(`_FusedMatMul for non non-float32 tensors not yet supported.`);
+  }
+  const { transposeA, transposeB, activation: activation2, leakyreluAlpha } = attrs;
+  const aId = backend2.dataIdMap.get(a.dataId).id;
+  const bId = backend2.dataIdMap.get(b.dataId).id;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    biasId = biasData.id;
+  }
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);
+  }
+  const leftDim = transposeA ? a.shape[2] : a.shape[1];
+  const rightDim = transposeB ? b.shape[1] : b.shape[2];
+  const batchDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const out = backend2.makeOutput([...batchDims, leftDim, rightDim], a.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);
+  const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);
+  wasmFusedMatMul(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, transposeA, transposeB, fusedActivation, biasId, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var _fusedMatMulConfig3 = {
+  kernelName: _FusedMatMul,
+  backendName: "wasm",
+  setupFunc: setup,
+  kernelFunc: fusedBatchMatMul
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/unary_kernel.js
+function createUnaryKernelConfig(kernelName, outType) {
+  let wasmFunc9;
+  function setupFunc3(backend2) {
+    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [
+      "number",
+      "number",
+      "number"
+    ]);
+  }
+  function kernelFunc3(args) {
+    const { backend: backend2, inputs: { x } } = args;
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const out = backend2.makeOutput(x.shape, outType || x.dtype);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    if (util_exports.sizeFromShape(out.shape) === 0) {
+      return out;
+    }
+    wasmFunc9(xId, CppDType[x.dtype], outId);
+    return out;
+  }
+  return { kernelName, backendName: "wasm", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Abs.js
+var absConfig3 = createUnaryKernelConfig(Abs);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/binary_kernel.js
+function createBinaryKernelConfig(kernelName, supportsFullBroadcast19, dtype) {
+  let wasmFunc9;
+  function setupFunc3(backend2) {
+    wasmFunc9 = backend2.wasm.cwrap(kernelName, null, [
+      "number",
+      "array",
+      "number",
+      "number",
+      "array",
+      "number",
+      "number",
+      "number"
+    ]);
+  }
+  function kernelFunc3(args) {
+    const { backend: backend2, inputs } = args;
+    const { a, b } = inputs;
+    const aId = backend2.dataIdMap.get(a.dataId).id;
+    const bId = backend2.dataIdMap.get(b.dataId).id;
+    const outputType = dtype != null ? dtype : a.dtype;
+    const newShape = backend_util_exports.assertAndGetBroadcastShape(a.shape, b.shape);
+    const out = backend2.makeOutput(newShape, outputType);
+    if (util_exports.sizeFromShape(newShape) === 0) {
+      return out;
+    }
+    const aShapeBytes = new Uint8Array(new Int32Array(a.shape).buffer);
+    const bShapeBytes = new Uint8Array(new Int32Array(b.shape).buffer);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    const kernelFunc4 = () => wasmFunc9(aId, aShapeBytes, a.shape.length, bId, bShapeBytes, b.shape.length, CppDType[a.dtype], outId);
+    kernelFunc4();
+    return out;
+  }
+  return { kernelName, backendName: "wasm", setupFunc: setupFunc3, kernelFunc: kernelFunc3 };
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Add.js
+var supportsFullBroadcast = true;
+var addConfig3 = createBinaryKernelConfig(Add, supportsFullBroadcast);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/AddN.js
+var wasmFunc;
+function setupFunc(backend2) {
+  wasmFunc = backend2.wasm.cwrap(AddN, null, [
+    "array",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function addn(args) {
+  const { inputs, backend: backend2 } = args;
+  const out = backend2.makeOutput(inputs[0].shape, inputs[0].dtype);
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  const inputIds = inputs.map((x) => backend2.dataIdMap.get(x.dataId).id);
+  const inputIdsBytes = new Uint8Array(new Int32Array(inputIds).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmFunc(inputIdsBytes, inputIds.length, CppDType[out.dtype], outId);
+  return out;
+}
+var addNConfig3 = {
+  kernelName: AddN,
+  backendName: "wasm",
+  setupFunc,
+  kernelFunc: addn
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Identity.js
+function identity4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  if (x.dtype === "string") {
+    return tensor(backend2.readSync(x.dataId), x.shape, x.dtype);
+  }
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const inVals = backend2.typedArrayFromHeap(x);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(inVals);
+  return out;
+}
+var identityConfig3 = {
+  kernelName: Identity,
+  backendName: "wasm",
+  kernelFunc: identity4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Transpose.js
+var wasmTranspose;
+function setup2(backend2) {
+  wasmTranspose = backend2.wasm.cwrap(Transpose, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function transpose4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const [reducedShape, perm] = removeOneSizeDims(inputs.x.shape, attrs.perm);
+  let permIsNoOp = true;
+  for (let i = 0; i < perm.length; i++) {
+    if (perm[i] !== i) {
+      permIsNoOp = false;
+    }
+  }
+  const outShape = computeOutShape4(inputs.x.shape, attrs.perm);
+  const x = {
+    dataId: inputs.x.dataId,
+    shape: reducedShape,
+    dtype: inputs.x.dtype
+  };
+  if (permIsNoOp) {
+    const cloned = identity4({ inputs, backend: backend2 });
+    cloned.shape = outShape;
+    return cloned;
+  }
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const permBytes = new Uint8Array(new Int32Array(perm).buffer);
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  wasmTranspose(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], outId, permBytes, perm.length);
+  return out;
+}
+function computeOutShape4(inShape, perm) {
+  const outShape = new Array(inShape.length);
+  for (let i = 0; i < outShape.length; i++) {
+    outShape[i] = inShape[perm[i]];
+  }
+  return outShape;
+}
+function removeOneSizeDims(shape, perm) {
+  const newShape = [];
+  const newPerm = [];
+  for (let i = 0; i < shape.length; ++i) {
+    if (shape[i] !== 1) {
+      newShape.push(shape[i]);
+    }
+    if (shape[perm[i]] !== 1) {
+      newPerm.push(perm[i]);
+    }
+  }
+  for (let i = 0; i < newPerm.length; ++i) {
+    let minValIdx = -1;
+    for (let j = 0; j < newPerm.length; ++j) {
+      if (newPerm[j] >= i && (minValIdx === -1 || newPerm[minValIdx] > newPerm[j])) {
+        minValIdx = j;
+      }
+    }
+    newPerm[minValIdx] = i;
+  }
+  return [newShape, newPerm];
+}
+var transposeConfig3 = {
+  kernelName: Transpose,
+  backendName: "wasm",
+  kernelFunc: transpose4,
+  setupFunc: setup2
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/kernel_utils.js
+function permuteAxesAndTranspose(x, axis, backend2) {
+  const xShape = x.shape;
+  const xRank = x.shape.length;
+  const originalAxes = util_exports.parseAxisParam(axis, xShape);
+  let axes = originalAxes;
+  const permutedAxes = backend_util_exports.getAxesPermutation(axes, xRank);
+  let xTransposed = null;
+  let inputWasTransposed = false;
+  if (permutedAxes != null) {
+    const newShape = new Array(xRank);
+    for (let i = 0; i < newShape.length; i++) {
+      newShape[i] = xShape[permutedAxes[i]];
+    }
+    axes = backend_util_exports.getInnerMostAxes(axes.length, xRank);
+    xTransposed = transpose4({ inputs: { x }, attrs: { perm: permutedAxes }, backend: backend2 });
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const transposedId = backend2.dataIdMap.get(xTransposed.dataId).id;
+    if (transposedId !== xId) {
+      inputWasTransposed = true;
+    }
+  }
+  return { transposed: xTransposed, originalAxes, axes, inputWasTransposed };
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/All.js
+var wasmAll;
+function setup3(backend2) {
+  wasmAll = backend2.wasm.cwrap(All, null, ["number, number, number"]);
+}
+function all4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("all", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmAll(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var allConfig3 = {
+  kernelName: All,
+  backendName: "wasm",
+  setupFunc: setup3,
+  kernelFunc: all4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Any.js
+var wasmAny;
+function setup4(backend2) {
+  wasmAny = backend2.wasm.cwrap(Any, null, ["number, number, number"]);
+}
+function any4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("any", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmAny(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var anyConfig3 = {
+  kernelName: Any,
+  backendName: "wasm",
+  setupFunc: setup4,
+  kernelFunc: any4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ArgMax.js
+var wasmFunc2;
+function setup5(backend2) {
+  wasmFunc2 = backend2.wasm.cwrap(ArgMax, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function argmax(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+    }
+  }
+  const outShape = input2.shape.slice(0, -1);
+  const out = backend2.makeOutput(outShape, "int32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const outerSize = util_exports.sizeFromShape(out.shape);
+  const innerSize = input2.shape[axes[0]];
+  wasmFunc2(inputId, CppDType[input2.dtype], outerSize, innerSize, outId);
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  return out;
+}
+var argMaxConfig3 = {
+  kernelName: ArgMax,
+  backendName: "wasm",
+  kernelFunc: argmax,
+  setupFunc: setup5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/AvgPool.js
+var wasmAvgPool;
+function setup6(backend2) {
+  wasmAvgPool = backend2.wasm.cwrap(AvgPool, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function avgPool4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const x = inputs.x;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const channels = convInfo.inChannels;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  if (convInfo.dilationWidth !== 1 || convInfo.dilationHeight !== 1) {
+    throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${convInfo.dilationHeight}, ${convInfo.dilationWidth}].`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmAvgPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, strideHeight, strideWidth, channels, outId);
+  return out;
+}
+var avgPoolConfig3 = {
+  kernelName: AvgPool,
+  backendName: "wasm",
+  setupFunc: setup6,
+  kernelFunc: avgPool4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Reshape.js
+function reshape5(args) {
+  const { inputs, attrs } = args;
+  const { x } = inputs;
+  const { shape } = attrs;
+  const xSize = util_exports.sizeFromShape(x.shape);
+  const $shape = util_exports.inferFromImplicitShape(shape, xSize);
+  util_exports.assert(xSize === util_exports.sizeFromShape($shape), () => `new shape: ${$shape}, old shape: ${x.shape}. New shape and old shape must have the same number of elements.`);
+  args.backend.incRef(x.dataId);
+  return { dataId: x.dataId, shape: $shape, dtype: x.dtype };
+}
+var reshapeConfig3 = {
+  kernelName: Reshape,
+  backendName: "wasm",
+  kernelFunc: reshape5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/BatchMatMul.js
+var wasmBatchMatMul;
+function setup7(backend2) {
+  wasmBatchMatMul = backend2.wasm.cwrap(BatchMatMul, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function batchMatMul3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { a, b } = inputs;
+  const { transposeA, transposeB } = attrs;
+  if (a.dtype !== "float32" || b.dtype !== "float32") {
+    throw new Error(`BatchMatMul for non non-float32 tensors not yet supported.`);
+  }
+  const aRank = a.shape.length;
+  const bRank = b.shape.length;
+  const innerShapeA = transposeA ? a.shape[aRank - 2] : a.shape[aRank - 1];
+  const innerShapeB = transposeB ? b.shape[bRank - 1] : b.shape[bRank - 2];
+  const outerShapeA = transposeA ? a.shape[aRank - 1] : a.shape[aRank - 2];
+  const outerShapeB = transposeB ? b.shape[bRank - 2] : b.shape[bRank - 1];
+  const outerDimsA = a.shape.slice(0, -2);
+  const outerDimsB = b.shape.slice(0, -2);
+  const batchDimA = util_exports.sizeFromShape(outerDimsA);
+  const batchDimB = util_exports.sizeFromShape(outerDimsB);
+  const outShapeOuterDims = broadcast_util_exports.assertAndGetBroadcastShape(a.shape.slice(0, -2), b.shape.slice(0, -2));
+  const outShape = outShapeOuterDims.concat([outerShapeA, outerShapeB]);
+  util_exports.assert(innerShapeA === innerShapeB, () => `Error in matMul: inner shapes (${innerShapeA}) and (${innerShapeB}) of Tensors with shapes ${a.shape} and ${b.shape} and transposeA=${transposeA} and transposeB=${transposeB} must match.`);
+  const a3dShape = transposeA ? [batchDimA, innerShapeA, outerShapeA] : [batchDimA, outerShapeA, innerShapeA];
+  const b3dShape = transposeB ? [batchDimB, outerShapeB, innerShapeB] : [batchDimB, innerShapeB, outerShapeB];
+  const a3d = reshape5({ inputs: { x: a }, backend: backend2, attrs: { shape: a3dShape } });
+  const b3d = reshape5({ inputs: { x: b }, backend: backend2, attrs: { shape: b3dShape } });
+  const a3dId = backend2.dataIdMap.get(a3d.dataId).id;
+  const b3dId = backend2.dataIdMap.get(b3d.dataId).id;
+  const leftDim = transposeA ? a3d.shape[2] : a3d.shape[1];
+  const rightDim = transposeB ? b3d.shape[1] : b3d.shape[2];
+  const batchDim = Math.max(batchDimA, batchDimB);
+  const out = backend2.makeOutput([batchDim, leftDim, rightDim], a3d.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const aShapeBytes = new Uint8Array(new Int32Array(a3d.shape).buffer);
+  const bShapeBytes = new Uint8Array(new Int32Array(b3d.shape).buffer);
+  wasmBatchMatMul(a3dId, aShapeBytes, a3d.shape.length, b3dId, bShapeBytes, b3d.shape.length, transposeA, transposeB, outId);
+  backend2.disposeData(a3d.dataId);
+  backend2.disposeData(b3d.dataId);
+  out.shape = outShape;
+  return out;
+}
+var batchMatMulConfig3 = {
+  kernelName: BatchMatMul,
+  backendName: "wasm",
+  setupFunc: setup7,
+  kernelFunc: batchMatMul3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Slice.js
+function slice4(args) {
+  const { inputs: { x }, attrs: { begin, size }, backend: backend2 } = args;
+  const [begin_, size_] = slice_util_exports.parseSliceParams(x, begin, size);
+  const isContinous = slice_util_exports.isSliceContinous(x.shape, begin_, size_);
+  const xVals = backend2.readSync(x.dataId);
+  const out = backend2.makeOutput(size_, x.dtype);
+  const xStrides = util_exports.computeStrides(x.shape);
+  const outData = backend2.dataIdMap.get(out.dataId);
+  if (isContinous) {
+    const flatOffset = slice_util_exports.computeFlatOffset(begin_, xStrides);
+    if (x.dtype === "string") {
+      outData.stringBytes = xVals.slice(flatOffset, flatOffset + util_exports.sizeFromShape(size_));
+    } else {
+      const outVals2 = backend2.typedArrayFromHeap(out);
+      outVals2.set(xVals.subarray(flatOffset, flatOffset + util_exports.sizeFromShape(size_)));
+    }
+    return out;
+  }
+  if (x.dtype === "string") {
+    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);
+    outData.stringBytes = res;
+    return out;
+  }
+  const outVals = backend2.typedArrayFromHeap(out);
+  const rank = x.shape.length;
+  if (rank === 2) {
+    slice2d2(xVals, xStrides[0], outVals, begin_, size_);
+  } else if (rank === 3) {
+    slice3d2(xVals, xStrides[0], xStrides[1], outVals, begin_, size_);
+  } else if (rank === 4) {
+    slice4d2(xVals, xStrides[0], xStrides[1], xStrides[2], outVals, begin_, size_);
+  } else {
+    const res = sliceImpl(xVals, begin_, size_, x.shape, x.dtype);
+    outVals.set(res);
+  }
+  return out;
+}
+function slice2d2(xVals, xStride, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const endI = beginI + size[0];
+  for (let i = beginI; i < endI; i++) {
+    const xOffset = i * xStride + beginJ;
+    outVals.set(xVals.subarray(xOffset, xOffset + size[1]), outOffset);
+    outOffset += size[1];
+  }
+}
+function slice3d2(xVals, xStride1, xStride2, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const beginK = begin[2];
+  const endI = beginI + size[0];
+  const endJ = beginJ + size[1];
+  for (let i = beginI; i < endI; i++) {
+    for (let j = beginJ; j < endJ; j++) {
+      const xOffset = i * xStride1 + j * xStride2 + beginK;
+      outVals.set(xVals.subarray(xOffset, xOffset + size[2]), outOffset);
+      outOffset += size[2];
+    }
+  }
+}
+function slice4d2(xVals, xStride1, xStride2, xStride3, outVals, begin, size) {
+  let outOffset = 0;
+  const beginI = begin[0];
+  const beginJ = begin[1];
+  const beginK = begin[2];
+  const endI = beginI + size[0];
+  const endJ = beginJ + size[1];
+  const endK = beginK + size[2];
+  const beginL = begin[3];
+  for (let i = beginI; i < endI; i++) {
+    for (let j = beginJ; j < endJ; j++) {
+      for (let k = beginK; k < endK; k++) {
+        const xOffset = i * xStride1 + j * xStride2 + k * xStride3 + beginL;
+        outVals.set(xVals.subarray(xOffset, xOffset + size[3]), outOffset);
+        outOffset += size[3];
+      }
+    }
+  }
+}
+var sliceConfig3 = {
+  kernelName: Slice,
+  backendName: "wasm",
+  kernelFunc: slice4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/BatchToSpaceND.js
+function batchToSpaceND4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, crops } = attrs;
+  const prod5 = blockShape.reduce((a, b) => a * b);
+  const reshaped = backend_util_exports.getReshaped(x.shape, blockShape, prod5);
+  const permuted = backend_util_exports.getPermuted(reshaped.length, blockShape.length);
+  const reshapedPermuted = backend_util_exports.getReshapedPermuted(x.shape, blockShape, prod5);
+  const sliceBeginCoords = backend_util_exports.getSliceBeginCoords(crops, blockShape.length);
+  const sliceSize = backend_util_exports.getSliceSize(reshapedPermuted, crops, blockShape.length);
+  const xReshaped = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: reshaped } });
+  const xTransposed = transpose4({ inputs: { x: xReshaped }, backend: backend2, attrs: { perm: permuted } });
+  const xTransposedReshaped = reshape5({ inputs: { x: xTransposed }, backend: backend2, attrs: { shape: reshapedPermuted } });
+  const result = slice4({
+    inputs: { x: xTransposedReshaped },
+    backend: backend2,
+    attrs: { begin: sliceBeginCoords, size: sliceSize }
+  });
+  backend2.disposeData(xReshaped.dataId);
+  backend2.disposeData(xTransposed.dataId);
+  backend2.disposeData(xReshaped.dataId);
+  return result;
+}
+var batchToSpaceNDConfig3 = {
+  kernelName: BatchToSpaceND,
+  backendName: "wasm",
+  kernelFunc: batchToSpaceND4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cast.js
+function cast5(args) {
+  const { inputs: { x }, attrs: { dtype }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, dtype);
+  const inVals = backend2.typedArrayFromHeap(x);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(inVals);
+  return out;
+}
+var castConfig3 = {
+  kernelName: Cast,
+  backendName: "wasm",
+  kernelFunc: cast5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Ceil.js
+var ceilConfig3 = createUnaryKernelConfig(Ceil);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ClipByValue.js
+var wasmClip;
+function setup8(backend2) {
+  wasmClip = backend2.wasm.cwrap(ClipByValue, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function clip(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { clipValueMin, clipValueMax } = attrs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmClip(xId, clipValueMin, clipValueMax, outId);
+  return out;
+}
+var clipByValueConfig3 = {
+  kernelName: ClipByValue,
+  backendName: "wasm",
+  setupFunc: setup8,
+  kernelFunc: clip
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Concat.js
+function concat4(args) {
+  const { inputs, backend: backend2 } = args;
+  const axis = util_exports.parseAxisParam(args.attrs.axis, inputs[0].shape)[0];
+  const shapes = inputs.map((t) => t.shape);
+  backend_util_exports.assertParamsConsistent(shapes, axis);
+  let outShape = backend_util_exports.computeOutShape(inputs.map((t) => t.shape), axis);
+  const $inputs = inputs.filter((t) => util_exports.sizeFromShape(t.shape) > 0);
+  if ($inputs.length === 1) {
+    return identity4({ inputs: { x: $inputs[0] }, backend: backend2 });
+  }
+  const out = backend2.makeOutput(outShape, inputs[0].dtype);
+  if (util_exports.sizeFromShape(outShape) === 0) {
+    return out;
+  }
+  if ($inputs[0].dtype === "string") {
+    const inputs2D = $inputs.map((t) => {
+      const innerSize = util_exports.sizeFromShape(t.shape.slice(axis));
+      const shape = [-1, innerSize];
+      return reshape5({ inputs: { x: t }, backend: backend2, attrs: { shape } });
+    });
+    const inputsValShapes = inputs2D.map((t) => {
+      return { vals: backend2.readSync(t.dataId), shape: t.shape };
+    });
+    outShape = backend_util_exports.computeOutShape(inputs2D.map((t) => t.shape), 1);
+    const simplyConcat = inputs2D[0].shape[0] === 1;
+    const outVals2 = concatImpl(inputsValShapes, outShape, inputs[0].dtype, simplyConcat);
+    const finalOutShape = backend_util_exports.computeOutShape($inputs.map((t) => t.shape), axis);
+    out.shape = finalOutShape;
+    const outData = backend2.dataIdMap.get(out.dataId);
+    outData.stringBytes = backend_util_exports.fromStringArrayToUint8(outVals2);
+    inputs2D.forEach((t) => backend2.disposeData(t.dataId));
+    return out;
+  }
+  const batchDim = util_exports.sizeFromShape($inputs[0].shape.slice(0, axis));
+  let sumInnerDims = 0;
+  const innerDims = $inputs.map((input2) => {
+    const innerDim = util_exports.sizeFromShape(input2.shape.slice(axis));
+    sumInnerDims += innerDim;
+    return innerDim;
+  });
+  const inVals = $inputs.map((input2) => backend2.typedArrayFromHeap(input2));
+  const outVals = backend2.typedArrayFromHeap(out);
+  for (let b = 0; b < batchDim; b++) {
+    let outOffset = b * sumInnerDims;
+    for (let i = 0; i < inVals.length; i++) {
+      const innerDim = innerDims[i];
+      const inOffset = b * innerDim;
+      const vals = inVals[i].subarray(inOffset, inOffset + innerDim);
+      outVals.set(vals, outOffset);
+      outOffset += innerDim;
+    }
+  }
+  return out;
+}
+var concatConfig3 = {
+  kernelName: Concat,
+  backendName: "wasm",
+  kernelFunc: concat4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Conv2D.js
+var wasmConv2d;
+function setup9(backend2) {
+  wasmConv2d = backend2.wasm.cwrap(Conv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function conv2d5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const { strides, dilations, pad: pad3, dimRoundingMode, dataFormat } = attrs;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend Conv2D does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var conv2DConfig3 = {
+  kernelName: Conv2D,
+  backendName: "wasm",
+  setupFunc: setup9,
+  kernelFunc: conv2d5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Conv2DBackpropInput.js
+var wasmConv2DBackpropInput;
+function setup10(backend2) {
+  wasmConv2DBackpropInput = backend2.wasm.cwrap(Conv2DBackpropInput, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function conv2DBackpropInput4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { dy, filter } = inputs;
+  const { strides, pad: pad3, dataFormat, dimRoundingMode, inputShape } = attrs;
+  const dilations = 1;
+  const $dataFormat = backend_util_exports.convertConv2DDataFormat(dataFormat);
+  const convInfo = backend_util_exports.computeConv2DInfo(inputShape, filter.shape, strides, dilations, pad3, dimRoundingMode, false, $dataFormat);
+  const { batchSize, filterHeight, filterWidth, inChannels, inHeight, inWidth, outChannels, outHeight, outWidth, strideHeight, strideWidth } = convInfo;
+  const topPad = filterHeight - 1 - convInfo.padInfo.top;
+  const leftPad = filterWidth - 1 - convInfo.padInfo.left;
+  const isChannelsLast = convInfo.dataFormat === "channelsLast";
+  const dxStrides = util_exports.computeStrides(convInfo.inShape);
+  const dyStrides = util_exports.computeStrides(dy.shape);
+  const [fltS0, fltS1, fltS2] = util_exports.computeStrides(filter.shape);
+  const xBatchStride = dxStrides[0];
+  const xRowStride = isChannelsLast ? dxStrides[1] : dxStrides[2];
+  const xColStride = isChannelsLast ? dxStrides[2] : 1;
+  const xChannelStride = isChannelsLast ? 1 : dxStrides[1];
+  const yBatchStride = dyStrides[0];
+  const yRowStride = isChannelsLast ? dyStrides[1] : dyStrides[2];
+  const yColStride = isChannelsLast ? dyStrides[2] : 1;
+  const yChannelStride = isChannelsLast ? 1 : dyStrides[1];
+  const out = backend2.makeOutput(convInfo.inShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const dyId = backend2.dataIdMap.get(dy.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  wasmConv2DBackpropInput(dyId, filterId, batchSize, filterHeight, filterWidth, inHeight, inWidth, inChannels, outHeight, outWidth, outChannels, strideHeight, strideWidth, topPad, leftPad, fltS0, fltS1, fltS2, xBatchStride, xRowStride, xColStride, xChannelStride, yBatchStride, yRowStride, yColStride, yChannelStride, outId);
+  return out;
+}
+var conv2DBackpropInputConfig3 = {
+  kernelName: Conv2DBackpropInput,
+  backendName: "wasm",
+  setupFunc: setup10,
+  kernelFunc: conv2DBackpropInput4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cos.js
+var cosConfig3 = createUnaryKernelConfig(Cos);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cosh.js
+var coshConfig3 = createUnaryKernelConfig(Cosh);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/CropAndResize.js
+var InterpolationMethod;
+(function(InterpolationMethod2) {
+  InterpolationMethod2[InterpolationMethod2["bilinear"] = 0] = "bilinear";
+  InterpolationMethod2[InterpolationMethod2["nearest"] = 1] = "nearest";
+})(InterpolationMethod || (InterpolationMethod = {}));
+var wasmCropAndResize;
+function setup11(backend2) {
+  wasmCropAndResize = backend2.wasm.cwrap(CropAndResize, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cropAndResize4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { method, extrapolationValue, cropSize } = attrs;
+  const { image: image2, boxes, boxInd } = inputs;
+  const numBoxes = boxes.shape[0];
+  const [cropHeight, cropWidth] = cropSize;
+  const outShape = [numBoxes, cropHeight, cropWidth, image2.shape[3]];
+  let imagesData = backend2.dataIdMap.get(image2.dataId);
+  let castedData;
+  if (image2.dtype !== "float32") {
+    castedData = cast5({ backend: backend2, inputs: { x: image2 }, attrs: { dtype: "float32" } });
+    imagesData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const imagesId = imagesData.id;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const boxIndId = backend2.dataIdMap.get(boxInd.dataId).id;
+  const out = backend2.makeOutput(outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const imagesShapeBytes = new Uint8Array(new Int32Array(image2.shape).buffer);
+  wasmCropAndResize(imagesId, boxesId, boxIndId, numBoxes, imagesShapeBytes, cropHeight, cropWidth, InterpolationMethod[method], extrapolationValue, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var cropAndResizeConfig3 = {
+  kernelName: CropAndResize,
+  backendName: "wasm",
+  setupFunc: setup11,
+  kernelFunc: cropAndResize4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cumprod.js
+var wasmCumprod;
+function setup12(backend2) {
+  wasmCumprod = backend2.wasm.cwrap(Cumprod, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cumprod4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  const xRank = x.shape.length;
+  util_exports.assert(x.dtype === "float32" || x.dtype === "int32", () => `cumprod does not support ${x.dtype} tensors in the WASM backend`);
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation !== null) {
+    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  backend_util_exports.assertAxesAreInnerMostDims("cumprod", [permutedAxis], xRank);
+  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);
+  const finalDim = permutedX.shape[permutedAxis];
+  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;
+  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;
+  wasmCumprod(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);
+  let out = permutedOut;
+  if (permutation !== null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });
+    backend2.disposeData(permutedX.dataId);
+    backend2.disposeData(permutedOut.dataId);
+  }
+  return out;
+}
+var cumprodConfig3 = {
+  kernelName: Cumprod,
+  backendName: "wasm",
+  setupFunc: setup12,
+  kernelFunc: cumprod4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Cumsum.js
+var wasmCumsum;
+function setup13(backend2) {
+  wasmCumsum = backend2.wasm.cwrap(Cumsum, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function cumsum4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { axis, exclusive, reverse: reverse5 } = attrs;
+  const xRank = x.shape.length;
+  util_exports.assert(x.dtype === "float32" || x.dtype === "int32", () => `cumsum does not support ${x.dtype} tensors in the WASM backend`);
+  const permutation = backend_util_exports.getAxesPermutation([axis], xRank);
+  let permutedX = x;
+  if (permutation !== null) {
+    permutedX = transpose4({ inputs: { x }, attrs: { perm: permutation }, backend: backend2 });
+  }
+  const permutedAxis = backend_util_exports.getInnerMostAxes(1, xRank)[0];
+  backend_util_exports.assertAxesAreInnerMostDims("cumsum", [permutedAxis], xRank);
+  const permutedOut = backend2.makeOutput(permutedX.shape, permutedX.dtype);
+  const finalDim = permutedX.shape[permutedAxis];
+  const permutedXId = backend2.dataIdMap.get(permutedX.dataId).id;
+  const permutedOutId = backend2.dataIdMap.get(permutedOut.dataId).id;
+  wasmCumsum(permutedXId, exclusive ? 1 : 0, reverse5 ? 1 : 0, finalDim, permutedOutId, CppDType[x.dtype]);
+  let out = permutedOut;
+  if (permutation !== null) {
+    const undoPermutation = backend_util_exports.getUndoAxesPermutation(permutation);
+    out = transpose4({ inputs: { x: permutedOut }, attrs: { perm: undoPermutation }, backend: backend2 });
+    backend2.disposeData(permutedX.dataId);
+    backend2.disposeData(permutedOut.dataId);
+  }
+  return out;
+}
+var cumsumConfig3 = {
+  kernelName: Cumsum,
+  backendName: "wasm",
+  setupFunc: setup13,
+  kernelFunc: cumsum4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/DepthToSpace.js
+var wasmDepthToSpace;
+function setup14(backend2) {
+  wasmDepthToSpace = backend2.wasm.cwrap(DepthToSpace, null, [
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function depthToSpace4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x } = inputs;
+  const { blockSize, dataFormat } = attrs;
+  const batchSize = x.shape[0];
+  const inputHeight = dataFormat === "NHWC" ? x.shape[1] : x.shape[2];
+  const inputWidth = dataFormat === "NHWC" ? x.shape[2] : x.shape[3];
+  const inputDepth = dataFormat === "NHWC" ? x.shape[3] : x.shape[1];
+  const outputHeight = inputHeight * blockSize;
+  const outputWidth = inputWidth * blockSize;
+  const outputDepth = inputDepth / (blockSize * blockSize);
+  const outputShape = dataFormat === "NHWC" ? [batchSize, outputHeight, outputWidth, outputDepth] : [batchSize, outputDepth, outputHeight, outputWidth];
+  const out = backend2.makeOutput(outputShape, "float32");
+  const xData = backend2.dataIdMap.get(x.dataId);
+  const xId = xData.id;
+  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);
+  const outputShapeBytes = new Uint8Array(new Int32Array(outputShape).buffer);
+  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(outputShape)).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const channelsLast = dataFormat === "NHWC" ? 1 : 0;
+  wasmDepthToSpace(xId, blockSize, channelsLast, xStridesBytes, x.shape.length - 1, outputShapeBytes, outStridesBytes, outputShape.length, outId);
+  return out;
+}
+var depthToSpaceConfig3 = {
+  kernelName: DepthToSpace,
+  backendName: "wasm",
+  setupFunc: setup14,
+  kernelFunc: depthToSpace4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/DepthwiseConv2dNative.js
+var wasmDepthwiseConv2d;
+function setup15(backend2) {
+  wasmDepthwiseConv2d = backend2.wasm.cwrap(DepthwiseConv2dNative, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function depthwiseConv2d5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const { strides, dilations, pad: pad3, dimRoundingMode } = attrs;
+  const $dilations = dilations == null ? [1, 1] : dilations;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, $dilations, pad3, dimRoundingMode, true);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmDepthwiseConv2d(xId, x.shape[0], x.shape[1], x.shape[2], filterId, filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var depthwiseConv2dNativeConfig3 = {
+  kernelName: DepthwiseConv2dNative,
+  backendName: "wasm",
+  setupFunc: setup15,
+  kernelFunc: depthwiseConv2d5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Elu.js
+var eluConfig3 = createUnaryKernelConfig(Elu);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Equal.js
+var supportsFullBroadcast2 = false;
+var equalConfig3 = createBinaryKernelConfig(Equal, supportsFullBroadcast2, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Exp.js
+var expConfig3 = createUnaryKernelConfig(Exp, "float32");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ExpandDims.js
+function expandDims5(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { input: input2 } = inputs;
+  const { dim } = attrs;
+  const inputRank = input2.shape.length;
+  const newShape = input2.shape.slice();
+  let $dim = dim;
+  if (dim < 0) {
+    util_exports.assert(-(inputRank + 1) <= dim, () => `Axis must be in the interval [${-(inputRank + 1)}, ${inputRank}]`);
+    $dim = inputRank + dim + 1;
+  }
+  newShape.splice($dim, 0, 1);
+  return reshape5({ inputs: { x: input2 }, backend: backend2, attrs: { shape: newShape } });
+}
+var expandDimsConfig3 = {
+  kernelName: ExpandDims,
+  backendName: "wasm",
+  kernelFunc: expandDims5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Fill.js
+function fill4(args) {
+  const { attrs: { shape, value, dtype }, backend: backend2 } = args;
+  const out = backend2.makeOutput(shape, dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(value);
+  return out;
+}
+var fillConfig3 = {
+  kernelName: Fill,
+  backendName: "wasm",
+  kernelFunc: fill4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FlipLeftRight.js
+var wasmFlipLeftRight;
+function setup16(backend2) {
+  wasmFlipLeftRight = backend2.wasm.cwrap(FlipLeftRight, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function flipLeftRight2(args) {
+  const { inputs, backend: backend2 } = args;
+  const { image: image2 } = inputs;
+  const out = backend2.makeOutput(image2.shape, image2.dtype);
+  const imageId = backend2.dataIdMap.get(image2.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  wasmFlipLeftRight(imageId, batch, imageHeight, imageWidth, numChannels, outId);
+  return out;
+}
+var flipLeftRightConfig3 = {
+  kernelName: FlipLeftRight,
+  backendName: "wasm",
+  kernelFunc: flipLeftRight2,
+  setupFunc: setup16
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Floor.js
+var floorConfig3 = createUnaryKernelConfig(Floor);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FloorDiv.js
+var supportsFullBroadcast3 = false;
+var floorDivConfig3 = createBinaryKernelConfig(FloorDiv, supportsFullBroadcast3);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedBatchNorm.js
+var wasmBatchNorm;
+function setup17(backend2) {
+  wasmBatchNorm = backend2.wasm.cwrap(FusedBatchNorm, null, ["number", "number", "number", "number", "number", "number", "number"]);
+}
+function fusedBatchNorm(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { varianceEpsilon } = attrs;
+  const { x, mean: mean4, variance, offset, scale: scale2 } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const meanId = backend2.dataIdMap.get(mean4.dataId).id;
+  const varianceId = backend2.dataIdMap.get(variance.dataId).id;
+  const offsetId = offset != null ? backend2.dataIdMap.get(offset.dataId).id : 0;
+  const scaleId = scale2 != null ? backend2.dataIdMap.get(scale2.dataId).id : 0;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return out;
+  }
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmBatchNorm(xId, meanId, varianceId, offsetId, scaleId, varianceEpsilon, outId);
+  return out;
+}
+var fusedBatchNormConfig = {
+  kernelName: FusedBatchNorm,
+  backendName: "wasm",
+  setupFunc: setup17,
+  kernelFunc: fusedBatchNorm
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedConv2D.js
+var wasmFusedConv2d;
+function setup18(backend2) {
+  wasmFusedConv2d = backend2.wasm.cwrap(FusedConv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedConv2d2(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode);
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedConv2D in the wasm backend.`);
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const outputChannels = convInfo.outChannels;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    if (biasData.shape[0] !== outputChannels) {
+      throw new Error(`FusedConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);
+    }
+    biasId = biasData.id;
+  }
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  const batchSize = convInfo.batchSize;
+  const inHeight = convInfo.inHeight;
+  const inWidth = convInfo.inWidth;
+  if (dataFormat !== "NHWC") {
+    throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  wasmFusedConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var fusedConv2DConfig3 = {
+  kernelName: FusedConv2D,
+  backendName: "wasm",
+  setupFunc: setup18,
+  kernelFunc: fusedConv2d2
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/FusedDepthwiseConv2D.js
+var wasmFusedDepthwiseConv2d;
+function setup19(backend2) {
+  wasmFusedDepthwiseConv2d = backend2.wasm.cwrap(FusedDepthwiseConv2D, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function fusedDepthwiseConv2d(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x, filter, bias, preluActivationWeights } = inputs;
+  const { strides, pad: pad3, dilations, dataFormat, dimRoundingMode, activation: activation2, leakyreluAlpha } = attrs;
+  const convInfo = backend_util_exports.computeConv2DInfo(x.shape, filter.shape, strides, dilations, pad3, dimRoundingMode, true);
+  const fusedActivation = FusableActivation[activation2];
+  if (fusedActivation == null) {
+    throw new Error(`${activation2} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const filterId = backend2.dataIdMap.get(filter.dataId).id;
+  const outputChannels = convInfo.outChannels;
+  let biasId = 0;
+  if (bias != null) {
+    const biasData = backend2.dataIdMap.get(bias.dataId);
+    if (biasData.shape.length !== 1) {
+      throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${biasData.shape.length}.`);
+    }
+    if (biasData.shape[0] !== outputChannels) {
+      throw new Error(`FusedDepthwiseConv2D bias shape (${biasData.shape}) does not match the number of output channels (${outputChannels})`);
+    }
+    biasId = biasData.id;
+  }
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const isSamePad = convInfo.padInfo.type === "SAME" ? 1 : 0;
+  const batchSize = convInfo.batchSize;
+  const inHeight = convInfo.inHeight;
+  const inWidth = convInfo.inWidth;
+  if (dataFormat !== "NHWC") {
+    throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${dataFormat}'. Please use 'NHWC'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const preluActivationWeightsId = preluActivationWeights == null ? 0 : backend2.dataIdMap.get(preluActivationWeights.dataId).id;
+  wasmFusedDepthwiseConv2d(xId, batchSize, inHeight, inWidth, filterId, filterHeight, filterWidth, biasId, padTop, padRight, padBottom, padLeft, isSamePad, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, fusedActivation, preluActivationWeightsId, leakyreluAlpha || 0, outId);
+  return out;
+}
+var fusedDepthwiseConv2DConfig3 = {
+  kernelName: FusedDepthwiseConv2D,
+  backendName: "wasm",
+  setupFunc: setup19,
+  kernelFunc: fusedDepthwiseConv2d
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GatherNd.js
+var wasmGatherNd;
+function setup20(backend2) {
+  wasmGatherNd = backend2.wasm.cwrap(GatherNd, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function gatherNd3(args) {
+  const { backend: backend2, inputs } = args;
+  const { params, indices } = inputs;
+  const [resultShape, numSlices, sliceSize, strides] = gather_nd_util_exports.prepareAndValidate(params, indices);
+  const out = backend2.makeOutput(resultShape, params.dtype);
+  if (numSlices === 0) {
+    return out;
+  }
+  const indicesShape = indices.shape;
+  const sliceRank = indicesShape[indicesShape.length - 1];
+  const xData = backend2.dataIdMap.get(params.dataId);
+  const xId = xData.id;
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmGatherNd(xId, CppDType[params.dtype], indicesId, numSlices, sliceRank, sliceSize, stridesBytes, outId);
+  return out;
+}
+var gatherNdConfig3 = {
+  kernelName: GatherNd,
+  backendName: "wasm",
+  setupFunc: setup20,
+  kernelFunc: gatherNd3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GatherV2.js
+var wasmGather;
+function setup21(backend2) {
+  wasmGather = backend2.wasm.cwrap("Gather", null, [
+    "number",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number"
+  ]);
+}
+function gatherV23(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x, indices } = inputs;
+  const { axis, batchDims } = attrs;
+  const parsedAxis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const indicesVals = backend2.readSync(indices.dataId);
+  const axisDim = x.shape[parsedAxis];
+  for (let i = 0; i < indicesVals.length; ++i) {
+    const index = indicesVals[i];
+    util_exports.assert(index <= axisDim - 1 && index >= 0, () => `GatherV2: the index value ${index} is not in [0, ${axisDim - 1}]`);
+  }
+  const shapeInfo = backend_util_exports.segment_util.collectGatherOpShapeInfo(x, indices, parsedAxis, batchDims);
+  const flattenX = reshape5({
+    inputs: { x },
+    attrs: {
+      shape: [
+        shapeInfo.batchSize,
+        shapeInfo.outerSize,
+        shapeInfo.dimSize,
+        shapeInfo.sliceSize
+      ]
+    },
+    backend: backend2
+  });
+  const indicesSize = util_exports.sizeFromShape(indices.shape);
+  const flattenIndex = reshape5({
+    inputs: { x: indices },
+    attrs: { shape: [shapeInfo.batchSize, indicesSize / shapeInfo.batchSize] },
+    backend: backend2
+  });
+  const flattenOutputShape = [
+    shapeInfo.batchSize,
+    shapeInfo.outerSize,
+    indicesSize / shapeInfo.batchSize,
+    shapeInfo.sliceSize
+  ];
+  const out = backend2.makeOutput(flattenOutputShape, x.dtype);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return out;
+  }
+  const stridesSize = flattenX.shape.length - 1;
+  const xData = backend2.dataIdMap.get(flattenX.dataId);
+  const xId = xData.id;
+  const indicesData = backend2.dataIdMap.get(flattenIndex.dataId);
+  const indicesId = indicesData.id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenX.shape)).buffer);
+  const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(flattenOutputShape)).buffer);
+  wasmGather(xId, CppDType[x.dtype], xStridesBytes, stridesSize, indicesId, shapeInfo.batchSize, outStridesBytes, outId);
+  backend2.disposeData(flattenX.dataId);
+  backend2.disposeData(flattenIndex.dataId);
+  out.shape = shapeInfo.outputShape;
+  return out;
+}
+var gatherV2Config3 = {
+  kernelName: GatherV2,
+  backendName: "wasm",
+  setupFunc: setup21,
+  kernelFunc: gatherV23
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Greater.js
+var supportsFullBroadcast4 = false;
+var greaterConfig3 = createBinaryKernelConfig(Greater, supportsFullBroadcast4, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/GreaterEqual.js
+var supportsFullBroadcast5 = false;
+var greaterEqualConfig3 = createBinaryKernelConfig(GreaterEqual, supportsFullBroadcast5, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LeakyRelu.js
+var wasmFunc3;
+function setupFunc2(backend2) {
+  wasmFunc3 = backend2.wasm.cwrap(LeakyRelu, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function leakyRelu4(args) {
+  const { inputs: { x }, attrs: { alpha }, backend: backend2 } = args;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, "float32");
+  if (util_exports.sizeFromShape(x.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmFunc3(xId, CppDType[x.dtype], alpha, outId);
+  }
+  return out;
+}
+var leakyReluConfig3 = {
+  kernelName: LeakyRelu,
+  backendName: "wasm",
+  setupFunc: setupFunc2,
+  kernelFunc: leakyRelu4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Less.js
+var supportsFullBroadcast6 = false;
+var lessConfig3 = createBinaryKernelConfig(Less, supportsFullBroadcast6, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LessEqual.js
+var supportsFullBroadcast7 = false;
+var lessEqualConfig3 = createBinaryKernelConfig(LessEqual, supportsFullBroadcast7, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Log.js
+var logConfig3 = createUnaryKernelConfig(Log);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalAnd.js
+var supportsFullBroadcast8 = false;
+var logicalAndConfig3 = createBinaryKernelConfig(LogicalAnd, supportsFullBroadcast8, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalNot.js
+var logicalNotConfig3 = createUnaryKernelConfig(LogicalNot);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalOr.js
+var supportsFullBroadcast9 = false;
+var logicalOrConfig3 = createBinaryKernelConfig(LogicalOr, supportsFullBroadcast9, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/LogicalXor.js
+var supportsFullBroadcast10 = false;
+var logicalXorConfig = createBinaryKernelConfig(LogicalXor, supportsFullBroadcast10, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Max.js
+var wasmMax;
+function setup22(backend2) {
+  wasmMax = backend2.wasm.cwrap(Max, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function max5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { reductionIndices: axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    input2 = transposed;
+    inputId = transposedId;
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("max", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, x.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMax(inputId, CppDType[x.dtype], reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var maxConfig3 = {
+  kernelName: Max,
+  backendName: "wasm",
+  setupFunc: setup22,
+  kernelFunc: max5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Maximum.js
+var supportsFullBroadcast11 = false;
+var maximumConfig3 = createBinaryKernelConfig(Maximum, supportsFullBroadcast11);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/MaxPool.js
+var wasmMaxPool;
+function setup23(backend2) {
+  wasmMaxPool = backend2.wasm.cwrap(MaxPool, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function maxPool4(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const x = inputs.x;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  util_exports.assert(x.dtype === "float32", () => `Error in MaxPool: only float32 input is supported. Got ${x.dtype}.`);
+  const { filterSize, strides, pad: pad3, dimRoundingMode } = attrs;
+  const convInfo = backend_util_exports.computePool2DInfo(x.shape, filterSize, strides, 1, pad3, dimRoundingMode);
+  const filterHeight = convInfo.filterHeight;
+  const filterWidth = convInfo.filterWidth;
+  const padTop = convInfo.padInfo.top;
+  const padRight = convInfo.padInfo.right;
+  const padBottom = convInfo.padInfo.bottom;
+  const padLeft = convInfo.padInfo.left;
+  const dilationHeight = convInfo.dilationHeight;
+  const dilationWidth = convInfo.dilationWidth;
+  const strideHeight = convInfo.strideHeight;
+  const strideWidth = convInfo.strideWidth;
+  const inputChannels = convInfo.inChannels;
+  const outputChannels = convInfo.outChannels;
+  if (convInfo.dataFormat !== "channelsLast") {
+    throw new Error(`wasm backend does not support dataFormat:'${convInfo.dataFormat}'. Please use 'channelsLast'.`);
+  }
+  const out = backend2.makeOutput(convInfo.outShape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmMaxPool(xId, x.shape[0], x.shape[1], x.shape[2], filterHeight, filterWidth, padTop, padRight, padBottom, padLeft, dilationHeight, dilationWidth, strideHeight, strideWidth, inputChannels, outputChannels, outId);
+  return out;
+}
+var maxPoolConfig3 = {
+  kernelName: MaxPool,
+  backendName: "wasm",
+  setupFunc: setup23,
+  kernelFunc: maxPool4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Mean.js
+var wasmMean;
+function setup24(backend2) {
+  wasmMean = backend2.wasm.cwrap(Mean, null, ["number, number, number"]);
+}
+function mean3(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("mean", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  let castedInput = input2;
+  if (input2.dtype !== "float32") {
+    castedInput = cast5({ backend: backend2, inputs: { x: input2 }, attrs: { dtype: "float32" } });
+    inputId = backend2.dataIdMap.get(castedInput.dataId).id;
+  }
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMean(inputId, reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  if (input2.dtype !== "float32") {
+    backend2.disposeData(castedInput.dataId);
+  }
+  return out;
+}
+var meanConfig3 = {
+  kernelName: Mean,
+  backendName: "wasm",
+  setupFunc: setup24,
+  kernelFunc: mean3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Min.js
+var wasmMin;
+function setup25(backend2) {
+  wasmMin = backend2.wasm.cwrap(Min, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function min5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+    }
+  }
+  const inputRank = input2.shape.length;
+  backend_util_exports.assertAxesAreInnerMostDims("min", axes, inputRank);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, axes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmMin(inputId, CppDType[x.dtype], reduceSize, outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var minConfig3 = {
+  kernelName: Min,
+  backendName: "wasm",
+  setupFunc: setup25,
+  kernelFunc: min5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Minimum.js
+var supportsFullBroadcast12 = false;
+var minimumConfig3 = createBinaryKernelConfig(Minimum, supportsFullBroadcast12);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/MirrorPad.js
+var MirrorPaddingMode;
+(function(MirrorPaddingMode2) {
+  MirrorPaddingMode2[MirrorPaddingMode2["reflect"] = 0] = "reflect";
+  MirrorPaddingMode2[MirrorPaddingMode2["symmetric"] = 1] = "symmetric";
+})(MirrorPaddingMode || (MirrorPaddingMode = {}));
+var wasmMirrorPad;
+function setup26(backend2) {
+  wasmMirrorPad = backend2.wasm.cwrap(MirrorPad, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function mirrorPad3(args) {
+  const { inputs: { x }, backend: backend2, attrs: { paddings, mode } } = args;
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);
+  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);
+  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);
+  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);
+  wasmMirrorPad(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, MirrorPaddingMode[mode], outId);
+  return out;
+}
+var mirrorPadConfig3 = {
+  kernelName: MirrorPad,
+  backendName: "wasm",
+  kernelFunc: mirrorPad3,
+  setupFunc: setup26
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Multiply.js
+var supportsFullBroadcast13 = true;
+var multiplyConfig3 = createBinaryKernelConfig(Multiply, supportsFullBroadcast13);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Neg.js
+var negConfig3 = createUnaryKernelConfig(Neg);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppression_util.js
+function parseResultStruct(backend2, resOffset) {
+  const result = new Int32Array(backend2.wasm.HEAPU8.buffer, resOffset, 4);
+  const pSelectedIndices = result[0];
+  const selectedSize = result[1];
+  const pSelectedScores = result[2];
+  const pValidOutputs = result[3];
+  backend2.wasm._free(resOffset);
+  return { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs };
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV3.js
+var wasmFunc4;
+function setup27(backend2) {
+  wasmFunc4 = backend2.wasm.cwrap(
+    NonMaxSuppressionV3,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number"
+    ]
+  );
+}
+function kernelFunc(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc4(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pSelectedScores);
+  backend2.wasm._free(pValidOutputs);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  return selectedIndicesTensor;
+}
+var nonMaxSuppressionV3Config3 = {
+  kernelName: NonMaxSuppressionV3,
+  backendName: "wasm",
+  setupFunc: setup27,
+  kernelFunc
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV4.js
+var wasmFunc5;
+function setup28(backend2) {
+  wasmFunc5 = backend2.wasm.cwrap(
+    NonMaxSuppressionV4,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number",
+      "bool"
+    ]
+  );
+}
+function nonMaxSuppressionV43(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold, padToMaxOutputSize } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc5(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, padToMaxOutputSize);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pSelectedScores);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  const validOutputsTensor = backend2.makeOutput([], "int32", pValidOutputs);
+  return [selectedIndicesTensor, validOutputsTensor];
+}
+var nonMaxSuppressionV4Config3 = {
+  kernelName: NonMaxSuppressionV4,
+  backendName: "wasm",
+  setupFunc: setup28,
+  kernelFunc: nonMaxSuppressionV43
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NonMaxSuppressionV5.js
+var wasmFunc6;
+function setup29(backend2) {
+  wasmFunc6 = backend2.wasm.cwrap(
+    NonMaxSuppressionV5,
+    "number",
+    [
+      "number",
+      "number",
+      "number",
+      "number",
+      "number",
+      "number"
+    ]
+  );
+}
+function kernelFunc2(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { iouThreshold, maxOutputSize, scoreThreshold, softNmsSigma } = attrs;
+  const { boxes, scores } = inputs;
+  const boxesId = backend2.dataIdMap.get(boxes.dataId).id;
+  const scoresId = backend2.dataIdMap.get(scores.dataId).id;
+  const resOffset = wasmFunc6(boxesId, scoresId, maxOutputSize, iouThreshold, scoreThreshold, softNmsSigma);
+  const { pSelectedIndices, selectedSize, pSelectedScores, pValidOutputs } = parseResultStruct(backend2, resOffset);
+  backend2.wasm._free(pValidOutputs);
+  const selectedIndicesTensor = backend2.makeOutput([selectedSize], "int32", pSelectedIndices);
+  const selectedScoresTensor = backend2.makeOutput([selectedSize], "float32", pSelectedScores);
+  return [selectedIndicesTensor, selectedScoresTensor];
+}
+var nonMaxSuppressionV5Config3 = {
+  kernelName: NonMaxSuppressionV5,
+  backendName: "wasm",
+  setupFunc: setup29,
+  kernelFunc: kernelFunc2
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/NotEqual.js
+var supportsFullBroadcast14 = false;
+var notEqualConfig3 = createBinaryKernelConfig(NotEqual, supportsFullBroadcast14, "bool");
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/OneHot.js
+var wasmOneHot;
+function setup30(backend2) {
+  wasmOneHot = backend2.wasm.cwrap(OneHot, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function oneHot4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { indices } = inputs;
+  const { dtype, depth, onValue, offValue } = attrs;
+  const out = backend2.makeOutput([...indices.shape, depth], dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  wasmOneHot(indicesId, depth, onValue, offValue, outId);
+  return out;
+}
+var oneHotConfig3 = {
+  kernelName: OneHot,
+  backendName: "wasm",
+  setupFunc: setup30,
+  kernelFunc: oneHot4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/OnesLike.js
+function onesLike4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(1);
+  return out;
+}
+var onesLikeConfig3 = {
+  kernelName: OnesLike,
+  backendName: "wasm",
+  kernelFunc: onesLike4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Pack.js
+function pack3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { axis } = attrs;
+  if (inputs.length === 1) {
+    return expandDims5({ inputs: { input: inputs[0] }, backend: backend2, attrs: { dim: axis } });
+  }
+  const shape = inputs[0].shape;
+  const dtype = inputs[0].dtype;
+  inputs.forEach((t) => {
+    util_exports.assertShapesMatch(shape, t.shape, "All tensors passed to stack must have matching shapes");
+    util_exports.assert(dtype === t.dtype, () => "All tensors passed to stack must have matching dtypes");
+  });
+  const intermediateTensorInfos = [];
+  const expandedTensors = inputs.map((t) => {
+    const expandedT = expandDims5({ inputs: { input: t }, backend: backend2, attrs: { dim: axis } });
+    intermediateTensorInfos.push(expandedT);
+    return expandedT;
+  });
+  const result = concat4({ inputs: expandedTensors, backend: backend2, attrs: { axis } });
+  intermediateTensorInfos.forEach((t) => backend2.disposeData(t.dataId));
+  return result;
+}
+var packConfig3 = {
+  kernelName: Pack,
+  backendName: "wasm",
+  kernelFunc: pack3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/PadV2.js
+var wasmPadV2;
+function setup31(backend2) {
+  wasmPadV2 = backend2.wasm.cwrap(PadV2, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function pad2(args) {
+  const { inputs: { x }, backend: backend2, attrs: { paddings, constantValue } } = args;
+  const outShape = paddings.map((p2, i) => p2[0] + x.shape[i] + p2[1]);
+  if (util_exports.sizeFromShape(x.shape) === 0) {
+    return fill4({
+      backend: backend2,
+      attrs: { shape: outShape, value: constantValue, dtype: x.dtype }
+    });
+  }
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(outShape, x.dtype);
+  const outTensorData = backend2.dataIdMap.get(out.dataId);
+  const outId = outTensorData.id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const prePaddingsFlat = paddings.map((padTuple) => padTuple[0]);
+  const postPaddingsFlat = paddings.map((padTuple) => padTuple[1]);
+  const prePaddingsBytes = new Uint8Array(new Int32Array(prePaddingsFlat).buffer);
+  const postPaddingsBytes = new Uint8Array(new Int32Array(postPaddingsFlat).buffer);
+  wasmPadV2(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], prePaddingsBytes, postPaddingsBytes, constantValue, outId);
+  return out;
+}
+var padV2Config3 = {
+  kernelName: PadV2,
+  backendName: "wasm",
+  kernelFunc: pad2,
+  setupFunc: setup31
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Pow.js
+var supportsFullBroadcast15 = false;
+var powConfig3 = createBinaryKernelConfig(Pow, supportsFullBroadcast15);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Prelu.js
+var wasmPrelu;
+function setup32(backend2) {
+  wasmPrelu = backend2.wasm.cwrap(Prelu, null, [
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function prelu5(args) {
+  const { inputs, backend: backend2 } = args;
+  const { x, alpha } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const weightsId = backend2.dataIdMap.get(alpha.dataId).id;
+  let inputId = xId;
+  const input2 = x;
+  let castedInput = input2;
+  if (input2.dtype !== "float32") {
+    castedInput = cast5({ backend: backend2, inputs: { x }, attrs: { dtype: "float32" } });
+    inputId = backend2.dataIdMap.get(castedInput.dataId).id;
+  }
+  const out = backend2.makeOutput(x.shape, "float32");
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmPrelu(inputId, weightsId, outId);
+  if (input2.dtype !== "float32") {
+    backend2.disposeData(castedInput.dataId);
+  }
+  return out;
+}
+var preluConfig3 = {
+  kernelName: Prelu,
+  backendName: "wasm",
+  setupFunc: setup32,
+  kernelFunc: prelu5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Prod.js
+var wasmProd;
+function setup33(backend2) {
+  wasmProd = backend2.wasm.cwrap(Prod, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function prod4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("prod", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmProd(inputId, reduceSize, CppDType[out.dtype], outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var prodConfig3 = {
+  kernelName: Prod,
+  backendName: "wasm",
+  setupFunc: setup33,
+  kernelFunc: prod4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Range.js
+var range5 = (args) => {
+  const { backend: backend2, attrs } = args;
+  const { start, stop, step: step5, dtype } = attrs;
+  const values = rangeImpl(start, stop, step5, dtype);
+  const out = backend2.makeOutput([values.length], dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(values);
+  return out;
+};
+var rangeConfig3 = {
+  kernelName: Range,
+  backendName: "wasm",
+  kernelFunc: range5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/RealDiv.js
+var supportsFullBroadcast16 = true;
+var realDivConfig3 = createBinaryKernelConfig(RealDiv, supportsFullBroadcast16);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Relu.js
+var reluConfig3 = createUnaryKernelConfig(Relu);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Relu6.js
+var relu6Config3 = createUnaryKernelConfig(Relu6);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ResizeBilinear.js
+var wasmResizeBilinear;
+function setup34(backend2) {
+  wasmResizeBilinear = backend2.wasm.cwrap(ResizeBilinear, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function resizeBilinear4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const outShape = [batch, newHeight, newWidth, numChannels];
+  let xData = backend2.dataIdMap.get(images.dataId);
+  let castedData;
+  if (xData.dtype !== "float32") {
+    castedData = cast5({ backend: backend2, inputs: { x: images }, attrs: { dtype: "float32" } });
+    xData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const xId = xData.id;
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(images.shape) === 0) {
+    return out;
+  }
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmResizeBilinear(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var resizeBilinearConfig3 = {
+  kernelName: ResizeBilinear,
+  backendName: "wasm",
+  setupFunc: setup34,
+  kernelFunc: resizeBilinear4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ResizeNearestNeighbor.js
+var wasmResizeNearestNeighbor;
+function setup35(backend2) {
+  wasmResizeNearestNeighbor = backend2.wasm.cwrap(ResizeNearestNeighbor, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function resizeNearestNeighbor4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { images } = inputs;
+  const { alignCorners, halfPixelCenters, size } = attrs;
+  const [newHeight, newWidth] = size;
+  const [batch, oldHeight, oldWidth, numChannels] = images.shape;
+  const outShape = [batch, newHeight, newWidth, numChannels];
+  const out = backend2.makeOutput(outShape, "float32");
+  if (util_exports.sizeFromShape(images.shape) === 0) {
+    return out;
+  }
+  let xData = backend2.dataIdMap.get(images.dataId);
+  let castedData;
+  if (xData.dtype !== "float32") {
+    castedData = cast5({
+      backend: backend2,
+      inputs: { x: images },
+      attrs: { dtype: "float32" }
+    });
+    xData = backend2.dataIdMap.get(castedData.dataId);
+  }
+  const xId = xData.id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmResizeNearestNeighbor(xId, batch, oldHeight, oldWidth, numChannels, newHeight, newWidth, alignCorners ? 1 : 0, halfPixelCenters ? 1 : 0, outId);
+  if (castedData != null) {
+    backend2.disposeData(castedData.dataId);
+  }
+  return out;
+}
+var resizeNearestNeighborConfig3 = {
+  kernelName: ResizeNearestNeighbor,
+  backendName: "wasm",
+  setupFunc: setup35,
+  kernelFunc: resizeNearestNeighbor4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Reverse.js
+var wasmReverse;
+function setup36(backend2) {
+  wasmReverse = backend2.wasm.cwrap(Reverse, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function reverse4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { dims } = attrs;
+  const axes = util_exports.parseAxisParam(dims, x.shape);
+  if (x.shape.length === 0) {
+    return identity4({ inputs: { x }, backend: backend2 });
+  }
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const axesBytes = new Uint8Array(new Int32Array(axes).buffer);
+  const outShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  wasmReverse(xId, axesBytes, axes.length, outShapeBytes, x.shape.length, outId);
+  const reshaped = reshape5({ inputs: { x: out }, attrs: { shape: x.shape }, backend: backend2 });
+  backend2.disposeData(out.dataId);
+  return reshaped;
+}
+var reverseConfig3 = {
+  kernelName: Reverse,
+  backendName: "wasm",
+  kernelFunc: reverse4,
+  setupFunc: setup36
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/RotateWithOffset.js
+var wasmRotate;
+function setup37(backend2) {
+  wasmRotate = backend2.wasm.cwrap(RotateWithOffset, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function rotateWithOffset2(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { image: image2 } = inputs;
+  const { radians, fillValue, center } = attrs;
+  const out = backend2.makeOutput(image2.shape, image2.dtype);
+  const imageId = backend2.dataIdMap.get(image2.dataId).id;
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [centerX, centerY] = backend_util_exports.getImageCenter(center, imageHeight, imageWidth);
+  const fillIsBlack = fillValue === 0;
+  const fullOpacityValue = 255;
+  const fillValues2 = typeof fillValue === "number" ? [fillValue, fillValue, fillValue, fillIsBlack ? 0 : fullOpacityValue] : [...fillValue, fullOpacityValue];
+  const fillBytes = new Uint8Array(new Int32Array(fillValues2).buffer);
+  wasmRotate(imageId, batch, imageHeight, imageWidth, numChannels, radians, centerX, centerY, fillBytes, fillValues2.length, outId);
+  return out;
+}
+var rotateWithOffsetConfig3 = {
+  kernelName: RotateWithOffset,
+  backendName: "wasm",
+  kernelFunc: rotateWithOffset2,
+  setupFunc: setup37
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Round.js
+var roundConfig3 = createUnaryKernelConfig(Round);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Rsqrt.js
+var rsqrtConfig3 = createUnaryKernelConfig(Rsqrt);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ScatterNd.js
+var wasmScatterNd;
+function setup38(backend2) {
+  wasmScatterNd = backend2.wasm.cwrap(ScatterNd, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function scatterNd3(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { indices, updates } = inputs;
+  const { shape } = attrs;
+  const out = backend2.makeOutput(shape, updates.dtype);
+  if (util_exports.sizeFromShape(shape) === 0) {
+    return out;
+  }
+  const { sliceRank, numUpdates, sliceSize, strides, outputSize } = scatter_nd_util_exports.calculateShapes(updates, indices, shape);
+  const indicesData = backend2.dataIdMap.get(indices.dataId);
+  const indicesId = indicesData.id;
+  const updatesData = backend2.dataIdMap.get(updates.dataId);
+  const updatesId = updatesData.id;
+  const stridesBytes = new Uint8Array(new Int32Array(strides).buffer);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmScatterNd(indicesId, updatesId, CppDType[updates.dtype], sliceRank, numUpdates, sliceSize, stridesBytes, outputSize, outId);
+  return out;
+}
+var scatterNdConfig3 = {
+  kernelName: ScatterNd,
+  backendName: "wasm",
+  setupFunc: setup38,
+  kernelFunc: scatterNd3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Select.js
+var wasmSelect;
+function setup39(backend2) {
+  wasmSelect = backend2.wasm.cwrap("SelectV2", null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function select4(args) {
+  const { inputs, backend: backend2 } = args;
+  const { condition, t, e } = inputs;
+  const conditionId = backend2.dataIdMap.get(condition.dataId).id;
+  const tId = backend2.dataIdMap.get(t.dataId).id;
+  const eId = backend2.dataIdMap.get(e.dataId).id;
+  const out = backend2.makeOutput(t.shape, t.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const cRank = condition.shape.length;
+  const tRank = t.shape.length;
+  const offset = cRank === 0 || cRank > 1 || tRank === 1 ? 1 : util_exports.sizeFromShape(t.shape.slice(1));
+  wasmSelect(conditionId, tId, eId, offset, outId);
+  return out;
+}
+var selectConfig3 = {
+  kernelName: Select,
+  backendName: "wasm",
+  kernelFunc: select4,
+  setupFunc: setup39
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sigmoid.js
+var wasmFunc7;
+function setup40(backend2) {
+  wasmFunc7 = backend2.wasm.cwrap(Sigmoid, null, ["number", "number"]);
+}
+function sigmoid4(args) {
+  const { backend: backend2, inputs: { x } } = args;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  wasmFunc7(xId, outId);
+  return out;
+}
+var sigmoidConfig3 = {
+  kernelName: "Sigmoid",
+  backendName: "wasm",
+  setupFunc: setup40,
+  kernelFunc: sigmoid4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sin.js
+var sinConfig3 = createUnaryKernelConfig(Sin);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Softmax.js
+var wasmFunc8;
+function setup41(backend2) {
+  wasmFunc8 = backend2.wasm.cwrap(Softmax, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function softmax5(args) {
+  const { backend: backend2, inputs: { logits }, attrs: { dim } } = args;
+  const xId = backend2.dataIdMap.get(logits.dataId).id;
+  const out = backend2.makeOutput(logits.shape, logits.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const channels = logits.shape[dim];
+  const batch = util_exports.sizeFromShape(logits.shape) / channels;
+  if (util_exports.sizeFromShape(out.shape) === 0) {
+    return out;
+  }
+  wasmFunc8(xId, outId, channels, batch);
+  return out;
+}
+var softmaxConfig3 = {
+  kernelName: Softmax,
+  backendName: "wasm",
+  setupFunc: setup41,
+  kernelFunc: softmax5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SpaceToBatchND.js
+function spaceToBatchND4(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const { blockShape, paddings } = attrs;
+  const prod5 = util_exports.sizeFromShape(blockShape);
+  const completePaddings = [[0, 0]];
+  completePaddings.push(...paddings);
+  for (let i = 1 + blockShape.length; i < x.shape.length; ++i) {
+    completePaddings.push([0, 0]);
+  }
+  const paddedX = padV2Config3.kernelFunc({
+    inputs: { x },
+    backend: backend2,
+    attrs: { paddings: completePaddings, constantValue: 0 }
+  });
+  const reshapedPaddedShape = backend_util_exports.getReshaped(paddedX.shape, blockShape, prod5, false);
+  const permutedReshapedPaddedPermutation = backend_util_exports.getPermuted(reshapedPaddedShape.length, blockShape.length, false);
+  const flattenShape = backend_util_exports.getReshapedPermuted(paddedX.shape, blockShape, prod5, false);
+  const reshapeInputs = { x: paddedX };
+  const reshapeAttrs = { shape: reshapedPaddedShape };
+  const paddedXReshaped = reshape5({ inputs: reshapeInputs, backend: backend2, attrs: reshapeAttrs });
+  const transposeInputs = { x: paddedXReshaped };
+  const transposeAttrs = { perm: permutedReshapedPaddedPermutation };
+  const paddedXT = transpose4({ inputs: transposeInputs, backend: backend2, attrs: transposeAttrs });
+  const resultReshapeInputs = { x: paddedXT };
+  const resultReshapeAttrs = { shape: flattenShape };
+  const result = reshape5({ inputs: resultReshapeInputs, backend: backend2, attrs: resultReshapeAttrs });
+  backend2.disposeData(paddedX.dataId);
+  backend2.disposeData(paddedXReshaped.dataId);
+  backend2.disposeData(paddedXT.dataId);
+  return result;
+}
+var spaceToBatchNDConfig3 = {
+  kernelName: SpaceToBatchND,
+  backendName: "wasm",
+  kernelFunc: spaceToBatchND4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseFillEmptyRows.js
+var wasmSparseFillEmptyRows;
+function setup42(backend2) {
+  wasmSparseFillEmptyRows = backend2.wasm.cwrap("SparseFillEmptyRows", "number", [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseFillEmptyRows4(args) {
+  const { backend: backend2, inputs } = args;
+  const { indices, values, denseShape, defaultValue } = inputs;
+  const indicesCount = indices.shape[0];
+  const rank = indices.shape[1];
+  const denseRows = backend2.readSync(denseShape.dataId)[0];
+  const maxOutputIndicesShape = [indicesCount + denseRows, rank];
+  const indicesId = backend2.dataIdMap.get(indices.dataId).id;
+  const valuesId = backend2.dataIdMap.get(values.dataId).id;
+  const defaultValueId = backend2.dataIdMap.get(defaultValue.dataId).id;
+  const outputIndices = backend2.makeOutput(maxOutputIndicesShape, indices.dtype);
+  const outputIndicesId = backend2.dataIdMap.get(outputIndices.dataId).id;
+  const outputValues = backend2.makeOutput(maxOutputIndicesShape.slice(0, 1), values.dtype);
+  const outputValuesId = backend2.dataIdMap.get(outputValues.dataId).id;
+  const emptyRowIndicator = backend2.makeOutput([denseRows], "bool");
+  const emptyRowIndicatorId = backend2.dataIdMap.get(emptyRowIndicator.dataId).id;
+  const reverseIndexMap = backend2.makeOutput([indicesCount], indices.dtype);
+  const reverseIndexMapId = backend2.dataIdMap.get(reverseIndexMap.dataId).id;
+  const exceptionValues = backend2.makeOutput([4], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  const outputRows = wasmSparseFillEmptyRows(indicesId, valuesId, CppDType[values.dtype], indicesCount, denseRows, rank, defaultValueId, outputIndicesId, outputValuesId, emptyRowIndicatorId, reverseIndexMapId, exceptionValuesId);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsIndicesDenseShapeMismatch(exceptionValuesArray[1]);
+      break;
+    }
+    case 2: {
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsNegativeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 3:
+      exceptionMessage = backend_util_exports.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);
+      break;
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(outputIndices.dataId);
+    backend2.disposeData(outputValues.dataId);
+    backend2.disposeData(emptyRowIndicator.dataId);
+    backend2.disposeData(reverseIndexMap.dataId);
+    throw new Error(exceptionMessage);
+  }
+  let resizedIndices = outputIndices;
+  let resizedValues = outputValues;
+  if (outputRows !== maxOutputIndicesShape[0]) {
+    resizedIndices = slice4({
+      inputs: { x: outputIndices },
+      attrs: { begin: 0, size: [outputRows, rank] },
+      backend: backend2
+    });
+    resizedValues = slice4({
+      inputs: { x: outputValues },
+      attrs: { begin: 0, size: outputRows },
+      backend: backend2
+    });
+    backend2.disposeData(outputIndices.dataId);
+    backend2.disposeData(outputValues.dataId);
+  }
+  return [resizedIndices, resizedValues, emptyRowIndicator, reverseIndexMap];
+}
+var sparseFillEmptyRowsConfig3 = {
+  kernelName: SparseFillEmptyRows,
+  backendName: "wasm",
+  setupFunc: setup42,
+  kernelFunc: sparseFillEmptyRows4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseReshape.js
+var wasmSparseReshape;
+function setup43(backend2) {
+  wasmSparseReshape = backend2.wasm.cwrap(SparseReshape, null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseReshape4(args) {
+  const { backend: backend2, inputs } = args;
+  const { inputIndices, inputShape, newShape } = inputs;
+  if (inputIndices.shape.length !== 2) {
+    throw new Error(`Input indices should be a matrix but received shape
+        ${inputIndices.shape}`);
+  }
+  if (inputShape.shape.length !== 1) {
+    throw new Error(`Input shape should be a vector but received shape
+        ${inputShape.shape}`);
+  }
+  if (newShape.shape.length !== 1) {
+    throw new Error(`Target shape should be a vector but received shape ${newShape.shape}`);
+  }
+  const inputIndicesId = backend2.dataIdMap.get(inputIndices.dataId).id;
+  const inputShapeId = backend2.dataIdMap.get(inputShape.dataId).id;
+  const newShapeId = backend2.dataIdMap.get(newShape.dataId).id;
+  const nnz = inputIndices.shape[0];
+  const outputRank = util_exports.sizeFromShape(newShape.shape);
+  const newIndices = backend2.makeOutput([nnz, outputRank], inputIndices.dtype);
+  const newIndicesId = backend2.dataIdMap.get(newIndices.dataId).id;
+  const outputShape = backend2.makeOutput([outputRank], newShape.dtype);
+  const outputShapeId = backend2.dataIdMap.get(outputShape.dataId).id;
+  const exceptionValues = backend2.makeOutput([3], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  wasmSparseReshape(inputIndicesId, inputShapeId, newShapeId, nnz, newIndicesId, outputShapeId, exceptionValuesId);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 0: {
+      exceptionMessage = backend_util_exports.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseReshapeNegativeOutputDimErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    }
+    case 2:
+      exceptionMessage = backend_util_exports.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();
+      break;
+    case 3: {
+      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));
+      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMultipleErrorMessage(inputShapeValues, outputShapeValues);
+      break;
+    }
+    case 4: {
+      const inputShapeValues = Array.from(backend2.readSync(inputShape.dataId)), outputShapeValues = Array.from(backend2.readSync(outputShape.dataId));
+      exceptionMessage = backend_util_exports.getSparseReshapeInputOutputMismatchErrorMessage(inputShapeValues, outputShapeValues);
+      break;
+    }
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(newIndices.dataId);
+    backend2.disposeData(outputShape.dataId);
+    throw new Error(exceptionMessage);
+  }
+  return [newIndices, outputShape];
+}
+var sparseReshapeConfig3 = {
+  kernelName: SparseReshape,
+  backendName: "wasm",
+  setupFunc: setup43,
+  kernelFunc: sparseReshape4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentReduction.js
+var wasmSparseSegmentReduction;
+function setup44(backend2) {
+  wasmSparseSegmentReduction = backend2.wasm.cwrap("SparseSegmentReduction", null, [
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sparseSegmentReduction(args, isMean) {
+  const { backend: backend2, inputs } = args;
+  const { data, indices, segmentIds } = inputs;
+  const numIndices = indices.shape[0];
+  const segmentIdsBack = backend2.readSync(segmentIds.dataId, numIndices - 1, numIndices)[0];
+  const lastSegmentIdPlusOne = numIndices > 0 ? segmentIdsBack + 1 : 0;
+  const outputRows = lastSegmentIdPlusOne;
+  if (outputRows < 0) {
+    throw new Error(backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());
+  }
+  const outputShape = data.shape.slice();
+  outputShape[0] = outputRows;
+  const dataId = backend2.dataIdMap.get(data.dataId).id;
+  const indicesId = backend2.dataIdMap.get(indices.dataId).id;
+  const segmentIdsId = backend2.dataIdMap.get(segmentIds.dataId).id;
+  const output = backend2.makeOutput(outputShape, data.dtype);
+  const outputId = backend2.dataIdMap.get(output.dataId).id;
+  const exceptionValues = backend2.makeOutput([4], "int32");
+  const exceptionValuesId = backend2.dataIdMap.get(exceptionValues.dataId).id;
+  wasmSparseSegmentReduction(dataId, CppDType[data.dtype], data.shape[0], indicesId, segmentIdsId, outputId, exceptionValuesId, isMean, 0);
+  const exceptionValuesArray = backend2.readSync(exceptionValues.dataId);
+  let exceptionMessage;
+  switch (exceptionValuesArray[0]) {
+    case 0: {
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();
+      break;
+    }
+    case 1: {
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();
+      break;
+    }
+    case 2:
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2]);
+      break;
+    case 3:
+      exceptionMessage = backend_util_exports.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(exceptionValuesArray[1], exceptionValuesArray[2], exceptionValuesArray[3]);
+      break;
+    default:
+      exceptionMessage = "";
+  }
+  backend2.disposeData(exceptionValues.dataId);
+  if (exceptionMessage) {
+    backend2.disposeData(output.dataId);
+    throw new Error(exceptionMessage);
+  }
+  return output;
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentMean.js
+function sparseSegmentMean4(args) {
+  return sparseSegmentReduction(args, true);
+}
+var sparseSegmentMeanConfig3 = {
+  kernelName: SparseSegmentMean,
+  backendName: "wasm",
+  setupFunc: setup44,
+  kernelFunc: sparseSegmentMean4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SparseSegmentSum.js
+function sparseSegmentSum4(args) {
+  return sparseSegmentReduction(args, false);
+}
+var sparseSegmentSumConfig3 = {
+  kernelName: SparseSegmentSum,
+  backendName: "wasm",
+  setupFunc: setup44,
+  kernelFunc: sparseSegmentSum4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SplitV.js
+function splitV3(args) {
+  const { inputs, attrs, backend: backend2 } = args;
+  const { x } = inputs;
+  const { numOrSizeSplits, axis } = attrs;
+  const $axis = util_exports.parseAxisParam(axis, x.shape)[0];
+  const splitSizes = backend_util_exports.prepareSplitSize(x, numOrSizeSplits, $axis);
+  const begin = new Array(x.shape.length).fill(0);
+  const size = x.shape.slice();
+  return splitSizes.map((s) => {
+    const xSliceSize = [...size];
+    xSliceSize[$axis] = s;
+    const xSlice = slice4({ inputs: { x }, attrs: { begin, size: xSliceSize }, backend: backend2 });
+    begin[$axis] += s;
+    return xSlice;
+  });
+}
+var splitVConfig3 = {
+  kernelName: SplitV,
+  backendName: "wasm",
+  kernelFunc: splitV3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sqrt.js
+var sqrtConfig3 = createUnaryKernelConfig(Sqrt);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Square.js
+var squareConfig3 = createUnaryKernelConfig(Square);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/SquaredDifference.js
+var supportsFullBroadcast17 = true;
+var squaredDifferenceConfig3 = createBinaryKernelConfig(SquaredDifference, supportsFullBroadcast17);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Step.js
+var wasmStep;
+function setup45(backend2) {
+  wasmStep = backend2.wasm.cwrap(Step, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function step4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { alpha } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmStep(xId, alpha, CppDType[x.dtype], outId);
+  return out;
+}
+var stepConfig3 = {
+  kernelName: Step,
+  backendName: "wasm",
+  setupFunc: setup45,
+  kernelFunc: step4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StridedSlice.js
+var wasmStridedSlice;
+function setup46(backend2) {
+  wasmStridedSlice = backend2.wasm.cwrap(StridedSlice, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "array",
+    "array",
+    "array",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function stridedSlice4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { x } = inputs;
+  const { begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask } = attrs;
+  const { finalShapeSparse, finalShape, isIdentity, sliceDim0, isSimpleSlice, begin: $begin, end: $end, strides: $strides } = slice_util_exports.sliceInfo(x.shape, begin, end, strides, beginMask, endMask, ellipsisMask, newAxisMask, shrinkAxisMask);
+  let result;
+  if (isIdentity) {
+    result = reshape5({ inputs: { x }, backend: backend2, attrs: { shape: finalShape } });
+  } else if (sliceDim0 || isSimpleSlice) {
+    util_exports.assert(x.shape.length >= 1, () => `Input must have rank at least 1, got: ${x.shape.length}`);
+    const size = slice_util_exports.computeOutShape($begin, $end, $strides);
+    const sliced = slice4({ inputs: { x }, backend: backend2, attrs: { begin: $begin, size } });
+    result = reshape5({ inputs: { x: sliced }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeData(sliced.dataId);
+  } else {
+    const out = backend2.makeOutput(finalShapeSparse, "float32");
+    const xId = backend2.dataIdMap.get(x.dataId).id;
+    const xStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(x.shape)).buffer);
+    const beginBytes = new Uint8Array(new Int32Array($begin).buffer);
+    const endBytes = new Uint8Array(new Int32Array($end).buffer);
+    const stridesBytes = new Uint8Array(new Int32Array($strides).buffer);
+    const outputShapeBytes = new Uint8Array(new Int32Array(finalShapeSparse).buffer);
+    const outStridesBytes = new Uint8Array(new Int32Array(util_exports.computeStrides(finalShapeSparse)).buffer);
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmStridedSlice(xId, xStridesBytes, x.shape.length, beginBytes, endBytes, stridesBytes, outputShapeBytes, outStridesBytes, finalShapeSparse.length, outId);
+    result = reshape5({ inputs: { x: out }, backend: backend2, attrs: { shape: finalShape } });
+    backend2.disposeData(out.dataId);
+  }
+  return result;
+}
+var stridedSliceConfig3 = {
+  kernelName: StridedSlice,
+  backendName: "wasm",
+  setupFunc: setup46,
+  kernelFunc: stridedSlice4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringNGrams.js
+function stringNGrams4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { data, dataSplits } = inputs;
+  const { separator, nGramWidths, leftPad, rightPad: rightPad2, padWidth, preserveShortSequences } = attrs;
+  const $data = backend2.readSync(data.dataId);
+  const $dataSplits = backend2.readSync(dataSplits.dataId);
+  const [nGrams, nGramsSplits] = stringNGramsImpl($data, $dataSplits, separator, nGramWidths, leftPad, rightPad2, padWidth, preserveShortSequences);
+  const nGramsOut = backend2.makeOutput([nGrams.length], "string");
+  const nGramsOutData = backend2.dataIdMap.get(nGramsOut.dataId);
+  nGramsOutData.stringBytes = nGrams;
+  const nGramsSplitsOut = backend2.makeOutput(dataSplits.shape, "int32");
+  const nGramsSplitsOutVals = backend2.typedArrayFromHeap(nGramsSplitsOut);
+  nGramsSplitsOutVals.set(nGramsSplits);
+  return [nGramsOut, nGramsSplitsOut];
+}
+var stringNGramsConfig3 = {
+  kernelName: StringNGrams,
+  backendName: "wasm",
+  kernelFunc: stringNGrams4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringSplit.js
+function stringSplit4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { input: input2, delimiter } = inputs;
+  const { skipEmpty } = attrs;
+  const inputVals = backend2.readSync(input2.dataId);
+  const delimiterVals = backend2.readSync(delimiter.dataId);
+  const [indices, values, shape] = stringSplitImpl(inputVals, delimiterVals[0], skipEmpty);
+  const outputSize = values.length;
+  const indicesOut = backend2.makeOutput([outputSize, 2], "int32");
+  const indicesOutVals = backend2.typedArrayFromHeap(indicesOut);
+  indicesOutVals.set(indices);
+  const valuesOut = backend2.makeOutput([outputSize], "string");
+  const valuesOutData = backend2.dataIdMap.get(valuesOut.dataId);
+  valuesOutData.stringBytes = values;
+  const shapeOut = backend2.makeOutput([2], "int32");
+  const shapeOutVals = backend2.typedArrayFromHeap(shapeOut);
+  shapeOutVals.set(shape);
+  return [indicesOut, valuesOut, shapeOut];
+}
+var stringSplitConfig3 = {
+  kernelName: StringSplit,
+  backendName: "wasm",
+  kernelFunc: stringSplit4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/StringToHashBucketFast.js
+function stringToHashBucketFast4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { input: input2 } = inputs;
+  const { numBuckets } = attrs;
+  const inputVals = backend2.readSync(input2.dataId);
+  const values = stringToHashBucketFastImpl(inputVals, numBuckets);
+  const out = backend2.makeOutput(input2.shape, "int32");
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.set(values);
+  return out;
+}
+var stringToHashBucketFastConfig3 = {
+  kernelName: StringToHashBucketFast,
+  backendName: "wasm",
+  kernelFunc: stringToHashBucketFast4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sub.js
+var supportsFullBroadcast18 = true;
+var subConfig3 = createBinaryKernelConfig(Sub, supportsFullBroadcast18);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Sum.js
+var wasmSum;
+function setup47(backend2) {
+  wasmSum = backend2.wasm.cwrap(Sum, null, [
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function sum5(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { axis, keepDims } = attrs;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  let inputId = xId;
+  let input2 = x;
+  const { transposed, axes, originalAxes, inputWasTransposed } = permuteAxesAndTranspose(x, axis, backend2);
+  let reductionAxes = axes;
+  if (inputWasTransposed) {
+    const transposedId = backend2.dataIdMap.get(transposed.dataId).id;
+    if (transposedId !== xId) {
+      input2 = transposed;
+      inputId = transposedId;
+      reductionAxes = backend_util_exports.getInnerMostAxes(reductionAxes.length, input2.shape.length);
+    }
+  }
+  backend_util_exports.assertAxesAreInnerMostDims("sum", reductionAxes, input2.shape.length);
+  const [outShape, reduceShape] = backend_util_exports.computeOutAndReduceShapes(input2.shape, reductionAxes);
+  const reduceSize = util_exports.sizeFromShape(reduceShape);
+  const out = backend2.makeOutput(outShape, input2.dtype);
+  if (util_exports.sizeFromShape(input2.shape) !== 0) {
+    const outId = backend2.dataIdMap.get(out.dataId).id;
+    wasmSum(inputId, reduceSize, CppDType[out.dtype], outId);
+  }
+  if (inputWasTransposed) {
+    backend2.disposeData(transposed.dataId);
+  }
+  if (keepDims) {
+    const newShape = backend_util_exports.expandShapeToKeepDim(out.shape, originalAxes);
+    out.shape = newShape;
+  }
+  return out;
+}
+var sumConfig3 = {
+  kernelName: Sum,
+  backendName: "wasm",
+  setupFunc: setup47,
+  kernelFunc: sum5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tan.js
+var tanConfig3 = createUnaryKernelConfig(Tan);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tanh.js
+var tanhConfig3 = createUnaryKernelConfig(Tanh);
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Tile.js
+var wasmTile;
+function setup48(backend2) {
+  wasmTile = backend2.wasm.cwrap(Tile, null, [
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number"
+  ]);
+}
+function tile5(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { x } = inputs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const { reps } = attrs;
+  const newShape = new Array(x.shape.length);
+  for (let i = 0; i < newShape.length; i++) {
+    newShape[i] = x.shape[i] * reps[i];
+  }
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const newShapeBytes = new Uint8Array(new Int32Array(newShape).buffer);
+  const out = backend2.makeOutput(newShape, x.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  wasmTile(xId, xShapeBytes, x.shape.length, newShapeBytes, newShape.length, CppDType[out.dtype], outId);
+  return out;
+}
+var tileConfig3 = {
+  kernelName: Tile,
+  backendName: "wasm",
+  setupFunc: setup48,
+  kernelFunc: tile5
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/TopK.js
+var wasmTopK;
+function setup49(backend2) {
+  wasmTopK = backend2.wasm.cwrap(TopK, null, [
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "bool",
+    "number",
+    "number"
+  ]);
+}
+var topk2 = ({ inputs, backend: backend2, attrs }) => {
+  const { x } = inputs;
+  const { k, sorted } = attrs;
+  const xId = backend2.dataIdMap.get(x.dataId).id;
+  const xShapeBytes = new Uint8Array(new Int32Array(x.shape).buffer);
+  const outputShape = x.shape.slice();
+  outputShape[outputShape.length - 1] = k;
+  const outValues = backend2.makeOutput(outputShape, x.dtype);
+  const outValuesId = backend2.dataIdMap.get(outValues.dataId).id;
+  const outIndices = backend2.makeOutput(outputShape, "int32");
+  const outIndicesId = backend2.dataIdMap.get(outIndices.dataId).id;
+  wasmTopK(xId, xShapeBytes, x.shape.length, CppDType[x.dtype], k, sorted, outValuesId, outIndicesId);
+  return [outValues, outIndices];
+};
+var topKConfig3 = {
+  kernelName: TopK,
+  backendName: "wasm",
+  setupFunc: setup49,
+  kernelFunc: topk2
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Transform.js
+var wasmTransform;
+function setup50(backend2) {
+  wasmTransform = backend2.wasm.cwrap(Transform, null, [
+    "number",
+    "number",
+    "bool",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number",
+    "array",
+    "number",
+    "array",
+    "number",
+    "number",
+    "number",
+    "number",
+    "number"
+  ]);
+}
+function transform4(args) {
+  const { backend: backend2, inputs, attrs } = args;
+  const { image: image2, transforms } = inputs;
+  const { interpolation, fillMode, fillValue, outputShape } = attrs;
+  const [batch, imageHeight, imageWidth, numChannels] = image2.shape;
+  const [outHeight, outWidth] = outputShape != null ? outputShape : [imageHeight, imageWidth];
+  const outShape = [
+    batch,
+    outHeight,
+    outWidth,
+    numChannels
+  ];
+  const inputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(image2.shape)).buffer);
+  const outputStrides = new Uint8Array(new Int32Array(util_exports.computeStrides(outShape)).buffer);
+  const out = backend2.makeOutput(outShape, image2.dtype);
+  const outId = backend2.dataIdMap.get(out.dataId).id;
+  const imageData = backend2.dataIdMap.get(image2.dataId);
+  const imageId = imageData.id;
+  const transformsData = backend2.dataIdMap.get(transforms.dataId);
+  const transformsId = transformsData.id;
+  const interpolationModeId = interpolation === "nearest" ? 1 : 2;
+  let fillModeId;
+  switch (fillMode) {
+    case "constant":
+      fillModeId = 1;
+      break;
+    case "reflect":
+      fillModeId = 2;
+      break;
+    case "wrap":
+      fillModeId = 3;
+      break;
+    case "nearest":
+      fillModeId = 4;
+      break;
+    default:
+      fillModeId = 1;
+      break;
+  }
+  wasmTransform(imageId, transformsId, transforms.shape[0] > 1, batch, outHeight, outWidth, numChannels, imageWidth, imageHeight, inputStrides, image2.shape.length - 1, outputStrides, outShape.length - 1, interpolationModeId, fillModeId, fillValue, outId);
+  return out;
+}
+var transformConfig3 = {
+  kernelName: Transform,
+  backendName: "wasm",
+  setupFunc: setup50,
+  kernelFunc: transform4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/Unpack.js
+function unpack3(args) {
+  const { inputs, backend: backend2, attrs } = args;
+  const { value } = inputs;
+  let { axis } = attrs;
+  if (axis < 0) {
+    axis += value.shape.length;
+  }
+  const numOutputs = value.shape[axis];
+  const rank = value.shape.length;
+  const outShape = new Array(rank - 1);
+  let outIndex = 0;
+  for (let i = 0; i < rank; i++) {
+    if (i !== axis) {
+      outShape[outIndex++] = value.shape[i];
+    }
+  }
+  const outs = new Array(numOutputs);
+  const begin = new Array(rank).fill(0);
+  const size = value.shape.slice();
+  size[axis] = 1;
+  for (let i = 0; i < outs.length; i++) {
+    begin[axis] = i;
+    outs[i] = slice4({ inputs: { x: value }, attrs: { begin, size }, backend: backend2 });
+  }
+  return outs.map(({ dataId, dtype }) => ({ dataId, dtype, shape: outShape }));
+}
+var unpackConfig3 = {
+  kernelName: Unpack,
+  backendName: "wasm",
+  kernelFunc: unpack3
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/kernels/ZerosLike.js
+function zerosLike4(args) {
+  const { inputs: { x }, backend: backend2 } = args;
+  const out = backend2.makeOutput(x.shape, x.dtype);
+  const outVals = backend2.typedArrayFromHeap(out);
+  outVals.fill(0);
+  return out;
+}
+var zerosLikeConfig3 = {
+  kernelName: ZerosLike,
+  backendName: "wasm",
+  kernelFunc: zerosLike4
+};
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/register_all_kernels.js
+var kernelConfigs3 = [
+  _fusedMatMulConfig3,
+  absConfig3,
+  addConfig3,
+  addNConfig3,
+  allConfig3,
+  anyConfig3,
+  argMaxConfig3,
+  avgPoolConfig3,
+  batchMatMulConfig3,
+  batchToSpaceNDConfig3,
+  castConfig3,
+  ceilConfig3,
+  clipByValueConfig3,
+  concatConfig3,
+  conv2DConfig3,
+  conv2DBackpropInputConfig3,
+  cosConfig3,
+  coshConfig3,
+  cropAndResizeConfig3,
+  cumprodConfig3,
+  cumsumConfig3,
+  depthToSpaceConfig3,
+  depthwiseConv2dNativeConfig3,
+  eluConfig3,
+  equalConfig3,
+  expConfig3,
+  expandDimsConfig3,
+  fillConfig3,
+  flipLeftRightConfig3,
+  floorConfig3,
+  floorDivConfig3,
+  fusedBatchNormConfig,
+  fusedConv2DConfig3,
+  fusedDepthwiseConv2DConfig3,
+  gatherNdConfig3,
+  gatherV2Config3,
+  greaterConfig3,
+  greaterEqualConfig3,
+  identityConfig3,
+  leakyReluConfig3,
+  lessConfig3,
+  lessEqualConfig3,
+  logConfig3,
+  logicalAndConfig3,
+  logicalNotConfig3,
+  logicalOrConfig3,
+  logicalXorConfig,
+  maxConfig3,
+  maximumConfig3,
+  maxPoolConfig3,
+  meanConfig3,
+  minConfig3,
+  minimumConfig3,
+  mirrorPadConfig3,
+  multiplyConfig3,
+  negConfig3,
+  nonMaxSuppressionV3Config3,
+  nonMaxSuppressionV4Config3,
+  nonMaxSuppressionV5Config3,
+  notEqualConfig3,
+  oneHotConfig3,
+  onesLikeConfig3,
+  packConfig3,
+  padV2Config3,
+  powConfig3,
+  preluConfig3,
+  prodConfig3,
+  rangeConfig3,
+  realDivConfig3,
+  reluConfig3,
+  relu6Config3,
+  reshapeConfig3,
+  resizeBilinearConfig3,
+  resizeNearestNeighborConfig3,
+  reverseConfig3,
+  rotateWithOffsetConfig3,
+  roundConfig3,
+  rsqrtConfig3,
+  scatterNdConfig3,
+  selectConfig3,
+  sigmoidConfig3,
+  sinConfig3,
+  sliceConfig3,
+  softmaxConfig3,
+  spaceToBatchNDConfig3,
+  sparseFillEmptyRowsConfig3,
+  sparseReshapeConfig3,
+  sparseSegmentMeanConfig3,
+  sparseSegmentSumConfig3,
+  splitVConfig3,
+  sqrtConfig3,
+  squareConfig3,
+  squaredDifferenceConfig3,
+  stepConfig3,
+  stridedSliceConfig3,
+  stringNGramsConfig3,
+  stringSplitConfig3,
+  stringToHashBucketFastConfig3,
+  subConfig3,
+  sumConfig3,
+  tanConfig3,
+  tanhConfig3,
+  tileConfig3,
+  topKConfig3,
+  transformConfig3,
+  transposeConfig3,
+  unpackConfig3,
+  zerosLikeConfig3
+];
+for (const kernelConfig of kernelConfigs3) {
+  registerKernel(kernelConfig);
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/flags_wasm.js
+var ENV6 = env();
+ENV6.registerFlag("WASM_HAS_SIMD_SUPPORT", async () => {
+  try {
+    return WebAssembly.validate(new Uint8Array([
+      0,
+      97,
+      115,
+      109,
+      1,
+      0,
+      0,
+      0,
+      1,
+      4,
+      1,
+      96,
+      0,
+      0,
+      3,
+      2,
+      1,
+      0,
+      10,
+      9,
+      1,
+      7,
+      0,
+      65,
+      0,
+      253,
+      15,
+      26,
+      11
+    ]));
+  } catch (e) {
+    return false;
+  }
+});
+ENV6.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT", async () => {
+  if (ENV6.get("IS_NODE")) {
+    return false;
+  }
+  try {
+    new MessageChannel().port1.postMessage(new SharedArrayBuffer(1));
+    return WebAssembly.validate(new Uint8Array([
+      0,
+      97,
+      115,
+      109,
+      1,
+      0,
+      0,
+      0,
+      1,
+      4,
+      1,
+      96,
+      0,
+      0,
+      3,
+      2,
+      1,
+      0,
+      5,
+      4,
+      1,
+      3,
+      1,
+      1,
+      10,
+      11,
+      1,
+      9,
+      0,
+      65,
+      0,
+      254,
+      16,
+      2,
+      0,
+      26,
+      11
+    ]));
+  } catch (e) {
+    return false;
+  }
+});
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/backend_wasm.js
+var wasmFactoryThreadedSimd_import = __toESM(require_tfjs_backend_wasm_threaded_simd());
+var import_tfjs_backend_wasm_threaded_simd_worker = __toESM(require_tfjs_backend_wasm_threaded_simd_worker());
+var wasmFactory_import = __toESM(require_tfjs_backend_wasm());
+var wasmFactoryThreadedSimd = wasmFactoryThreadedSimd_import.default || wasmFactoryThreadedSimd_import;
+var wasmFactory = wasmFactory_import.default || wasmFactory_import;
+var BackendWasm = class extends KernelBackend {
+  constructor(wasm) {
+    super();
+    this.wasm = wasm;
+    this.dataIdNextNumber = 1;
+    this.wasm.tfjs.initWithThreadsCount(threadsCount);
+    actualThreadsCount = this.wasm.tfjs.getThreadsCount();
+    this.dataIdMap = new DataStorage(this, engine());
+  }
+  write(values, shape, dtype) {
+    const dataId = { id: this.dataIdNextNumber++ };
+    this.move(dataId, values, shape, dtype, 1);
+    return dataId;
+  }
+  numDataIds() {
+    return this.dataIdMap.numDataIds();
+  }
+  async time(f) {
+    const start = util_exports.now();
+    f();
+    const kernelMs = util_exports.now() - start;
+    return { kernelMs };
+  }
+  move(dataId, values, shape, dtype, refCount) {
+    const id = this.dataIdNextNumber++;
+    if (dtype === "string") {
+      const stringBytes = values;
+      this.dataIdMap.set(dataId, { id, stringBytes, shape, dtype, memoryOffset: null, refCount });
+      return;
+    }
+    const size = util_exports.sizeFromShape(shape);
+    const numBytes = size * util_exports.bytesPerElement(dtype);
+    const memoryOffset = this.wasm._malloc(numBytes);
+    this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount });
+    this.wasm.tfjs.registerTensor(id, size, memoryOffset);
+    if (values != null) {
+      this.wasm.HEAPU8.set(new Uint8Array(values.buffer, values.byteOffset, numBytes), memoryOffset);
+    }
+  }
+  async read(dataId) {
+    return this.readSync(dataId);
+  }
+  readSync(dataId, start, end) {
+    const { memoryOffset, dtype, shape, stringBytes } = this.dataIdMap.get(dataId);
+    if (dtype === "string") {
+      if ((start == null || start === 0) && (end == null || end >= stringBytes.length)) {
+        return stringBytes;
+      }
+      return stringBytes.slice(start, end);
+    }
+    start = start || 0;
+    end = end || util_exports.sizeFromShape(shape);
+    const bytesPerElement2 = util_exports.bytesPerElement(dtype);
+    const bytes = this.wasm.HEAPU8.slice(memoryOffset + start * bytesPerElement2, memoryOffset + end * bytesPerElement2);
+    return typedArrayFromBuffer(bytes.buffer, dtype);
+  }
+  disposeData(dataId, force = false) {
+    if (this.dataIdMap.has(dataId)) {
+      const data = this.dataIdMap.get(dataId);
+      data.refCount--;
+      if (!force && data.refCount > 0) {
+        return false;
+      }
+      this.wasm._free(data.memoryOffset);
+      this.wasm.tfjs.disposeData(data.id);
+      this.dataIdMap.delete(dataId);
+    }
+    return true;
+  }
+  refCount(dataId) {
+    if (this.dataIdMap.has(dataId)) {
+      const tensorData = this.dataIdMap.get(dataId);
+      return tensorData.refCount;
+    }
+    return 0;
+  }
+  incRef(dataId) {
+    const data = this.dataIdMap.get(dataId);
+    if (data != null) {
+      data.refCount++;
+    }
+  }
+  floatPrecision() {
+    return 32;
+  }
+  getMemoryOffset(dataId) {
+    return this.dataIdMap.get(dataId).memoryOffset;
+  }
+  dispose() {
+    this.wasm.tfjs.dispose();
+    if ("PThread" in this.wasm) {
+      this.wasm.PThread.terminateAllThreads();
+    }
+    this.wasm = null;
+  }
+  memory() {
+    return { unreliable: false };
+  }
+  makeOutput(shape, dtype, memoryOffset) {
+    let dataId;
+    if (memoryOffset == null) {
+      dataId = this.write(null, shape, dtype);
+    } else {
+      const id = this.dataIdNextNumber++;
+      dataId = { id };
+      this.dataIdMap.set(dataId, { id, memoryOffset, shape, dtype, refCount: 1 });
+      const size = util_exports.sizeFromShape(shape);
+      this.wasm.tfjs.registerTensor(id, size, memoryOffset);
+    }
+    return { dataId, shape, dtype };
+  }
+  typedArrayFromHeap({ shape, dtype, dataId }) {
+    const buffer2 = this.wasm.HEAPU8.buffer;
+    const { memoryOffset } = this.dataIdMap.get(dataId);
+    const size = util_exports.sizeFromShape(shape);
+    switch (dtype) {
+      case "float32":
+        return new Float32Array(buffer2, memoryOffset, size);
+      case "int32":
+        return new Int32Array(buffer2, memoryOffset, size);
+      case "bool":
+        return new Uint8Array(buffer2, memoryOffset, size);
+      default:
+        throw new Error(`Unknown dtype ${dtype}`);
+    }
+  }
+};
+function createInstantiateWasmFunc(path) {
+  return (imports, callback) => {
+    util_exports.fetch(path, { credentials: "same-origin" }).then((response) => {
+      if (!response["ok"]) {
+        imports.env.a(`failed to load wasm binary file at '${path}'`);
+      }
+      response.arrayBuffer().then((binary) => {
+        WebAssembly.instantiate(binary, imports).then((output) => {
+          callback(output.instance, output.module);
+        });
+      });
+    });
+    return {};
+  };
+}
+function getPathToWasmBinary(simdSupported, threadsSupported, wasmModuleFolder) {
+  if (wasmPath != null) {
+    return wasmPath;
+  }
+  let path = "tfjs-backend-wasm.wasm";
+  if (simdSupported && threadsSupported) {
+    path = "tfjs-backend-wasm-threaded-simd.wasm";
+  } else if (simdSupported) {
+    path = "tfjs-backend-wasm-simd.wasm";
+  }
+  if (wasmFileMap != null) {
+    if (wasmFileMap[path] != null) {
+      return wasmFileMap[path];
+    }
+  }
+  return wasmModuleFolder + path;
+}
+async function init() {
+  const [simdSupported, threadsSupported] = await Promise.all([
+    env().getAsync("WASM_HAS_SIMD_SUPPORT"),
+    env().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")
+  ]);
+  return new Promise((resolve, reject) => {
+    const factoryConfig = {};
+    factoryConfig.locateFile = (path, prefix) => {
+      if (path.endsWith(".worker.js")) {
+        const response = import_tfjs_backend_wasm_threaded_simd_worker.wasmWorkerContents.replace(/\n/g, "\\n");
+        const blob = new Blob([response], { type: "application/javascript" });
+        return URL.createObjectURL(blob);
+      }
+      if (path.endsWith(".wasm")) {
+        return getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : prefix);
+      }
+      return prefix + path;
+    };
+    if (customFetch) {
+      factoryConfig.instantiateWasm = createInstantiateWasmFunc(getPathToWasmBinary(simdSupported, threadsSupported, wasmPathPrefix != null ? wasmPathPrefix : ""));
+    }
+    let initialized = false;
+    factoryConfig.onAbort = () => {
+      if (initialized) {
+        return;
+      }
+      if (initAborted) {
+        return;
+      }
+      initAborted = true;
+      const rejectMsg = "Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers";
+      reject({ message: rejectMsg });
+    };
+    let wasm;
+    if (threadsSupported && simdSupported && wasmPath == null) {
+      factoryConfig.mainScriptUrlOrBlob = new Blob([`var WasmBackendModuleThreadedSimd = ` + wasmFactoryThreadedSimd.toString()], { type: "text/javascript" });
+      wasm = wasmFactoryThreadedSimd(factoryConfig);
+    } else {
+      wasm = wasmFactory(factoryConfig);
+    }
+    wasm.then((module) => {
+      initialized = true;
+      initAborted = false;
+      const voidReturnType = null;
+      module.tfjs = {
+        init: module.cwrap("init", null, []),
+        initWithThreadsCount: module.cwrap("init_with_threads_count", null, ["number"]),
+        getThreadsCount: module.cwrap("get_threads_count", "number", []),
+        registerTensor: module.cwrap("register_tensor", null, [
+          "number",
+          "number",
+          "number"
+        ]),
+        disposeData: module.cwrap("dispose_data", voidReturnType, ["number"]),
+        dispose: module.cwrap("dispose", voidReturnType, [])
+      };
+      resolve({ wasm: module });
+    }).catch(reject);
+  });
+}
+function typedArrayFromBuffer(buffer2, dtype) {
+  switch (dtype) {
+    case "float32":
+      return new Float32Array(buffer2);
+    case "int32":
+      return new Int32Array(buffer2);
+    case "bool":
+      return new Uint8Array(buffer2);
+    default:
+      throw new Error(`Unknown dtype ${dtype}`);
+  }
+}
+var wasmBinaryNames = [
+  "tfjs-backend-wasm.wasm",
+  "tfjs-backend-wasm-simd.wasm",
+  "tfjs-backend-wasm-threaded-simd.wasm"
+];
+var wasmPath = null;
+var wasmPathPrefix = null;
+var wasmFileMap = {};
+var initAborted = false;
+var customFetch = false;
+function setWasmPath(path, usePlatformFetch = false) {
+  deprecationWarn("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release.");
+  if (initAborted) {
+    throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");
+  }
+  wasmPath = path;
+  customFetch = usePlatformFetch;
+}
+function setWasmPaths(prefixOrFileMap, usePlatformFetch = false) {
+  if (initAborted) {
+    throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");
+  }
+  if (typeof prefixOrFileMap === "string") {
+    wasmPathPrefix = prefixOrFileMap;
+  } else {
+    wasmFileMap = prefixOrFileMap;
+    const missingPaths = wasmBinaryNames.filter((name) => wasmFileMap[name] == null);
+    if (missingPaths.length > 0) {
+      throw new Error(`There were no entries found for the following binaries: ${missingPaths.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`);
+    }
+  }
+  customFetch = usePlatformFetch;
+}
+var threadsCount = -1;
+var actualThreadsCount = -1;
+function setThreadsCount(numThreads) {
+  threadsCount = numThreads;
+}
+function getThreadsCount() {
+  if (actualThreadsCount === -1) {
+    throw new Error(`WASM backend not initialized.`);
+  }
+  return actualThreadsCount;
+}
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/version.js
+var version8 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-wasm/dist/base.js
+var WASM_PRIORITY = 2;
+registerBackend("wasm", async () => {
+  const { wasm } = await init();
+  return new BackendWasm(wasm);
+}, WASM_PRIORITY);
+
+// dist/tfjs.version.js
+var version9 = "4.0.0";
+var version22 = "4.0.0";
+var version32 = "4.0.0";
+var version42 = "4.0.0";
+var version52 = "4.0.0";
+var version62 = {
+  tfjs: version9,
+  "tfjs-core": version9,
+  "tfjs-converter": version22,
+  "tfjs-backend-cpu": version32,
+  "tfjs-backend-webgl": version42,
+  "tfjs-backend-wasm": version52
+};
+export {
+  Abs,
+  Acos,
+  Acosh,
+  AdadeltaOptimizer,
+  AdagradOptimizer,
+  AdamOptimizer,
+  AdamaxOptimizer,
+  Add,
+  AddN,
+  All,
+  Any,
+  ArgMax,
+  ArgMin,
+  Asin,
+  Asinh,
+  Atan,
+  Atan2,
+  Atanh,
+  AvgPool,
+  AvgPool3D,
+  AvgPool3DGrad,
+  AvgPoolGrad,
+  BackendWasm,
+  BatchMatMul,
+  BatchToSpaceND,
+  Bincount,
+  BroadcastArgs,
+  BroadcastTo,
+  Callback,
+  CallbackList,
+  Cast,
+  Ceil,
+  ClipByValue,
+  Complex,
+  ComplexAbs,
+  Concat,
+  Conv2D,
+  Conv2DBackpropFilter,
+  Conv2DBackpropInput,
+  Conv3D,
+  Conv3DBackpropFilterV2,
+  Conv3DBackpropInputV2,
+  Cos,
+  Cosh,
+  CropAndResize,
+  Cumprod,
+  Cumsum,
+  CustomCallback,
+  DataStorage,
+  DenseBincount,
+  DepthToSpace,
+  DepthwiseConv2dNative,
+  DepthwiseConv2dNativeBackpropFilter,
+  DepthwiseConv2dNativeBackpropInput,
+  Diag,
+  Dilation2D,
+  Dilation2DBackpropFilter,
+  Dilation2DBackpropInput,
+  ENV,
+  EarlyStopping,
+  Einsum,
+  Elu,
+  EluGrad,
+  Environment,
+  Equal,
+  Erf,
+  Exp,
+  ExpandDims,
+  Expm1,
+  FFT,
+  Fill,
+  FlipLeftRight,
+  Floor,
+  FloorDiv,
+  FromPixels,
+  FusedBatchNorm,
+  FusedConv2D,
+  FusedDepthwiseConv2D,
+  GPGPUContext,
+  GatherNd,
+  GatherV2,
+  GraphModel,
+  Greater,
+  GreaterEqual,
+  History,
+  IFFT,
+  Identity,
+  Imag,
+  InputSpec,
+  IsFinite,
+  IsInf,
+  IsNan,
+  KernelBackend,
+  LRN,
+  LRNGrad,
+  LayerVariable,
+  LayersModel,
+  LeakyRelu,
+  Less,
+  LessEqual,
+  LinSpace,
+  Log,
+  Log1p,
+  LogSoftmax,
+  LogicalAnd,
+  LogicalNot,
+  LogicalOr,
+  LogicalXor,
+  LowerBound,
+  MathBackendWebGL,
+  Max,
+  MaxPool,
+  MaxPool3D,
+  MaxPool3DGrad,
+  MaxPoolGrad,
+  MaxPoolWithArgmax,
+  Maximum,
+  Mean,
+  Min,
+  Minimum,
+  MirrorPad,
+  Mod,
+  MomentumOptimizer,
+  Multinomial,
+  Multiply,
+  Neg,
+  NonMaxSuppressionV3,
+  NonMaxSuppressionV4,
+  NonMaxSuppressionV5,
+  NotEqual,
+  OP_SCOPE_SUFFIX,
+  OneHot,
+  OnesLike,
+  Optimizer,
+  OptimizerConstructors,
+  Pack,
+  PadV2,
+  Pool,
+  Pow,
+  Prelu,
+  Prod,
+  RMSPropOptimizer,
+  RNN,
+  RaggedGather,
+  RaggedRange,
+  RaggedTensorToTensor,
+  Range,
+  Rank,
+  Real,
+  RealDiv,
+  Reciprocal,
+  Reduction,
+  Relu,
+  Relu6,
+  Reshape,
+  ResizeBilinear,
+  ResizeBilinearGrad,
+  ResizeNearestNeighbor,
+  ResizeNearestNeighborGrad,
+  Reverse,
+  RotateWithOffset,
+  Round,
+  Rsqrt,
+  SGDOptimizer,
+  ScatterNd,
+  SearchSorted,
+  Select,
+  Selu,
+  Sequential,
+  Sigmoid,
+  Sign,
+  Sin,
+  Sinh,
+  Slice,
+  Softmax,
+  Softplus,
+  SpaceToBatchND,
+  SparseFillEmptyRows,
+  SparseReshape,
+  SparseSegmentMean,
+  SparseSegmentSum,
+  SparseToDense,
+  SplitV,
+  Sqrt,
+  Square,
+  SquaredDifference,
+  Step,
+  StridedSlice,
+  StringNGrams,
+  StringSplit,
+  StringToHashBucketFast,
+  Sub,
+  Sum,
+  SymbolicTensor,
+  Tan,
+  Tanh,
+  Tensor,
+  TensorBuffer,
+  Tile,
+  TopK,
+  Transform,
+  Transpose,
+  Unique,
+  Unpack,
+  UnsortedSegmentSum,
+  UpperBound,
+  Variable,
+  ZerosLike,
+  _FusedMatMul,
+  abs,
+  acos,
+  acosh,
+  add2 as add,
+  addN,
+  all,
+  any,
+  argMax,
+  argMin,
+  asin,
+  asinh,
+  atan,
+  atan2,
+  atanh,
+  avgPool,
+  avgPool3d,
+  backend,
+  backend_util_exports as backend_util,
+  basicLSTMCell,
+  batchNorm,
+  batchNorm2d,
+  batchNorm3d,
+  batchNorm4d,
+  batchToSpaceND,
+  bincount,
+  booleanMaskAsync,
+  broadcastArgs,
+  broadcastTo,
+  broadcast_util_exports as broadcast_util,
+  browser_exports as browser,
+  buffer,
+  callbacks,
+  cast,
+  ceil,
+  clipByValue,
+  clone,
+  complex,
+  concat,
+  concat1d,
+  concat2d,
+  concat3d,
+  concat4d,
+  exports_constraints_exports as constraints,
+  conv1d,
+  conv2d,
+  conv2dTranspose,
+  conv3d,
+  conv3dTranspose,
+  copyRegisteredKernels,
+  cos,
+  cosh,
+  cosineWindow,
+  cumprod,
+  cumsum,
+  customGrad,
+  dist_exports2 as data,
+  denseBincount,
+  deprecationWarn,
+  depthToSpace,
+  depthwiseConv2d,
+  deregisterOp,
+  device_util_exports as device_util,
+  diag,
+  dilation2d,
+  disableDeprecationWarnings,
+  dispose,
+  disposeVariables,
+  div,
+  divNoNan,
+  dot,
+  dropout,
+  einsum,
+  elu,
+  enableDebugMode,
+  enableProdMode,
+  enclosingPowerOfTwo,
+  engine,
+  env,
+  equal,
+  erf,
+  euclideanNorm,
+  exp,
+  expandDims,
+  expm1,
+  eye,
+  fft,
+  fill,
+  findBackend,
+  findBackendFactory,
+  floor,
+  floorDiv,
+  forceHalfFloat,
+  fused_ops_exports as fused,
+  gather,
+  gatherND,
+  gather_nd_util_exports as gather_util,
+  getBackend,
+  getGradient,
+  getKernel,
+  getKernelsForBackend,
+  getThreadsCount,
+  gpgpu_util_exports as gpgpu_util,
+  grad,
+  grads,
+  greater,
+  greaterEqual,
+  ifft,
+  imag,
+  image,
+  inTopKAsync,
+  exports_initializers_exports as initializers,
+  input,
+  io_exports as io,
+  irfft,
+  isFinite2 as isFinite,
+  isInf,
+  isNaN2 as isNaN,
+  keep,
+  kernel_impls_exports as kernel_impls,
+  exports_layers_exports as layers,
+  leakyRelu,
+  less,
+  lessEqual,
+  linalg,
+  linspace,
+  loadGraphModel,
+  loadGraphModelSync,
+  loadLayersModel,
+  localResponseNormalization,
+  log2 as log,
+  log1p,
+  logSigmoid,
+  logSoftmax,
+  logSumExp,
+  logicalAnd,
+  logicalNot,
+  logicalOr,
+  logicalXor,
+  losses,
+  lowerBound,
+  matMul,
+  math_exports as math,
+  max,
+  maxPool,
+  maxPool3d,
+  maxPoolWithArgmax,
+  maximum,
+  mean,
+  memory,
+  meshgrid,
+  exports_metrics_exports as metrics,
+  min,
+  minimum,
+  mirrorPad,
+  mod,
+  model,
+  exports_models_exports as models,
+  moments,
+  movingAverage,
+  mul,
+  multiRNNCell,
+  multinomial,
+  neg,
+  nextFrame,
+  norm,
+  notEqual,
+  oneHot,
+  ones2 as ones,
+  onesLike,
+  op,
+  outerProduct,
+  pad,
+  pad1d,
+  pad2d,
+  pad3d,
+  pad4d,
+  pool,
+  pow,
+  prelu,
+  print,
+  prod,
+  profile,
+  raggedGather,
+  raggedRange,
+  raggedTensorToTensor,
+  rand,
+  randomGamma,
+  randomNormal,
+  randomStandardNormal,
+  randomUniform,
+  range,
+  ready,
+  real,
+  reciprocal,
+  registerBackend,
+  registerCallbackConstructor,
+  registerGradient,
+  registerKernel,
+  registerOp,
+  exports_regularizers_exports as regularizers,
+  relu,
+  relu6,
+  removeBackend,
+  reshape,
+  reverse,
+  reverse1d,
+  reverse2d,
+  reverse3d,
+  reverse4d,
+  rfft,
+  round2 as round,
+  rsqrt,
+  scalar,
+  scatterND,
+  scatter_nd_util_exports as scatter_util,
+  searchSorted,
+  selu,
+  separableConv2d,
+  sequential,
+  serialization_exports as serialization,
+  setBackend,
+  setPlatform,
+  setThreadsCount,
+  setWasmPath,
+  setWasmPaths,
+  setWebGLContext,
+  setdiff1dAsync,
+  sigmoid,
+  sign,
+  signal,
+  sin,
+  sinh,
+  slice,
+  slice1d,
+  slice2d,
+  slice3d,
+  slice4d,
+  slice_util_exports as slice_util,
+  softmax,
+  softplus,
+  spaceToBatchND,
+  sparse,
+  sparseToDense,
+  spectral,
+  split,
+  sqrt,
+  square,
+  squaredDifference,
+  squeeze,
+  stack,
+  step,
+  stridedSlice,
+  string,
+  sub,
+  sum2 as sum,
+  sumOutType,
+  tan,
+  tanh2 as tanh,
+  tensor,
+  tensor1d,
+  tensor2d,
+  tensor3d,
+  tensor4d,
+  tensor5d,
+  tensor6d,
+  tensor_util_exports as tensor_util,
+  test_util_exports as test_util,
+  tidy,
+  tile,
+  time,
+  topk,
+  train,
+  transpose,
+  truncatedNormal,
+  unique,
+  unregisterGradient,
+  unregisterKernel,
+  unsortedSegmentSum,
+  unstack,
+  upcastType,
+  upperBound,
+  util_exports as util,
+  valueAndGrad,
+  valueAndGrads,
+  variable,
+  variableGrads,
+  version62 as version,
+  version3 as version_converter,
+  version as version_core,
+  version2 as version_layers,
+  version8 as version_wasm,
+  version6 as version_webgl,
+  webgl,
+  webgl_util_exports as webgl_util,
+  where,
+  whereAsync,
+  zeros,
+  zerosLike
+};
diff --git a/dist/tfjs.version.js b/dist/tfjs.version.js
index e2bb58a..25e3934 100644
--- a/dist/tfjs.version.js
+++ b/dist/tfjs.version.js
@@ -4,4 +4,31 @@
   author: <https://github.com/vladmandic>'
 */
 
-var e="4.0.0";var s="4.0.0";var t="4.0.0";var i="4.0.0";var n="4.0.0";var y={tfjs:e,"tfjs-core":e,"tfjs-converter":s,"tfjs-backend-cpu":t,"tfjs-backend-webgl":i,"tfjs-backend-wasm":n};export{y as version};
+
+// node_modules/@tensorflow/tfjs-core/package.json
+var version = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-converter/package.json
+var version2 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-cpu/package.json
+var version3 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-webgl/package.json
+var version4 = "4.0.0";
+
+// node_modules/@tensorflow/tfjs-backend-wasm/package.json
+var version5 = "4.0.0";
+
+// src/tfjs/tf-version.ts
+var version6 = {
+  tfjs: version,
+  "tfjs-core": version,
+  "tfjs-converter": version2,
+  "tfjs-backend-cpu": version3,
+  "tfjs-backend-webgl": version4,
+  "tfjs-backend-wasm": version5
+};
+export {
+  version6 as version
+};